crypto: lib/sha256 - Add helpers for block-based shash

Add an internal sha256_finup helper and move the finalisation code
from __sha256_final into it.

Also add sha256_choose_blocks and CRYPTO_ARCH_HAVE_LIB_SHA256_SIMD
so that the Crypto API can use the SIMD block function unconditionally.
The Crypto API must not be used in hard IRQs and there is no reason
to have a fallback path for hardirqs.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

include/crypto/internal/sha2.h

@@ -3,7 +3,12 @@
 #ifndef _CRYPTO_INTERNAL_SHA2_H
 #define _CRYPTO_INTERNAL_SHA2_H
 
+#include <crypto/internal/simd.h>
 #include <crypto/sha2.h>
+#include <linux/compiler_attributes.h>
+#include <linux/string.h>
+#include <linux/types.h>
+#include <linux/unaligned.h>
 
 void sha256_update_generic(struct sha256_state *sctx,
 			   const u8 *data, size_t len);
@@ -24,5 +29,45 @@ void sha256_blocks_generic(u32 state[SHA256_STATE_WORDS],
 			   const u8 *data, size_t nblocks);
 void sha256_blocks_arch(u32 state[SHA256_STATE_WORDS],
 			const u8 *data, size_t nblocks);
+void sha256_blocks_simd(u32 state[SHA256_STATE_WORDS],
+			const u8 *data, size_t nblocks);
+
+static inline void sha256_choose_blocks(
+	u32 state[SHA256_STATE_WORDS], const u8 *data, size_t nblocks,
+	bool force_generic, bool force_simd)
+{
+	if (!IS_ENABLED(CONFIG_CRYPTO_ARCH_HAVE_LIB_SHA256) || force_generic)
+		sha256_blocks_generic(state, data, nblocks);
+	else if (IS_ENABLED(CONFIG_CRYPTO_ARCH_HAVE_LIB_SHA256_SIMD) &&
+		 (force_simd || crypto_simd_usable()))
+		sha256_blocks_simd(state, data, nblocks);
+	else
+		sha256_blocks_arch(state, data, nblocks);
+}
+
+static __always_inline void sha256_finup(
+	struct crypto_sha256_state *sctx, u8 buf[SHA256_BLOCK_SIZE],
+	size_t len, u8 out[SHA256_DIGEST_SIZE], size_t digest_size,
+	bool force_generic, bool force_simd)
+{
+	const size_t bit_offset = SHA256_BLOCK_SIZE - 8;
+	__be64 *bits = (__be64 *)&buf[bit_offset];
+	int i;
+
+	buf[len++] = 0x80;
+	if (len > bit_offset) {
+		memset(&buf[len], 0, SHA256_BLOCK_SIZE - len);
+		sha256_choose_blocks(sctx->state, buf, 1, force_generic,
+				     force_simd);
+		len = 0;
+	}
+
+	memset(&buf[len], 0, bit_offset - len);
+	*bits = cpu_to_be64(sctx->count << 3);
+	sha256_choose_blocks(sctx->state, buf, 1, force_generic, force_simd);
+
+	for (i = 0; i < digest_size; i += 4)
+		put_unaligned_be32(sctx->state[i / 4], out + i);
+}
 
 #endif /* _CRYPTO_INTERNAL_SHA2_H */

lib/crypto/Kconfig

@@ -150,6 +150,14 @@ config CRYPTO_ARCH_HAVE_LIB_SHA256
 	  Declares whether the architecture provides an arch-specific
 	  accelerated implementation of the SHA-256 library interface.
 
+config CRYPTO_ARCH_HAVE_LIB_SHA256_SIMD
+	bool
+	help
+	  Declares whether the architecture provides an arch-specific
+	  accelerated implementation of the SHA-256 library interface
+	  that is SIMD-based and therefore not usable in hardirq
+	  context.
+
 config CRYPTO_LIB_SHA256_GENERIC
 	tristate
 	default CRYPTO_LIB_SHA256 if !CRYPTO_ARCH_HAVE_LIB_SHA256

lib/crypto/sha256.c

@@ -15,7 +15,6 @@
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/string.h>
-#include <linux/unaligned.h>
 
 /*
  * If __DISABLE_EXPORTS is defined, then this file is being compiled for a
@@ -26,14 +25,16 @@
 #include "sha256-generic.c"
 #endif
 
+static inline bool sha256_purgatory(void)
+{
+	return __is_defined(__DISABLE_EXPORTS);
+}
+
 static inline void sha256_blocks(u32 state[SHA256_STATE_WORDS], const u8 *data,
 				 size_t nblocks, bool force_generic)
 {
-#if IS_ENABLED(CONFIG_CRYPTO_ARCH_HAVE_LIB_SHA256) && !defined(__DISABLE_EXPORTS)
+	sha256_choose_blocks(state, data, nblocks,
-	if (!force_generic)
+			     force_generic || sha256_purgatory(), false);
-		return sha256_blocks_arch(state, data, nblocks);
-#endif
-	sha256_blocks_generic(state, data, nblocks);
 }
 
 static inline void __sha256_update(struct sha256_state *sctx, const u8 *data,
@@ -79,25 +80,10 @@ EXPORT_SYMBOL(sha256_update);
 static inline void __sha256_final(struct sha256_state *sctx, u8 *out,
 				  size_t digest_size, bool force_generic)
 {
-	const size_t bit_offset = SHA256_BLOCK_SIZE - sizeof(__be64);
-	__be64 *bits = (__be64 *)&sctx->buf[bit_offset];
 	size_t partial = sctx->count % SHA256_BLOCK_SIZE;
-	size_t i;
-
-	sctx->buf[partial++] = 0x80;
-	if (partial > bit_offset) {
-		memset(&sctx->buf[partial], 0, SHA256_BLOCK_SIZE - partial);
-		sha256_blocks(sctx->state, sctx->buf, 1, force_generic);
-		partial = 0;
-	}
-
-	memset(&sctx->buf[partial], 0, bit_offset - partial);
-	*bits = cpu_to_be64(sctx->count << 3);
-	sha256_blocks(sctx->state, sctx->buf, 1, force_generic);
-
-	for (i = 0; i < digest_size; i += 4)
-		put_unaligned_be32(sctx->state[i / 4], out + i);
 
+	sha256_finup(&sctx->ctx, sctx->buf, partial, out, digest_size,
+		     force_generic || sha256_purgatory(), false);
 	memzero_explicit(sctx, sizeof(*sctx));
 }