Re: [PATCH v6 10/16] mm: replace vm_lock and detached flag with a reference count

[Suren Baghdasaryan <surenb@google.com>]

On Sun, Dec 22, 2024 at 7:03 PM Suren Baghdasaryan <surenb@google.com> wrote:
>
> On Thu, Dec 19, 2024 at 10:55 AM Liam R. Howlett
> <Liam.Howlett@oracle.com> wrote:
> >
> > * Peter Zijlstra <peterz@infradead.org> [241219 13:47]:
> > > On Thu, Dec 19, 2024 at 01:18:23PM -0500, Liam R. Howlett wrote:
> > >
> > > > > For RCU lookups only the mas tree matters -- and its left present there.
> > > > >
> > > > > If you really want to block RCU readers, I would suggest punching a hole
> > > > > in the mm_mt. All the traditional code won't notice anyway, this is all
> > > > > with mmap_lock held for writing.
> > > >
> > > > We don't want to block all rcu readers, we want to block the rcu readers
> > > > that would see the problem - that is, anyone trying to read a particular
> > > > area.
> > > >
> > > > Right now we can page fault in unpopulated vmas while writing other vmas
> > > > to the tree.  We are also moving more users to rcu reading to use the
> > > > vmas they need without waiting on writes to finish.
> > > >
> > > > Maybe I don't understand your suggestion, but I would think punching a
> > > > hole would lose this advantage?
> > >
> > > My suggestion was to remove the range stuck in mas_detach from mm_mt.
> > > That is exactly the affected range, no?
> >
> > Yes.
> >
> > But then looping over the vmas will show a gap where there should not be
> > a gap.
> >
> > If we stop rcu readers entirely we lose the advantage.
> >
> > This is exactly the issue that the locking dance was working around :)
>
> IOW we write-lock the entire range before removing any part of it for
> the whole transaction to be atomic, correct?
>
>
> Peter, you suggested the following pattern for ensuring vma is
> detached with no possible readers:
>
> vma_iter_store()
> vma_start_write()
> vma_mark_detached()
>
> What do you think about this alternative?
>
> vma_start_write()
> ...
> vma_iter_store()
> vma_mark_detached()
>         vma_assert_write_locked(vma)
>         if (unlikely(!refcount_dec_and_test(&vma->vm_refcnt)))
>                 vma_start_write()
>
> The second vma_start_write() is unlikely to be executed because the
> vma is locked, vm_refcnt might be increased only temporarily by
> readers before they realize the vma is locked and that's a very narrow
> window. I think performance should not visibly suffer?
> OTOH this would let us keep current locking patterns and would
> guarantee that vma_mark_detached() always exits with a detached and
> unused vma (less possibilities for someone not following an exact
> pattern and ending up with a detached but still used vma).

I posted v7 of this patchset at
https://lore.kernel.org/all/20241226170710.1159679-1-surenb@google.com/
From the things we discussed, I didn't include the following:
- Changing vma locking patterns
- Changing do_vmi_align_munmap() to avoid reattach_vmas()
It seems we need more discussion for the first one and the second one
can be done completely independent from this patchset. I feel this
patchset is already quite large, so trying to keep its size
manageable.
Thanks,
Suren.

>
> >