Gentoo Archives: gentoo-dev

From: "Anthony G. Basile" <blueness@g.o>
To: gentoo-dev@l.g.o
Subject: Re: [gentoo-dev] The status of grsecurity upstream and hardened-sources downstream
Date: Sat, 24 Jun 2017 12:03:04
Message-Id: 3353b2f4-1080-7575-ed1a-cbeb5c94a37a@gentoo.org
In Reply to: Re: [gentoo-dev] The status of grsecurity upstream and hardened-sources downstream by Alexis Ballier
1 On 6/24/17 6:04 AM, Alexis Ballier wrote:
2 > On Fri, 23 Jun 2017 12:28:27 -0400
3 > "Anthony G. Basile" <blueness@g.o> wrote:
4 >
5 >> Hardened Gentoo has two sides to it, kernel hardening (done via
6 >> hardened-sources) and toolchain/executable hardening. The two are
7 >> interrelated but independent enough that toolchain hardening can
8 >> continue on its own. The hardened kernel, however, provided PaX
9 >> protection for executables and this will be lost. We did a lot of
10 >> work to properly maintain PaX markings in our package management
11 >> system and there was no part of Gentoo that wasn't touched by issues
12 >> stemming from PaX support.
13 >
14 >
15 > Good luck to them at providing a complete userland ecosystem for using
16 > pax protection. Good luck at getting people accept and review their
17 > often crashing asm patches at upstream projects that won't even be able
18 > to test their benefits.
19 >
20 > Maybe we should start a business for this ? :)
21 > http://static.sstic.org/videos2015/SSTIC_2015-06-03_P08_CLIP.mp4
22 > (This is for Patrice)
23
24 Correct. Zorry, myself and others on the hardened team did a lot to
25 make userland play nice with the hardened-kernel. It represents most of
26 my effort in Gentoo.
27
28 >
29 >
30 >
31 > We'll need to decide what to do with things like USE=pic. For media
32 > packages this is not something you usually want to enable as you can
33 > bear the 10Mb relocations at startup to have 10% or more performance
34 > improvement when reading your 2hours long movie.
35
36 It will be a mess going forward. We will necessarily have to start
37 dropping pax related stuff, if for no other reason than we can't support
38 making a package work under pax if we have no pax enabled kernel to test
39 on. Once this is gone, such bugs will float upstream to pipacs and
40 spender. "Good luck" is right.
41
42 >
43 >
44 > Alexis.
45 >
46
47
48 --
49 Anthony G. Basile, Ph.D.
50 Gentoo Linux Developer [Hardened]
51 E-Mail : blueness@g.o
52 GnuPG FP : 1FED FAD9 D82C 52A5 3BAB DC79 9384 FA6E F52D 4BBA
53 GnuPG ID : F52D4BBA