From 12fe5797ad3e58a365442751aed58c776c09e69f Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@gmail.com>
Date: Thu, 30 Dec 2021 21:23:06 -0500
Subject: [PATCH] Update bcachefs sources to 916d92b6b4 bcachefs: Add error
messages for memory allocation failures
---
.bcachefs_revision | 2 +-
libbcachefs/alloc_background.c | 3 +-
libbcachefs/bcachefs.h | 1 +
libbcachefs/btree_cache.c | 2 +
libbcachefs/btree_gc.c | 24 +++---
libbcachefs/btree_iter.c | 17 ++--
libbcachefs/btree_key_cache.c | 28 ++++---
libbcachefs/btree_update.h | 2 +-
libbcachefs/btree_update_interior.h | 4 +-
libbcachefs/btree_update_leaf.c | 14 +++-
libbcachefs/buckets.c | 16 ++--
libbcachefs/ec.c | 2 +-
libbcachefs/fsck.c | 37 +++++----
libbcachefs/inode.c | 57 ++++++++-----
libbcachefs/movinggc.c | 5 ++
libbcachefs/quota.c | 6 +-
libbcachefs/recovery.c | 121 +++++++++++++++++++---------
libbcachefs/recovery.h | 2 +
libbcachefs/super.c | 46 ++++++-----
libbcachefs/tests.c | 38 +++++----
libbcachefs/util.c | 2 +-
21 files changed, 268 insertions(+), 161 deletions(-)
diff --git a/.bcachefs_revision b/.bcachefs_revision
index e2012548..72264415 100644
--- a/.bcachefs_revision
+++ b/.bcachefs_revision
@@ -1 +1 @@
-078a1a596a74ade60db6eee0f0be927defb7abed
+916d92b6b46b13873118a608ff16212f966375ba
diff --git a/libbcachefs/alloc_background.c b/libbcachefs/alloc_background.c
index 2a36af5e..eb2e6642 100644
--- a/libbcachefs/alloc_background.c
+++ b/libbcachefs/alloc_background.c
@@ -859,7 +859,8 @@ static void discard_one_bucket(struct bch_fs *c, struct bch_dev *ca, u64 b)
static bool allocator_thread_running(struct bch_dev *ca)
{
unsigned state = ca->mi.state == BCH_MEMBER_STATE_rw &&
- test_bit(BCH_FS_ALLOCATOR_RUNNING, &ca->fs->flags)
+ test_bit(BCH_FS_ALLOCATOR_RUNNING, &ca->fs->flags) &&
+ test_bit(BCH_FS_ALLOC_REPLAY_DONE, &ca->fs->flags)
? ALLOCATOR_running
: ALLOCATOR_stopped;
alloc_thread_set_state(ca, state);
diff --git a/libbcachefs/bcachefs.h b/libbcachefs/bcachefs.h
index 3ada85ac..b21151ea 100644
--- a/libbcachefs/bcachefs.h
+++ b/libbcachefs/bcachefs.h
@@ -510,6 +510,7 @@ enum {
BCH_FS_INITIAL_GC_DONE,
BCH_FS_INITIAL_GC_UNFIXED,
BCH_FS_TOPOLOGY_REPAIR_DONE,
+ BCH_FS_ALLOC_REPLAY_DONE,
BCH_FS_BTREE_INTERIOR_REPLAY_DONE,
BCH_FS_FSCK_DONE,
BCH_FS_STARTED,
diff --git a/libbcachefs/btree_cache.c b/libbcachefs/btree_cache.c
index b13563d7..2788ba17 100644
--- a/libbcachefs/btree_cache.c
+++ b/libbcachefs/btree_cache.c
@@ -83,6 +83,8 @@ static int btree_node_data_alloc(struct bch_fs *c, struct btree *b, gfp_t gfp)
b->aux_data = mmap(NULL, btree_aux_data_bytes(b),
PROT_READ|PROT_WRITE|PROT_EXEC,
MAP_PRIVATE|MAP_ANONYMOUS, 0, 0);
+ if (b->aux_data == MAP_FAILED)
+ b->aux_data = NULL;
#endif
if (!b->aux_data) {
kvpfree(b->data, btree_bytes(c));
diff --git a/libbcachefs/btree_gc.c b/libbcachefs/btree_gc.c
index d1883701..9e3213b9 100644
--- a/libbcachefs/btree_gc.c
+++ b/libbcachefs/btree_gc.c
@@ -169,11 +169,11 @@ static int set_node_min(struct bch_fs *c, struct btree *b, struct bpos new_min)
new->v.min_key = new_min;
SET_BTREE_PTR_RANGE_UPDATED(&new->v, true);
- ret = bch2_journal_key_insert(c, b->c.btree_id, b->c.level + 1, &new->k_i);
- kfree(new);
-
- if (ret)
+ ret = bch2_journal_key_insert_take(c, b->c.btree_id, b->c.level + 1, &new->k_i);
+ if (ret) {
+ kfree(new);
return ret;
+ }
bch2_btree_node_drop_keys_outside_node(b);
@@ -198,11 +198,11 @@ static int set_node_max(struct bch_fs *c, struct btree *b, struct bpos new_max)
new->k.p = new_max;
SET_BTREE_PTR_RANGE_UPDATED(&new->v, true);
- ret = bch2_journal_key_insert(c, b->c.btree_id, b->c.level + 1, &new->k_i);
- kfree(new);
-
- if (ret)
+ ret = bch2_journal_key_insert_take(c, b->c.btree_id, b->c.level + 1, &new->k_i);
+ if (ret) {
+ kfree(new);
return ret;
+ }
bch2_btree_node_drop_keys_outside_node(b);
@@ -690,10 +690,10 @@ found:
}
}
- ret = bch2_journal_key_insert(c, btree_id, level, new);
- kfree(new);
-
- if (!ret)
+ ret = bch2_journal_key_insert_take(c, btree_id, level, new);
+ if (ret)
+ kfree(new);
+ else
*k = bkey_i_to_s_c(new);
}
fsck_err:
diff --git a/libbcachefs/btree_iter.c b/libbcachefs/btree_iter.c
index 65ab2cd6..9ebb81d7 100644
--- a/libbcachefs/btree_iter.c
+++ b/libbcachefs/btree_iter.c
@@ -2437,15 +2437,18 @@ struct bkey_s_c bch2_btree_iter_peek_slot(struct btree_iter *iter)
if (bkey_cmp(iter->pos, next) < 0) {
bkey_init(&iter->k);
iter->k.p = iter->pos;
- bch2_key_resize(&iter->k,
- min_t(u64, KEY_SIZE_MAX,
- (next.inode == iter->pos.inode
- ? next.offset
- : KEY_OFFSET_MAX) -
- iter->pos.offset));
+
+ if (iter->flags & BTREE_ITER_IS_EXTENTS) {
+ bch2_key_resize(&iter->k,
+ min_t(u64, KEY_SIZE_MAX,
+ (next.inode == iter->pos.inode
+ ? next.offset
+ : KEY_OFFSET_MAX) -
+ iter->pos.offset));
+ EBUG_ON(!iter->k.size);
+ }
k = (struct bkey_s_c) { &iter->k, NULL };
- EBUG_ON(!k.k->size);
}
}
diff --git a/libbcachefs/btree_key_cache.c b/libbcachefs/btree_key_cache.c
index 230a920a..80ed79b0 100644
--- a/libbcachefs/btree_key_cache.c
+++ b/libbcachefs/btree_key_cache.c
@@ -146,19 +146,23 @@ bkey_cached_reuse(struct btree_key_cache *c)
}
static struct bkey_cached *
-btree_key_cache_create(struct btree_key_cache *c,
+btree_key_cache_create(struct bch_fs *c,
enum btree_id btree_id,
struct bpos pos)
{
+ struct btree_key_cache *bc = &c->btree_key_cache;
struct bkey_cached *ck;
bool was_new = true;
- ck = bkey_cached_alloc(c);
+ ck = bkey_cached_alloc(bc);
if (unlikely(!ck)) {
- ck = bkey_cached_reuse(c);
- if (unlikely(!ck))
+ ck = bkey_cached_reuse(bc);
+ if (unlikely(!ck)) {
+ bch_err(c, "error allocating memory for key cache item, btree %s",
+ bch2_btree_ids[btree_id]);
return ERR_PTR(-ENOMEM);
+ }
was_new = false;
}
@@ -175,7 +179,7 @@ btree_key_cache_create(struct btree_key_cache *c,
ck->valid = false;
ck->flags = 1U << BKEY_CACHED_ACCESSED;
- if (unlikely(rhashtable_lookup_insert_fast(&c->table,
+ if (unlikely(rhashtable_lookup_insert_fast(&bc->table,
&ck->hash,
bch2_btree_key_cache_params))) {
/* We raced with another fill: */
@@ -185,15 +189,15 @@ btree_key_cache_create(struct btree_key_cache *c,
six_unlock_intent(&ck->c.lock);
kfree(ck);
} else {
- mutex_lock(&c->lock);
- bkey_cached_free(c, ck);
- mutex_unlock(&c->lock);
+ mutex_lock(&bc->lock);
+ bkey_cached_free(bc, ck);
+ mutex_unlock(&bc->lock);
}
return NULL;
}
- atomic_long_inc(&c->nr_keys);
+ atomic_long_inc(&bc->nr_keys);
six_unlock_write(&ck->c.lock);
@@ -204,6 +208,7 @@ static int btree_key_cache_fill(struct btree_trans *trans,
struct btree_path *ck_path,
struct bkey_cached *ck)
{
+ struct bch_fs *c = trans->c;
struct btree_iter iter;
struct bkey_s_c k;
unsigned new_u64s = 0;
@@ -233,6 +238,8 @@ static int btree_key_cache_fill(struct btree_trans *trans,
new_u64s = roundup_pow_of_two(new_u64s);
new_k = kmalloc(new_u64s * sizeof(u64), GFP_NOFS);
if (!new_k) {
+ bch_err(c, "error allocating memory for key cache key, btree %s u64s %u",
+ bch2_btree_ids[ck->key.btree_id], new_u64s);
ret = -ENOMEM;
goto err;
}
@@ -293,8 +300,7 @@ retry:
return 0;
}
- ck = btree_key_cache_create(&c->btree_key_cache,
- path->btree_id, path->pos);
+ ck = btree_key_cache_create(c, path->btree_id, path->pos);
ret = PTR_ERR_OR_ZERO(ck);
if (ret)
goto err;
diff --git a/libbcachefs/btree_update.h b/libbcachefs/btree_update.h
index 89f07e58..16ebf1a2 100644
--- a/libbcachefs/btree_update.h
+++ b/libbcachefs/btree_update.h
@@ -63,7 +63,7 @@ int bch2_btree_insert(struct bch_fs *, enum btree_id, struct bkey_i *,
int bch2_btree_delete_range_trans(struct btree_trans *, enum btree_id,
struct bpos, struct bpos, unsigned, u64 *);
int bch2_btree_delete_range(struct bch_fs *, enum btree_id,
- struct bpos, struct bpos, u64 *);
+ struct bpos, struct bpos, unsigned, u64 *);
int bch2_btree_node_rewrite(struct btree_trans *, struct btree_iter *,
struct btree *, unsigned);
diff --git a/libbcachefs/btree_update_interior.h b/libbcachefs/btree_update_interior.h
index 8cf59cee..8dc86fa6 100644
--- a/libbcachefs/btree_update_interior.h
+++ b/libbcachefs/btree_update_interior.h
@@ -82,12 +82,12 @@ struct btree_update {
/* Nodes being freed: */
struct keylist old_keys;
u64 _old_keys[BTREE_UPDATE_NODES_MAX *
- BKEY_BTREE_PTR_VAL_U64s_MAX];
