Loading...
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * thread-stack.c: Synthesize a thread's stack using call / return events
4 * Copyright (c) 2014, Intel Corporation.
5 */
6
7#include <linux/rbtree.h>
8#include <linux/list.h>
9#include <linux/log2.h>
10#include <linux/zalloc.h>
11#include <errno.h>
12#include <stdlib.h>
13#include <string.h>
14#include "thread.h"
15#include "event.h"
16#include "machine.h"
17#include "env.h"
18#include "debug.h"
19#include "symbol.h"
20#include "comm.h"
21#include "call-path.h"
22#include "thread-stack.h"
23
24#define STACK_GROWTH 2048
25
26/*
27 * State of retpoline detection.
28 *
29 * RETPOLINE_NONE: no retpoline detection
30 * X86_RETPOLINE_POSSIBLE: x86 retpoline possible
31 * X86_RETPOLINE_DETECTED: x86 retpoline detected
32 */
33enum retpoline_state_t {
34 RETPOLINE_NONE,
35 X86_RETPOLINE_POSSIBLE,
36 X86_RETPOLINE_DETECTED,
37};
38
39/**
40 * struct thread_stack_entry - thread stack entry.
41 * @ret_addr: return address
42 * @timestamp: timestamp (if known)
43 * @ref: external reference (e.g. db_id of sample)
44 * @branch_count: the branch count when the entry was created
45 * @insn_count: the instruction count when the entry was created
46 * @cyc_count the cycle count when the entry was created
47 * @db_id: id used for db-export
48 * @cp: call path
49 * @no_call: a 'call' was not seen
50 * @trace_end: a 'call' but trace ended
51 * @non_call: a branch but not a 'call' to the start of a different symbol
52 */
53struct thread_stack_entry {
54 u64 ret_addr;
55 u64 timestamp;
56 u64 ref;
57 u64 branch_count;
58 u64 insn_count;
59 u64 cyc_count;
60 u64 db_id;
61 struct call_path *cp;
62 bool no_call;
63 bool trace_end;
64 bool non_call;
65};
66
67/**
68 * struct thread_stack - thread stack constructed from 'call' and 'return'
69 * branch samples.
70 * @stack: array that holds the stack
71 * @cnt: number of entries in the stack
72 * @sz: current maximum stack size
73 * @trace_nr: current trace number
74 * @branch_count: running branch count
75 * @insn_count: running instruction count
76 * @cyc_count running cycle count
77 * @kernel_start: kernel start address
78 * @last_time: last timestamp
79 * @crp: call/return processor
80 * @comm: current comm
81 * @arr_sz: size of array if this is the first element of an array
82 * @rstate: used to detect retpolines
83 * @br_stack_rb: branch stack (ring buffer)
84 * @br_stack_sz: maximum branch stack size
85 * @br_stack_pos: current position in @br_stack_rb
86 * @mispred_all: mark all branches as mispredicted
87 */
88struct thread_stack {
89 struct thread_stack_entry *stack;
90 size_t cnt;
91 size_t sz;
92 u64 trace_nr;
93 u64 branch_count;
94 u64 insn_count;
95 u64 cyc_count;
96 u64 kernel_start;
97 u64 last_time;
98 struct call_return_processor *crp;
99 struct comm *comm;
100 unsigned int arr_sz;
101 enum retpoline_state_t rstate;
102 struct branch_stack *br_stack_rb;
103 unsigned int br_stack_sz;
104 unsigned int br_stack_pos;
105 bool mispred_all;
106};
107
108/*
109 * Assume pid == tid == 0 identifies the idle task as defined by
110 * perf_session__register_idle_thread(). The idle task is really 1 task per cpu,
111 * and therefore requires a stack for each cpu.
112 */
113static inline bool thread_stack__per_cpu(struct thread *thread)
114{
115 return !(thread->tid || thread->pid_);
116}
117
118static int thread_stack__grow(struct thread_stack *ts)
119{
120 struct thread_stack_entry *new_stack;
121 size_t sz, new_sz;
122
123 new_sz = ts->sz + STACK_GROWTH;
124 sz = new_sz * sizeof(struct thread_stack_entry);
125
126 new_stack = realloc(ts->stack, sz);
127 if (!new_stack)
128 return -ENOMEM;
129
130 ts->stack = new_stack;
131 ts->sz = new_sz;
132
133 return 0;
134}
135
136static int thread_stack__init(struct thread_stack *ts, struct thread *thread,
137 struct call_return_processor *crp,
138 bool callstack, unsigned int br_stack_sz)
139{
140 int err;
141
142 if (callstack) {
143 err = thread_stack__grow(ts);
144 if (err)
145 return err;
146 }
147
148 if (br_stack_sz) {
149 size_t sz = sizeof(struct branch_stack);
150
151 sz += br_stack_sz * sizeof(struct branch_entry);
152 ts->br_stack_rb = zalloc(sz);
153 if (!ts->br_stack_rb)
154 return -ENOMEM;
155 ts->br_stack_sz = br_stack_sz;
156 }
157
158 if (thread->maps && thread->maps->machine) {
159 struct machine *machine = thread->maps->machine;
160 const char *arch = perf_env__arch(machine->env);
161
162 ts->kernel_start = machine__kernel_start(machine);
163 if (!strcmp(arch, "x86"))
164 ts->rstate = X86_RETPOLINE_POSSIBLE;
165 } else {
166 ts->kernel_start = 1ULL << 63;
167 }
168 ts->crp = crp;
169
170 return 0;
171}
172
173static struct thread_stack *thread_stack__new(struct thread *thread, int cpu,
174 struct call_return_processor *crp,
175 bool callstack,
176 unsigned int br_stack_sz)
177{
178 struct thread_stack *ts = thread->ts, *new_ts;
179 unsigned int old_sz = ts ? ts->arr_sz : 0;
180 unsigned int new_sz = 1;
181
182 if (thread_stack__per_cpu(thread) && cpu > 0)
183 new_sz = roundup_pow_of_two(cpu + 1);
184
185 if (!ts || new_sz > old_sz) {
186 new_ts = calloc(new_sz, sizeof(*ts));
187 if (!new_ts)
188 return NULL;
189 if (ts)
190 memcpy(new_ts, ts, old_sz * sizeof(*ts));
191 new_ts->arr_sz = new_sz;
192 zfree(&thread->ts);
193 thread->ts = new_ts;
194 ts = new_ts;
195 }
196
197 if (thread_stack__per_cpu(thread) && cpu > 0 &&
198 (unsigned int)cpu < ts->arr_sz)
199 ts += cpu;
200
201 if (!ts->stack &&
202 thread_stack__init(ts, thread, crp, callstack, br_stack_sz))
203 return NULL;
204
205 return ts;
206}
207
208static struct thread_stack *thread__cpu_stack(struct thread *thread, int cpu)
209{
210 struct thread_stack *ts = thread->ts;
211
212 if (cpu < 0)
213 cpu = 0;
214
215 if (!ts || (unsigned int)cpu >= ts->arr_sz)
216 return NULL;
217
218 ts += cpu;
219
220 if (!ts->stack)
221 return NULL;
222
223 return ts;
224}
225
226static inline struct thread_stack *thread__stack(struct thread *thread,
227 int cpu)
228{
229 if (!thread)
230 return NULL;
231
232 if (thread_stack__per_cpu(thread))
233 return thread__cpu_stack(thread, cpu);
234
235 return thread->ts;
236}
237
238static int thread_stack__push(struct thread_stack *ts, u64 ret_addr,
239 bool trace_end)
240{
241 int err = 0;
242
243 if (ts->cnt == ts->sz) {
244 err = thread_stack__grow(ts);
245 if (err) {
246 pr_warning("Out of memory: discarding thread stack\n");
247 ts->cnt = 0;
248 }
249 }
250
251 ts->stack[ts->cnt].trace_end = trace_end;
252 ts->stack[ts->cnt++].ret_addr = ret_addr;
253
254 return err;
255}
256
257static void thread_stack__pop(struct thread_stack *ts, u64 ret_addr)
258{
259 size_t i;
260
261 /*
262 * In some cases there may be functions which are not seen to return.
263 * For example when setjmp / longjmp has been used. Or the perf context
264 * switch in the kernel which doesn't stop and start tracing in exactly
265 * the same code path. When that happens the return address will be
266 * further down the stack. If the return address is not found at all,
267 * we assume the opposite (i.e. this is a return for a call that wasn't
268 * seen for some reason) and leave the stack alone.
