[PARPORT] HP Colorado 5G and filemarks

soboroff (soboroff@xrce.xerox.com)
Thu, 22 Oct 1998 16:24:40 +0200

Hi... I'm new to the list, and haven't been able to find a list archive, so
forgive me if this is a FAQ.

I have a HP Colorado 5Gig tape drive connected to my Linux box (Gateway 2000
Solo 9100, Linux 2.1.125 (pt.c 1.04)). I load the drivers with modprobe,
first parport_pc, then epat, then pt, with no options. The pertinent dmesg
output is:

parport0: PC-style at 0x378 [SPP]
paride: version 1.03 installed
paride: epat registered as protocol 0
pt: pt version 1.04, major 96
pt0: Sharing parport0 at 0x378
pt0: epat 1.01, Shuttle EPAT chip c6 at 0x378, mode 5 (EPP-32), delay 1
pt0: HP COLORADO 5GBe, master, no media

I am using the amanda system for backups (ok, a bit much for just my
machine, but (a) I wanted the experience setting it up and (b) it's one of
the few backup options that don't look like alpha software, and come with
decent documentation). Taper dies with segmentation faults unceremoniously
at various places in a dump.

Amanda is a little more verbose... it labels the tape (amlabel) and checks
out (amcheck) just fine, but during the dump it seems to die while trying to
write its first filemark, and there is a dmesg output at that point:

pt0: HP COLORADO 5GBe, master, blocksize 512, 2458 MB
pt0: Unimplemented mt_op 5

This seems to agree... in /usr/include/linux/mtio.h, this op seems to be
MT_WFM, for writing a file mark. The pt error message is from an
unsupported mt ioctl call; the only supported ioctl is for rewind.

Why doesn't the pt driver support more of the mt ioctl's? Would it be
unreasonable to patch pt.c to handle MT_WFM with a call to pt_write_fm()? I
can't figure out exactly how else filemarks get written, except through the
ATAPI tape calls (drivers/block/ide-tape.c). So the backup software has to
know it's using an ATAPI tape drive, and can't use the more general mt

Anyway, tia,

-- To unsubscribe, send mail to: linux-parport-request@torque.net --
-- with the single word "unsubscribe" in the body of the message. --

This archive was generated by hypermail 2.0b3 on Wed 30 Dec 1998 - 10:18:39 EST