[PATCH] crypto: aesni - fix failing setkey for rfc4106-gcm-aesni

[PATCH] crypto: aesni - fix failing setkey for rfc4106-gcm-aesni

[Tadeusz Struk]

rfc4106(gcm(aes)) uses cbc(aes) to generate hash key. cbc(aes) needs
chainiv, but the chainiv gets initialized after aesni_intel when both
are statically linked so the setkey fails.
This patch forces aesni_intel to be initialized after chainiv.

Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com>
---
 arch/x86/crypto/aesni-intel_glue.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/crypto/aesni-intel_glue.c b/arch/x86/crypto/aesni-intel_glue.c
index ebcb981d..cb630a8 100644
--- a/arch/x86/crypto/aesni-intel_glue.c
+++ b/arch/x86/crypto/aesni-intel_glue.c
@@ -1537,7 +1537,7 @@ static void __exit aesni_exit(void)
 	crypto_fpu_exit();
 }
 
-module_init(aesni_init);
+late_initcall(aesni_init);
 module_exit(aesni_exit);
 
 MODULE_DESCRIPTION("Rijndael (AES) Cipher Algorithm, Intel AES-NI instructions optimized");

Re: [PATCH] crypto: aesni - fix failing setkey for rfc4106-gcm-aesni

[Herbert Xu]

On Wed, Jun 24, 2015 at 07:14:21AM -0700, Tadeusz Struk wrote:
> rfc4106(gcm(aes)) uses cbc(aes) to generate hash key. cbc(aes) needs
> chainiv, but the chainiv gets initialized after aesni_intel when both
> are statically linked so the setkey fails.
> This patch forces aesni_intel to be initialized after chainiv.
> 
> Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com>

Aha, this could indeed the explain the setkey error that Linus
saw.  Once the AEAD conversion is complete this would actually
become unnecessary because seqiv for blkcipher would disappear.

Linus, could you confirm that you have AESNI built into the kernel
and not as a module?

However, this is still brittle because you have the same ordering
issue with ctr.  IOW aesni may be registered before ctr.  In fact
you don't actually need ctr here.  You could just replace it with
plain aes plus a xor.

That should be more robust as you can then just use aesni for the
aes and you wouldn't depend on anything external to aesni.

Could you make a patch for that Tadeusz?

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

Re: [PATCH] crypto: aesni - fix failing setkey for rfc4106-gcm-aesni

[Tadeusz Struk]

On 06/25/2015 07:25 AM, Herbert Xu wrote:
> On Wed, Jun 24, 2015 at 07:14:21AM -0700, Tadeusz Struk wrote:
>> rfc4106(gcm(aes)) uses cbc(aes) to generate hash key. cbc(aes) needs
>> chainiv, but the chainiv gets initialized after aesni_intel when both
>> are statically linked so the setkey fails.
>> This patch forces aesni_intel to be initialized after chainiv.
>>
>> Signed-off-by: Tadeusz Struk <tadeusz.struk@intel.com>
> 
> Aha, this could indeed the explain the setkey error that Linus
> saw.  Once the AEAD conversion is complete this would actually
> become unnecessary because seqiv for blkcipher would disappear.

Yes, after the conversion is finished we can revert this patch.

> 
> Linus, could you confirm that you have AESNI built into the kernel
> and not as a module?
> 
> However, this is still brittle because you have the same ordering
> issue with ctr.  IOW aesni may be registered before ctr.  In fact
> you don't actually need ctr here.  You could just replace it with
> plain aes plus a xor.

So this will fix it for ctr too.

> 
> That should be more robust as you can then just use aesni for the
> aes and you wouldn't depend on anything external to aesni.
> 
> Could you make a patch for that Tadeusz?

I can start working on this after my vacation

Re: [PATCH] crypto: aesni - fix failing setkey for rfc4106-gcm-aesni

[Linus Torvalds]

On Thu, Jun 25, 2015 at 7:25 AM, Herbert Xu <herbert@gondor.apana.org.au> wrote:
>
> Linus, could you confirm that you have AESNI built into the kernel
> and not as a module?

Yes, confirmed.

                 Linus