2009-08-19 07:06:15

by Steffen Klassert

Subject: [PATCH] crypto: xcbc - Fix alignment calculation of xcbc_tfm_ctx

The alignment calculation of xcbc_tfm_ctx uses alg->cra_alignmask
and not alg->cra_alignmask + 1 as it should. This led to frequent
crashes during the selftest of xcbc(aes-asm) on x86_64
machines. This patch fixes this. Also we use the alignmask
of xcbc and not the alignmask of the underlying algorithm
for the alignmnent calculation in xcbc_create now.

Signed-off-by: Steffen Klassert <[email protected]>
---
crypto/xcbc.c | 8 +++++---
1 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/crypto/xcbc.c b/crypto/xcbc.c
index 1e30b31..bb7b67f 100644
--- a/crypto/xcbc.c
+++ b/crypto/xcbc.c
@@ -199,6 +199,7 @@ static int xcbc_create(struct crypto_template *tmpl, struct rtattr **tb)
{
struct shash_instance *inst;
struct crypto_alg *alg;
+ unsigned long alignmask;
int err;

err = crypto_check_attr_type(tb, CRYPTO_ALG_TYPE_SHASH);
@@ -228,19 +229,20 @@ static int xcbc_create(struct crypto_template *tmpl, struct rtattr **tb)
if (err)
goto out_free_inst;

+ alignmask = alg->cra_alignmask | 3;
+ inst->alg.base.cra_alignmask = alignmask;
inst->alg.base.cra_priority = alg->cra_priority;
inst->alg.base.cra_blocksize = alg->cra_blocksize;
- inst->alg.base.cra_alignmask = alg->cra_alignmask | 3;

inst->alg.digestsize = alg->cra_blocksize;
inst->alg.descsize = ALIGN(sizeof(struct xcbc_desc_ctx),
crypto_tfm_ctx_alignment()) +
- (alg->cra_alignmask &
+ (alignmask &
~(crypto_tfm_ctx_alignment() - 1)) +
alg->cra_blocksize * 2;

inst->alg.base.cra_ctxsize = ALIGN(sizeof(struct xcbc_tfm_ctx),
- alg->cra_alignmask) +
+ alignmask + 1) +
alg->cra_blocksize * 2;
inst->alg.base.cra_init = xcbc_init_tfm;
inst->alg.base.cra_exit = xcbc_exit_tfm;
--
1.5.4.2

2009-08-20 07:59:08

by Herbert Xu

Subject: Re: [PATCH] crypto: xcbc - Fix alignment calculation of xcbc_tfm_ctx

On Wed, Aug 19, 2009 at 09:09:02AM +0200, Steffen Klassert wrote:
> The alignment calculation of xcbc_tfm_ctx uses alg->cra_alignmask
> and not alg->cra_alignmask + 1 as it should. This led to frequent
> crashes during the selftest of xcbc(aes-asm) on x86_64
> machines. This patch fixes this. Also we use the alignmask
> of xcbc and not the alignmask of the underlying algorithm
> for the alignmnent calculation in xcbc_create now.
>
> Signed-off-by: Steffen Klassert <[email protected]>

Good catch!

I wonder why I didn't see it though as I too was testing this on
x86-64. Must've been luck I suppose as it depends on what follows
the context in memory. Anyway, patch applied to cryptodev-2.6.

Thanks,
--
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt