Jan Kiszka wrote:
Jan Kiszka wrote:

Hi all,

here is an improved version of the kgdb-over-ipipe patch. This version
specifically addresses the concerns Gilles brought up recently:

o No more dependencies on Xenomai, thus also no need to have the nucleus
 compiled into the kernel. There is now a registrable handler,
 ipipe_safe_current, which is invoked by the kgdb code to obtain
 "current". When the xeno nucleus arms its services, it overloads this
 handler to always provide a valid current (either the real one or
 init_task for kernel-only threads).

o Replaced smp_processor_id with ipipe_processor_id in critical kgdb
 code. I haven't tested this replacement, so no guarantees here. But so
 far it looks consistent - at least for me.

To use the kernel debugger with Xenomai, you need the latest kgdb
patches [1] and have to follow the attached patch series.

Oops, one patch was missing: the one to add the handler registration to
Xenomai. See attachment.



Index: ksrc/nucleus/pod.c
--- ksrc/nucleus/pod.c  (Revision 1095)
+++ ksrc/nucleus/pod.c  (Arbeitskopie)
@@ -43,6 +43,9 @@

Ok with the basic idea, but this code should definitely go to the arch-dependent section and not to the nucleus core directly, probably to include/asm-*/system.h, xnarch_init would be the proper place.


Xenomai-core mailing list

Reply via email to