Linux-mm Archive on lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH mm-unstable 1/1] fs/proc/task_mmu: advance seq_file position on interrupted VMA lookup
@ 2026-11-12 14:01 Adrian Huang (Lenovo)
  2026-11-13  2:37 ` Andrew Morton
  2026-11-16 14:37 ` Lorenzo Stoakes (ARM)
  0 siblings, 2 replies; 6+ messages in thread
From: Adrian Huang (Lenovo) @ 2026-11-12 14:01 UTC (permalink / raw)
  To: Suren Baghdasaryan, Lorenzo Stoakes
  Cc: Andrew Morton, Liam R . Howlett, Vlastimil Babka, Jann Horn,
	Pedro Falcato, linux-mm, linux-kernel, linux-fsdevel, ahuang12,
	Adrian Huang (Lenovo)

Commit 5631da56c9a8 ("fs/proc/task_mmu: read proc/pid/maps under
per-vma lock") introduced the possibility of returning -EINTR
from proc_get_vma() when the VMA lookup is interrupted.

When this happens in m_next(), proc_get_vma() returns an error
pointer without updating the position index. Consequently,
seq_read_iter() reports:

  seq_file: buggy .next function m_next did not update position index

The above buggy message is reproduced by the following command:

  # stress-ng --vma 0 -t 60

Update the position index when VMA lookup fails, as required by
the seq_file interface. Also save the original position in
priv->last_pos so that subsequent iteration can restart from
the correct address without skipping or repeating a VMA.

Fixes: 5631da56c9a8 ("fs/proc/task_mmu: read proc/pid/maps under per-vma lock")
Signed-off-by: Adrian Huang (Lenovo) <adrianhuang0701@gmail.com>
---
 fs/proc/task_mmu.c | 10 ++++++++++
 1 file changed, 10 insertions(+)

diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index 22361607020d..164b4efc5e09 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -221,6 +221,16 @@ static struct vm_area_struct *proc_get_vma(struct seq_file *m, loff_t *ppos)
 		if (PTR_ERR(vma) == -EAGAIN && fallback_to_mmap_lock(priv, *ppos))
 			goto retry;
 
+		/*
+		 * seq_file checks that the position advanced before it looks at
+		 * the returned pointer, so an error return that leaves *ppos
+		 * alone is reported as a buggy ->next. Advance it, and keep
+		 * last_pos at the pending position so m_start() would resume
+		 * the lookup from there.
+		 */
+		priv->last_pos = *ppos;
+		(*ppos)++;
+
 		return vma;
 	}
 
-- 
2.43.0



^ permalink raw reply related	[flat|nested] 6+ messages in thread

* Re: [PATCH mm-unstable 1/1] fs/proc/task_mmu: advance seq_file position on interrupted VMA lookup
  2026-11-12 14:01 [PATCH mm-unstable 1/1] fs/proc/task_mmu: advance seq_file position on interrupted VMA lookup Adrian Huang (Lenovo)
@ 2026-11-13  2:37 ` Andrew Morton
  2026-11-13  3:13   ` Suren Baghdasaryan
  2026-11-16 14:37 ` Lorenzo Stoakes (ARM)
  1 sibling, 1 reply; 6+ messages in thread
From: Andrew Morton @ 2026-11-13  2:37 UTC (permalink / raw)
  To: Adrian Huang (Lenovo)
  Cc: Suren Baghdasaryan, Lorenzo Stoakes, Liam R . Howlett,
	Vlastimil Babka, Jann Horn, Pedro Falcato, linux-mm, linux-kernel,
	linux-fsdevel, ahuang12

On Thu, 12 Nov 2026 22:01:01 +0800 "Adrian Huang (Lenovo)" <adrianhuang0701@gmail.com> wrote:

> Commit 5631da56c9a8 ("fs/proc/task_mmu: read proc/pid/maps under
> per-vma lock") introduced the possibility of returning -EINTR
> from proc_get_vma() when the VMA lookup is interrupted.

OK, but what's causing the interruption?

> When this happens in m_next(), proc_get_vma() returns an error
> pointer without updating the position index. Consequently,
> seq_read_iter() reports:
> 
>   seq_file: buggy .next function m_next did not update position index
> 
> The above buggy message is reproduced by the following command:
> 
>   # stress-ng --vma 0 -t 60

afaict "stress-ng -t" can send SIGALRM, but mmap_read_lock_killable()
only responds to fatal signals.

So please clarify: exactly what is causing this to happen?




