Message-ID: <201911270952.D66CD15AEC@keescook>
Date:   Wed, 27 Nov 2019 09:59:18 -0800
From:   Kees Cook <keescook@...omium.org>
To:     Dmitry Vyukov <dvyukov@...gle.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        Andrey Ryabinin <aryabinin@...tuozzo.com>,
        Elena Petrova <lenaptr@...gle.com>,
        Alexander Potapenko <glider@...gle.com>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Dan Carpenter <dan.carpenter@...cle.com>,
        "Gustavo A. R. Silva" <gustavo@...eddedor.com>,
        Arnd Bergmann <arnd@...db.de>,
        Ard Biesheuvel <ard.biesheuvel@...aro.org>,
        kasan-dev <kasan-dev@...glegroups.com>,
        LKML <linux-kernel@...r.kernel.org>,
        kernel-hardening@...ts.openwall.com,
        syzkaller <syzkaller@...glegroups.com>
Subject: Re: [PATCH v2 0/3] ubsan: Split out bounds checker

On Wed, Nov 27, 2019 at 10:34:24AM +0100, Dmitry Vyukov wrote:
> On Wed, Nov 27, 2019 at 7:54 AM Dmitry Vyukov <dvyukov@...gle.com> wrote:
> >
> > On Wed, Nov 27, 2019 at 6:42 AM Kees Cook <keescook@...omium.org> wrote:
> > >
> > > On Fri, Nov 22, 2019 at 10:07:29AM +0100, Dmitry Vyukov wrote:
> > > > On Thu, Nov 21, 2019 at 7:15 PM Kees Cook <keescook@...omium.org> wrote:
> > > > >
> > > > > v2:
> > > > >     - clarify Kconfig help text (aryabinin)
> > > > >     - add reviewed-by
> > > > >     - aim series at akpm, which seems to be where ubsan goes through?
> > > > > v1: https://lore.kernel.org/lkml/20191120010636.27368-1-keescook@chromium.org
> > > > >
> > > > > This splits out the bounds checker so it can be individually used. This
> > > > > is expected to be enabled in Android and hopefully for syzbot. Includes
> > > > > LKDTM tests for behavioral corner-cases (beyond just the bounds checker).
> > > > >
> > > > > -Kees
> > > >
> > > > +syzkaller mailing list
> > > >
> > > > This is great!
> > >
> > > BTW, can I consider this your Acked-by for these patches? :)
> > >
> > > > I wanted to enable UBSAN on syzbot for a long time. And it's
> > > > _probably_ not lots of work. But it was stuck on somebody actually
> > > > dedicating some time specifically for it.
> > >
> > > Do you have a general mechanism to test that syzkaller will actually
> > > pick up the kernel log splat of a new check?
> >
> > Yes. That's one of the most important and critical parts of syzkaller :)
> > The tests for different types of bugs are here:
> > https://github.com/google/syzkaller/tree/master/pkg/report/testdata/linux/report
> >
> > But have 3 for UBSAN, but they may be old and it would be useful to
> > have 1 example crash per bug type:
> >
> > syzkaller$ grep UBSAN pkg/report/testdata/linux/report/*
> > pkg/report/testdata/linux/report/40:TITLE: UBSAN: Undefined behaviour
> > in drivers/usb/core/devio.c:LINE
> > pkg/report/testdata/linux/report/40:[    4.556972] UBSAN: Undefined
> > behaviour in drivers/usb/core/devio.c:1517:25
> > pkg/report/testdata/linux/report/41:TITLE: UBSAN: Undefined behaviour
> > in ./arch/x86/include/asm/atomic.h:LINE
> > pkg/report/testdata/linux/report/41:[    3.805453] UBSAN: Undefined
> > behaviour in ./arch/x86/include/asm/atomic.h:156:2
> > pkg/report/testdata/linux/report/42:TITLE: UBSAN: Undefined behaviour
> > in kernel/time/hrtimer.c:LINE
> > pkg/report/testdata/linux/report/42:[   50.583499] UBSAN: Undefined
> > behaviour in kernel/time/hrtimer.c:310:16
> >
> > One of them is incomplete and is parsed as "corrupted kernel output"
> > (won't be reported):
> > https://github.com/google/syzkaller/blob/master/pkg/report/testdata/linux/report/42
> >
> > Also I see that report parsing just takes the first line, which
> > includes file name, which is suboptimal (too long, can't report 2 bugs
> > in the same file). We seem to converge on "bug-type in function-name"
> > format.
> > The thing about bug titles is that it's harder to change them later.
> > If syzbot already reported 100 bugs and we change titles, it will
> > start re-reporting the old one after new names and the old ones will
> > look stale, yet they still relevant, just detected under different
> > name.
> > So we also need to get this part right before enabling.

It Sounds like instead of "UBSAN: Undefined behaviour in $file", UBSAN
should report something like "UBSAN: $behavior in $file"?

e.g.
40: UBSAN: bad shift in drivers/usb/core/devio.c:1517:25"
41: UBSAN: signed integer overflow in ./arch/x86/include/asm/atomic.h:156:2

I'll add one for the bounds checker.

How are these reports used? (And is there a way to check a live kernel
crash? i.e. to tell syzkaller "echo ARRAY_BOUNDS >/.../lkdtm..." and
generate a report?

> > > I noticed a few things
> > > about the ubsan handlers: they don't use any of the common "warn"
> > > infrastructure (neither does kasan from what I can see), and was missing
> > > a check for panic_on_warn (kasan has this, but does it incorrectly).
> >
> > Yes, panic_on_warn we also need.
> >
> > I will look at the patches again for Acked-by.
> 
> 
> Acked-by: Dmitry Vyukov <dvyukov@...gle.com>
> for the series.

Thanks!

> 
> I see you extended the test module, do you have samples of all UBSAN
> report types that are triggered by these functions? Is so, please add
> them to:
> https://github.com/google/syzkaller/tree/master/pkg/report/testdata/linux/report

Okay, cool.

> with whatever titles they are detected now. Improving titles will then
> be the next step, but much simpler with a good collection of tests.
> 
> Will you send the panic_on_want patch as well?

Yes; I wanted to make sure it was needed first (which you've confirmed
now). I'll likely not send it until next week.

-- 
Kees Cook

Powered by blists - more mailing lists