fs/ceph/mds_client.c

   1 #include <linux/ceph/ceph_debug.h>
   2
   3 #include <linux/fs.h>
   4 #include <linux/wait.h>
   5 #include <linux/slab.h>
   6 #include <linux/gfp.h>
   7 #include <linux/sched.h>
   8 #include <linux/debugfs.h>
   9 #include <linux/seq_file.h>
  10 #include <linux/utsname.h>
  11 #include <linux/ratelimit.h>
  12
  13 #include "super.h"
  14 #include "mds_client.h"
  15
  16 #include <linux/ceph/ceph_features.h>
  17 #include <linux/ceph/messenger.h>
  18 #include <linux/ceph/decode.h>
  19 #include <linux/ceph/pagelist.h>
  20 #include <linux/ceph/auth.h>
  21 #include <linux/ceph/debugfs.h>
  22
  23 /*
  24  * A cluster of MDS (metadata server) daemons is responsible for
  25  * managing the file system namespace (the directory hierarchy and
  26  * inodes) and for coordinating shared access to storage.  Metadata is
  27  * partitioning hierarchically across a number of servers, and that
  28  * partition varies over time as the cluster adjusts the distribution
  29  * in order to balance load.
  30  *
  31  * The MDS client is primarily responsible to managing synchronous
  32  * metadata requests for operations like open, unlink, and so forth.
  33  * If there is a MDS failure, we find out about it when we (possibly
  34  * request and) receive a new MDS map, and can resubmit affected
  35  * requests.
  36  *
  37  * For the most part, though, we take advantage of a lossless
  38  * communications channel to the MDS, and do not need to worry about
  39  * timing out or resubmitting requests.
  40  *
  41  * We maintain a stateful "session" with each MDS we interact with.
  42  * Within each session, we sent periodic heartbeat messages to ensure
  43  * any capabilities or leases we have been issues remain valid.  If
  44  * the session times out and goes stale, our leases and capabilities
  45  * are no longer valid.
  46  */
  47
  48 struct ceph_reconnect_state {
  49         int nr_caps;
  50         struct ceph_pagelist *pagelist;
  51         bool flock;
  52 };
  53
  54 static void __wake_requests(struct ceph_mds_client *mdsc,
  55                             struct list_head *head);
  56
  57 static const struct ceph_connection_operations mds_con_ops;
  58
  59
  60 /*
  61  * mds reply parsing
  62  */
  63
  64 /*
  65  * parse individual inode info
  66  */
  67 static int parse_reply_info_in(void **p, void *end,
  68                                struct ceph_mds_reply_info_in *info,
  69                                u64 features)
  70 {
  71         int err = -EIO;
  72
  73         info->in = *p;
  74         *p += sizeof(struct ceph_mds_reply_inode) +
  75                 sizeof(*info->in->fragtree.splits) *
  76                 le32_to_cpu(info->in->fragtree.nsplits);
  77
  78         ceph_decode_32_safe(p, end, info->symlink_len, bad);
  79         ceph_decode_need(p, end, info->symlink_len, bad);
  80         info->symlink = *p;
  81         *p += info->symlink_len;
  82
  83         if (features & CEPH_FEATURE_DIRLAYOUTHASH)
  84                 ceph_decode_copy_safe(p, end, &info->dir_layout,
  85                                       sizeof(info->dir_layout), bad);
  86         else
  87                 memset(&info->dir_layout, 0, sizeof(info->dir_layout));
  88
  89         ceph_decode_32_safe(p, end, info->xattr_len, bad);
  90         ceph_decode_need(p, end, info->xattr_len, bad);
  91         info->xattr_data = *p;
  92         *p += info->xattr_len;
  93
  94         if (features & CEPH_FEATURE_MDS_INLINE_DATA) {
  95                 ceph_decode_64_safe(p, end, info->inline_version, bad);
  96                 ceph_decode_32_safe(p, end, info->inline_len, bad);
  97                 ceph_decode_need(p, end, info->inline_len, bad);
  98                 info->inline_data = *p;
  99                 *p += info->inline_len;
 100         } else
 101                 info->inline_version = CEPH_INLINE_NONE;
 102
 103         return 0;
 104 bad:
 105         return err;
 106 }
 107
 108 /*
 109  * parse a normal reply, which may contain a (dir+)dentry and/or a
 110  * target inode.
 111  */
 112 static int parse_reply_info_trace(void **p, void *end,
 113                                   struct ceph_mds_reply_info_parsed *info,
 114                                   u64 features)
 115 {
 116         int err;
 117
 118         if (info->head->is_dentry) {
 119                 err = parse_reply_info_in(p, end, &info->diri, features);
 120                 if (err < 0)
 121                         goto out_bad;
 122
 123                 if (unlikely(*p + sizeof(*info->dirfrag) > end))
 124                         goto bad;
 125                 info->dirfrag = *p;
 126                 *p += sizeof(*info->dirfrag) +
 127                         sizeof(u32)*le32_to_cpu(info->dirfrag->ndist);
 128                 if (unlikely(*p > end))
 129                         goto bad;
 130
 131                 ceph_decode_32_safe(p, end, info->dname_len, bad);
 132                 ceph_decode_need(p, end, info->dname_len, bad);
 133                 info->dname = *p;
 134                 *p += info->dname_len;
 135                 info->dlease = *p;
 136                 *p += sizeof(*info->dlease);
 137         }
 138
 139         if (info->head->is_target) {
 140                 err = parse_reply_info_in(p, end, &info->targeti, features);
 141                 if (err < 0)
 142                         goto out_bad;
 143         }
 144
 145         if (unlikely(*p != end))
 146                 goto bad;
 147         return 0;
 148
 149 bad:
 150         err = -EIO;
 151 out_bad:
 152         pr_err("problem parsing mds trace %d\n", err);
 153         return err;
 154 }
 155
 156 /*
 157  * parse readdir results
 158  */
 159 static int parse_reply_info_dir(void **p, void *end,
 160                                 struct ceph_mds_reply_info_parsed *info,
 161                                 u64 features)
 162 {
 163         u32 num, i = 0;
 164         int err;
 165
 166         info->dir_dir = *p;
 167         if (*p + sizeof(*info->dir_dir) > end)
 168                 goto bad;
 169         *p += sizeof(*info->dir_dir) +
 170                 sizeof(u32)*le32_to_cpu(info->dir_dir->ndist);
 171         if (*p > end)
 172                 goto bad;
 173
 174         ceph_decode_need(p, end, sizeof(num) + 2, bad);
 175         num = ceph_decode_32(p);
 176         info->dir_end = ceph_decode_8(p);
 177         info->dir_complete = ceph_decode_8(p);
 178         if (num == 0)
 179                 goto done;
 180
 181         BUG_ON(!info->dir_in);
 182         info->dir_dname = (void *)(info->dir_in + num);
 183         info->dir_dname_len = (void *)(info->dir_dname + num);
 184         info->dir_dlease = (void *)(info->dir_dname_len + num);
 185         if ((unsigned long)(info->dir_dlease + num) >
 186             (unsigned long)info->dir_in + info->dir_buf_size) {
 187                 pr_err("dir contents are larger than expected\n");
 188                 WARN_ON(1);
 189                 goto bad;
 190         }
 191
 192         info->dir_nr = num;
 193         while (num) {
 194                 /* dentry */
 195                 ceph_decode_need(p, end, sizeof(u32)*2, bad);
 196                 info->dir_dname_len[i] = ceph_decode_32(p);
 197                 ceph_decode_need(p, end, info->dir_dname_len[i], bad);
 198                 info->dir_dname[i] = *p;
 199                 *p += info->dir_dname_len[i];
 200                 dout("parsed dir dname '%.*s'\n", info->dir_dname_len[i],
 201                      info->dir_dname[i]);
 202                 info->dir_dlease[i] = *p;
 203                 *p += sizeof(struct ceph_mds_reply_lease);
 204
 205                 /* inode */
 206                 err = parse_reply_info_in(p, end, &info->dir_in[i], features);
 207                 if (err < 0)
 208                         goto out_bad;
 209                 i++;
 210                 num--;
 211         }
 212
 213 done:
 214         if (*p != end)
 215                 goto bad;
 216         return 0;
 217
 218 bad:
 219         err = -EIO;
 220 out_bad:
 221         pr_err("problem parsing dir contents %d\n", err);
 222         return err;
 223 }
 224
 225 /*
 226  * parse fcntl F_GETLK results
 227  */
 228 static int parse_reply_info_filelock(void **p, void *end,
 229                                      struct ceph_mds_reply_info_parsed *info,
 230                                      u64 features)
 231 {
 232         if (*p + sizeof(*info->filelock_reply) > end)
 233                 goto bad;
 234
 235         info->filelock_reply = *p;
 236         *p += sizeof(*info->filelock_reply);
 237
 238         if (unlikely(*p != end))
 239                 goto bad;
 240         return 0;
 241
 242 bad:
 243         return -EIO;
 244 }
 245
 246 /*
 247  * parse create results
 248  */
 249 static int parse_reply_info_create(void **p, void *end,
 250                                   struct ceph_mds_reply_info_parsed *info,
 251                                   u64 features)
 252 {
 253         if (features & CEPH_FEATURE_REPLY_CREATE_INODE) {
 254                 if (*p == end) {
 255                         info->has_create_ino = false;
 256                 } else {
 257                         info->has_create_ino = true;
 258                         info->ino = ceph_decode_64(p);
 259                 }
 260         }
 261
 262         if (unlikely(*p != end))
 263                 goto bad;
 264         return 0;
 265
 266 bad:
 267         return -EIO;
 268 }
 269
 270 /*
 271  * parse extra results
 272  */
 273 static int parse_reply_info_extra(void **p, void *end,
 274                                   struct ceph_mds_reply_info_parsed *info,
 275                                   u64 features)
 276 {
 277         if (info->head->op == CEPH_MDS_OP_GETFILELOCK)
 278                 return parse_reply_info_filelock(p, end, info, features);
 279         else if (info->head->op == CEPH_MDS_OP_READDIR ||
 280                  info->head->op == CEPH_MDS_OP_LSSNAP)
 281                 return parse_reply_info_dir(p, end, info, features);
 282         else if (info->head->op == CEPH_MDS_OP_CREATE)
 283                 return parse_reply_info_create(p, end, info, features);
 284         else
 285                 return -EIO;
 286 }
 287
 288 /*
 289  * parse entire mds reply
 290  */
 291 static int parse_reply_info(struct ceph_msg *msg,
 292                             struct ceph_mds_reply_info_parsed *info,
 293                             u64 features)
 294 {
 295         void *p, *end;
 296         u32 len;
 297         int err;
 298
 299         info->head = msg->front.iov_base;
 300         p = msg->front.iov_base + sizeof(struct ceph_mds_reply_head);
 301         end = p + msg->front.iov_len - sizeof(struct ceph_mds_reply_head);
 302
 303         /* trace */
 304         ceph_decode_32_safe(&p, end, len, bad);
 305         if (len > 0) {
 306                 ceph_decode_need(&p, end, len, bad);
 307                 err = parse_reply_info_trace(&p, p+len, info, features);
 308                 if (err < 0)
 309                         goto out_bad;
 310         }
 311
 312         /* extra */
 313         ceph_decode_32_safe(&p, end, len, bad);
 314         if (len > 0) {
 315                 ceph_decode_need(&p, end, len, bad);
 316                 err = parse_reply_info_extra(&p, p+len, info, features);
 317                 if (err < 0)
 318                         goto out_bad;
 319         }
 320
 321         /* snap blob */
 322         ceph_decode_32_safe(&p, end, len, bad);
 323         info->snapblob_len = len;
 324         info->snapblob = p;
 325         p += len;
 326
 327         if (p != end)
 328                 goto bad;
 329         return 0;
 330
 331 bad:
 332         err = -EIO;
 333 out_bad:
 334         pr_err("mds parse_reply err %d\n", err);
 335         return err;
 336 }
 337
 338 static void destroy_reply_info(struct ceph_mds_reply_info_parsed *info)
 339 {
 340         if (!info->dir_in)
 341                 return;
 342         free_pages((unsigned long)info->dir_in, get_order(info->dir_buf_size));
 343 }
 344
 345
 346 /*
 347  * sessions
 348  */
 349 const char *ceph_session_state_name(int s)
 350 {
 351         switch (s) {
 352         case CEPH_MDS_SESSION_NEW: return "new";
 353         case CEPH_MDS_SESSION_OPENING: return "opening";
 354         case CEPH_MDS_SESSION_OPEN: return "open";
 355         case CEPH_MDS_SESSION_HUNG: return "hung";
 356         case CEPH_MDS_SESSION_CLOSING: return "closing";
 357         case CEPH_MDS_SESSION_RESTARTING: return "restarting";
 358         case CEPH_MDS_SESSION_RECONNECTING: return "reconnecting";
 359         default: return "???";
 360         }
 361 }
 362
 363 static struct ceph_mds_session *get_session(struct ceph_mds_session *s)
 364 {
 365         if (atomic_inc_not_zero(&s->s_ref)) {
 366                 dout("mdsc get_session %p %d -> %d\n", s,
 367                      atomic_read(&s->s_ref)-1, atomic_read(&s->s_ref));
 368                 return s;
 369         } else {
 370                 dout("mdsc get_session %p 0 -- FAIL", s);
 371                 return NULL;
 372         }
 373 }
 374
 375 void ceph_put_mds_session(struct ceph_mds_session *s)
 376 {
 377         dout("mdsc put_session %p %d -> %d\n", s,
 378              atomic_read(&s->s_ref), atomic_read(&s->s_ref)-1);
 379         if (atomic_dec_and_test(&s->s_ref)) {
 380                 if (s->s_auth.authorizer)
 381                         ceph_auth_destroy_authorizer(
 382                                 s->s_mdsc->fsc->client->monc.auth,
 383                                 s->s_auth.authorizer);
 384                 kfree(s);
 385         }
 386 }
 387
 388 /*
 389  * called under mdsc->mutex
 390  */
 391 struct ceph_mds_session *__ceph_lookup_mds_session(struct ceph_mds_client *mdsc,
 392                                                    int mds)
 393 {
 394         struct ceph_mds_session *session;
 395
 396         if (mds >= mdsc->max_sessions || mdsc->sessions[mds] == NULL)
 397                 return NULL;
 398         session = mdsc->sessions[mds];
 399         dout("lookup_mds_session %p %d\n", session,
 400              atomic_read(&session->s_ref));
 401         get_session(session);
 402         return session;
 403 }
 404
 405 static bool __have_session(struct ceph_mds_client *mdsc, int mds)
 406 {
 407         if (mds >= mdsc->max_sessions)
 408                 return false;
 409         return mdsc->sessions[mds];
 410 }
 411
 412 static int __verify_registered_session(struct ceph_mds_client *mdsc,
 413                                        struct ceph_mds_session *s)
 414 {
 415         if (s->s_mds >= mdsc->max_sessions ||
 416             mdsc->sessions[s->s_mds] != s)
 417                 return -ENOENT;
 418         return 0;
 419 }
 420
 421 /*
 422  * create+register a new session for given mds.
 423  * called under mdsc->mutex.
 424  */
 425 static struct ceph_mds_session *register_session(struct ceph_mds_client *mdsc,
 426                                                  int mds)
 427 {
 428         struct ceph_mds_session *s;
 429
 430         if (mds >= mdsc->mdsmap->m_max_mds)
 431                 return ERR_PTR(-EINVAL);
 432
 433         s = kzalloc(sizeof(*s), GFP_NOFS);
 434         if (!s)
 435                 return ERR_PTR(-ENOMEM);
 436         s->s_mdsc = mdsc;
 437         s->s_mds = mds;
 438         s->s_state = CEPH_MDS_SESSION_NEW;
 439         s->s_ttl = 0;
 440         s->s_seq = 0;
 441         mutex_init(&s->s_mutex);
 442
 443         ceph_con_init(&s->s_con, s, &mds_con_ops, &mdsc->fsc->client->msgr);
 444
 445         spin_lock_init(&s->s_gen_ttl_lock);
 446         s->s_cap_gen = 0;
 447         s->s_cap_ttl = jiffies - 1;
 448
 449         spin_lock_init(&s->s_cap_lock);
 450         s->s_renew_requested = 0;
 451         s->s_renew_seq = 0;
 452         INIT_LIST_HEAD(&s->s_caps);
 453         s->s_nr_caps = 0;
 454         s->s_trim_caps = 0;
 455         atomic_set(&s->s_ref, 1);
 456         INIT_LIST_HEAD(&s->s_waiting);
 457         INIT_LIST_HEAD(&s->s_unsafe);
 458         s->s_num_cap_releases = 0;
 459         s->s_cap_reconnect = 0;
 460         s->s_cap_iterator = NULL;
 461         INIT_LIST_HEAD(&s->s_cap_releases);
 462         INIT_LIST_HEAD(&s->s_cap_flushing);
 463         INIT_LIST_HEAD(&s->s_cap_snaps_flushing);
 464
 465         dout("register_session mds%d\n", mds);
 466         if (mds >= mdsc->max_sessions) {
 467                 int newmax = 1 << get_count_order(mds+1);
 468                 struct ceph_mds_session **sa;
 469
 470                 dout("register_session realloc to %d\n", newmax);
 471                 sa = kcalloc(newmax, sizeof(void *), GFP_NOFS);
 472                 if (sa == NULL)
 473                         goto fail_realloc;
 474                 if (mdsc->sessions) {
 475                         memcpy(sa, mdsc->sessions,
 476                                mdsc->max_sessions * sizeof(void *));
 477                         kfree(mdsc->sessions);
 478                 }
 479                 mdsc->sessions = sa;
 480                 mdsc->max_sessions = newmax;
 481         }
 482         mdsc->sessions[mds] = s;
 483         atomic_inc(&mdsc->num_sessions);
 484         atomic_inc(&s->s_ref);  /* one ref to sessions[], one to caller */
 485
 486         ceph_con_open(&s->s_con, CEPH_ENTITY_TYPE_MDS, mds,
 487                       ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
 488
 489         return s;
 490
 491 fail_realloc:
 492         kfree(s);
 493         return ERR_PTR(-ENOMEM);
 494 }
 495
 496 /*
 497  * called under mdsc->mutex
 498  */
 499 static void __unregister_session(struct ceph_mds_client *mdsc,
 500                                struct ceph_mds_session *s)
 501 {
 502         dout("__unregister_session mds%d %p\n", s->s_mds, s);
 503         BUG_ON(mdsc->sessions[s->s_mds] != s);
 504         mdsc->sessions[s->s_mds] = NULL;
 505         ceph_con_close(&s->s_con);
 506         ceph_put_mds_session(s);
 507         atomic_dec(&mdsc->num_sessions);
 508 }
 509
 510 /*
 511  * drop session refs in request.
