Re: [PATCH v3 02/13] block: use return status of bdrv_append()

[Kevin Wolf]

Am 19.10.2020 um 17:45 hat Markus Armbruster geschrieben:
> Kevin Wolf <kwolf@redhat.com> writes:
> 
> > Am 16.10.2020 um 19:10 hat Vladimir Sementsov-Ogievskiy geschrieben:
> >> Now bdrv_append returns status and we can drop all the local_err things
> >> around it.
> >> 
> >> Signed-off-by: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
> >> Reviewed-by: Greg Kurz <groug@kaod.org>
> >> Reviewed-by: Alberto Garcia <berto@igalia.com>
> >> ---
> >>  block.c                     |  5 +----
> >>  block/backup-top.c          | 20 ++++++++------------
> >>  block/commit.c              |  5 +----
> >>  block/mirror.c              |  6 ++----
> >>  blockdev.c                  |  4 +---
> >>  tests/test-bdrv-graph-mod.c |  6 +++---
> >>  6 files changed, 16 insertions(+), 30 deletions(-)
> >> 
> >> diff --git a/block.c b/block.c
> >> index b05fbff42d..7b6818c681 100644
> >> --- a/block.c
> >> +++ b/block.c
> >> @@ -3161,7 +3161,6 @@ static BlockDriverState 
> >> *bdrv_append_temp_snapshot(BlockDriverState *bs,
> >>      int64_t total_size;
> >>      QemuOpts *opts = NULL;
> >>      BlockDriverState *bs_snapshot = NULL;
> >> -    Error *local_err = NULL;
> >>      int ret;
> >>  
> >>      /* if snapshot, we create a temporary backing file and open it
> >> @@ -3208,9 +3207,7 @@ static BlockDriverState 
> >> *bdrv_append_temp_snapshot(BlockDriverState *bs,
> >>       * order to be able to return one, we have to increase
> >>       * bs_snapshot's refcount here */
> >>      bdrv_ref(bs_snapshot);
> >> -    bdrv_append(bs_snapshot, bs, &local_err);
> >> -    if (local_err) {
> >> -        error_propagate(errp, local_err);
> >> +    if (bdrv_append(bs_snapshot, bs, errp) < 0) {
> >
> > We generally avoid calling functions with side effects inside a
> > comparison. Let's use the usual pattern:
> >
> >     ret = bdrv_append(bs_snapshot, bs, errp);
> >     if (ret < 0) {
> >         ...
> >     }
> 
> I'd also advice against buring side effects too deep, but calling a
> function in a failure-checking conditional is pretty benign.  It's also
> common:
> 
>     $ git-grep 'if ([a-z].*) < 0) {'
> 
> coughs up several hundred instances.
> 
> That said, there is none in block.c.  Local consistency matters.

Actually, after looking at the rest of the series, I also need to be a
bit more specific: We do have boolean if (!foo()), it's just the
negative errno ones that aren't very common in the block layer.

And the reason why it's not very common is probably that 0/-errno is
very common in the block layer and a check for < 0 just throws the
specific errno away. We had a lot of such code, which was wrong because
it lost information instead of passing the real error to the caller.
I guess this is the real origin of the block layer habits to avoid it.

Kevin