Add support for mmap'ing buffers via dmabuf. For handling mmap of
cached buffers correctly, fault handling and PTE shootdown are used
to track dirty pages and automagically handle cache flushes before
dma access to the buffer.
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
For now just implementing the exporting APIs, not yet importing. And
kmap is rejected on tiled buffers (although the usefulness of that seems
questionable, but could be added later if needed).
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Deal with the case of buffers with virtual stride larger than one
page in fault_2d().
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
This patch splits the DMM off into a separate sub-device, in order to
utilize the platform device information that was created as part of
the OMAP hwmod entry for the DMM.
The driver probe function queries the iomem resource and IRQ using
standard platform_get functions.
The OMAP DRM driver now calls the platform_driver_register() function
for the subordinate DMM driver inside its probe function. This
guarantees the required order for the DMM and ensures the DMM resources
are available for use by the DRM driver.
Signed-off-by: Andy Gross <andy.gross@ti.com>
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Add some additional debugfs file to aid in tracking buffer usage.
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
If fbcon calls us from atomic context, push the work off to the
workqueue to avoid calling into the gem/dmm code in an atomic
context.
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
The create/free mmap offset code must be synchronized. Yet only some
callers of omap_gem_mmap_offset() held struct_mutex. Leading to various
crashes around drm_mm_insert_helper_range(). (In the free-object path,
which is currently the only place we drm_gem_free_mmap_offset(),
struct_mutex is already held.)
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Add support in framebuffer objects for other color formats and multi-
planar YUV (NV12). Since this requires changing the API between the
plane and fb for getting scanout information (paddr, etc), take
advantage of the opportunity and put in place a way to allow fb's to
be unpinned when they are not being scanned out. Now, before start
of scanout the plane calls omap_framebuffer_pin() which takes care
to pin all the backing bo's, then omap_framebuffer_update_scanout()
however many times to update the scanout address(es), etc, and then
when finished omap_framebuffer_unpin().
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Uncached/writecombine buffers need to be clean before creating uc/wc
userspace mappings, lest dirty cache lines fall down and corrupt the
contents of the buffer.
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
omap_gem_roll() could be called by fbcon in atomic context or when
struct_mutext is held. Avoid aquiring mutex (deadlock), or calling
tiler_pin() (which itself is not safe for atomic context) in these
cases.
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Add support for YWRAP scrolling by shuffling pages around in DMM
instead of sw blits.
Note that fbcon only utilizes this mode if the y resolution is
divided evenly by the font height. So, for example, a 1920x1080
display using a 16 pixel tall font will not utilize this, but a
1280x1024 display would.
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
TILER/DMM provides two features for omapdrm GEM objects:
1) providing a physically contiguous view to discontiguous memory
for hw initiators that cannot otherwise support discontiguous
buffers (DSS scanout, IVAHD video decode/encode, etc)
2) providing untiling for 2d tiled buffers, which are used in some
cases to provide rotation and reduce memory bandwidth for hw
initiators that tend to access data in 2d block patterns.
For 2d tiled buffers, there are some additional complications when
it comes to userspace mmap'ings. For non-tiled buffers, the original
(potentially physically discontiguous) pages are used to back the
mmap. For tiled buffers, we need to mmap via the tiler/dmm region to
provide an unswizzled view of the buffer. But (a) the buffer is not
necessarily pinned in TILER all the time (it can be unmapped when
there is no DMA access to the buffer), and (b) when they are they
are pinned, they not necessarily page aligned from the perspective of
the CPU. And non-page aligned userspace buffer mapping is evil.
To solve this, we reserve one or more small regions in each of the 2d
containers when the driver is loaded to use as a "user-GART" where we
can create a second page-aligned mapping of parts of the buffer being
accessed from userspace. Page faulting is used to evict and remap
different regions of whichever buffers are being accessed from user-
space.
Signed-off-by: Rob Clark <rob@ti.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
A DRM display driver for TI OMAP platform. Similar to omapfb (fbdev)
and omap_vout (v4l2 display) drivers in the past, this driver uses the
DSS2 driver to access the display hardware, including support for
HDMI, DVI, and various types of LCD panels. And it implements GEM
support for buffer allocation (for KMS as well as offscreen buffers
used by the xf86-video-omap userspace xorg driver).
The driver maps CRTCs to overlays, encoders to overlay-managers, and
connectors to dssdev's. Note that this arrangement might change slightly
when support for drm_plane overlays is added.
For GEM support, non-scanout buffers are using the shmem backed pages
provided by GEM core (In drm_gem_object_init()). In the case of scanout
buffers, which need to be physically contiguous, those are allocated
with CMA and use drm_gem_private_object_init().
See userspace xorg driver:
git://github.com/robclark/xf86-video-omap.git
Refer to this link for CMA (Continuous Memory Allocator):
http://lkml.org/lkml/2011/8/19/302
Links to previous versions of the patch:
v1: http://lwn.net/Articles/458137/
v2: http://patches.linaro.org/4156/
v3: http://patches.linaro.org/4688/
v4: http://patches.linaro.org/4791/
History:
v5: move headers from include/drm at Greg KH's request, minor rebasing
on 3.2-rc1, pull in private copies of drm_gem_{get,put}_pages()
because "drm/gem: add functions to get/put pages" patch is not
merged yet
v4: bit of rework of encoder/connector _dpms() code, modeset_init()
rework to not use nested functions, update TODO.txt
v3: minor cleanups, improved error handling for dev_load(), some minor
API changes that will be needed later for tiled buffer support
v2: replace omap_vram with CMA for scanout buffer allocation, remove
unneeded functions, use dma_addr_t for physical addresses, error
handling cleanup, refactor attach/detach pages into common drm
functions, split non-userspace-facing API into omap_priv.h, remove
plugin API
v1: original
Signed-off-by: Rob Clark <rob@ti.com>
Acked-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Al Viro