Add a DDT prune operation

# This is the commit message #2:

fixed block leak found by ztest

also some code cleanups

# This is the commit message #3:

Address recent review feedback

 - Don’t' treat empty DDT as an error
 - Switch to power-of-two for histogram (was 7 day bins)
 - Use seconds internally for age
 - Remove cmn_err debugging messages
 - Cleanup code comments
 - Fix cstyle and compiler warnings

# This is the commit message openzfs#4:

increase histogram bin count to cover 4 years

Signed-off-by: Don Brady <don.brady@klarasystems.com>

cmd/zpool/zpool_main.c

@@ -130,6 +130,8 @@ static int zpool_do_version(int, char **);
 
 static int zpool_do_wait(int, char **);
 
+static int zpool_do_ddt_prune(int, char **);
+
 static int zpool_do_help(int argc, char **argv);
 
 static zpool_compat_status_t zpool_do_load_compat(
@@ -167,6 +169,7 @@ typedef enum {
 	HELP_CLEAR,
 	HELP_CREATE,
 	HELP_CHECKPOINT,
+	HELP_DDT_PRUNE,
 	HELP_DESTROY,
 	HELP_DETACH,
 	HELP_EXPORT,
@@ -343,6 +346,8 @@ static zpool_command_t command_table[] = {
 	{ "sync",	zpool_do_sync,		HELP_SYNC		},
 	{ NULL },
 	{ "wait",	zpool_do_wait,		HELP_WAIT		},
+	{ NULL },
+	{ "ddtprune",	zpool_do_ddt_prune,	HELP_DDT_PRUNE		},
 };
 
 #define	NCOMMAND	(ARRAY_SIZE(command_table))
@@ -456,6 +461,8 @@ get_usage(zpool_help_t idx)
 	case HELP_WAIT:
 		return (gettext("\twait [-Hp] [-T d|u] [-t <activity>[,...]] "
 		    "<pool> [interval]\n"));
+	case HELP_DDT_PRUNE:
+		return (gettext("\tddtprune -d|-p <amount> <pool>\n"));
 	default:
 		__builtin_unreachable();
 	}
@@ -11660,6 +11667,89 @@ found:;
 	return (error);
 }
 
+/*
+ * zpool ddtprune -d|-p <amount> <pool>
+ *
+ *       -d <days>	Prune entries <days> old and older
+ *       -p <percent>	Prune <percent> amount of entries
+ *
+ * Prune entries from DDT that have only a single reference to
+ * satisfy the amount specified.
+ */
+int
+zpool_do_ddt_prune(int argc, char **argv)
+{
+	zpool_ddt_prune_unit_t unit = ZPOOL_DDT_PRUNE_NONE;
+	uint_t amount = 0;
+	zpool_handle_t *zhp;
+	char *endptr;
+	int c;
+
+	while ((c = getopt(argc, argv, "d:p:")) != -1) {
+		switch (c) {
+		case 'd':
+			if (unit == ZPOOL_DDT_PRUNE_PERCENTAGE) {
+				(void) fprintf(stderr, gettext("-d cannot be "
+				    "combined with -p option\n"));
+				usage(B_FALSE);
+			}
+			errno = 0;
+			amount = strtoull(optarg, &endptr, 0);
+			if (errno != 0 || *endptr != '\0' || amount == 0) {
+				(void) fprintf(stderr,
+				    gettext("invalid days value\n"));
+				usage(B_FALSE);
+			}
+			amount *= 86400;	/* convert days to seconds */
+			unit = ZPOOL_DDT_PRUNE_AGE;
+			break;
+		case 'p':
+			if (unit == ZPOOL_DDT_PRUNE_AGE) {
+				(void) fprintf(stderr, gettext("-p cannot be "
+				    "combined with -d option\n"));
+				usage(B_FALSE);
+			}
+			errno = 0;
+			amount = strtoull(optarg, &endptr, 0);
+			if (errno != 0 || *endptr != '\0' ||
+			    amount == 0 || amount > 100) {
+				(void) fprintf(stderr,
+				    gettext("invalid percentage value\n"));
+				usage(B_FALSE);
+			}
+			unit = ZPOOL_DDT_PRUNE_PERCENTAGE;
+			break;
+		case '?':
+			(void) fprintf(stderr, gettext("invalid option '%c'\n"),
+			    optopt);
+			usage(B_FALSE);
+		}
+	}
+	argc -= optind;
+	argv += optind;
+
+	if (unit == ZPOOL_DDT_PRUNE_NONE) {
+		(void) fprintf(stderr,
+		    gettext("missing amount option (-d|-p <value>)\n"));
+		usage(B_FALSE);
+	} else if (argc < 1) {
+		(void) fprintf(stderr, gettext("missing pool argument\n"));
+		usage(B_FALSE);
+	} else if (argc > 1) {
+		(void) fprintf(stderr, gettext("too many arguments\n"));
+		usage(B_FALSE);
+	}
+	zhp = zpool_open(g_zfs, argv[0]);
+	if (zhp == NULL)
+		return (-1);
+
+	int error = zpool_ddt_prune(zhp, unit, amount);
+
+	zpool_close(zhp);
+
+	return (error);
+}
+
 static int
 find_command_idx(const char *command, int *idx)
 {

cmd/ztest.c

@@ -446,6 +446,7 @@ ztest_func_t ztest_fletcher;
 ztest_func_t ztest_fletcher_incr;
 ztest_func_t ztest_verify_dnode_bt;
 ztest_func_t ztest_pool_prefetch_ddt;
+ztest_func_t ztest_ddt_prune;
 
 static uint64_t zopt_always = 0ULL * NANOSEC;		/* all the time */
 static uint64_t zopt_incessant = 1ULL * NANOSEC / 10;	/* every 1/10 second */
@@ -502,6 +503,7 @@ static ztest_info_t ztest_info[] = {
 	ZTI_INIT(ztest_fletcher_incr, 1, &zopt_rarely),
 	ZTI_INIT(ztest_verify_dnode_bt, 1, &zopt_sometimes),
 	ZTI_INIT(ztest_pool_prefetch_ddt, 1, &zopt_rarely),
+	ZTI_INIT(ztest_ddt_prune, 1, &zopt_rarely),
 };
 
 #define	ZTEST_FUNCS	(sizeof (ztest_info) / sizeof (ztest_info_t))
@@ -7288,6 +7290,17 @@ ztest_trim(ztest_ds_t *zd, uint64_t id)
 	mutex_exit(&ztest_vdev_lock);
 }
 
+void
+ztest_ddt_prune(ztest_ds_t *zd, uint64_t id)
+{
+	(void) zd, (void) id;
+
+	spa_t *spa = ztest_spa;
+	int32_t pct = ztest_random(15) + 1;
+
+	(void) ddt_prune_unique_entries(spa, ZPOOL_DDT_PRUNE_PERCENTAGE, pct);
+}
+
 /*
  * Verify pool integrity by running zdb.
  */

include/libzfs.h

@@ -304,6 +304,9 @@ _LIBZFS_H int zpool_reopen_one(zpool_handle_t *, void *);
 
 _LIBZFS_H int zpool_sync_one(zpool_handle_t *, void *);
 
+_LIBZFS_H int zpool_ddt_prune(zpool_handle_t *, zpool_ddt_prune_unit_t,
+    uint64_t);
+
 _LIBZFS_H int zpool_vdev_online(zpool_handle_t *, const char *, int,
     vdev_state_t *);
 _LIBZFS_H int zpool_vdev_offline(zpool_handle_t *, const char *, boolean_t);

include/libzfs_core.h

@@ -160,6 +160,9 @@ _LIBZFS_CORE_H int lzc_set_vdev_prop(const char *, nvlist_t *, nvlist_t **);
 
 _LIBZFS_CORE_H int lzc_scrub(zfs_ioc_t, const char *, nvlist_t *, nvlist_t **);
 
+_LIBZFS_CORE_H int lzc_ddt_prune(const char *, zpool_ddt_prune_unit_t,
+    uint64_t);
+
 #ifdef	__cplusplus
 }
 #endif

include/sys/ddt.h

@@ -143,6 +143,7 @@ typedef struct {
 	dva_t		ddp_dva[SPA_DVAS_PER_BP];
 	uint64_t	ddp_refcnt;
 	uint64_t	ddp_phys_birth;
+	uint64_t	ddp_class_birth;
 } ddt_phys_t;
 
 typedef struct {
@@ -172,6 +173,7 @@ typedef struct {
 #define	DDE_FLAG_LOADED		(1 << 0)	/* entry ready for use */
 #define	DDE_FLAG_OVERQUOTA	(1 << 1)	/* entry unusable, no space */
 #define	DDE_FLAG_LOGGED		(1 << 2)	/* loaded from log */
+#define	DDE_FLAG_PRUNE_WANTED	(1 << 3)	/* prune has been requested */
 
 /*
  * Additional data to support entry update or repair. This is fixed size
@@ -358,6 +360,9 @@ extern int ddt_walk(spa_t *spa, ddt_bookmark_t *ddb,
 
 extern boolean_t ddt_addref(spa_t *spa, const blkptr_t *bp);
 
+extern int ddt_prune_unique_entries(spa_t *spa, zpool_ddt_prune_unit_t unit,
+    uint64_t amount);
+
 #ifdef	__cplusplus
 }
 #endif

include/sys/fs/zfs.h

@@ -1419,7 +1419,7 @@ typedef enum {
  */
 typedef enum zfs_ioc {
 	/*
-	 * Core features - 88/128 numbers reserved.
+	 * Core features - 89/128 numbers reserved.
 	 */
 #ifdef __FreeBSD__
 	ZFS_IOC_FIRST =	0,
@@ -1516,6 +1516,7 @@ typedef enum zfs_ioc {
 	ZFS_IOC_VDEV_SET_PROPS,			/* 0x5a56 */
 	ZFS_IOC_POOL_SCRUB,			/* 0x5a57 */
 	ZFS_IOC_POOL_PREFETCH,			/* 0x5a58 */
+	ZFS_IOC_DDT_PRUNE,			/* 0x5a59 */
 
 	/*
 	 * Per-platform (Optional) - 8/128 numbers reserved.
@@ -1652,6 +1653,12 @@ typedef enum {
 	ZPOOL_PREFETCH_DDT
 } zpool_prefetch_type_t;
 
+typedef enum {
+	ZPOOL_DDT_PRUNE_NONE,
+	ZPOOL_DDT_PRUNE_AGE,		/* in seconds */
+	ZPOOL_DDT_PRUNE_PERCENTAGE,	/* 1 - 100 */
+} zpool_ddt_prune_unit_t;
+
 /*
  * Bookmark name values.
  */
@@ -1745,6 +1752,12 @@ typedef enum {
  */
 #define	ZPOOL_PREFETCH_TYPE		"prefetch_type"
 
+/*
+ * The following are names used when invoking ZFS_IOC_DDT_PRUNE.
+ */
+#define	DDT_PRUNE_UNIT		"ddt_prune_unit"
+#define	DDT_PRUNE_AMOUNT	"ddt_prune_amount"
+
 /*
  * Flags for ZFS_IOC_VDEV_SET_STATE
  */

include/sys/spa_impl.h

@@ -411,6 +411,7 @@ struct spa {
 	uint64_t	spa_dedup_dspace;	/* Cache get_dedup_dspace() */
 	uint64_t	spa_dedup_checksum;	/* default dedup checksum */
 	uint64_t	spa_dspace;		/* dspace in normal class */
+	boolean_t	spa_active_ddt_prune;	/* ddt prune process active */
 	struct brt	*spa_brt;		/* in-core BRT */
 	kmutex_t	spa_vdev_top_lock;	/* dueling offline/remove */
 	kmutex_t	spa_proc_lock;		/* protects spa_proc* */

lib/libzfs/libzfs_pool.c

@@ -5549,3 +5549,31 @@ zpool_set_vdev_prop(zpool_handle_t *zhp, const char *vdevname,
 
 	return (ret);
 }
+
+/*
+ *
+ */
+int
+zpool_ddt_prune(zpool_handle_t *zhp, zpool_ddt_prune_unit_t unit,
+    uint64_t amount)
+{
+	int error = lzc_ddt_prune(zhp->zpool_name, unit, amount);
+	if (error != 0 && error != ENODATA) {
+		libzfs_handle_t *hdl = zhp->zpool_hdl;
+		char errbuf[ERRBUFLEN];
+
+		(void) snprintf(errbuf, sizeof (errbuf), dgettext(TEXT_DOMAIN,
+		    "cannot prune dedup table on '%s'"), zhp->zpool_name);
+
+		if (error == EALREADY) {
+			zfs_error_aux(hdl, dgettext(TEXT_DOMAIN,
+			    "a prune operation is already in progress"));
+			(void) zfs_error(hdl, EZFS_BUSY, errbuf);
+		} else {
+			(void) zpool_standard_error(hdl, errno, errbuf);
+		}
+		return (-1);
+	}
+
+	return (0);
+}

lib/libzfs_core/libzfs_core.c

@@ -1921,3 +1921,25 @@ lzc_get_bootenv(const char *pool, nvlist_t **outnvl)
 {
 	return (lzc_ioctl(ZFS_IOC_GET_BOOTENV, pool, NULL, outnvl));
 }
+
+/*
+ * Prune the specified amount from the pool's dedup table.
+ */
+int
+lzc_ddt_prune(const char *pool, zpool_ddt_prune_unit_t unit, uint64_t amount)
+{
+	int error;
+
+	nvlist_t *result = NULL;
+	nvlist_t *args = fnvlist_alloc();
+
+	fnvlist_add_int32(args, DDT_PRUNE_UNIT, unit);
+	fnvlist_add_uint64(args, DDT_PRUNE_AMOUNT, amount);
+
+	error = lzc_ioctl(ZFS_IOC_DDT_PRUNE, pool, args, &result);
+
+	fnvlist_free(args);
+	fnvlist_free(result);
+
+	return (error);
+}