Root/tools/perf/builtin-lock.c

1#include "builtin.h"
2#include "perf.h"
3
4#include "util/util.h"
5#include "util/cache.h"
6#include "util/symbol.h"
7#include "util/thread.h"
8#include "util/header.h"
9
10#include "util/parse-options.h"
11#include "util/trace-event.h"
12
13#include "util/debug.h"
14#include "util/session.h"
15#include "util/tool.h"
16
17#include <sys/types.h>
18#include <sys/prctl.h>
19#include <semaphore.h>
20#include <pthread.h>
21#include <math.h>
22#include <limits.h>
23
24#include <linux/list.h>
25#include <linux/hash.h>
26
27static struct perf_session *session;
28
29/* based on kernel/lockdep.c */
30#define LOCKHASH_BITS 12
31#define LOCKHASH_SIZE (1UL << LOCKHASH_BITS)
32
33static struct list_head lockhash_table[LOCKHASH_SIZE];
34
35#define __lockhashfn(key) hash_long((unsigned long)key, LOCKHASH_BITS)
36#define lockhashentry(key) (lockhash_table + __lockhashfn((key)))
37
38struct lock_stat {
39    struct list_head hash_entry;
40    struct rb_node rb; /* used for sorting */
41
42    /*
43     * FIXME: raw_field_value() returns unsigned long long,
44     * so address of lockdep_map should be dealed as 64bit.
45     * Is there more better solution?
46     */
47    void *addr; /* address of lockdep_map, used as ID */
48    char *name; /* for strcpy(), we cannot use const */
49
50    unsigned int nr_acquire;
51    unsigned int nr_acquired;
52    unsigned int nr_contended;
53    unsigned int nr_release;
54
55    unsigned int nr_readlock;
56    unsigned int nr_trylock;
57    /* these times are in nano sec. */
58    u64 wait_time_total;
59    u64 wait_time_min;
60    u64 wait_time_max;
61
62    int discard; /* flag of blacklist */
63};
64
65/*
66 * States of lock_seq_stat
67 *
68 * UNINITIALIZED is required for detecting first event of acquire.
69 * As the nature of lock events, there is no guarantee
70 * that the first event for the locks are acquire,
71 * it can be acquired, contended or release.
72 */
73#define SEQ_STATE_UNINITIALIZED 0 /* initial state */
74#define SEQ_STATE_RELEASED 1
75#define SEQ_STATE_ACQUIRING 2
76#define SEQ_STATE_ACQUIRED 3
77#define SEQ_STATE_READ_ACQUIRED 4
78#define SEQ_STATE_CONTENDED 5
79
80/*
81 * MAX_LOCK_DEPTH
82 * Imported from include/linux/sched.h.
83 * Should this be synchronized?
84 */
85#define MAX_LOCK_DEPTH 48
86
87/*
88 * struct lock_seq_stat:
89 * Place to put on state of one lock sequence
90 * 1) acquire -> acquired -> release
91 * 2) acquire -> contended -> acquired -> release
92 * 3) acquire (with read or try) -> release
93 * 4) Are there other patterns?
94 */
95struct lock_seq_stat {
96    struct list_head list;
97    int state;
98    u64 prev_event_time;
99    void *addr;
100
101    int read_count;
102};
103
104struct thread_stat {
105    struct rb_node rb;
106
107    u32 tid;
108    struct list_head seq_list;
109};
110
111static struct rb_root thread_stats;
112
113static struct thread_stat *thread_stat_find(u32 tid)
114{
115    struct rb_node *node;
116    struct thread_stat *st;
117
118    node = thread_stats.rb_node;
119    while (node) {
120        st = container_of(node, struct thread_stat, rb);
121        if (st->tid == tid)
122            return st;
123        else if (tid < st->tid)
124            node = node->rb_left;
125        else
126            node = node->rb_right;
127    }
128
129    return NULL;
130}
131
132static void thread_stat_insert(struct thread_stat *new)
133{
134    struct rb_node **rb = &thread_stats.rb_node;
135    struct rb_node *parent = NULL;
136    struct thread_stat *p;
137
138    while (*rb) {
139        p = container_of(*rb, struct thread_stat, rb);
140        parent = *rb;
141
142        if (new->tid < p->tid)
143            rb = &(*rb)->rb_left;
144        else if (new->tid > p->tid)
145            rb = &(*rb)->rb_right;
146        else
147            BUG_ON("inserting invalid thread_stat\n");
148    }
149
150    rb_link_node(&new->rb, parent, rb);
151    rb_insert_color(&new->rb, &thread_stats);
152}
153
154static struct thread_stat *thread_stat_findnew_after_first(u32 tid)
155{
156    struct thread_stat *st;
157
158    st = thread_stat_find(tid);
159    if (st)
160        return st;
161
162    st = zalloc(sizeof(struct thread_stat));
163    if (!st)
164        die("memory allocation failed\n");
165
166    st->tid = tid;
167    INIT_LIST_HEAD(&st->seq_list);
168
169    thread_stat_insert(st);
170
171    return st;
172}
173
174static struct thread_stat *thread_stat_findnew_first(u32 tid);
175static struct thread_stat *(*thread_stat_findnew)(u32 tid) =
176    thread_stat_findnew_first;
177
178static struct thread_stat *thread_stat_findnew_first(u32 tid)
179{
180    struct thread_stat *st;
181
182    st = zalloc(sizeof(struct thread_stat));
183    if (!st)
184        die("memory allocation failed\n");
185    st->tid = tid;
186    INIT_LIST_HEAD(&st->seq_list);
187
188    rb_link_node(&st->rb, NULL, &thread_stats.rb_node);
189    rb_insert_color(&st->rb, &thread_stats);
190
191    thread_stat_findnew = thread_stat_findnew_after_first;
192    return st;
193}
194
195/* build simple key function one is bigger than two */
196#define SINGLE_KEY(member) \
197    static int lock_stat_key_ ## member(struct lock_stat *one, \
198                     struct lock_stat *two) \
199    { \
200        return one->member > two->member; \
201    }
202
203SINGLE_KEY(nr_acquired)
204SINGLE_KEY(nr_contended)
205SINGLE_KEY(wait_time_total)
206SINGLE_KEY(wait_time_max)
207
208static int lock_stat_key_wait_time_min(struct lock_stat *one,
209                    struct lock_stat *two)
210{
211    u64 s1 = one->wait_time_min;
212    u64 s2 = two->wait_time_min;
213    if (s1 == ULLONG_MAX)
214        s1 = 0;
215    if (s2 == ULLONG_MAX)
216        s2 = 0;
217    return s1 > s2;
218}
219
220struct lock_key {
221    /*
222     * name: the value for specify by user
223     * this should be simpler than raw name of member
224     * e.g. nr_acquired -> acquired, wait_time_total -> wait_total
225     */
226    const char *name;
227    int (*key)(struct lock_stat*, struct lock_stat*);
228};
229
230static const char *sort_key = "acquired";
231
232static int (*compare)(struct lock_stat *, struct lock_stat *);
233
234static struct rb_root result; /* place to store sorted data */
235
236#define DEF_KEY_LOCK(name, fn_suffix) \
237    { #name, lock_stat_key_ ## fn_suffix }
238struct lock_key keys[] = {
239    DEF_KEY_LOCK(acquired, nr_acquired),
