modprobe fail and subsequent crash on rmmod ath9k D-Link DWA 652

modprobe fail and subsequent crash on rmmod ath9k D-Link DWA 652

[Russ Adams]

Hello all,

Environment:

I'm running 2.6.32 with a set of patches provided by Arch Linux:
ftp://ftp.archlinux.org/other/kernel26/patch-2.6.32.1-1-ARCH.bz2.  The
patches appear mostly cosmetic, with some changes applied to ext4.

Hardware:

I have a D-Link DWA 652 which reports as an AR5008 in lspci -vnn.  The
card is functional in kernel 2.6.28. (I'm using it to post this email)

The ath9k list indicated that the problem in rmmod may be in the
mac80211b module.

Here's the sequence of events:

% sudo modprobe ath9k # no pertinent dmesg output

% iwconfig # wlan0 not present
lo        no wireless extensions.

eth0      no wireless extensions.

% sudo rmmod ath9k # not sure why it's telling me "Killed" here
Killed

% dmesg | less
...
...
BUG: unable to handle kernel NULL pointer dereference at 0000000c
IP: [<c116de6a>] kref_put+0x1a/0x60
*pde = 00000000
Oops: 0002 [#1] PREEMPT SMP
last sysfs file:
/sys/devices/LNXSYSTM:00/LNXSYBUS:00/PNP0C0A:00/power_supply/BAT0/energy_full
Modules linked in: ipv6 fan nls_cp437 hid_microsoft vfat fat
snd_seq_dummy firewire_ohci mmc_block snd_seq_oss firewire_core
snd_seq_midi_event snd_seq snd_seq_device snd_pcm_oss sdhci_pci
snd_intel8x0 radeon snd_intel8x0m crc_itu_t ttm snd_ac97_codec sdhci
drm_kms_helper snd_pcm drm i2c_algo_bit snd_timer intel_agp
snd_mixer_oss video snd ac97_bus joydev iTCO_wdt i2c_core
iTCO_vendor_support soundcore shpchp output dell_laptop ac battery
button psmouse snd_page_alloc agpgart pci_hotplug processor thermal
dcdbas serio_raw evdev usbhid pcspkr sg hid rtc_cmos rtc_core rtc_lib
ext3 jbd mbcache ath9k(-) mac80211 ath cfg80211 rfkill led_class b44
ssb mmc_core pcmcia ipw2100 yenta_socket rsrc_nonstatic libipw
uhci_hcd mii lib80211 pcmcia_core sr_mod sd_mod cdrom ehci_hcd usbcore
ata_piix ata_generic pata_acpi libata scsi_mod

Pid: 2089, comm: rmmod Tainted: G        W  (2.6.32-ARCH #1) Inspiron
9200
EIP: 0060:[<c116de6a>] EFLAGS: 00010202 CPU: 0
EIP is at kref_put+0x1a/0x60
EAX: 0000000c EBX: 0000000c ECX: f70ce360 EDX: f8444410
ESI: f8444410 EDI: f732108c EBP: f70ce260 ESP: f559fec4
 DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
Process rmmod (pid: 2089, ti=f559e000 task=f54edd50 task.ti=f559e000)
Stack:
 f70ce260 f84ef8b0 f843814d f70cea54 f70cea54 f70cea54 f84cc40d f70cea54
<0> f84ef8b0 f732108c f84ef8e4 f84cc4d8 f7321000 c11832f6 f7321058 c11f631d
<0> f7321058 f84ef8b0 c11f6407 f84ef880 f84ef8b0 c13c4c80 c11f55eb f84ef880
Call Trace:
 [<f843814d>] ? ieee80211_unregister_hw+0x5d/0xd0 [mac80211]
 [<f84cc40d>] ? ath_detach+0x5d/0x120 [ath9k]
 [<f84cc4d8>] ? ath_cleanup+0x8/0x30 [ath9k]
 [<c11832f6>] ? pci_device_remove+0x16/0x40
 [<c11f631d>] ? __device_release_driver+0x4d/0xb0
 [<c11f6407>] ? driver_detach+0x87/0x90
 [<c11f55eb>] ? bus_remove_driver+0x5b/0xa0
 [<c118353b>] ? pci_unregister_driver+0x2b/0x80
 [<f84d431c>] ? ath9k_exit+0x8/0x31 [ath9k]
 [<c1070ef0>] ? sys_delete_module+0x190/0x240
 [<c10c6b47>] ? remove_vma+0x37/0x50
 [<c1025a92>] ? do_page_fault+0x102/0x300
 [<c1003913>] ? sysenter_do_call+0x12/0x28
Code: 00 c7 00 01 00 00 00 0f ae f0 89 f6 c3 8d 74 26 00 83 ec 08 85
d2 89 1c 24 89 c3 89 74 24 04 89 d6 74 39 81 fa a0 86 0d c1 74 20 <3e>
ff 0b 0f 94 c2 31 c0 84 d2 74 09 89 d8 ff d6 b8 01 00 00 00
EIP: [<c116de6a>] kref_put+0x1a/0x60 SS:ESP 0068:f559fec4
CR2: 000000000000000c
---[ end trace a7919e7f17c0a727 ]---

