Re: [PATCH] make INIT_FS use the __RW_LOCK_UNLOCKED initialization

[Al Viro <viro@ZenIV.linux.org.uk>]

On Wed, Dec 10, 2008 at 08:11:02PM -0500, Steven Rostedt wrote:
> 
> On Wed, 10 Dec 2008, Mike Frysinger wrote:
> 
> > On Wed, Dec 10, 2008 at 18:43, Al Viro wrote:
> > > On Wed, Dec 10, 2008 at 06:37:28PM -0500, Steven Rostedt wrote:
> > >>
> > >> The following patch is in:
> > >>
> > >>   git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace.git
> > >>
> > >>     branch: cleanups
> > >
> > > Would be better to take init_fs to something in fs/ or kernel/ and kill
> > > the duplicates, along with INIT_FS macro itself...
> > 
> > how about the next level: unify init_task.c.  a quick check shows that
> > like 90% of the stuff in there is the same for everyone ...
> > -mike
> 
> I could probably do it for x86 and PPC, but beyond that, each arch 
> maintainer will have to do it separately. That's a change that would need 
> testing.
> 
> These changes are trivial (something I need to get RT working).

Guys, let's not take it to init/main.c, please.  That file is basically
a trashcan and I'd rather take that stuff to more or less relevant places
in kernel/* and fs/*.  FWIW, I've done that with init_files (fs/file.c)
some time ago.  For init_fs...  Hell knows.  Might make sense to take
that to separate file, actually (kernel/exit.c and kernel/fork.c bits
related to fs_struct).

BTW, why does sparc64 export put_fs_struct?  And is there any good reason
for doing copy_fs_struct() in nfsd instead of just teaching kthread_create()
how to create threads without sharing fs_struct (or descriptors, while we
are at it)?