linux-kernel - Re: [PATCH 3/3] kcov: remove useless barrier()s

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <62a9c122-1527-3466-bbb9-4e2035a160e4@virtuozzo.com>
Date:   Tue, 19 Sep 2017 17:29:40 +0300
From:   Andrey Ryabinin <aryabinin@...tuozzo.com>
To:     Dmitry Vyukov <dvyukov@...gle.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>,
        Andrey Konovalov <andreyknvl@...gle.com>,
        Victor Chibotaru <tchibo@...gle.com>,
        syzkaller <syzkaller@...glegroups.com>,
        Mark Rutland <mark.rutland@....com>,
        LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH 3/3] kcov: remove useless barrier()s

On 09/19/2017 04:54 PM, Dmitry Vyukov wrote:
> On Tue, Sep 19, 2017 at 3:52 PM, Andrey Ryabinin
> <aryabinin@...tuozzo.com> wrote:
>>
>>
>> On 09/19/2017 03:57 PM, Dmitry Vyukov wrote:
>>> On Tue, Sep 19, 2017 at 2:46 PM, Andrey Ryabinin
>>> <aryabinin@...tuozzo.com> wrote:
>>>> As comment says barriers needed for preempt_schedule_irq() case
>>>> where in_interrupt() returns false. But we don't use in_interrupt()
>>>> since b274c0bb394c ("kcov: properly check if we are in an interrupt").
>>>>
>>>> Now we use in_task() which handles preempt_schedule_irq() case properly,
>>>> thus no barrier required.
>>>
>>>
>>> Are you sure in_task() handles preempt_schedule_irq() correctly?
>>> They seem to differ only by SOFTIRQ_MASK vs SOFTIRQ_OFFSET, and that
>>> only differs in local_bh_disable sections. But preempt_schedule_irq()
>>> does not seem to have anything to do softirq/local_bh_disable. It's
>>> called from real interrupts, right? So I would expect that in_task()
>>> returns true in preempt_schedule_irq().
>>
>> Indeed, you're right. I checked this only on !PREEMPT kernel, where this worked.
>>
>> Still, I think that barrier() in __sanitizer_cov_trace_pc() is not needed. AFAIU it needed
>> to make sure that load of t->kcov_area isn't moved before load of t->kcov_mode, but I don't
>> think that compiler is allowed to make such reorder. That would be a bug in the compiler.
> 

Ugh, it should have saied READ_ONCE(area[0]) instead of t->kcov_area.

> 
> Why? C compiler is allowed to fuse/reorder loads from the same base
> object. Also stores can be reordered.
> 

Ok, right.  t->kcov_area can be loaded before t->kcov_mode, and it's fine. But deference of the kcov_area 
(READ_ONCE(area[0])) can't be moved before kcov_mode check. And this barrier intended to prevent such move, right?



> 
>>>> Signed-off-by: Andrey Ryabinin <aryabinin@...tuozzo.com>
>>>> ---
>>>>  kernel/kcov.c | 10 ----------
>>>>  1 file changed, 10 deletions(-)
>>>>
>>>> diff --git a/kernel/kcov.c b/kernel/kcov.c
>>>> index 14cc8c1a7cad..b7fbcbef88c1 100644
>>>> --- a/kernel/kcov.c
>>>> +++ b/kernel/kcov.c
>>>> @@ -71,14 +71,6 @@ void notrace __sanitizer_cov_trace_pc(void)
>>>>
>>>>                 ip -= kaslr_offset();
>>>>
>>>> -               /*
>>>> -                * There is some code that runs in interrupts but for which
>>>> -                * in_interrupt() returns false (e.g. preempt_schedule_irq()).
>>>> -                * READ_ONCE()/barrier() effectively provides load-acquire wrt
>>>> -                * interrupts, there are paired barrier()/WRITE_ONCE() in
>>>> -                * kcov_ioctl_locked().
>>>> -                */
>>>> -               barrier();
>>>>                 area = t->kcov_area;
>>>>                 /* The first word is number of subsequent PCs. */
>>>>                 pos = READ_ONCE(area[0]) + 1;
>>>> @@ -228,8 +220,6 @@ static int kcov_ioctl_locked(struct kcov *kcov, unsigned int cmd,
>>>>                 /* Cache in task struct for performance. */
>>>>                 t->kcov_size = kcov->size;
>>>>                 t->kcov_area = kcov->area;
>>>> -               /* See comment in __sanitizer_cov_trace_pc(). */
>>>> -               barrier();
>>>>                 WRITE_ONCE(t->kcov_mode, kcov->mode);
>>>>                 t->kcov = kcov;
>>>>                 kcov->t = t;
>>>> --
>>>> 2.13.5
>>>>
>>