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         ret = ceph_compare_crush_locs(&opt1->crush_locs, &opt2->crush_locs);
 180         if (ret)
 181                 return ret;
 182
 183         /* any matching mon ip implies a match */
 184         for (i = 0; i < opt1->num_mon; i++) {
 185                 if (ceph_monmap_contains(client->monc.monmap,
 186                                  &opt1->mon_addr[i]))
 187                         return 0;
 188         }
 189         return -1;
 190 }
 191 EXPORT_SYMBOL(ceph_compare_options);
 192
 193 /*
 194  * kvmalloc() doesn't fall back to the vmalloc allocator unless flags are
 195  * compatible with (a superset of) GFP_KERNEL.  This is because while the
 196  * actual pages are allocated with the specified flags, the page table pages
 197  * are always allocated with GFP_KERNEL.
 198  *
 199  * ceph_kvmalloc() may be called with GFP_KERNEL, GFP_NOFS or GFP_NOIO.
 200  */
 201 void *ceph_kvmalloc(size_t size, gfp_t flags)
 202 {
 203         void *p;
 204
 205         if ((flags & (__GFP_IO | __GFP_FS)) == (__GFP_IO | __GFP_FS)) {
 206                 p = kvmalloc(size, flags);
 207         } else if ((flags & (__GFP_IO | __GFP_FS)) == __GFP_IO) {
 208                 unsigned int nofs_flag = memalloc_nofs_save();
 209                 p = kvmalloc(size, GFP_KERNEL);
 210                 memalloc_nofs_restore(nofs_flag);
 211         } else {
 212                 unsigned int noio_flag = memalloc_noio_save();
 213                 p = kvmalloc(size, GFP_KERNEL);
 214                 memalloc_noio_restore(noio_flag);
 215         }
 216
 217         return p;
 218 }
 219
 220 static int parse_fsid(const char *str, struct ceph_fsid *fsid)
 221 {
 222         int i = 0;
 223         char tmp[3];
 224         int err = -EINVAL;
 225         int d;
 226
 227         dout("parse_fsid '%s'\n", str);
 228         tmp[2] = 0;
 229         while (*str && i < 16) {
 230                 if (ispunct(*str)) {
 231                         str++;
 232                         continue;
 233                 }
 234                 if (!isxdigit(str[0]) || !isxdigit(str[1]))
 235                         break;
 236                 tmp[0] = str[0];
 237                 tmp[1] = str[1];
 238                 if (sscanf(tmp, "%x", &d) < 1)
 239                         break;
 240                 fsid->fsid[i] = d & 0xff;
 241                 i++;
 242                 str += 2;
 243         }
 244
 245         if (i == 16)
 246                 err = 0;
 247         dout("parse_fsid ret %d got fsid %pU\n", err, fsid);
 248         return err;
 249 }
 250
 251 /*
 252  * ceph options
 253  */
 254 enum {
 255         Opt_osdkeepalivetimeout,
 256         Opt_mount_timeout,
 257         Opt_osd_idle_ttl,
 258         Opt_osd_request_timeout,
 259         /* int args above */
 260         Opt_fsid,
 261         Opt_name,
 262         Opt_secret,
 263         Opt_key,
 264         Opt_ip,
 265         Opt_crush_location,
 266         Opt_read_from_replica,
 267         Opt_ms_mode,
 268         /* string args above */
 269         Opt_share,
 270         Opt_crc,
 271         Opt_cephx_require_signatures,
 272         Opt_cephx_sign_messages,
 273         Opt_tcp_nodelay,
 274         Opt_abort_on_full,
 275 };
 276
 277 enum {
 278         Opt_read_from_replica_no,
 279         Opt_read_from_replica_balance,
 280         Opt_read_from_replica_localize,
 281 };
 282
 283 static const struct constant_table ceph_param_read_from_replica[] = {
 284         {"no",          Opt_read_from_replica_no},
 285         {"balance",     Opt_read_from_replica_balance},
 286         {"localize",    Opt_read_from_replica_localize},
 287         {}
 288 };
 289
 290 enum ceph_ms_mode {
 291         Opt_ms_mode_legacy,
 292         Opt_ms_mode_crc,
 293         Opt_ms_mode_secure,
 294         Opt_ms_mode_prefer_crc,
 295         Opt_ms_mode_prefer_secure
 296 };
 297
 298 static const struct constant_table ceph_param_ms_mode[] = {
 299         {"legacy",              Opt_ms_mode_legacy},
 300         {"crc",                 Opt_ms_mode_crc},
 301         {"secure",              Opt_ms_mode_secure},
 302         {"prefer-crc",          Opt_ms_mode_prefer_crc},
 303         {"prefer-secure",       Opt_ms_mode_prefer_secure},
