Using compat-wireless w/ 2.6.27.26

Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

I'm the lead developer on the Astlinux project, which I might have
mentioned in a previous posting.  This means that I have my hands in
many pots, so I'm not an expert in the latest Linux kernel changes... be
patient with me.  Hard to keep track of 200+ individual projects.

I'm building 2.6.26.27 as I said for (amongst other things) an Alix 2D3
(GeodeLX) with EMP-8602 (AR-5413) as a test platform.

My kernel gets built with:

#
# Wireless
#
CONFIG_CFG80211=m
CONFIG_NL80211=y
CONFIG_WIRELESS_EXT=y
# CONFIG_WIRELESS_EXT_SYSFS is not set
CONFIG_MAC80211=m

#
# Rate control algorithm selection
#
CONFIG_MAC80211_RC_PID=y
CONFIG_MAC80211_RC_DEFAULT_PID=y
CONFIG_MAC80211_RC_DEFAULT="pid"
# CONFIG_MAC80211_MESH is not set
CONFIG_MAC80211_LEDS=y
# CONFIG_MAC80211_DEBUG_MENU is not set
# CONFIG_IEEE80211 is not set
CONFIG_RFKILL=m
CONFIG_RFKILL_INPUT=m
CONFIG_RFKILL_LEDS=y
# CONFIG_NET_9P is not set


I build compat-wireless-2.6.30 into /lib/modules/2.6.27.26-astlinux/kernel/... as a Kbuild.

The system also has iw-0.9.15 and hostapd-0.6.9.

All of this is fairly straightforward.

On boot, our /etc/init.d scripts do the following:

+ modprobe ath5k
ath5k phy0: Atheros AR5413 chip found (MAC: 0xa4, PHY: 0x61)

+ iw dev wlan0 interface add ap0 type managed
+ ip link set ap0 up
+ iw dev ap0 connect xxxx
command failed: Operation not supported (-95)



So doing some sanity checks by hand I get:

pbx ~ # lsmod | head
Module                  Size  Used by
ath5k                 104772  0 
mac80211              153304  1 ath5k
cfg80211               54584  2 ath5k,mac80211
binfmt_misc             7112  1 
aes_i586                7456  0 
aes_generic            28968  1 aes_i586
lm90                   11944  0 
hwmon                   2324  1 lm90
scx200_acb              4356  0 
pbx ~ # iw dev ap0 info
Interface ap0
	ifindex 9
	type managed
pbx ~ # iw phy phy0 info
command failed: No buffer space available (-105)
pbx ~ # 


Not sure why the "phy ... info" command fails.

pbx ~ # hostapd -v
hostapd v0.6.9
User space daemon for IEEE 802.11 AP management,
IEEE 802.1X/WPA/WPA2/EAP/RADIUS Authenticator
Copyright (c) 2002-2009, Jouni Malinen <j@w1.fi> and contributors
pbx ~ # cat /etc/hostapd.conf
# automatically generated. do not edit!

logger_syslog=-1
logger_syslog_level=2
logger_stdout=--1
logger_stdout_level=2

debug=0

ctrl_interface_group=0

# should this be per-ssid?
macaddr_acl=0

dump_file=/tmp/hostapd.dump

# should this be an option?
hw_mode=b

bridge=br1
interface=ap0
driver=nl80211
ssid=xxxxx
channel=0

auth_algs=1

wpa_passphrase=yyyyy
wpa=2
wpa_key_mgmt=WPA-PSK
wpa_pairwise=TKIP
rsn_pairwise=CCMP

ignore_broadcast_ssid=1

wpa_strict_rekey=1
wpa_group_rekey=600
wpa_gmk_rekey=14400

pbx ~ # ip link show ap0
9: ap0: <BROADCAST,MULTICAST> mtu 1500 qdisc pfifo_fast state DOWN qlen 1000
    link/ether 00:02:6f:5a:d2:2b brd ff:ff:ff:ff:ff:ff
pbx ~ # brctl show
bridge name	bridge id		STP enabled	interfaces
br1		8000.00026f5ad22b	no		ap0
							eth0
pbx ~ # 
pbx ~ # hostapd -dddd -P /var/run/hostapd.pid /etc/hostapd.conf
Configuration file: /etc/hostapd.conf
Line 8: DEPRECATED: 'debug' configuration variable is not used anymore
ctrl_interface_group=0
Failed to set interface ap0 to master mode.
nl80211 driver initialization failed.
ap0: Unable to setup interface.
ELOOP: remaining socket: sock=5 eloop_data=0x8098028 user_data=(nil) handler=0x806f503
pbx ~ # 
pbx ~ # ldd /usr/bin/hostapd
	libnl.so.1 => /usr/lib/libnl.so.1 (0xb7f04000)
	libcrypto.so.0.9.8 => /lib/libcrypto.so.0.9.8 (0xb7df8000)
	libgcc_s.so.1 => /lib/libgcc_s.so.1 (0xb7def000)
	libc.so.0 => /lib/libc.so.0 (0xb7da5000)
	libm.so.0 => /lib/libm.so.0 (0xb7d97000)
	libdl.so.0 => /lib/libdl.so.0 (0xb7d94000)
	ld-uClibc.so.0 => /lib/ld-uClibc.so.0 (0xb7f3d000)
pbx ~ # 


And on the host that this was cross-compiled:

[philipp@builder ~/kernel]$ cat build_i586/hostapd-0.6.9/hostapd/.config 
CFLAGS+=-Os -Wall -Os -pipe -fomit-frame-pointer -march=k6-2 -fno-align-functions -fno-align-loops -fno-align-jumps -fno-align-labels 
CFLAGS+= -DUSE_KERNEL_HEADERS -I/home/philipp/kernel/build_i586/linux-2.6.27.26-astlinux/include
CONFIG_IEEE80211N=y
CONFIG_DRIVER_NL80211=y
CFLAGS+=-I/home/philipp/kernel/build_i586/staging_dir/usr/include
LIBS+=-L/home/philipp/kernel/build_i586/root/usr/lib
CONFIG_DRIVER_NONE=y
CONFIG_DRIVER_WIRED=y
CONFIG_IAPP=y
CONFIG_RSN_PREAUTH=y
CONFIG_PEERKEY=y
CONFIG_EAP=y
CONFIG_EAP_MD5=y
CONFIG_EAP_MSCHAPV2=y
CONFIG_EAP_GTC=y
CONFIG_EAP_SIM=y
CONFIG_EAP_AKA=y
CONFIG_EAP_AKA_PRIME=y
CONFIG_EAP_PAX=y
CONFIG_EAP_PSK=y
CONFIG_EAP_SAKE=y
CONFIG_EAP_GPSK=y
CONFIG_EAP_GPSK_SHA256=y
CONFIG_EAP_IKEV2=y
CONFIG_PKCS12=y
CONFIG_IPV6=y
CONFIG_IEEE80211R=y
CONFIG_IEEE80211N=y
[philipp@builder ~/kernel]$ 



What am I missing?

