perf script: Add callindent option

Based on patches from Andi Kleen.

When printing PT instruction traces with perf script it is rather useful to see some indentation for the call tree. This patch adds a new callindent field to perf script that prints spaces for the function call stack depth.

We already have code to track the function call stack for PT, that we can reuse with minor modifications.

The resulting output is not quite as nice as ftrace yet, but a lot better than what was there before.

Note there are some corner cases when the thread stack gets code confused and prints incorrect indentation. Even with that it is fairly useful.

When displaying kernel code traces it is recommended to run as root, as otherwise perf doesn't understand the kernel addresses properly, and may not reset the call stack correctly on kernel boundaries.

Example output:

	sudo perf-with-kcore record eg2 -a -e intel_pt// -- sleep 1
	sudo perf-with-kcore script eg2 --ns -F callindent,time,comm,pid,sym,ip,addr,flags,cpu --itrace=cre | less
	...
         swapper     0 [000]  5830.389116586:   call        irq_exit                                                     ffffffff8104d620 smp_call_function_single_interrupt+0x30 => ffffffff8107e720 irq_exit
         swapper     0 [000]  5830.389116586:   call            idle_cpu                                                 ffffffff8107e769 irq_exit+0x49 => ffffffff810a3970 idle_cpu
         swapper     0 [000]  5830.389116586:   return          idle_cpu                                                 ffffffff810a39b7 idle_cpu+0x47 => ffffffff8107e76e irq_exit
         swapper     0 [000]  5830.389116586:   call            tick_nohz_irq_exit                                       ffffffff8107e7bd irq_exit+0x9d => ffffffff810f2fc0 tick_nohz_irq_exit
         swapper     0 [000]  5830.389116919:   call                __tick_nohz_idle_enter                               ffffffff810f2fe0 tick_nohz_irq_exit+0x20 => ffffffff810f28d0 __tick_nohz_idle_enter
         swapper     0 [000]  5830.389116919:   call                    ktime_get                                        ffffffff810f28f1 __tick_nohz_idle_enter+0x21 => ffffffff810e9ec0 ktime_get
         swapper     0 [000]  5830.389116919:   call                        read_tsc                                     ffffffff810e9ef6 ktime_get+0x36 => ffffffff81035070 read_tsc
         swapper     0 [000]  5830.389116919:   return                      read_tsc                                     ffffffff81035084 read_tsc+0x14 => ffffffff810e9efc ktime_get
         swapper     0 [000]  5830.389116919:   return                  ktime_get                                        ffffffff810e9f46 ktime_get+0x86 => ffffffff810f28f6 __tick_nohz_idle_enter
         swapper     0 [000]  5830.389116919:   call                    sched_clock_idle_sleep_event                     ffffffff810f290b __tick_nohz_idle_enter+0x3b => ffffffff810a7380 sched_clock_idle_sleep_event
         swapper     0 [000]  5830.389116919:   call                        sched_clock_cpu                              ffffffff810a738b sched_clock_idle_sleep_event+0xb => ffffffff810a72e0 sched_clock_cpu
         swapper     0 [000]  5830.389116919:   call                            sched_clock                              ffffffff810a734d sched_clock_cpu+0x6d => ffffffff81035750 sched_clock
         swapper     0 [000]  5830.389116919:   call                                native_sched_clock                   ffffffff81035754 sched_clock+0x4 => ffffffff81035640 native_sched_clock
         swapper     0 [000]  5830.389116919:   return                              native_sched_clock                   ffffffff8103568c native_sched_clock+0x4c => ffffffff81035759 sched_clock
         swapper     0 [000]  5830.389116919:   return                          sched_clock                              ffffffff8103575c sched_clock+0xc => ffffffff810a7352 sched_clock_cpu
         swapper     0 [000]  5830.389116919:   return                      sched_clock_cpu                              ffffffff810a7356 sched_clock_cpu+0x76 => ffffffff810a7390 sched_clock_idle_sleep_event
         swapper     0 [000]  5830.389116919:   return                  sched_clock_idle_sleep_event                     ffffffff810a7391 sched_clock_idle_sleep_event+0x11 => ffffffff810f2910 __tick_nohz_idle_enter
	...

