Skip to contentSkip to conversation

crypto: salsa20 - cleanup and convert to skcipher API

PATCH In review Eric Biggers 2 participants 5 messages Jan 5, 2018 – Jan 12, 2018

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

v2 cover crypto: salsa20 - cleanup and convert to skcipher API

From: Eric Biggers <[email protected]>

This series converts the Salsa20 implementations over to the skcipher API, in the process fixing a couple bugs and making them be more similar to the ChaCha20 implementations, rather than doing things slightly differently for no good reason. (Note, however, that Salsa20 still interprets the IV differently from ChaCha20; I didn't change that.)

I'm not sure who is actually using Salsa20 in the kernel, but either way if we're going to have it at all we might as well keep it up to date.

The Salsa20 algorithms still pass the self-tests after these changes.

Changed since v1:

- Use '__le32' type for keystream, as suggested by Stephan.

Eric Biggers (3):
crypto: salsa20-generic - cleanup and convert to skcipher API
crypto: salsa20 - export generic helpers
crypto: x86/salsa20 - cleanup and convert to skcipher API

 arch/x86/crypto/salsa20-i586-asm_32.S   | 184 +-----------------------
 arch/x86/crypto/salsa20-x86_64-asm_64.S | 114 ---------------
 arch/x86/crypto/salsa20_glue.c          | 105 ++++++--------
 crypto/Kconfig                          |   2 +
 crypto/salsa20_generic.c                | 240 ++++++++++++++------------------
 include/crypto/salsa20.h                |  27 ++++
 6 files changed, 174 insertions(+), 498 deletions(-)
 create mode 100644 include/crypto/salsa20.h
Signature · 1 line
2.16.0.rc0.223.g4a4ac83678-goog
#1 lore raw

PATCH v2 1/3 crypto: salsa20-generic - cleanup and convert to skcipher API

From: Eric Biggers <[email protected]>

Convert salsa20-generic from the deprecated "blkcipher" API to the "skcipher" API, in the process fixing it up to be thread-safe (as the crypto API expects) by maintaining each request's state separately from the transform context.

Also remove the unnecessary cra_alignmask and tighten validation of the key size by accepting only 16 or 32 bytes, not anything in between.

These changes bring the code close to the way chacha20-generic does things, so hopefully it will be easier to maintain in the future.

However, the way Salsa20 interprets the IV is still slightly different; that was not changed.

---
 crypto/salsa20_generic.c | 240 ++++++++++++++++++++---------------------------
 1 file changed, 104 insertions(+), 136 deletions(-)
crypto/salsa20_generic.c +104 -136
index d7da0eea5622..8c77bc78a09f 100644
@@ -19,49 +19,27 @@
  *
  */
 
-#include <linux/init.h>
+#include <asm/unaligned.h>
+#include <crypto/internal/skcipher.h>
 #include <linux/module.h>
-#include <linux/errno.h>
-#include <linux/crypto.h>
-#include <linux/types.h>
-#include <linux/bitops.h>
-#include <crypto/algapi.h>
-#include <asm/byteorder.h>
 
-#define SALSA20_IV_SIZE        8U
-#define SALSA20_MIN_KEY_SIZE  16U
-#define SALSA20_MAX_KEY_SIZE  32U
+#define SALSA20_IV_SIZE        8
+#define SALSA20_MIN_KEY_SIZE  16
+#define SALSA20_MAX_KEY_SIZE  32
+#define SALSA20_BLOCK_SIZE    64
 
-/*
- * Start of code taken from D. J. Bernstein's reference implementation.
- * With some modifications and optimizations made to suit our needs.
- */
-
-/*
-salsa20-ref.c version 20051118
-D. J. Bernstein
-Public domain.
-*/
-
-#define U32TO8_LITTLE(p, v) \
-	{ (p)[0] = (v >>  0) & 0xff; (p)[1] = (v >>  8) & 0xff; \
-	  (p)[2] = (v >> 16) & 0xff; (p)[3] = (v >> 24) & 0xff; }
-#define U8TO32_LITTLE(p)   \
-	(((u32)((p)[0])      ) | ((u32)((p)[1]) <<  8) | \
-	 ((u32)((p)[2]) << 16) | ((u32)((p)[3]) << 24)   )
-
-struct salsa20_ctx
-{
-	u32 input[16];
+struct salsa20_ctx {
+	u32 initial_state[16];
 };
 
