qemu-devel.nongnu.org archive mirror
 help / color / mirror / Atom feed
* [Qemu-devel] [PATCH 1/2] block: flush backing_hd in the right place
@ 2010-01-12 12:49 Christoph Hellwig
  2010-01-12 17:42 ` Kevin Wolf
                   ` (2 more replies)
  0 siblings, 3 replies; 9+ messages in thread
From: Christoph Hellwig @ 2010-01-12 12:49 UTC (permalink / raw)
  To: qemu-devel


The backing device is only modified from bdrv_commit.  So instead of
flushing it every time bdrv_flush is called for the front-end device
only flush it after we're written data to it in bdrv_commit.

Signed-off-by: Christoph Hellwig <hch@lst.de>

Index: qemu/block.c
===================================================================
--- qemu.orig/block.c	2010-01-12 11:34:35.549024986 +0100
+++ qemu/block.c	2010-01-12 11:43:28.965006129 +0100
@@ -623,6 +623,12 @@ int bdrv_commit(BlockDriverState *bs)
     if (drv->bdrv_make_empty)
 	return drv->bdrv_make_empty(bs);
 
+    /*
+     * Make sure all data we wrote to the backing device is actually
+     * stable on disk.
+     */
+    if (bs->backing_hd)
+        bdrv_flush(bs->backing_hd);
     return 0;
 }
 
@@ -1124,12 +1130,8 @@ const char *bdrv_get_device_name(BlockDr
 
 void bdrv_flush(BlockDriverState *bs)
 {
-    if (!bs->drv)
-        return;
-    if (bs->drv->bdrv_flush)
+    if (bs->drv && bs->drv->bdrv_flush)
         bs->drv->bdrv_flush(bs);
-    if (bs->backing_hd)
-        bdrv_flush(bs->backing_hd);
 }
 
 void bdrv_flush_all(void)
@@ -1806,11 +1808,6 @@ BlockDriverAIOCB *bdrv_aio_flush(BlockDr
 
     if (!drv)
         return NULL;
-
-    /*
-     * Note that unlike bdrv_flush the driver is reponsible for flushing a
-     * backing image if it exists.
-     */
     return drv->bdrv_aio_flush(bs, cb, opaque);
 }
 

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [Qemu-devel] [PATCH 1/2] block: flush backing_hd in the right place
  2010-01-12 12:49 [Qemu-devel] [PATCH 1/2] block: flush backing_hd in the right place Christoph Hellwig
@ 2010-01-12 17:42 ` Kevin Wolf
  2010-01-12 17:44   ` Christoph Hellwig
  2010-01-12 18:13 ` [Qemu-devel] [PATCH 1/2 v2] " Christoph Hellwig
  2010-01-13 23:26 ` [Qemu-devel] [PATCH 1/2] " Anthony Liguori
  2 siblings, 1 reply; 9+ messages in thread
From: Kevin Wolf @ 2010-01-12 17:42 UTC (permalink / raw)
  To: Christoph Hellwig; +Cc: qemu-devel

Am 12.01.2010 13:49, schrieb Christoph Hellwig:
> 
> The backing device is only modified from bdrv_commit.  So instead of
> flushing it every time bdrv_flush is called for the front-end device
> only flush it after we're written data to it in bdrv_commit.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> 
> Index: qemu/block.c
> ===================================================================
> --- qemu.orig/block.c	2010-01-12 11:34:35.549024986 +0100
> +++ qemu/block.c	2010-01-12 11:43:28.965006129 +0100
> @@ -623,6 +623,12 @@ int bdrv_commit(BlockDriverState *bs)
>      if (drv->bdrv_make_empty)
>  	return drv->bdrv_make_empty(bs);
>  
> +    /*
> +     * Make sure all data we wrote to the backing device is actually
> +     * stable on disk.
> +     */
> +    if (bs->backing_hd)
> +        bdrv_flush(bs->backing_hd);
>      return 0;
>  }

Format drivers with a bdrv_make_empty return before the flush, so it
won't work for qcow1. Looks good otherwise.

