linux-kernel - Re: [PATCH] asm-generic: Force inlining of get_order() to work around gcc10 poor decision

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: <188e00e1-ae41-693e-1d05-f8d87e7ee696@csgroup.eu>
Date:   Mon, 19 Oct 2020 10:54:40 +0200
From:   Christophe Leroy <christophe.leroy@...roup.eu>
To:     Segher Boessenkool <segher@...nel.crashing.org>
Cc:     Joel Stanley <joel@....id.au>, Arnd Bergmann <arnd@...db.de>,
        Masahiro Yamada <yamada.masahiro@...ionext.com>,
        Andrew Morton <akpm@...ux-foundation.org>,
        linux-arch@...r.kernel.org,
        linuxppc-dev <linuxppc-dev@...ts.ozlabs.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] asm-generic: Force inlining of get_order() to work around
 gcc10 poor decision



Le 19/10/2020 à 10:32, Segher Boessenkool a écrit :
> On Mon, Oct 19, 2020 at 07:50:41AM +0200, Christophe Leroy wrote:
>> Le 19/10/2020 à 06:55, Joel Stanley a écrit :
>>>> In the old days, marking a function 'static inline' was forcing
>>>> GCC to inline, but since commit ac7c3e4ff401 ("compiler: enable
>>>> CONFIG_OPTIMIZE_INLINING forcibly") GCC may decide to not inline
>>>> a function.
>>>>
>>>> It looks like GCC 10 is taking poor decisions on this.
> 
>>> 1952 bytes smaller with your patch applied. Did you raise this with
>>> anyone from GCC?
>>
>> Yes I did, see https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97445
>>
>> For the time being, it's at a standstill.
> 
> The kernel should just use __always_inline if that is what it *wants*;
> that is true here most likely.  GCC could perhaps improve its heuristics
> so that it no longer thinks these functions are often too big for
> inlining (they *are* pretty big, but not after basic optimisations with
> constant integer arguments).
> 

Yes I guess __always_inline is to be added on functions like this defined in headers for exactly 
that, and that's the purpose of this patch.

However I find it odd that get_order() is outlined by GCC even in some object files that don't use 
it at all, for instance in fs/pipe.o

Christophe