Patchwork [1/2,block] : Fix scsi-generic breakage in find_image_format()

login
register
mail settings
Submitter Nicholas A. Bellinger
Date May 15, 2010, 1:30 p.m.
Message ID <1273930252-12673-1-git-send-email-nab@linux-iscsi.org>
Download mbox | patch
Permalink /patch/52700/
State New
Headers show

Comments

Nicholas A. Bellinger - May 15, 2010, 1:30 p.m.
From: Nicholas Bellinger <nab@linux-iscsi.org>

This patch adds a special BlockDriverState->sg check in block.c:find_image_format()
after bdrv_file_open() -> block/raw-posix.c:hdev_open() has been called to determine
if we are dealing with a Linux host scsi-generic device or not.

The patch then returns the BlockDriver * from find_protocol(), skipping the subsequent
bdrv_read() and rest of find_image_format().

Signed-off-by: Nicholas A. Bellinger <nab@linux-iscsi.org>
---
 block.c |    7 +++++++
 1 files changed, 7 insertions(+), 0 deletions(-)
Christoph Hellwig - May 16, 2010, 1:29 p.m.
On Sat, May 15, 2010 at 06:30:52AM -0700, Nicholas A. Bellinger wrote:
> From: Nicholas Bellinger <nab@linux-iscsi.org>
> 
> This patch adds a special BlockDriverState->sg check in block.c:find_image_format()
> after bdrv_file_open() -> block/raw-posix.c:hdev_open() has been called to determine
> if we are dealing with a Linux host scsi-generic device or not.
> 
> The patch then returns the BlockDriver * from find_protocol(), skipping the subsequent
> bdrv_read() and rest of find_image_format().

That's not quite correct as we don't want to expose formats directly.
Returning

	bdrv_find_format("raw");

should fix it for now, although I really hate having these special
cases in block.c.
Nicholas A. Bellinger - May 17, 2010, 1:50 a.m.
On Sun, 2010-05-16 at 15:29 +0200, Christoph Hellwig wrote:
> On Sat, May 15, 2010 at 06:30:52AM -0700, Nicholas A. Bellinger wrote:
> > From: Nicholas Bellinger <nab@linux-iscsi.org>
> > 
> > This patch adds a special BlockDriverState->sg check in block.c:find_image_format()
> > after bdrv_file_open() -> block/raw-posix.c:hdev_open() has been called to determine
> > if we are dealing with a Linux host scsi-generic device or not.
> > 
> > The patch then returns the BlockDriver * from find_protocol(), skipping the subsequent
> > bdrv_read() and rest of find_image_format().
> 
> That's not quite correct as we don't want to expose formats directly.
> Returning
> 
> 	bdrv_find_format("raw");
> 
> should fix it for now,

Ahh, that makes more sense..  Attached is updated patch #1:

[PATCH 1/2] [block]: Make find_image_format() return 'raw' BlockDriver for SG_IO devices

>  although I really hate having these special
> cases in block.c.

Indeed having BlockDriverState->sg kinda feels kinda like a hack to
begin with, but it appears this is currently QEMU's way of telling block
code to disable internal SCSI CDB emulation for devices.

Best,

--nab

Patch

diff --git a/block.c b/block.c
index 48305b7..02a22f6 100644
--- a/block.c
+++ b/block.c
@@ -332,6 +332,13 @@  static BlockDriver *find_image_format(const char *filename)
     ret = bdrv_file_open(&bs, filename, 0);
     if (ret < 0)
         return NULL;
+    /*
+     * For scsi-generic we skip the read and return *drv from
+     * find_protocol().
+     */
+    if (bs->sg)
+	return drv;
+
     ret = bdrv_pread(bs, 0, buf, sizeof(buf));
     bdrv_delete(bs);
     if (ret < 0) {