#include "bcachefs.h"
#include "bkey_buf.h"
+#include "btree_cache.h"
#include "btree_update.h"
+#include "buckets.h"
#include "darray.h"
#include "dirent.h"
#include "error.h"
#include "fsck.h"
#include "inode.h"
#include "keylist.h"
-#include "subvolume.h"
+#include "recovery.h"
+#include "snapshot.h"
#include "super.h"
#include "xattr.h"
#define QSTR(n) { { { .len = strlen(n) } }, .name = n }
+/*
+ * XXX: this is handling transaction restarts without returning
+ * -BCH_ERR_transaction_restart_nested, this is not how we do things anymore:
+ */
static s64 bch2_count_inode_sectors(struct btree_trans *trans, u64 inum,
u32 snapshot)
{
u64 sectors = 0;
int ret;
- for_each_btree_key(trans, iter, BTREE_ID_extents,
- SPOS(inum, 0, snapshot), 0, k, ret) {
- if (k.k->p.inode != inum)
- break;
-
+ for_each_btree_key_upto(trans, iter, BTREE_ID_extents,
+ SPOS(inum, 0, snapshot),
+ POS(inum, U64_MAX),
+ 0, k, ret)
if (bkey_extent_is_allocation(k.k))
sectors += k.k->size;
- }
bch2_trans_iter_exit(trans, &iter);
u64 subdirs = 0;
int ret;
- for_each_btree_key(trans, iter, BTREE_ID_dirents,
- SPOS(inum, 0, snapshot), 0, k, ret) {
- if (k.k->p.inode != inum)
- break;
-
+ for_each_btree_key_upto(trans, iter, BTREE_ID_dirents,
+ SPOS(inum, 0, snapshot),
+ POS(inum, U64_MAX),
+ 0, k, ret) {
if (k.k->type != KEY_TYPE_dirent)
continue;
if (d.v->d_type == DT_DIR)
subdirs++;
}
-
bch2_trans_iter_exit(trans, &iter);
return ret ?: subdirs;
static int __snapshot_lookup_subvol(struct btree_trans *trans, u32 snapshot,
u32 *subvol)
{
- struct btree_iter iter;
- struct bkey_s_c k;
- int ret;
-
- bch2_trans_iter_init(trans, &iter, BTREE_ID_snapshots,
- POS(0, snapshot), 0);
- k = bch2_btree_iter_peek_slot(&iter);
- ret = bkey_err(k);
- if (ret)
- goto err;
-
- if (k.k->type != KEY_TYPE_snapshot) {
- bch_err(trans->c, "snapshot %u not fonud", snapshot);
- ret = -ENOENT;
- goto err;
- }
-
- *subvol = le32_to_cpu(bkey_s_c_to_snapshot(k).v->subvol);
-err:
- bch2_trans_iter_exit(trans, &iter);
+ struct bch_snapshot s;
+ int ret = bch2_bkey_get_val_typed(trans, BTREE_ID_snapshots,
+ POS(0, snapshot), 0,
+ snapshot, &s);
+ if (!ret)
+ *subvol = le32_to_cpu(s.subvol);
+ else if (bch2_err_matches(ret, ENOENT))
+ bch_err(trans->c, "snapshot %u not found", snapshot);
return ret;
}
if (ret)
goto err;
- if (!k.k || bkey_cmp(k.k->p, POS(0, inode_nr))) {
- ret = -ENOENT;
+ if (!k.k || !bkey_eq(k.k->p, POS(0, inode_nr))) {
+ ret = -BCH_ERR_ENOENT_inode;
goto err;
}
ret = bch2_inode_unpack(k, inode);
err:
- if (ret && ret != -EINTR)
- bch_err(trans->c, "error %i fetching inode %llu",
- ret, inode_nr);
+ bch_err_msg(trans->c, ret, "fetching inode %llu", inode_nr);
bch2_trans_iter_exit(trans, &iter);
return ret;
}
struct bkey_s_c k;
int ret;
- bch2_trans_iter_init(trans, &iter, BTREE_ID_inodes,
- SPOS(0, inode_nr, *snapshot), 0);
- k = bch2_btree_iter_peek_slot(&iter);
+ k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_inodes,
+ SPOS(0, inode_nr, *snapshot), 0);
ret = bkey_err(k);
if (ret)
goto err;
ret = bkey_is_inode(k.k)
? bch2_inode_unpack(k, inode)
- : -ENOENT;
+ : -BCH_ERR_ENOENT_inode;
if (!ret)
*snapshot = iter.pos.snapshot;
err:
- if (ret && ret != -EINTR)
- bch_err(trans->c, "error %i fetching inode %llu:%u",
- ret, inode_nr, *snapshot);
+ bch_err_msg(trans->c, ret, "fetching inode %llu:%u", inode_nr, *snapshot);
bch2_trans_iter_exit(trans, &iter);
return ret;
}
struct bch_inode_unpacked *inode,
u32 snapshot)
{
- struct btree_iter iter;
- int ret;
+ struct bkey_inode_buf *inode_p =
+ bch2_trans_kmalloc(trans, sizeof(*inode_p));
- bch2_trans_iter_init(trans, &iter, BTREE_ID_inodes,
- SPOS(0, inode->bi_inum, snapshot),
- BTREE_ITER_INTENT);
+ if (IS_ERR(inode_p))
+ return PTR_ERR(inode_p);
- ret = bch2_btree_iter_traverse(&iter) ?:
- bch2_inode_write(trans, &iter, inode);
- bch2_trans_iter_exit(trans, &iter);
- return ret;
+ bch2_inode_pack(inode_p, inode);
+ inode_p->inode.k.p.snapshot = snapshot;
+
+ return bch2_btree_insert_nonextent(trans, BTREE_ID_inodes,
+ &inode_p->inode.k_i,
+ BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
}
-static int write_inode(struct btree_trans *trans,
- struct bch_inode_unpacked *inode,
- u32 snapshot)
+static int fsck_write_inode(struct btree_trans *trans,
+ struct bch_inode_unpacked *inode,
+ u32 snapshot)
{
- int ret = __bch2_trans_do(trans, NULL, NULL,
+ int ret = commit_do(trans, NULL, NULL,
BTREE_INSERT_NOFAIL|
BTREE_INSERT_LAZY_RW,
__write_inode(trans, inode, snapshot));
if (ret)
- bch_err(trans->c, "error in fsck: error %i updating inode", ret);
- return ret;
-}
-
-static int fsck_inode_rm(struct btree_trans *trans, u64 inum, u32 snapshot)
-{
- struct btree_iter iter = { NULL };
- struct bkey_i_inode_generation delete;
- struct bch_inode_unpacked inode_u;
- struct bkey_s_c k;
- int ret;
-
- ret = bch2_btree_delete_range_trans(trans, BTREE_ID_extents,
- SPOS(inum, 0, snapshot),
- SPOS(inum, U64_MAX, snapshot),
- 0, NULL) ?:
- bch2_btree_delete_range_trans(trans, BTREE_ID_dirents,
- SPOS(inum, 0, snapshot),
- SPOS(inum, U64_MAX, snapshot),
- 0, NULL) ?:
- bch2_btree_delete_range_trans(trans, BTREE_ID_xattrs,
- SPOS(inum, 0, snapshot),
- SPOS(inum, U64_MAX, snapshot),
- 0, NULL);
- if (ret)
- goto err;
-retry:
- bch2_trans_begin(trans);
-
- bch2_trans_iter_init(trans, &iter, BTREE_ID_inodes,
- SPOS(0, inum, snapshot), BTREE_ITER_INTENT);
- k = bch2_btree_iter_peek_slot(&iter);
-
- ret = bkey_err(k);
- if (ret)
- goto err;
-
- if (!bkey_is_inode(k.k)) {
- bch2_fs_inconsistent(trans->c,
- "inode %llu:%u not found when deleting",
- inum, snapshot);
- ret = -EIO;
- goto err;
- }
-
- bch2_inode_unpack(k, &inode_u);
-
- /* Subvolume root? */
- if (inode_u.bi_subvol) {
- ret = bch2_subvolume_delete(trans, inode_u.bi_subvol);
- if (ret)
- goto err;
- }
-
- bkey_inode_generation_init(&delete.k_i);
- delete.k.p = iter.pos;
- delete.v.bi_generation = cpu_to_le32(inode_u.bi_generation + 1);
-
- ret = bch2_trans_update(trans, &iter, &delete.k_i, 0) ?:
- bch2_trans_commit(trans, NULL, NULL,
- BTREE_INSERT_NOFAIL);
-err:
- bch2_trans_iter_exit(trans, &iter);
- if (ret == -EINTR)
- goto retry;
-
+ bch_err_fn(trans->c, ret);
return ret;
}
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
bch2_trans_iter_exit(trans, &iter);
err:
- if (ret && ret != -EINTR)
- bch_err(c, "error %i from __remove_dirent()", ret);
+ bch_err_fn(c, ret);
return ret;
}
ret = __lookup_dirent(trans, root_hash_info, root_inum,
&lostfound_str, &inum, &d_type);
- if (ret == -ENOENT) {
+ if (bch2_err_matches(ret, ENOENT)) {
bch_notice(c, "creating lost+found");
goto create_lostfound;
}
- if (ret && ret != -EINTR)
- bch_err(c, "error looking up lost+found: %i", ret);
+ bch_err_fn(c, ret);
if (ret)
return ret;
if (d_type != DT_DIR) {
bch_err(c, "error looking up lost+found: not a directory");
- return ret;
+ return -BCH_ERR_ENOENT_not_directory;
}
/*
- * The check_dirents pass has already run, dangling dirents
+ * The bch2_check_dirents pass has already run, dangling dirents
* shouldn't exist here:
*/
return __lookup_inode(trans, inum, lostfound, &snapshot);
lostfound, &lostfound_str,
0, 0, S_IFDIR|0700, 0, NULL, NULL,
(subvol_inum) { }, 0);
- if (ret && ret != -EINTR)
- bch_err(c, "error creating lost+found: %i", ret);
+ bch_err_msg(c, ret, "creating lost+found");
return ret;
}
struct bch_inode_unpacked *inode,
u32 inode_snapshot)
{
- int ret = __bch2_trans_do(trans, NULL, NULL,
+ int ret = commit_do(trans, NULL, NULL,
BTREE_INSERT_LAZY_RW|
BTREE_INSERT_NOFAIL,
__reattach_inode(trans, inode, inode_snapshot));
- if (ret) {
- bch_err(trans->c, "error %i reattaching inode %llu",
- ret, inode->bi_inum);
- return ret;
- }
-
+ bch_err_msg(trans->c, ret, "reattaching inode %llu", inode->bi_inum);
return ret;
}
struct bch_inode_unpacked *inode)
{
struct btree_iter iter;
- struct bkey_s_c k;
+ struct bkey_s_c_dirent d;
int ret;
- bch2_trans_iter_init(trans, &iter, BTREE_ID_dirents,
- POS(inode->bi_dir, inode->bi_dir_offset), 0);
- k = bch2_btree_iter_peek_slot(&iter);
- ret = bkey_err(k);
- if (ret)
- goto out;
- if (k.k->type != KEY_TYPE_dirent) {
- ret = -ENOENT;
- goto out;
+ d = bch2_bkey_get_iter_typed(trans, &iter, BTREE_ID_dirents,
+ POS(inode->bi_dir, inode->bi_dir_offset), 0,
+ dirent);
+ ret = bkey_err(d) ?:
+ __remove_dirent(trans, d.k->p);
+ bch2_trans_iter_exit(trans, &iter);
+ return ret;
+}
+
+struct snapshots_seen_entry {
+ u32 id;
+ u32 equiv;
+};
+
+struct snapshots_seen {
+ struct bpos pos;
+ DARRAY(struct snapshots_seen_entry) ids;
+};
+
+static inline void snapshots_seen_exit(struct snapshots_seen *s)
+{
+ darray_exit(&s->ids);
+}
+
+static inline void snapshots_seen_init(struct snapshots_seen *s)
+{
+ memset(s, 0, sizeof(*s));
+}
+
+static int snapshots_seen_add_inorder(struct bch_fs *c, struct snapshots_seen *s, u32 id)
+{
+ struct snapshots_seen_entry *i, n = {
+ .id = id,
+ .equiv = bch2_snapshot_equiv(c, id),
+ };
+ int ret = 0;
+
+ darray_for_each(s->ids, i) {
