ctdb/tools/ctdb_killtcp.c

   1 /*
   2    CTDB TCP connection killing utility
   3
   4    Copyright (C) Martin Schwenke <martin@meltin.net> 2016
   5
   6    This program is free software; you can redistribute it and/or modify
   7    it under the terms of the GNU General Public License as published by
   8    the Free Software Foundation; either version 3 of the License, or
   9    (at your option) any later version.
  10
  11    This program is distributed in the hope that it will be useful,
  12    but WITHOUT ANY WARRANTY; without even the implied warranty of
  13    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14    GNU General Public License for more details.
  15
  16    You should have received a copy of the GNU General Public License
  17    along with this program; if not, see <http://www.gnu.org/licenses/>.
  18 */
  19
  20 #include "replace.h"
  21 #include "system/network.h"
  22
  23 #include <talloc.h>
  24 #include <tevent.h>
  25
  26 #include "lib/util/debug.h"
  27 #include "lib/util/tevent_unix.h"
  28
  29 #include "protocol/protocol.h"
  30 #include "protocol/protocol_util.h"
  31
  32 #include "common/db_hash.h"
  33 #include "common/system_socket.h"
  34 #include "common/logging.h"
  35
  36
  37 struct reset_connections_state {
  38         struct tevent_context *ev;
  39         int capture_fd;
  40         struct tevent_fd *fde;
  41         struct db_hash_context *connections;
  42         void *private_data;
  43         unsigned int attempts;
  44         unsigned int max_attempts;
  45         struct timeval retry_interval;
  46         unsigned int batch_count;
  47         unsigned int batch_size;
  48 };
  49
  50
  51 static void reset_connections_capture_tcp_handler(struct tevent_context *ev,
  52                                                   struct tevent_fd *fde,
  53                                                   uint16_t flags,
  54                                                   void *private_data);
  55 static void reset_connections_batch(struct tevent_req *subreq);
  56 static int reset_connections_tickle_connection(
  57                                         uint8_t *keybuf, size_t keylen,
  58                                         uint8_t *databuf, size_t datalen,
  59                                         void *private_data);
  60
  61 static struct tevent_req *reset_connections_send(
  62                               TALLOC_CTX *mem_ctx,
  63                               struct tevent_context *ev,
  64                               const char *iface,
  65                               struct ctdb_connection_list *conn_list)
  66 {
  67         struct tevent_req *req, *subreq;
  68         struct reset_connections_state *state;
  69         unsigned int i;
  70         int ret;
  71
  72         req = tevent_req_create(mem_ctx, &state,
  73                                 struct reset_connections_state);
  74         if (req == NULL) {
  75                 return NULL;
  76         }
  77
  78         state->ev = ev;
  79
  80         if (conn_list->num == 0) {
  81                 /* No connections, done! */
  82                 tevent_req_done(req);
  83                 return tevent_req_post(req, ev);
  84         }
  85
  86         ret = db_hash_init(state, "connections", 2048, DB_HASH_SIMPLE,
  87                            &state->connections);
  88         if (ret != 0) {
  89                 D_ERR("Failed to initialise connection hash (%s)\n",
  90                       strerror(ret));
  91                 tevent_req_error(req, ret);
  92                 return tevent_req_post(req, ev);
  93         }
  94
  95         DBG_DEBUG("Adding %u connections to hash\n", conn_list->num);
  96         for (i = 0; i < conn_list->num; i++) {
  97                 struct ctdb_connection *c = &conn_list->conn[i];
  98
  99                 DBG_DEBUG("Adding connection to hash: %s\n",
 100                           ctdb_connection_to_string(conn_list, c, true));
 101
 102                 /* Connection is stored as a key in the connections hash */
 103                 ret = db_hash_add(state->connections,
 104                                   (uint8_t *)discard_const(c), sizeof(*c),
 105                                   NULL, 0);
 106                 if (ret != 0) {
 107                         D_ERR("Error adding connection to hash (%s)\n",
 108                               strerror(ret));
 109                         tevent_req_error(req, ret);
 110                         return tevent_req_post(req, ev);
 111                 }
 112         }
 113
 114         state->attempts = 0;
 115         state->max_attempts = 50;
 116
 117         state->retry_interval.tv_sec = 0;
 118         state->retry_interval.tv_usec = 100 * 1000;
 119
 120         state->batch_count = 0;
 121         state->batch_size = 300;
 122
 123         state->capture_fd =
 124                 ctdb_sys_open_capture_socket(iface, &state->private_data);
 125         if (state->capture_fd == -1) {
 126                 D_ERR("Failed to open capture socket on iface '%s' (%s)\n",
 127                       iface, strerror(errno));
 128                         tevent_req_error(req, EIO);
 129                         return tevent_req_post(req, ev);
 130         }
 131
 132         state->fde = tevent_add_fd(ev, state, state->capture_fd,
 133                                    TEVENT_FD_READ,
