ctdb/server/ctdb_cluster_mutex.c

   1 /*
   2    CTDB cluster mutex handling
   3
   4    Copyright (C) Andrew Tridgell  2007
   5    Copyright (C) Ronnie Sahlberg  2007
   6    Copyright (C) Martin Schwenke  2016
   7
   8    This program is free software; you can redistribute it and/or modify
   9    it under the terms of the GNU General Public License as published by
  10    the Free Software Foundation; either version 3 of the License, or
  11    (at your option) any later version.
  12
  13    This program is distributed in the hope that it will be useful,
  14    but WITHOUT ANY WARRANTY; without even the implied warranty of
  15    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  16    GNU General Public License for more details.
  17
  18    You should have received a copy of the GNU General Public License
  19    along with this program; if not, see <http://www.gnu.org/licenses/>.
  20 */
  21
  22 #include <tevent.h>
  23
  24 #include "replace.h"
  25 #include "system/network.h"
  26
  27 #include "lib/util/debug.h"
  28 #include "lib/util/time.h"
  29 #include "lib/util/strv.h"
  30 #include "lib/util/strv_util.h"
  31 #include "lib/util/sys_rw.h"
  32 #include "lib/util/blocking.h"
  33
  34 #include "ctdb_private.h"
  35 #include "common/common.h"
  36 #include "common/logging.h"
  37
  38 #include "ctdb_cluster_mutex.h"
  39
  40 struct ctdb_cluster_mutex_handle {
  41         struct ctdb_context *ctdb;
  42         cluster_mutex_handler_t handler;
  43         void *private_data;
  44         cluster_mutex_lost_handler_t lost_handler;
  45         void *lost_data;
  46         int fd[2];
  47         struct tevent_timer *te;
  48         struct tevent_fd *fde;
  49         pid_t child;
  50         struct timeval start_time;
  51         bool have_response;
  52 };
  53
  54 static void cluster_mutex_timeout(struct tevent_context *ev,
  55                                   struct tevent_timer *te,
  56                                   struct timeval t, void *private_data)
  57 {
  58         struct ctdb_cluster_mutex_handle *h =
  59                 talloc_get_type(private_data, struct ctdb_cluster_mutex_handle);
  60         double latency = timeval_elapsed(&h->start_time);
  61
  62         if (h->handler != NULL) {
  63                 h->handler('2', latency, h->private_data);
  64         }
  65 }
  66
  67
  68 /* When the handle is freed it causes any child holding the mutex to
  69  * be killed, thus freeing the mutex */
  70 static int cluster_mutex_destructor(struct ctdb_cluster_mutex_handle *h)
  71 {
  72         if (h->fd[0] != -1) {
  73                 h->fd[0] = -1;
  74         }
  75         ctdb_kill(h->ctdb, h->child, SIGTERM);
  76         return 0;
  77 }
  78
  79 /* this is called when the client process has completed ctdb_recovery_lock()
  80    and has written data back to us through the pipe.
  81 */
  82 static void cluster_mutex_handler(struct tevent_context *ev,
  83                                   struct tevent_fd *fde,
  84                                   uint16_t flags, void *private_data)
  85 {
  86         struct ctdb_cluster_mutex_handle *h=
  87                 talloc_get_type(private_data, struct ctdb_cluster_mutex_handle);
  88         double latency = timeval_elapsed(&h->start_time);
  89         char c = '0';
  90         int ret;
  91
  92         /* Got response from child process so abort timeout */
  93         TALLOC_FREE(h->te);
  94
  95         ret = sys_read(h->fd[0], &c, 1);
  96
  97         /* Don't call the handler more than once.  It only exists to
  98          * process the initial response from the helper. */
  99         if (h->have_response) {
 100                 /* Only deal with EOF due to process exit.  Silently
 101                  * ignore any other output. */
 102                 if (ret == 0) {
 103                         if (h->lost_handler != NULL) {
 104                                 h->lost_handler(h->lost_data);
 105                         }
 106                 }
 107                 return;
 108         }
 109         h->have_response = true;
 110
 111         /* If the child wrote status then just pass it to the handler.
 112          * If no status was written then this is an unexpected error
 113          * so pass generic error code to handler. */
 114         if (h->handler != NULL) {
 115                 h->handler(ret == 1 ? c : '3', latency, h->private_data);
 116         }
 117 }
 118
 119 static char cluster_mutex_helper[PATH_MAX+1] = "";
 120
 121 static bool cluster_mutex_helper_args(TALLOC_CTX *mem_ctx,
 122                                       const char *argstring, char ***argv)
 123 {
 124         int nargs, i, ret, n;
 125         bool is_command = false;
 126         char **args = NULL;
 127         char *strv = NULL;
 128         char *t = NULL;
 129
 130         if (argstring != NULL && argstring[0] == '!') {
 131                 /* This is actually a full command */
 132                 is_command = true;
 133                 t = discard_const(&argstring[1]);
 134         } else {
 135                 is_command = false;
 136                 t = discard_const(argstring);
 137         }
 138
 139         ret = strv_split(mem_ctx, &strv, t, " \t");
 140         if (ret != 0) {
 141                 DEBUG(DEBUG_ERR,
 142                       ("Unable to parse mutex helper string \"%s\" (%s)\n",
 143                        argstring, strerror(ret)));
 144                 return false;
 145         }
 146         n = strv_count(strv);
 147
 148         args = talloc_array(mem_ctx, char *, n + (is_command ? 1 : 2));
 149
 150         if (args == NULL) {
 151                 DEBUG(DEBUG_ERR,(__location__ " out of memory\n"));
 152                 return false;
 153         }
 154
 155         nargs = 0;
 156
 157         if (! is_command) {
 158                 if (!ctdb_set_helper("cluster mutex helper",
 159                                      cluster_mutex_helper,
 160                                      sizeof(cluster_mutex_helper),
 161                                      "CTDB_CLUSTER_MUTEX_HELPER",
 162                                      CTDB_HELPER_BINDIR,
 163                                      "ctdb_mutex_fcntl_helper")) {
 164                         DEBUG(DEBUG_ERR,("ctdb exiting with error: %s\n",
 165                                          __location__
 166                                          " Unable to set cluster mutex helper\n"));
 167                         exit(1);
 168                 }
 169
 170                 args[nargs++] = cluster_mutex_helper;
 171         }
 172
 173         t = NULL;
 174         for (i = 0; i < n; i++) {
 175                 /* Don't copy, just keep cmd_args around */
 176                 t = strv_next(strv, t);
 177                 args[nargs++] = t;
 178         }
 179
 180         /* Make sure last argument is NULL */
 181         args[nargs] = NULL;
 182
 183         *argv = args;
 184         return true;
 185 }
 186
 187 struct ctdb_cluster_mutex_handle *
 188 ctdb_cluster_mutex(TALLOC_CTX *mem_ctx,
 189                    struct ctdb_context *ctdb,
 190                    const char *argstring,
 191                    int timeout,
 192                    cluster_mutex_handler_t handler,
 193                    void *private_data,
 194                    cluster_mutex_lost_handler_t lost_handler,
 195                    void *lost_data)
 196 {
 197         struct ctdb_cluster_mutex_handle *h;
 198         char **args;
 199         int ret;
 200
 201         h = talloc(mem_ctx, struct ctdb_cluster_mutex_handle);
 202         if (h == NULL) {
 203                 DEBUG(DEBUG_ERR, (__location__ " out of memory\n"));
 204                 return NULL;
 205         }
 206
 207         h->start_time = timeval_current();
 208         h->fd[0] = -1;
 209         h->fd[1] = -1;
 210         h->have_response = false;
 211
 212         ret = pipe(h->fd);
 213         if (ret != 0) {
 214                 talloc_free(h);
 215                 DEBUG(DEBUG_ERR, (__location__ " Failed to open pipe\n"));
 216                 return NULL;
 217         }
 218         set_close_on_exec(h->fd[0]);
 219
 220         /* Create arguments for lock helper */
 221         if (!cluster_mutex_helper_args(h, argstring, &args)) {
 222                 close(h->fd[0]);
 223                 close(h->fd[1]);
 224                 talloc_free(h);
 225                 return NULL;
 226         }
 227
 228         h->child = ctdb_fork(ctdb);
 229         if (h->child == (pid_t)-1) {
 230                 close(h->fd[0]);
 231                 close(h->fd[1]);
 232                 talloc_free(h);
 233                 return NULL;
 234         }
 235
 236         if (h->child == 0) {
 237                 /* Make stdout point to the pipe */
 238                 close(STDOUT_FILENO);
 239                 dup2(h->fd[1], STDOUT_FILENO);
 240                 close(h->fd[1]);
 241
 242                 execv(args[0], args);
 243
 244                 /* Only happens on error */
 245                 DEBUG(DEBUG_ERR, (__location__ "execv() failed\n"));
 246                 _exit(1);
 247         }
 248
 249         /* Parent */
 250
 251         DEBUG(DEBUG_DEBUG, (__location__ " Created PIPE FD:%d\n", h->fd[0]));
 252         set_close_on_exec(h->fd[0]);
 253
 254         close(h->fd[1]);
 255         h->fd[1] = -1;
 256
 257         talloc_set_destructor(h, cluster_mutex_destructor);
 258
 259         if (timeout != 0) {
 260                 h->te = tevent_add_timer(ctdb->ev, h,
 261                                          timeval_current_ofs(timeout, 0),
 262                                          cluster_mutex_timeout, h);
 263         } else {
 264                 h->te = NULL;
 265         }
 266
 267         h->fde = tevent_add_fd(ctdb->ev, h, h->fd[0], TEVENT_FD_READ,
 268                                cluster_mutex_handler, (void *)h);
 269
 270         if (h->fde == NULL) {
 271                 talloc_free(h);
 272                 return NULL;
 273         }
 274         tevent_fd_set_auto_close(h->fde);
 275
 276         h->ctdb = ctdb;
 277         h->handler = handler;
 278         h->private_data = private_data;
 279         h->lost_handler = lost_handler;
 280         h->lost_data = lost_data;
 281
 282         return h;
 283 }