How should I go about debugging this?

Thanks,

-Philip

Re: Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

Errr...  building 2.6.27.26 rather, as per the title...


Philip A. Prindeville wrote:
> I'm the lead developer on the Astlinux project, which I might have
> mentioned in a previous posting.  This means that I have my hands in
> many pots, so I'm not an expert in the latest Linux kernel changes... be
> patient with me.  Hard to keep track of 200+ individual projects.
>
> I'm building 2.6.26.27 as I said for (amongst other things) an Alix 2D3
> (GeodeLX) with EMP-8602 (AR-5413) as a test platform.
>
> My kernel gets built with:
>
> #
> # Wireless
> #
> CONFIG_CFG80211=m
> CONFIG_NL80211=y
> CONFIG_WIRELESS_EXT=y
> # CONFIG_WIRELESS_EXT_SYSFS is not set
> CONFIG_MAC80211=m
>
> #
> # Rate control algorithm selection
> #
> CONFIG_MAC80211_RC_PID=y
> CONFIG_MAC80211_RC_DEFAULT_PID=y
> CONFIG_MAC80211_RC_DEFAULT="pid"
> # CONFIG_MAC80211_MESH is not set
> CONFIG_MAC80211_LEDS=y
> # CONFIG_MAC80211_DEBUG_MENU is not set
> # CONFIG_IEEE80211 is not set
> CONFIG_RFKILL=m
> CONFIG_RFKILL_INPUT=m
> CONFIG_RFKILL_LEDS=y
> # CONFIG_NET_9P is not set
>
>
> I build compat-wireless-2.6.30 into /lib/modules/2.6.27.26-astlinux/kernel/... as a Kbuild.
>
> The system also has iw-0.9.15 and hostapd-0.6.9.
>
> All of this is fairly straightforward.
>
> On boot, our /etc/init.d scripts do the following:
>
> + modprobe ath5k
> ath5k phy0: Atheros AR5413 chip found (MAC: 0xa4, PHY: 0x61)
>
> + iw dev wlan0 interface add ap0 type managed
> + ip link set ap0 up
> + iw dev ap0 connect xxxx
> command failed: Operation not supported (-95)
>
>
>
> So doing some sanity checks by hand I get:
>
> pbx ~ # lsmod | head
> Module                  Size  Used by
> ath5k                 104772  0 
> mac80211              153304  1 ath5k
> cfg80211               54584  2 ath5k,mac80211
> binfmt_misc             7112  1 
> aes_i586                7456  0 
> aes_generic            28968  1 aes_i586
> lm90                   11944  0 
> hwmon                   2324  1 lm90
> scx200_acb              4356  0 
> pbx ~ # iw dev ap0 info
> Interface ap0
> 	ifindex 9
> 	type managed
> pbx ~ # iw phy phy0 info
> command failed: No buffer space available (-105)
> pbx ~ # 
>
>
> Not sure why the "phy ... info" command fails.
>
> pbx ~ # hostapd -v
> hostapd v0.6.9
> User space daemon for IEEE 802.11 AP management,
> IEEE 802.1X/WPA/WPA2/EAP/RADIUS Authenticator
> Copyright (c) 2002-2009, Jouni Malinen <j@w1.fi> and contributors
> pbx ~ # cat /etc/hostapd.conf
> # automatically generated. do not edit!
>
> logger_syslog=-1
> logger_syslog_level=2
> logger_stdout=--1
> logger_stdout_level=2
>
> debug=0
>
> ctrl_interface_group=0
>
> # should this be per-ssid?
> macaddr_acl=0
>
> dump_file=/tmp/hostapd.dump
>
> # should this be an option?
> hw_mode=b
>
> bridge=br1
> interface=ap0
> driver=nl80211
> ssid=xxxxx
> channel=0
>
> auth_algs=1
>
> wpa_passphrase=yyyyy
> wpa=2
> wpa_key_mgmt=WPA-PSK
> wpa_pairwise=TKIP
> rsn_pairwise=CCMP
>
> ignore_broadcast_ssid=1
>
> wpa_strict_rekey=1
> wpa_group_rekey=600
> wpa_gmk_rekey=14400
>
> pbx ~ # ip link show ap0
> 9: ap0: <BROADCAST,MULTICAST> mtu 1500 qdisc pfifo_fast state DOWN qlen 1000
>     link/ether 00:02:6f:5a:d2:2b brd ff:ff:ff:ff:ff:ff
> pbx ~ # brctl show
> bridge name	bridge id		STP enabled	interfaces
> br1		8000.00026f5ad22b	no		ap0
> 							eth0
> pbx ~ # 
> pbx ~ # hostapd -dddd -P /var/run/hostapd.pid /etc/hostapd.conf
> Configuration file: /etc/hostapd.conf
> Line 8: DEPRECATED: 'debug' configuration variable is not used anymore
> ctrl_interface_group=0
> Failed to set interface ap0 to master mode.
> nl80211 driver initialization failed.
> ap0: Unable to setup interface.
> ELOOP: remaining socket: sock=5 eloop_data=0x8098028 user_data=(nil) handler=0x806f503
> pbx ~ # 
> pbx ~ # ldd /usr/bin/hostapd
> 	libnl.so.1 => /usr/lib/libnl.so.1 (0xb7f04000)
> 	libcrypto.so.0.9.8 => /lib/libcrypto.so.0.9.8 (0xb7df8000)
> 	libgcc_s.so.1 => /lib/libgcc_s.so.1 (0xb7def000)
> 	libc.so.0 => /lib/libc.so.0 (0xb7da5000)
> 	libm.so.0 => /lib/libm.so.0 (0xb7d97000)
> 	libdl.so.0 => /lib/libdl.so.0 (0xb7d94000)
> 	ld-uClibc.so.0 => /lib/ld-uClibc.so.0 (0xb7f3d000)
> pbx ~ # 
>
>
> And on the host that this was cross-compiled:
>
> [philipp@builder ~/kernel]$ cat build_i586/hostapd-0.6.9/hostapd/.config 
> CFLAGS+=-Os -Wall -Os -pipe -fomit-frame-pointer -march=k6-2 -fno-align-functions -fno-align-loops -fno-align-jumps -fno-align-labels 
> CFLAGS+= -DUSE_KERNEL_HEADERS -I/home/philipp/kernel/build_i586/linux-2.6.27.26-astlinux/include
> CONFIG_IEEE80211N=y
> CONFIG_DRIVER_NL80211=y
> CFLAGS+=-I/home/philipp/kernel/build_i586/staging_dir/usr/include
> LIBS+=-L/home/philipp/kernel/build_i586/root/usr/lib
> CONFIG_DRIVER_NONE=y
> CONFIG_DRIVER_WIRED=y
> CONFIG_IAPP=y
> CONFIG_RSN_PREAUTH=y
> CONFIG_PEERKEY=y
> CONFIG_EAP=y
> CONFIG_EAP_MD5=y
> CONFIG_EAP_MSCHAPV2=y
> CONFIG_EAP_GTC=y
> CONFIG_EAP_SIM=y
> CONFIG_EAP_AKA=y
> CONFIG_EAP_AKA_PRIME=y
> CONFIG_EAP_PAX=y
> CONFIG_EAP_PSK=y
> CONFIG_EAP_SAKE=y
> CONFIG_EAP_GPSK=y
> CONFIG_EAP_GPSK_SHA256=y
> CONFIG_EAP_IKEV2=y
> CONFIG_PKCS12=y
> CONFIG_IPV6=y
> CONFIG_IEEE80211R=y
> CONFIG_IEEE80211N=y
> [philipp@builder ~/kernel]$ 
>
>
>
> What am I missing?
>
> How should I go about debugging this?
>
> Thanks,
>
> -Philip
>   

