linux/mm/page_alloc.c

2747 lines
68 KiB
C
Raw Normal View History

/*
* linux/mm/page_alloc.c
*
* Manages the free list, the system allocates free pages here.
* Note that kmalloc() lives in slab.c
*
* Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
* Swap reorganised 29.12.95, Stephen Tweedie
* Support of BIGMEM added by Gerhard Wichert, Siemens AG, July 1999
* Reshaped it to be a zoned allocator, Ingo Molnar, Red Hat, 1999
* Discontiguous memory support, Kanoj Sarcar, SGI, Nov 1999
* Zone balancing, Kanoj Sarcar, SGI, Jan 2000
* Per cpu hot/cold page lists, bulk allocation, Martin J. Bligh, Sept 2002
* (lots of bits borrowed from Ingo Molnar & Andrew Morton)
*/
#include <linux/config.h>
#include <linux/stddef.h>
#include <linux/mm.h>
#include <linux/swap.h>
#include <linux/interrupt.h>
#include <linux/pagemap.h>
#include <linux/bootmem.h>
#include <linux/compiler.h>
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/suspend.h>
#include <linux/pagevec.h>
#include <linux/blkdev.h>
#include <linux/slab.h>
#include <linux/notifier.h>
#include <linux/topology.h>
#include <linux/sysctl.h>
#include <linux/cpu.h>
#include <linux/cpuset.h>
#include <linux/memory_hotplug.h>
#include <linux/nodemask.h>
#include <linux/vmalloc.h>
#include <linux/mempolicy.h>
#include <asm/tlbflush.h>
#include "internal.h"
/*
* MCD - HACK: Find somewhere to initialize this EARLY, or make this
* initializer cleaner
*/
nodemask_t node_online_map __read_mostly = { { [0] = 1UL } };
EXPORT_SYMBOL(node_online_map);
nodemask_t node_possible_map __read_mostly = NODE_MASK_ALL;
EXPORT_SYMBOL(node_possible_map);
struct pglist_data *pgdat_list __read_mostly;
unsigned long totalram_pages __read_mostly;
unsigned long totalhigh_pages __read_mostly;
long nr_swap_pages;
int percpu_pagelist_fraction;
static void __free_pages_ok(struct page *page, unsigned int order);
/*
* results with 256, 32 in the lowmem_reserve sysctl:
* 1G machine -> (16M dma, 800M-16M normal, 1G-800M high)
* 1G machine -> (16M dma, 784M normal, 224M high)
* NORMAL allocation will leave 784M/256 of ram reserved in the ZONE_DMA
* HIGHMEM allocation will leave 224M/32 of ram reserved in ZONE_NORMAL
* HIGHMEM allocation will (224M+784M)/256 of ram reserved in ZONE_DMA
[PATCH] x86_64: Add 4GB DMA32 zone Add a new 4GB GFP_DMA32 zone between the GFP_DMA and GFP_NORMAL zones. As a bit of historical background: when the x86-64 port was originally designed we had some discussion if we should use a 16MB DMA zone like i386 or a 4GB DMA zone like IA64 or both. Both was ruled out at this point because it was in early 2.4 when VM is still quite shakey and had bad troubles even dealing with one DMA zone. We settled on the 16MB DMA zone mainly because we worried about older soundcards and the floppy. But this has always caused problems since then because device drivers had trouble getting enough DMA able memory. These days the VM works much better and the wide use of NUMA has proven it can deal with many zones successfully. So this patch adds both zones. This helps drivers who need a lot of memory below 4GB because their hardware is not accessing more (graphic drivers - proprietary and free ones, video frame buffer drivers, sound drivers etc.). Previously they could only use IOMMU+16MB GFP_DMA, which was not enough memory. Another common problem is that hardware who has full memory addressing for >4GB misses it for some control structures in memory (like transmit rings or other metadata). They tended to allocate memory in the 16MB GFP_DMA or the IOMMU/swiotlb then using pci_alloc_consistent, but that can tie up a lot of precious 16MB GFPDMA/IOMMU/swiotlb memory (even on AMD systems the IOMMU tends to be quite small) especially if you have many devices. With the new zone pci_alloc_consistent can just put this stuff into memory below 4GB which works better. One argument was still if the zone should be 4GB or 2GB. The main motivation for 2GB would be an unnamed not so unpopular hardware raid controller (mostly found in older machines from a particular four letter company) who has a strange 2GB restriction in firmware. But that one works ok with swiotlb/IOMMU anyways, so it doesn't really need GFP_DMA32. I chose 4GB to be compatible with IA64 and because it seems to be the most common restriction. The new zone is so far added only for x86-64. For other architectures who don't set up this new zone nothing changes. Architectures can set a compatibility define in Kconfig CONFIG_DMA_IS_DMA32 that will define GFP_DMA32 as GFP_DMA. Otherwise it's a nop because on 32bit architectures it's normally not needed because GFP_NORMAL (=0) is DMA able enough. One problem is still that GFP_DMA means different things on different architectures. e.g. some drivers used to have #ifdef ia64 use GFP_DMA (trusting it to be 4GB) #elif __x86_64__ (use other hacks like the swiotlb because 16MB is not enough) ... . This was quite ugly and is now obsolete. These should be now converted to use GFP_DMA32 unconditionally. I haven't done this yet. Or best only use pci_alloc_consistent/dma_alloc_coherent which will use GFP_DMA32 transparently. Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-11-05 16:25:53 +00:00
*
* TBD: should special case ZONE_DMA32 machines here - in those we normally
* don't need any ZONE_NORMAL reservation
*/
[PATCH] x86_64: Add 4GB DMA32 zone Add a new 4GB GFP_DMA32 zone between the GFP_DMA and GFP_NORMAL zones. As a bit of historical background: when the x86-64 port was originally designed we had some discussion if we should use a 16MB DMA zone like i386 or a 4GB DMA zone like IA64 or both. Both was ruled out at this point because it was in early 2.4 when VM is still quite shakey and had bad troubles even dealing with one DMA zone. We settled on the 16MB DMA zone mainly because we worried about older soundcards and the floppy. But this has always caused problems since then because device drivers had trouble getting enough DMA able memory. These days the VM works much better and the wide use of NUMA has proven it can deal with many zones successfully. So this patch adds both zones. This helps drivers who need a lot of memory below 4GB because their hardware is not accessing more (graphic drivers - proprietary and free ones, video frame buffer drivers, sound drivers etc.). Previously they could only use IOMMU+16MB GFP_DMA, which was not enough memory. Another common problem is that hardware who has full memory addressing for >4GB misses it for some control structures in memory (like transmit rings or other metadata). They tended to allocate memory in the 16MB GFP_DMA or the IOMMU/swiotlb then using pci_alloc_consistent, but that can tie up a lot of precious 16MB GFPDMA/IOMMU/swiotlb memory (even on AMD systems the IOMMU tends to be quite small) especially if you have many devices. With the new zone pci_alloc_consistent can just put this stuff into memory below 4GB which works better. One argument was still if the zone should be 4GB or 2GB. The main motivation for 2GB would be an unnamed not so unpopular hardware raid controller (mostly found in older machines from a particular four letter company) who has a strange 2GB restriction in firmware. But that one works ok with swiotlb/IOMMU anyways, so it doesn't really need GFP_DMA32. I chose 4GB to be compatible with IA64 and because it seems to be the most common restriction. The new zone is so far added only for x86-64. For other architectures who don't set up this new zone nothing changes. Architectures can set a compatibility define in Kconfig CONFIG_DMA_IS_DMA32 that will define GFP_DMA32 as GFP_DMA. Otherwise it's a nop because on 32bit architectures it's normally not needed because GFP_NORMAL (=0) is DMA able enough. One problem is still that GFP_DMA means different things on different architectures. e.g. some drivers used to have #ifdef ia64 use GFP_DMA (trusting it to be 4GB) #elif __x86_64__ (use other hacks like the swiotlb because 16MB is not enough) ... . This was quite ugly and is now obsolete. These should be now converted to use GFP_DMA32 unconditionally. I haven't done this yet. Or best only use pci_alloc_consistent/dma_alloc_coherent which will use GFP_DMA32 transparently. Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-11-05 16:25:53 +00:00
int sysctl_lowmem_reserve_ratio[MAX_NR_ZONES-1] = { 256, 256, 32 };
EXPORT_SYMBOL(totalram_pages);
/*
* Used by page_zone() to look up the address of the struct zone whose
* id is encoded in the upper bits of page->flags
*/
struct zone *zone_table[1 << ZONETABLE_SHIFT] __read_mostly;
EXPORT_SYMBOL(zone_table);
[PATCH] x86_64: Add 4GB DMA32 zone Add a new 4GB GFP_DMA32 zone between the GFP_DMA and GFP_NORMAL zones. As a bit of historical background: when the x86-64 port was originally designed we had some discussion if we should use a 16MB DMA zone like i386 or a 4GB DMA zone like IA64 or both. Both was ruled out at this point because it was in early 2.4 when VM is still quite shakey and had bad troubles even dealing with one DMA zone. We settled on the 16MB DMA zone mainly because we worried about older soundcards and the floppy. But this has always caused problems since then because device drivers had trouble getting enough DMA able memory. These days the VM works much better and the wide use of NUMA has proven it can deal with many zones successfully. So this patch adds both zones. This helps drivers who need a lot of memory below 4GB because their hardware is not accessing more (graphic drivers - proprietary and free ones, video frame buffer drivers, sound drivers etc.). Previously they could only use IOMMU+16MB GFP_DMA, which was not enough memory. Another common problem is that hardware who has full memory addressing for >4GB misses it for some control structures in memory (like transmit rings or other metadata). They tended to allocate memory in the 16MB GFP_DMA or the IOMMU/swiotlb then using pci_alloc_consistent, but that can tie up a lot of precious 16MB GFPDMA/IOMMU/swiotlb memory (even on AMD systems the IOMMU tends to be quite small) especially if you have many devices. With the new zone pci_alloc_consistent can just put this stuff into memory below 4GB which works better. One argument was still if the zone should be 4GB or 2GB. The main motivation for 2GB would be an unnamed not so unpopular hardware raid controller (mostly found in older machines from a particular four letter company) who has a strange 2GB restriction in firmware. But that one works ok with swiotlb/IOMMU anyways, so it doesn't really need GFP_DMA32. I chose 4GB to be compatible with IA64 and because it seems to be the most common restriction. The new zone is so far added only for x86-64. For other architectures who don't set up this new zone nothing changes. Architectures can set a compatibility define in Kconfig CONFIG_DMA_IS_DMA32 that will define GFP_DMA32 as GFP_DMA. Otherwise it's a nop because on 32bit architectures it's normally not needed because GFP_NORMAL (=0) is DMA able enough. One problem is still that GFP_DMA means different things on different architectures. e.g. some drivers used to have #ifdef ia64 use GFP_DMA (trusting it to be 4GB) #elif __x86_64__ (use other hacks like the swiotlb because 16MB is not enough) ... . This was quite ugly and is now obsolete. These should be now converted to use GFP_DMA32 unconditionally. I haven't done this yet. Or best only use pci_alloc_consistent/dma_alloc_coherent which will use GFP_DMA32 transparently. Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-11-05 16:25:53 +00:00
static char *zone_names[MAX_NR_ZONES] = { "DMA", "DMA32", "Normal", "HighMem" };
int min_free_kbytes = 1024;
unsigned long __initdata nr_kernel_pages;
unsigned long __initdata nr_all_pages;
#ifdef CONFIG_DEBUG_VM
static int page_outside_zone_boundaries(struct zone *zone, struct page *page)
{
int ret = 0;
unsigned seq;
unsigned long pfn = page_to_pfn(page);
do {
seq = zone_span_seqbegin(zone);
if (pfn >= zone->zone_start_pfn + zone->spanned_pages)
ret = 1;
else if (pfn < zone->zone_start_pfn)
ret = 1;
} while (zone_span_seqretry(zone, seq));
return ret;
}
static int page_is_consistent(struct zone *zone, struct page *page)
{
#ifdef CONFIG_HOLES_IN_ZONE
if (!pfn_valid(page_to_pfn(page)))
return 0;
#endif
if (zone != page_zone(page))
return 0;
return 1;
}
/*
* Temporary debugging check for pages not lying within a given zone.
*/
static int bad_range(struct zone *zone, struct page *page)
{
if (page_outside_zone_boundaries(zone, page))
return 1;
if (!page_is_consistent(zone, page))
return 1;
return 0;
}
#else
static inline int bad_range(struct zone *zone, struct page *page)
{
return 0;
}
#endif
static void bad_page(struct page *page)
{
printk(KERN_EMERG "Bad page state in process '%s'\n"
KERN_EMERG "page:%p flags:0x%0*lx mapping:%p mapcount:%d count:%d\n"
KERN_EMERG "Trying to fix it up, but a reboot is needed\n"
KERN_EMERG "Backtrace:\n",
current->comm, page, (int)(2*sizeof(unsigned long)),
(unsigned long)page->flags, page->mapping,
page_mapcount(page), page_count(page));
dump_stack();
page->flags &= ~(1 << PG_lru |
1 << PG_private |
1 << PG_locked |
1 << PG_active |
1 << PG_dirty |
1 << PG_reclaim |
1 << PG_slab |
1 << PG_swapcache |
1 << PG_writeback );
set_page_count(page, 0);
reset_page_mapcount(page);
page->mapping = NULL;
add_taint(TAINT_BAD_PAGE);
}
/*
* Higher-order pages are called "compound pages". They are structured thusly:
*
* The first PAGE_SIZE page is called the "head page".
*
* The remaining PAGE_SIZE pages are called "tail pages".
*
* All pages have PG_compound set. All pages have their ->private pointing at
* the head page (even the head page has this).
*
[PATCH] compound page: use page[1].lru If a compound page has its own put_page_testzero destructor (the only current example is free_huge_page), that is noted in page[1].mapping of the compound page. But that's rather a poor place to keep it: functions which call set_page_dirty_lock after get_user_pages (e.g. Infiniband's __ib_umem_release) ought to be checking first, otherwise set_page_dirty is liable to crash on what's not the address of a struct address_space. And now I'm about to make that worse: it turns out that every compound page needs a destructor, so we can no longer rely on hugetlb pages going their own special way, to avoid further problems of page->mapping reuse. For example, not many people know that: on 50% of i386 -Os builds, the first tail page of a compound page purports to be PageAnon (when its destructor has an odd address), which surprises page_add_file_rmap. Keep the compound page destructor in page[1].lru.next instead. And to free up the common pairing of mapping and index, also move compound page order from index to lru.prev. Slab reuses page->lru too: but if we ever need slab to use compound pages, it can easily stack its use above this. (akpm: decoded version of the above: the tail pages of a compound page now have ->mapping==NULL, so there's no need for the set_page_dirty[_lock]() caller to check that they're not compund pages before doing the dirty). Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-02-14 21:52:58 +00:00
* The first tail page's ->lru.next holds the address of the compound page's
* put_page() function. Its ->lru.prev holds the order of allocation.
* This usage means that zero-order pages may not be compound.
