linux/arch/x86/mm/numa_32.c

455 lines
14 KiB
C
Raw Normal View History

/*
* Written by: Patricia Gaughen <gone@us.ibm.com>, IBM Corporation
* August 2002: added remote node KVA remap - Martin J. Bligh
*
* 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.
*/
#include <linux/mm.h>
#include <linux/bootmem.h>
#include <linux/mmzone.h>
#include <linux/highmem.h>
#include <linux/initrd.h>
#include <linux/nodemask.h>
#include <linux/module.h>
#include <linux/kexec.h>
#include <linux/pfn.h>
#include <linux/swap.h>
#include <linux/acpi.h>
#include <asm/e820.h>
#include <asm/setup.h>
#include <asm/mmzone.h>
#include <asm/bios_ebda.h>
#include <asm/proto.h>
struct pglist_data *node_data[MAX_NUMNODES] __read_mostly;
EXPORT_SYMBOL(node_data);
/*
* numa interface - we expect the numa architecture specific code to have
* populated the following initialisation.
*
* 1) node_online_map - the map of all nodes configured (online) in the system
* 2) node_start_pfn - the starting page frame number for a node
* 3) node_end_pfn - the ending page fram number for a node
*/
unsigned long node_start_pfn[MAX_NUMNODES] __read_mostly;
unsigned long node_end_pfn[MAX_NUMNODES] __read_mostly;
#ifdef CONFIG_DISCONTIGMEM
/*
* 4) physnode_map - the mapping between a pfn and owning node
* physnode_map keeps track of the physical memory layout of a generic
* numa node on a 64Mb break (each element of the array will
* represent 64Mb of memory and will be marked by the node id. so,
* if the first gig is on node 0, and the second gig is on node 1
* physnode_map will contain:
*
* physnode_map[0-15] = 0;
* physnode_map[16-31] = 1;
* physnode_map[32- ] = -1;
*/
s8 physnode_map[MAX_ELEMENTS] __read_mostly = { [0 ... (MAX_ELEMENTS - 1)] = -1};
EXPORT_SYMBOL(physnode_map);
void memory_present(int nid, unsigned long start, unsigned long end)
{
unsigned long pfn;
printk(KERN_INFO "Node: %d, start_pfn: %lx, end_pfn: %lx\n",
nid, start, end);
printk(KERN_DEBUG " Setting physnode_map array to node %d for pfns:\n", nid);
printk(KERN_DEBUG " ");
for (pfn = start; pfn < end; pfn += PAGES_PER_ELEMENT) {
physnode_map[pfn / PAGES_PER_ELEMENT] = nid;
printk(KERN_CONT "%lx ", pfn);
}
printk(KERN_CONT "\n");
}
unsigned long node_memmap_size_bytes(int nid, unsigned long start_pfn,
unsigned long end_pfn)
{
unsigned long nr_pages = end_pfn - start_pfn;
if (!nr_pages)
return 0;
return (nr_pages + 1) * sizeof(struct page);
}
#endif
extern unsigned long find_max_low_pfn(void);
extern unsigned long highend_pfn, highstart_pfn;
#define LARGE_PAGE_BYTES (PTRS_PER_PTE * PAGE_SIZE)
unsigned long node_remap_size[MAX_NUMNODES];
static void *node_remap_start_vaddr[MAX_NUMNODES];
void set_pmd_pfn(unsigned long vaddr, unsigned long pfn, pgprot_t flags);
static unsigned long kva_start_pfn;
static unsigned long kva_pages;
/*
* FLAT - support for basic PC memory model with discontig enabled, essentially
* a single node with all available processors in it with a flat
* memory map.
*/
int __init get_memcfg_numa_flat(void)
{
printk(KERN_DEBUG "NUMA - single node, flat memory mode\n");
node_start_pfn[0] = 0;
node_end_pfn[0] = max_pfn;
e820_register_active_regions(0, 0, max_pfn);
memory_present(0, 0, max_pfn);
node_remap_size[0] = node_memmap_size_bytes(0, 0, max_pfn);
/* Indicate there is one node available. */
nodes_clear(node_online_map);
node_set_online(0);
return 1;
}
/*
* Find the highest page frame number we have available for the node
*/
static void __init propagate_e820_map_node(int nid)
{
if (node_end_pfn[nid] > max_pfn)
node_end_pfn[nid] = max_pfn;
/*
* if a user has given mem=XXXX, then we need to make sure
* that the node _starts_ before that, too, not just ends
*/
if (node_start_pfn[nid] > max_pfn)
node_start_pfn[nid] = max_pfn;
BUG_ON(node_start_pfn[nid] > node_end_pfn[nid]);
}
/*
* Allocate memory for the pg_data_t for this node via a crude pre-bootmem
* method. For node zero take this from the bottom of memory, for
* subsequent nodes place them at node_remap_start_vaddr which contains
* node local data in physically node local memory. See setup_memory()
* for details.
*/
static void __init allocate_pgdat(int nid)
{
char buf[16];
if (node_has_online_mem(nid) && node_remap_start_vaddr[nid])
NODE_DATA(nid) = (pg_data_t *)node_remap_start_vaddr[nid];
else {
unsigned long pgdat_phys;
pgdat_phys = find_e820_area(min_low_pfn<<PAGE_SHIFT,
max_pfn_mapped<<PAGE_SHIFT,
sizeof(pg_data_t),
PAGE_SIZE);
NODE_DATA(nid) = (pg_data_t *)(pfn_to_kaddr(pgdat_phys>>PAGE_SHIFT));
memset(buf, 0, sizeof(buf));
sprintf(buf, "NODE_DATA %d", nid);
reserve_early(pgdat_phys, pgdat_phys + sizeof(pg_data_t), buf);
}
printk(KERN_DEBUG "allocate_pgdat: node %d NODE_DATA %08lx\n",
nid, (unsigned long)NODE_DATA(nid));
}
/*
x86: reinstate numa remap for SPARSEMEM on x86 NUMA systems Recent kernels have been panic'ing trying to allocate memory early in boot, in __alloc_pages: BUG: unable to handle kernel paging request at 00001568 IP: [<c10407b6>] __alloc_pages+0x33/0x2cc *pdpt = 00000000013a5001 *pde = 0000000000000000 Oops: 0000 [#1] SMP Modules linked in: Pid: 1, comm: swapper Not tainted (2.6.25 #78) EIP: 0060:[<c10407b6>] EFLAGS: 00010246 CPU: 0 EIP is at __alloc_pages+0x33/0x2cc EAX: 00001564 EBX: 000412d0 ECX: 00001564 EDX: 000005c3 ESI: f78012a0 EDI: 00000001 EBP: 00001564 ESP: f7871e50 DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068 Process swapper (pid: 1, ti=f7870000 task=f786f670 task.ti=f7870000) Stack: 00000000 f786f670 00000010 00000000 0000b700 000412d0 f78012a0 00000001 00000000 c105b64d 00000000 000412d0 f78012a0 f7803120 00000000 c105c1c5 00000010 f7803144 000412d0 00000001 f7803130 f7803120 f78012a0 00000001 Call Trace: [<c105b64d>] kmem_getpages+0x94/0x129 [<c105c1c5>] cache_grow+0x8f/0x123 [<c105c689>] ____cache_alloc_node+0xb9/0xe4 [<c105c999>] kmem_cache_alloc_node+0x92/0xd2 [<c1018929>] build_sched_domains+0x536/0x70d [<c100b63c>] do_flush_tlb_all+0x0/0x3f [<c100b63c>] do_flush_tlb_all+0x0/0x3f [<c10572d6>] interleave_nodes+0x23/0x5a [<c105c44f>] alternate_node_alloc+0x43/0x5b [<c1018b47>] arch_init_sched_domains+0x46/0x51 [<c136e85e>] kernel_init+0x0/0x82 [<c137ac19>] sched_init_smp+0x10/0xbb [<c136e8a1>] kernel_init+0x43/0x82 [<c10035cf>] kernel_thread_helper+0x7/0x10 Debugging this showed that the NODE_DATA() for nodes other than node 0 were all NULL. Tracing this back showed that the NODE_DATA() pointers were being initialised to each nodes remap space. However under SPARSEMEM remap is disabled which leads to the pgdat's being placed incorrectly at kernel virtual address 0. Leading to the panic when attempting to allocate memory from these nodes. Numa remap was disabled in the commit below. This occured while fixing problems triggered when attempting to boot x86_32 NUMA SPARSEMEM kernels on non-numa hardware. x86: make NUMA work on 32-bit commit 1b000a5dbeb2f34bc03d45ebdf3f6d24a60c3aed The real problem is believed to be related to other alignment issues in the regions blocked out from the bootmem allocator for small memory systems, and has been fixed separately. Therefore re-enable remap for SPARSMEM, which fixes pgdat allocation issues. Testing confirms that SPARSMEM NUMA kernels will boot correctly with this part of the change reverted. Signed-off-by: Andy Whitcroft <apw@shadowen.org> Acked-by: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-05-20 18:01:08 +08:00
* In the DISCONTIGMEM and SPARSEMEM memory model, a portion of the kernel
* virtual address space (KVA) is reserved and portions of nodes are mapped
* using it. This is to allow node-local memory to be allocated for
* structures that would normally require ZONE_NORMAL. The memory is
* allocated with alloc_remap() and callers should be prepared to allocate
* from the bootmem allocator instead.
*/
static unsigned long node_remap_start_pfn[MAX_NUMNODES];
static void *node_remap_end_vaddr[MAX_NUMNODES];
static void *node_remap_alloc_vaddr[MAX_NUMNODES];
static unsigned long node_remap_offset[MAX_NUMNODES];
void *alloc_remap(int nid, unsigned long size)
{
void *allocation = node_remap_alloc_vaddr[nid];
size = ALIGN(size, L1_CACHE_BYTES);
if (!allocation || (allocation + size) >= node_remap_end_vaddr[nid])
return NULL;
node_remap_alloc_vaddr[nid] += size;
memset(allocation, 0, size);
return allocation;
}
static void __init remap_numa_kva(void)
{
void *vaddr;
unsigned long pfn;
int node;
for_each_online_node(node) {
printk(KERN_DEBUG "remap_numa_kva: node %d\n", node);
for (pfn=0; pfn < node_remap_size[node]; pfn += PTRS_PER_PTE) {
vaddr = node_remap_start_vaddr[node]+(pfn<<PAGE_SHIFT);
printk(KERN_DEBUG "remap_numa_kva: %08lx to pfn %08lx\n",
(unsigned long)vaddr,
node_remap_start_pfn[node] + pfn);
set_pmd_pfn((ulong) vaddr,
node_remap_start_pfn[node] + pfn,
PAGE_KERNEL_LARGE);
}
}
}
#ifdef CONFIG_HIBERNATION
/**
* resume_map_numa_kva - add KVA mapping to the temporary page tables created
* during resume from hibernation
* @pgd_base - temporary resume page directory
*/
void resume_map_numa_kva(pgd_t *pgd_base)
{
int node;
for_each_online_node(node) {
unsigned long start_va, start_pfn, size, pfn;
start_va = (unsigned long)node_remap_start_vaddr[node];
start_pfn = node_remap_start_pfn[node];
size = node_remap_size[node];
printk(KERN_DEBUG "%s: node %d\n", __func__, node);
for (pfn = 0; pfn < size; pfn += PTRS_PER_PTE) {
unsigned long vaddr = start_va + (pfn << PAGE_SHIFT);
pgd_t *pgd = pgd_base + pgd_index(vaddr);
pud_t *pud = pud_offset(pgd, vaddr);
pmd_t *pmd = pmd_offset(pud, vaddr);
set_pmd(pmd, pfn_pmd(start_pfn + pfn,
PAGE_KERNEL_LARGE_EXEC));
printk(KERN_DEBUG "%s: %08lx -> pfn %08lx\n",
__func__, vaddr, start_pfn + pfn);
}
}
}
#endif
static unsigned long calculate_numa_remap_pages(void)
{
int nid;
unsigned long size, reserve_pages = 0;
for_each_online_node(nid) {
u64 node_kva_target;
u64 node_kva_final;
/*
* The acpi/srat node info can show hot-add memroy zones
* where memory could be added but not currently present.
*/
printk(KERN_DEBUG "node %d pfn: [%lx - %lx]\n",
nid, node_start_pfn[nid], node_end_pfn[nid]);
if (node_start_pfn[nid] > max_pfn)
continue;
if (!node_end_pfn[nid])
continue;
if (node_end_pfn[nid] > max_pfn)
node_end_pfn[nid] = max_pfn;
/* ensure the remap includes space for the pgdat. */
size = node_remap_size[nid] + sizeof(pg_data_t);
/* convert size to large (pmd size) pages, rounding up */
size = (size + LARGE_PAGE_BYTES - 1) / LARGE_PAGE_BYTES;
/* now the roundup is correct, convert to PAGE_SIZE pages */
size = size * PTRS_PER_PTE;
node_kva_target = round_down(node_end_pfn[nid] - size,
PTRS_PER_PTE);
node_kva_target <<= PAGE_SHIFT;
do {
node_kva_final = find_e820_area(node_kva_target,
((u64)node_end_pfn[nid])<<PAGE_SHIFT,
((u64)size)<<PAGE_SHIFT,
LARGE_PAGE_BYTES);
node_kva_target -= LARGE_PAGE_BYTES;
} while (node_kva_final == -1ULL &&
(node_kva_target>>PAGE_SHIFT) > (node_start_pfn[nid]));
if (node_kva_final == -1ULL)
panic("Can not get kva ram\n");
node_remap_size[nid] = size;
node_remap_offset[nid] = reserve_pages;
reserve_pages += size;
printk(KERN_DEBUG "Reserving %ld pages of KVA for lmem_map of"
" node %d at %llx\n",
size, nid, node_kva_final>>PAGE_SHIFT);
/*
* prevent kva address below max_low_pfn want it on system
* with less memory later.
* layout will be: KVA address , KVA RAM
*
* we are supposed to only record the one less then max_low_pfn
* but we could have some hole in high memory, and it will only
* check page_is_ram(pfn) && !page_is_reserved_early(pfn) to decide
* to use it as free.
* So reserve_early here, hope we don't run out of that array
*/
reserve_early(node_kva_final,
node_kva_final+(((u64)size)<<PAGE_SHIFT),
"KVA RAM");
node_remap_start_pfn[nid] = node_kva_final>>PAGE_SHIFT;
remove_active_range(nid, node_remap_start_pfn[nid],
node_remap_start_pfn[nid] + size);
}
printk(KERN_INFO "Reserving total of %lx pages for numa KVA remap\n",
reserve_pages);
return reserve_pages;
}
static void init_remap_allocator(int nid)
{
node_remap_start_vaddr[nid] = pfn_to_kaddr(
kva_start_pfn + node_remap_offset[nid]);
node_remap_end_vaddr[nid] = node_remap_start_vaddr[nid] +
(node_remap_size[nid] * PAGE_SIZE);
node_remap_alloc_vaddr[nid] = node_remap_start_vaddr[nid] +
ALIGN(sizeof(pg_data_t), PAGE_SIZE);
printk(KERN_DEBUG "node %d will remap to vaddr %08lx - %08lx\n", nid,
(ulong) node_remap_start_vaddr[nid],
(ulong) node_remap_end_vaddr[nid]);
}
void __init initmem_init(unsigned long start_pfn,
unsigned long end_pfn)
{
int nid;
long kva_target_pfn;
/*
* When mapping a NUMA machine we allocate the node_mem_map arrays
* from node local memory. They are then mapped directly into KVA
* between zone normal and vmalloc space. Calculate the size of
* this space and use it to adjust the boundary between ZONE_NORMAL
* and ZONE_HIGHMEM.
