bcachefs: Make topology errors autofix

These repair paths are well tested, we can repair them without explicit
user intervention

This also tweaks bch2_topology_error() so that we run topology repair if
we're in recovery, not just fsck.

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

diff --git a/fs/bcachefs/btree_gc.c b/fs/bcachefs/btree_gc.c
index 2e8cfc4d326551858c4fcc69e17db24ecc7400f3..19db4d8aca88c4d1223fce01b4f4fe1bb5a27bb3 100644 (file)
--- a/fs/bcachefs/btree_gc.c
+++ b/fs/bcachefs/btree_gc.c
@@ -348,7 +348,7 @@ again:
                bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(cur_k.k));
 
                if (mustfix_fsck_err_on(bch2_err_matches(ret, EIO),
-                               trans, btree_node_unreadable,
+                               trans, btree_node_read_error,
                                "Topology repair: unreadable btree node at\n"
                                "  %s",
                                buf.buf)) {

diff --git a/fs/bcachefs/recovery.c b/fs/bcachefs/recovery.c
index e361057ffad4b22614a0f858c71b41f3b26c105e..64bb330eac861f738cb4f2b062edb55d7fb53982 100644 (file)
--- a/fs/bcachefs/recovery.c
+++ b/fs/bcachefs/recovery.c
@@ -40,19 +40,42 @@ int bch2_btree_lost_data(struct bch_fs *c, enum btree_id btree)
        int ret = 0;
 
        mutex_lock(&c->sb_lock);
+       struct bch_sb_field_ext *ext = bch2_sb_field_get(c->disk_sb.sb, ext);
 
        if (!(c->sb.btrees_lost_data & b)) {
                struct printbuf buf = PRINTBUF;
                bch2_btree_id_to_text(&buf, btree);
                bch_err(c, "flagging btree %s lost data", buf.buf);
                printbuf_exit(&buf);
-               bch2_sb_field_get(c->disk_sb.sb, ext)->btrees_lost_data |= cpu_to_le64(b);
+               ext->btrees_lost_data |= cpu_to_le64(b);
        }
 
+       /* Once we have runtime self healing for topology errors we won't need this: */
+       ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_topology) ?: ret;
+
+       /* Btree node accounting will be off: */
+       __set_bit_le64(BCH_FSCK_ERR_accounting_mismatch, ext->errors_silent);
+       ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_allocations) ?: ret;
+
+#ifdef CONFIG_BCACHEFS_DEBUG
+       /*
+        * These are much more minor, and don't need to be corrected right away,
+        * but in debug mode we want the next fsck run to be clean:
+        */
+       ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_lrus) ?: ret;
+       ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_backpointers_to_extents) ?: ret;
+#endif
+
        switch (btree) {
        case BTREE_ID_alloc:
-               ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_allocations) ?: ret;
                ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_alloc_info) ?: ret;
+
+               __set_bit_le64(BCH_FSCK_ERR_alloc_key_data_type_wrong, ext->errors_silent);
+               __set_bit_le64(BCH_FSCK_ERR_alloc_key_gen_wrong, ext->errors_silent);
+               __set_bit_le64(BCH_FSCK_ERR_alloc_key_dirty_sectors_wrong, ext->errors_silent);
+               __set_bit_le64(BCH_FSCK_ERR_alloc_key_cached_sectors_wrong, ext->errors_silent);
+               __set_bit_le64(BCH_FSCK_ERR_alloc_key_stripe_wrong, ext->errors_silent);
+               __set_bit_le64(BCH_FSCK_ERR_alloc_key_stripe_redundancy_wrong, ext->errors_silent);
                goto out;
        case BTREE_ID_backpointers:
                ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_btree_backpointers) ?: ret;
@@ -75,7 +98,6 @@ int bch2_btree_lost_data(struct bch_fs *c, enum btree_id btree)
                goto out;
        default:
                ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_scan_for_btree_nodes) ?: ret;
-               ret = bch2_run_explicit_recovery_pass_persistent_locked(c, BCH_RECOVERY_PASS_check_topology) ?: ret;
                goto out;
        }
 out:
@@ -748,9 +770,6 @@ int bch2_fs_recovery(struct bch_fs *c)
                bch2_write_super(c);
        mutex_unlock(&c->sb_lock);
 
-       if (c->opts.fsck && IS_ENABLED(CONFIG_BCACHEFS_DEBUG))
-               c->opts.recovery_passes |= BIT_ULL(BCH_RECOVERY_PASS_check_topology);
-
        if (c->opts.fsck)
                set_bit(BCH_FS_fsck_running, &c->flags);
        if (c->sb.clean)

diff --git a/fs/bcachefs/sb-errors_format.h b/fs/bcachefs/sb-errors_format.h
index 89d9dc2c859bb83fe031039280e2932a40cdd2ec..917ef6aa4a23a6295db55434863a99a547adb7e4 100644 (file)
--- a/fs/bcachefs/sb-errors_format.h
+++ b/fs/bcachefs/sb-errors_format.h
@@ -72,12 +72,12 @@ enum bch_fsck_flags {
        x(btree_root_read_error,                                 59,    FSCK_AUTOFIX)   \
        x(btree_root_bad_min_key,                                60,    0)              \
        x(btree_root_bad_max_key,                                61,    0)              \
-       x(btree_node_read_error,                                 62,    0)              \
-       x(btree_node_topology_bad_min_key,                       63,    0)              \
-       x(btree_node_topology_bad_max_key,                       64,    0)              \
-       x(btree_node_topology_overwritten_by_prev_node,          65,    0)              \
-       x(btree_node_topology_overwritten_by_next_node,          66,    0)              \
-       x(btree_node_topology_interior_node_empty,               67,    0)              \
+       x(btree_node_read_error,                                 62,    FSCK_AUTOFIX)   \
+       x(btree_node_topology_bad_min_key,                       63,    FSCK_AUTOFIX)   \
+       x(btree_node_topology_bad_max_key,                       64,    FSCK_AUTOFIX)   \
+       x(btree_node_topology_overwritten_by_prev_node,          65,    FSCK_AUTOFIX)   \
+       x(btree_node_topology_overwritten_by_next_node,          66,    FSCK_AUTOFIX)   \
+       x(btree_node_topology_interior_node_empty,               67,    FSCK_AUTOFIX)   \
        x(fs_usage_hidden_wrong,                                 68,    FSCK_AUTOFIX)   \
        x(fs_usage_btree_wrong,                                  69,    FSCK_AUTOFIX)   \
        x(fs_usage_data_wrong,                                   70,    FSCK_AUTOFIX)   \