Skip to contentSkip to conversation

block: don't set the same context

PATCH Superseded Denis Plotnikov 2 participants 3 messages Feb 15, 2019

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 block: don't set the same context

Adds a fast path on aio context setting preventing unnecessary context setting routine.
Also, it prevents issues with cyclic walk of child bds-es appeared because of registring aio walking notifiers:

Call stack:

0  __GI_raise
1  __GI_abort
2  __assert_fail_base
3  __GI___assert_fail
4  bdrv_detach_aio_context (bs=0x55f54d65c000)      <<<
5  bdrv_detach_aio_context (bs=0x55f54fc8a800)
6  bdrv_set_aio_context (bs=0x55f54fc8a800, ...)
7  block_job_attached_aio_context
8  bdrv_attach_aio_context (bs=0x55f54d65c000, ...) <<<
9  bdrv_set_aio_context (bs=0x55f54d65c000)

10 blk_set_aio_context
11 virtio_blk_data_plane_stop
12 virtio_bus_stop_ioeventfd
13 virtio_vmstate_change
14 vm_state_notify (running=0, state=RUN_STATE_SHUTDOWN) 15 do_vm_stop (state=RUN_STATE_SHUTDOWN, send_stop=true) 16 vm_stop (state=RUN_STATE_SHUTDOWN)
17 main_loop_should_exit
18 main_loop
19 main

This can happen because of "new" context attachment to VM disk bds. When attaching a new context the corresponding aio context handler is called for each of aio_notifiers registered on the VM disk bds context. Among those handlers, there is the block_job_attached_aio_context handler which sets a new aio context for the block job bds. When doing so, the old context is detached from all the block job bds children and one of them is the VM disk bds, serving as backing store for the blockjob bds, although the VM disk bds is actually the initializer of that process. Since the VM disk bds is protected with walking_aio_notifiers flag from double processing in recursive calls, the assert fires.

---
 block.c | 4 ++++
 1 file changed, 4 insertions(+)
block.c +4 -0
index adda221c2c..5742545e7a 100644
@@ -5263,6 +5263,10 @@ void bdrv_set_aio_context(BlockDriverState *bs, AioContext *new_context)
 {
     AioContext *ctx = bdrv_get_aio_context(bs);
 
+    if (ctx == new_context) {
+        return;
+    }
+
     aio_disable_external(ctx);
     bdrv_parent_drained_begin(bs, NULL, false);
     bdrv_drain(bs); /* ensure there are no in-flight requests */
Signature · 1 line
2.17.0
#1 lore raw
Quoted 4 lines · Denis Plotnikov

Adds a fast path on aio context setting preventing unnecessary context setting routine.
Also, it prevents issues with cyclic walk of child bds-es appeared because of registring aio walking

registering

Quoted 14 lines

notifiers:

This can happen because of "new" context attachment to VM disk bds. When attaching a new context the corresponding aio context handler is called for each of aio_notifiers registered on the VM disk bds context. Among those handlers, there is the block_job_attached_aio_context handler which sets a new aio context for the block job bds. When doing so, the old context is detached from all the block job bds children and one of them is the VM disk bds, serving as backing store for the blockjob bds, although the VM disk bds is actually the initializer of that process. Since the VM disk bds is protected with walking_aio_notifiers flag from double processing in recursive calls, the assert fires.

Quoted 5 lines
---
 block.c | 4 ++++
 1 file changed, 4 insertions(+)

Naive question - do we have to always call attach_aio_context/detach_aio_context in pairs? If so, will short-circuiting a nested attach cause reference count problems when we come to the matching detach calls?

Signature · 3 lines
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org
#2 lore raw
Quoted 7 lines · Eric Blake
Quoted 4 lines · Denis Plotnikov

Adds a fast path on aio context setting preventing unnecessary context setting routine.
Also, it prevents issues with cyclic walk of child bds-es appeared because of registring aio walking

registering

will fix it

Quoted 15 lines
Quoted 15 lines

notifiers:

This can happen because of "new" context attachment to VM disk bds. When attaching a new context the corresponding aio context handler is called for each of aio_notifiers registered on the VM disk bds context. Among those handlers, there is the block_job_attached_aio_context handler which sets a new aio context for the block job bds. When doing so, the old context is detached from all the block job bds children and one of them is the VM disk bds, serving as backing store for the blockjob bds, although the VM disk bds is actually the initializer of that process. Since the VM disk bds is protected with walking_aio_notifiers flag from double processing in recursive calls, the assert fires.

Quoted 8 lines
Quoted 5 lines
---
  block.c | 4 ++++
  1 file changed, 4 insertions(+)

Naive question - do we have to always call
attach_aio_context/detach_aio_context in pairs?

seems, to be so. the only user doing attach/detach is bdrv_set_aio_context
If so, will

Quoted 2 lines

short-circuiting a nested attach cause reference count problems when we come to the matching detach calls?

I don't quite understand. I suggested the short circuit for bdrv_set_aio_context which actually pairs attach/detach context. So, the pairing here should be a problem

Quoted 1 line
Signature · 2 lines
Best,
Denis