Re: [syzbot] [kernel?] WARNING in hrtimer_forward (4)

[Shrikanth Hegde <sshegde@linux.ibm.com>]

On 9/11/25 2:45 PM, Juri Lelli wrote:
> On 10/09/25 22:07, Thomas Gleixner wrote:
>> On Fri, Aug 29 2025 at 19:00, syzbot wrote:
>>
>>> HEAD commit:    b6add54ba618 Merge tag 'pinctrl-v6.17-2' of git://git.kern..
>>> git tree:       upstream
>>> console output: https://syzkaller.appspot.com/x/log.txt?x=1130eef0580000
>>> kernel config:  https://syzkaller.appspot.com/x/.config?x=e1e1566c7726877e
>>> dashboard link: https://syzkaller.appspot.com/bug?extid=8b3a2e23253b50098164
>>> compiler:       Debian clang version 20.1.7 (++20250616065708+6146a88f6049-1~exp1~20250616065826.132), Debian LLD 20.1.7
>>>
>>> Unfortunately, I don't have any reproducer for this issue yet.
>>>
>>> Downloadable assets:
>>> disk image: https://storage.googleapis.com/syzbot-assets/102656909b6f/disk-b6add54b.raw.xz
>>> vmlinux: https://storage.googleapis.com/syzbot-assets/fa30d1d80a47/vmlinux-b6add54b.xz
>>> kernel image: https://storage.googleapis.com/syzbot-assets/c25ee8abf30a/bzImage-b6add54b.xz
>>>
>>> IMPORTANT: if you fix the issue, please add the following tag to the commit:
>>> Reported-by: syzbot+8b3a2e23253b50098164@syzkaller.appspotmail.com
>>>
>>> ------------[ cut here ]------------
>>> WARNING: CPU: 1 PID: 1186 at kernel/time/hrtimer.c:1052 hrtimer_forward+0x1d6/0x2b0 kernel/time/hrtimer.c:1052
>>> Modules linked in:
>>> CPU: 1 UID: 0 PID: 1186 Comm: irq/33-virtio1- Not tainted syzkaller #0 PREEMPT_{RT,(full)}
>>> Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 07/12/2025
>>> RIP: 0010:hrtimer_forward+0x1d6/0x2b0 kernel/time/hrtimer.c:1052
>>
>> It compains that the timer is enqueued when it is attempted to be forwarded
>>
>>> Code: 4c 89 33 48 8b 04 24 eb 07 e8 86 34 12 00 31 c0 48 83 c4 18 5b 41 5c 41 5d 41 5e 41 5f 5d e9 01 d8 4d 09 cc e8 6b 34 12 00 90 <0f> 0b 90 eb df 48 89 e8 4c 09 f8 48 c1 e8 20 74 0a 48 89 e8 31 d2
>>> RSP: 0018:ffffc90000a78bd0 EFLAGS: 00010006
>>> RAX: ffffffff81ac27e5 RBX: ffff8880b883b508 RCX: ffff888026c19dc0
>>> RDX: 0000000000000100 RSI: 0000000000010000 RDI: 0000000000010100
>>> RBP: 000000000009d057 R08: 0000000000010000 R09: 0000000000010100
>>> R10: dffffc0000000000 R11: ffffffff8167a890 R12: ffff8880b883b520
>>> R13: 0000000000184487 R14: 1ffff110171076a4 R15: 0000000000000001
>>> FS:  0000000000000000(0000) GS:ffff8881269c2000(0000) knlGS:0000000000000000
>>> CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
>>> CR2: 00007f95323cbf98 CR3: 0000000064088000 CR4: 00000000003526f0
>>> Call Trace:
>>>   <IRQ>
>>>   hrtimer_forward_now include/linux/hrtimer.h:366 [inline]
>>>   dl_server_timer kernel/sched/deadline.c:1193 [inline]
>>
>> which is strange as this is with the timer callback itself, so it
>> shouldn't be enqueued, unless there is a possiblilty to have:
>>
>>     CPU0                       CPU1
>>                                  
>>     timer_expires()
>>        callback()              ????
>>          dl_task_timer()       rq_lock()
>>            rq_lock()             hrtimer_start()
>>                                rq_unlock()
>>             hrtimer_forward()
>>
>> No idea whether that's possible, but that's the only sensible
>> explanation.
> 
> So, a dl_server_start() could be your ????, but it should see
> dl_server_active and just return if the dl_server callback is running.
> Unless a dl_server_stop() somehow interleaved as well and cleared it.
> 

isn't dl_server timer per CPU?

> If we had a reproducer maybe some tracing would help reducing the
> guessing. :)
> 
> Thanks,
> Juri
>