Re: [syzbot] general protection fault in gadget_setup

[Alan Stern]

On Tue, Apr 13, 2021 at 07:11:11PM +0200, Dmitry Vyukov wrote:
> On Tue, Apr 13, 2021 at 6:57 PM Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote:
> >
> > On Tue, Apr 13, 2021 at 06:47:47PM +0200, Dmitry Vyukov wrote:
> > > On Tue, Apr 13, 2021 at 6:13 PM Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote:
> > > > Hopefully this patch will make the race a lot more likely to occur.  Is
> > > > there any way to tell syzkaller to test it, despite the fact that
> > > > syzkaller doesn't think it has a reproducer for this issue?
> > >
> > > If there is no reproducer the only way syzbot can test it is if it's
> > > in linux-next under CONFIG_DEBUG_AID_FOR_SYZBOT:
> > > http://bit.do/syzbot#no-custom-patches
> >
> > There _is_ a theoretical reproducer: the test that provoked syzkaller's
> > original bug report.  But syzkaller doesn't realize that it is (or may
> > be) a reproducer.
> >
> > It ought to be possible to ask syzkaller to run a particular test that
> > it has done before, with a patch applied -- and without having to add
> > anything to linux-next.
> 
> Yes, this is possible:
> http://bit.do/syzbot#syzkaller-reproducers

That's not really what I had in mind.  I don't want to spend the time 
and effort installing syskaller on my own system; I want to tell syzbot 
to run a particular syzkaller program (the one that originally led to 
this bug report) on a patched kernel.

The syzbot instructions say that it can test bugs with reproducers.  The 
problem here is that there doesn't seem to be any way to tell it to use 
a particular syzkaller program as a reproducer.

Alan Stern

> The log of tests executed before the crash is available under the
> "console output" link:
> console output: https://syzkaller.appspot.com/x/log.txt?x=124adbf6d00000
> And this log can be replayed using syz-execprog utility.