On 27.01.2014 04:10, Benoît Canet wrote:
Le Sunday 26 Jan 2014 à 20:02:40 (+0100), Max Reitz a écrit :
The fail paths of bdrv_file_open() and bdrv_open() naturally exhibit
similarities, thus it is possible to reuse the one from bdrv_open() and
shorten the one in bdrv_file_open() accordingly.
Signed-off-by: Max Reitz <mre...@redhat.com>
---
block.c | 17 +++++++----------
1 file changed, 7 insertions(+), 10 deletions(-)
diff --git a/block.c b/block.c
index 72eddd5..0f2cd3f 100644
--- a/block.c
+++ b/block.c
@@ -1038,9 +1038,6 @@ static int bdrv_file_open(BlockDriverState *bs, const
char *filename,
fail:
QDECREF(options);
- if (!bs->drv) {
- QDECREF(bs->options);
- }
return ret;
}
@@ -1240,17 +1237,17 @@ int bdrv_open(BlockDriverState **pbs, const char *filename,
if (flags & BDRV_O_PROTOCOL) {
assert(!drv);
ret = bdrv_file_open(bs, filename, options, flags & ~BDRV_O_PROTOCOL,
- errp);
+ &local_err);
+ options = NULL;
if (ret) {
- if (*pbs) {
- bdrv_close(bs);
+ if (bs->drv) {
+ goto close_and_fail;
} else {
- bdrv_unref(bs);
+ goto fail;
}
- } else {
- *pbs = bs;
}
Forget what I said about the goto chain in the previous mail.
But something like:
if(!ret) {
*pbs = bs;
return 0;
}
if (bs->drv) {
goto close_and_fail;
}
goto fail;
would keep the code in line.
Best regards
Benoît
Yes, this seems better to me as well. :-)
Or, after patch 9:
if (!ret) {
goto done;
} else if (bs->drv) {
goto close_and_fail;
} else {
goto fail;
}
(which I like a bit more, with the three gotos on the same indentation
level)
Max
+ *pbs = bs;
+ return 0;
}
bs->options = options;
--
1.8.5.3