Merge tag 'fsverity-for-linus' of git://git.kernel.org/pub/scm/fs/fscrypt/fscrypt

Pull fsverity updates from Eric Biggers:

 - Optimize fs-verity sequential read performance by implementing
   readahead of Merkle tree pages. This allows the Merkle tree to be
   read in larger chunks.

 - Optimize FS_IOC_ENABLE_VERITY performance in the uncached case by
   implementing readahead of data pages.

 - Allocate the hash requests from a mempool in order to eliminate the
   possibility of allocation failures during I/O.

* tag 'fsverity-for-linus' of git://git.kernel.org/pub/scm/fs/fscrypt/fscrypt:
  fs-verity: use u64_to_user_ptr()
  fs-verity: use mempool for hash requests
  fs-verity: implement readahead of Merkle tree pages
  fs-verity: implement readahead for FS_IOC_ENABLE_VERITY

fs/ext4/verity.c

@@ -342,12 +342,55 @@ static int ext4_get_verity_descriptor(struct inode *inode, void *buf,
 	return desc_size;
 }
 
+/*
+ * Prefetch some pages from the file's Merkle tree.
+ *
+ * This is basically a stripped-down version of __do_page_cache_readahead()
+ * which works on pages past i_size.
+ */
+static void ext4_merkle_tree_readahead(struct address_space *mapping,
+				       pgoff_t start_index, unsigned long count)
+{
+	LIST_HEAD(pages);
+	unsigned int nr_pages = 0;
+	struct page *page;
+	pgoff_t index;
+	struct blk_plug plug;
+
+	for (index = start_index; index < start_index + count; index++) {
+		page = xa_load(&mapping->i_pages, index);
+		if (!page || xa_is_value(page)) {
+			page = __page_cache_alloc(readahead_gfp_mask(mapping));
+			if (!page)
+				break;
+			page->index = index;
+			list_add(&page->lru, &pages);
+			nr_pages++;
+		}
+	}
+	blk_start_plug(&plug);
+	ext4_mpage_readpages(mapping, &pages, NULL, nr_pages, true);
+	blk_finish_plug(&plug);
+}
+
 static struct page *ext4_read_merkle_tree_page(struct inode *inode,
-					       pgoff_t index)
+					       pgoff_t index,
+					       unsigned long num_ra_pages)
 {
+	struct page *page;
+
 	index += ext4_verity_metadata_pos(inode) >> PAGE_SHIFT;
 
-	return read_mapping_page(inode->i_mapping, index, NULL);
+	page = find_get_page_flags(inode->i_mapping, index, FGP_ACCESSED);
+	if (!page || !PageUptodate(page)) {
+		if (page)
+			put_page(page);
+		else if (num_ra_pages > 1)
+			ext4_merkle_tree_readahead(inode->i_mapping, index,
+						   num_ra_pages);
+		page = read_mapping_page(inode->i_mapping, index, NULL);
+	}
+	return page;
 }
 
 static int ext4_write_merkle_tree_block(struct inode *inode, const void *buf,

fs/f2fs/data.c

@@ -1881,7 +1881,7 @@ static int f2fs_read_single_page(struct inode *inode, struct page *page,
  * use ->readpage() or do the necessary surgery to decouple ->readpages()
  * from read-ahead.
  */
-static int f2fs_mpage_readpages(struct address_space *mapping,
+int f2fs_mpage_readpages(struct address_space *mapping,
 			struct list_head *pages, struct page *page,
 			unsigned nr_pages, bool is_readahead)
 {

fs/f2fs/f2fs.h

@@ -3229,6 +3229,9 @@ int f2fs_reserve_new_block(struct dnode_of_data *dn);
 int f2fs_get_block(struct dnode_of_data *dn, pgoff_t index);
 int f2fs_preallocate_blocks(struct kiocb *iocb, struct iov_iter *from);
 int f2fs_reserve_block(struct dnode_of_data *dn, pgoff_t index);
+int f2fs_mpage_readpages(struct address_space *mapping,
+			struct list_head *pages, struct page *page,
+			unsigned nr_pages, bool is_readahead);
 struct page *f2fs_get_read_data_page(struct inode *inode, pgoff_t index,
 			int op_flags, bool for_write);
 struct page *f2fs_find_data_page(struct inode *inode, pgoff_t index);

fs/f2fs/verity.c

@@ -222,12 +222,55 @@ static int f2fs_get_verity_descriptor(struct inode *inode, void *buf,
 	return size;
 }
 
+/*
+ * Prefetch some pages from the file's Merkle tree.
+ *
+ * This is basically a stripped-down version of __do_page_cache_readahead()
+ * which works on pages past i_size.
+ */
+static void f2fs_merkle_tree_readahead(struct address_space *mapping,
+				       pgoff_t start_index, unsigned long count)
+{
+	LIST_HEAD(pages);
+	unsigned int nr_pages = 0;
+	struct page *page;
+	pgoff_t index;
+	struct blk_plug plug;
+
+	for (index = start_index; index < start_index + count; index++) {
+		page = xa_load(&mapping->i_pages, index);
+		if (!page || xa_is_value(page)) {
+			page = __page_cache_alloc(readahead_gfp_mask(mapping));
+			if (!page)
+				break;
+			page->index = index;
+			list_add(&page->lru, &pages);
+			nr_pages++;
+		}
+	}
+	blk_start_plug(&plug);
+	f2fs_mpage_readpages(mapping, &pages, NULL, nr_pages, true);
+	blk_finish_plug(&plug);
+}
+
 static struct page *f2fs_read_merkle_tree_page(struct inode *inode,
-					       pgoff_t index)
+					       pgoff_t index,
+					       unsigned long num_ra_pages)
 {
+	struct page *page;
+
 	index += f2fs_verity_metadata_pos(inode) >> PAGE_SHIFT;
 
-	return read_mapping_page(inode->i_mapping, index, NULL);
+	page = find_get_page_flags(inode->i_mapping, index, FGP_ACCESSED);
+	if (!page || !PageUptodate(page)) {
+		if (page)
+			put_page(page);
+		else if (num_ra_pages > 1)
+			f2fs_merkle_tree_readahead(inode->i_mapping, index,
+						   num_ra_pages);
+		page = read_mapping_page(inode->i_mapping, index, NULL);
+	}
+	return page;
 }
 
 static int f2fs_write_merkle_tree_block(struct inode *inode, const void *buf,

fs/verity/enable.c

@@ -8,18 +8,48 @@
 #include "fsverity_private.h"
 
 #include <crypto/hash.h>
+#include <linux/backing-dev.h>
 #include <linux/mount.h>
 #include <linux/pagemap.h>
 #include <linux/sched/signal.h>
 #include <linux/uaccess.h>
 
-static int build_merkle_tree_level(struct inode *inode, unsigned int level,
+/*
+ * Read a file data page for Merkle tree construction.  Do aggressive readahead,
+ * since we're sequentially reading the entire file.
+ */
+static struct page *read_file_data_page(struct file *filp, pgoff_t index,
+					struct file_ra_state *ra,
+					unsigned long remaining_pages)
+{
+	struct page *page;
+
+	page = find_get_page_flags(filp->f_mapping, index, FGP_ACCESSED);
+	if (!page || !PageUptodate(page)) {
+		if (page)
+			put_page(page);
+		else
+			page_cache_sync_readahead(filp->f_mapping, ra, filp,
+						  index, remaining_pages);
+		page = read_mapping_page(filp->f_mapping, index, NULL);
+		if (IS_ERR(page))
+			return page;
+	}
+	if (PageReadahead(page))
+		page_cache_async_readahead(filp->f_mapping, ra, filp, page,
+					   index, remaining_pages);
+	return page;
+}
+
+static int build_merkle_tree_level(struct file *filp, unsigned int level,
 				   u64 num_blocks_to_hash,
 				   const struct merkle_tree_params *params,
 				   u8 *pending_hashes,
 				   struct ahash_request *req)
 {
+	struct inode *inode = file_inode(filp);
 	const struct fsverity_operations *vops = inode->i_sb->s_vop;
+	struct file_ra_state ra = { 0 };
 	unsigned int pending_size = 0;
 	u64 dst_block_num;
 	u64 i;
@@ -36,6 +66,8 @@ static int build_merkle_tree_level(struct inode *inode, unsigned int level,
 		dst_block_num = 0; /* unused */
 	}
 
+	file_ra_state_init(&ra, filp->f_mapping);
+
 	for (i = 0; i < num_blocks_to_hash; i++) {
 		struct page *src_page;
 
