#include "journal.h"
#include "journal_reclaim.h"
#include "keylist.h"
+#include "recovery.h"
#include "replicas.h"
#include "super-io.h"
struct bkey_s_c k;
struct bkey_s_c_btree_ptr_v2 bp;
struct bkey unpacked;
- char buf1[100], buf2[100];
+ struct printbuf buf1 = PRINTBUF, buf2 = PRINTBUF;
BUG_ON(!b->c.level);
- if (!test_bit(BCH_FS_BTREE_INTERIOR_REPLAY_DONE, &c->flags))
+ if (!test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags))
return;
bch2_btree_node_iter_init_from_start(&iter, b);
if (bpos_cmp(next_node, bp.v->min_key)) {
bch2_dump_btree_node(c, b);
- panic("expected next min_key %s got %s\n",
- (bch2_bpos_to_text(&PBUF(buf1), next_node), buf1),
- (bch2_bpos_to_text(&PBUF(buf2), bp.v->min_key), buf2));
+ bch2_bpos_to_text(&buf1, next_node);
+ bch2_bpos_to_text(&buf2, bp.v->min_key);
+ panic("expected next min_key %s got %s\n", buf1.buf, buf2.buf);
}
bch2_btree_node_iter_advance(&iter, b);
if (bch2_btree_node_iter_end(&iter)) {
if (bpos_cmp(k.k->p, b->key.k.p)) {
bch2_dump_btree_node(c, b);
- panic("expected end %s got %s\n",
- (bch2_bpos_to_text(&PBUF(buf1), b->key.k.p), buf1),
- (bch2_bpos_to_text(&PBUF(buf2), k.k->p), buf2));
+ bch2_bpos_to_text(&buf1, b->key.k.p);
+ bch2_bpos_to_text(&buf2, k.k->p);
+ panic("expected end %s got %s\n", buf1.buf, buf2.buf);
}
break;
}
static struct btree *__bch2_btree_node_alloc(struct bch_fs *c,
struct disk_reservation *res,
struct closure *cl,
+ bool interior_node,
unsigned flags)
{
struct write_point *wp;
if (flags & BTREE_INSERT_USE_RESERVE) {
nr_reserve = 0;
- alloc_reserve = RESERVE_BTREE_MOVINGGC;
+ alloc_reserve = RESERVE_btree_movinggc;
} else {
nr_reserve = BTREE_NODE_RESERVE;
- alloc_reserve = RESERVE_BTREE;
+ alloc_reserve = RESERVE_btree;
}
mutex_lock(&c->btree_reserve_cache_lock);
if (IS_ERR(wp))
return ERR_CAST(wp);
- if (wp->sectors_free < c->opts.btree_node_size) {
+ if (wp->sectors_free < btree_sectors(c)) {
struct open_bucket *ob;
unsigned i;
open_bucket_for_each(c, &wp->ptrs, ob, i)
- if (ob->sectors_free < c->opts.btree_node_size)
+ if (ob->sectors_free < btree_sectors(c))
ob->sectors_free = 0;
bch2_alloc_sectors_done(c, wp);
}
bkey_btree_ptr_v2_init(&tmp.k);
- bch2_alloc_sectors_append_ptrs(c, wp, &tmp.k, c->opts.btree_node_size);
+ bch2_alloc_sectors_append_ptrs(c, wp, &tmp.k, btree_sectors(c), false);
bch2_open_bucket_get(c, wp, &ob);
bch2_alloc_sectors_done(c, wp);
mem_alloc:
- b = bch2_btree_node_mem_alloc(c);
+ b = bch2_btree_node_mem_alloc(c, interior_node);
+ six_unlock_write(&b->c.lock);
+ six_unlock_intent(&b->c.lock);
/* we hold cannibalize_lock: */
BUG_ON(IS_ERR(b));
{
struct bch_fs *c = as->c;
struct btree *b;
+ struct prealloc_nodes *p = &as->prealloc_nodes[!!level];
int ret;
BUG_ON(level >= BTREE_MAX_DEPTH);
- BUG_ON(!as->nr_prealloc_nodes);
+ BUG_ON(!p->nr);
- b = as->prealloc_nodes[--as->nr_prealloc_nodes];
+ b = p->b[--p->nr];
+
+ six_lock_intent(&b->c.lock, NULL, NULL);
+ six_lock_write(&b->c.lock, NULL, NULL);
set_btree_node_accessed(b);
- set_btree_node_dirty(c, b);
+ set_btree_node_dirty_acct(c, b);