*/
static void free_compound_page(struct page *page)
{
__free_pages_ok(page, (unsigned long)page[1].lru.prev);
}
static void prep_compound_page(struct page *page, unsigned long order)
{
int i;
int nr_pages = 1 << order;
page[1].lru.next = (void *)free_compound_page; /* set dtor */
[PATCH] compound page: use page[1].lru If a compound page has its own put_page_testzero destructor (the only current example is free_huge_page), that is noted in page[1].mapping of the compound page. But that's rather a poor place to keep it: functions which call set_page_dirty_lock after get_user_pages (e.g. Infiniband's __ib_umem_release) ought to be checking first, otherwise set_page_dirty is liable to crash on what's not the address of a struct address_space. And now I'm about to make that worse: it turns out that every compound page needs a destructor, so we can no longer rely on hugetlb pages going their own special way, to avoid further problems of page->mapping reuse. For example, not many people know that: on 50% of i386 -Os builds, the first tail page of a compound page purports to be PageAnon (when its destructor has an odd address), which surprises page_add_file_rmap. Keep the compound page destructor in page[1].lru.next instead. And to free up the common pairing of mapping and index, also move compound page order from index to lru.prev. Slab reuses page->lru too: but if we ever need slab to use compound pages, it can easily stack its use above this. (akpm: decoded version of the above: the tail pages of a compound page now have ->mapping==NULL, so there's no need for the set_page_dirty[_lock]() caller to check that they're not compund pages before doing the dirty). Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-02-14 21:52:58 +00:00
page[1].lru.prev = (void *)order;
for (i = 0; i < nr_pages; i++) {
struct page *p = page + i;
__SetPageCompound(p);
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
set_page_private(p, (unsigned long)page);
}
}
static void destroy_compound_page(struct page *page, unsigned long order)
{
int i;
int nr_pages = 1 << order;
[PATCH] compound page: use page[1].lru If a compound page has its own put_page_testzero destructor (the only current example is free_huge_page), that is noted in page[1].mapping of the compound page. But that's rather a poor place to keep it: functions which call set_page_dirty_lock after get_user_pages (e.g. Infiniband's __ib_umem_release) ought to be checking first, otherwise set_page_dirty is liable to crash on what's not the address of a struct address_space. And now I'm about to make that worse: it turns out that every compound page needs a destructor, so we can no longer rely on hugetlb pages going their own special way, to avoid further problems of page->mapping reuse. For example, not many people know that: on 50% of i386 -Os builds, the first tail page of a compound page purports to be PageAnon (when its destructor has an odd address), which surprises page_add_file_rmap. Keep the compound page destructor in page[1].lru.next instead. And to free up the common pairing of mapping and index, also move compound page order from index to lru.prev. Slab reuses page->lru too: but if we ever need slab to use compound pages, it can easily stack its use above this. (akpm: decoded version of the above: the tail pages of a compound page now have ->mapping==NULL, so there's no need for the set_page_dirty[_lock]() caller to check that they're not compund pages before doing the dirty). Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-02-14 21:52:58 +00:00
if (unlikely((unsigned long)page[1].lru.prev != order))
bad_page(page);
for (i = 0; i < nr_pages; i++) {
struct page *p = page + i;
if (unlikely(!PageCompound(p) |
(page_private(p) != (unsigned long)page)))
bad_page(page);
__ClearPageCompound(p);
}
}
static inline void prep_zero_page(struct page *page, int order, gfp_t gfp_flags)
{
int i;
BUG_ON((gfp_flags & (__GFP_WAIT | __GFP_HIGHMEM)) == __GFP_HIGHMEM);
/*
* clear_highpage() will use KM_USER0, so it's a bug to use __GFP_ZERO
* and __GFP_HIGHMEM from hard or soft interrupt context.
*/
BUG_ON((gfp_flags & __GFP_HIGHMEM) && in_interrupt());
for (i = 0; i < (1 << order); i++)
clear_highpage(page + i);
}
/*
* function for dealing with page's order in buddy system.
* zone->lock is already acquired when we use these.
* So, we don't need atomic page->flags operations here.
*/
static inline unsigned long page_order(struct page *page) {
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
return page_private(page);
}
static inline void set_page_order(struct page *page, int order) {
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
set_page_private(page, order);
__SetPagePrivate(page);
}
static inline void rmv_page_order(struct page *page)
{
__ClearPagePrivate(page);
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
set_page_private(page, 0);
}
/*
* Locate the struct page for both the matching buddy in our
* pair (buddy1) and the combined O(n+1) page they form (page).
*
* 1) Any buddy B1 will have an order O twin B2 which satisfies
* the following equation:
* B2 = B1 ^ (1 << O)
* For example, if the starting buddy (buddy2) is #8 its order
* 1 buddy is #10:
* B2 = 8 ^ (1 << 1) = 8 ^ 2 = 10
*
* 2) Any buddy B will have an order O+1 parent P which
* satisfies the following equation:
* P = B & ~(1 << O)
*
* Assumption: *_mem_map is contigious at least up to MAX_ORDER
*/
static inline struct page *
__page_find_buddy(struct page *page, unsigned long page_idx, unsigned int order)
{
unsigned long buddy_idx = page_idx ^ (1 << order);
return page + (buddy_idx - page_idx);
}
static inline unsigned long
__find_combined_index(unsigned long page_idx, unsigned int order)
{
return (page_idx & ~(1 << order));
}
/*
* This function checks whether a page is free && is the buddy
* we can do coalesce a page and its buddy if
* (a) the buddy is not in a hole &&
* (b) the buddy is free &&
* (c) the buddy is on the buddy system &&
* (d) a page and its buddy have the same order.
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
* for recording page's order, we use page_private(page) and PG_private.
*
*/
static inline int page_is_buddy(struct page *page, int order)
{
#ifdef CONFIG_HOLES_IN_ZONE
if (!pfn_valid(page_to_pfn(page)))
return 0;
#endif
if (PagePrivate(page) &&
(page_order(page) == order) &&
page_count(page) == 0)
return 1;
return 0;
}
/*
* Freeing function for a buddy system allocator.
*
* The concept of a buddy system is to maintain direct-mapped table
* (containing bit values) for memory blocks of various "orders".
* The bottom level table contains the map for the smallest allocatable
* units of memory (here, pages), and each level above it describes
* pairs of units from the levels below, hence, "buddies".
* At a high level, all that happens here is marking the table entry
* at the bottom level available, and propagating the changes upward
* as necessary, plus some accounting needed to play nicely with other
* parts of the VM system.
* At each level, we keep a list of pages, which are heads of continuous
* free pages of length of (1 << order) and marked with PG_Private.Page's
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
* order is recorded in page_private(page) field.
* So when we are allocating or freeing one, we can derive the state of the
* other. That is, if we allocate a small block, and both were
* free, the remainder of the region must be split into blocks.
* If a block is freed, and its buddy is also free, then this
* triggers coalescing into a block of larger size.
*
* -- wli
*/
static inline void __free_one_page(struct page *page,
struct zone *zone, unsigned int order)
{
unsigned long page_idx;
int order_size = 1 << order;
if (unlikely(PageCompound(page)))
destroy_compound_page(page, order);
page_idx = page_to_pfn(page) & ((1 << MAX_ORDER) - 1);
BUG_ON(page_idx & (order_size - 1));
BUG_ON(bad_range(zone, page));
zone->free_pages += order_size;
while (order < MAX_ORDER-1) {
unsigned long combined_idx;
struct free_area *area;
struct page *buddy;
buddy = __page_find_buddy(page, page_idx, order);
if (!page_is_buddy(buddy, order))
break; /* Move the buddy up one level. */
list_del(&buddy->lru);
area = zone->free_area + order;
area->nr_free--;
rmv_page_order(buddy);
combined_idx = __find_combined_index(page_idx, order);
page = page + (combined_idx - page_idx);
page_idx = combined_idx;
order++;
}
set_page_order(page, order);
list_add(&page->lru, &zone->free_area[order].free_list);
zone->free_area[order].nr_free++;
}
static inline int free_pages_check(struct page *page)
{
if (unlikely(page_mapcount(page) |
(page->mapping != NULL) |
(page_count(page) != 0) |
(page->flags & (
1 << PG_lru |
1 << PG_private |
1 << PG_locked |
1 << PG_active |
1 << PG_reclaim |
1 << PG_slab |
1 << PG_swapcache |
2005-10-30 01:16:12 +00:00
1 << PG_writeback |
1 << PG_reserved ))))
bad_page(page);
if (PageDirty(page))
__ClearPageDirty(page);
/*
* For now, we report if PG_reserved was found set, but do not
* clear it, and do not free the page. But we shall soon need
* to do more, for when the ZERO_PAGE count wraps negative.
*/
return PageReserved(page);
}
/*
* Frees a list of pages.
* Assumes all pages on list are in same zone, and of same order.
* count is the number of pages to free.
*
* If the zone was previously in an "all pages pinned" state then look to
* see if this freeing clears that state.
*
* And clear the zone's pages_scanned counter, to hold off the "all pages are
* pinned" detection logic.
*/
static void free_pages_bulk(struct zone *zone, int count,
struct list_head *list, int order)
{
spin_lock(&zone->lock);
zone->all_unreclaimable = 0;
zone->pages_scanned = 0;
while (count--) {
struct page *page;
BUG_ON(list_empty(list));
page = list_entry(list->prev, struct page, lru);
/* have to delete it as __free_one_page list manipulates */
list_del(&page->lru);
__free_one_page(page, zone, order);
}
spin_unlock(&zone->lock);
}
static void free_one_page(struct zone *zone, struct page *page, int order)
{
LIST_HEAD(list);
list_add(&page->lru, &list);
free_pages_bulk(zone, 1, &list, order);
}
static void __free_pages_ok(struct page *page, unsigned int order)
{
unsigned long flags;
int i;
int reserved = 0;
arch_free_page(page, order);
if (!PageHighMem(page))
mutex_debug_check_no_locks_freed(page_address(page),
PAGE_SIZE<<order);
for (i = 0 ; i < (1 << order) ; ++i)
reserved += free_pages_check(page + i);
if (reserved)
return;
kernel_map_pages(page, 1 << order, 0);
local_irq_save(flags);
__mod_page_state(pgfree, 1 << order);
free_one_page(page_zone(page), page, order);
local_irq_restore(flags);
}
/*
* permit the bootmem allocator to evade page validation on high-order frees
*/
void fastcall __init __free_pages_bootmem(struct page *page, unsigned int order)
{
if (order == 0) {
__ClearPageReserved(page);
set_page_count(page, 0);
set_page_refcounted(page);
__free_page(page);
} else {
int loop;
prefetchw(page);
for (loop = 0; loop < BITS_PER_LONG; loop++) {
struct page *p = &page[loop];
if (loop + 1 < BITS_PER_LONG)
prefetchw(p + 1);
__ClearPageReserved(p);
set_page_count(p, 0);
}
set_page_refcounted(page);
__free_pages(page, order);
}
}
/*
* The order of subdivision here is critical for the IO subsystem.
* Please do not alter this order without good reasons and regression
* testing. Specifically, as large blocks of memory are subdivided,
* the order in which smaller blocks are delivered depends on the order
* they're subdivided in this function. This is the primary factor
* influencing the order in which pages are delivered to the IO
* subsystem according to empirical testing, and this is also justified
* by considering the behavior of a buddy system containing a single
* large block of memory acted on by a series of small allocations.
* This behavior is a critical factor in sglist merging's success.
*
* -- wli
*/
static inline void expand(struct zone *zone, struct page *page,
int low, int high, struct free_area *area)
{
unsigned long size = 1 << high;
while (high > low) {
area--;
high--;
size >>= 1;
BUG_ON(bad_range(zone, &page[size]));
list_add(&page[size].lru, &area->free_list);
area->nr_free++;
set_page_order(&page[size], high);
}
}
/*
* This page is about to be returned from the page allocator
*/
static int prep_new_page(struct page *page, int order, gfp_t gfp_flags)
{
if (unlikely(page_mapcount(page) |
(page->mapping != NULL) |
(page_count(page) != 0) |
(page->flags & (
1 << PG_lru |
1 << PG_private |
1 << PG_locked |
1 << PG_active |
1 << PG_dirty |
1 << PG_reclaim |
1 << PG_slab |
1 << PG_swapcache |
2005-10-30 01:16:12 +00:00
1 << PG_writeback |
1 << PG_reserved ))))
bad_page(page);
/*
* For now, we report if PG_reserved was found set, but do not
* clear it, and do not allocate the page: as a safety net.
*/
if (PageReserved(page))
return 1;
page->flags &= ~(1 << PG_uptodate | 1 << PG_error |
1 << PG_referenced | 1 << PG_arch_1 |
1 << PG_checked | 1 << PG_mappedtodisk);
[PATCH] mm: split page table lock Christoph Lameter demonstrated very poor scalability on the SGI 512-way, with a many-threaded application which concurrently initializes different parts of a large anonymous area. This patch corrects that, by using a separate spinlock per page table page, to guard the page table entries in that page, instead of using the mm's single page_table_lock. (But even then, page_table_lock is still used to guard page table allocation, and anon_vma allocation.) In this implementation, the spinlock is tucked inside the struct page of the page table page: with a BUILD_BUG_ON in case it overflows - which it would in the case of 32-bit PA-RISC with spinlock debugging enabled. Splitting the lock is not quite for free: another cacheline access. Ideally, I suppose we would use split ptlock only for multi-threaded processes on multi-cpu machines; but deciding that dynamically would have its own costs. So for now enable it by config, at some number of cpus - since the Kconfig language doesn't support inequalities, let preprocessor compare that with NR_CPUS. But I don't think it's worth being user-configurable: for good testing of both split and unsplit configs, split now at 4 cpus, and perhaps change that to 8 later. There is a benefit even for singly threaded processes: kswapd can be attacking one part of the mm while another part is busy faulting. Signed-off-by: Hugh Dickins <hugh@veritas.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-10-30 01:16:40 +00:00
set_page_private(page, 0);
set_page_refcounted(page);
kernel_map_pages(page, 1 << order, 1);
if (gfp_flags & __GFP_ZERO)
prep_zero_page(page, order, gfp_flags);
if (order && (gfp_flags & __GFP_COMP))
prep_compound_page(page, order);
return 0;
}
/*
* Do the hard work of removing an element from the buddy allocator.
* Call me with the zone->lock already held.
*/
static struct page *__rmqueue(struct zone *zone, unsigned int order)
{
struct free_area * area;
unsigned int current_order;
struct page *page;
for (current_order = order; current_order < MAX_ORDER; ++current_order) {
area = zone->free_area + current_order;
if (list_empty(&area->free_list))
continue;
page = list_entry(area->free_list.next, struct page, lru);
list_del(&page->lru);
rmv_page_order(page);
area->nr_free--;
zone->free_pages -= 1UL << order;
expand(zone, page, order, current_order, area);
return page;
}
return NULL;
}
/*
* Obtain a specified number of elements from the buddy allocator, all under
* a single hold of the lock, for efficiency. Add them to the supplied list.
* Returns the number of new pages which were placed at *list.
*/
static int rmqueue_bulk(struct zone *zone, unsigned int order,
unsigned long count, struct list_head *list)
{
int i;
spin_lock(&zone->lock);
for (i = 0; i < count; ++i) {
struct page *page = __rmqueue(zone, order);
if (unlikely(page == NULL))
break;
list_add_tail(&page->lru, list);
}
spin_unlock(&zone->lock);
return i;
}
#ifdef CONFIG_NUMA
[PATCH] slab: Node rotor for freeing alien caches and remote per cpu pages. The cache reaper currently tries to free all alien caches and all remote per cpu pages in each pass of cache_reap. For a machines with large number of nodes (such as Altix) this may lead to sporadic delays of around ~10ms. Interrupts are disabled while reclaiming creating unacceptable delays. This patch changes that behavior by adding a per cpu reap_node variable. Instead of attempting to free all caches, we free only one alien cache and the per cpu pages from one remote node. That reduces the time spend in cache_reap. However, doing so will lengthen the time it takes to completely drain all remote per cpu pagesets and all alien caches. The time needed will grow with the number of nodes in the system. All caches are drained when they overflow their respective capacity. So the drawback here is only that a bit of memory may be wasted for awhile longer. Details: 1. Rename drain_remote_pages to drain_node_pages to allow the specification of the node to drain of pcp pages. 2. Add additional functions init_reap_node, next_reap_node for NUMA that manage a per cpu reap_node counter. 3. Add a reap_alien function that reaps only from the current reap_node. For us this seems to be a critical issue. Holdoffs of an average of ~7ms cause some HPC benchmarks to slow down significantly. F.e. NAS parallel slows down dramatically. NAS parallel has a 12-16 seconds runtime w/o rotor compared to 5.8 secs with the rotor patches. It gets down to 5.05 secs with the additional interrupt holdoff reductions. Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-10 01:33:54 +00:00
/*
* Called from the slab reaper to drain pagesets on a particular node that
* belong to the currently executing processor.
* Note that this function must be called with the thread pinned to
* a single processor.
[PATCH] slab: Node rotor for freeing alien caches and remote per cpu pages. The cache reaper currently tries to free all alien caches and all remote per cpu pages in each pass of cache_reap. For a machines with large number of nodes (such as Altix) this may lead to sporadic delays of around ~10ms. Interrupts are disabled while reclaiming creating unacceptable delays. This patch changes that behavior by adding a per cpu reap_node variable. Instead of attempting to free all caches, we free only one alien cache and the per cpu pages from one remote node. That reduces the time spend in cache_reap. However, doing so will lengthen the time it takes to completely drain all remote per cpu pagesets and all alien caches. The time needed will grow with the number of nodes in the system. All caches are drained when they overflow their respective capacity. So the drawback here is only that a bit of memory may be wasted for awhile longer. Details: 1. Rename drain_remote_pages to drain_node_pages to allow the specification of the node to drain of pcp pages. 2. Add additional functions init_reap_node, next_reap_node for NUMA that manage a per cpu reap_node counter. 3. Add a reap_alien function that reaps only from the current reap_node. For us this seems to be a critical issue. Holdoffs of an average of ~7ms cause some HPC benchmarks to slow down significantly. F.e. NAS parallel slows down dramatically. NAS parallel has a 12-16 seconds runtime w/o rotor compared to 5.8 secs with the rotor patches. It gets down to 5.05 secs with the additional interrupt holdoff reductions. Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-10 01:33:54 +00:00
*/
void drain_node_pages(int nodeid)
{
[PATCH] slab: Node rotor for freeing alien caches and remote per cpu pages. The cache reaper currently tries to free all alien caches and all remote per cpu pages in each pass of cache_reap. For a machines with large number of nodes (such as Altix) this may lead to sporadic delays of around ~10ms. Interrupts are disabled while reclaiming creating unacceptable delays. This patch changes that behavior by adding a per cpu reap_node variable. Instead of attempting to free all caches, we free only one alien cache and the per cpu pages from one remote node. That reduces the time spend in cache_reap. However, doing so will lengthen the time it takes to completely drain all remote per cpu pagesets and all alien caches. The time needed will grow with the number of nodes in the system. All caches are drained when they overflow their respective capacity. So the drawback here is only that a bit of memory may be wasted for awhile longer. Details: 1. Rename drain_remote_pages to drain_node_pages to allow the specification of the node to drain of pcp pages. 2. Add additional functions init_reap_node, next_reap_node for NUMA that manage a per cpu reap_node counter. 3. Add a reap_alien function that reaps only from the current reap_node. For us this seems to be a critical issue. Holdoffs of an average of ~7ms cause some HPC benchmarks to slow down significantly. F.e. NAS parallel slows down dramatically. NAS parallel has a 12-16 seconds runtime w/o rotor compared to 5.8 secs with the rotor patches. It gets down to 5.05 secs with the additional interrupt holdoff reductions. Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-10 01:33:54 +00:00
int i, z;
unsigned long flags;
[PATCH] slab: Node rotor for freeing alien caches and remote per cpu pages. The cache reaper currently tries to free all alien caches and all remote per cpu pages in each pass of cache_reap. For a machines with large number of nodes (such as Altix) this may lead to sporadic delays of around ~10ms. Interrupts are disabled while reclaiming creating unacceptable delays. This patch changes that behavior by adding a per cpu reap_node variable. Instead of attempting to free all caches, we free only one alien cache and the per cpu pages from one remote node. That reduces the time spend in cache_reap. However, doing so will lengthen the time it takes to completely drain all remote per cpu pagesets and all alien caches. The time needed will grow with the number of nodes in the system. All caches are drained when they overflow their respective capacity. So the drawback here is only that a bit of memory may be wasted for awhile longer. Details: 1. Rename drain_remote_pages to drain_node_pages to allow the specification of the node to drain of pcp pages. 2. Add additional functions init_reap_node, next_reap_node for NUMA that manage a per cpu reap_node counter. 3. Add a reap_alien function that reaps only from the current reap_node. For us this seems to be a critical issue. Holdoffs of an average of ~7ms cause some HPC benchmarks to slow down significantly. F.e. NAS parallel slows down dramatically. NAS parallel has a 12-16 seconds runtime w/o rotor compared to 5.8 secs with the rotor patches. It gets down to 5.05 secs with the additional interrupt holdoff reductions. Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-10 01:33:54 +00:00
for (z = 0; z < MAX_NR_ZONES; z++) {
struct zone *zone = NODE_DATA(nodeid)->node_zones + z;
struct per_cpu_pageset *pset;
pset = zone_pcp(zone, smp_processor_id());
for (i = 0; i < ARRAY_SIZE(pset->pcp); i++) {
struct per_cpu_pages *pcp;
pcp = &pset->pcp[i];
if (pcp->count) {
local_irq_save(flags);
free_pages_bulk(zone, pcp->count, &pcp->list, 0);
pcp->count = 0;
local_irq_restore(flags);
}
}
}
}
#endif
#if defined(CONFIG_PM) || defined(CONFIG_HOTPLUG_CPU)
static void __drain_pages(unsigned int cpu)
{
unsigned long flags;
struct zone *zone;
int i;
for_each_zone(zone) {
struct per_cpu_pageset *pset;
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
pset = zone_pcp(zone, cpu);
for (i = 0; i < ARRAY_SIZE(pset->pcp); i++) {
struct per_cpu_pages *pcp;
pcp = &pset->pcp[i];
local_irq_save(flags);
free_pages_bulk(zone, pcp->count, &pcp->list, 0);
pcp->count = 0;
local_irq_restore(flags);
}
}
}
#endif /* CONFIG_PM || CONFIG_HOTPLUG_CPU */
#ifdef CONFIG_PM
void mark_free_pages(struct zone *zone)
{
unsigned long zone_pfn, flags;
int order;
struct list_head *curr;
if (!zone->spanned_pages)
return;
spin_lock_irqsave(&zone->lock, flags);
for (zone_pfn = 0; zone_pfn < zone->spanned_pages; ++zone_pfn)
ClearPageNosaveFree(pfn_to_page(zone_pfn + zone->zone_start_pfn));
for (order = MAX_ORDER - 1; order >= 0; --order)
list_for_each(curr, &zone->free_area[order].free_list) {
unsigned long start_pfn, i;
start_pfn = page_to_pfn(list_entry(curr, struct page, lru));
for (i=0; i < (1<<order); i++)
SetPageNosaveFree(pfn_to_page(start_pfn+i));
}
spin_unlock_irqrestore(&zone->lock, flags);
}
/*
* Spill all of this CPU's per-cpu pages back into the buddy allocator.
*/
void drain_local_pages(void)
{
unsigned long flags;
local_irq_save(flags);
__drain_pages(smp_processor_id());
local_irq_restore(flags);
}
#endif /* CONFIG_PM */
static void zone_statistics(struct zonelist *zonelist, struct zone *z, int cpu)
{
#ifdef CONFIG_NUMA
pg_data_t *pg = z->zone_pgdat;
pg_data_t *orig = zonelist->zones[0]->zone_pgdat;
struct per_cpu_pageset *p;
p = zone_pcp(z, cpu);
if (pg == orig) {
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
p->numa_hit++;
} else {
p->numa_miss++;
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
zone_pcp(zonelist->zones[0], cpu)->numa_foreign++;
}
if (pg == NODE_DATA(numa_node_id()))
p->local_node++;
else
p->other_node++;
#endif
}
/*
* Free a 0-order page
*/
static void fastcall free_hot_cold_page(struct page *page, int cold)
{
struct zone *zone = page_zone(page);
struct per_cpu_pages *pcp;
unsigned long flags;
arch_free_page(page, 0);
if (PageAnon(page))
page->mapping = NULL;
if (free_pages_check(page))
return;
kernel_map_pages(page, 1, 0);
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
pcp = &zone_pcp(zone, get_cpu())->pcp[cold];
local_irq_save(flags);
__inc_page_state(pgfree);
list_add(&page->lru, &pcp->list);
pcp->count++;
if (pcp->count >= pcp->high) {
free_pages_bulk(zone, pcp->batch, &pcp->list, 0);
pcp->count -= pcp->batch;
}
local_irq_restore(flags);
put_cpu();
}
void fastcall free_hot_page(struct page *page)
{
free_hot_cold_page(page, 0);
}
void fastcall free_cold_page(struct page *page)
{
free_hot_cold_page(page, 1);
}
/*
* split_page takes a non-compound higher-order page, and splits it into
* n (1<<order) sub-pages: page[0..n]
* Each sub-page must be freed individually.
*
* Note: this is probably too low level an operation for use in drivers.
* Please consult with lkml before using this in your driver.
*/
void split_page(struct page *page, unsigned int order)
{
int i;
BUG_ON(PageCompound(page));
BUG_ON(!page_count(page));
for (i = 1; i < (1 << order); i++)
set_page_refcounted(page + i);
}
/*
* Really, prep_compound_page() should be called from __rmqueue_bulk(). But
* we cheat by calling it from here, in the order > 0 path. Saves a branch
* or two.
*/
static struct page *buffered_rmqueue(struct zonelist *zonelist,
struct zone *zone, int order, gfp_t gfp_flags)
{
unsigned long flags;
struct page *page;
int cold = !!(gfp_flags & __GFP_COLD);
int cpu;
again:
cpu = get_cpu();
if (likely(order == 0)) {
struct per_cpu_pages *pcp;
pcp = &zone_pcp(zone, cpu)->pcp[cold];
local_irq_save(flags);
if (!pcp->count) {
pcp->count += rmqueue_bulk(zone, 0,
pcp->batch, &pcp->list);
if (unlikely(!pcp->count))
goto failed;
}
page = list_entry(pcp->list.next, struct page, lru);
list_del(&page->lru);
pcp->count--;
} else {
spin_lock_irqsave(&zone->lock, flags);
page = __rmqueue(zone, order);
spin_unlock(&zone->lock);
if (!page)
goto failed;
}
__mod_page_state_zone(zone, pgalloc, 1 << order);
zone_statistics(zonelist, zone, cpu);
local_irq_restore(flags);
put_cpu();
BUG_ON(bad_range(zone, page));
if (prep_new_page(page, order, gfp_flags))
goto again;
return page;
failed:
local_irq_restore(flags);
put_cpu();
return NULL;
}
#define ALLOC_NO_WATERMARKS 0x01 /* don't check watermarks at all */
#define ALLOC_WMARK_MIN 0x02 /* use pages_min watermark */
#define ALLOC_WMARK_LOW 0x04 /* use pages_low watermark */
#define ALLOC_WMARK_HIGH 0x08 /* use pages_high watermark */
#define ALLOC_HARDER 0x10 /* try to alloc harder */
#define ALLOC_HIGH 0x20 /* __GFP_HIGH set */
#define ALLOC_CPUSET 0x40 /* check for correct cpuset */
/*
* Return 1 if free pages are above 'mark'. This takes into account the order
* of the allocation.
*/
int zone_watermark_ok(struct zone *z, int order, unsigned long mark,
int classzone_idx, int alloc_flags)
{
/* free_pages my go negative - that's OK */
long min = mark, free_pages = z->free_pages - (1 << order) + 1;
int o;
if (alloc_flags & ALLOC_HIGH)
min -= min / 2;
if (alloc_flags & ALLOC_HARDER)
min -= min / 4;
if (free_pages <= min + z->lowmem_reserve[classzone_idx])
return 0;
for (o = 0; o < order; o++) {
/* At the next order, this order's pages become unavailable */
free_pages -= z->free_area[o].nr_free << o;
/* Require fewer higher order pages to be free */
min >>= 1;
if (free_pages <= min)
return 0;
}
return 1;
}
/*
* get_page_from_freeliest goes through the zonelist trying to allocate
* a page.
*/
static struct page *
get_page_from_freelist(gfp_t gfp_mask, unsigned int order,
struct zonelist *zonelist, int alloc_flags)
[PATCH] VM: early zone reclaim This is the core of the (much simplified) early reclaim. The goal of this patch is to reclaim some easily-freed pages from a zone before falling back onto another zone. One of the major uses of this is NUMA machines. With the default allocator behavior the allocator would look for memory in another zone, which might be off-node, before trying to reclaim from the current zone. This adds a zone tuneable to enable early zone reclaim. It is selected on a per-zone basis and is turned on/off via syscall. Adding some extra throttling on the reclaim was also required (patch 4/4). Without the machine would grind to a crawl when doing a "make -j" kernel build. Even with this patch the System Time is higher on average, but it seems tolerable. Here are some numbers for kernbench runs on a 2-node, 4cpu, 8Gig RAM Altix in the "make -j" run: wall user sys %cpu ctx sw. sleeps ---- ---- --- ---- ------ ------ No patch 1009 1384 847 258 298170 504402 w/patch, no reclaim 880 1376 667 288 254064 396745 w/patch & reclaim 1079 1385 926 252 291625 548873 These numbers are the average of 2 runs of 3 "make -j" runs done right after system boot. Run-to-run variability for "make -j" is huge, so these numbers aren't terribly useful except to seee that with reclaim the benchmark still finishes in a reasonable amount of time. I also looked at the NUMA hit/miss stats for the "make -j" runs and the reclaim doesn't make any difference when the machine is thrashing away. Doing a "make -j8" on a single node that is filled with page cache pages takes 700 seconds with reclaim turned on and 735 seconds without reclaim (due to remote memory accesses). The simple zone_reclaim syscall program is at http://www.bork.org/~mort/sgi/zone_reclaim.c Signed-off-by: Martin Hicks <mort@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:41 +00:00
{
struct zone **z = zonelist->zones;
struct page *page = NULL;
int classzone_idx = zone_idx(*z);
/*
* Go through the zonelist once, looking for a zone with enough free.
* See also cpuset_zone_allowed() comment in kernel/cpuset.c.
*/
do {
if ((alloc_flags & ALLOC_CPUSET) &&
!cpuset_zone_allowed(*z, gfp_mask))
continue;
if (!(alloc_flags & ALLOC_NO_WATERMARKS)) {
unsigned long mark;
if (alloc_flags & ALLOC_WMARK_MIN)
mark = (*z)->pages_min;
else if (alloc_flags & ALLOC_WMARK_LOW)
mark = (*z)->pages_low;
else
mark = (*z)->pages_high;
if (!zone_watermark_ok(*z, order, mark,
classzone_idx, alloc_flags))
if (!zone_reclaim_mode ||
!zone_reclaim(*z, gfp_mask, order))
continue;
}
page = buffered_rmqueue(zonelist, *z, order, gfp_mask);
if (page) {
break;
}
} while (*(++z) != NULL);
return page;
[PATCH] VM: early zone reclaim This is the core of the (much simplified) early reclaim. The goal of this patch is to reclaim some easily-freed pages from a zone before falling back onto another zone. One of the major uses of this is NUMA machines. With the default allocator behavior the allocator would look for memory in another zone, which might be off-node, before trying to reclaim from the current zone. This adds a zone tuneable to enable early zone reclaim. It is selected on a per-zone basis and is turned on/off via syscall. Adding some extra throttling on the reclaim was also required (patch 4/4). Without the machine would grind to a crawl when doing a "make -j" kernel build. Even with this patch the System Time is higher on average, but it seems tolerable. Here are some numbers for kernbench runs on a 2-node, 4cpu, 8Gig RAM Altix in the "make -j" run: wall user sys %cpu ctx sw. sleeps ---- ---- --- ---- ------ ------ No patch 1009 1384 847 258 298170 504402 w/patch, no reclaim 880 1376 667 288 254064 396745 w/patch & reclaim 1079 1385 926 252 291625 548873 These numbers are the average of 2 runs of 3 "make -j" runs done right after system boot. Run-to-run variability for "make -j" is huge, so these numbers aren't terribly useful except to seee that with reclaim the benchmark still finishes in a reasonable amount of time. I also looked at the NUMA hit/miss stats for the "make -j" runs and the reclaim doesn't make any difference when the machine is thrashing away. Doing a "make -j8" on a single node that is filled with page cache pages takes 700 seconds with reclaim turned on and 735 seconds without reclaim (due to remote memory accesses). The simple zone_reclaim syscall program is at http://www.bork.org/~mort/sgi/zone_reclaim.c Signed-off-by: Martin Hicks <mort@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:41 +00:00
}
/*
* This is the 'heart' of the zoned buddy allocator.
*/
struct page * fastcall
__alloc_pages(gfp_t gfp_mask, unsigned int order,
struct zonelist *zonelist)
{
const gfp_t wait = gfp_mask & __GFP_WAIT;
struct zone **z;
struct page *page;
struct reclaim_state reclaim_state;
struct task_struct *p = current;
int do_retry;
int alloc_flags;
int did_some_progress;
might_sleep_if(wait);
restart:
z = zonelist->zones; /* the list of zones suitable for gfp_mask */
if (unlikely(*z == NULL)) {
/* Should this ever happen?? */
return NULL;
}
page = get_page_from_freelist(gfp_mask|__GFP_HARDWALL, order,
zonelist, ALLOC_WMARK_LOW|ALLOC_CPUSET);
if (page)
goto got_pg;
do {
wakeup_kswapd(*z, order);
} while (*(++z));
[PATCH] cpusets: formalize intermediate GFP_KERNEL containment This patch makes use of the previously underutilized cpuset flag 'mem_exclusive' to provide what amounts to another layer of memory placement resolution. With this patch, there are now the following four layers of memory placement available: 1) The whole system (interrupt and GFP_ATOMIC allocations can use this), 2) The nearest enclosing mem_exclusive cpuset (GFP_KERNEL allocations can use), 3) The current tasks cpuset (GFP_USER allocations constrained to here), and 4) Specific node placement, using mbind and set_mempolicy. These nest - each layer is a subset (same or within) of the previous. Layer (2) above is new, with this patch. The call used to check whether a zone (its node, actually) is in a cpuset (in its mems_allowed, actually) is extended to take a gfp_mask argument, and its logic is extended, in the case that __GFP_HARDWALL is not set in the flag bits, to look up the cpuset hierarchy for the nearest enclosing mem_exclusive cpuset, to determine if placement is allowed. The definition of GFP_USER, which used to be identical to GFP_KERNEL, is changed to also set the __GFP_HARDWALL bit, in the previous cpuset_gfp_hardwall_flag patch. GFP_ATOMIC and GFP_KERNEL allocations will stay within the current tasks cpuset, so long as any node therein is not too tight on memory, but will escape to the larger layer, if need be. The intended use is to allow something like a batch manager to handle several jobs, each job in its own cpuset, but using common kernel memory for caches and such. Swapper and oom_kill activity is also constrained to Layer (2). A task in or below one mem_exclusive cpuset should not cause swapping on nodes in another non-overlapping mem_exclusive cpuset, nor provoke oom_killing of a task in another such cpuset. Heavy use of kernel memory for i/o caching and such by one job should not impact the memory available to jobs in other non-overlapping mem_exclusive cpusets. This patch enables providing hardwall, inescapable cpusets for memory allocations of each job, while sharing kernel memory allocations between several jobs, in an enclosing mem_exclusive cpuset. Like Dinakar's patch earlier to enable administering sched domains using the cpu_exclusive flag, this patch also provides a useful meaning to a cpuset flag that had previously done nothing much useful other than restrict what cpuset configurations were allowed. Signed-off-by: Paul Jackson <pj@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-09-06 22:18:12 +00:00
/*
* OK, we're below the kswapd watermark and have kicked background
* reclaim. Now things get more complex, so set up alloc_flags according
* to how we want to proceed.
