/*
* Load the space map disk into the specified range tree. Segments of maptype
* are added to the range tree, other segment types are removed.
*
* Note: space_map_load() will drop sm_lock across dmu_read() calls.
* The caller must be OK with this.
*/
int
space_map_load(space_map_t *sm, range_tree_t *rt, maptype_t maptype)
{
uint64_t *entry, *entry_map, *entry_map_end;
uint64_t bufsize, size, offset, end, space;
int error = 0;
ASSERT(MUTEX_HELD(sm->sm_lock));
end = space_map_length(sm);
space = space_map_allocated(sm);
VERIFY0(range_tree_space(rt));
if (maptype == SM_FREE) {
range_tree_add(rt, sm->sm_start, sm->sm_size);
space = sm->sm_size - space;
}
bufsize = MAX(sm->sm_blksz, SPA_MINBLOCKSIZE);
entry_map = zio_buf_alloc(bufsize);
mutex_exit(sm->sm_lock);
if (end > bufsize) {
dmu_prefetch(sm->sm_os, space_map_object(sm), 0, bufsize,
end - bufsize, ZIO_PRIORITY_SYNC_READ);
}
mutex_enter(sm->sm_lock);
for (offset = 0; offset < end; offset += bufsize) {
size = MIN(end - offset, bufsize);
VERIFY(P2PHASE(size, sizeof (uint64_t)) == 0);
VERIFY(size != 0);
ASSERT3U(sm->sm_blksz, !=, 0);
dprintf("object=%llu offset=%llx size=%llx\n",
space_map_object(sm), offset, size);
mutex_exit(sm->sm_lock);
error = dmu_read(sm->sm_os, space_map_object(sm), offset, size,
entry_map, DMU_READ_PREFETCH);
mutex_enter(sm->sm_lock);
if (error != 0)
break;
entry_map_end = entry_map + (size / sizeof (uint64_t));
for (entry = entry_map; entry < entry_map_end; entry++) {
uint64_t e = *entry;
uint64_t offset, size;
if (SM_DEBUG_DECODE(e)) /* Skip debug entries */
continue;
offset = (SM_OFFSET_DECODE(e) << sm->sm_shift) +
sm->sm_start;
size = SM_RUN_DECODE(e) << sm->sm_shift;
VERIFY0(P2PHASE(offset, 1ULL << sm->sm_shift));
VERIFY0(P2PHASE(size, 1ULL << sm->sm_shift));
VERIFY3U(offset, >=, sm->sm_start);
VERIFY3U(offset + size, <=, sm->sm_start + sm->sm_size);
if (SM_TYPE_DECODE(e) == maptype) {
VERIFY3U(range_tree_space(rt) + size, <=,
sm->sm_size);
range_tree_add(rt, offset, size);
} else {
range_tree_remove(rt, offset, size);
}
}
boolean_t
vdev_indirect_should_condense(vdev_t *vd)
{
vdev_indirect_mapping_t *vim = vd->vdev_indirect_mapping;
spa_t *spa = vd->vdev_spa;
ASSERT(dsl_pool_sync_context(spa->spa_dsl_pool));
if (!zfs_condense_indirect_vdevs_enable)
return (B_FALSE);
/*
* We can only condense one indirect vdev at a time.
*/
if (spa->spa_condensing_indirect != NULL)
return (B_FALSE);
if (spa_shutting_down(spa))
return (B_FALSE);
/*
* The mapping object size must not change while we are
* condensing, so we can only condense indirect vdevs
* (not vdevs that are still in the middle of being removed).
*/
if (vd->vdev_ops != &vdev_indirect_ops)
return (B_FALSE);
/*
* If nothing new has been marked obsolete, there is no
* point in condensing.
*/
if (vd->vdev_obsolete_sm == NULL) {
ASSERT0(vdev_obsolete_sm_object(vd));
return (B_FALSE);
}
ASSERT(vd->vdev_obsolete_sm != NULL);
ASSERT3U(vdev_obsolete_sm_object(vd), ==,
space_map_object(vd->vdev_obsolete_sm));
uint64_t bytes_mapped = vdev_indirect_mapping_bytes_mapped(vim);
uint64_t bytes_obsolete = space_map_allocated(vd->vdev_obsolete_sm);
uint64_t mapping_size = vdev_indirect_mapping_size(vim);
uint64_t obsolete_sm_size = space_map_length(vd->vdev_obsolete_sm);
ASSERT3U(bytes_obsolete, <=, bytes_mapped);
/*
* If a high percentage of the bytes that are mapped have become
* obsolete, condense (unless the mapping is already small enough).
* This has a good chance of reducing the amount of memory used
* by the mapping.
*/
if (bytes_obsolete * 100 / bytes_mapped >=
zfs_indirect_condense_obsolete_pct &&
mapping_size > zfs_condense_min_mapping_bytes) {
zfs_dbgmsg("should condense vdev %llu because obsolete "
"spacemap covers %d%% of %lluMB mapping",
(u_longlong_t)vd->vdev_id,
(int)(bytes_obsolete * 100 / bytes_mapped),
(u_longlong_t)bytes_mapped / 1024 / 1024);
return (B_TRUE);
}
/*
* If the obsolete space map takes up too much space on disk,
* condense in order to free up this disk space.
*/
if (obsolete_sm_size >= zfs_condense_max_obsolete_bytes) {
zfs_dbgmsg("should condense vdev %llu because obsolete sm "
"length %lluMB >= max size %lluMB",
(u_longlong_t)vd->vdev_id,
(u_longlong_t)obsolete_sm_size / 1024 / 1024,
(u_longlong_t)zfs_condense_max_obsolete_bytes /
1024 / 1024);
return (B_TRUE);
}
return (B_FALSE);
}
