Archived
14
0
Fork 0
This repository has been archived on 2022-02-17.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/arch/i386/kernel/srat.c
keith mannthey 3b08606dc2 [PATCH] convert i386 Summit subarch to use SRAT info for apicid_to_node calls
Convert the i386 summit subarch apicid_to_node to use node information
provided by the SRAT.  It was discussed a little on LKML a few weeks ago
and was seen as an acceptable fix.  The current way of obtaining the nodeid

 static inline int apicid_to_node(int logical_apicid)
 {
   return logical_apicid >> 5;
 }

is just not correct for all summit systems/bios.  Assuming the apicid
matches the Linux node number require a leap of faith that the bios mapped
out the apicids a set way.  Modern summit HW (IBM x460) does not layout its
bios in the manner for various reasons and is unable to boot i386 numa.

The best way to get the correct apicid to node information is from the SRAT
table during boot.  It lays out what apicid belongs to what node.  I use
this information to create a table for use at run time.

Signed-off-by: Keith Mannthey <kmannth@us.ibm.com>
Cc: Andi Kleen <ak@suse.de>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-09-29 09:18:03 -07:00

363 lines
11 KiB
C

/*
* Some of the code in this file has been gleaned from the 64 bit
* discontigmem support code base.
*
* Copyright (C) 2002, IBM Corp.
*
* All rights reserved.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful, but
* WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE, GOOD TITLE or
* NON INFRINGEMENT. See the GNU General Public License for more
* details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*
* Send feedback to Pat Gaughen <gone@us.ibm.com>
*/
#include <linux/mm.h>
#include <linux/bootmem.h>
#include <linux/mmzone.h>
#include <linux/acpi.h>
#include <linux/nodemask.h>
#include <asm/srat.h>
#include <asm/topology.h>
#include <asm/smp.h>
/*
* proximity macros and definitions
*/
#define NODE_ARRAY_INDEX(x) ((x) / 8) /* 8 bits/char */
#define NODE_ARRAY_OFFSET(x) ((x) % 8) /* 8 bits/char */
#define BMAP_SET(bmap, bit) ((bmap)[NODE_ARRAY_INDEX(bit)] |= 1 << NODE_ARRAY_OFFSET(bit))
#define BMAP_TEST(bmap, bit) ((bmap)[NODE_ARRAY_INDEX(bit)] & (1 << NODE_ARRAY_OFFSET(bit)))
/* bitmap length; _PXM is at most 255 */
#define PXM_BITMAP_LEN (MAX_PXM_DOMAINS / 8)
static u8 pxm_bitmap[PXM_BITMAP_LEN]; /* bitmap of proximity domains */
#define MAX_CHUNKS_PER_NODE 3
#define MAXCHUNKS (MAX_CHUNKS_PER_NODE * MAX_NUMNODES)
struct node_memory_chunk_s {
unsigned long start_pfn;
unsigned long end_pfn;
u8 pxm; // proximity domain of node
u8 nid; // which cnode contains this chunk?
u8 bank; // which mem bank on this node
};
static struct node_memory_chunk_s node_memory_chunk[MAXCHUNKS];
static int num_memory_chunks; /* total number of memory chunks */
static u8 __initdata apicid_to_pxm[MAX_APICID];
extern void * boot_ioremap(unsigned long, unsigned long);
/* Identify CPU proximity domains */
static void __init parse_cpu_affinity_structure(char *p)
{
struct acpi_table_processor_affinity *cpu_affinity =
(struct acpi_table_processor_affinity *) p;
if (!cpu_affinity->flags.enabled)
return; /* empty entry */
/* mark this node as "seen" in node bitmap */
BMAP_SET(pxm_bitmap, cpu_affinity->proximity_domain);
apicid_to_pxm[cpu_affinity->apic_id] = cpu_affinity->proximity_domain;
printk("CPU 0x%02X in proximity domain 0x%02X\n",
cpu_affinity->apic_id, cpu_affinity->proximity_domain);
}
/*
* Identify memory proximity domains and hot-remove capabilities.
* Fill node memory chunk list structure.
*/
static void __init parse_memory_affinity_structure (char *sratp)
{
unsigned long long paddr, size;
unsigned long start_pfn, end_pfn;
u8 pxm;
struct node_memory_chunk_s *p, *q, *pend;
struct acpi_table_memory_affinity *memory_affinity =
(struct acpi_table_memory_affinity *) sratp;
if (!memory_affinity->flags.enabled)
return; /* empty entry */
/* mark this node as "seen" in node bitmap */
BMAP_SET(pxm_bitmap, memory_affinity->proximity_domain);
/* calculate info for memory chunk structure */
paddr = memory_affinity->base_addr_hi;
paddr = (paddr << 32) | memory_affinity->base_addr_lo;
size = memory_affinity->length_hi;
size = (size << 32) | memory_affinity->length_lo;
start_pfn = paddr >> PAGE_SHIFT;
end_pfn = (paddr + size) >> PAGE_SHIFT;
pxm = memory_affinity->proximity_domain;
if (num_memory_chunks >= MAXCHUNKS) {
printk("Too many mem chunks in SRAT. Ignoring %lld MBytes at %llx\n",
size/(1024*1024), paddr);
return;
}
/* Insertion sort based on base address */
pend = &node_memory_chunk[num_memory_chunks];
for (p = &node_memory_chunk[0]; p < pend; p++) {
if (start_pfn < p->start_pfn)
break;
}
if (p < pend) {
for (q = pend; q >= p; q--)
*(q + 1) = *q;
}
p->start_pfn = start_pfn;
p->end_pfn = end_pfn;
p->pxm = pxm;
num_memory_chunks++;
printk("Memory range 0x%lX to 0x%lX (type 0x%X) in proximity domain 0x%02X %s\n",
start_pfn, end_pfn,
memory_affinity->memory_type,
memory_affinity->proximity_domain,
(memory_affinity->flags.hot_pluggable ?
"enabled and removable" : "enabled" ) );
}
/*
* The SRAT table always lists ascending addresses, so can always
* assume that the first "start" address that you see is the real
* start of the node, and that the current "end" address is after
* the previous one.
*/
static __init void node_read_chunk(int nid, struct node_memory_chunk_s *memory_chunk)
{
/*
* Only add present memory as told by the e820.
* There is no guarantee from the SRAT that the memory it
* enumerates is present at boot time because it represents
* *possible* memory hotplug areas the same as normal RAM.
*/
if (memory_chunk->start_pfn >= max_pfn) {
printk (KERN_INFO "Ignoring SRAT pfns: 0x%08lx -> %08lx\n",
memory_chunk->start_pfn, memory_chunk->end_pfn);
return;
}
if (memory_chunk->nid != nid)
return;
if (!node_has_online_mem(nid))
node_start_pfn[nid] = memory_chunk->start_pfn;
if (node_start_pfn[nid] > memory_chunk->start_pfn)
node_start_pfn[nid] = memory_chunk->start_pfn;
if (node_end_pfn[nid] < memory_chunk->end_pfn)
node_end_pfn[nid] = memory_chunk->end_pfn;
}
/* Parse the ACPI Static Resource Affinity Table */
static int __init acpi20_parse_srat(struct acpi_table_srat *sratp)
{
u8 *start, *end, *p;
int i, j, nid;
start = (u8 *)(&(sratp->reserved) + 1); /* skip header */
p = start;
end = (u8 *)sratp + sratp->header.length;
memset(pxm_bitmap, 0, sizeof(pxm_bitmap)); /* init proximity domain bitmap */
memset(node_memory_chunk, 0, sizeof(node_memory_chunk));
num_memory_chunks = 0;
while (p < end) {
switch (*p) {
case ACPI_SRAT_PROCESSOR_AFFINITY:
parse_cpu_affinity_structure(p);
break;
case ACPI_SRAT_MEMORY_AFFINITY:
parse_memory_affinity_structure(p);
break;
default:
printk("ACPI 2.0 SRAT: unknown entry skipped: type=0x%02X, len=%d\n", p[0], p[1]);
break;
}
p += p[1];
if (p[1] == 0) {
printk("acpi20_parse_srat: Entry length value is zero;"
" can't parse any further!\n");
break;
}
}
if (num_memory_chunks == 0) {
printk("could not finy any ACPI SRAT memory areas.\n");
goto out_fail;
}
/* Calculate total number of nodes in system from PXM bitmap and create
* a set of sequential node IDs starting at zero. (ACPI doesn't seem
* to specify the range of _PXM values.)
*/
/*
* MCD - we no longer HAVE to number nodes sequentially. PXM domain
* numbers could go as high as 256, and MAX_NUMNODES for i386 is typically
* 32, so we will continue numbering them in this manner until MAX_NUMNODES
* approaches MAX_PXM_DOMAINS for i386.
*/
nodes_clear(node_online_map);