 304         {}
 305 };
 306
 307 static const struct fs_parameter_spec ceph_parameters[] = {
 308         fsparam_flag    ("abort_on_full",               Opt_abort_on_full),
 309         __fsparam       (NULL, "cephx_require_signatures", Opt_cephx_require_signatures,
 310                          fs_param_neg_with_no|fs_param_deprecated, NULL),
 311         fsparam_flag_no ("cephx_sign_messages",         Opt_cephx_sign_messages),
 312         fsparam_flag_no ("crc",                         Opt_crc),
 313         fsparam_string  ("crush_location",              Opt_crush_location),
 314         fsparam_string  ("fsid",                        Opt_fsid),
 315         fsparam_string  ("ip",                          Opt_ip),
 316         fsparam_string  ("key",                         Opt_key),
 317         fsparam_u32     ("mount_timeout",               Opt_mount_timeout),
 318         fsparam_string  ("name",                        Opt_name),
 319         fsparam_u32     ("osd_idle_ttl",                Opt_osd_idle_ttl),
 320         fsparam_u32     ("osd_request_timeout",         Opt_osd_request_timeout),
 321         fsparam_u32     ("osdkeepalive",                Opt_osdkeepalivetimeout),
 322         fsparam_enum    ("read_from_replica",           Opt_read_from_replica,
 323                          ceph_param_read_from_replica),
 324         fsparam_enum    ("ms_mode",                     Opt_ms_mode,
 325                          ceph_param_ms_mode),
 326         fsparam_string  ("secret",                      Opt_secret),
 327         fsparam_flag_no ("share",                       Opt_share),
 328         fsparam_flag_no ("tcp_nodelay",                 Opt_tcp_nodelay),
 329         {}
 330 };
 331
 332 struct ceph_options *ceph_alloc_options(void)
 333 {
 334         struct ceph_options *opt;
 335
 336         opt = kzalloc(sizeof(*opt), GFP_KERNEL);
 337         if (!opt)
 338                 return NULL;
 339
 340         opt->crush_locs = RB_ROOT;
 341         opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
 342                                 GFP_KERNEL);
 343         if (!opt->mon_addr) {
 344                 kfree(opt);
 345                 return NULL;
 346         }
 347
 348         opt->flags = CEPH_OPT_DEFAULT;
 349         opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
 350         opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT;
 351         opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT;
 352         opt->osd_request_timeout = CEPH_OSD_REQUEST_TIMEOUT_DEFAULT;
 353         opt->read_from_replica = CEPH_READ_FROM_REPLICA_DEFAULT;
 354         opt->con_modes[0] = CEPH_CON_MODE_UNKNOWN;
 355         opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
 356         return opt;
 357 }
 358 EXPORT_SYMBOL(ceph_alloc_options);
 359
 360 void ceph_destroy_options(struct ceph_options *opt)
 361 {
 362         dout("destroy_options %p\n", opt);
 363         if (!opt)
 364                 return;
 365
 366         ceph_clear_crush_locs(&opt->crush_locs);
 367         kfree(opt->name);
 368         if (opt->key) {
 369                 ceph_crypto_key_destroy(opt->key);
 370                 kfree(opt->key);
 371         }
 372         kfree(opt->mon_addr);
 373         kfree(opt);
 374 }
 375 EXPORT_SYMBOL(ceph_destroy_options);
 376
 377 /* get secret from key store */
 378 static int get_secret(struct ceph_crypto_key *dst, const char *name,
 379                       struct p_log *log)
 380 {
 381         struct key *ukey;
 382         int key_err;
 383         int err = 0;
 384         struct ceph_crypto_key *ckey;
 385
 386         ukey = request_key(&key_type_ceph, name, NULL);
 387         if (IS_ERR(ukey)) {
 388                 /* request_key errors don't map nicely to mount(2)
 389                    errors; don't even try, but still printk */
 390                 key_err = PTR_ERR(ukey);
 391                 switch (key_err) {
 392                 case -ENOKEY:
 393                         error_plog(log, "Failed due to key not found: %s",
 394                                name);
 395                         break;
 396                 case -EKEYEXPIRED:
 397                         error_plog(log, "Failed due to expired key: %s",
