Skip to contentSkip to conversation

mm: slub: fix flush_cpu_slab()/__free_slab() invocations in task context.

PATCH Superseded Maurizio Lombardi 3 participants 5 messages Sep 12, 2022 – Sep 16, 2022

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 mm: slub: fix flush_cpu_slab()/__free_slab() invocations in task context.

Commit 5a836bf6b09f ("mm: slub: move flush_cpu_slab() invocations __free_slab() invocations out of IRQ context") moved all flush_cpu_slab() invocations to the global workqueue to avoid a problem related with deactivate_slab()/__free_slab() being called from an IRQ context on PREEMPT_RT kernels.

When the flush_all_cpu_locked() function is called from a task context it may happen that a workqueue with WQ_MEM_RECLAIM bit set ends up flushing the global workqueue, this will cause a dependency issue.

workqueue: WQ_MEM_RECLAIM nvme-delete-wq:nvme_delete_ctrl_work [nvme_core]
is flushing !WQ_MEM_RECLAIM events:flush_cpu_slab

 WARNING: CPU: 37 PID: 410 at kernel/workqueue.c:2637
   check_flush_dependency+0x10a/0x120
 Workqueue: nvme-delete-wq nvme_delete_ctrl_work [nvme_core]
 RIP: 0010:check_flush_dependency+0x10a/0x120[  453.262125] Call Trace:
 __flush_work.isra.0+0xbf/0x220
 ? __queue_work+0x1dc/0x420
 flush_all_cpus_locked+0xfb/0x120
 __kmem_cache_shutdown+0x2b/0x320
 kmem_cache_destroy+0x49/0x100
 bioset_exit+0x143/0x190
 blk_release_queue+0xb9/0x100
 kobject_cleanup+0x37/0x130
 nvme_fc_ctrl_free+0xc6/0x150 [nvme_fc]
 nvme_free_ctrl+0x1ac/0x2b0 [nvme_core]

Fix this bug by creating a workqueue for the flush operation with the WQ_MEM_RECLAIM bit set.

v2: Create a workqueue with WQ_MEM_RECLAIM
instead of trying to revert the changes.

---
 mm/slub.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
mm/slub.c +8 -1
index 862dbd9af4f5..cdb8011884d0 100644
@@ -310,6 +310,11 @@ static inline void stat(const struct kmem_cache *s, enum stat_item si)
  */
 static nodemask_t slab_nodes;
 
+/*
+ * Workqueue used for flush_cpu_slab().
+ */
+static struct workqueue_struct *flushwq;
+
 /********************************************************************
  * 			Core slab cache functions
  *******************************************************************/
@@ -2730,7 +2735,7 @@ static void flush_all_cpus_locked(struct kmem_cache *s)
 		INIT_WORK(&sfw->work, flush_cpu_slab);
 		sfw->skip = false;
 		sfw->s = s;
-		schedule_work_on(cpu, &sfw->work);
+		queue_work_on(cpu, flushwq, &sfw->work);
 	}
 
 	for_each_online_cpu(cpu) {
@@ -4858,6 +4863,8 @@ void __init kmem_cache_init(void)
 
 void __init kmem_cache_init_late(void)
 {
+	flushwq = create_workqueue("slub_flushwq");
+	BUG_ON(!flushwq);
 }
 
 struct kmem_cache *
Signature · 1 line
2.31.1
#1 lore raw
Quoted 75 lines · Maurizio Lombardi

Commit 5a836bf6b09f ("mm: slub: move flush_cpu_slab() invocations __free_slab() invocations out of IRQ context") moved all flush_cpu_slab() invocations to the global workqueue to avoid a problem related with deactivate_slab()/__free_slab() being called from an IRQ context on PREEMPT_RT kernels.

When the flush_all_cpu_locked() function is called from a task context it may happen that a workqueue with WQ_MEM_RECLAIM bit set ends up flushing the global workqueue, this will cause a dependency issue.

workqueue: WQ_MEM_RECLAIM nvme-delete-wq:nvme_delete_ctrl_work [nvme_core]
is flushing !WQ_MEM_RECLAIM events:flush_cpu_slab

 WARNING: CPU: 37 PID: 410 at kernel/workqueue.c:2637
   check_flush_dependency+0x10a/0x120
 Workqueue: nvme-delete-wq nvme_delete_ctrl_work [nvme_core]
 RIP: 0010:check_flush_dependency+0x10a/0x120[  453.262125] Call Trace:
 __flush_work.isra.0+0xbf/0x220
 ? __queue_work+0x1dc/0x420
 flush_all_cpus_locked+0xfb/0x120
 __kmem_cache_shutdown+0x2b/0x320
 kmem_cache_destroy+0x49/0x100
 bioset_exit+0x143/0x190
 blk_release_queue+0xb9/0x100
 kobject_cleanup+0x37/0x130
 nvme_fc_ctrl_free+0xc6/0x150 [nvme_fc]
 nvme_free_ctrl+0x1ac/0x2b0 [nvme_core]

Fix this bug by creating a workqueue for the flush operation with the WQ_MEM_RECLAIM bit set.

v2: Create a workqueue with WQ_MEM_RECLAIM
instead of trying to revert the changes.

---
 mm/slub.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
mm/slub.c +8 -1
index 862dbd9af4f5..cdb8011884d0 100644
@@ -310,6 +310,11 @@ static inline void stat(const struct kmem_cache *s, enum stat_item si)
  */
 static nodemask_t slab_nodes;
 
+/*
+ * Workqueue used for flush_cpu_slab().
+ */
+static struct workqueue_struct *flushwq;
+
 /********************************************************************
  * 			Core slab cache functions
  *******************************************************************/
@@ -2730,7 +2735,7 @@ static void flush_all_cpus_locked(struct kmem_cache *s)
 		INIT_WORK(&sfw->work, flush_cpu_slab);
 		sfw->skip = false;
 		sfw->s = s;
-		schedule_work_on(cpu, &sfw->work);
+		queue_work_on(cpu, flushwq, &sfw->work);
 	}
 
 	for_each_online_cpu(cpu) {
@@ -4858,6 +4863,8 @@ void __init kmem_cache_init(void)
 
 void __init kmem_cache_init_late(void)
 {
+	flushwq = create_workqueue("slub_flushwq");
+	BUG_ON(!flushwq);
 }
 
 struct kmem_cache *
Signature · 1 line
2.31.1

No objection about this patch,
but it seems create_workqueue is deprecated?

Signature · 2 lines
Thanks,
Hyeonggon
#2 lore raw

čt 15. 9. 2022 v 14:48 odesílatel Hyeonggon Yoo <[email protected]> napsal:

Quoted 2 lines

No objection about this patch,
but it seems create_workqueue is deprecated?

You're right, the documentation says that create_workqueue() is deprecated, too bad there is not a comment in the source code.
Thanks for reporting this.

I will wait for the maintainers to review this one and eventually I will submit a V3.

Maurizio

Quoted 6 lines

--
Thanks,
Hyeonggon

#3 lore raw
Quoted 6 lines · Maurizio Lombardi

čt 15. 9. 2022 v 14:48 odesílatel Hyeonggon Yoo <[email protected]> napsal:

Quoted 2 lines

No objection about this patch,
but it seems create_workqueue is deprecated?

You're right, the documentation says that create_workqueue() is deprecated, too bad there is not a comment in the source code.

Let's make Tejun and Lai aware of this.

Quoted 16 lines

Thanks for reporting this.

I will wait for the maintainers to review this one and eventually I will submit a V3.

Maurizio

Quoted 6 lines

--
Thanks,
Hyeonggon

#4 lore raw
In reply to Maurizio Lombardi
Quoted 10 lines · Maurizio Lombardi
mm/slub.c +1 -1
@@ -2730,7 +2735,7 @@ static void flush_all_cpus_locked(struct kmem_cache *s)
 		INIT_WORK(&sfw->work, flush_cpu_slab);
 		sfw->skip = false;
 		sfw->s = s;
-		schedule_work_on(cpu, &sfw->work);
+		queue_work_on(cpu, flushwq, &sfw->work);
 	}
 
 	for_each_online_cpu(cpu) {

mm/slub.c void __init kmem_cache_init(void) · line 4867

Quoted 5 lines
mm/slub.c +2 -0
@@ -4858,6 +4863,8 @@ void __init kmem_cache_init(void)
 
 void __init kmem_cache_init_late(void)
 {
+	flushwq = create_workqueue("slub_flushwq");
+	BUG_ON(!flushwq);

Adding new BUG_ON is essentially prohibited these days, so please use WARN_ON(). Other than that deprecation issue, looks fine to me.

Quoted 3 lines
mm/slub.c +0 -0
 }
 
 struct kmem_cache *