@@ -45,7 +77,8 @@ static int build_merkle_tree_level(struct inode *inode, unsigned int level,
 
 		if (level == 0) {
 			/* Leaf: hashing a data block */
-			src_page = read_mapping_page(inode->i_mapping, i, NULL);
+			src_page = read_file_data_page(filp, i, &ra,
+						       num_blocks_to_hash - i);
 			if (IS_ERR(src_page)) {
 				err = PTR_ERR(src_page);
 				fsverity_err(inode,
@@ -54,9 +87,14 @@ static int build_merkle_tree_level(struct inode *inode, unsigned int level,
 				return err;
 			}
 		} else {
+			unsigned long num_ra_pages =
+				min_t(unsigned long, num_blocks_to_hash - i,
+				      inode->i_sb->s_bdi->io_pages);
+
 			/* Non-leaf: hashing hash block from level below */
 			src_page = vops->read_merkle_tree_page(inode,
-					params->level_start[level - 1] + i);
+					params->level_start[level - 1] + i,
+					num_ra_pages);
 			if (IS_ERR(src_page)) {
 				err = PTR_ERR(src_page);
 				fsverity_err(inode,
@@ -103,17 +141,18 @@ static int build_merkle_tree_level(struct inode *inode, unsigned int level,
 }
 
 /*
- * Build the Merkle tree for the given inode using the given parameters, and
+ * Build the Merkle tree for the given file using the given parameters, and
  * return the root hash in @root_hash.
  *
  * The tree is written to a filesystem-specific location as determined by the
  * ->write_merkle_tree_block() method.  However, the blocks that comprise the
  * tree are the same for all filesystems.
  */
-static int build_merkle_tree(struct inode *inode,
+static int build_merkle_tree(struct file *filp,
 			     const struct merkle_tree_params *params,
 			     u8 *root_hash)
 {
+	struct inode *inode = file_inode(filp);
 	u8 *pending_hashes;
 	struct ahash_request *req;
 	u64 blocks;
@@ -126,9 +165,11 @@ static int build_merkle_tree(struct inode *inode,
 		return 0;
 	}
 
+	/* This allocation never fails, since it's mempool-backed. */
+	req = fsverity_alloc_hash_request(params->hash_alg, GFP_KERNEL);
+
 	pending_hashes = kmalloc(params->block_size, GFP_KERNEL);
-	req = ahash_request_alloc(params->hash_alg->tfm, GFP_KERNEL);
-	if (!pending_hashes || !req)
+	if (!pending_hashes)
 		goto out;
 
 	/*
@@ -139,7 +180,7 @@ static int build_merkle_tree(struct inode *inode,
 	blocks = (inode->i_size + params->block_size - 1) >>
 		 params->log_blocksize;
 	for (level = 0; level <= params->num_levels; level++) {
-		err = build_merkle_tree_level(inode, level, blocks, params,
+		err = build_merkle_tree_level(filp, level, blocks, params,
 					      pending_hashes, req);
 		if (err)
 			goto out;
@@ -150,7 +191,7 @@ static int build_merkle_tree(struct inode *inode,
 	err = 0;
 out:
 	kfree(pending_hashes);
-	ahash_request_free(req);
+	fsverity_free_hash_request(params->hash_alg, req);
 	return err;
 }
 
@@ -175,8 +216,7 @@ static int enable_verity(struct file *filp,
 
 	/* Get the salt if the user provided one */
 	if (arg->salt_size &&
-	    copy_from_user(desc->salt,
-			   (const u8 __user *)(uintptr_t)arg->salt_ptr,
+	    copy_from_user(desc->salt, u64_to_user_ptr(arg->salt_ptr),
 			   arg->salt_size)) {
 		err = -EFAULT;
 		goto out;
@@ -185,8 +225,7 @@ static int enable_verity(struct file *filp,
 
