Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 12 Feb 2002 01:02:35 +0100 (CET)
From:      =?ISO-8859-1?Q?G=E9rard_Roudier?= <groudier@free.fr>
To:        Joerg Wunsch <joerg_wunsch@uriah.heep.sax.de>
Cc:        <freebsd-scsi@FreeBSD.ORG>
Subject:   Re: SCSI parity error detected
Message-ID:  <20020212004513.X2315-100000@gerard>
In-Reply-To: <200202121933.g1CJXCF87301@uriah.heep.sax.de>

next in thread | previous in thread | raw e-mail | index | archive | help


On Tue, 12 Feb 2002, Joerg Wunsch wrote:

> G=E9rard Roudier <groudier@free.fr> wrote:
>
> >> This Dawicontrol is 8-bit only btw., despite of using an 53c875.
>
> > But the driver cannot guess it. And it doesn't understand the
> > Dawicontrol settings from NVRAM (if some exists).
>
> But don't we have a problem here then?

May-be. If you can get the NVRAM layout and some data sheet about a
possible NVRAM type, let me know. ;)

Btw, if Dawicontrol uses PCI sub-vendor ids or some GPIO specific wiring
to identify their controller, I can let the driver probe such kind of
stuff. But unfortunately, I have exactly no info about those controllers.

> The controller uses a chip that could do 16 bit, but has only an 8-bit
> bus attached.  Since there's no NVRAM on it, there's no chance for the
> driver to tell this.  As soon as another 16-bit capable device
> appears, this device would also try to negotiate 16-bit when asked for
> (since it doesn't know about the 8-bitness of its bus adapter), and
> there's no chance for user intervention to tell the controller to
> restrict everything to 8 bit.

Btw, under Linux, this can be achieved using boot command parameters.
Under FreeBSD, such facility didn't exist at the time I worked on that.
Note that this can be achieved from the kernel config file using
appropriate SYM_* defines and building a custom driver.
(user can look into sym_conf.h file. There are tons of options)

> > OTOH, if there is no such message in the log, it could well be the
> > initial INQUIRY response that got a SCSI PARITY error. The problem
> > might then be the/a SCSI PARITY signal not being properly
> > driven/connected.
>
> What happens if the upper 8 bit are just terminated only on the
> controller?  Would this generate a parity error?

There is no risk as long as no device accepts a Wide Negotiation. If one
does, boom!!

> Sure, that controller looks fairly crappy by design to me (Frank
> meanwhile replaced it by a better one), but since we claim we
> support the entire Symbios Logic line, we IMHO should offer a
> method to fallback to safe defaults.  (For -current, could perhaps
> be done in the device hints, but that'd cause a chicken-and-egg
> problem when installing.)

The crappyness is to use undocumented stuff, in my opinion.
Will look into device hints again. Last time I did, this didn't fit.

The safe default it to use hardware correctly.

A SCA adapter that shrinks 16 bits BUS to 8 bit BUS, device being
connected on the 16 bit side and a Wide controller castrated to 8 bit
connected to the 8 bit is an horror that fails as expected.

  G=E9rard.


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-scsi" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020212004513.X2315-100000>