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/key.h>
   9 #include <keys/ceph-type.h>
  10 #include <linux/module.h>
  11 #include <linux/mount.h>
  12 #include <linux/parser.h>
  13 #include <linux/sched.h>
  14 #include <linux/seq_file.h>
  15 #include <linux/slab.h>
  16 #include <linux/statfs.h>
  17 #include <linux/string.h>
  18
  19
  20 #include <linux/ceph/ceph_features.h>
  21 #include <linux/ceph/libceph.h>
  22 #include <linux/ceph/debugfs.h>
  23 #include <linux/ceph/decode.h>
  24 #include <linux/ceph/mon_client.h>
  25 #include <linux/ceph/auth.h>
  26 #include "crypto.h"
  27
  28
  29 /*
  30  * Module compatibility interface.  For now it doesn't do anything,
  31  * but its existence signals a certain level of functionality.
  32  *
  33  * The data buffer is used to pass information both to and from
  34  * libceph.  The return value indicates whether libceph determines
  35  * it is compatible with the caller (from another kernel module),
  36  * given the provided data.
  37  *
  38  * The data pointer can be null.
  39  */
  40 bool libceph_compatible(void *data)
  41 {
  42         return true;
  43 }
  44 EXPORT_SYMBOL(libceph_compatible);
  45
  46 /*
  47  * find filename portion of a path (/foo/bar/baz -> baz)
  48  */
  49 const char *ceph_file_part(const char *s, int len)
  50 {
  51         const char *e = s + len;
  52
  53         while (e != s && *(e-1) != '/')
  54                 e--;
  55         return e;
  56 }
  57 EXPORT_SYMBOL(ceph_file_part);
  58
  59 const char *ceph_msg_type_name(int type)
  60 {
  61         switch (type) {
  62         case CEPH_MSG_SHUTDOWN: return "shutdown";
  63         case CEPH_MSG_PING: return "ping";
  64         case CEPH_MSG_AUTH: return "auth";
  65         case CEPH_MSG_AUTH_REPLY: return "auth_reply";
  66         case CEPH_MSG_MON_MAP: return "mon_map";
  67         case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
  68         case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
  69         case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
  70         case CEPH_MSG_STATFS: return "statfs";
  71         case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
  72         case CEPH_MSG_MDS_MAP: return "mds_map";
  73         case CEPH_MSG_CLIENT_SESSION: return "client_session";
  74         case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
  75         case CEPH_MSG_CLIENT_REQUEST: return "client_request";
  76         case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
  77         case CEPH_MSG_CLIENT_REPLY: return "client_reply";
  78         case CEPH_MSG_CLIENT_CAPS: return "client_caps";
  79         case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
  80         case CEPH_MSG_CLIENT_SNAP: return "client_snap";
  81         case CEPH_MSG_CLIENT_LEASE: return "client_lease";
  82         case CEPH_MSG_OSD_MAP: return "osd_map";
  83         case CEPH_MSG_OSD_OP: return "osd_op";
  84         case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
  85         case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
  86         default: return "unknown";
  87         }
  88 }
  89 EXPORT_SYMBOL(ceph_msg_type_name);
  90
  91 /*
  92  * Initially learn our fsid, or verify an fsid matches.
