Bug 40972 - brcmsmac module intermittently stops working with kernel: fifo 0: descriptor error / fatal error
Summary: brcmsmac module intermittently stops working with kernel: fifo 0: descriptor ...
Status: NEEDINFO
Alias: None
Product: Networking
Classification: Unclassified
Component: Wireless (show other bugs)
Hardware: All Linux
: P1 normal
Assignee: networking_wireless@kernel-bugs.osdl.org
URL: rvossen@broadcom.com
Keywords:
Depends on:
Blocks:
 
Reported: 2011-08-12 00:28 UTC by Pedram
Modified: 2012-08-30 09:46 UTC (History)
5 users (show)

See Also:
Kernel Version: 3.3
Subsystem:
Regression: No
Bisected commit-id:


Attachments

Description Pedram 2011-08-12 00:28:50 UTC
I am on a fresh Arch install with linux 3.0.1-1, using the brcmsmac module.
Initially, the wireless works for a few days. At some point, it will
stop, outputting the following error repeatedly:

localhost kernel: [  825.796697] ieee80211 phy0: wl0: fifo 0: descriptor error
localhost kernel: [  825.796713] ieee80211 phy0: wl0: fatal error, re-initializing

I can rmmod brcmsmac, reboot, shutdown, give it a few days, and start
over and it will sometimes work. Other times, it will persist with the
fifo errors. 

In other wods, a full shutdown may not resolve the issue and it reappears as soon as I boot and run ifconfig wlan0 up. I do not receive the error if I unload the module or if I bring down the interface using ifconfig wlan0 down.

My current network connection is a wpa2 Wireless N connection. Here is the output of iwconfig when it is able to connect:

wlan0     IEEE 802.11bgn  ESSID:"biteynet"  
          Mode:Managed  Frequency:2.412 GHz  Access Point: C0:C1:C0:EA:02:FF   
          Bit Rate=65 Mb/s   Tx-Power=19 dBm   
          Retry  long limit:7   RTS thr:off   Fragment thr:off
          Power Management:off
          Link Quality=70/70  Signal level=-18 dBm  
          Rx invalid nwid:0  Rx invalid crypt:0  Rx invalid frag:0
          Tx excessive retries:0  Invalid misc:14   Missed beacon:0

#    lspci -vnn | grep 14e4
05:00.0 Network controller [0280]: Broadcom Corporation
802.11b/g/n Wireless LAN Controller [14e4:4727] (rev 01)
       Subsystem: Broadcom Corporation Device [14e4:0510]

 #  lsmod 

Module                  Size  Used by
cryptd                  6893  0 
aes_i586                6940  2 
aes_generic            25670  1 aes_i586
appletalk              22087  0 
ipx                    16898  0 
p8022                    855  1 ipx
psnap                   1489  2 appletalk,ipx
llc                     3029  2 p8022,psnap
p8023                    784  1 ipx
ipv6                  249911  28 
ext2                   55531  1 
joydev                  7439  0 
lib80211                3270  0 
arc4                    1086  2 
brcmsmac              559153  0 
brcmutil                5471  1 brcmsmac
mac80211              191509  1 brcmsmac
snd_hda_codec_realtek   222124  1 
snd_hda_intel          19101  0 
uvcvideo               57363  0 
cfg80211              138620  2 brcmsmac,mac80211
snd_hda_codec          66954  2 snd_hda_codec_realtek,snd_hda_intel
snd_hwdep               4942  1 snd_hda_codec
snd_pcm                60015  2 snd_hda_intel,snd_hda_codec
snd_timer              15374  1 snd_pcm
snd                    43593  6 snd_hda_codec_realtek,snd_hda_intel,snd_hda_codec,snd_hwdep,snd_pcm,snd_timer
videodev               70542  1 uvcvideo
acer_wmi               15921  0 
iTCO_wdt               10293  0 
soundcore               4986  1 snd
serio_raw               3390  0 
psmouse                56431  0 
sparse_keymap           2628  1 acer_wmi
media                   8545  2 uvcvideo,videodev
rfkill                 12406  2 cfg80211,acer_wmi
pcspkr                  1375  0 
i2c_i801                7063  0 
iTCO_vendor_support     1545  1 iTCO_wdt
sg                     20982  0 
snd_page_alloc          5869  2 snd_hda_intel,snd_pcm
atl1c                  28528  0 
crc_ccitt               1095  1 brcmsmac
evdev                   7278  8 
battery                 8865  0 
wmi                     7295  1 acer_wmi
ac                      2505  0 
fan                     1902  0 
processor              21876  0 
thermal                 6595  0 
ext4                  337724  2 
mbcache                 4249  2 ext2,ext4
jbd2                   59488  1 ext4
crc16                   1069  1 ext4
sd_mod                 26147  5 
uhci_hcd               19609  0 
ahci                   18085  4 
libahci                16573  1 ahci
libata                154449  2 ahci,libahci
ehci_hcd               34935  0 
scsi_mod              111801  3 sg,sd_mod,libata
usbcore               119612  4 uvcvideo,uhci_hcd,ehci_hcd
i915                  612452  2 
drm_kms_helper         21997  1 i915
drm                   147613  3 i915,drm_kms_helper
intel_agp               8720  1 i915
i2c_algo_bit            4423  1 i915
button                  3646  1 i915
intel_gtt              11417  3 i915,intel_agp
agpgart                22096  3 drm,intel_agp,intel_gtt
i2c_core               16625  6 videodev,i2c_i801,i915,drm_kms_helper,drm,i2c_algo_bit
video                   9780  1 i915

# uname -a
Linux pineapple 3.0-ARCH #1 SMP PREEMPT Sat Aug 6 16:49:00 CEST 2011 i686 Intel(R) Atom(TM) CPU N455 @ 1.66GHz GenuineIntel GNU/Linux

# cat /etc/modprobe.d/modprobe.conf
blacklist bcma

I have run both 64-bit and 32-bit ARCH and experienced the issue on both. 

As this is intermittent, it may take some time for me to be able to reproduce the issue, but I will leave the laptop running to encourage its return. I have not seen the issue with the wl module through the broadcom-wl package in AUR, running 2.6.39 but have not tested this extensively.
Comment 1 Leho Kraav 2011-08-23 13:21:14 UTC
brcmsmac has been my bane since i put linux on this acer travelmate 8172T laptop, starting with 2.6.38. just updated my report with 3.0.1 update, or in other words, no change in being broken.

https://bugzilla.kernel.org/show_bug.cgi?id=34682

started trial period with broadcom-sta.
Comment 2 Pedram 2011-08-23 14:24:43 UTC
I'd like to report that this seems to have been auto-magically fixed running 32-bit. I have not reinstalled using 64-bit to see if the problem is still there.
Comment 3 Leho Kraav 2011-08-23 14:34:59 UTC
i for one am running 32-bit with PAE.
Comment 4 Roland Vossen 2011-08-23 14:39:01 UTC
Hi Pedram,

that is unsettling news to me :-) You didn't install something new during the past couple of days ?

Bye, Roland.
Comment 5 Pedram 2011-08-23 14:46:08 UTC
Hi Roland,

I do not know how to explain it. I did install kernel-netbook through AUR on ARch, which came with 2.6.39. I don't see how this would affect anything related to the 3.0 kernel or the brcmsmac module. It may be a coincidence but it is the only thing I can think of.

When I rebooted into 3.0 to test the module again, my brcmsmac module loaded fine and I have not had a problem since.

I will create a new partition and install a fresh Arch setup in 32-bit to see if I can recreate the problem for you.
Comment 6 Roland Vossen 2011-08-30 08:39:40 UTC
Hello Pedram,

did you already try the experiment of creating a new partition with a fresh Arch 32 bits install ? Curious to know if the problem is still there.

Bye, Roland.
Comment 7 Pedram 2011-08-30 13:01:40 UTC
Hi Roland,

I have reinstalled and unfortunately have not been able to replicate this problem on a 32-bit installation.
Comment 8 Roland Vossen 2011-08-30 13:13:16 UTC
Well, at least thank you for trying it. Shall we close this bug and reopen it if the issue pops up again ?

Bye, Roland.
Comment 9 Pedram 2011-08-31 17:06:20 UTC
Roland, great news! I was able to replicate the bug. 

What I did was install wicd, which also had the following dependencies: 
dbus-python
ethtool
python-urwid

Once installed, I started dbus using rc.d start dbus
and I started wicd using rc.d start wicd.

I then ran the wicd-client, tried to connect to a network using the tool and noticed my computer slow down to a crawl and then the fifo 0: descriptior error / fatal error messages in kernel.log.

I can reboot and if I do not load dbus or wicd and do not experience any issues. I am able to connect fine using netcfg. I can bring up the interface, run iwlist wlan0 scanning, connect to a network, and run dhcpcd to retrieve an ip address.
Comment 10 Roland Vossen 2011-09-01 07:19:19 UTC
That is indeed great news. I will try to replicate this right now.
Comment 11 Roland Vossen 2011-09-01 14:49:59 UTC
I am on a different distribution than you, so I tested on Ubuntu 11.04 (32 bits). I installed wicd but have not been able so far to replicate the problem you are seeing. Can you supply me with the wicd version you use ? Also, how did you get your kernel: did you pull it from git and if so, do you have a SHA1 for me ?

Also, do you have the right firmware:

$ md5sum /lib/firmware/brcm/*
c53608f5818b702c46a012c57b4196ee  /lib/firmware/brcm/bcm4329-fullmac-4.bin
b308d5bab6b373c2f1a550ed09573f3a  /lib/firmware/brcm/bcm4329-fullmac-4.txt
96cf06e4ff9f0c04a0f26ebefdf32e3d  /lib/firmware/brcm/bcm43xx-0.fw
48882412db63b4e2dd9c26571a29a799  /lib/firmware/brcm/bcm43xx_hdr-0.fw

Thanks, Roland.
Comment 12 Pedram 2011-09-01 15:04:18 UTC
I am running wicd 1.7.0-11
http://www.archlinux.org/packages/testing/any/wicd/

My firmware md5's match the ones you provided.

My kernel is pulled from my distribution's package manager. 
http://www.archlinux.org/packages/core/i686/linux/

Currently I am on 3.0.3-1
Comment 13 Roland Vossen 2011-09-02 14:33:47 UTC
Hello Pedram,

I tried very hard replicating this bug but did not succeed. I compiled the 3.0.3 kernel and installed it on a Ubuntu x86 machine. I also use the same wicd version as you.

Since you can replicate this very well and I can not, that calls for some debugging effort on your side, if you are up and willing to do this.

I can send you versions of brcm source files containing debug statements (printf's), and in this way we can zoom in onto the issue. Expect some iterations over email.

Are you willing to give this a try ?

Bye, Roland.
Comment 14 Pedram 2011-09-02 14:51:57 UTC
Absolutely, I'd be very happy to give it a try. Feel free to email me and we can work through them that way.
Comment 15 Roland Vossen 2011-09-02 15:06:58 UTC
Pedram kindly accepted the offer, we will debug this issue over email and
update this bug report when conclusions arise.
Comment 16 John W. Linville 2012-01-13 19:15:02 UTC
Does this problem persist with 3.2 kernels?
Comment 17 Patrick Stetter 2012-04-09 18:36:23 UTC
I still see this issue with brcmsmac loading inconsistently on 3.2.0-22-generic-pae on 32bit ubuntu.  I would be happy to provide any additional information necessary.
Comment 18 Pedram 2012-05-30 19:21:49 UTC
I am still seeing this error in 3.3..

Note You need to log in before you can comment on or make changes to this bug.