 398                                name);
 399                         break;
 400                 case -EKEYREVOKED:
 401                         error_plog(log, "Failed due to revoked key: %s",
 402                                name);
 403                         break;
 404                 default:
 405                         error_plog(log, "Failed due to key error %d: %s",
 406                                key_err, name);
 407                 }
 408                 err = -EPERM;
 409                 goto out;
 410         }
 411
 412         ckey = ukey->payload.data[0];
 413         err = ceph_crypto_key_clone(dst, ckey);
 414         if (err)
 415                 goto out_key;
 416         /* pass through, err is 0 */
 417
 418 out_key:
 419         key_put(ukey);
 420 out:
 421         return err;
 422 }
 423
 424 int ceph_parse_mon_ips(const char *buf, size_t len, struct ceph_options *opt,
 425                        struct fc_log *l, char delim)
 426 {
 427         struct p_log log = {.prefix = "libceph", .log = l};
 428         int ret;
 429
 430         /* ip1[:port1][<delim>ip2[:port2]...] */
 431         ret = ceph_parse_ips(buf, buf + len, opt->mon_addr, CEPH_MAX_MON,
 432                              &opt->num_mon, delim);
 433         if (ret) {
 434                 error_plog(&log, "Failed to parse monitor IPs: %d", ret);
 435                 return ret;
 436         }
 437
 438         return 0;
 439 }
 440 EXPORT_SYMBOL(ceph_parse_mon_ips);
 441
 442 int ceph_parse_param(struct fs_parameter *param, struct ceph_options *opt,
 443                      struct fc_log *l)
 444 {
 445         struct fs_parse_result result;
 446         int token, err;
 447         struct p_log log = {.prefix = "libceph", .log = l};
 448
 449         token = __fs_parse(&log, ceph_parameters, param, &result);
 450         dout("%s fs_parse '%s' token %d\n", __func__, param->key, token);
 451         if (token < 0)
 452                 return token;
 453
 454         switch (token) {
 455         case Opt_ip:
 456                 err = ceph_parse_ips(param->string,
 457                                      param->string + param->size,
 458                                      &opt->my_addr, 1, NULL, ',');
 459                 if (err) {
 460                         error_plog(&log, "Failed to parse ip: %d", err);
 461                         return err;
 462                 }
 463                 opt->flags |= CEPH_OPT_MYIP;
 464                 break;
 465
 466         case Opt_fsid:
 467                 err = parse_fsid(param->string, &opt->fsid);
 468                 if (err) {
 469                         error_plog(&log, "Failed to parse fsid: %d", err);
 470                         return err;
 471                 }
 472                 opt->flags |= CEPH_OPT_FSID;
 473                 break;
 474         case Opt_name:
 475                 kfree(opt->name);
 476                 opt->name = param->string;
 477                 param->string = NULL;
 478                 break;
 479         case Opt_secret:
 480                 ceph_crypto_key_destroy(opt->key);
 481                 kfree(opt->key);
 482
 483                 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
 484                 if (!opt->key)
 485                         return -ENOMEM;
 486                 err = ceph_crypto_key_unarmor(opt->key, param->string);
 487                 if (err) {
 488                         error_plog(&log, "Failed to parse secret: %d", err);
 489                         return err;
 490                 }
 491                 break;
 492         case Opt_key:
 493                 ceph_crypto_key_destroy(opt->key);
 494                 kfree(opt->key);
 495
 496                 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
 497                 if (!opt->key)
 498                         return -ENOMEM;
 499                 return get_secret(opt->key, param->string, &log);
 500         case Opt_crush_location:
 501                 ceph_clear_crush_locs(&opt->crush_locs);
 502                 err = ceph_parse_crush_location(param->string,
 503                                                 &opt->crush_locs);
 504                 if (err) {
 505                         error_plog(&log, "Failed to parse CRUSH location: %d",
