linux/mm/hugetlb.c

1238 lines
31 KiB
C
Raw Normal View History

/*
* Generic hugetlb support.
* (C) William Irwin, April 2004
*/
#include <linux/gfp.h>
#include <linux/list.h>
#include <linux/init.h>
#include <linux/module.h>
#include <linux/mm.h>
#include <linux/sysctl.h>
#include <linux/highmem.h>
#include <linux/nodemask.h>
#include <linux/pagemap.h>
#include <linux/mempolicy.h>
#include <linux/cpuset.h>
[PATCH] hugepage: serialize hugepage allocation and instantiation Currently, no lock or mutex is held between allocating a hugepage and inserting it into the pagetables / page cache. When we do go to insert the page into pagetables or page cache, we recheck and may free the newly allocated hugepage. However, since the number of hugepages in the system is strictly limited, and it's usualy to want to use all of them, this can still lead to spurious allocation failures. For example, suppose two processes are both mapping (MAP_SHARED) the same hugepage file, large enough to consume the entire available hugepage pool. If they race instantiating the last page in the mapping, they will both attempt to allocate the last available hugepage. One will fail, of course, returning OOM from the fault and thus causing the process to be killed, despite the fact that the entire mapping can, in fact, be instantiated. The patch fixes this race by the simple method of adding a (sleeping) mutex to serialize the hugepage fault path between allocation and insertion into pagetables and/or page cache. It would be possible to avoid the serialization by catching the allocation failures, waiting on some condition, then rechecking to see if someone else has instantiated the page for us. Given the likely frequency of hugepage instantiations, it seems very doubtful it's worth the extra complexity. This patch causes no regression on the libhugetlbfs testsuite, and one test, which can trigger this race now passes where it previously failed. Actually, the test still sometimes fails, though less often and only as a shmat() failure, rather processes getting OOM killed by the VM. The dodgy heuristic tests in fs/hugetlbfs/inode.c for whether there's enough hugepage space aren't protected by the new mutex, and would be ugly to do so, so there's still a race there. Another patch to replace those tests with something saner for this reason as well as others coming... Signed-off-by: David Gibson <dwg@au1.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:53 +00:00
#include <linux/mutex.h>
#include <asm/page.h>
#include <asm/pgtable.h>
#include <linux/hugetlb.h>
#include "internal.h"
const unsigned long hugetlb_zero = 0, hugetlb_infinity = ~0UL;
static unsigned long nr_huge_pages, free_huge_pages, resv_huge_pages;
static unsigned long surplus_huge_pages;
unsigned long max_huge_pages;
static struct list_head hugepage_freelists[MAX_NUMNODES];
static unsigned int nr_huge_pages_node[MAX_NUMNODES];
static unsigned int free_huge_pages_node[MAX_NUMNODES];
static unsigned int surplus_huge_pages_node[MAX_NUMNODES];
static gfp_t htlb_alloc_mask = GFP_HIGHUSER;
unsigned long hugepages_treat_as_movable;
hugetlb: introduce nr_overcommit_hugepages sysctl hugetlb: introduce nr_overcommit_hugepages sysctl While examining the code to support /proc/sys/vm/hugetlb_dynamic_pool, I became convinced that having a boolean sysctl was insufficient: 1) To support per-node control of hugepages, I have previously submitted patches to add a sysfs attribute related to nr_hugepages. However, with a boolean global value and per-mount quota enforcement constraining the dynamic pool, adding corresponding control of the dynamic pool on a per-node basis seems inconsistent to me. 2) Administration of the hugetlb dynamic pool with multiple hugetlbfs mount points is, arguably, more arduous than it needs to be. Each quota would need to be set separately, and the sum would need to be monitored. To ease the administration, and to help make the way for per-node control of the static & dynamic hugepage pool, I added a separate sysctl, nr_overcommit_hugepages. This value serves as a high watermark for the overall hugepage pool, while nr_hugepages serves as a low watermark. The boolean sysctl can then be removed, as the condition nr_overcommit_hugepages > 0 indicates the same administrative setting as hugetlb_dynamic_pool == 1 Quotas still serve as local enforcement of the size of the pool on a per-mount basis. A few caveats: 1) There is a race whereby the global surplus huge page counter is incremented before a hugepage has allocated. Another process could then try grow the pool, and fail to convert a surplus huge page to a normal huge page and instead allocate a fresh huge page. I believe this is benign, as no memory is leaked (the actual pages are still tracked correctly) and the counters won't go out of sync. 2) Shrinking the static pool while a surplus is in effect will allow the number of surplus huge pages to exceed the overcommit value. As long as this condition holds, however, no more surplus huge pages will be allowed on the system until one of the two sysctls are increased sufficiently, or the surplus huge pages go out of use and are freed. Successfully tested on x86_64 with the current libhugetlbfs snapshot, modified to use the new sysctl. Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Adam Litke <agl@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-12-18 00:20:12 +00:00
unsigned long nr_overcommit_huge_pages;
hugetlb: fix hugepage allocation with memoryless nodes Anton found a problem with the hugetlb pool allocation when some nodes have no memory (http://marc.info/?l=linux-mm&m=118133042025995&w=2). Lee worked on versions that tried to fix it, but none were accepted. Christoph has created a set of patches which allow for GFP_THISNODE allocations to fail if the node has no memory. Currently, alloc_fresh_huge_page() returns NULL when it is not able to allocate a huge page on the current node, as specified by its custom interleave variable. The callers of this function, though, assume that a failure in alloc_fresh_huge_page() indicates no hugepages can be allocated on the system period. This might not be the case, for instance, if we have an uneven NUMA system, and we happen to try to allocate a hugepage on a node with less memory and fail, while there is still plenty of free memory on the other nodes. To correct this, make alloc_fresh_huge_page() search through all online nodes before deciding no hugepages can be allocated. Add a helper function for actually allocating the hugepage. Use a new global nid iterator to control which nid to allocate on. Note: we expect particular semantics for __GFP_THISNODE, which are now enforced even for memoryless nodes. That is, there is should be no fallback to other nodes. Therefore, we rely on the nid passed into alloc_pages_node() to be the nid the page comes from. If this is incorrect, accounting will break. Tested on x86 !NUMA, x86 NUMA, x86_64 NUMA and ppc64 NUMA (with 2 memoryless nodes). Before on the ppc64 box: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 25 Node 1 HugePages_Free: 75 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 150 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free After: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 50 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 100 Node 1 HugePages_Free: 100 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Christoph Lameter <clameter@sgi.com> Cc: Adam Litke <agl@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:26:24 +00:00
static int hugetlb_next_nid;
[PATCH] hugepage: serialize hugepage allocation and instantiation Currently, no lock or mutex is held between allocating a hugepage and inserting it into the pagetables / page cache. When we do go to insert the page into pagetables or page cache, we recheck and may free the newly allocated hugepage. However, since the number of hugepages in the system is strictly limited, and it's usualy to want to use all of them, this can still lead to spurious allocation failures. For example, suppose two processes are both mapping (MAP_SHARED) the same hugepage file, large enough to consume the entire available hugepage pool. If they race instantiating the last page in the mapping, they will both attempt to allocate the last available hugepage. One will fail, of course, returning OOM from the fault and thus causing the process to be killed, despite the fact that the entire mapping can, in fact, be instantiated. The patch fixes this race by the simple method of adding a (sleeping) mutex to serialize the hugepage fault path between allocation and insertion into pagetables and/or page cache. It would be possible to avoid the serialization by catching the allocation failures, waiting on some condition, then rechecking to see if someone else has instantiated the page for us. Given the likely frequency of hugepage instantiations, it seems very doubtful it's worth the extra complexity. This patch causes no regression on the libhugetlbfs testsuite, and one test, which can trigger this race now passes where it previously failed. Actually, the test still sometimes fails, though less often and only as a shmat() failure, rather processes getting OOM killed by the VM. The dodgy heuristic tests in fs/hugetlbfs/inode.c for whether there's enough hugepage space aren't protected by the new mutex, and would be ugly to do so, so there's still a race there. Another patch to replace those tests with something saner for this reason as well as others coming... Signed-off-by: David Gibson <dwg@au1.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:53 +00:00
/*
* Protects updates to hugepage_freelists, nr_huge_pages, and free_huge_pages
*/
static DEFINE_SPINLOCK(hugetlb_lock);
static void clear_huge_page(struct page *page, unsigned long addr)
{
int i;
might_sleep();
for (i = 0; i < (HPAGE_SIZE/PAGE_SIZE); i++) {
cond_resched();
clear_user_highpage(page + i, addr + i * PAGE_SIZE);
}
}
static void copy_huge_page(struct page *dst, struct page *src,
unsigned long addr, struct vm_area_struct *vma)
{
int i;
might_sleep();
for (i = 0; i < HPAGE_SIZE/PAGE_SIZE; i++) {
cond_resched();
copy_user_highpage(dst + i, src + i, addr + i*PAGE_SIZE, vma);
}
}
static void enqueue_huge_page(struct page *page)
{
int nid = page_to_nid(page);
list_add(&page->lru, &hugepage_freelists[nid]);
free_huge_pages++;
free_huge_pages_node[nid]++;
}
static struct page *dequeue_huge_page(struct vm_area_struct *vma,
unsigned long address)
{
int nid;
struct page *page = NULL;
Fix NUMA Memory Policy Reference Counting This patch proposes fixes to the reference counting of memory policy in the page allocation paths and in show_numa_map(). Extracted from my "Memory Policy Cleanups and Enhancements" series as stand-alone. Shared policy lookup [shmem] has always added a reference to the policy, but this was never unrefed after page allocation or after formatting the numa map data. Default system policy should not require additional ref counting, nor should the current task's task policy. However, show_numa_map() calls get_vma_policy() to examine what may be [likely is] another task's policy. The latter case needs protection against freeing of the policy. This patch adds a reference count to a mempolicy returned by get_vma_policy() when the policy is a vma policy or another task's mempolicy. Again, shared policy is already reference counted on lookup. A matching "unref" [__mpol_free()] is performed in alloc_page_vma() for shared and vma policies, and in show_numa_map() for shared and another task's mempolicy. We can call __mpol_free() directly, saving an admittedly inexpensive inline NULL test, because we know we have a non-NULL policy. Handling policy ref counts for hugepages is a bit trickier. huge_zonelist() returns a zone list that might come from a shared or vma 'BIND policy. In this case, we should hold the reference until after the huge page allocation in dequeue_hugepage(). The patch modifies huge_zonelist() to return a pointer to the mempolicy if it needs to be unref'd after allocation. Kernel Build [16cpu, 32GB, ia64] - average of 10 runs: w/o patch w/ refcount patch Avg Std Devn Avg Std Devn Real: 100.59 0.38 100.63 0.43 User: 1209.60 0.37 1209.91 0.31 System: 81.52 0.42 81.64 0.34 Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com> Acked-by: Andi Kleen <ak@suse.de> Cc: Christoph Lameter <clameter@sgi.com> Acked-by: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-09-19 05:46:47 +00:00
struct mempolicy *mpol;
struct zonelist *zonelist = huge_zonelist(vma, address,
Fix NUMA Memory Policy Reference Counting This patch proposes fixes to the reference counting of memory policy in the page allocation paths and in show_numa_map(). Extracted from my "Memory Policy Cleanups and Enhancements" series as stand-alone. Shared policy lookup [shmem] has always added a reference to the policy, but this was never unrefed after page allocation or after formatting the numa map data. Default system policy should not require additional ref counting, nor should the current task's task policy. However, show_numa_map() calls get_vma_policy() to examine what may be [likely is] another task's policy. The latter case needs protection against freeing of the policy. This patch adds a reference count to a mempolicy returned by get_vma_policy() when the policy is a vma policy or another task's mempolicy. Again, shared policy is already reference counted on lookup. A matching "unref" [__mpol_free()] is performed in alloc_page_vma() for shared and vma policies, and in show_numa_map() for shared and another task's mempolicy. We can call __mpol_free() directly, saving an admittedly inexpensive inline NULL test, because we know we have a non-NULL policy. Handling policy ref counts for hugepages is a bit trickier. huge_zonelist() returns a zone list that might come from a shared or vma 'BIND policy. In this case, we should hold the reference until after the huge page allocation in dequeue_hugepage(). The patch modifies huge_zonelist() to return a pointer to the mempolicy if it needs to be unref'd after allocation. Kernel Build [16cpu, 32GB, ia64] - average of 10 runs: w/o patch w/ refcount patch Avg Std Devn Avg Std Devn Real: 100.59 0.38 100.63 0.43 User: 1209.60 0.37 1209.91 0.31 System: 81.52 0.42 81.64 0.34 Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com> Acked-by: Andi Kleen <ak@suse.de> Cc: Christoph Lameter <clameter@sgi.com> Acked-by: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-09-19 05:46:47 +00:00
htlb_alloc_mask, &mpol);
struct zone **z;
for (z = zonelist->zones; *z; z++) {
nid = zone_to_nid(*z);
if (cpuset_zone_allowed_softwall(*z, htlb_alloc_mask) &&
!list_empty(&hugepage_freelists[nid])) {
page = list_entry(hugepage_freelists[nid].next,
struct page, lru);
list_del(&page->lru);
free_huge_pages--;
free_huge_pages_node[nid]--;
if (vma && vma->vm_flags & VM_MAYSHARE)
resv_huge_pages--;
break;
}
}
Fix NUMA Memory Policy Reference Counting This patch proposes fixes to the reference counting of memory policy in the page allocation paths and in show_numa_map(). Extracted from my "Memory Policy Cleanups and Enhancements" series as stand-alone. Shared policy lookup [shmem] has always added a reference to the policy, but this was never unrefed after page allocation or after formatting the numa map data. Default system policy should not require additional ref counting, nor should the current task's task policy. However, show_numa_map() calls get_vma_policy() to examine what may be [likely is] another task's policy. The latter case needs protection against freeing of the policy. This patch adds a reference count to a mempolicy returned by get_vma_policy() when the policy is a vma policy or another task's mempolicy. Again, shared policy is already reference counted on lookup. A matching "unref" [__mpol_free()] is performed in alloc_page_vma() for shared and vma policies, and in show_numa_map() for shared and another task's mempolicy. We can call __mpol_free() directly, saving an admittedly inexpensive inline NULL test, because we know we have a non-NULL policy. Handling policy ref counts for hugepages is a bit trickier. huge_zonelist() returns a zone list that might come from a shared or vma 'BIND policy. In this case, we should hold the reference until after the huge page allocation in dequeue_hugepage(). The patch modifies huge_zonelist() to return a pointer to the mempolicy if it needs to be unref'd after allocation. Kernel Build [16cpu, 32GB, ia64] - average of 10 runs: w/o patch w/ refcount patch Avg Std Devn Avg Std Devn Real: 100.59 0.38 100.63 0.43 User: 1209.60 0.37 1209.91 0.31 System: 81.52 0.42 81.64 0.34 Signed-off-by: Lee Schermerhorn <lee.schermerhorn@hp.com> Acked-by: Andi Kleen <ak@suse.de> Cc: Christoph Lameter <clameter@sgi.com> Acked-by: Mel Gorman <mel@csn.ul.ie> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-09-19 05:46:47 +00:00
mpol_free(mpol); /* unref if mpol !NULL */
return page;
}
hugetlb: Move update_and_free_page Dynamic huge page pool resizing. In most real-world scenarios, configuring the size of the hugetlb pool correctly is a difficult task. If too few pages are allocated to the pool, applications using MAP_SHARED may fail to mmap() a hugepage region and applications using MAP_PRIVATE may receive SIGBUS. Isolating too much memory in the hugetlb pool means it is not available for other uses, especially those programs not using huge pages. The obvious answer is to let the hugetlb pool grow and shrink in response to the runtime demand for huge pages. The work Mel Gorman has been doing to establish a memory zone for movable memory allocations makes dynamically resizing the hugetlb pool reliable within the limits of that zone. This patch series implements dynamic pool resizing for private and shared mappings while being careful to maintain existing semantics. Please reply with your comments and feedback; even just to say whether it would be a useful feature to you. Thanks. How it works ============ Upon depletion of the hugetlb pool, rather than reporting an error immediately, first try and allocate the needed huge pages directly from the buddy allocator. Care must be taken to avoid unbounded growth of the hugetlb pool, so the hugetlb filesystem quota is used to limit overall pool size. The real work begins when we decide there is a shortage of huge pages. What happens next depends on whether the pages are for a private or shared mapping. Private mappings are straightforward. At fault time, if alloc_huge_page() fails, we allocate a page from the buddy allocator and increment the source node's surplus_huge_pages counter. When free_huge_page() is called for a page on a node with a surplus, the page is freed directly to the buddy allocator instead of the hugetlb pool. Because shared mappings require all of the pages to be reserved up front, some additional work must be done at mmap() to support them. We determine the reservation shortage and allocate the required number of pages all at once. These pages are then added to the hugetlb pool and marked reserved. Where that is not possible the mmap() will fail. As with private mappings, the appropriate surplus counters are updated. Since reserved huge pages won't necessarily be used by the process, we can't be sure that free_huge_page() will always be called to return surplus pages to the buddy allocator. To prevent the huge page pool from bloating, we must free unused surplus pages when their reservation has ended. Controlling it ============== With the entire patch series applied, pool resizing is off by default so unless specific action is taken, the semantics are unchanged. To take advantage of the flexibility afforded by this patch series one must tolerate a change in semantics. To control hugetlb pool growth, the following techniques can be employed: * A sysctl tunable to enable/disable the feature entirely * The size= mount option for hugetlbfs filesystems to limit pool size Performance =========== When contiguous memory is readily available, it is expected that the cost of dynamicly resizing the pool will be small. This series has been performance tested with 'stream' to measure this cost. Stream (http://www.cs.virginia.edu/stream/) was linked with libhugetlbfs to enable remapping of the text and data/bss segments into huge pages. Stream with small array ----------------------- Baseline: nr_hugepages = 0, No libhugetlbfs segment remapping Preallocated: nr_hugepages = 5, Text and data/bss remapping Dynamic: nr_hugepages = 0, Text and data/bss remapping Rate (MB/s) Function Baseline Preallocated Dynamic Copy: 4695.6266 5942.8371 5982.2287 Scale: 4451.5776 5017.1419 5658.7843 Add: 5815.8849 7927.7827 8119.3552 Triad: 5949.4144 8527.6492 8110.6903 Stream with large array ----------------------- Baseline: nr_hugepages = 0, No libhugetlbfs segment remapping Preallocated: nr_hugepages = 67, Text and data/bss remapping Dynamic: nr_hugepages = 0, Text and data/bss remapping Rate (MB/s) Function Baseline Preallocated Dynamic Copy: 2227.8281 2544.2732 2546.4947 Scale: 2136.3208 2430.7294 2421.2074 Add: 2773.1449 4004.0021 3999.4331 Triad: 2748.4502 3777.0109 3773.4970 * All numbers are averages taken from 10 consecutive runs with a maximum standard deviation of 1.3 percent noted. This patch: Simply move update_and_free_page() so that it can be reused later in this patch series. The implementation is not changed. Signed-off-by: Adam Litke <agl@us.ibm.com> Acked-by: Andy Whitcroft <apw@shadowen.org> Acked-by: Dave McCracken <dave.mccracken@oracle.com> Acked-by: William Irwin <bill.irwin@oracle.com> Cc: David Gibson <david@gibson.dropbear.id.au> Cc: Ken Chen <kenchen@google.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:26:16 +00:00
static void update_and_free_page(struct page *page)
{
int i;
nr_huge_pages--;
nr_huge_pages_node[page_to_nid(page)]--;
for (i = 0; i < (HPAGE_SIZE / PAGE_SIZE); i++) {
page[i].flags &= ~(1 << PG_locked | 1 << PG_error | 1 << PG_referenced |
1 << PG_dirty | 1 << PG_active | 1 << PG_reserved |
1 << PG_private | 1<< PG_writeback);
}
set_compound_page_dtor(page, NULL);
set_page_refcounted(page);
__free_pages(page, HUGETLB_PAGE_ORDER);
}
static void free_huge_page(struct page *page)
{
int nid = page_to_nid(page);
struct address_space *mapping;
mapping = (struct address_space *) page_private(page);
BUG_ON(page_count(page));
INIT_LIST_HEAD(&page->lru);
spin_lock(&hugetlb_lock);
if (surplus_huge_pages_node[nid]) {
update_and_free_page(page);
surplus_huge_pages--;
surplus_huge_pages_node[nid]--;
} else {
enqueue_huge_page(page);
}
spin_unlock(&hugetlb_lock);
if (mapping)
hugetlb_put_quota(mapping, 1);
set_page_private(page, 0);
}
/*
* Increment or decrement surplus_huge_pages. Keep node-specific counters
* balanced by operating on them in a round-robin fashion.
* Returns 1 if an adjustment was made.
*/
static int adjust_pool_surplus(int delta)
{
static int prev_nid;
int nid = prev_nid;
int ret = 0;
VM_BUG_ON(delta != -1 && delta != 1);
do {
nid = next_node(nid, node_online_map);
if (nid == MAX_NUMNODES)
nid = first_node(node_online_map);
/* To shrink on this node, there must be a surplus page */
if (delta < 0 && !surplus_huge_pages_node[nid])
continue;
/* Surplus cannot exceed the total number of pages */
if (delta > 0 && surplus_huge_pages_node[nid] >=
nr_huge_pages_node[nid])
continue;
surplus_huge_pages += delta;
surplus_huge_pages_node[nid] += delta;
ret = 1;
break;
} while (nid != prev_nid);
prev_nid = nid;
return ret;
}
hugetlb: fix hugepage allocation with memoryless nodes Anton found a problem with the hugetlb pool allocation when some nodes have no memory (http://marc.info/?l=linux-mm&m=118133042025995&w=2). Lee worked on versions that tried to fix it, but none were accepted. Christoph has created a set of patches which allow for GFP_THISNODE allocations to fail if the node has no memory. Currently, alloc_fresh_huge_page() returns NULL when it is not able to allocate a huge page on the current node, as specified by its custom interleave variable. The callers of this function, though, assume that a failure in alloc_fresh_huge_page() indicates no hugepages can be allocated on the system period. This might not be the case, for instance, if we have an uneven NUMA system, and we happen to try to allocate a hugepage on a node with less memory and fail, while there is still plenty of free memory on the other nodes. To correct this, make alloc_fresh_huge_page() search through all online nodes before deciding no hugepages can be allocated. Add a helper function for actually allocating the hugepage. Use a new global nid iterator to control which nid to allocate on. Note: we expect particular semantics for __GFP_THISNODE, which are now enforced even for memoryless nodes. That is, there is should be no fallback to other nodes. Therefore, we rely on the nid passed into alloc_pages_node() to be the nid the page comes from. If this is incorrect, accounting will break. Tested on x86 !NUMA, x86 NUMA, x86_64 NUMA and ppc64 NUMA (with 2 memoryless nodes). Before on the ppc64 box: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 25 Node 1 HugePages_Free: 75 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 150 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free After: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 50 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 100 Node 1 HugePages_Free: 100 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Christoph Lameter <clameter@sgi.com> Cc: Adam Litke <agl@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:26:24 +00:00
static struct page *alloc_fresh_huge_page_node(int nid)
{
struct page *page;
hugetlb: fix hugepage allocation with memoryless nodes Anton found a problem with the hugetlb pool allocation when some nodes have no memory (http://marc.info/?l=linux-mm&m=118133042025995&w=2). Lee worked on versions that tried to fix it, but none were accepted. Christoph has created a set of patches which allow for GFP_THISNODE allocations to fail if the node has no memory. Currently, alloc_fresh_huge_page() returns NULL when it is not able to allocate a huge page on the current node, as specified by its custom interleave variable. The callers of this function, though, assume that a failure in alloc_fresh_huge_page() indicates no hugepages can be allocated on the system period. This might not be the case, for instance, if we have an uneven NUMA system, and we happen to try to allocate a hugepage on a node with less memory and fail, while there is still plenty of free memory on the other nodes. To correct this, make alloc_fresh_huge_page() search through all online nodes before deciding no hugepages can be allocated. Add a helper function for actually allocating the hugepage. Use a new global nid iterator to control which nid to allocate on. Note: we expect particular semantics for __GFP_THISNODE, which are now enforced even for memoryless nodes. That is, there is should be no fallback to other nodes. Therefore, we rely on the nid passed into alloc_pages_node() to be the nid the page comes from. If this is incorrect, accounting will break. Tested on x86 !NUMA, x86 NUMA, x86_64 NUMA and ppc64 NUMA (with 2 memoryless nodes). Before on the ppc64 box: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 25 Node 1 HugePages_Free: 75 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 150 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free After: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 50 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 100 Node 1 HugePages_Free: 100 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Christoph Lameter <clameter@sgi.com> Cc: Adam Litke <agl@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:26:24 +00:00
page = alloc_pages_node(nid,
htlb_alloc_mask|__GFP_COMP|__GFP_THISNODE|__GFP_NOWARN,
HUGETLB_PAGE_ORDER);
if (page) {
set_compound_page_dtor(page, free_huge_page);
spin_lock(&hugetlb_lock);
nr_huge_pages++;
hugetlb: fix hugepage allocation with memoryless nodes Anton found a problem with the hugetlb pool allocation when some nodes have no memory (http://marc.info/?l=linux-mm&m=118133042025995&w=2). Lee worked on versions that tried to fix it, but none were accepted. Christoph has created a set of patches which allow for GFP_THISNODE allocations to fail if the node has no memory. Currently, alloc_fresh_huge_page() returns NULL when it is not able to allocate a huge page on the current node, as specified by its custom interleave variable. The callers of this function, though, assume that a failure in alloc_fresh_huge_page() indicates no hugepages can be allocated on the system period. This might not be the case, for instance, if we have an uneven NUMA system, and we happen to try to allocate a hugepage on a node with less memory and fail, while there is still plenty of free memory on the other nodes. To correct this, make alloc_fresh_huge_page() search through all online nodes before deciding no hugepages can be allocated. Add a helper function for actually allocating the hugepage. Use a new global nid iterator to control which nid to allocate on. Note: we expect particular semantics for __GFP_THISNODE, which are now enforced even for memoryless nodes. That is, there is should be no fallback to other nodes. Therefore, we rely on the nid passed into alloc_pages_node() to be the nid the page comes from. If this is incorrect, accounting will break. Tested on x86 !NUMA, x86 NUMA, x86_64 NUMA and ppc64 NUMA (with 2 memoryless nodes). Before on the ppc64 box: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 25 Node 1 HugePages_Free: 75 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 150 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free After: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 50 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 100 Node 1 HugePages_Free: 100 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Christoph Lameter <clameter@sgi.com> Cc: Adam Litke <agl@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:26:24 +00:00
nr_huge_pages_node[nid]++;
spin_unlock(&hugetlb_lock);
put_page(page); /* free it into the hugepage allocator */
}
hugetlb: fix hugepage allocation with memoryless nodes Anton found a problem with the hugetlb pool allocation when some nodes have no memory (http://marc.info/?l=linux-mm&m=118133042025995&w=2). Lee worked on versions that tried to fix it, but none were accepted. Christoph has created a set of patches which allow for GFP_THISNODE allocations to fail if the node has no memory. Currently, alloc_fresh_huge_page() returns NULL when it is not able to allocate a huge page on the current node, as specified by its custom interleave variable. The callers of this function, though, assume that a failure in alloc_fresh_huge_page() indicates no hugepages can be allocated on the system period. This might not be the case, for instance, if we have an uneven NUMA system, and we happen to try to allocate a hugepage on a node with less memory and fail, while there is still plenty of free memory on the other nodes. To correct this, make alloc_fresh_huge_page() search through all online nodes before deciding no hugepages can be allocated. Add a helper function for actually allocating the hugepage. Use a new global nid iterator to control which nid to allocate on. Note: we expect particular semantics for __GFP_THISNODE, which are now enforced even for memoryless nodes. That is, there is should be no fallback to other nodes. Therefore, we rely on the nid passed into alloc_pages_node() to be the nid the page comes from. If this is incorrect, accounting will break. Tested on x86 !NUMA, x86 NUMA, x86_64 NUMA and ppc64 NUMA (with 2 memoryless nodes). Before on the ppc64 box: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 25 Node 1 HugePages_Free: 75 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 150 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free After: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 50 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 100 Node 1 HugePages_Free: 100 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Christoph Lameter <clameter@sgi.com> Cc: Adam Litke <agl@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:26:24 +00:00
return page;
}
static int alloc_fresh_huge_page(void)
{
struct page *page;
int start_nid;
int next_nid;
int ret = 0;
start_nid = hugetlb_next_nid;
do {
page = alloc_fresh_huge_page_node(hugetlb_next_nid);
if (page)
ret = 1;
/*
* Use a helper variable to find the next node and then
* copy it back to hugetlb_next_nid afterwards:
* otherwise there's a window in which a racer might
* pass invalid nid MAX_NUMNODES to alloc_pages_node.
* But we don't need to use a spin_lock here: it really
* doesn't matter if occasionally a racer chooses the
* same nid as we do. Move nid forward in the mask even
* if we just successfully allocated a hugepage so that
* the next caller gets hugepages on the next node.
