tracing/ftrace: stop {irqs, preempt}soff tracers when tracing is stopped
authorFrederic Weisbecker <fweisbec@gmail.com>
Tue, 17 Mar 2009 21:38:58 +0000 (22:38 +0100)
committerIngo Molnar <mingo@elte.hu>
Wed, 18 Mar 2009 09:12:38 +0000 (10:12 +0100)
Impact: fix a selftest warning

In some cases, it's possible to see the following warning on irqsoff
tracer selftest:

[    4.640003] Testing tracer irqsoff: <4>------------[ cut here ]------------
[    4.653562] WARNING: at kernel/trace/trace.c:458 update_max_tr_single+0x9a/0xc4()
[    4.660000] Hardware name: System Product Name
[    4.660000] Modules linked in:
[    4.660000] Pid: 301, comm: kstop/1 Not tainted 2.6.29-rc8-tip #35837
[    4.660000] Call Trace:
[    4.660000]  [<4014b588>] warn_slowpath+0x79/0x8f
[    4.660000]  [<402d6949>] ? put_dec+0x64/0x6b
[    4.660000]  [<40162b56>] ? getnstimeofday+0x58/0xdd
[    4.660000]  [<40162210>] ? clocksource_read+0x3/0xf
[    4.660000]  [<4015eb44>] ? ktime_set+0x8/0x34
[    4.660000]  [<4014101a>] ? balance_runtime+0x8/0x56
[    4.660000]  [<405f6f11>] ? _spin_lock+0x3/0x10
[    4.660000]  [<4011f643>] ? ftrace_call+0x5/0x8
[    4.660000]  [<4015d0f1>] ? task_cputime_zero+0x3/0x27
[    4.660000]  [<40190ee7>] ? cpupri_set+0x90/0xcb
[    4.660000]  [<405f7208>] ? _spin_lock_irqsave+0x22/0x34
[    4.660000]  [<40190f12>] ? cpupri_set+0xbb/0xcb
[    4.660000]  [<405f7151>] ? _spin_unlock_irqrestore+0x23/0x35
[    4.660000]  [<4018493f>] ? ring_buffer_reset_cpu+0x27/0x51
[    4.660000]  [<405f7208>] ? _spin_lock_irqsave+0x22/0x34
[    4.660000]  [<40184962>] ? ring_buffer_reset_cpu+0x4a/0x51
[    4.660000]  [<405f7151>] ? _spin_unlock_irqrestore+0x23/0x35
[    4.660000]  [<4018cc29>] ? trace_hardirqs_off+0x1a/0x1c
[    4.660000]  [<405f7151>] ? _spin_unlock_irqrestore+0x23/0x35
[    4.660000]  [<40184962>] ? ring_buffer_reset_cpu+0x4a/0x51
[    4.660000]  [<401850f3>] ? cpumask_next+0x15/0x18
[    4.660000]  [<4018a41f>] update_max_tr_single+0x9a/0xc4
[    4.660000]  [<4014e5fe>] ? exit_notify+0x16/0xf2
[    4.660000]  [<4018cd13>] check_critical_timing+0xcc/0x11e
[    4.660000]  [<4014e5fe>] ? exit_notify+0x16/0xf2
[    4.660000]  [<4014e5fe>] ? exit_notify+0x16/0xf2
[    4.660000]  [<4018cdf1>] stop_critical_timing+0x8c/0x9f
[    4.660000]  [<4014e5c4>] ? forget_original_parent+0xac/0xd0
[    4.660000]  [<4018ce3a>] trace_hardirqs_on+0x1a/0x1c
[    4.660000]  [<4014e5c4>] forget_original_parent+0xac/0xd0
[    4.660000]  [<4014e5fe>] exit_notify+0x16/0xf2
[    4.660000]  [<4014e8a5>] do_exit+0x1cb/0x225
[    4.660000]  [<4015c72b>] ? kthread+0x0/0x69
[    4.660000]  [<4011f61d>] kernel_thread_helper+0xd/0x10
[    4.660000] ---[ end trace a7919e7f17c0a725 ]---
[    4.660164] .. no entries found ..FAILED!

During the selftest of irqsoff tracer, we do that:

/* disable interrupts for a bit */
local_irq_disable();
udelay(100);
local_irq_enable();
/* stop the tracing. */
tracing_stop();
/* check both trace buffers */
ret = trace_test_buffer(tr, NULL);

If a callsite performs a new max delay with irqs off just after
tracing_stop, update_max_tr_single() -> ring_buffer_swap_cpu()
will be called with the buffers disabled by tracing_stop(), hence
the warning, then ring_buffer_swap_cpu() return -EAGAIN and
update_max_tr_single() complains.

Fix it by also stopping the tracer before stopping the tracing globally.
A similar situation can happen with preemptoff and preemptirqsoff tracers
where we apply the same fix.

Reported-by: Ingo Molnar <mingo@elte.hu>
Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
Acked-by: Steven Rostedt <rostedt@goodmis.org>
LKML-Reference: <1237325938-5240-1-git-send-email-fweisbec@gmail.com>
Signed-off-by: Ingo Molnar <mingo@elte.hu>
kernel/trace/trace_selftest.c

index a2ca6f0fef9bb459d99038e2436f3e2a5f0a85bc..38856ba78a921e3251d7af08c74101fd4aa7edfc 100644 (file)
@@ -315,6 +315,14 @@ trace_selftest_startup_irqsoff(struct tracer *trace, struct trace_array *tr)
        local_irq_disable();
        udelay(100);
        local_irq_enable();
+
+       /*
+        * Stop the tracer to avoid a warning subsequent
+        * to buffer flipping failure because tracing_stop()
+        * disables the tr and max buffers, making flipping impossible
+        * in case of parallels max irqs off latencies.
+        */
+       trace->stop(tr);
        /* stop the tracing. */
        tracing_stop();
        /* check both trace buffers */
@@ -369,6 +377,14 @@ trace_selftest_startup_preemptoff(struct tracer *trace, struct trace_array *tr)
        preempt_disable();
        udelay(100);
        preempt_enable();
+
+       /*
+        * Stop the tracer to avoid a warning subsequent
+        * to buffer flipping failure because tracing_stop()
+        * disables the tr and max buffers, making flipping impossible
+        * in case of parallels max preempt off latencies.
+        */
+       trace->stop(tr);
        /* stop the tracing. */
        tracing_stop();
        /* check both trace buffers */
@@ -428,6 +444,13 @@ trace_selftest_startup_preemptirqsoff(struct tracer *trace, struct trace_array *
        /* reverse the order of preempt vs irqs */
        local_irq_enable();
 
+       /*
+        * Stop the tracer to avoid a warning subsequent
+        * to buffer flipping failure because tracing_stop()
+        * disables the tr and max buffers, making flipping impossible
+        * in case of parallels max irqs/preempt off latencies.
+        */
+       trace->stop(tr);
        /* stop the tracing. */
        tracing_stop();
        /* check both trace buffers */
@@ -448,6 +471,8 @@ trace_selftest_startup_preemptirqsoff(struct tracer *trace, struct trace_array *
        /* do the test by disabling interrupts first this time */
        tracing_max_latency = 0;
        tracing_start();
+       trace->start(tr);
+
        preempt_disable();
        local_irq_disable();
        udelay(100);
@@ -455,6 +480,7 @@ trace_selftest_startup_preemptirqsoff(struct tracer *trace, struct trace_array *
        /* reverse the order of preempt vs irqs */
        local_irq_enable();
 
+       trace->stop(tr);
        /* stop the tracing. */
        tracing_stop();
        /* check both trace buffers */