linux-kernel - Re: AMD / Memory Protection Keys

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

Message-ID: <01f3cf52-bb1e-f6a0-5810-c5e62ce5f302@amd.com>
Date:   Sat, 2 May 2020 06:50:25 +0700
From:   Suravee Suthikulpanit <suravee.suthikulpanit@....com>
To:     Dave Hansen <dave.hansen@...el.com>, linux-kernel@...r.kernel.org,
        x86@...nel.org
Cc:     tglx@...utronix.de, mingo@...hat.com, bp@...en8.de, hpa@...or.com,
        jon.grimm@....com,
        Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
        Thomas Lendacky <Thomas.Lendacky@....com>
Subject: Re: AMD / Memory Protection Keys

Dave,

On 4/30/20 2:18 AM, Dave Hansen wrote:
> The docs for AMD's implementation of protection keys showed up[1].
> Welcome to the party!
> 
> It all looks fine, except for this nugget:
> 
> 	The MPK mechanism is ignored in the following cases:
> 	... for pages marked in the paging structures as read-only
> 	(R/W=0) or as supervisor addresses (U/S=0)
> 
> That R/W=0 would mean that you can't access-disable read-only pages,
> which seems a bit goofy.  It's certainly a feature that I could imagine
> folks wanting to have.  Read-only pages might happen in unexpected
> places for things like mmap(PROT_WRITE)'d files or even CoW pages after
> fork().
> 
> Is this an error in the docs?
> 
> 1. https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.amd.com%2Fsystem%2Ffiles%2FTechDocs%2F24593.pdf&amp;data=02%7C01%7Csuravee.suthikulpanit%40amd.com%7C9ea21438bce6488589a308d7ec720d0d%7C3dd8961fe4884e608e11a82d994e183d%7C0%7C0%7C637237847630823928&amp;sdata=LZLQjvlVT5xSnjyRt0alN0lHleh8VD7Sxx53RL1UrX0%3D&amp;reserved=0
> 

Yes, we have checked with the hardware team and this is a mistake in the documentation.
The access disable bit (ADi) bit does apply to read-only pages. We will be correcting
the documentation. Thanks for pointing it out :)

Suravee