linux-toradex.git/mm/hugetlb.c, branch v3.2.34

hugetlb: do not use vma_hugecache_offset() for vma_prio_tree_foreach

2012-10-17T02:49:47+00:00

commit 36e4f20af833d1ce196e6a4ade05dc26c44652d1 upstream.

Commit 0c176d52b0b2 ("mm: hugetlb: fix pgoff computation when unmapping
page from vma") fixed pgoff calculation but it has replaced it by
vma_hugecache_offset() which is not approapriate for offsets used for
vma_prio_tree_foreach() because that one expects index in page units
rather than in huge_page_shift.

Johannes said:

: The resulting index may not be too big, but it can be too small: assume
: hpage size of 2M and the address to unmap to be 0x200000.  This is regular
: page index 512 and hpage index 1.  If you have a VMA that maps the file
: only starting at the second huge page, that VMAs vm_pgoff will be 512 but
: you ask for offset 1 and miss it even though it does map the page of
: interest.  hugetlb_cow() will try to unmap, miss the vma, and retry the
: cow until the allocation succeeds or the skipped vma(s) go away.

Signed-off-by: Michal Hocko 
Acked-by: Hillf Danton 
Cc: Mel Gorman 
Cc: KAMEZAWA Hiroyuki 
Cc: Andrea Arcangeli 
Cc: David Rientjes 
Acked-by: Johannes Weiner 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds 
Signed-off-by: Ben Hutchings

mm: hugetlb: fix pgoff computation when unmapping page from vma

2012-10-17T02:49:46+00:00

commit 0c176d52b0b2619f231b2bbf329b90c028134f58 upstream.

The computation for pgoff is incorrect, at least with

	(vma->vm_pgoff >> PAGE_SHIFT)

involved.  It is fixed with the available method if HPAGE_SIZE is
concerned in page cache lookup.

[akpm@linux-foundation.org: use vma_hugecache_offset() directly, per Michal]
Signed-off-by: Hillf Danton 
Cc: Mel Gorman 
Cc: Michal Hocko 
Reviewed-by: KAMEZAWA Hiroyuki 
Cc: Andrea Arcangeli 
Cc: David Rientjes 
Reviewed-by: Michal Hocko 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds 
[bwh: Backported to 3.2: adjust context]
Signed-off-by: Ben Hutchings

mm: hugetlbfs: close race during teardown of hugetlbfs shared page tables

2012-08-09T23:25:10+00:00

commit d833352a4338dc31295ed832a30c9ccff5c7a183 upstream.

If a process creates a large hugetlbfs mapping that is eligible for page
table sharing and forks heavily with children some of whom fault and
others which destroy the mapping then it is possible for page tables to
get corrupted.  Some teardowns of the mapping encounter a "bad pmd" and
output a message to the kernel log.  The final teardown will trigger a
BUG_ON in mm/filemap.c.

This was reproduced in 3.4 but is known to have existed for a long time
and goes back at least as far as 2.6.37.  It was probably was introduced
in 2.6.20 by [39dde65c: shared page table for hugetlb page].  The messages
look like this;

