X-Git-Url: http://git.lttng.org./?a=blobdiff_plain;f=liblttng-ust%2Flttng-ust-comm.c;h=256cbdc625ef19f6da306124542a6e0a08ab593f;hb=37dddb65504eff070a64fb4a8f1c56ee81c3173c;hp=5a4b6bd8693ed5b427f8baf693a903949d025547;hpb=32ce85691c17b331072b1c0df96f69e8b388d134;p=lttng-ust.git

diff --git a/liblttng-ust/lttng-ust-comm.c b/liblttng-ust/lttng-ust-comm.c
index 5a4b6bd8..256cbdc6 100644
--- a/liblttng-ust/lttng-ust-comm.c
+++ b/liblttng-ust/lttng-ust-comm.c
@@ -34,6 +34,7 @@
 #include <time.h>
 #include <assert.h>
 #include <signal.h>
+#include <dlfcn.h>
 #include <urcu/uatomic.h>
 #include <urcu/futex.h>
 #include <urcu/compiler.h>
@@ -51,6 +52,7 @@
 #include "lttng-tracer-core.h"
 #include "compat.h"
 #include "../libringbuffer/tlsfixup.h"
+#include "lttng-ust-baddr.h"
 
 /*
  * Has lttng ust comm constructor been called ?
@@ -106,6 +108,8 @@ struct sock_info {
 
 	char wait_shm_path[PATH_MAX];
 	char *wait_shm_mmap;
+	/* Keep track of lazy state dump not performed yet. */
+	int statedump_pending;
 };
 
 /* Socket from app (connect) to session daemon (listen) for communication */
@@ -122,6 +126,8 @@ struct sock_info global_apps = {
 	.notify_socket = -1,
 
 	.wait_shm_path = "/" LTTNG_UST_WAIT_FILENAME,
+
+	.statedump_pending = 0,
 };
 
 /* TODO: allow global_apps_sock_path override */
@@ -135,6 +141,8 @@ struct sock_info local_apps = {
 
 	.socket = -1,
 	.notify_socket = -1,
+
+	.statedump_pending = 0,
 };
 
 static int wait_poll_fallback;
@@ -172,15 +180,39 @@ static const char *cmd_name_mapping[] = {
 
 	/* Event FD commands */
 	[ LTTNG_UST_FILTER ] = "Create Filter",
+	[ LTTNG_UST_EXCLUSION ] = "Add exclusions to event",
 };
 
+static const char *str_timeout;
+static int got_timeout_env;
+
 extern void lttng_ring_buffer_client_overwrite_init(void);
+extern void lttng_ring_buffer_client_overwrite_rt_init(void);
 extern void lttng_ring_buffer_client_discard_init(void);
+extern void lttng_ring_buffer_client_discard_rt_init(void);
 extern void lttng_ring_buffer_metadata_client_init(void);
 extern void lttng_ring_buffer_client_overwrite_exit(void);
+extern void lttng_ring_buffer_client_overwrite_rt_exit(void);
 extern void lttng_ring_buffer_client_discard_exit(void);
+extern void lttng_ring_buffer_client_discard_rt_exit(void);
 extern void lttng_ring_buffer_metadata_client_exit(void);
 
+/*
+ * Returns the HOME directory path. Caller MUST NOT free(3) the returned
+ * pointer.
+ */
+static
+const char *get_lttng_home_dir(void)
+{
+       const char *val;
+
+       val = (const char *) getenv("LTTNG_HOME");
+       if (val != NULL) {
+               return val;
+       }
+       return (const char *) getenv("HOME");
+}
+
 /*
  * Force a read (imply TLS fixup for dlopen) of TLS variables.
  */
@@ -202,10 +234,12 @@ void print_cmd(int cmd, int handle)
 {
 	const char *cmd_name = "Unknown";
 
-	if (cmd_name_mapping[cmd]) {
+	if (cmd >= 0 && cmd < LTTNG_ARRAY_SIZE(cmd_name_mapping)
+			&& cmd_name_mapping[cmd]) {
 		cmd_name = cmd_name_mapping[cmd];
 	}
-	DBG("Message Received \"%s\", Handle \"%s\" (%d)", cmd_name,
+	DBG("Message Received \"%s\" (%d), Handle \"%s\" (%d)",
+		cmd_name, cmd,
 		lttng_ust_obj_get_name(handle), handle);
 }
 
@@ -223,7 +257,7 @@ int setup_local_apps(void)
 		assert(local_apps.allowed == 0);
 		return 0;
 	}
-	home_dir = (const char *) getenv("HOME");
+	home_dir = get_lttng_home_dir();
 	if (!home_dir) {
 		WARN("HOME environment variable not set. Disabling LTTng-UST per-user tracing.");
 		assert(local_apps.allowed == 0);
@@ -240,6 +274,66 @@ int setup_local_apps(void)
 	return 0;
 }
 
+/*
+ * Get notify_sock timeout, in ms.
+ * -1: don't wait. 0: wait forever. >0: timeout, in ms.
+ */
+static
+long get_timeout(void)
+{
+	long constructor_delay_ms = LTTNG_UST_DEFAULT_CONSTRUCTOR_TIMEOUT_MS;
+
+	if (!got_timeout_env) {
+		str_timeout = getenv("LTTNG_UST_REGISTER_TIMEOUT");
+		got_timeout_env = 1;
+	}
+	if (str_timeout)
+		constructor_delay_ms = strtol(str_timeout, NULL, 10);
+	return constructor_delay_ms;
+}
+
+static
+long get_notify_sock_timeout(void)
+{
+	return get_timeout();
+}
+
+/*
+ * Return values: -1: don't wait. 0: wait forever. 1: timeout wait.
+ */
+static
+int get_constructor_timeout(struct timespec *constructor_timeout)
+{
+	long constructor_delay_ms;
+	int ret;
+
+	constructor_delay_ms = get_timeout();
+
+	switch (constructor_delay_ms) {
+	case -1:/* fall-through */
+	case 0:
+		return constructor_delay_ms;
+	default:
+		break;
+	}
+
+	/*
+	 * If we are unable to find the current time, don't wait.
+	 */
+	ret = clock_gettime(CLOCK_REALTIME, constructor_timeout);
+	if (ret) {
+		return -1;
+	}
+	constructor_timeout->tv_sec += constructor_delay_ms / 1000UL;
+	constructor_timeout->tv_nsec +=
+		(constructor_delay_ms % 1000UL) * 1000000UL;
+	if (constructor_timeout->tv_nsec >= 1000000000UL) {
+		constructor_timeout->tv_sec++;
+		constructor_timeout->tv_nsec -= 1000000000UL;
+	}
+	return 1;
+}
+
 static
 int register_to_sessiond(int socket, enum ustctl_socket_type type)
 {
@@ -294,6 +388,28 @@ int handle_register_done(struct sock_info *sock_info)
 	return 0;
 }
 
+/*
+ * Only execute pending statedump after the constructor semaphore has
+ * been posted by each listener thread. This means statedump will only
+ * be performed after the "registration done" command is received from
+ * each session daemon the application is connected to.
+ *
+ * This ensures we don't run into deadlock issues with the dynamic
+ * loader mutex, which is held while the constructor is called and
+ * waiting on the constructor semaphore. All operations requiring this
+ * dynamic loader lock need to be postponed using this mechanism.
+ */
+static
+void handle_pending_statedump(struct sock_info *sock_info)
+{
+	int ctor_passed = sock_info->constructor_sem_posted;
+
+	if (ctor_passed && sock_info->statedump_pending) {
+		sock_info->statedump_pending = 0;
+		lttng_handle_pending_statedump(sock_info);
+	}
+}
+
 static
 int handle_message(struct sock_info *sock_info,
 		int sock, struct ustcomm_ust_msg *lum)
@@ -330,7 +446,7 @@ int handle_message(struct sock_info *sock_info,
 		if (lum->handle == LTTNG_UST_ROOT_HANDLE)
 			ret = -EPERM;
 		else
-			ret = lttng_ust_objd_unref(lum->handle);
+			ret = lttng_ust_objd_unref(lum->handle, 1);
 		break;
 	case LTTNG_UST_FILTER:
 	{
@@ -376,6 +492,7 @@ int handle_message(struct sock_info *sock_info,
 					goto error;
 				}
 				ret = len;
+				free(bytecode);
 				goto end;
 			} else {
 				DBG("incorrect filter data message size: %zd", len);
@@ -401,12 +518,76 @@ int handle_message(struct sock_info *sock_info,
 		}
 		break;
 	}
+	case LTTNG_UST_EXCLUSION:
+	{
+		/* Receive exclusion names */
+		struct lttng_ust_excluder_node *node;
+		unsigned int count;
+
+		count = lum->u.exclusion.count;
+		if (count == 0) {
+			/* There are no names to read */
+			ret = 0;
+			goto error;
+		}
+		node = zmalloc(sizeof(*node) +
+				count * LTTNG_UST_SYM_NAME_LEN);
+		if (!node) {
+			ret = -ENOMEM;
+			goto error;
+		}
+		node->excluder.count = count;
+		len = ustcomm_recv_unix_sock(sock, node->excluder.names,
+				count * LTTNG_UST_SYM_NAME_LEN);
+		switch (len) {
+		case 0:	/* orderly shutdown */
+			ret = 0;
+			free(node);
+			goto error;
+		default:
+			if (len == count * LTTNG_UST_SYM_NAME_LEN) {
+				DBG("Exclusion data received");
+				break;
+			} else if (len < 0) {
+				DBG("Receive failed from lttng-sessiond with errno %d", (int) -len);
+				if (len == -ECONNRESET) {
+					ERR("%s remote end closed connection", sock_info->name);
+					ret = len;
+					free(node);
+					goto error;
+				}
+				ret = len;
+				free(node);
+				goto end;
+			} else {
+				DBG("Incorrect exclusion data message size: %zd", len);
+				ret = -EINVAL;
+				free(node);
+				goto end;
+			}
+		}
+		if (ops->cmd) {
+			ret = ops->cmd(lum->handle, lum->cmd,
+					(unsigned long) node,
+					&args, sock_info);
+			if (ret) {
+				free(node);
+			}
+			/* Don't free exclusion data if everything went fine. */
+		} else {
+			ret = -ENOSYS;
+			free(node);
+		}
+		break;
+	}
 	case LTTNG_UST_CHANNEL:
 	{
 		void *chan_data;
+		int wakeup_fd;
 
 		len = ustcomm_recv_channel_from_sessiond(sock,
-				&chan_data, lum->u.channel.len);
+				&chan_data, lum->u.channel.len,
+				&wakeup_fd);
 		switch (len) {
 		case 0:	/* orderly shutdown */
 			ret = 0;
@@ -431,6 +612,7 @@ int handle_message(struct sock_info *sock_info,
 			}
 		}
 		args.channel.chan_data = chan_data;
+		args.channel.wakeup_fd = wakeup_fd;
 		if (ops->cmd)
 			ret = ops->cmd(lum->handle, lum->cmd,
 					(unsigned long) &lum->u,
@@ -546,6 +728,14 @@ end:
 
 error:
 	ust_unlock();
+
+	/*
+	 * Performed delayed statedump operations outside of the UST
+	 * lock. We need to take the dynamic loader lock before we take
+	 * the UST lock internally within handle_pending_statedump().
+	  */
+	handle_pending_statedump(sock_info);
+
 	return ret;
 }
 
@@ -554,6 +744,25 @@ void cleanup_sock_info(struct sock_info *sock_info, int exiting)
 {
 	int ret;
 
+	if (sock_info->root_handle != -1) {
+		ret = lttng_ust_objd_unref(sock_info->root_handle, 1);
+		if (ret) {
+			ERR("Error unref root handle");
+		}
+		sock_info->root_handle = -1;
+	}
+	sock_info->constructor_sem_posted = 0;
+
+	/*
+	 * wait_shm_mmap, socket and notify socket are used by listener
+	 * threads outside of the ust lock, so we cannot tear them down
+	 * ourselves, because we cannot join on these threads. Leave
+	 * responsibility of cleaning up these resources to the OS
+	 * process exit.
+	 */
+	if (exiting)
+		return;
+
 	if (sock_info->socket != -1) {
 		ret = ustcomm_close_unix_sock(sock_info->socket);
 		if (ret) {
@@ -568,21 +777,7 @@ void cleanup_sock_info(struct sock_info *sock_info, int exiting)
 		}
 		sock_info->notify_socket = -1;
 	}
-	if (sock_info->root_handle != -1) {
-		ret = lttng_ust_objd_unref(sock_info->root_handle);
-		if (ret) {
-			ERR("Error unref root handle");
-		}
-		sock_info->root_handle = -1;
-	}
-	sock_info->constructor_sem_posted = 0;
-	/*
-	 * wait_shm_mmap is used by listener threads outside of the
-	 * ust lock, so we cannot tear it down ourselves, because we
-	 * cannot join on these threads. Leave this task to the OS
-	 * process exit.
-	 */
-	if (!exiting && sock_info->wait_shm_mmap) {
+	if (sock_info->wait_shm_mmap) {
 		ret = munmap(sock_info->wait_shm_mmap, sysconf(_SC_PAGE_SIZE));
 		if (ret) {
 			ERR("Error unmapping wait shm");
@@ -611,6 +806,30 @@ int get_wait_shm(struct sock_info *sock_info, size_t mmap_size)
 	 */
 	wait_shm_fd = shm_open(sock_info->wait_shm_path, O_RDONLY, 0);
 	if (wait_shm_fd >= 0) {
+		int32_t tmp_read;
+		ssize_t len;
+		size_t bytes_read = 0;
+
+		/*
+		 * Try to read the fd. If unable to do so, try opening
+		 * it in write mode.
+		 */
+		do {
+			len = read(wait_shm_fd,
+				&((char *) &tmp_read)[bytes_read],
+				sizeof(tmp_read) - bytes_read);
+			if (len > 0) {
+				bytes_read += len;
+			}
+		} while ((len < 0 && errno == EINTR)
+			|| (len > 0 && bytes_read < sizeof(tmp_read)));
+		if (bytes_read != sizeof(tmp_read)) {
+			ret = close(wait_shm_fd);
+			if (ret) {
+				ERR("close wait_shm_fd");
+			}
+			goto open_write;
+		}
 		goto end;
 	} else if (wait_shm_fd < 0 && errno != ENOENT) {
 		/*
@@ -621,9 +840,11 @@ int get_wait_shm(struct sock_info *sock_info, size_t mmap_size)
 		ERR("Error opening shm %s", sock_info->wait_shm_path);
 		goto end;
 	}
+
+open_write:
 	/*
-	 * If the open failed because the file did not exist, try
-	 * creating it ourself.
+	 * If the open failed because the file did not exist, or because
+	 * the file was not truncated yet, try creating it ourself.
 	 */
 	URCU_TLS(lttng_ust_nest_count)++;
 	pid = fork();
@@ -818,8 +1039,7 @@ void *ust_listener_thread(void *arg)
 {
 	struct sock_info *sock_info = arg;
 	int sock, ret, prev_connect_failed = 0, has_waited = 0;
-	int open_sock[2];
-	int i;
+	long timeout;
 
 	/* Restart trying to connect to the session daemon */
 restart:
@@ -838,11 +1058,6 @@ restart:
 		has_waited = 1;
 		prev_connect_failed = 0;
 	}
-	ust_lock();
-
-	if (lttng_ust_comm_should_quit) {
-		goto quit;
-	}
 
 	if (sock_info->socket != -1) {
 		ret = ustcomm_close_unix_sock(sock_info->socket);
@@ -861,26 +1076,41 @@ restart:
 		sock_info->notify_socket = -1;
 	}
 
-	/* Register */
-	for (i = 0; i < 2; i++) {
-		ret = ustcomm_connect_unix_sock(sock_info->sock_path);
-		if (ret < 0) {
-			DBG("Info: sessiond not accepting connections to %s apps socket", sock_info->name);
-			prev_connect_failed = 1;
-			/*
-			 * If we cannot find the sessiond daemon, don't delay
-			 * constructor execution.
-			 */
-			ret = handle_register_done(sock_info);
-			assert(!ret);
-			ust_unlock();
-			goto restart;
+	/*
+	 * Register. We need to perform both connect and sending
+	 * registration message before doing the next connect otherwise
+	 * we may reach unix socket connect queue max limits and block
+	 * on the 2nd connect while the session daemon is awaiting the
+	 * first connect registration message.
+	 */
+	/* Connect cmd socket */
+	ret = ustcomm_connect_unix_sock(sock_info->sock_path);
+	if (ret < 0) {
+		DBG("Info: sessiond not accepting connections to %s apps socket", sock_info->name);
+		prev_connect_failed = 1;
+
+		ust_lock();
+
+		if (lttng_ust_comm_should_quit) {
+			goto quit;
 		}
-		open_sock[i] = ret;
+
+		/*
+		 * If we cannot find the sessiond daemon, don't delay
+		 * constructor execution.
+		 */
+		ret = handle_register_done(sock_info);
+		assert(!ret);
+		ust_unlock();
+		goto restart;
 	}
+	sock_info->socket = ret;
+
+	ust_lock();
 
-	sock_info->socket = open_sock[0];
-	sock_info->notify_socket = open_sock[1];
+	if (lttng_ust_comm_should_quit) {
+		goto quit;
+	}
 
 	/*
 	 * Create only one root handle per listener thread for the whole
@@ -910,6 +1140,60 @@ restart:
 		ust_unlock();
 		goto restart;
 	}
+
+	ust_unlock();
+
+	/* Connect notify socket */
+	ret = ustcomm_connect_unix_sock(sock_info->sock_path);
+	if (ret < 0) {
+		DBG("Info: sessiond not accepting connections to %s apps socket", sock_info->name);
+		prev_connect_failed = 1;
+
+		ust_lock();
+
+		if (lttng_ust_comm_should_quit) {
+			goto quit;
+		}
+
+		/*
+		 * If we cannot find the sessiond daemon, don't delay
+		 * constructor execution.
+		 */
+		ret = handle_register_done(sock_info);
+		assert(!ret);
+		ust_unlock();
+		goto restart;
+	}
+	sock_info->notify_socket = ret;
+
+	timeout = get_notify_sock_timeout();
+	if (timeout >= 0) {
+		/*
+		 * Give at least 10ms to sessiond to reply to
+		 * notifications.
+		 */
+		if (timeout < 10)
+			timeout = 10;
+		ret = ustcomm_setsockopt_rcv_timeout(sock_info->notify_socket,
+				timeout);
+		if (ret < 0) {
+			WARN("Error setting socket receive timeout");
+		}
+		ret = ustcomm_setsockopt_snd_timeout(sock_info->notify_socket,
+				timeout);
+		if (ret < 0) {
+			WARN("Error setting socket send timeout");
+		}
+	} else if (timeout < -1) {
+		WARN("Unsupported timeout value %ld", timeout);
+	}
+
+	ust_lock();
+
+	if (lttng_ust_comm_should_quit) {
+		goto quit;
+	}
+
 	ret = register_to_sessiond(sock_info->notify_socket,
 			USTCTL_SOCKET_NOTIFY);
 	if (ret < 0) {
@@ -992,46 +1276,6 @@ quit:
 	return NULL;
 }
 
-/*
- * Return values: -1: don't wait. 0: wait forever. 1: timeout wait.
- */
-static
-int get_timeout(struct timespec *constructor_timeout)
-{
-	long constructor_delay_ms = LTTNG_UST_DEFAULT_CONSTRUCTOR_TIMEOUT_MS;
-	char *str_delay;
-	int ret;
-
-	str_delay = getenv("LTTNG_UST_REGISTER_TIMEOUT");
-	if (str_delay) {
-		constructor_delay_ms = strtol(str_delay, NULL, 10);
-	}
-
-	switch (constructor_delay_ms) {
-	case -1:/* fall-through */
-	case 0:
-		return constructor_delay_ms;
-	default:
-		break;
-	}
-
-	/*
-	 * If we are unable to find the current time, don't wait.
-	 */
-	ret = clock_gettime(CLOCK_REALTIME, constructor_timeout);
-	if (ret) {
-		return -1;
-	}
-	constructor_timeout->tv_sec += constructor_delay_ms / 1000UL;
-	constructor_timeout->tv_nsec +=
-		(constructor_delay_ms % 1000UL) * 1000000UL;
-	if (constructor_timeout->tv_nsec >= 1000000000UL) {
-		constructor_timeout->tv_sec++;
-		constructor_timeout->tv_nsec -= 1000000000UL;
-	}
-	return 1;
-}
-
 /*
  * sessiond monitoring thread: monitor presence of global and per-user
  * sessiond by polling the application common named pipe.
@@ -1067,9 +1311,12 @@ void __attribute__((constructor)) lttng_ust_init(void)
 	init_tracepoint();
 	lttng_ring_buffer_metadata_client_init();
 	lttng_ring_buffer_client_overwrite_init();
+	lttng_ring_buffer_client_overwrite_rt_init();
 	lttng_ring_buffer_client_discard_init();
+	lttng_ring_buffer_client_discard_rt_init();
+	lttng_context_init();
 
-	timeout_mode = get_timeout(&constructor_timeout);
+	timeout_mode = get_constructor_timeout(&constructor_timeout);
 
 	ret = sem_init(&constructor_wait, 0, 0);
 	assert(!ret);
@@ -1170,7 +1417,10 @@ void lttng_ust_cleanup(int exiting)
 	 */
 	lttng_ust_abi_exit();
 	lttng_ust_events_exit();
+	lttng_context_exit();
+	lttng_ring_buffer_client_discard_rt_exit();
 	lttng_ring_buffer_client_discard_exit();
+	lttng_ring_buffer_client_overwrite_rt_exit();
 	lttng_ring_buffer_client_overwrite_exit();
 	lttng_ring_buffer_metadata_client_exit();
 	exit_tracepoint();
@@ -1307,3 +1557,9 @@ void ust_after_fork_child(sigset_t *restore_sigset)
 	ust_after_fork_common(restore_sigset);
 	lttng_ust_init();
 }
+
+void lttng_ust_sockinfo_session_enabled(void *owner)
+{
+	struct sock_info *sock_info = owner;
+	sock_info->statedump_pending = 1;
+}