net/ceph/ceph_common.c

   1 // SPDX-License-Identifier: GPL-2.0-only
   2
   3 #include <linux/ceph/ceph_debug.h>
   4 #include <linux/backing-dev.h>
   5 #include <linux/ctype.h>
   6 #include <linux/fs.h>
   7 #include <linux/inet.h>
   8 #include <linux/in6.h>
   9 #include <linux/key.h>
  10 #include <keys/ceph-type.h>
  11 #include <linux/module.h>
  12 #include <linux/mount.h>
  13 #include <linux/nsproxy.h>
  14 #include <linux/fs_parser.h>
  15 #include <linux/sched.h>
  16 #include <linux/sched/mm.h>
  17 #include <linux/seq_file.h>
  18 #include <linux/slab.h>
  19 #include <linux/statfs.h>
  20 #include <linux/string.h>
  21 #include <linux/vmalloc.h>
  22
  23
  24 #include <linux/ceph/ceph_features.h>
  25 #include <linux/ceph/libceph.h>
  26 #include <linux/ceph/debugfs.h>
  27 #include <linux/ceph/decode.h>
  28 #include <linux/ceph/mon_client.h>
  29 #include <linux/ceph/auth.h>
  30 #include "crypto.h"
  31
  32
  33 /*
  34  * Module compatibility interface.  For now it doesn't do anything,
  35  * but its existence signals a certain level of functionality.
  36  *
  37  * The data buffer is used to pass information both to and from
  38  * libceph.  The return value indicates whether libceph determines
  39  * it is compatible with the caller (from another kernel module),
  40  * given the provided data.
  41  *
  42  * The data pointer can be null.
  43  */
  44 bool libceph_compatible(void *data)
  45 {
  46         return true;
  47 }
  48 EXPORT_SYMBOL(libceph_compatible);
  49
  50 static int param_get_supported_features(char *buffer,
  51                                         const struct kernel_param *kp)
  52 {
  53         return sprintf(buffer, "0x%llx", CEPH_FEATURES_SUPPORTED_DEFAULT);
  54 }
  55 static const struct kernel_param_ops param_ops_supported_features = {
  56         .get = param_get_supported_features,
  57 };
  58 module_param_cb(supported_features, &param_ops_supported_features, NULL,
  59                 0444);
  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_MON_GET_VERSION: return "mon_get_version";
  75         case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply";
  76         case CEPH_MSG_MDS_MAP: return "mds_map";
  77         case CEPH_MSG_FS_MAP_USER: return "fs_map_user";
  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_QUOTA: return "client_quota";
  86         case CEPH_MSG_CLIENT_SNAP: return "client_snap";
  87         case CEPH_MSG_CLIENT_LEASE: return "client_lease";
  88         case CEPH_MSG_POOLOP_REPLY: return "poolop_reply";
  89         case CEPH_MSG_POOLOP: return "poolop";
  90         case CEPH_MSG_MON_COMMAND: return "mon_command";
  91         case CEPH_MSG_MON_COMMAND_ACK: return "mon_command_ack";
  92         case CEPH_MSG_OSD_MAP: return "osd_map";
  93         case CEPH_MSG_OSD_OP: return "osd_op";
  94         case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
  95         case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
  96         case CEPH_MSG_OSD_BACKOFF: return "osd_backoff";
  97         default: return "unknown";
  98         }
  99 }
 100 EXPORT_SYMBOL(ceph_msg_type_name);
 101
 102 /*
 103  * Initially learn our fsid, or verify an fsid matches.