 	/* Get the signature if the user provided one */
 	if (arg->sig_size &&
-	    copy_from_user(desc->signature,
-			   (const u8 __user *)(uintptr_t)arg->sig_ptr,
+	    copy_from_user(desc->signature, u64_to_user_ptr(arg->sig_ptr),
 			   arg->sig_size)) {
 		err = -EFAULT;
 		goto out;
@@ -227,7 +266,7 @@ static int enable_verity(struct file *filp,
 	 */
 	pr_debug("Building Merkle tree...\n");
 	BUILD_BUG_ON(sizeof(desc->root_hash) < FS_VERITY_MAX_DIGEST_SIZE);
-	err = build_merkle_tree(inode, &params, desc->root_hash);
+	err = build_merkle_tree(filp, &params, desc->root_hash);
 	if (err) {
 		fsverity_err(inode, "Error %d building Merkle tree", err);
 		goto rollback;

fs/verity/fsverity_private.h

@@ -16,6 +16,7 @@
 
 #include <crypto/sha.h>
 #include <linux/fsverity.h>
+#include <linux/mempool.h>
 
 struct ahash_request;
 
@@ -37,11 +38,12 @@ struct fsverity_hash_alg {
 	const char *name;	  /* crypto API name, e.g. sha256 */
 	unsigned int digest_size; /* digest size in bytes, e.g. 32 for SHA-256 */
 	unsigned int block_size;  /* block size in bytes, e.g. 64 for SHA-256 */
+	mempool_t req_pool;	  /* mempool with a preallocated hash request */
 };
 
 /* Merkle tree parameters: hash algorithm, initial hash state, and topology */
 struct merkle_tree_params {
-	const struct fsverity_hash_alg *hash_alg; /* the hash algorithm */
+	struct fsverity_hash_alg *hash_alg; /* the hash algorithm */
 	const u8 *hashstate;		/* initial hash state or NULL */
 	unsigned int digest_size;	/* same as hash_alg->digest_size */
 	unsigned int block_size;	/* size of data and tree blocks */
@@ -50,6 +52,7 @@ struct merkle_tree_params {
 	unsigned int log_arity;		/* log2(hashes_per_block) */
 	unsigned int num_levels;	/* number of levels in Merkle tree */
 	u64 tree_size;			/* Merkle tree size in bytes */
+	unsigned long level0_blocks;	/* number of blocks in tree level 0 */
 
 	/*
 	 * Starting block index for each tree level, ordered from leaf level (0)
@@ -114,14 +117,18 @@ struct fsverity_signed_digest {
 
 extern struct fsverity_hash_alg fsverity_hash_algs[];
 
-const struct fsverity_hash_alg *fsverity_get_hash_alg(const struct inode *inode,
+struct fsverity_hash_alg *fsverity_get_hash_alg(const struct inode *inode,
 						unsigned int num);
-const u8 *fsverity_prepare_hash_state(const struct fsverity_hash_alg *alg,
+struct ahash_request *fsverity_alloc_hash_request(struct fsverity_hash_alg *alg,
+						  gfp_t gfp_flags);
+void fsverity_free_hash_request(struct fsverity_hash_alg *alg,
+				struct ahash_request *req);
+const u8 *fsverity_prepare_hash_state(struct fsverity_hash_alg *alg,
 				      const u8 *salt, size_t salt_size);
 int fsverity_hash_page(const struct merkle_tree_params *params,
 		       const struct inode *inode,
 		       struct ahash_request *req, struct page *page, u8 *out);
-int fsverity_hash_buffer(const struct fsverity_hash_alg *alg,
+int fsverity_hash_buffer(struct fsverity_hash_alg *alg,
 			 const void *data, size_t size, u8 *out);
 void __init fsverity_check_hash_algs(void);
 

fs/verity/hash_algs.c

@@ -24,6 +24,8 @@ struct fsverity_hash_alg fsverity_hash_algs[] = {
 	},
 };
 
+static DEFINE_MUTEX(fsverity_hash_alg_init_mutex);
+
 /**
  * fsverity_get_hash_alg() - validate and prepare a hash algorithm
  * @inode: optional inode for logging purposes
@@ -36,7 +38,7 @@ struct fsverity_hash_alg fsverity_hash_algs[] = {
  *
  * Return: pointer to the hash alg on success, else an ERR_PTR()
  */
-const struct fsverity_hash_alg *fsverity_get_hash_alg(const struct inode *inode,
+struct fsverity_hash_alg *fsverity_get_hash_alg(const struct inode *inode,
 						unsigned int num)
 {
 	struct fsverity_hash_alg *alg;
@@ -50,10 +52,15 @@ const struct fsverity_hash_alg *fsverity_get_hash_alg(const struct inode *inode,
 	}
 	alg = &fsverity_hash_algs[num];
 