 134                                    reset_connections_capture_tcp_handler,
 135                                    state);
 136         if (tevent_req_nomem(state->fde, req)) {
 137                 return tevent_req_post(req, ev);
 138         }
 139         tevent_fd_set_auto_close(state->fde);
 140
 141         subreq = tevent_wakeup_send(state, ev, tevent_timeval_current_ofs(0,0));
 142         if (tevent_req_nomem(subreq, req)) {
 143                 return tevent_req_post(req, ev);
 144         }
 145         tevent_req_set_callback(subreq, reset_connections_batch, req);
 146
 147         return req;
 148 }
 149
 150 /*
 151   called when we get a read event on the raw socket
 152  */
 153 static void reset_connections_capture_tcp_handler(struct tevent_context *ev,
 154                                                   struct tevent_fd *fde,
 155                                                   uint16_t flags,
 156                                                   void *private_data)
 157 {
 158         struct reset_connections_state *state = talloc_get_type_abort(
 159                 private_data, struct reset_connections_state);
 160         /* 0 the parts that don't get set by ctdb_sys_read_tcp_packet */
 161         struct ctdb_connection conn;
 162         uint32_t ack_seq, seq;
 163         int rst;
 164         uint16_t window;
 165         int ret;
 166
 167         ret = ctdb_sys_read_tcp_packet(state->capture_fd,
 168                                        state->private_data,
 169                                        &conn.server, &conn.client,
 170                                        &ack_seq, &seq, &rst, &window);
 171         if (ret != 0) {
 172                 /* probably a non-tcp ACK packet */
 173                 return;
 174         }
 175
 176         if (window == htons(1234) && (rst || seq == 0)) {
 177                 /* Ignore packets that we sent! */
 178                 D_DEBUG("Ignoring packet: %s, "
 179                         "seq=%"PRIu32", ack_seq=%"PRIu32", "
 180                         "rst=%d, window=%"PRIu16"\n",
 181                         ctdb_connection_to_string(state, &conn, false),
 182                         seq, ack_seq, rst, ntohs(window));
 183                 return;
 184         }
 185
 186         /* Check if this connection is one being reset, if found then delete */
 187         ret = db_hash_delete(state->connections,
 188                              (uint8_t*)&conn, sizeof(conn));
 189         if (ret == ENOENT) {
 190                 /* Packet for some other connection, ignore */
 191                 DBG_DEBUG("Ignoring packet for unknown connection: %s\n",
 192                           ctdb_connection_to_string(state, &conn, true));
 193                 return;
 194         }
 195         if (ret != 0) {
 196                 DBG_WARNING("Internal error (%s)\n", strerror(ret));
 197                 return;
 198         }
 199
 200         D_INFO("Sending a TCP RST to for connection %s\n",
 201                ctdb_connection_to_string(state, &conn, true));
 202
 203         ret = ctdb_sys_send_tcp(&conn.server, &conn.client, ack_seq, seq, 1);
 204         if (ret != 0) {
 205                 DBG_ERR("Error sending TCP RST for connection\n");
 206         }
 207 }
 208
 209 /*
 210  * Called periodically until all sentenced connections have been reset
 211  * or enough attempts have been made
 212  */
 213 static void reset_connections_batch(struct tevent_req *subreq)
 214 {
 215         struct tevent_req *req = tevent_req_callback_data(
 216                 subreq, struct tevent_req);
 217         struct reset_connections_state *state = tevent_req_data(
 218                 req, struct reset_connections_state);
 219         bool status;
 220         int count, ret;
 221
 222         status = tevent_wakeup_recv(subreq);
 223         TALLOC_FREE(subreq);
 224
 225         if (! status) {
 226                 DBG_WARNING("Unexpected error on timer expiry\n");
 227                 /* Keep going... */
 228         }
 229
 230         /* loop over up to batch_size connections sending tickle ACKs */
 231         state->batch_count = 0;
 232         ret = db_hash_traverse(state->connections,
 233                                reset_connections_tickle_connection,
 234                                state, NULL);
 235         if (ret != 0) {
 236                 DBG_WARNING("Unexpected error traversing connections (%s)\n",
 237                             strerror(ret));
 238         }
 239
 240         state->attempts++;
 241
 242         /*
 243          * If there are no more connections to kill or we have tried
 244          * too many times we're finished
 245          */
 246         ret = db_hash_traverse(state->connections, NULL, NULL, &count);
 247         if (ret != 0) {
 248                 /* What now?  Try again until max_attempts reached */
 249                 DBG_WARNING("Unexpected error traversing connections (%s)\n",
 250                             strerror(ret));
 251                 count = 1;
 252         }
 253         if (count == 0 ||
 254             state->attempts >= state->max_attempts) {
 255                 tevent_req_done(req);
 256                 return;
 257         }
 258
 259         /* Schedule next attempt */
 260         subreq = tevent_wakeup_send(state, state->ev,
 261                                     tevent_timeval_current_ofs(
 262                                             state->retry_interval.tv_sec,
 263                                             state->retry_interval.tv_usec));
 264         if (tevent_req_nomem(subreq, req)) {
 265                 return;
 266         }
 267         tevent_req_set_callback(subreq, reset_connections_batch, req);
 268 }
 269
 270 static int reset_connections_tickle_connection(
 271                                         uint8_t *keybuf, size_t keylen,
 272                                         uint8_t *databuf, size_t datalen,
 273                                         void *private_data)
 274 {
 275         struct reset_connections_state *state = talloc_get_type_abort(
 276                 private_data, struct reset_connections_state);
 277         struct ctdb_connection *conn;
 278         int ret;
 279
 280         if (keylen != sizeof(*conn)) {
 281                 DBG_WARNING("Unexpected data in connection hash\n");
 282                 return 0;
 283         }
 284
 285         conn = (struct ctdb_connection *)keybuf;
 286
 287         state->batch_count++;
 288         if (state->batch_count > state->batch_size) {
 289                 /* Terminate the traverse */
 290                 return 1;
 291         }
 292
 293         DBG_DEBUG("Sending tickle ACK for connection '%s'\n",
 294                   ctdb_connection_to_string(state, conn, true));
 295         ret = ctdb_sys_send_tcp(&conn->server, &conn->client, 0, 0, 0);
 296         if (ret != 0) {
 297                 DBG_ERR("Error sending tickle ACK\n");
 298                 /* continue */
 299         }
 300
 301         return 0;
 302 }
 303
 304 static bool reset_connections_recv(struct tevent_req *req, int *perr)
 305 {
 306         int err;
 307
 308         if (tevent_req_is_unix_error(req, &err)) {
 309                 if (perr != NULL) {
 310                         *perr = err;
 311                 }
 312                 return false;
 313         }
 314
 315         return true;
 316 }
 317
 318 static void usage(const char *prog)
 319 {
 320         printf("usage: %s <interface> [ <srcip:port> <dstip:port> ]\n", prog);
 321         exit(1);
 322 }
 323
 324 int main(int argc, char **argv)
 325 {
 326         struct ctdb_connection conn;
 327         struct tevent_context *ev = NULL;
 328         TALLOC_CTX *mem_ctx = NULL;
 329         struct ctdb_connection_list *conn_list = NULL;
 330         const char *t;
 331         struct tevent_req *req;
 332         int debug_level;
 333         bool status;
 334         bool ok;
 335         int ret;
 336
 337         /* Set the debug level */
 338         t = getenv("CTDB_DEBUGLEVEL");
 339         if (t != NULL) {
 340                 ok = debug_level_parse(t, &debug_level);
 341                 if (!ok) {
 342                         debug_level = DEBUG_ERR;
 343                 }
 344                 debuglevel_set(debug_level);
 345         }
 346
 347         if (argc != 2 && argc != 4) {
 348                 usage(argv[0]);
 349         }
 350
 351         if (argc == 4) {
 352                 ret = ctdb_sock_addr_from_string(argv[2], &conn.client, true);
 353                 if (ret != 0) {
 354                         D_ERR("Bad IP:port '%s'\n", argv[2]);
 355                         goto fail;
 356                 }
 357
 358                 ret = ctdb_sock_addr_from_string(argv[3], &conn.server, true);
 359                 if (ret != 0) {
 360                         D_ERR("Bad IP:port '%s'\n", argv[3]);
 361                         goto fail;
 362                 }
 363
 364
 365                 conn_list = talloc_zero(mem_ctx, struct ctdb_connection_list);
 366                 if (conn_list == NULL) {
 367                         ret = ENOMEM;
 368                         DBG_ERR("Internal error (%s)\n", strerror(ret));
 369                         goto fail;
 370                 }
 371                 ret = ctdb_connection_list_add(conn_list, &conn);
 372                 if (ret != 0) {
 373                         DBG_ERR("Internal error (%s)\n", strerror(ret));
 374                         goto fail;
 375                 }
 376         } else {
 377                 ret = ctdb_connection_list_read(mem_ctx, 0, true, &conn_list);
 378                 if (ret != 0) {
 379                         D_ERR("Unable to parse connections (%s)\n",
 380                               strerror(ret));
 381                         goto fail;
 382                 }
 383         }
 384
 385         mem_ctx = talloc_new(NULL);
 386         if (mem_ctx == NULL) {
 387                 DEBUG(DEBUG_ERR, (__location__ " out of memory\n"));
 388                 goto fail;
 389         }
 390
 391         ev = tevent_context_init(mem_ctx);
 392         if (ev == NULL) {
 393                 DEBUG(DEBUG_ERR, ("Failed to initialise tevent\n"));
 394                 goto fail;
 395         }
 396
 397         req = reset_connections_send(mem_ctx, ev, argv[1], conn_list);
 398         if (req == NULL) {
 399                 goto fail;
 400         }
 401
 402         tevent_req_poll(req, ev);
 403
 404         status = reset_connections_recv(req, &ret);
 405         if (! status) {
 406                 D_ERR("Failed to kill connections (%s)\n", strerror(ret));
 407                 goto fail;
 408         }
 409
 410         talloc_free(mem_ctx);
 411
 412         return 0;
 413
 414 fail:
 415         TALLOC_FREE(mem_ctx);
 416         return -1;
 417 }