*/
get_memcfg_numa();
kva_pages = roundup(calculate_numa_remap_pages(), PTRS_PER_PTE);
kva_target_pfn = round_down(max_low_pfn - kva_pages, PTRS_PER_PTE);
do {
kva_start_pfn = find_e820_area(kva_target_pfn<<PAGE_SHIFT,
max_low_pfn<<PAGE_SHIFT,
kva_pages<<PAGE_SHIFT,
PTRS_PER_PTE<<PAGE_SHIFT) >> PAGE_SHIFT;
kva_target_pfn -= PTRS_PER_PTE;
} while (kva_start_pfn == -1UL && kva_target_pfn > min_low_pfn);
x86: make NUMA work on 32-bit again On 32-bit NUMA, the memmap representing struct pages on each node is allocated from node-local memory if possible. As only node-0 has memory from ZONE_NORMAL, the memmap must be mapped into low memory. This is done by reserving space in the Kernel Virtual Area (KVA) for the memmap belonging to other nodes by taking pages from the end of ZONE_NORMAL and remapping the other nodes memmap into those virtual addresses. The node boundaries are then adjusted so that the region of pages is not used and it is marked as reserved in the bootmem allocator. This reserved portion of the KVA is PMD aligned althought strictly speaking that requirement could be lifted (see thread at http://lkml.org/lkml/2007/8/24/220). The problem is that when aligned, there may be a portion of ZONE_NORMAL at the end that is not used for memmap and does not have an initialised memmap nor is it marked reserved in the bootmem allocator. Later in the boot process, these pages are freed and a storm of Bad page state messages result. This patch marks these pages reserved that are wasted due to alignment in the bootmem allocator so they are not accidently freed. It is worth noting that memory from node-0 is wasted where it could have been put into ZONE_HIGHMEM on NUMA machines. Worse, the KVA is always reserved from the location of real memory even when there is plenty of spare virtual address space. This patch also makes sure that reserve_bootmem() is not called with a 0-length size in numa_kva_reserve(). When this happens, it usually means that a kernel built for Summit is being booted on a normal machine. The resulting BUG_ON() is misleading so it is caught here. Signed-off-by: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Thomas Gleixner <tglx@linutronix.de> Signed-off-by: Ingo Molnar <mingo@elte.hu>
2008-01-30 20:32:54 +08:00
if (kva_start_pfn == -1UL)
panic("Can not get kva space\n");
printk(KERN_INFO "kva_start_pfn ~ %lx max_low_pfn ~ %lx\n",
kva_start_pfn, max_low_pfn);
printk(KERN_INFO "max_pfn = %lx\n", max_pfn);
/* avoid clash with initrd */
reserve_early(kva_start_pfn<<PAGE_SHIFT,
(kva_start_pfn + kva_pages)<<PAGE_SHIFT,
"KVA PG");
#ifdef CONFIG_HIGHMEM
highstart_pfn = highend_pfn = max_pfn;
if (max_pfn > max_low_pfn)
highstart_pfn = max_low_pfn;
printk(KERN_NOTICE "%ldMB HIGHMEM available.\n",
pages_to_mb(highend_pfn - highstart_pfn));
num_physpages = highend_pfn;
high_memory = (void *) __va(highstart_pfn * PAGE_SIZE - 1) + 1;
#else
num_physpages = max_low_pfn;
high_memory = (void *) __va(max_low_pfn * PAGE_SIZE - 1) + 1;
#endif
printk(KERN_NOTICE "%ldMB LOWMEM available.\n",
pages_to_mb(max_low_pfn));
printk(KERN_DEBUG "max_low_pfn = %lx, highstart_pfn = %lx\n",
max_low_pfn, highstart_pfn);
printk(KERN_DEBUG "Low memory ends at vaddr %08lx\n",
(ulong) pfn_to_kaddr(max_low_pfn));
for_each_online_node(nid) {
init_remap_allocator(nid);
allocate_pgdat(nid);
}
remap_numa_kva();
printk(KERN_DEBUG "High memory starts at vaddr %08lx\n",
(ulong) pfn_to_kaddr(highstart_pfn));
for_each_online_node(nid)
propagate_e820_map_node(nid);
for_each_online_node(nid) {
memset(NODE_DATA(nid), 0, sizeof(struct pglist_data));
NODE_DATA(nid)->bdata = &bootmem_node_data[nid];
}
setup_bootmem_allocator();
}
#ifdef CONFIG_MEMORY_HOTPLUG
static int paddr_to_nid(u64 addr)
{
int nid;
unsigned long pfn = PFN_DOWN(addr);
for_each_node(nid)
if (node_start_pfn[nid] <= pfn &&
pfn < node_end_pfn[nid])
return nid;
return -1;
}
/*
* This function is used to ask node id BEFORE memmap and mem_section's
* initialization (pfn_to_nid() can't be used yet).
* If _PXM is not defined on ACPI's DSDT, node id must be found by this.
*/
int memory_add_physaddr_to_nid(u64 addr)
{
int nid = paddr_to_nid(addr);
return (nid >= 0) ? nid : 0;
}
EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid);
#endif