Skip to contentSkip to conversation

multipathd: make uxlsnr errors really fatal

PATCH Superseded Martin Wilck 2 participants 16 messages Oct 23, 2018 – Oct 30, 2018 Device mapper

v1 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

cover multipathd: make uxlsnr errors really fatal

Hi Christophe,

this series, based on top of the recently submitted "various multipath-tools patches (v2)" and "checkers overhaul" series, fixes a problem that I recently observed: despite
ee01e841 "multipathd: handle errors in uxlsnr as fatal", multipathd sometimes doesn't quit when the socket setup fails.

While at that, I stumbled upon init_path_check_interval(), wondered about its purpose, and came to the conclusion that can be quite easily obsoleted.

Martin

Martin Wilck (7):
libmultipath: set pp->checkint in store_pathinfo()
multipathd: remove init_path_check_interval()
multipathd: print error message if checkint is not initialized
multipathd: open client socket early
multipathd: set DAEMON_CONFIGURE from uxlsnr thread
multipathd: make DAEMON_SHUTDOWN a terminal state
multipathd: only grab conf once for filter_path()

 libmultipath/defaults.h   |  1 +
 libmultipath/dict.c       | 14 +++++-
 libmultipath/discovery.c  |  1 +
 libmultipath/structs.c    |  1 +
 multipathd/cli_handlers.c | 10 ++--
 multipathd/main.c         | 99 +++++++++++++++++++++++++--------------
 multipathd/uxlsnr.c       | 11 +----
 multipathd/uxlsnr.h       |  2 +-
 8 files changed, 88 insertions(+), 51 deletions(-)
Signature · 1 line
2.19.1
#1 lore raw

PATCH 1/7 libmultipath: set pp->checkint in store_pathinfo()

store_pathinfo is called with valid conf pointer anyway, so checkint is available. pp->checkint is now valid for every path after path_discovery().

This fixes a bad conf access in cli_add_path().

Cc Bart Van Assche <[email protected]>
---
 libmultipath/discovery.c  | 1 +
 multipathd/cli_handlers.c | 1 -
 multipathd/main.c         | 2 --
 3 files changed, 1 insertion(+), 3 deletions(-)
libmultipath/discovery.c +1 -0
index 467ece7a..4ac3fde1 100644
@@ -103,6 +103,7 @@ store_pathinfo (vector pathvec, struct config *conf,
 	err = store_path(pathvec, pp);
 	if (err)
 		goto out;
+	pp->checkint = conf->checkint;
 
 out:
 	if (err)
multipathd/cli_handlers.c +0 -1
index 75000807..4aea4ce7 100644
@@ -736,7 +736,6 @@ cli_add_path (void * v, char ** reply, int * len, void * data)
 			condlog(0, "%s: failed to store path info", param);
 			return 1;
 		}
-		pp->checkint = conf->checkint;
 	}
 	return ev_add_path(pp, vecs, 1);
 blacklisted:
multipathd/main.c +0 -2
index e80ac906..b8423d89 100644
@@ -2322,8 +2322,6 @@ configure (struct vectors * vecs)
 			free_path(pp);
 			i--;
 		}
-		else
-			pp->checkint = conf->checkint;
 		pthread_cleanup_pop(1);
 	}
 	if (map_discovery(vecs)) {
Signature · 1 line
2.19.1
maintainer Reviewed-by on v1 1/7
#2 lore raw

Re: [PATCH 1/7] libmultipath: set pp->checkint in store_pathinfo()

Quoted 6 lines · Martin Wilck

store_pathinfo is called with valid conf pointer anyway, so checkint is available. pp->checkint is now valid for every path after path_discovery().

This fixes a bad conf access in cli_add_path().

Quoted 47 lines
Cc Bart Van Assche <[email protected]>
---
 libmultipath/discovery.c  | 1 +
 multipathd/cli_handlers.c | 1 -
 multipathd/main.c         | 2 --
 3 files changed, 1 insertion(+), 3 deletions(-)
libmultipath/discovery.c +1 -0
index 467ece7a..4ac3fde1 100644
@@ -103,6 +103,7 @@ store_pathinfo (vector pathvec, struct config *conf,
 	err = store_path(pathvec, pp);
 	if (err)
 		goto out;
+	pp->checkint = conf->checkint;
 
 out:
 	if (err)
multipathd/cli_handlers.c +0 -1
index 75000807..4aea4ce7 100644
@@ -736,7 +736,6 @@ cli_add_path (void * v, char ** reply, int * len, void * data)
 			condlog(0, "%s: failed to store path info", param);
 			return 1;
 		}
-		pp->checkint = conf->checkint;
 	}
 	return ev_add_path(pp, vecs, 1);
 blacklisted:
multipathd/main.c +0 -2
index e80ac906..b8423d89 100644
@@ -2322,8 +2322,6 @@ configure (struct vectors * vecs)
 			free_path(pp);
 			i--;
 		}
-		else
-			pp->checkint = conf->checkint;
 		pthread_cleanup_pop(1);
 	}
 	if (map_discovery(vecs)) {
Signature · 1 line
2.19.1
#3 lore raw

PATCH 2/7 multipathd: remove init_path_check_interval()

In reply to Martin Wilck

After "libmultipath: set pp->checkint in store_pathinfo()", pp-checkint should always be properly initialized, so this code is not needed any more.

Cc Bart Van Assche <[email protected]>
---
 multipathd/main.c | 14 --------------
 1 file changed, 14 deletions(-)
multipathd/main.c +0 -14
index b8423d89..3464ebfa 100644
@@ -2131,19 +2131,6 @@ check_path (struct vectors * vecs, struct path * pp, int ticks)
 	return 1;
 }
 
-static void init_path_check_interval(struct vectors *vecs)
-{
-	struct config *conf;
-	struct path *pp;
-	unsigned int i;
-
-	vector_foreach_slot (vecs->pathvec, pp, i) {
-		conf = get_multipath_config();
-		pp->checkint = conf->checkint;
-		put_multipath_config(conf);
-	}
-}
-
 static void *
 checkerloop (void *ap)
 {
@@ -2727,7 +2714,6 @@ child (void * param)
 	 */
 	post_config_state(DAEMON_CONFIGURE);
 
-	init_path_check_interval(vecs);
 
 	if (poll_dmevents) {
 		if (init_dmevent_waiter(vecs)) {
Signature · 1 line
2.19.1
maintainer Reviewed-by on v1 2/7
#4 lore raw

Re: [PATCH 2/7] multipathd: remove init_path_check_interval()

Quoted 4 lines · Martin Wilck

After "libmultipath: set pp->checkint in store_pathinfo()", pp-checkint should always be properly initialized, so this code is not needed any more.

Quoted 40 lines
Cc Bart Van Assche <[email protected]>
---
 multipathd/main.c | 14 --------------
 1 file changed, 14 deletions(-)
multipathd/main.c +0 -14
index b8423d89..3464ebfa 100644
@@ -2131,19 +2131,6 @@ check_path (struct vectors * vecs, struct path * pp, int ticks)
 	return 1;
 }
 
-static void init_path_check_interval(struct vectors *vecs)
-{
-	struct config *conf;
-	struct path *pp;
-	unsigned int i;
-
-	vector_foreach_slot (vecs->pathvec, pp, i) {
-		conf = get_multipath_config();
-		pp->checkint = conf->checkint;
-		put_multipath_config(conf);
-	}
-}
-
 static void *
 checkerloop (void *ap)
 {
@@ -2727,7 +2714,6 @@ child (void * param)
 	 */
 	post_config_state(DAEMON_CONFIGURE);
 
-	init_path_check_interval(vecs);
 
 	if (poll_dmevents) {
 		if (init_dmevent_waiter(vecs)) {
Signature · 1 line
2.19.1
#5 lore raw

PATCH 3/7 multipathd: print error message if checkint is not initialized

In reply to Martin Wilck

This is just a safety measure in case I overlooked something wrt the checkint initialization. It could be reverted once we know the error message isn't triggered.

---
 libmultipath/defaults.h |  1 +
 libmultipath/dict.c     | 14 ++++++++++++--
 libmultipath/structs.c  |  1 +
 multipathd/main.c       |  6 ++++++
 4 files changed, 20 insertions(+), 2 deletions(-)
libmultipath/defaults.h +1 -0
index 7f3839fc..65769398 100644
@@ -46,6 +46,7 @@
 #define DEFAULT_UNKNOWN_FIND_MULTIPATHS_TIMEOUT 1
 #define DEFAULT_ALL_TG_PT ALL_TG_PT_OFF
 
+#define CHECKINT_UNDEF		(~0U)
 #define DEFAULT_CHECKINT	5
 #define MAX_CHECKINT(a)		(a << 2)
 
libmultipath/dict.c +12 -2
index c3f5a6e6..a81c051f 100644
@@ -264,7 +264,17 @@ snprint_mp_ ## option (struct config *conf, char * buff, int len,	\
 	return function (buff, len, mpe->option);			\
 }
 
-declare_def_handler(checkint, set_int)
+static int checkint_handler(struct config *conf, vector strvec)
+{
+	int rc = set_int(strvec, &conf->checkint);
+
+	if (rc)
+		return rc;
+	if (conf->checkint == CHECKINT_UNDEF)
+		conf->checkint--;
+	return 0;
+}
+
 declare_def_snprint(checkint, print_int)
 
 declare_def_handler(max_checkint, set_int)
@@ -1563,7 +1573,7 @@ init_keywords(vector keywords)
 {
 	install_keyword_root("defaults", NULL);
 	install_keyword("verbosity", &def_verbosity_handler, &snprint_def_verbosity);
-	install_keyword("polling_interval", &def_checkint_handler, &snprint_def_checkint);
+	install_keyword("polling_interval", &checkint_handler, &snprint_def_checkint);
 	install_keyword("max_polling_interval", &def_max_checkint_handler, &snprint_def_max_checkint);
 	install_keyword("reassign_maps", &def_reassign_maps_handler, &snprint_def_reassign_maps);
 	install_keyword("multipath_dir", &def_multipath_dir_handler, &snprint_def_multipath_dir);
libmultipath/structs.c +1 -0
index caa178a6..fee899bd 100644
@@ -100,6 +100,7 @@ alloc_path (void)
 		pp->fd = -1;
 		pp->tpgs = TPGS_UNDEF;
 		pp->priority = PRIO_UNDEF;
+		pp->checkint = CHECKINT_UNDEF;
 		checker_clear(&pp->checker);
 		dm_path_to_gen(pp)->ops = &dm_gen_path_ops;
 		pp->hwe = vector_alloc();
multipathd/main.c +6 -0
index 3464ebfa..083abf28 100644
@@ -1840,6 +1840,12 @@ check_path (struct vectors * vecs, struct path * pp, int ticks)
 	max_checkint = conf->max_checkint;
 	verbosity = conf->verbosity;
 	put_multipath_config(conf);
+
+	if (pp->checkint == CHECKINT_UNDEF) {
+		condlog(0, "%s: BUG: checkint is not set", pp->dev);
+		pp->checkint = checkint;
+	};
+
 	if (!pp->mpp && pp->initialized == INIT_MISSING_UDEV) {
 		if (pp->retriggers < retrigger_tries) {
 			condlog(2, "%s: triggering change event to reinitialize",
Signature · 1 line
2.19.1
maintainer Reviewed-by on v1 3/7
#6 lore raw