-	/* pairs with cmpxchg() below */
-	tfm = READ_ONCE(alg->tfm);
-	if (likely(tfm != NULL))
+	/* pairs with smp_store_release() below */
+	if (likely(smp_load_acquire(&alg->tfm) != NULL))
 		return alg;
+
+	mutex_lock(&fsverity_hash_alg_init_mutex);
+
+	if (alg->tfm != NULL)
+		goto out_unlock;
+
 	/*
 	 * Using the shash API would make things a bit simpler, but the ahash
 	 * API is preferable as it allows the use of crypto accelerators.
@@ -64,12 +71,14 @@ const struct fsverity_hash_alg *fsverity_get_hash_alg(const struct inode *inode,
 			fsverity_warn(inode,
 				      "Missing crypto API support for hash algorithm \"%s\"",
 				      alg->name);
-			return ERR_PTR(-ENOPKG);
+			alg = ERR_PTR(-ENOPKG);
+			goto out_unlock;
 		}
 		fsverity_err(inode,
 			     "Error allocating hash algorithm \"%s\": %ld",
 			     alg->name, PTR_ERR(tfm));
-		return ERR_CAST(tfm);
+		alg = ERR_CAST(tfm);
+		goto out_unlock;
 	}
 
 	err = -EINVAL;
@@ -78,18 +87,61 @@ const struct fsverity_hash_alg *fsverity_get_hash_alg(const struct inode *inode,
 	if (WARN_ON(alg->block_size != crypto_ahash_blocksize(tfm)))
 		goto err_free_tfm;
 
+	err = mempool_init_kmalloc_pool(&alg->req_pool, 1,
+					sizeof(struct ahash_request) +
+					crypto_ahash_reqsize(tfm));
+	if (err)
+		goto err_free_tfm;
+
 	pr_info("%s using implementation \"%s\"\n",
 		alg->name, crypto_ahash_driver_name(tfm));
 
-	/* pairs with READ_ONCE() above */
-	if (cmpxchg(&alg->tfm, NULL, tfm) != NULL)
-		crypto_free_ahash(tfm);
-
-	return alg;
+	/* pairs with smp_load_acquire() above */
+	smp_store_release(&alg->tfm, tfm);
+	goto out_unlock;
 
 err_free_tfm:
 	crypto_free_ahash(tfm);
-	return ERR_PTR(err);
+	alg = ERR_PTR(err);
+out_unlock:
+	mutex_unlock(&fsverity_hash_alg_init_mutex);
+	return alg;
+}
+
+/**
+ * fsverity_alloc_hash_request() - allocate a hash request object
+ * @alg: the hash algorithm for which to allocate the request
+ * @gfp_flags: memory allocation flags
+ *
+ * This is mempool-backed, so this never fails if __GFP_DIRECT_RECLAIM is set in
+ * @gfp_flags.  However, in that case this might need to wait for all
+ * previously-allocated requests to be freed.  So to avoid deadlocks, callers
+ * must never need multiple requests at a time to make forward progress.
+ *
+ * Return: the request object on success; NULL on failure (but see above)
+ */
+struct ahash_request *fsverity_alloc_hash_request(struct fsverity_hash_alg *alg,
+						  gfp_t gfp_flags)
+{
+	struct ahash_request *req = mempool_alloc(&alg->req_pool, gfp_flags);
+
+	if (req)
+		ahash_request_set_tfm(req, alg->tfm);
+	return req;
+}
+
+/**
+ * fsverity_free_hash_request() - free a hash request object
+ * @alg: the hash algorithm
+ * @req: the hash request object to free
+ */
+void fsverity_free_hash_request(struct fsverity_hash_alg *alg,
+				struct ahash_request *req)
+{
+	if (req) {
+		ahash_request_zero(req);
+		mempool_free(req, &alg->req_pool);
+	}
 }
 
 /**
@@ -101,7 +153,7 @@ const struct fsverity_hash_alg *fsverity_get_hash_alg(const struct inode *inode,
  * Return: NULL if the salt is empty, otherwise the kmalloc()'ed precomputed
  *	   initial hash state on success or an ERR_PTR() on failure.
  */
-const u8 *fsverity_prepare_hash_state(const struct fsverity_hash_alg *alg,
+const u8 *fsverity_prepare_hash_state(struct fsverity_hash_alg *alg,
 				      const u8 *salt, size_t salt_size)
 {
 	u8 *hashstate = NULL;
@@ -119,11 +171,8 @@ const u8 *fsverity_prepare_hash_state(const struct fsverity_hash_alg *alg,
 	if (!hashstate)
 		return ERR_PTR(-ENOMEM);
 
