Re: [syzbot] [bpf?] possible deadlock in __bpf_ringbuf_reserve (2)

On 26/8/25 10:23, Alexei Starovoitov wrote:

On Mon, Aug 25, 2025 at 7:20 PM Leon Hwang [off-list ref] wrote:

quoted

On 26/8/25 01:39, syzbot wrote:

quoted

Hello,

syzbot found the following issue on:

HEAD commit:    dd9de524183a xsk: Fix immature cq descriptor production
git tree:       bpf
console output: https://syzkaller.appspot.com/x/log.txt?x=102da862580000
kernel config:  https://syzkaller.appspot.com/x/.config?x=c321f33e4545e2a1
dashboard link: https://syzkaller.appspot.com/bug?extid=fa5c2814795b5adca240
compiler:       Debian clang version 20.1.7 (++20250616065708+6146a88f6049-1~exp1~20250616065826.132), Debian LLD 20.1.7
syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=142da862580000
C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=1588aef0580000

Downloadable assets:
disk image: https://storage.googleapis.com/syzbot-assets/5a3389c1558f/disk-dd9de524.raw.xz
vmlinux: https://storage.googleapis.com/syzbot-assets/c97133192a27/vmlinux-dd9de524.xz
kernel image: https://storage.googleapis.com/syzbot-assets/3ae5a1a88637/bzImage-dd9de524.xz

IMPORTANT: if you fix the issue, please add the following tag to the commit:
Reported-by: syzbot+fa5c2814795b5adca240@syzkaller.appspotmail.com

============================================
WARNING: possible recursive locking detected
syzkaller #0 Not tainted
--------------------------------------------
syz-execprog/5866 is trying to acquire lock:
ffffc900048c10d8 (&rb->spinlock){-.-.}-{2:2}, at: __bpf_ringbuf_reserve+0x1c7/0x5a0 kernel/bpf/ringbuf.c:423

but task is already holding lock:
ffffc900048e90d8 (&rb->spinlock){-.-.}-{2:2}, at: __bpf_ringbuf_reserve+0x1c7/0x5a0 kernel/bpf/ringbuf.c:423

other info that might help us debug this:
 Possible unsafe locking scenario:

       CPU0
       ----
  lock(&rb->spinlock);
  lock(&rb->spinlock);

 *** DEADLOCK ***

 May be due to missing lock nesting notation

Confirmed.

I can reproduce this deadlock issue and will work on a fix.

Don't.

It's due to revert.
Once rqspinlock is fixed the revert will be reverted.

OK. I'll test it after the fixing.

Thanks,
Leon