-static void salsa20_wordtobyte(u8 output[64], const u32 input[16])
+static void salsa20_block(u32 *state, __le32 *stream)
 {
 	u32 x[16];
 	int i;
 
-	memcpy(x, input, sizeof(x));
-	for (i = 20; i > 0; i -= 2) {
+	memcpy(x, state, sizeof(x));
+
+	for (i = 0; i < 20; i += 2) {
 		x[ 4] ^= rol32((x[ 0] + x[12]),  7);
 		x[ 8] ^= rol32((x[ 4] + x[ 0]),  9);
 		x[12] ^= rol32((x[ 8] + x[ 4]), 13);
@@ -95,145 +73,135 @@ static void salsa20_wordtobyte(u8 output[64], const u32 input[16])
 		x[14] ^= rol32((x[13] + x[12]), 13);
 		x[15] ^= rol32((x[14] + x[13]), 18);
 	}
-	for (i = 0; i < 16; ++i)
-		x[i] += input[i];
-	for (i = 0; i < 16; ++i)
-		U32TO8_LITTLE(output + 4 * i,x[i]);
-}
 
-static const char sigma[16] = "expand 32-byte k";
-static const char tau[16] = "expand 16-byte k";
+	for (i = 0; i < 16; i++)
+		stream[i] = cpu_to_le32(x[i] + state[i]);
+
+	if (++state[8] == 0)
+		state[9]++;
+}
 
-static void salsa20_keysetup(struct salsa20_ctx *ctx, const u8 *k, u32 kbytes)
+static void salsa20_docrypt(u32 *state, u8 *dst, const u8 *src,
+			    unsigned int bytes)
 {
-	const char *constants;
+	__le32 stream[SALSA20_BLOCK_SIZE / sizeof(__le32)];
 
-	ctx->input[1] = U8TO32_LITTLE(k + 0);
-	ctx->input[2] = U8TO32_LITTLE(k + 4);
-	ctx->input[3] = U8TO32_LITTLE(k + 8);
-	ctx->input[4] = U8TO32_LITTLE(k + 12);
-	if (kbytes == 32) { /* recommended */
-		k += 16;
-		constants = sigma;
-	} else { /* kbytes == 16 */
-		constants = tau;
+	if (dst != src)
+		memcpy(dst, src, bytes);
+
+	while (bytes >= SALSA20_BLOCK_SIZE) {
+		salsa20_block(state, stream);
+		crypto_xor(dst, (const u8 *)stream, SALSA20_BLOCK_SIZE);
+		bytes -= SALSA20_BLOCK_SIZE;
+		dst += SALSA20_BLOCK_SIZE;
+	}
+	if (bytes) {
+		salsa20_block(state, stream);
+		crypto_xor(dst, (const u8 *)stream, bytes);
 	}
-	ctx->input[11] = U8TO32_LITTLE(k + 0);
-	ctx->input[12] = U8TO32_LITTLE(k + 4);
-	ctx->input[13] = U8TO32_LITTLE(k + 8);
-	ctx->input[14] = U8TO32_LITTLE(k + 12);
-	ctx->input[0] = U8TO32_LITTLE(constants + 0);
-	ctx->input[5] = U8TO32_LITTLE(constants + 4);
-	ctx->input[10] = U8TO32_LITTLE(constants + 8);
-	ctx->input[15] = U8TO32_LITTLE(constants + 12);
 }
 