-	req = ahash_request_alloc(alg->tfm, GFP_KERNEL);
-	if (!req) {
-		err = -ENOMEM;
-		goto err_free;
-	}
+	/* This allocation never fails, since it's mempool-backed. */
+	req = fsverity_alloc_hash_request(alg, GFP_KERNEL);
 
 	/*
 	 * Zero-pad the salt to the next multiple of the input size of the hash
@@ -158,7 +207,7 @@ const u8 *fsverity_prepare_hash_state(const struct fsverity_hash_alg *alg,
 	if (err)
 		goto err_free;
 out:
-	ahash_request_free(req);
+	fsverity_free_hash_request(alg, req);
 	kfree(padded_salt);
 	return hashstate;
 
@@ -229,7 +278,7 @@ int fsverity_hash_page(const struct merkle_tree_params *params,
  *
  * Return: 0 on success, -errno on failure
  */
-int fsverity_hash_buffer(const struct fsverity_hash_alg *alg,
+int fsverity_hash_buffer(struct fsverity_hash_alg *alg,
 			 const void *data, size_t size, u8 *out)
 {
 	struct ahash_request *req;
@@ -237,9 +286,8 @@ int fsverity_hash_buffer(const struct fsverity_hash_alg *alg,
 	DECLARE_CRYPTO_WAIT(wait);
 	int err;
 
-	req = ahash_request_alloc(alg->tfm, GFP_KERNEL);
-	if (!req)
-		return -ENOMEM;
+	/* This allocation never fails, since it's mempool-backed. */
+	req = fsverity_alloc_hash_request(alg, GFP_KERNEL);
 
 	sg_init_one(&sg, data, size);
 	ahash_request_set_callback(req, CRYPTO_TFM_REQ_MAY_SLEEP |
@@ -249,7 +297,7 @@ int fsverity_hash_buffer(const struct fsverity_hash_alg *alg,
 
 	err = crypto_wait_req(crypto_ahash_digest(req), &wait);
 
-	ahash_request_free(req);
+	fsverity_free_hash_request(alg, req);
 	return err;
 }
 

fs/verity/open.c

@@ -31,7 +31,7 @@ int fsverity_init_merkle_tree_params(struct merkle_tree_params *params,
 				     unsigned int log_blocksize,
 				     const u8 *salt, size_t salt_size)
 {
-	const struct fsverity_hash_alg *hash_alg;
+	struct fsverity_hash_alg *hash_alg;
 	int err;
 	u64 blocks;
 	u64 offset;
@@ -102,6 +102,7 @@ int fsverity_init_merkle_tree_params(struct merkle_tree_params *params,
 		/* temporarily using level_start[] to store blocks in level */
 		params->level_start[params->num_levels++] = blocks;
 	}
+	params->level0_blocks = params->level_start[0];
 
 	/* Compute the starting block of each level */
 	offset = 0;
@@ -126,7 +127,7 @@ int fsverity_init_merkle_tree_params(struct merkle_tree_params *params,
  * Compute the file measurement by hashing the fsverity_descriptor excluding the
  * signature and with the sig_size field set to 0.
  */
-static int compute_file_measurement(const struct fsverity_hash_alg *hash_alg,
+static int compute_file_measurement(struct fsverity_hash_alg *hash_alg,
 				    struct fsverity_descriptor *desc,
 				    u8 *measurement)
 {

fs/verity/verify.c

@@ -84,7 +84,8 @@ static inline int cmp_hashes(const struct fsverity_info *vi,
  * Return: true if the page is valid, else false.
  */
 static bool verify_page(struct inode *inode, const struct fsverity_info *vi,
-			struct ahash_request *req, struct page *data_page)
+			struct ahash_request *req, struct page *data_page,
+			unsigned long level0_ra_pages)
 {
 	const struct merkle_tree_params *params = &vi->tree_params;
 	const unsigned int hsize = params->digest_size;
@@ -117,8 +118,8 @@ static bool verify_page(struct inode *inode, const struct fsverity_info *vi,
 		pr_debug_ratelimited("Level %d: hindex=%lu, hoffset=%u\n",
 				     level, hindex, hoffset);
 
