fs/ceph/mds_client.c

   1 // SPDX-License-Identifier: GPL-2.0
   2 #include <linux/ceph/ceph_debug.h>
   3
   4 #include <linux/fs.h>
   5 #include <linux/wait.h>
   6 #include <linux/slab.h>
   7 #include <linux/gfp.h>
   8 #include <linux/sched.h>
   9 #include <linux/debugfs.h>
  10 #include <linux/seq_file.h>
  11 #include <linux/ratelimit.h>
  12 #include <linux/bits.h>
  13 #include <linux/ktime.h>
  14
  15 #include "super.h"
  16 #include "mds_client.h"
  17
  18 #include <linux/ceph/ceph_features.h>
  19 #include <linux/ceph/messenger.h>
  20 #include <linux/ceph/decode.h>
  21 #include <linux/ceph/pagelist.h>
  22 #include <linux/ceph/auth.h>
  23 #include <linux/ceph/debugfs.h>
  24
  25 #define RECONNECT_MAX_SIZE (INT_MAX - PAGE_SIZE)
  26
  27 /*
  28  * A cluster of MDS (metadata server) daemons is responsible for
  29  * managing the file system namespace (the directory hierarchy and
  30  * inodes) and for coordinating shared access to storage.  Metadata is
  31  * partitioning hierarchically across a number of servers, and that
  32  * partition varies over time as the cluster adjusts the distribution
  33  * in order to balance load.
  34  *
  35  * The MDS client is primarily responsible to managing synchronous
  36  * metadata requests for operations like open, unlink, and so forth.
  37  * If there is a MDS failure, we find out about it when we (possibly
  38  * request and) receive a new MDS map, and can resubmit affected
  39  * requests.
  40  *
  41  * For the most part, though, we take advantage of a lossless
  42  * communications channel to the MDS, and do not need to worry about
  43  * timing out or resubmitting requests.
  44  *
  45  * We maintain a stateful "session" with each MDS we interact with.
  46  * Within each session, we sent periodic heartbeat messages to ensure
  47  * any capabilities or leases we have been issues remain valid.  If
  48  * the session times out and goes stale, our leases and capabilities
  49  * are no longer valid.
  50  */
  51
  52 struct ceph_reconnect_state {
  53         struct ceph_mds_session *session;
  54         int nr_caps, nr_realms;
  55         struct ceph_pagelist *pagelist;
  56         unsigned msg_version;
  57         bool allow_multi;
  58 };
  59
  60 static void __wake_requests(struct ceph_mds_client *mdsc,
  61                             struct list_head *head);
  62 static void ceph_cap_release_work(struct work_struct *work);
  63 static void ceph_cap_reclaim_work(struct work_struct *work);
  64
  65 static const struct ceph_connection_operations mds_con_ops;
  66
  67
  68 /*
  69  * mds reply parsing
  70  */
  71
  72 static int parse_reply_info_quota(void **p, void *end,
  73                                   struct ceph_mds_reply_info_in *info)
  74 {
  75         u8 struct_v, struct_compat;
  76         u32 struct_len;
  77
  78         ceph_decode_8_safe(p, end, struct_v, bad);
  79         ceph_decode_8_safe(p, end, struct_compat, bad);
  80         /* struct_v is expected to be >= 1. we only
  81          * understand encoding with struct_compat == 1. */
  82         if (!struct_v || struct_compat != 1)
  83                 goto bad;
  84         ceph_decode_32_safe(p, end, struct_len, bad);
  85         ceph_decode_need(p, end, struct_len, bad);
  86         end = *p + struct_len;
  87         ceph_decode_64_safe(p, end, info->max_bytes, bad);
  88         ceph_decode_64_safe(p, end, info->max_files, bad);
  89         *p = end;
  90         return 0;
  91 bad:
  92         return -EIO;
  93 }
  94
  95 /*
  96  * parse individual inode info
  97  */
  98 static int parse_reply_info_in(void **p, void *end,
  99                                struct ceph_mds_reply_info_in *info,
 100                                u64 features)
 101 {
 102         int err = 0;
 103         u8 struct_v = 0;
 104
 105         if (features == (u64)-1) {
 106                 u32 struct_len;
 107                 u8 struct_compat;
 108                 ceph_decode_8_safe(p, end, struct_v, bad);
 109                 ceph_decode_8_safe(p, end, struct_compat, bad);
 110                 /* struct_v is expected to be >= 1. we only understand
 111                  * encoding with struct_compat == 1. */
 112                 if (!struct_v || struct_compat != 1)
 113                         goto bad;
 114                 ceph_decode_32_safe(p, end, struct_len, bad);
 115                 ceph_decode_need(p, end, struct_len, bad);
 116                 end = *p + struct_len;
 117         }
 118
 119         ceph_decode_need(p, end, sizeof(struct ceph_mds_reply_inode), bad);
 120         info->in = *p;
 121         *p += sizeof(struct ceph_mds_reply_inode) +
 122                 sizeof(*info->in->fragtree.splits) *
 123                 le32_to_cpu(info->in->fragtree.nsplits);
 124
 125         ceph_decode_32_safe(p, end, info->symlink_len, bad);
 126         ceph_decode_need(p, end, info->symlink_len, bad);
 127         info->symlink = *p;
 128         *p += info->symlink_len;
 129
 130         ceph_decode_copy_safe(p, end, &info->dir_layout,
 131                               sizeof(info->dir_layout), bad);
 132         ceph_decode_32_safe(p, end, info->xattr_len, bad);
 133         ceph_decode_need(p, end, info->xattr_len, bad);
 134         info->xattr_data = *p;
 135         *p += info->xattr_len;
 136
 137         if (features == (u64)-1) {
 138                 /* inline data */
 139                 ceph_decode_64_safe(p, end, info->inline_version, bad);
 140                 ceph_decode_32_safe(p, end, info->inline_len, bad);
 141                 ceph_decode_need(p, end, info->inline_len, bad);
 142                 info->inline_data = *p;
 143                 *p += info->inline_len;
 144                 /* quota */
 145                 err = parse_reply_info_quota(p, end, info);
 146                 if (err < 0)
 147                         goto out_bad;
 148                 /* pool namespace */
 149                 ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
 150                 if (info->pool_ns_len > 0) {
 151                         ceph_decode_need(p, end, info->pool_ns_len, bad);
 152                         info->pool_ns_data = *p;
 153                         *p += info->pool_ns_len;
 154                 }
 155
 156                 /* btime */
 157                 ceph_decode_need(p, end, sizeof(info->btime), bad);
 158                 ceph_decode_copy(p, &info->btime, sizeof(info->btime));
 159
 160                 /* change attribute */
 161                 ceph_decode_64_safe(p, end, info->change_attr, bad);
 162
 163                 /* dir pin */
 164                 if (struct_v >= 2) {
 165                         ceph_decode_32_safe(p, end, info->dir_pin, bad);
 166                 } else {
 167                         info->dir_pin = -ENODATA;
 168                 }
 169
 170                 /* snapshot birth time, remains zero for v<=2 */
 171                 if (struct_v >= 3) {
 172                         ceph_decode_need(p, end, sizeof(info->snap_btime), bad);
 173                         ceph_decode_copy(p, &info->snap_btime,
 174                                          sizeof(info->snap_btime));
 175                 } else {
 176                         memset(&info->snap_btime, 0, sizeof(info->snap_btime));
 177                 }
 178
 179                 *p = end;
 180         } else {
 181                 if (features & CEPH_FEATURE_MDS_INLINE_DATA) {
 182                         ceph_decode_64_safe(p, end, info->inline_version, bad);
 183                         ceph_decode_32_safe(p, end, info->inline_len, bad);
 184                         ceph_decode_need(p, end, info->inline_len, bad);
 185                         info->inline_data = *p;
 186                         *p += info->inline_len;
 187                 } else
 188                         info->inline_version = CEPH_INLINE_NONE;
 189
 190                 if (features & CEPH_FEATURE_MDS_QUOTA) {
 191                         err = parse_reply_info_quota(p, end, info);
 192                         if (err < 0)
 193                                 goto out_bad;
 194                 } else {
 195                         info->max_bytes = 0;
 196                         info->max_files = 0;
 197                 }
 198
 199                 info->pool_ns_len = 0;
 200                 info->pool_ns_data = NULL;
 201                 if (features & CEPH_FEATURE_FS_FILE_LAYOUT_V2) {
 202                         ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
 203                         if (info->pool_ns_len > 0) {
 204                                 ceph_decode_need(p, end, info->pool_ns_len, bad);
 205                                 info->pool_ns_data = *p;
 206                                 *p += info->pool_ns_len;
 207                         }
 208                 }
 209
 210                 if (features & CEPH_FEATURE_FS_BTIME) {
 211                         ceph_decode_need(p, end, sizeof(info->btime), bad);
 212                         ceph_decode_copy(p, &info->btime, sizeof(info->btime));
 213                         ceph_decode_64_safe(p, end, info->change_attr, bad);
 214                 }
 215
 216                 info->dir_pin = -ENODATA;
 217                 /* info->snap_btime remains zero */
 218         }
 219         return 0;
 220 bad:
 221         err = -EIO;
 222 out_bad:
 223         return err;
 224 }
 225
 226 static int parse_reply_info_dir(void **p, void *end,
 227                                 struct ceph_mds_reply_dirfrag **dirfrag,
 228                                 u64 features)
 229 {
 230         if (features == (u64)-1) {
 231                 u8 struct_v, struct_compat;
 232                 u32 struct_len;
 233                 ceph_decode_8_safe(p, end, struct_v, bad);
 234                 ceph_decode_8_safe(p, end, struct_compat, bad);
 235                 /* struct_v is expected to be >= 1. we only understand
 236                  * encoding whose struct_compat == 1. */
 237                 if (!struct_v || struct_compat != 1)
 238                         goto bad;
 239                 ceph_decode_32_safe(p, end, struct_len, bad);
 240                 ceph_decode_need(p, end, struct_len, bad);
 241                 end = *p + struct_len;
 242         }
 243
 244         ceph_decode_need(p, end, sizeof(**dirfrag), bad);
 245         *dirfrag = *p;
 246         *p += sizeof(**dirfrag) + sizeof(u32) * le32_to_cpu((*dirfrag)->ndist);
 247         if (unlikely(*p > end))
 248                 goto bad;
 249         if (features == (u64)-1)
 250                 *p = end;
 251         return 0;
 252 bad:
 253         return -EIO;
 254 }
 255
 256 static int parse_reply_info_lease(void **p, void *end,
 257                                   struct ceph_mds_reply_lease **lease,
 258                                   u64 features)
 259 {
 260         if (features == (u64)-1) {
 261                 u8 struct_v, struct_compat;
 262                 u32 struct_len;
 263                 ceph_decode_8_safe(p, end, struct_v, bad);
 264                 ceph_decode_8_safe(p, end, struct_compat, bad);
 265                 /* struct_v is expected to be >= 1. we only understand
 266                  * encoding whose struct_compat == 1. */
 267                 if (!struct_v || struct_compat != 1)
 268                         goto bad;
 269                 ceph_decode_32_safe(p, end, struct_len, bad);
 270                 ceph_decode_need(p, end, struct_len, bad);
 271                 end = *p + struct_len;
 272         }
 273
 274         ceph_decode_need(p, end, sizeof(**lease), bad);
 275         *lease = *p;
 276         *p += sizeof(**lease);
 277         if (features == (u64)-1)
 278                 *p = end;
 279         return 0;
 280 bad:
 281         return -EIO;
 282 }
 283
 284 /*
 285  * parse a normal reply, which may contain a (dir+)dentry and/or a
 286  * target inode.
 287  */
 288 static int parse_reply_info_trace(void **p, void *end,
 289                                   struct ceph_mds_reply_info_parsed *info,
 290                                   u64 features)
 291 {
 292         int err;
 293
 294         if (info->head->is_dentry) {
 295                 err = parse_reply_info_in(p, end, &info->diri, features);
 296                 if (err < 0)
 297                         goto out_bad;
 298
 299                 err = parse_reply_info_dir(p, end, &info->dirfrag, features);
 300                 if (err < 0)
 301                         goto out_bad;
 302
 303                 ceph_decode_32_safe(p, end, info->dname_len, bad);
 304                 ceph_decode_need(p, end, info->dname_len, bad);
 305                 info->dname = *p;
 306                 *p += info->dname_len;
 307
 308                 err = parse_reply_info_lease(p, end, &info->dlease, features);
 309                 if (err < 0)
 310                         goto out_bad;
 311         }
 312
 313         if (info->head->is_target) {
 314                 err = parse_reply_info_in(p, end, &info->targeti, features);
 315                 if (err < 0)
 316                         goto out_bad;
 317         }
 318
 319         if (unlikely(*p != end))
 320                 goto bad;
 321         return 0;
 322
 323 bad:
 324         err = -EIO;
 325 out_bad:
 326         pr_err("problem parsing mds trace %d\n", err);
 327         return err;
 328 }
 329
 330 /*
 331  * parse readdir results
 332  */
 333 static int parse_reply_info_readdir(void **p, void *end,
 334                                 struct ceph_mds_reply_info_parsed *info,
 335                                 u64 features)
 336 {
 337         u32 num, i = 0;
 338         int err;
 339
 340         err = parse_reply_info_dir(p, end, &info->dir_dir, features);
 341         if (err < 0)
 342                 goto out_bad;
 343
 344         ceph_decode_need(p, end, sizeof(num) + 2, bad);
 345         num = ceph_decode_32(p);
 346         {
 347                 u16 flags = ceph_decode_16(p);
 348                 info->dir_end = !!(flags & CEPH_READDIR_FRAG_END);
 349                 info->dir_complete = !!(flags & CEPH_READDIR_FRAG_COMPLETE);
 350                 info->hash_order = !!(flags & CEPH_READDIR_HASH_ORDER);
 351                 info->offset_hash = !!(flags & CEPH_READDIR_OFFSET_HASH);
 352         }
 353         if (num == 0)
 354                 goto done;
 355
 356         BUG_ON(!info->dir_entries);
 357         if ((unsigned long)(info->dir_entries + num) >
 358             (unsigned long)info->dir_entries + info->dir_buf_size) {
 359                 pr_err("dir contents are larger than expected\n");
 360                 WARN_ON(1);
 361                 goto bad;
 362         }
 363
 364         info->dir_nr = num;
 365         while (num) {
 366                 struct ceph_mds_reply_dir_entry *rde = info->dir_entries + i;
 367                 /* dentry */
 368                 ceph_decode_32_safe(p, end, rde->name_len, bad);
 369                 ceph_decode_need(p, end, rde->name_len, bad);
 370                 rde->name = *p;
 371                 *p += rde->name_len;
 372                 dout("parsed dir dname '%.*s'\n", rde->name_len, rde->name);
 373
 374                 /* dentry lease */
 375                 err = parse_reply_info_lease(p, end, &rde->lease, features);
 376                 if (err)
 377                         goto out_bad;
 378                 /* inode */
 379                 err = parse_reply_info_in(p, end, &rde->inode, features);
 380                 if (err < 0)
 381                         goto out_bad;
 382                 /* ceph_readdir_prepopulate() will update it */
 383                 rde->offset = 0;
 384                 i++;
 385                 num--;
 386         }
 387
 388 done:
 389         /* Skip over any unrecognized fields */
 390         *p = end;
 391         return 0;
 392
 393 bad:
 394         err = -EIO;
 395 out_bad:
 396         pr_err("problem parsing dir contents %d\n", err);
 397         return err;
 398 }
 399
 400 /*
 401  * parse fcntl F_GETLK results
 402  */
 403 static int parse_reply_info_filelock(void **p, void *end,
 404                                      struct ceph_mds_reply_info_parsed *info,
 405                                      u64 features)
 406 {
 407         if (*p + sizeof(*info->filelock_reply) > end)
 408                 goto bad;
 409
 410         info->filelock_reply = *p;
 411
 412         /* Skip over any unrecognized fields */
 413         *p = end;
 414         return 0;
 415 bad:
 416         return -EIO;
 417 }
 418
 419
 420 #if BITS_PER_LONG == 64
 421
 422 #define DELEGATED_INO_AVAILABLE         xa_mk_value(1)
 423
 424 static int ceph_parse_deleg_inos(void **p, void *end,
 425                                  struct ceph_mds_session *s)
 426 {
 427         u32 sets;
 428
 429         ceph_decode_32_safe(p, end, sets, bad);
 430         dout("got %u sets of delegated inodes\n", sets);
 431         while (sets--) {
 432                 u64 start, len, ino;
 433
 434                 ceph_decode_64_safe(p, end, start, bad);
 435                 ceph_decode_64_safe(p, end, len, bad);
 436                 while (len--) {
 437                         int err = xa_insert(&s->s_delegated_inos, ino = start++,
 438                                             DELEGATED_INO_AVAILABLE,
 439                                             GFP_KERNEL);
 440                         if (!err) {
 441                                 dout("added delegated inode 0x%llx\n",
 442                                      start - 1);
 443                         } else if (err == -EBUSY) {
 444                                 pr_warn("ceph: MDS delegated inode 0x%llx more than once.\n",
 445                                         start - 1);
 446                         } else {
 447                                 return err;
 448                         }
 449                 }
 450         }
 451         return 0;
 452 bad:
 453         return -EIO;
 454 }
 455
 456 u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
 457 {
 458         unsigned long ino;
 459         void *val;
 460
 461         xa_for_each(&s->s_delegated_inos, ino, val) {
 462                 val = xa_erase(&s->s_delegated_inos, ino);
 463                 if (val == DELEGATED_INO_AVAILABLE)
 464                         return ino;
 465         }
 466         return 0;
 467 }
 468
 469 int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
 470 {
 471         return xa_insert(&s->s_delegated_inos, ino, DELEGATED_INO_AVAILABLE,
 472                          GFP_KERNEL);
 473 }
 474 #else /* BITS_PER_LONG == 64 */
 475 /*
 476  * FIXME: xarrays can't handle 64-bit indexes on a 32-bit arch. For now, just
 477  * ignore delegated_inos on 32 bit arch. Maybe eventually add xarrays for top
 478  * and bottom words?
 479  */
 480 static int ceph_parse_deleg_inos(void **p, void *end,
 481                                  struct ceph_mds_session *s)
 482 {
 483         u32 sets;
 484
 485         ceph_decode_32_safe(p, end, sets, bad);
 486         if (sets)
 487                 ceph_decode_skip_n(p, end, sets * 2 * sizeof(__le64), bad);
 488         return 0;
 489 bad:
 490         return -EIO;
 491 }
 492
 493 u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
 494 {
 495         return 0;
 496 }
 497
 498 int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
 499 {
 500         return 0;
 501 }
 502 #endif /* BITS_PER_LONG == 64 */
 503
 504 /*
 505  * parse create results
 506  */
 507 static int parse_reply_info_create(void **p, void *end,
 508                                   struct ceph_mds_reply_info_parsed *info,
 509                                   u64 features, struct ceph_mds_session *s)
 510 {
 511         int ret;
 512
 513         if (features == (u64)-1 ||
 514             (features & CEPH_FEATURE_REPLY_CREATE_INODE)) {
 515                 if (*p == end) {
 516                         /* Malformed reply? */
 517                         info->has_create_ino = false;
 518                 } else if (test_bit(CEPHFS_FEATURE_DELEG_INO, &s->s_features)) {
 519                         u8 struct_v, struct_compat;
 520                         u32 len;
 521
 522                         info->has_create_ino = true;
 523                         ceph_decode_8_safe(p, end, struct_v, bad);
 524                         ceph_decode_8_safe(p, end, struct_compat, bad);
 525                         ceph_decode_32_safe(p, end, len, bad);
 526                         ceph_decode_64_safe(p, end, info->ino, bad);
 527                         ret = ceph_parse_deleg_inos(p, end, s);
 528                         if (ret)
 529                                 return ret;
 530                 } else {
 531                         /* legacy */
 532                         ceph_decode_64_safe(p, end, info->ino, bad);
 533                         info->has_create_ino = true;
 534                 }
 535         } else {
 536                 if (*p != end)
 537                         goto bad;
 538         }
 539
 540         /* Skip over any unrecognized fields */
 541         *p = end;
 542         return 0;
 543 bad:
 544         return -EIO;
 545 }
 546
 547 /*
 548  * parse extra results
 549  */
 550 static int parse_reply_info_extra(void **p, void *end,
 551                                   struct ceph_mds_reply_info_parsed *info,
 552                                   u64 features, struct ceph_mds_session *s)
 553 {
 554         u32 op = le32_to_cpu(info->head->op);
 555
 556         if (op == CEPH_MDS_OP_GETFILELOCK)
 557                 return parse_reply_info_filelock(p, end, info, features);
 558         else if (op == CEPH_MDS_OP_READDIR || op == CEPH_MDS_OP_LSSNAP)
 559                 return parse_reply_info_readdir(p, end, info, features);
 560         else if (op == CEPH_MDS_OP_CREATE)
 561                 return parse_reply_info_create(p, end, info, features, s);
 562         else
 563                 return -EIO;
 564 }
 565
 566 /*
 567  * parse entire mds reply
 568  */
 569 static int parse_reply_info(struct ceph_mds_session *s, struct ceph_msg *msg,
 570                             struct ceph_mds_reply_info_parsed *info,
 571                             u64 features)
 572 {
 573         void *p, *end;
 574         u32 len;
 575         int err;
 576
 577         info->head = msg->front.iov_base;
 578         p = msg->front.iov_base + sizeof(struct ceph_mds_reply_head);
 579         end = p + msg->front.iov_len - sizeof(struct ceph_mds_reply_head);
 580
 581         /* trace */
 582         ceph_decode_32_safe(&p, end, len, bad);
 583         if (len > 0) {
 584                 ceph_decode_need(&p, end, len, bad);
 585                 err = parse_reply_info_trace(&p, p+len, info, features);
 586                 if (err < 0)
 587                         goto out_bad;
 588         }
 589
 590         /* extra */
 591         ceph_decode_32_safe(&p, end, len, bad);
 592         if (len > 0) {
 593                 ceph_decode_need(&p, end, len, bad);
 594                 err = parse_reply_info_extra(&p, p+len, info, features, s);
 595                 if (err < 0)
 596                         goto out_bad;
 597         }
 598
 599         /* snap blob */
 600         ceph_decode_32_safe(&p, end, len, bad);
 601         info->snapblob_len = len;
 602         info->snapblob = p;
 603         p += len;
 604
 605         if (p != end)
 606                 goto bad;
 607         return 0;
 608
 609 bad:
 610         err = -EIO;
 611 out_bad:
 612         pr_err("mds parse_reply err %d\n", err);
 613         return err;
 614 }
 615
 616 static void destroy_reply_info(struct ceph_mds_reply_info_parsed *info)
 617 {
 618         if (!info->dir_entries)
 619                 return;
 620         free_pages((unsigned long)info->dir_entries, get_order(info->dir_buf_size));
 621 }
 622
 623
 624 /*
 625  * sessions
 626  */
 627 const char *ceph_session_state_name(int s)
 628 {
 629         switch (s) {
 630         case CEPH_MDS_SESSION_NEW: return "new";
 631         case CEPH_MDS_SESSION_OPENING: return "opening";
 632         case CEPH_MDS_SESSION_OPEN: return "open";
 633         case CEPH_MDS_SESSION_HUNG: return "hung";
 634         case CEPH_MDS_SESSION_CLOSING: return "closing";
 635         case CEPH_MDS_SESSION_CLOSED: return "closed";
 636         case CEPH_MDS_SESSION_RESTARTING: return "restarting";
 637         case CEPH_MDS_SESSION_RECONNECTING: return "reconnecting";
 638         case CEPH_MDS_SESSION_REJECTED: return "rejected";
 639         default: return "???";
 640         }
 641 }
 642
 643 struct ceph_mds_session *ceph_get_mds_session(struct ceph_mds_session *s)
 644 {
 645         if (refcount_inc_not_zero(&s->s_ref)) {
 646                 dout("mdsc get_session %p %d -> %d\n", s,
 647                      refcount_read(&s->s_ref)-1, refcount_read(&s->s_ref));
 648                 return s;
 649         } else {
 650                 dout("mdsc get_session %p 0 -- FAIL\n", s);
 651                 return NULL;
 652         }
 653 }
 654
 655 void ceph_put_mds_session(struct ceph_mds_session *s)
 656 {
 657         dout("mdsc put_session %p %d -> %d\n", s,
 658              refcount_read(&s->s_ref), refcount_read(&s->s_ref)-1);
 659         if (refcount_dec_and_test(&s->s_ref)) {
 660                 if (s->s_auth.authorizer)
 661                         ceph_auth_destroy_authorizer(s->s_auth.authorizer);
 662                 WARN_ON(mutex_is_locked(&s->s_mutex));
 663                 xa_destroy(&s->s_delegated_inos);
 664                 kfree(s);
 665         }
 666 }
 667
 668 /*
 669  * called under mdsc->mutex
 670  */
 671 struct ceph_mds_session *__ceph_lookup_mds_session(struct ceph_mds_client *mdsc,
 672                                                    int mds)
 673 {
 674         if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
 675                 return NULL;
 676         return ceph_get_mds_session(mdsc->sessions[mds]);
 677 }
 678
 679 static bool __have_session(struct ceph_mds_client *mdsc, int mds)
 680 {
 681         if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
 682                 return false;
 683         else
 684                 return true;
 685 }
 686
 687 static int __verify_registered_session(struct ceph_mds_client *mdsc,
 688                                        struct ceph_mds_session *s)
 689 {
 690         if (s->s_mds >= mdsc->max_sessions ||
 691             mdsc->sessions[s->s_mds] != s)
 692                 return -ENOENT;
 693         return 0;
 694 }
 695
 696 /*
 697  * create+register a new session for given mds.
