dm-devel.redhat.com archive mirror
 help / color / mirror / Atom feed
From: mwilck@suse.com
To: Christophe Varoqui <christophe.varoqui@opensvc.com>,
	Benjamin Marzinski <bmarzins@redhat.com>
Cc: lixiaokeng@huawei.com, dm-devel@redhat.com,
	Martin Wilck <mwilck@suse.com>
Subject: [PATCH 18/23] libmultipath: fix log_thread startup and teardown
Date: Thu, 24 Sep 2020 15:40:49 +0200	[thread overview]
Message-ID: <20200924134054.14632-19-mwilck@suse.com> (raw)
In-Reply-To: <20200924134054.14632-1-mwilck@suse.com>

From: Martin Wilck <mwilck@suse.com>

This fixes several issues with the log_thread. First, the running
flag logq_running should be set by the thread itself, not by
log_thread_start()/_stop(). Second, the thread was both cancelled and
terminated via a flag (again, logq_running). It's sufficient,
and better, to just cancel it and use logq_running as indication for
successful termination. Third, the locking wasn't cancel-safe in some
places. Forth, log_thread_start() and log_thread_stop() didn't wait for
startup/teardown properly. Fifth, using (pthread_t)0 is wrong (pthread_t is
opaque; there's no guarantee that 0 is not a valid pthread_t value). Finally,
pthread_cancel() was called under logq_lock, which doesn't make sense to
me at all.

Signed-off-by: Martin Wilck <mwilck@suse.com>
---
 libmultipath/log_pthread.c | 59 ++++++++++++++++++++++++++++----------
 1 file changed, 44 insertions(+), 15 deletions(-)

diff --git a/libmultipath/log_pthread.c b/libmultipath/log_pthread.c
index 0c327ff..5825e66 100644
--- a/libmultipath/log_pthread.c
+++ b/libmultipath/log_pthread.c
@@ -13,6 +13,7 @@
 #include "log_pthread.h"
 #include "log.h"
 #include "lock.h"
+#include "util.h"
 
 static pthread_t log_thr;
 
@@ -56,28 +57,45 @@ static void flush_logqueue (void)
 	} while (empty == 0);
 }
 
+static void cleanup_log_thread(__attribute((unused)) void *arg)
+{
+	logdbg(stderr, "log thread exiting");
+	pthread_mutex_lock(&logev_lock);
+	logq_running = 0;
+	pthread_cond_signal(&logev_cond);
+	pthread_mutex_unlock(&logev_lock);
+}
+
 static void * log_thread (__attribute__((unused)) void * et)
 {
 	int running;
 
 	pthread_mutex_lock(&logev_lock);
-	logq_running = 1;
+	running = logq_running;
+	if (!running)
+		logq_running = 1;
+	pthread_cond_signal(&logev_cond);
 	pthread_mutex_unlock(&logev_lock);
+	if (running)
+		/* already started */
+		return NULL;
+	pthread_cleanup_push(cleanup_log_thread, NULL);
 
 	mlockall(MCL_CURRENT | MCL_FUTURE);
 	logdbg(stderr,"enter log_thread\n");
 
 	while (1) {
 		pthread_mutex_lock(&logev_lock);
-		if (logq_running && !log_messages_pending)
+		pthread_cleanup_push(cleanup_mutex, &logev_lock);
+		while (!log_messages_pending)
+			/* this is a cancellation point */
 			pthread_cond_wait(&logev_cond, &logev_lock);
 		log_messages_pending = 0;
-		running = logq_running;
-		pthread_mutex_unlock(&logev_lock);
-		if (!running)
-			break;
+		pthread_cleanup_pop(1);
+
 		flush_logqueue();
 	}
+	pthread_cleanup_pop(1);
 	return NULL;
 }
 
@@ -98,6 +116,12 @@ void log_thread_start (pthread_attr_t *attr)
 		exit(1);
 	}
 
+	pthread_mutex_lock(&logev_lock);
+	pthread_cleanup_push(cleanup_mutex, &logev_lock);
+	/* wait for thread startup */
+	while (!logq_running)
+		pthread_cond_wait(&logev_cond, &logev_lock);
+	pthread_cleanup_pop(1);
 	return;
 }
 
@@ -112,21 +136,26 @@ void log_thread_reset (void)
 
 void log_thread_stop (void)
 {
+	int running;
+
 	if (!la)
 		return;
-
 	logdbg(stderr,"enter log_thread_stop\n");
 
 	pthread_mutex_lock(&logev_lock);
-	logq_running = 0;
-	pthread_cond_signal(&logev_cond);
-	pthread_mutex_unlock(&logev_lock);
+	pthread_cleanup_push(cleanup_mutex, &logev_lock);
+	running = logq_running;
+	if (running) {
+		pthread_cancel(log_thr);
+		pthread_cond_signal(&logev_cond);
+	}
+	while (logq_running)
+		pthread_cond_wait(&logev_cond, &logev_lock);
+	pthread_cleanup_pop(1);
+
+	if (running)
+		pthread_join(log_thr, NULL);
 