  93  */
  94 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
  95 {
  96         if (client->have_fsid) {
  97                 if (ceph_fsid_compare(&client->fsid, fsid)) {
  98                         pr_err("bad fsid, had %pU got %pU",
  99                                &client->fsid, fsid);
 100                         return -1;
 101                 }
 102         } else {
 103                 memcpy(&client->fsid, fsid, sizeof(*fsid));
 104         }
 105         return 0;
 106 }
 107 EXPORT_SYMBOL(ceph_check_fsid);
 108
 109 static int strcmp_null(const char *s1, const char *s2)
 110 {
 111         if (!s1 && !s2)
 112                 return 0;
 113         if (s1 && !s2)
 114                 return -1;
 115         if (!s1 && s2)
 116                 return 1;
 117         return strcmp(s1, s2);
 118 }
 119
 120 int ceph_compare_options(struct ceph_options *new_opt,
 121                          struct ceph_client *client)
 122 {
 123         struct ceph_options *opt1 = new_opt;
 124         struct ceph_options *opt2 = client->options;
 125         int ofs = offsetof(struct ceph_options, mon_addr);
 126         int i;
 127         int ret;
 128
 129         ret = memcmp(opt1, opt2, ofs);
 130         if (ret)
 131                 return ret;
 132
 133         ret = strcmp_null(opt1->name, opt2->name);
 134         if (ret)
 135                 return ret;
 136
 137         if (opt1->key && !opt2->key)
 138                 return -1;
 139         if (!opt1->key && opt2->key)
 140                 return 1;
 141         if (opt1->key && opt2->key) {
 142                 if (opt1->key->type != opt2->key->type)
 143                         return -1;
 144                 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
 145                         return -1;
 146                 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
 147                         return -1;
 148                 if (opt1->key->len != opt2->key->len)
 149                         return -1;
 150                 if (opt1->key->key && !opt2->key->key)
 151                         return -1;
 152                 if (!opt1->key->key && opt2->key->key)
 153                         return 1;
 154                 if (opt1->key->key && opt2->key->key) {
 155                         ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
 156                         if (ret)
 157                                 return ret;
 158                 }
 159         }
 160
 161         /* any matching mon ip implies a match */
 162         for (i = 0; i < opt1->num_mon; i++) {
 163                 if (ceph_monmap_contains(client->monc.monmap,
 164                                  &opt1->mon_addr[i]))
 165                         return 0;
 166         }
 167         return -1;
 168 }
 169 EXPORT_SYMBOL(ceph_compare_options);
 170
 171
 172 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
 173 {
 174         int i = 0;
 175         char tmp[3];
 176         int err = -EINVAL;
 177         int d;
 178
 179         dout("parse_fsid '%s'\n", str);
 180         tmp[2] = 0;
 181         while (*str && i < 16) {
 182                 if (ispunct(*str)) {
 183                         str++;
 184                         continue;
 185                 }
 186                 if (!isxdigit(str[0]) || !isxdigit(str[1]))
 187                         break;
 188                 tmp[0] = str[0];
 189                 tmp[1] = str[1];
 190                 if (sscanf(tmp, "%x", &d) < 1)
 191                         break;
 192                 fsid->fsid[i] = d & 0xff;
 193                 i++;
 194                 str += 2;
 195         }
 196
 197         if (i == 16)
 198                 err = 0;
 199         dout("parse_fsid ret %d got fsid %pU", err, fsid);
 200         return err;
 201 }
 202
 203 /*
 204  * ceph options
 205  */
 206 enum {
 207         Opt_osdtimeout,
 208         Opt_osdkeepalivetimeout,
 209         Opt_mount_timeout,
 210         Opt_osd_idle_ttl,
 211         Opt_last_int,
 212         /* int args above */
 213         Opt_fsid,
 214         Opt_name,
 215         Opt_secret,
 216         Opt_key,
 217         Opt_ip,
 218         Opt_last_string,
 219         /* string args above */
 220         Opt_share,
 221         Opt_noshare,
 222         Opt_crc,
 223         Opt_nocrc,
 224 };
 225
 226 static match_table_t opt_tokens = {
 227         {Opt_osdtimeout, "osdtimeout=%d"},
 228         {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
 229         {Opt_mount_timeout, "mount_timeout=%d"},
 230         {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
 231         /* int args above */
 232         {Opt_fsid, "fsid=%s"},
 233         {Opt_name, "name=%s"},
 234         {Opt_secret, "secret=%s"},
 235         {Opt_key, "key=%s"},
 236         {Opt_ip, "ip=%s"},
 237         /* string args above */
