Received-SPF: pass (google.com: domain of linux-crypto-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Fri, 3 Jun 2022 00:26:44 -0400
From:   Kent Overstreet <kent.overstreet@gmail.com>
To:     Eric Biggers <ebiggers@kernel.org>
Cc:     herbert@gondor.apana.org.au, linux-crypto@vger.kernel.org
Subject: Re: Memory corruption related to skcipher code
Message-ID: <20220603042644.jj7wcvp2lzox25vs@moria.home.lan>
References: <20220603021134.ehlnluenrz3adpxc@moria.home.lan>
 <YplxLIfUjFUgJA0Z@sol.localdomain>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <YplxLIfUjFUgJA0Z@sol.localdomain>
Precedence: bulk

On Thu, Jun 02, 2022 at 07:25:48PM -0700, Eric Biggers wrote:
> What if the address is in the vmalloc area and spans a page boundary?  If the
> next page isn't physically contiguous, this will write to some random page.  A
> multi-element scatterlist would be needed to describe all the physical pages.

That's exactly it :) Can't believe I missed that, or that that bug was lurking
for so long.

> Note that in v5.5 and later, ChaCha20 support is available via the library
> function chacha20_crypt(), in addition to the skcipher API.  You should consider
> using the library function, as it seems to match your use case.

I'll check that out - that looks like exactly what I was looking for when I
originally wrote this code.

Thanks, I owe you a beer :)