Received-SPF: pass (google.com: best guess record for domain of linux-crypto-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
From:   Christophe Leroy <christophe.leroy@c-s.fr>
Subject: Can scatterlist elements cross page boundary ?
To:     Eric Biggers <ebiggers@google.com>,
        Herbert Xu <herbert@gondor.apana.org.au>
Cc:     "linux-crypto@vger.kernel.org" <linux-crypto@vger.kernel.org>
Message-ID: <798a42c9-bcda-6612-088c-cb90c35a578f@c-s.fr>
Date:   Wed, 15 May 2019 13:58:13 +0200
User-Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:60.0) Gecko/20100101
 Thunderbird/60.6.1
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: fr
Content-Transfer-Encoding: 7bit
Sender: linux-crypto-owner@vger.kernel.org
Precedence: bulk

Hi,

I had understood that scatterlists where lists of buffers, each buffer 
being enclosed within a single memory PAGE. Isn't that right ?

As far as I understand, dma_map_sg() expects each entry to be DMA mapable.

But at the time being, I'm getting the following errors on the Talitos 
algorithmes:

[    2.382845] alg: aead: authenc-hmac-sha1-cbc-aes-talitos encryption 
test failed (wrong result) on test vector 0, cfg="misaligned splits 
crossing pages, inplace"

[    2.430178] alg: aead: authenc-hmac-sha1-cbc-aes-talitos encryption 
test failed (wrong result) on test vector 1, cfg="misaligned splits 
crossing pages, inplace"

[    2.509270] alg: aead: authenc-hmac-sha256-cbc-aes-talitos encryption 
test failed (wrong result) on test vector 0, cfg="uneven misaligned 
splits, may sleep"

When comparing the expected and actual results, I see (respectively for 
the 3 fails above tests) :

[    2.362271] 00000000: e3 53 77 9c 10 79 ae b8 27 08 94 2d be 77 18 1a
[    2.368529] 00000010: 1b 13 cb af 89 5e e1
[    2.372429] 00000000: e3 53 77 9c 10 79 ae b8 27 08 94 2d be 77 18 1a
[    2.378924] 00000010: fe fe fe fe fe fe fe

[    2.398908] 00000000: d2 96 cd 94 c2 cc cf 8a 3a 86 30 28 b5 e1 dc 0a
[    2.405185] 00000010: 75 86 60 2d 25 3c ff f9 1b 82 66 be a6 d6 1a b1
[    2.411407] 00000020: ad 9b 4c
[    2.414427] 00000000: d2 96 cd 94 c2 cc cf 8a 3a 86 30 28 b5 e1 dc 0a
[    2.420926] 00000010: 75 86 60 2d 25 3c ff f9 1b 82 66 be a6 d6 1a b1
[    2.427287] 00000020: fe fe fe

[    2.491701] 00000000: e3 53 77 9c 10 79 ae b8 27 08 94 2d be 77 18 1a
[    2.498125] 00000010: cc
[    2.500403] 00000000: e3 53 77 9c 10 79 ae b8 27 08 94 2d be 77 18 1a
[    2.507012] 00000010: fe


Looking at the test manager, I understand that it builds scatterlists 
with buffers that are using 2 pages. Am I correct ?

Then how do we expect the driver to behave ?

Thanks
Christophe