[PATCH] group_mapping: Avoid a talloc

[Volker Lendecke]

On Thu, Jan 09, 2014 at 09:22:42AM +1300, Andrew Bartlett wrote:
> I can see the argument for stuff in tight inner loops, but I'm very,
> very wary of going back to fixed length strings everywhere.  What's the
> whole argument of talloc_pool() that this was fast enough, without all
> the risks of fixed length strings?

No, talloc_pool is not fast enough. If you have to make a
Rasperry Pi level box saturate a gigabit link every single
cycle counts. In particular the case where we have
sid2string and we know the maximum length of the target
string, there is value in not doing
malloc/talloc/talloc_pool.

> There is a great value in consistency of talloc-based string handling,
> because it means that we can realistically examine our fixed length
> string use, because it's rare, not used anywhere we might save a cycle
> or two outside the hot path.

Please take a look at full_path_tos(): The normal case will
do it on the stack, but it does provide an overflow
mechanism with talloc.

By the way, one of the hottest routines is talloc_free these
days. Every talloc has to go through that, even if it comes
from a talloc_pool.

Volker

-- 
SerNet GmbH, Bahnhofsallee 1b, 37081 Göttingen
phone: +49-551-370000-0, fax: +49-551-370000-9
AG Göttingen, HRB 2816, GF: Dr. Johannes Loxen
http://www.sernet.de, mailto:kontakt at sernet.de