Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 31 Aug 1998 19:47:51 -0400 (EDT)
From:      Simon Shapiro <shimon@simon-shapiro.org>
To:        Andrew Gallatin <gallatin@cs.duke.edu>
Cc:        freebsd-alpha@FreeBSD.ORG, Doug Rabson <dfr@nlsystems.com>
Subject:   Re: Alpha Install - oops!
Message-ID:  <XFMail.980831194751.shimon@simon-shapiro.org>
In-Reply-To: <13802.61200.986416.952138@grasshopper.cs.duke.edu>

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

Andrew Gallatin, On 31-Aug-98 you wrote:

>  I think you may have more luck if you power-cycle the box.  When I've
>  seen this under NetBSD/alpha, I suspected it was caused by the IDE
>  controller acting as a noise generator, and was able to cure it by
>  disabling the ide controller.

How?

(I think there is a problem we may want to solve here, still)

...

>   > your mileage may vary.
>  
>  This happens if you attempt to put an 'unverified' card onto pci bus 0
>  (and thus expose it to the pyxis DMA bug present in non-GL miatas).
>  See my Miata brain-dump below for an explanation, bearing in mind that
>  a DPW {433,500,600}{a,au} == "Miata" or "MiataGL"
>  
>  A MiataGL can be distingiushed from a Miata by the presense of a
>  Cypress SIO (aka "PCI Peripheral Controller") chip in place of the
>  Intel SIO chip at Bus 00 Slot 07 (from SRM '>>>show config').  A
>  non-GL Miata has some nasty PCI bugs, including problems doing DMA
>  reads across page boundaries.  All Miatas are equipped with a 21x52
>  pci-pci bridge which breaks up DMA's at page boundaries.  The 3 32-bit
>  PCI slots sit behind the ppb.  The 2 64-bit slots sit directly on the
>  primary PCI bus. To avoid problems with cards which are not safe for
>  Digital UNIX (eg, they're not certain they don't do DMA reads across
>  page boundaries), DEC put a hack into the SRM console which prevents
>  you from placing unverified cards directly on the primary PCI bus.
>  
>  If you know a card & its driver are "safe", or if you're on a MiataGL,
>  you can get around this by setting the undocumented SRM console
>  variable 'pci_device_override' to the PCI device id followed by the
>  pci vendor id.  Eg, for a card with the VendorID 0x10e8 & Device ID
>  0x8043, one may use:
>       >>> set pci_device_override 804310e8
>  
>  In later versions of the firmware, I hear that you can just set it to
>  -1.
>  
>  On a Miata-GL that I just uncrated this afternoon (with version
>  V6.8-44 of the firmware), pci_device_override no longer needs to be
>  set. 

Wow!

Simon


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



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?XFMail.980831194751.shimon>