 104  */
 105 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
 106 {
 107         if (client->have_fsid) {
 108                 if (ceph_fsid_compare(&client->fsid, fsid)) {
 109                         pr_err("bad fsid, had %pU got %pU",
 110                                &client->fsid, fsid);
 111                         return -1;
 112                 }
 113         } else {
 114                 memcpy(&client->fsid, fsid, sizeof(*fsid));
 115         }
 116         return 0;
 117 }
 118 EXPORT_SYMBOL(ceph_check_fsid);
 119
 120 static int strcmp_null(const char *s1, const char *s2)
 121 {
 122         if (!s1 && !s2)
 123                 return 0;
 124         if (s1 && !s2)
 125                 return -1;
 126         if (!s1 && s2)
 127                 return 1;
 128         return strcmp(s1, s2);
 129 }
 130
 131 int ceph_compare_options(struct ceph_options *new_opt,
 132                          struct ceph_client *client)
 133 {
 134         struct ceph_options *opt1 = new_opt;
 135         struct ceph_options *opt2 = client->options;
 136         int ofs = offsetof(struct ceph_options, mon_addr);
 137         int i;
 138         int ret;
 139
 140         /*
 141          * Don't bother comparing options if network namespaces don't
 142          * match.
 143          */
 144         if (!net_eq(current->nsproxy->net_ns, read_pnet(&client->msgr.net)))
 145                 return -1;
 146
 147         ret = memcmp(opt1, opt2, ofs);
 148         if (ret)
 149                 return ret;
 150
 151         ret = strcmp_null(opt1->name, opt2->name);
 152         if (ret)
 153                 return ret;
 154
 155         if (opt1->key && !opt2->key)
 156                 return -1;
 157         if (!opt1->key && opt2->key)
 158                 return 1;
 159         if (opt1->key && opt2->key) {
 160                 if (opt1->key->type != opt2->key->type)
 161                         return -1;
 162                 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
 163                         return -1;
 164                 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
 165                         return -1;
 166                 if (opt1->key->len != opt2->key->len)
 167                         return -1;
 168                 if (opt1->key->key && !opt2->key->key)
 169                         return -1;
 170                 if (!opt1->key->key && opt2->key->key)
 171                         return 1;
 172                 if (opt1->key->key && opt2->key->key) {
 173                         ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
 174                         if (ret)
 175                                 return ret;
 176                 }
 177         }
 178
 179         /* any matching mon ip implies a match */
 180         for (i = 0; i < opt1->num_mon; i++) {
 181                 if (ceph_monmap_contains(client->monc.monmap,
 182                                  &opt1->mon_addr[i]))
 183                         return 0;
 184         }
 185         return -1;
 186 }
 187 EXPORT_SYMBOL(ceph_compare_options);
 188
 189 /*
 190  * kvmalloc() doesn't fall back to the vmalloc allocator unless flags are
 191  * compatible with (a superset of) GFP_KERNEL.  This is because while the
 192  * actual pages are allocated with the specified flags, the page table pages
 193  * are always allocated with GFP_KERNEL.
 194  *
 195  * ceph_kvmalloc() may be called with GFP_KERNEL, GFP_NOFS or GFP_NOIO.
 196  */
 197 void *ceph_kvmalloc(size_t size, gfp_t flags)
 198 {
 199         void *p;
 200
 201         if ((flags & (__GFP_IO | __GFP_FS)) == (__GFP_IO | __GFP_FS)) {
 202                 p = kvmalloc(size, flags);
 203         } else if ((flags & (__GFP_IO | __GFP_FS)) == __GFP_IO) {
 204                 unsigned int nofs_flag = memalloc_nofs_save();
 205                 p = kvmalloc(size, GFP_KERNEL);
 206                 memalloc_nofs_restore(nofs_flag);
 207         } else {
 208                 unsigned int noio_flag = memalloc_noio_save();
 209                 p = kvmalloc(size, GFP_KERNEL);
 210                 memalloc_noio_restore(noio_flag);
 211         }
 212
 213         return p;
 214 }
 215
 216 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
 217 {
 218         int i = 0;
 219         char tmp[3];
 220         int err = -EINVAL;
 221         int d;
 222
 223         dout("parse_fsid '%s'\n", str);
 224         tmp[2] = 0;
