net/ceph/ceph_common.c

   1
   2 #include <linux/ceph/ceph_debug.h>
   3 #include <linux/backing-dev.h>
   4 #include <linux/ctype.h>
   5 #include <linux/fs.h>
   6 #include <linux/inet.h>
   7 #include <linux/in6.h>
   8 #include <linux/module.h>
   9 #include <linux/mount.h>
  10 #include <linux/parser.h>
  11 #include <linux/sched.h>
  12 #include <linux/seq_file.h>
  13 #include <linux/slab.h>
  14 #include <linux/statfs.h>
  15 #include <linux/string.h>
  16
  17
  18 #include <linux/ceph/libceph.h>
  19 #include <linux/ceph/debugfs.h>
  20 #include <linux/ceph/decode.h>
  21 #include <linux/ceph/mon_client.h>
  22 #include <linux/ceph/auth.h>
  23 #include "crypto.h"
  24
  25
  26
  27 /*
  28  * find filename portion of a path (/foo/bar/baz -> baz)
  29  */
  30 const char *ceph_file_part(const char *s, int len)
  31 {
  32         const char *e = s + len;
  33
  34         while (e != s && *(e-1) != '/')
  35                 e--;
  36         return e;
  37 }
  38 EXPORT_SYMBOL(ceph_file_part);
  39
  40 const char *ceph_msg_type_name(int type)
  41 {
  42         switch (type) {
  43         case CEPH_MSG_SHUTDOWN: return "shutdown";
  44         case CEPH_MSG_PING: return "ping";
  45         case CEPH_MSG_AUTH: return "auth";
  46         case CEPH_MSG_AUTH_REPLY: return "auth_reply";
  47         case CEPH_MSG_MON_MAP: return "mon_map";
  48         case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
  49         case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
  50         case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
  51         case CEPH_MSG_STATFS: return "statfs";
  52         case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
  53         case CEPH_MSG_MDS_MAP: return "mds_map";
  54         case CEPH_MSG_CLIENT_SESSION: return "client_session";
  55         case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
  56         case CEPH_MSG_CLIENT_REQUEST: return "client_request";
  57         case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
  58         case CEPH_MSG_CLIENT_REPLY: return "client_reply";
  59         case CEPH_MSG_CLIENT_CAPS: return "client_caps";
  60         case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
  61         case CEPH_MSG_CLIENT_SNAP: return "client_snap";
  62         case CEPH_MSG_CLIENT_LEASE: return "client_lease";
  63         case CEPH_MSG_OSD_MAP: return "osd_map";
  64         case CEPH_MSG_OSD_OP: return "osd_op";
  65         case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
  66         case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
  67         default: return "unknown";
  68         }
  69 }
  70 EXPORT_SYMBOL(ceph_msg_type_name);
  71
  72 /*
  73  * Initially learn our fsid, or verify an fsid matches.
  74  */
  75 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
  76 {
  77         if (client->have_fsid) {
  78                 if (ceph_fsid_compare(&client->fsid, fsid)) {
  79                         pr_err("bad fsid, had %pU got %pU",
  80                                &client->fsid, fsid);
  81                         return -1;
  82                 }
  83         } else {
  84                 pr_info("client%lld fsid %pU\n", ceph_client_id(client), fsid);
  85                 memcpy(&client->fsid, fsid, sizeof(*fsid));
  86                 ceph_debugfs_client_init(client);
  87                 client->have_fsid = true;
  88         }
  89         return 0;
  90 }
  91 EXPORT_SYMBOL(ceph_check_fsid);
  92
  93 static int strcmp_null(const char *s1, const char *s2)
  94 {
  95         if (!s1 && !s2)
  96                 return 0;
  97         if (s1 && !s2)
  98                 return -1;
  99         if (!s1 && s2)
 100                 return 1;
 101         return strcmp(s1, s2);
 102 }
 103
 104 int ceph_compare_options(struct ceph_options *new_opt,
 105                          struct ceph_client *client)
 106 {
 107         struct ceph_options *opt1 = new_opt;
 108         struct ceph_options *opt2 = client->options;
 109         int ofs = offsetof(struct ceph_options, mon_addr);
 110         int i;
 111         int ret;
 112
 113         ret = memcmp(opt1, opt2, ofs);
 114         if (ret)
 115                 return ret;
 116
 117         ret = strcmp_null(opt1->name, opt2->name);
 118         if (ret)
 119                 return ret;
 120
 121         if (opt1->key && !opt2->key)
 122                 return -1;
 123         if (!opt1->key && opt2->key)
 124                 return 1;
 125         if (opt1->key && opt2->key) {
 126                 if (opt1->key->type != opt2->key->type)
 127                         return -1;
 128                 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
 129                         return -1;
 130                 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
 131                         return -1;
 132                 if (opt1->key->len != opt2->key->len)
 133                         return -1;
 134                 if (opt1->key->key && !opt2->key->key)
 135                         return -1;
 136                 if (!opt1->key->key && opt2->key->key)
 137                         return 1;
 138                 if (opt1->key->key && opt2->key->key) {
 139                         ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
 140                         if (ret)
 141                                 return ret;
 142                 }
 143         }
 144
 145         /* any matching mon ip implies a match */
 146         for (i = 0; i < opt1->num_mon; i++) {
 147                 if (ceph_monmap_contains(client->monc.monmap,
 148                                  &opt1->mon_addr[i]))
 149                         return 0;
 150         }
 151         return -1;
 152 }
 153 EXPORT_SYMBOL(ceph_compare_options);
 154
 155
 156 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
 157 {
 158         int i = 0;
 159         char tmp[3];
 160         int err = -EINVAL;
 161         int d;
 162
 163         dout("parse_fsid '%s'\n", str);
 164         tmp[2] = 0;
 165         while (*str && i < 16) {
 166                 if (ispunct(*str)) {
 167                         str++;
 168                         continue;
 169                 }
 170                 if (!isxdigit(str[0]) || !isxdigit(str[1]))
 171                         break;
 172                 tmp[0] = str[0];
 173                 tmp[1] = str[1];
 174                 if (sscanf(tmp, "%x", &d) < 1)
 175                         break;
 176                 fsid->fsid[i] = d & 0xff;
 177                 i++;
 178                 str += 2;
 179         }
 180
 181         if (i == 16)
 182                 err = 0;
 183         dout("parse_fsid ret %d got fsid %pU", err, fsid);
 184         return err;
 185 }
 186
 187 /*
 188  * ceph options
 189  */
 190 enum {
 191         Opt_osdtimeout,
 192         Opt_osdkeepalivetimeout,
 193         Opt_mount_timeout,
 194         Opt_osd_idle_ttl,
 195         Opt_last_int,
 196         /* int args above */
 197         Opt_fsid,
 198         Opt_name,
 199         Opt_secret,
 200         Opt_ip,
 201         Opt_last_string,
 202         /* string args above */
 203         Opt_noshare,
 204         Opt_nocrc,
 205 };
 206
 207 static match_table_t opt_tokens = {
 208         {Opt_osdtimeout, "osdtimeout=%d"},
 209         {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
 210         {Opt_mount_timeout, "mount_timeout=%d"},
 211         {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
 212         /* int args above */
 213         {Opt_fsid, "fsid=%s"},
 214         {Opt_name, "name=%s"},
 215         {Opt_secret, "secret=%s"},
 216         {Opt_ip, "ip=%s"},
 217         /* string args above */
 218         {Opt_noshare, "noshare"},
 219         {Opt_nocrc, "nocrc"},
 220         {-1, NULL}
 221 };
 222
 223 void ceph_destroy_options(struct ceph_options *opt)
 224 {
 225         dout("destroy_options %p\n", opt);
 226         kfree(opt->name);
 227         if (opt->key) {
 228                 ceph_crypto_key_destroy(opt->key);
 229                 kfree(opt->key);
 230         }
 231         kfree(opt);
 232 }
 233 EXPORT_SYMBOL(ceph_destroy_options);
 234
 235 int ceph_parse_options(struct ceph_options **popt, char *options,
 236                        const char *dev_name, const char *dev_name_end,
 237                        int (*parse_extra_token)(char *c, void *private),
 238                        void *private)
 239 {
 240         struct ceph_options *opt;
 241         const char *c;
 242         int err = -ENOMEM;
 243         substring_t argstr[MAX_OPT_ARGS];
 244
 245         opt = kzalloc(sizeof(*opt), GFP_KERNEL);
 246         if (!opt)
 247                 return err;
 248         opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
 249                                 GFP_KERNEL);
 250         if (!opt->mon_addr)
 251                 goto out;
 252
 253         dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
 254              dev_name);
 255
 256         /* start with defaults */
 257         opt->flags = CEPH_OPT_DEFAULT;
 258         opt->osd_timeout = CEPH_OSD_TIMEOUT_DEFAULT;
 259         opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
 260         opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
 261         opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;   /* seconds */
 262
 263         /* get mon ip(s) */
 264         /* ip1[:port1][,ip2[:port2]...] */
 265         err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
 266                              CEPH_MAX_MON, &opt->num_mon);
 267         if (err < 0)
 268                 goto out;
 269
 270         /* parse mount options */
 271         while ((c = strsep(&options, ",")) != NULL) {
 272                 int token, intval, ret;
 273                 if (!*c)
 274                         continue;
 275                 err = -EINVAL;
 276                 token = match_token((char *)c, opt_tokens, argstr);
 277                 if (token < 0 && parse_extra_token) {
 278                         /* extra? */
 279                         err = parse_extra_token((char *)c, private);
 280                         if (err < 0) {
 281                                 pr_err("bad option at '%s'\n", c);
 282                                 goto out;
 283                         }
 284                         continue;
 285                 }
 286                 if (token < Opt_last_int) {
 287                         ret = match_int(&argstr[0], &intval);
