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