Re: [RFC PATCH 2/6] jump label v3 - x86: Introduce generic jump patching without stop_machine

[Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>]

* Masami Hiramatsu (mhiramat@redhat.com) wrote:
> Mathieu Desnoyers wrote:
> > We should really do performance benchmarks comparing stop_machine() and
> > the int3-based approach rather than to try to come up with tricky
> > schemes. It's not a real problem until we prove there is indeed a
> > performance regression. I suspect that the combined effect of cache-line
> > bouncing, worker thread overhead and the IPI of stop_machine is probably
> > comparable to the two IPIs we propose for int3.
> 
> I assume that total latency of XMC is almost same on normal-size SMP.
> However,
> - stop_machine() can't support NMI/SMI.
> - stop_machine() stops all other processors while XMC.

I would also add that stop_machine() increases the system interrupt
latency of an amount O(num_online_cpus()), which I'd like to avoid given
the 90- to 128-core machines heading our way pretty quickly.

> 
> Anyway, int3-based approach still needs to be ensured its safeness
> by processor architects. So, until that, stop_machine() approach
> also useful for some cases.

True. This makes me think: If performance happens to be a problem, we
could do batched jump instruction modification. Using an hash table to
contain the pointers would allow us to only perform a single pair of IPI
for a whole bunch of instruction modifications.

Mathieu

> 
> Thank you,
 

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68