 698  * called under mdsc->mutex.
 699  */
 700 static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
 701                                                  int mds)
 702 {
 703         struct ceph_mds_session *s;
 704
 705         if (mds >= mdsc->mdsmap->possible_max_rank)
 706                 return ERR_PTR(-EINVAL);
 707
 708         s = kzalloc(sizeof(*s), GFP_NOFS);
 709         if (!s)
 710                 return ERR_PTR(-ENOMEM);
 711
 712         if (mds >= mdsc->max_sessions) {
 713                 int newmax = 1 << get_count_order(mds + 1);
 714                 struct ceph_mds_session **sa;
 715
 716                 dout("%s: realloc to %d\n", __func__, newmax);
 717                 sa = kcalloc(newmax, sizeof(void *), GFP_NOFS);
 718                 if (!sa)
 719                         goto fail_realloc;
 720                 if (mdsc->sessions) {
 721                         memcpy(sa, mdsc->sessions,
 722                                mdsc->max_sessions * sizeof(void *));
 723                         kfree(mdsc->sessions);
 724                 }
 725                 mdsc->sessions = sa;
 726                 mdsc->max_sessions = newmax;
 727         }
 728
 729         dout("%s: mds%d\n", __func__, mds);
 730         s->s_mdsc = mdsc;
 731         s->s_mds = mds;
 732         s->s_state = CEPH_MDS_SESSION_NEW;
 733         s->s_ttl = 0;
 734         s->s_seq = 0;
 735         mutex_init(&s->s_mutex);
 736
 737         ceph_con_init(&s->s_con, s, &mds_con_ops, &mdsc->fsc->client->msgr);
 738
 739         spin_lock_init(&s->s_gen_ttl_lock);
 740         s->s_cap_gen = 1;
 741         s->s_cap_ttl = jiffies - 1;
 742
 743         spin_lock_init(&s->s_cap_lock);
 744         s->s_renew_requested = 0;
 745         s->s_renew_seq = 0;
 746         INIT_LIST_HEAD(&s->s_caps);
 747         s->s_nr_caps = 0;
 748         refcount_set(&s->s_ref, 1);
 749         INIT_LIST_HEAD(&s->s_waiting);
 750         INIT_LIST_HEAD(&s->s_unsafe);
 751         xa_init(&s->s_delegated_inos);
 752         s->s_num_cap_releases = 0;
 753         s->s_cap_reconnect = 0;
 754         s->s_cap_iterator = NULL;
 755         INIT_LIST_HEAD(&s->s_cap_releases);
 756         INIT_WORK(&s->s_cap_release_work, ceph_cap_release_work);
 757
 758         INIT_LIST_HEAD(&s->s_cap_dirty);
 759         INIT_LIST_HEAD(&s->s_cap_flushing);
 760
 761         mdsc->sessions[mds] = s;
 762         atomic_inc(&mdsc->num_sessions);
 763         refcount_inc(&s->s_ref);  /* one ref to sessions[], one to caller */
 764
 765         ceph_con_open(&s->s_con, CEPH_ENTITY_TYPE_MDS, mds,
 766                       ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
 767
 768         return s;
 769
 770 fail_realloc:
 771         kfree(s);
 772         return ERR_PTR(-ENOMEM);
 773 }
 774
 775 /*
 776  * called under mdsc->mutex
 777  */
 778 static void __unregister_session(struct ceph_mds_client *mdsc,
 779                                struct ceph_mds_session *s)
 780 {
 781         dout("__unregister_session mds%d %p\n", s->s_mds, s);
 782         BUG_ON(mdsc->sessions[s->s_mds] != s);
 783         mdsc->sessions[s->s_mds] = NULL;
 784         ceph_con_close(&s->s_con);
 785         ceph_put_mds_session(s);
 786         atomic_dec(&mdsc->num_sessions);
 787 }
 788
 789 /*
 790  * drop session refs in request.
 791  *
 792  * should be last request ref, or hold mdsc->mutex
 793  */
 794 static void put_request_session(struct ceph_mds_request *req)
 795 {
 796         if (req->r_session) {
 797                 ceph_put_mds_session(req->r_session);
 798                 req->r_session = NULL;
 799         }
 800 }
 801
 802 void ceph_mdsc_release_request(struct kref *kref)
 803 {
 804         struct ceph_mds_request *req = container_of(kref,
 805                                                     struct ceph_mds_request,
 806                                                     r_kref);
 807         ceph_mdsc_release_dir_caps_no_check(req);
 808         destroy_reply_info(&req->r_reply_info);
 809         if (req->r_request)
 810                 ceph_msg_put(req->r_request);
 811         if (req->r_reply)
 812                 ceph_msg_put(req->r_reply);
 813         if (req->r_inode) {
 814                 ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
 815                 /* avoid calling iput_final() in mds dispatch threads */
 816                 ceph_async_iput(req->r_inode);
 817         }
 818         if (req->r_parent) {
 819                 ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
 820                 ceph_async_iput(req->r_parent);
 821         }
 822         ceph_async_iput(req->r_target_inode);
 823         if (req->r_dentry)
 824                 dput(req->r_dentry);
 825         if (req->r_old_dentry)
 826                 dput(req->r_old_dentry);
 827         if (req->r_old_dentry_dir) {
 828                 /*
 829                  * track (and drop pins for) r_old_dentry_dir
 830                  * separately, since r_old_dentry's d_parent may have
 831                  * changed between the dir mutex being dropped and
 832                  * this request being freed.
 833                  */
 834                 ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
 835                                   CEPH_CAP_PIN);
 836                 ceph_async_iput(req->r_old_dentry_dir);
 837         }
 838         kfree(req->r_path1);
 839         kfree(req->r_path2);
 840         if (req->r_pagelist)
 841                 ceph_pagelist_release(req->r_pagelist);
 842         put_request_session(req);
 843         ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);
 844         WARN_ON_ONCE(!list_empty(&req->r_wait));
 845         kmem_cache_free(ceph_mds_request_cachep, req);
 846 }
 847
 848 DEFINE_RB_FUNCS(request, struct ceph_mds_request, r_tid, r_node)
 849
 850 /*
 851  * lookup session, bump ref if found.
 852  *
 853  * called under mdsc->mutex.
 854  */
 855 static struct ceph_mds_request *
 856 lookup_get_request(struct ceph_mds_client *mdsc, u64 tid)
 857 {
 858         struct ceph_mds_request *req;
 859
 860         req = lookup_request(&mdsc->request_tree, tid);
 861         if (req)
 862                 ceph_mdsc_get_request(req);
 863
 864         return req;
 865 }
 866
 867 /*
 868  * Register an in-flight request, and assign a tid.  Link to directory
 869  * are modifying (if any).
 870  *
 871  * Called under mdsc->mutex.
 872  */
 873 static void __register_request(struct ceph_mds_client *mdsc,
 874                                struct ceph_mds_request *req,
 875                                struct inode *dir)
 876 {
 877         int ret = 0;
 878
 879         req->r_tid = ++mdsc->last_tid;
 880         if (req->r_num_caps) {
 881                 ret = ceph_reserve_caps(mdsc, &req->r_caps_reservation,
 882                                         req->r_num_caps);
 883                 if (ret < 0) {
 884                         pr_err("__register_request %p "
 885                                "failed to reserve caps: %d\n", req, ret);
 886                         /* set req->r_err to fail early from __do_request */
 887                         req->r_err = ret;
 888                         return;
 889                 }
 890         }
 891         dout("__register_request %p tid %lld\n", req, req->r_tid);
 892         ceph_mdsc_get_request(req);
 893         insert_request(&mdsc->request_tree, req);
 894
 895         req->r_uid = current_fsuid();
 896         req->r_gid = current_fsgid();
 897
 898         if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)
 899                 mdsc->oldest_tid = req->r_tid;
 900
 901         if (dir) {
 902                 struct ceph_inode_info *ci = ceph_inode(dir);
 903
 904                 ihold(dir);
 905                 req->r_unsafe_dir = dir;
 906                 spin_lock(&ci->i_unsafe_lock);
 907                 list_add_tail(&req->r_unsafe_dir_item, &ci->i_unsafe_dirops);
 908                 spin_unlock(&ci->i_unsafe_lock);
 909         }
 910 }
 911
 912 static void __unregister_request(struct ceph_mds_client *mdsc,
 913                                  struct ceph_mds_request *req)
 914 {
 915         dout("__unregister_request %p tid %lld\n", req, req->r_tid);
 916
 917         /* Never leave an unregistered request on an unsafe list! */
 918         list_del_init(&req->r_unsafe_item);
 919
 920         if (req->r_tid == mdsc->oldest_tid) {
 921                 struct rb_node *p = rb_next(&req->r_node);
 922                 mdsc->oldest_tid = 0;
 923                 while (p) {
 924                         struct ceph_mds_request *next_req =
 925                                 rb_entry(p, struct ceph_mds_request, r_node);
 926                         if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {
 927                                 mdsc->oldest_tid = next_req->r_tid;
 928                                 break;
 929                         }
 930                         p = rb_next(p);
 931                 }
 932         }
 933
 934         erase_request(&mdsc->request_tree, req);
 935
 936         if (req->r_unsafe_dir) {
 937                 struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
 938                 spin_lock(&ci->i_unsafe_lock);
 939                 list_del_init(&req->r_unsafe_dir_item);
 940                 spin_unlock(&ci->i_unsafe_lock);
 941         }
 942         if (req->r_target_inode &&
 943             test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
 944                 struct ceph_inode_info *ci = ceph_inode(req->r_target_inode);
 945                 spin_lock(&ci->i_unsafe_lock);
 946                 list_del_init(&req->r_unsafe_target_item);
 947                 spin_unlock(&ci->i_unsafe_lock);
 948         }
 949
 950         if (req->r_unsafe_dir) {
 951                 /* avoid calling iput_final() in mds dispatch threads */
 952                 ceph_async_iput(req->r_unsafe_dir);
 953                 req->r_unsafe_dir = NULL;
 954         }
 955
 956         complete_all(&req->r_safe_completion);
 957
 958         ceph_mdsc_put_request(req);
 959 }
 960
 961 /*
 962  * Walk back up the dentry tree until we hit a dentry representing a
 963  * non-snapshot inode. We do this using the rcu_read_lock (which must be held
 964  * when calling this) to ensure that the objects won't disappear while we're
 965  * working with them. Once we hit a candidate dentry, we attempt to take a
 966  * reference to it, and return that as the result.
 967  */
 968 static struct inode *get_nonsnap_parent(struct dentry *dentry)
 969 {
 970         struct inode *inode = NULL;
 971
 972         while (dentry && !IS_ROOT(dentry)) {
 973                 inode = d_inode_rcu(dentry);
 974                 if (!inode || ceph_snap(inode) == CEPH_NOSNAP)
 975                         break;
 976                 dentry = dentry->d_parent;
 977         }
 978         if (inode)
 979                 inode = igrab(inode);
 980         return inode;
 981 }
 982
 983 /*
 984  * Choose mds to send request to next.  If there is a hint set in the
 985  * request (e.g., due to a prior forward hint from the mds), use that.
 986  * Otherwise, consult frag tree and/or caps to identify the
 987  * appropriate mds.  If all else fails, choose randomly.
 988  *
 989  * Called under mdsc->mutex.
 990  */
 991 static int __choose_mds(struct ceph_mds_client *mdsc,
 992                         struct ceph_mds_request *req,
 993                         bool *random)
 994 {
 995         struct inode *inode;
 996         struct ceph_inode_info *ci;
 997         struct ceph_cap *cap;
 998         int mode = req->r_direct_mode;
 999         int mds = -1;
1000         u32 hash = req->r_direct_hash;
1001         bool is_hash = test_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags);
1002
1003         if (random)
1004                 *random = false;
1005
1006         /*
1007          * is there a specific mds we should try?  ignore hint if we have
1008          * no session and the mds is not up (active or recovering).
1009          */
1010         if (req->r_resend_mds >= 0 &&
1011             (__have_session(mdsc, req->r_resend_mds) ||
1012              ceph_mdsmap_get_state(mdsc->mdsmap, req->r_resend_mds) > 0)) {
1013                 dout("%s using resend_mds mds%d\n", __func__,
1014                      req->r_resend_mds);
1015                 return req->r_resend_mds;
1016         }
1017
1018         if (mode == USE_RANDOM_MDS)
1019                 goto random;
1020
1021         inode = NULL;
1022         if (req->r_inode) {
1023                 if (ceph_snap(req->r_inode) != CEPH_SNAPDIR) {
1024                         inode = req->r_inode;
1025                         ihold(inode);
1026                 } else {
1027                         /* req->r_dentry is non-null for LSSNAP request */
1028                         rcu_read_lock();
1029                         inode = get_nonsnap_parent(req->r_dentry);
1030                         rcu_read_unlock();
1031                         dout("%s using snapdir's parent %p\n", __func__, inode);
1032                 }
1033         } else if (req->r_dentry) {
1034                 /* ignore race with rename; old or new d_parent is okay */
1035                 struct dentry *parent;
1036                 struct inode *dir;
1037
1038                 rcu_read_lock();
1039                 parent = READ_ONCE(req->r_dentry->d_parent);
1040                 dir = req->r_parent ? : d_inode_rcu(parent);
1041
1042                 if (!dir || dir->i_sb != mdsc->fsc->sb) {
1043                         /*  not this fs or parent went negative */
1044                         inode = d_inode(req->r_dentry);
1045                         if (inode)
1046                                 ihold(inode);
1047                 } else if (ceph_snap(dir) != CEPH_NOSNAP) {
1048                         /* direct snapped/virtual snapdir requests
1049                          * based on parent dir inode */
1050                         inode = get_nonsnap_parent(parent);
1051                         dout("%s using nonsnap parent %p\n", __func__, inode);
1052                 } else {
1053                         /* dentry target */
1054                         inode = d_inode(req->r_dentry);
1055                         if (!inode || mode == USE_AUTH_MDS) {
1056                                 /* dir + name */
1057                                 inode = igrab(dir);
1058                                 hash = ceph_dentry_hash(dir, req->r_dentry);
1059                                 is_hash = true;
1060                         } else {
1061                                 ihold(inode);
1062                         }
1063                 }
1064                 rcu_read_unlock();
1065         }
1066
1067         dout("%s %p is_hash=%d (0x%x) mode %d\n", __func__, inode, (int)is_hash,
1068              hash, mode);
1069         if (!inode)
1070                 goto random;
1071         ci = ceph_inode(inode);
1072
1073         if (is_hash && S_ISDIR(inode->i_mode)) {
1074                 struct ceph_inode_frag frag;
1075                 int found;
1076
1077                 ceph_choose_frag(ci, hash, &frag, &found);
1078                 if (found) {
1079                         if (mode == USE_ANY_MDS && frag.ndist > 0) {
1080                                 u8 r;
1081
1082                                 /* choose a random replica */
1083                                 get_random_bytes(&r, 1);
1084                                 r %= frag.ndist;
1085                                 mds = frag.dist[r];
1086                                 dout("%s %p %llx.%llx frag %u mds%d (%d/%d)\n",
1087                                      __func__, inode, ceph_vinop(inode),
1088                                      frag.frag, mds, (int)r, frag.ndist);
1089                                 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
1090                                     CEPH_MDS_STATE_ACTIVE &&
1091                                     !ceph_mdsmap_is_laggy(mdsc->mdsmap, mds))
1092                                         goto out;
1093                         }
1094
1095                         /* since this file/dir wasn't known to be
1096                          * replicated, then we want to look for the
1097                          * authoritative mds. */
1098                         if (frag.mds >= 0) {
1099                                 /* choose auth mds */
1100                                 mds = frag.mds;
1101                                 dout("%s %p %llx.%llx frag %u mds%d (auth)\n",
1102                                      __func__, inode, ceph_vinop(inode),
1103                                      frag.frag, mds);
1104                                 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
1105                                     CEPH_MDS_STATE_ACTIVE) {
1106                                         if (mode == USE_ANY_MDS &&
1107                                             !ceph_mdsmap_is_laggy(mdsc->mdsmap,
1108                                                                   mds))
1109                                                 goto out;
1110                                 }
1111                         }
1112                         mode = USE_AUTH_MDS;
1113                 }
1114         }
1115
1116         spin_lock(&ci->i_ceph_lock);
1117         cap = NULL;
1118         if (mode == USE_AUTH_MDS)
1119                 cap = ci->i_auth_cap;
1120         if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
1121                 cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
1122         if (!cap) {
1123                 spin_unlock(&ci->i_ceph_lock);
1124                 ceph_async_iput(inode);
1125                 goto random;
1126         }
1127         mds = cap->session->s_mds;
1128         dout("%s %p %llx.%llx mds%d (%scap %p)\n", __func__,
1129              inode, ceph_vinop(inode), mds,
1130              cap == ci->i_auth_cap ? "auth " : "", cap);
1131         spin_unlock(&ci->i_ceph_lock);
1132 out:
1133         /* avoid calling iput_final() while holding mdsc->mutex or
1134          * in mds dispatch threads */
1135         ceph_async_iput(inode);
1136         return mds;
1137
1138 random:
1139         if (random)
1140                 *random = true;
1141
1142         mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);
1143         dout("%s chose random mds%d\n", __func__, mds);
1144         return mds;
1145 }
1146
1147
1148 /*
1149  * session messages
1150  */
1151 static struct ceph_msg *create_session_msg(u32 op, u64 seq)
1152 {
1153         struct ceph_msg *msg;
1154         struct ceph_mds_session_head *h;
1155
1156         msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h), GFP_NOFS,
1157                            false);
1158         if (!msg) {
1159                 pr_err("create_session_msg ENOMEM creating msg\n");
1160                 return NULL;
1161         }
1162         h = msg->front.iov_base;
1163         h->op = cpu_to_le32(op);
1164         h->seq = cpu_to_le64(seq);
1165
1166         return msg;
1167 }
1168
1169 static const unsigned char feature_bits[] = CEPHFS_FEATURES_CLIENT_SUPPORTED;
1170 #define FEATURE_BYTES(c) (DIV_ROUND_UP((size_t)feature_bits[c - 1] + 1, 64) * 8)
1171 static void encode_supported_features(void **p, void *end)
1172 {
1173         static const size_t count = ARRAY_SIZE(feature_bits);
1174
1175         if (count > 0) {
1176                 size_t i;
1177                 size_t size = FEATURE_BYTES(count);
1178
1179                 BUG_ON(*p + 4 + size > end);
1180                 ceph_encode_32(p, size);
1181                 memset(*p, 0, size);
1182                 for (i = 0; i < count; i++)
1183                         ((unsigned char*)(*p))[i / 8] |= BIT(feature_bits[i] % 8);
1184                 *p += size;
1185         } else {
1186                 BUG_ON(*p + 4 > end);
1187                 ceph_encode_32(p, 0);
1188         }
1189 }
1190
1191 /*
1192  * session message, specialization for CEPH_SESSION_REQUEST_OPEN
1193  * to include additional client metadata fields.
1194  */
1195 static struct ceph_msg *create_session_open_msg(struct ceph_mds_client *mdsc, u64 seq)
1196 {
1197         struct ceph_msg *msg;
1198         struct ceph_mds_session_head *h;
1199         int i = -1;
1200         int extra_bytes = 0;
1201         int metadata_key_count = 0;
1202         struct ceph_options *opt = mdsc->fsc->client->options;
1203         struct ceph_mount_options *fsopt = mdsc->fsc->mount_options;
1204         size_t size, count;
1205         void *p, *end;
1206
1207         const char* metadata[][2] = {
1208                 {"hostname", mdsc->nodename},
1209                 {"kernel_version", init_utsname()->release},
1210                 {"entity_id", opt->name ? : ""},
1211                 {"root", fsopt->server_path ? : "/"},
1212                 {NULL, NULL}
1213         };
1214
1215         /* Calculate serialized length of metadata */
1216         extra_bytes = 4;  /* map length */
1217         for (i = 0; metadata[i][0]; ++i) {
1218                 extra_bytes += 8 + strlen(metadata[i][0]) +
1219                         strlen(metadata[i][1]);
1220                 metadata_key_count++;
1221         }
1222
1223         /* supported feature */
1224         size = 0;
1225         count = ARRAY_SIZE(feature_bits);
1226         if (count > 0)
1227                 size = FEATURE_BYTES(count);
1228         extra_bytes += 4 + size;
1229
1230         /* Allocate the message */
1231         msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h) + extra_bytes,
1232                            GFP_NOFS, false);
1233         if (!msg) {
1234                 pr_err("create_session_msg ENOMEM creating msg\n");
1235                 return NULL;
1236         }
1237         p = msg->front.iov_base;
1238         end = p + msg->front.iov_len;
1239
1240         h = p;
1241         h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);
1242         h->seq = cpu_to_le64(seq);
1243
1244         /*
1245          * Serialize client metadata into waiting buffer space, using
1246          * the format that userspace expects for map<string, string>
1247          *
1248          * ClientSession messages with metadata are v3
1249          */
1250         msg->hdr.version = cpu_to_le16(3);
1251         msg->hdr.compat_version = cpu_to_le16(1);
1252
1253         /* The write pointer, following the session_head structure */
1254         p += sizeof(*h);
1255
1256         /* Number of entries in the map */
1257         ceph_encode_32(&p, metadata_key_count);
1258
1259         /* Two length-prefixed strings for each entry in the map */
1260         for (i = 0; metadata[i][0]; ++i) {
1261                 size_t const key_len = strlen(metadata[i][0]);
1262                 size_t const val_len = strlen(metadata[i][1]);
1263
1264                 ceph_encode_32(&p, key_len);
1265                 memcpy(p, metadata[i][0], key_len);
1266                 p += key_len;
1267                 ceph_encode_32(&p, val_len);
1268                 memcpy(p, metadata[i][1], val_len);
1269                 p += val_len;
1270         }
1271
1272         encode_supported_features(&p, end);
1273         msg->front.iov_len = p - msg->front.iov_base;
1274         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1275
1276         return msg;
1277 }
1278
1279 /*
1280  * send session open request.