for (i = 0; i < MAX_PXM_DOMAINS; i++) {
if (BMAP_TEST(pxm_bitmap, i)) {
int nid = acpi_map_pxm_to_node(i);
node_set_online(nid);
}
}
BUG_ON(num_online_nodes() == 0);
/* set cnode id in memory chunk structure */
for (i = 0; i < num_memory_chunks; i++)
node_memory_chunk[i].nid = pxm_to_node(node_memory_chunk[i].pxm);
printk("pxm bitmap: ");
for (i = 0; i < sizeof(pxm_bitmap); i++) {
printk("%02X ", pxm_bitmap[i]);
}
printk("\n");
printk("Number of logical nodes in system = %d\n", num_online_nodes());
printk("Number of memory chunks in system = %d\n", num_memory_chunks);
for (i = 0; i < MAX_APICID; i++)
apicid_2_node[i] = pxm_to_node(apicid_to_pxm[i]);
for (j = 0; j < num_memory_chunks; j++){
struct node_memory_chunk_s * chunk = &node_memory_chunk[j];
printk("chunk %d nid %d start_pfn %08lx end_pfn %08lx\n",
j, chunk->nid, chunk->start_pfn, chunk->end_pfn);
node_read_chunk(chunk->nid, chunk);
add_active_range(chunk->nid, chunk->start_pfn, chunk->end_pfn);
}
for_each_online_node(nid) {
unsigned long start = node_start_pfn[nid];
unsigned long end = node_end_pfn[nid];
memory_present(nid, start, end);
node_remap_size[nid] = node_memmap_size_bytes(nid, start, end);
}
return 1;
out_fail:
return 0;
}
int __init get_memcfg_from_srat(void)
{
struct acpi_table_header *header = NULL;
struct acpi_table_rsdp *rsdp = NULL;
struct acpi_table_rsdt *rsdt = NULL;
struct acpi_pointer *rsdp_address = NULL;
struct acpi_table_rsdt saved_rsdt;
int tables = 0;
int i = 0;
if (ACPI_FAILURE(acpi_find_root_pointer(ACPI_PHYSICAL_ADDRESSING,
rsdp_address))) {
printk("%s: System description tables not found\n",
__FUNCTION__);
goto out_err;
}
if (rsdp_address->pointer_type == ACPI_PHYSICAL_POINTER) {
printk("%s: assigning address to rsdp\n", __FUNCTION__);
rsdp = (struct acpi_table_rsdp *)
(u32)rsdp_address->pointer.physical;
} else {
printk("%s: rsdp_address is not a physical pointer\n", __FUNCTION__);
goto out_err;
}
if (!rsdp) {
printk("%s: Didn't find ACPI root!\n", __FUNCTION__);
goto out_err;
}
printk(KERN_INFO "%.8s v%d [%.6s]\n", rsdp->signature, rsdp->revision,
rsdp->oem_id);
if (strncmp(rsdp->signature, RSDP_SIG,strlen(RSDP_SIG))) {
printk(KERN_WARNING "%s: RSDP table signature incorrect\n", __FUNCTION__);
goto out_err;
}
rsdt = (struct acpi_table_rsdt *)
boot_ioremap(rsdp->rsdt_address, sizeof(struct acpi_table_rsdt));
if (!rsdt) {
printk(KERN_WARNING
"%s: ACPI: Invalid root system description tables (RSDT)\n",
__FUNCTION__);
goto out_err;
}
header = & rsdt->header;
if (strncmp(header->signature, RSDT_SIG, strlen(RSDT_SIG))) {
printk(KERN_WARNING "ACPI: RSDT signature incorrect\n");
goto out_err;
}
/*
* The number of tables is computed by taking the
* size of all entries (header size minus total
* size of RSDT) divided by the size of each entry
* (4-byte table pointers).
*/
tables = (header->length - sizeof(struct acpi_table_header)) / 4;
if (!tables)
goto out_err;
memcpy(&saved_rsdt, rsdt, sizeof(saved_rsdt));
if (saved_rsdt.header.length > sizeof(saved_rsdt)) {
printk(KERN_WARNING "ACPI: Too big length in RSDT: %d\n",
saved_rsdt.header.length);
goto out_err;
}
printk("Begin SRAT table scan....\n");
for (i = 0; i < tables; i++) {
/* Map in header, then map in full table length. */
header = (struct acpi_table_header *)
boot_ioremap(saved_rsdt.entry[i], sizeof(struct acpi_table_header));
if (!header)
break;
header = (struct acpi_table_header *)
boot_ioremap(saved_rsdt.entry[i], header->length);
if (!header)
break;
if (strncmp((char *) &header->signature, "SRAT", 4))
continue;
/* we've found the srat table. don't need to look at any more tables */
return acpi20_parse_srat((struct acpi_table_srat *)header);
}
out_err:
remove_all_active_ranges();
printk("failed to get NUMA memory information from SRAT table\n");
return 0;
}