Bug#265520: [i386][rc1][floppy,netinst] aic7xxx crash, partition table CHS damage
Package: installation-reports
INSTALL REPORT
Debian-installer-version: RC1 (20040801) from
http://wftp.tu-chemnitz.de/pub/linux/debian/debian/dists/sarge/main/installer-i386/rc1/images/floppy/{b,r}oot.img
uname -a: Linux guitar 2.4.26-1-586tsc #1 Thu Jul 22 12:53:42 JST 2004 i586
GNU/Linux
Date: 2004-08-12 16:00 CEST
Method:
- How did you install?
network
- What did you boot off?
2 floppies (boot,root), no additional driver disks
- If network install, from where?
ftp.de.debian.org
- Proxied?
yes, squid proxy, no authentication needed
Machine: custom-built PC
Processor: Pentium 100
Memory: 64MB
Root Device: /dev/hda5
Root Size/partition table:
---
Disk /dev/hda: 160.0 GB, 160041885696 bytes
255 heads, 63 sectors/track, 19457 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Device Boot Start End Blocks Id System
/dev/hda1 * 1 65 516064+ 6 FAT16
Partition 1 does not end on cylinder boundary.
/dev/hda2 65 19458 155774777 f W95 Ext'd (LBA)
/dev/hda5 65 130 524632+ 83 Linux
/dev/hda6 130 130 472+ 6 FAT16
/dev/hda7 130 391 2097112+ 6 FAT16
/dev/hda8 391 652 2097112+ 7 HPFS/NTFS
/dev/hda9 652 4096 27660496+ 8e Linux LVM
/dev/hda10 4096 4112 131008+ 82 Linux swap
/dev/hda11 4112 8311 33729664+ 8e Linux LVM
/dev/hda12 8311 12511 33730168+ 8e Linux LVM
/dev/hda13 12511 16710 33730168+ 8e Linux LVM
/dev/hda14 16710 19458 22073656+ 8e Linux LVM
---
Oops, the installer changed the CHS geometry! :-(
It used to be 16 heads/63 sectors, now it is 255 heads/63 sectors.
I already wondered during installation why the installer wanted to write
to the partition table at all, since I hadn't changed any of the
partitions.
This problem was also reported to occur in Fedora Core 2:
http://lwn.net/Articles/86835/
The sfdisk fix mentioned there didn't work for me, however.
I fixed it (hopefully without data loss) by removing all partitions,
changing the number of heads, and then re-creating every partition with
exactly the same size and type as before. Fortunately, I had kept a copy
of the original "fdisk -l" output, as before the installation.
This is the partition table as intended:
---
Disk /dev/hda: 160.0 GB, 160041885696 bytes
16 heads, 63 sectors/track, 310101 cylinders
Units = cylinders of 1008 * 512 = 516096 bytes
Device Boot Start End Blocks Id System mount point
/dev/hda1 * 1 1024 516064+ 6 FAT16 /mnt/dos
/dev/hda2 1025 310101 155774808 f W95 Ext'd (LBA)
/dev/hda5 1025 2065 524632+ 83 Linux /
/dev/hda6 2066 2066 472+ 6 FAT16 /mnt/spacer
/dev/hda7 2067 6227 2097112+ 6 FAT16 /mnt/win98
/dev/hda8 6228 10388 2097112+ 7 HPFS/NTFS /mnt/nt
/dev/hda9 10389 65270 27660496+ 8e Linux LVM
/dev/hda10 65271 65530 131008+ 82 Linux swap
/dev/hda11 65531 132454 33729664+ 8e Linux LVM
/dev/hda12 132455 199379 33730168+ 8e Linux LVM
/dev/hda13 199380 266304 33730168+ 8e Linux LVM
/dev/hda14 266305 310101 22073656+ 8e Linux LVM
---
logical volumes:
LV VG Attr LSize mount point
lv_afscache vg0 -wn-ao 224.00M /var/cache/openafs
lv_home vg0 -wn-ao 20.53G /home
lv_mirror vg0 -wn-ao 20.00G /mnt/mirror
lv_rescue vg0 -rn-ao 59.38G /mnt/rescue
lv_restore_2003-12-22 vg0 -rn-a- 1.09G /mnt/restore_2003-12-22
lv_tmp vg0 -wn-ao 256.00M /tmp
lv_usr vg0 -wn-ao 6.00G /usr
lv_usr_local vg0 -wn-ao 512.00M /usr/local
lv_usr_src vg0 -wn-ao 1.00G /usr/src
lv_var vg0 -wn-ao 1.00G /var
Output of lspci and lspci -n:
# lspci
0000:00:00.0 Host bridge: Intel Corp. 430FX - 82437FX TSC [Triton I] (rev 02)
0000:00:07.0 ISA bridge: Intel Corp. 82371FB PIIX ISA [Triton I] (rev 02)
0000:00:07.1 IDE interface: Intel Corp. 82371FB PIIX IDE [Triton I] (rev 02)
0000:00:09.0 Ethernet controller: Digital Equipment Corporation DECchip 21140
[FasterNet] (rev 22)
0000:00:0a.0 VGA compatible controller: Matrox Graphics, Inc. MGA 2164W
[Millennium II]
0000:00:0b.0 Multimedia audio controller: Creative Labs SB Live! EMU10k1 (rev
04)
0000:00:0b.1 Input device controller: Creative Labs SB Live! MIDI/Game Port (rev
01)
0000:00:0c.0 SCSI storage controller: Adaptec AHA-2940U/UW/D / AIC-7881U
# lspci -n
0000:00:00.0 0600: 8086:122d (rev 02)
0000:00:07.0 0601: 8086:122e (rev 02)
0000:00:07.1 0101: 8086:1230 (rev 02)
0000:00:09.0 0200: 1011:0009 (rev 22)
0000:00:0a.0 0300: 102b:051b
0000:00:0b.0 0401: 1102:0002 (rev 04)
0000:00:0b.1 0980: 1102:7002 (rev 01)
0000:00:0c.0 0100: 9004:8178
Base System Installation Checklist:
[O] = OK, [E] = Error (please elaborate below), [ ] = didn't try it
Initial boot worked: [E]
Configure network HW: [E]
Config network: [O]
Detect CD: [ ]
Load installer modules: [O]
Detect hard drives: [O]
Partition hard drives: [E]
Create file systems: [E]
Mount partitions: [E]
Install base system: [O]
Install boot loader: [O]
Reboot: [O]
Comments/Problems:
Initial boot, aic7xxx:
- several SCSI resets during device discovery, finally hung the machine
(not even console switching worked anymore), this was reproducible
- retried with all SCSI devices' power unplugged -> success
- once the install was finished, I shut down to re-plug the SCSI CDRW
drive, which works fine as always
network HW:
- installer used tulip.o -> no ethernet link
- needed to unplug and re-plug network cable after the interface is "up",
this restarts autonegotiation and establishes the ethernet link
- installer should have used de4x5.o instead (for this particular
ethernet card)
- I couldn't find an rmmod command on the installer shell
for removing the unwanted module tulip.o
partition hard drives:
CHS geometry changed, although I didn't want to repartition at all
see above under Root Size/partition table
filesystems/mount:
I prepared the harddisk in another machine (partitions, logical volumes)
and then told the installer where to mount these.
Result: some partitions already containing FAT 16 filesystems and
the read-only LVs were rejected as containing errors, although
fsck.msdos or fsck.ext2 now say they are OK.
Creating a FAT 16 filesystem also failed on one of the partions:
/dev/hda6, if I remember correctly.
my workaround: leave them unused during installation,
edit /etc/fstab afterwards.
additional system information:
# cat /proc/scsi/scsi
Attached devices:
Host: scsi0 Channel: 00 Id: 06 Lun: 00
Vendor: SONY Model: CD-RW CRX140S Rev: 1.0e
Type: CD-ROM ANSI SCSI revision: 04
# further hardware (not discovered automatically)
- ISA Soundcard, modprobe ad1816
isapnp: Scanning for PnP cards...
isapnp: Card 'Analog Devices AD1815'
isapnp: 1 Plug & Play card detected total
ad1816: Analog Devices AD1815 detected
ad1816: ISAPnP reports 'Analog Devices AD1815' at i/o 0x100, irq 5, dma 1, 3
ad1816: AD1816 sounddriver Copyright (C) 1998 by Thorsten Knabe
ad1816: io=0x100, irq=5, dma=1, dma2=3, clockfreq=33000, options=0 isadmabug=0
- ISA Ethernet card, modprobe ne.o io=0x300
ne.c:v1.10 9/23/94 Donald Becker (becker@scyld.com)
Last modified Nov 1, 2000 by Paul Gortmaker
NE*000 ethercard probe at 0x300: 00 80 ad 17 11 b8
eth1: NE2000 found at 0x300, using IRQ 10.
# cat /proc/interrupts
CPU0
0: 5889055 XT-PIC timer
1: 11178 XT-PIC keyboard
2: 0 XT-PIC cascade
4: 1557 XT-PIC serial
5: 0 XT-PIC SoundPort
8: 6 XT-PIC rtc
9: 1327 XT-PIC aic7xxx, EMU10K1, eth0
10: 1 XT-PIC NE2000
14: 43256 XT-PIC ide0
NMI: 0
LOC: 0
ERR: 0
MIS: 0
additional notes:
read-edid and mdetect were automatically installed, but when running
aptitude for the first time, they were atomatically removed again.
The partitioner showed only the physical partitions, not the logical
volumes. That was, until I discovered that I needed to scroll down. :-)
I find the display of logical volumes confusing. There are two
selectable lines for each LV, only the second of which is suitable for
defining the mount point - the first will give a strange error like
"this is a whole disk". Maybe the first line should be made
non-selectable if it doesn't have any other than informational purpose.
The partitioning interface feels slow.
But maybe this wasn't written for a 100MHz machine.
I couldn't find a way to use a local mirror for the first part of a
network installation. But since I had fast Internet connection,
I didn't look further and used one of the mirrors offered.
Not really installer related:
The debconf dialog of ca-certificates is wider than the 80-column
screen, but this is rather cosmetic, since mainly the frame is off-screen.
Thanks for reading this far.
Mirko
Reply to: