ddt: document the theory and the key data structures

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Rob Norris <rob.norris@klarasystems.com>
Sponsored-by: Klara, Inc.
Sponsored-by: iXsystems, Inc.
Closes #15887

include/sys/ddt.h

@@ -21,6 +21,7 @@
 /*
  * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2016 by Delphix. All rights reserved.
+ * Copyright (c) 2023, Klara Inc.
  */
 
 #ifndef _SYS_DDT_H
@@ -39,10 +40,16 @@ extern "C" {
 struct abd;
 
 /*
- * On-disk DDT formats, in the desired search order (newest version first).
+ * DDT on-disk storage object types. Each one corresponds to specific
+ * implementation, see ddt_ops_t. The value itself is not stored on disk.
+ *
+ * When searching for an entry, objects types will be searched in this order.
+ *
+ * Note that DDT_TYPES is used as the "no type" for new entries that have not
+ * yet been written to a storage object.
  */
 typedef enum {
-	DDT_TYPE_ZAP = 0,
+	DDT_TYPE_ZAP = 0,	/* ZAP storage object, ddt_zap */
 	DDT_TYPES
 } ddt_type_t;
 
@@ -53,12 +60,18 @@ _Static_assert(DDT_TYPES <= UINT8_MAX,
 #define	DDT_TYPE_DEFAULT	(DDT_TYPE_ZAP)
 
 /*
- * DDT classes, in the desired search order (highest replication level first).
+ * DDT storage classes. Each class has a separate storage object for each type.
+ * The value itself is not stored on disk.
+ *
+ * When search for an entry, object classes will be searched in this order.
+ *
+ * Note that DDT_CLASSES is used as the "no class" for new entries that have not
+ * yet been written to a storage object.
  */
 typedef enum {
-	DDT_CLASS_DITTO = 0,
+	DDT_CLASS_DITTO = 0,	/* entry has ditto blocks (obsolete) */
-	DDT_CLASS_DUPLICATE,
+	DDT_CLASS_DUPLICATE,	/* entry has multiple references */
-	DDT_CLASS_UNIQUE,
+	DDT_CLASS_UNIQUE,	/* entry has a single reference */
 	DDT_CLASSES
 } ddt_class_t;
 
@@ -66,7 +79,9 @@ _Static_assert(DDT_CLASSES < UINT8_MAX,
 	"ddt_class_t must fit in a uint8_t");
 
 /*
- * On-disk ddt entry:  key (name) and physical storage (value).
+ * The "key" part of an on-disk entry. This is the unique "name" for a block,
+ * that is, that parts of the block pointer that will always be the same for
+ * the same data.
  */
 typedef struct {
 	zio_cksum_t	ddk_cksum;	/* 256-bit block checksum */
@@ -80,6 +95,10 @@ typedef struct {
 	uint64_t	ddk_prop;
 } ddt_key_t;
 
+/*
+ * Macros for accessing parts of a ddt_key_t. These are similar to their BP_*
+ * counterparts.
+ */
 #define	DDK_GET_LSIZE(ddk)	\
 	BF64_GET_SB((ddk)->ddk_prop, 0, 16, SPA_MINBLOCKSHIFT, 1)
 #define	DDK_SET_LSIZE(ddk, x)	\
@@ -96,6 +115,16 @@ typedef struct {
 #define	DDK_GET_CRYPT(ddk)		BF64_GET((ddk)->ddk_prop, 39, 1)
 #define	DDK_SET_CRYPT(ddk, x)	BF64_SET((ddk)->ddk_prop, 39, 1, x)
 
+/*
+ * The "value" part for an on-disk entry. These are the "physical"
+ * characteristics of the stored block, such as its location on disk (DVAs),
+ * birth txg and ref count.
+ *
+ * Note that an entry has an array of four ddt_phys_t, one for each number of
+ * DVAs (copies= property) and another for additional "ditto" copies. Most
+ * users of ddt_phys_t will handle indexing into or counting the phys they
+ * want.
+ */
 typedef struct {
 	dva_t		ddp_dva[SPA_DVAS_PER_BP];
 	uint64_t	ddp_refcnt;
@@ -103,6 +132,8 @@ typedef struct {
 } ddt_phys_t;
 
 /*
+ * Named indexes into the ddt_phys_t array in each entry.
+ *
  * Note, we no longer generate new DDT_PHYS_DITTO-type blocks.  However,
  * we maintain the ability to free existing dedup-ditto blocks.
  */
@@ -115,7 +146,8 @@ enum ddt_phys_type {
 };
 
