bcachefs: Fix another deadlock in the btree interior update path

Can't take read locks on btree nodes while holding
btree_interior_update_lock. Also, fix a bug where we were leaking
journal prereservations.

Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

diff --git a/fs/bcachefs/btree_update_interior.c b/fs/bcachefs/btree_update_interior.c
index daa4c0716c05072a40e6d824aefc684f7cf6ca1a..772595b3da9f22d72b26098b717d5de9114f486e 100644 (file)
--- a/fs/bcachefs/btree_update_interior.c
+++ b/fs/bcachefs/btree_update_interior.c
@@ -609,33 +609,11 @@ static void bch2_btree_update_free(struct btree_update *as)
        mutex_unlock(&c->btree_interior_update_lock);
 }
 
-static void btree_update_nodes_reachable(struct btree_update *as, u64 seq)
-{
-       struct bch_fs *c = as->c;
-
-       while (as->nr_new_nodes) {
-               struct btree *b = as->new_nodes[--as->nr_new_nodes];
-
-               BUG_ON(b->will_make_reachable != (unsigned long) as);
-               b->will_make_reachable = 0;
-
-               /*
-                * b->will_make_reachable prevented it from being written, so
-                * write it now if it needs to be written:
-                */
-               btree_node_lock_type(c, b, SIX_LOCK_read);
-               bch2_btree_node_write_cond(c, b, btree_node_need_write(b));
-               six_unlock_read(&b->c.lock);
-       }
-
-       while (as->nr_pending)
-               bch2_btree_node_free_ondisk(c, &as->pending[--as->nr_pending],
-                                           seq);
-}
-
 static void btree_update_nodes_written(struct closure *cl)
 {
        struct btree_update *as = container_of(cl, struct btree_update, cl);
+       struct btree *new_nodes[BTREE_MAX_DEPTH * 2 + GC_MERGE_NODES];
+       unsigned nr_new_nodes;
        struct journal_res res = { 0 };
        struct bch_fs *c = as->c;
        struct btree *b;
@@ -650,6 +628,7 @@ static void btree_update_nodes_written(struct closure *cl)
        mutex_lock(&c->btree_interior_update_lock);
        as->nodes_written = true;
 again:
+       nr_new_nodes = 0;
        as = list_first_entry_or_null(&c->btree_interior_updates_unwritten,
                                      struct btree_update, unwritten_list);
        if (!as || !as->nodes_written) {
@@ -738,8 +717,23 @@ free_update:
                six_unlock_intent(&b->c.lock);
        }
 
-       if (!ret)
-               btree_update_nodes_reachable(as, res.seq);
+       if (!ret) {
+               nr_new_nodes = as->nr_new_nodes;
+               memcpy(new_nodes,
+                      as->new_nodes,
+                      as->nr_new_nodes * sizeof(struct btree *));
+
+               while (as->nr_new_nodes) {
+                       struct btree *b = as->new_nodes[--as->nr_new_nodes];
+
+                       BUG_ON(b->will_make_reachable != (unsigned long) as);
+                       b->will_make_reachable = 0;
+               }
+
+               while (as->nr_pending)
+                       bch2_btree_node_free_ondisk(c,
+                               &as->pending[--as->nr_pending], res.seq);
+       }
 
        __bch2_btree_update_free(as);
        /*
@@ -747,6 +741,20 @@ free_update:
         * nodes to be writeable:
         */
        closure_wake_up(&c->btree_interior_update_wait);
+
+       /*
+        * Can't take btree node locks while holding btree_interior_update_lock:
+        * */
+       mutex_unlock(&c->btree_interior_update_lock);
+
+       while (nr_new_nodes) {
+               struct btree *b = new_nodes[--nr_new_nodes];
+               btree_node_lock_type(c, b, SIX_LOCK_read);
+               bch2_btree_node_write_cond(c, b, btree_node_need_write(b));
+               six_unlock_read(&b->c.lock);
+       }
+
+       mutex_lock(&c->btree_interior_update_lock);
        goto again;
 }
 
@@ -963,11 +971,13 @@ bch2_btree_update_start(struct btree_trans *trans, enum btree_id id,
                                      BTREE_UPDATE_JOURNAL_RES,
                                      JOURNAL_RES_GET_NONBLOCK);
        if (ret == -EAGAIN) {
+               if (flags & BTREE_INSERT_NOUNLOCK)
+                       return ERR_PTR(-EINTR);
+
                bch2_trans_unlock(trans);
 
                ret = bch2_journal_preres_get(&c->journal, &journal_preres,
-                                             BTREE_UPDATE_JOURNAL_RES,
-                                             JOURNAL_RES_GET_NONBLOCK);
+                                             BTREE_UPDATE_JOURNAL_RES, 0);
                if (ret)
                        return ERR_PTR(ret);
 
@@ -978,8 +988,10 @@ bch2_btree_update_start(struct btree_trans *trans, enum btree_id id,
        }
 
        reserve = bch2_btree_reserve_get(c, nr_nodes, flags, cl);
-       if (IS_ERR(reserve))
+       if (IS_ERR(reserve)) {
+               bch2_journal_preres_put(&c->journal, &journal_preres);
                return ERR_CAST(reserve);
+       }
 
        as = mempool_alloc(&c->btree_interior_update_pool, GFP_NOIO);
        memset(as, 0, sizeof(*as));
@@ -1677,6 +1689,7 @@ retry:
 
        as = bch2_btree_update_start(trans, iter->btree_id,
                         btree_update_reserve_required(c, parent) + 1,
+                        flags|
                         BTREE_INSERT_NOFAIL|
                         BTREE_INSERT_USE_RESERVE,
                         !(flags & BTREE_INSERT_NOUNLOCK) ? &cl : NULL);