[100] in linux-scsi channel archive
Problems with sr.c in Linux 1.2.0 (fwd)
daemon@ATHENA.MIT.EDU (Roderick Smith)
Tue Mar 14 17:35:38 1995
Date: Tue, 14 Mar 1995 11:04:27 -0700 (MST)
From: Roderick Smith <rsmith@psych.colorado.edu>
To: linux-scsi@vger.rutgers.edu
I've sent the following message to people whose names appear in the
relevant source code, but I recently got a suggestion to send it to
linux-scsi@vger.rutgers.edu, so I have....
--Rod Smith
RSMITH@PSYCH.COLORADO.EDU
---------- Forwarded message ----------
Date: Tue, 14 Mar 1995 08:50:01 -0700 (MST)
From: Roderick Smith <rsmith@psych.colorado.edu>
To: kraxel@cs.tu-berlin.de, drew@colorado.edu, eric@aib.com
Subject: Problems with sr.c in Linux 1.2.0
I'm sorry to have to send this to three people, but all your names were
in the file in question. Anyhow, I've got a bug to report in the sr.c
file (or, at least, I fixed it by commenting out part of that file).
Here's a description of the problem:
First, my hardware:
Asus PCI-486SP3G motherboard, using the Intel Saturn II chipset and
NCR 53c810 SCSI controller
Two SCSI-2 hard disks, a Seagate ST3390N (ID0) and a Quantum LTS730S (ID1)
One SCSI-1 CD-ROM, a NEC CDR-84 (ID2)
One SCSI-1 tape drive, an Archive Viper 2525S (ID6)
One 3.5" floppy (connected to controller on the Asus board)
One SoundBlaster 8-bit board
16MB RAM
Linux boot and root partition on ID1, /usr partition on ID0, booting via
OS/2's Boot Manager, with LILO on the root partition
I've encountered no conspicuous crashes with this hardware under
1.1.59; it all works fine, including the CD-ROM drive (though I've
never tried using the sound card, FWIW, under Linux).
When I configured the kernel compilation using "make config", I
included support for the 53c810 controller (and specified support for
SCSI hard disks, CD-ROMs, tape drives, and "generic" devices), the
floppy drive, the SoundBlaster, and various filesystems (ext2fs, FAT,
HPFS, ISO9660, /proc, and maybe one or two others). Since I'd been
using the "default" Slackware kernel, I don't know precisely how my
1.1.59 system was configured.
Anyhow, when I try to mount a CD-ROM with the 1.2.0 kernel, the system
crashes, and reports filesystem errors on reboot (<Ctrl-Alt-Del> is
useless). When I try doing "mount /dev/scd0 /cdrom" as root, I get
the following mess of messages:
mount: block device /dev/scd0 is write-protected, mounting read-only
Oops: 0000
EIP: 0010:0018846f
EFLAGS: 00010002
eax: 00000000 ebx: 001bc4a4 exc: 00000004 edx: 0000000a
esi: 001bc460 edi: 00000000 ebp: fc000000 esp: 00215f50
ds: 0018 es: 0018 fs: 002b gs: 002b ss: 0018
Process klogd (pid: 34, process nr: 8, stackpage = 00215000)
Stack: 0000d002 00000000 001bc4a4 001bc460 0000000e 00000000 00fec400 00000008
0000004e bffff601 00126104 0091d000 00188be1 001bc460 00000000 00000400
000014d8 0000178c 00000000 00000000 0011d63e 00000001 bffff280 00000001
Call Trace: 00126104 00188be1 0011d63e 00112d0f 001124c3
Code: 8b 40 04 89 44 24 14 0f b6 50 09 89 54 24 1c 52 8b 4c 24 18
Aiee, killing interrupt handler
When I try to boot with a CD-ROM in the drive, I get something very
similar to the above (the esp: number is 001933c4, it's process
swapper, not klogd, and some of the stack and so on numbers are
different), but I also get some preceding messages:
scsi0: phase mismatch interrupt occurred with no current command
scsi0: DANGER: abort_connected() called
Unable to handle kernel NULL pointer dereference at virtual address c0000004
current->tss.cr3 = 00101000, %cr3 = 00101000
*pde = 00102067
*pte = 00000027
Following a tip I read on one of the Linux newsgroups, I commented out
the portion of the sr_photocd() function in this file that relates to NEC
CD-ROMs (case SCSI_MAN_NEC), and it now works. I find this a bit odd,
since I've got a NEC CD-ROM; but presumably there's a bug in there that
interacts with my specific model of NEC (CDR-84, SCSI-1 unit) or with my
SCSI controller (NCR 53c810).
If there's someplace else I should report this bug, please tell me.
Thanks.
--Rod Smith
RSMITH@PSYCH.COLORADO.EDU
