Re: ppc64 vDSO in mainline

[Steve Munroe <sjmunroe at us dot ibm dot com>]

Jakub Jelinek <jakub@redhat.com> wrote on 03/28/2005 11:23:45 PM:

> On Mon, Mar 28, 2005 at 03:59:36PM -0800, Ulrich Drepper wrote:
> > Steve Munroe wrote:
> > >3) a function which is currently
> > >exported by libc , but a better optimized version (with a different
> > >symbol) is also exported by the VDSO.
> > 
> > There is no reason to add any complications or dependency problems for
> > this.  Just using a pointer in libc itself, a test for NULL and if 
not,
> > jump to the function is enough.  The penalty for this extra indication
> > if minimal compared to all the other work involved.
> 
> Yeah.
> 
> Please when thinking about making vDSO symbols directly usable about
> prelink as well.  What prelink needs is that all libraries stay at the
> preferred address (otherwise it falls back to doing normal symbol 
lookups
> on program start up; but as vDSO is mapped by the kernel, kernel will 
not
> know if it should randomize the address or use a fixed one).
> If there are any direct symbol lookups into vDSO, prelink would need 
special
> hacks for vDSO and kernel help as it can't rewrite the vDSO even if vDSO
> address was always fixed, but if it is randomized (which is desirable),
> vDSO just makes prelink unusable on ppc*.
> 

It sounds like randomization (independent of VDSO) invalidates prelink!

Nothing I proposed implies that prelinking has to be disabled, only that 
prelinking has to exclude VDSO (VDSO symbols).

Steven J. Munroe
Linux on Power Toolchain Architect
IBM Corporation, Linux Technology Center