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 #include <linux/bitmap.h>
  15
  16 #include "super.h"
  17 #include "mds_client.h"
  18
  19 #include <linux/ceph/ceph_features.h>
  20 #include <linux/ceph/messenger.h>
  21 #include <linux/ceph/decode.h>
  22 #include <linux/ceph/pagelist.h>
  23 #include <linux/ceph/auth.h>
  24 #include <linux/ceph/debugfs.h>
  25
  26 #define RECONNECT_MAX_SIZE (INT_MAX - PAGE_SIZE)
  27
  28 /*
  29  * A cluster of MDS (metadata server) daemons is responsible for
  30  * managing the file system namespace (the directory hierarchy and
  31  * inodes) and for coordinating shared access to storage.  Metadata is
  32  * partitioning hierarchically across a number of servers, and that
  33  * partition varies over time as the cluster adjusts the distribution
  34  * in order to balance load.
  35  *
  36  * The MDS client is primarily responsible to managing synchronous
  37  * metadata requests for operations like open, unlink, and so forth.
  38  * If there is a MDS failure, we find out about it when we (possibly
  39  * request and) receive a new MDS map, and can resubmit affected
  40  * requests.
  41  *
  42  * For the most part, though, we take advantage of a lossless
  43  * communications channel to the MDS, and do not need to worry about
  44  * timing out or resubmitting requests.
  45  *
  46  * We maintain a stateful "session" with each MDS we interact with.
  47  * Within each session, we sent periodic heartbeat messages to ensure
  48  * any capabilities or leases we have been issues remain valid.  If
  49  * the session times out and goes stale, our leases and capabilities
  50  * are no longer valid.
  51  */
  52
  53 struct ceph_reconnect_state {
  54         struct ceph_mds_session *session;
  55         int nr_caps, nr_realms;
  56         struct ceph_pagelist *pagelist;
  57         unsigned msg_version;
  58         bool allow_multi;
  59 };
  60
  61 static void __wake_requests(struct ceph_mds_client *mdsc,
  62                             struct list_head *head);
  63 static void ceph_cap_release_work(struct work_struct *work);
  64 static void ceph_cap_reclaim_work(struct work_struct *work);
  65
  66 static const struct ceph_connection_operations mds_con_ops;
  67
  68
  69 /*
  70  * mds reply parsing
  71  */
  72
  73 static int parse_reply_info_quota(void **p, void *end,
  74                                   struct ceph_mds_reply_info_in *info)
  75 {
  76         u8 struct_v, struct_compat;
  77         u32 struct_len;
  78
  79         ceph_decode_8_safe(p, end, struct_v, bad);
  80         ceph_decode_8_safe(p, end, struct_compat, bad);
  81         /* struct_v is expected to be >= 1. we only
  82          * understand encoding with struct_compat == 1. */
  83         if (!struct_v || struct_compat != 1)
  84                 goto bad;
  85         ceph_decode_32_safe(p, end, struct_len, bad);
  86         ceph_decode_need(p, end, struct_len, bad);
  87         end = *p + struct_len;
  88         ceph_decode_64_safe(p, end, info->max_bytes, bad);
  89         ceph_decode_64_safe(p, end, info->max_files, bad);
  90         *p = end;
  91         return 0;
  92 bad:
  93         return -EIO;
  94 }
  95
  96 /*
  97  * parse individual inode info
  98  */
  99 static int parse_reply_info_in(void **p, void *end,
 100                                struct ceph_mds_reply_info_in *info,
 101                                u64 features)
 102 {
 103         int err = 0;
 104         u8 struct_v = 0;
 105
 106         if (features == (u64)-1) {
 107                 u32 struct_len;
 108                 u8 struct_compat;
 109                 ceph_decode_8_safe(p, end, struct_v, bad);
 110                 ceph_decode_8_safe(p, end, struct_compat, bad);
 111                 /* struct_v is expected to be >= 1. we only understand
 112                  * encoding with struct_compat == 1. */
 113                 if (!struct_v || struct_compat != 1)
 114                         goto bad;
 115                 ceph_decode_32_safe(p, end, struct_len, bad);
 116                 ceph_decode_need(p, end, struct_len, bad);
 117                 end = *p + struct_len;
 118         }
 119
 120         ceph_decode_need(p, end, sizeof(struct ceph_mds_reply_inode), bad);
 121         info->in = *p;
 122         *p += sizeof(struct ceph_mds_reply_inode) +
 123                 sizeof(*info->in->fragtree.splits) *
 124                 le32_to_cpu(info->in->fragtree.nsplits);
 125
 126         ceph_decode_32_safe(p, end, info->symlink_len, bad);
 127         ceph_decode_need(p, end, info->symlink_len, bad);
 128         info->symlink = *p;
 129         *p += info->symlink_len;
 130
 131         ceph_decode_copy_safe(p, end, &info->dir_layout,
 132                               sizeof(info->dir_layout), bad);
 133         ceph_decode_32_safe(p, end, info->xattr_len, bad);
 134         ceph_decode_need(p, end, info->xattr_len, bad);
 135         info->xattr_data = *p;
 136         *p += info->xattr_len;
 137
 138         if (features == (u64)-1) {
 139                 /* inline data */
 140                 ceph_decode_64_safe(p, end, info->inline_version, bad);
 141                 ceph_decode_32_safe(p, end, info->inline_len, bad);
 142                 ceph_decode_need(p, end, info->inline_len, bad);
 143                 info->inline_data = *p;
 144                 *p += info->inline_len;
 145                 /* quota */
 146                 err = parse_reply_info_quota(p, end, info);
 147                 if (err < 0)
 148                         goto out_bad;
 149                 /* pool namespace */
 150                 ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
 151                 if (info->pool_ns_len > 0) {
 152                         ceph_decode_need(p, end, info->pool_ns_len, bad);
 153                         info->pool_ns_data = *p;
 154                         *p += info->pool_ns_len;
 155                 }
 156
 157                 /* btime */
 158                 ceph_decode_need(p, end, sizeof(info->btime), bad);
 159                 ceph_decode_copy(p, &info->btime, sizeof(info->btime));
 160
 161                 /* change attribute */
 162                 ceph_decode_64_safe(p, end, info->change_attr, bad);
 163
 164                 /* dir pin */
 165                 if (struct_v >= 2) {
 166                         ceph_decode_32_safe(p, end, info->dir_pin, bad);
 167                 } else {
 168                         info->dir_pin = -ENODATA;
 169                 }
 170
 171                 /* snapshot birth time, remains zero for v<=2 */
 172                 if (struct_v >= 3) {
 173                         ceph_decode_need(p, end, sizeof(info->snap_btime), bad);
 174                         ceph_decode_copy(p, &info->snap_btime,
 175                                          sizeof(info->snap_btime));
 176                 } else {
 177                         memset(&info->snap_btime, 0, sizeof(info->snap_btime));
 178                 }
 179
 180                 /* snapshot count, remains zero for v<=3 */
 181                 if (struct_v >= 4) {
 182                         ceph_decode_64_safe(p, end, info->rsnaps, bad);
 183                 } else {
 184                         info->rsnaps = 0;
 185                 }
 186
 187                 *p = end;
 188         } else {
 189                 if (features & CEPH_FEATURE_MDS_INLINE_DATA) {
 190                         ceph_decode_64_safe(p, end, info->inline_version, bad);
 191                         ceph_decode_32_safe(p, end, info->inline_len, bad);
 192                         ceph_decode_need(p, end, info->inline_len, bad);
 193                         info->inline_data = *p;
 194                         *p += info->inline_len;
 195                 } else
 196                         info->inline_version = CEPH_INLINE_NONE;
 197
 198                 if (features & CEPH_FEATURE_MDS_QUOTA) {
 199                         err = parse_reply_info_quota(p, end, info);
 200                         if (err < 0)
 201                                 goto out_bad;
 202                 } else {
 203                         info->max_bytes = 0;
 204                         info->max_files = 0;
 205                 }
 206
 207                 info->pool_ns_len = 0;
 208                 info->pool_ns_data = NULL;
 209                 if (features & CEPH_FEATURE_FS_FILE_LAYOUT_V2) {
 210                         ceph_decode_32_safe(p, end, info->pool_ns_len, bad);
 211                         if (info->pool_ns_len > 0) {
 212                                 ceph_decode_need(p, end, info->pool_ns_len, bad);
 213                                 info->pool_ns_data = *p;
 214                                 *p += info->pool_ns_len;
 215                         }
 216                 }
 217
 218                 if (features & CEPH_FEATURE_FS_BTIME) {
 219                         ceph_decode_need(p, end, sizeof(info->btime), bad);
 220                         ceph_decode_copy(p, &info->btime, sizeof(info->btime));
 221                         ceph_decode_64_safe(p, end, info->change_attr, bad);
 222                 }
 223
 224                 info->dir_pin = -ENODATA;
 225                 /* info->snap_btime and info->rsnaps remain zero */
 226         }
 227         return 0;
 228 bad:
 229         err = -EIO;
 230 out_bad:
 231         return err;
 232 }
 233
 234 static int parse_reply_info_dir(void **p, void *end,
 235                                 struct ceph_mds_reply_dirfrag **dirfrag,
 236                                 u64 features)
 237 {
 238         if (features == (u64)-1) {
 239                 u8 struct_v, struct_compat;
 240                 u32 struct_len;
 241                 ceph_decode_8_safe(p, end, struct_v, bad);
 242                 ceph_decode_8_safe(p, end, struct_compat, bad);
 243                 /* struct_v is expected to be >= 1. we only understand
 244                  * encoding whose struct_compat == 1. */
 245                 if (!struct_v || struct_compat != 1)
 246                         goto bad;
 247                 ceph_decode_32_safe(p, end, struct_len, bad);
 248                 ceph_decode_need(p, end, struct_len, bad);
 249                 end = *p + struct_len;
 250         }
 251
 252         ceph_decode_need(p, end, sizeof(**dirfrag), bad);
 253         *dirfrag = *p;
 254         *p += sizeof(**dirfrag) + sizeof(u32) * le32_to_cpu((*dirfrag)->ndist);
 255         if (unlikely(*p > end))
 256                 goto bad;
 257         if (features == (u64)-1)
 258                 *p = end;
 259         return 0;
 260 bad:
 261         return -EIO;
 262 }
 263
 264 static int parse_reply_info_lease(void **p, void *end,
 265                                   struct ceph_mds_reply_lease **lease,
 266                                   u64 features)
 267 {
 268         if (features == (u64)-1) {
 269                 u8 struct_v, struct_compat;
 270                 u32 struct_len;
 271                 ceph_decode_8_safe(p, end, struct_v, bad);
 272                 ceph_decode_8_safe(p, end, struct_compat, bad);
 273                 /* struct_v is expected to be >= 1. we only understand
 274                  * encoding whose struct_compat == 1. */
 275                 if (!struct_v || struct_compat != 1)
 276                         goto bad;
 277                 ceph_decode_32_safe(p, end, struct_len, bad);
 278                 ceph_decode_need(p, end, struct_len, bad);
 279                 end = *p + struct_len;
 280         }
 281
 282         ceph_decode_need(p, end, sizeof(**lease), bad);
 283         *lease = *p;
 284         *p += sizeof(**lease);
 285         if (features == (u64)-1)
 286                 *p = end;
 287         return 0;
 288 bad:
 289         return -EIO;
 290 }
 291
 292 /*
 293  * parse a normal reply, which may contain a (dir+)dentry and/or a
 294  * target inode.
 295  */
 296 static int parse_reply_info_trace(void **p, void *end,
 297                                   struct ceph_mds_reply_info_parsed *info,
 298                                   u64 features)
 299 {
 300         int err;
 301
 302         if (info->head->is_dentry) {
 303                 err = parse_reply_info_in(p, end, &info->diri, features);
 304                 if (err < 0)
 305                         goto out_bad;
 306
 307                 err = parse_reply_info_dir(p, end, &info->dirfrag, features);
 308                 if (err < 0)
 309                         goto out_bad;
 310
 311                 ceph_decode_32_safe(p, end, info->dname_len, bad);
 312                 ceph_decode_need(p, end, info->dname_len, bad);
 313                 info->dname = *p;
 314                 *p += info->dname_len;
 315
 316                 err = parse_reply_info_lease(p, end, &info->dlease, features);
 317                 if (err < 0)
 318                         goto out_bad;
 319         }
 320
 321         if (info->head->is_target) {
 322                 err = parse_reply_info_in(p, end, &info->targeti, features);
 323                 if (err < 0)
 324                         goto out_bad;
 325         }
 326
 327         if (unlikely(*p != end))
 328                 goto bad;
 329         return 0;
 330
 331 bad:
 332         err = -EIO;
 333 out_bad:
 334         pr_err("problem parsing mds trace %d\n", err);
 335         return err;
 336 }
 337
 338 /*
 339  * parse readdir results
 340  */
 341 static int parse_reply_info_readdir(void **p, void *end,
 342                                 struct ceph_mds_reply_info_parsed *info,
 343                                 u64 features)
 344 {
 345         u32 num, i = 0;
 346         int err;
 347
 348         err = parse_reply_info_dir(p, end, &info->dir_dir, features);
 349         if (err < 0)
 350                 goto out_bad;
 351
 352         ceph_decode_need(p, end, sizeof(num) + 2, bad);
 353         num = ceph_decode_32(p);
 354         {
 355                 u16 flags = ceph_decode_16(p);
 356                 info->dir_end = !!(flags & CEPH_READDIR_FRAG_END);
 357                 info->dir_complete = !!(flags & CEPH_READDIR_FRAG_COMPLETE);
 358                 info->hash_order = !!(flags & CEPH_READDIR_HASH_ORDER);
 359                 info->offset_hash = !!(flags & CEPH_READDIR_OFFSET_HASH);
 360         }
 361         if (num == 0)
 362                 goto done;
 363
 364         BUG_ON(!info->dir_entries);
 365         if ((unsigned long)(info->dir_entries + num) >
 366             (unsigned long)info->dir_entries + info->dir_buf_size) {
 367                 pr_err("dir contents are larger than expected\n");
 368                 WARN_ON(1);
 369                 goto bad;
 370         }
 371
 372         info->dir_nr = num;
 373         while (num) {
 374                 struct ceph_mds_reply_dir_entry *rde = info->dir_entries + i;
 375                 /* dentry */
 376                 ceph_decode_32_safe(p, end, rde->name_len, bad);
 377                 ceph_decode_need(p, end, rde->name_len, bad);
 378                 rde->name = *p;
 379                 *p += rde->name_len;
 380                 dout("parsed dir dname '%.*s'\n", rde->name_len, rde->name);
 381
 382                 /* dentry lease */
 383                 err = parse_reply_info_lease(p, end, &rde->lease, features);
 384                 if (err)
 385                         goto out_bad;
 386                 /* inode */
 387                 err = parse_reply_info_in(p, end, &rde->inode, features);
 388                 if (err < 0)
 389                         goto out_bad;
 390                 /* ceph_readdir_prepopulate() will update it */
 391                 rde->offset = 0;
 392                 i++;
 393                 num--;
 394         }
 395
 396 done:
 397         /* Skip over any unrecognized fields */
 398         *p = end;
 399         return 0;
 400
 401 bad:
 402         err = -EIO;
 403 out_bad:
 404         pr_err("problem parsing dir contents %d\n", err);
 405         return err;
 406 }
 407
 408 /*
 409  * parse fcntl F_GETLK results
 410  */
 411 static int parse_reply_info_filelock(void **p, void *end,
 412                                      struct ceph_mds_reply_info_parsed *info,
 413                                      u64 features)
 414 {
 415         if (*p + sizeof(*info->filelock_reply) > end)
 416                 goto bad;
 417
 418         info->filelock_reply = *p;
 419
 420         /* Skip over any unrecognized fields */
 421         *p = end;
 422         return 0;
 423 bad:
 424         return -EIO;
 425 }
 426
 427
 428 #if BITS_PER_LONG == 64
 429
 430 #define DELEGATED_INO_AVAILABLE         xa_mk_value(1)
 431
 432 static int ceph_parse_deleg_inos(void **p, void *end,
 433                                  struct ceph_mds_session *s)
 434 {
 435         u32 sets;
 436
 437         ceph_decode_32_safe(p, end, sets, bad);
 438         dout("got %u sets of delegated inodes\n", sets);
 439         while (sets--) {
 440                 u64 start, len, ino;
 441
 442                 ceph_decode_64_safe(p, end, start, bad);
 443                 ceph_decode_64_safe(p, end, len, bad);
 444
 445                 /* Don't accept a delegation of system inodes */
 446                 if (start < CEPH_INO_SYSTEM_BASE) {
 447                         pr_warn_ratelimited("ceph: ignoring reserved inode range delegation (start=0x%llx len=0x%llx)\n",
 448                                         start, len);
 449                         continue;
 450                 }
 451                 while (len--) {
 452                         int err = xa_insert(&s->s_delegated_inos, ino = start++,
 453                                             DELEGATED_INO_AVAILABLE,
 454                                             GFP_KERNEL);
 455                         if (!err) {
 456                                 dout("added delegated inode 0x%llx\n",
 457                                      start - 1);
 458                         } else if (err == -EBUSY) {
 459                                 pr_warn("ceph: MDS delegated inode 0x%llx more than once.\n",
 460                                         start - 1);
 461                         } else {
 462                                 return err;
 463                         }
 464                 }
 465         }
 466         return 0;
 467 bad:
 468         return -EIO;
 469 }
 470
 471 u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
 472 {
 473         unsigned long ino;
 474         void *val;
 475
 476         xa_for_each(&s->s_delegated_inos, ino, val) {
 477                 val = xa_erase(&s->s_delegated_inos, ino);
 478                 if (val == DELEGATED_INO_AVAILABLE)
 479                         return ino;
 480         }
 481         return 0;
 482 }
 483
 484 int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
 485 {
 486         return xa_insert(&s->s_delegated_inos, ino, DELEGATED_INO_AVAILABLE,
 487                          GFP_KERNEL);
 488 }
 489 #else /* BITS_PER_LONG == 64 */
 490 /*
 491  * FIXME: xarrays can't handle 64-bit indexes on a 32-bit arch. For now, just
 492  * ignore delegated_inos on 32 bit arch. Maybe eventually add xarrays for top
 493  * and bottom words?
 494  */
 495 static int ceph_parse_deleg_inos(void **p, void *end,
 496                                  struct ceph_mds_session *s)
 497 {
 498         u32 sets;
 499
 500         ceph_decode_32_safe(p, end, sets, bad);
 501         if (sets)
 502                 ceph_decode_skip_n(p, end, sets * 2 * sizeof(__le64), bad);
 503         return 0;
 504 bad:
 505         return -EIO;
 506 }
 507
 508 u64 ceph_get_deleg_ino(struct ceph_mds_session *s)
 509 {
 510         return 0;
 511 }
 512
 513 int ceph_restore_deleg_ino(struct ceph_mds_session *s, u64 ino)
 514 {
 515         return 0;
 516 }
 517 #endif /* BITS_PER_LONG == 64 */
 518
 519 /*
 520  * parse create results
 521  */
 522 static int parse_reply_info_create(void **p, void *end,
 523                                   struct ceph_mds_reply_info_parsed *info,
 524                                   u64 features, struct ceph_mds_session *s)
 525 {
 526         int ret;
 527
 528         if (features == (u64)-1 ||
 529             (features & CEPH_FEATURE_REPLY_CREATE_INODE)) {
 530                 if (*p == end) {
 531                         /* Malformed reply? */
 532                         info->has_create_ino = false;
 533                 } else if (test_bit(CEPHFS_FEATURE_DELEG_INO, &s->s_features)) {
 534                         info->has_create_ino = true;
 535                         /* struct_v, struct_compat, and len */
 536                         ceph_decode_skip_n(p, end, 2 + sizeof(u32), bad);
 537                         ceph_decode_64_safe(p, end, info->ino, bad);
 538                         ret = ceph_parse_deleg_inos(p, end, s);
 539                         if (ret)
 540                                 return ret;
 541                 } else {
 542                         /* legacy */
 543                         ceph_decode_64_safe(p, end, info->ino, bad);
 544                         info->has_create_ino = true;
 545                 }
 546         } else {
 547                 if (*p != end)
 548                         goto bad;
 549         }
 550
 551         /* Skip over any unrecognized fields */
 552         *p = end;
 553         return 0;
 554 bad:
 555         return -EIO;
 556 }
 557
 558 static int parse_reply_info_getvxattr(void **p, void *end,
 559                                       struct ceph_mds_reply_info_parsed *info,
 560                                       u64 features)
 561 {
 562         u32 value_len;
 563
 564         ceph_decode_skip_8(p, end, bad); /* skip current version: 1 */
 565         ceph_decode_skip_8(p, end, bad); /* skip first version: 1 */
 566         ceph_decode_skip_32(p, end, bad); /* skip payload length */
 567
 568         ceph_decode_32_safe(p, end, value_len, bad);
 569
 570         if (value_len == end - *p) {
 571           info->xattr_info.xattr_value = *p;
 572           info->xattr_info.xattr_value_len = value_len;
 573           *p = end;
 574           return value_len;
 575         }
 576 bad:
 577         return -EIO;
 578 }
 579
 580 /*
 581  * parse extra results
 582  */
 583 static int parse_reply_info_extra(void **p, void *end,
 584                                   struct ceph_mds_reply_info_parsed *info,
 585                                   u64 features, struct ceph_mds_session *s)
 586 {
 587         u32 op = le32_to_cpu(info->head->op);
 588
 589         if (op == CEPH_MDS_OP_GETFILELOCK)
 590                 return parse_reply_info_filelock(p, end, info, features);
 591         else if (op == CEPH_MDS_OP_READDIR || op == CEPH_MDS_OP_LSSNAP)
 592                 return parse_reply_info_readdir(p, end, info, features);
 593         else if (op == CEPH_MDS_OP_CREATE)
 594                 return parse_reply_info_create(p, end, info, features, s);
 595         else if (op == CEPH_MDS_OP_GETVXATTR)
 596                 return parse_reply_info_getvxattr(p, end, info, features);
 597         else
 598                 return -EIO;
 599 }
 600
 601 /*
 602  * parse entire mds reply
 603  */
 604 static int parse_reply_info(struct ceph_mds_session *s, struct ceph_msg *msg,
 605                             struct ceph_mds_reply_info_parsed *info,
 606                             u64 features)
 607 {
 608         void *p, *end;
 609         u32 len;
 610         int err;
 611
 612         info->head = msg->front.iov_base;
 613         p = msg->front.iov_base + sizeof(struct ceph_mds_reply_head);
 614         end = p + msg->front.iov_len - sizeof(struct ceph_mds_reply_head);
 615
 616         /* trace */
 617         ceph_decode_32_safe(&p, end, len, bad);
 618         if (len > 0) {
 619                 ceph_decode_need(&p, end, len, bad);
 620                 err = parse_reply_info_trace(&p, p+len, info, features);
 621                 if (err < 0)
 622                         goto out_bad;
 623         }
 624
 625         /* extra */
 626         ceph_decode_32_safe(&p, end, len, bad);
 627         if (len > 0) {
 628                 ceph_decode_need(&p, end, len, bad);
 629                 err = parse_reply_info_extra(&p, p+len, info, features, s);
 630                 if (err < 0)
 631                         goto out_bad;
 632         }
 633
 634         /* snap blob */
 635         ceph_decode_32_safe(&p, end, len, bad);
 636         info->snapblob_len = len;
 637         info->snapblob = p;
 638         p += len;
 639
 640         if (p != end)
 641                 goto bad;
 642         return 0;
 643
 644 bad:
 645         err = -EIO;
 646 out_bad:
 647         pr_err("mds parse_reply err %d\n", err);
 648         return err;
 649 }
 650
 651 static void destroy_reply_info(struct ceph_mds_reply_info_parsed *info)
 652 {
 653         if (!info->dir_entries)
 654                 return;
 655         free_pages((unsigned long)info->dir_entries, get_order(info->dir_buf_size));
 656 }
 657
 658
 659 /*
 660  * sessions
 661  */
 662 const char *ceph_session_state_name(int s)
 663 {
 664         switch (s) {
 665         case CEPH_MDS_SESSION_NEW: return "new";
 666         case CEPH_MDS_SESSION_OPENING: return "opening";
 667         case CEPH_MDS_SESSION_OPEN: return "open";
 668         case CEPH_MDS_SESSION_HUNG: return "hung";
 669         case CEPH_MDS_SESSION_CLOSING: return "closing";
 670         case CEPH_MDS_SESSION_CLOSED: return "closed";
 671         case CEPH_MDS_SESSION_RESTARTING: return "restarting";
 672         case CEPH_MDS_SESSION_RECONNECTING: return "reconnecting";
 673         case CEPH_MDS_SESSION_REJECTED: return "rejected";
 674         default: return "???";
 675         }
 676 }
 677
 678 struct ceph_mds_session *ceph_get_mds_session(struct ceph_mds_session *s)
 679 {
 680         if (refcount_inc_not_zero(&s->s_ref))
 681                 return s;
 682         return NULL;
 683 }
 684
 685 void ceph_put_mds_session(struct ceph_mds_session *s)
 686 {
 687         if (IS_ERR_OR_NULL(s))
 688                 return;
 689
 690         if (refcount_dec_and_test(&s->s_ref)) {
 691                 if (s->s_auth.authorizer)
 692                         ceph_auth_destroy_authorizer(s->s_auth.authorizer);
 693                 WARN_ON(mutex_is_locked(&s->s_mutex));
 694                 xa_destroy(&s->s_delegated_inos);
 695                 kfree(s);
 696         }
 697 }
 698
 699 /*
 700  * called under mdsc->mutex
 701  */
 702 struct ceph_mds_session *__ceph_lookup_mds_session(struct ceph_mds_client *mdsc,
 703                                                    int mds)
 704 {
 705         if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
 706                 return NULL;
 707         return ceph_get_mds_session(mdsc->sessions[mds]);
 708 }
 709
 710 static bool __have_session(struct ceph_mds_client *mdsc, int mds)
 711 {
 712         if (mds >= mdsc->max_sessions || !mdsc->sessions[mds])
 713                 return false;
 714         else
 715                 return true;
 716 }
 717
 718 static int __verify_registered_session(struct ceph_mds_client *mdsc,
 719                                        struct ceph_mds_session *s)
 720 {
 721         if (s->s_mds >= mdsc->max_sessions ||
 722             mdsc->sessions[s->s_mds] != s)
 723                 return -ENOENT;
 724         return 0;
 725 }
 726
 727 /*
 728  * create+register a new session for given mds.
 729  * called under mdsc->mutex.
 730  */
 731 static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
 732                                                  int mds)
 733 {
 734         struct ceph_mds_session *s;
 735
 736         if (mds >= mdsc->mdsmap->possible_max_rank)
 737                 return ERR_PTR(-EINVAL);
 738
 739         s = kzalloc(sizeof(*s), GFP_NOFS);
 740         if (!s)
 741                 return ERR_PTR(-ENOMEM);
 742
 743         if (mds >= mdsc->max_sessions) {
 744                 int newmax = 1 << get_count_order(mds + 1);
 745                 struct ceph_mds_session **sa;
 746
 747                 dout("%s: realloc to %d\n", __func__, newmax);
 748                 sa = kcalloc(newmax, sizeof(void *), GFP_NOFS);
 749                 if (!sa)
 750                         goto fail_realloc;
 751                 if (mdsc->sessions) {
 752                         memcpy(sa, mdsc->sessions,
 753                                mdsc->max_sessions * sizeof(void *));
 754                         kfree(mdsc->sessions);
 755                 }
 756                 mdsc->sessions = sa;
 757                 mdsc->max_sessions = newmax;
 758         }
 759
 760         dout("%s: mds%d\n", __func__, mds);
 761         s->s_mdsc = mdsc;
 762         s->s_mds = mds;
 763         s->s_state = CEPH_MDS_SESSION_NEW;
 764         mutex_init(&s->s_mutex);
 765
 766         ceph_con_init(&s->s_con, s, &mds_con_ops, &mdsc->fsc->client->msgr);
 767
 768         atomic_set(&s->s_cap_gen, 1);
 769         s->s_cap_ttl = jiffies - 1;
 770
 771         spin_lock_init(&s->s_cap_lock);
 772         INIT_LIST_HEAD(&s->s_caps);
 773         refcount_set(&s->s_ref, 1);
 774         INIT_LIST_HEAD(&s->s_waiting);
 775         INIT_LIST_HEAD(&s->s_unsafe);
 776         xa_init(&s->s_delegated_inos);
 777         INIT_LIST_HEAD(&s->s_cap_releases);
 778         INIT_WORK(&s->s_cap_release_work, ceph_cap_release_work);
 779
 780         INIT_LIST_HEAD(&s->s_cap_dirty);
 781         INIT_LIST_HEAD(&s->s_cap_flushing);
 782
 783         mdsc->sessions[mds] = s;
 784         atomic_inc(&mdsc->num_sessions);
 785         refcount_inc(&s->s_ref);  /* one ref to sessions[], one to caller */
 786
 787         ceph_con_open(&s->s_con, CEPH_ENTITY_TYPE_MDS, mds,
 788                       ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
 789
 790         return s;
 791
 792 fail_realloc:
 793         kfree(s);
 794         return ERR_PTR(-ENOMEM);
 795 }
 796
 797 /*
 798  * called under mdsc->mutex
 799  */
 800 static void __unregister_session(struct ceph_mds_client *mdsc,
 801                                struct ceph_mds_session *s)
 802 {
 803         dout("__unregister_session mds%d %p\n", s->s_mds, s);
 804         BUG_ON(mdsc->sessions[s->s_mds] != s);
 805         mdsc->sessions[s->s_mds] = NULL;
 806         ceph_con_close(&s->s_con);
 807         ceph_put_mds_session(s);
 808         atomic_dec(&mdsc->num_sessions);
 809 }
 810
 811 /*
 812  * drop session refs in request.
 813  *
 814  * should be last request ref, or hold mdsc->mutex
 815  */
 816 static void put_request_session(struct ceph_mds_request *req)
 817 {
 818         if (req->r_session) {
 819                 ceph_put_mds_session(req->r_session);
 820                 req->r_session = NULL;
 821         }
 822 }
 823
 824 void ceph_mdsc_iterate_sessions(struct ceph_mds_client *mdsc,
 825                                 void (*cb)(struct ceph_mds_session *),
 826                                 bool check_state)
 827 {
 828         int mds;
 829
 830         mutex_lock(&mdsc->mutex);
 831         for (mds = 0; mds < mdsc->max_sessions; ++mds) {
 832                 struct ceph_mds_session *s;
 833
 834                 s = __ceph_lookup_mds_session(mdsc, mds);
 835                 if (!s)
 836                         continue;
 837
 838                 if (check_state && !check_session_state(s)) {
 839                         ceph_put_mds_session(s);
 840                         continue;
 841                 }
 842
 843                 mutex_unlock(&mdsc->mutex);
 844                 cb(s);
 845                 ceph_put_mds_session(s);
 846                 mutex_lock(&mdsc->mutex);
 847         }
 848         mutex_unlock(&mdsc->mutex);
 849 }
 850
 851 void ceph_mdsc_release_request(struct kref *kref)
 852 {
 853         struct ceph_mds_request *req = container_of(kref,
 854                                                     struct ceph_mds_request,
 855                                                     r_kref);
 856         ceph_mdsc_release_dir_caps_no_check(req);
 857         destroy_reply_info(&req->r_reply_info);
 858         if (req->r_request)
 859                 ceph_msg_put(req->r_request);
 860         if (req->r_reply)
 861                 ceph_msg_put(req->r_reply);
 862         if (req->r_inode) {
 863                 ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
 864                 iput(req->r_inode);
 865         }
 866         if (req->r_parent) {
 867                 ceph_put_cap_refs(ceph_inode(req->r_parent), CEPH_CAP_PIN);
 868                 iput(req->r_parent);
 869         }
 870         iput(req->r_target_inode);
 871         if (req->r_dentry)
 872                 dput(req->r_dentry);
 873         if (req->r_old_dentry)
 874                 dput(req->r_old_dentry);
 875         if (req->r_old_dentry_dir) {
 876                 /*
 877                  * track (and drop pins for) r_old_dentry_dir
 878                  * separately, since r_old_dentry's d_parent may have
 879                  * changed between the dir mutex being dropped and
 880                  * this request being freed.
 881                  */
 882                 ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
 883                                   CEPH_CAP_PIN);
 884                 iput(req->r_old_dentry_dir);
 885         }
 886         kfree(req->r_path1);
 887         kfree(req->r_path2);
 888         put_cred(req->r_cred);
 889         if (req->r_pagelist)
 890                 ceph_pagelist_release(req->r_pagelist);
 891         put_request_session(req);
 892         ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);
 893         WARN_ON_ONCE(!list_empty(&req->r_wait));
 894         kmem_cache_free(ceph_mds_request_cachep, req);
 895 }
 896
 897 DEFINE_RB_FUNCS(request, struct ceph_mds_request, r_tid, r_node)
 898
 899 /*
 900  * lookup session, bump ref if found.
 901  *
 902  * called under mdsc->mutex.
 903  */
 904 static struct ceph_mds_request *
 905 lookup_get_request(struct ceph_mds_client *mdsc, u64 tid)
 906 {
 907         struct ceph_mds_request *req;
 908
 909         req = lookup_request(&mdsc->request_tree, tid);
 910         if (req)
 911                 ceph_mdsc_get_request(req);
 912
 913         return req;
 914 }
 915
 916 /*
 917  * Register an in-flight request, and assign a tid.  Link to directory
 918  * are modifying (if any).
 919  *
 920  * Called under mdsc->mutex.
 921  */
 922 static void __register_request(struct ceph_mds_client *mdsc,
 923                                struct ceph_mds_request *req,
 924                                struct inode *dir)
 925 {
 926         int ret = 0;
 927
 928         req->r_tid = ++mdsc->last_tid;
 929         if (req->r_num_caps) {
 930                 ret = ceph_reserve_caps(mdsc, &req->r_caps_reservation,
 931                                         req->r_num_caps);
 932                 if (ret < 0) {
 933                         pr_err("__register_request %p "
 934                                "failed to reserve caps: %d\n", req, ret);
 935                         /* set req->r_err to fail early from __do_request */
 936                         req->r_err = ret;
 937                         return;
 938                 }
 939         }
 940         dout("__register_request %p tid %lld\n", req, req->r_tid);
 941         ceph_mdsc_get_request(req);
 942         insert_request(&mdsc->request_tree, req);
 943
 944         req->r_cred = get_current_cred();
 945
 946         if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)
 947                 mdsc->oldest_tid = req->r_tid;
 948
 949         if (dir) {
 950                 struct ceph_inode_info *ci = ceph_inode(dir);
 951
 952                 ihold(dir);
 953                 req->r_unsafe_dir = dir;
 954                 spin_lock(&ci->i_unsafe_lock);
 955                 list_add_tail(&req->r_unsafe_dir_item, &ci->i_unsafe_dirops);
 956                 spin_unlock(&ci->i_unsafe_lock);
 957         }
 958 }
 959
 960 static void __unregister_request(struct ceph_mds_client *mdsc,
 961                                  struct ceph_mds_request *req)
 962 {
 963         dout("__unregister_request %p tid %lld\n", req, req->r_tid);
 964
 965         /* Never leave an unregistered request on an unsafe list! */
 966         list_del_init(&req->r_unsafe_item);
 967
 968         if (req->r_tid == mdsc->oldest_tid) {
 969                 struct rb_node *p = rb_next(&req->r_node);
 970                 mdsc->oldest_tid = 0;
 971                 while (p) {
 972                         struct ceph_mds_request *next_req =
 973                                 rb_entry(p, struct ceph_mds_request, r_node);
 974                         if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {
 975                                 mdsc->oldest_tid = next_req->r_tid;
 976                                 break;
 977                         }
 978                         p = rb_next(p);
 979                 }
 980         }
 981
 982         erase_request(&mdsc->request_tree, req);
 983
 984         if (req->r_unsafe_dir) {
 985                 struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
 986                 spin_lock(&ci->i_unsafe_lock);
 987                 list_del_init(&req->r_unsafe_dir_item);
 988                 spin_unlock(&ci->i_unsafe_lock);
 989         }
 990         if (req->r_target_inode &&
 991             test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
 992                 struct ceph_inode_info *ci = ceph_inode(req->r_target_inode);
 993                 spin_lock(&ci->i_unsafe_lock);
 994                 list_del_init(&req->r_unsafe_target_item);
 995                 spin_unlock(&ci->i_unsafe_lock);
 996         }
 997
 998         if (req->r_unsafe_dir) {
 999                 iput(req->r_unsafe_dir);
1000                 req->r_unsafe_dir = NULL;
1001         }
1002
1003         complete_all(&req->r_safe_completion);
1004
1005         ceph_mdsc_put_request(req);
1006 }
1007
1008 /*
1009  * Walk back up the dentry tree until we hit a dentry representing a
1010  * non-snapshot inode. We do this using the rcu_read_lock (which must be held
1011  * when calling this) to ensure that the objects won't disappear while we're
1012  * working with them. Once we hit a candidate dentry, we attempt to take a
1013  * reference to it, and return that as the result.
1014  */
1015 static struct inode *get_nonsnap_parent(struct dentry *dentry)
1016 {
1017         struct inode *inode = NULL;
1018
1019         while (dentry && !IS_ROOT(dentry)) {
1020                 inode = d_inode_rcu(dentry);
1021                 if (!inode || ceph_snap(inode) == CEPH_NOSNAP)
1022                         break;
1023                 dentry = dentry->d_parent;
1024         }
1025         if (inode)
1026                 inode = igrab(inode);
1027         return inode;
1028 }
1029
1030 /*
1031  * Choose mds to send request to next.  If there is a hint set in the
1032  * request (e.g., due to a prior forward hint from the mds), use that.
1033  * Otherwise, consult frag tree and/or caps to identify the
1034  * appropriate mds.  If all else fails, choose randomly.
1035  *
1036  * Called under mdsc->mutex.
1037  */
1038 static int __choose_mds(struct ceph_mds_client *mdsc,
1039                         struct ceph_mds_request *req,
1040                         bool *random)
1041 {
1042         struct inode *inode;
1043         struct ceph_inode_info *ci;
1044         struct ceph_cap *cap;
1045         int mode = req->r_direct_mode;
1046         int mds = -1;
1047         u32 hash = req->r_direct_hash;
1048         bool is_hash = test_bit(CEPH_MDS_R_DIRECT_IS_HASH, &req->r_req_flags);
1049
1050         if (random)
1051                 *random = false;
1052
1053         /*
1054          * is there a specific mds we should try?  ignore hint if we have
1055          * no session and the mds is not up (active or recovering).
1056          */
1057         if (req->r_resend_mds >= 0 &&
1058             (__have_session(mdsc, req->r_resend_mds) ||
1059              ceph_mdsmap_get_state(mdsc->mdsmap, req->r_resend_mds) > 0)) {
1060                 dout("%s using resend_mds mds%d\n", __func__,
1061                      req->r_resend_mds);
1062                 return req->r_resend_mds;
1063         }
1064
1065         if (mode == USE_RANDOM_MDS)
1066                 goto random;
1067
1068         inode = NULL;
1069         if (req->r_inode) {
1070                 if (ceph_snap(req->r_inode) != CEPH_SNAPDIR) {
1071                         inode = req->r_inode;
1072                         ihold(inode);
1073                 } else {
1074                         /* req->r_dentry is non-null for LSSNAP request */
1075                         rcu_read_lock();
1076                         inode = get_nonsnap_parent(req->r_dentry);
1077                         rcu_read_unlock();
1078                         dout("%s using snapdir's parent %p\n", __func__, inode);
1079                 }
1080         } else if (req->r_dentry) {
1081                 /* ignore race with rename; old or new d_parent is okay */
1082                 struct dentry *parent;
1083                 struct inode *dir;
1084
1085                 rcu_read_lock();
1086                 parent = READ_ONCE(req->r_dentry->d_parent);
1087                 dir = req->r_parent ? : d_inode_rcu(parent);
1088
1089                 if (!dir || dir->i_sb != mdsc->fsc->sb) {
1090                         /*  not this fs or parent went negative */
1091                         inode = d_inode(req->r_dentry);
1092                         if (inode)
1093                                 ihold(inode);
1094                 } else if (ceph_snap(dir) != CEPH_NOSNAP) {
1095                         /* direct snapped/virtual snapdir requests
1096                          * based on parent dir inode */
1097                         inode = get_nonsnap_parent(parent);
1098                         dout("%s using nonsnap parent %p\n", __func__, inode);
1099                 } else {
1100                         /* dentry target */
1101                         inode = d_inode(req->r_dentry);
1102                         if (!inode || mode == USE_AUTH_MDS) {
1103                                 /* dir + name */
1104                                 inode = igrab(dir);
1105                                 hash = ceph_dentry_hash(dir, req->r_dentry);
1106                                 is_hash = true;
1107                         } else {
1108                                 ihold(inode);
1109                         }
1110                 }
1111                 rcu_read_unlock();
1112         }
1113
1114         dout("%s %p is_hash=%d (0x%x) mode %d\n", __func__, inode, (int)is_hash,
1115              hash, mode);
1116         if (!inode)
1117                 goto random;
1118         ci = ceph_inode(inode);
1119
1120         if (is_hash && S_ISDIR(inode->i_mode)) {
1121                 struct ceph_inode_frag frag;
1122                 int found;
1123
1124                 ceph_choose_frag(ci, hash, &frag, &found);
1125                 if (found) {
1126                         if (mode == USE_ANY_MDS && frag.ndist > 0) {
1127                                 u8 r;
1128
1129                                 /* choose a random replica */
1130                                 get_random_bytes(&r, 1);
1131                                 r %= frag.ndist;
1132                                 mds = frag.dist[r];
1133                                 dout("%s %p %llx.%llx frag %u mds%d (%d/%d)\n",
1134                                      __func__, inode, ceph_vinop(inode),
1135                                      frag.frag, mds, (int)r, frag.ndist);
1136                                 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
1137                                     CEPH_MDS_STATE_ACTIVE &&
1138                                     !ceph_mdsmap_is_laggy(mdsc->mdsmap, mds))
1139                                         goto out;
1140                         }
1141
1142                         /* since this file/dir wasn't known to be
1143                          * replicated, then we want to look for the
1144                          * authoritative mds. */
1145                         if (frag.mds >= 0) {
1146                                 /* choose auth mds */
1147                                 mds = frag.mds;
1148                                 dout("%s %p %llx.%llx frag %u mds%d (auth)\n",
1149                                      __func__, inode, ceph_vinop(inode),
1150                                      frag.frag, mds);
1151                                 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
1152                                     CEPH_MDS_STATE_ACTIVE) {
1153                                         if (!ceph_mdsmap_is_laggy(mdsc->mdsmap,
1154                                                                   mds))
1155                                                 goto out;
1156                                 }
1157                         }
1158                         mode = USE_AUTH_MDS;
1159                 }
1160         }
1161
1162         spin_lock(&ci->i_ceph_lock);
1163         cap = NULL;
1164         if (mode == USE_AUTH_MDS)
1165                 cap = ci->i_auth_cap;
1166         if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
1167                 cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
1168         if (!cap) {
1169                 spin_unlock(&ci->i_ceph_lock);
1170                 iput(inode);
1171                 goto random;
1172         }
1173         mds = cap->session->s_mds;
1174         dout("%s %p %llx.%llx mds%d (%scap %p)\n", __func__,
1175              inode, ceph_vinop(inode), mds,
1176              cap == ci->i_auth_cap ? "auth " : "", cap);
1177         spin_unlock(&ci->i_ceph_lock);
1178 out:
1179         iput(inode);
1180         return mds;
1181
1182 random:
1183         if (random)
1184                 *random = true;
1185
1186         mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);
1187         dout("%s chose random mds%d\n", __func__, mds);
1188         return mds;
1189 }
1190
1191
1192 /*
1193  * session messages
1194  */
1195 struct ceph_msg *ceph_create_session_msg(u32 op, u64 seq)
1196 {
1197         struct ceph_msg *msg;
1198         struct ceph_mds_session_head *h;
1199
1200         msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h), GFP_NOFS,
1201                            false);
1202         if (!msg) {
1203                 pr_err("ENOMEM creating session %s msg\n",
1204                        ceph_session_op_name(op));
1205                 return NULL;
1206         }
1207         h = msg->front.iov_base;
1208         h->op = cpu_to_le32(op);
1209         h->seq = cpu_to_le64(seq);
1210
1211         return msg;
1212 }
1213
1214 static const unsigned char feature_bits[] = CEPHFS_FEATURES_CLIENT_SUPPORTED;
1215 #define FEATURE_BYTES(c) (DIV_ROUND_UP((size_t)feature_bits[c - 1] + 1, 64) * 8)
1216 static int encode_supported_features(void **p, void *end)
1217 {
1218         static const size_t count = ARRAY_SIZE(feature_bits);
1219
1220         if (count > 0) {
1221                 size_t i;
1222                 size_t size = FEATURE_BYTES(count);
1223
1224                 if (WARN_ON_ONCE(*p + 4 + size > end))
1225                         return -ERANGE;
1226
1227                 ceph_encode_32(p, size);
1228                 memset(*p, 0, size);
1229                 for (i = 0; i < count; i++)
1230                         ((unsigned char*)(*p))[i / 8] |= BIT(feature_bits[i] % 8);
1231                 *p += size;
1232         } else {
1233                 if (WARN_ON_ONCE(*p + 4 > end))
1234                         return -ERANGE;
1235
1236                 ceph_encode_32(p, 0);
1237         }
1238
1239         return 0;
1240 }
1241
1242 static const unsigned char metric_bits[] = CEPHFS_METRIC_SPEC_CLIENT_SUPPORTED;
1243 #define METRIC_BYTES(cnt) (DIV_ROUND_UP((size_t)metric_bits[cnt - 1] + 1, 64) * 8)
1244 static int encode_metric_spec(void **p, void *end)
1245 {
1246         static const size_t count = ARRAY_SIZE(metric_bits);
1247
1248         /* header */
1249         if (WARN_ON_ONCE(*p + 2 > end))
1250                 return -ERANGE;
1251
1252         ceph_encode_8(p, 1); /* version */
1253         ceph_encode_8(p, 1); /* compat */
1254
1255         if (count > 0) {
1256                 size_t i;
1257                 size_t size = METRIC_BYTES(count);
1258
1259                 if (WARN_ON_ONCE(*p + 4 + 4 + size > end))
1260                         return -ERANGE;
1261
1262                 /* metric spec info length */
1263                 ceph_encode_32(p, 4 + size);
1264
1265                 /* metric spec */
1266                 ceph_encode_32(p, size);
1267                 memset(*p, 0, size);
1268                 for (i = 0; i < count; i++)
1269                         ((unsigned char *)(*p))[i / 8] |= BIT(metric_bits[i] % 8);
1270                 *p += size;
1271         } else {
1272                 if (WARN_ON_ONCE(*p + 4 + 4 > end))
1273                         return -ERANGE;
1274
1275                 /* metric spec info length */
1276                 ceph_encode_32(p, 4);
1277                 /* metric spec */
1278                 ceph_encode_32(p, 0);
1279         }
1280
1281         return 0;
1282 }
1283
1284 /*
1285  * session message, specialization for CEPH_SESSION_REQUEST_OPEN
1286  * to include additional client metadata fields.
1287  */
1288 static struct ceph_msg *create_session_open_msg(struct ceph_mds_client *mdsc, u64 seq)
1289 {
1290         struct ceph_msg *msg;
1291         struct ceph_mds_session_head *h;
1292         int i;
1293         int extra_bytes = 0;
1294         int metadata_key_count = 0;
1295         struct ceph_options *opt = mdsc->fsc->client->options;
1296         struct ceph_mount_options *fsopt = mdsc->fsc->mount_options;
1297         size_t size, count;
1298         void *p, *end;
1299         int ret;
1300
1301         const char* metadata[][2] = {
1302                 {"hostname", mdsc->nodename},
1303                 {"kernel_version", init_utsname()->release},
1304                 {"entity_id", opt->name ? : ""},
1305                 {"root", fsopt->server_path ? : "/"},
1306                 {NULL, NULL}
1307         };
1308
1309         /* Calculate serialized length of metadata */
1310         extra_bytes = 4;  /* map length */
1311         for (i = 0; metadata[i][0]; ++i) {
1312                 extra_bytes += 8 + strlen(metadata[i][0]) +
1313                         strlen(metadata[i][1]);
1314                 metadata_key_count++;
1315         }
1316
1317         /* supported feature */
1318         size = 0;
1319         count = ARRAY_SIZE(feature_bits);
1320         if (count > 0)
1321                 size = FEATURE_BYTES(count);
1322         extra_bytes += 4 + size;
1323
1324         /* metric spec */
1325         size = 0;
1326         count = ARRAY_SIZE(metric_bits);
1327         if (count > 0)
1328                 size = METRIC_BYTES(count);
1329         extra_bytes += 2 + 4 + 4 + size;
1330
1331         /* Allocate the message */
1332         msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h) + extra_bytes,
1333                            GFP_NOFS, false);
1334         if (!msg) {
1335                 pr_err("ENOMEM creating session open msg\n");
1336                 return ERR_PTR(-ENOMEM);
1337         }
1338         p = msg->front.iov_base;
1339         end = p + msg->front.iov_len;
1340
1341         h = p;
1342         h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);
1343         h->seq = cpu_to_le64(seq);
1344
1345         /*
1346          * Serialize client metadata into waiting buffer space, using
1347          * the format that userspace expects for map<string, string>
1348          *
1349          * ClientSession messages with metadata are v4
1350          */
1351         msg->hdr.version = cpu_to_le16(4);
1352         msg->hdr.compat_version = cpu_to_le16(1);
1353
1354         /* The write pointer, following the session_head structure */
1355         p += sizeof(*h);
1356
1357         /* Number of entries in the map */
1358         ceph_encode_32(&p, metadata_key_count);
1359
1360         /* Two length-prefixed strings for each entry in the map */
1361         for (i = 0; metadata[i][0]; ++i) {
1362                 size_t const key_len = strlen(metadata[i][0]);
1363                 size_t const val_len = strlen(metadata[i][1]);
1364
1365                 ceph_encode_32(&p, key_len);
1366                 memcpy(p, metadata[i][0], key_len);
1367                 p += key_len;
1368                 ceph_encode_32(&p, val_len);
1369                 memcpy(p, metadata[i][1], val_len);
1370                 p += val_len;
1371         }
1372
1373         ret = encode_supported_features(&p, end);
1374         if (ret) {
1375                 pr_err("encode_supported_features failed!\n");
1376                 ceph_msg_put(msg);
1377                 return ERR_PTR(ret);
1378         }
1379
1380         ret = encode_metric_spec(&p, end);
1381         if (ret) {
1382                 pr_err("encode_metric_spec failed!\n");
1383                 ceph_msg_put(msg);
1384                 return ERR_PTR(ret);
1385         }
1386
1387         msg->front.iov_len = p - msg->front.iov_base;
1388         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1389
1390         return msg;
1391 }
1392
1393 /*
1394  * send session open request.
1395  *
1396  * called under mdsc->mutex
1397  */
1398 static int __open_session(struct ceph_mds_client *mdsc,
1399                           struct ceph_mds_session *session)
1400 {
1401         struct ceph_msg *msg;
1402         int mstate;
1403         int mds = session->s_mds;
1404
1405         /* wait for mds to go active? */
1406         mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
1407         dout("open_session to mds%d (%s)\n", mds,
1408              ceph_mds_state_name(mstate));
1409         session->s_state = CEPH_MDS_SESSION_OPENING;
1410         session->s_renew_requested = jiffies;
1411
1412         /* send connect message */
1413         msg = create_session_open_msg(mdsc, session->s_seq);
1414         if (IS_ERR(msg))
1415                 return PTR_ERR(msg);
1416         ceph_con_send(&session->s_con, msg);
1417         return 0;
1418 }
1419
1420 /*
1421  * open sessions for any export targets for the given mds
1422  *
1423  * called under mdsc->mutex
1424  */
1425 static struct ceph_mds_session *
1426 __open_export_target_session(struct ceph_mds_client *mdsc, int target)
1427 {
1428         struct ceph_mds_session *session;
1429         int ret;
1430
1431         session = __ceph_lookup_mds_session(mdsc, target);
1432         if (!session) {
1433                 session = register_session(mdsc, target);
1434                 if (IS_ERR(session))
1435                         return session;
1436         }
1437         if (session->s_state == CEPH_MDS_SESSION_NEW ||
1438             session->s_state == CEPH_MDS_SESSION_CLOSING) {
1439                 ret = __open_session(mdsc, session);
1440                 if (ret)
1441                         return ERR_PTR(ret);
1442         }
1443
1444         return session;
1445 }
1446
1447 struct ceph_mds_session *
1448 ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target)
1449 {
1450         struct ceph_mds_session *session;
1451
1452         dout("open_export_target_session to mds%d\n", target);
1453
1454         mutex_lock(&mdsc->mutex);
1455         session = __open_export_target_session(mdsc, target);
1456         mutex_unlock(&mdsc->mutex);
1457
1458         return session;
1459 }
1460
1461 static void __open_export_target_sessions(struct ceph_mds_client *mdsc,
1462                                           struct ceph_mds_session *session)
1463 {
1464         struct ceph_mds_info *mi;
1465         struct ceph_mds_session *ts;
1466         int i, mds = session->s_mds;
1467
1468         if (mds >= mdsc->mdsmap->possible_max_rank)
1469                 return;
1470
1471         mi = &mdsc->mdsmap->m_info[mds];
1472         dout("open_export_target_sessions for mds%d (%d targets)\n",
1473              session->s_mds, mi->num_export_targets);
1474
1475         for (i = 0; i < mi->num_export_targets; i++) {
1476                 ts = __open_export_target_session(mdsc, mi->export_targets[i]);
1477                 ceph_put_mds_session(ts);
1478         }
1479 }
1480
1481 void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
1482                                            struct ceph_mds_session *session)
1483 {
1484         mutex_lock(&mdsc->mutex);
1485         __open_export_target_sessions(mdsc, session);
1486         mutex_unlock(&mdsc->mutex);
1487 }
1488
1489 /*
1490  * session caps
1491  */
1492
1493 static void detach_cap_releases(struct ceph_mds_session *session,
1494                                 struct list_head *target)
1495 {
1496         lockdep_assert_held(&session->s_cap_lock);
1497
1498         list_splice_init(&session->s_cap_releases, target);
1499         session->s_num_cap_releases = 0;
1500         dout("dispose_cap_releases mds%d\n", session->s_mds);
1501 }
1502
1503 static void dispose_cap_releases(struct ceph_mds_client *mdsc,
1504                                  struct list_head *dispose)
1505 {
1506         while (!list_empty(dispose)) {
1507                 struct ceph_cap *cap;
1508                 /* zero out the in-progress message */
1509                 cap = list_first_entry(dispose, struct ceph_cap, session_caps);
1510                 list_del(&cap->session_caps);
1511                 ceph_put_cap(mdsc, cap);
1512         }
1513 }
1514
1515 static void cleanup_session_requests(struct ceph_mds_client *mdsc,
1516                                      struct ceph_mds_session *session)
1517 {
1518         struct ceph_mds_request *req;
1519         struct rb_node *p;
1520
1521         dout("cleanup_session_requests mds%d\n", session->s_mds);
1522         mutex_lock(&mdsc->mutex);
1523         while (!list_empty(&session->s_unsafe)) {
1524                 req = list_first_entry(&session->s_unsafe,
1525                                        struct ceph_mds_request, r_unsafe_item);
1526                 pr_warn_ratelimited(" dropping unsafe request %llu\n",
1527                                     req->r_tid);
1528                 if (req->r_target_inode)
1529                         mapping_set_error(req->r_target_inode->i_mapping, -EIO);
1530                 if (req->r_unsafe_dir)
1531                         mapping_set_error(req->r_unsafe_dir->i_mapping, -EIO);
1532                 __unregister_request(mdsc, req);
1533         }
1534         /* zero r_attempts, so kick_requests() will re-send requests */
1535         p = rb_first(&mdsc->request_tree);
1536         while (p) {
1537                 req = rb_entry(p, struct ceph_mds_request, r_node);
1538                 p = rb_next(p);
1539                 if (req->r_session &&
1540                     req->r_session->s_mds == session->s_mds)
1541                         req->r_attempts = 0;
1542         }
1543         mutex_unlock(&mdsc->mutex);
1544 }
1545
1546 /*
1547  * Helper to safely iterate over all caps associated with a session, with
1548  * special care taken to handle a racing __ceph_remove_cap().
1549  *
1550  * Caller must hold session s_mutex.
1551  */
1552 int ceph_iterate_session_caps(struct ceph_mds_session *session,
1553                               int (*cb)(struct inode *, struct ceph_cap *,
1554                                         void *), void *arg)
1555 {
1556         struct list_head *p;
1557         struct ceph_cap *cap;
1558         struct inode *inode, *last_inode = NULL;
1559         struct ceph_cap *old_cap = NULL;
1560         int ret;
1561
1562         dout("iterate_session_caps %p mds%d\n", session, session->s_mds);
1563         spin_lock(&session->s_cap_lock);
1564         p = session->s_caps.next;
1565         while (p != &session->s_caps) {
1566                 cap = list_entry(p, struct ceph_cap, session_caps);
1567                 inode = igrab(&cap->ci->vfs_inode);
1568                 if (!inode) {
1569                         p = p->next;
1570                         continue;
1571                 }
1572                 session->s_cap_iterator = cap;
1573                 spin_unlock(&session->s_cap_lock);
1574
1575                 if (last_inode) {
1576                         iput(last_inode);
1577                         last_inode = NULL;
1578                 }
1579                 if (old_cap) {
1580                         ceph_put_cap(session->s_mdsc, old_cap);
1581                         old_cap = NULL;
1582                 }
1583
1584                 ret = cb(inode, cap, arg);
1585                 last_inode = inode;
1586
1587                 spin_lock(&session->s_cap_lock);
1588                 p = p->next;
1589                 if (!cap->ci) {
1590                         dout("iterate_session_caps  finishing cap %p removal\n",
1591                              cap);
1592                         BUG_ON(cap->session != session);
1593                         cap->session = NULL;
1594                         list_del_init(&cap->session_caps);
1595                         session->s_nr_caps--;
1596                         atomic64_dec(&session->s_mdsc->metric.total_caps);
1597                         if (cap->queue_release)
1598                                 __ceph_queue_cap_release(session, cap);
1599                         else
1600                                 old_cap = cap;  /* put_cap it w/o locks held */
1601                 }
1602                 if (ret < 0)
1603                         goto out;
1604         }
1605         ret = 0;
1606 out:
1607         session->s_cap_iterator = NULL;
1608         spin_unlock(&session->s_cap_lock);
1609
1610         iput(last_inode);
1611         if (old_cap)
1612                 ceph_put_cap(session->s_mdsc, old_cap);
1613
1614         return ret;
1615 }
1616
1617 static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
1618                                   void *arg)
1619 {
1620         struct ceph_inode_info *ci = ceph_inode(inode);
1621         bool invalidate = false;
1622         int iputs;
1623
1624         dout("removing cap %p, ci is %p, inode is %p\n",
1625              cap, ci, &ci->vfs_inode);
1626         spin_lock(&ci->i_ceph_lock);
1627         iputs = ceph_purge_inode_cap(inode, cap, &invalidate);
1628         spin_unlock(&ci->i_ceph_lock);
1629
1630         wake_up_all(&ci->i_cap_wq);
1631         if (invalidate)
1632                 ceph_queue_invalidate(inode);
1633         while (iputs--)
1634                 iput(inode);
1635         return 0;
1636 }
1637
1638 /*
1639  * caller must hold session s_mutex
1640  */
1641 static void remove_session_caps(struct ceph_mds_session *session)
1642 {
1643         struct ceph_fs_client *fsc = session->s_mdsc->fsc;
1644         struct super_block *sb = fsc->sb;
1645         LIST_HEAD(dispose);
1646
1647         dout("remove_session_caps on %p\n", session);
1648         ceph_iterate_session_caps(session, remove_session_caps_cb, fsc);
1649
1650         wake_up_all(&fsc->mdsc->cap_flushing_wq);
1651
1652         spin_lock(&session->s_cap_lock);
1653         if (session->s_nr_caps > 0) {
1654                 struct inode *inode;
1655                 struct ceph_cap *cap, *prev = NULL;
1656                 struct ceph_vino vino;
1657                 /*
1658                  * iterate_session_caps() skips inodes that are being
1659                  * deleted, we need to wait until deletions are complete.
1660                  * __wait_on_freeing_inode() is designed for the job,
1661                  * but it is not exported, so use lookup inode function
1662                  * to access it.
1663                  */
1664                 while (!list_empty(&session->s_caps)) {
1665                         cap = list_entry(session->s_caps.next,
1666                                          struct ceph_cap, session_caps);
1667                         if (cap == prev)
1668                                 break;
1669                         prev = cap;
1670                         vino = cap->ci->i_vino;
1671                         spin_unlock(&session->s_cap_lock);
1672
1673                         inode = ceph_find_inode(sb, vino);
1674                         iput(inode);
1675
1676                         spin_lock(&session->s_cap_lock);
1677                 }
1678         }
1679
1680         // drop cap expires and unlock s_cap_lock
1681         detach_cap_releases(session, &dispose);
1682
1683         BUG_ON(session->s_nr_caps > 0);
1684         BUG_ON(!list_empty(&session->s_cap_flushing));
1685         spin_unlock(&session->s_cap_lock);
1686         dispose_cap_releases(session->s_mdsc, &dispose);
1687 }
1688
1689 enum {
1690         RECONNECT,
1691         RENEWCAPS,
1692         FORCE_RO,
1693 };
1694
1695 /*
1696  * wake up any threads waiting on this session's caps.  if the cap is
1697  * old (didn't get renewed on the client reconnect), remove it now.
1698  *
1699  * caller must hold s_mutex.
1700  */
1701 static int wake_up_session_cb(struct inode *inode, struct ceph_cap *cap,
1702                               void *arg)
1703 {
1704         struct ceph_inode_info *ci = ceph_inode(inode);
1705         unsigned long ev = (unsigned long)arg;
1706
1707         if (ev == RECONNECT) {
1708                 spin_lock(&ci->i_ceph_lock);
1709                 ci->i_wanted_max_size = 0;
1710                 ci->i_requested_max_size = 0;
1711                 spin_unlock(&ci->i_ceph_lock);
1712         } else if (ev == RENEWCAPS) {
1713                 if (cap->cap_gen < atomic_read(&cap->session->s_cap_gen)) {
1714                         /* mds did not re-issue stale cap */
1715                         spin_lock(&ci->i_ceph_lock);
1716                         cap->issued = cap->implemented = CEPH_CAP_PIN;
1717                         spin_unlock(&ci->i_ceph_lock);
1718                 }
1719         } else if (ev == FORCE_RO) {
1720         }
1721         wake_up_all(&ci->i_cap_wq);
1722         return 0;
1723 }
1724
1725 static void wake_up_session_caps(struct ceph_mds_session *session, int ev)
1726 {
1727         dout("wake_up_session_caps %p mds%d\n", session, session->s_mds);
1728         ceph_iterate_session_caps(session, wake_up_session_cb,
1729                                   (void *)(unsigned long)ev);
1730 }
1731
1732 /*
1733  * Send periodic message to MDS renewing all currently held caps.  The
1734  * ack will reset the expiration for all caps from this session.
1735  *
1736  * caller holds s_mutex
1737  */
1738 static int send_renew_caps(struct ceph_mds_client *mdsc,
1739                            struct ceph_mds_session *session)
1740 {
1741         struct ceph_msg *msg;
1742         int state;
1743
1744         if (time_after_eq(jiffies, session->s_cap_ttl) &&
1745             time_after_eq(session->s_cap_ttl, session->s_renew_requested))
1746                 pr_info("mds%d caps stale\n", session->s_mds);
1747         session->s_renew_requested = jiffies;
1748
1749         /* do not try to renew caps until a recovering mds has reconnected
1750          * with its clients. */
1751         state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);
1752         if (state < CEPH_MDS_STATE_RECONNECT) {
1753                 dout("send_renew_caps ignoring mds%d (%s)\n",
1754                      session->s_mds, ceph_mds_state_name(state));
1755                 return 0;
1756         }
1757
1758         dout("send_renew_caps to mds%d (%s)\n", session->s_mds,
1759                 ceph_mds_state_name(state));
1760         msg = ceph_create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
1761                                       ++session->s_renew_seq);
1762         if (!msg)
1763                 return -ENOMEM;
1764         ceph_con_send(&session->s_con, msg);
1765         return 0;
1766 }
1767
1768 static int send_flushmsg_ack(struct ceph_mds_client *mdsc,
1769                              struct ceph_mds_session *session, u64 seq)
1770 {
1771         struct ceph_msg *msg;
1772
1773         dout("send_flushmsg_ack to mds%d (%s)s seq %lld\n",
1774              session->s_mds, ceph_session_state_name(session->s_state), seq);
1775         msg = ceph_create_session_msg(CEPH_SESSION_FLUSHMSG_ACK, seq);
1776         if (!msg)
1777                 return -ENOMEM;
1778         ceph_con_send(&session->s_con, msg);
1779         return 0;
1780 }
1781
1782
1783 /*
1784  * Note new cap ttl, and any transition from stale -> not stale (fresh?).
1785  *
1786  * Called under session->s_mutex
1787  */
1788 static void renewed_caps(struct ceph_mds_client *mdsc,
1789                          struct ceph_mds_session *session, int is_renew)
1790 {
1791         int was_stale;
1792         int wake = 0;
1793
1794         spin_lock(&session->s_cap_lock);
1795         was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);
1796
1797         session->s_cap_ttl = session->s_renew_requested +
1798                 mdsc->mdsmap->m_session_timeout*HZ;
1799
1800         if (was_stale) {
1801                 if (time_before(jiffies, session->s_cap_ttl)) {
1802                         pr_info("mds%d caps renewed\n", session->s_mds);
1803                         wake = 1;
1804                 } else {
1805                         pr_info("mds%d caps still stale\n", session->s_mds);
1806                 }
1807         }
1808         dout("renewed_caps mds%d ttl now %lu, was %s, now %s\n",
1809              session->s_mds, session->s_cap_ttl, was_stale ? "stale" : "fresh",
1810              time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");
1811         spin_unlock(&session->s_cap_lock);
1812
1813         if (wake)
1814                 wake_up_session_caps(session, RENEWCAPS);
1815 }
1816
1817 /*
1818  * send a session close request
1819  */
1820 static int request_close_session(struct ceph_mds_session *session)
1821 {
1822         struct ceph_msg *msg;
1823
1824         dout("request_close_session mds%d state %s seq %lld\n",
1825              session->s_mds, ceph_session_state_name(session->s_state),
1826              session->s_seq);
1827         msg = ceph_create_session_msg(CEPH_SESSION_REQUEST_CLOSE,
1828                                       session->s_seq);
1829         if (!msg)
1830                 return -ENOMEM;
1831         ceph_con_send(&session->s_con, msg);
1832         return 1;
1833 }
1834
1835 /*
1836  * Called with s_mutex held.
1837  */
1838 static int __close_session(struct ceph_mds_client *mdsc,
1839                          struct ceph_mds_session *session)
1840 {
1841         if (session->s_state >= CEPH_MDS_SESSION_CLOSING)
1842                 return 0;
1843         session->s_state = CEPH_MDS_SESSION_CLOSING;
1844         return request_close_session(session);
1845 }
1846
1847 static bool drop_negative_children(struct dentry *dentry)
1848 {
1849         struct dentry *child;
1850         bool all_negative = true;
1851
1852         if (!d_is_dir(dentry))
1853                 goto out;
1854
1855         spin_lock(&dentry->d_lock);
1856         list_for_each_entry(child, &dentry->d_subdirs, d_child) {
1857                 if (d_really_is_positive(child)) {
1858                         all_negative = false;
1859                         break;
1860                 }
1861         }
1862         spin_unlock(&dentry->d_lock);
1863
1864         if (all_negative)
1865                 shrink_dcache_parent(dentry);
1866 out:
1867         return all_negative;
1868 }
1869
1870 /*
1871  * Trim old(er) caps.
1872  *
1873  * Because we can't cache an inode without one or more caps, we do
1874  * this indirectly: if a cap is unused, we prune its aliases, at which
1875  * point the inode will hopefully get dropped to.
1876  *
1877  * Yes, this is a bit sloppy.  Our only real goal here is to respond to
1878  * memory pressure from the MDS, though, so it needn't be perfect.
1879  */
1880 static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
1881 {
1882         int *remaining = arg;
1883         struct ceph_inode_info *ci = ceph_inode(inode);
1884         int used, wanted, oissued, mine;
1885
1886         if (*remaining <= 0)
1887                 return -1;
1888
1889         spin_lock(&ci->i_ceph_lock);
1890         mine = cap->issued | cap->implemented;
1891         used = __ceph_caps_used(ci);
1892         wanted = __ceph_caps_file_wanted(ci);
1893         oissued = __ceph_caps_issued_other(ci, cap);
1894
1895         dout("trim_caps_cb %p cap %p mine %s oissued %s used %s wanted %s\n",
1896              inode, cap, ceph_cap_string(mine), ceph_cap_string(oissued),
1897              ceph_cap_string(used), ceph_cap_string(wanted));
1898         if (cap == ci->i_auth_cap) {
1899                 if (ci->i_dirty_caps || ci->i_flushing_caps ||
1900                     !list_empty(&ci->i_cap_snaps))
1901                         goto out;
1902                 if ((used | wanted) & CEPH_CAP_ANY_WR)
1903                         goto out;
1904                 /* Note: it's possible that i_filelock_ref becomes non-zero
1905                  * after dropping auth caps. It doesn't hurt because reply
1906                  * of lock mds request will re-add auth caps. */
1907                 if (atomic_read(&ci->i_filelock_ref) > 0)
1908                         goto out;
1909         }
1910         /* The inode has cached pages, but it's no longer used.
1911          * we can safely drop it */
1912         if (S_ISREG(inode->i_mode) &&
1913             wanted == 0 && used == CEPH_CAP_FILE_CACHE &&
1914             !(oissued & CEPH_CAP_FILE_CACHE)) {
1915           used = 0;
1916           oissued = 0;
1917         }
1918         if ((used | wanted) & ~oissued & mine)
1919                 goto out;   /* we need these caps */
1920
1921         if (oissued) {
1922                 /* we aren't the only cap.. just remove us */
1923                 ceph_remove_cap(cap, true);
1924                 (*remaining)--;
1925         } else {
1926                 struct dentry *dentry;
1927                 /* try dropping referring dentries */
1928                 spin_unlock(&ci->i_ceph_lock);
1929                 dentry = d_find_any_alias(inode);
1930                 if (dentry && drop_negative_children(dentry)) {
1931                         int count;
1932                         dput(dentry);
1933                         d_prune_aliases(inode);
1934                         count = atomic_read(&inode->i_count);
1935                         if (count == 1)
1936                                 (*remaining)--;
1937                         dout("trim_caps_cb %p cap %p pruned, count now %d\n",
1938                              inode, cap, count);
1939                 } else {
1940                         dput(dentry);
1941                 }
1942                 return 0;
1943         }
1944
1945 out:
1946         spin_unlock(&ci->i_ceph_lock);
1947         return 0;
1948 }
1949
1950 /*
1951  * Trim session cap count down to some max number.
1952  */
1953 int ceph_trim_caps(struct ceph_mds_client *mdsc,
1954                    struct ceph_mds_session *session,
1955                    int max_caps)
1956 {
1957         int trim_caps = session->s_nr_caps - max_caps;
1958
1959         dout("trim_caps mds%d start: %d / %d, trim %d\n",
1960              session->s_mds, session->s_nr_caps, max_caps, trim_caps);
1961         if (trim_caps > 0) {
1962                 int remaining = trim_caps;
1963
1964                 ceph_iterate_session_caps(session, trim_caps_cb, &remaining);
1965                 dout("trim_caps mds%d done: %d / %d, trimmed %d\n",
1966                      session->s_mds, session->s_nr_caps, max_caps,
1967                         trim_caps - remaining);
1968         }
1969
1970         ceph_flush_cap_releases(mdsc, session);
1971         return 0;
1972 }
1973
1974 static int check_caps_flush(struct ceph_mds_client *mdsc,
1975                             u64 want_flush_tid)
1976 {
1977         int ret = 1;
1978
1979         spin_lock(&mdsc->cap_dirty_lock);
1980         if (!list_empty(&mdsc->cap_flush_list)) {
1981                 struct ceph_cap_flush *cf =
1982                         list_first_entry(&mdsc->cap_flush_list,
1983                                          struct ceph_cap_flush, g_list);
1984                 if (cf->tid <= want_flush_tid) {
1985                         dout("check_caps_flush still flushing tid "
1986                              "%llu <= %llu\n", cf->tid, want_flush_tid);
1987                         ret = 0;
1988                 }
1989         }
1990         spin_unlock(&mdsc->cap_dirty_lock);
1991         return ret;
1992 }
1993
1994 /*
1995  * flush all dirty inode data to disk.
1996  *
1997  * returns true if we've flushed through want_flush_tid
1998  */
1999 static void wait_caps_flush(struct ceph_mds_client *mdsc,
2000                             u64 want_flush_tid)
2001 {
2002         dout("check_caps_flush want %llu\n", want_flush_tid);
2003
2004         wait_event(mdsc->cap_flushing_wq,
2005                    check_caps_flush(mdsc, want_flush_tid));
2006
2007         dout("check_caps_flush ok, flushed thru %llu\n", want_flush_tid);
2008 }
2009
2010 /*
2011  * called under s_mutex
2012  */
2013 static void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
2014                                    struct ceph_mds_session *session)
2015 {
2016         struct ceph_msg *msg = NULL;
2017         struct ceph_mds_cap_release *head;
2018         struct ceph_mds_cap_item *item;
2019         struct ceph_osd_client *osdc = &mdsc->fsc->client->osdc;
2020         struct ceph_cap *cap;
2021         LIST_HEAD(tmp_list);
2022         int num_cap_releases;
2023         __le32  barrier, *cap_barrier;
2024
2025         down_read(&osdc->lock);
2026         barrier = cpu_to_le32(osdc->epoch_barrier);
2027         up_read(&osdc->lock);
2028
2029         spin_lock(&session->s_cap_lock);
2030 again:
2031         list_splice_init(&session->s_cap_releases, &tmp_list);
2032         num_cap_releases = session->s_num_cap_releases;
2033         session->s_num_cap_releases = 0;
2034         spin_unlock(&session->s_cap_lock);
2035
2036         while (!list_empty(&tmp_list)) {
2037                 if (!msg) {
2038                         msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE,
2039                                         PAGE_SIZE, GFP_NOFS, false);
2040                         if (!msg)
2041                                 goto out_err;
2042                         head = msg->front.iov_base;
2043                         head->num = cpu_to_le32(0);
2044                         msg->front.iov_len = sizeof(*head);
2045
2046                         msg->hdr.version = cpu_to_le16(2);
2047                         msg->hdr.compat_version = cpu_to_le16(1);
2048                 }
2049
2050                 cap = list_first_entry(&tmp_list, struct ceph_cap,
2051                                         session_caps);
2052                 list_del(&cap->session_caps);
2053                 num_cap_releases--;
2054
2055                 head = msg->front.iov_base;
2056                 put_unaligned_le32(get_unaligned_le32(&head->num) + 1,
2057                                    &head->num);
2058                 item = msg->front.iov_base + msg->front.iov_len;
2059                 item->ino = cpu_to_le64(cap->cap_ino);
2060                 item->cap_id = cpu_to_le64(cap->cap_id);
2061                 item->migrate_seq = cpu_to_le32(cap->mseq);
2062                 item->seq = cpu_to_le32(cap->issue_seq);
2063                 msg->front.iov_len += sizeof(*item);
2064
2065                 ceph_put_cap(mdsc, cap);
2066
2067                 if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {
2068                         // Append cap_barrier field
2069                         cap_barrier = msg->front.iov_base + msg->front.iov_len;
2070                         *cap_barrier = barrier;
2071                         msg->front.iov_len += sizeof(*cap_barrier);
2072
2073                         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2074                         dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2075                         ceph_con_send(&session->s_con, msg);
2076                         msg = NULL;
2077                 }
2078         }
2079
2080         BUG_ON(num_cap_releases != 0);
2081
2082         spin_lock(&session->s_cap_lock);
2083         if (!list_empty(&session->s_cap_releases))
2084                 goto again;
2085         spin_unlock(&session->s_cap_lock);
2086
2087         if (msg) {
2088                 // Append cap_barrier field
2089                 cap_barrier = msg->front.iov_base + msg->front.iov_len;
2090                 *cap_barrier = barrier;
2091                 msg->front.iov_len += sizeof(*cap_barrier);
2092
2093                 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2094                 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
2095                 ceph_con_send(&session->s_con, msg);
2096         }
2097         return;
2098 out_err:
2099         pr_err("send_cap_releases mds%d, failed to allocate message\n",
2100                 session->s_mds);
2101         spin_lock(&session->s_cap_lock);
2102         list_splice(&tmp_list, &session->s_cap_releases);
2103         session->s_num_cap_releases += num_cap_releases;
2104         spin_unlock(&session->s_cap_lock);
2105 }
2106
2107 static void ceph_cap_release_work(struct work_struct *work)
2108 {
2109         struct ceph_mds_session *session =
2110                 container_of(work, struct ceph_mds_session, s_cap_release_work);
2111
2112         mutex_lock(&session->s_mutex);
2113         if (session->s_state == CEPH_MDS_SESSION_OPEN ||
2114             session->s_state == CEPH_MDS_SESSION_HUNG)
2115                 ceph_send_cap_releases(session->s_mdsc, session);
2116         mutex_unlock(&session->s_mutex);
2117         ceph_put_mds_session(session);
2118 }
2119
2120 void ceph_flush_cap_releases(struct ceph_mds_client *mdsc,
2121                              struct ceph_mds_session *session)
2122 {
2123         if (mdsc->stopping)
2124                 return;
2125
2126         ceph_get_mds_session(session);
2127         if (queue_work(mdsc->fsc->cap_wq,
2128                        &session->s_cap_release_work)) {
2129                 dout("cap release work queued\n");
2130         } else {
2131                 ceph_put_mds_session(session);
2132                 dout("failed to queue cap release work\n");
2133         }
2134 }
2135
2136 /*
2137  * caller holds session->s_cap_lock
2138  */
2139 void __ceph_queue_cap_release(struct ceph_mds_session *session,
2140                               struct ceph_cap *cap)
2141 {
2142         list_add_tail(&cap->session_caps, &session->s_cap_releases);
2143         session->s_num_cap_releases++;
2144
2145         if (!(session->s_num_cap_releases % CEPH_CAPS_PER_RELEASE))
2146                 ceph_flush_cap_releases(session->s_mdsc, session);
2147 }
2148
2149 static void ceph_cap_reclaim_work(struct work_struct *work)
2150 {
2151         struct ceph_mds_client *mdsc =
2152                 container_of(work, struct ceph_mds_client, cap_reclaim_work);
2153         int ret = ceph_trim_dentries(mdsc);
2154         if (ret == -EAGAIN)
2155                 ceph_queue_cap_reclaim_work(mdsc);
2156 }
2157
2158 void ceph_queue_cap_reclaim_work(struct ceph_mds_client *mdsc)
2159 {
2160         if (mdsc->stopping)
2161                 return;
2162
2163         if (queue_work(mdsc->fsc->cap_wq, &mdsc->cap_reclaim_work)) {
2164                 dout("caps reclaim work queued\n");
2165         } else {
2166                 dout("failed to queue caps release work\n");
2167         }
2168 }
2169
2170 void ceph_reclaim_caps_nr(struct ceph_mds_client *mdsc, int nr)
2171 {
2172         int val;
2173         if (!nr)
2174                 return;
2175         val = atomic_add_return(nr, &mdsc->cap_reclaim_pending);
2176         if ((val % CEPH_CAPS_PER_RELEASE) < nr) {
2177                 atomic_set(&mdsc->cap_reclaim_pending, 0);
2178                 ceph_queue_cap_reclaim_work(mdsc);
2179         }
2180 }
2181
2182 /*
2183  * requests
2184  */
2185
2186 int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
2187                                     struct inode *dir)
2188 {
2189         struct ceph_inode_info *ci = ceph_inode(dir);
2190         struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
2191         struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;
2192         size_t size = sizeof(struct ceph_mds_reply_dir_entry);
2193         unsigned int num_entries;
2194         int order;
2195
2196         spin_lock(&ci->i_ceph_lock);
2197         num_entries = ci->i_files + ci->i_subdirs;
2198         spin_unlock(&ci->i_ceph_lock);
2199         num_entries = max(num_entries, 1U);
2200         num_entries = min(num_entries, opt->max_readdir);
2201
2202         order = get_order(size * num_entries);
2203         while (order >= 0) {
2204                 rinfo->dir_entries = (void*)__get_free_pages(GFP_KERNEL |
2205                                                              __GFP_NOWARN |
2206                                                              __GFP_ZERO,
2207                                                              order);
2208                 if (rinfo->dir_entries)
2209                         break;
2210                 order--;
2211         }
2212         if (!rinfo->dir_entries)
2213                 return -ENOMEM;
2214
2215         num_entries = (PAGE_SIZE << order) / size;
2216         num_entries = min(num_entries, opt->max_readdir);
2217
2218         rinfo->dir_buf_size = PAGE_SIZE << order;
2219         req->r_num_caps = num_entries + 1;
2220         req->r_args.readdir.max_entries = cpu_to_le32(num_entries);
2221         req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);
2222         return 0;
2223 }
2224
2225 /*
2226  * Create an mds request.
2227  */
2228 struct ceph_mds_request *
2229 ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode)
2230 {
2231         struct ceph_mds_request *req;
2232
2233         req = kmem_cache_zalloc(ceph_mds_request_cachep, GFP_NOFS);
2234         if (!req)
2235                 return ERR_PTR(-ENOMEM);
2236
2237         mutex_init(&req->r_fill_mutex);
2238         req->r_mdsc = mdsc;
2239         req->r_started = jiffies;
2240         req->r_start_latency = ktime_get();
2241         req->r_resend_mds = -1;
2242         INIT_LIST_HEAD(&req->r_unsafe_dir_item);
2243         INIT_LIST_HEAD(&req->r_unsafe_target_item);
2244         req->r_fmode = -1;
2245         kref_init(&req->r_kref);
2246         RB_CLEAR_NODE(&req->r_node);
2247         INIT_LIST_HEAD(&req->r_wait);
2248         init_completion(&req->r_completion);
2249         init_completion(&req->r_safe_completion);
2250         INIT_LIST_HEAD(&req->r_unsafe_item);
2251
2252         ktime_get_coarse_real_ts64(&req->r_stamp);
2253
2254         req->r_op = op;
2255         req->r_direct_mode = mode;
2256         return req;
2257 }
2258
2259 /*
2260  * return oldest (lowest) request, tid in request tree, 0 if none.
2261  *
2262  * called under mdsc->mutex.
2263  */
2264 static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc)
2265 {
2266         if (RB_EMPTY_ROOT(&mdsc->request_tree))
2267                 return NULL;
2268         return rb_entry(rb_first(&mdsc->request_tree),
2269                         struct ceph_mds_request, r_node);
2270 }
2271
2272 static inline  u64 __get_oldest_tid(struct ceph_mds_client *mdsc)
2273 {
2274         return mdsc->oldest_tid;
2275 }
2276
2277 /*
2278  * Build a dentry's path.  Allocate on heap; caller must kfree.  Based
2279  * on build_path_from_dentry in fs/cifs/dir.c.
2280  *
2281  * If @stop_on_nosnap, generate path relative to the first non-snapped
2282  * inode.
2283  *
2284  * Encode hidden .snap dirs as a double /, i.e.
2285  *   foo/.snap/bar -> foo//bar
2286  */
2287 char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *pbase,
2288                            int stop_on_nosnap)
2289 {
2290         struct dentry *temp;
2291         char *path;
2292         int pos;
2293         unsigned seq;
2294         u64 base;
2295
2296         if (!dentry)
2297                 return ERR_PTR(-EINVAL);
2298
2299         path = __getname();
2300         if (!path)
2301                 return ERR_PTR(-ENOMEM);
2302 retry:
2303         pos = PATH_MAX - 1;
2304         path[pos] = '\0';
2305
2306         seq = read_seqbegin(&rename_lock);
2307         rcu_read_lock();
2308         temp = dentry;
2309         for (;;) {
2310                 struct inode *inode;
2311
2312                 spin_lock(&temp->d_lock);
2313                 inode = d_inode(temp);
2314                 if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
2315                         dout("build_path path+%d: %p SNAPDIR\n",
2316                              pos, temp);
2317                 } else if (stop_on_nosnap && inode && dentry != temp &&
2318                            ceph_snap(inode) == CEPH_NOSNAP) {
2319                         spin_unlock(&temp->d_lock);
2320                         pos++; /* get rid of any prepended '/' */
2321                         break;
2322                 } else {
2323                         pos -= temp->d_name.len;
2324                         if (pos < 0) {
2325                                 spin_unlock(&temp->d_lock);
2326                                 break;
2327                         }
2328                         memcpy(path + pos, temp->d_name.name, temp->d_name.len);
2329                 }
2330                 spin_unlock(&temp->d_lock);
2331                 temp = READ_ONCE(temp->d_parent);
2332
2333                 /* Are we at the root? */
2334                 if (IS_ROOT(temp))
2335                         break;
2336
2337                 /* Are we out of buffer? */
2338                 if (--pos < 0)
2339                         break;
2340
2341                 path[pos] = '/';
2342         }
2343         base = ceph_ino(d_inode(temp));
2344         rcu_read_unlock();
2345
2346         if (read_seqretry(&rename_lock, seq))
2347                 goto retry;
2348
2349         if (pos < 0) {
2350                 /*
2351                  * A rename didn't occur, but somehow we didn't end up where
2352                  * we thought we would. Throw a warning and try again.
2353                  */
2354                 pr_warn("build_path did not end path lookup where "
2355                         "expected, pos is %d\n", pos);
2356                 goto retry;
2357         }
2358
2359         *pbase = base;
2360         *plen = PATH_MAX - 1 - pos;
2361         dout("build_path on %p %d built %llx '%.*s'\n",
2362              dentry, d_count(dentry), base, *plen, path + pos);
2363         return path + pos;
2364 }
2365
2366 static int build_dentry_path(struct dentry *dentry, struct inode *dir,
2367                              const char **ppath, int *ppathlen, u64 *pino,
2368                              bool *pfreepath, bool parent_locked)
2369 {
2370         char *path;
2371
2372         rcu_read_lock();
2373         if (!dir)
2374                 dir = d_inode_rcu(dentry->d_parent);
2375         if (dir && parent_locked && ceph_snap(dir) == CEPH_NOSNAP) {
2376                 *pino = ceph_ino(dir);
2377                 rcu_read_unlock();
2378                 *ppath = dentry->d_name.name;
2379                 *ppathlen = dentry->d_name.len;
2380                 return 0;
2381         }
2382         rcu_read_unlock();
2383         path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2384         if (IS_ERR(path))
2385                 return PTR_ERR(path);
2386         *ppath = path;
2387         *pfreepath = true;
2388         return 0;
2389 }
2390
2391 static int build_inode_path(struct inode *inode,
2392                             const char **ppath, int *ppathlen, u64 *pino,
2393                             bool *pfreepath)
2394 {
2395         struct dentry *dentry;
2396         char *path;
2397
2398         if (ceph_snap(inode) == CEPH_NOSNAP) {
2399                 *pino = ceph_ino(inode);
2400                 *ppathlen = 0;
2401                 return 0;
2402         }
2403         dentry = d_find_alias(inode);
2404         path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
2405         dput(dentry);
2406         if (IS_ERR(path))
2407                 return PTR_ERR(path);
2408         *ppath = path;
2409         *pfreepath = true;
2410         return 0;
2411 }
2412
2413 /*
2414  * request arguments may be specified via an inode *, a dentry *, or
2415  * an explicit ino+path.
2416  */
2417 static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
2418                                   struct inode *rdiri, const char *rpath,
2419                                   u64 rino, const char **ppath, int *pathlen,
2420                                   u64 *ino, bool *freepath, bool parent_locked)
2421 {
2422         int r = 0;
2423
2424         if (rinode) {
2425                 r = build_inode_path(rinode, ppath, pathlen, ino, freepath);
2426                 dout(" inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
2427                      ceph_snap(rinode));
2428         } else if (rdentry) {
2429                 r = build_dentry_path(rdentry, rdiri, ppath, pathlen, ino,
2430                                         freepath, parent_locked);
2431                 dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
2432                      *ppath);
2433         } else if (rpath || rino) {
2434                 *ino = rino;
2435                 *ppath = rpath;
2436                 *pathlen = rpath ? strlen(rpath) : 0;
2437                 dout(" path %.*s\n", *pathlen, rpath);
2438         }
2439
2440         return r;
2441 }
2442
2443 static void encode_timestamp_and_gids(void **p,
2444                                       const struct ceph_mds_request *req)
2445 {
2446         struct ceph_timespec ts;
2447         int i;
2448
2449         ceph_encode_timespec64(&ts, &req->r_stamp);
2450         ceph_encode_copy(p, &ts, sizeof(ts));
2451
2452         /* gid_list */
2453         ceph_encode_32(p, req->r_cred->group_info->ngroups);
2454         for (i = 0; i < req->r_cred->group_info->ngroups; i++)
2455                 ceph_encode_64(p, from_kgid(&init_user_ns,
2456                                             req->r_cred->group_info->gid[i]));
2457 }
2458
2459 /*
2460  * called under mdsc->mutex
2461  */
2462 static struct ceph_msg *create_request_message(struct ceph_mds_session *session,
2463                                                struct ceph_mds_request *req,
2464                                                bool drop_cap_releases)
2465 {
2466         int mds = session->s_mds;
2467         struct ceph_mds_client *mdsc = session->s_mdsc;
2468         struct ceph_msg *msg;
2469         struct ceph_mds_request_head_old *head;
2470         const char *path1 = NULL;
2471         const char *path2 = NULL;
2472         u64 ino1 = 0, ino2 = 0;
2473         int pathlen1 = 0, pathlen2 = 0;
2474         bool freepath1 = false, freepath2 = false;
2475         int len;
2476         u16 releases;
2477         void *p, *end;
2478         int ret;
2479         bool legacy = !(session->s_con.peer_features & CEPH_FEATURE_FS_BTIME);
2480
2481         ret = set_request_path_attr(req->r_inode, req->r_dentry,
2482                               req->r_parent, req->r_path1, req->r_ino1.ino,
2483                               &path1, &pathlen1, &ino1, &freepath1,
2484                               test_bit(CEPH_MDS_R_PARENT_LOCKED,
2485                                         &req->r_req_flags));
2486         if (ret < 0) {
2487                 msg = ERR_PTR(ret);
2488                 goto out;
2489         }
2490
2491         /* If r_old_dentry is set, then assume that its parent is locked */
2492         ret = set_request_path_attr(NULL, req->r_old_dentry,
2493                               req->r_old_dentry_dir,
2494                               req->r_path2, req->r_ino2.ino,
2495                               &path2, &pathlen2, &ino2, &freepath2, true);
2496         if (ret < 0) {
2497                 msg = ERR_PTR(ret);
2498                 goto out_free1;
2499         }
2500
2501         len = legacy ? sizeof(*head) : sizeof(struct ceph_mds_request_head);
2502         len += pathlen1 + pathlen2 + 2*(1 + sizeof(u32) + sizeof(u64)) +
2503                 sizeof(struct ceph_timespec);
2504         len += sizeof(u32) + (sizeof(u64) * req->r_cred->group_info->ngroups);
2505
2506         /* calculate (max) length for cap releases */
2507         len += sizeof(struct ceph_mds_request_release) *
2508                 (!!req->r_inode_drop + !!req->r_dentry_drop +
2509                  !!req->r_old_inode_drop + !!req->r_old_dentry_drop);
2510
2511         if (req->r_dentry_drop)
2512                 len += pathlen1;
2513         if (req->r_old_dentry_drop)
2514                 len += pathlen2;
2515
2516         msg = ceph_msg_new2(CEPH_MSG_CLIENT_REQUEST, len, 1, GFP_NOFS, false);
2517         if (!msg) {
2518                 msg = ERR_PTR(-ENOMEM);
2519                 goto out_free2;
2520         }
2521
2522         msg->hdr.tid = cpu_to_le64(req->r_tid);
2523
2524         /*
2525          * The old ceph_mds_request_head didn't contain a version field, and
2526          * one was added when we moved the message version from 3->4.
2527          */
2528         if (legacy) {
2529                 msg->hdr.version = cpu_to_le16(3);
2530                 head = msg->front.iov_base;
2531                 p = msg->front.iov_base + sizeof(*head);
2532         } else {
2533                 struct ceph_mds_request_head *new_head = msg->front.iov_base;
2534
2535                 msg->hdr.version = cpu_to_le16(4);
2536                 new_head->version = cpu_to_le16(CEPH_MDS_REQUEST_HEAD_VERSION);
2537                 head = (struct ceph_mds_request_head_old *)&new_head->oldest_client_tid;
2538                 p = msg->front.iov_base + sizeof(*new_head);
2539         }
2540
2541         end = msg->front.iov_base + msg->front.iov_len;
2542
2543         head->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);
2544         head->op = cpu_to_le32(req->r_op);
2545         head->caller_uid = cpu_to_le32(from_kuid(&init_user_ns,
2546                                                  req->r_cred->fsuid));
2547         head->caller_gid = cpu_to_le32(from_kgid(&init_user_ns,
2548                                                  req->r_cred->fsgid));
2549         head->ino = cpu_to_le64(req->r_deleg_ino);
2550         head->args = req->r_args;
2551
2552         ceph_encode_filepath(&p, end, ino1, path1);
2553         ceph_encode_filepath(&p, end, ino2, path2);
2554
2555         /* make note of release offset, in case we need to replay */
2556         req->r_request_release_offset = p - msg->front.iov_base;
2557
2558         /* cap releases */
2559         releases = 0;
2560         if (req->r_inode_drop)
2561                 releases += ceph_encode_inode_release(&p,
2562                       req->r_inode ? req->r_inode : d_inode(req->r_dentry),
2563                       mds, req->r_inode_drop, req->r_inode_unless,
2564                       req->r_op == CEPH_MDS_OP_READDIR);
2565         if (req->r_dentry_drop)
2566                 releases += ceph_encode_dentry_release(&p, req->r_dentry,
2567                                 req->r_parent, mds, req->r_dentry_drop,
2568                                 req->r_dentry_unless);
2569         if (req->r_old_dentry_drop)
2570                 releases += ceph_encode_dentry_release(&p, req->r_old_dentry,
2571                                 req->r_old_dentry_dir, mds,
2572                                 req->r_old_dentry_drop,
2573                                 req->r_old_dentry_unless);
2574         if (req->r_old_inode_drop)
2575                 releases += ceph_encode_inode_release(&p,
2576                       d_inode(req->r_old_dentry),
2577                       mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
2578
2579         if (drop_cap_releases) {
2580                 releases = 0;
2581                 p = msg->front.iov_base + req->r_request_release_offset;
2582         }
2583
2584         head->num_releases = cpu_to_le16(releases);
2585
2586         encode_timestamp_and_gids(&p, req);
2587
2588         if (WARN_ON_ONCE(p > end)) {
2589                 ceph_msg_put(msg);
2590                 msg = ERR_PTR(-ERANGE);
2591                 goto out_free2;
2592         }
2593
2594         msg->front.iov_len = p - msg->front.iov_base;
2595         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2596
2597         if (req->r_pagelist) {
2598                 struct ceph_pagelist *pagelist = req->r_pagelist;
2599                 ceph_msg_data_add_pagelist(msg, pagelist);
2600                 msg->hdr.data_len = cpu_to_le32(pagelist->length);
2601         } else {
2602                 msg->hdr.data_len = 0;
2603         }
2604
2605         msg->hdr.data_off = cpu_to_le16(0);
2606
2607 out_free2:
2608         if (freepath2)
2609                 ceph_mdsc_free_path((char *)path2, pathlen2);
2610 out_free1:
2611         if (freepath1)
2612                 ceph_mdsc_free_path((char *)path1, pathlen1);
2613 out:
2614         return msg;
2615 }
2616
2617 /*
2618  * called under mdsc->mutex if error, under no mutex if
2619  * success.
2620  */
2621 static void complete_request(struct ceph_mds_client *mdsc,
2622                              struct ceph_mds_request *req)
2623 {
2624         req->r_end_latency = ktime_get();
2625
2626         if (req->r_callback)
2627                 req->r_callback(mdsc, req);
2628         complete_all(&req->r_completion);
2629 }
2630
2631 static struct ceph_mds_request_head_old *
2632 find_old_request_head(void *p, u64 features)
2633 {
2634         bool legacy = !(features & CEPH_FEATURE_FS_BTIME);
2635         struct ceph_mds_request_head *new_head;
2636
2637         if (legacy)
2638                 return (struct ceph_mds_request_head_old *)p;
2639         new_head = (struct ceph_mds_request_head *)p;
2640         return (struct ceph_mds_request_head_old *)&new_head->oldest_client_tid;
2641 }
2642
2643 /*
2644  * called under mdsc->mutex
2645  */
2646 static int __prepare_send_request(struct ceph_mds_session *session,
2647                                   struct ceph_mds_request *req,
2648                                   bool drop_cap_releases)
2649 {
2650         int mds = session->s_mds;
2651         struct ceph_mds_client *mdsc = session->s_mdsc;
2652         struct ceph_mds_request_head_old *rhead;
2653         struct ceph_msg *msg;
2654         int flags = 0;
2655
2656         req->r_attempts++;
2657         if (req->r_inode) {
2658                 struct ceph_cap *cap =
2659                         ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);
2660
2661                 if (cap)
2662                         req->r_sent_on_mseq = cap->mseq;
2663                 else
2664                         req->r_sent_on_mseq = -1;
2665         }
2666         dout("prepare_send_request %p tid %lld %s (attempt %d)\n", req,
2667              req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);
2668
2669         if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
2670                 void *p;
2671
2672                 /*
2673                  * Replay.  Do not regenerate message (and rebuild
2674                  * paths, etc.); just use the original message.
2675                  * Rebuilding paths will break for renames because
2676                  * d_move mangles the src name.
2677                  */
2678                 msg = req->r_request;
2679                 rhead = find_old_request_head(msg->front.iov_base,
2680                                               session->s_con.peer_features);
2681
2682                 flags = le32_to_cpu(rhead->flags);
2683                 flags |= CEPH_MDS_FLAG_REPLAY;
2684                 rhead->flags = cpu_to_le32(flags);
2685
2686                 if (req->r_target_inode)
2687                         rhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));
2688
2689                 rhead->num_retry = req->r_attempts - 1;
2690
2691                 /* remove cap/dentry releases from message */
2692                 rhead->num_releases = 0;
2693
2694                 p = msg->front.iov_base + req->r_request_release_offset;
2695                 encode_timestamp_and_gids(&p, req);
2696
2697                 msg->front.iov_len = p - msg->front.iov_base;
2698                 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2699                 return 0;
2700         }
2701
2702         if (req->r_request) {
2703                 ceph_msg_put(req->r_request);
2704                 req->r_request = NULL;
2705         }
2706         msg = create_request_message(session, req, drop_cap_releases);
2707         if (IS_ERR(msg)) {
2708                 req->r_err = PTR_ERR(msg);
2709                 return PTR_ERR(msg);
2710         }
2711         req->r_request = msg;
2712
2713         rhead = find_old_request_head(msg->front.iov_base,
2714                                       session->s_con.peer_features);
2715         rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
2716         if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2717                 flags |= CEPH_MDS_FLAG_REPLAY;
2718         if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags))
2719                 flags |= CEPH_MDS_FLAG_ASYNC;
2720         if (req->r_parent)
2721                 flags |= CEPH_MDS_FLAG_WANT_DENTRY;
2722         rhead->flags = cpu_to_le32(flags);
2723         rhead->num_fwd = req->r_num_fwd;
2724         rhead->num_retry = req->r_attempts - 1;
2725
2726         dout(" r_parent = %p\n", req->r_parent);
2727         return 0;
2728 }
2729
2730 /*
2731  * called under mdsc->mutex
2732  */
2733 static int __send_request(struct ceph_mds_session *session,
2734                           struct ceph_mds_request *req,
2735                           bool drop_cap_releases)
2736 {
2737         int err;
2738
2739         err = __prepare_send_request(session, req, drop_cap_releases);
2740         if (!err) {
2741                 ceph_msg_get(req->r_request);
2742                 ceph_con_send(&session->s_con, req->r_request);
2743         }
2744
2745         return err;
2746 }
2747
2748 /*
2749  * send request, or put it on the appropriate wait list.
2750  */
2751 static void __do_request(struct ceph_mds_client *mdsc,
2752                         struct ceph_mds_request *req)
2753 {
2754         struct ceph_mds_session *session = NULL;
2755         int mds = -1;
2756         int err = 0;
2757         bool random;
2758
2759         if (req->r_err || test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
2760                 if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags))
2761                         __unregister_request(mdsc, req);
2762                 return;
2763         }
2764
2765         if (req->r_timeout &&
2766             time_after_eq(jiffies, req->r_started + req->r_timeout)) {
2767                 dout("do_request timed out\n");
2768                 err = -ETIMEDOUT;
2769                 goto finish;
2770         }
2771         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
2772                 dout("do_request forced umount\n");
2773                 err = -EIO;
2774                 goto finish;
2775         }
2776         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_MOUNTING) {
2777                 if (mdsc->mdsmap_err) {
2778                         err = mdsc->mdsmap_err;
2779                         dout("do_request mdsmap err %d\n", err);
2780                         goto finish;
2781                 }
2782                 if (mdsc->mdsmap->m_epoch == 0) {
2783                         dout("do_request no mdsmap, waiting for map\n");
2784                         list_add(&req->r_wait, &mdsc->waiting_for_map);
2785                         return;
2786                 }
2787                 if (!(mdsc->fsc->mount_options->flags &
2788                       CEPH_MOUNT_OPT_MOUNTWAIT) &&
2789                     !ceph_mdsmap_is_cluster_available(mdsc->mdsmap)) {
2790                         err = -EHOSTUNREACH;
2791                         goto finish;
2792                 }
2793         }
2794
2795         put_request_session(req);
2796
2797         mds = __choose_mds(mdsc, req, &random);
2798         if (mds < 0 ||
2799             ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {
2800                 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2801                         err = -EJUKEBOX;
2802                         goto finish;
2803                 }
2804                 dout("do_request no mds or not active, waiting for map\n");
2805                 list_add(&req->r_wait, &mdsc->waiting_for_map);
2806                 return;
2807         }
2808
2809         /* get, open session */
2810         session = __ceph_lookup_mds_session(mdsc, mds);
2811         if (!session) {
2812                 session = register_session(mdsc, mds);
2813                 if (IS_ERR(session)) {
2814                         err = PTR_ERR(session);
2815                         goto finish;
2816                 }
2817         }
2818         req->r_session = ceph_get_mds_session(session);
2819
2820         dout("do_request mds%d session %p state %s\n", mds, session,
2821              ceph_session_state_name(session->s_state));
2822         if (session->s_state != CEPH_MDS_SESSION_OPEN &&
2823             session->s_state != CEPH_MDS_SESSION_HUNG) {
2824                 /*
2825                  * We cannot queue async requests since the caps and delegated
2826                  * inodes are bound to the session. Just return -EJUKEBOX and
2827                  * let the caller retry a sync request in that case.
2828                  */
2829                 if (test_bit(CEPH_MDS_R_ASYNC, &req->r_req_flags)) {
2830                         err = -EJUKEBOX;
2831                         goto out_session;
2832                 }
2833
2834                 /*
2835                  * If the session has been REJECTED, then return a hard error,
2836                  * unless it's a CLEANRECOVER mount, in which case we'll queue
2837                  * it to the mdsc queue.
2838                  */
2839                 if (session->s_state == CEPH_MDS_SESSION_REJECTED) {
2840                         if (ceph_test_mount_opt(mdsc->fsc, CLEANRECOVER))
2841                                 list_add(&req->r_wait, &mdsc->waiting_for_map);
2842                         else
2843                                 err = -EACCES;
2844                         goto out_session;
2845                 }
2846
2847                 if (session->s_state == CEPH_MDS_SESSION_NEW ||
2848                     session->s_state == CEPH_MDS_SESSION_CLOSING) {
2849                         err = __open_session(mdsc, session);
2850                         if (err)
2851                                 goto out_session;
2852                         /* retry the same mds later */
2853                         if (random)
2854                                 req->r_resend_mds = mds;
2855                 }
2856                 list_add(&req->r_wait, &session->s_waiting);
2857                 goto out_session;
2858         }
2859
2860         /* send request */
2861         req->r_resend_mds = -1;   /* forget any previous mds hint */
2862
2863         if (req->r_request_started == 0)   /* note request start time */
2864                 req->r_request_started = jiffies;
2865
2866         err = __send_request(session, req, false);
2867
2868 out_session:
2869         ceph_put_mds_session(session);
2870 finish:
2871         if (err) {
2872                 dout("__do_request early error %d\n", err);
2873                 req->r_err = err;
2874                 complete_request(mdsc, req);
2875                 __unregister_request(mdsc, req);
2876         }
2877         return;
2878 }
2879
2880 /*
2881  * called under mdsc->mutex
2882  */
2883 static void __wake_requests(struct ceph_mds_client *mdsc,
2884                             struct list_head *head)
2885 {
2886         struct ceph_mds_request *req;
2887         LIST_HEAD(tmp_list);
2888
2889         list_splice_init(head, &tmp_list);
2890
2891         while (!list_empty(&tmp_list)) {
2892                 req = list_entry(tmp_list.next,
2893                                  struct ceph_mds_request, r_wait);
2894                 list_del_init(&req->r_wait);
2895                 dout(" wake request %p tid %llu\n", req, req->r_tid);
2896                 __do_request(mdsc, req);
2897         }
2898 }
2899
2900 /*
2901  * Wake up threads with requests pending for @mds, so that they can
2902  * resubmit their requests to a possibly different mds.
2903  */
2904 static void kick_requests(struct ceph_mds_client *mdsc, int mds)
2905 {
2906         struct ceph_mds_request *req;
2907         struct rb_node *p = rb_first(&mdsc->request_tree);
2908
2909         dout("kick_requests mds%d\n", mds);
2910         while (p) {
2911                 req = rb_entry(p, struct ceph_mds_request, r_node);
2912                 p = rb_next(p);
2913                 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
2914                         continue;
2915                 if (req->r_attempts > 0)
2916                         continue; /* only new requests */
2917                 if (req->r_session &&
2918                     req->r_session->s_mds == mds) {
2919                         dout(" kicking tid %llu\n", req->r_tid);
2920                         list_del_init(&req->r_wait);
2921                         __do_request(mdsc, req);
2922                 }
2923         }
2924 }
2925
2926 int ceph_mdsc_submit_request(struct ceph_mds_client *mdsc, struct inode *dir,
2927                               struct ceph_mds_request *req)
2928 {
2929         int err = 0;
2930
2931         /* take CAP_PIN refs for r_inode, r_parent, r_old_dentry */
2932         if (req->r_inode)
2933                 ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
2934         if (req->r_parent) {
2935                 struct ceph_inode_info *ci = ceph_inode(req->r_parent);
2936                 int fmode = (req->r_op & CEPH_MDS_OP_WRITE) ?
2937                             CEPH_FILE_MODE_WR : CEPH_FILE_MODE_RD;
2938                 spin_lock(&ci->i_ceph_lock);
2939                 ceph_take_cap_refs(ci, CEPH_CAP_PIN, false);
2940                 __ceph_touch_fmode(ci, mdsc, fmode);
2941                 spin_unlock(&ci->i_ceph_lock);
2942         }
2943         if (req->r_old_dentry_dir)
2944                 ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
2945                                   CEPH_CAP_PIN);
2946
2947         if (req->r_inode) {
2948                 err = ceph_wait_on_async_create(req->r_inode);
2949                 if (err) {
2950                         dout("%s: wait for async create returned: %d\n",
2951                              __func__, err);
2952                         return err;
2953                 }
2954         }
2955
2956         if (!err && req->r_old_inode) {
2957                 err = ceph_wait_on_async_create(req->r_old_inode);
2958                 if (err) {
2959                         dout("%s: wait for async create returned: %d\n",
2960                              __func__, err);
2961                         return err;
2962                 }
2963         }
2964
2965         dout("submit_request on %p for inode %p\n", req, dir);
2966         mutex_lock(&mdsc->mutex);
2967         __register_request(mdsc, req, dir);
2968         __do_request(mdsc, req);
2969         err = req->r_err;
2970         mutex_unlock(&mdsc->mutex);
2971         return err;
2972 }
2973
2974 int ceph_mdsc_wait_request(struct ceph_mds_client *mdsc,
2975                            struct ceph_mds_request *req,
2976                            ceph_mds_request_wait_callback_t wait_func)
2977 {
2978         int err;
2979
2980         /* wait */
2981         dout("do_request waiting\n");
2982         if (wait_func) {
2983                 err = wait_func(mdsc, req);
2984         } else {
2985                 long timeleft = wait_for_completion_killable_timeout(
2986                                         &req->r_completion,
2987                                         ceph_timeout_jiffies(req->r_timeout));
2988                 if (timeleft > 0)
2989                         err = 0;
2990                 else if (!timeleft)
2991                         err = -ETIMEDOUT;  /* timed out */
2992                 else
2993                         err = timeleft;  /* killed */
2994         }
2995         dout("do_request waited, got %d\n", err);
2996         mutex_lock(&mdsc->mutex);
2997
2998         /* only abort if we didn't race with a real reply */
2999         if (test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags)) {
3000                 err = le32_to_cpu(req->r_reply_info.head->result);
3001         } else if (err < 0) {
3002                 dout("aborted request %lld with %d\n", req->r_tid, err);
3003
3004                 /*
3005                  * ensure we aren't running concurrently with
3006                  * ceph_fill_trace or ceph_readdir_prepopulate, which
3007                  * rely on locks (dir mutex) held by our caller.
3008                  */
3009                 mutex_lock(&req->r_fill_mutex);
3010                 req->r_err = err;
3011                 set_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags);
3012                 mutex_unlock(&req->r_fill_mutex);
3013
3014                 if (req->r_parent &&
3015                     (req->r_op & CEPH_MDS_OP_WRITE))
3016                         ceph_invalidate_dir_request(req);
3017         } else {
3018                 err = req->r_err;
3019         }
3020
3021         mutex_unlock(&mdsc->mutex);
3022         return err;
3023 }
3024
3025 /*
3026  * Synchrously perform an mds request.  Take care of all of the
3027  * session setup, forwarding, retry details.
3028  */
3029 int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
3030                          struct inode *dir,
3031                          struct ceph_mds_request *req)
3032 {
3033         int err;
3034
3035         dout("do_request on %p\n", req);
3036
3037         /* issue */
3038         err = ceph_mdsc_submit_request(mdsc, dir, req);
3039         if (!err)
3040                 err = ceph_mdsc_wait_request(mdsc, req, NULL);
3041         dout("do_request %p done, result %d\n", req, err);
3042         return err;
3043 }
3044
3045 /*
3046  * Invalidate dir's completeness, dentry lease state on an aborted MDS
3047  * namespace request.
3048  */
3049 void ceph_invalidate_dir_request(struct ceph_mds_request *req)
3050 {
3051         struct inode *dir = req->r_parent;
3052         struct inode *old_dir = req->r_old_dentry_dir;
3053
3054         dout("invalidate_dir_request %p %p (complete, lease(s))\n", dir, old_dir);
3055
3056         ceph_dir_clear_complete(dir);
3057         if (old_dir)
3058                 ceph_dir_clear_complete(old_dir);
3059         if (req->r_dentry)
3060                 ceph_invalidate_dentry_lease(req->r_dentry);
3061         if (req->r_old_dentry)
3062                 ceph_invalidate_dentry_lease(req->r_old_dentry);
3063 }
3064
3065 /*
3066  * Handle mds reply.
3067  *
3068  * We take the session mutex and parse and process the reply immediately.
3069  * This preserves the logical ordering of replies, capabilities, etc., sent
3070  * by the MDS as they are applied to our local cache.
3071  */
3072 static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
3073 {
3074         struct ceph_mds_client *mdsc = session->s_mdsc;
3075         struct ceph_mds_request *req;
3076         struct ceph_mds_reply_head *head = msg->front.iov_base;
3077         struct ceph_mds_reply_info_parsed *rinfo;  /* parsed reply info */
3078         struct ceph_snap_realm *realm;
3079         u64 tid;
3080         int err, result;
3081         int mds = session->s_mds;
3082
3083         if (msg->front.iov_len < sizeof(*head)) {
3084                 pr_err("mdsc_handle_reply got corrupt (short) reply\n");
3085                 ceph_msg_dump(msg);
3086                 return;
3087         }
3088
3089         /* get request, session */
3090         tid = le64_to_cpu(msg->hdr.tid);
3091         mutex_lock(&mdsc->mutex);
3092         req = lookup_get_request(mdsc, tid);
3093         if (!req) {
3094                 dout("handle_reply on unknown tid %llu\n", tid);
3095                 mutex_unlock(&mdsc->mutex);
3096                 return;
3097         }
3098         dout("handle_reply %p\n", req);
3099
3100         /* correct session? */
3101         if (req->r_session != session) {
3102                 pr_err("mdsc_handle_reply got %llu on session mds%d"
3103                        " not mds%d\n", tid, session->s_mds,
3104                        req->r_session ? req->r_session->s_mds : -1);
3105                 mutex_unlock(&mdsc->mutex);
3106                 goto out;
3107         }
3108
3109         /* dup? */
3110         if ((test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags) && !head->safe) ||
3111             (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags) && head->safe)) {
3112                 pr_warn("got a dup %s reply on %llu from mds%d\n",
3113                            head->safe ? "safe" : "unsafe", tid, mds);
3114                 mutex_unlock(&mdsc->mutex);
3115                 goto out;
3116         }
3117         if (test_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags)) {
3118                 pr_warn("got unsafe after safe on %llu from mds%d\n",
3119                            tid, mds);
3120                 mutex_unlock(&mdsc->mutex);
3121                 goto out;
3122         }
3123
3124         result = le32_to_cpu(head->result);
3125
3126         if (head->safe) {
3127                 set_bit(CEPH_MDS_R_GOT_SAFE, &req->r_req_flags);
3128                 __unregister_request(mdsc, req);
3129
3130                 /* last request during umount? */
3131                 if (mdsc->stopping && !__get_oldest_req(mdsc))
3132                         complete_all(&mdsc->safe_umount_waiters);
3133
3134                 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3135                         /*
3136                          * We already handled the unsafe response, now do the
3137                          * cleanup.  No need to examine the response; the MDS
3138                          * doesn't include any result info in the safe
3139                          * response.  And even if it did, there is nothing
3140                          * useful we could do with a revised return value.
3141                          */
3142                         dout("got safe reply %llu, mds%d\n", tid, mds);
3143
3144                         mutex_unlock(&mdsc->mutex);
3145                         goto out;
3146                 }
3147         } else {
3148                 set_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags);
3149                 list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
3150         }
3151
3152         dout("handle_reply tid %lld result %d\n", tid, result);
3153         rinfo = &req->r_reply_info;
3154         if (test_bit(CEPHFS_FEATURE_REPLY_ENCODING, &session->s_features))
3155                 err = parse_reply_info(session, msg, rinfo, (u64)-1);
3156         else
3157                 err = parse_reply_info(session, msg, rinfo, session->s_con.peer_features);
3158         mutex_unlock(&mdsc->mutex);
3159
3160         /* Must find target inode outside of mutexes to avoid deadlocks */
3161         if ((err >= 0) && rinfo->head->is_target) {
3162                 struct inode *in;
3163                 struct ceph_vino tvino = {
3164                         .ino  = le64_to_cpu(rinfo->targeti.in->ino),
3165                         .snap = le64_to_cpu(rinfo->targeti.in->snapid)
3166                 };
3167
3168                 in = ceph_get_inode(mdsc->fsc->sb, tvino);
3169                 if (IS_ERR(in)) {
3170                         err = PTR_ERR(in);
3171                         mutex_lock(&session->s_mutex);
3172                         goto out_err;
3173                 }
3174                 req->r_target_inode = in;
3175         }
3176
3177         mutex_lock(&session->s_mutex);
3178         if (err < 0) {
3179                 pr_err("mdsc_handle_reply got corrupt reply mds%d(tid:%lld)\n", mds, tid);
3180                 ceph_msg_dump(msg);
3181                 goto out_err;
3182         }
3183
3184         /* snap trace */
3185         realm = NULL;
3186         if (rinfo->snapblob_len) {
3187                 down_write(&mdsc->snap_rwsem);
3188                 ceph_update_snap_trace(mdsc, rinfo->snapblob,
3189                                 rinfo->snapblob + rinfo->snapblob_len,
3190                                 le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
3191                                 &realm);
3192                 downgrade_write(&mdsc->snap_rwsem);
3193         } else {
3194                 down_read(&mdsc->snap_rwsem);
3195         }
3196
3197         /* insert trace into our cache */
3198         mutex_lock(&req->r_fill_mutex);
3199         current->journal_info = req;
3200         err = ceph_fill_trace(mdsc->fsc->sb, req);
3201         if (err == 0) {
3202                 if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
3203                                     req->r_op == CEPH_MDS_OP_LSSNAP))
3204                         ceph_readdir_prepopulate(req, req->r_session);
3205         }
3206         current->journal_info = NULL;
3207         mutex_unlock(&req->r_fill_mutex);
3208
3209         up_read(&mdsc->snap_rwsem);
3210         if (realm)
3211                 ceph_put_snap_realm(mdsc, realm);
3212
3213         if (err == 0) {
3214                 if (req->r_target_inode &&
3215                     test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags)) {
3216                         struct ceph_inode_info *ci =
3217                                 ceph_inode(req->r_target_inode);
3218                         spin_lock(&ci->i_unsafe_lock);
3219                         list_add_tail(&req->r_unsafe_target_item,
3220                                       &ci->i_unsafe_iops);
3221                         spin_unlock(&ci->i_unsafe_lock);
3222                 }
3223
3224                 ceph_unreserve_caps(mdsc, &req->r_caps_reservation);
3225         }
3226 out_err:
3227         mutex_lock(&mdsc->mutex);
3228         if (!test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3229                 if (err) {
3230                         req->r_err = err;
3231                 } else {
3232                         req->r_reply =  ceph_msg_get(msg);
3233                         set_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags);
3234                 }
3235         } else {
3236                 dout("reply arrived after request %lld was aborted\n", tid);
3237         }
3238         mutex_unlock(&mdsc->mutex);
3239
3240         mutex_unlock(&session->s_mutex);
3241
3242         /* kick calling process */
3243         complete_request(mdsc, req);
3244
3245         ceph_update_metadata_metrics(&mdsc->metric, req->r_start_latency,
3246                                      req->r_end_latency, err);
3247 out:
3248         ceph_mdsc_put_request(req);
3249         return;
3250 }
3251
3252
3253
3254 /*
3255  * handle mds notification that our request has been forwarded.
3256  */
3257 static void handle_forward(struct ceph_mds_client *mdsc,
3258                            struct ceph_mds_session *session,
3259                            struct ceph_msg *msg)
3260 {
3261         struct ceph_mds_request *req;
3262         u64 tid = le64_to_cpu(msg->hdr.tid);
3263         u32 next_mds;
3264         u32 fwd_seq;
3265         int err = -EINVAL;
3266         void *p = msg->front.iov_base;
3267         void *end = p + msg->front.iov_len;
3268
3269         ceph_decode_need(&p, end, 2*sizeof(u32), bad);
3270         next_mds = ceph_decode_32(&p);
3271         fwd_seq = ceph_decode_32(&p);
3272
3273         mutex_lock(&mdsc->mutex);
3274         req = lookup_get_request(mdsc, tid);
3275         if (!req) {
3276                 dout("forward tid %llu to mds%d - req dne\n", tid, next_mds);
3277                 goto out;  /* dup reply? */
3278         }
3279
3280         if (test_bit(CEPH_MDS_R_ABORTED, &req->r_req_flags)) {
3281                 dout("forward tid %llu aborted, unregistering\n", tid);
3282                 __unregister_request(mdsc, req);
3283         } else if (fwd_seq <= req->r_num_fwd) {
3284                 dout("forward tid %llu to mds%d - old seq %d <= %d\n",
3285                      tid, next_mds, req->r_num_fwd, fwd_seq);
3286         } else {
3287                 /* resend. forward race not possible; mds would drop */
3288                 dout("forward tid %llu to mds%d (we resend)\n", tid, next_mds);
3289                 BUG_ON(req->r_err);
3290                 BUG_ON(test_bit(CEPH_MDS_R_GOT_RESULT, &req->r_req_flags));
3291                 req->r_attempts = 0;
3292                 req->r_num_fwd = fwd_seq;
3293                 req->r_resend_mds = next_mds;
3294                 put_request_session(req);
3295                 __do_request(mdsc, req);
3296         }
3297         ceph_mdsc_put_request(req);
3298 out:
3299         mutex_unlock(&mdsc->mutex);
3300         return;
3301
3302 bad:
3303         pr_err("mdsc_handle_forward decode error err=%d\n", err);
3304 }
3305
3306 static int __decode_session_metadata(void **p, void *end,
3307                                      bool *blocklisted)
3308 {
3309         /* map<string,string> */
3310         u32 n;
3311         bool err_str;
3312         ceph_decode_32_safe(p, end, n, bad);
3313         while (n-- > 0) {
3314                 u32 len;
3315                 ceph_decode_32_safe(p, end, len, bad);
3316                 ceph_decode_need(p, end, len, bad);
3317                 err_str = !strncmp(*p, "error_string", len);
3318                 *p += len;
3319                 ceph_decode_32_safe(p, end, len, bad);
3320                 ceph_decode_need(p, end, len, bad);
3321                 /*
3322                  * Match "blocklisted (blacklisted)" from newer MDSes,
3323                  * or "blacklisted" from older MDSes.
3324                  */
3325                 if (err_str && strnstr(*p, "blacklisted", len))
3326                         *blocklisted = true;
3327                 *p += len;
3328         }
3329         return 0;
3330 bad:
3331         return -1;
3332 }
3333
3334 /*
3335  * handle a mds session control message
3336  */
3337 static void handle_session(struct ceph_mds_session *session,
3338                            struct ceph_msg *msg)
3339 {
3340         struct ceph_mds_client *mdsc = session->s_mdsc;
3341         int mds = session->s_mds;
3342         int msg_version = le16_to_cpu(msg->hdr.version);
3343         void *p = msg->front.iov_base;
3344         void *end = p + msg->front.iov_len;
3345         struct ceph_mds_session_head *h;
3346         u32 op;
3347         u64 seq, features = 0;
3348         int wake = 0;
3349         bool blocklisted = false;
3350
3351         /* decode */
3352         ceph_decode_need(&p, end, sizeof(*h), bad);
3353         h = p;
3354         p += sizeof(*h);
3355
3356         op = le32_to_cpu(h->op);
3357         seq = le64_to_cpu(h->seq);
3358
3359         if (msg_version >= 3) {
3360                 u32 len;
3361                 /* version >= 2 and < 5, decode metadata, skip otherwise
3362                  * as it's handled via flags.
3363                  */
3364                 if (msg_version >= 5)
3365                         ceph_decode_skip_map(&p, end, string, string, bad);
3366                 else if (__decode_session_metadata(&p, end, &blocklisted) < 0)
3367                         goto bad;
3368
3369                 /* version >= 3, feature bits */
3370                 ceph_decode_32_safe(&p, end, len, bad);
3371                 if (len) {
3372                         ceph_decode_64_safe(&p, end, features, bad);
3373                         p += len - sizeof(features);
3374                 }
3375         }
3376
3377         if (msg_version >= 5) {
3378                 u32 flags;
3379                 /* version >= 4, struct_v, struct_cv, len, metric_spec */
3380                 ceph_decode_skip_n(&p, end, 2 + sizeof(u32) * 2, bad);
3381                 /* version >= 5, flags   */
3382                 ceph_decode_32_safe(&p, end, flags, bad);
3383                 if (flags & CEPH_SESSION_BLOCKLISTED) {
3384                         pr_warn("mds%d session blocklisted\n", session->s_mds);
3385                         blocklisted = true;
3386                 }
3387         }
3388
3389         mutex_lock(&mdsc->mutex);
3390         if (op == CEPH_SESSION_CLOSE) {
3391                 ceph_get_mds_session(session);
3392                 __unregister_session(mdsc, session);
3393         }
3394         /* FIXME: this ttl calculation is generous */
3395         session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;
3396         mutex_unlock(&mdsc->mutex);
3397
3398         mutex_lock(&session->s_mutex);
3399
3400         dout("handle_session mds%d %s %p state %s seq %llu\n",
3401              mds, ceph_session_op_name(op), session,
3402              ceph_session_state_name(session->s_state), seq);
3403
3404         if (session->s_state == CEPH_MDS_SESSION_HUNG) {
3405                 session->s_state = CEPH_MDS_SESSION_OPEN;
3406                 pr_info("mds%d came back\n", session->s_mds);
3407         }
3408
3409         switch (op) {
3410         case CEPH_SESSION_OPEN:
3411                 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3412                         pr_info("mds%d reconnect success\n", session->s_mds);
3413                 session->s_state = CEPH_MDS_SESSION_OPEN;
3414                 session->s_features = features;
3415                 renewed_caps(mdsc, session, 0);
3416                 if (test_bit(CEPHFS_FEATURE_METRIC_COLLECT, &session->s_features))
3417                         metric_schedule_delayed(&mdsc->metric);
3418                 wake = 1;
3419                 if (mdsc->stopping)
3420                         __close_session(mdsc, session);
3421                 break;
3422
3423         case CEPH_SESSION_RENEWCAPS:
3424                 if (session->s_renew_seq == seq)
3425                         renewed_caps(mdsc, session, 1);
3426                 break;
3427
3428         case CEPH_SESSION_CLOSE:
3429                 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
3430                         pr_info("mds%d reconnect denied\n", session->s_mds);
3431                 session->s_state = CEPH_MDS_SESSION_CLOSED;
3432                 cleanup_session_requests(mdsc, session);
3433                 remove_session_caps(session);
3434                 wake = 2; /* for good measure */
3435                 wake_up_all(&mdsc->session_close_wq);
3436                 break;
3437
3438         case CEPH_SESSION_STALE:
3439                 pr_info("mds%d caps went stale, renewing\n",
3440                         session->s_mds);
3441                 atomic_inc(&session->s_cap_gen);
3442                 session->s_cap_ttl = jiffies - 1;
3443                 send_renew_caps(mdsc, session);
3444                 break;
3445
3446         case CEPH_SESSION_RECALL_STATE:
3447                 ceph_trim_caps(mdsc, session, le32_to_cpu(h->max_caps));
3448                 break;
3449
3450         case CEPH_SESSION_FLUSHMSG:
3451                 send_flushmsg_ack(mdsc, session, seq);
3452                 break;
3453
3454         case CEPH_SESSION_FORCE_RO:
3455                 dout("force_session_readonly %p\n", session);
3456                 spin_lock(&session->s_cap_lock);
3457                 session->s_readonly = true;
3458                 spin_unlock(&session->s_cap_lock);
3459                 wake_up_session_caps(session, FORCE_RO);
3460                 break;
3461
3462         case CEPH_SESSION_REJECT:
3463                 WARN_ON(session->s_state != CEPH_MDS_SESSION_OPENING);
3464                 pr_info("mds%d rejected session\n", session->s_mds);
3465                 session->s_state = CEPH_MDS_SESSION_REJECTED;
3466                 cleanup_session_requests(mdsc, session);
3467                 remove_session_caps(session);
3468                 if (blocklisted)
3469                         mdsc->fsc->blocklisted = true;
3470                 wake = 2; /* for good measure */
3471                 break;
3472
3473         default:
3474                 pr_err("mdsc_handle_session bad op %d mds%d\n", op, mds);
3475                 WARN_ON(1);
3476         }
3477
3478         mutex_unlock(&session->s_mutex);
3479         if (wake) {
3480                 mutex_lock(&mdsc->mutex);
3481                 __wake_requests(mdsc, &session->s_waiting);
3482                 if (wake == 2)
3483                         kick_requests(mdsc, mds);
3484                 mutex_unlock(&mdsc->mutex);
3485         }
3486         if (op == CEPH_SESSION_CLOSE)
3487                 ceph_put_mds_session(session);
3488         return;
3489
3490 bad:
3491         pr_err("mdsc_handle_session corrupt message mds%d len %d\n", mds,
3492                (int)msg->front.iov_len);
3493         ceph_msg_dump(msg);
3494         return;
3495 }
3496
3497 void ceph_mdsc_release_dir_caps(struct ceph_mds_request *req)
3498 {
3499         int dcaps;
3500
3501         dcaps = xchg(&req->r_dir_caps, 0);
3502         if (dcaps) {
3503                 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3504                 ceph_put_cap_refs(ceph_inode(req->r_parent), dcaps);
3505         }
3506 }
3507
3508 void ceph_mdsc_release_dir_caps_no_check(struct ceph_mds_request *req)
3509 {
3510         int dcaps;
3511
3512         dcaps = xchg(&req->r_dir_caps, 0);
3513         if (dcaps) {
3514                 dout("releasing r_dir_caps=%s\n", ceph_cap_string(dcaps));
3515                 ceph_put_cap_refs_no_check_caps(ceph_inode(req->r_parent),
3516                                                 dcaps);
3517         }
3518 }
3519
3520 /*
3521  * called under session->mutex.
3522  */
3523 static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
3524                                    struct ceph_mds_session *session)
3525 {
3526         struct ceph_mds_request *req, *nreq;
3527         struct rb_node *p;
3528
3529         dout("replay_unsafe_requests mds%d\n", session->s_mds);
3530
3531         mutex_lock(&mdsc->mutex);
3532         list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item)
3533                 __send_request(session, req, true);
3534
3535         /*
3536          * also re-send old requests when MDS enters reconnect stage. So that MDS
3537          * can process completed request in clientreplay stage.
3538          */
3539         p = rb_first(&mdsc->request_tree);
3540         while (p) {
3541                 req = rb_entry(p, struct ceph_mds_request, r_node);
3542                 p = rb_next(p);
3543                 if (test_bit(CEPH_MDS_R_GOT_UNSAFE, &req->r_req_flags))
3544                         continue;
3545                 if (req->r_attempts == 0)
3546                         continue; /* only old requests */
3547                 if (!req->r_session)
3548                         continue;
3549                 if (req->r_session->s_mds != session->s_mds)
3550                         continue;
3551
3552                 ceph_mdsc_release_dir_caps_no_check(req);
3553
3554                 __send_request(session, req, true);
3555         }
3556         mutex_unlock(&mdsc->mutex);
3557 }
3558
3559 static int send_reconnect_partial(struct ceph_reconnect_state *recon_state)
3560 {
3561         struct ceph_msg *reply;
3562         struct ceph_pagelist *_pagelist;
3563         struct page *page;
3564         __le32 *addr;
3565         int err = -ENOMEM;
3566
3567         if (!recon_state->allow_multi)
3568                 return -ENOSPC;
3569
3570         /* can't handle message that contains both caps and realm */
3571         BUG_ON(!recon_state->nr_caps == !recon_state->nr_realms);
3572
3573         /* pre-allocate new pagelist */
3574         _pagelist = ceph_pagelist_alloc(GFP_NOFS);
3575         if (!_pagelist)
3576                 return -ENOMEM;
3577
3578         reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3579         if (!reply)
3580                 goto fail_msg;
3581
3582         /* placeholder for nr_caps */
3583         err = ceph_pagelist_encode_32(_pagelist, 0);
3584         if (err < 0)
3585                 goto fail;
3586
3587         if (recon_state->nr_caps) {
3588                 /* currently encoding caps */
3589                 err = ceph_pagelist_encode_32(recon_state->pagelist, 0);
3590                 if (err)
3591                         goto fail;
3592         } else {
3593                 /* placeholder for nr_realms (currently encoding relams) */
3594                 err = ceph_pagelist_encode_32(_pagelist, 0);
3595                 if (err < 0)
3596                         goto fail;
3597         }
3598
3599         err = ceph_pagelist_encode_8(recon_state->pagelist, 1);
3600         if (err)
3601                 goto fail;
3602
3603         page = list_first_entry(&recon_state->pagelist->head, struct page, lru);
3604         addr = kmap_atomic(page);
3605         if (recon_state->nr_caps) {
3606                 /* currently encoding caps */
3607                 *addr = cpu_to_le32(recon_state->nr_caps);
3608         } else {
3609                 /* currently encoding relams */
3610                 *(addr + 1) = cpu_to_le32(recon_state->nr_realms);
3611         }
3612         kunmap_atomic(addr);
3613
3614         reply->hdr.version = cpu_to_le16(5);
3615         reply->hdr.compat_version = cpu_to_le16(4);
3616
3617         reply->hdr.data_len = cpu_to_le32(recon_state->pagelist->length);
3618         ceph_msg_data_add_pagelist(reply, recon_state->pagelist);
3619
3620         ceph_con_send(&recon_state->session->s_con, reply);
3621         ceph_pagelist_release(recon_state->pagelist);
3622
3623         recon_state->pagelist = _pagelist;
3624         recon_state->nr_caps = 0;
3625         recon_state->nr_realms = 0;
3626         recon_state->msg_version = 5;
3627         return 0;
3628 fail:
3629         ceph_msg_put(reply);
3630 fail_msg:
3631         ceph_pagelist_release(_pagelist);
3632         return err;
3633 }
3634
3635 static struct dentry* d_find_primary(struct inode *inode)
3636 {
3637         struct dentry *alias, *dn = NULL;
3638
3639         if (hlist_empty(&inode->i_dentry))
3640                 return NULL;
3641
3642         spin_lock(&inode->i_lock);
3643         if (hlist_empty(&inode->i_dentry))
3644                 goto out_unlock;
3645
3646         if (S_ISDIR(inode->i_mode)) {
3647                 alias = hlist_entry(inode->i_dentry.first, struct dentry, d_u.d_alias);
3648                 if (!IS_ROOT(alias))
3649                         dn = dget(alias);
3650                 goto out_unlock;
3651         }
3652
3653         hlist_for_each_entry(alias, &inode->i_dentry, d_u.d_alias) {
3654                 spin_lock(&alias->d_lock);
3655                 if (!d_unhashed(alias) &&
3656                     (ceph_dentry(alias)->flags & CEPH_DENTRY_PRIMARY_LINK)) {
3657                         dn = dget_dlock(alias);
3658                 }
3659                 spin_unlock(&alias->d_lock);
3660                 if (dn)
3661                         break;
3662         }
3663 out_unlock:
3664         spin_unlock(&inode->i_lock);
3665         return dn;
3666 }
3667
3668 /*
3669  * Encode information about a cap for a reconnect with the MDS.
3670  */
3671 static int reconnect_caps_cb(struct inode *inode, struct ceph_cap *cap,
3672                           void *arg)
3673 {
3674         union {
3675                 struct ceph_mds_cap_reconnect v2;
3676                 struct ceph_mds_cap_reconnect_v1 v1;
3677         } rec;
3678         struct ceph_inode_info *ci = cap->ci;
3679         struct ceph_reconnect_state *recon_state = arg;
3680         struct ceph_pagelist *pagelist = recon_state->pagelist;
3681         struct dentry *dentry;
3682         char *path;
3683         int pathlen = 0, err;
3684         u64 pathbase;
3685         u64 snap_follows;
3686
3687         dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
3688              inode, ceph_vinop(inode), cap, cap->cap_id,
3689              ceph_cap_string(cap->issued));
3690
3691         dentry = d_find_primary(inode);
3692         if (dentry) {
3693                 /* set pathbase to parent dir when msg_version >= 2 */
3694                 path = ceph_mdsc_build_path(dentry, &pathlen, &pathbase,
3695                                             recon_state->msg_version >= 2);
3696                 dput(dentry);
3697                 if (IS_ERR(path)) {
3698                         err = PTR_ERR(path);
3699                         goto out_err;
3700                 }
3701         } else {
3702                 path = NULL;
3703                 pathbase = 0;
3704         }
3705
3706         spin_lock(&ci->i_ceph_lock);
3707         cap->seq = 0;        /* reset cap seq */
3708         cap->issue_seq = 0;  /* and issue_seq */
3709         cap->mseq = 0;       /* and migrate_seq */
3710         cap->cap_gen = atomic_read(&cap->session->s_cap_gen);
3711
3712         /* These are lost when the session goes away */
3713         if (S_ISDIR(inode->i_mode)) {
3714                 if (cap->issued & CEPH_CAP_DIR_CREATE) {
3715                         ceph_put_string(rcu_dereference_raw(ci->i_cached_layout.pool_ns));
3716                         memset(&ci->i_cached_layout, 0, sizeof(ci->i_cached_layout));
3717                 }
3718                 cap->issued &= ~CEPH_CAP_ANY_DIR_OPS;
3719         }
3720
3721         if (recon_state->msg_version >= 2) {
3722                 rec.v2.cap_id = cpu_to_le64(cap->cap_id);
3723                 rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3724                 rec.v2.issued = cpu_to_le32(cap->issued);
3725                 rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3726                 rec.v2.pathbase = cpu_to_le64(pathbase);
3727                 rec.v2.flock_len = (__force __le32)
3728                         ((ci->i_ceph_flags & CEPH_I_ERROR_FILELOCK) ? 0 : 1);
3729         } else {
3730                 rec.v1.cap_id = cpu_to_le64(cap->cap_id);
3731                 rec.v1.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
3732                 rec.v1.issued = cpu_to_le32(cap->issued);
3733                 rec.v1.size = cpu_to_le64(i_size_read(inode));
3734                 ceph_encode_timespec64(&rec.v1.mtime, &inode->i_mtime);
3735                 ceph_encode_timespec64(&rec.v1.atime, &inode->i_atime);
3736                 rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
3737                 rec.v1.pathbase = cpu_to_le64(pathbase);
3738         }
3739
3740         if (list_empty(&ci->i_cap_snaps)) {
3741                 snap_follows = ci->i_head_snapc ? ci->i_head_snapc->seq : 0;
3742         } else {
3743                 struct ceph_cap_snap *capsnap =
3744                         list_first_entry(&ci->i_cap_snaps,
3745                                          struct ceph_cap_snap, ci_item);
3746                 snap_follows = capsnap->follows;
3747         }
3748         spin_unlock(&ci->i_ceph_lock);
3749
3750         if (recon_state->msg_version >= 2) {
3751                 int num_fcntl_locks, num_flock_locks;
3752                 struct ceph_filelock *flocks = NULL;
3753                 size_t struct_len, total_len = sizeof(u64);
3754                 u8 struct_v = 0;
3755
3756 encode_again:
3757                 if (rec.v2.flock_len) {
3758                         ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
3759                 } else {
3760                         num_fcntl_locks = 0;
3761                         num_flock_locks = 0;
3762                 }
3763                 if (num_fcntl_locks + num_flock_locks > 0) {
3764                         flocks = kmalloc_array(num_fcntl_locks + num_flock_locks,
3765                                                sizeof(struct ceph_filelock),
3766                                                GFP_NOFS);
3767                         if (!flocks) {
3768                                 err = -ENOMEM;
3769                                 goto out_err;
3770                         }
3771                         err = ceph_encode_locks_to_buffer(inode, flocks,
3772                                                           num_fcntl_locks,
3773                                                           num_flock_locks);
3774                         if (err) {
3775                                 kfree(flocks);
3776                                 flocks = NULL;
3777                                 if (err == -ENOSPC)
3778                                         goto encode_again;
3779                                 goto out_err;
3780                         }
3781                 } else {
3782                         kfree(flocks);
3783                         flocks = NULL;
3784                 }
3785
3786                 if (recon_state->msg_version >= 3) {
3787                         /* version, compat_version and struct_len */
3788                         total_len += 2 * sizeof(u8) + sizeof(u32);
3789                         struct_v = 2;
3790                 }
3791                 /*
3792                  * number of encoded locks is stable, so copy to pagelist
3793                  */
3794                 struct_len = 2 * sizeof(u32) +
3795                             (num_fcntl_locks + num_flock_locks) *
3796                             sizeof(struct ceph_filelock);
3797                 rec.v2.flock_len = cpu_to_le32(struct_len);
3798
3799                 struct_len += sizeof(u32) + pathlen + sizeof(rec.v2);
3800
3801                 if (struct_v >= 2)
3802                         struct_len += sizeof(u64); /* snap_follows */
3803
3804                 total_len += struct_len;
3805
3806                 if (pagelist->length + total_len > RECONNECT_MAX_SIZE) {
3807                         err = send_reconnect_partial(recon_state);
3808                         if (err)
3809                                 goto out_freeflocks;
3810                         pagelist = recon_state->pagelist;
3811                 }
3812
3813                 err = ceph_pagelist_reserve(pagelist, total_len);
3814                 if (err)
3815                         goto out_freeflocks;
3816
3817                 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
3818                 if (recon_state->msg_version >= 3) {
3819                         ceph_pagelist_encode_8(pagelist, struct_v);
3820                         ceph_pagelist_encode_8(pagelist, 1);
3821                         ceph_pagelist_encode_32(pagelist, struct_len);
3822                 }
3823                 ceph_pagelist_encode_string(pagelist, path, pathlen);
3824                 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v2));
3825                 ceph_locks_to_pagelist(flocks, pagelist,
3826                                        num_fcntl_locks, num_flock_locks);
3827                 if (struct_v >= 2)
3828                         ceph_pagelist_encode_64(pagelist, snap_follows);
3829 out_freeflocks:
3830                 kfree(flocks);
3831         } else {
3832                 err = ceph_pagelist_reserve(pagelist,
3833                                             sizeof(u64) + sizeof(u32) +
3834                                             pathlen + sizeof(rec.v1));
3835                 if (err)
3836                         goto out_err;
3837
3838                 ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
3839                 ceph_pagelist_encode_string(pagelist, path, pathlen);
3840                 ceph_pagelist_append(pagelist, &rec, sizeof(rec.v1));
3841         }
3842
3843 out_err:
3844         ceph_mdsc_free_path(path, pathlen);
3845         if (!err)
3846                 recon_state->nr_caps++;
3847         return err;
3848 }
3849
3850 static int encode_snap_realms(struct ceph_mds_client *mdsc,
3851                               struct ceph_reconnect_state *recon_state)
3852 {
3853         struct rb_node *p;
3854         struct ceph_pagelist *pagelist = recon_state->pagelist;
3855         int err = 0;
3856
3857         if (recon_state->msg_version >= 4) {
3858                 err = ceph_pagelist_encode_32(pagelist, mdsc->num_snap_realms);
3859                 if (err < 0)
3860                         goto fail;
3861         }
3862
3863         /*
3864          * snaprealms.  we provide mds with the ino, seq (version), and
3865          * parent for all of our realms.  If the mds has any newer info,
3866          * it will tell us.
3867          */
3868         for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {
3869                 struct ceph_snap_realm *realm =
3870                        rb_entry(p, struct ceph_snap_realm, node);
3871                 struct ceph_mds_snaprealm_reconnect sr_rec;
3872
3873                 if (recon_state->msg_version >= 4) {
3874                         size_t need = sizeof(u8) * 2 + sizeof(u32) +
3875                                       sizeof(sr_rec);
3876
3877                         if (pagelist->length + need > RECONNECT_MAX_SIZE) {
3878                                 err = send_reconnect_partial(recon_state);
3879                                 if (err)
3880                                         goto fail;
3881                                 pagelist = recon_state->pagelist;
3882                         }
3883
3884                         err = ceph_pagelist_reserve(pagelist, need);
3885                         if (err)
3886                                 goto fail;
3887
3888                         ceph_pagelist_encode_8(pagelist, 1);
3889                         ceph_pagelist_encode_8(pagelist, 1);
3890                         ceph_pagelist_encode_32(pagelist, sizeof(sr_rec));
3891                 }
3892
3893                 dout(" adding snap realm %llx seq %lld parent %llx\n",
3894                      realm->ino, realm->seq, realm->parent_ino);
3895                 sr_rec.ino = cpu_to_le64(realm->ino);
3896                 sr_rec.seq = cpu_to_le64(realm->seq);
3897                 sr_rec.parent = cpu_to_le64(realm->parent_ino);
3898
3899                 err = ceph_pagelist_append(pagelist, &sr_rec, sizeof(sr_rec));
3900                 if (err)
3901                         goto fail;
3902
3903                 recon_state->nr_realms++;
3904         }
3905 fail:
3906         return err;
3907 }
3908
3909
3910 /*
3911  * If an MDS fails and recovers, clients need to reconnect in order to
3912  * reestablish shared state.  This includes all caps issued through
3913  * this session _and_ the snap_realm hierarchy.  Because it's not
3914  * clear which snap realms the mds cares about, we send everything we
3915  * know about.. that ensures we'll then get any new info the
3916  * recovering MDS might have.
3917  *
3918  * This is a relatively heavyweight operation, but it's rare.
3919  */
3920 static void send_mds_reconnect(struct ceph_mds_client *mdsc,
3921                                struct ceph_mds_session *session)
3922 {
3923         struct ceph_msg *reply;
3924         int mds = session->s_mds;
3925         int err = -ENOMEM;
3926         struct ceph_reconnect_state recon_state = {
3927                 .session = session,
3928         };
3929         LIST_HEAD(dispose);
3930
3931         pr_info("mds%d reconnect start\n", mds);
3932
3933         recon_state.pagelist = ceph_pagelist_alloc(GFP_NOFS);
3934         if (!recon_state.pagelist)
3935                 goto fail_nopagelist;
3936
3937         reply = ceph_msg_new2(CEPH_MSG_CLIENT_RECONNECT, 0, 1, GFP_NOFS, false);
3938         if (!reply)
3939                 goto fail_nomsg;
3940
3941         xa_destroy(&session->s_delegated_inos);
3942
3943         mutex_lock(&session->s_mutex);
3944         session->s_state = CEPH_MDS_SESSION_RECONNECTING;
3945         session->s_seq = 0;
3946
3947         dout("session %p state %s\n", session,
3948              ceph_session_state_name(session->s_state));
3949
3950         atomic_inc(&session->s_cap_gen);
3951
3952         spin_lock(&session->s_cap_lock);
3953         /* don't know if session is readonly */
3954         session->s_readonly = 0;
3955         /*
3956          * notify __ceph_remove_cap() that we are composing cap reconnect.
3957          * If a cap get released before being added to the cap reconnect,
3958          * __ceph_remove_cap() should skip queuing cap release.