 506                                    err);
 507                         return err;
 508                 }
 509                 break;
 510         case Opt_read_from_replica:
 511                 switch (result.uint_32) {
 512                 case Opt_read_from_replica_no:
 513                         opt->read_from_replica = 0;
 514                         break;
 515                 case Opt_read_from_replica_balance:
 516                         opt->read_from_replica = CEPH_OSD_FLAG_BALANCE_READS;
 517                         break;
 518                 case Opt_read_from_replica_localize:
 519                         opt->read_from_replica = CEPH_OSD_FLAG_LOCALIZE_READS;
 520                         break;
 521                 default:
 522                         BUG();
 523                 }
 524                 break;
 525         case Opt_ms_mode:
 526                 switch (result.uint_32) {
 527                 case Opt_ms_mode_legacy:
 528                         opt->con_modes[0] = CEPH_CON_MODE_UNKNOWN;
 529                         opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
 530                         break;
 531                 case Opt_ms_mode_crc:
 532                         opt->con_modes[0] = CEPH_CON_MODE_CRC;
 533                         opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
 534                         break;
 535                 case Opt_ms_mode_secure:
 536                         opt->con_modes[0] = CEPH_CON_MODE_SECURE;
 537                         opt->con_modes[1] = CEPH_CON_MODE_UNKNOWN;
 538                         break;
 539                 case Opt_ms_mode_prefer_crc:
 540                         opt->con_modes[0] = CEPH_CON_MODE_CRC;
 541                         opt->con_modes[1] = CEPH_CON_MODE_SECURE;
 542                         break;
 543                 case Opt_ms_mode_prefer_secure:
 544                         opt->con_modes[0] = CEPH_CON_MODE_SECURE;
 545                         opt->con_modes[1] = CEPH_CON_MODE_CRC;
 546                         break;
 547                 default:
 548                         BUG();
 549                 }
 550                 break;
 551
 552         case Opt_osdkeepalivetimeout:
 553                 /* 0 isn't well defined right now, reject it */
 554                 if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
 555                         goto out_of_range;
 556                 opt->osd_keepalive_timeout =
 557                     msecs_to_jiffies(result.uint_32 * 1000);
 558                 break;
 559         case Opt_osd_idle_ttl:
 560                 /* 0 isn't well defined right now, reject it */
 561                 if (result.uint_32 < 1 || result.uint_32 > INT_MAX / 1000)
 562                         goto out_of_range;
 563                 opt->osd_idle_ttl = msecs_to_jiffies(result.uint_32 * 1000);
 564                 break;
 565         case Opt_mount_timeout:
 566                 /* 0 is "wait forever" (i.e. infinite timeout) */
 567                 if (result.uint_32 > INT_MAX / 1000)
 568                         goto out_of_range;
 569                 opt->mount_timeout = msecs_to_jiffies(result.uint_32 * 1000);
 570                 break;
 571         case Opt_osd_request_timeout:
 572                 /* 0 is "wait forever" (i.e. infinite timeout) */
 573                 if (result.uint_32 > INT_MAX / 1000)
 574                         goto out_of_range;
 575                 opt->osd_request_timeout =
 576                     msecs_to_jiffies(result.uint_32 * 1000);
 577                 break;
 578
 579         case Opt_share:
 580                 if (!result.negated)
 581                         opt->flags &= ~CEPH_OPT_NOSHARE;
 582                 else
 583                         opt->flags |= CEPH_OPT_NOSHARE;
 584                 break;
 585         case Opt_crc:
 586                 if (!result.negated)
 587                         opt->flags &= ~CEPH_OPT_NOCRC;
 588                 else
 589                         opt->flags |= CEPH_OPT_NOCRC;
 590                 break;
 591         case Opt_cephx_require_signatures:
 592                 if (!result.negated)
 593                         warn_plog(&log, "Ignoring cephx_require_signatures");
 594                 else
