Re: [GIT PULL] Block driver changes for 5.20-rc1

From: Jens Axboe <axboe@kernel.dk>
To: Nick Desaulniers <ndesaulniers@google.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>,
	Mark Rutland <mark.rutland@arm.com>,
	Josh Poimboeuf <jpoimboe@redhat.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Jiri Slaby <jirislaby@kernel.org>,
	Keith Busch <kbusch@kernel.org>, Christoph Hellwig <hch@lst.de>,
	Sagi Grimberg <sagi@grimberg.me>, Hannes Reinecke <hare@suse.de>,
	"linux-block@vger.kernel.org" <linux-block@vger.kernel.org>,
	linux-toolchains <linux-toolchains@vger.kernel.org>,
	Nick Clifton <nickc@redhat.com>
Subject: Re: [GIT PULL] Block driver changes for 5.20-rc1
Date: Thu, 4 Aug 2022 10:17:59 -0600	[thread overview]
Message-ID: <69b3bb50-e07f-cd7d-ff10-01856dd50053@kernel.dk> (raw)
In-Reply-To: <CAKwvOdmD10yK0r1H-M2PcnZgy3M0aA9gdkY0BErDOQ+KpBRxVQ@mail.gmail.com>

On 8/3/22 12:06 PM, Nick Desaulniers wrote:
>> ld: warning: arch/x86/boot/compressed/efi_thunk_64.o: missing .note.GNU-stack section implies executable stack
>> ld: NOTE: This behaviour is deprecated and will be removed in a future version of the linker
> 
> Right, arch/x86/boot/compressed (and arch/x86/boot/) discard/reset
> KBUILD_AFLAGS (via the := assignment operator).
> 
> So arch/x86/boot/Makefile and arch/x86/boot/compressed/Makefile also
> will need changes similar to the one you did below.
> 
> Finally, if those parts of the code actually expect the stack to be
> executable (probably depends on some inline asm), I suspect we might
> get some kind of fault at runtime (though I don't know how the kernel
> handles segment permissions or even uses ELF segments).  Point being
> that -Wa,--noexecstack is somewhat a promise that we wont try to
> execute data on the stack as if it were code.  -Wa,--execstack is
> useful if we need to be able to do so, but we should be careful to
> limit that to individual translation units if they really truly need
> that.  The linker warning is more so about the current ambiguity since
> if the implicit default changes in the future, that could break code.
> Better for us to be explicit here, and disable executable stack unless
> strictly necessary and only as necessary IMO.  Personally, I think the
> current implicit default is wrong, but pragmatically I recognize that
> people have been used to the status quo for years, and changing it
> could break existing codebases.
> 
> If you send the below diff with the two other additions I suggest to
> the x86 ML, the x86 maintainers might be able to comment if they're
> familiar with any possible cases where the stack is expected to be
> executable.

I'd be happy for you to take this over, I'm really just the reporter
here...

-- 
Jens Axboe