This is the mail archive of the
libc-alpha@sources.redhat.com
mailing list for the glibc project.
Re: new ppc32 GOT/PLT support for mcount
- From: Alan Modra <amodra at bigpond dot net dot au>
- To: Jakub Jelinek <jakub at redhat dot com>
- Cc: Steve Munroe <sjmunroe at us dot ibm dot com>, libc-alpha at sources dot redhat dot com
- Date: Mon, 13 Jun 2005 13:10:33 +0930
- Subject: Re: new ppc32 GOT/PLT support for mcount
- References: <42A8C35C.1050209@us.ibm.com> <20050611085852.GF7663@sunsite.mff.cuni.cz>
On Sat, Jun 11, 2005 at 10:58:52AM +0200, Jakub Jelinek wrote:
> On Thu, Jun 09, 2005 at 05:31:56PM -0500, Steve Munroe wrote:
[about _mcount]
> This looks wrong, for multiple reasons:
Yes, my fault. When looking at _mcount for ppc, I noticed that gcc
carefully sets up r0 to point at a count var, and the ppc _mcount stub
in glibc commented about the r0 param too. Of course, glibc's _mcount
needs no count var, but I'd forgotten that. Now, if glibc's _mcount
really did need a pointer to the counter, then we'd be in trouble
because the new lazy sym resolver call stub trashes r0. ie. we couldn't
call _mcount via the plt. So I thought we'd need to move _mcount to
libc_nonshared.a. I put a fixme in my new plt/got libc patch, and told
Steve about the _mcount "problem". Really, all that needs doing is
fixing the comment in ppc-mcount.S, and cleaning up
sysdeps/powerpc/powerpc32/sysdep.h. Sorry for any wasted effort. :-(
I'll generate a new glibc patch.
--
Alan Modra
IBM OzLabs - Linux Technology Centre