[Bug 567137] New: 2.6.25.20-0.5 implementation of ath5k driver refuses to identify hardware correctly
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c0
Summary: 2.6.25.20-0.5 implementation of ath5k driver refuses
to identify hardware correctly
Classification: openSUSE
Product: openSUSE 11.0
Version: Final
Platform: IA64
OS/Version: openSUSE 11.0
Status: NEW
Severity: Major
Priority: P5 - None
Component: Network
AssignedTo: bnc-team-screening@forge.provo.novell.com
ReportedBy: drankinatty@suddenlinkmail.com
QAContact: qa@suse.de
Found By: ---
Blocker: ---
User-Agent: Mozilla/5.0 (X11; U; Linux x86_64; en-US; rv:1.9.1.6)
Gecko/20091216 Shiretoko/3.5.6
I have probably installed the madwifi driver at least 10 time on my laptop
on SuSE 10.2 through 11.2, but after a new 11.0 install, I am unable to find a
madwifi driver that will install and "stay" installed. The funny thing is that
the madwifi driver has no problems on the same box when installed on Arch
Linux.
Even more perplexing, I still have the 11.0 rpms that 'I WAS USING ON THIS
BOX' that no longer work, so I know the difficulty is something openSuSE
specific with the latest kernel or two. I say that, because the existing
drivers I have -- have always worked when I installed them with the original
kernel and then updated through each successive kernel. Now, jumping from
2.6.25.5-1.1 to 2.6.25.20-0.5 in one step has left 11.0 in a condition where it
will not load the driver properly.
I have built and installed from the driver for the following sources all of
which worked on this box in the past:
madwifi-0.9.4-1.src.rpm # The standard 0.9.4-1 driver
madwifi-0.9.4-1-current.src.rpm # The standard 0.9.4-1 driver with the
madwifi-project suggested snapshot for 2.6.25+ kernels.
madwifi-svn-122109 # this module worked briefly (limited to 1M/s), but then
died and despite a number of reinstalls will not configure properly.
madwifi-ng_r3366+ar5007-1.4.src.rpm # always worked in the past
madwifi-ng_r3366+ar5007-1.9.src.rpm # always worked...
madwifi-r3725+AR5007EG-2.2.src.rpm # what I was running on before the 11.2
install
madwifi-trunk+r3717-1.2.src.rpm # worked in the past...
and I have one or two more around... I even got to the point I was making
sure the kernel was 'properly prepared' and that 'all' remnants of the prior
installs were removed from the module tree and unloaded by reboot. I would:
make mrproper
make cloneconfig
make modules_prepare
make clean
in between the package installs.
Two notes of interest. (1) the packages are never consistent between where
they put the modules in the module tree under
/lib/modules/2.6.25.20-0.5-default/.. Sometimes the modules go under ../net,
sometimes under ../updates, sometimes under ../weak-updates, but until now it
always worked.
Second, under openSuSE, I have always had to use the 'ath_pci' module. No
matter what, the ath5k driver has never worked -- period. That always seemed
strange. Well I confirmed it was strange, because when I checked the module
used by Arch Linux on this box, low-and-behold, it was using ath5k??? How does
that happen?
The errors are pretty much of the same flavor. After the driver is loaded,
it complains that the chip doesn't respond:
Dec 22 15:00:11 alchemy kernel: PCI: Enabling device 0000:17:00.0 (0000 ->
0002)
Dec 22 15:00:11 alchemy kernel: ACPI: PCI Interrupt 0000:17:00.0[A] -> GSI 19
(level, low) -> IRQ 19
Dec 22 15:00:11 alchemy kernel: PCI: Setting latency timer of device
0000:17:00.0 to 64
Dec 22 15:00:11 alchemy kernel: MadWifi: unable to attach hardware: 'Hardware
didn't respond as expected' (HAL status 3)
Dec 22 15:00:11 alchemy kernel: ACPI: PCI interrupt for device 0000:17:00.0
disabled
Dec 22 15:02:18 alchemy kernel: ACPI: PCI Interrupt 0000:17:00.0[A] -> GSI 19
(level, low) -> IRQ 19
Dec 22 15:02:18 alchemy kernel: PCI: Setting latency timer of device
0000:17:00.0 to 64
Dec 22 15:02:18 alchemy kernel: ath5k_pci 0000:17:00.0: registered as 'phy1'
Dec 22 15:02:19 alchemy kernel: ath5k phy1: failed to wakeup the MAC Chip
Dec 22 15:02:19 alchemy kernel: ACPI: PCI interrupt for device 0000:17:00.0
disabled
Dec 22 15:02:19 alchemy kernel: ath5k_pci: probe of 0000:17:00.0 failed with
error -5
dmesg:
ath5k_pci 0000:17:00.0: registered as 'phy0'
ath5k phy0: failed to resume the MAC Chip
ath5k_pci: probe of 0000:17:00.0 failed with error -5
ath_hal: 0.9.18.0 (AR5210, AR5211, AR5212, RF5111, RF5112, RF2413, RF5413)
wlan: 0.9.4
ath_pci: 0.9.4
eth0: RTL8101e at 0xffffc2000065a000, 00:1b:38:af:36:b8, XID 34200000 IRQ 4348
I don't see any logical reason why what has always worked on this box now
refuses to work. I have probably uninstalled, confirmed the uninstall was
complete, built, rebuilt, reinstalled until I'm blue in the face.
I don't know what suse is doing with the module tree and how it is
assigning modules/devices. From what I can tell the modules.alias handling of
ath5k versus ath_pci may be an issue. In the past, I have always just
blacklisted ath5k in /etc/modprobe.d/blacklist to prevent it from loading, but
that doesn't work any more. I have attempted to insure that ath_pci is loaded
before ath5k with MODULES_LOADED_ON_BOOT`as shown in
http://en.opensuse.org/Kernel_module_configuration. Still doesn't help.
For comparison that will hopefully help in figuring out what is going on.
Currently I have my arch drive installed in this same laptop. Here is the
config for madwifi (ath5k driver) that is working fine:
syslog:
Dec 23 04:30:35 alchemy kernel: ath5k 0000:17:00.0: PCI INT A -> GSI 19 (level,
low) -> IRQ 19
Dec 23 04:30:35 alchemy kernel: ath5k 0000:17:00.0: registered as 'phy0'
Dec 23 04:30:35 alchemy kernel: ath5k phy0: Atheros AR2425 chip found (MAC:
0xe2, PHY: 0x70)
Dec 23 04:30:51 alchemy avahi-daemon[4161]: Joining mDNS multicast group on
interface wlan0.IPv4 with address 192.168.6.102.
Dec 23 04:30:51 alchemy avahi-daemon[4161]: New relevant interface wlan0.IPv4
for mDNS.
Dec 23 04:30:51 alchemy avahi-daemon[4161]: Registering new address record for
192.168.6.102 on wlan0.IPv4.
Dec 23 04:30:52 alchemy ntpd[4170]: Listen normally on 3 wlan0 192.168.6.102
UDP 123
Dec 23 04:30:53 alchemy avahi-daemon[4161]: Registering new address record for
fe80::21b:9eff:fe7c:f6e7 on wlan0.*.
Dec 23 04:30:54 alchemy ntpd[4170]: Listen normally on 7 wlan0
fe80::21b:9eff:fe7c:f6e7 UDP 123
dmesg:
18:25 alchemy:~> sudo dmesg | grep ath
ath5k 0000:17:00.0: PCI INT A -> GSI 19 (level, low) -> IRQ 19
ath5k 0000:17:00.0: setting latency timer to 64
ath5k 0000:17:00.0: registered as 'phy0'
ath: EEPROM regdomain: 0x64
ath: EEPROM indicates we should expect a direct regpair map
ath: Country alpha2 being used: 00
ath: Regpair used: 0x64
ath5k phy0: Atheros AR2425 chip found (MAC: 0xe2, PHY: 0x70)
ath5k phy0: unsupported jumbo
ath5k phy0: unsupported jumbo
ath5k phy0: unsupported jumbo
madwifi-0.9.4 package:
18:24 alchemy:~> pacman -Q madwifi
madwifi 0.9.4.4100-1
18:24 alchemy:~> pacman -Ql madwifi
madwifi /lib/
madwifi /lib/modules/
madwifi /lib/modules/2.6.32-ARCH/
madwifi /lib/modules/2.6.32-ARCH/kernel/
madwifi /lib/modules/2.6.32-ARCH/kernel/drivers/
madwifi /lib/modules/2.6.32-ARCH/kernel/drivers/net/
madwifi /lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/
madwifi /lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/
madwifi /lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/ath_hal.ko
madwifi /lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/ath_pci.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/ath_rate_amrr.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/ath_rate_minstrel.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/ath_rate_onoe.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/ath_rate_sample.ko
madwifi /lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/wlan.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/wlan_acl.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/wlan_ccmp.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/wlan_scan_ap.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/wlan_scan_sta.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/wlan_tkip.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/wlan_wep.ko
madwifi
/lib/modules/2.6.32-ARCH/kernel/drivers/net/wireless/madwifi/wlan_xauth.ko
Why doesn't openSuSE activate the driver successfully in the manner that
Arch does? I have looked at modinfo to see if there are any special module
loading options being used in Arch, but there aren't any. Hopefully it is just
a bug in hardware recognition that can be easily fixed. Let me know if you need
anything else. Thanks. (See the Additional information below)
Reproducible: Always
Steps to Reproduce:
1. install 11.0, update kernel to 2.6.25.20-0.5-default
2. then try to install the madwifi driver
3.
Actual Results:
ath5k doesn't recognize hardware correctly and prevents ath_pci from loading.
Blacklisting ath5k has no effect, ath5k still loads.
Expected Results:
working madwifi install.
It looks like the problem may be opensuse incorrectly recognizing the card. I
will have to swap drives to load opensuse again. It looks like opensuse fails
to recognize the hardware correctly. The dmesg error from opensuse is:
ath5k_pci 0000:17:00.0: registered as 'phy0'
ath5k phy0: failed to resume the MAC Chip
ath5k_pci: probe of 0000:17:00.0 failed with error -5
There is no recognigition of the AR2425 chip. Comparing, Arch Linux dmesg
information shows the chip as:
ath5k phy0: Atheros AR2425 chip found (MAC: 0xe2, PHY: 0x70)
The complete lspci from the Arch Linux install is:
17:00.0 Ethernet controller: Atheros Communications Inc. AR5001 Wireless
Network Adapter (rev 01)
Subsystem: Askey Computer Corp. Device 7128
Control: I/O+ Mem+ BusMaster+ SpecCycle- MemWINV- VGASnoop- ParErr-
Stepping- SERR- FastB2B- DisINTx-
Status: Cap+ 66MHz- UDF- FastB2B- ParErr- DEVSEL=fast >TAbort-
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c
Marcus Meissner
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c1
Jeff Mahoney
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c2
David Rankin
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c
Jeff Mahoney
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c3
Jiri Slaby
Yes, I'm blacklisting ath5k (have to -- no choice)
Why no choice? 11.0 kernel is pretty old. Could you update to openSUSE 11.2 or at least to a kernel from the Kernel:HEAD repository? There were problems with 2425 chip in the early ath5k.
The problem here is in the opensuse kernel module table that coordinates card hardware ID with the proper module.
Sorry I don't understand here. What do you mean by that?
It was just a fluke that I loaded Arch Linux on the spare drive for the laptop, and presto! the same ath5k driver worked perfectly because the lookup table properly identified the card assigned ath5k correctly.
What kernel version is in Arch? -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c4
David Rankin
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c5
Jiri Slaby
I can't update to 11.2, I tried, but opensuse does not have a fglrx driver for 11.2 for my card (x1200) and without the fglrx driver desktop effects crashes and spontaneously shuts down. The radeonhd driver is a no-go as well due to July 09 changes that left it blackscreening for my card. So 11.0 with the 8-9 ATI driver (fglrx 8.532) is the only workable solution.
11.0 is near the end of its lifecycle. Please report any problems in 11.2 you have to have fixes and support after 11.0 ceases to exist.
I am no expect of kernel module loading, but I believe the problem with the madwifi driver is the openSuSE /lib/modules/<kernel>/modules.alias file which causes the proper module to fail to load. Specifically, I think suse has a problem with module.alias because it has duplicate entries for the same madwifi device that points to different modules (that can't be right).
It is pretty OK to have double entries in alias list.
My Atheros card has the following ID:
ath_pci: r3725+AR5007EG ACPI: PCI Interrupt 0000:17:00.0[A] -> GSI 19 (level, low) -> IRQ 19 ^^^^^^^^^^^^^
These are just bus identification. Use lspci -nn to see PCI IDs.
SuSE module.alias has: alias pci:v0000168Cd00000017sv*sd*bc*sc*i* ath5k ^^^^^^ ^^^^^ .. alias pci:v0000168Cd00000017sv*sd*bc*sc*i* ath_pci ^^^^^^ ^^^^^ .. That is duplicate entries for 0000:17:00.0[A] with one pointing to ath5k and one pointing at ath_pci. Consequently it doesn't know what module to load.
It loads both. And the earlier loaded binds to the device.
No with Arch, the Arch module.alias only has one entry for 0000:17:00.0[A]: .. alias pci:v0000168Cd00000017sv*sd*bc*sc*i* ath5k ^^^^^^ ^^^^^
So, you don't use madwifi at all in Arch? (You shouldn't have to.)
If you look at the dmesg and syslog messages posted above, I think this explains why my Atheros card is nearly impossible to get working on SuSE without compiling custom madwifi packages, standing on one foot, and wiggling your nose during the driver install and boot :p
Sorry, your card (better, the 2425 chip) is not supported in 11.0. You have to use later kernel (e.g. 2.6.32 in 11.2 or 2.6.33 in Arch). Or use compat-wireless. Then you can get rid of madwifi and use solely ath5k which should work just fine. Though, the blacklisting should work. What's the output of modprobe -bnv ath5k ? -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c6
--- Comment #6 from David Rankin
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c7
David Rankin
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c
Jiri Slaby
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c8
Jiri Slaby
17:00.0 Ethernet controller [0200]: Atheros Communications Inc. AR242x 802.11abg Wireless PCI Express Adapter [168c:001c] (rev 01)
So you should look at v168Cp001C lines.
As far as the card not being supported -- I'm confused? I have used this laptop with 10.2 -> 11.2 and back to 11.0. In 10.2 & 10.3, the card was no problem to configure with madwifi. Just blacklist ath5k and madwifi worked instantly.
Yeah, by supported I meant supported by ath5k in opensuse, not madwifi.
IIRC, I had to physically remove /lib/modules/<kernel>/(any ath5k reference) before the init process would skip the ath5k module and actually load the ath_pci module. This wasn't a problem with 10.2-10.3. This is a problem with 11.0-11.1.
If you remove the module, it cannot be loaded. Have you checked, that your initrd doesn't contain the module as well, so that init will load it there. What is output of zcat /boot/initrd|cpio -t|grep ath But you say that if you remove the module from /lib/modules/, it doesn't load, so I'm confused.
On 11.2, the card is recognized and ath5k works fine. Same with Arch, ath5k just works fine.
Great, as expected.
I still have madwifi installed for the encryption tools (wpa_passphrase, iwlist, etc.)
I don't know wpa_passphrase. Can it do something what wpa_supplicant can't? iwlist works with ath5k.
[11:21 alchemy:/home/david] # modprobe -bnv ath5k
So it doesn't load the module actually. Either it's removed from /lib/modules/... or properly blacklisted. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c9
--- Comment #9 from David Rankin
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c10
David Rankin
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c11
Jiri Slaby
So in sum, the problem is that ath5k in the standard kernel incorrectly recognizes the hardware and blacklisting ath5k is ignored leaving the wireless user screwed unless you go through the steps outlined above.
Michal, do you see immediately any reason for this, please? This is 11.0, so I think we're not going to invest much time into that unless it affects also 11.1/11.2. -- Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c12
Michal Marek
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c
Jiri Slaby
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c13
David Rankin
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c14
Jiri Slaby
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c15
Michal Marek
http://bugzilla.novell.com/show_bug.cgi?id=567137
http://bugzilla.novell.com/show_bug.cgi?id=567137#c16
Jiri Slaby
participants (1)
-
bugzilla_noreply@novell.com