linux/tools/perf/tests/hists_link.c

351 lines
8.4 KiB
C
Raw Normal View History

#include "perf.h"
#include "tests.h"
#include "debug.h"
#include "symbol.h"
#include "sort.h"
#include "evsel.h"
#include "evlist.h"
#include "machine.h"
#include "thread.h"
#include "parse-events.h"
#include "hists_common.h"
struct sample {
u32 pid;
u64 ip;
struct thread *thread;
struct map *map;
struct symbol *sym;
};
/* For the numbers, see hists_common.c */
static struct sample fake_common_samples[] = {
/* perf [kernel] schedule() */
{ .pid = FAKE_PID_PERF1, .ip = FAKE_IP_KERNEL_SCHEDULE, },
/* perf [perf] main() */
{ .pid = FAKE_PID_PERF2, .ip = FAKE_IP_PERF_MAIN, },
/* perf [perf] cmd_record() */
{ .pid = FAKE_PID_PERF2, .ip = FAKE_IP_PERF_CMD_RECORD, },
/* bash [bash] xmalloc() */
{ .pid = FAKE_PID_BASH, .ip = FAKE_IP_BASH_XMALLOC, },
/* bash [libc] malloc() */
{ .pid = FAKE_PID_BASH, .ip = FAKE_IP_LIBC_MALLOC, },
};
static struct sample fake_samples[][5] = {
{
/* perf [perf] run_command() */
{ .pid = FAKE_PID_PERF1, .ip = FAKE_IP_PERF_RUN_COMMAND, },
/* perf [libc] malloc() */
{ .pid = FAKE_PID_PERF1, .ip = FAKE_IP_LIBC_MALLOC, },
/* perf [kernel] page_fault() */
{ .pid = FAKE_PID_PERF1, .ip = FAKE_IP_KERNEL_PAGE_FAULT, },
/* perf [kernel] sys_perf_event_open() */
{ .pid = FAKE_PID_PERF2, .ip = FAKE_IP_KERNEL_SYS_PERF_EVENT_OPEN, },
/* bash [libc] free() */
{ .pid = FAKE_PID_BASH, .ip = FAKE_IP_LIBC_FREE, },
},
{
/* perf [libc] free() */
{ .pid = FAKE_PID_PERF2, .ip = FAKE_IP_LIBC_FREE, },
/* bash [libc] malloc() */
{ .pid = FAKE_PID_BASH, .ip = FAKE_IP_LIBC_MALLOC, }, /* will be merged */
/* bash [bash] xfee() */
{ .pid = FAKE_PID_BASH, .ip = FAKE_IP_BASH_XFREE, },
/* bash [libc] realloc() */
{ .pid = FAKE_PID_BASH, .ip = FAKE_IP_LIBC_REALLOC, },
/* bash [kernel] page_fault() */
{ .pid = FAKE_PID_BASH, .ip = FAKE_IP_KERNEL_PAGE_FAULT, },
},
};
static int add_hist_entries(struct perf_evlist *evlist, struct machine *machine)
{
struct perf_evsel *evsel;
struct addr_location al;
struct hist_entry *he;
struct perf_sample sample = { .period = 1, .weight = 1, };
size_t i = 0, k;
/*
* each evsel will have 10 samples - 5 common and 5 distinct.
* However the second evsel also has a collapsed entry for
* "bash [libc] malloc" so total 9 entries will be in the tree.
*/
evlist__for_each(evlist, evsel) {
struct hists *hists = evsel__hists(evsel);
for (k = 0; k < ARRAY_SIZE(fake_common_samples); k++) {
const union perf_event event = {
.header = {
.misc = PERF_RECORD_MISC_USER,
},
};
sample.pid = fake_common_samples[k].pid;
sample.tid = fake_common_samples[k].pid;
sample.ip = fake_common_samples[k].ip;
if (perf_event__preprocess_sample(&event, machine, &al,
&sample) < 0)
goto out;
he = __hists__add_entry(hists, &al, NULL,
NULL, NULL, &sample, true);
perf machine: Protect the machine->threads with a rwlock In addition to using refcounts for the struct thread lifetime management, we need to protect access to machine->threads from concurrent access. That happens in 'perf top', where a thread processes events, inserting and deleting entries from that rb_tree while another thread decays hist_entries, that end up dropping references and ultimately deleting threads from the rb_tree and releasing its resources when no further hist_entry (or other data structures, like in 'perf sched') references it. So the rule is the same for refcounts + protected trees in the kernel, get the tree lock, find object, bump the refcount, drop the tree lock, return, use object, drop the refcount if no more use of it is needed, keep it if storing it in some other data structure, drop when releasing that data structure. I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and "perf_event__preprocess_sample(&al)" with "addr_location__put(&al)". The addr_location__put() one is because as we return references to several data structures, we may end up adding more reference counting for the other data structures and then we'll drop it at addr_location__put() time. Acked-by: David Ahern <dsahern@gmail.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Borislav Petkov <bp@suse.de> Cc: Don Zickus <dzickus@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
if (he == NULL) {
addr_location__put(&al);
goto out;
perf machine: Protect the machine->threads with a rwlock In addition to using refcounts for the struct thread lifetime management, we need to protect access to machine->threads from concurrent access. That happens in 'perf top', where a thread processes events, inserting and deleting entries from that rb_tree while another thread decays hist_entries, that end up dropping references and ultimately deleting threads from the rb_tree and releasing its resources when no further hist_entry (or other data structures, like in 'perf sched') references it. So the rule is the same for refcounts + protected trees in the kernel, get the tree lock, find object, bump the refcount, drop the tree lock, return, use object, drop the refcount if no more use of it is needed, keep it if storing it in some other data structure, drop when releasing that data structure. I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and "perf_event__preprocess_sample(&al)" with "addr_location__put(&al)". The addr_location__put() one is because as we return references to several data structures, we may end up adding more reference counting for the other data structures and then we'll drop it at addr_location__put() time. Acked-by: David Ahern <dsahern@gmail.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Borislav Petkov <bp@suse.de> Cc: Don Zickus <dzickus@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
}
fake_common_samples[k].thread = al.thread;
fake_common_samples[k].map = al.map;
fake_common_samples[k].sym = al.sym;
}
for (k = 0; k < ARRAY_SIZE(fake_samples[i]); k++) {
const union perf_event event = {
.header = {
.misc = PERF_RECORD_MISC_USER,
},
};
sample.pid = fake_samples[i][k].pid;
sample.tid = fake_samples[i][k].pid;
sample.ip = fake_samples[i][k].ip;
if (perf_event__preprocess_sample(&event, machine, &al,
&sample) < 0)
goto out;
he = __hists__add_entry(hists, &al, NULL,
NULL, NULL, &sample, true);
perf machine: Protect the machine->threads with a rwlock In addition to using refcounts for the struct thread lifetime management, we need to protect access to machine->threads from concurrent access. That happens in 'perf top', where a thread processes events, inserting and deleting entries from that rb_tree while another thread decays hist_entries, that end up dropping references and ultimately deleting threads from the rb_tree and releasing its resources when no further hist_entry (or other data structures, like in 'perf sched') references it. So the rule is the same for refcounts + protected trees in the kernel, get the tree lock, find object, bump the refcount, drop the tree lock, return, use object, drop the refcount if no more use of it is needed, keep it if storing it in some other data structure, drop when releasing that data structure. I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and "perf_event__preprocess_sample(&al)" with "addr_location__put(&al)". The addr_location__put() one is because as we return references to several data structures, we may end up adding more reference counting for the other data structures and then we'll drop it at addr_location__put() time. Acked-by: David Ahern <dsahern@gmail.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Borislav Petkov <bp@suse.de> Cc: Don Zickus <dzickus@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
if (he == NULL) {
addr_location__put(&al);
goto out;
perf machine: Protect the machine->threads with a rwlock In addition to using refcounts for the struct thread lifetime management, we need to protect access to machine->threads from concurrent access. That happens in 'perf top', where a thread processes events, inserting and deleting entries from that rb_tree while another thread decays hist_entries, that end up dropping references and ultimately deleting threads from the rb_tree and releasing its resources when no further hist_entry (or other data structures, like in 'perf sched') references it. So the rule is the same for refcounts + protected trees in the kernel, get the tree lock, find object, bump the refcount, drop the tree lock, return, use object, drop the refcount if no more use of it is needed, keep it if storing it in some other data structure, drop when releasing that data structure. I.e. pair "t = machine__find(new)_thread()" with a "thread__put(t)", and "perf_event__preprocess_sample(&al)" with "addr_location__put(&al)". The addr_location__put() one is because as we return references to several data structures, we may end up adding more reference counting for the other data structures and then we'll drop it at addr_location__put() time. Acked-by: David Ahern <dsahern@gmail.com> Cc: Adrian Hunter <adrian.hunter@intel.com> Cc: Borislav Petkov <bp@suse.de> Cc: Don Zickus <dzickus@redhat.com> Cc: Frederic Weisbecker <fweisbec@gmail.com> Cc: Jiri Olsa <jolsa@redhat.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Stephane Eranian <eranian@google.com> Link: http://lkml.kernel.org/n/tip-bs9rt4n0jw3hi9f3zxyy3xln@git.kernel.org Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-07 07:43:22 +08:00
}
fake_samples[i][k].thread = al.thread;
fake_samples[i][k].map = al.map;
fake_samples[i][k].sym = al.sym;
}
i++;
}
return 0;
out:
pr_debug("Not enough memory for adding a hist entry\n");
return -1;
}
static int find_sample(struct sample *samples, size_t nr_samples,
struct thread *t, struct map *m, struct symbol *s)
{
while (nr_samples--) {
if (samples->thread == t && samples->map == m &&
samples->sym == s)
return 1;
samples++;
}
return 0;
}
static int __validate_match(struct hists *hists)
{
size_t count = 0;
struct rb_root *root;
struct rb_node *node;
/*
* Only entries from fake_common_samples should have a pair.
*/
if (sort__need_collapse)
root = &hists->entries_collapsed;
else
root = hists->entries_in;
node = rb_first(root);
while (node) {
struct hist_entry *he;
he = rb_entry(node, struct hist_entry, rb_node_in);
if (hist_entry__has_pairs(he)) {
if (find_sample(fake_common_samples,
ARRAY_SIZE(fake_common_samples),
he->thread, he->ms.map, he->ms.sym)) {
count++;
} else {
pr_debug("Can't find the matched entry\n");
return -1;
}
}
node = rb_next(node);
}
if (count != ARRAY_SIZE(fake_common_samples)) {
pr_debug("Invalid count for matched entries: %zd of %zd\n",
count, ARRAY_SIZE(fake_common_samples));
return -1;
}
return 0;
}
static int validate_match(struct hists *leader, struct hists *other)
{
return __validate_match(leader) || __validate_match(other);
}
static int __validate_link(struct hists *hists, int idx)
{
size_t count = 0;
size_t count_pair = 0;
size_t count_dummy = 0;
struct rb_root *root;
struct rb_node *node;
/*
* Leader hists (idx = 0) will have dummy entries from other,
* and some entries will have no pair. However every entry
* in other hists should have (dummy) pair.
*/
if (sort__need_collapse)
root = &hists->entries_collapsed;
else
root = hists->entries_in;
node = rb_first(root);
while (node) {
struct hist_entry *he;
he = rb_entry(node, struct hist_entry, rb_node_in);
if (hist_entry__has_pairs(he)) {
if (!find_sample(fake_common_samples,
ARRAY_SIZE(fake_common_samples),
he->thread, he->ms.map, he->ms.sym) &&
!find_sample(fake_samples[idx],
ARRAY_SIZE(fake_samples[idx]),
he->thread, he->ms.map, he->ms.sym)) {
count_dummy++;
}
count_pair++;
} else if (idx) {
pr_debug("A entry from the other hists should have pair\n");
return -1;
}
count++;
node = rb_next(node);
}
/*
* Note that we have a entry collapsed in the other (idx = 1) hists.
*/
if (idx == 0) {
if (count_dummy != ARRAY_SIZE(fake_samples[1]) - 1) {
pr_debug("Invalid count of dummy entries: %zd of %zd\n",
count_dummy, ARRAY_SIZE(fake_samples[1]) - 1);
return -1;
}
if (count != count_pair + ARRAY_SIZE(fake_samples[0])) {
pr_debug("Invalid count of total leader entries: %zd of %zd\n",
count, count_pair + ARRAY_SIZE(fake_samples[0]));
return -1;
}
} else {
if (count != count_pair) {
pr_debug("Invalid count of total other entries: %zd of %zd\n",
count, count_pair);
return -1;
}
if (count_dummy > 0) {
pr_debug("Other hists should not have dummy entries: %zd\n",
count_dummy);
return -1;
}
}
return 0;
}
static int validate_link(struct hists *leader, struct hists *other)
{
return __validate_link(leader, 0) || __validate_link(other, 1);
}
int test__hists_link(int subtest __maybe_unused)
{
int err = -1;
struct hists *hists, *first_hists;
struct machines machines;
struct machine *machine = NULL;
struct perf_evsel *evsel, *first;
struct perf_evlist *evlist = perf_evlist__new();
if (evlist == NULL)
return -ENOMEM;
perf tools: Add parse_events_error interface Adding support to return error information from parse_events function. Following struct will be populated by parse_events function on return: struct parse_events_error { int idx; char *str; char *help; }; where 'idx' is the position in the string where the parsing failed, 'str' contains dynamically allocated error string describing the error and 'help' is optional help string. The change contains reporting function, which currently does not display anything. The code changes to supply error data for specific event types are coming in next patches. However this is what the expected output is: $ sudo perf record -e 'sched:krava' ls event syntax error: 'sched:krava' \___ unknown tracepoint ... $ perf record -e 'cpu/even=0x1/' ls event syntax error: 'cpu/even=0x1/' \___ unknown term valid terms: pc,any,inv,edge,cmask,event,in_tx,ldlat,umask,in_tx_cp,offcore_rsp,config,config1,config2,name,period,branch_type ... $ perf record -e cycles,cache-mises ls event syntax error: '..es,cache-mises' \___ parser error ... The output functions cut the beginning of the event string so the error starts up to 10th character and cut the end of the string of it crosses the terminal width. Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: David Ahern <dsahern@gmail.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1429729824-13932-2-git-send-email-jolsa@kernel.org [ Renamed 'error' variables to 'err', not to clash with util.h error() ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-23 03:10:16 +08:00
err = parse_events(evlist, "cpu-clock", NULL);
if (err)
goto out;
perf tools: Add parse_events_error interface Adding support to return error information from parse_events function. Following struct will be populated by parse_events function on return: struct parse_events_error { int idx; char *str; char *help; }; where 'idx' is the position in the string where the parsing failed, 'str' contains dynamically allocated error string describing the error and 'help' is optional help string. The change contains reporting function, which currently does not display anything. The code changes to supply error data for specific event types are coming in next patches. However this is what the expected output is: $ sudo perf record -e 'sched:krava' ls event syntax error: 'sched:krava' \___ unknown tracepoint ... $ perf record -e 'cpu/even=0x1/' ls event syntax error: 'cpu/even=0x1/' \___ unknown term valid terms: pc,any,inv,edge,cmask,event,in_tx,ldlat,umask,in_tx_cp,offcore_rsp,config,config1,config2,name,period,branch_type ... $ perf record -e cycles,cache-mises ls event syntax error: '..es,cache-mises' \___ parser error ... The output functions cut the beginning of the event string so the error starts up to 10th character and cut the end of the string of it crosses the terminal width. Signed-off-by: Jiri Olsa <jolsa@kernel.org> Cc: David Ahern <dsahern@gmail.com> Cc: Namhyung Kim <namhyung@kernel.org> Cc: Paul Mackerras <paulus@samba.org> Cc: Peter Zijlstra <a.p.zijlstra@chello.nl> Link: http://lkml.kernel.org/r/1429729824-13932-2-git-send-email-jolsa@kernel.org [ Renamed 'error' variables to 'err', not to clash with util.h error() ] Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2015-04-23 03:10:16 +08:00
err = parse_events(evlist, "task-clock", NULL);
if (err)
goto out;
/* default sort order (comm,dso,sym) will be used */
if (setup_sorting() < 0)
goto out;
machines__init(&machines);
/* setup threads/dso/map/symbols also */
machine = setup_fake_machine(&machines);
if (!machine)
goto out;
if (verbose > 1)
machine__fprintf(machine, stderr);
/* process sample events */
err = add_hist_entries(evlist, machine);
if (err < 0)
goto out;
evlist__for_each(evlist, evsel) {
hists = evsel__hists(evsel);
hists__collapse_resort(hists, NULL);
if (verbose > 2)
print_hists_in(hists);
}
first = perf_evlist__first(evlist);
evsel = perf_evlist__last(evlist);
first_hists = evsel__hists(first);
hists = evsel__hists(evsel);
/* match common entries */
hists__match(first_hists, hists);
err = validate_match(first_hists, hists);
if (err)
goto out;
/* link common and/or dummy entries */
hists__link(first_hists, hists);
err = validate_link(first_hists, hists);
if (err)
goto out;
err = 0;
out:
/* tear down everything */
perf_evlist__delete(evlist);
reset_output_field();
machines__exit(&machines);
return err;
}