269 */
270 for (i = ts->cnt; i; ) {
271 if (ts->stack[--i].ret_addr == ret_addr) {
272 ts->cnt = i;
273 return;
274 }
275 }
276}
277
278static void thread_stack__pop_trace_end(struct thread_stack *ts)
279{
280 size_t i;
281
282 for (i = ts->cnt; i; ) {
283 if (ts->stack[--i].trace_end)
284 ts->cnt = i;
285 else
286 return;
287 }
288}
289
290static bool thread_stack__in_kernel(struct thread_stack *ts)
291{
292 if (!ts->cnt)
293 return false;
294
295 return ts->stack[ts->cnt - 1].cp->in_kernel;
296}
297
298static int thread_stack__call_return(struct thread *thread,
299 struct thread_stack *ts, size_t idx,
300 u64 timestamp, u64 ref, bool no_return)
301{
302 struct call_return_processor *crp = ts->crp;
303 struct thread_stack_entry *tse;
304 struct call_return cr = {
305 .thread = thread,
306 .comm = ts->comm,
307 .db_id = 0,
308 };
309 u64 *parent_db_id;
310
311 tse = &ts->stack[idx];
312 cr.cp = tse->cp;
313 cr.call_time = tse->timestamp;
314 cr.return_time = timestamp;
315 cr.branch_count = ts->branch_count - tse->branch_count;
316 cr.insn_count = ts->insn_count - tse->insn_count;
317 cr.cyc_count = ts->cyc_count - tse->cyc_count;
318 cr.db_id = tse->db_id;
319 cr.call_ref = tse->ref;
320 cr.return_ref = ref;
321 if (tse->no_call)
322 cr.flags |= CALL_RETURN_NO_CALL;
323 if (no_return)
324 cr.flags |= CALL_RETURN_NO_RETURN;
325 if (tse->non_call)
326 cr.flags |= CALL_RETURN_NON_CALL;
327
328 /*
329 * The parent db_id must be assigned before exporting the child. Note
330 * it is not possible to export the parent first because its information
331 * is not yet complete because its 'return' has not yet been processed.
332 */
333 parent_db_id = idx ? &(tse - 1)->db_id : NULL;
334
335 return crp->process(&cr, parent_db_id, crp->data);
336}
337
338static int __thread_stack__flush(struct thread *thread, struct thread_stack *ts)
339{
340 struct call_return_processor *crp = ts->crp;
341 int err;
342
343 if (!crp) {
344 ts->cnt = 0;
345 ts->br_stack_pos = 0;
346 if (ts->br_stack_rb)
347 ts->br_stack_rb->nr = 0;
348 return 0;
349 }
350
351 while (ts->cnt) {
352 err = thread_stack__call_return(thread, ts, --ts->cnt,
353 ts->last_time, 0, true);
354 if (err) {
355 pr_err("Error flushing thread stack!\n");
356 ts->cnt = 0;
357 return err;
358 }
359 }
360
361 return 0;
362}
363
364int thread_stack__flush(struct thread *thread)
365{
366 struct thread_stack *ts = thread->ts;
367 unsigned int pos;
368 int err = 0;
369
370 if (ts) {
371 for (pos = 0; pos < ts->arr_sz; pos++) {
372 int ret = __thread_stack__flush(thread, ts + pos);
373
374 if (ret)
375 err = ret;
376 }
377 }
378
379 return err;
380}
381
382static void thread_stack__update_br_stack(struct thread_stack *ts, u32 flags,
383 u64 from_ip, u64 to_ip)
384{
385 struct branch_stack *bs = ts->br_stack_rb;
386 struct branch_entry *be;
387
388 if (!ts->br_stack_pos)
389 ts->br_stack_pos = ts->br_stack_sz;
390
391 ts->br_stack_pos -= 1;
392
393 be = &bs->entries[ts->br_stack_pos];
394 be->from = from_ip;
395 be->to = to_ip;
396 be->flags.value = 0;
397 be->flags.abort = !!(flags & PERF_IP_FLAG_TX_ABORT);
398 be->flags.in_tx = !!(flags & PERF_IP_FLAG_IN_TX);
399 /* No support for mispredict */
400 be->flags.mispred = ts->mispred_all;
401
402 if (bs->nr < ts->br_stack_sz)
403 bs->nr += 1;
404}
405
406int thread_stack__event(struct thread *thread, int cpu, u32 flags, u64 from_ip,
407 u64 to_ip, u16 insn_len, u64 trace_nr, bool callstack,
408 unsigned int br_stack_sz, bool mispred_all)
409{
410 struct thread_stack *ts = thread__stack(thread, cpu);
411
412 if (!thread)
413 return -EINVAL;
414
415 if (!ts) {
416 ts = thread_stack__new(thread, cpu, NULL, callstack, br_stack_sz);
417 if (!ts) {
418 pr_warning("Out of memory: no thread stack\n");
419 return -ENOMEM;
420 }
421 ts->trace_nr = trace_nr;
422 ts->mispred_all = mispred_all;
423 }
424
425 /*
426 * When the trace is discontinuous, the trace_nr changes. In that case
427 * the stack might be completely invalid. Better to report nothing than
428 * to report something misleading, so flush the stack.
429 */
430 if (trace_nr != ts->trace_nr) {
431 if (ts->trace_nr)
432 __thread_stack__flush(thread, ts);
433 ts->trace_nr = trace_nr;
434 }
435
436 if (br_stack_sz)
437 thread_stack__update_br_stack(ts, flags, from_ip, to_ip);
438
439 /*
440 * Stop here if thread_stack__process() is in use, or not recording call
441 * stack.
442 */
443 if (ts->crp || !callstack)
444 return 0;
445
446 if (flags & PERF_IP_FLAG_CALL) {
447 u64 ret_addr;
448
449 if (!to_ip)
450 return 0;
451 ret_addr = from_ip + insn_len;
452 if (ret_addr == to_ip)
453 return 0; /* Zero-length calls are excluded */
454 return thread_stack__push(ts, ret_addr,
455 flags & PERF_IP_FLAG_TRACE_END);
456 } else if (flags & PERF_IP_FLAG_TRACE_BEGIN) {
457 /*
458 * If the caller did not change the trace number (which would
459 * have flushed the stack) then try to make sense of the stack.
460 * Possibly, tracing began after returning to the current
461 * address, so try to pop that. Also, do not expect a call made
462 * when the trace ended, to return, so pop that.
463 */
464 thread_stack__pop(ts, to_ip);
465 thread_stack__pop_trace_end(ts);
466 } else if ((flags & PERF_IP_FLAG_RETURN) && from_ip) {
467 thread_stack__pop(ts, to_ip);
468 }
469
470 return 0;
471}
472
473void thread_stack__set_trace_nr(struct thread *thread, int cpu, u64 trace_nr)
474{
475 struct thread_stack *ts = thread__stack(thread, cpu);
476
477 if (!ts)
478 return;
479
480 if (trace_nr != ts->trace_nr) {
481 if (ts->trace_nr)
482 __thread_stack__flush(thread, ts);
483 ts->trace_nr = trace_nr;
484 }
485}
486
487static void __thread_stack__free(struct thread *thread, struct thread_stack *ts)
488{
489 __thread_stack__flush(thread, ts);
490 zfree(&ts->stack);
491 zfree(&ts->br_stack_rb);
492}
493
494static void thread_stack__reset(struct thread *thread, struct thread_stack *ts)
495{
496 unsigned int arr_sz = ts->arr_sz;
497
498 __thread_stack__free(thread, ts);
499 memset(ts, 0, sizeof(*ts));
500 ts->arr_sz = arr_sz;
501}
502
503void thread_stack__free(struct thread *thread)
504{
505 struct thread_stack *ts = thread->ts;
506 unsigned int pos;
507
508 if (ts) {
509 for (pos = 0; pos < ts->arr_sz; pos++)
510 __thread_stack__free(thread, ts + pos);
511 zfree(&thread->ts);
512 }
513}
514
515static inline u64 callchain_context(u64 ip, u64 kernel_start)
516{
517 return ip < kernel_start ? PERF_CONTEXT_USER : PERF_CONTEXT_KERNEL;
518}
519
520void thread_stack__sample(struct thread *thread, int cpu,
521 struct ip_callchain *chain,
522 size_t sz, u64 ip, u64 kernel_start)
523{
524 struct thread_stack *ts = thread__stack(thread, cpu);
525 u64 context = callchain_context(ip, kernel_start);
526 u64 last_context;
527 size_t i, j;
528
529 if (sz < 2) {
530 chain->nr = 0;
531 return;
532 }
533
534 chain->ips[0] = context;
535 chain->ips[1] = ip;
536
537 if (!ts) {
538 chain->nr = 2;
539 return;
540 }
541
542 last_context = context;
543
544 for (i = 2, j = 1; i < sz && j <= ts->cnt; i++, j++) {
545 ip = ts->stack[ts->cnt - j].ret_addr;
546 context = callchain_context(ip, kernel_start);
547 if (context != last_context) {
548 if (i >= sz - 1)
549 break;
550 chain->ips[i++] = context;
551 last_context = context;
552 }
553 chain->ips[i] = ip;
554 }
555
556 chain->nr = i;
557}
558
559/*
560 * Hardware sample records, created some time after the event occurred, need to
561 * have subsequent addresses removed from the call chain.
