Re: [RFC v2 0/7] arm64: return address signing

From: Kees Cook <keescook@chromium.org>
To: Will Deacon <will.deacon@arm.com>
Cc: Mark Rutland <Mark.Rutland@arm.com>,
	Luke Cheeseman <Luke.Cheeseman2@arm.com>,
	Diogo Sampaio <Diogo.Sampaio@arm.com>,
	Luke Cheeseman <luke.cheeseman@arm.com>,
	Catalin Marinas <Catalin.Marinas@arm.com>,
	Ard Biesheuvel <ard.biesheuvel@linaro.org>,
	Kristina Martsenko <Kristina.Martsenko@arm.com>,
	Ramana Radhakrishnan <Ramana.Radhakrishnan@arm.com>,
	Amit Kachhap <Amit.Kachhap@arm.com>,
	Kristof Beyls <Kristof.Beyls@arm.com>,
	Christof Douma <Christof.Douma@arm.com>,
	Suzuki Poulose <Suzuki.Poulose@arm.com>,
	Dave P Martin <Dave.Martin@arm.com>,
	"linux-arm-kernel@lists.infradead.org"
	<linux-arm-kernel@lists.infradead.org>
Subject: Re: [RFC v2 0/7] arm64: return address signing
Date: Sun, 2 Jun 2019 08:43:55 -0700	[thread overview]
Message-ID: <201906020843.140EC55FB@keescook> (raw)
In-Reply-To: <20190531092202.GA19208@fuggles.cambridge.arm.com>

On Fri, May 31, 2019 at 10:22:02AM +0100, Will Deacon wrote:
> On Thu, May 30, 2019 at 11:05:15AM -0700, Kees Cook wrote:
> > On Thu, May 30, 2019 at 04:55:08PM +0000, Luke Cheeseman wrote:
> > > The semantics of this attribute are straightforward enough but it
> > > raises some questions. One question being why would I want to turn off
> > > BTI (also controlled by this option) for one function in a file? Which
> > > gets a bit odd.
> > 
> > It's about leaving very early CPU startup functions in the kernel from
> > getting marked up (since they are running before or during the PAC setup).
> > 
> > > I don't know if the alternatives have been suggested but it's
> > > possible to achieve the result you seem to be after (a function without
> > > return address signing) in a couple of ways. First and foremost,
> > > separating the function out into it's own file and compiling with
> > > -mbranch-protection=none. Alternatively, writing the function in assembly
> > > or perhaps even a naked function with inline assembly.
> > 
> > Fair enough. :) Thanks for the clarification. Yeah, split on compilation
> > unit could work. (In the future, though, having the attribute flexibility
> > would be nice.)
> > 
> > Kristina, would it be feasible to split these functions into a separate
> > source file? (There doesn't seem to be a need to inline them, given
> > they're not performance sensitive and only used once, etc?)
> 
> Right, and we could call it kernel.c
> 
> Sarcasm aside, please fix this in the toolchain. Moving logically unrelated
> functions into one file just because the toolchain doesn't yet support this
> feature just messes up the codebase and removes the incentive to get this
> implemented properly. After all, you need something to do now that asm goto
> is out of the way, right? ;)

LLVM tracking bug created...
https://bugs.llvm.org/show_bug.cgi?id=42095

-- 
Kees Cook

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel