At 10:07 AM -0500 1/7/2001, Allen Briggs wrote:
>  > esp0: unexpected disconnect; sending REQUEST SENSE <----
>>  probe(esp0:0:0):  Check Condition on CDB: 0x00 00 00 00 00 00 <----
>>       SENSE KEY:  Hardware Error
>>        ASC/ASCQ:  Message Error
>
>Hmmm...  That looks like it might be rejecting the tag message.  Can you
>try recompiling a kernel with ncr53c9x_notag = 1 ?  I actually would
>have expected more failures, though.  I'm kind of surprised to see the
>errors on your FIREBALL1080S as that's one of the drives that I've been
>testing with.  :-P

OK, I'll do this.

Just to try it out, I hooked up a 6100 with ZTerm over serial. Even 
in a kernel that doesn't give me problems (1.5.1_ALPHA) I get 
bombarded with debugging info when I do "write/l ncr53c9x_debug 
0xffff"... Is there someway to turn some of this off?

I turned off the debugging, ran bonnie, and when I saw the error I 
started it again. The bombardment restarts. I had over 100K logged 
when I last checked!

If I compile the kernel with the "ncr53c9x_notag = 1", will some of 
these non-error debugging messages go away? It is very difficult to 
start bonnie with them going. Maybe if I made syslogd more picky 
about what it sent to console?

>If you can make this fail reliably, can you break into the debugger,
>enter the command "write/l ncr53c9x_debug 0xffff", enter "c" to continue,
>make it fail, then enter the command "write/l ncr53c9x_debug 0" (and
>"c" to continue again) and send me the data around the failure?

How much do you want? It would be a lot easier if I could get rid of 
the debugging messages that come up before I start bonnie.

>  It might also be useful to see the output with a "write/l mac68k_esp_debug",
>but that won't be available unless you've compiled with "options DEBUG".

OK, I'll compile with "ncr53c9x_notag = 1" and "options DEBUG". After 
that, you want me to do this:

Reboot into new kernel, break into debugger.
write/l ncr53c9x_debug 0xffff
write/l mac68k_esp_debug 0xffff
c
run bonnie -d /tmp -s 60 ... same as before
After it fails (how long?) break into debugger.
write/l ncr53c9x_debug 0
write/l mac68k_esp_debug 0
c
I guess I'll have to reboot from here as the machine is unusable 
after it fails.

Sorry, I've never used [gd]db for debugging before, and only compiled 
my first kernel a few days ago...

Thanks for you help and patience,
Daniel