dm-verity: use SHA-256 library for SHA-256

When the hash algorithm is SHA-256 and the verity version is not 0, use
the SHA-256 library instead of crypto_shash.

This is a prerequisite for making dm-verity interleave the computation
of SHA-256 hashes for increased performance.  That optimization is
available in the SHA-256 library but not in crypto_shash.

Even without interleaved hashing, switching to the library also slightly
improves performance by itself because it avoids the overhead of
crypto_shash, including indirect calls and other API overhead.
(Benchmark on x86_64, AMD Zen 5: hashing 4K blocks gets 2.1% faster.)

SHA-256 is by far the most common hash algorithm used with dm-verity.
It makes sense to optimize for the common case and fall back to the
generic crypto layer for uncommon cases, as suggested by Linus:
https://lore.kernel.org/r/CAHk-=wgp-fOSsZsYrbyzqCAfEvrt5jQs1jL-97Wc4seMNTUyng@mail.gmail.com

Signed-off-by: Eric Biggers <ebiggers@kernel.org>
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>

drivers/md/Kconfig

@@ -547,6 +547,7 @@ config DM_VERITY
 	depends on BLK_DEV_DM
 	select CRYPTO
 	select CRYPTO_HASH
+	select CRYPTO_LIB_SHA256
 	select DM_BUFIO
 	help
 	  This device-mapper target creates a read-only device that

drivers/md/dm-verity-target.c

@@ -117,11 +117,25 @@ static sector_t verity_position_at_level(struct dm_verity *v, sector_t block,
 int verity_hash(struct dm_verity *v, struct dm_verity_io *io,
 		const u8 *data, size_t len, u8 *digest)
 {
-	struct shash_desc *desc = &io->hash_desc;
+	struct shash_desc *desc;
 	int r;
 
+	if (likely(v->use_sha256_lib)) {
+		struct sha256_ctx *ctx = &io->hash_ctx.sha256;
+
+		/*
+		 * Fast path using SHA-256 library.  This is enabled only for
+		 * verity version 1, where the salt is at the beginning.
+		 */
+		*ctx = *v->initial_hashstate.sha256;
+		sha256_update(ctx, data, len);
+		sha256_final(ctx, digest);
+		return 0;
+	}
+
+	desc = &io->hash_ctx.shash;
 	desc->tfm = v->shash_tfm;
-	if (unlikely(v->initial_hashstate == NULL)) {
+	if (unlikely(v->initial_hashstate.shash == NULL)) {
 		/* Version 0: salt at end */
 		r = crypto_shash_init(desc) ?:
 		    crypto_shash_update(desc, data, len) ?:
@@ -129,7 +143,7 @@ int verity_hash(struct dm_verity *v, struct dm_verity_io *io,
 		    crypto_shash_final(desc, digest);
 	} else {
 		/* Version 1: salt at beginning */
-		r = crypto_shash_import(desc, v->initial_hashstate) ?:
+		r = crypto_shash_import(desc, v->initial_hashstate.shash) ?:
 		    crypto_shash_finup(desc, data, len, digest);
 	}
 	if (unlikely(r))
@@ -1004,7 +1018,7 @@ static void verity_dtr(struct dm_target *ti)
 
 	kvfree(v->validated_blocks);
 	kfree(v->salt);
-	kfree(v->initial_hashstate);
+	kfree(v->initial_hashstate.shash);
 	kfree(v->root_digest);
 	kfree(v->zero_digest);
 	verity_free_sig(v);
@@ -1069,8 +1083,7 @@ static int verity_alloc_zero_digest(struct dm_verity *v)
 	if (!v->zero_digest)
 		return r;
 
-	io = kmalloc(sizeof(*io) + crypto_shash_descsize(v->shash_tfm),
-		     GFP_KERNEL);
+	io = kmalloc(v->ti->per_io_data_size, GFP_KERNEL);
 
 	if (!io)
 		return r; /* verity_dtr will free zero_digest */
@@ -1256,6 +1269,20 @@ static int verity_setup_hash_alg(struct dm_verity *v, const char *alg_name)
 		ti->error = "Digest size too big";
 		return -EINVAL;
 	}
+	if (likely(v->version && strcmp(alg_name, "sha256") == 0)) {
+		/*
+		 * Fast path: use the library API for reduced overhead and
+		 * interleaved hashing support.
+		 */
+		v->use_sha256_lib = true;
+		ti->per_io_data_size =
+			offsetofend(struct dm_verity_io, hash_ctx.sha256);
+	} else {
+		/* Fallback case: use the generic crypto API. */
+		ti->per_io_data_size =
+			offsetofend(struct dm_verity_io, hash_ctx.shash) +
+			crypto_shash_descsize(shash);
+	}
 	return 0;
 }
 
@@ -1276,7 +1303,18 @@ static int verity_setup_salt_and_hashstate(struct dm_verity *v, const char *arg)
 			return -EINVAL;
 		}
 	}
-	if (v->version) { /* Version 1: salt at beginning */
+	if (likely(v->use_sha256_lib)) {
+		/* Implies version 1: salt at beginning */
+		v->initial_hashstate.sha256 =
+			kmalloc(sizeof(struct sha256_ctx), GFP_KERNEL);
+		if (!v->initial_hashstate.sha256) {
+			ti->error = "Cannot allocate initial hash state";
+			return -ENOMEM;
+		}
+		sha256_init(v->initial_hashstate.sha256);
+		sha256_update(v->initial_hashstate.sha256,
+			      v->salt, v->salt_size);
+	} else if (v->version) { /* Version 1: salt at beginning */
 		SHASH_DESC_ON_STACK(desc, v->shash_tfm);
 		int r;
 
@@ -1284,16 +1322,16 @@ static int verity_setup_salt_and_hashstate(struct dm_verity *v, const char *arg)
 		 * Compute the pre-salted hash state that can be passed to
 		 * crypto_shash_import() for each block later.
 		 */
-		v->initial_hashstate = kmalloc(
+		v->initial_hashstate.shash = kmalloc(
 			crypto_shash_statesize(v->shash_tfm), GFP_KERNEL);
-		if (!v->initial_hashstate) {
+		if (!v->initial_hashstate.shash) {
 			ti->error = "Cannot allocate initial hash state";
 			return -ENOMEM;
 		}
 		desc->tfm = v->shash_tfm;
 		r = crypto_shash_init(desc) ?:
 		    crypto_shash_update(desc, v->salt, v->salt_size) ?:
-		    crypto_shash_export(desc, v->initial_hashstate);
+		    crypto_shash_export(desc, v->initial_hashstate.shash);
 		if (r) {
 			ti->error = "Cannot set up initial hash state";
 			return r;
@@ -1555,9 +1593,6 @@ static int verity_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		goto bad;
 	}
 
-	ti->per_io_data_size = sizeof(struct dm_verity_io) +
-			       crypto_shash_descsize(v->shash_tfm);
-
 	r = verity_fec_ctr(v);
 	if (r)
 		goto bad;

drivers/md/dm-verity.h

@@ -16,6 +16,7 @@
 #include <linux/device-mapper.h>
 #include <linux/interrupt.h>
 #include <crypto/hash.h>
+#include <crypto/sha2.h>
 
 #define DM_VERITY_MAX_LEVELS		63
 
@@ -42,7 +43,10 @@ struct dm_verity {
 	struct crypto_shash *shash_tfm;
 	u8 *root_digest;	/* digest of the root block */
 	u8 *salt;		/* salt: its size is salt_size */
-	u8 *initial_hashstate;	/* salted initial state, if version >= 1 */
+	union {
+		struct sha256_ctx *sha256;	/* for use_sha256_lib=1 */
+		u8 *shash;			/* for use_sha256_lib=0 */
+	} initial_hashstate; /* salted initial state, if version >= 1 */
 	u8 *zero_digest;	/* digest for a zero block */
 #ifdef CONFIG_SECURITY
 	u8 *root_digest_sig;	/* signature of the root digest */
@@ -59,6 +63,7 @@ struct dm_verity {
 	unsigned char version;
 	bool hash_failed:1;	/* set if hash of any block failed */
 	bool use_bh_wq:1;	/* try to verify in BH wq before normal work-queue */
+	bool use_sha256_lib:1;	/* use SHA-256 library instead of generic crypto API */
 	unsigned int digest_size;	/* digest size for the current hash algorithm */
 	enum verity_mode mode;	/* mode for handling verification errors */
 	enum verity_mode error_mode;/* mode for handling I/O errors */
@@ -98,11 +103,16 @@ struct dm_verity_io {
 	u8 want_digest[HASH_MAX_DIGESTSIZE];
 
 	/*
-	 * Temporary space for hashing.  This is variable-length and must be at
-	 * the end of the struct.  struct shash_desc is just the fixed part;
-	 * it's followed by a context of size crypto_shash_descsize(shash_tfm).
+	 * Temporary space for hashing.  Either sha256 or shash is used,
+	 * depending on the value of use_sha256_lib.  If shash is used,
+	 * then this field is variable-length, with total size
+	 * sizeof(struct shash_desc) + crypto_shash_descsize(shash_tfm).
+	 * For this reason, this field must be the end of the struct.
 	 */
-	struct shash_desc hash_desc;
+	union {
+		struct sha256_ctx sha256;
+		struct shash_desc shash;
+	} hash_ctx;
 };
 
 static inline u8 *verity_io_real_digest(struct dm_verity *v,