+ BKEY_BTREE_PTR_U64s_MAX];
/* Nodes being added: */
struct keylist new_keys;
u64 _new_keys[BTREE_UPDATE_NODES_MAX *
- BKEY_BTREE_PTR_VAL_U64s_MAX];
+ BKEY_BTREE_PTR_U64s_MAX];
/* New nodes, that will be made reachable by this update: */
struct btree *new_nodes[BTREE_UPDATE_NODES_MAX];
diff --git a/libbcachefs/btree_update_leaf.c b/libbcachefs/btree_update_leaf.c
index 1966441b..f561e09c 100644
--- a/libbcachefs/btree_update_leaf.c
+++ b/libbcachefs/btree_update_leaf.c
@@ -308,6 +308,7 @@ btree_key_can_insert_cached(struct btree_trans *trans,
struct btree_path *path,
unsigned u64s)
{
+ struct bch_fs *c = trans->c;
struct bkey_cached *ck = (void *) path->l[0].b;
unsigned new_u64s;
struct bkey_i *new_k;
@@ -315,7 +316,7 @@ btree_key_can_insert_cached(struct btree_trans *trans,
EBUG_ON(path->level);
if (!test_bit(BKEY_CACHED_DIRTY, &ck->flags) &&
- bch2_btree_key_cache_must_wait(trans->c) &&
+ bch2_btree_key_cache_must_wait(c) &&
!(trans->flags & BTREE_INSERT_JOURNAL_RECLAIM))
return BTREE_INSERT_NEED_JOURNAL_RECLAIM;
@@ -330,8 +331,11 @@ btree_key_can_insert_cached(struct btree_trans *trans,
new_u64s = roundup_pow_of_two(u64s);
new_k = krealloc(ck->k, new_u64s * sizeof(u64), GFP_NOFS);
- if (!new_k)
+ if (!new_k) {
+ bch_err(c, "error allocating memory for key cache key, btree %s u64s %u",
+ bch2_btree_ids[path->btree_id], new_u64s);
return -ENOMEM;
+ }
ck->u64s = new_u64s;
ck->k = new_k;
@@ -1463,7 +1467,7 @@ retry:
*/
delete.k.p = iter.pos;
- if (btree_node_type_is_extents(id)) {
+ if (iter.flags & BTREE_ITER_IS_EXTENTS) {
unsigned max_sectors =
KEY_SIZE_MAX & (~0 << trans->c->block_bits);
@@ -1500,8 +1504,10 @@ retry:
*/
int bch2_btree_delete_range(struct bch_fs *c, enum btree_id id,
struct bpos start, struct bpos end,
+ unsigned iter_flags,
u64 *journal_seq)
{
return bch2_trans_do(c, NULL, journal_seq, 0,
- bch2_btree_delete_range_trans(&trans, id, start, end, 0, journal_seq));
+ bch2_btree_delete_range_trans(&trans, id, start, end,
+ iter_flags, journal_seq));
}
diff --git a/libbcachefs/buckets.c b/libbcachefs/buckets.c
index 738ce67d..709df569 100644
--- a/libbcachefs/buckets.c
+++ b/libbcachefs/buckets.c
@@ -940,9 +940,11 @@ static int bch2_mark_stripe_ptr(struct btree_trans *trans,
BUG_ON(!(flags & BTREE_TRIGGER_GC));
m = genradix_ptr_alloc(&c->gc_stripes, p.idx, GFP_KERNEL);
-
- if (!m)
+ if (!m) {
+ bch_err(c, "error allocating memory for gc_stripes, idx %llu",
+ (u64) p.idx);
return -ENOMEM;
+ }
spin_lock(&c->ec_stripes_heap_lock);
@@ -1053,7 +1055,7 @@ static int bch2_mark_stripe(struct btree_trans *trans,
bool gc = flags & BTREE_TRIGGER_GC;
u64 journal_seq = trans->journal_res.seq;
struct bch_fs *c = trans->c;
- size_t idx = new.k->p.offset;
+ u64 idx = new.k->p.offset;
const struct bch_stripe *old_s = old.k->type == KEY_TYPE_stripe
? bkey_s_c_to_stripe(old).v : NULL;
const struct bch_stripe *new_s = new.k->type == KEY_TYPE_stripe
@@ -1071,7 +1073,7 @@ static int bch2_mark_stripe(struct btree_trans *trans,
bch2_bkey_val_to_text(&PBUF(buf1), c, old);
bch2_bkey_val_to_text(&PBUF(buf2), c, new);
- bch_err_ratelimited(c, "error marking nonexistent stripe %zu while marking\n"
+ bch_err_ratelimited(c, "error marking nonexistent stripe %llu while marking\n"
"old %s\n"
"new %s", idx, buf1, buf2);
bch2_inconsistent_error(c);
@@ -1103,9 +1105,11 @@ static int bch2_mark_stripe(struct btree_trans *trans,
struct gc_stripe *m =
genradix_ptr_alloc(&c->gc_stripes, idx, GFP_KERNEL);
- if (!m)
+ if (!m) {
+ bch_err(c, "error allocating memory for gc_stripes, idx %llu",
+ idx);
return -ENOMEM;
-
+ }
/*
* This will be wrong when we bring back runtime gc: we should
* be unmarking the old key and then marking the new key
diff --git a/libbcachefs/ec.c b/libbcachefs/ec.c
index 3cccd1fa..9a1751d4 100644
--- a/libbcachefs/ec.c
+++ b/libbcachefs/ec.c
@@ -677,7 +677,7 @@ static int ec_stripe_delete(struct bch_fs *c, size_t idx)
return bch2_btree_delete_range(c, BTREE_ID_stripes,
POS(0, idx),
POS(0, idx + 1),
- NULL);
+ 0, NULL);
}
static void ec_stripe_delete_work(struct work_struct *work)
diff --git a/libbcachefs/fsck.c b/libbcachefs/fsck.c
