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  49  * are no longer valid.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 (!ceph_mdsmap_is_laggy(mdsc->mdsmap,
1107                                                                   mds))
1108                                                 goto out;
1109                                 }
1110                         }
1111                         mode = USE_AUTH_MDS;
1112                 }
1113         }
1114
1115         spin_lock(&ci->i_ceph_lock);
1116         cap = NULL;
1117         if (mode == USE_AUTH_MDS)
1118                 cap = ci->i_auth_cap;
1119         if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
1120                 cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
1121         if (!cap) {
1122                 spin_unlock(&ci->i_ceph_lock);
1123                 ceph_async_iput(inode);
1124                 goto random;
1125         }
1126         mds = cap->session->s_mds;
1127         dout("%s %p %llx.%llx mds%d (%scap %p)\n", __func__,
1128              inode, ceph_vinop(inode), mds,
1129              cap == ci->i_auth_cap ? "auth " : "", cap);
1130         spin_unlock(&ci->i_ceph_lock);
1131 out:
1132         /* avoid calling iput_final() while holding mdsc->mutex or
1133          * in mds dispatch threads */
1134         ceph_async_iput(inode);
1135         return mds;
1136
1137 random:
1138         if (random)
1139                 *random = true;
1140
1141         mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);
1142         dout("%s chose random mds%d\n", __func__, mds);
1143         return mds;
1144 }
1145
1146
1147 /*
1148  * session messages
1149  */
1150 static struct ceph_msg *create_session_msg(u32 op, u64 seq)
1151 {
1152         struct ceph_msg *msg;
1153         struct ceph_mds_session_head *h;
1154
1155         msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h), GFP_NOFS,
1156                            false);
1157         if (!msg) {
1158                 pr_err("create_session_msg ENOMEM creating msg\n");
1159                 return NULL;
1160         }
1161         h = msg->front.iov_base;
1162         h->op = cpu_to_le32(op);
1163         h->seq = cpu_to_le64(seq);
1164
1165         return msg;
1166 }
1167
1168 static const unsigned char feature_bits[] = CEPHFS_FEATURES_CLIENT_SUPPORTED;
1169 #define FEATURE_BYTES(c) (DIV_ROUND_UP((size_t)feature_bits[c - 1] + 1, 64) * 8)
1170 static int encode_supported_features(void **p, void *end)
1171 {
1172         static const size_t count = ARRAY_SIZE(feature_bits);
1173
1174         if (count > 0) {
1175                 size_t i;
1176                 size_t size = FEATURE_BYTES(count);
1177
1178                 if (WARN_ON_ONCE(*p + 4 + size > end))
1179                         return -ERANGE;
1180
1181                 ceph_encode_32(p, size);
1182                 memset(*p, 0, size);
1183                 for (i = 0; i < count; i++)
1184                         ((unsigned char*)(*p))[i / 8] |= BIT(feature_bits[i] % 8);
1185                 *p += size;
1186         } else {
1187                 if (WARN_ON_ONCE(*p + 4 > end))
1188                         return -ERANGE;
1189
1190                 ceph_encode_32(p, 0);
1191         }
1192
1193         return 0;
1194 }
1195
1196 static const unsigned char metric_bits[] = CEPHFS_METRIC_SPEC_CLIENT_SUPPORTED;
1197 #define METRIC_BYTES(cnt) (DIV_ROUND_UP((size_t)metric_bits[cnt - 1] + 1, 64) * 8)
1198 static int encode_metric_spec(void **p, void *end)
1199 {
1200         static const size_t count = ARRAY_SIZE(metric_bits);
1201
1202         /* header */
1203         if (WARN_ON_ONCE(*p + 2 > end))
1204                 return -ERANGE;
1205
1206         ceph_encode_8(p, 1); /* version */
1207         ceph_encode_8(p, 1); /* compat */
1208
1209         if (count > 0) {
1210                 size_t i;
1211                 size_t size = METRIC_BYTES(count);
1212
1213                 if (WARN_ON_ONCE(*p + 4 + 4 + size > end))
1214                         return -ERANGE;
1215
1216                 /* metric spec info length */
1217                 ceph_encode_32(p, 4 + size);
1218
1219                 /* metric spec */
1220                 ceph_encode_32(p, size);
1221                 memset(*p, 0, size);
1222                 for (i = 0; i < count; i++)
1223                         ((unsigned char *)(*p))[i / 8] |= BIT(metric_bits[i] % 8);
1224                 *p += size;
1225         } else {
1226                 if (WARN_ON_ONCE(*p + 4 + 4 > end))
1227                         return -ERANGE;
1228
1229                 /* metric spec info length */
1230                 ceph_encode_32(p, 4);
1231                 /* metric spec */
1232                 ceph_encode_32(p, 0);
1233         }
1234
1235         return 0;
1236 }
1237
1238 /*
1239  * session message, specialization for CEPH_SESSION_REQUEST_OPEN
1240  * to include additional client metadata fields.
1241  */
1242 static struct ceph_msg *create_session_open_msg(struct ceph_mds_client *mdsc, u64 seq)
1243 {
1244         struct ceph_msg *msg;
1245         struct ceph_mds_session_head *h;
1246         int i = -1;
1247         int extra_bytes = 0;
1248         int metadata_key_count = 0;
1249         struct ceph_options *opt = mdsc->fsc->client->options;
1250         struct ceph_mount_options *fsopt = mdsc->fsc->mount_options;
1251         size_t size, count;
1252         void *p, *end;
1253         int ret;
1254
1255         const char* metadata[][2] = {
1256                 {"hostname", mdsc->nodename},
1257                 {"kernel_version", init_utsname()->release},
1258                 {"entity_id", opt->name ? : ""},
1259                 {"root", fsopt->server_path ? : "/"},
1260                 {NULL, NULL}
1261         };
1262
1263         /* Calculate serialized length of metadata */
1264         extra_bytes = 4;  /* map length */
1265         for (i = 0; metadata[i][0]; ++i) {
1266                 extra_bytes += 8 + strlen(metadata[i][0]) +
1267                         strlen(metadata[i][1]);
1268                 metadata_key_count++;
1269         }
1270
1271         /* supported feature */
1272         size = 0;
1273         count = ARRAY_SIZE(feature_bits);
1274         if (count > 0)
1275                 size = FEATURE_BYTES(count);
1276         extra_bytes += 4 + size;
1277
1278         /* metric spec */
1279         size = 0;
1280         count = ARRAY_SIZE(metric_bits);
1281         if (count > 0)
1282                 size = METRIC_BYTES(count);
1283         extra_bytes += 2 + 4 + 4 + size;
1284
1285         /* Allocate the message */
1286         msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h) + extra_bytes,
1287                            GFP_NOFS, false);
1288         if (!msg) {
1289                 pr_err("create_session_msg ENOMEM creating msg\n");
1290                 return ERR_PTR(-ENOMEM);
1291         }
1292         p = msg->front.iov_base;
1293         end = p + msg->front.iov_len;
1294
1295         h = p;
1296         h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);
1297         h->seq = cpu_to_le64(seq);
1298
1299         /*
1300          * Serialize client metadata into waiting buffer space, using
1301          * the format that userspace expects for map<string, string>
1302          *
1303          * ClientSession messages with metadata are v4
1304          */
1305         msg->hdr.version = cpu_to_le16(4);
1306         msg->hdr.compat_version = cpu_to_le16(1);
1307
1308         /* The write pointer, following the session_head structure */
1309         p += sizeof(*h);
1310
1311         /* Number of entries in the map */
1312         ceph_encode_32(&p, metadata_key_count);
1313
1314         /* Two length-prefixed strings for each entry in the map */
1315         for (i = 0; metadata[i][0]; ++i) {
1316                 size_t const key_len = strlen(metadata[i][0]);
1317                 size_t const val_len = strlen(metadata[i][1]);
1318
1319                 ceph_encode_32(&p, key_len);
1320                 memcpy(p, metadata[i][0], key_len);
1321                 p += key_len;
1322                 ceph_encode_32(&p, val_len);
1323                 memcpy(p, metadata[i][1], val_len);
1324                 p += val_len;
1325         }
1326
1327         ret = encode_supported_features(&p, end);
1328         if (ret) {
1329                 pr_err("encode_supported_features failed!\n");
1330                 ceph_msg_put(msg);
1331                 return ERR_PTR(ret);
1332         }
1333
1334         ret = encode_metric_spec(&p, end);
1335         if (ret) {
1336                 pr_err("encode_metric_spec failed!\n");
1337                 ceph_msg_put(msg);
1338                 return ERR_PTR(ret);
1339         }
1340
1341         msg->front.iov_len = p - msg->front.iov_base;
1342         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1343
1344         return msg;
1345 }
1346
1347 /*
1348  * send session open request.
1349  *
1350  * called under mdsc->mutex
1351  */
1352 static int __open_session(struct ceph_mds_client *mdsc,
1353                           struct ceph_mds_session *session)
1354 {
1355         struct ceph_msg *msg;
1356         int mstate;
1357         int mds = session->s_mds;
1358
1359         /* wait for mds to go active? */
1360         mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
1361         dout("open_session to mds%d (%s)\n", mds,
1362              ceph_mds_state_name(mstate));
1363         session->s_state = CEPH_MDS_SESSION_OPENING;
1364         session->s_renew_requested = jiffies;
1365
1366         /* send connect message */
1367         msg = create_session_open_msg(mdsc, session->s_seq);
1368         if (IS_ERR(msg))
1369                 return PTR_ERR(msg);
1370         ceph_con_send(&session->s_con, msg);
1371         return 0;
1372 }
1373
1374 /*
1375  * open sessions for any export targets for the given mds
1376  *
1377  * called under mdsc->mutex
1378  */
1379 static struct ceph_mds_session *
1380 __open_export_target_session(struct ceph_mds_client *mdsc, int target)
1381 {
1382         struct ceph_mds_session *session;
1383         int ret;
1384
1385         session = __ceph_lookup_mds_session(mdsc, target);
1386         if (!session) {
1387                 session = register_session(mdsc, target);
1388                 if (IS_ERR(session))
1389                         return session;
1390         }
1391         if (session->s_state == CEPH_MDS_SESSION_NEW ||
1392             session->s_state == CEPH_MDS_SESSION_CLOSING) {
1393                 ret = __open_session(mdsc, session);
1394                 if (ret)
1395                         return ERR_PTR(ret);
1396         }
1397
1398         return session;
1399 }
1400
1401 struct ceph_mds_session *
1402 ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target)
1403 {
1404         struct ceph_mds_session *session;
1405
1406         dout("open_export_target_session to mds%d\n", target);
1407
1408         mutex_lock(&mdsc->mutex);
1409         session = __open_export_target_session(mdsc, target);
1410         mutex_unlock(&mdsc->mutex);
1411
1412         return session;
1413 }
1414
1415 static void __open_export_target_sessions(struct ceph_mds_client *mdsc,
1416                                           struct ceph_mds_session *session)
1417 {
1418         struct ceph_mds_info *mi;
1419         struct ceph_mds_session *ts;
1420         int i, mds = session->s_mds;
1421
1422         if (mds >= mdsc->mdsmap->possible_max_rank)
1423                 return;
1424
1425         mi = &mdsc->mdsmap->m_info[mds];
1426         dout("open_export_target_sessions for mds%d (%d targets)\n",
1427              session->s_mds, mi->num_export_targets);
1428
1429         for (i = 0; i < mi->num_export_targets; i++) {
1430                 ts = __open_export_target_session(mdsc, mi->export_targets[i]);
1431                 if (!IS_ERR(ts))
1432                         ceph_put_mds_session(ts);
1433         }
1434 }
1435
1436 void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
1437                                            struct ceph_mds_session *session)
1438 {
1439         mutex_lock(&mdsc->mutex);
1440         __open_export_target_sessions(mdsc, session);
1441         mutex_unlock(&mdsc->mutex);
1442 }
1443
1444 /*
1445  * session caps
1446  */
1447
1448 static void detach_cap_releases(struct ceph_mds_session *session,
1449                                 struct list_head *target)
1450 {
1451         lockdep_assert_held(&session->s_cap_lock);
1452
1453         list_splice_init(&session->s_cap_releases, target);
1454         session->s_num_cap_releases = 0;
1455         dout("dispose_cap_releases mds%d\n", session->s_mds);
1456 }
1457
1458 static void dispose_cap_releases(struct ceph_mds_client *mdsc,
1459                                  struct list_head *dispose)
1460 {
1461         while (!list_empty(dispose)) {
1462                 struct ceph_cap *cap;
1463                 /* zero out the in-progress message */
1464                 cap = list_first_entry(dispose, struct ceph_cap, session_caps);
1465                 list_del(&cap->session_caps);
1466                 ceph_put_cap(mdsc, cap);
1467         }
1468 }
1469
1470 static void cleanup_session_requests(struct ceph_mds_client *mdsc,
1471                                      struct ceph_mds_session *session)
1472 {
1473         struct ceph_mds_request *req;
1474         struct rb_node *p;
1475         struct ceph_inode_info *ci;
1476
1477         dout("cleanup_session_requests mds%d\n", session->s_mds);
1478         mutex_lock(&mdsc->mutex);
1479         while (!list_empty(&session->s_unsafe)) {
1480                 req = list_first_entry(&session->s_unsafe,
1481                                        struct ceph_mds_request, r_unsafe_item);
1482                 pr_warn_ratelimited(" dropping unsafe request %llu\n",
1483                                     req->r_tid);
1484                 if (req->r_target_inode) {
1485                         /* dropping unsafe change of inode's attributes */
1486                         ci = ceph_inode(req->r_target_inode);
1487                         errseq_set(&ci->i_meta_err, -EIO);
1488                 }
1489                 if (req->r_unsafe_dir) {
1490                         /* dropping unsafe directory operation */
1491                         ci = ceph_inode(req->r_unsafe_dir);
1492                         errseq_set(&ci->i_meta_err, -EIO);
1493                 }
1494                 __unregister_request(mdsc, req);
1495         }
1496         /* zero r_attempts, so kick_requests() will re-send requests */
1497         p = rb_first(&mdsc->request_tree);
1498         while (p) {
1499                 req = rb_entry(p, struct ceph_mds_request, r_node);
1500                 p = rb_next(p);
1501                 if (req->r_session &&
1502                     req->r_session->s_mds == session->s_mds)
1503                         req->r_attempts = 0;
1504         }
1505         mutex_unlock(&mdsc->mutex);
1506 }
1507
1508 /*
1509  * Helper to safely iterate over all caps associated with a session, with
1510  * special care taken to handle a racing __ceph_remove_cap().
1511  *
1512  * Caller must hold session s_mutex.
1513  */
1514 int ceph_iterate_session_caps(struct ceph_mds_session *session,
1515                               int (*cb)(struct inode *, struct ceph_cap *,
1516                                         void *), void *arg)
1517 {
1518         struct list_head *p;
1519         struct ceph_cap *cap;
1520         struct inode *inode, *last_inode = NULL;
1521         struct ceph_cap *old_cap = NULL;
1522         int ret;
1523
1524         dout("iterate_session_caps %p mds%d\n", session, session->s_mds);
1525         spin_lock(&session->s_cap_lock);
1526         p = session->s_caps.next;
1527         while (p != &session->s_caps) {
1528                 cap = list_entry(p, struct ceph_cap, session_caps);
1529                 inode = igrab(&cap->ci->vfs_inode);
1530                 if (!inode) {
1531                         p = p->next;
1532                         continue;
1533                 }
1534                 session->s_cap_iterator = cap;
1535                 spin_unlock(&session->s_cap_lock);
1536
1537                 if (last_inode) {
1538                         /* avoid calling iput_final() while holding
1539                          * s_mutex or in mds dispatch threads */
1540                         ceph_async_iput(last_inode);
1541                         last_inode = NULL;
1542                 }
1543                 if (old_cap) {
1544                         ceph_put_cap(session->s_mdsc, old_cap);
1545                         old_cap = NULL;
1546                 }
1547
1548                 ret = cb(inode, cap, arg);
1549                 last_inode = inode;
1550
1551                 spin_lock(&session->s_cap_lock);
1552                 p = p->next;
1553                 if (!cap->ci) {
1554                         dout("iterate_session_caps  finishing cap %p removal\n",
1555                              cap);
1556                         BUG_ON(cap->session != session);
1557                         cap->session = NULL;
1558                         list_del_init(&cap->session_caps);
1559                         session->s_nr_caps--;
1560                         atomic64_dec(&session->s_mdsc->metric.total_caps);
1561                         if (cap->queue_release)
1562                                 __ceph_queue_cap_release(session, cap);
1563                         else
1564                                 old_cap = cap;  /* put_cap it w/o locks held */
1565                 }
1566                 if (ret < 0)
1567                         goto out;
1568         }
1569         ret = 0;
1570 out:
1571         session->s_cap_iterator = NULL;
1572         spin_unlock(&session->s_cap_lock);
1573
1574         ceph_async_iput(last_inode);
1575         if (old_cap)
1576                 ceph_put_cap(session->s_mdsc, old_cap);
1577
1578         return ret;
1579 }
1580
1581 static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
1582                                   void *arg)
1583 {
1584         struct ceph_fs_client *fsc = (struct ceph_fs_client *)arg;
1585         struct ceph_inode_info *ci = ceph_inode(inode);
1586         LIST_HEAD(to_remove);
1587         bool dirty_dropped = false;
1588         bool invalidate = false;
1589
1590         dout("removing cap %p, ci is %p, inode is %p\n",
1591              cap, ci, &ci->vfs_inode);
1592         spin_lock(&ci->i_ceph_lock);
1593         __ceph_remove_cap(cap, false);
1594         if (!ci->i_auth_cap) {
1595                 struct ceph_cap_flush *cf;
1596                 struct ceph_mds_client *mdsc = fsc->mdsc;
1597
1598                 if (READ_ONCE(fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
1599                         if (inode->i_data.nrpages > 0)
1600                                 invalidate = true;
1601                         if (ci->i_wrbuffer_ref > 0)
1602                                 mapping_set_error(&inode->i_data, -EIO);
1603                 }
1604
1605                 while (!list_empty(&ci->i_cap_flush_list)) {
1606                         cf = list_first_entry(&ci->i_cap_flush_list,
1607                                               struct ceph_cap_flush, i_list);
1608                         list_move(&cf->i_list, &to_remove);
1609                 }
1610
1611                 spin_lock(&mdsc->cap_dirty_lock);
1612
1613                 list_for_each_entry(cf, &to_remove, i_list)
1614                         list_del(&cf->g_list);
1615
1616                 if (!list_empty(&ci->i_dirty_item)) {
1617                         pr_warn_ratelimited(
1618                                 " dropping dirty %s state for %p %lld\n",
1619                                 ceph_cap_string(ci->i_dirty_caps),
1620                                 inode, ceph_ino(inode));
1621                         ci->i_dirty_caps = 0;
1622                         list_del_init(&ci->i_dirty_item);
1623                         dirty_dropped = true;
1624                 }
1625                 if (!list_empty(&ci->i_flushing_item)) {
1626                         pr_warn_ratelimited(
1627                                 " dropping dirty+flushing %s state for %p %lld\n",
1628                                 ceph_cap_string(ci->i_flushing_caps),
1629                                 inode, ceph_ino(inode));
1630                         ci->i_flushing_caps = 0;
1631                         list_del_init(&ci->i_flushing_item);
1632                         mdsc->num_cap_flushing--;
1633                         dirty_dropped = true;
1634                 }
1635                 spin_unlock(&mdsc->cap_dirty_lock);
1636
1637                 if (dirty_dropped) {
1638                         errseq_set(&ci->i_meta_err, -EIO);
1639
1640                         if (ci->i_wrbuffer_ref_head == 0 &&
1641                             ci->i_wr_ref == 0 &&
1642                             ci->i_dirty_caps == 0 &&
1643                             ci->i_flushing_caps == 0) {
1644                                 ceph_put_snap_context(ci->i_head_snapc);
1645                                 ci->i_head_snapc = NULL;
1646                         }
1647                 }
1648
1649                 if (atomic_read(&ci->i_filelock_ref) > 0) {
1650                         /* make further file lock syscall return -EIO */
1651                         ci->i_ceph_flags |= CEPH_I_ERROR_FILELOCK;
1652                         pr_warn_ratelimited(" dropping file locks for %p %lld\n",
1653                                             inode, ceph_ino(inode));
1654                 }
1655
1656                 if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {
1657                         list_add(&ci->i_prealloc_cap_flush->i_list, &to_remove);
1658                         ci->i_prealloc_cap_flush = NULL;
1659                 }
1660         }
1661         spin_unlock(&ci->i_ceph_lock);
1662         while (!list_empty(&to_remove)) {
1663                 struct ceph_cap_flush *cf;
1664                 cf = list_first_entry(&to_remove,
1665                                       struct ceph_cap_flush, i_list);
1666                 list_del(&cf->i_list);
1667                 ceph_free_cap_flush(cf);
1668         }
1669
1670         wake_up_all(&ci->i_cap_wq);
1671         if (invalidate)
1672                 ceph_queue_invalidate(inode);
1673         if (dirty_dropped)
1674                 iput(inode);
1675         return 0;
1676 }
1677
1678 /*
1679  * caller must hold session s_mutex
1680  */
1681 static void remove_session_caps(struct ceph_mds_session *session)
1682 {
1683         struct ceph_fs_client *fsc = session->s_mdsc->fsc;
1684         struct super_block *sb = fsc->sb;
1685         LIST_HEAD(dispose);
1686
1687         dout("remove_session_caps on %p\n", session);
1688         ceph_iterate_session_caps(session, remove_session_caps_cb, fsc);
1689
1690         wake_up_all(&fsc->mdsc->cap_flushing_wq);
1691
1692         spin_lock(&session->s_cap_lock);
1693         if (session->s_nr_caps > 0) {
1694                 struct inode *inode;
1695                 struct ceph_cap *cap, *prev = NULL;
1696                 struct ceph_vino vino;
1697                 /*
1698                  * iterate_session_caps() skips inodes that are being
1699                  * deleted, we need to wait until deletions are complete.
