Kernel panic on booting with nVidia MCP55 [10de:037f]

[GoogleCodeExporter]

There is a another problem: kernel panic.
(the older kernel didn't do so)

Original issue reported on code.google.com by gleb.ax...@gmail.com on 31 Jan 2012 at 9:48

[GoogleCodeExporter]

can you make a picture of the panic, or set
console=0 from ">"-prompt / plan9.ini and capture 
it on the serial line?

or you might just write down the pc values...

is this 9pcf or 9pccpuf kernel? single processor or
multiprocessor system? there are just 4 changes to
the kernel. the "exec free(file0);", "procalloc race",
"added another ICH6 vid/did to sdide" and your "MPC55"
one.

except the MPC55 one, the others should not affect you.
but can you verity this by reverting to previous versions
and checking?

Original comment by cinap_le...@felloff.net on 3 Feb 2012 at 12:17

[GoogleCodeExporter]

9pcf, multiprocessor system (x86_64 AMD Athlon(tm) 64 X2 Dual Core
Processor 5000+ AuthenticAMD), *nomp=1
Making a picture of the panic is problematic.

Original comment by gleb.ax...@gmail.com on 3 Feb 2012 at 3:41

[GoogleCodeExporter]

...
panic: boot process died: main
dumpstack
ktrace /kernel/path f0109631 f004b910 <<EOF
estackx f004bbe0
f004b8b0=f01093c6 f004b8c8=f01e35cf f004b8d0=f01e35cf f004b8d4=f021080b
f004b8e8=f020db00 f004b8f8=f0109631 f004b90c=f0109631 f004b910=f01093ca
f004b918=f015af4a f004b934=f020df30 f004b95c=f020df30 f004b964=f020db19
f004b974=f020f6f0 f004b980=f020de65 f004b984=f020ebd4 f004b98c=f020ebd4
f004b998=f020ebd4 f004b9a8=f020dd2f f004b9d0=f020f6f0 f004b9dc=f01e3593
f004b9e4=f01e359f f004b9ec=f01fbba6 f004b9f8=f020fba0 f004ba04=f01f341e
f004ba0c=f01f3677 f004ba10=f01e3925 f004ba20=f01ed39d f004ba40=f01efdd6
f004ba54=f01e359f f004ba60=f01eb7a2 f004ba9c=f01fa856 f004bac0=f01eb83a
f004bad8=f01f8432 f004bb18=f01f848e f004bb20=f0108b20 f004bb3c=f0109b2f
f004bb54=f01f08b1 f004bb70=f01021b0 f004bb78=f01e31c0 f004bb8c=f0100bc2
f004bbc4=00000040 f004bbc8=f010070d f004bbcc=000038c5 f004bbd0=00000023
f004bbd4=00000286 f004bbd8=dfffef48 f004bbdc=0000001b
EOF

Original comment by gleb.ax...@gmail.com on 4 Feb 2012 at 1:37

[GoogleCodeExporter]

oh, this is interesting.

the important message is "boot process died".

this means that bootrc exits/crashes for some reason. you
can try *nodumpstack= so previous messages that might get
printed before the panic wont scroll off.

how mutch memory was detected? if memory detection fails,
rc executing bootrc might just get killed by oom killer.

whats the exact version you'r using? maybe i fucked up the
build and some files in the bootfs miss the execute bit...

Original comment by cinap_le...@felloff.net on 4 Feb 2012 at 4:02

[GoogleCodeExporter]

9pcf kernel from 9front-1320.1ad64419a229.iso boots just fine.

Original comment by cinap_le...@felloff.net on 4 Feb 2012 at 10:48

[GoogleCodeExporter]

Well, I have built new kernel (before R3b71bbc4bf62) without MCP55
support, and it didn't kernel panic. But after I have added the line
in sdide.c — it does. So, how i/o to sata hard disks can kill boot
process?

Original comment by gleb.ax...@gmail.com on 5 Feb 2012 at 9:32

[GoogleCodeExporter]

it shouldnt. my guess is that something else is wrong. maybe not all
memory is detected, causing it to run out of memory when we run
partdisk or something. what does the kernel print before panicing?
how mutch memory is deteted? whats the e820= line printed by the
bootloader?

Original comment by cinap_le...@felloff.net on 5 Feb 2012 at 8:28

[GoogleCodeExporter]

e820=0000000000000000 000000000009fc00 0000000000100000 00000000bffd0000
Plan 9
126 holes free
00018000 0009f000 552960
00352000 10000000 264953856
265506816 bytes free
cpu0: 2600MHz AuthenticAMD AMD-K8 (cpuid: AX 0x40F32 DX 0x178BFBFF)
ELCR: 0CA0
pcirouting: ignoring south bridge PCI.0.1.0 10DE/0362
3072M memory: 256M kernel data, 2815M user, 3440M swap
panic: boot process died: main

Original comment by gleb.ax...@gmail.com on 5 Feb 2012 at 9:18

[GoogleCodeExporter]

wired. try commenting out the showlocaldevs line in /sys/src/9/boot/bootrc
rc script. maybe add some debug echos to see how far the script comes.

then:

cd /sys/src/9/pc
mk clean
mk install
9fs 9fat
cp 9pcf /n/9fat

if it goes to the bootargs quesiton, type !rc and inspect /dev/sd*,
cat the ctl files. maybe theres something that confuses the script.

Original comment by cinap_le...@felloff.net on 5 Feb 2012 at 11:57

[GoogleCodeExporter]

in 9/boot/bootrc gives no effect. This is strange.

Original comment by gleb.ax...@gmail.com on 7 Feb 2012 at 1:59

[GoogleCodeExporter]

maybe theres a problem how you build the kernel? forgot
to install new kernel in 9fat? try mk clean in /sys/src/9/pc

Original comment by cinap_le...@felloff.net on 8 Feb 2012 at 12:11

[GoogleCodeExporter]

% cd pc/
% mk clean
% mk install
compile in 9vx.

Original comment by gleb.ax...@gmail.com on 8 Feb 2012 at 1:04

[GoogleCodeExporter]

9vx builds are broken. because a bug in 9vx wstat implementation.
it fails when building the bootfs.paq file when copying the files
needed for the bootfs.

see Issue 63:
see 
http://code.google.com/p/plan9front/issues/detail?id=63&colspec=ID%20Type%20Stat
us%20Priority%20Milestone%20Owner%20Summary%20Subsystem

Original comment by cinap_le...@felloff.net on 8 Feb 2012 at 1:51

[GoogleCodeExporter]

So, I have to use qemu to build new kernel? Ok, I shall try do this.

Original comment by gleb.ax...@gmail.com on 8 Feb 2012 at 5:55

[GoogleCodeExporter]

fixed: I used an broken build (one of it has got by hg?).

Original comment by gleb.ax...@gmail.com on 8 Feb 2012 at 7:55

[GoogleCodeExporter]

or patch 9vx. :)

Original comment by cinap_le...@felloff.net on 8 Feb 2012 at 9:29

[GoogleCodeExporter]

Original comment by cinap_le...@felloff.net on 9 Feb 2012 at 2:39

• Changed state: Fixed