dect
/
linux-2.6
Archived
13
0
Fork 0
Code Releases Activity
This repository has been archived on 2022-02-17. You can view files and clone it, but cannot push or open issues or pull requests.
linux-2.6/drivers/gpu/drm/radeon/r600_cs.c

2498 lines
73 KiB
C
Raw Normal View History

drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
/*
* Copyright 2008 Advanced Micro Devices, Inc.
* Copyright 2008 Red Hat Inc.
* Copyright 2009 Jerome Glisse.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
* OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
* ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
* OTHER DEALINGS IN THE SOFTWARE.
*
* Authors: Dave Airlie
* Alex Deucher
* Jerome Glisse
*/
drm/radeon/kms: fix CS alignment checking for tiling (v2) Covers depth, cb, and textures. Hopefully I got this right. v2: - fix bugs: https://bugs.freedesktop.org/show_bug.cgi?id=28327 https://bugs.freedesktop.org/show_bug.cgi?id=28381 - use ALIGNED(), IS_ALIGNED() macros Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-06-04 22:41:42 +00:00
#include <linux/kernel.h>
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
#include "drmP.h"
#include "radeon.h"
#include "r600d.h"
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
#include "r600_reg_safe.h"
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
struct radeon_cs_reloc **cs_reloc);
static int r600_cs_packet_next_reloc_nomm(struct radeon_cs_parser *p,
struct radeon_cs_reloc **cs_reloc);
typedef int (*next_reloc_t)(struct radeon_cs_parser*, struct radeon_cs_reloc**);
static next_reloc_t r600_cs_packet_next_reloc = &r600_cs_packet_next_reloc_mm;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
extern void r600_cs_legacy_get_tiling_conf(struct drm_device *dev, u32 *npipes, u32 *nbanks, u32 *group_size);
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
drm/radeon: r6xx/r7xx possible security issue, system ram access This patch workaround a possible security issue which can allow user to abuse drm on r6xx/r7xx hw to access any system ram memory. This patch doesn't break userspace, it detect "valid" old use of CB_COLOR[0-7]_FRAG & CB_COLOR[0-7]_TILE registers and overwritte the address these registers are pointing to with the one of the last color buffer. This workaround will work for old mesa & xf86-video-ati and any old user which did use similar register programming pattern as those (we expect that there is no others user of those ioctl except possibly a malicious one). This patch add a warning if it detects such usage, warning encourage people to update their mesa & xf86-video-ati. New userspace will submit proper relocation. Fix for xf86-video-ati / mesa (this kernel patch is enough to prevent abuse, fix for userspace are to set proper cs stream and avoid kernel warning) : http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=95d63e408cc88b6934bec84a0b1ef94dfe8bee7b http://cgit.freedesktop.org/mesa/mesa/commit/?id=46dc6fd3ed5ef96cda53641a97bc68c3bc104a9f Abusing this register to perform system ram memory is not easy, here is outline on how it could be achieve. First attacker must have access to the drm device and be able to submit command stream throught cs ioctl. Then attacker must build a proper command stream for r6xx/r7xx hw which will abuse the FRAG or TILE buffer to overwrite the GPU GART which is in VRAM. To achieve so attacker as to setup CB_COLOR[0-7]_FRAG or CB_COLOR[0-7]_TILE to point to the GPU GART, then it has to find a way to write predictable value into those buffer (with little cleverness i believe this can be done but this is an hard task). Once attacker have such program it can overwritte GPU GART to program GPU gart to point anywhere in system memory. It then can reusse same method as he used to reprogram GART to overwritte the system ram through the GART mapping. In the process the attacker has to be carefull to not overwritte any sensitive area of the GART table, like ring or IB gart entry as it will more then likely lead to GPU lockup. Bottom line is that i think it's very hard to use this flaw to get system ram access but in theory one can achieve so. Side note: I am not aware of anyone ever using the GPU as an attack vector, nevertheless we take great care in the opensource driver to try to detect and forbid malicious use of GPU. I don't think the closed source driver are as cautious as we are. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2010-01-18 12:01:36 +00:00
struct r600_cs_track {
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
/* configuration we miror so that we use same code btw kms/ums */
u32 group_size;
u32 nbanks;
u32 npipes;
/* value we track */
drm/radeon/r6xx/r7xx: CS parser fixes - Drop some more safe regs taht userspace shouldn't hit - Constant base regs need relocs. This allows us to use constant buffers rather than the constant register file. Also we don't want userspace to be able to set arbitrary mc base values for the const caches. - Track SQ_CONFIG so we know whether userspace is using the cfile or constant buffers. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-03-26 18:52:32 +00:00
u32 sq_config;
drm/radeon: allow CMASK and FMASK in the CS checker on r600-r700 MSAA is impossible without them. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
2012-08-19 00:22:09 +00:00
u32 log_nsamples;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
u32 nsamples;
u32 cb_color_base_last[8];
struct radeon_bo *cb_color_bo[8];
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
u64 cb_color_bo_mc[8];
drm/radeon: allow CMASK and FMASK in the CS checker on r600-r700 MSAA is impossible without them. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
2012-08-19 00:22:09 +00:00
u64 cb_color_bo_offset[8];
struct radeon_bo *cb_color_frag_bo[8];
u64 cb_color_frag_offset[8];
struct radeon_bo *cb_color_tile_bo[8];
u64 cb_color_tile_offset[8];
u32 cb_color_mask[8];
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
u32 cb_color_info[8];
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
u32 cb_color_view[8];
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
u32 cb_color_size_idx[8]; /* unused */
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
u32 cb_target_mask;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
u32 cb_shader_mask; /* unused */
drm/radeon: add proper checking of RESOLVE_BOX command for r600-r700 Checking of the second colorbuffer was skipped on r700, because CB_TARGET_MASK was 0xf. With r600, CB_TARGET_MASK is changed to 0xff, so we must set the number of samples of the second colorbuffer to 1 in order to pass the CS checker. The DRM version is bumped, because RESOLVE_BOX is always rejected without this fix on r600. Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-24 12:27:36 +00:00
bool is_resolve;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
u32 cb_color_size[8];
u32 vgt_strmout_en;
u32 vgt_strmout_buffer_en;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
struct radeon_bo *vgt_strmout_bo[4];
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
u64 vgt_strmout_bo_mc[4]; /* unused */
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
u32 vgt_strmout_bo_offset[4];
u32 vgt_strmout_size[4];
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
u32 db_depth_control;
u32 db_depth_info;
u32 db_depth_size_idx;
u32 db_depth_view;
u32 db_depth_size;
u32 db_offset;
struct radeon_bo *db_bo;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
u64 db_bo_mc;
drm/radeon/kms: skip cb/db checking if SX_MISC is 1 on r600+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-07 23:56:00 +00:00
bool sx_misc_kill_all_prims;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
bool cb_dirty;
bool db_dirty;
bool streamout_dirty;
drm/radeon/kms: add htile support to the cs checker v3 For 6xx+. Required for mesa to use htile support for HiZ/HiS. Userspace will check radeon version 2.14 with is bumped either by tiling patch or stream out patch. This patch only add support for htile relocation which should be enough for any userspace to implement the hyperz (using htile buffer) feature. v2: Jerome: Fix size checking for htile buffer. v3: Jerome: Adapt on top of r600/evergreen cs checker changes, also check htile surface in case only stencil is present. Signed-off-by: Pierre-Eric Pelloux-Prayer <pelloux@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-21 23:18:21 +00:00
struct radeon_bo *htile_bo;
u64 htile_offset;
u32 htile_surface;
drm/radeon: r6xx/r7xx possible security issue, system ram access This patch workaround a possible security issue which can allow user to abuse drm on r6xx/r7xx hw to access any system ram memory. This patch doesn't break userspace, it detect "valid" old use of CB_COLOR[0-7]_FRAG & CB_COLOR[0-7]_TILE registers and overwritte the address these registers are pointing to with the one of the last color buffer. This workaround will work for old mesa & xf86-video-ati and any old user which did use similar register programming pattern as those (we expect that there is no others user of those ioctl except possibly a malicious one). This patch add a warning if it detects such usage, warning encourage people to update their mesa & xf86-video-ati. New userspace will submit proper relocation. Fix for xf86-video-ati / mesa (this kernel patch is enough to prevent abuse, fix for userspace are to set proper cs stream and avoid kernel warning) : http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=95d63e408cc88b6934bec84a0b1ef94dfe8bee7b http://cgit.freedesktop.org/mesa/mesa/commit/?id=46dc6fd3ed5ef96cda53641a97bc68c3bc104a9f Abusing this register to perform system ram memory is not easy, here is outline on how it could be achieve. First attacker must have access to the drm device and be able to submit command stream throught cs ioctl. Then attacker must build a proper command stream for r6xx/r7xx hw which will abuse the FRAG or TILE buffer to overwrite the GPU GART which is in VRAM. To achieve so attacker as to setup CB_COLOR[0-7]_FRAG or CB_COLOR[0-7]_TILE to point to the GPU GART, then it has to find a way to write predictable value into those buffer (with little cleverness i believe this can be done but this is an hard task). Once attacker have such program it can overwritte GPU GART to program GPU gart to point anywhere in system memory. It then can reusse same method as he used to reprogram GART to overwritte the system ram through the GART mapping. In the process the attacker has to be carefull to not overwritte any sensitive area of the GART table, like ring or IB gart entry as it will more then likely lead to GPU lockup. Bottom line is that i think it's very hard to use this flaw to get system ram access but in theory one can achieve so. Side note: I am not aware of anyone ever using the GPU as an attack vector, nevertheless we take great care in the opensource driver to try to detect and forbid malicious use of GPU. I don't think the closed source driver are as cautious as we are. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2010-01-18 12:01:36 +00:00
};
drm/radeon/kms: add missing Evergreen texture formats to the CS parser BC6 and BC7 are described in ARB_texture_compression_bptc. No idea what FMT_32_AS_32_32_32_32 is good for. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-05-06 23:09:57 +00:00
#define FMT_8_BIT(fmt, vc) [fmt] = { 1, 1, 1, vc, CHIP_R600 }
#define FMT_16_BIT(fmt, vc) [fmt] = { 1, 1, 2, vc, CHIP_R600 }
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
#define FMT_24_BIT(fmt) [fmt] = { 1, 1, 4, 0, CHIP_R600 }
drm/radeon/kms: add missing Evergreen texture formats to the CS parser BC6 and BC7 are described in ARB_texture_compression_bptc. No idea what FMT_32_AS_32_32_32_32 is good for. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-05-06 23:09:57 +00:00
#define FMT_32_BIT(fmt, vc) [fmt] = { 1, 1, 4, vc, CHIP_R600 }
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
#define FMT_48_BIT(fmt) [fmt] = { 1, 1, 8, 0, CHIP_R600 }
drm/radeon/kms: add missing Evergreen texture formats to the CS parser BC6 and BC7 are described in ARB_texture_compression_bptc. No idea what FMT_32_AS_32_32_32_32 is good for. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-05-06 23:09:57 +00:00
#define FMT_64_BIT(fmt, vc) [fmt] = { 1, 1, 8, vc, CHIP_R600 }
#define FMT_96_BIT(fmt) [fmt] = { 1, 1, 12, 0, CHIP_R600 }
#define FMT_128_BIT(fmt, vc) [fmt] = { 1, 1, 16,vc, CHIP_R600 }
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
struct gpu_formats {
unsigned blockwidth;
unsigned blockheight;
unsigned blocksize;
unsigned valid_color;
drm/radeon/kms: add missing Evergreen texture formats to the CS parser BC6 and BC7 are described in ARB_texture_compression_bptc. No idea what FMT_32_AS_32_32_32_32 is good for. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-05-06 23:09:57 +00:00
enum radeon_family min_family;
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
};
static const struct gpu_formats color_formats_table[] = {
/* 8 bit */
FMT_8_BIT(V_038004_COLOR_8, 1),
FMT_8_BIT(V_038004_COLOR_4_4, 1),
FMT_8_BIT(V_038004_COLOR_3_3_2, 1),
FMT_8_BIT(V_038004_FMT_1, 0),
/* 16-bit */
FMT_16_BIT(V_038004_COLOR_16, 1),
FMT_16_BIT(V_038004_COLOR_16_FLOAT, 1),
FMT_16_BIT(V_038004_COLOR_8_8, 1),
FMT_16_BIT(V_038004_COLOR_5_6_5, 1),
FMT_16_BIT(V_038004_COLOR_6_5_5, 1),
FMT_16_BIT(V_038004_COLOR_1_5_5_5, 1),
FMT_16_BIT(V_038004_COLOR_4_4_4_4, 1),
FMT_16_BIT(V_038004_COLOR_5_5_5_1, 1),
/* 24-bit */
FMT_24_BIT(V_038004_FMT_8_8_8),
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
/* 32-bit */
FMT_32_BIT(V_038004_COLOR_32, 1),
FMT_32_BIT(V_038004_COLOR_32_FLOAT, 1),
FMT_32_BIT(V_038004_COLOR_16_16, 1),
FMT_32_BIT(V_038004_COLOR_16_16_FLOAT, 1),
FMT_32_BIT(V_038004_COLOR_8_24, 1),
FMT_32_BIT(V_038004_COLOR_8_24_FLOAT, 1),
FMT_32_BIT(V_038004_COLOR_24_8, 1),
FMT_32_BIT(V_038004_COLOR_24_8_FLOAT, 1),
FMT_32_BIT(V_038004_COLOR_10_11_11, 1),
FMT_32_BIT(V_038004_COLOR_10_11_11_FLOAT, 1),
FMT_32_BIT(V_038004_COLOR_11_11_10, 1),
FMT_32_BIT(V_038004_COLOR_11_11_10_FLOAT, 1),
FMT_32_BIT(V_038004_COLOR_2_10_10_10, 1),
FMT_32_BIT(V_038004_COLOR_8_8_8_8, 1),
FMT_32_BIT(V_038004_COLOR_10_10_10_2, 1),
FMT_32_BIT(V_038004_FMT_5_9_9_9_SHAREDEXP, 0),
FMT_32_BIT(V_038004_FMT_32_AS_8, 0),
FMT_32_BIT(V_038004_FMT_32_AS_8_8, 0),
/* 48-bit */
FMT_48_BIT(V_038004_FMT_16_16_16),
FMT_48_BIT(V_038004_FMT_16_16_16_FLOAT),
/* 64-bit */
FMT_64_BIT(V_038004_COLOR_X24_8_32_FLOAT, 1),
FMT_64_BIT(V_038004_COLOR_32_32, 1),
FMT_64_BIT(V_038004_COLOR_32_32_FLOAT, 1),
FMT_64_BIT(V_038004_COLOR_16_16_16_16, 1),
FMT_64_BIT(V_038004_COLOR_16_16_16_16_FLOAT, 1),
FMT_96_BIT(V_038004_FMT_32_32_32),
FMT_96_BIT(V_038004_FMT_32_32_32_FLOAT),
/* 128-bit */
FMT_128_BIT(V_038004_COLOR_32_32_32_32, 1),
FMT_128_BIT(V_038004_COLOR_32_32_32_32_FLOAT, 1),
[V_038004_FMT_GB_GR] = { 2, 1, 4, 0 },
[V_038004_FMT_BG_RG] = { 2, 1, 4, 0 },
/* block compressed formats */
[V_038004_FMT_BC1] = { 4, 4, 8, 0 },
[V_038004_FMT_BC2] = { 4, 4, 16, 0 },
[V_038004_FMT_BC3] = { 4, 4, 16, 0 },
[V_038004_FMT_BC4] = { 4, 4, 8, 0 },
[V_038004_FMT_BC5] = { 4, 4, 16, 0},
drm/radeon/kms: add missing Evergreen texture formats to the CS parser BC6 and BC7 are described in ARB_texture_compression_bptc. No idea what FMT_32_AS_32_32_32_32 is good for. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-05-06 23:09:57 +00:00
[V_038004_FMT_BC6] = { 4, 4, 16, 0, CHIP_CEDAR}, /* Evergreen-only */
[V_038004_FMT_BC7] = { 4, 4, 16, 0, CHIP_CEDAR}, /* Evergreen-only */
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
drm/radeon/kms: add missing Evergreen texture formats to the CS parser BC6 and BC7 are described in ARB_texture_compression_bptc. No idea what FMT_32_AS_32_32_32_32 is good for. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-05-06 23:09:57 +00:00
/* The other Evergreen formats */
[V_038004_FMT_32_AS_32_32_32_32] = { 1, 1, 4, 0, CHIP_CEDAR},
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
};
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
bool r600_fmt_is_valid_color(u32 format)
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
{
drm/radeon/r600_cs: off by one errors There are a bunch of off by one errors in the sanity checks here. Signed-off-by: Dan Carpenter <error27@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-26 01:48:18 +00:00
if (format >= ARRAY_SIZE(color_formats_table))
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
return false;
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
if (color_formats_table[format].valid_color)
return true;
return false;
}
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
bool r600_fmt_is_valid_texture(u32 format, enum radeon_family family)
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
{
drm/radeon/r600_cs: off by one errors There are a bunch of off by one errors in the sanity checks here. Signed-off-by: Dan Carpenter <error27@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-26 01:48:18 +00:00
if (format >= ARRAY_SIZE(color_formats_table))
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
return false;
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
drm/radeon/kms: add missing Evergreen texture formats to the CS parser BC6 and BC7 are described in ARB_texture_compression_bptc. No idea what FMT_32_AS_32_32_32_32 is good for. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-05-06 23:09:57 +00:00
if (family < color_formats_table[format].min_family)
return false;
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
if (color_formats_table[format].blockwidth > 0)
return true;
return false;
}
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
int r600_fmt_get_blocksize(u32 format)
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
{
drm/radeon/r600_cs: off by one errors There are a bunch of off by one errors in the sanity checks here. Signed-off-by: Dan Carpenter <error27@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-26 01:48:18 +00:00
if (format >= ARRAY_SIZE(color_formats_table))
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
return 0;
return color_formats_table[format].blocksize;
}
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
int r600_fmt_get_nblocksx(u32 format, u32 w)
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
{
unsigned bw;
drm/radeon/r600_cs: off by one errors There are a bunch of off by one errors in the sanity checks here. Signed-off-by: Dan Carpenter <error27@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-26 01:48:18 +00:00
if (format >= ARRAY_SIZE(color_formats_table))
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
return 0;
bw = color_formats_table[format].blockwidth;
if (bw == 0)
return 0;
return (w + bw - 1) / bw;
}
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
int r600_fmt_get_nblocksy(u32 format, u32 h)
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
{
unsigned bh;
drm/radeon/r600_cs: off by one errors There are a bunch of off by one errors in the sanity checks here. Signed-off-by: Dan Carpenter <error27@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-26 01:48:18 +00:00
if (format >= ARRAY_SIZE(color_formats_table))
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
return 0;
bh = color_formats_table[format].blockheight;
if (bh == 0)
return 0;
return (h + bh - 1) / bh;
}
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
struct array_mode_checker {
int array_mode;
u32 group_size;
u32 nbanks;
u32 npipes;
u32 nsamples;
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
u32 blocksize;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
};
/* returns alignment in pixels for pitch/height/depth and bytes for base */
drm/radeon: Drop inlines from evergreen_cs.c / r600_cs.c Fixes evergreen_cs_parse 4080 23124 +19044 and others compared to a non force inline kernel. Signed-off-by: Andi Kleen <ak@linux.intel.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-13 23:08:41 +00:00
static int r600_get_array_mode_alignment(struct array_mode_checker *values,
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
u32 *pitch_align,
u32 *height_align,
u32 *depth_align,
u64 *base_align)
{
u32 tile_width = 8;
u32 tile_height = 8;
u32 macro_tile_width = values->nbanks;
u32 macro_tile_height = values->npipes;
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
u32 tile_bytes = tile_width * tile_height * values->blocksize * values->nsamples;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
u32 macro_tile_bytes = macro_tile_width * macro_tile_height * tile_bytes;
switch (values->array_mode) {
case ARRAY_LINEAR_GENERAL:
/* technically tile_width/_height for pitch/height */
*pitch_align = 1; /* tile_width */
*height_align = 1; /* tile_height */
*depth_align = 1;
*base_align = 1;
break;
case ARRAY_LINEAR_ALIGNED:
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
*pitch_align = max((u32)64, (u32)(values->group_size / values->blocksize));
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
*height_align = 1;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
*depth_align = 1;
*base_align = values->group_size;
break;
case ARRAY_1D_TILED_THIN1:
*pitch_align = max((u32)tile_width,
(u32)(values->group_size /
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
(tile_height * values->blocksize * values->nsamples)));
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
*height_align = tile_height;
*depth_align = 1;
*base_align = values->group_size;
break;
case ARRAY_2D_TILED_THIN1:
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
*pitch_align = max((u32)macro_tile_width * tile_width,
(u32)((values->group_size * values->nbanks) /
(values->blocksize * values->nsamples * tile_width)));
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
*height_align = macro_tile_height * tile_height;
*depth_align = 1;
*base_align = max(macro_tile_bytes,
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
(*pitch_align) * values->blocksize * (*height_align) * values->nsamples);
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
break;
default:
return -EINVAL;
}
return 0;
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
static void r600_cs_track_init(struct r600_cs_track *track)
{
int i;
drm/radeon/r6xx/r7xx: CS parser fixes - Drop some more safe regs taht userspace shouldn't hit - Constant base regs need relocs. This allows us to use constant buffers rather than the constant register file. Also we don't want userspace to be able to set arbitrary mc base values for the const caches. - Track SQ_CONFIG so we know whether userspace is using the cfile or constant buffers. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-03-26 18:52:32 +00:00
/* assume DX9 mode */
track->sq_config = DX9_CONSTS;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
for (i = 0; i < 8; i++) {
track->cb_color_base_last[i] = 0;
track->cb_color_size[i] = 0;
track->cb_color_size_idx[i] = 0;
track->cb_color_info[i] = 0;
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
track->cb_color_view[i] = 0xFFFFFFFF;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
track->cb_color_bo[i] = NULL;
track->cb_color_bo_offset[i] = 0xFFFFFFFF;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
track->cb_color_bo_mc[i] = 0xFFFFFFFF;
drm/radeon: initialize tracked CS state This should help catch uninitialized registers and reject commands because of that. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-22 15:02:43 +00:00
track->cb_color_frag_bo[i] = NULL;
track->cb_color_frag_offset[i] = 0xFFFFFFFF;
track->cb_color_tile_bo[i] = NULL;
track->cb_color_tile_offset[i] = 0xFFFFFFFF;
track->cb_color_mask[i] = 0xFFFFFFFF;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
drm/radeon: add proper checking of RESOLVE_BOX command for r600-r700 Checking of the second colorbuffer was skipped on r700, because CB_TARGET_MASK was 0xf. With r600, CB_TARGET_MASK is changed to 0xff, so we must set the number of samples of the second colorbuffer to 1 in order to pass the CS checker. The DRM version is bumped, because RESOLVE_BOX is always rejected without this fix on r600. Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-24 12:27:36 +00:00
track->is_resolve = false;
drm/radeon: initialize tracked CS state This should help catch uninitialized registers and reject commands because of that. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-22 15:02:43 +00:00
track->nsamples = 16;
track->log_nsamples = 4;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
track->cb_target_mask = 0xFFFFFFFF;
track->cb_shader_mask = 0xFFFFFFFF;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->cb_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
track->db_bo = NULL;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
track->db_bo_mc = 0xFFFFFFFF;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
/* assume the biggest format and that htile is enabled */
track->db_depth_info = 7 | (1 << 25);
track->db_depth_view = 0xFFFFC000;
track->db_depth_size = 0xFFFFFFFF;
track->db_depth_size_idx = 0;
track->db_depth_control = 0xFFFFFFFF;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->db_dirty = true;
drm/radeon/kms: add htile support to the cs checker v3 For 6xx+. Required for mesa to use htile support for HiZ/HiS. Userspace will check radeon version 2.14 with is bumped either by tiling patch or stream out patch. This patch only add support for htile relocation which should be enough for any userspace to implement the hyperz (using htile buffer) feature. v2: Jerome: Fix size checking for htile buffer. v3: Jerome: Adapt on top of r600/evergreen cs checker changes, also check htile surface in case only stencil is present. Signed-off-by: Pierre-Eric Pelloux-Prayer <pelloux@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-21 23:18:21 +00:00
track->htile_bo = NULL;
track->htile_offset = 0xFFFFFFFF;
track->htile_surface = 0;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
for (i = 0; i < 4; i++) {
track->vgt_strmout_size[i] = 0;
track->vgt_strmout_bo[i] = NULL;
track->vgt_strmout_bo_offset[i] = 0xFFFFFFFF;
track->vgt_strmout_bo_mc[i] = 0xFFFFFFFF;
}
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->streamout_dirty = true;
drm/radeon/kms: skip cb/db checking if SX_MISC is 1 on r600+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-07 23:56:00 +00:00
track->sx_misc_kill_all_prims = false;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
drm/radeon: Drop inlines from evergreen_cs.c / r600_cs.c Fixes evergreen_cs_parse 4080 23124 +19044 and others compared to a non force inline kernel. Signed-off-by: Andi Kleen <ak@linux.intel.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-13 23:08:41 +00:00
static int r600_cs_track_validate_cb(struct radeon_cs_parser *p, int i)
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
{
struct r600_cs_track *track = p->track;
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
u32 slice_tile_max, size, tmp;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
u32 height, height_align, pitch, pitch_align, depth_align;
u64 base_offset, base_align;
struct array_mode_checker array_check;
drm/radeon: make the ib an inline object No need to malloc it any more. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Christian König <deathsimple@vodafone.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-05-09 13:35:02 +00:00
volatile u32 *ib = p->ib.ptr;
drm/radeon/r600: fix tiling issues in CS checker. The CS checker had some incorrect alignment requirements for 2D surfaces, this made rendering to mipmap levels that were 2D broken. Also the CB height was being worked out from the BO size, this doesn't work at all when rendering mipmap levels, instead we work out what height userspace wanted from slice max and use that to check it fits inside the BO, however the DDX send the wrong slice max for an unaligned buffer so we have to workaround for that even though its a userspace bug. Reviewed-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-21 03:55:40 +00:00
unsigned array_mode;
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
u32 format;
drm/radeon: add proper checking of RESOLVE_BOX command for r600-r700 Checking of the second colorbuffer was skipped on r700, because CB_TARGET_MASK was 0xf. With r600, CB_TARGET_MASK is changed to 0xff, so we must set the number of samples of the second colorbuffer to 1 in order to pass the CS checker. The DRM version is bumped, because RESOLVE_BOX is always rejected without this fix on r600. Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-24 12:27:36 +00:00
/* When resolve is used, the second colorbuffer has always 1 sample. */
unsigned nsamples = track->is_resolve && i == 1 ? 1 : track->nsamples;
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Andre Maasikas <amaasikas@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-06 06:54:05 +00:00
size = radeon_bo_size(track->cb_color_bo[i]) - track->cb_color_bo_offset[i];
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
format = G_0280A0_FORMAT(track->cb_color_info[i]);
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
if (!r600_fmt_is_valid_color(format)) {
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
dev_warn(p->dev, "%s:%d cb invalid format %d for %d (0x%08X)\n",
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
__func__, __LINE__, format,
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
i, track->cb_color_info[i]);
return -EINVAL;
}
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
/* pitch in pixels */
pitch = (G_028060_PITCH_TILE_MAX(track->cb_color_size[i]) + 1) * 8;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
slice_tile_max = G_028060_SLICE_TILE_MAX(track->cb_color_size[i]) + 1;
drm/radeon/r600: fix tiling issues in CS checker. The CS checker had some incorrect alignment requirements for 2D surfaces, this made rendering to mipmap levels that were 2D broken. Also the CB height was being worked out from the BO size, this doesn't work at all when rendering mipmap levels, instead we work out what height userspace wanted from slice max and use that to check it fits inside the BO, however the DDX send the wrong slice max for an unaligned buffer so we have to workaround for that even though its a userspace bug. Reviewed-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-21 03:55:40 +00:00
slice_tile_max *= 64;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
height = slice_tile_max / pitch;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
if (height > 8192)
height = 8192;
drm/radeon/r600: fix tiling issues in CS checker. The CS checker had some incorrect alignment requirements for 2D surfaces, this made rendering to mipmap levels that were 2D broken. Also the CB height was being worked out from the BO size, this doesn't work at all when rendering mipmap levels, instead we work out what height userspace wanted from slice max and use that to check it fits inside the BO, however the DDX send the wrong slice max for an unaligned buffer so we have to workaround for that even though its a userspace bug. Reviewed-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-21 03:55:40 +00:00
array_mode = G_0280A0_ARRAY_MODE(track->cb_color_info[i]);
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
base_offset = track->cb_color_bo_mc[i] + track->cb_color_bo_offset[i];
array_check.array_mode = array_mode;
array_check.group_size = track->group_size;
array_check.nbanks = track->nbanks;
array_check.npipes = track->npipes;
drm/radeon: add proper checking of RESOLVE_BOX command for r600-r700 Checking of the second colorbuffer was skipped on r700, because CB_TARGET_MASK was 0xf. With r600, CB_TARGET_MASK is changed to 0xff, so we must set the number of samples of the second colorbuffer to 1 in order to pass the CS checker. The DRM version is bumped, because RESOLVE_BOX is always rejected without this fix on r600. Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-24 12:27:36 +00:00
array_check.nsamples = nsamples;
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
array_check.blocksize = r600_fmt_get_blocksize(format);
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
if (r600_get_array_mode_alignment(&array_check,
&pitch_align, &height_align, &depth_align, &base_align)) {
dev_warn(p->dev, "%s invalid tiling %d for %d (0x%08X)\n", __func__,
G_0280A0_ARRAY_MODE(track->cb_color_info[i]), i,
track->cb_color_info[i]);
return -EINVAL;
}
drm/radeon/r600: fix tiling issues in CS checker. The CS checker had some incorrect alignment requirements for 2D surfaces, this made rendering to mipmap levels that were 2D broken. Also the CB height was being worked out from the BO size, this doesn't work at all when rendering mipmap levels, instead we work out what height userspace wanted from slice max and use that to check it fits inside the BO, however the DDX send the wrong slice max for an unaligned buffer so we have to workaround for that even though its a userspace bug. Reviewed-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-21 03:55:40 +00:00
switch (array_mode) {
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
case V_0280A0_ARRAY_LINEAR_GENERAL:
drm/radeon/kms: fix CS alignment checking for tiling (v2) Covers depth, cb, and textures. Hopefully I got this right. v2: - fix bugs: https://bugs.freedesktop.org/show_bug.cgi?id=28327 https://bugs.freedesktop.org/show_bug.cgi?id=28381 - use ALIGNED(), IS_ALIGNED() macros Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-06-04 22:41:42 +00:00
break;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
case V_0280A0_ARRAY_LINEAR_ALIGNED:
break;
case V_0280A0_ARRAY_1D_TILED_THIN1:
drm/radeon/kms: fix r6xx/7xx 1D tiling CS checker v2 broken by: drm/radeon/r600: fix tiling issues in CS checker. v2: only apply it to 1D tiling case. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-27 00:22:42 +00:00
/* avoid breaking userspace */
if (height > 7)
height &= ~0x7;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
case V_0280A0_ARRAY_2D_TILED_THIN1:
break;
default:
dev_warn(p->dev, "%s invalid tiling %d for %d (0x%08X)\n", __func__,
G_0280A0_ARRAY_MODE(track->cb_color_info[i]), i,
track->cb_color_info[i]);
return -EINVAL;
}
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
if (!IS_ALIGNED(pitch, pitch_align)) {
drm/radeon/kms: improve 6xx/7xx CS error output Makes debugging CS rejections much easier. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: stable@kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-13 23:42:41 +00:00
dev_warn(p->dev, "%s:%d cb pitch (%d, 0x%x, %d) invalid\n",
__func__, __LINE__, pitch, pitch_align, array_mode);
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
return -EINVAL;
}
if (!IS_ALIGNED(height, height_align)) {
drm/radeon/kms: improve 6xx/7xx CS error output Makes debugging CS rejections much easier. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: stable@kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-13 23:42:41 +00:00
dev_warn(p->dev, "%s:%d cb height (%d, 0x%x, %d) invalid\n",
__func__, __LINE__, height, height_align, array_mode);
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
return -EINVAL;
}
if (!IS_ALIGNED(base_offset, base_align)) {
drm/radeon/kms: improve 6xx/7xx CS error output Makes debugging CS rejections much easier. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: stable@kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-13 23:42:41 +00:00
dev_warn(p->dev, "%s offset[%d] 0x%llx 0x%llx, %d not aligned\n", __func__, i,
base_offset, base_align, array_mode);
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
return -EINVAL;
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
/* check offset */
drm/radeon: fix checking of MSAA renderbuffers on r600-r700 The MSAA checking was mostly unimplemented on r600-r700. The userspace submits GPU commands and the kernel driver computes how much memory the GPU will access and checks if it's all within buffer bounds the userspace allocated. This patch fixes the computations of the size of MSAA surfaces in memory. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-19 19:23:26 +00:00
tmp = r600_fmt_get_nblocksy(format, height) * r600_fmt_get_nblocksx(format, pitch) *
drm/radeon: add proper checking of RESOLVE_BOX command for r600-r700 Checking of the second colorbuffer was skipped on r700, because CB_TARGET_MASK was 0xf. With r600, CB_TARGET_MASK is changed to 0xff, so we must set the number of samples of the second colorbuffer to 1 in order to pass the CS checker. The DRM version is bumped, because RESOLVE_BOX is always rejected without this fix on r600. Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-24 12:27:36 +00:00
r600_fmt_get_blocksize(format) * nsamples;
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
switch (array_mode) {
default:
case V_0280A0_ARRAY_LINEAR_GENERAL:
case V_0280A0_ARRAY_LINEAR_ALIGNED:
tmp += track->cb_color_view[i] & 0xFF;
break;
case V_0280A0_ARRAY_1D_TILED_THIN1:
case V_0280A0_ARRAY_2D_TILED_THIN1:
tmp += G_028080_SLICE_MAX(track->cb_color_view[i]) * tmp;
break;
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
if ((tmp + track->cb_color_bo_offset[i]) > radeon_bo_size(track->cb_color_bo[i])) {
drm/radeon/r600: fix tiling issues in CS checker. The CS checker had some incorrect alignment requirements for 2D surfaces, this made rendering to mipmap levels that were 2D broken. Also the CB height was being worked out from the BO size, this doesn't work at all when rendering mipmap levels, instead we work out what height userspace wanted from slice max and use that to check it fits inside the BO, however the DDX send the wrong slice max for an unaligned buffer so we have to workaround for that even though its a userspace bug. Reviewed-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-21 03:55:40 +00:00
if (array_mode == V_0280A0_ARRAY_LINEAR_GENERAL) {
/* the initial DDX does bad things with the CB size occasionally */
/* it rounds up height too far for slice tile max but the BO is smaller */
drm/radeon/kms: disable the r600 cb offset checker for linear surfaces There are too many strange corner cases triggered in old userspace drivers out there to that it's nearly impossible to not break some obscure app. Cc: Jean Delvare <khali@linux-fr.org> Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-12-13 19:03:09 +00:00
/* r600c,g also seem to flush at bad times in some apps resulting in
* bogus values here. So for linear just allow anything to avoid breaking
* broken userspace.
*/
drm/radeon/r600: fix tiling issues in CS checker. The CS checker had some incorrect alignment requirements for 2D surfaces, this made rendering to mipmap levels that were 2D broken. Also the CB height was being worked out from the BO size, this doesn't work at all when rendering mipmap levels, instead we work out what height userspace wanted from slice max and use that to check it fits inside the BO, however the DDX send the wrong slice max for an unaligned buffer so we have to workaround for that even though its a userspace bug. Reviewed-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-21 03:55:40 +00:00
} else {
drm/radeon: allow CMASK and FMASK in the CS checker on r600-r700 MSAA is impossible without them. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
2012-08-19 00:22:09 +00:00
dev_warn(p->dev, "%s offset[%d] %d %llu %d %lu too big (%d %d) (%d %d %d)\n",
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
__func__, i, array_mode,
drm/radeon/kms: improve 6xx/7xx CS error output Makes debugging CS rejections much easier. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: stable@kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-13 23:42:41 +00:00
track->cb_color_bo_offset[i], tmp,
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
radeon_bo_size(track->cb_color_bo[i]),
pitch, height, r600_fmt_get_nblocksx(format, pitch),
r600_fmt_get_nblocksy(format, height),
r600_fmt_get_blocksize(format));
drm/radeon/r600: fix tiling issues in CS checker. The CS checker had some incorrect alignment requirements for 2D surfaces, this made rendering to mipmap levels that were 2D broken. Also the CB height was being worked out from the BO size, this doesn't work at all when rendering mipmap levels, instead we work out what height userspace wanted from slice max and use that to check it fits inside the BO, however the DDX send the wrong slice max for an unaligned buffer so we have to workaround for that even though its a userspace bug. Reviewed-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-10-21 03:55:40 +00:00
return -EINVAL;
}
drm/radeon/kms: fix CS alignment checking for tiling (v2) Covers depth, cb, and textures. Hopefully I got this right. v2: - fix bugs: https://bugs.freedesktop.org/show_bug.cgi?id=28327 https://bugs.freedesktop.org/show_bug.cgi?id=28381 - use ALIGNED(), IS_ALIGNED() macros Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-06-04 22:41:42 +00:00
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
/* limit max tile */
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
tmp = (height * pitch) >> 6;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
if (tmp < slice_tile_max)
slice_tile_max = tmp;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
tmp = S_028060_PITCH_TILE_MAX((pitch / 8) - 1) |
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
S_028060_SLICE_TILE_MAX(slice_tile_max - 1);
ib[track->cb_color_size_idx[i]] = tmp;
drm/radeon: allow CMASK and FMASK in the CS checker on r600-r700 MSAA is impossible without them. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
2012-08-19 00:22:09 +00:00
/* FMASK/CMASK */
switch (G_0280A0_TILE_MODE(track->cb_color_info[i])) {
case V_0280A0_TILE_DISABLE:
break;
case V_0280A0_FRAG_ENABLE:
if (track->nsamples > 1) {
uint32_t tile_max = G_028100_FMASK_TILE_MAX(track->cb_color_mask[i]);
/* the tile size is 8x8, but the size is in units of bits.
* for bytes, do just * 8. */
uint32_t bytes = track->nsamples * track->log_nsamples * 8 * (tile_max + 1);
if (bytes + track->cb_color_frag_offset[i] >
radeon_bo_size(track->cb_color_frag_bo[i])) {
dev_warn(p->dev, "%s FMASK_TILE_MAX too large "
"(tile_max=%u, bytes=%u, offset=%llu, bo_size=%lu)\n",
__func__, tile_max, bytes,
track->cb_color_frag_offset[i],
radeon_bo_size(track->cb_color_frag_bo[i]));
return -EINVAL;
}
}
/* fall through */
case V_0280A0_CLEAR_ENABLE:
{
uint32_t block_max = G_028100_CMASK_BLOCK_MAX(track->cb_color_mask[i]);
/* One block = 128x128 pixels, one 8x8 tile has 4 bits..
