[v3,1/2] fork: lock VMAs of the parent process when forking

Commit Message

Suren Baghdasaryan July 5, 2023, 5:12 p.m. UTC

When forking a child process, parent write-protects an anonymous page
and COW-shares it with the child being forked using copy_present_pte().
Parent's TLB is flushed right before we drop the parent's mmap_lock in
dup_mmap(). If we get a write-fault before that TLB flush in the parent,
and we end up replacing that anonymous page in the parent process in
do_wp_page() (because, COW-shared with the child), this might lead to
some stale writable TLB entries targeting the wrong (old) page.
Similar issue happened in the past with userfaultfd (see flush_tlb_page()
call inside do_wp_page()).
Lock VMAs of the parent process when forking a child, which prevents
concurrent page faults during fork operation and avoids this issue.
This fix can potentially regress some fork-heavy workloads. Kernel build
time did not show noticeable regression on a 56-core machine while a
stress test mapping 10000 VMAs and forking 5000 times in a tight loop
shows ~5% regression. If such fork time regression is unacceptable,
disabling CONFIG_PER_VMA_LOCK should restore its performance. Further
optimizations are possible if this regression proves to be problematic.

Suggested-by: David Hildenbrand <[e-mail]>
Reported-by: Jiri Slaby <[e-mail]>
Closes: https://lore.kernel.org/all/[e-mail]/
Reported-by: Holger Hoffstätte <[e-mail]>
Closes: https://lore.kernel.org/all/[e-mail]/
Reported-by: Jacob Young <[e-mail]>
Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217624
Fixes: 0bff0aaea03e ("x86/mm: try VMA lock-based page fault handling first")
Cc: [e-mail]
Signed-off-by: Suren Baghdasaryan <[e-mail]>
---
 kernel/fork.c | 6 ++++++
 1 file changed, 6 insertions(+)

Comments

David Hildenbrand July 5, 2023, 5:14 p.m. UTC | #1

On 05.07.23 19:12, Suren Baghdasaryan wrote:
> When forking a child process, parent write-protects an anonymous page
> and COW-shares it with the child being forked using copy_present_pte().
> Parent's TLB is flushed right before we drop the parent's mmap_lock in
> dup_mmap(). If we get a write-fault before that TLB flush in the parent,
> and we end up replacing that anonymous page in the parent process in
> do_wp_page() (because, COW-shared with the child), this might lead to
> some stale writable TLB entries targeting the wrong (old) page.
> Similar issue happened in the past with userfaultfd (see flush_tlb_page()
> call inside do_wp_page()).
> Lock VMAs of the parent process when forking a child, which prevents
> concurrent page faults during fork operation and avoids this issue.
> This fix can potentially regress some fork-heavy workloads. Kernel build
> time did not show noticeable regression on a 56-core machine while a
> stress test mapping 10000 VMAs and forking 5000 times in a tight loop
> shows ~5% regression. If such fork time regression is unacceptable,
> disabling CONFIG_PER_VMA_LOCK should restore its performance. Further
> optimizations are possible if this regression proves to be problematic.
> 
> Suggested-by: David Hildenbrand <[e-mail]>
> Reported-by: Jiri Slaby <[e-mail]>
> Closes: https://lore.kernel.org/all/[e-mail]/
> Reported-by: Holger Hoffstätte <[e-mail]>
> Closes: https://lore.kernel.org/all/[e-mail]/
> Reported-by: Jacob Young <[e-mail]>
> Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217624
> Fixes: 0bff0aaea03e ("x86/mm: try VMA lock-based page fault handling first")
> Cc: [e-mail]
> Signed-off-by: Suren Baghdasaryan <[e-mail]>
> ---
>   kernel/fork.c | 6 ++++++
>   1 file changed, 6 insertions(+)
> 
> diff --git a/kernel/fork.c b/kernel/fork.c
> index b85814e614a5..403bc2b72301 100644
> --- a/kernel/fork.c
> +++ b/kernel/fork.c
> @@ -658,6 +658,12 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
>   		retval = -EINTR;
>   		goto fail_uprobe_end;
>   	}
> +#ifdef CONFIG_PER_VMA_LOCK
> +	/* Disallow any page faults before calling flush_cache_dup_mm */
> +	for_each_vma(old_vmi, mpnt)
> +		vma_start_write(mpnt);
> +	vma_iter_init(&old_vmi, oldmm, 0);
> +#endif
>   	flush_cache_dup_mm(oldmm);
>   	uprobe_dup_mmap(oldmm, mm);
>   	/*

The old version was most probably fine as well, but this certainly looks 
even safer.

Acked-by: David Hildenbrand <[e-mail]>

Suren Baghdasaryan July 5, 2023, 5:23 p.m. UTC | #2

On Wed, Jul 5, 2023 at 10:14 AM David Hildenbrand <[e-mail]> wrote:
>
> On 05.07.23 19:12, Suren Baghdasaryan wrote:
> > When forking a child process, parent write-protects an anonymous page
> > and COW-shares it with the child being forked using copy_present_pte().
> > Parent's TLB is flushed right before we drop the parent's mmap_lock in
> > dup_mmap(). If we get a write-fault before that TLB flush in the parent,
> > and we end up replacing that anonymous page in the parent process in
> > do_wp_page() (because, COW-shared with the child), this might lead to
> > some stale writable TLB entries targeting the wrong (old) page.
> > Similar issue happened in the past with userfaultfd (see flush_tlb_page()
> > call inside do_wp_page()).
> > Lock VMAs of the parent process when forking a child, which prevents
> > concurrent page faults during fork operation and avoids this issue.
> > This fix can potentially regress some fork-heavy workloads. Kernel build
> > time did not show noticeable regression on a 56-core machine while a
> > stress test mapping 10000 VMAs and forking 5000 times in a tight loop
> > shows ~5% regression. If such fork time regression is unacceptable,
> > disabling CONFIG_PER_VMA_LOCK should restore its performance. Further
> > optimizations are possible if this regression proves to be problematic.
> >
> > Suggested-by: David Hildenbrand <[e-mail]>
> > Reported-by: Jiri Slaby <[e-mail]>
> > Closes: https://lore.kernel.org/all/[e-mail]/
> > Reported-by: Holger Hoffstätte <[e-mail]>
> > Closes: https://lore.kernel.org/all/[e-mail]/
> > Reported-by: Jacob Young <[e-mail]>
> > Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217624
> > Fixes: 0bff0aaea03e ("x86/mm: try VMA lock-based page fault handling first")
> > Cc: [e-mail]
> > Signed-off-by: Suren Baghdasaryan <[e-mail]>
> > ---
> >   kernel/fork.c | 6 ++++++
> >   1 file changed, 6 insertions(+)
> >
> > diff --git a/kernel/fork.c b/kernel/fork.c
> > index b85814e614a5..403bc2b72301 100644
> > --- a/kernel/fork.c
> > +++ b/kernel/fork.c
> > @@ -658,6 +658,12 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
> >               retval = -EINTR;
> >               goto fail_uprobe_end;
> >       }
> > +#ifdef CONFIG_PER_VMA_LOCK
> > +     /* Disallow any page faults before calling flush_cache_dup_mm */
> > +     for_each_vma(old_vmi, mpnt)
> > +             vma_start_write(mpnt);
> > +     vma_iter_init(&old_vmi, oldmm, 0);
> > +#endif
> >       flush_cache_dup_mm(oldmm);
> >       uprobe_dup_mmap(oldmm, mm);
> >       /*
>
> The old version was most probably fine as well, but this certainly looks
> even safer.
>
> Acked-by: David Hildenbrand <[e-mail]>

Thanks!

