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/vmalloc.h>
  19 #include <linux/nsproxy.h>
  20 #include <net/net_namespace.h>
  21
  22
  23 #include <linux/ceph/ceph_features.h>
  24 #include <linux/ceph/libceph.h>
  25 #include <linux/ceph/debugfs.h>
  26 #include <linux/ceph/decode.h>
  27 #include <linux/ceph/mon_client.h>
  28 #include <linux/ceph/auth.h>
  29 #include "crypto.h"
  30
  31
  32 /*
  33  * Module compatibility interface.  For now it doesn't do anything,
  34  * but its existence signals a certain level of functionality.
  35  *
  36  * The data buffer is used to pass information both to and from
  37  * libceph.  The return value indicates whether libceph determines
  38  * it is compatible with the caller (from another kernel module),
  39  * given the provided data.
  40  *
  41  * The data pointer can be null.
  42  */
  43 bool libceph_compatible(void *data)
  44 {
  45         return true;
  46 }
  47 EXPORT_SYMBOL(libceph_compatible);
  48
  49 /*
  50  * find filename portion of a path (/foo/bar/baz -> baz)
  51  */
  52 const char *ceph_file_part(const char *s, int len)
  53 {
  54         const char *e = s + len;
  55
  56         while (e != s && *(e-1) != '/')
  57                 e--;
  58         return e;
  59 }
  60 EXPORT_SYMBOL(ceph_file_part);
  61
  62 const char *ceph_msg_type_name(int type)
  63 {
  64         switch (type) {
  65         case CEPH_MSG_SHUTDOWN: return "shutdown";
  66         case CEPH_MSG_PING: return "ping";
  67         case CEPH_MSG_AUTH: return "auth";
  68         case CEPH_MSG_AUTH_REPLY: return "auth_reply";
  69         case CEPH_MSG_MON_MAP: return "mon_map";
  70         case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
  71         case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
  72         case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
  73         case CEPH_MSG_STATFS: return "statfs";
  74         case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
  75         case CEPH_MSG_MON_GET_VERSION: return "mon_get_version";
  76         case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
  77         case CEPH_MSG_MDS_MAP: return "mds_map";
  78         case CEPH_MSG_CLIENT_SESSION: return "client_session";
  79         case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
  80         case CEPH_MSG_CLIENT_REQUEST: return "client_request";
  81         case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
  82         case CEPH_MSG_CLIENT_REPLY: return "client_reply";
  83         case CEPH_MSG_CLIENT_CAPS: return "client_caps";
  84         case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
  85         case CEPH_MSG_CLIENT_SNAP: return "client_snap";
  86         case CEPH_MSG_CLIENT_LEASE: return "client_lease";
  87         case CEPH_MSG_OSD_MAP: return "osd_map";
  88         case CEPH_MSG_OSD_OP: return "osd_op";
  89         case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
  90         case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
  91         default: return "unknown";
  92         }
  93 }
  94 EXPORT_SYMBOL(ceph_msg_type_name);
  95
  96 /*
  97  * Initially learn our fsid, or verify an fsid matches.
  98  */
  99 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
 100 {
 101         if (client->have_fsid) {
 102                 if (ceph_fsid_compare(&client->fsid, fsid)) {
 103                         pr_err("bad fsid, had %pU got %pU",
 104                                &client->fsid, fsid);
 105                         return -1;
 106                 }
 107         } else {
 108                 memcpy(&client->fsid, fsid, sizeof(*fsid));
 109         }
 110         return 0;
 111 }
 112 EXPORT_SYMBOL(ceph_check_fsid);
 113
 114 static int strcmp_null(const char *s1, const char *s2)
 115 {
 116         if (!s1 && !s2)
 117                 return 0;
 118         if (s1 && !s2)
 119                 return -1;
 120         if (!s1 && s2)
 121                 return 1;
 122         return strcmp(s1, s2);
 123 }
 124
 125 int ceph_compare_options(struct ceph_options *new_opt,
 126                          struct ceph_client *client)
 127 {
 128         struct ceph_options *opt1 = new_opt;
 129         struct ceph_options *opt2 = client->options;
 130         int ofs = offsetof(struct ceph_options, mon_addr);
 131         int i;
 132         int ret;
 133
 134         ret = memcmp(opt1, opt2, ofs);
 135         if (ret)
 136                 return ret;
 137
 138         ret = strcmp_null(opt1->name, opt2->name);
 139         if (ret)
 140                 return ret;
 141
 142         if (opt1->key && !opt2->key)
 143                 return -1;
 144         if (!opt1->key && opt2->key)
 145                 return 1;
 146         if (opt1->key && opt2->key) {
 147                 if (opt1->key->type != opt2->key->type)
 148                         return -1;
 149                 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
 150                         return -1;
 151                 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
 152                         return -1;
 153                 if (opt1->key->len != opt2->key->len)
 154                         return -1;
 155                 if (opt1->key->key && !opt2->key->key)
 156                         return -1;
 157                 if (!opt1->key->key && opt2->key->key)
 158                         return 1;
 159                 if (opt1->key->key && opt2->key->key) {
 160                         ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
 161                         if (ret)
 162                                 return ret;
 163                 }
 164         }
 165
 166         /* any matching mon ip implies a match */
 167         for (i = 0; i < opt1->num_mon; i++) {
 168                 if (ceph_monmap_contains(client->monc.monmap,
 169                                  &opt1->mon_addr[i]))
 170                         return 0;
 171         }
 172         return -1;
 173 }
 174 EXPORT_SYMBOL(ceph_compare_options);
 175
 176 void *ceph_kvmalloc(size_t size, gfp_t flags)
 177 {
 178         if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) {
 179                 void *ptr = kmalloc(size, flags | __GFP_NOWARN);
 180                 if (ptr)
 181                         return ptr;
 182         }
 183
 184         return __vmalloc(size, flags | __GFP_HIGHMEM, PAGE_KERNEL);
 185 }
 186
 187
 188 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
 189 {
 190         int i = 0;
 191         char tmp[3];
 192         int err = -EINVAL;
 193         int d;
 194
 195         dout("parse_fsid '%s'\n", str);
 196         tmp[2] = 0;
 197         while (*str && i < 16) {
 198                 if (ispunct(*str)) {
 199                         str++;
 200                         continue;
 201                 }
 202                 if (!isxdigit(str[0]) || !isxdigit(str[1]))
 203                         break;
 204                 tmp[0] = str[0];
 205                 tmp[1] = str[1];
 206                 if (sscanf(tmp, "%x", &d) < 1)
 207                         break;
 208                 fsid->fsid[i] = d & 0xff;
 209                 i++;
 210                 str += 2;
 211         }
 212
 213         if (i == 16)
 214                 err = 0;
 215         dout("parse_fsid ret %d got fsid %pU", err, fsid);
 216         return err;
 217 }
 218
 219 /*
 220  * ceph options
 221  */
 222 enum {
 223         Opt_osdtimeout,
 224         Opt_osdkeepalivetimeout,
 225         Opt_mount_timeout,
 226         Opt_osd_idle_ttl,
 227         Opt_last_int,
 228         /* int args above */
 229         Opt_fsid,
 230         Opt_name,
 231         Opt_secret,
 232         Opt_key,
 233         Opt_ip,
 234         Opt_last_string,
 235         /* string args above */
 236         Opt_share,
 237         Opt_noshare,
 238         Opt_crc,
 239         Opt_nocrc,
 240         Opt_cephx_require_signatures,
 241         Opt_nocephx_require_signatures,
 242 };
 243
 244 static match_table_t opt_tokens = {
 245         {Opt_osdtimeout, "osdtimeout=%d"},
 246         {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
 247         {Opt_mount_timeout, "mount_timeout=%d"},
 248         {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
 249         /* int args above */
 250         {Opt_fsid, "fsid=%s"},
 251         {Opt_name, "name=%s"},
 252         {Opt_secret, "secret=%s"},
 253         {Opt_key, "key=%s"},
 254         {Opt_ip, "ip=%s"},
 255         /* string args above */
 256         {Opt_share, "share"},
 257         {Opt_noshare, "noshare"},
 258         {Opt_crc, "crc"},
 259         {Opt_nocrc, "nocrc"},
 260         {Opt_cephx_require_signatures, "cephx_require_signatures"},
 261         {Opt_nocephx_require_signatures, "nocephx_require_signatures"},
 262         {-1, NULL}
 263 };
 264
 265 void ceph_destroy_options(struct ceph_options *opt)
 266 {
 267         dout("destroy_options %p\n", opt);
 268         kfree(opt->name);
 269         if (opt->key) {
 270                 ceph_crypto_key_destroy(opt->key);
 271                 kfree(opt->key);
 272         }
 273         kfree(opt->mon_addr);
 274         kfree(opt);
 275 }
 276 EXPORT_SYMBOL(ceph_destroy_options);
 277
 278 /* get secret from key store */
 279 static int get_secret(struct ceph_crypto_key *dst, const char *name) {
 280         struct key *ukey;
 281         int key_err;
 282         int err = 0;
 283         struct ceph_crypto_key *ckey;
 284
 285         ukey = request_key(&key_type_ceph, name, NULL);
 286         if (!ukey || IS_ERR(ukey)) {
 287                 /* request_key errors don't map nicely to mount(2)
 288                    errors; don't even try, but still printk */
 289                 key_err = PTR_ERR(ukey);
 290                 switch (key_err) {
 291                 case -ENOKEY:
 292                         pr_warn("ceph: Mount failed due to key not found: %s\n",
 293                                 name);
 294                         break;
 295                 case -EKEYEXPIRED:
 296                         pr_warn("ceph: Mount failed due to expired key: %s\n",
 297                                 name);
 298                         break;
 299                 case -EKEYREVOKED:
 300                         pr_warn("ceph: Mount failed due to revoked key: %s\n",
 301                                 name);
 302                         break;
 303                 default:
 304                         pr_warn("ceph: Mount failed due to unknown key error %d: %s\n",
 305                                 key_err, name);
 306                 }
 307                 err = -EPERM;
 308                 goto out;
 309         }
 310
 311         ckey = ukey->payload.data;
 312         err = ceph_crypto_key_clone(dst, ckey);
 313         if (err)
 314                 goto out_key;
 315         /* pass through, err is 0 */
 316
 317 out_key:
 318         key_put(ukey);
 319 out:
 320         return err;
 321 }
 322
 323 struct ceph_options *
 324 ceph_parse_options(char *options, const char *dev_name,
 325                         const char *dev_name_end,
 326                         int (*parse_extra_token)(char *c, void *private),
 327                         void *private)
 328 {
 329         struct ceph_options *opt;
 330         const char *c;
 331         int err = -ENOMEM;
 332         substring_t argstr[MAX_OPT_ARGS];
 333
 334         if (current->nsproxy->net_ns != &init_net)
 335                 return ERR_PTR(-EINVAL);
 336
 337         opt = kzalloc(sizeof(*opt), GFP_KERNEL);
 338         if (!opt)
 339                 return ERR_PTR(-ENOMEM);
 340         opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
 341                                 GFP_KERNEL);
 342         if (!opt->mon_addr)
 343                 goto out;
 344
 345         dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
 346              dev_name);
 347
 348         /* start with defaults */
 349         opt->flags = CEPH_OPT_DEFAULT;
 350         opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
 351         opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
 352         opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;   /* seconds */
 353
 354         /* get mon ip(s) */
 355         /* ip1[:port1][,ip2[:port2]...] */
 356         err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
 357                              CEPH_MAX_MON, &opt->num_mon);
 358         if (err < 0)
 359                 goto out;
 360
 361         /* parse mount options */
 362         while ((c = strsep(&options, ",")) != NULL) {
 363                 int token, intval, ret;
 364                 if (!*c)
 365                         continue;
 366                 err = -EINVAL;
 367                 token = match_token((char *)c, opt_tokens, argstr);
 368                 if (token < 0 && parse_extra_token) {
 369                         /* extra? */
 370                         err = parse_extra_token((char *)c, private);
 371                         if (err < 0) {
 372                                 pr_err("bad option at '%s'\n", c);
 373                                 goto out;
 374                         }
 375                         continue;
 376                 }
 377                 if (token < Opt_last_int) {
 378                         ret = match_int(&argstr[0], &intval);
 379                         if (ret < 0) {
 380                                 pr_err("bad mount option arg (not int) "
 381                                        "at '%s'\n", c);
 382                                 continue;
 383                         }
 384                         dout("got int token %d val %d\n", token, intval);
 385                 } else if (token > Opt_last_int && token < Opt_last_string) {
 386                         dout("got string token %d val %s\n", token,
 387                              argstr[0].from);
 388                 } else {
 389                         dout("got token %d\n", token);
 390                 }
 391                 switch (token) {
 392                 case Opt_ip:
 393                         err = ceph_parse_ips(argstr[0].from,
 394                                              argstr[0].to,
 395                                              &opt->my_addr,
 396                                              1, NULL);
 397                         if (err < 0)
