*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
+ * Copyright (c) 2012 by Delphix. All rights reserved.
*/
#include <sys/zfs_context.h>
static int free_range_compar(const void *node1, const void *node2);
static kmem_cache_t *dnode_cache;
+/*
+ * Define DNODE_STATS to turn on statistic gathering. By default, it is only
+ * turned on when DEBUG is also defined.
+ */
+#ifdef DEBUG
+#define DNODE_STATS
+#endif /* DEBUG */
+
+#ifdef DNODE_STATS
+#define DNODE_STAT_ADD(stat) ((stat)++)
+#else
+#define DNODE_STAT_ADD(stat) /* nothing */
+#endif /* DNODE_STATS */
-static dnode_phys_t dnode_phys_zero;
+ASSERTV(static dnode_phys_t dnode_phys_zero);
int zfs_default_bs = SPA_MINBLOCKSHIFT;
int zfs_default_ibs = DN_MAX_INDBLKSHIFT;
+#ifdef _KERNEL
+static kmem_cbrc_t dnode_move(void *, void *, size_t, void *);
+#endif /* _KERNEL */
+
/* ARGSUSED */
static int
dnode_cons(void *arg, void *unused, int kmflag)
{
- int i;
dnode_t *dn = arg;
- bzero(dn, sizeof (dnode_t));
+ int i;
rw_init(&dn->dn_struct_rwlock, NULL, RW_DEFAULT, NULL);
mutex_init(&dn->dn_mtx, NULL, MUTEX_DEFAULT, NULL);
refcount_create(&dn->dn_holds);
refcount_create(&dn->dn_tx_holds);
+ list_link_init(&dn->dn_link);
+
+ bzero(&dn->dn_next_nblkptr[0], sizeof (dn->dn_next_nblkptr));
+ bzero(&dn->dn_next_nlevels[0], sizeof (dn->dn_next_nlevels));
+ bzero(&dn->dn_next_indblkshift[0], sizeof (dn->dn_next_indblkshift));
+ bzero(&dn->dn_next_bonustype[0], sizeof (dn->dn_next_bonustype));
+ bzero(&dn->dn_rm_spillblk[0], sizeof (dn->dn_rm_spillblk));
+ bzero(&dn->dn_next_bonuslen[0], sizeof (dn->dn_next_bonuslen));
+ bzero(&dn->dn_next_blksz[0], sizeof (dn->dn_next_blksz));
for (i = 0; i < TXG_SIZE; i++) {
+ list_link_init(&dn->dn_dirty_link[i]);
avl_create(&dn->dn_ranges[i], free_range_compar,
sizeof (free_range_t),
offsetof(struct free_range, fr_node));
offsetof(dbuf_dirty_record_t, dr_dirty_node));
}
+ dn->dn_allocated_txg = 0;
+ dn->dn_free_txg = 0;
+ dn->dn_assigned_txg = 0;
+ dn->dn_dirtyctx = 0;
+ dn->dn_dirtyctx_firstset = NULL;
+ dn->dn_bonus = NULL;
+ dn->dn_have_spill = B_FALSE;
+ dn->dn_zio = NULL;
+ dn->dn_oldused = 0;
+ dn->dn_oldflags = 0;
+ dn->dn_olduid = 0;
+ dn->dn_oldgid = 0;
+ dn->dn_newuid = 0;
+ dn->dn_newgid = 0;
+ dn->dn_id_flags = 0;
+
+ dn->dn_dbufs_count = 0;
list_create(&dn->dn_dbufs, sizeof (dmu_buf_impl_t),
offsetof(dmu_buf_impl_t, db_link));
+ dn->dn_moved = 0;
return (0);
}
cv_destroy(&dn->dn_notxholds);
refcount_destroy(&dn->dn_holds);
refcount_destroy(&dn->dn_tx_holds);
+ ASSERT(!list_link_active(&dn->dn_link));
for (i = 0; i < TXG_SIZE; i++) {
+ ASSERT(!list_link_active(&dn->dn_dirty_link[i]));
avl_destroy(&dn->dn_ranges[i]);
list_destroy(&dn->dn_dirty_records[i]);
+ ASSERT0(dn->dn_next_nblkptr[i]);
+ ASSERT0(dn->dn_next_nlevels[i]);
+ ASSERT0(dn->dn_next_indblkshift[i]);
+ ASSERT0(dn->dn_next_bonustype[i]);
+ ASSERT0(dn->dn_rm_spillblk[i]);
+ ASSERT0(dn->dn_next_bonuslen[i]);
+ ASSERT0(dn->dn_next_blksz[i]);
}
+ ASSERT0(dn->dn_allocated_txg);
+ ASSERT0(dn->dn_free_txg);
+ ASSERT0(dn->dn_assigned_txg);
+ ASSERT0(dn->dn_dirtyctx);
+ ASSERT3P(dn->dn_dirtyctx_firstset, ==, NULL);
+ ASSERT3P(dn->dn_bonus, ==, NULL);
+ ASSERT(!dn->dn_have_spill);
+ ASSERT3P(dn->dn_zio, ==, NULL);
+ ASSERT0(dn->dn_oldused);
+ ASSERT0(dn->dn_oldflags);
+ ASSERT0(dn->dn_olduid);
+ ASSERT0(dn->dn_oldgid);
+ ASSERT0(dn->dn_newuid);
+ ASSERT0(dn->dn_newgid);
+ ASSERT0(dn->dn_id_flags);
+
+ ASSERT0(dn->dn_dbufs_count);
list_destroy(&dn->dn_dbufs);
}
void
dnode_init(void)
{
- dnode_cache = kmem_cache_create("dnode_t",
- sizeof (dnode_t),
- 0, dnode_cons, dnode_dest, NULL, NULL, NULL, 0);
+ ASSERT(dnode_cache == NULL);
+ dnode_cache = kmem_cache_create("dnode_t", sizeof (dnode_t),
+ 0, dnode_cons, dnode_dest, NULL, NULL, NULL, KMC_KMEM);
+ kmem_cache_set_move(dnode_cache, dnode_move);
}
void
dnode_fini(void)
{
kmem_cache_destroy(dnode_cache);
+ dnode_cache = NULL;
}
ASSERT(dn->dn_phys);
ASSERT(dn->dn_objset);
+ ASSERT(dn->dn_handle->dnh_dnode == dn);
- ASSERT(dn->dn_phys->dn_type < DMU_OT_NUMTYPES);
+ ASSERT(DMU_OT_IS_VALID(dn->dn_phys->dn_type));
if (!(zfs_flags & ZFS_DEBUG_DNODE_VERIFY))
return;
}
if (dn->dn_phys->dn_type != DMU_OT_NONE || dn->dn_allocated_txg != 0) {
int i;
- ASSERT3U(dn->dn_indblkshift, >=, 0);
ASSERT3U(dn->dn_indblkshift, <=, SPA_MAXBLOCKSHIFT);
if (dn->dn_datablkshift) {
ASSERT3U(dn->dn_datablkshift, >=, SPA_MINBLOCKSHIFT);
ASSERT3U(1<<dn->dn_datablkshift, ==, dn->dn_datablksz);
}
ASSERT3U(dn->dn_nlevels, <=, 30);
- ASSERT3U(dn->dn_type, <=, DMU_OT_NUMTYPES);
+ ASSERT(DMU_OT_IS_VALID(dn->dn_type));
ASSERT3U(dn->dn_nblkptr, >=, 1);
ASSERT3U(dn->dn_nblkptr, <=, DN_MAX_NBLKPTR);
ASSERT3U(dn->dn_bonuslen, <=, DN_MAX_BONUSLEN);
*/
int off = (dnp->dn_nblkptr-1) * sizeof (blkptr_t);
size_t len = DN_MAX_BONUSLEN - off;
- ASSERT3U(dnp->dn_bonustype, <, DMU_OT_NUMTYPES);
- dmu_ot[dnp->dn_bonustype].ot_byteswap(dnp->dn_bonus + off, len);
+ dmu_object_byteswap_t byteswap;
+ ASSERT(DMU_OT_IS_VALID(dnp->dn_bonustype));
+ byteswap = DMU_OT_BYTESWAP(dnp->dn_bonustype);
+ dmu_ot_byteswap[byteswap].ob_func(dnp->dn_bonus + off, len);
}
/* Swap SPILL block if we have one */
static void
dnode_setdblksz(dnode_t *dn, int size)
{
- ASSERT3U(P2PHASE(size, SPA_MINBLOCKSIZE), ==, 0);
+ ASSERT0(P2PHASE(size, SPA_MINBLOCKSIZE));
ASSERT3U(size, <=, SPA_MAXBLOCKSIZE);
ASSERT3U(size, >=, SPA_MINBLOCKSIZE);
ASSERT3U(size >> SPA_MINBLOCKSHIFT, <,
static dnode_t *
dnode_create(objset_t *os, dnode_phys_t *dnp, dmu_buf_impl_t *db,
- uint64_t object)
+ uint64_t object, dnode_handle_t *dnh)
{
- dnode_t *dn = kmem_cache_alloc(dnode_cache, KM_SLEEP);
- (void) dnode_cons(dn, NULL, 0); /* XXX */
+ dnode_t *dn = kmem_cache_alloc(dnode_cache, KM_PUSHPAGE);
- dn->dn_objset = os;
+ ASSERT(!POINTER_IS_VALID(dn->dn_objset));
+ dn->dn_moved = 0;
+
+ /*
+ * Defer setting dn_objset until the dnode is ready to be a candidate
+ * for the dnode_move() callback.
+ */
dn->dn_object = object;
dn->dn_dbuf = db;
+ dn->dn_handle = dnh;
dn->dn_phys = dnp;
- if (dnp->dn_datablkszsec)
+ if (dnp->dn_datablkszsec) {
dnode_setdblksz(dn, dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);
+ } else {
+ dn->dn_datablksz = 0;
+ dn->dn_datablkszsec = 0;
+ dn->dn_datablkshift = 0;
+ }
dn->dn_indblkshift = dnp->dn_indblkshift;
dn->dn_nlevels = dnp->dn_nlevels;
dn->dn_type = dnp->dn_type;
dmu_zfetch_init(&dn->dn_zfetch, dn);
- ASSERT(dn->dn_phys->dn_type < DMU_OT_NUMTYPES);
+ ASSERT(DMU_OT_IS_VALID(dn->dn_phys->dn_type));
+
mutex_enter(&os->os_lock);
list_insert_head(&os->os_dnodes, dn);
+ membar_producer();
+ /*
+ * Everything else must be valid before assigning dn_objset makes the
+ * dnode eligible for dnode_move().
+ */
+ dn->dn_objset = os;
mutex_exit(&os->os_lock);
arc_space_consume(sizeof (dnode_t), ARC_SPACE_OTHER);
return (dn);
}
+/*
+ * Caller must be holding the dnode handle, which is released upon return.
