This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] PowerPC thread pointer bias requires powerpc/lowlevelrobustlock.sym


On Wed, Nov 01, 2006 at 06:14:48PM -0600, Ryan S. Arnold wrote:
> I've been sitting on this patch for way too long.  
> 
> The TID implementation in the default lowlevelrobustlock.sym is wrong
> for PowerPC.  The thread point on PowerPC is biased by -0x7000 to extend
> the size of the FAST TLS to 64KB.  This patch provides a powerpc
> implementation to take this into account.

2006-11-01  Ryan S. Arnold  <rsa@us.ibm.com>

        * sysdeps/unix/sysv/linux/powerpc/lowlevelrobustlock.sym: New file.
        Powerpc thread pointer bias requires separate implementation of TID
        for powerpc.

Can you explain how this patch can change a single bit in the powerpc libraries?
lowlevelrobustlock.sym is the source for glibc buildtime generated
lowlevelrobustlock.h header.
This header is used in exactly 3 files in glibc sources:

nptl/sysdeps/unix/sysv/linux/sh/lowlevelrobustlock.S:#include <lowlevelrobustlock.h>
nptl/sysdeps/unix/sysv/linux/i386/i486/lowlevelrobustlock.S:#include <lowlevelrobustlock.h>
nptl/sysdeps/unix/sysv/linux/x86_64/lowlevelrobustlock.S:#include <lowlevelrobustlock.h>

as the directories clearly suggest, these have nothing to do with powerpc.

Uli, as TID is defined in tcb-offsets.{sym,h}, I think we should just nuke
lowlevelrobustlock.sym and s/<lowlevelrobustlock.h>/<tcb-offsets.h>/
in those 3 files.

	Jakub


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]