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  42 * are no longer valid.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