set_btree_node_need_write(b);
bch2_bset_init_first(b, &b->data->keys);
static void bch2_btree_reserve_put(struct btree_update *as)
{
struct bch_fs *c = as->c;
+ struct prealloc_nodes *p;
mutex_lock(&c->btree_reserve_cache_lock);
- while (as->nr_prealloc_nodes) {
- struct btree *b = as->prealloc_nodes[--as->nr_prealloc_nodes];
+ for (p = as->prealloc_nodes;
+ p < as->prealloc_nodes + ARRAY_SIZE(as->prealloc_nodes);
+ p++) {
+ while (p->nr) {
+ struct btree *b = p->b[--p->nr];
- six_unlock_write(&b->c.lock);
-
- if (c->btree_reserve_cache_nr <
- ARRAY_SIZE(c->btree_reserve_cache)) {
- struct btree_alloc *a =
- &c->btree_reserve_cache[c->btree_reserve_cache_nr++];
+ six_lock_intent(&b->c.lock, NULL, NULL);
+ six_lock_write(&b->c.lock, NULL, NULL);
- a->ob = b->ob;
- b->ob.nr = 0;
- bkey_copy(&a->k, &b->key);
- } else {
- bch2_open_buckets_put(c, &b->ob);
- }
+ if (c->btree_reserve_cache_nr <
+ ARRAY_SIZE(c->btree_reserve_cache)) {
+ struct btree_alloc *a =
+ &c->btree_reserve_cache[c->btree_reserve_cache_nr++];
- btree_node_lock_type(c, b, SIX_LOCK_write);
- __btree_node_free(c, b);
- six_unlock_write(&b->c.lock);
+ a->ob = b->ob;
+ b->ob.nr = 0;
+ bkey_copy(&a->k, &b->key);
+ } else {
+ bch2_open_buckets_put(c, &b->ob);
+ }
- six_unlock_intent(&b->c.lock);
+ __btree_node_free(c, b);
+ six_unlock_write(&b->c.lock);
+ six_unlock_intent(&b->c.lock);
+ }
}
mutex_unlock(&c->btree_reserve_cache_lock);
}
-static int bch2_btree_reserve_get(struct btree_update *as, unsigned nr_nodes,
- unsigned flags, struct closure *cl)
+static int bch2_btree_reserve_get(struct btree_update *as,
+ unsigned nr_nodes[2],
+ unsigned flags)
{
struct bch_fs *c = as->c;
+ struct closure cl;
struct btree *b;
+ unsigned interior;
int ret;
- BUG_ON(nr_nodes > BTREE_RESERVE_MAX);
+ closure_init_stack(&cl);
+retry:
+
+ BUG_ON(nr_nodes[0] + nr_nodes[1] > BTREE_RESERVE_MAX);
/*
* Protects reaping from the btree node cache and using the btree node
* open bucket reserve:
+ *
+ * BTREE_INSERT_NOWAIT only applies to btree node allocation, not
+ * blocking on this lock:
*/
- ret = bch2_btree_cache_cannibalize_lock(c, cl);
+ ret = bch2_btree_cache_cannibalize_lock(c, &cl);
if (ret)
- return ret;
-
- while (as->nr_prealloc_nodes < nr_nodes) {
- b = __bch2_btree_node_alloc(c, &as->disk_res,
- flags & BTREE_INSERT_NOWAIT
- ? NULL : cl, flags);
- if (IS_ERR(b)) {
- ret = PTR_ERR(b);
- goto err_free;
- }
+ goto err;
- as->prealloc_nodes[as->nr_prealloc_nodes++] = b;
+ for (interior = 0; interior < 2; interior++) {
+ struct prealloc_nodes *p = as->prealloc_nodes + interior;
+
+ while (p->nr < nr_nodes[interior]) {
+ b = __bch2_btree_node_alloc(c, &as->disk_res,
+ flags & BTREE_INSERT_NOWAIT
+ ? NULL : &cl,
+ interior, flags);
+ if (IS_ERR(b)) {
+ ret = PTR_ERR(b);
+ goto err;
+ }
+
+ p->b[p->nr++] = b;
+ }
}
bch2_btree_cache_cannibalize_unlock(c);
+ closure_sync(&cl);
return 0;
-err_free:
+err:
bch2_btree_cache_cannibalize_unlock(c);
- trace_btree_reserve_get_fail(c, nr_nodes, cl);
+ closure_sync(&cl);
+
+ if (ret == -EAGAIN)
