[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

DS20E, SMP



Hi!

I recently got a new dual processor Compaq DS20E, and am trying
to get Debian running.  I've gotten virtually everything working,
but I'm having trouble with SMP.  I tried compiling 2.4.2 and 2.4.3pre3
(GENERIC processor) and both of these kernels panic on boot.  Next,
I compiled 2.2.19pre13 (from a Debian source package) with SMP support.
It seems to compile in, and it sees both processors:

    $ uname -a
    Linux tonka 2.2.19pre13 #1 SMP Fri Mar 9 07:02:02 AKST 2001     \
    alpha unknown

    $ cat /proc/cpuinfo
    cpu			: Alpha
    cpu model		: EV67
    cpu variation		: 7
    cpu revision		: 0
    cpu serial number	: 
    system type		: Tsunami
    system variation	: Catamaran
    system revision		: 0
    cycle frequency [Hz]	: 666666666 
    timer frequency [Hz]	: 1024.00
    page size [bytes]	: 8192
    phys. address bits	: 44
    max. addr. space #	: 255
    BogoMIPS		: 1325.40
    kernel unaligned acc	: 0 (pc=0,va=0)
    user unaligned acc	: 0 (pc=0,va=0)
    platform string		: COMPAQ AlphaServer DS20E 666 MHz
    cpus detected		: 2
    CPUs probed 2 active 2 map 0x3 IPIs 985

But the data in /proc/stat seems pretty bizzare (CPU1 is doing something,
but it's not user, system or nice CPU time?):

    $ cat /proc/stat
    cpu  15209 0 3468 288257
    cpu0 15209 0 3468 134790
    cpu1 0 0 0 153467

When I compile a kernel with 'MAKE=make' it takes a wall clock time
of 5m3.822s and user time of 2m16.566s.  With 'MAKE=make -j3' it's
wall clock time is 2m58.493s and user time is 2m17.061s.

So, the kernel seems to be using both processors (a kernel compile
can take almost half as long), but the information the kernel provides
seems to be wrong.  'top' also shows strange behavior, probably because
it's trying to get information from /proc.  I've got the latest version
of procps, which is supposed to support multiple CPU's:

    $ dpkg --list procps
    ii  procps      2.0.7-3    The /proc file system utilities

Is this what I should be seeing?

Any idea on what's wrong with the 2.4 kernels, or if a solution to
the panic is imminant?

Thanks!

Chris
-- 
Christopher S. Swingley         930 Koyukuk Drive
System / Network Manager        University of Alaska Fairbanks
IARC -- Frontier Program        Fairbanks, AK 99775

phone: 907-474-2689             fax: 907-474-2643
email: cswingle@iarc.uaf.edu    GNUPG and PGP2 keys at my web site
  web: http://www.frontier.iarc.uaf.edu/~cswingle



Reply to: