All of lore.kernel.org
 help / color / mirror / Atom feed
From: Nadav Amit <namit@vmware.com>
To: Huang Ying <ying.huang@intel.com>
Cc: Andrew Morton <akpm@linux-foundation.org>,
	Linux-MM <linux-mm@kvack.org>,
	LKML <linux-kernel@vger.kernel.org>,
	syzbot <syzbot+aa5bebed695edaccf0df@syzkaller.appspotmail.com>,
	Mel Gorman <mgorman@techsingularity.net>,
	Andrea Arcangeli <aarcange@redhat.com>,
	Andy Lutomirski <luto@kernel.org>,
	Dave Hansen <dave.hansen@linux.intel.com>,
	Will Deacon <will@kernel.org>, Yu Zhao <yuzhao@google.com>,
	Marco Elver <elver@google.com>
Subject: Re: [PATCH -V2] mm/rmap: fix potential batched TLB flush race
Date: Thu, 25 Nov 2021 08:39:31 +0000	[thread overview]
Message-ID: <7127B0FC-A00D-48C3-9C01-0B85FB1BD63F@vmware.com> (raw)
In-Reply-To: <20211125065026.142735-1-ying.huang@intel.com>



> On Nov 24, 2021, at 10:50 PM, Huang Ying <ying.huang@intel.com> wrote:
> 
> In theory, the following race is possible for batched TLB flushing.
> 
> CPU0                               CPU1
> ----                               ----
> shrink_page_list()
>                                   unmap
>                                     zap_pte_range()
>                                       flush_tlb_batched_pending()
>                                         flush_tlb_mm()
>  try_to_unmap()
>    set_tlb_ubc_flush_pending()
>      mm->tlb_flush_batched = true
>                                         mm->tlb_flush_batched = false
> 
> After the TLB is flushed on CPU1 via flush_tlb_mm() and before
> mm->tlb_flush_batched is set to false, some PTE is unmapped on CPU0
> and the TLB flushing is pended.  Then the pended TLB flushing will be
> lost.  Although both set_tlb_ubc_flush_pending() and
> flush_tlb_batched_pending() are called with PTL locked, different PTL
> instances may be used.
> 
> Because the race window is really small, and the lost TLB flushing
> will cause problem only if a TLB entry is inserted before the
> unmapping in the race window, the race is only theoretical.  But the
> fix is simple and cheap too.
> 
> Syzbot has reported this too as follows,
> 
> ==================================================================
> BUG: KCSAN: data-race in flush_tlb_batched_pending / try_to_unmap_one
> 
> write to 0xffff8881072cfbbc of 1 bytes by task 17406 on cpu 1:
> flush_tlb_batched_pending+0x5f/0x80 mm/rmap.c:691
> madvise_free_pte_range+0xee/0x7d0 mm/madvise.c:594
> walk_pmd_range mm/pagewalk.c:128 [inline]
> walk_pud_range mm/pagewalk.c:205 [inline]
> walk_p4d_range mm/pagewalk.c:240 [inline]
> walk_pgd_range mm/pagewalk.c:277 [inline]
> __walk_page_range+0x981/0x1160 mm/pagewalk.c:379
> walk_page_range+0x131/0x300 mm/pagewalk.c:475
> madvise_free_single_vma mm/madvise.c:734 [inline]
> madvise_dontneed_free mm/madvise.c:822 [inline]
> madvise_vma mm/madvise.c:996 [inline]
> do_madvise+0xe4a/0x1140 mm/madvise.c:1202
> __do_sys_madvise mm/madvise.c:1228 [inline]
> __se_sys_madvise mm/madvise.c:1226 [inline]
