From d1c366ee42bd3d2abb4fd855ac4a496b720d8bb6 Mon Sep 17 00:00:00 2001
From: David Sterba <dsterba@suse.com>
Date: Thu, 16 Feb 2023 03:22:58 +0100
Subject: [PATCH] btrfs-progs: crypto: call blake2 implementations by pointer

Change how blake2 implementation is selected. Instead of an if-else check
inside blake2b_compress each time, select the best one and assign the
function pointer. This is slightly faster.

At this point the selection is not implemented properly in
hash-speedtest so all results are from the fastest version. This will
be fixed once all algorithms are converted.

Signed-off-by: David Sterba <dsterba@suse.com>
---
 crypto/blake2.h      |  2 ++
 crypto/blake2b-ref.c | 20 ++++++++++++--------
 crypto/hash.c        |  6 ++++++
 crypto/hash.h        |  1 +
 4 files changed, 21 insertions(+), 8 deletions(-)

diff --git a/crypto/blake2.h b/crypto/blake2.h
index ad62f260..e51d8696 100644
--- a/crypto/blake2.h
+++ b/crypto/blake2.h
@@ -188,6 +188,8 @@ extern "C" {
   /* This is simply an alias for blake2b */
   int blake2( void *out, size_t outlen, const void *in, size_t inlen, const void *key, size_t keylen );
 
+  void blake2_init_accel(void);
+
 #if defined(__cplusplus)
 }
 #endif
diff --git a/crypto/blake2b-ref.c b/crypto/blake2b-ref.c
index 19589451..b39cd59d 100644
--- a/crypto/blake2b-ref.c
+++ b/crypto/blake2b-ref.c
@@ -224,21 +224,25 @@ void blake2b_compress_sse2( blake2b_state *S, const uint8_t block[BLAKE2B_BLOCKB
 void blake2b_compress_sse41( blake2b_state *S, const uint8_t block[BLAKE2B_BLOCKBYTES] );
 void blake2b_compress_avx2( blake2b_state *S, const uint8_t block[BLAKE2B_BLOCKBYTES] );
 
-static void blake2b_compress( blake2b_state *S, const uint8_t block[BLAKE2B_BLOCKBYTES] )
+static void (*blake2b_compress)( blake2b_state *S, const uint8_t block[BLAKE2B_BLOCKBYTES] ) = blake2b_compress_ref;
+
+void blake2_init_accel(void)
 {
+	if (0);
 #if HAVE_AVX2
-	if (cpu_has_feature(CPU_FLAG_AVX2))
-		return blake2b_compress_avx2(S, block);
+	else if (cpu_has_feature(CPU_FLAG_AVX2))
+		blake2b_compress = blake2b_compress_avx2;
 #endif
 #if HAVE_SSE41
-	if (cpu_has_feature(CPU_FLAG_SSE41))
-		return blake2b_compress_sse41(S, block);
+	else if (cpu_has_feature(CPU_FLAG_SSE41))
+		blake2b_compress = blake2b_compress_sse41;
 #endif
 #if HAVE_SSE2
-	if (cpu_has_feature(CPU_FLAG_SSE2))
-		return blake2b_compress_sse2(S, block);
+	else if (cpu_has_feature(CPU_FLAG_SSE2))
+		blake2b_compress = blake2b_compress_sse2;
 #endif
-	return blake2b_compress_ref(S, block);
+	else
+		blake2b_compress = blake2b_compress_ref;
 }
 
 int blake2b_update( blake2b_state *S, const void *pin, size_t inlen )
diff --git a/crypto/hash.c b/crypto/hash.c
index 4ee65385..0ccdab75 100644
--- a/crypto/hash.c
+++ b/crypto/hash.c
@@ -23,6 +23,12 @@
 void hash_init_accel(void)
 {
 	crc32c_optimization_init();
+	blake2_init_accel();
+}
+
+void hash_init_blake2(void)
+{
+	blake2_init_accel();
 }
 
 /*
diff --git a/crypto/hash.h b/crypto/hash.h
index 40a579ce..e1e073c3 100644
--- a/crypto/hash.h
+++ b/crypto/hash.h
@@ -27,5 +27,6 @@ int hash_sha256(const u8 *buf, size_t length, u8 *out);
 int hash_blake2b(const u8 *buf, size_t length, u8 *out);
 
 void hash_init_accel(void);
+void hash_init_blake2(void);
 
 #endif