2.6.0-test8: cpufreq longhaul probs..still :-(

Hi,

With test8 there are still problems with the longhaul cpufreq stuff:

longhaul: VIA C3 'Ezra' [C5C] CPU detected. Longhaul v1 supported.
longhaul: MinMult=3.0x MaxMult=6.0x
longhaul: FSB: 0MHz Lowestspeed=0MHz Highestspeed=0MHz

------------[ cut here ]------------
kernel BUG at drivers/cpufreq/cpufreq_userspace.c:502!
invalid operand: 0000 [#1]
CPU:    0
EIP:    0060:[<c022a203>]    Not tainted
EFLAGS: 00010246

EIP is at cpufreq_governor_userspace+0xb3/0x1b0
eax: 00000000 ebx: 00000001 ecx: cf2ef0a0 edx: 00000000
esi: cf2ef6a0 edi: 00000001 ebp: 00000000 esp: ccc0fe64
ds: 007b es: 007b ss: 0068
Process modprobe (pid: 1941, threadinfo=ccc0e000 task=cd250ca0)

Stack: 00000001 cf2ef6a0 00000001 cf2ef6a0 c0229778 cf2ef6a0 00000001
00000000
       00000000 ccc0fee4 cf2ef6a0 c0229abe cf2ef6a0 00000001 cf2ef6c4
ccc0fee4
       cf2ef6a0 cf2ef6e4 c0229b59 cf2ef6a0 ccc0fee4 cf2ef6a0 00000000
ccc0e000
Call Trace:
 [<c0229778>] __cpufreq_governor+0x68/0x120
 [<c0229abe>] __cpufreq_set_policy+0xde/0x140
 [<c0229b59>] cpufreq_set_policy+0x39/0x70
 [<c02292c5>] cpufreq_add_dev+0x185/0x2a0
 [<c014d74c>] unmap_vm_area+0x2c/0x80
 [<c01d8a83>] sysdev_driver_register+0x83/0xe0
 [<c0229e21>] cpufreq_register_driver+0x91/0xa0
 [<d082e4e4>] longhaul_init+0x54/0x56 [longhaul]
 [<c013677d>] sys_init_module+0x10d/0x230
 [<c010b247>] syscall_call+0x7/0xb

Code: 0f 0b f6 01 60 66 2a c0 eb 82 0f 0b f4 01 60 66 2a c0 e9 6e

/sys/devices/system/cpu/cpu0/cpufreq:

[root@paradox cpufreq]# more *
::::::::::::::
cpuinfo_max_freq
::::::::::::::
0
::::::::::::::
cpuinfo_min_freq
::::::::::::::
0
::::::::::::::
scaling_available_governors
::::::::::::::
userspace
::::::::::::::
scaling_driver
::::::::::::::
longhaul
::::::::::::::
scaling_governor
::::::::::::::
userspace
::::::::::::::
scaling_max_freq
::::::::::::::
0
::::::::::::::
scaling_min_freq
::::::::::::::
0

I'll check if I can fix the min/max freq. Is Longhaul v1 correct for this CPU? With cpufreq from 2.4 it always displays v2.

Jurgen

*boggle*, we had these fixed I thought..

I'm going to need you to litter the code with printk's to find out why your MSRs are returning zeros. Contact me off list if you want a patch to do this, otherwise.. don't be afraid to go overboard, the more debug info the better.

According to the documentation, yes.

Dave

 Dave Jones     http://www.codemonkey.org.uk