[nrf fromlist] zms: Initial support for 64 bit IDs

Upstream PR #: 94330

Allow the ZMS API to optionally accept 64 bit IDs. A typedef `zms_id_t`
is added, so that the maximum ID width can be controlled using Kconfig.

The current ATE structure is already large enough that it is possible to
reserve 64 bits for IDs without increasing its total size (128 bits).
This makes the feature a natural, low footprint alternative to Settings,
for cases where the supported key namespace must be larger than 32 bit
but not arbitrarily large.

The ATE format does have to be altered to accommodate larger IDs, but
the default "32 bit" format is left as is. Now, the `struct zms_ate`
describes one of two supported formats, selected by an `#if` condition.
In the future, it may be possible to support multiple ATE formats at
runtime, in which case the structure can be turned into a union.

In the new, "64 bit" ATEs, the `offset` and `metadata` fields are moved
into a union, because they are found to be mutually exclusive. With the
old format, the same fields are in different locations, but one of them
always gets filled with a dummy value, depending on the given ATE type.
To cover both cases, a `memset` is used, which should be optimized away
by the compiler when appropriate.

The only limitation is that the new ATE format has no room for data CRC,
but an alternative integrity check can be implemented by the caller.

Signed-off-by: Grzegorz Swiderski <[email protected]>
(cherry picked from commit 019a9eb)

Author: 57300

include/zephyr/fs/zms.h

@@ -77,6 +77,17 @@ struct zms_fs {
  * @{
  */
 
+/**
+ * @brief ID type used in the ZMS API.
+ *
+ * @note The width of this type depends on @kconfig{CONFIG_ZMS_ID_64BIT}.
+ */
+#if CONFIG_ZMS_ID_64BIT
+typedef uint64_t zms_id_t;
+#else
+typedef uint32_t zms_id_t;
+#endif
+
 /**
  * @brief Mount a ZMS file system onto the device specified in `fs`.
  *
@@ -128,7 +139,7 @@ int zms_clear(struct zms_fs *fs);
  * @retval -EINVAL if `fs` is NULL or `len` is invalid.
  * @retval -ENOSPC if no space is left on the device.
  */
-ssize_t zms_write(struct zms_fs *fs, uint32_t id, const void *data, size_t len);
+ssize_t zms_write(struct zms_fs *fs, zms_id_t id, const void *data, size_t len);
 
 /**
  * @brief Delete an entry from the file system
@@ -142,7 +153,7 @@ ssize_t zms_write(struct zms_fs *fs, uint32_t id, const void *data, size_t len);
  * @retval -EIO if there is a memory read/write error.
  * @retval -EINVAL if `fs` is NULL.
  */
-int zms_delete(struct zms_fs *fs, uint32_t id);
+int zms_delete(struct zms_fs *fs, zms_id_t id);
 
 /**
  * @brief Read an entry from the file system.
@@ -161,7 +172,7 @@ int zms_delete(struct zms_fs *fs, uint32_t id);
  * @retval -ENOENT if there is no entry with the given `id`.
  * @retval -EINVAL if `fs` is NULL.
  */
-ssize_t zms_read(struct zms_fs *fs, uint32_t id, void *data, size_t len);
+ssize_t zms_read(struct zms_fs *fs, zms_id_t id, void *data, size_t len);
 
 /**
  * @brief Read a history entry from the file system.
@@ -182,7 +193,7 @@ ssize_t zms_read(struct zms_fs *fs, uint32_t id, void *data, size_t len);
  * @retval -ENOENT if there is no entry with the given `id` and history counter.
  * @retval -EINVAL if `fs` is NULL.
  */
-ssize_t zms_read_hist(struct zms_fs *fs, uint32_t id, void *data, size_t len, uint32_t cnt);
+ssize_t zms_read_hist(struct zms_fs *fs, zms_id_t id, void *data, size_t len, uint32_t cnt);
 
 /**
  * @brief Gets the length of the data that is stored in an entry with a given `id`
@@ -198,7 +209,7 @@ ssize_t zms_read_hist(struct zms_fs *fs, uint32_t id, void *data, size_t len, ui
  * @retval -ENOENT if there is no entry with the given id.
  * @retval -EINVAL if `fs` is NULL.
  */
-ssize_t zms_get_data_length(struct zms_fs *fs, uint32_t id);
+ssize_t zms_get_data_length(struct zms_fs *fs, zms_id_t id);
 
 /**
  * @brief Calculate the available free space in the file system.

subsys/fs/zms/Kconfig

@@ -16,6 +16,15 @@ config ZMS
 
 if ZMS
 
+config ZMS_ID_64BIT
+	bool "64 bit ZMS IDs"
+	help
+	  When this option is true, the `zms_id_t` values passed to ZMS APIs will be 64 bit,
+	  as opposed to the default 32 bit.
+	  This option will also change the format of allocation table entries (ATEs) in memory
+	  to accommodate larger IDs. Currently, this will make ZMS unable to mount an existing
+	  file system if it has been initialized with a different ATE format.
+
 config ZMS_LOOKUP_CACHE
 	bool "ZMS lookup cache"
 	help
@@ -34,6 +43,7 @@ config ZMS_LOOKUP_CACHE_SIZE
 
 config ZMS_DATA_CRC
 	bool "ZMS data CRC"
+	depends on !ZMS_ID_64BIT
 
 config ZMS_CUSTOMIZE_BLOCK_SIZE
 	bool "Customize the size of the buffer used internally for reads and writes"
@@ -54,6 +64,7 @@ config ZMS_CUSTOM_BLOCK_SIZE
 config ZMS_LOOKUP_CACHE_FOR_SETTINGS
 	bool "ZMS Storage lookup cache optimized for settings"
 	depends on ZMS_LOOKUP_CACHE && SETTINGS_ZMS
+	depends on !ZMS_ID_64BIT
 	help
 	  Enable usage of lookup cache based on hashes to get, the best ZMS performance,
 	  provided that the ZMS is used only for the purpose of providing the settings

subsys/fs/zms/zms.c

@@ -29,10 +29,8 @@ static int zms_ate_valid_different_sector(struct zms_fs *fs, const struct zms_at
 
 #ifdef CONFIG_ZMS_LOOKUP_CACHE
 
-static inline size_t zms_lookup_cache_pos(uint32_t id)
+static inline size_t zms_lookup_cache_pos(zms_id_t id)
 {
-	uint32_t hash = id;
-
 #ifdef CONFIG_ZMS_LOOKUP_CACHE_FOR_SETTINGS
 	/*
 	 * 1. Settings subsystem is storing the name ID and the linked list node ID
@@ -52,14 +50,27 @@ static inline size_t zms_lookup_cache_pos(uint32_t id)
 	uint32_t key_value_bit;
 	uint32_t key_value_hash;
 	uint32_t key_value_ll;
+	uint32_t hash;
 
 	key_value_bit = (id >> LOG2(ZMS_DATA_ID_OFFSET)) & 1;
 	key_value_hash = (id & ZMS_HASH_MASK) >> (CONFIG_SETTINGS_ZMS_MAX_COLLISIONS_BITS + 1);
 	key_value_ll = id & BIT(0);
 
 	hash = (key_value_hash << 2) | (key_value_bit << 1) | key_value_ll;
+
+#elif defined(CONFIG_ZMS_ID_64BIT)
+	/* 64-bit integer hash function found by https://github.com/skeeto/hash-prospector. */
+	uint64_t hash = id;
+
+	hash ^= hash >> 32;
+	hash *= 0x42ab4abe4c475039ULL;
+	hash ^= hash >> 31;
+	hash *= 0xfa90c4424c537791ULL;
+	hash ^= hash >> 32;
 #else
 	/* 32-bit integer hash function found by https://github.com/skeeto/hash-prospector. */
+	uint32_t hash = id;
+
 	hash ^= hash >> 16;
 	hash *= 0x7feb352dU;
 	hash ^= hash >> 15;
@@ -239,7 +250,7 @@ static int zms_flash_ate_wrt(struct zms_fs *fs, const struct zms_ate *entry)
 		goto end;
 	}
 #ifdef CONFIG_ZMS_LOOKUP_CACHE
-	/* 0xFFFFFFFF is a special-purpose identifier. Exclude it from the cache */
+	/* ZMS_HEAD_ID is a special-purpose identifier. Exclude it from the cache */
 	if (entry->id != ZMS_HEAD_ID) {
 		fs->lookup_cache[zms_lookup_cache_pos(entry->id)] = fs->ate_wra;
 	}
@@ -487,7 +498,7 @@ static bool zms_close_ate_valid(struct zms_fs *fs, const struct zms_ate *entry)
 /* zms_empty_ate_valid validates an sector empty ate.
  * A valid sector empty ate should be:
  * - a valid ate
- * - with len = 0xffff and id = 0xffffffff
+ * - with len = 0xffff and id = ZMS_HEAD_ID
  * return true if valid, false otherwise
  */
 static bool zms_empty_ate_valid(struct zms_fs *fs, const struct zms_ate *entry)
@@ -500,7 +511,7 @@ static bool zms_empty_ate_valid(struct zms_fs *fs, const struct zms_ate *entry)
  * Valid gc_done_ate:
  * - valid ate
  * - len = 0
- * - id = 0xffffffff
+ * - id = ZMS_HEAD_ID
  * return true if valid, false otherwise
  */
 static bool zms_gc_done_ate_valid(struct zms_fs *fs, const struct zms_ate *entry)
@@ -509,6 +520,18 @@ static bool zms_gc_done_ate_valid(struct zms_fs *fs, const struct zms_ate *entry
 		(entry->id == ZMS_HEAD_ID));
 }
 
+/* zms_sector_closed checks whether the current sector is closed, which would imply
+ * that the empty ATE and close ATE are both valid and have matching cycle counters
+ *
+ * return true if closed, false otherwise
+ */
+static bool zms_sector_closed(struct zms_fs *fs, struct zms_ate *empty_ate,
+			      struct zms_ate *close_ate)
+{
+	return (zms_empty_ate_valid(fs, empty_ate) && zms_close_ate_valid(fs, close_ate) &&
+		(empty_ate->cycle_cnt == close_ate->cycle_cnt));
+}
+
 /* Read empty and close ATE of the sector where belongs address "addr" and
  * validates that the sector is closed.
  * retval: 0 if sector is not close
@@ -526,8 +549,7 @@ static int zms_validate_closed_sector(struct zms_fs *fs, uint64_t addr, struct z
 		return rc;
 	}
 
-	if (zms_empty_ate_valid(fs, empty_ate) && zms_close_ate_valid(fs, close_ate) &&
-	    (empty_ate->cycle_cnt == close_ate->cycle_cnt)) {
+	if (zms_sector_closed(fs, empty_ate, close_ate)) {
 		/* Closed sector validated */
 		return 1;
 	}
@@ -536,7 +558,7 @@ static int zms_validate_closed_sector(struct zms_fs *fs, uint64_t addr, struct z
 }
 
 /* store an entry in flash */
-static int zms_flash_write_entry(struct zms_fs *fs, uint32_t id, const void *data, size_t len)
+static int zms_flash_write_entry(struct zms_fs *fs, zms_id_t id, const void *data, size_t len)
 {
 	int rc;
 	struct zms_ate entry;
@@ -549,13 +571,13 @@ static int zms_flash_write_entry(struct zms_fs *fs, uint32_t id, const void *dat
 	entry.cycle_cnt = fs->sector_cycle;
 
 	if (len > ZMS_DATA_IN_ATE_SIZE) {
-		/* only compute CRC if len is greater than 8 bytes */
-		if (IS_ENABLED(CONFIG_ZMS_DATA_CRC)) {
-			entry.data_crc = crc32_ieee(data, len);
-		}
+#ifdef CONFIG_ZMS_DATA_CRC
+		/* only compute CRC if data is to be stored outside of entry */
+		entry.data_crc = crc32_ieee(data, len);
+#endif
 		entry.offset = (uint32_t)SECTOR_OFFSET(fs->data_wra);
 	} else if ((len > 0) && (len <= ZMS_DATA_IN_ATE_SIZE)) {
-		/* Copy data into entry for small data ( < 8B) */
+		/* Copy data into entry for small data (at most ZMS_DATA_IN_ATE_SIZE bytes) */
 		memcpy(&entry.data, data, len);
 	}
 
@@ -688,10 +710,12 @@ static int zms_sector_close(struct zms_fs *fs)
 	struct zms_ate close_ate;
 	struct zms_ate garbage_ate;
 
+	/* Initialize all members to 0xff */
+	memset(&close_ate, 0xff, sizeof(struct zms_ate));
+
 	close_ate.id = ZMS_HEAD_ID;
 	close_ate.len = 0U;
 	close_ate.offset = (uint32_t)SECTOR_OFFSET(fs->ate_wra + fs->ate_size);
-	close_ate.metadata = 0xffffffff;
 	close_ate.cycle_cnt = fs->sector_cycle;
 
 	/* When we close the sector, we must write all non used ATE with
@@ -740,11 +764,13 @@ static int zms_add_gc_done_ate(struct zms_fs *fs)
 {
 	struct zms_ate gc_done_ate;
 
+	/* Initialize all members to 0xff */
+	memset(&gc_done_ate, 0xff, sizeof(struct zms_ate));
+
 	LOG_DBG("Adding gc done ate at %llx", fs->ate_wra);
 	gc_done_ate.id = ZMS_HEAD_ID;
 	gc_done_ate.len = 0U;
 	gc_done_ate.offset = (uint32_t)SECTOR_OFFSET(fs->data_wra);
-	gc_done_ate.metadata = 0xffffffff;
 	gc_done_ate.cycle_cnt = fs->sector_cycle;
 
 	zms_ate_crc8_update(&gc_done_ate);
@@ -793,14 +819,17 @@ static int zms_add_empty_ate(struct zms_fs *fs, uint64_t addr)
 	int rc = 0;
 	uint64_t previous_ate_wra;
 
+	/* Initialize all members to 0 */
+	memset(&empty_ate, 0, sizeof(struct zms_ate));
+
 	addr &= ADDR_SECT_MASK;
 
 	LOG_DBG("Adding empty ate at %llx", (uint64_t)(addr + fs->sector_size - fs->ate_size));
 	empty_ate.id = ZMS_HEAD_ID;
 	empty_ate.len = 0xffff;
-	empty_ate.offset = 0U;
-	empty_ate.metadata =
-		FIELD_PREP(ZMS_MAGIC_NUMBER_MASK, ZMS_MAGIC_NUMBER) | ZMS_DEFAULT_VERSION;
+	empty_ate.metadata = FIELD_PREP(ZMS_VERSION_MASK, ZMS_DEFAULT_VERSION) |
+			     FIELD_PREP(ZMS_MAGIC_NUMBER_MASK, ZMS_MAGIC_NUMBER) |
+			     FIELD_PREP(ZMS_ATE_FORMAT_MASK, ZMS_DEFAULT_ATE_FORMAT);
 
 	rc = zms_get_sector_cycle(fs, addr, &cycle_cnt);
 	if (rc == -ENOENT) {
@@ -893,7 +922,7 @@ static int zms_get_sector_header(struct zms_fs *fs, uint64_t addr, struct zms_at
  * @retval 1 valid ATE with same ID found
  * @retval < 0 An error happened
  */
-static int zms_find_ate_with_id(struct zms_fs *fs, uint32_t id, uint64_t start_addr,
+static int zms_find_ate_with_id(struct zms_fs *fs, zms_id_t id, uint64_t start_addr,
 				uint64_t end_addr, struct zms_ate *ate, uint64_t *ate_addr)
 {
 	int rc;
@@ -1044,10 +1073,10 @@ static int zms_gc(struct zms_fs *fs)
 		 */
 		if (wlk_prev_addr == gc_prev_addr) {
 			/* copy needed */
-			LOG_DBG("Moving %d, len %d", gc_ate.id, gc_ate.len);
+			LOG_DBG("Moving %lld, len %d", (long long)gc_ate.id, gc_ate.len);
 
 			if (gc_ate.len > ZMS_DATA_IN_ATE_SIZE) {
-				/* Copy Data only when len > 8
+				/* Copy Data only when len > ZMS_DATA_IN_ATE_SIZE
 				 * Otherwise, Data is already inside ATE
 				 */
 				data_addr = (gc_prev_addr & ADDR_SECT_MASK);
@@ -1156,16 +1185,28 @@ static int zms_init(struct zms_fs *fs)
 	for (i = 0; i < fs->sector_count; i++) {
 		addr = zms_close_ate_addr(fs, ((uint64_t)i << ADDR_SECT_SHIFT));
 
-		/* verify if the sector is closed */
-		sec_closed = zms_validate_closed_sector(fs, addr, &empty_ate, &close_ate);
-		if (sec_closed < 0) {
-			rc = sec_closed;
+		/* read the header ATEs */
+		rc = zms_get_sector_header(fs, addr, &empty_ate, &close_ate);
+		if (rc) {
 			goto end;
 		}
 		/* update cycle count */
 		fs->sector_cycle = empty_ate.cycle_cnt;
 
-		if (sec_closed == 1) {
+		/* Check the ATE format indicator so that we know how to validate ATEs.
+		 * The metadata field has the same offset and size in all ATE formats
+		 * (the same is guaranteed for crc8 and cycle_cnt).
+		 * Currently, ZMS can only recognize one of its supported ATE formats
+		 * (the one chosen at build time), so their indicators are defined for
+		 * the possibility of a future extension.
+		 * If this indicator is unknown, then consider the header ATEs invalid,
+		 * because we might not be dealing with ZMS contents at all.
+		 */
+		if (ZMS_GET_ATE_FORMAT(empty_ate.metadata) != ZMS_DEFAULT_ATE_FORMAT) {
+			continue;
+		}
+
+		if (zms_sector_closed(fs, &empty_ate, &close_ate)) {
 			/* closed sector */
 			closed_sectors++;
 			/* Let's verify that this is a ZMS storage system */
@@ -1223,7 +1264,8 @@ static int zms_init(struct zms_fs *fs)
 			goto end;
 		}
 
-		if (zms_empty_ate_valid(fs, &empty_ate)) {
+		if ((ZMS_GET_ATE_FORMAT(empty_ate.metadata) == ZMS_DEFAULT_ATE_FORMAT) &&
+		    zms_empty_ate_valid(fs, &empty_ate)) {
 			/* Empty ATE is valid, let's verify that this is a ZMS storage system */
 			if (ZMS_GET_MAGIC_NUMBER(empty_ate.metadata) == ZMS_MAGIC_NUMBER) {
 				zms_magic_exist = true;
@@ -1468,7 +1510,7 @@ int zms_mount(struct zms_fs *fs)
 	return 0;
 }
 
-ssize_t zms_write(struct zms_fs *fs, uint32_t id, const void *data, size_t len)
+ssize_t zms_write(struct zms_fs *fs, zms_id_t id, const void *data, size_t len)
 {
 	int rc;
 	size_t data_size;
@@ -1613,12 +1655,12 @@ ssize_t zms_write(struct zms_fs *fs, uint32_t id, const void *data, size_t len)
 	return rc;
 }
 
-int zms_delete(struct zms_fs *fs, uint32_t id)
+int zms_delete(struct zms_fs *fs, zms_id_t id)
 {
 	return zms_write(fs, id, NULL, 0);
 }
 
-ssize_t zms_read_hist(struct zms_fs *fs, uint32_t id, void *data, size_t len, uint32_t cnt)
+ssize_t zms_read_hist(struct zms_fs *fs, zms_id_t id, void *data, size_t len, uint32_t cnt)
 {
 	int rc;
 	int prev_found = 0;
@@ -1719,7 +1761,7 @@ ssize_t zms_read_hist(struct zms_fs *fs, uint32_t id, void *data, size_t len, ui
 	return rc;
 }
 
-ssize_t zms_read(struct zms_fs *fs, uint32_t id, void *data, size_t len)
+ssize_t zms_read(struct zms_fs *fs, zms_id_t id, void *data, size_t len)
 {
 	int rc;
 
@@ -1732,7 +1774,7 @@ ssize_t zms_read(struct zms_fs *fs, uint32_t id, void *data, size_t len)
 	return MIN(rc, len);
 }
 
-ssize_t zms_get_data_length(struct zms_fs *fs, uint32_t id)
+ssize_t zms_get_data_length(struct zms_fs *fs, zms_id_t id)
 {
 	int rc;