[  ..........] Lots of bad pmd messages followed by this
[  127.164256] mm/memory.c:391: bad pmd ffff880412e04fe8(80000003de4000e7).
[  127.164257] mm/memory.c:391: bad pmd ffff880412e04ff0(80000003de6000e7).
[  127.164258] mm/memory.c:391: bad pmd ffff880412e04ff8(80000003de0000e7).
[  127.186778] ------------[ cut here ]------------
[  127.186781] kernel BUG at mm/filemap.c:134!
[  127.186782] invalid opcode: 0000 [#1] SMP
[  127.186783] CPU 7
[  127.186784] Modules linked in: af_packet cpufreq_conservative cpufreq_userspace cpufreq_powersave acpi_cpufreq mperf ext3 jbd dm_mod coretemp crc32c_intel usb_storage ghash_clmulni_intel aesni_intel i2c_i801 r8169 mii uas sr_mod cdrom sg iTCO_wdt iTCO_vendor_support shpchp serio_raw cryptd aes_x86_64 e1000e pci_hotplug dcdbas aes_generic container microcode ext4 mbcache jbd2 crc16 sd_mod crc_t10dif i915 drm_kms_helper drm i2c_algo_bit ehci_hcd ahci libahci usbcore rtc_cmos usb_common button i2c_core intel_agp video intel_gtt fan processor thermal thermal_sys hwmon ata_generic pata_atiixp libata scsi_mod
[  127.186801]
[  127.186802] Pid: 9017, comm: hugetlbfs-test Not tainted 3.4.0-autobuild #53 Dell Inc. OptiPlex 990/06D7TR
[  127.186804] RIP: 0010:[]  [] __delete_from_page_cache+0x15e/0x160
[  127.186809] RSP: 0000:ffff8804144b5c08  EFLAGS: 00010002
[  127.186810] RAX: 0000000000000001 RBX: ffffea000a5c9000 RCX: 00000000ffffffc0
[  127.186811] RDX: 0000000000000000 RSI: 0000000000000009 RDI: ffff88042dfdad00
[  127.186812] RBP: ffff8804144b5c18 R08: 0000000000000009 R09: 0000000000000003
[  127.186813] R10: 0000000000000000 R11: 000000000000002d R12: ffff880412ff83d8
[  127.186814] R13: ffff880412ff83d8 R14: 0000000000000000 R15: ffff880412ff83d8
[  127.186815] FS:  00007fe18ed2c700(0000) GS:ffff88042dce0000(0000) knlGS:0000000000000000
[  127.186816] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[  127.186817] CR2: 00007fe340000503 CR3: 0000000417a14000 CR4: 00000000000407e0
[  127.186818] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
[  127.186819] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7: 0000000000000400
[  127.186820] Process hugetlbfs-test (pid: 9017, threadinfo ffff8804144b4000, task ffff880417f803c0)
[  127.186821] Stack:
[  127.186822]  ffffea000a5c9000 0000000000000000 ffff8804144b5c48 ffffffff810ed83b
[  127.186824]  ffff8804144b5c48 000000000000138a 0000000000001387 ffff8804144b5c98
[  127.186825]  ffff8804144b5d48 ffffffff811bc925 ffff8804144b5cb8 0000000000000000
[  127.186827] Call Trace:
[  127.186829]  [] delete_from_page_cache+0x3b/0x80
[  127.186832]  [] truncate_hugepages+0x115/0x220
[  127.186834]  [] hugetlbfs_evict_inode+0x13/0x30
[  127.186837]  [] evict+0xa7/0x1b0
[  127.186839]  [] iput_final+0xd3/0x1f0
[  127.186840]  [] iput+0x39/0x50
[  127.186842]  [] d_kill+0xf8/0x130
[  127.186843]  [] dput+0xd2/0x1a0
[  127.186845]  [] __fput+0x170/0x230
[  127.186848]  [] ? rb_erase+0xce/0x150
[  127.186849]  [] fput+0x1d/0x30
[  127.186851]  [] remove_vma+0x37/0x80
[  127.186853]  [] do_munmap+0x2d2/0x360
[  127.186855]  [] sys_shmdt+0xc9/0x170
[  127.186857]  [] system_call_fastpath+0x16/0x1b
[  127.186858] Code: 0f 1f 44 00 00 48 8b 43 08 48 8b 00 48 8b 40 28 8b b0 40 03 00 00 85 f6 0f 88 df fe ff ff 48 89 df e8 e7 cb 05 00 e9 d2 fe ff ff <0f> 0b 55 83 e2 fd 48 89 e5 48 83 ec 30 48 89 5d d8 4c 89 65 e0
[  127.186868] RIP  [] __delete_from_page_cache+0x15e/0x160
[  127.186870]  RSP 
[  127.186871] ---[ end trace 7cbac5d1db69f426 ]---

The bug is a race and not always easy to reproduce.  To reproduce it I was
doing the following on a single socket I7-based machine with 16G of RAM.

