bcachefs: bch2_fpunch_snapshot()

Add a new version of fpunch for operating on a snapshot ID, not a
subvolume - and use it for "extent past end of inode" repair.

Previously, repair would try to delete everything at once, but deleting
too many extents at once can overflow the btree_trans bump allocator, as
well as causing other problems - the new helper properly uses
bch2_extent_trim_atomic().

Reported-and-tested-by: Edoardo Codeglia <bcachefs@404.blue>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/fsck.c

@@ -12,6 +12,7 @@
 #include "fs.h"
 #include "fsck.h"
 #include "inode.h"
+#include "io_misc.h"
 #include "keylist.h"
 #include "namei.h"
 #include "recovery_passes.h"
@@ -1919,33 +1920,11 @@ static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
 					"extent type past end of inode %llu:%u, i_size %llu\n%s",
 					i->inode.bi_inum, i->inode.bi_snapshot, i->inode.bi_size,
 					(bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
-				struct bkey_i *whiteout = bch2_trans_kmalloc(trans, sizeof(*whiteout));
+				ret = bch2_fpunch_snapshot(trans,
-				ret = PTR_ERR_OR_ZERO(whiteout);
+							   SPOS(i->inode.bi_inum,
-				if (ret)
+								last_block,
-					goto err;
+								i->inode.bi_snapshot),
-
+							   POS(i->inode.bi_inum, U64_MAX));
-				bkey_init(&whiteout->k);
-				whiteout->k.p = SPOS(k.k->p.inode,
-						     last_block,
-						     i->inode.bi_snapshot);
-				bch2_key_resize(&whiteout->k,
-						min(KEY_SIZE_MAX & (~0 << c->block_bits),
-						    U64_MAX - whiteout->k.p.offset));
-
-
-				/*
-				 * Need a normal (not BTREE_ITER_all_snapshots)
-				 * iterator, if we're deleting in a different
-				 * snapshot and need to emit a whiteout
-				 */
-				struct btree_iter iter2;
-				bch2_trans_iter_init(trans, &iter2, BTREE_ID_extents,
-						     bkey_start_pos(&whiteout->k),
-						     BTREE_ITER_intent);
-				ret =   bch2_btree_iter_traverse(trans, &iter2) ?:
-					bch2_trans_update(trans, &iter2, whiteout,
-						BTREE_UPDATE_internal_snapshot_node);
-				bch2_trans_iter_exit(trans, &iter2);
 				if (ret)
 					goto err;
 

fs/bcachefs/io_misc.c

@@ -135,6 +135,33 @@ err_noprint:
 	return ret;
 }
 
+/* For fsck */
+int bch2_fpunch_snapshot(struct btree_trans *trans, struct bpos start, struct bpos end)
+{
+	u32 restart_count = trans->restart_count;
+	struct bch_fs *c = trans->c;
+	struct disk_reservation disk_res = bch2_disk_reservation_init(c, 0);
+	unsigned max_sectors	= KEY_SIZE_MAX & (~0 << c->block_bits);
+	struct bkey_i delete;
+
+	int ret = for_each_btree_key_max_commit(trans, iter, BTREE_ID_extents,
+			start, end, 0, k,
+			&disk_res, NULL, BCH_TRANS_COMMIT_no_enospc, ({
+		bkey_init(&delete.k);
+		delete.k.p = iter.pos;
+
+		/* create the biggest key we can */
+		bch2_key_resize(&delete.k, max_sectors);
+		bch2_cut_back(end, &delete);
+
+		bch2_extent_trim_atomic(trans, &iter, &delete) ?:
+		bch2_trans_update(trans, &iter, &delete, 0);
+	}));
+
+	bch2_disk_reservation_put(c, &disk_res);
+	return ret ?: trans_was_restarted(trans, restart_count);
+}
+
 /*
  * Returns -BCH_ERR_transacton_restart if we had to drop locks:
  */

fs/bcachefs/io_misc.h

@@ -5,6 +5,8 @@
 int bch2_extent_fallocate(struct btree_trans *, subvol_inum, struct btree_iter *,
 			  u64, struct bch_io_opts, s64 *,
 			  struct write_point_specifier);
+
+int bch2_fpunch_snapshot(struct btree_trans *, struct bpos, struct bpos);
 int bch2_fpunch_at(struct btree_trans *, struct btree_iter *,
 		   subvol_inum, u64, s64 *);
 int bch2_fpunch(struct bch_fs *c, subvol_inum, u64, u64, s64 *);