X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fbtree_io.h;h=c43fb60b8c82c2ec4f066ec0ff752e4a4baf7401;hb=ba88873432ec9621b9913c00b31d7401357b5e55;hp=1f16394fd5c35249b835ef5ce3116b70d5a4a1b7;hpb=fad8236b812f795993b88804065d950709a6c13c;p=bcachefs-tools-debian

diff --git a/libbcachefs/btree_io.h b/libbcachefs/btree_io.h
index 1f16394..c43fb60 100644
--- a/libbcachefs/btree_io.h
+++ b/libbcachefs/btree_io.h
@@ -5,6 +5,7 @@
 #include "bkey_methods.h"
 #include "bset.h"
 #include "btree_locking.h"
+#include "checksum.h"
 #include "extents.h"
 #include "io_types.h"
 
@@ -12,46 +13,54 @@ struct bch_fs;
 struct btree_write;
 struct btree;
 struct btree_iter;
+struct btree_node_read_all;
+
+static inline void set_btree_node_dirty_acct(struct bch_fs *c, struct btree *b)
+{
+	if (!test_and_set_bit(BTREE_NODE_dirty, &b->flags))
+		atomic_inc(&c->btree_cache.dirty);
+}
+
+static inline void clear_btree_node_dirty_acct(struct bch_fs *c, struct btree *b)
+{
+	if (test_and_clear_bit(BTREE_NODE_dirty, &b->flags))
+		atomic_dec(&c->btree_cache.dirty);
+}
+
+static inline unsigned btree_ptr_sectors_written(struct bkey_i *k)
+{
+	return k->k.type == KEY_TYPE_btree_ptr_v2
+		? le16_to_cpu(bkey_i_to_btree_ptr_v2(k)->v.sectors_written)
+		: 0;
+}
 
 struct btree_read_bio {
 	struct bch_fs		*c;
+	struct btree		*b;
+	struct btree_node_read_all *ra;
 	u64			start_time;
 	unsigned		have_ioref:1;
+	unsigned		idx:7;
 	struct extent_ptr_decoded	pick;
 	struct work_struct	work;
 	struct bio		bio;
 };
 
 struct btree_write_bio {
-	void			*data;
 	struct work_struct	work;
+	__BKEY_PADDED(key, BKEY_BTREE_PTR_VAL_U64s_MAX);
+	void			*data;
+	unsigned		data_bytes;
+	unsigned		sector_offset;
 	struct bch_write_bio	wbio;
 };
 
-static inline void btree_node_io_unlock(struct btree *b)
-{
-	EBUG_ON(!btree_node_write_in_flight(b));
-	clear_btree_node_write_in_flight(b);
-	wake_up_bit(&b->flags, BTREE_NODE_write_in_flight);
-}
-
-static inline void btree_node_io_lock(struct btree *b)
-{
-	wait_on_bit_lock_io(&b->flags, BTREE_NODE_write_in_flight,
-			    TASK_UNINTERRUPTIBLE);
-}
-
-static inline void btree_node_wait_on_io(struct btree *b)
-{
-	wait_on_bit_io(&b->flags, BTREE_NODE_write_in_flight,
-		       TASK_UNINTERRUPTIBLE);
-}
-
-static inline bool btree_node_may_write(struct btree *b)
-{
-	return list_empty_careful(&b->write_blocked) &&
-		(!b->written || !b->will_make_reachable);
-}
+void bch2_btree_node_io_unlock(struct btree *);
+void bch2_btree_node_io_lock(struct btree *);
+void __bch2_btree_node_wait_on_read(struct btree *);
+void __bch2_btree_node_wait_on_write(struct btree *);
+void bch2_btree_node_wait_on_read(struct btree *);
+void bch2_btree_node_wait_on_write(struct btree *);
 
 enum compact_mode {
 	COMPACT_LAZY,
@@ -81,77 +90,96 @@ static inline bool bch2_maybe_compact_whiteouts(struct bch_fs *c, struct btree *
 	return false;
 }
 
+static inline struct nonce btree_nonce(struct bset *i, unsigned offset)
+{
+	return (struct nonce) {{
+		[0] = cpu_to_le32(offset),
+		[1] = ((__le32 *) &i->seq)[0],
+		[2] = ((__le32 *) &i->seq)[1],
+		[3] = ((__le32 *) &i->journal_seq)[0]^BCH_NONCE_BTREE,
+	}};
+}
+
+static inline int bset_encrypt(struct bch_fs *c, struct bset *i, unsigned offset)
+{
+	struct nonce nonce = btree_nonce(i, offset);
+	int ret;
+
+	if (!offset) {
+		struct btree_node *bn = container_of(i, struct btree_node, keys);
+		unsigned bytes = (void *) &bn->keys - (void *) &bn->flags;
+
+		ret = bch2_encrypt(c, BSET_CSUM_TYPE(i), nonce,
+				   &bn->flags, bytes);
+		if (ret)
+			return ret;
+
+		nonce = nonce_add(nonce, round_up(bytes, CHACHA_BLOCK_SIZE));
+	}
+
+	return bch2_encrypt(c, BSET_CSUM_TYPE(i), nonce, i->_data,
+			    vstruct_end(i) - (void *) i->_data);
+}
+
 void bch2_btree_sort_into(struct bch_fs *, struct btree *, struct btree *);
 
+void bch2_btree_node_drop_keys_outside_node(struct btree *);
+
 void bch2_btree_build_aux_trees(struct btree *);
-void bch2_btree_init_next(struct bch_fs *, struct btree *,
-			 struct btree_iter *);
+void bch2_btree_init_next(struct btree_trans *, struct btree *);
 
-int bch2_btree_node_read_done(struct bch_fs *, struct btree *, bool);
+int bch2_btree_node_read_done(struct bch_fs *, struct bch_dev *,
+			      struct btree *, bool, bool *);
 void bch2_btree_node_read(struct bch_fs *, struct btree *, bool);
 int bch2_btree_root_read(struct bch_fs *, enum btree_id,
 			 const struct bkey_i *, unsigned);
 
 void bch2_btree_complete_write(struct bch_fs *, struct btree *,
 			      struct btree_write *);
-void bch2_btree_write_error_work(struct work_struct *);
 
-void __bch2_btree_node_write(struct bch_fs *, struct btree *,
-			    enum six_lock_type);
 bool bch2_btree_post_write_cleanup(struct bch_fs *, struct btree *);
 
+enum btree_write_flags {
+	__BTREE_WRITE_ONLY_IF_NEED = BTREE_WRITE_TYPE_BITS,
+	__BTREE_WRITE_ALREADY_STARTED,
+};
+#define BTREE_WRITE_ONLY_IF_NEED	(1U << __BTREE_WRITE_ONLY_IF_NEED )
+#define BTREE_WRITE_ALREADY_STARTED	(1U << __BTREE_WRITE_ALREADY_STARTED)
+
+void __bch2_btree_node_write(struct bch_fs *, struct btree *, unsigned);
 void bch2_btree_node_write(struct bch_fs *, struct btree *,
-			  enum six_lock_type);
+			   enum six_lock_type, unsigned);
 
 static inline void btree_node_write_if_need(struct bch_fs *c, struct btree *b,
 					    enum six_lock_type lock_held)
 {
-	while (b->written &&
-	       btree_node_need_write(b) &&
-	       btree_node_may_write(b)) {
-		if (!btree_node_write_in_flight(b)) {
-			bch2_btree_node_write(c, b, lock_held);
-			break;
-		}
-
-		six_unlock_read(&b->lock);
-		btree_node_wait_on_io(b);
-		btree_node_lock_type(c, b, lock_held);
-	}
+	bch2_btree_node_write(c, b, lock_held, BTREE_WRITE_ONLY_IF_NEED);
 }
 
-#define bch2_btree_node_write_cond(_c, _b, cond)			\
-do {									\
-	unsigned long old, new, v = READ_ONCE((_b)->flags);		\
-									\
-	do {								\
-		old = new = v;						\
-									\
-		if (!(old & (1 << BTREE_NODE_dirty)) || !(cond))	\
-			break;						\
-									\
-		new |= (1 << BTREE_NODE_need_write);			\
-	} while ((v = cmpxchg(&(_b)->flags, old, new)) != old);		\
-									\
-	btree_node_write_if_need(_c, _b, SIX_LOCK_read);		\
-} while (0)
-
-void bch2_btree_flush_all_reads(struct bch_fs *);
-void bch2_btree_flush_all_writes(struct bch_fs *);
-void bch2_btree_verify_flushed(struct bch_fs *);
-ssize_t bch2_dirty_btree_nodes_print(struct bch_fs *, char *);
+bool bch2_btree_flush_all_reads(struct bch_fs *);
+bool bch2_btree_flush_all_writes(struct bch_fs *);
 
 static inline void compat_bformat(unsigned level, enum btree_id btree_id,
-				 unsigned version, unsigned big_endian,
-				 int write, struct bkey_format *f)
+				  unsigned version, unsigned big_endian,
+				  int write, struct bkey_format *f)
 {
 	if (version < bcachefs_metadata_version_inode_btree_change &&
-	    btree_id == BTREE_ID_INODES) {
+	    btree_id == BTREE_ID_inodes) {
 		swap(f->bits_per_field[BKEY_FIELD_INODE],
 		     f->bits_per_field[BKEY_FIELD_OFFSET]);
 		swap(f->field_offset[BKEY_FIELD_INODE],
 		     f->field_offset[BKEY_FIELD_OFFSET]);
 	}
+
+	if (version < bcachefs_metadata_version_snapshot &&
+	    (level || btree_type_has_snapshots(btree_id))) {
+		u64 max_packed =
+			~(~0ULL << f->bits_per_field[BKEY_FIELD_SNAPSHOT]);
+
+		f->field_offset[BKEY_FIELD_SNAPSHOT] = write
+			? 0
+			: U32_MAX - max_packed;
+	}
 }
 
 static inline void compat_bpos(unsigned level, enum btree_id btree_id,
@@ -162,7 +190,7 @@ static inline void compat_bpos(unsigned level, enum btree_id btree_id,
 		bch2_bpos_swab(p);
 
 	if (version < bcachefs_metadata_version_inode_btree_change &&
-	    btree_id == BTREE_ID_INODES)
+	    btree_id == BTREE_ID_inodes)
 		swap(p->inode, p->offset);
 }
 
@@ -173,18 +201,28 @@ static inline void compat_btree_node(unsigned level, enum btree_id btree_id,
 {
 	if (version < bcachefs_metadata_version_inode_btree_change &&
 	    btree_node_type_is_extents(btree_id) &&
-	    bkey_cmp(bn->min_key, POS_MIN) &&
+	    !bpos_eq(bn->min_key, POS_MIN) &&
 	    write)
-		bn->min_key = bkey_predecessor(bn->min_key);
+		bn->min_key = bpos_nosnap_predecessor(bn->min_key);
+
+	if (version < bcachefs_metadata_version_snapshot &&
+	    write)
+		bn->max_key.snapshot = 0;
 
 	compat_bpos(level, btree_id, version, big_endian, write, &bn->min_key);
 	compat_bpos(level, btree_id, version, big_endian, write, &bn->max_key);
 
+	if (version < bcachefs_metadata_version_snapshot &&
+	    !write)
+		bn->max_key.snapshot = U32_MAX;
+
 	if (version < bcachefs_metadata_version_inode_btree_change &&
 	    btree_node_type_is_extents(btree_id) &&
-	    bkey_cmp(bn->min_key, POS_MIN) &&
+	    !bpos_eq(bn->min_key, POS_MIN) &&
 	    !write)
-		bn->min_key = bkey_successor(bn->min_key);
+		bn->min_key = bpos_nosnap_successor(bn->min_key);
 }
 
+void bch2_btree_write_stats_to_text(struct printbuf *, struct bch_fs *);
+
 #endif /* _BCACHEFS_BTREE_IO_H */