 288                         if (ret < 0) {
 289                                 pr_err("bad mount option arg (not int) "
 290                                        "at '%s'\n", c);
 291                                 continue;
 292                         }
 293                         dout("got int token %d val %d\n", token, intval);
 294                 } else if (token > Opt_last_int && token < Opt_last_string) {
 295                         dout("got string token %d val %s\n", token,
 296                              argstr[0].from);
 297                 } else {
 298                         dout("got token %d\n", token);
 299                 }
 300                 switch (token) {
 301                 case Opt_ip:
 302                         err = ceph_parse_ips(argstr[0].from,
 303                                              argstr[0].to,
 304                                              &opt->my_addr,
 305                                              1, NULL);
 306                         if (err < 0)
 307                                 goto out;
 308                         opt->flags |= CEPH_OPT_MYIP;
 309                         break;
 310
 311                 case Opt_fsid:
 312                         err = parse_fsid(argstr[0].from, &opt->fsid);
 313                         if (err == 0)
 314                                 opt->flags |= CEPH_OPT_FSID;
 315                         break;
 316                 case Opt_name:
 317                         opt->name = kstrndup(argstr[0].from,
 318                                               argstr[0].to-argstr[0].from,
 319                                               GFP_KERNEL);
 320                         break;
 321                 case Opt_secret:
 322                         opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
 323                         if (!opt->key) {
 324                                 err = -ENOMEM;
 325                                 goto out;
 326                         }
 327                         err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
 328                         if (err < 0)
 329                                 goto out;
 330                         break;
 331
 332                         /* misc */
 333                 case Opt_osdtimeout:
 334                         opt->osd_timeout = intval;
 335                         break;
 336                 case Opt_osdkeepalivetimeout:
 337                         opt->osd_keepalive_timeout = intval;
 338                         break;
 339                 case Opt_osd_idle_ttl:
 340                         opt->osd_idle_ttl = intval;
 341                         break;
 342                 case Opt_mount_timeout:
 343                         opt->mount_timeout = intval;
 344                         break;
 345
 346                 case Opt_noshare:
 347                         opt->flags |= CEPH_OPT_NOSHARE;
 348                         break;
 349
 350                 case Opt_nocrc:
 351                         opt->flags |= CEPH_OPT_NOCRC;
 352                         break;
 353
 354                 default:
 355                         BUG_ON(token);
 356                 }
 357         }
 358
 359         /* success */
 360         *popt = opt;
 361         return 0;
 362
 363 out:
 364         ceph_destroy_options(opt);
 365         return err;
 366 }
 367 EXPORT_SYMBOL(ceph_parse_options);
 368
 369 u64 ceph_client_id(struct ceph_client *client)
 370 {
 371         return client->monc.auth->global_id;
 372 }
 373 EXPORT_SYMBOL(ceph_client_id);
 374
 375 /*
 376  * create a fresh client instance
 377  */
 378 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
 379 {
 380         struct ceph_client *client;
 381         int err = -ENOMEM;
 382
 383         client = kzalloc(sizeof(*client), GFP_KERNEL);
 384         if (client == NULL)
 385                 return ERR_PTR(-ENOMEM);
 386
 387         client->private = private;
 388         client->options = opt;
 389
 390         mutex_init(&client->mount_mutex);
 391         init_waitqueue_head(&client->auth_wq);
 392         client->auth_err = 0;
 393
 394         client->extra_mon_dispatch = NULL;
 395         client->supported_features = CEPH_FEATURE_SUPPORTED_DEFAULT;
 396         client->required_features = CEPH_FEATURE_REQUIRED_DEFAULT;
 397
 398         client->msgr = NULL;
 399
 400         /* subsystems */
 401         err = ceph_monc_init(&client->monc, client);
 402         if (err < 0)
 403                 goto fail;
 404         err = ceph_osdc_init(&client->osdc, client);
 405         if (err < 0)
 406                 goto fail_monc;
 407
 408         return client;
 409
 410 fail_monc:
 411         ceph_monc_stop(&client->monc);
 412 fail:
 413         kfree(client);
 414         return ERR_PTR(err);
 415 }
 416 EXPORT_SYMBOL(ceph_create_client);
 417
 418 void ceph_destroy_client(struct ceph_client *client)
 419 {
 420         dout("destroy_client %p\n", client);
 421
 422         /* unmount */
 423         ceph_osdc_stop(&client->osdc);
 424
 425         /*
 426          * make sure osd connections close out before destroying the
 427          * auth module, which is needed to free those connections'
 428          * ceph_authorizers.
