linux-kernel - Re: stop messing with set_fs in arm

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: <20200420102259.GA7862@lst.de>
Date:   Mon, 20 Apr 2020 12:22:59 +0200
From:   Christoph Hellwig <hch@....de>
To:     James Morse <james.morse@....com>
Cc:     Christoph Hellwig <hch@....de>,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org
Subject: Re: stop messing with set_fs in arm_sdei

On Tue, Apr 14, 2020 at 04:59:16PM +0100, James Morse wrote:
> Hi Christoph,
> 
> On 14/04/2020 15:23, Christoph Hellwig wrote:
> > can you take a look at this series?  I've been trying to figure out
> > what the set_fs in arm_sdei is good for, and could not find any
> > good reason.  But I don't have any hardware implementing this interface,
> > so the changes are entirely untested.
> 
> Its a firmware thing, think of it as a firmware assisted software NMI.
> 
> The arch code save/restores set_fs() because the entry code does that when taking an
> exception from EL1. SDEI does the same because it doesn't come via the same entry code. It
> does it in C because that C is always run before the handler, something that isn't true
> for the regular assembly version.
> 
> The regular entry code does this because any exception may have interrupted code that had
> addr_limit set to something else:
> https://bugs.chromium.org/p/project-zero/issues/detail?id=822
> 
> and the patch that fixed it: commit e19a6ee2460b "arm64: kernel: Save and restore UAO and
> addr_limit on exception entry"

Can you throw in a comment documenting this better?  And pick up the
first patch while we're at it - no need to expose such low-level
mechanisms to modules.