Re: new execve/kernel_thread design

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

On Thu, Nov 15, 2012 at 05:41:16PM +0100, Michal Simek wrote:
> Here is the branch based on rc5 (information below)
> and here is giweb.
> http://developer.petalogix.com/git/gitweb.cgi?p=linux-2.6-microblaze.git;a=shortlog;h=refs/heads/viro/arch-microblaze-rc5
> 
> I have also looked at your sys_fork / sys_vfork / sys_clone unification
> and I have fixed it for Microblaze.
> 
> Also I have done some tests on it for sure.
> 
> I would add sys_execve/kernel_execve/kernel_thread patches to my next branch.
> Are you OK with that?

Umm...  In principle - yes, but I've a couple of question abouts those.

1) What's that set_fs(USER_DS) in start_thread() for?  Note that we do the same
thing in flush_old_exec(), at the same time we remove PF_KTHREAD from
current->flags.

While we are at it, if we *ever* hit do_signal() with KERNEL_DS, we are
very deep in trouble.  set_fs(USER_DS) in setup_{rt_,}frame() is pointless.

2) your definition of current_pt_regs() is an exact copy of on in
include/linux/ptrace.h; why is "microblaze: Define current_pt_regs"
needed at all?  IOW, I'd rather added #include <linux/ptrace.h> to
arch/microblaze/kernel/process.c instead...