crypto: padlock-aes: Use the correct mask when checking whether copying is required
Masking with PAGE_SIZE is just wrong... and there's no need for the braces and return
statement either.
Signed-off-by: Chuck Ebbert <[email protected]>
---
Can we get this in 2.6.32 and -stable? It fixes a fatal oops in the driver.
--- a/drivers/crypto/padlock-aes.c
+++ b/drivers/crypto/padlock-aes.c
@@ -236,19 +236,17 @@ static inline void ecb_crypt(const u8 *i
/* Padlock in ECB mode fetches at least ecb_fetch_bytes of data.
* We could avoid some copying here but it's probably not worth it.
*/
- if (unlikely(((unsigned long)in & PAGE_SIZE) + ecb_fetch_bytes > PAGE_SIZE)) {
+ if (unlikely(((unsigned long)in & ~PAGE_MASK) + ecb_fetch_bytes > PAGE_SIZE))
ecb_crypt_copy(in, out, key, cword, count);
- return;
- }
On Tue, Oct 27, 2009 at 07:25:33PM -0400, Chuck Ebbert wrote:
> crypto: padlock-aes: Use the correct mask when checking whether copying is required
>
> Masking with PAGE_SIZE is just wrong... and there's no need for the braces and return
Thanks for the patch, that does indeed look wrong.
> statement either.
But can we please keep unrelated stylistic changes in a separate
patch? In this case I'm not even sure that it is better than the
original.
Cheers,
--
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
crypto: padlock-aes: Use the correct mask when checking whether copying is required
Masking with PAGE_SIZE is just wrong...
Signed-off-by: Chuck Ebbert <[email protected]>
--- vanilla-2.6.31.orig/drivers/crypto/padlock-aes.c
+++ vanilla-2.6.31/drivers/crypto/padlock-aes.c
@@ -236,7 +236,7 @@ static inline void ecb_crypt(const u8 *i
/* Padlock in ECB mode fetches at least ecb_fetch_bytes of data.
* We could avoid some copying here but it's probably not worth it.
*/
- if (unlikely(((unsigned long)in & PAGE_SIZE) + ecb_fetch_bytes > PAGE_SIZE)) {
+ if (unlikely(((unsigned long)in & ~PAGE_MASK) + ecb_fetch_bytes > PAGE_SIZE)) {
ecb_crypt_copy(in, out, key, cword, count);
return;
}
@@ -248,7 +248,7 @@ static inline u8 *cbc_crypt(const u8 *in
u8 *iv, struct cword *cword, int count)
{
/* Padlock in CBC mode fetches at least cbc_fetch_bytes of data. */
- if (unlikely(((unsigned long)in & PAGE_SIZE) + cbc_fetch_bytes > PAGE_SIZE))
+ if (unlikely(((unsigned long)in & ~PAGE_MASK) + cbc_fetch_bytes > PAGE_SIZE))
return cbc_crypt_copy(in, out, key, iv, cword, count);
return rep_xcrypt_cbc(in, out, key, iv, cword, count);
On Mon, Nov 02, 2009 at 05:26:50PM -0500, Chuck Ebbert wrote:
> crypto: padlock-aes: Use the correct mask when checking whether copying is required
>
> Masking with PAGE_SIZE is just wrong...
>
> Signed-off-by: Chuck Ebbert <[email protected]>
Applied to crypto-2.6. Thanks Chuck!
--
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