 512  *
 513  * should be last request ref, or hold mdsc->mutex
 514  */
 515 static void put_request_session(struct ceph_mds_request *req)
 516 {
 517         if (req->r_session) {
 518                 ceph_put_mds_session(req->r_session);
 519                 req->r_session = NULL;
 520         }
 521 }
 522
 523 void ceph_mdsc_release_request(struct kref *kref)
 524 {
 525         struct ceph_mds_request *req = container_of(kref,
 526                                                     struct ceph_mds_request,
 527                                                     r_kref);
 528         destroy_reply_info(&req->r_reply_info);
 529         if (req->r_request)
 530                 ceph_msg_put(req->r_request);
 531         if (req->r_reply)
 532                 ceph_msg_put(req->r_reply);
 533         if (req->r_inode) {
 534                 ceph_put_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
 535                 iput(req->r_inode);
 536         }
 537         if (req->r_locked_dir)
 538                 ceph_put_cap_refs(ceph_inode(req->r_locked_dir), CEPH_CAP_PIN);
 539         iput(req->r_target_inode);
 540         if (req->r_dentry)
 541                 dput(req->r_dentry);
 542         if (req->r_old_dentry)
 543                 dput(req->r_old_dentry);
 544         if (req->r_old_dentry_dir) {
 545                 /*
 546                  * track (and drop pins for) r_old_dentry_dir
 547                  * separately, since r_old_dentry's d_parent may have
 548                  * changed between the dir mutex being dropped and
 549                  * this request being freed.
 550                  */
 551                 ceph_put_cap_refs(ceph_inode(req->r_old_dentry_dir),
 552                                   CEPH_CAP_PIN);
 553                 iput(req->r_old_dentry_dir);
 554         }
 555         kfree(req->r_path1);
 556         kfree(req->r_path2);
 557         if (req->r_pagelist)
 558                 ceph_pagelist_release(req->r_pagelist);
 559         put_request_session(req);
 560         ceph_unreserve_caps(req->r_mdsc, &req->r_caps_reservation);
 561         kfree(req);
 562 }
 563
 564 /*
 565  * lookup session, bump ref if found.
 566  *
 567  * called under mdsc->mutex.
 568  */
 569 static struct ceph_mds_request *__lookup_request(struct ceph_mds_client *mdsc,
 570                                              u64 tid)
 571 {
 572         struct ceph_mds_request *req;
 573         struct rb_node *n = mdsc->request_tree.rb_node;
 574
 575         while (n) {
 576                 req = rb_entry(n, struct ceph_mds_request, r_node);
 577                 if (tid < req->r_tid)
 578                         n = n->rb_left;
 579                 else if (tid > req->r_tid)
 580                         n = n->rb_right;
 581                 else {
 582                         ceph_mdsc_get_request(req);
 583                         return req;
 584                 }
 585         }
 586         return NULL;
 587 }
 588
 589 static void __insert_request(struct ceph_mds_client *mdsc,
 590                              struct ceph_mds_request *new)
 591 {
 592         struct rb_node **p = &mdsc->request_tree.rb_node;
 593         struct rb_node *parent = NULL;
 594         struct ceph_mds_request *req = NULL;
 595
 596         while (*p) {
 597                 parent = *p;
 598                 req = rb_entry(parent, struct ceph_mds_request, r_node);
 599                 if (new->r_tid < req->r_tid)
 600                         p = &(*p)->rb_left;
 601                 else if (new->r_tid > req->r_tid)
 602                         p = &(*p)->rb_right;
 603                 else
 604                         BUG();
 605         }
 606
 607         rb_link_node(&new->r_node, parent, p);
 608         rb_insert_color(&new->r_node, &mdsc->request_tree);
 609 }
 610
 611 /*
 612  * Register an in-flight request, and assign a tid.  Link to directory
 613  * are modifying (if any).
 614  *
 615  * Called under mdsc->mutex.
 616  */
 617 static void __register_request(struct ceph_mds_client *mdsc,
 618                                struct ceph_mds_request *req,
 619                                struct inode *dir)
 620 {
 621         req->r_tid = ++mdsc->last_tid;
 622         if (req->r_num_caps)
 623                 ceph_reserve_caps(mdsc, &req->r_caps_reservation,
 624                                   req->r_num_caps);
 625         dout("__register_request %p tid %lld\n", req, req->r_tid);
 626         ceph_mdsc_get_request(req);
 627         __insert_request(mdsc, req);
 628
 629         req->r_uid = current_fsuid();
 630         req->r_gid = current_fsgid();
 631
 632         if (mdsc->oldest_tid == 0 && req->r_op != CEPH_MDS_OP_SETFILELOCK)
 633                 mdsc->oldest_tid = req->r_tid;
 634
 635         if (dir) {
 636                 struct ceph_inode_info *ci = ceph_inode(dir);
 637
 638                 ihold(dir);
 639                 spin_lock(&ci->i_unsafe_lock);
 640                 req->r_unsafe_dir = dir;
 641                 list_add_tail(&req->r_unsafe_dir_item, &ci->i_unsafe_dirops);
 642                 spin_unlock(&ci->i_unsafe_lock);
 643         }
 644 }
 645
 646 static void __unregister_request(struct ceph_mds_client *mdsc,
 647                                  struct ceph_mds_request *req)
 648 {
 649         dout("__unregister_request %p tid %lld\n", req, req->r_tid);
 650
 651         if (req->r_tid == mdsc->oldest_tid) {
 652                 struct rb_node *p = rb_next(&req->r_node);
 653                 mdsc->oldest_tid = 0;
 654                 while (p) {
 655                         struct ceph_mds_request *next_req =
 656                                 rb_entry(p, struct ceph_mds_request, r_node);
 657                         if (next_req->r_op != CEPH_MDS_OP_SETFILELOCK) {
 658                                 mdsc->oldest_tid = next_req->r_tid;
 659                                 break;
 660                         }
 661                         p = rb_next(p);
 662                 }
 663         }
 664
 665         rb_erase(&req->r_node, &mdsc->request_tree);
 666         RB_CLEAR_NODE(&req->r_node);
 667
 668         if (req->r_unsafe_dir) {
 669                 struct ceph_inode_info *ci = ceph_inode(req->r_unsafe_dir);
 670
 671                 spin_lock(&ci->i_unsafe_lock);
 672                 list_del_init(&req->r_unsafe_dir_item);
 673                 spin_unlock(&ci->i_unsafe_lock);
 674
 675                 iput(req->r_unsafe_dir);
 676                 req->r_unsafe_dir = NULL;
 677         }
 678
 679         complete_all(&req->r_safe_completion);
 680
 681         ceph_mdsc_put_request(req);
 682 }
 683
 684 /*
 685  * Choose mds to send request to next.  If there is a hint set in the
 686  * request (e.g., due to a prior forward hint from the mds), use that.
 687  * Otherwise, consult frag tree and/or caps to identify the
 688  * appropriate mds.  If all else fails, choose randomly.
 689  *
 690  * Called under mdsc->mutex.
 691  */
 692 static struct dentry *get_nonsnap_parent(struct dentry *dentry)
 693 {
 694         /*
 695          * we don't need to worry about protecting the d_parent access
 696          * here because we never renaming inside the snapped namespace
 697          * except to resplice to another snapdir, and either the old or new
 698          * result is a valid result.
 699          */
 700         while (!IS_ROOT(dentry) && ceph_snap(d_inode(dentry)) != CEPH_NOSNAP)
 701                 dentry = dentry->d_parent;
 702         return dentry;
 703 }
 704
 705 static int __choose_mds(struct ceph_mds_client *mdsc,
 706                         struct ceph_mds_request *req)
 707 {
 708         struct inode *inode;
 709         struct ceph_inode_info *ci;
 710         struct ceph_cap *cap;
 711         int mode = req->r_direct_mode;
 712         int mds = -1;
 713         u32 hash = req->r_direct_hash;
 714         bool is_hash = req->r_direct_is_hash;
 715
 716         /*
 717          * is there a specific mds we should try?  ignore hint if we have
 718          * no session and the mds is not up (active or recovering).
 719          */
 720         if (req->r_resend_mds >= 0 &&
 721             (__have_session(mdsc, req->r_resend_mds) ||
 722              ceph_mdsmap_get_state(mdsc->mdsmap, req->r_resend_mds) > 0)) {
 723                 dout("choose_mds using resend_mds mds%d\n",
 724                      req->r_resend_mds);
 725                 return req->r_resend_mds;
 726         }
 727
 728         if (mode == USE_RANDOM_MDS)
 729                 goto random;
 730
 731         inode = NULL;
 732         if (req->r_inode) {
 733                 inode = req->r_inode;
 734         } else if (req->r_dentry) {
 735                 /* ignore race with rename; old or new d_parent is okay */
 736                 struct dentry *parent = req->r_dentry->d_parent;
 737                 struct inode *dir = d_inode(parent);
 738
 739                 if (dir->i_sb != mdsc->fsc->sb) {
 740                         /* not this fs! */
 741                         inode = d_inode(req->r_dentry);
 742                 } else if (ceph_snap(dir) != CEPH_NOSNAP) {
 743                         /* direct snapped/virtual snapdir requests
 744                          * based on parent dir inode */
 745                         struct dentry *dn = get_nonsnap_parent(parent);
 746                         inode = d_inode(dn);
 747                         dout("__choose_mds using nonsnap parent %p\n", inode);
 748                 } else {
 749                         /* dentry target */
 750                         inode = d_inode(req->r_dentry);
 751                         if (!inode || mode == USE_AUTH_MDS) {
 752                                 /* dir + name */
 753                                 inode = dir;
 754                                 hash = ceph_dentry_hash(dir, req->r_dentry);
 755                                 is_hash = true;
 756                         }
 757                 }
 758         }
 759
 760         dout("__choose_mds %p is_hash=%d (%d) mode %d\n", inode, (int)is_hash,
 761              (int)hash, mode);
 762         if (!inode)
 763                 goto random;
 764         ci = ceph_inode(inode);
 765
 766         if (is_hash && S_ISDIR(inode->i_mode)) {
 767                 struct ceph_inode_frag frag;
 768                 int found;
 769
 770                 ceph_choose_frag(ci, hash, &frag, &found);
 771                 if (found) {
 772                         if (mode == USE_ANY_MDS && frag.ndist > 0) {
 773                                 u8 r;
 774
 775                                 /* choose a random replica */
 776                                 get_random_bytes(&r, 1);
 777                                 r %= frag.ndist;
 778                                 mds = frag.dist[r];
 779                                 dout("choose_mds %p %llx.%llx "
 780                                      "frag %u mds%d (%d/%d)\n",
 781                                      inode, ceph_vinop(inode),
 782                                      frag.frag, mds,
 783                                      (int)r, frag.ndist);
 784                                 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
 785                                     CEPH_MDS_STATE_ACTIVE)
 786                                         return mds;
 787                         }
 788
 789                         /* since this file/dir wasn't known to be
 790                          * replicated, then we want to look for the
 791                          * authoritative mds. */
 792                         mode = USE_AUTH_MDS;
 793                         if (frag.mds >= 0) {
 794                                 /* choose auth mds */
 795                                 mds = frag.mds;
 796                                 dout("choose_mds %p %llx.%llx "
 797                                      "frag %u mds%d (auth)\n",
 798                                      inode, ceph_vinop(inode), frag.frag, mds);
 799                                 if (ceph_mdsmap_get_state(mdsc->mdsmap, mds) >=
 800                                     CEPH_MDS_STATE_ACTIVE)
 801                                         return mds;
 802                         }
 803                 }
 804         }
 805
 806         spin_lock(&ci->i_ceph_lock);
 807         cap = NULL;
 808         if (mode == USE_AUTH_MDS)
 809                 cap = ci->i_auth_cap;
 810         if (!cap && !RB_EMPTY_ROOT(&ci->i_caps))
 811                 cap = rb_entry(rb_first(&ci->i_caps), struct ceph_cap, ci_node);
 812         if (!cap) {
 813                 spin_unlock(&ci->i_ceph_lock);
 814                 goto random;
 815         }
 816         mds = cap->session->s_mds;
 817         dout("choose_mds %p %llx.%llx mds%d (%scap %p)\n",
 818              inode, ceph_vinop(inode), mds,
 819              cap == ci->i_auth_cap ? "auth " : "", cap);
 820         spin_unlock(&ci->i_ceph_lock);
 821         return mds;
 822
 823 random:
 824         mds = ceph_mdsmap_get_random_mds(mdsc->mdsmap);
 825         dout("choose_mds chose random mds%d\n", mds);
 826         return mds;
 827 }
 828
 829
 830 /*
 831  * session messages
 832  */
 833 static struct ceph_msg *create_session_msg(u32 op, u64 seq)
 834 {
 835         struct ceph_msg *msg;
 836         struct ceph_mds_session_head *h;
 837
 838         msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h), GFP_NOFS,
 839                            false);
 840         if (!msg) {
 841                 pr_err("create_session_msg ENOMEM creating msg\n");
 842                 return NULL;
 843         }
 844         h = msg->front.iov_base;
 845         h->op = cpu_to_le32(op);
 846         h->seq = cpu_to_le64(seq);
 847
 848         return msg;
 849 }
 850
 851 /*
 852  * session message, specialization for CEPH_SESSION_REQUEST_OPEN
 853  * to include additional client metadata fields.
 854  */
 855 static struct ceph_msg *create_session_open_msg(struct ceph_mds_client *mdsc, u64 seq)
 856 {
 857         struct ceph_msg *msg;
 858         struct ceph_mds_session_head *h;
 859         int i = -1;
 860         int metadata_bytes = 0;
 861         int metadata_key_count = 0;
 862         struct ceph_options *opt = mdsc->fsc->client->options;
 863         void *p;
 864
 865         const char* metadata[][2] = {
 866                 {"hostname", utsname()->nodename},
 867                 {"kernel_version", utsname()->release},
 868                 {"entity_id", opt->name ? opt->name : ""},
 869                 {NULL, NULL}
 870         };
 871
 872         /* Calculate serialized length of metadata */
 873         metadata_bytes = 4;  /* map length */
 874         for (i = 0; metadata[i][0] != NULL; ++i) {
 875                 metadata_bytes += 8 + strlen(metadata[i][0]) +
 876                         strlen(metadata[i][1]);
 877                 metadata_key_count++;
 878         }
 879
 880         /* Allocate the message */
 881         msg = ceph_msg_new(CEPH_MSG_CLIENT_SESSION, sizeof(*h) + metadata_bytes,
 882                            GFP_NOFS, false);
 883         if (!msg) {
 884                 pr_err("create_session_msg ENOMEM creating msg\n");
 885                 return NULL;
 886         }
 887         h = msg->front.iov_base;
 888         h->op = cpu_to_le32(CEPH_SESSION_REQUEST_OPEN);
 889         h->seq = cpu_to_le64(seq);
 890
 891         /*
 892          * Serialize client metadata into waiting buffer space, using
 893          * the format that userspace expects for map<string, string>
 894          *
 895          * ClientSession messages with metadata are v2
 896          */
 897         msg->hdr.version = cpu_to_le16(2);
 898         msg->hdr.compat_version = cpu_to_le16(1);
 899
 900         /* The write pointer, following the session_head structure */
 901         p = msg->front.iov_base + sizeof(*h);
 902
 903         /* Number of entries in the map */
 904         ceph_encode_32(&p, metadata_key_count);
 905
 906         /* Two length-prefixed strings for each entry in the map */
 907         for (i = 0; metadata[i][0] != NULL; ++i) {
 908                 size_t const key_len = strlen(metadata[i][0]);
 909                 size_t const val_len = strlen(metadata[i][1]);
 910
 911                 ceph_encode_32(&p, key_len);
 912                 memcpy(p, metadata[i][0], key_len);
 913                 p += key_len;
 914                 ceph_encode_32(&p, val_len);
 915                 memcpy(p, metadata[i][1], val_len);
 916                 p += val_len;
 917         }
 918
 919         return msg;
 920 }
 921
 922 /*
 923  * send session open request.
 924  *
 925  * called under mdsc->mutex
 926  */
 927 static int __open_session(struct ceph_mds_client *mdsc,
 928                           struct ceph_mds_session *session)
 929 {
 930         struct ceph_msg *msg;
 931         int mstate;
 932         int mds = session->s_mds;
 933
 934         /* wait for mds to go active? */
 935         mstate = ceph_mdsmap_get_state(mdsc->mdsmap, mds);
 936         dout("open_session to mds%d (%s)\n", mds,
 937              ceph_mds_state_name(mstate));
 938         session->s_state = CEPH_MDS_SESSION_OPENING;
 939         session->s_renew_requested = jiffies;
 940
 941         /* send connect message */
 942         msg = create_session_open_msg(mdsc, session->s_seq);
 943         if (!msg)
 944                 return -ENOMEM;
 945         ceph_con_send(&session->s_con, msg);
 946         return 0;
 947 }
 948
 949 /*
 950  * open sessions for any export targets for the given mds
 951  *
 952  * called under mdsc->mutex
 953  */
 954 static struct ceph_mds_session *
 955 __open_export_target_session(struct ceph_mds_client *mdsc, int target)
 956 {
 957         struct ceph_mds_session *session;
 958
 959         session = __ceph_lookup_mds_session(mdsc, target);
 960         if (!session) {
 961                 session = register_session(mdsc, target);
 962                 if (IS_ERR(session))
 963                         return session;
 964         }
 965         if (session->s_state == CEPH_MDS_SESSION_NEW ||
 966             session->s_state == CEPH_MDS_SESSION_CLOSING)
 967                 __open_session(mdsc, session);
 968
 969         return session;
 970 }
 971
 972 struct ceph_mds_session *
 973 ceph_mdsc_open_export_target_session(struct ceph_mds_client *mdsc, int target)
 974 {
 975         struct ceph_mds_session *session;
 976
 977         dout("open_export_target_session to mds%d\n", target);
 978
 979         mutex_lock(&mdsc->mutex);
 980         session = __open_export_target_session(mdsc, target);
 981         mutex_unlock(&mdsc->mutex);
 982
 983         return session;
 984 }
 985
 986 static void __open_export_target_sessions(struct ceph_mds_client *mdsc,
 987                                           struct ceph_mds_session *session)
 988 {
 989         struct ceph_mds_info *mi;
 990         struct ceph_mds_session *ts;
 991         int i, mds = session->s_mds;
 992
 993         if (mds >= mdsc->mdsmap->m_max_mds)
 994                 return;
 995
 996         mi = &mdsc->mdsmap->m_info[mds];
 997         dout("open_export_target_sessions for mds%d (%d targets)\n",
 998              session->s_mds, mi->num_export_targets);
 999
1000         for (i = 0; i < mi->num_export_targets; i++) {
1001                 ts = __open_export_target_session(mdsc, mi->export_targets[i]);
1002                 if (!IS_ERR(ts))
1003                         ceph_put_mds_session(ts);
1004         }
1005 }
1006
1007 void ceph_mdsc_open_export_target_sessions(struct ceph_mds_client *mdsc,
1008                                            struct ceph_mds_session *session)
1009 {
1010         mutex_lock(&mdsc->mutex);
1011         __open_export_target_sessions(mdsc, session);
1012         mutex_unlock(&mdsc->mutex);
1013 }
1014
1015 /*
1016  * session caps
1017  */
1018
1019 /* caller holds s_cap_lock, we drop it */
1020 static void cleanup_cap_releases(struct ceph_mds_client *mdsc,
1021                                  struct ceph_mds_session *session)
1022         __releases(session->s_cap_lock)
1023 {
1024         LIST_HEAD(tmp_list);
1025         list_splice_init(&session->s_cap_releases, &tmp_list);
1026         session->s_num_cap_releases = 0;
1027         spin_unlock(&session->s_cap_lock);
1028
1029         dout("cleanup_cap_releases mds%d\n", session->s_mds);
1030         while (!list_empty(&tmp_list)) {
1031                 struct ceph_cap *cap;
1032                 /* zero out the in-progress message */
1033                 cap = list_first_entry(&tmp_list,
1034                                         struct ceph_cap, session_caps);
1035                 list_del(&cap->session_caps);
1036                 ceph_put_cap(mdsc, cap);
1037         }
1038 }
1039
1040 static void cleanup_session_requests(struct ceph_mds_client *mdsc,
1041                                      struct ceph_mds_session *session)
1042 {
1043         struct ceph_mds_request *req;
1044         struct rb_node *p;
1045
1046         dout("cleanup_session_requests mds%d\n", session->s_mds);
1047         mutex_lock(&mdsc->mutex);
1048         while (!list_empty(&session->s_unsafe)) {
1049                 req = list_first_entry(&session->s_unsafe,
1050                                        struct ceph_mds_request, r_unsafe_item);
1051                 list_del_init(&req->r_unsafe_item);
1052                 pr_warn_ratelimited(" dropping unsafe request %llu\n",
1053                                     req->r_tid);
1054                 __unregister_request(mdsc, req);
1055         }
1056         /* zero r_attempts, so kick_requests() will re-send requests */
1057         p = rb_first(&mdsc->request_tree);
1058         while (p) {
1059                 req = rb_entry(p, struct ceph_mds_request, r_node);
1060                 p = rb_next(p);
1061                 if (req->r_session &&
1062                     req->r_session->s_mds == session->s_mds)
1063                         req->r_attempts = 0;
1064         }
1065         mutex_unlock(&mdsc->mutex);
1066 }
1067
1068 /*
1069  * Helper to safely iterate over all caps associated with a session, with
1070  * special care taken to handle a racing __ceph_remove_cap().
1071  *
1072  * Caller must hold session s_mutex.
1073  */
1074 static int iterate_session_caps(struct ceph_mds_session *session,
1075                                  int (*cb)(struct inode *, struct ceph_cap *,
1076                                             void *), void *arg)
1077 {
1078         struct list_head *p;
1079         struct ceph_cap *cap;
1080         struct inode *inode, *last_inode = NULL;
1081         struct ceph_cap *old_cap = NULL;
1082         int ret;
1083
1084         dout("iterate_session_caps %p mds%d\n", session, session->s_mds);
1085         spin_lock(&session->s_cap_lock);
1086         p = session->s_caps.next;
1087         while (p != &session->s_caps) {
1088                 cap = list_entry(p, struct ceph_cap, session_caps);
1089                 inode = igrab(&cap->ci->vfs_inode);
1090                 if (!inode) {
1091                         p = p->next;
1092                         continue;
1093                 }
1094                 session->s_cap_iterator = cap;
1095                 spin_unlock(&session->s_cap_lock);
1096
1097                 if (last_inode) {
1098                         iput(last_inode);
1099                         last_inode = NULL;
1100                 }
1101                 if (old_cap) {
1102                         ceph_put_cap(session->s_mdsc, old_cap);
1103                         old_cap = NULL;
1104                 }
1105
1106                 ret = cb(inode, cap, arg);
1107                 last_inode = inode;
1108
1109                 spin_lock(&session->s_cap_lock);
1110                 p = p->next;
1111                 if (cap->ci == NULL) {
1112                         dout("iterate_session_caps  finishing cap %p removal\n",
1113                              cap);
1114                         BUG_ON(cap->session != session);
1115                         cap->session = NULL;
1116                         list_del_init(&cap->session_caps);
1117                         session->s_nr_caps--;
1118                         if (cap->queue_release) {
1119                                 list_add_tail(&cap->session_caps,
1120                                               &session->s_cap_releases);
1121                                 session->s_num_cap_releases++;
1122                         } else {
1123                                 old_cap = cap;  /* put_cap it w/o locks held */
1124                         }
1125                 }
1126                 if (ret < 0)
1127                         goto out;
1128         }
1129         ret = 0;
1130 out:
1131         session->s_cap_iterator = NULL;
1132         spin_unlock(&session->s_cap_lock);
1133
1134         iput(last_inode);
1135         if (old_cap)
1136                 ceph_put_cap(session->s_mdsc, old_cap);
1137
1138         return ret;
1139 }
1140
1141 static int remove_session_caps_cb(struct inode *inode, struct ceph_cap *cap,
1142                                   void *arg)
1143 {
1144         struct ceph_inode_info *ci = ceph_inode(inode);
1145         LIST_HEAD(to_remove);
1146         int drop = 0;
1147
1148         dout("removing cap %p, ci is %p, inode is %p\n",
1149              cap, ci, &ci->vfs_inode);
1150         spin_lock(&ci->i_ceph_lock);
1151         __ceph_remove_cap(cap, false);
1152         if (!ci->i_auth_cap) {
1153                 struct ceph_cap_flush *cf;
1154                 struct ceph_mds_client *mdsc =
1155                         ceph_sb_to_client(inode->i_sb)->mdsc;
1156
1157                 while (true) {
1158                         struct rb_node *n = rb_first(&ci->i_cap_flush_tree);
1159                         if (!n)
1160                                 break;
1161                         cf = rb_entry(n, struct ceph_cap_flush, i_node);
1162                         rb_erase(&cf->i_node, &ci->i_cap_flush_tree);
1163                         list_add(&cf->list, &to_remove);
1164                 }
1165
1166                 spin_lock(&mdsc->cap_dirty_lock);
1167
1168                 list_for_each_entry(cf, &to_remove, list)
1169                         rb_erase(&cf->g_node, &mdsc->cap_flush_tree);
1170
1171                 if (!list_empty(&ci->i_dirty_item)) {
1172                         pr_warn_ratelimited(
1173                                 " dropping dirty %s state for %p %lld\n",
1174                                 ceph_cap_string(ci->i_dirty_caps),
1175                                 inode, ceph_ino(inode));
1176                         ci->i_dirty_caps = 0;
1177                         list_del_init(&ci->i_dirty_item);
1178                         drop = 1;
1179                 }
1180                 if (!list_empty(&ci->i_flushing_item)) {
1181                         pr_warn_ratelimited(
1182                                 " dropping dirty+flushing %s state for %p %lld\n",
1183                                 ceph_cap_string(ci->i_flushing_caps),
1184                                 inode, ceph_ino(inode));
1185                         ci->i_flushing_caps = 0;
1186                         list_del_init(&ci->i_flushing_item);
1187                         mdsc->num_cap_flushing--;
1188                         drop = 1;
1189                 }
1190                 spin_unlock(&mdsc->cap_dirty_lock);
1191
1192                 if (!ci->i_dirty_caps && ci->i_prealloc_cap_flush) {
1193                         list_add(&ci->i_prealloc_cap_flush->list, &to_remove);
1194                         ci->i_prealloc_cap_flush = NULL;
1195                 }
1196         }
1197         spin_unlock(&ci->i_ceph_lock);
1198         while (!list_empty(&to_remove)) {
1199                 struct ceph_cap_flush *cf;
1200                 cf = list_first_entry(&to_remove,
1201                                       struct ceph_cap_flush, list);
1202                 list_del(&cf->list);
1203                 ceph_free_cap_flush(cf);
1204         }
1205         while (drop--)
1206                 iput(inode);
1207         return 0;
1208 }
1209
1210 /*
1211  * caller must hold session s_mutex
1212  */
1213 static void remove_session_caps(struct ceph_mds_session *session)
1214 {
1215         dout("remove_session_caps on %p\n", session);
1216         iterate_session_caps(session, remove_session_caps_cb, NULL);
1217
1218         spin_lock(&session->s_cap_lock);
1219         if (session->s_nr_caps > 0) {
1220                 struct super_block *sb = session->s_mdsc->fsc->sb;
1221                 struct inode *inode;
1222                 struct ceph_cap *cap, *prev = NULL;
1223                 struct ceph_vino vino;
1224                 /*
1225                  * iterate_session_caps() skips inodes that are being
1226                  * deleted, we need to wait until deletions are complete.
1227                  * __wait_on_freeing_inode() is designed for the job,
1228                  * but it is not exported, so use lookup inode function
1229                  * to access it.
1230                  */
1231                 while (!list_empty(&session->s_caps)) {
1232                         cap = list_entry(session->s_caps.next,
1233                                          struct ceph_cap, session_caps);
1234                         if (cap == prev)
1235                                 break;
1236                         prev = cap;
1237                         vino = cap->ci->i_vino;
1238                         spin_unlock(&session->s_cap_lock);
1239
1240                         inode = ceph_find_inode(sb, vino);
1241                         iput(inode);
1242
1243                         spin_lock(&session->s_cap_lock);
1244                 }
1245         }
1246
1247         // drop cap expires and unlock s_cap_lock
1248         cleanup_cap_releases(session->s_mdsc, session);
1249
1250         BUG_ON(session->s_nr_caps > 0);
1251         BUG_ON(!list_empty(&session->s_cap_flushing));
1252 }
1253
1254 /*
1255  * wake up any threads waiting on this session's caps.  if the cap is
1256  * old (didn't get renewed on the client reconnect), remove it now.
1257  *
1258  * caller must hold s_mutex.
1259  */
1260 static int wake_up_session_cb(struct inode *inode, struct ceph_cap *cap,
1261                               void *arg)
1262 {
1263         struct ceph_inode_info *ci = ceph_inode(inode);
1264
1265         wake_up_all(&ci->i_cap_wq);
1266         if (arg) {
1267                 spin_lock(&ci->i_ceph_lock);
1268                 ci->i_wanted_max_size = 0;
1269                 ci->i_requested_max_size = 0;
1270                 spin_unlock(&ci->i_ceph_lock);
1271         }
1272         return 0;
1273 }
1274
1275 static void wake_up_session_caps(struct ceph_mds_session *session,
1276                                  int reconnect)
1277 {
1278         dout("wake_up_session_caps %p mds%d\n", session, session->s_mds);
1279         iterate_session_caps(session, wake_up_session_cb,
1280                              (void *)(unsigned long)reconnect);
1281 }
1282
1283 /*
1284  * Send periodic message to MDS renewing all currently held caps.  The
1285  * ack will reset the expiration for all caps from this session.
1286  *
1287  * caller holds s_mutex
1288  */
1289 static int send_renew_caps(struct ceph_mds_client *mdsc,
1290                            struct ceph_mds_session *session)
1291 {
1292         struct ceph_msg *msg;
1293         int state;
1294
1295         if (time_after_eq(jiffies, session->s_cap_ttl) &&
1296             time_after_eq(session->s_cap_ttl, session->s_renew_requested))
1297                 pr_info("mds%d caps stale\n", session->s_mds);
1298         session->s_renew_requested = jiffies;
1299
1300         /* do not try to renew caps until a recovering mds has reconnected
1301          * with its clients. */
1302         state = ceph_mdsmap_get_state(mdsc->mdsmap, session->s_mds);
1303         if (state < CEPH_MDS_STATE_RECONNECT) {
1304                 dout("send_renew_caps ignoring mds%d (%s)\n",
1305                      session->s_mds, ceph_mds_state_name(state));
1306                 return 0;
1307         }
1308
1309         dout("send_renew_caps to mds%d (%s)\n", session->s_mds,
1310                 ceph_mds_state_name(state));
1311         msg = create_session_msg(CEPH_SESSION_REQUEST_RENEWCAPS,
1312                                  ++session->s_renew_seq);
1313         if (!msg)
1314                 return -ENOMEM;
1315         ceph_con_send(&session->s_con, msg);
1316         return 0;
1317 }
1318
1319 static int send_flushmsg_ack(struct ceph_mds_client *mdsc,
1320                              struct ceph_mds_session *session, u64 seq)
1321 {
1322         struct ceph_msg *msg;
1323
1324         dout("send_flushmsg_ack to mds%d (%s)s seq %lld\n",
1325              session->s_mds, ceph_session_state_name(session->s_state), seq);
1326         msg = create_session_msg(CEPH_SESSION_FLUSHMSG_ACK, seq);
1327         if (!msg)
1328                 return -ENOMEM;
1329         ceph_con_send(&session->s_con, msg);
1330         return 0;
1331 }
1332
1333
1334 /*
1335  * Note new cap ttl, and any transition from stale -> not stale (fresh?).
1336  *
1337  * Called under session->s_mutex
1338  */
1339 static void renewed_caps(struct ceph_mds_client *mdsc,
1340                          struct ceph_mds_session *session, int is_renew)
1341 {
1342         int was_stale;
1343         int wake = 0;
1344
1345         spin_lock(&session->s_cap_lock);
1346         was_stale = is_renew && time_after_eq(jiffies, session->s_cap_ttl);
1347
1348         session->s_cap_ttl = session->s_renew_requested +
1349                 mdsc->mdsmap->m_session_timeout*HZ;
1350
1351         if (was_stale) {
1352                 if (time_before(jiffies, session->s_cap_ttl)) {
1353                         pr_info("mds%d caps renewed\n", session->s_mds);
1354                         wake = 1;
1355                 } else {
1356                         pr_info("mds%d caps still stale\n", session->s_mds);
1357                 }
1358         }
1359         dout("renewed_caps mds%d ttl now %lu, was %s, now %s\n",
1360              session->s_mds, session->s_cap_ttl, was_stale ? "stale" : "fresh",
1361              time_before(jiffies, session->s_cap_ttl) ? "stale" : "fresh");
1362         spin_unlock(&session->s_cap_lock);
1363
1364         if (wake)
1365                 wake_up_session_caps(session, 0);
1366 }
1367
1368 /*
1369  * send a session close request
1370  */
1371 static int request_close_session(struct ceph_mds_client *mdsc,
1372                                  struct ceph_mds_session *session)
1373 {
1374         struct ceph_msg *msg;
1375
1376         dout("request_close_session mds%d state %s seq %lld\n",
1377              session->s_mds, ceph_session_state_name(session->s_state),
1378              session->s_seq);
1379         msg = create_session_msg(CEPH_SESSION_REQUEST_CLOSE, session->s_seq);
1380         if (!msg)
1381                 return -ENOMEM;
1382         ceph_con_send(&session->s_con, msg);
1383         return 0;
1384 }
1385
1386 /*
1387  * Called with s_mutex held.
1388  */
1389 static int __close_session(struct ceph_mds_client *mdsc,
1390                          struct ceph_mds_session *session)
1391 {
1392         if (session->s_state >= CEPH_MDS_SESSION_CLOSING)
1393                 return 0;
1394         session->s_state = CEPH_MDS_SESSION_CLOSING;
1395         return request_close_session(mdsc, session);
1396 }
1397
1398 /*
1399  * Trim old(er) caps.
1400  *
1401  * Because we can't cache an inode without one or more caps, we do
1402  * this indirectly: if a cap is unused, we prune its aliases, at which
1403  * point the inode will hopefully get dropped to.
1404  *
1405  * Yes, this is a bit sloppy.  Our only real goal here is to respond to
1406  * memory pressure from the MDS, though, so it needn't be perfect.
1407  */
1408 static int trim_caps_cb(struct inode *inode, struct ceph_cap *cap, void *arg)
1409 {
1410         struct ceph_mds_session *session = arg;
1411         struct ceph_inode_info *ci = ceph_inode(inode);
1412         int used, wanted, oissued, mine;
1413
1414         if (session->s_trim_caps <= 0)
1415                 return -1;
1416
1417         spin_lock(&ci->i_ceph_lock);
1418         mine = cap->issued | cap->implemented;
1419         used = __ceph_caps_used(ci);
1420         wanted = __ceph_caps_file_wanted(ci);
1421         oissued = __ceph_caps_issued_other(ci, cap);
1422
1423         dout("trim_caps_cb %p cap %p mine %s oissued %s used %s wanted %s\n",
1424              inode, cap, ceph_cap_string(mine), ceph_cap_string(oissued),
1425              ceph_cap_string(used), ceph_cap_string(wanted));
1426         if (cap == ci->i_auth_cap) {
1427                 if (ci->i_dirty_caps || ci->i_flushing_caps ||
1428                     !list_empty(&ci->i_cap_snaps))
1429                         goto out;
1430                 if ((used | wanted) & CEPH_CAP_ANY_WR)
1431                         goto out;
1432         }
1433         if ((used | wanted) & ~oissued & mine)
1434                 goto out;   /* we need these caps */
1435
1436         session->s_trim_caps--;
1437         if (oissued) {
1438                 /* we aren't the only cap.. just remove us */
1439                 __ceph_remove_cap(cap, true);
1440         } else {
1441                 /* try to drop referring dentries */
1442                 spin_unlock(&ci->i_ceph_lock);
1443                 d_prune_aliases(inode);
1444                 dout("trim_caps_cb %p cap %p  pruned, count now %d\n",
1445                      inode, cap, atomic_read(&inode->i_count));
1446                 return 0;
1447         }
1448
1449 out:
1450         spin_unlock(&ci->i_ceph_lock);
1451         return 0;
1452 }
1453
1454 /*
1455  * Trim session cap count down to some max number.
1456  */
1457 static int trim_caps(struct ceph_mds_client *mdsc,
1458                      struct ceph_mds_session *session,
1459                      int max_caps)
1460 {
1461         int trim_caps = session->s_nr_caps - max_caps;
1462
1463         dout("trim_caps mds%d start: %d / %d, trim %d\n",
1464              session->s_mds, session->s_nr_caps, max_caps, trim_caps);
1465         if (trim_caps > 0) {
1466                 session->s_trim_caps = trim_caps;
1467                 iterate_session_caps(session, trim_caps_cb, session);
1468                 dout("trim_caps mds%d done: %d / %d, trimmed %d\n",
1469                      session->s_mds, session->s_nr_caps, max_caps,
1470                         trim_caps - session->s_trim_caps);
1471                 session->s_trim_caps = 0;
1472         }
1473
1474         ceph_send_cap_releases(mdsc, session);
1475         return 0;
1476 }
1477
1478 static int check_capsnap_flush(struct ceph_inode_info *ci,
1479                                u64 want_snap_seq)
1480 {
1481         int ret = 1;
1482         spin_lock(&ci->i_ceph_lock);
1483         if (want_snap_seq > 0 && !list_empty(&ci->i_cap_snaps)) {
1484                 struct ceph_cap_snap *capsnap =
1485                         list_first_entry(&ci->i_cap_snaps,
1486                                          struct ceph_cap_snap, ci_item);
1487                 ret = capsnap->follows >= want_snap_seq;
1488         }
1489         spin_unlock(&ci->i_ceph_lock);
1490         return ret;
1491 }
1492
1493 static int check_caps_flush(struct ceph_mds_client *mdsc,
1494                             u64 want_flush_tid)
1495 {
1496         struct rb_node *n;
1497         struct ceph_cap_flush *cf;
1498         int ret = 1;
1499
1500         spin_lock(&mdsc->cap_dirty_lock);
1501         n = rb_first(&mdsc->cap_flush_tree);
1502         cf = n ? rb_entry(n, struct ceph_cap_flush, g_node) : NULL;
1503         if (cf && cf->tid <= want_flush_tid) {
1504                 dout("check_caps_flush still flushing tid %llu <= %llu\n",
1505                      cf->tid, want_flush_tid);
1506                 ret = 0;
1507         }
1508         spin_unlock(&mdsc->cap_dirty_lock);
1509         return ret;
1510 }
1511
1512 /*
1513  * flush all dirty inode data to disk.
