On Wed, Apr 02, 2008 at 03:22:58PM +0800, Herbert Xu wrote: > On Wed, Apr 02, 2008 at 09:17:22AM +0200, Sebastian Siewior wrote: > > I'm not sure if they have SMP or multicore out / plans but I gave a > > little thought on this early morning: The only problem is if we get > > swapped to another cores before encryption process beginns and after key > > reset thing. If the scheduler decides to swap cores than EFLAGS should > > be reloaded and wer are fine. The only problem that I see with cpus > 1 > > is that we might reload the key if it is not required. This could be > > solved with old_ctx pointer beeing a percpu. > > Yeah I think having a percpu would be good even if it isn't needed > right now. I've coded this up with the per-cpu structure and it seems to work. It doesn't make much of a difference for ECB/CBC but with LRW/XTS it gives a huge boost. Cheers, -- Visit Openswan at http://www.openswan.org/ Email: Herbert Xu ~{PmV>HI~} <herbert@xxxxxxxxxxxxxxxxxxx> Home Page: http://gondor.apana.org.au/~herbert/ PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt -- diff --git a/drivers/crypto/padlock-aes.c b/drivers/crypto/padlock-aes.c index 54a2a16..c71256a 100644 --- a/drivers/crypto/padlock-aes.c +++ b/drivers/crypto/padlock-aes.c @@ -15,6 +15,8 @@ #include <linux/errno.h> #include <linux/interrupt.h> #include <linux/kernel.h> +#include <linux/percpu.h> +#include <linux/smp.h> #include <asm/byteorder.h> #include "padlock.h" @@ -48,6 +50,8 @@ struct aes_ctx { u32 *D; }; +static DEFINE_PER_CPU(struct cword *, last_cword); + /* Tells whether the ACE is capable to generate the extended key for a given key_len. */ static inline int @@ -88,6 +92,7 @@ static int aes_set_key(struct crypto_tfm *tfm, const u8 *in_key, const __le32 *key = (const __le32 *)in_key; u32 *flags = &tfm->crt_flags; struct crypto_aes_ctx gen_aes; + int cpu; if (key_len % 8) { *flags |= CRYPTO_TFM_RES_BAD_KEY_LEN; @@ -117,7 +122,7 @@ static int aes_set_key(struct crypto_tfm *tfm, const u8 *in_key, /* Don't generate extended keys if the hardware can do it. */ if (aes_hw_extkey_available(key_len)) - return 0; + goto ok; ctx->D = ctx->d_data; ctx->cword.encrypt.keygen = 1; @@ -130,19 +135,34 @@ static int aes_set_key(struct crypto_tfm *tfm, const u8 *in_key, memcpy(ctx->E, gen_aes.key_enc, AES_MAX_KEYLENGTH); memcpy(ctx->D, gen_aes.key_dec, AES_MAX_KEYLENGTH); + +ok: + for_each_online_cpu(cpu) + if (&ctx->cword.encrypt == per_cpu(last_cword, cpu) || + &ctx->cword.decrypt == per_cpu(last_cword, cpu)) + per_cpu(last_cword, cpu) = NULL; + return 0; } /* ====== Encryption/decryption routines ====== */ /* These are the real call to PadLock. */ -static inline void padlock_reset_key(void) +static inline void padlock_reset_key(struct cword *cword) { - asm volatile ("pushfl; popfl"); + int cpu = raw_smp_processor_id(); + + if (cword != per_cpu(last_cword, cpu)) + asm volatile ("pushfl; popfl"); +} + +static inline void padlock_store_cword(struct cword *cword) +{ + per_cpu(last_cword, raw_smp_processor_id()) = cword; } static inline void padlock_xcrypt(const u8 *input, u8 *output, void *key, - void *control_word) + struct cword *control_word) { asm volatile (".byte 0xf3,0x0f,0xa7,0xc8" /* rep xcryptecb */ : "+S"(input), "+D"(output) @@ -205,15 +225,17 @@ static inline u8 *padlock_xcrypt_cbc(const u8 *input, u8 *output, void *key, static void aes_encrypt(struct crypto_tfm *tfm, u8 *out, const u8 *in) { struct aes_ctx *ctx = aes_ctx(tfm); - padlock_reset_key(); + padlock_reset_key(&ctx->cword.encrypt); aes_crypt(in, out, ctx->E, &ctx->cword.encrypt); + padlock_store_cword(&ctx->cword.encrypt); } static void aes_decrypt(struct crypto_tfm *tfm, u8 *out, const u8 *in) { struct aes_ctx *ctx = aes_ctx(tfm); - padlock_reset_key(); + padlock_reset_key(&ctx->cword.decrypt); aes_crypt(in, out, ctx->D, &ctx->cword.decrypt); + padlock_store_cword(&ctx->cword.decrypt); } static struct crypto_alg aes_alg = { @@ -245,7 +267,7 @@ static int ecb_aes_encrypt(struct blkcipher_desc *desc, struct blkcipher_walk walk; int err; - padlock_reset_key(); + padlock_reset_key(&ctx->cword.encrypt); blkcipher_walk_init(&walk, dst, src, nbytes); err = blkcipher_walk_virt(desc, &walk); @@ -258,6 +280,8 @@ static int ecb_aes_encrypt(struct blkcipher_desc *desc, err = blkcipher_walk_done(desc, &walk, nbytes); } + padlock_store_cword(&ctx->cword.encrypt); + return err; } @@ -269,7 +293,7 @@ static int ecb_aes_decrypt(struct blkcipher_desc *desc, struct blkcipher_walk walk; int err; - padlock_reset_key(); + padlock_reset_key(&ctx->cword.decrypt); blkcipher_walk_init(&walk, dst, src, nbytes); err = blkcipher_walk_virt(desc, &walk); @@ -282,6 +306,8 @@ static int ecb_aes_decrypt(struct blkcipher_desc *desc, err = blkcipher_walk_done(desc, &walk, nbytes); } + padlock_store_cword(&ctx->cword.decrypt); + return err; } @@ -315,7 +341,7 @@ static int cbc_aes_encrypt(struct blkcipher_desc *desc, struct blkcipher_walk walk; int err; - padlock_reset_key(); + padlock_reset_key(&ctx->cword.encrypt); blkcipher_walk_init(&walk, dst, src, nbytes); err = blkcipher_walk_virt(desc, &walk); @@ -330,6 +356,8 @@ static int cbc_aes_encrypt(struct blkcipher_desc *desc, err = blkcipher_walk_done(desc, &walk, nbytes); } + padlock_store_cword(&ctx->cword.encrypt); + return err; } @@ -341,7 +369,7 @@ static int cbc_aes_decrypt(struct blkcipher_desc *desc, struct blkcipher_walk walk; int err; - padlock_reset_key(); + padlock_reset_key(&ctx->cword.decrypt); blkcipher_walk_init(&walk, dst, src, nbytes); err = blkcipher_walk_virt(desc, &walk); @@ -354,6 +382,8 @@ static int cbc_aes_decrypt(struct blkcipher_desc *desc, err = blkcipher_walk_done(desc, &walk, nbytes); } + padlock_store_cword(&ctx->cword.decrypt); + return err; } -- To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