 /*
- * In-core ddt entry
+ * A "live" entry, holding changes to an entry made this txg, and other data to
+ * support loading, updating and repairing the entry.
  */
 
 /* State flags for dde_flags */
@@ -123,36 +155,56 @@ enum ddt_phys_type {
 
 typedef struct {
 	/* key must be first for ddt_key_compare */
-	ddt_key_t	dde_key;
+	ddt_key_t	dde_key;			/* ddt_tree key */
-	ddt_phys_t	dde_phys[DDT_PHYS_TYPES];
+	ddt_phys_t	dde_phys[DDT_PHYS_TYPES];	/* on-disk data */
+
+	/* in-flight update IOs */
 	zio_t		*dde_lead_zio[DDT_PHYS_TYPES];
+
+	/* copy of data after a repair read, to be rewritten */
 	struct abd	*dde_repair_abd;
+
+	/* storage type and class the entry was loaded from */
 	ddt_type_t	dde_type;
 	ddt_class_t	dde_class;
-	uint8_t		dde_flags;
+
-	kcondvar_t	dde_cv;
+	uint8_t		dde_flags;	/* load state flags */
-	avl_node_t	dde_node;
+	kcondvar_t	dde_cv;		/* signaled when load completes */
+
+	avl_node_t	dde_node;	/* ddt_tree node */
 } ddt_entry_t;
 
 /*
- * In-core ddt
+ * In-core DDT object. This covers all entries and stats for a the whole pool
+ * for a given checksum type.
  */
 typedef struct {
-	kmutex_t	ddt_lock;
+	kmutex_t	ddt_lock;	/* protects changes to all fields */
-	avl_tree_t	ddt_tree;
+
-	avl_tree_t	ddt_repair_tree;
+	avl_tree_t	ddt_tree;	/* "live" (changed) entries this txg */
-	enum zio_checksum ddt_checksum;
+
-	spa_t		*ddt_spa;
+	avl_tree_t	ddt_repair_tree;	/* entries being repaired */
-	objset_t	*ddt_os;
+
-	uint64_t	ddt_stat_object;
+	enum zio_checksum ddt_checksum;		/* checksum algorithm in use */
+	spa_t		*ddt_spa;		/* pool this ddt is on */
+	objset_t	*ddt_os;		/* ddt objset (always MOS) */
+
+	/* per-type/per-class entry store objects */
 	uint64_t	ddt_object[DDT_TYPES][DDT_CLASSES];
+
+	/* object ids for whole-ddt and per-type/per-class stats */
+	uint64_t	ddt_stat_object;
+	ddt_object_t	ddt_object_stats[DDT_TYPES][DDT_CLASSES];
+
+	/* type/class stats by power-2-sized referenced blocks */
 	ddt_histogram_t	ddt_histogram[DDT_TYPES][DDT_CLASSES];
 	ddt_histogram_t	ddt_histogram_cache[DDT_TYPES][DDT_CLASSES];
-	ddt_object_t	ddt_object_stats[DDT_TYPES][DDT_CLASSES];
 } ddt_t;
 
 /*
- * In-core and on-disk bookmark for DDT walks
+ * In-core and on-disk bookmark for DDT walks. This is a cursor for ddt_walk(),
+ * and is stable across calls, even if the DDT is updated, the pool is
+ * restarted or loaded on another system, or OpenZFS is upgraded.
  */
 typedef struct {
 	uint64_t	ddb_class;

module/zfs/ddt.c

@@ -23,6 +23,7 @@
  * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2012, 2016 by Delphix. All rights reserved.
  * Copyright (c) 2022 by Pawel Jakub Dawidek
+ * Copyright (c) 2023, Klara Inc.
  */
 
 #include <sys/zfs_context.h>
@@ -39,6 +40,101 @@
 #include <sys/dsl_scan.h>
 #include <sys/abd.h>
 
+/*
+ * # DDT: Deduplication tables
+ *
+ * The dedup subsystem provides block-level deduplication. When enabled, blocks
+ * to be written will have the dedup (D) bit set, which causes them to be
+ * tracked in a "dedup table", or DDT. If a block has been seen before (exists
+ * in the DDT), instead of being written, it will instead be made to reference
+ * the existing on-disk data, and a refcount bumped in the DDT instead.
+ *
+ * ## Dedup tables and entries
+ *
+ * Conceptually, a DDT is a dictionary or map. Each entry has a "key"
+ * (ddt_key_t) made up a block's checksum and certian properties, and a "value"
+ * (one or more ddt_phys_t) containing valid DVAs for the block's data, birth
+ * time and refcount. Together these are enough to track references to a
+ * specific block, to build a valid block pointer to reference that block (for
+ * freeing, scrubbing, etc), and to fill a new block pointer with the missing
+ * pieces to make it seem like it was written.
+ *
+ * There's a single DDT (ddt_t) for each checksum type, held in spa_ddt[].
+ * Within each DDT, there can be multiple storage "types" (ddt_type_t, on-disk
+ * object data formats, each with their own implementations) and "classes"
+ * (ddt_class_t, instance of a storage type object, for entries with a specific
+ * characteristic). An entry (key) will only ever exist on one of these objects
+ * at any given time, but may be moved from one to another if their type or
+ * class changes.
+ *
+ * The DDT is driven by the write IO pipeline (zio_ddt_write()). When a block
+ * is to be written, before DVAs have been allocated, ddt_lookup() is called to
+ * see if the block has been seen before. If its not found, the write proceeds
+ * as normal, and after it succeeds, a new entry is created. If it is found, we
+ * fill the BP with the DVAs from the entry, increment the refcount and cause
+ * the write IO to return immediately.
+ *
+ * Each ddt_phys_t slot in the entry represents a separate dedup block for the
+ * same content/checksum. The slot is selected based on the zp_copies parameter
+ * the block is written with, that is, the number of DVAs in the block. The
+ * "ditto" slot (DDT_PHYS_DITTO) used to be used for now-removed "dedupditto"
+ * feature. These are no longer written, and will be freed if encountered on
+ * old pools.
+ *
+ * ## Lifetime of an entry
+ *
+ * A DDT can be enormous, and typically is not held in memory all at once.
+ * Instead, the changes to an entry are tracked in memory, and written down to
+ * disk at the end of each txg.
+ *
+ * A "live" in-memory entry (ddt_entry_t) is a node on the live tree
+ * (ddt_tree).  At the start of a txg, ddt_tree is empty. When an entry is
+ * required for IO, ddt_lookup() is called. If an entry already exists on
+ * ddt_tree, it is returned. Otherwise, a new one is created, and the
+ * type/class objects for the DDT are searched for that key. If its found, its
+ * value is copied into the live entry. If not, an empty entry is created.
+ *
+ * The live entry will be modified during the txg, usually by modifying the
+ * refcount, but sometimes by adding or updating DVAs. At the end of the txg
+ * (during spa_sync()), type and class are recalculated for entry (see
+ * ddt_sync_entry()), and the entry is written to the appropriate storage
+ * object and (if necessary), removed from an old one. ddt_tree is cleared and
+ * the next txg can start.
+ *
+ * ## Repair IO
+ *
+ * If a read on a dedup block fails, but there are other copies of the block in
+ * the other ddt_phys_t slots, reads will be issued for those instead
+ * (zio_ddt_read_start()). If one of those succeeds, the read is returned to
+ * the caller, and a copy is stashed on the entry's dde_repair_abd.
+ *
+ * During the end-of-txg sync, any entries with a dde_repair_abd get a
+ * "rewrite" write issued for the original block pointer, with the data read
+ * from the alternate block. If the block is actually damaged, this will invoke
+ * the pool's "self-healing" mechanism, and repair the block.
+ *
+ * ## Scanning (scrub/resilver)
+ *
+ * If dedup is active, the scrub machinery will walk the dedup table first, and
+ * scrub all blocks with refcnt > 1 first. After that it will move on to the
+ * regular top-down scrub, and exclude the refcnt > 1 blocks when it sees them.
+ * In this way, heavily deduplicated blocks are only scrubbed once. See the
+ * commentary on dsl_scan_ddt() for more details.
+ *
+ * Walking the DDT is done via ddt_walk(). The current position is stored in a
+ * ddt_bookmark_t, which represents a stable position in the storage object.
+ * This bookmark is stored by the scan machinery, and must reference the same
+ * position on the object even if the object changes, the pool is exported, or
+ * OpenZFS is upgraded.
+ *
+ * ## Interaction with block cloning
+ *
+ * If block cloning and dedup are both enabled on a pool, BRT will look for the
+ * dedup bit on an incoming block pointer. If set, it will call into the DDT
+ * (ddt_addref()) to add a reference to the block, instead of adding a
+ * reference to the BRT. See brt_pending_apply().
+ */
+
 /*
  * These are the only checksums valid for dedup. They must match the list
  * from dedup_table in zfs_prop.c