Re: [PATCH 3/3] Use conditional

[Tony Jones]

On Sun, Jul 03, 2005 at 05:44:05PM +0200, Kurt Garloff wrote:

Agree with James, pls resend to linux-security-module@xxxxxxxxxx

The topic of replacing dummy (with capability) was discussed there
last week, in the context of stacker, but a common solution for both
cases would be needed.

Also, I was going to ask where 4/5 and 5/5 were :-)

If you are claiming a perf increase it would be nice to get an idea
what these patches were even though you believe most of the gain was
in patch #3.

Thanks

> Hi,
> 
> this optimizes the case where no LSM is loaded and the (new) default 
> capablities is used. These are not called via indirect calls but 
> called as hardcoded calls and might thus be inlined; the price for
> this is a conditional -- benchmarks done by hp showed this to be
> beneficial (on ia64).
> 
> Enjoy,
> -- 
> Kurt Garloff, Director SUSE Labs, Novell Inc.

> From: Kurt Garloff <garloff@xxxxxxx>
> Subject: Replace indirect calls by a branch
> References: SUSE40217, SUSE39439
> 
> In the LSM stub collection, rather do a branch than an indirect
> call. Many of the functions called do only return 0 or do nothing
> for the default (capability) case.
> This is a fast-path optimization; a branch is faster than an
> indirect call, even more so if correctly predicted.
> This shows a >3% perf. increase in netperf -t TCP_RR benchmark on IA64.
> (More exactly: The benchmark was taken with the next two patches
>  applied as well, but I attribute the main effect to this patch.)
> 
> This is patch 3/5 of the LSM overhaul.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/