Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip
Pull perf fixes from Thomas Gleixner:
"A pile of perf updates:
- Fix broken sanity check in the /proc/sys/kernel/perf_cpu_time_max_percent
write handler
- Cure a perf script crash which caused by an unitinialized data
structure
- Highlight the hottest instruction in perf top and not a random one
- Cure yet another clang issue when building perf python
- Handle topology entries with no CPU correctly in the tools
- Handle perf data which contains both tracepoints and performance
counter entries correctly.
- Add a missing NULL pointer check in perf ordered_events_free()"
* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
perf script: Fix crash when processing recorded stat data
perf top: Fix wrong hottest instruction highlighted
perf tools: Handle TOPOLOGY headers with no CPU
perf python: Remove -fstack-clash-protection when building with some clang versions
perf core: Fix perf_proc_update_handler() bug
perf script: Fix crash with printing mixed trace point and other events
perf ordered_events: Fix crash in ordered_events__free
+35
-23
+7
-7
kernel/events/core.c
+7
-7
kernel/events/core.c
···
436
436
void __user *buffer, size_t *lenp,
437
437
loff_t *ppos)
438
438
{
439
-
int ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
440
-
441
-
if (ret || !write)
442
-
return ret;
443
-
439
+
int ret;
440
+
int perf_cpu = sysctl_perf_cpu_time_max_percent;
444
441
/*
445
442
* If throttling is disabled don't allow the write:
446
443
*/
447
-
if (sysctl_perf_cpu_time_max_percent == 100 ||
448
-
sysctl_perf_cpu_time_max_percent == 0)
444
+
if (write && (perf_cpu == 100 || perf_cpu == 0))
449
445
return -EINVAL;
446
+
447
+
ret = proc_dointvec_minmax(table, write, buffer, lenp, ppos);
448
+
if (ret || !write)
449
+
return ret;
450
450
451
451
max_samples_per_tick = DIV_ROUND_UP(sysctl_perf_event_sample_rate, HZ);
452
452
perf_sample_period_ns = NSEC_PER_SEC / sysctl_perf_event_sample_rate;
+3
-6
tools/perf/builtin-script.c
+3
-6
tools/perf/builtin-script.c
···
1681
1681
.force_header = false,
1682
1682
};
1683
1683
struct perf_evsel *ev2;
1684
-
static bool init;
1685
1684
u64 val;
1686
1685
1687
-
if (!init) {
1688
-
perf_stat__init_shadow_stats();
1689
-
init = true;
1690
-
}
1691
1686
if (!evsel->stats)
1692
1687
perf_evlist__alloc_stats(script->session->evlist, false);
1693
1688
if (evsel_script(evsel->leader)->gnum++ == 0)
···
1789
1794
return;
1790
1795
}
1791
1796
1792
-
if (PRINT_FIELD(TRACE)) {
1797
+
if (PRINT_FIELD(TRACE) && sample->raw_data) {
1793
1798
event_format__fprintf(evsel->tp_format, sample->cpu,
1794
1799
sample->raw_data, sample->raw_size, fp);
1795
1800
}
···
2353
2358
int ret;
2354
2359
2355
2360
signal(SIGINT, sig_handler);
2361
+
2362
+
perf_stat__init_shadow_stats();
2356
2363
2357
2364
/* override event processing functions */
2358
2365
if (script->show_task_events) {
+10
-6
tools/perf/ui/browsers/annotate.c
+10
-6
tools/perf/ui/browsers/annotate.c
···
224
224
return ret;
225
225
}
226
226
227
-
static int disasm__cmp(struct annotation_line *a, struct annotation_line *b)
227
+
static double disasm__cmp(struct annotation_line *a, struct annotation_line *b,
228
+
int percent_type)
228
229
{
229
230
int i;
230
231
231
232
for (i = 0; i < a->data_nr; i++) {
232
-
if (a->data[i].percent == b->data[i].percent)
233
+
if (a->data[i].percent[percent_type] == b->data[i].percent[percent_type])
233
234
continue;
234
-
return a->data[i].percent < b->data[i].percent;
235
+
return a->data[i].percent[percent_type] -
236
+
b->data[i].percent[percent_type];
235
237
}
236
238
return 0;
237
239
}
238
240
239
-
static void disasm_rb_tree__insert(struct rb_root *root, struct annotation_line *al)
241
+
static void disasm_rb_tree__insert(struct annotate_browser *browser,
242
+
struct annotation_line *al)
240
243
{
244
+
struct rb_root *root = &browser->entries;
241
245
struct rb_node **p = &root->rb_node;
242
246
struct rb_node *parent = NULL;
243
247
struct annotation_line *l;
···
250
246
parent = *p;
251
247
l = rb_entry(parent, struct annotation_line, rb_node);
252
248
253
-
if (disasm__cmp(al, l))
249
+
if (disasm__cmp(al, l, browser->opts->percent_type) < 0)
254
250
p = &(*p)->rb_left;
255
251
else
256
252
p = &(*p)->rb_right;
···
333
329
RB_CLEAR_NODE(&pos->al.rb_node);
334
330
continue;
335
331
}
336
-
disasm_rb_tree__insert(&browser->entries, &pos->al);
332
+
disasm_rb_tree__insert(browser, &pos->al);
337
333
}
338
334
pthread_mutex_unlock(¬es->lock);
339
335
+9
-2
tools/perf/util/cpumap.c
+9
-2
tools/perf/util/cpumap.c
···
134
134
if (!cpu_list)
135
135
return cpu_map__read_all_cpu_map();
136
136
137
-
if (!isdigit(*cpu_list))
137
+
/*
138
+
* must handle the case of empty cpumap to cover
139
+
* TOPOLOGY header for NUMA nodes with no CPU
140
+
* ( e.g., because of CPU hotplug)
141
+
*/
142
+
if (!isdigit(*cpu_list) && *cpu_list != '\0')
138
143
goto out;
139
144
140
145
while (isdigit(*cpu_list)) {
···
186
181
187
182
if (nr_cpus > 0)
188
183
cpus = cpu_map__trim_new(nr_cpus, tmp_cpus);
189
-
else
184
+
else if (*cpu_list != '\0')
190
185
cpus = cpu_map__default_new();
186
+
else
187
+
cpus = cpu_map__dummy_new();
191
188
invalid:
192
189
free(tmp_cpus);
193
190
out:
+4
-2
tools/perf/util/ordered-events.c
+4
-2
tools/perf/util/ordered-events.c
···
391
391
* Current buffer might not have all the events allocated
392
392
* yet, we need to free only allocated ones ...
393
393
*/
394
-
list_del(&oe->buffer->list);
395
-
ordered_events_buffer__free(oe->buffer, oe->buffer_idx, oe);
394
+
if (oe->buffer) {
395
+
list_del(&oe->buffer->list);
396
+
ordered_events_buffer__free(oe->buffer, oe->buffer_idx, oe);
397
+
}
396
398
397
399
/* ... and continue with the rest */
398
400
list_for_each_entry_safe(buffer, tmp, &oe->to_free, list) {
+2
tools/perf/util/setup.py
+2
tools/perf/util/setup.py
···
17
17
vars[var] = sub("-mcet", "", vars[var])
18
18
if not clang_has_option("-fcf-protection"):
19
19
vars[var] = sub("-fcf-protection", "", vars[var])
20
+
if not clang_has_option("-fstack-clash-protection"):
21
+
vars[var] = sub("-fstack-clash-protection", "", vars[var])
20
22
21
23
from distutils.core import setup, Extension
22
24