Re: [Xenomai-core] latency -t 1 crashes on SMP.

[Jan Kiszka <jan.kiszka@domain.hid>]

[-- Attachment #1: Type: text/plain, Size: 735 bytes --]

Gilles Chanteperdrix wrote:
> Jan Kiszka wrote:
>  > o instrument the xenomai exception handler with an ipipe_trace_freeze()
>  >   (something which should be merged into SVN later)
> 
> The I-pipe tracer as well as Linux NMI handling have the same issue
> which is the cause for the exceptions 14: they rely on a correct
> "current" to read the processor id. The first attached patch, to be
> applied after the ipipe patch, add an ipipe_processor_id_notrace
> function, which is called instead of raw_smp_processor_id() in the
> second, I-pipe tracer, attached patch.
> 

Ah, good point! But why not make the original __ipipe_*_cpuid functions
non-traced? I don't think we would loose significant information.

Jan


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 250 bytes --]