Red Hat Bugzilla – Bug 52023
system hangs when mounting IDE CD writer
Last modified: 2007-04-18 12:36:02 EDT
Description of Problem:
I have two IDE CD drives in my system, the first (secondary master - hdc)
is a Mitsumi FX-240 CD reader, and the second (secondary slave - hdd/sr0)
is a Ricoh MP-7083A CD writer.
Mounting, unmounting, and reading using the Mitsumi reader works fine,
but trying to mount a disk on the Ricoh writer causes the system to hang.
I first noticed this when inserting a CD in KDE 2, where the system
attempted to automount the CD, but issuing "mount /mnt/cdrom1" also hangs
(as does "mount /dev/sr0 /mnt/cdram").
The CD writer is an IDE drive using SCSI emulation, as configured by Red
Note that in trying to diagnose this problem, I renamed /dev/cdrom1 to
/dev/cdram (which makes more sense to me), and removed the "kudzu"
parameter to the CD-ROM drives from /etc/fstab. The hang occurred before
(and after) making these changes.
LABEL=/ / ext2 defaults 1
LABEL=/boot /boot ext2 defaults 1
/dev/fd0 /mnt/floppy auto noauto,owner 0
LABEL=/var /var ext2 defaults 1
LABEL=/home /home ext2 defaults 1
none /proc proc defaults 0
none /dev/pts devpts gid=5,mode=620 0
/dev/hda7 swap swap defaults 0
/dev/cdrom /mnt/cdrom auto noauto,owner,ro 0
/dev/cdram /mnt/cdram auto noauto,owner,ro 0
/etc/sysconfig/hwconf previously contained something like:
desc: Mitsumi FX-240
desc: Ricoh MP-7083A CD-R/CD-RW
However I deleted hwconf, and kudzu has not recreated entries for either
/etc/lilo.conf contains the line:
ppp_async 6768 0 (autoclean)
ppp_generic 18032 0 (autoclean) [ppp_async]
sb 7824 1 (autoclean)
sb_lib 35760 0 (autoclean) [sb]
uart401 6608 0 (autoclean) [sb_lib]
sound 62112 1 (autoclean) [sb_lib uart401]
soundcore 4432 5 (autoclean) [sb_lib sound]
tdfx 57840 1
autofs 11136 1 (autoclean)
ne2k-pci 4864 1 (autoclean)
8390 6752 0 (autoclean) [ne2k-pci]
ipchains 38944 0 (unused)
ide-scsi 8352 0
scsi_mod 94336 1 [ide-scsi]
ide-cd 27104 1
cdrom 27392 0 [ide-cd]
usb-uhci 20848 0 (unused)
usbcore 49632 1 [usb-uhci]
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. issue "mount /mnt/cdrom1"
the system hangs (does not respond to local mouse or keyboard events)
the filesystem is mounted
this bug may lie in
running "cdrecord -scanbus" shows:
Cdrecord 1.9 (i686-pc-linux-gnu) Copyright (C) 1995-2000 Jvrg Schilling
Linux sg driver version: 3.1.17
Using libscg version 'schily-0.1'
0,0,0 0) 'RICOH ' 'CD-R/RW MP7083A ' '1.20' Removable CD-ROM
0,1,0 1) *
0,2,0 2) *
0,3,0 3) *
0,4,0 4) *
0,5,0 5) *
0,6,0 6) *
0,7,0 7) *
i would like to try using the CD writer as an IDE drive, but am not sure
how to (kudzu seems to always set the drive as a SCSI-IDE drive).
i will try to add more information as i find it, but it's annoying since
reproducing the bug hangs the system :-)
I just tried to issue mount with no CD in the drive (trying to avoid crashing
the system), and noticed this in the syslog:
Aug 19 10:37:16 air kernel: Attached scsi CD-ROM sr0 at scsi0, channel 0, id
0, lun 0
Aug 19 10:37:16 air kernel: sr0: scsi3-mmc drive: 32x/32x writer cd/rw
xa/form2 cdda tray
Aug 19 10:37:19 air kernel: cdrom: open failed.
"mount -t iso9660 /dev/hdd /mnt/cdram" gives:
mount: wrong fs type, bad option, bad superblock on /dev/hdd,
or too many mounted file systems
(could this be the IDE device where you in fact use
ide-scsi so that sr0 or sda or so is needed?)
/dev/cdram (previously /dev/cdrom1) is a symlink to /dev/scd0
primary master - Quantum UDMA 33 hard disk
secondary master - Mitsumi FX240 CD reader
secondary slave - Ricoh MP7083A CD reader/writer
mount itself can't lock up the system - this must be happening in the mount
system call, assigning to kernel
As a test, could you boot the system with "ide=nodma" and see if it still hangs?
I added "ide=nodma" to the kernel parameters, and rebooted.
After the reboot, I logged in as root and did "mount /mnt/cdrom1", and the
filesystem was successfully mounted, so it looks very much like this worked.
Is this known to only affect certain drives or configurations, so that this
condition might be automatically detected and worked-around (by kudzu)?
The 2.4 kernel will automatically enable (U)DMA on devices that claim to support
it. Several CDROM drives seem to advertise to support this but don't actually
do so. The kernel has a list of "known bad" devices and, surprise surprise,
the RICOH MP7083A was added to that list only a few weeks ago due to someone
else also seeing this. This means that the next kernel Red Hat will release will
just work without this parameter.
Also, "ide=nodma" will also disable DMA for your harddisks, this will cost
significant performance. You can achieve the same effect by adding a
"hdparm -d0 /dev/hdd" in one of the startupscripts
which is specific to your cdrw device, without also affecting the harddisks.
I'll close this bug as "fixed in rawhide" as the kernel in the rawhide (eg
experimental) version has this device already blacklisted. If you object to
that, or that kernel somehow doesn't fix it, please reopen this bug.