^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH mm-unstable 1/1] fs/proc/task_mmu: advance seq_file position on interrupted VMA lookup
  2026-11-13  2:37 ` Andrew Morton
@ 2026-11-13  3:13   ` Suren Baghdasaryan
  2026-11-13  3:17     ` Suren Baghdasaryan
  0 siblings, 1 reply; 6+ messages in thread
From: Suren Baghdasaryan @ 2026-11-13  3:13 UTC (permalink / raw)
  To: Andrew Morton
  Cc: Adrian Huang (Lenovo), Lorenzo Stoakes, Liam R . Howlett,
	Vlastimil Babka, Jann Horn, Pedro Falcato, linux-mm, linux-kernel,
	linux-fsdevel, ahuang12

On Thu, Nov 12, 2026 at 6:37 PM Andrew Morton <akpm@linux-foundation.org> wrote:
>
> On Thu, 12 Nov 2026 22:01:01 +0800 "Adrian Huang (Lenovo)" <adrianhuang0701@gmail.com> wrote:
>
> > Commit 5631da56c9a8 ("fs/proc/task_mmu: read proc/pid/maps under
> > per-vma lock") introduced the possibility of returning -EINTR
> > from proc_get_vma() when the VMA lookup is interrupted.
>
> OK, but what's causing the interruption?
>
> > When this happens in m_next(), proc_get_vma() returns an error
> > pointer without updating the position index. Consequently,
> > seq_read_iter() reports:
> >
> >   seq_file: buggy .next function m_next did not update position index
> >
> > The above buggy message is reproduced by the following command:
> >
> >   # stress-ng --vma 0 -t 60
>
> afaict "stress-ng -t" can send SIGALRM, but mmap_read_lock_killable()
> only responds to fatal signals.
>
> So please clarify: exactly what is causing this to happen?

Also I don't think stress-ng --vma 0 -t 60 command reads maps, smaps
or smaps_rollup files by itself but I might be mistaken. It would be
nice to get more details on how this gets reproduced.

>
>


^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH mm-unstable 1/1] fs/proc/task_mmu: advance seq_file position on interrupted VMA lookup
  2026-11-13  3:13   ` Suren Baghdasaryan
@ 2026-11-13  3:17     ` Suren Baghdasaryan
  2026-11-16  6:26       ` Huang Adrian
  0 siblings, 1 reply; 6+ messages in thread
From: Suren Baghdasaryan @ 2026-11-13  3:17 UTC (permalink / raw)
  To: Andrew Morton
  Cc: Adrian Huang (Lenovo), Lorenzo Stoakes, Liam R . Howlett,
	Vlastimil Babka, Jann Horn, Pedro Falcato, linux-mm, linux-kernel,
	linux-fsdevel, ahuang12

On Thu, Nov 12, 2026 at 7:13 PM Suren Baghdasaryan <surenb@google.com> wrote:
>
> On Thu, Nov 12, 2026 at 6:37 PM Andrew Morton <akpm@linux-foundation.org> wrote:
> >
> > On Thu, 12 Nov 2026 22:01:01 +0800 "Adrian Huang (Lenovo)" <adrianhuang0701@gmail.com> wrote:
> >
> > > Commit 5631da56c9a8 ("fs/proc/task_mmu: read proc/pid/maps under
> > > per-vma lock") introduced the possibility of returning -EINTR
> > > from proc_get_vma() when the VMA lookup is interrupted.
> >
> > OK, but what's causing the interruption?
> >
> > > When this happens in m_next(), proc_get_vma() returns an error
> > > pointer without updating the position index. Consequently,
> > > seq_read_iter() reports:
> > >
> > >   seq_file: buggy .next function m_next did not update position index
> > >
> > > The above buggy message is reproduced by the following command:
> > >
> > >   # stress-ng --vma 0 -t 60
> >
> > afaict "stress-ng -t" can send SIGALRM, but mmap_read_lock_killable()
> > only responds to fatal signals.
> >
> > So please clarify: exactly what is causing this to happen?
>
> Also I don't think stress-ng --vma 0 -t 60 command reads maps, smaps
> or smaps_rollup files by itself but I might be mistaken. It would be
> nice to get more details on how this gets reproduced.

Just after sending this I found
https://github.com/ColinIanKing/stress-ng/blob/master/stress-vma.c#L536,
so I stand corrected. Still, it would be nice to know how exactly this
issue is reproduced.

>
> >
> >


^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH mm-unstable 1/1] fs/proc/task_mmu: advance seq_file position on interrupted VMA lookup
  2026-11-13  3:17     ` Suren Baghdasaryan
@ 2026-11-16  6:26       ` Huang Adrian
  0 siblings, 0 replies; 6+ messages in thread
From: Huang Adrian @ 2026-11-16  6:26 UTC (permalink / raw)
  To: Suren Baghdasaryan
  Cc: Andrew Morton, Lorenzo Stoakes, Liam R . Howlett, Vlastimil Babka,
	Jann Horn, Pedro Falcato, linux-mm, linux-kernel, linux-fsdevel,
	ahuang12

Hi Andrew and Suren,

On Fri, Nov 13, 2026 at 11:17 AM Suren Baghdasaryan <surenb@google.com> wrote:
>
> On Thu, Nov 12, 2026 at 7:13 PM Suren Baghdasaryan <surenb@google.com> wrote:
> >
> > On Thu, Nov 12, 2026 at 6:37 PM Andrew Morton <akpm@linux-foundation.org> wrote:
> > >
> > > On Thu, 12 Nov 2026 22:01:01 +0800 "Adrian Huang (Lenovo)" <adrianhuang0701@gmail.com> wrote:
> > >
> > > > Commit 5631da56c9a8 ("fs/proc/task_mmu: read proc/pid/maps under
> > > > per-vma lock") introduced the possibility of returning -EINTR
> > > > from proc_get_vma() when the VMA lookup is interrupted.
> > >
> > > OK, but what's causing the interruption?
> > >
> > > > When this happens in m_next(), proc_get_vma() returns an error
> > > > pointer without updating the position index. Consequently,
> > > > seq_read_iter() reports:
> > > >
> > > >   seq_file: buggy .next function m_next did not update position index
> > > >
> > > > The above buggy message is reproduced by the following command:
> > > >
> > > >   # stress-ng --vma 0 -t 60
> > >
> > > afaict "stress-ng -t" can send SIGALRM, but mmap_read_lock_killable()
> > > only responds to fatal signals.
> > >
> > > So please clarify: exactly what is causing this to happen?
> >
> > Also I don't think stress-ng --vma 0 -t 60 command reads maps, smaps
> > or smaps_rollup files by itself but I might be mistaken. It would be
> > nice to get more details on how this gets reproduced.
>
> Just after sending this I found
> https://github.com/ColinIanKing/stress-ng/blob/master/stress-vma.c#L536,
> so I stand corrected. Still, it would be nice to know how exactly this
> issue is reproduced.

The --vma stressor performs the following jobs:
  * It creates a thread that repeatedly reads /proc/self/maps while other
    threads modify the process's VMAs[1][2].
  * It sends SIGKILL (fatal signal) to its own threads approximately
    every 10 seconds[3]. This happens independently of the 60-second
    timeout for the command 'stress-ng --vma 0 -t 60'.

As a result, down_read_killable() in Linux kernel returns -EINTR.

This finding is based on code inspection and analysis. To validate the
behavior, I collected ftrace logs and added debug patches to trace the
relevant task PID/TID information. The full explanation is fairly
detailed, so I have not included it here, but I can provide it if
there's interest.

[1] https://github.com/ColinIanKing/stress-ng/blob/master/stress-vma.c#L748
[2] https://github.com/ColinIanKing/stress-ng/blob/master/stress-vma.c#L658
[3] https://github.com/ColinIanKing/stress-ng/blob/master/stress-vma.c#L768

-- Adrian


^ permalink raw reply	[flat|nested] 6+ messages in thread

* Re: [PATCH mm-unstable 1/1] fs/proc/task_mmu: advance seq_file position on interrupted VMA lookup
  2026-11-12 14:01 [PATCH mm-unstable 1/1] fs/proc/task_mmu: advance seq_file position on interrupted VMA lookup Adrian Huang (Lenovo)
  2026-11-13  2:37 ` Andrew Morton
@ 2026-11-16 14:37 ` Lorenzo Stoakes (ARM)
  1 sibling, 0 replies; 6+ messages in thread
From: Lorenzo Stoakes (ARM) @ 2026-11-16 14:37 UTC (permalink / raw)
  To: Adrian Huang (Lenovo)
  Cc: Suren Baghdasaryan, Andrew Morton, Liam R . Howlett,
	Vlastimil Babka, Jann Horn, Pedro Falcato, linux-mm, linux-kernel,
	linux-fsdevel, ahuang12

On Thu, Nov 12, 2026 at 10:01:01PM +0800, Adrian Huang (Lenovo) wrote:
> Commit 5631da56c9a8 ("fs/proc/task_mmu: read proc/pid/maps under
> per-vma lock") introduced the possibility of returning -EINTR
> from proc_get_vma() when the VMA lookup is interrupted.
>
> When this happens in m_next(), proc_get_vma() returns an error
> pointer without updating the position index. Consequently,
> seq_read_iter() reports:
>
>   seq_file: buggy .next function m_next did not update position index

What does seq_read_iter() 'reporting' something mean? It feels like you're
trying to make this into something bigger than it is.

