[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200703222516.GW2786714@ZenIV.linux.org.uk>
Date: Fri, 3 Jul 2020 23:25:16 +0100
From: Al Viro <viro@...iv.linux.org.uk>
To: Andy Lutomirski <luto@...nel.org>
Cc: Linus Torvalds <torvalds@...ux-foundation.org>,
Michael Ellerman <mpe@...erman.id.au>,
Christophe Leroy <christophe.leroy@....fr>,
Josh Poimboeuf <jpoimboe@...hat.com>,
Peter Zijlstra <peterz@...radead.org>,
the arch/x86 maintainers <x86@...nel.org>,
Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: objtool clac/stac handling change..
On Fri, Jul 03, 2020 at 02:41:43PM -0700, Andy Lutomirski wrote:
> I still feel like the ex_handler-automatically-does-CLAC thing is an
> optimization that isn't worth it. Once we pull our heads out of the
> giant pile of macros and inlined functions, we're talking about
> changing:
> clac; jmp. But on the flip side, the jump folding pattern looks
> better like this:
>
> unsafe_uaccess_begin();
> if (unsafe_get_user(...))
> goto fail;
> if (unsafe_get_user(...))
> goto fail;
> unsafe_uaccess_end();
>
> fail:
> unsafe_uaccess_end();
>
> than like:
>
> unsafe_uaccess_begin();
> if (unsafe_get_user(...))
> goto fail;
> if (unsafe_get_user(...))
> goto fail;
> unsafe_uaccess_end();
>
> fail:
> /* not unsafe_uaccess_end(); because unsafe_get_user() has
> conditional-CLAC semantics */
First of all, user_access_begin() itself can bloody well fail. So you need
to handle that as well. And then it becomes nowhere near as pretty.
We can pretend that it's normal C; however, that's not true at all - there
are shitloads of things you can't do in such areas, starting with "call anything
other than a very small list of functions". It's not a normal C environment
at all.
My problem is not with having AC turned off in exception handler - it leads
to saner patterns, no arguments here. I'm not happy with doing doing that
on *every* exception, with no way to specify whether it should or should not
be done. It's not like it would've cost us anything to be able to specify
that - we have the third argument of _ASM_EXTABLE_HANDLE(), after all.
Powered by blists - more mailing lists