>
> --
> Cheers,
>
> David / dhildenb
>

Liam R. Howlett July 5, 2023, 11:06 p.m. UTC | #3

* Suren Baghdasaryan <[e-mail]> [230705 13:24]:
> On Wed, Jul 5, 2023 at 10:14 AM David Hildenbrand <[e-mail]> wrote:
> >
> > On 05.07.23 19:12, Suren Baghdasaryan wrote:
> > > When forking a child process, parent write-protects an anonymous page
> > > and COW-shares it with the child being forked using copy_present_pte().
> > > Parent's TLB is flushed right before we drop the parent's mmap_lock in
> > > dup_mmap(). If we get a write-fault before that TLB flush in the parent,
> > > and we end up replacing that anonymous page in the parent process in
> > > do_wp_page() (because, COW-shared with the child), this might lead to
> > > some stale writable TLB entries targeting the wrong (old) page.
> > > Similar issue happened in the past with userfaultfd (see flush_tlb_page()
> > > call inside do_wp_page()).
> > > Lock VMAs of the parent process when forking a child, which prevents
> > > concurrent page faults during fork operation and avoids this issue.
> > > This fix can potentially regress some fork-heavy workloads. Kernel build
> > > time did not show noticeable regression on a 56-core machine while a
> > > stress test mapping 10000 VMAs and forking 5000 times in a tight loop
> > > shows ~5% regression. If such fork time regression is unacceptable,
> > > disabling CONFIG_PER_VMA_LOCK should restore its performance. Further
> > > optimizations are possible if this regression proves to be problematic.
> > >
> > > Suggested-by: David Hildenbrand <[e-mail]>
> > > Reported-by: Jiri Slaby <[e-mail]>
> > > Closes: https://lore.kernel.org/all/[e-mail]/
> > > Reported-by: Holger Hoffstätte <[e-mail]>
> > > Closes: https://lore.kernel.org/all/[e-mail]/
> > > Reported-by: Jacob Young <[e-mail]>
> > > Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217624
> > > Fixes: 0bff0aaea03e ("x86/mm: try VMA lock-based page fault handling first")
> > > Cc: [e-mail]
> > > Signed-off-by: Suren Baghdasaryan <[e-mail]>
> > > ---
> > >   kernel/fork.c | 6 ++++++
> > >   1 file changed, 6 insertions(+)
> > >
> > > diff --git a/kernel/fork.c b/kernel/fork.c
> > > index b85814e614a5..403bc2b72301 100644
> > > --- a/kernel/fork.c
> > > +++ b/kernel/fork.c
> > > @@ -658,6 +658,12 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
> > >               retval = -EINTR;
> > >               goto fail_uprobe_end;
> > >       }
> > > +#ifdef CONFIG_PER_VMA_LOCK
> > > +     /* Disallow any page faults before calling flush_cache_dup_mm */
> > > +     for_each_vma(old_vmi, mpnt)
> > > +             vma_start_write(mpnt);
> > > +     vma_iter_init(&old_vmi, oldmm, 0);

vma_iter_set(&old_vmi, 0) is probably what you want here.

> > > +#endif
> > >       flush_cache_dup_mm(oldmm);
> > >       uprobe_dup_mmap(oldmm, mm);
> > >       /*
> >
> > The old version was most probably fine as well, but this certainly looks
> > even safer.
> >
> > Acked-by: David Hildenbrand <[e-mail]>

I think this is overkill and believe setting the vma_start_write() will
synchronize with any readers since it's using the per-vma rw semaphore
in write mode. Anything faulting will need to finish before the fork
continues and faults during the fork will fall back to a read lock of
the mmap_lock.  Is there a possibility of populate happening outside the
mmap_write lock/vma_lock?

Was your benchmarking done with this loop at the start?

Thanks,
Liam

Suren Baghdasaryan July 6, 2023, 12:20 a.m. UTC | #4

On Wed, Jul 5, 2023 at 4:07 PM Liam R. Howlett <[e-mail]> wrote:
>
> * Suren Baghdasaryan <[e-mail]> [230705 13:24]:
> > On Wed, Jul 5, 2023 at 10:14 AM David Hildenbrand <[e-mail]> wrote:
> > >
> > > On 05.07.23 19:12, Suren Baghdasaryan wrote:
> > > > When forking a child process, parent write-protects an anonymous page
> > > > and COW-shares it with the child being forked using copy_present_pte().
> > > > Parent's TLB is flushed right before we drop the parent's mmap_lock in
> > > > dup_mmap(). If we get a write-fault before that TLB flush in the parent,
> > > > and we end up replacing that anonymous page in the parent process in
> > > > do_wp_page() (because, COW-shared with the child), this might lead to
> > > > some stale writable TLB entries targeting the wrong (old) page.
> > > > Similar issue happened in the past with userfaultfd (see flush_tlb_page()
> > > > call inside do_wp_page()).
> > > > Lock VMAs of the parent process when forking a child, which prevents
> > > > concurrent page faults during fork operation and avoids this issue.
> > > > This fix can potentially regress some fork-heavy workloads. Kernel build
> > > > time did not show noticeable regression on a 56-core machine while a
> > > > stress test mapping 10000 VMAs and forking 5000 times in a tight loop
> > > > shows ~5% regression. If such fork time regression is unacceptable,
> > > > disabling CONFIG_PER_VMA_LOCK should restore its performance. Further
> > > > optimizations are possible if this regression proves to be problematic.
> > > >
> > > > Suggested-by: David Hildenbrand <[e-mail]>
> > > > Reported-by: Jiri Slaby <[e-mail]>
> > > > Closes: https://lore.kernel.org/all/[e-mail]/
> > > > Reported-by: Holger Hoffstätte <[e-mail]>
> > > > Closes: https://lore.kernel.org/all/[e-mail]/
> > > > Reported-by: Jacob Young <[e-mail]>
> > > > Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217624
> > > > Fixes: 0bff0aaea03e ("x86/mm: try VMA lock-based page fault handling first")
> > > > Cc: [e-mail]
> > > > Signed-off-by: Suren Baghdasaryan <[e-mail]>
> > > > ---
> > > >   kernel/fork.c | 6 ++++++
> > > >   1 file changed, 6 insertions(+)
> > > >
> > > > diff --git a/kernel/fork.c b/kernel/fork.c
> > > > index b85814e614a5..403bc2b72301 100644
> > > > --- a/kernel/fork.c
> > > > +++ b/kernel/fork.c
> > > > @@ -658,6 +658,12 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
> > > >               retval = -EINTR;
> > > >               goto fail_uprobe_end;
> > > >       }
> > > > +#ifdef CONFIG_PER_VMA_LOCK
> > > > +     /* Disallow any page faults before calling flush_cache_dup_mm */
> > > > +     for_each_vma(old_vmi, mpnt)
> > > > +             vma_start_write(mpnt);
> > > > +     vma_iter_init(&old_vmi, oldmm, 0);
>
> vma_iter_set(&old_vmi, 0) is probably what you want here.

Ok, I send another version with that.

>
> > > > +#endif
> > > >       flush_cache_dup_mm(oldmm);
> > > >       uprobe_dup_mmap(oldmm, mm);
> > > >       /*
> > >
> > > The old version was most probably fine as well, but this certainly looks
> > > even safer.
> > >
> > > Acked-by: David Hildenbrand <[e-mail]>
>
> I think this is overkill and believe setting the vma_start_write() will
> synchronize with any readers since it's using the per-vma rw semaphore
> in write mode. Anything faulting will need to finish before the fork
> continues and faults during the fork will fall back to a read lock of
> the mmap_lock.  Is there a possibility of populate happening outside the
> mmap_write lock/vma_lock?

Yes, I think we understand the loss of concurrency in the parent's
ability to fault pages while forking. Is that a real problem though?

>
> Was your benchmarking done with this loop at the start?

No, it was done with the initial version where the lock was inside the
existing loop. I just reran the benchmark and while kernel compilation
times did not change, the stress test shows ~7% regression now,
probably due to that additional tree walk. I'll update that number in
the new patch.
Thanks!

>
> Thanks,
> Liam

Liam R. Howlett July 6, 2023, 12:32 a.m. UTC | #5

* Suren Baghdasaryan <[e-mail]> [230705 20:20]:
> On Wed, Jul 5, 2023 at 4:07 PM Liam R. Howlett <[e-mail]> wrote:
> >
> > * Suren Baghdasaryan <[e-mail]> [230705 13:24]:
> > > On Wed, Jul 5, 2023 at 10:14 AM David Hildenbrand <[e-mail]> wrote:
> > > >
> > > > On 05.07.23 19:12, Suren Baghdasaryan wrote:
> > > > > When forking a child process, parent write-protects an anonymous page
> > > > > and COW-shares it with the child being forked using copy_present_pte().
> > > > > Parent's TLB is flushed right before we drop the parent's mmap_lock in
> > > > > dup_mmap(). If we get a write-fault before that TLB flush in the parent,
> > > > > and we end up replacing that anonymous page in the parent process in
> > > > > do_wp_page() (because, COW-shared with the child), this might lead to
> > > > > some stale writable TLB entries targeting the wrong (old) page.
> > > > > Similar issue happened in the past with userfaultfd (see flush_tlb_page()
> > > > > call inside do_wp_page()).
> > > > > Lock VMAs of the parent process when forking a child, which prevents
> > > > > concurrent page faults during fork operation and avoids this issue.
> > > > > This fix can potentially regress some fork-heavy workloads. Kernel build
> > > > > time did not show noticeable regression on a 56-core machine while a
> > > > > stress test mapping 10000 VMAs and forking 5000 times in a tight loop
> > > > > shows ~5% regression. If such fork time regression is unacceptable,
> > > > > disabling CONFIG_PER_VMA_LOCK should restore its performance. Further
> > > > > optimizations are possible if this regression proves to be problematic.
> > > > >
> > > > > Suggested-by: David Hildenbrand <[e-mail]>
> > > > > Reported-by: Jiri Slaby <[e-mail]>
> > > > > Closes: https://lore.kernel.org/all/[e-mail]/
> > > > > Reported-by: Holger Hoffstätte <[e-mail]>
> > > > > Closes: https://lore.kernel.org/all/[e-mail]/
> > > > > Reported-by: Jacob Young <[e-mail]>
> > > > > Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217624
> > > > > Fixes: 0bff0aaea03e ("x86/mm: try VMA lock-based page fault handling first")
> > > > > Cc: [e-mail]
> > > > > Signed-off-by: Suren Baghdasaryan <[e-mail]>
> > > > > ---
> > > > >   kernel/fork.c | 6 ++++++
> > > > >   1 file changed, 6 insertions(+)
> > > > >
> > > > > diff --git a/kernel/fork.c b/kernel/fork.c
> > > > > index b85814e614a5..403bc2b72301 100644
> > > > > --- a/kernel/fork.c
> > > > > +++ b/kernel/fork.c
> > > > > @@ -658,6 +658,12 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
> > > > >               retval = -EINTR;
> > > > >               goto fail_uprobe_end;
> > > > >       }
> > > > > +#ifdef CONFIG_PER_VMA_LOCK
> > > > > +     /* Disallow any page faults before calling flush_cache_dup_mm */
> > > > > +     for_each_vma(old_vmi, mpnt)
> > > > > +             vma_start_write(mpnt);
> > > > > +     vma_iter_init(&old_vmi, oldmm, 0);
> >
> > vma_iter_set(&old_vmi, 0) is probably what you want here.
> 
> Ok, I send another version with that.
> 
> >
> > > > > +#endif
> > > > >       flush_cache_dup_mm(oldmm);
> > > > >       uprobe_dup_mmap(oldmm, mm);
> > > > >       /*
> > > >
> > > > The old version was most probably fine as well, but this certainly looks
> > > > even safer.
> > > >
> > > > Acked-by: David Hildenbrand <[e-mail]>
> >
> > I think this is overkill and believe setting the vma_start_write() will
> > synchronize with any readers since it's using the per-vma rw semaphore
> > in write mode. Anything faulting will need to finish before the fork
> > continues and faults during the fork will fall back to a read lock of
> > the mmap_lock.  Is there a possibility of populate happening outside the
> > mmap_write lock/vma_lock?
> 
> Yes, I think we understand the loss of concurrency in the parent's
> ability to fault pages while forking. Is that a real problem though?

