Since 2.2.9[1], I have had problems with linux dying with the error
"Timeout on hda waiting for DMA". By dying I mean that nothing but the
reset button responds (I have an ATX mobo, and even the poer button
doesn't respond!). The problem has usually (but not always) happened
during the 4 am cron updates. Of the 8 or so times it has happened,
only twice did it happen at a different time. Once was a seemingly
random time during the day, and the other was while I was logged in (but
I don't remember what I was doing when it occurred). The DMA timeout
has happened on both hda and hde (my two hard drives)., although only
once did hde timeout. The only other systematic thing I have noticed
about this is that the crashes have, as of yet, always occurred after
the system has been up for at least a day and a half or so (i.e. the
first 4 am cron run never seems to kill it).
[1] This did not occur with 2.2.6, I don't think it happened with
2.2.7, although I am not entirely sure, and I never had 2.2.8
installed. I have not tried any of the 2.3.x nor any of the ac
kernels. It has certainly occured in both 2.2.9 and 2.2.10.
Nitty gritty of the hardware:
Motherboard is a SuperMicro P6DBE with two PII/300 chips. I use both
the on-board piix4 IDE controller, and an offboard Promise Ultra33
controller. While the crashes usually happen on hda, the only linux
files on hda are the ftp and http documents. The IDE devices are as
follows:
hda: Maxtor 8.4 GB UltraATA/DMA drive (purchased August 1998)
hdb: Panasonic 24x CD-ROM (same purchase time)
hdc: Seagate Travan tape drive (STT-8000A, I believe)
hdd: Sony CDU-928E CD-R
hde: MAxtor 8.4 GB UltraATA/DMA drive (purchased January 1999)
To facilitate cd recording and tape backup, I have all the IDE device
types except the ATAPI Floppy enabled, as well as all the SCSI device
types. I boot the kernel with "hdb=ide-scsi hdc=ide-scsi hdd=ide-scsi"
in addition to other standard options (read-only, etc.)
I use gcc 2.7.2.3 (that *is* the compiler that comes with rh 5.2,
right?) to compile the kernel.
I hope this info helps track down the problem,
-Matt
P.S.: Due to size restrictions on my e-inbox, it is impractical for me
to subscribe to the mailing list, so please cc me if you have more
questions, or other things you would like me to try.
-
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/