sched/fair: don't NUMA balance for kthreads
PATCH 1/1 sched/fair: don't NUMA balance for kthreads
Stefano reported a crash with using SQPOLL with io_uring:
BUG: kernel NULL pointer dereference, address: 00000000000003b0 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page PGD 800000046c042067 P4D 800000046c042067 PUD 461fcf067 PMD 0 Oops: 0000 [#1] SMP PTI CPU: 2 PID: 1307 Comm: io_uring-sq Not tainted 5.7.0-rc7 #11 Hardware name: Dell Inc. PowerEdge R430/03XKDV, BIOS 1.2.6 06/08/2015 RIP: 0010:task_numa_work+0x4f/0x2c0 Code: 18 4c 8b 25 e3 f0 8e 01 49 8b 9f 00 08 00 00 4d 8b af c8 00 00 00 49 39 c7 0f 85 e8 01 00 00 48 89 6d 00 41 f6 47 24 04 75 67 <48> 8b ab b0 03 00 00 48 85 ed 75 16 8b 3d 6f 68 94 01 e8 aa fb 04 RSP: 0018:ffffaaa98415be10 EFLAGS: 00010246 RAX: ffff953ee36b8000 RBX: 0000000000000000 RCX: 0000000000000000 RDX: 0000000000000001 RSI: ffff953ee36b8000 RDI: ffff953ee36b8dc8 RBP: ffff953ee36b8dc8 R08: 00000000001200db R09: ffff9542e3ad2e08 R10: ffff9542ecd20070 R11: 0000000000000000 R12: 00000000fffca35b R13: 000000012a06a949 R14: ffff9542e3ad2c00 R15: ffff953ee36b8000 FS: 0000000000000000(0000) GS:ffff953eefc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000000003b0 CR3: 000000046bbd0002 CR4: 00000000001606e0 Call Trace: task_work_run+0x68/0xa0 io_sq_thread+0x252/0x3d0 ? finish_wait+0x80/0x80 kthread+0xf9/0x130 ? __ia32_sys_io_uring_enter+0x370/0x370 ? kthread_park+0x90/0x90 ret_from_fork+0x35/0x40
which is task_numa_work() oopsing on current->mm being NULL. The task work is queued by task_tick_numa(), which checks if current->mm is NULL at the time of the call. But this state isn't necessarily persistent, if the kthread is using use_mm() to temporarily adopt the mm of a task.
Change the task_tick_numa() check to exclude kernel threads in general, as it doesn't make sense to attempt ot balance for kthreads anyway.
---
This should go into 5.7 imho
kernel/sched/fair.c +1 -1
@@ -2908,7 +2908,7 @@ static void task_tick_numa(struct rq *rq, struct task_struct *curr)/* * We don't care about NUMA placement if we don't have memory. */ -if (!curr->mm || (curr->flags &PF_EXITING ) || work->next != work) +if ((curr->flags & (PF_EXITING |PF_KTHREAD )) || work->next != work)return ;/*
Signature · 1 line
Jens Axboe
[tip: sched/urgent] sched/fair: Don't NUMA balance for kthreads
The following commit has been merged into the sched/urgent branch of tip:
The following commit has been merged into the sched/urgent branch of tip:
Commit-ID: 18f855e574d9799a0e7489f8ae6fd8447d0dd74a Gitweb: https://git.kernel.org/tip/18f855e574d9799a0e7489f8ae6fd8447d0dd74a Author: Jens Axboe <[email protected]> AuthorDate: Tue, 26 May 2020 09:38:31 -06:00 Committer: Ingo Molnar <[email protected]> CommitterDate: Tue, 26 May 2020 18:34:58 +02:00
sched/fair: Don't NUMA balance for kthreads
Stefano reported a crash with using SQPOLL with io_uring:
BUG: kernel NULL pointer dereference, address: 00000000000003b0 CPU: 2 PID: 1307 Comm: io_uring-sq Not tainted 5.7.0-rc7 #11 RIP: 0010:task_numa_work+0x4f/0x2c0 Call Trace: task_work_run+0x68/0xa0 io_sq_thread+0x252/0x3d0 kthread+0xf9/0x130 ret_from_fork+0x35/0x40
which is task_numa_work() oopsing on current->mm being NULL.
The task work is queued by task_tick_numa(), which checks if current->mm is NULL at the time of the call. But this state isn't necessarily persistent, if the kthread is using use_mm() to temporarily adopt the mm of a task.
Change the task_tick_numa() check to exclude kernel threads in general, as it doesn't make sense to attempt ot balance for kthreads anyway.