Re: Using compat-wireless w/ 2.6.27.26

[Bob Copeland]

On Fri, Jul 17, 2009 at 9:27 PM, Philip A.
Prindeville<philipp_subx@redfish-solutions.com> wrote:
> + modprobe ath5k
> ath5k phy0: Atheros AR5413 chip found (MAC: 0xa4, PHY: 0x61)
>
> + iw dev wlan0 interface add ap0 type managed
> + ip link set ap0 up
> + iw dev ap0 connect xxxx
> command failed: Operation not supported (-95)

Why are you adding another interface?  Should be able to use "iw dev wlan0..."

> pbx ~ # iw phy phy0 info
> command failed: No buffer space available (-105)

Means the netlink buffer was exceeded.  In the past this was because there
were too many channels -- are you sure you're loading the compat-wireless
ath5k instead of the .29 one (and not using all_channels modparam)?

-- 
Bob Copeland %% www.bobcopeland.com

Re: Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

Bob Copeland wrote:
> On Fri, Jul 17, 2009 at 9:27 PM, Philip A.
> Prindeville<philipp_subx@redfish-solutions.com> wrote:
>> + modprobe ath5k
>> ath5k phy0: Atheros AR5413 chip found (MAC: 0xa4, PHY: 0x61)
>>
>> + iw dev wlan0 interface add ap0 type managed
>> + ip link set ap0 up
>> + iw dev ap0 connect xxxx
>> command failed: Operation not supported (-95)
> 
> Why are you adding another interface?  Should be able to use "iw dev wlan0..."

Because I might want to run two access points (two SSIDs, anyway) on the same radio...  one with WEP on one VLAN (for Wifi SIP handsets), and one with WPA-PSK2 on another VLAN (for laptops, etc).

>> pbx ~ # iw phy phy0 info
>> command failed: No buffer space available (-105)
> 
> Means the netlink buffer was exceeded.  In the past this was because there
> were too many channels -- are you sure you're loading the compat-wireless
> ath5k instead of the .29 one (and not using all_channels modparam)?
> 

Reasonably sure, yes.  I stopped the distro build and did an "ls -R" after building linux, and there were no drivers other than cfg80211.ko and mac80211.ko that get duplicated.

After resuming the rest of the build, there were drivers from compat-wireless including ath5k, and cfg80211.ko and mac80211.ko also had refreshed timestamps matching everything else.

But, if you think there's still a doubt, what to look for in the modinfo?

-Philip

Re: Using compat-wireless w/ 2.6.27.26

[Bob Copeland]

On Sun, Jul 19, 2009 at 10:55:16AM -0700, Philip A. Prindeville wrote:
> Bob Copeland wrote:
> > On Fri, Jul 17, 2009 at 9:27 PM, Philip A.
> >> + iw dev wlan0 interface add ap0 type managed
> >> + ip link set ap0 up
> >> + iw dev ap0 connect xxxx
> >> command failed: Operation not supported (-95)
> > 
> > Why are you adding another interface?  Should be able to use "iw dev wlan0..."
> 
> Because I might want to run two access points (two SSIDs, anyway) on the same
> radio...  one with WEP on one VLAN (for Wifi SIP handsets), and one with
> WPA-PSK2 on another VLAN (for laptops, etc).

Hmm, I don't know if it matters, but you were creating it as a managed
interface.  Anyway I don't believe ath5k supports that yet - ath9k might.
Someone on ath5k-devel was porting the ath9k virtual wiphy stuff to
ath5k but I haven't heard anything recently.

> But, if you think there's still a doubt, what to look for in the modinfo?

If it says "depends: mac80211,cfg80211,ath" then it is probably the right
one.  It just sounds like this bug:

http://bugzilla.kernel.org/show_bug.cgi?id=12315

The fix was merged after 2.6.30, IIRC, but it should definitely be in
compat-wireless.  Also check your modules directory for an old ath5k.ko
sitting in drivers/.../wireless/ath5k.  It should now be in ath/ath5k.

HTH!

-- 
Bob Copeland %% www.bobcopeland.com

Re: Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

Bob Copeland wrote:
> On Sun, Jul 19, 2009 at 10:55:16AM -0700, Philip A. Prindeville wrote:
>> Bob Copeland wrote:
>>> On Fri, Jul 17, 2009 at 9:27 PM, Philip A.
>>>> + iw dev wlan0 interface add ap0 type managed
>>>> + ip link set ap0 up
>>>> + iw dev ap0 connect xxxx
>>>> command failed: Operation not supported (-95)
>>> Why are you adding another interface?  Should be able to use "iw dev wlan0..."
>> Because I might want to run two access points (two SSIDs, anyway) on the same
>> radio...  one with WEP on one VLAN (for Wifi SIP handsets), and one with
>> WPA-PSK2 on another VLAN (for laptops, etc).
> 
> Hmm, I don't know if it matters, but you were creating it as a managed
> interface.  Anyway I don't believe ath5k supports that yet - ath9k might.
> Someone on ath5k-devel was porting the ath9k virtual wiphy stuff to
> ath5k but I haven't heard anything recently.
> 
>> But, if you think there's still a doubt, what to look for in the modinfo?
> 
> If it says "depends: mac80211,cfg80211,ath" then it is probably the right
> one.  It just sounds like this bug:
> 
> http://bugzilla.kernel.org/show_bug.cgi?id=12315

Actually, it's saying:

filename:       /lib/modules/2.6.27.26-astlinux/kernel/drivers/net/wireless/ath5k/ath5k.ko
version:        0.6.0 (EXPERIMENTAL)
license:        Dual BSD/GPL
description:    Support for 5xxx series of Atheros 802.11 wireless LAN cards.
author:         Nick Kossifidis
author:         Jiri Slaby
srcversion:     E33724BC42EBC4E7E7714E1
...
depends:        mac80211,cfg80211
vermagic:       2.6.27.26-astlinux preempt mod_unload modversions GEODE
parm:           nohwcrypt:Disable hardware encryption. (int)

"nohwcrypt" is present in the 2.6.30 wireless-compat sources but not in the 2.6.27 drivers, so I'm pretty sure we're looking at the right ones.