* (128*128) / (8*8) / 2 = 128 bytes per block. */
uint32_t bytes = (block_max + 1) * 128;
if (bytes + track->cb_color_tile_offset[i] >
radeon_bo_size(track->cb_color_tile_bo[i])) {
dev_warn(p->dev, "%s CMASK_BLOCK_MAX too large "
"(block_max=%u, bytes=%u, offset=%llu, bo_size=%lu)\n",
__func__, block_max, bytes,
track->cb_color_tile_offset[i],
radeon_bo_size(track->cb_color_tile_bo[i]));
return -EINVAL;
}
break;
}
default:
dev_warn(p->dev, "%s invalid tile mode\n", __func__);
return -EINVAL;
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
return 0;
}
drm/radeon/kms: add htile support to the cs checker v3 For 6xx+. Required for mesa to use htile support for HiZ/HiS. Userspace will check radeon version 2.14 with is bumped either by tiling patch or stream out patch. This patch only add support for htile relocation which should be enough for any userspace to implement the hyperz (using htile buffer) feature. v2: Jerome: Fix size checking for htile buffer. v3: Jerome: Adapt on top of r600/evergreen cs checker changes, also check htile surface in case only stencil is present. Signed-off-by: Pierre-Eric Pelloux-Prayer <pelloux@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-21 23:18:21 +00:00
static int r600_cs_track_validate_db(struct radeon_cs_parser *p)
{
struct r600_cs_track *track = p->track;
u32 nviews, bpe, ntiles, size, slice_tile_max, tmp;
u32 height_align, pitch_align, depth_align;
u32 pitch = 8192;
u32 height = 8192;
u64 base_offset, base_align;
struct array_mode_checker array_check;
int array_mode;
drm/radeon: make the ib an inline object No need to malloc it any more. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Christian König <deathsimple@vodafone.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-05-09 13:35:02 +00:00
volatile u32 *ib = p->ib.ptr;
drm/radeon/kms: add htile support to the cs checker v3 For 6xx+. Required for mesa to use htile support for HiZ/HiS. Userspace will check radeon version 2.14 with is bumped either by tiling patch or stream out patch. This patch only add support for htile relocation which should be enough for any userspace to implement the hyperz (using htile buffer) feature. v2: Jerome: Fix size checking for htile buffer. v3: Jerome: Adapt on top of r600/evergreen cs checker changes, also check htile surface in case only stencil is present. Signed-off-by: Pierre-Eric Pelloux-Prayer <pelloux@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-21 23:18:21 +00:00
if (track->db_bo == NULL) {
dev_warn(p->dev, "z/stencil with no depth buffer\n");
return -EINVAL;
}
switch (G_028010_FORMAT(track->db_depth_info)) {
case V_028010_DEPTH_16:
bpe = 2;
break;
case V_028010_DEPTH_X8_24:
case V_028010_DEPTH_8_24:
case V_028010_DEPTH_X8_24_FLOAT:
case V_028010_DEPTH_8_24_FLOAT:
case V_028010_DEPTH_32_FLOAT:
bpe = 4;
break;
case V_028010_DEPTH_X24_8_32_FLOAT:
bpe = 8;
break;
default:
dev_warn(p->dev, "z/stencil with invalid format %d\n", G_028010_FORMAT(track->db_depth_info));
return -EINVAL;
}
if ((track->db_depth_size & 0xFFFFFC00) == 0xFFFFFC00) {
if (!track->db_depth_size_idx) {
dev_warn(p->dev, "z/stencil buffer size not set\n");
return -EINVAL;
}
tmp = radeon_bo_size(track->db_bo) - track->db_offset;
tmp = (tmp / bpe) >> 6;
if (!tmp) {
dev_warn(p->dev, "z/stencil buffer too small (0x%08X %d %d %ld)\n",
track->db_depth_size, bpe, track->db_offset,
radeon_bo_size(track->db_bo));
return -EINVAL;
}
ib[track->db_depth_size_idx] = S_028000_SLICE_TILE_MAX(tmp - 1) | (track->db_depth_size & 0x3FF);
} else {
size = radeon_bo_size(track->db_bo);
/* pitch in pixels */
pitch = (G_028000_PITCH_TILE_MAX(track->db_depth_size) + 1) * 8;
slice_tile_max = G_028000_SLICE_TILE_MAX(track->db_depth_size) + 1;
slice_tile_max *= 64;
height = slice_tile_max / pitch;
if (height > 8192)
height = 8192;
base_offset = track->db_bo_mc + track->db_offset;
array_mode = G_028010_ARRAY_MODE(track->db_depth_info);
array_check.array_mode = array_mode;
array_check.group_size = track->group_size;
array_check.nbanks = track->nbanks;
array_check.npipes = track->npipes;
array_check.nsamples = track->nsamples;
array_check.blocksize = bpe;
if (r600_get_array_mode_alignment(&array_check,
&pitch_align, &height_align, &depth_align, &base_align)) {
dev_warn(p->dev, "%s invalid tiling %d (0x%08X)\n", __func__,
G_028010_ARRAY_MODE(track->db_depth_info),
track->db_depth_info);
return -EINVAL;
}
switch (array_mode) {
case V_028010_ARRAY_1D_TILED_THIN1:
/* don't break userspace */
height &= ~0x7;
break;
case V_028010_ARRAY_2D_TILED_THIN1:
break;
default:
dev_warn(p->dev, "%s invalid tiling %d (0x%08X)\n", __func__,
G_028010_ARRAY_MODE(track->db_depth_info),
track->db_depth_info);
return -EINVAL;
}
if (!IS_ALIGNED(pitch, pitch_align)) {
dev_warn(p->dev, "%s:%d db pitch (%d, 0x%x, %d) invalid\n",
__func__, __LINE__, pitch, pitch_align, array_mode);
return -EINVAL;
}
if (!IS_ALIGNED(height, height_align)) {
dev_warn(p->dev, "%s:%d db height (%d, 0x%x, %d) invalid\n",
__func__, __LINE__, height, height_align, array_mode);
return -EINVAL;
}
if (!IS_ALIGNED(base_offset, base_align)) {
dev_warn(p->dev, "%s offset 0x%llx, 0x%llx, %d not aligned\n", __func__,
base_offset, base_align, array_mode);
return -EINVAL;
}
ntiles = G_028000_SLICE_TILE_MAX(track->db_depth_size) + 1;
nviews = G_028004_SLICE_MAX(track->db_depth_view) + 1;
drm/radeon: fix checking of MSAA renderbuffers on r600-r700 The MSAA checking was mostly unimplemented on r600-r700. The userspace submits GPU commands and the kernel driver computes how much memory the GPU will access and checks if it's all within buffer bounds the userspace allocated. This patch fixes the computations of the size of MSAA surfaces in memory. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-19 19:23:26 +00:00
tmp = ntiles * bpe * 64 * nviews * track->nsamples;
drm/radeon/kms: add htile support to the cs checker v3 For 6xx+. Required for mesa to use htile support for HiZ/HiS. Userspace will check radeon version 2.14 with is bumped either by tiling patch or stream out patch. This patch only add support for htile relocation which should be enough for any userspace to implement the hyperz (using htile buffer) feature. v2: Jerome: Fix size checking for htile buffer. v3: Jerome: Adapt on top of r600/evergreen cs checker changes, also check htile surface in case only stencil is present. Signed-off-by: Pierre-Eric Pelloux-Prayer <pelloux@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-21 23:18:21 +00:00
if ((tmp + track->db_offset) > radeon_bo_size(track->db_bo)) {
dev_warn(p->dev, "z/stencil buffer (%d) too small (0x%08X %d %d %d -> %u have %lu)\n",
array_mode,
track->db_depth_size, ntiles, nviews, bpe, tmp + track->db_offset,
radeon_bo_size(track->db_bo));
return -EINVAL;
}
}
/* hyperz */
if (G_028010_TILE_SURFACE_ENABLE(track->db_depth_info)) {
unsigned long size;
unsigned nbx, nby;
if (track->htile_bo == NULL) {
dev_warn(p->dev, "%s:%d htile enabled without htile surface 0x%08x\n",
__func__, __LINE__, track->db_depth_info);
return -EINVAL;
}
if ((track->db_depth_size & 0xFFFFFC00) == 0xFFFFFC00) {
dev_warn(p->dev, "%s:%d htile can't be enabled with bogus db_depth_size 0x%08x\n",
__func__, __LINE__, track->db_depth_size);
return -EINVAL;
}
nbx = pitch;
nby = height;
if (G_028D24_LINEAR(track->htile_surface)) {
/* nbx must be 16 htiles aligned == 16 * 8 pixel aligned */
nbx = round_up(nbx, 16 * 8);
/* nby is npipes htiles aligned == npipes * 8 pixel aligned */
nby = round_up(nby, track->npipes * 8);
} else {
/* htile widht & nby (8 or 4) make 2 bits number */
tmp = track->htile_surface & 3;
/* align is htile align * 8, htile align vary according to
* number of pipe and tile width and nby
*/
switch (track->npipes) {
case 8:
switch (tmp) {
case 3: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 8*/
nbx = round_up(nbx, 64 * 8);
nby = round_up(nby, 64 * 8);
break;
case 2: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 8*/
case 1: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 4*/
nbx = round_up(nbx, 64 * 8);
nby = round_up(nby, 32 * 8);
break;
case 0: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 4*/
nbx = round_up(nbx, 32 * 8);
nby = round_up(nby, 32 * 8);
break;
default:
return -EINVAL;
}
break;
case 4:
switch (tmp) {
case 3: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 8*/
nbx = round_up(nbx, 64 * 8);
nby = round_up(nby, 32 * 8);
break;
case 2: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 8*/
case 1: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 4*/
nbx = round_up(nbx, 32 * 8);
nby = round_up(nby, 32 * 8);
break;
case 0: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 4*/
nbx = round_up(nbx, 32 * 8);
nby = round_up(nby, 16 * 8);
break;
default:
return -EINVAL;
}
break;
case 2:
switch (tmp) {
case 3: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 8*/
nbx = round_up(nbx, 32 * 8);
nby = round_up(nby, 32 * 8);
break;
case 2: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 8*/
case 1: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 4*/
nbx = round_up(nbx, 32 * 8);
nby = round_up(nby, 16 * 8);
break;
case 0: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 4*/
nbx = round_up(nbx, 16 * 8);
nby = round_up(nby, 16 * 8);
break;
default:
return -EINVAL;
}
break;
case 1:
switch (tmp) {
case 3: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 8*/
nbx = round_up(nbx, 32 * 8);
nby = round_up(nby, 16 * 8);
break;
case 2: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 8*/
case 1: /* HTILE_WIDTH = 8 & HTILE_HEIGHT = 4*/
nbx = round_up(nbx, 16 * 8);
nby = round_up(nby, 16 * 8);
break;
case 0: /* HTILE_WIDTH = 4 & HTILE_HEIGHT = 4*/
nbx = round_up(nbx, 16 * 8);
nby = round_up(nby, 8 * 8);
break;
default:
return -EINVAL;
}
break;
default:
dev_warn(p->dev, "%s:%d invalid num pipes %d\n",
__func__, __LINE__, track->npipes);
return -EINVAL;
}
}
/* compute number of htile */
nbx = G_028D24_HTILE_WIDTH(track->htile_surface) ? nbx / 8 : nbx / 4;
nby = G_028D24_HTILE_HEIGHT(track->htile_surface) ? nby / 8 : nby / 4;
size = nbx * nby * 4;
size += track->htile_offset;
if (size > radeon_bo_size(track->htile_bo)) {
dev_warn(p->dev, "%s:%d htile surface too small %ld for %ld (%d %d)\n",
__func__, __LINE__, radeon_bo_size(track->htile_bo),
size, nbx, nby);
return -EINVAL;
}
}
track->db_dirty = false;
return 0;
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
static int r600_cs_track_check(struct radeon_cs_parser *p)
{
struct r600_cs_track *track = p->track;
u32 tmp;
int r, i;
/* on legacy kernel we don't perform advanced check */
if (p->rdev == NULL)
return 0;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
/* check streamout */
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
if (track->streamout_dirty && track->vgt_strmout_en) {
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
for (i = 0; i < 4; i++) {
if (track->vgt_strmout_buffer_en & (1 << i)) {
if (track->vgt_strmout_bo[i]) {
u64 offset = (u64)track->vgt_strmout_bo_offset[i] +
(u64)track->vgt_strmout_size[i];
if (offset > radeon_bo_size(track->vgt_strmout_bo[i])) {
DRM_ERROR("streamout %d bo too small: 0x%llx, 0x%lx\n",
i, offset,
radeon_bo_size(track->vgt_strmout_bo[i]));
return -EINVAL;
}
} else {
dev_warn(p->dev, "No buffer for streamout %d\n", i);
return -EINVAL;
}
}
}
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->streamout_dirty = false;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
drm/radeon/kms: skip cb/db checking if SX_MISC is 1 on r600+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-07 23:56:00 +00:00
if (track->sx_misc_kill_all_prims)
return 0;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
/* check that we have a cb for each enabled target, we don't check
* shader_mask because it seems mesa isn't always setting it :(
*/
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
if (track->cb_dirty) {
tmp = track->cb_target_mask;
drm/radeon: add proper checking of RESOLVE_BOX command for r600-r700 Checking of the second colorbuffer was skipped on r700, because CB_TARGET_MASK was 0xf. With r600, CB_TARGET_MASK is changed to 0xff, so we must set the number of samples of the second colorbuffer to 1 in order to pass the CS checker. The DRM version is bumped, because RESOLVE_BOX is always rejected without this fix on r600. Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-24 12:27:36 +00:00
/* We must check both colorbuffers for RESOLVE. */
if (track->is_resolve) {
tmp |= 0xff;
}
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
for (i = 0; i < 8; i++) {
if ((tmp >> (i * 4)) & 0xF) {
/* at least one component is enabled */
if (track->cb_color_bo[i] == NULL) {
dev_warn(p->dev, "%s:%d mask 0x%08X | 0x%08X no cb for %d\n",
__func__, __LINE__, track->cb_target_mask, track->cb_shader_mask, i);
return -EINVAL;
}
/* perform rewrite of CB_COLOR[0-7]_SIZE */
r = r600_cs_track_validate_cb(p, i);
if (r)
return r;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
}
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->cb_dirty = false;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
drm/radeon/kms: add htile support to the cs checker v3 For 6xx+. Required for mesa to use htile support for HiZ/HiS. Userspace will check radeon version 2.14 with is bumped either by tiling patch or stream out patch. This patch only add support for htile relocation which should be enough for any userspace to implement the hyperz (using htile buffer) feature. v2: Jerome: Fix size checking for htile buffer. v3: Jerome: Adapt on top of r600/evergreen cs checker changes, also check htile surface in case only stencil is present. Signed-off-by: Pierre-Eric Pelloux-Prayer <pelloux@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-21 23:18:21 +00:00
/* Check depth buffer */
drm/radeon/kms: allow "invalid" DB formats as a means to disable DB Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Cc: stable@vger.kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-07-29 14:24:57 +00:00
if (track->db_dirty &&
G_028010_FORMAT(track->db_depth_info) != V_028010_DEPTH_INVALID &&
(G_028800_STENCIL_ENABLE(track->db_depth_control) ||
G_028800_Z_ENABLE(track->db_depth_control))) {
drm/radeon/kms: add htile support to the cs checker v3 For 6xx+. Required for mesa to use htile support for HiZ/HiS. Userspace will check radeon version 2.14 with is bumped either by tiling patch or stream out patch. This patch only add support for htile relocation which should be enough for any userspace to implement the hyperz (using htile buffer) feature. v2: Jerome: Fix size checking for htile buffer. v3: Jerome: Adapt on top of r600/evergreen cs checker changes, also check htile surface in case only stencil is present. Signed-off-by: Pierre-Eric Pelloux-Prayer <pelloux@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-21 23:18:21 +00:00
r = r600_cs_track_validate_db(p);
if (r)
return r;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
drm/radeon/kms: add htile support to the cs checker v3 For 6xx+. Required for mesa to use htile support for HiZ/HiS. Userspace will check radeon version 2.14 with is bumped either by tiling patch or stream out patch. This patch only add support for htile relocation which should be enough for any userspace to implement the hyperz (using htile buffer) feature. v2: Jerome: Fix size checking for htile buffer. v3: Jerome: Adapt on top of r600/evergreen cs checker changes, also check htile surface in case only stencil is present. Signed-off-by: Pierre-Eric Pelloux-Prayer <pelloux@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-21 23:18:21 +00:00
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
return 0;
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
/**
* r600_cs_packet_parse() - parse cp packet and point ib index to next packet
* @parser: parser structure holding parsing context.
* @pkt: where to store packet informations
*
* Assume that chunk_ib_index is properly set. Will return -EINVAL
* if packet is bigger than remaining ib size. or if packets is unknown.
**/
int r600_cs_packet_parse(struct radeon_cs_parser *p,
struct radeon_cs_packet *pkt,
unsigned idx)
{
struct radeon_cs_chunk *ib_chunk = &p->chunks[p->chunk_ib_idx];
uint32_t header;
if (idx >= ib_chunk->length_dw) {
DRM_ERROR("Can not parse packet at %d after CS end %d !\n",
idx, ib_chunk->length_dw);
return -EINVAL;
}
drm/radeon/kms: don't require up to 64k allocations. (v2) This avoids needing to do a kmalloc > PAGE_SIZE for the main indirect buffer chunk, it adds an accessor for all reads from the chunk and caches a single page at a time for subsequent reads. changes since v1: Use a two page pool which should be the most common case a single packet spanning > PAGE_SIZE will be hit, but I'm having trouble seeing anywhere we currently generate anything like that. hopefully proper short page copying at end added parser_error flag to set deep errors instead of having to test every ib value fetch. fixed bug in patch that went to list. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-23 06:56:27 +00:00
header = radeon_get_ib_value(p, idx);
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
pkt->idx = idx;
pkt->type = CP_PACKET_GET_TYPE(header);
pkt->count = CP_PACKET_GET_COUNT(header);
pkt->one_reg_wr = 0;
switch (pkt->type) {
case PACKET_TYPE0:
pkt->reg = CP_PACKET0_GET_REG(header);
break;
case PACKET_TYPE3:
pkt->opcode = CP_PACKET3_GET_OPCODE(header);
break;
case PACKET_TYPE2:
pkt->count = -1;
break;
default:
DRM_ERROR("Unknown packet type %d at %d !\n", pkt->type, idx);
return -EINVAL;
}
if ((pkt->count + 1 + pkt->idx) >= ib_chunk->length_dw) {
DRM_ERROR("Packet (%d:%d:%d) end after CS buffer (%d) !\n",
pkt->idx, pkt->type, pkt->count, ib_chunk->length_dw);
return -EINVAL;
}
return 0;
}
/**
* r600_cs_packet_next_reloc_mm() - parse next packet which should be reloc packet3
* @parser: parser structure holding parsing context.
* @data: pointer to relocation data
* @offset_start: starting offset
* @offset_mask: offset mask (to align start offset on)
* @reloc: reloc informations
*
* Check next packet is relocation packet3, do bo validation and compute
* GPU offset using the provided start.
**/
static int r600_cs_packet_next_reloc_mm(struct radeon_cs_parser *p,
struct radeon_cs_reloc **cs_reloc)
{
struct radeon_cs_chunk *relocs_chunk;
struct radeon_cs_packet p3reloc;
unsigned idx;
int r;
if (p->chunk_relocs_idx == -1) {
DRM_ERROR("No relocation chunk !\n");
return -EINVAL;
}
*cs_reloc = NULL;
relocs_chunk = &p->chunks[p->chunk_relocs_idx];
r = r600_cs_packet_parse(p, &p3reloc, p->idx);
if (r) {
return r;
}
p->idx += p3reloc.count + 2;
if (p3reloc.type != PACKET_TYPE3 || p3reloc.opcode != PACKET3_NOP) {
DRM_ERROR("No packet3 for relocation for packet at %d.\n",
p3reloc.idx);
return -EINVAL;
}
drm/radeon/kms: don't require up to 64k allocations. (v2) This avoids needing to do a kmalloc > PAGE_SIZE for the main indirect buffer chunk, it adds an accessor for all reads from the chunk and caches a single page at a time for subsequent reads. changes since v1: Use a two page pool which should be the most common case a single packet spanning > PAGE_SIZE will be hit, but I'm having trouble seeing anywhere we currently generate anything like that. hopefully proper short page copying at end added parser_error flag to set deep errors instead of having to test every ib value fetch. fixed bug in patch that went to list. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-23 06:56:27 +00:00
idx = radeon_get_ib_value(p, p3reloc.idx + 1);
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
if (idx >= relocs_chunk->length_dw) {
DRM_ERROR("Relocs at %d after relocations chunk end %d !\n",
idx, relocs_chunk->length_dw);
return -EINVAL;
}
/* FIXME: we assume reloc size is 4 dwords */
*cs_reloc = p->relocs_ptr[(idx / 4)];
return 0;
}
/**
* r600_cs_packet_next_reloc_nomm() - parse next packet which should be reloc packet3
* @parser: parser structure holding parsing context.
* @data: pointer to relocation data
* @offset_start: starting offset
* @offset_mask: offset mask (to align start offset on)
* @reloc: reloc informations
*
* Check next packet is relocation packet3, do bo validation and compute
* GPU offset using the provided start.
**/
static int r600_cs_packet_next_reloc_nomm(struct radeon_cs_parser *p,
struct radeon_cs_reloc **cs_reloc)
{
struct radeon_cs_chunk *relocs_chunk;
struct radeon_cs_packet p3reloc;
unsigned idx;
int r;
if (p->chunk_relocs_idx == -1) {
DRM_ERROR("No relocation chunk !\n");
return -EINVAL;
}
*cs_reloc = NULL;
relocs_chunk = &p->chunks[p->chunk_relocs_idx];
r = r600_cs_packet_parse(p, &p3reloc, p->idx);
if (r) {
return r;
}
p->idx += p3reloc.count + 2;
if (p3reloc.type != PACKET_TYPE3 || p3reloc.opcode != PACKET3_NOP) {
DRM_ERROR("No packet3 for relocation for packet at %d.\n",
p3reloc.idx);
return -EINVAL;
}
drm/radeon/kms: don't require up to 64k allocations. (v2) This avoids needing to do a kmalloc > PAGE_SIZE for the main indirect buffer chunk, it adds an accessor for all reads from the chunk and caches a single page at a time for subsequent reads. changes since v1: Use a two page pool which should be the most common case a single packet spanning > PAGE_SIZE will be hit, but I'm having trouble seeing anywhere we currently generate anything like that. hopefully proper short page copying at end added parser_error flag to set deep errors instead of having to test every ib value fetch. fixed bug in patch that went to list. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-23 06:56:27 +00:00
idx = radeon_get_ib_value(p, p3reloc.idx + 1);
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
if (idx >= relocs_chunk->length_dw) {
DRM_ERROR("Relocs at %d after relocations chunk end %d !\n",
idx, relocs_chunk->length_dw);
return -EINVAL;
}
drivers/gpu: Use kzalloc for allocating only one thing Use kzalloc rather than kcalloc(1,...) The use of the allocated memory that looks like an array is &p->relocs[0], but this should be the same as p->relocs. The semantic patch that makes this change is as follows: (http://coccinelle.lip6.fr/) // <smpl> @@ @@ - kcalloc(1, + kzalloc( ...) // </smpl> Signed-off-by: Julia Lawall <julia@diku.dk> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-12-19 07:16:33 +00:00
*cs_reloc = p->relocs;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
(*cs_reloc)->lobj.gpu_offset = (u64)relocs_chunk->kdata[idx + 3] << 32;
(*cs_reloc)->lobj.gpu_offset |= relocs_chunk->kdata[idx + 0];
return 0;
}
drm/radeon: r6xx/r7xx possible security issue, system ram access This patch workaround a possible security issue which can allow user to abuse drm on r6xx/r7xx hw to access any system ram memory. This patch doesn't break userspace, it detect "valid" old use of CB_COLOR[0-7]_FRAG & CB_COLOR[0-7]_TILE registers and overwritte the address these registers are pointing to with the one of the last color buffer. This workaround will work for old mesa & xf86-video-ati and any old user which did use similar register programming pattern as those (we expect that there is no others user of those ioctl except possibly a malicious one). This patch add a warning if it detects such usage, warning encourage people to update their mesa & xf86-video-ati. New userspace will submit proper relocation. Fix for xf86-video-ati / mesa (this kernel patch is enough to prevent abuse, fix for userspace are to set proper cs stream and avoid kernel warning) : http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=95d63e408cc88b6934bec84a0b1ef94dfe8bee7b http://cgit.freedesktop.org/mesa/mesa/commit/?id=46dc6fd3ed5ef96cda53641a97bc68c3bc104a9f Abusing this register to perform system ram memory is not easy, here is outline on how it could be achieve. First attacker must have access to the drm device and be able to submit command stream throught cs ioctl. Then attacker must build a proper command stream for r6xx/r7xx hw which will abuse the FRAG or TILE buffer to overwrite the GPU GART which is in VRAM. To achieve so attacker as to setup CB_COLOR[0-7]_FRAG or CB_COLOR[0-7]_TILE to point to the GPU GART, then it has to find a way to write predictable value into those buffer (with little cleverness i believe this can be done but this is an hard task). Once attacker have such program it can overwritte GPU GART to program GPU gart to point anywhere in system memory. It then can reusse same method as he used to reprogram GART to overwritte the system ram through the GART mapping. In the process the attacker has to be carefull to not overwritte any sensitive area of the GART table, like ring or IB gart entry as it will more then likely lead to GPU lockup. Bottom line is that i think it's very hard to use this flaw to get system ram access but in theory one can achieve so. Side note: I am not aware of anyone ever using the GPU as an attack vector, nevertheless we take great care in the opensource driver to try to detect and forbid malicious use of GPU. I don't think the closed source driver are as cautious as we are. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2010-01-18 12:01:36 +00:00
/**
* r600_cs_packet_next_is_pkt3_nop() - test if next packet is packet3 nop for reloc
* @parser: parser structure holding parsing context.