-static void salsa20_ivsetup(struct salsa20_ctx *ctx, const u8 *iv)
+static void salsa20_init(u32 *state, const struct salsa20_ctx *ctx,
+			 const u8 *iv)
 {
-	ctx->input[6] = U8TO32_LITTLE(iv + 0);
-	ctx->input[7] = U8TO32_LITTLE(iv + 4);
-	ctx->input[8] = 0;
-	ctx->input[9] = 0;
+	memcpy(state, ctx->initial_state, sizeof(ctx->initial_state));
+	state[6] = get_unaligned_le32(iv + 0);
+	state[7] = get_unaligned_le32(iv + 4);
 }
 
-static void salsa20_encrypt_bytes(struct salsa20_ctx *ctx, u8 *dst,
-				  const u8 *src, unsigned int bytes)
+static int salsa20_setkey(struct crypto_skcipher *tfm, const u8 *key,
+			  unsigned int keysize)
 {
-	u8 buf[64];
-
-	if (dst != src)
-		memcpy(dst, src, bytes);
-
-	while (bytes) {
-		salsa20_wordtobyte(buf, ctx->input);
-
-		ctx->input[8]++;
-		if (!ctx->input[8])
-			ctx->input[9]++;
+	static const char sigma[16] = "expand 32-byte k";
+	static const char tau[16] = "expand 16-byte k";
+	struct salsa20_ctx *ctx = crypto_skcipher_ctx(tfm);
+	const char *constants;
 
-		if (bytes <= 64) {
-			crypto_xor(dst, buf, bytes);
-			return;
-		}
+	if (keysize != SALSA20_MIN_KEY_SIZE &&
+	    keysize != SALSA20_MAX_KEY_SIZE)
+		return -EINVAL;
 
-		crypto_xor(dst, buf, 64);
-		bytes -= 64;
-		dst += 64;
+	ctx->initial_state[1] = get_unaligned_le32(key + 0);
+	ctx->initial_state[2] = get_unaligned_le32(key + 4);
+	ctx->initial_state[3] = get_unaligned_le32(key + 8);
+	ctx->initial_state[4] = get_unaligned_le32(key + 12);
+	if (keysize == 32) { /* recommended */
+		key += 16;
+		constants = sigma;
+	} else { /* keysize == 16 */
+		constants = tau;
 	}
-}
-
-/*
- * End of code taken from D. J. Bernstein's reference implementation.
- */
+	ctx->initial_state[11] = get_unaligned_le32(key + 0);
+	ctx->initial_state[12] = get_unaligned_le32(key + 4);
+	ctx->initial_state[13] = get_unaligned_le32(key + 8);
+	ctx->initial_state[14] = get_unaligned_le32(key + 12);
+	ctx->initial_state[0]  = get_unaligned_le32(constants + 0);
+	ctx->initial_state[5]  = get_unaligned_le32(constants + 4);
+	ctx->initial_state[10] = get_unaligned_le32(constants + 8);
+	ctx->initial_state[15] = get_unaligned_le32(constants + 12);
+
+	/* space for the nonce; it will be overridden for each request */
+	ctx->initial_state[6] = 0;
+	ctx->initial_state[7] = 0;
+
+	/* initial block number */
+	ctx->initial_state[8] = 0;
+	ctx->initial_state[9] = 0;
 
-static int setkey(struct crypto_tfm *tfm, const u8 *key,
-		  unsigned int keysize)
-{
-	struct salsa20_ctx *ctx = crypto_tfm_ctx(tfm);
-	salsa20_keysetup(ctx, key, keysize);
 	return 0;
 }
 
-static int encrypt(struct blkcipher_desc *desc,
-		   struct scatterlist *dst, struct scatterlist *src,
-		   unsigned int nbytes)
+static int salsa20_crypt(struct skcipher_request *req)
 {
-	struct blkcipher_walk walk;
-	struct crypto_blkcipher *tfm = desc->tfm;
-	struct salsa20_ctx *ctx = crypto_blkcipher_ctx(tfm);
+	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
+	const struct salsa20_ctx *ctx = crypto_skcipher_ctx(tfm);
+	struct skcipher_walk walk;
+	u32 state[16];
 	int err;
 