1281  *
1282  * called under mdsc->mutex
1283  */
1284 static int __open_session(struct ceph_mds_client *mdsc,
1285                           struct ceph_mds_session *session)
1286 {
1287         struct ceph_msg *msg;
1288         int mstate;
1289         int mds = session->s_mds;
1290
1291         /* wait for mds to go active? */
1292         mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
1293         dout("open_session to mds%d (%s)\n", mds,
1294              ceph_mds_state_name(mstate));
1295         session->s_state = CEPH_MDS_SESSION_OPENING;
1296         session->s_renew_requested = jiffies;
1297
1298         /* send connect message */
1299         msg = create_session_open_msg(mdsc, session->s_seq);
1300         if (!msg)
1301                 return -ENOMEM;
1302         ceph_con_send(&session->s_con, msg);
1303         return 0;
1304 }
1305
1306 /*
1307  * open sessions for any export targets for the given mds
1308  *
1309  * called under mdsc->mutex
1310  */
1311 static struct ceph_mds_session *
1312 __open_export_target_session(struct ceph_mds_client *mdsc, int target)
1313 {
1314         struct ceph_mds_session *session;
1315
1316         session = __ceph_lookup_mds_session(mdsc, target);
1317         if (!session) {
1318                 session = register_session(mdsc, target);
1319                 if (IS_ERR(session))
1320                         return session;
1321         }
1322         if (session->s_state == CEPH_MDS_SESSION_NEW ||
1323             session->s_state == CEPH_MDS_SESSION_CLOSING)
1324                 __open_session(mdsc, session);
1325
1326         return session;
1327 }
1328
1329 struct ceph_mds_session *
1330 ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target)
1331 {
1332         struct ceph_mds_session *session;
1333
1334         dout("open_export_target_session to mds%d\n", target);
1335
1336         mutex_lock(&mdsc->mutex);
1337         session = __open_export_target_session(mdsc, target);
1338         mutex_unlock(&mdsc->mutex);
1339
1340         return session;
1341 }
1342
1343 static void __open_export_target_sessions(struct ceph_mds_client *mdsc,
1344                                           struct ceph_mds_session *session)
1345 {
1346         struct ceph_mds_info *mi;
1347         struct ceph_mds_session *ts;
1348         int i, mds = session->s_mds;
1349
1350         if (mds >= mdsc->mdsmap->possible_max_rank)
1351                 return;
1352
1353         mi = &mdsc->mdsmap->m_info[mds];
1354         dout("open_export_target_sessions for mds%d (%d targets)\n",
1355              session->s_mds, mi->num_export_targets);
1356
1357         for (i = 0; i < mi->num_export_targets; i++) {
1358                 ts = __open_export_target_session(mdsc, mi->export_targets[i]);
1359                 if (!IS_ERR(ts))
1360                         ceph_put_mds_session(ts);
1361         }
1362 }
1363
1364 void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
1365                                            struct ceph_mds_session *session)
1366 {
1367         mutex_lock(&mdsc->mutex);
1368         __open_export_target_sessions(mdsc, session);
1369         mutex_unlock(&mdsc->mutex);
1370 }
1371
1372 /*
1373  * session caps
1374  */
1375
1376 static void detach_cap_releases(struct ceph_mds_session *session,
1377                                 struct list_head *target)
1378 {
1379         lockdep_assert_held(&session->s_cap_lock);
1380
1381         list_splice_init(&session->s_cap_releases, target);
1382         session->s_num_cap_releases = 0;
1383         dout("dispose_cap_releases mds%d\n", session->s_mds);
1384 }
1385
1386 static void dispose_cap_releases(struct ceph_mds_client *mdsc,
1387                                  struct list_head *dispose)
1388 {
1389         while (!list_empty(dispose)) {
1390                 struct ceph_cap *cap;
1391                 /* zero out the in-progress message */
1392                 cap = list_first_entry(dispose, struct ceph_cap, session_caps);
1393                 list_del(&cap->session_caps);
1394                 ceph_put_cap(mdsc, cap);
1395         }
1396 }
1397
1398 static void cleanup_session_requests(struct ceph_mds_client *mdsc,
1399                                      struct ceph_mds_session *session)
1400 {
1401         struct ceph_mds_request *req;
1402         struct rb_node *p;
1403         struct ceph_inode_info *ci;
1404
1405         dout("cleanup_session_requests mds%d\n", session->s_mds);
1406         mutex_lock(&mdsc->mutex);
1407         while (!list_empty(&session->s_unsafe)) {
1408                 req = list_first_entry(&session->s_unsafe,
1409                                        struct ceph_mds_request, r_unsafe_item);
1410                 pr_warn_ratelimited(" dropping unsafe request %llu\n",
1411                                     req->r_tid);
1412                 if (req->r_target_inode) {
1413                         /* dropping unsafe change of inode's attributes */
1414                         ci = ceph_inode(req->r_target_inode);
1415                         errseq_set(&ci->i_meta_err, -EIO);
1416                 }
1417                 if (req->r_unsafe_dir) {
1418                         /* dropping unsafe directory operation */
1419                         ci = ceph_inode(req->r_unsafe_dir);
1420                         errseq_set(&ci->i_meta_err, -EIO);
1421                 }
1422                 __unregister_request(mdsc, req);
1423         }
1424         /* zero r_attempts, so kick_requests() will re-send requests */
1425         p = rb_first(&mdsc->request_tree);
1426         while (p) {
1427                 req = rb_entry(p, struct ceph_mds_request, r_node);
1428                 p = rb_next(p);
1429                 if (req->r_session &&
1430                     req->r_session->s_mds == session->s_mds)
1431                         req->r_attempts = 0;
1432         }
1433         mutex_unlock(&mdsc->mutex);
1434 }
1435
1436 /*
1437  * Helper to safely iterate over all caps associated with a session, with
1438  * special care taken to handle a racing __ceph_remove_cap().
1439  *
1440  * Caller must hold session s_mutex.
1441  */
1442 int ceph_iterate_session_caps(struct ceph_mds_session *session,
1443                               int (*cb)(struct inode *, struct ceph_cap *,
1444                                         void *), void *arg)
1445 {
1446         struct list_head *p;
1447         struct ceph_cap *cap;
1448         struct inode *inode, *last_inode = NULL;
1449         struct ceph_cap *old_cap = NULL;
1450         int ret;
1451
1452         dout("iterate_session_caps %p mds%d\n", session, session->s_mds);
1453         spin_lock(&session->s_cap_lock);
1454         p = session->s_caps.next;
1455         while (p != &session->s_caps) {
1456                 cap = list_entry(p, struct ceph_cap, session_caps);
1457                 inode = igrab(&cap->ci->vfs_inode);
1458                 if (!inode) {
1459                         p = p->next;
1460                         continue;
1461                 }
1462                 session->s_cap_iterator = cap;
1463                 spin_unlock(&session->s_cap_lock);
1464
1465                 if (last_inode) {
1466                         /* avoid calling iput_final() while holding
1467                          * s_mutex or in mds dispatch threads */
1468                         ceph_async_iput(last_inode);
1469                         last_inode = NULL;
1470                 }
1471                 if (old_cap) {
1472                         ceph_put_cap(session->s_mdsc, old_cap);
1473                         old_cap = NULL;
1474                 }
1475
1476                 ret = cb(inode, cap, arg);
1477                 last_inode = inode;
1478
1479                 spin_lock(&session->s_cap_lock);
1480                 p = p->next;
1481                 if (!cap->ci) {
1482                         dout("iterate_session_caps  finishing cap %p removal\n",
1483                              cap);
1484                         BUG_ON(cap->session != session);
1485                         cap->session = NULL;
1486                         list_del_init(&cap->session_caps);
1487                         session->s_nr_caps--;
1488                         if (cap->queue_release)
1489                                 __ceph_queue_cap_release(session, cap);
1490                         else
1491                                 old_cap = cap;  /* put_cap it w/o locks held */
1492                 }
1493                 if (ret < 0)
1494                         goto out;
1495         }
1496         ret = 0;
1497 out:
1498         session->s_cap_iterator = NULL;
1499         spin_unlock(&session->s_cap_lock);
1500
1501         ceph_async_iput(last_inode);
1502         if (old_cap)
1503                 ceph_put_cap(session->s_mdsc, old_cap);
1504
1505         return ret;
1506 }
1507
1508 static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
1509                                   void *arg)
1510 {
1511         struct ceph_fs_client *fsc = (struct ceph_fs_client *)arg;
1512         struct ceph_inode_info *ci = ceph_inode(inode);
1513         LIST_HEAD(to_remove);
1514         bool dirty_dropped = false;
1515         bool invalidate = false;
1516
1517         dout("removing cap %p, ci is %p, inode is %p\n",
1518              cap, ci, &ci->vfs_inode);
1519         spin_lock(&ci->i_ceph_lock);
1520         __ceph_remove_cap(cap, false);
1521         if (!ci->i_auth_cap) {
1522                 struct ceph_cap_flush *cf;
1523                 struct ceph_mds_client *mdsc = fsc->mdsc;
1524
1525                 if (READ_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
1526                         if (inode->i_data.nrpages > 0)
1527                                 invalidate = true;
1528                         if (ci->i_wrbuffer_ref > 0)
1529                                 mapping_set_error(&inode->i_data, -EIO);
1530                 }
1531
1532                 while (!list_empty(&ci->i_cap_flush_list)) {
1533                         cf = list_first_entry(&ci->i_cap_flush_list,
1534                                               struct ceph_cap_flush, i_list);
1535                         list_move(&cf->i_list, &to_remove);
1536                 }
1537
1538                 spin_lock(&mdsc->cap_dirty_lock);
1539
1540                 list_for_each_entry(cf, &to_remove, i_list)
1541                         list_del(&cf->g_list);
1542
1543                 if (!list_empty(&ci->i_dirty_item)) {
1544                         pr_warn_ratelimited(
1545                                 " dropping dirty %s state for %p %lld\n",
1546                                 ceph_cap_string(ci->i_dirty_caps),
1547                                 inode, ceph_ino(inode));
1548                         ci->i_dirty_caps = 0;
1549                         list_del_init(&ci->i_dirty_item);
1550                         dirty_dropped = true;
1551                 }
1552                 if (!list_empty(&ci->i_flushing_item)) {
1553                         pr_warn_ratelimited(
1554                                 " dropping dirty+flushing %s state for %p %lld\n",
1555                                 ceph_cap_string(ci->i_flushing_caps),
1556                                 inode, ceph_ino(inode));
1557                         ci->i_flushing_caps = 0;
1558                         list_del_init(&ci->i_flushing_item);
1559                         mdsc->num_cap_flushing--;
1560                         dirty_dropped = true;
1561                 }
1562                 spin_unlock(&mdsc->cap_dirty_lock);
1563
1564                 if (dirty_dropped) {
1565                         errseq_set(&ci->i_meta_err, -EIO);
1566
1567                         if (ci->i_wrbuffer_ref_head == 0 &&
1568                             ci->i_wr_ref == 0 &&
1569                             ci->i_dirty_caps == 0 &&
1570                             ci->i_flushing_caps == 0) {
1571                                 ceph_put_snap_context(ci->i_head_snapc);
1572                                 ci->i_head_snapc = NULL;
1573                         }
1574                 }
1575
1576                 if (atomic_read(&ci->i_filelock_ref) > 0) {
1577                         /* make further file lock syscall return -EIO */
1578                         ci->i_ceph_flags |= CEPH_I_ERROR_FILELOCK;
1579                         pr_warn_ratelimited(" dropping file locks for %p %lld\n",
1580                                             inode, ceph_ino(inode));
1581                 }
1582
1583                 if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {
1584                         list_add(&ci->i_prealloc_cap_flush->i_list, &to_remove);
1585                         ci->i_prealloc_cap_flush = NULL;
1586                 }
1587         }
1588         spin_unlock(&ci->i_ceph_lock);
1589         while (!list_empty(&to_remove)) {
1590                 struct ceph_cap_flush *cf;
1591                 cf = list_first_entry(&to_remove,
1592                                       struct ceph_cap_flush, i_list);
1593                 list_del(&cf->i_list);
1594                 ceph_free_cap_flush(cf);
1595         }
1596
1597         wake_up_all(&ci->i_cap_wq);
1598         if (invalidate)
1599                 ceph_queue_invalidate(inode);
1600         if (dirty_dropped)
1601                 iput(inode);
1602         return 0;
1603 }
1604
1605 /*
1606  * caller must hold session s_mutex
1607  */
1608 static void remove_session_caps(struct ceph_mds_session *session)
1609 {
1610         struct ceph_fs_client *fsc = session->s_mdsc->fsc;
1611         struct super_block *sb = fsc->sb;
1612         LIST_HEAD(dispose);
1613
1614         dout("remove_session_caps on %p\n", session);
1615         ceph_iterate_session_caps(session, remove_session_caps_cb, fsc);
1616
1617         wake_up_all(&fsc->mdsc->cap_flushing_wq);
1618
1619         spin_lock(&session->s_cap_lock);
1620         if (session->s_nr_caps > 0) {
1621                 struct inode *inode;
1622                 struct ceph_cap *cap, *prev = NULL;
1623                 struct ceph_vino vino;
1624                 /*
1625                  * iterate_session_caps() skips inodes that are being
1626                  * deleted, we need to wait until deletions are complete.
1627                  * __wait_on_freeing_inode() is designed for the job,
1628                  * but it is not exported, so use lookup inode function
1629                  * to access it.
1630                  */
1631                 while (!list_empty(&session->s_caps)) {
1632                         cap = list_entry(session->s_caps.next,
1633                                          struct ceph_cap, session_caps);
1634                         if (cap == prev)
1635                                 break;
1636                         prev = cap;
1637                         vino = cap->ci->i_vino;
1638                         spin_unlock(&session->s_cap_lock);
1639
1640                         inode = ceph_find_inode(sb, vino);
1641                          /* avoid calling iput_final() while holding s_mutex */
1642                         ceph_async_iput(inode);
1643
1644                         spin_lock(&session->s_cap_lock);
1645                 }
1646         }
1647
1648         // drop cap expires and unlock s_cap_lock
1649         detach_cap_releases(session, &dispose);
1650
1651         BUG_ON(session->s_nr_caps > 0);
1652         BUG_ON(!list_empty(&session->s_cap_flushing));
1653         spin_unlock(&session->s_cap_lock);
1654         dispose_cap_releases(session->s_mdsc, &dispose);
1655 }
1656
1657 enum {
1658         RECONNECT,
1659         RENEWCAPS,
1660         FORCE_RO,
1661 };
1662
1663 /*
1664  * wake up any threads waiting on this session's caps.  if the cap is
1665  * old (didn't get renewed on the client reconnect), remove it now.
1666  *
1667  * caller must hold s_mutex.
1668  */
1669 static int wake_up_session_cb(struct inode *inode, struct ceph_cap *cap,
1670                               void *arg)
1671 {
1672         struct ceph_inode_info *ci = ceph_inode(inode);
1673         unsigned long ev = (unsigned long)arg;
1674
1675         if (ev == RECONNECT) {
1676                 spin_lock(&ci->i_ceph_lock);
1677                 ci->i_wanted_max_size = 0;
1678                 ci->i_requested_max_size = 0;
1679                 spin_unlock(&ci->i_ceph_lock);
1680         } else if (ev == RENEWCAPS) {
1681                 if (cap->cap_gen < cap->session->s_cap_gen) {
1682                         /* mds did not re-issue stale cap */
1683                         spin_lock(&ci->i_ceph_lock);
1684                         cap->issued = cap->implemented = CEPH_CAP_PIN;
1685                         spin_unlock(&ci->i_ceph_lock);
1686                 }
1687         } else if (ev == FORCE_RO) {
1688         }
1689         wake_up_all(&ci->i_cap_wq);
1690         return 0;
1691 }
1692
1693 static void wake_up_session_caps(struct ceph_mds_session *session, int ev)
1694 {
1695         dout("wake_up_session_caps %p mds%d\n", session, session->s_mds);
1696         ceph_iterate_session_caps(session, wake_up_session_cb,
1697                                   (void *)(unsigned long)ev);
1698 }
1699
1700 /*
1701  * Send periodic message to MDS renewing all currently held caps.  The
1702  * ack will reset the expiration for all caps from this session.
1703  *
1704  * caller holds s_mutex
1705  */
1706 static int send_renew_caps(struct ceph_mds_client *mdsc,
1707                            struct ceph_mds_session *session)
1708 {
1709         struct ceph_msg *msg;
1710         int state;
1711
1712         if (time_after_eq(jiffies, session->s_cap_ttl) &&
1713             time_after_eq(session->s_cap_ttl, session->s_renew_requested))
1714                 pr_info("mds%d caps stale\n", session->s_mds);
1715         session->s_renew_requested = jiffies;
1716
1717         /* do not try to renew caps until a recovering mds has reconnected
1718          * with its clients. */
1719         state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);
1720         if (state < CEPH_MDS_STATE_RECONNECT) {
1721                 dout("send_renew_caps ignoring mds%d (%s)\n",
1722                      session->s_mds, ceph_mds_state_name(state));
1723                 return 0;
1724         }
1725
1726         dout("send_renew_caps to mds%d (%s)\n", session->s_mds,
1727                 ceph_mds_state_name(state));
1728         msg = create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
1729                                  ++session->s_renew_seq);
1730         if (!msg)
1731                 return -ENOMEM;
1732         ceph_con_send(&session->s_con, msg);
1733         return 0;
1734 }
1735
1736 static int send_flushmsg_ack(struct ceph_mds_client *mdsc,
1737                              struct ceph_mds_session *session, u64 seq)
1738 {
1739         struct ceph_msg *msg;
1740
1741         dout("send_flushmsg_ack to mds%d (%s)s seq %lld\n",
1742              session->s_mds, ceph_session_state_name(session->s_state), seq);
1743         msg = create_session_msg(CEPH_SESSION_FLUSHMSG_ACK, seq);
1744         if (!msg)
1745                 return -ENOMEM;
1746         ceph_con_send(&session->s_con, msg);
1747         return 0;
1748 }
1749
1750
1751 /*
1752  * Note new cap ttl, and any transition from stale -> not stale (fresh?).
1753  *
1754  * Called under session->s_mutex
1755  */
1756 static void renewed_caps(struct ceph_mds_client *mdsc,
1757                          struct ceph_mds_session *session, int is_renew)
1758 {
1759         int was_stale;
1760         int wake = 0;
1761
1762         spin_lock(&session->s_cap_lock);
1763         was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);
1764
1765         session->s_cap_ttl = session->s_renew_requested +
1766                 mdsc->mdsmap->m_session_timeout*HZ;
1767
1768         if (was_stale) {
1769                 if (time_before(jiffies, session->s_cap_ttl)) {
1770                         pr_info("mds%d caps renewed\n", session->s_mds);
1771                         wake = 1;
1772                 } else {
1773                         pr_info("mds%d caps still stale\n", session->s_mds);
1774                 }
1775         }
1776         dout("renewed_caps mds%d ttl now %lu, was %s, now %s\n",
1777              session->s_mds, session->s_cap_ttl, was_stale ? "stale" : "fresh",
1778              time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");
1779         spin_unlock(&session->s_cap_lock);
1780
1781         if (wake)
1782                 wake_up_session_caps(session, RENEWCAPS);
1783 }
1784
1785 /*
1786  * send a session close request
1787  */
1788 static int request_close_session(struct ceph_mds_client *mdsc,
1789                                  struct ceph_mds_session *session)
1790 {
1791         struct ceph_msg *msg;
1792
1793         dout("request_close_session mds%d state %s seq %lld\n",
1794              session->s_mds, ceph_session_state_name(session->s_state),
1795              session->s_seq);
1796         msg = create_session_msg(CEPH_SESSION_REQUEST_CLOSE, session->s_seq);
1797         if (!msg)
1798                 return -ENOMEM;
1799         ceph_con_send(&session->s_con, msg);
1800         return 1;
1801 }
1802
1803 /*
1804  * Called with s_mutex held.
1805  */
1806 static int __close_session(struct ceph_mds_client *mdsc,
1807                          struct ceph_mds_session *session)
1808 {
1809         if (session->s_state >= CEPH_MDS_SESSION_CLOSING)
1810                 return 0;
1811         session->s_state = CEPH_MDS_SESSION_CLOSING;
1812         return request_close_session(mdsc, session);
1813 }
1814
1815 static bool drop_negative_children(struct dentry *dentry)
1816 {
1817         struct dentry *child;
1818         bool all_negative = true;
1819
1820         if (!d_is_dir(dentry))
1821                 goto out;
1822
1823         spin_lock(&dentry->d_lock);
1824         list_for_each_entry(child, &dentry->d_subdirs, d_child) {
1825                 if (d_really_is_positive(child)) {
1826                         all_negative = false;
1827                         break;
1828                 }
1829         }
1830         spin_unlock(&dentry->d_lock);
1831
1832         if (all_negative)
1833                 shrink_dcache_parent(dentry);
1834 out:
1835         return all_negative;
1836 }
1837
1838 /*
1839  * Trim old(er) caps.
1840  *
1841  * Because we can't cache an inode without one or more caps, we do
1842  * this indirectly: if a cap is unused, we prune its aliases, at which
1843  * point the inode will hopefully get dropped to.
1844  *
1845  * Yes, this is a bit sloppy.  Our only real goal here is to respond to
1846  * memory pressure from the MDS, though, so it needn't be perfect.
1847  */
1848 static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
1849 {
1850         int *remaining = arg;
1851         struct ceph_inode_info *ci = ceph_inode(inode);
1852         int used, wanted, oissued, mine;
1853
1854         if (*remaining <= 0)
1855                 return -1;
1856
1857         spin_lock(&ci->i_ceph_lock);
1858         mine = cap->issued | cap->implemented;
1859         used = __ceph_caps_used(ci);
1860         wanted = __ceph_caps_file_wanted(ci);
1861         oissued = __ceph_caps_issued_other(ci, cap);
1862
1863         dout("trim_caps_cb %p cap %p mine %s oissued %s used %s wanted %s\n",
1864              inode, cap, ceph_cap_string(mine), ceph_cap_string(oissued),
1865              ceph_cap_string(used), ceph_cap_string(wanted));
1866         if (cap == ci->i_auth_cap) {
1867                 if (ci->i_dirty_caps || ci->i_flushing_caps ||
1868                     !list_empty(&ci->i_cap_snaps))
1869                         goto out;
1870                 if ((used | wanted) & CEPH_CAP_ANY_WR)
1871                         goto out;
1872                 /* Note: it's possible that i_filelock_ref becomes non-zero
1873                  * after dropping auth caps. It doesn't hurt because reply
1874                  * of lock mds request will re-add auth caps. */
1875                 if (atomic_read(&ci->i_filelock_ref) > 0)
1876                         goto out;
1877         }
1878         /* The inode has cached pages, but it's no longer used.
