Re: KMSAN: uninit-value in linear_transfer (2)

[Takashi Iwai]

On Fri, 09 Nov 2018 13:17:25 +0100,
Alexander Potapenko wrote:
> 
> > It's still strange that the conversion function gets called for the
> > uninitialized source.  But we should clear the vmalloc page in anyway
> > for avoiding such a problem.  And even better would be to use
> > kvzalloc() for a better performance.
> >
> > Could you check whether the patch works?
> The patch works and fixes the problem for me.

OK, I'm going to queue it, then.

> > I forgot the way to trigger the test for kmsan stuff.  IIRC, just
> > pushing to my tree and triggering syz test won't work for KMSAN,
> > right?
> Right, you need to test against KMSAN tree, see
> https://github.com/google/syzkaller/blob/master/docs/syzbot.md#kmsan-bugs

Thanks, that's the note I forgot :)


Takashi