*
* The caller may dip into page reserves a bit more if the caller
* cannot run direct reclaim, or if the caller has realtime scheduling
* policy or is asking for __GFP_HIGH memory. GFP_ATOMIC requests will
* set both ALLOC_HARDER (!wait) and ALLOC_HIGH (__GFP_HIGH).
[PATCH] cpusets: formalize intermediate GFP_KERNEL containment This patch makes use of the previously underutilized cpuset flag 'mem_exclusive' to provide what amounts to another layer of memory placement resolution. With this patch, there are now the following four layers of memory placement available: 1) The whole system (interrupt and GFP_ATOMIC allocations can use this), 2) The nearest enclosing mem_exclusive cpuset (GFP_KERNEL allocations can use), 3) The current tasks cpuset (GFP_USER allocations constrained to here), and 4) Specific node placement, using mbind and set_mempolicy. These nest - each layer is a subset (same or within) of the previous. Layer (2) above is new, with this patch. The call used to check whether a zone (its node, actually) is in a cpuset (in its mems_allowed, actually) is extended to take a gfp_mask argument, and its logic is extended, in the case that __GFP_HARDWALL is not set in the flag bits, to look up the cpuset hierarchy for the nearest enclosing mem_exclusive cpuset, to determine if placement is allowed. The definition of GFP_USER, which used to be identical to GFP_KERNEL, is changed to also set the __GFP_HARDWALL bit, in the previous cpuset_gfp_hardwall_flag patch. GFP_ATOMIC and GFP_KERNEL allocations will stay within the current tasks cpuset, so long as any node therein is not too tight on memory, but will escape to the larger layer, if need be. The intended use is to allow something like a batch manager to handle several jobs, each job in its own cpuset, but using common kernel memory for caches and such. Swapper and oom_kill activity is also constrained to Layer (2). A task in or below one mem_exclusive cpuset should not cause swapping on nodes in another non-overlapping mem_exclusive cpuset, nor provoke oom_killing of a task in another such cpuset. Heavy use of kernel memory for i/o caching and such by one job should not impact the memory available to jobs in other non-overlapping mem_exclusive cpusets. This patch enables providing hardwall, inescapable cpusets for memory allocations of each job, while sharing kernel memory allocations between several jobs, in an enclosing mem_exclusive cpuset. Like Dinakar's patch earlier to enable administering sched domains using the cpu_exclusive flag, this patch also provides a useful meaning to a cpuset flag that had previously done nothing much useful other than restrict what cpuset configurations were allowed. Signed-off-by: Paul Jackson <pj@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-09-06 22:18:12 +00:00
*/
alloc_flags = ALLOC_WMARK_MIN;
if ((unlikely(rt_task(p)) && !in_interrupt()) || !wait)
alloc_flags |= ALLOC_HARDER;
if (gfp_mask & __GFP_HIGH)
alloc_flags |= ALLOC_HIGH;
alloc_flags |= ALLOC_CPUSET;
/*
* Go through the zonelist again. Let __GFP_HIGH and allocations
* coming from realtime tasks go deeper into reserves.
*
* This is the last chance, in general, before the goto nopage.
* Ignore cpuset if GFP_ATOMIC (!wait) rather than fail alloc.
[PATCH] cpusets: formalize intermediate GFP_KERNEL containment This patch makes use of the previously underutilized cpuset flag 'mem_exclusive' to provide what amounts to another layer of memory placement resolution. With this patch, there are now the following four layers of memory placement available: 1) The whole system (interrupt and GFP_ATOMIC allocations can use this), 2) The nearest enclosing mem_exclusive cpuset (GFP_KERNEL allocations can use), 3) The current tasks cpuset (GFP_USER allocations constrained to here), and 4) Specific node placement, using mbind and set_mempolicy. These nest - each layer is a subset (same or within) of the previous. Layer (2) above is new, with this patch. The call used to check whether a zone (its node, actually) is in a cpuset (in its mems_allowed, actually) is extended to take a gfp_mask argument, and its logic is extended, in the case that __GFP_HARDWALL is not set in the flag bits, to look up the cpuset hierarchy for the nearest enclosing mem_exclusive cpuset, to determine if placement is allowed. The definition of GFP_USER, which used to be identical to GFP_KERNEL, is changed to also set the __GFP_HARDWALL bit, in the previous cpuset_gfp_hardwall_flag patch. GFP_ATOMIC and GFP_KERNEL allocations will stay within the current tasks cpuset, so long as any node therein is not too tight on memory, but will escape to the larger layer, if need be. The intended use is to allow something like a batch manager to handle several jobs, each job in its own cpuset, but using common kernel memory for caches and such. Swapper and oom_kill activity is also constrained to Layer (2). A task in or below one mem_exclusive cpuset should not cause swapping on nodes in another non-overlapping mem_exclusive cpuset, nor provoke oom_killing of a task in another such cpuset. Heavy use of kernel memory for i/o caching and such by one job should not impact the memory available to jobs in other non-overlapping mem_exclusive cpusets. This patch enables providing hardwall, inescapable cpusets for memory allocations of each job, while sharing kernel memory allocations between several jobs, in an enclosing mem_exclusive cpuset. Like Dinakar's patch earlier to enable administering sched domains using the cpu_exclusive flag, this patch also provides a useful meaning to a cpuset flag that had previously done nothing much useful other than restrict what cpuset configurations were allowed. Signed-off-by: Paul Jackson <pj@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-09-06 22:18:12 +00:00
* See also cpuset_zone_allowed() comment in kernel/cpuset.c.
*/
page = get_page_from_freelist(gfp_mask, order, zonelist, alloc_flags);
if (page)
goto got_pg;
/* This allocation should allow future memory freeing. */
if (((p->flags & PF_MEMALLOC) || unlikely(test_thread_flag(TIF_MEMDIE)))
&& !in_interrupt()) {
if (!(gfp_mask & __GFP_NOMEMALLOC)) {
nofail_alloc:
/* go through the zonelist yet again, ignoring mins */
page = get_page_from_freelist(gfp_mask, order,
zonelist, ALLOC_NO_WATERMARKS);
if (page)
goto got_pg;
if (gfp_mask & __GFP_NOFAIL) {
blk_congestion_wait(WRITE, HZ/50);
goto nofail_alloc;
}
}
goto nopage;
}
/* Atomic allocations - we can't balance anything */
if (!wait)
goto nopage;
rebalance:
cond_resched();
/* We now go into synchronous reclaim */
[PATCH] cpuset: memory pressure meter Provide a simple per-cpuset metric of memory pressure, tracking the -rate- that the tasks in a cpuset call try_to_free_pages(), the synchronous (direct) memory reclaim code. This enables batch managers monitoring jobs running in dedicated cpusets to efficiently detect what level of memory pressure that job is causing. This is useful both on tightly managed systems running a wide mix of submitted jobs, which may choose to terminate or reprioritize jobs that are trying to use more memory than allowed on the nodes assigned them, and with tightly coupled, long running, massively parallel scientific computing jobs that will dramatically fail to meet required performance goals if they start to use more memory than allowed to them. This patch just provides a very economical way for the batch manager to monitor a cpuset for signs of memory pressure. It's up to the batch manager or other user code to decide what to do about it and take action. ==> Unless this feature is enabled by writing "1" to the special file /dev/cpuset/memory_pressure_enabled, the hook in the rebalance code of __alloc_pages() for this metric reduces to simply noticing that the cpuset_memory_pressure_enabled flag is zero. So only systems that enable this feature will compute the metric. Why a per-cpuset, running average: Because this meter is per-cpuset, rather than per-task or mm, the system load imposed by a batch scheduler monitoring this metric is sharply reduced on large systems, because a scan of the tasklist can be avoided on each set of queries. Because this meter is a running average, instead of an accumulating counter, a batch scheduler can detect memory pressure with a single read, instead of having to read and accumulate results for a period of time. Because this meter is per-cpuset rather than per-task or mm, the batch scheduler can obtain the key information, memory pressure in a cpuset, with a single read, rather than having to query and accumulate results over all the (dynamically changing) set of tasks in the cpuset. A per-cpuset simple digital filter (requires a spinlock and 3 words of data per-cpuset) is kept, and updated by any task attached to that cpuset, if it enters the synchronous (direct) page reclaim code. A per-cpuset file provides an integer number representing the recent (half-life of 10 seconds) rate of direct page reclaims caused by the tasks in the cpuset, in units of reclaims attempted per second, times 1000. Signed-off-by: Paul Jackson <pj@sgi.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-01-08 09:01:49 +00:00
cpuset_memory_pressure_bump();
p->flags |= PF_MEMALLOC;
reclaim_state.reclaimed_slab = 0;
p->reclaim_state = &reclaim_state;
did_some_progress = try_to_free_pages(zonelist->zones, gfp_mask);
p->reclaim_state = NULL;
p->flags &= ~PF_MEMALLOC;
cond_resched();
if (likely(did_some_progress)) {
page = get_page_from_freelist(gfp_mask, order,
zonelist, alloc_flags);
if (page)
goto got_pg;
} else if ((gfp_mask & __GFP_FS) && !(gfp_mask & __GFP_NORETRY)) {
/*
* Go through the zonelist yet one more time, keep
* very high watermark here, this is only to catch
* a parallel oom killing, we must fail if we're still
* under heavy pressure.
*/
page = get_page_from_freelist(gfp_mask|__GFP_HARDWALL, order,
zonelist, ALLOC_WMARK_HIGH|ALLOC_CPUSET);
if (page)
goto got_pg;
out_of_memory(zonelist, gfp_mask, order);
goto restart;
}
/*
* Don't let big-order allocations loop unless the caller explicitly
* requests that. Wait for some write requests to complete then retry.
*
* In this implementation, __GFP_REPEAT means __GFP_NOFAIL for order
* <= 3, but that may not be true in other implementations.
*/
do_retry = 0;
if (!(gfp_mask & __GFP_NORETRY)) {
if ((order <= 3) || (gfp_mask & __GFP_REPEAT))
do_retry = 1;
if (gfp_mask & __GFP_NOFAIL)
do_retry = 1;
}
if (do_retry) {
blk_congestion_wait(WRITE, HZ/50);
goto rebalance;
}
nopage:
if (!(gfp_mask & __GFP_NOWARN) && printk_ratelimit()) {
printk(KERN_WARNING "%s: page allocation failure."
" order:%d, mode:0x%x\n",
p->comm, order, gfp_mask);
dump_stack();
show_mem();
}
got_pg:
return page;
}
EXPORT_SYMBOL(__alloc_pages);
/*
* Common helper functions.
*/
fastcall unsigned long __get_free_pages(gfp_t gfp_mask, unsigned int order)
{
struct page * page;
page = alloc_pages(gfp_mask, order);
if (!page)
return 0;
return (unsigned long) page_address(page);
}
EXPORT_SYMBOL(__get_free_pages);
fastcall unsigned long get_zeroed_page(gfp_t gfp_mask)
{
struct page * page;
/*
* get_zeroed_page() returns a 32-bit address, which cannot represent
* a highmem page
*/
BUG_ON((gfp_mask & __GFP_HIGHMEM) != 0);
page = alloc_pages(gfp_mask | __GFP_ZERO, 0);
if (page)
return (unsigned long) page_address(page);
return 0;
}
EXPORT_SYMBOL(get_zeroed_page);
void __pagevec_free(struct pagevec *pvec)
{
int i = pagevec_count(pvec);
while (--i >= 0)
free_hot_cold_page(pvec->pages[i], pvec->cold);
}
fastcall void __free_pages(struct page *page, unsigned int order)
{
2005-10-30 01:16:12 +00:00
if (put_page_testzero(page)) {
if (order == 0)
free_hot_page(page);
else
__free_pages_ok(page, order);
}
}
EXPORT_SYMBOL(__free_pages);
fastcall void free_pages(unsigned long addr, unsigned int order)
{
if (addr != 0) {
BUG_ON(!virt_addr_valid((void *)addr));
__free_pages(virt_to_page((void *)addr), order);
}
}
EXPORT_SYMBOL(free_pages);
/*
* Total amount of free (allocatable) RAM:
*/
unsigned int nr_free_pages(void)
{
unsigned int sum = 0;
struct zone *zone;
for_each_zone(zone)
sum += zone->free_pages;
return sum;
}
EXPORT_SYMBOL(nr_free_pages);
#ifdef CONFIG_NUMA
unsigned int nr_free_pages_pgdat(pg_data_t *pgdat)
{
unsigned int i, sum = 0;
for (i = 0; i < MAX_NR_ZONES; i++)
sum += pgdat->node_zones[i].free_pages;
return sum;
}
#endif
static unsigned int nr_free_zone_pages(int offset)
{
/* Just pick one node, since fallback list is circular */
pg_data_t *pgdat = NODE_DATA(numa_node_id());
unsigned int sum = 0;
struct zonelist *zonelist = pgdat->node_zonelists + offset;
struct zone **zonep = zonelist->zones;
struct zone *zone;
for (zone = *zonep++; zone; zone = *zonep++) {
unsigned long size = zone->present_pages;
unsigned long high = zone->pages_high;
if (size > high)
sum += size - high;
}
return sum;
}
/*
* Amount of free RAM allocatable within ZONE_DMA and ZONE_NORMAL
*/
unsigned int nr_free_buffer_pages(void)
{
return nr_free_zone_pages(gfp_zone(GFP_USER));
}
/*
* Amount of free RAM allocatable within all zones
*/
unsigned int nr_free_pagecache_pages(void)
{
return nr_free_zone_pages(gfp_zone(GFP_HIGHUSER));
}
#ifdef CONFIG_HIGHMEM
unsigned int nr_free_highpages (void)
{
pg_data_t *pgdat;
unsigned int pages = 0;
for_each_pgdat(pgdat)
pages += pgdat->node_zones[ZONE_HIGHMEM].free_pages;
return pages;
}
#endif
#ifdef CONFIG_NUMA
static void show_node(struct zone *zone)
{
printk("Node %d ", zone->zone_pgdat->node_id);
}
#else
#define show_node(zone) do { } while (0)
#endif
/*
* Accumulate the page_state information across all CPUs.
* The result is unavoidably approximate - it can change
* during and after execution of this function.
*/
static DEFINE_PER_CPU(struct page_state, page_states) = {0};
atomic_t nr_pagecache = ATOMIC_INIT(0);
EXPORT_SYMBOL(nr_pagecache);
#ifdef CONFIG_SMP
DEFINE_PER_CPU(long, nr_pagecache_local) = 0;
#endif
static void __get_page_state(struct page_state *ret, int nr, cpumask_t *cpumask)
{
unsigned cpu;
memset(ret, 0, nr * sizeof(unsigned long));
cpus_and(*cpumask, *cpumask, cpu_online_map);
for_each_cpu_mask(cpu, *cpumask) {
unsigned long *in;
unsigned long *out;
unsigned off;
unsigned next_cpu;
in = (unsigned long *)&per_cpu(page_states, cpu);
next_cpu = next_cpu(cpu, *cpumask);
if (likely(next_cpu < NR_CPUS))
prefetch(&per_cpu(page_states, next_cpu));
out = (unsigned long *)ret;
for (off = 0; off < nr; off++)
*out++ += *in++;
}
}
void get_page_state_node(struct page_state *ret, int node)
{
int nr;
cpumask_t mask = node_to_cpumask(node);
nr = offsetof(struct page_state, GET_PAGE_STATE_LAST);
nr /= sizeof(unsigned long);
__get_page_state(ret, nr+1, &mask);
}
void get_page_state(struct page_state *ret)
{
int nr;
cpumask_t mask = CPU_MASK_ALL;
nr = offsetof(struct page_state, GET_PAGE_STATE_LAST);
nr /= sizeof(unsigned long);
__get_page_state(ret, nr + 1, &mask);
}
void get_full_page_state(struct page_state *ret)
{
cpumask_t mask = CPU_MASK_ALL;
__get_page_state(ret, sizeof(*ret) / sizeof(unsigned long), &mask);
}
unsigned long read_page_state_offset(unsigned long offset)
{
unsigned long ret = 0;
int cpu;
for_each_online_cpu(cpu) {
unsigned long in;
in = (unsigned long)&per_cpu(page_states, cpu) + offset;
ret += *((unsigned long *)in);
}
return ret;
}
void __mod_page_state_offset(unsigned long offset, unsigned long delta)
{
void *ptr;
ptr = &__get_cpu_var(page_states);
*(unsigned long *)(ptr + offset) += delta;
}
EXPORT_SYMBOL(__mod_page_state_offset);
void mod_page_state_offset(unsigned long offset, unsigned long delta)
{
unsigned long flags;
void *ptr;
local_irq_save(flags);
ptr = &__get_cpu_var(page_states);
*(unsigned long *)(ptr + offset) += delta;
local_irq_restore(flags);
}
EXPORT_SYMBOL(mod_page_state_offset);
void __get_zone_counts(unsigned long *active, unsigned long *inactive,
unsigned long *free, struct pglist_data *pgdat)
{
struct zone *zones = pgdat->node_zones;
int i;
*active = 0;
*inactive = 0;
*free = 0;
for (i = 0; i < MAX_NR_ZONES; i++) {
*active += zones[i].nr_active;
*inactive += zones[i].nr_inactive;
*free += zones[i].free_pages;
}
}
void get_zone_counts(unsigned long *active,
unsigned long *inactive, unsigned long *free)
{
struct pglist_data *pgdat;
*active = 0;
*inactive = 0;
*free = 0;
for_each_pgdat(pgdat) {
unsigned long l, m, n;
__get_zone_counts(&l, &m, &n, pgdat);
*active += l;
*inactive += m;
*free += n;
}
}
void si_meminfo(struct sysinfo *val)
{
val->totalram = totalram_pages;
val->sharedram = 0;
val->freeram = nr_free_pages();
val->bufferram = nr_blockdev_pages();
#ifdef CONFIG_HIGHMEM
val->totalhigh = totalhigh_pages;
val->freehigh = nr_free_highpages();
#else
val->totalhigh = 0;
val->freehigh = 0;
#endif
val->mem_unit = PAGE_SIZE;
}
EXPORT_SYMBOL(si_meminfo);
#ifdef CONFIG_NUMA
void si_meminfo_node(struct sysinfo *val, int nid)
{
pg_data_t *pgdat = NODE_DATA(nid);
val->totalram = pgdat->node_present_pages;
val->freeram = nr_free_pages_pgdat(pgdat);
val->totalhigh = pgdat->node_zones[ZONE_HIGHMEM].present_pages;
val->freehigh = pgdat->node_zones[ZONE_HIGHMEM].free_pages;
val->mem_unit = PAGE_SIZE;
}
#endif
#define K(x) ((x) << (PAGE_SHIFT-10))
/*
* Show free area list (used inside shift_scroll-lock stuff)
* We also calculate the percentage fragmentation. We do this by counting the
* memory on each free list with the exception of the first item on the list.
*/
void show_free_areas(void)
{
struct page_state ps;
int cpu, temperature;
unsigned long active;
unsigned long inactive;
unsigned long free;
struct zone *zone;
for_each_zone(zone) {
show_node(zone);
printk("%s per-cpu:", zone->name);
if (!populated_zone(zone)) {
printk(" empty\n");
continue;
} else
printk("\n");
for_each_online_cpu(cpu) {
struct per_cpu_pageset *pageset;
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
pageset = zone_pcp(zone, cpu);
for (temperature = 0; temperature < 2; temperature++)
printk("cpu %d %s: high %d, batch %d used:%d\n",
cpu,
temperature ? "cold" : "hot",
pageset->pcp[temperature].high,
pageset->pcp[temperature].batch,
pageset->pcp[temperature].count);
}
}
get_page_state(&ps);
get_zone_counts(&active, &inactive, &free);
printk("Free pages: %11ukB (%ukB HighMem)\n",
K(nr_free_pages()),
K(nr_free_highpages()));
printk("Active:%lu inactive:%lu dirty:%lu writeback:%lu "
"unstable:%lu free:%u slab:%lu mapped:%lu pagetables:%lu\n",
active,
inactive,
ps.nr_dirty,
ps.nr_writeback,
ps.nr_unstable,
nr_free_pages(),
ps.nr_slab,
ps.nr_mapped,
ps.nr_page_table_pages);
for_each_zone(zone) {
int i;
show_node(zone);
printk("%s"
" free:%lukB"
" min:%lukB"
" low:%lukB"
" high:%lukB"
" active:%lukB"
" inactive:%lukB"
" present:%lukB"
" pages_scanned:%lu"
" all_unreclaimable? %s"
"\n",
zone->name,
K(zone->free_pages),
K(zone->pages_min),
K(zone->pages_low),
K(zone->pages_high),
K(zone->nr_active),
K(zone->nr_inactive),
K(zone->present_pages),
zone->pages_scanned,
(zone->all_unreclaimable ? "yes" : "no")
);
printk("lowmem_reserve[]:");
for (i = 0; i < MAX_NR_ZONES; i++)
printk(" %lu", zone->lowmem_reserve[i]);
printk("\n");
}
for_each_zone(zone) {
unsigned long nr, flags, order, total = 0;
show_node(zone);
printk("%s: ", zone->name);
if (!populated_zone(zone)) {
printk("empty\n");
continue;
}
spin_lock_irqsave(&zone->lock, flags);
for (order = 0; order < MAX_ORDER; order++) {
nr = zone->free_area[order].nr_free;
total += nr << order;
printk("%lu*%lukB ", nr, K(1UL) << order);
}
spin_unlock_irqrestore(&zone->lock, flags);
printk("= %lukB\n", K(total));
}
show_swap_cache_info();
}
/*
* Builds allocation fallback zone lists.
*
* Add all populated zones of a node to the zonelist.
*/
static int __init build_zonelists_node(pg_data_t *pgdat,
struct zonelist *zonelist, int nr_zones, int zone_type)
{
struct zone *zone;
BUG_ON(zone_type > ZONE_HIGHMEM);
do {
zone = pgdat->node_zones + zone_type;
if (populated_zone(zone)) {
#ifndef CONFIG_HIGHMEM
BUG_ON(zone_type > ZONE_NORMAL);
#endif
zonelist->zones[nr_zones++] = zone;
check_highest_zone(zone_type);
}
zone_type--;
} while (zone_type >= 0);
return nr_zones;
}
static inline int highest_zone(int zone_bits)
{
int res = ZONE_NORMAL;
if (zone_bits & (__force int)__GFP_HIGHMEM)
res = ZONE_HIGHMEM;
[PATCH] x86_64: Add 4GB DMA32 zone Add a new 4GB GFP_DMA32 zone between the GFP_DMA and GFP_NORMAL zones. As a bit of historical background: when the x86-64 port was originally designed we had some discussion if we should use a 16MB DMA zone like i386 or a 4GB DMA zone like IA64 or both. Both was ruled out at this point because it was in early 2.4 when VM is still quite shakey and had bad troubles even dealing with one DMA zone. We settled on the 16MB DMA zone mainly because we worried about older soundcards and the floppy. But this has always caused problems since then because device drivers had trouble getting enough DMA able memory. These days the VM works much better and the wide use of NUMA has proven it can deal with many zones successfully. So this patch adds both zones. This helps drivers who need a lot of memory below 4GB because their hardware is not accessing more (graphic drivers - proprietary and free ones, video frame buffer drivers, sound drivers etc.). Previously they could only use IOMMU+16MB GFP_DMA, which was not enough memory. Another common problem is that hardware who has full memory addressing for >4GB misses it for some control structures in memory (like transmit rings or other metadata). They tended to allocate memory in the 16MB GFP_DMA or the IOMMU/swiotlb then using pci_alloc_consistent, but that can tie up a lot of precious 16MB GFPDMA/IOMMU/swiotlb memory (even on AMD systems the IOMMU tends to be quite small) especially if you have many devices. With the new zone pci_alloc_consistent can just put this stuff into memory below 4GB which works better. One argument was still if the zone should be 4GB or 2GB. The main motivation for 2GB would be an unnamed not so unpopular hardware raid controller (mostly found in older machines from a particular four letter company) who has a strange 2GB restriction in firmware. But that one works ok with swiotlb/IOMMU anyways, so it doesn't really need GFP_DMA32. I chose 4GB to be compatible with IA64 and because it seems to be the most common restriction. The new zone is so far added only for x86-64. For other architectures who don't set up this new zone nothing changes. Architectures can set a compatibility define in Kconfig CONFIG_DMA_IS_DMA32 that will define GFP_DMA32 as GFP_DMA. Otherwise it's a nop because on 32bit architectures it's normally not needed because GFP_NORMAL (=0) is DMA able enough. One problem is still that GFP_DMA means different things on different architectures. e.g. some drivers used to have #ifdef ia64 use GFP_DMA (trusting it to be 4GB) #elif __x86_64__ (use other hacks like the swiotlb because 16MB is not enough) ... . This was quite ugly and is now obsolete. These should be now converted to use GFP_DMA32 unconditionally. I haven't done this yet. Or best only use pci_alloc_consistent/dma_alloc_coherent which will use GFP_DMA32 transparently. Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-11-05 16:25:53 +00:00
if (zone_bits & (__force int)__GFP_DMA32)
res = ZONE_DMA32;
if (zone_bits & (__force int)__GFP_DMA)
res = ZONE_DMA;
return res;
}
#ifdef CONFIG_NUMA
#define MAX_NODE_LOAD (num_online_nodes())
static int __initdata node_load[MAX_NUMNODES];
/**
[PATCH] DocBook: changes and extensions to the kernel documentation I have recompiled Linux kernel 2.6.11.5 documentation for me and our university students again. The documentation could be extended for more sources which are equipped by structured comments for recent 2.6 kernels. I have tried to proceed with that task. I have done that more times from 2.6.0 time and it gets boring to do same changes again and again. Linux kernel compiles after changes for i386 and ARM targets. I have added references to some more files into kernel-api book, I have added some section names as well. So please, check that changes do not break something and that categories are not too much skewed. I have changed kernel-doc to accept "fastcall" and "asmlinkage" words reserved by kernel convention. Most of the other changes are modifications in the comments to make kernel-doc happy, accept some parameters description and do not bail out on errors. Changed <pid> to @pid in the description, moved some #ifdef before comments to correct function to comments bindings, etc. You can see result of the modified documentation build at http://cmp.felk.cvut.cz/~pisa/linux/lkdb-2.6.11.tar.gz Some more sources are ready to be included into kernel-doc generated documentation. Sources has been added into kernel-api for now. Some more section names added and probably some more chaos introduced as result of quick cleanup work. Signed-off-by: Pavel Pisa <pisa@cmp.felk.cvut.cz> Signed-off-by: Martin Waitz <tali@admingilde.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-05-01 15:59:25 +00:00
* find_next_best_node - find the next node that should appear in a given node's fallback list
* @node: node whose fallback list we're appending
* @used_node_mask: nodemask_t of already used nodes
*
* We use a number of factors to determine which is the next node that should
* appear on a given node's fallback list. The node should not have appeared
* already in @node's fallback list, and it should be the next closest node
* according to the distance array (which contains arbitrary distance values
* from each node to each node in the system), and should also prefer nodes
* with no CPUs, since presumably they'll have very little allocation pressure
* on them otherwise.
* It returns -1 if no node is found.
*/
static int __init find_next_best_node(int node, nodemask_t *used_node_mask)
{
int n, val;
int min_val = INT_MAX;
int best_node = -1;
/* Use the local node if we haven't already */
if (!node_isset(node, *used_node_mask)) {
node_set(node, *used_node_mask);
return node;
}
for_each_online_node(n) {
cpumask_t tmp;
/* Don't want a node to appear more than once */
if (node_isset(n, *used_node_mask))
continue;
/* Use the distance array to find the distance */
val = node_distance(node, n);
/* Penalize nodes under us ("prefer the next node") */
val += (n < node);
/* Give preference to headless and unused nodes */
tmp = node_to_cpumask(n);
if (!cpus_empty(tmp))
val += PENALTY_FOR_NODE_WITH_CPUS;
/* Slight preference for less loaded node */
val *= (MAX_NODE_LOAD*MAX_NUMNODES);
val += node_load[n];
if (val < min_val) {
min_val = val;
best_node = n;
}
}
if (best_node >= 0)
node_set(best_node, *used_node_mask);
return best_node;
}
static void __init build_zonelists(pg_data_t *pgdat)
{
int i, j, k, node, local_node;
int prev_node, load;
struct zonelist *zonelist;
nodemask_t used_mask;
/* initialize zonelists */
for (i = 0; i < GFP_ZONETYPES; i++) {
zonelist = pgdat->node_zonelists + i;
zonelist->zones[0] = NULL;
}
/* NUMA-aware ordering of nodes */
local_node = pgdat->node_id;
load = num_online_nodes();
prev_node = local_node;
nodes_clear(used_mask);
while ((node = find_next_best_node(local_node, &used_mask)) >= 0) {
int distance = node_distance(local_node, node);
/*
* If another node is sufficiently far away then it is better
* to reclaim pages in a zone before going off node.
*/
if (distance > RECLAIM_DISTANCE)
zone_reclaim_mode = 1;
/*
* We don't want to pressure a particular node.
* So adding penalty to the first node in same
* distance group to make it round-robin.
*/
if (distance != node_distance(local_node, prev_node))
node_load[node] += load;
prev_node = node;
load--;
for (i = 0; i < GFP_ZONETYPES; i++) {
zonelist = pgdat->node_zonelists + i;
for (j = 0; zonelist->zones[j] != NULL; j++);
k = highest_zone(i);
j = build_zonelists_node(NODE_DATA(node), zonelist, j, k);
zonelist->zones[j] = NULL;
}
}
}
#else /* CONFIG_NUMA */
static void __init build_zonelists(pg_data_t *pgdat)
{
int i, j, k, node, local_node;
local_node = pgdat->node_id;
for (i = 0; i < GFP_ZONETYPES; i++) {
struct zonelist *zonelist;
zonelist = pgdat->node_zonelists + i;
j = 0;
k = highest_zone(i);
j = build_zonelists_node(pgdat, zonelist, j, k);
/*
* Now we build the zonelist so that it contains the zones
* of all the other nodes.
* We don't want to pressure a particular node, so when
* building the zones for node N, we make sure that the
* zones coming right after the local ones are those from
* node N+1 (modulo N)
*/
for (node = local_node + 1; node < MAX_NUMNODES; node++) {
if (!node_online(node))
continue;
j = build_zonelists_node(NODE_DATA(node), zonelist, j, k);
}
for (node = 0; node < local_node; node++) {
if (!node_online(node))
continue;
j = build_zonelists_node(NODE_DATA(node), zonelist, j, k);
}
zonelist->zones[j] = NULL;
}
}
#endif /* CONFIG_NUMA */
void __init build_all_zonelists(void)
{
int i;
for_each_online_node(i)
build_zonelists(NODE_DATA(i));
printk("Built %i zonelists\n", num_online_nodes());
cpuset_init_current_mems_allowed();
}
/*
* Helper functions to size the waitqueue hash table.
* Essentially these want to choose hash table sizes sufficiently
* large so that collisions trying to wait on pages are rare.
* But in fact, the number of active page waitqueues on typical
* systems is ridiculously low, less than 200. So this is even
* conservative, even though it seems large.
*
* The constant PAGES_PER_WAITQUEUE specifies the ratio of pages to
* waitqueues, i.e. the size of the waitq table given the number of pages.
*/
#define PAGES_PER_WAITQUEUE 256
static inline unsigned long wait_table_size(unsigned long pages)
{
unsigned long size = 1;
pages /= PAGES_PER_WAITQUEUE;
while (size < pages)
size <<= 1;
/*
* Once we have dozens or even hundreds of threads sleeping
* on IO we've got bigger problems than wait queue collision.
* Limit the size of the wait table to a reasonable size.
*/
size = min(size, 4096UL);
return max(size, 4UL);
}
/*
* This is an integer logarithm so that shifts can be used later
* to extract the more random high bits from the multiplicative
* hash function before the remainder is taken.