-	blkcipher_walk_init(&walk, dst, src, nbytes);
-	err = blkcipher_walk_virt_block(desc, &walk, 64);
+	err = skcipher_walk_virt(&walk, req, true);
 
-	salsa20_ivsetup(ctx, walk.iv);
+	salsa20_init(state, ctx, walk.iv);
 
-	while (walk.nbytes >= 64) {
-		salsa20_encrypt_bytes(ctx, walk.dst.virt.addr,
-				      walk.src.virt.addr,
-				      walk.nbytes - (walk.nbytes % 64));
-		err = blkcipher_walk_done(desc, &walk, walk.nbytes % 64);
-	}
+	while (walk.nbytes > 0) {
+		unsigned int nbytes = walk.nbytes;
 
-	if (walk.nbytes) {
-		salsa20_encrypt_bytes(ctx, walk.dst.virt.addr,
-				      walk.src.virt.addr, walk.nbytes);
-		err = blkcipher_walk_done(desc, &walk, 0);
+		if (nbytes < walk.total)
+			nbytes = round_down(nbytes, walk.stride);
+
+		salsa20_docrypt(state, walk.dst.virt.addr, walk.src.virt.addr,
+				nbytes);
+		err = skcipher_walk_done(&walk, walk.nbytes - nbytes);
 	}
 
 	return err;
 }
 
-static struct crypto_alg alg = {
-	.cra_name           =   "salsa20",
-	.cra_driver_name    =   "salsa20-generic",
-	.cra_priority       =   100,
-	.cra_flags          =   CRYPTO_ALG_TYPE_BLKCIPHER,
-	.cra_type           =   &crypto_blkcipher_type,
-	.cra_blocksize      =   1,
-	.cra_ctxsize        =   sizeof(struct salsa20_ctx),
-	.cra_alignmask      =	3,
-	.cra_module         =   THIS_MODULE,
-	.cra_u              =   {
-		.blkcipher = {
-			.setkey         =   setkey,
-			.encrypt        =   encrypt,
-			.decrypt        =   encrypt,
-			.min_keysize    =   SALSA20_MIN_KEY_SIZE,
-			.max_keysize    =   SALSA20_MAX_KEY_SIZE,
-			.ivsize         =   SALSA20_IV_SIZE,
-		}
-	}
+static struct skcipher_alg alg = {
+	.base.cra_name		= "salsa20",
+	.base.cra_driver_name	= "salsa20-generic",
+	.base.cra_priority	= 100,
+	.base.cra_blocksize	= 1,
+	.base.cra_ctxsize	= sizeof(struct salsa20_ctx),
+	.base.cra_module	= THIS_MODULE,
+
+	.min_keysize		= SALSA20_MIN_KEY_SIZE,
+	.max_keysize		= SALSA20_MAX_KEY_SIZE,
+	.ivsize			= SALSA20_IV_SIZE,
+	.chunksize		= SALSA20_BLOCK_SIZE,
+	.setkey			= salsa20_setkey,
+	.encrypt		= salsa20_crypt,
+	.decrypt		= salsa20_crypt,
 };
 
 static int __init salsa20_generic_mod_init(void)
 {
-	return crypto_register_alg(&alg);
+	return crypto_register_skcipher(&alg);
 }
 
 static void __exit salsa20_generic_mod_fini(void)
 {
-	crypto_unregister_alg(&alg);
+	crypto_unregister_skcipher(&alg);
 }
 
 module_init(salsa20_generic_mod_init);
Signature · 1 line
2.16.0.rc0.223.g4a4ac83678-goog
#2 lore raw

PATCH v2 2/3 crypto: salsa20 - export generic helpers

In reply to Eric Biggers

From: Eric Biggers <[email protected]>

Export the Salsa20 constants, transform context, and initialization functions so that they can be reused by the x86 implementation.

---
 crypto/salsa20_generic.c | 20 +++++++-------------
 include/crypto/salsa20.h | 27 +++++++++++++++++++++++++++
 2 files changed, 34 insertions(+), 13 deletions(-)
 create mode 100644 include/crypto/salsa20.h
crypto/salsa20_generic.c +7 -13
index 8c77bc78a09f..5074006a56c3 100644
@@ -21,17 +21,9 @@
 
 #include <asm/unaligned.h>
 #include <crypto/internal/skcipher.h>
+#include <crypto/salsa20.h>
 #include <linux/module.h>
 
-#define SALSA20_IV_SIZE        8
-#define SALSA20_MIN_KEY_SIZE  16
-#define SALSA20_MAX_KEY_SIZE  32
-#define SALSA20_BLOCK_SIZE    64
-
-struct salsa20_ctx {
-	u32 initial_state[16];
-};
-
 static void salsa20_block(u32 *state, __le32 *stream)
 {
 	u32 x[16];
@@ -101,15 +93,16 @@ static void salsa20_docrypt(u32 *state, u8 *dst, const u8 *src,
 	}
 }
 
-static void salsa20_init(u32 *state, const struct salsa20_ctx *ctx,
+void crypto_salsa20_init(u32 *state, const struct salsa20_ctx *ctx,
 			 const u8 *iv)
 {
 	memcpy(state, ctx->initial_state, sizeof(ctx->initial_state));
 	state[6] = get_unaligned_le32(iv + 0);
 	state[7] = get_unaligned_le32(iv + 4);
 }
+EXPORT_SYMBOL_GPL(crypto_salsa20_init);
 
-static int salsa20_setkey(struct crypto_skcipher *tfm, const u8 *key,
+int crypto_salsa20_setkey(struct crypto_skcipher *tfm, const u8 *key,
 			  unsigned int keysize)
 {
 	static const char sigma[16] = "expand 32-byte k";
@@ -150,6 +143,7 @@ static int salsa20_setkey(struct crypto_skcipher *tfm, const u8 *key,
 
 	return 0;
 }
+EXPORT_SYMBOL_GPL(crypto_salsa20_setkey);
 
 static int salsa20_crypt(struct skcipher_request *req)
 {
@@ -161,7 +155,7 @@ static int salsa20_crypt(struct skcipher_request *req)
 
 	err = skcipher_walk_virt(&walk, req, true);
 
-	salsa20_init(state, ctx, walk.iv);
+	crypto_salsa20_init(state, ctx, walk.iv);
 
 	while (walk.nbytes > 0) {
 		unsigned int nbytes = walk.nbytes;
@@ -189,7 +183,7 @@ static struct skcipher_alg alg = {
 	.max_keysize		= SALSA20_MAX_KEY_SIZE,
 	.ivsize			= SALSA20_IV_SIZE,
 	.chunksize		= SALSA20_BLOCK_SIZE,
-	.setkey			= salsa20_setkey,
+	.setkey			= crypto_salsa20_setkey,
 	.encrypt		= salsa20_crypt,
 	.decrypt		= salsa20_crypt,
 };
include/crypto/salsa20.h new file +27 -0
new file mode 100644
index 000000000000..19ed48aefc86
@@ -0,0 +1,27 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Common values for the Salsa20 algorithm
+ */
+
+#ifndef _CRYPTO_SALSA20_H
+#define _CRYPTO_SALSA20_H
+
+#include <linux/types.h>
+
+#define SALSA20_IV_SIZE		8
+#define SALSA20_MIN_KEY_SIZE	16
+#define SALSA20_MAX_KEY_SIZE	32
+#define SALSA20_BLOCK_SIZE	64
+
+struct crypto_skcipher;
+
+struct salsa20_ctx {
+	u32 initial_state[16];
+};
+
+void crypto_salsa20_init(u32 *state, const struct salsa20_ctx *ctx,
+			 const u8 *iv);
+int crypto_salsa20_setkey(struct crypto_skcipher *tfm, const u8 *key,
+			  unsigned int keysize);
+
+#endif /* _CRYPTO_SALSA20_H */
Signature · 1 line
2.16.0.rc0.223.g4a4ac83678-goog
#3 lore raw

