From mboxrd@z Thu Jan  1 00:00:00 1970
Received: from eggs.gnu.org ([208.118.235.92]:43137)
	by lists.gnu.org with esmtp (Exim 4.71)
	(envelope-from <paolo.bonzini@gmail.com>) id 1TYYhg-0003B5-Gy
	for qemu-devel@nongnu.org; Wed, 14 Nov 2012 03:52:15 -0500
Received: from Debian-exim by eggs.gnu.org with spam-scanned (Exim 4.71)
	(envelope-from <paolo.bonzini@gmail.com>) id 1TYYhd-0007yC-EU
	for qemu-devel@nongnu.org; Wed, 14 Nov 2012 03:52:12 -0500
Received: from mail-ee0-f45.google.com ([74.125.83.45]:54718)
	by eggs.gnu.org with esmtp (Exim 4.71)
	(envelope-from <paolo.bonzini@gmail.com>) id 1TYYhd-0007y3-7P
	for qemu-devel@nongnu.org; Wed, 14 Nov 2012 03:52:09 -0500
Received: by mail-ee0-f45.google.com with SMTP id d49so111313eek.4
	for <qemu-devel@nongnu.org>; Wed, 14 Nov 2012 00:52:07 -0800 (PST)
Sender: Paolo Bonzini <paolo.bonzini@gmail.com>
Message-ID: <50A35BA7.7090900@redhat.com>
Date: Wed, 14 Nov 2012 09:51:51 +0100
From: Paolo Bonzini <pbonzini@redhat.com>
MIME-Version: 1.0
References: <1352816095-14051-1-git-send-email-kwolf@redhat.com>
	<1352816095-14051-3-git-send-email-kwolf@redhat.com>
	<20121114083201.GA23826@stefanha-thinkpad.redhat.com>
In-Reply-To: <20121114083201.GA23826@stefanha-thinkpad.redhat.com>
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Subject: Re: [Qemu-devel] [PATCH 2/2] block: Avoid second open for format
	probing
List-Id: <qemu-devel.nongnu.org>
List-Unsubscribe: <https://lists.nongnu.org/mailman/options/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=unsubscribe>
List-Archive: <http://lists.nongnu.org/archive/html/qemu-devel>
List-Post: <mailto:qemu-devel@nongnu.org>
List-Help: <mailto:qemu-devel-request@nongnu.org?subject=help>
List-Subscribe: <https://lists.nongnu.org/mailman/listinfo/qemu-devel>,
	<mailto:qemu-devel-request@nongnu.org?subject=subscribe>
To: Stefan Hajnoczi <stefanha@gmail.com>
Cc: Kevin Wolf <kwolf@redhat.com>, qemu-devel@nongnu.org

Il 14/11/2012 09:32, Stefan Hajnoczi ha scritto:
> On Tue, Nov 13, 2012 at 03:14:55PM +0100, Kevin Wolf wrote:
>> @@ -691,12 +685,15 @@ static int bdrv_open_common(BlockDriverState *bs, const char *filename,
>>  
>>      /* Open the image, either directly or using a protocol */
>>      if (drv->bdrv_file_open) {
>> +        if (file != NULL) {
>> +            bdrv_swap(file, bs);
>> +            bdrv_delete(file);
>> +        }
>>          ret = drv->bdrv_file_open(bs, filename, open_flags);
>>      } else {
> [...]
>>      /* Open the image */
>> -    ret = bdrv_open_common(bs, filename, flags, drv);
>> +    ret = bdrv_open_common(bs, file, filename, flags, drv);
>>      if (ret < 0) {
>>          goto unlink_and_fail;
>>      }
>> @@ -894,6 +895,9 @@ int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
>>      return 0;
>>  
>>  unlink_and_fail:
>> +    if (file != NULL) {
>> +        bdrv_delete(file);
>> +    }
> 
> Not sure I understand this code path.
> 
> We have a protocol (the driver implements .bdrv_file_open()) so we swap
> file and bs, then delete old bs.  Then we call .bdrv_file_open() on the
> already open file BDS.
> 
> Is it okay to call .bdrv_file_open() on an already open BDS?

I don't think so.  But do the cases where this happen make sense?  Can
we just fail if drv is not equal to bs->drv if we reach the "if
(drv->bdrv_file_open)" case?  That would be for cases like "-drive
file=test.img,format=host_device" but how does that make sense?...
(Plus, it fails to stack the raw format on top).

So perhaps we could even assert(drv == bs->drv) if protocols had no
.format_name?

> Now if .bdrv_file_open() fails we will bdrv_delete() the already deleted
> file BDS.  This is a double-free.

Right, always better to NULL out whatever you delete (which means
passing a BDS** to bdrv_open_common.

Paolo