*/
static inline unsigned long wait_table_bits(unsigned long size)
{
return ffz(~size);
}
#define LONG_ALIGN(x) (((x)+(sizeof(long))-1)&~((sizeof(long))-1))
static void __init calculate_zone_totalpages(struct pglist_data *pgdat,
unsigned long *zones_size, unsigned long *zholes_size)
{
unsigned long realtotalpages, totalpages = 0;
int i;
for (i = 0; i < MAX_NR_ZONES; i++)
totalpages += zones_size[i];
pgdat->node_spanned_pages = totalpages;
realtotalpages = totalpages;
if (zholes_size)
for (i = 0; i < MAX_NR_ZONES; i++)
realtotalpages -= zholes_size[i];
pgdat->node_present_pages = realtotalpages;
printk(KERN_DEBUG "On node %d totalpages: %lu\n", pgdat->node_id, realtotalpages);
}
/*
* Initially all pages are reserved - free ones are freed
* up by free_all_bootmem() once the early boot process is
* done. Non-atomic initialization, single-pass.
*/
void __meminit memmap_init_zone(unsigned long size, int nid, unsigned long zone,
unsigned long start_pfn)
{
struct page *page;
[PATCH] sparsemem hotplug base Make sparse's initalization be accessible at runtime. This allows sparse mappings to be created after boot in a hotplug situation. This patch is separated from the previous one just to give an indication how much of the sparse infrastructure is *just* for hotplug memory. The section_mem_map doesn't really store a pointer. It stores something that is convenient to do some math against to get a pointer. It isn't valid to just do *section_mem_map, so I don't think it should be stored as a pointer. There are a couple of things I'd like to store about a section. First of all, the fact that it is !NULL does not mean that it is present. There could be such a combination where section_mem_map *is* NULL, but the math gets you properly to a real mem_map. So, I don't think that check is safe. Since we're storing 32-bit-aligned structures, we have a few bits in the bottom of the pointer to play with. Use one bit to encode whether there's really a mem_map there, and the other one to tell whether there's a valid section there. We need to distinguish between the two because sometimes there's a gap between when a section is discovered to be present and when we can get the mem_map for it. Signed-off-by: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Jack Steiner <steiner@sgi.com> Signed-off-by: Bob Picco <bob.picco@hp.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 07:08:00 +00:00
unsigned long end_pfn = start_pfn + size;
unsigned long pfn;
for (pfn = start_pfn; pfn < end_pfn; pfn++) {
[PATCH] sparsemem memory model Sparsemem abstracts the use of discontiguous mem_maps[]. This kind of mem_map[] is needed by discontiguous memory machines (like in the old CONFIG_DISCONTIGMEM case) as well as memory hotplug systems. Sparsemem replaces DISCONTIGMEM when enabled, and it is hoped that it can eventually become a complete replacement. A significant advantage over DISCONTIGMEM is that it's completely separated from CONFIG_NUMA. When producing this patch, it became apparent in that NUMA and DISCONTIG are often confused. Another advantage is that sparse doesn't require each NUMA node's ranges to be contiguous. It can handle overlapping ranges between nodes with no problems, where DISCONTIGMEM currently throws away that memory. Sparsemem uses an array to provide different pfn_to_page() translations for each SECTION_SIZE area of physical memory. This is what allows the mem_map[] to be chopped up. In order to do quick pfn_to_page() operations, the section number of the page is encoded in page->flags. Part of the sparsemem infrastructure enables sharing of these bits more dynamically (at compile-time) between the page_zone() and sparsemem operations. However, on 32-bit architectures, the number of bits is quite limited, and may require growing the size of the page->flags type in certain conditions. Several things might force this to occur: a decrease in the SECTION_SIZE (if you want to hotplug smaller areas of memory), an increase in the physical address space, or an increase in the number of used page->flags. One thing to note is that, once sparsemem is present, the NUMA node information no longer needs to be stored in the page->flags. It might provide speed increases on certain platforms and will be stored there if there is room. But, if out of room, an alternate (theoretically slower) mechanism is used. This patch introduces CONFIG_FLATMEM. It is used in almost all cases where there used to be an #ifndef DISCONTIG, because SPARSEMEM and DISCONTIGMEM often have to compile out the same areas of code. Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Martin Bligh <mbligh@aracnet.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com> Signed-off-by: Bob Picco <bob.picco@hp.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 07:07:54 +00:00
if (!early_pfn_valid(pfn))
continue;
page = pfn_to_page(pfn);
set_page_links(page, zone, nid, pfn);
init_page_count(page);
reset_page_mapcount(page);
SetPageReserved(page);
INIT_LIST_HEAD(&page->lru);
#ifdef WANT_PAGE_VIRTUAL
/* The shift won't overflow because ZONE_NORMAL is below 4G. */
if (!is_highmem_idx(zone))
set_page_address(page, __va(pfn << PAGE_SHIFT));
#endif
}
}
void zone_init_free_lists(struct pglist_data *pgdat, struct zone *zone,
unsigned long size)
{
int order;
for (order = 0; order < MAX_ORDER ; order++) {
INIT_LIST_HEAD(&zone->free_area[order].free_list);
zone->free_area[order].nr_free = 0;
}
}
[PATCH] sparsemem memory model Sparsemem abstracts the use of discontiguous mem_maps[]. This kind of mem_map[] is needed by discontiguous memory machines (like in the old CONFIG_DISCONTIGMEM case) as well as memory hotplug systems. Sparsemem replaces DISCONTIGMEM when enabled, and it is hoped that it can eventually become a complete replacement. A significant advantage over DISCONTIGMEM is that it's completely separated from CONFIG_NUMA. When producing this patch, it became apparent in that NUMA and DISCONTIG are often confused. Another advantage is that sparse doesn't require each NUMA node's ranges to be contiguous. It can handle overlapping ranges between nodes with no problems, where DISCONTIGMEM currently throws away that memory. Sparsemem uses an array to provide different pfn_to_page() translations for each SECTION_SIZE area of physical memory. This is what allows the mem_map[] to be chopped up. In order to do quick pfn_to_page() operations, the section number of the page is encoded in page->flags. Part of the sparsemem infrastructure enables sharing of these bits more dynamically (at compile-time) between the page_zone() and sparsemem operations. However, on 32-bit architectures, the number of bits is quite limited, and may require growing the size of the page->flags type in certain conditions. Several things might force this to occur: a decrease in the SECTION_SIZE (if you want to hotplug smaller areas of memory), an increase in the physical address space, or an increase in the number of used page->flags. One thing to note is that, once sparsemem is present, the NUMA node information no longer needs to be stored in the page->flags. It might provide speed increases on certain platforms and will be stored there if there is room. But, if out of room, an alternate (theoretically slower) mechanism is used. This patch introduces CONFIG_FLATMEM. It is used in almost all cases where there used to be an #ifndef DISCONTIG, because SPARSEMEM and DISCONTIGMEM often have to compile out the same areas of code. Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Martin Bligh <mbligh@aracnet.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com> Signed-off-by: Bob Picco <bob.picco@hp.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 07:07:54 +00:00
#define ZONETABLE_INDEX(x, zone_nr) ((x << ZONES_SHIFT) | zone_nr)
void zonetable_add(struct zone *zone, int nid, int zid, unsigned long pfn,
unsigned long size)
{
unsigned long snum = pfn_to_section_nr(pfn);
unsigned long end = pfn_to_section_nr(pfn + size);
if (FLAGS_HAS_NODE)
zone_table[ZONETABLE_INDEX(nid, zid)] = zone;
else
for (; snum <= end; snum++)
zone_table[ZONETABLE_INDEX(snum, zid)] = zone;
}
#ifndef __HAVE_ARCH_MEMMAP_INIT
#define memmap_init(size, nid, zone, start_pfn) \
memmap_init_zone((size), (nid), (zone), (start_pfn))
#endif
static int __cpuinit zone_batchsize(struct zone *zone)
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
{
int batch;
/*
* The per-cpu-pages pools are set to around 1000th of the
* size of the zone. But no more than 1/2 of a meg.
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
*
* OK, so we don't know how big the cache is. So guess.
*/
batch = zone->present_pages / 1024;
if (batch * PAGE_SIZE > 512 * 1024)
batch = (512 * 1024) / PAGE_SIZE;
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
batch /= 4; /* We effectively *= 4 below */
if (batch < 1)
batch = 1;
/*
* Clamp the batch to a 2^n - 1 value. Having a power
* of 2 value was found to be more likely to have
* suboptimal cache aliasing properties in some cases.
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
*
* For example if 2 tasks are alternately allocating
* batches of pages, one task can end up with a lot
* of pages of one half of the possible page colors
* and the other with pages of the other colors.
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
*/
batch = (1 << (fls(batch + batch/2)-1)) - 1;
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
return batch;
}
inline void setup_pageset(struct per_cpu_pageset *p, unsigned long batch)
{
struct per_cpu_pages *pcp;
memset(p, 0, sizeof(*p));
pcp = &p->pcp[0]; /* hot */
pcp->count = 0;
pcp->high = 6 * batch;
pcp->batch = max(1UL, 1 * batch);
INIT_LIST_HEAD(&pcp->list);
pcp = &p->pcp[1]; /* cold*/
pcp->count = 0;
pcp->high = 2 * batch;
pcp->batch = max(1UL, batch/2);
INIT_LIST_HEAD(&pcp->list);
}
/*
* setup_pagelist_highmark() sets the high water mark for hot per_cpu_pagelist
* to the value high for the pageset p.
*/
static void setup_pagelist_highmark(struct per_cpu_pageset *p,
unsigned long high)
{
struct per_cpu_pages *pcp;
pcp = &p->pcp[0]; /* hot list */
pcp->high = high;
pcp->batch = max(1UL, high/4);
if ((high/4) > (PAGE_SHIFT * 8))
pcp->batch = PAGE_SHIFT * 8;
}
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
#ifdef CONFIG_NUMA
/*
* Boot pageset table. One per cpu which is going to be used for all
* zones and all nodes. The parameters will be set in such a way
* that an item put on a list will immediately be handed over to
* the buddy list. This is safe since pageset manipulation is done
* with interrupts disabled.
*
* Some NUMA counter updates may also be caught by the boot pagesets.
*
* The boot_pagesets must be kept even after bootup is complete for
* unused processors and/or zones. They do play a role for bootstrapping
* hotplugged processors.
*
* zoneinfo_show() and maybe other functions do
* not check if the processor is online before following the pageset pointer.
* Other parts of the kernel may not check if the zone is available.
*/
static struct per_cpu_pageset boot_pageset[NR_CPUS];
/*
* Dynamically allocate memory for the
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
* per cpu pageset array in struct zone.
*/
static int __cpuinit process_zones(int cpu)
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
{
struct zone *zone, *dzone;
for_each_zone(zone) {
zone_pcp(zone, cpu) = kmalloc_node(sizeof(struct per_cpu_pageset),
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
GFP_KERNEL, cpu_to_node(cpu));
if (!zone_pcp(zone, cpu))
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
goto bad;
setup_pageset(zone_pcp(zone, cpu), zone_batchsize(zone));
if (percpu_pagelist_fraction)
setup_pagelist_highmark(zone_pcp(zone, cpu),
(zone->present_pages / percpu_pagelist_fraction));
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
}
return 0;
bad:
for_each_zone(dzone) {
if (dzone == zone)
break;
kfree(zone_pcp(dzone, cpu));
zone_pcp(dzone, cpu) = NULL;
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
}
return -ENOMEM;
}
static inline void free_zone_pagesets(int cpu)
{
struct zone *zone;
for_each_zone(zone) {
struct per_cpu_pageset *pset = zone_pcp(zone, cpu);
zone_pcp(zone, cpu) = NULL;
kfree(pset);
}
}
static int __cpuinit pageset_cpuup_callback(struct notifier_block *nfb,
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
unsigned long action,
void *hcpu)
{
int cpu = (long)hcpu;
int ret = NOTIFY_OK;
switch (action) {
case CPU_UP_PREPARE:
if (process_zones(cpu))
ret = NOTIFY_BAD;
break;
case CPU_UP_CANCELED:
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
case CPU_DEAD:
free_zone_pagesets(cpu);
break;
default:
break;
}
return ret;
}
static struct notifier_block pageset_notifier =
{ &pageset_cpuup_callback, NULL, 0 };
void __init setup_per_cpu_pageset(void)
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
{
int err;
/* Initialize per_cpu_pageset for cpu 0.
* A cpuup callback will do this for every cpu
* as it comes online
*/
err = process_zones(smp_processor_id());
BUG_ON(err);
register_cpu_notifier(&pageset_notifier);
}
#endif
static __meminit
void zone_wait_table_init(struct zone *zone, unsigned long zone_size_pages)
{
int i;
struct pglist_data *pgdat = zone->zone_pgdat;
/*
* The per-page waitqueue mechanism uses hashed waitqueues
* per zone.
