Re: [PATCH 0/4] block: Do not call BlockDriver.bdrv_make_empty() directly

[Eric Blake <eblake@redhat.com>]

On 4/28/20 8:26 AM, Max Reitz wrote:
> Branch: https://github.com/XanClic/qemu.git fix-bdrv_make_empty-v1
> Branch: https://git.xanclic.moe/XanClic/qemu.git fix-bdrv_make_empty-v1
> 
> Hi,
> 
> Right now, there is no centralized bdrv_make_empty() function.  Not only
> is it bad style to call BlockDriver methods directly, it is also wrong,
> unless the caller has a BdrvChild with BLK_PERM_WRITE taken.

I'm also in the middle of writing a patch series that adds a 
corresponding .bdrv_make_empty driver callback.  I'll rebase that work 
on top of this, as part of my efforts at fixing more code to rely on 
bdrv_make_empty rather than directly querying bdrv_has_zero_init[_truncate].

> 
> This series fixes that.
> 
> Note that as far as I’m aware this series shouldn’t visibly fix anything
> at this point; but “block: Introduce real BdrvChildRole”
> (https://lists.nongnu.org/archive/html/qemu-block/2020-02/msg00737.html)
> makes the iotest break when run with -o data_file=$SOMETHING, without
> this series applied beforehand.  (That is because without that series,
> external data files are treated much like metadata children, so the
> format driver always takes the WRITE permission if the file is writable;
> but after that series, it only does so when it itself has a parent
> requestion the WRITE permission.)
> 

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org