Re: [PATCH] clean up PPC for private futex changes.

[Steven Munroe <munroesj at us dot ibm dot com>]

Jakub Jelinek wrote:

> On Mon, Jul 23, 2007 at 09:10:27AM -0700, Ulrich Drepper wrote:
>  
>
> > -----BEGIN PGP SIGNED MESSAGE-----
> > Hash: SHA1
> >
> > Applied.
> >    
>
> Unfortunately that patch contains several important bugs:
> 1) lll_futex_wait etc. had only correct definition when
>   __ASSUME_PRIVATE_FUTEX, when that is not true, it would
>   call futex syscall with bit 7 set even when libpthread.so
>   init determined it is not supported
>  
Ya, I was looking at the x86_64 code which did seem incomplete to me. So 
I tried to go one step better for PPC.

But how can header.private_futex be set in the absence of  
__ASSUME_PRIVATE_FUTEX?

> 2) lll_private_futex_wait etc. macros were shared when
>   not __ASSUME_PRIVATE_FUTEX
> 3) FUTEX_WAKE_OP had the wake operation argument also ored
>   with 128 in some cases
>  

Ya, I was concerned about that, but the examples from i386 and x86_64 
where not clear to me...

> Here is how it IMHO should look like (built and tested on ppc64-linux,
> unfortunately not wiht a 2.6.23ish kernel).
>  

This looks better. Thanks