Re: NETDEV WATCHDOG: eth0 (atl1c): transmit queue 0 timed out

[Eric Dumazet <eric.dumazet@gmail.com>]

Le mardi 14 février 2012 à 08:36 +0100, Eric Dumazet a écrit :
> Le lundi 13 février 2012 à 16:00 -0500, Josh Boyer a écrit :
> > On Sun, Feb 12, 2012 at 11:16 AM, Thomas Meyer <thomas@m3y3r.de> wrote:
> > > abrt_version:   2.0.7
> > > cmdline:        BOOT_IMAGE=/vmlinuz-3.2.3 root=/dev/sda2 rootfstype=ext4 rootflags=data=writeback ro quiet rhgb
> > > kernel:         3.2.3
> > > reason:         WARNING: at net/sched/sch_generic.c:255 dev_watchdog+0x1e7/0x1f0()
> > > time:           Sa 04 Feb 2012 15:31:22 CET
> > >
> > > backtrace:
> > > :WARNING: at net/sched/sch_generic.c:255 dev_watchdog+0x1e7/0x1f0()
> > > :Hardware name: Aspire 1810T
> > > :NETDEV WATCHDOG: eth0 (atl1c): transmit queue 0 timed out
> > > :Modules linked in: vfat fat fuse bluetooth nf_conntrack_ipv6 nf_defrag_ipv6 ip6t_REJECT ip6table_filter ip6_tables snd_hda_codec_hdmi snd_hda_codec_realtek snd_hda_intel snd_hda_codec snd_hwdep uvcvideo snd_seq arc4 videodev usb_storage snd_seq_device v4l2_compat_ioctl32 snd_pcm snd_timer acer_wmi iwlwifi snd mac80211 sparse_keymap cfg80211 rfkill soundcore snd_page_alloc joydev wmi acerhdf pcspkr virtio_net virtio atl1c virtio_ring kvm_intel kvm uinput ipv6 [last unloaded: scsi_wait_scan]
> > > :Pid: 0, comm: swapper Not tainted 3.2.3 #5
> > > :Call Trace:
> > > : <IRQ>  [<ffffffff810322ca>] warn_slowpath_common+0x7a/0xb0
> > > : [<ffffffff81018a18>] ? lapic_next_event+0x18/0x20
> > > : [<ffffffff810323a1>] warn_slowpath_fmt+0x41/0x50
> > > : [<ffffffff8104e9be>] ? hrtimer_interrupt+0xfe/0x1f0
> > > : [<ffffffff8102c350>] ? wake_up_process+0x10/0x20
> > > : [<ffffffff813e9397>] dev_watchdog+0x1e7/0x1f0
> > > : [<ffffffff8103cfcf>] run_timer_softirq+0xef/0x210
> > > : [<ffffffff813e91b0>] ? qdisc_reset+0x50/0x50
> > > : [<ffffffff81037827>] __do_softirq+0x87/0x110
> > > : [<ffffffff814945fa>] call_softirq+0x1a/0x30
> > > : [<ffffffff810038bd>] do_softirq+0x4d/0x80
> > > : [<ffffffff81037a8e>] irq_exit+0x7e/0xa0
> > > : [<ffffffff8100375b>] do_IRQ+0x5b/0xd0
> > > : [<ffffffff8148de29>] common_interrupt+0x69/0x69
> > > : <EOI>  [<ffffffff8124e115>] ? acpi_idle_enter_bm+0x1c0/0x1ff
> > > : [<ffffffff8124e110>] ? acpi_idle_enter_bm+0x1bb/0x1ff
> > > : [<ffffffff813a24b4>] ? menu_select+0xd4/0x390
> > > : [<ffffffff813a150b>] cpuidle_idle_call+0x8b/0xe0
> > > : [<ffffffff810011ed>] cpu_idle+0x9d/0xe0
> > > : [<ffffffff8147d08e>] rest_init+0x62/0x64
> > > : [<ffffffff81699b54>] start_kernel+0x358/0x363
> > > : [<ffffffff81699322>] x86_64_start_reservations+0x132/0x136
> > > : [<ffffffff81699416>] x86_64_start_kernel+0xf0/0xf7
> > > : [<ffffffff81017e70>] ? acpi_suspend_lowlevel+0x1b0/0x1b0
> > 
> > We've seen this in Fedora for a long time now.  As far as I know, nobody
> > really knows what is going on.
> > 
> > https://bugzilla.redhat.com/show_bug.cgi?id=717211
> > 
> > (Also, you probably should have sent this to netdev.)
> 
> CC netdev, Jay Cliburn <jcliburn@gmail.com>, 
> 
> It seems this driver has partial support for two TX rings.
> 
> (TX completion only drains the first ring)
> 
> Please try following patch.
> 
> diff --git a/drivers/net/ethernet/atheros/atl1c/atl1c_main.c b/drivers/net/ethernet/atheros/atl1c/atl1c_main.c
> index b859124..1ff3c6d 100644
> --- a/drivers/net/ethernet/atheros/atl1c/atl1c_main.c
> +++ b/drivers/net/ethernet/atheros/atl1c/atl1c_main.c
> @@ -2244,10 +2244,6 @@ static netdev_tx_t atl1c_xmit_frame(struct sk_buff *skb,
>  			dev_info(&adapter->pdev->dev, "tx locked\n");
>  		return NETDEV_TX_LOCKED;
>  	}
> -	if (skb->mark == 0x01)
> -		type = atl1c_trans_high;
> -	else
> -		type = atl1c_trans_normal;
>  
>  	if (atl1c_tpd_avail(adapter, type) < tpd_req) {
>  		/* no enough descriptor, just stop queue */
> 


Thomas, have you had the chance to test this patch ?

Thanks !