Re: [PATCH 3/7] multipathd: print error message if checkint is not initialized

Quoted 4 lines · Martin Wilck

This is just a safety measure in case I overlooked something wrt the checkint initialization. It could be reverted once we know the error message isn't triggered.

Quoted 83 lines
---
 libmultipath/defaults.h |  1 +
 libmultipath/dict.c     | 14 ++++++++++++--
 libmultipath/structs.c  |  1 +
 multipathd/main.c       |  6 ++++++
 4 files changed, 20 insertions(+), 2 deletions(-)
libmultipath/defaults.h +1 -0
index 7f3839fc..65769398 100644
@@ -46,6 +46,7 @@
 #define DEFAULT_UNKNOWN_FIND_MULTIPATHS_TIMEOUT 1
 #define DEFAULT_ALL_TG_PT ALL_TG_PT_OFF
 
+#define CHECKINT_UNDEF		(~0U)
 #define DEFAULT_CHECKINT	5
 #define MAX_CHECKINT(a)		(a << 2)
 
libmultipath/dict.c +12 -2
index c3f5a6e6..a81c051f 100644
@@ -264,7 +264,17 @@ snprint_mp_ ## option (struct config *conf, char * buff, int len,	\
 	return function (buff, len, mpe->option);			\
 }
 
-declare_def_handler(checkint, set_int)
+static int checkint_handler(struct config *conf, vector strvec)
+{
+	int rc = set_int(strvec, &conf->checkint);
+
+	if (rc)
+		return rc;
+	if (conf->checkint == CHECKINT_UNDEF)
+		conf->checkint--;
+	return 0;
+}
+
 declare_def_snprint(checkint, print_int)
 
 declare_def_handler(max_checkint, set_int)
@@ -1563,7 +1573,7 @@ init_keywords(vector keywords)
 {
 	install_keyword_root("defaults", NULL);
 	install_keyword("verbosity", &def_verbosity_handler, &snprint_def_verbosity);
-	install_keyword("polling_interval", &def_checkint_handler, &snprint_def_checkint);
+	install_keyword("polling_interval", &checkint_handler, &snprint_def_checkint);
 	install_keyword("max_polling_interval", &def_max_checkint_handler, &snprint_def_max_checkint);
 	install_keyword("reassign_maps", &def_reassign_maps_handler, &snprint_def_reassign_maps);
 	install_keyword("multipath_dir", &def_multipath_dir_handler, &snprint_def_multipath_dir);
libmultipath/structs.c +1 -0
index caa178a6..fee899bd 100644
@@ -100,6 +100,7 @@ alloc_path (void)
 		pp->fd = -1;
 		pp->tpgs = TPGS_UNDEF;
 		pp->priority = PRIO_UNDEF;
+		pp->checkint = CHECKINT_UNDEF;
 		checker_clear(&pp->checker);
 		dm_path_to_gen(pp)->ops = &dm_gen_path_ops;
 		pp->hwe = vector_alloc();
multipathd/main.c +6 -0
index 3464ebfa..083abf28 100644
@@ -1840,6 +1840,12 @@ check_path (struct vectors * vecs, struct path * pp, int ticks)
 	max_checkint = conf->max_checkint;
 	verbosity = conf->verbosity;
 	put_multipath_config(conf);
+
+	if (pp->checkint == CHECKINT_UNDEF) {
+		condlog(0, "%s: BUG: checkint is not set", pp->dev);
+		pp->checkint = checkint;
+	};
+
 	if (!pp->mpp && pp->initialized == INIT_MISSING_UDEV) {
 		if (pp->retriggers < retrigger_tries) {
 			condlog(2, "%s: triggering change event to reinitialize",
Signature · 1 line
2.19.1
#7 lore raw

PATCH 4/7 multipathd: open client socket early

In reply to Martin Wilck

Open the unix socket in multipathd code and pass the fd to uxsock_listen(). This will enable us to make the main thread wait for successful socket initialization in a follow-up patch.

---
 multipathd/main.c   | 17 ++++++++++++++---
 multipathd/uxlsnr.c | 11 ++---------
 multipathd/uxlsnr.h |  2 +-
 3 files changed, 17 insertions(+), 13 deletions(-)
multipathd/main.c +14 -3
index 083abf28..50f69171 100644
@@ -66,6 +66,7 @@ static int use_watchdog;
 #include "pgpolicies.h"
 #include "uevent.h"
 #include "log.h"
+#include "uxsock.h"
 
 #include "mpath_cmd.h"
 #include "mpath_persist.h"
@@ -1494,12 +1495,22 @@ uevqloop (void * ap)
 static void *
 uxlsnrloop (void * ap)
 {
+	int ux_sock;
+
+	pthread_cleanup_push(rcu_unregister, NULL);
+	rcu_register_thread();
+
+	ux_sock = ux_socket_listen(DEFAULT_SOCKET);
+	if (ux_sock == -1)  {
+		condlog(1, "could not create uxsock: %d", errno);
+		exit_daemon();
+		return NULL;
+	}
+
 	if (cli_init()) {
 		condlog(1, "Failed to init uxsock listener");
 		return NULL;
 	}
-	pthread_cleanup_push(rcu_unregister, NULL);
-	rcu_register_thread();
 	set_handler_callback(LIST+PATHS, cli_list_paths);
 	set_handler_callback(LIST+PATHS+FMT, cli_list_paths_fmt);
 	set_handler_callback(LIST+PATHS+RAW+FMT, cli_list_paths_raw);
@@ -1554,7 +1565,7 @@ uxlsnrloop (void * ap)
 	set_handler_callback(UNSETPRKEY+MAP, cli_unsetprkey);
 
 	umask(077);
-	uxsock_listen(&uxsock_trigger, ap);
+	uxsock_listen(&uxsock_trigger, ux_sock, ap);
 	pthread_cleanup_pop(1);
 	return NULL;
 }
multipathd/uxlsnr.c +2 -9
index 6f666663..04967e9a 100644
@@ -165,22 +165,15 @@ void uxsock_cleanup(void *arg)
 /*
  * entry point
  */
-void * uxsock_listen(uxsock_trigger_fn uxsock_trigger, void * trigger_data)
+void * uxsock_listen(uxsock_trigger_fn uxsock_trigger, long ux_sock,
+		     void * trigger_data)
 {
-	long ux_sock;
 	int rlen;
 	char *inbuf;
 	char *reply;
 	sigset_t mask;
 	int old_clients = MIN_POLLS;
 
-	ux_sock = ux_socket_listen(DEFAULT_SOCKET);
-
-	if (ux_sock == -1) {
-		condlog(1, "could not create uxsock: %d", errno);
-		exit_daemon();
-	}
-
 	pthread_cleanup_push(uxsock_cleanup, (void *)ux_sock);
 
 	condlog(3, "uxsock: startup listener");
multipathd/uxlsnr.h +1 -1
index d51a8f99..c7f25b2b 100644
@@ -5,7 +5,7 @@
 
 typedef int (uxsock_trigger_fn)(char *, char **, int *, bool, void *);
 
-void * uxsock_listen(uxsock_trigger_fn uxsock_trigger,
+void * uxsock_listen(uxsock_trigger_fn uxsock_trigger, long ux_sock,
 		     void * trigger_data);
 
 #endif
Signature · 1 line
2.19.1
#8 lore raw

Re: [PATCH 4/7] multipathd: open client socket early

Quoted 26 lines · Martin Wilck

Open the unix socket in multipathd code and pass the fd to uxsock_listen(). This will enable us to make the main thread wait for successful socket initialization in a follow-up patch.

---
 multipathd/main.c   | 17 ++++++++++++++---
 multipathd/uxlsnr.c | 11 ++---------
 multipathd/uxlsnr.h |  2 +-
 3 files changed, 17 insertions(+), 13 deletions(-)
multipathd/main.c +1 -0
index 083abf28..50f69171 100644
@@ -66,6 +66,7 @@ static int use_watchdog;
 #include "pgpolicies.h"
 #include "uevent.h"
 #include "log.h"
+#include "uxsock.h"
 
 #include "mpath_cmd.h"
 #include "mpath_persist.h"
@@ -1494,12 +1495,22 @@ uevqloop (void * ap)
 static void *
 uxlsnrloop (void * ap)

multipathd/main.c uevqloop (void * ap) · line 1506

Quoted 10 lines
multipathd/main.c +9 -0
 {
+	int ux_sock;
+
+	pthread_cleanup_push(rcu_unregister, NULL);
+	rcu_register_thread();
+
+	ux_sock = ux_socket_listen(DEFAULT_SOCKET);
+	if (ux_sock == -1)  {
+		condlog(1, "could not create uxsock: %d", errno);
+		exit_daemon();

I'm pretty sure you want to goto the end of the function, where you pthread_cleanup_pop(), instead of returning NULL right here. According the to the pthread_cleanup_push() man page "Clean-up handlers are not called if the thread terminates by performing a return from the thread start function."