And the 0.6.0 version number matches ath5k/base.c in the compat-wireless sources as well.

Not seeing the dependency on "ath", which is strange.  Is that common code that's shared in 2.6.30 for the Atheros drivers?

> The fix was merged after 2.6.30, IIRC, but it should definitely be in
> compat-wireless.  Also check your modules directory for an old ath5k.ko
> sitting in drivers/.../wireless/ath5k.  It should now be in ath/ath5k.
> 
> HTH!
> 

Re: Using compat-wireless w/ 2.6.27.26

[Bob Copeland]

On Sun, Jul 19, 2009 at 08:05:47PM -0700, Philip A. Prindeville wrote:
> Actually, it's saying:
> 
> filename:       /lib/modules/2.6.27.26-astlinux/kernel/drivers/net/wireless/ath5k/ath5k.ko
> depends:        mac80211,cfg80211
> 
> And the 0.6.0 version number matches ath5k/base.c in the compat-wireless
> sources as well.

Yeah, ignore the module version number, no one ever changes it.

> Not seeing the dependency on "ath", which is strange.  Is that common code
> that's shared in 2.6.30 for the Atheros drivers?

In 2.6.31, yes.

> > The fix was merged after 2.6.30, IIRC, but it should definitely be in
> > compat-wireless.  Also check your modules directory for an old ath5k.ko
> > sitting in drivers/.../wireless/ath5k.  It should now be in ath/ath5k.

See above, the fix was post-2.6.30, however compat-wireless should have
it.  The module should now be in drivers/net/wireless/ath/ath5k, not
drivers/net/wireless/ath5k.

-- 
Bob Copeland %% www.bobcopeland.com

Re: Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

Bob Copeland wrote:
> On Sun, Jul 19, 2009 at 08:05:47PM -0700, Philip A. Prindeville wrote:
>> Actually, it's saying:
>>
>> filename:       /lib/modules/2.6.27.26-astlinux/kernel/drivers/net/wireless/ath5k/ath5k.ko
>> depends:        mac80211,cfg80211
>>
>> And the 0.6.0 version number matches ath5k/base.c in the compat-wireless
>> sources as well.
> 
> Yeah, ignore the module version number, no one ever changes it.
> 
>> Not seeing the dependency on "ath", which is strange.  Is that common code
>> that's shared in 2.6.30 for the Atheros drivers?
> 
> In 2.6.31, yes.
> 
>>> The fix was merged after 2.6.30, IIRC, but it should definitely be in
>>> compat-wireless.  Also check your modules directory for an old ath5k.ko
>>> sitting in drivers/.../wireless/ath5k.  It should now be in ath/ath5k.
> 
> See above, the fix was post-2.6.30, however compat-wireless should have
> it.  The module should now be in drivers/net/wireless/ath/ath5k, not
> drivers/net/wireless/ath5k.
> 

I'm using the tarball out of:

http://www.orbit-lab.org/kernel/compat-wireless-2.6-stable/v2.6.30

perhaps I should be using a different one?  I tried 2.6.31-rc1, but it doesn't build against 2.6.27.

-Philip

Re: Using compat-wireless w/ 2.6.27.26

[Luis R. Rodriguez]

On Mon, Jul 20, 2009 at 12:16 PM, Philip A.
Prindeville<philipp_subx@redfish-solutions.com> wrote:
> Bob Copeland wrote:
>> On Sun, Jul 19, 2009 at 08:05:47PM -0700, Philip A. Prindeville wrote:
>>> Actually, it's saying:
>>>
>>> filename:       /lib/modules/2.6.27.26-astlinux/kernel/drivers/net/wireless/ath5k/ath5k.ko
>>> depends:        mac80211,cfg80211
>>>
>>> And the 0.6.0 version number matches ath5k/base.c in the compat-wireless
>>> sources as well.
>>
>> Yeah, ignore the module version number, no one ever changes it.
>>
>>> Not seeing the dependency on "ath", which is strange.  Is that common code
>>> that's shared in 2.6.30 for the Atheros drivers?
>>
>> In 2.6.31, yes.
>>
>>>> The fix was merged after 2.6.30, IIRC, but it should definitely be in
>>>> compat-wireless.  Also check your modules directory for an old ath5k.ko
>>>> sitting in drivers/.../wireless/ath5k.  It should now be in ath/ath5k.
>>
>> See above, the fix was post-2.6.30, however compat-wireless should have
>> it.  The module should now be in drivers/net/wireless/ath/ath5k, not
>> drivers/net/wireless/ath5k.
>>
>
> I'm using the tarball out of:
>
> http://www.orbit-lab.org/kernel/compat-wireless-2.6-stable/v2.6.30
>
> perhaps I should be using a different one?  I tried 2.6.31-rc1, but it doesn't build against 2.6.27.

I have to fix this, will do so in a bit, will also update it to
account for the new rc3.

  Luis

Re: Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

Luis R. Rodriguez wrote:
> On Mon, Jul 20, 2009 at 12:16 PM, Philip A.
> Prindeville<philipp_subx@redfish-solutions.com> wrote:
>> Bob Copeland wrote:
>>> On Sun, Jul 19, 2009 at 08:05:47PM -0700, Philip A. Prindeville wrote:
>>>> Actually, it's saying:
>>>>
>>>> filename:       /lib/modules/2.6.27.26-astlinux/kernel/drivers/net/wireless/ath5k/ath5k.ko
>>>> depends:        mac80211,cfg80211
>>>>
>>>> And the 0.6.0 version number matches ath5k/base.c in the compat-wireless
>>>> sources as well.
>>> Yeah, ignore the module version number, no one ever changes it.
>>>
>>>> Not seeing the dependency on "ath", which is strange.  Is that common code
>>>> that's shared in 2.6.30 for the Atheros drivers?
>>> In 2.6.31, yes.
>>>
>>>>> The fix was merged after 2.6.30, IIRC, but it should definitely be in
>>>>> compat-wireless.  Also check your modules directory for an old ath5k.ko
>>>>> sitting in drivers/.../wireless/ath5k.  It should now be in ath/ath5k.
>>> See above, the fix was post-2.6.30, however compat-wireless should have
>>> it.  The module should now be in drivers/net/wireless/ath/ath5k, not
>>> drivers/net/wireless/ath5k.
>>>
>> I'm using the tarball out of:
>>
>> http://www.orbit-lab.org/kernel/compat-wireless-2.6-stable/v2.6.30
>>
>> perhaps I should be using a different one?  I tried 2.6.31-rc1, but it doesn't build against 2.6.27.
> 
> I have to fix this, will do so in a bit, will also update it to
> account for the new rc3.
> 
>   Luis

