[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20250129-quiet-ermine-of-growth-0ee6b6@leitao>
Date: Wed, 29 Jan 2025 09:12:33 -0800
From: Breno Leitao <leitao@...ian.org>
To: Michael Chan <michael.chan@...adcom.com>
Cc: pavan.chebbi@...adcom.com, netdev@...r.kernel.org, kuba@...nel.org,
kernel-team@...a.com
Subject: Re: bnxt_en: NETDEV WATCHDOG in 6.13-rc7
Hello Michael,
On Fri, Jan 17, 2025 at 09:44:12AM -0800, Michael Chan wrote:
> On Fri, Jan 17, 2025 at 4:08 AM Breno Leitao <leitao@...ian.org> wrote:
>
> > Showing all locks held in the system:
>
> > 7 locks held by kworker/u144:3/208:
> > 4 locks held by kworker/u144:4/290:
> > #0: ffff88811db39948 ((wq_completion)bnxt_pf_wq){+.+.}-{0:0}, at: process_one_work+0x1090/0x1950
> > #1: ffffc9000303fda0 ((work_completion)(&bp->sp_task)){+.+.}-{0:0}, at: process_one_work+0x7eb/0x1950
> > #2: ffffffff86f71208 (rtnl_mutex){+.+.}-{4:4}, at: bnxt_reset+0x30/0xa0
> > #3: ffff88811e41d160 (&bp->hwrm_cmd_lock){+.+.}-{4:4}, at: __hwrm_send+0x2f6/0x28d0
>
> Since there is TX timeout, we will call bnxt_reset() from
> bnxt_sp_task() workqueue. rtnl_lock will be held and we will hold the
> hwrm_cmd_lock mutex for every command we send to the firmware.
> Perhaps there is a problem communicating with the firmware. This will
> cause the firmware command to timeout in about a second with these
> locks held. We send many commands to the firmware and this can take a
> while if firmware is not responding.
>
> > 3 locks held by kworker/u144:6/322:
> > #0: ffff88810812a948 ((wq_completion)events_unbound){+.+.}-{0:0}, at: process_one_work+0x1090/0x1950
> > #1: ffffc90003a4fda0 ((linkwatch_work).work){+.+.}-{0:0}, at: process_one_work+0x7eb/0x1950
> > #2: ffffffff86f71208 (rtnl_mutex){+.+.}-{4:4}, at: linkwatch_event+0xe/0x60
>
> Meanwhile linkwatch is trying to get the rtnl_lock.
>
> >
> >
> > Full log at https://pastebin.com/4pWmaayt
> >
>
> I will take a closer look at the full log today. Thanks.
Thanks. Have you had a chance to look at it?
Do you suggest anything I can do on my side to get more data?
Thanks
--breno
Powered by blists - more mailing lists