Please advise.  Thank you!

Russ Adams

Re: modprobe fail and subsequent crash on rmmod ath9k D-Link DWA 652

[Russ Adams]

Nothing new on this just yet.

Tonight I will try 2.6.32-2 to see if that fixes the problem.  Does
anyone have any debugging steps I can try?

On Sat, Dec 19, 2009 at 10:27 AM, Russ Adams <russ.adams@gmail.com> wrote:
> Hello all,
>
> Environment:
>
> I'm running 2.6.32 with a set of patches provided by Arch Linux:
> ftp://ftp.archlinux.org/other/kernel26/patch-2.6.32.1-1-ARCH.bz2.  The
> patches appear mostly cosmetic, with some changes applied to ext4.
>
> Hardware:
>
> I have a D-Link DWA 652 which reports as an AR5008 in lspci -vnn.  The
> card is functional in kernel 2.6.28. (I'm using it to post this email)
>
> The ath9k list indicated that the problem in rmmod may be in the
> mac80211b module.
>
> Here's the sequence of events:
>
> % sudo modprobe ath9k # no pertinent dmesg output
>
> % iwconfig # wlan0 not present
> lo        no wireless extensions.
>
> eth0      no wireless extensions.
>
> % sudo rmmod ath9k # not sure why it's telling me "Killed" here
> Killed
>
> % dmesg | less
> ...
> ...
> BUG: unable to handle kernel NULL pointer dereference at 0000000c
> IP: [<c116de6a>] kref_put+0x1a/0x60
> *pde = 00000000
> Oops: 0002 [#1] PREEMPT SMP
> last sysfs file:
> /sys/devices/LNXSYSTM:00/LNXSYBUS:00/PNP0C0A:00/power_supply/BAT0/energy_full
> Modules linked in: ipv6 fan nls_cp437 hid_microsoft vfat fat
> snd_seq_dummy firewire_ohci mmc_block snd_seq_oss firewire_core
> snd_seq_midi_event snd_seq snd_seq_device snd_pcm_oss sdhci_pci
> snd_intel8x0 radeon snd_intel8x0m crc_itu_t ttm snd_ac97_codec sdhci
> drm_kms_helper snd_pcm drm i2c_algo_bit snd_timer intel_agp
> snd_mixer_oss video snd ac97_bus joydev iTCO_wdt i2c_core
> iTCO_vendor_support soundcore shpchp output dell_laptop ac battery
> button psmouse snd_page_alloc agpgart pci_hotplug processor thermal
> dcdbas serio_raw evdev usbhid pcspkr sg hid rtc_cmos rtc_core rtc_lib
> ext3 jbd mbcache ath9k(-) mac80211 ath cfg80211 rfkill led_class b44
> ssb mmc_core pcmcia ipw2100 yenta_socket rsrc_nonstatic libipw
> uhci_hcd mii lib80211 pcmcia_core sr_mod sd_mod cdrom ehci_hcd usbcore
> ata_piix ata_generic pata_acpi libata scsi_mod
>
> Pid: 2089, comm: rmmod Tainted: G        W  (2.6.32-ARCH #1) Inspiron
