Message-ID: <20120724093406.GO9222@suse.de>
Date:	Tue, 24 Jul 2012 10:34:06 +0100
From:	Mel Gorman <mgorman@...e.de>
To:	Hugh Dickins <hughd@...gle.com>
Cc:	Michal Hocko <mhocko@...e.cz>, Linux-MM <linux-mm@...ck.org>,
	David Gibson <david@...son.dropbear.id.au>,
	Ken Chen <kenchen@...gle.com>,
	Cong Wang <xiyou.wangcong@...il.com>,
	LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH -alternative] mm: hugetlbfs: Close race during teardown
 of hugetlbfs shared page tables V2 (resend)

On Mon, Jul 23, 2012 at 06:08:05PM -0700, Hugh Dickins wrote:
> On Mon, 23 Jul 2012, Mel Gorman wrote:
> > On Sun, Jul 22, 2012 at 09:04:33PM -0700, Hugh Dickins wrote:
> > > On Fri, 20 Jul 2012, Mel Gorman wrote:
> > > > On Fri, Jul 20, 2012 at 04:36:35PM +0200, Michal Hocko wrote:
> > 
> > I like it in that it's simple and I can confirm it works for the test case
> > of interest.
> 
> Phew, I'm glad to hear that, thanks.
> 
> > 
> > However, is your patch not vunerable to truncate issues?
> > madvise()/truncate() issues was the main reason why I was wary of VMA tricks
> > as a solution. As it turns out, madvise(DONTNEED) is not a problem as it is
> > ignored for hugetlbfs but I think truncate is still problematic. Lets say
> > we mmap(MAP_SHARED) a hugetlbfs file and then truncate for whatever reason.
> > 
> > invalidate_inode_pages2
> >   invalidate_inode_pages2_range
> >     unmap_mapping_range_vma
> >       zap_page_range_single
> >         unmap_single_vma
> > 	  __unmap_hugepage_range (removes VM_MAYSHARE)
> > 
> > The VMA still exists so the consequences for this would be varied but
> > minimally fault is going to be "interesting".
> 
> You had me worried there, I hadn't considered truncation or invalidation2
> at all.
> 
> But actually, I don't think they do pose any problem for my patch.  They
> would indeed if I were removing VM_MAYSHARE in __unmap_hugepage_range()
> as you show above; but no, I'm removing it in unmap_hugepage_range().
> 

True, I hadn't considered the distinction to be relevant. My train of
thought was "we potentially end up calling these functions and affect
VMAs that are not being torn down". This set off an alarm but I didn't
follow through properly.

> That's only called by unmap_single_vma(): which is called via
> unmap_vmas() by unmap_region() or exit_mmap() just before free_pgtables()
> (the problem cases); or by madvise_dontneed() via zap_page_range(), which
> as you note is disallowed on VM_HUGETLB; or by zap_page_range_single().
> 

The madvise(DONTNEED) case also made me worry that if this fix worked, it
worked by co-incidence. If someone "fixes" hugetlbfs to support DONTNEED
(which would be nonsensical) or direct IO (which would be weird) then
this fix potentially regressed.

> zap_page_range_single() is called by zap_vma_ptes(), which is only
> allowed on VM_PFNMAP; or by unmap_mapping_range_vma(), which looked
> like it was going to deadlock on i_mmap_mutex (with or without my
> patch) until I realized that hugetlbfs has its own hugetlbfs_setattr()
> and hugetlb_vmtruncate() which don't use unmap_mapping_range() at all.
> 

Yep, it happens to work.

> invalidate_inode_pages2() (and _range()) do use unmap_mapping_range(),
> but hugetlbfs doesn't support direct_IO, and otherwise I think they're
> called by a filesystem directly on its own inodes, which hugetlbfs
> does not. Anyway, if there's a deadlock on i_mmap_mutex somewhere
> in there, it's not introduced by the proposed patch.
> 
> So, unmap_hugepage_range() is only being called in the problem cases,
> just before free_pgtables(), when unmapping a vma (with mmap_sem held),
> or when exiting (when we have the last reference to mm): in each case,
> the vma is on its way out, and VM_MAYSHARE no longer of interest to others.
> 
> I spent a while concerned that I'd overlooked the truncation case, before
> realizing that it's not a problem: the issue comes when we free_pgtables(),
> which truncation makes no attempt to do.
> 
> So, after a bout of anxiety, I think my &= ~VM_MAYSHARE remains good.
> 

I agree with you. When I was thinking about the potential problems, I was
thinking of them in the general context of the core VM and what we normally
take into account.

I confess that I really find this working-by-coincidence very icky and am
uncomfortable with it but your patch is the only patch that contains the
mess to hugetlbfs. I fixed exit_mmap() for my version but only by changing
the core to introduce exit_vmas() to take mmap_sem for write if a hugetlb
VMA is found so I also affected the core.

So, lets go with your patch but with all this documented! I stuck a
changelog and an additional comment onto your patch and this is the end
result.

Do you want to pick this up and send it to Andrew or will I?

Thanks Hugh!

---8<---
mm: hugetlbfs: Close race during teardown of hugetlbfs shared page tables

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:[<ffffffff810ed6ce>]  [<ffffffff810ed6ce>] __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]  [<ffffffff810ed83b>] delete_from_page_cache+0x3b/0x80
[  127.186832]  [<ffffffff811bc925>] truncate_hugepages+0x115/0x220
[  127.186834]  [<ffffffff811bca43>] hugetlbfs_evict_inode+0x13/0x30
[  127.186837]  [<ffffffff811655c7>] evict+0xa7/0x1b0
[  127.186839]  [<ffffffff811657a3>] iput_final+0xd3/0x1f0
[  127.186840]  [<ffffffff811658f9>] iput+0x39/0x50
[  127.186842]  [<ffffffff81162708>] d_kill+0xf8/0x130
[  127.186843]  [<ffffffff81162812>] dput+0xd2/0x1a0
[  127.186845]  [<ffffffff8114e2d0>] __fput+0x170/0x230
[  127.186848]  [<ffffffff81236e0e>] ? rb_erase+0xce/0x150
[  127.186849]  [<ffffffff8114e3ad>] fput+0x1d/0x30
[  127.186851]  [<ffffffff81117db7>] remove_vma+0x37/0x80
[  127.186853]  [<ffffffff81119182>] do_munmap+0x2d2/0x360
[  127.186855]  [<ffffffff811cc639>] sys_shmdt+0xc9/0x170
[  127.186857]  [<ffffffff81410a39>] 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  [<ffffffff810ed6ce>] __delete_from_page_cache+0x15e/0x160
[  127.186870]  RSP <ffff8804144b5c08>
[  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 clearing 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 problems but
this is avoided by the limitations of hugetlbfs.

madvise and trunctate would be problems if removing VM_MAYSHARE in
__unmap_hugepage_range() but it is removed in unmap_hugepage_range().
This is only called by unmap_single_vma(): which is called via unmap_vmas()
by unmap_region() or exit_mmap() just before free_pgtables() (the problem
cases); or by madvise_dontneed() via zap_page_range(), which is disallowed
on VM_HUGETLB; or by zap_page_range_single().

zap_page_range_single() is called by zap_vma_ptes(), which is only allowed
on VM_PFNMAP; or by unmap_mapping_range_vma(), which looked like it was
going to deadlock on i_mmap_mutex (with or without my patch) but does
not as hugetlbfs has its own hugetlbfs_setattr() and hugetlb_vmtruncate()
which don't use unmap_mapping_range() at all.

invalidate_inode_pages2() (and _range()) do use unmap_mapping_range(),
but hugetlbfs doesn't support direct_IO, and otherwise they're called by a
filesystem directly on its own inodes, which hugetlbfs does not.  If there's
a deadlock on i_mmap_mutex somewhere in there, it's not introduced by the
proposed patch.

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;
}

Signed-off-by: Hugh Dickins <hughd@...gle.com>
Reviewed-by: Michal Hocko <mhocko@...e.cz>
Reviewed-by: Mel Gorman <mgorman@...e.de>
---
 mm/hugetlb.c |   25 +++++++++++++++++++++++--
 1 file changed, 23 insertions(+), 2 deletions(-)

diff --git a/mm/hugetlb.c b/mm/hugetlb.c
index ae8f708..d488476 100644
--- a/mm/hugetlb.c
+++ b/mm/hugetlb.c
@@ -2383,6 +2383,22 @@ void unmap_hugepage_range(struct vm_area_struct *vma, unsigned long start,
 {
 	mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
 	__unmap_hugepage_range(vma, start, end, ref_page);
+	/*
+	 * Clear this flag so that x86's huge_pmd_share page_table_shareable
+	 * test will fail on a vma being torn down, and not grab a page table
+	 * on its way out.  We're lucky that the flag has such an appropriate
+	 * name, and can in fact be safely cleared here. We could clear it
+	 * before the __unmap_hugepage_range above, but all that's necessary
+	 * is to clear it before releasing the i_mmap_mutex below.
+	 *
+	 * This works because in the contexts this is called, the VMA is
+	 * going to be destroyed. It is not vunerable to madvise(DONTNEED)
+	 * because madvise is not supported on hugetlbfs. The same applies
+	 * for direct IO. unmap_hugepage_range() is only being called just
+	 * before free_pgtables() so clearing VM_MAYSHARE will not cause
+	 * surprises later.
+	 */
+	vma->vm_flags &= ~VM_MAYSHARE;
 	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
 }
 
@@ -2949,9 +2965,14 @@ void hugetlb_change_protection(struct vm_area_struct *vma,
 		}
 	}
 	spin_unlock(&mm->page_table_lock);
-	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
-
+	/*
+	 * Must flush TLB before releasing i_mmap_mutex: x86's huge_pmd_unshare
+	 * may have cleared our pud entry and done put_page on the page table:
+	 * once we release i_mmap_mutex, another task can do the final put_page
+	 * and that page table be reused and filled with junk.
+	 */
 	flush_tlb_range(vma, start, end);
+	mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
 }
 
 int hugetlb_reserve_pages(struct inode *inode,
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists