diff --git a/include/linux/ftrace_event.h b/include/linux/ftrace_event.h
index 43360c1..47bbdf9 100644
--- a/include/linux/ftrace_event.h
+++ b/include/linux/ftrace_event.h
@@ -137,13 +137,8 @@
 
 #define FTRACE_MAX_PROFILE_SIZE	2048
 
-struct perf_trace_buf {
-	char	buf[FTRACE_MAX_PROFILE_SIZE];
-	int	recursion;
-};
-
-extern struct perf_trace_buf	*perf_trace_buf;
-extern struct perf_trace_buf	*perf_trace_buf_nmi;
+extern char *perf_trace_buf;
+extern char *perf_trace_buf_nmi;
 
 #define MAX_FILTER_PRED		32
 #define MAX_FILTER_STR_VAL	256	/* Should handle KSYM_SYMBOL_LEN */
diff --git a/include/linux/perf_event.h b/include/linux/perf_event.h
index 36fe89f..74e98b1 100644
--- a/include/linux/perf_event.h
+++ b/include/linux/perf_event.h
@@ -874,6 +874,8 @@
 extern void perf_output_end(struct perf_output_handle *handle);
 extern void perf_output_copy(struct perf_output_handle *handle,
 			     const void *buf, unsigned int len);
+extern int perf_swevent_get_recursion_context(int **recursion);
+extern void perf_swevent_put_recursion_context(int *recursion);
 #else
 static inline void
 perf_event_task_sched_in(struct task_struct *task, int cpu)		{ }
@@ -902,6 +904,8 @@
 static inline void perf_event_comm(struct task_struct *tsk)		{ }
 static inline void perf_event_fork(struct task_struct *tsk)		{ }
 static inline void perf_event_init(void)				{ }
+static int perf_swevent_get_recursion_context(int **recursion)	{ return -1; }
+static void perf_swevent_put_recursion_context(int *recursion)		{ }
 
 #endif
 
diff --git a/include/trace/ftrace.h b/include/trace/ftrace.h
index 4945d1c9..c222ef5 100644
--- a/include/trace/ftrace.h
+++ b/include/trace/ftrace.h
@@ -724,16 +724,19 @@
 static void ftrace_profile_##call(proto)				\
 {									\
 	struct ftrace_data_offsets_##call __maybe_unused __data_offsets;\
+	extern int perf_swevent_get_recursion_context(int **recursion); \
+	extern void perf_swevent_put_recursion_context(int *recursion); \
 	struct ftrace_event_call *event_call = &event_##call;		\
 	extern void perf_tp_event(int, u64, u64, void *, int);		\
 	struct ftrace_raw_##call *entry;				\
-	struct perf_trace_buf *trace_buf;				\
 	u64 __addr = 0, __count = 1;					\
 	unsigned long irq_flags;					\
 	struct trace_entry *ent;					\
 	int __entry_size;						\
 	int __data_size;						\
+	char *trace_buf;						\
 	char *raw_data;							\
+	int *recursion;							\
 	int __cpu;							\
 	int pc;								\
 									\
@@ -749,6 +752,10 @@
 		return;							\
 									\
 	local_irq_save(irq_flags);					\
+									\
+	if (perf_swevent_get_recursion_context(&recursion))		\
+		goto end_recursion;						\
+									\
 	__cpu = smp_processor_id();					\
 									\
 	if (in_nmi())							\
@@ -759,13 +766,7 @@
 	if (!trace_buf)							\
 		goto end;						\
 									\
-	trace_buf = per_cpu_ptr(trace_buf, __cpu);			\
-	if (trace_buf->recursion++)					\
-		goto end_recursion;					\
-									\
-	barrier();							\
-									\
-	raw_data = trace_buf->buf;					\
+	raw_data = per_cpu_ptr(trace_buf, __cpu);			\
 									\
 	*(u64 *)(&raw_data[__entry_size - sizeof(u64)]) = 0ULL;		\
 	entry = (struct ftrace_raw_##call *)raw_data;			\
@@ -780,9 +781,9 @@
 	perf_tp_event(event_call->id, __addr, __count, entry,		\
 			     __entry_size);				\
 									\
-end_recursion:								\
-	trace_buf->recursion--;						\
-end:									\
+end:								\
+	perf_swevent_put_recursion_context(recursion);			\
+end_recursion:									\
 	local_irq_restore(irq_flags);					\
 									\
 }