Re: [Qemu-devel] [PATCH] block: Remove -errno return value from bdrv_assign_node_name

[Markus Armbruster]

Kevin Wolf <address@hidden> writes:

> Am 18.03.2014 um 11:09 hat Markus Armbruster geschrieben:
>> Kevin Wolf <address@hidden> writes:
>> 
>> > It takes an errp argument. That's enough for error handling.
>> >
>> > Signed-off-by: Kevin Wolf <address@hidden>
>> > ---
>> >  block.c | 21 ++++++++++-----------
>> >  1 file changed, 10 insertions(+), 11 deletions(-)
>> >
>> > diff --git a/block.c b/block.c
>> > index acb70fd..b4f3f77 100644
>> > --- a/block.c
>> > +++ b/block.c
>> > @@ -783,18 +783,18 @@ static int bdrv_open_flags(BlockDriverState *bs, int 
>> > flags)
>> >      return open_flags;
>> >  }
>> >  
>> > -static int bdrv_assign_node_name(BlockDriverState *bs,
>> > -                                 const char *node_name,
>> > -                                 Error **errp)
>> > +static void bdrv_assign_node_name(BlockDriverState *bs,
>> > +                                  const char *node_name,
>> > +                                  Error **errp)
>> >  {
>> >      if (!node_name) {
>> > -        return 0;
>> > +        return;
>> >      }
>> >  
>> >      /* empty string node name is invalid */
>> >      if (node_name[0] == '\0') {
>> >          error_setg(errp, "Empty node name");
>> > -        return -EINVAL;
>> > +        return;
>> >      }
>> >  
>> >      /* takes care of avoiding namespaces collisions */
>> > @@ -807,14 +807,12 @@ static int bdrv_assign_node_name(BlockDriverState 
>> > *bs,
>> >      /* takes care of avoiding duplicates node names */
>> >      if (bdrv_find_node(node_name)) {
>> >          error_setg(errp, "Duplicate node name");
>> > -        return -EINVAL;
>> > +        return;
>> >      }
>> >  
>> >      /* copy node name into the bs and insert it into the graph list */
>> >      pstrcpy(bs->node_name, sizeof(bs->node_name), node_name);
>> >      QTAILQ_INSERT_TAIL(&graph_bdrv_states, bs, node_list);
>> > -
>> > -    return 0;
>> >  }
>> >  
>> >  /*
>> > @@ -849,9 +847,10 @@ static int bdrv_open_common(BlockDriverState *bs, 
>> > BlockDriverState *file,
>> >      trace_bdrv_open_common(bs, filename ?: "", flags, drv->format_name);
>> >  
>> >      node_name = qdict_get_try_str(options, "node-name");
>> > -    ret = bdrv_assign_node_name(bs, node_name, errp);
>> > -    if (ret < 0) {
>> > -        return ret;
>> > +    bdrv_assign_node_name(bs, node_name, &local_err);
>> > +    if (error_is_set(&local_err)) {
>> > +        error_propagate(errp, local_err);
>> > +        return -EINVAL;
>> >      }
>> >      qdict_del(options, "node-name");
>> 
>> Please use 'if (local_err)' instead of 'if (error_is_set(&local_err))'.
>> See commit 84d18f0.
>
> I can do that; this patch predates your change and I just dug it out
> from an old branch.

No harm done :)

>                     Note that there are still many places that use
> error_is_set() where it's not necessary, including documentation that
> should show how to do things right, e.g. docs/writing-qmp-commands.txt.

The documentation I simply missed.  The others probably crept in after I
prepared the patch that became commit 84d18f0.  Such things usually take
an extra iteration or two.

> (By the way, would the actual problem with Coverity in commit 84d18f0
> not be fixed more easily and for all instances by making error_is_set()
> a static inline function in the header?)

I wouldn't bet on it.  Coverity is pretty good at looking through
compilation unit boundaries, but it gets easily spooked by null tests.

> What was the proper use case for error_is_set() again? Or can we get rid
> of it? As long as it's there, you'll keep getting new offenders.

I don't like it myself, and I think we can get rid of it.