1514  *
1515  * returns true if we've flushed through want_flush_tid
1516  */
1517 static void wait_caps_flush(struct ceph_mds_client *mdsc,
1518                             u64 want_flush_tid, u64 want_snap_seq)
1519 {
1520         int mds;
1521
1522         dout("check_caps_flush want %llu snap want %llu\n",
1523              want_flush_tid, want_snap_seq);
1524         mutex_lock(&mdsc->mutex);
1525         for (mds = 0; mds < mdsc->max_sessions; ) {
1526                 struct ceph_mds_session *session = mdsc->sessions[mds];
1527                 struct inode *inode = NULL;
1528
1529                 if (!session) {
1530                         mds++;
1531                         continue;
1532                 }
1533                 get_session(session);
1534                 mutex_unlock(&mdsc->mutex);
1535
1536                 mutex_lock(&session->s_mutex);
1537                 if (!list_empty(&session->s_cap_snaps_flushing)) {
1538                         struct ceph_cap_snap *capsnap =
1539                                 list_first_entry(&session->s_cap_snaps_flushing,
1540                                                  struct ceph_cap_snap,
1541                                                  flushing_item);
1542                         struct ceph_inode_info *ci = capsnap->ci;
1543                         if (!check_capsnap_flush(ci, want_snap_seq)) {
1544                                 dout("check_cap_flush still flushing snap %p "
1545                                      "follows %lld <= %lld to mds%d\n",
1546                                      &ci->vfs_inode, capsnap->follows,
1547                                      want_snap_seq, mds);
1548                                 inode = igrab(&ci->vfs_inode);
1549                         }
1550                 }
1551                 mutex_unlock(&session->s_mutex);
1552                 ceph_put_mds_session(session);
1553
1554                 if (inode) {
1555                         wait_event(mdsc->cap_flushing_wq,
1556                                    check_capsnap_flush(ceph_inode(inode),
1557                                                        want_snap_seq));
1558                         iput(inode);
1559                 } else {
1560                         mds++;
1561                 }
1562
1563                 mutex_lock(&mdsc->mutex);
1564         }
1565         mutex_unlock(&mdsc->mutex);
1566
1567         wait_event(mdsc->cap_flushing_wq,
1568                    check_caps_flush(mdsc, want_flush_tid));
1569
1570         dout("check_caps_flush ok, flushed thru %llu\n", want_flush_tid);
1571 }
1572
1573 /*
1574  * called under s_mutex
1575  */
1576 void ceph_send_cap_releases(struct ceph_mds_client *mdsc,
1577                             struct ceph_mds_session *session)
1578 {
1579         struct ceph_msg *msg = NULL;
1580         struct ceph_mds_cap_release *head;
1581         struct ceph_mds_cap_item *item;
1582         struct ceph_cap *cap;
1583         LIST_HEAD(tmp_list);
1584         int num_cap_releases;
1585
1586         spin_lock(&session->s_cap_lock);
1587 again:
1588         list_splice_init(&session->s_cap_releases, &tmp_list);
1589         num_cap_releases = session->s_num_cap_releases;
1590         session->s_num_cap_releases = 0;
1591         spin_unlock(&session->s_cap_lock);
1592
1593         while (!list_empty(&tmp_list)) {
1594                 if (!msg) {
1595                         msg = ceph_msg_new(CEPH_MSG_CLIENT_CAPRELEASE,
1596                                         PAGE_CACHE_SIZE, GFP_NOFS, false);
1597                         if (!msg)
1598                                 goto out_err;
1599                         head = msg->front.iov_base;
1600                         head->num = cpu_to_le32(0);
1601                         msg->front.iov_len = sizeof(*head);
1602                 }
1603                 cap = list_first_entry(&tmp_list, struct ceph_cap,
1604                                         session_caps);
1605                 list_del(&cap->session_caps);
1606                 num_cap_releases--;
1607
1608                 head = msg->front.iov_base;
1609                 le32_add_cpu(&head->num, 1);
1610                 item = msg->front.iov_base + msg->front.iov_len;
1611                 item->ino = cpu_to_le64(cap->cap_ino);
1612                 item->cap_id = cpu_to_le64(cap->cap_id);
1613                 item->migrate_seq = cpu_to_le32(cap->mseq);
1614                 item->seq = cpu_to_le32(cap->issue_seq);
1615                 msg->front.iov_len += sizeof(*item);
1616
1617                 ceph_put_cap(mdsc, cap);
1618
1619                 if (le32_to_cpu(head->num) == CEPH_CAPS_PER_RELEASE) {
1620                         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1621                         dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
1622                         ceph_con_send(&session->s_con, msg);
1623                         msg = NULL;
1624                 }
1625         }
1626
1627         BUG_ON(num_cap_releases != 0);
1628
1629         spin_lock(&session->s_cap_lock);
1630         if (!list_empty(&session->s_cap_releases))
1631                 goto again;
1632         spin_unlock(&session->s_cap_lock);
1633
1634         if (msg) {
1635                 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
1636                 dout("send_cap_releases mds%d %p\n", session->s_mds, msg);
1637                 ceph_con_send(&session->s_con, msg);
1638         }
1639         return;
1640 out_err:
1641         pr_err("send_cap_releases mds%d, failed to allocate message\n",
1642                 session->s_mds);
1643         spin_lock(&session->s_cap_lock);
1644         list_splice(&tmp_list, &session->s_cap_releases);
1645         session->s_num_cap_releases += num_cap_releases;
1646         spin_unlock(&session->s_cap_lock);
1647 }
1648
1649 /*
1650  * requests
1651  */
1652
1653 int ceph_alloc_readdir_reply_buffer(struct ceph_mds_request *req,
1654                                     struct inode *dir)
1655 {
1656         struct ceph_inode_info *ci = ceph_inode(dir);
1657         struct ceph_mds_reply_info_parsed *rinfo = &req->r_reply_info;
1658         struct ceph_mount_options *opt = req->r_mdsc->fsc->mount_options;
1659         size_t size = sizeof(*rinfo->dir_in) + sizeof(*rinfo->dir_dname_len) +
1660                       sizeof(*rinfo->dir_dname) + sizeof(*rinfo->dir_dlease);
1661         int order, num_entries;
1662
1663         spin_lock(&ci->i_ceph_lock);
1664         num_entries = ci->i_files + ci->i_subdirs;
1665         spin_unlock(&ci->i_ceph_lock);
1666         num_entries = max(num_entries, 1);
1667         num_entries = min(num_entries, opt->max_readdir);
1668
1669         order = get_order(size * num_entries);
1670         while (order >= 0) {
1671                 rinfo->dir_in = (void*)__get_free_pages(GFP_KERNEL |
1672                                                         __GFP_NOWARN,
1673                                                         order);
1674                 if (rinfo->dir_in)
1675                         break;
1676                 order--;
1677         }
1678         if (!rinfo->dir_in)
1679                 return -ENOMEM;
1680
1681         num_entries = (PAGE_SIZE << order) / size;
1682         num_entries = min(num_entries, opt->max_readdir);
1683
1684         rinfo->dir_buf_size = PAGE_SIZE << order;
1685         req->r_num_caps = num_entries + 1;
1686         req->r_args.readdir.max_entries = cpu_to_le32(num_entries);
1687         req->r_args.readdir.max_bytes = cpu_to_le32(opt->max_readdir_bytes);
1688         return 0;
1689 }
1690
1691 /*
1692  * Create an mds request.
1693  */
1694 struct ceph_mds_request *
1695 ceph_mdsc_create_request(struct ceph_mds_client *mdsc, int op, int mode)
1696 {
1697         struct ceph_mds_request *req = kzalloc(sizeof(*req), GFP_NOFS);
1698
1699         if (!req)
1700                 return ERR_PTR(-ENOMEM);
1701
1702         mutex_init(&req->r_fill_mutex);
1703         req->r_mdsc = mdsc;
1704         req->r_started = jiffies;
1705         req->r_resend_mds = -1;
1706         INIT_LIST_HEAD(&req->r_unsafe_dir_item);
1707         req->r_fmode = -1;
1708         kref_init(&req->r_kref);
1709         INIT_LIST_HEAD(&req->r_wait);
1710         init_completion(&req->r_completion);
1711         init_completion(&req->r_safe_completion);
1712         INIT_LIST_HEAD(&req->r_unsafe_item);
1713
1714         req->r_stamp = CURRENT_TIME;
1715
1716         req->r_op = op;
1717         req->r_direct_mode = mode;
1718         return req;
1719 }
1720
1721 /*
1722  * return oldest (lowest) request, tid in request tree, 0 if none.
1723  *
1724  * called under mdsc->mutex.
1725  */
1726 static struct ceph_mds_request *__get_oldest_req(struct ceph_mds_client *mdsc)
1727 {
1728         if (RB_EMPTY_ROOT(&mdsc->request_tree))
1729                 return NULL;
1730         return rb_entry(rb_first(&mdsc->request_tree),
1731                         struct ceph_mds_request, r_node);
1732 }
1733
1734 static inline  u64 __get_oldest_tid(struct ceph_mds_client *mdsc)
1735 {
1736         return mdsc->oldest_tid;
1737 }
1738
1739 /*
1740  * Build a dentry's path.  Allocate on heap; caller must kfree.  Based
1741  * on build_path_from_dentry in fs/cifs/dir.c.
1742  *
1743  * If @stop_on_nosnap, generate path relative to the first non-snapped
1744  * inode.
1745  *
1746  * Encode hidden .snap dirs as a double /, i.e.
1747  *   foo/.snap/bar -> foo//bar
1748  */
1749 char *ceph_mdsc_build_path(struct dentry *dentry, int *plen, u64 *base,
1750                            int stop_on_nosnap)
1751 {
1752         struct dentry *temp;
1753         char *path;
1754         int len, pos;
1755         unsigned seq;
1756
1757         if (dentry == NULL)
1758                 return ERR_PTR(-EINVAL);
1759
1760 retry:
1761         len = 0;
1762         seq = read_seqbegin(&rename_lock);
1763         rcu_read_lock();
1764         for (temp = dentry; !IS_ROOT(temp);) {
1765                 struct inode *inode = d_inode(temp);
1766                 if (inode && ceph_snap(inode) == CEPH_SNAPDIR)
1767                         len++;  /* slash only */
1768                 else if (stop_on_nosnap && inode &&
1769                          ceph_snap(inode) == CEPH_NOSNAP)
1770                         break;
1771                 else
1772                         len += 1 + temp->d_name.len;
1773                 temp = temp->d_parent;
1774         }
1775         rcu_read_unlock();
1776         if (len)
1777                 len--;  /* no leading '/' */
1778
1779         path = kmalloc(len+1, GFP_NOFS);
1780         if (path == NULL)
1781                 return ERR_PTR(-ENOMEM);
1782         pos = len;
1783         path[pos] = 0;  /* trailing null */
1784         rcu_read_lock();
1785         for (temp = dentry; !IS_ROOT(temp) && pos != 0; ) {
1786                 struct inode *inode;
1787
1788                 spin_lock(&temp->d_lock);
1789                 inode = d_inode(temp);
1790                 if (inode && ceph_snap(inode) == CEPH_SNAPDIR) {
1791                         dout("build_path path+%d: %p SNAPDIR\n",
1792                              pos, temp);
1793                 } else if (stop_on_nosnap && inode &&
1794                            ceph_snap(inode) == CEPH_NOSNAP) {
1795                         spin_unlock(&temp->d_lock);
1796                         break;
1797                 } else {
1798                         pos -= temp->d_name.len;
1799                         if (pos < 0) {
1800                                 spin_unlock(&temp->d_lock);
1801                                 break;
1802                         }
1803                         strncpy(path + pos, temp->d_name.name,
1804                                 temp->d_name.len);
1805                 }
1806                 spin_unlock(&temp->d_lock);
1807                 if (pos)
1808                         path[--pos] = '/';
1809                 temp = temp->d_parent;
1810         }
1811         rcu_read_unlock();
1812         if (pos != 0 || read_seqretry(&rename_lock, seq)) {
1813                 pr_err("build_path did not end path lookup where "
1814                        "expected, namelen is %d, pos is %d\n", len, pos);
1815                 /* presumably this is only possible if racing with a
1816                    rename of one of the parent directories (we can not
1817                    lock the dentries above us to prevent this, but
1818                    retrying should be harmless) */
1819                 kfree(path);
1820                 goto retry;
1821         }
1822
1823         *base = ceph_ino(d_inode(temp));
1824         *plen = len;
1825         dout("build_path on %p %d built %llx '%.*s'\n",
1826              dentry, d_count(dentry), *base, len, path);
1827         return path;
1828 }
1829
1830 static int build_dentry_path(struct dentry *dentry,
1831                              const char **ppath, int *ppathlen, u64 *pino,
1832                              int *pfreepath)
1833 {
1834         char *path;
1835
1836         if (ceph_snap(d_inode(dentry->d_parent)) == CEPH_NOSNAP) {
1837                 *pino = ceph_ino(d_inode(dentry->d_parent));
1838                 *ppath = dentry->d_name.name;
1839                 *ppathlen = dentry->d_name.len;
1840                 return 0;
1841         }
1842         path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
1843         if (IS_ERR(path))
1844                 return PTR_ERR(path);
1845         *ppath = path;
1846         *pfreepath = 1;
1847         return 0;
1848 }
1849
1850 static int build_inode_path(struct inode *inode,
1851                             const char **ppath, int *ppathlen, u64 *pino,
1852                             int *pfreepath)
1853 {
1854         struct dentry *dentry;
1855         char *path;
1856
1857         if (ceph_snap(inode) == CEPH_NOSNAP) {
1858                 *pino = ceph_ino(inode);
1859                 *ppathlen = 0;
1860                 return 0;
1861         }
1862         dentry = d_find_alias(inode);
1863         path = ceph_mdsc_build_path(dentry, ppathlen, pino, 1);
1864         dput(dentry);
1865         if (IS_ERR(path))
1866                 return PTR_ERR(path);
1867         *ppath = path;
1868         *pfreepath = 1;
1869         return 0;
1870 }
1871
1872 /*
1873  * request arguments may be specified via an inode *, a dentry *, or
1874  * an explicit ino+path.
1875  */
1876 static int set_request_path_attr(struct inode *rinode, struct dentry *rdentry,
1877                                   const char *rpath, u64 rino,
1878                                   const char **ppath, int *pathlen,
1879                                   u64 *ino, int *freepath)
1880 {
1881         int r = 0;
1882
1883         if (rinode) {
1884                 r = build_inode_path(rinode, ppath, pathlen, ino, freepath);
1885                 dout(" inode %p %llx.%llx\n", rinode, ceph_ino(rinode),
1886                      ceph_snap(rinode));
1887         } else if (rdentry) {
1888                 r = build_dentry_path(rdentry, ppath, pathlen, ino, freepath);
1889                 dout(" dentry %p %llx/%.*s\n", rdentry, *ino, *pathlen,
1890                      *ppath);
1891         } else if (rpath || rino) {
1892                 *ino = rino;
1893                 *ppath = rpath;
1894                 *pathlen = rpath ? strlen(rpath) : 0;
1895                 dout(" path %.*s\n", *pathlen, rpath);
1896         }
1897
1898         return r;
1899 }
1900
1901 /*
1902  * called under mdsc->mutex
1903  */
1904 static struct ceph_msg *create_request_message(struct ceph_mds_client *mdsc,
1905                                                struct ceph_mds_request *req,
1906                                                int mds, bool drop_cap_releases)
1907 {
1908         struct ceph_msg *msg;
1909         struct ceph_mds_request_head *head;
1910         const char *path1 = NULL;
1911         const char *path2 = NULL;
1912         u64 ino1 = 0, ino2 = 0;
1913         int pathlen1 = 0, pathlen2 = 0;
1914         int freepath1 = 0, freepath2 = 0;
1915         int len;
1916         u16 releases;
1917         void *p, *end;
1918         int ret;
1919
1920         ret = set_request_path_attr(req->r_inode, req->r_dentry,
1921                               req->r_path1, req->r_ino1.ino,
1922                               &path1, &pathlen1, &ino1, &freepath1);
1923         if (ret < 0) {
1924                 msg = ERR_PTR(ret);
1925                 goto out;
1926         }
1927
1928         ret = set_request_path_attr(NULL, req->r_old_dentry,
1929                               req->r_path2, req->r_ino2.ino,
1930                               &path2, &pathlen2, &ino2, &freepath2);
1931         if (ret < 0) {
1932                 msg = ERR_PTR(ret);
1933                 goto out_free1;
1934         }
1935
1936         len = sizeof(*head) +
1937                 pathlen1 + pathlen2 + 2*(1 + sizeof(u32) + sizeof(u64)) +
1938                 sizeof(struct timespec);
1939
1940         /* calculate (max) length for cap releases */
1941         len += sizeof(struct ceph_mds_request_release) *
1942                 (!!req->r_inode_drop + !!req->r_dentry_drop +
1943                  !!req->r_old_inode_drop + !!req->r_old_dentry_drop);
1944         if (req->r_dentry_drop)
1945                 len += req->r_dentry->d_name.len;
1946         if (req->r_old_dentry_drop)
1947                 len += req->r_old_dentry->d_name.len;
1948
1949         msg = ceph_msg_new(CEPH_MSG_CLIENT_REQUEST, len, GFP_NOFS, false);
1950         if (!msg) {
1951                 msg = ERR_PTR(-ENOMEM);
1952                 goto out_free2;
1953         }
1954
1955         msg->hdr.version = cpu_to_le16(2);
1956         msg->hdr.tid = cpu_to_le64(req->r_tid);
1957
1958         head = msg->front.iov_base;
1959         p = msg->front.iov_base + sizeof(*head);
1960         end = msg->front.iov_base + msg->front.iov_len;
1961
1962         head->mdsmap_epoch = cpu_to_le32(mdsc->mdsmap->m_epoch);
1963         head->op = cpu_to_le32(req->r_op);
1964         head->caller_uid = cpu_to_le32(from_kuid(&init_user_ns, req->r_uid));
1965         head->caller_gid = cpu_to_le32(from_kgid(&init_user_ns, req->r_gid));
1966         head->args = req->r_args;
1967
1968         ceph_encode_filepath(&p, end, ino1, path1);
1969         ceph_encode_filepath(&p, end, ino2, path2);
1970
1971         /* make note of release offset, in case we need to replay */
1972         req->r_request_release_offset = p - msg->front.iov_base;
1973
1974         /* cap releases */
1975         releases = 0;
1976         if (req->r_inode_drop)
1977                 releases += ceph_encode_inode_release(&p,
1978                       req->r_inode ? req->r_inode : d_inode(req->r_dentry),
1979                       mds, req->r_inode_drop, req->r_inode_unless, 0);
1980         if (req->r_dentry_drop)
1981                 releases += ceph_encode_dentry_release(&p, req->r_dentry,
1982                        mds, req->r_dentry_drop, req->r_dentry_unless);
1983         if (req->r_old_dentry_drop)
1984                 releases += ceph_encode_dentry_release(&p, req->r_old_dentry,
1985                        mds, req->r_old_dentry_drop, req->r_old_dentry_unless);
1986         if (req->r_old_inode_drop)
1987                 releases += ceph_encode_inode_release(&p,
1988                       d_inode(req->r_old_dentry),
1989                       mds, req->r_old_inode_drop, req->r_old_inode_unless, 0);
1990
1991         if (drop_cap_releases) {
1992                 releases = 0;
1993                 p = msg->front.iov_base + req->r_request_release_offset;
1994         }
1995
1996         head->num_releases = cpu_to_le16(releases);
1997
1998         /* time stamp */
1999         {
2000                 struct ceph_timespec ts;
2001                 ceph_encode_timespec(&ts, &req->r_stamp);
2002                 ceph_encode_copy(&p, &ts, sizeof(ts));
2003         }
2004
2005         BUG_ON(p > end);
2006         msg->front.iov_len = p - msg->front.iov_base;
2007         msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2008
2009         if (req->r_pagelist) {
2010                 struct ceph_pagelist *pagelist = req->r_pagelist;
2011                 atomic_inc(&pagelist->refcnt);
2012                 ceph_msg_data_add_pagelist(msg, pagelist);
2013                 msg->hdr.data_len = cpu_to_le32(pagelist->length);
2014         } else {
2015                 msg->hdr.data_len = 0;
2016         }
2017
2018         msg->hdr.data_off = cpu_to_le16(0);
2019
2020 out_free2:
2021         if (freepath2)
2022                 kfree((char *)path2);
2023 out_free1:
2024         if (freepath1)
2025                 kfree((char *)path1);
2026 out:
2027         return msg;
2028 }
2029
2030 /*
2031  * called under mdsc->mutex if error, under no mutex if
2032  * success.
