Received-SPF: pass (google.com: domain of linux-kernel+bounces-209154-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240608193504.429644-2-torvalds@linux-foundation.org>
 <20240610104352.GT8774@noisy.programming.kicks-ass.net> <20240610120201.GAZmbrOYmcA21kD8NB@fat_crate.local>
 <CAHk-=wgb98nSCvJ-gL42mt+jt6Eyp-0QSMJLovmAoJOkQ_G3gQ@mail.gmail.com>
 <71FE7A14-62F6-45D3-9BC4-BE09E06F7863@zytor.com> <CAHk-=wjTzFYo2+eQJpb56Df8sNDW7JEV=_6Di2v-M5x2kv06_g@mail.gmail.com>
In-Reply-To: <CAHk-=wjTzFYo2+eQJpb56Df8sNDW7JEV=_6Di2v-M5x2kv06_g@mail.gmail.com>
From: Linus Torvalds <torvalds@linux-foundation.org>
Date: Mon, 10 Jun 2024 18:24:19 -0700
Message-ID: <CAHk-=wjdsN=dH41MO+gASWZkexCgrwK6CGT=NvpA3xsVXEhxBw@mail.gmail.com>
Subject: Re: [PATCH] x86: add 'runtime constant' infrastructure
To: "H. Peter Anvin" <hpa@zytor.com>
Cc: Borislav Petkov <bp@alien8.de>, Peter Zijlstra <peterz@infradead.org>, Ingo Molnar <mingo@kernel.org>, 
	Thomas Gleixner <tglx@linutronix.de>, Rasmus Villemoes <linux@rasmusvillemoes.dk>, 
	Josh Poimboeuf <jpoimboe@kernel.org>, 
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>, "the arch/x86 maintainers" <x86@kernel.org>, 
	linux-arch <linux-arch@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"

On Mon, 10 Jun 2024 at 18:09, Linus Torvalds
<torvalds@linux-foundation.org> wrote:
>
> Doing it in general is actually very very painful. Feel free to try -
> but I can almost guarantee that you will throw out the "Keep It Simple
> Stupid" approach and your patch will be twice the size if you do some
> "rewrite the whole instruction" stuff.
>
> I really think there's a fundamental advantage to keeping things simple.

I guess the KISS approach would be to have a debug mode that just adds
an 'int3' instruction *after* the constant. And then the constant
rewriting rewrites the constant and just changes the 'int3' into the
standard single-byte 'nop' instruction.

That wouldn't be complicated, and the cost would be minimal. But I
don't see it being worth it, at least not for the current use where
the unrewritten constant will just cause an oops on use.

                Linus