Quoted 65 lines
multipathd/main.c +4 -3
+		return NULL;
+	}
+
 	if (cli_init()) {
 		condlog(1, "Failed to init uxsock listener");
 		return NULL;
 	}
-	pthread_cleanup_push(rcu_unregister, NULL);
-	rcu_register_thread();
 	set_handler_callback(LIST+PATHS, cli_list_paths);
 	set_handler_callback(LIST+PATHS+FMT, cli_list_paths_fmt);
 	set_handler_callback(LIST+PATHS+RAW+FMT, cli_list_paths_raw);
@@ -1554,7 +1565,7 @@ uxlsnrloop (void * ap)
 	set_handler_callback(UNSETPRKEY+MAP, cli_unsetprkey);
 
 	umask(077);
-	uxsock_listen(&uxsock_trigger, ap);
+	uxsock_listen(&uxsock_trigger, ux_sock, ap);
 	pthread_cleanup_pop(1);
 	return NULL;
 }
multipathd/uxlsnr.c +2 -9
index 6f666663..04967e9a 100644
@@ -165,22 +165,15 @@ void uxsock_cleanup(void *arg)
 /*
  * entry point
  */
-void * uxsock_listen(uxsock_trigger_fn uxsock_trigger, void * trigger_data)
+void * uxsock_listen(uxsock_trigger_fn uxsock_trigger, long ux_sock,
+		     void * trigger_data)
 {
-	long ux_sock;
 	int rlen;
 	char *inbuf;
 	char *reply;
 	sigset_t mask;
 	int old_clients = MIN_POLLS;
 
-	ux_sock = ux_socket_listen(DEFAULT_SOCKET);
-
-	if (ux_sock == -1) {
-		condlog(1, "could not create uxsock: %d", errno);
-		exit_daemon();
-	}
-
 	pthread_cleanup_push(uxsock_cleanup, (void *)ux_sock);
 
 	condlog(3, "uxsock: startup listener");
multipathd/uxlsnr.h +1 -1
index d51a8f99..c7f25b2b 100644
@@ -5,7 +5,7 @@
 
 typedef int (uxsock_trigger_fn)(char *, char **, int *, bool, void *);
 
-void * uxsock_listen(uxsock_trigger_fn uxsock_trigger,
+void * uxsock_listen(uxsock_trigger_fn uxsock_trigger, long ux_sock,
 		     void * trigger_data);
 
 #endif
Signature · 1 line
2.19.1
#9 lore raw

PATCH 5/7 multipathd: set DAEMON_CONFIGURE from uxlsnr thread

In reply to Martin Wilck

Commit ee01e841 had the intention to make multipathd quit if the client socket couldn't be set up, because the unix socket listener is vital for signal handling in multipathd. But during startup, this condition might be lost if the main thread doesn't wait for the unix listener to initialize.

Fixes ee01e841 "multipathd: handle errors in uxlsnr as fatal"
---
 multipathd/main.c | 43 ++++++++++++++++++++++++++++++++-----------
 1 file changed, 32 insertions(+), 11 deletions(-)
multipathd/main.c +32 -11
index 50f69171..c71e7d03 100644
@@ -205,9 +205,8 @@ static void config_cleanup(void *arg)
 	pthread_mutex_unlock(&config_lock);
 }
 
-void post_config_state(enum daemon_status state)
+static void __post_config_state(enum daemon_status state)
 {
-	pthread_mutex_lock(&config_lock);
 	if (state != running_state) {
 		enum daemon_status old_state = running_state;
 
@@ -217,7 +216,14 @@ void post_config_state(enum daemon_status state)
 		do_sd_notify(old_state);
 #endif
 	}
-	pthread_mutex_unlock(&config_lock);
+}
+
+void post_config_state(enum daemon_status state)
+{
+	pthread_mutex_lock(&config_lock);
+	pthread_cleanup_push(config_cleanup, NULL);
+	__post_config_state(state);
+	pthread_cleanup_pop(1);
 }
 
 int set_config_state(enum daemon_status state)
@@ -1511,6 +1517,10 @@ uxlsnrloop (void * ap)
 		condlog(1, "Failed to init uxsock listener");
 		return NULL;
 	}
+
+	/* Tell main thread that thread has started */
+	post_config_state(DAEMON_CONFIGURE);
+
 	set_handler_callback(LIST+PATHS, cli_list_paths);
 	set_handler_callback(LIST+PATHS+FMT, cli_list_paths_fmt);
 	set_handler_callback(LIST+PATHS+RAW+FMT, cli_list_paths_raw);
@@ -2726,11 +2736,26 @@ child (void * param)
 	 */
 	conf = NULL;
 
-	/*
-	 * Signal start of configuration
-	 */
-	post_config_state(DAEMON_CONFIGURE);
+	pthread_cleanup_push(config_cleanup, NULL);
+	pthread_mutex_lock(&config_lock);
 
+	__post_config_state(DAEMON_IDLE);
+	rc = pthread_create(&uxlsnr_thr, &misc_attr, uxlsnrloop, vecs);
+	if (!rc) {
+		/* Wait for uxlsnr startup */
+		while (running_state == DAEMON_IDLE)
+			pthread_cond_wait(&config_cond, &config_lock);
+	}
+	pthread_cleanup_pop(1);
+
+	if (rc) {
+		condlog(0, "failed to create cli listener: %d", rc);
+		goto failed;
+	}
+	else if (running_state != DAEMON_CONFIGURE) {
+		condlog(0, "cli listener failed to start");
+		goto failed;
+	}
 
 	if (poll_dmevents) {
 		if (init_dmevent_waiter(vecs)) {
@@ -2753,10 +2778,6 @@ child (void * param)
 		goto failed;
 	}
 	pthread_attr_destroy(&uevent_attr);
-	if ((rc = pthread_create(&uxlsnr_thr, &misc_attr, uxlsnrloop, vecs))) {
-		condlog(0, "failed to create cli listener: %d", rc);
-		goto failed;
-	}
 
 	/*
 	 * start threads
Signature · 1 line
2.19.1
maintainer Reviewed-by on v1 5/7
#10 lore raw

Re: [PATCH 5/7] multipathd: set DAEMON_CONFIGURE from uxlsnr thread

Quoted 56 lines · Martin Wilck

Commit ee01e841 had the intention to make multipathd quit if the client socket couldn't be set up, because the unix socket listener is vital for signal handling in multipathd. But during startup, this condition might be lost if the main thread doesn't wait for the unix listener to initialize.

Fixes ee01e841 "multipathd: handle errors in uxlsnr as fatal"
---
 multipathd/main.c | 43 ++++++++++++++++++++++++++++++++-----------
 1 file changed, 32 insertions(+), 11 deletions(-)
multipathd/main.c +13 -3
index 50f69171..c71e7d03 100644
@@ -205,9 +205,8 @@ static void config_cleanup(void *arg)
 	pthread_mutex_unlock(&config_lock);
 }
 
-void post_config_state(enum daemon_status state)
+static void __post_config_state(enum daemon_status state)
 {
-	pthread_mutex_lock(&config_lock);
 	if (state != running_state) {
 		enum daemon_status old_state = running_state;
 
@@ -217,7 +216,14 @@ void post_config_state(enum daemon_status state)
 		do_sd_notify(old_state);
 #endif
 	}
-	pthread_mutex_unlock(&config_lock);
+}
+
+void post_config_state(enum daemon_status state)
+{
+	pthread_mutex_lock(&config_lock);
+	pthread_cleanup_push(config_cleanup, NULL);
+	__post_config_state(state);
+	pthread_cleanup_pop(1);
 }
 
 int set_config_state(enum daemon_status state)
@@ -1511,6 +1517,10 @@ uxlsnrloop (void * ap)
 		condlog(1, "Failed to init uxsock listener");
 		return NULL;
 	}
+
+	/* Tell main thread that thread has started */
+	post_config_state(DAEMON_CONFIGURE);
+
 	set_handler_callback(LIST+PATHS, cli_list_paths);
 	set_handler_callback(LIST+PATHS+FMT, cli_list_paths_fmt);
 	set_handler_callback(LIST+PATHS+RAW+FMT, cli_list_paths_raw);
@@ -2726,11 +2736,26 @@ child (void * param)
 	 */

multipathd/main.c child (void * param) · line 2742

Quoted 10 lines
multipathd/main.c +3 -4
 	conf = NULL;
 
-	/*
-	 * Signal start of configuration
-	 */
-	post_config_state(DAEMON_CONFIGURE);
+	pthread_cleanup_push(config_cleanup, NULL);
+	pthread_mutex_lock(&config_lock);
 
+	__post_config_state(DAEMON_IDLE);

I don't really understand the need for __post_config_state(). Couldn't you just call post_config_state(DAEMON_IDLE), and then grab the config_lock after checking that the thread was created successfully. If the scheduler decides to run the uxlsnrloop thread before continuing with the child thread, it would be better if uxlsnrloop thread was free to grab the config lock anyway.

At any rate, what you have is perfectly correct, so

Quoted 32 lines
multipathd/main.c +16 -4
+	rc = pthread_create(&uxlsnr_thr, &misc_attr, uxlsnrloop, vecs);
+	if (!rc) {
+		/* Wait for uxlsnr startup */
+		while (running_state == DAEMON_IDLE)
+			pthread_cond_wait(&config_cond, &config_lock);
+	}
+	pthread_cleanup_pop(1);
+
+	if (rc) {
+		condlog(0, "failed to create cli listener: %d", rc);
+		goto failed;
+	}
+	else if (running_state != DAEMON_CONFIGURE) {
+		condlog(0, "cli listener failed to start");
+		goto failed;
+	}
 
 	if (poll_dmevents) {
 		if (init_dmevent_waiter(vecs)) {
@@ -2753,10 +2778,6 @@ child (void * param)
 		goto failed;
 	}
 	pthread_attr_destroy(&uevent_attr);
-	if ((rc = pthread_create(&uxlsnr_thr, &misc_attr, uxlsnrloop, vecs))) {
-		condlog(0, "failed to create cli listener: %d", rc);
-		goto failed;
-	}
 
 	/*
 	 * start threads
Signature · 1 line
2.19.1
#11 lore raw

Re: [PATCH 5/7] multipathd: set DAEMON_CONFIGURE from uxlsnr thread

Quoted 28 lines · Benjamin Marzinski
Quoted 27 lines · Martin Wilck

Commit ee01e841 had the intention to make multipathd quit if the client socket couldn't be set up, because the unix socket listener is vital for signal handling in multipathd. But during startup, this condition might be lost if the main thread doesn't wait for the unix listener to initialize.

Fixes ee01e841 "multipathd: handle errors in uxlsnr as fatal"
---
 multipathd/main.c | 43 ++++++++++++++++++++++++++++++++-----------
 1 file changed, 32 insertions(+), 11 deletions(-)
multipathd/main.c +2 -4
index 50f69171..c71e7d03 100644
@@ -2726,11 +2736,26 @@ child (void * param)
 	 */
 	conf = NULL;
 
-	/*
-	 * Signal start of configuration
-	 */
-	post_config_state(DAEMON_CONFIGURE);
+	pthread_cleanup_push(config_cleanup, NULL);
+	pthread_mutex_lock(&config_lock);
 
Quoted 11 lines

multipathd/main.c child (void * param) · line 2742

Quoted 1 line
multipathd/main.c +1 -0
+	__post_config_state(DAEMON_IDLE);

I don't really understand the need for __post_config_state(). Couldn't
you just call post_config_state(DAEMON_IDLE), and then grab the config_lock after checking that the thread was created successfully. If
the scheduler decides to run the uxlsnrloop thread before continuing with the child thread, it would be better if uxlsnrloop thread was free
to grab the config lock anyway.

I wanted to enforce ordering here. The uxlsnr won't be able to grab the lock and change running_state until the child enters pthread_cond_wait() below, which is intentional. running_state shouldn't be read or written without holding the lock, so this actually saves us an unlock/lock operation, as child() would otherwise need to release the lock before calling pthread_create(), and reacquire it afterwards. This way the startup sequence is fully deterministic and doesn't depend on scheduler decisions. I like it that way :-)

Quoted 3 lines

At any rate, what you have is perfectly correct, so

Thanks a lot.

Martin

Quoted 34 lines
Quoted 34 lines
multipathd/main.c +16 -4
+	rc = pthread_create(&uxlsnr_thr, &misc_attr, uxlsnrloop, vecs);
+	if (!rc) {
+		/* Wait for uxlsnr startup */
+		while (running_state == DAEMON_IDLE)
+			pthread_cond_wait(&config_cond, &config_lock);
+	}
+	pthread_cleanup_pop(1);
+
+	if (rc) {
+		condlog(0, "failed to create cli listener: %d", rc);
+		goto failed;
+	}
+	else if (running_state != DAEMON_CONFIGURE) {
+		condlog(0, "cli listener failed to start");
+		goto failed;
+	}
 
 	if (poll_dmevents) {
 		if (init_dmevent_waiter(vecs)) {
@@ -2753,10 +2778,6 @@ child (void * param)
 		goto failed;
 	}
 	pthread_attr_destroy(&uevent_attr);
-	if ((rc = pthread_create(&uxlsnr_thr, &misc_attr, uxlsnrloop,
 vecs))) {
-		condlog(0, "failed to create cli listener: %d", rc);
-		goto failed;
-	}
 
 	/*

* start threads

Signature · 1 line
2.19.1
Signature · 9 lines
Dr. Martin Wilck <[email protected]>, Tel. +49 (0)911 74053 2107
SUSE Linux GmbH, GF: Felix Imendörffer, Jane Smithard, Graham Norton
HRB 21284 (AG Nürnberg)


--
dm-devel mailing list
[email protected]
https://www.redhat.com/mailman/listinfo/dm-devel
#12 lore raw

PATCH 6/7 multipathd: make DAEMON_SHUTDOWN a terminal state

In reply to Martin Wilck

It can happen that, before the main thread reacts on DAEMON_SHUTDOWN and starts cancelling threads, another thread resets the state to something else. Fix that.

---
 multipathd/cli_handlers.c |  9 +++++++--
 multipathd/main.c         | 10 +++++++---
 2 files changed, 14 insertions(+), 5 deletions(-)
multipathd/cli_handlers.c +7 -2
index 4aea4ce7..a0d57a53 100644
@@ -1124,12 +1124,17 @@ cli_switch_group(void * v, char ** reply, int * len, void * data)
 int
 cli_reconfigure(void * v, char ** reply, int * len, void * data)
 {
+	int rc;
+
 	condlog(2, "reconfigure (operator)");
 
-	if (set_config_state(DAEMON_CONFIGURE) == ETIMEDOUT) {
+	rc = set_config_state(DAEMON_CONFIGURE); 
+	if (rc == ETIMEDOUT) {
 		condlog(2, "timeout starting reconfiguration");
 		return 1;
-	}
+	} else if (rc == EINVAL)
+		/* daemon shutting down */
+		return 1;
 	return 0;
 }
 
multipathd/main.c +7 -3
index c71e7d03..768da8da 100644
@@ -207,7 +207,7 @@ static void config_cleanup(void *arg)
 
 static void __post_config_state(enum daemon_status state)
 {
-	if (state != running_state) {
+	if (state != running_state && running_state != DAEMON_SHUTDOWN) {
 		enum daemon_status old_state = running_state;
 
 		running_state = state;
@@ -235,7 +235,9 @@ int set_config_state(enum daemon_status state)
 	if (running_state != state) {
 		enum daemon_status old_state = running_state;
 
-		if (running_state != DAEMON_IDLE) {
+		if (running_state == DAEMON_SHUTDOWN)
+			rc = EINVAL;
+		else if (running_state != DAEMON_IDLE) {
 			struct timespec ts;
 
 			clock_gettime(CLOCK_MONOTONIC, &ts);
@@ -2204,7 +2206,9 @@ checkerloop (void *ap)
 		if (rc == ETIMEDOUT) {
 			condlog(4, "timeout waiting for DAEMON_IDLE");
 			continue;
-		}
+		} else if (rc == EINVAL)
+			/* daemon shutdown */
+			break;
 
 		pthread_cleanup_push(cleanup_lock, &vecs->lock);
 		lock(&vecs->lock);
Signature · 1 line
2.19.1
maintainer Reviewed-by on v1 6/7
#13 lore raw

Re: [PATCH 6/7] multipathd: make DAEMON_SHUTDOWN a terminal state

Quoted 4 lines · Martin Wilck

It can happen that, before the main thread reacts on DAEMON_SHUTDOWN and starts cancelling threads, another thread resets the state to something else. Fix that.

