esp(4) SCSI (Re: MI SONIC Ethernet driver for mac68k)

haukeEspresso.Rhein-Neckar.DE wrote:

> At 9:35 Uhr +0900 2.6.2007, Izumi Tsutsui wrote:
> >BTW, does esp(4) SCSI work without problem on your
Q650?
> >
> >On my LC630, "dd if=/dev/rsd0a of=/dev/null
bs=64k" doesn't work
> >while "bs=63k" works:
> 
> No problems here:
> 
> fattie# dmesg | grep sd0
> sd0 at scsibus0 target 0 lun 0: <SEAGATE,
ST336706LW, 010A> disk fixed
> sd0: 35003 MB, 26302 cyl, 4 head, 681 sec, 512
bytes/sect x 71687370 sectors
> sd0: async, 8-bit transfers, tagged queueing
> boot device: sd0
> root on sd0a dumps on sd0b
> fattie# uname -a
> NetBSD fattie 4.0_BETA2 NetBSD 4.0_BETA2 (FATTIE) #0:
Fri Apr 20 18:38:05
> CEST 2007  haukefattie:/usr/src/sys/arch/mac68k/compile/FATTIE
mac68k
> fattie# dd if=/dev/rsd0a of=/dev/null bs=64k
> ^C706+0 records in
> 706+0 records out
> 46268416 bytes transferred in 18.627 secs (2483943
bytes/sec)
> fattie#

Hmm, then LC630 (Q630) might have some quirk around esp DMA
and should we handle it in mac68k/obio/esp.c?

> I noted, though, that the Seagate Cheetah is only
recognized as
> asynchronous scsi device.

If you don't have any "flags" at esp(4) lines in
your config,
you could add "PQUIRK_CAP_SYNC" quirk flag to
scsi_quirk_patterns[]
in sys/dev/scsipi/scsiconf.c.
---
Izumi Tsutsui

At 2:27 Uhr +0900 5.6.2007, Izumi Tsutsui wrote:
>> >On my LC630, "dd if=/dev/rsd0a
of=/dev/null bs=64k" doesn't work
>> >while "bs=63k" works:
>>
>> No problems here:

[...]

>Hmm, then LC630 (Q630) might have some quirk around esp
DMA
>and should we handle it in mac68k/obio/esp.c?

Sounds like it... maybe the issue hasn't shown up so far
because most
people stick with IDE disks for that Mac?

>> I noted, though, that the Seagate Cheetah is only
recognized as
>> asynchronous scsi device.
>
>If you don't have any "flags" at esp(4) lines
in your config,
>you could add "PQUIRK_CAP_SYNC" quirk flag to
scsi_quirk_patterns[]
>in sys/dev/scsipi/scsiconf.c.

Interesting. I had the Cheetah in a PowerMac G3/400 before
on an Adaptec
U2W, and it would fall back to async there, as soon as
another disk was on
the bus with it. In addition, the other disk (a 9 GB IBM U2W
that the Mac
shipped with) would then fall back to narrow (8 Bit) SCSI.
So there's
definitely something fishy there.

OTOH, the Cheetah ran on a Sun sbus fas adapter for years
without problems.
And a

Index: scsiconf.c
============================================================
=======
RCS file: /cvsroot/src/sys/dev/scsipi/scsiconf.c,v
retrieving revision 1.241.2.1
diff -u -p -u -r1.241.2.1 scsiconf.c
--- scsiconf.c  19 Feb 2007 20:03:17 -0000      1.241.2.1
+++ scsiconf.c  4 Jun 2007 22:28:40 -0000
 -644,6
+644,8  static const struct scsi_quirk_inquiry_p
         "FUJITSU ", "M2624S-512      ",
""},     PQUIRK_CAP_SYNC},
        {{T_DIRECT, T_FIXED,
         "SEAGATE ", "SX336704LC"   ,
""}, PQUIRK_CAP_SYNC |
PQUIRK_CAP_WIDE16},
+       {{T_DIRECT, T_FIXED,
+        "SEAGATE ", "ST336706LW"   ,
""}, PQUIRK_CAP_SYNC},

        {{T_DIRECT, T_REMOV,
         "IOMEGA", "ZIP 100",           
"J.03"}, PQUIRK_NOLUNS},

doesn't make a difference for the Quadra, with or without a
PQUIRK_CAP_WIDE16. The disk has a 16-to-8 bit adapter, maybe
that leads the
kernel to assume async operation?

	hauke

--
"It's never straight up and down"     (DEVO)