Signed-off-by Adrian Hunter <[email protected]>

---
 tools/perf/Documentation/perf-script.txt |  4 ++
 tools/perf/builtin-script.c              | 68 +++++++++++++++++++++++++++++++-
 tools/perf/util/thread-stack.c           |  7 ++++
 tools/perf/util/thread-stack.h           |  1 +
 4 files changed, 79 insertions(+), 1 deletion(-)

tools/perf/Documentation/perf-script.txt +4 -0

index a46030d8962d..1f6c70594f0f 100644
@@ -177,6 +177,10 @@ OPTIONS
 	"tr end" for "bE". However the "x" flag will be display separately in those
 	cases e.g. "jcc     (x)" for a condition branch within a transaction.
 
+	The callindent field is synthesized and may have a value when
+	Instruction Trace decoding. For calls and returns, it will display the
+	name of the symbol indented with spaces to reflect the stack depth.
+
 	Finally, a user may not set fields to none for all event types.
 	i.e., -F "" is not allowed.
 

tools/perf/builtin-script.c +67 -1

index 5a7633274d0a..c5648fd86229 100644
@@ -21,6 +21,7 @@
 #include "util/cpumap.h"
 #include "util/thread_map.h"
 #include "util/stat.h"
+#include "util/thread-stack.h"
 #include <linux/bitmap.h>
 #include <linux/stringify.h>
 #include "asm/bug.h"

@@ -63,6 +64,7 @@ enum perf_output_field {
 	PERF_OUTPUT_DATA_SRC	    = 1U << 17,
 	PERF_OUTPUT_WEIGHT	    = 1U << 18,
 	PERF_OUTPUT_BPF_OUTPUT	    = 1U << 19,
+	PERF_OUTPUT_CALLINDENT	    = 1U << 20,
 };
 
 struct output_option {

@@ -89,6 +91,7 @@ struct output_option {
 	{.str = "data_src", .field = PERF_OUTPUT_DATA_SRC},
 	{.str = "weight",   .field = PERF_OUTPUT_WEIGHT},
 	{.str = "bpf-output",   .field = PERF_OUTPUT_BPF_OUTPUT},
+	{.str = "callindent", .field = PERF_OUTPUT_CALLINDENT},
 };
 
 /* default set to maintain compatibility with current format */

@@ -562,6 +565,62 @@ static void print_sample_addr(struct perf_sample *sample,
 	}
 }
 
+static void print_sample_callindent(struct perf_sample *sample,
+				    struct perf_evsel *evsel,
+				    struct thread *thread,
+				    struct addr_location *al)
+{
+	struct perf_event_attr *attr = &evsel->attr;
+	size_t depth = thread_stack__depth(thread);
+	struct addr_location addr_al;
+	const char *name = NULL;
+	static int spacing;
+	int len = 0;
+	u64 ip = 0;
+
+	/*
+	 * The 'return' has already been popped off the stack so the depth has
+	 * to be adjusted to match the 'call'.
+	 */
+	if (thread->ts && sample->flags & PERF_IP_FLAG_RETURN)
+		depth += 1;
+
+	if (sample->flags & (PERF_IP_FLAG_CALL | PERF_IP_FLAG_TRACE_BEGIN)) {
+		if (sample_addr_correlates_sym(attr)) {
+			thread__resolve(thread, &addr_al, sample);
+			if (addr_al.sym)
+				name = addr_al.sym->name;
+			else
+				ip = sample->addr;
+		} else {
+			ip = sample->addr;
+		}
+	} else if (sample->flags & (PERF_IP_FLAG_RETURN | PERF_IP_FLAG_TRACE_END)) {
+		if (al->sym)
+			name = al->sym->name;
+		else
+			ip = sample->ip;
+	}
+
+	if (name)
+		len = printf("%*s%s", (int)depth * 4, "", name);
+	else if (ip)
+		len = printf("%*s%16" PRIx64, (int)depth * 4, "", ip);
+
+	if (len < 0)
+		return;
+
+	/*
+	 * Try to keep the output length from changing frequently so that the
+	 * output lines up more nicely.
+	 */
+	if (len > spacing || (len && len < spacing - 52))
+		spacing = round_up(len + 4, 32);
+
+	if (len < spacing)
+		printf("%*s", spacing - len, "");
+}
+
 static void print_sample_bts(struct perf_sample *sample,
 			     struct perf_evsel *evsel,
 			     struct thread *thread,

@@ -570,6 +629,9 @@ static void print_sample_bts(struct perf_sample *sample,
 	struct perf_event_attr *attr = &evsel->attr;
 	bool print_srcline_last = false;
 
+	if (PRINT_FIELD(CALLINDENT))
+		print_sample_callindent(sample, evsel, thread, al);
+
 	/* print branch_from information */
 	if (PRINT_FIELD(IP)) {
 		unsigned int print_opts = output[attr->type].print_ip_opts;

@@ -2053,7 +2115,8 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)
 		     "comma separated output fields prepend with 'type:'. "
 		     "Valid types: hw,sw,trace,raw. "
 		     "Fields: comm,tid,pid,time,cpu,event,trace,ip,sym,dso,"
-		     "addr,symoff,period,iregs,brstack,brstacksym,flags", parse_output_fields),
+		     "addr,symoff,period,iregs,brstack,brstacksym,flags,"
+		     "callindent", parse_output_fields),
 	OPT_BOOLEAN('a', "all-cpus", &system_wide,
 		    "system-wide collection from all CPUs"),
 	OPT_STRING('S', "symbols", &symbol_conf.sym_list_str, "symbol[,symbol...]",

@@ -2292,6 +2355,9 @@ int cmd_script(int argc, const char **argv, const char *prefix __maybe_unused)
 	script.session = session;
 	script__setup_sample_type(&script);
 
+	if (output[PERF_TYPE_HARDWARE].fields & PERF_OUTPUT_CALLINDENT)
+		itrace_synth_opts.thread_stack = true;
+
 	session->itrace_synth_opts = &itrace_synth_opts;
 
 	if (cpu_list) {

tools/perf/util/thread-stack.c +7 -0

index 825086aa9a08..d3301529f6a7 100644
@@ -616,3 +616,10 @@ int thread_stack__process(struct thread *thread, struct comm *comm,
 
 	return err;
 }
+
+size_t thread_stack__depth(struct thread *thread)
+{
+	if (!thread->ts)
+		return 0;
+	return thread->ts->cnt;
+}

tools/perf/util/thread-stack.h +1 -0

index ad44c7944b8e..b7e41c4ebfdd 100644
@@ -87,6 +87,7 @@ void thread_stack__sample(struct thread *thread, struct ip_callchain *chain,
 			  size_t sz, u64 ip);
 int thread_stack__flush(struct thread *thread);
 void thread_stack__free(struct thread *thread);
+size_t thread_stack__depth(struct thread *thread);
 
 struct call_return_processor *
 call_return_processor__new(int (*process)(struct call_return *cr, void *data),

Signature · 1 line

1.9.1

Here I think we should try to make the output more compact, so perhaps we should have some way to ask 'perf script' to show just timestamp deltas from the previous entry, for instance, i.e. replace that 'time' with a new 'delta' -F entry.

Quoted 4 lines

Here I think we should try to make the output more compact, so perhaps we should have some way to ask 'perf script' to show just timestamp deltas from the previous entry, for instance, i.e. replace that 'time' with a new 'delta' -F entry.

Agreed, also have a way to hide the numerical addresses, and some summary options to make it easier to use.

However this can be all done as followups later, no need to hold up this basic feature series for it.

Hi

Andi Kleen sent a couple of patches to add a callindent option to perf script. Andi was agreeable to this alternative implementation.

While there are some differences in the resulting output, the main difference is:

1. Tell the decoder to feed branches to the thread stack, which has the advantage that it happens before branch filtering and so can be used with different itrace options (e.g. it still works when only showing calls, even though the thread stack needs to see calls and returns). Also it does not conflict with using the thread stack to get callchains.