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  41  * are no longer valid.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