index 835ef114..43b6159b 100644
--- a/libbcachefs/fsck.c
+++ b/libbcachefs/fsck.c
@@ -564,14 +564,17 @@ static struct inode_walker inode_walker_init(void)
return (struct inode_walker) { 0, };
}
-static int inode_walker_realloc(struct inode_walker *w)
+static int inode_walker_realloc(struct bch_fs *c, struct inode_walker *w)
{
if (w->nr == w->size) {
size_t new_size = max_t(size_t, 8UL, w->size * 2);
void *d = krealloc(w->d, new_size * sizeof(w->d[0]),
GFP_KERNEL);
- if (!d)
+ if (!d) {
+ bch_err(c, "fsck: error allocating memory for inode_walker, size %zu",
+ new_size);
return -ENOMEM;
+ }
w->d = d;
w->size = new_size;
@@ -586,7 +589,7 @@ static int add_inode(struct bch_fs *c, struct inode_walker *w,
struct bch_inode_unpacked u;
int ret;
- ret = inode_walker_realloc(w);
+ ret = inode_walker_realloc(c, w);
if (ret)
return ret;
@@ -647,7 +650,7 @@ found:
while (i && w->d[i - 1].snapshot > pos.snapshot)
--i;
- ret = inode_walker_realloc(w);
+ ret = inode_walker_realloc(c, w);
if (ret)
return ret;
@@ -1132,9 +1135,9 @@ static int check_i_sectors(struct btree_trans *trans, struct inode_walker *w)
count2 = lockrestart_do(trans,
bch2_count_inode_sectors(trans, w->cur_inum, i->snapshot));
- if (fsck_err_on(i->count != count2, c,
- "fsck counted i_sectors wrong: got %llu should be %llu",
- i->count, count2)) {
+ 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;
@@ -1316,9 +1319,9 @@ static int check_subdir_count(struct btree_trans *trans, struct inode_walker *w)
count2 = lockrestart_do(trans,
bch2_count_subdirs(trans, w->cur_inum, i->snapshot));
- if (fsck_err_on(i->count != count2, c,
- "directory %llu:%u: fsck counted subdirectories wrong, got %llu should be %llu",
- w->cur_inum, i->snapshot, i->count, count2)) {
+ if (i->count != count2) {
+ bch_err(c, "fsck counted subdirectories wrong: got %llu should be %llu",
+ i->count, count2);
i->count = count2;
if (i->inode.bi_nlink == i->count)
continue;
@@ -1812,7 +1815,8 @@ static bool path_is_dup(struct pathbuf *p, u64 inum, u32 snapshot)
return false;
}
-static int path_down(struct pathbuf *p, u64 inum, u32 snapshot)
+static int path_down(struct bch_fs *c, struct pathbuf *p,
+ u64 inum, u32 snapshot)
{
if (p->nr == p->size) {
size_t new_size = max_t(size_t, 256UL, p->size * 2);
@@ -1820,6 +1824,8 @@ static int path_down(struct pathbuf *p, u64 inum, u32 snapshot)
new_size * sizeof(p->entries[0]),
GFP_KERNEL);
if (!n) {
+ bch_err(c, "fsck: error allocating memory for pathbuf, size %zu",
+ new_size);
return -ENOMEM;
}
@@ -1893,7 +1899,7 @@ static int check_path(struct btree_trans *trans,
if (!S_ISDIR(inode->bi_mode))
break;
- ret = path_down(p, inode->bi_inum, snapshot);
+ ret = path_down(c, p, inode->bi_inum, snapshot);
if (ret) {
bch_err(c, "memory allocation failure");
return ret;
@@ -1998,12 +2004,15 @@ struct nlink_table {
} *d;
};
-static int add_nlink(struct nlink_table *t, u64 inum, u32 snapshot)
+static int add_nlink(struct bch_fs *c, struct nlink_table *t,
+ u64 inum, u32 snapshot)
{
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);
if (!d) {
+ bch_err(c, "fsck: error allocating memory for nlink_table, size %zu",
+ new_size);
return -ENOMEM;
}
@@ -2093,7 +2102,7 @@ static int check_nlinks_find_hardlinks(struct bch_fs *c,
if (!u.bi_nlink)
continue;
- ret = add_nlink(t, k.k->p.offset, k.k->p.snapshot);
+ ret = add_nlink(c, t, k.k->p.offset, k.k->p.snapshot);
if (ret) {
*end = k.k->p.offset;
ret = 0;
diff --git a/libbcachefs/inode.c b/libbcachefs/inode.c
index 3a7c1468..ef6da535 100644
--- a/libbcachefs/inode.c
+++ b/libbcachefs/inode.c
@@ -585,49 +585,62 @@ found_slot:
static int bch2_inode_delete_keys(struct btree_trans *trans,
subvol_inum inum, enum btree_id id)
{
- struct btree_iter iter;
- struct bkey_s_c k;
- struct bkey_i delete;
- u32 snapshot;
+ u64 offset = 0;
int ret = 0;
- /*
- * We're never going to be deleting extents, no need to use an extent
- * iterator:
- */
- bch2_trans_iter_init(trans, &iter, id, POS(inum.inum, 0),
- BTREE_ITER_NOT_EXTENTS|
- BTREE_ITER_INTENT);
+ while (!ret || ret == -EINTR) {
