Hello,
In AEAD mode (or in any case, in IPSec ESP IPv4 – esp4.c), in kernel
versions prior to 4.2 the cryptographic driver is expected to generate
an IV.
What if my driver is unable to generate an IV?
Thanks,
Dennis.
Inline comments.
________________________________________
From: [email protected] <[email protected]> on behalf of Denis B <[email protected]>
Sent: Wednesday, May 18, 2016 3:06 PM
To: [email protected]
Subject: IV generation in cryptographic driver in AEAD
Hello,
In AEAD mode (or in any case, in IPSec ESP IPv4 ? esp4.c), in kernel
versions prior to 4.2 the cryptographic driver is expected to generate
an IV.
What if my driver is unable to generate an IV?
[Catalin Vasile]
Simple: You do not implement the givcrypt() primitive. The kernel will generate the IV in software and then call your encrypt() primitive.
Thanks,
Dennis.
Forgive my dumbness, but in:
.cra_type = &crypto_aead_type,
.cra_u = {
.aead = {
.setkey = pp_crypto_aead_setkey,
.setauthsize = pp_crypto_aead_setauthsize,
.decrypt = pp_crypto_aead_dec,
.encrypt = pp_crypto_aead_enc,
.givencrypt = pp_crypto_aead_genivencrypt,
.givdecrypt = pp_crypto_aead_genivdecrypt,
.ivsize = AES_BLOCK_SIZE,
.maxauthsize = SHA256_DIGEST_SIZE,
},
},
Just delete the givencrypt and givdecrypt lines (or set to NULL)?
On Wed, May 18, 2016 at 4:56 PM, Catalin Vasile <[email protected]> wrote:
> Inline comments.
>
> ________________________________________
> From: [email protected] <[email protected]> on behalf of Denis B <[email protected]>
> Sent: Wednesday, May 18, 2016 3:06 PM
> To: [email protected]
> Subject: IV generation in cryptographic driver in AEAD
>
> Hello,
>
> In AEAD mode (or in any case, in IPSec ESP IPv4 – esp4.c), in kernel
> versions prior to 4.2 the cryptographic driver is expected to generate
> an IV.
>
> What if my driver is unable to generate an IV?
> [Catalin Vasile]
> Simple: You do not implement the givcrypt() primitive. The kernel will generate the IV in software and then call your encrypt() primitive.
>
> Thanks,
> Dennis.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-crypto" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
Am Mittwoch, 18. Mai 2016, 15:06:19 schrieb Denis B:
Hi Denis,
> Hello,
>
> In AEAD mode (or in any case, in IPSec ESP IPv4 – esp4.c), in kernel
> versions prior to 4.2 the cryptographic driver is expected to generate
> an IV.
The driver is not expected to generate the IV. Please see the ASCII art at
[1]. The IV generation is done with the seqiv component. So, you driver can
implement all of the logic of GCM, but pull the IV generation from the C
implementation provided by the kernel crypto API.
[1] http://www.chronox.de/crypto-API/ch02s07.html
>
> What if my driver is unable to generate an IV?
>
> Thanks,
> Dennis.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-crypto" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
Ciao
Stephan
Denis B <[email protected]> wrote:
> Forgive my dumbness, but in:
>
> .cra_type = &crypto_aead_type,
> .cra_u = {
> .aead = {
> .setkey = pp_crypto_aead_setkey,
> .setauthsize = pp_crypto_aead_setauthsize,
> .decrypt = pp_crypto_aead_dec,
> .encrypt = pp_crypto_aead_enc,
> .givencrypt = pp_crypto_aead_genivencrypt,
> .givdecrypt = pp_crypto_aead_genivdecrypt,
> .ivsize = AES_BLOCK_SIZE,
> .maxauthsize = SHA256_DIGEST_SIZE,
> },
> },
>
> Just delete the givencrypt and givdecrypt lines (or set to NULL)?
You also need to set your cra_type to nivaead.
Cheers,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
Inline comments
________________________________________
From: Denis B <[email protected]>
Sent: Wednesday, May 18, 2016 5:21 PM
To: Catalin Vasile
Cc: [email protected]
Subject: Re: IV generation in cryptographic driver in AEAD
Forgive my dumbness, but in:
.cra_type = &crypto_aead_type,
.cra_u = {
.aead = {
.setkey = pp_crypto_aead_setkey,
.setauthsize = pp_crypto_aead_setauthsize,
.decrypt = pp_crypto_aead_dec,
.encrypt = pp_crypto_aead_enc,
.givencrypt = pp_crypto_aead_genivencrypt,
.givdecrypt = pp_crypto_aead_genivdecrypt,
.ivsize = AES_BLOCK_SIZE,
.maxauthsize = SHA256_DIGEST_SIZE,
},
},
Just delete the givencrypt and givdecrypt lines (or set to NULL)?
[Catalin Vasile]
You need to set it to NULL. If your struct is declared globally, the parts
that load your code will be set any uninitialized struct elements to NULL.
On Wed, May 18, 2016 at 4:56 PM, Catalin Vasile <[email protected]> wrote:
> Inline comments.
