Re: [PATCH v4 0/6] kasan: add workqueue and timer stack for generic KASAN

[Walter Wu]

On Tue, 2020-12-01 at 15:02 +0100, 'Dmitry Vyukov' via kasan-dev wrote:
> On Tue, Dec 1, 2020 at 12:17 PM Walter Wu <walter-zh.wu@xxxxxxxxxxxx> wrote:
> >
> > Hi Dmitry,
> >
> > On Tue, 2020-12-01 at 08:59 +0100, 'Dmitry Vyukov' via kasan-dev wrote:
> > > On Wed, Sep 30, 2020 at 5:29 PM Thomas Gleixner <tglx@xxxxxxxxxxxxx> wrote:
> > > >
> > > > On Thu, Sep 24 2020 at 12:01, Walter Wu wrote:
> > > > > Syzbot reports many UAF issues for workqueue or timer, see [1] and [2].
> > > > > In some of these access/allocation happened in process_one_work(),
> > > > > we see the free stack is useless in KASAN report, it doesn't help
> > > > > programmers to solve UAF on workqueue. The same may stand for times.
> > > > >
> > > > > This patchset improves KASAN reports by making them to have workqueue
> > > > > queueing stack and timer stack information. It is useful for programmers
> > > > > to solve use-after-free or double-free memory issue.
> > > > >
> > > > > Generic KASAN also records the last two workqueue and timer stacks and
> > > > > prints them in KASAN report. It is only suitable for generic KASAN.
> > >
> > > Walter, did you mail v5?
> > > Checking statuses of KASAN issues and this seems to be not in linux-next.
> > >
> >
> > Sorry for the delay in responding to this patch. I'm busy these few
> > months, so that suspend processing it.
> > Yes, I will send it next week. But v4 need to confirm the timer stack is
> > useful. I haven't found an example. Do you have some suggestion about
> > timer?
> 
> Good question.
> 
> We had some use-after-free's what mention call_timer_fn:
> https://groups.google.com/g/syzkaller-bugs/search?q=%22kasan%22%20%22use-after-free%22%20%22expire_timers%22%20%22call_timer_fn%22%20
> In the reports I checked call_timer_fn appears in the "access" stack
> rather in the "free" stack.
> 
Yes, call stack already is useful for it in KASAN report.

> Looking at these reports I cannot conclude that do_init_timer stack
> would be useful.
> I am mildly leaning towards not memorizing do_init_timer stack for now
> (until we have clear use cases) as the number of aux stacks is very
> limited (2).
> 
Got it. I will remove timer patch and send v5.
Thanks for your suggestion.

Walter