+ if (i->id == id)
+ return 0;
+ if (i->id > id)
+ break;
}
- ret = __remove_dirent(trans, k.k->p);
-out:
- bch2_trans_iter_exit(trans, &iter);
+ ret = darray_insert_item(&s->ids, i - s->ids.data, n);
+ if (ret)
+ bch_err(c, "error reallocating snapshots_seen table (size %zu)",
+ s->ids.size);
return ret;
}
-static int snapshots_seen_update(struct bch_fs *c, struct snapshots_seen *s, struct bpos pos)
+static int snapshots_seen_update(struct bch_fs *c, struct snapshots_seen *s,
+ enum btree_id btree_id, struct bpos pos)
{
- pos.snapshot = snapshot_t(c, pos.snapshot)->equiv;
+ struct snapshots_seen_entry *i, n = {
+ .id = pos.snapshot,
+ .equiv = bch2_snapshot_equiv(c, pos.snapshot),
+ };
+ int ret = 0;
- if (bkey_cmp(s->pos, pos))
+ if (!bkey_eq(s->pos, pos))
s->ids.nr = 0;
+
s->pos = pos;
+ s->pos.snapshot = n.equiv;
- /* Might get called multiple times due to lock restarts */
- if (s->ids.nr && s->ids.data[s->ids.nr - 1] == pos.snapshot)
- return 0;
+ darray_for_each(s->ids, i) {
+ if (i->id == n.id)
+ return 0;
+
+ /*
+ * We currently don't rigorously track for snapshot cleanup
+ * needing to be run, so it shouldn't be a fsck error yet:
+ */
+ if (i->equiv == n.equiv) {
+ bch_err(c, "snapshot deletion did not finish:\n"
+ " duplicate keys in btree %s at %llu:%llu snapshots %u, %u (equiv %u)\n",
+ bch2_btree_id_str(btree_id),
+ pos.inode, pos.offset,
+ i->id, n.id, n.equiv);
+ set_bit(BCH_FS_NEED_DELETE_DEAD_SNAPSHOTS, &c->flags);
+ return bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_delete_dead_snapshots);
+ }
+ }
- return snapshots_seen_add(c, s, pos.snapshot);
+ ret = darray_push(&s->ids, n);
+ if (ret)
+ bch_err(c, "error reallocating snapshots_seen table (size %zu)",
+ s->ids.size);
+ return ret;
}
/**
* key_visible_in_snapshot - returns true if @id is a descendent of @ancestor,
* and @ancestor hasn't been overwritten in @seen
*
- * That is, returns whether key in @ancestor snapshot is visible in @id snapshot
+ * @c: filesystem handle
+ * @seen: list of snapshot ids already seen at current position
+ * @id: descendent snapshot id
+ * @ancestor: ancestor snapshot id
+ *
+ * Returns: whether key in @ancestor snapshot is visible in @id snapshot
*/
static bool key_visible_in_snapshot(struct bch_fs *c, struct snapshots_seen *seen,
u32 id, u32 ancestor)
{
ssize_t i;
- BUG_ON(id > ancestor);
-
- id = snapshot_t(c, id)->equiv;
- ancestor = snapshot_t(c, ancestor)->equiv;
+ EBUG_ON(id > ancestor);
+ EBUG_ON(!bch2_snapshot_is_equiv(c, id));
+ EBUG_ON(!bch2_snapshot_is_equiv(c, ancestor));
/* @ancestor should be the snapshot most recently added to @seen */
- BUG_ON(!seen->ids.nr || seen->ids.data[seen->ids.nr - 1] != ancestor);
- BUG_ON(seen->pos.snapshot != ancestor);
+ EBUG_ON(ancestor != seen->pos.snapshot);
+ EBUG_ON(ancestor != seen->ids.data[seen->ids.nr - 1].equiv);
if (id == ancestor)
return true;
if (!bch2_snapshot_is_ancestor(c, id, ancestor))
return false;
+ /*
+ * We know that @id is a descendant of @ancestor, we're checking if
+ * we've seen a key that overwrote @ancestor - i.e. also a descendent of
+ * @ascestor and with @id as a descendent.
+ *
+ * But we already know that we're scanning IDs between @id and @ancestor
+ * numerically, since snapshot ID lists are kept sorted, so if we find
+ * an id that's an ancestor of @id we're done:
+ */
+
for (i = seen->ids.nr - 2;
- i >= 0 && seen->ids.data[i] >= id;
+ i >= 0 && seen->ids.data[i].equiv >= id;
--i)
- if (bch2_snapshot_is_ancestor(c, id, seen->ids.data[i]) &&
- bch2_snapshot_is_ancestor(c, seen->ids.data[i], ancestor))
+ if (bch2_snapshot_is_ancestor(c, id, seen->ids.data[i].equiv))
return false;
return true;
* snapshot id @dst, test whether there is some snapshot in which @dst is
* visible.
*
- * This assumes we're visiting @src keys in natural key order.
+ * @c: filesystem handle
+ * @s: list of snapshot IDs already seen at @src
+ * @src: snapshot ID of src key
+ * @dst: snapshot ID of dst key
+ * Returns: true if there is some snapshot in which @dst is visible
*
- * @s - list of snapshot IDs already seen at @src
- * @src - snapshot ID of src key
- * @dst - snapshot ID of dst key
+ * Assumes we're visiting @src keys in natural key order
*/
-static int ref_visible(struct bch_fs *c, struct snapshots_seen *s,
- u32 src, u32 dst)
+static bool ref_visible(struct bch_fs *c, struct snapshots_seen *s,
+ u32 src, u32 dst)
{
return dst <= src
? key_visible_in_snapshot(c, s, dst, src)
: bch2_snapshot_is_ancestor(c, src, dst);
}
-#define for_each_visible_inode(_c, _s, _w, _snapshot, _i) \
- for (_i = (_w)->inodes.data; _i < (_w)->inodes.data + (_w)->inodes.nr && (_i)->snapshot <= (_snapshot); _i++)\
+static int ref_visible2(struct bch_fs *c,
+ u32 src, struct snapshots_seen *src_seen,
+ u32 dst, struct snapshots_seen *dst_seen)
+{
+ src = bch2_snapshot_equiv(c, src);
+ dst = bch2_snapshot_equiv(c, dst);
+
+ if (dst > src) {
+ swap(dst, src);
+ swap(dst_seen, src_seen);
+ }
+ return key_visible_in_snapshot(c, src_seen, dst, src);
+}
+
+#define for_each_visible_inode(_c, _s, _w, _snapshot, _i) \
+ for (_i = (_w)->inodes.data; _i < (_w)->inodes.data + (_w)->inodes.nr && \
+ (_i)->snapshot <= (_snapshot); _i++) \
if (key_visible_in_snapshot(_c, _s, _i->snapshot, _snapshot))
struct inode_walker_entry {
struct bch_inode_unpacked inode;
u32 snapshot;
+ bool seen_this_pos;
u64 count;
};
struct inode_walker {
bool first_this_inode;
- u64 cur_inum;
+ bool recalculate_sums;
+ struct bpos last_pos;
DARRAY(struct inode_walker_entry) inodes;
};
return darray_push(&w->inodes, ((struct inode_walker_entry) {
.inode = u,
- .snapshot = snapshot_t(c, inode.k->p.snapshot)->equiv,
+ .snapshot = bch2_snapshot_equiv(c, inode.k->p.snapshot),
}));
}
-static int __walk_inode(struct btree_trans *trans,
- struct inode_walker *w, struct bpos pos)
+static int get_inodes_all_snapshots(struct btree_trans *trans,
+ struct inode_walker *w, u64 inum)
{
struct bch_fs *c = trans->c;
struct btree_iter iter;
struct bkey_s_c k;
- unsigned i, ancestor_pos;
+ u32 restart_count = trans->restart_count;
int ret;
- pos.snapshot = snapshot_t(c, pos.snapshot)->equiv;
-
- if (pos.inode == w->cur_inum) {
- w->first_this_inode = false;
- goto lookup_snapshot;
- }
-
+ w->recalculate_sums = false;
w->inodes.nr = 0;
- for_each_btree_key(trans, iter, BTREE_ID_inodes, POS(0, pos.inode),
+ for_each_btree_key(trans, iter, BTREE_ID_inodes, POS(0, inum),
BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
- if (k.k->p.offset != pos.inode)
+ if (k.k->p.offset != inum)
break;
if (bkey_is_inode(k.k))
if (ret)
return ret;
- w->cur_inum = pos.inode;
- w->first_this_inode = true;
-lookup_snapshot:
- for (i = 0; i < w->inodes.nr; i++)
- if (bch2_snapshot_is_ancestor(c, pos.snapshot, w->inodes.data[i].snapshot))
+ w->first_this_inode = true;
+
+ return trans_was_restarted(trans, restart_count);
+}
+
+static struct inode_walker_entry *
+lookup_inode_for_snapshot(struct bch_fs *c, struct inode_walker *w,
+ u32 snapshot, bool is_whiteout)
+{
+ struct inode_walker_entry *i;
+
+ snapshot = bch2_snapshot_equiv(c, snapshot);
+
+ darray_for_each(w->inodes, i)
+ if (bch2_snapshot_is_ancestor(c, snapshot, i->snapshot))
goto found;
- return INT_MAX;
+
+ return NULL;
found:
- BUG_ON(pos.snapshot > w->inodes.data[i].snapshot);
+ BUG_ON(snapshot > i->snapshot);
- if (pos.snapshot != w->inodes.data[i].snapshot) {
- ancestor_pos = i;
+ if (snapshot != i->snapshot && !is_whiteout) {
+ struct inode_walker_entry new = *i;
+ size_t pos;
+ int ret;
- while (i && w->inodes.data[i - 1].snapshot > pos.snapshot)
+ new.snapshot = snapshot;
+ new.count = 0;
+
+ bch_info(c, "have key for inode %llu:%u but have inode in ancestor snapshot %u",
+ w->last_pos.inode, snapshot, i->snapshot);
+
+ while (i > w->inodes.data && i[-1].snapshot > snapshot)
--i;
- ret = darray_insert_item(&w->inodes, i, w->inodes.data[ancestor_pos]);
+ pos = i - w->inodes.data;
+ ret = darray_insert_item(&w->inodes, pos, new);
if (ret)
- return ret;
+ return ERR_PTR(ret);
- w->inodes.data[i].snapshot = pos.snapshot;
- w->inodes.data[i].count = 0;
+ i = w->inodes.data + pos;
}
return i;
}
+static struct inode_walker_entry *walk_inode(struct btree_trans *trans,
+ struct inode_walker *w, struct bpos pos,
+ bool is_whiteout)
+{
+ if (w->last_pos.inode != pos.inode) {
+ int ret = get_inodes_all_snapshots(trans, w, pos.inode);
+ if (ret)
+ return ERR_PTR(ret);
+ } else if (bkey_cmp(w->last_pos, pos)) {
+ struct inode_walker_entry *i;
+
+ darray_for_each(w->inodes, i)
+ i->seen_this_pos = false;
+
+ }
+
+ w->last_pos = pos;
+
+ return lookup_inode_for_snapshot(trans->c, w, pos.snapshot, is_whiteout);
+}
+
static int __get_visible_inodes(struct btree_trans *trans,
struct inode_walker *w,
struct snapshots_seen *s,
w->inodes.nr = 0;
- for_each_btree_key(trans, iter, BTREE_ID_inodes, POS(0, inum),
+ for_each_btree_key_norestart(trans, iter, BTREE_ID_inodes, POS(0, inum),
BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
+ u32 equiv = bch2_snapshot_equiv(c, k.k->p.snapshot);
+
if (k.k->p.offset != inum)
break;
- if (!bkey_is_inode(k.k))
+ if (!ref_visible(c, s, s->pos.snapshot, equiv))
continue;
- if (ref_visible(c, s, s->pos.snapshot, k.k->p.snapshot)) {
+ if (bkey_is_inode(k.k))
add_inode(c, w, k);
- if (k.k->p.snapshot >= s->pos.snapshot)
- break;
- }
+
+ if (equiv >= s->pos.snapshot)
+ break;
}
bch2_trans_iter_exit(trans, &iter);
struct printbuf buf = PRINTBUF;
int ret = 0;
- if (mustfix_fsck_err_on(!snapshot_t(c, k.k->p.snapshot)->equiv, c,
- "key in missing snapshot: %s",
- (bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+ if (mustfix_fsck_err_on(!bch2_snapshot_equiv(c, k.k->p.snapshot), c,
+ bkey_in_missing_snapshot,
+ "key in missing snapshot: %s",
+ (bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
ret = bch2_btree_delete_at(trans, iter,
BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?: 1;
fsck_err:
struct bch_hash_info *hash_info,
struct btree_iter *k_iter, struct bkey_s_c k)
{
- bch_err(trans->c, "hash_redo_key() not implemented yet");
- return -EINVAL;
-#if 0
struct bkey_i *delete;
struct bkey_i *tmp;
if (IS_ERR(delete))
return PTR_ERR(delete);
- tmp = bch2_trans_kmalloc(trans, bkey_bytes(k.k));
+ tmp = bch2_bkey_make_mut_noupdate(trans, k);
if (IS_ERR(tmp))
return PTR_ERR(tmp);
- bkey_reassemble(tmp, k);
-
bkey_init(&delete->k);
delete->k.p = k_iter->pos;
return bch2_btree_iter_traverse(k_iter) ?:
bch2_trans_update(trans, k_iter, delete, 0) ?:
- bch2_hash_set(trans, desc, hash_info, k_iter->pos.inode, tmp, 0);
-#endif
+ bch2_hash_set_snapshot(trans, desc, hash_info,
+ (subvol_inum) { 0, k.k->p.inode },
+ k.k->p.snapshot, tmp,
+ BCH_HASH_SET_MUST_CREATE,
+ BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE) ?:
+ bch2_trans_commit(trans, NULL, NULL,
+ BTREE_INSERT_NOFAIL|
+ BTREE_INSERT_LAZY_RW);
}
static int hash_check_key(struct btree_trans *trans,
goto bad_hash;
for_each_btree_key_norestart(trans, iter, desc.btree_id,
- POS(hash_k.k->p.inode, hash),
+ SPOS(hash_k.k->p.inode, hash, hash_k.k->p.snapshot),
BTREE_ITER_SLOTS, k, ret) {
- if (!bkey_cmp(k.k->p, hash_k.k->p))
+ if (bkey_eq(k.k->p, hash_k.k->p))
break;
if (fsck_err_on(k.k->type == desc.key_type &&
!desc.cmp_bkey(k, hash_k), c,
+ hash_table_key_duplicate,
"duplicate hash table keys:\n%s",
(printbuf_reset(&buf),
bch2_bkey_val_to_text(&buf, c, hash_k),
printbuf_exit(&buf);
return ret;
bad_hash:
- if (fsck_err(c, "hash table key at wrong offset: btree %s inode %llu offset %llu, "
- "hashed to %llu\n%s",
- bch2_btree_ids[desc.btree_id], hash_k.k->p.inode, hash_k.k->p.offset, hash,
+ if (fsck_err(c, hash_table_key_wrong_offset,
+ "hash table key at wrong offset: btree %s inode %llu offset %llu, hashed to %llu\n%s",
+ bch2_btree_id_str(desc.btree_id), hash_k.k->p.inode, hash_k.k->p.offset, hash,
(printbuf_reset(&buf),
- bch2_bkey_val_to_text(&buf, c, hash_k), buf.buf)) == FSCK_ERR_IGNORE)
- return 0;
-
- ret = hash_redo_key(trans, desc, hash_info, k_iter, hash_k);
- if (ret) {
- bch_err(c, "hash_redo_key err %i", ret);
- return ret;
+ bch2_bkey_val_to_text(&buf, c, hash_k), buf.buf))) {
+ ret = hash_redo_key(trans, desc, hash_info, k_iter, hash_k);
+ bch_err_fn(c, ret);
+ if (ret)
+ return ret;
+ ret = -BCH_ERR_transaction_restart_nested;
}
- ret = -EINTR;
fsck_err:
goto out;
}
static int check_inode(struct btree_trans *trans,
struct btree_iter *iter,
+ struct bkey_s_c k,
struct bch_inode_unpacked *prev,
+ struct snapshots_seen *s,
bool full)
{
struct bch_fs *c = trans->c;
- struct bkey_s_c k;
struct bch_inode_unpacked u;
bool do_update = false;
int ret;
- k = bch2_btree_iter_peek(iter);
- if (!k.k)
- return 0;
-
- ret = bkey_err(k);
- if (ret)
- return ret;
-
ret = check_key_has_snapshot(trans, iter, k);
if (ret < 0)
goto err;
if (ret)
return 0;
- /*
- * if snapshot id isn't a leaf node, skip it - deletion in
- * particular is not atomic, so on the internal snapshot nodes
- * we can see inodes marked for deletion after a clean shutdown
- */
- if (bch2_snapshot_internal_node(c, k.k->p.snapshot))
- return 0;
+ ret = snapshots_seen_update(c, s, iter->btree_id, k.k->p);
+ if (ret)
+ goto err;
if (!bkey_is_inode(k.k))
return 0;
BUG_ON(bch2_inode_unpack(k, &u));
if (!full &&
- !(u.bi_flags & (BCH_INODE_I_SIZE_DIRTY|
- BCH_INODE_I_SECTORS_DIRTY|
- BCH_INODE_UNLINKED)))
+ !(u.bi_flags & (BCH_INODE_i_size_dirty|
+ BCH_INODE_i_sectors_dirty|
+ BCH_INODE_unlinked)))
return 0;
if (prev->bi_inum != u.bi_inum)
*prev = u;
if (fsck_err_on(prev->bi_hash_seed != u.bi_hash_seed ||
- inode_d_type(prev) != inode_d_type(&u), c,
+ inode_d_type(prev) != inode_d_type(&u),
+ c, inode_snapshot_mismatch,
"inodes in different snapshots don't match")) {
bch_err(c, "repair not implemented yet");
return -EINVAL;
}
- if (u.bi_flags & BCH_INODE_UNLINKED &&
+ if ((u.bi_flags & (BCH_INODE_i_size_dirty|BCH_INODE_unlinked)) &&
+ bch2_key_has_snapshot_overwrites(trans, BTREE_ID_inodes, k.k->p)) {
+ struct bpos new_min_pos;
+
+ ret = bch2_propagate_key_to_snapshot_leaves(trans, iter->btree_id, k, &new_min_pos);
+ if (ret)
+ goto err;
+
+ u.bi_flags &= ~BCH_INODE_i_size_dirty|BCH_INODE_unlinked;
+
+ ret = __write_inode(trans, &u, iter->pos.snapshot);
+ bch_err_msg(c, ret, "in fsck updating inode");
+ if (ret)
+ return ret;
+
+ if (!bpos_eq(new_min_pos, POS_MIN))
+ bch2_btree_iter_set_pos(iter, bpos_predecessor(new_min_pos));
+ return 0;
+ }
+
+ if (u.bi_flags & BCH_INODE_unlinked &&
(!c->sb.clean ||
- fsck_err(c, "filesystem marked clean, but inode %llu unlinked",
+ fsck_err(c, inode_unlinked_but_clean,
+ "filesystem marked clean, but inode %llu unlinked",
u.bi_inum))) {
bch2_trans_unlock(trans);
bch2_fs_lazy_rw(c);
- ret = fsck_inode_rm(trans, u.bi_inum, iter->pos.snapshot);
- if (ret)
- bch_err(c, "error in fsck: error %i while deleting inode", ret);
+ ret = bch2_inode_rm_snapshot(trans, u.bi_inum, iter->pos.snapshot);
+ bch_err_msg(c, ret, "in fsck deleting inode");
return ret;
}
- if (u.bi_flags & BCH_INODE_I_SIZE_DIRTY &&
+ if (u.bi_flags & BCH_INODE_i_size_dirty &&
(!c->sb.clean ||
- fsck_err(c, "filesystem marked clean, but inode %llu has i_size dirty",
+ fsck_err(c, inode_i_size_dirty_but_clean,
+ "filesystem marked clean, but inode %llu has i_size dirty",
u.bi_inum))) {
bch_verbose(c, "truncating inode %llu", u.bi_inum);
iter->pos.snapshot),
POS(u.bi_inum, U64_MAX),
0, NULL);
- if (ret) {
- bch_err(c, "error in fsck: error %i truncating inode", ret);
+ bch_err_msg(c, ret, "in fsck truncating inode");
+ if (ret)
return ret;
- }
/*
* We truncated without our normal sector accounting hook, just
* make sure we recalculate it:
*/
- u.bi_flags |= BCH_INODE_I_SECTORS_DIRTY;
+ u.bi_flags |= BCH_INODE_i_sectors_dirty;
- u.bi_flags &= ~BCH_INODE_I_SIZE_DIRTY;
+ u.bi_flags &= ~BCH_INODE_i_size_dirty;
do_update = true;
}
- if (u.bi_flags & BCH_INODE_I_SECTORS_DIRTY &&
+ if (u.bi_flags & BCH_INODE_i_sectors_dirty &&
(!c->sb.clean ||
- fsck_err(c, "filesystem marked clean, but inode %llu has i_sectors dirty",
+ fsck_err(c, inode_i_sectors_dirty_but_clean,
+ "filesystem marked clean, but inode %llu has i_sectors dirty",
u.bi_inum))) {
s64 sectors;
sectors = bch2_count_inode_sectors(trans, u.bi_inum, iter->pos.snapshot);
if (sectors < 0) {
- bch_err(c, "error in fsck: error %i recounting inode sectors",
- (int) sectors);
+ bch_err_msg(c, sectors, "in fsck recounting inode sectors");
return sectors;
}
u.bi_sectors = sectors;
- u.bi_flags &= ~BCH_INODE_I_SECTORS_DIRTY;
+ u.bi_flags &= ~BCH_INODE_i_sectors_dirty;
do_update = true;
}
- if (u.bi_flags & BCH_INODE_BACKPTR_UNTRUSTED) {
+ if (u.bi_flags & BCH_INODE_backptr_untrusted) {
u.bi_dir = 0;
u.bi_dir_offset = 0;
- u.bi_flags &= ~BCH_INODE_BACKPTR_UNTRUSTED;
+ u.bi_flags &= ~BCH_INODE_backptr_untrusted;
do_update = true;
}
if (do_update) {
- ret = write_inode(trans, &u, iter->pos.snapshot);
+ ret = __write_inode(trans, &u, iter->pos.snapshot);
+ bch_err_msg(c, ret, "in fsck updating inode");
if (ret)
- bch_err(c, "error in fsck: error %i "
- "updating inode", ret);
+ return ret;
}
err:
fsck_err:
- if (ret)
- bch_err(c, "error %i from check_inode()", ret);
+ bch_err_fn(c, ret);
return ret;
}
noinline_for_stack
-static int check_inodes(struct bch_fs *c, bool full)
+int bch2_check_inodes(struct bch_fs *c)
{
- struct btree_trans trans;
+ bool full = c->opts.fsck;
+ struct btree_trans *trans = bch2_trans_get(c);
struct btree_iter iter;
struct bch_inode_unpacked prev = { 0 };
+ struct snapshots_seen s;
+ struct bkey_s_c k;
int ret;
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 0);
+ snapshots_seen_init(&s);
- bch2_trans_iter_init(&trans, &iter, BTREE_ID_inodes, POS_MIN,
- BTREE_ITER_INTENT|
- BTREE_ITER_PREFETCH|
- BTREE_ITER_ALL_SNAPSHOTS);
+ ret = for_each_btree_key_commit(trans, iter, BTREE_ID_inodes,
+ POS_MIN,
+ BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+ NULL, NULL, BTREE_INSERT_LAZY_RW|BTREE_INSERT_NOFAIL,