$ hugeadm --pool-pages-max DEFAULT:13G
$ echo $((18*1048576*1024)) > /proc/sys/kernel/shmmax
$ echo $((18*1048576*1024)) > /proc/sys/kernel/shmall
$ for i in `seq 1 9000`; do ./hugetlbfs-test; done

On my particular machine, it usually triggers within 10 minutes but
enabling debug options can change the timing such that it never hits.
Once the bug is triggered, the machine is in trouble and needs to be
rebooted.  The machine will respond but processes accessing proc like "ps
aux" will hang due to the BUG_ON.  shutdown will also hang and needs a
hard reset or a sysrq-b.

The basic problem is a race between page table sharing and teardown.  For
the most part page table sharing depends on i_mmap_mutex.  In some cases,
it is also taking the mm->page_table_lock for the PTE updates but with
shared page tables, it is the i_mmap_mutex that is more important.

Unfortunately it appears to be also insufficient. Consider the following
situation

Process A					Process B
---------					---------
hugetlb_fault					shmdt
  						LockWrite(mmap_sem)
    						  do_munmap
						    unmap_region
						      unmap_vmas
						        unmap_single_vma
						          unmap_hugepage_range
      						            Lock(i_mmap_mutex)
							    Lock(mm->page_table_lock)
							    huge_pmd_unshare/unmap tables <--- (1)
							    Unlock(mm->page_table_lock)
      						            Unlock(i_mmap_mutex)
  huge_pte_alloc				      ...
    Lock(i_mmap_mutex)				      ...
    vma_prio_walk, find svma, spte		      ...
    Lock(mm->page_table_lock)			      ...
    share spte					      ...
    Unlock(mm->page_table_lock)			      ...
    Unlock(i_mmap_mutex)			      ...
  hugetlb_no_page									  <--- (2)
						      free_pgtables
						        unlink_file_vma
							hugetlb_free_pgd_range
						    remove_vma_list

In this scenario, it is possible for Process A to share page tables with
Process B that is trying to tear them down.  The i_mmap_mutex on its own
does not prevent Process A walking Process B's page tables.  At (1) above,
the page tables are not shared yet so it unmaps the PMDs.  Process A sets
up page table sharing and at (2) faults a new entry.  Process B then trips
up on it in free_pgtables.

This patch fixes the problem by adding a new function
__unmap_hugepage_range_final that is only called when the VMA is about to
be destroyed.  This function clears VM_MAYSHARE during
unmap_hugepage_range() under the i_mmap_mutex.  This makes the VMA
ineligible for sharing and avoids the race.  Superficially this looks like
it would then be vunerable to truncate and madvise issues but hugetlbfs
has its own truncate handlers so does not use unmap_mapping_range() and
does not support madvise(DONTNEED).

This should be treated as a -stable candidate if it is merged.

Test program is as follows. The test case was mostly written by Michal
Hocko with a few minor changes to reproduce this bug.

==== CUT HERE ====

static size_t huge_page_size = (2UL << 20);
static size_t nr_huge_page_A = 512;
static size_t nr_huge_page_B = 5632;

unsigned int get_random(unsigned int max)
{
	struct timeval tv;

	gettimeofday(&tv, NULL);
	srandom(tv.tv_usec);
	return random() % max;
}

static void play(void *addr, size_t size)
{
	unsigned char *start = addr,
		      *end = start + size,
		      *a;
	start += get_random(size/2);