 238         {Opt_share, "share"},
 239         {Opt_noshare, "noshare"},
 240         {Opt_crc, "crc"},
 241         {Opt_nocrc, "nocrc"},
 242         {-1, NULL}
 243 };
 244
 245 void ceph_destroy_options(struct ceph_options *opt)
 246 {
 247         dout("destroy_options %p\n", opt);
 248         kfree(opt->name);
 249         if (opt->key) {
 250                 ceph_crypto_key_destroy(opt->key);
 251                 kfree(opt->key);
 252         }
 253         kfree(opt->mon_addr);
 254         kfree(opt);
 255 }
 256 EXPORT_SYMBOL(ceph_destroy_options);
 257
 258 /* get secret from key store */
 259 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
 260         struct key *ukey;
 261         int key_err;
 262         int err = 0;
 263         struct ceph_crypto_key *ckey;
 264
 265         ukey = request_key(&key_type_ceph, name, NULL);
 266         if (!ukey || IS_ERR(ukey)) {
 267                 /* request_key errors don't map nicely to mount(2)
 268                    errors; don't even try, but still printk */
 269                 key_err = PTR_ERR(ukey);
 270                 switch (key_err) {
 271                 case -ENOKEY:
 272                         pr_warning("ceph: Mount failed due to key not found: %s\n", name);
 273                         break;
 274                 case -EKEYEXPIRED:
 275                         pr_warning("ceph: Mount failed due to expired key: %s\n", name);
 276                         break;
 277                 case -EKEYREVOKED:
 278                         pr_warning("ceph: Mount failed due to revoked key: %s\n", name);
 279                         break;
 280                 default:
 281                         pr_warning("ceph: Mount failed due to unknown key error"
 282                                " %d: %s\n", key_err, name);
 283                 }
 284                 err = -EPERM;
 285                 goto out;
 286         }
 287
 288         ckey = ukey->payload.data;
 289         err = ceph_crypto_key_clone(dst, ckey);
 290         if (err)
 291                 goto out_key;
 292         /* pass through, err is 0 */
 293
 294 out_key:
 295         key_put(ukey);
 296 out:
 297         return err;
 298 }
 299
 300 struct ceph_options *
 301 ceph_parse_options(char *options, const char *dev_name,
 302                         const char *dev_name_end,
 303                         int (*parse_extra_token)(char *c, void *private),
 304                         void *private)
 305 {
 306         struct ceph_options *opt;
 307         const char *c;
 308         int err = -ENOMEM;
 309         substring_t argstr[MAX_OPT_ARGS];
 310
 311         opt = kzalloc(sizeof(*opt), GFP_KERNEL);
 312         if (!opt)
 313                 return ERR_PTR(-ENOMEM);
 314         opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
 315                                 GFP_KERNEL);
 316         if (!opt->mon_addr)
 317                 goto out;
 318
 319         dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
 320              dev_name);
 321
 322         /* start with defaults */
 323         opt->flags = CEPH_OPT_DEFAULT;
 324         opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
 325         opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
 326         opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;   /* seconds */
 327
 328         /* get mon ip(s) */
 329         /* ip1[:port1][,ip2[:port2]...] */
 330         err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
 331                              CEPH_MAX_MON, &opt->num_mon);
 332         if (err < 0)
 333                 goto out;
 334
 335         /* parse mount options */
 336         while ((c = strsep(&options, ",")) != NULL) {
 337                 int token, intval, ret;
 338                 if (!*c)
 339                         continue;
 340                 err = -EINVAL;
 341                 token = match_token((char *)c, opt_tokens, argstr);
 342                 if (token < 0 && parse_extra_token) {
 343                         /* extra? */
 344                         err = parse_extra_token((char *)c, private);
 345                         if (err < 0) {
 346                                 pr_err("bad option at '%s'\n", c);
 347                                 goto out;
 348                         }
 349                         continue;
 350                 }
 351                 if (token < Opt_last_int) {
 352                         ret = match_int(&argstr[0], &intval);
 353                         if (ret < 0) {
 354                                 pr_err("bad mount option arg (not int) "
 355                                        "at '%s'\n", c);
 356                                 continue;
 357                         }
 358                         dout("got int token %d val %d\n", token, intval);