240    DEF_KEY_LOCK(contended, nr_contended),
241    DEF_KEY_LOCK(wait_total, wait_time_total),
242    DEF_KEY_LOCK(wait_min, wait_time_min),
243    DEF_KEY_LOCK(wait_max, wait_time_max),
244
245    /* extra comparisons much complicated should be here */
246
247    { NULL, NULL }
248};
249
250static void select_key(void)
251{
252    int i;
253
254    for (i = 0; keys[i].name; i++) {
255        if (!strcmp(keys[i].name, sort_key)) {
256            compare = keys[i].key;
257            return;
258        }
259    }
260
261    die("Unknown compare key:%s\n", sort_key);
262}
263
264static void insert_to_result(struct lock_stat *st,
265                 int (*bigger)(struct lock_stat *, struct lock_stat *))
266{
267    struct rb_node **rb = &result.rb_node;
268    struct rb_node *parent = NULL;
269    struct lock_stat *p;
270
271    while (*rb) {
272        p = container_of(*rb, struct lock_stat, rb);
273        parent = *rb;
274
275        if (bigger(st, p))
276            rb = &(*rb)->rb_left;
277        else
278            rb = &(*rb)->rb_right;
279    }
280
281    rb_link_node(&st->rb, parent, rb);
282    rb_insert_color(&st->rb, &result);
283}
284
285/* returns left most element of result, and erase it */
286static struct lock_stat *pop_from_result(void)
287{
288    struct rb_node *node = result.rb_node;
289
290    if (!node)
291        return NULL;
292
293    while (node->rb_left)
294        node = node->rb_left;
295
296    rb_erase(node, &result);
297    return container_of(node, struct lock_stat, rb);
298}
299
300static struct lock_stat *lock_stat_findnew(void *addr, const char *name)
301{
302    struct list_head *entry = lockhashentry(addr);
303    struct lock_stat *ret, *new;
304
305    list_for_each_entry(ret, entry, hash_entry) {
306        if (ret->addr == addr)
307            return ret;
308    }
309
310    new = zalloc(sizeof(struct lock_stat));
311    if (!new)
312        goto alloc_failed;
313
314    new->addr = addr;
315    new->name = zalloc(sizeof(char) * strlen(name) + 1);
316    if (!new->name)
317        goto alloc_failed;
318    strcpy(new->name, name);
319
320    new->wait_time_min = ULLONG_MAX;
321
322    list_add(&new->hash_entry, entry);
323    return new;
324
325alloc_failed:
326    die("memory allocation failed\n");
327}
328
329static const char *input_name;
330
331struct raw_event_sample {
332    u32 size;
333    char data[0];
334};
335
336struct trace_acquire_event {
337    void *addr;
338    const char *name;
339    int flag;
340};
341
342struct trace_acquired_event {
343    void *addr;
344    const char *name;
345};
346
347struct trace_contended_event {
348    void *addr;
349    const char *name;
350};
351
352struct trace_release_event {
353    void *addr;
354    const char *name;
355};
356
357struct trace_lock_handler {
358    void (*acquire_event)(struct trace_acquire_event *,
359                  struct event_format *,
360                  int cpu,
361                  u64 timestamp,
362                  struct thread *thread);
363
364    void (*acquired_event)(struct trace_acquired_event *,
365                   struct event_format *,
366                   int cpu,
367                   u64 timestamp,
368                   struct thread *thread);
369
370    void (*contended_event)(struct trace_contended_event *,
371                struct event_format *,
372                int cpu,
373                u64 timestamp,
374                struct thread *thread);
375
376    void (*release_event)(struct trace_release_event *,
377                  struct event_format *,
378                  int cpu,
379                  u64 timestamp,
380                  struct thread *thread);
381};
382
383static struct lock_seq_stat *get_seq(struct thread_stat *ts, void *addr)
384{
385    struct lock_seq_stat *seq;
386
387    list_for_each_entry(seq, &ts->seq_list, list) {
388        if (seq->addr == addr)
389            return seq;
390    }
391
392    seq = zalloc(sizeof(struct lock_seq_stat));
393    if (!seq)
394        die("Not enough memory\n");
395    seq->state = SEQ_STATE_UNINITIALIZED;
396    seq->addr = addr;
397
398    list_add(&seq->list, &ts->seq_list);
399    return seq;
400}
401
402enum broken_state {
403    BROKEN_ACQUIRE,
404    BROKEN_ACQUIRED,
405    BROKEN_CONTENDED,
406    BROKEN_RELEASE,
407    BROKEN_MAX,
408};
409
410static int bad_hist[BROKEN_MAX];
411
412enum acquire_flags {
413    TRY_LOCK = 1,
414    READ_LOCK = 2,
415};
416
417static void
418report_lock_acquire_event(struct trace_acquire_event *acquire_event,
419            struct event_format *__event __used,
420            int cpu __used,
421            u64 timestamp __used,
422            struct thread *thread __used)
423{
424    struct lock_stat *ls;
425    struct thread_stat *ts;
426    struct lock_seq_stat *seq;
427
428    ls = lock_stat_findnew(acquire_event->addr, acquire_event->name);
429    if (ls->discard)
430        return;
431
432    ts = thread_stat_findnew(thread->pid);
433    seq = get_seq(ts, acquire_event->addr);
434
435    switch (seq->state) {
436    case SEQ_STATE_UNINITIALIZED:
437    case SEQ_STATE_RELEASED:
438        if (!acquire_event->flag) {
439            seq->state = SEQ_STATE_ACQUIRING;
440        } else {
441            if (acquire_event->flag & TRY_LOCK)
442                ls->nr_trylock++;
443            if (acquire_event->flag & READ_LOCK)
444                ls->nr_readlock++;
445            seq->state = SEQ_STATE_READ_ACQUIRED;
446            seq->read_count = 1;
447            ls->nr_acquired++;
448        }
449        break;
450    case SEQ_STATE_READ_ACQUIRED:
451        if (acquire_event->flag & READ_LOCK) {
452            seq->read_count++;
453            ls->nr_acquired++;
454            goto end;
455        } else {
456            goto broken;
457        }
458        break;
459    case SEQ_STATE_ACQUIRED:
460    case SEQ_STATE_ACQUIRING:
461    case SEQ_STATE_CONTENDED:
462broken:
463        /* broken lock sequence, discard it */
464        ls->discard = 1;
465        bad_hist[BROKEN_ACQUIRE]++;
466        list_del(&seq->list);
467        free(seq);
468        goto end;
469        break;
470    default:
471        BUG_ON("Unknown state of lock sequence found!\n");
472        break;
473    }
474
475    ls->nr_acquire++;
476    seq->prev_event_time = timestamp;
477end:
478    return;
479}
480
481static void
482report_lock_acquired_event(struct trace_acquired_event *acquired_event,
483             struct event_format *__event __used,
484             int cpu __used,
485             u64 timestamp __used,
486             struct thread *thread __used)
487{
488    struct lock_stat *ls;
489    struct thread_stat *ts;
490    struct lock_seq_stat *seq;
491    u64 contended_term;
492
493    ls = lock_stat_findnew(acquired_event->addr, acquired_event->name);
494    if (ls->discard)
495        return;
496
497    ts = thread_stat_findnew(thread->pid);
498    seq = get_seq(ts, acquired_event->addr);
499
500    switch (seq->state) {
501    case SEQ_STATE_UNINITIALIZED:
502        /* orphan event, do nothing */
503        return;
504    case SEQ_STATE_ACQUIRING:
505        break;
506    case SEQ_STATE_CONTENDED:
507        contended_term = timestamp - seq->prev_event_time;
508        ls->wait_time_total += contended_term;
509        if (contended_term < ls->wait_time_min)
510            ls->wait_time_min = contended_term;
511        if (ls->wait_time_max < contended_term)
512            ls->wait_time_max = contended_term;
513        break;
514    case SEQ_STATE_RELEASED:
515    case SEQ_STATE_ACQUIRED:
516    case SEQ_STATE_READ_ACQUIRED:
517        /* broken lock sequence, discard it */
518        ls->discard = 1;
519        bad_hist[BROKEN_ACQUIRED]++;
520        list_del(&seq->list);
521        free(seq);
522        goto end;
523        break;
524
525    default:
526        BUG_ON("Unknown state of lock sequence found!\n");
527        break;
528    }
529
530    seq->state = SEQ_STATE_ACQUIRED;
531    ls->nr_acquired++;
532    seq->prev_event_time = timestamp;
533end:
534    return;
535}
536
537static void
538report_lock_contended_event(struct trace_contended_event *contended_event,
539              struct event_format *__event __used,
540              int cpu __used,
541              u64 timestamp __used,
542              struct thread *thread __used)
543{
544    struct lock_stat *ls;
545    struct thread_stat *ts;
546    struct lock_seq_stat *seq;
547
548    ls = lock_stat_findnew(contended_event->addr, contended_event->name);
549    if (ls->discard)
550        return;
551
552    ts = thread_stat_findnew(thread->pid);
553    seq = get_seq(ts, contended_event->addr);
554
555    switch (seq->state) {
556    case SEQ_STATE_UNINITIALIZED:
557        /* orphan event, do nothing */
558        return;
559    case SEQ_STATE_ACQUIRING:
560        break;
561    case SEQ_STATE_RELEASED:
562    case SEQ_STATE_ACQUIRED:
563    case SEQ_STATE_READ_ACQUIRED:
564    case SEQ_STATE_CONTENDED:
565        /* broken lock sequence, discard it */
566        ls->discard = 1;
567        bad_hist[BROKEN_CONTENDED]++;
568        list_del(&seq->list);
569        free(seq);
570        goto end;
571        break;
572    default:
573        BUG_ON("Unknown state of lock sequence found!\n");
574        break;
575    }
576
577    seq->state = SEQ_STATE_CONTENDED;
578    ls->nr_contended++;
579    seq->prev_event_time = timestamp;
580end:
581    return;
582}
583
584static void
585report_lock_release_event(struct trace_release_event *release_event,
586            struct event_format *__event __used,
587            int cpu __used,
588            u64 timestamp __used,
589            struct thread *thread __used)
590{
591    struct lock_stat *ls;
592    struct thread_stat *ts;
593    struct lock_seq_stat *seq;
594
595    ls = lock_stat_findnew(release_event->addr, release_event->name);
596    if (ls->discard)
597        return;
598
599    ts = thread_stat_findnew(thread->pid);
600    seq = get_seq(ts, release_event->addr);
601
602    switch (seq->state) {
603    case SEQ_STATE_UNINITIALIZED:
604        goto end;
605        break;
606    case SEQ_STATE_ACQUIRED:
607        break;
608    case SEQ_STATE_READ_ACQUIRED:
609        seq->read_count--;
610        BUG_ON(seq->read_count < 0);
611        if (!seq->read_count) {
612            ls->nr_release++;
613            goto end;
614        }
615        break;
616    case SEQ_STATE_ACQUIRING:
617    case SEQ_STATE_CONTENDED:
618    case SEQ_STATE_RELEASED:
619        /* broken lock sequence, discard it */
620        ls->discard = 1;
621        bad_hist[BROKEN_RELEASE]++;
622        goto free_seq;
623        break;
624    default:
625        BUG_ON("Unknown state of lock sequence found!\n");
626        break;
627    }
628
629    ls->nr_release++;
630free_seq:
631    list_del(&seq->list);
632    free(seq);
633end:
634    return;
635}
636
637/* lock oriented handlers */
638/* TODO: handlers for CPU oriented, thread oriented */
639static struct trace_lock_handler report_lock_ops = {
640    .acquire_event = report_lock_acquire_event,
641    .acquired_event = report_lock_acquired_event,
642    .contended_event = report_lock_contended_event,
643    .release_event = report_lock_release_event,
644};
645
646static struct trace_lock_handler *trace_handler;
647
648static void
649process_lock_acquire_event(void *data,
650               struct event_format *event __used,
651               int cpu __used,
652               u64 timestamp __used,
653               struct thread *thread __used)
654{
655    struct trace_acquire_event acquire_event;
656    u64 tmp; /* this is required for casting... */
657
658    tmp = raw_field_value(event, "lockdep_addr", data);
659    memcpy(&acquire_event.addr, &tmp, sizeof(void *));
660    acquire_event.name = (char *)raw_field_ptr(event, "name", data);
661    acquire_event.flag = (int)raw_field_value(event, "flag", data);
662
663    if (trace_handler->acquire_event)
664        trace_handler->acquire_event(&acquire_event, event, cpu, timestamp, thread);
665}
666
667static void
668process_lock_acquired_event(void *data,
669                struct event_format *event __used,
670                int cpu __used,
671                u64 timestamp __used,
672                struct thread *thread __used)
673{
674    struct trace_acquired_event acquired_event;
675    u64 tmp; /* this is required for casting... */
676
677    tmp = raw_field_value(event, "lockdep_addr", data);
678    memcpy(&acquired_event.addr, &tmp, sizeof(void *));
679    acquired_event.name = (char *)raw_field_ptr(event, "name", data);
680
681    if (trace_handler->acquire_event)
682        trace_handler->acquired_event(&acquired_event, event, cpu, timestamp, thread);
683}
684
685static void
686process_lock_contended_event(void *data,
687                 struct event_format *event __used,
688                 int cpu __used,
689                 u64 timestamp __used,
690                 struct thread *thread __used)
691{
692    struct trace_contended_event contended_event;
693    u64 tmp; /* this is required for casting... */
694
695    tmp = raw_field_value(event, "lockdep_addr", data);
696    memcpy(&contended_event.addr, &tmp, sizeof(void *));
697    contended_event.name = (char *)raw_field_ptr(event, "name", data);
698
699    if (trace_handler->acquire_event)
700        trace_handler->contended_event(&contended_event, event, cpu, timestamp, thread);
701}
702
703static void
704process_lock_release_event(void *data,
705               struct event_format *event __used,
706               int cpu __used,
707               u64 timestamp __used,
708               struct thread *thread __used)
709{
710    struct trace_release_event release_event;
711    u64 tmp; /* this is required for casting... */
712
713    tmp = raw_field_value(event, "lockdep_addr", data);
714    memcpy(&release_event.addr, &tmp, sizeof(void *));
715    release_event.name = (char *)raw_field_ptr(event, "name", data);
716
717    if (trace_handler->acquire_event)
718        trace_handler->release_event(&release_event, event, cpu, timestamp, thread);
719}
720
721static void
722process_raw_event(void *data, int cpu, u64 timestamp, struct thread *thread)
723{
724    struct event_format *event;
725    int type;
726
727    type = trace_parse_common_type(data);
728    event = trace_find_event(type);
729
730    if (!strcmp(event->name, "lock_acquire"))
731        process_lock_acquire_event(data, event, cpu, timestamp, thread);
732    if (!strcmp(event->name, "lock_acquired"))
733        process_lock_acquired_event(data, event, cpu, timestamp, thread);
734    if (!strcmp(event->name, "lock_contended"))
735        process_lock_contended_event(data, event, cpu, timestamp, thread);
736    if (!strcmp(event->name, "lock_release"))
737        process_lock_release_event(data, event, cpu, timestamp, thread);
738}
739
740static void print_bad_events(int bad, int total)
741{
742    /* Output for debug, this have to be removed */
743    int i;
744    const char *name[4] =
745        { "acquire", "acquired", "contended", "release" };
746
747    pr_info("\n=== output for debug===\n\n");
748    pr_info("bad: %d, total: %d\n", bad, total);
749    pr_info("bad rate: %f %%\n", (double)bad / (double)total * 100);
750    pr_info("histogram of events caused bad sequence\n");
751    for (i = 0; i < BROKEN_MAX; i++)
752        pr_info(" %10s: %d\n", name[i], bad_hist[i]);
753}
754
755/* TODO: various way to print, coloring, nano or milli sec */
756static void print_result(void)
757{
758    struct lock_stat *st;
759    char cut_name[20];
760    int bad, total;
761
762    pr_info("%20s ", "Name");
763    pr_info("%10s ", "acquired");
764    pr_info("%10s ", "contended");
765
766    pr_info("%15s ", "total wait (ns)");
767    pr_info("%15s ", "max wait (ns)");
768    pr_info("%15s ", "min wait (ns)");
769
770    pr_info("\n\n");
771
772    bad = total = 0;
773    while ((st = pop_from_result())) {
774        total++;
775        if (st->discard) {
776            bad++;
777            continue;
778        }
779        bzero(cut_name, 20);
780
781        if (strlen(st->name) < 16) {
782            /* output raw name */
783            pr_info("%20s ", st->name);
784        } else {
785            strncpy(cut_name, st->name, 16);
786            cut_name[16] = '.';
787            cut_name[17] = '.';
788            cut_name[18] = '.';
789            cut_name[19] = '\0';
790            /* cut off name for saving output style */
791            pr_info("%20s ", cut_name);
792        }
793
794        pr_info("%10u ", st->nr_acquired);
795        pr_info("%10u ", st->nr_contended);
796
797        pr_info("%15" PRIu64 " ", st->wait_time_total);
798        pr_info("%15" PRIu64 " ", st->wait_time_max);
799        pr_info("%15" PRIu64 " ", st->wait_time_min == ULLONG_MAX ?
800               0 : st->wait_time_min);
801        pr_info("\n");
802    }
803
804    print_bad_events(bad, total);
805}
806
807static bool info_threads, info_map;
808
809static void dump_threads(void)
810{
811    struct thread_stat *st;
812    struct rb_node *node;
813    struct thread *t;
814
815    pr_info("%10s: comm\n", "Thread ID");
816
817    node = rb_first(&thread_stats);
818    while (node) {
819        st = container_of(node, struct thread_stat, rb);
820        t = perf_session__findnew(session, st->tid);
821        pr_info("%10d: %s\n", st->tid, t->comm);
822        node = rb_next(node);
823    };
824}
825
826static void dump_map(void)
827{
828    unsigned int i;
829    struct lock_stat *st;
830
831    pr_info("Address of instance: name of class\n");
832    for (i = 0; i < LOCKHASH_SIZE; i++) {
833        list_for_each_entry(st, &lockhash_table[i], hash_entry) {
834            pr_info(" %p: %s\n", st->addr, st->name);
835        }
836    }
837}
838
839static void dump_info(void)
840{
841    if (info_threads)
842        dump_threads();
843    else if (info_map)
844        dump_map();
845    else
846        die("Unknown type of information\n");
847}
848
849static int process_sample_event(struct perf_tool *tool __used,
850                union perf_event *event,
851                struct perf_sample *sample,
852                struct perf_evsel *evsel __used,
853                struct machine *machine)
854{
855    struct thread *thread = machine__findnew_thread(machine, sample->tid);
856
857    if (thread == NULL) {
858        pr_debug("problem processing %d event, skipping it.\n",
859            event->header.type);
860        return -1;
861    }
862
863    process_raw_event(sample->raw_data, sample->cpu, sample->time, thread);
864
865    return 0;
866}
867
868static struct perf_tool eops = {
869    .sample = process_sample_event,
870    .comm = perf_event__process_comm,
871    .ordered_samples = true,
872};
873
874static int read_events(void)
875{
876    session = perf_session__new(input_name, O_RDONLY, 0, false, &eops);
877    if (!session)
878        die("Initializing perf session failed\n");
879
880    return perf_session__process_events(session, &eops);
881}
882
883static void sort_result(void)
884{
885    unsigned int i;
886    struct lock_stat *st;
887
888    for (i = 0; i < LOCKHASH_SIZE; i++) {
889        list_for_each_entry(st, &lockhash_table[i], hash_entry) {
890            insert_to_result(st, compare);
891        }
892    }
893}
894
895static void __cmd_report(void)
896{
897    setup_pager();
898    select_key();
899    read_events();
900    sort_result();
901    print_result();
902}
903
904static const char * const report_usage[] = {
905    "perf lock report [<options>]",
906    NULL
907};
908
909static const struct option report_options[] = {
910    OPT_STRING('k', "key", &sort_key, "acquired",
911            "key for sorting (acquired / contended / wait_total / wait_max / wait_min)"),
912    /* TODO: type */
913    OPT_END()
914};
915
916static const char * const info_usage[] = {
917    "perf lock info [<options>]",
918    NULL
919};
920
921static const struct option info_options[] = {
922    OPT_BOOLEAN('t', "threads", &info_threads,
923            "dump thread list in perf.data"),
924    OPT_BOOLEAN('m', "map", &info_map,
925            "map of lock instances (address:name table)"),
926    OPT_END()
927};
928
929static const char * const lock_usage[] = {
930    "perf lock [<options>] {record|report|script|info}",
931    NULL
932};
933
934static const struct option lock_options[] = {
935    OPT_STRING('i', "input", &input_name, "file", "input file name"),
936    OPT_INCR('v', "verbose", &verbose, "be more verbose (show symbol address, etc)"),
937    OPT_BOOLEAN('D', "dump-raw-trace", &dump_trace, "dump raw trace in ASCII"),
938    OPT_END()
939};
940
941static const char *record_args[] = {
942    "record",
943    "-R",
944    "-f",
945    "-m", "1024",
946    "-c", "1",
947    "-e", "lock:lock_acquire",
948    "-e", "lock:lock_acquired",
949    "-e", "lock:lock_contended",
950    "-e", "lock:lock_release",
951};
952
953static int __cmd_record(int argc, const char **argv)
954{
955    unsigned int rec_argc, i, j;
956    const char **rec_argv;
957
958    rec_argc = ARRAY_SIZE(record_args) + argc - 1;
959    rec_argv = calloc(rec_argc + 1, sizeof(char *));
960
961    if (rec_argv == NULL)
962        return -ENOMEM;
963
964    for (i = 0; i < ARRAY_SIZE(record_args); i++)
965        rec_argv[i] = strdup(record_args[i]);
966
967    for (j = 1; j < (unsigned int)argc; j++, i++)
968        rec_argv[i] = argv[j];
969
970    BUG_ON(i != rec_argc);
971
972    return cmd_record(i, rec_argv, NULL);
973}
974
975int cmd_lock(int argc, const char **argv, const char *prefix __used)
976{
977    unsigned int i;
978
979    symbol__init();
980    for (i = 0; i < LOCKHASH_SIZE; i++)
981        INIT_LIST_HEAD(lockhash_table + i);
982
983    argc = parse_options(argc, argv, lock_options, lock_usage,
984                 PARSE_OPT_STOP_AT_NON_OPTION);
985    if (!argc)
986        usage_with_options(lock_usage, lock_options);
987
988    if (!strncmp(argv[0], "rec", 3)) {
989        return __cmd_record(argc, argv);
990    } else if (!strncmp(argv[0], "report", 6)) {
991        trace_handler = &report_lock_ops;
992        if (argc) {
993            argc = parse_options(argc, argv,
994                         report_options, report_usage, 0);
995            if (argc)
996                usage_with_options(report_usage, report_options);
997        }
998        __cmd_report();
999    } else if (!strcmp(argv[0], "script")) {
1000        /* Aliased to 'perf script' */
1001        return cmd_script(argc, argv, prefix);
1002    } else if (!strcmp(argv[0], "info")) {
1003        if (argc) {
1004            argc = parse_options(argc, argv,
1005                         info_options, info_usage, 0);
1006            if (argc)
1007                usage_with_options(info_usage, info_options);
1008        }
1009        /* recycling report_lock_ops */
1010        trace_handler = &report_lock_ops;
1011        setup_pager();
1012        read_events();
1013        dump_info();
1014    } else {
1015        usage_with_options(lock_usage, lock_options);
1016    }
1017
1018    return 0;
1019}
1020

Archive Download this file



interactive