diff --git a/.bcachefs_revision b/.bcachefs_revision
index 562beb39..f5052e09 100644
--- a/.bcachefs_revision
+++ b/.bcachefs_revision
@@ -1 +1 @@
-0389c09b2fb702ca7924ddf550ce0c8af708b8be
+035773e55b52cd87410b2cc319feb1981c74dab0
diff --git a/libbcachefs/btree_key_cache.c b/libbcachefs/btree_key_cache.c
index 88a3582a..e8c1c530 100644
--- a/libbcachefs/btree_key_cache.c
+++ b/libbcachefs/btree_key_cache.c
@@ -842,8 +842,6 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
* Newest freed entries are at the end of the list - once we hit one
* that's too new to be freed, we can bail out:
*/
- scanned += bc->nr_freed_nonpcpu;
-
list_for_each_entry_safe(ck, t, &bc->freed_nonpcpu, list) {
if (!poll_state_synchronize_srcu(&c->btree_trans_barrier,
ck->btree_trans_barrier_seq))
@@ -857,11 +855,6 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
bc->nr_freed_nonpcpu--;
}
- if (scanned >= nr)
- goto out;
-
- scanned += bc->nr_freed_pcpu;
-
list_for_each_entry_safe(ck, t, &bc->freed_pcpu, list) {
if (!poll_state_synchronize_srcu(&c->btree_trans_barrier,
ck->btree_trans_barrier_seq))
@@ -875,9 +868,6 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
bc->nr_freed_pcpu--;
}
- if (scanned >= nr)
- goto out;
-
rcu_read_lock();
tbl = rht_dereference_rcu(bc->table.tbl, &bc->table);
if (bc->shrink_iter >= tbl->size)
@@ -893,12 +883,12 @@ static unsigned long bch2_btree_key_cache_scan(struct shrinker *shrink,
next = rht_dereference_bucket_rcu(pos->next, tbl, bc->shrink_iter);
ck = container_of(pos, struct bkey_cached, hash);
- if (test_bit(BKEY_CACHED_DIRTY, &ck->flags))
+ if (test_bit(BKEY_CACHED_DIRTY, &ck->flags)) {
goto next;
-
- if (test_bit(BKEY_CACHED_ACCESSED, &ck->flags))
+ } else if (test_bit(BKEY_CACHED_ACCESSED, &ck->flags)) {
clear_bit(BKEY_CACHED_ACCESSED, &ck->flags);
- else if (bkey_cached_lock_for_evict(ck)) {
+ goto next;
+ } else if (bkey_cached_lock_for_evict(ck)) {
bkey_cached_evict(bc, ck);
bkey_cached_free(bc, ck);
}
@@ -916,7 +906,6 @@ next:
} while (scanned < nr && bc->shrink_iter != start);
rcu_read_unlock();
-out:
memalloc_nofs_restore(flags);
srcu_read_unlock(&c->btree_trans_barrier, srcu_idx);
mutex_unlock(&bc->lock);
diff --git a/libbcachefs/btree_node_scan.c b/libbcachefs/btree_node_scan.c
index c6079426..45cb8149 100644
--- a/libbcachefs/btree_node_scan.c
+++ b/libbcachefs/btree_node_scan.c
@@ -57,13 +57,14 @@ static void found_btree_node_to_key(struct bkey_i *k, const struct found_btree_n
bp->v.seq = cpu_to_le64(f->cookie);
bp->v.sectors_written = 0;
bp->v.flags = 0;
+ bp->v.sectors_written = cpu_to_le16(f->sectors_written);
bp->v.min_key = f->min_key;
SET_BTREE_PTR_RANGE_UPDATED(&bp->v, f->range_updated);
memcpy(bp->v.start, f->ptrs, sizeof(struct bch_extent_ptr) * f->nr_ptrs);
}
static bool found_btree_node_is_readable(struct btree_trans *trans,
- const struct found_btree_node *f)
+ struct found_btree_node *f)
{
struct { __BKEY_PADDED(k, BKEY_BTREE_PTR_VAL_U64s_MAX); } k;
@@ -71,8 +72,10 @@ static bool found_btree_node_is_readable(struct btree_trans *trans,
struct btree *b = bch2_btree_node_get_noiter(trans, &k.k, f->btree_id, f->level, false);
bool ret = !IS_ERR_OR_NULL(b);
- if (ret)
+ if (ret) {
+ f->sectors_written = b->written;
six_unlock_read(&b->c.lock);
+ }
/*
* We might update this node's range; if that happens, we need the node
diff --git a/libbcachefs/btree_node_scan_types.h b/libbcachefs/btree_node_scan_types.h
index abb7b27d..5cfaeb5a 100644
--- a/libbcachefs/btree_node_scan_types.h
+++ b/libbcachefs/btree_node_scan_types.h
