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 <talloc.h>
  21 #include <tevent.h>
  22
  23 #include "replace.h"
  24 #include "system/network.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.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         int i, ret;
  70
  71         req = tevent_req_create(mem_ctx, &state,
  72                                 struct reset_connections_state);
  73         if (req == NULL) {
  74                 return NULL;
  75         }
  76
  77         state->ev = ev;
  78
  79         ret = db_hash_init(state, "connections", 2048, DB_HASH_SIMPLE,
  80                            &state->connections);
  81         if (ret != 0) {
  82                 D_ERR("Failed to initialise connection hash (%s)\n",
  83                       strerror(ret));
  84                 tevent_req_error(req, ret);
  85                 return tevent_req_post(req, ev);
  86         }
  87
  88         for (i = 0; i < conn_list->num; i++) {
  89                 struct ctdb_connection *c = &conn_list->conn[i];
  90
  91                 /* Connection is stored as a key in the connections hash */
  92                 ret = db_hash_add(state->connections,
  93                                   (uint8_t *)discard_const(c), sizeof(*c),
  94                                   NULL, 0);
  95                 if (ret != 0) {
  96                         D_ERR("Error adding connection to hash (%s)\n",
  97                               strerror(ret));
  98                         tevent_req_error(req, ret);
  99                         return tevent_req_post(req, ev);
 100                 }
 101         }
 102
 103         state->attempts = 0;
 104         state->max_attempts = 50;
 105
 106         state->retry_interval.tv_sec = 0;
 107         state->retry_interval.tv_usec = 100 * 1000;
 108
 109         state->batch_count = 0;
 110         state->batch_size = 300;
 111
 112         state->capture_fd =
 113                 ctdb_sys_open_capture_socket(iface, &state->private_data);
 114         if (state->capture_fd == -1) {
 115                 D_ERR("Failed to open capture socket on iface '%s' (%s)\n",
 116                       iface, strerror(errno));
 117                         tevent_req_error(req, EIO);
 118                         return tevent_req_post(req, ev);
 119         }
 120
 121         state->fde = tevent_add_fd(ev, state, state->capture_fd,
 122                                    TEVENT_FD_READ,
 123                                    reset_connections_capture_tcp_handler,
 124                                    state);
 125         if (tevent_req_nomem(state->fde, req)) {
 126                 return tevent_req_post(req, ev);
 127         }
 128         tevent_fd_set_auto_close(state->fde);
 129
 130         subreq = tevent_wakeup_send(state, ev, tevent_timeval_current_ofs(0,0));
 131         if (tevent_req_nomem(subreq, req)) {
 132                 return tevent_req_post(req, ev);
 133         }
 134         tevent_req_set_callback(subreq, reset_connections_batch, req);
 135
 136         return req;
 137 }
 138
 139 /*
 140   called when we get a read event on the raw socket
 141  */
 142 static void reset_connections_capture_tcp_handler(struct tevent_context *ev,
 143                                                   struct tevent_fd *fde,
 144                                                   uint16_t flags,
 145                                                   void *private_data)
 146 {
 147         struct reset_connections_state *state = talloc_get_type_abort(
 148                 private_data, struct reset_connections_state);
 149         /* 0 the parts that don't get set by ctdb_sys_read_tcp_packet */
 150         struct ctdb_connection conn;
 151         uint32_t ack_seq, seq;
 152         int rst;
 153         uint16_t window;
 154         int ret;
 155
 156         ret = ctdb_sys_read_tcp_packet(state->capture_fd,
 157                                        state->private_data,
 158                                        &conn.server, &conn.client,
 159                                        &ack_seq, &seq, &rst, &window);
 160         if (ret != 0) {
 161                 /* probably a non-tcp ACK packet */
 162                 return;
 163         }
 164
 165         if (window == htons(1234) && (rst || seq == 0)) {
 166                 /* Ignore packets that we sent! */
