Ticket #1645 (new defect)

Opened 1 year ago

Last modified 8 months ago

ath_pci panic when memory remapping enabled in bios (4Gb RAM)

Reported by: claudio@c3systems.com.br Assigned to:
Priority: major Milestone:
Component: madwifi: other Version: v0.9.3.3
Keywords: memory remapping 4Gb ASUS Atheros Cc:
Patch is attached: 0 Pending:

Description (Last modified by mrenzmann)

Hi there !

I had been using madwifi for years without any problems until now. Recently, I replaced my mobo with ASUS Commando and 4Gb of RAM. After the first boot, everything works fine, but I´m able to see only 2Gb of RAM. After some research, I found that I need to enable Memory Remapping in BIOS in order to see all the 4Gb of RAM. After that, in every single boot, kernel hangs when running udev. If I disabled memory remapping in BIOS, I´m able to boot again. Then, I try to remove each module trying to identify which one is causing the problem. After remove madwifi completely I was able to boot WITH Memory Remapping enabled. I´m running kernel 2.6.20-1.2320_FC5xen0 (64bits) with madwifi 0.9.3.3 (locally compiled). Distro is FC5-x86_64. Below follow lspci: 

00:00.0 Host bridge: Intel Corporation Unknown device 29a0 (rev 02)
00:01.0 PCI bridge: Intel Corporation Unknown device 29a1 (rev 02)
00:1a.0 USB Controller: Intel Corporation Unknown device 2834 (rev 02)
00:1a.1 USB Controller: Intel Corporation Unknown device 2835 (rev 02)
00:1a.7 USB Controller: Intel Corporation Unknown device 283a (rev 02)
00:1b.0 Audio device: Intel Corporation Unknown device 284b (rev 02)
00:1c.0 PCI bridge: Intel Corporation Unknown device 283f (rev 02)
00:1c.4 PCI bridge: Intel Corporation Unknown device 2847 (rev 02)
00:1c.5 PCI bridge: Intel Corporation Unknown device 2849 (rev 02)
00:1d.0 USB Controller: Intel Corporation Unknown device 2830 (rev 02)
00:1d.1 USB Controller: Intel Corporation Unknown device 2831 (rev 02)
00:1d.2 USB Controller: Intel Corporation Unknown device 2832 (rev 02)
00:1d.7 USB Controller: Intel Corporation Unknown device 2836 (rev 02)
00:1e.0 PCI bridge: Intel Corporation 82801 PCI Bridge (rev f2)
00:1f.0 ISA bridge: Intel Corporation Unknown device 2810 (rev 02)
00:1f.2 SATA controller: Intel Corporation Unknown device 2821 (rev 02)
00:1f.3 SMBus: Intel Corporation Unknown device 283e (rev 02)
01:00.0 VGA compatible controller: nVidia Corporation NV43 [GeForce 6600 GT] (rev a2)
02:00.0 Ethernet controller: Marvell Technology Group Ltd. Unknown device 4364 (rev 12)
03:00.0 IDE interface: JMicron Technologies, Inc. JMicron 20360/20363 AHCI Controller (rev 03)
03:00.1 IDE interface: JMicron Technologies, Inc. JMicron 20360/20363 AHCI Controller (rev 03)
04:00.0 VGA compatible controller: nVidia Corporation NV43 [GeForce 6600 GT] (rev a2)
05:00.0 Multimedia video controller: Internext Compression Inc iTVC15 MPEG-2 Encoder (rev 01)
05:02.0 Communication controller: Tiger Jet Network Inc. Tiger3XX Modem/ISDN interface
05:03.0 FireWire (IEEE 1394): Texas Instruments TSB43AB22/A IEEE-1394a-2000 Controller (PHY/Link)
05:04.0 Ethernet controller: Marvell Technology Group Ltd. 88E8001 Gigabit Ethernet Controller (rev 13)
05:05.0 Ethernet controller: Atheros Communications, Inc. AR5212 802.11abg NIC (rev 01)

I´m using the LAST Mobo BIOS. Processor is Intel Q6700 (kernel has SMP enabled).

Any ideas ?

Regards,

Claudio Cuqui

Attachments

madwifi.debug (2.1 kB) - added by claudio@c3systems.com.br on 11/15/07 09:46:05.

Change History

11/14/07 12:12:44 changed by mrenzmann

  • description changed.

11/14/07 15:28:05 changed by mentor

  • priority changed from blocker to major.

Would you provide the kernel oops log? DevDocs/KernelOops

11/14/07 16:42:30 changed by claudio@c3systems.com.br

I will try this today at night (this server is in production with memory remapping off - Unfortunately I can only run tests at night :-/)

Looks like there are other people with same problem, using the same solution (second paragraph - ticket 1159):

"Another bug which merits a separate ticket is that the atheros driver is not compatible with the IOMMU kernel option on the x86_64 architecture. In a nutshell since I have more than 3GB RAM in my box if I enable the "remap memory" BIOS (northbridge) option, every standard 2.6.20 driver seems to be IOMMU compliant, but if the latest (svn trunk, version Feb. 17) atheros driver is installed I get a lockup (i.e. CAPS lock key doesn't turn light off/on) on startup when the driver is loaded. So for now I disable memory remapping in the BIOS as a workaround. IOMMU and suboptions seem to be applicable to late series kernels, e.g. definitely 2.6.20 but it doesn't seem to be obeyed by drivers in older kernels like 2.6.17-11. In other words, don't bother investigating this issue on older kernels."

Regards !

11/15/07 09:44:41 changed by claudio@c3systems.com.br

Hi there,

Since the last night was a long one, now, I have some new information. Since is almost impossible to debug anything when using xen kernel (problems with console, interactions between driver, etc), I decide to save some precious time and use the last vanilla kernel (2.6.23.1) and the last madwifi driver to do my tests and built them on my own. I enable every single debug option in kernel, reboot with it, and then, built the driver. After try to load the driver, I have the same problem: the server just hang. Completly freezed ! Even SysRq? doesn´t work here (any combination). Attached below, you will find the messages returned after loading of each module (every module loads ok until I try to load ath_pci). PS: I forgot to tell that I was using autocreate=ap option in modprobe.conf. Even removing it, didn´t change anything.

11/15/07 09:46:05 changed by claudio@c3systems.com.br

  • attachment madwifi.debug added.

(follow-up: ↓ 6 ) 11/15/07 09:48:01 changed by strasak@bubakov.net

try to turn on nmi_watchdog, it helps most of the time

(in reply to: ↑ 5 ; follow-up: ↓ 7 ) 11/15/07 10:48:31 changed by anonymous

Replying to strasak@bubakov.net:

try to turn on nmi_watchdog, it helps most of the time

Doesn´t help (or I´m doing something wrong). As fas as I know, nmi_watchdog is compiled by default on x86_64 archs with SMP (this servers is a quad core processor and the kernel was compiled with SMP enabled - I read this in nmi_watchdog.txt in kernel/Documentation), so, it´s only necessary to enable it in grub, as I did: 

title Fedora Core (2.6.23.1-C3Systems-Intel)
        root (hd0,0)
        kernel /vmlinuz-2.6.23.1-C3Systems-Intel ro root=/dev/bluegene/root rhgb nmi_watchdog=1 single
        initrd /initrd-2.6.23.1-C3Systems-Intel.img

Any other suggestions ?

(in reply to: ↑ 6 ) 11/15/07 10:53:58 changed by claudio@c3systems.com.br

Replying to anonymous:

Replying to strasak@bubakov.net:

try to turn on nmi_watchdog, it helps most of the time

Doesn´t help (or I´m doing something wrong). As fas as I know, nmi_watchdog is compiled by default on x86_64 archs with SMP (this servers is a quad core processor and the kernel was compiled with SMP enabled - I read this in nmi_watchdog.txt in kernel/Documentation), so, it´s only necessary to enable it in grub, as I did: {{{ title Fedora Core (2.6.23.1-C3Systems-Intel) root (hd0,0) kernel /vmlinuz-2.6.23.1-C3Systems-Intel ro root=/dev/bluegene/root rhgb nmi_watchdog=1 single initrd /initrd-2.6.23.1-C3Systems-Intel.img }}} Any other suggestions ?

Just to clarify: Even after doing this, when I tried to modprobe ath_pci the server hangs exactly as before and even after waiting for almost 2 minutes for the nmi_watchdog/oops come in place, the servers keep freezed.

04/08/08 16:11:29 changed by claudio@c3systems.com.br

Hi there !

I just tried 0.9.4 yesterday and have exactly the same problems. I really like to know if there is anybody using madwifi driver (any version) with a P965 motherboard with 4Gb and memory remap enabled with success... I saw many reports describing the same issues that I´m facing.....

Regards !

Claudio Cuqui

Add/Change #1645 (ath_pci panic when memory remapping enabled in bios (4Gb RAM))