-eabde7cb370040434991209cc3644c45079900da
+f1c9030ccbf6d7b5c46f08f92ee878bfc9f6ee6b
return 0;
}
+/*
+ * This synthesizes deleted extents for holes, similar to BTREE_ITER_SLOTS for
+ * extents style btrees, but works on non-extents btrees:
+ */
+struct bkey_s_c bch2_get_key_or_hole(struct btree_iter *iter, struct bpos end, struct bkey *hole)
+{
+ struct bkey_s_c k = bch2_btree_iter_peek_slot(iter);
+
+ if (bkey_err(k))
+ return k;
+
+ if (k.k->type) {
+ return k;
+ } else {
+ struct btree_iter iter2;
+ struct bpos next;
+
+ bch2_trans_copy_iter(&iter2, iter);
+ k = bch2_btree_iter_peek_upto(&iter2,
+ bkey_min(bkey_min(end,
+ iter->path->l[0].b->key.k.p),
+ POS(iter->pos.inode, iter->pos.offset + U32_MAX - 1)));
+ next = iter2.pos;
+ bch2_trans_iter_exit(iter->trans, &iter2);
+
+ BUG_ON(next.offset >= iter->pos.offset + U32_MAX);
+
+ if (bkey_err(k))
+ return k;
+
+ bkey_init(hole);
+ hole->p = iter->pos;
+
+ bch2_key_resize(hole, next.offset - iter->pos.offset);
+ return (struct bkey_s_c) { hole, NULL };
+ }
+}
+
+static bool next_bucket(struct bch_fs *c, struct bpos *bucket)
+{
+ struct bch_dev *ca;
+ unsigned iter;
+
+ if (bch2_dev_bucket_exists(c, *bucket))
+ return true;
+
+ if (bch2_dev_exists2(c, bucket->inode)) {
+ ca = bch_dev_bkey_exists(c, bucket->inode);
+
+ if (bucket->offset < ca->mi.first_bucket) {
+ bucket->offset = ca->mi.first_bucket;
+ return true;
+ }
+
+ bucket->inode++;
+ bucket->offset = 0;
+ }
+
+ rcu_read_lock();
+ iter = bucket->inode;
+ ca = __bch2_next_dev(c, &iter, NULL);
+ if (ca)
+ bucket->offset = ca->mi.first_bucket;
+ rcu_read_unlock();
+
+ return ca != NULL;
+}
+
+struct bkey_s_c bch2_get_key_or_real_bucket_hole(struct btree_iter *iter, struct bkey *hole)
+{
+ struct bch_fs *c = iter->trans->c;
+ struct bkey_s_c k;
+again:
+ k = bch2_get_key_or_hole(iter, POS_MAX, hole);
+ if (bkey_err(k))
+ return k;
+
+ if (!k.k->type) {
+ struct bpos bucket = bkey_start_pos(k.k);
+
+ if (!bch2_dev_bucket_exists(c, bucket)) {
+ if (!next_bucket(c, &bucket))
+ return bkey_s_c_null;
+
+ bch2_btree_iter_set_pos(iter, bucket);
+ goto again;
+ }
+
+ if (!bch2_dev_bucket_exists(c, k.k->p)) {
+ struct bch_dev *ca = bch_dev_bkey_exists(c, bucket.inode);
+
+ bch2_key_resize(hole, ca->mi.nbuckets - bucket.offset);
+ }
+ }
+
+ return k;
+}
+
static int bch2_check_alloc_key(struct btree_trans *trans,
+ struct bkey_s_c alloc_k,
struct btree_iter *alloc_iter,
struct btree_iter *discard_iter,
struct btree_iter *freespace_iter)
struct bch_alloc_v4 a_convert;
const struct bch_alloc_v4 *a;
unsigned discard_key_type, freespace_key_type;
- struct bkey_s_c alloc_k, k;
+ struct bkey_s_c k;
struct printbuf buf = PRINTBUF;
int ret;
- alloc_k = bch2_dev_bucket_exists(c, alloc_iter->pos)
- ? bch2_btree_iter_peek_slot(alloc_iter)
- : bch2_btree_iter_peek(alloc_iter);
- if (!alloc_k.k)
- return 1;
-
- ret = bkey_err(alloc_k);
- if (ret)
- return ret;
-
if (fsck_err_on(!bch2_dev_bucket_exists(c, alloc_k.k->p), c,
"alloc key for invalid device:bucket %llu:%llu",
alloc_k.k->p.inode, alloc_k.k->p.offset))
return ret;
}
+static int bch2_check_alloc_hole(struct btree_trans *trans,
+ struct bpos start,
+ struct bpos *end,
+ struct btree_iter *freespace_iter)
+{