+ goto retry;
+
+ trace_btree_reserve_get_fail(c, nr_nodes[0] + nr_nodes[1], &cl);
return ret;
}
bch2_disk_reservation_put(c, &as->disk_res);
bch2_btree_reserve_put(as);
+ bch2_time_stats_update(&c->times[BCH_TIME_btree_interior_update_total],
+ as->start_time);
+
mutex_lock(&c->btree_interior_update_lock);
list_del(&as->unwritten_list);
list_del(&as->list);
- mutex_unlock(&c->btree_interior_update_lock);
closure_debug_destroy(&as->cl);
mempool_free(as, &c->btree_interior_update_pool);
+ /*
+ * Have to do the wakeup with btree_interior_update_lock still held,
+ * since being on btree_interior_update_list is our ref on @c:
+ */
closure_wake_up(&c->btree_interior_update_wait);
+
+ mutex_unlock(&c->btree_interior_update_lock);
}
static void btree_update_will_delete_key(struct btree_update *as,
trans->journal_pin = &as->journal;
for_each_keylist_key(&as->new_keys, k) {
- ret = bch2_trans_mark_key(trans,
- bkey_s_c_null,
- bkey_i_to_s_c(k),
- BTREE_TRIGGER_INSERT);
+ ret = bch2_trans_mark_new(trans, k, 0);
if (ret)
return ret;
}
for_each_keylist_key(&as->old_keys, k) {
- ret = bch2_trans_mark_key(trans,
- bkey_i_to_s_c(k),
- bkey_s_c_null,
- BTREE_TRIGGER_OVERWRITE);
+ ret = bch2_trans_mark_old(trans, bkey_i_to_s_c(k), 0);
if (ret)
return ret;
}
if (ret)
goto err;
- BUG_ON(!journal_pin_active(&as->journal));
-
/*
* Wait for any in flight writes to finish before we free the old nodes
* on disk:
BTREE_INSERT_NOFAIL|
BTREE_INSERT_NOCHECK_RW|
BTREE_INSERT_JOURNAL_RECLAIM|
- BTREE_INSERT_JOURNAL_RESERVED,
+ JOURNAL_WATERMARK_reserved,
btree_update_nodes_written_trans(&trans, as));
bch2_trans_exit(&trans);
* we're in journal error state:
*/
- btree_node_lock_type(c, b, SIX_LOCK_intent);
- btree_node_lock_type(c, b, SIX_LOCK_write);
+ six_lock_intent(&b->c.lock, NULL, NULL);
+ six_lock_write(&b->c.lock, NULL, NULL);
mutex_lock(&c->btree_interior_update_lock);
list_del(&as->write_blocked_list);
+ if (list_empty(&b->write_blocked))
+ clear_btree_node_write_blocked(b);
/*
* Node might have been freed, recheck under
BUG_ON(b->will_make_reachable != (unsigned long) as);
b->will_make_reachable = 0;
+ clear_btree_node_will_make_reachable(b);
}
mutex_unlock(&c->btree_interior_update_lock);
for (i = 0; i < as->nr_new_nodes; i++) {
b = as->new_nodes[i];
- btree_node_lock_type(c, b, SIX_LOCK_read);
+ six_lock_read(&b->c.lock, NULL, NULL);
btree_node_write_if_need(c, b, SIX_LOCK_read);
six_unlock_read(&b->c.lock);
}
as->mode = BTREE_INTERIOR_UPDATING_NODE;
as->b = b;
+
+ set_btree_node_write_blocked(b);
list_add(&as->write_blocked_list, &b->write_blocked);
mutex_unlock(&c->btree_interior_update_lock);
as->new_nodes[as->nr_new_nodes++] = b;
b->will_make_reachable = 1UL|(unsigned long) as;
+ set_btree_node_will_make_reachable(b);
mutex_unlock(&c->btree_interior_update_lock);
* xchg() is for synchronization with bch2_btree_complete_write:
*/
v = xchg(&b->will_make_reachable, 0);
+ clear_btree_node_will_make_reachable(b);
as = (struct btree_update *) (v & ~1UL);
if (!as) {
closure_wake_up(&c->btree_interior_update_wait);
