On 18 Dec 2005 at 17:38, Thor Lancelot Simon wrote:
> There are several issues.  The most obvious one, it seems to me, is that
> this is going to repeatedly flush and reload libc when it ought to stay
> resident in the cache, since it will be at a different virtual address in
> each process.  If that's not the case, I'd like to know why it's not the
> case.

let's turn the question around: what makes you think that a VIVT
cache does not need to be flushed on a context switch? it's an
instant local root if it's not (i modify my libc, incoming suid
app happily executes my code). and as i said before, if an arch
cannot have randomization (e.g., transmeta/CMS is known to suffer),
then it won't, still nothing is lost for the rest.