linux-kernel - Re: WARNING: locking bug in lock

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]

Date:   Wed, 9 Jan 2019 17:46:03 -0500
From:   Waiman Long <longman@...hat.com>
To:     Tetsuo Handa <penguin-kernel@...ove.SAKURA.ne.jp>,
        peterz@...radead.org, "mingo@...hat.com" <mingo@...hat.com>
Cc:     Yang Shi <yang.shi@...ux.alibaba.com>,
        syzbot <syzbot+53383ae265fb161ef488@...kaller.appspotmail.com>,
        akpm@...ux-foundation.org, kirill.shutemov@...ux.intel.com,
        linux-kernel@...r.kernel.org, linux-mm@...ck.org,
        linux@...inikbrodowski.net, mhocko@...e.com, rientjes@...gle.com,
        syzkaller-bugs@...glegroups.com, vbabka@...e.cz,
        boqun.feng@...il.com
Subject: Re: WARNING: locking bug in lock_downgrade

On 01/09/2019 09:18 AM, Tetsuo Handa wrote:
> On 2018/12/14 4:46, Waiman Long wrote:
>> On 12/12/2018 08:14 PM, Yang Shi wrote:
>>> By looking into lockdep code, I'm not sure if lockdep may get confused
>>> by such sequence or not?
>>>
>>>
>>> Any hint is appreciated.
>>>
>>>
>>> Regards,
>>>
>>> Yang 
>> The warning was printed because hlock->read was set when doing the
>> downgrade_write(). So it is either downgrade_write() was called a second
>> time or a read lock was held originally. It is hard to tell what is the
>> root cause without a reproducer.
>>
>> Cheers,
>> Longman
>>
> Comparing with output from
>
>         struct rw_semaphore *sem = &current->mm->mmap_sem;
>
>         down_write(sem);
>         pr_warn("mmap_sem: count=%ld current=%px, owner=%px\n", atomic_long_read(&sem->count), current, READ_ONCE(sem->owner));
>         /* mmap_sem: count=-4294967295 current=ffff88813095ca80, owner=ffff88813095ca80 */
>         downgrade_write(sem);
>         pr_warn("mmap_sem: count=%ld current=%px, owner=%px\n", atomic_long_read(&sem->count), current, READ_ONCE(sem->owner));
>         /* mmap_sem: count=1 current=ffff88813095ca80, owner=ffff88813095ca83 */
>         up_read(sem);
>         pr_warn("mmap_sem: count=%ld current=%px, owner=%px\n", atomic_long_read(&sem->count), current, READ_ONCE(sem->owner));
>         /* mmap_sem: count=0 current=ffff88813095ca80, owner=0000000000000003 */

The behavior is correct. The current code will leave the reader task
structure pointer in owner even if it is a read lock. You have to look
at bit 0 to know if the owner is a reader or writer.

> what we got with debug printk() patch
>
>   https://syzkaller.appspot.com/text?tag=CrashLog&x=169dbb9b400000
>
>   [ 2580.337550][ T3645] mmap_sem: hlock->read=1 count=-4294967295 current=ffff888050e04140, owner=ffff888050e04140
>   [ 2580.353526][ T3645] ------------[ cut here ]------------
>   [ 2580.367859][ T3645] downgrading a read lock
>   [ 2580.367935][ T3645] WARNING: CPU: 1 PID: 3645 at kernel/locking/lockdep.c:3572 lock_downgrade+0x35d/0xbe0
>   [ 2580.382206][ T3645] Kernel panic - not syncing: panic_on_warn set ...
>
>   https://syzkaller.appspot.com/text?tag=CrashLog&x=1542da4f400000
>
>   [  386.342585][T16698] mmap_sem: hlock->read=1 count=-4294967295 current=ffff8880512ae180, owner=ffff8880512ae180
>   [  386.348586][T16698] ------------[ cut here ]------------
>   [  386.357203][T16698] downgrading a read lock
>   [  386.357294][T16698] WARNING: CPU: 1 PID: 16698 at kernel/locking/lockdep.c:3572 lock_downgrade+0x35d/0xbe0
>   [  386.372148][T16698] Kernel panic - not syncing: panic_on_warn set ...
>
> indicates that lockdep is saying that "current->mm->mmap_sem is held for read"
> while "struct rw_semaphore" is saying that "current->mm->mmap_sem is held for write".
> Something made lockdep confused. Possibly a lockdep bug.
>
It could be a bug in lockdep regarding downgrade. Someone else has
reported similar problem before.

-Longman