On Thu, Jan 12, 2023 at 2:06 AM Yang Shi <shy828301@xxxxxxxxx> wrote:
>
> On Wed, Jan 11, 2023 at 5:33 AM Jann Horn <jannh@xxxxxxxxxx> wrote:
> >
> > If an ->anon_vma is attached to the VMA, collapse_and_free_pmd() requires
> > it to be locked. retract_page_tables() bails out if an ->anon_vma is
> > attached, but does this check before holding the mmap lock (as the comment
> > above the check explains).
> >
> > If we racily merge an existing ->anon_vma (shared with a child process)
> > from a neighboring VMA, subsequent rmap traversals on pages belonging to
> > the child will be able to see the page tables that we are concurrently
> > removing while assuming that nothing else can access them.
> >
> > Repeat the ->anon_vma check once we hold the mmap lock to ensure that there
> > really is no concurrent page table access.
> >
> > Reported-by: Zach O'Keefe <zokeefe@xxxxxxxxxx>
> > Fixes: f3f0e1d2150b ("khugepaged: add support of collapse for tmpfs/shmem pages")
> > Cc: stable@xxxxxxxxxxxxxxx
> > Signed-off-by: Jann Horn <jannh@xxxxxxxxxx>
> > ---
> > zokeefe@ pointed out to me that the current code (after my last round of patches)
> > can hit a lockdep assert by racing, and after staring at it a bit I've
>
> I'm supposed the lockdep is the one in collapse_and_free_pmd(). It is
> better to have the splat included in the commit log.
I pasted the splat in
https://lore.kernel.org/linux-mm/CAG48ez3434wZBKFFbdx4M9j6eUwSUVPd4dxhzW_k_POneSDF+A@xxxxxxxxxxxxxx/
- which part do you think should go into the commit log? Just the
"WARNING: CPU: 14 PID: 116 at
mm/khugepaged.c:1406 collapse_and_free_pmd+0x364/0x420"? Or the whole
ASAN splat below the lockdep complaint with all three backtraces?
> > convinced myself that this is a real, preexisting bug.
> > (I haven't written a reproducer for it though. One way to hit it might be
> > something along the lines of:
> >
> > - set up a process A with a private-file-mapping VMA V1
> > - let A fork() to create process B, thereby copying V1 in A to V1' in B
> > - let B extend the end of V1'
> > - let B put some anon pages into the extended part of V1'
>
> I don't quite get why we need this step. A cow fault on A's V1 isn't
> enough to have anon_vma for V1? This should not prevent V1 and V2 from
> sharing anon_vma. Did I miss something?
You're right, these steps don't work.
