Re: [CRYPTO] is it really optimized ?

["Francis Moreau" <francis.moro@gmail.com>]

On 4/17/07, Herbert Xu <herbert@gondor.apana.org.au> wrote:
> On Mon, Apr 16, 2007 at 10:37:01AM +0200, Francis Moreau wrote:
> >
> > BTW, here are figures I got with 2 different versions of the driver
> > when using tcrypt module. The second being the result with the
> > optimized driver (no key reloading on each block):
> >
> > normal version:
> > test 4 (128 bit key, 8192 byte blocks): 1 operation in 67991 cycles (8192
> > bytes)
> >
> > optimized version:
> > test 4 (128 bit key, 8192 byte blocks): 1 operation in 51783 cycles (8192
> > bytes)
> >
> > So the gain is 16000 cycles which seems to worth the change, isn't it ?
>
> Sounds like it would.  It would help of course if you posted the patch :)
>

OK, I tried to cook up something very simple. Since I don't know this
code, please be indulgent when reading the following patch ;)

diff --git a/crypto/api.c b/crypto/api.c
index 55af8bb..f067de8 100644
--- a/crypto/api.c
+++ b/crypto/api.c
@@ -71,6 +71,10 @@ struct crypto_alg *__crypto_alg_lookup(const char
*name, u32 type, u32 mask)
 		    ((struct crypto_larval *)q)->mask != mask)
 			continue;

+		if (alg->cra_flags & CRYPTO_ALG_EXCLUSIVE &&
+		    atomic_read(&alg->cra_refcnt) > 0)
+			continue;
+
 		exact = !strcmp(q->cra_driver_name, name);
 		fuzzy = !strcmp(q->cra_name, name);
 		if (!exact && !(fuzzy && q->cra_priority > best))
diff --git a/include/linux/crypto.h b/include/linux/crypto.h
index 779aa78..278d386 100644
--- a/include/linux/crypto.h
+++ b/include/linux/crypto.h
@@ -41,6 +41,7 @@
 #define CRYPTO_ALG_DEAD			0x00000020
 #define CRYPTO_ALG_DYING		0x00000040
 #define CRYPTO_ALG_ASYNC		0x00000080
+#define CRYPTO_ALG_EXCLUSIVE		0x00000100

 /*
  * Set this bit if and only if the algorithm requires another algorithm of

-- 
Francis