linux-kernel - Re: [PATCH] Partially revert "compiler: enable CONFIG_OPTIMIZE

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20191001161453.GO42880@e119886-lin.cambridge.arm.com>
Date:   Tue, 1 Oct 2019 17:14:53 +0100
From:   Andrew Murray <andrew.murray@....com>
To:     Russell King - ARM Linux admin <linux@...linux.org.uk>
Cc:     Will Deacon <will@...nel.org>,
        Masahiro Yamada <yamada.masahiro@...ionext.com>,
        Arnd Bergmann <arnd@...db.de>,
        Catalin Marinas <catalin.marinas@....com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        linux-arm-kernel <linux-arm-kernel@...ts.infradead.org>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Nicolas Saenz Julienne <nsaenzjulienne@...e.de>
Subject: Re: [PATCH] Partially revert "compiler: enable
 CONFIG_OPTIMIZE_INLINING forcibly"

On Tue, Oct 01, 2019 at 04:48:14PM +0100, Russell King - ARM Linux admin wrote:
> On Tue, Oct 01, 2019 at 04:28:27PM +0100, Andrew Murray wrote:
> > I hadn't noticed the use of __OPTIMIZE__ - indeed if __compiletime_assert
> > is no-op'd and you reach it then you won't have a build error - but you
> > may get uninitialised values instead.
> > 
> > Presumably the purpose of __OPTIMIZE__ in this case is to prevent getting
> > an undefined function error for the __compiletime_assert line, even though
> > it doesn't get called (when using a compiler that doesn't optimize out the
> > call to the unused function).
> > 
> > Why is the call to __get_user_bad not guarded in this way for when
> > __OPTIMIZE__ isn't set, i.e. why doesn't it suffer from the issue
> > that the following fixes?
> 
> Officially, the kernel does not support building with -O0.  To start
> with, the top level makefile has:
> 
> ifdef CONFIG_CC_OPTIMIZE_FOR_SIZE
> KBUILD_CFLAGS   += -Os
> else
> KBUILD_CFLAGS   += -O2
> endif
> 
> and we've said for years that the kernel relies upon the compiler
> optimiser to build correctly.  You may be lucky if you pass it via
> some method to 'make' but that's going to rely on the argument order
> to the compiler, and the order in which the compiler processes its
> arguments, and whether it (for example) correctly disables all
> optimisations if it encounters -O0 somewhere.

So in practice, __OPTIMIZE__ will likely always be set, as far as I
can tell it's supported in GCC, Clang and Intel compilers.  Though
the exception to this is for crypto/jitterentropy.c where the -O0 flag
is unconditionally set.

Are there other exceptions to this in terms of compilers?

Perhaps it may be possible to use BUILD_BUG after all.

Thanks,

Andrew Murray

> 
> -- 
> RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
> FTTC broadband for 0.8mile line in suburbia: sync at 12.1Mbps down 622kbps up
> According to speedtest.net: 11.9Mbps down 500kbps up