+ check_inode(trans, &iter, k, &prev, &s, full));
- do {
- ret = __bch2_trans_do(&trans, NULL, NULL,
- BTREE_INSERT_LAZY_RW|
- BTREE_INSERT_NOFAIL,
- check_inode(&trans, &iter, &prev, full));
- if (ret)
- break;
- } while (bch2_btree_iter_advance(&iter));
- bch2_trans_iter_exit(&trans, &iter);
-
- bch2_trans_exit(&trans);
- if (ret)
- bch_err(c, "error %i from check_inodes()", ret);
+ snapshots_seen_exit(&s);
+ bch2_trans_put(trans);
+ bch_err_fn(c, ret);
return ret;
}
-static int check_subvol(struct btree_trans *trans,
- struct btree_iter *iter)
+static struct bkey_s_c_dirent dirent_get_by_pos(struct btree_trans *trans,
+ struct btree_iter *iter,
+ struct bpos pos)
{
- struct bkey_s_c k;
- struct bkey_s_c_subvolume subvol;
- int ret;
-
- k = bch2_btree_iter_peek(iter);
- if (!k.k)
- return 0;
-
- ret = bkey_err(k);
- if (ret)
- return ret;
-
- if (k.k->type != KEY_TYPE_subvolume)
- return 0;
-
- subvol = bkey_s_c_to_subvolume(k);
+ return bch2_bkey_get_iter_typed(trans, iter, BTREE_ID_dirents, pos, 0, dirent);
+}
- if (BCH_SUBVOLUME_UNLINKED(subvol.v)) {
- ret = bch2_subvolume_delete(trans, iter->pos.offset);
- if (ret && ret != -EINTR)
- bch_err(trans->c, "error deleting subvolume %llu: %i",
- iter->pos.offset, ret);
- if (ret)
- return ret;
- }
+static bool inode_points_to_dirent(struct bch_inode_unpacked *inode,
+ struct bkey_s_c_dirent d)
+{
+ return inode->bi_dir == d.k->p.inode &&
+ inode->bi_dir_offset == d.k->p.offset;
+}
- return 0;
+static bool dirent_points_to_inode(struct bkey_s_c_dirent d,
+ struct bch_inode_unpacked *inode)
+{
+ return d.v->d_type == DT_SUBVOL
+ ? le32_to_cpu(d.v->d_child_subvol) == inode->bi_subvol
+ : le64_to_cpu(d.v->d_inum) == inode->bi_inum;
}
-noinline_for_stack
-static int check_subvols(struct bch_fs *c)
+static int inode_backpointer_exists(struct btree_trans *trans,
+ struct bch_inode_unpacked *inode,
+ u32 snapshot)
{
- struct btree_trans trans;
struct btree_iter iter;
+ struct bkey_s_c_dirent d;
int ret;
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 0);
-
- bch2_trans_iter_init(&trans, &iter, BTREE_ID_subvolumes,
- POS_MIN,
- BTREE_ITER_INTENT|
- BTREE_ITER_PREFETCH);
-
- do {
- ret = __bch2_trans_do(&trans, NULL, NULL,
- BTREE_INSERT_LAZY_RW|
- BTREE_INSERT_NOFAIL,
- check_subvol(&trans, &iter));
- if (ret)
- break;
- } while (bch2_btree_iter_advance(&iter));
- bch2_trans_iter_exit(&trans, &iter);
+ d = dirent_get_by_pos(trans, &iter,
+ SPOS(inode->bi_dir, inode->bi_dir_offset, snapshot));
+ ret = bkey_err(d);
+ if (ret)
+ return bch2_err_matches(ret, ENOENT) ? 0 : ret;
- bch2_trans_exit(&trans);
+ ret = dirent_points_to_inode(d, inode);
+ bch2_trans_iter_exit(trans, &iter);
return ret;
}
-/*
- * Checking for overlapping extents needs to be reimplemented
- */
-#if 0
-static int fix_overlapping_extent(struct btree_trans *trans,
- struct bkey_s_c k, struct bpos cut_at)
+static int check_i_sectors(struct btree_trans *trans, struct inode_walker *w)
{
- struct btree_iter iter;
- struct bkey_i *u;
- int ret;
+ struct bch_fs *c = trans->c;
+ struct inode_walker_entry *i;
+ u32 restart_count = trans->restart_count;
+ int ret = 0;
+ s64 count2;
- u = bch2_trans_kmalloc(trans, bkey_bytes(k.k));
- ret = PTR_ERR_OR_ZERO(u);
- if (ret)
- return ret;
+ darray_for_each(w->inodes, i) {
+ if (i->inode.bi_sectors == i->count)
+ continue;
- bkey_reassemble(u, k);
- bch2_cut_front(cut_at, u);
+ count2 = bch2_count_inode_sectors(trans, w->last_pos.inode, i->snapshot);
+ if (w->recalculate_sums)
+ i->count = count2;
- /*
- * We don't want to go through the extent_handle_overwrites path:
- *
- * XXX: this is going to screw up disk accounting, extent triggers
- * assume things about extent overwrites - we should be running the
- * triggers manually here
- */
- bch2_trans_iter_init(trans, &iter, BTREE_ID_extents, u->k.p,
- BTREE_ITER_INTENT|BTREE_ITER_NOT_EXTENTS);
+ if (i->count != count2) {
+ bch_err(c, "fsck counted i_sectors wrong for inode %llu:%u: got %llu should be %llu",
+ w->last_pos.inode, i->snapshot, i->count, count2);
+ return -BCH_ERR_internal_fsck_err;
+ }
- BUG_ON(iter.flags & BTREE_ITER_IS_EXTENTS);
- ret = bch2_btree_iter_traverse(&iter) ?:
- bch2_trans_update(trans, &iter, u, BTREE_TRIGGER_NORUN) ?:
- bch2_trans_commit(trans, NULL, NULL,
- BTREE_INSERT_NOFAIL|
- BTREE_INSERT_LAZY_RW);
- bch2_trans_iter_exit(trans, &iter);
- return ret;
+ if (fsck_err_on(!(i->inode.bi_flags & BCH_INODE_i_sectors_dirty),
+ c, inode_i_sectors_wrong,
+ "inode %llu:%u has incorrect i_sectors: got %llu, should be %llu",
+ w->last_pos.inode, i->snapshot,
+ i->inode.bi_sectors, i->count)) {
+ i->inode.bi_sectors = i->count;
+ ret = fsck_write_inode(trans, &i->inode, i->snapshot);
+ if (ret)
+ break;
+ }
+ }
+fsck_err:
+ bch_err_fn(c, ret);
+ return ret ?: trans_was_restarted(trans, restart_count);
}
-#endif
-static struct bkey_s_c_dirent dirent_get_by_pos(struct btree_trans *trans,
- struct btree_iter *iter,
- struct bpos pos)
+struct extent_end {
+ u32 snapshot;
+ u64 offset;
+ struct snapshots_seen seen;
+};
+
+struct extent_ends {
+ struct bpos last_pos;
+ DARRAY(struct extent_end) e;
+};
+
+static void extent_ends_reset(struct extent_ends *extent_ends)
{
- struct bkey_s_c k;
- int ret;
+ struct extent_end *i;
- bch2_trans_iter_init(trans, iter, BTREE_ID_dirents, pos, 0);
- k = bch2_btree_iter_peek_slot(iter);
- ret = bkey_err(k);
- if (!ret && k.k->type != KEY_TYPE_dirent)
- ret = -ENOENT;
- if (ret) {
- bch2_trans_iter_exit(trans, iter);
- return (struct bkey_s_c_dirent) { .k = ERR_PTR(ret) };
- }
+ darray_for_each(extent_ends->e, i)
+ snapshots_seen_exit(&i->seen);
- return bkey_s_c_to_dirent(k);
+ extent_ends->e.nr = 0;
}
-static bool inode_points_to_dirent(struct bch_inode_unpacked *inode,
- struct bkey_s_c_dirent d)
+static void extent_ends_exit(struct extent_ends *extent_ends)
{
- return inode->bi_dir == d.k->p.inode &&
- inode->bi_dir_offset == d.k->p.offset;
+ extent_ends_reset(extent_ends);
+ darray_exit(&extent_ends->e);
}
-static bool dirent_points_to_inode(struct bkey_s_c_dirent d,
- struct bch_inode_unpacked *inode)
+static void extent_ends_init(struct extent_ends *extent_ends)
{
- return d.v->d_type == DT_SUBVOL
- ? le32_to_cpu(d.v->d_child_subvol) == inode->bi_subvol
- : le64_to_cpu(d.v->d_inum) == inode->bi_inum;
+ memset(extent_ends, 0, sizeof(*extent_ends));
}
-static int inode_backpointer_exists(struct btree_trans *trans,
- struct bch_inode_unpacked *inode,
- u32 snapshot)
+static int extent_ends_at(struct bch_fs *c,
+ struct extent_ends *extent_ends,
+ struct snapshots_seen *seen,
+ struct bkey_s_c k)
{
- struct btree_iter iter;
- struct bkey_s_c_dirent d;
+ struct extent_end *i, n = (struct extent_end) {
+ .offset = k.k->p.offset,
+ .snapshot = k.k->p.snapshot,
+ .seen = *seen,
+ };
+
+ n.seen.ids.data = kmemdup(seen->ids.data,
+ sizeof(seen->ids.data[0]) * seen->ids.size,
+ GFP_KERNEL);
+ if (!n.seen.ids.data)
+ return -BCH_ERR_ENOMEM_fsck_extent_ends_at;
+
+ darray_for_each(extent_ends->e, i) {
+ if (i->snapshot == k.k->p.snapshot) {
+ snapshots_seen_exit(&i->seen);
+ *i = n;
+ return 0;
+ }
+
+ if (i->snapshot >= k.k->p.snapshot)
+ break;
+ }
+
+ return darray_insert_item(&extent_ends->e, i - extent_ends->e.data, n);
+}
+
+static int overlapping_extents_found(struct btree_trans *trans,
+ enum btree_id btree,
+ struct bpos pos1, struct snapshots_seen *pos1_seen,
+ struct bkey pos2,
+ bool *fixed,
+ struct extent_end *extent_end)
+{
+ struct bch_fs *c = trans->c;
+ struct printbuf buf = PRINTBUF;
+ struct btree_iter iter1, iter2 = { NULL };
+ struct bkey_s_c k1, k2;
int ret;
- d = dirent_get_by_pos(trans, &iter,
- SPOS(inode->bi_dir, inode->bi_dir_offset, snapshot));
- ret = bkey_err(d.s_c);
+ BUG_ON(bkey_le(pos1, bkey_start_pos(&pos2)));
+
+ bch2_trans_iter_init(trans, &iter1, btree, pos1,
+ BTREE_ITER_ALL_SNAPSHOTS|
+ BTREE_ITER_NOT_EXTENTS);
+ k1 = bch2_btree_iter_peek_upto(&iter1, POS(pos1.inode, U64_MAX));
+ ret = bkey_err(k1);
if (ret)
- return ret == -ENOENT ? 0 : ret;
+ goto err;
+
+ prt_str(&buf, "\n ");
+ bch2_bkey_val_to_text(&buf, c, k1);
+
+ if (!bpos_eq(pos1, k1.k->p)) {
+ prt_str(&buf, "\n wanted\n ");
+ bch2_bpos_to_text(&buf, pos1);
+ prt_str(&buf, "\n ");
+ bch2_bkey_to_text(&buf, &pos2);
+
+ bch_err(c, "%s: error finding first overlapping extent when repairing, got%s",
+ __func__, buf.buf);
+ ret = -BCH_ERR_internal_fsck_err;
+ goto err;
+ }
+
+ bch2_trans_copy_iter(&iter2, &iter1);
+
+ while (1) {
+ bch2_btree_iter_advance(&iter2);
+
+ k2 = bch2_btree_iter_peek_upto(&iter2, POS(pos1.inode, U64_MAX));
+ ret = bkey_err(k2);
+ if (ret)
+ goto err;
+
+ if (bpos_ge(k2.k->p, pos2.p))
+ break;
+ }
+
+ prt_str(&buf, "\n ");
+ bch2_bkey_val_to_text(&buf, c, k2);
+
+ if (bpos_gt(k2.k->p, pos2.p) ||
+ pos2.size != k2.k->size) {
+ bch_err(c, "%s: error finding seconding overlapping extent when repairing%s",
+ __func__, buf.buf);
+ ret = -BCH_ERR_internal_fsck_err;