PATCH v2 3/3 crypto: x86/salsa20 - cleanup and convert to skcipher API

In reply to Eric Biggers

From: Eric Biggers <[email protected]>

Convert salsa20-asm from the deprecated "blkcipher" API to the "skcipher" API, in the process fixing it up to use the generic helpers. This allows removing the salsa20_keysetup() and salsa20_ivsetup() assembly functions, which aren't performance critical; the C versions do just fine.

This also fixes the same bug that salsa20-generic had, where the state array was being maintained directly in the transform context rather than on the stack or in the request context. Thus, if multiple threads used the same Salsa20 transform concurrently they produced the wrong results.

---
 arch/x86/crypto/salsa20-i586-asm_32.S   | 184 +-------------------------------
 arch/x86/crypto/salsa20-x86_64-asm_64.S | 114 --------------------
 arch/x86/crypto/salsa20_glue.c          | 105 +++++++-----------
 crypto/Kconfig                          |   2 +
 4 files changed, 46 insertions(+), 359 deletions(-)
arch/x86/crypto/salsa20-i586-asm_32.S +4 -180
index 329452b8f794..6014b7b9e52a 100644
@@ -1,6 +1,7 @@
-# salsa20_pm.s version 20051229
-# D. J. Bernstein
-# Public domain.
+# Derived from:
+#	salsa20_pm.s version 20051229
+#	D. J. Bernstein
+#	Public domain.
 
 #include <linux/linkage.h>
 
@@ -935,180 +936,3 @@ ENTRY(salsa20_encrypt_bytes)
 	# goto bytesatleast1
 	jmp	._bytesatleast1
 ENDPROC(salsa20_encrypt_bytes)
-
-# enter salsa20_keysetup
-ENTRY(salsa20_keysetup)
-	mov	%esp,%eax
-	and	$31,%eax
-	add	$256,%eax
-	sub	%eax,%esp
-	#   eax_stack = eax
-	movl	%eax,64(%esp)
-	#   ebx_stack = ebx
-	movl	%ebx,68(%esp)
-	#   esi_stack = esi
-	movl	%esi,72(%esp)
-	#   edi_stack = edi
-	movl	%edi,76(%esp)
-	#   ebp_stack = ebp
-	movl	%ebp,80(%esp)
-	#   k = arg2
-	movl	8(%esp,%eax),%ecx
-	#   kbits = arg3
-	movl	12(%esp,%eax),%edx
-	#   x = arg1
-	movl	4(%esp,%eax),%eax
-	#   in1 = *(uint32 *) (k + 0)
-	movl	0(%ecx),%ebx
-	#   in2 = *(uint32 *) (k + 4)
-	movl	4(%ecx),%esi
-	#   in3 = *(uint32 *) (k + 8)
-	movl	8(%ecx),%edi
-	#   in4 = *(uint32 *) (k + 12)
-	movl	12(%ecx),%ebp
-	#   *(uint32 *) (x + 4) = in1
-	movl	%ebx,4(%eax)
-	#   *(uint32 *) (x + 8) = in2
-	movl	%esi,8(%eax)
-	#   *(uint32 *) (x + 12) = in3
-	movl	%edi,12(%eax)
-	#   *(uint32 *) (x + 16) = in4
-	movl	%ebp,16(%eax)
-	#   kbits - 256
-	cmp	$256,%edx
-	#   goto kbits128 if unsigned<
-	jb	._kbits128
-._kbits256:
-	#     in11 = *(uint32 *) (k + 16)
-	movl	16(%ecx),%edx
-	#     in12 = *(uint32 *) (k + 20)
-	movl	20(%ecx),%ebx
-	#     in13 = *(uint32 *) (k + 24)
-	movl	24(%ecx),%esi
-	#     in14 = *(uint32 *) (k + 28)
-	movl	28(%ecx),%ecx
-	#     *(uint32 *) (x + 44) = in11
-	movl	%edx,44(%eax)
-	#     *(uint32 *) (x + 48) = in12
-	movl	%ebx,48(%eax)
-	#     *(uint32 *) (x + 52) = in13
-	movl	%esi,52(%eax)
-	#     *(uint32 *) (x + 56) = in14
-	movl	%ecx,56(%eax)
-	#     in0 = 1634760805
-	mov	$1634760805,%ecx
-	#     in5 = 857760878
-	mov	$857760878,%edx
-	#     in10 = 2036477234
-	mov	$2036477234,%ebx
-	#     in15 = 1797285236
-	mov	$1797285236,%esi
-	#     *(uint32 *) (x + 0) = in0
-	movl	%ecx,0(%eax)
-	#     *(uint32 *) (x + 20) = in5
-	movl	%edx,20(%eax)
-	#     *(uint32 *) (x + 40) = in10
-	movl	%ebx,40(%eax)
-	#     *(uint32 *) (x + 60) = in15
-	movl	%esi,60(%eax)
-	#   goto keysetupdone
-	jmp	._keysetupdone
-._kbits128:
-	#     in11 = *(uint32 *) (k + 0)
-	movl	0(%ecx),%edx
-	#     in12 = *(uint32 *) (k + 4)
-	movl	4(%ecx),%ebx
-	#     in13 = *(uint32 *) (k + 8)
-	movl	8(%ecx),%esi
-	#     in14 = *(uint32 *) (k + 12)
-	movl	12(%ecx),%ecx
-	#     *(uint32 *) (x + 44) = in11
-	movl	%edx,44(%eax)
-	#     *(uint32 *) (x + 48) = in12
-	movl	%ebx,48(%eax)
-	#     *(uint32 *) (x + 52) = in13
-	movl	%esi,52(%eax)
-	#     *(uint32 *) (x + 56) = in14
-	movl	%ecx,56(%eax)
-	#     in0 = 1634760805
-	mov	$1634760805,%ecx
-	#     in5 = 824206446
-	mov	$824206446,%edx
-	#     in10 = 2036477238
-	mov	$2036477238,%ebx
-	#     in15 = 1797285236
-	mov	$1797285236,%esi
-	#     *(uint32 *) (x + 0) = in0
-	movl	%ecx,0(%eax)
-	#     *(uint32 *) (x + 20) = in5
-	movl	%edx,20(%eax)
-	#     *(uint32 *) (x + 40) = in10
-	movl	%ebx,40(%eax)
-	#     *(uint32 *) (x + 60) = in15
-	movl	%esi,60(%eax)
-._keysetupdone:
-	#   eax = eax_stack
-	movl	64(%esp),%eax
-	#   ebx = ebx_stack
-	movl	68(%esp),%ebx
-	#   esi = esi_stack
-	movl	72(%esp),%esi
-	#   edi = edi_stack
-	movl	76(%esp),%edi
-	#   ebp = ebp_stack
-	movl	80(%esp),%ebp
-	# leave
-	add	%eax,%esp
-	ret
-ENDPROC(salsa20_keysetup)
-
-# enter salsa20_ivsetup
-ENTRY(salsa20_ivsetup)
-	mov	%esp,%eax
-	and	$31,%eax
-	add	$256,%eax
-	sub	%eax,%esp
-	#   eax_stack = eax
-	movl	%eax,64(%esp)
-	#   ebx_stack = ebx
-	movl	%ebx,68(%esp)
-	#   esi_stack = esi
-	movl	%esi,72(%esp)
-	#   edi_stack = edi
-	movl	%edi,76(%esp)
-	#   ebp_stack = ebp
-	movl	%ebp,80(%esp)
-	#   iv = arg2
-	movl	8(%esp,%eax),%ecx
-	#   x = arg1
-	movl	4(%esp,%eax),%eax
-	#   in6 = *(uint32 *) (iv + 0)
-	movl	0(%ecx),%edx
-	#   in7 = *(uint32 *) (iv + 4)
-	movl	4(%ecx),%ecx
-	#   in8 = 0
-	mov	$0,%ebx
-	#   in9 = 0
-	mov	$0,%esi
-	#   *(uint32 *) (x + 24) = in6
-	movl	%edx,24(%eax)
-	#   *(uint32 *) (x + 28) = in7
-	movl	%ecx,28(%eax)
-	#   *(uint32 *) (x + 32) = in8
-	movl	%ebx,32(%eax)
-	#   *(uint32 *) (x + 36) = in9
-	movl	%esi,36(%eax)
-	#   eax = eax_stack
-	movl	64(%esp),%eax
-	#   ebx = ebx_stack
-	movl	68(%esp),%ebx
-	#   esi = esi_stack
-	movl	72(%esp),%esi
-	#   edi = edi_stack
-	movl	76(%esp),%edi
-	#   ebp = ebp_stack
-	movl	80(%esp),%ebp
-	# leave
-	add	%eax,%esp
-	ret
-ENDPROC(salsa20_ivsetup)
arch/x86/crypto/salsa20-x86_64-asm_64.S +0 -114