@@ -9,6 +9,7 @@ struct found_btree_node {
bool overwritten:1;
u8 btree_id;
u8 level;
+ unsigned sectors_written;
u32 seq;
u64 cookie;
diff --git a/libbcachefs/btree_types.h b/libbcachefs/btree_types.h
index e0c982a4..c69b233c 100644
--- a/libbcachefs/btree_types.h
+++ b/libbcachefs/btree_types.h
@@ -321,9 +321,9 @@ struct bkey_cached {
struct btree_bkey_cached_common c;
unsigned long flags;
+ unsigned long btree_trans_barrier_seq;
u16 u64s;
bool valid;
- u32 btree_trans_barrier_seq;
struct bkey_cached_key key;
struct rhash_head hash;
diff --git a/libbcachefs/btree_update_interior.c b/libbcachefs/btree_update_interior.c
index 6030c396..b4efd8cc 100644
--- a/libbcachefs/btree_update_interior.c
+++ b/libbcachefs/btree_update_interior.c
@@ -1960,7 +1960,11 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
if ((flags & BCH_WATERMARK_MASK) == BCH_WATERMARK_interior_updates)
return 0;
- flags &= ~BCH_WATERMARK_MASK;
+ if ((flags & BCH_WATERMARK_MASK) <= BCH_WATERMARK_reclaim) {
+ flags &= ~BCH_WATERMARK_MASK;
+ flags |= BCH_WATERMARK_btree;
+ flags |= BCH_TRANS_COMMIT_journal_reclaim;
+ }
b = trans->paths[path].l[level].b;
diff --git a/libbcachefs/buckets.c b/libbcachefs/buckets.c
index 941401a2..8a3a12db 100644
--- a/libbcachefs/buckets.c
+++ b/libbcachefs/buckets.c
@@ -629,7 +629,6 @@ int bch2_check_bucket_ref(struct btree_trans *trans,
bch2_data_type_str(ptr_data_type),
(printbuf_reset(&buf),
bch2_bkey_val_to_text(&buf, c, k), buf.buf));
- BUG();
ret = -EIO;
goto err;
}
diff --git a/libbcachefs/fs.c b/libbcachefs/fs.c
index b5ea9fa1..fce69000 100644
--- a/libbcachefs/fs.c
+++ b/libbcachefs/fs.c
@@ -188,7 +188,8 @@ static struct bch_inode_info *bch2_inode_insert(struct bch_fs *c, struct bch_ino
BUG_ON(!old);
if (unlikely(old != inode)) {
- discard_new_inode(&inode->v);
+ __destroy_inode(&inode->v);
+ kmem_cache_free(bch2_inode_cache, inode);
inode = old;
} else {
mutex_lock(&c->vfs_inodes_lock);
@@ -225,8 +226,10 @@ static struct bch_inode_info *bch2_new_inode(struct btree_trans *trans)
if (unlikely(!inode)) {
int ret = drop_locks_do(trans, (inode = to_bch_ei(new_inode(c->vfs_sb))) ? 0 : -ENOMEM);
- if (ret && inode)
- discard_new_inode(&inode->v);
+ if (ret && inode) {
+ __destroy_inode(&inode->v);
+ kmem_cache_free(bch2_inode_cache, inode);
+ }
if (ret)
return ERR_PTR(ret);
}
diff --git a/libbcachefs/inode.c b/libbcachefs/inode.c
index ca4a066e..0f95d7fb 100644
--- a/libbcachefs/inode.c
+++ b/libbcachefs/inode.c
@@ -606,7 +606,7 @@ int bch2_trigger_inode(struct btree_trans *trans,
struct bkey_s new,
unsigned flags)
{
- s64 nr = bkey_is_inode(new.k) - bkey_is_inode(old.k);
+ s64 nr = (s64) bkey_is_inode(new.k) - (s64) bkey_is_inode(old.k);
if (flags & BTREE_TRIGGER_TRANSACTIONAL) {
if (nr) {
diff --git a/libbcachefs/journal_io.c b/libbcachefs/journal_io.c
index 9aa28b52..eb1f9d6f 100644
--- a/libbcachefs/journal_io.c
+++ b/libbcachefs/journal_io.c
@@ -1723,7 +1723,7 @@ static void journal_write_endio(struct bio *bio)
percpu_ref_put(&ca->io_ref);
}
-static CLOSURE_CALLBACK(do_journal_write)
+static CLOSURE_CALLBACK(journal_write_submit)
{
closure_type(w, struct journal_buf, io);
struct journal *j = container_of(w, struct journal, buf[w->idx]);
@@ -1768,6 +1768,44 @@ static CLOSURE_CALLBACK(do_journal_write)
continue_at(cl, journal_write_done, j->wq);
}
+static CLOSURE_CALLBACK(journal_write_preflush)
+{
+ closure_type(w, struct journal_buf, io);
+ struct journal *j = container_of(w, struct journal, buf[w->idx]);