>
> ________________________________________
> From: [email protected] <[email protected]> on behalf of Denis B <[email protected]>
> Sent: Wednesday, May 18, 2016 3:06 PM
> To: [email protected]
> Subject: IV generation in cryptographic driver in AEAD
>
> Hello,
>
> In AEAD mode (or in any case, in IPSec ESP IPv4 ? esp4.c), in kernel
> versions prior to 4.2 the cryptographic driver is expected to generate
> an IV.
>
> What if my driver is unable to generate an IV?
> [Catalin Vasile]
> Simple: You do not implement the givcrypt() primitive. The kernel will generate the IV in software and then call your encrypt() primitive.
>
> Thanks,
> Dennis.
> --
> To unsubscribe from this list: send the line "unsubscribe linux-crypto" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
My algs struct now looks like this:
static struct crypto_alg pp_crypto_algs[] = {
{
.cra_name = "authenc(hmac(sha256),cbc(aes))",
.cra_driver_name = "pp_crypto_cbc_hmac_sha256",
.cra_priority = 1, /**TODO set to highest values after
implementing encrypt decrypt functions */
.cra_flags = CRYPTO_ALG_TYPE_AEAD | CRYPTO_ALG_ASYNC,
.cra_blocksize = AES_BLOCK_SIZE,
.cra_ctxsize = sizeof(struct pp_crypto_aead_ctx),
.cra_type = &crypto_nivaead_type,
.cra_u = {
.aead = {
.setkey = pp_crypto_aead_setkey,
.setauthsize = pp_crypto_aead_setauthsize,
.decrypt = pp_crypto_aead_dec,
.encrypt = pp_crypto_aead_enc,
.givencrypt = NULL,
.givdecrypt = NULL,
.ivsize = AES_BLOCK_SIZE,
.maxauthsize = SHA256_DIGEST_SIZE,
},
},
.cra_init = pp_crypto_aead_sha256_init,
.cra_exit = pp_crypto_aead_exit,
.cra_module = THIS_MODULE,
},
{
.cra_name = "authenc(hmac(sha512),cbc(aes))",
.cra_driver_name = "pp_crypto_cbc_hmac_sha512",
.cra_priority = 1, /**TODO set to highest values after
implementing encrypt decrypt functions */
.cra_flags = CRYPTO_ALG_TYPE_AEAD | CRYPTO_ALG_ASYNC,
.cra_blocksize = AES_BLOCK_SIZE,
.cra_ctxsize = sizeof(struct pp_crypto_aead_ctx),
.cra_type = &crypto_nivaead_type,
.cra_u = {
.aead = {
.setkey = pp_crypto_aead_setkey,
.decrypt = pp_crypto_aead_dec,
.encrypt = pp_crypto_aead_enc,
.givencrypt = NULL,
.givdecrypt = NULL,
.ivsize = AES_BLOCK_SIZE,
.maxauthsize = SHA512_DIGEST_SIZE,
},
},
.cra_init = pp_crypto_aead_sha512_init,
.cra_exit = pp_crypto_aead_exit,
.cra_module = THIS_MODULE,
}
};
I issue a ping and see esp_output() getting called. After that, I'm
used to seeing pp_crypto_givaead_enc() in my driver called. Now I see:
ping: sendto: Function not implemented
Please note that I am working with kernel 3.12, where in esp4.c line
266 we have:
err = crypto_aead_givencrypt(req);
On Thu, May 19, 2016 at 9:44 AM, Catalin Vasile <[email protected]> wrote:
> Inline comments
>
> ________________________________________
> From: Denis B <[email protected]>
> Sent: Wednesday, May 18, 2016 5:21 PM
> To: Catalin Vasile
> Cc: [email protected]
> Subject: Re: IV generation in cryptographic driver in AEAD
>
> Forgive my dumbness, but in:
>
> .cra_type = &crypto_aead_type,
> .cra_u = {
> .aead = {
> .setkey = pp_crypto_aead_setkey,
> .setauthsize = pp_crypto_aead_setauthsize,
> .decrypt = pp_crypto_aead_dec,
> .encrypt = pp_crypto_aead_enc,
> .givencrypt = pp_crypto_aead_genivencrypt,
> .givdecrypt = pp_crypto_aead_genivdecrypt,
> .ivsize = AES_BLOCK_SIZE,
> .maxauthsize = SHA256_DIGEST_SIZE,
> },
> },
>
> Just delete the givencrypt and givdecrypt lines (or set to NULL)?
> [Catalin Vasile]
> You need to set it to NULL. If your struct is declared globally, the parts
> that load your code will be set any uninitialized struct elements to NULL.
>
> On Wed, May 18, 2016 at 4:56 PM, Catalin Vasile <[email protected]> wrote:
>> Inline comments.
>>
>> ________________________________________
>> From: [email protected] <[email protected]> on behalf of Denis B <[email protected]>
>> Sent: Wednesday, May 18, 2016 3:06 PM
>> To: [email protected]
>> Subject: IV generation in cryptographic driver in AEAD
>>
>> Hello,
>>
>> In AEAD mode (or in any case, in IPSec ESP IPv4 – esp4.c), in kernel
>> versions prior to 4.2 the cryptographic driver is expected to generate
>> an IV.