If it has done a bdrv_make_empty we might also want to flush bs?

Kevin

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [Qemu-devel] [PATCH 1/2] block: flush backing_hd in the right place
  2010-01-12 17:42 ` Kevin Wolf
@ 2010-01-12 17:44   ` Christoph Hellwig
  0 siblings, 0 replies; 9+ messages in thread
From: Christoph Hellwig @ 2010-01-12 17:44 UTC (permalink / raw)
  To: Kevin Wolf; +Cc: Christoph Hellwig, qemu-devel

On Tue, Jan 12, 2010 at 06:42:20PM +0100, Kevin Wolf wrote:
> > @@ -623,6 +623,12 @@ int bdrv_commit(BlockDriverState *bs)
> >      if (drv->bdrv_make_empty)
> >  	return drv->bdrv_make_empty(bs);
> >  
> > +    /*
> > +     * Make sure all data we wrote to the backing device is actually
> > +     * stable on disk.
> > +     */
> > +    if (bs->backing_hd)
> > +        bdrv_flush(bs->backing_hd);
> >      return 0;
> >  }
> 
> Format drivers with a bdrv_make_empty return before the flush, so it
> won't work for qcow1. Looks good otherwise.

Oh, okay.

> If it has done a bdrv_make_empty we might also want to flush bs?

Indeed.  I'll spin a new version.

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [Qemu-devel] [PATCH 1/2 v2] block: flush backing_hd in the right place
  2010-01-12 12:49 [Qemu-devel] [PATCH 1/2] block: flush backing_hd in the right place Christoph Hellwig
  2010-01-12 17:42 ` Kevin Wolf
@ 2010-01-12 18:13 ` Christoph Hellwig
  2010-01-13  9:50   ` Kevin Wolf
  2010-01-13 23:26 ` [Qemu-devel] [PATCH 1/2] " Anthony Liguori
  2 siblings, 1 reply; 9+ messages in thread
From: Christoph Hellwig @ 2010-01-12 18:13 UTC (permalink / raw)
  To: qemu-devel

The backing device is only modified from bdrv_commit.  So instead of
flushing it every time bdrv_flush is called for the front-end device
only flush it after we're written data to it in bdrv_commit.

Also flush the frontend image if we have a make_empty method that
possibly writes to it.

Signed-off-by: Christoph Hellwig <hch@lst.de>

Index: qemu/block.c
===================================================================
--- qemu.orig/block.c	2010-01-12 19:08:07.363003775 +0100
+++ qemu/block.c	2010-01-12 19:09:10.836255948 +0100
@@ -589,6 +589,7 @@ int bdrv_commit(BlockDriverState *bs)
     BlockDriver *drv = bs->drv;
     int64_t i, total_sectors;
     int n, j;
+    int ret = 0;
     unsigned char sector[512];
 
     if (!drv)
@@ -620,10 +621,18 @@ int bdrv_commit(BlockDriverState *bs)
         }
     }
 
-    if (drv->bdrv_make_empty)
-	return drv->bdrv_make_empty(bs);
+    if (drv->bdrv_make_empty) {
+	ret = drv->bdrv_make_empty(bs);
+        bdrv_flush(bs);
+    }
 
-    return 0;
+    /*
+     * Make sure all data we wrote to the backing device is actually
+     * stable on disk.
+     */
+    if (bs->backing_hd)
+        bdrv_flush(bs->backing_hd);
+    return ret;
 }
 
 static int bdrv_check_byte_request(BlockDriverState *bs, int64_t offset,
@@ -1124,12 +1133,8 @@ const char *bdrv_get_device_name(BlockDr
 
 void bdrv_flush(BlockDriverState *bs)
 {
-    if (!bs->drv)
-        return;
-    if (bs->drv->bdrv_flush)
+    if (bs->drv && bs->drv->bdrv_flush)
         bs->drv->bdrv_flush(bs);
-    if (bs->backing_hd)
-        bdrv_flush(bs->backing_hd);
 }
 
 void bdrv_flush_all(void)
@@ -1806,11 +1811,6 @@ BlockDriverAIOCB *bdrv_aio_flush(BlockDr
 
     if (!drv)
         return NULL;
-
-    /*
-     * Note that unlike bdrv_flush the driver is reponsible for flushing a
-     * backing image if it exists.
-     */
     return drv->bdrv_aio_flush(bs, cb, opaque);
 }
 

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [Qemu-devel] [PATCH 1/2 v2] block: flush backing_hd in the right place
  2010-01-12 18:13 ` [Qemu-devel] [PATCH 1/2 v2] " Christoph Hellwig
@ 2010-01-13  9:50   ` Kevin Wolf
  0 siblings, 0 replies; 9+ messages in thread
From: Kevin Wolf @ 2010-01-13  9:50 UTC (permalink / raw)
  To: Christoph Hellwig; +Cc: qemu-devel

Am 12.01.2010 19:13, schrieb Christoph Hellwig:
> The backing device is only modified from bdrv_commit.  So instead of
> flushing it every time bdrv_flush is called for the front-end device
> only flush it after we're written data to it in bdrv_commit.
> 
> Also flush the frontend image if we have a make_empty method that
> possibly writes to it.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> 
> Index: qemu/block.c
> ===================================================================
> --- qemu.orig/block.c	2010-01-12 19:08:07.363003775 +0100
> +++ qemu/block.c	2010-01-12 19:09:10.836255948 +0100
> @@ -589,6 +589,7 @@ int bdrv_commit(BlockDriverState *bs)
>      BlockDriver *drv = bs->drv;
>      int64_t i, total_sectors;
>      int n, j;
> +    int ret = 0;
>      unsigned char sector[512];
>  
>      if (!drv)
> @@ -620,10 +621,18 @@ int bdrv_commit(BlockDriverState *bs)
>          }
>      }
>  
> -    if (drv->bdrv_make_empty)
> -	return drv->bdrv_make_empty(bs);
> +    if (drv->bdrv_make_empty) {
> +	ret = drv->bdrv_make_empty(bs);
> +        bdrv_flush(bs);
> +    }

Indentation is off here (but it already was before the patch). The logic
looks good to me now.

Kevin

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [Qemu-devel] [PATCH 1/2] block: flush backing_hd in the right place
  2010-01-12 12:49 [Qemu-devel] [PATCH 1/2] block: flush backing_hd in the right place Christoph Hellwig
  2010-01-12 17:42 ` Kevin Wolf
  2010-01-12 18:13 ` [Qemu-devel] [PATCH 1/2 v2] " Christoph Hellwig
@ 2010-01-13 23:26 ` Anthony Liguori
  2010-01-14 10:21   ` Kevin Wolf
  2 siblings, 1 reply; 9+ messages in thread
From: Anthony Liguori @ 2010-01-13 23:26 UTC (permalink / raw)
  To: Christoph Hellwig; +Cc: qemu-devel

On 01/12/2010 06:49 AM, Christoph Hellwig wrote:
> The backing device is only modified from bdrv_commit.  So instead of
> flushing it every time bdrv_flush is called for the front-end device
> only flush it after we're written data to it in bdrv_commit.
>
> Signed-off-by: Christoph Hellwig<hch@lst.de>
>    

Applied.  Thanks.

Regards,

Anthony Liguori

> Index: qemu/block.c
> ===================================================================
> --- qemu.orig/block.c	2010-01-12 11:34:35.549024986 +0100
> +++ qemu/block.c	2010-01-12 11:43:28.965006129 +0100
> @@ -623,6 +623,12 @@ int bdrv_commit(BlockDriverState *bs)
>       if (drv->bdrv_make_empty)
>   	return drv->bdrv_make_empty(bs);
>
> +    /*
> +     * Make sure all data we wrote to the backing device is actually
> +     * stable on disk.
> +     */
> +    if (bs->backing_hd)
> +        bdrv_flush(bs->backing_hd);
>       return 0;
>   }
>
> @@ -1124,12 +1130,8 @@ const char *bdrv_get_device_name(BlockDr
>
>   void bdrv_flush(BlockDriverState *bs)
>   {
> -    if (!bs->drv)
> -        return;
> -    if (bs->drv->bdrv_flush)
> +    if (bs->drv&&  bs->drv->bdrv_flush)
>           bs->drv->bdrv_flush(bs);
> -    if (bs->backing_hd)
> -        bdrv_flush(bs->backing_hd);
>   }
>
>   void bdrv_flush_all(void)
> @@ -1806,11 +1808,6 @@ BlockDriverAIOCB *bdrv_aio_flush(BlockDr
>
>       if (!drv)
>           return NULL;
> -
> -    /*
> -     * Note that unlike bdrv_flush the driver is reponsible for flushing a
> -     * backing image if it exists.
> -     */
>       return drv->bdrv_aio_flush(bs, cb, opaque);
>   }
>
>
>
>
>    

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [Qemu-devel] [PATCH 1/2] block: flush backing_hd in the right place
  2010-01-13 23:26 ` [Qemu-devel] [PATCH 1/2] " Anthony Liguori
@ 2010-01-14 10:21   ` Kevin Wolf
  2010-01-17 11:32     ` [Qemu-devel] [PATCH] block: fix cache flushing in bdrv_commit Christoph Hellwig
  0 siblings, 1 reply; 9+ messages in thread
From: Kevin Wolf @ 2010-01-14 10:21 UTC (permalink / raw)
  To: Anthony Liguori; +Cc: Christoph Hellwig, qemu-devel

Am 14.01.2010 00:26, schrieb Anthony Liguori:
> On 01/12/2010 06:49 AM, Christoph Hellwig wrote:
>> The backing device is only modified from bdrv_commit.  So instead of
>> flushing it every time bdrv_flush is called for the front-end device
>> only flush it after we're written data to it in bdrv_commit.
>>
>> Signed-off-by: Christoph Hellwig<hch@lst.de>
>>    
> 
> Applied.  Thanks.

Anthony, you seem to have missed the v2 patch that considered my review
comments. Can you please apply the diff between v1 and v2 on top?

Kevin

^ permalink raw reply	[flat|nested] 9+ messages in thread

* [Qemu-devel] [PATCH] block: fix cache flushing in bdrv_commit
  2010-01-14 10:21   ` Kevin Wolf
@ 2010-01-17 11:32     ` Christoph Hellwig
  2010-01-20 14:58       ` Anthony Liguori
  0 siblings, 1 reply; 9+ messages in thread
From: Christoph Hellwig @ 2010-01-17 11:32 UTC (permalink / raw)
  To: Kevin Wolf; +Cc: Christoph Hellwig, qemu-devel

On Thu, Jan 14, 2010 at 11:21:15AM +0100, Kevin Wolf wrote:
> Anthony, you seem to have missed the v2 patch that considered my review
> comments. Can you please apply the diff between v1 and v2 on top?

Here is the differences in patch form:

---
From: Christoph Hellwig <hch@lst.de>
Subject: block: fix cache flushing in bdrv_commit

As pointed out by Kevin Wolf the previous patch returned early if we
there is a bdrv_make_empty method and it missed a cache flush for the
frontend device after the bdev_make_empty call.  This patch fixes it
up.

Signed-off-by: Christoph Hellwig <hch@lst.de>


Index: qemu/block.c
===================================================================
--- qemu.orig/block.c	2010-01-17 12:27:03.589006970 +0100
+++ qemu/block.c	2010-01-17 12:27:55.718008519 +0100
@@ -589,6 +589,7 @@ int bdrv_commit(BlockDriverState *bs)
     BlockDriver *drv = bs->drv;
     int64_t i, total_sectors;
     int n, j;
+    int ret = 0;
     unsigned char sector[512];
 
     if (!drv)
@@ -620,8 +621,10 @@ int bdrv_commit(BlockDriverState *bs)
         }
     }
 
-    if (drv->bdrv_make_empty)
-	return drv->bdrv_make_empty(bs);
+    if (drv->bdrv_make_empty) {
+        ret = drv->bdrv_make_empty(bs);
+        bdrv_flush(bs);
+    }
 
