bcachefs: btree write buffer now slurps keys from journal

Previosuly, the transaction commit path would have to add keys to the
btree write buffer as a separate operation, requiring additional global
synchronization.

This patch introduces a new journal entry type, which indicates that the
keys need to be copied into the btree write buffer prior to being
written out. We switch the journal entry type back to
JSET_ENTRY_btree_keys prior to write, so this is not an on disk format
change.

Flushing the btree write buffer may require pulling keys out of journal
entries yet to be written, and quiescing outstanding journal
reservations; we previously added journal->buf_lock for synchronization
with the journal write path.

We also can't put strict bounds on the number of keys in the journal
destined for the write buffer, which means we might overflow the size of
the preallocated buffer and have to reallocate - this introduces a
potentially fatal memory allocation failure. This is something we'll
have to watch for, if it becomes an issue in practice we can do
additional mitigation.

The transaction commit path no longer has to explicitly check if the
write buffer is full and wait on flushing; this is another performance
optimization. Instead, when the btree write buffer is close to full we
change the journal watermark, so that only reservations for journal
reclaim are allowed.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

1 files changed, 38 insertions, 25 deletions

diff --git a/fs/bcachefs/btree_write_buffer_types.h b/fs/bcachefs/btree_write_buffer_types.h
index 99993ba77aea..8758d6adabf4 100644
--- a/fs/bcachefs/btree_write_buffer_types.h
+++ b/fs/bcachefs/btree_write_buffer_types.h
@@ -2,43 +2,56 @@
 #ifndef _BCACHEFS_BTREE_WRITE_BUFFER_TYPES_H
 #define _BCACHEFS_BTREE_WRITE_BUFFER_TYPES_H
 
+#include "darray.h"
 #include "journal_types.h"
 
 #define BTREE_WRITE_BUFERED_VAL_U64s_MAX	4
 #define BTREE_WRITE_BUFERED_U64s_MAX	(BKEY_U64s + BTREE_WRITE_BUFERED_VAL_U64s_MAX)
 
-struct btree_write_buffered_key {
-	u64			journal_seq;
-	unsigned		journal_offset;
-	enum btree_id		btree;
-	__BKEY_PADDED(k, BTREE_WRITE_BUFERED_VAL_U64s_MAX);
-};
-
-union btree_write_buffer_state {
+struct wb_key_ref {
+union {
 	struct {
-		atomic64_t	counter;
-	};
-
-	struct {
-		u64		v;
-	};
-
+#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
+		unsigned			idx:24;
+		struct bpos			pos;
+		enum btree_id			btree:8;
+#else
+		enum btree_id			btree:8;
+		struct bpos			pos;
+		unsigned			idx:24;
+#endif
+	} __packed;
 	struct {
-		u64			nr:23;
-		u64			idx:1;
-		u64			ref0:20;
-		u64			ref1:20;
+#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
+		u64 lo;
+		u64 mi;
+		u64 hi;
+#else
+		u64 hi;
+		u64 mi;
+		u64 lo;
+#endif
 	};
 };
+};
 
-struct btree_write_buffer {
-	struct mutex			flush_lock;
-	struct journal_entry_pin	journal_pin;
+struct btree_write_buffered_key {
+	enum btree_id			btree:8;
+	u64				journal_seq:56;
+	__BKEY_PADDED(k, BTREE_WRITE_BUFERED_VAL_U64s_MAX);
+};
 
-	union btree_write_buffer_state	state;
-	size_t				size;
+struct btree_write_buffer_keys {
+	DARRAY(struct btree_write_buffered_key) keys;
+	struct journal_entry_pin	pin;
+	struct mutex			lock;
+};
 
-	struct btree_write_buffered_key	*keys[2];
+struct btree_write_buffer {
+	DARRAY(struct wb_key_ref)	sorted;
+	struct btree_write_buffer_keys	inc;
+	struct btree_write_buffer_keys	flushing;
+	struct work_struct		flush_work;
 };
 
 #endif /* _BCACHEFS_BTREE_WRITE_BUFFER_TYPES_H */