-	pthread_mutex_lock(&logq_lock);
-	pthread_cancel(log_thr);
-	pthread_mutex_unlock(&logq_lock);
-	pthread_join(log_thr, NULL);
-	log_thr = (pthread_t)0;
 
 	flush_logqueue();
 
-- 
2.28.0

  parent reply	other threads:[~2020-09-24 13:40 UTC|newest]

Thread overview: 37+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-09-24 13:40 [PATCH 00/23] libmultipath: improve cleanup on exit mwilck
2020-09-24 13:40 ` [PATCH 01/23] multipathd: uxlsnr: avoid deadlock " mwilck
2020-09-26  1:52   ` Benjamin Marzinski
2020-09-26 13:00     ` Martin Wilck
2020-09-24 13:40 ` [PATCH 02/23] multipathd: Fix liburcu memory leak mwilck
2020-09-24 13:40 ` [PATCH 03/23] multipathd: move handling of io_err_stat_attr into libmultipath mwilck
2020-09-24 13:40 ` [PATCH 04/23] multipathd: move vecs desctruction into cleanup function mwilck
2020-09-24 13:40 ` [PATCH 05/23] multipathd: make some globals static mwilck
2020-09-24 13:40 ` [PATCH 06/23] multipathd: move threads destruction into separate function mwilck
2020-09-24 13:40 ` [PATCH 07/23] multipathd: move conf " mwilck
2020-09-28 18:03   ` Benjamin Marzinski
2020-09-29  9:12     ` Martin Wilck
2020-09-24 13:40 ` [PATCH 08/23] multipathd: move pid " mwilck
2020-09-24 13:40 ` [PATCH 09/23] multipathd: close pidfile on exit mwilck
2020-09-24 13:40 ` [PATCH 10/23] multipathd: add helper for systemd notification at exit mwilck
2020-09-24 13:40 ` [PATCH 11/23] multipathd: child(): call cleanups in failure case, too mwilck
2020-09-24 13:40 ` [PATCH 12/23] multipathd: unwatch_all_dmevents: check if waiter is initialized mwilck
2020-09-24 13:40 ` [PATCH 13/23] multipathd: print error message if config can't be loaded mwilck
2020-09-24 13:40 ` [PATCH 14/23] libmultipath: add libmp_dm_exit() mwilck
2020-09-28 18:41   ` Benjamin Marzinski
2020-09-24 13:40 ` [PATCH 15/23] multipathd: fixup libdm deinitialization mwilck
2020-09-24 13:40 ` [PATCH 16/23] libmultipath: log_thread_stop(): check if logarea is initialized mwilck
2020-09-24 13:40 ` [PATCH 17/23] multipathd: add cleanup_child() exit handler mwilck
2020-09-24 13:40 ` mwilck [this message]
2020-09-28 20:15   ` [PATCH 18/23] libmultipath: fix log_thread startup and teardown Benjamin Marzinski
2020-09-29  9:18     ` Martin Wilck
2020-09-24 13:40 ` [PATCH 19/23] multipathd: move cleanup_{prio, checkers, foreign} to libmultipath_exit mwilck
2020-09-28 20:26   ` Benjamin Marzinski
2020-09-29  9:31     ` Martin Wilck
2020-09-29 17:50       ` Benjamin Marzinski
2020-09-24 13:40 ` [PATCH 20/23] multipath: use atexit() for cleanup handlers mwilck
2020-09-24 13:40 ` [PATCH 21/23] mpathpersist: " mwilck
2020-09-24 13:40 ` [PATCH 22/23] multipath: fix leaks in check_path_valid() mwilck
2020-09-24 13:40 ` [PATCH 23/23] multipath-tools: mpath-tools.supp: file with valgrind suppressions mwilck
2020-09-24 20:58   ` Benjamin Marzinski
2020-09-25  9:53     ` Martin Wilck
2020-09-28 20:51 ` [PATCH 00/23] libmultipath: improve cleanup on exit Benjamin Marzinski

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200924134054.14632-19-mwilck@suse.com \
    --to=mwilck@suse.com \
    --cc=bmarzins@redhat.com \
    --cc=christophe.varoqui@opensvc.com \
    --cc=dm-devel@redhat.com \
    --cc=lixiaokeng@huawei.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).