*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
* Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
*/
#include <sys/refcount.h>
#include <sys/dmu_zfetch.h>
#include <sys/zrlock.h>
+#include <sys/multilist.h>
#ifdef __cplusplus
extern "C" {
* Fixed constants.
*/
#define DNODE_SHIFT 9 /* 512 bytes */
-#define DN_MIN_INDBLKSHIFT 10 /* 1k */
-#define DN_MAX_INDBLKSHIFT 14 /* 16k */
+#define DN_MIN_INDBLKSHIFT 12 /* 4k */
+/*
+ * If we ever increase this value beyond 20, we need to revisit all logic that
+ * does x << level * ebps to handle overflow. With a 1M indirect block size,
+ * 4 levels of indirect blocks would not be able to guarantee addressing an
+ * entire object, so 5 levels will be used, but 5 * (20 - 7) = 65.
+ */
+#define DN_MAX_INDBLKSHIFT 17 /* 128k */
#define DNODE_BLOCK_SHIFT 14 /* 16k */
#define DNODE_CORE_SIZE 64 /* 64 bytes for dnode sans blkptrs */
#define DN_MAX_OBJECT_SHIFT 48 /* 256 trillion (zfs_fid_t limit) */
/*
* Derived constants.
*/
-#define DNODE_SIZE (1 << DNODE_SHIFT)
-#define DN_MAX_NBLKPTR ((DNODE_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT)
-#define DN_MAX_BONUSLEN (DNODE_SIZE - DNODE_CORE_SIZE - (1 << SPA_BLKPTRSHIFT))
+#define DNODE_MIN_SIZE (1 << DNODE_SHIFT)
+#define DNODE_MAX_SIZE (1 << DNODE_BLOCK_SHIFT)
+#define DNODE_BLOCK_SIZE (1 << DNODE_BLOCK_SHIFT)
+#define DNODE_MIN_SLOTS (DNODE_MIN_SIZE >> DNODE_SHIFT)
+#define DNODE_MAX_SLOTS (DNODE_MAX_SIZE >> DNODE_SHIFT)
+#define DN_BONUS_SIZE(dnsize) ((dnsize) - DNODE_CORE_SIZE - \
+ (1 << SPA_BLKPTRSHIFT))
+#define DN_SLOTS_TO_BONUSLEN(slots) DN_BONUS_SIZE((slots) << DNODE_SHIFT)
+#define DN_OLD_MAX_BONUSLEN (DN_BONUS_SIZE(DNODE_MIN_SIZE))
+#define DN_MAX_NBLKPTR ((DNODE_MIN_SIZE - DNODE_CORE_SIZE) >> SPA_BLKPTRSHIFT)
#define DN_MAX_OBJECT (1ULL << DN_MAX_OBJECT_SHIFT)
-#define DN_ZERO_BONUSLEN (DN_MAX_BONUSLEN + 1)
+#define DN_ZERO_BONUSLEN (DN_BONUS_SIZE(DNODE_MAX_SIZE) + 1)
#define DN_KILL_SPILLBLK (1)
+#define DN_SLOT_UNINIT ((void *)NULL) /* Uninitialized */
+#define DN_SLOT_FREE ((void *)1UL) /* Free slot */
+#define DN_SLOT_ALLOCATED ((void *)2UL) /* Allocated slot */
+#define DN_SLOT_INTERIOR ((void *)3UL) /* Interior allocated slot */
+#define DN_SLOT_IS_PTR(dn) ((void *)dn > DN_SLOT_INTERIOR)
+#define DN_SLOT_IS_VALID(dn) ((void *)dn != NULL)
+
#define DNODES_PER_BLOCK_SHIFT (DNODE_BLOCK_SHIFT - DNODE_SHIFT)
#define DNODES_PER_BLOCK (1ULL << DNODES_PER_BLOCK_SHIFT)
+
+/*
+ * This is inaccurate if the indblkshift of the particular object is not the
+ * max. But it's only used by userland to calculate the zvol reservation.
+ */
#define DNODES_PER_LEVEL_SHIFT (DN_MAX_INDBLKSHIFT - SPA_BLKPTRSHIFT)
#define DNODES_PER_LEVEL (1ULL << DNODES_PER_LEVEL_SHIFT)
-/* The +2 here is a cheesy way to round up */
-#define DN_MAX_LEVELS (2 + ((DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT) / \
- (DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT)))
+#define DN_MAX_LEVELS (DIV_ROUND_UP(DN_MAX_OFFSET_SHIFT - SPA_MINBLOCKSHIFT, \
+ DN_MIN_INDBLKSHIFT - SPA_BLKPTRSHIFT) + 1)
#define DN_BONUS(dnp) ((void*)((dnp)->dn_bonus + \
(((dnp)->dn_nblkptr - 1) * sizeof (blkptr_t))))
};
/* Is dn_used in bytes? if not, it's in multiples of SPA_MINBLOCKSIZE */
-#define DNODE_FLAG_USED_BYTES (1<<0)
-#define DNODE_FLAG_USERUSED_ACCOUNTED (1<<1)
+#define DNODE_FLAG_USED_BYTES (1 << 0)
+#define DNODE_FLAG_USERUSED_ACCOUNTED (1 << 1)
/* Does dnode have a SA spill blkptr in bonus? */
-#define DNODE_FLAG_SPILL_BLKPTR (1<<2)
+#define DNODE_FLAG_SPILL_BLKPTR (1 << 2)
+
+/* User/Group dnode accounting */
+#define DNODE_FLAG_USEROBJUSED_ACCOUNTED (1 << 3)
typedef struct dnode_phys {
uint8_t dn_type; /* dmu_object_type_t */
uint8_t dn_flags; /* DNODE_FLAG_* */
uint16_t dn_datablkszsec; /* data block size in 512b sectors */
uint16_t dn_bonuslen; /* length of dn_bonus */
- uint8_t dn_pad2[4];
+ uint8_t dn_extra_slots; /* # of subsequent slots consumed */
+ uint8_t dn_pad2[3];
/* accounting is protected by dn_dirty_mtx */
uint64_t dn_maxblkid; /* largest allocated block ID */
uint64_t dn_pad3[4];
/*
- * The tail region is 448 bytes, and there are three ways to
- * look at it.
+ * The tail region is 448 bytes for a 512 byte dnode, and
+ * correspondingly larger for larger dnode sizes. The spill
+ * block pointer, when present, is always at the end of the tail
+ * region. There are three ways this space may be used, using
+ * a 512 byte dnode for this diagram:
*
* 0 64 128 192 256 320 384 448 (offset)
* +---------------+---------------+---------------+-------+
* +---------------+---------------+---------------+-------+
* | dn_blkptr[0] | dn_bonus[0..319] |
* +---------------+-----------------------+---------------+
- * | dn_blkptr[0] | / | dn_spill |
+ * | dn_blkptr[0] | dn_bonus[0..191] | dn_spill |
* +---------------+-----------------------+---------------+
*/
union {
- blkptr_t dn_blkptr[1+DN_MAX_BONUSLEN/sizeof (blkptr_t)];
+ blkptr_t dn_blkptr[1+DN_OLD_MAX_BONUSLEN/sizeof (blkptr_t)];
struct {
blkptr_t __dn_ignore1;
- uint8_t dn_bonus[DN_MAX_BONUSLEN];
+ uint8_t dn_bonus[DN_OLD_MAX_BONUSLEN];
};
struct {
blkptr_t __dn_ignore2;
- uint8_t __dn_ignore3[DN_MAX_BONUSLEN-sizeof (blkptr_t)];
+ uint8_t __dn_ignore3[DN_OLD_MAX_BONUSLEN -
+ sizeof (blkptr_t)];
blkptr_t dn_spill;
};
};
} dnode_phys_t;
-typedef struct dnode {
+#define DN_SPILL_BLKPTR(dnp) (blkptr_t *)((char *)(dnp) + \
+ (((dnp)->dn_extra_slots + 1) << DNODE_SHIFT) - (1 << SPA_BLKPTRSHIFT))
+
+struct dnode {
/*
* Protects the structure of the dnode, including the number of levels
* of indirection (dn_nlevels), dn_maxblkid, and dn_next_*
uint32_t dn_datablksz; /* in bytes */
uint64_t dn_maxblkid;
uint8_t dn_next_type[TXG_SIZE];
+ uint8_t dn_num_slots; /* metadnode slots consumed on disk */
uint8_t dn_next_nblkptr[TXG_SIZE];
uint8_t dn_next_nlevels[TXG_SIZE];
uint8_t dn_next_indblkshift[TXG_SIZE];
/* protected by dn_dbufs_mtx; declared here to fill 32-bit hole */
uint32_t dn_dbufs_count; /* count of dn_dbufs */
- /* There are no level-0 blocks of this blkid or higher in dn_dbufs */
- uint64_t dn_unlisted_l0_blkid;
/* protected by os_lock: */
- list_node_t dn_dirty_link[TXG_SIZE]; /* next on dataset's dirty */
+ multilist_node_t dn_dirty_link[TXG_SIZE]; /* next on dataset's dirty */
/* protected by dn_mtx: */
kmutex_t dn_mtx;
* duplicate entries, we order the dbufs by an arbitrary value -
* their address in memory. This means that dn_dbufs cannot be used to
* directly look up a dbuf. Instead, callers must use avl_walk, have
- * a reference to the dbuf, or look up a non-existant node with
+ * a reference to the dbuf, or look up a non-existent node with
* db_state = DB_SEARCH (see dbuf_free_range for an example).
