reentrant probes

[Seth, Rohit]

Frank Ch. Eigler <mailto:fche@redhat.com> wrote on Monday, May 02, 2005
10:33 AM:

> Hi -
> 
> rohit.seth wrote:
> 
>> [...]  I think it will be quite useful to allow renentrancy in
>> probes.  Not sure why would you want to restrict this.  [...]
> 
> It seems hazardous, for example by permitting infinite regress, should
> a breakpoint be placed into the normal execution path of the kprobes
> layer itself.  It would also require supporting nested int3 faults,
> definitely requiring extra stack space.
> 

You will in all probability need to support a nested int3 faults (and
yes that increases the pressure on limited stack space further).  For
example, a scenario like where a probe handler is executing and then
gets an interrupt and as part of handling that interrupt, the execution
hits another probe....

> Maybe these are surmountable by some combination of run-time nesting
> limits (like the ones we'd have in translated recursive scripts), and
> blacklisting some modules (e.g. kprobes itself) from becoming
> systemtap probing targets.
> 

I agree some static time sanity check (to find infinte regress) will be
useful.



-rohit