3959          */
3960         session->s_cap_reconnect = 1;
3961         /* drop old cap expires; we're about to reestablish that state */
3962         detach_cap_releases(session, &dispose);
3963         spin_unlock(&session->s_cap_lock);
3964         dispose_cap_releases(mdsc, &dispose);
3965
3966         /* trim unused caps to reduce MDS's cache rejoin time */
3967         if (mdsc->fsc->sb->s_root)
3968                 shrink_dcache_parent(mdsc->fsc->sb->s_root);
3969
3970         ceph_con_close(&session->s_con);
3971         ceph_con_open(&session->s_con,
3972                       CEPH_ENTITY_TYPE_MDS, mds,
3973                       ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
3974
3975         /* replay unsafe requests */
3976         replay_unsafe_requests(mdsc, session);
3977
3978         ceph_early_kick_flushing_caps(mdsc, session);
3979
3980         down_read(&mdsc->snap_rwsem);
3981
3982         /* placeholder for nr_caps */
3983         err = ceph_pagelist_encode_32(recon_state.pagelist, 0);
3984         if (err)
3985                 goto fail;
3986
3987         if (test_bit(CEPHFS_FEATURE_MULTI_RECONNECT, &session->s_features)) {
3988                 recon_state.msg_version = 3;
3989                 recon_state.allow_multi = true;
3990         } else if (session->s_con.peer_features & CEPH_FEATURE_MDSENC) {
3991                 recon_state.msg_version = 3;
3992         } else {
3993                 recon_state.msg_version = 2;
3994         }
3995         /* trsaverse this session's caps */
3996         err = ceph_iterate_session_caps(session, reconnect_caps_cb, &recon_state);
3997
3998         spin_lock(&session->s_cap_lock);
3999         session->s_cap_reconnect = 0;
4000         spin_unlock(&session->s_cap_lock);
4001
4002         if (err < 0)
4003                 goto fail;
4004
4005         /* check if all realms can be encoded into current message */
4006         if (mdsc->num_snap_realms) {
4007                 size_t total_len =
4008                         recon_state.pagelist->length +
4009                         mdsc->num_snap_realms *
4010                         sizeof(struct ceph_mds_snaprealm_reconnect);
4011                 if (recon_state.msg_version >= 4) {
4012                         /* number of realms */
4013                         total_len += sizeof(u32);
4014                         /* version, compat_version and struct_len */
4015                         total_len += mdsc->num_snap_realms *
4016                                      (2 * sizeof(u8) + sizeof(u32));
4017                 }
4018                 if (total_len > RECONNECT_MAX_SIZE) {
4019                         if (!recon_state.allow_multi) {
4020                                 err = -ENOSPC;
4021                                 goto fail;
4022                         }
4023                         if (recon_state.nr_caps) {
4024                                 err = send_reconnect_partial(&recon_state);
4025                                 if (err)
4026                                         goto fail;
4027                         }
4028                         recon_state.msg_version = 5;
4029                 }
4030         }
4031
4032         err = encode_snap_realms(mdsc, &recon_state);
4033         if (err < 0)
4034                 goto fail;
4035
4036         if (recon_state.msg_version >= 5) {
4037                 err = ceph_pagelist_encode_8(recon_state.pagelist, 0);
4038                 if (err < 0)
4039                         goto fail;
4040         }
4041
4042         if (recon_state.nr_caps || recon_state.nr_realms) {
4043                 struct page *page =
4044                         list_first_entry(&recon_state.pagelist->head,
4045                                         struct page, lru);
4046                 __le32 *addr = kmap_atomic(page);
4047                 if (recon_state.nr_caps) {
4048                         WARN_ON(recon_state.nr_realms != mdsc->num_snap_realms);
4049                         *addr = cpu_to_le32(recon_state.nr_caps);
4050                 } else if (recon_state.msg_version >= 4) {
4051                         *(addr + 1) = cpu_to_le32(recon_state.nr_realms);
4052                 }
4053                 kunmap_atomic(addr);
4054         }
4055
4056         reply->hdr.version = cpu_to_le16(recon_state.msg_version);
4057         if (recon_state.msg_version >= 4)
4058                 reply->hdr.compat_version = cpu_to_le16(4);
4059
4060         reply->hdr.data_len = cpu_to_le32(recon_state.pagelist->length);
4061         ceph_msg_data_add_pagelist(reply, recon_state.pagelist);
4062
4063         ceph_con_send(&session->s_con, reply);
4064
4065         mutex_unlock(&session->s_mutex);
4066
4067         mutex_lock(&mdsc->mutex);
4068         __wake_requests(mdsc, &session->s_waiting);
4069         mutex_unlock(&mdsc->mutex);
4070
4071         up_read(&mdsc->snap_rwsem);
4072         ceph_pagelist_release(recon_state.pagelist);
4073         return;
4074
4075 fail:
4076         ceph_msg_put(reply);
4077         up_read(&mdsc->snap_rwsem);
4078         mutex_unlock(&session->s_mutex);
4079 fail_nomsg:
4080         ceph_pagelist_release(recon_state.pagelist);
4081 fail_nopagelist:
4082         pr_err("error %d preparing reconnect for mds%d\n", err, mds);
4083         return;
4084 }
4085
4086
4087 /*
4088  * compare old and new mdsmaps, kicking requests
4089  * and closing out old connections as necessary
4090  *
4091  * called under mdsc->mutex.
4092  */
4093 static void check_new_map(struct ceph_mds_client *mdsc,
4094                           struct ceph_mdsmap *newmap,
4095                           struct ceph_mdsmap *oldmap)
4096 {
4097         int i, j, err;
4098         int oldstate, newstate;
4099         struct ceph_mds_session *s;
4100         unsigned long targets[DIV_ROUND_UP(CEPH_MAX_MDS, sizeof(unsigned long))] = {0};
4101
4102         dout("check_new_map new %u old %u\n",
4103              newmap->m_epoch, oldmap->m_epoch);
4104
4105         if (newmap->m_info) {
4106                 for (i = 0; i < newmap->possible_max_rank; i++) {
4107                         for (j = 0; j < newmap->m_info[i].num_export_targets; j++)
4108                                 set_bit(newmap->m_info[i].export_targets[j], targets);
4109                 }
4110         }
4111
4112         for (i = 0; i < oldmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4113                 if (!mdsc->sessions[i])
4114                         continue;
4115                 s = mdsc->sessions[i];
4116                 oldstate = ceph_mdsmap_get_state(oldmap, i);
4117                 newstate = ceph_mdsmap_get_state(newmap, i);
4118
4119                 dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",
4120                      i, ceph_mds_state_name(oldstate),
4121                      ceph_mdsmap_is_laggy(oldmap, i) ? " (laggy)" : "",
4122                      ceph_mds_state_name(newstate),
4123                      ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
4124                      ceph_session_state_name(s->s_state));
4125
4126                 if (i >= newmap->possible_max_rank) {
4127                         /* force close session for stopped mds */
4128                         ceph_get_mds_session(s);
4129                         __unregister_session(mdsc, s);
4130                         __wake_requests(mdsc, &s->s_waiting);
4131                         mutex_unlock(&mdsc->mutex);
4132
4133                         mutex_lock(&s->s_mutex);
4134                         cleanup_session_requests(mdsc, s);
4135                         remove_session_caps(s);
4136                         mutex_unlock(&s->s_mutex);
4137
4138                         ceph_put_mds_session(s);
4139
4140                         mutex_lock(&mdsc->mutex);
4141                         kick_requests(mdsc, i);
4142                         continue;
4143                 }
4144
4145                 if (memcmp(ceph_mdsmap_get_addr(oldmap, i),
4146                            ceph_mdsmap_get_addr(newmap, i),
4147                            sizeof(struct ceph_entity_addr))) {
4148                         /* just close it */
4149                         mutex_unlock(&mdsc->mutex);
4150                         mutex_lock(&s->s_mutex);
4151                         mutex_lock(&mdsc->mutex);
4152                         ceph_con_close(&s->s_con);
4153                         mutex_unlock(&s->s_mutex);
4154                         s->s_state = CEPH_MDS_SESSION_RESTARTING;
4155                 } else if (oldstate == newstate) {
4156                         continue;  /* nothing new with this mds */
4157                 }
4158
4159                 /*
4160                  * send reconnect?
4161                  */
4162                 if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&
4163                     newstate >= CEPH_MDS_STATE_RECONNECT) {
4164                         mutex_unlock(&mdsc->mutex);
4165                         clear_bit(i, targets);
4166                         send_mds_reconnect(mdsc, s);
4167                         mutex_lock(&mdsc->mutex);
4168                 }
4169
4170                 /*
4171                  * kick request on any mds that has gone active.
4172                  */
4173                 if (oldstate < CEPH_MDS_STATE_ACTIVE &&
4174                     newstate >= CEPH_MDS_STATE_ACTIVE) {
4175                         if (oldstate != CEPH_MDS_STATE_CREATING &&
4176                             oldstate != CEPH_MDS_STATE_STARTING)
4177                                 pr_info("mds%d recovery completed\n", s->s_mds);
4178                         kick_requests(mdsc, i);
4179                         mutex_unlock(&mdsc->mutex);
4180                         mutex_lock(&s->s_mutex);
4181                         mutex_lock(&mdsc->mutex);
4182                         ceph_kick_flushing_caps(mdsc, s);
4183                         mutex_unlock(&s->s_mutex);
4184                         wake_up_session_caps(s, RECONNECT);
4185                 }
4186         }
4187
4188         /*
4189          * Only open and reconnect sessions that don't exist yet.
4190          */
4191         for (i = 0; i < newmap->possible_max_rank; i++) {
4192                 /*
4193                  * In case the import MDS is crashed just after
4194                  * the EImportStart journal is flushed, so when
4195                  * a standby MDS takes over it and is replaying
4196                  * the EImportStart journal the new MDS daemon
4197                  * will wait the client to reconnect it, but the
4198                  * client may never register/open the session yet.
4199                  *
4200                  * Will try to reconnect that MDS daemon if the
4201                  * rank number is in the export targets array and
4202                  * is the up:reconnect state.
4203                  */
4204                 newstate = ceph_mdsmap_get_state(newmap, i);
4205                 if (!test_bit(i, targets) || newstate != CEPH_MDS_STATE_RECONNECT)
4206                         continue;
4207
4208                 /*
4209                  * The session maybe registered and opened by some
4210                  * requests which were choosing random MDSes during
4211                  * the mdsc->mutex's unlock/lock gap below in rare
4212                  * case. But the related MDS daemon will just queue
4213                  * that requests and be still waiting for the client's
4214                  * reconnection request in up:reconnect state.
4215                  */
4216                 s = __ceph_lookup_mds_session(mdsc, i);
4217                 if (likely(!s)) {
4218                         s = __open_export_target_session(mdsc, i);
4219                         if (IS_ERR(s)) {
4220                                 err = PTR_ERR(s);
4221                                 pr_err("failed to open export target session, err %d\n",
4222                                        err);
4223                                 continue;
4224                         }
4225                 }
4226                 dout("send reconnect to export target mds.%d\n", i);
4227                 mutex_unlock(&mdsc->mutex);
4228                 send_mds_reconnect(mdsc, s);
4229                 ceph_put_mds_session(s);
4230                 mutex_lock(&mdsc->mutex);
4231         }
4232
4233         for (i = 0; i < newmap->possible_max_rank && i < mdsc->max_sessions; i++) {
4234                 s = mdsc->sessions[i];
4235                 if (!s)
4236                         continue;
4237                 if (!ceph_mdsmap_is_laggy(newmap, i))
4238                         continue;
4239                 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4240                     s->s_state == CEPH_MDS_SESSION_HUNG ||
4241                     s->s_state == CEPH_MDS_SESSION_CLOSING) {
4242                         dout(" connecting to export targets of laggy mds%d\n",
4243                              i);
4244                         __open_export_target_sessions(mdsc, s);
4245                 }
4246         }
4247 }
4248
4249
4250
4251 /*
4252  * leases
4253  */
4254
4255 /*
4256  * caller must hold session s_mutex, dentry->d_lock
4257  */
4258 void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry)
4259 {
4260         struct ceph_dentry_info *di = ceph_dentry(dentry);
4261
4262         ceph_put_mds_session(di->lease_session);
4263         di->lease_session = NULL;
4264 }
4265
4266 static void handle_lease(struct ceph_mds_client *mdsc,
4267                          struct ceph_mds_session *session,
4268                          struct ceph_msg *msg)
4269 {
4270         struct super_block *sb = mdsc->fsc->sb;
4271         struct inode *inode;
4272         struct dentry *parent, *dentry;
4273         struct ceph_dentry_info *di;
4274         int mds = session->s_mds;
4275         struct ceph_mds_lease *h = msg->front.iov_base;
4276         u32 seq;
4277         struct ceph_vino vino;
4278         struct qstr dname;
4279         int release = 0;
4280
4281         dout("handle_lease from mds%d\n", mds);
4282
4283         /* decode */
4284         if (msg->front.iov_len < sizeof(*h) + sizeof(u32))
4285                 goto bad;
4286         vino.ino = le64_to_cpu(h->ino);
4287         vino.snap = CEPH_NOSNAP;
4288         seq = le32_to_cpu(h->seq);
4289         dname.len = get_unaligned_le32(h + 1);
4290         if (msg->front.iov_len < sizeof(*h) + sizeof(u32) + dname.len)
4291                 goto bad;
4292         dname.name = (void *)(h + 1) + sizeof(u32);
4293
4294         /* lookup inode */
4295         inode = ceph_find_inode(sb, vino);
4296         dout("handle_lease %s, ino %llx %p %.*s\n",
4297              ceph_lease_op_name(h->action), vino.ino, inode,
4298              dname.len, dname.name);
4299
4300         mutex_lock(&session->s_mutex);
4301         inc_session_sequence(session);
4302
4303         if (!inode) {
4304                 dout("handle_lease no inode %llx\n", vino.ino);
4305                 goto release;
4306         }
4307
4308         /* dentry */
4309         parent = d_find_alias(inode);
4310         if (!parent) {
4311                 dout("no parent dentry on inode %p\n", inode);
4312                 WARN_ON(1);
4313                 goto release;  /* hrm... */
4314         }
4315         dname.hash = full_name_hash(parent, dname.name, dname.len);
4316         dentry = d_lookup(parent, &dname);
4317         dput(parent);
4318         if (!dentry)
4319                 goto release;
4320
4321         spin_lock(&dentry->d_lock);
4322         di = ceph_dentry(dentry);
4323         switch (h->action) {
4324         case CEPH_MDS_LEASE_REVOKE:
4325                 if (di->lease_session == session) {
4326                         if (ceph_seq_cmp(di->lease_seq, seq) > 0)
4327                                 h->seq = cpu_to_le32(di->lease_seq);
4328                         __ceph_mdsc_drop_dentry_lease(dentry);
4329                 }
4330                 release = 1;
4331                 break;
4332
4333         case CEPH_MDS_LEASE_RENEW:
4334                 if (di->lease_session == session &&
4335                     di->lease_gen == atomic_read(&session->s_cap_gen) &&
4336                     di->lease_renew_from &&
4337                     di->lease_renew_after == 0) {
4338                         unsigned long duration =
4339                                 msecs_to_jiffies(le32_to_cpu(h->duration_ms));
4340
4341                         di->lease_seq = seq;
4342                         di->time = di->lease_renew_from + duration;
4343                         di->lease_renew_after = di->lease_renew_from +
4344                                 (duration >> 1);
4345                         di->lease_renew_from = 0;
4346                 }
4347                 break;
4348         }
4349         spin_unlock(&dentry->d_lock);
4350         dput(dentry);
4351
4352         if (!release)
4353                 goto out;
4354
4355 release:
4356         /* let's just reuse the same message */
4357         h->action = CEPH_MDS_LEASE_REVOKE_ACK;
4358         ceph_msg_get(msg);
4359         ceph_con_send(&session->s_con, msg);
4360
4361 out:
4362         mutex_unlock(&session->s_mutex);
4363         iput(inode);
4364         return;
4365
4366 bad:
4367         pr_err("corrupt lease message\n");
4368         ceph_msg_dump(msg);
4369 }
4370
4371 void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
4372                               struct dentry *dentry, char action,
4373                               u32 seq)
4374 {
4375         struct ceph_msg *msg;
4376         struct ceph_mds_lease *lease;
4377         struct inode *dir;
4378         int len = sizeof(*lease) + sizeof(u32) + NAME_MAX;
4379
4380         dout("lease_send_msg identry %p %s to mds%d\n",
4381              dentry, ceph_lease_op_name(action), session->s_mds);
4382
4383         msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
4384         if (!msg)
4385                 return;
4386         lease = msg->front.iov_base;
4387         lease->action = action;
4388         lease->seq = cpu_to_le32(seq);
4389
4390         spin_lock(&dentry->d_lock);
4391         dir = d_inode(dentry->d_parent);
4392         lease->ino = cpu_to_le64(ceph_ino(dir));
4393         lease->first = lease->last = cpu_to_le64(ceph_snap(dir));
4394
4395         put_unaligned_le32(dentry->d_name.len, lease + 1);
4396         memcpy((void *)(lease + 1) + 4,
4397                dentry->d_name.name, dentry->d_name.len);
4398         spin_unlock(&dentry->d_lock);
4399         /*
4400          * if this is a preemptive lease RELEASE, no need to
4401          * flush request stream, since the actual request will
4402          * soon follow.
4403          */
4404         msg->more_to_follow = (action == CEPH_MDS_LEASE_RELEASE);
4405
4406         ceph_con_send(&session->s_con, msg);
4407 }
4408
4409 /*
4410  * lock unlock the session, to wait ongoing session activities
4411  */
4412 static void lock_unlock_session(struct ceph_mds_session *s)
4413 {
4414         mutex_lock(&s->s_mutex);
4415         mutex_unlock(&s->s_mutex);
4416 }
4417
4418 static void maybe_recover_session(struct ceph_mds_client *mdsc)
4419 {
4420         struct ceph_fs_client *fsc = mdsc->fsc;
4421
4422         if (!ceph_test_mount_opt(fsc, CLEANRECOVER))
4423                 return;
4424
4425         if (READ_ONCE(fsc->mount_state) != CEPH_MOUNT_MOUNTED)
4426                 return;
4427
4428         if (!READ_ONCE(fsc->blocklisted))
4429                 return;
4430
4431         pr_info("auto reconnect after blocklisted\n");
4432         ceph_force_reconnect(fsc->sb);
4433 }
4434
4435 bool check_session_state(struct ceph_mds_session *s)
4436 {
4437         struct ceph_fs_client *fsc = s->s_mdsc->fsc;
4438
4439         switch (s->s_state) {
4440         case CEPH_MDS_SESSION_OPEN:
4441                 if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
4442                         s->s_state = CEPH_MDS_SESSION_HUNG;
4443                         pr_info("mds%d hung\n", s->s_mds);
4444                 }
4445                 break;
4446         case CEPH_MDS_SESSION_CLOSING:
4447                 /* Should never reach this when not force unmounting */
4448                 WARN_ON_ONCE(s->s_ttl &&
4449                              READ_ONCE(fsc->mount_state) != CEPH_MOUNT_SHUTDOWN);
4450                 fallthrough;
4451         case CEPH_MDS_SESSION_NEW:
4452         case CEPH_MDS_SESSION_RESTARTING:
4453         case CEPH_MDS_SESSION_CLOSED:
4454         case CEPH_MDS_SESSION_REJECTED:
4455                 return false;
4456         }
4457
4458         return true;
4459 }
4460
4461 /*
4462  * If the sequence is incremented while we're waiting on a REQUEST_CLOSE reply,
4463  * then we need to retransmit that request.
4464  */
4465 void inc_session_sequence(struct ceph_mds_session *s)
4466 {
4467         lockdep_assert_held(&s->s_mutex);
4468
4469         s->s_seq++;
4470
4471         if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
4472                 int ret;
4473
4474                 dout("resending session close request for mds%d\n", s->s_mds);
4475                 ret = request_close_session(s);
4476                 if (ret < 0)
4477                         pr_err("unable to close session to mds%d: %d\n",
4478                                s->s_mds, ret);
4479         }
4480 }
4481
4482 /*
4483  * delayed work -- periodically trim expired leases, renew caps with mds.  If
4484  * the @delay parameter is set to 0 or if it's more than 5 secs, the default
4485  * workqueue delay value of 5 secs will be used.
4486  */
4487 static void schedule_delayed(struct ceph_mds_client *mdsc, unsigned long delay)
4488 {
4489         unsigned long max_delay = HZ * 5;
4490
4491         /* 5 secs default delay */
4492         if (!delay || (delay > max_delay))
4493                 delay = max_delay;
4494         schedule_delayed_work(&mdsc->delayed_work,
4495                               round_jiffies_relative(delay));
4496 }
4497
4498 static void delayed_work(struct work_struct *work)
4499 {
4500         struct ceph_mds_client *mdsc =
4501                 container_of(work, struct ceph_mds_client, delayed_work.work);
4502         unsigned long delay;
4503         int renew_interval;
4504         int renew_caps;
4505         int i;
4506
4507         dout("mdsc delayed_work\n");
4508
4509         if (mdsc->stopping)
4510                 return;
4511
4512         mutex_lock(&mdsc->mutex);
4513         renew_interval = mdsc->mdsmap->m_session_timeout >> 2;
4514         renew_caps = time_after_eq(jiffies, HZ*renew_interval +
4515                                    mdsc->last_renew_caps);
4516         if (renew_caps)
4517                 mdsc->last_renew_caps = jiffies;
4518
4519         for (i = 0; i < mdsc->max_sessions; i++) {
4520                 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
4521                 if (!s)
4522                         continue;
4523
4524                 if (!check_session_state(s)) {
4525                         ceph_put_mds_session(s);
4526                         continue;
4527                 }
4528                 mutex_unlock(&mdsc->mutex);
4529
4530                 mutex_lock(&s->s_mutex);
4531                 if (renew_caps)
4532                         send_renew_caps(mdsc, s);
4533                 else
4534                         ceph_con_keepalive(&s->s_con);
4535                 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
4536                     s->s_state == CEPH_MDS_SESSION_HUNG)
4537                         ceph_send_cap_releases(mdsc, s);
4538                 mutex_unlock(&s->s_mutex);
4539                 ceph_put_mds_session(s);
4540
4541                 mutex_lock(&mdsc->mutex);
4542         }
4543         mutex_unlock(&mdsc->mutex);
4544
4545         delay = ceph_check_delayed_caps(mdsc);
4546
4547         ceph_queue_cap_reclaim_work(mdsc);
4548
4549         ceph_trim_snapid_map(mdsc);
4550
4551         maybe_recover_session(mdsc);
4552
4553         schedule_delayed(mdsc, delay);
4554 }
4555
4556 int ceph_mdsc_init(struct ceph_fs_client *fsc)
4557
4558 {
4559         struct ceph_mds_client *mdsc;
4560         int err;
4561
4562         mdsc = kzalloc(sizeof(struct ceph_mds_client), GFP_NOFS);
4563         if (!mdsc)
4564                 return -ENOMEM;
4565         mdsc->fsc = fsc;
4566         mutex_init(&mdsc->mutex);
4567         mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);
4568         if (!mdsc->mdsmap) {
4569                 err = -ENOMEM;
4570                 goto err_mdsc;
4571         }
4572
4573         init_completion(&mdsc->safe_umount_waiters);
4574         init_waitqueue_head(&mdsc->session_close_wq);
4575         INIT_LIST_HEAD(&mdsc->waiting_for_map);
4576         mdsc->quotarealms_inodes = RB_ROOT;
4577         mutex_init(&mdsc->quotarealms_inodes_mutex);
4578         init_rwsem(&mdsc->snap_rwsem);
4579         mdsc->snap_realms = RB_ROOT;
4580         INIT_LIST_HEAD(&mdsc->snap_empty);
4581         spin_lock_init(&mdsc->snap_empty_lock);
4582         mdsc->request_tree = RB_ROOT;
4583         INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);
4584         mdsc->last_renew_caps = jiffies;
4585         INIT_LIST_HEAD(&mdsc->cap_delay_list);
4586         INIT_LIST_HEAD(&mdsc->cap_wait_list);
4587         spin_lock_init(&mdsc->cap_delay_lock);
4588         INIT_LIST_HEAD(&mdsc->snap_flush_list);
4589         spin_lock_init(&mdsc->snap_flush_lock);
4590         mdsc->last_cap_flush_tid = 1;
4591         INIT_LIST_HEAD(&mdsc->cap_flush_list);
4592         INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
4593         spin_lock_init(&mdsc->cap_dirty_lock);
4594         init_waitqueue_head(&mdsc->cap_flushing_wq);
4595         INIT_WORK(&mdsc->cap_reclaim_work, ceph_cap_reclaim_work);
4596         err = ceph_metric_init(&mdsc->metric);
4597         if (err)
4598                 goto err_mdsmap;
4599
4600         spin_lock_init(&mdsc->dentry_list_lock);
4601         INIT_LIST_HEAD(&mdsc->dentry_leases);
4602         INIT_LIST_HEAD(&mdsc->dentry_dir_leases);
4603
4604         ceph_caps_init(mdsc);
4605         ceph_adjust_caps_max_min(mdsc, fsc->mount_options);
4606
4607         spin_lock_init(&mdsc->snapid_map_lock);
4608         mdsc->snapid_map_tree = RB_ROOT;
4609         INIT_LIST_HEAD(&mdsc->snapid_map_lru);
4610
4611         init_rwsem(&mdsc->pool_perm_rwsem);
4612         mdsc->pool_perm_tree = RB_ROOT;
4613
4614         strscpy(mdsc->nodename, utsname()->nodename,
4615                 sizeof(mdsc->nodename));
4616
4617         fsc->mdsc = mdsc;
4618         return 0;
4619
4620 err_mdsmap:
4621         kfree(mdsc->mdsmap);
4622 err_mdsc:
4623         kfree(mdsc);
4624         return err;
4625 }
4626
4627 /*
4628  * Wait for safe replies on open mds requests.  If we time out, drop
4629  * all requests from the tree to avoid dangling dentry refs.
4630  */
4631 static void wait_requests(struct ceph_mds_client *mdsc)
4632 {
4633         struct ceph_options *opts = mdsc->fsc->client->options;
4634         struct ceph_mds_request *req;
4635
4636         mutex_lock(&mdsc->mutex);
4637         if (__get_oldest_req(mdsc)) {
4638                 mutex_unlock(&mdsc->mutex);
4639
4640                 dout("wait_requests waiting for requests\n");
4641                 wait_for_completion_timeout(&mdsc->safe_umount_waiters,
4642                                     ceph_timeout_jiffies(opts->mount_timeout));
4643
4644                 /* tear down remaining requests */
4645                 mutex_lock(&mdsc->mutex);
4646                 while ((req = __get_oldest_req(mdsc))) {
4647                         dout("wait_requests timed out on tid %llu\n",
4648                              req->r_tid);
4649                         list_del_init(&req->r_wait);
4650                         __unregister_request(mdsc, req);
4651                 }
4652         }
4653         mutex_unlock(&mdsc->mutex);
4654         dout("wait_requests done\n");
4655 }
4656
4657 void send_flush_mdlog(struct ceph_mds_session *s)
4658 {
4659         struct ceph_msg *msg;
4660
4661         /*
4662          * Pre-luminous MDS crashes when it sees an unknown session request
4663          */
4664         if (!CEPH_HAVE_FEATURE(s->s_con.peer_features, SERVER_LUMINOUS))
4665                 return;
4666
4667         mutex_lock(&s->s_mutex);
4668         dout("request mdlog flush to mds%d (%s)s seq %lld\n", s->s_mds,
4669              ceph_session_state_name(s->s_state), s->s_seq);
4670         msg = ceph_create_session_msg(CEPH_SESSION_REQUEST_FLUSH_MDLOG,
4671                                       s->s_seq);
4672         if (!msg) {
4673                 pr_err("failed to request mdlog flush to mds%d (%s) seq %lld\n",
4674                        s->s_mds, ceph_session_state_name(s->s_state), s->s_seq);
4675         } else {
4676                 ceph_con_send(&s->s_con, msg);
4677         }
4678         mutex_unlock(&s->s_mutex);
4679 }
4680
4681 /*
4682  * called before mount is ro, and before dentries are torn down.
