5.2.1 SMP problems

[Toomas Aas]

Hi!

>     Installed 5.2.1 on a high load server - albeit not recommended - and 
> it was working perfectly with SMP support. It picked up 4 cpu's and 
> appeared to be using them all. (The box is a dual 2.4gig Xeon machine). 
> Anyway, the box was as I said under quite a bit of load, and would die 
> about once a day - just reboot. I looked into it a bit and found 
> whenever I put more load on (e.g. a make install in ports) it would 
> panic - vm_load_page or something like that - sorry about being so 
> vague. Anyway I took out SMP support, and now it boots with 2 cpu's - 
> one normal Xeon + HT for the second (no idea how to disable that - but I 
> don't need to)... its been running perfect for 4 days now and I put it 
> under SERIOUS load.. has anyone else had SMP problems on 5.2.1? 

I put 5.2.1-RELEASE on an IBM Netfinity 5000 (dual Pentium III) and I 
had problems, although different from yours. This system has some 
warning LEDs on the motherboard which light up when something is wrong. 
Quite often, when booting, the NMI and PCI1 LEDs would light up, the 
message "NMI ISA 34, EISA ff" appears among the boot messages and 
system reboots shortly after that. PCI1 LED indicates a problem with 
the first PCI bus, which is where I have IBM ServeRAID controller.

Interesting thing is that sometimes this problem during booting does 
not happen and once the system boots successfully there are never any 
problems.

When I removed the second CPU, the problem disappeared. I am not quite 
convinced yet that the problem is completely gone (haven't rebooted too 
many times) but it certainly has been greatly reduced.

Downgrading the firmware of ServeRAID controller from 6.10 to 5.10 (as 
recommended by another poster on this list) may also have contributed 
to resolving the problem. But the first thing I did was removing the 
2nd CPU and I successfully booted several times between that and 
downgrading the ServeRAID firmware. I *seem* to remember one NMI-boot 
after removing the second CPU, but I'm not quite sure my memory isn't 
failing me - it has been quite a rough ride with this NF5000 during the 
past few days. Anyway, *before* removing the second CPU, the boot 
failure rate was >50%, so I'm sure removing the second CPU did help, 
even if downgrading the ServeRAID firmware may have helped further.
--
Toomas Aas | toomas.aas at raad.tartu.ee | http://www.raad.tartu.ee/~toomas/
* Feeling compressed ARJ you?