+ */
static void
dnode_destroy(dnode_t *dn)
{
objset_t *os = dn->dn_objset;
-#ifdef ZFS_DEBUG
- int i;
-
- for (i = 0; i < TXG_SIZE; i++) {
- ASSERT(!list_link_active(&dn->dn_dirty_link[i]));
- ASSERT(NULL == list_head(&dn->dn_dirty_records[i]));
- ASSERT(0 == avl_numnodes(&dn->dn_ranges[i]));
- }
- ASSERT(NULL == list_head(&dn->dn_dbufs));
-#endif
ASSERT((dn->dn_id_flags & DN_ID_NEW_EXIST) == 0);
mutex_enter(&os->os_lock);
+ POINTER_INVALIDATE(&dn->dn_objset);
list_remove(&os->os_dnodes, dn);
mutex_exit(&os->os_lock);
- if (dn->dn_dirtyctx_firstset) {
+ /* the dnode can no longer move, so we can release the handle */
+ zrl_remove(&dn->dn_handle->dnh_zrlock);
+
+ dn->dn_allocated_txg = 0;
+ dn->dn_free_txg = 0;
+ dn->dn_assigned_txg = 0;
+
+ dn->dn_dirtyctx = 0;
+ if (dn->dn_dirtyctx_firstset != NULL) {
kmem_free(dn->dn_dirtyctx_firstset, 1);
dn->dn_dirtyctx_firstset = NULL;
}
- dmu_zfetch_rele(&dn->dn_zfetch);
- if (dn->dn_bonus) {
+ if (dn->dn_bonus != NULL) {
mutex_enter(&dn->dn_bonus->db_mtx);
dbuf_evict(dn->dn_bonus);
dn->dn_bonus = NULL;
}
+ dn->dn_zio = NULL;
+
+ dn->dn_have_spill = B_FALSE;
+ dn->dn_oldused = 0;
+ dn->dn_oldflags = 0;
+ dn->dn_olduid = 0;
+ dn->dn_oldgid = 0;
+ dn->dn_newuid = 0;
+ dn->dn_newgid = 0;
+ dn->dn_id_flags = 0;
+
+ dmu_zfetch_rele(&dn->dn_zfetch);
kmem_cache_free(dnode_cache, dn);
arc_space_return(sizeof (dnode_t), ARC_SPACE_OTHER);
}
ASSERT(bcmp(dn->dn_phys, &dnode_phys_zero, sizeof (dnode_phys_t)) == 0);
ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE);
ASSERT(ot != DMU_OT_NONE);
- ASSERT3U(ot, <, DMU_OT_NUMTYPES);
+ ASSERT(DMU_OT_IS_VALID(ot));
ASSERT((bonustype == DMU_OT_NONE && bonuslen == 0) ||
(bonustype == DMU_OT_SA && bonuslen == 0) ||
(bonustype != DMU_OT_NONE && bonuslen != 0));
- ASSERT3U(bonustype, <, DMU_OT_NUMTYPES);
+ ASSERT(DMU_OT_IS_VALID(bonustype));
ASSERT3U(bonuslen, <=, DN_MAX_BONUSLEN);
ASSERT(dn->dn_type == DMU_OT_NONE);
- ASSERT3U(dn->dn_maxblkid, ==, 0);
- ASSERT3U(dn->dn_allocated_txg, ==, 0);
- ASSERT3U(dn->dn_assigned_txg, ==, 0);
+ ASSERT0(dn->dn_maxblkid);
+ ASSERT0(dn->dn_allocated_txg);
+ ASSERT0(dn->dn_assigned_txg);
ASSERT(refcount_is_zero(&dn->dn_tx_holds));
ASSERT3U(refcount_count(&dn->dn_holds), <=, 1);
ASSERT3P(list_head(&dn->dn_dbufs), ==, NULL);
for (i = 0; i < TXG_SIZE; i++) {
- ASSERT3U(dn->dn_next_nlevels[i], ==, 0);
- ASSERT3U(dn->dn_next_indblkshift[i], ==, 0);
- ASSERT3U(dn->dn_next_bonuslen[i], ==, 0);
- ASSERT3U(dn->dn_next_bonustype[i], ==, 0);
- ASSERT3U(dn->dn_rm_spillblk[i], ==, 0);
- ASSERT3U(dn->dn_next_blksz[i], ==, 0);
+ ASSERT0(dn->dn_next_nblkptr[i]);
+ ASSERT0(dn->dn_next_nlevels[i]);
+ ASSERT0(dn->dn_next_indblkshift[i]);
+ ASSERT0(dn->dn_next_bonuslen[i]);
+ ASSERT0(dn->dn_next_bonustype[i]);
+ ASSERT0(dn->dn_rm_spillblk[i]);
+ ASSERT0(dn->dn_next_blksz[i]);
ASSERT(!list_link_active(&dn->dn_dirty_link[i]));
ASSERT3P(list_head(&dn->dn_dirty_records[i]), ==, NULL);
- ASSERT3U(avl_numnodes(&dn->dn_ranges[i]), ==, 0);
+ ASSERT0(avl_numnodes(&dn->dn_ranges[i]));
}
dn->dn_type = ot;
ASSERT3U(blocksize, >=, SPA_MINBLOCKSIZE);
ASSERT3U(blocksize, <=, SPA_MAXBLOCKSIZE);
- ASSERT3U(blocksize % SPA_MINBLOCKSIZE, ==, 0);
+ ASSERT0(blocksize % SPA_MINBLOCKSIZE);
ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT || dmu_tx_private_ok(tx));
ASSERT(tx->tx_txg != 0);
ASSERT((bonustype == DMU_OT_NONE && bonuslen == 0) ||
(bonustype != DMU_OT_NONE && bonuslen != 0) ||
(bonustype == DMU_OT_SA && bonuslen == 0));
- ASSERT3U(bonustype, <, DMU_OT_NUMTYPES);
