Re: Tagged cmnd queueing trouble in aic7xxx 5.0.19

Doug Ledford (dledford@dialnet.net)
Tue, 16 Jun 1998 16:47:40 -0500

Leonard N. Zubkoff wrote:
>
> From: Matthias Andree <mandree@dosis.uni-dortmund.de>
> Date: Sun, 14 Jun 1998 13:23:14 +0200 (CEST)
>
> Hello,
>
> I am using AIC-7XXX v5.0.19 (Adaptec 2842VL) with both Linux 2.0.34 and
> 2.1.105.
>
> Drive information is below. To put it short:
>
> sda - Micropolis 4345WS WIDE ULTRA QUEUEING
> sdb - IBM DCAS 32160 ULTRA QUEUEING LINKING
>
> Activating tagged command queueing for the Micropolis drive -
> aic7xxx=tag_info:{{4}} - gives LOADS of timeout (I cannot currently paste
> in, sorry), and a lot of set/reset loops, the system is not usable. On the
> other hand, activating tagged command queueing for the IBM -
> aic7xxx=tag_info:{{,0}} works just fine, even under heavy bus load.
>
> My question: is the Micropolis tagged queueing failure in any way related to
> * firmware bugs in the Micropolis drive?
>
> It's a Micropolis. Firmware bugs in the tagged queuing implementation was one
> of their specialities. They were also especially competent at firmware bugs
> where the drive became wedged so badly that only a power cycle would recover.

Aka, Micrapolis drives should be their name.

> * it being a wide drive on a narrow adaptor (I remember 5.0.18 negotiating
> wide transfers in spite of that)?
>
> Unlikely.

5.0.18 did have a bug that would allow wide devices to negotiate wide even
if they are on a narrow controller. 5.0.19 fixed that.

> * it not supporting linked commands?
>
> No. Nothing must really uses linked commands any more.

The aic7xxx driver in particular never uses linked commands, only regular
tags and ordered tags. But, we also have code in place to detect problems
with ordered tag commands and disable them if the drive hoses them up, so
that shouldn't be the problem either.

-- 


 Doug Ledford  <dledford@dialnet.net>
  Opinions expressed are my own, but
     they should be everybody's.


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu