bcachefs: Fix locking in data move path

We need to ensure we don't have any btree locks held when calling
do_pending_writes() - besides issuing IOs, upcoming allocator changes
will have allocations doing btree lookups directly.

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

diff --git a/fs/bcachefs/move.c b/fs/bcachefs/move.c
index f428e2ff99f64817b0cf5122f1024402ae1a4a41..04971bf847bf811e622552fcae023ca71388cca7 100644 (file)
--- a/fs/bcachefs/move.c
+++ b/fs/bcachefs/move.c
@@ -486,19 +486,22 @@ static void move_read_endio(struct bio *bio)
        closure_put(&ctxt->cl);
 }
 
-static void do_pending_writes(struct moving_context *ctxt)
+static void do_pending_writes(struct moving_context *ctxt, struct btree_trans *trans)
 {
        struct moving_io *io;
 
+       if (trans)
+               bch2_trans_unlock(trans);
+
        while ((io = next_pending_write(ctxt))) {
                list_del(&io->list);
                move_write(io);
        }
 }
 
-#define move_ctxt_wait_event(_ctxt, _cond)                     \
+#define move_ctxt_wait_event(_ctxt, _trans, _cond)             \
 do {                                                           \
-       do_pending_writes(_ctxt);                               \
+       do_pending_writes(_ctxt, _trans);                       \
                                                                \
        if (_cond)                                              \
                break;                                          \
@@ -506,11 +509,12 @@ do {                                                              \
                     next_pending_write(_ctxt) || (_cond));     \
 } while (1)
 
-static void bch2_move_ctxt_wait_for_io(struct moving_context *ctxt)
+static void bch2_move_ctxt_wait_for_io(struct moving_context *ctxt,
+                                      struct btree_trans *trans)
 {
        unsigned sectors_pending = atomic_read(&ctxt->write_sectors);
 
-       move_ctxt_wait_event(ctxt,
+       move_ctxt_wait_event(ctxt, trans,
                !atomic_read(&ctxt->write_sectors) ||
                atomic_read(&ctxt->write_sectors) != sectors_pending);
 }
@@ -532,14 +536,6 @@ static int bch2_move_extent(struct btree_trans *trans,
        unsigned sectors = k.k->size, pages;
        int ret = -ENOMEM;
 
-       move_ctxt_wait_event(ctxt,
-               atomic_read(&ctxt->write_sectors) <
-               SECTORS_IN_FLIGHT_PER_DEVICE);
-
-       move_ctxt_wait_event(ctxt,
-               atomic_read(&ctxt->read_sectors) <
-               SECTORS_IN_FLIGHT_PER_DEVICE);
-
        /* write path might have to decompress data: */
        bkey_for_each_ptr_decode(k.k, ptrs, p, entry)
                sectors = max_t(unsigned, sectors, p.crc.uncompressed_size);
@@ -692,12 +688,19 @@ static int __bch2_move_data(struct bch_fs *c,
                                schedule_timeout(delay);
 
                        if (unlikely(freezing(current))) {
-                               bch2_trans_unlock(&trans);
-                               move_ctxt_wait_event(ctxt, list_empty(&ctxt->reads));
+                               move_ctxt_wait_event(ctxt, &trans, list_empty(&ctxt->reads));
                                try_to_freeze();
                        }
                } while (delay);
 
+               move_ctxt_wait_event(ctxt, &trans,
+                       atomic_read(&ctxt->write_sectors) <
+                       SECTORS_IN_FLIGHT_PER_DEVICE);
+
+               move_ctxt_wait_event(ctxt, &trans,
+                       atomic_read(&ctxt->read_sectors) <
+                       SECTORS_IN_FLIGHT_PER_DEVICE);
+
                bch2_trans_begin(&trans);
 
                k = bch2_btree_iter_peek(&iter);
@@ -762,7 +765,7 @@ static int __bch2_move_data(struct bch_fs *c,
 
                        if (ret2 == -ENOMEM) {
                                /* memory allocation failure, wait for some IO to finish */
-                               bch2_move_ctxt_wait_for_io(ctxt);
+                               bch2_move_ctxt_wait_for_io(ctxt, &trans);
                                continue;
                        }
 
@@ -847,7 +850,7 @@ int bch2_move_data(struct bch_fs *c,
        }
 
 
-       move_ctxt_wait_event(&ctxt, list_empty(&ctxt.reads));
+       move_ctxt_wait_event(&ctxt, NULL, list_empty(&ctxt.reads));
        closure_sync(&ctxt.cl);
 
        EBUG_ON(atomic_read(&ctxt.write_sectors));