+ struct bch_fs *c = trans->c;
+ struct bch_dev *ca;
+ struct bkey_s_c k;
+ struct printbuf buf = PRINTBUF;
+ int ret;
+
+ ca = bch_dev_bkey_exists(c, start.inode);
+ if (!ca->mi.freespace_initialized)
+ return 0;
+
+ bch2_btree_iter_set_pos(freespace_iter, start);
+
+ k = bch2_btree_iter_peek_slot(freespace_iter);
+ ret = bkey_err(k);
+ if (ret)
+ goto err;
+
+ *end = bkey_min(k.k->p, *end);
+
+ if (k.k->type != KEY_TYPE_set &&
+ (c->opts.reconstruct_alloc ||
+ fsck_err(c, "hole in alloc btree missing in freespace btree\n"
+ " device %llu buckets %llu-%llu",
+ freespace_iter->pos.inode,
+ freespace_iter->pos.offset,
+ end->offset))) {
+ struct bkey_i *update =
+ bch2_trans_kmalloc(trans, sizeof(*update));
+
+ ret = PTR_ERR_OR_ZERO(update);
+ if (ret)
+ goto err;
+
+ bkey_init(&update->k);
+ update->k.type = KEY_TYPE_set;
+ update->k.p = freespace_iter->pos;
+ bch2_key_resize(&update->k,
+ min_t(u64, U32_MAX, end->offset -
+ freespace_iter->pos.offset));
+
+ ret = bch2_trans_update(trans, freespace_iter, update, 0);
+ if (ret)
+ goto err;
+ }
+err:
+fsck_err:
+ printbuf_exit(&buf);
+ return ret;
+}
+
static int bch2_check_discard_freespace_key(struct btree_trans *trans,
struct btree_iter *iter)
{
{
struct btree_trans trans;
struct btree_iter iter, discard_iter, freespace_iter;
+ struct bkey hole;
struct bkey_s_c k;
int ret = 0;
BTREE_ITER_PREFETCH);
bch2_trans_iter_init(&trans, &freespace_iter, BTREE_ID_freespace, POS_MIN,
BTREE_ITER_PREFETCH);
+
while (1) {
- ret = commit_do(&trans, NULL, NULL,
- BTREE_INSERT_NOFAIL|
- BTREE_INSERT_LAZY_RW,
- bch2_check_alloc_key(&trans, &iter,
- &discard_iter,
- &freespace_iter));
+ struct bpos next;
+
+ bch2_trans_begin(&trans);
+
+ k = bch2_get_key_or_real_bucket_hole(&iter, &hole);
+ ret = bkey_err(k);
if (ret)
+ goto bkey_err;
+
+ if (!k.k)
break;
- bch2_btree_iter_advance(&iter);
+ if (k.k->type) {
+ next = bpos_nosnap_successor(k.k->p);
+
+ ret = bch2_check_alloc_key(&trans,
+ k, &iter,
+ &discard_iter,
+ &freespace_iter);
+ if (ret)
+ break;
+ } else {
+ next = k.k->p;
+
+ ret = bch2_check_alloc_hole(&trans,
+ bkey_start_pos(k.k),
+ &next,
+ &freespace_iter);
+ if (ret)
+ goto bkey_err;
+ }
+
+ ret = bch2_trans_commit(&trans, NULL, NULL,
+ BTREE_INSERT_NOFAIL|
+ BTREE_INSERT_LAZY_RW);
+ if (ret)
+ goto bkey_err;
+
+ bch2_btree_iter_set_pos(&iter, next);
+bkey_err:
+ if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
+ continue;
+ if (ret)
+ break;
}
bch2_trans_iter_exit(&trans, &freespace_iter);
bch2_trans_iter_exit(&trans, &discard_iter);
struct btree_trans trans;
struct btree_iter iter;
struct bkey_s_c k;
+ struct bkey hole;
struct bpos end = POS(ca->dev_idx, ca->mi.nbuckets);
struct bch_member *m;
int ret;
*/
while (1) {
bch2_trans_begin(&trans);
- ret = 0;
- if (bkey_ge(iter.pos, end))
+ if (bkey_ge(iter.pos, end)) {
+ ret = 0;
break;
+ }
- k = bch2_btree_iter_peek_slot(&iter);
+ k = bch2_get_key_or_hole(&iter, end, &hole);
ret = bkey_err(k);
if (ret)
goto bkey_err;
bch2_btree_iter_advance(&iter);
} else {
- /*
- * When there's a hole, process a whole range of keys
- * all at once:
- *
- * This is similar to how extent btree iterators in
- * slots mode will synthesize a whole range - a
- * KEY_TYPE_deleted extent.
- *
- * But alloc keys aren't extents (they have zero size),
- * so we're open coding it here:
- */
- struct btree_iter iter2;
struct bkey_i *freespace;
- struct bpos next;
-
- bch2_trans_copy_iter(&iter2, &iter);
- k = bch2_btree_iter_peek_upto(&iter2,
- bkey_min(bkey_min(end,
- iter.path->l[0].b->key.k.p),
- POS(iter.pos.inode, iter.pos.offset + U32_MAX - 1)));
- next = iter2.pos;
- ret = bkey_err(k);
- bch2_trans_iter_exit(&trans, &iter2);
-
- BUG_ON(next.offset >= iter.pos.offset + U32_MAX);
-
- if (ret)
- goto bkey_err;
freespace = bch2_trans_kmalloc(&trans, sizeof(*freespace));
ret = PTR_ERR_OR_ZERO(freespace);
goto bkey_err;
bkey_init(&freespace->k);
- freespace->k.type = KEY_TYPE_set;
- freespace->k.p = iter.pos;
-
- bch2_key_resize(&freespace->k, next.offset - iter.pos.offset);
+ freespace->k.type = KEY_TYPE_set;
+ freespace->k.p = k.k->p;
+ freespace->k.size = k.k->size;
ret = __bch2_btree_insert(&trans, BTREE_ID_freespace, freespace) ?:
bch2_trans_commit(&trans, NULL, NULL,
if (ret)
goto bkey_err;
- bch2_btree_iter_set_pos(&iter, next);
+ bch2_btree_iter_set_pos(&iter, k.k->p);
}
bkey_err:
if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
/* Btree path: fixups after btree node updates: */
+static void bch2_trans_revalidate_updates_in_node(struct btree_trans *trans, struct btree *b)
+{
+ struct bch_fs *c = trans->c;
+ struct btree_insert_entry *i;
+
+ trans_for_each_update(trans, i)
+ if (!i->cached &&
+ i->level == b->c.level &&
+ i->btree_id == b->c.btree_id &&
+ bpos_cmp(i->k->k.p, b->data->min_key) >= 0 &&
+ bpos_cmp(i->k->k.p, b->data->max_key) <= 0) {
+ i->old_v = bch2_btree_path_peek_slot(i->path, &i->old_k).v;
+
+ if (unlikely(trans->journal_replay_not_finished)) {
+ struct bkey_i *j_k =
+ bch2_journal_keys_peek_slot(c, i->btree_id, i->level,
+ i->k->k.p);
+
+ if (j_k) {
+ i->old_k = j_k->k;
+ i->old_v = &j_k->v;
+ }
+ }
+ }
+}
+
/*
* A btree node is being replaced - update the iterator to point to the new
* node:
bch2_btree_path_level_init(trans, path, b);
}
+
+ bch2_trans_revalidate_updates_in_node(trans, b);
}
/*
trans_for_each_path_with_node(trans, b, path)
__btree_path_level_init(path, b->c.level);
+
+ bch2_trans_revalidate_updates_in_node(trans, b);
}
/* Btree path: traverse, set_pos: */
unsigned i;
EBUG_ON(path->sorted_idx >= trans->nr_sorted);
-#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
- trans->nr_sorted--;
- memmove_u64s_down_small(trans->sorted + path->sorted_idx,
- trans->sorted + path->sorted_idx + 1,
- DIV_ROUND_UP(trans->nr_sorted - path->sorted_idx, 8));
-#else
+
array_remove_item(trans->sorted, trans->nr_sorted, path->sorted_idx);
-#endif
+
for (i = path->sorted_idx; i < trans->nr_sorted; i++)
trans->paths[trans->sorted[i]].sorted_idx = i;
trans->traverse_all_idx >= path->sorted_idx)
trans->traverse_all_idx++;
-#ifdef CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS
- memmove_u64s_up_small(trans->sorted + path->sorted_idx + 1,
- trans->sorted + path->sorted_idx,
- DIV_ROUND_UP(trans->nr_sorted - path->sorted_idx, 8));
- trans->nr_sorted++;
- trans->sorted[path->sorted_idx] = path->idx;
-#else
array_insert_item(trans->sorted, trans->nr_sorted, path->sorted_idx, path->idx);