562 */
563void thread_stack__sample_late(struct thread *thread, int cpu,
564 struct ip_callchain *chain, size_t sz,
565 u64 sample_ip, u64 kernel_start)
566{
567 struct thread_stack *ts = thread__stack(thread, cpu);
568 u64 sample_context = callchain_context(sample_ip, kernel_start);
569 u64 last_context, context, ip;
570 size_t nr = 0, j;
571
572 if (sz < 2) {
573 chain->nr = 0;
574 return;
575 }
576
577 if (!ts)
578 goto out;
579
580 /*
581 * When tracing kernel space, kernel addresses occur at the top of the
582 * call chain after the event occurred but before tracing stopped.
583 * Skip them.
584 */
585 for (j = 1; j <= ts->cnt; j++) {
586 ip = ts->stack[ts->cnt - j].ret_addr;
587 context = callchain_context(ip, kernel_start);
588 if (context == PERF_CONTEXT_USER ||
589 (context == sample_context && ip == sample_ip))
590 break;
591 }
592
593 last_context = sample_ip; /* Use sample_ip as an invalid context */
594
595 for (; nr < sz && j <= ts->cnt; nr++, j++) {
596 ip = ts->stack[ts->cnt - j].ret_addr;
597 context = callchain_context(ip, kernel_start);
598 if (context != last_context) {
599 if (nr >= sz - 1)
600 break;
601 chain->ips[nr++] = context;
602 last_context = context;
603 }
604 chain->ips[nr] = ip;
605 }
606out:
607 if (nr) {
608 chain->nr = nr;
609 } else {
610 chain->ips[0] = sample_context;
611 chain->ips[1] = sample_ip;
612 chain->nr = 2;
613 }
614}
615
616void thread_stack__br_sample(struct thread *thread, int cpu,
617 struct branch_stack *dst, unsigned int sz)
618{
619 struct thread_stack *ts = thread__stack(thread, cpu);
620 const size_t bsz = sizeof(struct branch_entry);
621 struct branch_stack *src;
622 struct branch_entry *be;
623 unsigned int nr;
624
625 dst->nr = 0;
626
627 if (!ts)
628 return;
629
630 src = ts->br_stack_rb;
631 if (!src->nr)
632 return;
633
634 dst->nr = min((unsigned int)src->nr, sz);
635
636 be = &dst->entries[0];
637 nr = min(ts->br_stack_sz - ts->br_stack_pos, (unsigned int)dst->nr);
638 memcpy(be, &src->entries[ts->br_stack_pos], bsz * nr);
639
640 if (src->nr >= ts->br_stack_sz) {
641 sz -= nr;
642 be = &dst->entries[nr];
643 nr = min(ts->br_stack_pos, sz);
644 memcpy(be, &src->entries[0], bsz * ts->br_stack_pos);
645 }
646}
647
648/* Start of user space branch entries */
649static bool us_start(struct branch_entry *be, u64 kernel_start, bool *start)
650{
651 if (!*start)
652 *start = be->to && be->to < kernel_start;
653
654 return *start;
655}
656
657/*
658 * Start of branch entries after the ip fell in between 2 branches, or user
659 * space branch entries.
660 */
661static bool ks_start(struct branch_entry *be, u64 sample_ip, u64 kernel_start,
662 bool *start, struct branch_entry *nb)
663{
664 if (!*start) {
665 *start = (nb && sample_ip >= be->to && sample_ip <= nb->from) ||
666 be->from < kernel_start ||
667 (be->to && be->to < kernel_start);
668 }
669
670 return *start;
671}
672
673/*
674 * Hardware sample records, created some time after the event occurred, need to
675 * have subsequent addresses removed from the branch stack.
676 */
677void thread_stack__br_sample_late(struct thread *thread, int cpu,
678 struct branch_stack *dst, unsigned int sz,
679 u64 ip, u64 kernel_start)
680{
681 struct thread_stack *ts = thread__stack(thread, cpu);
682 struct branch_entry *d, *s, *spos, *ssz;
683 struct branch_stack *src;
684 unsigned int nr = 0;
685 bool start = false;
686
687 dst->nr = 0;
688
689 if (!ts)
690 return;
691
692 src = ts->br_stack_rb;
693 if (!src->nr)
694 return;
695
696 spos = &src->entries[ts->br_stack_pos];
697 ssz = &src->entries[ts->br_stack_sz];
698
699 d = &dst->entries[0];
700 s = spos;
701
702 if (ip < kernel_start) {
703 /*
704 * User space sample: start copying branch entries when the
705 * branch is in user space.
706 */
707 for (s = spos; s < ssz && nr < sz; s++) {
708 if (us_start(s, kernel_start, &start)) {
709 *d++ = *s;
710 nr += 1;
711 }
712 }
713
714 if (src->nr >= ts->br_stack_sz) {
715 for (s = &src->entries[0]; s < spos && nr < sz; s++) {
716 if (us_start(s, kernel_start, &start)) {
717 *d++ = *s;
718 nr += 1;
719 }
720 }
721 }
722 } else {
723 struct branch_entry *nb = NULL;
724
725 /*
726 * Kernel space sample: start copying branch entries when the ip
727 * falls in between 2 branches (or the branch is in user space
728 * because then the start must have been missed).
729 */
730 for (s = spos; s < ssz && nr < sz; s++) {
731 if (ks_start(s, ip, kernel_start, &start, nb)) {
732 *d++ = *s;
733 nr += 1;
734 }
735 nb = s;
736 }
737
738 if (src->nr >= ts->br_stack_sz) {
739 for (s = &src->entries[0]; s < spos && nr < sz; s++) {
740 if (ks_start(s, ip, kernel_start, &start, nb)) {
741 *d++ = *s;
742 nr += 1;
743 }
744 nb = s;
745 }
746 }
747 }
748
749 dst->nr = nr;
750}
751
752struct call_return_processor *
753call_return_processor__new(int (*process)(struct call_return *cr, u64 *parent_db_id, void *data),
754 void *data)
755{
756 struct call_return_processor *crp;
757
758 crp = zalloc(sizeof(struct call_return_processor));
759 if (!crp)
760 return NULL;
761 crp->cpr = call_path_root__new();
762 if (!crp->cpr)
763 goto out_free;
764 crp->process = process;
765 crp->data = data;
766 return crp;
767
768out_free:
769 free(crp);
770 return NULL;
771}
772
773void call_return_processor__free(struct call_return_processor *crp)
774{
775 if (crp) {
776 call_path_root__free(crp->cpr);
777 free(crp);
778 }
779}
780
781static int thread_stack__push_cp(struct thread_stack *ts, u64 ret_addr,
782 u64 timestamp, u64 ref, struct call_path *cp,
783 bool no_call, bool trace_end)
784{
785 struct thread_stack_entry *tse;
786 int err;
787
788 if (!cp)
789 return -ENOMEM;
790
791 if (ts->cnt == ts->sz) {
792 err = thread_stack__grow(ts);
793 if (err)
794 return err;
795 }
796
797 tse = &ts->stack[ts->cnt++];
798 tse->ret_addr = ret_addr;
799 tse->timestamp = timestamp;
800 tse->ref = ref;
801 tse->branch_count = ts->branch_count;
802 tse->insn_count = ts->insn_count;
803 tse->cyc_count = ts->cyc_count;
804 tse->cp = cp;
805 tse->no_call = no_call;
806 tse->trace_end = trace_end;
807 tse->non_call = false;
808 tse->db_id = 0;
809
810 return 0;
811}
812
813static int thread_stack__pop_cp(struct thread *thread, struct thread_stack *ts,
814 u64 ret_addr, u64 timestamp, u64 ref,
815 struct symbol *sym)
816{
817 int err;
818
819 if (!ts->cnt)
820 return 1;
821
822 if (ts->cnt == 1) {
823 struct thread_stack_entry *tse = &ts->stack[0];
824
825 if (tse->cp->sym == sym)
826 return thread_stack__call_return(thread, ts, --ts->cnt,
827 timestamp, ref, false);
828 }
829
830 if (ts->stack[ts->cnt - 1].ret_addr == ret_addr &&
831 !ts->stack[ts->cnt - 1].non_call) {
832 return thread_stack__call_return(thread, ts, --ts->cnt,
833 timestamp, ref, false);
834 } else {
835 size_t i = ts->cnt - 1;
836
837 while (i--) {
838 if (ts->stack[i].ret_addr != ret_addr ||
839 ts->stack[i].non_call)
840 continue;
841 i += 1;
842 while (ts->cnt > i) {
843 err = thread_stack__call_return(thread, ts,
844 --ts->cnt,
845 timestamp, ref,
846 true);
847 if (err)
848 return err;
849 }
850 return thread_stack__call_return(thread, ts, --ts->cnt,
851 timestamp, ref, false);
852 }
853 }
854
855 return 1;
856}
857
858static int thread_stack__bottom(struct thread_stack *ts,
859 struct perf_sample *sample,
860 struct addr_location *from_al,
861 struct addr_location *to_al, u64 ref)
862{
863 struct call_path_root *cpr = ts->crp->cpr;
864 struct call_path *cp;
865 struct symbol *sym;
866 u64 ip;
867
868 if (sample->ip) {
869 ip = sample->ip;
870 sym = from_al->sym;
871 } else if (sample->addr) {
872 ip = sample->addr;
873 sym = to_al->sym;
874 } else {
875 return 0;
876 }
877
878 cp = call_path__findnew(cpr, &cpr->call_path, sym, ip,
879 ts->kernel_start);
880
881 return thread_stack__push_cp(ts, ip, sample->time, ref, cp,
882 true, false);
883}
884
885static int thread_stack__pop_ks(struct thread *thread, struct thread_stack *ts,
886 struct perf_sample *sample, u64 ref)
887{
888 u64 tm = sample->time;
889 int err;
890
891 /* Return to userspace, so pop all kernel addresses */
892 while (thread_stack__in_kernel(ts)) {
893 err = thread_stack__call_return(thread, ts, --ts->cnt,
894 tm, ref, true);
895 if (err)
896 return err;
897 }
898
899 return 0;
900}
901
902static int thread_stack__no_call_return(struct thread *thread,
903 struct thread_stack *ts,
904 struct perf_sample *sample,
905 struct addr_location *from_al,
906 struct addr_location *to_al, u64 ref)
907{
908 struct call_path_root *cpr = ts->crp->cpr;
909 struct call_path *root = &cpr->call_path;
910 struct symbol *fsym = from_al->sym;
911 struct symbol *tsym = to_al->sym;
912 struct call_path *cp, *parent;
913 u64 ks = ts->kernel_start;
914 u64 addr = sample->addr;
915 u64 tm = sample->time;
916 u64 ip = sample->ip;
917 int err;
918
919 if (ip >= ks && addr < ks) {
920 /* Return to userspace, so pop all kernel addresses */
921 err = thread_stack__pop_ks(thread, ts, sample, ref);
922 if (err)
923 return err;
924
925 /* If the stack is empty, push the userspace address */
926 if (!ts->cnt) {
927 cp = call_path__findnew(cpr, root, tsym, addr, ks);
928 return thread_stack__push_cp(ts, 0, tm, ref, cp, true,
929 false);
930 }
931 } else if (thread_stack__in_kernel(ts) && ip < ks) {
932 /* Return to userspace, so pop all kernel addresses */
933 err = thread_stack__pop_ks(thread, ts, sample, ref);
934 if (err)
935 return err;
936 }
937
938 if (ts->cnt)
939 parent = ts->stack[ts->cnt - 1].cp;
940 else
941 parent = root;
942
943 if (parent->sym == from_al->sym) {
944 /*
945 * At the bottom of the stack, assume the missing 'call' was
946 * before the trace started. So, pop the current symbol and push
947 * the 'to' symbol.
948 */
949 if (ts->cnt == 1) {
950 err = thread_stack__call_return(thread, ts, --ts->cnt,
951 tm, ref, false);
952 if (err)
953 return err;
954 }
955
956 if (!ts->cnt) {
957 cp = call_path__findnew(cpr, root, tsym, addr, ks);
958
959 return thread_stack__push_cp(ts, addr, tm, ref, cp,
960 true, false);
961 }
962
963 /*
964 * Otherwise assume the 'return' is being used as a jump (e.g.
965 * retpoline) and just push the 'to' symbol.
966 */
967 cp = call_path__findnew(cpr, parent, tsym, addr, ks);
968
969 err = thread_stack__push_cp(ts, 0, tm, ref, cp, true, false);
970 if (!err)
971 ts->stack[ts->cnt - 1].non_call = true;
972
973 return err;
974 }
975
976 /*
977 * Assume 'parent' has not yet returned, so push 'to', and then push and
978 * pop 'from'.
979 */
980
981 cp = call_path__findnew(cpr, parent, tsym, addr, ks);
982
983 err = thread_stack__push_cp(ts, addr, tm, ref, cp, true, false);
984 if (err)
985 return err;
986
987 cp = call_path__findnew(cpr, cp, fsym, ip, ks);
988
989 err = thread_stack__push_cp(ts, ip, tm, ref, cp, true, false);
990 if (err)
991 return err;
992
993 return thread_stack__call_return(thread, ts, --ts->cnt, tm, ref, false);
994}
995
996static int thread_stack__trace_begin(struct thread *thread,
997 struct thread_stack *ts, u64 timestamp,
998 u64 ref)
999{
1000 struct thread_stack_entry *tse;
1001 int err;
1002
1003 if (!ts->cnt)
1004 return 0;
1005
1006 /* Pop trace end */
1007 tse = &ts->stack[ts->cnt - 1];
1008 if (tse->trace_end) {
1009 err = thread_stack__call_return(thread, ts, --ts->cnt,
1010 timestamp, ref, false);
1011 if (err)
1012 return err;
1013 }
1014
1015 return 0;
1016}
1017
1018static int thread_stack__trace_end(struct thread_stack *ts,
1019 struct perf_sample *sample, u64 ref)
1020{
1021 struct call_path_root *cpr = ts->crp->cpr;
1022 struct call_path *cp;
1023 u64 ret_addr;
1024
1025 /* No point having 'trace end' on the bottom of the stack */
1026 if (!ts->cnt || (ts->cnt == 1 && ts->stack[0].ref == ref))
1027 return 0;
1028
1029 cp = call_path__findnew(cpr, ts->stack[ts->cnt - 1].cp, NULL, 0,
1030 ts->kernel_start);
1031
1032 ret_addr = sample->ip + sample->insn_len;
1033
1034 return thread_stack__push_cp(ts, ret_addr, sample->time, ref, cp,
1035 false, true);
1036}
1037
1038static bool is_x86_retpoline(const char *name)
1039{
1040 const char *p = strstr(name, "__x86_indirect_thunk_");
1041
1042 return p == name || !strcmp(name, "__indirect_thunk_start");
1043}
1044
1045/*
1046 * x86 retpoline functions pollute the call graph. This function removes them.
1047 * This does not handle function return thunks, nor is there any improvement
1048 * for the handling of inline thunks or extern thunks.
1049 */
1050static int thread_stack__x86_retpoline(struct thread_stack *ts,
1051 struct perf_sample *sample,
1052 struct addr_location *to_al)
1053{
1054 struct thread_stack_entry *tse = &ts->stack[ts->cnt - 1];
1055 struct call_path_root *cpr = ts->crp->cpr;
1056 struct symbol *sym = tse->cp->sym;
1057 struct symbol *tsym = to_al->sym;
1058 struct call_path *cp;
1059
1060 if (sym && is_x86_retpoline(sym->name)) {
1061 /*
1062 * This is a x86 retpoline fn. It pollutes the call graph by
1063 * showing up everywhere there is an indirect branch, but does
1064 * not itself mean anything. Here the top-of-stack is removed,
1065 * by decrementing the stack count, and then further down, the
1066 * resulting top-of-stack is replaced with the actual target.
1067 * The result is that the retpoline functions will no longer
1068 * appear in the call graph. Note this only affects the call
1069 * graph, since all the original branches are left unchanged.
1070 */
1071 ts->cnt -= 1;
1072 sym = ts->stack[ts->cnt - 2].cp->sym;
1073 if (sym && sym == tsym && to_al->addr != tsym->start) {
1074 /*
1075 * Target is back to the middle of the symbol we came
1076 * from so assume it is an indirect jmp and forget it
1077 * altogether.
1078 */
1079 ts->cnt -= 1;
1080 return 0;
1081 }
1082 } else if (sym && sym == tsym) {
1083 /*
1084 * Target is back to the symbol we came from so assume it is an
1085 * indirect jmp and forget it altogether.
1086 */
1087 ts->cnt -= 1;
1088 return 0;
1089 }
1090
1091 cp = call_path__findnew(cpr, ts->stack[ts->cnt - 2].cp, tsym,
1092 sample->addr, ts->kernel_start);
1093 if (!cp)
1094 return -ENOMEM;
1095
1096 /* Replace the top-of-stack with the actual target */
1097 ts->stack[ts->cnt - 1].cp = cp;
1098
1099 return 0;
1100}
1101
1102int thread_stack__process(struct thread *thread, struct comm *comm,
1103 struct perf_sample *sample,
1104 struct addr_location *from_al,
1105 struct addr_location *to_al, u64 ref,
1106 struct call_return_processor *crp)
1107{
1108 struct thread_stack *ts = thread__stack(thread, sample->cpu);
1109 enum retpoline_state_t rstate;
1110 int err = 0;
1111
1112 if (ts && !ts->crp) {
1113 /* Supersede thread_stack__event() */
1114 thread_stack__reset(thread, ts);
1115 ts = NULL;
1116 }
1117
1118 if (!ts) {
1119 ts = thread_stack__new(thread, sample->cpu, crp, true, 0);
1120 if (!ts)
1121 return -ENOMEM;
1122 ts->comm = comm;
1123 }
1124
1125 rstate = ts->rstate;
1126 if (rstate == X86_RETPOLINE_DETECTED)
1127 ts->rstate = X86_RETPOLINE_POSSIBLE;
1128
1129 /* Flush stack on exec */
1130 if (ts->comm != comm && thread->pid_ == thread->tid) {
1131 err = __thread_stack__flush(thread, ts);
1132 if (err)
1133 return err;
1134 ts->comm = comm;
1135 }
1136
1137 /* If the stack is empty, put the current symbol on the stack */
1138 if (!ts->cnt) {
1139 err = thread_stack__bottom(ts, sample, from_al, to_al, ref);
1140 if (err)
1141 return err;
1142 }
1143
1144 ts->branch_count += 1;
1145 ts->insn_count += sample->insn_cnt;
1146 ts->cyc_count += sample->cyc_cnt;
1147 ts->last_time = sample->time;
1148
1149 if (sample->flags & PERF_IP_FLAG_CALL) {
1150 bool trace_end = sample->flags & PERF_IP_FLAG_TRACE_END;
1151 struct call_path_root *cpr = ts->crp->cpr;
1152 struct call_path *cp;
1153 u64 ret_addr;
1154
1155 if (!sample->ip || !sample->addr)
1156 return 0;
1157
1158 ret_addr = sample->ip + sample->insn_len;
1159 if (ret_addr == sample->addr)
1160 return 0; /* Zero-length calls are excluded */
1161
1162 cp = call_path__findnew(cpr, ts->stack[ts->cnt - 1].cp,
1163 to_al->sym, sample->addr,
1164 ts->kernel_start);
1165 err = thread_stack__push_cp(ts, ret_addr, sample->time, ref,
1166 cp, false, trace_end);
1167
1168 /*
1169 * A call to the same symbol but not the start of the symbol,
1170 * may be the start of a x86 retpoline.
1171 */
1172 if (!err && rstate == X86_RETPOLINE_POSSIBLE && to_al->sym &&
1173 from_al->sym == to_al->sym &&
1174 to_al->addr != to_al->sym->start)
1175 ts->rstate = X86_RETPOLINE_DETECTED;
1176
1177 } else if (sample->flags & PERF_IP_FLAG_RETURN) {
1178 if (!sample->addr) {
1179 u32 return_from_kernel = PERF_IP_FLAG_SYSCALLRET |
1180 PERF_IP_FLAG_INTERRUPT;
1181
1182 if (!(sample->flags & return_from_kernel))
1183 return 0;
1184
1185 /* Pop kernel stack */
1186 return thread_stack__pop_ks(thread, ts, sample, ref);
1187 }
1188
1189 if (!sample->ip)
1190 return 0;
1191
1192 /* x86 retpoline 'return' doesn't match the stack */
1193 if (rstate == X86_RETPOLINE_DETECTED && ts->cnt > 2 &&
1194 ts->stack[ts->cnt - 1].ret_addr != sample->addr)
1195 return thread_stack__x86_retpoline(ts, sample, to_al);
1196
1197 err = thread_stack__pop_cp(thread, ts, sample->addr,
1198 sample->time, ref, from_al->sym);
1199 if (err) {
1200 if (err < 0)
1201 return err;
1202 err = thread_stack__no_call_return(thread, ts, sample,
1203 from_al, to_al, ref);
1204 }
1205 } else if (sample->flags & PERF_IP_FLAG_TRACE_BEGIN) {
1206 err = thread_stack__trace_begin(thread, ts, sample->time, ref);
1207 } else if (sample->flags & PERF_IP_FLAG_TRACE_END) {
1208 err = thread_stack__trace_end(ts, sample, ref);
1209 } else if (sample->flags & PERF_IP_FLAG_BRANCH &&
1210 from_al->sym != to_al->sym && to_al->sym &&
1211 to_al->addr == to_al->sym->start) {
1212 struct call_path_root *cpr = ts->crp->cpr;
1213 struct call_path *cp;
1214
1215 /*
1216 * The compiler might optimize a call/ret combination by making
1217 * it a jmp. Make that visible by recording on the stack a
1218 * branch to the start of a different symbol. Note, that means
1219 * when a ret pops the stack, all jmps must be popped off first.
1220 */
1221 cp = call_path__findnew(cpr, ts->stack[ts->cnt - 1].cp,
1222 to_al->sym, sample->addr,
1223 ts->kernel_start);
1224 err = thread_stack__push_cp(ts, 0, sample->time, ref, cp, false,
1225 false);
1226 if (!err)
1227 ts->stack[ts->cnt - 1].non_call = true;
1228 }
1229
1230 return err;
1231}
1232
1233size_t thread_stack__depth(struct thread *thread, int cpu)
1234{
1235 struct thread_stack *ts = thread__stack(thread, cpu);
1236
1237 if (!ts)
1238 return 0;
1239 return ts->cnt;
1240}
1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * thread-stack.c: Synthesize a thread's stack using call / return events
4 * Copyright (c) 2014, Intel Corporation.
5 */
6
7#include <linux/rbtree.h>
8#include <linux/list.h>
9#include <linux/log2.h>
10#include <linux/zalloc.h>
11#include <errno.h>
12#include <stdlib.h>
13#include <string.h>
14#include "thread.h"
15#include "event.h"
16#include "machine.h"
17#include "env.h"
18#include "debug.h"
19#include "symbol.h"
20#include "comm.h"
21#include "call-path.h"
22#include "thread-stack.h"
23
24#define STACK_GROWTH 2048
25
26/*
27 * State of retpoline detection.
28 *
29 * RETPOLINE_NONE: no retpoline detection
30 * X86_RETPOLINE_POSSIBLE: x86 retpoline possible
31 * X86_RETPOLINE_DETECTED: x86 retpoline detected
32 */
33enum retpoline_state_t {
34 RETPOLINE_NONE,
35 X86_RETPOLINE_POSSIBLE,
36 X86_RETPOLINE_DETECTED,
37};
38
39/**
40 * struct thread_stack_entry - thread stack entry.
41 * @ret_addr: return address
42 * @timestamp: timestamp (if known)
43 * @ref: external reference (e.g. db_id of sample)
44 * @branch_count: the branch count when the entry was created
45 * @insn_count: the instruction count when the entry was created
46 * @cyc_count the cycle count when the entry was created
47 * @db_id: id used for db-export
48 * @cp: call path
49 * @no_call: a 'call' was not seen
50 * @trace_end: a 'call' but trace ended
51 * @non_call: a branch but not a 'call' to the start of a different symbol
52 */
53struct thread_stack_entry {
54 u64 ret_addr;
55 u64 timestamp;
56 u64 ref;
57 u64 branch_count;
58 u64 insn_count;
59 u64 cyc_count;
60 u64 db_id;
61 struct call_path *cp;
62 bool no_call;
63 bool trace_end;
64 bool non_call;
65};
66
67/**
68 * struct thread_stack - thread stack constructed from 'call' and 'return'
69 * branch samples.
70 * @stack: array that holds the stack
71 * @cnt: number of entries in the stack
72 * @sz: current maximum stack size
73 * @trace_nr: current trace number
74 * @branch_count: running branch count
75 * @insn_count: running instruction count
76 * @cyc_count running cycle count
77 * @kernel_start: kernel start address
78 * @last_time: last timestamp
79 * @crp: call/return processor
80 * @comm: current comm
81 * @arr_sz: size of array if this is the first element of an array
82 * @rstate: used to detect retpolines
83 */
84struct thread_stack {
85 struct thread_stack_entry *stack;
86 size_t cnt;
87 size_t sz;
88 u64 trace_nr;
89 u64 branch_count;
90 u64 insn_count;
91 u64 cyc_count;
92 u64 kernel_start;
93 u64 last_time;
94 struct call_return_processor *crp;
95 struct comm *comm;
96 unsigned int arr_sz;
97 enum retpoline_state_t rstate;
98};
99
100/*
101 * Assume pid == tid == 0 identifies the idle task as defined by
102 * perf_session__register_idle_thread(). The idle task is really 1 task per cpu,
103 * and therefore requires a stack for each cpu.
104 */
105static inline bool thread_stack__per_cpu(struct thread *thread)
106{
107 return !(thread->tid || thread->pid_);
108}
109
110static int thread_stack__grow(struct thread_stack *ts)
111{
112 struct thread_stack_entry *new_stack;
113 size_t sz, new_sz;
114
115 new_sz = ts->sz + STACK_GROWTH;
116 sz = new_sz * sizeof(struct thread_stack_entry);
117
118 new_stack = realloc(ts->stack, sz);
119 if (!new_stack)
120 return -ENOMEM;
121
122 ts->stack = new_stack;
123 ts->sz = new_sz;
124
125 return 0;
126}
127
128static int thread_stack__init(struct thread_stack *ts, struct thread *thread,
129 struct call_return_processor *crp)
130{
131 int err;
132
133 err = thread_stack__grow(ts);
134 if (err)
135 return err;
136
137 if (thread->mg && thread->mg->machine) {
138 struct machine *machine = thread->mg->machine;
139 const char *arch = perf_env__arch(machine->env);
140
141 ts->kernel_start = machine__kernel_start(machine);
142 if (!strcmp(arch, "x86"))
143 ts->rstate = X86_RETPOLINE_POSSIBLE;
144 } else {
145 ts->kernel_start = 1ULL << 63;
146 }
147 ts->crp = crp;
148
149 return 0;
150}
151
152static struct thread_stack *thread_stack__new(struct thread *thread, int cpu,
153 struct call_return_processor *crp)
154{
155 struct thread_stack *ts = thread->ts, *new_ts;
156 unsigned int old_sz = ts ? ts->arr_sz : 0;
157 unsigned int new_sz = 1;
158
159 if (thread_stack__per_cpu(thread) && cpu > 0)
160 new_sz = roundup_pow_of_two(cpu + 1);
161
162 if (!ts || new_sz > old_sz) {
163 new_ts = calloc(new_sz, sizeof(*ts));
164 if (!new_ts)
165 return NULL;
166 if (ts)
167 memcpy(new_ts, ts, old_sz * sizeof(*ts));
168 new_ts->arr_sz = new_sz;
169 zfree(&thread->ts);
170 thread->ts = new_ts;
171 ts = new_ts;
172 }
173
174 if (thread_stack__per_cpu(thread) && cpu > 0 &&
175 (unsigned int)cpu < ts->arr_sz)
176 ts += cpu;
177
178 if (!ts->stack &&
179 thread_stack__init(ts, thread, crp))
180 return NULL;
181
182 return ts;
183}
184
185static struct thread_stack *thread__cpu_stack(struct thread *thread, int cpu)
186{
187 struct thread_stack *ts = thread->ts;
188
189 if (cpu < 0)
190 cpu = 0;
191
192 if (!ts || (unsigned int)cpu >= ts->arr_sz)
193 return NULL;
194
195 ts += cpu;
196
197 if (!ts->stack)
198 return NULL;
199
200 return ts;
201}
202
203static inline struct thread_stack *thread__stack(struct thread *thread,
204 int cpu)
205{
206 if (!thread)
207 return NULL;
208
209 if (thread_stack__per_cpu(thread))
210 return thread__cpu_stack(thread, cpu);
211
212 return thread->ts;
213}
214
215static int thread_stack__push(struct thread_stack *ts, u64 ret_addr,
216 bool trace_end)
217{
218 int err = 0;
219
220 if (ts->cnt == ts->sz) {
221 err = thread_stack__grow(ts);
222 if (err) {
223 pr_warning("Out of memory: discarding thread stack\n");
224 ts->cnt = 0;
225 }
226 }
227
228 ts->stack[ts->cnt].trace_end = trace_end;
229 ts->stack[ts->cnt++].ret_addr = ret_addr;
230
231 return err;
232}
233
234static void thread_stack__pop(struct thread_stack *ts, u64 ret_addr)
235{
236 size_t i;
237
238 /*
239 * In some cases there may be functions which are not seen to return.
240 * For example when setjmp / longjmp has been used. Or the perf context
241 * switch in the kernel which doesn't stop and start tracing in exactly
242 * the same code path. When that happens the return address will be
243 * further down the stack. If the return address is not found at all,
244 * we assume the opposite (i.e. this is a return for a call that wasn't
245 * seen for some reason) and leave the stack alone.
246 */
247 for (i = ts->cnt; i; ) {
248 if (ts->stack[--i].ret_addr == ret_addr) {
249 ts->cnt = i;
250 return;
251 }
252 }
253}
254
255static void thread_stack__pop_trace_end(struct thread_stack *ts)
256{
257 size_t i;
258
259 for (i = ts->cnt; i; ) {
260 if (ts->stack[--i].trace_end)
261 ts->cnt = i;
262 else
263 return;
264 }
265}
266
267static bool thread_stack__in_kernel(struct thread_stack *ts)
268{
269 if (!ts->cnt)
270 return false;
271
272 return ts->stack[ts->cnt - 1].cp->in_kernel;
273}
274
275static int thread_stack__call_return(struct thread *thread,
276 struct thread_stack *ts, size_t idx,
277 u64 timestamp, u64 ref, bool no_return)
278{
279 struct call_return_processor *crp = ts->crp;
280 struct thread_stack_entry *tse;
281 struct call_return cr = {
282 .thread = thread,
283 .comm = ts->comm,
284 .db_id = 0,
285 };
286 u64 *parent_db_id;
287
288 tse = &ts->stack[idx];
289 cr.cp = tse->cp;
290 cr.call_time = tse->timestamp;
291 cr.return_time = timestamp;
292 cr.branch_count = ts->branch_count - tse->branch_count;
293 cr.insn_count = ts->insn_count - tse->insn_count;
294 cr.cyc_count = ts->cyc_count - tse->cyc_count;
295 cr.db_id = tse->db_id;
296 cr.call_ref = tse->ref;
297 cr.return_ref = ref;
298 if (tse->no_call)
299 cr.flags |= CALL_RETURN_NO_CALL;
300 if (no_return)
301 cr.flags |= CALL_RETURN_NO_RETURN;
302 if (tse->non_call)
303 cr.flags |= CALL_RETURN_NON_CALL;
304
305 /*
306 * The parent db_id must be assigned before exporting the child. Note
307 * it is not possible to export the parent first because its information
308 * is not yet complete because its 'return' has not yet been processed.
309 */
310 parent_db_id = idx ? &(tse - 1)->db_id : NULL;
311
312 return crp->process(&cr, parent_db_id, crp->data);
313}
314
315static int __thread_stack__flush(struct thread *thread, struct thread_stack *ts)
316{
317 struct call_return_processor *crp = ts->crp;
318 int err;
319
320 if (!crp) {
321 ts->cnt = 0;
322 return 0;
323 }
324
325 while (ts->cnt) {
326 err = thread_stack__call_return(thread, ts, --ts->cnt,
327 ts->last_time, 0, true);
328 if (err) {
329 pr_err("Error flushing thread stack!\n");
330 ts->cnt = 0;
331 return err;
332 }
333 }
334
335 return 0;
336}
337
338int thread_stack__flush(struct thread *thread)
339{
340 struct thread_stack *ts = thread->ts;
341 unsigned int pos;
342 int err = 0;
343
344 if (ts) {
345 for (pos = 0; pos < ts->arr_sz; pos++) {
346 int ret = __thread_stack__flush(thread, ts + pos);
347
348 if (ret)
349 err = ret;
350 }
351 }
352
353 return err;
354}
355
356int thread_stack__event(struct thread *thread, int cpu, u32 flags, u64 from_ip,
357 u64 to_ip, u16 insn_len, u64 trace_nr)
358{
359 struct thread_stack *ts = thread__stack(thread, cpu);
360
361 if (!thread)
362 return -EINVAL;
363
364 if (!ts) {
365 ts = thread_stack__new(thread, cpu, NULL);
366 if (!ts) {
367 pr_warning("Out of memory: no thread stack\n");
368 return -ENOMEM;
369 }
370 ts->trace_nr = trace_nr;
371 }
372
373 /*
374 * When the trace is discontinuous, the trace_nr changes. In that case
375 * the stack might be completely invalid. Better to report nothing than
376 * to report something misleading, so flush the stack.
377 */
378 if (trace_nr != ts->trace_nr) {
379 if (ts->trace_nr)
380 __thread_stack__flush(thread, ts);
381 ts->trace_nr = trace_nr;
382 }
383
384 /* Stop here if thread_stack__process() is in use */
385 if (ts->crp)
386 return 0;
387
388 if (flags & PERF_IP_FLAG_CALL) {
389 u64 ret_addr;
390
391 if (!to_ip)
392 return 0;
393 ret_addr = from_ip + insn_len;
394 if (ret_addr == to_ip)
395 return 0; /* Zero-length calls are excluded */
396 return thread_stack__push(ts, ret_addr,
397 flags & PERF_IP_FLAG_TRACE_END);
398 } else if (flags & PERF_IP_FLAG_TRACE_BEGIN) {
399 /*
400 * If the caller did not change the trace number (which would
401 * have flushed the stack) then try to make sense of the stack.
402 * Possibly, tracing began after returning to the current
403 * address, so try to pop that. Also, do not expect a call made
404 * when the trace ended, to return, so pop that.
405 */
406 thread_stack__pop(ts, to_ip);
407 thread_stack__pop_trace_end(ts);
408 } else if ((flags & PERF_IP_FLAG_RETURN) && from_ip) {
409 thread_stack__pop(ts, to_ip);
410 }
411
412 return 0;
413}
414
415void thread_stack__set_trace_nr(struct thread *thread, int cpu, u64 trace_nr)
416{
417 struct thread_stack *ts = thread__stack(thread, cpu);
418
419 if (!ts)
420 return;
421
422 if (trace_nr != ts->trace_nr) {
423 if (ts->trace_nr)
424 __thread_stack__flush(thread, ts);
425 ts->trace_nr = trace_nr;
426 }
427}
428
429static void __thread_stack__free(struct thread *thread, struct thread_stack *ts)
430{
431 __thread_stack__flush(thread, ts);
432 zfree(&ts->stack);
433}
434
435static void thread_stack__reset(struct thread *thread, struct thread_stack *ts)
436{
437 unsigned int arr_sz = ts->arr_sz;
438
439 __thread_stack__free(thread, ts);
440 memset(ts, 0, sizeof(*ts));
441 ts->arr_sz = arr_sz;
442}
443
444void thread_stack__free(struct thread *thread)
445{
446 struct thread_stack *ts = thread->ts;
447 unsigned int pos;
448
449 if (ts) {
450 for (pos = 0; pos < ts->arr_sz; pos++)
451 __thread_stack__free(thread, ts + pos);
452 zfree(&thread->ts);
453 }
454}
455
456static inline u64 callchain_context(u64 ip, u64 kernel_start)
457{
458 return ip < kernel_start ? PERF_CONTEXT_USER : PERF_CONTEXT_KERNEL;
459}
460
461void thread_stack__sample(struct thread *thread, int cpu,
462 struct ip_callchain *chain,
463 size_t sz, u64 ip, u64 kernel_start)
464{
465 struct thread_stack *ts = thread__stack(thread, cpu);
466 u64 context = callchain_context(ip, kernel_start);
467 u64 last_context;
468 size_t i, j;
469
470 if (sz < 2) {
471 chain->nr = 0;
472 return;
473 }
474
475 chain->ips[0] = context;
476 chain->ips[1] = ip;
477
478 if (!ts) {
479 chain->nr = 2;
480 return;
481 }
482
483 last_context = context;
484
485 for (i = 2, j = 1; i < sz && j <= ts->cnt; i++, j++) {
486 ip = ts->stack[ts->cnt - j].ret_addr;
487 context = callchain_context(ip, kernel_start);
488 if (context != last_context) {
489 if (i >= sz - 1)
490 break;
491 chain->ips[i++] = context;
492 last_context = context;
493 }
494 chain->ips[i] = ip;
495 }
496
497 chain->nr = i;
498}
499
500struct call_return_processor *
501call_return_processor__new(int (*process)(struct call_return *cr, u64 *parent_db_id, void *data),
502 void *data)
503{
504 struct call_return_processor *crp;
505
506 crp = zalloc(sizeof(struct call_return_processor));
507 if (!crp)
508 return NULL;
509 crp->cpr = call_path_root__new();
510 if (!crp->cpr)
511 goto out_free;
512 crp->process = process;
513 crp->data = data;
514 return crp;
515
516out_free:
517 free(crp);
518 return NULL;
519}
520
521void call_return_processor__free(struct call_return_processor *crp)
522{
523 if (crp) {
524 call_path_root__free(crp->cpr);
525 free(crp);
526 }
527}
528
529static int thread_stack__push_cp(struct thread_stack *ts, u64 ret_addr,
530 u64 timestamp, u64 ref, struct call_path *cp,
531 bool no_call, bool trace_end)
532{
533 struct thread_stack_entry *tse;
534 int err;
535
536 if (!cp)
537 return -ENOMEM;
538
539 if (ts->cnt == ts->sz) {
540 err = thread_stack__grow(ts);
541 if (err)
542 return err;
543 }
544
545 tse = &ts->stack[ts->cnt++];
546 tse->ret_addr = ret_addr;
547 tse->timestamp = timestamp;
548 tse->ref = ref;
549 tse->branch_count = ts->branch_count;
550 tse->insn_count = ts->insn_count;
551 tse->cyc_count = ts->cyc_count;
552 tse->cp = cp;
553 tse->no_call = no_call;
554 tse->trace_end = trace_end;
555 tse->non_call = false;
556 tse->db_id = 0;
557
558 return 0;
559}
560
561static int thread_stack__pop_cp(struct thread *thread, struct thread_stack *ts,
562 u64 ret_addr, u64 timestamp, u64 ref,
563 struct symbol *sym)
564{
565 int err;
566
567 if (!ts->cnt)
568 return 1;
569
570 if (ts->cnt == 1) {
571 struct thread_stack_entry *tse = &ts->stack[0];
572
573 if (tse->cp->sym == sym)
574 return thread_stack__call_return(thread, ts, --ts->cnt,
575 timestamp, ref, false);
576 }
577
578 if (ts->stack[ts->cnt - 1].ret_addr == ret_addr &&
579 !ts->stack[ts->cnt - 1].non_call) {
580 return thread_stack__call_return(thread, ts, --ts->cnt,
581 timestamp, ref, false);
582 } else {
583 size_t i = ts->cnt - 1;
584
585 while (i--) {
586 if (ts->stack[i].ret_addr != ret_addr ||
587 ts->stack[i].non_call)
588 continue;
589 i += 1;
590 while (ts->cnt > i) {
591 err = thread_stack__call_return(thread, ts,
592 --ts->cnt,
593 timestamp, ref,
594 true);
595 if (err)
596 return err;
597 }
598 return thread_stack__call_return(thread, ts, --ts->cnt,
599 timestamp, ref, false);
600 }
601 }
602
603 return 1;
604}
605
606static int thread_stack__bottom(struct thread_stack *ts,
607 struct perf_sample *sample,
608 struct addr_location *from_al,
609 struct addr_location *to_al, u64 ref)
610{
611 struct call_path_root *cpr = ts->crp->cpr;
612 struct call_path *cp;
613 struct symbol *sym;
614 u64 ip;
615
616 if (sample->ip) {
617 ip = sample->ip;
618 sym = from_al->sym;
619 } else if (sample->addr) {
620 ip = sample->addr;
621 sym = to_al->sym;
622 } else {
623 return 0;
624 }
625
626 cp = call_path__findnew(cpr, &cpr->call_path, sym, ip,
627 ts->kernel_start);
628
629 return thread_stack__push_cp(ts, ip, sample->time, ref, cp,
630 true, false);
631}
632
633static int thread_stack__pop_ks(struct thread *thread, struct thread_stack *ts,
634 struct perf_sample *sample, u64 ref)
635{
636 u64 tm = sample->time;
637 int err;
638
639 /* Return to userspace, so pop all kernel addresses */
640 while (thread_stack__in_kernel(ts)) {
641 err = thread_stack__call_return(thread, ts, --ts->cnt,
642 tm, ref, true);
643 if (err)
644 return err;
645 }
646
647 return 0;
648}
649
650static int thread_stack__no_call_return(struct thread *thread,
651 struct thread_stack *ts,
652 struct perf_sample *sample,
653 struct addr_location *from_al,
654 struct addr_location *to_al, u64 ref)
655{
656 struct call_path_root *cpr = ts->crp->cpr;
657 struct call_path *root = &cpr->call_path;
658 struct symbol *fsym = from_al->sym;
659 struct symbol *tsym = to_al->sym;
660 struct call_path *cp, *parent;
661 u64 ks = ts->kernel_start;
662 u64 addr = sample->addr;
663 u64 tm = sample->time;
664 u64 ip = sample->ip;
665 int err;
666
667 if (ip >= ks && addr < ks) {
668 /* Return to userspace, so pop all kernel addresses */
669 err = thread_stack__pop_ks(thread, ts, sample, ref);
670 if (err)
671 return err;
672
673 /* If the stack is empty, push the userspace address */
674 if (!ts->cnt) {
675 cp = call_path__findnew(cpr, root, tsym, addr, ks);
676 return thread_stack__push_cp(ts, 0, tm, ref, cp, true,
677 false);
678 }
679 } else if (thread_stack__in_kernel(ts) && ip < ks) {
680 /* Return to userspace, so pop all kernel addresses */
681 err = thread_stack__pop_ks(thread, ts, sample, ref);
682 if (err)
683 return err;
684 }
685
686 if (ts->cnt)
687 parent = ts->stack[ts->cnt - 1].cp;
688 else
689 parent = root;
690
691 if (parent->sym == from_al->sym) {
692 /*
693 * At the bottom of the stack, assume the missing 'call' was
694 * before the trace started. So, pop the current symbol and push
695 * the 'to' symbol.
696 */
697 if (ts->cnt == 1) {
698 err = thread_stack__call_return(thread, ts, --ts->cnt,
699 tm, ref, false);
700 if (err)
701 return err;
702 }
703
704 if (!ts->cnt) {
705 cp = call_path__findnew(cpr, root, tsym, addr, ks);
706
707 return thread_stack__push_cp(ts, addr, tm, ref, cp,
708 true, false);
709 }
710
711 /*
712 * Otherwise assume the 'return' is being used as a jump (e.g.
713 * retpoline) and just push the 'to' symbol.
714 */
715 cp = call_path__findnew(cpr, parent, tsym, addr, ks);
716
717 err = thread_stack__push_cp(ts, 0, tm, ref, cp, true, false);
718 if (!err)
719 ts->stack[ts->cnt - 1].non_call = true;
720
721 return err;
722 }
723
724 /*
725 * Assume 'parent' has not yet returned, so push 'to', and then push and
726 * pop 'from'.
727 */
728
729 cp = call_path__findnew(cpr, parent, tsym, addr, ks);
730
731 err = thread_stack__push_cp(ts, addr, tm, ref, cp, true, false);
732 if (err)
733 return err;
734
735 cp = call_path__findnew(cpr, cp, fsym, ip, ks);
736
737 err = thread_stack__push_cp(ts, ip, tm, ref, cp, true, false);
738 if (err)
739 return err;
740
741 return thread_stack__call_return(thread, ts, --ts->cnt, tm, ref, false);
742}
743
744static int thread_stack__trace_begin(struct thread *thread,
745 struct thread_stack *ts, u64 timestamp,
746 u64 ref)
747{
748 struct thread_stack_entry *tse;
749 int err;
750
751 if (!ts->cnt)
752 return 0;
753
754 /* Pop trace end */
755 tse = &ts->stack[ts->cnt - 1];
756 if (tse->trace_end) {
757 err = thread_stack__call_return(thread, ts, --ts->cnt,
758 timestamp, ref, false);
759 if (err)
760 return err;
761 }
762
763 return 0;
764}
765
766static int thread_stack__trace_end(struct thread_stack *ts,
767 struct perf_sample *sample, u64 ref)
768{
769 struct call_path_root *cpr = ts->crp->cpr;
770 struct call_path *cp;
771 u64 ret_addr;
772
773 /* No point having 'trace end' on the bottom of the stack */
774 if (!ts->cnt || (ts->cnt == 1 && ts->stack[0].ref == ref))
775 return 0;
776
777 cp = call_path__findnew(cpr, ts->stack[ts->cnt - 1].cp, NULL, 0,
778 ts->kernel_start);
779
780 ret_addr = sample->ip + sample->insn_len;
781
782 return thread_stack__push_cp(ts, ret_addr, sample->time, ref, cp,
783 false, true);
784}
785
786static bool is_x86_retpoline(const char *name)
787{
788 const char *p = strstr(name, "__x86_indirect_thunk_");
789
790 return p == name || !strcmp(name, "__indirect_thunk_start");
791}
792
793/*
794 * x86 retpoline functions pollute the call graph. This function removes them.
795 * This does not handle function return thunks, nor is there any improvement
796 * for the handling of inline thunks or extern thunks.
797 */
798static int thread_stack__x86_retpoline(struct thread_stack *ts,
799 struct perf_sample *sample,
800 struct addr_location *to_al)
801{
802 struct thread_stack_entry *tse = &ts->stack[ts->cnt - 1];
803 struct call_path_root *cpr = ts->crp->cpr;
804 struct symbol *sym = tse->cp->sym;
805 struct symbol *tsym = to_al->sym;
806 struct call_path *cp;
807
808 if (sym && is_x86_retpoline(sym->name)) {
809 /*
810 * This is a x86 retpoline fn. It pollutes the call graph by
811 * showing up everywhere there is an indirect branch, but does
812 * not itself mean anything. Here the top-of-stack is removed,
813 * by decrementing the stack count, and then further down, the
814 * resulting top-of-stack is replaced with the actual target.
815 * The result is that the retpoline functions will no longer
816 * appear in the call graph. Note this only affects the call
817 * graph, since all the original branches are left unchanged.
818 */
819 ts->cnt -= 1;
820 sym = ts->stack[ts->cnt - 2].cp->sym;
821 if (sym && sym == tsym && to_al->addr != tsym->start) {
822 /*
823 * Target is back to the middle of the symbol we came
824 * from so assume it is an indirect jmp and forget it
825 * altogether.
826 */
827 ts->cnt -= 1;
828 return 0;
829 }
830 } else if (sym && sym == tsym) {
831 /*
832 * Target is back to the symbol we came from so assume it is an
833 * indirect jmp and forget it altogether.
834 */
835 ts->cnt -= 1;
836 return 0;
837 }
838
839 cp = call_path__findnew(cpr, ts->stack[ts->cnt - 2].cp, tsym,
840 sample->addr, ts->kernel_start);
841 if (!cp)
842 return -ENOMEM;
843
844 /* Replace the top-of-stack with the actual target */
845 ts->stack[ts->cnt - 1].cp = cp;
846
847 return 0;
848}
849
850int thread_stack__process(struct thread *thread, struct comm *comm,
851 struct perf_sample *sample,
852 struct addr_location *from_al,
853 struct addr_location *to_al, u64 ref,
854 struct call_return_processor *crp)
855{
856 struct thread_stack *ts = thread__stack(thread, sample->cpu);
857 enum retpoline_state_t rstate;
858 int err = 0;
859
860 if (ts && !ts->crp) {
861 /* Supersede thread_stack__event() */
862 thread_stack__reset(thread, ts);
863 ts = NULL;
864 }
865
866 if (!ts) {
867 ts = thread_stack__new(thread, sample->cpu, crp);
868 if (!ts)
869 return -ENOMEM;
870 ts->comm = comm;
871 }
872
873 rstate = ts->rstate;
874 if (rstate == X86_RETPOLINE_DETECTED)
875 ts->rstate = X86_RETPOLINE_POSSIBLE;
876
877 /* Flush stack on exec */
878 if (ts->comm != comm && thread->pid_ == thread->tid) {
879 err = __thread_stack__flush(thread, ts);
880 if (err)
881 return err;
882 ts->comm = comm;
883 }
884
885 /* If the stack is empty, put the current symbol on the stack */
886 if (!ts->cnt) {
887 err = thread_stack__bottom(ts, sample, from_al, to_al, ref);
888 if (err)
889 return err;
890 }
891
892 ts->branch_count += 1;
893 ts->insn_count += sample->insn_cnt;
894 ts->cyc_count += sample->cyc_cnt;
895 ts->last_time = sample->time;
896
897 if (sample->flags & PERF_IP_FLAG_CALL) {
898 bool trace_end = sample->flags & PERF_IP_FLAG_TRACE_END;
899 struct call_path_root *cpr = ts->crp->cpr;
900 struct call_path *cp;
901 u64 ret_addr;
902
903 if (!sample->ip || !sample->addr)
904 return 0;
905
906 ret_addr = sample->ip + sample->insn_len;
907 if (ret_addr == sample->addr)
908 return 0; /* Zero-length calls are excluded */
909
910 cp = call_path__findnew(cpr, ts->stack[ts->cnt - 1].cp,
911 to_al->sym, sample->addr,
912 ts->kernel_start);
913 err = thread_stack__push_cp(ts, ret_addr, sample->time, ref,
914 cp, false, trace_end);
915
916 /*
917 * A call to the same symbol but not the start of the symbol,
918 * may be the start of a x86 retpoline.
919 */
920 if (!err && rstate == X86_RETPOLINE_POSSIBLE && to_al->sym &&
921 from_al->sym == to_al->sym &&
922 to_al->addr != to_al->sym->start)
923 ts->rstate = X86_RETPOLINE_DETECTED;
924
925 } else if (sample->flags & PERF_IP_FLAG_RETURN) {
926 if (!sample->addr) {
927 u32 return_from_kernel = PERF_IP_FLAG_SYSCALLRET |
928 PERF_IP_FLAG_INTERRUPT;
929
930 if (!(sample->flags & return_from_kernel))
931 return 0;
932
933 /* Pop kernel stack */
934 return thread_stack__pop_ks(thread, ts, sample, ref);
935 }
936
937 if (!sample->ip)
938 return 0;
939
940 /* x86 retpoline 'return' doesn't match the stack */
941 if (rstate == X86_RETPOLINE_DETECTED && ts->cnt > 2 &&
942 ts->stack[ts->cnt - 1].ret_addr != sample->addr)
943 return thread_stack__x86_retpoline(ts, sample, to_al);
944
945 err = thread_stack__pop_cp(thread, ts, sample->addr,
946 sample->time, ref, from_al->sym);
947 if (err) {
948 if (err < 0)
949 return err;
950 err = thread_stack__no_call_return(thread, ts, sample,
951 from_al, to_al, ref);
952 }
953 } else if (sample->flags & PERF_IP_FLAG_TRACE_BEGIN) {
954 err = thread_stack__trace_begin(thread, ts, sample->time, ref);
955 } else if (sample->flags & PERF_IP_FLAG_TRACE_END) {
956 err = thread_stack__trace_end(ts, sample, ref);
957 } else if (sample->flags & PERF_IP_FLAG_BRANCH &&
958 from_al->sym != to_al->sym && to_al->sym &&
959 to_al->addr == to_al->sym->start) {
960 struct call_path_root *cpr = ts->crp->cpr;
961 struct call_path *cp;
962
963 /*
964 * The compiler might optimize a call/ret combination by making
965 * it a jmp. Make that visible by recording on the stack a
966 * branch to the start of a different symbol. Note, that means
967 * when a ret pops the stack, all jmps must be popped off first.
968 */
969 cp = call_path__findnew(cpr, ts->stack[ts->cnt - 1].cp,
970 to_al->sym, sample->addr,
971 ts->kernel_start);
972 err = thread_stack__push_cp(ts, 0, sample->time, ref, cp, false,
973 false);
974 if (!err)
975 ts->stack[ts->cnt - 1].non_call = true;
976 }
977
978 return err;
979}
980
981size_t thread_stack__depth(struct thread *thread, int cpu)
982{
983 struct thread_stack *ts = thread__stack(thread, cpu);
984
985 if (!ts)
986 return 0;
987 return ts->cnt;
988}