Help w/ SuSE 6.3 install crash - or any jobs open?

Help w/ SuSE 6.3 install crash - or any jobs open?

Post by Chris Wre » Sun, 16 Apr 2000 04:00:00



I begged my boss to let me set up a new Linux box
as a workstation under the assumption that we could
replace my aging p75 (32m ram 400m hd) and also
use it  as a lightweight server (mostly for CVS).
He bought me a nice shiny new box after I shot down
two or three WinJunked configurations he found - and
now, I can't get SuSE 6.3 to even install on this
beast without it crashing, currently during
HD partitioning step.

Things I know:

1) the box is at least somewhat functional - as
   a desperation test around three this morning
   I temporarily installed Win95 - worked fine.
   (yes, that's not the same, but it does demonstrate
   some functionality)

2) I may need to come stay with you for a few
   weeks until I find another job after my boss
   gets in Monday morning and finds we do not
   have the promised Linux wonderland.

3) Having maybe a dozen painless installs under
   my belt does NOT make me the Linux install
   whiz kid I thought I was.

The System in question:
=============================
 MotherBoard: MicroStar K7 Pro
 on-board IDE  AMD 756 chipset
 CPU:         AMD Athlon 600
 Ram:         256m
 Video:       Matrox G400 32meg
           (yeah, I know, 1 head unless I buy Xi)
 HD: two 20g IDE FA520S60 Maxtor 92041U4 ultraATA/66
              UDMA mode 4
 Sound: SB PCI16 (Ensoniq AudioPCI ES1371)
 Ethernet : Intel EtherExpressPro 100+ (82557 ctrlr)
=============================

Some of my BIOS settings (I'm probably
leaving out the ones you need):
==============================
 AMI (c) 1998
 BIOS Drive Autodect -
                             LBA   Blk  PIO 23Bit
 Size  Cyln  Head WpCom Sec  Mode Mode Mode Mode
 20492 39703 16   0     63   On   On   4    On

 Quick Boot                Enabled
 Add-On ROM Init           Force BIOS
 Hard Disk Access Control  Read-Write
 S.M.A.R.T. for Hard Disks Enabled
 Internal Cache WriteBack
 External Cache WriteBack
 L2 Cache ECC   TableDefault
 System BIOS Cacheable DISABLED
 all Shadows are       DISABLED

 SDRAM Timing

 Configure Timing by SPD   Enabled
 PH Limit                  23 Cycles
 Idle Cycle Limit           8 Cycles
 TRP SRAS Precharge         3 Cycles
 TRAS Timing                5 Cycles
 CAS Latency                3 Cycles

 TRC Bank Cycle Timing      3 Cycles
 TRCD Tiing                 4 Cycles
  Super Bypass Mode        Disabled

 DRAM Integrity Mode       Disabled
 DRAM Burst Refresh        Enabled
 Memory Hole               Disabled
 Graphics Aperture Size    128MB
 MDA Support               No
 USB Function              Enabled
 USB KB/Mouse Support      Enabled
========================================

Raw (no boot parameters) Suse 6.3 boot died
struggling to come up fairly quickly:
==============================================
 Calibrating delay loop... 602.93 BogoMIPS
 Memory: 63124k/66556k available (1260k kernel code, \
      412k reserved, 1004k data, 44k init, 0k bigmem
 DENTRY hash table entries: 16384 (order: 5, 131072 bytes)
 Buffer-cache has table entries: 65536 (order: 6, 262144 bytes)
 Unable to handle kernel paging request at virtual address c6205f4f
 current->tss.cr3 = 00101000, %cr3 = 00101000
 #pde = 00000000
 Oops = 0000
 CPU:  0
 EIP:  0010:[<c0120a37>]
 EFLAGS: 00010206

 register dump....
 Process swapper (pid: 0, process nr: 0, stackpage=c025f000)
 Stack dump....
 Call Trace: ....
 Code: ....
 Kernel panic: Attempted to kill the idle task!
 In swapper task - not syncing
========================================

OK, right off the top, I didn't like the memory
it was reporting - I popped the case and looked,
and sure enough, I still had the two 128m sticks
in there that we paid for - and the BIOS and W95
had reported - So I tried booting with:

linux mem=256M

which got me a little further.  This time I
got through to initting the swap partition -
and that's as far as I've been able to get it.
When the system inits the swap partition, it
writes to the HD for a few seconds, then dies.
Sometimes it's a complete lockup - nothing will
do but the three-finger salute. (or reset button)
A couple of  times I was able to ctrl-alt to
another console and get the following information:
========================================
 Unable to find swap-space signature
 Unable to handle kernel paging request at virtual address f0248980
 current->tss.cr3 = 0f688000, %cr3 = 0f688000
 *pde = 00000000
 Oops: 0002
 CPU: 0
 EIP: 0010: [<c0126776>]
 EFLAGS: 00010217
 eax: 30000000    ebx: 0000000    ecx: c6353010       edx: 0804cd68
 esi: c635e010    edi 0c000000    ebp: cf685e9c       esp: cf685d04
 ds:  0018   es: 0018   ss: 0018
 Process mkswap (pid: 93, process nr: 14, stackpage=cf685000)

 stack dump...
 Call Trace...
 Code: ....

 Unable to find swap-space signature
=================================================

I have noticed that the SuSE info screen always misreports
the drive geometry in the HardDisks / CD-Roms system info
selection but then the Kernel messages (I believe this
screen is a dmesg capture) and Linux fdisk report an entirely
different geometry!
Without touching anything:
==================================
                Cyl     Head  Sec
BIOS:           39703    16    63
HD/CDrom Info:  16383    16    63
dmesg:           2491   255    63
fdisk:           2491   255    63
=================================

Well, apparently the mem= trick worked, so I why not
make a good param list a little longer?  This time,

linux mem=256M hda=39703,16,63 hdb=39703,16,63

This yielded:
==================================
                Cyl     Head  Sec
BIOS:           39703    16    63
HD/CDrom Info:  16383    16    63
dmesg:          39703    16    63
fdisk:          39703   255    63
=================================

That looked somewhat better, although I have
no idea where SuSE is getting the info for the
HD/CDrom Info panel.  It may have LOOKED better,
but it didn't WORK any better.  Still crashed on
either swap init (most common) or sometimes it
will live long enough to die formatting one of
the ext2 partitions.

I have also tried forcing my BIOS to:
    Cyl     Head  Sec
     2491   255    63

just to try for a little agreement from the
other direction.  This was also NDG.

Folks, I need a clue.  I'd like to show these
WinZealots what a Linux box can do - of course,
I'd also kind of like to be employed after Monday.

Any ideas?

Chris -

 wChris (at) ubginc.com

 
 
 

Help w/ SuSE 6.3 install crash - or any jobs open?

Post by Peter T. Breu » Mon, 17 Apr 2000 04:00:00


: two or three WinJunked configurations he found - and
: now, I can't get SuSE 6.3 to even install on this
: beast without it crashing, currently during

Whilst your post is generally very informative, it doesn't provide any
clues. We'd need to see all the logging output you can get from a
variety of differnt kernel probes, not just the one. As it is, the only
advice I can give is to yank out your memory and try again. Also yank
out your cards and try again.

: The System in question:
: =============================
:  MotherBoard: MicroStar K7 Pro
:  on-board IDE  AMD 756 chipset
:  CPU:         AMD Athlon 600
:  Ram:         256m
:  Video:       Matrox G400 32meg
:            (yeah, I know, 1 head unless I buy Xi)
:  HD: two 20g IDE FA520S60 Maxtor 92041U4 ultraATA/66
:               UDMA mode 4
:  Sound: SB PCI16 (Ensoniq AudioPCI ES1371)
:  Ethernet : Intel EtherExpressPro 100+ (82557 ctrlr)
: =============================

: ==============================================
:  Calibrating delay loop... 602.93 BogoMIPS
:  Memory: 63124k/66556k available (1260k kernel code, \

Looks like a full 64MiB used. Drop it to 32MiB via mem=32M at boot
and try again. Swap the sticks around.

:       412k reserved, 1004k data, 44k init, 0k bigmem
:  DENTRY hash table entries: 16384 (order: 5, 131072 bytes)
:  Buffer-cache has table entries: 65536 (order: 6, 262144 bytes)
:  Unable to handle kernel paging request at virtual address c6205f4f

This appears to be a genuine kernel space address.

:  current->tss.cr3 = 00101000, %cr3 = 00101000
:  #pde = 00000000
:  Oops = 0000
:  CPU:  0
:  EIP:  0010:[<c0120a37>]
:  EFLAGS: 00010206

:  register dump....
:  Process swapper (pid: 0, process nr: 0, stackpage=c025f000)
:  Stack dump....
:  Call Trace: ....
:  Code: ....
:  Kernel panic: Attempted to kill the idle task!
:  In swapper task - not syncing
: ========================================

: OK, right off the top, I didn't like the memory
: it was reporting - I popped the case and looked,
: and sure enough, I still had the two 128m sticks
: in there that we paid for - and the BIOS and W95
: had reported - So I tried booting with:

: linux mem=256M

: which got me a little further.  This time I

I'd do the opposite! It looks clear that part of that memory is not
responding properly. Take it out.

: got through to initting the swap partition -
: and that's as far as I've been able to get it.
: When the system inits the swap partition, it
: writes to the HD for a few seconds, then dies.
: Sometimes it's a complete lockup - nothing will

Fairly clear memory or bus problem. Drop settings to more conservative
values and put in one element at a time and test it.

: do but the three-finger salute. (or reset button)
: A couple of  times I was able to ctrl-alt to
: another console and get the following information:
: ========================================
:  Unable to find swap-space signature
:  Unable to handle kernel paging request at virtual address f0248980
:  current->tss.cr3 = 0f688000, %cr3 = 0f688000
:  *pde = 00000000
:  Oops: 0002
:  CPU: 0
:  EIP: 0010: [<c0126776>]
:  EFLAGS: 00010217
:  eax: 30000000    ebx: 0000000    ecx: c6353010       edx: 0804cd68
:  esi: c635e010    edi 0c000000    ebp: cf685e9c       esp: cf685d04
:  ds:  0018   es: 0018   ss: 0018
:  Process mkswap (pid: 93, process nr: 14, stackpage=cf685000)

:  stack dump...
:  Call Trace...
:  Code: ....

: I have noticed that the SuSE info screen always misreports
: the drive geometry in the HardDisks / CD-Roms system info

Don't worry about it.

: selection but then the Kernel messages (I believe this
: screen is a dmesg capture) and Linux fdisk report an entirely
: different geometry!

Choose the one you like best.

: Without touching anything:
: ==================================
:                 Cyl     Head  Sec
: BIOS:           39703    16    63
: HD/CDrom Info:  16383    16    63
: dmesg:           2491   255    63
: fdisk:           2491   255    63
: =================================

: Well, apparently the mem= trick worked, so I why not
: make a good param list a little longer?  This time,

: linux mem=256M hda=39703,16,63 hdb=39703,16,63

: This yielded:
: ==================================
:                 Cyl     Head  Sec
: BIOS:           39703    16    63
: HD/CDrom Info:  16383    16    63
: dmesg:          39703    16    63
: fdisk:          39703   255    63
: =================================

: That looked somewhat better, although I have
: no idea where SuSE is getting the info for the
: HD/CDrom Info panel.  It may have LOOKED better,
: but it didn't WORK any better.  Still crashed on
: either swap init (most common) or sometimes it
: will live long enough to die formatting one of
: the ext2 partitions.

Don't bother with this. First verify your ram and bus. Then start
on the disk.

Peter

 
 
 

Help w/ SuSE 6.3 install crash - or any jobs open?

Post by Doc Shiple » Tue, 18 Apr 2000 04:00:00


Chris,
 Try sticking one DIMM in slot1 and one in slot3 (or 0 and 2; I don't
remember how the BioStar numbers 'em), leaving the middle slot empty. I
know that sounds crazy, but that's the only way my boss' Biostar with
Athlon 750 will boot Linux. It IS a memory issue, but both RAM and board
are solid. The Athlon boards are all notorious for memory-related
quirks. It's no surprise that Win95 doesn't notice. Win95 is only using
about 25% of your physical RAM.
 As far as the kernel finding only 64M without the append line, that's a
problem I ran into on a K6-2 FIC board, and did some research. Some of
the newer BIOS's do that, depending on the kernel version. But nobody I
found or heard from reported crash/malfuntion problems with it.
 I'd turn of the S.M.A.R.T disk in BIOS, too. I don't think the boot
kernel supports it. Other than that, I'd let the BIOS auto-detect the
drive and partition it normally.

 FWIW, any number of painless installs teaches ya pretty much nothing at
all. It's the dozen or so really agony-ridden installs that'll make you
the Linux-install Whiz Kid. Welcome to class!

--
 Doc Shipley
 Network Support
 TARL Labs, UT
 Austin, Texas