     /*
      * Make sure all data we wrote to the backing device is actually
@@ -629,7 +632,7 @@ int bdrv_commit(BlockDriverState *bs)
      */
     if (bs->backing_hd)
         bdrv_flush(bs->backing_hd);
-    return 0;
+    return ret;
 }
 
 /*

^ permalink raw reply	[flat|nested] 9+ messages in thread

* Re: [Qemu-devel] [PATCH] block: fix cache flushing in bdrv_commit
  2010-01-17 11:32     ` [Qemu-devel] [PATCH] block: fix cache flushing in bdrv_commit Christoph Hellwig
@ 2010-01-20 14:58       ` Anthony Liguori
  0 siblings, 0 replies; 9+ messages in thread
From: Anthony Liguori @ 2010-01-20 14:58 UTC (permalink / raw)
  To: Christoph Hellwig; +Cc: Kevin Wolf, qemu-devel

On 01/17/2010 05:32 AM, Christoph Hellwig wrote:
> On Thu, Jan 14, 2010 at 11:21:15AM +0100, Kevin Wolf wrote:
>    
>> Anthony, you seem to have missed the v2 patch that considered my review
>> comments. Can you please apply the diff between v1 and v2 on top?
>>      
> Here is the differences in patch form:
>
> ---
> From: Christoph Hellwig<hch@lst.de>
> Subject: block: fix cache flushing in bdrv_commit
>
> As pointed out by Kevin Wolf the previous patch returned early if we
> there is a bdrv_make_empty method and it missed a cache flush for the
> frontend device after the bdev_make_empty call.  This patch fixes it
> up.
>
> Signed-off-by: Christoph Hellwig<hch@lst.de>
>    

Applied.  Thanks.

Regards,

Anthony Liguori
> Index: qemu/block.c
> ===================================================================
> --- qemu.orig/block.c	2010-01-17 12:27:03.589006970 +0100
> +++ qemu/block.c	2010-01-17 12:27:55.718008519 +0100
> @@ -589,6 +589,7 @@ int bdrv_commit(BlockDriverState *bs)
>       BlockDriver *drv = bs->drv;
>       int64_t i, total_sectors;
>       int n, j;
> +    int ret = 0;
>       unsigned char sector[512];
>
>       if (!drv)
> @@ -620,8 +621,10 @@ int bdrv_commit(BlockDriverState *bs)
>           }
>       }
>
> -    if (drv->bdrv_make_empty)
> -	return drv->bdrv_make_empty(bs);
> +    if (drv->bdrv_make_empty) {
> +        ret = drv->bdrv_make_empty(bs);
> +        bdrv_flush(bs);
> +    }
>
>       /*
>        * Make sure all data we wrote to the backing device is actually
> @@ -629,7 +632,7 @@ int bdrv_commit(BlockDriverState *bs)
>        */
>       if (bs->backing_hd)
>           bdrv_flush(bs->backing_hd);
> -    return 0;
> +    return ret;
>   }
>
>   /*
>
>
>
>    

^ permalink raw reply	[flat|nested] 9+ messages in thread

end of thread, other threads:[~2010-01-20 14:59 UTC | newest]

Thread overview: 9+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-01-12 12:49 [Qemu-devel] [PATCH 1/2] block: flush backing_hd in the right place Christoph Hellwig
2010-01-12 17:42 ` Kevin Wolf
2010-01-12 17:44   ` Christoph Hellwig
2010-01-12 18:13 ` [Qemu-devel] [PATCH 1/2 v2] " Christoph Hellwig
2010-01-13  9:50   ` Kevin Wolf
2010-01-13 23:26 ` [Qemu-devel] [PATCH 1/2] " Anthony Liguori
2010-01-14 10:21   ` Kevin Wolf
2010-01-17 11:32     ` [Qemu-devel] [PATCH] block: fix cache flushing in bdrv_commit Christoph Hellwig
2010-01-20 14:58       ` Anthony Liguori

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).