+ ASSERT(DMU_OT_IS_VALID(bonustype));
ASSERT3U(bonuslen, <=, DN_MAX_BONUSLEN);
/* clean up any unreferenced dbufs */
mutex_exit(&dn->dn_mtx);
}
+#ifdef _KERNEL
+#ifdef DNODE_STATS
+static struct {
+ uint64_t dms_dnode_invalid;
+ uint64_t dms_dnode_recheck1;
+ uint64_t dms_dnode_recheck2;
+ uint64_t dms_dnode_special;
+ uint64_t dms_dnode_handle;
+ uint64_t dms_dnode_rwlock;
+ uint64_t dms_dnode_active;
+} dnode_move_stats;
+#endif /* DNODE_STATS */
+
+static void
+dnode_move_impl(dnode_t *odn, dnode_t *ndn)
+{
+ int i;
+
+ ASSERT(!RW_LOCK_HELD(&odn->dn_struct_rwlock));
+ ASSERT(MUTEX_NOT_HELD(&odn->dn_mtx));
+ ASSERT(MUTEX_NOT_HELD(&odn->dn_dbufs_mtx));
+ ASSERT(!RW_LOCK_HELD(&odn->dn_zfetch.zf_rwlock));
+
+ /* Copy fields. */
+ ndn->dn_objset = odn->dn_objset;
+ ndn->dn_object = odn->dn_object;
+ ndn->dn_dbuf = odn->dn_dbuf;
+ ndn->dn_handle = odn->dn_handle;
+ ndn->dn_phys = odn->dn_phys;
+ ndn->dn_type = odn->dn_type;
+ ndn->dn_bonuslen = odn->dn_bonuslen;
+ ndn->dn_bonustype = odn->dn_bonustype;
+ ndn->dn_nblkptr = odn->dn_nblkptr;
+ ndn->dn_checksum = odn->dn_checksum;
+ ndn->dn_compress = odn->dn_compress;
+ ndn->dn_nlevels = odn->dn_nlevels;
+ ndn->dn_indblkshift = odn->dn_indblkshift;
+ ndn->dn_datablkshift = odn->dn_datablkshift;
+ ndn->dn_datablkszsec = odn->dn_datablkszsec;
+ ndn->dn_datablksz = odn->dn_datablksz;
+ ndn->dn_maxblkid = odn->dn_maxblkid;
+ bcopy(&odn->dn_next_nblkptr[0], &ndn->dn_next_nblkptr[0],
+ sizeof (odn->dn_next_nblkptr));
+ bcopy(&odn->dn_next_nlevels[0], &ndn->dn_next_nlevels[0],
+ sizeof (odn->dn_next_nlevels));
+ bcopy(&odn->dn_next_indblkshift[0], &ndn->dn_next_indblkshift[0],
+ sizeof (odn->dn_next_indblkshift));
+ bcopy(&odn->dn_next_bonustype[0], &ndn->dn_next_bonustype[0],
+ sizeof (odn->dn_next_bonustype));
+ bcopy(&odn->dn_rm_spillblk[0], &ndn->dn_rm_spillblk[0],
+ sizeof (odn->dn_rm_spillblk));
+ bcopy(&odn->dn_next_bonuslen[0], &ndn->dn_next_bonuslen[0],
+ sizeof (odn->dn_next_bonuslen));
+ bcopy(&odn->dn_next_blksz[0], &ndn->dn_next_blksz[0],
+ sizeof (odn->dn_next_blksz));
+ for (i = 0; i < TXG_SIZE; i++) {
+ list_move_tail(&ndn->dn_dirty_records[i],
+ &odn->dn_dirty_records[i]);
+ }
+ bcopy(&odn->dn_ranges[0], &ndn->dn_ranges[0], sizeof (odn->dn_ranges));
+ ndn->dn_allocated_txg = odn->dn_allocated_txg;
+ ndn->dn_free_txg = odn->dn_free_txg;
+ ndn->dn_assigned_txg = odn->dn_assigned_txg;
+ ndn->dn_dirtyctx = odn->dn_dirtyctx;
+ ndn->dn_dirtyctx_firstset = odn->dn_dirtyctx_firstset;
+ ASSERT(refcount_count(&odn->dn_tx_holds) == 0);
+ refcount_transfer(&ndn->dn_holds, &odn->dn_holds);
+ ASSERT(list_is_empty(&ndn->dn_dbufs));
+ list_move_tail(&ndn->dn_dbufs, &odn->dn_dbufs);
+ ndn->dn_dbufs_count = odn->dn_dbufs_count;
+ ndn->dn_bonus = odn->dn_bonus;
+ ndn->dn_have_spill = odn->dn_have_spill;
+ ndn->dn_zio = odn->dn_zio;
+ ndn->dn_oldused = odn->dn_oldused;
+ ndn->dn_oldflags = odn->dn_oldflags;
+ ndn->dn_olduid = odn->dn_olduid;
+ ndn->dn_oldgid = odn->dn_oldgid;
+ ndn->dn_newuid = odn->dn_newuid;
+ ndn->dn_newgid = odn->dn_newgid;
+ ndn->dn_id_flags = odn->dn_id_flags;
+ dmu_zfetch_init(&ndn->dn_zfetch, NULL);
+ list_move_tail(&ndn->dn_zfetch.zf_stream, &odn->dn_zfetch.zf_stream);
+ ndn->dn_zfetch.zf_dnode = odn->dn_zfetch.zf_dnode;
+ ndn->dn_zfetch.zf_stream_cnt = odn->dn_zfetch.zf_stream_cnt;
+ ndn->dn_zfetch.zf_alloc_fail = odn->dn_zfetch.zf_alloc_fail;
+
+ /*
+ * Update back pointers. Updating the handle fixes the back pointer of
+ * every descendant dbuf as well as the bonus dbuf.
+ */
+ ASSERT(ndn->dn_handle->dnh_dnode == odn);
+ ndn->dn_handle->dnh_dnode = ndn;
+ if (ndn->dn_zfetch.zf_dnode == odn) {
+ ndn->dn_zfetch.zf_dnode = ndn;
+ }
+
+ /*
+ * Invalidate the original dnode by clearing all of its back pointers.
+ */
+ odn->dn_dbuf = NULL;
+ odn->dn_handle = NULL;
+ list_create(&odn->dn_dbufs, sizeof (dmu_buf_impl_t),
+ offsetof(dmu_buf_impl_t, db_link));
+ odn->dn_dbufs_count = 0;
+ odn->dn_bonus = NULL;
+ odn->dn_zfetch.zf_dnode = NULL;
+
+ /*
+ * Set the low bit of the objset pointer to ensure that dnode_move()
+ * recognizes the dnode as invalid in any subsequent callback.
+ */
+ POINTER_INVALIDATE(&odn->dn_objset);
+
+ /*
+ * Satisfy the destructor.
+ */
+ for (i = 0; i < TXG_SIZE; i++) {
+ list_create(&odn->dn_dirty_records[i],
+ sizeof (dbuf_dirty_record_t),
+ offsetof(dbuf_dirty_record_t, dr_dirty_node));
+ odn->dn_ranges[i].avl_root = NULL;
+ odn->dn_ranges[i].avl_numnodes = 0;
+ odn->dn_next_nlevels[i] = 0;
+ odn->dn_next_indblkshift[i] = 0;
+ odn->dn_next_bonustype[i] = 0;
+ odn->dn_rm_spillblk[i] = 0;
+ odn->dn_next_bonuslen[i] = 0;
+ odn->dn_next_blksz[i] = 0;
+ }
+ odn->dn_allocated_txg = 0;
+ odn->dn_free_txg = 0;
+ odn->dn_assigned_txg = 0;
+ odn->dn_dirtyctx = 0;
+ odn->dn_dirtyctx_firstset = NULL;
+ odn->dn_have_spill = B_FALSE;
+ odn->dn_zio = NULL;
+ odn->dn_oldused = 0;
+ odn->dn_oldflags = 0;
+ odn->dn_olduid = 0;
+ odn->dn_oldgid = 0;
+ odn->dn_newuid = 0;
+ odn->dn_newgid = 0;
+ odn->dn_id_flags = 0;
+
+ /*
+ * Mark the dnode.
+ */
+ ndn->dn_moved = 1;
+ odn->dn_moved = (uint8_t)-1;
+}
+
+/*ARGSUSED*/
+static kmem_cbrc_t
+dnode_move(void *buf, void *newbuf, size_t size, void *arg)
+{
+ dnode_t *odn = buf, *ndn = newbuf;
+ objset_t *os;
+ int64_t refcount;
+ uint32_t dbufs;
+
+ /*
+ * The dnode is on the objset's list of known dnodes if the objset
+ * pointer is valid. We set the low bit of the objset pointer when
+ * freeing the dnode to invalidate it, and the memory patterns written
+ * by kmem (baddcafe and deadbeef) set at least one of the two low bits.
+ * A newly created dnode sets the objset pointer last of all to indicate
+ * that the dnode is known and in a valid state to be moved by this
+ * function.
+ */
+ os = odn->dn_objset;
+ if (!POINTER_IS_VALID(os)) {
+ DNODE_STAT_ADD(dnode_move_stats.dms_dnode_invalid);
+ return (KMEM_CBRC_DONT_KNOW);
+ }
+
+ /*
+ * Ensure that the objset does not go away during the move.
+ */
+ rw_enter(&os_lock, RW_WRITER);
+ if (os != odn->dn_objset) {
+ rw_exit(&os_lock);
+ DNODE_STAT_ADD(dnode_move_stats.dms_dnode_recheck1);
+ return (KMEM_CBRC_DONT_KNOW);
+ }
+
+ /*
+ * If the dnode is still valid, then so is the objset. We know that no
+ * valid objset can be freed while we hold os_lock, so we can safely
+ * ensure that the objset remains in use.
+ */
+ mutex_enter(&os->os_lock);
+
+ /*
+ * Recheck the objset pointer in case the dnode was removed just before
+ * acquiring the lock.
+ */
+ if (os != odn->dn_objset) {
+ mutex_exit(&os->os_lock);
+ rw_exit(&os_lock);
+ DNODE_STAT_ADD(dnode_move_stats.dms_dnode_recheck2);
+ return (KMEM_CBRC_DONT_KNOW);
+ }
+
+ /*
+ * At this point we know that as long as we hold os->os_lock, the dnode
+ * cannot be freed and fields within the dnode can be safely accessed.
+ * The objset listing this dnode cannot go away as long as this dnode is
+ * on its list.
+ */
+ rw_exit(&os_lock);
+ if (DMU_OBJECT_IS_SPECIAL(odn->dn_object)) {
+ mutex_exit(&os->os_lock);
+ DNODE_STAT_ADD(dnode_move_stats.dms_dnode_special);
+ return (KMEM_CBRC_NO);
+ }
+ ASSERT(odn->dn_dbuf != NULL); /* only "special" dnodes have no parent */
+
+ /*
+ * Lock the dnode handle to prevent the dnode from obtaining any new
+ * holds. This also prevents the descendant dbufs and the bonus dbuf
+ * from accessing the dnode, so that we can discount their holds. The
+ * handle is safe to access because we know that while the dnode cannot
+ * go away, neither can its handle. Once we hold dnh_zrlock, we can
+ * safely move any dnode referenced only by dbufs.
+ */
+ if (!zrl_tryenter(&odn->dn_handle->dnh_zrlock)) {
+ mutex_exit(&os->os_lock);
+ DNODE_STAT_ADD(dnode_move_stats.dms_dnode_handle);
+ return (KMEM_CBRC_LATER);
+ }
+
+ /*
+ * Ensure a consistent view of the dnode's holds and the dnode's dbufs.
+ * We need to guarantee that there is a hold for every dbuf in order to
+ * determine whether the dnode is actively referenced. Falsely matching
+ * a dbuf to an active hold would lead to an unsafe move. It's possible
+ * that a thread already having an active dnode hold is about to add a
+ * dbuf, and we can't compare hold and dbuf counts while the add is in
+ * progress.
+ */
+ if (!rw_tryenter(&odn->dn_struct_rwlock, RW_WRITER)) {
+ zrl_exit(&odn->dn_handle->dnh_zrlock);
+ mutex_exit(&os->os_lock);
+ DNODE_STAT_ADD(dnode_move_stats.dms_dnode_rwlock);
+ return (KMEM_CBRC_LATER);
+ }
+
+ /*
+ * A dbuf may be removed (evicted) without an active dnode hold. In that
+ * case, the dbuf count is decremented under the handle lock before the
+ * dbuf's hold is released. This order ensures that if we count the hold
+ * after the dbuf is removed but before its hold is released, we will
+ * treat the unmatched hold as active and exit safely. If we count the
+ * hold before the dbuf is removed, the hold is discounted, and the
+ * removal is blocked until the move completes.
+ */
+ refcount = refcount_count(&odn->dn_holds);
+ ASSERT(refcount >= 0);
+ dbufs = odn->dn_dbufs_count;
+
+ /* We can't have more dbufs than dnode holds. */
+ ASSERT3U(dbufs, <=, refcount);
+ DTRACE_PROBE3(dnode__move, dnode_t *, odn, int64_t, refcount,
+ uint32_t, dbufs);
+
+ if (refcount > dbufs) {
+ rw_exit(&odn->dn_struct_rwlock);
+ zrl_exit(&odn->dn_handle->dnh_zrlock);
+ mutex_exit(&os->os_lock);
+ DNODE_STAT_ADD(dnode_move_stats.dms_dnode_active);
+ return (KMEM_CBRC_LATER);
+ }
+
+ rw_exit(&odn->dn_struct_rwlock);
+
+ /*
+ * At this point we know that anyone with a hold on the dnode is not
+ * actively referencing it. The dnode is known and in a valid state to
+ * move. We're holding the locks needed to execute the critical section.
+ */
+ dnode_move_impl(odn, ndn);
+
+ list_link_replace(&odn->dn_link, &ndn->dn_link);
+ /* If the dnode was safe to move, the refcount cannot have changed. */
+ ASSERT(refcount == refcount_count(&ndn->dn_holds));
+ ASSERT(dbufs == ndn->dn_dbufs_count);
+ zrl_exit(&ndn->dn_handle->dnh_zrlock); /* handle has moved */
+ mutex_exit(&os->os_lock);
+
+ return (KMEM_CBRC_YES);
+}
+#endif /* _KERNEL */
+
void
-dnode_special_close(dnode_t *dn)
+dnode_special_close(dnode_handle_t *dnh)
{
+ dnode_t *dn = dnh->dnh_dnode;
+
/*
* Wait for final references to the dnode to clear. This can
* only happen if the arc is asyncronously evicting state that
*/
while (refcount_count(&dn->dn_holds) > 0)
delay(1);
- dnode_destroy(dn);
+ zrl_add(&dnh->dnh_zrlock);
+ dnode_destroy(dn); /* implicit zrl_remove() */
+ zrl_destroy(&dnh->dnh_zrlock);
+ dnh->dnh_dnode = NULL;
}
dnode_t *
-dnode_special_open(objset_t *os, dnode_phys_t *dnp, uint64_t object)
+dnode_special_open(objset_t *os, dnode_phys_t *dnp, uint64_t object,
+ dnode_handle_t *dnh)
{
- dnode_t *dn = dnode_create(os, dnp, NULL, object);
+ dnode_t *dn = dnode_create(os, dnp, NULL, object, dnh);
+ dnh->dnh_dnode = dn;
+ zrl_init(&dnh->dnh_zrlock);
DNODE_VERIFY(dn);
return (dn);
}
static void
dnode_buf_pageout(dmu_buf_t *db, void *arg)
{
- dnode_t **children_dnodes = arg;
+ dnode_children_t *children_dnodes = arg;
int i;
int epb = db->db_size >> DNODE_SHIFT;
+ ASSERT(epb == children_dnodes->dnc_count);
+
for (i = 0; i < epb; i++) {
- dnode_t *dn = children_dnodes[i];
- int n;
+ dnode_handle_t *dnh = &children_dnodes->dnc_children[i];
+ dnode_t *dn;
- if (dn == NULL)
+ /*
+ * The dnode handle lock guards against the dnode moving to
+ * another valid address, so there is no need here to guard
+ * against changes to or from NULL.
+ */
+ if (dnh->dnh_dnode == NULL) {
+ zrl_destroy(&dnh->dnh_zrlock);
continue;
-#ifdef ZFS_DEBUG
+ }
+
+ zrl_add(&dnh->dnh_zrlock);
+ dn = dnh->dnh_dnode;
/*
* If there are holds on this dnode, then there should
* be holds on the dnode's containing dbuf as well; thus
- * it wouldn't be eligable for eviction and this function
+ * it wouldn't be eligible for eviction and this function
* would not have been called.
*/
ASSERT(refcount_is_zero(&dn->dn_holds));
- ASSERT(list_head(&dn->dn_dbufs) == NULL);
ASSERT(refcount_is_zero(&dn->dn_tx_holds));
- for (n = 0; n < TXG_SIZE; n++)
- ASSERT(!list_link_active(&dn->dn_dirty_link[n]));
-#endif
- children_dnodes[i] = NULL;
- dnode_destroy(dn);
+ dnode_destroy(dn); /* implicit zrl_remove() */
+ zrl_destroy(&dnh->dnh_zrlock);
+ dnh->dnh_dnode = NULL;
}
- kmem_free(children_dnodes, epb * sizeof (dnode_t *));
+ kmem_free(children_dnodes, sizeof (dnode_children_t) +
+ (epb - 1) * sizeof (dnode_handle_t));
}
/*
uint64_t blk;
dnode_t *mdn, *dn;
dmu_buf_impl_t *db;
- dnode_t **children_dnodes;
+ dnode_children_t *children_dnodes;
+ dnode_handle_t *dnh;
/*
* If you are holding the spa config lock as writer, you shouldn't
*/
ASSERT(spa_config_held(os->os_spa, SCL_ALL, RW_WRITER) == 0 ||
(spa_is_root(os->os_spa) &&
- spa_config_held(os->os_spa, SCL_STATE, RW_WRITER) &&
- !spa_config_held(os->os_spa, SCL_ZIO, RW_WRITER)));
+ spa_config_held(os->os_spa, SCL_STATE, RW_WRITER)));
if (object == DMU_USERUSED_OBJECT || object == DMU_GROUPUSED_OBJECT) {
dn = (object == DMU_USERUSED_OBJECT) ?
- os->os_userused_dnode : os->os_groupused_dnode;
+ DMU_USERUSED_DNODE(os) : DMU_GROUPUSED_DNODE(os);
if (dn == NULL)
return (ENOENT);
type = dn->dn_type;
if (object == 0 || object >= DN_MAX_OBJECT)
return (EINVAL);
- mdn = os->os_meta_dnode;
+ mdn = DMU_META_DNODE(os);
+ ASSERT(mdn->dn_object == DMU_META_DNODE_OBJECT);
DNODE_VERIFY(mdn);
idx = object & (epb-1);
+ ASSERT(DB_DNODE(db)->dn_type == DMU_OT_DNODE);
children_dnodes = dmu_buf_get_user(&db->db);
if (children_dnodes == NULL) {
- dnode_t **winner;
- children_dnodes = kmem_zalloc(epb * sizeof (dnode_t *),
- KM_SLEEP);
- if (winner = dmu_buf_set_user(&db->db, children_dnodes, NULL,
- dnode_buf_pageout)) {
- kmem_free(children_dnodes, epb * sizeof (dnode_t *));
+ int i;
+ dnode_children_t *winner;
+ children_dnodes = kmem_alloc(sizeof (dnode_children_t) +
+ (epb - 1) * sizeof (dnode_handle_t),
+ KM_PUSHPAGE | KM_NODEBUG);
+ children_dnodes->dnc_count = epb;
+ dnh = &children_dnodes->dnc_children[0];
+ for (i = 0; i < epb; i++) {
+ zrl_init(&dnh[i].dnh_zrlock);
+ dnh[i].dnh_dnode = NULL;
+ }
+ if ((winner = dmu_buf_set_user(&db->db, children_dnodes, NULL,
+ dnode_buf_pageout))) {
+ kmem_free(children_dnodes, sizeof (dnode_children_t) +
+ (epb - 1) * sizeof (dnode_handle_t));
children_dnodes = winner;
}
}
+ ASSERT(children_dnodes->dnc_count == epb);
- if ((dn = children_dnodes[idx]) == NULL) {
- dnode_phys_t *dnp = (dnode_phys_t *)db->db.db_data+idx;
+ dnh = &children_dnodes->dnc_children[idx];
+ zrl_add(&dnh->dnh_zrlock);
+ if ((dn = dnh->dnh_dnode) == NULL) {
+ dnode_phys_t *phys = (dnode_phys_t *)db->db.db_data+idx;
dnode_t *winner;
- dn = dnode_create(os, dnp, db, object);
- winner = atomic_cas_ptr(&children_dnodes[idx], NULL, dn);
+ dn = dnode_create(os, phys, db, object, dnh);
+ winner = atomic_cas_ptr(&dnh->dnh_dnode, NULL, dn);
if (winner != NULL) {
- dnode_destroy(dn);
+ zrl_add(&dnh->dnh_zrlock);
+ dnode_destroy(dn); /* implicit zrl_remove() */
dn = winner;
}
}
((flag & DNODE_MUST_BE_FREE) &&
(type != DMU_OT_NONE || !refcount_is_zero(&dn->dn_holds)))) {
mutex_exit(&dn->dn_mtx);
+ zrl_remove(&dnh->dnh_zrlock);
dbuf_rele(db, FTAG);
return (type == DMU_OT_NONE ? ENOENT : EEXIST);
}
mutex_exit(&dn->dn_mtx);
if (refcount_add(&dn->dn_holds, tag) == 1)
- dbuf_add_ref(db, dn);
+ dbuf_add_ref(db, dnh);
+ /* Now we can rely on the hold to prevent the dnode from moving. */
+ zrl_remove(&dnh->dnh_zrlock);
DNODE_VERIFY(dn);
ASSERT3P(dn->dn_dbuf, ==, db);
dnode_rele(dnode_t *dn, void *tag)
{
uint64_t refs;
+ /* Get while the hold prevents the dnode from moving. */
+ dmu_buf_impl_t *db = dn->dn_dbuf;
+ dnode_handle_t *dnh = dn->dn_handle;
mutex_enter(&dn->dn_mtx);
refs = refcount_remove(&dn->dn_holds, tag);
mutex_exit(&dn->dn_mtx);
+
+ /*
+ * It's unsafe to release the last hold on a dnode by dnode_rele() or
+ * indirectly by dbuf_rele() while relying on the dnode handle to
+ * prevent the dnode from moving, since releasing the last hold could
+ * result in the dnode's parent dbuf evicting its dnode handles. For
+ * that reason anyone calling dnode_rele() or dbuf_rele() without some
+ * other direct or indirect hold on the dnode must first drop the dnode
+ * handle.
+ */
+ ASSERT(refs > 0 || dnh->dnh_zrlock.zr_owner != curthread);
+
/* NOTE: the DNODE_DNODE does not have a dn_dbuf */
- if (refs == 0 && dn->dn_dbuf)
- dbuf_rele(dn->dn_dbuf, dn);
+ if (refs == 0 && db != NULL) {
+ /*
+ * Another thread could add a hold to the dnode handle in
+ * dnode_hold_impl() while holding the parent dbuf. Since the
+ * hold on the parent dbuf prevents the handle from being
+ * destroyed, the hold on the handle is OK. We can't yet assert
+ * that the handle has zero references, but that will be
+ * asserted anyway when the handle gets destroyed.
+ */
+ dbuf_rele(db, dnh);
+ }
}
void
#ifdef ZFS_DEBUG
mutex_enter(&dn->dn_mtx);
ASSERT(dn->dn_phys->dn_type || dn->dn_allocated_txg);
- /* ASSERT(dn->dn_free_txg == 0 || dn->dn_free_txg >= txg); */
+ ASSERT(dn->dn_free_txg == 0 || dn->dn_free_txg >= txg);
mutex_exit(&dn->dn_mtx);
#endif
ASSERT(!refcount_is_zero(&dn->dn_holds) || list_head(&dn->dn_dbufs));
ASSERT(dn->dn_datablksz != 0);
- ASSERT3U(dn->dn_next_bonuslen[txg&TXG_MASK], ==, 0);
- ASSERT3U(dn->dn_next_blksz[txg&TXG_MASK], ==, 0);
- ASSERT3U(dn->dn_next_bonustype[txg&TXG_MASK], ==, 0);
+ ASSERT0(dn->dn_next_bonuslen[txg&TXG_MASK]);
+ ASSERT0(dn->dn_next_blksz[txg&TXG_MASK]);
+ ASSERT0(dn->dn_next_bonustype[txg&TXG_MASK]);
dprintf_ds(os->os_dsl_dataset, "obj=%llu txg=%llu\n",
dn->dn_object, txg);
/*
* The dnode maintains a hold on its containing dbuf as
* long as there are holds on it. Each instantiated child
- * dbuf maintaines a hold on the dnode. When the last child
+ * dbuf maintains a hold on the dnode. When the last child
* drops its hold, the dnode will drop its hold on the
* containing dbuf. We add a "dirty hold" here so that the
* dnode will hang around after we finish processing its
} else if (blkid > rp->fr_blkid && endblk < fr_endblk) {
/* clear a chunk out of this range */
free_range_t *new_rp =
- kmem_alloc(sizeof (free_range_t), KM_SLEEP);
+ kmem_alloc(sizeof (free_range_t), KM_PUSHPAGE);
new_rp->fr_blkid = endblk;
new_rp->fr_nblks = fr_endblk - endblk;
else
tail = P2PHASE(len, blksz);
- ASSERT3U(P2PHASE(off, blksz), ==, 0);
+ ASSERT0(P2PHASE(off, blksz));
/* zero out any partial block data at the end of the range */
if (tail) {
if (len < tail)
int shift = epbs + dn->dn_datablkshift;
first = blkid >> epbs;
- if (db = dbuf_hold_level(dn, 1, first, FTAG)) {
+ if ((db = dbuf_hold_level(dn, 1, first, FTAG))) {
dbuf_will_dirty(db, tx);
dbuf_rele(db, FTAG);
}
avl_tree_t *tree = &dn->dn_ranges[tx->tx_txg&TXG_MASK];
/* Add new range to dn_ranges */
- rp = kmem_alloc(sizeof (free_range_t), KM_SLEEP);
+ rp = kmem_alloc(sizeof (free_range_t), KM_PUSHPAGE);
rp->fr_blkid = blkid;
rp->fr_nblks = nblks;
found = avl_find(tree, rp, &where);
space += delta;
if (spa_version(dn->dn_objset->os_spa) < SPA_VERSION_DNODE_BYTES) {
ASSERT((dn->dn_phys->dn_flags & DNODE_FLAG_USED_BYTES) == 0);
- ASSERT3U(P2PHASE(space, 1<<DEV_BSHIFT), ==, 0);
+ ASSERT0(P2PHASE(space, 1<<DEV_BSHIFT));
dn->dn_phys->dn_used = space >> DEV_BSHIFT;
} else {
dn->dn_phys->dn_used = space;
git://git.camperquake.de / zfs.git / blobdiff