4683  * (hmm, does this still race with new lookups?)
4684  */
4685 void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
4686 {
4687         dout("pre_umount\n");
4688         mdsc->stopping = 1;
4689
4690         ceph_mdsc_iterate_sessions(mdsc, send_flush_mdlog, true);
4691         ceph_mdsc_iterate_sessions(mdsc, lock_unlock_session, false);
4692         ceph_flush_dirty_caps(mdsc);
4693         wait_requests(mdsc);
4694
4695         /*
4696          * wait for reply handlers to drop their request refs and
4697          * their inode/dcache refs
4698          */
4699         ceph_msgr_flush();
4700
4701         ceph_cleanup_quotarealms_inodes(mdsc);
4702 }
4703
4704 /*
4705  * wait for all write mds requests to flush.
4706  */
4707 static void wait_unsafe_requests(struct ceph_mds_client *mdsc, u64 want_tid)
4708 {
4709         struct ceph_mds_request *req = NULL, *nextreq;
4710         struct rb_node *n;
4711
4712         mutex_lock(&mdsc->mutex);
4713         dout("wait_unsafe_requests want %lld\n", want_tid);
4714 restart:
4715         req = __get_oldest_req(mdsc);
4716         while (req && req->r_tid <= want_tid) {
4717                 /* find next request */
4718                 n = rb_next(&req->r_node);
4719                 if (n)
4720                         nextreq = rb_entry(n, struct ceph_mds_request, r_node);
4721                 else
4722                         nextreq = NULL;
4723                 if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&
4724                     (req->r_op & CEPH_MDS_OP_WRITE)) {
4725                         /* write op */
4726                         ceph_mdsc_get_request(req);
4727                         if (nextreq)
4728                                 ceph_mdsc_get_request(nextreq);
4729                         mutex_unlock(&mdsc->mutex);
4730                         dout("wait_unsafe_requests  wait on %llu (want %llu)\n",
4731                              req->r_tid, want_tid);
4732                         wait_for_completion(&req->r_safe_completion);
4733                         mutex_lock(&mdsc->mutex);
4734                         ceph_mdsc_put_request(req);
4735                         if (!nextreq)
4736                                 break;  /* next dne before, so we're done! */
4737                         if (RB_EMPTY_NODE(&nextreq->r_node)) {
4738                                 /* next request was removed from tree */
4739                                 ceph_mdsc_put_request(nextreq);
4740                                 goto restart;
4741                         }
4742                         ceph_mdsc_put_request(nextreq);  /* won't go away */
4743                 }
4744                 req = nextreq;
4745         }
4746         mutex_unlock(&mdsc->mutex);
4747         dout("wait_unsafe_requests done\n");
4748 }
4749
4750 void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
4751 {
4752         u64 want_tid, want_flush;
4753
4754         if (READ_ONCE(mdsc->fsc->mount_state) >= CEPH_MOUNT_SHUTDOWN)
4755                 return;
4756
4757         dout("sync\n");
4758         mutex_lock(&mdsc->mutex);
4759         want_tid = mdsc->last_tid;
4760         mutex_unlock(&mdsc->mutex);
4761
4762         ceph_flush_dirty_caps(mdsc);
4763         spin_lock(&mdsc->cap_dirty_lock);
4764         want_flush = mdsc->last_cap_flush_tid;
4765         if (!list_empty(&mdsc->cap_flush_list)) {
4766                 struct ceph_cap_flush *cf =
4767                         list_last_entry(&mdsc->cap_flush_list,
4768                                         struct ceph_cap_flush, g_list);
4769                 cf->wake = true;
4770         }
4771         spin_unlock(&mdsc->cap_dirty_lock);
4772
4773         dout("sync want tid %lld flush_seq %lld\n",
4774              want_tid, want_flush);
4775
4776         wait_unsafe_requests(mdsc, want_tid);
4777         wait_caps_flush(mdsc, want_flush);
4778 }
4779
4780 /*
4781  * true if all sessions are closed, or we force unmount
4782  */
4783 static bool done_closing_sessions(struct ceph_mds_client *mdsc, int skipped)
4784 {
4785         if (READ_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
4786                 return true;
4787         return atomic_read(&mdsc->num_sessions) <= skipped;
4788 }
4789
4790 /*
4791  * called after sb is ro.
4792  */
4793 void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
4794 {
4795         struct ceph_options *opts = mdsc->fsc->client->options;
4796         struct ceph_mds_session *session;
4797         int i;
4798         int skipped = 0;
4799
4800         dout("close_sessions\n");
4801
4802         /* close sessions */
4803         mutex_lock(&mdsc->mutex);
4804         for (i = 0; i < mdsc->max_sessions; i++) {
4805                 session = __ceph_lookup_mds_session(mdsc, i);
4806                 if (!session)
4807                         continue;
4808                 mutex_unlock(&mdsc->mutex);
4809                 mutex_lock(&session->s_mutex);
4810                 if (__close_session(mdsc, session) <= 0)
4811                         skipped++;
4812                 mutex_unlock(&session->s_mutex);
4813                 ceph_put_mds_session(session);
4814                 mutex_lock(&mdsc->mutex);
4815         }
4816         mutex_unlock(&mdsc->mutex);
4817
4818         dout("waiting for sessions to close\n");
4819         wait_event_timeout(mdsc->session_close_wq,
4820                            done_closing_sessions(mdsc, skipped),
4821                            ceph_timeout_jiffies(opts->mount_timeout));
4822
4823         /* tear down remaining sessions */
4824         mutex_lock(&mdsc->mutex);
4825         for (i = 0; i < mdsc->max_sessions; i++) {
4826                 if (mdsc->sessions[i]) {
4827                         session = ceph_get_mds_session(mdsc->sessions[i]);
4828                         __unregister_session(mdsc, session);
4829                         mutex_unlock(&mdsc->mutex);
4830                         mutex_lock(&session->s_mutex);
4831                         remove_session_caps(session);
4832                         mutex_unlock(&session->s_mutex);
4833                         ceph_put_mds_session(session);
4834                         mutex_lock(&mdsc->mutex);
4835                 }
4836         }
4837         WARN_ON(!list_empty(&mdsc->cap_delay_list));
4838         mutex_unlock(&mdsc->mutex);
4839
4840         ceph_cleanup_snapid_map(mdsc);
4841         ceph_cleanup_global_and_empty_realms(mdsc);
4842
4843         cancel_work_sync(&mdsc->cap_reclaim_work);
4844         cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
4845
4846         dout("stopped\n");
4847 }
4848
4849 void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc)
4850 {
4851         struct ceph_mds_session *session;
4852         int mds;
4853
4854         dout("force umount\n");
4855
4856         mutex_lock(&mdsc->mutex);
4857         for (mds = 0; mds < mdsc->max_sessions; mds++) {
4858                 session = __ceph_lookup_mds_session(mdsc, mds);
4859                 if (!session)
4860                         continue;
4861
4862                 if (session->s_state == CEPH_MDS_SESSION_REJECTED)
4863                         __unregister_session(mdsc, session);
4864                 __wake_requests(mdsc, &session->s_waiting);
4865                 mutex_unlock(&mdsc->mutex);
4866
4867                 mutex_lock(&session->s_mutex);
4868                 __close_session(mdsc, session);
4869                 if (session->s_state == CEPH_MDS_SESSION_CLOSING) {
4870                         cleanup_session_requests(mdsc, session);
4871                         remove_session_caps(session);
4872                 }
4873                 mutex_unlock(&session->s_mutex);
4874                 ceph_put_mds_session(session);
4875
4876                 mutex_lock(&mdsc->mutex);
4877                 kick_requests(mdsc, mds);
4878         }
4879         __wake_requests(mdsc, &mdsc->waiting_for_map);
4880         mutex_unlock(&mdsc->mutex);
4881 }
4882
4883 static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
4884 {
4885         dout("stop\n");
4886         /*
4887          * Make sure the delayed work stopped before releasing
4888          * the resources.
4889          *
4890          * Because the cancel_delayed_work_sync() will only
4891          * guarantee that the work finishes executing. But the
4892          * delayed work will re-arm itself again after that.
4893          */
4894         flush_delayed_work(&mdsc->delayed_work);
4895
4896         if (mdsc->mdsmap)
4897                 ceph_mdsmap_destroy(mdsc->mdsmap);
4898         kfree(mdsc->sessions);
4899         ceph_caps_finalize(mdsc);
4900         ceph_pool_perm_destroy(mdsc);
4901 }
4902
4903 void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
4904 {
4905         struct ceph_mds_client *mdsc = fsc->mdsc;
4906         dout("mdsc_destroy %p\n", mdsc);
4907
4908         if (!mdsc)
4909                 return;
4910
4911         /* flush out any connection work with references to us */
4912         ceph_msgr_flush();
4913
4914         ceph_mdsc_stop(mdsc);
4915
4916         ceph_metric_destroy(&mdsc->metric);
4917
4918         fsc->mdsc = NULL;
4919         kfree(mdsc);
4920         dout("mdsc_destroy %p done\n", mdsc);
4921 }
4922
4923 void ceph_mdsc_handle_fsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4924 {
4925         struct ceph_fs_client *fsc = mdsc->fsc;
4926         const char *mds_namespace = fsc->mount_options->mds_namespace;
4927         void *p = msg->front.iov_base;
4928         void *end = p + msg->front.iov_len;
4929         u32 epoch;
4930         u32 num_fs;
4931         u32 mount_fscid = (u32)-1;
4932         int err = -EINVAL;
4933
4934         ceph_decode_need(&p, end, sizeof(u32), bad);
4935         epoch = ceph_decode_32(&p);
4936
4937         dout("handle_fsmap epoch %u\n", epoch);
4938
4939         /* struct_v, struct_cv, map_len, epoch, legacy_client_fscid */
4940         ceph_decode_skip_n(&p, end, 2 + sizeof(u32) * 3, bad);
4941
4942         ceph_decode_32_safe(&p, end, num_fs, bad);
4943         while (num_fs-- > 0) {
4944                 void *info_p, *info_end;
4945                 u32 info_len;
4946                 u32 fscid, namelen;
4947
4948                 ceph_decode_need(&p, end, 2 + sizeof(u32), bad);
4949                 p += 2;         // info_v, info_cv
4950                 info_len = ceph_decode_32(&p);
4951                 ceph_decode_need(&p, end, info_len, bad);
4952                 info_p = p;
4953                 info_end = p + info_len;
4954                 p = info_end;
4955
4956                 ceph_decode_need(&info_p, info_end, sizeof(u32) * 2, bad);
4957                 fscid = ceph_decode_32(&info_p);
4958                 namelen = ceph_decode_32(&info_p);
4959                 ceph_decode_need(&info_p, info_end, namelen, bad);
4960
4961                 if (mds_namespace &&
4962                     strlen(mds_namespace) == namelen &&
4963                     !strncmp(mds_namespace, (char *)info_p, namelen)) {
4964                         mount_fscid = fscid;
4965                         break;
4966                 }
4967         }
4968
4969         ceph_monc_got_map(&fsc->client->monc, CEPH_SUB_FSMAP, epoch);
4970         if (mount_fscid != (u32)-1) {
4971                 fsc->client->monc.fs_cluster_id = mount_fscid;
4972                 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
4973                                    0, true);
4974                 ceph_monc_renew_subs(&fsc->client->monc);
4975         } else {
4976                 err = -ENOENT;
4977                 goto err_out;
4978         }
4979         return;
4980
4981 bad:
4982         pr_err("error decoding fsmap %d. Shutting down mount.\n", err);
4983         ceph_umount_begin(mdsc->fsc->sb);
4984 err_out:
4985         mutex_lock(&mdsc->mutex);
4986         mdsc->mdsmap_err = err;
4987         __wake_requests(mdsc, &mdsc->waiting_for_map);
4988         mutex_unlock(&mdsc->mutex);
4989 }
4990
4991 /*
4992  * handle mds map update.
4993  */
4994 void ceph_mdsc_handle_mdsmap(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
4995 {
4996         u32 epoch;
4997         u32 maplen;
4998         void *p = msg->front.iov_base;
4999         void *end = p + msg->front.iov_len;
5000         struct ceph_mdsmap *newmap, *oldmap;
5001         struct ceph_fsid fsid;
5002         int err = -EINVAL;
5003
5004         ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
5005         ceph_decode_copy(&p, &fsid, sizeof(fsid));
5006         if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)
5007                 return;
5008         epoch = ceph_decode_32(&p);
5009         maplen = ceph_decode_32(&p);
5010         dout("handle_map epoch %u len %d\n", epoch, (int)maplen);
5011
5012         /* do we need it? */
5013         mutex_lock(&mdsc->mutex);
5014         if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {
5015                 dout("handle_map epoch %u <= our %u\n",
5016                      epoch, mdsc->mdsmap->m_epoch);
5017                 mutex_unlock(&mdsc->mutex);
5018                 return;
5019         }
5020
5021         newmap = ceph_mdsmap_decode(&p, end, ceph_msgr2(mdsc->fsc->client));
5022         if (IS_ERR(newmap)) {
5023                 err = PTR_ERR(newmap);
5024                 goto bad_unlock;
5025         }
5026
5027         /* swap into place */
5028         if (mdsc->mdsmap) {
5029                 oldmap = mdsc->mdsmap;
5030                 mdsc->mdsmap = newmap;
5031                 check_new_map(mdsc, newmap, oldmap);
5032                 ceph_mdsmap_destroy(oldmap);
5033         } else {
5034                 mdsc->mdsmap = newmap;  /* first mds map */
5035         }
5036         mdsc->fsc->max_file_size = min((loff_t)mdsc->mdsmap->m_max_file_size,
5037                                         MAX_LFS_FILESIZE);
5038
5039         __wake_requests(mdsc, &mdsc->waiting_for_map);
5040         ceph_monc_got_map(&mdsc->fsc->client->monc, CEPH_SUB_MDSMAP,
5041                           mdsc->mdsmap->m_epoch);
5042
5043         mutex_unlock(&mdsc->mutex);
5044         schedule_delayed(mdsc, 0);
5045         return;
5046
5047 bad_unlock:
5048         mutex_unlock(&mdsc->mutex);
5049 bad:
5050         pr_err("error decoding mdsmap %d. Shutting down mount.\n", err);
5051         ceph_umount_begin(mdsc->fsc->sb);
5052         return;
5053 }
5054
5055 static struct ceph_connection *mds_get_con(struct ceph_connection *con)
5056 {
5057         struct ceph_mds_session *s = con->private;
5058
5059         if (ceph_get_mds_session(s))
5060                 return con;
5061         return NULL;
5062 }
5063
5064 static void mds_put_con(struct ceph_connection *con)
5065 {
5066         struct ceph_mds_session *s = con->private;
5067
5068         ceph_put_mds_session(s);
5069 }
5070
5071 /*
5072  * if the client is unresponsive for long enough, the mds will kill
5073  * the session entirely.
5074  */
5075 static void mds_peer_reset(struct ceph_connection *con)
5076 {
5077         struct ceph_mds_session *s = con->private;
5078         struct ceph_mds_client *mdsc = s->s_mdsc;
5079
5080         pr_warn("mds%d closed our session\n", s->s_mds);
5081         send_mds_reconnect(mdsc, s);
5082 }
5083
5084 static void mds_dispatch(struct ceph_connection *con, struct ceph_msg *msg)
5085 {
5086         struct ceph_mds_session *s = con->private;
5087         struct ceph_mds_client *mdsc = s->s_mdsc;
5088         int type = le16_to_cpu(msg->hdr.type);
5089
5090         mutex_lock(&mdsc->mutex);
5091         if (__verify_registered_session(mdsc, s) < 0) {
5092                 mutex_unlock(&mdsc->mutex);
5093                 goto out;
5094         }
5095         mutex_unlock(&mdsc->mutex);
5096
5097         switch (type) {
5098         case CEPH_MSG_MDS_MAP:
5099                 ceph_mdsc_handle_mdsmap(mdsc, msg);
5100                 break;
5101         case CEPH_MSG_FS_MAP_USER:
5102                 ceph_mdsc_handle_fsmap(mdsc, msg);
5103                 break;
5104         case CEPH_MSG_CLIENT_SESSION:
5105                 handle_session(s, msg);
5106                 break;
5107         case CEPH_MSG_CLIENT_REPLY:
5108                 handle_reply(s, msg);
5109                 break;
5110         case CEPH_MSG_CLIENT_REQUEST_FORWARD:
5111                 handle_forward(mdsc, s, msg);
5112                 break;
5113         case CEPH_MSG_CLIENT_CAPS:
5114                 ceph_handle_caps(s, msg);
5115                 break;
5116         case CEPH_MSG_CLIENT_SNAP:
5117                 ceph_handle_snap(mdsc, s, msg);
5118                 break;
5119         case CEPH_MSG_CLIENT_LEASE:
5120                 handle_lease(mdsc, s, msg);
5121                 break;
5122         case CEPH_MSG_CLIENT_QUOTA:
5123                 ceph_handle_quota(mdsc, s, msg);
5124                 break;
5125
5126         default:
5127                 pr_err("received unknown message type %d %s\n", type,
5128                        ceph_msg_type_name(type));
5129         }
5130 out:
5131         ceph_msg_put(msg);
5132 }
5133
5134 /*
5135  * authentication
5136  */
5137
5138 /*
5139  * Note: returned pointer is the address of a structure that's
5140  * managed separately.  Caller must *not* attempt to free it.
5141  */
5142 static struct ceph_auth_handshake *
5143 mds_get_authorizer(struct ceph_connection *con, int *proto, int force_new)
5144 {
5145         struct ceph_mds_session *s = con->private;
5146         struct ceph_mds_client *mdsc = s->s_mdsc;
5147         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5148         struct ceph_auth_handshake *auth = &s->s_auth;
5149         int ret;
5150
5151         ret = __ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_MDS,
5152                                          force_new, proto, NULL, NULL);
5153         if (ret)
5154                 return ERR_PTR(ret);
5155
5156         return auth;
5157 }
5158
5159 static int mds_add_authorizer_challenge(struct ceph_connection *con,
5160                                     void *challenge_buf, int challenge_buf_len)
5161 {
5162         struct ceph_mds_session *s = con->private;
5163         struct ceph_mds_client *mdsc = s->s_mdsc;
5164         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5165
5166         return ceph_auth_add_authorizer_challenge(ac, s->s_auth.authorizer,
5167                                             challenge_buf, challenge_buf_len);
5168 }
5169
5170 static int mds_verify_authorizer_reply(struct ceph_connection *con)
5171 {
5172         struct ceph_mds_session *s = con->private;
5173         struct ceph_mds_client *mdsc = s->s_mdsc;
5174         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5175         struct ceph_auth_handshake *auth = &s->s_auth;
5176
5177         return ceph_auth_verify_authorizer_reply(ac, auth->authorizer,
5178                 auth->authorizer_reply_buf, auth->authorizer_reply_buf_len,
5179                 NULL, NULL, NULL, NULL);
5180 }
5181
5182 static int mds_invalidate_authorizer(struct ceph_connection *con)
5183 {
5184         struct ceph_mds_session *s = con->private;
5185         struct ceph_mds_client *mdsc = s->s_mdsc;
5186         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
5187
5188         ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_MDS);
5189
5190         return ceph_monc_validate_auth(&mdsc->fsc->client->monc);
5191 }
5192
5193 static int mds_get_auth_request(struct ceph_connection *con,
5194                                 void *buf, int *buf_len,
5195                                 void **authorizer, int *authorizer_len)
5196 {
5197         struct ceph_mds_session *s = con->private;
5198         struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
5199         struct ceph_auth_handshake *auth = &s->s_auth;
5200         int ret;
5201
5202         ret = ceph_auth_get_authorizer(ac, auth, CEPH_ENTITY_TYPE_MDS,
5203                                        buf, buf_len);
5204         if (ret)
5205                 return ret;
5206
5207         *authorizer = auth->authorizer_buf;
5208         *authorizer_len = auth->authorizer_buf_len;
5209         return 0;
5210 }
5211
5212 static int mds_handle_auth_reply_more(struct ceph_connection *con,
5213                                       void *reply, int reply_len,
5214                                       void *buf, int *buf_len,
5215                                       void **authorizer, int *authorizer_len)
5216 {
5217         struct ceph_mds_session *s = con->private;
5218         struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
5219         struct ceph_auth_handshake *auth = &s->s_auth;
5220         int ret;
5221
5222         ret = ceph_auth_handle_svc_reply_more(ac, auth, reply, reply_len,
5223                                               buf, buf_len);
5224         if (ret)
5225                 return ret;
5226
5227         *authorizer = auth->authorizer_buf;
5228         *authorizer_len = auth->authorizer_buf_len;
5229         return 0;
5230 }
5231
5232 static int mds_handle_auth_done(struct ceph_connection *con,
5233                                 u64 global_id, void *reply, int reply_len,
5234                                 u8 *session_key, int *session_key_len,
5235                                 u8 *con_secret, int *con_secret_len)
5236 {
5237         struct ceph_mds_session *s = con->private;
5238         struct ceph_auth_client *ac = s->s_mdsc->fsc->client->monc.auth;
5239         struct ceph_auth_handshake *auth = &s->s_auth;
5240
5241         return ceph_auth_handle_svc_reply_done(ac, auth, reply, reply_len,
5242                                                session_key, session_key_len,
5243                                                con_secret, con_secret_len);
5244 }
5245
5246 static int mds_handle_auth_bad_method(struct ceph_connection *con,
5247                                       int used_proto, int result,
5248                                       const int *allowed_protos, int proto_cnt,
5249                                       const int *allowed_modes, int mode_cnt)
5250 {
5251         struct ceph_mds_session *s = con->private;
5252         struct ceph_mon_client *monc = &s->s_mdsc->fsc->client->monc;
5253         int ret;
5254
5255         if (ceph_auth_handle_bad_authorizer(monc->auth, CEPH_ENTITY_TYPE_MDS,
5256                                             used_proto, result,
5257                                             allowed_protos, proto_cnt,
5258                                             allowed_modes, mode_cnt)) {
5259                 ret = ceph_monc_validate_auth(monc);
5260                 if (ret)
5261                         return ret;
5262         }
5263
5264         return -EACCES;
5265 }
5266
5267 static struct ceph_msg *mds_alloc_msg(struct ceph_connection *con,
5268                                 struct ceph_msg_header *hdr, int *skip)
5269 {
5270         struct ceph_msg *msg;
5271         int type = (int) le16_to_cpu(hdr->type);
5272         int front_len = (int) le32_to_cpu(hdr->front_len);
5273
5274         if (con->in_msg)
5275                 return con->in_msg;
5276
5277         *skip = 0;
5278         msg = ceph_msg_new(type, front_len, GFP_NOFS, false);
5279         if (!msg) {
5280                 pr_err("unable to allocate msg type %d len %d\n",
5281                        type, front_len);
5282                 return NULL;
5283         }
5284
5285         return msg;
5286 }
5287
5288 static int mds_sign_message(struct ceph_msg *msg)
5289 {
5290        struct ceph_mds_session *s = msg->con->private;
5291        struct ceph_auth_handshake *auth = &s->s_auth;
5292
5293        return ceph_auth_sign_message(auth, msg);
5294 }
5295
5296 static int mds_check_message_signature(struct ceph_msg *msg)
5297 {
5298        struct ceph_mds_session *s = msg->con->private;
5299        struct ceph_auth_handshake *auth = &s->s_auth;
5300
5301        return ceph_auth_check_message_signature(auth, msg);
5302 }
5303
5304 static const struct ceph_connection_operations mds_con_ops = {
5305         .get = mds_get_con,
5306         .put = mds_put_con,
5307         .alloc_msg = mds_alloc_msg,
5308         .dispatch = mds_dispatch,
5309         .peer_reset = mds_peer_reset,
5310         .get_authorizer = mds_get_authorizer,
5311         .add_authorizer_challenge = mds_add_authorizer_challenge,
5312         .verify_authorizer_reply = mds_verify_authorizer_reply,
5313         .invalidate_authorizer = mds_invalidate_authorizer,
5314         .sign_message = mds_sign_message,
5315         .check_message_signature = mds_check_message_signature,
5316         .get_auth_request = mds_get_auth_request,
5317         .handle_auth_reply_more = mds_handle_auth_reply_more,
5318         .handle_auth_done = mds_handle_auth_done,
5319         .handle_auth_bad_method = mds_handle_auth_bad_method,
5320 };
5321
5322 /* eof */