 595                         warn_plog(&log, "Ignoring nocephx_require_signatures, use nocephx_sign_messages");
 596                 break;
 597         case Opt_cephx_sign_messages:
 598                 if (!result.negated)
 599                         opt->flags &= ~CEPH_OPT_NOMSGSIGN;
 600                 else
 601                         opt->flags |= CEPH_OPT_NOMSGSIGN;
 602                 break;
 603         case Opt_tcp_nodelay:
 604                 if (!result.negated)
 605                         opt->flags |= CEPH_OPT_TCP_NODELAY;
 606                 else
 607                         opt->flags &= ~CEPH_OPT_TCP_NODELAY;
 608                 break;
 609
 610         case Opt_abort_on_full:
 611                 opt->flags |= CEPH_OPT_ABORT_ON_FULL;
 612                 break;
 613
 614         default:
 615                 BUG();
 616         }
 617
 618         return 0;
 619
 620 out_of_range:
 621         return inval_plog(&log, "%s out of range", param->key);
 622 }
 623 EXPORT_SYMBOL(ceph_parse_param);
 624
 625 int ceph_print_client_options(struct seq_file *m, struct ceph_client *client,
 626                               bool show_all)
 627 {
 628         struct ceph_options *opt = client->options;
 629         size_t pos = m->count;
 630         struct rb_node *n;
 631
 632         if (opt->name) {
 633                 seq_puts(m, "name=");
 634                 seq_escape(m, opt->name, ", \t\n\\");
 635                 seq_putc(m, ',');
 636         }
 637         if (opt->key)
 638                 seq_puts(m, "secret=<hidden>,");
 639
 640         if (!RB_EMPTY_ROOT(&opt->crush_locs)) {
 641                 seq_puts(m, "crush_location=");
 642                 for (n = rb_first(&opt->crush_locs); ; ) {
 643                         struct crush_loc_node *loc =
 644                             rb_entry(n, struct crush_loc_node, cl_node);
 645
 646                         seq_printf(m, "%s:%s", loc->cl_loc.cl_type_name,
 647                                    loc->cl_loc.cl_name);
 648                         n = rb_next(n);
 649                         if (!n)
 650                                 break;
 651
 652                         seq_putc(m, '|');
 653                 }
 654                 seq_putc(m, ',');
 655         }
 656         if (opt->read_from_replica == CEPH_OSD_FLAG_BALANCE_READS) {
 657                 seq_puts(m, "read_from_replica=balance,");
 658         } else if (opt->read_from_replica == CEPH_OSD_FLAG_LOCALIZE_READS) {
 659                 seq_puts(m, "read_from_replica=localize,");
 660         }
 661         if (opt->con_modes[0] != CEPH_CON_MODE_UNKNOWN) {
 662                 if (opt->con_modes[0] == CEPH_CON_MODE_CRC &&
 663                     opt->con_modes[1] == CEPH_CON_MODE_UNKNOWN) {
 664                         seq_puts(m, "ms_mode=crc,");
 665                 } else if (opt->con_modes[0] == CEPH_CON_MODE_SECURE &&
 666                            opt->con_modes[1] == CEPH_CON_MODE_UNKNOWN) {
 667                         seq_puts(m, "ms_mode=secure,");
 668                 } else if (opt->con_modes[0] == CEPH_CON_MODE_CRC &&
 669                            opt->con_modes[1] == CEPH_CON_MODE_SECURE) {
 670                         seq_puts(m, "ms_mode=prefer-crc,");
 671                 } else if (opt->con_modes[0] == CEPH_CON_MODE_SECURE &&
 672                            opt->con_modes[1] == CEPH_CON_MODE_CRC) {
 673                         seq_puts(m, "ms_mode=prefer-secure,");
 674                 }
 675         }
 676
 677         if (opt->flags & CEPH_OPT_FSID)
 678                 seq_printf(m, "fsid=%pU,", &opt->fsid);
 679         if (opt->flags & CEPH_OPT_NOSHARE)
 680                 seq_puts(m, "noshare,");
 681         if (opt->flags & CEPH_OPT_NOCRC)
 682                 seq_puts(m, "nocrc,");
 683         if (opt->flags & CEPH_OPT_NOMSGSIGN)
 684                 seq_puts(m, "nocephx_sign_messages,");
 685         if ((opt->flags & CEPH_OPT_TCP_NODELAY) == 0)
 686                 seq_puts(m, "notcp_nodelay,");
 687         if (show_all && (opt->flags & CEPH_OPT_ABORT_ON_FULL))
 688                 seq_puts(m, "abort_on_full,");
 689
 690         if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
 691                 seq_printf(m, "mount_timeout=%d,",
