aboutsummaryrefslogtreecommitdiffstats
path: root/kernel/trace/trace.c
diff options
context:
space:
mode:
authorSteven Rostedt <srostedt@redhat.com>2009-09-01 11:06:29 -0400
committerSteven Rostedt <rostedt@goodmis.org>2009-09-04 18:44:22 -0400
commit2f26ebd549b9ab55ac756b836ec759c11fe93f81 (patch)
tree0f6fb154e2b5e9233a683b3267e5efe862a846f4 /kernel/trace/trace.c
parent76f0d07376388f32698ba51b6090a26b90c1342f (diff)
tracing: use timestamp to determine start of latency traces
Currently the latency tracers reset the ring buffer. Unfortunately if a commit is in process (due to a trace event), this can corrupt the ring buffer. When this happens, the ring buffer will detect the corruption and then permanently disable the ring buffer. The bug does not crash the system, but it does prevent further tracing after the bug is hit. Instead of reseting the trace buffers, the timestamp of the start of the trace is used instead. The buffers will still contain the previous data, but the output will not count any data that is before the timestamp of the trace. Note, this only affects the static trace output (trace) and not the runtime trace output (trace_pipe). The runtime trace output does not make sense for the latency tracers anyway. Reported-by: Arnaldo Carvalho de Melo <acme@redhat.com> Signed-off-by: Steven Rostedt <rostedt@goodmis.org>
Diffstat (limited to 'kernel/trace/trace.c')
-rw-r--r--kernel/trace/trace.c80
1 files changed, 63 insertions, 17 deletions
diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
index 54517a889791..7daf372e319a 100644
--- a/kernel/trace/trace.c
+++ b/kernel/trace/trace.c
@@ -454,10 +454,6 @@ update_max_tr(struct trace_array *tr, struct task_struct *tsk, int cpu)
454 tr->buffer = max_tr.buffer; 454 tr->buffer = max_tr.buffer;
455 max_tr.buffer = buf; 455 max_tr.buffer = buf;
456 456
457 ftrace_disable_cpu();
458 ring_buffer_reset(tr->buffer);
459 ftrace_enable_cpu();
460
461 __update_max_tr(tr, tsk, cpu); 457 __update_max_tr(tr, tsk, cpu);
462 __raw_spin_unlock(&ftrace_max_lock); 458 __raw_spin_unlock(&ftrace_max_lock);
463} 459}
@@ -483,7 +479,6 @@ update_max_tr_single(struct trace_array *tr, struct task_struct *tsk, int cpu)
483 479
484 ftrace_disable_cpu(); 480 ftrace_disable_cpu();
485 481
486 ring_buffer_reset(max_tr.buffer);
487 ret = ring_buffer_swap_cpu(max_tr.buffer, tr->buffer, cpu); 482 ret = ring_buffer_swap_cpu(max_tr.buffer, tr->buffer, cpu);
488 483
489 ftrace_enable_cpu(); 484 ftrace_enable_cpu();
@@ -1374,6 +1369,37 @@ static void *s_next(struct seq_file *m, void *v, loff_t *pos)
1374 return ent; 1369 return ent;
1375} 1370}
1376 1371
1372static void tracing_iter_reset(struct trace_iterator *iter, int cpu)
1373{
1374 struct trace_array *tr = iter->tr;
1375 struct ring_buffer_event *event;
1376 struct ring_buffer_iter *buf_iter;
1377 unsigned long entries = 0;
1378 u64 ts;
1379
1380 tr->data[cpu]->skipped_entries = 0;
1381
1382 if (!iter->buffer_iter[cpu])
1383 return;
1384
1385 buf_iter = iter->buffer_iter[cpu];
1386 ring_buffer_iter_reset(buf_iter);
1387
1388 /*
1389 * We could have the case with the max latency tracers
1390 * that a reset never took place on a cpu. This is evident
1391 * by the timestamp being before the start of the buffer.
1392 */
1393 while ((event = ring_buffer_iter_peek(buf_iter, &ts))) {
1394 if (ts >= iter->tr->time_start)
1395 break;
1396 entries++;
1397 ring_buffer_read(buf_iter, NULL);
1398 }
1399
1400 tr->data[cpu]->skipped_entries = entries;
1401}
1402
1377/* 1403/*
1378 * No necessary locking here. The worst thing which can 1404 * No necessary locking here. The worst thing which can
1379 * happen is loosing events consumed at the same time 1405 * happen is loosing events consumed at the same time
@@ -1412,10 +1438,9 @@ static void *s_start(struct seq_file *m, loff_t *pos)
1412 1438
1413 if (cpu_file == TRACE_PIPE_ALL_CPU) { 1439 if (cpu_file == TRACE_PIPE_ALL_CPU) {
1414 for_each_tracing_cpu(cpu) 1440 for_each_tracing_cpu(cpu)
1415 ring_buffer_iter_reset(iter->buffer_iter[cpu]); 1441 tracing_iter_reset(iter, cpu);
1416 } else 1442 } else
1417 ring_buffer_iter_reset(iter->buffer_iter[cpu_file]); 1443 tracing_iter_reset(iter, cpu_file);
1418
1419 1444
1420 ftrace_enable_cpu(); 1445 ftrace_enable_cpu();
1421 1446
@@ -1464,16 +1489,32 @@ print_trace_header(struct seq_file *m, struct trace_iterator *iter)
1464 struct trace_array *tr = iter->tr; 1489 struct trace_array *tr = iter->tr;
1465 struct trace_array_cpu *data = tr->data[tr->cpu]; 1490 struct trace_array_cpu *data = tr->data[tr->cpu];
1466 struct tracer *type = current_trace; 1491 struct tracer *type = current_trace;
1467 unsigned long total; 1492 unsigned long entries = 0;
1468 unsigned long entries; 1493 unsigned long total = 0;
1494 unsigned long count;
1469 const char *name = "preemption"; 1495 const char *name = "preemption";
1496 int cpu;
1470 1497
1471 if (type) 1498 if (type)
1472 name = type->name; 1499 name = type->name;
1473 1500
1474 entries = ring_buffer_entries(iter->tr->buffer); 1501
1475 total = entries + 1502 for_each_tracing_cpu(cpu) {
1476 ring_buffer_overruns(iter->tr->buffer); 1503 count = ring_buffer_entries_cpu(tr->buffer, cpu);
1504 /*
1505 * If this buffer has skipped entries, then we hold all
1506 * entries for the trace and we need to ignore the
1507 * ones before the time stamp.