1879          * we can safely drop it */
1880         if (S_ISREG(inode->i_mode) &&
1881             wanted == 0 && used == CEPH_CAP_FILE_CACHE &&
1882             !(oissued & CEPH_CAP_FILE_CACHE)) {
1883           used = 0;
1884           oissued = 0;
1885         }
1886         if ((used | wanted) & ~oissued & mine)
1887                 goto out;   /* we need these caps */
1888
1889         if (oissued) {
1890                 /* we aren't the only cap.. just remove us */
1891                 __ceph_remove_cap(cap, true);
1892                 (*remaining)--;
1893         } else {
1894                 struct dentry *dentry;
1895                 /* try dropping referring dentries */
1896                 spin_unlock(&ci->i_ceph_lock);
1897                 dentry = d_find_any_alias(inode);
1898                 if (dentry && drop_negative_children(dentry)) {
1899                         int count;
1900                         dput(dentry);
1901                         d_prune_aliases(inode);
1902                         count = atomic_read(&inode->i_count);
1903                         if (count == 1)
1904                                 (*remaining)--;
1905                         dout("trim_caps_cb %p cap %p pruned, count now %d\n",
1906                              inode, cap, count);
1907                 } else {
1908                         dput(dentry);
1909                 }
1910                 return 0;
1911         }
1912
1913 out:
1914         spin_unlock(&ci->i_ceph_lock);
1915         return 0;
1916 }
1917
1918 /*
1919  * Trim session cap count down to some max number.
1920  */
1921 int ceph_trim_caps(struct ceph_mds_client *mdsc,
1922                    struct ceph_mds_session *session,
1923                    int max_caps)
1924 {
1925         int trim_caps = session->s_nr_caps - max_caps;
1926
1927         dout("trim_caps mds%d start: %d / %d, trim %d\n",
1928              session->s_mds, session->s_nr_caps, max_caps, trim_caps);
1929         if (trim_caps > 0) {
1930                 int remaining = trim_caps;
1931
1932                 ceph_iterate_session_caps(session, trim_caps_cb, &remaining);
1933                 dout("trim_caps mds%d done: %d / %d, trimmed %d\n",
1934                      session->s_mds, session->s_nr_caps, max_caps,
1935                         trim_caps - remaining);
1936         }
1937
1938         ceph_flush_cap_releases(mdsc, session);
1939         return 0;
1940 }
1941
1942 static int check_caps_flush(struct ceph_mds_client *mdsc,
1943                             u64 want_flush_tid)
1944 {
1945         int ret = 1;
1946
1947         spin_lock(&mdsc->cap_dirty_lock);
1948         if (!list_empty(&mdsc->cap_flush_list)) {
1949                 struct ceph_cap_flush *cf =
1950                         list_first_entry(&mdsc->cap_flush_list,
1951                                          struct ceph_cap_flush, g_list);
1952                 if (cf->tid <= want_flush_tid) {
1953                         dout("check_caps_flush still flushing tid "
1954                              "%llu <= %llu\n", cf->tid, want_flush_tid);
1955                         ret = 0;
1956                 }
1957         }
1958         spin_unlock(&mdsc->cap_dirty_lock);
1959         return ret;
1960 }
1961
1962 /*
1963  * flush all dirty inode data to disk.
1964  *
1965  * returns true if we've flushed through want_flush_tid
1966  */
1967 static void wait_caps_flush(struct ceph_mds_client *mdsc,
1968                             u64 want_flush_tid)
1969 {
1970         dout("check_caps_flush want %llu\n", want_flush_tid);
1971
1972         wait_event(mdsc->cap_flushing_wq,
1973                    check_caps_flush(mdsc, want_flush_tid));
1974
1975         dout("check_caps_flush ok, flushed thru %llu\n", want_flush_tid);
1976 }
1977
1978 /*
1979  * called under s_mutex
1980  */
1981 static void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
1982                                    struct ceph_mds_session *session)
1983 {
1984         struct ceph_msg *msg = NULL;
1985         struct ceph_mds_cap_release *head;
1986         struct ceph_mds_cap_item *item;
1987         struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;
1988         struct ceph_cap *cap;
1989         LIST_HEAD(tmp_list);
1990         int num_cap_releases;
1991         __le32  barrier, *cap_barrier;
1992
1993         down_read(&osdc->lock);
1994         barrier = cpu_to_le32(osdc->epoch_barrier);
1995         up_read(&osdc->lock);
1996
1997         spin_lock(&session->s_cap_lock);
1998 again:
1999         list_splice_init(&session->s_cap_releases, &tmp_list);
2000         num_cap_releases = session->s_num_cap_releases;
2001         session->s_num_cap_releases = 0;
2002         spin_unlock(&session->s_cap_lock);
2003
2004         while (!list_empty(&tmp_list)) {
2005                 if (!msg) {
2006                         msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE,
2007                                         PAGE_SIZE, GFP_NOFS, false);
2008                         if (!msg)
2009                                 goto out_err;
2010                         head = msg->front.iov_base;
2011                         head->num = cpu_to_le32(0);
2012                         msg->front.iov_len = sizeof(*head);
2013
2014                         msg->hdr.version = cpu_to_le16(2);
2015                         msg->hdr.compat_version = cpu_to_le16(1);
2016                 }
2017
2018                 cap = list_first_entry(&tmp_list, struct ceph_cap,
2019                                         session_caps);
2020                 list_del(&cap->session_caps);
2021                 num_cap_releases--;
2022
2023                 head = msg->front.iov_base;
2024                 put_unaligned_le32(get_unaligned_le32(&head->num) + 1,
2025                                    &head->num);
2026                 item = msg->front.iov_base + msg->front.iov_len;
2027                 item->ino = cpu_to_le64(cap->cap_ino);
2028                 item->cap_id = cpu_to_le64(cap->cap_id);
2029                 item->migrate_seq = cpu_to_le32(cap->mseq);
2030                 item->seq = cpu_to_le32(cap->issue_seq);
2031                 msg->front.iov_len += sizeof(*item);
2032
2033                 ceph_put_cap(mdsc, cap);
2034
2035                 if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {
2036                         // Append cap_barrier field
2037                         cap_barrier = msg->front.iov_base + msg->front.iov_len;
2038                         *cap_barrier = barrier;
2039                         msg->front.iov_len += sizeof(*cap_barrier);
2040
2041                         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2042                         dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2043                         ceph_con_send(&session->s_con, msg);
2044                         msg = NULL;
2045                 }
2046         }
2047
2048         BUG_ON(num_cap_releases != 0);
2049
2050         spin_lock(&session->s_cap_lock);
2051         if (!list_empty(&session->s_cap_releases))
2052                 goto again;
2053         spin_unlock(&session->s_cap_lock);
2054
2055         if (msg) {
2056                 // Append cap_barrier field
2057                 cap_barrier = msg->front.iov_base + msg->front.iov_len;
2058                 *cap_barrier = barrier;
2059                 msg->front.iov_len += sizeof(*cap_barrier);
2060
2061                 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2062                 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2063                 ceph_con_send(&session->s_con, msg);
2064         }
2065         return;
2066 out_err:
2067         pr_err("send_cap_releases mds%d, failed to allocate message\n",
2068                 session->s_mds);
2069         spin_lock(&session->s_cap_lock);
2070         list_splice(&tmp_list, &session->s_cap_releases);
2071         session->s_num_cap_releases += num_cap_releases;
2072         spin_unlock(&session->s_cap_lock);
2073 }
2074
2075 static void ceph_cap_release_work(struct work_struct *work)
2076 {
2077         struct ceph_mds_session *session =
2078                 container_of(work, struct ceph_mds_session, s_cap_release_work);
2079
2080         mutex_lock(&session->s_mutex);
2081         if (session->s_state == CEPH_MDS_SESSION_OPEN ||
2082             session->s_state == CEPH_MDS_SESSION_HUNG)
2083                 ceph_send_cap_releases(session->s_mdsc, session);
2084         mutex_unlock(&session->s_mutex);
2085         ceph_put_mds_session(session);
2086 }
2087
2088 void ceph_flush_cap_releases(struct ceph_mds_client *mdsc,
2089                              struct ceph_mds_session *session)
2090 {
2091         if (mdsc->stopping)
2092                 return;
2093
2094         ceph_get_mds_session(session);
2095         if (queue_work(mdsc->fsc->cap_wq,
2096                        &session->s_cap_release_work)) {
2097                 dout("cap release work queued\n");
2098         } else {
2099                 ceph_put_mds_session(session);
2100                 dout("failed to queue cap release work\n");
2101         }
2102 }
2103
2104 /*
2105  * caller holds session->s_cap_lock
2106  */
2107 void __ceph_queue_cap_release(struct ceph_mds_session *session,
2108                               struct ceph_cap *cap)
2109 {
2110         list_add_tail(&cap->session_caps, &session->s_cap_releases);
2111         session->s_num_cap_releases++;
2112
2113         if (!(session->s_num_cap_releases % CEPH_CAPS_PER_RELEASE))
2114                 ceph_flush_cap_releases(session->s_mdsc, session);
2115 }
2116
2117 static void ceph_cap_reclaim_work(struct work_struct *work)
2118 {
2119         struct ceph_mds_client *mdsc =
2120                 container_of(work, struct ceph_mds_client, cap_reclaim_work);
2121         int ret = ceph_trim_dentries(mdsc);
2122         if (ret == -EAGAIN)
2123                 ceph_queue_cap_reclaim_work(mdsc);
2124 }
2125
2126 void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc)
2127 {
2128         if (mdsc->stopping)
2129                 return;
2130
2131         if (queue_work(mdsc->fsc->cap_wq, &mdsc->cap_reclaim_work)) {
2132                 dout("caps reclaim work queued\n");
2133         } else {
2134                 dout("failed to queue caps release work\n");
2135         }
2136 }
2137
2138 void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr)
2139 {
2140         int val;
2141         if (!nr)
2142                 return;
2143         val = atomic_add_return(nr, &mdsc->cap_reclaim_pending);
2144         if ((val % CEPH_CAPS_PER_RELEASE) < nr) {
2145                 atomic_set(&mdsc->cap_reclaim_pending, 0);
2146                 ceph_queue_cap_reclaim_work(mdsc);
2147         }
2148 }
2149
2150 /*
2151  * requests
2152  */
2153
2154 int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
2155                                     struct inode *dir)
2156 {
2157         struct ceph_inode_info *ci = ceph_inode(dir);
2158         struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
2159         struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;
2160         size_t size = sizeof(struct ceph_mds_reply_dir_entry);
2161         unsigned int num_entries;
2162         int order;
2163
2164         spin_lock(&ci->i_ceph_lock);
2165         num_entries = ci->i_files + ci->i_subdirs;
2166         spin_unlock(&ci->i_ceph_lock);
2167         num_entries = max(num_entries, 1U);
2168         num_entries = min(num_entries, opt->max_readdir);
2169
2170         order = get_order(size * num_entries);
2171         while (order >= 0) {
2172                 rinfo->dir_entries = (void*)__get_free_pages(GFP_KERNEL |
2173                                                              __GFP_NOWARN,
2174                                                              order);
2175                 if (rinfo->dir_entries)
2176                         break;
2177                 order--;
2178         }
2179         if (!rinfo->dir_entries)
2180                 return -ENOMEM;
2181
2182         num_entries = (PAGE_SIZE << order) / size;
2183         num_entries = min(num_entries, opt->max_readdir);
2184
2185         rinfo->dir_buf_size = PAGE_SIZE << order;
2186         req->r_num_caps = num_entries + 1;
2187         req->r_args.readdir.max_entries = cpu_to_le32(num_entries);
2188         req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);
2189         return 0;
2190 }
2191
2192 /*
2193  * Create an mds request.
2194  */
2195 struct ceph_mds_request *
2196 ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode)
2197 {
2198         struct ceph_mds_request *req;
2199
2200         req = kmem_cache_zalloc(ceph_mds_request_cachep, GFP_NOFS);
2201         if (!req)
2202                 return ERR_PTR(-ENOMEM);
2203
2204         mutex_init(&req->r_fill_mutex);
2205         req->r_mdsc = mdsc;
2206         req->r_started = jiffies;
2207         req->r_start_latency = ktime_get();
2208         req->r_resend_mds = -1;
2209         INIT_LIST_HEAD(&req->r_unsafe_dir_item);
2210         INIT_LIST_HEAD(&req->r_unsafe_target_item);
2211         req->r_fmode = -1;
2212         kref_init(&req->r_kref);
2213         RB_CLEAR_NODE(&req->r_node);
2214         INIT_LIST_HEAD(&req->r_wait);
2215         init_completion(&req->r_completion);
2216         init_completion(&req->r_safe_completion);
2217         INIT_LIST_HEAD(&req->r_unsafe_item);
2218
2219         ktime_get_coarse_real_ts64(&req->r_stamp);
2220
2221         req->r_op = op;
2222         req->r_direct_mode = mode;
2223         return req;
2224 }
2225
2226 /*
2227  * return oldest (lowest) request, tid in request tree, 0 if none.
2228  *
2229  * called under mdsc->mutex.
2230  */
2231 static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc)
2232 {
2233         if (RB_EMPTY_ROOT(&mdsc->request_tree))
2234                 return NULL;
2235         return rb_entry(rb_first(&mdsc->request_tree),
2236                         struct ceph_mds_request, r_node);
2237 }
2238
2239 static inline  u64 __get_oldest_tid(struct ceph_mds_client *mdsc)
2240 {
2241         return mdsc->oldest_tid;
2242 }
2243
2244 /*
2245  * Build a dentry's path.  Allocate on heap; caller must kfree.  Based
2246  * on build_path_from_dentry in fs/cifs/dir.c.
2247  *
2248  * If @stop_on_nosnap, generate path relative to the first non-snapped
2249  * inode.
2250  *
2251  * Encode hidden .snap dirs as a double /, i.e.
2252  *   foo/.snap/bar -> foo//bar
2253  */
2254 char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *pbase,
2255                            int stop_on_nosnap)
2256 {
2257         struct dentry *temp;
2258         char *path;
2259         int pos;
2260         unsigned seq;
2261         u64 base;
2262
2263         if (!dentry)
2264                 return ERR_PTR(-EINVAL);
2265
2266         path = __getname();
2267         if (!path)
2268                 return ERR_PTR(-ENOMEM);
2269 retry:
2270         pos = PATH_MAX - 1;
2271         path[pos] = '\0';
2272
2273         seq = read_seqbegin(&rename_lock);
2274         rcu_read_lock();
2275         temp = dentry;
2276         for (;;) {
2277                 struct inode *inode;
2278
2279                 spin_lock(&temp->d_lock);
2280                 inode = d_inode(temp);
2281                 if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
2282                         dout("build_path path+%d: %p SNAPDIR\n",
2283                              pos, temp);
2284                 } else if (stop_on_nosnap && inode && dentry != temp &&
2285                            ceph_snap(inode) == CEPH_NOSNAP) {
2286                         spin_unlock(&temp->d_lock);
2287                         pos++; /* get rid of any prepended '/' */
2288                         break;
2289                 } else {
2290                         pos -= temp->d_name.len;
2291                         if (pos < 0) {
2292                                 spin_unlock(&temp->d_lock);
2293                                 break;
2294                         }
2295                         memcpy(path + pos, temp->d_name.name, temp->d_name.len);
2296                 }
2297                 spin_unlock(&temp->d_lock);
2298                 temp = READ_ONCE(temp->d_parent);
2299
2300                 /* Are we at the root? */
2301                 if (IS_ROOT(temp))
2302                         break;
2303
2304                 /* Are we out of buffer? */
2305                 if (--pos < 0)
2306                         break;
2307
2308                 path[pos] = '/';
2309         }
2310         base = ceph_ino(d_inode(temp));
2311         rcu_read_unlock();
2312
2313         if (read_seqretry(&rename_lock, seq))
2314                 goto retry;
2315
2316         if (pos < 0) {
2317                 /*
2318                  * A rename didn't occur, but somehow we didn't end up where
2319                  * we thought we would. Throw a warning and try again.
2320                  */
2321                 pr_warn("build_path did not end path lookup where "
2322                         "expected, pos is %d\n", pos);
2323                 goto retry;
2324         }
2325
2326         *pbase = base;
2327         *plen = PATH_MAX - 1 - pos;
2328         dout("build_path on %p %d built %llx '%.*s'\n",
2329              dentry, d_count(dentry), base, *plen, path + pos);
2330         return path + pos;
2331 }
2332
2333 static int build_dentry_path(struct dentry *dentry, struct inode *dir,
2334                              const char **ppath, int *ppathlen, u64 *pino,
2335                              bool *pfreepath, bool parent_locked)
2336 {
2337         char *path;
2338
2339         rcu_read_lock();
2340         if (!dir)
2341                 dir = d_inode_rcu(dentry->d_parent);
2342         if (dir && parent_locked && ceph_snap(dir) == CEPH_NOSNAP) {
2343                 *pino = ceph_ino(dir);
2344                 rcu_read_unlock();
2345                 *ppath = dentry->d_name.name;
2346                 *ppathlen = dentry->d_name.len;
2347                 return 0;
2348         }
2349         rcu_read_unlock();
2350         path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2351         if (IS_ERR(path))
2352                 return PTR_ERR(path);
2353         *ppath = path;
2354         *pfreepath = true;
2355         return 0;
2356 }
2357
2358 static int build_inode_path(struct inode *inode,
2359                             const char **ppath, int *ppathlen, u64 *pino,
2360                             bool *pfreepath)
2361 {
2362         struct dentry *dentry;
2363         char *path;
2364
2365         if (ceph_snap(inode) == CEPH_NOSNAP) {
2366                 *pino = ceph_ino(inode);
2367                 *ppathlen = 0;
2368                 return 0;
2369         }
2370         dentry = d_find_alias(inode);
2371         path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2372         dput(dentry);
2373         if (IS_ERR(path))
2374                 return PTR_ERR(path);
2375         *ppath = path;
2376         *pfreepath = true;
2377         return 0;
2378 }
2379
2380 /*
2381  * request arguments may be specified via an inode *, a dentry *, or
2382  * an explicit ino+path.
2383  */
2384 static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
2385                                   struct inode *rdiri, const char *rpath,
2386                                   u64 rino, const char **ppath, int *pathlen,
2387                                   u64 *ino, bool *freepath, bool parent_locked)
2388 {
2389         int r = 0;
2390
2391         if (rinode) {
2392                 r = build_inode_path(rinode, ppath, pathlen, ino, freepath);
2393                 dout(" inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
2394                      ceph_snap(rinode));
2395         } else if (rdentry) {
2396                 r = build_dentry_path(rdentry, rdiri, ppath, pathlen, ino,
2397                                         freepath, parent_locked);
2398                 dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
2399                      *ppath);
2400         } else if (rpath || rino) {
2401                 *ino = rino;
2402                 *ppath = rpath;
2403                 *pathlen = rpath ? strlen(rpath) : 0;
2404                 dout(" path %.*s\n", *pathlen, rpath);
2405         }
2406
2407         return r;
2408 }
2409
2410 /*
2411  * called under mdsc->mutex
2412  */
2413 static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
2414                                                struct ceph_mds_request *req,
2415                                                int mds, bool drop_cap_releases)
2416 {
2417         struct ceph_msg *msg;
2418         struct ceph_mds_request_head *head;
2419         const char *path1 = NULL;
2420         const char *path2 = NULL;
2421         u64 ino1 = 0, ino2 = 0;
2422         int pathlen1 = 0, pathlen2 = 0;
2423         bool freepath1 = false, freepath2 = false;
2424         int len;
2425         u16 releases;
2426         void *p, *end;
2427         int ret;
2428
2429         ret = set_request_path_attr(req->r_inode, req->r_dentry,
2430                               req->r_parent, req->r_path1, req->r_ino1.ino,
2431                               &path1, &pathlen1, &ino1, &freepath1,
2432                               test_bit(CEPH_MDS_R_PARENT_LOCKED,
2433                                         &req->r_req_flags));
2434         if (ret < 0) {
2435                 msg = ERR_PTR(ret);
2436                 goto out;
2437         }
2438
2439         /* If r_old_dentry is set, then assume that its parent is locked */
2440         ret = set_request_path_attr(NULL, req->r_old_dentry,
2441                               req->r_old_dentry_dir,
2442                               req->r_path2, req->r_ino2.ino,
2443                               &path2, &pathlen2, &ino2, &freepath2, true);
2444         if (ret < 0) {
2445                 msg = ERR_PTR(ret);
2446                 goto out_free1;
2447         }
2448
2449         len = sizeof(*head) +
2450                 pathlen1 + pathlen2 + 2*(1 + sizeof(u32) + sizeof(u64)) +
2451                 sizeof(struct ceph_timespec);
2452
2453         /* calculate (max) length for cap releases */
2454         len += sizeof(struct ceph_mds_request_release) *
2455                 (!!req->r_inode_drop + !!req->r_dentry_drop +
2456                  !!req->r_old_inode_drop + !!req->r_old_dentry_drop);
2457         if (req->r_dentry_drop)
2458                 len += pathlen1;
2459         if (req->r_old_dentry_drop)
2460                 len += pathlen2;
2461
2462         msg = ceph_msg_new2(CEPH_MSG_CLIENT_REQUEST, len, 1, GFP_NOFS, false);
2463         if (!msg) {
2464                 msg = ERR_PTR(-ENOMEM);
2465                 goto out_free2;
2466         }
2467
2468         msg->hdr.version = cpu_to_le16(2);
2469         msg->hdr.tid = cpu_to_le64(req->r_tid);
2470
2471         head = msg->front.iov_base;
2472         p = msg->front.iov_base + sizeof(*head);
2473         end = msg->front.iov_base + msg->front.iov_len;
2474
2475         head->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);
2476         head->op = cpu_to_le32(req->r_op);
2477         head->caller_uid = cpu_to_le32(from_kuid(&init_user_ns, req->r_uid));
2478         head->caller_gid = cpu_to_le32(from_kgid(&init_user_ns, req->r_gid));
2479         head->ino = cpu_to_le64(req->r_deleg_ino);
2480         head->args = req->r_args;
2481
2482         ceph_encode_filepath(&p, end, ino1, path1);
2483         ceph_encode_filepath(&p, end, ino2, path2);
2484
2485         /* make note of release offset, in case we need to replay */
2486         req->r_request_release_offset = p - msg->front.iov_base;
2487
2488         /* cap releases */
2489         releases = 0;
2490         if (req->r_inode_drop)
2491                 releases += ceph_encode_inode_release(&p,
2492                       req->r_inode ? req->r_inode : d_inode(req->r_dentry),
2493                       mds, req->r_inode_drop, req->r_inode_unless,
2494                       req->r_op == CEPH_MDS_OP_READDIR);
2495         if (req->r_dentry_drop)
2496                 releases += ceph_encode_dentry_release(&p, req->r_dentry,
2497                                 req->r_parent, mds, req->r_dentry_drop,
2498                                 req->r_dentry_unless);
2499         if (req->r_old_dentry_drop)
2500                 releases += ceph_encode_dentry_release(&p, req->r_old_dentry,
2501                                 req->r_old_dentry_dir, mds,
2502                                 req->r_old_dentry_drop,
2503                                 req->r_old_dentry_unless);
2504         if (req->r_old_inode_drop)
2505                 releases += ceph_encode_inode_release(&p,
2506                       d_inode(req->r_old_dentry),
2507                       mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
2508
2509         if (drop_cap_releases) {
2510                 releases = 0;
2511                 p = msg->front.iov_base + req->r_request_release_offset;
2512         }
2513
2514         head->num_releases = cpu_to_le16(releases);
2515
2516         /* time stamp */
2517         {
2518                 struct ceph_timespec ts;
2519                 ceph_encode_timespec64(&ts, &req->r_stamp);
2520                 ceph_encode_copy(&p, &ts, sizeof(ts));
2521         }
2522
2523         BUG_ON(p > end);
2524         msg->front.iov_len = p - msg->front.iov_base;
2525         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2526
2527         if (req->r_pagelist) {
2528                 struct ceph_pagelist *pagelist = req->r_pagelist;
2529                 ceph_msg_data_add_pagelist(msg, pagelist);
2530                 msg->hdr.data_len = cpu_to_le32(pagelist->length);
2531         } else {
2532                 msg->hdr.data_len = 0;
2533         }
2534
2535         msg->hdr.data_off = cpu_to_le16(0);
2536
2537 out_free2:
2538         if (freepath2)
2539                 ceph_mdsc_free_path((char *)path2, pathlen2);
2540 out_free1:
2541         if (freepath1)
2542                 ceph_mdsc_free_path((char *)path1, pathlen1);
2543 out:
2544         return msg;
2545 }
2546
2547 /*
2548  * called under mdsc->mutex if error, under no mutex if
2549  * success.
2550  */
2551 static void complete_request(struct ceph_mds_client *mdsc,
2552                              struct ceph_mds_request *req)
2553 {
2554         req->r_end_latency = ktime_get();
2555
2556         if (req->r_callback)
2557                 req->r_callback(mdsc, req);
2558         complete_all(&req->r_completion);
2559 }
2560
2561 /*
2562  * called under mdsc->mutex
2563  */
2564 static int __prepare_send_request(struct ceph_mds_client *mdsc,
2565                                   struct ceph_mds_request *req,
2566                                   int mds, bool drop_cap_releases)
2567 {
2568         struct ceph_mds_request_head *rhead;
2569         struct ceph_msg *msg;
2570         int flags = 0;
2571
2572         req->r_attempts++;
2573         if (req->r_inode) {
2574                 struct ceph_cap *cap =
2575                         ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);
2576
2577                 if (cap)
2578                         req->r_sent_on_mseq = cap->mseq;
2579                 else
2580                         req->r_sent_on_mseq = -1;
2581         }
2582         dout("prepare_send_request %p tid %lld %s (attempt %d)\n", req,
2583              req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);
2584
2585         if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
2586                 void *p;
2587                 /*
2588                  * Replay.  Do not regenerate message (and rebuild
2589                  * paths, etc.); just use the original message.
2590                  * Rebuilding paths will break for renames because
2591                  * d_move mangles the src name.
2592                  */
2593                 msg = req->r_request;
2594                 rhead = msg->front.iov_base;
2595
2596                 flags = le32_to_cpu(rhead->flags);
2597                 flags |= CEPH_MDS_FLAG_REPLAY;
2598                 rhead->flags = cpu_to_le32(flags);
2599
2600                 if (req->r_target_inode)
2601                         rhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));
2602
2603                 rhead->num_retry = req->r_attempts - 1;
2604
2605                 /* remove cap/dentry releases from message */
2606                 rhead->num_releases = 0;
2607
2608                 /* time stamp */
2609                 p = msg->front.iov_base + req->r_request_release_offset;
2610                 {
2611                         struct ceph_timespec ts;
2612                         ceph_encode_timespec64(&ts, &req->r_stamp);
2613                         ceph_encode_copy(&p, &ts, sizeof(ts));
2614                 }
2615
2616                 msg->front.iov_len = p - msg->front.iov_base;
2617                 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2618                 return 0;
2619         }
2620
2621         if (req->r_request) {
2622                 ceph_msg_put(req->r_request);
2623                 req->r_request = NULL;
2624         }
2625         msg = create_request_message(mdsc, req, mds, drop_cap_releases);
2626         if (IS_ERR(msg)) {
2627                 req->r_err = PTR_ERR(msg);
2628                 return PTR_ERR(msg);
2629         }
2630         req->r_request = msg;
2631
2632         rhead = msg->front.iov_base;
2633         rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
2634         if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2635                 flags |= CEPH_MDS_FLAG_REPLAY;
2636         if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags))
2637                 flags |= CEPH_MDS_FLAG_ASYNC;
2638         if (req->r_parent)
2639                 flags |= CEPH_MDS_FLAG_WANT_DENTRY;
2640         rhead->flags = cpu_to_le32(flags);
2641         rhead->num_fwd = req->r_num_fwd;
2642         rhead->num_retry = req->r_attempts - 1;
2643
2644         dout(" r_parent = %p\n", req->r_parent);
2645         return 0;
2646 }
2647
2648 /*
2649  * called under mdsc->mutex
2650  */
2651 static int __send_request(struct ceph_mds_client *mdsc,
2652                           struct ceph_mds_session *session,
2653                           struct ceph_mds_request *req,
2654                           bool drop_cap_releases)
2655 {
2656         int err;
2657
2658         err = __prepare_send_request(mdsc, req, session->s_mds,
2659                                      drop_cap_releases);
2660         if (!err) {
2661                 ceph_msg_get(req->r_request);
2662                 ceph_con_send(&session->s_con, req->r_request);
2663         }
2664
2665         return err;
2666 }
2667
2668 /*
2669  * send request, or put it on the appropriate wait list.
2670  */
2671 static void __do_request(struct ceph_mds_client *mdsc,
2672                         struct ceph_mds_request *req)
2673 {
2674         struct ceph_mds_session *session = NULL;
2675         int mds = -1;
2676         int err = 0;
2677         bool random;
2678
2679         if (req->r_err || test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
2680                 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))
2681                         __unregister_request(mdsc, req);
2682                 return;
2683         }
2684
2685         if (req->r_timeout &&
2686             time_after_eq(jiffies, req->r_started + req->r_timeout)) {
2687                 dout("do_request timed out\n");
2688                 err = -ETIMEDOUT;
2689                 goto finish;
2690         }
2691         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
2692                 dout("do_request forced umount\n");
2693                 err = -EIO;
2694                 goto finish;
2695         }
2696         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {
2697                 if (mdsc->mdsmap_err) {
2698                         err = mdsc->mdsmap_err;
2699                         dout("do_request mdsmap err %d\n", err);
2700                         goto finish;
2701                 }
2702                 if (mdsc->mdsmap->m_epoch == 0) {
2703                         dout("do_request no mdsmap, waiting for map\n");
2704                         list_add(&req->r_wait, &mdsc->waiting_for_map);
2705                         return;
2706                 }
2707                 if (!(mdsc->fsc->mount_options->flags &
2708                       CEPH_MOUNT_OPT_MOUNTWAIT) &&
2709                     !ceph_mdsmap_is_cluster_available(mdsc->mdsmap)) {
2710                         err = -EHOSTUNREACH;
2711                         goto finish;
2712                 }
2713         }
2714
2715         put_request_session(req);
2716
2717         mds = __choose_mds(mdsc, req, &random);
2718         if (mds < 0 ||
2719             ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {
2720                 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2721                         err = -EJUKEBOX;
2722                         goto finish;
2723                 }
2724                 dout("do_request no mds or not active, waiting for map\n");
2725                 list_add(&req->r_wait, &mdsc->waiting_for_map);
2726                 return;
2727         }
2728
2729         /* get, open session */
2730         session = __ceph_lookup_mds_session(mdsc, mds);
2731         if (!session) {
2732                 session = register_session(mdsc, mds);
2733                 if (IS_ERR(session)) {
2734                         err = PTR_ERR(session);
2735                         goto finish;
2736                 }
2737         }
2738         req->r_session = ceph_get_mds_session(session);
2739
2740         dout("do_request mds%d session %p state %s\n", mds, session,
2741              ceph_session_state_name(session->s_state));
2742         if (session->s_state != CEPH_MDS_SESSION_OPEN &&
2743             session->s_state != CEPH_MDS_SESSION_HUNG) {
2744                 if (session->s_state == CEPH_MDS_SESSION_REJECTED) {
2745                         err = -EACCES;
2746                         goto out_session;
2747                 }
2748                 /*
2749                  * We cannot queue async requests since the caps and delegated
2750                  * inodes are bound to the session. Just return -EJUKEBOX and
2751                  * let the caller retry a sync request in that case.
2752                  */
2753                 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2754                         err = -EJUKEBOX;
2755                         goto out_session;
2756                 }
2757                 if (session->s_state == CEPH_MDS_SESSION_NEW ||
2758                     session->s_state == CEPH_MDS_SESSION_CLOSING) {
2759                         __open_session(mdsc, session);
2760                         /* retry the same mds later */
2761                         if (random)
2762                                 req->r_resend_mds = mds;
2763                 }
2764                 list_add(&req->r_wait, &session->s_waiting);
2765                 goto out_session;
2766         }
2767
2768         /* send request */
2769         req->r_resend_mds = -1;   /* forget any previous mds hint */
2770
2771         if (req->r_request_started == 0)   /* note request start time */
2772                 req->r_request_started = jiffies;
2773
2774         err = __send_request(mdsc, session, req, false);
2775
2776 out_session:
2777         ceph_put_mds_session(session);
2778 finish:
2779         if (err) {
2780                 dout("__do_request early error %d\n", err);
2781                 req->r_err = err;
2782                 complete_request(mdsc, req);
2783                 __unregister_request(mdsc, req);
2784         }
2785         return;
2786 }
2787
2788 /*
2789  * called under mdsc->mutex
2790  */
2791 static void __wake_requests(struct ceph_mds_client *mdsc,
2792                             struct list_head *head)
2793 {
2794         struct ceph_mds_request *req;
2795         LIST_HEAD(tmp_list);
2796
2797         list_splice_init(head, &tmp_list);
2798
2799         while (!list_empty(&tmp_list)) {
2800                 req = list_entry(tmp_list.next,
2801                                  struct ceph_mds_request, r_wait);
2802                 list_del_init(&req->r_wait);
2803                 dout(" wake request %p tid %llu\n", req, req->r_tid);
2804                 __do_request(mdsc, req);
2805         }
2806 }
2807
2808 /*
2809  * Wake up threads with requests pending for @mds, so that they can
2810  * resubmit their requests to a possibly different mds.
2811  */
2812 static void kick_requests(struct ceph_mds_client *mdsc, int mds)
2813 {
2814         struct ceph_mds_request *req;
2815         struct rb_node *p = rb_first(&mdsc->request_tree);
2816
2817         dout("kick_requests mds%d\n", mds);
2818         while (p) {
2819                 req = rb_entry(p, struct ceph_mds_request, r_node);
2820                 p = rb_next(p);
2821                 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2822                         continue;
2823                 if (req->r_attempts > 0)
2824                         continue; /* only new requests */
2825                 if (req->r_session &&
2826                     req->r_session->s_mds == mds) {
2827                         dout(" kicking tid %llu\n", req->r_tid);
2828                         list_del_init(&req->r_wait);
2829                         __do_request(mdsc, req);
2830                 }
2831         }
2832 }
2833
2834 int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, struct inode *dir,
2835                               struct ceph_mds_request *req)
2836 {
2837         int err = 0;
2838
2839         /* take CAP_PIN refs for r_inode, r_parent, r_old_dentry */
2840         if (req->r_inode)
2841                 ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
2842         if (req->r_parent) {
2843                 struct ceph_inode_info *ci = ceph_inode(req->r_parent);
2844                 int fmode = (req->r_op & CEPH_MDS_OP_WRITE) ?
2845                             CEPH_FILE_MODE_WR : CEPH_FILE_MODE_RD;
2846                 spin_lock(&ci->i_ceph_lock);
2847                 ceph_take_cap_refs(ci, CEPH_CAP_PIN, false);
2848                 __ceph_touch_fmode(ci, mdsc, fmode);
2849                 spin_unlock(&ci->i_ceph_lock);
2850                 ihold(req->r_parent);
2851         }
2852         if (req->r_old_dentry_dir)
2853                 ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
2854                                   CEPH_CAP_PIN);
2855
2856         if (req->r_inode) {
2857                 err = ceph_wait_on_async_create(req->r_inode);
2858                 if (err) {
2859                         dout("%s: wait for async create returned: %d\n",
2860                              __func__, err);
2861                         return err;
2862                 }
2863         }
2864
2865         if (!err && req->r_old_inode) {
2866                 err = ceph_wait_on_async_create(req->r_old_inode);
2867                 if (err) {
2868                         dout("%s: wait for async create returned: %d\n",
2869                              __func__, err);
2870                         return err;
2871                 }
2872         }
2873
2874         dout("submit_request on %p for inode %p\n", req, dir);
2875         mutex_lock(&mdsc->mutex);
2876         __register_request(mdsc, req, dir);
2877         __do_request(mdsc, req);
2878         err = req->r_err;
2879         mutex_unlock(&mdsc->mutex);
2880         return err;
2881 }
2882
2883 static int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc,
2884                                   struct ceph_mds_request *req)
2885 {
2886         int err;
2887
2888         /* wait */
2889         dout("do_request waiting\n");
2890         if (!req->r_timeout && req->r_wait_for_completion) {
2891                 err = req->r_wait_for_completion(mdsc, req);
2892         } else {
2893                 long timeleft = wait_for_completion_killable_timeout(
2894                                         &req->r_completion,
2895                                         ceph_timeout_jiffies(req->r_timeout));
2896                 if (timeleft > 0)
2897                         err = 0;
2898                 else if (!timeleft)
2899                         err = -ETIMEDOUT;  /* timed out */
2900                 else
2901                         err = timeleft;  /* killed */
2902         }
2903         dout("do_request waited, got %d\n", err);
2904         mutex_lock(&mdsc->mutex);
2905
2906         /* only abort if we didn't race with a real reply */
2907         if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
2908                 err = le32_to_cpu(req->r_reply_info.head->result);
2909         } else if (err < 0) {
2910                 dout("aborted request %lld with %d\n", req->r_tid, err);
2911
2912                 /*
2913                  * ensure we aren't running concurrently with
2914                  * ceph_fill_trace or ceph_readdir_prepopulate, which
2915                  * rely on locks (dir mutex) held by our caller.
2916                  */
2917                 mutex_lock(&req->r_fill_mutex);
2918                 req->r_err = err;
2919                 set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
2920                 mutex_unlock(&req->r_fill_mutex);
2921
2922                 if (req->r_parent &&
2923                     (req->r_op & CEPH_MDS_OP_WRITE))
2924                         ceph_invalidate_dir_request(req);
2925         } else {
2926                 err = req->r_err;
2927         }
2928
2929         mutex_unlock(&mdsc->mutex);
2930         return err;
2931 }
2932
2933 /*
2934  * Synchrously perform an mds request.  Take care of all of the
2935  * session setup, forwarding, retry details.
2936  */
2937 int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
2938                          struct inode *dir,
2939                          struct ceph_mds_request *req)
2940 {
2941         int err;
2942
2943         dout("do_request on %p\n", req);
2944
2945         /* issue */
2946         err = ceph_mdsc_submit_request(mdsc, dir, req);
2947         if (!err)
2948                 err = ceph_mdsc_wait_request(mdsc, req);
2949         dout("do_request %p done, result %d\n", req, err);
2950         return err;
2951 }
2952
2953 /*
2954  * Invalidate dir's completeness, dentry lease state on an aborted MDS
2955  * namespace request.
2956  */
2957 void ceph_invalidate_dir_request(struct ceph_mds_request *req)
2958 {
2959         struct inode *dir = req->r_parent;
2960         struct inode *old_dir = req->r_old_dentry_dir;
2961
2962         dout("invalidate_dir_request %p %p (complete, lease(s))\n", dir, old_dir);
2963
2964         ceph_dir_clear_complete(dir);
2965         if (old_dir)
2966                 ceph_dir_clear_complete(old_dir);
2967         if (req->r_dentry)
2968                 ceph_invalidate_dentry_lease(req->r_dentry);
2969         if (req->r_old_dentry)
2970                 ceph_invalidate_dentry_lease(req->r_old_dentry);
2971 }
2972
2973 /*
2974  * Handle mds reply.
2975  *
2976  * We take the session mutex and parse and process the reply immediately.
2977  * This preserves the logical ordering of replies, capabilities, etc., sent
2978  * by the MDS as they are applied to our local cache.
2979  */
2980 static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
2981 {
2982         struct ceph_mds_client *mdsc = session->s_mdsc;
2983         struct ceph_mds_request *req;
2984         struct ceph_mds_reply_head *head = msg->front.iov_base;
2985         struct ceph_mds_reply_info_parsed *rinfo;  /* parsed reply info */
2986         struct ceph_snap_realm *realm;
2987         u64 tid;
2988         int err, result;
2989         int mds = session->s_mds;
2990
2991         if (msg->front.iov_len < sizeof(*head)) {
2992                 pr_err("mdsc_handle_reply got corrupt (short) reply\n");
2993                 ceph_msg_dump(msg);
2994                 return;
2995         }
2996
2997         /* get request, session */
2998         tid = le64_to_cpu(msg->hdr.tid);
2999         mutex_lock(&mdsc->mutex);
3000         req = lookup_get_request(mdsc, tid);
3001         if (!req) {
3002                 dout("handle_reply on unknown tid %llu\n", tid);
3003                 mutex_unlock(&mdsc->mutex);
3004                 return;
3005         }
3006         dout("handle_reply %p\n", req);
3007
3008         /* correct session? */
3009         if (req->r_session != session) {
3010                 pr_err("mdsc_handle_reply got %llu on session mds%d"
3011                        " not mds%d\n", tid, session->s_mds,
3012                        req->r_session ? req->r_session->s_mds : -1);
3013                 mutex_unlock(&mdsc->mutex);
3014                 goto out;
3015         }
3016
3017         /* dup? */
3018         if ((test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags) && !head->safe) ||
3019             (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags) && head->safe)) {
3020                 pr_warn("got a dup %s reply on %llu from mds%d\n",
3021                            head->safe ? "safe" : "unsafe", tid, mds);
3022                 mutex_unlock(&mdsc->mutex);
3023                 goto out;
3024         }
3025         if (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags)) {
3026                 pr_warn("got unsafe after safe on %llu from mds%d\n",
3027                            tid, mds);
3028                 mutex_unlock(&mdsc->mutex);
3029                 goto out;
3030         }
3031
3032         result = le32_to_cpu(head->result);
3033
3034         /*
3035          * Handle an ESTALE
3036          * if we're not talking to the authority, send to them
3037          * if the authority has changed while we weren't looking,
3038          * send to new authority
3039          * Otherwise we just have to return an ESTALE
3040          */
3041         if (result == -ESTALE) {
3042                 dout("got ESTALE on request %llu\n", req->r_tid);
3043                 req->r_resend_mds = -1;
3044                 if (req->r_direct_mode != USE_AUTH_MDS) {
3045                         dout("not using auth, setting for that now\n");
3046                         req->r_direct_mode = USE_AUTH_MDS;
3047                         __do_request(mdsc, req);
3048                         mutex_unlock(&mdsc->mutex);
3049                         goto out;
3050                 } else  {
3051                         int mds = __choose_mds(mdsc, req, NULL);
3052                         if (mds >= 0 && mds != req->r_session->s_mds) {
3053                                 dout("but auth changed, so resending\n");
3054                                 __do_request(mdsc, req);
3055                                 mutex_unlock(&mdsc->mutex);
3056                                 goto out;
3057                         }
3058                 }
3059                 dout("have to return ESTALE on request %llu\n", req->r_tid);
3060         }
3061
3062
3063         if (head->safe) {
3064                 set_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags);
3065                 __unregister_request(mdsc, req);
3066
3067                 /* last request during umount? */
3068                 if (mdsc->stopping && !__get_oldest_req(mdsc))
3069                         complete_all(&mdsc->safe_umount_waiters);
3070
3071                 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3072                         /*
3073                          * We already handled the unsafe response, now do the
3074                          * cleanup.  No need to examine the response; the MDS
3075                          * doesn't include any result info in the safe
3076                          * response.  And even if it did, there is nothing
3077                          * useful we could do with a revised return value.
3078                          */
3079                         dout("got safe reply %llu, mds%d\n", tid, mds);
3080
3081                         mutex_unlock(&mdsc->mutex);
3082                         goto out;
3083                 }
3084         } else {
3085                 set_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags);
3086                 list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
3087         }
3088
3089         dout("handle_reply tid %lld result %d\n", tid, result);
3090         rinfo = &req->r_reply_info;
3091         if (test_bit(CEPHFS_FEATURE_REPLY_ENCODING, &session->s_features))
3092                 err = parse_reply_info(session, msg, rinfo, (u64)-1);
3093         else
3094                 err = parse_reply_info(session, msg, rinfo, session->s_con.peer_features);
3095         mutex_unlock(&mdsc->mutex);
3096
3097         mutex_lock(&session->s_mutex);
3098         if (err < 0) {
3099                 pr_err("mdsc_handle_reply got corrupt reply mds%d(tid:%lld)\n", mds, tid);
3100                 ceph_msg_dump(msg);
3101                 goto out_err;
3102         }
3103
3104         /* snap trace */
3105         realm = NULL;
3106         if (rinfo->snapblob_len) {
3107                 down_write(&mdsc->snap_rwsem);
3108                 ceph_update_snap_trace(mdsc, rinfo->snapblob,
3109                                 rinfo->snapblob + rinfo->snapblob_len,
3110                                 le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
3111                                 &realm);
3112                 downgrade_write(&mdsc->snap_rwsem);
3113         } else {
3114                 down_read(&mdsc->snap_rwsem);
3115         }
3116
3117         /* insert trace into our cache */
3118         mutex_lock(&req->r_fill_mutex);
3119         current->journal_info = req;
3120         err = ceph_fill_trace(mdsc->fsc->sb, req);
3121         if (err == 0) {
3122                 if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
3123                                     req->r_op == CEPH_MDS_OP_LSSNAP))
3124                         ceph_readdir_prepopulate(req, req->r_session);
3125         }
3126         current->journal_info = NULL;
3127         mutex_unlock(&req->r_fill_mutex);
3128
3129         up_read(&mdsc->snap_rwsem);
3130         if (realm)
3131                 ceph_put_snap_realm(mdsc, realm);
3132
3133         if (err == 0) {
3134                 if (req->r_target_inode &&
3135                     test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3136                         struct ceph_inode_info *ci =
3137                                 ceph_inode(req->r_target_inode);
3138                         spin_lock(&ci->i_unsafe_lock);
3139                         list_add_tail(&req->r_unsafe_target_item,
3140                                       &ci->i_unsafe_iops);
3141                         spin_unlock(&ci->i_unsafe_lock);
3142                 }
3143
3144                 ceph_unreserve_caps(mdsc, &req->r_caps_reservation);
3145         }
3146 out_err:
3147         mutex_lock(&mdsc->mutex);
3148         if (!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3149                 if (err) {
3150                         req->r_err = err;
3151                 } else {
3152                         req->r_reply =  ceph_msg_get(msg);
3153                         set_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags);
3154                 }
3155         } else {
3156                 dout("reply arrived after request %lld was aborted\n", tid);
3157         }
3158         mutex_unlock(&mdsc->mutex);
3159
3160         mutex_unlock(&session->s_mutex);
3161
3162         /* kick calling process */
3163         complete_request(mdsc, req);
3164
3165         ceph_update_metadata_latency(&mdsc->metric, req->r_start_latency,
3166                                      req->r_end_latency, err);
3167 out:
3168         ceph_mdsc_put_request(req);
3169         return;
3170 }
3171
3172
3173
3174 /*
3175  * handle mds notification that our request has been forwarded.