-#endif
for (i = path->sorted_idx; i < trans->nr_sorted; i++)
trans->paths[trans->sorted[i]].sorted_idx = i;
return k;
}
+static inline struct bkey_s_c
+__bch2_btree_iter_peek_upto_and_restart(struct btree_trans *trans,
+ struct btree_iter *iter,
+ struct bpos end,
+ unsigned flags)
+{
+ struct bkey_s_c k;
+
+ while (btree_trans_too_many_iters(trans) ||
+ (k = bch2_btree_iter_peek_upto_type(iter, end, flags),
+ bch2_err_matches(bkey_err(k), BCH_ERR_transaction_restart)))
+ bch2_trans_begin(trans);
+
+ return k;
+}
+
#define lockrestart_do(_trans, _do) \
({ \
u32 _restart_count; \
!((_ret) = bkey_err(_k)) && (_k).k; \
bch2_btree_iter_advance(&(_iter)))
+#define for_each_btree_key_upto(_trans, _iter, _btree_id, \
+ _start, _end, _flags, _k, _ret) \
+ for (bch2_trans_iter_init((_trans), &(_iter), (_btree_id), \
+ (_start), (_flags)); \
+ (_k) = __bch2_btree_iter_peek_upto_and_restart((_trans), \
+ &(_iter), _end, _flags),\
+ !((_ret) = bkey_err(_k)) && (_k).k; \
+ bch2_btree_iter_advance(&(_iter)))
+
#define for_each_btree_key_norestart(_trans, _iter, _btree_id, \
_start, _flags, _k, _ret) \
for (bch2_trans_iter_init((_trans), &(_iter), (_btree_id), \
#include <linux/sort.h>
#include <trace/events/bcachefs.h>
+static void verify_update_old_key(struct btree_trans *trans, struct btree_insert_entry *i)
+{
+#ifdef CONFIG_BCACHEFS_DEBUG
+ struct bch_fs *c = trans->c;
+ struct bkey u;
+ struct bkey_s_c k = bch2_btree_path_peek_slot(i->path, &u);
+
+ if (unlikely(trans->journal_replay_not_finished)) {
+ struct bkey_i *j_k =
+ bch2_journal_keys_peek_slot(c, i->btree_id, i->level, i->k->k.p);
+
+ if (j_k)
+ k = bkey_i_to_s_c(j_k);
+ }
+
+ i->old_k.needs_whiteout = k.k->needs_whiteout;
+
+ BUG_ON(memcmp(&i->old_k, k.k, sizeof(struct bkey)));
+ BUG_ON(i->old_v != k.v);
+#endif
+}
+
static int __must_check
bch2_trans_update_by_path(struct btree_trans *, struct btree_path *,
struct bkey_i *, enum btree_update_flags);
{
struct bch_fs *c = trans->c;
struct bkey_cached *ck = (void *) path->l[0].b;
+ struct btree_insert_entry *i;
unsigned new_u64s;
struct bkey_i *new_k;
return -ENOMEM;
}
+ trans_for_each_update(trans, i)
+ if (i->old_v == &ck->k->v)
+ i->old_v = &new_k->v;
+
ck->u64s = new_u64s;
ck->k = new_k;
return 0;
struct bkey_i *new = i->k;
int ret;
+ verify_update_old_key(trans, i);
+
if (unlikely(flags & BTREE_TRIGGER_NORUN))
return 0;
struct bkey old_k = i->old_k;
struct bkey_s_c old = { &old_k, i->old_v };
+ verify_update_old_key(trans, i);
+
if ((i->flags & BTREE_TRIGGER_NORUN) ||
!(BTREE_NODE_TYPE_HAS_TRANS_TRIGGERS & (1U << i->bkey_type)))
return 0;
if (btree_node_type_needs_gc(i->bkey_type))
marking = true;
-
- /*
- * Revalidate before calling mem triggers - XXX, ugly:
- *
- * - successful btree node splits don't cause transaction
- * restarts and will have invalidated the pointer to the bkey
- * value
- * - btree_node_lock_for_insert() -> btree_node_prep_for_write()
- * when it has to resort
- * - btree_key_can_insert_cached() when it has to reallocate
- *
- * Ugly because we currently have no way to tell if the
- * pointer's been invalidated, which means it's debatabale
- * whether we should be stashing the old key at all.
