[RFC][PATCH] retprobes - remove orig_ret_addr usage

[Ananth N Mavinakayanahalli]

Lynch, Rusty wrote:
>>Hi,
>>
>>Here is a patch against 2.6.12-rc6-mm1 that removes usage of the static
>>void *orig_ret_addr. We now populate the ret_addr from the previous
>>kretprobe_instance in cases where the return address on stack has been
>>changed to kretprobe_trampoline().
>>
>>Changes only done to i386. x86_64 may be similar. Rusty?
>>
>>Ananth
> 
>  
> Yes, this should work for x86_64.  I'm not so sure it is a good idea to
> include the likely() wrapping comparison of *sara to
> kretprobe_trampoline since we don't really know how this will be used.
> (If for what ever reason a usage model tended to always have multiple
> return probes registered for the same function, then the likely() call
> would have a negative impact to performance.)
> 
>     --rusty

Agreed. I will make this change as also the x86_64 modification and send
an updated patch.

Ananth