-		hpage = inode->i_sb->s_vop->read_merkle_tree_page(inode,
-								  hindex);
+		hpage = inode->i_sb->s_vop->read_merkle_tree_page(inode, hindex,
+				level == 0 ? level0_ra_pages : 0);
 		if (IS_ERR(hpage)) {
 			err = PTR_ERR(hpage);
 			fsverity_err(inode,
@@ -191,13 +192,12 @@ bool fsverity_verify_page(struct page *page)
 	struct ahash_request *req;
 	bool valid;
 
-	req = ahash_request_alloc(vi->tree_params.hash_alg->tfm, GFP_NOFS);
-	if (unlikely(!req))
-		return false;
+	/* This allocation never fails, since it's mempool-backed. */
+	req = fsverity_alloc_hash_request(vi->tree_params.hash_alg, GFP_NOFS);
 
-	valid = verify_page(inode, vi, req, page);
+	valid = verify_page(inode, vi, req, page, 0);
 
-	ahash_request_free(req);
+	fsverity_free_hash_request(vi->tree_params.hash_alg, req);
 
 	return valid;
 }
@@ -222,25 +222,42 @@ void fsverity_verify_bio(struct bio *bio)
 {
 	struct inode *inode = bio_first_page_all(bio)->mapping->host;
 	const struct fsverity_info *vi = inode->i_verity_info;
+	const struct merkle_tree_params *params = &vi->tree_params;
 	struct ahash_request *req;
 	struct bio_vec *bv;
 	struct bvec_iter_all iter_all;
+	unsigned long max_ra_pages = 0;
+
+	/* This allocation never fails, since it's mempool-backed. */
+	req = fsverity_alloc_hash_request(params->hash_alg, GFP_NOFS);
 
-	req = ahash_request_alloc(vi->tree_params.hash_alg->tfm, GFP_NOFS);
-	if (unlikely(!req)) {
+	if (bio->bi_opf & REQ_RAHEAD) {
+		/*
+		 * If this bio is for data readahead, then we also do readahead
+		 * of the first (largest) level of the Merkle tree.  Namely,
+		 * when a Merkle tree page is read, we also try to piggy-back on
+		 * some additional pages -- up to 1/4 the number of data pages.
+		 *
+		 * This improves sequential read performance, as it greatly
+		 * reduces the number of I/O requests made to the Merkle tree.
+		 */
 		bio_for_each_segment_all(bv, bio, iter_all)
-			SetPageError(bv->bv_page);
-		return;
+			max_ra_pages++;
+		max_ra_pages /= 4;
 	}
 
 	bio_for_each_segment_all(bv, bio, iter_all) {
 		struct page *page = bv->bv_page;
+		unsigned long level0_index = page->index >> params->log_arity;
+		unsigned long level0_ra_pages =
+			min(max_ra_pages, params->level0_blocks - level0_index);
 
-		if (!PageError(page) && !verify_page(inode, vi, req, page))
+		if (!PageError(page) &&
+		    !verify_page(inode, vi, req, page, level0_ra_pages))
 			SetPageError(page);
 	}
 
-	ahash_request_free(req);
+	fsverity_free_hash_request(params->hash_alg, req);
 }
 EXPORT_SYMBOL_GPL(fsverity_verify_bio);
 #endif /* CONFIG_BLOCK */

include/linux/fsverity.h

@@ -77,6 +77,10 @@ struct fsverity_operations {
 	 *
 	 * @inode: the inode
 	 * @index: 0-based index of the page within the Merkle tree
+	 * @num_ra_pages: The number of Merkle tree pages that should be
+	 *		  prefetched starting at @index if the page at @index
+	 *		  isn't already cached.  Implementations may ignore this
+	 *		  argument; it's only a performance optimization.
 	 *
 	 * This can be called at any time on an open verity file, as well as
 	 * between ->begin_enable_verity() and ->end_enable_verity().  It may be
@@ -87,7 +91,8 @@ struct fsverity_operations {
 	 * Return: the page on success, ERR_PTR() on failure
 	 */
 	struct page *(*read_merkle_tree_page)(struct inode *inode,
-					      pgoff_t index);
+					      pgoff_t index,
+					      unsigned long num_ra_pages);
 
 	/**
 	 * Write a Merkle tree block to the given inode.