qemu-block
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-block] [PATCH 0/2] block: Let blockdev-create return 0 on succ


From: Kevin Wolf
Subject: Re: [Qemu-block] [PATCH 0/2] block: Let blockdev-create return 0 on success
Date: Mon, 2 Sep 2019 15:32:20 +0200
User-agent: Mutt/1.11.3 (2019-02-01)

Am 23.08.2019 um 20:47 hat Max Reitz geschrieben:
> Jobs are expected to return 0 on success.  .bdrv_co_create() on the
> other hand is a block layer function, and as such returns a
> non-negative value on success.

I don't agree that >= 0 is the block layer way. The block layer uses
0/-errno for the largest part of its interfaces; and I think the
BlockDriver callbacks might even be consistent in this. Of course, we
never documented this anywhere, maybe we should...

The only historical exceptions I'm aware of are blk/bdrv_pread/pwrite(),
which return the byte count instead of 0. They should be fixed
eventually, but it just never seemed important enough, even though it
did cause bugs every now and then.

> blockdev_create_run() should translate between the two (patch 1).
> 
> Without patch 1, blockdev-create is likely to fail for VPC images.
> Hence patch 2.

I'd argue this is a VPC bug. In the success path, it shouldn't return
ret as it happens to be at the end (it comes from bdrv_pwrite()), but
set it to 0 right before the 'fail:' label.

This is really a regression Jeff introduced in commit fef6070eff2,
though the bug was only latent then (five years ago).

Kevin



reply via email to

[Prev in Thread] Current Thread [Next in Thread]