global/pidfile: do not start two daemons with a single pid-file

add functions named pidfile_open and pidfile_verify to avoid starting two daemons by a single pid-file

Fixes: #13422
Signed-off-by: shun song <song.shun3@zte.com.cn>

src/global/global_init.cc

@@ -261,14 +261,17 @@ int global_init_prefork(CephContext *cct)
 {
   if (g_code_env != CODE_ENVIRONMENT_DAEMON)
     return -1;
+  if (pidfile_open(g_conf) < 0) {
+     exit(1);
+  } 
+
   const md_config_t *conf = cct->_conf;
   if (!conf->daemonize) {
     if (atexit(pidfile_remove_void)) {
       derr << "global_init_daemonize: failed to set pidfile_remove function "
 	   << "to run at exit." << dendl;
     }
-
-    pidfile_write(g_conf);
+    pidfile_write();
 
     return -1;
   }
@@ -292,7 +295,11 @@ void global_init_daemonize(CephContext *cct)
 	 << cpp_strerror(ret) << dendl;
     exit(1);
   }
-
+
+ if (pidfile_open(g_conf) < 0) {
+    exit(1);
+ }
+
   global_init_postfork_start(cct);
   global_init_postfork_finish(cct);
 #else
@@ -333,7 +340,7 @@ void global_init_postfork_start(CephContext *cct)
     exit(1);
   }
 
-  pidfile_write(g_conf);
+  pidfile_write();
 }
 
 void global_init_postfork_finish(CephContext *cct)

src/global/pidfile.cc

@@ -31,68 +31,131 @@
 
 #define dout_prefix *_dout
 
-static char pid_file[PATH_MAX] = "";
+struct pidfh {
+        int     pf_fd;
+        char    pf_path[PATH_MAX + 1];
+        dev_t   pf_dev;
+        ino_t   pf_ino;
+
+        pidfh() : pf_fd(-1), pf_dev(0), pf_ino(0) {
+                  memset(pf_path, 0, sizeof(pf_path));
+        }
+
+        void close() {
+                pf_fd = -1;
+                pf_path[0] = '\0';
+                pf_dev = 0;
+                pf_ino = 0;
+        }
+};
+static struct pidfh pfh;
 
-int pidfile_write(const md_config_t *conf)
-{
-  int ret, fd;
+static int pidfile_verify() {
+     struct stat sb;
 
-  if (conf->pid_file.empty()) {
-    return pidfile_remove();
-  }
-  snprintf(pid_file, PATH_MAX, "%s", conf->pid_file.c_str());
+     if (pfh.pf_fd == -1)
+            return -EINVAL;
+     /*
+      * Check remembered descriptor
+      */ 
+    if (fstat(pfh.pf_fd, &sb) == -1)
+            return -errno;
+    if (sb.st_dev != pfh.pf_dev || sb.st_ino != pfh.pf_ino)
+           return -ESTALE;
+    return 0;
+}
 
-  fd = TEMP_FAILURE_RETRY(::open(pid_file,
-				 O_CREAT|O_TRUNC|O_WRONLY, 0644));
-  if (fd < 0) {
-    int err = errno;
-    derr << "write_pid_file: failed to open pid file '"
-	 << pid_file << "': " << cpp_strerror(err) << dendl;
-    return err;
-  }
+int pidfile_write()
+{
+  int ret;
+  if (!pfh.pf_path[0])
+      return 0;
 
-  char buf[20];
+  if ((ret=pidfile_verify()) < 0) {
+    return ret;
+  }
+  char buf[32];
   int len = snprintf(buf, sizeof(buf), "%d\n", getpid());
-  ret = safe_write(fd, buf, len);
+  ret = safe_write(pfh.pf_fd, buf, len);
   if (ret < 0) {
     derr << "write_pid_file: failed to write to pid file '"
-	 << pid_file << "': " << cpp_strerror(ret) << dendl;
-    VOID_TEMP_FAILURE_RETRY(::close(fd));
+	 << pfh.pf_path << "': " << cpp_strerror(ret) << dendl;
+    pfh.close(); 
     return ret;
   }
-  if (TEMP_FAILURE_RETRY(::close(fd))) {
-    ret = errno;
-    derr << "SimpleMessenger::write_pid_file: failed to close to pid file '"
-	 << pid_file << "': " << cpp_strerror(ret) << dendl;
-    return -ret;
-  }
 
   return 0;
 }
 
 int pidfile_remove(void)
 {
-  if (!pid_file[0])
+  int ret;
+
+  if (!pfh.pf_path[0])
     return 0;
+  if ( (ret=pidfile_verify()) < 0) {
+      VOID_TEMP_FAILURE_RETRY(::close(pfh.pf_fd));
+      return ret;
+  }
 
   // only remove it if it has OUR pid in it!
-  int fd = TEMP_FAILURE_RETRY(::open(pid_file, O_RDONLY));
-  if (fd < 0)
-    return -errno;
   char buf[32];
   memset(buf, 0, sizeof(buf));
-  ssize_t res = safe_read(fd, buf, sizeof(buf));
-  VOID_TEMP_FAILURE_RETRY(::close(fd));
+  ssize_t res = safe_read(pfh.pf_fd, buf, sizeof(buf));
+  VOID_TEMP_FAILURE_RETRY(::close(pfh.pf_fd));
   if (res < 0)
-    return res;
+     return res;
   int a = atoi(buf);
   if (a != getpid())
     return -EDOM;
-
-  res = ::unlink(pid_file);
+  
+  res = ::unlink(pfh.pf_path);
   if (res)
-    return res;
-
-  pid_file[0] = '\0';
+     return res;
+  pfh.close();
   return 0;
 }
+
+int pidfile_open(const md_config_t *conf) 
+{
+  int fd;
+  struct stat sb;
+  if (conf->pid_file.empty()) {
+    return 0;
+  }
+  int len = snprintf(pfh.pf_path, sizeof(pfh.pf_path),
+                    "%s", conf->pid_file.c_str());
+
+  if (len >= (int)sizeof(pfh.pf_path)) { 
+      return -ENAMETOOLONG;
+  }
+
+  fd = ::open(pfh.pf_path, O_CREAT|O_WRONLY, 0644);
+  if (fd < 0) {
+     int err = errno;
+     derr << "write_pid_file: failed to open pid file '"
+          << pfh.pf_path << "': " << cpp_strerror(err) << dendl;
+     pfh.close();
+     return -errno;
+   }
+
+   if (fstat(fd, &sb) == -1) {
+         close(fd);
+         pfh.close();  
+         return -errno;
+   }
+   pfh.pf_fd = fd;
+   pfh.pf_dev = sb.st_dev;
+   pfh.pf_ino = sb.st_ino;
+
+   struct flock l = { F_WRLCK, SEEK_SET, 0, 0, 0 };
+   int r = ::fcntl(pfh.pf_fd, F_SETLK, &l);
+   if (r < 0) {
+       derr << "failed to lock" << pfh.pf_path << ".is there another process in using?" << dendl;
+       close(pfh.pf_fd);
+       pfh.close();
+       return -errno;
+   }
+   return 0;
+}
+

src/global/pidfile.h

@@ -19,10 +19,13 @@ struct md_config_t;
 
 // Write a pidfile with the current pid, using the configuration in the
 // provided conf structure.
-int pidfile_write(const md_config_t *conf);
+int pidfile_write();
 
 // Remove the pid file that was previously written by pidfile_write.
 // This is safe to call in a signal handler context.
 int pidfile_remove(void);
 
+//test whether the pid_file is being used by another process
+int pidfile_open(const md_config_t *conf);
+
 #endif