*/
next_nid = next_node(hugetlb_next_nid, node_online_map);
if (next_nid == MAX_NUMNODES)
next_nid = first_node(node_online_map);
hugetlb_next_nid = next_nid;
} while (!page && hugetlb_next_nid != start_nid);
return ret;
}
static struct page *alloc_buddy_huge_page(struct vm_area_struct *vma,
unsigned long address)
{
struct page *page;
hugetlb: introduce nr_overcommit_hugepages sysctl hugetlb: introduce nr_overcommit_hugepages sysctl While examining the code to support /proc/sys/vm/hugetlb_dynamic_pool, I became convinced that having a boolean sysctl was insufficient: 1) To support per-node control of hugepages, I have previously submitted patches to add a sysfs attribute related to nr_hugepages. However, with a boolean global value and per-mount quota enforcement constraining the dynamic pool, adding corresponding control of the dynamic pool on a per-node basis seems inconsistent to me. 2) Administration of the hugetlb dynamic pool with multiple hugetlbfs mount points is, arguably, more arduous than it needs to be. Each quota would need to be set separately, and the sum would need to be monitored. To ease the administration, and to help make the way for per-node control of the static & dynamic hugepage pool, I added a separate sysctl, nr_overcommit_hugepages. This value serves as a high watermark for the overall hugepage pool, while nr_hugepages serves as a low watermark. The boolean sysctl can then be removed, as the condition nr_overcommit_hugepages > 0 indicates the same administrative setting as hugetlb_dynamic_pool == 1 Quotas still serve as local enforcement of the size of the pool on a per-mount basis. A few caveats: 1) There is a race whereby the global surplus huge page counter is incremented before a hugepage has allocated. Another process could then try grow the pool, and fail to convert a surplus huge page to a normal huge page and instead allocate a fresh huge page. I believe this is benign, as no memory is leaked (the actual pages are still tracked correctly) and the counters won't go out of sync. 2) Shrinking the static pool while a surplus is in effect will allow the number of surplus huge pages to exceed the overcommit value. As long as this condition holds, however, no more surplus huge pages will be allowed on the system until one of the two sysctls are increased sufficiently, or the surplus huge pages go out of use and are freed. Successfully tested on x86_64 with the current libhugetlbfs snapshot, modified to use the new sysctl. Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Adam Litke <agl@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-12-18 00:20:12 +00:00
unsigned int nid;
hugetlb: introduce nr_overcommit_hugepages sysctl hugetlb: introduce nr_overcommit_hugepages sysctl While examining the code to support /proc/sys/vm/hugetlb_dynamic_pool, I became convinced that having a boolean sysctl was insufficient: 1) To support per-node control of hugepages, I have previously submitted patches to add a sysfs attribute related to nr_hugepages. However, with a boolean global value and per-mount quota enforcement constraining the dynamic pool, adding corresponding control of the dynamic pool on a per-node basis seems inconsistent to me. 2) Administration of the hugetlb dynamic pool with multiple hugetlbfs mount points is, arguably, more arduous than it needs to be. Each quota would need to be set separately, and the sum would need to be monitored. To ease the administration, and to help make the way for per-node control of the static & dynamic hugepage pool, I added a separate sysctl, nr_overcommit_hugepages. This value serves as a high watermark for the overall hugepage pool, while nr_hugepages serves as a low watermark. The boolean sysctl can then be removed, as the condition nr_overcommit_hugepages > 0 indicates the same administrative setting as hugetlb_dynamic_pool == 1 Quotas still serve as local enforcement of the size of the pool on a per-mount basis. A few caveats: 1) There is a race whereby the global surplus huge page counter is incremented before a hugepage has allocated. Another process could then try grow the pool, and fail to convert a surplus huge page to a normal huge page and instead allocate a fresh huge page. I believe this is benign, as no memory is leaked (the actual pages are still tracked correctly) and the counters won't go out of sync. 2) Shrinking the static pool while a surplus is in effect will allow the number of surplus huge pages to exceed the overcommit value. As long as this condition holds, however, no more surplus huge pages will be allowed on the system until one of the two sysctls are increased sufficiently, or the surplus huge pages go out of use and are freed. Successfully tested on x86_64 with the current libhugetlbfs snapshot, modified to use the new sysctl. Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Adam Litke <agl@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-12-18 00:20:12 +00:00
/*
* Assume we will successfully allocate the surplus page to
* prevent racing processes from causing the surplus to exceed
* overcommit
*
* This however introduces a different race, where a process B
* tries to grow the static hugepage pool while alloc_pages() is
* called by process A. B will only examine the per-node
* counters in determining if surplus huge pages can be
* converted to normal huge pages in adjust_pool_surplus(). A
* won't be able to increment the per-node counter, until the
* lock is dropped by B, but B doesn't drop hugetlb_lock until
* no more huge pages can be converted from surplus to normal
* state (and doesn't try to convert again). Thus, we have a
* case where a surplus huge page exists, the pool is grown, and
* the surplus huge page still exists after, even though it
* should just have been converted to a normal huge page. This
* does not leak memory, though, as the hugepage will be freed
* once it is out of use. It also does not allow the counters to
* go out of whack in adjust_pool_surplus() as we don't modify
* the node values until we've gotten the hugepage and only the
* per-node value is checked there.
*/
spin_lock(&hugetlb_lock);
if (surplus_huge_pages >= nr_overcommit_huge_pages) {
spin_unlock(&hugetlb_lock);
return NULL;
} else {
nr_huge_pages++;
surplus_huge_pages++;
}
spin_unlock(&hugetlb_lock);
page = alloc_pages(htlb_alloc_mask|__GFP_COMP|__GFP_NOWARN,
HUGETLB_PAGE_ORDER);
hugetlb: introduce nr_overcommit_hugepages sysctl hugetlb: introduce nr_overcommit_hugepages sysctl While examining the code to support /proc/sys/vm/hugetlb_dynamic_pool, I became convinced that having a boolean sysctl was insufficient: 1) To support per-node control of hugepages, I have previously submitted patches to add a sysfs attribute related to nr_hugepages. However, with a boolean global value and per-mount quota enforcement constraining the dynamic pool, adding corresponding control of the dynamic pool on a per-node basis seems inconsistent to me. 2) Administration of the hugetlb dynamic pool with multiple hugetlbfs mount points is, arguably, more arduous than it needs to be. Each quota would need to be set separately, and the sum would need to be monitored. To ease the administration, and to help make the way for per-node control of the static & dynamic hugepage pool, I added a separate sysctl, nr_overcommit_hugepages. This value serves as a high watermark for the overall hugepage pool, while nr_hugepages serves as a low watermark. The boolean sysctl can then be removed, as the condition nr_overcommit_hugepages > 0 indicates the same administrative setting as hugetlb_dynamic_pool == 1 Quotas still serve as local enforcement of the size of the pool on a per-mount basis. A few caveats: 1) There is a race whereby the global surplus huge page counter is incremented before a hugepage has allocated. Another process could then try grow the pool, and fail to convert a surplus huge page to a normal huge page and instead allocate a fresh huge page. I believe this is benign, as no memory is leaked (the actual pages are still tracked correctly) and the counters won't go out of sync. 2) Shrinking the static pool while a surplus is in effect will allow the number of surplus huge pages to exceed the overcommit value. As long as this condition holds, however, no more surplus huge pages will be allowed on the system until one of the two sysctls are increased sufficiently, or the surplus huge pages go out of use and are freed. Successfully tested on x86_64 with the current libhugetlbfs snapshot, modified to use the new sysctl. Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Adam Litke <agl@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-12-18 00:20:12 +00:00
spin_lock(&hugetlb_lock);
if (page) {
hugetlb: introduce nr_overcommit_hugepages sysctl hugetlb: introduce nr_overcommit_hugepages sysctl While examining the code to support /proc/sys/vm/hugetlb_dynamic_pool, I became convinced that having a boolean sysctl was insufficient: 1) To support per-node control of hugepages, I have previously submitted patches to add a sysfs attribute related to nr_hugepages. However, with a boolean global value and per-mount quota enforcement constraining the dynamic pool, adding corresponding control of the dynamic pool on a per-node basis seems inconsistent to me. 2) Administration of the hugetlb dynamic pool with multiple hugetlbfs mount points is, arguably, more arduous than it needs to be. Each quota would need to be set separately, and the sum would need to be monitored. To ease the administration, and to help make the way for per-node control of the static & dynamic hugepage pool, I added a separate sysctl, nr_overcommit_hugepages. This value serves as a high watermark for the overall hugepage pool, while nr_hugepages serves as a low watermark. The boolean sysctl can then be removed, as the condition nr_overcommit_hugepages > 0 indicates the same administrative setting as hugetlb_dynamic_pool == 1 Quotas still serve as local enforcement of the size of the pool on a per-mount basis. A few caveats: 1) There is a race whereby the global surplus huge page counter is incremented before a hugepage has allocated. Another process could then try grow the pool, and fail to convert a surplus huge page to a normal huge page and instead allocate a fresh huge page. I believe this is benign, as no memory is leaked (the actual pages are still tracked correctly) and the counters won't go out of sync. 2) Shrinking the static pool while a surplus is in effect will allow the number of surplus huge pages to exceed the overcommit value. As long as this condition holds, however, no more surplus huge pages will be allowed on the system until one of the two sysctls are increased sufficiently, or the surplus huge pages go out of use and are freed. Successfully tested on x86_64 with the current libhugetlbfs snapshot, modified to use the new sysctl. Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Adam Litke <agl@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-12-18 00:20:12 +00:00
nid = page_to_nid(page);
set_compound_page_dtor(page, free_huge_page);
hugetlb: introduce nr_overcommit_hugepages sysctl hugetlb: introduce nr_overcommit_hugepages sysctl While examining the code to support /proc/sys/vm/hugetlb_dynamic_pool, I became convinced that having a boolean sysctl was insufficient: 1) To support per-node control of hugepages, I have previously submitted patches to add a sysfs attribute related to nr_hugepages. However, with a boolean global value and per-mount quota enforcement constraining the dynamic pool, adding corresponding control of the dynamic pool on a per-node basis seems inconsistent to me. 2) Administration of the hugetlb dynamic pool with multiple hugetlbfs mount points is, arguably, more arduous than it needs to be. Each quota would need to be set separately, and the sum would need to be monitored. To ease the administration, and to help make the way for per-node control of the static & dynamic hugepage pool, I added a separate sysctl, nr_overcommit_hugepages. This value serves as a high watermark for the overall hugepage pool, while nr_hugepages serves as a low watermark. The boolean sysctl can then be removed, as the condition nr_overcommit_hugepages > 0 indicates the same administrative setting as hugetlb_dynamic_pool == 1 Quotas still serve as local enforcement of the size of the pool on a per-mount basis. A few caveats: 1) There is a race whereby the global surplus huge page counter is incremented before a hugepage has allocated. Another process could then try grow the pool, and fail to convert a surplus huge page to a normal huge page and instead allocate a fresh huge page. I believe this is benign, as no memory is leaked (the actual pages are still tracked correctly) and the counters won't go out of sync. 2) Shrinking the static pool while a surplus is in effect will allow the number of surplus huge pages to exceed the overcommit value. As long as this condition holds, however, no more surplus huge pages will be allowed on the system until one of the two sysctls are increased sufficiently, or the surplus huge pages go out of use and are freed. Successfully tested on x86_64 with the current libhugetlbfs snapshot, modified to use the new sysctl. Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Adam Litke <agl@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-12-18 00:20:12 +00:00
/*
* We incremented the global counters already
*/
nr_huge_pages_node[nid]++;
surplus_huge_pages_node[nid]++;
} else {
nr_huge_pages--;
surplus_huge_pages--;
}
hugetlb: introduce nr_overcommit_hugepages sysctl hugetlb: introduce nr_overcommit_hugepages sysctl While examining the code to support /proc/sys/vm/hugetlb_dynamic_pool, I became convinced that having a boolean sysctl was insufficient: 1) To support per-node control of hugepages, I have previously submitted patches to add a sysfs attribute related to nr_hugepages. However, with a boolean global value and per-mount quota enforcement constraining the dynamic pool, adding corresponding control of the dynamic pool on a per-node basis seems inconsistent to me. 2) Administration of the hugetlb dynamic pool with multiple hugetlbfs mount points is, arguably, more arduous than it needs to be. Each quota would need to be set separately, and the sum would need to be monitored. To ease the administration, and to help make the way for per-node control of the static & dynamic hugepage pool, I added a separate sysctl, nr_overcommit_hugepages. This value serves as a high watermark for the overall hugepage pool, while nr_hugepages serves as a low watermark. The boolean sysctl can then be removed, as the condition nr_overcommit_hugepages > 0 indicates the same administrative setting as hugetlb_dynamic_pool == 1 Quotas still serve as local enforcement of the size of the pool on a per-mount basis. A few caveats: 1) There is a race whereby the global surplus huge page counter is incremented before a hugepage has allocated. Another process could then try grow the pool, and fail to convert a surplus huge page to a normal huge page and instead allocate a fresh huge page. I believe this is benign, as no memory is leaked (the actual pages are still tracked correctly) and the counters won't go out of sync. 2) Shrinking the static pool while a surplus is in effect will allow the number of surplus huge pages to exceed the overcommit value. As long as this condition holds, however, no more surplus huge pages will be allowed on the system until one of the two sysctls are increased sufficiently, or the surplus huge pages go out of use and are freed. Successfully tested on x86_64 with the current libhugetlbfs snapshot, modified to use the new sysctl. Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Adam Litke <agl@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-12-18 00:20:12 +00:00
spin_unlock(&hugetlb_lock);
return page;
}
/*
* Increase the hugetlb pool such that it can accomodate a reservation
* of size 'delta'.
*/
static int gather_surplus_pages(int delta)
{
struct list_head surplus_list;
struct page *page, *tmp;
int ret, i;
int needed, allocated;
needed = (resv_huge_pages + delta) - free_huge_pages;
if (needed <= 0)
return 0;
allocated = 0;
INIT_LIST_HEAD(&surplus_list);
ret = -ENOMEM;
retry:
spin_unlock(&hugetlb_lock);
for (i = 0; i < needed; i++) {
page = alloc_buddy_huge_page(NULL, 0);
if (!page) {
/*
* We were not able to allocate enough pages to
* satisfy the entire reservation so we free what
* we've allocated so far.
*/
spin_lock(&hugetlb_lock);
needed = 0;
goto free;
}
list_add(&page->lru, &surplus_list);
}
allocated += needed;
/*
* After retaking hugetlb_lock, we need to recalculate 'needed'
* because either resv_huge_pages or free_huge_pages may have changed.
*/
spin_lock(&hugetlb_lock);
needed = (resv_huge_pages + delta) - (free_huge_pages + allocated);
if (needed > 0)
goto retry;
/*
* The surplus_list now contains _at_least_ the number of extra pages
* needed to accomodate the reservation. Add the appropriate number
* of pages to the hugetlb pool and free the extras back to the buddy
* allocator.
*/
needed += allocated;
ret = 0;
free:
list_for_each_entry_safe(page, tmp, &surplus_list, lru) {
list_del(&page->lru);
if ((--needed) >= 0)
enqueue_huge_page(page);
else {
/*
* Decrement the refcount and free the page using its
* destructor. This must be done with hugetlb_lock
* unlocked which is safe because free_huge_page takes
* hugetlb_lock before deciding how to free the page.
*/
spin_unlock(&hugetlb_lock);
put_page(page);
spin_lock(&hugetlb_lock);
}
}
return ret;
}
/*
* When releasing a hugetlb pool reservation, any surplus pages that were
* allocated to satisfy the reservation must be explicitly freed if they were
* never used.
*/
static void return_unused_surplus_pages(unsigned long unused_resv_pages)
{
static int nid = -1;
struct page *page;
unsigned long nr_pages;
nr_pages = min(unused_resv_pages, surplus_huge_pages);
while (nr_pages) {
nid = next_node(nid, node_online_map);
if (nid == MAX_NUMNODES)
nid = first_node(node_online_map);
if (!surplus_huge_pages_node[nid])
continue;
if (!list_empty(&hugepage_freelists[nid])) {
page = list_entry(hugepage_freelists[nid].next,
struct page, lru);
list_del(&page->lru);
update_and_free_page(page);
free_huge_pages--;
free_huge_pages_node[nid]--;
surplus_huge_pages--;
surplus_huge_pages_node[nid]--;
nr_pages--;
}
}
}
hugetlb: split alloc_huge_page into private and shared components Hugetlbfs implements a quota system which can limit the amount of memory that can be used by the filesystem. Before allocating a new huge page for a file, the quota is checked and debited. The quota is then credited when truncating the file. I found a few bugs in the code for both MAP_PRIVATE and MAP_SHARED mappings. Before detailing the problems and my proposed solutions, we should agree on a definition of quotas that properly addresses both private and shared pages. Since the purpose of quotas is to limit total memory consumption on a per-filesystem basis, I argue that all pages allocated by the fs (private and shared) should be charged against quota. Private Mappings ================ The current code will debit quota for private pages sometimes, but will never credit it. At a minimum, this causes a leak in the quota accounting which renders the accounting essentially useless as it is. Shared pages have a one to one mapping with a hugetlbfs file and are easy to account by debiting on allocation and crediting on truncate. Private pages are anonymous in nature and have a many to one relationship with their hugetlbfs files (due to copy on write). Because private pages are not indexed by the mapping's radix tree, thier quota cannot be credited at file truncation time. Crediting must be done when the page is unmapped and freed. Shared Pages ============ I discovered an issue concerning the interaction between the MAP_SHARED reservation system and quotas. Since quota is not checked until page instantiation, an over-quota mmap/reservation will initially succeed. When instantiating the first over-quota page, the program will receive SIGBUS. This is inconsistent since the reservation is supposed to be a guarantee. The solution is to debit the full amount of quota at reservation time and credit the unused portion when the reservation is released. This patch series brings quotas back in line by making the following modifications: * Private pages - Debit quota in alloc_huge_page() - Credit quota in free_huge_page() * Shared pages - Debit quota for entire reservation at mmap time - Credit quota for instantiated pages in free_huge_page() - Credit quota for unused reservation at munmap time This patch: The shared page reservation and dynamic pool resizing features have made the allocation of private vs. shared huge pages quite different. By splitting out the private/shared-specific portions of the process into their own functions, readability is greatly improved. alloc_huge_page now calls the proper helper and performs common operations. [akpm@linux-foundation.org: coding-style cleanups] Signed-off-by: Adam Litke <agl@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-11-15 00:59:37 +00:00
static struct page *alloc_huge_page_shared(struct vm_area_struct *vma,
unsigned long addr)
{
hugetlb: split alloc_huge_page into private and shared components Hugetlbfs implements a quota system which can limit the amount of memory that can be used by the filesystem. Before allocating a new huge page for a file, the quota is checked and debited. The quota is then credited when truncating the file. I found a few bugs in the code for both MAP_PRIVATE and MAP_SHARED mappings. Before detailing the problems and my proposed solutions, we should agree on a definition of quotas that properly addresses both private and shared pages. Since the purpose of quotas is to limit total memory consumption on a per-filesystem basis, I argue that all pages allocated by the fs (private and shared) should be charged against quota. Private Mappings ================ The current code will debit quota for private pages sometimes, but will never credit it. At a minimum, this causes a leak in the quota accounting which renders the accounting essentially useless as it is. Shared pages have a one to one mapping with a hugetlbfs file and are easy to account by debiting on allocation and crediting on truncate. Private pages are anonymous in nature and have a many to one relationship with their hugetlbfs files (due to copy on write). Because private pages are not indexed by the mapping's radix tree, thier quota cannot be credited at file truncation time. Crediting must be done when the page is unmapped and freed. Shared Pages ============ I discovered an issue concerning the interaction between the MAP_SHARED reservation system and quotas. Since quota is not checked until page instantiation, an over-quota mmap/reservation will initially succeed. When instantiating the first over-quota page, the program will receive SIGBUS. This is inconsistent since the reservation is supposed to be a guarantee. The solution is to debit the full amount of quota at reservation time and credit the unused portion when the reservation is released. This patch series brings quotas back in line by making the following modifications: * Private pages - Debit quota in alloc_huge_page() - Credit quota in free_huge_page() * Shared pages - Debit quota for entire reservation at mmap time - Credit quota for instantiated pages in free_huge_page() - Credit quota for unused reservation at munmap time This patch: The shared page reservation and dynamic pool resizing features have made the allocation of private vs. shared huge pages quite different. By splitting out the private/shared-specific portions of the process into their own functions, readability is greatly improved. alloc_huge_page now calls the proper helper and performs common operations. [akpm@linux-foundation.org: coding-style cleanups] Signed-off-by: Adam Litke <agl@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-11-15 00:59:37 +00:00
struct page *page;
spin_lock(&hugetlb_lock);
[PATCH] hugepage: Strict page reservation for hugepage inodes These days, hugepages are demand-allocated at first fault time. There's a somewhat dubious (and racy) heuristic when making a new mmap() to check if there are enough available hugepages to fully satisfy that mapping. A particularly obvious case where the heuristic breaks down is where a process maps its hugepages not as a single chunk, but as a bunch of individually mmap()ed (or shmat()ed) blocks without touching and instantiating the pages in between allocations. In this case the size of each block is compared against the total number of available hugepages. It's thus easy for the process to become overcommitted, because each block mapping will succeed, although the total number of hugepages required by all blocks exceeds the number available. In particular, this defeats such a program which will detect a mapping failure and adjust its hugepage usage downward accordingly. The patch below addresses this problem, by strictly reserving a number of physical hugepages for hugepage inodes which have been mapped, but not instatiated. MAP_SHARED mappings are thus "safe" - they will fail on mmap(), not later with an OOM SIGKILL. MAP_PRIVATE mappings can still trigger an OOM. (Actually SHARED mappings can technically still OOM, but only if the sysadmin explicitly reduces the hugepage pool between mapping and instantiation) This patch appears to address the problem at hand - it allows DB2 to start correctly, for instance, which previously suffered the failure described above. This patch causes no regressions on the libhugetblfs testsuite, and makes a test (designed to catch this problem) pass which previously failed (ppc64, POWER5). Signed-off-by: David Gibson <dwg@au1.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:55 +00:00
page = dequeue_huge_page(vma, addr);
spin_unlock(&hugetlb_lock);
return page ? page : ERR_PTR(-VM_FAULT_OOM);
hugetlb: split alloc_huge_page into private and shared components Hugetlbfs implements a quota system which can limit the amount of memory that can be used by the filesystem. Before allocating a new huge page for a file, the quota is checked and debited. The quota is then credited when truncating the file. I found a few bugs in the code for both MAP_PRIVATE and MAP_SHARED mappings. Before detailing the problems and my proposed solutions, we should agree on a definition of quotas that properly addresses both private and shared pages. Since the purpose of quotas is to limit total memory consumption on a per-filesystem basis, I argue that all pages allocated by the fs (private and shared) should be charged against quota. Private Mappings ================ The current code will debit quota for private pages sometimes, but will never credit it. At a minimum, this causes a leak in the quota accounting which renders the accounting essentially useless as it is. Shared pages have a one to one mapping with a hugetlbfs file and are easy to account by debiting on allocation and crediting on truncate. Private pages are anonymous in nature and have a many to one relationship with their hugetlbfs files (due to copy on write). Because private pages are not indexed by the mapping's radix tree, thier quota cannot be credited at file truncation time. Crediting must be done when the page is unmapped and freed. Shared Pages ============ I discovered an issue concerning the interaction between the MAP_SHARED reservation system and quotas. Since quota is not checked until page instantiation, an over-quota mmap/reservation will initially succeed. When instantiating the first over-quota page, the program will receive SIGBUS. This is inconsistent since the reservation is supposed to be a guarantee. The solution is to debit the full amount of quota at reservation time and credit the unused portion when the reservation is released. This patch series brings quotas back in line by making the following modifications: * Private pages - Debit quota in alloc_huge_page() - Credit quota in free_huge_page() * Shared pages - Debit quota for entire reservation at mmap time - Credit quota for instantiated pages in free_huge_page() - Credit quota for unused reservation at munmap time This patch: The shared page reservation and dynamic pool resizing features have made the allocation of private vs. shared huge pages quite different. By splitting out the private/shared-specific portions of the process into their own functions, readability is greatly improved. alloc_huge_page now calls the proper helper and performs common operations. [akpm@linux-foundation.org: coding-style cleanups] Signed-off-by: Adam Litke <agl@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-11-15 00:59:37 +00:00
}
[PATCH] hugepage: Strict page reservation for hugepage inodes These days, hugepages are demand-allocated at first fault time. There's a somewhat dubious (and racy) heuristic when making a new mmap() to check if there are enough available hugepages to fully satisfy that mapping. A particularly obvious case where the heuristic breaks down is where a process maps its hugepages not as a single chunk, but as a bunch of individually mmap()ed (or shmat()ed) blocks without touching and instantiating the pages in between allocations. In this case the size of each block is compared against the total number of available hugepages. It's thus easy for the process to become overcommitted, because each block mapping will succeed, although the total number of hugepages required by all blocks exceeds the number available. In particular, this defeats such a program which will detect a mapping failure and adjust its hugepage usage downward accordingly. The patch below addresses this problem, by strictly reserving a number of physical hugepages for hugepage inodes which have been mapped, but not instatiated. MAP_SHARED mappings are thus "safe" - they will fail on mmap(), not later with an OOM SIGKILL. MAP_PRIVATE mappings can still trigger an OOM. (Actually SHARED mappings can technically still OOM, but only if the sysadmin explicitly reduces the hugepage pool between mapping and instantiation) This patch appears to address the problem at hand - it allows DB2 to start correctly, for instance, which previously suffered the failure described above. This patch causes no regressions on the libhugetblfs testsuite, and makes a test (designed to catch this problem) pass which previously failed (ppc64, POWER5). Signed-off-by: David Gibson <dwg@au1.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:55 +00:00
hugetlb: split alloc_huge_page into private and shared components Hugetlbfs implements a quota system which can limit the amount of memory that can be used by the filesystem. Before allocating a new huge page for a file, the quota is checked and debited. The quota is then credited when truncating the file. I found a few bugs in the code for both MAP_PRIVATE and MAP_SHARED mappings. Before detailing the problems and my proposed solutions, we should agree on a definition of quotas that properly addresses both private and shared pages. Since the purpose of quotas is to limit total memory consumption on a per-filesystem basis, I argue that all pages allocated by the fs (private and shared) should be charged against quota. Private Mappings ================ The current code will debit quota for private pages sometimes, but will never credit it. At a minimum, this causes a leak in the quota accounting which renders the accounting essentially useless as it is. Shared pages have a one to one mapping with a hugetlbfs file and are easy to account by debiting on allocation and crediting on truncate. Private pages are anonymous in nature and have a many to one relationship with their hugetlbfs files (due to copy on write). Because private pages are not indexed by the mapping's radix tree, thier quota cannot be credited at file truncation time. Crediting must be done when the page is unmapped and freed. Shared Pages ============ I discovered an issue concerning the interaction between the MAP_SHARED reservation system and quotas. Since quota is not checked until page instantiation, an over-quota mmap/reservation will initially succeed. When instantiating the first over-quota page, the program will receive SIGBUS. This is inconsistent since the reservation is supposed to be a guarantee. The solution is to debit the full amount of quota at reservation time and credit the unused portion when the reservation is released. This patch series brings quotas back in line by making the following modifications: * Private pages - Debit quota in alloc_huge_page() - Credit quota in free_huge_page() * Shared pages - Debit quota for entire reservation at mmap time - Credit quota for instantiated pages in free_huge_page() - Credit quota for unused reservation at munmap time This patch: The shared page reservation and dynamic pool resizing features have made the allocation of private vs. shared huge pages quite different. By splitting out the private/shared-specific portions of the process into their own functions, readability is greatly improved. alloc_huge_page now calls the proper helper and performs common operations. [akpm@linux-foundation.org: coding-style cleanups] Signed-off-by: Adam Litke <agl@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-11-15 00:59:37 +00:00
static struct page *alloc_huge_page_private(struct vm_area_struct *vma,
unsigned long addr)
{
struct page *page = NULL;
if (hugetlb_get_quota(vma->vm_file->f_mapping, 1))
return ERR_PTR(-VM_FAULT_SIGBUS);
hugetlb: split alloc_huge_page into private and shared components Hugetlbfs implements a quota system which can limit the amount of memory that can be used by the filesystem. Before allocating a new huge page for a file, the quota is checked and debited. The quota is then credited when truncating the file. I found a few bugs in the code for both MAP_PRIVATE and MAP_SHARED mappings. Before detailing the problems and my proposed solutions, we should agree on a definition of quotas that properly addresses both private and shared pages. Since the purpose of quotas is to limit total memory consumption on a per-filesystem basis, I argue that all pages allocated by the fs (private and shared) should be charged against quota. Private Mappings ================ The current code will debit quota for private pages sometimes, but will never credit it. At a minimum, this causes a leak in the quota accounting which renders the accounting essentially useless as it is. Shared pages have a one to one mapping with a hugetlbfs file and are easy to account by debiting on allocation and crediting on truncate. Private pages are anonymous in nature and have a many to one relationship with their hugetlbfs files (due to copy on write). Because private pages are not indexed by the mapping's radix tree, thier quota cannot be credited at file truncation time. Crediting must be done when the page is unmapped and freed. Shared Pages ============ I discovered an issue concerning the interaction between the MAP_SHARED reservation system and quotas. Since quota is not checked until page instantiation, an over-quota mmap/reservation will initially succeed. When instantiating the first over-quota page, the program will receive SIGBUS. This is inconsistent since the reservation is supposed to be a guarantee. The solution is to debit the full amount of quota at reservation time and credit the unused portion when the reservation is released. This patch series brings quotas back in line by making the following modifications: * Private pages - Debit quota in alloc_huge_page() - Credit quota in free_huge_page() * Shared pages - Debit quota for entire reservation at mmap time - Credit quota for instantiated pages in free_huge_page() - Credit quota for unused reservation at munmap time This patch: The shared page reservation and dynamic pool resizing features have made the allocation of private vs. shared huge pages quite different. By splitting out the private/shared-specific portions of the process into their own functions, readability is greatly improved. alloc_huge_page now calls the proper helper and performs common operations. [akpm@linux-foundation.org: coding-style cleanups] Signed-off-by: Adam Litke <agl@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-11-15 00:59:37 +00:00
