[PATCH] sched/numa: use down_read_trylock for mmap_sem

[PATCH] sched/numa: use down_read_trylock for mmap_sem

[Vlastimil Babka]

A customer has reported a soft-lockup when running a proprietary intensive
memory stress test, where the trace on multiple CPU's looks like this:

 RIP: 0010:[<ffffffff810c53fe>]
  [<ffffffff810c53fe>] native_queued_spin_lock_slowpath+0x10e/0x190
...
 Call Trace:
  [<ffffffff81182d07>] queued_spin_lock_slowpath+0x7/0xa
  [<ffffffff811bc331>] change_protection_range+0x3b1/0x930
  [<ffffffff811d4be8>] change_prot_numa+0x18/0x30
  [<ffffffff810adefe>] task_numa_work+0x1fe/0x310
  [<ffffffff81098322>] task_work_run+0x72/0x90

Further investigation showed that the lock contention here is pmd_lock().

The task_numa_work() function makes sure that only one thread is let to perform
the work in a single scan period (via cmpxchg), but if there's a thread with
mmap_sem locked for writing for several periods, multiple threads in
task_numa_work() can build up a convoy waiting for mmap_sem for read and then
all get unblocked at once.

This patch changes the down_read() to the trylock version, which prevents the
build up. For a workload experiencing mmap_sem contention, it's probably better
to postpone the NUMA balancing work anyway. This seems to have fixed the soft
lockups involving pmd_lock(), which is in line with the convoy theory.

Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
---
 kernel/sched/fair.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index dea138964b91..d70f9026defc 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2475,7 +2475,8 @@ void task_numa_work(struct callback_head *work)
 		return;
 
 
-	down_read(&mm->mmap_sem);
+	if (!down_read_trylock(&mm->mmap_sem))
+		return;
 	vma = find_vma(mm, start);
 	if (!vma) {
 		reset_ptenuma_scan(p);
-- 
2.12.2

Re: [PATCH] sched/numa: use down_read_trylock for mmap_sem

[Rik van Riel]

On Mon, 2017-05-15 at 15:13 +0200, Vlastimil Babka wrote:
> A customer has reported a soft-lockup when running a proprietary
> intensive
> memory stress test, where the trace on multiple CPU's looks like
> this:
> 
>  RIP: 0010:[<ffffffff810c53fe>]
>   [<ffffffff810c53fe>] native_queued_spin_lock_slowpath+0x10e/0x190
> ...
>  Call Trace:
>   [<ffffffff81182d07>] queued_spin_lock_slowpath+0x7/0xa
>   [<ffffffff811bc331>] change_protection_range+0x3b1/0x930
>   [<ffffffff811d4be8>] change_prot_numa+0x18/0x30
>   [<ffffffff810adefe>] task_numa_work+0x1fe/0x310
>   [<ffffffff81098322>] task_work_run+0x72/0x90
> 
> Further investigation showed that the lock contention here is
> pmd_lock().
> 
> The task_numa_work() function makes sure that only one thread is let
> to perform
> the work in a single scan period (via cmpxchg), but if there's a
> thread with
> mmap_sem locked for writing for several periods, multiple threads in
> task_numa_work() can build up a convoy waiting for mmap_sem for read
> and then
> all get unblocked at once.
> 
> This patch changes the down_read() to the trylock version, which
> prevents the
> build up. For a workload experiencing mmap_sem contention, it's
> probably better
> to postpone the NUMA balancing work anyway. This seems to have fixed
> the soft
> lockups involving pmd_lock(), which is in line with the convoy
> theory.
> 
> Signed-off-by: Vlastimil Babka <vbabka@suse.cz>

Acked-by: Rik van Riel <riel@redhat.com>

Re: [PATCH] sched/numa: use down_read_trylock for mmap_sem

[Mel Gorman]

On Mon, May 15, 2017 at 03:13:16PM +0200, Vlastimil Babka wrote:
> A customer has reported a soft-lockup when running a proprietary intensive
> memory stress test, where the trace on multiple CPU's looks like this:
> 
>  RIP: 0010:[<ffffffff810c53fe>]
>   [<ffffffff810c53fe>] native_queued_spin_lock_slowpath+0x10e/0x190
> ...
>  Call Trace:
>   [<ffffffff81182d07>] queued_spin_lock_slowpath+0x7/0xa
>   [<ffffffff811bc331>] change_protection_range+0x3b1/0x930
>   [<ffffffff811d4be8>] change_prot_numa+0x18/0x30
>   [<ffffffff810adefe>] task_numa_work+0x1fe/0x310
>   [<ffffffff81098322>] task_work_run+0x72/0x90
> 
> Further investigation showed that the lock contention here is pmd_lock().
> 
> The task_numa_work() function makes sure that only one thread is let to perform
> the work in a single scan period (via cmpxchg), but if there's a thread with
> mmap_sem locked for writing for several periods, multiple threads in
> task_numa_work() can build up a convoy waiting for mmap_sem for read and then
> all get unblocked at once.
> 
> This patch changes the down_read() to the trylock version, which prevents the
> build up. For a workload experiencing mmap_sem contention, it's probably better
> to postpone the NUMA balancing work anyway. This seems to have fixed the soft
> lockups involving pmd_lock(), which is in line with the convoy theory.
> 
> Signed-off-by: Vlastimil Babka <vbabka@suse.cz>

