Print this page
4101 metaslab_debug should allow for fine-grained control
4102 space_maps should store more information about themselves
4103 space map object blocksize should be increased
4104 ::spa_space no longer works
4105 removing a mirrored log device results in a leaked object
4106 asynchronously load metaslab
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Adam Leventhal <ahl@delphix.com>
Reviewed by: Sebastien Roy <seb@delphix.com>
@@ -34,10 +34,11 @@
#include <sys/vdev_impl.h>
#include <sys/uberblock_impl.h>
#include <sys/metaslab.h>
#include <sys/metaslab_impl.h>
#include <sys/space_map.h>
+#include <sys/space_reftree.h>
#include <sys/zio.h>
#include <sys/zap.h>
#include <sys/fs/zfs.h>
#include <sys/arc.h>
#include <sys/zil.h>
@@ -317,11 +318,11 @@
mutex_init(&vd->vdev_dtl_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&vd->vdev_stat_lock, NULL, MUTEX_DEFAULT, NULL);
mutex_init(&vd->vdev_probe_lock, NULL, MUTEX_DEFAULT, NULL);
for (int t = 0; t < DTL_TYPES; t++) {
- space_map_create(&vd->vdev_dtl[t], 0, -1ULL, 0,
+ vd->vdev_dtl[t] = range_tree_create(NULL, NULL,
&vd->vdev_dtl_lock);
}
txg_list_create(&vd->vdev_ms_list,
offsetof(struct metaslab, ms_txg_node));
txg_list_create(&vd->vdev_dtl_list,
@@ -503,11 +504,11 @@
if (vd->vdev_ops->vdev_op_leaf &&
(alloctype == VDEV_ALLOC_LOAD || alloctype == VDEV_ALLOC_L2CACHE ||
alloctype == VDEV_ALLOC_ROOTPOOL)) {
if (alloctype == VDEV_ALLOC_LOAD) {
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_DTL,
- &vd->vdev_dtl_smo.smo_object);
+ &vd->vdev_dtl_object);
(void) nvlist_lookup_uint64(nv, ZPOOL_CONFIG_UNSPARE,
&vd->vdev_unspare);
}
if (alloctype == VDEV_ALLOC_ROOTPOOL) {
@@ -625,13 +626,14 @@
txg_list_destroy(&vd->vdev_ms_list);
txg_list_destroy(&vd->vdev_dtl_list);
mutex_enter(&vd->vdev_dtl_lock);
+ space_map_close(vd->vdev_dtl_sm);
for (int t = 0; t < DTL_TYPES; t++) {
- space_map_unload(&vd->vdev_dtl[t]);
- space_map_destroy(&vd->vdev_dtl[t]);
+ range_tree_vacate(vd->vdev_dtl[t], NULL, NULL);
+ range_tree_destroy(vd->vdev_dtl[t]);
}
mutex_exit(&vd->vdev_dtl_lock);
mutex_destroy(&vd->vdev_dtl_lock);
mutex_destroy(&vd->vdev_stat_lock);
@@ -838,32 +840,21 @@
vd->vdev_ms = mspp;
vd->vdev_ms_count = newc;
for (m = oldc; m < newc; m++) {
- space_map_obj_t smo = { 0, 0, 0 };
- if (txg == 0) {
uint64_t object = 0;
+
+ if (txg == 0) {
error = dmu_read(mos, vd->vdev_ms_array,
m * sizeof (uint64_t), sizeof (uint64_t), &object,
DMU_READ_PREFETCH);
if (error)
return (error);
- if (object != 0) {
- dmu_buf_t *db;
- error = dmu_bonus_hold(mos, object, FTAG, &db);
- if (error)
- return (error);
- ASSERT3U(db->db_size, >=, sizeof (smo));
- bcopy(db->db_data, &smo, sizeof (smo));
- ASSERT3U(smo.smo_object, ==, object);
- dmu_buf_rele(db, FTAG);
}
+ vd->vdev_ms[m] = metaslab_init(vd->vdev_mg, m, object, txg);
}
- vd->vdev_ms[m] = metaslab_init(vd->vdev_mg, &smo,
- m << vd->vdev_ms_shift, 1ULL << vd->vdev_ms_shift, txg);
- }
if (txg == 0)
spa_config_enter(spa, SCL_ALLOC, FTAG, RW_WRITER);
/*
@@ -886,13 +877,16 @@
uint64_t m;
uint64_t count = vd->vdev_ms_count;
if (vd->vdev_ms != NULL) {
metaslab_group_passivate(vd->vdev_mg);
- for (m = 0; m < count; m++)
- if (vd->vdev_ms[m] != NULL)
- metaslab_fini(vd->vdev_ms[m]);
+ for (m = 0; m < count; m++) {
+ metaslab_t *msp = vd->vdev_ms[m];
+
+ if (msp != NULL)
+ metaslab_fini(msp);
+ }
kmem_free(vd->vdev_ms, count * sizeof (metaslab_t *));
vd->vdev_ms = NULL;
}
}
@@ -1539,13 +1533,14 @@
vdev_close(vd);
return (error ? error : ENXIO);
}
/*
- * Recursively initialize all labels.
+ * Recursively load DTLs and initialize all labels.
*/
- if ((error = vdev_label_init(vd, txg, isreplacing ?
+ if ((error = vdev_dtl_load(vd)) != 0 ||
+ (error = vdev_label_init(vd, txg, isreplacing ?
VDEV_LABEL_REPLACE : VDEV_LABEL_CREATE)) != 0) {
vdev_close(vd);
return (error);
}
@@ -1577,10 +1572,20 @@
(void) txg_list_add(&vd->vdev_dtl_list, arg, txg);
(void) txg_list_add(&vd->vdev_spa->spa_vdev_txg_list, vd, txg);
}
+void
+vdev_dirty_leaves(vdev_t *vd, int flags, uint64_t txg)
+{
+ for (int c = 0; c < vd->vdev_children; c++)
+ vdev_dirty_leaves(vd->vdev_child[c], flags, txg);
+
+ if (vd->vdev_ops->vdev_op_leaf)
+ vdev_dirty(vd->vdev_top, flags, vd, txg);
+}
+
/*
* DTLs.
*
* A vdev's DTL (dirty time log) is the set of transaction groups for which
* the vdev has less than perfect replication. There are four kinds of DTL:
@@ -1618,48 +1623,48 @@
* a configuration change, we generate all other DTLs from first principles.
*/
void
vdev_dtl_dirty(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
{
- space_map_t *sm = &vd->vdev_dtl[t];
+ range_tree_t *rt = vd->vdev_dtl[t];
ASSERT(t < DTL_TYPES);
ASSERT(vd != vd->vdev_spa->spa_root_vdev);
ASSERT(spa_writeable(vd->vdev_spa));
- mutex_enter(sm->sm_lock);
- if (!space_map_contains(sm, txg, size))
- space_map_add(sm, txg, size);
- mutex_exit(sm->sm_lock);
+ mutex_enter(rt->rt_lock);
+ if (!range_tree_contains(rt, txg, size))
+ range_tree_add(rt, txg, size);
+ mutex_exit(rt->rt_lock);
}
boolean_t
vdev_dtl_contains(vdev_t *vd, vdev_dtl_type_t t, uint64_t txg, uint64_t size)
{
- space_map_t *sm = &vd->vdev_dtl[t];
+ range_tree_t *rt = vd->vdev_dtl[t];
boolean_t dirty = B_FALSE;
ASSERT(t < DTL_TYPES);
ASSERT(vd != vd->vdev_spa->spa_root_vdev);
- mutex_enter(sm->sm_lock);
- if (sm->sm_space != 0)
- dirty = space_map_contains(sm, txg, size);
- mutex_exit(sm->sm_lock);
+ mutex_enter(rt->rt_lock);
+ if (range_tree_space(rt) != 0)
+ dirty = range_tree_contains(rt, txg, size);
+ mutex_exit(rt->rt_lock);
return (dirty);
}
boolean_t
vdev_dtl_empty(vdev_t *vd, vdev_dtl_type_t t)
{
- space_map_t *sm = &vd->vdev_dtl[t];
+ range_tree_t *rt = vd->vdev_dtl[t];
boolean_t empty;
- mutex_enter(sm->sm_lock);
- empty = (sm->sm_space == 0);
- mutex_exit(sm->sm_lock);
+ mutex_enter(rt->rt_lock);
+ empty = (range_tree_space(rt) == 0);
+ mutex_exit(rt->rt_lock);
return (empty);
}
/*
@@ -1666,34 +1671,34 @@
* Returns the lowest txg in the DTL range.
*/
static uint64_t
vdev_dtl_min(vdev_t *vd)
{
- space_seg_t *ss;
+ range_seg_t *rs;
ASSERT(MUTEX_HELD(&vd->vdev_dtl_lock));
- ASSERT3U(vd->vdev_dtl[DTL_MISSING].sm_space, !=, 0);
+ ASSERT3U(range_tree_space(vd->vdev_dtl[DTL_MISSING]), !=, 0);
ASSERT0(vd->vdev_children);
- ss = avl_first(&vd->vdev_dtl[DTL_MISSING].sm_root);
- return (ss->ss_start - 1);
+ rs = avl_first(&vd->vdev_dtl[DTL_MISSING]->rt_root);
+ return (rs->rs_start - 1);
}
/*
* Returns the highest txg in the DTL.
*/
static uint64_t
vdev_dtl_max(vdev_t *vd)
{
- space_seg_t *ss;
+ range_seg_t *rs;
ASSERT(MUTEX_HELD(&vd->vdev_dtl_lock));
- ASSERT3U(vd->vdev_dtl[DTL_MISSING].sm_space, !=, 0);
+ ASSERT3U(range_tree_space(vd->vdev_dtl[DTL_MISSING]), !=, 0);
ASSERT0(vd->vdev_children);
- ss = avl_last(&vd->vdev_dtl[DTL_MISSING].sm_root);
- return (ss->ss_end);
+ rs = avl_last(&vd->vdev_dtl[DTL_MISSING]->rt_root);
+ return (rs->rs_end);
}
/*
* Determine if a resilvering vdev should remove any DTL entries from
* its range. If the vdev was resilvering for the entire duration of the
@@ -1710,11 +1715,11 @@
ASSERT0(scn->scn_phys.scn_errors);
ASSERT0(vd->vdev_children);
if (vd->vdev_resilver_txg == 0 ||
- vd->vdev_dtl[DTL_MISSING].sm_space == 0)
+ range_tree_space(vd->vdev_dtl[DTL_MISSING]) == 0)
return (B_TRUE);
/*
* When a resilver is initiated the scan will assign the scn_max_txg
* value to the highest txg value that exists in all DTLs. If this
@@ -1780,39 +1785,39 @@
* We then add DTL_SCRUB with a refcnt of 2, so that
* entries in the range [0, scrub_txg) will have a
* positive refcnt -- either 1 or 2. We then convert
* the reference tree into the new DTL_MISSING map.
*/
- space_map_ref_create(&reftree);
- space_map_ref_add_map(&reftree,
- &vd->vdev_dtl[DTL_MISSING], 1);
- space_map_ref_add_seg(&reftree, 0, scrub_txg, -1);
- space_map_ref_add_map(&reftree,
- &vd->vdev_dtl[DTL_SCRUB], 2);
- space_map_ref_generate_map(&reftree,
- &vd->vdev_dtl[DTL_MISSING], 1);
- space_map_ref_destroy(&reftree);
+ space_reftree_create(&reftree);
+ space_reftree_add_map(&reftree,
+ vd->vdev_dtl[DTL_MISSING], 1);
+ space_reftree_add_seg(&reftree, 0, scrub_txg, -1);
+ space_reftree_add_map(&reftree,
+ vd->vdev_dtl[DTL_SCRUB], 2);
+ space_reftree_generate_map(&reftree,
+ vd->vdev_dtl[DTL_MISSING], 1);
+ space_reftree_destroy(&reftree);
}
- space_map_vacate(&vd->vdev_dtl[DTL_PARTIAL], NULL, NULL);
- space_map_walk(&vd->vdev_dtl[DTL_MISSING],
- space_map_add, &vd->vdev_dtl[DTL_PARTIAL]);
+ range_tree_vacate(vd->vdev_dtl[DTL_PARTIAL], NULL, NULL);
+ range_tree_walk(vd->vdev_dtl[DTL_MISSING],
+ range_tree_add, vd->vdev_dtl[DTL_PARTIAL]);
if (scrub_done)
- space_map_vacate(&vd->vdev_dtl[DTL_SCRUB], NULL, NULL);
- space_map_vacate(&vd->vdev_dtl[DTL_OUTAGE], NULL, NULL);
+ range_tree_vacate(vd->vdev_dtl[DTL_SCRUB], NULL, NULL);
+ range_tree_vacate(vd->vdev_dtl[DTL_OUTAGE], NULL, NULL);
if (!vdev_readable(vd))
- space_map_add(&vd->vdev_dtl[DTL_OUTAGE], 0, -1ULL);
+ range_tree_add(vd->vdev_dtl[DTL_OUTAGE], 0, -1ULL);
else
- space_map_walk(&vd->vdev_dtl[DTL_MISSING],
- space_map_add, &vd->vdev_dtl[DTL_OUTAGE]);
+ range_tree_walk(vd->vdev_dtl[DTL_MISSING],
+ range_tree_add, vd->vdev_dtl[DTL_OUTAGE]);
/*
* If the vdev was resilvering and no longer has any
* DTLs then reset its resilvering flag.
*/
if (vd->vdev_resilver_txg != 0 &&
- vd->vdev_dtl[DTL_MISSING].sm_space == 0 &&
- vd->vdev_dtl[DTL_OUTAGE].sm_space == 0)
+ range_tree_space(vd->vdev_dtl[DTL_MISSING]) == 0 &&
+ range_tree_space(vd->vdev_dtl[DTL_OUTAGE]) == 0)
vd->vdev_resilver_txg = 0;
mutex_exit(&vd->vdev_dtl_lock);
if (txg != 0)
@@ -1830,117 +1835,135 @@
minref = 1; /* i.e. non-zero */
else if (vd->vdev_nparity != 0)
minref = vd->vdev_nparity + 1; /* RAID-Z */
else
minref = vd->vdev_children; /* any kind of mirror */
- space_map_ref_create(&reftree);
+ space_reftree_create(&reftree);
for (int c = 0; c < vd->vdev_children; c++) {
vdev_t *cvd = vd->vdev_child[c];
mutex_enter(&cvd->vdev_dtl_lock);
- space_map_ref_add_map(&reftree, &cvd->vdev_dtl[s], 1);
+ space_reftree_add_map(&reftree, cvd->vdev_dtl[s], 1);
mutex_exit(&cvd->vdev_dtl_lock);
}
- space_map_ref_generate_map(&reftree, &vd->vdev_dtl[t], minref);
- space_map_ref_destroy(&reftree);
+ space_reftree_generate_map(&reftree, vd->vdev_dtl[t], minref);
+ space_reftree_destroy(&reftree);
}
mutex_exit(&vd->vdev_dtl_lock);
}
-static int
+int
vdev_dtl_load(vdev_t *vd)
{
spa_t *spa = vd->vdev_spa;
- space_map_obj_t *smo = &vd->vdev_dtl_smo;
objset_t *mos = spa->spa_meta_objset;
- dmu_buf_t *db;
- int error;
+ int error = 0;
- ASSERT(vd->vdev_children == 0);
-
- if (smo->smo_object == 0)
- return (0);
-
+ if (vd->vdev_ops->vdev_op_leaf && vd->vdev_dtl_object != 0) {
ASSERT(!vd->vdev_ishole);
- if ((error = dmu_bonus_hold(mos, smo->smo_object, FTAG, &db)) != 0)
+ error = space_map_open(&vd->vdev_dtl_sm, mos,
+ vd->vdev_dtl_object, 0, -1ULL, 0, &vd->vdev_dtl_lock);
+ if (error)
return (error);
+ ASSERT(vd->vdev_dtl_sm != NULL);
- ASSERT3U(db->db_size, >=, sizeof (*smo));
- bcopy(db->db_data, smo, sizeof (*smo));
- dmu_buf_rele(db, FTAG);
-
mutex_enter(&vd->vdev_dtl_lock);
- error = space_map_load(&vd->vdev_dtl[DTL_MISSING],
- NULL, SM_ALLOC, smo, mos);
+
+ /*
+ * Now that we've opened the space_map we need to update
+ * the in-core DTL.
+ */
+ space_map_update(vd->vdev_dtl_sm);
+
+ error = space_map_load(vd->vdev_dtl_sm,
+ vd->vdev_dtl[DTL_MISSING], SM_ALLOC);
mutex_exit(&vd->vdev_dtl_lock);
return (error);
+ }
+
+ for (int c = 0; c < vd->vdev_children; c++) {
+ error = vdev_dtl_load(vd->vdev_child[c]);
+ if (error != 0)
+ break;
+ }
+
+ return (error);
}
void
vdev_dtl_sync(vdev_t *vd, uint64_t txg)
{
spa_t *spa = vd->vdev_spa;
- space_map_obj_t *smo = &vd->vdev_dtl_smo;
- space_map_t *sm = &vd->vdev_dtl[DTL_MISSING];
+ range_tree_t *rt = vd->vdev_dtl[DTL_MISSING];
objset_t *mos = spa->spa_meta_objset;
- space_map_t smsync;
- kmutex_t smlock;
- dmu_buf_t *db;
+ range_tree_t *rtsync;
+ kmutex_t rtlock;
dmu_tx_t *tx;
+ uint64_t object = space_map_object(vd->vdev_dtl_sm);
ASSERT(!vd->vdev_ishole);
+ ASSERT(vd->vdev_ops->vdev_op_leaf);
tx = dmu_tx_create_assigned(spa->spa_dsl_pool, txg);
- if (vd->vdev_detached) {
- if (smo->smo_object != 0) {
- int err = dmu_object_free(mos, smo->smo_object, tx);
- ASSERT0(err);
- smo->smo_object = 0;
- }
+ if (vd->vdev_detached || vd->vdev_top->vdev_removing) {
+ mutex_enter(&vd->vdev_dtl_lock);
+ space_map_free(vd->vdev_dtl_sm, tx);
+ space_map_close(vd->vdev_dtl_sm);
+ vd->vdev_dtl_sm = NULL;
+ mutex_exit(&vd->vdev_dtl_lock);
dmu_tx_commit(tx);
return;
}
- if (smo->smo_object == 0) {
- ASSERT(smo->smo_objsize == 0);
- ASSERT(smo->smo_alloc == 0);
- smo->smo_object = dmu_object_alloc(mos,
- DMU_OT_SPACE_MAP, 1 << SPACE_MAP_BLOCKSHIFT,
- DMU_OT_SPACE_MAP_HEADER, sizeof (*smo), tx);
- ASSERT(smo->smo_object != 0);
- vdev_config_dirty(vd->vdev_top);
+ if (vd->vdev_dtl_sm == NULL) {
+ uint64_t new_object;
+
+ new_object = space_map_alloc(mos, tx);
+ VERIFY3U(new_object, !=, 0);
+
+ VERIFY0(space_map_open(&vd->vdev_dtl_sm, mos, new_object,
+ 0, -1ULL, 0, &vd->vdev_dtl_lock));
+ ASSERT(vd->vdev_dtl_sm != NULL);
}
- mutex_init(&smlock, NULL, MUTEX_DEFAULT, NULL);
+ mutex_init(&rtlock, NULL, MUTEX_DEFAULT, NULL);
- space_map_create(&smsync, sm->sm_start, sm->sm_size, sm->sm_shift,
- &smlock);
+ rtsync = range_tree_create(NULL, NULL, &rtlock);
- mutex_enter(&smlock);
+ mutex_enter(&rtlock);
mutex_enter(&vd->vdev_dtl_lock);
- space_map_walk(sm, space_map_add, &smsync);
+ range_tree_walk(rt, range_tree_add, rtsync);
mutex_exit(&vd->vdev_dtl_lock);
- space_map_truncate(smo, mos, tx);
- space_map_sync(&smsync, SM_ALLOC, smo, mos, tx);
- space_map_vacate(&smsync, NULL, NULL);
+ space_map_truncate(vd->vdev_dtl_sm, tx);
+ space_map_write(vd->vdev_dtl_sm, rtsync, SM_ALLOC, tx);
+ range_tree_vacate(rtsync, NULL, NULL);
- space_map_destroy(&smsync);
+ range_tree_destroy(rtsync);
- mutex_exit(&smlock);
- mutex_destroy(&smlock);
+ mutex_exit(&rtlock);
+ mutex_destroy(&rtlock);
- VERIFY(0 == dmu_bonus_hold(mos, smo->smo_object, FTAG, &db));
- dmu_buf_will_dirty(db, tx);
- ASSERT3U(db->db_size, >=, sizeof (*smo));
- bcopy(smo, db->db_data, sizeof (*smo));
- dmu_buf_rele(db, FTAG);
+ /*
+ * If the object for the space map has changed then dirty
+ * the top level so that we update the config.
+ */
+ if (object != space_map_object(vd->vdev_dtl_sm)) {
+ zfs_dbgmsg("txg %llu, spa %s, DTL old object %llu, "
+ "new object %llu", txg, spa_name(spa), object,
+ space_map_object(vd->vdev_dtl_sm));
+ vdev_config_dirty(vd->vdev_top);
+ }
dmu_tx_commit(tx);
+
+ mutex_enter(&vd->vdev_dtl_lock);
+ space_map_update(vd->vdev_dtl_sm);
+ mutex_exit(&vd->vdev_dtl_lock);
}
/*
* Determine whether the specified vdev can be offlined/detached/removed
* without losing data.
@@ -1985,11 +2008,11 @@
uint64_t thismin = UINT64_MAX;
uint64_t thismax = 0;
if (vd->vdev_children == 0) {
mutex_enter(&vd->vdev_dtl_lock);
- if (vd->vdev_dtl[DTL_MISSING].sm_space != 0 &&
+ if (range_tree_space(vd->vdev_dtl[DTL_MISSING]) != 0 &&
vdev_writeable(vd)) {
thismin = vdev_dtl_min(vd);
thismax = vdev_dtl_max(vd);
needed = B_TRUE;
@@ -2090,33 +2113,29 @@
objset_t *mos = spa->spa_meta_objset;
dmu_tx_t *tx;
tx = dmu_tx_create_assigned(spa_get_dsl(spa), txg);
- if (vd->vdev_dtl_smo.smo_object) {
- ASSERT0(vd->vdev_dtl_smo.smo_alloc);
- (void) dmu_object_free(mos, vd->vdev_dtl_smo.smo_object, tx);
- vd->vdev_dtl_smo.smo_object = 0;
- }
-
if (vd->vdev_ms != NULL) {
for (int m = 0; m < vd->vdev_ms_count; m++) {
metaslab_t *msp = vd->vdev_ms[m];
- if (msp == NULL || msp->ms_smo.smo_object == 0)
+ if (msp == NULL || msp->ms_sm == NULL)
continue;
- ASSERT0(msp->ms_smo.smo_alloc);
- (void) dmu_object_free(mos, msp->ms_smo.smo_object, tx);
- msp->ms_smo.smo_object = 0;
+ mutex_enter(&msp->ms_lock);
+ VERIFY0(space_map_allocated(msp->ms_sm));
+ space_map_free(msp->ms_sm, tx);
+ space_map_close(msp->ms_sm);
+ msp->ms_sm = NULL;
+ mutex_exit(&msp->ms_lock);
}
}
if (vd->vdev_ms_array) {
(void) dmu_object_free(mos, vd->vdev_ms_array, tx);
vd->vdev_ms_array = 0;
- vd->vdev_ms_shift = 0;
}
dmu_tx_commit(tx);
}
void