 225         while (*str && i < 16) {
 226                 if (ispunct(*str)) {
 227                         str++;
 228                         continue;
 229                 }
 230                 if (!isxdigit(str[0]) || !isxdigit(str[1]))
 231                         break;
 232                 tmp[0] = str[0];
 233                 tmp[1] = str[1];
 234                 if (sscanf(tmp, "%x", &d) < 1)
 235                         break;
 236                 fsid->fsid[i] = d & 0xff;
 237                 i++;
 238                 str += 2;
 239         }
 240
 241         if (i == 16)
 242                 err = 0;
 243         dout("parse_fsid ret %d got fsid %pU\n", err, fsid);
 244         return err;
 245 }
 246
 247 /*
 248  * ceph options
 249  */
 250 enum {
 251         Opt_osdtimeout,
 252         Opt_osdkeepalivetimeout,
 253         Opt_mount_timeout,
 254         Opt_osd_idle_ttl,
 255         Opt_osd_request_timeout,
 256         /* int args above */
 257         Opt_fsid,
 258         Opt_name,
 259         Opt_secret,
 260         Opt_key,
 261         Opt_ip,
 262         /* string args above */
 263         Opt_share,
 264         Opt_crc,
 265         Opt_cephx_require_signatures,
 266         Opt_cephx_sign_messages,
 267         Opt_tcp_nodelay,
 268         Opt_abort_on_full,
 269 };
 270
 271 static const struct fs_parameter_spec ceph_parameters[] = {
 272         fsparam_flag    ("abort_on_full",               Opt_abort_on_full),
 273         fsparam_flag_no ("cephx_require_signatures",    Opt_cephx_require_signatures),
 274         fsparam_flag_no ("cephx_sign_messages",         Opt_cephx_sign_messages),
 275         fsparam_flag_no ("crc",                         Opt_crc),
 276         fsparam_string  ("fsid",                        Opt_fsid),
 277         fsparam_string  ("ip",                          Opt_ip),
 278         fsparam_string  ("key",                         Opt_key),
 279         fsparam_u32     ("mount_timeout",               Opt_mount_timeout),
 280         fsparam_string  ("name",                        Opt_name),
 281         fsparam_u32     ("osd_idle_ttl",                Opt_osd_idle_ttl),
 282         fsparam_u32     ("osd_request_timeout",         Opt_osd_request_timeout),
 283         fsparam_u32     ("osdkeepalive",                Opt_osdkeepalivetimeout),
 284         __fsparam       (fs_param_is_s32, "osdtimeout", Opt_osdtimeout,
 285                          fs_param_deprecated, NULL),
 286         fsparam_string  ("secret",                      Opt_secret),
 287         fsparam_flag_no ("share",                       Opt_share),
 288         fsparam_flag_no ("tcp_nodelay",                 Opt_tcp_nodelay),
 289         {}
 290 };
 291
 292 struct ceph_options *ceph_alloc_options(void)
 293 {
 294         struct ceph_options *opt;
 295
 296         opt = kzalloc(sizeof(*opt), GFP_KERNEL);
 297         if (!opt)
 298                 return NULL;
 299
 300         opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
 301                                 GFP_KERNEL);
 302         if (!opt->mon_addr) {
 303                 kfree(opt);
 304                 return NULL;
 305         }
 306
 307         opt->flags = CEPH_OPT_DEFAULT;
 308         opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
 309         opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
 310         opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
 311         opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
 312         return opt;
 313 }
 314 EXPORT_SYMBOL(ceph_alloc_options);
 315
 316 void ceph_destroy_options(struct ceph_options *opt)
 317 {
 318         dout("destroy_options %p\n", opt);
 319         if (!opt)
 320                 return;
 321
 322         kfree(opt->name);
 323         if (opt->key) {
 324                 ceph_crypto_key_destroy(opt->key);
 325                 kfree(opt->key);
 326         }
 327         kfree(opt->mon_addr);
 328         kfree(opt);
 329 }
 330 EXPORT_SYMBOL(ceph_destroy_options);
 331
 332 /* get secret from key store */
 333 static int get_secret(struct ceph_crypto_key *dst, const char *name,
 334                       struct p_log *log)
 335 {
 336         struct key *ukey;
 337         int key_err;
 338         int err = 0;
 339         struct ceph_crypto_key *ckey;
 340
 341         ukey = request_key(&key_type_ceph, name, NULL);
 342         if (IS_ERR(ukey)) {
