>I'm not even sure that it's been determined that the crashes are during
>write ops.  It could be that the same thing would happen in a large scale
>read.  Has anyone tried to untar a large archive from an EZ135 to a fixed
>drive of known caliber?

Just raw experimentation shows that it's fairly stable on reads. Writes are
OK if they're bursty. Long writes just bite it and sometimes too many bursty
writes make the drive nose-dive too.

>Would be nice.  Anybody with an EZ drive and an interrupt button manage to
>figure out where it's crashing?  Oh that's right.  It's not crashing, it's
>"freezing".  If I remember right, turning the drive off and back on wakes
>the thing up?  Weird.

It's not a question of where it's crashing or freezing anymore. That doesn't
happen in the latest kernels. In GENERIC5 it used to freeze but under 6
the kernel reports "target requested too much data" and just keeps chugging.
Of course, in the process the data gets botched.

My idea was that when these errors are encountered (i.e. rather than printing
the message) that the driver do a bus resync and re-send the faulty block.
Surely this can't be that hard? This may just be a band-aid but I'd like
to test it to see if it works. Are the people who are working on the SCSI
code hearing this? :)

>About the SCSI driver, here's a possible approach.  Apple has a SCSI
>Manager and driver structure already built into the ROMs.  Now I
>understand that nobody wants to call ROM routines, but what if somebody
>examined the ROM routines themselves and described how they worked so that
>somebody could rewrite the code.  You know, the clean room thing.

I think there's enough code in the kernel already that does a good job.
It just needs a little more work. That's all.

On a somewhat related note, I have also been working with the CD driver which
sits atop the kernel scsi driver. This is also tempremental. Working on
occasion and then failing. These problems are almost certainly interrelated.

_______________________________________________________________________
 Michael Zucca - mrz5149@rit.cs.rit.edu - http://www.rit.edu/~mrz5149/
 "I will choose a path that's clear. I will choose Freewill. "
  --Rush, Freewill
_______________________________________________________________________