RE: [Bug translator/1276] support more timer varieties

["Stone, Joshua I" <joshua dot i dot stone at intel dot com>]

Keshavamurthy, Anil S wrote:
> 	We fixed similar return probe bug in RHEL4-U3 kernel. So, Can
you
> please try the same test on  RHEL4-U3 Beta and get back to the
> mailing list. 

I updated only the kernel-* rpms - that's all that matters here, right?
I'm now on 2.6.9-24.ELsmp...

I get the same result, only with the line number changed: "Kernel BUG at
kprobes:440".  This line 440 is still the same BUG_ON check as I posted
before.

Here is the script I use to trigger it:

  global i
  probe begin { log("starting probe") }
  probe kernel.function("__switch_to").return { ++i }
  probe end { printf("ending probe (%d)\n", i) }

I also trimmed it down to a very small C file that still exhibits the
behavior, included below.

Let me know if there's more info I can provide...

Thanks,

Josh


===================================================================
#include <linux/module.h>
#include <linux/kprobes.h>

#ifdef ARCH_SUPPORTS_KRETPROBES
static int
probe_enter (struct kretprobe_instance *probe_instance, struct pt_regs
*regs) {
  return 0;
}

//
kernel.function("__switch_to@arch/x86_64/kernel/process.c:433").return
static struct kretprobe probe = {
  .kp.addr= (void *) 0xffffffff8010ebea,
  .handler = &probe_enter,
  .maxactive = 0,
};
#endif /* ARCH_SUPPORTS_KRETPROBES */

int init_module(void) {
  #ifdef ARCH_SUPPORTS_KRETPROBES
  return register_kretprobe (&probe);
  #else
  return -1;
  #endif
}

void cleanup_module(void) {
  #ifdef ARCH_SUPPORTS_KRETPROBES
  unregister_kretprobe (&probe);
  #endif
}

MODULE_DESCRIPTION("systemtap probe");
MODULE_LICENSE("GPL");