[1550] in SIPB_Linux_Development
[Gary A Weston: Re: [Patrick McCormick: Re: IRQ share w/ network and SCSI
daemon@ATHENA.MIT.EDU (Theodore Y. Ts'o)
Mon Dec 23 16:55:03 1996
Date: Mon, 23 Dec 1996 16:54:48 -0500
From: "Theodore Y. Ts'o" <tytso@MIT.EDU>
To: linux-dev@MIT.EDU
For what it's worth...
------- Forwarded Message
X-Sender: gweston@po7.mit.edu
X-Mailer: Windows Eudora Pro Version 2.2 (32)
Mime-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Date: Mon, 23 Dec 1996 09:09:50 -0500
To: "Theodore Y. Ts'o" <tytso@MIT.EDU>
From: Gary A Weston <gweston@MIT.EDU>
Subject: Re: [Patrick McCormick: Re: IRQ share w/ network and SCSI
cards]
Cc: gweston@MIT.EDU
Hi,
I haven't personally played a whole lot with the GXPro systems, but i've
played with quite a number of other system boards in my time. You may want
to recheck the system BIOS settings. On most PC BIOS's there is an option
for configuring the PCI slots. The slots are usually labelled A,B,C etc and
they can be configured a number of ways. Also some BIOS's are plug&play
(plug&pray). Each PCI 'slot' can most likely be configured to use a
specific IRQ, or set to AUTO. Set as AUTO the BIOS then can automatically
choose an IRQ for each slot by bouncing down a list of IRQ's and taking
whatever is available at the time. You may also try disabling the plug and
play bios if it is enabled. This should be in the same menu. I have
noticed that Linux sometimes may not like plug and play enabled (this may be
the problem with your linux situation). If you have a questions, or if the
above seemed unclear please contact me if you wish. If all this has been
tried, then I guess you can ignore me ;P Best of luck!
At 03:08 AM 12/21/96 -0500, you wrote:
>For your information,
>
> It sounds like the Dell Optiplex has a cheesy architecture where
>all PCI cards have to use the same IRQ channel. I understand that
>making Linux work on recommended hardware isn't a high priority (if it's
>a priority at all), and while it's true that operating systems can work
>around this sort of design flaw, it can't possibly be good for the
>overall performance of the machine..... This might be something to
>consider for future choices of machines.
>
> People at the MCC might want to take a note of this in case they
>have students coming to complain; SIPB does have a workaround.
>
> - Ted
######################################
Gary Weston Consultant
MIT Computer Connection
617/253-7686 gweston@mit.edu
######################################
------- End Forwarded Message
