// SPDX-License-Identifier: GPL-2.0
#include "bcachefs.h"
+#include "backpointers.h"
#include "bkey_buf.h"
#include "alloc_background.h"
#include "btree_gc.h"
#include "buckets.h"
#include "dirent.h"
#include "ec.h"
+#include "errcode.h"
#include "error.h"
#include "fs-common.h"
#include "fsck.h"
#include "journal_io.h"
#include "journal_reclaim.h"
#include "journal_seq_blacklist.h"
+#include "lru.h"
#include "move.h"
#include "quota.h"
#include "recovery.h"
return __journal_key_cmp(l->btree_id, l->level, l->k->k.p, r);
}
-size_t bch2_journal_key_search(struct journal_keys *journal_keys,
- enum btree_id id, unsigned level,
- struct bpos pos)
+static inline size_t idx_to_pos(struct journal_keys *keys, size_t idx)
{
- size_t l = 0, r = journal_keys->nr, m;
+ size_t gap_size = keys->size - keys->nr;
+
+ if (idx >= keys->gap)
+ idx += gap_size;
+ return idx;
+}
+
+static inline struct journal_key *idx_to_key(struct journal_keys *keys, size_t idx)
+{
+ return keys->d + idx_to_pos(keys, idx);
+}
+
+static size_t __bch2_journal_key_search(struct journal_keys *keys,
+ enum btree_id id, unsigned level,
+ struct bpos pos)
+{
+ size_t l = 0, r = keys->nr, m;
while (l < r) {
m = l + ((r - l) >> 1);
- if (__journal_key_cmp(id, level, pos, &journal_keys->d[m]) > 0)
+ if (__journal_key_cmp(id, level, pos, idx_to_key(keys, m)) > 0)
l = m + 1;
else
r = m;
}
- BUG_ON(l < journal_keys->nr &&
- __journal_key_cmp(id, level, pos, &journal_keys->d[l]) > 0);
+ BUG_ON(l < keys->nr &&
+ __journal_key_cmp(id, level, pos, idx_to_key(keys, l)) > 0);
BUG_ON(l &&
- __journal_key_cmp(id, level, pos, &journal_keys->d[l - 1]) <= 0);
+ __journal_key_cmp(id, level, pos, idx_to_key(keys, l - 1)) <= 0);
return l;
}
-static void journal_iter_fix(struct bch_fs *c, struct journal_iter *iter, unsigned idx)
+static size_t bch2_journal_key_search(struct journal_keys *keys,
+ enum btree_id id, unsigned level,
+ struct bpos pos)
{
- struct bkey_i *n = iter->keys->d[idx].k;
- struct btree_and_journal_iter *biter =
- container_of(iter, struct btree_and_journal_iter, journal);
-
- if (iter->idx > idx ||
- (iter->idx == idx &&
- biter->last &&
- bpos_cmp(n->k.p, biter->unpacked.p) <= 0))
- iter->idx++;
+ return idx_to_pos(keys, __bch2_journal_key_search(keys, id, level, pos));
+}
+
+struct bkey_i *bch2_journal_keys_peek_upto(struct bch_fs *c, enum btree_id btree_id,
+ unsigned level, struct bpos pos,
+ struct bpos end_pos, size_t *idx)
+{
+ struct journal_keys *keys = &c->journal_keys;
+ unsigned iters = 0;
+ struct journal_key *k;
+search:
+ if (!*idx)
+ *idx = __bch2_journal_key_search(keys, btree_id, level, pos);
+
+ while (*idx < keys->nr &&
+ (k = idx_to_key(keys, *idx),
+ k->btree_id == btree_id &&
+ k->level == level &&
+ bpos_cmp(k->k->k.p, end_pos) <= 0)) {
+ if (bpos_cmp(k->k->k.p, pos) >= 0 &&
+ !k->overwritten)
+ return k->k;
+
+ (*idx)++;
+ iters++;
+ if (iters == 10) {
+ *idx = 0;
+ goto search;
+ }
+ }
+
+ return NULL;
+}
+
+struct bkey_i *bch2_journal_keys_peek_slot(struct bch_fs *c, enum btree_id btree_id,
+ unsigned level, struct bpos pos)
+{
+ size_t idx = 0;
+
+ return bch2_journal_keys_peek_upto(c, btree_id, level, pos, pos, &idx);
+}
+
+static void journal_iters_fix(struct bch_fs *c)
+{
+ struct journal_keys *keys = &c->journal_keys;
+ /* The key we just inserted is immediately before the gap: */
+ size_t gap_end = keys->gap + (keys->size - keys->nr);