- */
- i->old_v = bch2_btree_path_peek_slot(i->path, &i->old_k).v;
-
- if (unlikely(trans->journal_replay_not_finished)) {
- struct bkey_i *j_k =
- bch2_journal_keys_peek_slot(c, i->btree_id, i->level,
- i->k->k.p);
-
- if (j_k) {
- i->old_k = j_k->k;
- i->old_v = &j_k->v;
- }
- }
}
/*
if (i->key_cache_already_flushed)
continue;
+ verify_update_old_key(trans, i);
+
entry = bch2_journal_add_entry(j, &trans->journal_res,
BCH_JSET_ENTRY_overwrite,
i->btree_id, i->level,
u64 sectors = 0;
int ret;
- for_each_btree_key(trans, iter, BTREE_ID_extents,
- SPOS(inum, 0, snapshot), 0, k, ret) {
- if (k.k->p.inode != inum)
- break;
-
+ for_each_btree_key_upto(trans, iter, BTREE_ID_extents,
+ SPOS(inum, 0, snapshot),
+ POS(inum, U64_MAX),
+ 0, k, ret)
if (bkey_extent_is_allocation(k.k))
sectors += k.k->size;
- }
bch2_trans_iter_exit(trans, &iter);
u64 subdirs = 0;
int ret;
- for_each_btree_key(trans, iter, BTREE_ID_dirents,
- SPOS(inum, 0, snapshot), 0, k, ret) {
- if (k.k->p.inode != inum)
- break;
-
+ for_each_btree_key_upto(trans, iter, BTREE_ID_dirents,
+ SPOS(inum, 0, snapshot),
+ POS(inum, U64_MAX),
+ 0, k, ret) {
if (k.k->type != KEY_TYPE_dirent)
continue;
if (d.v->d_type == DT_DIR)
subdirs++;
}
-
bch2_trans_iter_exit(trans, &iter);
return ret ?: subdirs;
if (!last_seq) {
fsck_err(c, "journal read done, but no entries found after dropping non-flushes");
- ret = -1;
- goto err;
+ return 0;
}
bch_info(c, "journal read done, replaying entries %llu-%llu",
if (ret)
goto err;
+ /*
+ * note: cmd_list_journal needs the blacklist table fully up to date so
+ * it can asterisk ignored journal entries:
+ */
+ if (c->opts.read_journal_only)
+ goto out;
+
genradix_for_each_reverse(&c->journal_entries, iter, i)
if (*i && !(*i)->ignore) {
last_journal_entry = &(*i)->j;
}
}
- /*
- * note: cmd_list_journal needs the blacklist table fully up to date so
- * it can asterisk ignored journal entries:
- */
- if (c->opts.read_journal_only)
- goto out;
-
ret = bch2_fs_journal_start(&c->journal, journal_seq);
if (ret)
goto err;
__memmove_u64s_down(dst, src, u64s);
}
-static inline void __memmove_u64s_down_small(void *dst, const void *src,
- unsigned u64s)
-{
- memcpy_u64s_small(dst, src, u64s);
-}
-
-static inline void memmove_u64s_down_small(void *dst, const void *src,
- unsigned u64s)
-{
- EBUG_ON(dst > src);
-
- __memmove_u64s_down_small(dst, src, u64s);
-}
-
static inline void __memmove_u64s_up_small(void *_dst, const void *_src,
unsigned u64s)
{
projects / bcachefs-tools-debian / commitdiff