Re: Deadlock in n_hdlc_buf_put

[Jiri Slaby <jslaby@suse.cz>]

On 11/26/2015, 01:37 PM, Dmitry Vyukov wrote:
> [  341.376188] =============================================
> [  341.376607] [ INFO: possible recursive locking detected ]
> [  341.376607] 4.4.0-rc1+ #117 Not tainted
> [  341.376607] ---------------------------------------------
> [  341.376607] syzkaller_execu/14066 is trying to acquire lock:
> [  341.376607]  (&(&list->spinlock)->rlock){......}, at:
> [<ffffffff82a9f548>] n_hdlc_buf_put+0x28/0x170
> [  341.376607]
> [  341.376607] but task is already holding lock:
> [  341.376607]  (&(&list->spinlock)->rlock){......}, at:
> [<ffffffff82aa1368>] n_hdlc_tty_ioctl+0x2b8/0x3f0
> [  341.376607]
> [  341.376607] other info that might help us debug this:
> [  341.376607]  Possible unsafe locking scenario:
> [  341.376607]
> [  341.376607]        CPU0
> [  341.376607]        ----
> [  341.376607]   lock(&(&list->spinlock)->rlock);
> [  341.376607]   lock(&(&list->spinlock)->rlock);

Hi,

this is a lockdep false positive. The first one is tx_buf_list.spinlock,
the latter tx_free_buf_list.spinlock, both in flush_tx_queue. So we need
a lockdep annotation here.

thanks,
-- 
js
suse labs