spin_lock(&hugetlb_lock);
if (free_huge_pages > resv_huge_pages)
page = dequeue_huge_page(vma, addr);
spin_unlock(&hugetlb_lock);
if (!page) {
page = alloc_buddy_huge_page(vma, addr);
if (!page) {
hugetlb_put_quota(vma->vm_file->f_mapping, 1);
return ERR_PTR(-VM_FAULT_OOM);
}
}
return page;
hugetlb: split alloc_huge_page into private and shared components Hugetlbfs implements a quota system which can limit the amount of memory that can be used by the filesystem. Before allocating a new huge page for a file, the quota is checked and debited. The quota is then credited when truncating the file. I found a few bugs in the code for both MAP_PRIVATE and MAP_SHARED mappings. Before detailing the problems and my proposed solutions, we should agree on a definition of quotas that properly addresses both private and shared pages. Since the purpose of quotas is to limit total memory consumption on a per-filesystem basis, I argue that all pages allocated by the fs (private and shared) should be charged against quota. Private Mappings ================ The current code will debit quota for private pages sometimes, but will never credit it. At a minimum, this causes a leak in the quota accounting which renders the accounting essentially useless as it is. Shared pages have a one to one mapping with a hugetlbfs file and are easy to account by debiting on allocation and crediting on truncate. Private pages are anonymous in nature and have a many to one relationship with their hugetlbfs files (due to copy on write). Because private pages are not indexed by the mapping's radix tree, thier quota cannot be credited at file truncation time. Crediting must be done when the page is unmapped and freed. Shared Pages ============ I discovered an issue concerning the interaction between the MAP_SHARED reservation system and quotas. Since quota is not checked until page instantiation, an over-quota mmap/reservation will initially succeed. When instantiating the first over-quota page, the program will receive SIGBUS. This is inconsistent since the reservation is supposed to be a guarantee. The solution is to debit the full amount of quota at reservation time and credit the unused portion when the reservation is released. This patch series brings quotas back in line by making the following modifications: * Private pages - Debit quota in alloc_huge_page() - Credit quota in free_huge_page() * Shared pages - Debit quota for entire reservation at mmap time - Credit quota for instantiated pages in free_huge_page() - Credit quota for unused reservation at munmap time This patch: The shared page reservation and dynamic pool resizing features have made the allocation of private vs. shared huge pages quite different. By splitting out the private/shared-specific portions of the process into their own functions, readability is greatly improved. alloc_huge_page now calls the proper helper and performs common operations. [akpm@linux-foundation.org: coding-style cleanups] Signed-off-by: Adam Litke <agl@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-11-15 00:59:37 +00:00
}
static struct page *alloc_huge_page(struct vm_area_struct *vma,
unsigned long addr)
{
struct page *page;
struct address_space *mapping = vma->vm_file->f_mapping;
hugetlb: split alloc_huge_page into private and shared components Hugetlbfs implements a quota system which can limit the amount of memory that can be used by the filesystem. Before allocating a new huge page for a file, the quota is checked and debited. The quota is then credited when truncating the file. I found a few bugs in the code for both MAP_PRIVATE and MAP_SHARED mappings. Before detailing the problems and my proposed solutions, we should agree on a definition of quotas that properly addresses both private and shared pages. Since the purpose of quotas is to limit total memory consumption on a per-filesystem basis, I argue that all pages allocated by the fs (private and shared) should be charged against quota. Private Mappings ================ The current code will debit quota for private pages sometimes, but will never credit it. At a minimum, this causes a leak in the quota accounting which renders the accounting essentially useless as it is. Shared pages have a one to one mapping with a hugetlbfs file and are easy to account by debiting on allocation and crediting on truncate. Private pages are anonymous in nature and have a many to one relationship with their hugetlbfs files (due to copy on write). Because private pages are not indexed by the mapping's radix tree, thier quota cannot be credited at file truncation time. Crediting must be done when the page is unmapped and freed. Shared Pages ============ I discovered an issue concerning the interaction between the MAP_SHARED reservation system and quotas. Since quota is not checked until page instantiation, an over-quota mmap/reservation will initially succeed. When instantiating the first over-quota page, the program will receive SIGBUS. This is inconsistent since the reservation is supposed to be a guarantee. The solution is to debit the full amount of quota at reservation time and credit the unused portion when the reservation is released. This patch series brings quotas back in line by making the following modifications: * Private pages - Debit quota in alloc_huge_page() - Credit quota in free_huge_page() * Shared pages - Debit quota for entire reservation at mmap time - Credit quota for instantiated pages in free_huge_page() - Credit quota for unused reservation at munmap time This patch: The shared page reservation and dynamic pool resizing features have made the allocation of private vs. shared huge pages quite different. By splitting out the private/shared-specific portions of the process into their own functions, readability is greatly improved. alloc_huge_page now calls the proper helper and performs common operations. [akpm@linux-foundation.org: coding-style cleanups] Signed-off-by: Adam Litke <agl@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-11-15 00:59:37 +00:00
if (vma->vm_flags & VM_MAYSHARE)
page = alloc_huge_page_shared(vma, addr);
else
page = alloc_huge_page_private(vma, addr);
if (!IS_ERR(page)) {
hugetlb: split alloc_huge_page into private and shared components Hugetlbfs implements a quota system which can limit the amount of memory that can be used by the filesystem. Before allocating a new huge page for a file, the quota is checked and debited. The quota is then credited when truncating the file. I found a few bugs in the code for both MAP_PRIVATE and MAP_SHARED mappings. Before detailing the problems and my proposed solutions, we should agree on a definition of quotas that properly addresses both private and shared pages. Since the purpose of quotas is to limit total memory consumption on a per-filesystem basis, I argue that all pages allocated by the fs (private and shared) should be charged against quota. Private Mappings ================ The current code will debit quota for private pages sometimes, but will never credit it. At a minimum, this causes a leak in the quota accounting which renders the accounting essentially useless as it is. Shared pages have a one to one mapping with a hugetlbfs file and are easy to account by debiting on allocation and crediting on truncate. Private pages are anonymous in nature and have a many to one relationship with their hugetlbfs files (due to copy on write). Because private pages are not indexed by the mapping's radix tree, thier quota cannot be credited at file truncation time. Crediting must be done when the page is unmapped and freed. Shared Pages ============ I discovered an issue concerning the interaction between the MAP_SHARED reservation system and quotas. Since quota is not checked until page instantiation, an over-quota mmap/reservation will initially succeed. When instantiating the first over-quota page, the program will receive SIGBUS. This is inconsistent since the reservation is supposed to be a guarantee. The solution is to debit the full amount of quota at reservation time and credit the unused portion when the reservation is released. This patch series brings quotas back in line by making the following modifications: * Private pages - Debit quota in alloc_huge_page() - Credit quota in free_huge_page() * Shared pages - Debit quota for entire reservation at mmap time - Credit quota for instantiated pages in free_huge_page() - Credit quota for unused reservation at munmap time This patch: The shared page reservation and dynamic pool resizing features have made the allocation of private vs. shared huge pages quite different. By splitting out the private/shared-specific portions of the process into their own functions, readability is greatly improved. alloc_huge_page now calls the proper helper and performs common operations. [akpm@linux-foundation.org: coding-style cleanups] Signed-off-by: Adam Litke <agl@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: Andy Whitcroft <apw@shadowen.org> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-11-15 00:59:37 +00:00
set_page_refcounted(page);
set_page_private(page, (unsigned long) mapping);
}
return page;
[PATCH] hugepage: Strict page reservation for hugepage inodes These days, hugepages are demand-allocated at first fault time. There's a somewhat dubious (and racy) heuristic when making a new mmap() to check if there are enough available hugepages to fully satisfy that mapping. A particularly obvious case where the heuristic breaks down is where a process maps its hugepages not as a single chunk, but as a bunch of individually mmap()ed (or shmat()ed) blocks without touching and instantiating the pages in between allocations. In this case the size of each block is compared against the total number of available hugepages. It's thus easy for the process to become overcommitted, because each block mapping will succeed, although the total number of hugepages required by all blocks exceeds the number available. In particular, this defeats such a program which will detect a mapping failure and adjust its hugepage usage downward accordingly. The patch below addresses this problem, by strictly reserving a number of physical hugepages for hugepage inodes which have been mapped, but not instatiated. MAP_SHARED mappings are thus "safe" - they will fail on mmap(), not later with an OOM SIGKILL. MAP_PRIVATE mappings can still trigger an OOM. (Actually SHARED mappings can technically still OOM, but only if the sysadmin explicitly reduces the hugepage pool between mapping and instantiation) This patch appears to address the problem at hand - it allows DB2 to start correctly, for instance, which previously suffered the failure described above. This patch causes no regressions on the libhugetblfs testsuite, and makes a test (designed to catch this problem) pass which previously failed (ppc64, POWER5). Signed-off-by: David Gibson <dwg@au1.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:55 +00:00
}
static int __init hugetlb_init(void)
{
unsigned long i;
if (HPAGE_SHIFT == 0)
return 0;
for (i = 0; i < MAX_NUMNODES; ++i)
INIT_LIST_HEAD(&hugepage_freelists[i]);
hugetlb: fix hugepage allocation with memoryless nodes Anton found a problem with the hugetlb pool allocation when some nodes have no memory (http://marc.info/?l=linux-mm&m=118133042025995&w=2). Lee worked on versions that tried to fix it, but none were accepted. Christoph has created a set of patches which allow for GFP_THISNODE allocations to fail if the node has no memory. Currently, alloc_fresh_huge_page() returns NULL when it is not able to allocate a huge page on the current node, as specified by its custom interleave variable. The callers of this function, though, assume that a failure in alloc_fresh_huge_page() indicates no hugepages can be allocated on the system period. This might not be the case, for instance, if we have an uneven NUMA system, and we happen to try to allocate a hugepage on a node with less memory and fail, while there is still plenty of free memory on the other nodes. To correct this, make alloc_fresh_huge_page() search through all online nodes before deciding no hugepages can be allocated. Add a helper function for actually allocating the hugepage. Use a new global nid iterator to control which nid to allocate on. Note: we expect particular semantics for __GFP_THISNODE, which are now enforced even for memoryless nodes. That is, there is should be no fallback to other nodes. Therefore, we rely on the nid passed into alloc_pages_node() to be the nid the page comes from. If this is incorrect, accounting will break. Tested on x86 !NUMA, x86 NUMA, x86_64 NUMA and ppc64 NUMA (with 2 memoryless nodes). Before on the ppc64 box: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 25 Node 1 HugePages_Free: 75 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 150 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free After: Trying to clear the hugetlb pool Done. 0 free Trying to resize the pool to 100 Node 0 HugePages_Free: 50 Node 1 HugePages_Free: 50 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. Initially 100 free Trying to resize the pool to 200 Node 0 HugePages_Free: 100 Node 1 HugePages_Free: 100 Node 2 HugePages_Free: 0 Node 3 HugePages_Free: 0 Done. 200 free Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Christoph Lameter <clameter@sgi.com> Cc: Adam Litke <agl@us.ibm.com> Cc: David Gibson <hermes@gibson.dropbear.id.au> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Lee Schermerhorn <lee.schermerhorn@hp.com> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-10-16 08:26:24 +00:00
hugetlb_next_nid = first_node(node_online_map);
for (i = 0; i < max_huge_pages; ++i) {
if (!alloc_fresh_huge_page())
break;
}
max_huge_pages = free_huge_pages = nr_huge_pages = i;
printk("Total HugeTLB memory allocated, %ld\n", free_huge_pages);
return 0;
}
module_init(hugetlb_init);
static int __init hugetlb_setup(char *s)
{
if (sscanf(s, "%lu", &max_huge_pages) <= 0)
max_huge_pages = 0;
return 1;
}
__setup("hugepages=", hugetlb_setup);
static unsigned int cpuset_mems_nr(unsigned int *array)
{
int node;
unsigned int nr = 0;
for_each_node_mask(node, cpuset_current_mems_allowed)
nr += array[node];
return nr;
}
#ifdef CONFIG_SYSCTL
#ifdef CONFIG_HIGHMEM
static void try_to_free_low(unsigned long count)
{
int i;
for (i = 0; i < MAX_NUMNODES; ++i) {
struct page *page, *next;
list_for_each_entry_safe(page, next, &hugepage_freelists[i], lru) {
if (count >= nr_huge_pages)
return;
if (PageHighMem(page))
continue;
list_del(&page->lru);
update_and_free_page(page);
free_huge_pages--;
free_huge_pages_node[page_to_nid(page)]--;
}
}
}
#else
static inline void try_to_free_low(unsigned long count)
{
}
#endif
#define persistent_huge_pages (nr_huge_pages - surplus_huge_pages)
static unsigned long set_max_huge_pages(unsigned long count)
{
unsigned long min_count, ret;
/*
* Increase the pool size
* First take pages out of surplus state. Then make up the
* remaining difference by allocating fresh huge pages.
hugetlb: introduce nr_overcommit_hugepages sysctl hugetlb: introduce nr_overcommit_hugepages sysctl While examining the code to support /proc/sys/vm/hugetlb_dynamic_pool, I became convinced that having a boolean sysctl was insufficient: 1) To support per-node control of hugepages, I have previously submitted patches to add a sysfs attribute related to nr_hugepages. However, with a boolean global value and per-mount quota enforcement constraining the dynamic pool, adding corresponding control of the dynamic pool on a per-node basis seems inconsistent to me. 2) Administration of the hugetlb dynamic pool with multiple hugetlbfs mount points is, arguably, more arduous than it needs to be. Each quota would need to be set separately, and the sum would need to be monitored. To ease the administration, and to help make the way for per-node control of the static & dynamic hugepage pool, I added a separate sysctl, nr_overcommit_hugepages. This value serves as a high watermark for the overall hugepage pool, while nr_hugepages serves as a low watermark. The boolean sysctl can then be removed, as the condition nr_overcommit_hugepages > 0 indicates the same administrative setting as hugetlb_dynamic_pool == 1 Quotas still serve as local enforcement of the size of the pool on a per-mount basis. A few caveats: 1) There is a race whereby the global surplus huge page counter is incremented before a hugepage has allocated. Another process could then try grow the pool, and fail to convert a surplus huge page to a normal huge page and instead allocate a fresh huge page. I believe this is benign, as no memory is leaked (the actual pages are still tracked correctly) and the counters won't go out of sync. 2) Shrinking the static pool while a surplus is in effect will allow the number of surplus huge pages to exceed the overcommit value. As long as this condition holds, however, no more surplus huge pages will be allowed on the system until one of the two sysctls are increased sufficiently, or the surplus huge pages go out of use and are freed. Successfully tested on x86_64 with the current libhugetlbfs snapshot, modified to use the new sysctl. Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Adam Litke <agl@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-12-18 00:20:12 +00:00
*
* We might race with alloc_buddy_huge_page() here and be unable
* to convert a surplus huge page to a normal huge page. That is
* not critical, though, it just means the overall size of the
* pool might be one hugepage larger than it needs to be, but
* within all the constraints specified by the sysctls.
*/
spin_lock(&hugetlb_lock);
while (surplus_huge_pages && count > persistent_huge_pages) {
if (!adjust_pool_surplus(-1))
break;
}
while (count > persistent_huge_pages) {
int ret;
/*
* If this allocation races such that we no longer need the
* page, free_huge_page will handle it by freeing the page
* and reducing the surplus.
*/
spin_unlock(&hugetlb_lock);
ret = alloc_fresh_huge_page();
spin_lock(&hugetlb_lock);
if (!ret)
goto out;
}
/*
* Decrease the pool size
* First return free pages to the buddy allocator (being careful
* to keep enough around to satisfy reservations). Then place
* pages into surplus state as needed so the pool will shrink
* to the desired size as pages become free.
hugetlb: introduce nr_overcommit_hugepages sysctl hugetlb: introduce nr_overcommit_hugepages sysctl While examining the code to support /proc/sys/vm/hugetlb_dynamic_pool, I became convinced that having a boolean sysctl was insufficient: 1) To support per-node control of hugepages, I have previously submitted patches to add a sysfs attribute related to nr_hugepages. However, with a boolean global value and per-mount quota enforcement constraining the dynamic pool, adding corresponding control of the dynamic pool on a per-node basis seems inconsistent to me. 2) Administration of the hugetlb dynamic pool with multiple hugetlbfs mount points is, arguably, more arduous than it needs to be. Each quota would need to be set separately, and the sum would need to be monitored. To ease the administration, and to help make the way for per-node control of the static & dynamic hugepage pool, I added a separate sysctl, nr_overcommit_hugepages. This value serves as a high watermark for the overall hugepage pool, while nr_hugepages serves as a low watermark. The boolean sysctl can then be removed, as the condition nr_overcommit_hugepages > 0 indicates the same administrative setting as hugetlb_dynamic_pool == 1 Quotas still serve as local enforcement of the size of the pool on a per-mount basis. A few caveats: 1) There is a race whereby the global surplus huge page counter is incremented before a hugepage has allocated. Another process could then try grow the pool, and fail to convert a surplus huge page to a normal huge page and instead allocate a fresh huge page. I believe this is benign, as no memory is leaked (the actual pages are still tracked correctly) and the counters won't go out of sync. 2) Shrinking the static pool while a surplus is in effect will allow the number of surplus huge pages to exceed the overcommit value. As long as this condition holds, however, no more surplus huge pages will be allowed on the system until one of the two sysctls are increased sufficiently, or the surplus huge pages go out of use and are freed. Successfully tested on x86_64 with the current libhugetlbfs snapshot, modified to use the new sysctl. Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Acked-by: Adam Litke <agl@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: Dave Hansen <haveblue@us.ibm.com> Cc: David Gibson <david@gibson.dropbear.id.au> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-12-18 00:20:12 +00:00
*
* By placing pages into the surplus state independent of the
* overcommit value, we are allowing the surplus pool size to
* exceed overcommit. There are few sane options here. Since
* alloc_buddy_huge_page() is checking the global counter,
* though, we'll note that we're not allowed to exceed surplus
* and won't grow the pool anywhere else. Not until one of the
* sysctls are changed, or the surplus pages go out of use.
*/
min_count = resv_huge_pages + nr_huge_pages - free_huge_pages;
min_count = max(count, min_count);
try_to_free_low(min_count);
while (min_count < persistent_huge_pages) {
struct page *page = dequeue_huge_page(NULL, 0);
if (!page)
break;
update_and_free_page(page);
}
while (count < persistent_huge_pages) {
if (!adjust_pool_surplus(1))
break;
}
out:
ret = persistent_huge_pages;
spin_unlock(&hugetlb_lock);
return ret;
}
int hugetlb_sysctl_handler(struct ctl_table *table, int write,
struct file *file, void __user *buffer,
size_t *length, loff_t *ppos)
{
proc_doulongvec_minmax(table, write, file, buffer, length, ppos);
max_huge_pages = set_max_huge_pages(max_huge_pages);
return 0;
}
int hugetlb_treat_movable_handler(struct 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);
if (hugepages_treat_as_movable)
htlb_alloc_mask = GFP_HIGHUSER_MOVABLE;
else
htlb_alloc_mask = GFP_HIGHUSER;
return 0;
}
#endif /* CONFIG_SYSCTL */
int hugetlb_report_meminfo(char *buf)
{
return sprintf(buf,
"HugePages_Total: %5lu\n"
"HugePages_Free: %5lu\n"
"HugePages_Rsvd: %5lu\n"
"HugePages_Surp: %5lu\n"
"Hugepagesize: %5lu kB\n",
nr_huge_pages,
free_huge_pages,
resv_huge_pages,
surplus_huge_pages,
HPAGE_SIZE/1024);
}
int hugetlb_report_node_meminfo(int nid, char *buf)
{
return sprintf(buf,
"Node %d HugePages_Total: %5u\n"
"Node %d HugePages_Free: %5u\n",
nid, nr_huge_pages_node[nid],
nid, free_huge_pages_node[nid]);
}
/* Return the number pages of memory we physically have, in PAGE_SIZE units. */
unsigned long hugetlb_total_pages(void)
{
return nr_huge_pages * (HPAGE_SIZE / PAGE_SIZE);
}
/*
* We cannot handle pagefaults against hugetlb pages at all. They cause
* handle_mm_fault() to try to instantiate regular-sized pages in the
* hugegpage VMA. do_page_fault() is supposed to trap this, so BUG is we get
* this far.
*/
static int hugetlb_vm_op_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
{
BUG();
return 0;
}
struct vm_operations_struct hugetlb_vm_ops = {
.fault = hugetlb_vm_op_fault,
};
static pte_t make_huge_pte(struct vm_area_struct *vma, struct page *page,
int writable)
{
pte_t entry;
if (writable) {
entry =
pte_mkwrite(pte_mkdirty(mk_pte(page, vma->vm_page_prot)));
} else {
entry = pte_wrprotect(mk_pte(page, vma->vm_page_prot));
}
entry = pte_mkyoung(entry);
entry = pte_mkhuge(entry);
return entry;
}
static void set_huge_ptep_writable(struct vm_area_struct *vma,
unsigned long address, pte_t *ptep)
{
pte_t entry;
entry = pte_mkwrite(pte_mkdirty(*ptep));
if (ptep_set_access_flags(vma, address, ptep, entry, 1)) {
update_mmu_cache(vma, address, entry);
}
}
int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src,
struct vm_area_struct *vma)
{
pte_t *src_pte, *dst_pte, entry;
struct page *ptepage;
unsigned long addr;
int cow;
cow = (vma->vm_flags & (VM_SHARED | VM_MAYWRITE)) == VM_MAYWRITE;
for (addr = vma->vm_start; addr < vma->vm_end; addr += HPAGE_SIZE) {
src_pte = huge_pte_offset(src, addr);
if (!src_pte)
continue;
dst_pte = huge_pte_alloc(dst, addr);
if (!dst_pte)
goto nomem;
fix hugepages leak due to pagetable page sharing The shared page table code for hugetlb memory on x86 and x86_64 is causing a leak. When a user of hugepages exits using this code the system leaks some of the hugepages. ------------------------------------------------------- Part of /proc/meminfo just before database startup: HugePages_Total: 5500 HugePages_Free: 5500 HugePages_Rsvd: 0 Hugepagesize: 2048 kB Just before shutdown: HugePages_Total: 5500 HugePages_Free: 4475 HugePages_Rsvd: 0 Hugepagesize: 2048 kB After shutdown: HugePages_Total: 5500 HugePages_Free: 4988 HugePages_Rsvd: 0 Hugepagesize: 2048 kB ---------------------------------------------------------- The problem occurs durring a fork, in copy_hugetlb_page_range(). It locates the dst_pte using huge_pte_alloc(). Since huge_pte_alloc() calls huge_pmd_share() it will share the pmd page if can, yet the main loop in copy_hugetlb_page_range() does a get_page() on every hugepage. This is a violation of the shared hugepmd pagetable protocol and creates additional referenced to the hugepages causing a leak when the unmap of the VMA occurs. We can skip the entire replication of the ptes when the hugepage pagetables are shared. The attached patch skips copying the ptes and the get_page() calls if the hugetlbpage pagetable is shared. [akpm@linux-foundation.org: coding-style cleanups] Signed-off-by: Larry Woodman <lwoodman@redhat.com> Signed-off-by: Adam Litke <agl@us.ibm.com> Cc: Badari Pulavarty <pbadari@us.ibm.com> Cc: Ken Chen <kenchen@google.com> Cc: David Gibson <david@gibson.dropbear.id.au> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2008-01-24 13:49:25 +00:00
/* If the pagetables are shared don't copy or take references */
if (dst_pte == src_pte)
continue;
spin_lock(&dst->page_table_lock);
spin_lock(&src->page_table_lock);
if (!pte_none(*src_pte)) {
if (cow)
ptep_set_wrprotect(src, addr, src_pte);
entry = *src_pte;
ptepage = pte_page(entry);
get_page(ptepage);
set_huge_pte_at(dst, addr, dst_pte, entry);
}
spin_unlock(&src->page_table_lock);
spin_unlock(&dst->page_table_lock);
}
return 0;
nomem:
return -ENOMEM;
}
void __unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
unsigned long end)
{
struct mm_struct *mm = vma->vm_mm;
unsigned long address;
pte_t *ptep;
pte_t pte;
struct page *page;
struct page *tmp;
/*
* A page gathering list, protected by per file i_mmap_lock. The
* lock is used to avoid list corruption from multiple unmapping
* of the same page since we are using page->lru.
*/
LIST_HEAD(page_list);
WARN_ON(!is_vm_hugetlb_page(vma));
BUG_ON(start & ~HPAGE_MASK);
BUG_ON(end & ~HPAGE_MASK);
spin_lock(&mm->page_table_lock);
for (address = start; address < end; address += HPAGE_SIZE) {
ptep = huge_pte_offset(mm, address);
if (!ptep)
continue;
[PATCH] shared page table for hugetlb page Following up with the work on shared page table done by Dave McCracken. This set of patch target shared page table for hugetlb memory only. The shared page table is particular useful in the situation of large number of independent processes sharing large shared memory segments. In the normal page case, the amount of memory saved from process' page table is quite significant. For hugetlb, the saving on page table memory is not the primary objective (as hugetlb itself already cuts down page table overhead significantly), instead, the purpose of using shared page table on hugetlb is to allow faster TLB refill and smaller cache pollution upon TLB miss. With PT sharing, pte entries are shared among hundreds of processes, the cache consumption used by all the page table is smaller and in return, application gets much higher cache hit ratio. One other effect is that cache hit ratio with hardware page walker hitting on pte in cache will be higher and this helps to reduce tlb miss latency. These two effects contribute to higher application performance. Signed-off-by: Ken Chen <kenneth.w.chen@intel.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Dave McCracken <dmccr@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Gibson <david@gibson.dropbear.id.au> Cc: Adam Litke <agl@us.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:32:03 +00:00
if (huge_pmd_unshare(mm, &address, ptep))
continue;
pte = huge_ptep_get_and_clear(mm, address, ptep);
if (pte_none(pte))
continue;
page = pte_page(pte);
if (pte_dirty(pte))
set_page_dirty(page);
list_add(&page->lru, &page_list);
}
spin_unlock(&mm->page_table_lock);
flush_tlb_range(vma, start, end);
list_for_each_entry_safe(page, tmp, &page_list, lru) {
list_del(&page->lru);
put_page(page);
}
}
void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
unsigned long end)
{
/*
* It is undesirable to test vma->vm_file as it should be non-null
* for valid hugetlb area. However, vm_file will be NULL in the error
* cleanup path of do_mmap_pgoff. When hugetlbfs ->mmap method fails,
* do_mmap_pgoff() nullifies vma->vm_file before calling this function
* to clean up. Since no pte has actually been setup, it is safe to
* do nothing in this case.
*/
if (vma->vm_file) {
spin_lock(&vma->vm_file->f_mapping->i_mmap_lock);
__unmap_hugepage_range(vma, start, end);
spin_unlock(&vma->vm_file->f_mapping->i_mmap_lock);
}
}
static int hugetlb_cow(struct mm_struct *mm, struct vm_area_struct *vma,
unsigned long address, pte_t *ptep, pte_t pte)
{
struct page *old_page, *new_page;
int avoidcopy;
old_page = pte_page(pte);
/* If no-one else is actually using this page, avoid the copy
* and just make the page writable */
avoidcopy = (page_count(old_page) == 1);
if (avoidcopy) {
set_huge_ptep_writable(vma, address, ptep);
mm: fault feedback #2 This patch completes Linus's wish that the fault return codes be made into bit flags, which I agree makes everything nicer. This requires requires all handle_mm_fault callers to be modified (possibly the modifications should go further and do things like fault accounting in handle_mm_fault -- however that would be for another patch). [akpm@linux-foundation.org: fix alpha build] [akpm@linux-foundation.org: fix s390 build] [akpm@linux-foundation.org: fix sparc build] [akpm@linux-foundation.org: fix sparc64 build] [akpm@linux-foundation.org: fix ia64 build] Signed-off-by: Nick Piggin <npiggin@suse.de> Cc: Richard Henderson <rth@twiddle.net> Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Ian Molton <spyro@f2s.com> Cc: Bryan Wu <bryan.wu@analog.com> Cc: Mikael Starvik <starvik@axis.com> Cc: David Howells <dhowells@redhat.com> Cc: Yoshinori Sato <ysato@users.sourceforge.jp> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Hirokazu Takata <takata@linux-m32r.org> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Roman Zippel <zippel@linux-m68k.org> Cc: Greg Ungerer <gerg@uclinux.org> Cc: Matthew Wilcox <willy@debian.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp> Cc: Richard Curnow <rc@rc0.org.uk> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Jeff Dike <jdike@addtoit.com> Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp> Cc: Chris Zankel <chris@zankel.net> Acked-by: Kyle McMartin <kyle@mcmartin.ca> Acked-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Acked-by: Ralf Baechle <ralf@linux-mips.org> Acked-by: Andi Kleen <ak@muc.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> [ Still apparently needs some ARM and PPC loving - Linus ] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 08:47:05 +00:00
return 0;
}
page_cache_get(old_page);
new_page = alloc_huge_page(vma, address);
if (IS_ERR(new_page)) {
page_cache_release(old_page);
return -PTR_ERR(new_page);
}
spin_unlock(&mm->page_table_lock);
copy_huge_page(new_page, old_page, address, vma);
spin_lock(&mm->page_table_lock);
ptep = huge_pte_offset(mm, address & HPAGE_MASK);
if (likely(pte_same(*ptep, pte))) {
/* Break COW */
set_huge_pte_at(mm, address, ptep,
make_huge_pte(vma, new_page, 1));
/* Make the old page be freed below */
new_page = old_page;
}
page_cache_release(new_page);
page_cache_release(old_page);
mm: fault feedback #2 This patch completes Linus's wish that the fault return codes be made into bit flags, which I agree makes everything nicer. This requires requires all handle_mm_fault callers to be modified (possibly the modifications should go further and do things like fault accounting in handle_mm_fault -- however that would be for another patch). [akpm@linux-foundation.org: fix alpha build] [akpm@linux-foundation.org: fix s390 build] [akpm@linux-foundation.org: fix sparc build] [akpm@linux-foundation.org: fix sparc64 build] [akpm@linux-foundation.org: fix ia64 build] Signed-off-by: Nick Piggin <npiggin@suse.de> Cc: Richard Henderson <rth@twiddle.net> Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Ian Molton <spyro@f2s.com> Cc: Bryan Wu <bryan.wu@analog.com> Cc: Mikael Starvik <starvik@axis.com> Cc: David Howells <dhowells@redhat.com> Cc: Yoshinori Sato <ysato@users.sourceforge.jp> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Hirokazu Takata <takata@linux-m32r.org> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Roman Zippel <zippel@linux-m68k.org> Cc: Greg Ungerer <gerg@uclinux.org> Cc: Matthew Wilcox <willy@debian.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp> Cc: Richard Curnow <rc@rc0.org.uk> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Jeff Dike <jdike@addtoit.com> Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp> Cc: Chris Zankel <chris@zankel.net> Acked-by: Kyle McMartin <kyle@mcmartin.ca> Acked-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Acked-by: Ralf Baechle <ralf@linux-mips.org> Acked-by: Andi Kleen <ak@muc.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> [ Still apparently needs some ARM and PPC loving - Linus ] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 08:47:05 +00:00
return 0;
}
static int hugetlb_no_page(struct mm_struct *mm, struct vm_area_struct *vma,
unsigned long address, pte_t *ptep, int write_access)
{
int ret = VM_FAULT_SIGBUS;
unsigned long idx;
unsigned long size;
struct page *page;
struct address_space *mapping;
pte_t new_pte;
mapping = vma->vm_file->f_mapping;
idx = ((address - vma->vm_start) >> HPAGE_SHIFT)
+ (vma->vm_pgoff >> (HPAGE_SHIFT - PAGE_SHIFT));
/*
* Use page lock to guard against racing truncation
* before we get page_table_lock.