 343                 /* request_key errors don't map nicely to mount(2)
 344                    errors; don't even try, but still printk */
 345                 key_err = PTR_ERR(ukey);
 346                 switch (key_err) {
 347                 case -ENOKEY:
 348                         error_plog(log, "Failed due to key not found: %s",
 349                                name);
 350                         break;
 351                 case -EKEYEXPIRED:
 352                         error_plog(log, "Failed due to expired key: %s",
 353                                name);
 354                         break;
 355                 case -EKEYREVOKED:
 356                         error_plog(log, "Failed due to revoked key: %s",
 357                                name);
 358                         break;
 359                 default:
 360                         error_plog(log, "Failed due to key error %d: %s",
 361                                key_err, name);
 362                 }
 363                 err = -EPERM;
 364                 goto out;
 365         }
 366
 367         ckey = ukey->payload.data[0];
 368         err = ceph_crypto_key_clone(dst, ckey);
 369         if (err)
 370                 goto out_key;
 371         /* pass through, err is 0 */
 372
 373 out_key:
 374         key_put(ukey);
 375 out:
 376         return err;
 377 }
 378
 379 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
 380                        struct fc_log *l)
 381 {
 382         struct p_log log = {.prefix = "libceph", .log = l};
 383         int ret;
 384
 385         /* ip1[:port1][,ip2[:port2]...] */
 386         ret = ceph_parse_ips(buf, buf + len, opt->mon_addr, CEPH_MAX_MON,
 387                              &opt->num_mon);
 388         if (ret) {
 389                 error_plog(&log, "Failed to parse monitor IPs: %d", ret);
 390                 return ret;
 391         }
 392
 393         return 0;
 394 }
 395 EXPORT_SYMBOL(ceph_parse_mon_ips);
 396
 397 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
 398                      struct fc_log *l)
 399 {
 400         struct fs_parse_result result;
 401         int token, err;
 402         struct p_log log = {.prefix = "libceph", .log = l};
 403
 404         token = __fs_parse(&log, ceph_parameters, param, &result);
 405         dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
 406         if (token < 0)
 407                 return token;
 408
 409         switch (token) {
 410         case Opt_ip:
 411                 err = ceph_parse_ips(param->string,
 412                                      param->string + param->size,
 413                                      &opt->my_addr,
 414                                      1, NULL);
 415                 if (err) {
 416                         error_plog(&log, "Failed to parse ip: %d", err);
 417                         return err;
 418                 }
 419                 opt->flags |= CEPH_OPT_MYIP;
 420                 break;
 421
 422         case Opt_fsid:
 423                 err = parse_fsid(param->string, &opt->fsid);
 424                 if (err) {
 425                         error_plog(&log, "Failed to parse fsid: %d", err);
 426                         return err;
 427                 }
 428                 opt->flags |= CEPH_OPT_FSID;
 429                 break;
 430         case Opt_name:
 431                 kfree(opt->name);
 432                 opt->name = param->string;
 433                 param->string = NULL;
 434                 break;
 435         case Opt_secret:
 436                 ceph_crypto_key_destroy(opt->key);
 437                 kfree(opt->key);
 438
 439                 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
 440                 if (!opt->key)
 441                         return -ENOMEM;
 442                 err = ceph_crypto_key_unarmor(opt->key, param->string);
 443                 if (err) {
 444                         error_plog(&log, "Failed to parse secret: %d", err);
 445                         return err;
 446                 }
 447                 break;
 448         case Opt_key:
 449                 ceph_crypto_key_destroy(opt->key);
 450                 kfree(opt->key);
 451
 452                 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
 453                 if (!opt->key)
 454                         return -ENOMEM;
 455                 return get_secret(opt->key, param->string, &log);
 456
 457         case Opt_osdtimeout:
 458                 warn_plog(&log, "Ignoring osdtimeout");
 459                 break;
 460         case Opt_osdkeepalivetimeout:
 461                 /* 0 isn't well defined right now, reject it */
 462                 if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
 463                         goto out_of_range;
 464                 opt->osd_keepalive_timeout =
 465                     msecs_to_jiffies(result.uint_32 * 1000);
 466                 break;
 467         case Opt_osd_idle_ttl:
 468                 /* 0 isn't well defined right now, reject it */
 469                 if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
 470                         goto out_of_range;
 471                 opt->osd_idle_ttl = msecs_to_jiffies(result.uint_32 * 1000);
 472                 break;
 473         case Opt_mount_timeout:
 474                 /* 0 is "wait forever" (i.e. infinite timeout) */
 475                 if (result.uint_32 > INT_MAX / 1000)
 476                         goto out_of_range;
 477                 opt->mount_timeout = msecs_to_jiffies(result.uint_32 * 1000);
 478                 break;
 479         case Opt_osd_request_timeout:
 480                 /* 0 is "wait forever" (i.e. infinite timeout) */
 481                 if (result.uint_32 > INT_MAX / 1000)
 482                         goto out_of_range;
 483                 opt->osd_request_timeout =
 484                     msecs_to_jiffies(result.uint_32 * 1000);
 485                 break;
 486
 487         case Opt_share:
 488                 if (!result.negated)
 489                         opt->flags &= ~CEPH_OPT_NOSHARE;
 490                 else
 491                         opt->flags |= CEPH_OPT_NOSHARE;
 492                 break;
 493         case Opt_crc:
 494                 if (!result.negated)
 495                         opt->flags &= ~CEPH_OPT_NOCRC;
 496                 else
 497                         opt->flags |= CEPH_OPT_NOCRC;
 498                 break;
 499         case Opt_cephx_require_signatures:
 500                 if (!result.negated)
 501                         opt->flags &= ~CEPH_OPT_NOMSGAUTH;
 502                 else
 503                         opt->flags |= CEPH_OPT_NOMSGAUTH;
 504                 break;
 505         case Opt_cephx_sign_messages:
 506                 if (!result.negated)
 507                         opt->flags &= ~CEPH_OPT_NOMSGSIGN;
 508                 else
 509                         opt->flags |= CEPH_OPT_NOMSGSIGN;
 510                 break;
 511         case Opt_tcp_nodelay:
 512                 if (!result.negated)
 513                         opt->flags |= CEPH_OPT_TCP_NODELAY;
 514                 else
 515                         opt->flags &= ~CEPH_OPT_TCP_NODELAY;
 516                 break;
 517
 518         case Opt_abort_on_full:
 519                 opt->flags |= CEPH_OPT_ABORT_ON_FULL;
 520                 break;
 521
 522         default:
 523                 BUG();
 524         }
 525
 526         return 0;
 527
 528 out_of_range:
 529         return inval_plog(&log, "%s out of range", param->key);
 530 }
 531 EXPORT_SYMBOL(ceph_parse_param);
 532
 533 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
 534                               bool show_all)
 535 {
 536         struct ceph_options *opt = client->options;
 537         size_t pos = m->count;
 538
 539         if (opt->name) {
 540                 seq_puts(m, "name=");
 541                 seq_escape(m, opt->name, ", \t\n\\");
 542                 seq_putc(m, ',');
 543         }
 544         if (opt->key)
 545                 seq_puts(m, "secret=<hidden>,");
 546
 547         if (opt->flags & CEPH_OPT_FSID)
 548                 seq_printf(m, "fsid=%pU,", &opt->fsid);
 549         if (opt->flags & CEPH_OPT_NOSHARE)
 550                 seq_puts(m, "noshare,");
 551         if (opt->flags & CEPH_OPT_NOCRC)
 552                 seq_puts(m, "nocrc,");
 553         if (opt->flags & CEPH_OPT_NOMSGAUTH)
 554                 seq_puts(m, "nocephx_require_signatures,");
 555         if (opt->flags & CEPH_OPT_NOMSGSIGN)
 556                 seq_puts(m, "nocephx_sign_messages,");
 557         if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
 558                 seq_puts(m, "notcp_nodelay,");
 559         if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
 560                 seq_puts(m, "abort_on_full,");