This skips the entire scan window and defers to the next once.
Potentially, with constant contention, it'll never make progress and
there could be other disruption. However, I cannot see any way how
that's worse than waiting on mmap_sem so

Acked-by: Mel Gorman <mgorman@techsingularity.net>

-- 
Mel Gorman
SUSE Labs

Re: [PATCH] sched/numa: use down_read_trylock for mmap_sem

[Peter Zijlstra]

On Mon, May 15, 2017 at 03:13:16PM +0200, Vlastimil Babka wrote:
> A customer has reported a soft-lockup when running a proprietary intensive
> memory stress test, where the trace on multiple CPU's looks like this:
> 
>  RIP: 0010:[<ffffffff810c53fe>]
>   [<ffffffff810c53fe>] native_queued_spin_lock_slowpath+0x10e/0x190
> ...
>  Call Trace:
>   [<ffffffff81182d07>] queued_spin_lock_slowpath+0x7/0xa
>   [<ffffffff811bc331>] change_protection_range+0x3b1/0x930
>   [<ffffffff811d4be8>] change_prot_numa+0x18/0x30
>   [<ffffffff810adefe>] task_numa_work+0x1fe/0x310
>   [<ffffffff81098322>] task_work_run+0x72/0x90
> 
> Further investigation showed that the lock contention here is pmd_lock().
> 
> The task_numa_work() function makes sure that only one thread is let to perform
> the work in a single scan period (via cmpxchg), but if there's a thread with
> mmap_sem locked for writing for several periods, multiple threads in
> task_numa_work() can build up a convoy waiting for mmap_sem for read and then
> all get unblocked at once.
> 
> This patch changes the down_read() to the trylock version, which prevents the
> build up. For a workload experiencing mmap_sem contention, it's probably better
> to postpone the NUMA balancing work anyway. This seems to have fixed the soft
> lockups involving pmd_lock(), which is in line with the convoy theory.
> 
> Signed-off-by: Vlastimil Babka <vbabka@suse.cz>

Thanks!

[tip:sched/core] sched/numa: Use down_read_trylock() for the mmap_sem

[tip-bot for Vlastimil Babka]

Commit-ID:  8655d5497735b288f8a9b458bd22e7d1bf95bb61
Gitweb:     http://git.kernel.org/tip/8655d5497735b288f8a9b458bd22e7d1bf95bb61
Author:     Vlastimil Babka <vbabka@suse.cz>
AuthorDate: Mon, 15 May 2017 15:13:16 +0200
Committer:  Ingo Molnar <mingo@kernel.org>
CommitDate: Tue, 23 May 2017 10:01:34 +0200

sched/numa: Use down_read_trylock() for the mmap_sem

A customer has reported a soft-lockup when running an intensive
memory stress test, where the trace on multiple CPU's looks like this:

 RIP: 0010:[<ffffffff810c53fe>]
  [<ffffffff810c53fe>] native_queued_spin_lock_slowpath+0x10e/0x190
...
 Call Trace:
  [<ffffffff81182d07>] queued_spin_lock_slowpath+0x7/0xa
  [<ffffffff811bc331>] change_protection_range+0x3b1/0x930
  [<ffffffff811d4be8>] change_prot_numa+0x18/0x30
  [<ffffffff810adefe>] task_numa_work+0x1fe/0x310
  [<ffffffff81098322>] task_work_run+0x72/0x90

Further investigation showed that the lock contention here is pmd_lock().

The task_numa_work() function makes sure that only one thread is let to perform
the work in a single scan period (via cmpxchg), but if there's a thread with
mmap_sem locked for writing for several periods, multiple threads in
task_numa_work() can build up a convoy waiting for mmap_sem for read and then
all get unblocked at once.

This patch changes the down_read() to the trylock version, which prevents the
build up. For a workload experiencing mmap_sem contention, it's probably better
to postpone the NUMA balancing work anyway. This seems to have fixed the soft
lockups involving pmd_lock(), which is in line with the convoy theory.

Signed-off-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Acked-by: Rik van Riel <riel@redhat.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Link: http://lkml.kernel.org/r/20170515131316.21909-1-vbabka@suse.cz
Signed-off-by: Ingo Molnar <mingo@kernel.org>
---
 kernel/sched/fair.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 219fe58..47a0c55 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -2470,7 +2470,8 @@ void task_numa_work(struct callback_head *work)
 		return;
 
 
-	down_read(&mm->mmap_sem);
+	if (!down_read_trylock(&mm->mmap_sem))
+		return;
 	vma = find_vma(mm, start);
 	if (!vma) {
 		reset_ptenuma_scan(p);