 167                 D_DEBUG("Ignoring packet: %s, "
 168                         "seq=%"PRIu32", ack_seq=%"PRIu32", "
 169                         "rst=%d, window=%"PRIu16"\n",
 170                         ctdb_connection_to_string(state, &conn, false),
 171                         seq, ack_seq, rst, ntohs(window));
 172                 return;
 173         }
 174
 175         /* Check if this connection is one being reset, if found then delete */
 176         ret = db_hash_delete(state->connections,
 177                              (uint8_t*)&conn, sizeof(conn));
 178         if (ret == ENOENT) {
 179                 /* Packet for some other connection, ignore */
 180                 return;
 181         }
 182         if (ret != 0) {
 183                 DBG_WARNING("Internal error (%s)\n", strerror(ret));
 184                 return;
 185         }
 186
 187         D_INFO("Sending a TCP RST to for connection %s\n",
 188                ctdb_connection_to_string(state, &conn, true));
 189
 190         ret = ctdb_sys_send_tcp(&conn.server, &conn.client, ack_seq, seq, 1);
 191         if (ret != 0) {
 192                 DBG_ERR("Error sending TCP RST for connection\n");
 193         }
 194 }
 195
 196 /*
 197  * Called periodically until all sentenced connections have been reset
 198  * or enough attempts have been made
 199  */
 200 static void reset_connections_batch(struct tevent_req *subreq)
 201 {
 202         struct tevent_req *req = tevent_req_callback_data(
 203                 subreq, struct tevent_req);
 204         struct reset_connections_state *state = tevent_req_data(
 205                 req, struct reset_connections_state);
 206         bool status;
 207         int count, ret;
 208
 209         status = tevent_wakeup_recv(subreq);
 210         TALLOC_FREE(subreq);
 211
 212         if (! status) {
 213                 DBG_WARNING("Unexpected error on timer expiry\n");
 214                 /* Keep going... */
 215         }
 216
 217         /* loop over up to batch_size connections sending tickle ACKs */
 218         state->batch_count = 0;
 219         ret = db_hash_traverse(state->connections,
 220                                reset_connections_tickle_connection,
 221                                state, NULL);
 222         if (ret != 0) {
 223                 DBG_WARNING("Unexpected error traversing connections (%s)\n",
 224                             strerror(ret));
 225         }
 226
 227         state->attempts++;
 228
 229         /*
 230          * If there are no more connections to kill or we have tried
 231          * too many times we're finished
 232          */
 233         ret = db_hash_traverse(state->connections, NULL, NULL, &count);
 234         if (ret != 0) {
 235                 /* What now?  Try again until max_attempts reached */
 236                 DBG_WARNING("Unexpected error traversing connections (%s)\n",
 237                             strerror(ret));
 238                 count = 1;
 239         }
 240         if (count == 0 ||
 241             state->attempts >= state->max_attempts) {
 242                 tevent_req_done(req);
 243                 return;
 244         }
 245
 246         /* Schedule next attempt */
 247         subreq = tevent_wakeup_send(state, state->ev,
 248                                     tevent_timeval_current_ofs(
 249                                             state->retry_interval.tv_sec,
 250                                             state->retry_interval.tv_usec));
 251         if (tevent_req_nomem(subreq, req)) {
 252                 return;
 253         }
 254         tevent_req_set_callback(subreq, reset_connections_batch, req);
 255 }
 256
 257 static int reset_connections_tickle_connection(
 258                                         uint8_t *keybuf, size_t keylen,
 259                                         uint8_t *databuf, size_t datalen,
 260                                         void *private_data)
 261 {
 262         struct reset_connections_state *state = talloc_get_type_abort(
 263                 private_data, struct reset_connections_state);
 264         struct ctdb_connection *conn;
 265         int ret;
 266
 267         if (keylen != sizeof(*conn)) {
 268                 DBG_WARNING("Unexpected data in connection hash\n");
 269                 return 0;
 270         }
 271
 272         conn = (struct ctdb_connection *)keybuf;
 273
 274         state->batch_count++;
 275         if (state->batch_count > state->batch_size) {
 276                 /* Terminate the traverse */
 277                 return 1;
 278         }
 279
 280         ret = ctdb_sys_send_tcp(&conn->server, &conn->client, 0, 0, 0);
 281         if (ret != 0) {
 282                 DBG_ERR("Error sending tickle ACK\n");
 283                 /* continue */
 284         }
 285
 286         return 0;
 287 }
 288
 289 static bool reset_connections_recv(struct tevent_req *req, int *perr)
 290 {
 291         int err;
 292
 293         if (tevent_req_is_unix_error(req, &err)) {
 294                 if (perr != NULL) {
 295                         *perr = err;
 296                 }
 297                 return false;
 298         }
 299
 300         return true;
 301 }
 302
 303 static void usage(const char *prog)
 304 {
 305         printf("usage: %s <interface> [ <srcip:port> <dstip:port> ]\n", prog);
 306         exit(1);
 307 }
 308
 309 int main(int argc, char **argv)
 310 {
 311         struct ctdb_connection conn;
 312         struct tevent_context *ev = NULL;
 313         struct TALLOC_CONTEXT *mem_ctx = NULL;
 314         struct ctdb_connection_list *conn_list = NULL;
 315         const char *t;
 316         struct tevent_req *req;
 317         int debug_level;
 318         bool status;
 319         int ret;
 320
 321         /* Set the debug level */
 322         t = getenv("CTDB_DEBUGLEVEL");
 323         if (t != NULL) {
 324                 if (debug_level_parse(t, &debug_level)) {
 325                         DEBUGLEVEL = debug_level;
 326                 } else {
 327                         DEBUGLEVEL = DEBUG_ERR;
 328                 }
 329         }
 330
 331         if (argc != 2 && argc != 4) {
 332                 usage(argv[0]);
 333         }
 334
 335         if (argc == 4) {
 336                 ret = ctdb_sock_addr_from_string(argv[2], &conn.client, true);
 337                 if (ret != 0) {
 338                         D_ERR("Bad IP:port '%s'\n", argv[2]);
 339                         goto fail;
 340                 }
 341
 342                 ret = ctdb_sock_addr_from_string(argv[3], &conn.server, true);
 343                 if (ret != 0) {
 344                         D_ERR("Bad IP:port '%s'\n", argv[3]);
 345                         goto fail;
 346                 }
 347
 348
 349                 conn_list = talloc_zero(mem_ctx, struct ctdb_connection_list);
 350                 if (conn_list == NULL) {
 351                         ret = ENOMEM;
 352                         DBG_ERR("Internal error (%s)\n", strerror(ret));
 353                         goto fail;
 354                 }
 355                 ret = ctdb_connection_list_add(conn_list, &conn);
 356                 if (ret != 0) {
 357                         DBG_ERR("Internal error (%s)\n", strerror(ret));
 358                         goto fail;
 359                 }
 360         } else {
 361                 ret = ctdb_connection_list_read(mem_ctx, true, &conn_list);
 362                 if (ret != 0) {
 363                         D_ERR("Unable to parse connections (%s)\n",
 364                               strerror(ret));
 365                         goto fail;
 366                 }
 367         }
 368
 369         mem_ctx = talloc_new(NULL);
 370         if (mem_ctx == NULL) {
 371                 DEBUG(DEBUG_ERR, (__location__ " out of memory\n"));
 372                 goto fail;
 373         }
 374
 375         ev = tevent_context_init(mem_ctx);
 376         if (ev == NULL) {
 377                 DEBUG(DEBUG_ERR, ("Failed to initialise tevent\n"));
 378                 goto fail;
 379         }
 380
 381         if (conn_list->num == 0) {
 382                 /* No connections, done! */
 383                 talloc_free(mem_ctx);
 384                 return 0;
 385         }
 386
 387         req = reset_connections_send(mem_ctx, ev, argv[1], conn_list);
 388         if (req == NULL) {
 389                 goto fail;
 390         }
 391
 392         tevent_req_poll(req, ev);
 393
 394         status = reset_connections_recv(req, &ret);
 395         if (! status) {
 396                 D_ERR("Failed to kill connections (%s)\n", strerror(ret));
 397                 goto fail;
 398         }
 399
 400         talloc_free(mem_ctx);
 401
 402         return 0;
 403
 404 fail:
 405         TALLOC_FREE(mem_ctx);
 406         return -1;
 407 }