2033  */
2034 static void complete_request(struct ceph_mds_client *mdsc,
2035                              struct ceph_mds_request *req)
2036 {
2037         if (req->r_callback)
2038                 req->r_callback(mdsc, req);
2039         else
2040                 complete_all(&req->r_completion);
2041 }
2042
2043 /*
2044  * called under mdsc->mutex
2045  */
2046 static int __prepare_send_request(struct ceph_mds_client *mdsc,
2047                                   struct ceph_mds_request *req,
2048                                   int mds, bool drop_cap_releases)
2049 {
2050         struct ceph_mds_request_head *rhead;
2051         struct ceph_msg *msg;
2052         int flags = 0;
2053
2054         req->r_attempts++;
2055         if (req->r_inode) {
2056                 struct ceph_cap *cap =
2057                         ceph_get_cap_for_mds(ceph_inode(req->r_inode), mds);
2058
2059                 if (cap)
2060                         req->r_sent_on_mseq = cap->mseq;
2061                 else
2062                         req->r_sent_on_mseq = -1;
2063         }
2064         dout("prepare_send_request %p tid %lld %s (attempt %d)\n", req,
2065              req->r_tid, ceph_mds_op_name(req->r_op), req->r_attempts);
2066
2067         if (req->r_got_unsafe) {
2068                 void *p;
2069                 /*
2070                  * Replay.  Do not regenerate message (and rebuild
2071                  * paths, etc.); just use the original message.
2072                  * Rebuilding paths will break for renames because
2073                  * d_move mangles the src name.
2074                  */
2075                 msg = req->r_request;
2076                 rhead = msg->front.iov_base;
2077
2078                 flags = le32_to_cpu(rhead->flags);
2079                 flags |= CEPH_MDS_FLAG_REPLAY;
2080                 rhead->flags = cpu_to_le32(flags);
2081
2082                 if (req->r_target_inode)
2083                         rhead->ino = cpu_to_le64(ceph_ino(req->r_target_inode));
2084
2085                 rhead->num_retry = req->r_attempts - 1;
2086
2087                 /* remove cap/dentry releases from message */
2088                 rhead->num_releases = 0;
2089
2090                 /* time stamp */
2091                 p = msg->front.iov_base + req->r_request_release_offset;
2092                 {
2093                         struct ceph_timespec ts;
2094                         ceph_encode_timespec(&ts, &req->r_stamp);
2095                         ceph_encode_copy(&p, &ts, sizeof(ts));
2096                 }
2097
2098                 msg->front.iov_len = p - msg->front.iov_base;
2099                 msg->hdr.front_len = cpu_to_le32(msg->front.iov_len);
2100                 return 0;
2101         }
2102
2103         if (req->r_request) {
2104                 ceph_msg_put(req->r_request);
2105                 req->r_request = NULL;
2106         }
2107         msg = create_request_message(mdsc, req, mds, drop_cap_releases);
2108         if (IS_ERR(msg)) {
2109                 req->r_err = PTR_ERR(msg);
2110                 return PTR_ERR(msg);
2111         }
2112         req->r_request = msg;
2113
2114         rhead = msg->front.iov_base;
2115         rhead->oldest_client_tid = cpu_to_le64(__get_oldest_tid(mdsc));
2116         if (req->r_got_unsafe)
2117                 flags |= CEPH_MDS_FLAG_REPLAY;
2118         if (req->r_locked_dir)
2119                 flags |= CEPH_MDS_FLAG_WANT_DENTRY;
2120         rhead->flags = cpu_to_le32(flags);
2121         rhead->num_fwd = req->r_num_fwd;
2122         rhead->num_retry = req->r_attempts - 1;
2123         rhead->ino = 0;
2124
2125         dout(" r_locked_dir = %p\n", req->r_locked_dir);
2126         return 0;
2127 }
2128
2129 /*
2130  * send request, or put it on the appropriate wait list.
2131  */
2132 static int __do_request(struct ceph_mds_client *mdsc,
2133                         struct ceph_mds_request *req)
2134 {
2135         struct ceph_mds_session *session = NULL;
2136         int mds = -1;
2137         int err = 0;
2138
2139         if (req->r_err || req->r_got_result) {
2140                 if (req->r_aborted)
2141                         __unregister_request(mdsc, req);
2142                 goto out;
2143         }
2144
2145         if (req->r_timeout &&
2146             time_after_eq(jiffies, req->r_started + req->r_timeout)) {
2147                 dout("do_request timed out\n");
2148                 err = -EIO;
2149                 goto finish;
2150         }
2151         if (ACCESS_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN) {
2152                 dout("do_request forced umount\n");
2153                 err = -EIO;
2154                 goto finish;
2155         }
2156
2157         put_request_session(req);
2158
2159         mds = __choose_mds(mdsc, req);
2160         if (mds < 0 ||
2161             ceph_mdsmap_get_state(mdsc->mdsmap, mds) < CEPH_MDS_STATE_ACTIVE) {
2162                 dout("do_request no mds or not active, waiting for map\n");
2163                 list_add(&req->r_wait, &mdsc->waiting_for_map);
2164                 goto out;
2165         }
2166
2167         /* get, open session */
2168         session = __ceph_lookup_mds_session(mdsc, mds);
2169         if (!session) {
2170                 session = register_session(mdsc, mds);
2171                 if (IS_ERR(session)) {
2172                         err = PTR_ERR(session);
2173                         goto finish;
2174                 }
2175         }
2176         req->r_session = get_session(session);
2177
2178         dout("do_request mds%d session %p state %s\n", mds, session,
2179              ceph_session_state_name(session->s_state));
2180         if (session->s_state != CEPH_MDS_SESSION_OPEN &&
2181             session->s_state != CEPH_MDS_SESSION_HUNG) {
2182                 if (session->s_state == CEPH_MDS_SESSION_NEW ||
2183                     session->s_state == CEPH_MDS_SESSION_CLOSING)
2184                         __open_session(mdsc, session);
2185                 list_add(&req->r_wait, &session->s_waiting);
2186                 goto out_session;
2187         }
2188
2189         /* send request */
2190         req->r_resend_mds = -1;   /* forget any previous mds hint */
2191
2192         if (req->r_request_started == 0)   /* note request start time */
2193                 req->r_request_started = jiffies;
2194
2195         err = __prepare_send_request(mdsc, req, mds, false);
2196         if (!err) {
2197                 ceph_msg_get(req->r_request);
2198                 ceph_con_send(&session->s_con, req->r_request);
2199         }
2200
2201 out_session:
2202         ceph_put_mds_session(session);
2203 finish:
2204         if (err) {
2205                 dout("__do_request early error %d\n", err);
2206                 req->r_err = err;
2207                 complete_request(mdsc, req);
2208                 __unregister_request(mdsc, req);
2209         }
2210 out:
2211         return err;
2212 }
2213
2214 /*
2215  * called under mdsc->mutex
2216  */
2217 static void __wake_requests(struct ceph_mds_client *mdsc,
2218                             struct list_head *head)
2219 {
2220         struct ceph_mds_request *req;
2221         LIST_HEAD(tmp_list);
2222
2223         list_splice_init(head, &tmp_list);
2224
2225         while (!list_empty(&tmp_list)) {
2226                 req = list_entry(tmp_list.next,
2227                                  struct ceph_mds_request, r_wait);
2228                 list_del_init(&req->r_wait);
2229                 dout(" wake request %p tid %llu\n", req, req->r_tid);
2230                 __do_request(mdsc, req);
2231         }
2232 }
2233
2234 /*
2235  * Wake up threads with requests pending for @mds, so that they can
2236  * resubmit their requests to a possibly different mds.
2237  */
2238 static void kick_requests(struct ceph_mds_client *mdsc, int mds)
2239 {
2240         struct ceph_mds_request *req;
2241         struct rb_node *p = rb_first(&mdsc->request_tree);
2242
2243         dout("kick_requests mds%d\n", mds);
2244         while (p) {
2245                 req = rb_entry(p, struct ceph_mds_request, r_node);
2246                 p = rb_next(p);
2247                 if (req->r_got_unsafe)
2248                         continue;
2249                 if (req->r_attempts > 0)
2250                         continue; /* only new requests */
2251                 if (req->r_session &&
2252                     req->r_session->s_mds == mds) {
2253                         dout(" kicking tid %llu\n", req->r_tid);
2254                         list_del_init(&req->r_wait);
2255                         __do_request(mdsc, req);
2256                 }
2257         }
2258 }
2259
2260 void ceph_mdsc_submit_request(struct ceph_mds_client *mdsc,
2261                               struct ceph_mds_request *req)
2262 {
2263         dout("submit_request on %p\n", req);
2264         mutex_lock(&mdsc->mutex);
2265         __register_request(mdsc, req, NULL);
2266         __do_request(mdsc, req);
2267         mutex_unlock(&mdsc->mutex);
2268 }
2269
2270 /*
2271  * Synchrously perform an mds request.  Take care of all of the
2272  * session setup, forwarding, retry details.
2273  */
2274 int ceph_mdsc_do_request(struct ceph_mds_client *mdsc,
2275                          struct inode *dir,
2276                          struct ceph_mds_request *req)
2277 {
2278         int err;
2279
2280         dout("do_request on %p\n", req);
2281
2282         /* take CAP_PIN refs for r_inode, r_locked_dir, r_old_dentry */
2283         if (req->r_inode)
2284                 ceph_get_cap_refs(ceph_inode(req->r_inode), CEPH_CAP_PIN);
2285         if (req->r_locked_dir)
2286                 ceph_get_cap_refs(ceph_inode(req->r_locked_dir), CEPH_CAP_PIN);
2287         if (req->r_old_dentry_dir)
2288                 ceph_get_cap_refs(ceph_inode(req->r_old_dentry_dir),
2289                                   CEPH_CAP_PIN);
2290
2291         /* issue */
2292         mutex_lock(&mdsc->mutex);
2293         __register_request(mdsc, req, dir);
2294         __do_request(mdsc, req);
2295
2296         if (req->r_err) {
2297                 err = req->r_err;
2298                 goto out;
2299         }
2300
2301         /* wait */
2302         mutex_unlock(&mdsc->mutex);
2303         dout("do_request waiting\n");
2304         if (!req->r_timeout && req->r_wait_for_completion) {
2305                 err = req->r_wait_for_completion(mdsc, req);
2306         } else {
2307                 long timeleft = wait_for_completion_killable_timeout(
2308                                         &req->r_completion,
2309                                         ceph_timeout_jiffies(req->r_timeout));
2310                 if (timeleft > 0)
2311                         err = 0;
2312                 else if (!timeleft)
2313                         err = -EIO;  /* timed out */
2314                 else
2315                         err = timeleft;  /* killed */
2316         }
2317         dout("do_request waited, got %d\n", err);
2318         mutex_lock(&mdsc->mutex);
2319
2320         /* only abort if we didn't race with a real reply */
2321         if (req->r_got_result) {
2322                 err = le32_to_cpu(req->r_reply_info.head->result);
2323         } else if (err < 0) {
2324                 dout("aborted request %lld with %d\n", req->r_tid, err);
2325
2326                 /*
2327                  * ensure we aren't running concurrently with
2328                  * ceph_fill_trace or ceph_readdir_prepopulate, which
2329                  * rely on locks (dir mutex) held by our caller.
2330                  */
2331                 mutex_lock(&req->r_fill_mutex);
2332                 req->r_err = err;
2333                 req->r_aborted = true;
2334                 mutex_unlock(&req->r_fill_mutex);
2335
2336                 if (req->r_locked_dir &&
2337                     (req->r_op & CEPH_MDS_OP_WRITE))
2338                         ceph_invalidate_dir_request(req);
2339         } else {
2340                 err = req->r_err;
2341         }
2342
2343 out:
2344         mutex_unlock(&mdsc->mutex);
2345         dout("do_request %p done, result %d\n", req, err);
2346         return err;
2347 }
2348
2349 /*
2350  * Invalidate dir's completeness, dentry lease state on an aborted MDS
2351  * namespace request.
2352  */
2353 void ceph_invalidate_dir_request(struct ceph_mds_request *req)
2354 {
2355         struct inode *inode = req->r_locked_dir;
2356
2357         dout("invalidate_dir_request %p (complete, lease(s))\n", inode);
2358
2359         ceph_dir_clear_complete(inode);
2360         if (req->r_dentry)
2361                 ceph_invalidate_dentry_lease(req->r_dentry);
2362         if (req->r_old_dentry)
2363                 ceph_invalidate_dentry_lease(req->r_old_dentry);
2364 }
2365
2366 /*
2367  * Handle mds reply.
2368  *
2369  * We take the session mutex and parse and process the reply immediately.
2370  * This preserves the logical ordering of replies, capabilities, etc., sent
2371  * by the MDS as they are applied to our local cache.
2372  */
2373 static void handle_reply(struct ceph_mds_session *session, struct ceph_msg *msg)
2374 {
2375         struct ceph_mds_client *mdsc = session->s_mdsc;
2376         struct ceph_mds_request *req;
2377         struct ceph_mds_reply_head *head = msg->front.iov_base;
2378         struct ceph_mds_reply_info_parsed *rinfo;  /* parsed reply info */
2379         struct ceph_snap_realm *realm;
2380         u64 tid;
2381         int err, result;
2382         int mds = session->s_mds;
2383
2384         if (msg->front.iov_len < sizeof(*head)) {
2385                 pr_err("mdsc_handle_reply got corrupt (short) reply\n");
2386                 ceph_msg_dump(msg);
2387                 return;
2388         }
2389
2390         /* get request, session */
2391         tid = le64_to_cpu(msg->hdr.tid);
2392         mutex_lock(&mdsc->mutex);
2393         req = __lookup_request(mdsc, tid);
2394         if (!req) {
2395                 dout("handle_reply on unknown tid %llu\n", tid);
2396                 mutex_unlock(&mdsc->mutex);
2397                 return;
2398         }
2399         dout("handle_reply %p\n", req);
2400
2401         /* correct session? */
2402         if (req->r_session != session) {
2403                 pr_err("mdsc_handle_reply got %llu on session mds%d"
2404                        " not mds%d\n", tid, session->s_mds,
2405                        req->r_session ? req->r_session->s_mds : -1);
2406                 mutex_unlock(&mdsc->mutex);
2407                 goto out;
2408         }
2409
2410         /* dup? */
2411         if ((req->r_got_unsafe && !head->safe) ||
2412             (req->r_got_safe && head->safe)) {
2413                 pr_warn("got a dup %s reply on %llu from mds%d\n",
2414                            head->safe ? "safe" : "unsafe", tid, mds);
2415                 mutex_unlock(&mdsc->mutex);
2416                 goto out;
2417         }
2418         if (req->r_got_safe && !head->safe) {
2419                 pr_warn("got unsafe after safe on %llu from mds%d\n",
2420                            tid, mds);
2421                 mutex_unlock(&mdsc->mutex);
2422                 goto out;
2423         }
2424
2425         result = le32_to_cpu(head->result);
2426
2427         /*
2428          * Handle an ESTALE
2429          * if we're not talking to the authority, send to them
2430          * if the authority has changed while we weren't looking,
2431          * send to new authority
2432          * Otherwise we just have to return an ESTALE
2433          */
2434         if (result == -ESTALE) {
2435                 dout("got ESTALE on request %llu", req->r_tid);
2436                 req->r_resend_mds = -1;
2437                 if (req->r_direct_mode != USE_AUTH_MDS) {
2438                         dout("not using auth, setting for that now");
2439                         req->r_direct_mode = USE_AUTH_MDS;
2440                         __do_request(mdsc, req);
2441                         mutex_unlock(&mdsc->mutex);
2442                         goto out;
2443                 } else  {
2444                         int mds = __choose_mds(mdsc, req);
2445                         if (mds >= 0 && mds != req->r_session->s_mds) {
2446                                 dout("but auth changed, so resending");
2447                                 __do_request(mdsc, req);
2448                                 mutex_unlock(&mdsc->mutex);
2449                                 goto out;
2450                         }
2451                 }
2452                 dout("have to return ESTALE on request %llu", req->r_tid);
2453         }
2454
2455
2456         if (head->safe) {
2457                 req->r_got_safe = true;
2458                 __unregister_request(mdsc, req);
2459
2460                 if (req->r_got_unsafe) {
2461                         /*
2462                          * We already handled the unsafe response, now do the
2463                          * cleanup.  No need to examine the response; the MDS
2464                          * doesn't include any result info in the safe
2465                          * response.  And even if it did, there is nothing
2466                          * useful we could do with a revised return value.
2467                          */
2468                         dout("got safe reply %llu, mds%d\n", tid, mds);
2469                         list_del_init(&req->r_unsafe_item);
2470
2471                         /* last unsafe request during umount? */
2472                         if (mdsc->stopping && !__get_oldest_req(mdsc))
2473                                 complete_all(&mdsc->safe_umount_waiters);
2474                         mutex_unlock(&mdsc->mutex);
2475                         goto out;
2476                 }
2477         } else {
2478                 req->r_got_unsafe = true;
2479                 list_add_tail(&req->r_unsafe_item, &req->r_session->s_unsafe);
2480         }
2481
2482         dout("handle_reply tid %lld result %d\n", tid, result);
2483         rinfo = &req->r_reply_info;
2484         err = parse_reply_info(msg, rinfo, session->s_con.peer_features);
2485         mutex_unlock(&mdsc->mutex);
2486
2487         mutex_lock(&session->s_mutex);
2488         if (err < 0) {
2489                 pr_err("mdsc_handle_reply got corrupt reply mds%d(tid:%lld)\n", mds, tid);
2490                 ceph_msg_dump(msg);
2491                 goto out_err;
2492         }
2493
2494         /* snap trace */
2495         realm = NULL;
2496         if (rinfo->snapblob_len) {
2497                 down_write(&mdsc->snap_rwsem);
2498                 ceph_update_snap_trace(mdsc, rinfo->snapblob,
2499                                 rinfo->snapblob + rinfo->snapblob_len,
2500                                 le32_to_cpu(head->op) == CEPH_MDS_OP_RMSNAP,
2501                                 &realm);
2502                 downgrade_write(&mdsc->snap_rwsem);
2503         } else {
2504                 down_read(&mdsc->snap_rwsem);
2505         }
2506
2507         /* insert trace into our cache */
2508         mutex_lock(&req->r_fill_mutex);
2509         err = ceph_fill_trace(mdsc->fsc->sb, req, req->r_session);
2510         if (err == 0) {
2511                 if (result == 0 && (req->r_op == CEPH_MDS_OP_READDIR ||
2512                                     req->r_op == CEPH_MDS_OP_LSSNAP))
2513                         ceph_readdir_prepopulate(req, req->r_session);
2514                 ceph_unreserve_caps(mdsc, &req->r_caps_reservation);
2515         }
2516         mutex_unlock(&req->r_fill_mutex);
2517
2518         up_read(&mdsc->snap_rwsem);
2519         if (realm)
2520                 ceph_put_snap_realm(mdsc, realm);
2521 out_err:
2522         mutex_lock(&mdsc->mutex);
2523         if (!req->r_aborted) {
2524                 if (err) {
2525                         req->r_err = err;
2526                 } else {
2527                         req->r_reply = msg;
2528                         ceph_msg_get(msg);
2529                         req->r_got_result = true;
2530                 }
2531         } else {
2532                 dout("reply arrived after request %lld was aborted\n", tid);
2533         }
2534         mutex_unlock(&mdsc->mutex);
2535
2536         mutex_unlock(&session->s_mutex);
2537
2538         /* kick calling process */
2539         complete_request(mdsc, req);
2540 out:
2541         ceph_mdsc_put_request(req);
2542         return;
2543 }
2544
2545
2546
2547 /*
2548  * handle mds notification that our request has been forwarded.
2549  */
2550 static void handle_forward(struct ceph_mds_client *mdsc,
2551                            struct ceph_mds_session *session,
2552                            struct ceph_msg *msg)
2553 {
2554         struct ceph_mds_request *req;
2555         u64 tid = le64_to_cpu(msg->hdr.tid);
2556         u32 next_mds;
2557         u32 fwd_seq;
2558         int err = -EINVAL;
2559         void *p = msg->front.iov_base;
2560         void *end = p + msg->front.iov_len;
2561
2562         ceph_decode_need(&p, end, 2*sizeof(u32), bad);
2563         next_mds = ceph_decode_32(&p);
2564         fwd_seq = ceph_decode_32(&p);
2565
2566         mutex_lock(&mdsc->mutex);
2567         req = __lookup_request(mdsc, tid);
2568         if (!req) {
2569                 dout("forward tid %llu to mds%d - req dne\n", tid, next_mds);
2570                 goto out;  /* dup reply? */
2571         }
2572
2573         if (req->r_aborted) {
2574                 dout("forward tid %llu aborted, unregistering\n", tid);
2575                 __unregister_request(mdsc, req);
2576         } else if (fwd_seq <= req->r_num_fwd) {
2577                 dout("forward tid %llu to mds%d - old seq %d <= %d\n",
2578                      tid, next_mds, req->r_num_fwd, fwd_seq);
2579         } else {
2580                 /* resend. forward race not possible; mds would drop */
2581                 dout("forward tid %llu to mds%d (we resend)\n", tid, next_mds);
2582                 BUG_ON(req->r_err);
2583                 BUG_ON(req->r_got_result);
2584                 req->r_attempts = 0;
2585                 req->r_num_fwd = fwd_seq;
2586                 req->r_resend_mds = next_mds;
2587                 put_request_session(req);
2588                 __do_request(mdsc, req);
2589         }
2590         ceph_mdsc_put_request(req);
2591 out:
2592         mutex_unlock(&mdsc->mutex);
2593         return;
2594
2595 bad:
2596         pr_err("mdsc_handle_forward decode error err=%d\n", err);
2597 }
2598
2599 /*
2600  * handle a mds session control message
2601  */
2602 static void handle_session(struct ceph_mds_session *session,
2603                            struct ceph_msg *msg)
2604 {
2605         struct ceph_mds_client *mdsc = session->s_mdsc;
2606         u32 op;
2607         u64 seq;
2608         int mds = session->s_mds;
2609         struct ceph_mds_session_head *h = msg->front.iov_base;
2610         int wake = 0;
2611
2612         /* decode */
2613         if (msg->front.iov_len != sizeof(*h))
2614                 goto bad;
2615         op = le32_to_cpu(h->op);
2616         seq = le64_to_cpu(h->seq);
2617
2618         mutex_lock(&mdsc->mutex);
2619         if (op == CEPH_SESSION_CLOSE)
2620                 __unregister_session(mdsc, session);
2621         /* FIXME: this ttl calculation is generous */
2622         session->s_ttl = jiffies + HZ*mdsc->mdsmap->m_session_autoclose;
2623         mutex_unlock(&mdsc->mutex);
2624
2625         mutex_lock(&session->s_mutex);
2626
2627         dout("handle_session mds%d %s %p state %s seq %llu\n",
2628              mds, ceph_session_op_name(op), session,
2629              ceph_session_state_name(session->s_state), seq);
2630
2631         if (session->s_state == CEPH_MDS_SESSION_HUNG) {
2632                 session->s_state = CEPH_MDS_SESSION_OPEN;
2633                 pr_info("mds%d came back\n", session->s_mds);
2634         }
2635
2636         switch (op) {
2637         case CEPH_SESSION_OPEN:
2638                 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
2639                         pr_info("mds%d reconnect success\n", session->s_mds);
2640                 session->s_state = CEPH_MDS_SESSION_OPEN;
2641                 renewed_caps(mdsc, session, 0);
2642                 wake = 1;
2643                 if (mdsc->stopping)
2644                         __close_session(mdsc, session);
2645                 break;
2646
2647         case CEPH_SESSION_RENEWCAPS:
2648                 if (session->s_renew_seq == seq)
2649                         renewed_caps(mdsc, session, 1);
2650                 break;
2651
2652         case CEPH_SESSION_CLOSE:
2653                 if (session->s_state == CEPH_MDS_SESSION_RECONNECTING)
2654                         pr_info("mds%d reconnect denied\n", session->s_mds);
2655                 cleanup_session_requests(mdsc, session);
2656                 remove_session_caps(session);
2657                 wake = 2; /* for good measure */
2658                 wake_up_all(&mdsc->session_close_wq);
2659                 break;
2660
2661         case CEPH_SESSION_STALE:
2662                 pr_info("mds%d caps went stale, renewing\n",
2663                         session->s_mds);
2664                 spin_lock(&session->s_gen_ttl_lock);
2665                 session->s_cap_gen++;
2666                 session->s_cap_ttl = jiffies - 1;
2667                 spin_unlock(&session->s_gen_ttl_lock);
2668                 send_renew_caps(mdsc, session);
2669                 break;
2670
2671         case CEPH_SESSION_RECALL_STATE:
2672                 trim_caps(mdsc, session, le32_to_cpu(h->max_caps));
2673                 break;
2674
2675         case CEPH_SESSION_FLUSHMSG:
2676                 send_flushmsg_ack(mdsc, session, seq);
2677                 break;
2678
2679         case CEPH_SESSION_FORCE_RO:
2680                 dout("force_session_readonly %p\n", session);
2681                 spin_lock(&session->s_cap_lock);
2682                 session->s_readonly = true;
2683                 spin_unlock(&session->s_cap_lock);
2684                 wake_up_session_caps(session, 0);
2685                 break;
2686
2687         default:
2688                 pr_err("mdsc_handle_session bad op %d mds%d\n", op, mds);
2689                 WARN_ON(1);
2690         }
2691
2692         mutex_unlock(&session->s_mutex);
2693         if (wake) {
2694                 mutex_lock(&mdsc->mutex);
2695                 __wake_requests(mdsc, &session->s_waiting);
2696                 if (wake == 2)
2697                         kick_requests(mdsc, mds);
2698                 mutex_unlock(&mdsc->mutex);
2699         }
2700         return;
2701
2702 bad:
2703         pr_err("mdsc_handle_session corrupt message mds%d len %d\n", mds,
2704                (int)msg->front.iov_len);
2705         ceph_msg_dump(msg);
2706         return;
2707 }
2708
2709
2710 /*
2711  * called under session->mutex.
2712  */
2713 static void replay_unsafe_requests(struct ceph_mds_client *mdsc,
2714                                    struct ceph_mds_session *session)
2715 {
2716         struct ceph_mds_request *req, *nreq;
2717         struct rb_node *p;
2718         int err;
2719
2720         dout("replay_unsafe_requests mds%d\n", session->s_mds);
2721
2722         mutex_lock(&mdsc->mutex);
2723         list_for_each_entry_safe(req, nreq, &session->s_unsafe, r_unsafe_item) {
2724                 err = __prepare_send_request(mdsc, req, session->s_mds, true);
2725                 if (!err) {
2726                         ceph_msg_get(req->r_request);
2727                         ceph_con_send(&session->s_con, req->r_request);
2728                 }
2729         }
2730
2731         /*
2732          * also re-send old requests when MDS enters reconnect stage. So that MDS
2733          * can process completed request in clientreplay stage.
2734          */
2735         p = rb_first(&mdsc->request_tree);
2736         while (p) {
2737                 req = rb_entry(p, struct ceph_mds_request, r_node);
2738                 p = rb_next(p);
2739                 if (req->r_got_unsafe)
2740                         continue;
2741                 if (req->r_attempts == 0)
2742                         continue; /* only old requests */
2743                 if (req->r_session &&
2744                     req->r_session->s_mds == session->s_mds) {
2745                         err = __prepare_send_request(mdsc, req,
2746                                                      session->s_mds, true);
2747                         if (!err) {
2748                                 ceph_msg_get(req->r_request);
2749                                 ceph_con_send(&session->s_con, req->r_request);
2750                         }
2751                 }
2752         }
2753         mutex_unlock(&mdsc->mutex);
2754 }
2755
2756 /*
2757  * Encode information about a cap for a reconnect with the MDS.
2758  */
2759 static int encode_caps_cb(struct inode *inode, struct ceph_cap *cap,
2760                           void *arg)
2761 {
2762         union {
2763                 struct ceph_mds_cap_reconnect v2;
2764                 struct ceph_mds_cap_reconnect_v1 v1;
2765         } rec;
2766         size_t reclen;
2767         struct ceph_inode_info *ci;
2768         struct ceph_reconnect_state *recon_state = arg;
2769         struct ceph_pagelist *pagelist = recon_state->pagelist;
2770         char *path;
2771         int pathlen, err;
2772         u64 pathbase;
2773         struct dentry *dentry;
2774
2775         ci = cap->ci;
2776
2777         dout(" adding %p ino %llx.%llx cap %p %lld %s\n",
2778              inode, ceph_vinop(inode), cap, cap->cap_id,
2779              ceph_cap_string(cap->issued));
2780         err = ceph_pagelist_encode_64(pagelist, ceph_ino(inode));
2781         if (err)
2782                 return err;
2783
2784         dentry = d_find_alias(inode);
2785         if (dentry) {
2786                 path = ceph_mdsc_build_path(dentry, &pathlen, &pathbase, 0);
2787                 if (IS_ERR(path)) {
2788                         err = PTR_ERR(path);
2789                         goto out_dput;
2790                 }
2791         } else {
2792                 path = NULL;
2793                 pathlen = 0;
2794         }
2795         err = ceph_pagelist_encode_string(pagelist, path, pathlen);
2796         if (err)
2797                 goto out_free;
2798
2799         spin_lock(&ci->i_ceph_lock);
2800         cap->seq = 0;        /* reset cap seq */
2801         cap->issue_seq = 0;  /* and issue_seq */
2802         cap->mseq = 0;       /* and migrate_seq */
2803         cap->cap_gen = cap->session->s_cap_gen;
2804
2805         if (recon_state->flock) {
2806                 rec.v2.cap_id = cpu_to_le64(cap->cap_id);
2807                 rec.v2.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
2808                 rec.v2.issued = cpu_to_le32(cap->issued);
2809                 rec.v2.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
2810                 rec.v2.pathbase = cpu_to_le64(pathbase);
2811                 rec.v2.flock_len = 0;
2812                 reclen = sizeof(rec.v2);
2813         } else {
2814                 rec.v1.cap_id = cpu_to_le64(cap->cap_id);
2815                 rec.v1.wanted = cpu_to_le32(__ceph_caps_wanted(ci));
2816                 rec.v1.issued = cpu_to_le32(cap->issued);
2817                 rec.v1.size = cpu_to_le64(inode->i_size);
2818                 ceph_encode_timespec(&rec.v1.mtime, &inode->i_mtime);
2819                 ceph_encode_timespec(&rec.v1.atime, &inode->i_atime);
2820                 rec.v1.snaprealm = cpu_to_le64(ci->i_snap_realm->ino);
2821                 rec.v1.pathbase = cpu_to_le64(pathbase);
2822                 reclen = sizeof(rec.v1);
2823         }
2824         spin_unlock(&ci->i_ceph_lock);
2825
2826         if (recon_state->flock) {
2827                 int num_fcntl_locks, num_flock_locks;
2828                 struct ceph_filelock *flocks;
2829
2830 encode_again:
2831                 ceph_count_locks(inode, &num_fcntl_locks, &num_flock_locks);
2832                 flocks = kmalloc((num_fcntl_locks+num_flock_locks) *
2833                                  sizeof(struct ceph_filelock), GFP_NOFS);
2834                 if (!flocks) {
2835                         err = -ENOMEM;
2836                         goto out_free;
2837                 }
2838                 err = ceph_encode_locks_to_buffer(inode, flocks,
2839                                                   num_fcntl_locks,
2840                                                   num_flock_locks);
2841                 if (err) {
2842                         kfree(flocks);
2843                         if (err == -ENOSPC)
2844                                 goto encode_again;
2845                         goto out_free;
2846                 }
2847                 /*
2848                  * number of encoded locks is stable, so copy to pagelist
2849                  */
2850                 rec.v2.flock_len = cpu_to_le32(2*sizeof(u32) +
2851                                     (num_fcntl_locks+num_flock_locks) *
2852                                     sizeof(struct ceph_filelock));
2853                 err = ceph_pagelist_append(pagelist, &rec, reclen);
2854                 if (!err)
2855                         err = ceph_locks_to_pagelist(flocks, pagelist,
2856                                                      num_fcntl_locks,
2857                                                      num_flock_locks);
2858                 kfree(flocks);
2859         } else {
2860                 err = ceph_pagelist_append(pagelist, &rec, reclen);
2861         }
2862
2863         recon_state->nr_caps++;
2864 out_free:
2865         kfree(path);
2866 out_dput:
2867         dput(dentry);
2868         return err;
2869 }
2870
2871
2872 /*
2873  * If an MDS fails and recovers, clients need to reconnect in order to
2874  * reestablish shared state.  This includes all caps issued through
2875  * this session _and_ the snap_realm hierarchy.  Because it's not
2876  * clear which snap realms the mds cares about, we send everything we
2877  * know about.. that ensures we'll then get any new info the
2878  * recovering MDS might have.
2879  *
2880  * This is a relatively heavyweight operation, but it's rare.
2881  *
2882  * called with mdsc->mutex held.
2883  */
2884 static void send_mds_reconnect(struct ceph_mds_client *mdsc,
2885                                struct ceph_mds_session *session)
2886 {
2887         struct ceph_msg *reply;
2888         struct rb_node *p;
2889         int mds = session->s_mds;
2890         int err = -ENOMEM;
2891         int s_nr_caps;
2892         struct ceph_pagelist *pagelist;
2893         struct ceph_reconnect_state recon_state;
2894
2895         pr_info("mds%d reconnect start\n", mds);
2896
2897         pagelist = kmalloc(sizeof(*pagelist), GFP_NOFS);
2898         if (!pagelist)
2899                 goto fail_nopagelist;
2900         ceph_pagelist_init(pagelist);
2901
2902         reply = ceph_msg_new(CEPH_MSG_CLIENT_RECONNECT, 0, GFP_NOFS, false);
2903         if (!reply)
2904                 goto fail_nomsg;
2905
2906         mutex_lock(&session->s_mutex);
2907         session->s_state = CEPH_MDS_SESSION_RECONNECTING;
2908         session->s_seq = 0;
2909
2910         dout("session %p state %s\n", session,
2911              ceph_session_state_name(session->s_state));
2912
2913         spin_lock(&session->s_gen_ttl_lock);
2914         session->s_cap_gen++;
2915         spin_unlock(&session->s_gen_ttl_lock);
2916
2917         spin_lock(&session->s_cap_lock);
2918         /* don't know if session is readonly */
2919         session->s_readonly = 0;
2920         /*
2921          * notify __ceph_remove_cap() that we are composing cap reconnect.
2922          * If a cap get released before being added to the cap reconnect,
2923          * __ceph_remove_cap() should skip queuing cap release.
2924          */
2925         session->s_cap_reconnect = 1;
2926         /* drop old cap expires; we're about to reestablish that state */
2927         cleanup_cap_releases(mdsc, session);
2928
2929         /* trim unused caps to reduce MDS's cache rejoin time */
2930         if (mdsc->fsc->sb->s_root)
2931                 shrink_dcache_parent(mdsc->fsc->sb->s_root);
2932
2933         ceph_con_close(&session->s_con);
2934         ceph_con_open(&session->s_con,
2935                       CEPH_ENTITY_TYPE_MDS, mds,
2936                       ceph_mdsmap_get_addr(mdsc->mdsmap, mds));
2937
2938         /* replay unsafe requests */
2939         replay_unsafe_requests(mdsc, session);
2940
2941         down_read(&mdsc->snap_rwsem);
2942
2943         /* traverse this session's caps */
2944         s_nr_caps = session->s_nr_caps;
2945         err = ceph_pagelist_encode_32(pagelist, s_nr_caps);
2946         if (err)
2947                 goto fail;
2948
2949         recon_state.nr_caps = 0;
2950         recon_state.pagelist = pagelist;
2951         recon_state.flock = session->s_con.peer_features & CEPH_FEATURE_FLOCK;
2952         err = iterate_session_caps(session, encode_caps_cb, &recon_state);
2953         if (err < 0)
2954                 goto fail;
2955
2956         spin_lock(&session->s_cap_lock);
2957         session->s_cap_reconnect = 0;
2958         spin_unlock(&session->s_cap_lock);
2959
2960         /*
2961          * snaprealms.  we provide mds with the ino, seq (version), and
2962          * parent for all of our realms.  If the mds has any newer info,
2963          * it will tell us.
