Skip to contentSkip to conversation

revert: loop: queue_lock NULL pointer derefence in blk_throtl_exit V1

PATCH New Sergey Senozhatsky 2 participants 3 messages Feb 11, 2011 Block layer

lore
Share & tools
lore
#0 lore raw

Hello Jens, Andrew

We have accidentally commit "loop: queue_lock NULL pointer
derefence in blk_throtl_exit V1" applied: ee71a968672a9951aee6014c55511007596425bc

It's not a huge problem, but I think, it's better to have V3 applied
https://lkml.org/lkml/2011/1/21/158

since V3 targets all the throttle users, not only loop (by moving `fall-back' from loop_free to blk_release_queue).

V3 is in Andrew's mmotm as for now loop-queue_lock-null-pointer-derefence-in-blk_throtl_exit-v3.patch

Could we please revert ee71a968672a9951aee6014c55511007596425bc when loop-queue_lock-null-pointer-derefence-in-blk_throtl_exit-v3 will land in the kernel?

Sorry.

Thanks in advance,

Sergey

#1 lore raw
Quoted 11 lines · Sergey Senozhatsky

Hello Jens, Andrew

We have accidentally commit "loop: queue_lock NULL pointer
derefence in blk_throtl_exit V1" applied: ee71a968672a9951aee6014c55511007596425bc

It's not a huge problem, but I think, it's better to have V3 applied
https://lkml.org/lkml/2011/1/21/158

since V3 targets all the throttle users, not only loop (by moving `fall-back' from loop_free to blk_release_queue).

Quoted 6 lines

V3 is in Andrew's mmotm as for now loop-queue_lock-null-pointer-derefence-in-blk_throtl_exit-v3.patch

Could we please revert ee71a968672a9951aee6014c55511007596425bc when loop-queue_lock-null-pointer-derefence-in-blk_throtl_exit-v3 will land in the kernel?

Send an incremental patch if there's overlap, otherwise we'll do the revert. I'm on my way out the door and I'll be away from the computer all of next week, so perhaps Andrew can assist you in upstreaming such a patch?

Signature · 1 line
Jens Axboe
#2 lore raw

PATCH 1/1 block: fix queue_lock NULL pointer derefence in blk_throtl_exit (v4)

block: fix queue_lock NULL pointer derefence in blk_throtl_exit

BUG: unable to handle kernel NULL pointer dereference at 0000000000000004
 IP: [<ffffffff812479d4>] do_raw_spin_lock+0x14/0x122
 Process modprobe (pid: 6189, threadinfo ffff88009a898000, task
 ffff880154a88000)                                                        
 Call Trace:                                         
  [<ffffffff81486788>] _raw_spin_lock_irq+0x4a/0x51                         
  [<ffffffff8123404b>] ? blk_throtl_exit+0x3b/0xa0
  [<ffffffff8105b120>] ? cancel_delayed_work_sync+0xd/0xf
  [<ffffffff8123404b>] blk_throtl_exit+0x3b/0xa0   
  [<ffffffff81229bc8>] blk_release_queue+0x21/0x65
  [<ffffffff8123bb06>] kobject_release+0x51/0x66         
  [<ffffffff8123bab5>] ? kobject_release+0x0/0x66
  [<ffffffff8123ce1e>] kref_put+0x43/0x4d         
  [<ffffffff8123ba27>] kobject_put+0x47/0x4b    
  [<ffffffff8122717c>] blk_cleanup_queue+0x56/0x5b
  [<ffffffffa01c3824>] loop_exit+0x68/0x844 [loop]
  [<ffffffff8107cccc>] sys_delete_module+0x1e8/0x25b
  [<ffffffff814864c9>] ? trace_hardirqs_on_thunk+0x3a/0x3f
  [<ffffffff81002112>] system_call_fastpath+0x16/0x1b

because of an attempt to acquire NULL queue_lock.
Added the same lines as in blk_queue_make_request:
'fall back to embedded per-queue lock' - when call blk_release_queue on allocated but never initialized queue.

v4:
Incremental patch to move fall back to embedded per-queue lock from loop_free to blk_release_queue, because v1 patch was accidentally applied.

v3: According to comment by Vivek Goyal, queue_lock NULL check and fix moved out from loop driver code to blk_release_queue.

---

 block/blk-sysfs.c    |    6 ++++++
 drivers/block/loop.c |    3 ---
 2 files changed, 6 insertions(+), 3 deletions(-)
block/blk-sysfs.c +6 -0
index 41fb691..2b426ac 100644
@@ -471,6 +471,12 @@ static void blk_release_queue(struct kobject *kobj)
 
 	blk_sync_queue(q);
 
+	/* It's possible that blk_release_queue will be called on allocated
+	 * but never initialized queue. Fall back to our embedded per-queue 
+	 * locks in this case. */
+	if (!q->queue_lock)
+		q->queue_lock = &q->__queue_lock;
+
 	blk_throtl_exit(q);
 
 	if (rl->rq_pool)
drivers/block/loop.c +0 -3
index 49e6a54..44e18c0 100644
@@ -1641,9 +1641,6 @@ out:
 
 static void loop_free(struct loop_device *lo)
 {
-	if (!lo->lo_queue->queue_lock)
-		lo->lo_queue->queue_lock = &lo->lo_queue->__queue_lock;
-
 	blk_cleanup_queue(lo->lo_queue);
 	put_disk(lo->lo_disk);
 	list_del(&lo->lo_list);