}
- clear_btree_node_dirty(c, b);
+ clear_btree_node_dirty_acct(c, b);
clear_btree_node_need_write(b);
/*
static void bch2_btree_update_done(struct btree_update *as)
{
+ struct bch_fs *c = as->c;
+ u64 start_time = as->start_time;
+
BUG_ON(as->mode == BTREE_INTERIOR_NO_UPDATE);
if (as->took_gc_lock)
continue_at(&as->cl, btree_update_set_nodes_written,
as->c->btree_interior_update_worker);
+
+ bch2_time_stats_update(&c->times[BCH_TIME_btree_interior_update_foreground],
+ start_time);
}
static struct btree_update *
bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
- unsigned level, unsigned nr_nodes, unsigned flags)
+ unsigned level, bool split, unsigned flags)
{
struct bch_fs *c = trans->c;
struct btree_update *as;
- struct closure cl;
+ u64 start_time = local_clock();
int disk_res_flags = (flags & BTREE_INSERT_NOFAIL)
? BCH_DISK_RESERVATION_NOFAIL : 0;
- int journal_flags = 0;
+ unsigned nr_nodes[2] = { 0, 0 };
+ unsigned update_level = level;
+ int journal_flags = flags & JOURNAL_WATERMARK_MASK;
int ret = 0;
BUG_ON(!path->should_be_locked);
- if (flags & BTREE_INSERT_JOURNAL_RESERVED)
- journal_flags |= JOURNAL_RES_GET_RESERVED;
+ if (flags & BTREE_INSERT_JOURNAL_RECLAIM)
+ journal_flags |= JOURNAL_RES_GET_NONBLOCK;
- closure_init_stack(&cl);
-retry:
+ while (1) {
+ nr_nodes[!!update_level] += 1 + split;
+ update_level++;
+
+ if (!btree_path_node(path, update_level))
+ break;
+
+ /*
+ * XXX: figure out how far we might need to split,
+ * instead of locking/reserving all the way to the root:
+ */
+ split = update_level + 1 < BTREE_MAX_DEPTH;
+ }
+
+ /* Might have to allocate a new root: */
+ if (update_level < BTREE_MAX_DEPTH)
+ nr_nodes[1] += 1;
- /*
- * XXX: figure out how far we might need to split,
- * instead of locking/reserving all the way to the root:
- */
if (!bch2_btree_path_upgrade(trans, path, U8_MAX)) {
- trace_trans_restart_iter_upgrade(trans->ip, _RET_IP_,
+ trace_trans_restart_iter_upgrade(trans->fn, _RET_IP_,
path->btree_id, &path->pos);
ret = btree_trans_restart(trans);
return ERR_PTR(ret);
memset(as, 0, sizeof(*as));
closure_init(&as->cl, NULL);
as->c = c;
+ as->start_time = start_time;
as->mode = BTREE_INTERIOR_NO_UPDATE;
as->took_gc_lock = !(flags & BTREE_INSERT_GC_LOCK_HELD);
as->btree_id = path->btree_id;
if (ret)
goto err;
+ bch2_trans_unlock(trans);
+
ret = bch2_journal_preres_get(&c->journal, &as->journal_preres,
BTREE_UPDATE_JOURNAL_RES,
- journal_flags|JOURNAL_RES_GET_NONBLOCK);
- if (ret == -EAGAIN) {
- bch2_trans_unlock(trans);
-
- if (flags & BTREE_INSERT_JOURNAL_RECLAIM) {
- bch2_btree_update_free(as);
- btree_trans_restart(trans);
- return ERR_PTR(ret);
- }
-
- ret = bch2_journal_preres_get(&c->journal, &as->journal_preres,
- BTREE_UPDATE_JOURNAL_RES,
- journal_flags);
- if (ret) {
- trace_trans_restart_journal_preres_get(trans->ip, _RET_IP_);
- goto err;
- }
-
- if (!bch2_trans_relock(trans)) {
- ret = -EINTR;
- goto err;
- }
+ journal_flags);
+ if (ret) {
+ bch2_btree_update_free(as);
+ trace_trans_restart_journal_preres_get(trans->fn, _RET_IP_);
+ btree_trans_restart(trans);
+ return ERR_PTR(ret);
}
ret = bch2_disk_reservation_get(c, &as->disk_res,
- nr_nodes * c->opts.btree_node_size,
+ (nr_nodes[0] + nr_nodes[1]) * btree_sectors(c),
c->opts.metadata_replicas,
disk_res_flags);
if (ret)
goto err;
- ret = bch2_btree_reserve_get(as, nr_nodes, flags, &cl);
+ ret = bch2_btree_reserve_get(as, nr_nodes, flags);
if (ret)
goto err;
- bch2_journal_pin_add(&c->journal,
- atomic64_read(&c->journal.seq),
- &as->journal, NULL);
+ if (!bch2_trans_relock(trans)) {
+ ret = -EINTR;
+ goto err;
+ }
return as;
err:
bch2_btree_update_free(as);
-
- if (ret == -EAGAIN) {
- bch2_trans_unlock(trans);
- closure_sync(&cl);
- ret = -EINTR;
- }
-
- if (ret == -EINTR && bch2_trans_relock(trans))
- goto retry;
-
return ERR_PTR(ret);
}
struct btree *old;
trace_btree_set_root(c, b);
- BUG_ON(!b->written &&
- !test_bit(BCH_FS_HOLD_BTREE_WRITES, &c->flags));
+ BUG_ON(!b->written);
old = btree_node_root(c, b);
BUG_ON(insert->k.type == KEY_TYPE_btree_ptr_v2 &&
!btree_ptr_sectors_written(insert));
+ if (unlikely(!test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags)))
+ bch2_journal_key_overwritten(c, b->c.btree_id, b->c.level, insert->k.p);
+
invalid = bch2_bkey_invalid(c, bkey_i_to_s_c(insert), btree_node_type(b)) ?:
bch2_bkey_in_btree_node(b, bkey_i_to_s_c(insert));
if (invalid) {
- char buf[160];
+ struct printbuf buf = PRINTBUF;
- bch2_bkey_val_to_text(&PBUF(buf), c, bkey_i_to_s_c(insert));
- bch2_fs_inconsistent(c, "inserting invalid bkey %s: %s", buf, invalid);
+ bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(insert));
+ bch2_fs_inconsistent(c, "inserting invalid bkey %s: %s", buf.buf, invalid);
+ printbuf_exit(&buf);
dump_stack();
}
bch2_btree_node_iter_advance(node_iter, b);
bch2_btree_bset_insert_key(trans, path, b, node_iter, insert);
- set_btree_node_dirty(c, b);
+ set_btree_node_dirty_acct(c, b);
set_btree_node_need_write(b);
}
six_unlock_write(&n2->c.lock);
six_unlock_write(&n1->c.lock);
- bch2_btree_node_write(c, n1, SIX_LOCK_intent);
- bch2_btree_node_write(c, n2, SIX_LOCK_intent);
+ bch2_btree_node_write(c, n1, SIX_LOCK_intent, 0);
+ bch2_btree_node_write(c, n2, SIX_LOCK_intent, 0);
/*
* Note that on recursive parent_keys == keys, so we
btree_split_insert_keys(as, trans, path, n3, &as->parent_keys);
- bch2_btree_node_write(c, n3, SIX_LOCK_intent);
+ bch2_btree_node_write(c, n3, SIX_LOCK_intent, 0);
}
} else {
trace_btree_compact(c, b);
bch2_btree_build_aux_trees(n1);
six_unlock_write(&n1->c.lock);
- bch2_btree_node_write(c, n1, SIX_LOCK_intent);
+ bch2_btree_node_write(c, n1, SIX_LOCK_intent, 0);
if (parent)
bch2_keylist_add(&as->parent_keys, &n1->key);
bch2_trans_verify_locks(trans);
- bch2_time_stats_update(&c->times[BCH_TIME_btree_node_split],
+ bch2_time_stats_update(&c->times[n2
+ ? BCH_TIME_btree_node_split
+ : BCH_TIME_btree_node_compact],
start_time);
}
struct btree_path *path,
unsigned flags)
{
- struct bch_fs *c = trans->c;
struct btree *b = path_l(path)->b;
struct btree_update *as;
unsigned l;
int ret = 0;
