On Fri, Mar 17, 2000 at 05:09:51PM -0800, Jean Tourrilhes wrote:
> Hi everybody...
>
> <Note : I'm not suscribed to the mailing list>
>
> My trusty P90 (54MB, CMD640) is not very happy with the latest
> 2.3.X kernels (2.3.40 -> 2.3.99-pre1). The following bugs have been
> verified with kernel 2.3.99-pre1 just a few minute ago...
Hi again,
I decided to give my poor P90 a well deserved rest, and I
upgraded to a PP150.
The beast :
o Pentium Pro 150
o 64 MB RAM ECC
o 450KX/GX chipset
o Adaptec AIC-7880U SCSI controller
I've compiled a 2.3.99-pre5 kernel with mostly usual options,
most as modules, but nothing fancy (no SMP, no devfs, ...). Distrib is
Debian 2.2 (meet all the required changes).
1) If I compile the kernel with support for uniprocessor APIC,
the kernel hang at boot. Simple.
2) If I compile the kernel without support for uniprocessor
APIC, I boot and run Linux with following problems :
2a) Error boot messages (also under 2.2.X) :
---------------------
Apr 25 18:57:40 lagaffe kernel: PCI: Device 00:a0 not found by BIOS
Apr 25 18:57:40 lagaffe kernel: PCI: Device 00:c8 not found by BIOS
---------------------
Probably harmless...
2b) If I leave the box overnight, it crashes :
-----------------------------------
stack segment: 0000
CPU: 0
EIP: 0010:[kmem_cache_grow+811/1036]
EFLAGS: 00010296
eax: 0000001f ebx: c00eff00 ecx: c013bcfd edx: 00000008
esi: c1127be0 edi: c00ef17c ebp: ffffffff esp: c2df7e60
ds: 0018 es: 0018 ss: 0018
Process find (pid: 310, stackpage=c2df7000)
Stack: c1127be8 c1127be0 00000007 00000282 c00ef2f8 c1127be8 c00ef000 00000008
00000000 00000206 00000001 00000007 00000000 c0125dcb c1127be0 00000007
00000000 c02751e0 c02751e0 c1172c00 c013c7cd c1127be0 00000007 00000000
Call Trace: [kmem_cache_alloc+375/456] [get_new_inode+25/312] [iget4+174/188] [ext2_lookup+91/132] [real_lookup+143/324] [walk_name+843/1220] [lookup_dentry+44/72]
[__namei+42/104] [sys_newlstat+14/100] [system_call+52/56]
Code: 89 45 00 8b 6d 00 83 6c 24 1c 01 0f 83 04 ff ff ff c7 45 00
stack segment: 0000
CPU: 0
EIP: 0010:[kmem_cache_grow+811/1036]
EFLAGS: 00010296
eax: 00000073 ebx: c00ecfc0 ecx: 00000000 edx: 0000001d
esi: c11279e0 edi: c00ec040 ebp: ffffffff esp: c2de5e0c
ds: 0018 es: 0018 ss: 0018
Process sort (pid: 308, stackpage=c2de5000)
Stack: c11279e8 c11279e0 00000003 00000282 c00ec138 c11279e8 c00ec040 0000001d
00000000 00000202 00000001 00000003 00000040 c0125dcb c11279e0 00000003
c2ddef60 00000400 c2ddef60 00000400 c012ca8d c11279e0 00000003 c2ddef60
Call Trace: [kmem_cache_alloc+375/456] [get_unused_buffer_head+57/184] [create_buffers+32/784] [timer_bh+166/648] [create_empty_buffers+24/112] [__block_prepare_write+84/520] [block_prepare_write+34/60]
[ext2_get_block+0/1164] [ext2_prepare_write+25/32] [ext2_get_block+0/1164] [generic_file_write+830/1224] [sys_write+192/224] [system_call+52/56]
Code: 89 45 00 8b 6d 00 83 6c 24 1c 01 0f 83 04 ff ff ff c7 45 00
...
-----------------------------------
In fact, this is almost identical to the crash I reported
earlier on my P90 (March), which is quite a different achitecture (P5
vs P6, IDE vs SCSI).
2c) When rebooting, it crashes in the fsck and can never
complete a single fsck of the first partition (1GB). At this point,
modules are not loaded and daemons not running, so there is not much
stuff we can blame. 2.2.14 does the fsck without any fuss.
2d) If I do some kernel compilation, big ftp or other stuff, I
can also crash the box without effort. Oops look the same.
2e) It seems that the kernel refuse to use the swap
space. "cat /proc/meminfo" always show use of 0B. 2.2.14 use the swap
without any trouble (and swap < 128 MB). Anyway, when the crash occur,
I still have some free memory :
------------------
total: used: free: shared: buffers: cached:
Mem: 64036864 48467968 15568896 0 720896 37167104
Swap: 115146752 0 115146752
MemTotal: 62536 kB
MemFree: 15204 kB
MemShared: 0 kB
Buffers: 704 kB
Cached: 36296 kB
HighTotal: 0 kB
HighFree: 0 kB
LowTotal: 62536 kB
LowFree: 15204 kB
SwapTotal: 112448 kB
SwapFree: 112448 kB
------------------
Note that after the crash, the amount of "cached" doesn't want
to go down and seem to only increase. Weird.
3) 2.2.14 running without any trouble on this box. I need to
go back to 2.2.14 to complete fsck and my kernel compilations. In
fact, I'm running 2.2.14 for months on a similar box.
I'm willing to provide more information and help investigate
the crash, but remember that I'm not on the mailing list...
Meanwhile, have fun...
Jean
P.S. : I didn't tried Pcmcia yet on this box. That was enoug bad news
for the week...
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Sun Apr 30 2000 - 21:00:11 EST