[Qemu-devel] [PATCH] Call qemu_bh_delete at bdrv_aio_bh_cb

[Qemu-devel] [PATCH] Call qemu_bh_delete at bdrv_aio_bh_cb

[Dor Laor]

[-- Attachment #1: Type: text/plain, Size: 1 bytes --]



[-- Attachment #2: 0001-Call-qemu_bh_delete-at-bdrv_aio_bh_cb.patch --]
[-- Type: text/plain, Size: 1135 bytes --]

>From 09ad82d8b8ebd1670aadf21eae8021dfad373c61 Mon Sep 17 00:00:00 2001
From: Dor Laor <dor@redhat.com>
Date: Mon, 1 Jun 2009 12:07:23 +0300
Subject: [PATCH] Call qemu_bh_delete at bdrv_aio_bh_cb.
 Also replave qemu_bh_cancel with qemu_bh_delete in bdrv_aio_cancel_em.
 Otherwise the bh will live forever in the bh list.

Signed-off-by: Dor Laor <dor@redhat.com>
---
 block.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/block.c b/block.c
index c80d4b9..4a0b642 100644
--- a/block.c
+++ b/block.c
@@ -1402,7 +1402,7 @@ static void bdrv_aio_bh_cb(void *opaque)
         qemu_iovec_from_buffer(acb->qiov, acb->bounce, acb->qiov->size);
     qemu_vfree(acb->bounce);
     acb->common.cb(acb->common.opaque, acb->ret);
-
+    qemu_bh_delete(acb->bh);
     qemu_aio_release(acb);
 }
 
@@ -1455,7 +1455,7 @@ static BlockDriverAIOCB *bdrv_aio_writev_em(BlockDriverState *bs,
 static void bdrv_aio_cancel_em(BlockDriverAIOCB *blockacb)
 {
     BlockDriverAIOCBSync *acb = (BlockDriverAIOCBSync *)blockacb;
-    qemu_bh_cancel(acb->bh);
+    qemu_bh_delete(acb->bh);
     qemu_aio_release(acb);
 }
 
-- 
1.5.6.6

Re: [Qemu-devel] [PATCH] Call qemu_bh_delete at bdrv_aio_bh_cb

[Christoph Hellwig]

On Mon, Jun 01, 2009 at 12:10:02PM +0300, Dor Laor wrote:
> 

> >From 09ad82d8b8ebd1670aadf21eae8021dfad373c61 Mon Sep 17 00:00:00 2001
> From: Dor Laor <dor@redhat.com>
> Date: Mon, 1 Jun 2009 12:07:23 +0300
> Subject: [PATCH] Call qemu_bh_delete at bdrv_aio_bh_cb.
>  Also replave qemu_bh_cancel with qemu_bh_delete in bdrv_aio_cancel_em.
>  Otherwise the bh will live forever in the bh list.

Looks good to me.  Those lifetime rules for QEMUBH objects look
quite awfull to me, though - but this whole don't directly free objects
but rather wait for $event (or forever) scheme seems pretty common in
qemu.

Re: [Qemu-devel] [PATCH] Call qemu_bh_delete at bdrv_aio_bh_cb

[Avi Kivity]

On 06/01/2009 12:10 PM, Dor Laor wrote:
> Subject: [PATCH] Call qemu_bh_delete at bdrv_aio_bh_cb.
>   Also replave qemu_bh_cancel with qemu_bh_delete in bdrv_aio_cancel_em.
>   Otherwise the bh will live forever in the bh list.
>    

This breaks vmdk.  Qemu hangs while starting up.  Trying to see why.

-- 
error compiling committee.c: too many arguments to function