1700                  * __wait_on_freeing_inode() is designed for the job,
1701                  * but it is not exported, so use lookup inode function
1702                  * to access it.
1703                  */
1704                 while (!list_empty(&session->s_caps)) {
1705                         cap = list_entry(session->s_caps.next,
1706                                          struct ceph_cap, session_caps);
1707                         if (cap == prev)
1708                                 break;
1709                         prev = cap;
1710                         vino = cap->ci->i_vino;
1711                         spin_unlock(&session->s_cap_lock);
1712
1713                         inode = ceph_find_inode(sb, vino);
1714                          /* avoid calling iput_final() while holding s_mutex */
1715                         ceph_async_iput(inode);
1716
1717                         spin_lock(&session->s_cap_lock);
1718                 }
1719         }
1720
1721         // drop cap expires and unlock s_cap_lock
1722         detach_cap_releases(session, &dispose);
1723
1724         BUG_ON(session->s_nr_caps > 0);
1725         BUG_ON(!list_empty(&session->s_cap_flushing));
1726         spin_unlock(&session->s_cap_lock);
1727         dispose_cap_releases(session->s_mdsc, &dispose);
1728 }
1729
1730 enum {
1731         RECONNECT,
1732         RENEWCAPS,
1733         FORCE_RO,
1734 };
1735
1736 /*
1737  * wake up any threads waiting on this session's caps.  if the cap is
1738  * old (didn't get renewed on the client reconnect), remove it now.
1739  *
1740  * caller must hold s_mutex.
1741  */
1742 static int wake_up_session_cb(struct inode *inode, struct ceph_cap *cap,
1743                               void *arg)
1744 {
1745         struct ceph_inode_info *ci = ceph_inode(inode);
1746         unsigned long ev = (unsigned long)arg;
1747
1748         if (ev == RECONNECT) {
1749                 spin_lock(&ci->i_ceph_lock);
1750                 ci->i_wanted_max_size = 0;
1751                 ci->i_requested_max_size = 0;
1752                 spin_unlock(&ci->i_ceph_lock);
1753         } else if (ev == RENEWCAPS) {
1754                 if (cap->cap_gen < cap->session->s_cap_gen) {
1755                         /* mds did not re-issue stale cap */
1756                         spin_lock(&ci->i_ceph_lock);
1757                         cap->issued = cap->implemented = CEPH_CAP_PIN;
1758                         spin_unlock(&ci->i_ceph_lock);
1759                 }
1760         } else if (ev == FORCE_RO) {
1761         }
1762         wake_up_all(&ci->i_cap_wq);
1763         return 0;
1764 }
1765
1766 static void wake_up_session_caps(struct ceph_mds_session *session, int ev)
1767 {
1768         dout("wake_up_session_caps %p mds%d\n", session, session->s_mds);
1769         ceph_iterate_session_caps(session, wake_up_session_cb,
1770                                   (void *)(unsigned long)ev);
1771 }
1772
1773 /*
1774  * Send periodic message to MDS renewing all currently held caps.  The
1775  * ack will reset the expiration for all caps from this session.
1776  *
1777  * caller holds s_mutex
1778  */
1779 static int send_renew_caps(struct ceph_mds_client *mdsc,
1780                            struct ceph_mds_session *session)
1781 {
1782         struct ceph_msg *msg;
1783         int state;
1784
1785         if (time_after_eq(jiffies, session->s_cap_ttl) &&
1786             time_after_eq(session->s_cap_ttl, session->s_renew_requested))
1787                 pr_info("mds%d caps stale\n", session->s_mds);
1788         session->s_renew_requested = jiffies;
1789
1790         /* do not try to renew caps until a recovering mds has reconnected
1791          * with its clients. */
1792         state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);
1793         if (state < CEPH_MDS_STATE_RECONNECT) {
1794                 dout("send_renew_caps ignoring mds%d (%s)\n",
1795                      session->s_mds, ceph_mds_state_name(state));
1796                 return 0;
1797         }
1798
1799         dout("send_renew_caps to mds%d (%s)\n", session->s_mds,
1800                 ceph_mds_state_name(state));
1801         msg = create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
1802                                  ++session->s_renew_seq);
1803         if (!msg)
1804                 return -ENOMEM;
1805         ceph_con_send(&session->s_con, msg);
1806         return 0;
1807 }
1808
1809 static int send_flushmsg_ack(struct ceph_mds_client *mdsc,
1810                              struct ceph_mds_session *session, u64 seq)
1811 {
1812         struct ceph_msg *msg;
1813
1814         dout("send_flushmsg_ack to mds%d (%s)s seq %lld\n",
1815              session->s_mds, ceph_session_state_name(session->s_state), seq);
1816         msg = create_session_msg(CEPH_SESSION_FLUSHMSG_ACK, seq);
1817         if (!msg)
1818                 return -ENOMEM;
1819         ceph_con_send(&session->s_con, msg);
1820         return 0;
1821 }
1822
1823
1824 /*
1825  * Note new cap ttl, and any transition from stale -> not stale (fresh?).
1826  *
1827  * Called under session->s_mutex
1828  */
1829 static void renewed_caps(struct ceph_mds_client *mdsc,
1830                          struct ceph_mds_session *session, int is_renew)
1831 {
1832         int was_stale;
1833         int wake = 0;
1834
1835         spin_lock(&session->s_cap_lock);
1836         was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);
1837
1838         session->s_cap_ttl = session->s_renew_requested +
1839                 mdsc->mdsmap->m_session_timeout*HZ;
1840
1841         if (was_stale) {
1842                 if (time_before(jiffies, session->s_cap_ttl)) {
1843                         pr_info("mds%d caps renewed\n", session->s_mds);
1844                         wake = 1;
1845                 } else {
1846                         pr_info("mds%d caps still stale\n", session->s_mds);
1847                 }
1848         }
1849         dout("renewed_caps mds%d ttl now %lu, was %s, now %s\n",
1850              session->s_mds, session->s_cap_ttl, was_stale ? "stale" : "fresh",
1851              time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");
1852         spin_unlock(&session->s_cap_lock);
1853
1854         if (wake)
1855                 wake_up_session_caps(session, RENEWCAPS);
1856 }
1857
1858 /*
1859  * send a session close request
1860  */
1861 static int request_close_session(struct ceph_mds_session *session)
1862 {
1863         struct ceph_msg *msg;
1864
1865         dout("request_close_session mds%d state %s seq %lld\n",
1866              session->s_mds, ceph_session_state_name(session->s_state),
1867              session->s_seq);
1868         msg = create_session_msg(CEPH_SESSION_REQUEST_CLOSE, session->s_seq);
1869         if (!msg)
1870                 return -ENOMEM;
1871         ceph_con_send(&session->s_con, msg);
1872         return 1;
1873 }
1874
1875 /*
1876  * Called with s_mutex held.
1877  */
1878 static int __close_session(struct ceph_mds_client *mdsc,
1879                          struct ceph_mds_session *session)
1880 {
1881         if (session->s_state >= CEPH_MDS_SESSION_CLOSING)
1882                 return 0;
1883         session->s_state = CEPH_MDS_SESSION_CLOSING;
1884         return request_close_session(session);
1885 }
1886
1887 static bool drop_negative_children(struct dentry *dentry)
1888 {
1889         struct dentry *child;
1890         bool all_negative = true;
1891
1892         if (!d_is_dir(dentry))
1893                 goto out;
1894
1895         spin_lock(&dentry->d_lock);
1896         list_for_each_entry(child, &dentry->d_subdirs, d_child) {
1897                 if (d_really_is_positive(child)) {
1898                         all_negative = false;
1899                         break;
1900                 }
1901         }
1902         spin_unlock(&dentry->d_lock);
1903
1904         if (all_negative)
1905                 shrink_dcache_parent(dentry);
1906 out:
1907         return all_negative;
1908 }
1909
1910 /*
1911  * Trim old(er) caps.
1912  *
1913  * Because we can't cache an inode without one or more caps, we do
1914  * this indirectly: if a cap is unused, we prune its aliases, at which
1915  * point the inode will hopefully get dropped to.
1916  *
1917  * Yes, this is a bit sloppy.  Our only real goal here is to respond to
1918  * memory pressure from the MDS, though, so it needn't be perfect.
1919  */
1920 static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
1921 {
1922         int *remaining = arg;
1923         struct ceph_inode_info *ci = ceph_inode(inode);
1924         int used, wanted, oissued, mine;
1925
1926         if (*remaining <= 0)
1927                 return -1;
1928
1929         spin_lock(&ci->i_ceph_lock);
1930         mine = cap->issued | cap->implemented;
1931         used = __ceph_caps_used(ci);
1932         wanted = __ceph_caps_file_wanted(ci);
1933         oissued = __ceph_caps_issued_other(ci, cap);
1934
1935         dout("trim_caps_cb %p cap %p mine %s oissued %s used %s wanted %s\n",
1936              inode, cap, ceph_cap_string(mine), ceph_cap_string(oissued),
1937              ceph_cap_string(used), ceph_cap_string(wanted));
1938         if (cap == ci->i_auth_cap) {
1939                 if (ci->i_dirty_caps || ci->i_flushing_caps ||
1940                     !list_empty(&ci->i_cap_snaps))
1941                         goto out;
1942                 if ((used | wanted) & CEPH_CAP_ANY_WR)
1943                         goto out;
1944                 /* Note: it's possible that i_filelock_ref becomes non-zero
1945                  * after dropping auth caps. It doesn't hurt because reply
1946                  * of lock mds request will re-add auth caps. */
1947                 if (atomic_read(&ci->i_filelock_ref) > 0)
1948                         goto out;
1949         }
1950         /* The inode has cached pages, but it's no longer used.
1951          * we can safely drop it */
1952         if (S_ISREG(inode->i_mode) &&
1953             wanted == 0 && used == CEPH_CAP_FILE_CACHE &&
1954             !(oissued & CEPH_CAP_FILE_CACHE)) {
1955           used = 0;
1956           oissued = 0;
1957         }
1958         if ((used | wanted) & ~oissued & mine)
1959                 goto out;   /* we need these caps */
1960
1961         if (oissued) {
1962                 /* we aren't the only cap.. just remove us */
1963                 __ceph_remove_cap(cap, true);
1964                 (*remaining)--;
1965         } else {
1966                 struct dentry *dentry;
1967                 /* try dropping referring dentries */
1968                 spin_unlock(&ci->i_ceph_lock);
1969                 dentry = d_find_any_alias(inode);
1970                 if (dentry && drop_negative_children(dentry)) {
1971                         int count;
1972                         dput(dentry);
1973                         d_prune_aliases(inode);
1974                         count = atomic_read(&inode->i_count);
1975                         if (count == 1)
1976                                 (*remaining)--;
1977                         dout("trim_caps_cb %p cap %p pruned, count now %d\n",
1978                              inode, cap, count);
1979                 } else {
1980                         dput(dentry);
1981                 }
1982                 return 0;
1983         }
1984
1985 out:
1986         spin_unlock(&ci->i_ceph_lock);
1987         return 0;
1988 }
1989
1990 /*
1991  * Trim session cap count down to some max number.
1992  */
1993 int ceph_trim_caps(struct ceph_mds_client *mdsc,
1994                    struct ceph_mds_session *session,
1995                    int max_caps)
1996 {
1997         int trim_caps = session->s_nr_caps - max_caps;
1998
1999         dout("trim_caps mds%d start: %d / %d, trim %d\n",
2000              session->s_mds, session->s_nr_caps, max_caps, trim_caps);
2001         if (trim_caps > 0) {
2002                 int remaining = trim_caps;
2003
2004                 ceph_iterate_session_caps(session, trim_caps_cb, &remaining);
2005                 dout("trim_caps mds%d done: %d / %d, trimmed %d\n",
2006                      session->s_mds, session->s_nr_caps, max_caps,
2007                         trim_caps - remaining);
2008         }
2009
2010         ceph_flush_cap_releases(mdsc, session);
2011         return 0;
2012 }
2013
2014 static int check_caps_flush(struct ceph_mds_client *mdsc,
2015                             u64 want_flush_tid)
2016 {
2017         int ret = 1;
2018
2019         spin_lock(&mdsc->cap_dirty_lock);
2020         if (!list_empty(&mdsc->cap_flush_list)) {
2021                 struct ceph_cap_flush *cf =
2022                         list_first_entry(&mdsc->cap_flush_list,
2023                                          struct ceph_cap_flush, g_list);
2024                 if (cf->tid <= want_flush_tid) {
2025                         dout("check_caps_flush still flushing tid "
2026                              "%llu <= %llu\n", cf->tid, want_flush_tid);
2027                         ret = 0;
2028                 }
2029         }
2030         spin_unlock(&mdsc->cap_dirty_lock);
2031         return ret;
2032 }
2033
2034 /*
2035  * flush all dirty inode data to disk.
2036  *
2037  * returns true if we've flushed through want_flush_tid
2038  */
2039 static void wait_caps_flush(struct ceph_mds_client *mdsc,
2040                             u64 want_flush_tid)
2041 {
2042         dout("check_caps_flush want %llu\n", want_flush_tid);
2043
2044         wait_event(mdsc->cap_flushing_wq,
2045                    check_caps_flush(mdsc, want_flush_tid));
2046
2047         dout("check_caps_flush ok, flushed thru %llu\n", want_flush_tid);
2048 }
2049
2050 /*
2051  * called under s_mutex
2052  */
2053 static void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
2054                                    struct ceph_mds_session *session)
2055 {
2056         struct ceph_msg *msg = NULL;
2057         struct ceph_mds_cap_release *head;
2058         struct ceph_mds_cap_item *item;
2059         struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;
2060         struct ceph_cap *cap;
2061         LIST_HEAD(tmp_list);
2062         int num_cap_releases;
2063         __le32  barrier, *cap_barrier;
2064
2065         down_read(&osdc->lock);
2066         barrier = cpu_to_le32(osdc->epoch_barrier);
2067         up_read(&osdc->lock);
2068
2069         spin_lock(&session->s_cap_lock);
2070 again:
2071         list_splice_init(&session->s_cap_releases, &tmp_list);
2072         num_cap_releases = session->s_num_cap_releases;
2073         session->s_num_cap_releases = 0;
2074         spin_unlock(&session->s_cap_lock);
2075
2076         while (!list_empty(&tmp_list)) {
2077                 if (!msg) {
2078                         msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE,
2079                                         PAGE_SIZE, GFP_NOFS, false);
2080                         if (!msg)
2081                                 goto out_err;
2082                         head = msg->front.iov_base;
2083                         head->num = cpu_to_le32(0);
2084                         msg->front.iov_len = sizeof(*head);
2085
2086                         msg->hdr.version = cpu_to_le16(2);
2087                         msg->hdr.compat_version = cpu_to_le16(1);
2088                 }
2089
2090                 cap = list_first_entry(&tmp_list, struct ceph_cap,
2091                                         session_caps);
2092                 list_del(&cap->session_caps);
2093                 num_cap_releases--;
2094
2095                 head = msg->front.iov_base;
2096                 put_unaligned_le32(get_unaligned_le32(&head->num) + 1,
2097                                    &head->num);
2098                 item = msg->front.iov_base + msg->front.iov_len;
2099                 item->ino = cpu_to_le64(cap->cap_ino);
2100                 item->cap_id = cpu_to_le64(cap->cap_id);
2101                 item->migrate_seq = cpu_to_le32(cap->mseq);
2102                 item->seq = cpu_to_le32(cap->issue_seq);
2103                 msg->front.iov_len += sizeof(*item);
2104
2105                 ceph_put_cap(mdsc, cap);
2106
2107                 if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {
2108                         // Append cap_barrier field
2109                         cap_barrier = msg->front.iov_base + msg->front.iov_len;
2110                         *cap_barrier = barrier;
2111                         msg->front.iov_len += sizeof(*cap_barrier);
2112
2113                         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2114                         dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2115                         ceph_con_send(&session->s_con, msg);
2116                         msg = NULL;
2117                 }
2118         }
2119
2120         BUG_ON(num_cap_releases != 0);
2121
2122         spin_lock(&session->s_cap_lock);
2123         if (!list_empty(&session->s_cap_releases))
2124                 goto again;
2125         spin_unlock(&session->s_cap_lock);
2126
2127         if (msg) {
2128                 // Append cap_barrier field
2129                 cap_barrier = msg->front.iov_base + msg->front.iov_len;
2130                 *cap_barrier = barrier;
2131                 msg->front.iov_len += sizeof(*cap_barrier);
2132
2133                 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2134                 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2135                 ceph_con_send(&session->s_con, msg);
2136         }
2137         return;
2138 out_err:
2139         pr_err("send_cap_releases mds%d, failed to allocate message\n",
2140                 session->s_mds);
2141         spin_lock(&session->s_cap_lock);
2142         list_splice(&tmp_list, &session->s_cap_releases);
2143         session->s_num_cap_releases += num_cap_releases;
2144         spin_unlock(&session->s_cap_lock);
2145 }
2146
2147 static void ceph_cap_release_work(struct work_struct *work)
2148 {
2149         struct ceph_mds_session *session =
2150                 container_of(work, struct ceph_mds_session, s_cap_release_work);
2151
2152         mutex_lock(&session->s_mutex);
2153         if (session->s_state == CEPH_MDS_SESSION_OPEN ||
2154             session->s_state == CEPH_MDS_SESSION_HUNG)
2155                 ceph_send_cap_releases(session->s_mdsc, session);
2156         mutex_unlock(&session->s_mutex);
2157         ceph_put_mds_session(session);
2158 }
2159
2160 void ceph_flush_cap_releases(struct ceph_mds_client *mdsc,
2161                              struct ceph_mds_session *session)
2162 {
2163         if (mdsc->stopping)
2164                 return;
2165
2166         ceph_get_mds_session(session);
2167         if (queue_work(mdsc->fsc->cap_wq,
2168                        &session->s_cap_release_work)) {
2169                 dout("cap release work queued\n");
2170         } else {
2171                 ceph_put_mds_session(session);
2172                 dout("failed to queue cap release work\n");
2173         }
2174 }
2175
2176 /*
2177  * caller holds session->s_cap_lock
2178  */
2179 void __ceph_queue_cap_release(struct ceph_mds_session *session,
2180                               struct ceph_cap *cap)
2181 {
2182         list_add_tail(&cap->session_caps, &session->s_cap_releases);
2183         session->s_num_cap_releases++;
2184
2185         if (!(session->s_num_cap_releases % CEPH_CAPS_PER_RELEASE))
2186                 ceph_flush_cap_releases(session->s_mdsc, session);
2187 }
2188
2189 static void ceph_cap_reclaim_work(struct work_struct *work)
2190 {
2191         struct ceph_mds_client *mdsc =
2192                 container_of(work, struct ceph_mds_client, cap_reclaim_work);
2193         int ret = ceph_trim_dentries(mdsc);
2194         if (ret == -EAGAIN)
2195                 ceph_queue_cap_reclaim_work(mdsc);
2196 }
2197
2198 void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc)
2199 {
2200         if (mdsc->stopping)
2201                 return;
2202
2203         if (queue_work(mdsc->fsc->cap_wq, &mdsc->cap_reclaim_work)) {
2204                 dout("caps reclaim work queued\n");
2205         } else {
2206                 dout("failed to queue caps release work\n");
2207         }
2208 }
2209
2210 void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr)
2211 {
2212         int val;
2213         if (!nr)
2214                 return;
2215         val = atomic_add_return(nr, &mdsc->cap_reclaim_pending);
2216         if ((val % CEPH_CAPS_PER_RELEASE) < nr) {
2217                 atomic_set(&mdsc->cap_reclaim_pending, 0);
2218                 ceph_queue_cap_reclaim_work(mdsc);
2219         }
2220 }
2221
2222 /*
2223  * requests
2224  */
2225
2226 int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
2227                                     struct inode *dir)
2228 {
2229         struct ceph_inode_info *ci = ceph_inode(dir);
2230         struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
2231         struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;
2232         size_t size = sizeof(struct ceph_mds_reply_dir_entry);
2233         unsigned int num_entries;
2234         int order;
2235
2236         spin_lock(&ci->i_ceph_lock);
2237         num_entries = ci->i_files + ci->i_subdirs;
2238         spin_unlock(&ci->i_ceph_lock);
2239         num_entries = max(num_entries, 1U);
2240         num_entries = min(num_entries, opt->max_readdir);
2241
2242         order = get_order(size * num_entries);
2243         while (order >= 0) {
2244                 rinfo->dir_entries = (void*)__get_free_pages(GFP_KERNEL |
2245                                                              __GFP_NOWARN,
2246                                                              order);
2247                 if (rinfo->dir_entries)
2248                         break;
2249                 order--;
2250         }
2251         if (!rinfo->dir_entries)
2252                 return -ENOMEM;
2253
2254         num_entries = (PAGE_SIZE << order) / size;
2255         num_entries = min(num_entries, opt->max_readdir);
2256
2257         rinfo->dir_buf_size = PAGE_SIZE << order;
2258         req->r_num_caps = num_entries + 1;
2259         req->r_args.readdir.max_entries = cpu_to_le32(num_entries);
2260         req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);
2261         return 0;
2262 }
2263
2264 /*
2265  * Create an mds request.
2266  */
2267 struct ceph_mds_request *
2268 ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode)
2269 {
2270         struct ceph_mds_request *req;
2271
2272         req = kmem_cache_zalloc(ceph_mds_request_cachep, GFP_NOFS);
2273         if (!req)
2274                 return ERR_PTR(-ENOMEM);
2275
2276         mutex_init(&req->r_fill_mutex);
2277         req->r_mdsc = mdsc;
2278         req->r_started = jiffies;
2279         req->r_start_latency = ktime_get();
2280         req->r_resend_mds = -1;
2281         INIT_LIST_HEAD(&req->r_unsafe_dir_item);
2282         INIT_LIST_HEAD(&req->r_unsafe_target_item);
2283         req->r_fmode = -1;
2284         kref_init(&req->r_kref);
2285         RB_CLEAR_NODE(&req->r_node);
2286         INIT_LIST_HEAD(&req->r_wait);
2287         init_completion(&req->r_completion);
2288         init_completion(&req->r_safe_completion);
2289         INIT_LIST_HEAD(&req->r_unsafe_item);
2290
2291         ktime_get_coarse_real_ts64(&req->r_stamp);
2292
2293         req->r_op = op;
2294         req->r_direct_mode = mode;
2295         return req;
2296 }
2297
2298 /*
2299  * return oldest (lowest) request, tid in request tree, 0 if none.
2300  *
2301  * called under mdsc->mutex.
2302  */
2303 static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc)
2304 {
2305         if (RB_EMPTY_ROOT(&mdsc->request_tree))
2306                 return NULL;
2307         return rb_entry(rb_first(&mdsc->request_tree),
2308                         struct ceph_mds_request, r_node);
2309 }
2310
2311 static inline  u64 __get_oldest_tid(struct ceph_mds_client *mdsc)
2312 {
2313         return mdsc->oldest_tid;
2314 }
2315
2316 /*
2317  * Build a dentry's path.  Allocate on heap; caller must kfree.  Based
2318  * on build_path_from_dentry in fs/cifs/dir.c.
2319  *
2320  * If @stop_on_nosnap, generate path relative to the first non-snapped
2321  * inode.
2322  *
2323  * Encode hidden .snap dirs as a double /, i.e.
2324  *   foo/.snap/bar -> foo//bar
2325  */
2326 char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *pbase,
2327                            int stop_on_nosnap)
2328 {
2329         struct dentry *temp;
2330         char *path;
2331         int pos;
2332         unsigned seq;
2333         u64 base;
2334
2335         if (!dentry)
2336                 return ERR_PTR(-EINVAL);
2337
2338         path = __getname();
2339         if (!path)
2340                 return ERR_PTR(-ENOMEM);
2341 retry:
2342         pos = PATH_MAX - 1;
2343         path[pos] = '\0';
2344
2345         seq = read_seqbegin(&rename_lock);
2346         rcu_read_lock();
2347         temp = dentry;
2348         for (;;) {
2349                 struct inode *inode;
2350
2351                 spin_lock(&temp->d_lock);
2352                 inode = d_inode(temp);
2353                 if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
2354                         dout("build_path path+%d: %p SNAPDIR\n",
2355                              pos, temp);
2356                 } else if (stop_on_nosnap && inode && dentry != temp &&
2357                            ceph_snap(inode) == CEPH_NOSNAP) {
2358                         spin_unlock(&temp->d_lock);
2359                         pos++; /* get rid of any prepended '/' */
2360                         break;
2361                 } else {
2362                         pos -= temp->d_name.len;
2363                         if (pos < 0) {
2364                                 spin_unlock(&temp->d_lock);
2365                                 break;
2366                         }
2367                         memcpy(path + pos, temp->d_name.name, temp->d_name.len);
2368                 }
2369                 spin_unlock(&temp->d_lock);
2370                 temp = READ_ONCE(temp->d_parent);
2371
2372                 /* Are we at the root? */
2373                 if (IS_ROOT(temp))
2374                         break;
2375
2376                 /* Are we out of buffer? */
2377                 if (--pos < 0)
2378                         break;
2379
2380                 path[pos] = '/';
2381         }
2382         base = ceph_ino(d_inode(temp));
2383         rcu_read_unlock();
2384
2385         if (read_seqretry(&rename_lock, seq))
2386                 goto retry;
2387
2388         if (pos < 0) {
2389                 /*
2390                  * A rename didn't occur, but somehow we didn't end up where
2391                  * we thought we would. Throw a warning and try again.
2392                  */
2393                 pr_warn("build_path did not end path lookup where "
2394                         "expected, pos is %d\n", pos);
2395                 goto retry;
2396         }
2397
2398         *pbase = base;
2399         *plen = PATH_MAX - 1 - pos;
2400         dout("build_path on %p %d built %llx '%.*s'\n",
2401              dentry, d_count(dentry), base, *plen, path + pos);
2402         return path + pos;
2403 }
2404
2405 static int build_dentry_path(struct dentry *dentry, struct inode *dir,
2406                              const char **ppath, int *ppathlen, u64 *pino,
2407                              bool *pfreepath, bool parent_locked)
2408 {
2409         char *path;
2410
2411         rcu_read_lock();
2412         if (!dir)
2413                 dir = d_inode_rcu(dentry->d_parent);
2414         if (dir && parent_locked && ceph_snap(dir) == CEPH_NOSNAP) {
2415                 *pino = ceph_ino(dir);
2416                 rcu_read_unlock();
2417                 *ppath = dentry->d_name.name;
2418                 *ppathlen = dentry->d_name.len;
2419                 return 0;
2420         }
2421         rcu_read_unlock();
2422         path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2423         if (IS_ERR(path))
2424                 return PTR_ERR(path);
2425         *ppath = path;
2426         *pfreepath = true;
2427         return 0;
2428 }
2429
2430 static int build_inode_path(struct inode *inode,
2431                             const char **ppath, int *ppathlen, u64 *pino,
2432                             bool *pfreepath)
2433 {
2434         struct dentry *dentry;
2435         char *path;
2436
2437         if (ceph_snap(inode) == CEPH_NOSNAP) {
2438                 *pino = ceph_ino(inode);
2439                 *ppathlen = 0;
2440                 return 0;
2441         }
2442         dentry = d_find_alias(inode);
2443         path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2444         dput(dentry);
2445         if (IS_ERR(path))
2446                 return PTR_ERR(path);
2447         *ppath = path;
2448         *pfreepath = true;
2449         return 0;
2450 }
2451
2452 /*
2453  * request arguments may be specified via an inode *, a dentry *, or
2454  * an explicit ino+path.
2455  */
2456 static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
2457                                   struct inode *rdiri, const char *rpath,
2458                                   u64 rino, const char **ppath, int *pathlen,
2459                                   u64 *ino, bool *freepath, bool parent_locked)
2460 {
2461         int r = 0;
2462
2463         if (rinode) {
2464                 r = build_inode_path(rinode, ppath, pathlen, ino, freepath);
2465                 dout(" inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
2466                      ceph_snap(rinode));
2467         } else if (rdentry) {
2468                 r = build_dentry_path(rdentry, rdiri, ppath, pathlen, ino,
2469                                         freepath, parent_locked);
2470                 dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
2471                      *ppath);
2472         } else if (rpath || rino) {
2473                 *ino = rino;
2474                 *ppath = rpath;
2475                 *pathlen = rpath ? strlen(rpath) : 0;
2476                 dout(" path %.*s\n", *pathlen, rpath);
2477         }
2478
2479         return r;
2480 }
2481
2482 /*
2483  * called under mdsc->mutex
2484  */
2485 static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
2486                                                struct ceph_mds_request *req,
2487                                                int mds, bool drop_cap_releases)
2488 {
2489         struct ceph_msg *msg;
2490         struct ceph_mds_request_head *head;
2491         const char *path1 = NULL;
2492         const char *path2 = NULL;
2493         u64 ino1 = 0, ino2 = 0;
2494         int pathlen1 = 0, pathlen2 = 0;
2495         bool freepath1 = false, freepath2 = false;
2496         int len;
2497         u16 releases;
2498         void *p, *end;
2499         int ret;
2500
2501         ret = set_request_path_attr(req->r_inode, req->r_dentry,
2502                               req->r_parent, req->r_path1, req->r_ino1.ino,
2503                               &path1, &pathlen1, &ino1, &freepath1,
2504                               test_bit(CEPH_MDS_R_PARENT_LOCKED,
2505                                         &req->r_req_flags));
2506         if (ret < 0) {
2507                 msg = ERR_PTR(ret);
2508                 goto out;
2509         }
2510
2511         /* If r_old_dentry is set, then assume that its parent is locked */
2512         ret = set_request_path_attr(NULL, req->r_old_dentry,
2513                               req->r_old_dentry_dir,
2514                               req->r_path2, req->r_ino2.ino,
2515                               &path2, &pathlen2, &ino2, &freepath2, true);
2516         if (ret < 0) {
2517                 msg = ERR_PTR(ret);
2518                 goto out_free1;
2519         }
2520
2521         len = sizeof(*head) +
2522                 pathlen1 + pathlen2 + 2*(1 + sizeof(u32) + sizeof(u64)) +
2523                 sizeof(struct ceph_timespec);
2524
2525         /* calculate (max) length for cap releases */
2526         len += sizeof(struct ceph_mds_request_release) *
2527                 (!!req->r_inode_drop + !!req->r_dentry_drop +
2528                  !!req->r_old_inode_drop + !!req->r_old_dentry_drop);
2529         if (req->r_dentry_drop)
2530                 len += pathlen1;
2531         if (req->r_old_dentry_drop)
2532                 len += pathlen2;
2533
2534         msg = ceph_msg_new2(CEPH_MSG_CLIENT_REQUEST, len, 1, GFP_NOFS, false);
2535         if (!msg) {
2536                 msg = ERR_PTR(-ENOMEM);
2537                 goto out_free2;
2538         }
2539
2540         msg->hdr.version = cpu_to_le16(2);
2541         msg->hdr.tid = cpu_to_le64(req->r_tid);
2542
2543         head = msg->front.iov_base;
2544         p = msg->front.iov_base + sizeof(*head);
2545         end = msg->front.iov_base + msg->front.iov_len;
2546
2547         head->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);
2548         head->op = cpu_to_le32(req->r_op);
2549         head->caller_uid = cpu_to_le32(from_kuid(&init_user_ns, req->r_uid));
2550         head->caller_gid = cpu_to_le32(from_kgid(&init_user_ns, req->r_gid));
2551         head->ino = cpu_to_le64(req->r_deleg_ino);
2552         head->args = req->r_args;
2553
2554         ceph_encode_filepath(&p, end, ino1, path1);
2555         ceph_encode_filepath(&p, end, ino2, path2);
2556
2557         /* make note of release offset, in case we need to replay */
2558         req->r_request_release_offset = p - msg->front.iov_base;
2559
2560         /* cap releases */
2561         releases = 0;
2562         if (req->r_inode_drop)
2563                 releases += ceph_encode_inode_release(&p,
2564                       req->r_inode ? req->r_inode : d_inode(req->r_dentry),
2565                       mds, req->r_inode_drop, req->r_inode_unless,
2566                       req->r_op == CEPH_MDS_OP_READDIR);
2567         if (req->r_dentry_drop)
2568                 releases += ceph_encode_dentry_release(&p, req->r_dentry,
2569                                 req->r_parent, mds, req->r_dentry_drop,
2570                                 req->r_dentry_unless);
2571         if (req->r_old_dentry_drop)
2572                 releases += ceph_encode_dentry_release(&p, req->r_old_dentry,
2573                                 req->r_old_dentry_dir, mds,
2574                                 req->r_old_dentry_drop,
2575                                 req->r_old_dentry_unless);
2576         if (req->r_old_inode_drop)
2577                 releases += ceph_encode_inode_release(&p,
2578                       d_inode(req->r_old_dentry),
2579                       mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
2580
2581         if (drop_cap_releases) {
2582                 releases = 0;
2583                 p = msg->front.iov_base + req->r_request_release_offset;
2584         }
2585
2586         head->num_releases = cpu_to_le16(releases);
2587
2588         /* time stamp */
2589         {
2590                 struct ceph_timespec ts;
2591                 ceph_encode_timespec64(&ts, &req->r_stamp);
2592                 ceph_encode_copy(&p, &ts, sizeof(ts));
2593         }
2594
2595         if (WARN_ON_ONCE(p > end)) {
2596                 ceph_msg_put(msg);
2597                 msg = ERR_PTR(-ERANGE);
2598                 goto out_free2;
2599         }
2600
2601         msg->front.iov_len = p - msg->front.iov_base;
2602         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2603
2604         if (req->r_pagelist) {
2605                 struct ceph_pagelist *pagelist = req->r_pagelist;
2606                 ceph_msg_data_add_pagelist(msg, pagelist);
2607                 msg->hdr.data_len = cpu_to_le32(pagelist->length);
2608         } else {
2609                 msg->hdr.data_len = 0;
2610         }
2611
2612         msg->hdr.data_off = cpu_to_le16(0);
2613
2614 out_free2:
2615         if (freepath2)
2616                 ceph_mdsc_free_path((char *)path2, pathlen2);
2617 out_free1:
2618         if (freepath1)
2619                 ceph_mdsc_free_path((char *)path1, pathlen1);
2620 out:
2621         return msg;
2622 }
2623
2624 /*
2625  * called under mdsc->mutex if error, under no mutex if
2626  * success.