3176  */
3177 static void handle_forward(struct ceph_mds_client *mdsc,
3178                            struct ceph_mds_session *session,
3179                            struct ceph_msg *msg)
3180 {
3181         struct ceph_mds_request *req;
3182         u64 tid = le64_to_cpu(msg->hdr.tid);
3183         u32 next_mds;
3184         u32 fwd_seq;
3185         int err = -EINVAL;
3186         void *p = msg->front.iov_base;
3187         void *end = p + msg->front.iov_len;
3188
3189         ceph_decode_need(&p, end, 2*sizeof(u32), bad);
3190         next_mds = ceph_decode_32(&p);
3191         fwd_seq = ceph_decode_32(&p);
3192
3193         mutex_lock(&mdsc->mutex);
3194         req = lookup_get_request(mdsc, tid);
3195         if (!req) {
3196                 dout("forward tid %llu to mds%d - req dne\n", tid, next_mds);
3197                 goto out;  /* dup reply? */
3198         }
3199
3200         if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3201                 dout("forward tid %llu aborted, unregistering\n", tid);
3202                 __unregister_request(mdsc, req);
3203         } else if (fwd_seq <= req->r_num_fwd) {
3204                 dout("forward tid %llu to mds%d - old seq %d <= %d\n",
3205                      tid, next_mds, req->r_num_fwd, fwd_seq);
3206         } else {
3207                 /* resend. forward race not possible; mds would drop */
3208                 dout("forward tid %llu to mds%d (we resend)\n", tid, next_mds);
3209                 BUG_ON(req->r_err);
3210                 BUG_ON(test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags));
3211                 req->r_attempts = 0;
3212                 req->r_num_fwd = fwd_seq;
3213                 req->r_resend_mds = next_mds;
3214                 put_request_session(req);
3215                 __do_request(mdsc, req);
3216         }
3217         ceph_mdsc_put_request(req);
3218 out:
3219         mutex_unlock(&mdsc->mutex);
3220         return;
3221
3222 bad:
3223         pr_err("mdsc_handle_forward decode error err=%d\n", err);
3224 }
3225
3226 static int __decode_session_metadata(void **p, void *end,
3227                                      bool *blacklisted)
3228 {
3229         /* map<string,string> */
3230         u32 n;
3231         bool err_str;
3232         ceph_decode_32_safe(p, end, n, bad);
3233         while (n-- > 0) {
3234                 u32 len;
3235                 ceph_decode_32_safe(p, end, len, bad);
3236                 ceph_decode_need(p, end, len, bad);
3237                 err_str = !strncmp(*p, "error_string", len);
3238                 *p += len;
3239                 ceph_decode_32_safe(p, end, len, bad);
3240                 ceph_decode_need(p, end, len, bad);
3241                 if (err_str && strnstr(*p, "blacklisted", len))
3242                         *blacklisted = true;
3243                 *p += len;
3244         }
3245         return 0;
3246 bad:
3247         return -1;
3248 }
3249
3250 /*
3251  * handle a mds session control message
3252  */
3253 static void handle_session(struct ceph_mds_session *session,
3254                            struct ceph_msg *msg)
3255 {
3256         struct ceph_mds_client *mdsc = session->s_mdsc;
3257         int mds = session->s_mds;
3258         int msg_version = le16_to_cpu(msg->hdr.version);
3259         void *p = msg->front.iov_base;
3260         void *end = p + msg->front.iov_len;
3261         struct ceph_mds_session_head *h;
3262         u32 op;
3263         u64 seq, features = 0;
3264         int wake = 0;
3265         bool blacklisted = false;
3266
3267         /* decode */
3268         ceph_decode_need(&p, end, sizeof(*h), bad);
3269         h = p;
3270         p += sizeof(*h);
3271
3272         op = le32_to_cpu(h->op);
3273         seq = le64_to_cpu(h->seq);
3274
3275         if (msg_version >= 3) {
3276                 u32 len;
3277                 /* version >= 2, metadata */
3278                 if (__decode_session_metadata(&p, end, &blacklisted) < 0)
3279                         goto bad;
3280                 /* version >= 3, feature bits */
3281                 ceph_decode_32_safe(&p, end, len, bad);
3282                 ceph_decode_64_safe(&p, end, features, bad);
3283                 p += len - sizeof(features);
3284         }
3285
3286         mutex_lock(&mdsc->mutex);
3287         if (op == CEPH_SESSION_CLOSE) {
3288                 ceph_get_mds_session(session);
3289                 __unregister_session(mdsc, session);
3290         }
3291         /* FIXME: this ttl calculation is generous */
3292         session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;
3293         mutex_unlock(&mdsc->mutex);
3294
3295         mutex_lock(&session->s_mutex);
3296
3297         dout("handle_session mds%d %s %p state %s seq %llu\n",
3298              mds, ceph_session_op_name(op), session,
3299              ceph_session_state_name(session->s_state), seq);
3300
3301         if (session->s_state == CEPH_MDS_SESSION_HUNG) {
3302                 session->s_state = CEPH_MDS_SESSION_OPEN;
3303                 pr_info("mds%d came back\n", session->s_mds);
3304         }
3305
3306         switch (op) {
3307         case CEPH_SESSION_OPEN:
3308                 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3309                         pr_info("mds%d reconnect success\n", session->s_mds);
3310                 session->s_state = CEPH_MDS_SESSION_OPEN;
3311                 session->s_features = features;
3312                 renewed_caps(mdsc, session, 0);
3313                 wake = 1;
3314                 if (mdsc->stopping)
3315                         __close_session(mdsc, session);
3316                 break;
3317
3318         case CEPH_SESSION_RENEWCAPS:
3319                 if (session->s_renew_seq == seq)
3320                         renewed_caps(mdsc, session, 1);
3321                 break;
3322
3323         case CEPH_SESSION_CLOSE:
3324                 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3325                         pr_info("mds%d reconnect denied\n", session->s_mds);
3326                 session->s_state = CEPH_MDS_SESSION_CLOSED;
3327                 cleanup_session_requests(mdsc, session);
3328                 remove_session_caps(session);
3329                 wake = 2; /* for good measure */
3330                 wake_up_all(&mdsc->session_close_wq);
3331                 break;
3332
3333         case CEPH_SESSION_STALE:
3334                 pr_info("mds%d caps went stale, renewing\n",
3335                         session->s_mds);
3336                 spin_lock(&session->s_gen_ttl_lock);
3337                 session->s_cap_gen++;
3338                 session->s_cap_ttl = jiffies - 1;
3339                 spin_unlock(&session->s_gen_ttl_lock);
3340                 send_renew_caps(mdsc, session);
3341                 break;
3342
3343         case CEPH_SESSION_RECALL_STATE:
3344                 ceph_trim_caps(mdsc, session, le32_to_cpu(h->max_caps));
3345                 break;
3346
3347         case CEPH_SESSION_FLUSHMSG:
3348                 send_flushmsg_ack(mdsc, session, seq);
3349                 break;
3350
3351         case CEPH_SESSION_FORCE_RO:
3352                 dout("force_session_readonly %p\n", session);
3353                 spin_lock(&session->s_cap_lock);
3354                 session->s_readonly = true;
3355                 spin_unlock(&session->s_cap_lock);
3356                 wake_up_session_caps(session, FORCE_RO);
3357                 break;
3358
3359         case CEPH_SESSION_REJECT:
3360                 WARN_ON(session->s_state != CEPH_MDS_SESSION_OPENING);
3361                 pr_info("mds%d rejected session\n", session->s_mds);
3362                 session->s_state = CEPH_MDS_SESSION_REJECTED;
3363                 cleanup_session_requests(mdsc, session);
3364                 remove_session_caps(session);
3365                 if (blacklisted)
3366                         mdsc->fsc->blacklisted = true;
3367                 wake = 2; /* for good measure */
3368                 break;
3369
3370         default:
3371                 pr_err("mdsc_handle_session bad op %d mds%d\n", op, mds);
3372                 WARN_ON(1);
3373         }
3374
3375         mutex_unlock(&session->s_mutex);
3376         if (wake) {
3377                 mutex_lock(&mdsc->mutex);
3378                 __wake_requests(mdsc, &session->s_waiting);
3379                 if (wake == 2)
3380                         kick_requests(mdsc, mds);
3381                 mutex_unlock(&mdsc->mutex);
3382         }
3383         if (op == CEPH_SESSION_CLOSE)
3384                 ceph_put_mds_session(session);
3385         return;
3386
3387 bad:
3388         pr_err("mdsc_handle_session corrupt message mds%d len %d\n", mds,
3389                (int)msg->front.iov_len);
3390         ceph_msg_dump(msg);
3391         return;
3392 }
3393
3394 void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req)
3395 {
3396         int dcaps;
3397
3398         dcaps = xchg(&req->r_dir_caps, 0);
3399         if (dcaps) {
3400                 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3401                 ceph_put_cap_refs(ceph_inode(req->r_parent), dcaps);
3402         }
3403 }
3404
3405 void ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request *req)
3406 {
3407         int dcaps;
3408
3409         dcaps = xchg(&req->r_dir_caps, 0);
3410         if (dcaps) {
3411                 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3412                 ceph_put_cap_refs_no_check_caps(ceph_inode(req->r_parent),
3413                                                 dcaps);
3414         }
3415 }
3416
3417 /*
3418  * called under session->mutex.
3419  */
3420 static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
3421                                    struct ceph_mds_session *session)
3422 {
3423         struct ceph_mds_request *req, *nreq;
3424         struct rb_node *p;
3425
3426         dout("replay_unsafe_requests mds%d\n", session->s_mds);
3427
3428         mutex_lock(&mdsc->mutex);
3429         list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item)
3430                 __send_request(mdsc, session, req, true);
3431
3432         /*
3433          * also re-send old requests when MDS enters reconnect stage. So that MDS
3434          * can process completed request in clientreplay stage.
3435          */
3436         p = rb_first(&mdsc->request_tree);
3437         while (p) {
3438                 req = rb_entry(p, struct ceph_mds_request, r_node);
3439                 p = rb_next(p);
3440                 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
3441                         continue;
3442                 if (req->r_attempts == 0)
3443                         continue; /* only old requests */
3444                 if (!req->r_session)
3445                         continue;
3446                 if (req->r_session->s_mds != session->s_mds)
3447                         continue;
3448
3449                 ceph_mdsc_release_dir_caps_no_check(req);
3450
3451                 __send_request(mdsc, session, req, true);
3452         }
3453         mutex_unlock(&mdsc->mutex);
3454 }
3455
3456 static int send_reconnect_partial(struct ceph_reconnect_state *recon_state)
3457 {
3458         struct ceph_msg *reply;
3459         struct ceph_pagelist *_pagelist;
3460         struct page *page;
3461         __le32 *addr;
3462         int err = -ENOMEM;
3463
3464         if (!recon_state->allow_multi)
3465                 return -ENOSPC;
3466
3467         /* can't handle message that contains both caps and realm */
3468         BUG_ON(!recon_state->nr_caps == !recon_state->nr_realms);
3469
3470         /* pre-allocate new pagelist */
3471         _pagelist = ceph_pagelist_alloc(GFP_NOFS);
3472         if (!_pagelist)
3473                 return -ENOMEM;
3474
3475         reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3476         if (!reply)
3477                 goto fail_msg;
3478
3479         /* placeholder for nr_caps */
3480         err = ceph_pagelist_encode_32(_pagelist, 0);
3481         if (err < 0)
3482                 goto fail;
3483
3484         if (recon_state->nr_caps) {
3485                 /* currently encoding caps */
3486                 err = ceph_pagelist_encode_32(recon_state->pagelist, 0);
3487                 if (err)
3488                         goto fail;
3489         } else {
3490                 /* placeholder for nr_realms (currently encoding relams) */
3491                 err = ceph_pagelist_encode_32(_pagelist, 0);
3492                 if (err < 0)
3493                         goto fail;
3494         }
3495
3496         err = ceph_pagelist_encode_8(recon_state->pagelist, 1);
3497         if (err)
3498                 goto fail;
3499
3500         page = list_first_entry(&recon_state->pagelist->head, struct page, lru);
3501         addr = kmap_atomic(page);
3502         if (recon_state->nr_caps) {
3503                 /* currently encoding caps */
3504                 *addr = cpu_to_le32(recon_state->nr_caps);
3505         } else {
3506                 /* currently encoding relams */
3507                 *(addr + 1) = cpu_to_le32(recon_state->nr_realms);
3508         }
3509         kunmap_atomic(addr);
3510
3511         reply->hdr.version = cpu_to_le16(5);
3512         reply->hdr.compat_version = cpu_to_le16(4);
3513
3514         reply->hdr.data_len = cpu_to_le32(recon_state->pagelist->length);
3515         ceph_msg_data_add_pagelist(reply, recon_state->pagelist);
3516
3517         ceph_con_send(&recon_state->session->s_con, reply);
3518         ceph_pagelist_release(recon_state->pagelist);
3519
3520         recon_state->pagelist = _pagelist;
3521         recon_state->nr_caps = 0;
3522         recon_state->nr_realms = 0;
3523         recon_state->msg_version = 5;
3524         return 0;
3525 fail:
3526         ceph_msg_put(reply);
3527 fail_msg:
3528         ceph_pagelist_release(_pagelist);
3529         return err;
3530 }
3531
3532 /*
3533  * Encode information about a cap for a reconnect with the MDS.
3534  */
3535 static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
3536                           void *arg)
3537 {
3538         union {
3539                 struct ceph_mds_cap_reconnect v2;
3540                 struct ceph_mds_cap_reconnect_v1 v1;
3541         } rec;
3542         struct ceph_inode_info *ci = cap->ci;
3543         struct ceph_reconnect_state *recon_state = arg;
3544         struct ceph_pagelist *pagelist = recon_state->pagelist;
3545         int err;
3546         u64 snap_follows;
3547
3548         dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
3549              inode, ceph_vinop(inode), cap, cap->cap_id,
3550              ceph_cap_string(cap->issued));
3551
3552         spin_lock(&ci->i_ceph_lock);
3553         cap->seq = 0;        /* reset cap seq */
3554         cap->issue_seq = 0;  /* and issue_seq */
3555         cap->mseq = 0;       /* and migrate_seq */
3556         cap->cap_gen = cap->session->s_cap_gen;
3557
3558         /* These are lost when the session goes away */
3559         if (S_ISDIR(inode->i_mode)) {
3560                 if (cap->issued & CEPH_CAP_DIR_CREATE) {
3561                         ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns));
3562                         memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout));
3563                 }
3564                 cap->issued &= ~CEPH_CAP_ANY_DIR_OPS;
3565         }
3566
3567         if (recon_state->msg_version >= 2) {
3568                 rec.v2.cap_id = cpu_to_le64(cap->cap_id);
3569                 rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3570                 rec.v2.issued = cpu_to_le32(cap->issued);
3571                 rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3572                 rec.v2.pathbase = 0;
3573                 rec.v2.flock_len = (__force __le32)
3574                         ((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1);
3575         } else {
3576                 rec.v1.cap_id = cpu_to_le64(cap->cap_id);
3577                 rec.v1.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3578                 rec.v1.issued = cpu_to_le32(cap->issued);
3579                 rec.v1.size = cpu_to_le64(inode->i_size);
3580                 ceph_encode_timespec64(&rec.v1.mtime, &inode->i_mtime);
3581                 ceph_encode_timespec64(&rec.v1.atime, &inode->i_atime);
3582                 rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3583                 rec.v1.pathbase = 0;
3584         }
3585
3586         if (list_empty(&ci->i_cap_snaps)) {
3587                 snap_follows = ci->i_head_snapc ? ci->i_head_snapc->seq : 0;
3588         } else {
3589                 struct ceph_cap_snap *capsnap =
3590                         list_first_entry(&ci->i_cap_snaps,
3591                                          struct ceph_cap_snap, ci_item);
3592                 snap_follows = capsnap->follows;
3593         }
3594         spin_unlock(&ci->i_ceph_lock);
3595
3596         if (recon_state->msg_version >= 2) {
3597                 int num_fcntl_locks, num_flock_locks;
3598                 struct ceph_filelock *flocks = NULL;
3599                 size_t struct_len, total_len = sizeof(u64);
3600                 u8 struct_v = 0;
3601
3602 encode_again:
3603                 if (rec.v2.flock_len) {
3604                         ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
3605                 } else {
3606                         num_fcntl_locks = 0;
3607                         num_flock_locks = 0;
3608                 }
3609                 if (num_fcntl_locks + num_flock_locks > 0) {
3610                         flocks = kmalloc_array(num_fcntl_locks + num_flock_locks,
3611                                                sizeof(struct ceph_filelock),
3612                                                GFP_NOFS);
3613                         if (!flocks) {
3614                                 err = -ENOMEM;
3615                                 goto out_err;
3616                         }
3617                         err = ceph_encode_locks_to_buffer(inode, flocks,
3618                                                           num_fcntl_locks,
3619                                                           num_flock_locks);
3620                         if (err) {
3621                                 kfree(flocks);
3622                                 flocks = NULL;
3623                                 if (err == -ENOSPC)
3624                                         goto encode_again;
3625                                 goto out_err;
3626                         }
3627                 } else {
3628                         kfree(flocks);
3629                         flocks = NULL;
3630                 }
3631
3632                 if (recon_state->msg_version >= 3) {
3633                         /* version, compat_version and struct_len */
3634                         total_len += 2 * sizeof(u8) + sizeof(u32);
3635                         struct_v = 2;
3636                 }
3637                 /*
3638                  * number of encoded locks is stable, so copy to pagelist
3639                  */
3640                 struct_len = 2 * sizeof(u32) +
3641                             (num_fcntl_locks + num_flock_locks) *
3642                             sizeof(struct ceph_filelock);
3643                 rec.v2.flock_len = cpu_to_le32(struct_len);
3644
3645                 struct_len += sizeof(u32) + sizeof(rec.v2);
3646
3647                 if (struct_v >= 2)
3648                         struct_len += sizeof(u64); /* snap_follows */
3649
3650                 total_len += struct_len;
3651
3652                 if (pagelist->length + total_len > RECONNECT_MAX_SIZE) {
3653                         err = send_reconnect_partial(recon_state);
3654                         if (err)
3655                                 goto out_freeflocks;
3656                         pagelist = recon_state->pagelist;
3657                 }
3658
3659                 err = ceph_pagelist_reserve(pagelist, total_len);
3660                 if (err)
3661                         goto out_freeflocks;
3662
3663                 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
3664                 if (recon_state->msg_version >= 3) {
3665                         ceph_pagelist_encode_8(pagelist, struct_v);
3666                         ceph_pagelist_encode_8(pagelist, 1);
3667                         ceph_pagelist_encode_32(pagelist, struct_len);
3668                 }
3669                 ceph_pagelist_encode_string(pagelist, NULL, 0);
3670                 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v2));
3671                 ceph_locks_to_pagelist(flocks, pagelist,
3672                                        num_fcntl_locks, num_flock_locks);
3673                 if (struct_v >= 2)
3674                         ceph_pagelist_encode_64(pagelist, snap_follows);
3675 out_freeflocks:
3676                 kfree(flocks);
3677         } else {
3678                 u64 pathbase = 0;
3679                 int pathlen = 0;
3680                 char *path = NULL;
3681                 struct dentry *dentry;
3682
3683                 dentry = d_find_alias(inode);
3684                 if (dentry) {
3685                         path = ceph_mdsc_build_path(dentry,
3686                                                 &pathlen, &pathbase, 0);
3687                         dput(dentry);
3688                         if (IS_ERR(path)) {
3689                                 err = PTR_ERR(path);
3690                                 goto out_err;
3691                         }
3692                         rec.v1.pathbase = cpu_to_le64(pathbase);
3693                 }
3694
3695                 err = ceph_pagelist_reserve(pagelist,
3696                                             sizeof(u64) + sizeof(u32) +
3697                                             pathlen + sizeof(rec.v1));
3698                 if (err) {
3699                         goto out_freepath;
3700                 }
3701
3702                 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
3703                 ceph_pagelist_encode_string(pagelist, path, pathlen);
3704                 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v1));
3705 out_freepath:
3706                 ceph_mdsc_free_path(path, pathlen);
3707         }
3708
3709 out_err:
3710         if (err >= 0)
3711                 recon_state->nr_caps++;
3712         return err;
3713 }
3714
3715 static int encode_snap_realms(struct ceph_mds_client *mdsc,
3716                               struct ceph_reconnect_state *recon_state)
3717 {
3718         struct rb_node *p;
3719         struct ceph_pagelist *pagelist = recon_state->pagelist;
3720         int err = 0;
3721
3722         if (recon_state->msg_version >= 4) {
3723                 err = ceph_pagelist_encode_32(pagelist, mdsc->num_snap_realms);
3724                 if (err < 0)
3725                         goto fail;
3726         }
3727
3728         /*
3729          * snaprealms.  we provide mds with the ino, seq (version), and
3730          * parent for all of our realms.  If the mds has any newer info,
3731          * it will tell us.
3732          */
3733         for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {
3734                 struct ceph_snap_realm *realm =
3735                        rb_entry(p, struct ceph_snap_realm, node);
3736                 struct ceph_mds_snaprealm_reconnect sr_rec;
3737
3738                 if (recon_state->msg_version >= 4) {
3739                         size_t need = sizeof(u8) * 2 + sizeof(u32) +
3740                                       sizeof(sr_rec);
3741
3742                         if (pagelist->length + need > RECONNECT_MAX_SIZE) {
3743                                 err = send_reconnect_partial(recon_state);
3744                                 if (err)
3745                                         goto fail;
3746                                 pagelist = recon_state->pagelist;
3747                         }
3748
3749                         err = ceph_pagelist_reserve(pagelist, need);
3750                         if (err)
3751                                 goto fail;
3752
3753                         ceph_pagelist_encode_8(pagelist, 1);
3754                         ceph_pagelist_encode_8(pagelist, 1);
3755                         ceph_pagelist_encode_32(pagelist, sizeof(sr_rec));
3756                 }
3757
3758                 dout(" adding snap realm %llx seq %lld parent %llx\n",
3759                      realm->ino, realm->seq, realm->parent_ino);
3760                 sr_rec.ino = cpu_to_le64(realm->ino);
3761                 sr_rec.seq = cpu_to_le64(realm->seq);
3762                 sr_rec.parent = cpu_to_le64(realm->parent_ino);
3763
3764                 err = ceph_pagelist_append(pagelist, &sr_rec, sizeof(sr_rec));
3765                 if (err)
3766                         goto fail;
3767
3768                 recon_state->nr_realms++;
3769         }
3770 fail:
3771         return err;
3772 }
3773
3774
3775 /*
3776  * If an MDS fails and recovers, clients need to reconnect in order to
3777  * reestablish shared state.  This includes all caps issued through
3778  * this session _and_ the snap_realm hierarchy.  Because it's not
3779  * clear which snap realms the mds cares about, we send everything we
3780  * know about.. that ensures we'll then get any new info the
3781  * recovering MDS might have.
3782  *
3783  * This is a relatively heavyweight operation, but it's rare.
3784  */
3785 static void send_mds_reconnect(struct ceph_mds_client *mdsc,
3786                                struct ceph_mds_session *session)
3787 {
3788         struct ceph_msg *reply;
3789         int mds = session->s_mds;
3790         int err = -ENOMEM;
3791         struct ceph_reconnect_state recon_state = {
3792                 .session = session,
3793         };
3794         LIST_HEAD(dispose);
3795
3796         pr_info("mds%d reconnect start\n", mds);
3797
3798         recon_state.pagelist = ceph_pagelist_alloc(GFP_NOFS);
3799         if (!recon_state.pagelist)
3800                 goto fail_nopagelist;
3801
3802         reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3803         if (!reply)
3804                 goto fail_nomsg;
3805
3806         xa_destroy(&session->s_delegated_inos);
3807
3808         mutex_lock(&session->s_mutex);
3809         session->s_state = CEPH_MDS_SESSION_RECONNECTING;
3810         session->s_seq = 0;
3811
3812         dout("session %p state %s\n", session,
3813              ceph_session_state_name(session->s_state));
3814
3815         spin_lock(&session->s_gen_ttl_lock);
3816         session->s_cap_gen++;
3817         spin_unlock(&session->s_gen_ttl_lock);
3818
3819         spin_lock(&session->s_cap_lock);
3820         /* don't know if session is readonly */
3821         session->s_readonly = 0;
3822         /*
3823          * notify __ceph_remove_cap() that we are composing cap reconnect.
3824          * If a cap get released before being added to the cap reconnect,
3825          * __ceph_remove_cap() should skip queuing cap release.
3826          */
3827         session->s_cap_reconnect = 1;
3828         /* drop old cap expires; we're about to reestablish that state */
3829         detach_cap_releases(session, &dispose);
3830         spin_unlock(&session->s_cap_lock);
3831         dispose_cap_releases(mdsc, &dispose);
3832
3833         /* trim unused caps to reduce MDS's cache rejoin time */
3834         if (mdsc->fsc->sb->s_root)
3835                 shrink_dcache_parent(mdsc->fsc->sb->s_root);
3836
3837         ceph_con_close(&session->s_con);
3838         ceph_con_open(&session->s_con,
3839                       CEPH_ENTITY_TYPE_MDS, mds,
3840                       ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
3841
3842         /* replay unsafe requests */
3843         replay_unsafe_requests(mdsc, session);
3844
3845         ceph_early_kick_flushing_caps(mdsc, session);
3846
3847         down_read(&mdsc->snap_rwsem);
3848
3849         /* placeholder for nr_caps */
3850         err = ceph_pagelist_encode_32(recon_state.pagelist, 0);
3851         if (err)
3852                 goto fail;
3853
3854         if (test_bit(CEPHFS_FEATURE_MULTI_RECONNECT, &session->s_features)) {
3855                 recon_state.msg_version = 3;
3856                 recon_state.allow_multi = true;
3857         } else if (session->s_con.peer_features & CEPH_FEATURE_MDSENC) {
3858                 recon_state.msg_version = 3;
3859         } else {
3860                 recon_state.msg_version = 2;
3861         }
3862         /* trsaverse this session's caps */
3863         err = ceph_iterate_session_caps(session, reconnect_caps_cb, &recon_state);
3864
3865         spin_lock(&session->s_cap_lock);
3866         session->s_cap_reconnect = 0;
3867         spin_unlock(&session->s_cap_lock);
3868
3869         if (err < 0)
3870                 goto fail;
3871
3872         /* check if all realms can be encoded into current message */
3873         if (mdsc->num_snap_realms) {
3874                 size_t total_len =
3875                         recon_state.pagelist->length +
3876                         mdsc->num_snap_realms *
3877                         sizeof(struct ceph_mds_snaprealm_reconnect);
3878                 if (recon_state.msg_version >= 4) {
3879                         /* number of realms */
3880                         total_len += sizeof(u32);
3881                         /* version, compat_version and struct_len */
3882                         total_len += mdsc->num_snap_realms *
3883                                      (2 * sizeof(u8) + sizeof(u32));
3884                 }
3885                 if (total_len > RECONNECT_MAX_SIZE) {
3886                         if (!recon_state.allow_multi) {
3887                                 err = -ENOSPC;
3888                                 goto fail;
3889                         }
3890                         if (recon_state.nr_caps) {
3891                                 err = send_reconnect_partial(&recon_state);
3892                                 if (err)
3893                                         goto fail;
3894                         }
3895                         recon_state.msg_version = 5;
3896                 }
3897         }
3898
3899         err = encode_snap_realms(mdsc, &recon_state);
3900         if (err < 0)
3901                 goto fail;
3902
3903         if (recon_state.msg_version >= 5) {
3904                 err = ceph_pagelist_encode_8(recon_state.pagelist, 0);
3905                 if (err < 0)
3906                         goto fail;
3907         }
3908
3909         if (recon_state.nr_caps || recon_state.nr_realms) {
3910                 struct page *page =
3911                         list_first_entry(&recon_state.pagelist->head,
3912                                         struct page, lru);
3913                 __le32 *addr = kmap_atomic(page);
3914                 if (recon_state.nr_caps) {
3915                         WARN_ON(recon_state.nr_realms != mdsc->num_snap_realms);
3916                         *addr = cpu_to_le32(recon_state.nr_caps);
3917                 } else if (recon_state.msg_version >= 4) {
3918                         *(addr + 1) = cpu_to_le32(recon_state.nr_realms);
3919                 }
3920                 kunmap_atomic(addr);
3921         }
3922
3923         reply->hdr.version = cpu_to_le16(recon_state.msg_version);
3924         if (recon_state.msg_version >= 4)
3925                 reply->hdr.compat_version = cpu_to_le16(4);
3926
3927         reply->hdr.data_len = cpu_to_le32(recon_state.pagelist->length);
3928         ceph_msg_data_add_pagelist(reply, recon_state.pagelist);
3929
3930         ceph_con_send(&session->s_con, reply);
3931
3932         mutex_unlock(&session->s_mutex);
3933
3934         mutex_lock(&mdsc->mutex);
3935         __wake_requests(mdsc, &session->s_waiting);
3936         mutex_unlock(&mdsc->mutex);
3937
3938         up_read(&mdsc->snap_rwsem);
3939         ceph_pagelist_release(recon_state.pagelist);
3940         return;
3941
3942 fail:
3943         ceph_msg_put(reply);
3944         up_read(&mdsc->snap_rwsem);
3945         mutex_unlock(&session->s_mutex);
3946 fail_nomsg:
3947         ceph_pagelist_release(recon_state.pagelist);
3948 fail_nopagelist:
3949         pr_err("error %d preparing reconnect for mds%d\n", err, mds);
3950         return;
3951 }
3952
3953
3954 /*
3955  * compare old and new mdsmaps, kicking requests
3956  * and closing out old connections as necessary
3957  *
3958  * called under mdsc->mutex.
3959  */
3960 static void check_new_map(struct ceph_mds_client *mdsc,
3961                           struct ceph_mdsmap *newmap,
3962                           struct ceph_mdsmap *oldmap)
3963 {
3964         int i;
3965         int oldstate, newstate;
3966         struct ceph_mds_session *s;
3967
3968         dout("check_new_map new %u old %u\n",
3969              newmap->m_epoch, oldmap->m_epoch);
3970
3971         for (i = 0; i < oldmap->possible_max_rank && i < mdsc->max_sessions; i++) {
3972                 if (!mdsc->sessions[i])
3973                         continue;
3974                 s = mdsc->sessions[i];
3975                 oldstate = ceph_mdsmap_get_state(oldmap, i);
3976                 newstate = ceph_mdsmap_get_state(newmap, i);
3977
3978                 dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",
3979                      i, ceph_mds_state_name(oldstate),
3980                      ceph_mdsmap_is_laggy(oldmap, i) ? " (laggy)" : "",
3981                      ceph_mds_state_name(newstate),
3982                      ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
3983                      ceph_session_state_name(s->s_state));
3984
3985                 if (i >= newmap->possible_max_rank) {
3986                         /* force close session for stopped mds */
3987                         ceph_get_mds_session(s);
3988                         __unregister_session(mdsc, s);
3989                         __wake_requests(mdsc, &s->s_waiting);
3990                         mutex_unlock(&mdsc->mutex);
3991
3992                         mutex_lock(&s->s_mutex);
3993                         cleanup_session_requests(mdsc, s);
3994                         remove_session_caps(s);
3995                         mutex_unlock(&s->s_mutex);
3996
3997                         ceph_put_mds_session(s);
3998
3999                         mutex_lock(&mdsc->mutex);
4000                         kick_requests(mdsc, i);
4001                         continue;
4002                 }
4003
4004                 if (memcmp(ceph_mdsmap_get_addr(oldmap, i),
4005                            ceph_mdsmap_get_addr(newmap, i),
4006                            sizeof(struct ceph_entity_addr))) {
4007                         /* just close it */
4008                         mutex_unlock(&mdsc->mutex);
4009                         mutex_lock(&s->s_mutex);
4010                         mutex_lock(&mdsc->mutex);
4011                         ceph_con_close(&s->s_con);
4012                         mutex_unlock(&s->s_mutex);
4013                         s->s_state = CEPH_MDS_SESSION_RESTARTING;
4014                 } else if (oldstate == newstate) {
4015                         continue;  /* nothing new with this mds */
4016                 }
4017
4018                 /*
4019                  * send reconnect?
4020                  */
4021                 if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&
4022                     newstate >= CEPH_MDS_STATE_RECONNECT) {
4023                         mutex_unlock(&mdsc->mutex);
4024                         send_mds_reconnect(mdsc, s);
4025                         mutex_lock(&mdsc->mutex);
4026                 }
4027
4028                 /*
4029                  * kick request on any mds that has gone active.
4030                  */
4031                 if (oldstate < CEPH_MDS_STATE_ACTIVE &&
4032                     newstate >= CEPH_MDS_STATE_ACTIVE) {
4033                         if (oldstate != CEPH_MDS_STATE_CREATING &&
4034                             oldstate != CEPH_MDS_STATE_STARTING)
4035                                 pr_info("mds%d recovery completed\n", s->s_mds);
4036                         kick_requests(mdsc, i);
4037                         mutex_unlock(&mdsc->mutex);
4038                         mutex_lock(&s->s_mutex);
4039                         mutex_lock(&mdsc->mutex);
4040                         ceph_kick_flushing_caps(mdsc, s);
4041                         mutex_unlock(&s->s_mutex);
4042                         wake_up_session_caps(s, RECONNECT);
4043                 }
4044         }
4045
4046         for (i = 0; i < newmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4047                 s = mdsc->sessions[i];
4048                 if (!s)
4049                         continue;
4050                 if (!ceph_mdsmap_is_laggy(newmap, i))
4051                         continue;
4052                 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4053                     s->s_state == CEPH_MDS_SESSION_HUNG ||
4054                     s->s_state == CEPH_MDS_SESSION_CLOSING) {
4055                         dout(" connecting to export targets of laggy mds%d\n",
4056                              i);
4057                         __open_export_target_sessions(mdsc, s);
4058                 }
4059         }
4060 }
4061
4062
4063
4064 /*
4065  * leases
4066  */
4067
4068 /*
4069  * caller must hold session s_mutex, dentry->d_lock
4070  */
4071 void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry)
4072 {
4073         struct ceph_dentry_info *di = ceph_dentry(dentry);
4074
4075         ceph_put_mds_session(di->lease_session);
4076         di->lease_session = NULL;
4077 }
4078
4079 static void handle_lease(struct ceph_mds_client *mdsc,
4080                          struct ceph_mds_session *session,
4081                          struct ceph_msg *msg)
4082 {
4083         struct super_block *sb = mdsc->fsc->sb;
4084         struct inode *inode;
4085         struct dentry *parent, *dentry;
4086         struct ceph_dentry_info *di;
4087         int mds = session->s_mds;
4088         struct ceph_mds_lease *h = msg->front.iov_base;
4089         u32 seq;
4090         struct ceph_vino vino;
4091         struct qstr dname;
4092         int release = 0;
4093
4094         dout("handle_lease from mds%d\n", mds);
4095
4096         /* decode */
4097         if (msg->front.iov_len < sizeof(*h) + sizeof(u32))
4098                 goto bad;
4099         vino.ino = le64_to_cpu(h->ino);
4100         vino.snap = CEPH_NOSNAP;
4101         seq = le32_to_cpu(h->seq);
4102         dname.len = get_unaligned_le32(h + 1);
4103         if (msg->front.iov_len < sizeof(*h) + sizeof(u32) + dname.len)
4104                 goto bad;
4105         dname.name = (void *)(h + 1) + sizeof(u32);
4106
4107         /* lookup inode */
4108         inode = ceph_find_inode(sb, vino);
4109         dout("handle_lease %s, ino %llx %p %.*s\n",
4110              ceph_lease_op_name(h->action), vino.ino, inode,
4111              dname.len, dname.name);
4112
4113         mutex_lock(&session->s_mutex);
4114         session->s_seq++;
4115
4116         if (!inode) {
4117                 dout("handle_lease no inode %llx\n", vino.ino);
4118                 goto release;
4119         }
4120
4121         /* dentry */
4122         parent = d_find_alias(inode);
4123         if (!parent) {
4124                 dout("no parent dentry on inode %p\n", inode);
4125                 WARN_ON(1);
4126                 goto release;  /* hrm... */
4127         }
4128         dname.hash = full_name_hash(parent, dname.name, dname.len);
4129         dentry = d_lookup(parent, &dname);
4130         dput(parent);
4131         if (!dentry)
4132                 goto release;
4133
4134         spin_lock(&dentry->d_lock);
4135         di = ceph_dentry(dentry);
4136         switch (h->action) {
4137         case CEPH_MDS_LEASE_REVOKE:
4138                 if (di->lease_session == session) {
4139                         if (ceph_seq_cmp(di->lease_seq, seq) > 0)
4140                                 h->seq = cpu_to_le32(di->lease_seq);
4141                         __ceph_mdsc_drop_dentry_lease(dentry);
4142                 }
4143                 release = 1;
4144                 break;
4145
4146         case CEPH_MDS_LEASE_RENEW:
4147                 if (di->lease_session == session &&
4148                     di->lease_gen == session->s_cap_gen &&
4149                     di->lease_renew_from &&
4150                     di->lease_renew_after == 0) {
4151                         unsigned long duration =
4152                                 msecs_to_jiffies(le32_to_cpu(h->duration_ms));
4153
4154                         di->lease_seq = seq;
4155                         di->time = di->lease_renew_from + duration;
4156                         di->lease_renew_after = di->lease_renew_from +
4157                                 (duration >> 1);
4158                         di->lease_renew_from = 0;
4159                 }
4160                 break;
4161         }
4162         spin_unlock(&dentry->d_lock);
4163         dput(dentry);
4164
4165         if (!release)
4166                 goto out;
4167
4168 release:
4169         /* let's just reuse the same message */
4170         h->action = CEPH_MDS_LEASE_REVOKE_ACK;
4171         ceph_msg_get(msg);
4172         ceph_con_send(&session->s_con, msg);
4173
4174 out:
4175         mutex_unlock(&session->s_mutex);
4176         /* avoid calling iput_final() in mds dispatch threads */
4177         ceph_async_iput(inode);
4178         return;
4179
4180 bad:
4181         pr_err("corrupt lease message\n");
4182         ceph_msg_dump(msg);
4183 }
4184
4185 void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
4186                               struct dentry *dentry, char action,
4187                               u32 seq)
4188 {
4189         struct ceph_msg *msg;
4190         struct ceph_mds_lease *lease;
4191         struct inode *dir;
4192         int len = sizeof(*lease) + sizeof(u32) + NAME_MAX;
4193
4194         dout("lease_send_msg identry %p %s to mds%d\n",
4195              dentry, ceph_lease_op_name(action), session->s_mds);
4196
4197         msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
4198         if (!msg)
4199                 return;
4200         lease = msg->front.iov_base;
4201         lease->action = action;
4202         lease->seq = cpu_to_le32(seq);
4203
4204         spin_lock(&dentry->d_lock);
4205         dir = d_inode(dentry->d_parent);
4206         lease->ino = cpu_to_le64(ceph_ino(dir));
4207         lease->first = lease->last = cpu_to_le64(ceph_snap(dir));
4208
4209         put_unaligned_le32(dentry->d_name.len, lease + 1);
4210         memcpy((void *)(lease + 1) + 4,
4211                dentry->d_name.name, dentry->d_name.len);
4212         spin_unlock(&dentry->d_lock);
4213         /*
4214          * if this is a preemptive lease RELEASE, no need to
4215          * flush request stream, since the actual request will
4216          * soon follow.
4217          */
4218         msg->more_to_follow = (action == CEPH_MDS_LEASE_RELEASE);
4219
4220         ceph_con_send(&session->s_con, msg);
4221 }
4222
4223 /*
4224  * lock unlock sessions, to wait ongoing session activities
4225  */
4226 static void lock_unlock_sessions(struct ceph_mds_client *mdsc)
4227 {
4228         int i;
4229
4230         mutex_lock(&mdsc->mutex);
4231         for (i = 0; i < mdsc->max_sessions; i++) {
4232                 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
4233                 if (!s)
4234                         continue;
4235                 mutex_unlock(&mdsc->mutex);
4236                 mutex_lock(&s->s_mutex);
4237                 mutex_unlock(&s->s_mutex);
4238                 ceph_put_mds_session(s);
4239                 mutex_lock(&mdsc->mutex);
4240         }
4241         mutex_unlock(&mdsc->mutex);
4242 }
4243
4244 static void maybe_recover_session(struct ceph_mds_client *mdsc)
4245 {
4246         struct ceph_fs_client *fsc = mdsc->fsc;
4247
4248         if (!ceph_test_mount_opt(fsc, CLEANRECOVER))
4249                 return;
4250
4251         if (READ_ONCE(fsc->mount_state) != CEPH_MOUNT_MOUNTED)
4252                 return;
4253
4254         if (!READ_ONCE(fsc->blacklisted))
4255                 return;
4256
4257         if (fsc->last_auto_reconnect &&
4258             time_before(jiffies, fsc->last_auto_reconnect + HZ * 60 * 30))
4259                 return;
4260
4261         pr_info("auto reconnect after blacklisted\n");
4262         fsc->last_auto_reconnect = jiffies;
4263         ceph_force_reconnect(fsc->sb);
4264 }
4265
4266 /*
4267  * delayed work -- periodically trim expired leases, renew caps with mds
4268  */
4269 static void schedule_delayed(struct ceph_mds_client *mdsc)
4270 {
4271         int delay = 5;
4272         unsigned hz = round_jiffies_relative(HZ * delay);
4273         schedule_delayed_work(&mdsc->delayed_work, hz);
4274 }
4275
4276 static void delayed_work(struct work_struct *work)
4277 {
4278         int i;
4279         struct ceph_mds_client *mdsc =
4280                 container_of(work, struct ceph_mds_client, delayed_work.work);
4281         int renew_interval;
4282         int renew_caps;
4283
4284         dout("mdsc delayed_work\n");
4285
4286         mutex_lock(&mdsc->mutex);
4287         renew_interval = mdsc->mdsmap->m_session_timeout >> 2;
4288         renew_caps = time_after_eq(jiffies, HZ*renew_interval +
4289                                    mdsc->last_renew_caps);
4290         if (renew_caps)
4291                 mdsc->last_renew_caps = jiffies;
4292
4293         for (i = 0; i < mdsc->max_sessions; i++) {
4294                 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
4295                 if (!s)
4296                         continue;
4297                 if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
4298                         dout("resending session close request for mds%d\n",
4299                              s->s_mds);
4300                         request_close_session(mdsc, s);
4301                         ceph_put_mds_session(s);
4302                         continue;
4303                 }
4304                 if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
4305                         if (s->s_state == CEPH_MDS_SESSION_OPEN) {
4306                                 s->s_state = CEPH_MDS_SESSION_HUNG;
4307                                 pr_info("mds%d hung\n", s->s_mds);
4308                         }
4309                 }
4310                 if (s->s_state == CEPH_MDS_SESSION_NEW ||
4311                     s->s_state == CEPH_MDS_SESSION_RESTARTING ||
4312                     s->s_state == CEPH_MDS_SESSION_REJECTED) {
4313                         /* this mds is failed or recovering, just wait */
4314                         ceph_put_mds_session(s);
4315                         continue;
4316                 }
4317                 mutex_unlock(&mdsc->mutex);
4318
4319                 mutex_lock(&s->s_mutex);
4320                 if (renew_caps)
4321                         send_renew_caps(mdsc, s);
4322                 else
4323                         ceph_con_keepalive(&s->s_con);
4324                 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4325                     s->s_state == CEPH_MDS_SESSION_HUNG)
4326                         ceph_send_cap_releases(mdsc, s);
4327                 mutex_unlock(&s->s_mutex);
4328                 ceph_put_mds_session(s);
4329
4330                 mutex_lock(&mdsc->mutex);
4331         }
4332         mutex_unlock(&mdsc->mutex);
4333
4334         ceph_check_delayed_caps(mdsc);
4335
4336         ceph_queue_cap_reclaim_work(mdsc);
4337
4338         ceph_trim_snapid_map(mdsc);
4339
4340         maybe_recover_session(mdsc);
4341
4342         schedule_delayed(mdsc);
4343 }
4344
4345 int ceph_mdsc_init(struct ceph_fs_client *fsc)
4346
4347 {
4348         struct ceph_mds_client *mdsc;
4349         int err;
4350
4351         mdsc = kzalloc(sizeof(struct ceph_mds_client), GFP_NOFS);
4352         if (!mdsc)
4353                 return -ENOMEM;
4354         mdsc->fsc = fsc;
4355         mutex_init(&mdsc->mutex);
4356         mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);
4357         if (!mdsc->mdsmap) {
4358                 err = -ENOMEM;
4359                 goto err_mdsc;
4360         }
4361
4362         fsc->mdsc = mdsc;
4363         init_completion(&mdsc->safe_umount_waiters);
4364         init_waitqueue_head(&mdsc->session_close_wq);
4365         INIT_LIST_HEAD(&mdsc->waiting_for_map);
4366         mdsc->sessions = NULL;
4367         atomic_set(&mdsc->num_sessions, 0);
4368         mdsc->max_sessions = 0;
4369         mdsc->stopping = 0;
4370         atomic64_set(&mdsc->quotarealms_count, 0);
4371         mdsc->quotarealms_inodes = RB_ROOT;
4372         mutex_init(&mdsc->quotarealms_inodes_mutex);
4373         mdsc->last_snap_seq = 0;
4374         init_rwsem(&mdsc->snap_rwsem);
4375         mdsc->snap_realms = RB_ROOT;
4376         INIT_LIST_HEAD(&mdsc->snap_empty);
4377         mdsc->num_snap_realms = 0;
4378         spin_lock_init(&mdsc->snap_empty_lock);
4379         mdsc->last_tid = 0;
4380         mdsc->oldest_tid = 0;
4381         mdsc->request_tree = RB_ROOT;
4382         INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);
4383         mdsc->last_renew_caps = jiffies;
4384         INIT_LIST_HEAD(&mdsc->cap_delay_list);
4385         INIT_LIST_HEAD(&mdsc->cap_wait_list);
4386         spin_lock_init(&mdsc->cap_delay_lock);
4387         INIT_LIST_HEAD(&mdsc->snap_flush_list);
4388         spin_lock_init(&mdsc->snap_flush_lock);
4389         mdsc->last_cap_flush_tid = 1;
4390         INIT_LIST_HEAD(&mdsc->cap_flush_list);
4391         INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
4392         mdsc->num_cap_flushing = 0;
4393         spin_lock_init(&mdsc->cap_dirty_lock);
4394         init_waitqueue_head(&mdsc->cap_flushing_wq);
4395         INIT_WORK(&mdsc->cap_reclaim_work, ceph_cap_reclaim_work);
4396         atomic_set(&mdsc->cap_reclaim_pending, 0);
4397         err = ceph_metric_init(&mdsc->metric);
4398         if (err)
4399                 goto err_mdsmap;
4400
4401         spin_lock_init(&mdsc->dentry_list_lock);
4402         INIT_LIST_HEAD(&mdsc->dentry_leases);
4403         INIT_LIST_HEAD(&mdsc->dentry_dir_leases);
4404
4405         ceph_caps_init(mdsc);
4406         ceph_adjust_caps_max_min(mdsc, fsc->mount_options);
4407
4408         spin_lock_init(&mdsc->snapid_map_lock);
4409         mdsc->snapid_map_tree = RB_ROOT;
4410         INIT_LIST_HEAD(&mdsc->snapid_map_lru);
4411
4412         init_rwsem(&mdsc->pool_perm_rwsem);
4413         mdsc->pool_perm_tree = RB_ROOT;
4414
4415         strscpy(mdsc->nodename, utsname()->nodename,
4416                 sizeof(mdsc->nodename));
4417         return 0;
4418
4419 err_mdsmap:
4420         kfree(mdsc->mdsmap);
4421 err_mdsc:
4422         kfree(mdsc);
4423         return err;
4424 }
4425
4426 /*
4427  * Wait for safe replies on open mds requests.  If we time out, drop
4428  * all requests from the tree to avoid dangling dentry refs.