 561
 562         if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
 563                 seq_printf(m, "mount_timeout=%d,",
 564                            jiffies_to_msecs(opt->mount_timeout) / 1000);
 565         if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
 566                 seq_printf(m, "osd_idle_ttl=%d,",
 567                            jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
 568         if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
 569                 seq_printf(m, "osdkeepalivetimeout=%d,",
 570                     jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
 571         if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
 572                 seq_printf(m, "osd_request_timeout=%d,",
 573                            jiffies_to_msecs(opt->osd_request_timeout) / 1000);
 574
 575         /* drop redundant comma */
 576         if (m->count != pos)
 577                 m->count--;
 578
 579         return 0;
 580 }
 581 EXPORT_SYMBOL(ceph_print_client_options);
 582
 583 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
 584 {
 585         return &client->msgr.inst.addr;
 586 }
 587 EXPORT_SYMBOL(ceph_client_addr);
 588
 589 u64 ceph_client_gid(struct ceph_client *client)
 590 {
 591         return client->monc.auth->global_id;
 592 }
 593 EXPORT_SYMBOL(ceph_client_gid);
 594
 595 /*
 596  * create a fresh client instance
 597  */
 598 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
 599 {
 600         struct ceph_client *client;
 601         struct ceph_entity_addr *myaddr = NULL;
 602         int err;
 603
 604         err = wait_for_random_bytes();
 605         if (err < 0)
 606                 return ERR_PTR(err);
 607
 608         client = kzalloc(sizeof(*client), GFP_KERNEL);
 609         if (client == NULL)
 610                 return ERR_PTR(-ENOMEM);
 611
 612         client->private = private;
 613         client->options = opt;
 614
 615         mutex_init(&client->mount_mutex);
 616         init_waitqueue_head(&client->auth_wq);
 617         client->auth_err = 0;
 618
 619         client->extra_mon_dispatch = NULL;
 620         client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
 621         client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
 622
 623         if (!ceph_test_opt(client, NOMSGAUTH))
 624                 client->required_features |= CEPH_FEATURE_MSG_AUTH;
 625
 626         /* msgr */
 627         if (ceph_test_opt(client, MYIP))
 628                 myaddr = &client->options->my_addr;
 629
 630         ceph_messenger_init(&client->msgr, myaddr);
 631
 632         /* subsystems */
 633         err = ceph_monc_init(&client->monc, client);
 634         if (err < 0)
 635                 goto fail;
 636         err = ceph_osdc_init(&client->osdc, client);
 637         if (err < 0)
 638                 goto fail_monc;
 639
 640         return client;
 641
 642 fail_monc:
 643         ceph_monc_stop(&client->monc);
 644 fail:
 645         ceph_messenger_fini(&client->msgr);
 646         kfree(client);
 647         return ERR_PTR(err);
 648 }
 649 EXPORT_SYMBOL(ceph_create_client);
 650
 651 void ceph_destroy_client(struct ceph_client *client)
 652 {
 653         dout("destroy_client %p\n", client);
 654
 655         atomic_set(&client->msgr.stopping, 1);
 656
 657         /* unmount */
 658         ceph_osdc_stop(&client->osdc);
 659         ceph_monc_stop(&client->monc);
 660         ceph_messenger_fini(&client->msgr);
 661
 662         ceph_debugfs_client_cleanup(client);
 663
 664         ceph_destroy_options(client->options);
 665
 666         kfree(client);
 667         dout("destroy_client %p done\n", client);
 668 }
 669 EXPORT_SYMBOL(ceph_destroy_client);
 670
 671 void ceph_reset_client_addr(struct ceph_client *client)
 672 {
 673         ceph_messenger_reset_nonce(&client->msgr);
 674         ceph_monc_reopen_session(&client->monc);
 675         ceph_osdc_reopen_osds(&client->osdc);
 676 }
 677 EXPORT_SYMBOL(ceph_reset_client_addr);
 678
 679 /*
 680  * true if we have the mon map (and have thus joined the cluster)
 681  */
 682 static bool have_mon_and_osd_map(struct ceph_client *client)
 683 {
 684         return client->monc.monmap && client->monc.monmap->epoch &&
 685                client->osdc.osdmap && client->osdc.osdmap->epoch;
 686 }
 687
 688 /*
 689  * mount: join the ceph cluster, and open root directory.
