Subject: Re: O2 issues.
To: josh lynch <strike@bitstream.net>
From: Rafal Boni <rafal.boni@eDial.com>
List: port-sgimips
Date: 11/26/2002 19:54:35
--==_Exmh_-1687705038P
Content-Type: text/plain; charset=us-ascii

In message <p05200d05ba09babea0ad@[192.168.0.207]>, you write: 

-> I'm running into a few weird issues with my 180Mhz O2 w/o secondary
-> cache. 

Bonus: you may actually have some luck with a SC-less O2.  We have no
code to drive the SC correctly, so all SC models blow chunks early on.

->       First of all when the machine boots cold, the kernel comes all
-> the way up perfectly. On any warm reboot, it panic's in 
-> macepci_conf_read(), apparently the PERR_MEMORY_ADDR bit is set in
-> the PCI_ERROR_FLAGS register. Which leads me to believe somewhere in
-> PCI land there is something not getting (re)initialized properly?
-> I have an fxp card installed into the slot. Has anyone else seen this?

There is much broken in the current O2 code; Chris Sekiya re-did much
of the PCI code and I've hacked on that to the point where I actually
mounted root via a PCI-X Intel GigE card the other day (I've been using
a RealTek for a while; a 3c905 was slightly harder to get working, and
I don't recall if the fxp just worked or not).  Unfortunately, none of
that code is in the tree yet.  Maybe I'll try and do a mini-sync over
the Thanksgiving holiday (ie, this weekend for the non-US folks out
there 8-).

-> Also, when it does boot successfully (/ on NFS), after an unknown
-> period of sitting idle, the machine becomes almost totally unresponsive,
-> but it's not completely locked up (I wrote a routine to cycle though
-> colors on the front panel led so I could see if interrupts were
-> still being delivered). Has anyone run into this?

You're probably not getting KB/serial/... interrupts, as the interrupt
code currently in-tree (some have said "what interrupt code" 8-) is
lacking to say the least.  I don't yet have better code, but I do have
some icky hacks to get serial mostly working with the current code.

The O2 code as is is in the "rolling up your sleeves and getting down
and dirty" stage, so getting anything to work semi-reliably takes work...

--rafal

----
Rafal Boni                                              rafal.boni@eDial.com
 PGP key C7D3024C, print EA49 160D F5E4 C46A 9E91  524E 11E0 7133 C7D3 024C


--==_Exmh_-1687705038P
Content-Type: application/pgp-signature

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.6 (GNU/Linux)
Comment: Exmh version 2.4 06/23/2000

iD8DBQE95BfLEeBxM8fTAkwRAgL/AJ4rpfIhTxZY6GJawuUpCqKDarlVdACfcvh6
/jItdIR8O94ypgjivba/TRY=
=tMVd
-----END PGP SIGNATURE-----

--==_Exmh_-1687705038P--