+ struct btree_and_journal_iter *iter;
+
+ /*
+ * If an iterator points one after the key we just inserted, decrement
+ * the iterator so it points at the key we just inserted - if the
+ * decrement was unnecessary, bch2_btree_and_journal_iter_peek() will
+ * handle that:
+ */
+ list_for_each_entry(iter, &c->journal_iters, journal.list)
+ if (iter->journal.idx == gap_end)
+ iter->journal.idx = keys->gap - 1;
+}
+
+static void journal_iters_move_gap(struct bch_fs *c, size_t old_gap, size_t new_gap)
+{
+ struct journal_keys *keys = &c->journal_keys;
+ struct journal_iter *iter;
+ size_t gap_size = keys->size - keys->nr;
+
+ list_for_each_entry(iter, &c->journal_iters, list) {
+ if (iter->idx > old_gap)
+ iter->idx -= gap_size;
+ if (iter->idx >= new_gap)
+ iter->idx += gap_size;
+ }
}
int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id,
.journal_seq = U32_MAX,
};
struct journal_keys *keys = &c->journal_keys;
- struct journal_iter *iter;
size_t idx = bch2_journal_key_search(keys, id, level, k->k.p);
BUG_ON(test_bit(BCH_FS_RW, &c->flags));
- if (idx < keys->nr &&
+ if (idx < keys->size &&
journal_key_cmp(&n, &keys->d[idx]) == 0) {
if (keys->d[idx].allocated)
kfree(keys->d[idx].k);
return 0;
}
+ if (idx > keys->gap)
+ idx -= keys->size - keys->nr;
+
if (keys->nr == keys->size) {
struct journal_keys new_keys = {
.nr = keys->nr,
- .size = keys->size * 2,
- .journal_seq_base = keys->journal_seq_base,
+ .size = max_t(size_t, keys->size, 8) * 2,
};
- new_keys.d = kvmalloc(sizeof(new_keys.d[0]) * new_keys.size, GFP_KERNEL);
+ new_keys.d = kvmalloc_array(new_keys.size, sizeof(new_keys.d[0]), GFP_KERNEL);
if (!new_keys.d) {
bch_err(c, "%s: error allocating new key array (size %zu)",
__func__, new_keys.size);
return -ENOMEM;
}
+ /* Since @keys was full, there was no gap: */
memcpy(new_keys.d, keys->d, sizeof(keys->d[0]) * keys->nr);
kvfree(keys->d);
*keys = new_keys;
+
+ /* And now the gap is at the end: */
+ keys->gap = keys->nr;
}
- array_insert_item(keys->d, keys->nr, idx, n);
+ journal_iters_move_gap(c, keys->gap, idx);
- list_for_each_entry(iter, &c->journal_iters, list)
- journal_iter_fix(c, iter, idx);
+ move_gap(keys->d, keys->nr, keys->size, keys->gap, idx);
+ keys->gap = idx;
+
+ keys->nr++;
+ keys->d[keys->gap++] = n;
+
+ journal_iters_fix(c);
return 0;
}
struct journal_keys *keys = &c->journal_keys;
size_t idx = bch2_journal_key_search(keys, btree, level, pos);
- if (idx < keys->nr &&
+ if (idx < keys->size &&
keys->d[idx].btree_id == btree &&
keys->d[idx].level == level &&
!bpos_cmp(keys->d[idx].k->k.p, pos))
keys->d[idx].overwritten = true;
}
-static struct bkey_i *bch2_journal_iter_peek(struct journal_iter *iter)
+static void bch2_journal_iter_advance(struct journal_iter *iter)
+{
+ if (iter->idx < iter->keys->size) {
+ iter->idx++;
+ if (iter->idx == iter->keys->gap)
+ iter->idx += iter->keys->size - iter->keys->nr;
+ }
+}
+
+struct bkey_s_c bch2_journal_iter_peek(struct journal_iter *iter)
{
struct journal_key *k = iter->keys->d + iter->idx;
- while (k < iter->keys->d + iter->keys->nr &&
+ while (k < iter->keys->d + iter->keys->size &&
k->btree_id == iter->btree_id &&
k->level == iter->level) {
if (!k->overwritten)
- return k->k;
+ return bkey_i_to_s_c(k->k);
- iter->idx++;
+ bch2_journal_iter_advance(iter);
k = iter->keys->d + iter->idx;
}
- return NULL;
-}
-
-static void bch2_journal_iter_advance(struct journal_iter *iter)
-{
- if (iter->idx < iter->keys->nr)
- iter->idx++;
+ return bkey_s_c_null;
}
static void bch2_journal_iter_exit(struct journal_iter *iter)
void bch2_btree_and_journal_iter_advance(struct btree_and_journal_iter *iter)
{
- switch (iter->last) {
- case none:
- break;
- case btree:
- bch2_journal_iter_advance_btree(iter);
- break;
- case journal:
- bch2_journal_iter_advance(&iter->journal);
- break;
- }
-
- iter->last = none;
+ if (!bpos_cmp(iter->pos, SPOS_MAX))
+ iter->at_end = true;
+ else
+ iter->pos = bpos_successor(iter->pos);
}
struct bkey_s_c bch2_btree_and_journal_iter_peek(struct btree_and_journal_iter *iter)
{
- struct bkey_s_c ret;
-
- while (1) {
- struct bkey_s_c btree_k =
- bch2_journal_iter_peek_btree(iter);
- struct bkey_s_c journal_k =
- bkey_i_to_s_c(bch2_journal_iter_peek(&iter->journal));
+ struct bkey_s_c btree_k, journal_k, ret;
+again:
+ if (iter->at_end)
+ return bkey_s_c_null;
- if (btree_k.k && journal_k.k) {
- int cmp = bpos_cmp(btree_k.k->p, journal_k.k->p);
+ while ((btree_k = bch2_journal_iter_peek_btree(iter)).k &&
+ bpos_cmp(btree_k.k->p, iter->pos) < 0)
+ bch2_journal_iter_advance_btree(iter);
- if (!cmp)
- bch2_journal_iter_advance_btree(iter);
+ while ((journal_k = bch2_journal_iter_peek(&iter->journal)).k &&
+ bpos_cmp(journal_k.k->p, iter->pos) < 0)
+ bch2_journal_iter_advance(&iter->journal);
- iter->last = cmp < 0 ? btree : journal;
- } else if (btree_k.k) {
- iter->last = btree;
- } else if (journal_k.k) {
- iter->last = journal;
- } else {
- iter->last = none;
- return bkey_s_c_null;
- }
+ ret = journal_k.k &&
+ (!btree_k.k || bpos_cmp(journal_k.k->p, btree_k.k->p) <= 0)
+ ? journal_k
+ : btree_k;
- ret = iter->last == journal ? journal_k : btree_k;
+ if (ret.k && iter->b && bpos_cmp(ret.k->p, iter->b->data->max_key) > 0)
+ ret = bkey_s_c_null;
- if (iter->b &&
- bpos_cmp(ret.k->p, iter->b->data->max_key) > 0) {
- iter->journal.idx = iter->journal.keys->nr;
- iter->last = none;
- return bkey_s_c_null;
+ if (ret.k) {
+ iter->pos = ret.k->p;
+ if (bkey_deleted(ret.k)) {
+ bch2_btree_and_journal_iter_advance(iter);
+ goto again;
}
-
- if (!bkey_deleted(ret.k))
- break;
-
- bch2_btree_and_journal_iter_advance(iter);
+ } else {
+ iter->pos = SPOS_MAX;
+ iter->at_end = true;
}
return ret;
}
-struct bkey_s_c bch2_btree_and_journal_iter_next(struct btree_and_journal_iter *iter)
-{
- bch2_btree_and_journal_iter_advance(iter);
-
- return bch2_btree_and_journal_iter_peek(iter);
-}
-
void bch2_btree_and_journal_iter_exit(struct btree_and_journal_iter *iter)
{
bch2_journal_iter_exit(&iter->journal);
iter->node_iter = node_iter;
bch2_journal_iter_init(c, &iter->journal, b->c.btree_id, b->c.level, pos);
INIT_LIST_HEAD(&iter->journal.list);
+ iter->pos = b->data->min_key;
+ iter->at_end = false;
}
/*
/* sort and dedup all keys in the journal: */
-void bch2_journal_entries_free(struct list_head *list)
+void bch2_journal_entries_free(struct bch_fs *c)
{
-
- while (!list_empty(list)) {
- struct journal_replay *i =
- list_first_entry(list, struct journal_replay, list);
- list_del(&i->list);
- kvpfree(i, offsetof(struct journal_replay, j) +
- vstruct_bytes(&i->j));
- }
+ struct journal_replay **i;
+ struct genradix_iter iter;
+
+ genradix_for_each(&c->journal_entries, iter, i)
+ if (*i)
+ kvpfree(*i, offsetof(struct journal_replay, j) +
+ vstruct_bytes(&(*i)->j));
+ genradix_free(&c->journal_entries);
}
/*
{
