linux-kernel - Re: Linux 4.10-rc1

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] [day] [month] [year] [list]

Message-ID: <20161226151422.GQ1555@ZenIV.linux.org.uk>
Date:   Mon, 26 Dec 2016 15:14:22 +0000
From:   Al Viro <viro@...IV.linux.org.uk>
To:     Hanjun Guo <guohanjun@...wei.com>
Cc:     Kefeng Wang <wangkefeng.wang@...wei.com>,
        Linus Torvalds <torvalds@...ux-foundation.org>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Will Deacon <will.deacon@....com>,
        Catalin Marinas <catalin.marinas@....com>
Subject: Re: Linux 4.10-rc1

On Mon, Dec 26, 2016 at 02:23:43PM +0000, Al Viro wrote:
> On Mon, Dec 26, 2016 at 05:45:10PM +0800, Hanjun Guo wrote:
> > On 2016/12/26 17:18, Al Viro wrote:
> > > On Mon, Dec 26, 2016 at 05:05:37PM +0800, Kefeng Wang wrote:
> > >> It looks like we need revert the changes from assembly files.
> > 
> > I tested Kefeng's patch and works. more comments below.
> > 
> > > Better yet, split the damn thing in two and include the asm-only part.
> > 
> > split will cause other places for building errors, because there is no
> > 
> > -#ifndef __ASSEMBLY__
> 
> Huh?  It's included only from assembler files, so what would use the other
> parts?

FWIW, my arguments for splitting it are
	* asm and non-asm parts have almost no overlap - only
uaccess_{en,dis}able_not_uao (as asm macro and static inline
resp.), but that's it.  It's not as if there had been arseloads of
constants shared between C and assembler, etc.
	* having no asm/uaccess.h includes left allows to consolidate stuff
into linux/uaccess.h; sure, in this case the stuff getting moved there would be
under ifndef anyway, but "no includes outside of linux/uaccess.h" is easier
to verify than "no includes outside of linux/uaccess.h and arch/arm64/.../*.S"

I can live with reverting those several includes to asm/uaccess.h (all
interesting stuff is under that ifndef), but I think splitting the asm
part away would be cleaner.

PS: if that variant does cause any build errors, I would very much like to
see .config.  Hanjun, could you post one that is triggering those?