	/* we could itterate on huge pages but let's give it more time. */
	for (a = start; a < end; a += 4096)
		*a = 0;
}

int main(int argc, char **argv)
{
	key_t key = IPC_PRIVATE;
	size_t sizeA = nr_huge_page_A * huge_page_size;
	size_t sizeB = nr_huge_page_B * huge_page_size;
	int shmidA, shmidB;
	void *addrA = NULL, *addrB = NULL;
	int nr_children = 300, n = 0;

	if ((shmidA = shmget(key, sizeA, IPC_CREAT|SHM_HUGETLB|0660)) == -1) {
		perror("shmget:");
		return 1;
	}

	if ((addrA = shmat(shmidA, addrA, SHM_R|SHM_W)) == (void *)-1UL) {
		perror("shmat");
		return 1;
	}
	if ((shmidB = shmget(key, sizeB, IPC_CREAT|SHM_HUGETLB|0660)) == -1) {
		perror("shmget:");
		return 1;
	}

	if ((addrB = shmat(shmidB, addrB, SHM_R|SHM_W)) == (void *)-1UL) {
		perror("shmat");
		return 1;
	}

fork_child:
	switch(fork()) {
		case 0:
			switch (n%3) {
			case 0:
				play(addrA, sizeA);
				break;
			case 1:
				play(addrB, sizeB);
				break;
			case 2:
				break;
			}
			break;
		case -1:
			perror("fork:");
			break;
		default:
			if (++n < nr_children)
				goto fork_child;
			play(addrA, sizeA);
			break;
	}
	shmdt(addrA);
	shmdt(addrB);
	do {
		wait(NULL);
	} while (--n > 0);
	shmctl(shmidA, IPC_RMID, NULL);
	shmctl(shmidB, IPC_RMID, NULL);
	return 0;
}

[akpm@linux-foundation.org: name the declaration's args, fix CONFIG_HUGETLBFS=n build]
Signed-off-by: Hugh Dickins 
Reviewed-by: Michal Hocko 
Signed-off-by: Mel Gorman 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds 
[bwh: Backported to 3.2:
 - Adjust context
 - Drop the mmu_gather * parameters]
Signed-off-by: Ben Hutchings

mm/hugetlb: fix warning in alloc_huge_page/dequeue_huge_page_vma

2012-08-02T13:37:37+00:00

commit b1c12cbcd0a02527c180a862e8971e249d3b347d upstream.

Stable note: Not tracked in Bugzilla. [get|put]_mems_allowed() is extremely
	expensive and severely impacted page allocator performance. This
	is part of a series of patches that reduce page allocator overhead.

Fix a gcc warning (and bug?) introduced in cc9a6c877 ("cpuset: mm: reduce
large amounts of memory barrier related damage v3")

Local variable "page" can be uninitialized if the nodemask from vma policy
does not intersects with nodemask from cpuset.  Even if it doesn't happens
it is better to initialize this variable explicitly than to introduce
a kernel oops in a weird corner case.

mm/hugetlb.c: In function `alloc_huge_page':
mm/hugetlb.c:1135:5: warning: `page' may be used uninitialized in this function

Signed-off-by: Konstantin Khlebnikov 
Acked-by: Mel Gorman 
Acked-by: David Rientjes 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds 
Signed-off-by: Mel Gorman 
Signed-off-by: Ben Hutchings

cpuset: mm: reduce large amounts of memory barrier related damage v3

2012-08-02T13:37:36+00:00

commit cc9a6c8776615f9c194ccf0b63a0aa5628235545 upstream.

Stable note:  Not tracked in Bugzilla. [get|put]_mems_allowed() is extremely
	expensive and severely impacted page allocator performance. This
	is part of a series of patches that reduce page allocator overhead.

Commit c0ff7453bb5c ("cpuset,mm: fix no node to alloc memory when
changing cpuset's mems") wins a super prize for the largest number of
memory barriers entered into fast paths for one commit.

[get|put]_mems_allowed is incredibly heavy with pairs of full memory
barriers inserted into a number of hot paths.  This was detected while
investigating at large page allocator slowdown introduced some time
after 2.6.32.  The largest portion of this overhead was shown by
oprofile to be at an mfence introduced by this commit into the page
allocator hot path.

For extra style points, the commit introduced the use of yield() in an
implementation of what looks like a spinning mutex.

This patch replaces the full memory barriers on both read and write
sides with a sequence counter with just read barriers on the fast path
side.  This is much cheaper on some architectures, including x86.  The
main bulk of the patch is the retry logic if the nodemask changes in a
manner that can cause a false failure.