struct journal_key *i;
+ move_gap(keys->d, keys->nr, keys->size, keys->gap, keys->nr);
+ keys->gap = keys->nr;
+
for (i = keys->d; i < keys->d + keys->nr; i++)
if (i->allocated)
kfree(i->k);
kvfree(keys->d);
keys->d = NULL;
- keys->nr = 0;
+ keys->nr = keys->gap = keys->size = 0;
}
-static struct journal_keys journal_keys_sort(struct list_head *journal_entries)
+static int journal_keys_sort(struct bch_fs *c)
{
- struct journal_replay *i;
+ struct genradix_iter iter;
+ struct journal_replay *i, **_i;
struct jset_entry *entry;
struct bkey_i *k, *_n;
- struct journal_keys keys = { NULL };
+ struct journal_keys *keys = &c->journal_keys;
struct journal_key *src, *dst;
size_t nr_keys = 0;
- if (list_empty(journal_entries))
- return keys;
+ genradix_for_each(&c->journal_entries, iter, _i) {
+ i = *_i;
- list_for_each_entry(i, journal_entries, list) {
- if (i->ignore)
+ if (!i || i->ignore)
continue;
- if (!keys.journal_seq_base)
- keys.journal_seq_base = le64_to_cpu(i->j.seq);
-
for_each_jset_key(k, _n, entry, &i->j)
nr_keys++;
}
- keys.size = roundup_pow_of_two(nr_keys);
+ if (!nr_keys)
+ return 0;
- keys.d = kvmalloc(sizeof(keys.d[0]) * keys.size, GFP_KERNEL);
- if (!keys.d)
- goto err;
+ keys->size = roundup_pow_of_two(nr_keys);
- list_for_each_entry(i, journal_entries, list) {
- if (i->ignore)
- continue;
+ keys->d = kvmalloc_array(keys->size, sizeof(keys->d[0]), GFP_KERNEL);
+ if (!keys->d)
+ return -ENOMEM;
- BUG_ON(le64_to_cpu(i->j.seq) - keys.journal_seq_base > U32_MAX);
+ genradix_for_each(&c->journal_entries, iter, _i) {
+ i = *_i;
+
+ if (!i || i->ignore)
+ continue;
for_each_jset_key(k, _n, entry, &i->j)
- keys.d[keys.nr++] = (struct journal_key) {
+ keys->d[keys->nr++] = (struct journal_key) {
.btree_id = entry->btree_id,
.level = entry->level,
.k = k,
- .journal_seq = le64_to_cpu(i->j.seq) -
- keys.journal_seq_base,
+ .journal_seq = le64_to_cpu(i->j.seq),
.journal_offset = k->_data - i->j._data,
};
}
- sort(keys.d, keys.nr, sizeof(keys.d[0]), journal_sort_key_cmp, NULL);
+ sort(keys->d, keys->nr, sizeof(keys->d[0]), journal_sort_key_cmp, NULL);
- src = dst = keys.d;
- while (src < keys.d + keys.nr) {
- while (src + 1 < keys.d + keys.nr &&
+ src = dst = keys->d;
+ while (src < keys->d + keys->nr) {
+ while (src + 1 < keys->d + keys->nr &&
src[0].btree_id == src[1].btree_id &&
src[0].level == src[1].level &&
!bpos_cmp(src[0].k->k.p, src[1].k->k.p))
*dst++ = *src++;
}
- keys.nr = dst - keys.d;
-err:
- return keys;
+ keys->nr = dst - keys->d;
+ keys->gap = keys->nr;
+ return 0;
}
/* journal replay: */
static void replay_now_at(struct journal *j, u64 seq)
{
BUG_ON(seq < j->replay_journal_seq);
- BUG_ON(seq > j->replay_journal_seq_end);
+
+ seq = min(seq, j->replay_journal_seq_end);
while (j->replay_journal_seq < seq)
bch2_journal_pin_put(j, j->replay_journal_seq++);
size_t i;
int ret;
+ move_gap(keys->d, keys->nr, keys->size, keys->gap, keys->nr);
+ keys->gap = keys->nr;
+
keys_sorted = kvmalloc_array(sizeof(*keys_sorted), keys->nr, GFP_KERNEL);
if (!keys_sorted)
return -ENOMEM;
sizeof(keys_sorted[0]),
journal_sort_seq_cmp, NULL);
- if (keys->nr)
- replay_now_at(j, keys->journal_seq_base);
-
for (i = 0; i < keys->nr; i++) {
k = keys_sorted[i];
cond_resched();
- if (!k->allocated)
- replay_now_at(j, keys->journal_seq_base + k->journal_seq);
+ replay_now_at(j, k->journal_seq);
ret = bch2_trans_do(c, NULL, NULL,
BTREE_INSERT_LAZY_RW|
BTREE_INSERT_NOFAIL|
- BTREE_INSERT_JOURNAL_RESERVED|
- (!k->allocated ? BTREE_INSERT_JOURNAL_REPLAY : 0),
+ (!k->allocated
+ ? BTREE_INSERT_JOURNAL_REPLAY|JOURNAL_WATERMARK_reserved
+ : 0),
bch2_journal_replay_key(&trans, k));
if (ret) {
bch_err(c, "journal replay: error %d while replaying key at btree %s level %u",
bch2_journal_set_replay_done(j);
bch2_journal_flush_all_pins(j);
ret = bch2_journal_error(j);
+
+ if (keys->nr && !ret)
+ bch2_journal_log_msg(&c->journal, "journal replay finished");
err:
kvfree(keys_sorted);
return ret;
unsigned i, nr_types = jset_entry_dev_usage_nr_types(u);
ca->usage_base->buckets_ec = le64_to_cpu(u->buckets_ec);
- ca->usage_base->buckets_unavailable = le64_to_cpu(u->buckets_unavailable);
for (i = 0; i < min_t(unsigned, nr_types, BCH_DATA_NR); i++) {
ca->usage_base->d[i].buckets = le64_to_cpu(u->d[i].buckets);
}
static int journal_replay_early(struct bch_fs *c,
- struct bch_sb_field_clean *clean,
- struct list_head *journal)
+ struct bch_sb_field_clean *clean)
{
- struct journal_replay *i;
struct jset_entry *entry;
int ret;
return ret;
}
} else {
- list_for_each_entry(i, journal, list) {
- if (i->ignore)
+ struct genradix_iter iter;
+ struct journal_replay *i, **_i;
+
+ genradix_for_each(&c->journal_entries, iter, _i) {
+ i = *_i;
+
+ if (!i || i->ignore)
continue;
vstruct_for_each(&i->j, entry) {
{
unsigned i;
struct bch_sb_field_clean *clean = *cleanp;
+ struct printbuf buf1 = PRINTBUF;
+ struct printbuf buf2 = PRINTBUF;
int ret = 0;
if (mustfix_fsck_err_on(j->seq != clean->journal_seq, c,
}
for (i = 0; i < BTREE_ID_NR; i++) {
- char buf1[200], buf2[200];
struct bkey_i *k1, *k2;
unsigned l1 = 0, l2 = 0;
if (!k1 && !k2)
continue;
+ printbuf_reset(&buf1);
+ printbuf_reset(&buf2);
+
+ if (k1)
+ bch2_bkey_val_to_text(&buf1, c, bkey_i_to_s_c(k1));
+ else
+ prt_printf(&buf1, "(none)");
+
+ if (k2)
+ bch2_bkey_val_to_text(&buf2, c, bkey_i_to_s_c(k2));
+ else
+ prt_printf(&buf2, "(none)");
+
mustfix_fsck_err_on(!k1 || !k2 ||
IS_ERR(k1) ||
IS_ERR(k2) ||
"superblock btree root %u doesn't match journal after clean shutdown\n"
"sb: l=%u %s\n"
"journal: l=%u %s\n", i,
- l1, (bch2_bkey_val_to_text(&PBUF(buf1), c, bkey_i_to_s_c(k1)), buf1),
- l2, (bch2_bkey_val_to_text(&PBUF(buf2), c, bkey_i_to_s_c(k2)), buf2));
+ l1, buf1.buf,
+ l2, buf2.buf);
}
fsck_err:
+ printbuf_exit(&buf2);
+ printbuf_exit(&buf1);
return ret;
}
return ERR_PTR(-ENOMEM);
}
- ret = bch2_sb_clean_validate(c, clean, READ);
+ ret = bch2_sb_clean_validate_late(c, clean, READ);
if (ret) {
mutex_unlock(&c->sb_lock);
return ERR_PTR(ret);
return ERR_PTR(ret);
}
+static bool btree_id_is_alloc(enum btree_id id)
+{
+ switch (id) {
+ case BTREE_ID_alloc:
+ case BTREE_ID_backpointers:
+ case BTREE_ID_need_discard:
+ case BTREE_ID_freespace:
+ return true;
+ default:
+ return false;
+ }
+}
+
static int read_btree_roots(struct bch_fs *c)
{
unsigned i;
if (!r->alive)
continue;
- if (i == BTREE_ID_alloc &&
+ if (btree_id_is_alloc(i) &&
c->opts.reconstruct_alloc) {
c->sb.compat &= ~(1ULL << BCH_COMPAT_alloc_info);
continue;
}
if (r->error) {
- __fsck_err(c, i == BTREE_ID_alloc
+ __fsck_err(c, btree_id_is_alloc(i)
? FSCK_CAN_IGNORE : 0,
"invalid btree root %s",
bch2_btree_ids[i]);
ret = bch2_btree_root_read(c, i, &r->key, r->level);
if (ret) {
- __fsck_err(c, i == BTREE_ID_alloc
+ __fsck_err(c,
+ btree_id_is_alloc(i)
? FSCK_CAN_IGNORE : 0,
"error reading btree root %s",
bch2_btree_ids[i]);
if (ret)
return ret;
-
bkey_subvolume_init(&root_volume.k_i);
root_volume.k.p.offset = BCACHEFS_ROOT_SUBVOL;
root_volume.v.flags = 0;
c->opts.fix_errors = FSCK_OPT_YES;
}
- if (!c->replicas.entries ||
- c->opts.rebuild_replicas) {
- bch_info(c, "building replicas info");
- set_bit(BCH_FS_REBUILD_REPLICAS, &c->flags);
- }
-
if (!c->opts.nochanges) {
- if (c->sb.version < bcachefs_metadata_version_inode_backpointers) {
- bch_info(c, "version prior to inode backpointers, upgrade and fsck required");
+ if (c->sb.version < bcachefs_metadata_version_backpointers) {
+ bch_info(c, "version prior to backpointers, upgrade and fsck required");
c->opts.version_upgrade = true;
c->opts.fsck = true;
c->opts.fix_errors = FSCK_OPT_YES;
- } else if (c->sb.version < bcachefs_metadata_version_subvol_dirent) {
- bch_info(c, "filesystem version is prior to subvol_dirent - upgrading");
- c->opts.version_upgrade = true;
- c->opts.fsck = true;
- } else if (c->sb.version < bcachefs_metadata_version_inode_v2) {
- bch_info(c, "filesystem version is prior to inode_v2 - upgrading");
- c->opts.version_upgrade = true;
+ } else if (c->sb.version < bcachefs_metadata_version_inode_v3) {
+ bch_info(c, "version prior to inode_v3, upgrade required");
+ c->opts.version_upgrade = true;
}
}
+ if (c->opts.fsck && c->opts.norecovery) {
+ bch_err(c, "cannot select both norecovery and fsck");
+ ret = -EINVAL;
+ goto err;
+ }
+
ret = bch2_blacklist_table_initialize(c);
if (ret) {
bch_err(c, "error initializing blacklist table");
}
if (!c->sb.clean || c->opts.fsck || c->opts.keep_journal) {
- struct journal_replay *i;
+ struct genradix_iter iter;
+ struct journal_replay **i;
bch_verbose(c, "starting journal read");
- ret = bch2_journal_read(c, &c->journal_entries,
- &blacklist_seq, &journal_seq);
+ ret = bch2_journal_read(c, &blacklist_seq, &journal_seq);
if (ret)
goto err;
- list_for_each_entry_reverse(i, &c->journal_entries, list)
- if (!i->ignore) {
- last_journal_entry = &i->j;
+ genradix_for_each_reverse(&c->journal_entries, iter, i)
+ if (*i && !(*i)->ignore) {
+ last_journal_entry = &(*i)->j;
break;
}
goto use_clean;
}
- c->journal_keys = journal_keys_sort(&c->journal_entries);
- if (!c->journal_keys.d) {
- ret = -ENOMEM;
+ ret = journal_keys_sort(c);
+ if (ret)
goto err;
- }
if (c->sb.clean && last_journal_entry) {
ret = verify_superblock_clean(c, &clean,
use_clean:
if (!clean) {
bch_err(c, "no superblock clean section found");
- ret = BCH_FSCK_REPAIR_IMPOSSIBLE;
+ ret = -BCH_ERR_fsck_repair_impossible;
goto err;
}
zero_out_btree_mem_ptr(&c->journal_keys);
- ret = journal_replay_early(c, clean, &c->journal_entries);
+ ret = journal_replay_early(c, clean);
if (ret)
goto err;
}
}
- ret = bch2_fs_journal_start(&c->journal, journal_seq,
- &c->journal_entries);
+ /*
+ * note: cmd_list_journal needs the blacklist table fully up to date so
+ * it can asterisk ignored journal entries:
+ */
+ if (c->opts.read_journal_only)
+ goto out;
+
+ ret = bch2_fs_journal_start(&c->journal, journal_seq);
if (ret)
goto err;
+ /*
+ * Skip past versions that might have possibly been used (as nonces),
+ * but hadn't had their pointers written:
+ */
+ if (c->sb.encryption_type && !c->sb.clean)
+ atomic64_add(1 << 16, &c->key_version);
+
ret = read_btree_roots(c);
if (ret)
goto err;
err = "error reading allocation information";
down_read(&c->gc_lock);
- ret = bch2_alloc_read(c, false, false);
+ ret = bch2_alloc_read(c);
up_read(&c->gc_lock);
if (ret)
goto err;
bch_verbose(c, "stripes_read done");
- set_bit(BCH_FS_ALLOC_READ_DONE, &c->flags);
-
- /*
- * If we're not running fsck, this ensures bch2_fsck_err() calls are
- * instead interpreted as bch2_inconsistent_err() calls:
- */
- if (!c->opts.fsck)
- set_bit(BCH_FS_FSCK_DONE, &c->flags);
+ bch2_stripes_heap_start(c);
- if (c->opts.fsck ||
- !(c->sb.compat & (1ULL << BCH_COMPAT_alloc_info)) ||
- !(c->sb.compat & (1ULL << BCH_COMPAT_alloc_metadata)) ||
- test_bit(BCH_FS_REBUILD_REPLICAS, &c->flags)) {
+ if (c->opts.fsck) {
bool metadata_only = c->opts.norecovery;
bch_info(c, "checking allocations");
- err = "error in mark and sweep";
+ err = "error checking allocations";
ret = bch2_gc(c, true, metadata_only);
if (ret)
goto err;
bch_verbose(c, "done checking allocations");
- }
- bch2_stripes_heap_start(c);
+ set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags);
- clear_bit(BCH_FS_REBUILD_REPLICAS, &c->flags);
- set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags);
+ bch_info(c, "checking need_discard and freespace btrees");
+ err = "error checking need_discard and freespace btrees";
+ ret = bch2_check_alloc_info(c);
+ if (ret)
+ goto err;
+ bch_verbose(c, "done checking need_discard and freespace btrees");
- /*
- * Skip past versions that might have possibly been used (as nonces),
- * but hadn't had their pointers written:
- */
- if (c->sb.encryption_type && !c->sb.clean)
- atomic64_add(1 << 16, &c->key_version);
+ set_bit(BCH_FS_MAY_GO_RW, &c->flags);
- if (c->opts.norecovery)
- goto out;
+ bch_info(c, "starting journal replay, %zu keys", c->journal_keys.nr);
+ err = "journal replay failed";
+ ret = bch2_journal_replay(c);
+ if (ret)
+ goto err;
+ if (c->opts.verbose || !c->sb.clean)
+ bch_info(c, "journal replay done");
+
+ bch_info(c, "checking lrus");
+ err = "error checking lrus";
+ ret = bch2_check_lrus(c);
+ if (ret)
+ goto err;
+ bch_verbose(c, "done checking lrus");
+ set_bit(BCH_FS_CHECK_LRUS_DONE, &c->flags);
+
+ bch_info(c, "checking backpointers to alloc keys");
+ err = "error checking backpointers to alloc keys";
+ ret = bch2_check_btree_backpointers(c);
+ if (ret)
+ goto err;
+ bch_verbose(c, "done checking backpointers to alloc keys");
+
+ bch_info(c, "checking backpointers to extents");
+ err = "error checking backpointers to extents";
+ ret = bch2_check_backpointers_to_extents(c);
+ if (ret)
+ goto err;
+ bch_verbose(c, "done checking backpointers to extents");
+
+ bch_info(c, "checking extents to backpointers");
+ err = "error checking extents to backpointers";
+ ret = bch2_check_extents_to_backpointers(c);
+ if (ret)
+ goto err;
+ bch_verbose(c, "done checking extents to backpointers");
+ set_bit(BCH_FS_CHECK_BACKPOINTERS_DONE, &c->flags);
+
+ bch_info(c, "checking alloc to lru refs");
+ err = "error checking alloc to lru refs";
+ ret = bch2_check_alloc_to_lru_refs(c);
+ if (ret)
+ goto err;
+ bch_verbose(c, "done checking alloc to lru refs");
+ set_bit(BCH_FS_CHECK_ALLOC_TO_LRU_REFS_DONE, &c->flags);
+ } else {
+ set_bit(BCH_FS_MAY_GO_RW, &c->flags);
+ set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags);
+ set_bit(BCH_FS_CHECK_LRUS_DONE, &c->flags);
+ set_bit(BCH_FS_CHECK_BACKPOINTERS_DONE, &c->flags);
+ set_bit(BCH_FS_CHECK_ALLOC_TO_LRU_REFS_DONE, &c->flags);
+ set_bit(BCH_FS_FSCK_DONE, &c->flags);
+
+ if (c->opts.norecovery)
+ goto out;
+
+ bch_verbose(c, "starting journal replay, %zu keys", c->journal_keys.nr);
+ err = "journal replay failed";
+ ret = bch2_journal_replay(c);
+ if (ret)
+ goto err;
+ if (c->opts.verbose || !c->sb.clean)
+ bch_info(c, "journal replay done");
+ }
- bch_verbose(c, "starting journal replay, %zu keys", c->journal_keys.nr);
- err = "journal replay failed";
- ret = bch2_journal_replay(c);
+ err = "error initializing freespace";
+ ret = bch2_fs_freespace_init(c);
if (ret)
goto err;
- if (c->opts.verbose || !c->sb.clean)
- bch_info(c, "journal replay done");
if (c->sb.version < bcachefs_metadata_version_snapshot_2) {
bch2_fs_lazy_rw(c);
if (!c->opts.keep_journal) {
bch2_journal_keys_free(&c->journal_keys);
- bch2_journal_entries_free(&c->journal_entries);
+ bch2_journal_entries_free(c);
}
kfree(clean);
+
+ if (!ret && test_bit(BCH_FS_HAVE_DELETED_SNAPSHOTS, &c->flags)) {
+ bch2_fs_read_write_early(c);
+ bch2_delete_dead_snapshots_async(c);
+ }
+
if (ret)
- bch_err(c, "Error in recovery: %s (%i)", err, ret);
+ bch_err(c, "Error in recovery: %s (%s)", err, bch2_err_str(ret));
else
- bch_verbose(c, "ret %i", ret);
+ bch_verbose(c, "ret %s", bch2_err_str(ret));
return ret;
err:
fsck_err:
struct qstr lostfound = QSTR("lost+found");
const char *err = "cannot allocate memory";
struct bch_dev *ca;
- LIST_HEAD(journal);
unsigned i;
int ret;
c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_extents_above_btree_updates_done);
c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_bformat_overflow_done);
+ if (c->sb.version < bcachefs_metadata_version_inode_v3)
+ c->opts.version_upgrade = true;
+
if (c->opts.version_upgrade) {
c->disk_sb.sb->version = cpu_to_le16(bcachefs_metadata_version_current);
c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALL);
}
mutex_unlock(&c->sb_lock);
- set_bit(BCH_FS_ALLOC_READ_DONE, &c->flags);
set_bit(BCH_FS_INITIAL_GC_DONE, &c->flags);
+ set_bit(BCH_FS_MAY_GO_RW, &c->flags);
set_bit(BCH_FS_FSCK_DONE, &c->flags);
for (i = 0; i < BTREE_ID_NR; i++)
bch2_btree_root_alloc(c, i);
+ for_each_online_member(ca, c, i)
+ bch2_dev_usage_init(ca);
+
err = "unable to allocate journal buckets";
for_each_online_member(ca, c, i) {
ret = bch2_dev_journal_alloc(ca);
* journal_res_get() will crash if called before this has
* set up the journal.pin FIFO and journal.cur pointer:
*/
- bch2_fs_journal_start(&c->journal, 1, &journal);
+ bch2_fs_journal_start(&c->journal, 1);
bch2_journal_set_replay_done(&c->journal);
err = "error going read-write";
* Write out the superblock and journal buckets, now that we can do
* btree updates
*/
+ bch_verbose(c, "marking superblocks");
err = "error marking superblock and journal";
for_each_member_device(ca, c, i) {
ret = bch2_trans_mark_dev_sb(c, ca);
ca->new_fs_bucket_idx = 0;
}
+ bch_verbose(c, "initializing freespace");
+ err = "error initializing freespace";
+ ret = bch2_fs_freespace_init(c);
+ if (ret)
+ goto err;
+
err = "error creating root snapshot node";
ret = bch2_fs_initialize_subvolumes(c);
if (ret)
S_IFDIR|S_IRWXU|S_IRUGO|S_IXUGO, 0, NULL);
root_inode.bi_inum = BCACHEFS_ROOT_INO;
root_inode.bi_subvol = BCACHEFS_ROOT_SUBVOL;
- bch2_inode_pack(c, &packed_inode, &root_inode);
+ bch2_inode_pack(&packed_inode, &root_inode);
packed_inode.inode.k.p.snapshot = U32_MAX;
err = "error creating root directory";
projects / bcachefs-tools-debian / blobdiff