RDMA/core: Fix WARNING in gid_table_release_one
PATCH 1/1 RDMA/core: Fix WARNING in gid_table_release_one
From: "Zhu Yanjun" <[email protected]>
GID entry ref leak for dev syz1 index 2 ref=615
...
Call Trace: <TASK> ib_device_release+0xd2/0x1c0 drivers/infiniband/core/device.c:509 device_release+0x99/0x1c0 drivers/base/core.c:-1 kobject_cleanup lib/kobject.c:689 [inline] kobject_release lib/kobject.c:720 [inline] kref_put include/linux/kref.h:65 [inline] kobject_put+0x228/0x480 lib/kobject.c:737 process_one_work kernel/workqueue.c:3263 [inline] process_scheduled_works+0xae1/0x17b0 kernel/workqueue.c:3346 worker_thread+0x8a0/0xda0 kernel/workqueue.c:3427 kthread+0x711/0x8a0 kernel/kthread.c:463 ret_from_fork+0x47c/0x820 arch/x86/kernel/process.c:158 ret_from_fork_asm+0x1a/0x30 arch/x86/entry/entry_64.S:245 </TASK>
When the state of a GID is GID_TABLE_ENTRY_PENDING_DEL, it indicates that the GID is about to be released soon. Therefore, it does not appear to be a leak.
The release_gid_table() function also waits for a short period; if the GID still exists after that, a GID leak warning will be triggered.
--- drivers/infiniband/core/cache.c | 17 ++++++++++++++--- 1 file changed, 14 insertions(+), 3 deletions(-)
drivers/infiniband/core/cache.c +14 -3
@@ -800,13 +800,24 @@ static void release_gid_table(struct ib_device *device,return ;for (i =0 ; i < table->sz; i++) { +int cnt =200 ; +if (is_gid_entry_free (table->data_vec[i]))continue ; -WARN_ONCE (true, -"GID entry ref leak for dev %s index %d ref=%u\n" , +WARN_ONCE (table->data_vec[i]->state !=GID_TABLE_ENTRY_PENDING_DEL , +"GID entry ref leak for dev %s index %d ref=%u, state: %d\n" ,dev_name (&device->dev), i, -kref_read (&table->data_vec[i]->kref)); +kref_read (&table->data_vec[i]->kref), table->data_vec[i]->state); + +while ((kref_read (&table->data_vec[i]->kref) >0 ) && (cnt >0 )) { + cnt--; +msleep (10 ); + } + +if (cnt <=0 ) +pr_warn_ratelimited ("Possibly kref leak, ref:%u, state: %d\n" , +kref_read (&table->data_vec[i]->kref), table->data_vec[i]->state); }mutex_destroy (&table->lock);
Signature · 1 line
2.51.0
Quoted 11 lines · Zhu Yanjun
(quoted diff) +2
-2@@ -800,13 +800,24 @@ static void release_gid_table(struct ib_device *device, return; for (i = 0; i < table->sz; i++) { + int cnt = 200; + if (is_gid_entry_free(table->data_vec[i])) continue; - WARN_ONCE(true, - "GID entry ref leak for dev %s index %d ref=%u\n",
Review static void release_gid_table(struct ib_device *device, · line 816
Quoted 10 lines
(quoted diff) +8
-1+ WARN_ONCE(table->data_vec[i]->state != GID_TABLE_ENTRY_PENDING_DEL, + "GID entry ref leak for dev %s index %d ref=%u, state: %d\n", dev_name(&device->dev), i, - kref_read(&table->data_vec[i]->kref)); + kref_read(&table->data_vec[i]->kref), table->data_vec[i]->state); + + while ((kref_read(&table->data_vec[i]->kref) > 0) && (cnt > 0)) { + cnt--; + msleep(10); + }
Definately don't want to see this looping.
If it is waiting for the work queue then maybe this should flush the work queue.
Something like this?