Quoted 67 lines
---
 multipathd/cli_handlers.c |  9 +++++++--
 multipathd/main.c         | 10 +++++++---
 2 files changed, 14 insertions(+), 5 deletions(-)
multipathd/cli_handlers.c +7 -2
index 4aea4ce7..a0d57a53 100644
@@ -1124,12 +1124,17 @@ cli_switch_group(void * v, char ** reply, int * len, void * data)
 int
 cli_reconfigure(void * v, char ** reply, int * len, void * data)
 {
+	int rc;
+
 	condlog(2, "reconfigure (operator)");
 
-	if (set_config_state(DAEMON_CONFIGURE) == ETIMEDOUT) {
+	rc = set_config_state(DAEMON_CONFIGURE); 
+	if (rc == ETIMEDOUT) {
 		condlog(2, "timeout starting reconfiguration");
 		return 1;
-	}
+	} else if (rc == EINVAL)
+		/* daemon shutting down */
+		return 1;
 	return 0;
 }
 
multipathd/main.c +7 -3
index c71e7d03..768da8da 100644
@@ -207,7 +207,7 @@ static void config_cleanup(void *arg)
 
 static void __post_config_state(enum daemon_status state)
 {
-	if (state != running_state) {
+	if (state != running_state && running_state != DAEMON_SHUTDOWN) {
 		enum daemon_status old_state = running_state;
 
 		running_state = state;
@@ -235,7 +235,9 @@ int set_config_state(enum daemon_status state)
 	if (running_state != state) {
 		enum daemon_status old_state = running_state;
 
-		if (running_state != DAEMON_IDLE) {
+		if (running_state == DAEMON_SHUTDOWN)
+			rc = EINVAL;
+		else if (running_state != DAEMON_IDLE) {
 			struct timespec ts;
 
 			clock_gettime(CLOCK_MONOTONIC, &ts);
@@ -2204,7 +2206,9 @@ checkerloop (void *ap)
 		if (rc == ETIMEDOUT) {
 			condlog(4, "timeout waiting for DAEMON_IDLE");
 			continue;
-		}
+		} else if (rc == EINVAL)
+			/* daemon shutdown */
+			break;
 
 		pthread_cleanup_push(cleanup_lock, &vecs->lock);
 		lock(&vecs->lock);
Signature · 1 line
2.19.1
#14 lore raw

PATCH 7/7 multipathd: only grab conf once for filter_path()

In reply to Martin Wilck

This saves a possibly large number of cleanup push/pop calls and slightly improves readability.

---
 multipathd/main.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
multipathd/main.c +4 -3
index 768da8da..a7cd3b09 100644
@@ -2332,16 +2332,17 @@ configure (struct vectors * vecs)
 		goto fail;
 	}
 
+	conf = get_multipath_config();
+	pthread_cleanup_push(put_multipath_config, conf);
 	vector_foreach_slot (vecs->pathvec, pp, i){
-		conf = get_multipath_config();
-		pthread_cleanup_push(put_multipath_config, conf);
 		if (filter_path(conf, pp) > 0){
 			vector_del_slot(vecs->pathvec, i);
 			free_path(pp);
 			i--;
 		}
-		pthread_cleanup_pop(1);
 	}
+	pthread_cleanup_pop(1);
+
 	if (map_discovery(vecs)) {
 		condlog(0, "configure failed at map discovery");
 		goto fail;
Signature · 1 line
2.19.1
maintainer Reviewed-by on v1 7/7
#15 lore raw

Re: [PATCH 7/7] multipathd: only grab conf once for filter_path()

Quoted 3 lines · Martin Wilck

This saves a possibly large number of cleanup push/pop calls and slightly improves readability.

Quoted 32 lines
---
 multipathd/main.c | 7 ++++---
 1 file changed, 4 insertions(+), 3 deletions(-)
multipathd/main.c +4 -3
index 768da8da..a7cd3b09 100644
@@ -2332,16 +2332,17 @@ configure (struct vectors * vecs)
 		goto fail;
 	}
 
+	conf = get_multipath_config();
+	pthread_cleanup_push(put_multipath_config, conf);
 	vector_foreach_slot (vecs->pathvec, pp, i){
-		conf = get_multipath_config();
-		pthread_cleanup_push(put_multipath_config, conf);
 		if (filter_path(conf, pp) > 0){
 			vector_del_slot(vecs->pathvec, i);
 			free_path(pp);
 			i--;
 		}
-		pthread_cleanup_pop(1);
 	}
+	pthread_cleanup_pop(1);
+
 	if (map_discovery(vecs)) {
 		condlog(0, "configure failed at map discovery");
 		goto fail;
Signature · 1 line
2.19.1