+ goto err;
+ }
+
+ prt_printf(&buf, "\n overwriting %s extent",
+ pos1.snapshot >= pos2.p.snapshot ? "first" : "second");
+
+ if (fsck_err(c, extent_overlapping,
+ "overlapping extents%s", buf.buf)) {
+ struct btree_iter *old_iter = &iter1;
+ struct disk_reservation res = { 0 };
+
+ if (pos1.snapshot < pos2.p.snapshot) {
+ old_iter = &iter2;
+ swap(k1, k2);
+ }
+
+ trans->extra_journal_res += bch2_bkey_sectors_compressed(k2);
- ret = dirent_points_to_inode(d, inode);
- bch2_trans_iter_exit(trans, &iter);
+ ret = bch2_trans_update_extent_overwrite(trans, old_iter,
+ BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE,
+ k1, k2) ?:
+ bch2_trans_commit(trans, &res, NULL,
+ BTREE_INSERT_LAZY_RW|BTREE_INSERT_NOFAIL);
+ bch2_disk_reservation_put(c, &res);
+
+ if (ret)
+ goto err;
+
+ *fixed = true;
+
+ if (pos1.snapshot == pos2.p.snapshot) {
+ /*
+ * We overwrote the first extent, and did the overwrite
+ * in the same snapshot:
+ */
+ extent_end->offset = bkey_start_offset(&pos2);
+ } else if (pos1.snapshot > pos2.p.snapshot) {
+ /*
+ * We overwrote the first extent in pos2's snapshot:
+ */
+ ret = snapshots_seen_add_inorder(c, pos1_seen, pos2.p.snapshot);
+ } else {
+ /*
+ * We overwrote the second extent - restart
+ * check_extent() from the top:
+ */
+ ret = -BCH_ERR_transaction_restart_nested;
+ }
+ }
+fsck_err:
+err:
+ bch2_trans_iter_exit(trans, &iter2);
+ bch2_trans_iter_exit(trans, &iter1);
+ printbuf_exit(&buf);
return ret;
}
-static int check_i_sectors(struct btree_trans *trans, struct inode_walker *w)
+static int check_overlapping_extents(struct btree_trans *trans,
+ struct snapshots_seen *seen,
+ struct extent_ends *extent_ends,
+ struct bkey_s_c k,
+ u32 equiv,
+ struct btree_iter *iter,
+ bool *fixed)
{
struct bch_fs *c = trans->c;
- struct inode_walker_entry *i;
- int ret = 0, ret2 = 0;
- s64 count2;
+ struct extent_end *i;
+ int ret = 0;
- darray_for_each(w->inodes, i) {
- if (i->inode.bi_sectors == i->count)
- continue;
+ /* transaction restart, running again */
+ if (bpos_eq(extent_ends->last_pos, k.k->p))
+ return 0;
- count2 = lockrestart_do(trans,
- bch2_count_inode_sectors(trans, w->cur_inum, i->snapshot));
+ if (extent_ends->last_pos.inode != k.k->p.inode)
+ extent_ends_reset(extent_ends);
- if (i->count != count2) {
- bch_err(c, "fsck counted i_sectors wrong: got %llu should be %llu",
- i->count, count2);
- i->count = count2;
- if (i->inode.bi_sectors == i->count)
- continue;
- }
+ darray_for_each(extent_ends->e, i) {
+ if (i->offset <= bkey_start_offset(k.k))
+ continue;
- if (fsck_err_on(!(i->inode.bi_flags & BCH_INODE_I_SECTORS_DIRTY), c,
- "inode %llu:%u has incorrect i_sectors: got %llu, should be %llu",
- w->cur_inum, i->snapshot,
- i->inode.bi_sectors, i->count) == FSCK_ERR_IGNORE)
+ if (!ref_visible2(c,
+ k.k->p.snapshot, seen,
+ i->snapshot, &i->seen))
continue;
- i->inode.bi_sectors = i->count;
- ret = write_inode(trans, &i->inode, i->snapshot);
+ ret = overlapping_extents_found(trans, iter->btree_id,
+ SPOS(iter->pos.inode,
+ i->offset,
+ i->snapshot),
+ &i->seen,
+ *k.k, fixed, i);
if (ret)
- break;
- ret2 = -EINTR;
+ goto err;
}
-fsck_err:
+
+ ret = extent_ends_at(c, extent_ends, seen, k);
if (ret)
- bch_err(c, "error %i from check_i_sectors()", ret);
- return ret ?: ret2;
+ goto err;
+
+ extent_ends->last_pos = k.k->p;
+err:
+ return ret;
+}
+
+static int check_extent_overbig(struct btree_trans *trans, struct btree_iter *iter,
+ struct bkey_s_c k)
+{
+ struct bch_fs *c = trans->c;
+ struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
+ struct bch_extent_crc_unpacked crc;
+ const union bch_extent_entry *i;
+ unsigned encoded_extent_max_sectors = c->opts.encoded_extent_max >> 9;
+
+ bkey_for_each_crc(k.k, ptrs, crc, i)
+ if (crc_is_encoded(crc) &&
+ crc.uncompressed_size > encoded_extent_max_sectors) {
+ struct printbuf buf = PRINTBUF;
+
+ bch2_bkey_val_to_text(&buf, c, k);
+ bch_err(c, "overbig encoded extent, please report this:\n %s", buf.buf);
+ printbuf_exit(&buf);
+ }
+
+ return 0;
}
static int check_extent(struct btree_trans *trans, struct btree_iter *iter,
+ struct bkey_s_c k,
struct inode_walker *inode,
- struct snapshots_seen *s)
+ struct snapshots_seen *s,
+ struct extent_ends *extent_ends)
{
struct bch_fs *c = trans->c;
- struct bkey_s_c k;
struct inode_walker_entry *i;
struct printbuf buf = PRINTBUF;
+ struct bpos equiv = k.k->p;
int ret = 0;
-peek:
- k = bch2_btree_iter_peek(iter);
- if (!k.k)
- goto out;
- ret = bkey_err(k);
- if (ret)
- goto err;
+ equiv.snapshot = bch2_snapshot_equiv(c, k.k->p.snapshot);
ret = check_key_has_snapshot(trans, iter, k);
if (ret) {
goto out;
}
- ret = snapshots_seen_update(c, s, k.k->p);
- if (ret)
- goto err;
-
- if (k.k->type == KEY_TYPE_whiteout)
- goto out;
-
- if (inode->cur_inum != k.k->p.inode) {
+ if (inode->last_pos.inode != k.k->p.inode) {
ret = check_i_sectors(trans, inode);
if (ret)
goto err;
}
- if (!iter->path->should_be_locked) {
- /*
- * hack: check_i_sectors may have handled a transaction restart,
- * it shouldn't be but we need to fix the new i_sectors check
- * code and delete the old bch2_count_inode_sectors() first
- */
- goto peek;
- }
-#if 0
- if (bkey_cmp(prev.k->k.p, bkey_start_pos(k.k)) > 0) {
- char buf1[200];
- char buf2[200];
-
- bch2_bkey_val_to_text(&PBUF(buf1), c, bkey_i_to_s_c(prev.k));
- bch2_bkey_val_to_text(&PBUF(buf2), c, k);
+ i = walk_inode(trans, inode, equiv, k.k->type == KEY_TYPE_whiteout);
+ ret = PTR_ERR_OR_ZERO(i);
+ if (ret)
+ goto err;
- if (fsck_err(c, "overlapping extents:\n%s\n%s", buf1, buf2)) {
- ret = fix_overlapping_extent(trans, k, prev.k->k.p) ?: -EINTR;
- goto out;
- }
- }
-#endif
- ret = __walk_inode(trans, inode, k.k->p);
- if (ret < 0)
+ ret = snapshots_seen_update(c, s, iter->btree_id, k.k->p);
+ if (ret)
goto err;
- if (fsck_err_on(ret == INT_MAX, c,
- "extent in missing inode:\n %s",
- (printbuf_reset(&buf),
- bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
- ret = bch2_btree_delete_at(trans, iter,
- BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
- goto out;
- }
+ if (k.k->type != KEY_TYPE_whiteout) {
+ if (fsck_err_on(!i, c, extent_in_missing_inode,
+ "extent in missing inode:\n %s",
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+ goto delete;
+
+ if (fsck_err_on(i &&
+ !S_ISREG(i->inode.bi_mode) &&
+ !S_ISLNK(i->inode.bi_mode),
+ c, extent_in_non_reg_inode,
+ "extent in non regular inode mode %o:\n %s",
+ i->inode.bi_mode,
+ (printbuf_reset(&buf),
+ bch2_bkey_val_to_text(&buf, c, k), buf.buf)))
+ goto delete;
- if (ret == INT_MAX) {
- ret = 0;
- goto out;
+ ret = check_overlapping_extents(trans, s, extent_ends, k,
+ equiv.snapshot, iter,
+ &inode->recalculate_sums);
+ if (ret)
+ goto err;
}
- i = inode->inodes.data + ret;
- ret = 0;
-
- if (fsck_err_on(!S_ISREG(i->inode.bi_mode) &&
- !S_ISLNK(i->inode.bi_mode), c,
- "extent in non regular inode mode %o:\n %s",
- i->inode.bi_mode,
- (printbuf_reset(&buf),
- bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
- ret = bch2_btree_delete_at(trans, iter,
- BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
- goto out;
- }
+ /*
+ * Check inodes in reverse order, from oldest snapshots to newest,
+ * starting from the inode that matches this extent's snapshot. If we
+ * didn't have one, iterate over all inodes:
+ */
+ if (!i)
+ i = inode->inodes.data + inode->inodes.nr - 1;
+
+ for (;
+ inode->inodes.data && i >= inode->inodes.data;
+ --i) {
+ if (i->snapshot > equiv.snapshot ||
+ !key_visible_in_snapshot(c, s, i->snapshot, equiv.snapshot))
+ continue;
- if (!bch2_snapshot_internal_node(c, k.k->p.snapshot)) {
- for_each_visible_inode(c, s, inode, k.k->p.snapshot, i) {
- if (fsck_err_on(!(i->inode.bi_flags & BCH_INODE_I_SIZE_DIRTY) &&
- k.k->type != KEY_TYPE_reservation &&
- k.k->p.offset > round_up(i->inode.bi_size, block_bytes(c)) >> 9, c,
- "extent type %u offset %llu past end of inode %llu, i_size %llu",
- k.k->type, k.k->p.offset, k.k->p.inode, i->inode.bi_size)) {
- bch2_fs_lazy_rw(c);
- ret = bch2_btree_delete_range_trans(trans, BTREE_ID_extents,
- SPOS(k.k->p.inode, round_up(i->inode.bi_size, block_bytes(c)) >> 9,
- k.k->p.snapshot),
- POS(k.k->p.inode, U64_MAX),
- 0, NULL) ?: -EINTR;
- goto out;
+ if (k.k->type != KEY_TYPE_whiteout) {
+ if (fsck_err_on(!(i->inode.bi_flags & BCH_INODE_i_size_dirty) &&
+ k.k->p.offset > round_up(i->inode.bi_size, block_bytes(c)) >> 9 &&
+ !bkey_extent_is_reservation(k),
+ c, extent_past_end_of_inode,
+ "extent type past end of inode %llu:%u, i_size %llu\n %s",
+ i->inode.bi_inum, i->snapshot, i->inode.bi_size,
+ (bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
+ struct btree_iter iter2;
+
+ bch2_trans_copy_iter(&iter2, iter);
+ bch2_btree_iter_set_snapshot(&iter2, i->snapshot);
+ ret = bch2_btree_iter_traverse(&iter2) ?:
+ bch2_btree_delete_at(trans, &iter2,
+ BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+ bch2_trans_iter_exit(trans, &iter2);
+ if (ret)
+ goto err;
+
+ iter->k.type = KEY_TYPE_whiteout;
}
- }
- }
- if (bkey_extent_is_allocation(k.k))
- for_each_visible_inode(c, s, inode, k.k->p.snapshot, i)
- i->count += k.k->size;
-#if 0
- bch2_bkey_buf_reassemble(&prev, c, k);
-#endif
+ if (bkey_extent_is_allocation(k.k))
+ i->count += k.k->size;
+ }
+ i->seen_this_pos = true;
+ }
out:
err:
fsck_err:
printbuf_exit(&buf);
-
- if (ret && ret != -EINTR)
- bch_err(c, "error %i from check_extent()", ret);
+ bch_err_fn(c, ret);
return ret;
+delete:
+ ret = bch2_btree_delete_at(trans, iter, BTREE_UPDATE_INTERNAL_SNAPSHOT_NODE);
+ goto out;
}
/*
* Walk extents: verify that extents have a corresponding S_ISREG inode, and
* that i_size an i_sectors are consistent
*/
-noinline_for_stack
-static int check_extents(struct bch_fs *c)
+int bch2_check_extents(struct bch_fs *c)
{
struct inode_walker w = inode_walker_init();
struct snapshots_seen s;
- struct btree_trans trans;
+ struct btree_trans *trans = bch2_trans_get(c);
struct btree_iter iter;
+ struct bkey_s_c k;
+ struct extent_ends extent_ends;
+ struct disk_reservation res = { 0 };
int ret = 0;
-#if 0
- struct bkey_buf prev;
- bch2_bkey_buf_init(&prev);
- prev.k->k = KEY(0, 0, 0);
-#endif
snapshots_seen_init(&s);
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 0);
+ extent_ends_init(&extent_ends);
+
+ ret = for_each_btree_key_commit(trans, iter, BTREE_ID_extents,
+ POS(BCACHEFS_ROOT_INO, 0),
+ BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+ &res, NULL,
+ BTREE_INSERT_LAZY_RW|BTREE_INSERT_NOFAIL, ({
+ bch2_disk_reservation_put(c, &res);
+ check_extent(trans, &iter, k, &w, &s, &extent_ends) ?:
+ check_extent_overbig(trans, &iter, k);
+ })) ?:
+ check_i_sectors(trans, &w);
+
+ bch2_disk_reservation_put(c, &res);
+ extent_ends_exit(&extent_ends);
+ inode_walker_exit(&w);
+ snapshots_seen_exit(&s);
+ bch2_trans_put(trans);
- bch_verbose(c, "checking extents");
+ bch_err_fn(c, ret);
+ return ret;
+}
- bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents,
- POS(BCACHEFS_ROOT_INO, 0),
- BTREE_ITER_INTENT|
- BTREE_ITER_PREFETCH|
- BTREE_ITER_ALL_SNAPSHOTS);
+int bch2_check_indirect_extents(struct bch_fs *c)
+{
+ struct btree_trans *trans = bch2_trans_get(c);
+ struct btree_iter iter;
+ struct bkey_s_c k;
+ struct disk_reservation res = { 0 };
+ int ret = 0;
- do {
- ret = __bch2_trans_do(&trans, NULL, NULL,
- BTREE_INSERT_LAZY_RW|
- BTREE_INSERT_NOFAIL,
- check_extent(&trans, &iter, &w, &s));
- if (ret)
- break;
- } while (bch2_btree_iter_advance(&iter));
- bch2_trans_iter_exit(&trans, &iter);
-#if 0
- bch2_bkey_buf_exit(&prev, c);
-#endif
- inode_walker_exit(&w);
- bch2_trans_exit(&trans);
- snapshots_seen_exit(&s);
+ ret = for_each_btree_key_commit(trans, iter, BTREE_ID_reflink,
+ POS_MIN,
+ BTREE_ITER_PREFETCH, k,
+ &res, NULL,
+ BTREE_INSERT_LAZY_RW|BTREE_INSERT_NOFAIL, ({
+ bch2_disk_reservation_put(c, &res);
+ check_extent_overbig(trans, &iter, k);
+ }));
- if (ret)
- bch_err(c, "error %i from check_extents()", ret);
+ bch2_disk_reservation_put(c, &res);
+ bch2_trans_put(trans);
+
+ bch_err_fn(c, ret);
return ret;
}
{
struct bch_fs *c = trans->c;
struct inode_walker_entry *i;
- int ret = 0, ret2 = 0;
+ u32 restart_count = trans->restart_count;
+ int ret = 0;
s64 count2;
darray_for_each(w->inodes, i) {
if (i->inode.bi_nlink == i->count)
continue;
- count2 = bch2_count_subdirs(trans, w->cur_inum, i->snapshot);
+ count2 = bch2_count_subdirs(trans, w->last_pos.inode, i->snapshot);
if (count2 < 0)
return count2;
continue;
}
- if (fsck_err_on(i->inode.bi_nlink != i->count, c,
+ if (fsck_err_on(i->inode.bi_nlink != i->count,
+ c, inode_dir_wrong_nlink,
"directory %llu:%u with wrong i_nlink: got %u, should be %llu",
- w->cur_inum, i->snapshot, i->inode.bi_nlink, i->count)) {
+ w->last_pos.inode, i->snapshot, i->inode.bi_nlink, i->count)) {
i->inode.bi_nlink = i->count;
- ret = write_inode(trans, &i->inode, i->snapshot);
+ ret = fsck_write_inode(trans, &i->inode, i->snapshot);
if (ret)
break;
- ret2 = -EINTR;
}
}
fsck_err:
- if (ret)
- bch_err(c, "error %i from check_subdir_count()", ret);
- return ret ?: ret2;
+ bch_err_fn(c, ret);
+ return ret ?: trans_was_restarted(trans, restart_count);
}
static int check_dirent_target(struct btree_trans *trans,
backpointer_exists = ret;
ret = 0;
- if (fsck_err_on(S_ISDIR(target->bi_mode) &&
- backpointer_exists, c,
+ if (fsck_err_on(S_ISDIR(target->bi_mode) && backpointer_exists,
+ c, inode_dir_multiple_links,
"directory %llu with multiple links",
target->bi_inum)) {
ret = __remove_dirent(trans, d.k->p);
goto out;
}
- if (fsck_err_on(backpointer_exists &&
- !target->bi_nlink, c,
+ if (fsck_err_on(backpointer_exists && !target->bi_nlink,
+ c, inode_multiple_links_but_nlink_0,
"inode %llu type %s has multiple links but i_nlink 0",
target->bi_inum, bch2_d_types[d.v->d_type])) {
target->bi_nlink++;
- target->bi_flags &= ~BCH_INODE_UNLINKED;
+ target->bi_flags &= ~BCH_INODE_unlinked;
ret = __write_inode(trans, target, target_snapshot);
if (ret)
goto err;
}
- if (fsck_err_on(!backpointer_exists, c,
+ if (fsck_err_on(!backpointer_exists,
+ c, inode_wrong_backpointer,
"inode %llu:%u has wrong backpointer:\n"
"got %llu:%llu\n"
"should be %llu:%llu",
}
}
- if (fsck_err_on(d.v->d_type != inode_d_type(target), c,
+ if (fsck_err_on(d.v->d_type != inode_d_type(target),
+ c, dirent_d_type_wrong,
"incorrect d_type: got %s, should be %s:\n%s",
bch2_d_type_str(d.v->d_type),
bch2_d_type_str(inode_d_type(target)),
if (d.v->d_type == DT_SUBVOL &&
target->bi_parent_subvol != le32_to_cpu(d.v->d_parent_subvol) &&
(c->sb.version < bcachefs_metadata_version_subvol_dirent ||
- fsck_err(c, "dirent has wrong d_parent_subvol field: got %u, should be %u",
+ fsck_err(c, dirent_d_parent_subvol_wrong,
+ "dirent has wrong d_parent_subvol field: got %u, should be %u",
le32_to_cpu(d.v->d_parent_subvol),
target->bi_parent_subvol))) {
n = bch2_trans_kmalloc(trans, bkey_bytes(d.k));
err:
fsck_err:
printbuf_exit(&buf);
-
- if (ret && ret != -EINTR)
- bch_err(c, "error %i from check_target()", ret);
+ bch_err_fn(c, ret);
return ret;
}
static int check_dirent(struct btree_trans *trans, struct btree_iter *iter,
+ struct bkey_s_c k,
struct bch_hash_info *hash_info,
struct inode_walker *dir,
struct inode_walker *target,
struct snapshots_seen *s)
{
struct bch_fs *c = trans->c;
- struct bkey_s_c k;
struct bkey_s_c_dirent d;
struct inode_walker_entry *i;
struct printbuf buf = PRINTBUF;
+ struct bpos equiv;
int ret = 0;
-peek:
- k = bch2_btree_iter_peek(iter);
- if (!k.k)
- goto out;
-
- ret = bkey_err(k);
- if (ret)
- goto err;
ret = check_key_has_snapshot(trans, iter, k);
if (ret) {
goto out;
}
- ret = snapshots_seen_update(c, s, k.k->p);
+ equiv = k.k->p;
+ equiv.snapshot = bch2_snapshot_equiv(c, k.k->p.snapshot);
+
+ ret = snapshots_seen_update(c, s, iter->btree_id, k.k->p);
if (ret)
goto err;
if (k.k->type == KEY_TYPE_whiteout)
goto out;
- if (dir->cur_inum != k.k->p.inode) {
+ if (dir->last_pos.inode != k.k->p.inode) {
ret = check_subdir_count(trans, dir);
if (ret)
goto err;
}
- if (!iter->path->should_be_locked) {
- /* hack: see check_extent() */
- goto peek;
- }
+ BUG_ON(!iter->path->should_be_locked);
- ret = __walk_inode(trans, dir, k.k->p);
+ i = walk_inode(trans, dir, equiv, k.k->type == KEY_TYPE_whiteout);
+ ret = PTR_ERR_OR_ZERO(i);
if (ret < 0)
goto err;
- if (fsck_err_on(ret == INT_MAX, c,
+ if (dir->first_this_inode && dir->inodes.nr)
+ *hash_info = bch2_hash_info_init(c, &dir->inodes.data[0].inode);
+ dir->first_this_inode = false;
+
+ if (fsck_err_on(!i, c, dirent_in_missing_dir_inode,
"dirent in nonexisting directory:\n%s",
(printbuf_reset(&buf),
bch2_bkey_val_to_text(&buf, c, k), buf.buf))) {
goto out;
}
- if (ret == INT_MAX) {
- ret = 0;
+ if (!i)
goto out;
- }
-
- i = dir->inodes.data + ret;
- ret = 0;
- if (fsck_err_on(!S_ISDIR(i->inode.bi_mode), c,
+ if (fsck_err_on(!S_ISDIR(i->inode.bi_mode),
+ c, dirent_in_non_dir_inode,
"dirent in non directory inode type %s:\n%s",
bch2_d_type_str(inode_d_type(&i->inode)),
(printbuf_reset(&buf),
goto out;
}
- if (dir->first_this_inode)
- *hash_info = bch2_hash_info_init(c, &dir->inodes.data[0].inode);
-
- ret = hash_check_key(trans, bch2_dirent_hash_desc,
- hash_info, iter, k);
+ ret = hash_check_key(trans, bch2_dirent_hash_desc, hash_info, iter, k);
if (ret < 0)
goto err;
if (ret) {
ret = __subvol_lookup(trans, target_subvol,
&target_snapshot, &target_inum);
- if (ret && ret != -ENOENT)
+ if (ret && !bch2_err_matches(ret, ENOENT))
goto err;
- if (fsck_err_on(ret, c,
- "dirent points to missing subvolume %llu",
- le64_to_cpu(d.v->d_child_subvol))) {
+ if (fsck_err_on(ret, c, dirent_to_missing_subvol,
+ "dirent points to missing subvolume %u",
+ le32_to_cpu(d.v->d_child_subvol))) {
ret = __remove_dirent(trans, d.k->p);
goto err;
}
ret = __lookup_inode(trans, target_inum,
&subvol_root, &target_snapshot);
- if (ret && ret != -ENOENT)
+ if (ret && !bch2_err_matches(ret, ENOENT))
goto err;
- if (fsck_err_on(ret, c,
+ if (fsck_err_on(ret, c, subvol_to_missing_root,
"subvolume %u points to missing subvolume root %llu",
target_subvol,
target_inum)) {
goto err;
}
- if (fsck_err_on(subvol_root.bi_subvol != target_subvol, c,
+ if (fsck_err_on(subvol_root.bi_subvol != target_subvol,
+ c, subvol_root_wrong_bi_subvol,
"subvol root %llu has wrong bi_subvol field: got %u, should be %u",
target_inum,
subvol_root.bi_subvol, target_subvol)) {
if (ret)
goto err;
- if (fsck_err_on(!target->inodes.nr, c,
- "dirent points to missing inode:\n%s",
+ if (fsck_err_on(!target->inodes.nr,
+ c, dirent_to_missing_inode,
+ "dirent points to missing inode: (equiv %u)\n%s",
+ equiv.snapshot,
(printbuf_reset(&buf),
bch2_bkey_val_to_text(&buf, c, k),
buf.buf))) {
}
if (d.v->d_type == DT_DIR)
- for_each_visible_inode(c, s, dir, d.k->p.snapshot, i)
+ for_each_visible_inode(c, s, dir, equiv.snapshot, i)
i->count++;
out:
err:
fsck_err:
printbuf_exit(&buf);
-
- if (ret && ret != -EINTR)
- bch_err(c, "error %i from check_dirent()", ret);
+ bch_err_fn(c, ret);
return ret;
}
* Walk dirents: verify that they all have a corresponding S_ISDIR inode,
* validate d_type
*/
-noinline_for_stack
-static int check_dirents(struct bch_fs *c)
+int bch2_check_dirents(struct bch_fs *c)
{
struct inode_walker dir = inode_walker_init();
struct inode_walker target = inode_walker_init();
struct snapshots_seen s;
struct bch_hash_info hash_info;
- struct btree_trans trans;
+ struct btree_trans *trans = bch2_trans_get(c);
struct btree_iter iter;
+ struct bkey_s_c k;
int ret = 0;
- bch_verbose(c, "checking dirents");
-
snapshots_seen_init(&s);
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 0);
-
- bch2_trans_iter_init(&trans, &iter, BTREE_ID_dirents,
- POS(BCACHEFS_ROOT_INO, 0),
- BTREE_ITER_INTENT|
- BTREE_ITER_PREFETCH|
- BTREE_ITER_ALL_SNAPSHOTS);
- do {
- ret = __bch2_trans_do(&trans, NULL, NULL,
- BTREE_INSERT_LAZY_RW|
- BTREE_INSERT_NOFAIL,
- check_dirent(&trans, &iter, &hash_info,
- &dir, &target, &s));
- if (ret)
- break;
- } while (bch2_btree_iter_advance(&iter));
- bch2_trans_iter_exit(&trans, &iter);
+ ret = for_each_btree_key_commit(trans, iter, BTREE_ID_dirents,
+ POS(BCACHEFS_ROOT_INO, 0),
+ BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS,
+ k,
+ NULL, NULL,
+ BTREE_INSERT_LAZY_RW|BTREE_INSERT_NOFAIL,
+ check_dirent(trans, &iter, k, &hash_info, &dir, &target, &s));
- bch2_trans_exit(&trans);
+ bch2_trans_put(trans);
snapshots_seen_exit(&s);
inode_walker_exit(&dir);
inode_walker_exit(&target);
-
- if (ret)
- bch_err(c, "error %i from check_dirents()", ret);
+ bch_err_fn(c, ret);
return ret;
}
static int check_xattr(struct btree_trans *trans, struct btree_iter *iter,
+ struct bkey_s_c k,
struct bch_hash_info *hash_info,
struct inode_walker *inode)
{
struct bch_fs *c = trans->c;
- struct bkey_s_c k;
+ struct inode_walker_entry *i;
int ret;
- k = bch2_btree_iter_peek(iter);
- if (!k.k)
- return 0;
-
- ret = bkey_err(k);
+ ret = check_key_has_snapshot(trans, iter, k);
if (ret)
return ret;
- ret = check_key_has_snapshot(trans, iter, k);
+ i = walk_inode(trans, inode, k.k->p, k.k->type == KEY_TYPE_whiteout);
+ ret = PTR_ERR_OR_ZERO(i);
if (ret)
return ret;
- ret = __walk_inode(trans, inode, k.k->p);
- if (ret < 0)
- return ret;
+ if (inode->first_this_inode && inode->inodes.nr)
+ *hash_info = bch2_hash_info_init(c, &inode->inodes.data[0].inode);
+ inode->first_this_inode = false;
- if (fsck_err_on(ret == INT_MAX, c,
+ if (fsck_err_on(!i, c, xattr_in_missing_inode,
"xattr for missing inode %llu",
k.k->p.inode))
return bch2_btree_delete_at(trans, iter, 0);
- if (ret == INT_MAX)
+ if (!i)
return 0;
- ret = 0;
-
- if (inode->first_this_inode)
- *hash_info = bch2_hash_info_init(c, &inode->inodes.data[0].inode);
-
ret = hash_check_key(trans, bch2_xattr_hash_desc, hash_info, iter, k);
fsck_err:
- if (ret && ret != -EINTR)
- bch_err(c, "error %i from check_xattr()", ret);
+ bch_err_fn(c, ret);
return ret;
}
/*
* Walk xattrs: verify that they all have a corresponding inode
*/
-noinline_for_stack
-static int check_xattrs(struct bch_fs *c)
+int bch2_check_xattrs(struct bch_fs *c)
{
struct inode_walker inode = inode_walker_init();
struct bch_hash_info hash_info;
- struct btree_trans trans;
struct btree_iter iter;
+ struct bkey_s_c k;
int ret = 0;
- bch_verbose(c, "checking xattrs");
-
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 0);
-
- bch2_trans_iter_init(&trans, &iter, BTREE_ID_xattrs,
- POS(BCACHEFS_ROOT_INO, 0),
- BTREE_ITER_INTENT|
- BTREE_ITER_PREFETCH|
- BTREE_ITER_ALL_SNAPSHOTS);
-
- do {
- ret = __bch2_trans_do(&trans, NULL, NULL,
- BTREE_INSERT_LAZY_RW|
- BTREE_INSERT_NOFAIL,
- check_xattr(&trans, &iter, &hash_info,
- &inode));
- if (ret)
- break;
- } while (bch2_btree_iter_advance(&iter));
- bch2_trans_iter_exit(&trans, &iter);
-
- bch2_trans_exit(&trans);
-
- if (ret)
- bch_err(c, "error %i from check_xattrs()", ret);
+ ret = bch2_trans_run(c,
+ for_each_btree_key_commit(trans, iter, BTREE_ID_xattrs,
+ POS(BCACHEFS_ROOT_INO, 0),
+ BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS,
+ k,
+ NULL, NULL,
+ BTREE_INSERT_LAZY_RW|BTREE_INSERT_NOFAIL,
+ check_xattr(trans, &iter, k, &hash_info, &inode)));
+ bch_err_fn(c, ret);
return ret;
}
int ret;
ret = __subvol_lookup(trans, BCACHEFS_ROOT_SUBVOL, &snapshot, &inum);
- if (ret && ret != -ENOENT)
+ if (ret && !bch2_err_matches(ret, ENOENT))
return ret;
- if (mustfix_fsck_err_on(ret, c, "root subvol missing")) {
+ if (mustfix_fsck_err_on(ret, c, root_subvol_missing,
+ "root subvol missing")) {
struct bkey_i_subvolume root_subvol;
snapshot = U32_MAX;
root_subvol.v.flags = 0;
root_subvol.v.snapshot = cpu_to_le32(snapshot);
root_subvol.v.inode = cpu_to_le64(inum);
- ret = __bch2_trans_do(trans, NULL, NULL,
+ ret = commit_do(trans, NULL, NULL,
BTREE_INSERT_NOFAIL|
BTREE_INSERT_LAZY_RW,
- __bch2_btree_insert(trans, BTREE_ID_subvolumes, &root_subvol.k_i));
- if (ret) {
- bch_err(c, "error writing root subvol: %i", ret);
+ bch2_btree_insert_trans(trans, BTREE_ID_subvolumes,
+ &root_subvol.k_i, 0));
+ bch_err_msg(c, ret, "writing root subvol");
+ if (ret)
goto err;
- }
}
ret = __lookup_inode(trans, BCACHEFS_ROOT_INO, &root_inode, &snapshot);
- if (ret && ret != -ENOENT)
+ if (ret && !bch2_err_matches(ret, ENOENT))
return ret;
- if (mustfix_fsck_err_on(ret, c, "root directory missing") ||
- mustfix_fsck_err_on(!S_ISDIR(root_inode.bi_mode), c,
+ if (mustfix_fsck_err_on(ret, c, root_dir_missing,
+ "root directory missing") ||
+ mustfix_fsck_err_on(!S_ISDIR(root_inode.bi_mode),
+ c, root_inode_not_dir,
"root inode not a directory")) {
bch2_inode_init(c, &root_inode, 0, 0, S_IFDIR|0755,
0, NULL);
root_inode.bi_inum = inum;
ret = __write_inode(trans, &root_inode, snapshot);
- if (ret)
- bch_err(c, "error writing root inode: %i", ret);
+ bch_err_msg(c, ret, "writing root inode");
}
err:
fsck_err:
}
/* Get root directory, create if it doesn't exist: */
-noinline_for_stack
-static int check_root(struct bch_fs *c)
+int bch2_check_root(struct bch_fs *c)
{
- bch_verbose(c, "checking root directory");
+ int ret;
- return bch2_trans_do(c, NULL, NULL,
+ ret = bch2_trans_do(c, NULL, NULL,
BTREE_INSERT_NOFAIL|
BTREE_INSERT_LAZY_RW,
- check_root_trans(&trans));
+ check_root_trans(trans));
+ bch_err_fn(c, ret);
+ return ret;
}
struct pathbuf_entry {
struct bch_fs *c = trans->c;
int ret = 0;
- snapshot = snapshot_t(c, snapshot)->equiv;
+ snapshot = bch2_snapshot_equiv(c, snapshot);
p->nr = 0;
while (!(inode->bi_inum == BCACHEFS_ROOT_INO &&
PTR_ERR_OR_ZERO((d = dirent_get_by_pos(trans, &dirent_iter,
SPOS(inode->bi_dir, inode->bi_dir_offset,
parent_snapshot))).k));
- if (ret && ret != -ENOENT)
+ if (ret && !bch2_err_matches(ret, ENOENT))
break;
if (!ret && !dirent_points_to_inode(d, inode)) {
bch2_trans_iter_exit(trans, &dirent_iter);
- ret = -ENOENT;
+ ret = -BCH_ERR_ENOENT_dirent_doesnt_match_inode;
}
- if (ret == -ENOENT) {
- if (fsck_err(c, "unreachable inode %llu:%u, type %s nlink %u backptr %llu:%llu",
+ if (bch2_err_matches(ret, ENOENT)) {
+ if (fsck_err(c, inode_unreachable,
+ "unreachable inode %llu:%u, type %s nlink %u backptr %llu:%llu",
inode->bi_inum, snapshot,
bch2_d_type_str(inode_d_type(inode)),
inode->bi_nlink,
pr_err("%llu:%u", i->inum, i->snapshot);
pr_err("%llu:%u", inode->bi_inum, snapshot);
- if (!fsck_err(c, "directory structure loop"))
+ if (!fsck_err(c, dir_loop,
+ "directory structure loop"))
return 0;
- ret = __bch2_trans_do(trans, NULL, NULL,
+ ret = commit_do(trans, NULL, NULL,
BTREE_INSERT_NOFAIL|
BTREE_INSERT_LAZY_RW,
remove_backpointer(trans, inode));
}
}
fsck_err:
- if (ret)
- bch_err(c, "%s: err %i", __func__, ret);
+ bch_err_fn(c, ret);
return ret;
}
/*
* Check for unreachable inodes, as well as loops in the directory structure:
- * After check_dirents(), if an inode backpointer doesn't exist that means it's
+ * After bch2_check_dirents(), if an inode backpointer doesn't exist that means it's