drivers/infiniband/core/cache.c +13 -0
@@ -799,7 +799,19 @@ static void release_gid_table(struct ib_device *device,if (!table)return ; +mutex_lock (&table->lock);for (i =0 ; i < table->sz; i++) { +if (is_gid_entry_free (table->data_vec[i])) +continue ; + +/* +* The entry may be sitting in the WQ waiting for +* free_gid_work(), flush it to try to clean it. +*/ +mutex_unlock (&table->lock); +flush_workqueue (ib_wq); +mutex_lock (&table->lock); +if (is_gid_entry_free (table->data_vec[i]))continue ;
@@ -808,6 +820,7 @@ static void release_gid_table(struct ib_device *device,dev_name (&device->dev), i, kref_read(&table->data_vec[i]->kref)); } +mutex_unlock (&table->lock);mutex_destroy (&table->lock);kfree (table->data_vec);
Quoted 24 lines · Jason Gunthorpe
Quoted 11 lines · Zhu Yanjun
(quoted diff) +2
-2@@ -800,13 +800,24 @@ static void release_gid_table(struct ib_device *device, return; for (i = 0; i < table->sz; i++) { + int cnt = 200; + if (is_gid_entry_free(table->data_vec[i])) continue; - WARN_ONCE(true, - "GID entry ref leak for dev %s index %d ref=%u\n",Review static void release_gid_table(struct ib_device *device, · line 816
Quoted 10 lines
(quoted diff) +8
-1+ WARN_ONCE(table->data_vec[i]->state != GID_TABLE_ENTRY_PENDING_DEL, + "GID entry ref leak for dev %s index %d ref=%u, state: %d\n", dev_name(&device->dev), i, - kref_read(&table->data_vec[i]->kref)); + kref_read(&table->data_vec[i]->kref), table->data_vec[i]->state); + + while ((kref_read(&table->data_vec[i]->kref) > 0) && (cnt > 0)) { + cnt--; + msleep(10); + }Definately don't want to see this looping.
Quoted 2 lines
If it is waiting for the work queue then maybe this should flush the work queue.
Thanks a lot, Jason. I will delve into your suggestions and follow your advice to send a new patch for this problem.
Yanjun.Zhu
Quoted 35 lines
Something like this?
drivers/infiniband/core/cache.c +13
-0@@ -799,7 +799,19 @@ static void release_gid_table(struct ib_device *device,if (!table)return ; +mutex_lock (&table->lock);for (i =0 ; i < table->sz; i++) { +if (is_gid_entry_free (table->data_vec[i])) +continue ; + +/* +* The entry may be sitting in the WQ waiting for +* free_gid_work(), flush it to try to clean it. +*/ +mutex_unlock (&table->lock); +flush_workqueue (ib_wq); +mutex_lock (&table->lock); +if (is_gid_entry_free (table->data_vec[i]))continue ;@@ -808,6 +820,7 @@ static void release_gid_table(struct ib_device *device,dev_name (&device->dev), i, kref_read(&table->data_vec[i]->kref)); } +mutex_unlock (&table->lock);mutex_destroy (&table->lock);kfree (table->data_vec);
Signature · 2 lines
Best Regards, Yanjun.Zhu
Quoted 38 lines · Jason Gunthorpe
Quoted 11 lines · Zhu Yanjun
(quoted diff) +2
-2@@ -800,13 +800,24 @@ static void release_gid_table(struct ib_device *device, return; for (i = 0; i < table->sz; i++) { + int cnt = 200; + if (is_gid_entry_free(table->data_vec[i])) continue; - WARN_ONCE(true, - "GID entry ref leak for dev %s index %d ref=%u\n",Review static void release_gid_table(struct ib_device *device, · line 816
Quoted 10 lines
(quoted diff) +8
-1+ WARN_ONCE(table->data_vec[i]->state != GID_TABLE_ENTRY_PENDING_DEL, + "GID entry ref leak for dev %s index %d ref=%u, state: %d\n", dev_name(&device->dev), i, - kref_read(&table->data_vec[i]->kref)); + kref_read(&table->data_vec[i]->kref), table->data_vec[i]->state); + + while ((kref_read(&table->data_vec[i]->kref) > 0) && (cnt > 0)) { + cnt--; + msleep(10); + }Definately don't want to see this looping.
If it is waiting for the work queue then maybe this should flush the work queue.
Something like this?
drivers/infiniband/core/cache.c +1
-0@@ -799,7 +799,19 @@ static void release_gid_table(struct ib_device *device,if (!table)return ; +mutex_lock (&table->lock);for (i =0 ; i < table->sz; i++) {
drivers/infiniband/core/cache.c static void release_gid_table(struct ib_device *device, · line 813
Quoted 10 lines
drivers/infiniband/core/cache.c +10
-0+if (is_gid_entry_free (table->data_vec[i])) +continue ; + +/* +* The entry may be sitting in the WQ waiting for +* free_gid_work(), flush it to try to clean it. +*/ +mutex_unlock (&table->lock); +flush_workqueue (ib_wq); +mutex_lock (&table->lock);
I don't think that this is right thing to do. If you want, you can call to flush_workqueue(ib_wq) in ib_cache_release_one().
Thanks
Quoted 15 lines
drivers/infiniband/core/cache.c +2
-0+if (is_gid_entry_free (table->data_vec[i]))continue ;@@ -808,6 +820,7 @@ static void release_gid_table(struct ib_device *device,dev_name (&device->dev), i, kref_read(&table->data_vec[i]->kref)); } +mutex_unlock (&table->lock);mutex_destroy (&table->lock);kfree (table->data_vec);