+ struct disk_reservation disk_res =
+ bch2_disk_reservation_init(trans->c, 0);
+ struct btree_iter iter;
+ struct bkey_s_c k;
+ struct bkey_i delete;
+ u32 snapshot;
- while (1) {
bch2_trans_begin(trans);
ret = bch2_subvolume_get_snapshot(trans, inum.subvol, &snapshot);
if (ret)
- goto err;
-
- bch2_btree_iter_set_snapshot(&iter, snapshot);
+ continue;
+ bch2_trans_iter_init(trans, &iter, id,
+ SPOS(inum.inum, offset, snapshot),
+ BTREE_ITER_INTENT);
k = bch2_btree_iter_peek(&iter);
+
+ if (!k.k || iter.pos.inode != inum.inum) {
+ bch2_trans_iter_exit(trans, &iter);
+ break;
+ }
+
ret = bkey_err(k);
if (ret)
goto err;
- if (!k.k || iter.pos.inode != inum.inum)
- break;
-
bkey_init(&delete.k);
delete.k.p = iter.pos;
+ if (btree_node_type_is_extents(iter.btree_id)) {
+ unsigned max_sectors =
+ min_t(u64, U64_MAX - iter.pos.offset,
+ KEY_SIZE_MAX & (~0 << trans->c->block_bits));
+
+ /* create the biggest key we can */
+ bch2_key_resize(&delete.k, max_sectors);
+
+ ret = bch2_extent_trim_atomic(trans, &iter, &delete);
+ if (ret)
+ goto err;
+ }
+
ret = bch2_trans_update(trans, &iter, &delete, 0) ?:
- bch2_trans_commit(trans, NULL, NULL,
+ bch2_trans_commit(trans, &disk_res, NULL,
BTREE_INSERT_NOFAIL);
+ bch2_disk_reservation_put(trans->c, &disk_res);
err:
- if (ret && ret != -EINTR)
- break;
+ offset = iter.pos.offset;
+ bch2_trans_iter_exit(trans, &iter);
}
- bch2_trans_iter_exit(trans, &iter);
return ret;
}
diff --git a/libbcachefs/movinggc.c b/libbcachefs/movinggc.c
index 7b7eee9b..7cd1b0cf 100644
--- a/libbcachefs/movinggc.c
+++ b/libbcachefs/movinggc.c
@@ -205,6 +205,11 @@ static int bch2_copygc(struct bch_fs *c)
up_read(&ca->bucket_lock);
}
+ if (!h->used) {
+ bch_err_ratelimited(c, "copygc requested to run but found no buckets to move!");
+ return 0;
+ }
+
/*
* Our btree node allocations also come out of RESERVE_MOVINGGC:
*/
diff --git a/libbcachefs/quota.c b/libbcachefs/quota.c
index 8f8f4b0a..54bb2a45 100644
--- a/libbcachefs/quota.c
+++ b/libbcachefs/quota.c
@@ -570,7 +570,7 @@ static int bch2_quota_remove(struct super_block *sb, unsigned uflags)
ret = bch2_btree_delete_range(c, BTREE_ID_quotas,
POS(QTYP_USR, 0),
POS(QTYP_USR + 1, 0),
- NULL);
+ 0, NULL);
if (ret)
return ret;
}
@@ -582,7 +582,7 @@ static int bch2_quota_remove(struct super_block *sb, unsigned uflags)
ret = bch2_btree_delete_range(c, BTREE_ID_quotas,
POS(QTYP_GRP, 0),
POS(QTYP_GRP + 1, 0),
- NULL);
+ 0, NULL);
if (ret)
return ret;
}
@@ -594,7 +594,7 @@ static int bch2_quota_remove(struct super_block *sb, unsigned uflags)
ret = bch2_btree_delete_range(c, BTREE_ID_quotas,
POS(QTYP_PRJ, 0),
POS(QTYP_PRJ + 1, 0),
- NULL);
+ 0, NULL);
if (ret)
return ret;
}
diff --git a/libbcachefs/recovery.c b/libbcachefs/recovery.c
index 8b0e468f..0b923037 100644
--- a/libbcachefs/recovery.c
+++ b/libbcachefs/recovery.c
@@ -109,18 +109,27 @@ static void journal_iter_fix(struct bch_fs *c, struct journal_iter *iter, unsign
iter->idx++;
}
-int bch2_journal_key_insert(struct bch_fs *c, enum btree_id id,
- unsigned level, struct bkey_i *k)
+int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id,
+ unsigned level, struct bkey_i *k)
{
struct journal_key n = {
.btree_id = id,
.level = level,
+ .k = k,
.allocated = true
};
struct journal_keys *keys = &c->journal_keys;
struct journal_iter *iter;
unsigned idx = journal_key_search(keys, id, level, k->k.p);
+ if (idx < keys->nr &&
+ journal_key_cmp(&n, &keys->d[idx]) == 0) {
+ if (keys->d[idx].allocated)
+ kfree(keys->d[idx].k);
+ keys->d[idx] = n;
+ return 0;
+ }
+
if (keys->nr == keys->size) {
struct journal_keys new_keys = {
.nr = keys->nr,
@@ -140,27 +149,31 @@ int bch2_journal_key_insert(struct bch_fs *c, enum btree_id id,
*keys = new_keys;
}
- n.k = kmalloc(bkey_bytes(&k->k), GFP_KERNEL);
- if (!n.k)
- return -ENOMEM;
+ array_insert_item(keys->d, keys->nr, idx, n);
- bkey_copy(n.k, k);
-
- if (idx < keys->nr &&
- journal_key_cmp(&n, &keys->d[idx]) == 0) {
- if (keys->d[idx].allocated)
- kfree(keys->d[idx].k);
- keys->d[idx] = n;
- } else {
- array_insert_item(keys->d, keys->nr, idx, n);
-
