Martin Schwidefsky wrote:

>>It seems to me that if we do:
>>* s390x defines CONFIG_UID16
>>* typedef __kernel_old_gid_t to u16
>>* get rid of all the sys32_*16 stuff and just call the uid16.c function
> 
> I checked the code and didn't find any reason why this shouldn't work.
> In fact with the 31 bit emulation layer the 64 bit kernel does need the
> 16 bit uid/gid system calls although only for the emulation. To make it
> really perfect you could define CONFIG_UID16 dependent on
> CONFIG_S390_SUPPORT. This saves a few bytes in the image if the emulation
> support is not enabled.

Ok, done.  I don't know if you guys prefer BK or plain-old patch, so 
I'll do this:

Attached are the two patches (one as sent before, the other is just the 
above change).  You can apply the patches, or if you prefer, ping me 
back and I'll put a bk repository up publicly for you.  The patches are 
probably easiest.

This change will make my life easier for the rest of the ngroups crap I 
am fixing.  I will commit this locally on teh assumption it will turn up 
sooner or later on your end.

Now, for 2.5.x - do you want me to do another patchset, or are these 
close enough?

Thanks.




-- 
Tim Hockin
Systems Software Engineer
Sun Microsystems, Linux Kernel Engineering
thockin@sun.com