No, I don't think that part is an issue at all.  I wanted to be sure I
didn't miss something.

> 
> >
> > Was your benchmarking done with this loop at the start?
> 
> No, it was done with the initial version where the lock was inside the
> existing loop. I just reran the benchmark and while kernel compilation
> times did not change, the stress test shows ~7% regression now,
> probably due to that additional tree walk. I'll update that number in
> the new patch.

..but I expected a performance hit and didn't understand why you updated
the patch this way.  It would probably only happen on really big trees
though and, ah, the largest trees I see are from the android side.  I'd
wager the impact will be felt more when larger trees encounter smaller
CPU cache.

Thanks,
Liam

Suren Baghdasaryan July 6, 2023, 12:42 a.m. UTC | #6

On Wed, Jul 5, 2023 at 5:33 PM Liam R. Howlett <[e-mail]> wrote:
>
> * Suren Baghdasaryan <[e-mail]> [230705 20:20]:
> > On Wed, Jul 5, 2023 at 4:07 PM Liam R. Howlett <[e-mail]> wrote:
> > >
> > > * Suren Baghdasaryan <[e-mail]> [230705 13:24]:
> > > > On Wed, Jul 5, 2023 at 10:14 AM David Hildenbrand <[e-mail]> wrote:
> > > > >
> > > > > On 05.07.23 19:12, Suren Baghdasaryan wrote:
> > > > > > When forking a child process, parent write-protects an anonymous page
> > > > > > and COW-shares it with the child being forked using copy_present_pte().
> > > > > > Parent's TLB is flushed right before we drop the parent's mmap_lock in
> > > > > > dup_mmap(). If we get a write-fault before that TLB flush in the parent,
> > > > > > and we end up replacing that anonymous page in the parent process in
> > > > > > do_wp_page() (because, COW-shared with the child), this might lead to
> > > > > > some stale writable TLB entries targeting the wrong (old) page.
> > > > > > Similar issue happened in the past with userfaultfd (see flush_tlb_page()
> > > > > > call inside do_wp_page()).
> > > > > > Lock VMAs of the parent process when forking a child, which prevents
> > > > > > concurrent page faults during fork operation and avoids this issue.
> > > > > > This fix can potentially regress some fork-heavy workloads. Kernel build
> > > > > > time did not show noticeable regression on a 56-core machine while a
> > > > > > stress test mapping 10000 VMAs and forking 5000 times in a tight loop
> > > > > > shows ~5% regression. If such fork time regression is unacceptable,
> > > > > > disabling CONFIG_PER_VMA_LOCK should restore its performance. Further
> > > > > > optimizations are possible if this regression proves to be problematic.
> > > > > >
> > > > > > Suggested-by: David Hildenbrand <[e-mail]>
> > > > > > Reported-by: Jiri Slaby <[e-mail]>
> > > > > > Closes: https://lore.kernel.org/all/[e-mail]/
> > > > > > Reported-by: Holger Hoffstätte <[e-mail]>
> > > > > > Closes: https://lore.kernel.org/all/[e-mail]/
> > > > > > Reported-by: Jacob Young <[e-mail]>
> > > > > > Closes: https://bugzilla.kernel.org/show_bug.cgi?id=217624
> > > > > > Fixes: 0bff0aaea03e ("x86/mm: try VMA lock-based page fault handling first")
> > > > > > Cc: [e-mail]
> > > > > > Signed-off-by: Suren Baghdasaryan <[e-mail]>
> > > > > > ---
> > > > > >   kernel/fork.c | 6 ++++++
> > > > > >   1 file changed, 6 insertions(+)
> > > > > >
> > > > > > diff --git a/kernel/fork.c b/kernel/fork.c