* unreachable:
*/
-noinline_for_stack
-static int check_directory_structure(struct bch_fs *c)
+int bch2_check_directory_structure(struct bch_fs *c)
{
- struct btree_trans trans;
+ struct btree_trans *trans = bch2_trans_get(c);
struct btree_iter iter;
struct bkey_s_c k;
struct bch_inode_unpacked u;
pathbuf path = { 0, };
int ret;
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 0);
-
- for_each_btree_key(&trans, iter, BTREE_ID_inodes, POS_MIN,
+ for_each_btree_key(trans, iter, BTREE_ID_inodes, POS_MIN,
BTREE_ITER_INTENT|
BTREE_ITER_PREFETCH|
BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
break;
}
- if (u.bi_flags & BCH_INODE_UNLINKED)
+ if (u.bi_flags & BCH_INODE_unlinked)
continue;
- ret = check_path(&trans, &path, &u, iter.pos.snapshot);
+ ret = check_path(trans, &path, &u, iter.pos.snapshot);
if (ret)
break;
}
- bch2_trans_iter_exit(&trans, &iter);
-
- BUG_ON(ret == -EINTR);
-
+ bch2_trans_iter_exit(trans, &iter);
+ bch2_trans_put(trans);
darray_exit(&path);
-
- bch2_trans_exit(&trans);
+ bch_err_fn(c, ret);
return ret;
}
{
if (t->nr == t->size) {
size_t new_size = max_t(size_t, 128UL, t->size * 2);
- void *d = kvmalloc(new_size * sizeof(t->d[0]), GFP_KERNEL);
+ void *d = kvmalloc_array(new_size, sizeof(t->d[0]), GFP_KERNEL);
+
if (!d) {
bch_err(c, "fsck: error allocating memory for nlink_table, size %zu",
new_size);
- return -ENOMEM;
+ return -BCH_ERR_ENOMEM_fsck_add_nlink;
}
if (t->d)
struct nlink_table *t,
u64 start, u64 *end)
{
- struct btree_trans trans;
+ struct btree_trans *trans = bch2_trans_get(c);
struct btree_iter iter;
struct bkey_s_c k;
struct bch_inode_unpacked u;
int ret = 0;
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 0);
-
- for_each_btree_key(&trans, iter, BTREE_ID_inodes,
+ for_each_btree_key(trans, iter, BTREE_ID_inodes,
POS(0, start),
BTREE_ITER_INTENT|
BTREE_ITER_PREFETCH|
* Backpointer and directory structure checks are sufficient for
* directories, since they can't have hardlinks:
*/
- if (S_ISDIR(le16_to_cpu(u.bi_mode)))
+ if (S_ISDIR(u.bi_mode))
continue;
if (!u.bi_nlink)
}
}
- bch2_trans_iter_exit(&trans, &iter);
- bch2_trans_exit(&trans);
+ bch2_trans_iter_exit(trans, &iter);
+ bch2_trans_put(trans);
if (ret)
bch_err(c, "error in fsck: btree error %i while walking inodes", ret);
static int check_nlinks_walk_dirents(struct bch_fs *c, struct nlink_table *links,
u64 range_start, u64 range_end)
{
- struct btree_trans trans;
+ struct btree_trans *trans = bch2_trans_get(c);
struct snapshots_seen s;
struct btree_iter iter;
struct bkey_s_c k;
snapshots_seen_init(&s);
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 0);
-
- for_each_btree_key(&trans, iter, BTREE_ID_dirents, POS_MIN,
+ for_each_btree_key(trans, iter, BTREE_ID_dirents, POS_MIN,
BTREE_ITER_INTENT|
BTREE_ITER_PREFETCH|
BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
- ret = snapshots_seen_update(c, &s, k.k->p);
+ ret = snapshots_seen_update(c, &s, iter.btree_id, k.k->p);
if (ret)
break;
d.v->d_type != DT_SUBVOL)
inc_link(c, &s, links, range_start, range_end,
le64_to_cpu(d.v->d_inum),
- d.k->p.snapshot);
+ bch2_snapshot_equiv(c, d.k->p.snapshot));
break;
}
}
- bch2_trans_iter_exit(&trans, &iter);
+ bch2_trans_iter_exit(trans, &iter);
if (ret)
bch_err(c, "error in fsck: btree error %i while walking dirents", ret);
- bch2_trans_exit(&trans);
+ bch2_trans_put(trans);
snapshots_seen_exit(&s);
return ret;
}
-noinline_for_stack
-static int check_nlinks_update_hardlinks(struct bch_fs *c,
- struct nlink_table *links,
- u64 range_start, u64 range_end)
+static int check_nlinks_update_inode(struct btree_trans *trans, struct btree_iter *iter,
+ struct bkey_s_c k,
+ struct nlink_table *links,
+ size_t *idx, u64 range_end)
{
- struct btree_trans trans;
- struct btree_iter iter;
- struct bkey_s_c k;
+ struct bch_fs *c = trans->c;
struct bch_inode_unpacked u;
- struct nlink *link = links->d;
+ struct nlink *link = &links->d[*idx];
int ret = 0;
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 0);
-
- for_each_btree_key(&trans, iter, BTREE_ID_inodes,
- POS(0, range_start),
- BTREE_ITER_INTENT|
- BTREE_ITER_PREFETCH|
- BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
- if (k.k->p.offset >= range_end)
- break;
-
- if (!bkey_is_inode(k.k))
- continue;
+ if (k.k->p.offset >= range_end)
+ return 1;
- BUG_ON(bch2_inode_unpack(k, &u));
+ if (!bkey_is_inode(k.k))
+ return 0;
- if (S_ISDIR(le16_to_cpu(u.bi_mode)))
- continue;
+ BUG_ON(bch2_inode_unpack(k, &u));
- if (!u.bi_nlink)
- continue;
+ if (S_ISDIR(u.bi_mode))
+ return 0;
- while ((cmp_int(link->inum, k.k->p.offset) ?:
- cmp_int(link->snapshot, k.k->p.snapshot)) < 0) {
- link++;
- BUG_ON(link >= links->d + links->nr);
- }
+ if (!u.bi_nlink)
+ return 0;
- if (fsck_err_on(bch2_inode_nlink_get(&u) != link->count, c,
- "inode %llu type %s has wrong i_nlink (%u, should be %u)",
- u.bi_inum, bch2_d_types[mode_to_type(u.bi_mode)],
- bch2_inode_nlink_get(&u), link->count)) {
- bch2_inode_nlink_set(&u, link->count);
+ while ((cmp_int(link->inum, k.k->p.offset) ?:
+ cmp_int(link->snapshot, k.k->p.snapshot)) < 0) {
+ BUG_ON(*idx == links->nr);
+ link = &links->d[++*idx];
+ }
- ret = write_inode(&trans, &u, k.k->p.snapshot);
- if (ret)
- bch_err(c, "error in fsck: error %i updating inode", ret);
- }
+ if (fsck_err_on(bch2_inode_nlink_get(&u) != link->count,
+ c, inode_wrong_nlink,
+ "inode %llu type %s has wrong i_nlink (%u, should be %u)",
+ u.bi_inum, bch2_d_types[mode_to_type(u.bi_mode)],
+ bch2_inode_nlink_get(&u), link->count)) {
+ bch2_inode_nlink_set(&u, link->count);
+ ret = __write_inode(trans, &u, k.k->p.snapshot);
}
fsck_err:
- bch2_trans_iter_exit(&trans, &iter);
- bch2_trans_exit(&trans);
+ return ret;
+}
- if (ret)
+noinline_for_stack
+static int check_nlinks_update_hardlinks(struct bch_fs *c,
+ struct nlink_table *links,
+ u64 range_start, u64 range_end)
+{
+ struct btree_iter iter;
+ struct bkey_s_c k;
+ size_t idx = 0;
+ int ret = 0;
+
+ ret = bch2_trans_run(c,
+ for_each_btree_key_commit(trans, iter, BTREE_ID_inodes,
+ POS(0, range_start),
+ BTREE_ITER_INTENT|BTREE_ITER_PREFETCH|BTREE_ITER_ALL_SNAPSHOTS, k,
+ NULL, NULL, BTREE_INSERT_LAZY_RW|BTREE_INSERT_NOFAIL,
+ check_nlinks_update_inode(trans, &iter, k, links, &idx, range_end)));
+ if (ret < 0) {
bch_err(c, "error in fsck: btree error %i while walking inodes", ret);
+ return ret;
+ }
- return ret;
+ return 0;
}
-noinline_for_stack
-static int check_nlinks(struct bch_fs *c)
+int bch2_check_nlinks(struct bch_fs *c)
{
struct nlink_table links = { 0 };
u64 this_iter_range_start, next_iter_range_start = 0;
int ret = 0;
- bch_verbose(c, "checking inode nlinks");
-
do {
this_iter_range_start = next_iter_range_start;
next_iter_range_start = U64_MAX;
} while (next_iter_range_start != U64_MAX);
kvfree(links.d);
-
+ bch_err_fn(c, ret);
return ret;
}
-static int fix_reflink_p_key(struct btree_trans *trans, struct btree_iter *iter)
+static int fix_reflink_p_key(struct btree_trans *trans, struct btree_iter *iter,
+ struct bkey_s_c k)
{
- struct bkey_s_c k;
struct bkey_s_c_reflink_p p;
struct bkey_i_reflink_p *u;
int ret;
- k = bch2_btree_iter_peek(iter);
- if (!k.k)
- return 0;
-
- ret = bkey_err(k);
- if (ret)
- return ret;
-
if (k.k->type != KEY_TYPE_reflink_p)
return 0;
return bch2_trans_update(trans, iter, &u->k_i, BTREE_TRIGGER_NORUN);
}
-noinline_for_stack
-static int fix_reflink_p(struct bch_fs *c)
+int bch2_fix_reflink_p(struct bch_fs *c)
{
- struct btree_trans trans;
struct btree_iter iter;
struct bkey_s_c k;
int ret;
if (c->sb.version >= bcachefs_metadata_version_reflink_p_fix)
return 0;
- bch_verbose(c, "fixing reflink_p keys");
-
- bch2_trans_init(&trans, c, BTREE_ITER_MAX, 0);
-
- for_each_btree_key(&trans, iter, BTREE_ID_extents, POS_MIN,
- BTREE_ITER_INTENT|
- BTREE_ITER_PREFETCH|
- BTREE_ITER_ALL_SNAPSHOTS, k, ret) {
- if (k.k->type == KEY_TYPE_reflink_p) {
- ret = __bch2_trans_do(&trans, NULL, NULL,
- BTREE_INSERT_NOFAIL|
- BTREE_INSERT_LAZY_RW,
- fix_reflink_p_key(&trans, &iter));
- if (ret)
- break;
- }
- }
- bch2_trans_iter_exit(&trans, &iter);
-
- bch2_trans_exit(&trans);
+ ret = bch2_trans_run(c,
+ for_each_btree_key_commit(trans, iter,
+ BTREE_ID_extents, POS_MIN,
+ BTREE_ITER_INTENT|BTREE_ITER_PREFETCH|
+ BTREE_ITER_ALL_SNAPSHOTS, k,
+ NULL, NULL, BTREE_INSERT_NOFAIL|BTREE_INSERT_LAZY_RW,
+ fix_reflink_p_key(trans, &iter, k)));
+ bch_err_fn(c, ret);
return ret;
}
-
-/*
- * Checks for inconsistencies that shouldn't happen, unless we have a bug.
- * Doesn't fix them yet, mainly because they haven't yet been observed:
- */
-int bch2_fsck_full(struct bch_fs *c)
-{
- return bch2_fs_snapshots_check(c) ?:
- check_inodes(c, true) ?:
- check_subvols(c) ?:
- check_extents(c) ?:
- check_dirents(c) ?:
- check_xattrs(c) ?:
- check_root(c) ?:
- check_directory_structure(c) ?:
- check_nlinks(c) ?:
- fix_reflink_p(c);
-}
-
-int bch2_fsck_walk_inodes_only(struct bch_fs *c)
-{
- return check_inodes(c, false);
-}