*/
avl_tree_t dn_dbufs;
/* holds prefetch structure */
struct zfetch dn_zfetch;
-} dnode_t;
+};
/*
* Adds a level of indirection between the dbuf and the dnode to avoid
int dnode_hold(struct objset *dd, uint64_t object,
void *ref, dnode_t **dnp);
-int dnode_hold_impl(struct objset *dd, uint64_t object, int flag,
+int dnode_hold_impl(struct objset *dd, uint64_t object, int flag, int dn_slots,
void *ref, dnode_t **dnp);
boolean_t dnode_add_ref(dnode_t *dn, void *ref);
void dnode_rele(dnode_t *dn, void *ref);
void dnode_setdirty(dnode_t *dn, dmu_tx_t *tx);
void dnode_sync(dnode_t *dn, dmu_tx_t *tx);
void dnode_allocate(dnode_t *dn, dmu_object_type_t ot, int blocksize, int ibs,
- dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
+ dmu_object_type_t bonustype, int bonuslen, int dn_slots, dmu_tx_t *tx);
void dnode_reallocate(dnode_t *dn, dmu_object_type_t ot, int blocksize,
- dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx);
+ dmu_object_type_t bonustype, int bonuslen, int dn_slots, dmu_tx_t *tx);
void dnode_free(dnode_t *dn, dmu_tx_t *tx);
void dnode_byteswap(dnode_phys_t *dnp);
void dnode_buf_byteswap(void *buf, size_t size);
int dnode_set_blksz(dnode_t *dn, uint64_t size, int ibs, dmu_tx_t *tx);
void dnode_free_range(dnode_t *dn, uint64_t off, uint64_t len, dmu_tx_t *tx);
void dnode_diduse_space(dnode_t *dn, int64_t space);
-void dnode_willuse_space(dnode_t *dn, int64_t space, dmu_tx_t *tx);
void dnode_new_blkid(dnode_t *dn, uint64_t blkid, dmu_tx_t *tx, boolean_t);
uint64_t dnode_block_freed(dnode_t *dn, uint64_t blkid);
void dnode_init(void);
void dnode_evict_dbufs(dnode_t *dn);
void dnode_evict_bonus(dnode_t *dn);
+#define DNODE_IS_CACHEABLE(_dn) \
+ ((_dn)->dn_objset->os_primary_cache == ZFS_CACHE_ALL || \
+ (DMU_OT_IS_METADATA((_dn)->dn_type) && \
+ (_dn)->dn_objset->os_primary_cache == ZFS_CACHE_METADATA))
+
+#define DNODE_META_IS_CACHEABLE(_dn) \
+ ((_dn)->dn_objset->os_primary_cache == ZFS_CACHE_ALL || \
+ (_dn)->dn_objset->os_primary_cache == ZFS_CACHE_METADATA)
+
+/*
+ * Used for dnodestats kstat.
+ */
+typedef struct dnode_stats {
+ /*
+ * Number of failed attempts to hold a meta dnode dbuf.
+ */
+ kstat_named_t dnode_hold_dbuf_hold;
+ /*
+ * Number of failed attempts to read a meta dnode dbuf.
+ */
+ kstat_named_t dnode_hold_dbuf_read;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_ALLOCATED) was able
+ * to hold the requested object number which was allocated. This is
+ * the common case when looking up any allocated object number.
+ */
+ kstat_named_t dnode_hold_alloc_hits;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_ALLOCATED) was not
+ * able to hold the request object number because it was not allocated.
+ */
+ kstat_named_t dnode_hold_alloc_misses;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_ALLOCATED) was not
+ * able to hold the request object number because the object number
+ * refers to an interior large dnode slot.
+ */
+ kstat_named_t dnode_hold_alloc_interior;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_ALLOCATED) needed
+ * to retry acquiring slot zrl locks due to contention.
+ */
+ kstat_named_t dnode_hold_alloc_lock_retry;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_ALLOCATED) did not
+ * need to create the dnode because another thread did so after
+ * dropping the read lock but before acquiring the write lock.
+ */
+ kstat_named_t dnode_hold_alloc_lock_misses;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_ALLOCATED) found
+ * a free dnode instantiated by dnode_create() but not yet allocated
+ * by dnode_allocate().
+ */
+ kstat_named_t dnode_hold_alloc_type_none;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_FREE) was able
+ * to hold the requested range of free dnode slots.
+ */
+ kstat_named_t dnode_hold_free_hits;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_FREE) was not
+ * able to hold the requested range of free dnode slots because
+ * at least one slot was allocated.
+ */
+ kstat_named_t dnode_hold_free_misses;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_FREE) was not
+ * able to hold the requested range of free dnode slots because
+ * after acquiring the zrl lock at least one slot was allocated.
+ */
+ kstat_named_t dnode_hold_free_lock_misses;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_FREE) needed
+ * to retry acquiring slot zrl locks due to contention.
+ */
+ kstat_named_t dnode_hold_free_lock_retry;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_FREE) requested
+ * a range of dnode slots which were held by another thread.
+ */
+ kstat_named_t dnode_hold_free_refcount;
+ /*
+ * Number of times dnode_hold(..., DNODE_MUST_BE_FREE) requested
+ * a range of dnode slots which would overflow the dnode_phys_t.
+ */
+ kstat_named_t dnode_hold_free_overflow;
+ /*
+ * Number of times a dnode_hold(...) was attempted on a dnode
+ * which had already been unlinked in an earlier txg.
+ */
+ kstat_named_t dnode_hold_free_txg;
+ /*
+ * Number of new dnodes allocated by dnode_allocate().
+ */
+ kstat_named_t dnode_allocate;
+ /*
+ * Number of dnodes re-allocated by dnode_reallocate().
+ */
+ kstat_named_t dnode_reallocate;
+ /*
+ * Number of meta dnode dbufs evicted.
+ */
+ kstat_named_t dnode_buf_evict;
+ /*
+ * Number of times dmu_object_alloc*() reached the end of the existing
+ * object ID chunk and advanced to a new one.
+ */
+ kstat_named_t dnode_alloc_next_chunk;
+ /*
+ * Number of times multiple threads attempted to allocate a dnode
+ * from the same block of free dnodes.
+ */
+ kstat_named_t dnode_alloc_race;
+ /*
+ * Number of times dmu_object_alloc*() was forced to advance to the
+ * next meta dnode dbuf due to an error from dmu_object_next().
+ */
+ kstat_named_t dnode_alloc_next_block;
+ /*
+ * Statistics for tracking dnodes which have been moved.
+ */
+ kstat_named_t dnode_move_invalid;
+ kstat_named_t dnode_move_recheck1;
+ kstat_named_t dnode_move_recheck2;
+ kstat_named_t dnode_move_special;
+ kstat_named_t dnode_move_handle;
+ kstat_named_t dnode_move_rwlock;
+ kstat_named_t dnode_move_active;
+} dnode_stats_t;
+
+extern dnode_stats_t dnode_stats;
+
+#define DNODE_STAT_INCR(stat, val) \
+ atomic_add_64(&dnode_stats.stat.value.ui64, (val));
+#define DNODE_STAT_BUMP(stat) \
+ DNODE_STAT_INCR(stat, 1);
+
#ifdef ZFS_DEBUG
/*