bcachefs: Improve btree write buffer tracepoints

 - add a tracepoint for write_buffer_flush_sync; this is expensive
 - fix the write_buffer_flush_slowpath tracepoint
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>

fs/bcachefs/bcachefs_format.h

@@ -1578,7 +1578,9 @@ struct bch_sb_field_disk_groups {
 	x(write_super,					73)	\
 	x(trans_restart_would_deadlock_recursion_limit,	74)	\
 	x(trans_restart_write_buffer_flush,		75)	\
-	x(trans_restart_split_race,			76)
+	x(trans_restart_split_race,			76)	\
+	x(write_buffer_flush_slowpath,			77)	\
+	x(write_buffer_flush_sync,			78)
 
 enum bch_persistent_counters {
 #define x(t, n, ...) BCH_COUNTER_##t,

fs/bcachefs/btree_write_buffer.c

@@ -241,7 +241,7 @@ int __bch2_btree_write_buffer_flush(struct btree_trans *trans, unsigned commit_f
 	mutex_unlock(&wb->flush_lock);
 	return ret;
 slowpath:
-	trace_write_buffer_flush_slowpath(trans, i - keys, nr);
+	trace_and_count(c, write_buffer_flush_slowpath, trans, slowpath, nr);
 
 	/*
 	 * Now sort the rest by journal seq and bump the journal pin as we go.
@@ -277,8 +277,12 @@ int __bch2_btree_write_buffer_flush(struct btree_trans *trans, unsigned commit_f
 
 int bch2_btree_write_buffer_flush_sync(struct btree_trans *trans)
 {
+	struct bch_fs *c = trans->c;
+
+	trace_and_count(c, write_buffer_flush_sync, trans, _RET_IP_);
+
 	bch2_trans_unlock(trans);
-	mutex_lock(&trans->c->btree_write_buffer.flush_lock);
+	mutex_lock(&c->btree_write_buffer.flush_lock);
 	return __bch2_btree_write_buffer_flush(trans, 0, true);
 }
 

fs/bcachefs/trace.h

@@ -1334,21 +1334,38 @@ TRACE_EVENT(write_buffer_flush,
 		  __entry->nr, __entry->size, __entry->skipped, __entry->fast)
 );
 
+TRACE_EVENT(write_buffer_flush_sync,
+	TP_PROTO(struct btree_trans *trans, unsigned long caller_ip),
+	TP_ARGS(trans, caller_ip),
+
+	TP_STRUCT__entry(
+		__array(char,			trans_fn, 32	)
+		__field(unsigned long,		caller_ip	)
+	),
+
+	TP_fast_assign(
+		strscpy(__entry->trans_fn, trans->fn, sizeof(__entry->trans_fn));
+		__entry->caller_ip		= caller_ip;
+	),
+
+	TP_printk("%s %pS", __entry->trans_fn, (void *) __entry->caller_ip)
+);
+
 TRACE_EVENT(write_buffer_flush_slowpath,
-	TP_PROTO(struct btree_trans *trans, size_t nr, size_t size),
-	TP_ARGS(trans, nr, size),
+	TP_PROTO(struct btree_trans *trans, size_t slowpath, size_t total),
+	TP_ARGS(trans, slowpath, total),
 
 	TP_STRUCT__entry(
-		__field(size_t,		nr		)
-		__field(size_t,		size		)
+		__field(size_t,		slowpath	)
+		__field(size_t,		total		)
 	),
 
 	TP_fast_assign(
-		__entry->nr	= nr;
-		__entry->size	= size;
+		__entry->slowpath	= slowpath;
+		__entry->total		= total;
 	),
 
-	TP_printk("%zu/%zu", __entry->nr, __entry->size)
+	TP_printk("%zu/%zu", __entry->slowpath, __entry->total)
 );
 
 DEFINE_EVENT(str, rebalance_extent,