 398                                 goto out;
 399                         opt->flags |= CEPH_OPT_MYIP;
 400                         break;
 401
 402                 case Opt_fsid:
 403                         err = parse_fsid(argstr[0].from, &opt->fsid);
 404                         if (err == 0)
 405                                 opt->flags |= CEPH_OPT_FSID;
 406                         break;
 407                 case Opt_name:
 408                         opt->name = kstrndup(argstr[0].from,
 409                                               argstr[0].to-argstr[0].from,
 410                                               GFP_KERNEL);
 411                         break;
 412                 case Opt_secret:
 413                         opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
 414                         if (!opt->key) {
 415                                 err = -ENOMEM;
 416                                 goto out;
 417                         }
 418                         err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
 419                         if (err < 0)
 420                                 goto out;
 421                         break;
 422                 case Opt_key:
 423                         opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
 424                         if (!opt->key) {
 425                                 err = -ENOMEM;
 426                                 goto out;
 427                         }
 428                         err = get_secret(opt->key, argstr[0].from);
 429                         if (err < 0)
 430                                 goto out;
 431                         break;
 432
 433                         /* misc */
 434                 case Opt_osdtimeout:
 435                         pr_warn("ignoring deprecated osdtimeout option\n");
 436                         break;
 437                 case Opt_osdkeepalivetimeout:
 438                         opt->osd_keepalive_timeout = intval;
 439                         break;
 440                 case Opt_osd_idle_ttl:
 441                         opt->osd_idle_ttl = intval;
 442                         break;
 443                 case Opt_mount_timeout:
 444                         opt->mount_timeout = intval;
 445                         break;
 446
 447                 case Opt_share:
 448                         opt->flags &= ~CEPH_OPT_NOSHARE;
 449                         break;
 450                 case Opt_noshare:
 451                         opt->flags |= CEPH_OPT_NOSHARE;
 452                         break;
 453
 454                 case Opt_crc:
 455                         opt->flags &= ~CEPH_OPT_NOCRC;
 456                         break;
 457                 case Opt_nocrc:
 458                         opt->flags |= CEPH_OPT_NOCRC;
 459                         break;
 460                 case Opt_cephx_require_signatures:
 461                         opt->flags &= ~CEPH_OPT_NOMSGAUTH;
 462                         break;
 463                 case Opt_nocephx_require_signatures:
 464                         opt->flags |= CEPH_OPT_NOMSGAUTH;
 465                         break;
 466
 467                 default:
 468                         BUG_ON(token);
 469                 }
 470         }
 471
 472         /* success */
 473         return opt;
 474
 475 out:
 476         ceph_destroy_options(opt);
 477         return ERR_PTR(err);
 478 }
 479 EXPORT_SYMBOL(ceph_parse_options);
 480
 481 u64 ceph_client_id(struct ceph_client *client)
 482 {
 483         return client->monc.auth->global_id;
 484 }
 485 EXPORT_SYMBOL(ceph_client_id);
 486
 487 /*
 488  * create a fresh client instance
 489  */
 490 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private,
 491                                        u64 supported_features,
 492                                        u64 required_features)
 493 {
 494         struct ceph_client *client;
 495         struct ceph_entity_addr *myaddr = NULL;
 496         int err = -ENOMEM;
 497
 498         client = kzalloc(sizeof(*client), GFP_KERNEL);
 499         if (client == NULL)
 500                 return ERR_PTR(-ENOMEM);
 501
 502         client->private = private;
 503         client->options = opt;
 504
 505         mutex_init(&client->mount_mutex);
 506         init_waitqueue_head(&client->auth_wq);
 507         client->auth_err = 0;
 508
 509         if (!ceph_test_opt(client, NOMSGAUTH))
 510                 required_features |= CEPH_FEATURE_MSG_AUTH;
 511
 512         client->extra_mon_dispatch = NULL;
 513         client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT |
 514                 supported_features;
 515         client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT |
 516                 required_features;
 517
 518         /* msgr */