 690  */
 691 int __ceph_open_session(struct ceph_client *client, unsigned long started)
 692 {
 693         unsigned long timeout = client->options->mount_timeout;
 694         long err;
 695
 696         /* open session, and wait for mon and osd maps */
 697         err = ceph_monc_open_session(&client->monc);
 698         if (err < 0)
 699                 return err;
 700
 701         while (!have_mon_and_osd_map(client)) {
 702                 if (timeout && time_after_eq(jiffies, started + timeout))
 703                         return -ETIMEDOUT;
 704
 705                 /* wait */
 706                 dout("mount waiting for mon_map\n");
 707                 err = wait_event_interruptible_timeout(client->auth_wq,
 708                         have_mon_and_osd_map(client) || (client->auth_err < 0),
 709                         ceph_timeout_jiffies(timeout));
 710                 if (err < 0)
 711                         return err;
 712                 if (client->auth_err < 0)
 713                         return client->auth_err;
 714         }
 715
 716         pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
 717                 &client->fsid);
 718         ceph_debugfs_client_init(client);
 719
 720         return 0;
 721 }
 722 EXPORT_SYMBOL(__ceph_open_session);
 723
 724 int ceph_open_session(struct ceph_client *client)
 725 {
 726         int ret;
 727         unsigned long started = jiffies;  /* note the start time */
 728
 729         dout("open_session start\n");
 730         mutex_lock(&client->mount_mutex);
 731
 732         ret = __ceph_open_session(client, started);
 733
 734         mutex_unlock(&client->mount_mutex);
 735         return ret;
 736 }
 737 EXPORT_SYMBOL(ceph_open_session);
 738
 739 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
 740                                 unsigned long timeout)
 741 {
 742         u64 newest_epoch;
 743         int ret;
 744
 745         ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
 746         if (ret)
 747                 return ret;
 748
 749         if (client->osdc.osdmap->epoch >= newest_epoch)
 750                 return 0;
 751
 752         ceph_osdc_maybe_request_map(&client->osdc);
 753         return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
 754 }
 755 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
 756
 757 static int __init init_ceph_lib(void)
 758 {
 759         int ret = 0;
 760
 761         ceph_debugfs_init();
 762
 763         ret = ceph_crypto_init();
 764         if (ret < 0)
 765                 goto out_debugfs;
 766
 767         ret = ceph_msgr_init();
 768         if (ret < 0)
 769                 goto out_crypto;
 770
 771         ret = ceph_osdc_setup();
 772         if (ret < 0)
 773                 goto out_msgr;
 774
 775         pr_info("loaded (mon/osd proto %d/%d)\n",
 776                 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
 777
 778         return 0;
 779
 780 out_msgr:
 781         ceph_msgr_exit();
 782 out_crypto:
 783         ceph_crypto_shutdown();
 784 out_debugfs:
 785         ceph_debugfs_cleanup();
 786         return ret;
 787 }
 788
 789 static void __exit exit_ceph_lib(void)
 790 {
 791         dout("exit_ceph_lib\n");
 792         WARN_ON(!ceph_strings_empty());
 793
 794         ceph_osdc_cleanup();
 795         ceph_msgr_exit();
 796         ceph_crypto_shutdown();
 797         ceph_debugfs_cleanup();
 798 }
 799
 800 module_init(init_ceph_lib);
 801 module_exit(exit_ceph_lib);
 802
 803 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
 804 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
 805 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
 806 MODULE_DESCRIPTION("Ceph core library");
 807 MODULE_LICENSE("GPL");