Hi,
I've just pulled latest -git and tried this morning in on one of my ALIX

mobos equipped with an AMD Geode LX800 + CS5536. HEAD is a7da60f41551*.
Last kernel which runs on it was 2.6.22.14. I've not yet tried intermediate

kernels nor tried to bisect yet. Last messages are :
Dentry cache hash table entries: 32768 (order: 5, 131072 bytes)

Inode-cache hash table entries: 16384 (order: 4, 65536 bytes)

Memory: 257304k/262144k available (1506k kernel code, 4400k reserved, 370k data,

 136k init, 0k highmem)

virtual kernel memory layout:

    fixmap  : 0xffffb000 - 0xfffff000   (  16 kB)

    vmalloc : 0xd0800000 - 0xffff9000   ( 759 MB)

    lowmem  : 0xc0000000 - 0xd0000000   ( 256 MB)

      .init : 0xc02d8000 - 0xc02fa000   ( 136 kB)

      .data : 0xc0278970 - 0xc02d5504   ( 370 kB)

      .text : 0xc0100000 - 0xc0278970   (1506 kB)

Checking if this processor honours the WP bit even in supervisor mode... Ok.

SLUB: Genslabs=11, HWalign=32, Order=0-1, MinObjects=4, CPUs=1, Nodes=1

Calibrating delay using timer specific routine.. 997.59 BogoMIPS (lpj=1995180)

Mount-cache hash table entries: 512

CPU: After generic identify, caps: 0088a93d c0c0a13d 00000000 00000000 00000000

00000000 00000000 00000000

CPU: L1 I Cache: 64K (32 bytes/line), D cache 64K (32 bytes/line)

CPU: L2 Cache: 128K (32 bytes/line)

CPU: After all inits, caps: 0088a93d c0c0a13d 00000000 00000000 00000000 0000000

0 00000000 00000000

Compat vDSO mapped to ffffe000.      ...
[ message continues ]

On Sun, 20 Jan 2008 09:14:42 +0100
Hi Willy,
If this is in fact true (that GEODE_MFGPT_TIMER was disabled), the

machine hung, and with nomfgpt it didn't hang..   That sounds like

our geode_mfgpt_detect call hung.  Do you have "MFGPT Workaround"

selected in the BIOS?  I'm curious whether enabling or disabling

that makes a difference..
--

Hi,
Did you try to boot with nomfgpt?

In my experiments disabling CONFIG_GEODE_MFGPT_TIMER won't be enough to

boot 2.6.24-rc8 on a v0.98 bios or on a v0.99 bios with MFGPT workaround

turned on in the bios menu.

Otherwise the system will hang in arch/x86/kernel/mfgpt_32.c while

trying to detect MFGPT timers.

To clearify, for us the following works:

v0.99 bios, MFGPT workaround turned off in bios and

CONFIG_GEODE_MFGPT_TIMER=n
--- snip ---
Best regards,

Arnd Hannemann

--

Hi Arnd,
Just did, it works perfectly that way, thanks very much!
I will try to investigate why this is necessary now. Last 2.6.22.X I used

was patched with a geode-mfgpt patch I found which enabled the watchdog
I will also check BIOS v0.99 sometime later.
Thanks very much for your quick and insightful response!
Willy
--

Hi Arnd,
I don"t remember and am not at home right now. I had 1.1 in mind but I may
OK thanks very much. I'll try this ASAP.
Cheers!

Willy
--

just a hint: this is near the first piece of code that makes active use

of a timer wakeup, on a typical x86 bootup. So sometimes if timers are

not working at all, hangs happen around here. But the PCI related thing

you suspect could be the reason too. Getting a SysRq-T or an NMI

printout (if that's possible) of the hung site would be nice.
	Ingo

--

Hi Ingo,
OK Thanks for the hint, I'll try things like this when I get back home.
Cheers,

Willy
--