~hc/RK356X_SDK_RELEASE.git

..	..	@@ -8,15 +8,20 @@
8	8	*/
9	9	#include "builtin.h"
10	10
11		-#include "util/util.h"
12	11	#include "util/config.h"
13	12
14	13	#include "util/annotate.h"
15	14	#include "util/color.h"
	15	+#include "util/dso.h"
16	16	#include <linux/list.h>
17	17	#include <linux/rbtree.h>
18	18	#include <linux/err.h>
	19	+#include <linux/zalloc.h>
	20	+#include "util/map.h"
19	21	#include "util/symbol.h"
	22	+#include "util/map_symbol.h"
	23	+#include "util/mem-events.h"
	24	+#include "util/branch.h"
20	25	#include "util/callchain.h"
21	26	#include "util/values.h"
22	27
..	..	@@ -24,8 +29,10 @@
24	29	#include "util/debug.h"
25	30	#include "util/evlist.h"
26	31	#include "util/evsel.h"
	32	+#include "util/evswitch.h"
27	33	#include "util/header.h"
28	34	#include "util/session.h"
	35	+#include "util/srcline.h"
29	36	#include "util/tool.h"
30	37
31	38	#include <subcmd/parse-options.h>
..	..	@@ -40,15 +47,21 @@
40	47	#include "util/time-utils.h"
41	48	#include "util/auxtrace.h"
42	49	#include "util/units.h"
43		-#include "util/branch.h"
	50	+#include "util/util.h" // perf_tip()
	51	+#include "ui/ui.h"
	52	+#include "ui/progress.h"
	53	+#include "util/block-info.h"
44	54
45	55	#include <dlfcn.h>
46	56	#include <errno.h>
47	57	#include <inttypes.h>
48	58	#include <regex.h>
	59	+#include <linux/ctype.h>
49	60	#include <signal.h>
50	61	#include <linux/bitmap.h>
	62	+#include <linux/string.h>
51	63	#include <linux/stringify.h>
	64	+#include <linux/time64.h>
52	65	#include <sys/types.h>
53	66	#include <sys/stat.h>
54	67	#include <unistd.h>
..	..	@@ -57,6 +70,7 @@
57	70	struct report {
58	71	struct perf_tool tool;
59	72	struct perf_session *session;
	73	+ struct evswitch evswitch;
60	74	bool use_tui, use_gtk, use_stdio;
61	75	bool show_full_info;
62	76	bool show_threads;
..	..	@@ -69,6 +83,7 @@
69	83	bool header_only;
70	84	bool nonany_branch_mode;
71	85	bool group_set;
	86	+ bool stitch_lbr;
72	87	int max_stack;
73	88	struct perf_read_values show_threads_values;
74	89	struct annotation_options annotation_opts;
..	..	@@ -82,9 +97,14 @@
82	97	float min_percent;
83	98	u64 nr_entries;
84	99	u64 queue_size;
	100	+ u64 total_cycles;
85	101	int socket_filter;
86	102	DECLARE_BITMAP(cpu_bitmap, MAX_NR_CPUS);
87	103	struct branch_type_stat brtype_stat;
	104	+ bool symbol_ipc;
	105	+ bool total_cycles_mode;
	106	+ struct block_report *block_reports;
	107	+ int nr_block_reports;
88	108	};
89	109
90	110	static int report__config(const char var, const char value, void *cb)
..	..	@@ -124,16 +144,13 @@
124	144	int err = 0;
125	145	struct report *rep = arg;
126	146	struct hist_entry *he = iter->he;
127		- struct perf_evsel *evsel = iter->evsel;
	147	+ struct evsel *evsel = iter->evsel;
128	148	struct perf_sample *sample = iter->sample;
129	149	struct mem_info *mi;
130	150	struct branch_info *bi;
131	151
132		- if (!ui__has_annotation())
	152	+ if (!ui__has_annotation() && !rep->symbol_ipc)
133	153	return 0;
134		-
135		- hist__account_cycles(sample->branch_stack, al, sample,
136		- rep->nonany_branch_mode);
137	154
138	155	if (sort__mode == SORT_MODE__BRANCH) {
139	156	bi = he->branch_info;
..	..	@@ -169,49 +186,44 @@
169	186	{
170	187	struct hist_entry *he = iter->he;
171	188	struct report *rep = arg;
172		- struct branch_info *bi;
	189	+ struct branch_info *bi = he->branch_info;
173	190	struct perf_sample *sample = iter->sample;
174		- struct perf_evsel *evsel = iter->evsel;
	191	+ struct evsel *evsel = iter->evsel;
175	192	int err;
176	193
177		- if (!ui__has_annotation())
	194	+ branch_type_count(&rep->brtype_stat, &bi->flags,
	195	+ bi->from.addr, bi->to.addr);
	196	+
	197	+ if (!ui__has_annotation() && !rep->symbol_ipc)
178	198	return 0;
179	199
180		- hist__account_cycles(sample->branch_stack, al, sample,
181		- rep->nonany_branch_mode);
182		-
183		- bi = he->branch_info;
184	200	err = addr_map_symbol__inc_samples(&bi->from, sample, evsel);
185	201	if (err)
186	202	goto out;
187	203
188	204	err = addr_map_symbol__inc_samples(&bi->to, sample, evsel);
189	205
190		- branch_type_count(&rep->brtype_stat, &bi->flags,
191		- bi->from.addr, bi->to.addr);
192		-
193	206	out:
194	207	return err;
195	208	}
196	209
197	210	static void setup_forced_leader(struct report *report,
198		- struct perf_evlist *evlist)
	211	+ struct evlist *evlist)
199	212	{
200	213	if (report->group_set)
201	214	perf_evlist__force_leader(evlist);
202	215	}
203	216
204		-static int process_feature_event(struct perf_tool *tool,
205		- union perf_event *event,
206		- struct perf_session *session __maybe_unused)
	217	+static int process_feature_event(struct perf_session *session,
	218	+ union perf_event *event)
207	219	{
208		- struct report *rep = container_of(tool, struct report, tool);
	220	+ struct report *rep = container_of(session->tool, struct report, tool);
209	221
210	222	if (event->feat.feat_id < HEADER_LAST_FEATURE)
211		- return perf_event__process_feature(tool, event, session);
	223	+ return perf_event__process_feature(session, event);
212	224
213	225	if (event->feat.feat_id != HEADER_LAST_FEATURE) {
214		- pr_err("failed: wrong feature ID: %" PRIu64 "\n",
	226	+ pr_err("failed: wrong feature ID: %" PRI_lu64 "\n",
215	227	event->feat.feat_id);
216	228	return -1;
217	229	}
..	..	@@ -228,7 +240,7 @@
228	240	static int process_sample_event(struct perf_tool *tool,
229	241	union perf_event *event,
230	242	struct perf_sample *sample,
231		- struct perf_evsel *evsel,
	243	+ struct evsel *evsel,
232	244	struct machine *machine)
233	245	{
234	246	struct report *rep = container_of(tool, struct report, tool);
..	..	@@ -246,11 +258,17 @@
246	258	return 0;
247	259	}
248	260
	261	+ if (evswitch__discard(&rep->evswitch, evsel))
	262	+ return 0;
	263	+
249	264	if (machine__resolve(machine, &al, sample) < 0) {
250	265	pr_debug("problem processing %d event, skipping it.\n",
251	266	event->header.type);
252	267	return -1;
253	268	}
	269	+
	270	+ if (rep->stitch_lbr)
	271	+ al.thread->lbr_stitch_enable = true;
254	272
255	273	if (symbol_conf.hide_unresolved && al.sym == NULL)
256	274	goto out_put;
..	..	@@ -279,6 +297,12 @@
279	297	if (al.map != NULL)
280	298	al.map->dso->hit = 1;
281	299
	300	+ if (ui__has_annotation() \|\| rep->symbol_ipc \|\| rep->total_cycles_mode) {
	301	+ hist__account_cycles(sample->branch_stack, &al, sample,
	302	+ rep->nonany_branch_mode,
	303	+ &rep->total_cycles);
	304	+ }
	305	+
282	306	ret = hist_entry_iter__add(&iter, &al, rep->max_stack, rep);
283	307	if (ret < 0)
284	308	pr_debug("problem adding hist entry, skipping event\n");
..	..	@@ -290,13 +314,13 @@
290	314	static int process_read_event(struct perf_tool *tool,
291	315	union perf_event *event,
292	316	struct perf_sample *sample __maybe_unused,
293		- struct perf_evsel *evsel,
	317	+ struct evsel *evsel,
294	318	struct machine *machine __maybe_unused)
295	319	{
296	320	struct report *rep = container_of(tool, struct report, tool);
297	321
298	322	if (rep->show_threads) {
299		- const char *name = evsel ? perf_evsel__name(evsel) : "unknown";
	323	+ const char *name = evsel__name(evsel);
300	324	int err = perf_read_values_add_value(&rep->show_threads_values,
301	325	event->read.pid, event->read.tid,
302	326	evsel->idx,
..	..	@@ -314,16 +338,19 @@
314	338	static int report__setup_sample_type(struct report *rep)
315	339	{
316	340	struct perf_session *session = rep->session;
317		- u64 sample_type = perf_evlist__combined_sample_type(session->evlist);
	341	+ u64 sample_type = evlist__combined_sample_type(session->evlist);
318	342	bool is_pipe = perf_data__is_pipe(session->data);
	343	+ struct evsel *evsel;
319	344
320	345	if (session->itrace_synth_opts->callchain \|\|
	346	+ session->itrace_synth_opts->add_callchain \|\|
321	347	(!is_pipe &&
322	348	perf_header__has_feat(&session->header, HEADER_AUXTRACE) &&
323	349	!session->itrace_synth_opts->set))
324	350	sample_type \|= PERF_SAMPLE_CALLCHAIN;
325	351
326		- if (session->itrace_synth_opts->last_branch)
	352	+ if (session->itrace_synth_opts->last_branch \|\|
	353	+ session->itrace_synth_opts->add_last_branch)
327	354	sample_type \|= PERF_SAMPLE_BRANCH_STACK;
328	355
329	356	if (!is_pipe && !(sample_type & PERF_SAMPLE_CALLCHAIN)) {
..	..	@@ -367,20 +394,37 @@
367	394	}
368	395	}
369	396
370		- if (symbol_conf.use_callchain \|\| symbol_conf.cumulate_callchain) {
371		- if ((sample_type & PERF_SAMPLE_REGS_USER) &&
372		- (sample_type & PERF_SAMPLE_STACK_USER)) {
373		- callchain_param.record_mode = CALLCHAIN_DWARF;
374		- dwarf_callchain_users = true;
375		- } else if (sample_type & PERF_SAMPLE_BRANCH_STACK)
376		- callchain_param.record_mode = CALLCHAIN_LBR;
377		- else
378		- callchain_param.record_mode = CALLCHAIN_FP;
	397	+ if (sort__mode == SORT_MODE__MEMORY) {
	398	+ /*
	399	+ * FIXUP: prior to kernel 5.18, Arm SPE missed to set
	400	+ * PERF_SAMPLE_DATA_SRC bit in sample type. For backward
	401	+ * compatibility, set the bit if it's an old perf data file.
	402	+ */
	403	+ evlist__for_each_entry(session->evlist, evsel) {
	404	+ if (strstr(evsel->name, "arm_spe") &&
	405	+ !(sample_type & PERF_SAMPLE_DATA_SRC)) {
	406	+ evsel->core.attr.sample_type \|= PERF_SAMPLE_DATA_SRC;
	407	+ sample_type \|= PERF_SAMPLE_DATA_SRC;
	408	+ }
	409	+ }
	410	+
	411	+ if (!is_pipe && !(sample_type & PERF_SAMPLE_DATA_SRC)) {
	412	+ ui__error("Selected --mem-mode but no mem data. "
	413	+ "Did you call perf record without -d?\n");
	414	+ return -1;
	415	+ }
	416	+ }
	417	+
	418	+ callchain_param_setup(sample_type);
	419	+
	420	+ if (rep->stitch_lbr && (callchain_param.record_mode != CALLCHAIN_LBR)) {
	421	+ ui__warning("Can't find LBR callchain. Switch off --stitch-lbr.\n"
	422	+ "Please apply --call-graph lbr when recording.\n");
	423	+ rep->stitch_lbr = false;
379	424	}
380	425
381	426	/* ??? handle more cases than just ANY? */
382		- if (!(perf_evlist__combined_branch_type(session->evlist) &
383		- PERF_SAMPLE_BRANCH_ANY))
	427	+ if (!(evlist__combined_branch_type(session->evlist) & PERF_SAMPLE_BRANCH_ANY))
384	428	rep->nonany_branch_mode = true;
385	429
386	430	#if !defined(HAVE_LIBUNWIND_SUPPORT) && !defined(HAVE_DWARF_SUPPORT)
..	..	@@ -405,7 +449,7 @@
405	449	char unit;
406	450	unsigned long nr_samples = hists->stats.nr_events[PERF_RECORD_SAMPLE];
407	451	u64 nr_events = hists->stats.total_period;
408		- struct perf_evsel *evsel = hists_to_evsel(hists);
	452	+ struct evsel *evsel = hists_to_evsel(hists);
409	453	char buf[512];
410	454	size_t size = sizeof(buf);
411	455	int socked_id = hists->socket_filter;
..	..	@@ -418,10 +462,10 @@
418	462	nr_events = hists->stats.total_non_filtered_period;
419	463	}
420	464
421		- if (perf_evsel__is_group_event(evsel)) {
422		- struct perf_evsel *pos;
	465	+ if (evsel__is_group_event(evsel)) {
	466	+ struct evsel *pos;
423	467
424		- perf_evsel__group_desc(evsel, buf, size);
	468	+ evsel__group_desc(evsel, buf, size);
425	469	evname = buf;
426	470
427	471	for_each_group_member(pos, evsel) {
..	..	@@ -441,7 +485,7 @@
441	485	ret = fprintf(fp, "# Samples: %lu%c", nr_samples, unit);
442	486	if (evname != NULL) {
443	487	ret += fprintf(fp, " of event%s '%s'",
444		- evsel->nr_members > 1 ? "s" : "", evname);
	488	+ evsel->core.nr_members > 1 ? "s" : "", evname);
445	489	}
446	490
447	491	if (rep->time_str)
..	..	@@ -463,11 +507,30 @@
463	507	return ret + fprintf(fp, "\n#\n");
464	508	}
465	509
466		-static int perf_evlist__tty_browse_hists(struct perf_evlist *evlist,
	510	+static int perf_evlist__tui_block_hists_browse(struct evlist *evlist,
	511	+ struct report *rep)
	512	+{
	513	+ struct evsel *pos;
	514	+ int i = 0, ret;
	515	+
	516	+ evlist__for_each_entry(evlist, pos) {
	517	+ ret = report__browse_block_hists(&rep->block_reports[i++].hist,
	518	+ rep->min_percent, pos,
	519	+ &rep->session->header.env,
	520	+ &rep->annotation_opts);
	521	+ if (ret != 0)
	522	+ return ret;
	523	+ }
	524	+
	525	+ return 0;
	526	+}
	527	+
	528	+static int perf_evlist__tty_browse_hists(struct evlist *evlist,
467	529	struct report *rep,
468	530	const char *help)
469	531	{
470		- struct perf_evsel *pos;
	532	+ struct evsel *pos;
	533	+ int i = 0;
471	534
472	535	if (!quiet) {
473	536	fprintf(stdout, "#\n# Total Lost Samples: %" PRIu64 "\n#\n",
..	..	@@ -476,13 +539,20 @@
476	539
477	540	evlist__for_each_entry(evlist, pos) {
478	541	struct hists *hists = evsel__hists(pos);
479		- const char *evname = perf_evsel__name(pos);
	542	+ const char *evname = evsel__name(pos);
480	543
481		- if (symbol_conf.event_group &&
482		- !perf_evsel__is_group_leader(pos))
	544	+ if (symbol_conf.event_group && !evsel__is_group_leader(pos))
483	545	continue;
484	546
485	547	hists__fprintf_nr_sample_events(hists, rep, evname, stdout);
	548	+
	549	+ if (rep->total_cycles_mode) {
	550	+ report__browse_block_hists(&rep->block_reports[i++].hist,
	551	+ rep->min_percent, pos,
	552	+ NULL, NULL);
	553	+ continue;
	554	+ }
	555	+
486	556	hists__fprintf(hists, !quiet, 0, 0, rep->min_percent, stdout,
487	557	!(symbol_conf.use_callchain \|\|
488	558	symbol_conf.show_branchflag_count));
..	..	@@ -536,7 +606,7 @@
536	606
537	607	static int report__gtk_browse_hists(struct report rep, const char help)
538	608	{
539		- int (hist_browser)(struct perf_evlist evlist, const char *help,
	609	+ int (hist_browser)(struct evlist evlist, const char *help,
540	610	struct hist_browser_timer *timer, float min_pcnt);
541	611
542	612	hist_browser = dlsym(perf_gtk_handle, "perf_evlist__gtk_browse_hists");
..	..	@@ -553,18 +623,26 @@
553	623	{
554	624	int ret;
555	625	struct perf_session *session = rep->session;
556		- struct perf_evlist *evlist = session->evlist;
557		- const char *help = perf_tip(system_path(TIPDIR));
	626	+ struct evlist *evlist = session->evlist;
	627	+ char help = NULL, path = NULL;
558	628
559		- if (help == NULL) {
	629	+ path = system_path(TIPDIR);
	630	+ if (perf_tip(&help, path) \|\| help == NULL) {
560	631	/* fallback for people who don't install perf ;-) */
561		- help = perf_tip(DOCDIR);
562		- if (help == NULL)
563		- help = "Cannot load tips.txt file, please install perf!";
	632	+ free(path);
	633	+ path = system_path(DOCDIR);
	634	+ if (perf_tip(&help, path) \|\| help == NULL)
	635	+ help = strdup("Cannot load tips.txt file, please install perf!");
564	636	}
	637	+ free(path);
565	638
566	639	switch (use_browser) {
567	640	case 1:
	641	+ if (rep->total_cycles_mode) {
	642	+ ret = perf_evlist__tui_block_hists_browse(evlist, rep);
	643	+ break;
	644	+ }
	645	+
568	646	ret = perf_evlist__tui_browse_hists(evlist, help, NULL,
569	647	rep->min_percent,
570	648	&session->header.env,
..	..	@@ -573,7 +651,7 @@
573	651	* Usually "ret" is the last pressed key, and we only
574	652	* care if the key notifies us to switch data file.
