Re: esp vs. MSG_IGN_WIDE_RESIDUE

[Michael <macallan%netbsd.org@localhost>]

Hello,

On Sun, 11 Jun 2017 19:54:07 -0000 (UTC)
mlelstv%serpens.de@localhost (Michael van Elst) wrote:

> macallan%netbsd.org@localhost (Michael) writes:
> 
> >I keep getting messages like this with my new old toy, and Ultra 2:
> >sd(0,0,0): unrecognized MESSAGE; sending REJECT  
> 
> >The disk is hooked to an esp/hme SBus card, so wide SCSI. I've been
> >using those on sparc with all sorts of disks and never saw this.  
> 
> The message says that not all bytes of the last wide transfer cycle
> are valid. The message is necessary for a read command, because that
> is the only way to indicate the correct transfer length.

Ok, that makes sense.

> Disks rarely have odd-sized disk blocks, but an odd-sized transfer
> might occur for other data, but there the length is usually transmitted
> or implied, the correct transfer length rarely matters.

Mounting filesystems triggered these messages, didn't see them
afterwards.

> Tapes on the other hand may support odd-sized blocks. Ignoring the
> message there is a problem.

The previous behaviour was to reject & reset, which I would assume is
worse.

How would one correctly handle this - check if any read is followed by
one of these? Comments in other drivers indicate that the SCSI
controller might handle them on its own ( I think I've seen that in the
aha driver, possibly esiop )

have fun
Michael