*/
zone->wait_table_size = wait_table_size(zone_size_pages);
zone->wait_table_bits = wait_table_bits(zone->wait_table_size);
zone->wait_table = (wait_queue_head_t *)
alloc_bootmem_node(pgdat, zone->wait_table_size
* sizeof(wait_queue_head_t));
for(i = 0; i < zone->wait_table_size; ++i)
init_waitqueue_head(zone->wait_table + i);
}
static __meminit void zone_pcp_init(struct zone *zone)
{
int cpu;
unsigned long batch = zone_batchsize(zone);
for (cpu = 0; cpu < NR_CPUS; cpu++) {
#ifdef CONFIG_NUMA
/* Early boot. Slab allocator not functional yet */
zone_pcp(zone, cpu) = &boot_pageset[cpu];
setup_pageset(&boot_pageset[cpu],0);
#else
setup_pageset(zone_pcp(zone,cpu), batch);
#endif
}
printk(KERN_DEBUG " %s zone: %lu pages, LIFO batch:%lu\n",
zone->name, zone->present_pages, batch);
}
static __meminit void init_currently_empty_zone(struct zone *zone,
unsigned long zone_start_pfn, unsigned long size)
{
struct pglist_data *pgdat = zone->zone_pgdat;
zone_wait_table_init(zone, size);
pgdat->nr_zones = zone_idx(zone) + 1;
zone->zone_mem_map = pfn_to_page(zone_start_pfn);
zone->zone_start_pfn = zone_start_pfn;
memmap_init(size, pgdat->node_id, zone_idx(zone), zone_start_pfn);
zone_init_free_lists(pgdat, zone, zone->spanned_pages);
}
/*
* Set up the zone data structures:
* - mark all pages reserved
* - mark all memory queues empty
* - clear the memory bitmaps
*/
static void __init free_area_init_core(struct pglist_data *pgdat,
unsigned long *zones_size, unsigned long *zholes_size)
{
unsigned long j;
int nid = pgdat->node_id;
unsigned long zone_start_pfn = pgdat->node_start_pfn;
pgdat_resize_init(pgdat);
pgdat->nr_zones = 0;
init_waitqueue_head(&pgdat->kswapd_wait);
pgdat->kswapd_max_order = 0;
for (j = 0; j < MAX_NR_ZONES; j++) {
struct zone *zone = pgdat->node_zones + j;
unsigned long size, realsize;
realsize = size = zones_size[j];
if (zholes_size)
realsize -= zholes_size[j];
[PATCH] x86_64: Add 4GB DMA32 zone Add a new 4GB GFP_DMA32 zone between the GFP_DMA and GFP_NORMAL zones. As a bit of historical background: when the x86-64 port was originally designed we had some discussion if we should use a 16MB DMA zone like i386 or a 4GB DMA zone like IA64 or both. Both was ruled out at this point because it was in early 2.4 when VM is still quite shakey and had bad troubles even dealing with one DMA zone. We settled on the 16MB DMA zone mainly because we worried about older soundcards and the floppy. But this has always caused problems since then because device drivers had trouble getting enough DMA able memory. These days the VM works much better and the wide use of NUMA has proven it can deal with many zones successfully. So this patch adds both zones. This helps drivers who need a lot of memory below 4GB because their hardware is not accessing more (graphic drivers - proprietary and free ones, video frame buffer drivers, sound drivers etc.). Previously they could only use IOMMU+16MB GFP_DMA, which was not enough memory. Another common problem is that hardware who has full memory addressing for >4GB misses it for some control structures in memory (like transmit rings or other metadata). They tended to allocate memory in the 16MB GFP_DMA or the IOMMU/swiotlb then using pci_alloc_consistent, but that can tie up a lot of precious 16MB GFPDMA/IOMMU/swiotlb memory (even on AMD systems the IOMMU tends to be quite small) especially if you have many devices. With the new zone pci_alloc_consistent can just put this stuff into memory below 4GB which works better. One argument was still if the zone should be 4GB or 2GB. The main motivation for 2GB would be an unnamed not so unpopular hardware raid controller (mostly found in older machines from a particular four letter company) who has a strange 2GB restriction in firmware. But that one works ok with swiotlb/IOMMU anyways, so it doesn't really need GFP_DMA32. I chose 4GB to be compatible with IA64 and because it seems to be the most common restriction. The new zone is so far added only for x86-64. For other architectures who don't set up this new zone nothing changes. Architectures can set a compatibility define in Kconfig CONFIG_DMA_IS_DMA32 that will define GFP_DMA32 as GFP_DMA. Otherwise it's a nop because on 32bit architectures it's normally not needed because GFP_NORMAL (=0) is DMA able enough. One problem is still that GFP_DMA means different things on different architectures. e.g. some drivers used to have #ifdef ia64 use GFP_DMA (trusting it to be 4GB) #elif __x86_64__ (use other hacks like the swiotlb because 16MB is not enough) ... . This was quite ugly and is now obsolete. These should be now converted to use GFP_DMA32 unconditionally. I haven't done this yet. Or best only use pci_alloc_consistent/dma_alloc_coherent which will use GFP_DMA32 transparently. Signed-off-by: Andi Kleen <ak@suse.de> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-11-05 16:25:53 +00:00
if (j < ZONE_HIGHMEM)
nr_kernel_pages += realsize;
nr_all_pages += realsize;
zone->spanned_pages = size;
zone->present_pages = realsize;
zone->name = zone_names[j];
spin_lock_init(&zone->lock);
spin_lock_init(&zone->lru_lock);
zone_seqlock_init(zone);
zone->zone_pgdat = pgdat;
zone->free_pages = 0;
zone->temp_priority = zone->prev_priority = DEF_PRIORITY;
zone_pcp_init(zone);
INIT_LIST_HEAD(&zone->active_list);
INIT_LIST_HEAD(&zone->inactive_list);
zone->nr_scan_active = 0;
zone->nr_scan_inactive = 0;
zone->nr_active = 0;
zone->nr_inactive = 0;
atomic_set(&zone->reclaim_in_progress, 0);
if (!size)
continue;
[PATCH] sparsemem memory model Sparsemem abstracts the use of discontiguous mem_maps[]. This kind of mem_map[] is needed by discontiguous memory machines (like in the old CONFIG_DISCONTIGMEM case) as well as memory hotplug systems. Sparsemem replaces DISCONTIGMEM when enabled, and it is hoped that it can eventually become a complete replacement. A significant advantage over DISCONTIGMEM is that it's completely separated from CONFIG_NUMA. When producing this patch, it became apparent in that NUMA and DISCONTIG are often confused. Another advantage is that sparse doesn't require each NUMA node's ranges to be contiguous. It can handle overlapping ranges between nodes with no problems, where DISCONTIGMEM currently throws away that memory. Sparsemem uses an array to provide different pfn_to_page() translations for each SECTION_SIZE area of physical memory. This is what allows the mem_map[] to be chopped up. In order to do quick pfn_to_page() operations, the section number of the page is encoded in page->flags. Part of the sparsemem infrastructure enables sharing of these bits more dynamically (at compile-time) between the page_zone() and sparsemem operations. However, on 32-bit architectures, the number of bits is quite limited, and may require growing the size of the page->flags type in certain conditions. Several things might force this to occur: a decrease in the SECTION_SIZE (if you want to hotplug smaller areas of memory), an increase in the physical address space, or an increase in the number of used page->flags. One thing to note is that, once sparsemem is present, the NUMA node information no longer needs to be stored in the page->flags. It might provide speed increases on certain platforms and will be stored there if there is room. But, if out of room, an alternate (theoretically slower) mechanism is used. This patch introduces CONFIG_FLATMEM. It is used in almost all cases where there used to be an #ifndef DISCONTIG, because SPARSEMEM and DISCONTIGMEM often have to compile out the same areas of code. Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Martin Bligh <mbligh@aracnet.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com> Signed-off-by: Bob Picco <bob.picco@hp.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 07:07:54 +00:00
zonetable_add(zone, nid, j, zone_start_pfn, size);
init_currently_empty_zone(zone, zone_start_pfn, size);
zone_start_pfn += size;
}
}
static void __init alloc_node_mem_map(struct pglist_data *pgdat)
{
/* Skip empty nodes */
if (!pgdat->node_spanned_pages)
return;
[PATCH] sparsemem memory model Sparsemem abstracts the use of discontiguous mem_maps[]. This kind of mem_map[] is needed by discontiguous memory machines (like in the old CONFIG_DISCONTIGMEM case) as well as memory hotplug systems. Sparsemem replaces DISCONTIGMEM when enabled, and it is hoped that it can eventually become a complete replacement. A significant advantage over DISCONTIGMEM is that it's completely separated from CONFIG_NUMA. When producing this patch, it became apparent in that NUMA and DISCONTIG are often confused. Another advantage is that sparse doesn't require each NUMA node's ranges to be contiguous. It can handle overlapping ranges between nodes with no problems, where DISCONTIGMEM currently throws away that memory. Sparsemem uses an array to provide different pfn_to_page() translations for each SECTION_SIZE area of physical memory. This is what allows the mem_map[] to be chopped up. In order to do quick pfn_to_page() operations, the section number of the page is encoded in page->flags. Part of the sparsemem infrastructure enables sharing of these bits more dynamically (at compile-time) between the page_zone() and sparsemem operations. However, on 32-bit architectures, the number of bits is quite limited, and may require growing the size of the page->flags type in certain conditions. Several things might force this to occur: a decrease in the SECTION_SIZE (if you want to hotplug smaller areas of memory), an increase in the physical address space, or an increase in the number of used page->flags. One thing to note is that, once sparsemem is present, the NUMA node information no longer needs to be stored in the page->flags. It might provide speed increases on certain platforms and will be stored there if there is room. But, if out of room, an alternate (theoretically slower) mechanism is used. This patch introduces CONFIG_FLATMEM. It is used in almost all cases where there used to be an #ifndef DISCONTIG, because SPARSEMEM and DISCONTIGMEM often have to compile out the same areas of code. Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Martin Bligh <mbligh@aracnet.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com> Signed-off-by: Bob Picco <bob.picco@hp.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 07:07:54 +00:00
#ifdef CONFIG_FLAT_NODE_MEM_MAP
/* ia64 gets its own node_mem_map, before this, without bootmem */
if (!pgdat->node_mem_map) {
[PATCH] sparsemem memory model Sparsemem abstracts the use of discontiguous mem_maps[]. This kind of mem_map[] is needed by discontiguous memory machines (like in the old CONFIG_DISCONTIGMEM case) as well as memory hotplug systems. Sparsemem replaces DISCONTIGMEM when enabled, and it is hoped that it can eventually become a complete replacement. A significant advantage over DISCONTIGMEM is that it's completely separated from CONFIG_NUMA. When producing this patch, it became apparent in that NUMA and DISCONTIG are often confused. Another advantage is that sparse doesn't require each NUMA node's ranges to be contiguous. It can handle overlapping ranges between nodes with no problems, where DISCONTIGMEM currently throws away that memory. Sparsemem uses an array to provide different pfn_to_page() translations for each SECTION_SIZE area of physical memory. This is what allows the mem_map[] to be chopped up. In order to do quick pfn_to_page() operations, the section number of the page is encoded in page->flags. Part of the sparsemem infrastructure enables sharing of these bits more dynamically (at compile-time) between the page_zone() and sparsemem operations. However, on 32-bit architectures, the number of bits is quite limited, and may require growing the size of the page->flags type in certain conditions. Several things might force this to occur: a decrease in the SECTION_SIZE (if you want to hotplug smaller areas of memory), an increase in the physical address space, or an increase in the number of used page->flags. One thing to note is that, once sparsemem is present, the NUMA node information no longer needs to be stored in the page->flags. It might provide speed increases on certain platforms and will be stored there if there is room. But, if out of room, an alternate (theoretically slower) mechanism is used. This patch introduces CONFIG_FLATMEM. It is used in almost all cases where there used to be an #ifndef DISCONTIG, because SPARSEMEM and DISCONTIGMEM often have to compile out the same areas of code. Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Martin Bligh <mbligh@aracnet.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com> Signed-off-by: Bob Picco <bob.picco@hp.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 07:07:54 +00:00
unsigned long size;
struct page *map;
size = (pgdat->node_spanned_pages + 1) * sizeof(struct page);
map = alloc_remap(pgdat->node_id, size);
if (!map)
map = alloc_bootmem_node(pgdat, size);
pgdat->node_mem_map = map;
}
[PATCH] sparsemem memory model Sparsemem abstracts the use of discontiguous mem_maps[]. This kind of mem_map[] is needed by discontiguous memory machines (like in the old CONFIG_DISCONTIGMEM case) as well as memory hotplug systems. Sparsemem replaces DISCONTIGMEM when enabled, and it is hoped that it can eventually become a complete replacement. A significant advantage over DISCONTIGMEM is that it's completely separated from CONFIG_NUMA. When producing this patch, it became apparent in that NUMA and DISCONTIG are often confused. Another advantage is that sparse doesn't require each NUMA node's ranges to be contiguous. It can handle overlapping ranges between nodes with no problems, where DISCONTIGMEM currently throws away that memory. Sparsemem uses an array to provide different pfn_to_page() translations for each SECTION_SIZE area of physical memory. This is what allows the mem_map[] to be chopped up. In order to do quick pfn_to_page() operations, the section number of the page is encoded in page->flags. Part of the sparsemem infrastructure enables sharing of these bits more dynamically (at compile-time) between the page_zone() and sparsemem operations. However, on 32-bit architectures, the number of bits is quite limited, and may require growing the size of the page->flags type in certain conditions. Several things might force this to occur: a decrease in the SECTION_SIZE (if you want to hotplug smaller areas of memory), an increase in the physical address space, or an increase in the number of used page->flags. One thing to note is that, once sparsemem is present, the NUMA node information no longer needs to be stored in the page->flags. It might provide speed increases on certain platforms and will be stored there if there is room. But, if out of room, an alternate (theoretically slower) mechanism is used. This patch introduces CONFIG_FLATMEM. It is used in almost all cases where there used to be an #ifndef DISCONTIG, because SPARSEMEM and DISCONTIGMEM often have to compile out the same areas of code. Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Martin Bligh <mbligh@aracnet.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com> Signed-off-by: Bob Picco <bob.picco@hp.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 07:07:54 +00:00
#ifdef CONFIG_FLATMEM
/*
* With no DISCONTIG, the global mem_map is just set as node 0's
*/
if (pgdat == NODE_DATA(0))
mem_map = NODE_DATA(0)->node_mem_map;
#endif
[PATCH] sparsemem memory model Sparsemem abstracts the use of discontiguous mem_maps[]. This kind of mem_map[] is needed by discontiguous memory machines (like in the old CONFIG_DISCONTIGMEM case) as well as memory hotplug systems. Sparsemem replaces DISCONTIGMEM when enabled, and it is hoped that it can eventually become a complete replacement. A significant advantage over DISCONTIGMEM is that it's completely separated from CONFIG_NUMA. When producing this patch, it became apparent in that NUMA and DISCONTIG are often confused. Another advantage is that sparse doesn't require each NUMA node's ranges to be contiguous. It can handle overlapping ranges between nodes with no problems, where DISCONTIGMEM currently throws away that memory. Sparsemem uses an array to provide different pfn_to_page() translations for each SECTION_SIZE area of physical memory. This is what allows the mem_map[] to be chopped up. In order to do quick pfn_to_page() operations, the section number of the page is encoded in page->flags. Part of the sparsemem infrastructure enables sharing of these bits more dynamically (at compile-time) between the page_zone() and sparsemem operations. However, on 32-bit architectures, the number of bits is quite limited, and may require growing the size of the page->flags type in certain conditions. Several things might force this to occur: a decrease in the SECTION_SIZE (if you want to hotplug smaller areas of memory), an increase in the physical address space, or an increase in the number of used page->flags. One thing to note is that, once sparsemem is present, the NUMA node information no longer needs to be stored in the page->flags. It might provide speed increases on certain platforms and will be stored there if there is room. But, if out of room, an alternate (theoretically slower) mechanism is used. This patch introduces CONFIG_FLATMEM. It is used in almost all cases where there used to be an #ifndef DISCONTIG, because SPARSEMEM and DISCONTIGMEM often have to compile out the same areas of code. Signed-off-by: Andy Whitcroft <apw@shadowen.org> Signed-off-by: Dave Hansen <haveblue@us.ibm.com> Signed-off-by: Martin Bligh <mbligh@aracnet.com> Signed-off-by: Adrian Bunk <bunk@stusta.de> Signed-off-by: Yasunori Goto <y-goto@jp.fujitsu.com> Signed-off-by: Bob Picco <bob.picco@hp.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-23 07:07:54 +00:00
#endif /* CONFIG_FLAT_NODE_MEM_MAP */
}
void __init free_area_init_node(int nid, struct pglist_data *pgdat,
unsigned long *zones_size, unsigned long node_start_pfn,
unsigned long *zholes_size)
{
pgdat->node_id = nid;
pgdat->node_start_pfn = node_start_pfn;
calculate_zone_totalpages(pgdat, zones_size, zholes_size);
alloc_node_mem_map(pgdat);
free_area_init_core(pgdat, zones_size, zholes_size);
}
#ifndef CONFIG_NEED_MULTIPLE_NODES
static bootmem_data_t contig_bootmem_data;
struct pglist_data contig_page_data = { .bdata = &contig_bootmem_data };
EXPORT_SYMBOL(contig_page_data);
#endif
void __init free_area_init(unsigned long *zones_size)
{
free_area_init_node(0, NODE_DATA(0), zones_size,
__pa(PAGE_OFFSET) >> PAGE_SHIFT, NULL);
}
#ifdef CONFIG_PROC_FS
#include <linux/seq_file.h>
static void *frag_start(struct seq_file *m, loff_t *pos)
{
pg_data_t *pgdat;
loff_t node = *pos;
for (pgdat = pgdat_list; pgdat && node; pgdat = pgdat->pgdat_next)
--node;
return pgdat;
}
static void *frag_next(struct seq_file *m, void *arg, loff_t *pos)
{
pg_data_t *pgdat = (pg_data_t *)arg;
(*pos)++;
return pgdat->pgdat_next;
}
static void frag_stop(struct seq_file *m, void *arg)
{
}
/*
* This walks the free areas for each zone.
*/
static int frag_show(struct seq_file *m, void *arg)
{
pg_data_t *pgdat = (pg_data_t *)arg;
struct zone *zone;
struct zone *node_zones = pgdat->node_zones;
unsigned long flags;
int order;
for (zone = node_zones; zone - node_zones < MAX_NR_ZONES; ++zone) {
if (!populated_zone(zone))
continue;
spin_lock_irqsave(&zone->lock, flags);
seq_printf(m, "Node %d, zone %8s ", pgdat->node_id, zone->name);
for (order = 0; order < MAX_ORDER; ++order)
seq_printf(m, "%6lu ", zone->free_area[order].nr_free);
spin_unlock_irqrestore(&zone->lock, flags);
seq_putc(m, '\n');
}
return 0;
}
struct seq_operations fragmentation_op = {
.start = frag_start,
.next = frag_next,
.stop = frag_stop,
.show = frag_show,
};
/*
* Output information about zones in @pgdat.
*/
static int zoneinfo_show(struct seq_file *m, void *arg)
{
pg_data_t *pgdat = arg;
struct zone *zone;
struct zone *node_zones = pgdat->node_zones;
unsigned long flags;
for (zone = node_zones; zone - node_zones < MAX_NR_ZONES; zone++) {
int i;
if (!populated_zone(zone))
continue;
spin_lock_irqsave(&zone->lock, flags);
seq_printf(m, "Node %d, zone %8s", pgdat->node_id, zone->name);
seq_printf(m,
"\n pages free %lu"
"\n min %lu"
"\n low %lu"
"\n high %lu"
"\n active %lu"
"\n inactive %lu"
"\n scanned %lu (a: %lu i: %lu)"
"\n spanned %lu"
"\n present %lu",
zone->free_pages,
zone->pages_min,
zone->pages_low,
zone->pages_high,
zone->nr_active,
zone->nr_inactive,
zone->pages_scanned,
zone->nr_scan_active, zone->nr_scan_inactive,
zone->spanned_pages,
zone->present_pages);
seq_printf(m,
"\n protection: (%lu",
zone->lowmem_reserve[0]);
for (i = 1; i < ARRAY_SIZE(zone->lowmem_reserve); i++)
seq_printf(m, ", %lu", zone->lowmem_reserve[i]);
seq_printf(m,
")"
"\n pagesets");
for_each_online_cpu(i) {
struct per_cpu_pageset *pageset;
int j;
[PATCH] node local per-cpu-pages This patch modifies the way pagesets in struct zone are managed. Each zone has a per-cpu array of pagesets. So any particular CPU has some memory in each zone structure which belongs to itself. Even if that CPU is not local to that zone. So the patch relocates the pagesets for each cpu to the node that is nearest to the cpu instead of allocating the pagesets in the (possibly remote) target zone. This means that the operations to manage pages on remote zone can be done with information available locally. We play a macro trick so that non-NUMA pmachines avoid the additional pointer chase on the page allocator fastpath. AIM7 benchmark on a 32 CPU SGI Altix w/o patches: Tasks jobs/min jti jobs/min/task real cpu 1 484.68 100 484.6769 12.01 1.97 Fri Mar 25 11:01:42 2005 100 27140.46 89 271.4046 21.44 148.71 Fri Mar 25 11:02:04 2005 200 30792.02 82 153.9601 37.80 296.72 Fri Mar 25 11:02:42 2005 300 32209.27 81 107.3642 54.21 451.34 Fri Mar 25 11:03:37 2005 400 34962.83 78 87.4071 66.59 588.97 Fri Mar 25 11:04:44 2005 500 31676.92 75 63.3538 91.87 742.71 Fri Mar 25 11:06:16 2005 600 36032.69 73 60.0545 96.91 885.44 Fri Mar 25 11:07:54 2005 700 35540.43 77 50.7720 114.63 1024.28 Fri Mar 25 11:09:49 2005 800 33906.70 74 42.3834 137.32 1181.65 Fri Mar 25 11:12:06 2005 900 34120.67 73 37.9119 153.51 1325.26 Fri Mar 25 11:14:41 2005 1000 34802.37 74 34.8024 167.23 1465.26 Fri Mar 25 11:17:28 2005 with slab API changes and pageset patch: Tasks jobs/min jti jobs/min/task real cpu 1 485.00 100 485.0000 12.00 1.96 Fri Mar 25 11:46:18 2005 100 28000.96 89 280.0096 20.79 150.45 Fri Mar 25 11:46:39 2005 200 32285.80 79 161.4290 36.05 293.37 Fri Mar 25 11:47:16 2005 300 40424.15 84 134.7472 43.19 438.42 Fri Mar 25 11:47:59 2005 400 39155.01 79 97.8875 59.46 590.05 Fri Mar 25 11:48:59 2005 500 37881.25 82 75.7625 76.82 730.19 Fri Mar 25 11:50:16 2005 600 39083.14 78 65.1386 89.35 872.79 Fri Mar 25 11:51:46 2005 700 38627.83 77 55.1826 105.47 1022.46 Fri Mar 25 11:53:32 2005 800 39631.94 78 49.5399 117.48 1169.94 Fri Mar 25 11:55:30 2005 900 36903.70 79 41.0041 141.94 1310.78 Fri Mar 25 11:57:53 2005 1000 36201.23 77 36.2012 160.77 1458.31 Fri Mar 25 12:00:34 2005 Signed-off-by: Christoph Lameter <clameter@sgi.com> Signed-off-by: Shobhit Dayal <shobhit@calsoftinc.com> Signed-off-by: Shai Fultheim <Shai@Scalex86.org> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2005-06-22 00:14:47 +00:00
pageset = zone_pcp(zone, i);
for (j = 0; j < ARRAY_SIZE(pageset->pcp); j++) {
if (pageset->pcp[j].count)
break;
}
if (j == ARRAY_SIZE(pageset->pcp))
continue;
for (j = 0; j < ARRAY_SIZE(pageset->pcp); j++) {
seq_printf(m,
"\n cpu: %i pcp: %i"
"\n count: %i"
"\n high: %i"
"\n batch: %i",
i, j,
pageset->pcp[j].count,
pageset->pcp[j].high,
pageset->pcp[j].batch);
}
#ifdef CONFIG_NUMA
seq_printf(m,
"\n numa_hit: %lu"
"\n numa_miss: %lu"
"\n numa_foreign: %lu"
"\n interleave_hit: %lu"
"\n local_node: %lu"
"\n other_node: %lu",
pageset->numa_hit,
pageset->numa_miss,
pageset->numa_foreign,
pageset->interleave_hit,
pageset->local_node,
pageset->other_node);
#endif
}
seq_printf(m,
"\n all_unreclaimable: %u"
"\n prev_priority: %i"
"\n temp_priority: %i"
"\n start_pfn: %lu",
zone->all_unreclaimable,
zone->prev_priority,
zone->temp_priority,
zone->zone_start_pfn);
spin_unlock_irqrestore(&zone->lock, flags);
seq_putc(m, '\n');
}
return 0;
}
struct seq_operations zoneinfo_op = {
.start = frag_start, /* iterate over all zones. The same as in
* fragmentation. */
.next = frag_next,
.stop = frag_stop,
.show = zoneinfo_show,
};
static char *vmstat_text[] = {
"nr_dirty",
"nr_writeback",
"nr_unstable",
"nr_page_table_pages",
"nr_mapped",
"nr_slab",
"pgpgin",
"pgpgout",
"pswpin",
"pswpout",
"pgalloc_high",
"pgalloc_normal",
"pgalloc_dma32",
"pgalloc_dma",
"pgfree",
"pgactivate",
"pgdeactivate",
"pgfault",
"pgmajfault",
"pgrefill_high",
"pgrefill_normal",
"pgrefill_dma32",
"pgrefill_dma",
"pgsteal_high",
"pgsteal_normal",
"pgsteal_dma32",
"pgsteal_dma",
"pgscan_kswapd_high",
"pgscan_kswapd_normal",
"pgscan_kswapd_dma32",
"pgscan_kswapd_dma",
"pgscan_direct_high",
"pgscan_direct_normal",
"pgscan_direct_dma32",
"pgscan_direct_dma",
"pginodesteal",
"slabs_scanned",
"kswapd_steal",
"kswapd_inodesteal",
"pageoutrun",
"allocstall",
"pgrotated",
"nr_bounce",
};
static void *vmstat_start(struct seq_file *m, loff_t *pos)
{
struct page_state *ps;
if (*pos >= ARRAY_SIZE(vmstat_text))
return NULL;
ps = kmalloc(sizeof(*ps), GFP_KERNEL);
m->private = ps;
if (!ps)
return ERR_PTR(-ENOMEM);
get_full_page_state(ps);
ps->pgpgin /= 2; /* sectors -> kbytes */
ps->pgpgout /= 2;
return (unsigned long *)ps + *pos;
}
static void *vmstat_next(struct seq_file *m, void *arg, loff_t *pos)
{
(*pos)++;
if (*pos >= ARRAY_SIZE(vmstat_text))
return NULL;
return (unsigned long *)m->private + *pos;
}
static int vmstat_show(struct seq_file *m, void *arg)
{
unsigned long *l = arg;
unsigned long off = l - (unsigned long *)m->private;
seq_printf(m, "%s %lu\n", vmstat_text[off], *l);
return 0;
}
static void vmstat_stop(struct seq_file *m, void *arg)
{
kfree(m->private);
m->private = NULL;
}
struct seq_operations vmstat_op = {
.start = vmstat_start,
.next = vmstat_next,
.stop = vmstat_stop,
.show = vmstat_show,
};
#endif /* CONFIG_PROC_FS */
#ifdef CONFIG_HOTPLUG_CPU
static int page_alloc_cpu_notify(struct notifier_block *self,
unsigned long action, void *hcpu)
{
int cpu = (unsigned long)hcpu;
long *count;
unsigned long *src, *dest;
if (action == CPU_DEAD) {
int i;
/* Drain local pagecache count. */
count = &per_cpu(nr_pagecache_local, cpu);
atomic_add(*count, &nr_pagecache);
*count = 0;
local_irq_disable();
__drain_pages(cpu);
/* Add dead cpu's page_states to our own. */
dest = (unsigned long *)&__get_cpu_var(page_states);
src = (unsigned long *)&per_cpu(page_states, cpu);
for (i = 0; i < sizeof(struct page_state)/sizeof(unsigned long);
i++) {
dest[i] += src[i];
src[i] = 0;
}
local_irq_enable();
}
return NOTIFY_OK;
}
#endif /* CONFIG_HOTPLUG_CPU */
void __init page_alloc_init(void)
{
hotcpu_notifier(page_alloc_cpu_notify, 0);
}
/*
* setup_per_zone_lowmem_reserve - called whenever
* sysctl_lower_zone_reserve_ratio changes. Ensures that each zone
* has a correct pages reserved value, so an adequate number of
* pages are left in the zone after a successful __alloc_pages().
*/
static void setup_per_zone_lowmem_reserve(void)
{
struct pglist_data *pgdat;
int j, idx;
for_each_pgdat(pgdat) {
for (j = 0; j < MAX_NR_ZONES; j++) {
struct zone *zone = pgdat->node_zones + j;
unsigned long present_pages = zone->present_pages;
zone->lowmem_reserve[j] = 0;
for (idx = j-1; idx >= 0; idx--) {
struct zone *lower_zone;
if (sysctl_lowmem_reserve_ratio[idx] < 1)
sysctl_lowmem_reserve_ratio[idx] = 1;
lower_zone = pgdat->node_zones + idx;
lower_zone->lowmem_reserve[j] = present_pages /
sysctl_lowmem_reserve_ratio[idx];
present_pages += lower_zone->present_pages;
}
}
}
}
/*
* setup_per_zone_pages_min - called when min_free_kbytes changes. Ensures
* that the pages_{min,low,high} values for each zone are set correctly
* with respect to min_free_kbytes.
*/
void setup_per_zone_pages_min(void)
{
unsigned long pages_min = min_free_kbytes >> (PAGE_SHIFT - 10);
unsigned long lowmem_pages = 0;
struct zone *zone;
unsigned long flags;
/* Calculate total number of !ZONE_HIGHMEM pages */
for_each_zone(zone) {
if (!is_highmem(zone))
lowmem_pages += zone->present_pages;
}
for_each_zone(zone) {
unsigned long tmp;
spin_lock_irqsave(&zone->lru_lock, flags);
tmp = (pages_min * zone->present_pages) / lowmem_pages;
if (is_highmem(zone)) {
/*
* __GFP_HIGH and PF_MEMALLOC allocations usually don't
* need highmem pages, so cap pages_min to a small
* value here.
*
* The (pages_high-pages_low) and (pages_low-pages_min)
* deltas controls asynch page reclaim, and so should
* not be capped for highmem.
*/
int min_pages;
min_pages = zone->present_pages / 1024;
if (min_pages < SWAP_CLUSTER_MAX)
min_pages = SWAP_CLUSTER_MAX;
if (min_pages > 128)
min_pages = 128;
zone->pages_min = min_pages;
} else {
/*
* If it's a lowmem zone, reserve a number of pages
* proportionate to the zone's size.
*/
zone->pages_min = tmp;
}
zone->pages_low = zone->pages_min + tmp / 4;
zone->pages_high = zone->pages_min + tmp / 2;
spin_unlock_irqrestore(&zone->lru_lock, flags);
}
}
/*
* Initialise min_free_kbytes.
*
* For small machines we want it small (128k min). For large machines
* we want it large (64MB max). But it is not linear, because network
* bandwidth does not increase linearly with machine size. We use
*
* min_free_kbytes = 4 * sqrt(lowmem_kbytes), for better accuracy:
* min_free_kbytes = sqrt(lowmem_kbytes * 16)
*
* which yields
*
* 16MB: 512k
* 32MB: 724k
* 64MB: 1024k
* 128MB: 1448k
* 256MB: 2048k
* 512MB: 2896k
* 1024MB: 4096k
* 2048MB: 5792k
* 4096MB: 8192k
* 8192MB: 11584k
* 16384MB: 16384k
*/
static int __init init_per_zone_pages_min(void)
{
unsigned long lowmem_kbytes;
lowmem_kbytes = nr_free_buffer_pages() * (PAGE_SIZE >> 10);
min_free_kbytes = int_sqrt(lowmem_kbytes * 16);
if (min_free_kbytes < 128)
min_free_kbytes = 128;
if (min_free_kbytes > 65536)
min_free_kbytes = 65536;
setup_per_zone_pages_min();
setup_per_zone_lowmem_reserve();
return 0;
}
module_init(init_per_zone_pages_min)
/*
* min_free_kbytes_sysctl_handler - just a wrapper around proc_dointvec() so
* that we can call two helper functions whenever min_free_kbytes
* changes.
*/
int min_free_kbytes_sysctl_handler(ctl_table *table, int write,
struct file *file, void __user *buffer, size_t *length, loff_t *ppos)
{
proc_dointvec(table, write, file, buffer, length, ppos);
setup_per_zone_pages_min();
return 0;
}
/*
* lowmem_reserve_ratio_sysctl_handler - just a wrapper around
* proc_dointvec() so that we can call setup_per_zone_lowmem_reserve()
* whenever sysctl_lowmem_reserve_ratio changes.
*
* The reserve ratio obviously has absolutely no relation with the
* pages_min watermarks. The lowmem reserve ratio can only make sense
* if in function of the boot time zone sizes.
*/
int lowmem_reserve_ratio_sysctl_handler(ctl_table *table, int write,
struct file *file, void __user *buffer, size_t *length, loff_t *ppos)
{
proc_dointvec_minmax(table, write, file, buffer, length, ppos);
setup_per_zone_lowmem_reserve();
return 0;
}
/*
* percpu_pagelist_fraction - changes the pcp->high for each zone on each
* cpu. It is the fraction of total pages in each zone that a hot per cpu pagelist
* can have before it gets flushed back to buddy allocator.
*/
int percpu_pagelist_fraction_sysctl_handler(ctl_table *table, int write,
struct file *file, void __user *buffer, size_t *length, loff_t *ppos)
{
struct zone *zone;
unsigned int cpu;
int ret;
ret = proc_dointvec_minmax(table, write, file, buffer, length, ppos);
if (!write || (ret == -EINVAL))
return ret;
for_each_zone(zone) {
for_each_online_cpu(cpu) {
unsigned long high;
high = zone->present_pages / percpu_pagelist_fraction;
setup_pagelist_highmark(zone_pcp(zone, cpu), high);
}
}
return 0;
}
__initdata int hashdist = HASHDIST_DEFAULT;
#ifdef CONFIG_NUMA
static int __init set_hashdist(char *str)
{
if (!str)
return 0;
hashdist = simple_strtoul(str, &str, 0);
return 1;
}
__setup("hashdist=", set_hashdist);
#endif
/*
* allocate a large system hash table from bootmem
* - it is assumed that the hash table must contain an exact power-of-2
* quantity of entries
* - limit is the number of hash buckets, not the total allocation size
*/
void *__init alloc_large_system_hash(const char *tablename,
unsigned long bucketsize,
unsigned long numentries,
int scale,
int flags,
unsigned int *_hash_shift,
unsigned int *_hash_mask,
unsigned long limit)
{
unsigned long long max = limit;
unsigned long log2qty, size;
void *table = NULL;
/* allow the kernel cmdline to have a say */
if (!numentries) {
/* round applicable memory size up to nearest megabyte */
numentries = (flags & HASH_HIGHMEM) ? nr_all_pages : nr_kernel_pages;
numentries += (1UL << (20 - PAGE_SHIFT)) - 1;
numentries >>= 20 - PAGE_SHIFT;
numentries <<= 20 - PAGE_SHIFT;
/* limit to 1 bucket per 2^scale bytes of low memory */
if (scale > PAGE_SHIFT)
numentries >>= (scale - PAGE_SHIFT);
else
numentries <<= (PAGE_SHIFT - scale);
}
/* rounded up to nearest power of 2 in size */
numentries = 1UL << (long_log2(numentries) + 1);
/* limit allocation size to 1/16 total memory by default */
if (max == 0) {
max = ((unsigned long long)nr_all_pages << PAGE_SHIFT) >> 4;
do_div(max, bucketsize);
}
if (numentries > max)
numentries = max;
log2qty = long_log2(numentries);
do {
size = bucketsize << log2qty;
if (flags & HASH_EARLY)
table = alloc_bootmem(size);
else if (hashdist)
table = __vmalloc(size, GFP_ATOMIC, PAGE_KERNEL);
else {
unsigned long order;
for (order = 0; ((1UL << order) << PAGE_SHIFT) < size; order++)
;
table = (void*) __get_free_pages(GFP_ATOMIC, order);
}
} while (!table && size > PAGE_SIZE && --log2qty);
if (!table)
panic("Failed to allocate %s hash table\n", tablename);
printk("%s hash table entries: %d (order: %d, %lu bytes)\n",
tablename,
(1U << log2qty),
long_log2(size) - PAGE_SHIFT,
size);
if (_hash_shift)
*_hash_shift = log2qty;
if (_hash_mask)
*_hash_mask = (1 << log2qty) - 1;
return table;
}