While updating the nodemask, a check is made to see if a false failure
is a risk.  If it is, the sequence number gets bumped and parallel
allocators will briefly stall while the nodemask update takes place.

In a page fault test microbenchmark, oprofile samples from
__alloc_pages_nodemask went from 4.53% of all samples to 1.15%.  The
actual results were

                             3.3.0-rc3          3.3.0-rc3
                             rc3-vanilla        nobarrier-v2r1
    Clients   1 UserTime       0.07 (  0.00%)   0.08 (-14.19%)
    Clients   2 UserTime       0.07 (  0.00%)   0.07 (  2.72%)
    Clients   4 UserTime       0.08 (  0.00%)   0.07 (  3.29%)
    Clients   1 SysTime        0.70 (  0.00%)   0.65 (  6.65%)
    Clients   2 SysTime        0.85 (  0.00%)   0.82 (  3.65%)
    Clients   4 SysTime        1.41 (  0.00%)   1.41 (  0.32%)
    Clients   1 WallTime       0.77 (  0.00%)   0.74 (  4.19%)
    Clients   2 WallTime       0.47 (  0.00%)   0.45 (  3.73%)
    Clients   4 WallTime       0.38 (  0.00%)   0.37 (  1.58%)
    Clients   1 Flt/sec/cpu  497620.28 (  0.00%) 520294.53 (  4.56%)
    Clients   2 Flt/sec/cpu  414639.05 (  0.00%) 429882.01 (  3.68%)
    Clients   4 Flt/sec/cpu  257959.16 (  0.00%) 258761.48 (  0.31%)
    Clients   1 Flt/sec      495161.39 (  0.00%) 517292.87 (  4.47%)
    Clients   2 Flt/sec      820325.95 (  0.00%) 850289.77 (  3.65%)
    Clients   4 Flt/sec      1020068.93 (  0.00%) 1022674.06 (  0.26%)
    MMTests Statistics: duration
    Sys Time Running Test (seconds)             135.68    132.17
    User+Sys Time Running Test (seconds)         164.2    160.13
    Total Elapsed Time (seconds)                123.46    120.87

The overall improvement is small but the System CPU time is much
improved and roughly in correlation to what oprofile reported (these
performance figures are without profiling so skew is expected).  The
actual number of page faults is noticeably improved.

For benchmarks like kernel builds, the overall benefit is marginal but
the system CPU time is slightly reduced.

To test the actual bug the commit fixed I opened two terminals.  The
first ran within a cpuset and continually ran a small program that
faulted 100M of anonymous data.  In a second window, the nodemask of the
cpuset was continually randomised in a loop.

Without the commit, the program would fail every so often (usually
within 10 seconds) and obviously with the commit everything worked fine.
With this patch applied, it also worked fine so the fix should be
functionally equivalent.

Signed-off-by: Mel Gorman 
Cc: Miao Xie 
Cc: David Rientjes 
Cc: Peter Zijlstra 
Cc: Christoph Lameter 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds 
Signed-off-by: Mel Gorman 
[bwh: Forward-ported from 3.0 to 3.2: apply the upstream changes
 to get_any_partial()]
Signed-off-by: Ben Hutchings

hugepages: fix use after free bug in "quota" handling

2012-07-25T03:11:01+00:00

commit 90481622d75715bfcb68501280a917dbfe516029 upstream.

hugetlbfs_{get,put}_quota() are badly named.  They don't interact with the
general quota handling code, and they don't much resemble its behaviour.
Rather than being about maintaining limits on on-disk block usage by
particular users, they are instead about maintaining limits on in-memory
page usage (including anonymous MAP_PRIVATE copied-on-write pages)
associated with a particular hugetlbfs filesystem instance.

Worse, they work by having callbacks to the hugetlbfs filesystem code from
the low-level page handling code, in particular from free_huge_page().
This is a layering violation of itself, but more importantly, if the
kernel does a get_user_pages() on hugepages (which can happen from KVM
amongst others), then the free_huge_page() can be delayed until after the
associated inode has already been freed.  If an unmount occurs at the
wrong time, even the hugetlbfs superblock where the "quota" limits are
stored may have been freed.

Andrew Barry proposed a patch to fix this by having hugepages, instead of
storing a pointer to their address_space and reaching the superblock from
there, had the hugepages store pointers directly to the superblock,
bumping the reference count as appropriate to avoid it being freed.
Andrew Morton rejected that version, however, on the grounds that it made
the existing layering violation worse.

This is a reworked version of Andrew's patch, which removes the extra, and
some of the existing, layering violation.  It works by introducing the
concept of a hugepage "subpool" at the lower hugepage mm layer - that is a
finite logical pool of hugepages to allocate from.  hugetlbfs now creates
a subpool for each filesystem instance with a page limit set, and a
pointer to the subpool gets added to each allocated hugepage, instead of
the address_space pointer used now.  The subpool has its own lifetime and
is only freed once all pages in it _and_ all other references to it (i.e.
superblocks) are gone.

subpools are optional - a NULL subpool pointer is taken by the code to
mean that no subpool limits are in effect.

Previous discussion of this bug found in:  "Fix refcounting in hugetlbfs
quota handling.". See:  https://lkml.org/lkml/2011/8/11/28 or
http://marc.info/?l=linux-mm&m=126928970510627&w=1

v2: Fixed a bug spotted by Hillf Danton, and removed the extra parameter to
alloc_huge_page() - since it already takes the vma, it is not necessary.

Signed-off-by: Andrew Barry 
Signed-off-by: David Gibson 
Cc: Hugh Dickins 
Cc: Mel Gorman 
Cc: Minchan Kim 
Cc: Hillf Danton 
Cc: Paul Mackerras 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds 
[bwh: Backported to 3.2: adjust context to apply after commit
 c50ac050811d6485616a193eb0f37bfbd191cc89 'hugetlb: fix resv_map leak in
 error path', backported in 3.2.20]
Signed-off-by: Ben Hutchings

mm: fix vma_resv_map() NULL pointer

2012-06-10T13:41:47+00:00

commit 4523e1458566a0e8ecfaff90f380dd23acc44d27 upstream.

hugetlb_reserve_pages() can be used for either normal file-backed
hugetlbfs mappings, or MAP_HUGETLB.  In the MAP_HUGETLB, semi-anonymous
mode, there is not a VMA around.  The new call to resv_map_put() assumed
that there was, and resulted in a NULL pointer dereference:

  BUG: unable to handle kernel NULL pointer dereference at 0000000000000030
  IP: vma_resv_map+0x9/0x30
  PGD 141453067 PUD 1421e1067 PMD 0
  Oops: 0000 [#1] PREEMPT SMP
  ...
  Pid: 14006, comm: trinity-child6 Not tainted 3.4.0+ #36
  RIP: vma_resv_map+0x9/0x30
  ...
  Process trinity-child6 (pid: 14006, threadinfo ffff8801414e0000, task ffff8801414f26b0)
  Call Trace:
    resv_map_put+0xe/0x40
    hugetlb_reserve_pages+0xa6/0x1d0
    hugetlb_file_setup+0x102/0x2c0
    newseg+0x115/0x360
    ipcget+0x1ce/0x310
    sys_shmget+0x5a/0x60
    system_call_fastpath+0x16/0x1b

This was reported by Dave Jones, but was reproducible with the
libhugetlbfs test cases, so shame on me for not running them in the
first place.

With this, the oops is gone, and the output of libhugetlbfs's
run_tests.py is identical to plain 3.4 again.