*/
retry:
page = find_lock_page(mapping, idx);
if (!page) {
size = i_size_read(mapping->host) >> HPAGE_SHIFT;
if (idx >= size)
goto out;
page = alloc_huge_page(vma, address);
if (IS_ERR(page)) {
ret = -PTR_ERR(page);
goto out;
}
clear_huge_page(page, address);
if (vma->vm_flags & VM_SHARED) {
int err;
struct inode *inode = mapping->host;
err = add_to_page_cache(page, mapping, idx, GFP_KERNEL);
if (err) {
put_page(page);
if (err == -EEXIST)
goto retry;
goto out;
}
spin_lock(&inode->i_lock);
inode->i_blocks += BLOCKS_PER_HUGEPAGE;
spin_unlock(&inode->i_lock);
} else
lock_page(page);
}
spin_lock(&mm->page_table_lock);
size = i_size_read(mapping->host) >> HPAGE_SHIFT;
if (idx >= size)
goto backout;
mm: fault feedback #2 This patch completes Linus's wish that the fault return codes be made into bit flags, which I agree makes everything nicer. This requires requires all handle_mm_fault callers to be modified (possibly the modifications should go further and do things like fault accounting in handle_mm_fault -- however that would be for another patch). [akpm@linux-foundation.org: fix alpha build] [akpm@linux-foundation.org: fix s390 build] [akpm@linux-foundation.org: fix sparc build] [akpm@linux-foundation.org: fix sparc64 build] [akpm@linux-foundation.org: fix ia64 build] Signed-off-by: Nick Piggin <npiggin@suse.de> Cc: Richard Henderson <rth@twiddle.net> Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Ian Molton <spyro@f2s.com> Cc: Bryan Wu <bryan.wu@analog.com> Cc: Mikael Starvik <starvik@axis.com> Cc: David Howells <dhowells@redhat.com> Cc: Yoshinori Sato <ysato@users.sourceforge.jp> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Hirokazu Takata <takata@linux-m32r.org> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Roman Zippel <zippel@linux-m68k.org> Cc: Greg Ungerer <gerg@uclinux.org> Cc: Matthew Wilcox <willy@debian.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp> Cc: Richard Curnow <rc@rc0.org.uk> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Jeff Dike <jdike@addtoit.com> Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp> Cc: Chris Zankel <chris@zankel.net> Acked-by: Kyle McMartin <kyle@mcmartin.ca> Acked-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Acked-by: Ralf Baechle <ralf@linux-mips.org> Acked-by: Andi Kleen <ak@muc.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> [ Still apparently needs some ARM and PPC loving - Linus ] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 08:47:05 +00:00
ret = 0;
if (!pte_none(*ptep))
goto backout;
new_pte = make_huge_pte(vma, page, ((vma->vm_flags & VM_WRITE)
&& (vma->vm_flags & VM_SHARED)));
set_huge_pte_at(mm, address, ptep, new_pte);
if (write_access && !(vma->vm_flags & VM_SHARED)) {
/* Optimization, do the COW without a second fault */
ret = hugetlb_cow(mm, vma, address, ptep, new_pte);
}
spin_unlock(&mm->page_table_lock);
unlock_page(page);
out:
return ret;
backout:
spin_unlock(&mm->page_table_lock);
unlock_page(page);
put_page(page);
goto out;
}
int hugetlb_fault(struct mm_struct *mm, struct vm_area_struct *vma,
unsigned long address, int write_access)
{
pte_t *ptep;
pte_t entry;
int ret;
[PATCH] hugepage: serialize hugepage allocation and instantiation Currently, no lock or mutex is held between allocating a hugepage and inserting it into the pagetables / page cache. When we do go to insert the page into pagetables or page cache, we recheck and may free the newly allocated hugepage. However, since the number of hugepages in the system is strictly limited, and it's usualy to want to use all of them, this can still lead to spurious allocation failures. For example, suppose two processes are both mapping (MAP_SHARED) the same hugepage file, large enough to consume the entire available hugepage pool. If they race instantiating the last page in the mapping, they will both attempt to allocate the last available hugepage. One will fail, of course, returning OOM from the fault and thus causing the process to be killed, despite the fact that the entire mapping can, in fact, be instantiated. The patch fixes this race by the simple method of adding a (sleeping) mutex to serialize the hugepage fault path between allocation and insertion into pagetables and/or page cache. It would be possible to avoid the serialization by catching the allocation failures, waiting on some condition, then rechecking to see if someone else has instantiated the page for us. Given the likely frequency of hugepage instantiations, it seems very doubtful it's worth the extra complexity. This patch causes no regression on the libhugetlbfs testsuite, and one test, which can trigger this race now passes where it previously failed. Actually, the test still sometimes fails, though less often and only as a shmat() failure, rather processes getting OOM killed by the VM. The dodgy heuristic tests in fs/hugetlbfs/inode.c for whether there's enough hugepage space aren't protected by the new mutex, and would be ugly to do so, so there's still a race there. Another patch to replace those tests with something saner for this reason as well as others coming... Signed-off-by: David Gibson <dwg@au1.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:53 +00:00
static DEFINE_MUTEX(hugetlb_instantiation_mutex);
ptep = huge_pte_alloc(mm, address);
if (!ptep)
return VM_FAULT_OOM;
[PATCH] hugepage: serialize hugepage allocation and instantiation Currently, no lock or mutex is held between allocating a hugepage and inserting it into the pagetables / page cache. When we do go to insert the page into pagetables or page cache, we recheck and may free the newly allocated hugepage. However, since the number of hugepages in the system is strictly limited, and it's usualy to want to use all of them, this can still lead to spurious allocation failures. For example, suppose two processes are both mapping (MAP_SHARED) the same hugepage file, large enough to consume the entire available hugepage pool. If they race instantiating the last page in the mapping, they will both attempt to allocate the last available hugepage. One will fail, of course, returning OOM from the fault and thus causing the process to be killed, despite the fact that the entire mapping can, in fact, be instantiated. The patch fixes this race by the simple method of adding a (sleeping) mutex to serialize the hugepage fault path between allocation and insertion into pagetables and/or page cache. It would be possible to avoid the serialization by catching the allocation failures, waiting on some condition, then rechecking to see if someone else has instantiated the page for us. Given the likely frequency of hugepage instantiations, it seems very doubtful it's worth the extra complexity. This patch causes no regression on the libhugetlbfs testsuite, and one test, which can trigger this race now passes where it previously failed. Actually, the test still sometimes fails, though less often and only as a shmat() failure, rather processes getting OOM killed by the VM. The dodgy heuristic tests in fs/hugetlbfs/inode.c for whether there's enough hugepage space aren't protected by the new mutex, and would be ugly to do so, so there's still a race there. Another patch to replace those tests with something saner for this reason as well as others coming... Signed-off-by: David Gibson <dwg@au1.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:53 +00:00
/*
* Serialize hugepage allocation and instantiation, so that we don't
* get spurious allocation failures if two CPUs race to instantiate
* the same page in the page cache.
*/
mutex_lock(&hugetlb_instantiation_mutex);
entry = *ptep;
[PATCH] hugepage: serialize hugepage allocation and instantiation Currently, no lock or mutex is held between allocating a hugepage and inserting it into the pagetables / page cache. When we do go to insert the page into pagetables or page cache, we recheck and may free the newly allocated hugepage. However, since the number of hugepages in the system is strictly limited, and it's usualy to want to use all of them, this can still lead to spurious allocation failures. For example, suppose two processes are both mapping (MAP_SHARED) the same hugepage file, large enough to consume the entire available hugepage pool. If they race instantiating the last page in the mapping, they will both attempt to allocate the last available hugepage. One will fail, of course, returning OOM from the fault and thus causing the process to be killed, despite the fact that the entire mapping can, in fact, be instantiated. The patch fixes this race by the simple method of adding a (sleeping) mutex to serialize the hugepage fault path between allocation and insertion into pagetables and/or page cache. It would be possible to avoid the serialization by catching the allocation failures, waiting on some condition, then rechecking to see if someone else has instantiated the page for us. Given the likely frequency of hugepage instantiations, it seems very doubtful it's worth the extra complexity. This patch causes no regression on the libhugetlbfs testsuite, and one test, which can trigger this race now passes where it previously failed. Actually, the test still sometimes fails, though less often and only as a shmat() failure, rather processes getting OOM killed by the VM. The dodgy heuristic tests in fs/hugetlbfs/inode.c for whether there's enough hugepage space aren't protected by the new mutex, and would be ugly to do so, so there's still a race there. Another patch to replace those tests with something saner for this reason as well as others coming... Signed-off-by: David Gibson <dwg@au1.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:53 +00:00
if (pte_none(entry)) {
ret = hugetlb_no_page(mm, vma, address, ptep, write_access);
mutex_unlock(&hugetlb_instantiation_mutex);
return ret;
}
mm: fault feedback #2 This patch completes Linus's wish that the fault return codes be made into bit flags, which I agree makes everything nicer. This requires requires all handle_mm_fault callers to be modified (possibly the modifications should go further and do things like fault accounting in handle_mm_fault -- however that would be for another patch). [akpm@linux-foundation.org: fix alpha build] [akpm@linux-foundation.org: fix s390 build] [akpm@linux-foundation.org: fix sparc build] [akpm@linux-foundation.org: fix sparc64 build] [akpm@linux-foundation.org: fix ia64 build] Signed-off-by: Nick Piggin <npiggin@suse.de> Cc: Richard Henderson <rth@twiddle.net> Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru> Cc: Russell King <rmk@arm.linux.org.uk> Cc: Ian Molton <spyro@f2s.com> Cc: Bryan Wu <bryan.wu@analog.com> Cc: Mikael Starvik <starvik@axis.com> Cc: David Howells <dhowells@redhat.com> Cc: Yoshinori Sato <ysato@users.sourceforge.jp> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Hirokazu Takata <takata@linux-m32r.org> Cc: Geert Uytterhoeven <geert@linux-m68k.org> Cc: Roman Zippel <zippel@linux-m68k.org> Cc: Greg Ungerer <gerg@uclinux.org> Cc: Matthew Wilcox <willy@debian.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Heiko Carstens <heiko.carstens@de.ibm.com> Cc: Martin Schwidefsky <schwidefsky@de.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: Kazumoto Kojima <kkojima@rr.iij4u.or.jp> Cc: Richard Curnow <rc@rc0.org.uk> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: "David S. Miller" <davem@davemloft.net> Cc: Jeff Dike <jdike@addtoit.com> Cc: Paolo 'Blaisorblade' Giarrusso <blaisorblade@yahoo.it> Cc: Miles Bader <uclinux-v850@lsi.nec.co.jp> Cc: Chris Zankel <chris@zankel.net> Acked-by: Kyle McMartin <kyle@mcmartin.ca> Acked-by: Haavard Skinnemoen <hskinnemoen@atmel.com> Acked-by: Ralf Baechle <ralf@linux-mips.org> Acked-by: Andi Kleen <ak@muc.de> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> [ Still apparently needs some ARM and PPC loving - Linus ] Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2007-07-19 08:47:05 +00:00
ret = 0;
spin_lock(&mm->page_table_lock);
/* Check for a racing update before calling hugetlb_cow */
if (likely(pte_same(entry, *ptep)))
if (write_access && !pte_write(entry))
ret = hugetlb_cow(mm, vma, address, ptep, entry);
spin_unlock(&mm->page_table_lock);
[PATCH] hugepage: serialize hugepage allocation and instantiation Currently, no lock or mutex is held between allocating a hugepage and inserting it into the pagetables / page cache. When we do go to insert the page into pagetables or page cache, we recheck and may free the newly allocated hugepage. However, since the number of hugepages in the system is strictly limited, and it's usualy to want to use all of them, this can still lead to spurious allocation failures. For example, suppose two processes are both mapping (MAP_SHARED) the same hugepage file, large enough to consume the entire available hugepage pool. If they race instantiating the last page in the mapping, they will both attempt to allocate the last available hugepage. One will fail, of course, returning OOM from the fault and thus causing the process to be killed, despite the fact that the entire mapping can, in fact, be instantiated. The patch fixes this race by the simple method of adding a (sleeping) mutex to serialize the hugepage fault path between allocation and insertion into pagetables and/or page cache. It would be possible to avoid the serialization by catching the allocation failures, waiting on some condition, then rechecking to see if someone else has instantiated the page for us. Given the likely frequency of hugepage instantiations, it seems very doubtful it's worth the extra complexity. This patch causes no regression on the libhugetlbfs testsuite, and one test, which can trigger this race now passes where it previously failed. Actually, the test still sometimes fails, though less often and only as a shmat() failure, rather processes getting OOM killed by the VM. The dodgy heuristic tests in fs/hugetlbfs/inode.c for whether there's enough hugepage space aren't protected by the new mutex, and would be ugly to do so, so there's still a race there. Another patch to replace those tests with something saner for this reason as well as others coming... Signed-off-by: David Gibson <dwg@au1.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:53 +00:00
mutex_unlock(&hugetlb_instantiation_mutex);
return ret;
}
int follow_hugetlb_page(struct mm_struct *mm, struct vm_area_struct *vma,
struct page **pages, struct vm_area_struct **vmas,
unsigned long *position, int *length, int i,
int write)
{
unsigned long pfn_offset;
unsigned long vaddr = *position;
int remainder = *length;
spin_lock(&mm->page_table_lock);
while (vaddr < vma->vm_end && remainder) {
pte_t *pte;
struct page *page;
/*
* Some archs (sparc64, sh*) have multiple pte_ts to
* each hugepage. We have to make * sure we get the
* first, for the page indexing below to work.
*/
pte = huge_pte_offset(mm, vaddr & HPAGE_MASK);
if (!pte || pte_none(*pte) || (write && !pte_write(*pte))) {
int ret;
spin_unlock(&mm->page_table_lock);
ret = hugetlb_fault(mm, vma, vaddr, write);
spin_lock(&mm->page_table_lock);
if (!(ret & VM_FAULT_ERROR))
continue;
remainder = 0;
if (!i)
i = -EFAULT;
break;
}
pfn_offset = (vaddr & ~HPAGE_MASK) >> PAGE_SHIFT;
page = pte_page(*pte);
same_page:
if (pages) {
get_page(page);
pages[i] = page + pfn_offset;
}
if (vmas)
vmas[i] = vma;
vaddr += PAGE_SIZE;
++pfn_offset;
--remainder;
++i;
if (vaddr < vma->vm_end && remainder &&
pfn_offset < HPAGE_SIZE/PAGE_SIZE) {
/*
* We use pfn_offset to avoid touching the pageframes
* of this compound page.
*/
goto same_page;
}
}
spin_unlock(&mm->page_table_lock);
*length = remainder;
*position = vaddr;
return i;
}
[PATCH] Enable mprotect on huge pages 2.6.16-rc3 uses hugetlb on-demand paging, but it doesn_t support hugetlb mprotect. From: David Gibson <david@gibson.dropbear.id.au> Remove a test from the mprotect() path which checks that the mprotect()ed range on a hugepage VMA is hugepage aligned (yes, really, the sense of is_aligned_hugepage_range() is the opposite of what you'd guess :-/). In fact, we don't need this test. If the given addresses match the beginning/end of a hugepage VMA they must already be suitably aligned. If they don't, then mprotect_fixup() will attempt to split the VMA. The very first test in split_vma() will check for a badly aligned address on a hugepage VMA and return -EINVAL if necessary. From: "Chen, Kenneth W" <kenneth.w.chen@intel.com> On i386 and x86-64, pte flag _PAGE_PSE collides with _PAGE_PROTNONE. The identify of hugetlb pte is lost when changing page protection via mprotect. A page fault occurs later will trigger a bug check in huge_pte_alloc(). The fix is to always make new pte a hugetlb pte and also to clean up legacy code where _PAGE_PRESENT is forced on in the pre-faulting day. Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com> Cc: David Gibson <david@gibson.dropbear.id.au> Cc: "David S. Miller" <davem@davemloft.net> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Ken Chen <kenneth.w.chen@intel.com> Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Cc: Andi Kleen <ak@muc.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:50 +00:00
void hugetlb_change_protection(struct vm_area_struct *vma,
unsigned long address, unsigned long end, pgprot_t newprot)
{
struct mm_struct *mm = vma->vm_mm;
unsigned long start = address;
pte_t *ptep;
pte_t pte;
BUG_ON(address >= end);
flush_cache_range(vma, address, end);
[PATCH] shared page table for hugetlb page Following up with the work on shared page table done by Dave McCracken. This set of patch target shared page table for hugetlb memory only. The shared page table is particular useful in the situation of large number of independent processes sharing large shared memory segments. In the normal page case, the amount of memory saved from process' page table is quite significant. For hugetlb, the saving on page table memory is not the primary objective (as hugetlb itself already cuts down page table overhead significantly), instead, the purpose of using shared page table on hugetlb is to allow faster TLB refill and smaller cache pollution upon TLB miss. With PT sharing, pte entries are shared among hundreds of processes, the cache consumption used by all the page table is smaller and in return, application gets much higher cache hit ratio. One other effect is that cache hit ratio with hardware page walker hitting on pte in cache will be higher and this helps to reduce tlb miss latency. These two effects contribute to higher application performance. Signed-off-by: Ken Chen <kenneth.w.chen@intel.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Dave McCracken <dmccr@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Gibson <david@gibson.dropbear.id.au> Cc: Adam Litke <agl@us.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:32:03 +00:00
spin_lock(&vma->vm_file->f_mapping->i_mmap_lock);
[PATCH] Enable mprotect on huge pages 2.6.16-rc3 uses hugetlb on-demand paging, but it doesn_t support hugetlb mprotect. From: David Gibson <david@gibson.dropbear.id.au> Remove a test from the mprotect() path which checks that the mprotect()ed range on a hugepage VMA is hugepage aligned (yes, really, the sense of is_aligned_hugepage_range() is the opposite of what you'd guess :-/). In fact, we don't need this test. If the given addresses match the beginning/end of a hugepage VMA they must already be suitably aligned. If they don't, then mprotect_fixup() will attempt to split the VMA. The very first test in split_vma() will check for a badly aligned address on a hugepage VMA and return -EINVAL if necessary. From: "Chen, Kenneth W" <kenneth.w.chen@intel.com> On i386 and x86-64, pte flag _PAGE_PSE collides with _PAGE_PROTNONE. The identify of hugetlb pte is lost when changing page protection via mprotect. A page fault occurs later will trigger a bug check in huge_pte_alloc(). The fix is to always make new pte a hugetlb pte and also to clean up legacy code where _PAGE_PRESENT is forced on in the pre-faulting day. Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com> Cc: David Gibson <david@gibson.dropbear.id.au> Cc: "David S. Miller" <davem@davemloft.net> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Ken Chen <kenneth.w.chen@intel.com> Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Cc: Andi Kleen <ak@muc.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:50 +00:00
spin_lock(&mm->page_table_lock);
for (; address < end; address += HPAGE_SIZE) {
ptep = huge_pte_offset(mm, address);
if (!ptep)
continue;
[PATCH] shared page table for hugetlb page Following up with the work on shared page table done by Dave McCracken. This set of patch target shared page table for hugetlb memory only. The shared page table is particular useful in the situation of large number of independent processes sharing large shared memory segments. In the normal page case, the amount of memory saved from process' page table is quite significant. For hugetlb, the saving on page table memory is not the primary objective (as hugetlb itself already cuts down page table overhead significantly), instead, the purpose of using shared page table on hugetlb is to allow faster TLB refill and smaller cache pollution upon TLB miss. With PT sharing, pte entries are shared among hundreds of processes, the cache consumption used by all the page table is smaller and in return, application gets much higher cache hit ratio. One other effect is that cache hit ratio with hardware page walker hitting on pte in cache will be higher and this helps to reduce tlb miss latency. These two effects contribute to higher application performance. Signed-off-by: Ken Chen <kenneth.w.chen@intel.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Dave McCracken <dmccr@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Gibson <david@gibson.dropbear.id.au> Cc: Adam Litke <agl@us.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:32:03 +00:00
if (huge_pmd_unshare(mm, &address, ptep))
continue;
[PATCH] Enable mprotect on huge pages 2.6.16-rc3 uses hugetlb on-demand paging, but it doesn_t support hugetlb mprotect. From: David Gibson <david@gibson.dropbear.id.au> Remove a test from the mprotect() path which checks that the mprotect()ed range on a hugepage VMA is hugepage aligned (yes, really, the sense of is_aligned_hugepage_range() is the opposite of what you'd guess :-/). In fact, we don't need this test. If the given addresses match the beginning/end of a hugepage VMA they must already be suitably aligned. If they don't, then mprotect_fixup() will attempt to split the VMA. The very first test in split_vma() will check for a badly aligned address on a hugepage VMA and return -EINVAL if necessary. From: "Chen, Kenneth W" <kenneth.w.chen@intel.com> On i386 and x86-64, pte flag _PAGE_PSE collides with _PAGE_PROTNONE. The identify of hugetlb pte is lost when changing page protection via mprotect. A page fault occurs later will trigger a bug check in huge_pte_alloc(). The fix is to always make new pte a hugetlb pte and also to clean up legacy code where _PAGE_PRESENT is forced on in the pre-faulting day. Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com> Cc: David Gibson <david@gibson.dropbear.id.au> Cc: "David S. Miller" <davem@davemloft.net> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Ken Chen <kenneth.w.chen@intel.com> Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Cc: Andi Kleen <ak@muc.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:50 +00:00
if (!pte_none(*ptep)) {
pte = huge_ptep_get_and_clear(mm, address, ptep);
pte = pte_mkhuge(pte_modify(pte, newprot));
set_huge_pte_at(mm, address, ptep, pte);
}
}
spin_unlock(&mm->page_table_lock);
[PATCH] shared page table for hugetlb page Following up with the work on shared page table done by Dave McCracken. This set of patch target shared page table for hugetlb memory only. The shared page table is particular useful in the situation of large number of independent processes sharing large shared memory segments. In the normal page case, the amount of memory saved from process' page table is quite significant. For hugetlb, the saving on page table memory is not the primary objective (as hugetlb itself already cuts down page table overhead significantly), instead, the purpose of using shared page table on hugetlb is to allow faster TLB refill and smaller cache pollution upon TLB miss. With PT sharing, pte entries are shared among hundreds of processes, the cache consumption used by all the page table is smaller and in return, application gets much higher cache hit ratio. One other effect is that cache hit ratio with hardware page walker hitting on pte in cache will be higher and this helps to reduce tlb miss latency. These two effects contribute to higher application performance. Signed-off-by: Ken Chen <kenneth.w.chen@intel.com> Acked-by: Hugh Dickins <hugh@veritas.com> Cc: Dave McCracken <dmccr@us.ibm.com> Cc: William Lee Irwin III <wli@holomorphy.com> Cc: "Luck, Tony" <tony.luck@intel.com> Cc: Paul Mackerras <paulus@samba.org> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: David Gibson <david@gibson.dropbear.id.au> Cc: Adam Litke <agl@us.ibm.com> Cc: Paul Mundt <lethal@linux-sh.org> Cc: "David S. Miller" <davem@davemloft.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-12-07 04:32:03 +00:00
spin_unlock(&vma->vm_file->f_mapping->i_mmap_lock);
[PATCH] Enable mprotect on huge pages 2.6.16-rc3 uses hugetlb on-demand paging, but it doesn_t support hugetlb mprotect. From: David Gibson <david@gibson.dropbear.id.au> Remove a test from the mprotect() path which checks that the mprotect()ed range on a hugepage VMA is hugepage aligned (yes, really, the sense of is_aligned_hugepage_range() is the opposite of what you'd guess :-/). In fact, we don't need this test. If the given addresses match the beginning/end of a hugepage VMA they must already be suitably aligned. If they don't, then mprotect_fixup() will attempt to split the VMA. The very first test in split_vma() will check for a badly aligned address on a hugepage VMA and return -EINVAL if necessary. From: "Chen, Kenneth W" <kenneth.w.chen@intel.com> On i386 and x86-64, pte flag _PAGE_PSE collides with _PAGE_PROTNONE. The identify of hugetlb pte is lost when changing page protection via mprotect. A page fault occurs later will trigger a bug check in huge_pte_alloc(). The fix is to always make new pte a hugetlb pte and also to clean up legacy code where _PAGE_PRESENT is forced on in the pre-faulting day. Signed-off-by: Zhang Yanmin <yanmin.zhang@intel.com> Cc: David Gibson <david@gibson.dropbear.id.au> Cc: "David S. Miller" <davem@davemloft.net> Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mackerras <paulus@samba.org> Cc: William Lee Irwin III <wli@holomorphy.com> Signed-off-by: Ken Chen <kenneth.w.chen@intel.com> Signed-off-by: Nishanth Aravamudan <nacc@us.ibm.com> Cc: Andi Kleen <ak@muc.de> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Linus Torvalds <torvalds@osdl.org>
2006-03-22 08:08:50 +00:00
flush_tlb_range(vma, start, end);
}
struct file_region {
struct list_head link;
long from;
long to;
};
static long region_add(struct list_head *head, long f, long t)
{
struct file_region *rg, *nrg, *trg;
/* Locate the region we are either in or before. */
list_for_each_entry(rg, head, link)
if (f <= rg->to)
break;
/* Round our left edge to the current segment if it encloses us. */
if (f > rg->from)
f = rg->from;
/* Check for and consume any regions we now overlap with. */
nrg = rg;
list_for_each_entry_safe(rg, trg, rg->link.prev, link) {
if (&rg->link == head)
break;
if (rg->from > t)
break;
/* If this area reaches higher then extend our area to
* include it completely. If this is not the first area
* which we intend to reuse, free it. */
if (rg->to > t)
t = rg->to;
if (rg != nrg) {
list_del(&rg->link);
kfree(rg);
}
}
nrg->from = f;
nrg->to = t;
return 0;
}
static long region_chg(struct list_head *head, long f, long t)
{
struct file_region *rg, *nrg;
long chg = 0;
/* Locate the region we are before or in. */
list_for_each_entry(rg, head, link)
if (f <= rg->to)
break;
/* If we are below the current region then a new region is required.
* Subtle, allocate a new region at the position but make it zero
* size such that we can guarantee to record the reservation. */
if (&rg->link == head || t < rg->from) {
nrg = kmalloc(sizeof(*nrg), GFP_KERNEL);
if (!nrg)
return -ENOMEM;
nrg->from = f;
nrg->to = f;
INIT_LIST_HEAD(&nrg->link);
list_add(&nrg->link, rg->link.prev);
return t - f;
}
/* Round our left edge to the current segment if it encloses us. */
if (f > rg->from)
f = rg->from;
chg = t - f;
/* Check for and consume any regions we now overlap with. */
list_for_each_entry(rg, rg->link.prev, link) {
if (&rg->link == head)
break;
if (rg->from > t)
return chg;
/* We overlap with this area, if it extends futher than
* us then we must extend ourselves. Account for its
* existing reservation. */
if (rg->to > t) {
chg += rg->to - t;
t = rg->to;
}
chg -= rg->to - rg->from;
}
return chg;
}
static long region_truncate(struct list_head *head, long end)
{
struct file_region *rg, *trg;
long chg = 0;
/* Locate the region we are either in or before. */
list_for_each_entry(rg, head, link)
if (end <= rg->to)
break;
if (&rg->link == head)
return 0;
/* If we are in the middle of a region then adjust it. */
if (end > rg->from) {
chg = rg->to - end;
rg->to = end;
rg = list_entry(rg->link.next, typeof(*rg), link);
}
/* Drop any remaining regions. */
list_for_each_entry_safe(rg, trg, rg->link.prev, link) {
if (&rg->link == head)
break;
chg += rg->to - rg->from;
list_del(&rg->link);
kfree(rg);
}
return chg;
}
static int hugetlb_acct_memory(long delta)
{
int ret = -ENOMEM;
spin_lock(&hugetlb_lock);
/*
* When cpuset is configured, it breaks the strict hugetlb page
* reservation as the accounting is done on a global variable. Such
* reservation is completely rubbish in the presence of cpuset because
* the reservation is not checked against page availability for the
* current cpuset. Application can still potentially OOM'ed by kernel
* with lack of free htlb page in cpuset that the task is in.
* Attempt to enforce strict accounting with cpuset is almost
* impossible (or too ugly) because cpuset is too fluid that
* task or memory node can be dynamically moved between cpusets.
*
* The change of semantics for shared hugetlb mapping with cpuset is
* undesirable. However, in order to preserve some of the semantics,
* we fall back to check against current free page availability as
* a best attempt and hopefully to minimize the impact of changing
* semantics that cpuset has.
*/
if (delta > 0) {
if (gather_surplus_pages(delta) < 0)
goto out;
if (delta > cpuset_mems_nr(free_huge_pages_node))
goto out;
}
ret = 0;
resv_huge_pages += delta;
if (delta < 0)
return_unused_surplus_pages((unsigned long) -delta);
out:
spin_unlock(&hugetlb_lock);
return ret;
}
int hugetlb_reserve_pages(struct inode *inode, long from, long to)
{
long ret, chg;
chg = region_chg(&inode->i_mapping->private_list, from, to);
if (chg < 0)
return chg;
if (hugetlb_get_quota(inode->i_mapping, chg))
return -ENOSPC;
ret = hugetlb_acct_memory(chg);
if (ret < 0) {
hugetlb_put_quota(inode->i_mapping, chg);
return ret;
}
region_add(&inode->i_mapping->private_list, from, to);
return 0;
}
void hugetlb_unreserve_pages(struct inode *inode, long offset, long freed)
{
long chg = region_truncate(&inode->i_mapping->private_list, offset);
spin_lock(&inode->i_lock);
inode->i_blocks -= BLOCKS_PER_HUGEPAGE * freed;
spin_unlock(&inode->i_lock);
hugetlb_put_quota(inode->i_mapping, (chg - freed));
hugetlb_acct_memory(-(chg - freed));
}