linux-kernel - Re: [PATCH] x86/mm/pkeys: Do not enable them by default

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 for Android: free password hash cracker in your pocket

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

Message-ID: <56C7558E.2060603@sr71.net>
Date:	Fri, 19 Feb 2016 09:49:02 -0800
From:	Dave Hansen <dave@...1.net>
To:	Borislav Petkov <bp@...en8.de>
Cc:	Dave Hansen <dave.hansen@...ux.intel.com>,
	linux-kernel@...r.kernel.org, brgerst@...il.com,
	tglx@...utronix.de, torvalds@...ux-foundation.org, hpa@...or.com,
	akpm@...ux-foundation.org, peterz@...radead.org, mingo@...nel.org,
	luto@...capital.net, dvlasenk@...hat.com, riel@...hat.com,
	Linus Torvalds <torvalds@...ux-foundation.org>
Subject: Re: [PATCH] x86/mm/pkeys: Do not enable them by default

Borislav, we're talking about 1566 bytes of text here:

>    text	   data	    bss	    dec	    hex	filename
> 13874312	2633704	3014656	19522672	129e470	64bit-pkey/vmlinux
> 13872746	2633648	3014656	19521050	129de1a	64bit-nopkey/vmlinux

For gains that small, we should barely even allow this thing to be
configurable, much less default it to off.

On 02/19/2016 09:23 AM, Borislav Petkov wrote:
> On Fri, Feb 19, 2016 at 09:11:03AM -0800, Dave Hansen wrote:
>> I'd really prefer this be left on by default.  This is a feature that I
>> expect to be widely enabled in distribution kernels.
> 
> Distribution kernels can enable it without defaulting to y here.

Yes, agreed.  Distros _can_ override things.  But, In general,
*especially* with user-visible effects, I'd really like defconfig (or
other build defaults) to be reasonably close to what distributions do.

> Also, this code doesn't need to be built on the majority of x86 boxes
> out there because they don't have the hw support.

My view has always been that the folks that really care about binary
size are the ones that will be the ones digging through their .configs
turning things off.

BTW, what percentage of x86 boxes must have a feature before we can
enable it by default?