SUMMARY: cpu pairs - two becomes one?

From: Matt Fountain <routeunix_at_yahoo.co.uk>
Date: Fri Jul 27 2001 - 09:47:11 EDT
Apologies for late summary. Busy times. It turned out
after a visual inspection of the boards that the boxes
were telling the truth and that, unbeknown to me, the
other CPU's had indeed been removed and re-located to
pastures new. 
Thanks to:-
"V. Ray Krebs III" <rayk@tera.teralink.com> 
Mike Kiernan <mkiernan@onet.pl> 
Dirk Jansen <dirk@abcomp.be> 
for very helpful suggestions on doing further checks.

Original q's and shortened replies follow:-

Thanks again.

Matt

****
   I have a few dual processor E250's (Solaris
2.6)none of which claim to acknowledge the presence of
the second processor with the commands:-
Blake# psrinfo
0       on-line   since 06/11/01 10:21:51
Blake# psrinfo -v
Status of processor 0 as of: 07/03/01 16:27:43
  Processor has been on-line since 06/11/01 10:21:51.
        and has a sparc floating point processor.
Blake# psrinfo -s 1
psrinfo: processor 1: Invalid argument
Blake# mpstat
CPU minf mjf xcal  intr ithr  csw icsw migr smtx  srw
syscl  usr sys  wt idl
  0    1   0    0   311  111  158    0    0    0    0 
1013    0   1   1  98

Blake# /usr/platform/`uname -i`/sbin/prtdiag -v
etc 

Any ideas? I did find reference to a patch (104595) -
but not for this chip. Also "Summary: UE450 with
disappearing CPU" (5 Jul 99) refers to setting
upa-noprobe-list to 0 at the OK prompt, but this
variable does not exist for this platform.

Also how does one find out the revision number of the
cpu? A while back Sun reported that some cpu's had a
prob with cache memory and I thought it was possible
to check this against the cpu revision number.


*****
.  The cpu revision will be the 'Mask' column in your
prtdiag -v - other than that the speed MHz and L2
cache (Ecache) size should help determine the part
number.

Have a hack with adb and see if the kernel is seeing
the other cpu at all:
# adb -k  [doesn't give you a prompt, just type the
commands]
ncpus/X
$<cpus
prom_cpu/X
$q

If it's not - chances are the prom doesn't see it
either.  All I can think of is firmware - check your
OBP rev is at the latest.  Pull the cpus out,
swap em around etc - if they have DC converters like
the E450 make sure they're installed in the correct
slot also.

cheers,
Mike
OK-prompt. If you then type "banner", the system will
show the number of cpu's, the total memory installed,
the ethernet-address... If the cpu already doesn't
show up here it will definitely not show up under
solaris.(if the system can't see it, the OS can't see
it)
The cpu's should apear as "(2x 400Mhz UltrasparcII)"
for example.

Once solaris is running:
The two cpu's are not seen as one. Mpstat should
produce an output like:

CPU minf mjf xcal  intr ithr  csw icsw migr smtx  srw
syscl  usr sys  
wt idl
  1   24  14  223   312  235  463  123   58   17    0 
 408   39  19  
16  26
  3   27  18  278   866  611  520  131   58   17    0 
1096   34  13  
19  34
CPU minf mjf xcal  intr ithr  csw icsw migr smtx  srw
syscl  usr sys  
wt idl
  1   36  19  660   886  815  330   93   64   15    0
26158   41  32  
22   4
  3   29  20 3211  1292 1043  340   65   62   13    0
13893   25  29  
36  10


Best regards,

Dirk
prtconf -V will give you the OBP version you're
running. check this
against the
firmware patch for this machine:  106503-06 - latest
rev is listed in
the README
file for the patch -  OBP_3.22.0. The readme also
explains how to
install it.

cheers,
Mike

btw: part no for your cpu is most likely 501-445.



____________________________________________________________
Do You Yahoo!?
Get your free @yahoo.co.uk address at http://mail.yahoo.co.uk
or your free @yahoo.ie address at http://mail.yahoo.ie
Received on Fri Jul 27 14:47:11 2001

This archive was generated by hypermail 2.1.8 : Wed Mar 23 2016 - 16:25:00 EDT