 359                 } else if (token > Opt_last_int && token < Opt_last_string) {
 360                         dout("got string token %d val %s\n", token,
 361                              argstr[0].from);
 362                 } else {
 363                         dout("got token %d\n", token);
 364                 }
 365                 switch (token) {
 366                 case Opt_ip:
 367                         err = ceph_parse_ips(argstr[0].from,
 368                                              argstr[0].to,
 369                                              &opt->my_addr,
 370                                              1, NULL);
 371                         if (err < 0)
 372                                 goto out;
 373                         opt->flags |= CEPH_OPT_MYIP;
 374                         break;
 375
 376                 case Opt_fsid:
 377                         err = parse_fsid(argstr[0].from, &opt->fsid);
 378                         if (err == 0)
 379                                 opt->flags |= CEPH_OPT_FSID;
 380                         break;
 381                 case Opt_name:
 382                         opt->name = kstrndup(argstr[0].from,
 383                                               argstr[0].to-argstr[0].from,
 384                                               GFP_KERNEL);
 385                         break;
 386                 case Opt_secret:
 387                         opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
 388                         if (!opt->key) {
 389                                 err = -ENOMEM;
 390                                 goto out;
 391                         }
 392                         err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
 393                         if (err < 0)
 394                                 goto out;
 395                         break;
 396                 case Opt_key:
 397                         opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
 398                         if (!opt->key) {
 399                                 err = -ENOMEM;
 400                                 goto out;
 401                         }
 402                         err = get_secret(opt->key, argstr[0].from);
 403                         if (err < 0)
 404                                 goto out;
 405                         break;
 406
 407                         /* misc */
 408                 case Opt_osdtimeout:
 409                         pr_warning("ignoring deprecated osdtimeout option\n");
 410                         break;
 411                 case Opt_osdkeepalivetimeout:
 412                         opt->osd_keepalive_timeout = intval;
 413                         break;
 414                 case Opt_osd_idle_ttl:
 415                         opt->osd_idle_ttl = intval;
 416                         break;
 417                 case Opt_mount_timeout:
 418                         opt->mount_timeout = intval;
 419                         break;
 420
 421                 case Opt_share:
 422                         opt->flags &= ~CEPH_OPT_NOSHARE;
 423                         break;
 424                 case Opt_noshare:
 425                         opt->flags |= CEPH_OPT_NOSHARE;
 426                         break;
 427
 428                 case Opt_crc:
 429                         opt->flags &= ~CEPH_OPT_NOCRC;
 430                         break;
 431                 case Opt_nocrc:
 432                         opt->flags |= CEPH_OPT_NOCRC;
 433                         break;
 434
 435                 default:
 436                         BUG_ON(token);
 437                 }
 438         }
 439
 440         /* success */
 441         return opt;
 442
 443 out:
 444         ceph_destroy_options(opt);
 445         return ERR_PTR(err);
 446 }
 447 EXPORT_SYMBOL(ceph_parse_options);
 448
 449 u64 ceph_client_id(struct ceph_client *client)
 450 {
 451         return client->monc.auth->global_id;
 452 }
 453 EXPORT_SYMBOL(ceph_client_id);
 454
 455 /*
 456  * create a fresh client instance
 457  */
 458 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private,
 459                                        unsigned int supported_features,
 460                                        unsigned int required_features)
 461 {
 462         struct ceph_client *client;
 463         struct ceph_entity_addr *myaddr = NULL;
 464         int err = -ENOMEM;
 465
 466         client = kzalloc(sizeof(*client), GFP_KERNEL);
 467         if (client == NULL)
 468                 return ERR_PTR(-ENOMEM);
 469
 470         client->private = private;
 471         client->options = opt;
 472
 473         mutex_init(&client->mount_mutex);
 474         init_waitqueue_head(&client->auth_wq);
 475         client->auth_err = 0;
 476
 477         client->extra_mon_dispatch = NULL;
