Re: [2.2.17] oops in /proc/scsi/scsi

[Kurt Garloff <garloff@suse.de>]

[-- Attachment #1: Type: text/plain, Size: 1699 bytes --]

On Mon, Nov 27, 2000 at 08:57:03AM +0100, Matthias Andree wrote:
> Good thing first: I can't trigger the oopses with 2.0e6. I tried hard, didn't
> manage to.

Nice to learn.

> However, trying to trigger a race condition, I found that your script suffers
> from too small a kernel API in this place, since it uses the
> remove-single-device and all-single-devices in a non-locked manner, so if you
> run several instances of your script in parallel (simple rescan-scsi-bus.sh -r
> & will do) it will give a mess and may leave you with missing devices.
> 
> I don't currently recall if your web site states that your script may only be
> run once at the same time; lock files would be nice :-]

Well, it's one of these root-only admin tools.
I assume, root knows about what (s)he's doing.
So, I'll add some line to the documentation that tells you not to run the
scripts in parallel. Locking is really overkill, IMHO.

> A subsequent rescan-scsi-bus.sh will find the PX-32TS and add it, this time,
> without bus resets and aborts.
> 
> Should the reset->inquiry delay be applied after ANY reset? Is it actually
> applied but too short for my Plextor?

Well, the default for tmscsim is 1.5s time after a reset before it start
command processing again. This can be changed by echoing DelayReset values tp
the proc file. If your Plextor requires 5s, just do 
echo "delayreset 5" >/proc/scsi/tmscsim/? and the driver will wait for 5.5s
...

Regards,
-- 
Kurt Garloff  <garloff@suse.de>                          Eindhoven, NL
GPG key: See mail header, key servers         Linux kernel development
SuSE GmbH, Nuernberg, FRG                               SCSI, Security

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]