Received-SPF: pass (google.com: domain of linux-kernel+bounces-210704-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Precedence: bulk
MIME-Version: 1.0
References: <20240610204821.230388-1-torvalds@linux-foundation.org>
 <20240610204821.230388-7-torvalds@linux-foundation.org> <20240611215556.GA3021057@thelio-3990X>
In-Reply-To: <20240611215556.GA3021057@thelio-3990X>
From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Tue, 11 Jun 2024 16:29:17 -0700
Message-ID: <CAHk-=wg4mwfbUQwEVDhEsA9MxZKTdTyOKTQa7n8M68_6fxo5Aw@mail.gmail.com>
Subject: Re: [PATCH 6/7] arm64: start using 'asm goto' for put_user() when available
To: Nathan Chancellor <nathan@kernel.org>
Cc: Peter Anvin <hpa@zytor.com>, Ingo Molnar <mingo@kernel.org>, Borislav Petkov <bp@alien8.de>, 
	Thomas Gleixner <tglx@linutronix.de>, Rasmus Villemoes <linux@rasmusvillemoes.dk>, 
	Josh Poimboeuf <jpoimboe@kernel.org>, Catalin Marinas <catalin.marinas@arm.com>, 
	Will Deacon <will@kernel.org>, Linux Kernel Mailing List <linux-kernel@vger.kernel.org>, 
	"the arch/x86 maintainers" <x86@kernel.org>, linux-arm-kernel@lists.infradead.org, 
	linux-arch <linux-arch@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"

On Tue, 11 Jun 2024 at 14:56, Nathan Chancellor <nathan@kernel.org> wrote:
>
> > +#ifdef CONFIG_CC_HAS_ASM_GOTO
>
> This symbol was eliminated two years ago with commit a0a12c3ed057 ("asm
> goto: eradicate CC_HAS_ASM_GOTO") since all supported compilers have
> support for it.

Hah. And I was trying to be a good boy and keep the old setup working.

Instead - because the HAS_ASM_GOTO config variable no longer exists -
I didn't actually test the new case at all, and it only worked because
the old case did in fact work.

Because fixing the broken #ifdef also showed that the

  +       _ASM_EXTABLE_##type##ACCESS_ZERO(1b, %l2)

line was wrong and was a copy-and-paste error from the get_user case
(that zeroes the result register on error).

It should be just

  +       _ASM_EXTABLE_##type##ACCESS(1b, %l2)

and to make the nasty copy_to_kernel_nofault_loop() build I also need
to do the proper _ASM_EXTABLE_KACCESS macro without the zeroing that
didn't exist.

Oops.

It would be nice to get rid of the CC_HAS_ASM_GOTO_OUTPUT thing too,
but that's probably a decade away ;(

But at least  this made me now go and actually test the _actual_ old
compiler case (no asm goto output). Perhaps ironically, I did get
*that* one right. That's the case where I had actually checked the new
code for get_user().

             Linus