What you mean to say is a rate-limited pr_info() output occurs, and when a
process that _has received a fatal signal_ (!!) somehow continues long
enough to try to output more data, you get an pr_info().

The relevant code from seq_read_iter():

		if (pos == m->index) {
			pr_info_ratelimited("buggy .next function %ps did not update position index\n",
					    m->op->next);
			m->index++;
		}

Anyway you should mention the fatal signal requirement clearly.

>
> The above buggy message is reproduced by the following command:
>
>   # stress-ng --vma 0 -t 60

OK I could reproduce the output locally. It seems that this run results in
fatal signals being sent to the processes triggering this output.

> Update the position index when VMA lookup fails, as required by
> the seq_file interface. Also save the original position in
> priv->last_pos so that subsequent iteration can restart from
> the correct address without skipping or repeating a VMA.

We don't really care about skipping or repeating a VMA for a process that's
about to die :)

We may as well do this but better to just hoist that to share code.

>
> Fixes: 5631da56c9a8 ("fs/proc/task_mmu: read proc/pid/maps under per-vma lock")

This isn't really a bug, it's really cosmetic to be honest. So not sure if
we need a cc: stable hmm.

> Signed-off-by: Adrian Huang (Lenovo) <adrianhuang0701@gmail.com>

See attached suggested patch that makes the comment sane + hoists the shared
bit (dropping a silly existing comment).

With that and an updated commit message should be ok.

> ---
>  fs/proc/task_mmu.c | 10 ++++++++++
>  1 file changed, 10 insertions(+)
>
> diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
> index 22361607020d..164b4efc5e09 100644
> --- a/fs/proc/task_mmu.c
> +++ b/fs/proc/task_mmu.c
> @@ -221,6 +221,16 @@ static struct vm_area_struct *proc_get_vma(struct seq_file *m, loff_t *ppos)
>  		if (PTR_ERR(vma) == -EAGAIN && fallback_to_mmap_lock(priv, *ppos))
>  			goto retry;
>
> +		/*
> +		 * seq_file checks that the position advanced before it looks at
> +		 * the returned pointer, so an error return that leaves *ppos
> +		 * alone is reported as a buggy ->next. Advance it, and keep
> +		 * last_pos at the pending position so m_start() would resume
> +		 * the lookup from there.
> +		 */

We don't want (schloppy?) comments that go into painful detail about super
specific elements of the code.

This is asking for bit rot and is too complicated to be in any way useful.

> +		priv->last_pos = *ppos;

This is duplicative with code below and strictly not necessary for the fix,
better to hoist I think.

> +		(*ppos)++;
> +
>  		return vma;
>  	}
>
> --
> 2.43.0
>

Suggested:

----8<----
From 860d936c61bb641b72135e95c291aa6dfbf5fb0f Mon Sep 17 00:00:00 2001
From: "Lorenzo Stoakes (ARM)" <ljs@kernel.org>
Date: Mon, 16 Nov 2026 14:33:12 +0000
Subject: [PATCH] tmp

---
 fs/proc/task_mmu.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/fs/proc/task_mmu.c b/fs/proc/task_mmu.c
index 22361607020d..564e52f7ddc6 100644
--- a/fs/proc/task_mmu.c
+++ b/fs/proc/task_mmu.c
@@ -216,16 +216,17 @@ static struct vm_area_struct *proc_get_vma(struct seq_file *m, loff_t *ppos)

 retry:
 	vma = get_next_vma(priv, *ppos);
-	/* EINTR of EAGAIN is possible */
+	/* Store previous position to be able to restart if needed */
+	priv->last_pos = *ppos;
 	if (IS_ERR(vma)) {
 		if (PTR_ERR(vma) == -EAGAIN && fallback_to_mmap_lock(priv, *ppos))
 			goto retry;

+		/* Advance to avoid spurious buggy seq warning in dmesg */
+		(*ppos)++;
 		return vma;
 	}

-	/* Store previous position to be able to restart if needed */
-	priv->last_pos = *ppos;
 	if (vma) {
 		/*
 		 * Track the end of the reported vma to ensure position changes
--
2.56.0


^ permalink raw reply related	[flat|nested] 6+ messages in thread

end of thread, other threads:[~2026-11-16 14:37 UTC | newest]

Thread overview: 6+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-12 14:01 [PATCH mm-unstable 1/1] fs/proc/task_mmu: advance seq_file position on interrupted VMA lookup Adrian Huang (Lenovo)
2026-11-13  2:37 ` Andrew Morton
2026-11-13  3:13   ` Suren Baghdasaryan
2026-11-13  3:17     ` Suren Baghdasaryan
2026-11-16  6:26       ` Huang Adrian
2026-11-16 14:37 ` Lorenzo Stoakes (ARM)

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox