On Sat, May 13, 2000 at 11:43:38AM -0400, Eric Youngdale wrote:
>
> I don't have any good ideas here - are you saying that with the logging
> enabled the timeouts don't occur?
With logging enabled, the result code in scan_scsis_done is 0x25040000,
and no error recovery is performed. Total time to scan one device ID
is about 0.5secs.
With no (or less) logging, the result code is 26030000, and scsi_error.c
goes into error recovery mode. Total time to scan is about 5 secs.
> The implication to me is that the
> TEST_UNIT_READY command is taking a bit longer to complete than the timeout
> value for the command. With the logging, it finishes before the timeout
> hits.
>
> You could play with the timeout for TEST_UNIT_READY and see whether this
> changes matters.
But the timeout for TEST_UNIT_READY is SCSI_TIMEOUT + 4 * HZ, which is
at least 4 seconds. I don't think adjusting this will make any difference.
If the timeout was marginal, we'd expect to see the scan take much longer
when logging is enabled.
I guess I should hack around a bit more and see if I can get a dump
of exactly what the ISR sees under both these conditions. It's tough
when you're fighting with a Heisenbug like this, though :-)
Later,
Kenn
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to [EMAIL PROTECTED]