- list_for_each_entry(iter, &c->journal_iters, list)
- journal_iter_fix(c, iter, idx);
- }
+ list_for_each_entry(iter, &c->journal_iters, list)
+ journal_iter_fix(c, iter, idx);
return 0;
}
+int bch2_journal_key_insert(struct bch_fs *c, enum btree_id id,
+ unsigned level, struct bkey_i *k)
+{
+ struct bkey_i *n;
+ int ret;
+
+ n = kmalloc(bkey_bytes(&k->k), GFP_KERNEL);
+ if (!n)
+ return -ENOMEM;
+
+ bkey_copy(n, k);
+ ret = bch2_journal_key_insert_take(c, id, level, n);
+ if (ret)
+ kfree(n);
+ return ret;
+}
+
int bch2_journal_key_delete(struct bch_fs *c, enum btree_id id,
unsigned level, struct bpos pos)
{
@@ -548,8 +561,8 @@ static int bch2_journal_replay_key(struct bch_fs *c, struct journal_key *k)
static int journal_sort_seq_cmp(const void *_l, const void *_r)
{
- const struct journal_key *l = _l;
- const struct journal_key *r = _r;
+ const struct journal_key *l = *((const struct journal_key **)_l);
+ const struct journal_key *r = *((const struct journal_key **)_r);
return cmp_int(r->level, l->level) ?:
cmp_int(l->journal_seq, r->journal_seq) ?:
@@ -557,18 +570,30 @@ static int journal_sort_seq_cmp(const void *_l, const void *_r)
bpos_cmp(l->k->k.p, r->k->k.p);
}
-static int bch2_journal_replay(struct bch_fs *c,
- struct journal_keys keys)
+static int bch2_journal_replay(struct bch_fs *c)
{
+ struct journal_keys *keys = &c->journal_keys;
+ struct journal_key **keys_sorted, *k;
struct journal *j = &c->journal;
- struct journal_key *i;
+ struct bch_dev *ca;
+ unsigned idx;
+ size_t i;
u64 seq;
int ret;
- sort(keys.d, keys.nr, sizeof(keys.d[0]), journal_sort_seq_cmp, NULL);
+ keys_sorted = kmalloc_array(sizeof(*keys_sorted), keys->nr, GFP_KERNEL);
+ if (!keys_sorted)
+ return -ENOMEM;
- if (keys.nr)
- replay_now_at(j, keys.journal_seq_base);
+ for (i = 0; i < keys->nr; i++)
+ keys_sorted[i] = &keys->d[i];
+
+ sort(keys_sorted, keys->nr,
+ sizeof(keys_sorted[0]),
+ journal_sort_seq_cmp, NULL);
+
+ if (keys->nr)
+ replay_now_at(j, keys->journal_seq_base);
seq = j->replay_journal_seq;
@@ -576,26 +601,35 @@ static int bch2_journal_replay(struct bch_fs *c,
* First replay updates to the alloc btree - these will only update the
* btree key cache:
*/
- for_each_journal_key(keys, i) {
+ for (i = 0; i < keys->nr; i++) {
+ k = keys_sorted[i];
+
cond_resched();
- if (!i->level && i->btree_id == BTREE_ID_alloc) {
- j->replay_journal_seq = keys.journal_seq_base + i->journal_seq;
- ret = bch2_journal_replay_key(c, i);
+ if (!k->level && k->btree_id == BTREE_ID_alloc) {
+ j->replay_journal_seq = keys->journal_seq_base + k->journal_seq;
+ ret = bch2_journal_replay_key(c, k);
if (ret)
goto err;
}
}
+ /* Now we can start the allocator threads: */
+ set_bit(BCH_FS_ALLOC_REPLAY_DONE, &c->flags);
+ for_each_member_device(ca, c, idx)
+ bch2_wake_allocator(ca);
+
/*
* Next replay updates to interior btree nodes:
*/
- for_each_journal_key(keys, i) {
+ for (i = 0; i < keys->nr; i++) {
+ k = keys_sorted[i];
+
cond_resched();
- if (i->level) {
- j->replay_journal_seq = keys.journal_seq_base + i->journal_seq;
- ret = bch2_journal_replay_key(c, i);
+ if (k->level) {
+ j->replay_journal_seq = keys->journal_seq_base + k->journal_seq;
+ ret = bch2_journal_replay_key(c, k);
if (ret)
goto err;
}
@@ -615,15 +649,17 @@ static int bch2_journal_replay(struct bch_fs *c,
/*
* Now replay leaf node updates:
*/
- for_each_journal_key(keys, i) {
+ for (i = 0; i < keys->nr; i++) {
+ k = keys_sorted[i];
+
cond_resched();
- if (i->level || i->btree_id == BTREE_ID_alloc)
+ if (k->level || k->btree_id == BTREE_ID_alloc)
continue;
- replay_now_at(j, keys.journal_seq_base + i->journal_seq);
+ replay_now_at(j, keys->journal_seq_base + k->journal_seq);
- ret = bch2_journal_replay_key(c, i);
+ ret = bch2_journal_replay_key(c, k);
if (ret)
goto err;
}
@@ -633,10 +669,14 @@ static int bch2_journal_replay(struct bch_fs *c,
bch2_journal_set_replay_done(j);
bch2_journal_flush_all_pins(j);
+ kfree(keys_sorted);
+
return bch2_journal_error(j);
err:
bch_err(c, "journal replay: error %d while replaying key at btree %s level %u",
- ret, bch2_btree_ids[i->btree_id], i->level);
+ ret, bch2_btree_ids[k->btree_id], k->level);
+ kfree(keys_sorted);
+
return ret;
}
@@ -1208,7 +1248,7 @@ use_clean:
bch_verbose(c, "starting journal replay");
err = "journal replay failed";
- ret = bch2_journal_replay(c, c->journal_keys);
+ ret = bch2_journal_replay(c);
if (ret)
goto err;
bch_verbose(c, "journal replay done");
@@ -1380,6 +1420,7 @@ int bch2_fs_initialize(struct bch_fs *c)
for (i = 0; i < BTREE_ID_NR; i++)
bch2_btree_root_alloc(c, i);
+ set_bit(BCH_FS_ALLOC_REPLAY_DONE, &c->flags);
set_bit(BCH_FS_BTREE_INTERIOR_REPLAY_DONE, &c->flags);
set_bit(JOURNAL_RECLAIM_STARTED, &c->journal.flags);
diff --git a/libbcachefs/recovery.h b/libbcachefs/recovery.h
index e45c70b3..1504e0bd 100644
--- a/libbcachefs/recovery.h
+++ b/libbcachefs/recovery.h
@@ -31,6 +31,8 @@ struct btree_and_journal_iter {
} last;
};
+int bch2_journal_key_insert_take(struct bch_fs *, enum btree_id,
+ unsigned, struct bkey_i *);
int bch2_journal_key_insert(struct bch_fs *, enum btree_id,
unsigned, struct bkey_i *);
int bch2_journal_key_delete(struct bch_fs *, enum btree_id,
diff --git a/libbcachefs/super.c b/libbcachefs/super.c
index df6bffef..3afa7ebd 100644
--- a/libbcachefs/super.c
+++ b/libbcachefs/super.c
@@ -1478,7 +1478,7 @@ static int bch2_dev_remove_alloc(struct bch_fs *c, struct bch_dev *ca)
return bch2_btree_delete_range(c, BTREE_ID_alloc,
POS(ca->dev_idx, 0),
POS(ca->dev_idx + 1, 0),
- NULL);
+ 0, NULL);
}
int bch2_dev_remove(struct bch_fs *c, struct bch_dev *ca, int flags)
@@ -1599,18 +1599,24 @@ int bch2_dev_add(struct bch_fs *c, const char *path)
int ret;
ret = bch2_read_super(path, &opts, &sb);
- if (ret)
+ if (ret) {
+ bch_err(c, "device add error: error reading super: %i", ret);
return ret;
+ }
err = bch2_sb_validate(&sb);
- if (err)
+ if (err) {
+ bch_err(c, "device add error: error validating super: %s", err);
return -EINVAL;
+ }
dev_mi = bch2_sb_get_members(sb.sb)->members[sb.sb->dev_idx];
err = bch2_dev_may_add(sb.sb, c);
- if (err)
+ if (err) {
+ bch_err(c, "device add error: %s", err);
return -EINVAL;
+ }
ca = __bch2_dev_alloc(c, &dev_mi);
if (!ca) {
@@ -1624,24 +1630,27 @@ int bch2_dev_add(struct bch_fs *c, const char *path)
return ret;
}
- err = "journal alloc failed";
ret = bch2_dev_journal_alloc(ca);
- if (ret)
+ if (ret) {
+ bch_err(c, "device add error: journal alloc failed");
goto err;
+ }
down_write(&c->state_lock);
mutex_lock(&c->sb_lock);
- err = "insufficient space in new superblock";
ret = bch2_sb_from_fs(c, ca);
- if (ret)
+ if (ret) {
+ bch_err(c, "device add error: new device superblock too small");
goto err_unlock;
+ }
mi = bch2_sb_get_members(ca->disk_sb.sb);
if (!bch2_sb_resize_members(&ca->disk_sb,
le32_to_cpu(mi->field.u64s) +
sizeof(dev_mi) / sizeof(u64))) {
+ bch_err(c, "device add error: new device superblock too small");
ret = -ENOSPC;
goto err_unlock;
}
@@ -1654,7 +1663,7 @@ int bch2_dev_add(struct bch_fs *c, const char *path)
if (!bch2_dev_exists(c->disk_sb.sb, mi, dev_idx))
goto have_slot;
no_slot:
- err = "no slots available in superblock";
+ bch_err(c, "device add error: already have maximum number of devices");
ret = -ENOSPC;
goto err_unlock;
@@ -1663,12 +1672,12 @@ have_slot:
u64s = (sizeof(struct bch_sb_field_members) +
sizeof(struct bch_member) * nr_devices) / sizeof(u64);
- err = "no space in superblock for member info";
- ret = -ENOSPC;
-
mi = bch2_sb_resize_members(&c->disk_sb, u64s);
- if (!mi)
+ if (!mi) {
+ bch_err(c, "device add error: no room in superblock for member info");
+ ret = -ENOSPC;
goto err_unlock;
+ }
/* success: */
@@ -1684,17 +1693,20 @@ have_slot:
bch2_dev_usage_journal_reserve(c);
- err = "error marking superblock";
ret = bch2_trans_mark_dev_sb(c, ca);
- if (ret)
+ if (ret) {
+ bch_err(c, "device add error: error marking new superblock: %i", ret);
goto err_late;
+ }
ca->new_fs_bucket_idx = 0;
if (ca->mi.state == BCH_MEMBER_STATE_rw) {
ret = __bch2_dev_read_write(c, ca);
- if (ret)
+ if (ret) {
+ bch_err(c, "device add error: error going RW on new device: %i", ret);
goto err_late;
+ }
}
up_write(&c->state_lock);
@@ -1707,11 +1719,9 @@ err:
if (ca)