2964          */
2965         for (p = rb_first(&mdsc->snap_realms); p; p = rb_next(p)) {
2966                 struct ceph_snap_realm *realm =
2967                         rb_entry(p, struct ceph_snap_realm, node);
2968                 struct ceph_mds_snaprealm_reconnect sr_rec;
2969
2970                 dout(" adding snap realm %llx seq %lld parent %llx\n",
2971                      realm->ino, realm->seq, realm->parent_ino);
2972                 sr_rec.ino = cpu_to_le64(realm->ino);
2973                 sr_rec.seq = cpu_to_le64(realm->seq);
2974                 sr_rec.parent = cpu_to_le64(realm->parent_ino);
2975                 err = ceph_pagelist_append(pagelist, &sr_rec, sizeof(sr_rec));
2976                 if (err)
2977                         goto fail;
2978         }
2979
2980         if (recon_state.flock)
2981                 reply->hdr.version = cpu_to_le16(2);
2982
2983         /* raced with cap release? */
2984         if (s_nr_caps != recon_state.nr_caps) {
2985                 struct page *page = list_first_entry(&pagelist->head,
2986                                                      struct page, lru);
2987                 __le32 *addr = kmap_atomic(page);
2988                 *addr = cpu_to_le32(recon_state.nr_caps);
2989                 kunmap_atomic(addr);
2990         }
2991
2992         reply->hdr.data_len = cpu_to_le32(pagelist->length);
2993         ceph_msg_data_add_pagelist(reply, pagelist);
2994
2995         ceph_early_kick_flushing_caps(mdsc, session);
2996
2997         ceph_con_send(&session->s_con, reply);
2998
2999         mutex_unlock(&session->s_mutex);
3000
3001         mutex_lock(&mdsc->mutex);
3002         __wake_requests(mdsc, &session->s_waiting);
3003         mutex_unlock(&mdsc->mutex);
3004
3005         up_read(&mdsc->snap_rwsem);
3006         return;
3007
3008 fail:
3009         ceph_msg_put(reply);
3010         up_read(&mdsc->snap_rwsem);
3011         mutex_unlock(&session->s_mutex);
3012 fail_nomsg:
3013         ceph_pagelist_release(pagelist);
3014 fail_nopagelist:
3015         pr_err("error %d preparing reconnect for mds%d\n", err, mds);
3016         return;
3017 }
3018
3019
3020 /*
3021  * compare old and new mdsmaps, kicking requests
3022  * and closing out old connections as necessary
3023  *
3024  * called under mdsc->mutex.
3025  */
3026 static void check_new_map(struct ceph_mds_client *mdsc,
3027                           struct ceph_mdsmap *newmap,
3028                           struct ceph_mdsmap *oldmap)
3029 {
3030         int i;
3031         int oldstate, newstate;
3032         struct ceph_mds_session *s;
3033
3034         dout("check_new_map new %u old %u\n",
3035              newmap->m_epoch, oldmap->m_epoch);
3036
3037         for (i = 0; i < oldmap->m_max_mds && i < mdsc->max_sessions; i++) {
3038                 if (mdsc->sessions[i] == NULL)
3039                         continue;
3040                 s = mdsc->sessions[i];
3041                 oldstate = ceph_mdsmap_get_state(oldmap, i);
3042                 newstate = ceph_mdsmap_get_state(newmap, i);
3043
3044                 dout("check_new_map mds%d state %s%s -> %s%s (session %s)\n",
3045                      i, ceph_mds_state_name(oldstate),
3046                      ceph_mdsmap_is_laggy(oldmap, i) ? " (laggy)" : "",
3047                      ceph_mds_state_name(newstate),
3048                      ceph_mdsmap_is_laggy(newmap, i) ? " (laggy)" : "",
3049                      ceph_session_state_name(s->s_state));
3050
3051                 if (i >= newmap->m_max_mds ||
3052                     memcmp(ceph_mdsmap_get_addr(oldmap, i),
3053                            ceph_mdsmap_get_addr(newmap, i),
3054                            sizeof(struct ceph_entity_addr))) {
3055                         if (s->s_state == CEPH_MDS_SESSION_OPENING) {
3056                                 /* the session never opened, just close it
3057                                  * out now */
3058                                 __wake_requests(mdsc, &s->s_waiting);
3059                                 __unregister_session(mdsc, s);
3060                         } else {
3061                                 /* just close it */
3062                                 mutex_unlock(&mdsc->mutex);
3063                                 mutex_lock(&s->s_mutex);
3064                                 mutex_lock(&mdsc->mutex);
3065                                 ceph_con_close(&s->s_con);
3066                                 mutex_unlock(&s->s_mutex);
3067                                 s->s_state = CEPH_MDS_SESSION_RESTARTING;
3068                         }
3069                 } else if (oldstate == newstate) {
3070                         continue;  /* nothing new with this mds */
3071                 }
3072
3073                 /*
3074                  * send reconnect?
3075                  */
3076                 if (s->s_state == CEPH_MDS_SESSION_RESTARTING &&
3077                     newstate >= CEPH_MDS_STATE_RECONNECT) {
3078                         mutex_unlock(&mdsc->mutex);
3079                         send_mds_reconnect(mdsc, s);
3080                         mutex_lock(&mdsc->mutex);
3081                 }
3082
3083                 /*
3084                  * kick request on any mds that has gone active.
3085                  */
3086                 if (oldstate < CEPH_MDS_STATE_ACTIVE &&
3087                     newstate >= CEPH_MDS_STATE_ACTIVE) {
3088                         if (oldstate != CEPH_MDS_STATE_CREATING &&
3089                             oldstate != CEPH_MDS_STATE_STARTING)
3090                                 pr_info("mds%d recovery completed\n", s->s_mds);
3091                         kick_requests(mdsc, i);
3092                         ceph_kick_flushing_caps(mdsc, s);
3093                         wake_up_session_caps(s, 1);
3094                 }
3095         }
3096
3097         for (i = 0; i < newmap->m_max_mds && i < mdsc->max_sessions; i++) {
3098                 s = mdsc->sessions[i];
3099                 if (!s)
3100                         continue;
3101                 if (!ceph_mdsmap_is_laggy(newmap, i))
3102                         continue;
3103                 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
3104                     s->s_state == CEPH_MDS_SESSION_HUNG ||
3105                     s->s_state == CEPH_MDS_SESSION_CLOSING) {
3106                         dout(" connecting to export targets of laggy mds%d\n",
3107                              i);
3108                         __open_export_target_sessions(mdsc, s);
3109                 }
3110         }
3111 }
3112
3113
3114
3115 /*
3116  * leases
3117  */
3118
3119 /*
3120  * caller must hold session s_mutex, dentry->d_lock
3121  */
3122 void __ceph_mdsc_drop_dentry_lease(struct dentry *dentry)
3123 {
3124         struct ceph_dentry_info *di = ceph_dentry(dentry);
3125
3126         ceph_put_mds_session(di->lease_session);
3127         di->lease_session = NULL;
3128 }
3129
3130 static void handle_lease(struct ceph_mds_client *mdsc,
3131                          struct ceph_mds_session *session,
3132                          struct ceph_msg *msg)
3133 {
3134         struct super_block *sb = mdsc->fsc->sb;
3135         struct inode *inode;
3136         struct dentry *parent, *dentry;
3137         struct ceph_dentry_info *di;
3138         int mds = session->s_mds;
3139         struct ceph_mds_lease *h = msg->front.iov_base;
3140         u32 seq;
3141         struct ceph_vino vino;
3142         struct qstr dname;
3143         int release = 0;
3144
3145         dout("handle_lease from mds%d\n", mds);
3146
3147         /* decode */
3148         if (msg->front.iov_len < sizeof(*h) + sizeof(u32))
3149                 goto bad;
3150         vino.ino = le64_to_cpu(h->ino);
3151         vino.snap = CEPH_NOSNAP;
3152         seq = le32_to_cpu(h->seq);
3153         dname.name = (void *)h + sizeof(*h) + sizeof(u32);
3154         dname.len = msg->front.iov_len - sizeof(*h) - sizeof(u32);
3155         if (dname.len != get_unaligned_le32(h+1))
3156                 goto bad;
3157
3158         /* lookup inode */
3159         inode = ceph_find_inode(sb, vino);
3160         dout("handle_lease %s, ino %llx %p %.*s\n",
3161              ceph_lease_op_name(h->action), vino.ino, inode,
3162              dname.len, dname.name);
3163
3164         mutex_lock(&session->s_mutex);
3165         session->s_seq++;
3166
3167         if (inode == NULL) {
3168                 dout("handle_lease no inode %llx\n", vino.ino);
3169                 goto release;
3170         }
3171
3172         /* dentry */
3173         parent = d_find_alias(inode);
3174         if (!parent) {
3175                 dout("no parent dentry on inode %p\n", inode);
3176                 WARN_ON(1);
3177                 goto release;  /* hrm... */
3178         }
3179         dname.hash = full_name_hash(dname.name, dname.len);
3180         dentry = d_lookup(parent, &dname);
3181         dput(parent);
3182         if (!dentry)
3183                 goto release;
3184
3185         spin_lock(&dentry->d_lock);
3186         di = ceph_dentry(dentry);
3187         switch (h->action) {
3188         case CEPH_MDS_LEASE_REVOKE:
3189                 if (di->lease_session == session) {
3190                         if (ceph_seq_cmp(di->lease_seq, seq) > 0)
3191                                 h->seq = cpu_to_le32(di->lease_seq);
3192                         __ceph_mdsc_drop_dentry_lease(dentry);
3193                 }
3194                 release = 1;
3195                 break;
3196
3197         case CEPH_MDS_LEASE_RENEW:
3198                 if (di->lease_session == session &&
3199                     di->lease_gen == session->s_cap_gen &&
3200                     di->lease_renew_from &&
3201                     di->lease_renew_after == 0) {
3202                         unsigned long duration =
3203                                 msecs_to_jiffies(le32_to_cpu(h->duration_ms));
3204
3205                         di->lease_seq = seq;
3206                         dentry->d_time = di->lease_renew_from + duration;
3207                         di->lease_renew_after = di->lease_renew_from +
3208                                 (duration >> 1);
3209                         di->lease_renew_from = 0;
3210                 }
3211                 break;
3212         }
3213         spin_unlock(&dentry->d_lock);
3214         dput(dentry);
3215
3216         if (!release)
3217                 goto out;
3218
3219 release:
3220         /* let's just reuse the same message */
3221         h->action = CEPH_MDS_LEASE_REVOKE_ACK;
3222         ceph_msg_get(msg);
3223         ceph_con_send(&session->s_con, msg);
3224
3225 out:
3226         iput(inode);
3227         mutex_unlock(&session->s_mutex);
3228         return;
3229
3230 bad:
3231         pr_err("corrupt lease message\n");
3232         ceph_msg_dump(msg);
3233 }
3234
3235 void ceph_mdsc_lease_send_msg(struct ceph_mds_session *session,
3236                               struct inode *inode,
3237                               struct dentry *dentry, char action,
3238                               u32 seq)
3239 {
3240         struct ceph_msg *msg;
3241         struct ceph_mds_lease *lease;
3242         int len = sizeof(*lease) + sizeof(u32);
3243         int dnamelen = 0;
3244
3245         dout("lease_send_msg inode %p dentry %p %s to mds%d\n",
3246              inode, dentry, ceph_lease_op_name(action), session->s_mds);
3247         dnamelen = dentry->d_name.len;
3248         len += dnamelen;
3249
3250         msg = ceph_msg_new(CEPH_MSG_CLIENT_LEASE, len, GFP_NOFS, false);
3251         if (!msg)
3252                 return;
3253         lease = msg->front.iov_base;
3254         lease->action = action;
3255         lease->ino = cpu_to_le64(ceph_vino(inode).ino);
3256         lease->first = lease->last = cpu_to_le64(ceph_vino(inode).snap);
3257         lease->seq = cpu_to_le32(seq);
3258         put_unaligned_le32(dnamelen, lease + 1);
3259         memcpy((void *)(lease + 1) + 4, dentry->d_name.name, dnamelen);
3260
3261         /*
3262          * if this is a preemptive lease RELEASE, no need to
3263          * flush request stream, since the actual request will
3264          * soon follow.
3265          */
3266         msg->more_to_follow = (action == CEPH_MDS_LEASE_RELEASE);
3267
3268         ceph_con_send(&session->s_con, msg);
3269 }
3270
3271 /*
3272  * Preemptively release a lease we expect to invalidate anyway.
3273  * Pass @inode always, @dentry is optional.
3274  */
3275 void ceph_mdsc_lease_release(struct ceph_mds_client *mdsc, struct inode *inode,
3276                              struct dentry *dentry)
3277 {
3278         struct ceph_dentry_info *di;
3279         struct ceph_mds_session *session;
3280         u32 seq;
3281
3282         BUG_ON(inode == NULL);
3283         BUG_ON(dentry == NULL);
3284
3285         /* is dentry lease valid? */
3286         spin_lock(&dentry->d_lock);
3287         di = ceph_dentry(dentry);
3288         if (!di || !di->lease_session ||
3289             di->lease_session->s_mds < 0 ||
3290             di->lease_gen != di->lease_session->s_cap_gen ||
3291             !time_before(jiffies, dentry->d_time)) {
3292                 dout("lease_release inode %p dentry %p -- "
3293                      "no lease\n",
3294                      inode, dentry);
3295                 spin_unlock(&dentry->d_lock);
3296                 return;
3297         }
3298
3299         /* we do have a lease on this dentry; note mds and seq */
3300         session = ceph_get_mds_session(di->lease_session);
3301         seq = di->lease_seq;
3302         __ceph_mdsc_drop_dentry_lease(dentry);
3303         spin_unlock(&dentry->d_lock);
3304
3305         dout("lease_release inode %p dentry %p to mds%d\n",
3306              inode, dentry, session->s_mds);
3307         ceph_mdsc_lease_send_msg(session, inode, dentry,
3308                                  CEPH_MDS_LEASE_RELEASE, seq);
3309         ceph_put_mds_session(session);
3310 }
3311
3312 /*
3313  * drop all leases (and dentry refs) in preparation for umount
3314  */
3315 static void drop_leases(struct ceph_mds_client *mdsc)
3316 {
3317         int i;
3318
3319         dout("drop_leases\n");
3320         mutex_lock(&mdsc->mutex);
3321         for (i = 0; i < mdsc->max_sessions; i++) {
3322                 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
3323                 if (!s)
3324                         continue;
3325                 mutex_unlock(&mdsc->mutex);
3326                 mutex_lock(&s->s_mutex);
3327                 mutex_unlock(&s->s_mutex);
3328                 ceph_put_mds_session(s);
3329                 mutex_lock(&mdsc->mutex);
3330         }
3331         mutex_unlock(&mdsc->mutex);
3332 }
3333
3334
3335
3336 /*
3337  * delayed work -- periodically trim expired leases, renew caps with mds
3338  */
3339 static void schedule_delayed(struct ceph_mds_client *mdsc)
3340 {
3341         int delay = 5;
3342         unsigned hz = round_jiffies_relative(HZ * delay);
3343         schedule_delayed_work(&mdsc->delayed_work, hz);
3344 }
3345
3346 static void delayed_work(struct work_struct *work)
3347 {
3348         int i;
3349         struct ceph_mds_client *mdsc =
3350                 container_of(work, struct ceph_mds_client, delayed_work.work);
3351         int renew_interval;
3352         int renew_caps;
3353
3354         dout("mdsc delayed_work\n");
3355         ceph_check_delayed_caps(mdsc);
3356
3357         mutex_lock(&mdsc->mutex);
3358         renew_interval = mdsc->mdsmap->m_session_timeout >> 2;
3359         renew_caps = time_after_eq(jiffies, HZ*renew_interval +
3360                                    mdsc->last_renew_caps);
3361         if (renew_caps)
3362                 mdsc->last_renew_caps = jiffies;
3363
3364         for (i = 0; i < mdsc->max_sessions; i++) {
3365                 struct ceph_mds_session *s = __ceph_lookup_mds_session(mdsc, i);
3366                 if (s == NULL)
3367                         continue;
3368                 if (s->s_state == CEPH_MDS_SESSION_CLOSING) {
3369                         dout("resending session close request for mds%d\n",
3370                              s->s_mds);
3371                         request_close_session(mdsc, s);
3372                         ceph_put_mds_session(s);
3373                         continue;
3374                 }
3375                 if (s->s_ttl && time_after(jiffies, s->s_ttl)) {
3376                         if (s->s_state == CEPH_MDS_SESSION_OPEN) {
3377                                 s->s_state = CEPH_MDS_SESSION_HUNG;
3378                                 pr_info("mds%d hung\n", s->s_mds);
3379                         }
3380                 }
3381                 if (s->s_state < CEPH_MDS_SESSION_OPEN) {
3382                         /* this mds is failed or recovering, just wait */
3383                         ceph_put_mds_session(s);
3384                         continue;
3385                 }
3386                 mutex_unlock(&mdsc->mutex);
3387
3388                 mutex_lock(&s->s_mutex);
3389                 if (renew_caps)
3390                         send_renew_caps(mdsc, s);
3391                 else
3392                         ceph_con_keepalive(&s->s_con);
3393                 if (s->s_state == CEPH_MDS_SESSION_OPEN ||
3394                     s->s_state == CEPH_MDS_SESSION_HUNG)
3395                         ceph_send_cap_releases(mdsc, s);
3396                 mutex_unlock(&s->s_mutex);
3397                 ceph_put_mds_session(s);
3398
3399                 mutex_lock(&mdsc->mutex);
3400         }
3401         mutex_unlock(&mdsc->mutex);
3402
3403         schedule_delayed(mdsc);
3404 }
3405
3406 int ceph_mdsc_init(struct ceph_fs_client *fsc)
3407
3408 {
3409         struct ceph_mds_client *mdsc;
3410
3411         mdsc = kzalloc(sizeof(struct ceph_mds_client), GFP_NOFS);
3412         if (!mdsc)
3413                 return -ENOMEM;
3414         mdsc->fsc = fsc;
3415         fsc->mdsc = mdsc;
3416         mutex_init(&mdsc->mutex);
3417         mdsc->mdsmap = kzalloc(sizeof(*mdsc->mdsmap), GFP_NOFS);
3418         if (mdsc->mdsmap == NULL) {
3419                 kfree(mdsc);
3420                 return -ENOMEM;
3421         }
3422
3423         init_completion(&mdsc->safe_umount_waiters);
3424         init_waitqueue_head(&mdsc->session_close_wq);
3425         INIT_LIST_HEAD(&mdsc->waiting_for_map);
3426         mdsc->sessions = NULL;
3427         atomic_set(&mdsc->num_sessions, 0);
3428         mdsc->max_sessions = 0;
3429         mdsc->stopping = 0;
3430         mdsc->last_snap_seq = 0;
3431         init_rwsem(&mdsc->snap_rwsem);
3432         mdsc->snap_realms = RB_ROOT;
3433         INIT_LIST_HEAD(&mdsc->snap_empty);
3434         spin_lock_init(&mdsc->snap_empty_lock);
3435         mdsc->last_tid = 0;
3436         mdsc->oldest_tid = 0;
3437         mdsc->request_tree = RB_ROOT;
3438         INIT_DELAYED_WORK(&mdsc->delayed_work, delayed_work);
3439         mdsc->last_renew_caps = jiffies;
3440         INIT_LIST_HEAD(&mdsc->cap_delay_list);
3441         spin_lock_init(&mdsc->cap_delay_lock);
3442         INIT_LIST_HEAD(&mdsc->snap_flush_list);
3443         spin_lock_init(&mdsc->snap_flush_lock);
3444         mdsc->last_cap_flush_tid = 1;
3445         mdsc->cap_flush_tree = RB_ROOT;
3446         INIT_LIST_HEAD(&mdsc->cap_dirty);
3447         INIT_LIST_HEAD(&mdsc->cap_dirty_migrating);
3448         mdsc->num_cap_flushing = 0;
3449         spin_lock_init(&mdsc->cap_dirty_lock);
3450         init_waitqueue_head(&mdsc->cap_flushing_wq);
3451         spin_lock_init(&mdsc->dentry_lru_lock);
3452         INIT_LIST_HEAD(&mdsc->dentry_lru);
3453
3454         ceph_caps_init(mdsc);
3455         ceph_adjust_min_caps(mdsc, fsc->min_caps);
3456
3457         init_rwsem(&mdsc->pool_perm_rwsem);
3458         mdsc->pool_perm_tree = RB_ROOT;
3459
3460         return 0;
3461 }
3462
3463 /*
3464  * Wait for safe replies on open mds requests.  If we time out, drop
3465  * all requests from the tree to avoid dangling dentry refs.
3466  */
3467 static void wait_requests(struct ceph_mds_client *mdsc)
3468 {
3469         struct ceph_options *opts = mdsc->fsc->client->options;
3470         struct ceph_mds_request *req;
3471
3472         mutex_lock(&mdsc->mutex);
3473         if (__get_oldest_req(mdsc)) {
3474                 mutex_unlock(&mdsc->mutex);
3475
3476                 dout("wait_requests waiting for requests\n");
3477                 wait_for_completion_timeout(&mdsc->safe_umount_waiters,
3478                                     ceph_timeout_jiffies(opts->mount_timeout));
3479
3480                 /* tear down remaining requests */
3481                 mutex_lock(&mdsc->mutex);
3482                 while ((req = __get_oldest_req(mdsc))) {
3483                         dout("wait_requests timed out on tid %llu\n",
3484                              req->r_tid);
3485                         __unregister_request(mdsc, req);
3486                 }
3487         }
3488         mutex_unlock(&mdsc->mutex);
3489         dout("wait_requests done\n");
3490 }
3491
3492 /*
3493  * called before mount is ro, and before dentries are torn down.
3494  * (hmm, does this still race with new lookups?)
3495  */
3496 void ceph_mdsc_pre_umount(struct ceph_mds_client *mdsc)
3497 {
3498         dout("pre_umount\n");
3499         mdsc->stopping = 1;
3500
3501         drop_leases(mdsc);
3502         ceph_flush_dirty_caps(mdsc);
3503         wait_requests(mdsc);
3504
3505         /*
3506          * wait for reply handlers to drop their request refs and
3507          * their inode/dcache refs
3508          */
3509         ceph_msgr_flush();
3510 }
3511
3512 /*
3513  * wait for all write mds requests to flush.
3514  */
3515 static void wait_unsafe_requests(struct ceph_mds_client *mdsc, u64 want_tid)
3516 {
3517         struct ceph_mds_request *req = NULL, *nextreq;
3518         struct rb_node *n;
3519
3520         mutex_lock(&mdsc->mutex);
3521         dout("wait_unsafe_requests want %lld\n", want_tid);
3522 restart:
3523         req = __get_oldest_req(mdsc);
3524         while (req && req->r_tid <= want_tid) {
3525                 /* find next request */
3526                 n = rb_next(&req->r_node);
3527                 if (n)
3528                         nextreq = rb_entry(n, struct ceph_mds_request, r_node);
3529                 else
3530                         nextreq = NULL;
3531                 if (req->r_op != CEPH_MDS_OP_SETFILELOCK &&
3532                     (req->r_op & CEPH_MDS_OP_WRITE)) {
3533                         /* write op */
3534                         ceph_mdsc_get_request(req);
3535                         if (nextreq)
3536                                 ceph_mdsc_get_request(nextreq);
3537                         mutex_unlock(&mdsc->mutex);
3538                         dout("wait_unsafe_requests  wait on %llu (want %llu)\n",
3539                              req->r_tid, want_tid);
3540                         wait_for_completion(&req->r_safe_completion);
3541                         mutex_lock(&mdsc->mutex);
3542                         ceph_mdsc_put_request(req);
3543                         if (!nextreq)
3544                                 break;  /* next dne before, so we're done! */
3545                         if (RB_EMPTY_NODE(&nextreq->r_node)) {
3546                                 /* next request was removed from tree */
3547                                 ceph_mdsc_put_request(nextreq);
3548                                 goto restart;
3549                         }
3550                         ceph_mdsc_put_request(nextreq);  /* won't go away */
3551                 }
3552                 req = nextreq;
3553         }
3554         mutex_unlock(&mdsc->mutex);
3555         dout("wait_unsafe_requests done\n");
3556 }
3557
3558 void ceph_mdsc_sync(struct ceph_mds_client *mdsc)
3559 {
3560         u64 want_tid, want_flush, want_snap;
3561
3562         if (ACCESS_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
3563                 return;
3564
3565         dout("sync\n");
3566         mutex_lock(&mdsc->mutex);
3567         want_tid = mdsc->last_tid;
3568         mutex_unlock(&mdsc->mutex);
3569
3570         ceph_flush_dirty_caps(mdsc);
3571         spin_lock(&mdsc->cap_dirty_lock);
3572         want_flush = mdsc->last_cap_flush_tid;
3573         spin_unlock(&mdsc->cap_dirty_lock);
3574
3575         down_read(&mdsc->snap_rwsem);
3576         want_snap = mdsc->last_snap_seq;
3577         up_read(&mdsc->snap_rwsem);
3578
3579         dout("sync want tid %lld flush_seq %lld snap_seq %lld\n",
3580              want_tid, want_flush, want_snap);
3581
3582         wait_unsafe_requests(mdsc, want_tid);
3583         wait_caps_flush(mdsc, want_flush, want_snap);
3584 }
3585
3586 /*
3587  * true if all sessions are closed, or we force unmount
3588  */
3589 static bool done_closing_sessions(struct ceph_mds_client *mdsc)
3590 {
3591         if (ACCESS_ONCE(mdsc->fsc->mount_state) == CEPH_MOUNT_SHUTDOWN)
3592                 return true;
3593         return atomic_read(&mdsc->num_sessions) == 0;
3594 }
3595
3596 /*
3597  * called after sb is ro.
3598  */
3599 void ceph_mdsc_close_sessions(struct ceph_mds_client *mdsc)
3600 {
3601         struct ceph_options *opts = mdsc->fsc->client->options;
3602         struct ceph_mds_session *session;
3603         int i;
3604
3605         dout("close_sessions\n");
3606
3607         /* close sessions */
3608         mutex_lock(&mdsc->mutex);
3609         for (i = 0; i < mdsc->max_sessions; i++) {
3610                 session = __ceph_lookup_mds_session(mdsc, i);
3611                 if (!session)
3612                         continue;
3613                 mutex_unlock(&mdsc->mutex);
3614                 mutex_lock(&session->s_mutex);
3615                 __close_session(mdsc, session);
3616                 mutex_unlock(&session->s_mutex);
3617                 ceph_put_mds_session(session);
3618                 mutex_lock(&mdsc->mutex);
3619         }
3620         mutex_unlock(&mdsc->mutex);
3621
3622         dout("waiting for sessions to close\n");
3623         wait_event_timeout(mdsc->session_close_wq, done_closing_sessions(mdsc),
3624                            ceph_timeout_jiffies(opts->mount_timeout));
3625
3626         /* tear down remaining sessions */
3627         mutex_lock(&mdsc->mutex);
3628         for (i = 0; i < mdsc->max_sessions; i++) {
3629                 if (mdsc->sessions[i]) {
3630                         session = get_session(mdsc->sessions[i]);
3631                         __unregister_session(mdsc, session);
3632                         mutex_unlock(&mdsc->mutex);
3633                         mutex_lock(&session->s_mutex);
3634                         remove_session_caps(session);
3635                         mutex_unlock(&session->s_mutex);
3636                         ceph_put_mds_session(session);
3637                         mutex_lock(&mdsc->mutex);
3638                 }
3639         }
3640         WARN_ON(!list_empty(&mdsc->cap_delay_list));
3641         mutex_unlock(&mdsc->mutex);
3642
3643         ceph_cleanup_empty_realms(mdsc);
3644
3645         cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
3646
3647         dout("stopped\n");
3648 }
3649
3650 void ceph_mdsc_force_umount(struct ceph_mds_client *mdsc)
3651 {
3652         struct ceph_mds_session *session;
3653         int mds;
3654
3655         dout("force umount\n");
3656
3657         mutex_lock(&mdsc->mutex);
3658         for (mds = 0; mds < mdsc->max_sessions; mds++) {
3659                 session = __ceph_lookup_mds_session(mdsc, mds);
3660                 if (!session)
3661                         continue;
3662                 mutex_unlock(&mdsc->mutex);
3663                 mutex_lock(&session->s_mutex);
3664                 __close_session(mdsc, session);
3665                 if (session->s_state == CEPH_MDS_SESSION_CLOSING) {
3666                         cleanup_session_requests(mdsc, session);
3667                         remove_session_caps(session);
3668                 }
3669                 mutex_unlock(&session->s_mutex);
3670                 ceph_put_mds_session(session);
3671                 mutex_lock(&mdsc->mutex);
3672                 kick_requests(mdsc, mds);
3673         }
3674         __wake_requests(mdsc, &mdsc->waiting_for_map);
3675         mutex_unlock(&mdsc->mutex);
3676 }
3677
3678 static void ceph_mdsc_stop(struct ceph_mds_client *mdsc)
3679 {
3680         dout("stop\n");
3681         cancel_delayed_work_sync(&mdsc->delayed_work); /* cancel timer */
3682         if (mdsc->mdsmap)
3683                 ceph_mdsmap_destroy(mdsc->mdsmap);
3684         kfree(mdsc->sessions);
3685         ceph_caps_finalize(mdsc);
3686         ceph_pool_perm_destroy(mdsc);
3687 }
3688
3689 void ceph_mdsc_destroy(struct ceph_fs_client *fsc)
3690 {
3691         struct ceph_mds_client *mdsc = fsc->mdsc;
3692
3693         dout("mdsc_destroy %p\n", mdsc);
3694         ceph_mdsc_stop(mdsc);
3695
3696         /* flush out any connection work with references to us */
3697         ceph_msgr_flush();
3698
3699         fsc->mdsc = NULL;
3700         kfree(mdsc);
3701         dout("mdsc_destroy %p done\n", mdsc);
3702 }
3703
3704
3705 /*
3706  * handle mds map update.
3707  */
3708 void ceph_mdsc_handle_map(struct ceph_mds_client *mdsc, struct ceph_msg *msg)
3709 {
3710         u32 epoch;
3711         u32 maplen;
3712         void *p = msg->front.iov_base;
3713         void *end = p + msg->front.iov_len;
3714         struct ceph_mdsmap *newmap, *oldmap;
3715         struct ceph_fsid fsid;
3716         int err = -EINVAL;
3717
3718         ceph_decode_need(&p, end, sizeof(fsid)+2*sizeof(u32), bad);
3719         ceph_decode_copy(&p, &fsid, sizeof(fsid));
3720         if (ceph_check_fsid(mdsc->fsc->client, &fsid) < 0)
3721                 return;
3722         epoch = ceph_decode_32(&p);
3723         maplen = ceph_decode_32(&p);
3724         dout("handle_map epoch %u len %d\n", epoch, (int)maplen);
3725
3726         /* do we need it? */
3727         ceph_monc_got_mdsmap(&mdsc->fsc->client->monc, epoch);
3728         mutex_lock(&mdsc->mutex);
3729         if (mdsc->mdsmap && epoch <= mdsc->mdsmap->m_epoch) {
3730                 dout("handle_map epoch %u <= our %u\n",
3731                      epoch, mdsc->mdsmap->m_epoch);
3732                 mutex_unlock(&mdsc->mutex);
3733                 return;
3734         }
3735
3736         newmap = ceph_mdsmap_decode(&p, end);
3737         if (IS_ERR(newmap)) {
3738                 err = PTR_ERR(newmap);
3739                 goto bad_unlock;
3740         }
3741
3742         /* swap into place */
3743         if (mdsc->mdsmap) {
3744                 oldmap = mdsc->mdsmap;
3745                 mdsc->mdsmap = newmap;
3746                 check_new_map(mdsc, newmap, oldmap);
3747                 ceph_mdsmap_destroy(oldmap);
3748         } else {
3749                 mdsc->mdsmap = newmap;  /* first mds map */
3750         }
3751         mdsc->fsc->sb->s_maxbytes = mdsc->mdsmap->m_max_file_size;
3752
3753         __wake_requests(mdsc, &mdsc->waiting_for_map);
3754
3755         mutex_unlock(&mdsc->mutex);
3756         schedule_delayed(mdsc);
3757         return;
3758
3759 bad_unlock:
3760         mutex_unlock(&mdsc->mutex);
3761 bad:
3762         pr_err("error decoding mdsmap %d\n", err);
3763         return;
3764 }
3765
3766 static struct ceph_connection *con_get(struct ceph_connection *con)
3767 {
3768         struct ceph_mds_session *s = con->private;
3769
3770         if (get_session(s)) {
3771                 dout("mdsc con_get %p ok (%d)\n", s, atomic_read(&s->s_ref));
3772                 return con;
3773         }
3774         dout("mdsc con_get %p FAIL\n", s);
3775         return NULL;
3776 }
3777
3778 static void con_put(struct ceph_connection *con)
3779 {
3780         struct ceph_mds_session *s = con->private;
3781
3782         dout("mdsc con_put %p (%d)\n", s, atomic_read(&s->s_ref) - 1);
3783         ceph_put_mds_session(s);
3784 }
3785
3786 /*
3787  * if the client is unresponsive for long enough, the mds will kill
3788  * the session entirely.
3789  */
3790 static void peer_reset(struct ceph_connection *con)
3791 {
3792         struct ceph_mds_session *s = con->private;
3793         struct ceph_mds_client *mdsc = s->s_mdsc;
3794
3795         pr_warn("mds%d closed our session\n", s->s_mds);
3796         send_mds_reconnect(mdsc, s);
3797 }
3798
3799 static void dispatch(struct ceph_connection *con, struct ceph_msg *msg)
3800 {
3801         struct ceph_mds_session *s = con->private;
3802         struct ceph_mds_client *mdsc = s->s_mdsc;
3803         int type = le16_to_cpu(msg->hdr.type);
3804
3805         mutex_lock(&mdsc->mutex);
3806         if (__verify_registered_session(mdsc, s) < 0) {
3807                 mutex_unlock(&mdsc->mutex);
3808                 goto out;
3809         }
3810         mutex_unlock(&mdsc->mutex);
3811
3812         switch (type) {
3813         case CEPH_MSG_MDS_MAP:
3814                 ceph_mdsc_handle_map(mdsc, msg);
3815                 break;
3816         case CEPH_MSG_CLIENT_SESSION:
3817                 handle_session(s, msg);
3818                 break;
3819         case CEPH_MSG_CLIENT_REPLY:
3820                 handle_reply(s, msg);
3821                 break;
3822         case CEPH_MSG_CLIENT_REQUEST_FORWARD:
3823                 handle_forward(mdsc, s, msg);
3824                 break;
3825         case CEPH_MSG_CLIENT_CAPS:
3826                 ceph_handle_caps(s, msg);
3827                 break;
3828         case CEPH_MSG_CLIENT_SNAP:
3829                 ceph_handle_snap(mdsc, s, msg);
3830                 break;
3831         case CEPH_MSG_CLIENT_LEASE:
3832                 handle_lease(mdsc, s, msg);
3833                 break;
3834
3835         default:
3836                 pr_err("received unknown message type %d %s\n", type,
3837                        ceph_msg_type_name(type));
3838         }
3839 out:
3840         ceph_msg_put(msg);
3841 }
3842
3843 /*
3844  * authentication
3845  */
3846
3847 /*
3848  * Note: returned pointer is the address of a structure that's
3849  * managed separately.  Caller must *not* attempt to free it.
3850  */
3851 static struct ceph_auth_handshake *get_authorizer(struct ceph_connection *con,
3852                                         int *proto, int force_new)
3853 {
3854         struct ceph_mds_session *s = con->private;
3855         struct ceph_mds_client *mdsc = s->s_mdsc;
3856         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
3857         struct ceph_auth_handshake *auth = &s->s_auth;
3858
3859         if (force_new && auth->authorizer) {
3860                 ceph_auth_destroy_authorizer(ac, auth->authorizer);
3861                 auth->authorizer = NULL;
3862         }
3863         if (!auth->authorizer) {
3864                 int ret = ceph_auth_create_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
3865                                                       auth);
3866                 if (ret)
3867                         return ERR_PTR(ret);
3868         } else {
3869                 int ret = ceph_auth_update_authorizer(ac, CEPH_ENTITY_TYPE_MDS,
3870                                                       auth);
3871                 if (ret)
3872                         return ERR_PTR(ret);
3873         }
3874         *proto = ac->protocol;
3875
3876         return auth;
3877 }
3878
3879
3880 static int verify_authorizer_reply(struct ceph_connection *con, int len)
3881 {
3882         struct ceph_mds_session *s = con->private;
3883         struct ceph_mds_client *mdsc = s->s_mdsc;
3884         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
3885
3886         return ceph_auth_verify_authorizer_reply(ac, s->s_auth.authorizer, len);
3887 }
3888
3889 static int invalidate_authorizer(struct ceph_connection *con)
3890 {
3891         struct ceph_mds_session *s = con->private;
3892         struct ceph_mds_client *mdsc = s->s_mdsc;
3893         struct ceph_auth_client *ac = mdsc->fsc->client->monc.auth;
3894
3895         ceph_auth_invalidate_authorizer(ac, CEPH_ENTITY_TYPE_MDS);
3896
3897         return ceph_monc_validate_auth(&mdsc->fsc->client->monc);
3898 }
3899
3900 static struct ceph_msg *mds_alloc_msg(struct ceph_connection *con,
3901                                 struct ceph_msg_header *hdr, int *skip)
3902 {
3903         struct ceph_msg *msg;
3904         int type = (int) le16_to_cpu(hdr->type);
3905         int front_len = (int) le32_to_cpu(hdr->front_len);
3906
3907         if (con->in_msg)
3908                 return con->in_msg;
3909
3910         *skip = 0;
3911         msg = ceph_msg_new(type, front_len, GFP_NOFS, false);
3912         if (!msg) {
3913                 pr_err("unable to allocate msg type %d len %d\n",
3914                        type, front_len);
3915                 return NULL;
3916         }
3917
3918         return msg;
3919 }
3920
3921 static int sign_message(struct ceph_connection *con, struct ceph_msg *msg)
3922 {
3923        struct ceph_mds_session *s = con->private;
3924        struct ceph_auth_handshake *auth = &s->s_auth;
3925        return ceph_auth_sign_message(auth, msg);
3926 }
3927
3928 static int check_message_signature(struct ceph_connection *con, struct ceph_msg *msg)
3929 {
3930        struct ceph_mds_session *s = con->private;
3931        struct ceph_auth_handshake *auth = &s->s_auth;
3932        return ceph_auth_check_message_signature(auth, msg);
3933 }
3934
3935 static const struct ceph_connection_operations mds_con_ops = {
3936         .get = con_get,
3937         .put = con_put,
3938         .dispatch = dispatch,
3939         .get_authorizer = get_authorizer,
3940         .verify_authorizer_reply = verify_authorizer_reply,
3941         .invalidate_authorizer = invalidate_authorizer,
3942         .peer_reset = peer_reset,
3943         .alloc_msg = mds_alloc_msg,
3944         .sign_message = sign_message,
3945         .check_message_signature = check_message_signature,
3946 };
3947
3948 /* eof */