 692                            jiffies_to_msecs(opt->mount_timeout) / 1000);
 693         if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
 694                 seq_printf(m, "osd_idle_ttl=%d,",
 695                            jiffies_to_msecs(opt->osd_idle_ttl) / 1000);
 696         if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
 697                 seq_printf(m, "osdkeepalivetimeout=%d,",
 698                     jiffies_to_msecs(opt->osd_keepalive_timeout) / 1000);
 699         if (opt->osd_request_timeout != CEPH_OSD_REQUEST_TIMEOUT_DEFAULT)
 700                 seq_printf(m, "osd_request_timeout=%d,",
 701                            jiffies_to_msecs(opt->osd_request_timeout) / 1000);
 702
 703         /* drop redundant comma */
 704         if (m->count != pos)
 705                 m->count--;
 706
 707         return 0;
 708 }
 709 EXPORT_SYMBOL(ceph_print_client_options);
 710
 711 struct ceph_entity_addr *ceph_client_addr(struct ceph_client *client)
 712 {
 713         return &client->msgr.inst.addr;
 714 }
 715 EXPORT_SYMBOL(ceph_client_addr);
 716
 717 u64 ceph_client_gid(struct ceph_client *client)
 718 {
 719         return client->monc.auth->global_id;
 720 }
 721 EXPORT_SYMBOL(ceph_client_gid);
 722
 723 /*
 724  * create a fresh client instance
 725  */
 726 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
 727 {
 728         struct ceph_client *client;
 729         struct ceph_entity_addr *myaddr = NULL;
 730         int err;
 731
 732         err = wait_for_random_bytes();
 733         if (err < 0)
 734                 return ERR_PTR(err);
 735
 736         client = kzalloc(sizeof(*client), GFP_KERNEL);
 737         if (client == NULL)
 738                 return ERR_PTR(-ENOMEM);
 739
 740         client->private = private;
 741         client->options = opt;
 742
 743         mutex_init(&client->mount_mutex);
 744         init_waitqueue_head(&client->auth_wq);
 745         client->auth_err = 0;
 746
 747         client->extra_mon_dispatch = NULL;
 748         client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT;
 749         client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT;
 750
 751         if (!ceph_test_opt(client, NOMSGSIGN))
 752                 client->required_features |= CEPH_FEATURE_MSG_AUTH;
 753
 754         /* msgr */
 755         if (ceph_test_opt(client, MYIP))
 756                 myaddr = &client->options->my_addr;
 757
 758         ceph_messenger_init(&client->msgr, myaddr);
 759
 760         /* subsystems */
 761         err = ceph_monc_init(&client->monc, client);
 762         if (err < 0)
 763                 goto fail;
 764         err = ceph_osdc_init(&client->osdc, client);
 765         if (err < 0)
 766                 goto fail_monc;
 767
 768         return client;
 769
 770 fail_monc:
 771         ceph_monc_stop(&client->monc);
 772 fail:
 773         ceph_messenger_fini(&client->msgr);
 774         kfree(client);
 775         return ERR_PTR(err);
 776 }
 777 EXPORT_SYMBOL(ceph_create_client);
 778
 779 void ceph_destroy_client(struct ceph_client *client)
 780 {
 781         dout("destroy_client %p\n", client);
 782
 783         atomic_set(&client->msgr.stopping, 1);
 784
 785         /* unmount */
 786         ceph_osdc_stop(&client->osdc);
 787         ceph_monc_stop(&client->monc);
 788         ceph_messenger_fini(&client->msgr);
 789
 790         ceph_debugfs_client_cleanup(client);
 791
 792         ceph_destroy_options(client->options);
 793
 794         kfree(client);
 795         dout("destroy_client %p done\n", client);
 796 }
 797 EXPORT_SYMBOL(ceph_destroy_client);
 798
 799 void ceph_reset_client_addr(struct ceph_client *client)
 800 {
 801         ceph_messenger_reset_nonce(&client->msgr);
 802         ceph_monc_reopen_session(&client->monc);
 803         ceph_osdc_reopen_osds(&client->osdc);
 804 }
 805 EXPORT_SYMBOL(ceph_reset_client_addr);
 806
 807 /*
 808  * true if we have the mon map (and have thus joined the cluster)
 809  */
 810 static bool have_mon_and_osd_map(struct ceph_client *client)
 811 {
 812         return client->monc.monmap && client->monc.monmap->epoch &&
 813                client->osdc.osdmap && client->osdc.osdmap->epoch;
 814 }
 815
 816 /*
 817  * mount: join the ceph cluster, and open root directory.
 818  */
 819 int __ceph_open_session(struct ceph_client *client, unsigned long started)
 820 {
 821         unsigned long timeout = client->options->mount_timeout;
 822         long err;
 823
 824         /* open session, and wait for mon and osd maps */
 825         err = ceph_monc_open_session(&client->monc);
 826         if (err < 0)
 827                 return err;
 828
 829         while (!have_mon_and_osd_map(client)) {
 830                 if (timeout && time_after_eq(jiffies, started + timeout))
 831                         return -ETIMEDOUT;
 832
 833                 /* wait */
 834                 dout("mount waiting for mon_map\n");
 835                 err = wait_event_interruptible_timeout(client->auth_wq,
 836                         have_mon_and_osd_map(client) || (client->auth_err < 0),
 837                         ceph_timeout_jiffies(timeout));
 838                 if (err < 0)
 839                         return err;
 840                 if (client->auth_err < 0)
 841                         return client->auth_err;
 842         }
 843
 844         pr_info("client%llu fsid %pU\n", ceph_client_gid(client),
 845                 &client->fsid);
 846         ceph_debugfs_client_init(client);
 847
 848         return 0;
 849 }
 850 EXPORT_SYMBOL(__ceph_open_session);
 851
 852 int ceph_open_session(struct ceph_client *client)
 853 {
 854         int ret;
 855         unsigned long started = jiffies;  /* note the start time */
 856
 857         dout("open_session start\n");
 858         mutex_lock(&client->mount_mutex);
 859
 860         ret = __ceph_open_session(client, started);
 861
 862         mutex_unlock(&client->mount_mutex);
 863         return ret;
 864 }
 865 EXPORT_SYMBOL(ceph_open_session);
 866
 867 int ceph_wait_for_latest_osdmap(struct ceph_client *client,
 868                                 unsigned long timeout)
 869 {
 870         u64 newest_epoch;
 871         int ret;
 872
 873         ret = ceph_monc_get_version(&client->monc, "osdmap", &newest_epoch);
 874         if (ret)
 875                 return ret;
 876
 877         if (client->osdc.osdmap->epoch >= newest_epoch)
 878                 return 0;
 879
 880         ceph_osdc_maybe_request_map(&client->osdc);
 881         return ceph_monc_wait_osdmap(&client->monc, newest_epoch, timeout);
 882 }
 883 EXPORT_SYMBOL(ceph_wait_for_latest_osdmap);
 884
 885 static int __init init_ceph_lib(void)
 886 {
 887         int ret = 0;
 888
 889         ceph_debugfs_init();
 890
 891         ret = ceph_crypto_init();
 892         if (ret < 0)
 893                 goto out_debugfs;
 894
 895         ret = ceph_msgr_init();
 896         if (ret < 0)
 897                 goto out_crypto;
 898
 899         ret = ceph_osdc_setup();
 900         if (ret < 0)
 901                 goto out_msgr;
 902
 903         pr_info("loaded (mon/osd proto %d/%d)\n",
 904                 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL);
 905
 906         return 0;
 907
 908 out_msgr:
 909         ceph_msgr_exit();
 910 out_crypto:
 911         ceph_crypto_shutdown();
 912 out_debugfs:
 913         ceph_debugfs_cleanup();
 914         return ret;
 915 }
 916
 917 static void __exit exit_ceph_lib(void)
 918 {
 919         dout("exit_ceph_lib\n");
 920         WARN_ON(!ceph_strings_empty());
 921
 922         ceph_osdc_cleanup();
 923         ceph_msgr_exit();
 924         ceph_crypto_shutdown();
 925         ceph_debugfs_cleanup();
 926 }
 927
 928 module_init(init_ceph_lib);
 929 module_exit(exit_ceph_lib);
 930
 931 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
 932 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
 933 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
 934 MODULE_DESCRIPTION("Ceph core library");
 935 MODULE_LICENSE("GPL");