Well, while waiting for a "blessed" (or at least, 'stable') tarball to be pushed out, I tried running 2009-06-30, and I'm seeing a lot of:

mgmt::proberesp cb
Probe Request from 00:21:29:69:3e:a0 for broadcast SSID ignored
STA 00:21:29:69:3e:a0 sent probe request for our SSID
MGMT (TX callback) fail
mgmt::proberesp cb
Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
...
STA 00:21:e9:6e:a7:4f sent probe request for our SSID
MGMT (TX callback) ACK
mgmt::proberesp cb
STA 00:21:e9:6e:a7:4f sent probe request for our SSID
MGMT (TX callback) ACK
mgmt::proberesp cb
STA 00:21:e9:6e:a7:4f sent probe request for our SSID
MGMT (TX callback) fail
mgmt::proberesp cb
STA 00:21:e9:6e:a7:4f sent probe request for our SSID
MGMT (TX callback) fail
mgmt::proberesp cb
STA 00:21:e9:6e:a7:4f sent probe request for our SSID
MGMT (TX callback) fail
mgmt::proberesp cb
STA 00:21:e9:6e:a7:4f sent probe request for our SSID
MGMT (TX callback) fail
mgmt::proberesp cb
STA 00:21:e9:6e:a7:4f sent probe request for our SSID
MGMT (TX callback) fail
mgmt::proberesp cb
STA 00:21:e9:6e:a7:4f sent probe request for our SSID
MGMT (TX callback) fail
mgmt::proberesp cb
Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
STA 00:21:29:69:3e:a0 sent probe request for our SSID
Probe Request from 00:21:29:69:3e:a0 for broadcast SSID ignored
MGMT (TX callback) ACK
mgmt::proberesp cb
STA 00:21:29:69:3e:a0 sent probe request for our SSID
MGMT (TX callback) fail
mgmt::proberesp cb



00:21:29:69:3e:a0 is a PC running Vista SP2.

00:21:e9:6e:a7:4f is an iPhone running 3.0.

The iPhone should be scanning for available networks, but it's also been explicitly configured with this particular AP (and SSID).

Is this a known issue?

-Philip

Re: Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

Philip A. Prindeville wrote:
> Luis R. Rodriguez wrote:
>   
>> On Mon, Jul 20, 2009 at 12:16 PM, Philip A.
>> Prindeville<philipp_subx@redfish-solutions.com> wrote:
>>     
>>> Bob Copeland wrote:
>>>       
>>>> On Sun, Jul 19, 2009 at 08:05:47PM -0700, Philip A. Prindeville wrote:
>>>>         
>>>>> Actually, it's saying:
>>>>>
>>>>> filename:       /lib/modules/2.6.27.26-astlinux/kernel/drivers/net/wireless/ath5k/ath5k.ko
>>>>> depends:        mac80211,cfg80211
>>>>>
>>>>> And the 0.6.0 version number matches ath5k/base.c in the compat-wireless
>>>>> sources as well.
>>>>>           
>>>> Yeah, ignore the module version number, no one ever changes it.
>>>>
>>>>         
>>>>> Not seeing the dependency on "ath", which is strange.  Is that common code
>>>>> that's shared in 2.6.30 for the Atheros drivers?
>>>>>           
>>>> In 2.6.31, yes.
>>>>
>>>>         
>>>>>> The fix was merged after 2.6.30, IIRC, but it should definitely be in
>>>>>> compat-wireless.  Also check your modules directory for an old ath5k.ko
>>>>>> sitting in drivers/.../wireless/ath5k.  It should now be in ath/ath5k.
>>>>>>             
>>>> See above, the fix was post-2.6.30, however compat-wireless should have
>>>> it.  The module should now be in drivers/net/wireless/ath/ath5k, not
>>>> drivers/net/wireless/ath5k.
>>>>
>>>>         
>>> I'm using the tarball out of:
>>>
>>> http://www.orbit-lab.org/kernel/compat-wireless-2.6-stable/v2.6.30
>>>
>>> perhaps I should be using a different one?  I tried 2.6.31-rc1, but it doesn't build against 2.6.27.
>>>       
>> I have to fix this, will do so in a bit, will also update it to
>> account for the new rc3.
>>
>>   Luis
>>     
>
> Well, while waiting for a "blessed" (or at least, 'stable') tarball to be pushed out, I tried running 2009-06-30, and I'm seeing a lot of:
>
> mgmt::proberesp cb
> Probe Request from 00:21:29:69:3e:a0 for broadcast SSID ignored
> STA 00:21:29:69:3e:a0 sent probe request for our SSID
> MGMT (TX callback) fail
> mgmt::proberesp cb
> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
> ...
> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
> MGMT (TX callback) ACK
> mgmt::proberesp cb
> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
> MGMT (TX callback) ACK
> mgmt::proberesp cb
> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
> MGMT (TX callback) fail
> mgmt::proberesp cb
> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
> MGMT (TX callback) fail
> mgmt::proberesp cb
> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
> MGMT (TX callback) fail
> mgmt::proberesp cb
> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
> MGMT (TX callback) fail
> mgmt::proberesp cb
> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
> MGMT (TX callback) fail
> mgmt::proberesp cb
> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
> MGMT (TX callback) fail
> mgmt::proberesp cb
> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
> STA 00:21:29:69:3e:a0 sent probe request for our SSID
> Probe Request from 00:21:29:69:3e:a0 for broadcast SSID ignored
> MGMT (TX callback) ACK
> mgmt::proberesp cb
> STA 00:21:29:69:3e:a0 sent probe request for our SSID
> MGMT (TX callback) fail
> mgmt::proberesp cb
>
>
>
> 00:21:29:69:3e:a0 is a PC running Vista SP2.
>
> 00:21:e9:6e:a7:4f is an iPhone running 3.0.
>
> The iPhone should be scanning for available networks, but it's also been explicitly configured with this particular AP (and SSID).
>
> Is this a known issue?
>
> -Philip
>   

A little progress:  I build 2009-07-10 with 2.6.27.26, and I can get the
Vista box to associate, but not an iPhone running 3.0...

I have traces if anyone is interested.

-Philip

Re: Using compat-wireless w/ 2.6.27.26

[Luis R. Rodriguez]

On Tue, Jul 21, 2009 at 1:10 PM, Philip A.
Prindeville<philipp_subx@redfish-solutions.com> wrote:
> Philip A. Prindeville wrote:
>> Luis R. Rodriguez wrote:
>>
>>> On Mon, Jul 20, 2009 at 12:16 PM, Philip A.
>>> Prindeville<philipp_subx@redfish-solutions.com> wrote:
>>>
>>>> Bob Copeland wrote:
>>>>
>>>>> On Sun, Jul 19, 2009 at 08:05:47PM -0700, Philip A. Prindeville wrote:
>>>>>
>>>>>> Actually, it's saying:
>>>>>>
>>>>>> filename:       /lib/modules/2.6.27.26-astlinux/kernel/drivers/net/wireless/ath5k/ath5k.ko
>>>>>> depends:        mac80211,cfg80211
>>>>>>
>>>>>> And the 0.6.0 version number matches ath5k/base.c in the compat-wireless
>>>>>> sources as well.
>>>>>>
>>>>> Yeah, ignore the module version number, no one ever changes it.
>>>>>
>>>>>
>>>>>> Not seeing the dependency on "ath", which is strange.  Is that common code
>>>>>> that's shared in 2.6.30 for the Atheros drivers?
>>>>>>
>>>>> In 2.6.31, yes.
>>>>>
>>>>>
>>>>>>> The fix was merged after 2.6.30, IIRC, but it should definitely be in
>>>>>>> compat-wireless.  Also check your modules directory for an old ath5k.ko
>>>>>>> sitting in drivers/.../wireless/ath5k.  It should now be in ath/ath5k.
>>>>>>>
>>>>> See above, the fix was post-2.6.30, however compat-wireless should have
>>>>> it.  The module should now be in drivers/net/wireless/ath/ath5k, not
>>>>> drivers/net/wireless/ath5k.
>>>>>
>>>>>
>>>> I'm using the tarball out of:
>>>>
>>>> http://www.orbit-lab.org/kernel/compat-wireless-2.6-stable/v2.6.30
>>>>
>>>> perhaps I should be using a different one?  I tried 2.6.31-rc1, but it doesn't build against 2.6.27.
>>>>
>>> I have to fix this, will do so in a bit, will also update it to
>>> account for the new rc3.
>>>
>>>   Luis
>>>
>>
>> Well, while waiting for a "blessed" (or at least, 'stable') tarball to be pushed out, I tried running 2009-06-30, and I'm seeing a lot of:
>>
>> mgmt::proberesp cb
>> Probe Request from 00:21:29:69:3e:a0 for broadcast SSID ignored
>> STA 00:21:29:69:3e:a0 sent probe request for our SSID
>> MGMT (TX callback) fail
>> mgmt::proberesp cb
>> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
>> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
>> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
>> ...
>> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
>> MGMT (TX callback) ACK
>> mgmt::proberesp cb
>> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
>> MGMT (TX callback) ACK
>> mgmt::proberesp cb
>> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
>> MGMT (TX callback) fail
>> mgmt::proberesp cb
>> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
>> MGMT (TX callback) fail
>> mgmt::proberesp cb
>> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
>> MGMT (TX callback) fail
>> mgmt::proberesp cb
>> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
>> MGMT (TX callback) fail
>> mgmt::proberesp cb
>> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
>> MGMT (TX callback) fail
>> mgmt::proberesp cb
>> STA 00:21:e9:6e:a7:4f sent probe request for our SSID
>> MGMT (TX callback) fail
>> mgmt::proberesp cb
>> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
>> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
>> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
>> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
>> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
>> Probe Request from 00:21:e9:6e:a7:4f for broadcast SSID ignored
>> STA 00:21:29:69:3e:a0 sent probe request for our SSID
>> Probe Request from 00:21:29:69:3e:a0 for broadcast SSID ignored
>> MGMT (TX callback) ACK
>> mgmt::proberesp cb
>> STA 00:21:29:69:3e:a0 sent probe request for our SSID
>> MGMT (TX callback) fail
>> mgmt::proberesp cb
>>
>>
>>
>> 00:21:29:69:3e:a0 is a PC running Vista SP2.
>>
>> 00:21:e9:6e:a7:4f is an iPhone running 3.0.
>>
>> The iPhone should be scanning for available networks, but it's also been explicitly configured with this particular AP (and SSID).
>>
>> Is this a known issue?
>>
>> -Philip
>>
>
> A little progress:  I build 2009-07-10 with 2.6.27.26, and I can get the
> Vista box to associate, but not an iPhone running 3.0...

BTW new shiny compat-wireless is out based on John's latest merge from today.

> I have traces if anyone is interested.
>
> -Philip
>
>

Re: Using compat-wireless w/ 2.6.27.26

[Luis R. Rodriguez]

On Mon, Jul 20, 2009 at 12:47 PM, Luis R. Rodriguez<mcgrof@gmail.com> wrote:
> On Mon, Jul 20, 2009 at 12:16 PM, Philip A.
> Prindeville<philipp_subx@redfish-solutions.com> wrote:
>> Bob Copeland wrote:
>>> On Sun, Jul 19, 2009 at 08:05:47PM -0700, Philip A. Prindeville wrote:
>>>> Actually, it's saying:
>>>>
>>>> filename:       /lib/modules/2.6.27.26-astlinux/kernel/drivers/net/wireless/ath5k/ath5k.ko
>>>> depends:        mac80211,cfg80211
>>>>
>>>> And the 0.6.0 version number matches ath5k/base.c in the compat-wireless
>>>> sources as well.
>>>
>>> Yeah, ignore the module version number, no one ever changes it.
>>>
>>>> Not seeing the dependency on "ath", which is strange.  Is that common code
>>>> that's shared in 2.6.30 for the Atheros drivers?
>>>
>>> In 2.6.31, yes.
>>>
>>>>> The fix was merged after 2.6.30, IIRC, but it should definitely be in
>>>>> compat-wireless.  Also check your modules directory for an old ath5k.ko
>>>>> sitting in drivers/.../wireless/ath5k.  It should now be in ath/ath5k.
>>>
>>> See above, the fix was post-2.6.30, however compat-wireless should have
>>> it.  The module should now be in drivers/net/wireless/ath/ath5k, not
>>> drivers/net/wireless/ath5k.
>>>
>>
>> I'm using the tarball out of:
>>
>> http://www.orbit-lab.org/kernel/compat-wireless-2.6-stable/v2.6.30
>>
>> perhaps I should be using a different one?  I tried 2.6.31-rc1, but it doesn't build against 2.6.27.
>
> I have to fix this, will do so in a bit, will also update it to
> account for the new rc3.

Updated:

http://www.orbit-lab.org/kernel/compat-wireless-2.6-stable/v2.6.31/compat-wireless-2.6.31-rc3.tar.bz2

  Luis

Re: Using compat-wireless w/ 2.6.27.26

[Bob Copeland]

On Tue, Jul 21, 2009 at 01:10:33PM -0700, Philip A. Prindeville wrote:
> A little progress:  I build 2009-07-10 with 2.6.27.26, and I can get the
> Vista box to associate, but not an iPhone running 3.0...

Yeah there's a bug, I can replicate it and am trying to track it
down right now.

