the subject says it. I have an "all SCSI" system, controlled by an Adaptec
2940U on Linux 2.1.128SMP with two Pentium 100 prozessors.
This morning there was a read error on the CD-ROM (id3, "6plex" Plextor) .
Not uncommon, the errors were like this:
scsi0: MEDIUM ERROR on channel 0, id 3, lun 0,
CDB: Read (10) 00 00 02 4a 0e 00 00 04 00
Current error sr0b:00: sense key Medium Error
Additional sense indicates No seek complete
CD-ROM I/O error: dev 0b:00, sector 600120
One error was so wired, that the CDROM timeouted:
01:15:46 [..] scsi : aborting command due to timeout : pid 40090,
scsi0, channel 0, id 3, lun 0 Read (10) 00 00 02 4a 0e 00 00 02 00
Because of the timeout the SCSI bus hung and got reset by the kernel.
01:15:49 [..] SCSI bus is being reset for host 0 channel 0.
But the hard disc (id 0, Micropolis 4343NS 4,3 GB drive) got also reset
during reading and writing. Of course there is no chance for the drive to
complete a command if the command is "killed" by a bus reset:
01:16:01 [..] scsi : aborting command due to timeout : pid 4059
0, scsi0, channel 0, id 0, lun 0 Read (10) 00 00 59 06 80 00 00 08 00
01:16:01 [..] scsi : aborting command due to timeout : pid 4059
2, scsi0, channel 0, id 0, lun 0 Write (10) 00 00 42 18 3a 00 00 02 00
After this 2 processes were stuck in "wait_on_page" and process status "D".
These processes - a mpg123 accessing the hard disc and a bash running in a
kvt (KDE rxvt replacement) with a child process accessing the disc. This
child process was up and running and completed okay (and was a zombie after
this) only the bash and the mpg123 were "dead". Even the process accessing
the CDROM terminated with a read error. After rebooting some files were
killed/removed by the fsck (*arg* at least my ~/.newsrc) and the kernel
syslog contains about 500 byte ASCII-NUL characters.
Regards
Carsten
-- -rw-r--r-- 1 carsten muser 198 Nov 20 21:24 .signature Carsten Gross carsten@sol.wohnheim.uni-ulm.de Wohnheim Heilmeyersteige Sebastian Kneipp Weg 6, 89075 Ulm- 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/