bch2_dev_free(ca);
bch2_free_super(&sb);
- bch_err(c, "Unable to add device: %s", err);
return ret;
err_late:
up_write(&c->state_lock);
- bch_err(c, "Error going rw after adding device: %s", err);
return -EINVAL;
}
diff --git a/libbcachefs/tests.c b/libbcachefs/tests.c
index 60ccb94e..16d67eb6 100644
--- a/libbcachefs/tests.c
+++ b/libbcachefs/tests.c
@@ -14,14 +14,14 @@ static void delete_test_keys(struct bch_fs *c)
int ret;
ret = bch2_btree_delete_range(c, BTREE_ID_extents,
- SPOS(0, 0, U32_MAX),
- SPOS(0, U64_MAX, U32_MAX),
+ POS_MIN, SPOS_MAX,
+ BTREE_ITER_ALL_SNAPSHOTS,
NULL);
BUG_ON(ret);
ret = bch2_btree_delete_range(c, BTREE_ID_xattrs,
- SPOS(0, 0, U32_MAX),
- SPOS(0, U64_MAX, U32_MAX),
+ POS_MIN, SPOS_MAX,
+ BTREE_ITER_ALL_SNAPSHOTS,
NULL);
BUG_ON(ret);
}
@@ -146,7 +146,7 @@ static int test_iterate(struct bch_fs *c, u64 nr)
i = 0;
for_each_btree_key(&trans, iter, BTREE_ID_xattrs,
- POS_MIN, 0, k, ret) {
+ SPOS(0, 0, U32_MAX), 0, k, ret) {
if (k.k->p.inode)
break;
@@ -202,7 +202,7 @@ static int test_iterate_extents(struct bch_fs *c, u64 nr)
i = 0;
for_each_btree_key(&trans, iter, BTREE_ID_extents,
- POS_MIN, 0, k, ret) {
+ SPOS(0, 0, U32_MAX), 0, k, ret) {
BUG_ON(bkey_start_offset(k.k) != i);
i = k.k->p.offset;
}
@@ -256,8 +256,8 @@ static int test_iterate_slots(struct bch_fs *c, u64 nr)
i = 0;
- for_each_btree_key(&trans, iter, BTREE_ID_xattrs, POS_MIN,
- 0, k, ret) {
+ for_each_btree_key(&trans, iter, BTREE_ID_xattrs,
+ SPOS(0, 0, U32_MAX), 0, k, ret) {
if (k.k->p.inode)
break;
@@ -272,7 +272,8 @@ static int test_iterate_slots(struct bch_fs *c, u64 nr)
i = 0;
- for_each_btree_key(&trans, iter, BTREE_ID_xattrs, POS_MIN,
+ for_each_btree_key(&trans, iter, BTREE_ID_xattrs,
+ SPOS(0, 0, U32_MAX),
BTREE_ITER_SLOTS, k, ret) {
BUG_ON(k.k->p.offset != i);
BUG_ON(bkey_deleted(k.k) != (i & 1));
@@ -321,8 +322,8 @@ static int test_iterate_slots_extents(struct bch_fs *c, u64 nr)
i = 0;
- for_each_btree_key(&trans, iter, BTREE_ID_extents, POS_MIN,
- 0, k, ret) {
+ for_each_btree_key(&trans, iter, BTREE_ID_extents,
+ SPOS(0, 0, U32_MAX), 0, k, ret) {
BUG_ON(bkey_start_offset(k.k) != i + 8);
BUG_ON(k.k->size != 8);
i += 16;
@@ -335,7 +336,8 @@ static int test_iterate_slots_extents(struct bch_fs *c, u64 nr)
i = 0;
- for_each_btree_key(&trans, iter, BTREE_ID_extents, POS_MIN,
+ for_each_btree_key(&trans, iter, BTREE_ID_extents,
+ SPOS(0, 0, U32_MAX),
BTREE_ITER_SLOTS, k, ret) {
BUG_ON(bkey_deleted(k.k) != !(i % 16));
@@ -363,7 +365,8 @@ static int test_peek_end(struct bch_fs *c, u64 nr)
struct bkey_s_c k;
bch2_trans_init(&trans, c, 0, 0);
- bch2_trans_iter_init(&trans, &iter, BTREE_ID_xattrs, POS_MIN, 0);
+ bch2_trans_iter_init(&trans, &iter, BTREE_ID_xattrs,
+ SPOS(0, 0, U32_MAX), 0);
k = bch2_btree_iter_peek(&iter);
BUG_ON(k.k);
@@ -383,7 +386,8 @@ static int test_peek_end_extents(struct bch_fs *c, u64 nr)
struct bkey_s_c k;
bch2_trans_init(&trans, c, 0, 0);
- bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents, POS_MIN, 0);
+ bch2_trans_iter_init(&trans, &iter, BTREE_ID_extents,
+ SPOS(0, 0, U32_MAX), 0);
k = bch2_btree_iter_peek(&iter);
BUG_ON(k.k);
@@ -406,8 +410,6 @@ static int insert_test_extent(struct bch_fs *c,
struct bkey_i_cookie k;
int ret;
- //pr_info("inserting %llu-%llu v %llu", start, end, test_version);
-
bkey_cookie_init(&k.k_i);
k.k_i.k.p.offset = end;
k.k_i.k.p.snapshot = U32_MAX;
@@ -747,7 +749,9 @@ static int seq_delete(struct bch_fs *c, u64 nr)
int ret;
ret = bch2_btree_delete_range(c, BTREE_ID_xattrs,
- SPOS(0, 0, U32_MAX), POS_MAX, NULL);
+ POS_MIN, SPOS_MAX,
+ BTREE_ITER_ALL_SNAPSHOTS,
+ NULL);
if (ret)
bch_err(c, "error in seq_delete: %i", ret);
return ret;
diff --git a/libbcachefs/util.c b/libbcachefs/util.c
index 52de7c49..0bbea332 100644
--- a/libbcachefs/util.c
+++ b/libbcachefs/util.c
@@ -114,7 +114,7 @@ void bch2_hprint(struct printbuf *buf, s64 v)
* 103 is magic: t is in the range [-1023, 1023] and we want
* to turn it into [-9, 9]
*/
- if (u && v < 100 && v > -100)
+ if (u && t && v < 100 && v > -100)
pr_buf(buf, ".%i", t / 103);
if (u)
pr_buf(buf, "%c", si_units[u]);