--- kernel/sched/fair.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
kernel/sched/fair.c +1 -1
@@ -2908,7 +2908,7 @@ static void task_tick_numa(struct rq *rq, struct task_struct *curr)/* * We don't care about NUMA placement if we don't have memory. */ -if (!curr->mm || (curr->flags &PF_EXITING ) || work->next != work) +if ((curr->flags & (PF_EXITING |PF_KTHREAD )) || work->next != work)return ;/*
Quoted 29 lines · Jens Axboe
Stefano reported a crash with using SQPOLL with
io_uring:BUG: kernel NULL pointer dereference, address: 00000000000003b0 #PF: supervisor read access in kernel mode #PF: error_code(0x0000) - not-present page PGD 800000046c042067 P4D 800000046c042067 PUD 461fcf067 PMD 0 Oops: 0000 [#1] SMP PTI CPU: 2 PID: 1307 Comm: io_uring-sq Not tainted 5.7.0-rc7 #11 Hardware name: Dell Inc. PowerEdge R430/03XKDV, BIOS 1.2.6 06/08/2015 RIP: 0010:task_numa_work+0x4f/0x2c0 Code: 18 4c 8b 25 e3 f0 8e 01 49 8b 9f 00 08 00 00 4d 8b af c8 00 00 00 49 39 c7 0f 85 e8 01 00 00 48 89 6d 00 41 f6 47 24 04 75 67 <48> 8b ab b0 03 00 00 48 85 ed 75 16 8b 3d 6f 68 94 01 e8 aa fb 04 RSP: 0018:ffffaaa98415be10 EFLAGS: 00010246 RAX: ffff953ee36b8000 RBX: 0000000000000000 RCX: 0000000000000000 RDX: 0000000000000001 RSI: ffff953ee36b8000 RDI: ffff953ee36b8dc8 RBP: ffff953ee36b8dc8 R08: 00000000001200db R09: ffff9542e3ad2e08 R10: ffff9542ecd20070 R11: 0000000000000000 R12: 00000000fffca35b R13: 000000012a06a949 R14: ffff9542e3ad2c00 R15: ffff953ee36b8000 FS: 0000000000000000(0000) GS:ffff953eefc40000(0000) knlGS:0000000000000000 CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 CR2: 00000000000003b0 CR3: 000000046bbd0002 CR4: 00000000001606e0 Call Trace: task_work_run+0x68/0xa0 io_sq_thread+0x252/0x3d0 ? finish_wait+0x80/0x80 kthread+0xf9/0x130 ? __ia32_sys_io_uring_enter+0x370/0x370 ? kthread_park+0x90/0x90 ret_from_fork+0x35/0x40
Quoted 8 lines
which is
task_numa_work()oopsing oncurrent->mmbeing NULL. The task work is queued bytask_tick_numa(), which checks ifcurrent->mmis NULL at the time of the call. But this state isn't necessarily persistent, if the kthread is usinguse_mm()to temporarily adopt the mm of a task.Change the
task_tick_numa()check to exclude kernel threads in general, as it doesn't make sense to attempt ot balance for kthreads anyway.
Does it? (this isn't a rethorical question)
Suppose a given kthread ends up doing more accesses to some pages (via use_mm()) than the other threads that access them, wouldn't it make sense to take that into account when it comes to NUMA balancing?
AFAIA that's not something that really happens in the wild, and as you say we don't have a (guaranteed) persistent state to work with in those cases, so it's more of a pipe dream. With that in mind, I think the proposed change is fine.
Quoted 20 lines
Reported-by Stefano Garzarella <[email protected]>Signed-off-by Jens Axboe <[email protected]>---
This should go into 5.7 imho
kernel/sched/fair.c +1
-1@@ -2908,7 +2908,7 @@ static void task_tick_numa(struct rq *rq, struct task_struct *curr)/* * We don't care about NUMA placement if we don't have memory. */ -if (!curr->mm || (curr->flags &PF_EXITING ) || work->next != work) +if ((curr->flags & (PF_EXITING |PF_KTHREAD )) || work->next != work)return ;/*
Peter Zijlstra
maintainerQuoted 9 lines · Valentin Schneider
Quoted 3 lines
Change the
task_tick_numa()check to exclude kernel threads in general, as it doesn't make sense to attempt ot balance for kthreads anyway.Does it? (this isn't a rethorical question)
Suppose a given kthread ends up doing more accesses to some pages (via
use_mm()) than the other threads that access them, wouldn't it make sense to take that into account when it comes to NUMA balancing?
Well, task_tick_numa() tries and farm off a bunch of actual work to task_work_add(), and there's so very little userspace for a kernel thread to return to... :-)
Quoted 8 lines · Peter Zijlstra
Quoted 6 lines · Valentin Schneider
Quoted 3 lines
Change the
task_tick_numa()check to exclude kernel threads in general, as it doesn't make sense to attempt ot balance for kthreads anyway.Does it? (this isn't a rethorical question)
Quoted 7 lines
Quoted 3 lines
Suppose a given kthread ends up doing more accesses to some pages (via
use_mm()) than the other threads that access them, wouldn't it make sense to take that into account when it comes to NUMA balancing?Well,
task_tick_numa()tries and farm off a bunch of actual work totask_work_add(), and there's so very little userspace for a kernel thread to return to... :-)
Err, true... I did say pipe dreams!
I had only really taken note of the exit / return to userspace callbacks, but I see io_uring has its own task_work_run() calls, which (I think) explains how we can end up with a kthread actually running task_numa_work().
I'm also thinking we really don't want that task_numa_work() to be left hanging on the task_work list, because that self-looping thing will not play nice to whatever else has been queued (which AFAICT shouldn't happen under normal conditions, i.e. !kthreads).