core/path: install inotify watches top-down instead of bottom-up

When watches are installed from the bottom, it is always possible
to race, and miss a file creation event. The race can be avoided
if a watch is first established for a parent directory, and then for
the file in the directory. If the file is created in the time between,
the watch on the parent directory will fire.

Some messages (mostly at debug level) are added to help diagnose
pidfile issues.

Should fix https://bugzilla.redhat.com/show_bug.cgi?id=917075.

diff --git a/src/core/path.c b/src/core/path.c
index fc101280a1e7c53df895726137f3a3a27dbb80cc..01a2b0810eacb7398700e97d27c70c66b592019f 100644 (file)
--- a/src/core/path.c
+++ b/src/core/path.c
@@ -53,8 +53,8 @@ int path_spec_watch(PathSpec *s, Unit *u) {
         };
 
         bool exists = false;
-        char _cleanup_free_ *k = NULL;
-        char *slash;
+        char _cleanup_free_ *path = NULL;
+        char *slash, *oldslash = NULL;
         int r;
 
         assert(u);
@@ -62,8 +62,8 @@ int path_spec_watch(PathSpec *s, Unit *u) {
 
         path_spec_unwatch(s, u);
 
-        k = strdup(s->path);
-        if (!k)
+        path = strdup(s->path);
+        if (!path)
                 return -ENOMEM;
 
         s->inotify_fd = inotify_init1(IN_NONBLOCK|IN_CLOEXEC);
@@ -76,43 +76,61 @@ int path_spec_watch(PathSpec *s, Unit *u) {
         if (r < 0)
                 goto fail;
 
-        s->primary_wd = inotify_add_watch(s->inotify_fd, k, flags_table[s->type]);
-        if (s->primary_wd >= 0)
-                exists = true;
-        else if (errno != EACCES && errno != ENOENT) {
-                log_error("Failed to add watch on %s: %m", k);
-                r = -errno;
-                goto fail;
-        }
+        /* This assumes the path was passed through path_kill_slashes()! */
 
-        do {
+        for(slash = strchr(path, '/'); ; slash = strchr(slash+1, '/')) {
                 int flags;
+                char tmp;
+
+                if (slash) {
+                        tmp = slash[slash == path];
+                        slash[slash == path] = '\0';
+                        flags = IN_MOVE_SELF | IN_DELETE_SELF | IN_ATTRIB | IN_CREATE | IN_MOVED_TO;
+                } else {
+                        flags = flags_table[s->type];
+                }
 
-                /* This assumes the path was passed through path_kill_slashes()! */
-                slash = strrchr(k, '/');
-                if (!slash)
-                        break;
-
-                /* Trim the path at the last slash. Keep the slash if it's the root dir. */
-                slash[slash == k] = 0;
-
-                flags = IN_MOVE_SELF;
-                if (!exists)
-                        flags |= IN_DELETE_SELF | IN_ATTRIB | IN_CREATE | IN_MOVED_TO;
+                r = inotify_add_watch(s->inotify_fd, path, flags);
+                if (r < 0) {
+                        if (errno == EACCES || errno == ENOENT)
+                                break;
 
-                if (inotify_add_watch(s->inotify_fd, k, flags) >= 0)
-                        exists = true;
-                else if (errno != EACCES && errno != ENOENT) {
-                        log_error("Failed to add watch on %s: %m", k);
+                        log_warning("Failed to add watch on %s: %m", path);
                         r = -errno;
                         goto fail;
+                } else {
+                        exists = true;
+
+                        /* Path exists, we don't need to watch parent
+                           too closely. */
+                        if (oldslash) {
+                                char tmp2 = oldslash[oldslash == path];
+                                oldslash[oldslash == path] = '\0';
+
+                                inotify_add_watch(s->inotify_fd, path, IN_MOVE_SELF);
+                                /* Error is ignored, the worst can happen is
+                                   we get spurious events. */
+
+                                oldslash[oldslash == path] = tmp2;
+                        }
                 }
-        } while (slash != k);
+
+                if (slash) {
+                        slash[slash == path] = tmp;
+                        oldslash = slash;
+                } else {
+                        /* whole path has been iterated over */
+                        s->primary_wd = r;
+                        break;
+                }
+        }
+
+        assert(errno == EACCES || errno == ENOENT || streq(path, s->path));
 
         if (!exists) {
                 log_error("Failed to add watch on any of the components of %s: %m",
                           s->path);
-                r = -errno;
+                r = -errno; /* either EACCESS or ENOENT */
                 goto fail;
         }
 

diff --git a/src/core/service.c b/src/core/service.c
index 3f8aabcf45b69e8b137aa19182903feeefdbac30..61b150cba8e55fff42109bce0bea6e520d4e6754 100644 (file)
--- a/src/core/service.c
+++ b/src/core/service.c
@@ -1400,8 +1400,13 @@ static int service_load_pid_file(Service *s, bool may_warn) {
         }
 
         r = parse_pid(k, &pid);
-        if (r < 0)
+        if (r < 0) {
+                if (may_warn)
+                        log_info_unit(UNIT(s)->id,
+                                      "Failed to read PID from file %s: %s",
+                                      s->pid_file, strerror(-r));
                 return r;
+        }
 
         if (kill(pid, 0) < 0 && errno != EPERM) {
                 if (may_warn)
@@ -1429,9 +1434,13 @@ static int service_load_pid_file(Service *s, bool may_warn) {
                 return r;
 
         r = unit_watch_pid(UNIT(s), pid);
-        if (r < 0)
+        if (r < 0) {
                 /* FIXME: we need to do something here */
+                log_warning_unit(UNIT(s)->id,
+                                 "Failed to watch PID %lu from service %s",
+                                 (unsigned long) pid, UNIT(s)->id);
                 return r;
+        }
 
         return 0;
 }
@@ -2824,6 +2833,9 @@ static int service_watch_pid_file(Service *s) {
                 goto fail;
 
         /* the pidfile might have appeared just before we set the watch */
+        log_debug_unit(UNIT(s)->id,
+                       "Trying to read %s's PID file %s in case it changed",
+                       UNIT(s)->id, s->pid_file_pathspec->path);
         service_retry_pid_file(s);
 
         return 0;