Re: Potential Null dereference

[Kevin Wolf]

Am 24.03.2020 um 08:14 hat Philippe Mathieu-Daudé geschrieben:
> On 3/24/20 4:05 AM, Mansour Ahmadi wrote:
> > Hi,
> > 
> > Nullness of  needs to be checked here:
> > https://github.com/qemu/qemu/blob/c532b954d96f96d361ca31308f75f1b95bd4df76/block/commit.c#L221
> > 
> > pstrcpy(bs->exact_filename, sizeof(bs->exact_filename),...

Do you have a reproducer? It's not obvious to me how bs->backing could
be NULL here.

> > 
> > While it is done at 2 other locations:
> > https://github.com/qemu/qemu/blob/c532b954d96f96d361ca31308f75f1b95bd4df76/block/backup-top.c#L113
> > https://github.com/qemu/qemu/blob/c532b954d96f96d361ca31308f75f1b95bd4df76/block/mirror.c#L1477

Commit 18775ff3269 made the change for mirror, however its commit
message is terse and doesn't say anything about the scenario where it
would happen. We also didn't add a test case for it. I would have
expected that failure to add the backing file would immediately error
out and not try to refresh the filename first.

backup-top.c has the check from the beginning. I assume it just copied
it from mirror.

Vladimir, do you remember the details?

Kevin