117 lines not on this page · raw message

arch/x86/crypto/salsa20_glue.c +40 -65

139 lines not on this page · raw message

crypto/Kconfig +2 -0
index 9327fbfccf5a..b44c0ae04eb2 100644
@@ -1339,6 +1339,7 @@ config CRYPTO_SALSA20_586
 	tristate "Salsa20 stream cipher algorithm (i586)"
 	depends on (X86 || UML_X86) && !64BIT
 	select CRYPTO_BLKCIPHER
+	select CRYPTO_SALSA20
 	help
 	  Salsa20 stream cipher algorithm.
 
@@ -1352,6 +1353,7 @@ config CRYPTO_SALSA20_X86_64
 	tristate "Salsa20 stream cipher algorithm (x86_64)"
 	depends on (X86 || UML_X86) && 64BIT
 	select CRYPTO_BLKCIPHER
+	select CRYPTO_SALSA20
 	help
 	  Salsa20 stream cipher algorithm.
 
Signature · 1 line
2.16.0.rc0.223.g4a4ac83678-goog
maintainer
#4 lore raw
In reply to Eric Biggers
Quoted 22 lines · Eric Biggers

From: Eric Biggers <[email protected]>

This series converts the Salsa20 implementations over to the skcipher API, in the process fixing a couple bugs and making them be more similar to the ChaCha20 implementations, rather than doing things slightly differently for no good reason. (Note, however, that Salsa20 still interprets the IV differently from ChaCha20; I didn't change that.)

I'm not sure who is actually using Salsa20 in the kernel, but either way if we're going to have it at all we might as well keep it up to date.

The Salsa20 algorithms still pass the self-tests after these changes.

Changed since v1:

- Use '__le32' type for keystream, as suggested by Stephan.

Eric Biggers (3):
crypto: salsa20-generic - cleanup and convert to skcipher API
crypto: salsa20 - export generic helpers
crypto: x86/salsa20 - cleanup and convert to skcipher API

Quoted 8 lines
 arch/x86/crypto/salsa20-i586-asm_32.S   | 184 +-----------------------
 arch/x86/crypto/salsa20-x86_64-asm_64.S | 114 ---------------
 arch/x86/crypto/salsa20_glue.c          | 105 ++++++--------
 crypto/Kconfig                          |   2 +
 crypto/salsa20_generic.c                | 240 ++++++++++++++------------------
 include/crypto/salsa20.h                |  27 ++++
 6 files changed, 174 insertions(+), 498 deletions(-)
 create mode 100644 include/crypto/salsa20.h

All applied. Thanks.

Signature · 3 lines
Email: Herbert Xu <[email protected]>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt