[Gllug] the joy of SCSI

[Christian Smith]

On Mon, 27 Feb 2006, Nix wrote:

>/dev/sda:
> Timing buffered disk reads:   70 MB in  3.07 seconds =  22.80 MB/sec
>
>/dev/sdb:
> Timing buffered disk reads:   70 MB in  3.00 seconds =  23.33 MB/sec
>
>Then, quietly the syslog murmurs
>
>Feb 25 16:17:56 loki warning: kernel: sym0: SCSI parity error detected: SCR1=132 DBC=50000000 SBCL=0
>Feb 25 16:17:56 loki warning: kernel: sym0:2: ERROR (81:0) (8-0-0) (10/9d/0) @ (mem c2800038:ffffffff).


>From the source (drivers/scsi/sym53c8xx_2/sym_hipd.c):
/*
 *  log message for real hard errors
 *
 *  sym0 targ 0?: ERROR (ds:si) (so-si-sd) (sx/s3/s4) @ name (dsp:dbc).
 *            reg: r0 r1 r2 r3 r4 r5 r6 ..... rf.
 *
 *  exception register:
 *      ds:     dstat
 *      si:     sist
 *
 *  SCSI bus lines:
 *      so:     control lines as driven by chip.
 *      si:     control lines as seen by chip.
 *      sd:     scsi data lines as seen by chip.
 *
 *  wide/fastmode:
 *      sx:     sxfer  (see the manual)
 *      s3:     scntl3 (see the manual)
 *      s4:     scntl4 (see the manual)
 *
 *  current script command:
 *      dsp:    script address (relative to start of script).
 *      dbc:    first word of script command.
 *
 *  First 24 register of the chip:
 *      r0..rf
 */

So, (8-0-0) are some representation of SCSI control lines. Check your
cabling.



>Feb 25 16:17:56 loki warning: kernel: sym0: regdump: da 00 00 9d 47 10 02 0e 00 08 82 00 80 00 0f 0a 30 a6 3d 01 02 ff ff ff.
>Feb 25 16:17:56 loki warning: kernel: sym0: SCSI BUS reset detected.
>Feb 25 16:17:56 loki notice: kernel: sym0: SCSI BUS has been reset.
>
>/dev/sda:
> Timing buffered disk reads:   6.2 MB in  3.03 seconds =  2.04 MB/sec
>
>/dev/sdb:
> Timing buffered disk reads:   14.5 MB in  3.00 seconds =  4.84 MB/sec
>
>Gee, thanks, SCSI layer. One stupid parity error ON A HOST I DON'T EVEN
>HAVE ATTACHED and you divide the disk speed by ten.
>
>(The adapter is on ID 11; the disks on 0 and 1. There's nothing on
>8:0:0.)


Adaptor on 11? Why?


>
>I guess it's a transient cable problem (odd, the cable is a SCSI III
>160Mb/s cable and I'm chucking 40Mb/s over it at most). But this isn't
>a very nice way to respond to a transient cable problem, really.
>I mean I didn't even know you could *make* SCSI buses run that slowly.


You've used an old SCSI DAT drive, yes?

It's quite a reasonable way of responding to cable problems, if it "fixes"
the problems.

Christian


-- 
    /"\
    \ /    ASCII RIBBON CAMPAIGN - AGAINST HTML MAIL
     X                           - AGAINST MS ATTACHMENTS
    / \
-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug