From 5aaa6422b6fdc9cda8f808f7152df2073626b1bd Mon Sep 17 00:00:00 2001
From: Kent Overstreet <kent.overstreet@linux.dev>
Date: Mon, 15 Apr 2024 14:10:51 -0400
Subject: [PATCH] Update bcachefs sources to ad29cf999a91 bcachefs:
set_btree_iter_dontneed also clears should_be_locked
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
---
.bcachefs_revision | 2 +-
libbcachefs/acl.c | 32 ++---
libbcachefs/backpointers.c | 8 +-
libbcachefs/backpointers.h | 9 +-
libbcachefs/bcachefs.h | 2 +
libbcachefs/bcachefs_format.h | 25 +++-
libbcachefs/bkey.h | 6 +
libbcachefs/bkey_methods.c | 8 +-
libbcachefs/btree_cache.c | 75 ++++++----
libbcachefs/btree_gc.c | 26 +++-
libbcachefs/btree_io.c | 19 +--
libbcachefs/btree_iter.h | 11 +-
libbcachefs/btree_journal_iter.c | 67 +++++++--
libbcachefs/btree_key_cache.c | 4 +-
libbcachefs/btree_locking.c | 28 +---
libbcachefs/btree_node_scan.c | 30 +++-
libbcachefs/btree_trans_commit.c | 34 ++---
libbcachefs/btree_types.h | 14 ++
libbcachefs/btree_update_interior.c | 209 +++++++++++++++++-----------
libbcachefs/btree_update_interior.h | 3 +-
libbcachefs/btree_write_buffer.c | 14 +-
libbcachefs/buckets.h | 8 --
libbcachefs/chardev.c | 98 +++++++------
libbcachefs/checksum.c | 23 +--
libbcachefs/checksum.h | 5 +-
libbcachefs/compress.h | 8 --
libbcachefs/data_update.c | 17 ++-
libbcachefs/debug.c | 75 +++++++---
libbcachefs/ec.c | 50 +++----
libbcachefs/ec.h | 2 +
libbcachefs/extents.c | 11 +-
libbcachefs/eytzinger.c | 8 +-
libbcachefs/eytzinger.h | 26 +++-
libbcachefs/fs-io-direct.c | 19 ++-
libbcachefs/fs-io.c | 16 +--
libbcachefs/journal_io.c | 17 +--
libbcachefs/journal_reclaim.c | 2 +
libbcachefs/journal_types.h | 1 +
libbcachefs/opts.c | 29 +++-
libbcachefs/opts.h | 10 +-
libbcachefs/recovery.c | 14 --
libbcachefs/recovery_passes.c | 2 +-
libbcachefs/sb-downgrade.c | 5 +-
libbcachefs/sb-errors_types.h | 5 +-
libbcachefs/sb-members.c | 53 +++++++
libbcachefs/sb-members.h | 21 +++
libbcachefs/snapshot.c | 19 ++-
libbcachefs/super-io.c | 7 +-
libbcachefs/super.c | 5 +
libbcachefs/super_types.h | 2 +
libbcachefs/sysfs.c | 17 ++-
libbcachefs/tests.c | 2 +-
libbcachefs/util.h | 10 +-
53 files changed, 792 insertions(+), 421 deletions(-)
diff --git a/.bcachefs_revision b/.bcachefs_revision
index 0370dc87..cd9472dd 100644
--- a/.bcachefs_revision
+++ b/.bcachefs_revision
@@ -1 +1 @@
-09d4c2acbf4c864fef0f520bbcba256c9a19102e
+ad29cf999a9161e7849aa229d2028854f90728c2
diff --git a/libbcachefs/acl.c b/libbcachefs/acl.c
index 3640f417..5c180fdc 100644
--- a/libbcachefs/acl.c
+++ b/libbcachefs/acl.c
@@ -281,7 +281,6 @@ struct posix_acl *bch2_get_acl(struct mnt_idmap *idmap,
struct xattr_search_key search = X_SEARCH(acl_to_xattr_type(type), "", 0);
struct btree_trans *trans = bch2_trans_get(c);
struct btree_iter iter = { NULL };
- struct bkey_s_c_xattr xattr;
struct posix_acl *acl = NULL;
struct bkey_s_c k;
int ret;
@@ -290,29 +289,28 @@ retry:
ret = bch2_hash_lookup(trans, &iter, bch2_xattr_hash_desc,
&hash, inode_inum(inode), &search, 0);
- if (ret) {
- if (!bch2_err_matches(ret, ENOENT))
- acl = ERR_PTR(ret);
- goto out;
- }
+ if (ret)
+ goto err;
k = bch2_btree_iter_peek_slot(&iter);
ret = bkey_err(k);
- if (ret) {
- acl = ERR_PTR(ret);
- goto out;
- }
+ if (ret)
+ goto err;
- xattr = bkey_s_c_to_xattr(k);
+ struct bkey_s_c_xattr xattr = bkey_s_c_to_xattr(k);
acl = bch2_acl_from_disk(trans, xattr_val(xattr.v),
- le16_to_cpu(xattr.v->x_val_len));
-
- if (!IS_ERR(acl))
- set_cached_acl(&inode->v, type, acl);
-out:
- if (bch2_err_matches(PTR_ERR_OR_ZERO(acl), BCH_ERR_transaction_restart))
+ le16_to_cpu(xattr.v->x_val_len));
+ ret = PTR_ERR_OR_ZERO(acl);
+err:
+ if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
goto retry;
+ if (ret)
+ acl = !bch2_err_matches(ret, ENOENT) ? ERR_PTR(ret) : NULL;
+
+ if (!IS_ERR_OR_NULL(acl))
+ set_cached_acl(&inode->v, type, acl);
+
bch2_trans_iter_exit(trans, &iter);
bch2_trans_put(trans);
return acl;
diff --git a/libbcachefs/backpointers.c b/libbcachefs/backpointers.c
index 114328ac..fadb1078 100644
--- a/libbcachefs/backpointers.c
+++ b/libbcachefs/backpointers.c
@@ -49,13 +49,15 @@ int bch2_backpointer_invalid(struct bch_fs *c, struct bkey_s_c k,
if (!bch2_dev_exists2(c, bp.k->p.inode))
return 0;
+ struct bch_dev *ca = bch_dev_bkey_exists(c, bp.k->p.inode);
struct bpos bucket = bp_pos_to_bucket(c, bp.k->p);
int ret = 0;
- bkey_fsck_err_on(!bpos_eq(bp.k->p, bucket_pos_to_bp(c, bucket, bp.v->bucket_offset)),
+ bkey_fsck_err_on((bp.v->bucket_offset >> MAX_EXTENT_COMPRESS_RATIO_SHIFT) >= ca->mi.bucket_size ||
+ !bpos_eq(bp.k->p, bucket_pos_to_bp(c, bucket, bp.v->bucket_offset)),
c, err,
- backpointer_pos_wrong,
- "backpointer at wrong pos");
+ backpointer_bucket_offset_wrong,
+ "backpointer bucket_offset wrong");
fsck_err:
return ret;
}
diff --git a/libbcachefs/backpointers.h b/libbcachefs/backpointers.h
index da012ca7..85949b9f 100644
--- a/libbcachefs/backpointers.h
+++ b/libbcachefs/backpointers.h
@@ -53,14 +53,11 @@ static inline struct bpos bucket_pos_to_bp(const struct bch_fs *c,
u64 bucket_offset)
{
struct bch_dev *ca = bch_dev_bkey_exists(c, bucket.inode);
- struct bpos ret;
-
- ret = POS(bucket.inode,
- (bucket_to_sector(ca, bucket.offset) <<
- MAX_EXTENT_COMPRESS_RATIO_SHIFT) + bucket_offset);
+ struct bpos ret = POS(bucket.inode,
+ (bucket_to_sector(ca, bucket.offset) <<
+ MAX_EXTENT_COMPRESS_RATIO_SHIFT) + bucket_offset);
EBUG_ON(!bkey_eq(bucket, bp_pos_to_bucket(c, ret)));
-
return ret;
}
diff --git a/libbcachefs/bcachefs.h b/libbcachefs/bcachefs.h
index a31a5f70..91c3c1fe 100644
--- a/libbcachefs/bcachefs.h
+++ b/libbcachefs/bcachefs.h
@@ -709,6 +709,8 @@ struct btree_trans_buf {
x(stripe_delete) \
x(reflink) \
x(fallocate) \
+ x(fsync) \
+ x(dio_write) \
x(discard) \
x(discard_fast) \
x(invalidate) \
diff --git a/libbcachefs/bcachefs_format.h b/libbcachefs/bcachefs_format.h
index 63102992..08598743 100644
--- a/libbcachefs/bcachefs_format.h
+++ b/libbcachefs/bcachefs_format.h
@@ -578,7 +578,8 @@ struct bch_member {
__le64 nbuckets; /* device size */
__le16 first_bucket; /* index of first bucket used */
__le16 bucket_size; /* sectors */
- __le32 pad;
+ __u8 btree_bitmap_shift;
+ __u8 pad[3];
__le64 last_mount; /* time_t */
__le64 flags;
@@ -587,6 +588,7 @@ struct bch_member {
__le64 errors_at_reset[BCH_MEMBER_ERROR_NR];
__le64 errors_reset_time;
__le64 seq;
+ __le64 btree_allocated_bitmap;
};
#define BCH_MEMBER_V1_BYTES 56
@@ -876,7 +878,8 @@ struct bch_sb_field_downgrade {
x(rebalance_work, BCH_VERSION(1, 3)) \
x(member_seq, BCH_VERSION(1, 4)) \
x(subvolume_fs_parent, BCH_VERSION(1, 5)) \
- x(btree_subvolume_children, BCH_VERSION(1, 6))
+ x(btree_subvolume_children, BCH_VERSION(1, 6)) \
+ x(mi_btree_bitmap, BCH_VERSION(1, 7))
enum bcachefs_metadata_version {
bcachefs_metadata_version_min = 9,
@@ -1314,7 +1317,7 @@ static inline __u64 __bset_magic(struct bch_sb *sb)
x(write_buffer_keys, 11) \
x(datetime, 12)
-enum {
+enum bch_jset_entry_type {
#define x(f, nr) BCH_JSET_ENTRY_##f = nr,
BCH_JSET_ENTRY_TYPES()
#undef x
@@ -1360,7 +1363,7 @@ struct jset_entry_blacklist_v2 {
x(inodes, 1) \
x(key_version, 2)
-enum {
+enum bch_fs_usage_type {
#define x(f, nr) BCH_FS_USAGE_##f = nr,
BCH_FS_USAGE_TYPES()
#undef x
@@ -1535,6 +1538,20 @@ enum btree_id {
BTREE_ID_NR
};
+static inline 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:
+ case BTREE_ID_bucket_gens:
+ return true;
+ default:
+ return false;
+ }
+}
+
#define BTREE_MAX_DEPTH 4U
/* Btree nodes */
diff --git a/libbcachefs/bkey.h b/libbcachefs/bkey.h
index cf23ff47..3a45d128 100644
--- a/libbcachefs/bkey.h
+++ b/libbcachefs/bkey.h
@@ -314,6 +314,12 @@ static inline unsigned bkeyp_key_u64s(const struct bkey_format *format,
return bkey_packed(k) ? format->key_u64s : BKEY_U64s;
}
+static inline bool bkeyp_u64s_valid(const struct bkey_format *f,
+ const struct bkey_packed *k)
+{
+ return ((unsigned) k->u64s - bkeyp_key_u64s(f, k) <= U8_MAX - BKEY_U64s);
+}
+
static inline unsigned bkeyp_key_bytes(const struct bkey_format *format,
const struct bkey_packed *k)
{
diff --git a/libbcachefs/bkey_methods.c b/libbcachefs/bkey_methods.c
index 5e526847..db336a43 100644
--- a/libbcachefs/bkey_methods.c
+++ b/libbcachefs/bkey_methods.c
@@ -171,11 +171,15 @@ int __bch2_bkey_invalid(struct bch_fs *c, struct bkey_s_c k,
if (type >= BKEY_TYPE_NR)
return 0;
- bkey_fsck_err_on((flags & BKEY_INVALID_COMMIT) &&
+ bkey_fsck_err_on((type == BKEY_TYPE_btree ||
+ (flags & BKEY_INVALID_COMMIT)) &&
!(bch2_key_types_allowed[type] & BIT_ULL(k.k->type)), c, err,
bkey_invalid_type_for_btree,
"invalid key type for btree %s (%s)",
- bch2_btree_node_type_str(type), bch2_bkey_types[k.k->type]);
+ bch2_btree_node_type_str(type),
+ k.k->type < KEY_TYPE_MAX
+ ? bch2_bkey_types[k.k->type]
+ : "(unknown)");
if (btree_node_type_is_extents(type) && !bkey_whiteout(k.k)) {
bkey_fsck_err_on(k.k->size == 0, c, err,
diff --git a/libbcachefs/btree_cache.c b/libbcachefs/btree_cache.c
index 84474324..02c70e81 100644
--- a/libbcachefs/btree_cache.c
+++ b/libbcachefs/btree_cache.c
@@ -709,9 +709,31 @@ static noinline struct btree *bch2_btree_node_fill(struct btree_trans *trans,
struct bch_fs *c = trans->c;
struct btree_cache *bc = &c->btree_cache;
struct btree *b;
- u32 seq;
- BUG_ON(level + 1 >= BTREE_MAX_DEPTH);
+ if (unlikely(level >= BTREE_MAX_DEPTH)) {
+ int ret = bch2_fs_topology_error(c, "attempting to get btree node at level %u, >= max depth %u",
+ level, BTREE_MAX_DEPTH);
+ return ERR_PTR(ret);
+ }
+
+ if (unlikely(!bkey_is_btree_ptr(&k->k))) {
+ struct printbuf buf = PRINTBUF;
+ bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(k));
+
+ int ret = bch2_fs_topology_error(c, "attempting to get btree node with non-btree key %s", buf.buf);
+ printbuf_exit(&buf);
+ return ERR_PTR(ret);
+ }
+
+ if (unlikely(k->k.u64s > BKEY_BTREE_PTR_U64s_MAX)) {
+ struct printbuf buf = PRINTBUF;
+ bch2_bkey_val_to_text(&buf, c, bkey_i_to_s_c(k));
+
+ int ret = bch2_fs_topology_error(c, "attempting to get btree node with too big key %s", buf.buf);
+ printbuf_exit(&buf);
+ return ERR_PTR(ret);
+ }
+
/*
* Parent node must be locked, else we could read in a btree node that's
* been freed:
@@ -752,34 +774,26 @@ static noinline struct btree *bch2_btree_node_fill(struct btree_trans *trans,
}
set_btree_node_read_in_flight(b);
-
six_unlock_write(&b->c.lock);
- seq = six_lock_seq(&b->c.lock);
- six_unlock_intent(&b->c.lock);
-
- /* Unlock before doing IO: */
- if (path && sync)
- bch2_trans_unlock_noassert(trans);
-
- bch2_btree_node_read(trans, b, sync);
-
- if (!sync)
- return NULL;
if (path) {
- int ret = bch2_trans_relock(trans) ?:
- bch2_btree_path_relock_intent(trans, path);
- if (ret) {
- BUG_ON(!trans->restarted);
- return ERR_PTR(ret);
- }
- }
+ u32 seq = six_lock_seq(&b->c.lock);
- if (!six_relock_type(&b->c.lock, lock_type, seq)) {
- BUG_ON(!path);
+ /* Unlock before doing IO: */
+ six_unlock_intent(&b->c.lock);
+ bch2_trans_unlock_noassert(trans);
- trace_and_count(c, trans_restart_relock_after_fill, trans, _THIS_IP_, path);
- return ERR_PTR(btree_trans_restart(trans, BCH_ERR_transaction_restart_relock_after_fill));
+ bch2_btree_node_read(trans, b, sync);
+
+ if (!sync)
+ return NULL;
+
+ if (!six_relock_type(&b->c.lock, lock_type, seq))
+ b = NULL;
+ } else {
+ bch2_btree_node_read(trans, b, sync);
+ if (lock_type == SIX_LOCK_read)
+ six_lock_downgrade(&b->c.lock);
}
return b;
@@ -1112,18 +1126,19 @@ int bch2_btree_node_prefetch(struct btree_trans *trans,
{
struct bch_fs *c = trans->c;
struct btree_cache *bc = &c->btree_cache;
- struct btree *b;
BUG_ON(path && !btree_node_locked(path, level + 1));
BUG_ON(level >= BTREE_MAX_DEPTH);
- b = btree_cache_find(bc, k);
+ struct btree *b = btree_cache_find(bc, k);
if (b)
return 0;
b = bch2_btree_node_fill(trans, path, k, btree_id,
level, SIX_LOCK_read, false);
- return PTR_ERR_OR_ZERO(b);
+ if (!IS_ERR_OR_NULL(b))
+ six_unlock_read(&b->c.lock);
+ return bch2_trans_relock(trans) ?: PTR_ERR_OR_ZERO(b);
}
void bch2_btree_node_evict(struct btree_trans *trans, const struct bkey_i *k)
@@ -1148,6 +1163,8 @@ wait_on_io:
btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_intent);
btree_node_lock_nopath_nofail(trans, &b->c, SIX_LOCK_write);
+ if (unlikely(b->hash_val != btree_ptr_hash_val(k)))
+ goto out;
if (btree_node_dirty(b)) {
__bch2_btree_node_write(c, b, BTREE_WRITE_cache_reclaim);
@@ -1162,7 +1179,7 @@ wait_on_io:
btree_node_data_free(c, b);
bch2_btree_node_hash_remove(bc, b);
mutex_unlock(&bc->lock);
-
+out:
six_unlock_write(&b->c.lock);
six_unlock_intent(&b->c.lock);
}
diff --git a/libbcachefs/btree_gc.c b/libbcachefs/btree_gc.c
index 6280da12..ecbd9598 100644
--- a/libbcachefs/btree_gc.c
+++ b/libbcachefs/btree_gc.c
@@ -368,11 +368,16 @@ again:
buf.buf)) {
bch2_btree_node_evict(trans, cur_k.k);
cur = NULL;
- ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_scan_for_btree_nodes) ?:
- bch2_journal_key_delete(c, b->c.btree_id,
- b->c.level, cur_k.k->k.p);
+ ret = bch2_journal_key_delete(c, b->c.btree_id,
+ b->c.level, cur_k.k->k.p);
if (ret)
break;
+
+ if (!btree_id_is_alloc(b->c.btree_id)) {
+ ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_scan_for_btree_nodes);
+ if (ret)
+ break;
+ }
continue;
}
@@ -544,12 +549,12 @@ reconstruct_root:
bch2_btree_root_alloc_fake(c, i, 0);
} else {
bch2_btree_root_alloc_fake(c, i, 1);
+ bch2_shoot_down_journal_keys(c, i, 1, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX);
ret = bch2_get_scanned_nodes(c, i, 0, POS_MIN, SPOS_MAX);
if (ret)
break;
}
- bch2_shoot_down_journal_keys(c, i, 1, BTREE_MAX_DEPTH, POS_MIN, SPOS_MAX);
reconstructed_root = true;
}
@@ -823,6 +828,7 @@ static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id,
struct bch_fs *c = trans->c;
struct bkey deleted = KEY(0, 0, 0);
struct bkey_s_c old = (struct bkey_s_c) { &deleted, NULL };
+ struct printbuf buf = PRINTBUF;
int ret = 0;
deleted.p = k->k->p;
@@ -843,11 +849,23 @@ static int bch2_gc_mark_key(struct btree_trans *trans, enum btree_id btree_id,
if (ret)
goto err;
+ if (mustfix_fsck_err_on(level && !bch2_dev_btree_bitmap_marked(c, *k),
+ c, btree_bitmap_not_marked,
+ "btree ptr not marked in member info btree allocated bitmap\n %s",
+ (bch2_bkey_val_to_text(&buf, c, *k),
+ buf.buf))) {
+ mutex_lock(&c->sb_lock);
+ bch2_dev_btree_bitmap_mark(c, *k);
+ bch2_write_super(c);
+ mutex_unlock(&c->sb_lock);
+ }
+
ret = commit_do(trans, NULL, NULL, 0,
bch2_key_trigger(trans, btree_id, level, old,
unsafe_bkey_s_c_to_s(*k), BTREE_TRIGGER_GC));
fsck_err:
err:
+ printbuf_exit(&buf);
bch_err_fn(c, ret);
return ret;
}
diff --git a/libbcachefs/btree_io.c b/libbcachefs/btree_io.c
index d7de82ac..9678b237 100644
--- a/libbcachefs/btree_io.c
+++ b/libbcachefs/btree_io.c
@@ -831,7 +831,7 @@ static int bset_key_invalid(struct bch_fs *c, struct btree *b,
(rw == WRITE ? bch2_bkey_val_invalid(c, k, READ, err) : 0);
}
-static bool __bkey_valid(struct bch_fs *c, struct btree *b,
+static bool bkey_packed_valid(struct bch_fs *c, struct btree *b,
struct bset *i, struct bkey_packed *k)
{
if (bkey_p_next(k) > vstruct_last(i))
@@ -840,7 +840,7 @@ static bool __bkey_valid(struct bch_fs *c, struct btree *b,
if (k->format > KEY_FORMAT_CURRENT)
return false;
- if (k->u64s < bkeyp_key_u64s(&b->format, k))
+ if (!bkeyp_u64s_valid(&b->format, k))
return false;
struct printbuf buf = PRINTBUF;
@@ -884,11 +884,13 @@ static int validate_bset_keys(struct bch_fs *c, struct btree *b,
"invalid bkey format %u", k->format))
goto drop_this_key;
- if (btree_err_on(k->u64s < bkeyp_key_u64s(&b->format, k),
+ if (btree_err_on(!bkeyp_u64s_valid(&b->format, k),
-BCH_ERR_btree_node_read_err_fixable,
c, NULL, b, i,
btree_node_bkey_bad_u64s,
- "k->u64s too small (%u < %u)", k->u64s, bkeyp_key_u64s(&b->format, k)))
+ "bad k->u64s %u (min %u max %lu)", k->u64s,
+ bkeyp_key_u64s(&b->format, k),
+ U8_MAX - BKEY_U64s + bkeyp_key_u64s(&b->format, k)))
goto drop_this_key;
if (!write)
@@ -947,13 +949,12 @@ drop_this_key:
* do
*/
- if (!__bkey_valid(c, b, i, (void *) ((u64 *) k + next_good_key))) {
+ if (!bkey_packed_valid(c, b, i, (void *) ((u64 *) k + next_good_key))) {
for (next_good_key = 1;
next_good_key < (u64 *) vstruct_last(i) - (u64 *) k;
next_good_key++)
- if (__bkey_valid(c, b, i, (void *) ((u64 *) k + next_good_key)))
+ if (bkey_packed_valid(c, b, i, (void *) ((u64 *) k + next_good_key)))
goto got_good_key;
-
}
/*
@@ -1339,7 +1340,9 @@ start:
rb->start_time);
bio_put(&rb->bio);
- if (saw_error && !btree_node_read_error(b)) {
+ if (saw_error &&
+ !btree_node_read_error(b) &&
+ c->curr_recovery_pass != BCH_RECOVERY_PASS_scan_for_btree_nodes) {
printbuf_reset(&buf);
bch2_bpos_to_text(&buf, b->key.k.p);
bch_err_ratelimited(c, "%s: rewriting btree node at btree=%s level=%u %s due to error",
diff --git a/libbcachefs/btree_iter.h b/libbcachefs/btree_iter.h
index 24772538..1c70836d 100644
--- a/libbcachefs/btree_iter.h
+++ b/libbcachefs/btree_iter.h
@@ -498,8 +498,13 @@ static inline void set_btree_iter_dontneed(struct btree_iter *iter)
{
struct btree_trans *trans = iter->trans;
- if (!trans->restarted)
- btree_iter_path(trans, iter)->preserve = false;
+ if (!iter->path || trans->restarted)
+ return;
+
+ struct btree_path *path = btree_iter_path(trans, iter);
+ path->preserve = false;
+ if (path->ref == 1)
+ path->should_be_locked = false;
}
void *__bch2_trans_kmalloc(struct btree_trans *, size_t);
@@ -642,7 +647,7 @@ int __bch2_btree_trans_too_many_iters(struct btree_trans *);
static inline int btree_trans_too_many_iters(struct btree_trans *trans)
{
- if (bitmap_weight(trans->paths_allocated, trans->nr_paths) > BTREE_ITER_INITIAL - 8)
+ if (bitmap_weight(trans->paths_allocated, trans->nr_paths) > BTREE_ITER_NORMAL_LIMIT - 8)
return __bch2_btree_trans_too_many_iters(trans);
return 0;
diff --git a/libbcachefs/btree_journal_iter.c b/libbcachefs/btree_journal_iter.c
index 5cbcbfe8..1e8cf49a 100644
--- a/libbcachefs/btree_journal_iter.c
+++ b/libbcachefs/btree_journal_iter.c
@@ -130,12 +130,30 @@ struct bkey_i *bch2_journal_keys_peek_slot(struct bch_fs *c, enum btree_id btree
return bch2_journal_keys_peek_upto(c, btree_id, level, pos, pos, &idx);
}
+static void journal_iter_verify(struct journal_iter *iter)
+{
+ struct journal_keys *keys = iter->keys;
+ size_t gap_size = keys->size - keys->nr;
+
+ BUG_ON(iter->idx >= keys->gap &&
+ iter->idx < keys->gap + gap_size);
+
+ if (iter->idx < keys->size) {
+ struct journal_key *k = keys->data + iter->idx;
+
+ int cmp = cmp_int(k->btree_id, iter->btree_id) ?:
+ cmp_int(k->level, iter->level);
+ BUG_ON(cmp < 0);
+ }
+}
+
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;
+ struct journal_key *new_key = &keys->data[keys->gap - 1];
+ struct journal_iter *iter;
/*
* If an iterator points one after the key we just inserted, decrement
@@ -143,9 +161,14 @@ static void journal_iters_fix(struct bch_fs *c)
* 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;
+ list_for_each_entry(iter, &c->journal_iters, list) {
+ journal_iter_verify(iter);
+ if (iter->idx == gap_end &&
+ new_key->btree_id == iter->btree_id &&
+ new_key->level == iter->level)
+ iter->idx = keys->gap - 1;
+ journal_iter_verify(iter);
+ }
}
static void journal_iters_move_gap(struct bch_fs *c, size_t old_gap, size_t new_gap)
@@ -192,7 +215,12 @@ int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id,
if (idx > keys->gap)
idx -= keys->size - keys->nr;
+ size_t old_gap = keys->gap;
+
if (keys->nr == keys->size) {
+ journal_iters_move_gap(c, old_gap, keys->size);
+ old_gap = keys->size;
+
struct journal_keys new_keys = {
.nr = keys->nr,
.size = max_t(size_t, keys->size, 8) * 2,
@@ -216,7 +244,7 @@ int bch2_journal_key_insert_take(struct bch_fs *c, enum btree_id id,
keys->gap = keys->nr;
}
- journal_iters_move_gap(c, keys->gap, idx);
+ journal_iters_move_gap(c, old_gap, idx);
move_gap(keys, idx);
@@ -301,16 +329,21 @@ static void bch2_journal_iter_advance(struct journal_iter *iter)
static struct bkey_s_c bch2_journal_iter_peek(struct journal_iter *iter)
{
- struct journal_key *k = iter->keys->data + iter->idx;
+ journal_iter_verify(iter);
+
+ while (iter->idx < iter->keys->size) {
+ struct journal_key *k = iter->keys->data + iter->idx;
+
+ int cmp = cmp_int(k->btree_id, iter->btree_id) ?:
+ cmp_int(k->level, iter->level);
+ if (cmp > 0)
+ break;
+ BUG_ON(cmp);
- while (k < iter->keys->data + iter->keys->size &&
- k->btree_id == iter->btree_id &&
- k->level == iter->level) {
if (!k->overwritten)
return bkey_i_to_s_c(k->k);
bch2_journal_iter_advance(iter);
- k = iter->keys->data + iter->idx;
}
return bkey_s_c_null;
@@ -330,6 +363,8 @@ static void bch2_journal_iter_init(struct bch_fs *c,
iter->level = level;
iter->keys = &c->journal_keys;
iter->idx = bch2_journal_key_search(&c->journal_keys, id, level, pos);
+
+ journal_iter_verify(iter);
}
static struct bkey_s_c bch2_journal_iter_peek_btree(struct btree_and_journal_iter *iter)
@@ -434,10 +469,15 @@ void __bch2_btree_and_journal_iter_init_node_iter(struct btree_trans *trans,
iter->trans = trans;
iter->b = b;
iter->node_iter = node_iter;
- bch2_journal_iter_init(trans->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;
+ INIT_LIST_HEAD(&iter->journal.list);
+
+ if (trans->journal_replay_not_finished) {
+ bch2_journal_iter_init(trans->c, &iter->journal, b->c.btree_id, b->c.level, pos);
+ if (!test_bit(BCH_FS_may_go_rw, &trans->c->flags))
+ list_add(&iter->journal.list, &trans->c->journal_iters);
+ }
}
/*
@@ -452,9 +492,6 @@ void bch2_btree_and_journal_iter_init_node_iter(struct btree_trans *trans,
bch2_btree_node_iter_init_from_start(&node_iter, b);
__bch2_btree_and_journal_iter_init_node_iter(trans, iter, b, node_iter, b->data->min_key);
- if (trans->journal_replay_not_finished &&
- !test_bit(BCH_FS_may_go_rw, &trans->c->flags))
- list_add(&iter->journal.list, &trans->c->journal_iters);
}
/* sort and dedup all keys in the journal: */
diff --git a/libbcachefs/btree_key_cache.c b/libbcachefs/btree_key_cache.c
index 581edcb0..88a3582a 100644
--- a/libbcachefs/btree_key_cache.c
+++ b/libbcachefs/btree_key_cache.c
@@ -169,6 +169,7 @@ static void bkey_cached_move_to_freelist(struct btree_key_cache *bc,
} else {
mutex_lock(&bc->lock);
list_move_tail(&ck->list, &bc->freed_pcpu);
+ bc->nr_freed_pcpu++;
mutex_unlock(&bc->lock);
}
}
@@ -245,6 +246,7 @@ bkey_cached_alloc(struct btree_trans *trans, struct btree_path *path,
if (!list_empty(&bc->freed_pcpu)) {
ck = list_last_entry(&bc->freed_pcpu, struct bkey_cached, list);
list_del_init(&ck->list);
+ bc->nr_freed_pcpu--;
}
mutex_unlock(&bc->lock);
}
@@ -659,7 +661,7 @@ static int btree_key_cache_flush_pos(struct btree_trans *trans,
commit_flags |= BCH_WATERMARK_reclaim;
if (ck->journal.seq != journal_last_seq(j) ||
- j->watermark == BCH_WATERMARK_stripe)
+ !test_bit(JOURNAL_SPACE_LOW, &c->journal.flags))
commit_flags |= BCH_TRANS_COMMIT_no_journal_res;
ret = bch2_btree_iter_traverse(&b_iter) ?:
diff --git a/libbcachefs/btree_locking.c b/libbcachefs/btree_locking.c
index b9b151e6..f2caf491 100644
--- a/libbcachefs/btree_locking.c
+++ b/libbcachefs/btree_locking.c
@@ -440,33 +440,7 @@ void bch2_btree_node_lock_write_nofail(struct btree_trans *trans,
struct btree_path *path,
struct btree_bkey_cached_common *b)
{
- struct btree_path *linked;
- unsigned i, iter;
- int ret;
-
- /*
- * XXX BIG FAT NOTICE
- *
- * Drop all read locks before taking a write lock:
- *
- * This is a hack, because bch2_btree_node_lock_write_nofail() is a
- * hack - but by dropping read locks first, this should never fail, and
- * we only use this in code paths where whatever read locks we've
- * already taken are no longer needed:
- */
-
- trans_for_each_path(trans, linked, iter) {
- if (!linked->nodes_locked)
- continue;
-
- for (i = 0; i < BTREE_MAX_DEPTH; i++)
- if (btree_node_read_locked(linked, i)) {
- btree_node_unlock(trans, linked, i);
- btree_path_set_dirty(linked, BTREE_ITER_NEED_RELOCK);
- }
- }
-
- ret = __btree_node_lock_write(trans, path, b, true);
+ int ret = __btree_node_lock_write(trans, path, b, true);
BUG_ON(ret);
}
diff --git a/libbcachefs/btree_node_scan.c b/libbcachefs/btree_node_scan.c
index 3f33be7e..866bd278 100644
--- a/libbcachefs/btree_node_scan.c
+++ b/libbcachefs/btree_node_scan.c
@@ -133,6 +133,19 @@ static void try_read_btree_node(struct find_btree_nodes *f, struct bch_dev *ca,
if (le64_to_cpu(bn->magic) != bset_magic(c))
return;
+ if (bch2_csum_type_is_encryption(BSET_CSUM_TYPE(&bn->keys))) {
+ struct nonce nonce = btree_nonce(&bn->keys, 0);
+ unsigned bytes = (void *) &bn->keys - (void *) &bn->flags;
+
+ bch2_encrypt(c, BSET_CSUM_TYPE(&bn->keys), nonce, &bn->flags, bytes);
+ }
+
+ if (btree_id_is_alloc(BTREE_NODE_ID(bn)))
+ return;
+
+ if (BTREE_NODE_LEVEL(bn) >= BTREE_MAX_DEPTH)
+ return;
+
rcu_read_lock();
struct found_btree_node n = {
.btree_id = BTREE_NODE_ID(bn),
@@ -192,8 +205,13 @@ static int read_btree_nodes_worker(void *p)
last_print = jiffies;
}
- try_read_btree_node(w->f, ca, bio, buf,
- bucket * ca->mi.bucket_size + bucket_offset);
+ u64 sector = bucket * ca->mi.bucket_size + bucket_offset;
+
+ if (c->sb.version_upgrade_complete >= bcachefs_metadata_version_mi_btree_bitmap &&
+ !bch2_dev_btree_bitmap_marked_sectors(ca, sector, btree_sectors(c)))
+ continue;
+
+ try_read_btree_node(w->f, ca, bio, buf, sector);
}
err:
bio_put(bio);
@@ -213,6 +231,9 @@ static int read_btree_nodes(struct find_btree_nodes *f)
closure_init_stack(&cl);
for_each_online_member(c, ca) {
+ if (!(ca->mi.data_allowed & BIT(BCH_DATA_btree)))
+ continue;
+
struct find_btree_nodes_worker *w = kmalloc(sizeof(*w), GFP_KERNEL);
struct task_struct *t;
@@ -290,7 +311,7 @@ again:
found_btree_node_to_text(&buf, c, n);
bch_err(c, "%s", buf.buf);
printbuf_exit(&buf);
- return -1;
+ return -BCH_ERR_fsck_repair_unimplemented;
}
}
@@ -436,6 +457,9 @@ bool bch2_btree_has_scanned_nodes(struct bch_fs *c, enum btree_id btree)
int bch2_get_scanned_nodes(struct bch_fs *c, enum btree_id btree,
unsigned level, struct bpos node_min, struct bpos node_max)
{
+ if (btree_id_is_alloc(btree))
+ return 0;
+
struct find_btree_nodes *f = &c->found_btree_nodes;
int ret = bch2_run_explicit_recovery_pass(c, BCH_RECOVERY_PASS_scan_for_btree_nodes);
diff --git a/libbcachefs/btree_trans_commit.c b/libbcachefs/btree_trans_commit.c
index aa9da497..bbec91e8 100644
--- a/libbcachefs/btree_trans_commit.c
+++ b/libbcachefs/btree_trans_commit.c
@@ -397,12 +397,13 @@ static int btree_key_can_insert_cached(struct btree_trans *trans, unsigned flags
struct bkey_cached *ck = (void *) path->l[0].b;
unsigned new_u64s;
struct bkey_i *new_k;
+ unsigned watermark = flags & BCH_WATERMARK_MASK;
EBUG_ON(path->level);
- if (!test_bit(BKEY_CACHED_DIRTY, &ck->flags) &&
- bch2_btree_key_cache_must_wait(c) &&
- !(flags & BCH_TRANS_COMMIT_journal_reclaim))
+ if (watermark < BCH_WATERMARK_reclaim &&
+ !test_bit(BKEY_CACHED_DIRTY, &ck->flags) &&
+ bch2_btree_key_cache_must_wait(c))
return -BCH_ERR_btree_insert_need_journal_reclaim;
/*
@@ -499,9 +500,8 @@ static int run_one_trans_trigger(struct btree_trans *trans, struct btree_insert_
}
static int run_btree_triggers(struct btree_trans *trans, enum btree_id btree_id,
- struct btree_insert_entry *btree_id_start)
+ unsigned btree_id_start)
{
- struct btree_insert_entry *i;
bool trans_trigger_run;
int ret, overwrite;
@@ -514,13 +514,13 @@ static int run_btree_triggers(struct btree_trans *trans, enum btree_id btree_id,
do {
trans_trigger_run = false;
- for (i = btree_id_start;
- i < trans->updates + trans->nr_updates && i->btree_id <= btree_id;
+ for (unsigned i = btree_id_start;
+ i < trans->nr_updates && trans->updates[i].btree_id <= btree_id;
i++) {
- if (i->btree_id != btree_id)
+ if (trans->updates[i].btree_id != btree_id)
continue;
- ret = run_one_trans_trigger(trans, i, overwrite);
+ ret = run_one_trans_trigger(trans, trans->updates + i, overwrite);
if (ret < 0)
return ret;
if (ret)
@@ -534,8 +534,7 @@ static int run_btree_triggers(struct btree_trans *trans, enum btree_id btree_id,
static int bch2_trans_commit_run_triggers(struct btree_trans *trans)
{
- struct btree_insert_entry *btree_id_start = trans->updates;
- unsigned btree_id = 0;
+ unsigned btree_id = 0, btree_id_start = 0;
int ret = 0;
/*
@@ -549,8 +548,8 @@ static int bch2_trans_commit_run_triggers(struct btree_trans *trans)
if (btree_id == BTREE_ID_alloc)
continue;
- while (btree_id_start < trans->updates + trans->nr_updates &&
- btree_id_start->btree_id < btree_id)
+ while (btree_id_start < trans->nr_updates &&
+ trans->updates[btree_id_start].btree_id < btree_id)
btree_id_start++;
ret = run_btree_triggers(trans, btree_id, btree_id_start);
@@ -558,11 +557,13 @@ static int bch2_trans_commit_run_triggers(struct btree_trans *trans)
return ret;
}
- trans_for_each_update(trans, i) {
+ for (unsigned idx = 0; idx < trans->nr_updates; idx++) {
+ struct btree_insert_entry *i = trans->updates + idx;
+
if (i->btree_id > BTREE_ID_alloc)
break;
if (i->btree_id == BTREE_ID_alloc) {
- ret = run_btree_triggers(trans, BTREE_ID_alloc, i);
+ ret = run_btree_triggers(trans, BTREE_ID_alloc, idx);
if (ret)
return ret;
break;
@@ -826,7 +827,8 @@ static inline int do_bch2_trans_commit(struct btree_trans *trans, unsigned flags
struct bch_fs *c = trans->c;
int ret = 0, u64s_delta = 0;
- trans_for_each_update(trans, i) {
+ for (unsigned idx = 0; idx < trans->nr_updates; idx++) {
+ struct btree_insert_entry *i = trans->updates + idx;
if (i->cached)
continue;
diff --git a/libbcachefs/btree_types.h b/libbcachefs/btree_types.h
index 9404d96c..e0c982a4 100644
--- a/libbcachefs/btree_types.h
+++ b/libbcachefs/btree_types.h
@@ -364,7 +364,21 @@ struct btree_insert_entry {
unsigned long ip_allocated;
};
+/* Number of btree paths we preallocate, usually enough */
#define BTREE_ITER_INITIAL 64
+/*
+ * Lmiit for btree_trans_too_many_iters(); this is enough that almost all code
+ * paths should run inside this limit, and if they don't it usually indicates a
+ * bug (leaking/duplicated btree paths).
+ *
+ * exception: some fsck paths
+ *
+ * bugs with excessive path usage seem to have possibly been eliminated now, so
+ * we might consider eliminating this (and btree_trans_too_many_iter()) at some
+ * point.
+ */
+#define BTREE_ITER_NORMAL_LIMIT 256
+/* never exceed limit */
#define BTREE_ITER_MAX (1U << 10)
struct btree_trans_commit_hook;
diff --git a/libbcachefs/btree_update_interior.c b/libbcachefs/btree_update_interior.c
index 32397b99..6030c396 100644
--- a/libbcachefs/btree_update_interior.c
+++ b/libbcachefs/btree_update_interior.c
@@ -21,14 +21,15 @@
#include "keylist.h"
#include "recovery_passes.h"
#include "replicas.h"
+#include "sb-members.h"
#include "super-io.h"
#include "trace.h"
#include <linux/random.h>
-const char * const bch2_btree_update_modes[] = {
+static const char * const bch2_btree_update_modes[] = {
#define x(t) #t,
- BCH_WATERMARKS()
+ BTREE_UPDATE_MODES()
#undef x
NULL
};
@@ -605,6 +606,26 @@ static void btree_update_add_key(struct btree_update *as,
bch2_keylist_push(keys);
}
+static bool btree_update_new_nodes_marked_sb(struct btree_update *as)
+{
+ for_each_keylist_key(&as->new_keys, k)
+ if (!bch2_dev_btree_bitmap_marked(as->c, bkey_i_to_s_c(k)))
+ return false;
+ return true;
+}
+
+static void btree_update_new_nodes_mark_sb(struct btree_update *as)
+{
+ struct bch_fs *c = as->c;
+
+ mutex_lock(&c->sb_lock);
+ for_each_keylist_key(&as->new_keys, k)
+ bch2_dev_btree_bitmap_mark(c, bkey_i_to_s_c(k));
+
+ bch2_write_super(c);
+ mutex_unlock(&c->sb_lock);
+}
+
/*
* The transactional part of an interior btree node update, where we journal the
* update we did to the interior node and update alloc info:
@@ -662,6 +683,9 @@ static void btree_update_nodes_written(struct btree_update *as)
if (ret)
goto err;
+ if (!btree_update_new_nodes_marked_sb(as))
+ btree_update_new_nodes_mark_sb(as);
+
/*
* Wait for any in flight writes to finish before we free the old nodes
* on disk:
@@ -704,9 +728,13 @@ static void btree_update_nodes_written(struct btree_update *as)
bch2_fs_fatal_err_on(ret && !bch2_journal_error(&c->journal), c,
"%s", bch2_err_str(ret));
err:
- if (as->b) {
-
- b = as->b;
+ /*
+ * We have to be careful because another thread might be getting ready
+ * to free as->b and calling btree_update_reparent() on us - we'll
+ * recheck under btree_update_lock below:
+ */
+ b = READ_ONCE(as->b);
+ if (b) {
btree_path_idx_t path_idx = get_unlocked_mut_path(trans,
as->btree_id, b->c.level, b->key.k.p);
struct btree_path *path = trans->paths + path_idx;
@@ -850,15 +878,17 @@ static void btree_update_updated_node(struct btree_update *as, struct btree *b)
{
struct bch_fs *c = as->c;
- mutex_lock(&c->btree_interior_update_lock);
- list_add_tail(&as->unwritten_list, &c->btree_interior_updates_unwritten);
-
BUG_ON(as->mode != BTREE_UPDATE_none);
+ BUG_ON(as->update_level_end < b->c.level);
BUG_ON(!btree_node_dirty(b));
BUG_ON(!b->c.level);
+ mutex_lock(&c->btree_interior_update_lock);
+ list_add_tail(&as->unwritten_list, &c->btree_interior_updates_unwritten);
+
as->mode = BTREE_UPDATE_node;
as->b = b;
+ as->update_level_end = b->c.level;
set_btree_node_write_blocked(b);
list_add(&as->write_blocked_list, &b->write_blocked);
@@ -1100,7 +1130,7 @@ static void bch2_btree_update_done(struct btree_update *as, struct btree_trans *
static struct btree_update *
bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
- unsigned level, bool split, unsigned flags)
+ unsigned level_start, bool split, unsigned flags)
{
struct bch_fs *c = trans->c;
struct btree_update *as;
@@ -1108,7 +1138,7 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
int disk_res_flags = (flags & BCH_TRANS_COMMIT_no_enospc)
? BCH_DISK_RESERVATION_NOFAIL : 0;
unsigned nr_nodes[2] = { 0, 0 };
- unsigned update_level = level;
+ unsigned level_end = level_start;
enum bch_watermark watermark = flags & BCH_WATERMARK_MASK;
int ret = 0;
u32 restart_count = trans->restart_count;
@@ -1123,34 +1153,30 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
flags &= ~BCH_WATERMARK_MASK;
flags |= watermark;
- if (watermark < c->journal.watermark) {
- struct journal_res res = { 0 };
- unsigned journal_flags = watermark|JOURNAL_RES_GET_CHECK;
+ if (watermark < BCH_WATERMARK_reclaim &&
+ test_bit(JOURNAL_SPACE_LOW, &c->journal.flags)) {
+ if (flags & BCH_TRANS_COMMIT_journal_reclaim)
+ return ERR_PTR(-BCH_ERR_journal_reclaim_would_deadlock);
- if ((flags & BCH_TRANS_COMMIT_journal_reclaim) &&
- watermark < BCH_WATERMARK_reclaim)
- journal_flags |= JOURNAL_RES_GET_NONBLOCK;
-
- ret = drop_locks_do(trans,
- bch2_journal_res_get(&c->journal, &res, 1, journal_flags));
- if (bch2_err_matches(ret, BCH_ERR_operation_blocked))
- ret = -BCH_ERR_journal_reclaim_would_deadlock;
+ bch2_trans_unlock(trans);
+ wait_event(c->journal.wait, !test_bit(JOURNAL_SPACE_LOW, &c->journal.flags));
+ ret = bch2_trans_relock(trans);
if (ret)
return ERR_PTR(ret);
}
while (1) {
- nr_nodes[!!update_level] += 1 + split;
- update_level++;
+ nr_nodes[!!level_end] += 1 + split;
+ level_end++;
- ret = bch2_btree_path_upgrade(trans, path, update_level + 1);
+ ret = bch2_btree_path_upgrade(trans, path, level_end + 1);
if (ret)
return ERR_PTR(ret);
- if (!btree_path_node(path, update_level)) {
+ if (!btree_path_node(path, level_end)) {
/* Allocating new root? */
nr_nodes[1] += split;
- update_level = BTREE_MAX_DEPTH;
+ level_end = BTREE_MAX_DEPTH;
break;
}
@@ -1158,11 +1184,11 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
* Always check for space for two keys, even if we won't have to
* split at prior level - it might have been a merge instead:
*/
- if (bch2_btree_node_insert_fits(path->l[update_level].b,
+ if (bch2_btree_node_insert_fits(path->l[level_end].b,
BKEY_BTREE_PTR_U64s_MAX * 2))
break;
- split = path->l[update_level].b->nr.live_u64s > BTREE_SPLIT_THRESHOLD(c);
+ split = path->l[level_end].b->nr.live_u64s > BTREE_SPLIT_THRESHOLD(c);
}
if (!down_read_trylock(&c->gc_lock)) {
@@ -1176,14 +1202,15 @@ bch2_btree_update_start(struct btree_trans *trans, struct btree_path *path,
as = mempool_alloc(&c->btree_interior_update_pool, GFP_NOFS);
memset(as, 0, sizeof(*as));
closure_init(&as->cl, NULL);
- as->c = c;
- as->start_time = start_time;
- as->ip_started = _RET_IP_;
- as->mode = BTREE_UPDATE_none;
- as->watermark = watermark;
- as->took_gc_lock = true;
- as->btree_id = path->btree_id;
- as->update_level = update_level;
+ as->c = c;
+ as->start_time = start_time;
+ as->ip_started = _RET_IP_;
+ as->mode = BTREE_UPDATE_none;
+ as->watermark = watermark;
+ as->took_gc_lock = true;
+ as->btree_id = path->btree_id;
+ as->update_level_start = level_start;
+ as->update_level_end = level_end;
INIT_LIST_HEAD(&as->list);
INIT_LIST_HEAD(&as->unwritten_list);
INIT_LIST_HEAD(&as->write_blocked_list);
@@ -1277,23 +1304,29 @@ static void bch2_btree_set_root_inmem(struct bch_fs *c, struct btree *b)
bch2_recalc_btree_reserve(c);
}
-static void bch2_btree_set_root(struct btree_update *as,
- struct btree_trans *trans,
- struct btree_path *path,
- struct btree *b)
+static int bch2_btree_set_root(struct btree_update *as,
+ struct btree_trans *trans,
+ struct btree_path *path,
+ struct btree *b,
+ bool nofail)
{
struct bch_fs *c = as->c;
- struct btree *old;
trace_and_count(c, btree_node_set_root, trans, b);
- old = btree_node_root(c, b);
+ struct btree *old = btree_node_root(c, b);
/*
* Ensure no one is using the old root while we switch to the
* new root:
*/
- bch2_btree_node_lock_write_nofail(trans, path, &old->c);
+ if (nofail) {
+ bch2_btree_node_lock_write_nofail(trans, path, &old->c);
+ } else {
+ int ret = bch2_btree_node_lock_write(trans, path, &old->c);
+ if (ret)
+ return ret;
+ }
bch2_btree_set_root_inmem(c, b);
@@ -1307,6 +1340,7 @@ static void bch2_btree_set_root(struct btree_update *as,
* depend on the new root would have to update the new root.
*/
bch2_btree_node_unlock_write(trans, path, old);
+ return 0;
}
/* Interior node updates: */
@@ -1373,12 +1407,12 @@ static void bch2_insert_fixup_btree_ptr(struct btree_update *as,
}
static void
-__bch2_btree_insert_keys_interior(struct btree_update *as,
- struct btree_trans *trans,
- struct btree_path *path,
- struct btree *b,
- struct btree_node_iter node_iter,
- struct keylist *keys)
+bch2_btree_insert_keys_interior(struct btree_update *as,
+ struct btree_trans *trans,
+ struct btree_path *path,
+ struct btree *b,
+ struct btree_node_iter node_iter,
+ struct keylist *keys)
{
struct bkey_i *insert = bch2_keylist_front(keys);
struct bkey_packed *k;
@@ -1534,7 +1568,7 @@ static void btree_split_insert_keys(struct btree_update *as,
bch2_btree_node_iter_init(&node_iter, b, &bch2_keylist_front(keys)->k.p);
- __bch2_btree_insert_keys_interior(as, trans, path, b, node_iter, keys);
+ bch2_btree_insert_keys_interior(as, trans, path, b, node_iter, keys);
BUG_ON(bch2_btree_node_check_topology(trans, b));
}
@@ -1649,15 +1683,16 @@ static int btree_split(struct btree_update *as, struct btree_trans *trans,
if (parent) {
/* Split a non root node */
ret = bch2_btree_insert_node(as, trans, path, parent, &as->parent_keys);
- if (ret)
- goto err;
} else if (n3) {
- bch2_btree_set_root(as, trans, trans->paths + path, n3);
+ ret = bch2_btree_set_root(as, trans, trans->paths + path, n3, false);
} else {
/* Root filled up but didn't need to be split */
- bch2_btree_set_root(as, trans, trans->paths + path, n1);
+ ret = bch2_btree_set_root(as, trans, trans->paths + path, n1, false);
}
+ if (ret)
+ goto err;
+
if (n3) {
bch2_btree_update_get_open_buckets(as, n3);
bch2_btree_node_write(c, n3, SIX_LOCK_intent, 0);
@@ -1714,27 +1749,6 @@ err:
goto out;
}
-static void
-bch2_btree_insert_keys_interior(struct btree_update *as,
- struct btree_trans *trans,
- struct btree_path *path,
- struct btree *b,
- struct keylist *keys)
-{
- struct btree_path *linked;
- unsigned i;
-
- __bch2_btree_insert_keys_interior(as, trans, path, b,
- path->l[b->c.level].iter, keys);
-
- btree_update_updated_node(as, b);
-
- trans_for_each_path_with_node(trans, b, linked, i)
- bch2_btree_node_iter_peek(&linked->l[b->c.level].iter, b);
-
- bch2_trans_verify_paths(trans);
-}
-
/**
* bch2_btree_insert_node - insert bkeys into a given btree node
*
@@ -1755,7 +1769,8 @@ static int bch2_btree_insert_node(struct btree_update *as, struct btree_trans *t
struct keylist *keys)
{
struct bch_fs *c = as->c;
- struct btree_path *path = trans->paths + path_idx;
+ struct btree_path *path = trans->paths + path_idx, *linked;
+ unsigned i;
int old_u64s = le16_to_cpu(btree_bset_last(b)->u64s);
int old_live_u64s = b->nr.live_u64s;
int live_u64s_added, u64s_added;
@@ -1784,7 +1799,13 @@ static int bch2_btree_insert_node(struct btree_update *as, struct btree_trans *t
return ret;
}
- bch2_btree_insert_keys_interior(as, trans, path, b, keys);
+ bch2_btree_insert_keys_interior(as, trans, path, b,
+ path->l[b->c.level].iter, keys);
+
+ trans_for_each_path_with_node(trans, b, linked, i)
+ bch2_btree_node_iter_peek(&linked->l[b->c.level].iter, b);
+
+ bch2_trans_verify_paths(trans);
live_u64s_added = (int) b->nr.live_u64s - old_live_u64s;
u64s_added = (int) le16_to_cpu(btree_bset_last(b)->u64s) - old_u64s;
@@ -1798,6 +1819,7 @@ static int bch2_btree_insert_node(struct btree_update *as, struct btree_trans *t
bch2_maybe_compact_whiteouts(c, b))
bch2_trans_node_reinit_iter(trans, b);
+ btree_update_updated_node(as, b);
bch2_btree_node_unlock_write(trans, path, b);
BUG_ON(bch2_btree_node_check_topology(trans, b));
@@ -1807,7 +1829,7 @@ split:
* We could attempt to avoid the transaction restart, by calling
* bch2_btree_path_upgrade() and allocating more nodes:
*/
- if (b->c.level >= as->update_level) {
+ if (b->c.level >= as->update_level_end) {
trace_and_count(c, trans_restart_split_race, trans, _THIS_IP_, b);
return btree_trans_restart(trans, BCH_ERR_transaction_restart_split_race);
}
@@ -1873,7 +1895,9 @@ static void __btree_increase_depth(struct btree_update *as, struct btree_trans *
bch2_keylist_add(&as->parent_keys, &b->key);
btree_split_insert_keys(as, trans, path_idx, n, &as->parent_keys);
- bch2_btree_set_root(as, trans, path, n);
+ int ret = bch2_btree_set_root(as, trans, path, n, true);
+ BUG_ON(ret);
+
bch2_btree_update_get_open_buckets(as, n);
bch2_btree_node_write(c, n, SIX_LOCK_intent, 0);
bch2_trans_node_add(trans, path, n);
@@ -1926,6 +1950,18 @@ int __bch2_foreground_maybe_merge(struct btree_trans *trans,
BUG_ON(!trans->paths[path].should_be_locked);
BUG_ON(!btree_node_locked(&trans->paths[path], level));
+ /*
+ * Work around a deadlock caused by the btree write buffer not doing
+ * merges and leaving tons of merges for us to do - we really don't need
+ * to be doing merges at all from the interior update path, and if the
+ * interior update path is generating too many new interior updates we
+ * deadlock:
+ */
+ if ((flags & BCH_WATERMARK_MASK) == BCH_WATERMARK_interior_updates)
+ return 0;
+
+ flags &= ~BCH_WATERMARK_MASK;
+
b = trans->paths[path].l[level].b;
if ((sib == btree_prev_sib && bpos_eq(b->data->min_key, POS_MIN)) ||
@@ -2071,6 +2107,10 @@ err:
bch2_path_put(trans, new_path, true);
bch2_path_put(trans, sib_path, true);
bch2_trans_verify_locks(trans);
+ if (ret == -BCH_ERR_journal_reclaim_would_deadlock)
+ ret = 0;
+ if (!ret)
+ ret = bch2_trans_relock(trans);
return ret;
err_free_update:
bch2_btree_node_free_never_used(as, trans, n);
@@ -2116,12 +2156,13 @@ int bch2_btree_node_rewrite(struct btree_trans *trans,
if (parent) {
bch2_keylist_add(&as->parent_keys, &n->key);
ret = bch2_btree_insert_node(as, trans, iter->path, parent, &as->parent_keys);
- if (ret)
- goto err;
} else {
- bch2_btree_set_root(as, trans, btree_iter_path(trans, iter), n);
+ ret = bch2_btree_set_root(as, trans, btree_iter_path(trans, iter), n, false);
}
+ if (ret)
+ goto err;
+
bch2_btree_update_get_open_buckets(as, n);
bch2_btree_node_write(c, n, SIX_LOCK_intent, 0);
@@ -2519,9 +2560,11 @@ void bch2_btree_root_alloc_fake(struct bch_fs *c, enum btree_id id, unsigned lev
static void bch2_btree_update_to_text(struct printbuf *out, struct btree_update *as)
{
- prt_printf(out, "%ps: btree=%s watermark=%s mode=%s nodes_written=%u cl.remaining=%u journal_seq=%llu\n",
+ prt_printf(out, "%ps: btree=%s l=%u-%u watermark=%s mode=%s nodes_written=%u cl.remaining=%u journal_seq=%llu\n",
(void *) as->ip_started,
bch2_btree_id_str(as->btree_id),
+ as->update_level_start,
+ as->update_level_end,
bch2_watermarks[as->watermark],
bch2_btree_update_modes[as->mode],
as->nodes_written,
diff --git a/libbcachefs/btree_update_interior.h b/libbcachefs/btree_update_interior.h
index 88dcf5a2..c1a479eb 100644
--- a/libbcachefs/btree_update_interior.h
+++ b/libbcachefs/btree_update_interior.h
@@ -57,7 +57,8 @@ struct btree_update {
unsigned took_gc_lock:1;
enum btree_id btree_id;
- unsigned update_level;
+ unsigned update_level_start;
+ unsigned update_level_end;
struct disk_reservation disk_res;
diff --git a/libbcachefs/btree_write_buffer.c b/libbcachefs/btree_write_buffer.c
index baf63e2f..36a6f42a 100644
--- a/libbcachefs/btree_write_buffer.c
+++ b/libbcachefs/btree_write_buffer.c
@@ -316,6 +316,16 @@ static int bch2_btree_write_buffer_flush_locked(struct btree_trans *trans)
bpos_gt(k->k.k.p, path->l[0].b->key.k.p)) {
bch2_btree_node_unlock_write(trans, path, path->l[0].b);
write_locked = false;
+
+ ret = lockrestart_do(trans,
+ bch2_btree_iter_traverse(&iter) ?:
+ bch2_foreground_maybe_merge(trans, iter.path, 0,
+ BCH_WATERMARK_reclaim|
+ BCH_TRANS_COMMIT_journal_reclaim|
+ BCH_TRANS_COMMIT_no_check_rw|
+ BCH_TRANS_COMMIT_no_enospc));
+ if (ret)
+ goto err;
}
}
@@ -382,10 +392,10 @@ static int bch2_btree_write_buffer_flush_locked(struct btree_trans *trans)
ret = commit_do(trans, NULL, NULL,
BCH_WATERMARK_reclaim|
+ BCH_TRANS_COMMIT_journal_reclaim|
BCH_TRANS_COMMIT_no_check_rw|
BCH_TRANS_COMMIT_no_enospc|
- BCH_TRANS_COMMIT_no_journal_res|
- BCH_TRANS_COMMIT_journal_reclaim,
+ BCH_TRANS_COMMIT_no_journal_res ,
btree_write_buffered_insert(trans, i));
if (ret)
goto err;
diff --git a/libbcachefs/buckets.h b/libbcachefs/buckets.h
index 00aaf4bb..f9af5ada 100644
--- a/libbcachefs/buckets.h
+++ b/libbcachefs/buckets.h
@@ -395,14 +395,6 @@ static inline const char *bch2_data_type_str(enum bch_data_type type)
: "(invalid data type)";
}
-static inline void bch2_prt_data_type(struct printbuf *out, enum bch_data_type type)
-{
- if (type < BCH_DATA_NR)
- prt_str(out, __bch2_data_types[type]);
- else
- prt_printf(out, "(invalid data type %u)", type);
-}
-
/* disk reservations: */
static inline void bch2_disk_reservation_put(struct bch_fs *c,
diff --git a/libbcachefs/chardev.c b/libbcachefs/chardev.c
index cbfa6459..72781aad 100644
--- a/libbcachefs/chardev.c
+++ b/libbcachefs/chardev.c
@@ -134,42 +134,38 @@ static long bch2_ioctl_incremental(struct bch_ioctl_incremental __user *user_arg
struct fsck_thread {
struct thread_with_stdio thr;
struct bch_fs *c;
- char **devs;
- size_t nr_devs;
struct bch_opts opts;
};
static void bch2_fsck_thread_exit(struct thread_with_stdio *_thr)
{
struct fsck_thread *thr = container_of(_thr, struct fsck_thread, thr);
- if (thr->devs)
- for (size_t i = 0; i < thr->nr_devs; i++)
- kfree(thr->devs[i]);
- kfree(thr->devs);
kfree(thr);
}
static int bch2_fsck_offline_thread_fn(struct thread_with_stdio *stdio)
{
struct fsck_thread *thr = container_of(stdio, struct fsck_thread, thr);
- struct bch_fs *c = bch2_fs_open(thr->devs, thr->nr_devs, thr->opts);
+ struct bch_fs *c = thr->c;
- if (IS_ERR(c))
- return PTR_ERR(c);
+ int ret = PTR_ERR_OR_ZERO(c);
+ if (ret)
+ return ret;
- int ret = 0;
- if (test_bit(BCH_FS_errors_fixed, &c->flags))
- ret |= 1;
- if (test_bit(BCH_FS_error, &c->flags))
- ret |= 4;
+ ret = bch2_fs_start(thr->c);
+ if (ret)
+ goto err;
- bch2_fs_stop(c);
-
- if (ret & 1)
+ if (test_bit(BCH_FS_errors_fixed, &c->flags)) {
bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: errors fixed\n", c->name);
- if (ret & 4)
+ ret |= 1;
+ }
+ if (test_bit(BCH_FS_error, &c->flags)) {
bch2_stdio_redirect_printf(&stdio->stdio, false, "%s: still has errors\n", c->name);
-
+ ret |= 4;
+ }
+err:
+ bch2_fs_stop(c);
return ret;
}
@@ -182,7 +178,7 @@ static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_a
{
struct bch_ioctl_fsck_offline arg;
struct fsck_thread *thr = NULL;
- u64 *devs = NULL;
+ darray_str(devs) = {};
long ret = 0;
if (copy_from_user(&arg, user_arg, sizeof(arg)))
@@ -194,28 +190,31 @@ static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_a
if (!capable(CAP_SYS_ADMIN))
return -EPERM;
- if (!(devs = kcalloc(arg.nr_devs, sizeof(*devs), GFP_KERNEL)) ||
- !(thr = kzalloc(sizeof(*thr), GFP_KERNEL)) ||
- !(thr->devs = kcalloc(arg.nr_devs, sizeof(*thr->devs), GFP_KERNEL))) {
+ for (size_t i = 0; i < arg.nr_devs; i++) {
+ u64 dev_u64;
+ ret = copy_from_user_errcode(&dev_u64, &user_arg->devs[i], sizeof(u64));
+ if (ret)
+ goto err;
+
+ char *dev_str = strndup_user((char __user *)(unsigned long) dev_u64, PATH_MAX);
+ ret = PTR_ERR_OR_ZERO(dev_str);
+ if (ret)
+ goto err;
+
+ ret = darray_push(&devs, dev_str);
+ if (ret) {
+ kfree(dev_str);
+ goto err;
+ }
+ }
+
+ thr = kzalloc(sizeof(*thr), GFP_KERNEL);
+ if (!thr) {
ret = -ENOMEM;
goto err;
}
thr->opts = bch2_opts_empty();
- thr->nr_devs = arg.nr_devs;
-
- if (copy_from_user(devs, &user_arg->devs[0],
- array_size(sizeof(user_arg->devs[0]), arg.nr_devs))) {
- ret = -EINVAL;
- goto err;
- }
-
- for (size_t i = 0; i < arg.nr_devs; i++) {
- thr->devs[i] = strndup_user((char __user *)(unsigned long) devs[i], PATH_MAX);
- ret = PTR_ERR_OR_ZERO(thr->devs[i]);
- if (ret)
- goto err;
- }
if (arg.opts) {
char *optstr = strndup_user((char __user *)(unsigned long) arg.opts, 1 << 16);
@@ -230,15 +229,26 @@ static long bch2_ioctl_fsck_offline(struct bch_ioctl_fsck_offline __user *user_a
opt_set(thr->opts, stdio, (u64)(unsigned long)&thr->thr.stdio);
+ /* We need request_key() to be called before we punt to kthread: */
+ opt_set(thr->opts, nostart, true);
+
+ thr->c = bch2_fs_open(devs.data, arg.nr_devs, thr->opts);
+
+ if (!IS_ERR(thr->c) &&
+ thr->c->opts.errors == BCH_ON_ERROR_panic)
+ thr->c->opts.errors = BCH_ON_ERROR_ro;
+
ret = bch2_run_thread_with_stdio(&thr->thr, &bch2_offline_fsck_ops);
-err:
- if (ret < 0) {
- if (thr)
- bch2_fsck_thread_exit(&thr->thr);
- pr_err("ret %s", bch2_err_str(ret));
- }
- kfree(devs);
+out:
+ darray_for_each(devs, i)
+ kfree(*i);
+ darray_exit(&devs);
return ret;
+err:
+ if (thr)
+ bch2_fsck_thread_exit(&thr->thr);
+ pr_err("ret %s", bch2_err_str(ret));
+ goto out;
}
static long bch2_global_ioctl(unsigned cmd, void __user *arg)
diff --git a/libbcachefs/checksum.c b/libbcachefs/checksum.c
index 4701457f..7ed779b4 100644
--- a/libbcachefs/checksum.c
+++ b/libbcachefs/checksum.c
@@ -429,15 +429,20 @@ int bch2_rechecksum_bio(struct bch_fs *c, struct bio *bio,
extent_nonce(version, crc_old), bio);
if (bch2_crc_cmp(merged, crc_old.csum) && !c->opts.no_data_io) {
- bch_err(c, "checksum error in %s() (memory corruption or bug?)\n"
- "expected %0llx:%0llx got %0llx:%0llx (old type %s new type %s)",
- __func__,
- crc_old.csum.hi,
- crc_old.csum.lo,
- merged.hi,
- merged.lo,
- bch2_csum_types[crc_old.csum_type],
- bch2_csum_types[new_csum_type]);
+ struct printbuf buf = PRINTBUF;
+ prt_printf(&buf, "checksum error in %s() (memory corruption or bug?)\n"
+ "expected %0llx:%0llx got %0llx:%0llx (old type ",
+ __func__,
+ crc_old.csum.hi,
+ crc_old.csum.lo,
+ merged.hi,
+ merged.lo);
+ bch2_prt_csum_type(&buf, crc_old.csum_type);
+ prt_str(&buf, " new type ");
+ bch2_prt_csum_type(&buf, new_csum_type);
+ prt_str(&buf, ")");
+ bch_err(c, "%s", buf.buf);
+ printbuf_exit(&buf);
return -EIO;
}
diff --git a/libbcachefs/checksum.h b/libbcachefs/checksum.h
index 1b8c2c10..e40499fd 100644
--- a/libbcachefs/checksum.h
+++ b/libbcachefs/checksum.h
@@ -61,11 +61,12 @@ static inline void bch2_csum_err_msg(struct printbuf *out,
struct bch_csum expected,
struct bch_csum got)
{
- prt_printf(out, "checksum error: got ");
+ prt_str(out, "checksum error, type ");
+ bch2_prt_csum_type(out, type);
+ prt_str(out, ": got ");
bch2_csum_to_text(out, type, got);
prt_str(out, " should be ");
bch2_csum_to_text(out, type, expected);
- prt_printf(out, " type %s", bch2_csum_types[type]);
}
int bch2_chacha_encrypt_key(struct bch_key *, struct nonce, void *, size_t);
diff --git a/libbcachefs/compress.h b/libbcachefs/compress.h
index 58c2eb45..607fd5e2 100644
--- a/libbcachefs/compress.h
+++ b/libbcachefs/compress.h
@@ -47,14 +47,6 @@ static inline enum bch_compression_type bch2_compression_opt_to_type(unsigned v)
return __bch2_compression_opt_to_type[bch2_compression_decode(v).type];
}
-static inline void bch2_prt_compression_type(struct printbuf *out, enum bch_compression_type type)
-{
- if (type < BCH_COMPRESSION_TYPE_NR)
- prt_str(out, __bch2_compression_types[type]);
- else
- prt_printf(out, "(invalid compression type %u)", type);
-}
-
int bch2_bio_uncompress_inplace(struct bch_fs *, struct bio *,
struct bch_extent_crc_unpacked *);
int bch2_bio_uncompress(struct bch_fs *, struct bio *, struct bio *,
diff --git a/libbcachefs/data_update.c b/libbcachefs/data_update.c
index 34731ee0..0022b51c 100644
--- a/libbcachefs/data_update.c
+++ b/libbcachefs/data_update.c
@@ -598,6 +598,8 @@ int bch2_data_update_init(struct btree_trans *trans,
i++;
}
+ unsigned durability_required = max(0, (int) (io_opts.data_replicas - durability_have));
+
/*
* If current extent durability is less than io_opts.data_replicas,
* we're not trying to rereplicate the extent up to data_replicas here -
@@ -607,7 +609,7 @@ int bch2_data_update_init(struct btree_trans *trans,
* rereplicate, currently, so that users don't get an unexpected -ENOSPC
*/
if (!(m->data_opts.write_flags & BCH_WRITE_CACHED) &&
- durability_have >= io_opts.data_replicas) {
+ !durability_required) {
m->data_opts.kill_ptrs |= m->data_opts.rewrite_ptrs;
m->data_opts.rewrite_ptrs = 0;
/* if iter == NULL, it's just a promote */
@@ -616,11 +618,18 @@ int bch2_data_update_init(struct btree_trans *trans,
goto done;
}
- m->op.nr_replicas = min(durability_removing, io_opts.data_replicas - durability_have) +
+ m->op.nr_replicas = min(durability_removing, durability_required) +
m->data_opts.extra_replicas;
- m->op.nr_replicas_required = m->op.nr_replicas;
- BUG_ON(!m->op.nr_replicas);
+ /*
+ * If device(s) were set to durability=0 after data was written to them
+ * we can end up with a duribilty=0 extent, and the normal algorithm
+ * that tries not to increase durability doesn't work:
+ */
+ if (!(durability_have + durability_removing))
+ m->op.nr_replicas = max((unsigned) m->op.nr_replicas, 1);
+
+ m->op.nr_replicas_required = m->op.nr_replicas;
if (reserve_sectors) {
ret = bch2_disk_reservation_add(c, &m->op.res, reserve_sectors,
diff --git a/libbcachefs/debug.c b/libbcachefs/debug.c
index 208ce6f0..cd99b739 100644
--- a/libbcachefs/debug.c
+++ b/libbcachefs/debug.c
@@ -13,6 +13,7 @@
#include "btree_iter.h"
#include "btree_locking.h"
#include "btree_update.h"
+#include "btree_update_interior.h"
#include "buckets.h"
#include "debug.h"
#include "error.h"
@@ -668,7 +669,7 @@ static ssize_t bch2_journal_pins_read(struct file *file, char __user *buf,
i->size = size;
i->ret = 0;
- do {
+ while (1) {
err = flush_buf(i);
if (err)
return err;
@@ -676,9 +677,12 @@ static ssize_t bch2_journal_pins_read(struct file *file, char __user *buf,
if (!i->size)
break;
+ if (done)
+ break;
+
done = bch2_journal_seq_pins_to_text(&i->buf, &c->journal, &i->iter);
i->iter++;
- } while (!done);
+ }
if (i->buf.allocation_failure)
return -ENOMEM;
@@ -693,13 +697,45 @@ static const struct file_operations journal_pins_ops = {
.read = bch2_journal_pins_read,
};
+static ssize_t bch2_btree_updates_read(struct file *file, char __user *buf,
+ size_t size, loff_t *ppos)
+{
+ struct dump_iter *i = file->private_data;
+ struct bch_fs *c = i->c;
+ int err;
+
+ i->ubuf = buf;
+ i->size = size;
+ i->ret = 0;
+
+ if (!i->iter) {
+ bch2_btree_updates_to_text(&i->buf, c);
+ i->iter++;
+ }
+
+ err = flush_buf(i);
+ if (err)
+ return err;
+
+ if (i->buf.allocation_failure)
+ return -ENOMEM;
+
+ return i->ret;
+}
+
+static const struct file_operations btree_updates_ops = {
+ .owner = THIS_MODULE,
+ .open = bch2_dump_open,
+ .release = bch2_dump_release,
+ .read = bch2_btree_updates_read,
+};
+
static int btree_transaction_stats_open(struct inode *inode, struct file *file)
{
struct bch_fs *c = inode->i_private;
struct dump_iter *i;
i = kzalloc(sizeof(struct dump_iter), GFP_KERNEL);
-
if (!i)
return -ENOMEM;
@@ -866,6 +902,20 @@ void bch2_fs_debug_exit(struct bch_fs *c)
debugfs_remove_recursive(c->fs_debug_dir);
}
+static void bch2_fs_debug_btree_init(struct bch_fs *c, struct btree_debug *bd)
+{
+ struct dentry *d;
+
+ d = debugfs_create_dir(bch2_btree_id_str(bd->id), c->btree_debug_dir);
+
+ debugfs_create_file("keys", 0400, d, bd, &btree_debug_ops);
+
+ debugfs_create_file("formats", 0400, d, bd, &btree_format_debug_ops);
+
+ debugfs_create_file("bfloat-failed", 0400, d, bd,
+ &bfloat_failed_debug_ops);
+}
+
void bch2_fs_debug_init(struct bch_fs *c)
{
struct btree_debug *bd;
@@ -888,6 +938,9 @@ void bch2_fs_debug_init(struct bch_fs *c)
debugfs_create_file("journal_pins", 0400, c->fs_debug_dir,
c->btree_debug, &journal_pins_ops);
+ debugfs_create_file("btree_updates", 0400, c->fs_debug_dir,
+ c->btree_debug, &btree_updates_ops);
+
debugfs_create_file("btree_transaction_stats", 0400, c->fs_debug_dir,
c, &btree_transaction_stats_op);
@@ -902,21 +955,7 @@ void bch2_fs_debug_init(struct bch_fs *c)
bd < c->btree_debug + ARRAY_SIZE(c->btree_debug);
bd++) {
bd->id = bd - c->btree_debug;
- debugfs_create_file(bch2_btree_id_str(bd->id),
- 0400, c->btree_debug_dir, bd,
- &btree_debug_ops);
-
- snprintf(name, sizeof(name), "%s-formats",
- bch2_btree_id_str(bd->id));
-
- debugfs_create_file(name, 0400, c->btree_debug_dir, bd,
- &btree_format_debug_ops);
-
- snprintf(name, sizeof(name), "%s-bfloat-failed",
- bch2_btree_id_str(bd->id));
-
- debugfs_create_file(name, 0400, c->btree_debug_dir, bd,
- &bfloat_failed_debug_ops);
+ bch2_fs_debug_btree_init(c, bd);
}
}
diff --git a/libbcachefs/ec.c b/libbcachefs/ec.c
index 08207524..556a2171 100644
--- a/libbcachefs/ec.c
+++ b/libbcachefs/ec.c
@@ -131,29 +131,33 @@ fsck_err:
void bch2_stripe_to_text(struct printbuf *out, struct bch_fs *c,
struct bkey_s_c k)
{
- const struct bch_stripe *s = bkey_s_c_to_stripe(k).v;
- unsigned i, nr_data = s->nr_blocks - s->nr_redundant;
+ const struct bch_stripe *sp = bkey_s_c_to_stripe(k).v;
+ struct bch_stripe s = {};
- prt_printf(out, "algo %u sectors %u blocks %u:%u csum %u gran %u",
- s->algorithm,
- le16_to_cpu(s->sectors),
- nr_data,
- s->nr_redundant,
- s->csum_type,
- 1U << s->csum_granularity_bits);
+ memcpy(&s, sp, min(sizeof(s), bkey_val_bytes(k.k)));
- for (i = 0; i < s->nr_blocks; i++) {
- const struct bch_extent_ptr *ptr = s->ptrs + i;
- struct bch_dev *ca = bch_dev_bkey_exists(c, ptr->dev);
- u32 offset;
- u64 b = sector_to_bucket_and_offset(ca, ptr->offset, &offset);
+ unsigned nr_data = s.nr_blocks - s.nr_redundant;
- prt_printf(out, " %u:%llu:%u", ptr->dev, b, offset);
- if (i < nr_data)
- prt_printf(out, "#%u", stripe_blockcount_get(s, i));
- prt_printf(out, " gen %u", ptr->gen);
- if (ptr_stale(ca, ptr))
- prt_printf(out, " stale");
+ prt_printf(out, "algo %u sectors %u blocks %u:%u csum ",
+ s.algorithm,
+ le16_to_cpu(s.sectors),
+ nr_data,
+ s.nr_redundant);
+ bch2_prt_csum_type(out, s.csum_type);
+ prt_printf(out, " gran %u", 1U << s.csum_granularity_bits);
+
+ for (unsigned i = 0; i < s.nr_blocks; i++) {
+ const struct bch_extent_ptr *ptr = sp->ptrs + i;
+
+ if ((void *) ptr >= bkey_val_end(k))
+ break;
+
+ bch2_extent_ptr_to_text(out, c, ptr);
+
+ if (s.csum_type < BCH_CSUM_NR &&
+ i < nr_data &&
+ stripe_blockcount_offset(&s, i) < bkey_val_bytes(k.k))
+ prt_printf(out, "#%u", stripe_blockcount_get(sp, i));
}
}
@@ -607,10 +611,8 @@ static void ec_validate_checksums(struct bch_fs *c, struct ec_stripe_buf *buf)
struct printbuf err = PRINTBUF;
struct bch_dev *ca = bch_dev_bkey_exists(c, v->ptrs[i].dev);
- prt_printf(&err, "stripe checksum error: expected %0llx:%0llx got %0llx:%0llx (type %s)\n",
- want.hi, want.lo,
- got.hi, got.lo,
- bch2_csum_types[v->csum_type]);
+ prt_str(&err, "stripe ");
+ bch2_csum_err_msg(&err, v->csum_type, want, got);
prt_printf(&err, " for %ps at %u of\n ", (void *) _RET_IP_, i);
bch2_bkey_val_to_text(&err, c, bkey_i_to_s_c(&buf->key));
bch_err_ratelimited(ca, "%s", err.buf);
diff --git a/libbcachefs/ec.h b/libbcachefs/ec.h
index f4369b02..f0426168 100644
--- a/libbcachefs/ec.h
+++ b/libbcachefs/ec.h
@@ -32,6 +32,8 @@ static inline unsigned stripe_csums_per_device(const struct bch_stripe *s)
static inline unsigned stripe_csum_offset(const struct bch_stripe *s,
unsigned dev, unsigned csum_idx)
{
+ EBUG_ON(s->csum_type >= BCH_CSUM_NR);
+
unsigned csum_bytes = bch_crc_bytes[s->csum_type];
return sizeof(struct bch_stripe) +
diff --git a/libbcachefs/extents.c b/libbcachefs/extents.c
index 0e3ca99f..1a331e53 100644
--- a/libbcachefs/extents.c
+++ b/libbcachefs/extents.c
@@ -998,7 +998,9 @@ void bch2_extent_ptr_to_text(struct printbuf *out, struct bch_fs *c, const struc
prt_str(out, " cached");
if (ptr->unwritten)
prt_str(out, " unwritten");
- if (ca && ptr_stale(ca, ptr))
+ if (b >= ca->mi.first_bucket &&
+ b < ca->mi.nbuckets &&
+ ptr_stale(ca, ptr))
prt_printf(out, " stale");
}
}
@@ -1028,11 +1030,12 @@ void bch2_bkey_ptrs_to_text(struct printbuf *out, struct bch_fs *c,
struct bch_extent_crc_unpacked crc =
bch2_extent_crc_unpack(k.k, entry_to_crc(entry));
- prt_printf(out, "crc: c_size %u size %u offset %u nonce %u csum %s compress ",
+ prt_printf(out, "crc: c_size %u size %u offset %u nonce %u csum ",
crc.compressed_size,
crc.uncompressed_size,
- crc.offset, crc.nonce,
- bch2_csum_types[crc.csum_type]);
+ crc.offset, crc.nonce);
+ bch2_prt_csum_type(out, crc.csum_type);
+ prt_str(out, " compress ");
bch2_prt_compression_type(out, crc.compression_type);
break;
}
diff --git a/libbcachefs/eytzinger.c b/libbcachefs/eytzinger.c
index 4ce5e957..0f955c3c 100644
--- a/libbcachefs/eytzinger.c
+++ b/libbcachefs/eytzinger.c
@@ -115,7 +115,7 @@ static void swap_bytes(void *a, void *b, size_t n)
struct wrapper {
cmp_func_t cmp;
- swap_func_t swap;
+ swap_func_t swap_func;
};
/*
@@ -125,7 +125,7 @@ struct wrapper {
static void do_swap(void *a, void *b, size_t size, swap_r_func_t swap_func, const void *priv)
{
if (swap_func == SWAP_WRAPPER) {
- ((const struct wrapper *)priv)->swap(a, b, (int)size);
+ ((const struct wrapper *)priv)->swap_func(a, b, (int)size);
return;
}
@@ -174,7 +174,7 @@ void eytzinger0_sort_r(void *base, size_t n, size_t size,
int i, c, r;
/* called from 'sort' without swap function, let's pick the default */
- if (swap_func == SWAP_WRAPPER && !((struct wrapper *)priv)->swap)
+ if (swap_func == SWAP_WRAPPER && !((struct wrapper *)priv)->swap_func)
swap_func = NULL;
if (!swap_func) {
@@ -227,7 +227,7 @@ void eytzinger0_sort(void *base, size_t n, size_t size,
{
struct wrapper w = {
.cmp = cmp_func,
- .swap = swap_func,
+ .swap_func = swap_func,
};
return eytzinger0_sort_r(base, n, size, _CMP_WRAPPER, SWAP_WRAPPER, &w);
diff --git a/libbcachefs/eytzinger.h b/libbcachefs/eytzinger.h
index ee0e2df3..24840aee 100644
--- a/libbcachefs/eytzinger.h
+++ b/libbcachefs/eytzinger.h
@@ -242,8 +242,8 @@ static inline unsigned inorder_to_eytzinger0(unsigned i, unsigned size)
(_i) = eytzinger0_next((_i), (_size)))
/* return greatest node <= @search, or -1 if not found */
-static inline ssize_t eytzinger0_find_le(void *base, size_t nr, size_t size,
- cmp_func_t cmp, const void *search)
+static inline int eytzinger0_find_le(void *base, size_t nr, size_t size,
+ cmp_func_t cmp, const void *search)
{
unsigned i, n = 0;
@@ -256,18 +256,32 @@ static inline ssize_t eytzinger0_find_le(void *base, size_t nr, size_t size,
} while (n < nr);
if (n & 1) {
- /* @i was greater than @search, return previous node: */
+ /*
+ * @i was greater than @search, return previous node:
+ *
+ * if @i was leftmost/smallest element,
+ * eytzinger0_prev(eytzinger0_first())) returns -1, as expected
+ */
return eytzinger0_prev(i, nr);
} else {
return i;
}
}
-static inline ssize_t eytzinger0_find_gt(void *base, size_t nr, size_t size,
- cmp_func_t cmp, const void *search)
+static inline int eytzinger0_find_gt(void *base, size_t nr, size_t size,
+ cmp_func_t cmp, const void *search)
{
ssize_t idx = eytzinger0_find_le(base, nr, size, cmp, search);
- return eytzinger0_next(idx, size);
+
+ /*
+ * if eytitzinger0_find_le() returned -1 - no element was <= search - we
+ * want to return the first element; next/prev identities mean this work
+ * as expected
+ *
+ * similarly if find_le() returns last element, we should return -1;
+ * identities mean this all works out:
+ */
+ return eytzinger0_next(idx, nr);
}
#define eytzinger0_find(base, nr, size, _cmp, search) \
diff --git a/libbcachefs/fs-io-direct.c b/libbcachefs/fs-io-direct.c
index f49e6c0f..b889370a 100644
--- a/libbcachefs/fs-io-direct.c
+++ b/libbcachefs/fs-io-direct.c
@@ -387,6 +387,8 @@ static __always_inline long bch2_dio_write_done(struct dio_write *dio)
ret = dio->op.error ?: ((long) dio->written << 9);
bio_put(&dio->op.wbio.bio);
+ bch2_write_ref_put(dio->op.c, BCH_WRITE_REF_dio_write);
+
/* inode->i_dio_count is our ref on inode and thus bch_fs */
inode_dio_end(&inode->v);
@@ -590,22 +592,25 @@ ssize_t bch2_direct_write(struct kiocb *req, struct iov_iter *iter)
prefetch(&inode->ei_inode);
prefetch((void *) &inode->ei_inode + 64);
+ if (!bch2_write_ref_tryget(c, BCH_WRITE_REF_dio_write))
+ return -EROFS;
+
inode_lock(&inode->v);
ret = generic_write_checks(req, iter);
if (unlikely(ret <= 0))
- goto err;
+ goto err_put_write_ref;
ret = file_remove_privs(file);
if (unlikely(ret))
- goto err;
+ goto err_put_write_ref;
ret = file_update_time(file);
if (unlikely(ret))
- goto err;
+ goto err_put_write_ref;
if (unlikely((req->ki_pos|iter->count) & (block_bytes(c) - 1)))
- goto err;
+ goto err_put_write_ref;
inode_dio_begin(&inode->v);
bch2_pagecache_block_get(inode);
@@ -645,7 +650,7 @@ ssize_t bch2_direct_write(struct kiocb *req, struct iov_iter *iter)
}
ret = bch2_dio_write_loop(dio);
-err:
+out:
if (locked)
inode_unlock(&inode->v);
return ret;
@@ -653,7 +658,9 @@ err_put_bio:
bch2_pagecache_block_put(inode);
bio_put(bio);
inode_dio_end(&inode->v);
- goto err;
+err_put_write_ref:
+ bch2_write_ref_put(c, BCH_WRITE_REF_dio_write);
+ goto out;
}
void bch2_fs_fs_io_direct_exit(struct bch_fs *c)
diff --git a/libbcachefs/fs-io.c b/libbcachefs/fs-io.c
index 8c70123b..20b40477 100644
--- a/libbcachefs/fs-io.c
+++ b/libbcachefs/fs-io.c
@@ -174,18 +174,18 @@ void __bch2_i_sectors_acct(struct bch_fs *c, struct bch_inode_info *inode,
static int bch2_flush_inode(struct bch_fs *c,
struct bch_inode_info *inode)
{
- struct bch_inode_unpacked u;
- int ret;
-
if (c->opts.journal_flush_disabled)
return 0;
- ret = bch2_inode_find_by_inum(c, inode_inum(inode), &u);
- if (ret)
- return ret;
+ if (!bch2_write_ref_tryget(c, BCH_WRITE_REF_fsync))
+ return -EROFS;
- return bch2_journal_flush_seq(&c->journal, u.bi_journal_seq) ?:
- bch2_inode_flush_nocow_writes(c, inode);
+ struct bch_inode_unpacked u;
+ int ret = bch2_inode_find_by_inum(c, inode_inum(inode), &u) ?:
+ bch2_journal_flush_seq(&c->journal, u.bi_journal_seq) ?:
+ bch2_inode_flush_nocow_writes(c, inode);
+ bch2_write_ref_put(c, BCH_WRITE_REF_fsync);
+ return ret;
}
int bch2_fsync(struct file *file, loff_t start, loff_t end, int datasync)
diff --git a/libbcachefs/journal_io.c b/libbcachefs/journal_io.c
index 725fcf46..9aa28b52 100644
--- a/libbcachefs/journal_io.c
+++ b/libbcachefs/journal_io.c
@@ -247,7 +247,7 @@ static void journal_entry_err_msg(struct printbuf *out,
if (entry) {
prt_str(out, " type=");
- prt_str(out, bch2_jset_entry_types[entry->type]);
+ bch2_prt_jset_entry_type(out, entry->type);
}
if (!jset) {
@@ -403,7 +403,8 @@ static void journal_entry_btree_keys_to_text(struct printbuf *out, struct bch_fs
jset_entry_for_each_key(entry, k) {
if (!first) {
prt_newline(out);
- prt_printf(out, "%s: ", bch2_jset_entry_types[entry->type]);
+ bch2_prt_jset_entry_type(out, entry->type);
+ prt_str(out, ": ");
}
prt_printf(out, "btree=%s l=%u ", bch2_btree_id_str(entry->btree_id), entry->level);
bch2_bkey_val_to_text(out, c, bkey_i_to_s_c(k));
@@ -563,9 +564,9 @@ static void journal_entry_usage_to_text(struct printbuf *out, struct bch_fs *c,
struct jset_entry_usage *u =
container_of(entry, struct jset_entry_usage, entry);
- prt_printf(out, "type=%s v=%llu",
- bch2_fs_usage_types[u->entry.btree_id],
- le64_to_cpu(u->v));
+ prt_str(out, "type=");
+ bch2_prt_fs_usage_type(out, u->entry.btree_id);
+ prt_printf(out, " v=%llu", le64_to_cpu(u->v));
}
static int journal_entry_data_usage_validate(struct bch_fs *c,
@@ -827,11 +828,11 @@ int bch2_journal_entry_validate(struct bch_fs *c,
void bch2_journal_entry_to_text(struct printbuf *out, struct bch_fs *c,
struct jset_entry *entry)
{
+ bch2_prt_jset_entry_type(out, entry->type);
+
if (entry->type < BCH_JSET_ENTRY_NR) {
- prt_printf(out, "%s: ", bch2_jset_entry_types[entry->type]);
+ prt_str(out, ": ");
bch2_jset_entry_ops[entry->type].to_text(out, c, entry);
- } else {
- prt_printf(out, "(unknown type %u)", entry->type);
}
}
diff --git a/libbcachefs/journal_reclaim.c b/libbcachefs/journal_reclaim.c
index ab811c0d..04a57784 100644
--- a/libbcachefs/journal_reclaim.c
+++ b/libbcachefs/journal_reclaim.c
@@ -67,6 +67,8 @@ void bch2_journal_set_watermark(struct journal *j)
track_event_change(&c->times[BCH_TIME_blocked_write_buffer_full], low_on_wb))
trace_and_count(c, journal_full, c);
+ mod_bit(JOURNAL_SPACE_LOW, &j->flags, low_on_space || low_on_pin);
+
swap(watermark, j->watermark);
if (watermark > j->watermark)
journal_wake(j);
diff --git a/libbcachefs/journal_types.h b/libbcachefs/journal_types.h
index 8c053cb6..b5161b5d 100644
--- a/libbcachefs/journal_types.h
+++ b/libbcachefs/journal_types.h
@@ -134,6 +134,7 @@ enum journal_flags {
JOURNAL_STARTED,
JOURNAL_MAY_SKIP_FLUSH,
JOURNAL_NEED_FLUSH_WRITE,
+ JOURNAL_SPACE_LOW,
};
/* Reasons we may fail to get a journal reservation: */
diff --git a/libbcachefs/opts.c b/libbcachefs/opts.c
index e1800c41..bb068fd7 100644
--- a/libbcachefs/opts.c
+++ b/libbcachefs/opts.c
@@ -43,7 +43,7 @@ const char * const __bch2_btree_ids[] = {
NULL
};
-const char * const bch2_csum_types[] = {
+static const char * const __bch2_csum_types[] = {
BCH_CSUM_TYPES()
NULL
};
@@ -53,7 +53,7 @@ const char * const bch2_csum_opts[] = {
NULL
};
-const char * const __bch2_compression_types[] = {
+static const char * const __bch2_compression_types[] = {
BCH_COMPRESSION_TYPES()
NULL
};
@@ -83,18 +83,39 @@ const char * const bch2_member_states[] = {
NULL
};
-const char * const bch2_jset_entry_types[] = {
+static const char * const __bch2_jset_entry_types[] = {
BCH_JSET_ENTRY_TYPES()
NULL
};
-const char * const bch2_fs_usage_types[] = {
+static const char * const __bch2_fs_usage_types[] = {
BCH_FS_USAGE_TYPES()
NULL
};
#undef x
+static void prt_str_opt_boundscheck(struct printbuf *out, const char * const opts[],
+ unsigned nr, const char *type, unsigned idx)
+{
+ if (idx < nr)
+ prt_str(out, opts[idx]);
+ else
+ prt_printf(out, "(unknown %s %u)", type, idx);
+}
+
+#define PRT_STR_OPT_BOUNDSCHECKED(name, type) \
+void bch2_prt_##name(struct printbuf *out, type t) \
+{ \
+ prt_str_opt_boundscheck(out, __bch2_##name##s, ARRAY_SIZE(__bch2_##name##s) - 1, #name, t);\
+}
+
+PRT_STR_OPT_BOUNDSCHECKED(jset_entry_type, enum bch_jset_entry_type);
+PRT_STR_OPT_BOUNDSCHECKED(fs_usage_type, enum bch_fs_usage_type);
+PRT_STR_OPT_BOUNDSCHECKED(data_type, enum bch_data_type);
+PRT_STR_OPT_BOUNDSCHECKED(csum_type, enum bch_csum_type);
+PRT_STR_OPT_BOUNDSCHECKED(compression_type, enum bch_compression_type);
+
static int bch2_opt_fix_errors_parse(struct bch_fs *c, const char *val, u64 *res,
struct printbuf *err)
{
diff --git a/libbcachefs/opts.h b/libbcachefs/opts.h
index 1ac4135c..84e45283 100644
--- a/libbcachefs/opts.h
+++ b/libbcachefs/opts.h
@@ -16,18 +16,20 @@ extern const char * const bch2_version_upgrade_opts[];
extern const char * const bch2_sb_features[];
extern const char * const bch2_sb_compat[];
extern const char * const __bch2_btree_ids[];
-extern const char * const bch2_csum_types[];
extern const char * const bch2_csum_opts[];
-extern const char * const __bch2_compression_types[];
extern const char * const bch2_compression_opts[];
extern const char * const bch2_str_hash_types[];
extern const char * const bch2_str_hash_opts[];
extern const char * const __bch2_data_types[];
extern const char * const bch2_member_states[];
-extern const char * const bch2_jset_entry_types[];
-extern const char * const bch2_fs_usage_types[];
extern const char * const bch2_d_types[];
+void bch2_prt_jset_entry_type(struct printbuf *, enum bch_jset_entry_type);
+void bch2_prt_fs_usage_type(struct printbuf *, enum bch_fs_usage_type);
+void bch2_prt_data_type(struct printbuf *, enum bch_data_type);
+void bch2_prt_csum_type(struct printbuf *, enum bch_csum_type);
+void bch2_prt_compression_type(struct printbuf *, enum bch_compression_type);
+
static inline const char *bch2_d_type_str(unsigned d_type)
{
return (d_type < BCH_DT_MAX ? bch2_d_types[d_type] : NULL) ?: "(bad d_type)";
diff --git a/libbcachefs/recovery.c b/libbcachefs/recovery.c
index b76c1615..0f328aba 100644
--- a/libbcachefs/recovery.c
+++ b/libbcachefs/recovery.c
@@ -47,20 +47,6 @@ void bch2_btree_lost_data(struct bch_fs *c, enum btree_id btree)
}
}
-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:
- case BTREE_ID_bucket_gens:
- return true;
- default:
- return false;
- }
-}
-
/* for -o reconstruct_alloc: */
static void bch2_reconstruct_alloc(struct bch_fs *c)
{
diff --git a/libbcachefs/recovery_passes.c b/libbcachefs/recovery_passes.c
index cb501460..0cec0f7d 100644
--- a/libbcachefs/recovery_passes.c
+++ b/libbcachefs/recovery_passes.c
@@ -44,7 +44,7 @@ static int bch2_set_may_go_rw(struct bch_fs *c)
set_bit(BCH_FS_may_go_rw, &c->flags);
- if (keys->nr || c->opts.fsck || !c->sb.clean)
+ if (keys->nr || c->opts.fsck || !c->sb.clean || c->recovery_passes_explicit)
return bch2_fs_read_write_early(c);
return 0;
}
diff --git a/libbcachefs/sb-downgrade.c b/libbcachefs/sb-downgrade.c
index d6f81179..a98ef940 100644
--- a/libbcachefs/sb-downgrade.c
+++ b/libbcachefs/sb-downgrade.c
@@ -51,7 +51,10 @@
BCH_FSCK_ERR_subvol_fs_path_parent_wrong) \
x(btree_subvolume_children, \
BIT_ULL(BCH_RECOVERY_PASS_check_subvols), \
- BCH_FSCK_ERR_subvol_children_not_set)
+ BCH_FSCK_ERR_subvol_children_not_set) \
+ x(mi_btree_bitmap, \
+ BIT_ULL(BCH_RECOVERY_PASS_check_allocations), \
+ BCH_FSCK_ERR_btree_bitmap_not_marked)
#define DOWNGRADE_TABLE()
diff --git a/libbcachefs/sb-errors_types.h b/libbcachefs/sb-errors_types.h
index d7d60913..4ca6e7b0 100644
--- a/libbcachefs/sb-errors_types.h
+++ b/libbcachefs/sb-errors_types.h
@@ -130,7 +130,7 @@
x(bucket_gens_nonzero_for_invalid_buckets, 122) \
x(need_discard_freespace_key_to_invalid_dev_bucket, 123) \
x(need_discard_freespace_key_bad, 124) \
- x(backpointer_pos_wrong, 125) \
+ x(backpointer_bucket_offset_wrong, 125) \
x(backpointer_to_missing_device, 126) \
x(backpointer_to_missing_alloc, 127) \
x(backpointer_to_missing_ptr, 128) \
@@ -270,7 +270,8 @@
x(btree_ptr_v2_min_key_bad, 262) \
x(btree_root_unreadable_and_scan_found_nothing, 263) \
x(snapshot_node_missing, 264) \
- x(dup_backpointer_to_bad_csum_extent, 265)
+ x(dup_backpointer_to_bad_csum_extent, 265) \
+ x(btree_bitmap_not_marked, 266)
enum bch_sb_error_id {
#define x(t, n) BCH_FSCK_ERR_##t = n,
diff --git a/libbcachefs/sb-members.c b/libbcachefs/sb-members.c
index eff5ce18..522a9693 100644
--- a/libbcachefs/sb-members.c
+++ b/libbcachefs/sb-members.c
@@ -1,6 +1,7 @@
// SPDX-License-Identifier: GPL-2.0
#include "bcachefs.h"
+#include "btree_cache.h"
#include "disk_groups.h"
#include "opts.h"
#include "replicas.h"
@@ -426,3 +427,55 @@ void bch2_dev_errors_reset(struct bch_dev *ca)
bch2_write_super(c);
mutex_unlock(&c->sb_lock);
}
+
+/*
+ * Per member "range has btree nodes" bitmap:
+ *
+ * This is so that if we ever have to run the btree node scan to repair we don't
+ * have to scan full devices:
+ */
+
+bool bch2_dev_btree_bitmap_marked(struct bch_fs *c, struct bkey_s_c k)
+{
+ bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr)
+ if (!bch2_dev_btree_bitmap_marked_sectors(bch_dev_bkey_exists(c, ptr->dev),
+ ptr->offset, btree_sectors(c)))
+ return false;
+ return true;
+}
+
+static void __bch2_dev_btree_bitmap_mark(struct bch_sb_field_members_v2 *mi, unsigned dev,
+ u64 start, unsigned sectors)
+{
+ struct bch_member *m = __bch2_members_v2_get_mut(mi, dev);
+ u64 bitmap = le64_to_cpu(m->btree_allocated_bitmap);
+
+ u64 end = start + sectors;
+
+ int resize = ilog2(roundup_pow_of_two(end)) - (m->btree_bitmap_shift + 6);
+ if (resize > 0) {
+ u64 new_bitmap = 0;
+
+ for (unsigned i = 0; i < 64; i++)
+ if (bitmap & BIT_ULL(i))
+ new_bitmap |= BIT_ULL(i >> resize);
+ bitmap = new_bitmap;
+ m->btree_bitmap_shift += resize;
+ }
+
+ for (unsigned bit = sectors >> m->btree_bitmap_shift;
+ bit << m->btree_bitmap_shift < end;
+ bit++)
+ bitmap |= BIT_ULL(bit);
+
+ m->btree_allocated_bitmap = cpu_to_le64(bitmap);
+}
+
+void bch2_dev_btree_bitmap_mark(struct bch_fs *c, struct bkey_s_c k)
+{
+ lockdep_assert_held(&c->sb_lock);
+
+ struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
+ bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr)
+ __bch2_dev_btree_bitmap_mark(mi, ptr->dev, ptr->offset, btree_sectors(c));
+}
diff --git a/libbcachefs/sb-members.h b/libbcachefs/sb-members.h
index be0a9418..b27c3e44 100644
--- a/libbcachefs/sb-members.h
+++ b/libbcachefs/sb-members.h
@@ -3,6 +3,7 @@
#define _BCACHEFS_SB_MEMBERS_H
#include "darray.h"
+#include "bkey_types.h"
extern char * const bch2_member_error_strs[];
@@ -220,6 +221,8 @@ static inline struct bch_member_cpu bch2_mi_to_cpu(struct bch_member *mi)
: 1,
.freespace_initialized = BCH_MEMBER_FREESPACE_INITIALIZED(mi),
.valid = bch2_member_exists(mi),
+ .btree_bitmap_shift = mi->btree_bitmap_shift,
+ .btree_allocated_bitmap = le64_to_cpu(mi->btree_allocated_bitmap),
};
}
@@ -228,4 +231,22 @@ void bch2_sb_members_from_cpu(struct bch_fs *);
void bch2_dev_io_errors_to_text(struct printbuf *, struct bch_dev *);
void bch2_dev_errors_reset(struct bch_dev *);
+static inline bool bch2_dev_btree_bitmap_marked_sectors(struct bch_dev *ca, u64 start, unsigned sectors)
+{
+ u64 end = start + sectors;
+
+ if (end > 64 << ca->mi.btree_bitmap_shift)
+ return false;
+
+ for (unsigned bit = sectors >> ca->mi.btree_bitmap_shift;
+ bit << ca->mi.btree_bitmap_shift < end;
+ bit++)
+ if (!(ca->mi.btree_allocated_bitmap & BIT_ULL(bit)))
+ return false;
+ return true;
+}
+
+bool bch2_dev_btree_bitmap_marked(struct bch_fs *, struct bkey_s_c);
+void bch2_dev_btree_bitmap_mark(struct bch_fs *, struct bkey_s_c);
+
#endif /* _BCACHEFS_SB_MEMBERS_H */
diff --git a/libbcachefs/snapshot.c b/libbcachefs/snapshot.c
index 0e806f04..544322d5 100644
--- a/libbcachefs/snapshot.c
+++ b/libbcachefs/snapshot.c
@@ -125,6 +125,15 @@ static inline u32 get_ancestor_below(struct snapshot_table *t, u32 id, u32 ances
return s->parent;
}
+static bool test_ancestor_bitmap(struct snapshot_table *t, u32 id, u32 ancestor)
+{
+ const struct snapshot_t *s = __snapshot_t(t, id);
+ if (!s)
+ return false;
+
+ return test_bit(ancestor - id - 1, s->is_ancestor);
+}
+
bool __bch2_snapshot_is_ancestor(struct bch_fs *c, u32 id, u32 ancestor)
{
bool ret;
@@ -140,13 +149,11 @@ bool __bch2_snapshot_is_ancestor(struct bch_fs *c, u32 id, u32 ancestor)
while (id && id < ancestor - IS_ANCESTOR_BITMAP)
id = get_ancestor_below(t, id, ancestor);
- if (id && id < ancestor) {
- ret = test_bit(ancestor - id - 1, __snapshot_t(t, id)->is_ancestor);
+ ret = id && id < ancestor
+ ? test_ancestor_bitmap(t, id, ancestor)
+ : id == ancestor;
- EBUG_ON(ret != __bch2_snapshot_is_ancestor_early(t, id, ancestor));
- } else {
- ret = id == ancestor;
- }
+ EBUG_ON(ret != __bch2_snapshot_is_ancestor_early(t, id, ancestor));
out:
rcu_read_unlock();
diff --git a/libbcachefs/super-io.c b/libbcachefs/super-io.c
index e0aa3655..648986a7 100644
--- a/libbcachefs/super-io.c
+++ b/libbcachefs/super-io.c
@@ -700,8 +700,11 @@ retry:
return -ENOMEM;
sb->sb_name = kstrdup(path, GFP_KERNEL);
- if (!sb->sb_name)
- return -ENOMEM;
+ if (!sb->sb_name) {
+ ret = -ENOMEM;
+ prt_printf(&err, "error allocating memory for sb_name");
+ goto err;
+ }
#ifndef __KERNEL__
if (opt_get(*opts, direct_io) == false)
diff --git a/libbcachefs/super.c b/libbcachefs/super.c
index ed63018f..8daf80a3 100644
--- a/libbcachefs/super.c
+++ b/libbcachefs/super.c
@@ -288,8 +288,13 @@ static void __bch2_fs_read_only(struct bch_fs *c)
if (test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags) &&
!test_bit(BCH_FS_emergency_ro, &c->flags))
set_bit(BCH_FS_clean_shutdown, &c->flags);
+
bch2_fs_journal_stop(&c->journal);
+ bch_info(c, "%sshutdown complete, journal seq %llu",
+ test_bit(BCH_FS_clean_shutdown, &c->flags) ? "" : "un",
+ c->journal.seq_ondisk);
+
/*
* After stopping journal:
*/
diff --git a/libbcachefs/super_types.h b/libbcachefs/super_types.h
index ec784d97..11bcef17 100644
--- a/libbcachefs/super_types.h
+++ b/libbcachefs/super_types.h
@@ -37,6 +37,8 @@ struct bch_member_cpu {
u8 durability;
u8 freespace_initialized;
u8 valid;
+ u8 btree_bitmap_shift;
+ u64 btree_allocated_bitmap;
};
#endif /* _BCACHEFS_SUPER_TYPES_H */
diff --git a/libbcachefs/sysfs.c b/libbcachefs/sysfs.c
index c86a93a8..5be92fe3 100644
--- a/libbcachefs/sysfs.c
+++ b/libbcachefs/sysfs.c
@@ -17,7 +17,6 @@
#include "btree_iter.h"
#include "btree_key_cache.h"
#include "btree_update.h"
-#include "btree_update_interior.h"
#include "btree_gc.h"
#include "buckets.h"
#include "clock.h"
@@ -26,6 +25,7 @@
#include "ec.h"
#include "inode.h"
#include "journal.h"
+#include "journal_reclaim.h"
#include "keylist.h"
#include "move.h"
#include "movinggc.h"
@@ -139,6 +139,7 @@ do { \
write_attribute(trigger_gc);
write_attribute(trigger_discards);
write_attribute(trigger_invalidates);
+write_attribute(trigger_journal_flush);
write_attribute(prune_cache);
write_attribute(btree_wakeup);
rw_attribute(btree_gc_periodic);
@@ -166,7 +167,6 @@ read_attribute(btree_write_stats);
read_attribute(btree_cache_size);
read_attribute(compression_stats);
read_attribute(journal_debug);
-read_attribute(btree_updates);
read_attribute(btree_cache);
read_attribute(btree_key_cache);
read_attribute(stripes_heap);
@@ -415,9 +415,6 @@ SHOW(bch2_fs)
if (attr == &sysfs_journal_debug)
bch2_journal_debug_to_text(out, &c->journal);
- if (attr == &sysfs_btree_updates)
- bch2_btree_updates_to_text(out, c);
-
if (attr == &sysfs_btree_cache)
bch2_btree_cache_to_text(out, c);
@@ -505,7 +502,7 @@ STORE(bch2_fs)
/* Debugging: */
- if (!test_bit(BCH_FS_rw, &c->flags))
+ if (!bch2_write_ref_tryget(c, BCH_WRITE_REF_sysfs))
return -EROFS;
if (attr == &sysfs_prune_cache) {
@@ -538,6 +535,11 @@ STORE(bch2_fs)
if (attr == &sysfs_trigger_invalidates)
bch2_do_invalidates(c);
+ if (attr == &sysfs_trigger_journal_flush) {
+ bch2_journal_flush_all_pins(&c->journal);
+ bch2_journal_meta(&c->journal);
+ }
+
#ifdef CONFIG_BCACHEFS_TESTS
if (attr == &sysfs_perf_test) {
char *tmp = kstrdup(buf, GFP_KERNEL), *p = tmp;
@@ -558,6 +560,7 @@ STORE(bch2_fs)
size = ret;
}
#endif
+ bch2_write_ref_put(c, BCH_WRITE_REF_sysfs);
return size;
}
SYSFS_OPS(bch2_fs);
@@ -639,7 +642,6 @@ SYSFS_OPS(bch2_fs_internal);
struct attribute *bch2_fs_internal_files[] = {
&sysfs_flags,
&sysfs_journal_debug,
- &sysfs_btree_updates,
&sysfs_btree_cache,
&sysfs_btree_key_cache,
&sysfs_new_stripes,
@@ -657,6 +659,7 @@ struct attribute *bch2_fs_internal_files[] = {
&sysfs_trigger_gc,
&sysfs_trigger_discards,
&sysfs_trigger_invalidates,
+ &sysfs_trigger_journal_flush,
&sysfs_prune_cache,
&sysfs_btree_wakeup,
diff --git a/libbcachefs/tests.c b/libbcachefs/tests.c
index b3fe9fc5..bfec656f 100644
--- a/libbcachefs/tests.c
+++ b/libbcachefs/tests.c
@@ -672,7 +672,7 @@ static int __do_delete(struct btree_trans *trans, struct bpos pos)
bch2_trans_iter_init(trans, &iter, BTREE_ID_xattrs, pos,
BTREE_ITER_INTENT);
- k = bch2_btree_iter_peek(&iter);
+ k = bch2_btree_iter_peek_upto(&iter, POS(0, U64_MAX));
ret = bkey_err(k);
if (ret)
goto err;
diff --git a/libbcachefs/util.h b/libbcachefs/util.h
index b7e7c292..5cf885b0 100644
--- a/libbcachefs/util.h
+++ b/libbcachefs/util.h
@@ -788,6 +788,14 @@ static inline int copy_from_user_errcode(void *to, const void __user *from, unsi
#endif
+static inline void mod_bit(long nr, volatile unsigned long *addr, bool v)
+{
+ if (v)
+ set_bit(nr, addr);
+ else
+ clear_bit(nr, addr);
+}
+
static inline void __set_bit_le64(size_t bit, __le64 *addr)
{
addr[bit / 64] |= cpu_to_le64(BIT_ULL(bit % 64));
@@ -795,7 +803,7 @@ static inline void __set_bit_le64(size_t bit, __le64 *addr)
static inline void __clear_bit_le64(size_t bit, __le64 *addr)
{
- addr[bit / 64] &= !cpu_to_le64(BIT_ULL(bit % 64));
+ addr[bit / 64] &= ~cpu_to_le64(BIT_ULL(bit % 64));
}
static inline bool test_bit_le64(size_t bit, __le64 *addr)