 478         client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT |
 479                 supported_features;
 480         client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT |
 481                 required_features;
 482
 483         /* msgr */
 484         if (ceph_test_opt(client, MYIP))
 485                 myaddr = &client->options->my_addr;
 486         ceph_messenger_init(&client->msgr, myaddr,
 487                 client->supported_features,
 488                 client->required_features,
 489                 ceph_test_opt(client, NOCRC));
 490
 491         /* subsystems */
 492         err = ceph_monc_init(&client->monc, client);
 493         if (err < 0)
 494                 goto fail;
 495         err = ceph_osdc_init(&client->osdc, client);
 496         if (err < 0)
 497                 goto fail_monc;
 498
 499         return client;
 500
 501 fail_monc:
 502         ceph_monc_stop(&client->monc);
 503 fail:
 504         kfree(client);
 505         return ERR_PTR(err);
 506 }
 507 EXPORT_SYMBOL(ceph_create_client);
 508
 509 void ceph_destroy_client(struct ceph_client *client)
 510 {
 511         dout("destroy_client %p\n", client);
 512
 513         atomic_set(&client->msgr.stopping, 1);
 514
 515         /* unmount */
 516         ceph_osdc_stop(&client->osdc);
 517
 518         ceph_monc_stop(&client->monc);
 519
 520         ceph_debugfs_client_cleanup(client);
 521
 522         ceph_destroy_options(client->options);
 523
 524         kfree(client);
 525         dout("destroy_client %p done\n", client);
 526 }
 527 EXPORT_SYMBOL(ceph_destroy_client);
 528
 529 /*
 530  * true if we have the mon map (and have thus joined the cluster)
 531  */
 532 static int have_mon_and_osd_map(struct ceph_client *client)
 533 {
 534         return client->monc.monmap && client->monc.monmap->epoch &&
 535                client->osdc.osdmap && client->osdc.osdmap->epoch;
 536 }
 537
 538 /*
 539  * mount: join the ceph cluster, and open root directory.
 540  */
 541 int __ceph_open_session(struct ceph_client *client, unsigned long started)
 542 {
 543         int err;
 544         unsigned long timeout = client->options->mount_timeout * HZ;
 545
 546         /* open session, and wait for mon and osd maps */
 547         err = ceph_monc_open_session(&client->monc);
 548         if (err < 0)
 549                 return err;
 550
 551         while (!have_mon_and_osd_map(client)) {
 552                 err = -EIO;
 553                 if (timeout && time_after_eq(jiffies, started + timeout))
 554                         return err;
 555
 556                 /* wait */
 557                 dout("mount waiting for mon_map\n");
 558                 err = wait_event_interruptible_timeout(client->auth_wq,
 559                         have_mon_and_osd_map(client) || (client->auth_err < 0),
 560                         timeout);
 561                 if (err == -EINTR || err == -ERESTARTSYS)
 562                         return err;
 563                 if (client->auth_err < 0)
 564                         return client->auth_err;
 565         }
 566
 567         return 0;
 568 }
 569 EXPORT_SYMBOL(__ceph_open_session);
 570
 571
 572 int ceph_open_session(struct ceph_client *client)
 573 {
 574         int ret;
 575         unsigned long started = jiffies;  /* note the start time */
 576
 577         dout("open_session start\n");
 578         mutex_lock(&client->mount_mutex);
 579
 580         ret = __ceph_open_session(client, started);
 581
 582         mutex_unlock(&client->mount_mutex);
 583         return ret;
 584 }
 585 EXPORT_SYMBOL(ceph_open_session);
 586
 587
 588 static int __init init_ceph_lib(void)
 589 {
 590         int ret = 0;
 591
 592         ret = ceph_debugfs_init();
 593         if (ret < 0)
 594                 goto out;
 595
 596         ret = ceph_crypto_init();
 597         if (ret < 0)
 598                 goto out_debugfs;
 599
 600         ret = ceph_msgr_init();
 601         if (ret < 0)
 602                 goto out_crypto;
 603
 604         pr_info("loaded (mon/osd proto %d/%d)\n",
 605                 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
 606
 607         return 0;
 608
 609 out_crypto:
 610         ceph_crypto_shutdown();
 611 out_debugfs:
 612         ceph_debugfs_cleanup();
 613 out:
 614         return ret;
 615 }
 616
 617 static void __exit exit_ceph_lib(void)
 618 {
 619         dout("exit_ceph_lib\n");
 620         ceph_msgr_exit();
 621         ceph_crypto_shutdown();
 622         ceph_debugfs_cleanup();
 623 }
 624
 625 module_init(init_ceph_lib);
 626 module_exit(exit_ceph_lib);
 627
 628 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
 629 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
 630 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
 631 MODULE_DESCRIPTION("Ceph filesystem for Linux");
 632 MODULE_LICENSE("GPL");