+ struct bch_fs *c = container_of(j, struct bch_fs, journal);
+
+ if (j->seq_ondisk + 1 != le64_to_cpu(w->data->seq)) {
+ spin_lock(&j->lock);
+ closure_wait(&j->async_wait, cl);
+ spin_unlock(&j->lock);
+
+ continue_at(cl, journal_write_preflush, j->wq);
+ return;
+ }
+
+ if (w->separate_flush) {
+ for_each_rw_member(c, ca) {
+ percpu_ref_get(&ca->io_ref);
+
+ struct journal_device *ja = &ca->journal;
+ struct bio *bio = &ja->bio[w->idx]->bio;
+ bio_reset(bio, ca->disk_sb.bdev,
+ REQ_OP_WRITE|REQ_SYNC|REQ_META|REQ_PREFLUSH);
+ bio->bi_end_io = journal_write_endio;
+ bio->bi_private = ca;
+ closure_bio_submit(bio, cl);
+ }
+
+ continue_at(cl, journal_write_submit, j->wq);
+ } else {
+ /*
+ * no need to punt to another work item if we're not waiting on
+ * preflushes
+ */
+ journal_write_submit(&cl->work);
+ }
+}
+
static int bch2_journal_write_prep(struct journal *j, struct journal_buf *w)
{
struct bch_fs *c = container_of(j, struct bch_fs, journal);
@@ -2033,23 +2071,9 @@ CLOSURE_CALLBACK(bch2_journal_write)
goto err;
if (!JSET_NO_FLUSH(w->data))
- closure_wait_event(&j->async_wait, j->seq_ondisk + 1 == le64_to_cpu(w->data->seq));
-
- if (!JSET_NO_FLUSH(w->data) && w->separate_flush) {
- for_each_rw_member(c, ca) {
- percpu_ref_get(&ca->io_ref);
-
- struct journal_device *ja = &ca->journal;
- struct bio *bio = &ja->bio[w->idx]->bio;
- bio_reset(bio, ca->disk_sb.bdev,
- REQ_OP_WRITE|REQ_SYNC|REQ_META|REQ_PREFLUSH);
- bio->bi_end_io = journal_write_endio;
- bio->bi_private = ca;
- closure_bio_submit(bio, cl);
- }
- }
-
- continue_at(cl, do_journal_write, j->wq);
+ continue_at(cl, journal_write_preflush, j->wq);
+ else
+ continue_at(cl, journal_write_submit, j->wq);
return;
no_io:
continue_at(cl, journal_write_done, j->wq);
diff --git a/libbcachefs/sb-clean.c b/libbcachefs/sb-clean.c
index 5980ba25..35ca3f13 100644
--- a/libbcachefs/sb-clean.c
+++ b/libbcachefs/sb-clean.c
@@ -29,6 +29,14 @@ int bch2_sb_clean_validate_late(struct bch_fs *c, struct bch_sb_field_clean *cle
for (entry = clean->start;
entry < (struct jset_entry *) vstruct_end(&clean->field);
entry = vstruct_next(entry)) {
+ if (vstruct_end(entry) > vstruct_end(&clean->field)) {
+ bch_err(c, "journal entry (u64s %u) overran end of superblock clean section (u64s %u) by %zu",
+ le16_to_cpu(entry->u64s), le32_to_cpu(clean->field.u64s),
+ (u64 *) vstruct_end(entry) - (u64 *) vstruct_end(&clean->field));
+ bch2_sb_error_count(c, BCH_FSCK_ERR_sb_clean_entry_overrun);
+ return -BCH_ERR_fsck_repair_unimplemented;
+ }
+
ret = bch2_journal_entry_validate(c, NULL, entry,
le16_to_cpu(c->disk_sb.sb->version),
BCH_SB_BIG_ENDIAN(c->disk_sb.sb),
diff --git a/libbcachefs/sb-errors_types.h b/libbcachefs/sb-errors_types.h
index 4ca6e7b0..06c7a644 100644
--- a/libbcachefs/sb-errors_types.h
+++ b/libbcachefs/sb-errors_types.h
@@ -271,7 +271,8 @@
x(btree_root_unreadable_and_scan_found_nothing, 263) \
x(snapshot_node_missing, 264) \
x(dup_backpointer_to_bad_csum_extent, 265) \
- x(btree_bitmap_not_marked, 266)
+ x(btree_bitmap_not_marked, 266) \
+ x(sb_clean_entry_overrun, 267)
enum bch_sb_error_id {
#define x(t, n) BCH_FSCK_ERR_##t = n,
diff --git a/libbcachefs/super.c b/libbcachefs/super.c
index 8daf80a3..88e214c6 100644
--- a/libbcachefs/super.c
+++ b/libbcachefs/super.c
@@ -544,6 +544,7 @@ static void __bch2_fs_free(struct bch_fs *c)
bch2_find_btree_nodes_exit(&c->found_btree_nodes);
bch2_free_pending_node_rewrites(c);
+ bch2_fs_allocator_background_exit(c);
bch2_fs_sb_errors_exit(c);
bch2_fs_counters_exit(c);
bch2_fs_snapshots_exit(c);