575	653	*/
576		- if (ret != K_SWITCH_INPUT_DATA)
	654	+ if (ret != K_SWITCH_INPUT_DATA && ret != K_RELOAD)
577	655	ret = 0;
578	656	break;
579	657	case 2:
..	..	@@ -583,14 +661,14 @@
583	661	ret = perf_evlist__tty_browse_hists(evlist, rep, help);
584	662	break;
585	663	}
586		-
	664	+ free(help);
587	665	return ret;
588	666	}
589	667
590	668	static int report__collapse_hists(struct report *rep)
591	669	{
592	670	struct ui_progress prog;
593		- struct perf_evsel *pos;
	671	+ struct evsel *pos;
594	672	int ret = 0;
595	673
596	674	ui_progress__init(&prog, rep->nr_entries, "Merging related events...");
..	..	@@ -608,8 +686,7 @@
608	686	break;
609	687
610	688	/* Non-group events are considered as leader */
611		- if (symbol_conf.event_group &&
612		- !perf_evsel__is_group_leader(pos)) {
	689	+ if (symbol_conf.event_group && !evsel__is_group_leader(pos)) {
613	690	struct hists *leader_hists = evsel__hists(pos->leader);
614	691
615	692	hists__match(leader_hists, hists);
..	..	@@ -621,15 +698,31 @@
621	698	return ret;
622	699	}
623	700
	701	+static int hists__resort_cb(struct hist_entry he, void arg)
	702	+{
	703	+ struct report *rep = arg;
	704	+ struct symbol *sym = he->ms.sym;
	705	+
	706	+ if (rep->symbol_ipc && sym && !sym->annotate2) {
	707	+ struct evsel *evsel = hists_to_evsel(he->hists);
	708	+
	709	+ symbol__annotate2(&he->ms, evsel,
	710	+ &annotation__default_options, NULL);
	711	+ }
	712	+
	713	+ return 0;
	714	+}
	715	+
624	716	static void report__output_resort(struct report *rep)
625	717	{
626	718	struct ui_progress prog;
627		- struct perf_evsel *pos;
	719	+ struct evsel *pos;
628	720
629	721	ui_progress__init(&prog, rep->nr_entries, "Sorting events for output...");
630	722
631		- evlist__for_each_entry(rep->session->evlist, pos)
632		- perf_evsel__output_resort(pos, &prog);
	723	+ evlist__for_each_entry(rep->session->evlist, pos) {
	724	+ evsel__output_resort_cb(pos, &prog, hists__resort_cb, rep);
	725	+ }
633	726
634	727	ui_progress__finish();
635	728	}
..	..	@@ -693,11 +786,9 @@
693	786	static size_t maps__fprintf_task(struct maps maps, int indent, FILE fp)
694	787	{
695	788	size_t printed = 0;
696		- struct rb_node *nd;
	789	+ struct map *map;
697	790
698		- for (nd = rb_first(&maps->entries); nd; nd = rb_next(nd)) {
699		- struct map *map = rb_entry(nd, struct map, rb_node);
700		-
	791	+ maps__for_each_entry(maps, map) {
701	792	printed += fprintf(fp, "%*s %" PRIx64 "-%" PRIx64 " %c%c%c%c %08" PRIx64 " %" PRIu64 " %s\n",
702	793	indent, "", map->start, map->end,
703	794	map->prot & PROT_READ ? 'r' : '-',
..	..	@@ -705,15 +796,10 @@
705	796	map->prot & PROT_EXEC ? 'x' : '-',
706	797	map->flags & MAP_SHARED ? 's' : 'p',
707	798	map->pgoff,
708		- map->ino, map->dso->name);
	799	+ map->dso->id.ino, map->dso->name);
709	800	}
710	801
711	802	return printed;
712		-}
713		-
714		-static int map_groups__fprintf_task(struct map_groups mg, int indent, FILE fp)
715		-{
716		- return maps__fprintf_task(&mg->maps, indent, fp);
717	803	}
718	804
719	805	static void task__print_level(struct task task, FILE fp, int level)
..	..	@@ -726,7 +812,7 @@
726	812
727	813	fprintf(fp, "%s\n", thread__comm_str(thread));
728	814
729		- map_groups__fprintf_task(thread->mg, comm_indent, fp);
	815	+ maps__fprintf_task(thread->maps, comm_indent, fp);
730	816
731	817	if (!list_empty(&task->children)) {
732	818	list_for_each_entry(child, &task->children, list)
..	..	@@ -759,7 +845,8 @@
759	845	for (i = 0; i < THREADS__TABLE_SIZE; i++) {
760	846	struct threads *threads = &machine->threads[i];
761	847
762		- for (nd = rb_first(&threads->entries); nd; nd = rb_next(nd)) {
	848	+ for (nd = rb_first_cached(&threads->entries); nd;
	849	+ nd = rb_next(nd)) {
763	850	task = tasks + itask++;
764	851
765	852	task->thread = rb_entry(nd, struct thread, rb_node);
..	..	@@ -804,7 +891,7 @@
804	891	{
805	892	int ret;
806	893	struct perf_session *session = rep->session;
807		- struct perf_evsel *pos;
	894	+ struct evsel *pos;
808	895	struct perf_data *data = session->data;
809	896
810	897	signal(SIGINT, sig_handler);
..	..	@@ -886,11 +973,29 @@
886	973	rep->nr_entries += evsel__hists(pos)->nr_entries;
887	974
888	975	if (rep->nr_entries == 0) {
889		- ui__error("The %s file has no samples!\n", data->file.path);
	976	+ ui__error("The %s data has no samples!\n", data->path);
890	977	return 0;
891	978	}
892	979
893	980	report__output_resort(rep);
	981	+
	982	+ if (rep->total_cycles_mode) {
	983	+ int block_hpps[6] = {
	984	+ PERF_HPP_REPORT__BLOCK_TOTAL_CYCLES_PCT,
	985	+ PERF_HPP_REPORT__BLOCK_LBR_CYCLES,
	986	+ PERF_HPP_REPORT__BLOCK_CYCLES_PCT,
	987	+ PERF_HPP_REPORT__BLOCK_AVG_CYCLES,
	988	+ PERF_HPP_REPORT__BLOCK_RANGE,
	989	+ PERF_HPP_REPORT__BLOCK_DSO,
	990	+ };
	991	+
	992	+ rep->block_reports = block_info__create_report(session->evlist,
	993	+ rep->total_cycles,
	994	+ block_hpps, 6,
	995	+ &rep->nr_block_reports);
	996	+ if (!rep->block_reports)
	997	+ return -1;
	998	+ }
894	999
895	1000	return report__browse_hists(rep);
896	1001	}
..	..	@@ -911,6 +1016,42 @@
911	1016	}
912	1017
913	1018	return parse_callchain_report_opt(arg);
	1019	+}
	1020	+
	1021	+static int
	1022	+parse_time_quantum(const struct option opt, const char arg,
	1023	+ int unset __maybe_unused)
	1024	+{
	1025	+ unsigned long *time_q = opt->value;
	1026	+ char *end;
	1027	+
	1028	+ *time_q = strtoul(arg, &end, 0);
	1029	+ if (end == arg)
	1030	+ goto parse_err;
	1031	+ if (*time_q == 0) {
	1032	+ pr_err("time quantum cannot be 0");
	1033	+ return -1;
	1034	+ }
	1035	+ end = skip_spaces(end);
	1036	+ if (*end == 0)
	1037	+ return 0;
	1038	+ if (!strcmp(end, "s")) {
	1039	+ time_q = NSEC_PER_SEC;
	1040	+ return 0;
	1041	+ }
	1042	+ if (!strcmp(end, "ms")) {
	1043	+ time_q = NSEC_PER_MSEC;
	1044	+ return 0;
	1045	+ }
	1046	+ if (!strcmp(end, "us")) {
	1047	+ time_q = NSEC_PER_USEC;
	1048	+ return 0;
	1049	+ }
	1050	+ if (!strcmp(end, "ns"))
	1051	+ return 0;
	1052	+parse_err:
	1053	+ pr_err("Cannot parse time quantum `%s'\n", arg);
	1054	+ return -1;
914	1055	}
915	1056
916	1057	int
..	..	@@ -953,6 +1094,26 @@
953	1094	return 0;
954	1095	}
955	1096
	1097	+static int process_attr(struct perf_tool *tool __maybe_unused,
	1098	+ union perf_event *event,
	1099	+ struct evlist **pevlist)
	1100	+{
	1101	+ u64 sample_type;
	1102	+ int err;
	1103	+
	1104	+ err = perf_event__process_attr(tool, event, pevlist);
	1105	+ if (err)
	1106	+ return err;
	1107	+
	1108	+ /*
	1109	+ * Check if we need to enable callchains based
	1110	+ * on events sample_type.
	1111	+ */
	1112	+ sample_type = evlist__combined_sample_type(*pevlist);
	1113	+ callchain_param_setup(sample_type);
	1114	+ return 0;
	1115	+}
	1116	+
956	1117	int cmd_report(int argc, const char **argv)
957	1118	{
958	1119	struct perf_session *session;
..	..	@@ -963,9 +1124,9 @@
963	1124	int last_key = 0;
964	1125	bool branch_call_mode = false;
965	1126	#define CALLCHAIN_DEFAULT_OPT "graph,0.5,caller,function,percent"
966		- const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
967		- CALLCHAIN_REPORT_HELP
968		- "\n\t\t\t\tDefault: " CALLCHAIN_DEFAULT_OPT;
	1127	+ static const char report_callchain_help[] = "Display call graph (stack chain/backtrace):\n\n"
	1128	+ CALLCHAIN_REPORT_HELP
	1129	+ "\n\t\t\t\tDefault: " CALLCHAIN_DEFAULT_OPT;
969	1130	char callchain_default_opt[] = CALLCHAIN_DEFAULT_OPT;
970	1131	const char * const report_usage[] = {
971	1132	"perf report [<options>]",
..	..	@@ -978,11 +1139,12 @@
978	1139	.mmap2 = perf_event__process_mmap2,
979	1140	.comm = perf_event__process_comm,
980	1141	.namespaces = perf_event__process_namespaces,
	1142	+ .cgroup = perf_event__process_cgroup,
981	1143	.exit = perf_event__process_exit,
982	1144	.fork = perf_event__process_fork,
983	1145	.lost = perf_event__process_lost,
984	1146	.read = process_read_event,
985		- .attr = perf_event__process_attr,
	1147	+ .attr = process_attr,
986	1148	.tracing_data = perf_event__process_tracing_data,
987	1149	.build_id = perf_event__process_build_id,
988	1150	.id_index = perf_event__process_id_index,
..	..	@@ -998,6 +1160,8 @@
998	1160	.socket_filter = -1,
999	1161	.annotation_opts = annotation__default_options,
1000	1162	};
	1163	+ char *sort_order_help = sort_help("sort by key(s):");
	1164	+ char *field_order_help = sort_help("output field(s): overhead period sample ");
1001	1165	const struct option options[] = {
1002	1166	OPT_STRING('i', "input", &input_name, "file",
1003	1167	"input file name"),
..	..	@@ -1032,10 +1196,9 @@
1032	1196	OPT_BOOLEAN(0, "header-only", &report.header_only,
1033	1197	"Show only data header."),
1034	1198	OPT_STRING('s', "sort", &sort_order, "key[,key2...]",
1035		- "sort by key(s): pid, comm, dso, symbol, parent, cpu, srcline, ..."
1036		- " Please refer the man page for the complete list."),
	1199	+ sort_order_help),
1037	1200	OPT_STRING('F', "fields", &field_order, "key[,keys...]",
1038		- "output field(s): overhead, period, sample plus all of sort keys"),
	1201	+ field_order_help),
1039	1202	OPT_BOOLEAN(0, "show-cpu-utilization", &symbol_conf.show_cpu_utilization,
1040	1203	"Show sample percentage for different cpu modes"),
1041	1204	OPT_BOOLEAN_FLAG(0, "showcpuutilization", &symbol_conf.show_cpu_utilization,
..	..	@@ -1049,7 +1212,8 @@
1049	1212	report_callchain_help, &report_parse_callchain_opt,
1050	1213	callchain_default_opt),
1051	1214	OPT_BOOLEAN(0, "children", &symbol_conf.cumulate_callchain,
1052		- "Accumulate callchains of children and show total overhead as well"),
	1215	+ "Accumulate callchains of children and show total overhead as well. "
	1216	+ "Enabled by default, use --no-children to disable."),
1053	1217	OPT_INTEGER(0, "max-stack", &report.max_stack,
1054	1218	"Set the maximum stack depth when parsing the callchain, "
1055	1219	"anything beyond the specified depth will be ignored. "
..	..	@@ -1092,10 +1256,18 @@
1092	1256	"Display raw encoding of assembly instructions (default)"),
1093	1257	OPT_STRING('M', "disassembler-style", &report.annotation_opts.disassembler_style, "disassembler style",
1094	1258	"Specify disassembler style (e.g. -M intel for intel syntax)"),
	1259	+ OPT_STRING(0, "prefix", &report.annotation_opts.prefix, "prefix",
	1260	+ "Add prefix to source file path names in programs (with --prefix-strip)"),
	1261	+ OPT_STRING(0, "prefix-strip", &report.annotation_opts.prefix_strip, "N",
	1262	+ "Strip first N entries of source file path name in programs (with --prefix)"),
1095	1263	OPT_BOOLEAN(0, "show-total-period", &symbol_conf.show_total_period,
1096	1264	"Show a column with the sum of periods"),
1097	1265	OPT_BOOLEAN_SET(0, "group", &symbol_conf.event_group, &report.group_set,
1098	1266	"Show event group information together"),
	1267	+ OPT_INTEGER(0, "group-sort-idx", &symbol_conf.group_sort_idx,
	1268	+ "Sort the output by the event at the index n in group. "
	1269	+ "If n is invalid, sort by the first event. "
	1270	+ "WARNING: should be used on grouped events."),
1099	1271	OPT_CALLBACK_NOOPT('b', "branch-stack", &branch_mode, "",
1100	1272	"use branch records for per branch histogram filling",
1101	1273	parse_branch_mode),
..	..	@@ -1108,17 +1280,21 @@
1108	1280	OPT_BOOLEAN(0, "demangle-kernel", &symbol_conf.demangle_kernel,
1109	1281	"Enable kernel symbol demangling"),
1110	1282	OPT_BOOLEAN(0, "mem-mode", &report.mem_mode, "mem access profile"),
	1283	+ OPT_INTEGER(0, "samples", &symbol_conf.res_sample,
	1284	+ "Number of samples to save per histogram entry for individual browsing"),
1111	1285	OPT_CALLBACK(0, "percent-limit", &report, "percent",
1112	1286	"Don't show entries under that percent", parse_percent_limit),
1113	1287	OPT_CALLBACK(0, "percentage", NULL, "relative\|absolute",
1114	1288	"how to display percentage of filtered entries", parse_filter_percentage),
1115	1289	OPT_CALLBACK_OPTARG(0, "itrace", &itrace_synth_opts, NULL, "opts",
1116		- "Instruction Tracing options",
	1290	+ "Instruction Tracing options\n" ITRACE_HELP,
1117	1291	itrace_parse_synth_opts),
1118	1292	OPT_BOOLEAN(0, "full-source-path", &srcline_full_filename,
1119	1293	"Show full source file name path for source lines"),
1120	1294	OPT_BOOLEAN(0, "show-ref-call-graph", &symbol_conf.show_ref_callgraph,
1121	1295	"Show callgraph from reference event"),
	1296	+ OPT_BOOLEAN(0, "stitch-lbr", &report.stitch_lbr,
	1297	+ "Enable LBR callgraph stitching approach"),
1122	1298	OPT_INTEGER(0, "socket-filter", &report.socket_filter,
1123	1299	"only show processor socket that match with this filter"),
1124	1300	OPT_BOOLEAN(0, "raw-trace", &symbol_conf.raw_trace,
..	..	@@ -1135,19 +1311,27 @@
1135	1311	OPT_CALLBACK(0, "percent-type", &report.annotation_opts, "local-period",
1136	1312	"Set percent type local/global-period/hits",
1137	1313	annotate_parse_percent_type),
	1314	+ OPT_BOOLEAN(0, "ns", &symbol_conf.nanosecs, "Show times in nanosecs"),
	1315	+ OPT_CALLBACK(0, "time-quantum", &symbol_conf.time_quantum, "time (ms\|us\|ns\|s)",
	1316	+ "Set time quantum for time sort key (default 100ms)",
	1317	+ parse_time_quantum),
	1318	+ OPTS_EVSWITCH(&report.evswitch),
	1319	+ OPT_BOOLEAN(0, "total-cycles", &report.total_cycles_mode,
	1320	+ "Sort all blocks by 'Sampled Cycles%'"),
1138	1321	OPT_END()
1139	1322	};
1140	1323	struct perf_data data = {
1141	1324	.mode = PERF_DATA_MODE_READ,
1142	1325	};
1143	1326	int ret = hists__init();
	1327	+ char sort_tmp[128];
1144	1328
1145	1329	if (ret < 0)
1146		- return ret;
	1330	+ goto exit;
1147	1331
1148	1332	ret = perf_config(report__config, &report);
1149	1333	if (ret)
1150		- return ret;
	1334	+ goto exit;
1151	1335
1152	1336	argc = parse_options(argc, argv, options, report_usage, 0);
1153	1337	if (argc) {
..	..	@@ -1161,8 +1345,16 @@
1161	1345	report.symbol_filter_str = argv[0];
1162	1346	}
1163	1347
	1348	+ if (annotate_check_args(&report.annotation_opts) < 0) {
	1349	+ ret = -EINVAL;
	1350	+ goto exit;
	1351	+ }
	1352	+
1164	1353	if (report.mmaps_mode)
1165	1354	report.tasks_mode = true;
	1355	+
	1356	+ if (dump_trace)
	1357	+ report.tool.ordered_events = false;
1166	1358
1167	1359	if (quiet)
1168	1360	perf_quiet_option();
..	..	@@ -1170,12 +1362,14 @@
1170	1362	if (symbol_conf.vmlinux_name &&
1171	1363	access(symbol_conf.vmlinux_name, R_OK)) {
1172	1364	pr_err("Invalid file: %s\n", symbol_conf.vmlinux_name);
1173		- return -EINVAL;
	1365	+ ret = -EINVAL;
	1366	+ goto exit;
1174	1367	}
1175	1368	if (symbol_conf.kallsyms_name &&
1176	1369	access(symbol_conf.kallsyms_name, R_OK)) {
1177	1370	pr_err("Invalid file: %s\n", symbol_conf.kallsyms_name);
1178		- return -EINVAL;
	1371	+ ret = -EINVAL;
	1372	+ goto exit;
1179	1373	}
1180	1374
1181	1375	if (report.inverted_callchain)
..	..	@@ -1183,7 +1377,7 @@
1183	1377	if (symbol_conf.cumulate_callchain && !callchain_param.order_set)
1184	1378	callchain_param.order = ORDER_CALLER;
1185	1379
1186		- if (itrace_synth_opts.callchain &&
	1380	+ if ((itrace_synth_opts.callchain \|\| itrace_synth_opts.add_callchain) &&
1187	1381	(int)itrace_synth_opts.callchain_sz > report.max_stack)
1188	1382	report.max_stack = itrace_synth_opts.callchain_sz;
1189	1383
..	..	@@ -1194,13 +1388,22 @@
1194	1388	input_name = "perf.data";
1195	1389	}
1196	1390
1197		- data.file.path = input_name;
1198		- data.force = symbol_conf.force;
	1391	+ data.path = input_name;
	1392	+ data.force = symbol_conf.force;
1199	1393
1200	1394	repeat:
1201	1395	session = perf_session__new(&data, false, &report.tool);
1202		- if (session == NULL)
1203		- return -1;
	1396	+ if (IS_ERR(session)) {
	1397	+ ret = PTR_ERR(session);
	1398	+ goto exit;
	1399	+ }
	1400	+
	1401	+ ret = evswitch__init(&report.evswitch, session->evlist, stderr);
	1402	+ if (ret)
	1403	+ goto exit;
	1404	+
	1405	+ if (zstd_init(&(session->zstd_data), 0) < 0)
	1406	+ pr_warning("Decompression initialization failed. Reported data may be incomplete.\n");
1204	1407
1205	1408	if (report.queue_size) {
1206	1409	ordered_events__set_alloc_size(&session->ordered_events,
..	..	@@ -1213,10 +1416,18 @@
1213	1416
1214	1417	has_br_stack = perf_header__has_feat(&session->header,
1215	1418	HEADER_BRANCH_STACK);
	1419	+ if (evlist__combined_sample_type(session->evlist) & PERF_SAMPLE_STACK_USER)
	1420	+ has_br_stack = false;
1216	1421
1217	1422	setup_forced_leader(&report, session->evlist);
1218	1423
1219		- if (itrace_synth_opts.last_branch)
	1424	+ if (symbol_conf.group_sort_idx && !session->evlist->nr_groups) {
	1425	+ parse_options_usage(NULL, options, "group-sort-idx", 0);
	1426	+ ret = -EINVAL;
	1427	+ goto error;
	1428	+ }
	1429	+
	1430	+ if (itrace_synth_opts.last_branch \|\| itrace_synth_opts.add_last_branch)
1220	1431	has_br_stack = true;
1221	1432
1222	1433	if (has_br_stack && branch_call_mode)
..	..	@@ -1236,7 +1447,7 @@
1236	1447	}
1237	1448	if (branch_call_mode) {
1238	1449	callchain_param.key = CCKEY_ADDRESS;
1239		- callchain_param.branch_callstack = 1;
	1450	+ callchain_param.branch_callstack = true;
1240	1451	symbol_conf.use_callchain = true;
1241	1452	callchain_register_param(&callchain_param);
1242	1453	if (sort_order == NULL)
..	..	@@ -1287,12 +1498,37 @@
1287	1498	goto error;
1288	1499	}
1289	1500
	1501	+ if (report.total_cycles_mode) {
	1502	+ if (sort__mode != SORT_MODE__BRANCH)
	1503	+ report.total_cycles_mode = false;
	1504	+ else
	1505	+ sort_order = NULL;
	1506	+ }
	1507	+
1290	1508	if (strcmp(input_name, "-") != 0)
1291	1509	setup_browser(true);
1292	1510	else
1293	1511	use_browser = 0;
1294	1512
1295		- if ((last_key != K_SWITCH_INPUT_DATA) &&
	1513	+ if (sort_order && strstr(sort_order, "ipc")) {
	1514	+ parse_options_usage(report_usage, options, "s", 1);
	1515	+ goto error;
	1516	+ }
	1517	+
	1518	+ if (sort_order && strstr(sort_order, "symbol")) {
	1519	+ if (sort__mode == SORT_MODE__BRANCH) {
	1520	+ snprintf(sort_tmp, sizeof(sort_tmp), "%s,%s",
	1521	+ sort_order, "ipc_lbr");
	1522	+ report.symbol_ipc = true;
	1523	+ } else {
	1524	+ snprintf(sort_tmp, sizeof(sort_tmp), "%s,%s",
	1525	+ sort_order, "ipc_null");
	1526	+ }
	1527	+
	1528	+ sort_order = sort_tmp;
	1529	+ }
	1530	+
	1531	+ if ((last_key != K_SWITCH_INPUT_DATA && last_key != K_RELOAD) &&
1296	1532	(setup_sorting(session->evlist) < 0)) {
1297	1533	if (sort_order)
1298	1534	parse_options_usage(report_usage, options, "s", 1);
..	..	@@ -1320,7 +1556,8 @@
1320	1556	* so don't allocate extra space that won't be used in the stdio
1321	1557	* implementation.
1322	1558	*/
1323		- if (ui__has_annotation()) {
	1559	+ if (ui__has_annotation() \|\| report.symbol_ipc \|\|
	1560	+ report.total_cycles_mode) {
1324	1561	ret = symbol__annotation_init();
1325	1562	if (ret < 0)
1326	1563	goto error;
..	..	@@ -1339,42 +1576,23 @@
1339	1576	symbol_conf.priv_size += sizeof(u32);
1340	1577	symbol_conf.sort_by_name = true;
1341	1578	}
1342		- annotation_config__init();
	1579	+ annotation_config__init(&report.annotation_opts);
1343	1580	}
1344	1581
1345	1582	if (symbol__init(&session->header.env) < 0)
1346	1583	goto error;
1347	1584
1348		- report.ptime_range = perf_time__range_alloc(report.time_str,
1349		- &report.range_size);
1350		- if (!report.ptime_range) {
1351		- ret = -ENOMEM;
1352		- goto error;
1353		- }
1354		-
1355		- if (perf_time__parse_str(report.ptime_range, report.time_str) != 0) {
1356		- if (session->evlist->first_sample_time == 0 &&
1357		- session->evlist->last_sample_time == 0) {
1358		- pr_err("HINT: no first/last sample time found in perf data.\n"
1359		- "Please use latest perf binary to execute 'perf record'\n"
1360		- "(if '--buildid-all' is enabled, please set '--timestamp-boundary').\n");
1361		- ret = -EINVAL;
	1585	+ if (report.time_str) {
	1586	+ ret = perf_time__parse_for_ranges(report.time_str, session,
	1587	+ &report.ptime_range,
	1588	+ &report.range_size,
	1589	+ &report.range_num);
	1590	+ if (ret < 0)
1362	1591	goto error;
1363		- }
1364	1592
1365		- report.range_num = perf_time__percent_parse_str(
1366		- report.ptime_range, report.range_size,
1367		- report.time_str,
1368		- session->evlist->first_sample_time,
1369		- session->evlist->last_sample_time);
1370		-
1371		- if (report.range_num < 0) {
1372		- pr_err("Invalid time string\n");
1373		- ret = -EINVAL;
1374		- goto error;
1375		- }
1376		- } else {
1377		- report.range_num = 1;
	1593	+ itrace_synth_opts__set_time_range(&itrace_synth_opts,
	1594	+ report.ptime_range,
	1595	+ report.range_num);
1378	1596	}
1379	1597
1380	1598	if (session->tevent.pevent &&
..	..	@@ -1389,7 +1607,7 @@
1389	1607	sort__setup_elide(stdout);
1390	1608
1391	1609	ret = __cmd_report(&report);
1392		- if (ret == K_SWITCH_INPUT_DATA) {
	1610	+ if (ret == K_SWITCH_INPUT_DATA \|\| ret == K_RELOAD) {
1393	1611	perf_session__delete(session);
1394	1612	last_key = K_SWITCH_INPUT_DATA;
1395	1613	goto repeat;
..	..	@@ -1397,8 +1615,21 @@
1397	1615	ret = 0;
1398	1616
1399	1617	error:
1400		- zfree(&report.ptime_range);
	1618	+ if (report.ptime_range) {
	1619	+ itrace_synth_opts__clear_time_range(&itrace_synth_opts);
	1620	+ zfree(&report.ptime_range);
	1621	+ }
1401	1622
	1623	+ if (report.block_reports) {
	1624	+ block_info__free_report(report.block_reports,
	1625	+ report.nr_block_reports);
	1626	+ report.block_reports = NULL;
	1627	+ }
	1628	+
	1629	+ zstd_fini(&(session->zstd_data));
1402	1630	perf_session__delete(session);
	1631	+exit:
	1632	+ free(sort_order_help);
	1633	+ free(field_order_help);
1403	1634	return ret;
1404	1635	}