as = bch2_btree_update_start(trans, path, path->level,
- btree_update_reserve_required(c, b), flags);
+ true, flags);
if (IS_ERR(as))
return PTR_ERR(as);
struct btree *b, *m, *n, *prev, *next, *parent;
struct bpos sib_pos;
size_t sib_u64s;
+ u64 start_time = local_clock();
int ret = 0;
BUG_ON(!path->should_be_locked);
? bpos_predecessor(b->data->min_key)
: bpos_successor(b->data->max_key);
- sib_path = bch2_path_get(trans, false, path->btree_id,
- sib_pos, U8_MAX, level, true);
+ sib_path = bch2_path_get(trans, path->btree_id, sib_pos,
+ U8_MAX, level, BTREE_ITER_INTENT, _THIS_IP_);
ret = bch2_btree_path_traverse(trans, sib_path, false);
if (ret)
goto err;
}
if (bkey_cmp(bpos_successor(prev->data->max_key), next->data->min_key)) {
- char buf1[100], buf2[100];
+ struct printbuf buf1 = PRINTBUF, buf2 = PRINTBUF;
- bch2_bpos_to_text(&PBUF(buf1), prev->data->max_key);
- bch2_bpos_to_text(&PBUF(buf2), next->data->min_key);
+ bch2_bpos_to_text(&buf1, prev->data->max_key);
+ bch2_bpos_to_text(&buf2, next->data->min_key);
bch_err(c,
"btree topology error in btree merge:\n"
" prev ends at %s\n"
" next starts at %s",
- buf1, buf2);
+ buf1.buf, buf2.buf);
+ printbuf_exit(&buf1);
+ printbuf_exit(&buf2);
bch2_topology_error(c);
ret = -EIO;
goto err;
goto out;
parent = btree_node_parent(path, b);
- as = bch2_btree_update_start(trans, path, level,
- btree_update_reserve_required(c, parent) + 1,
- flags|
+ as = bch2_btree_update_start(trans, path, level, false,
BTREE_INSERT_NOFAIL|
- BTREE_INSERT_USE_RESERVE);
+ BTREE_INSERT_USE_RESERVE|
+ flags);
ret = PTR_ERR_OR_ZERO(as);
if (ret)
goto err;
n = bch2_btree_node_alloc(as, b->c.level);
bch2_btree_update_add_new_node(as, n);
+ SET_BTREE_NODE_SEQ(n->data,
+ max(BTREE_NODE_SEQ(b->data),
+ BTREE_NODE_SEQ(m->data)) + 1);
+
btree_set_min(n, prev->data->min_key);
btree_set_max(n, next->data->max_key);
n->data->format = new_f;
bch2_btree_build_aux_trees(n);
six_unlock_write(&n->c.lock);
- bch2_btree_node_write(c, n, SIX_LOCK_intent);
+ bch2_btree_node_write(c, n, SIX_LOCK_intent, 0);
bkey_init(&delete.k);
delete.k.p = prev->key.k.p;
six_unlock_intent(&n->c.lock);
bch2_btree_update_done(as);
+
+ bch2_time_stats_update(&c->times[BCH_TIME_btree_node_merge], start_time);
out:
err:
bch2_path_put(trans, sib_path, true);
*/
int bch2_btree_node_rewrite(struct btree_trans *trans,
struct btree_iter *iter,
- __le64 seq, unsigned flags)
+ struct btree *b,
+ unsigned flags)
{
struct bch_fs *c = trans->c;
- struct btree *b, *n, *parent;
+ struct btree *n, *parent;
struct btree_update *as;
int ret;
flags |= BTREE_INSERT_NOFAIL;
-retry:
- ret = bch2_btree_iter_traverse(iter);
- if (ret)
- goto out;
-
- b = bch2_btree_iter_peek_node(iter);
- ret = PTR_ERR_OR_ZERO(b);
- if (ret)
- goto out;
-
- if (!b || b->data->keys.seq != seq)
- goto out;
parent = btree_node_parent(iter->path, b);
as = bch2_btree_update_start(trans, iter->path, b->c.level,
- (parent
- ? btree_update_reserve_required(c, parent)
- : 0) + 1,
- flags);
+ false, flags);
ret = PTR_ERR_OR_ZERO(as);
- if (ret == -EINTR)
