Hello Amadeo,
...
0x819dbff4:generic/src/proc/thread.o:cushion()+0x00000059
Thanks, Amadeo, this is interesting. I would say that the boot process
actually got quite far. But to be able to say anything for sure, I would
have to see the rest of the output (i.e. above this line).
Especially if there is some specific panic message (assertion message),
that would be of particular importance. But even the complete stack
trace might tell us much more.
If you have troubles to see the complete stack trace, then try to
disable the SMP support in compile time, just for the sake of debugging.
I have tried with gPXE as well, with same result.
OK, I'll try PXE booting myself, as my time permits.
Nevertheless, I have been able to run HelenOS, from CD, in a 48-cores
AMD Opteron Magny-Cours server, which unblocks me for now. However, when
I execute 'cpus' in the kconsole, it tells me only the first eight cores
are active. Is this a HelenOS limitation? Is there some flag I need to
turn on in order to have all the 48 cores active?
There are no explicit limitations on the number of CPUs, but there might
be some implicit limitations. There might be a problem with the
enumeration of the CPUs in your machine or a problem with waking them
up. Again, it is hard to tell without a detailed investigation.
In user space, try to run
edit /log/klog
and examine the entire kernel log. This might give us some hints why
only 8 CPUs were detected.
My suspicion is that your machine has actually 6 NUMA nodes with 8 cores
each, and it is not configured to show the other CPUs to non-NUMA-aware
OSes (i.e. it is not in the "transparent" mode).
HelenOS has NUMA support thanks to Vojtech Horky [1], but it hasn't been
integrated into the mainline yet.
[1] http://www.helenos.org/doc/theses/vh-thesis.pdf
M.D.
_______________________________________________
HelenOS-devel mailing list
[email protected]
http://lists.modry.cz/cgi-bin/listinfo/helenos-devel
