WARNING: at net/mac80211/iface.c:162

WARNING: at net/mac80211/iface.c:162

[Larry Finger]

Hi,

I am getting the following warning from the latest wireless-testing system:

  ------------[ cut here ]------------
  WARNING: at net/mac80211/iface.c:162 ieee80211_check_queues+0x119/0x130 
[mac80211]()
  Hardware name: MS-7309
  Modules linked in: nouveau ttm drm_kms_helper drm hwmon mxm_wmi video wmi 
ipt_REJECT iptable_raw iptable_filter nf_conntrack_netbios_ns 
nf_conntrack_broadcast nf_conntrack_ipv4 nf_conntrack nf_defrag_ipv4 ip_tables 
x_tables ext3 jbd arc4 b43legacy rng_core rtl8187 acx_mac80211(O+) mac80211 fuse 
snd_hda_codec_realtek snd_hda_intel snd_hda_codec snd_pcm cfg80211 ssb mmc_core 
eeprom_93cx6 sr_mod cdrom ata_generic snd_timer snd pata_amd soundcore rfkill 
pcmcia snd_page_alloc pcmcia_core powernow_k8 i2c_nforce2 button i2c_core mperf 
forcedeth ipv6 autofs4 ext4 mbcache jbd2 crc16 ohci_hcd ehci_hcd sata_nv usbcore 
usb_common processor scsi_dh_hp_sw scsi_dh_emc scsi_dh_rdac scsi_dh_alua scsi_dh
  Pid: 2299, comm: NetworkManager Tainted: G           O 3.4.0-rc5-wl+ #39
  Call Trace:
   [<ffffffff81037f6a>] warn_slowpath_common+0x7a/0xb0
   [<ffffffff81037fb5>] warn_slowpath_null+0x15/0x20
   [<ffffffffa038c169>] ieee80211_check_queues+0x119/0x130 [mac80211]
   [<ffffffffa038ed79>] ieee80211_do_open+0x179/0x1200 [mac80211]
   [<ffffffff810657c4>] ? notifier_call_chain+0x54/0x110
   [<ffffffffa038fe6d>] ieee80211_open+0x6d/0x80 [mac80211]
   [<ffffffff8132f397>] __dev_open+0x97/0xe0
   [<ffffffff8132f61c>] __dev_change_flags+0x9c/0x180
   [<ffffffff8132f7b3>] dev_change_flags+0x23/0x70
   [<ffffffff8133d99f>] do_setlink+0x25f/0x9b0
   [<ffffffff8120e420>] ? nla_parse+0x30/0xe0
   [<ffffffff8133eaa5>] rtnl_newlink+0x345/0x550
   [<ffffffff810975f1>] ? mark_held_locks+0x61/0x140
   [<ffffffff81043c44>] ? ns_capable+0x44/0x80
   [<ffffffff8133e344>] rtnetlink_rcv_msg+0x114/0x2e0
   [<ffffffff8133d532>] ? rtnl_lock+0x12/0x20
   [<ffffffff8133e230>] ? __rtnl_unlock+0x20/0x20
   [<ffffffff813519e9>] netlink_rcv_skb+0xa9/0xd0
   [<ffffffff8133d560>] rtnetlink_rcv+0x20/0x30
   [<ffffffff813513ff>] netlink_unicast+0x19f/0x1f0
   [<ffffffff813517cd>] netlink_sendmsg+0x2bd/0x320
   [<ffffffff813191c6>] ? sock_update_classid+0x86/0x3a0
   [<ffffffff81312b52>] sock_sendmsg+0xe2/0x120
   [<ffffffff81113817>] ? might_fault+0x57/0xb0
   [<ffffffff81113860>] ? might_fault+0xa0/0xb0
   [<ffffffff81113817>] ? might_fault+0x57/0xb0
   [<ffffffff81321a86>] ? verify_iovec+0x56/0xd0
   [<ffffffff81312fe6>] __sys_sendmsg+0x376/0x380
   [<ffffffff8113f773>] ? fget_light+0x43/0x4e0
   [<ffffffff8113f7fd>] ? fget_light+0xcd/0x4e0
   [<ffffffff8113f773>] ? fget_light+0x43/0x4e0
   [<ffffffff81316204>] sys_sendmsg+0x44/0x80
   [<ffffffff813d70e2>] system_call_fastpath+0x16/0x1b
  ---[ end trace 415685f07ec6c955 ]---

This warning seems to be the same as the one reported at

https://bugzilla.redhat.com/show_bug.cgi?id=817490

The line in question is

      if (WARN_ON_ONCE(sdata->vif.hw_queue[i] >= n_queues))

The warning is triggered for i, n_queues, and sdata->vif.hw_queue[i] all equal to 1.

Larry

Re: WARNING: at net/mac80211/iface.c:162

[Johannes Berg]

On Sun, 2012-05-06 at 18:30 -0500, Larry Finger wrote:
> Hi,
> 
> I am getting the following warning from the latest wireless-testing system:
> 
>   ------------[ cut here ]------------
>   WARNING: at net/mac80211/iface.c:162 ieee80211_check_queues+0x119/0x130 
> [mac80211]()


>       if (WARN_ON_ONCE(sdata->vif.hw_queue[i] >= n_queues))
> 
> The warning is triggered for i, n_queues, and sdata->vif.hw_queue[i] all equal to 1.

Is this a driver with just a single queue by any chance? I never tested
that and probably forgot to set those up correctly.

johannes

Re: WARNING: at net/mac80211/iface.c:162

[Arend van Spriel]

On 05/07/2012 08:34 AM, Johannes Berg wrote:
> On Sun, 2012-05-06 at 18:30 -0500, Larry Finger wrote:
>> Hi,
>>
>> I am getting the following warning from the latest wireless-testing system:
>>
>>   ------------[ cut here ]------------
>>   WARNING: at net/mac80211/iface.c:162 ieee80211_check_queues+0x119/0x130 
>> [mac80211]()
> 
> 
>>       if (WARN_ON_ONCE(sdata->vif.hw_queue[i] >= n_queues))
>>
>> The warning is triggered for i, n_queues, and sdata->vif.hw_queue[i] all equal to 1.
> 
> Is this a driver with just a single queue by any chance? I never tested
> that and probably forgot to set those up correctly.
> 

The bug report Larry mentions is with iwlwifi driver:

pci, iwlwifi, NETWORK, Centrino Wireless-N 130 BGN

Gr. AvS

Re: WARNING: at net/mac80211/iface.c:162

[Johannes Berg]

On Mon, 2012-05-07 at 09:55 +0200, Arend van Spriel wrote:
> On 05/07/2012 08:34 AM, Johannes Berg wrote:
> > On Sun, 2012-05-06 at 18:30 -0500, Larry Finger wrote:
> >> Hi,
> >>
> >> I am getting the following warning from the latest wireless-testing system:
> >>
> >>   ------------[ cut here ]------------
> >>   WARNING: at net/mac80211/iface.c:162 ieee80211_check_queues+0x119/0x130 
> >> [mac80211]()
> > 
> > 
> >>       if (WARN_ON_ONCE(sdata->vif.hw_queue[i] >= n_queues))
> >>
> >> The warning is triggered for i, n_queues, and sdata->vif.hw_queue[i] all equal to 1.
> > 
> > Is this a driver with just a single queue by any chance? I never tested
> > that and probably forgot to set those up correctly.
> > 
> 
> The bug report Larry mentions is with iwlwifi driver:
> 
> pci, iwlwifi, NETWORK, Centrino Wireless-N 130 BGN

No, the bug report says rtl8187, the user said this happened when
plugging in the USB device.

I think this will fix it:

--- a/net/mac80211/iface.c
+++ b/net/mac80211/iface.c
@@ -206,8 +206,10 @@ static void ieee80211_set_default_queues(struct ieee80211_sub_if_data *sdata)
 	for (i = 0; i < IEEE80211_NUM_ACS; i++) {
 		if (local->hw.flags & IEEE80211_HW_QUEUE_CONTROL)
 			sdata->vif.hw_queue[i] = IEEE80211_INVAL_HW_QUEUE;
-		else
+		else if (local->hw.queues >= IEEE80211_NUM_ACS)
 			sdata->vif.hw_queue[i] = i;
+		else
+			sdata->vif.hw_queue[i] = 0;
 	}
 	sdata->vif.cab_queue = IEEE80211_INVAL_HW_QUEUE;
 }


johannes

Re: WARNING: at net/mac80211/iface.c:162

[Larry Finger]

On 05/07/2012 01:34 AM, Johannes Berg wrote:
> On Sun, 2012-05-06 at 18:30 -0500, Larry Finger wrote:
>> Hi,
>>
>> I am getting the following warning from the latest wireless-testing system:
>>
>>    ------------[ cut here ]------------
>>    WARNING: at net/mac80211/iface.c:162 ieee80211_check_queues+0x119/0x130
>> [mac80211]()
>
>
>>        if (WARN_ON_ONCE(sdata->vif.hw_queue[i]>= n_queues))
>>
>> The warning is triggered for i, n_queues, and sdata->vif.hw_queue[i] all equal to 1.
>
> Is this a driver with just a single queue by any chance? I never tested
> that and probably forgot to set those up correctly.

Yes, the device has only one hardware queue. Is this something the driver needs 
to do?

Larry

Re: WARNING: at net/mac80211/iface.c:162

[Larry Finger]

On 05/07/2012 06:29 AM, Johannes Berg wrote:
> On Mon, 2012-05-07 at 09:55 +0200, Arend van Spriel wrote:
>> On 05/07/2012 08:34 AM, Johannes Berg wrote:
>>> On Sun, 2012-05-06 at 18:30 -0500, Larry Finger wrote:
>>>> Hi,
>>>>
>>>> I am getting the following warning from the latest wireless-testing system:
>>>>
>>>>    ------------[ cut here ]------------
>>>>    WARNING: at net/mac80211/iface.c:162 ieee80211_check_queues+0x119/0x130
>>>> [mac80211]()
>>>
>>>
>>>>        if (WARN_ON_ONCE(sdata->vif.hw_queue[i]>= n_queues))
>>>>
>>>> The warning is triggered for i, n_queues, and sdata->vif.hw_queue[i] all equal to 1.
>>>
>>> Is this a driver with just a single queue by any chance? I never tested
>>> that and probably forgot to set those up correctly.
>>>
>>
>> The bug report Larry mentions is with iwlwifi driver:
>>
>> pci, iwlwifi, NETWORK, Centrino Wireless-N 130 BGN
>
> No, the bug report says rtl8187, the user said this happened when
> plugging in the USB device.
>
> I think this will fix it:
>
> --- a/net/mac80211/iface.c
> +++ b/net/mac80211/iface.c
> @@ -206,8 +206,10 @@ static void ieee80211_set_default_queues(struct ieee80211_sub_if_data *sdata)
>   	for (i = 0; i<  IEEE80211_NUM_ACS; i++) {
>   		if (local->hw.flags&  IEEE80211_HW_QUEUE_CONTROL)
>   			sdata->vif.hw_queue[i] = IEEE80211_INVAL_HW_QUEUE;
> -		else
> +		else if (local->hw.queues>= IEEE80211_NUM_ACS)
>   			sdata->vif.hw_queue[i] = i;
> +		else
> +			sdata->vif.hw_queue[i] = 0;
>   	}
>   	sdata->vif.cab_queue = IEEE80211_INVAL_HW_QUEUE;
>   }

Yes, this patch fixes the problem.

Reported-and-Tested-by: Larry Finger <Larry.Finger@lwfinger.net>

Once you post the patch, I will update that entry at bugzilla.redhat.com.

Thanks,

Larry

Re: WARNING: at net/mac80211/iface.c:162

[Johannes Berg]

On Mon, 2012-05-07 at 10:12 -0500, Larry Finger wrote:
> On 05/07/2012 06:29 AM, Johannes Berg wrote:
> > On Mon, 2012-05-07 at 09:55 +0200, Arend van Spriel wrote:
> >> On 05/07/2012 08:34 AM, Johannes Berg wrote:
> >>> On Sun, 2012-05-06 at 18:30 -0500, Larry Finger wrote:
> >>>> Hi,
> >>>>
> >>>> I am getting the following warning from the latest wireless-testing system:
> >>>>
> >>>>    ------------[ cut here ]------------
> >>>>    WARNING: at net/mac80211/iface.c:162 ieee80211_check_queues+0x119/0x130
> >>>> [mac80211]()
> >>>
> >>>
> >>>>        if (WARN_ON_ONCE(sdata->vif.hw_queue[i]>= n_queues))
> >>>>
> >>>> The warning is triggered for i, n_queues, and sdata->vif.hw_queue[i] all equal to 1.
> >>>
> >>> Is this a driver with just a single queue by any chance? I never tested
> >>> that and probably forgot to set those up correctly.
> >>>
> >>
> >> The bug report Larry mentions is with iwlwifi driver:
> >>
> >> pci, iwlwifi, NETWORK, Centrino Wireless-N 130 BGN
> >
> > No, the bug report says rtl8187, the user said this happened when
> > plugging in the USB device.
> >
> > I think this will fix it:
> >
> > --- a/net/mac80211/iface.c
> > +++ b/net/mac80211/iface.c
> > @@ -206,8 +206,10 @@ static void ieee80211_set_default_queues(struct ieee80211_sub_if_data *sdata)
> >   	for (i = 0; i<  IEEE80211_NUM_ACS; i++) {
> >   		if (local->hw.flags&  IEEE80211_HW_QUEUE_CONTROL)
> >   			sdata->vif.hw_queue[i] = IEEE80211_INVAL_HW_QUEUE;
> > -		else
> > +		else if (local->hw.queues>= IEEE80211_NUM_ACS)
> >   			sdata->vif.hw_queue[i] = i;
> > +		else
> > +			sdata->vif.hw_queue[i] = 0;
> >   	}
> >   	sdata->vif.cab_queue = IEEE80211_INVAL_HW_QUEUE;
> >   }
> 
> Yes, this patch fixes the problem.
> 
> Reported-and-Tested-by: Larry Finger <Larry.Finger@lwfinger.net>
> 
> Once you post the patch, I will update that entry at bugzilla.redhat.com.

Great, thanks.

johannes