- goto retry;
if (ret) {
trace_btree_gc_rewrite_node_fail(c, b);
goto out;
trace_btree_gc_rewrite_node(c, b);
- bch2_btree_node_write(c, n, SIX_LOCK_intent);
+ bch2_btree_node_write(c, n, SIX_LOCK_intent, 0);
if (parent) {
bch2_keylist_add(&as->parent_keys, &n->key);
__le64 seq;
};
+static int async_btree_node_rewrite_trans(struct btree_trans *trans,
+ struct async_btree_rewrite *a)
+{
+ struct btree_iter iter;
+ struct btree *b;
+ int ret;
+
+ bch2_trans_node_iter_init(trans, &iter, a->btree_id, a->pos,
+ BTREE_MAX_DEPTH, a->level, 0);
+ b = bch2_btree_iter_peek_node(&iter);
+ ret = PTR_ERR_OR_ZERO(b);
+ if (ret)
+ goto out;
+
+ if (!b || b->data->keys.seq != a->seq)
+ goto out;
+
+ ret = bch2_btree_node_rewrite(trans, &iter, b, 0);
+out :
+ bch2_trans_iter_exit(trans, &iter);
+
+ return ret;
+}
+
void async_btree_node_rewrite_work(struct work_struct *work)
{
struct async_btree_rewrite *a =
container_of(work, struct async_btree_rewrite, work);
struct bch_fs *c = a->c;
- struct btree_trans trans;
- struct btree_iter iter;
- bch2_trans_init(&trans, c, 0, 0);
- bch2_trans_node_iter_init(&trans, &iter, a->btree_id, a->pos,
- BTREE_MAX_DEPTH, a->level, 0);
- bch2_btree_node_rewrite(&trans, &iter, a->seq, 0);
- bch2_trans_iter_exit(&trans, &iter);
- bch2_trans_exit(&trans);
+ bch2_trans_do(c, NULL, NULL, 0,
+ async_btree_node_rewrite_trans(&trans, a));
percpu_ref_put(&c->writes);
kfree(a);
}
{
struct async_btree_rewrite *a;
- if (!test_bit(BCH_FS_BTREE_INTERIOR_REPLAY_DONE, &c->flags))
- return;
-
if (!percpu_ref_tryget(&c->writes))
return;
int ret;
if (!skip_triggers) {
- ret = bch2_trans_mark_key(trans,
- bkey_s_c_null,
- bkey_i_to_s_c(new_key),
- BTREE_TRIGGER_INSERT);
+ ret = bch2_trans_mark_new(trans, new_key, 0);
if (ret)
return ret;
- ret = bch2_trans_mark_key(trans,
- bkey_i_to_s_c(&b->key),
- bkey_s_c_null,
- BTREE_TRIGGER_OVERWRITE);
+ ret = bch2_trans_mark_old(trans, bkey_i_to_s_c(&b->key), 0);
if (ret)
return ret;
}
bch2_trans_copy_iter(&iter2, iter);
iter2.path = bch2_btree_path_make_mut(trans, iter2.path,
- iter2.flags & BTREE_ITER_INTENT);
+ iter2.flags & BTREE_ITER_INTENT,
+ _THIS_IP_);
BUG_ON(iter2.path->level != b->c.level);
BUG_ON(bpos_cmp(iter2.path->pos, new_key->k.p));
path_l(iter2.path)->b = BTREE_ITER_NO_NODE_UP;
iter2.path->level++;
+ bch2_btree_path_check_sort(trans, iter2.path, 0);
+
ret = bch2_btree_iter_traverse(&iter2) ?:
bch2_trans_update(trans, &iter2, new_key, BTREE_TRIGGER_NORUN);
if (ret)
ret = bch2_trans_commit(trans, NULL, NULL,
BTREE_INSERT_NOFAIL|
BTREE_INSERT_NOCHECK_RW|
+ BTREE_INSERT_USE_RESERVE|
BTREE_INSERT_JOURNAL_RECLAIM|
- BTREE_INSERT_JOURNAL_RESERVED);
+ JOURNAL_WATERMARK_reserved);
if (ret)
goto err;
return -EINTR;
}
- new_hash = bch2_btree_node_mem_alloc(c);
+ new_hash = bch2_btree_node_mem_alloc(c, false);
}
path->intent_ref++;
closure_sync(&cl);
} while (ret);
- b = bch2_btree_node_mem_alloc(c);
+ b = bch2_btree_node_mem_alloc(c, false);
bch2_btree_cache_cannibalize_unlock(c);
set_btree_node_fake(b);
projects / bcachefs-tools-debian / blobdiff