until recently, I had a bunch of (narrow fast-SCSI) devices,
among them a scanner "Umax Astra 1200S" connected to a
NCR810-based SCSI-controller (Asus SC-200). Except for occasional
mysterious problems (which might have been caused by the cables)
everything worked fairly well.
Now I got a new motherboard (Asus P2B-S), which has an onboard
Adaptec (Ultra2) SCSI-controller. Since I thought that this would
cause the least problems, I moved all the other devices to the
Adaptec controller. The NCR810 now has the scanner as the only
device on it (termination on the controller & scanner is on).
Ever since then, only the very first access to the scanner after
booting the system will work. (Actually, more the first program,
that accesses the scanner: when I start "xscanimage" from the
SANE package (version 0.74) for the first time, I can do multiple
scans and the scanner works alright until I finish the program).
After the first program is finished, any attempt to access the
scanner will fail until rebooting the system. At this time the
message "tagged command queue depth set to 2" shows up in the
system logs. From there on, the NCR driver obviously is busy
resetting the SCSI bus:
Oct 4 16:29:02 swamp kernel: ncr53c810-0-<5,0>: tagged command queue depth set to 2
Oct 4 16:30:05 swamp kernel: ncr53c810-0: restart (scsi reset).
Oct 4 16:30:05 swamp kernel: ncr53c810-0-<5,*>: asynchronous.
Oct 4 16:31:06 swamp kernel: ncr53c810-0: restart (scsi reset).
Oct 4 16:31:06 swamp kernel: ncr53c810-0-<5,*>: asynchronous.
Oct 4 16:32:09 swamp kernel: ncr53c810-0: restart (scsi reset).
Oct 4 16:32:09 swamp kernel: ncr53c810-0-<5,*>: asynchronous.
Oct 4 16:33:11 swamp kernel: ncr53c810-0: restart (scsi reset).
Oct 4 16:33:11 swamp kernel: ncr53c810-0-<5,*>: asynchronous.
Oct 4 16:34:12 swamp kernel: ncr53c810-0: restart (scsi reset).
Oct 4 16:34:12 swamp kernel: ncr53c810-0-<5,*>: asynchronous.
Oct 4 16:34:56 swamp kernel: ncr53c810-0-<5,0>: COMMAND FAILED (8a ff) @c009a820.
Oct 4 16:36:34 swamp kernel: ncr53c810-0-<5,0>: M_REJECT received (75:8).
Oct 4 16:36:34 swamp kernel: ncr53c810-0-<5,0>: M_REJECT received (75:8).
Oct 4 16:36:49 swamp kernel: ncr53c810-0-<5,0>: tagged command queueing disabled
When I disable tagged command queueing, I don't get the "depth
set to 2" - message, but otherwise things are the same. The only
workaround I could find is to remove the scanner via
"remove-single-device" and immediately add it again after every
program run. The problem is obviously caused by the current
version of the ncr driver (Kernel 2.1.125) - when I boot an
ancient kernel (2.1.59), the scanner works alright.
Regards,
Peter
--
| Peter Daum (gator@cs.tu-berlin.de) __o (+49 30)
| http://www.cs.tu-berlin.de/~gator _'\<_ 691 46 35
| pgp messages welcome! _(_)/(_) ___________
-
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/