[ Marked for stable, since this was introduced by commit c50ac050811d
  ("hugetlb: fix resv_map leak in error path") which was also marked for
  stable ]

Reported-by: Dave Jones 
Cc: Mel Gorman 
Cc: KOSAKI Motohiro 
Cc: Christoph Lameter 
Cc: Andrea Arcangeli 
Cc: Andrew Morton 
Signed-off-by: Linus Torvalds 
Signed-off-by: Ben Hutchings

hugetlb: fix resv_map leak in error path

2012-06-10T13:41:45+00:00

commit c50ac050811d6485616a193eb0f37bfbd191cc89 upstream.

When called for anonymous (non-shared) mappings, hugetlb_reserve_pages()
does a resv_map_alloc().  It depends on code in hugetlbfs's
vm_ops->close() to release that allocation.

However, in the mmap() failure path, we do a plain unmap_region() without
the remove_vma() which actually calls vm_ops->close().

This is a decent fix.  This leak could get reintroduced if new code (say,
after hugetlb_reserve_pages() in hugetlbfs_file_mmap()) decides to return
an error.  But, I think it would have to unroll the reservation anyway.

Christoph's test case:

	http://marc.info/?l=linux-mm&m=133728900729735

Signed-off-by: Dave Hansen 
[Christoph Lameter: I have rediffed the patch against 2.6.32 and 3.2.0.]
Signed-off-by: Ben Hutchings

hugetlb: prevent BUG_ON in hugetlb_fault() -> hugetlb_cow()

2012-05-20T21:56:32+00:00

commit 4998a6c0edce7fae9c0a5463f6ec3fa585258ee7 upstream.

Commit 66aebce747eaf ("hugetlb: fix race condition in hugetlb_fault()")
added code to avoid a race condition by elevating the page refcount in
hugetlb_fault() while calling hugetlb_cow().

However, one code path in hugetlb_cow() includes an assertion that the
page count is 1, whereas it may now also have the value 2 in this path.

The consensus is that this BUG_ON has served its purpose, so rather than
extending it to cover both cases, we just remove it.

Signed-off-by: Chris Metcalf 
Acked-by: Mel Gorman 
Acked-by: Hillf Danton 
Acked-by: Hugh Dickins 
Cc: Michal Hocko 
Cc: KAMEZAWA Hiroyuki 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds 
Signed-off-by: Ben Hutchings

hugetlb: fix race condition in hugetlb_fault()

2012-04-22T22:30:59+00:00

commit 66aebce747eaf9bc456bf1f1b217d8db843031d0 upstream.

The race is as follows:

Suppose a multi-threaded task forks a new process (on cpu A), thus
bumping up the ref count on all the pages.  While the fork is occurring
(and thus we have marked all the PTEs as read-only), another thread in
the original process (on cpu B) tries to write to a huge page, taking an
access violation from the write-protect and calling hugetlb_cow().  Now,
suppose the fork() fails.  It will undo the COW and decrement the ref
count on the pages, so the ref count on the huge page drops back to 1.
Meanwhile hugetlb_cow() also decrements the ref count by one on the
original page, since the original address space doesn't need it any
more, having copied a new page to replace the original page.  This
leaves the ref count at zero, and when we call unlock_page(), we panic.

	fork on CPU A				fault on CPU B
	=============				==============
	...
	down_write(&parent->mmap_sem);
	down_write_nested(&child->mmap_sem);
	...
	while duplicating vmas
		if error
			break;
	...
	up_write(&child->mmap_sem);
	up_write(&parent->mmap_sem);		...
						down_read(&parent->mmap_sem);
						...
						lock_page(page);
						handle COW
						page_mapcount(old_page) == 2
						alloc and prepare new_page
	...
	handle error
	page_remove_rmap(page);
	put_page(page);
	...
						fold new_page into pte
						page_remove_rmap(page);
						put_page(page);
						...
				oops ==>	unlock_page(page);
						up_read(&parent->mmap_sem);

The solution is to take an extra reference to the page while we are
holding the lock on it.

Signed-off-by: Chris Metcalf 
Cc: Hillf Danton 
Cc: Michal Hocko 
Cc: KAMEZAWA Hiroyuki 
Cc: Hugh Dickins 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds 
Signed-off-by: Greg Kroah-Hartman