Hi,
I have an AlphaPC164 with a 4 GB IDE disk (NetBSD-1.3.3 on it),
a Symbios 8750SP ultra narrow (i.e. 8bit) hostadapter, and an IBM DDRS U2W
disk attached via a narrow/wide adapter.

NetBSD (both 1.3.3 and the 1.4 installation kernel) detects:

  ncr0 at pci0 dev 7 function 0: ncr 53c875 fast20 wide scsi
  ncr0: interrupting at eb164 irq 1
  ncr0: minsync=25, maxsync=254, maxoffs=16, 128 dwords burst, large dma fifo
  ncr0: single-ended, open drain IRQ driver, using on-chip SRAM
  ncr0: restart (scsi reset).
  scsibus0 at ncr0: 16 targets

Then it scans the bus, finds the DDRS, decides to switch it to wide
and query its parameters.  That doesn't work with a narrow cable
though. :)

It starts writing endless error messages on the console (see below).

After adding "options SCSI_NCR_MAX_WIDE=0" to the kernel config file,
it detects and uses the DDRS without problems:

  sd0 at scsibus0 targ 0 lun 0: <IBM, DDRS-34560D, DC1B> SCSI2 0/direct fixed
  sd0: sd0(ncr0:0:0): 10.0 MB/s (100 ns, offset 15)
  4357MB, 8387 cyl, 5 head, 212 sec, 512 bytes/sect x 8925000 sectors

(Is this correct for ultra SCSI or are there more problems lurking?
I'm using a rather short cable and an active terminator, so the hardware
should be ok.)


Then I put the NetBSD-1.4/alpha/installation/instkernel/netbsd.gz and a
/boot on sd0a and did an installboot.  I can boot the kernel, but
it also switches the DDRS to wide mode.  After booting 1.3.3 (on wd0)
again, I found some of the console messages in the dmesg output
(I saved more of it, but it doesn't look much different):

  30000a8).
  ncr0: script cmd = 878b0000
  ncr0: regdump: da 00 00 03 47 00 00 0f 31 08 00 a6 80 00 00 02.
  ncr0: restart (fatal error).
  sd0(ncr0:0:0): COMMAND0: ERROR (90:0) (0-a6-0) (0/3) @ (script a0:430000a8).
  ncr0: script cmd = 878b0000
  ncr0: regdump: da 00 00 03 47 00 00 0f 31 08 00 a6 80 00 00 02.
  ncr0: restart (fatal error).
  sd0(ncr0:0:0): COMMAND FAILED (9 ff) @0xfffffe0000052400.
  ncr0: aborting job ...
  ncr0:0: ERROR (90:0) (0-a6-0) (0/3) @ (script a0:430000a8).
  ncr0: script cmd = 878b0000
  ncr0: regdump: da 00 00 03 47 00 00 0f 31 08 00 a6 80 00 00 02.
  ncr0: restart (fatal error).
  sd0(ncr0:0:0): COMMAND FAILED (9 ff) @0xfffffe0000052400.
  ncr0: aborting job ...
  [SNIP]

Is it a bug?  Is there anything I can do about it (short of upgrading
NetBSD on wd0)?  Can I build an 1.4 installation kernel under NetBSD 1.3.3
without disturbing the 1.3.3 installation?
If I can't, is somebody willing to build a kernel with SCSI_NCR_MAX_WIDE=0
for me?

joerch