> > > > > > index b85814e614a5..403bc2b72301 100644
> > > > > > --- a/kernel/fork.c
> > > > > > +++ b/kernel/fork.c
> > > > > > @@ -658,6 +658,12 @@ static __latent_entropy int dup_mmap(struct mm_struct *mm,
> > > > > >               retval = -EINTR;
> > > > > >               goto fail_uprobe_end;
> > > > > >       }
> > > > > > +#ifdef CONFIG_PER_VMA_LOCK
> > > > > > +     /* Disallow any page faults before calling flush_cache_dup_mm */
> > > > > > +     for_each_vma(old_vmi, mpnt)
> > > > > > +             vma_start_write(mpnt);
> > > > > > +     vma_iter_init(&old_vmi, oldmm, 0);
> > >
> > > vma_iter_set(&old_vmi, 0) is probably what you want here.
> >
> > Ok, I send another version with that.
> >
> > >
> > > > > > +#endif
> > > > > >       flush_cache_dup_mm(oldmm);
> > > > > >       uprobe_dup_mmap(oldmm, mm);
> > > > > >       /*
> > > > >
> > > > > The old version was most probably fine as well, but this certainly looks
> > > > > even safer.
> > > > >
> > > > > Acked-by: David Hildenbrand <[e-mail]>
> > >
> > > I think this is overkill and believe setting the vma_start_write() will
> > > synchronize with any readers since it's using the per-vma rw semaphore
> > > in write mode. Anything faulting will need to finish before the fork
> > > continues and faults during the fork will fall back to a read lock of
> > > the mmap_lock.  Is there a possibility of populate happening outside the
> > > mmap_write lock/vma_lock?
> >
> > Yes, I think we understand the loss of concurrency in the parent's
> > ability to fault pages while forking. Is that a real problem though?
>
> No, I don't think that part is an issue at all.  I wanted to be sure I
> didn't miss something.
>
> >
> > >
> > > Was your benchmarking done with this loop at the start?
> >
> > No, it was done with the initial version where the lock was inside the
> > existing loop. I just reran the benchmark and while kernel compilation
> > times did not change, the stress test shows ~7% regression now,
> > probably due to that additional tree walk. I'll update that number in
> > the new patch.
>
> ..but I expected a performance hit and didn't understand why you updated
> the patch this way.  It would probably only happen on really big trees
> though and, ah, the largest trees I see are from the android side.  I'd
> wager the impact will be felt more when larger trees encounter smaller
> CPU cache.

My test has 10000 vmas and even for Android that's a stretch (the
highest number I've seen was ~4000).
We can think of a less restrictive solution if this proves to be a
problem for some workloads but for now I would prefer to fix this in a
safe way and possibly improve that later. The alternative is to revert
this completely and we get no more testing until the next release.

>
> Thanks,
> Liam

Patch

diff --git a/kernel/fork.c b/kernel/fork.c
index b85814e614a5..403bc2b72301 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -658,6 +658,12 @@  static __latent_entropy int dup_mmap(struct mm_struct *mm,
 		retval = -EINTR;
 		goto fail_uprobe_end;
 	}
+#ifdef CONFIG_PER_VMA_LOCK
+	/* Disallow any page faults before calling flush_cache_dup_mm */
+	for_each_vma(old_vmi, mpnt)
+		vma_start_write(mpnt);
+	vma_iter_init(&old_vmi, oldmm, 0);
+#endif
 	flush_cache_dup_mm(oldmm);
 	uprobe_dup_mmap(oldmm, mm);
 	/*