selection timeout

[Matthew Jacob]

* From the T10 Reflector (t10 at t10.org), posted by:
* Matthew Jacob <mjacob at feral.com>
*


On Tue, 7 May 2002 jmasker at attotech.com wrote:

> * From the T10 Reflector (t10 at t10.org), posted by: * jmasker at attotech.com
> *
> 
> *** Rant mode on ***
>    THANK YOU for asking this question.  This has been a hot button with me
> for quite some time.
>    To me it is always PAINFUL to watch a system boot with multiple SCSI
> adapters.  Some adapter setup utilities don't even have the ability to
> change this parameter.  Whenever I can, I always back this down to 64 ms
> and have never seen a problem.
>    So to scan an empty bus takes almost 4 seconds.  Add to this the stupid
> 3 SECOND bus reset delay and we're up to 7 ... PER CHANNEL.  Now look at
> the fact that the BIOS driver comes up first and must scan all channels,
> typically one at a time, then the OS comes up and scans the channels one
> at a time, and we're at almost 30 seconds for a motherboard with a
> dual-channel SCSI built in.
>    Granted, in the typical every day application environment, you aren't
> sitting there booting all the time.  But in a test environment, I wonder
> just how many man-hours are lost waiting. *** Rant mode off ***

This is an absurd rant.

a) Fix your OS to not reset the bus

b) Use an adult system- not a toy with a broken BIOS.

c) 3.75 seconds is not unreasonable. If all busses done in parallel, what's
the problem?>

d) In either specialized production systems or test environments, shorten the
timeout as needed. The 250ms delay is a 'reasonable' amount of time for an
unknown target on an open back of system connection.

*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org