> __x64_sys_madvise+0x5d/0x70 mm/madvise.c:1226
> do_syscall_x64 arch/x86/entry/common.c:50 [inline]
> do_syscall_64+0x44/0xd0 arch/x86/entry/common.c:80
> entry_SYSCALL_64_after_hwframe+0x44/0xae
> 
> write to 0xffff8881072cfbbc of 1 bytes by task 71 on cpu 0:
> set_tlb_ubc_flush_pending mm/rmap.c:636 [inline]
> try_to_unmap_one+0x60e/0x1220 mm/rmap.c:1515
> rmap_walk_anon+0x2fb/0x470 mm/rmap.c:2301
> try_to_unmap+0xec/0x110
> shrink_page_list+0xe91/0x2620 mm/vmscan.c:1719
> shrink_inactive_list+0x3fb/0x730 mm/vmscan.c:2394
> shrink_list mm/vmscan.c:2621 [inline]
> shrink_lruvec+0x3c9/0x710 mm/vmscan.c:2940
> shrink_node_memcgs+0x23e/0x410 mm/vmscan.c:3129
> shrink_node+0x8f6/0x1190 mm/vmscan.c:3252
> kswapd_shrink_node mm/vmscan.c:4022 [inline]
> balance_pgdat+0x702/0xd30 mm/vmscan.c:4213
> kswapd+0x200/0x340 mm/vmscan.c:4473
> kthread+0x2c7/0x2e0 kernel/kthread.c:327
> ret_from_fork+0x1f/0x30
> 
> value changed: 0x01 -> 0x00
> 
> Reported by Kernel Concurrency Sanitizer on:
> CPU: 0 PID: 71 Comm: kswapd0 Not tainted 5.16.0-rc1-syzkaller #0
> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
> ==================================================================
> 
> Signed-off-by: "Huang, Ying" <ying.huang@intel.com>
> Reported-by: syzbot+aa5bebed695edaccf0df@syzkaller.appspotmail.com
> Cc: Nadav Amit <namit@vmware.com>
> Cc: Mel Gorman <mgorman@techsingularity.net>
> Cc: Andrea Arcangeli <aarcange@redhat.com>
> Cc: Andy Lutomirski <luto@kernel.org>
> Cc: Dave Hansen <dave.hansen@linux.intel.com>
> Cc: Will Deacon <will@kernel.org>
> Cc: Yu Zhao <yuzhao@google.com>
> Cc: Marco Elver <elver@google.com>
> ---
> include/linux/mm_types.h |  2 +-
> mm/rmap.c                | 30 +++++++++++++++++++++++-------
> 2 files changed, 24 insertions(+), 8 deletions(-)
> 
> diff --git a/include/linux/mm_types.h b/include/linux/mm_types.h
> index c3a6e6209600..789778067db9 100644
> --- a/include/linux/mm_types.h
> +++ b/include/linux/mm_types.h
> @@ -632,7 +632,7 @@ struct mm_struct {
> 		atomic_t tlb_flush_pending;
> #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH
> 		/* See flush_tlb_batched_pending() */
> -		bool tlb_flush_batched;
> +		atomic_t tlb_flush_batched;
> #endif
> 		struct uprobes_state uprobes_state;
> #ifdef CONFIG_PREEMPT_RT
> diff --git a/mm/rmap.c b/mm/rmap.c
> index 163ac4e6bcee..2e6b19be5a18 100644
> --- a/mm/rmap.c
> +++ b/mm/rmap.c
> @@ -621,6 +621,18 @@ void try_to_unmap_flush_dirty(void)
> 		try_to_unmap_flush();
> }
> 
> +/* The upper 15 bit of mm->tlb_flush_batched records pending flushes */

Why 15? I think it will be easier to swallow if it was 32-bit (which
correspond to number of PIDs?)

What would happen on an overflow? If you regarded each pneding/flushed
counter as a separate atomic, that would have been easier. But anyhow,
a comment is necessary IMHO.

> +#define TLB_FLUSH_BATCH_PENDING_SHIFT	16
> +#define TLB_FLUSH_BATCH_COUNT_MASK	0x7f

0x7f is not 15 bits the last time I checked.

> +#define TLB_FLUSH_BATCH_PENDING_ONE	(1 << TLB_FLUSH_BATCH_PENDING_SHIFT)
> +
> +#define TLB_FLUSH_BATCH_PENDING(cnt)					\
> +	(((cnt) >> TLB_FLUSH_BATCH_PENDING_SHIFT) & TLB_FLUSH_BATCH_COUNT_MASK)
> +#define TLB_FLUSH_BATCH_FLUSHED(cnt)					\
> +	((cnt) & TLB_FLUSH_BATCH_COUNT_MASK)
> +#define TLB_FLUSH_BATCH_PACK(pending, flushed)				\
> +	(((pending) << TLB_FLUSH_BATCH_PENDING_SHIFT) | (flushed))

I would have preferred, when possible to avoid such macros. It just makes
reading the code harder.

> +
> static void set_tlb_ubc_flush_pending(struct mm_struct *mm, bool writable)
> {
> 	struct tlbflush_unmap_batch *tlb_ubc = &current->tlb_ubc;
> @@ -633,7 +645,7 @@ static void set_tlb_ubc_flush_pending(struct mm_struct *mm, bool writable)
> 	 * before the PTE is cleared.
> 	 */
> 	barrier();
> -	mm->tlb_flush_batched = true;
> +	atomic_add(TLB_FLUSH_BATCH_PENDING_ONE, &mm->tlb_flush_batched);

Any reason to put the pending in the top bits instead of the low ones?
It can at least simplify the code a bit. As for the barrier, I would
change it for smp_mb__before_atomic() or smp_wmb(). You want the
PTE changes and the counter updates to be ordered.

> 
> 	/*
> 	 * If the PTE was dirty then it's best to assume it's writable. The
> @@ -680,15 +692,19 @@ static bool should_defer_flush(struct mm_struct *mm, enum ttu_flags flags)
>  */
> void flush_tlb_batched_pending(struct mm_struct *mm)
> {
> -	if (data_race(mm->tlb_flush_batched)) {
> -		flush_tlb_mm(mm);

The previous smp_mb__before_atomic() or smp_wmb() should be matched
with an smp_mb__before_atomic() or smp_rmb() here, I think.

> +	int batched = atomic_read(&mm->tlb_flush_batched);
> +	int pending = TLB_FLUSH_BATCH_PENDING(batched);
> +	int flushed = TLB_FLUSH_BATCH_FLUSHED(batched);

I would prefer them being unsigned.

> 
> +	if (pending != flushed) {
> +		flush_tlb_mm(mm);
> 		/*
> -		 * Do not allow the compiler to re-order the clearing of
> -		 * tlb_flush_batched before the tlb is flushed.
> +		 * If the new TLB flushing is pended during flushing,
> +		 * leave mm->tlb_flush_batched as is, to avoid to lose
> +		 * flushing.
> 		 */
> -		barrier();
> -		mm->tlb_flush_batched = false;
> +		atomic_cmpxchg(&mm->tlb_flush_batched, batched,
> +			       TLB_FLUSH_BATCH_PACK(pending, pending));
> 	}

Overall, I am not overly excited about the fact the the mm generation
and this batching mechanism remain separated. This makes reasoning
about TLB flushes harder and can lead to unnecessary flushes.

I understand the complexity in changing the code to get there,
and perhaps your approach is reasonable for now, if you at least
manage to deal with overflows.

Thanks for doing all of that, sorry for being negative. I guess I 
suffer from NIH ("not invented here”).

  reply	other threads:[~2021-11-25  8:41 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-11-25  6:50 [PATCH -V2] mm/rmap: fix potential batched TLB flush race Huang Ying
2021-11-25  8:39 ` Nadav Amit [this message]
2021-11-29  6:39 Huang, Ying
2021-11-29 16:51 ` Nadav Amit
2021-11-30  1:38   ` Huang, Ying

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=7127B0FC-A00D-48C3-9C01-0B85FB1BD63F@vmware.com \
    --to=namit@vmware.com \
    --cc=aarcange@redhat.com \
    --cc=akpm@linux-foundation.org \
    --cc=dave.hansen@linux.intel.com \
    --cc=elver@google.com \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-mm@kvack.org \
    --cc=luto@kernel.org \
    --cc=mgorman@techsingularity.net \
    --cc=syzbot+aa5bebed695edaccf0df@syzkaller.appspotmail.com \
    --cc=will@kernel.org \
    --cc=ying.huang@intel.com \
    --cc=yuzhao@google.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.