 429          */
 430         ceph_msgr_flush();
 431
 432         ceph_monc_stop(&client->monc);
 433
 434         ceph_debugfs_client_cleanup(client);
 435
 436         if (client->msgr)
 437                 ceph_messenger_destroy(client->msgr);
 438
 439         ceph_destroy_options(client->options);
 440
 441         kfree(client);
 442         dout("destroy_client %p done\n", client);
 443 }
 444 EXPORT_SYMBOL(ceph_destroy_client);
 445
 446 /*
 447  * true if we have the mon map (and have thus joined the cluster)
 448  */
 449 static int have_mon_and_osd_map(struct ceph_client *client)
 450 {
 451         return client->monc.monmap && client->monc.monmap->epoch &&
 452                client->osdc.osdmap && client->osdc.osdmap->epoch;
 453 }
 454
 455 /*
 456  * mount: join the ceph cluster, and open root directory.
 457  */
 458 int __ceph_open_session(struct ceph_client *client, unsigned long started)
 459 {
 460         struct ceph_entity_addr *myaddr = NULL;
 461         int err;
 462         unsigned long timeout = client->options->mount_timeout * HZ;
 463
 464         /* initialize the messenger */
 465         if (client->msgr == NULL) {
 466                 if (ceph_test_opt(client, MYIP))
 467                         myaddr = &client->options->my_addr;
 468                 client->msgr = ceph_messenger_create(myaddr,
 469                                         client->supported_features,
 470                                         client->required_features);
 471                 if (IS_ERR(client->msgr)) {
 472                         client->msgr = NULL;
 473                         return PTR_ERR(client->msgr);
 474                 }
 475                 client->msgr->nocrc = ceph_test_opt(client, NOCRC);
 476         }
 477
 478         /* open session, and wait for mon and osd maps */
 479         err = ceph_monc_open_session(&client->monc);
 480         if (err < 0)
 481                 return err;
 482
 483         while (!have_mon_and_osd_map(client)) {
 484                 err = -EIO;
 485                 if (timeout && time_after_eq(jiffies, started + timeout))
 486                         return err;
 487
 488                 /* wait */
 489                 dout("mount waiting for mon_map\n");
 490                 err = wait_event_interruptible_timeout(client->auth_wq,
 491                         have_mon_and_osd_map(client) || (client->auth_err < 0),
 492                         timeout);
 493                 if (err == -EINTR || err == -ERESTARTSYS)
 494                         return err;
 495                 if (client->auth_err < 0)
 496                         return client->auth_err;
 497         }
 498
 499         return 0;
 500 }
 501 EXPORT_SYMBOL(__ceph_open_session);
 502
 503
 504 int ceph_open_session(struct ceph_client *client)
 505 {
 506         int ret;
 507         unsigned long started = jiffies;  /* note the start time */
 508
 509         dout("open_session start\n");
 510         mutex_lock(&client->mount_mutex);
 511
 512         ret = __ceph_open_session(client, started);
 513
 514         mutex_unlock(&client->mount_mutex);
 515         return ret;
 516 }
 517 EXPORT_SYMBOL(ceph_open_session);
 518
 519
 520 static int __init init_ceph_lib(void)
 521 {
 522         int ret = 0;
 523
 524         ret = ceph_debugfs_init();
 525         if (ret < 0)
 526                 goto out;
 527
 528         ret = ceph_msgr_init();
 529         if (ret < 0)
 530                 goto out_debugfs;
 531
 532         pr_info("loaded (mon/osd proto %d/%d, osdmap %d/%d %d/%d)\n",
 533                 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL,
 534                 CEPH_OSDMAP_VERSION, CEPH_OSDMAP_VERSION_EXT,
 535                 CEPH_OSDMAP_INC_VERSION, CEPH_OSDMAP_INC_VERSION_EXT);
 536
 537         return 0;
 538
 539 out_debugfs:
 540         ceph_debugfs_cleanup();
 541 out:
 542         return ret;
 543 }
 544
 545 static void __exit exit_ceph_lib(void)
 546 {
 547         dout("exit_ceph_lib\n");
 548         ceph_msgr_exit();
 549         ceph_debugfs_cleanup();
 550 }
 551
 552 module_init(init_ceph_lib);
 553 module_exit(exit_ceph_lib);
 554
 555 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
 556 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
 557 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
 558 MODULE_DESCRIPTION("Ceph filesystem for Linux");
 559 MODULE_LICENSE("GPL");