yellowdog-general digest, Vol 1 #43 - 13 msgs

[Endymion]

On Wednesday, April 10, 2002, at 12:47 PM, yellowdog-general-
request@lists.terrasoftsolutions.com wrote:

> Does YDL "see" G3 upgrade processors?  I have a 6500 at home with a
> Sonnet Crescendo G3/300 upgrade.  The last time I tried Linux on it
> (LinuxPPC2000) it was unusable...Linux didn't see/use the G3, just
> the (now cacheless) built-in processor.

Linux does indeed see G3 upgrade processors, but the 6500, 6400, and=20
6360 are a special case. Software is absolutely needed to make these G3=20=

CPUs function--not just their cache, but to function at all. So the only=20=

way to get it working is to make sure the Sonnet Crescendo extension=20
loads before you start BootX.

Even so, this is no guarantee that Linux, via either YDL or any other=20
distribution will work. I and one other list member (Drew Lane) have=20
feverishly attempted to figure out the situations under which the Sonnet=20=

G3 is enabled and working, and we have had varying results for varying=20=

reasons. Those times that we have managed to make the G3 work, the L2=20
cache was not enabled unless PowerLogix Cache Profiler extension was=20
used. This involves altering the file type of the Sonnet Crescendo=20
extension to "scri," rather than "INIT," to make it load first, then the=20=

PowerLogix extension, and then finally BootX (either the BootX app or=20
the BootX extension), and that will enable the L2 cache.

But, all of that is useless if you are unable to get the G3 activated in=20=

the first place, and in the last YDL installation I have attempted, I=20
could not get either YDL 2.1 or 2.0 installed while the G3 was=20
activated. Additionally, once I made a G3'less installation, Linux would=20=

never start from the hard disk if I enabled the G3.

We suspect three factors are involved.

1. The Sonnet Crescendo extension itself. When Linux fails to start it=20=

either stops at a "booting..." message, or with several "hda: lost=20
interrupt" messages. It's as if it can't read the hda partitions, so=20
something is pretty darn whacked out when the Crescendo extension is on,=20=

remember, it is telling everything to look at the L2 slot for the CPU=20
now.
2. The scsi mesh driver may have changed. Some of the older kernels in=20=

older distributions are known to work with the Sonnet Crescendo L2 G3=20
just fine. The last one I know of to have no problems was YDL 1.2.=20
LinuxPPC 2000Q4 also started for me with my G3 activated, but Drew could=20=

not get it to work. (Like I said, we have had differing results in=20
trying to figure this out with almost the same identical hardware and=20
setups.) But at any rate, the hda: lost interrupt messages come=20
immediately after a long "pause" as the startup is trying to detect scsi=20=

devices. Although this may be a part of the problem though, it is=20
directly related to problem #1--because when we disable the Crescendo=20
extension, this problem goes away completely.
3. Certain PCI cards may cause problems. Some think that it's just=20
graphics cards. On the other hand I have had almost no problems from PCI=20=

graphics cards in my 6400 and 6500--but Drew couldn't get the installer=20=

CD or HDD install to start up until he removed his ATI Rage128 card. A=20=

Voodoo 3 also caused the startup to hang. And curiously, so did a USB=20
PCI card, later. So it looks like the mere presence of a PCI card can=20
cause some problems. Replacing that USB card with a completely different=20=

USB card solved his problem. If this is the case, I think that it is=20
much the same as problem #2 above, which of course is ultimately again=20=

the result of problem #1 above.

This is something that I am pretty sure is out of the domain of=20
Terrasoft and we are going to have to rely on some seriously savvy=20
kernel hackers to help us get this problem fixed. I have both a 6400,=20
and a 6500, and if I can get ahold of a spare power supply I might even=20=

have a second 6400 to play with. (It'll make testing configurations a=20
lot easier by having more than one unit to work with.) I had been=20
waiting for that to happen before seeking help on this matter--because=20=

that's about the most I can do. I'm not a kernel hacker so I can't do=20
much beyond trying fixes and things out.

But you asked, and that's what I know. I hope someone can lend a hand to=20=

this because these particular Macs are seriously hurting in  almost any=20=

OS (except for Be maybe) if you cannot have a G3 in them.

~Alek
--
Lime and limpid green the sound surrounds the icy waters under
Lime and limpid green the sound surrounds the icy waters
Underground
=A91968 Westminster Music Ltd., written by Syd Barrett