*
* Check next packet is relocation packet3, do bo validation and compute
* GPU offset using the provided start.
**/
drm/radeon: Drop inlines from evergreen_cs.c / r600_cs.c Fixes evergreen_cs_parse 4080 23124 +19044 and others compared to a non force inline kernel. Signed-off-by: Andi Kleen <ak@linux.intel.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-13 23:08:41 +00:00
static int r600_cs_packet_next_is_pkt3_nop(struct radeon_cs_parser *p)
drm/radeon: r6xx/r7xx possible security issue, system ram access This patch workaround a possible security issue which can allow user to abuse drm on r6xx/r7xx hw to access any system ram memory. This patch doesn't break userspace, it detect "valid" old use of CB_COLOR[0-7]_FRAG & CB_COLOR[0-7]_TILE registers and overwritte the address these registers are pointing to with the one of the last color buffer. This workaround will work for old mesa & xf86-video-ati and any old user which did use similar register programming pattern as those (we expect that there is no others user of those ioctl except possibly a malicious one). This patch add a warning if it detects such usage, warning encourage people to update their mesa & xf86-video-ati. New userspace will submit proper relocation. Fix for xf86-video-ati / mesa (this kernel patch is enough to prevent abuse, fix for userspace are to set proper cs stream and avoid kernel warning) : http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=95d63e408cc88b6934bec84a0b1ef94dfe8bee7b http://cgit.freedesktop.org/mesa/mesa/commit/?id=46dc6fd3ed5ef96cda53641a97bc68c3bc104a9f Abusing this register to perform system ram memory is not easy, here is outline on how it could be achieve. First attacker must have access to the drm device and be able to submit command stream throught cs ioctl. Then attacker must build a proper command stream for r6xx/r7xx hw which will abuse the FRAG or TILE buffer to overwrite the GPU GART which is in VRAM. To achieve so attacker as to setup CB_COLOR[0-7]_FRAG or CB_COLOR[0-7]_TILE to point to the GPU GART, then it has to find a way to write predictable value into those buffer (with little cleverness i believe this can be done but this is an hard task). Once attacker have such program it can overwritte GPU GART to program GPU gart to point anywhere in system memory. It then can reusse same method as he used to reprogram GART to overwritte the system ram through the GART mapping. In the process the attacker has to be carefull to not overwritte any sensitive area of the GART table, like ring or IB gart entry as it will more then likely lead to GPU lockup. Bottom line is that i think it's very hard to use this flaw to get system ram access but in theory one can achieve so. Side note: I am not aware of anyone ever using the GPU as an attack vector, nevertheless we take great care in the opensource driver to try to detect and forbid malicious use of GPU. I don't think the closed source driver are as cautious as we are. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2010-01-18 12:01:36 +00:00
{
struct radeon_cs_packet p3reloc;
int r;
r = r600_cs_packet_parse(p, &p3reloc, p->idx);
if (r) {
return 0;
}
if (p3reloc.type != PACKET_TYPE3 || p3reloc.opcode != PACKET3_NOP) {
return 0;
}
return 1;
}
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
/**
* r600_cs_packet_next_vline() - parse userspace VLINE packet
* @parser: parser structure holding parsing context.
*
* Userspace sends a special sequence for VLINE waits.
* PACKET0 - VLINE_START_END + value
* PACKET3 - WAIT_REG_MEM poll vline status reg
* RELOC (P3) - crtc_id in reloc.
*
* This function parses this and relocates the VLINE START END
* and WAIT_REG_MEM packets to the correct crtc.
* It also detects a switched off crtc and nulls out the
* wait in that case.
*/
static int r600_cs_packet_parse_vline(struct radeon_cs_parser *p)
{
struct drm_mode_object *obj;
struct drm_crtc *crtc;
struct radeon_crtc *radeon_crtc;
struct radeon_cs_packet p3reloc, wait_reg_mem;
int crtc_id;
int r;
uint32_t header, h_idx, reg, wait_reg_mem_info;
volatile uint32_t *ib;
drm/radeon: make the ib an inline object No need to malloc it any more. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Christian König <deathsimple@vodafone.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-05-09 13:35:02 +00:00
ib = p->ib.ptr;
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
/* parse the WAIT_REG_MEM */
r = r600_cs_packet_parse(p, &wait_reg_mem, p->idx);
if (r)
return r;
/* check its a WAIT_REG_MEM */
if (wait_reg_mem.type != PACKET_TYPE3 ||
wait_reg_mem.opcode != PACKET3_WAIT_REG_MEM) {
DRM_ERROR("vline wait missing WAIT_REG_MEM segment\n");
drm: radeon: *_cs_packet_parse_vline() cleanup Simplify the way the return value is set a number of times (mostly on error). Signed-off-by: Paul Bolle <pebolle@tiscali.nl> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-16 21:10:06 +00:00
return -EINVAL;
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
}
wait_reg_mem_info = radeon_get_ib_value(p, wait_reg_mem.idx + 1);
/* bit 4 is reg (0) or mem (1) */
if (wait_reg_mem_info & 0x10) {
DRM_ERROR("vline WAIT_REG_MEM waiting on MEM rather than REG\n");
drm: radeon: *_cs_packet_parse_vline() cleanup Simplify the way the return value is set a number of times (mostly on error). Signed-off-by: Paul Bolle <pebolle@tiscali.nl> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-16 21:10:06 +00:00
return -EINVAL;
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
}
/* waiting for value to be equal */
if ((wait_reg_mem_info & 0x7) != 0x3) {
DRM_ERROR("vline WAIT_REG_MEM function not equal\n");
drm: radeon: *_cs_packet_parse_vline() cleanup Simplify the way the return value is set a number of times (mostly on error). Signed-off-by: Paul Bolle <pebolle@tiscali.nl> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-16 21:10:06 +00:00
return -EINVAL;
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
}
if ((radeon_get_ib_value(p, wait_reg_mem.idx + 2) << 2) != AVIVO_D1MODE_VLINE_STATUS) {
DRM_ERROR("vline WAIT_REG_MEM bad reg\n");
drm: radeon: *_cs_packet_parse_vline() cleanup Simplify the way the return value is set a number of times (mostly on error). Signed-off-by: Paul Bolle <pebolle@tiscali.nl> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-16 21:10:06 +00:00
return -EINVAL;
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
}
if (radeon_get_ib_value(p, wait_reg_mem.idx + 5) != AVIVO_D1MODE_VLINE_STAT) {
DRM_ERROR("vline WAIT_REG_MEM bad bit mask\n");
drm: radeon: *_cs_packet_parse_vline() cleanup Simplify the way the return value is set a number of times (mostly on error). Signed-off-by: Paul Bolle <pebolle@tiscali.nl> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-16 21:10:06 +00:00
return -EINVAL;
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
}
/* jump over the NOP */
r = r600_cs_packet_parse(p, &p3reloc, p->idx + wait_reg_mem.count + 2);
if (r)
return r;
h_idx = p->idx - 2;
p->idx += wait_reg_mem.count + 2;
p->idx += p3reloc.count + 2;
header = radeon_get_ib_value(p, h_idx);
crtc_id = radeon_get_ib_value(p, h_idx + 2 + 7 + 1);
drm/radeon/kms: fix vline register for second head. Both r100/r600 had this wrong, use the macro to extract the register to relocate. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-10-08 01:32:49 +00:00
reg = CP_PACKET0_GET_REG(header);
drm/radeon/kms: drop taking lock around crtc lookup. We only add/remove crtcs at driver load, you cannot remove when the GPU is running a CS packet since the fd is open, when GPU hotplugging on radeons actually is needed all this locking needs a review and I've started re-working kms core locking to deal with this better. But for now avoid long delays in CS processing when hotplug detect is happening in a different thread. this fixes a regression introduced with hotplug detection. Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-07-21 23:57:13 +00:00
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
obj = drm_mode_object_find(p->rdev->ddev, crtc_id, DRM_MODE_OBJECT_CRTC);
if (!obj) {
DRM_ERROR("cannot find crtc %d\n", crtc_id);
drm: radeon: *_cs_packet_parse_vline() cleanup Simplify the way the return value is set a number of times (mostly on error). Signed-off-by: Paul Bolle <pebolle@tiscali.nl> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-16 21:10:06 +00:00
return -EINVAL;
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
}
crtc = obj_to_crtc(obj);
radeon_crtc = to_radeon_crtc(crtc);
crtc_id = radeon_crtc->crtc_id;
if (!crtc->enabled) {
/* if the CRTC isn't enabled - we need to nop out the WAIT_REG_MEM */
ib[h_idx + 2] = PACKET2(0);
ib[h_idx + 3] = PACKET2(0);
ib[h_idx + 4] = PACKET2(0);
ib[h_idx + 5] = PACKET2(0);
ib[h_idx + 6] = PACKET2(0);
ib[h_idx + 7] = PACKET2(0);
ib[h_idx + 8] = PACKET2(0);
} else if (crtc_id == 1) {
switch (reg) {
case AVIVO_D1MODE_VLINE_START_END:
header &= ~R600_CP_PACKET0_REG_MASK;
header |= AVIVO_D2MODE_VLINE_START_END >> 2;
break;
default:
DRM_ERROR("unknown crtc reloc\n");
drm: radeon: *_cs_packet_parse_vline() cleanup Simplify the way the return value is set a number of times (mostly on error). Signed-off-by: Paul Bolle <pebolle@tiscali.nl> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-16 21:10:06 +00:00
return -EINVAL;
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
}
ib[h_idx] = header;
ib[h_idx + 4] = AVIVO_D2MODE_VLINE_STATUS >> 2;
}
drm: radeon: *_cs_packet_parse_vline() cleanup Simplify the way the return value is set a number of times (mostly on error). Signed-off-by: Paul Bolle <pebolle@tiscali.nl> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-03-16 21:10:06 +00:00
return 0;
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
static int r600_packet0_check(struct radeon_cs_parser *p,
struct radeon_cs_packet *pkt,
unsigned idx, unsigned reg)
{
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
int r;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
switch (reg) {
case AVIVO_D1MODE_VLINE_START_END:
drm/radeon/kms/r600: add support for vline relocs Provides support for anti-tearing functionality in the ddx. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2009-09-25 20:35:11 +00:00
r = r600_cs_packet_parse_vline(p);
if (r) {
DRM_ERROR("No reloc for ib[%d]=0x%04X\n",
idx, reg);
return r;
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
break;
default:
printk(KERN_ERR "Forbidden register 0x%04X in cs at %d\n",
reg, idx);
return -EINVAL;
}
return 0;
}
static int r600_cs_parse_packet0(struct radeon_cs_parser *p,
struct radeon_cs_packet *pkt)
{
unsigned reg, i;
unsigned idx;
int r;
idx = pkt->idx + 1;
reg = pkt->reg;
for (i = 0; i <= pkt->count; i++, idx++, reg += 4) {
r = r600_packet0_check(p, pkt, idx, reg);
if (r) {
return r;
}
}
return 0;
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
/**
* r600_cs_check_reg() - check if register is authorized or not
* @parser: parser structure holding parsing context
* @reg: register we are testing
* @idx: index into the cs buffer
*
* This function will test against r600_reg_safe_bm and return 0
* if register is safe. If register is not flag as safe this function
* will test it against a list of register needind special handling.
*/
drm/radeon: Drop inlines from evergreen_cs.c / r600_cs.c Fixes evergreen_cs_parse 4080 23124 +19044 and others compared to a non force inline kernel. Signed-off-by: Andi Kleen <ak@linux.intel.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-13 23:08:41 +00:00
static int r600_cs_check_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
{
struct r600_cs_track *track = (struct r600_cs_track *)p->track;
struct radeon_cs_reloc *reloc;
u32 m, i, tmp, *ib;
int r;
i = (reg >> 7);
drm/radeon: off by one in check_reg() functions This off by one range check was copy and pasted a couple places. It's not really harmful, but we should fix it anyway. Signed-off-by: Dan Carpenter <error27@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-07-27 09:53:40 +00:00
if (i >= ARRAY_SIZE(r600_reg_safe_bm)) {
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
dev_warn(p->dev, "forbidden register 0x%08x at %d\n", reg, idx);
return -EINVAL;
}
m = 1 << ((reg >> 2) & 31);
if (!(r600_reg_safe_bm[i] & m))
return 0;
drm/radeon: make the ib an inline object No need to malloc it any more. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Christian König <deathsimple@vodafone.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-05-09 13:35:02 +00:00
ib = p->ib.ptr;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
switch (reg) {
Fix common misspellings Fixes generated by 'codespell' and manually reviewed. Signed-off-by: Lucas De Marchi <lucas.demarchi@profusion.mobi>
2011-03-31 01:57:33 +00:00
/* force following reg to 0 in an attempt to disable out buffer
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
* which will need us to better understand how it works to perform
* security check on it (Jerome)
*/
case R_0288A8_SQ_ESGS_RING_ITEMSIZE:
case R_008C44_SQ_ESGS_RING_SIZE:
case R_0288B0_SQ_ESTMP_RING_ITEMSIZE:
case R_008C54_SQ_ESTMP_RING_SIZE:
case R_0288C0_SQ_FBUF_RING_ITEMSIZE:
case R_008C74_SQ_FBUF_RING_SIZE:
case R_0288B4_SQ_GSTMP_RING_ITEMSIZE:
case R_008C5C_SQ_GSTMP_RING_SIZE:
case R_0288AC_SQ_GSVS_RING_ITEMSIZE:
case R_008C4C_SQ_GSVS_RING_SIZE:
case R_0288BC_SQ_PSTMP_RING_ITEMSIZE:
case R_008C6C_SQ_PSTMP_RING_SIZE:
case R_0288C4_SQ_REDUC_RING_ITEMSIZE:
case R_008C7C_SQ_REDUC_RING_SIZE:
case R_0288B8_SQ_VSTMP_RING_ITEMSIZE:
case R_008C64_SQ_VSTMP_RING_SIZE:
case R_0288C8_SQ_GS_VERT_ITEMSIZE:
/* get value to populate the IB don't remove */
tmp =radeon_get_ib_value(p, idx);
ib[idx] = 0;
break;
drm/radeon/r6xx/r7xx: CS parser fixes - Drop some more safe regs taht userspace shouldn't hit - Constant base regs need relocs. This allows us to use constant buffers rather than the constant register file. Also we don't want userspace to be able to set arbitrary mc base values for the const caches. - Track SQ_CONFIG so we know whether userspace is using the cfile or constant buffers. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-03-26 18:52:32 +00:00
case SQ_CONFIG:
track->sq_config = radeon_get_ib_value(p, idx);
break;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
case R_028800_DB_DEPTH_CONTROL:
track->db_depth_control = radeon_get_ib_value(p, idx);
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->db_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
case R_028010_DB_DEPTH_INFO:
drm/radeon: GPU virtual memory support v22 Virtual address space are per drm client (opener of /dev/drm). Client are in charge of virtual address space, they need to map bo into it by calling DRM_RADEON_GEM_VA ioctl. First 16M of virtual address space is reserved by the kernel. Once using 2 level page table we should be able to have a small vram memory footprint for each pt (there would be one pt for all gart, one for all vram and then one first level for each virtual address space). Plan include using the sub allocator for a common vm page table area and using memcpy to copy vm page table in & out. Or use a gart object and copy things in & out using dma. v2: agd5f fixes: - Add vram base offset for vram pages. The GPU physical address of a vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete cards and the physical bus address of the stolen memory on integrated chips. - VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1 v3: agd5f: - integrate with the semaphore/multi-ring stuff v4: - rebase on top ttm dma & multi-ring stuff - userspace is now in charge of the address space - no more specific cs vm ioctl, instead cs ioctl has a new chunk v5: - properly handle mem == NULL case from move_notify callback - fix the vm cleanup path v6: - fix update of page table to only happen on valid mem placement v7: - add tlb flush for each vm context - add flags to define mapping property (readable, writeable, snooped) - make ring id implicit from ib->fence->ring, up to each asic callback to then do ring specific scheduling if vm ib scheduling function v8: - add query for ib limit and kernel reserved virtual space - rename vm->size to max_pfn (maximum number of page) - update gem_va ioctl to also allow unmap operation - bump kernel version to allow userspace to query for vm support v9: - rebuild page table only when bind and incrementaly depending on bo referenced by cs and that have been moved - allow virtual address space to grow - use sa allocator for vram page table - return invalid when querying vm limit on non cayman GPU - dump vm fault register on lockup v10: agd5f: - Move the vm schedule_ib callback to a standalone function, remove the callback and use the existing ib_execute callback for VM IBs. v11: - rebase on top of lastest Linus v12: agd5f: - remove spurious backslash - set IB vm_id to 0 in radeon_ib_get() v13: agd5f: - fix handling of RADEON_CHUNK_ID_FLAGS v14: - fix va destruction - fix suspend resume - forbid bo to have several different va in same vm v15: - rebase v16: - cleanup left over of vm init/fini v17: agd5f: - cs checker v18: agd5f: - reworks the CS ioctl to better support multiple rings and VM. Rather than adding a new chunk id for VM, just re-use the IB chunk id and add a new flags for VM mode. Also define additional dwords for the flags chunk id to define the what ring we want to use (gfx, compute, uvd, etc.) and the priority. v19: - fix cs fini in weird case of no ib - semi working flush fix for ni - rebase on top of sa allocator changes v20: agd5f: - further CS ioctl cleanups from Christian's comments v21: agd5f: - integrate CS checker improvements v22: agd5f: - final cleanups for release, only allow VM CS on cayman Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 03:11:05 +00:00
if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS) &&
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
r600_cs_packet_next_is_pkt3_nop(p)) {
drm/radeon/kms: add tiling support to the cs checker for r6xx/r7xx Check for relocs for DB_DEPTH_INFO, CB_COLOR*_INFO, and texture resources. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-05-20 16:43:52 +00:00
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
track->db_depth_info = radeon_get_ib_value(p, idx);
ib[idx] &= C_028010_ARRAY_MODE;
track->db_depth_info &= C_028010_ARRAY_MODE;
if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
ib[idx] |= S_028010_ARRAY_MODE(V_028010_ARRAY_2D_TILED_THIN1);
track->db_depth_info |= S_028010_ARRAY_MODE(V_028010_ARRAY_2D_TILED_THIN1);
} else {
ib[idx] |= S_028010_ARRAY_MODE(V_028010_ARRAY_1D_TILED_THIN1);
track->db_depth_info |= S_028010_ARRAY_MODE(V_028010_ARRAY_1D_TILED_THIN1);
}
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
} else {
drm/radeon/kms: add tiling support to the cs checker for r6xx/r7xx Check for relocs for DB_DEPTH_INFO, CB_COLOR*_INFO, and texture resources. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-05-20 16:43:52 +00:00
track->db_depth_info = radeon_get_ib_value(p, idx);
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
}
track->db_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
case R_028004_DB_DEPTH_VIEW:
track->db_depth_view = radeon_get_ib_value(p, idx);
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->db_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
case R_028000_DB_DEPTH_SIZE:
track->db_depth_size = radeon_get_ib_value(p, idx);
track->db_depth_size_idx = idx;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->db_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
case R_028AB0_VGT_STRMOUT_EN:
track->vgt_strmout_en = radeon_get_ib_value(p, idx);
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->streamout_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
case R_028B20_VGT_STRMOUT_BUFFER_EN:
track->vgt_strmout_buffer_en = radeon_get_ib_value(p, idx);
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->streamout_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
case VGT_STRMOUT_BUFFER_BASE_0:
case VGT_STRMOUT_BUFFER_BASE_1:
case VGT_STRMOUT_BUFFER_BASE_2:
case VGT_STRMOUT_BUFFER_BASE_3:
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
tmp = (reg - VGT_STRMOUT_BUFFER_BASE_0) / 16;
track->vgt_strmout_bo_offset[tmp] = radeon_get_ib_value(p, idx) << 8;
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->vgt_strmout_bo[tmp] = reloc->robj;
track->vgt_strmout_bo_mc[tmp] = reloc->lobj.gpu_offset;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->streamout_dirty = true;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
break;
case VGT_STRMOUT_BUFFER_SIZE_0:
case VGT_STRMOUT_BUFFER_SIZE_1:
case VGT_STRMOUT_BUFFER_SIZE_2:
case VGT_STRMOUT_BUFFER_SIZE_3:
tmp = (reg - VGT_STRMOUT_BUFFER_SIZE_0) / 16;
/* size in register is DWs, convert to bytes */
track->vgt_strmout_size[tmp] = radeon_get_ib_value(p, idx) * 4;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->streamout_dirty = true;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
break;
case CP_COHER_BASE:
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "missing reloc for CP_COHER_BASE "
"0x%04X\n", reg);
return -EINVAL;
}
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
break;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
case R_028238_CB_TARGET_MASK:
track->cb_target_mask = radeon_get_ib_value(p, idx);
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->cb_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
case R_02823C_CB_SHADER_MASK:
track->cb_shader_mask = radeon_get_ib_value(p, idx);
break;
case R_028C04_PA_SC_AA_CONFIG:
tmp = G_028C04_MSAA_NUM_SAMPLES(radeon_get_ib_value(p, idx));
drm/radeon: allow CMASK and FMASK in the CS checker on r600-r700 MSAA is impossible without them. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
2012-08-19 00:22:09 +00:00
track->log_nsamples = tmp;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
track->nsamples = 1 << tmp;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->cb_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
drm/radeon: add proper checking of RESOLVE_BOX command for r600-r700 Checking of the second colorbuffer was skipped on r700, because CB_TARGET_MASK was 0xf. With r600, CB_TARGET_MASK is changed to 0xff, so we must set the number of samples of the second colorbuffer to 1 in order to pass the CS checker. The DRM version is bumped, because RESOLVE_BOX is always rejected without this fix on r600. Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-24 12:27:36 +00:00
case R_028808_CB_COLOR_CONTROL:
tmp = G_028808_SPECIAL_OP(radeon_get_ib_value(p, idx));
track->is_resolve = tmp == V_028808_SPECIAL_RESOLVE_BOX;
track->cb_dirty = true;
break;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
case R_0280A0_CB_COLOR0_INFO:
case R_0280A4_CB_COLOR1_INFO:
case R_0280A8_CB_COLOR2_INFO:
case R_0280AC_CB_COLOR3_INFO:
case R_0280B0_CB_COLOR4_INFO:
case R_0280B4_CB_COLOR5_INFO:
case R_0280B8_CB_COLOR6_INFO:
case R_0280BC_CB_COLOR7_INFO:
drm/radeon: GPU virtual memory support v22 Virtual address space are per drm client (opener of /dev/drm). Client are in charge of virtual address space, they need to map bo into it by calling DRM_RADEON_GEM_VA ioctl. First 16M of virtual address space is reserved by the kernel. Once using 2 level page table we should be able to have a small vram memory footprint for each pt (there would be one pt for all gart, one for all vram and then one first level for each virtual address space). Plan include using the sub allocator for a common vm page table area and using memcpy to copy vm page table in & out. Or use a gart object and copy things in & out using dma. v2: agd5f fixes: - Add vram base offset for vram pages. The GPU physical address of a vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete cards and the physical bus address of the stolen memory on integrated chips. - VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1 v3: agd5f: - integrate with the semaphore/multi-ring stuff v4: - rebase on top ttm dma & multi-ring stuff - userspace is now in charge of the address space - no more specific cs vm ioctl, instead cs ioctl has a new chunk v5: - properly handle mem == NULL case from move_notify callback - fix the vm cleanup path v6: - fix update of page table to only happen on valid mem placement v7: - add tlb flush for each vm context - add flags to define mapping property (readable, writeable, snooped) - make ring id implicit from ib->fence->ring, up to each asic callback to then do ring specific scheduling if vm ib scheduling function v8: - add query for ib limit and kernel reserved virtual space - rename vm->size to max_pfn (maximum number of page) - update gem_va ioctl to also allow unmap operation - bump kernel version to allow userspace to query for vm support v9: - rebuild page table only when bind and incrementaly depending on bo referenced by cs and that have been moved - allow virtual address space to grow - use sa allocator for vram page table - return invalid when querying vm limit on non cayman GPU - dump vm fault register on lockup v10: agd5f: - Move the vm schedule_ib callback to a standalone function, remove the callback and use the existing ib_execute callback for VM IBs. v11: - rebase on top of lastest Linus v12: agd5f: - remove spurious backslash - set IB vm_id to 0 in radeon_ib_get() v13: agd5f: - fix handling of RADEON_CHUNK_ID_FLAGS v14: - fix va destruction - fix suspend resume - forbid bo to have several different va in same vm v15: - rebase v16: - cleanup left over of vm init/fini v17: agd5f: - cs checker v18: agd5f: - reworks the CS ioctl to better support multiple rings and VM. Rather than adding a new chunk id for VM, just re-use the IB chunk id and add a new flags for VM mode. Also define additional dwords for the flags chunk id to define the what ring we want to use (gfx, compute, uvd, etc.) and the priority. v19: - fix cs fini in weird case of no ib - semi working flush fix for ni - rebase on top of sa allocator changes v20: agd5f: - further CS ioctl cleanups from Christian's comments v21: agd5f: - integrate CS checker improvements v22: agd5f: - final cleanups for release, only allow VM CS on cayman Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 03:11:05 +00:00
if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS) &&
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
r600_cs_packet_next_is_pkt3_nop(p)) {
drm/radeon/kms: add tiling support to the cs checker for r6xx/r7xx Check for relocs for DB_DEPTH_INFO, CB_COLOR*_INFO, and texture resources. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-05-20 16:43:52 +00:00
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_err(p->dev, "bad SET_CONTEXT_REG 0x%04X\n", reg);
return -EINVAL;
}
tmp = (reg - R_0280A0_CB_COLOR0_INFO) / 4;
track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO) {
ib[idx] |= S_0280A0_ARRAY_MODE(V_0280A0_ARRAY_2D_TILED_THIN1);
track->cb_color_info[tmp] |= S_0280A0_ARRAY_MODE(V_0280A0_ARRAY_2D_TILED_THIN1);
} else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO) {
ib[idx] |= S_0280A0_ARRAY_MODE(V_0280A0_ARRAY_1D_TILED_THIN1);
track->cb_color_info[tmp] |= S_0280A0_ARRAY_MODE(V_0280A0_ARRAY_1D_TILED_THIN1);
}
} else {
tmp = (reg - R_0280A0_CB_COLOR0_INFO) / 4;
track->cb_color_info[tmp] = radeon_get_ib_value(p, idx);
}
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->cb_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
case R_028080_CB_COLOR0_VIEW:
case R_028084_CB_COLOR1_VIEW:
case R_028088_CB_COLOR2_VIEW:
case R_02808C_CB_COLOR3_VIEW:
case R_028090_CB_COLOR4_VIEW:
case R_028094_CB_COLOR5_VIEW:
case R_028098_CB_COLOR6_VIEW:
case R_02809C_CB_COLOR7_VIEW:
tmp = (reg - R_028080_CB_COLOR0_VIEW) / 4;
track->cb_color_view[tmp] = radeon_get_ib_value(p, idx);
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->cb_dirty = true;
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
break;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
case R_028060_CB_COLOR0_SIZE:
case R_028064_CB_COLOR1_SIZE:
case R_028068_CB_COLOR2_SIZE:
case R_02806C_CB_COLOR3_SIZE:
case R_028070_CB_COLOR4_SIZE:
case R_028074_CB_COLOR5_SIZE:
case R_028078_CB_COLOR6_SIZE:
case R_02807C_CB_COLOR7_SIZE:
tmp = (reg - R_028060_CB_COLOR0_SIZE) / 4;
track->cb_color_size[tmp] = radeon_get_ib_value(p, idx);
track->cb_color_size_idx[tmp] = idx;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->cb_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
/* This register were added late, there is userspace
* which does provide relocation for those but set
* 0 offset. In order to avoid breaking old userspace
* we detect this and set address to point to last
* CB_COLOR0_BASE, note that if userspace doesn't set
* CB_COLOR0_BASE before this register we will report
* error. Old userspace always set CB_COLOR0_BASE
* before any of this.
*/
case R_0280E0_CB_COLOR0_FRAG:
case R_0280E4_CB_COLOR1_FRAG:
case R_0280E8_CB_COLOR2_FRAG:
case R_0280EC_CB_COLOR3_FRAG:
case R_0280F0_CB_COLOR4_FRAG:
case R_0280F4_CB_COLOR5_FRAG:
case R_0280F8_CB_COLOR6_FRAG:
case R_0280FC_CB_COLOR7_FRAG:
tmp = (reg - R_0280E0_CB_COLOR0_FRAG) / 4;
if (!r600_cs_packet_next_is_pkt3_nop(p)) {
if (!track->cb_color_base_last[tmp]) {
dev_err(p->dev, "Broken old userspace ? no cb_color0_base supplied before trying to write 0x%08X\n", reg);
return -EINVAL;
}
track->cb_color_frag_bo[tmp] = track->cb_color_bo[tmp];
drm/radeon: allow CMASK and FMASK in the CS checker on r600-r700 MSAA is impossible without them. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
2012-08-19 00:22:09 +00:00
track->cb_color_frag_offset[tmp] = track->cb_color_bo_offset[tmp];
ib[idx] = track->cb_color_base_last[tmp];
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
} else {
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_err(p->dev, "bad SET_CONTEXT_REG 0x%04X\n", reg);
return -EINVAL;
}
track->cb_color_frag_bo[tmp] = reloc->robj;
drm/radeon: allow CMASK and FMASK in the CS checker on r600-r700 MSAA is impossible without them. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
2012-08-19 00:22:09 +00:00
track->cb_color_frag_offset[tmp] = (u64)ib[idx] << 8;
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
}
if (G_0280A0_TILE_MODE(track->cb_color_info[tmp])) {
track->cb_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
break;
case R_0280C0_CB_COLOR0_TILE:
case R_0280C4_CB_COLOR1_TILE:
case R_0280C8_CB_COLOR2_TILE:
case R_0280CC_CB_COLOR3_TILE:
case R_0280D0_CB_COLOR4_TILE:
case R_0280D4_CB_COLOR5_TILE:
case R_0280D8_CB_COLOR6_TILE:
case R_0280DC_CB_COLOR7_TILE:
tmp = (reg - R_0280C0_CB_COLOR0_TILE) / 4;
if (!r600_cs_packet_next_is_pkt3_nop(p)) {
if (!track->cb_color_base_last[tmp]) {
dev_err(p->dev, "Broken old userspace ? no cb_color0_base supplied before trying to write 0x%08X\n", reg);
return -EINVAL;
}
track->cb_color_tile_bo[tmp] = track->cb_color_bo[tmp];
drm/radeon: allow CMASK and FMASK in the CS checker on r600-r700 MSAA is impossible without them. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
2012-08-19 00:22:09 +00:00
track->cb_color_tile_offset[tmp] = track->cb_color_bo_offset[tmp];
ib[idx] = track->cb_color_base_last[tmp];
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
} else {
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_err(p->dev, "bad SET_CONTEXT_REG 0x%04X\n", reg);
return -EINVAL;
}
track->cb_color_tile_bo[tmp] = reloc->robj;
drm/radeon: allow CMASK and FMASK in the CS checker on r600-r700 MSAA is impossible without them. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
2012-08-19 00:22:09 +00:00
track->cb_color_tile_offset[tmp] = (u64)ib[idx] << 8;
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
}
if (G_0280A0_TILE_MODE(track->cb_color_info[tmp])) {
track->cb_dirty = true;
}
break;
case R_028100_CB_COLOR0_MASK:
case R_028104_CB_COLOR1_MASK:
case R_028108_CB_COLOR2_MASK:
case R_02810C_CB_COLOR3_MASK:
case R_028110_CB_COLOR4_MASK:
case R_028114_CB_COLOR5_MASK:
case R_028118_CB_COLOR6_MASK:
case R_02811C_CB_COLOR7_MASK:
tmp = (reg - R_028100_CB_COLOR0_MASK) / 4;
drm/radeon: fix reading CB_COLORn_MASK from the CS Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-22 15:02:42 +00:00
track->cb_color_mask[tmp] = radeon_get_ib_value(p, idx);
drm/radeon: allow CMASK and FMASK in the CS checker on r600-r700 MSAA is impossible without them. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com>
2012-08-19 00:22:09 +00:00
if (G_0280A0_TILE_MODE(track->cb_color_info[tmp])) {
track->cb_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
break;
case CB_COLOR0_BASE:
case CB_COLOR1_BASE:
case CB_COLOR2_BASE:
case CB_COLOR3_BASE:
case CB_COLOR4_BASE:
case CB_COLOR5_BASE:
case CB_COLOR6_BASE:
case CB_COLOR7_BASE:
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
drm/radeon/kms: fix r600/r700 cs checker to avoid double kfree radeon_cs kfree the tracker structure but for r6xx/r7xx we want to kfree it inside the parse function because we share it with the UMS code path. Set tracker to NULL after freeing it will avoid double free. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-11 11:44:32 +00:00
tmp = (reg - CB_COLOR0_BASE) / 4;
drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Andre Maasikas <amaasikas@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-06 06:54:05 +00:00
track->cb_color_bo_offset[tmp] = radeon_get_ib_value(p, idx) << 8;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->cb_color_base_last[tmp] = ib[idx];
track->cb_color_bo[tmp] = reloc->robj;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
track->cb_color_bo_mc[tmp] = reloc->lobj.gpu_offset;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->cb_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
case DB_DEPTH_BASE:
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Andre Maasikas <amaasikas@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-06 06:54:05 +00:00
track->db_offset = radeon_get_ib_value(p, idx) << 8;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->db_bo = reloc->robj;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
track->db_bo_mc = reloc->lobj.gpu_offset;
drm/radeon/kms: skip db/cb/streamout checking when possible on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:38 +00:00
track->db_dirty = true;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
case DB_HTILE_DATA_BASE:
drm/radeon/kms: add htile support to the cs checker v3 For 6xx+. Required for mesa to use htile support for HiZ/HiS. Userspace will check radeon version 2.14 with is bumped either by tiling patch or stream out patch. This patch only add support for htile relocation which should be enough for any userspace to implement the hyperz (using htile buffer) feature. v2: Jerome: Fix size checking for htile buffer. v3: Jerome: Adapt on top of r600/evergreen cs checker changes, also check htile surface in case only stencil is present. Signed-off-by: Pierre-Eric Pelloux-Prayer <pelloux@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-21 23:18:21 +00:00
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
}
track->htile_offset = radeon_get_ib_value(p, idx) << 8;
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
track->htile_bo = reloc->robj;
track->db_dirty = true;
break;
case DB_HTILE_SURFACE:
track->htile_surface = radeon_get_ib_value(p, idx);
track->db_dirty = true;
break;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
case SQ_PGM_START_FS:
case SQ_PGM_START_ES:
case SQ_PGM_START_VS:
case SQ_PGM_START_GS:
case SQ_PGM_START_PS:
drm/radeon/r6xx/r7xx: CS parser fixes - Drop some more safe regs taht userspace shouldn't hit - Constant base regs need relocs. This allows us to use constant buffers rather than the constant register file. Also we don't want userspace to be able to set arbitrary mc base values for the const caches. - Track SQ_CONFIG so we know whether userspace is using the cfile or constant buffers. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-03-26 18:52:32 +00:00
case SQ_ALU_CONST_CACHE_GS_0:
case SQ_ALU_CONST_CACHE_GS_1:
case SQ_ALU_CONST_CACHE_GS_2:
case SQ_ALU_CONST_CACHE_GS_3:
case SQ_ALU_CONST_CACHE_GS_4:
case SQ_ALU_CONST_CACHE_GS_5:
case SQ_ALU_CONST_CACHE_GS_6:
case SQ_ALU_CONST_CACHE_GS_7:
case SQ_ALU_CONST_CACHE_GS_8:
case SQ_ALU_CONST_CACHE_GS_9:
case SQ_ALU_CONST_CACHE_GS_10:
case SQ_ALU_CONST_CACHE_GS_11:
case SQ_ALU_CONST_CACHE_GS_12:
case SQ_ALU_CONST_CACHE_GS_13:
case SQ_ALU_CONST_CACHE_GS_14:
case SQ_ALU_CONST_CACHE_GS_15:
case SQ_ALU_CONST_CACHE_PS_0:
case SQ_ALU_CONST_CACHE_PS_1:
case SQ_ALU_CONST_CACHE_PS_2:
case SQ_ALU_CONST_CACHE_PS_3:
case SQ_ALU_CONST_CACHE_PS_4:
case SQ_ALU_CONST_CACHE_PS_5:
case SQ_ALU_CONST_CACHE_PS_6:
case SQ_ALU_CONST_CACHE_PS_7:
case SQ_ALU_CONST_CACHE_PS_8:
case SQ_ALU_CONST_CACHE_PS_9:
case SQ_ALU_CONST_CACHE_PS_10:
case SQ_ALU_CONST_CACHE_PS_11:
case SQ_ALU_CONST_CACHE_PS_12:
case SQ_ALU_CONST_CACHE_PS_13:
case SQ_ALU_CONST_CACHE_PS_14:
case SQ_ALU_CONST_CACHE_PS_15:
case SQ_ALU_CONST_CACHE_VS_0:
case SQ_ALU_CONST_CACHE_VS_1:
case SQ_ALU_CONST_CACHE_VS_2:
case SQ_ALU_CONST_CACHE_VS_3:
case SQ_ALU_CONST_CACHE_VS_4:
case SQ_ALU_CONST_CACHE_VS_5:
case SQ_ALU_CONST_CACHE_VS_6:
case SQ_ALU_CONST_CACHE_VS_7:
case SQ_ALU_CONST_CACHE_VS_8:
case SQ_ALU_CONST_CACHE_VS_9:
case SQ_ALU_CONST_CACHE_VS_10:
case SQ_ALU_CONST_CACHE_VS_11:
case SQ_ALU_CONST_CACHE_VS_12:
case SQ_ALU_CONST_CACHE_VS_13:
case SQ_ALU_CONST_CACHE_VS_14:
case SQ_ALU_CONST_CACHE_VS_15:
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONTEXT_REG "
"0x%04X\n", reg);
return -EINVAL;
drm/radeon/kms: add initial CS checker support for compute - Add some new compute regs - Add new dispatch packets for evergreen/cayman Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-06-08 19:26:45 +00:00
}
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
break;
case SX_MEMORY_EXPORT_BASE:
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
dev_warn(p->dev, "bad SET_CONFIG_REG "
"0x%04X\n", reg);
return -EINVAL;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
ib[idx] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
break;
drm/radeon/kms: skip cb/db checking if SX_MISC is 1 on r600+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-07 23:56:00 +00:00
case SX_MISC:
track->sx_misc_kill_all_prims = (radeon_get_ib_value(p, idx) & 0x1) != 0;
break;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
default:
dev_warn(p->dev, "forbidden register 0x%08x at %d\n", reg, idx);
return -EINVAL;
}
return 0;
}
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
unsigned r600_mip_minify(unsigned size, unsigned level)
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
{
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
unsigned val;
val = max(1U, size >> level);
if (level > 0)
val = roundup_pow_of_two(val);
return val;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
static void r600_texture_size(unsigned nfaces, unsigned blevel, unsigned llevel,
drm/radeon: fix checking of MSAA renderbuffers on r600-r700 The MSAA checking was mostly unimplemented on r600-r700. The userspace submits GPU commands and the kernel driver computes how much memory the GPU will access and checks if it's all within buffer bounds the userspace allocated. This patch fixes the computations of the size of MSAA surfaces in memory. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-19 19:23:26 +00:00
unsigned w0, unsigned h0, unsigned d0, unsigned nsamples, unsigned format,
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
unsigned block_align, unsigned height_align, unsigned base_align,
drm/radeon/kms: fix CS alignment checking for tiling (v2) Covers depth, cb, and textures. Hopefully I got this right. v2: - fix bugs: https://bugs.freedesktop.org/show_bug.cgi?id=28327 https://bugs.freedesktop.org/show_bug.cgi?id=28381 - use ALIGNED(), IS_ALIGNED() macros Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-06-04 22:41:42 +00:00
unsigned *l0_size, unsigned *mipmap_size)
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
{
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
unsigned offset, i, level;
unsigned width, height, depth, size;
unsigned blocksize;
unsigned nbx, nby;
unsigned nlevels = llevel - blevel + 1;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
*l0_size = -1;
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
blocksize = r600_fmt_get_blocksize(format);
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
w0 = r600_mip_minify(w0, 0);
h0 = r600_mip_minify(h0, 0);
d0 = r600_mip_minify(d0, 0);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
for(i = 0, offset = 0, level = blevel; i < nlevels; i++, level++) {
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
width = r600_mip_minify(w0, i);
nbx = r600_fmt_get_nblocksx(format, width);
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
nbx = round_up(nbx, block_align);
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
height = r600_mip_minify(h0, i);
nby = r600_fmt_get_nblocksy(format, height);
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
nby = round_up(nby, height_align);
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
depth = r600_mip_minify(d0, i);
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
drm/radeon: fix checking of MSAA renderbuffers on r600-r700 The MSAA checking was mostly unimplemented on r600-r700. The userspace submits GPU commands and the kernel driver computes how much memory the GPU will access and checks if it's all within buffer bounds the userspace allocated. This patch fixes the computations of the size of MSAA surfaces in memory. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-19 19:23:26 +00:00
size = nbx * nby * blocksize * nsamples;
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
if (nfaces)
size *= nfaces;
else
size *= depth;
if (i == 0)
*l0_size = size;
if (i == 0 || i == 1)
offset = round_up(offset, base_align);
offset += size;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
*mipmap_size = offset;
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
if (llevel == 0)
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
*mipmap_size = *l0_size;
drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Andre Maasikas <amaasikas@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-06 06:54:05 +00:00
if (!blevel)
*mipmap_size -= *l0_size;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
/**
* r600_check_texture_resource() - check if register is authorized or not
* @p: parser structure holding parsing context
* @idx: index into the cs buffer
* @texture: texture's bo structure
* @mipmap: mipmap's bo structure
*
* This function will check that the resource has valid field and that
* the texture and mipmap bo object are big enough to cover this resource.
*/
drm/radeon: Drop inlines from evergreen_cs.c / r600_cs.c Fixes evergreen_cs_parse 4080 23124 +19044 and others compared to a non force inline kernel. Signed-off-by: Andi Kleen <ak@linux.intel.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-13 23:08:41 +00:00
static int r600_check_texture_resource(struct radeon_cs_parser *p, u32 idx,
drm/radeon/kms: add tiling support to the cs checker for r6xx/r7xx Check for relocs for DB_DEPTH_INFO, CB_COLOR*_INFO, and texture resources. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-05-20 16:43:52 +00:00
struct radeon_bo *texture,
struct radeon_bo *mipmap,
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
u64 base_offset,
u64 mip_offset,
drm/radeon/kms: add tiling support to the cs checker for r6xx/r7xx Check for relocs for DB_DEPTH_INFO, CB_COLOR*_INFO, and texture resources. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-05-20 16:43:52 +00:00
u32 tiling_flags)
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
{
drm/radeon/kms: fix CS alignment checking for tiling (v2) Covers depth, cb, and textures. Hopefully I got this right. v2: - fix bugs: https://bugs.freedesktop.org/show_bug.cgi?id=28327 https://bugs.freedesktop.org/show_bug.cgi?id=28381 - use ALIGNED(), IS_ALIGNED() macros Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-06-04 22:41:42 +00:00
struct r600_cs_track *track = p->track;
drm/radeon/kms: reorder code in r600_check_texture_resource Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-09 14:34:15 +00:00
u32 dim, nfaces, llevel, blevel, w0, h0, d0;
u32 word0, word1, l0_size, mipmap_size, word2, word3, word4, word5;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
u32 height_align, pitch, pitch_align, depth_align;
drm/radeon/kms: reorder code in r600_check_texture_resource Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-09 14:34:15 +00:00
u32 barray, larray;
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
u64 base_align;
struct array_mode_checker array_check;
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
u32 format;
drm/radeon/kms: reorder code in r600_check_texture_resource Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-09 14:34:15 +00:00
bool is_array;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
/* on legacy kernel we don't perform advanced check */
if (p->rdev == NULL)
return 0;
drm/radeon/kms: add tiling support to the cs checker for r6xx/r7xx Check for relocs for DB_DEPTH_INFO, CB_COLOR*_INFO, and texture resources. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-05-20 16:43:52 +00:00
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
/* convert to bytes */
base_offset <<= 8;
mip_offset <<= 8;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
word0 = radeon_get_ib_value(p, idx + 0);
drm/radeon: GPU virtual memory support v22 Virtual address space are per drm client (opener of /dev/drm). Client are in charge of virtual address space, they need to map bo into it by calling DRM_RADEON_GEM_VA ioctl. First 16M of virtual address space is reserved by the kernel. Once using 2 level page table we should be able to have a small vram memory footprint for each pt (there would be one pt for all gart, one for all vram and then one first level for each virtual address space). Plan include using the sub allocator for a common vm page table area and using memcpy to copy vm page table in & out. Or use a gart object and copy things in & out using dma. v2: agd5f fixes: - Add vram base offset for vram pages. The GPU physical address of a vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete cards and the physical bus address of the stolen memory on integrated chips. - VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1 v3: agd5f: - integrate with the semaphore/multi-ring stuff v4: - rebase on top ttm dma & multi-ring stuff - userspace is now in charge of the address space - no more specific cs vm ioctl, instead cs ioctl has a new chunk v5: - properly handle mem == NULL case from move_notify callback - fix the vm cleanup path v6: - fix update of page table to only happen on valid mem placement v7: - add tlb flush for each vm context - add flags to define mapping property (readable, writeable, snooped) - make ring id implicit from ib->fence->ring, up to each asic callback to then do ring specific scheduling if vm ib scheduling function v8: - add query for ib limit and kernel reserved virtual space - rename vm->size to max_pfn (maximum number of page) - update gem_va ioctl to also allow unmap operation - bump kernel version to allow userspace to query for vm support v9: - rebuild page table only when bind and incrementaly depending on bo referenced by cs and that have been moved - allow virtual address space to grow - use sa allocator for vram page table - return invalid when querying vm limit on non cayman GPU - dump vm fault register on lockup v10: agd5f: - Move the vm schedule_ib callback to a standalone function, remove the callback and use the existing ib_execute callback for VM IBs. v11: - rebase on top of lastest Linus v12: agd5f: - remove spurious backslash - set IB vm_id to 0 in radeon_ib_get() v13: agd5f: - fix handling of RADEON_CHUNK_ID_FLAGS v14: - fix va destruction - fix suspend resume - forbid bo to have several different va in same vm v15: - rebase v16: - cleanup left over of vm init/fini v17: agd5f: - cs checker v18: agd5f: - reworks the CS ioctl to better support multiple rings and VM. Rather than adding a new chunk id for VM, just re-use the IB chunk id and add a new flags for VM mode. Also define additional dwords for the flags chunk id to define the what ring we want to use (gfx, compute, uvd, etc.) and the priority. v19: - fix cs fini in weird case of no ib - semi working flush fix for ni - rebase on top of sa allocator changes v20: agd5f: - further CS ioctl cleanups from Christian's comments v21: agd5f: - integrate CS checker improvements v22: agd5f: - final cleanups for release, only allow VM CS on cayman Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 03:11:05 +00:00
if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) {
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
if (tiling_flags & RADEON_TILING_MACRO)
word0 |= S_038000_TILE_MODE(V_038000_ARRAY_2D_TILED_THIN1);
else if (tiling_flags & RADEON_TILING_MICRO)
word0 |= S_038000_TILE_MODE(V_038000_ARRAY_1D_TILED_THIN1);
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
word1 = radeon_get_ib_value(p, idx + 1);
drm/radeon/kms: reorder code in r600_check_texture_resource Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-09 14:34:15 +00:00
word2 = radeon_get_ib_value(p, idx + 2) << 8;
word3 = radeon_get_ib_value(p, idx + 3) << 8;
word4 = radeon_get_ib_value(p, idx + 4);
word5 = radeon_get_ib_value(p, idx + 5);
dim = G_038000_DIM(word0);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
w0 = G_038000_TEX_WIDTH(word0) + 1;
drm/radeon/kms: reorder code in r600_check_texture_resource Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-09 14:34:15 +00:00
pitch = (G_038000_PITCH(word0) + 1) * 8;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
h0 = G_038004_TEX_HEIGHT(word1) + 1;
d0 = G_038004_TEX_DEPTH(word1);
drm/radeon/kms: reorder code in r600_check_texture_resource Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-09 14:34:15 +00:00
format = G_038004_DATA_FORMAT(word1);
blevel = G_038010_BASE_LEVEL(word4);
llevel = G_038014_LAST_LEVEL(word5);
/* pitch in texels */
array_check.array_mode = G_038000_TILE_MODE(word0);
array_check.group_size = track->group_size;
array_check.nbanks = track->nbanks;
array_check.npipes = track->npipes;
array_check.nsamples = 1;
array_check.blocksize = r600_fmt_get_blocksize(format);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
nfaces = 1;
drm/radeon/kms: reorder code in r600_check_texture_resource Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-09 14:34:15 +00:00
is_array = false;
switch (dim) {
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
case V_038000_SQ_TEX_DIM_1D:
case V_038000_SQ_TEX_DIM_2D:
case V_038000_SQ_TEX_DIM_3D:
break;
case V_038000_SQ_TEX_DIM_CUBEMAP:
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
if (p->family >= CHIP_RV770)
nfaces = 8;
else
nfaces = 6;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
break;
case V_038000_SQ_TEX_DIM_1D_ARRAY:
case V_038000_SQ_TEX_DIM_2D_ARRAY:
drm/radeon/kms: reorder code in r600_check_texture_resource Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-09 14:34:15 +00:00
is_array = true;
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
break;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
case V_038000_SQ_TEX_DIM_2D_ARRAY_MSAA:
drm/radeon/kms: add MSAA texture support for r600-evergreen Most of the checking seems to be in place already. As you can see, log2(number of samples) resides in LAST_LEVEL. This is required for MSAA support (namely for depth-stencil resolve and blitting between MSAA resources). Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-09 14:34:16 +00:00
is_array = true;
/* fall through */
case V_038000_SQ_TEX_DIM_2D_MSAA:
array_check.nsamples = 1 << llevel;
llevel = 0;
break;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
default:
dev_warn(p->dev, "this kernel doesn't support %d texture dim\n", G_038000_DIM(word0));
return -EINVAL;
}
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
if (!r600_fmt_is_valid_texture(format, p->family)) {
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
dev_warn(p->dev, "%s:%d texture invalid format %d\n",
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
__func__, __LINE__, format);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
return -EINVAL;
}
drm/radeon/kms: fix CS alignment checking for tiling (v2) Covers depth, cb, and textures. Hopefully I got this right. v2: - fix bugs: https://bugs.freedesktop.org/show_bug.cgi?id=28327 https://bugs.freedesktop.org/show_bug.cgi?id=28381 - use ALIGNED(), IS_ALIGNED() macros Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-06-04 22:41:42 +00:00
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
if (r600_get_array_mode_alignment(&array_check,
&pitch_align, &height_align, &depth_align, &base_align)) {
dev_warn(p->dev, "%s:%d tex array mode (%d) invalid\n",
__func__, __LINE__, G_038000_TILE_MODE(word0));
return -EINVAL;
}
/* XXX check height as well... */
if (!IS_ALIGNED(pitch, pitch_align)) {
drm/radeon/kms: improve 6xx/7xx CS error output Makes debugging CS rejections much easier. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: stable@kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-13 23:42:41 +00:00
dev_warn(p->dev, "%s:%d tex pitch (%d, 0x%x, %d) invalid\n",
__func__, __LINE__, pitch, pitch_align, G_038000_TILE_MODE(word0));
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
return -EINVAL;
}
if (!IS_ALIGNED(base_offset, base_align)) {
drm/radeon/kms: improve 6xx/7xx CS error output Makes debugging CS rejections much easier. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: stable@kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-13 23:42:41 +00:00
dev_warn(p->dev, "%s:%d tex base offset (0x%llx, 0x%llx, %d) invalid\n",
__func__, __LINE__, base_offset, base_align, G_038000_TILE_MODE(word0));
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
return -EINVAL;
}
if (!IS_ALIGNED(mip_offset, base_align)) {
drm/radeon/kms: improve 6xx/7xx CS error output Makes debugging CS rejections much easier. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: stable@kernel.org Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-13 23:42:41 +00:00
dev_warn(p->dev, "%s:%d tex mip offset (0x%llx, 0x%llx, %d) invalid\n",
__func__, __LINE__, mip_offset, base_align, G_038000_TILE_MODE(word0));
drm/radeon/kms: fix CS alignment checking for tiling (v2) Covers depth, cb, and textures. Hopefully I got this right. v2: - fix bugs: https://bugs.freedesktop.org/show_bug.cgi?id=28327 https://bugs.freedesktop.org/show_bug.cgi?id=28381 - use ALIGNED(), IS_ALIGNED() macros Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-06-04 22:41:42 +00:00
return -EINVAL;
}
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
if (blevel > llevel) {
dev_warn(p->dev, "texture blevel %d > llevel %d\n",
blevel, llevel);
}
drm/radeon/kms: reorder code in r600_check_texture_resource Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-09 14:34:15 +00:00
if (is_array) {
barray = G_038014_BASE_ARRAY(word5);
larray = G_038014_LAST_ARRAY(word5);
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
nfaces = larray - barray + 1;
}
drm/radeon: fix checking of MSAA renderbuffers on r600-r700 The MSAA checking was mostly unimplemented on r600-r700. The userspace submits GPU commands and the kernel driver computes how much memory the GPU will access and checks if it's all within buffer bounds the userspace allocated. This patch fixes the computations of the size of MSAA surfaces in memory. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2012-08-19 19:23:26 +00:00
r600_texture_size(nfaces, blevel, llevel, w0, h0, d0, array_check.nsamples, format,
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
pitch_align, height_align, base_align,
drm/radeon/kms: fix CS alignment checking for tiling (v2) Covers depth, cb, and textures. Hopefully I got this right. v2: - fix bugs: https://bugs.freedesktop.org/show_bug.cgi?id=28327 https://bugs.freedesktop.org/show_bug.cgi?id=28381 - use ALIGNED(), IS_ALIGNED() macros Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-06-04 22:41:42 +00:00
&l0_size, &mipmap_size);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
/* using get ib will give us the offset into the texture bo */
drm/radeon: make sure ib reads are in-order. have to read values from the IB in order as we could cross a page boundary at any time and won't be able to go backwards. Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-28 04:27:03 +00:00
if ((l0_size + word2) > radeon_bo_size(texture)) {
drm/radeon: add support for evergreen/ni tiling informations v11 evergreen and northern island gpu needs more informations for 2D tiling than previous r6xx/r7xx. Add field to tiling ioctl to allow userspace to provide those. The v8 cs checking change to track color view on r6xx/r7xx doesn't affect old userspace as old userspace always emited 0 for this register. v2 fix r6xx/r7xx 2D tiling computation v3 fix r6xx/r7xx height align for untiled surface & add support for tile split on evergreen and newer v4 improve tiling debugging output v5 fix tile split code for evergreen and newer v6 set proper tile split for crtc register v7 fix tile split limit value v8 add COLOR_VIEW checking to r6xx/r7xx checker, add evergreen cs checking, update safe reg for r600, evergreen and cayman. Evergreen checking need some work around for stencil alignment issues v9 fix tile split value range, fix compressed texture handling and mipmap calculation, allow evergreen check to be silencious in front of current broken userspace (depth/stencil alignment issue) v10 fix eg 3d texture and compressed texture, fix r600 depth array, fix r600 color view computation, add support for evergreen stencil split v11 more verbose debugging in some case Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-12-16 22:03:42 +00:00
dev_warn(p->dev, "texture bo too small ((%d %d) (%d %d) %d %d %d -> %d have %ld)\n",
w0, h0, pitch_align, height_align,
array_check.array_mode, format, word2,
l0_size, radeon_bo_size(texture));
drm/radeon: overhaul texture checking. (v3) the texture checking code didn't work for block formats like s3tc, this overhauls it to work for all types. v2: add texture array support. v3: add subsampled formats Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-18 05:51:58 +00:00
dev_warn(p->dev, "alignments %d %d %d %lld\n", pitch, pitch_align, height_align, base_align);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
return -EINVAL;
}
/* using get ib will give us the offset into the mipmap bo */
drm/radeon: make sure ib reads are in-order. have to read values from the IB in order as we could cross a page boundary at any time and won't be able to go backwards. Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-28 04:27:03 +00:00
if ((mipmap_size + word3) > radeon_bo_size(mipmap)) {
drm/radeon/kms: only warn on mipmap size checks in r600 cs checker (v2) The texture base address registers are in units of 256 bytes. The original CS checker treated these offsets as bytes, so the original check was wrong. I fixed the units in a patch during the 2.6.36 cycle, but this ended up breaking some existing userspace (probably due to a bug in either userspace texture allocation or the drm texture mipmap checker). So for now, until we come up with a better fix, just warn if the mipmap size it too large. This will keep existing userspace working and it should be just as safe as before when we were checking the wrong units. These are GPU MC addresses, so if they fall outside of the VRAM or GART apertures, they end up at the GPU default page, so this should be safe from a security perspective. v2: Just disable the warning. It just spams the log and there's nothing the user can do about it. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Jerome Glisse <glisse@freedesktop.org> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-09-14 14:10:47 +00:00
/*dev_warn(p->dev, "mipmap bo too small (%d %d %d %d %d %d -> %d have %ld)\n",
drm/radeon: make sure ib reads are in-order. have to read values from the IB in order as we could cross a page boundary at any time and won't be able to go backwards. Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-28 04:27:03 +00:00
w0, h0, format, blevel, nlevels, word3, mipmap_size, radeon_bo_size(texture));*/
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
return 0;
}
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
static bool r600_is_safe_reg(struct radeon_cs_parser *p, u32 reg, u32 idx)
{
u32 m, i;
i = (reg >> 7);
if (i >= ARRAY_SIZE(r600_reg_safe_bm)) {
dev_warn(p->dev, "forbidden register 0x%08x at %d\n", reg, idx);
return false;
}
m = 1 << ((reg >> 2) & 31);
if (!(r600_reg_safe_bm[i] & m))
return true;
dev_warn(p->dev, "forbidden register 0x%08x at %d\n", reg, idx);
return false;
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
static int r600_packet3_check(struct radeon_cs_parser *p,
struct radeon_cs_packet *pkt)
{
struct radeon_cs_reloc *reloc;
drm/radeon: r6xx/r7xx possible security issue, system ram access This patch workaround a possible security issue which can allow user to abuse drm on r6xx/r7xx hw to access any system ram memory. This patch doesn't break userspace, it detect "valid" old use of CB_COLOR[0-7]_FRAG & CB_COLOR[0-7]_TILE registers and overwritte the address these registers are pointing to with the one of the last color buffer. This workaround will work for old mesa & xf86-video-ati and any old user which did use similar register programming pattern as those (we expect that there is no others user of those ioctl except possibly a malicious one). This patch add a warning if it detects such usage, warning encourage people to update their mesa & xf86-video-ati. New userspace will submit proper relocation. Fix for xf86-video-ati / mesa (this kernel patch is enough to prevent abuse, fix for userspace are to set proper cs stream and avoid kernel warning) : http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=95d63e408cc88b6934bec84a0b1ef94dfe8bee7b http://cgit.freedesktop.org/mesa/mesa/commit/?id=46dc6fd3ed5ef96cda53641a97bc68c3bc104a9f Abusing this register to perform system ram memory is not easy, here is outline on how it could be achieve. First attacker must have access to the drm device and be able to submit command stream throught cs ioctl. Then attacker must build a proper command stream for r6xx/r7xx hw which will abuse the FRAG or TILE buffer to overwrite the GPU GART which is in VRAM. To achieve so attacker as to setup CB_COLOR[0-7]_FRAG or CB_COLOR[0-7]_TILE to point to the GPU GART, then it has to find a way to write predictable value into those buffer (with little cleverness i believe this can be done but this is an hard task). Once attacker have such program it can overwritte GPU GART to program GPU gart to point anywhere in system memory. It then can reusse same method as he used to reprogram GART to overwritte the system ram through the GART mapping. In the process the attacker has to be carefull to not overwritte any sensitive area of the GART table, like ring or IB gart entry as it will more then likely lead to GPU lockup. Bottom line is that i think it's very hard to use this flaw to get system ram access but in theory one can achieve so. Side note: I am not aware of anyone ever using the GPU as an attack vector, nevertheless we take great care in the opensource driver to try to detect and forbid malicious use of GPU. I don't think the closed source driver are as cautious as we are. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2010-01-18 12:01:36 +00:00
struct r600_cs_track *track;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
volatile u32 *ib;
unsigned idx;
unsigned i;
unsigned start_reg, end_reg, reg;
int r;
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
u32 idx_value;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
drm/radeon: r6xx/r7xx possible security issue, system ram access This patch workaround a possible security issue which can allow user to abuse drm on r6xx/r7xx hw to access any system ram memory. This patch doesn't break userspace, it detect "valid" old use of CB_COLOR[0-7]_FRAG & CB_COLOR[0-7]_TILE registers and overwritte the address these registers are pointing to with the one of the last color buffer. This workaround will work for old mesa & xf86-video-ati and any old user which did use similar register programming pattern as those (we expect that there is no others user of those ioctl except possibly a malicious one). This patch add a warning if it detects such usage, warning encourage people to update their mesa & xf86-video-ati. New userspace will submit proper relocation. Fix for xf86-video-ati / mesa (this kernel patch is enough to prevent abuse, fix for userspace are to set proper cs stream and avoid kernel warning) : http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=95d63e408cc88b6934bec84a0b1ef94dfe8bee7b http://cgit.freedesktop.org/mesa/mesa/commit/?id=46dc6fd3ed5ef96cda53641a97bc68c3bc104a9f Abusing this register to perform system ram memory is not easy, here is outline on how it could be achieve. First attacker must have access to the drm device and be able to submit command stream throught cs ioctl. Then attacker must build a proper command stream for r6xx/r7xx hw which will abuse the FRAG or TILE buffer to overwrite the GPU GART which is in VRAM. To achieve so attacker as to setup CB_COLOR[0-7]_FRAG or CB_COLOR[0-7]_TILE to point to the GPU GART, then it has to find a way to write predictable value into those buffer (with little cleverness i believe this can be done but this is an hard task). Once attacker have such program it can overwritte GPU GART to program GPU gart to point anywhere in system memory. It then can reusse same method as he used to reprogram GART to overwritte the system ram through the GART mapping. In the process the attacker has to be carefull to not overwritte any sensitive area of the GART table, like ring or IB gart entry as it will more then likely lead to GPU lockup. Bottom line is that i think it's very hard to use this flaw to get system ram access but in theory one can achieve so. Side note: I am not aware of anyone ever using the GPU as an attack vector, nevertheless we take great care in the opensource driver to try to detect and forbid malicious use of GPU. I don't think the closed source driver are as cautious as we are. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2010-01-18 12:01:36 +00:00
track = (struct r600_cs_track *)p->track;
drm/radeon: make the ib an inline object No need to malloc it any more. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Christian König <deathsimple@vodafone.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-05-09 13:35:02 +00:00
ib = p->ib.ptr;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
idx = pkt->idx + 1;
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
idx_value = radeon_get_ib_value(p, idx);
drm/radeon/kms: don't require up to 64k allocations. (v2) This avoids needing to do a kmalloc > PAGE_SIZE for the main indirect buffer chunk, it adds an accessor for all reads from the chunk and caches a single page at a time for subsequent reads. changes since v1: Use a two page pool which should be the most common case a single packet spanning > PAGE_SIZE will be hit, but I'm having trouble seeing anywhere we currently generate anything like that. hopefully proper short page copying at end added parser_error flag to set deep errors instead of having to test every ib value fetch. fixed bug in patch that went to list. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-23 06:56:27 +00:00
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
switch (pkt->opcode) {
drm/r600: parse the set predication command. (v2) This is required for NV_conditional_render and EXT_transform_feedback. v2: add evergreen support. Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-28 06:11:48 +00:00
case PACKET3_SET_PREDICATION:
{
int pred_op;
int tmp;
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
uint64_t offset;
drm/r600: parse the set predication command. (v2) This is required for NV_conditional_render and EXT_transform_feedback. v2: add evergreen support. Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-28 06:11:48 +00:00
if (pkt->count != 1) {
DRM_ERROR("bad SET PREDICATION\n");
return -EINVAL;
}
tmp = radeon_get_ib_value(p, idx + 1);
pred_op = (tmp >> 16) & 0x7;
/* for the clear predicate operation */
if (pred_op == 0)
return 0;
if (pred_op > 2) {
DRM_ERROR("bad SET PREDICATION operation %d\n", pred_op);
return -EINVAL;
}
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad SET PREDICATION\n");
return -EINVAL;
}
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
offset = reloc->lobj.gpu_offset +
(idx_value & 0xfffffff0) +
((u64)(tmp & 0xff) << 32);
ib[idx + 0] = offset;
ib[idx + 1] = (tmp & 0xffffff00) | (upper_32_bits(offset) & 0xff);
drm/r600: parse the set predication command. (v2) This is required for NV_conditional_render and EXT_transform_feedback. v2: add evergreen support. Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-02-28 06:11:48 +00:00
}
break;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
case PACKET3_START_3D_CMDBUF:
if (p->family >= CHIP_RV770 || pkt->count) {
DRM_ERROR("bad START_3D\n");
return -EINVAL;
}
break;
case PACKET3_CONTEXT_CONTROL:
if (pkt->count != 1) {
DRM_ERROR("bad CONTEXT_CONTROL\n");
return -EINVAL;
}
break;
case PACKET3_INDEX_TYPE:
case PACKET3_NUM_INSTANCES:
if (pkt->count) {
DRM_ERROR("bad INDEX_TYPE/NUM_INSTANCES\n");
return -EINVAL;
}
break;
case PACKET3_DRAW_INDEX:
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
{
uint64_t offset;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
if (pkt->count != 3) {
DRM_ERROR("bad DRAW_INDEX\n");
return -EINVAL;
}
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad DRAW_INDEX\n");
return -EINVAL;
}
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
offset = reloc->lobj.gpu_offset +
idx_value +
((u64)(radeon_get_ib_value(p, idx+1) & 0xff) << 32);
ib[idx+0] = offset;
ib[idx+1] = upper_32_bits(offset) & 0xff;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
r = r600_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
return r;
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
break;
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
case PACKET3_DRAW_INDEX_AUTO:
if (pkt->count != 1) {
DRM_ERROR("bad DRAW_INDEX_AUTO\n");
return -EINVAL;
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
r = r600_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream %d\n", __func__, __LINE__, idx);
return r;
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
break;
case PACKET3_DRAW_INDEX_IMMD_BE:
case PACKET3_DRAW_INDEX_IMMD:
if (pkt->count < 2) {
DRM_ERROR("bad DRAW_INDEX_IMMD\n");
return -EINVAL;
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
r = r600_cs_track_check(p);
if (r) {
dev_warn(p->dev, "%s:%d invalid cmd stream\n", __func__, __LINE__);
return r;
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
break;
case PACKET3_WAIT_REG_MEM:
if (pkt->count != 5) {
DRM_ERROR("bad WAIT_REG_MEM\n");
return -EINVAL;
}
/* bit 4 is reg (0) or mem (1) */
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
if (idx_value & 0x10) {
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
uint64_t offset;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad WAIT_REG_MEM\n");
return -EINVAL;
}
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
offset = reloc->lobj.gpu_offset +
(radeon_get_ib_value(p, idx+1) & 0xfffffff0) +
((u64)(radeon_get_ib_value(p, idx+2) & 0xff) << 32);
ib[idx+1] = (ib[idx+1] & 0x3) | (offset & 0xfffffff0);
ib[idx+2] = upper_32_bits(offset) & 0xff;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
}
break;
case PACKET3_SURFACE_SYNC:
if (pkt->count != 3) {
DRM_ERROR("bad SURFACE_SYNC\n");
return -EINVAL;
}
/* 0xffffffff/0x0 is flush all cache flag */
drm/radeon/kms: don't require up to 64k allocations. (v2) This avoids needing to do a kmalloc > PAGE_SIZE for the main indirect buffer chunk, it adds an accessor for all reads from the chunk and caches a single page at a time for subsequent reads. changes since v1: Use a two page pool which should be the most common case a single packet spanning > PAGE_SIZE will be hit, but I'm having trouble seeing anywhere we currently generate anything like that. hopefully proper short page copying at end added parser_error flag to set deep errors instead of having to test every ib value fetch. fixed bug in patch that went to list. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-23 06:56:27 +00:00
if (radeon_get_ib_value(p, idx + 1) != 0xffffffff ||
radeon_get_ib_value(p, idx + 2) != 0) {
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad SURFACE_SYNC\n");
return -EINVAL;
}
ib[idx+2] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
}
break;
case PACKET3_EVENT_WRITE:
if (pkt->count != 2 && pkt->count != 0) {
DRM_ERROR("bad EVENT_WRITE\n");
return -EINVAL;
}
if (pkt->count) {
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
uint64_t offset;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad EVENT_WRITE\n");
return -EINVAL;
}
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
offset = reloc->lobj.gpu_offset +
(radeon_get_ib_value(p, idx+1) & 0xfffffff8) +
((u64)(radeon_get_ib_value(p, idx+2) & 0xff) << 32);
ib[idx+1] = offset & 0xfffffff8;
ib[idx+2] = upper_32_bits(offset) & 0xff;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
}
break;
case PACKET3_EVENT_WRITE_EOP:
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
{
uint64_t offset;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
if (pkt->count != 4) {
DRM_ERROR("bad EVENT_WRITE_EOP\n");
return -EINVAL;
}
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad EVENT_WRITE\n");
return -EINVAL;
}
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
offset = reloc->lobj.gpu_offset +
(radeon_get_ib_value(p, idx+1) & 0xfffffffc) +
((u64)(radeon_get_ib_value(p, idx+2) & 0xff) << 32);
ib[idx+1] = offset & 0xfffffffc;
ib[idx+2] = (ib[idx+2] & 0xffffff00) | (upper_32_bits(offset) & 0xff);
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
break;
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
case PACKET3_SET_CONFIG_REG:
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
start_reg = (idx_value << 2) + PACKET3_SET_CONFIG_REG_OFFSET;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_CONFIG_REG_OFFSET) ||
(start_reg >= PACKET3_SET_CONFIG_REG_END) ||
(end_reg >= PACKET3_SET_CONFIG_REG_END)) {
DRM_ERROR("bad PACKET3_SET_CONFIG_REG\n");
return -EINVAL;
}
for (i = 0; i < pkt->count; i++) {
reg = start_reg + (4 * i);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
r = r600_cs_check_reg(p, reg, idx+1+i);
if (r)
return r;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
}
break;
case PACKET3_SET_CONTEXT_REG:
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
start_reg = (idx_value << 2) + PACKET3_SET_CONTEXT_REG_OFFSET;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_CONTEXT_REG_OFFSET) ||
(start_reg >= PACKET3_SET_CONTEXT_REG_END) ||
(end_reg >= PACKET3_SET_CONTEXT_REG_END)) {
DRM_ERROR("bad PACKET3_SET_CONTEXT_REG\n");
return -EINVAL;
}
for (i = 0; i < pkt->count; i++) {
reg = start_reg + (4 * i);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
r = r600_cs_check_reg(p, reg, idx+1+i);
if (r)
return r;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
}
break;
case PACKET3_SET_RESOURCE:
if (pkt->count % 7) {
DRM_ERROR("bad SET_RESOURCE\n");
return -EINVAL;
}
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
start_reg = (idx_value << 2) + PACKET3_SET_RESOURCE_OFFSET;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_RESOURCE_OFFSET) ||
(start_reg >= PACKET3_SET_RESOURCE_END) ||
(end_reg >= PACKET3_SET_RESOURCE_END)) {
DRM_ERROR("bad SET_RESOURCE\n");
return -EINVAL;
}
for (i = 0; i < (pkt->count / 7); i++) {
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
struct radeon_bo *texture, *mipmap;
drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Andre Maasikas <amaasikas@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-06 06:54:05 +00:00
u32 size, offset, base_offset, mip_offset;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
switch (G__SQ_VTX_CONSTANT_TYPE(radeon_get_ib_value(p, idx+(i*7)+6+1))) {
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
case SQ_TEX_VTX_VALID_TEXTURE:
/* tex base */
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad SET_RESOURCE\n");
return -EINVAL;
}
drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Andre Maasikas <amaasikas@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-06 06:54:05 +00:00
base_offset = (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
drm/radeon: GPU virtual memory support v22 Virtual address space are per drm client (opener of /dev/drm). Client are in charge of virtual address space, they need to map bo into it by calling DRM_RADEON_GEM_VA ioctl. First 16M of virtual address space is reserved by the kernel. Once using 2 level page table we should be able to have a small vram memory footprint for each pt (there would be one pt for all gart, one for all vram and then one first level for each virtual address space). Plan include using the sub allocator for a common vm page table area and using memcpy to copy vm page table in & out. Or use a gart object and copy things in & out using dma. v2: agd5f fixes: - Add vram base offset for vram pages. The GPU physical address of a vram page is FB_OFFSET + page offset. FB_OFFSET is 0 on discrete cards and the physical bus address of the stolen memory on integrated chips. - VM_CONTEXT1_PROTECTION_FAULT_DEFAULT_ADDR covers all vmid's >= 1 v3: agd5f: - integrate with the semaphore/multi-ring stuff v4: - rebase on top ttm dma & multi-ring stuff - userspace is now in charge of the address space - no more specific cs vm ioctl, instead cs ioctl has a new chunk v5: - properly handle mem == NULL case from move_notify callback - fix the vm cleanup path v6: - fix update of page table to only happen on valid mem placement v7: - add tlb flush for each vm context - add flags to define mapping property (readable, writeable, snooped) - make ring id implicit from ib->fence->ring, up to each asic callback to then do ring specific scheduling if vm ib scheduling function v8: - add query for ib limit and kernel reserved virtual space - rename vm->size to max_pfn (maximum number of page) - update gem_va ioctl to also allow unmap operation - bump kernel version to allow userspace to query for vm support v9: - rebuild page table only when bind and incrementaly depending on bo referenced by cs and that have been moved - allow virtual address space to grow - use sa allocator for vram page table - return invalid when querying vm limit on non cayman GPU - dump vm fault register on lockup v10: agd5f: - Move the vm schedule_ib callback to a standalone function, remove the callback and use the existing ib_execute callback for VM IBs. v11: - rebase on top of lastest Linus v12: agd5f: - remove spurious backslash - set IB vm_id to 0 in radeon_ib_get() v13: agd5f: - fix handling of RADEON_CHUNK_ID_FLAGS v14: - fix va destruction - fix suspend resume - forbid bo to have several different va in same vm v15: - rebase v16: - cleanup left over of vm init/fini v17: agd5f: - cs checker v18: agd5f: - reworks the CS ioctl to better support multiple rings and VM. Rather than adding a new chunk id for VM, just re-use the IB chunk id and add a new flags for VM mode. Also define additional dwords for the flags chunk id to define the what ring we want to use (gfx, compute, uvd, etc.) and the priority. v19: - fix cs fini in weird case of no ib - semi working flush fix for ni - rebase on top of sa allocator changes v20: agd5f: - further CS ioctl cleanups from Christian's comments v21: agd5f: - integrate CS checker improvements v22: agd5f: - final cleanups for release, only allow VM CS on cayman Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-06 03:11:05 +00:00
if (!(p->cs_flags & RADEON_CS_KEEP_TILING_FLAGS)) {
drm/radeon/kms: add a CS ioctl flag not to rewrite tiling flags in the CS This adds a new optional chunk to the CS ioctl that specifies optional flags to the CS parser. Why this is useful is explained below. Note that some regs no longer need the NOP relocation packet if this feature is enabled. Tested on r300g and r600g with this flag disabled and enabled. Assume there are two contexts sharing the same mipmapped tiled texture. One context wants to render into the first mipmap and the other one wants to render into the last mipmap. As you probably know, the hardware has a MACRO_SWITCH feature, which turns off macro tiling for small mipmaps, but that only applies to samplers. (at least on r300-r500, though later hardware likely behaves the same) So we want to just re-set the tiling flags before rendering (writing packets), right? ... No. The contexts run in parallel, so they may set the tiling flags simultaneously and then fire their command streams also simultaneously. The last one setting the flags wins, the other one loses. Another problem is when one context wants to render into the first and the last mipmap in one CS. Impossible. It must flush before changing tiling flags and do the rendering into the smaller mipmaps in another CS. Yet another problem is that writing copy_blit in userspace would be a mess involving re-setting tiling flags to please the kernel, and causing races with other contexts at the same time. The only way out of this is to send tiling flags with each CS, ideally with each relocation. But we already do that through the registers. So let's just use what we have in the registers. Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2011-10-24 23:38:45 +00:00
if (reloc->lobj.tiling_flags & RADEON_TILING_MACRO)
ib[idx+1+(i*7)+0] |= S_038000_TILE_MODE(V_038000_ARRAY_2D_TILED_THIN1);
else if (reloc->lobj.tiling_flags & RADEON_TILING_MICRO)
ib[idx+1+(i*7)+0] |= S_038000_TILE_MODE(V_038000_ARRAY_1D_TILED_THIN1);
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
texture = reloc->robj;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
/* tex mip base */
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad SET_RESOURCE\n");
return -EINVAL;
}
drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Andre Maasikas <amaasikas@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-06 06:54:05 +00:00
mip_offset = (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
mipmap = reloc->robj;
r = r600_check_texture_resource(p, idx+(i*7)+1,
drm/radeon/kms: fix and unify tiled buffer alignment checking for r6xx/7xx Tiled buffers have the same alignment requirements regardless of whether the surface is for db, cb, or textures. Previously, the calculations where inconsistent for each buffer type. - Unify the alignment calculations in a common function - Standardize the alignment units (pixels for pitch/height/depth, bytes for base) - properly check the buffer base alignments Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-11-15 01:24:35 +00:00
texture, mipmap,
base_offset + radeon_get_ib_value(p, idx+1+(i*7)+2),
mip_offset + radeon_get_ib_value(p, idx+1+(i*7)+3),
reloc->lobj.tiling_flags);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
if (r)
return r;
drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Andre Maasikas <amaasikas@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-06 06:54:05 +00:00
ib[idx+1+(i*7)+2] += base_offset;
ib[idx+1+(i*7)+3] += mip_offset;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
break;
case SQ_TEX_VTX_VALID_BUFFER:
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
{
uint64_t offset64;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
/* vtx base */
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad SET_RESOURCE\n");
return -EINVAL;
}
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
offset = radeon_get_ib_value(p, idx+1+(i*7)+0);
drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Andre Maasikas <amaasikas@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-06 06:54:05 +00:00
size = radeon_get_ib_value(p, idx+1+(i*7)+1) + 1;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
if (p->rdev && (size + offset) > radeon_bo_size(reloc->robj)) {
/* force size to size of the buffer */
drm/radeon/kms: r600 CS parser fixes - buffer offsets in the base regs are 256b aligned so shift properly when comparing, fixed by Andre Maasikas - mipmap size was calculated wrong when nlevel=0 - texture bo offsets were used after the bo base address was added - vertex resource size register is size - 1, not size Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Cc: Andre Maasikas <amaasikas@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-08-06 06:54:05 +00:00
dev_warn(p->dev, "vbo resource seems too big (%d) for the bo (%ld)\n",
size + offset, radeon_bo_size(reloc->robj));
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
ib[idx+1+(i*7)+1] = radeon_bo_size(reloc->robj) - offset;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
}
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
offset64 = reloc->lobj.gpu_offset + offset;
ib[idx+1+(i*8)+0] = offset64;
ib[idx+1+(i*8)+2] = (ib[idx+1+(i*8)+2] & 0xffffff00) |
(upper_32_bits(offset64) & 0xff);
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
break;
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
case SQ_TEX_VTX_INVALID_TEXTURE:
case SQ_TEX_VTX_INVALID_BUFFER:
default:
DRM_ERROR("bad SET_RESOURCE\n");
return -EINVAL;
}
}
break;
case PACKET3_SET_ALU_CONST:
drm/radeon/r6xx/r7xx: CS parser fixes - Drop some more safe regs taht userspace shouldn't hit - Constant base regs need relocs. This allows us to use constant buffers rather than the constant register file. Also we don't want userspace to be able to set arbitrary mc base values for the const caches. - Track SQ_CONFIG so we know whether userspace is using the cfile or constant buffers. Signed-off-by: Alex Deucher <alexdeucher@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-03-26 18:52:32 +00:00
if (track->sq_config & DX9_CONSTS) {
start_reg = (idx_value << 2) + PACKET3_SET_ALU_CONST_OFFSET;
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_ALU_CONST_OFFSET) ||
(start_reg >= PACKET3_SET_ALU_CONST_END) ||
(end_reg >= PACKET3_SET_ALU_CONST_END)) {
DRM_ERROR("bad SET_ALU_CONST\n");
return -EINVAL;
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
}
break;
case PACKET3_SET_BOOL_CONST:
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
start_reg = (idx_value << 2) + PACKET3_SET_BOOL_CONST_OFFSET;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_BOOL_CONST_OFFSET) ||
(start_reg >= PACKET3_SET_BOOL_CONST_END) ||
(end_reg >= PACKET3_SET_BOOL_CONST_END)) {
DRM_ERROR("bad SET_BOOL_CONST\n");
return -EINVAL;
}
break;
case PACKET3_SET_LOOP_CONST:
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
start_reg = (idx_value << 2) + PACKET3_SET_LOOP_CONST_OFFSET;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_LOOP_CONST_OFFSET) ||
(start_reg >= PACKET3_SET_LOOP_CONST_END) ||
(end_reg >= PACKET3_SET_LOOP_CONST_END)) {
DRM_ERROR("bad SET_LOOP_CONST\n");
return -EINVAL;
}
break;
case PACKET3_SET_CTL_CONST:
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
start_reg = (idx_value << 2) + PACKET3_SET_CTL_CONST_OFFSET;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_CTL_CONST_OFFSET) ||
(start_reg >= PACKET3_SET_CTL_CONST_END) ||
(end_reg >= PACKET3_SET_CTL_CONST_END)) {
DRM_ERROR("bad SET_CTL_CONST\n");
return -EINVAL;
}
break;
case PACKET3_SET_SAMPLER:
if (pkt->count % 3) {
DRM_ERROR("bad SET_SAMPLER\n");
return -EINVAL;
}
drm/r600: get values from the passed in IB not the copy. this avoids reading back the IB on AGP, also it avoids the race where since we haven't fetched the page from the main IB and written it to the gpu one, reading back fetches 0. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-25 04:23:47 +00:00
start_reg = (idx_value << 2) + PACKET3_SET_SAMPLER_OFFSET;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
end_reg = 4 * pkt->count + start_reg - 4;
if ((start_reg < PACKET3_SET_SAMPLER_OFFSET) ||
(start_reg >= PACKET3_SET_SAMPLER_END) ||
(end_reg >= PACKET3_SET_SAMPLER_END)) {
DRM_ERROR("bad SET_SAMPLER\n");
return -EINVAL;
}
break;
drm/radeon: add support for STRMOUT_BASE_UPDATE on 7xx Required for streamout. Bump drm minor. Marek v2: fix pkt->count check Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Reviewed-by: Michel Dänzer <michel.daenzer@amd.com> Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-06-14 20:06:37 +00:00
case PACKET3_STRMOUT_BASE_UPDATE:
if (p->family < CHIP_RV770) {
DRM_ERROR("STRMOUT_BASE_UPDATE only supported on 7xx\n");
return -EINVAL;
}
if (pkt->count != 1) {
DRM_ERROR("bad STRMOUT_BASE_UPDATE packet count\n");
return -EINVAL;
}
if (idx_value > 3) {
DRM_ERROR("bad STRMOUT_BASE_UPDATE index\n");
return -EINVAL;
}
{
u64 offset;
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad STRMOUT_BASE_UPDATE reloc\n");
return -EINVAL;
}
if (reloc->robj != track->vgt_strmout_bo[idx_value]) {
DRM_ERROR("bad STRMOUT_BASE_UPDATE, bo does not match\n");
return -EINVAL;
}
offset = radeon_get_ib_value(p, idx+1) << 8;
if (offset != track->vgt_strmout_bo_offset[idx_value]) {
DRM_ERROR("bad STRMOUT_BASE_UPDATE, bo offset does not match: 0x%llx, 0x%x\n",
offset, track->vgt_strmout_bo_offset[idx_value]);
return -EINVAL;
}
if ((offset + 4) > radeon_bo_size(reloc->robj)) {
DRM_ERROR("bad STRMOUT_BASE_UPDATE bo too small: 0x%llx, 0x%lx\n",
offset + 4, radeon_bo_size(reloc->robj));
return -EINVAL;
}
ib[idx+1] += (u32)((reloc->lobj.gpu_offset >> 8) & 0xffffffff);
}
break;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
case PACKET3_SURFACE_BASE_UPDATE:
if (p->family >= CHIP_RV770 || p->family == CHIP_R600) {
DRM_ERROR("bad SURFACE_BASE_UPDATE\n");
return -EINVAL;
}
if (pkt->count) {
DRM_ERROR("bad SURFACE_BASE_UPDATE\n");
return -EINVAL;
}
break;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
case PACKET3_STRMOUT_BUFFER_UPDATE:
if (pkt->count != 4) {
DRM_ERROR("bad STRMOUT_BUFFER_UPDATE (invalid count)\n");
return -EINVAL;
}
/* Updating memory at DST_ADDRESS. */
if (idx_value & 0x1) {
u64 offset;
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad STRMOUT_BUFFER_UPDATE (missing dst reloc)\n");
return -EINVAL;
}
offset = radeon_get_ib_value(p, idx+1);
offset += ((u64)(radeon_get_ib_value(p, idx+2) & 0xff)) << 32;
if ((offset + 4) > radeon_bo_size(reloc->robj)) {
DRM_ERROR("bad STRMOUT_BUFFER_UPDATE dst bo too small: 0x%llx, 0x%lx\n",
offset + 4, radeon_bo_size(reloc->robj));
return -EINVAL;
}
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
offset += reloc->lobj.gpu_offset;
ib[idx+1] = offset;
ib[idx+2] = upper_32_bits(offset) & 0xff;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
}
/* Reading data from SRC_ADDRESS. */
if (((idx_value >> 1) & 0x3) == 2) {
u64 offset;
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad STRMOUT_BUFFER_UPDATE (missing src reloc)\n");
return -EINVAL;
}
offset = radeon_get_ib_value(p, idx+3);
offset += ((u64)(radeon_get_ib_value(p, idx+4) & 0xff)) << 32;
if ((offset + 4) > radeon_bo_size(reloc->robj)) {
DRM_ERROR("bad STRMOUT_BUFFER_UPDATE src bo too small: 0x%llx, 0x%lx\n",
offset + 4, radeon_bo_size(reloc->robj));
return -EINVAL;
}
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
offset += reloc->lobj.gpu_offset;
ib[idx+3] = offset;
ib[idx+4] = upper_32_bits(offset) & 0xff;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
}
break;
case PACKET3_COPY_DW:
if (pkt->count != 4) {
DRM_ERROR("bad COPY_DW (invalid count)\n");
return -EINVAL;
}
if (idx_value & 0x1) {
u64 offset;
/* SRC is memory. */
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad COPY_DW (missing src reloc)\n");
return -EINVAL;
}
offset = radeon_get_ib_value(p, idx+1);
offset += ((u64)(radeon_get_ib_value(p, idx+2) & 0xff)) << 32;
if ((offset + 4) > radeon_bo_size(reloc->robj)) {
DRM_ERROR("bad COPY_DW src bo too small: 0x%llx, 0x%lx\n",
offset + 4, radeon_bo_size(reloc->robj));
return -EINVAL;
}
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
offset += reloc->lobj.gpu_offset;
ib[idx+1] = offset;
ib[idx+2] = upper_32_bits(offset) & 0xff;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
} else {
/* SRC is a reg. */
reg = radeon_get_ib_value(p, idx+1) << 2;
if (!r600_is_safe_reg(p, reg, idx+1))
return -EINVAL;
}
if (idx_value & 0x2) {
u64 offset;
/* DST is memory. */
r = r600_cs_packet_next_reloc(p, &reloc);
if (r) {
DRM_ERROR("bad COPY_DW (missing dst reloc)\n");
return -EINVAL;
}
offset = radeon_get_ib_value(p, idx+3);
offset += ((u64)(radeon_get_ib_value(p, idx+4) & 0xff)) << 32;
if ((offset + 4) > radeon_bo_size(reloc->robj)) {
DRM_ERROR("bad COPY_DW dst bo too small: 0x%llx, 0x%lx\n",
offset + 4, radeon_bo_size(reloc->robj));
return -EINVAL;
}
drm/radeon/kms: compute GPU addresses correctly on r600 Signed-off-by: Marek Olšák <maraeo@gmail.com> Reviewed-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-03-19 02:09:37 +00:00
offset += reloc->lobj.gpu_offset;
ib[idx+3] = offset;
ib[idx+4] = upper_32_bits(offset) & 0xff;
drm/radeon/kms: add support for streamout v7 v2: agd5f: add strmout CS checking, copy_dw register checking v3: agd5f: don't use cs_check_reg() for copy_dw checking as it will incorrectly patch the command stream for certain regs. v4: agd5f: add warning if safe reg check fails for copy_dw v5: agd5f: add stricter checking for 6xx/7xx v6: agd5f: add range checking for copy_dw on eg+, add sx_surface_sync to safe reg list for 7xx. v7: agd5f: add stricter checking for eg+ Signed-off-by: Marek Olšák <maraeo@gmail.com> Signed-off-by: Alex Deucher <alexander.deucher@amd.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-01-27 17:17:59 +00:00
} else {
/* DST is a reg. */
reg = radeon_get_ib_value(p, idx+3) << 2;
if (!r600_is_safe_reg(p, reg, idx+3))
return -EINVAL;
}
break;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
case PACKET3_NOP:
break;
default:
DRM_ERROR("Packet3 opcode %x not supported\n", pkt->opcode);
return -EINVAL;
}
return 0;
}
int r600_cs_parse(struct radeon_cs_parser *p)
{
struct radeon_cs_packet pkt;
drm/radeon: r6xx/r7xx possible security issue, system ram access This patch workaround a possible security issue which can allow user to abuse drm on r6xx/r7xx hw to access any system ram memory. This patch doesn't break userspace, it detect "valid" old use of CB_COLOR[0-7]_FRAG & CB_COLOR[0-7]_TILE registers and overwritte the address these registers are pointing to with the one of the last color buffer. This workaround will work for old mesa & xf86-video-ati and any old user which did use similar register programming pattern as those (we expect that there is no others user of those ioctl except possibly a malicious one). This patch add a warning if it detects such usage, warning encourage people to update their mesa & xf86-video-ati. New userspace will submit proper relocation. Fix for xf86-video-ati / mesa (this kernel patch is enough to prevent abuse, fix for userspace are to set proper cs stream and avoid kernel warning) : http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=95d63e408cc88b6934bec84a0b1ef94dfe8bee7b http://cgit.freedesktop.org/mesa/mesa/commit/?id=46dc6fd3ed5ef96cda53641a97bc68c3bc104a9f Abusing this register to perform system ram memory is not easy, here is outline on how it could be achieve. First attacker must have access to the drm device and be able to submit command stream throught cs ioctl. Then attacker must build a proper command stream for r6xx/r7xx hw which will abuse the FRAG or TILE buffer to overwrite the GPU GART which is in VRAM. To achieve so attacker as to setup CB_COLOR[0-7]_FRAG or CB_COLOR[0-7]_TILE to point to the GPU GART, then it has to find a way to write predictable value into those buffer (with little cleverness i believe this can be done but this is an hard task). Once attacker have such program it can overwritte GPU GART to program GPU gart to point anywhere in system memory. It then can reusse same method as he used to reprogram GART to overwritte the system ram through the GART mapping. In the process the attacker has to be carefull to not overwritte any sensitive area of the GART table, like ring or IB gart entry as it will more then likely lead to GPU lockup. Bottom line is that i think it's very hard to use this flaw to get system ram access but in theory one can achieve so. Side note: I am not aware of anyone ever using the GPU as an attack vector, nevertheless we take great care in the opensource driver to try to detect and forbid malicious use of GPU. I don't think the closed source driver are as cautious as we are. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2010-01-18 12:01:36 +00:00
struct r600_cs_track *track;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
int r;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
if (p->track == NULL) {
/* initialize tracker, we are in kms */
track = kzalloc(sizeof(*track), GFP_KERNEL);
if (track == NULL)
return -ENOMEM;
r600_cs_track_init(track);
if (p->rdev->family < CHIP_RV770) {
track->npipes = p->rdev->config.r600.tiling_npipes;
track->nbanks = p->rdev->config.r600.tiling_nbanks;
track->group_size = p->rdev->config.r600.tiling_group_size;
} else if (p->rdev->family <= CHIP_RV740) {
track->npipes = p->rdev->config.rv770.tiling_npipes;
track->nbanks = p->rdev->config.rv770.tiling_nbanks;
track->group_size = p->rdev->config.rv770.tiling_group_size;
}
p->track = track;
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
do {
r = r600_cs_packet_parse(p, &pkt, p->idx);
if (r) {
drm/radeon/kms: fix r600/r700 cs checker to avoid double kfree radeon_cs kfree the tracker structure but for r6xx/r7xx we want to kfree it inside the parse function because we share it with the UMS code path. Set tracker to NULL after freeing it will avoid double free. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-11 11:44:32 +00:00
kfree(p->track);
p->track = NULL;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
return r;
}
p->idx += pkt.count + 2;
switch (pkt.type) {
case PACKET_TYPE0:
r = r600_cs_parse_packet0(p, &pkt);
break;
case PACKET_TYPE2:
break;
case PACKET_TYPE3:
r = r600_packet3_check(p, &pkt);
break;
default:
DRM_ERROR("Unknown packet type %d !\n", pkt.type);
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
kfree(p->track);
drm/radeon/kms: fix r600/r700 cs checker to avoid double kfree radeon_cs kfree the tracker structure but for r6xx/r7xx we want to kfree it inside the parse function because we share it with the UMS code path. Set tracker to NULL after freeing it will avoid double free. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-11 11:44:32 +00:00
p->track = NULL;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
return -EINVAL;
}
if (r) {
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
kfree(p->track);
drm/radeon/kms: fix r600/r700 cs checker to avoid double kfree radeon_cs kfree the tracker structure but for r6xx/r7xx we want to kfree it inside the parse function because we share it with the UMS code path. Set tracker to NULL after freeing it will avoid double free. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-11 11:44:32 +00:00
p->track = NULL;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
return r;
}
} while (p->idx < p->chunks[p->chunk_ib_idx].length_dw);
#if 0
drm/radeon: make the ib an inline object No need to malloc it any more. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Christian König <deathsimple@vodafone.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-05-09 13:35:02 +00:00
for (r = 0; r < p->ib.length_dw; r++) {
printk(KERN_INFO "%05d 0x%08X\n", r, p->ib.ptr[r]);
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
mdelay(1);
}
#endif
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
kfree(p->track);
drm/radeon/kms: fix r600/r700 cs checker to avoid double kfree radeon_cs kfree the tracker structure but for r6xx/r7xx we want to kfree it inside the parse function because we share it with the UMS code path. Set tracker to NULL after freeing it will avoid double free. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-11 11:44:32 +00:00
p->track = NULL;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
return 0;
}
static int r600_cs_parser_relocs_legacy(struct radeon_cs_parser *p)
{
if (p->chunk_relocs_idx == -1) {
return 0;
}
drivers/gpu: Use kzalloc for allocating only one thing Use kzalloc rather than kcalloc(1,...) The use of the allocated memory that looks like an array is &p->relocs[0], but this should be the same as p->relocs. The semantic patch that makes this change is as follows: (http://coccinelle.lip6.fr/) // <smpl> @@ @@ - kcalloc(1, + kzalloc( ...) // </smpl> Signed-off-by: Julia Lawall <julia@diku.dk> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-12-19 07:16:33 +00:00
p->relocs = kzalloc(sizeof(struct radeon_cs_reloc), GFP_KERNEL);
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
if (p->relocs == NULL) {
return -ENOMEM;
}
return 0;
}
/**
* cs_parser_fini() - clean parser states
* @parser: parser structure holding parsing context.
* @error: error number
*
* If error is set than unvalidate buffer, otherwise just free memory
* used by parsing context.
**/
static void r600_cs_parser_fini(struct radeon_cs_parser *parser, int error)
{
unsigned i;
kfree(parser->relocs);
for (i = 0; i < parser->nchunks; i++) {
kfree(parser->chunks[i].kdata);
drm/r600: fix memory leak introduced with 64k malloc avoidance fix. The legacy r600 path shares code, but doesn't share quite enough to get the freeing correct. Free the pages here also. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-28 05:37:25 +00:00
kfree(parser->chunks[i].kpage[0]);
kfree(parser->chunks[i].kpage[1]);
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
}
kfree(parser->chunks);
kfree(parser->chunks_array);
}
int r600_cs_legacy(struct drm_device *dev, void *data, struct drm_file *filp,
unsigned family, u32 *ib, int *l)
{
struct radeon_cs_parser parser;
struct radeon_cs_chunk *ib_chunk;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
struct r600_cs_track *track;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
int r;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
/* initialize tracker */
track = kzalloc(sizeof(*track), GFP_KERNEL);
if (track == NULL)
return -ENOMEM;
r600_cs_track_init(track);
r600_cs_legacy_get_tiling_conf(dev, &track->npipes, &track->nbanks, &track->group_size);
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
/* initialize parser */
memset(&parser, 0, sizeof(struct radeon_cs_parser));
parser.filp = filp;
drm/radeon: r6xx/r7xx possible security issue, system ram access This patch workaround a possible security issue which can allow user to abuse drm on r6xx/r7xx hw to access any system ram memory. This patch doesn't break userspace, it detect "valid" old use of CB_COLOR[0-7]_FRAG & CB_COLOR[0-7]_TILE registers and overwritte the address these registers are pointing to with the one of the last color buffer. This workaround will work for old mesa & xf86-video-ati and any old user which did use similar register programming pattern as those (we expect that there is no others user of those ioctl except possibly a malicious one). This patch add a warning if it detects such usage, warning encourage people to update their mesa & xf86-video-ati. New userspace will submit proper relocation. Fix for xf86-video-ati / mesa (this kernel patch is enough to prevent abuse, fix for userspace are to set proper cs stream and avoid kernel warning) : http://cgit.freedesktop.org/xorg/driver/xf86-video-ati/commit/?id=95d63e408cc88b6934bec84a0b1ef94dfe8bee7b http://cgit.freedesktop.org/mesa/mesa/commit/?id=46dc6fd3ed5ef96cda53641a97bc68c3bc104a9f Abusing this register to perform system ram memory is not easy, here is outline on how it could be achieve. First attacker must have access to the drm device and be able to submit command stream throught cs ioctl. Then attacker must build a proper command stream for r6xx/r7xx hw which will abuse the FRAG or TILE buffer to overwrite the GPU GART which is in VRAM. To achieve so attacker as to setup CB_COLOR[0-7]_FRAG or CB_COLOR[0-7]_TILE to point to the GPU GART, then it has to find a way to write predictable value into those buffer (with little cleverness i believe this can be done but this is an hard task). Once attacker have such program it can overwritte GPU GART to program GPU gart to point anywhere in system memory. It then can reusse same method as he used to reprogram GART to overwritte the system ram through the GART mapping. In the process the attacker has to be carefull to not overwritte any sensitive area of the GART table, like ring or IB gart entry as it will more then likely lead to GPU lockup. Bottom line is that i think it's very hard to use this flaw to get system ram access but in theory one can achieve so. Side note: I am not aware of anyone ever using the GPU as an attack vector, nevertheless we take great care in the opensource driver to try to detect and forbid malicious use of GPU. I don't think the closed source driver are as cautious as we are. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@linux.ie>
2010-01-18 12:01:36 +00:00
parser.dev = &dev->pdev->dev;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
parser.rdev = NULL;
parser.family = family;
drm/radeon/kms: r600/r700 command stream checker This patch add cs checker to r600/r700 hw. Command stream checking will rewrite some of the cs value in order to restrict GPU access to BO size. This doesn't break old userspace but just enforce safe value. It should break any things that was using the r600/r700 cs ioctl to do forbidden things (malicious software), though we are not aware of such things. Here is the list of thing we check : - enforcing resource size - enforcing color buffer slice tile max, will restrict cb access - enforcing db buffer slice tile max, will restrict db access We don't check for shader bigger than the BO in which they are supposed to be, such use would lead to GPU lockup and is harmless from security POV, as far as we can tell (note that even checking for this wouldn't prevent someone to write bogus shader that lead to lockup). This patch has received as much testing as humanly possible with old userspace to check that it didn't break such configuration. However not all the applications out there were tested, thus it might broke some odd, rare applications. [airlied: fix rules for cs checker for parallel builds] Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2010-02-10 22:30:05 +00:00
parser.track = track;
drm/radeon: make the ib an inline object No need to malloc it any more. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Christian König <deathsimple@vodafone.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-05-09 13:35:02 +00:00
parser.ib.ptr = ib;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
r = radeon_cs_parser_init(&parser, data);
if (r) {
DRM_ERROR("Failed to initialize parser !\n");
r600_cs_parser_fini(&parser, r);
return r;
}
r = r600_cs_parser_relocs_legacy(&parser);
if (r) {
DRM_ERROR("Failed to parse relocation !\n");
r600_cs_parser_fini(&parser, r);
return r;
}
/* Copy the packet into the IB, the parser will read from the
* input memory (cached) and write to the IB (which can be
* uncached). */
ib_chunk = &parser.chunks[parser.chunk_ib_idx];
drm/radeon: make the ib an inline object No need to malloc it any more. Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Christian König <deathsimple@vodafone.de> Signed-off-by: Dave Airlie <airlied@redhat.com>
2012-05-09 13:35:02 +00:00
parser.ib.length_dw = ib_chunk->length_dw;
*l = parser.ib.length_dw;
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
r = r600_cs_parse(&parser);
if (r) {
DRM_ERROR("Invalid command stream !\n");
r600_cs_parser_fini(&parser, r);
return r;
}
drm/radeon/kms: don't require up to 64k allocations. (v2) This avoids needing to do a kmalloc > PAGE_SIZE for the main indirect buffer chunk, it adds an accessor for all reads from the chunk and caches a single page at a time for subsequent reads. changes since v1: Use a two page pool which should be the most common case a single packet spanning > PAGE_SIZE will be hit, but I'm having trouble seeing anywhere we currently generate anything like that. hopefully proper short page copying at end added parser_error flag to set deep errors instead of having to test every ib value fetch. fixed bug in patch that went to list. Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-23 06:56:27 +00:00
r = radeon_cs_finish_pages(&parser);
if (r) {
DRM_ERROR("Invalid command stream !\n");
r600_cs_parser_fini(&parser, r);
return r;
}
drm/radeon/kms: add r600 KMS support This adds the r600 KMS + CS support to the Linux kernel. The r600 TTM support is quite basic and still needs more work esp around using interrupts, but the polled fencing should work okay for now. Also currently TTM is using memcpy to do VRAM moves, the code is here to use a 3D blit to do this, but isn't fully debugged yet. Authors: Alex Deucher <alexdeucher@gmail.com> Dave Airlie <airlied@redhat.com> Jerome Glisse <jglisse@redhat.com> Signed-off-by: Jerome Glisse <jglisse@redhat.com> Signed-off-by: Dave Airlie <airlied@redhat.com>
2009-09-08 00:10:24 +00:00
r600_cs_parser_fini(&parser, r);
return r;
}
void r600_cs_legacy_init(void)
{
r600_cs_packet_next_reloc = &r600_cs_packet_next_reloc_nomm;
}