lkml.org 
[lkml]   [2016]   [Dec]   [16]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
Subjectcrash during oom reaper (was: Re: [PATCH 4/4] [RFC!] mm: 'struct mm_struct' reference counting debugging)
On Fri 16-12-16 10:43:52, Vegard Nossum wrote:
[...]
> I don't think it's a bug in the OOM reaper itself, but either of the
> following two patches will fix the problem (without my understand how or
> why):
>
> diff --git a/mm/oom_kill.c b/mm/oom_kill.c
> index ec9f11d4f094..37b14b2e2af4 100644
> --- a/mm/oom_kill.c
> +++ b/mm/oom_kill.c
> @@ -485,7 +485,7 @@ static bool __oom_reap_task_mm(struct task_struct *tsk,
> struct mm_struct *mm)
> */
> mutex_lock(&oom_lock);
>
> - if (!down_read_trylock(&mm->mmap_sem)) {
> + if (!down_write_trylock(&mm->mmap_sem)) {

__oom_reap_task_mm is basically the same thing as MADV_DONTNEED and that
doesn't require the exlusive mmap_sem. So this looks correct to me.
[...]

> --OR--
>
> diff --git a/mm/oom_kill.c b/mm/oom_kill.c
> index ec9f11d4f094..559aec0acd21 100644
> --- a/mm/oom_kill.c
> +++ b/mm/oom_kill.c
> @@ -508,6 +508,7 @@ static bool __oom_reap_task_mm(struct task_struct *tsk,
> struct mm_struct *mm)
> */
> set_bit(MMF_UNSTABLE, &mm->flags);
>
> +#if 0
> tlb_gather_mmu(&tlb, mm, 0, -1);
> for (vma = mm->mmap ; vma; vma = vma->vm_next) {
> if (is_vm_hugetlb_page(vma))
> @@ -535,6 +536,7 @@ static bool __oom_reap_task_mm(struct task_struct *tsk,
> struct mm_struct *mm)
> &details);
> }
> tlb_finish_mmu(&tlb, 0, -1);
> +#endif

same here, nothing different from the madvise... Well, except for the
MMF_UNSTABLE part which will force any page fault on this mm to SEGV.

> pr_info("oom_reaper: reaped process %d (%s), now anon-rss:%lukB,
> file-rss:%lukB, shmem-rss:%lukB\n",
> task_pid_nr(tsk), tsk->comm,
> K(get_mm_counter(mm, MM_ANONPAGES)),
>
> Maybe it's just the fact that we're not releasing the memory and so some
> other bit of code is not able to make enough progress to trigger the
> bug, although curiously, if I just move the #if 0..#endif inside
> tlb_gather_mmu()..tlb_finish_mmu() itself (so just calling tlb_*()
> without doing the for-loop), it still reproduces the crash.

What is the atual crash?

> Another clue, although it might just be a coincidence, is that it seems
> the VMA/file in question is always a mapping for the exe file itself
> (the reason I think this might be a coincidence is that the exe file
> mapping is the first one and we usually traverse VMAs starting with this
> one, that doesn't mean the other VMAs aren't affected by the same
> problem, just that we never hit them).

You can experiment a bit and exclude PROT_EXEC vmas...
--
Michal Hocko
SUSE Labs

\
 
 \ /
  Last update: 2016-12-16 11:12    [W:0.089 / U:0.928 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site