unix sockets on Compaq PW 5000 SMP

Zow (zow@pensive.llnl.gov)
Tue, 29 Sep 1998 18:18:16 -0700


Good day,
I've searched high and low for a solution to this, and have become convinced
that it is a conflict with 2.1, or at least that this is the only place that
someone will know the solution to my problem. I have a Compaq Professional
Workstation 5000 with two (2) PPro 200Mhz 256k processors, 160 MB ECC EDO
DRAM, built in TLAN networking, NCR53C8XX SCSI, etc (that's should cover
anything relevant). When I run 2.0.35 in SMP mode, everything is fine. But, of
course, I want 2.1 for the better drivers, performance, features, etc. But
when I run a 2.1 SMP kernel (115, 119 and 122 all do this) and do a "reboot"
or ctrl-alt-del, I get:

Processor 1 Initialized
212 - Processor 0 Initialization Failed

Now, here's the cincher: if I run a monolithic kernel, it always happens; but
if I run a modular kernel and kmod can't load the unix.o module, or it (unix)
never gets loaded b/c I reboot the system before anything uses anything from
unix (like a pipe), then Processor 0 works just fine.

What this means and what can be done to help: Something isn't perfect in the
unix module. I won't say that something's broken because I doubt that it's
something as simple as a static getting trampled or any "normal"
multi-processor type problem. My research into the problem indicates that
Compaqs do funky stuff with SMP. The "SMP Linux on XXX Compaq" guides that
I've seen note that something has to be changed in the BIOS, like the OS to
Unixware or the APIC mode to full, but I can't find any settings like these in
the Setup utilities on the 5000. So does anyone have the magic formula to get
a 5000 to run 2.1 SMP? Failing that, I make no claims to kernel debugging
experience, but I'm a competent applications developer, so if someone could
tell me what to try to do to debug this from my end, I would be most
appreciative.

Thanks in advance for any leads!
"Zow" Terry Brugger
zow@acm.org
zow@llnl.gov

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/