linux-kernel - Re: [PATCH] crypto: gcm

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

Date:   Fri, 31 May 2019 14:14:04 +0800
From:   Herbert Xu <herbert@...dor.apana.org.au>
To:     Horia Geanta <horia.geanta@....com>
Cc:     Ard Biesheuvel <ard.biesheuvel@...aro.org>,
        Eric Biggers <ebiggers@...nel.org>,
        Iuliana Prodan <iuliana.prodan@....com>,
        "David S. Miller" <davem@...emloft.net>,
        Sascha Hauer <s.hauer@...gutronix.de>,
        "linux-crypto@...r.kernel.org" <linux-crypto@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        dl-linux-imx <linux-imx@....com>
Subject: Re: [PATCH] crypto: gcm - fix cacheline sharing

On Fri, May 31, 2019 at 06:10:51AM +0000, Horia Geanta wrote:
>
> Driver is not touching the DMA mapped areas, the DMA API conventions are
> carefully followed.
> It's touching a virtual pointer that is not DMA mapped, that just happens to be
> on the same cache line with a DMA mapped buffer.

Well you can't control what the users give you so you must assume
that the virtual address always share a cacheline with the DMA
buffer.

That's why you must only operate on that virtual address either
before you DMA map or after you DMA unmap.  Virtual addresses
that you allocate yourself (including ones on the stack) are
obviously not subject to this restriction.

Cheers,
-- 
Email: Herbert Xu <herbert@...dor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt