linux-kernel - Re: [PATCH v2 1/1] mm: zswap - Add crypto acomp/scomp framework support

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:   Thu, 16 Mar 2017 11:54:43 -0400
From:   Dan Streetman <ddstreet@...e.org>
To:     Herbert Xu <herbert@...dor.apana.org.au>
Cc:     Mahipal Reddy <mahipalreddy2006@...il.com>,
        Mahipal Challa <Mahipal.Challa@...ium.com>,
        Seth Jennings <sjenning@...hat.com>,
        Linux-MM <linux-mm@...ck.org>,
        linux-kernel <linux-kernel@...r.kernel.org>, pathreya@...ium.com,
        Vishnu Nair <Vishnu.Nair@...ium.com>
Subject: Re: [PATCH v2 1/1] mm: zswap - Add crypto acomp/scomp framework support

On Thu, Mar 9, 2017 at 4:39 AM, Herbert Xu <herbert@...dor.apana.org.au> wrote:
> On Wed, Mar 08, 2017 at 12:38:40PM -0500, Dan Streetman wrote:
>>
>> It looks like the crypto_scomp interface is buried under
>> include/crypto/internal/scompress.h, however that's exactly what zswap
>> should be using.  We don't need to switch to an asynchronous interface
>> that's rather significantly more complicated, and then use it in a
>> synchronous way.  The crypto_scomp interface should probably be made
>> public, not an implementation internal.
>
> No scomp is not meant to be used externally.  We provide exactly
> one compression interface and it's acomp.  acomp can be used
> synchronously by setting the CRYPTO_ALG_ASYNC bit in the mask
> field when allocating the algorithm.

setting the ASYNC bit makes it synchronous?  that seems backwards...?

Anyway, I have a few concerns about moving over to using that first,
specifically:

- no docs on acomp in Documentation/crypto/ that I can see
- no place in the crypto code that I can see that parses ALG_ASYNC to
make the crypto_acomp_compress() call synchronous
- no synchronous test in crypto/testmgr.c

Maybe I'm reading the code wrong, but it looks like any compression
backend that is actually scomp, actually does the (de)compression
synchronously.  In crypto_acomp_init_tfm(), if the tfm is not
crypto_acomp_type (and I assume because all the current
implementations register as scomp, they aren't acomp_type) it calls
crypto_init_scomp_ops_async(), which then sets ->compress to
scomp_acomp_compress() and that function appears to directly call the
scomp compression function.  This is just after a very quick look, so
maybe I'm reading it wrong.  I'll look some more, and also add a
synchronous testmgr test so i can understand how it works better.

Is the acomp interface fully ready for use?

>
> The existing compression interface will be phased out.
>
> 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