2627  */
2628 static void complete_request(struct ceph_mds_client *mdsc,
2629                              struct ceph_mds_request *req)
2630 {
2631         req->r_end_latency = ktime_get();
2632
2633         if (req->r_callback)
2634                 req->r_callback(mdsc, req);
2635         complete_all(&req->r_completion);
2636 }
2637
2638 /*
2639  * called under mdsc->mutex
2640  */
2641 static int __prepare_send_request(struct ceph_mds_client *mdsc,
2642                                   struct ceph_mds_request *req,
2643                                   int mds, bool drop_cap_releases)
2644 {
2645         struct ceph_mds_request_head *rhead;
2646         struct ceph_msg *msg;
2647         int flags = 0;
2648
2649         req->r_attempts++;
2650         if (req->r_inode) {
2651                 struct ceph_cap *cap =
2652                         ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);
2653
2654                 if (cap)
2655                         req->r_sent_on_mseq = cap->mseq;
2656                 else
2657                         req->r_sent_on_mseq = -1;
2658         }
2659         dout("prepare_send_request %p tid %lld %s (attempt %d)\n", req,
2660              req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);
2661
2662         if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
2663                 void *p;
2664                 /*
2665                  * Replay.  Do not regenerate message (and rebuild
2666                  * paths, etc.); just use the original message.
2667                  * Rebuilding paths will break for renames because
2668                  * d_move mangles the src name.
2669                  */
2670                 msg = req->r_request;
2671                 rhead = msg->front.iov_base;
2672
2673                 flags = le32_to_cpu(rhead->flags);
2674                 flags |= CEPH_MDS_FLAG_REPLAY;
2675                 rhead->flags = cpu_to_le32(flags);
2676
2677                 if (req->r_target_inode)
2678                         rhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));
2679
2680                 rhead->num_retry = req->r_attempts - 1;
2681
2682                 /* remove cap/dentry releases from message */
2683                 rhead->num_releases = 0;
2684
2685                 /* time stamp */
2686                 p = msg->front.iov_base + req->r_request_release_offset;
2687                 {
2688                         struct ceph_timespec ts;
2689                         ceph_encode_timespec64(&ts, &req->r_stamp);
2690                         ceph_encode_copy(&p, &ts, sizeof(ts));
2691                 }
2692
2693                 msg->front.iov_len = p - msg->front.iov_base;
2694                 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2695                 return 0;
2696         }
2697
2698         if (req->r_request) {
2699                 ceph_msg_put(req->r_request);
2700                 req->r_request = NULL;
2701         }
2702         msg = create_request_message(mdsc, req, mds, drop_cap_releases);
2703         if (IS_ERR(msg)) {
2704                 req->r_err = PTR_ERR(msg);
2705                 return PTR_ERR(msg);
2706         }
2707         req->r_request = msg;
2708
2709         rhead = msg->front.iov_base;
2710         rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
2711         if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2712                 flags |= CEPH_MDS_FLAG_REPLAY;
2713         if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags))
2714                 flags |= CEPH_MDS_FLAG_ASYNC;
2715         if (req->r_parent)
2716                 flags |= CEPH_MDS_FLAG_WANT_DENTRY;
2717         rhead->flags = cpu_to_le32(flags);
2718         rhead->num_fwd = req->r_num_fwd;
2719         rhead->num_retry = req->r_attempts - 1;
2720
2721         dout(" r_parent = %p\n", req->r_parent);
2722         return 0;
2723 }
2724
2725 /*
2726  * called under mdsc->mutex
2727  */
2728 static int __send_request(struct ceph_mds_client *mdsc,
2729                           struct ceph_mds_session *session,
2730                           struct ceph_mds_request *req,
2731                           bool drop_cap_releases)
2732 {
2733         int err;
2734
2735         err = __prepare_send_request(mdsc, req, session->s_mds,
2736                                      drop_cap_releases);
2737         if (!err) {
2738                 ceph_msg_get(req->r_request);
2739                 ceph_con_send(&session->s_con, req->r_request);
2740         }
2741
2742         return err;
2743 }
2744
2745 /*
2746  * send request, or put it on the appropriate wait list.
2747  */
2748 static void __do_request(struct ceph_mds_client *mdsc,
2749                         struct ceph_mds_request *req)
2750 {
2751         struct ceph_mds_session *session = NULL;
2752         int mds = -1;
2753         int err = 0;
2754         bool random;
2755
2756         if (req->r_err || test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
2757                 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))
2758                         __unregister_request(mdsc, req);
2759                 return;
2760         }
2761
2762         if (req->r_timeout &&
2763             time_after_eq(jiffies, req->r_started + req->r_timeout)) {
2764                 dout("do_request timed out\n");
2765                 err = -ETIMEDOUT;
2766                 goto finish;
2767         }
2768         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
2769                 dout("do_request forced umount\n");
2770                 err = -EIO;
2771                 goto finish;
2772         }
2773         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {
2774                 if (mdsc->mdsmap_err) {
2775                         err = mdsc->mdsmap_err;
2776                         dout("do_request mdsmap err %d\n", err);
2777                         goto finish;
2778                 }
2779                 if (mdsc->mdsmap->m_epoch == 0) {
2780                         dout("do_request no mdsmap, waiting for map\n");
2781                         list_add(&req->r_wait, &mdsc->waiting_for_map);
2782                         return;
2783                 }
2784                 if (!(mdsc->fsc->mount_options->flags &
2785                       CEPH_MOUNT_OPT_MOUNTWAIT) &&
2786                     !ceph_mdsmap_is_cluster_available(mdsc->mdsmap)) {
2787                         err = -EHOSTUNREACH;
2788                         goto finish;
2789                 }
2790         }
2791
2792         put_request_session(req);
2793
2794         mds = __choose_mds(mdsc, req, &random);
2795         if (mds < 0 ||
2796             ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {
2797                 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2798                         err = -EJUKEBOX;
2799                         goto finish;
2800                 }
2801                 dout("do_request no mds or not active, waiting for map\n");
2802                 list_add(&req->r_wait, &mdsc->waiting_for_map);
2803                 return;
2804         }
2805
2806         /* get, open session */
2807         session = __ceph_lookup_mds_session(mdsc, mds);
2808         if (!session) {
2809                 session = register_session(mdsc, mds);
2810                 if (IS_ERR(session)) {
2811                         err = PTR_ERR(session);
2812                         goto finish;
2813                 }
2814         }
2815         req->r_session = ceph_get_mds_session(session);
2816
2817         dout("do_request mds%d session %p state %s\n", mds, session,
2818              ceph_session_state_name(session->s_state));
2819         if (session->s_state != CEPH_MDS_SESSION_OPEN &&
2820             session->s_state != CEPH_MDS_SESSION_HUNG) {
2821                 if (session->s_state == CEPH_MDS_SESSION_REJECTED) {
2822                         err = -EACCES;
2823                         goto out_session;
2824                 }
2825                 /*
2826                  * We cannot queue async requests since the caps and delegated
2827                  * inodes are bound to the session. Just return -EJUKEBOX and
2828                  * let the caller retry a sync request in that case.
2829                  */
2830                 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2831                         err = -EJUKEBOX;
2832                         goto out_session;
2833                 }
2834                 if (session->s_state == CEPH_MDS_SESSION_NEW ||
2835                     session->s_state == CEPH_MDS_SESSION_CLOSING) {
2836                         err = __open_session(mdsc, session);
2837                         if (err)
2838                                 goto out_session;
2839                         /* retry the same mds later */
2840                         if (random)
2841                                 req->r_resend_mds = mds;
2842                 }
2843                 list_add(&req->r_wait, &session->s_waiting);
2844                 goto out_session;
2845         }
2846
2847         /* send request */
2848         req->r_resend_mds = -1;   /* forget any previous mds hint */
2849
2850         if (req->r_request_started == 0)   /* note request start time */
2851                 req->r_request_started = jiffies;
2852
2853         err = __send_request(mdsc, session, req, false);
2854
2855 out_session:
2856         ceph_put_mds_session(session);
2857 finish:
2858         if (err) {
2859                 dout("__do_request early error %d\n", err);
2860                 req->r_err = err;
2861                 complete_request(mdsc, req);
2862                 __unregister_request(mdsc, req);
2863         }
2864         return;
2865 }
2866
2867 /*
2868  * called under mdsc->mutex
2869  */
2870 static void __wake_requests(struct ceph_mds_client *mdsc,
2871                             struct list_head *head)
2872 {
2873         struct ceph_mds_request *req;
2874         LIST_HEAD(tmp_list);
2875
2876         list_splice_init(head, &tmp_list);
2877
2878         while (!list_empty(&tmp_list)) {
2879                 req = list_entry(tmp_list.next,
2880                                  struct ceph_mds_request, r_wait);
2881                 list_del_init(&req->r_wait);
2882                 dout(" wake request %p tid %llu\n", req, req->r_tid);
2883                 __do_request(mdsc, req);
2884         }
2885 }
2886
2887 /*
2888  * Wake up threads with requests pending for @mds, so that they can
2889  * resubmit their requests to a possibly different mds.
2890  */
2891 static void kick_requests(struct ceph_mds_client *mdsc, int mds)
2892 {
2893         struct ceph_mds_request *req;
2894         struct rb_node *p = rb_first(&mdsc->request_tree);
2895
2896         dout("kick_requests mds%d\n", mds);
2897         while (p) {
2898                 req = rb_entry(p, struct ceph_mds_request, r_node);
2899                 p = rb_next(p);
2900                 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2901                         continue;
2902                 if (req->r_attempts > 0)
2903                         continue; /* only new requests */
2904                 if (req->r_session &&
2905                     req->r_session->s_mds == mds) {
2906                         dout(" kicking tid %llu\n", req->r_tid);
2907                         list_del_init(&req->r_wait);
2908                         __do_request(mdsc, req);
2909                 }
2910         }
2911 }
2912
2913 int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, struct inode *dir,
2914                               struct ceph_mds_request *req)
2915 {
2916         int err = 0;
2917
2918         /* take CAP_PIN refs for r_inode, r_parent, r_old_dentry */
2919         if (req->r_inode)
2920                 ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
2921         if (req->r_parent) {
2922                 struct ceph_inode_info *ci = ceph_inode(req->r_parent);
2923                 int fmode = (req->r_op & CEPH_MDS_OP_WRITE) ?
2924                             CEPH_FILE_MODE_WR : CEPH_FILE_MODE_RD;
2925                 spin_lock(&ci->i_ceph_lock);
2926                 ceph_take_cap_refs(ci, CEPH_CAP_PIN, false);
2927                 __ceph_touch_fmode(ci, mdsc, fmode);
2928                 spin_unlock(&ci->i_ceph_lock);
2929                 ihold(req->r_parent);
2930         }
2931         if (req->r_old_dentry_dir)
2932                 ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
2933                                   CEPH_CAP_PIN);
2934
2935         if (req->r_inode) {
2936                 err = ceph_wait_on_async_create(req->r_inode);
2937                 if (err) {
2938                         dout("%s: wait for async create returned: %d\n",
2939                              __func__, err);
2940                         return err;
2941                 }
2942         }
2943
2944         if (!err && req->r_old_inode) {
2945                 err = ceph_wait_on_async_create(req->r_old_inode);
2946                 if (err) {
2947                         dout("%s: wait for async create returned: %d\n",
2948                              __func__, err);
2949                         return err;
2950                 }
2951         }
2952
2953         dout("submit_request on %p for inode %p\n", req, dir);
2954         mutex_lock(&mdsc->mutex);
2955         __register_request(mdsc, req, dir);
2956         __do_request(mdsc, req);
2957         err = req->r_err;
2958         mutex_unlock(&mdsc->mutex);
2959         return err;
2960 }
2961
2962 static int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc,
2963                                   struct ceph_mds_request *req)
2964 {
2965         int err;
2966
2967         /* wait */
2968         dout("do_request waiting\n");
2969         if (!req->r_timeout && req->r_wait_for_completion) {
2970                 err = req->r_wait_for_completion(mdsc, req);
2971         } else {
2972                 long timeleft = wait_for_completion_killable_timeout(
2973                                         &req->r_completion,
2974                                         ceph_timeout_jiffies(req->r_timeout));
2975                 if (timeleft > 0)
2976                         err = 0;
2977                 else if (!timeleft)
2978                         err = -ETIMEDOUT;  /* timed out */
2979                 else
2980                         err = timeleft;  /* killed */
2981         }
2982         dout("do_request waited, got %d\n", err);
2983         mutex_lock(&mdsc->mutex);
2984
2985         /* only abort if we didn't race with a real reply */
2986         if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
2987                 err = le32_to_cpu(req->r_reply_info.head->result);
2988         } else if (err < 0) {
2989                 dout("aborted request %lld with %d\n", req->r_tid, err);
2990
2991                 /*
2992                  * ensure we aren't running concurrently with
2993                  * ceph_fill_trace or ceph_readdir_prepopulate, which
2994                  * rely on locks (dir mutex) held by our caller.
2995                  */
2996                 mutex_lock(&req->r_fill_mutex);
2997                 req->r_err = err;
2998                 set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
2999                 mutex_unlock(&req->r_fill_mutex);
3000
3001                 if (req->r_parent &&
3002                     (req->r_op & CEPH_MDS_OP_WRITE))
3003                         ceph_invalidate_dir_request(req);
3004         } else {
3005                 err = req->r_err;
3006         }
3007
3008         mutex_unlock(&mdsc->mutex);
3009         return err;
3010 }
3011
3012 /*
3013  * Synchrously perform an mds request.  Take care of all of the
3014  * session setup, forwarding, retry details.
3015  */
3016 int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
3017                          struct inode *dir,
3018                          struct ceph_mds_request *req)
3019 {
3020         int err;
3021
3022         dout("do_request on %p\n", req);
3023
3024         /* issue */
3025         err = ceph_mdsc_submit_request(mdsc, dir, req);
3026         if (!err)
3027                 err = ceph_mdsc_wait_request(mdsc, req);
3028         dout("do_request %p done, result %d\n", req, err);
3029         return err;
3030 }
3031
3032 /*
3033  * Invalidate dir's completeness, dentry lease state on an aborted MDS
3034  * namespace request.
3035  */
3036 void ceph_invalidate_dir_request(struct ceph_mds_request *req)
3037 {
3038         struct inode *dir = req->r_parent;
3039         struct inode *old_dir = req->r_old_dentry_dir;
3040
3041         dout("invalidate_dir_request %p %p (complete, lease(s))\n", dir, old_dir);
3042
3043         ceph_dir_clear_complete(dir);
3044         if (old_dir)
3045                 ceph_dir_clear_complete(old_dir);
3046         if (req->r_dentry)
3047                 ceph_invalidate_dentry_lease(req->r_dentry);
3048         if (req->r_old_dentry)
3049                 ceph_invalidate_dentry_lease(req->r_old_dentry);
3050 }
3051
3052 /*
3053  * Handle mds reply.
3054  *
3055  * We take the session mutex and parse and process the reply immediately.
3056  * This preserves the logical ordering of replies, capabilities, etc., sent
3057  * by the MDS as they are applied to our local cache.
3058  */
3059 static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
3060 {
3061         struct ceph_mds_client *mdsc = session->s_mdsc;
3062         struct ceph_mds_request *req;
3063         struct ceph_mds_reply_head *head = msg->front.iov_base;
3064         struct ceph_mds_reply_info_parsed *rinfo;  /* parsed reply info */
3065         struct ceph_snap_realm *realm;
3066         u64 tid;
3067         int err, result;
3068         int mds = session->s_mds;
3069
3070         if (msg->front.iov_len < sizeof(*head)) {
3071                 pr_err("mdsc_handle_reply got corrupt (short) reply\n");
3072                 ceph_msg_dump(msg);
3073                 return;
3074         }
3075
3076         /* get request, session */
3077         tid = le64_to_cpu(msg->hdr.tid);
3078         mutex_lock(&mdsc->mutex);
3079         req = lookup_get_request(mdsc, tid);
3080         if (!req) {
3081                 dout("handle_reply on unknown tid %llu\n", tid);
3082                 mutex_unlock(&mdsc->mutex);
3083                 return;
3084         }
3085         dout("handle_reply %p\n", req);
3086
3087         /* correct session? */
3088         if (req->r_session != session) {
3089                 pr_err("mdsc_handle_reply got %llu on session mds%d"
3090                        " not mds%d\n", tid, session->s_mds,
3091                        req->r_session ? req->r_session->s_mds : -1);
3092                 mutex_unlock(&mdsc->mutex);
3093                 goto out;
3094         }
3095
3096         /* dup? */
3097         if ((test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags) && !head->safe) ||
3098             (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags) && head->safe)) {
3099                 pr_warn("got a dup %s reply on %llu from mds%d\n",
3100                            head->safe ? "safe" : "unsafe", tid, mds);
3101                 mutex_unlock(&mdsc->mutex);
3102                 goto out;
3103         }
3104         if (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags)) {
3105                 pr_warn("got unsafe after safe on %llu from mds%d\n",
3106                            tid, mds);
3107                 mutex_unlock(&mdsc->mutex);
3108                 goto out;
3109         }
3110
3111         result = le32_to_cpu(head->result);
3112
3113         /*
3114          * Handle an ESTALE
3115          * if we're not talking to the authority, send to them
3116          * if the authority has changed while we weren't looking,
3117          * send to new authority
3118          * Otherwise we just have to return an ESTALE
3119          */
3120         if (result == -ESTALE) {
3121                 dout("got ESTALE on request %llu\n", req->r_tid);
3122                 req->r_resend_mds = -1;
3123                 if (req->r_direct_mode != USE_AUTH_MDS) {
3124                         dout("not using auth, setting for that now\n");
3125                         req->r_direct_mode = USE_AUTH_MDS;
3126                         __do_request(mdsc, req);
3127                         mutex_unlock(&mdsc->mutex);
3128                         goto out;
3129                 } else  {
3130                         int mds = __choose_mds(mdsc, req, NULL);
3131                         if (mds >= 0 && mds != req->r_session->s_mds) {
3132                                 dout("but auth changed, so resending\n");
3133                                 __do_request(mdsc, req);
3134                                 mutex_unlock(&mdsc->mutex);
3135                                 goto out;
3136                         }
3137                 }
3138                 dout("have to return ESTALE on request %llu\n", req->r_tid);
3139         }
3140
3141
3142         if (head->safe) {
3143                 set_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags);
3144                 __unregister_request(mdsc, req);
3145
3146                 /* last request during umount? */
3147                 if (mdsc->stopping && !__get_oldest_req(mdsc))
3148                         complete_all(&mdsc->safe_umount_waiters);
3149
3150                 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3151                         /*
3152                          * We already handled the unsafe response, now do the
3153                          * cleanup.  No need to examine the response; the MDS
3154                          * doesn't include any result info in the safe
3155                          * response.  And even if it did, there is nothing
3156                          * useful we could do with a revised return value.
3157                          */
3158                         dout("got safe reply %llu, mds%d\n", tid, mds);
3159
3160                         mutex_unlock(&mdsc->mutex);
3161                         goto out;
3162                 }
3163         } else {
3164                 set_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags);
3165                 list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
3166         }
3167
3168         dout("handle_reply tid %lld result %d\n", tid, result);
3169         rinfo = &req->r_reply_info;
3170         if (test_bit(CEPHFS_FEATURE_REPLY_ENCODING, &session->s_features))
3171                 err = parse_reply_info(session, msg, rinfo, (u64)-1);
3172         else
3173                 err = parse_reply_info(session, msg, rinfo, session->s_con.peer_features);
3174         mutex_unlock(&mdsc->mutex);
3175
3176         mutex_lock(&session->s_mutex);
3177         if (err < 0) {
3178                 pr_err("mdsc_handle_reply got corrupt reply mds%d(tid:%lld)\n", mds, tid);
3179                 ceph_msg_dump(msg);
3180                 goto out_err;
3181         }
3182
3183         /* snap trace */
3184         realm = NULL;
3185         if (rinfo->snapblob_len) {
3186                 down_write(&mdsc->snap_rwsem);
3187                 ceph_update_snap_trace(mdsc, rinfo->snapblob,
3188                                 rinfo->snapblob + rinfo->snapblob_len,
3189                                 le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
3190                                 &realm);
3191                 downgrade_write(&mdsc->snap_rwsem);
3192         } else {
3193                 down_read(&mdsc->snap_rwsem);
3194         }
3195
3196         /* insert trace into our cache */
3197         mutex_lock(&req->r_fill_mutex);
3198         current->journal_info = req;
3199         err = ceph_fill_trace(mdsc->fsc->sb, req);
3200         if (err == 0) {
3201                 if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
3202                                     req->r_op == CEPH_MDS_OP_LSSNAP))
3203                         ceph_readdir_prepopulate(req, req->r_session);
3204         }
3205         current->journal_info = NULL;
3206         mutex_unlock(&req->r_fill_mutex);
3207
3208         up_read(&mdsc->snap_rwsem);
3209         if (realm)
3210                 ceph_put_snap_realm(mdsc, realm);
3211
3212         if (err == 0) {
3213                 if (req->r_target_inode &&
3214                     test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3215                         struct ceph_inode_info *ci =
3216                                 ceph_inode(req->r_target_inode);
3217                         spin_lock(&ci->i_unsafe_lock);
3218                         list_add_tail(&req->r_unsafe_target_item,
3219                                       &ci->i_unsafe_iops);
3220                         spin_unlock(&ci->i_unsafe_lock);
3221                 }
3222
3223                 ceph_unreserve_caps(mdsc, &req->r_caps_reservation);
3224         }
3225 out_err:
3226         mutex_lock(&mdsc->mutex);
3227         if (!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3228                 if (err) {
3229                         req->r_err = err;
3230                 } else {
3231                         req->r_reply =  ceph_msg_get(msg);
3232                         set_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags);
3233                 }
3234         } else {
3235                 dout("reply arrived after request %lld was aborted\n", tid);
3236         }
3237         mutex_unlock(&mdsc->mutex);
3238
3239         mutex_unlock(&session->s_mutex);
3240
3241         /* kick calling process */
3242         complete_request(mdsc, req);
3243
3244         ceph_update_metadata_latency(&mdsc->metric, req->r_start_latency,
3245                                      req->r_end_latency, err);
3246 out:
3247         ceph_mdsc_put_request(req);
3248         return;
3249 }
3250
3251
3252
3253 /*
3254  * handle mds notification that our request has been forwarded.
3255  */
3256 static void handle_forward(struct ceph_mds_client *mdsc,
3257                            struct ceph_mds_session *session,
3258                            struct ceph_msg *msg)
3259 {
3260         struct ceph_mds_request *req;
3261         u64 tid = le64_to_cpu(msg->hdr.tid);
3262         u32 next_mds;
3263         u32 fwd_seq;
3264         int err = -EINVAL;
3265         void *p = msg->front.iov_base;
3266         void *end = p + msg->front.iov_len;
3267
3268         ceph_decode_need(&p, end, 2*sizeof(u32), bad);
3269         next_mds = ceph_decode_32(&p);
3270         fwd_seq = ceph_decode_32(&p);
3271
3272         mutex_lock(&mdsc->mutex);
3273         req = lookup_get_request(mdsc, tid);
3274         if (!req) {
3275                 dout("forward tid %llu to mds%d - req dne\n", tid, next_mds);
3276                 goto out;  /* dup reply? */
3277         }
3278
3279         if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3280                 dout("forward tid %llu aborted, unregistering\n", tid);
3281                 __unregister_request(mdsc, req);
3282         } else if (fwd_seq <= req->r_num_fwd) {
3283                 dout("forward tid %llu to mds%d - old seq %d <= %d\n",
3284                      tid, next_mds, req->r_num_fwd, fwd_seq);
3285         } else {
3286                 /* resend. forward race not possible; mds would drop */
3287                 dout("forward tid %llu to mds%d (we resend)\n", tid, next_mds);
3288                 BUG_ON(req->r_err);
3289                 BUG_ON(test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags));
3290                 req->r_attempts = 0;
3291                 req->r_num_fwd = fwd_seq;
3292                 req->r_resend_mds = next_mds;
3293                 put_request_session(req);
3294                 __do_request(mdsc, req);
3295         }
3296         ceph_mdsc_put_request(req);
3297 out:
3298         mutex_unlock(&mdsc->mutex);
3299         return;
3300
3301 bad:
3302         pr_err("mdsc_handle_forward decode error err=%d\n", err);
3303 }
3304
3305 static int __decode_session_metadata(void **p, void *end,
3306                                      bool *blacklisted)
3307 {
3308         /* map<string,string> */
3309         u32 n;
3310         bool err_str;
3311         ceph_decode_32_safe(p, end, n, bad);
3312         while (n-- > 0) {
3313                 u32 len;
3314                 ceph_decode_32_safe(p, end, len, bad);
3315                 ceph_decode_need(p, end, len, bad);
3316                 err_str = !strncmp(*p, "error_string", len);
3317                 *p += len;
3318                 ceph_decode_32_safe(p, end, len, bad);
3319                 ceph_decode_need(p, end, len, bad);
3320                 if (err_str && strnstr(*p, "blacklisted", len))
3321                         *blacklisted = true;
3322                 *p += len;
3323         }
3324         return 0;
3325 bad:
3326         return -1;
3327 }
3328
3329 /*
3330  * handle a mds session control message
3331  */
3332 static void handle_session(struct ceph_mds_session *session,
3333                            struct ceph_msg *msg)
3334 {
3335         struct ceph_mds_client *mdsc = session->s_mdsc;
3336         int mds = session->s_mds;
3337         int msg_version = le16_to_cpu(msg->hdr.version);
3338         void *p = msg->front.iov_base;
3339         void *end = p + msg->front.iov_len;
3340         struct ceph_mds_session_head *h;
3341         u32 op;
3342         u64 seq, features = 0;
3343         int wake = 0;
3344         bool blacklisted = false;
3345
3346         /* decode */
3347         ceph_decode_need(&p, end, sizeof(*h), bad);
3348         h = p;
3349         p += sizeof(*h);
3350
3351         op = le32_to_cpu(h->op);
3352         seq = le64_to_cpu(h->seq);
3353
3354         if (msg_version >= 3) {
3355                 u32 len;
3356                 /* version >= 2, metadata */
3357                 if (__decode_session_metadata(&p, end, &blacklisted) < 0)
3358                         goto bad;
3359                 /* version >= 3, feature bits */
3360                 ceph_decode_32_safe(&p, end, len, bad);
3361                 ceph_decode_64_safe(&p, end, features, bad);
3362                 p += len - sizeof(features);
3363         }
3364
3365         mutex_lock(&mdsc->mutex);
3366         if (op == CEPH_SESSION_CLOSE) {
3367                 ceph_get_mds_session(session);
3368                 __unregister_session(mdsc, session);
3369         }
3370         /* FIXME: this ttl calculation is generous */
3371         session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;
3372         mutex_unlock(&mdsc->mutex);
3373
3374         mutex_lock(&session->s_mutex);
3375
3376         dout("handle_session mds%d %s %p state %s seq %llu\n",
3377              mds, ceph_session_op_name(op), session,
3378              ceph_session_state_name(session->s_state), seq);
3379
3380         if (session->s_state == CEPH_MDS_SESSION_HUNG) {
3381                 session->s_state = CEPH_MDS_SESSION_OPEN;
3382                 pr_info("mds%d came back\n", session->s_mds);
3383         }
3384
3385         switch (op) {
3386         case CEPH_SESSION_OPEN:
3387                 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3388                         pr_info("mds%d reconnect success\n", session->s_mds);
3389                 session->s_state = CEPH_MDS_SESSION_OPEN;
3390                 session->s_features = features;
3391                 renewed_caps(mdsc, session, 0);
3392                 if (test_bit(CEPHFS_FEATURE_METRIC_COLLECT, &session->s_features))
3393                         metric_schedule_delayed(&mdsc->metric);
3394                 wake = 1;
3395                 if (mdsc->stopping)
3396                         __close_session(mdsc, session);
3397                 break;
3398
3399         case CEPH_SESSION_RENEWCAPS:
3400                 if (session->s_renew_seq == seq)
3401                         renewed_caps(mdsc, session, 1);
3402                 break;
3403
3404         case CEPH_SESSION_CLOSE:
3405                 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3406                         pr_info("mds%d reconnect denied\n", session->s_mds);
3407                 session->s_state = CEPH_MDS_SESSION_CLOSED;
3408                 cleanup_session_requests(mdsc, session);
3409                 remove_session_caps(session);
3410                 wake = 2; /* for good measure */
3411                 wake_up_all(&mdsc->session_close_wq);
3412                 break;
3413
3414         case CEPH_SESSION_STALE:
3415                 pr_info("mds%d caps went stale, renewing\n",
3416                         session->s_mds);
3417                 spin_lock(&session->s_gen_ttl_lock);
3418                 session->s_cap_gen++;
3419                 session->s_cap_ttl = jiffies - 1;
3420                 spin_unlock(&session->s_gen_ttl_lock);
3421                 send_renew_caps(mdsc, session);
3422                 break;
3423
3424         case CEPH_SESSION_RECALL_STATE:
3425                 ceph_trim_caps(mdsc, session, le32_to_cpu(h->max_caps));
3426                 break;
3427
3428         case CEPH_SESSION_FLUSHMSG:
3429                 send_flushmsg_ack(mdsc, session, seq);
3430                 break;
3431
3432         case CEPH_SESSION_FORCE_RO:
3433                 dout("force_session_readonly %p\n", session);
3434                 spin_lock(&session->s_cap_lock);
3435                 session->s_readonly = true;
3436                 spin_unlock(&session->s_cap_lock);
3437                 wake_up_session_caps(session, FORCE_RO);
3438                 break;
3439
3440         case CEPH_SESSION_REJECT:
3441                 WARN_ON(session->s_state != CEPH_MDS_SESSION_OPENING);
3442                 pr_info("mds%d rejected session\n", session->s_mds);
3443                 session->s_state = CEPH_MDS_SESSION_REJECTED;
3444                 cleanup_session_requests(mdsc, session);
3445                 remove_session_caps(session);
3446                 if (blacklisted)
3447                         mdsc->fsc->blacklisted = true;
3448                 wake = 2; /* for good measure */
3449                 break;
3450
3451         default:
3452                 pr_err("mdsc_handle_session bad op %d mds%d\n", op, mds);
3453                 WARN_ON(1);
3454         }
3455
3456         mutex_unlock(&session->s_mutex);
3457         if (wake) {
3458                 mutex_lock(&mdsc->mutex);
3459                 __wake_requests(mdsc, &session->s_waiting);
3460                 if (wake == 2)
3461                         kick_requests(mdsc, mds);
3462                 mutex_unlock(&mdsc->mutex);
3463         }
3464         if (op == CEPH_SESSION_CLOSE)
3465                 ceph_put_mds_session(session);
3466         return;
3467
3468 bad:
3469         pr_err("mdsc_handle_session corrupt message mds%d len %d\n", mds,
3470                (int)msg->front.iov_len);
3471         ceph_msg_dump(msg);
3472         return;
3473 }
3474
3475 void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req)
3476 {
3477         int dcaps;
3478
3479         dcaps = xchg(&req->r_dir_caps, 0);
3480         if (dcaps) {
3481                 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3482                 ceph_put_cap_refs(ceph_inode(req->r_parent), dcaps);
3483         }
3484 }
3485
3486 void ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request *req)
3487 {
3488         int dcaps;
3489
3490         dcaps = xchg(&req->r_dir_caps, 0);
3491         if (dcaps) {
3492                 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3493                 ceph_put_cap_refs_no_check_caps(ceph_inode(req->r_parent),
3494                                                 dcaps);
3495         }
3496 }
3497
3498 /*
3499  * called under session->mutex.
3500  */
3501 static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
3502                                    struct ceph_mds_session *session)
3503 {
3504         struct ceph_mds_request *req, *nreq;
3505         struct rb_node *p;
3506
3507         dout("replay_unsafe_requests mds%d\n", session->s_mds);
3508
3509         mutex_lock(&mdsc->mutex);
3510         list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item)
3511                 __send_request(mdsc, session, req, true);
3512
3513         /*
3514          * also re-send old requests when MDS enters reconnect stage. So that MDS
3515          * can process completed request in clientreplay stage.
3516          */
3517         p = rb_first(&mdsc->request_tree);
3518         while (p) {
3519                 req = rb_entry(p, struct ceph_mds_request, r_node);
3520                 p = rb_next(p);
3521                 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
3522                         continue;
3523                 if (req->r_attempts == 0)
3524                         continue; /* only old requests */
3525                 if (!req->r_session)
3526                         continue;
3527                 if (req->r_session->s_mds != session->s_mds)
3528                         continue;
3529
3530                 ceph_mdsc_release_dir_caps_no_check(req);
3531
3532                 __send_request(mdsc, session, req, true);
3533         }
3534         mutex_unlock(&mdsc->mutex);
3535 }
3536
3537 static int send_reconnect_partial(struct ceph_reconnect_state *recon_state)
3538 {
3539         struct ceph_msg *reply;
3540         struct ceph_pagelist *_pagelist;
3541         struct page *page;
3542         __le32 *addr;
3543         int err = -ENOMEM;
3544
3545         if (!recon_state->allow_multi)
3546                 return -ENOSPC;
3547
3548         /* can't handle message that contains both caps and realm */
3549         BUG_ON(!recon_state->nr_caps == !recon_state->nr_realms);
3550
3551         /* pre-allocate new pagelist */
3552         _pagelist = ceph_pagelist_alloc(GFP_NOFS);
3553         if (!_pagelist)
3554                 return -ENOMEM;
3555
3556         reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3557         if (!reply)
3558                 goto fail_msg;
3559
3560         /* placeholder for nr_caps */
3561         err = ceph_pagelist_encode_32(_pagelist, 0);
3562         if (err < 0)
3563                 goto fail;
3564
3565         if (recon_state->nr_caps) {
3566                 /* currently encoding caps */
3567                 err = ceph_pagelist_encode_32(recon_state->pagelist, 0);
3568                 if (err)
3569                         goto fail;
3570         } else {
3571                 /* placeholder for nr_realms (currently encoding relams) */
3572                 err = ceph_pagelist_encode_32(_pagelist, 0);
3573                 if (err < 0)
3574                         goto fail;
3575         }
3576
3577         err = ceph_pagelist_encode_8(recon_state->pagelist, 1);
3578         if (err)
3579                 goto fail;
3580
3581         page = list_first_entry(&recon_state->pagelist->head, struct page, lru);
3582         addr = kmap_atomic(page);
3583         if (recon_state->nr_caps) {
3584                 /* currently encoding caps */
3585                 *addr = cpu_to_le32(recon_state->nr_caps);
3586         } else {
3587                 /* currently encoding relams */
3588                 *(addr + 1) = cpu_to_le32(recon_state->nr_realms);
3589         }
3590         kunmap_atomic(addr);
3591
3592         reply->hdr.version = cpu_to_le16(5);
3593         reply->hdr.compat_version = cpu_to_le16(4);
3594
3595         reply->hdr.data_len = cpu_to_le32(recon_state->pagelist->length);
3596         ceph_msg_data_add_pagelist(reply, recon_state->pagelist);
3597
3598         ceph_con_send(&recon_state->session->s_con, reply);
3599         ceph_pagelist_release(recon_state->pagelist);
3600
3601         recon_state->pagelist = _pagelist;
3602         recon_state->nr_caps = 0;
3603         recon_state->nr_realms = 0;
3604         recon_state->msg_version = 5;
3605         return 0;
3606 fail:
3607         ceph_msg_put(reply);
3608 fail_msg:
3609         ceph_pagelist_release(_pagelist);
3610         return err;
3611 }
3612
3613 /*
3614  * Encode information about a cap for a reconnect with the MDS.
3615  */
3616 static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
3617                           void *arg)
3618 {
3619         union {
3620                 struct ceph_mds_cap_reconnect v2;
3621                 struct ceph_mds_cap_reconnect_v1 v1;
3622         } rec;
3623         struct ceph_inode_info *ci = cap->ci;
3624         struct ceph_reconnect_state *recon_state = arg;
3625         struct ceph_pagelist *pagelist = recon_state->pagelist;
3626         int err;
3627         u64 snap_follows;
3628
3629         dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
3630              inode, ceph_vinop(inode), cap, cap->cap_id,
3631              ceph_cap_string(cap->issued));
3632
3633         spin_lock(&ci->i_ceph_lock);
3634         cap->seq = 0;        /* reset cap seq */
3635         cap->issue_seq = 0;  /* and issue_seq */
3636         cap->mseq = 0;       /* and migrate_seq */
3637         cap->cap_gen = cap->session->s_cap_gen;
3638
3639         /* These are lost when the session goes away */
3640         if (S_ISDIR(inode->i_mode)) {
3641                 if (cap->issued & CEPH_CAP_DIR_CREATE) {
3642                         ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns));
3643                         memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout));
3644                 }
3645                 cap->issued &= ~CEPH_CAP_ANY_DIR_OPS;
3646         }
3647
3648         if (recon_state->msg_version >= 2) {
3649                 rec.v2.cap_id = cpu_to_le64(cap->cap_id);
3650                 rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3651                 rec.v2.issued = cpu_to_le32(cap->issued);
3652                 rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3653                 rec.v2.pathbase = 0;
3654                 rec.v2.flock_len = (__force __le32)
3655                         ((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1);
3656         } else {
3657                 rec.v1.cap_id = cpu_to_le64(cap->cap_id);
3658                 rec.v1.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3659                 rec.v1.issued = cpu_to_le32(cap->issued);
3660                 rec.v1.size = cpu_to_le64(inode->i_size);
3661                 ceph_encode_timespec64(&rec.v1.mtime, &inode->i_mtime);
3662                 ceph_encode_timespec64(&rec.v1.atime, &inode->i_atime);
3663                 rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3664                 rec.v1.pathbase = 0;
3665         }
3666
3667         if (list_empty(&ci->i_cap_snaps)) {
3668                 snap_follows = ci->i_head_snapc ? ci->i_head_snapc->seq : 0;
3669         } else {
3670                 struct ceph_cap_snap *capsnap =
3671                         list_first_entry(&ci->i_cap_snaps,
3672                                          struct ceph_cap_snap, ci_item);
3673                 snap_follows = capsnap->follows;
3674         }
3675         spin_unlock(&ci->i_ceph_lock);
3676
3677         if (recon_state->msg_version >= 2) {
3678                 int num_fcntl_locks, num_flock_locks;
3679                 struct ceph_filelock *flocks = NULL;
3680                 size_t struct_len, total_len = sizeof(u64);
3681                 u8 struct_v = 0;
3682
3683 encode_again:
3684                 if (rec.v2.flock_len) {
3685                         ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
3686                 } else {
3687                         num_fcntl_locks = 0;
3688                         num_flock_locks = 0;
3689                 }
3690                 if (num_fcntl_locks + num_flock_locks > 0) {
3691                         flocks = kmalloc_array(num_fcntl_locks + num_flock_locks,
3692                                                sizeof(struct ceph_filelock),
3693                                                GFP_NOFS);
3694                         if (!flocks) {
3695                                 err = -ENOMEM;
3696                                 goto out_err;
3697                         }
3698                         err = ceph_encode_locks_to_buffer(inode, flocks,
3699                                                           num_fcntl_locks,
3700                                                           num_flock_locks);
3701                         if (err) {
3702                                 kfree(flocks);
3703                                 flocks = NULL;
3704                                 if (err == -ENOSPC)
3705                                         goto encode_again;
3706                                 goto out_err;
3707                         }
3708                 } else {
3709                         kfree(flocks);
3710                         flocks = NULL;
3711                 }
3712
3713                 if (recon_state->msg_version >= 3) {
3714                         /* version, compat_version and struct_len */
3715                         total_len += 2 * sizeof(u8) + sizeof(u32);
3716                         struct_v = 2;
3717                 }
3718                 /*
3719                  * number of encoded locks is stable, so copy to pagelist
3720                  */
3721                 struct_len = 2 * sizeof(u32) +
3722                             (num_fcntl_locks + num_flock_locks) *
3723                             sizeof(struct ceph_filelock);
3724                 rec.v2.flock_len = cpu_to_le32(struct_len);
3725
3726                 struct_len += sizeof(u32) + sizeof(rec.v2);
3727
3728                 if (struct_v >= 2)
3729                         struct_len += sizeof(u64); /* snap_follows */
3730
3731                 total_len += struct_len;
3732
3733                 if (pagelist->length + total_len > RECONNECT_MAX_SIZE) {
3734                         err = send_reconnect_partial(recon_state);
3735                         if (err)
3736                                 goto out_freeflocks;
3737                         pagelist = recon_state->pagelist;
3738                 }
3739
3740                 err = ceph_pagelist_reserve(pagelist, total_len);
3741                 if (err)
3742                         goto out_freeflocks;
3743
3744                 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
3745                 if (recon_state->msg_version >= 3) {
3746                         ceph_pagelist_encode_8(pagelist, struct_v);
3747                         ceph_pagelist_encode_8(pagelist, 1);
3748                         ceph_pagelist_encode_32(pagelist, struct_len);
3749                 }
3750                 ceph_pagelist_encode_string(pagelist, NULL, 0);
3751                 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v2));
3752                 ceph_locks_to_pagelist(flocks, pagelist,
3753                                        num_fcntl_locks, num_flock_locks);
3754                 if (struct_v >= 2)
3755                         ceph_pagelist_encode_64(pagelist, snap_follows);
3756 out_freeflocks:
3757                 kfree(flocks);
3758         } else {
3759                 u64 pathbase = 0;
3760                 int pathlen = 0;
3761                 char *path = NULL;
3762                 struct dentry *dentry;
3763
3764                 dentry = d_find_alias(inode);
3765                 if (dentry) {
3766                         path = ceph_mdsc_build_path(dentry,
3767                                                 &pathlen, &pathbase, 0);
3768                         dput(dentry);
3769                         if (IS_ERR(path)) {
3770                                 err = PTR_ERR(path);
3771                                 goto out_err;
3772                         }
3773                         rec.v1.pathbase = cpu_to_le64(pathbase);
3774                 }
3775
3776                 err = ceph_pagelist_reserve(pagelist,
3777                                             sizeof(u64) + sizeof(u32) +
3778                                             pathlen + sizeof(rec.v1));
3779                 if (err) {
3780                         goto out_freepath;
3781                 }
3782
3783                 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
3784                 ceph_pagelist_encode_string(pagelist, path, pathlen);
3785                 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v1));
3786 out_freepath:
3787                 ceph_mdsc_free_path(path, pathlen);
3788         }
3789
3790 out_err:
3791         if (err >= 0)
3792                 recon_state->nr_caps++;
3793         return err;
3794 }
3795
3796 static int encode_snap_realms(struct ceph_mds_client *mdsc,
3797                               struct ceph_reconnect_state *recon_state)
3798 {
3799         struct rb_node *p;
3800         struct ceph_pagelist *pagelist = recon_state->pagelist;
3801         int err = 0;
3802
3803         if (recon_state->msg_version >= 4) {
3804                 err = ceph_pagelist_encode_32(pagelist, mdsc->num_snap_realms);
3805                 if (err < 0)
3806                         goto fail;
3807         }
3808
3809         /*
3810          * snaprealms.  we provide mds with the ino, seq (version), and
3811          * parent for all of our realms.  If the mds has any newer info,
3812          * it will tell us.
3813          */
3814         for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {
3815                 struct ceph_snap_realm *realm =
3816                        rb_entry(p, struct ceph_snap_realm, node);
3817                 struct ceph_mds_snaprealm_reconnect sr_rec;
3818
3819                 if (recon_state->msg_version >= 4) {
3820                         size_t need = sizeof(u8) * 2 + sizeof(u32) +
3821                                       sizeof(sr_rec);
3822
3823                         if (pagelist->length + need > RECONNECT_MAX_SIZE) {
3824                                 err = send_reconnect_partial(recon_state);
3825                                 if (err)
3826                                         goto fail;
3827                                 pagelist = recon_state->pagelist;
3828                         }
3829
3830                         err = ceph_pagelist_reserve(pagelist, need);
3831                         if (err)
3832                                 goto fail;
3833
3834                         ceph_pagelist_encode_8(pagelist, 1);
3835                         ceph_pagelist_encode_8(pagelist, 1);
3836                         ceph_pagelist_encode_32(pagelist, sizeof(sr_rec));
3837                 }
3838
3839                 dout(" adding snap realm %llx seq %lld parent %llx\n",
3840                      realm->ino, realm->seq, realm->parent_ino);
3841                 sr_rec.ino = cpu_to_le64(realm->ino);
3842                 sr_rec.seq = cpu_to_le64(realm->seq);
3843                 sr_rec.parent = cpu_to_le64(realm->parent_ino);
3844
3845                 err = ceph_pagelist_append(pagelist, &sr_rec, sizeof(sr_rec));
3846                 if (err)
3847                         goto fail;
3848
3849                 recon_state->nr_realms++;
3850         }
3851 fail:
3852         return err;
3853 }
3854
3855
3856 /*
3857  * If an MDS fails and recovers, clients need to reconnect in order to
3858  * reestablish shared state.  This includes all caps issued through
3859  * this session _and_ the snap_realm hierarchy.  Because it's not
3860  * clear which snap realms the mds cares about, we send everything we
3861  * know about.. that ensures we'll then get any new info the
3862  * recovering MDS might have.
3863  *
3864  * This is a relatively heavyweight operation, but it's rare.
3865  */
3866 static void send_mds_reconnect(struct ceph_mds_client *mdsc,
3867                                struct ceph_mds_session *session)
3868 {
3869         struct ceph_msg *reply;
3870         int mds = session->s_mds;
3871         int err = -ENOMEM;
3872         struct ceph_reconnect_state recon_state = {
3873                 .session = session,
3874         };
3875         LIST_HEAD(dispose);
3876
3877         pr_info("mds%d reconnect start\n", mds);
3878
3879         recon_state.pagelist = ceph_pagelist_alloc(GFP_NOFS);
3880         if (!recon_state.pagelist)
3881                 goto fail_nopagelist;
3882
3883         reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3884         if (!reply)
3885                 goto fail_nomsg;
3886
3887         xa_destroy(&session->s_delegated_inos);
3888
3889         mutex_lock(&session->s_mutex);
3890         session->s_state = CEPH_MDS_SESSION_RECONNECTING;
3891         session->s_seq = 0;
3892
3893         dout("session %p state %s\n", session,
3894              ceph_session_state_name(session->s_state));
3895
3896         spin_lock(&session->s_gen_ttl_lock);
3897         session->s_cap_gen++;
3898         spin_unlock(&session->s_gen_ttl_lock);
3899
3900         spin_lock(&session->s_cap_lock);
3901         /* don't know if session is readonly */
3902         session->s_readonly = 0;
3903         /*
3904          * notify __ceph_remove_cap() that we are composing cap reconnect.
3905          * If a cap get released before being added to the cap reconnect,
3906          * __ceph_remove_cap() should skip queuing cap release.
3907          */
3908         session->s_cap_reconnect = 1;
3909         /* drop old cap expires; we're about to reestablish that state */
3910         detach_cap_releases(session, &dispose);
3911         spin_unlock(&session->s_cap_lock);
3912         dispose_cap_releases(mdsc, &dispose);
3913
3914         /* trim unused caps to reduce MDS's cache rejoin time */
3915         if (mdsc->fsc->sb->s_root)
3916                 shrink_dcache_parent(mdsc->fsc->sb->s_root);
3917
3918         ceph_con_close(&session->s_con);
3919         ceph_con_open(&session->s_con,
3920                       CEPH_ENTITY_TYPE_MDS, mds,
3921                       ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
3922
3923         /* replay unsafe requests */
3924         replay_unsafe_requests(mdsc, session);
3925
3926         ceph_early_kick_flushing_caps(mdsc, session);
3927
3928         down_read(&mdsc->snap_rwsem);
3929
3930         /* placeholder for nr_caps */
3931         err = ceph_pagelist_encode_32(recon_state.pagelist, 0);
3932         if (err)
3933                 goto fail;
3934
3935         if (test_bit(CEPHFS_FEATURE_MULTI_RECONNECT, &session->s_features)) {
3936                 recon_state.msg_version = 3;
3937                 recon_state.allow_multi = true;
3938         } else if (session->s_con.peer_features & CEPH_FEATURE_MDSENC) {
3939                 recon_state.msg_version = 3;
3940         } else {
3941                 recon_state.msg_version = 2;
3942         }
3943         /* trsaverse this session's caps */
3944         err = ceph_iterate_session_caps(session, reconnect_caps_cb, &recon_state);
3945
3946         spin_lock(&session->s_cap_lock);
3947         session->s_cap_reconnect = 0;
3948         spin_unlock(&session->s_cap_lock);
3949
3950         if (err < 0)
3951                 goto fail;
3952
3953         /* check if all realms can be encoded into current message */
3954         if (mdsc->num_snap_realms) {
3955                 size_t total_len =
3956                         recon_state.pagelist->length +
3957                         mdsc->num_snap_realms *
3958                         sizeof(struct ceph_mds_snaprealm_reconnect);
3959                 if (recon_state.msg_version >= 4) {
3960                         /* number of realms */
3961                         total_len += sizeof(u32);
3962                         /* version, compat_version and struct_len */
3963                         total_len += mdsc->num_snap_realms *
3964                                      (2 * sizeof(u8) + sizeof(u32));
3965                 }
3966                 if (total_len > RECONNECT_MAX_SIZE) {
3967                         if (!recon_state.allow_multi) {
3968                                 err = -ENOSPC;
3969                                 goto fail;
3970                         }
3971                         if (recon_state.nr_caps) {
3972                                 err = send_reconnect_partial(&recon_state);
3973                                 if (err)
3974                                         goto fail;
3975                         }
3976                         recon_state.msg_version = 5;
3977                 }
3978         }
3979
3980         err = encode_snap_realms(mdsc, &recon_state);
3981         if (err < 0)
3982                 goto fail;
3983
3984         if (recon_state.msg_version >= 5) {
3985                 err = ceph_pagelist_encode_8(recon_state.pagelist, 0);
3986                 if (err < 0)
3987                         goto fail;
3988         }
3989
3990         if (recon_state.nr_caps || recon_state.nr_realms) {
3991                 struct page *page =
3992                         list_first_entry(&recon_state.pagelist->head,
3993                                         struct page, lru);
3994                 __le32 *addr = kmap_atomic(page);
3995                 if (recon_state.nr_caps) {
3996                         WARN_ON(recon_state.nr_realms != mdsc->num_snap_realms);
3997                         *addr = cpu_to_le32(recon_state.nr_caps);
3998                 } else if (recon_state.msg_version >= 4) {
3999                         *(addr + 1) = cpu_to_le32(recon_state.nr_realms);
4000                 }
4001                 kunmap_atomic(addr);
4002         }
4003
4004         reply->hdr.version = cpu_to_le16(recon_state.msg_version);
4005         if (recon_state.msg_version >= 4)
4006                 reply->hdr.compat_version = cpu_to_le16(4);
4007
4008         reply->hdr.data_len = cpu_to_le32(recon_state.pagelist->length);
4009         ceph_msg_data_add_pagelist(reply, recon_state.pagelist);
4010
4011         ceph_con_send(&session->s_con, reply);
4012
4013         mutex_unlock(&session->s_mutex);
4014
4015         mutex_lock(&mdsc->mutex);
4016         __wake_requests(mdsc, &session->s_waiting);
4017         mutex_unlock(&mdsc->mutex);
4018
4019         up_read(&mdsc->snap_rwsem);
4020         ceph_pagelist_release(recon_state.pagelist);
4021         return;
4022
4023 fail:
4024         ceph_msg_put(reply);
4025         up_read(&mdsc->snap_rwsem);
4026         mutex_unlock(&session->s_mutex);
4027 fail_nomsg:
4028         ceph_pagelist_release(recon_state.pagelist);
4029 fail_nopagelist:
4030         pr_err("error %d preparing reconnect for mds%d\n", err, mds);
4031         return;
4032 }
4033
4034
4035 /*
4036  * compare old and new mdsmaps, kicking requests
4037  * and closing out old connections as necessary
4038  *
4039  * called under mdsc->mutex.
4040  */
4041 static void check_new_map(struct ceph_mds_client *mdsc,
4042                           struct ceph_mdsmap *newmap,
4043                           struct ceph_mdsmap *oldmap)
4044 {
4045         int i;
4046         int oldstate, newstate;
4047         struct ceph_mds_session *s;
4048
4049         dout("check_new_map new %u old %u\n",
4050              newmap->m_epoch, oldmap->m_epoch);
4051
4052         for (i = 0; i < oldmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4053                 if (!mdsc->sessions[i])
4054                         continue;
4055                 s = mdsc->sessions[i];
4056                 oldstate = ceph_mdsmap_get_state(oldmap, i);
4057                 newstate = ceph_mdsmap_get_state(newmap, i);
4058
4059                 dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",
4060                      i, ceph_mds_state_name(oldstate),
4061                      ceph_mdsmap_is_laggy(oldmap, i) ? " (laggy)" : "",
4062                      ceph_mds_state_name(newstate),
4063                      ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
4064                      ceph_session_state_name(s->s_state));
4065
4066                 if (i >= newmap->possible_max_rank) {
4067                         /* force close session for stopped mds */
4068                         ceph_get_mds_session(s);
4069                         __unregister_session(mdsc, s);
4070                         __wake_requests(mdsc, &s->s_waiting);
4071                         mutex_unlock(&mdsc->mutex);
4072
4073                         mutex_lock(&s->s_mutex);
4074                         cleanup_session_requests(mdsc, s);
4075                         remove_session_caps(s);
4076                         mutex_unlock(&s->s_mutex);
4077
4078                         ceph_put_mds_session(s);
4079
4080                         mutex_lock(&mdsc->mutex);
4081                         kick_requests(mdsc, i);
4082                         continue;
4083                 }
4084
4085                 if (memcmp(ceph_mdsmap_get_addr(oldmap, i),
4086                            ceph_mdsmap_get_addr(newmap, i),
4087                            sizeof(struct ceph_entity_addr))) {
4088                         /* just close it */
4089                         mutex_unlock(&mdsc->mutex);
4090                         mutex_lock(&s->s_mutex);
4091                         mutex_lock(&mdsc->mutex);
4092                         ceph_con_close(&s->s_con);
4093                         mutex_unlock(&s->s_mutex);
4094                         s->s_state = CEPH_MDS_SESSION_RESTARTING;
4095                 } else if (oldstate == newstate) {
4096                         continue;  /* nothing new with this mds */
4097                 }
4098
4099                 /*
4100                  * send reconnect?
4101                  */
4102                 if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&
4103                     newstate >= CEPH_MDS_STATE_RECONNECT) {
4104                         mutex_unlock(&mdsc->mutex);
4105                         send_mds_reconnect(mdsc, s);
4106                         mutex_lock(&mdsc->mutex);
4107                 }
4108
4109                 /*
4110                  * kick request on any mds that has gone active.
4111                  */
4112                 if (oldstate < CEPH_MDS_STATE_ACTIVE &&
4113                     newstate >= CEPH_MDS_STATE_ACTIVE) {
4114                         if (oldstate != CEPH_MDS_STATE_CREATING &&
4115                             oldstate != CEPH_MDS_STATE_STARTING)
4116                                 pr_info("mds%d recovery completed\n", s->s_mds);
4117                         kick_requests(mdsc, i);
4118                         mutex_unlock(&mdsc->mutex);
4119                         mutex_lock(&s->s_mutex);
4120                         mutex_lock(&mdsc->mutex);
4121                         ceph_kick_flushing_caps(mdsc, s);
4122                         mutex_unlock(&s->s_mutex);
4123                         wake_up_session_caps(s, RECONNECT);
4124                 }
4125         }
4126
4127         for (i = 0; i < newmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4128                 s = mdsc->sessions[i];
4129                 if (!s)
4130                         continue;
4131                 if (!ceph_mdsmap_is_laggy(newmap, i))
4132                         continue;
4133                 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4134                     s->s_state == CEPH_MDS_SESSION_HUNG ||
4135                     s->s_state == CEPH_MDS_SESSION_CLOSING) {
4136                         dout(" connecting to export targets of laggy mds%d\n",
4137                              i);
4138                         __open_export_target_sessions(mdsc, s);
4139                 }
4140         }
4141 }
4142
4143
4144
4145 /*
4146  * leases
4147  */
4148
4149 /*
4150  * caller must hold session s_mutex, dentry->d_lock
4151  */
4152 void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry)
4153 {
4154         struct ceph_dentry_info *di = ceph_dentry(dentry);
4155
4156         ceph_put_mds_session(di->lease_session);
4157         di->lease_session = NULL;
4158 }
4159
4160 static void handle_lease(struct ceph_mds_client *mdsc,
4161                          struct ceph_mds_session *session,
4162                          struct ceph_msg *msg)
4163 {
4164         struct super_block *sb = mdsc->fsc->sb;
4165         struct inode *inode;
4166         struct dentry *parent, *dentry;
4167         struct ceph_dentry_info *di;
4168         int mds = session->s_mds;
4169         struct ceph_mds_lease *h = msg->front.iov_base;
4170         u32 seq;
4171         struct ceph_vino vino;
4172         struct qstr dname;
4173         int release = 0;
4174
4175         dout("handle_lease from mds%d\n", mds);
4176
4177         /* decode */
4178         if (msg->front.iov_len < sizeof(*h) + sizeof(u32))
4179                 goto bad;
4180         vino.ino = le64_to_cpu(h->ino);
4181         vino.snap = CEPH_NOSNAP;
4182         seq = le32_to_cpu(h->seq);
4183         dname.len = get_unaligned_le32(h + 1);
4184         if (msg->front.iov_len < sizeof(*h) + sizeof(u32) + dname.len)
4185                 goto bad;
4186         dname.name = (void *)(h + 1) + sizeof(u32);
4187
4188         /* lookup inode */
4189         inode = ceph_find_inode(sb, vino);
4190         dout("handle_lease %s, ino %llx %p %.*s\n",
4191              ceph_lease_op_name(h->action), vino.ino, inode,
4192              dname.len, dname.name);
4193
4194         mutex_lock(&session->s_mutex);
4195         session->s_seq++;
4196
4197         if (!inode) {
4198                 dout("handle_lease no inode %llx\n", vino.ino);
4199                 goto release;
4200         }
4201
4202         /* dentry */
4203         parent = d_find_alias(inode);
4204         if (!parent) {
4205                 dout("no parent dentry on inode %p\n", inode);
4206                 WARN_ON(1);
4207                 goto release;  /* hrm... */
4208         }
4209         dname.hash = full_name_hash(parent, dname.name, dname.len);
4210         dentry = d_lookup(parent, &dname);
4211         dput(parent);
4212         if (!dentry)
4213                 goto release;
4214
4215         spin_lock(&dentry->d_lock);
4216         di = ceph_dentry(dentry);
4217         switch (h->action) {
4218         case CEPH_MDS_LEASE_REVOKE:
4219                 if (di->lease_session == session) {
4220                         if (ceph_seq_cmp(di->lease_seq, seq) > 0)
4221                                 h->seq = cpu_to_le32(di->lease_seq);
4222                         __ceph_mdsc_drop_dentry_lease(dentry);
4223                 }
4224                 release = 1;
4225                 break;
4226
4227         case CEPH_MDS_LEASE_RENEW:
4228                 if (di->lease_session == session &&
4229                     di->lease_gen == session->s_cap_gen &&
4230                     di->lease_renew_from &&
4231                     di->lease_renew_after == 0) {
4232                         unsigned long duration =
4233                                 msecs_to_jiffies(le32_to_cpu(h->duration_ms));
4234
4235                         di->lease_seq = seq;
4236                         di->time = di->lease_renew_from + duration;
4237                         di->lease_renew_after = di->lease_renew_from +
4238                                 (duration >> 1);
4239                         di->lease_renew_from = 0;
4240                 }
4241                 break;
4242         }
4243         spin_unlock(&dentry->d_lock);
4244         dput(dentry);
4245
4246         if (!release)
4247                 goto out;
4248
4249 release:
4250         /* let's just reuse the same message */
4251         h->action = CEPH_MDS_LEASE_REVOKE_ACK;
4252         ceph_msg_get(msg);
4253         ceph_con_send(&session->s_con, msg);
4254
4255 out:
4256         mutex_unlock(&session->s_mutex);
4257         /* avoid calling iput_final() in mds dispatch threads */
4258         ceph_async_iput(inode);
4259         return;
4260
4261 bad:
4262         pr_err("corrupt lease message\n");
4263         ceph_msg_dump(msg);
4264 }
4265
4266 void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
4267                               struct dentry *dentry, char action,
4268                               u32 seq)
4269 {
4270         struct ceph_msg *msg;
4271         struct ceph_mds_lease *lease;
4272         struct inode *dir;
4273         int len = sizeof(*lease) + sizeof(u32) + NAME_MAX;
4274
4275         dout("lease_send_msg identry %p %s to mds%d\n",
4276              dentry, ceph_lease_op_name(action), session->s_mds);
4277
4278         msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
4279         if (!msg)
4280                 return;
4281         lease = msg->front.iov_base;
4282         lease->action = action;
4283         lease->seq = cpu_to_le32(seq);
4284
4285         spin_lock(&dentry->d_lock);
4286         dir = d_inode(dentry->d_parent);
4287         lease->ino = cpu_to_le64(ceph_ino(dir));
4288         lease->first = lease->last = cpu_to_le64(ceph_snap(dir));
4289
4290         put_unaligned_le32(dentry->d_name.len, lease + 1);
4291         memcpy((void *)(lease + 1) + 4,
4292                dentry->d_name.name, dentry->d_name.len);
4293         spin_unlock(&dentry->d_lock);
4294         /*
4295          * if this is a preemptive lease RELEASE, no need to
4296          * flush request stream, since the actual request will
4297          * soon follow.
4298          */
4299         msg->more_to_follow = (action == CEPH_MDS_LEASE_RELEASE);
4300
4301         ceph_con_send(&session->s_con, msg);
4302 }
4303
4304 /*
4305  * lock unlock sessions, to wait ongoing session activities
4306  */
4307 static void lock_unlock_sessions(struct ceph_mds_client *mdsc)
4308 {
4309         int i;
4310
4311         mutex_lock(&mdsc->mutex);
4312         for (i = 0; i < mdsc->max_sessions; i++) {
4313                 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
4314                 if (!s)
4315                         continue;
4316                 mutex_unlock(&mdsc->mutex);
4317                 mutex_lock(&s->s_mutex);
4318                 mutex_unlock(&s->s_mutex);
4319                 ceph_put_mds_session(s);
4320                 mutex_lock(&mdsc->mutex);
4321         }
4322         mutex_unlock(&mdsc->mutex);
4323 }
4324
4325 static void maybe_recover_session(struct ceph_mds_client *mdsc)
4326 {
4327         struct ceph_fs_client *fsc = mdsc->fsc;
4328
4329         if (!ceph_test_mount_opt(fsc, CLEANRECOVER))
4330                 return;
4331
4332         if (READ_ONCE(fsc->mount_state) != CEPH_MOUNT_MOUNTED)
4333                 return;
4334
4335         if (!READ_ONCE(fsc->blacklisted))
4336                 return;
4337
4338         if (fsc->last_auto_reconnect &&
4339             time_before(jiffies, fsc->last_auto_reconnect + HZ * 60 * 30))
4340                 return;
4341
4342         pr_info("auto reconnect after blacklisted\n");
4343         fsc->last_auto_reconnect = jiffies;
4344         ceph_force_reconnect(fsc->sb);
4345 }
4346
4347 bool check_session_state(struct ceph_mds_session *s)
4348 {
4349         if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
4350                 dout("resending session close request for mds%d\n",
4351                                 s->s_mds);
4352                 request_close_session(s);
4353                 return false;
4354         }
4355         if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
4356                 if (s->s_state == CEPH_MDS_SESSION_OPEN) {
4357                         s->s_state = CEPH_MDS_SESSION_HUNG;
4358                         pr_info("mds%d hung\n", s->s_mds);
4359                 }
4360         }
4361         if (s->s_state == CEPH_MDS_SESSION_NEW ||
4362             s->s_state == CEPH_MDS_SESSION_RESTARTING ||
4363             s->s_state == CEPH_MDS_SESSION_CLOSED ||
4364             s->s_state == CEPH_MDS_SESSION_REJECTED)
4365                 /* this mds is failed or recovering, just wait */
4366                 return false;
4367
4368         return true;
4369 }
4370
4371 /*
4372  * delayed work -- periodically trim expired leases, renew caps with mds
4373  */
4374 static void schedule_delayed(struct ceph_mds_client *mdsc)
4375 {
4376         int delay = 5;
4377         unsigned hz = round_jiffies_relative(HZ * delay);
4378         schedule_delayed_work(&mdsc->delayed_work, hz);
4379 }
4380
4381 static void delayed_work(struct work_struct *work)
4382 {
4383         int i;
4384         struct ceph_mds_client *mdsc =
4385                 container_of(work, struct ceph_mds_client, delayed_work.work);
4386         int renew_interval;
4387         int renew_caps;
4388
4389         dout("mdsc delayed_work\n");
4390
4391         if (mdsc->stopping)
4392                 return;
4393
4394         mutex_lock(&mdsc->mutex);
4395         renew_interval = mdsc->mdsmap->m_session_timeout >> 2;
4396         renew_caps = time_after_eq(jiffies, HZ*renew_interval +
4397                                    mdsc->last_renew_caps);
4398         if (renew_caps)
4399                 mdsc->last_renew_caps = jiffies;
4400
4401         for (i = 0; i < mdsc->max_sessions; i++) {
4402                 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
4403                 if (!s)
4404                         continue;
4405
4406                 if (!check_session_state(s)) {
4407                         ceph_put_mds_session(s);
4408                         continue;
4409                 }
4410                 mutex_unlock(&mdsc->mutex);
4411
4412                 mutex_lock(&s->s_mutex);
4413                 if (renew_caps)
4414                         send_renew_caps(mdsc, s);
4415                 else
4416                         ceph_con_keepalive(&s->s_con);
4417                 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4418                     s->s_state == CEPH_MDS_SESSION_HUNG)
4419                         ceph_send_cap_releases(mdsc, s);
4420                 mutex_unlock(&s->s_mutex);
4421                 ceph_put_mds_session(s);
4422
4423                 mutex_lock(&mdsc->mutex);
4424         }
4425         mutex_unlock(&mdsc->mutex);
4426
4427         ceph_check_delayed_caps(mdsc);
4428
4429         ceph_queue_cap_reclaim_work(mdsc);
4430
4431         ceph_trim_snapid_map(mdsc);
4432
4433         maybe_recover_session(mdsc);
4434
4435         schedule_delayed(mdsc);
4436 }
4437
4438 int ceph_mdsc_init(struct ceph_fs_client *fsc)
4439
4440 {
4441         struct ceph_mds_client *mdsc;
4442         int err;
4443
4444         mdsc = kzalloc(sizeof(struct ceph_mds_client), GFP_NOFS);
4445         if (!mdsc)
4446                 return -ENOMEM;
4447         mdsc->fsc = fsc;
4448         mutex_init(&mdsc->mutex);
4449         mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);
4450         if (!mdsc->mdsmap) {
4451                 err = -ENOMEM;
4452                 goto err_mdsc;
4453         }
4454
4455         init_completion(&mdsc->safe_umount_waiters);
4456         init_waitqueue_head(&mdsc->session_close_wq);
4457         INIT_LIST_HEAD(&mdsc->waiting_for_map);
4458         mdsc->sessions = NULL;
4459         atomic_set(&mdsc->num_sessions, 0);
4460         mdsc->max_sessions = 0;
4461         mdsc->stopping = 0;
4462         atomic64_set(&mdsc->quotarealms_count, 0);
4463         mdsc->quotarealms_inodes = RB_ROOT;
4464         mutex_init(&mdsc->quotarealms_inodes_mutex);
4465         mdsc->last_snap_seq = 0;
4466         init_rwsem(&mdsc->snap_rwsem);
4467         mdsc->snap_realms = RB_ROOT;
4468         INIT_LIST_HEAD(&mdsc->snap_empty);
4469         mdsc->num_snap_realms = 0;
4470         spin_lock_init(&mdsc->snap_empty_lock);
4471         mdsc->last_tid = 0;
4472         mdsc->oldest_tid = 0;
4473         mdsc->request_tree = RB_ROOT;
4474         INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);
4475         mdsc->last_renew_caps = jiffies;
4476         INIT_LIST_HEAD(&mdsc->cap_delay_list);
4477         INIT_LIST_HEAD(&mdsc->cap_wait_list);
4478         spin_lock_init(&mdsc->cap_delay_lock);
4479         INIT_LIST_HEAD(&mdsc->snap_flush_list);
4480         spin_lock_init(&mdsc->snap_flush_lock);
4481         mdsc->last_cap_flush_tid = 1;
4482         INIT_LIST_HEAD(&mdsc->cap_flush_list);
4483         INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
4484         mdsc->num_cap_flushing = 0;
4485         spin_lock_init(&mdsc->cap_dirty_lock);
4486         init_waitqueue_head(&mdsc->cap_flushing_wq);
4487         INIT_WORK(&mdsc->cap_reclaim_work, ceph_cap_reclaim_work);
4488         atomic_set(&mdsc->cap_reclaim_pending, 0);
4489         err = ceph_metric_init(&mdsc->metric);
4490         if (err)
4491                 goto err_mdsmap;
4492
4493         spin_lock_init(&mdsc->dentry_list_lock);
4494         INIT_LIST_HEAD(&mdsc->dentry_leases);
4495         INIT_LIST_HEAD(&mdsc->dentry_dir_leases);
4496
4497         ceph_caps_init(mdsc);
4498         ceph_adjust_caps_max_min(mdsc, fsc->mount_options);
4499
4500         spin_lock_init(&mdsc->snapid_map_lock);
4501         mdsc->snapid_map_tree = RB_ROOT;
4502         INIT_LIST_HEAD(&mdsc->snapid_map_lru);
4503
4504         init_rwsem(&mdsc->pool_perm_rwsem);
4505         mdsc->pool_perm_tree = RB_ROOT;
4506
4507         strscpy(mdsc->nodename, utsname()->nodename,
4508                 sizeof(mdsc->nodename));
4509
4510         fsc->mdsc = mdsc;
4511         return 0;
4512
4513 err_mdsmap:
4514         kfree(mdsc->mdsmap);
4515 err_mdsc:
4516         kfree(mdsc);
4517         return err;
4518 }
4519
4520 /*
4521  * Wait for safe replies on open mds requests.  If we time out, drop
4522  * all requests from the tree to avoid dangling dentry refs.
4523  */
4524 static void wait_requests(struct ceph_mds_client *mdsc)
4525 {
4526         struct ceph_options *opts = mdsc->fsc->client->options;
4527         struct ceph_mds_request *req;
4528
4529         mutex_lock(&mdsc->mutex);
4530         if (__get_oldest_req(mdsc)) {
4531                 mutex_unlock(&mdsc->mutex);
4532
4533                 dout("wait_requests waiting for requests\n");
4534                 wait_for_completion_timeout(&mdsc->safe_umount_waiters,
4535                                     ceph_timeout_jiffies(opts->mount_timeout));
4536
4537                 /* tear down remaining requests */
4538                 mutex_lock(&mdsc->mutex);
4539                 while ((req = __get_oldest_req(mdsc))) {
4540                         dout("wait_requests timed out on tid %llu\n",
4541                              req->r_tid);
4542                         list_del_init(&req->r_wait);
4543                         __unregister_request(mdsc, req);
4544                 }
4545         }
4546         mutex_unlock(&mdsc->mutex);
4547         dout("wait_requests done\n");
4548 }
4549
4550 /*
4551  * called before mount is ro, and before dentries are torn down.
4552  * (hmm, does this still race with new lookups?)
4553  */
4554 void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
4555 {
4556         dout("pre_umount\n");
4557         mdsc->stopping = 1;
4558
4559         lock_unlock_sessions(mdsc);
4560         ceph_flush_dirty_caps(mdsc);
4561         wait_requests(mdsc);
4562
4563         /*
4564          * wait for reply handlers to drop their request refs and
4565          * their inode/dcache refs
4566          */
4567         ceph_msgr_flush();
4568
4569         ceph_cleanup_quotarealms_inodes(mdsc);
4570 }
4571
4572 /*
4573  * wait for all write mds requests to flush.
