* [Qemu-devel] [PATCH v2] trace/simple: fix hang in child after fork(2)
@ 2018-07-13 18:42 Stefan Hajnoczi
2018-07-15 12:42 ` Paolo Bonzini
0 siblings, 1 reply; 3+ messages in thread
From: Stefan Hajnoczi @ 2018-07-13 18:42 UTC (permalink / raw)
To: qemu-devel; +Cc: Cornelia Huck, Paolo Bonzini, Stefan Hajnoczi
The simple trace backend spawns a write-out thread which is used to
asynchronously flush the in-memory ring buffer to disk.
fork(2) does not clone all threads, only the thread that invoked
fork(2). As a result there is no write-out thread in the child process!
This causes a hang during shutdown when atexit(3) handler installed by
the simple trace backend waits for the non-existent write-out thread.
This patch uses pthread_atfork(3) to terminate the write-out thread
before fork and restart it in both the parent and child after fork.
This solves a hang in qemu-iotests 147 due to qemu-nbd --fork usage.
Reported-by: Cornelia Huck <cohuck@redhat.com>
Suggested-by: Paolo Bonzini <pbonzini@redhat.com>
Signed-off-by: Stefan Hajnoczi <stefanha@redhat.com>
---
trace/simple.c | 67 +++++++++++++++++++++++++++++++++++++++++++-------
1 file changed, 58 insertions(+), 9 deletions(-)
diff --git a/trace/simple.c b/trace/simple.c
index 701dec639c..9f6ac9ef24 100644
--- a/trace/simple.c
+++ b/trace/simple.c
@@ -39,9 +39,11 @@
static GMutex trace_lock;
static GCond trace_available_cond;
static GCond trace_empty_cond;
+static GThread *trace_writeout_thread;
static bool trace_available;
static bool trace_writeout_enabled;
+static bool trace_writeout_running;
enum {
TRACE_BUF_LEN = 4096 * 64,
@@ -142,15 +144,34 @@ static void flush_trace_file(bool wait)
g_mutex_unlock(&trace_lock);
}
-static void wait_for_trace_records_available(void)
+/**
+ * Wait to be kicked by flush_trace_file()
+ *
+ * Returns: true if the writeout thread should continue
+ * false if the writeout thread should terminate
+ */
+static bool wait_for_trace_records_available(void)
{
+ bool running;
+
g_mutex_lock(&trace_lock);
- while (!(trace_available && trace_writeout_enabled)) {
+ for (;;) {
+ running = trace_writeout_running;
+ if (!running) {
+ break;
+ }
+
+ if (trace_available && trace_writeout_enabled) {
+ break;
+ }
+
g_cond_signal(&trace_empty_cond);
g_cond_wait(&trace_available_cond, &trace_lock);
}
trace_available = false;
g_mutex_unlock(&trace_lock);
+
+ return running;
}
static gpointer writeout_thread(gpointer opaque)
@@ -165,9 +186,7 @@ static gpointer writeout_thread(gpointer opaque)
size_t unused __attribute__ ((unused));
uint64_t type = TRACE_RECORD_TYPE_EVENT;
- for (;;) {
- wait_for_trace_records_available();
-
+ while (wait_for_trace_records_available()) {
if (g_atomic_int_get(&dropped_events)) {
dropped.rec.event = DROPPED_EVENT_ID,
dropped.rec.timestamp_ns = get_clock();
@@ -398,18 +417,48 @@ static GThread *trace_thread_create(GThreadFunc fn)
return thread;
}
+#ifndef _WIN32
+static void stop_writeout_thread(void)
+{
+ g_mutex_lock(&trace_lock);
+ trace_writeout_running = false;
+ g_cond_signal(&trace_available_cond);
+ g_mutex_unlock(&trace_lock);
+
+ g_thread_join(trace_writeout_thread);
+ trace_writeout_thread = NULL;
+}
+
+static void restart_writeout_thread(void)
+{
+ trace_writeout_running = true;
+ trace_writeout_thread = trace_thread_create(writeout_thread);
+ if (!trace_writeout_thread) {
+ warn_report("unable to initialize simple trace backend");
+ }
+}
+#endif /* !_WIN32 */
+
bool st_init(void)
{
- GThread *thread;
-
trace_pid = getpid();
+ trace_writeout_running = true;
- thread = trace_thread_create(writeout_thread);
- if (!thread) {
+ trace_writeout_thread = trace_thread_create(writeout_thread);
+ if (!trace_writeout_thread) {
warn_report("unable to initialize simple trace backend");
return false;
}
+#ifndef _WIN32
+ /* Terminate writeout thread across fork and restart it in parent and
+ * child afterwards.
+ */
+ pthread_atfork(stop_writeout_thread,
+ restart_writeout_thread,
+ restart_writeout_thread);
+#endif
+
atexit(st_flush_trace_buffer);
return true;
}
--
2.17.1
^ permalink raw reply related [flat|nested] 3+ messages in thread
* Re: [Qemu-devel] [PATCH v2] trace/simple: fix hang in child after fork(2)
2018-07-13 18:42 [Qemu-devel] [PATCH v2] trace/simple: fix hang in child after fork(2) Stefan Hajnoczi
@ 2018-07-15 12:42 ` Paolo Bonzini
2018-07-16 11:09 ` Cornelia Huck
0 siblings, 1 reply; 3+ messages in thread
From: Paolo Bonzini @ 2018-07-15 12:42 UTC (permalink / raw)
To: Stefan Hajnoczi, qemu-devel; +Cc: Cornelia Huck
On 13/07/2018 20:42, Stefan Hajnoczi wrote:
> +#ifndef _WIN32
> +static void stop_writeout_thread(void)
> +{
> + g_mutex_lock(&trace_lock);
> + trace_writeout_running = false;
> + g_cond_signal(&trace_available_cond);
> + g_mutex_unlock(&trace_lock);
> +
> + g_thread_join(trace_writeout_thread);
> + trace_writeout_thread = NULL;
> +}
After stop_writeout_thread returns, another could start a write to the
shared data structure---and the write would never finish, because the
thread disappears after fork(2) returns. This would leave the mutex
locked, causing a deadlock soon after the fork. So you need to lock
trace_lock again here, and unlock it in restart_writeout_thread.
Apart from this, it looks good!
Thanks,
Paolo
> +static void restart_writeout_thread(void)
> +{
> + trace_writeout_running = true;
> + trace_writeout_thread = trace_thread_create(writeout_thread);
> + if (!trace_writeout_thread) {
> + warn_report("unable to initialize simple trace backend");
> + }
> +}
> +#endif /* !_WIN32 */
> +
> bool st_init(void)
> {
> - GThread *thread;
> -
> trace_pid = getpid();
> + trace_writeout_running = true;
>
> - thread = trace_thread_create(writeout_thread);
> - if (!thread) {
> + trace_writeout_thread = trace_thread_create(writeout_thread);
> + if (!trace_writeout_thread) {
> warn_report("unable to initialize simple trace backend");
> return false;
> }
>
> +#ifndef _WIN32
> + /* Terminate writeout thread across fork and restart it in parent and
> + * child afterwards.
> + */
> + pthread_atfork(stop_writeout_thread,
> + restart_writeout_thread,
> + restart_writeout_thread);
> +#endif
> +
> atexit(st_flush_trace_buffer);
> return true;
> }
>
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [Qemu-devel] [PATCH v2] trace/simple: fix hang in child after fork(2)
2018-07-15 12:42 ` Paolo Bonzini
@ 2018-07-16 11:09 ` Cornelia Huck
0 siblings, 0 replies; 3+ messages in thread
From: Cornelia Huck @ 2018-07-16 11:09 UTC (permalink / raw)
To: Paolo Bonzini; +Cc: Stefan Hajnoczi, qemu-devel
On Sun, 15 Jul 2018 14:42:23 +0200
Paolo Bonzini <pbonzini@redhat.com> wrote:
> On 13/07/2018 20:42, Stefan Hajnoczi wrote:
> > +#ifndef _WIN32
> > +static void stop_writeout_thread(void)
> > +{
> > + g_mutex_lock(&trace_lock);
> > + trace_writeout_running = false;
> > + g_cond_signal(&trace_available_cond);
> > + g_mutex_unlock(&trace_lock);
> > +
> > + g_thread_join(trace_writeout_thread);
> > + trace_writeout_thread = NULL;
> > +}
>
> After stop_writeout_thread returns, another could start a write to the
> shared data structure---and the write would never finish, because the
> thread disappears after fork(2) returns. This would leave the mutex
> locked, causing a deadlock soon after the fork. So you need to lock
> trace_lock again here, and unlock it in restart_writeout_thread.
So, I suppose there will be a v3, right?
>
> Apart from this, it looks good!
Did a quick run and it fixed the problems for me.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2018-07-16 11:09 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2018-07-13 18:42 [Qemu-devel] [PATCH v2] trace/simple: fix hang in child after fork(2) Stefan Hajnoczi
2018-07-15 12:42 ` Paolo Bonzini
2018-07-16 11:09 ` Cornelia Huck
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.