> 9200
> EIP: 0060:[<c116de6a>] EFLAGS: 00010202 CPU: 0
> EIP is at kref_put+0x1a/0x60
> EAX: 0000000c EBX: 0000000c ECX: f70ce360 EDX: f8444410
> ESI: f8444410 EDI: f732108c EBP: f70ce260 ESP: f559fec4
>  DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
> Process rmmod (pid: 2089, ti=f559e000 task=f54edd50 task.ti=f559e000)
> Stack:
>  f70ce260 f84ef8b0 f843814d f70cea54 f70cea54 f70cea54 f84cc40d f70cea54
> <0> f84ef8b0 f732108c f84ef8e4 f84cc4d8 f7321000 c11832f6 f7321058 c11f631d
> <0> f7321058 f84ef8b0 c11f6407 f84ef880 f84ef8b0 c13c4c80 c11f55eb f84ef880
> Call Trace:
>  [<f843814d>] ? ieee80211_unregister_hw+0x5d/0xd0 [mac80211]
>  [<f84cc40d>] ? ath_detach+0x5d/0x120 [ath9k]
>  [<f84cc4d8>] ? ath_cleanup+0x8/0x30 [ath9k]
>  [<c11832f6>] ? pci_device_remove+0x16/0x40
>  [<c11f631d>] ? __device_release_driver+0x4d/0xb0
>  [<c11f6407>] ? driver_detach+0x87/0x90
>  [<c11f55eb>] ? bus_remove_driver+0x5b/0xa0
>  [<c118353b>] ? pci_unregister_driver+0x2b/0x80
>  [<f84d431c>] ? ath9k_exit+0x8/0x31 [ath9k]
>  [<c1070ef0>] ? sys_delete_module+0x190/0x240
>  [<c10c6b47>] ? remove_vma+0x37/0x50
>  [<c1025a92>] ? do_page_fault+0x102/0x300
>  [<c1003913>] ? sysenter_do_call+0x12/0x28
> Code: 00 c7 00 01 00 00 00 0f ae f0 89 f6 c3 8d 74 26 00 83 ec 08 85
> d2 89 1c 24 89 c3 89 74 24 04 89 d6 74 39 81 fa a0 86 0d c1 74 20 <3e>
> ff 0b 0f 94 c2 31 c0 84 d2 74 09 89 d8 ff d6 b8 01 00 00 00
> EIP: [<c116de6a>] kref_put+0x1a/0x60 SS:ESP 0068:f559fec4
> CR2: 000000000000000c
> ---[ end trace a7919e7f17c0a727 ]---
>
> Please advise.  Thank you!
>
> Russ Adams
>

Re: modprobe fail and subsequent crash on rmmod ath9k D-Link DWA 652

[Pavel Roskin]

On Wed, 2009-12-23 at 12:11 -0500, Russ Adams wrote:
> Nothing new on this just yet.
> 
> Tonight I will try 2.6.32-2 to see if that fixes the problem.  Does
> anyone have any debugging steps I can try?

Remember, I asked you to show the debug messages after the ath9k module
is loaded.  If it's loaded on startup, you can blacklist it and then
load manually with modprobe.

It seems to me that something fails during the initialization, which
causes a crash at the unload time.  But since ath9k is working properly
for most of us, it's hard to figure out what is failing and how to model
the problem on functioning hardware.

-- 
Regards,
Pavel Roskin

Re: modprobe fail and subsequent crash on rmmod ath9k D-Link DWA 652

[Luis R. Rodriguez]

On Wed, Dec 23, 2009 at 9:30 AM, Pavel Roskin <proski@gnu.org> wrote:
> On Wed, 2009-12-23 at 12:11 -0500, Russ Adams wrote:
>> Nothing new on this just yet.
>>
>> Tonight I will try 2.6.32-2 to see if that fixes the problem.  Does
>> anyone have any debugging steps I can try?
>
> Remember, I asked you to show the debug messages after the ath9k module
> is loaded.  If it's loaded on startup, you can blacklist it and then
> load manually with modprobe.
>
> It seems to me that something fails during the initialization, which
> causes a crash at the unload time.  But since ath9k is working properly
> for most of us, it's hard to figure out what is failing and how to model
> the problem on functioning hardware.

Also, please see:

http://wireless.kernel.org/en/users/Documentation/Reporting_bugs

specifically the parts about the latest stable kernels.

 Luis

Re: modprobe fail and subsequent crash on rmmod ath9k D-Link DWA 652

[Russ Adams]

Apologies on the delay here.  I will make a priority to compile a
vanilla kernel this week and do suggested debugging steps against it.

On Wed, Dec 23, 2009 at 12:33 PM, Luis R. Rodriguez <mcgrof@gmail.com> wrote:
> On Wed, Dec 23, 2009 at 9:30 AM, Pavel Roskin <proski@gnu.org> wrote:
>> On Wed, 2009-12-23 at 12:11 -0500, Russ Adams wrote:
>>> Nothing new on this just yet.
>>>
>>> Tonight I will try 2.6.32-2 to see if that fixes the problem.  Does
>>> anyone have any debugging steps I can try?
>>
>> Remember, I asked you to show the debug messages after the ath9k module
>> is loaded.  If it's loaded on startup, you can blacklist it and then
>> load manually with modprobe.
>>
>> It seems to me that something fails during the initialization, which
>> causes a crash at the unload time.  But since ath9k is working properly
>> for most of us, it's hard to figure out what is failing and how to model
>> the problem on functioning hardware.
>
> Also, please see:
>
> http://wireless.kernel.org/en/users/Documentation/Reporting_bugs
>
> specifically the parts about the latest stable kernels.
>
>  Luis
>

Re: modprobe fail and subsequent crash on rmmod ath9k D-Link DWA 652

[Russ Adams]

I've recompiled a vanilla kernel, 2.6.32.3 with debug for ATH9K module
enabled.  I found two main things:

1. When I booted, ath9k was being inserted without the debug parameter
via my initrd.

  * Subsequent calls to rmmod ath9k would trigger the Killed message.
  * Manually inserting the ath9k driver with debug symbol (modprobe
debug=0xffffffff) would fail with no error in dmesg.

2. If I explicitly disabled ath9k by adding disablemodules=ath9k to my
kernel arguments in /boot/grub/menu.lst ath9k is *not* inserted
automatically at boot. (or for that matter if I simply physically
removed the card at boot).  Now with debug symbols on or off, the
driver correctly installs itself and functions properly thereafter.
rmmod ath9k does not cause the Killed message.

Now I'm suspicious that my initrd was loading ath9k before some
dependency was filled.  Any thoughts in that direction?

Below is the dmesg output when rmmod ath9k fails:

BUG: unable to handle kernel NULL pointer dereference at 0000000c
IP: [<c117338a>] kref_put+0x1a/0x60
*pde = 00000000
Oops: 0002 [#1] PREEMPT SMP
last sysfs file:
/sys/devices/pci0000:00/0000:00:1f.1/host0/target0:0:0/0:0:0:0/block/sda/size
Modules linked in: thermal processor fan button battery ac mmc_block
sdhci_pci sdhci snd_intel8x0m snd_intel8x0 snd_ac97_codec snd_pcm_oss
snd_pcm snd_timer snd_page_alloc snd_mixer_oss snd soundcore ac97_bus
psmouse evdev rtc_cmos rtc_core rtc_lib ext3 jbd mbcache hid_microsoft
ath9k(-) usbhid hid mac80211 ath cfg80211 rfkill led_class b44 ssb
mmc_core pcmcia ipw2100 libipw uhci_hcd yenta_socket rsrc_nonstatic
lib80211 pcmcia_core mii sr_mod sd_mod cdrom ehci_hcd usbcore ata_piix
ata_generic pata_acpi libata scsi_mod

Pid: 1790, comm: rmmod Not tainted (2.6.32.3-ARCH #1) Inspiron 9200
EIP: 0060:[<c117338a>] EFLAGS: 00010202 CPU: 0
EIP is at kref_put+0x1a/0x60
EAX: 0000000c EBX: 0000000c ECX: f695c360 EDX: f84504c0
ESI: f84504c0 EDI: f732188c EBP: f695c260 ESP: f622feb8
DS: 007b ES: 007b FS: 00d8 GS: 00e0 SS: 0068
Process rmmod (pid: 1790, ti=f622e000 task=f61025b0 task.ti=f622e000)
Stack:
f695c260 f858c910 f844414d f695ca54 f695ca54 f695ca54 f85640ed f695ca54
<0> 00000200 f858b960 f695ca54 f858c910 f732188c f858c944 f85641c8 f7321800
<0> c11887b6 f7321858 c11fb94d f7321858 f858c910 c11fba37 f858c8e0 f858c910
Call Trace:
[<f844414d>] ? ieee80211_unregister_hw+0x5d/0xd0 [mac80211]
[<f85640ed>] ? ath_detach+0x7d/0x150 [ath9k]
[<f85641c8>] ? ath_cleanup+0x8/0x30 [ath9k]
[<c11887b6>] ? pci_device_remove+0x16/0x40
[<c11fb94d>] ? __device_release_driver+0x4d/0xb0
[<c11fba37>] ? driver_detach+0x87/0x90
[<c11fac1b>] ? bus_remove_driver+0x5b/0xa0
[<c11889fb>] ? pci_unregister_driver+0x2b/0x80
[<f856e298>] ? ath9k_exit+0x8/0x36 [ath9k]
[<c1072c30>] ? sys_delete_module+0x190/0x240
[<c10cb9e7>] ? remove_vma+0x37/0x50
[<c1025bd1>] ? do_page_fault+0x111/0x330
[<c10039f3>] ? sysenter_do_call+0x12/0x28
Code: 00 c7 00 01 00 00 00 0f ae f0 89 f6 c3 8d 74 26 00 83 ec 08 85
d2 89 1c 24 89 c3 89 74 24 04 89 d6 74 39 81 fa 50 d5 0d c1 74 20 <3e>
ff 0b 0f 94 c2 31 c0 84 d2 74 09 89 d8 ff d6 b8 01 00 00 00
EIP: [<c117338a>] kref_put+0x1a/0x60 SS:ESP 0068:f622feb8
CR2: 000000000000000c
---[ end trace 4ce8923552844d86 ]---


On Tue, Jan 12, 2010 at 12:58 PM, Russ Adams <russ.adams@gmail.com> wrote:
> Apologies on the delay here.  I will make a priority to compile a
> vanilla kernel this week and do suggested debugging steps against it.
>
> On Wed, Dec 23, 2009 at 12:33 PM, Luis R. Rodriguez <mcgrof@gmail.com> wrote:
>> On Wed, Dec 23, 2009 at 9:30 AM, Pavel Roskin <proski@gnu.org> wrote:
>>> On Wed, 2009-12-23 at 12:11 -0500, Russ Adams wrote:
>>>> Nothing new on this just yet.
>>>>
>>>> Tonight I will try 2.6.32-2 to see if that fixes the problem.  Does
>>>> anyone have any debugging steps I can try?
>>>
>>> Remember, I asked you to show the debug messages after the ath9k module
>>> is loaded.  If it's loaded on startup, you can blacklist it and then
>>> load manually with modprobe.
>>>
>>> It seems to me that something fails during the initialization, which
>>> causes a crash at the unload time.  But since ath9k is working properly
>>> for most of us, it's hard to figure out what is failing and how to model
>>> the problem on functioning hardware.
>>
>> Also, please see:
>>
>> http://wireless.kernel.org/en/users/Documentation/Reporting_bugs
>>
>> specifically the parts about the latest stable kernels.
>>
>>  Luis
>>
>