4574  */
4575 static void wait_unsafe_requests(struct ceph_mds_client *mdsc, u64 want_tid)
4576 {
4577         struct ceph_mds_request *req = NULL, *nextreq;
4578         struct rb_node *n;
4579
4580         mutex_lock(&mdsc->mutex);
4581         dout("wait_unsafe_requests want %lld\n", want_tid);
4582 restart:
4583         req = __get_oldest_req(mdsc);
4584         while (req && req->r_tid <= want_tid) {
4585                 /* find next request */
4586                 n = rb_next(&req->r_node);
4587                 if (n)
4588                         nextreq = rb_entry(n, struct ceph_mds_request, r_node);
4589                 else
4590                         nextreq = NULL;
4591                 if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&
4592                     (req->r_op & CEPH_MDS_OP_WRITE)) {
4593                         /* write op */
4594                         ceph_mdsc_get_request(req);
4595                         if (nextreq)
4596                                 ceph_mdsc_get_request(nextreq);
4597                         mutex_unlock(&mdsc->mutex);
4598                         dout("wait_unsafe_requests  wait on %llu (want %llu)\n",
4599                              req->r_tid, want_tid);
4600                         wait_for_completion(&req->r_safe_completion);
4601                         mutex_lock(&mdsc->mutex);
4602                         ceph_mdsc_put_request(req);
4603                         if (!nextreq)
4604                                 break;  /* next dne before, so we're done! */
4605                         if (RB_EMPTY_NODE(&nextreq->r_node)) {
4606                                 /* next request was removed from tree */
4607                                 ceph_mdsc_put_request(nextreq);
4608                                 goto restart;
4609                         }
4610                         ceph_mdsc_put_request(nextreq);  /* won't go away */
4611                 }
4612                 req = nextreq;
4613         }
4614         mutex_unlock(&mdsc->mutex);
4615         dout("wait_unsafe_requests done\n");
4616 }
4617
4618 void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
4619 {
4620         u64 want_tid, want_flush;
4621
4622         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
4623                 return;
4624
4625         dout("sync\n");
4626         mutex_lock(&mdsc->mutex);
4627         want_tid = mdsc->last_tid;
4628         mutex_unlock(&mdsc->mutex);
4629
4630         ceph_flush_dirty_caps(mdsc);
4631         spin_lock(&mdsc->cap_dirty_lock);
4632         want_flush = mdsc->last_cap_flush_tid;
4633         if (!list_empty(&mdsc->cap_flush_list)) {
4634                 struct ceph_cap_flush *cf =
4635                         list_last_entry(&mdsc->cap_flush_list,
4636                                         struct ceph_cap_flush, g_list);
4637                 cf->wake = true;
4638         }
4639         spin_unlock(&mdsc->cap_dirty_lock);
4640
4641         dout("sync want tid %lld flush_seq %lld\n",
4642              want_tid, want_flush);
4643
4644         wait_unsafe_requests(mdsc, want_tid);
4645         wait_caps_flush(mdsc, want_flush);
4646 }
4647
4648 /*
4649  * true if all sessions are closed, or we force unmount
4650  */
4651 static bool done_closing_sessions(struct ceph_mds_client *mdsc, int skipped)
4652 {
4653         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
4654                 return true;
4655         return atomic_read(&mdsc->num_sessions) <= skipped;
4656 }
4657
4658 /*
4659  * called after sb is ro.
4660  */
4661 void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
4662 {
4663         struct ceph_options *opts = mdsc->fsc->client->options;
4664         struct ceph_mds_session *session;
4665         int i;
4666         int skipped = 0;
4667
4668         dout("close_sessions\n");
4669
4670         /* close sessions */
4671         mutex_lock(&mdsc->mutex);
4672         for (i = 0; i < mdsc->max_sessions; i++) {
4673                 session = __ceph_lookup_mds_session(mdsc, i);
4674                 if (!session)
4675                         continue;
4676                 mutex_unlock(&mdsc->mutex);
4677                 mutex_lock(&session->s_mutex);
4678                 if (__close_session(mdsc, session) <= 0)
4679                         skipped++;
4680                 mutex_unlock(&session->s_mutex);
4681                 ceph_put_mds_session(session);
4682                 mutex_lock(&mdsc->mutex);
4683         }
4684         mutex_unlock(&mdsc->mutex);
4685
4686         dout("waiting for sessions to close\n");
4687         wait_event_timeout(mdsc->session_close_wq,
4688                            done_closing_sessions(mdsc, skipped),
4689                            ceph_timeout_jiffies(opts->mount_timeout));
4690
4691         /* tear down remaining sessions */
4692         mutex_lock(&mdsc->mutex);
4693         for (i = 0; i < mdsc->max_sessions; i++) {
4694                 if (mdsc->sessions[i]) {
4695                         session = ceph_get_mds_session(mdsc->sessions[i]);
4696                         __unregister_session(mdsc, session);
4697                         mutex_unlock(&mdsc->mutex);
4698                         mutex_lock(&session->s_mutex);
4699                         remove_session_caps(session);
4700                         mutex_unlock(&session->s_mutex);
4701                         ceph_put_mds_session(session);
4702                         mutex_lock(&mdsc->mutex);
4703                 }
4704         }
4705         WARN_ON(!list_empty(&mdsc->cap_delay_list));
4706         mutex_unlock(&mdsc->mutex);
4707
4708         ceph_cleanup_snapid_map(mdsc);
4709         ceph_cleanup_empty_realms(mdsc);
4710
4711         cancel_work_sync(&mdsc->cap_reclaim_work);
4712         cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
4713
4714         dout("stopped\n");
4715 }
4716
4717 void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc)
4718 {
4719         struct ceph_mds_session *session;
4720         int mds;
4721
4722         dout("force umount\n");
4723
4724         mutex_lock(&mdsc->mutex);
4725         for (mds = 0; mds < mdsc->max_sessions; mds++) {
4726                 session = __ceph_lookup_mds_session(mdsc, mds);
4727                 if (!session)
4728                         continue;
4729
4730                 if (session->s_state == CEPH_MDS_SESSION_REJECTED)
4731                         __unregister_session(mdsc, session);
4732                 __wake_requests(mdsc, &session->s_waiting);
4733                 mutex_unlock(&mdsc->mutex);
4734
4735                 mutex_lock(&session->s_mutex);
4736                 __close_session(mdsc, session);
4737                 if (session->s_state == CEPH_MDS_SESSION_CLOSING) {
4738                         cleanup_session_requests(mdsc, session);
4739                         remove_session_caps(session);
4740                 }
4741                 mutex_unlock(&session->s_mutex);
4742                 ceph_put_mds_session(session);
4743
4744                 mutex_lock(&mdsc->mutex);
4745                 kick_requests(mdsc, mds);
4746         }
4747         __wake_requests(mdsc, &mdsc->waiting_for_map);
4748         mutex_unlock(&mdsc->mutex);
4749 }
4750
4751 static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
4752 {
4753         dout("stop\n");
4754         /*
4755          * Make sure the delayed work stopped before releasing
4756          * the resources.
4757          *
4758          * Because the cancel_delayed_work_sync() will only
4759          * guarantee that the work finishes executing. But the
4760          * delayed work will re-arm itself again after that.
4761          */
4762         flush_delayed_work(&mdsc->delayed_work);
4763
4764         if (mdsc->mdsmap)
4765                 ceph_mdsmap_destroy(mdsc->mdsmap);
4766         kfree(mdsc->sessions);
4767         ceph_caps_finalize(mdsc);
4768         ceph_pool_perm_destroy(mdsc);
4769 }
4770
4771 void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
4772 {
4773         struct ceph_mds_client *mdsc = fsc->mdsc;
4774         dout("mdsc_destroy %p\n", mdsc);
4775
4776         if (!mdsc)
4777                 return;
4778
4779         /* flush out any connection work with references to us */
4780         ceph_msgr_flush();
4781
4782         ceph_mdsc_stop(mdsc);
4783
4784         ceph_metric_destroy(&mdsc->metric);
4785
4786         flush_delayed_work(&mdsc->metric.delayed_work);
4787         fsc->mdsc = NULL;
4788         kfree(mdsc);
4789         dout("mdsc_destroy %p done\n", mdsc);
4790 }
4791
4792 void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4793 {
4794         struct ceph_fs_client *fsc = mdsc->fsc;
4795         const char *mds_namespace = fsc->mount_options->mds_namespace;
4796         void *p = msg->front.iov_base;
4797         void *end = p + msg->front.iov_len;
4798         u32 epoch;
4799         u32 map_len;
4800         u32 num_fs;
4801         u32 mount_fscid = (u32)-1;
4802         u8 struct_v, struct_cv;
4803         int err = -EINVAL;
4804
4805         ceph_decode_need(&p, end, sizeof(u32), bad);
4806         epoch = ceph_decode_32(&p);
4807
4808         dout("handle_fsmap epoch %u\n", epoch);
4809
4810         ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
4811         struct_v = ceph_decode_8(&p);
4812         struct_cv = ceph_decode_8(&p);
4813         map_len = ceph_decode_32(&p);
4814
4815         ceph_decode_need(&p, end, sizeof(u32) * 3, bad);
4816         p += sizeof(u32) * 2; /* skip epoch and legacy_client_fscid */
4817
4818         num_fs = ceph_decode_32(&p);
4819         while (num_fs-- > 0) {
4820                 void *info_p, *info_end;
4821                 u32 info_len;
4822                 u8 info_v, info_cv;
4823                 u32 fscid, namelen;
4824
4825                 ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
4826                 info_v = ceph_decode_8(&p);
4827                 info_cv = ceph_decode_8(&p);
4828                 info_len = ceph_decode_32(&p);
4829                 ceph_decode_need(&p, end, info_len, bad);
4830                 info_p = p;
4831                 info_end = p + info_len;
4832                 p = info_end;
4833
4834                 ceph_decode_need(&info_p, info_end, sizeof(u32) * 2, bad);
4835                 fscid = ceph_decode_32(&info_p);
4836                 namelen = ceph_decode_32(&info_p);
4837                 ceph_decode_need(&info_p, info_end, namelen, bad);
4838
4839                 if (mds_namespace &&
4840                     strlen(mds_namespace) == namelen &&
4841                     !strncmp(mds_namespace, (char *)info_p, namelen)) {
4842                         mount_fscid = fscid;
4843                         break;
4844                 }
4845         }
4846
4847         ceph_monc_got_map(&fsc->client->monc, CEPH_SUB_FSMAP, epoch);
4848         if (mount_fscid != (u32)-1) {
4849                 fsc->client->monc.fs_cluster_id = mount_fscid;
4850                 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
4851                                    0, true);
4852                 ceph_monc_renew_subs(&fsc->client->monc);
4853         } else {
4854                 err = -ENOENT;
4855                 goto err_out;
4856         }
4857         return;
4858
4859 bad:
4860         pr_err("error decoding fsmap\n");
4861 err_out:
4862         mutex_lock(&mdsc->mutex);
4863         mdsc->mdsmap_err = err;
4864         __wake_requests(mdsc, &mdsc->waiting_for_map);
4865         mutex_unlock(&mdsc->mutex);
4866 }
4867
4868 /*
4869  * handle mds map update.
4870  */
4871 void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4872 {
4873         u32 epoch;
4874         u32 maplen;
4875         void *p = msg->front.iov_base;
4876         void *end = p + msg->front.iov_len;
4877         struct ceph_mdsmap *newmap, *oldmap;
4878         struct ceph_fsid fsid;
4879         int err = -EINVAL;
4880
4881         ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
4882         ceph_decode_copy(&p, &fsid, sizeof(fsid));
4883         if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)
4884                 return;
4885         epoch = ceph_decode_32(&p);
4886         maplen = ceph_decode_32(&p);
4887         dout("handle_map epoch %u len %d\n", epoch, (int)maplen);
4888
4889         /* do we need it? */
4890         mutex_lock(&mdsc->mutex);
4891         if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {
4892                 dout("handle_map epoch %u <= our %u\n",
4893                      epoch, mdsc->mdsmap->m_epoch);
4894                 mutex_unlock(&mdsc->mutex);
4895                 return;
4896         }
4897
4898         newmap = ceph_mdsmap_decode(&p, end);
4899         if (IS_ERR(newmap)) {
4900                 err = PTR_ERR(newmap);
4901                 goto bad_unlock;
4902         }
4903
4904         /* swap into place */
4905         if (mdsc->mdsmap) {
4906                 oldmap = mdsc->mdsmap;
4907                 mdsc->mdsmap = newmap;
4908                 check_new_map(mdsc, newmap, oldmap);
4909                 ceph_mdsmap_destroy(oldmap);
4910         } else {
4911                 mdsc->mdsmap = newmap;  /* first mds map */
4912         }
4913         mdsc->fsc->max_file_size = min((loff_t)mdsc->mdsmap->m_max_file_size,
4914                                         MAX_LFS_FILESIZE);
4915
4916         __wake_requests(mdsc, &mdsc->waiting_for_map);
4917         ceph_monc_got_map(&mdsc->fsc->client->monc, CEPH_SUB_MDSMAP,
4918                           mdsc->mdsmap->m_epoch);
4919
4920         mutex_unlock(&mdsc->mutex);
4921         schedule_delayed(mdsc);
4922         return;
4923
4924 bad_unlock:
4925         mutex_unlock(&mdsc->mutex);
4926 bad:
4927         pr_err("error decoding mdsmap %d\n", err);
4928         return;
4929 }
4930
4931 static struct ceph_connection *con_get(struct ceph_connection *con)
4932 {
4933         struct ceph_mds_session *s = con->private;
4934
4935         if (ceph_get_mds_session(s))
4936                 return con;
4937         return NULL;
4938 }
4939
4940 static void con_put(struct ceph_connection *con)
4941 {
4942         struct ceph_mds_session *s = con->private;
4943
4944         ceph_put_mds_session(s);
4945 }
4946
4947 /*
4948  * if the client is unresponsive for long enough, the mds will kill
4949  * the session entirely.
4950  */
4951 static void peer_reset(struct ceph_connection *con)
4952 {
4953         struct ceph_mds_session *s = con->private;
4954         struct ceph_mds_client *mdsc = s->s_mdsc;
4955
4956         pr_warn("mds%d closed our session\n", s->s_mds);
4957         send_mds_reconnect(mdsc, s);
4958 }
4959
4960 static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
4961 {
4962         struct ceph_mds_session *s = con->private;
4963         struct ceph_mds_client *mdsc = s->s_mdsc;
4964         int type = le16_to_cpu(msg->hdr.type);
4965
4966         mutex_lock(&mdsc->mutex);
4967         if (__verify_registered_session(mdsc, s) < 0) {
4968                 mutex_unlock(&mdsc->mutex);
4969                 goto out;
4970         }
4971         mutex_unlock(&mdsc->mutex);
4972
4973         switch (type) {
4974         case CEPH_MSG_MDS_MAP:
4975                 ceph_mdsc_handle_mdsmap(mdsc, msg);
4976                 break;
4977         case CEPH_MSG_FS_MAP_USER:
4978                 ceph_mdsc_handle_fsmap(mdsc, msg);
4979                 break;
4980         case CEPH_MSG_CLIENT_SESSION:
4981                 handle_session(s, msg);
4982                 break;
4983         case CEPH_MSG_CLIENT_REPLY:
4984                 handle_reply(s, msg);
4985                 break;
4986         case CEPH_MSG_CLIENT_REQUEST_FORWARD:
4987                 handle_forward(mdsc, s, msg);
4988                 break;
4989         case CEPH_MSG_CLIENT_CAPS:
4990                 ceph_handle_caps(s, msg);
4991                 break;
4992         case CEPH_MSG_CLIENT_SNAP:
4993                 ceph_handle_snap(mdsc, s, msg);
4994                 break;
4995         case CEPH_MSG_CLIENT_LEASE:
4996                 handle_lease(mdsc, s, msg);
4997                 break;
4998         case CEPH_MSG_CLIENT_QUOTA:
4999                 ceph_handle_quota(mdsc, s, msg);
5000                 break;
5001
5002         default:
5003                 pr_err("received unknown message type %d %s\n", type,
5004                        ceph_msg_type_name(type));
5005         }
5006 out:
5007         ceph_msg_put(msg);
5008 }
5009
5010 /*
5011  * authentication
5012  */
5013
5014 /*
5015  * Note: returned pointer is the address of a structure that's
5016  * managed separately.  Caller must *not* attempt to free it.
5017  */
5018 static struct ceph_auth_handshake *get_authorizer(struct ceph_connection *con,
5019                                         int *proto, int force_new)
5020 {
5021         struct ceph_mds_session *s = con->private;
5022         struct ceph_mds_client *mdsc = s->s_mdsc;
5023         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5024         struct ceph_auth_handshake *auth = &s->s_auth;
5025
5026         if (force_new && auth->authorizer) {
5027                 ceph_auth_destroy_authorizer(auth->authorizer);
5028                 auth->authorizer = NULL;
5029         }
5030         if (!auth->authorizer) {
5031                 int ret = ceph_auth_create_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
5032                                                       auth);
5033                 if (ret)
5034                         return ERR_PTR(ret);
5035         } else {
5036                 int ret = ceph_auth_update_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
5037                                                       auth);
5038                 if (ret)
5039                         return ERR_PTR(ret);
5040         }
5041         *proto = ac->protocol;
5042
5043         return auth;
5044 }
5045
5046 static int add_authorizer_challenge(struct ceph_connection *con,
5047                                     void *challenge_buf, int challenge_buf_len)
5048 {
5049         struct ceph_mds_session *s = con->private;
5050         struct ceph_mds_client *mdsc = s->s_mdsc;
5051         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5052
5053         return ceph_auth_add_authorizer_challenge(ac, s->s_auth.authorizer,
5054                                             challenge_buf, challenge_buf_len);
5055 }
5056
5057 static int verify_authorizer_reply(struct ceph_connection *con)
5058 {
5059         struct ceph_mds_session *s = con->private;
5060         struct ceph_mds_client *mdsc = s->s_mdsc;
5061         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5062
5063         return ceph_auth_verify_authorizer_reply(ac, s->s_auth.authorizer);
5064 }
5065
5066 static int invalidate_authorizer(struct ceph_connection *con)
5067 {
5068         struct ceph_mds_session *s = con->private;
5069         struct ceph_mds_client *mdsc = s->s_mdsc;
5070         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5071
5072         ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_MDS);
5073
5074         return ceph_monc_validate_auth(&mdsc->fsc->client->monc);
5075 }
5076
5077 static struct ceph_msg *mds_alloc_msg(struct ceph_connection *con,
5078                                 struct ceph_msg_header *hdr, int *skip)
5079 {
5080         struct ceph_msg *msg;
5081         int type = (int) le16_to_cpu(hdr->type);
5082         int front_len = (int) le32_to_cpu(hdr->front_len);
5083
5084         if (con->in_msg)
5085                 return con->in_msg;
5086
5087         *skip = 0;
5088         msg = ceph_msg_new(type, front_len, GFP_NOFS, false);
5089         if (!msg) {
5090                 pr_err("unable to allocate msg type %d len %d\n",
5091                        type, front_len);
5092                 return NULL;
5093         }
5094
5095         return msg;
5096 }
5097
5098 static int mds_sign_message(struct ceph_msg *msg)
5099 {
5100        struct ceph_mds_session *s = msg->con->private;
5101        struct ceph_auth_handshake *auth = &s->s_auth;
5102
5103        return ceph_auth_sign_message(auth, msg);
5104 }
5105
5106 static int mds_check_message_signature(struct ceph_msg *msg)
5107 {
5108        struct ceph_mds_session *s = msg->con->private;
5109        struct ceph_auth_handshake *auth = &s->s_auth;
5110
5111        return ceph_auth_check_message_signature(auth, msg);
5112 }
5113
5114 static const struct ceph_connection_operations mds_con_ops = {
5115         .get = con_get,
5116         .put = con_put,
5117         .dispatch = dispatch,
5118         .get_authorizer = get_authorizer,
5119         .add_authorizer_challenge = add_authorizer_challenge,
5120         .verify_authorizer_reply = verify_authorizer_reply,
5121         .invalidate_authorizer = invalidate_authorizer,
5122         .peer_reset = peer_reset,
5123         .alloc_msg = mds_alloc_msg,
5124         .sign_message = mds_sign_message,
5125         .check_message_signature = mds_check_message_signature,
5126 };
5127
5128 /* eof */