-- 
Bob Copeland %% www.bobcopeland.com

Re: Using compat-wireless w/ 2.6.27.26

[Bob Copeland]

On Tue, Jul 21, 2009 at 01:10:33PM -0700, Philip A. Prindeville wrote:
> A little progress:  I build 2009-07-10 with 2.6.27.26, and I can get the
> Vista box to associate, but not an iPhone running 3.0...

Here's a start, but I still need to find and fix another issue.

Ath5k seems to have had an implicit assumption that config() would
be called before beacon setup and the opmode would get set in HW then.
Not sure if the order changed when beacon stuff was moved to bss_conf,
but it's fragile anyway so I think we should configure the mode up
front so that the timers actually work.  Also we need to process all
of the TX queue's status descriptors.

diff --git a/drivers/net/wireless/ath/ath5k/base.c b/drivers/net/wireless/ath/ath5k/base.c
index c00f83f..168649d 100644
--- a/drivers/net/wireless/ath/ath5k/base.c
+++ b/drivers/net/wireless/ath/ath5k/base.c
@@ -1117,6 +1117,8 @@ ath5k_mode_setup(struct ath5k_softc *sc)
 	struct ath5k_hw *ah = sc->ah;
 	u32 rfilt;
 
+	ah->ah_op_mode = sc->opmode;
+
 	/* configure rx filter */
 	rfilt = sc->filter_flags;
 	ath5k_hw_set_rx_filter(ah, rfilt);
@@ -1999,8 +2001,13 @@ static void
 ath5k_tasklet_tx(unsigned long data)
 {
 	struct ath5k_softc *sc = (void *)data;
+	int i;
 
-	ath5k_tx_processq(sc, sc->txq);
+	for (i=0; i < AR5K_NUM_TX_QUEUES; i++) {
+		if (!sc->txqs[i].setup)
+			continue;
+		ath5k_tx_processq(sc, sc->txq);
+	}
 }
 
 
@@ -2768,6 +2775,7 @@ static int ath5k_add_interface(struct ieee80211_hw *hw,
 	}
 
 	ath5k_hw_set_lladdr(sc->ah, conf->mac_addr);
+	ath5k_mode_setup(sc);
 
 	ret = 0;
 end:

-- 
Bob Copeland %% www.bobcopeland.com

Re: Using compat-wireless w/ 2.6.27.26

[Bob Copeland]

On Thu, Jul 23, 2009 at 7:24 AM, Bob Copeland<me@bobcopeland.com> wrote:
> -       ath5k_tx_processq(sc, sc->txq);
> +       for (i=0; i < AR5K_NUM_TX_QUEUES; i++) {
> +               if (!sc->txqs[i].setup)
> +                       continue;
> +               ath5k_tx_processq(sc, sc->txq);

oops, should be:

ath5k_tx_processq(sc, sc->txqs[i]);

> +       }
>  }
>
-- 
Bob Copeland %% www.bobcopeland.com

Re: Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

Bob Copeland wrote:
> On Thu, Jul 23, 2009 at 7:24 AM, Bob Copeland<me@bobcopeland.com> wrote:
>> -       ath5k_tx_processq(sc, sc->txq);
>> +       for (i=0; i < AR5K_NUM_TX_QUEUES; i++) {
>> +               if (!sc->txqs[i].setup)
>> +                       continue;
>> +               ath5k_tx_processq(sc, sc->txq);
> 
> oops, should be:
> 
> ath5k_tx_processq(sc, sc->txqs[i]);
> 
>> +       }
>>  }
>>

Getting closer :-)

ath5k_tx_processq(sc, &sc->txqs[i]);

Re: Using compat-wireless w/ 2.6.27.26

[Bob Copeland]

On Thu, Jul 23, 2009 at 4:29 PM, Philip A.
Prindeville<philipp_subx@redfish-solutions.com> wrote:

> Getting closer :-)
>
> ath5k_tx_processq(sc, &sc->txqs[i]);

Yeah I spotted it on my next compile but then didn't want to send
yet-another-email... Oh well :)  Thanks!


-- 
Bob Copeland %% www.bobcopeland.com

Re: Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

Philip A. Prindeville wrote:
> Bob Copeland wrote:
>   
>> On Thu, Jul 23, 2009 at 7:24 AM, Bob Copeland<me@bobcopeland.com> wrote:
>>     
>>> -       ath5k_tx_processq(sc, sc->txq);
>>> +       for (i=0; i < AR5K_NUM_TX_QUEUES; i++) {
>>> +               if (!sc->txqs[i].setup)
>>> +                       continue;
>>> +               ath5k_tx_processq(sc, sc->txq);
>>>       
>> oops, should be:
>>
>> ath5k_tx_processq(sc, sc->txqs[i]);
>>
>>     
>>> +       }
>>>  }
>>>
>>>       
>
> Getting closer :-)
>
> ath5k_tx_processq(sc, &sc->txqs[i]);
>   

Didn't work with rc4:

BUG: unable to handle kernel NULL pointer dereference at 00000000
IP: [<d0935658>] :ath5k:ath5k_tasklet_tx+0x5e/0x27b
*pde = 00000000 
Oops: 0000 [#1] PREEMPT 
Modules linked in: lm90 hwmon scx200_acb i2c_core bridge stp llc dummy ath5k mac80211 ath cfg80211 rfkill_backport dahdi_dummy dahdi sha512_generic sha256_generic deflate zlib_deflate arc4 ecb sha1_generic blowfish des_generic cbc cryptosoft cryptodev(P) ocf(P) geodewdt geode_rng geode_aes crypto_blkcipher via_rhine rtc

Pid: 1585, comm: hostapd Tainted: P          (2.6.27.26-astlinux #1)
EIP: 0060:[<d0935658>] EFLAGS: 00010202 CPU: 0
EIP is at ath5k_tasklet_tx+0x5e/0x27b [ath5k]
EAX: c5712000 EBX: cf994000 ECX: 00000000 EDX: 00000000
ESI: ffffff8d EDI: c5713d00 EBP: c5488920 ESP: c5713ccc
 DS: 007b ES: 007b FS: 0000 GS: 0000 SS: 0068
Process hostapd (pid: 1585, ti=c5712000 task=cf86a0a0 task.ti=c5712000)
Stack: c5713cec c548bf9c c548bfb0 c548bfd0 c011a1d7 00000000 00000028 c548bfb0 
       00000000 00000000 00000000 00000000 00000000 00000000 c039d214 0000000a 
       00000009 c011a3f9 00000001 c011a111 00000046 c03473f4 00000000 c011a161 
Call Trace:
 [<c011a1d7>] _local_bh_enable_ip+0x72/0x7c
 [<c011a3f9>] tasklet_action+0x3c/0x62
 [<c011a111>] __do_softirq+0x30/0x5e
 [<c011a161>] do_softirq+0x22/0x26
 [<c011a4fa>] irq_exit+0x25/0x33
 [<c0104a14>] do_IRQ+0x4d/0x5d
 [<c0103a1b>] common_interrupt+0x23/0x28
 [<c02445ec>] __sock_sendmsg+0x20/0x24
 [<c0245331>] sock_sendmsg+0x95/0xad
 [<c0125714>] autoremove_wake_function+0x0/0x2b
 [<c0125714>] autoremove_wake_function+0x0/0x2b
 [<c013aca6>] get_pageblock_flags_group+0xf/0x50
 [<c024b551>] verify_iovec+0x3e/0x6d
 [<c02454d6>] sys_sendmsg+0x18d/0x1ed
 [<c0245d72>] sys_recvmsg+0x144/0x1c6
 [<c0245de0>] sys_recvmsg+0x1b2/0x1c6
 [<c0113a7b>] __wake_up+0x1d/0x45
 [<c0113a93>] __wake_up+0x35/0x45
 [<c024dc8e>] dev_name_hash+0x13/0x42
 [<c024dc8e>] dev_name_hash+0x13/0x42
 [<c024dcec>] __dev_get_by_name+0x2f/0x39
 [<c011a1d7>] _local_bh_enable_ip+0x72/0x7c
 [<c0246dd0>] lock_sock_nested+0xb1/0xb8
 [<c011a1d7>] _local_bh_enable_ip+0x72/0x7c
 [<c0246377>] sys_socketcall+0x15b/0x193
 [<c0152703>] sys_close+0x75/0xc5
 [<c01038b6>] syscall_call+0x7/0xb
 =======================
Code: 00 03 85 b0 36 00 00 8b 3c 24 fc 89 44 24 0c 31 c0 ab ab ab ab ab 89 e0 25 00 e0 ff ff ff 40 14 8b 54 24 0c 8b 52 08 89 54 24 14 <8b> 02 8b 54 24 0c 83 c2 08 89 44 24 1c 89 54 24 08 e9 96 01 00 
EIP: [<d0935658>] ath5k_tasklet_tx+0x5e/0x27b [ath5k] SS:ESP 0068:c5713ccc
Kernel panic - not syncing: Fatal exception in interrupt

Re: Using compat-wireless w/ 2.6.27.26

[Bob Copeland]

On Thu, Jul 23, 2009 at 5:53 PM, Philip A.
Prindeville<philipp_subx@redfish-solutions.com> wrote:
>> Getting closer :-)
>>
>> ath5k_tx_processq(sc, &sc->txqs[i]);
>>
>
> Didn't work with rc4:

What did you apply against?  It's supposed to be against
wireless-testing (or, at worst, latest compat-wireless).

> BUG: unable to handle kernel NULL pointer dereference at 00000000
> IP: [<d0935658>] :ath5k:ath5k_tasklet_tx+0x5e/0x27b

Can you run scripts/markup_oops.pl on that?

-- 
Bob Copeland %% www.bobcopeland.com

Re: Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

Bob Copeland wrote:
> On Thu, Jul 23, 2009 at 5:53 PM, Philip A.
> Prindeville<philipp_subx@redfish-solutions.com> wrote:
>>> Getting closer :-)
>>>
>>> ath5k_tx_processq(sc, &sc->txqs[i]);
>>>
>> Didn't work with rc4:
> 
> What did you apply against?  It's supposed to be against
> wireless-testing (or, at worst, latest compat-wireless).
> 
>> BUG: unable to handle kernel NULL pointer dereference at 00000000
>> IP: [<d0935658>] :ath5k:ath5k_tasklet_tx+0x5e/0x27b
> 
> Can you run scripts/markup_oops.pl on that?
> 

This was against compat-wireless-2.6.31-rc4.

Alas, I already blew away the build directory...

Re: Using compat-wireless w/ 2.6.27.26

[Bob Copeland]

On Thu, Jul 23, 2009 at 6:45 PM, Philip A.
Prindeville<philipp_subx@redfish-solutions.com> wrote:
> Bob Copeland wrote:
>> On Thu, Jul 23, 2009 at 5:53 PM, Philip A.
>> Prindeville<philipp_subx@redfish-solutions.com> wrote:
>>>> Getting closer :-)
>>>>
>>>> ath5k_tx_processq(sc, &sc->txqs[i]);
>>>>
>>> Didn't work with rc4:
>>
>> What did you apply against?  It's supposed to be against
>> wireless-testing (or, at worst, latest compat-wireless).
>>
>>> BUG: unable to handle kernel NULL pointer dereference at 00000000
>>> IP: [<d0935658>] :ath5k:ath5k_tasklet_tx+0x5e/0x27b
>>

> This was against compat-wireless-2.6.31-rc4.
>
> Alas, I already blew away the build directory...

I think you need to use the unstable compat-wireless snapshots;
I have a feeling that the above doesn't have all of the tx queues.

Check out ath5k.h, the txq array in struct ath5k_softc should have
NUM_TX_QUEUES entries, an earlier version only had 2.

Although your compiler generates rather different code from mine
(different config or whatever), it looks like the oops happened
in the list_for_each_entry(... txq->q ...) -- but I don't see how
that can happen in my copy, since txq[i].q is initialized at the
same time as txq[i].setup.

-- 
Bob Copeland %% www.bobcopeland.com

Re: Using compat-wireless w/ 2.6.27.26

[Philip A. Prindeville]

Bob Copeland wrote:
> On Thu, Jul 23, 2009 at 6:45 PM, Philip A.
> Prindeville<philipp_subx@redfish-solutions.com> wrote:
>> Bob Copeland wrote:
>>> On Thu, Jul 23, 2009 at 5:53 PM, Philip A.
>>> Prindeville<philipp_subx@redfish-solutions.com> wrote:
>>>>> Getting closer :-)
>>>>>
>>>>> ath5k_tx_processq(sc, &sc->txqs[i]);
>>>>>
>>>> Didn't work with rc4:
>>> What did you apply against?  It's supposed to be against
>>> wireless-testing (or, at worst, latest compat-wireless).
>>>
>>>> BUG: unable to handle kernel NULL pointer dereference at 00000000
>>>> IP: [<d0935658>] :ath5k:ath5k_tasklet_tx+0x5e/0x27b
> 
>> This was against compat-wireless-2.6.31-rc4.
>>
>> Alas, I already blew away the build directory...
> 
> I think you need to use the unstable compat-wireless snapshots;
> I have a feeling that the above doesn't have all of the tx queues.
> 
> Check out ath5k.h, the txq array in struct ath5k_softc should have
> NUM_TX_QUEUES entries, an earlier version only had 2.
> 
> Although your compiler generates rather different code from mine
> (different config or whatever), it looks like the oops happened
> in the list_for_each_entry(... txq->q ...) -- but I don't see how
> that can happen in my copy, since txq[i].q is initialized at the
> same time as txq[i].setup.
> 

Give me a shout when this patch makes its way into compat-wireless-2.6.31-rcN and I'll try again with that.

-Philip