 519         if (ceph_test_opt(client, MYIP))
 520                 myaddr = &client->options->my_addr;
 521         ceph_messenger_init(&client->msgr, myaddr,
 522                 client->supported_features,
 523                 client->required_features,
 524                 ceph_test_opt(client, NOCRC));
 525
 526         /* subsystems */
 527         err = ceph_monc_init(&client->monc, client);
 528         if (err < 0)
 529                 goto fail;
 530         err = ceph_osdc_init(&client->osdc, client);
 531         if (err < 0)
 532                 goto fail_monc;
 533
 534         return client;
 535
 536 fail_monc:
 537         ceph_monc_stop(&client->monc);
 538 fail:
 539         kfree(client);
 540         return ERR_PTR(err);
 541 }
 542 EXPORT_SYMBOL(ceph_create_client);
 543
 544 void ceph_destroy_client(struct ceph_client *client)
 545 {
 546         dout("destroy_client %p\n", client);
 547
 548         atomic_set(&client->msgr.stopping, 1);
 549
 550         /* unmount */
 551         ceph_osdc_stop(&client->osdc);
 552
 553         ceph_monc_stop(&client->monc);
 554
 555         ceph_debugfs_client_cleanup(client);
 556
 557         ceph_destroy_options(client->options);
 558
 559         kfree(client);
 560         dout("destroy_client %p done\n", client);
 561 }
 562 EXPORT_SYMBOL(ceph_destroy_client);
 563
 564 /*
 565  * true if we have the mon map (and have thus joined the cluster)
 566  */
 567 static int have_mon_and_osd_map(struct ceph_client *client)
 568 {
 569         return client->monc.monmap && client->monc.monmap->epoch &&
 570                client->osdc.osdmap && client->osdc.osdmap->epoch;
 571 }
 572
 573 /*
 574  * mount: join the ceph cluster, and open root directory.
 575  */
 576 int __ceph_open_session(struct ceph_client *client, unsigned long started)
 577 {
 578         int err;
 579         unsigned long timeout = client->options->mount_timeout * HZ;
 580
 581         /* open session, and wait for mon and osd maps */
 582         err = ceph_monc_open_session(&client->monc);
 583         if (err < 0)
 584                 return err;
 585
 586         while (!have_mon_and_osd_map(client)) {
 587                 err = -EIO;
 588                 if (timeout && time_after_eq(jiffies, started + timeout))
 589                         return err;
 590
 591                 /* wait */
 592                 dout("mount waiting for mon_map\n");
 593                 err = wait_event_interruptible_timeout(client->auth_wq,
 594                         have_mon_and_osd_map(client) || (client->auth_err < 0),
 595                         timeout);
 596                 if (err == -EINTR || err == -ERESTARTSYS)
 597                         return err;
 598                 if (client->auth_err < 0)
 599                         return client->auth_err;
 600         }
 601
 602         return 0;
 603 }
 604 EXPORT_SYMBOL(__ceph_open_session);
 605
 606
 607 int ceph_open_session(struct ceph_client *client)
 608 {
 609         int ret;
 610         unsigned long started = jiffies;  /* note the start time */
 611
 612         dout("open_session start\n");
 613         mutex_lock(&client->mount_mutex);
 614
 615         ret = __ceph_open_session(client, started);
 616
 617         mutex_unlock(&client->mount_mutex);
 618         return ret;
 619 }
 620 EXPORT_SYMBOL(ceph_open_session);
 621
 622
 623 static int __init init_ceph_lib(void)
 624 {
 625         int ret = 0;
 626
 627         ret = ceph_debugfs_init();
 628         if (ret < 0)
 629                 goto out;
 630
 631         ret = ceph_crypto_init();
 632         if (ret < 0)
 633                 goto out_debugfs;
 634
 635         ret = ceph_msgr_init();
 636         if (ret < 0)
 637                 goto out_crypto;
 638
 639         ret = ceph_osdc_setup();
 640         if (ret < 0)
 641                 goto out_msgr;
 642
 643         pr_info("loaded (mon/osd proto %d/%d)\n",
 644                 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
 645
 646         return 0;
 647
 648 out_msgr:
 649         ceph_msgr_exit();
 650 out_crypto:
 651         ceph_crypto_shutdown();
 652 out_debugfs:
 653         ceph_debugfs_cleanup();
 654 out:
 655         return ret;
 656 }
 657
 658 static void __exit exit_ceph_lib(void)
 659 {
 660         dout("exit_ceph_lib\n");
 661         ceph_osdc_cleanup();
 662         ceph_msgr_exit();
 663         ceph_crypto_shutdown();
 664         ceph_debugfs_cleanup();
 665 }
 666
 667 module_init(init_ceph_lib);
 668 module_exit(exit_ceph_lib);
 669
 670 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
 671 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
 672 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
 673 MODULE_DESCRIPTION("Ceph filesystem for Linux");
 674 MODULE_LICENSE("GPL");