4429  */
4430 static void wait_requests(struct ceph_mds_client *mdsc)
4431 {
4432         struct ceph_options *opts = mdsc->fsc->client->options;
4433         struct ceph_mds_request *req;
4434
4435         mutex_lock(&mdsc->mutex);
4436         if (__get_oldest_req(mdsc)) {
4437                 mutex_unlock(&mdsc->mutex);
4438
4439                 dout("wait_requests waiting for requests\n");
4440                 wait_for_completion_timeout(&mdsc->safe_umount_waiters,
4441                                     ceph_timeout_jiffies(opts->mount_timeout));
4442
4443                 /* tear down remaining requests */
4444                 mutex_lock(&mdsc->mutex);
4445                 while ((req = __get_oldest_req(mdsc))) {
4446                         dout("wait_requests timed out on tid %llu\n",
4447                              req->r_tid);
4448                         list_del_init(&req->r_wait);
4449                         __unregister_request(mdsc, req);
4450                 }
4451         }
4452         mutex_unlock(&mdsc->mutex);
4453         dout("wait_requests done\n");
4454 }
4455
4456 /*
4457  * called before mount is ro, and before dentries are torn down.
4458  * (hmm, does this still race with new lookups?)
4459  */
4460 void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
4461 {
4462         dout("pre_umount\n");
4463         mdsc->stopping = 1;
4464
4465         lock_unlock_sessions(mdsc);
4466         ceph_flush_dirty_caps(mdsc);
4467         wait_requests(mdsc);
4468
4469         /*
4470          * wait for reply handlers to drop their request refs and
4471          * their inode/dcache refs
4472          */
4473         ceph_msgr_flush();
4474
4475         ceph_cleanup_quotarealms_inodes(mdsc);
4476 }
4477
4478 /*
4479  * wait for all write mds requests to flush.
4480  */
4481 static void wait_unsafe_requests(struct ceph_mds_client *mdsc, u64 want_tid)
4482 {
4483         struct ceph_mds_request *req = NULL, *nextreq;
4484         struct rb_node *n;
4485
4486         mutex_lock(&mdsc->mutex);
4487         dout("wait_unsafe_requests want %lld\n", want_tid);
4488 restart:
4489         req = __get_oldest_req(mdsc);
4490         while (req && req->r_tid <= want_tid) {
4491                 /* find next request */
4492                 n = rb_next(&req->r_node);
4493                 if (n)
4494                         nextreq = rb_entry(n, struct ceph_mds_request, r_node);
4495                 else
4496                         nextreq = NULL;
4497                 if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&
4498                     (req->r_op & CEPH_MDS_OP_WRITE)) {
4499                         /* write op */
4500                         ceph_mdsc_get_request(req);
4501                         if (nextreq)
4502                                 ceph_mdsc_get_request(nextreq);
4503                         mutex_unlock(&mdsc->mutex);
4504                         dout("wait_unsafe_requests  wait on %llu (want %llu)\n",
4505                              req->r_tid, want_tid);
4506                         wait_for_completion(&req->r_safe_completion);
4507                         mutex_lock(&mdsc->mutex);
4508                         ceph_mdsc_put_request(req);
4509                         if (!nextreq)
4510                                 break;  /* next dne before, so we're done! */
4511                         if (RB_EMPTY_NODE(&nextreq->r_node)) {
4512                                 /* next request was removed from tree */
4513                                 ceph_mdsc_put_request(nextreq);
4514                                 goto restart;
4515                         }
4516                         ceph_mdsc_put_request(nextreq);  /* won't go away */
4517                 }
4518                 req = nextreq;
4519         }
4520         mutex_unlock(&mdsc->mutex);
4521         dout("wait_unsafe_requests done\n");
4522 }
4523
4524 void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
4525 {
4526         u64 want_tid, want_flush;
4527
4528         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
4529                 return;
4530
4531         dout("sync\n");
4532         mutex_lock(&mdsc->mutex);
4533         want_tid = mdsc->last_tid;
4534         mutex_unlock(&mdsc->mutex);
4535
4536         ceph_flush_dirty_caps(mdsc);
4537         spin_lock(&mdsc->cap_dirty_lock);
4538         want_flush = mdsc->last_cap_flush_tid;
4539         if (!list_empty(&mdsc->cap_flush_list)) {
4540                 struct ceph_cap_flush *cf =
4541                         list_last_entry(&mdsc->cap_flush_list,
4542                                         struct ceph_cap_flush, g_list);
4543                 cf->wake = true;
4544         }
4545         spin_unlock(&mdsc->cap_dirty_lock);
4546
4547         dout("sync want tid %lld flush_seq %lld\n",
4548              want_tid, want_flush);
4549
4550         wait_unsafe_requests(mdsc, want_tid);
4551         wait_caps_flush(mdsc, want_flush);
4552 }
4553
4554 /*
4555  * true if all sessions are closed, or we force unmount
4556  */
4557 static bool done_closing_sessions(struct ceph_mds_client *mdsc, int skipped)
4558 {
4559         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
4560                 return true;
4561         return atomic_read(&mdsc->num_sessions) <= skipped;
4562 }
4563
4564 /*
4565  * called after sb is ro.
4566  */
4567 void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
4568 {
4569         struct ceph_options *opts = mdsc->fsc->client->options;
4570         struct ceph_mds_session *session;
4571         int i;
4572         int skipped = 0;
4573
4574         dout("close_sessions\n");
4575
4576         /* close sessions */
4577         mutex_lock(&mdsc->mutex);
4578         for (i = 0; i < mdsc->max_sessions; i++) {
4579                 session = __ceph_lookup_mds_session(mdsc, i);
4580                 if (!session)
4581                         continue;
4582                 mutex_unlock(&mdsc->mutex);
4583                 mutex_lock(&session->s_mutex);
4584                 if (__close_session(mdsc, session) <= 0)
4585                         skipped++;
4586                 mutex_unlock(&session->s_mutex);
4587                 ceph_put_mds_session(session);
4588                 mutex_lock(&mdsc->mutex);
4589         }
4590         mutex_unlock(&mdsc->mutex);
4591
4592         dout("waiting for sessions to close\n");
4593         wait_event_timeout(mdsc->session_close_wq,
4594                            done_closing_sessions(mdsc, skipped),
4595                            ceph_timeout_jiffies(opts->mount_timeout));
4596
4597         /* tear down remaining sessions */
4598         mutex_lock(&mdsc->mutex);
4599         for (i = 0; i < mdsc->max_sessions; i++) {
4600                 if (mdsc->sessions[i]) {
4601                         session = ceph_get_mds_session(mdsc->sessions[i]);
4602                         __unregister_session(mdsc, session);
4603                         mutex_unlock(&mdsc->mutex);
4604                         mutex_lock(&session->s_mutex);
4605                         remove_session_caps(session);
4606                         mutex_unlock(&session->s_mutex);
4607                         ceph_put_mds_session(session);
4608                         mutex_lock(&mdsc->mutex);
4609                 }
4610         }
4611         WARN_ON(!list_empty(&mdsc->cap_delay_list));
4612         mutex_unlock(&mdsc->mutex);
4613
4614         ceph_cleanup_snapid_map(mdsc);
4615         ceph_cleanup_empty_realms(mdsc);
4616
4617         cancel_work_sync(&mdsc->cap_reclaim_work);
4618         cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
4619
4620         dout("stopped\n");
4621 }
4622
4623 void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc)
4624 {
4625         struct ceph_mds_session *session;
4626         int mds;
4627
4628         dout("force umount\n");
4629
4630         mutex_lock(&mdsc->mutex);
4631         for (mds = 0; mds < mdsc->max_sessions; mds++) {
4632                 session = __ceph_lookup_mds_session(mdsc, mds);
4633                 if (!session)
4634                         continue;
4635
4636                 if (session->s_state == CEPH_MDS_SESSION_REJECTED)
4637                         __unregister_session(mdsc, session);
4638                 __wake_requests(mdsc, &session->s_waiting);
4639                 mutex_unlock(&mdsc->mutex);
4640
4641                 mutex_lock(&session->s_mutex);
4642                 __close_session(mdsc, session);
4643                 if (session->s_state == CEPH_MDS_SESSION_CLOSING) {
4644                         cleanup_session_requests(mdsc, session);
4645                         remove_session_caps(session);
4646                 }
4647                 mutex_unlock(&session->s_mutex);
4648                 ceph_put_mds_session(session);
4649
4650                 mutex_lock(&mdsc->mutex);
4651                 kick_requests(mdsc, mds);
4652         }
4653         __wake_requests(mdsc, &mdsc->waiting_for_map);
4654         mutex_unlock(&mdsc->mutex);
4655 }
4656
4657 static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
4658 {
4659         dout("stop\n");
4660         cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
4661         if (mdsc->mdsmap)
4662                 ceph_mdsmap_destroy(mdsc->mdsmap);
4663         kfree(mdsc->sessions);
4664         ceph_caps_finalize(mdsc);
4665         ceph_pool_perm_destroy(mdsc);
4666 }
4667
4668 void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
4669 {
4670         struct ceph_mds_client *mdsc = fsc->mdsc;
4671         dout("mdsc_destroy %p\n", mdsc);
4672
4673         if (!mdsc)
4674                 return;
4675
4676         /* flush out any connection work with references to us */
4677         ceph_msgr_flush();
4678
4679         ceph_mdsc_stop(mdsc);
4680
4681         ceph_metric_destroy(&mdsc->metric);
4682
4683         fsc->mdsc = NULL;
4684         kfree(mdsc);
4685         dout("mdsc_destroy %p done\n", mdsc);
4686 }
4687
4688 void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4689 {
4690         struct ceph_fs_client *fsc = mdsc->fsc;
4691         const char *mds_namespace = fsc->mount_options->mds_namespace;
4692         void *p = msg->front.iov_base;
4693         void *end = p + msg->front.iov_len;
4694         u32 epoch;
4695         u32 map_len;
4696         u32 num_fs;
4697         u32 mount_fscid = (u32)-1;
4698         u8 struct_v, struct_cv;
4699         int err = -EINVAL;
4700
4701         ceph_decode_need(&p, end, sizeof(u32), bad);
4702         epoch = ceph_decode_32(&p);
4703
4704         dout("handle_fsmap epoch %u\n", epoch);
4705
4706         ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
4707         struct_v = ceph_decode_8(&p);
4708         struct_cv = ceph_decode_8(&p);
4709         map_len = ceph_decode_32(&p);
4710
4711         ceph_decode_need(&p, end, sizeof(u32) * 3, bad);
4712         p += sizeof(u32) * 2; /* skip epoch and legacy_client_fscid */
4713
4714         num_fs = ceph_decode_32(&p);
4715         while (num_fs-- > 0) {
4716                 void *info_p, *info_end;
4717                 u32 info_len;
4718                 u8 info_v, info_cv;
4719                 u32 fscid, namelen;
4720
4721                 ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
4722                 info_v = ceph_decode_8(&p);
4723                 info_cv = ceph_decode_8(&p);
4724                 info_len = ceph_decode_32(&p);
4725                 ceph_decode_need(&p, end, info_len, bad);
4726                 info_p = p;
4727                 info_end = p + info_len;
4728                 p = info_end;
4729
4730                 ceph_decode_need(&info_p, info_end, sizeof(u32) * 2, bad);
4731                 fscid = ceph_decode_32(&info_p);
4732                 namelen = ceph_decode_32(&info_p);
4733                 ceph_decode_need(&info_p, info_end, namelen, bad);
4734
4735                 if (mds_namespace &&
4736                     strlen(mds_namespace) == namelen &&
4737                     !strncmp(mds_namespace, (char *)info_p, namelen)) {
4738                         mount_fscid = fscid;
4739                         break;
4740                 }
4741         }
4742
4743         ceph_monc_got_map(&fsc->client->monc, CEPH_SUB_FSMAP, epoch);
4744         if (mount_fscid != (u32)-1) {
4745                 fsc->client->monc.fs_cluster_id = mount_fscid;
4746                 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
4747                                    0, true);
4748                 ceph_monc_renew_subs(&fsc->client->monc);
4749         } else {
4750                 err = -ENOENT;
4751                 goto err_out;
4752         }
4753         return;
4754
4755 bad:
4756         pr_err("error decoding fsmap\n");
4757 err_out:
4758         mutex_lock(&mdsc->mutex);
4759         mdsc->mdsmap_err = err;
4760         __wake_requests(mdsc, &mdsc->waiting_for_map);
4761         mutex_unlock(&mdsc->mutex);
4762 }
4763
4764 /*
4765  * handle mds map update.
4766  */
4767 void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4768 {
4769         u32 epoch;
4770         u32 maplen;
4771         void *p = msg->front.iov_base;
4772         void *end = p + msg->front.iov_len;
4773         struct ceph_mdsmap *newmap, *oldmap;
4774         struct ceph_fsid fsid;
4775         int err = -EINVAL;
4776
4777         ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
4778         ceph_decode_copy(&p, &fsid, sizeof(fsid));
4779         if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)
4780                 return;
4781         epoch = ceph_decode_32(&p);
4782         maplen = ceph_decode_32(&p);
4783         dout("handle_map epoch %u len %d\n", epoch, (int)maplen);
4784
4785         /* do we need it? */
4786         mutex_lock(&mdsc->mutex);
4787         if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {
4788                 dout("handle_map epoch %u <= our %u\n",
4789                      epoch, mdsc->mdsmap->m_epoch);
4790                 mutex_unlock(&mdsc->mutex);
4791                 return;
4792         }
4793
4794         newmap = ceph_mdsmap_decode(&p, end);
4795         if (IS_ERR(newmap)) {
4796                 err = PTR_ERR(newmap);
4797                 goto bad_unlock;
4798         }
4799
4800         /* swap into place */
4801         if (mdsc->mdsmap) {
4802                 oldmap = mdsc->mdsmap;
4803                 mdsc->mdsmap = newmap;
4804                 check_new_map(mdsc, newmap, oldmap);
4805                 ceph_mdsmap_destroy(oldmap);
4806         } else {
4807                 mdsc->mdsmap = newmap;  /* first mds map */
4808         }
4809         mdsc->fsc->max_file_size = min((loff_t)mdsc->mdsmap->m_max_file_size,
4810                                         MAX_LFS_FILESIZE);
4811
4812         __wake_requests(mdsc, &mdsc->waiting_for_map);
4813         ceph_monc_got_map(&mdsc->fsc->client->monc, CEPH_SUB_MDSMAP,
4814                           mdsc->mdsmap->m_epoch);
4815
4816         mutex_unlock(&mdsc->mutex);
4817         schedule_delayed(mdsc);
4818         return;
4819
4820 bad_unlock:
4821         mutex_unlock(&mdsc->mutex);
4822 bad:
4823         pr_err("error decoding mdsmap %d\n", err);
4824         return;
4825 }
4826
4827 static struct ceph_connection *con_get(struct ceph_connection *con)
4828 {
4829         struct ceph_mds_session *s = con->private;
4830
4831         if (ceph_get_mds_session(s))
4832                 return con;
4833         return NULL;
4834 }
4835
4836 static void con_put(struct ceph_connection *con)
4837 {
4838         struct ceph_mds_session *s = con->private;
4839
4840         ceph_put_mds_session(s);
4841 }
4842
4843 /*
4844  * if the client is unresponsive for long enough, the mds will kill
4845  * the session entirely.
4846  */
4847 static void peer_reset(struct ceph_connection *con)
4848 {
4849         struct ceph_mds_session *s = con->private;
4850         struct ceph_mds_client *mdsc = s->s_mdsc;
4851
4852         pr_warn("mds%d closed our session\n", s->s_mds);
4853         send_mds_reconnect(mdsc, s);
4854 }
4855
4856 static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
4857 {
4858         struct ceph_mds_session *s = con->private;
4859         struct ceph_mds_client *mdsc = s->s_mdsc;
4860         int type = le16_to_cpu(msg->hdr.type);
4861
4862         mutex_lock(&mdsc->mutex);
4863         if (__verify_registered_session(mdsc, s) < 0) {
4864                 mutex_unlock(&mdsc->mutex);
4865                 goto out;
4866         }
4867         mutex_unlock(&mdsc->mutex);
4868
4869         switch (type) {
4870         case CEPH_MSG_MDS_MAP:
4871                 ceph_mdsc_handle_mdsmap(mdsc, msg);
4872                 break;
4873         case CEPH_MSG_FS_MAP_USER:
4874                 ceph_mdsc_handle_fsmap(mdsc, msg);
4875                 break;
4876         case CEPH_MSG_CLIENT_SESSION:
4877                 handle_session(s, msg);
4878                 break;
4879         case CEPH_MSG_CLIENT_REPLY:
4880                 handle_reply(s, msg);
4881                 break;
4882         case CEPH_MSG_CLIENT_REQUEST_FORWARD:
4883                 handle_forward(mdsc, s, msg);
4884                 break;
4885         case CEPH_MSG_CLIENT_CAPS:
4886                 ceph_handle_caps(s, msg);
4887                 break;
4888         case CEPH_MSG_CLIENT_SNAP:
4889                 ceph_handle_snap(mdsc, s, msg);
4890                 break;
4891         case CEPH_MSG_CLIENT_LEASE:
4892                 handle_lease(mdsc, s, msg);
4893                 break;
4894         case CEPH_MSG_CLIENT_QUOTA:
4895                 ceph_handle_quota(mdsc, s, msg);
4896                 break;
4897
4898         default:
4899                 pr_err("received unknown message type %d %s\n", type,
4900                        ceph_msg_type_name(type));
4901         }
4902 out:
4903         ceph_msg_put(msg);
4904 }
4905
4906 /*
4907  * authentication
4908  */
4909
4910 /*
4911  * Note: returned pointer is the address of a structure that's
4912  * managed separately.  Caller must *not* attempt to free it.
4913  */
4914 static struct ceph_auth_handshake *get_authorizer(struct ceph_connection *con,
4915                                         int *proto, int force_new)
4916 {
4917         struct ceph_mds_session *s = con->private;
4918         struct ceph_mds_client *mdsc = s->s_mdsc;
4919         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
4920         struct ceph_auth_handshake *auth = &s->s_auth;
4921
4922         if (force_new && auth->authorizer) {
4923                 ceph_auth_destroy_authorizer(auth->authorizer);
4924                 auth->authorizer = NULL;
4925         }
4926         if (!auth->authorizer) {
4927                 int ret = ceph_auth_create_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
4928                                                       auth);
4929                 if (ret)
4930                         return ERR_PTR(ret);
4931         } else {
4932                 int ret = ceph_auth_update_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
4933                                                       auth);
4934                 if (ret)
4935                         return ERR_PTR(ret);
4936         }
4937         *proto = ac->protocol;
4938
4939         return auth;
4940 }
4941
4942 static int add_authorizer_challenge(struct ceph_connection *con,
4943                                     void *challenge_buf, int challenge_buf_len)
4944 {
4945         struct ceph_mds_session *s = con->private;
4946         struct ceph_mds_client *mdsc = s->s_mdsc;
4947         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
4948
4949         return ceph_auth_add_authorizer_challenge(ac, s->s_auth.authorizer,
4950                                             challenge_buf, challenge_buf_len);
4951 }
4952
4953 static int verify_authorizer_reply(struct ceph_connection *con)
4954 {
4955         struct ceph_mds_session *s = con->private;
4956         struct ceph_mds_client *mdsc = s->s_mdsc;
4957         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
4958
4959         return ceph_auth_verify_authorizer_reply(ac, s->s_auth.authorizer);
4960 }
4961
4962 static int invalidate_authorizer(struct ceph_connection *con)
4963 {
4964         struct ceph_mds_session *s = con->private;
4965         struct ceph_mds_client *mdsc = s->s_mdsc;
4966         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
4967
4968         ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_MDS);
4969
4970         return ceph_monc_validate_auth(&mdsc->fsc->client->monc);
4971 }
4972
4973 static struct ceph_msg *mds_alloc_msg(struct ceph_connection *con,
4974                                 struct ceph_msg_header *hdr, int *skip)
4975 {
4976         struct ceph_msg *msg;
4977         int type = (int) le16_to_cpu(hdr->type);
4978         int front_len = (int) le32_to_cpu(hdr->front_len);
4979
4980         if (con->in_msg)
4981                 return con->in_msg;
4982
4983         *skip = 0;
4984         msg = ceph_msg_new(type, front_len, GFP_NOFS, false);
4985         if (!msg) {
4986                 pr_err("unable to allocate msg type %d len %d\n",
4987                        type, front_len);
4988                 return NULL;
4989         }
4990
4991         return msg;
4992 }
4993
4994 static int mds_sign_message(struct ceph_msg *msg)
4995 {
4996        struct ceph_mds_session *s = msg->con->private;
4997        struct ceph_auth_handshake *auth = &s->s_auth;
4998
4999        return ceph_auth_sign_message(auth, msg);
5000 }
5001
5002 static int mds_check_message_signature(struct ceph_msg *msg)
5003 {
5004        struct ceph_mds_session *s = msg->con->private;
5005        struct ceph_auth_handshake *auth = &s->s_auth;
5006
5007        return ceph_auth_check_message_signature(auth, msg);
5008 }
5009
5010 static const struct ceph_connection_operations mds_con_ops = {
5011         .get = con_get,
5012         .put = con_put,
5013         .dispatch = dispatch,
5014         .get_authorizer = get_authorizer,
5015         .add_authorizer_challenge = add_authorizer_challenge,
5016         .verify_authorizer_reply = verify_authorizer_reply,
5017         .invalidate_authorizer = invalidate_authorizer,
5018         .peer_reset = peer_reset,
5019         .alloc_msg = mds_alloc_msg,
5020         .sign_message = mds_sign_message,
5021         .check_message_signature = mds_check_message_signature,
5022 };
5023
5024 /* eof */