Re: [EXTERNAL] Re: [GIT PULL] Please pull powerpc/linux.git powerpc-5.9-1 tag

On 8/8/20 12:44 AM, Al Viro wrote:

On Fri, Aug 07, 2020 at 10:46:13AM -0700, Linus Torvalds wrote:

quoted

On Fri, Aug 7, 2020 at 6:14 AM Michael Ellerman [off-list ref] wrote:

quoted

Just one minor conflict, in a comment in drivers/misc/ocxl/config.c.

Well, this morning I merged the ptrace ->regset_get() updates from Al,
and that brought in a different conflict.

I _think_ I resolved it correctly, but while the new model is fairly
readable, the old one sure wasn't, and who knows how messed up my
attempt to sort it out was. I don't know the pkey details on powerpc..

So I'd appreciate it if both Al and Aneesh Kumar would check that what
I did to pkey_get() in arch/powerpc/kernel/ptrace/ptrace-view.c makes
sense and works..

Grabbing...

Looks sane and yes, 3 membuf_store() instead of membuf_write() + membuf_store()
would make sense (might even yield better code).  Up to ppc folks...

quoted

Side note - it might have been cleaner to just make it do

         membuf_store(&to, target->thread.amr);
         membuf_store(&to, target->thread.iamr);
         return membuf_store(&to, default_uamor);

That will also allow to get rid of

	BUILD_BUG_ON(TSO(amr) + sizeof(unsigned long) != TSO(iamr));

I will followup with a cleanup patch.

quoted

instead of doing that membuf_write() for the first two ones and then
the membuf_store() for the uamor field, but I did what I did to keep
the logic as close to what it used to be as possible.

If I messed up, I apologize.

And if you agree that making it three membuf_store() instead of that
odd "depend on the exact order of the thread struct and pick two
consecutive values", I'll leave that to you as a separate cleanup.

                    Linus

-aneesh