Subject: Re: anyone seen this with -current (of recent vintage)?
To: Jason Thorpe <thorpej@nas.nasa.gov>
From: Matthew Jacob <mjacob@feral.com>
List: port-alpha
Date: 07/01/1998 16:20:50
>  It's very rare, and I can't really reproduce
> it on demand, either... sorry, I've just been a space cadet :-)
> 
 

's okay, but data integrity issues are pretty important
(Jason- umm, all the MegaDrives at NASA/Ames are going
to be on ISP 1040s.... umm......)


Well- I'm *somewhat* pleased that it isn't FibreChannel related-
at least it isn't the clusterf**k of FC F/W revs and ISP 21XX F/W
revs...

There are a couple of areas I'm currently looking at and am
aware of in the ISP driver (some I filed PRs on already):

1) Error Recovery (is pathetic). If commands timeout and continue
to timeout or you get jammed up trying to do a mailbox command,
things just go to hell.

2) I'm not reading NVRAM settings like I should be. On PCI cards
that are driven by the BIOS, you can get important items like
SYNC mode and whether you need to do termination or not.

3) I'm not really quite happy with what I've done with tagging.
I suspect (in the case of the mangled filesystem) perhaps some
suboptimal resultant interactions with various OS platforms (yes,
I mean "various"- this driver as I currently have it at my site
runs on FreeBSD, FreeBSD-CAM, Linux and NetBSD), and all have their
various breakages. Why I like working on devices, I do not know-
it's a constant lesson in humility- either that or I'm just not
that smart... sigh....

4) Not as efficient as it should be....

Well, now that I know that people other than me really see
significat problems, I'd guess I had better budget more time
to this....

Keep those cards && letters coming...

-matt