1508 */
1509 if (tr->data[cpu]->skipped_entries) {
1510 count -= tr->data[cpu]->skipped_entries;
1511 /* total is the same as the entries */
1512 total += count;
1513 } else
1514 total += count +
1515 ring_buffer_overrun_cpu(tr->buffer, cpu);
1516 entries += count;
1517 }
1477 1518
1478 seq_printf(m, "# %s latency trace v1.1.5 on %s\n", 1519 seq_printf(m, "# %s latency trace v1.1.5 on %s\n",
1479 name, UTS_RELEASE); 1520 name, UTS_RELEASE);
@@ -1534,6 +1575,9 @@ static void test_cpu_buff_start(struct trace_iterator *iter)
1534 if (cpumask_test_cpu(iter->cpu, iter->started)) 1575 if (cpumask_test_cpu(iter->cpu, iter->started))
1535 return; 1576 return;
1536 1577
1578 if (iter->tr->data[iter->cpu]->skipped_entries)
1579 return;
1580
1537 cpumask_set_cpu(iter->cpu, iter->started); 1581 cpumask_set_cpu(iter->cpu, iter->started);
1538 1582
1539 /* Don't print started cpu buffer for the first entry of the trace */ 1583 /* Don't print started cpu buffer for the first entry of the trace */
@@ -1796,19 +1840,23 @@ __tracing_open(struct inode *inode, struct file *file)
1796 if (ring_buffer_overruns(iter->tr->buffer)) 1840 if (ring_buffer_overruns(iter->tr->buffer))
1797 iter->iter_flags |= TRACE_FILE_ANNOTATE; 1841 iter->iter_flags |= TRACE_FILE_ANNOTATE;
1798 1842
1843 /* stop the trace while dumping */
1844 tracing_stop();
1845
1799 if (iter->cpu_file == TRACE_PIPE_ALL_CPU) { 1846 if (iter->cpu_file == TRACE_PIPE_ALL_CPU) {
1800 for_each_tracing_cpu(cpu) { 1847 for_each_tracing_cpu(cpu) {
1801 1848
1802 iter->buffer_iter[cpu] = 1849 iter->buffer_iter[cpu] =
1803 ring_buffer_read_start(iter->tr->buffer, cpu); 1850 ring_buffer_read_start(iter->tr->buffer, cpu);
1851 tracing_iter_reset(iter, cpu);
1804 } 1852 }
1805 } else { 1853 } else {
1806 cpu = iter->cpu_file; 1854 cpu = iter->cpu_file;
1807 iter->buffer_iter[cpu] = 1855 iter->buffer_iter[cpu] =
1808 ring_buffer_read_start(iter->tr->buffer, cpu); 1856 ring_buffer_read_start(iter->tr->buffer, cpu);
1857 tracing_iter_reset(iter, cpu);
1809 } 1858 }
1810 1859
1811 /* TODO stop tracer */
1812 ret = seq_open(file, &tracer_seq_ops); 1860 ret = seq_open(file, &tracer_seq_ops);
1813 if (ret < 0) { 1861 if (ret < 0) {
1814 fail_ret = ERR_PTR(ret); 1862 fail_ret = ERR_PTR(ret);
@@ -1818,9 +1866,6 @@ __tracing_open(struct inode *inode, struct file *file)
1818 m = file->private_data; 1866 m = file->private_data;
1819 m->private = iter; 1867 m->private = iter;
1820 1868
1821 /* stop the trace while dumping */
1822 tracing_stop();
1823
1824 mutex_unlock(&trace_types_lock); 1869 mutex_unlock(&trace_types_lock);
1825 1870
1826 return iter; 1871 return iter;
@@ -1831,6 +1876,7 @@ __tracing_open(struct inode *inode, struct file *file)
1831 ring_buffer_read_finish(iter->buffer_iter[cpu]); 1876 ring_buffer_read_finish(iter->buffer_iter[cpu]);
1832 } 1877 }
1833 free_cpumask_var(iter->started); 1878 free_cpumask_var(iter->started);
1879 tracing_start();
1834 fail: 1880 fail:
1835 mutex_unlock(&trace_types_lock); 1881 mutex_unlock(&trace_types_lock);
1836 kfree(iter->trace); 1882 kfree(iter->trace);