Re: [AD] More "mprotect" troubles (4.2.1 showstopper?)

[Hans de Goede <j.w.r.degoede@xxxxxxxxxx>]

Catatonic Porpoise wrote:
> Hans de Goede <j.w.r.degoede@xxxxxxxxxx> wrote:
>
> > See:
> > https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=185214
> >
> > I'm working on a nice clean fix for this involving a temp file and mmap 
> > as described in the above url.
>
> That's really silly. I doubt there will be much of a
> performance gain after that, and the complexity involved is
> ridiculous. Just build without assembly and use the C
> routines on SELinux.

1) The fix is almost finished
2) The complexity isn't all that great
3) mremap is probably faster then or atleast as fast as realloc, so
   there is no speed loose
4) I do wonder about the speed gain with modern processors as the code
   generator also contains a loop, wouldn't it be better to put the loop
   in the actual asm code instead of unrolling it, that what make the
   code quite a bit smaller, which could be a real gain with modern
   processors intensive reliance on cache.

Regards,

Hans