>>
>> What if my driver is unable to generate an IV?
>> [Catalin Vasile]
>> Simple: You do not implement the givcrypt() primitive. The kernel will generate the IV in software and then call your encrypt() primitive.
>>
>> Thanks,
>> Dennis.
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-crypto" in
>> the body of a message to [email protected]
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
Denis B <[email protected]> wrote:
> My algs struct now looks like this:
>
> static struct crypto_alg pp_crypto_algs[] = {
> {
> .cra_name = "authenc(hmac(sha256),cbc(aes))",
> .cra_driver_name = "pp_crypto_cbc_hmac_sha256",
> .cra_priority = 1, /**TODO set to highest values after
> implementing encrypt decrypt functions */
> .cra_flags = CRYPTO_ALG_TYPE_AEAD | CRYPTO_ALG_ASYNC,
> .cra_blocksize = AES_BLOCK_SIZE,
> .cra_ctxsize = sizeof(struct pp_crypto_aead_ctx),
> .cra_type = &crypto_nivaead_type,
> .cra_u = {
> .aead = {
> .setkey = pp_crypto_aead_setkey,
> .setauthsize = pp_crypto_aead_setauthsize,
> .decrypt = pp_crypto_aead_dec,
> .encrypt = pp_crypto_aead_enc,
> .givencrypt = NULL,
> .givdecrypt = NULL,
> .ivsize = AES_BLOCK_SIZE,
> .maxauthsize = SHA256_DIGEST_SIZE,
You also need to set geniv to "eseqiv".
Cheers,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
On 05/19/2016 11:19 PM, Herbert Xu wrote:
> Denis B <[email protected]> wrote:
>> My algs struct now looks like this:
>>
>> static struct crypto_alg pp_crypto_algs[] = {
>> {
>> .cra_name = "authenc(hmac(sha256),cbc(aes))",
>> .cra_driver_name = "pp_crypto_cbc_hmac_sha256",
>> .cra_priority = 1, /**TODO set to highest values after
>> implementing encrypt decrypt functions */
>> .cra_flags = CRYPTO_ALG_TYPE_AEAD | CRYPTO_ALG_ASYNC,
>> .cra_blocksize = AES_BLOCK_SIZE,
>> .cra_ctxsize = sizeof(struct pp_crypto_aead_ctx),
>> .cra_type = &crypto_nivaead_type,
>> .cra_u = {
>> .aead = {
>> .setkey = pp_crypto_aead_setkey,
>> .setauthsize = pp_crypto_aead_setauthsize,
>> .decrypt = pp_crypto_aead_dec,
>> .encrypt = pp_crypto_aead_enc,
>> .givencrypt = NULL,
>> .givdecrypt = NULL,
>> .ivsize = AES_BLOCK_SIZE,
>> .maxauthsize = SHA256_DIGEST_SIZE,
>
> You also need to set geniv to "eseqiv".
I just gotta ask. Forgive my ignorance.
Why is (or should) setting geniv (be) required?
crypto_givcipher_default() appears to call crypto_default_geniv() if the
geniv member
is NULL. That function returns "eseqiv" or "chainiv" (under certain
conditions). If an
implementation isn't generating its own IVs, shouldn't the default
happen anyway? Or is
this more a matter of populating the structure with known, intentional
values?
Thank you for any illumination provided.
Gary
On Fri, May 20, 2016 at 10:50:38AM -0500, Gary R Hook wrote:
>
> Why is (or should) setting geniv (be) required?
>
> crypto_givcipher_default() appears to call crypto_default_geniv() if
> the geniv member
> is NULL. That function returns "eseqiv" or "chainiv" (under certain
> conditions). If an
> implementation isn't generating its own IVs, shouldn't the default
> happen anyway? Or is
> this more a matter of populating the structure with known,
> intentional values?
>
> Thank you for any illumination provided.
In the upstream kernel AEAD geniv has been completely phased out
and no longer exists. Denis is working on an old kernel that still
has it.
We haven't yet phased it out for skcipher but I'm working on it.
Cheers,
--
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt
On 05/20/2016 06:31 PM, Herbert Xu wrote:
> On Fri, May 20, 2016 at 10:50:38AM -0500, Gary R Hook wrote:
>>
>> Why is (or should) setting geniv (be) required?
>>
>> crypto_givcipher_default() appears to call crypto_default_geniv() if
>> the geniv member
>> is NULL. That function returns "eseqiv" or "chainiv" (under certain
>> conditions). If an
>> implementation isn't generating its own IVs, shouldn't the default
>> happen anyway? Or is
>> this more a matter of populating the structure with known,
>> intentional values?
>>
>> Thank you for any illumination provided.
>
> In the upstream kernel AEAD geniv has been completely phased out
> and no longer exists. Denis is working on an old kernel that still
> has it.
>
> We haven't yet phased it out for skcipher but I'm working on it.
...and there was light.
Thank you; very helpful.
(I'll work harder on my line wraps... If someone knows how to get
Thunderbird
to do it for me, I'd love to know.)