ftrace: take advantage of variable length entries

Now that the underlining ring buffer for ftrace now hold variable length
entries, we can take advantage of this by only storing the size of the
actual event into the buffer. This happens to increase the number of
entries in the buffer dramatically.

We can also get rid of the "trace_cont" operation, but I'm keeping that
until we have no more users. Some of the ftrace tracers can now change
their code to adapt to this new feature.
Signed-off-by: Steven Rostedt <srostedt@redhat.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>

kernel/trace/trace.h

@@ -26,10 +26,25 @@ enum trace_type {
 	__TRACE_LAST_TYPE
 };
 
+/*
+ * The trace entry - the most basic unit of tracing. This is what
+ * is printed in the end as a single line in the trace output, such as:
+ *
+ *     bash-15816 [01]   235.197585: idle_cpu <- irq_enter
+ */
+struct trace_entry {
+	unsigned char		type;
+	unsigned char		cpu;
+	unsigned char		flags;
+	unsigned char		preempt_count;
+	int			pid;
+};
+
 /*
  * Function trace entry - function address and parent function addres:
  */
 struct ftrace_entry {
+	struct trace_entry	ent;
 	unsigned long		ip;
 	unsigned long		parent_ip;
 };
@@ -39,6 +54,7 @@ extern struct tracer boot_tracer;
  * Context switch trace entry - which task (and prio) we switched from/to:
  */
 struct ctx_switch_entry {
+	struct trace_entry	ent;
 	unsigned int		prev_pid;
 	unsigned char		prev_prio;
 	unsigned char		prev_state;
@@ -52,6 +68,7 @@ struct ctx_switch_entry {
  * Special (free-form) trace entry:
  */
 struct special_entry {
+	struct trace_entry	ent;
 	unsigned long		arg1;
 	unsigned long		arg2;
 	unsigned long		arg3;
@@ -64,6 +81,7 @@ struct special_entry {
 #define FTRACE_STACK_ENTRIES	8
 
 struct stack_entry {
+	struct trace_entry	ent;
 	unsigned long		caller[FTRACE_STACK_ENTRIES];
 };
 
@@ -71,10 +89,34 @@ struct stack_entry {
  * ftrace_printk entry:
  */
 struct print_entry {
+	struct trace_entry	ent;
 	unsigned long		ip;
 	char			buf[];
 };
 
+#define TRACE_OLD_SIZE		88
+
+struct trace_field_cont {
+	unsigned char		type;
+	/* Temporary till we get rid of this completely */
+	char			buf[TRACE_OLD_SIZE - 1];
+};
+
+struct trace_mmiotrace_rw {
+	struct trace_entry	ent;
+	struct mmiotrace_rw	rw;
+};
+
+struct trace_mmiotrace_map {
+	struct trace_entry	ent;
+	struct mmiotrace_map	map;
+};
+
+struct trace_boot {
+	struct trace_entry	ent;
+	struct boot_trace	initcall;
+};
+
 /*
  * trace_flag_type is an enumeration that holds different
  * states when a trace occurs. These are:
@@ -92,46 +134,7 @@ enum trace_flag_type {
 	TRACE_FLAG_CONT			= 0x10,
 };
 
-/*
- * The trace field - the most basic unit of tracing. This is what
- * is printed in the end as a single line in the trace output, such as:
- *
- *     bash-15816 [01]   235.197585: idle_cpu <- irq_enter
- */
-struct trace_field {
-	char			cpu;
-	char			flags;
-	char			preempt_count;
-	int			pid;
-	union {
-		struct ftrace_entry		fn;
-		struct ctx_switch_entry		ctx;
-		struct special_entry		special;
-		struct stack_entry		stack;
-		struct print_entry		print;
-		struct mmiotrace_rw		mmiorw;
-		struct mmiotrace_map		mmiomap;
-		struct boot_trace		initcall;
-	};
-};
-
-struct trace_field_cont {
-	char				buf[sizeof(struct trace_field)];
-};
-
-struct trace_entry {
-	char 				type;
-	union {
-		struct trace_field	field;
-		struct trace_field_cont	cont;
-	};
-};
-
-#define TRACE_ENTRY_SIZE	sizeof(struct trace_entry)
 #define TRACE_BUF_SIZE		1024
-#define TRACE_PRINT_BUF_SIZE \
-	(sizeof(struct trace_field) - offsetof(struct trace_field, print.buf))
-#define TRACE_CONT_BUF_SIZE	sizeof(struct trace_field)
 
 /*
  * The CPU trace array - it consists of thousands of trace entries

kernel/trace/trace_boot.c

@@ -49,10 +49,11 @@ static int initcall_print_line(struct trace_iterator *iter)
 {
 	int ret = 0;
 	struct trace_entry *entry = iter->ent;
-	struct boot_trace *it = &entry->field.initcall;
+	struct trace_boot *field = (struct trace_boot *)entry;
+	struct boot_trace *it = &field->initcall;
 	struct trace_seq *s = &iter->seq;
 
-	if (iter->ent->type == TRACE_BOOT)
+	if (entry->type == TRACE_BOOT)
 		ret = trace_seq_printf(s, "%pF called from %i "
 				       "returned %d after %lld msecs\n",
 				       it->func, it->caller, it->result,
@@ -75,7 +76,7 @@ struct tracer boot_tracer __read_mostly =
 void trace_boot(struct boot_trace *it)
 {
 	struct ring_buffer_event *event;
-	struct trace_entry *entry;
+	struct trace_boot *entry;
 	struct trace_array_cpu *data;
 	unsigned long irq_flags;
 	struct trace_array *tr = boot_trace;
@@ -91,9 +92,9 @@ void trace_boot(struct boot_trace *it)
 	if (!event)
 		goto out;
 	entry	= ring_buffer_event_data(event);
-	tracing_generic_entry_update(entry, 0);
-	entry->type = TRACE_BOOT;
-	entry->field.initcall = *it;
+	tracing_generic_entry_update(&entry->ent, 0);
+	entry->ent.type = TRACE_BOOT;
+	entry->initcall = *it;
 	ring_buffer_unlock_commit(tr->buffer, event, irq_flags);
 
 	trace_wake_up();

kernel/trace/trace_mmiotrace.c

@@ -178,14 +178,16 @@ static ssize_t mmio_read(struct trace_iterator *iter, struct file *filp,
 static int mmio_print_rw(struct trace_iterator *iter)
 {
 	struct trace_entry *entry = iter->ent;
-	struct mmiotrace_rw *rw	= &entry->field.mmiorw;
+	struct trace_mmiotrace_rw *field =
+		(struct trace_mmiotrace_rw *)entry;
+	struct mmiotrace_rw *rw	= &field->rw;
 	struct trace_seq *s	= &iter->seq;
 	unsigned long long t	= ns2usecs(iter->ts);
 	unsigned long usec_rem	= do_div(t, 1000000ULL);
 	unsigned secs		= (unsigned long)t;
 	int ret = 1;
 
-	switch (entry->field.mmiorw.opcode) {
+	switch (rw->opcode) {
 	case MMIO_READ:
 		ret = trace_seq_printf(s,
 			"R %d %lu.%06lu %d 0x%llx 0x%lx 0x%lx %d\n",
@@ -220,14 +222,14 @@ static int mmio_print_rw(struct trace_iterator *iter)
 static int mmio_print_map(struct trace_iterator *iter)
 {
 	struct trace_entry *entry = iter->ent;
-	struct mmiotrace_map *m	= &entry->field.mmiomap;
+	struct mmiotrace_map *m	= (struct mmiotrace_map *)entry;
 	struct trace_seq *s	= &iter->seq;
 	unsigned long long t	= ns2usecs(iter->ts);
 	unsigned long usec_rem	= do_div(t, 1000000ULL);
 	unsigned secs		= (unsigned long)t;
 	int ret = 1;
 
-	switch (entry->field.mmiorw.opcode) {
+	switch (m->opcode) {
 	case MMIO_PROBE:
 		ret = trace_seq_printf(s,
 			"MAP %lu.%06lu %d 0x%llx 0x%lx 0x%lx 0x%lx %d\n",
@@ -252,7 +254,8 @@ static int mmio_print_map(struct trace_iterator *iter)
 static int mmio_print_mark(struct trace_iterator *iter)
 {
 	struct trace_entry *entry = iter->ent;
-	const char *msg		= entry->field.print.buf;
+	struct print_entry *print = (struct print_entry *)entry;
+	const char *msg		= print->buf;
 	struct trace_seq *s	= &iter->seq;
 	unsigned long long t	= ns2usecs(iter->ts);
 	unsigned long usec_rem	= do_div(t, 1000000ULL);
@@ -264,7 +267,7 @@ static int mmio_print_mark(struct trace_iterator *iter)
 	if (!ret)
 		return 0;
 
-	if (entry->field.flags & TRACE_FLAG_CONT)
+	if (entry->flags & TRACE_FLAG_CONT)
 		trace_seq_print_cont(s, iter);
 
 	return 1;
@@ -308,7 +311,7 @@ static void __trace_mmiotrace_rw(struct trace_array *tr,
 				struct mmiotrace_rw *rw)
 {
 	struct ring_buffer_event *event;
-	struct trace_entry *entry;
+	struct trace_mmiotrace_rw *entry;
 	unsigned long irq_flags;
 
 	event	= ring_buffer_lock_reserve(tr->buffer, sizeof(*entry),
@@ -316,9 +319,9 @@ static void __trace_mmiotrace_rw(struct trace_array *tr,
 	if (!event)
 		return;
 	entry	= ring_buffer_event_data(event);
-	tracing_generic_entry_update(entry, 0);
-	entry->type			= TRACE_MMIO_RW;
-	entry->field.mmiorw		= *rw;
+	tracing_generic_entry_update(&entry->ent, 0);
+	entry->ent.type			= TRACE_MMIO_RW;
+	entry->rw			= *rw;
 	ring_buffer_unlock_commit(tr->buffer, event, irq_flags);
 
 	trace_wake_up();
@@ -336,7 +339,7 @@ static void __trace_mmiotrace_map(struct trace_array *tr,
 				struct mmiotrace_map *map)
 {
 	struct ring_buffer_event *event;
-	struct trace_entry *entry;
+	struct trace_mmiotrace_map *entry;
 	unsigned long irq_flags;
 
 	event	= ring_buffer_lock_reserve(tr->buffer, sizeof(*entry),
@@ -344,9 +347,9 @@ static void __trace_mmiotrace_map(struct trace_array *tr,
 	if (!event)
 		return;
 	entry	= ring_buffer_event_data(event);
-	tracing_generic_entry_update(entry, 0);
-	entry->type			= TRACE_MMIO_MAP;
-	entry->field.mmiomap		= *map;
+	tracing_generic_entry_update(&entry->ent, 0);
+	entry->ent.type			= TRACE_MMIO_MAP;
+	entry->map			= *map;
 	ring_buffer_unlock_commit(tr->buffer, event, irq_flags);
 
 	trace_wake_up();