crypto: arm64/chacha - depend on generic chacha library instead of crypto driver

Depend on the generic ChaCha library routines instead of pulling in the
generic ChaCha skcipher driver, which is more than we need, and makes
managing the dependencies between the generic library, generic driver,
accelerated library and driver more complicated.

While at it, drop the logic to prefer the scalar code on short inputs.
Turning the NEON on and off is cheap these days, and one major use case
for ChaCha20 is ChaCha20-Poly1305, which is guaranteed to hit the scalar
path upon every invocation  (when doing the Poly1305 nonce generation)
Signed-off-by: Ard Biesheuvel <ardb@kernel.org>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

arch/arm64/crypto/Kconfig

@@ -103,7 +103,7 @@ config CRYPTO_CHACHA20_NEON
 	tristate "ChaCha20, XChaCha20, and XChaCha12 stream ciphers using NEON instructions"
 	depends on KERNEL_MODE_NEON
 	select CRYPTO_SKCIPHER
-	select CRYPTO_CHACHA20
+	select CRYPTO_LIB_CHACHA_GENERIC
 
 config CRYPTO_NHPOLY1305_NEON
 	tristate "NHPoly1305 hash function using NEON instructions (for Adiantum)"

arch/arm64/crypto/chacha-neon-glue.c

@@ -68,7 +68,7 @@ static int chacha_neon_stream_xor(struct skcipher_request *req,
 
 	err = skcipher_walk_virt(&walk, req, false);
 
-	crypto_chacha_init(state, ctx, iv);
+	chacha_init_generic(state, ctx->key, iv);
 
 	while (walk.nbytes > 0) {
 		unsigned int nbytes = walk.nbytes;
@@ -76,10 +76,16 @@ static int chacha_neon_stream_xor(struct skcipher_request *req,
 		if (nbytes < walk.total)
 			nbytes = rounddown(nbytes, walk.stride);
 
-		kernel_neon_begin();
-		chacha_doneon(state, walk.dst.virt.addr, walk.src.virt.addr,
-			      nbytes, ctx->nrounds);
-		kernel_neon_end();
+		if (!crypto_simd_usable()) {
+			chacha_crypt_generic(state, walk.dst.virt.addr,
+					     walk.src.virt.addr, nbytes,
+					     ctx->nrounds);
+		} else {
+			kernel_neon_begin();
+			chacha_doneon(state, walk.dst.virt.addr,
+				      walk.src.virt.addr, nbytes, ctx->nrounds);
+			kernel_neon_end();
+		}
 		err = skcipher_walk_done(&walk, walk.nbytes - nbytes);
 	}
 
@@ -91,9 +97,6 @@ static int chacha_neon(struct skcipher_request *req)
 	struct crypto_skcipher *tfm = crypto_skcipher_reqtfm(req);
 	struct chacha_ctx *ctx = crypto_skcipher_ctx(tfm);
 
-	if (req->cryptlen <= CHACHA_BLOCK_SIZE || !crypto_simd_usable())
-		return crypto_chacha_crypt(req);
-
 	return chacha_neon_stream_xor(req, ctx, req->iv);
 }
 
@@ -105,14 +108,15 @@ static int xchacha_neon(struct skcipher_request *req)
 	u32 state[16];
 	u8 real_iv[16];
 
-	if (req->cryptlen <= CHACHA_BLOCK_SIZE || !crypto_simd_usable())
-		return crypto_xchacha_crypt(req);
+	chacha_init_generic(state, ctx->key, req->iv);
 
-	crypto_chacha_init(state, ctx, req->iv);
-
-	kernel_neon_begin();
-	hchacha_block_neon(state, subctx.key, ctx->nrounds);
-	kernel_neon_end();
+	if (crypto_simd_usable()) {
+		kernel_neon_begin();
+		hchacha_block_neon(state, subctx.key, ctx->nrounds);
+		kernel_neon_end();
+	} else {
+		hchacha_block_generic(state, subctx.key, ctx->nrounds);
+	}
 	subctx.nrounds = ctx->nrounds;
 
 	memcpy(&real_iv[0], req->iv + 24, 8);
@@ -134,7 +138,7 @@ static struct skcipher_alg algs[] = {
 		.ivsize			= CHACHA_IV_SIZE,
 		.chunksize		= CHACHA_BLOCK_SIZE,
 		.walksize		= 5 * CHACHA_BLOCK_SIZE,
-		.setkey			= crypto_chacha20_setkey,
+		.setkey			= chacha20_setkey,
 		.encrypt		= chacha_neon,
 		.decrypt		= chacha_neon,
 	}, {
@@ -150,7 +154,7 @@ static struct skcipher_alg algs[] = {
 		.ivsize			= XCHACHA_IV_SIZE,
 		.chunksize		= CHACHA_BLOCK_SIZE,
 		.walksize		= 5 * CHACHA_BLOCK_SIZE,
-		.setkey			= crypto_chacha20_setkey,
+		.setkey			= chacha20_setkey,
 		.encrypt		= xchacha_neon,
 		.decrypt		= xchacha_neon,
 	}, {
@@ -166,7 +170,7 @@ static struct skcipher_alg algs[] = {
 		.ivsize			= XCHACHA_IV_SIZE,
 		.chunksize		= CHACHA_BLOCK_SIZE,
 		.walksize		= 5 * CHACHA_BLOCK_SIZE,
-		.setkey			= crypto_chacha12_setkey,
+		.setkey			= chacha12_setkey,
 		.encrypt		= xchacha_neon,
 		.decrypt		= xchacha_neon,
 	}