linux/fs/nfsd/nfs4state.c
<<
>>
Prefs
   1/*
   2*  Copyright (c) 2001 The Regents of the University of Michigan.
   3*  All rights reserved.
   4*
   5*  Kendrick Smith <kmsmith@umich.edu>
   6*  Andy Adamson <kandros@umich.edu>
   7*
   8*  Redistribution and use in source and binary forms, with or without
   9*  modification, are permitted provided that the following conditions
  10*  are met:
  11*
  12*  1. Redistributions of source code must retain the above copyright
  13*     notice, this list of conditions and the following disclaimer.
  14*  2. Redistributions in binary form must reproduce the above copyright
  15*     notice, this list of conditions and the following disclaimer in the
  16*     documentation and/or other materials provided with the distribution.
  17*  3. Neither the name of the University nor the names of its
  18*     contributors may be used to endorse or promote products derived
  19*     from this software without specific prior written permission.
  20*
  21*  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  22*  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  23*  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  24*  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  25*  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  26*  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  27*  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  28*  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  29*  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  30*  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  31*  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  32*
  33*/
  34
  35#include <linux/file.h>
  36#include <linux/fs.h>
  37#include <linux/slab.h>
  38#include <linux/namei.h>
  39#include <linux/swap.h>
  40#include <linux/pagemap.h>
  41#include <linux/ratelimit.h>
  42#include <linux/sunrpc/svcauth_gss.h>
  43#include <linux/sunrpc/addr.h>
  44#include <linux/jhash.h>
  45#include <linux/string_helpers.h>
  46#include <linux/fsnotify.h>
  47#include <linux/nfs_ssc.h>
  48#include "xdr4.h"
  49#include "xdr4cb.h"
  50#include "vfs.h"
  51#include "current_stateid.h"
  52
  53#include "netns.h"
  54#include "pnfs.h"
  55#include "filecache.h"
  56#include "trace.h"
  57
  58#define NFSDDBG_FACILITY                NFSDDBG_PROC
  59
  60#define all_ones {{~0,~0},~0}
  61static const stateid_t one_stateid = {
  62        .si_generation = ~0,
  63        .si_opaque = all_ones,
  64};
  65static const stateid_t zero_stateid = {
  66        /* all fields zero */
  67};
  68static const stateid_t currentstateid = {
  69        .si_generation = 1,
  70};
  71static const stateid_t close_stateid = {
  72        .si_generation = 0xffffffffU,
  73};
  74
  75static u64 current_sessionid = 1;
  76
  77#define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
  78#define ONE_STATEID(stateid)  (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
  79#define CURRENT_STATEID(stateid) (!memcmp((stateid), &currentstateid, sizeof(stateid_t)))
  80#define CLOSE_STATEID(stateid)  (!memcmp((stateid), &close_stateid, sizeof(stateid_t)))
  81
  82/* forward declarations */
  83static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
  84static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
  85void nfsd4_end_grace(struct nfsd_net *nn);
  86static void _free_cpntf_state_locked(struct nfsd_net *nn, struct nfs4_cpntf_state *cps);
  87
  88/* Locking: */
  89
  90/*
  91 * Currently used for the del_recall_lru and file hash table.  In an
  92 * effort to decrease the scope of the client_mutex, this spinlock may
  93 * eventually cover more:
  94 */
  95static DEFINE_SPINLOCK(state_lock);
  96
  97enum nfsd4_st_mutex_lock_subclass {
  98        OPEN_STATEID_MUTEX = 0,
  99        LOCK_STATEID_MUTEX = 1,
 100};
 101
 102/*
 103 * A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
 104 * the refcount on the open stateid to drop.
 105 */
 106static DECLARE_WAIT_QUEUE_HEAD(close_wq);
 107
 108/*
 109 * A waitqueue where a writer to clients/#/ctl destroying a client can
 110 * wait for cl_rpc_users to drop to 0 and then for the client to be
 111 * unhashed.
 112 */
 113static DECLARE_WAIT_QUEUE_HEAD(expiry_wq);
 114
 115static struct kmem_cache *client_slab;
 116static struct kmem_cache *openowner_slab;
 117static struct kmem_cache *lockowner_slab;
 118static struct kmem_cache *file_slab;
 119static struct kmem_cache *stateid_slab;
 120static struct kmem_cache *deleg_slab;
 121static struct kmem_cache *odstate_slab;
 122
 123static void free_session(struct nfsd4_session *);
 124
 125static const struct nfsd4_callback_ops nfsd4_cb_recall_ops;
 126static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops;
 127
 128static struct workqueue_struct *laundry_wq;
 129
 130int nfsd4_create_laundry_wq(void)
 131{
 132        int rc = 0;
 133
 134        laundry_wq = alloc_workqueue("%s", WQ_UNBOUND, 0, "nfsd4");
 135        if (laundry_wq == NULL)
 136                rc = -ENOMEM;
 137        return rc;
 138}
 139
 140void nfsd4_destroy_laundry_wq(void)
 141{
 142        destroy_workqueue(laundry_wq);
 143}
 144
 145static bool is_session_dead(struct nfsd4_session *ses)
 146{
 147        return ses->se_flags & NFS4_SESSION_DEAD;
 148}
 149
 150static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
 151{
 152        if (atomic_read(&ses->se_ref) > ref_held_by_me)
 153                return nfserr_jukebox;
 154        ses->se_flags |= NFS4_SESSION_DEAD;
 155        return nfs_ok;
 156}
 157
 158static bool is_client_expired(struct nfs4_client *clp)
 159{
 160        return clp->cl_time == 0;
 161}
 162
 163static __be32 get_client_locked(struct nfs4_client *clp)
 164{
 165        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 166
 167        lockdep_assert_held(&nn->client_lock);
 168
 169        if (is_client_expired(clp))
 170                return nfserr_expired;
 171        atomic_inc(&clp->cl_rpc_users);
 172        clp->cl_state = NFSD4_ACTIVE;
 173        return nfs_ok;
 174}
 175
 176/* must be called under the client_lock */
 177static inline void
 178renew_client_locked(struct nfs4_client *clp)
 179{
 180        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 181
 182        if (is_client_expired(clp)) {
 183                WARN_ON(1);
 184                printk("%s: client (clientid %08x/%08x) already expired\n",
 185                        __func__,
 186                        clp->cl_clientid.cl_boot,
 187                        clp->cl_clientid.cl_id);
 188                return;
 189        }
 190
 191        list_move_tail(&clp->cl_lru, &nn->client_lru);
 192        clp->cl_time = ktime_get_boottime_seconds();
 193        clp->cl_state = NFSD4_ACTIVE;
 194}
 195
 196static void put_client_renew_locked(struct nfs4_client *clp)
 197{
 198        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 199
 200        lockdep_assert_held(&nn->client_lock);
 201
 202        if (!atomic_dec_and_test(&clp->cl_rpc_users))
 203                return;
 204        if (!is_client_expired(clp))
 205                renew_client_locked(clp);
 206        else
 207                wake_up_all(&expiry_wq);
 208}
 209
 210static void put_client_renew(struct nfs4_client *clp)
 211{
 212        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 213
 214        if (!atomic_dec_and_lock(&clp->cl_rpc_users, &nn->client_lock))
 215                return;
 216        if (!is_client_expired(clp))
 217                renew_client_locked(clp);
 218        else
 219                wake_up_all(&expiry_wq);
 220        spin_unlock(&nn->client_lock);
 221}
 222
 223static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
 224{
 225        __be32 status;
 226
 227        if (is_session_dead(ses))
 228                return nfserr_badsession;
 229        status = get_client_locked(ses->se_client);
 230        if (status)
 231                return status;
 232        atomic_inc(&ses->se_ref);
 233        return nfs_ok;
 234}
 235
 236static void nfsd4_put_session_locked(struct nfsd4_session *ses)
 237{
 238        struct nfs4_client *clp = ses->se_client;
 239        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 240
 241        lockdep_assert_held(&nn->client_lock);
 242
 243        if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
 244                free_session(ses);
 245        put_client_renew_locked(clp);
 246}
 247
 248static void nfsd4_put_session(struct nfsd4_session *ses)
 249{
 250        struct nfs4_client *clp = ses->se_client;
 251        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 252
 253        spin_lock(&nn->client_lock);
 254        nfsd4_put_session_locked(ses);
 255        spin_unlock(&nn->client_lock);
 256}
 257
 258static struct nfsd4_blocked_lock *
 259find_blocked_lock(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
 260                        struct nfsd_net *nn)
 261{
 262        struct nfsd4_blocked_lock *cur, *found = NULL;
 263
 264        spin_lock(&nn->blocked_locks_lock);
 265        list_for_each_entry(cur, &lo->lo_blocked, nbl_list) {
 266                if (fh_match(fh, &cur->nbl_fh)) {
 267                        list_del_init(&cur->nbl_list);
 268                        WARN_ON(list_empty(&cur->nbl_lru));
 269                        list_del_init(&cur->nbl_lru);
 270                        found = cur;
 271                        break;
 272                }
 273        }
 274        spin_unlock(&nn->blocked_locks_lock);
 275        if (found)
 276                locks_delete_block(&found->nbl_lock);
 277        return found;
 278}
 279
 280static struct nfsd4_blocked_lock *
 281find_or_allocate_block(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
 282                        struct nfsd_net *nn)
 283{
 284        struct nfsd4_blocked_lock *nbl;
 285
 286        nbl = find_blocked_lock(lo, fh, nn);
 287        if (!nbl) {
 288                nbl= kmalloc(sizeof(*nbl), GFP_KERNEL);
 289                if (nbl) {
 290                        INIT_LIST_HEAD(&nbl->nbl_list);
 291                        INIT_LIST_HEAD(&nbl->nbl_lru);
 292                        fh_copy_shallow(&nbl->nbl_fh, fh);
 293                        locks_init_lock(&nbl->nbl_lock);
 294                        kref_init(&nbl->nbl_kref);
 295                        nfsd4_init_cb(&nbl->nbl_cb, lo->lo_owner.so_client,
 296                                        &nfsd4_cb_notify_lock_ops,
 297                                        NFSPROC4_CLNT_CB_NOTIFY_LOCK);
 298                }
 299        }
 300        return nbl;
 301}
 302
 303static void
 304free_nbl(struct kref *kref)
 305{
 306        struct nfsd4_blocked_lock *nbl;
 307
 308        nbl = container_of(kref, struct nfsd4_blocked_lock, nbl_kref);
 309        kfree(nbl);
 310}
 311
 312static void
 313free_blocked_lock(struct nfsd4_blocked_lock *nbl)
 314{
 315        locks_delete_block(&nbl->nbl_lock);
 316        locks_release_private(&nbl->nbl_lock);
 317        kref_put(&nbl->nbl_kref, free_nbl);
 318}
 319
 320static void
 321remove_blocked_locks(struct nfs4_lockowner *lo)
 322{
 323        struct nfs4_client *clp = lo->lo_owner.so_client;
 324        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 325        struct nfsd4_blocked_lock *nbl;
 326        LIST_HEAD(reaplist);
 327
 328        /* Dequeue all blocked locks */
 329        spin_lock(&nn->blocked_locks_lock);
 330        while (!list_empty(&lo->lo_blocked)) {
 331                nbl = list_first_entry(&lo->lo_blocked,
 332                                        struct nfsd4_blocked_lock,
 333                                        nbl_list);
 334                list_del_init(&nbl->nbl_list);
 335                WARN_ON(list_empty(&nbl->nbl_lru));
 336                list_move(&nbl->nbl_lru, &reaplist);
 337        }
 338        spin_unlock(&nn->blocked_locks_lock);
 339
 340        /* Now free them */
 341        while (!list_empty(&reaplist)) {
 342                nbl = list_first_entry(&reaplist, struct nfsd4_blocked_lock,
 343                                        nbl_lru);
 344                list_del_init(&nbl->nbl_lru);
 345                free_blocked_lock(nbl);
 346        }
 347}
 348
 349static void
 350nfsd4_cb_notify_lock_prepare(struct nfsd4_callback *cb)
 351{
 352        struct nfsd4_blocked_lock       *nbl = container_of(cb,
 353                                                struct nfsd4_blocked_lock, nbl_cb);
 354        locks_delete_block(&nbl->nbl_lock);
 355}
 356
 357static int
 358nfsd4_cb_notify_lock_done(struct nfsd4_callback *cb, struct rpc_task *task)
 359{
 360        /*
 361         * Since this is just an optimization, we don't try very hard if it
 362         * turns out not to succeed. We'll requeue it on NFS4ERR_DELAY, and
 363         * just quit trying on anything else.
 364         */
 365        switch (task->tk_status) {
 366        case -NFS4ERR_DELAY:
 367                rpc_delay(task, 1 * HZ);
 368                return 0;
 369        default:
 370                return 1;
 371        }
 372}
 373
 374static void
 375nfsd4_cb_notify_lock_release(struct nfsd4_callback *cb)
 376{
 377        struct nfsd4_blocked_lock       *nbl = container_of(cb,
 378                                                struct nfsd4_blocked_lock, nbl_cb);
 379
 380        free_blocked_lock(nbl);
 381}
 382
 383static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops = {
 384        .prepare        = nfsd4_cb_notify_lock_prepare,
 385        .done           = nfsd4_cb_notify_lock_done,
 386        .release        = nfsd4_cb_notify_lock_release,
 387};
 388
 389/*
 390 * We store the NONE, READ, WRITE, and BOTH bits separately in the
 391 * st_{access,deny}_bmap field of the stateid, in order to track not
 392 * only what share bits are currently in force, but also what
 393 * combinations of share bits previous opens have used.  This allows us
 394 * to enforce the recommendation in
 395 * https://datatracker.ietf.org/doc/html/rfc7530#section-16.19.4 that
 396 * the server return an error if the client attempt to downgrade to a
 397 * combination of share bits not explicable by closing some of its
 398 * previous opens.
 399 *
 400 * This enforcement is arguably incomplete, since we don't keep
 401 * track of access/deny bit combinations; so, e.g., we allow:
 402 *
 403 *      OPEN allow read, deny write
 404 *      OPEN allow both, deny none
 405 *      DOWNGRADE allow read, deny none
 406 *
 407 * which we should reject.
 408 *
 409 * But you could also argue that our current code is already overkill,
 410 * since it only exists to return NFS4ERR_INVAL on incorrect client
 411 * behavior.
 412 */
 413static unsigned int
 414bmap_to_share_mode(unsigned long bmap)
 415{
 416        int i;
 417        unsigned int access = 0;
 418
 419        for (i = 1; i < 4; i++) {
 420                if (test_bit(i, &bmap))
 421                        access |= i;
 422        }
 423        return access;
 424}
 425
 426/* set share access for a given stateid */
 427static inline void
 428set_access(u32 access, struct nfs4_ol_stateid *stp)
 429{
 430        unsigned char mask = 1 << access;
 431
 432        WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
 433        stp->st_access_bmap |= mask;
 434}
 435
 436/* clear share access for a given stateid */
 437static inline void
 438clear_access(u32 access, struct nfs4_ol_stateid *stp)
 439{
 440        unsigned char mask = 1 << access;
 441
 442        WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
 443        stp->st_access_bmap &= ~mask;
 444}
 445
 446/* test whether a given stateid has access */
 447static inline bool
 448test_access(u32 access, struct nfs4_ol_stateid *stp)
 449{
 450        unsigned char mask = 1 << access;
 451
 452        return (bool)(stp->st_access_bmap & mask);
 453}
 454
 455/* set share deny for a given stateid */
 456static inline void
 457set_deny(u32 deny, struct nfs4_ol_stateid *stp)
 458{
 459        unsigned char mask = 1 << deny;
 460
 461        WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
 462        stp->st_deny_bmap |= mask;
 463}
 464
 465/* clear share deny for a given stateid */
 466static inline void
 467clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
 468{
 469        unsigned char mask = 1 << deny;
 470
 471        WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
 472        stp->st_deny_bmap &= ~mask;
 473}
 474
 475/* test whether a given stateid is denying specific access */
 476static inline bool
 477test_deny(u32 deny, struct nfs4_ol_stateid *stp)
 478{
 479        unsigned char mask = 1 << deny;
 480
 481        return (bool)(stp->st_deny_bmap & mask);
 482}
 483
 484static int nfs4_access_to_omode(u32 access)
 485{
 486        switch (access & NFS4_SHARE_ACCESS_BOTH) {
 487        case NFS4_SHARE_ACCESS_READ:
 488                return O_RDONLY;
 489        case NFS4_SHARE_ACCESS_WRITE:
 490                return O_WRONLY;
 491        case NFS4_SHARE_ACCESS_BOTH:
 492                return O_RDWR;
 493        }
 494        WARN_ON_ONCE(1);
 495        return O_RDONLY;
 496}
 497
 498static inline int
 499access_permit_read(struct nfs4_ol_stateid *stp)
 500{
 501        return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
 502                test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
 503                test_access(NFS4_SHARE_ACCESS_WRITE, stp);
 504}
 505
 506static inline int
 507access_permit_write(struct nfs4_ol_stateid *stp)
 508{
 509        return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
 510                test_access(NFS4_SHARE_ACCESS_BOTH, stp);
 511}
 512
 513static inline struct nfs4_stateowner *
 514nfs4_get_stateowner(struct nfs4_stateowner *sop)
 515{
 516        atomic_inc(&sop->so_count);
 517        return sop;
 518}
 519
 520static int
 521same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner)
 522{
 523        return (sop->so_owner.len == owner->len) &&
 524                0 == memcmp(sop->so_owner.data, owner->data, owner->len);
 525}
 526
 527static struct nfs4_openowner *
 528find_openstateowner_str_locked(unsigned int hashval, struct nfsd4_open *open,
 529                        struct nfs4_client *clp)
 530{
 531        struct nfs4_stateowner *so;
 532
 533        lockdep_assert_held(&clp->cl_lock);
 534
 535        list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[hashval],
 536                            so_strhash) {
 537                if (!so->so_is_open_owner)
 538                        continue;
 539                if (same_owner_str(so, &open->op_owner))
 540                        return openowner(nfs4_get_stateowner(so));
 541        }
 542        return NULL;
 543}
 544
 545static struct nfs4_openowner *
 546find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
 547                        struct nfs4_client *clp)
 548{
 549        struct nfs4_openowner *oo;
 550
 551        spin_lock(&clp->cl_lock);
 552        oo = find_openstateowner_str_locked(hashval, open, clp);
 553        spin_unlock(&clp->cl_lock);
 554        return oo;
 555}
 556
 557static inline u32
 558opaque_hashval(const void *ptr, int nbytes)
 559{
 560        unsigned char *cptr = (unsigned char *) ptr;
 561
 562        u32 x = 0;
 563        while (nbytes--) {
 564                x *= 37;
 565                x += *cptr++;
 566        }
 567        return x;
 568}
 569
 570static void nfsd4_free_file_rcu(struct rcu_head *rcu)
 571{
 572        struct nfs4_file *fp = container_of(rcu, struct nfs4_file, fi_rcu);
 573
 574        kmem_cache_free(file_slab, fp);
 575}
 576
 577void
 578put_nfs4_file(struct nfs4_file *fi)
 579{
 580        might_lock(&state_lock);
 581
 582        if (refcount_dec_and_lock(&fi->fi_ref, &state_lock)) {
 583                hlist_del_rcu(&fi->fi_hash);
 584                spin_unlock(&state_lock);
 585                WARN_ON_ONCE(!list_empty(&fi->fi_clnt_odstate));
 586                WARN_ON_ONCE(!list_empty(&fi->fi_delegations));
 587                call_rcu(&fi->fi_rcu, nfsd4_free_file_rcu);
 588        }
 589}
 590
 591static struct nfsd_file *
 592__nfs4_get_fd(struct nfs4_file *f, int oflag)
 593{
 594        if (f->fi_fds[oflag])
 595                return nfsd_file_get(f->fi_fds[oflag]);
 596        return NULL;
 597}
 598
 599static struct nfsd_file *
 600find_writeable_file_locked(struct nfs4_file *f)
 601{
 602        struct nfsd_file *ret;
 603
 604        lockdep_assert_held(&f->fi_lock);
 605
 606        ret = __nfs4_get_fd(f, O_WRONLY);
 607        if (!ret)
 608                ret = __nfs4_get_fd(f, O_RDWR);
 609        return ret;
 610}
 611
 612static struct nfsd_file *
 613find_writeable_file(struct nfs4_file *f)
 614{
 615        struct nfsd_file *ret;
 616
 617        spin_lock(&f->fi_lock);
 618        ret = find_writeable_file_locked(f);
 619        spin_unlock(&f->fi_lock);
 620
 621        return ret;
 622}
 623
 624static struct nfsd_file *
 625find_readable_file_locked(struct nfs4_file *f)
 626{
 627        struct nfsd_file *ret;
 628
 629        lockdep_assert_held(&f->fi_lock);
 630
 631        ret = __nfs4_get_fd(f, O_RDONLY);
 632        if (!ret)
 633                ret = __nfs4_get_fd(f, O_RDWR);
 634        return ret;
 635}
 636
 637static struct nfsd_file *
 638find_readable_file(struct nfs4_file *f)
 639{
 640        struct nfsd_file *ret;
 641
 642        spin_lock(&f->fi_lock);
 643        ret = find_readable_file_locked(f);
 644        spin_unlock(&f->fi_lock);
 645
 646        return ret;
 647}
 648
 649struct nfsd_file *
 650find_any_file(struct nfs4_file *f)
 651{
 652        struct nfsd_file *ret;
 653
 654        if (!f)
 655                return NULL;
 656        spin_lock(&f->fi_lock);
 657        ret = __nfs4_get_fd(f, O_RDWR);
 658        if (!ret) {
 659                ret = __nfs4_get_fd(f, O_WRONLY);
 660                if (!ret)
 661                        ret = __nfs4_get_fd(f, O_RDONLY);
 662        }
 663        spin_unlock(&f->fi_lock);
 664        return ret;
 665}
 666
 667static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
 668{
 669        struct nfsd_file *ret = NULL;
 670
 671        spin_lock(&f->fi_lock);
 672        if (f->fi_deleg_file)
 673                ret = nfsd_file_get(f->fi_deleg_file);
 674        spin_unlock(&f->fi_lock);
 675        return ret;
 676}
 677
 678static atomic_long_t num_delegations;
 679unsigned long max_delegations;
 680
 681/*
 682 * Open owner state (share locks)
 683 */
 684
 685/* hash tables for lock and open owners */
 686#define OWNER_HASH_BITS              8
 687#define OWNER_HASH_SIZE             (1 << OWNER_HASH_BITS)
 688#define OWNER_HASH_MASK             (OWNER_HASH_SIZE - 1)
 689
 690static unsigned int ownerstr_hashval(struct xdr_netobj *ownername)
 691{
 692        unsigned int ret;
 693
 694        ret = opaque_hashval(ownername->data, ownername->len);
 695        return ret & OWNER_HASH_MASK;
 696}
 697
 698/* hash table for nfs4_file */
 699#define FILE_HASH_BITS                   8
 700#define FILE_HASH_SIZE                  (1 << FILE_HASH_BITS)
 701
 702static unsigned int file_hashval(struct svc_fh *fh)
 703{
 704        struct inode *inode = d_inode(fh->fh_dentry);
 705
 706        /* XXX: why not (here & in file cache) use inode? */
 707        return (unsigned int)hash_long(inode->i_ino, FILE_HASH_BITS);
 708}
 709
 710static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
 711
 712/*
 713 * Check if courtesy clients have conflicting access and resolve it if possible
 714 *
 715 * access:  is op_share_access if share_access is true.
 716 *          Check if access mode, op_share_access, would conflict with
 717 *          the current deny mode of the file 'fp'.
 718 * access:  is op_share_deny if share_access is false.
 719 *          Check if the deny mode, op_share_deny, would conflict with
 720 *          current access of the file 'fp'.
 721 * stp:     skip checking this entry.
 722 * new_stp: normal open, not open upgrade.
 723 *
 724 * Function returns:
 725 *      false - access/deny mode conflict with normal client.
 726 *      true  - no conflict or conflict with courtesy client(s) is resolved.
 727 */
 728static bool
 729nfs4_resolve_deny_conflicts_locked(struct nfs4_file *fp, bool new_stp,
 730                struct nfs4_ol_stateid *stp, u32 access, bool share_access)
 731{
 732        struct nfs4_ol_stateid *st;
 733        bool resolvable = true;
 734        unsigned char bmap;
 735        struct nfsd_net *nn;
 736        struct nfs4_client *clp;
 737
 738        lockdep_assert_held(&fp->fi_lock);
 739        list_for_each_entry(st, &fp->fi_stateids, st_perfile) {
 740                /* ignore lock stateid */
 741                if (st->st_openstp)
 742                        continue;
 743                if (st == stp && new_stp)
 744                        continue;
 745                /* check file access against deny mode or vice versa */
 746                bmap = share_access ? st->st_deny_bmap : st->st_access_bmap;
 747                if (!(access & bmap_to_share_mode(bmap)))
 748                        continue;
 749                clp = st->st_stid.sc_client;
 750                if (try_to_expire_client(clp))
 751                        continue;
 752                resolvable = false;
 753                break;
 754        }
 755        if (resolvable) {
 756                clp = stp->st_stid.sc_client;
 757                nn = net_generic(clp->net, nfsd_net_id);
 758                mod_delayed_work(laundry_wq, &nn->laundromat_work, 0);
 759        }
 760        return resolvable;
 761}
 762
 763static void
 764__nfs4_file_get_access(struct nfs4_file *fp, u32 access)
 765{
 766        lockdep_assert_held(&fp->fi_lock);
 767
 768        if (access & NFS4_SHARE_ACCESS_WRITE)
 769                atomic_inc(&fp->fi_access[O_WRONLY]);
 770        if (access & NFS4_SHARE_ACCESS_READ)
 771                atomic_inc(&fp->fi_access[O_RDONLY]);
 772}
 773
 774static __be32
 775nfs4_file_get_access(struct nfs4_file *fp, u32 access)
 776{
 777        lockdep_assert_held(&fp->fi_lock);
 778
 779        /* Does this access mode make sense? */
 780        if (access & ~NFS4_SHARE_ACCESS_BOTH)
 781                return nfserr_inval;
 782
 783        /* Does it conflict with a deny mode already set? */
 784        if ((access & fp->fi_share_deny) != 0)
 785                return nfserr_share_denied;
 786
 787        __nfs4_file_get_access(fp, access);
 788        return nfs_ok;
 789}
 790
 791static __be32 nfs4_file_check_deny(struct nfs4_file *fp, u32 deny)
 792{
 793        /* Common case is that there is no deny mode. */
 794        if (deny) {
 795                /* Does this deny mode make sense? */
 796                if (deny & ~NFS4_SHARE_DENY_BOTH)
 797                        return nfserr_inval;
 798
 799                if ((deny & NFS4_SHARE_DENY_READ) &&
 800                    atomic_read(&fp->fi_access[O_RDONLY]))
 801                        return nfserr_share_denied;
 802
 803                if ((deny & NFS4_SHARE_DENY_WRITE) &&
 804                    atomic_read(&fp->fi_access[O_WRONLY]))
 805                        return nfserr_share_denied;
 806        }
 807        return nfs_ok;
 808}
 809
 810static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
 811{
 812        might_lock(&fp->fi_lock);
 813
 814        if (atomic_dec_and_lock(&fp->fi_access[oflag], &fp->fi_lock)) {
 815                struct nfsd_file *f1 = NULL;
 816                struct nfsd_file *f2 = NULL;
 817
 818                swap(f1, fp->fi_fds[oflag]);
 819                if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
 820                        swap(f2, fp->fi_fds[O_RDWR]);
 821                spin_unlock(&fp->fi_lock);
 822                if (f1)
 823                        nfsd_file_put(f1);
 824                if (f2)
 825                        nfsd_file_put(f2);
 826        }
 827}
 828
 829static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
 830{
 831        WARN_ON_ONCE(access & ~NFS4_SHARE_ACCESS_BOTH);
 832
 833        if (access & NFS4_SHARE_ACCESS_WRITE)
 834                __nfs4_file_put_access(fp, O_WRONLY);
 835        if (access & NFS4_SHARE_ACCESS_READ)
 836                __nfs4_file_put_access(fp, O_RDONLY);
 837}
 838
 839/*
 840 * Allocate a new open/delegation state counter. This is needed for
 841 * pNFS for proper return on close semantics.
 842 *
 843 * Note that we only allocate it for pNFS-enabled exports, otherwise
 844 * all pointers to struct nfs4_clnt_odstate are always NULL.
 845 */
 846static struct nfs4_clnt_odstate *
 847alloc_clnt_odstate(struct nfs4_client *clp)
 848{
 849        struct nfs4_clnt_odstate *co;
 850
 851        co = kmem_cache_zalloc(odstate_slab, GFP_KERNEL);
 852        if (co) {
 853                co->co_client = clp;
 854                refcount_set(&co->co_odcount, 1);
 855        }
 856        return co;
 857}
 858
 859static void
 860hash_clnt_odstate_locked(struct nfs4_clnt_odstate *co)
 861{
 862        struct nfs4_file *fp = co->co_file;
 863
 864        lockdep_assert_held(&fp->fi_lock);
 865        list_add(&co->co_perfile, &fp->fi_clnt_odstate);
 866}
 867
 868static inline void
 869get_clnt_odstate(struct nfs4_clnt_odstate *co)
 870{
 871        if (co)
 872                refcount_inc(&co->co_odcount);
 873}
 874
 875static void
 876put_clnt_odstate(struct nfs4_clnt_odstate *co)
 877{
 878        struct nfs4_file *fp;
 879
 880        if (!co)
 881                return;
 882
 883        fp = co->co_file;
 884        if (refcount_dec_and_lock(&co->co_odcount, &fp->fi_lock)) {
 885                list_del(&co->co_perfile);
 886                spin_unlock(&fp->fi_lock);
 887
 888                nfsd4_return_all_file_layouts(co->co_client, fp);
 889                kmem_cache_free(odstate_slab, co);
 890        }
 891}
 892
 893static struct nfs4_clnt_odstate *
 894find_or_hash_clnt_odstate(struct nfs4_file *fp, struct nfs4_clnt_odstate *new)
 895{
 896        struct nfs4_clnt_odstate *co;
 897        struct nfs4_client *cl;
 898
 899        if (!new)
 900                return NULL;
 901
 902        cl = new->co_client;
 903
 904        spin_lock(&fp->fi_lock);
 905        list_for_each_entry(co, &fp->fi_clnt_odstate, co_perfile) {
 906                if (co->co_client == cl) {
 907                        get_clnt_odstate(co);
 908                        goto out;
 909                }
 910        }
 911        co = new;
 912        co->co_file = fp;
 913        hash_clnt_odstate_locked(new);
 914out:
 915        spin_unlock(&fp->fi_lock);
 916        return co;
 917}
 918
 919struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab,
 920                                  void (*sc_free)(struct nfs4_stid *))
 921{
 922        struct nfs4_stid *stid;
 923        int new_id;
 924
 925        stid = kmem_cache_zalloc(slab, GFP_KERNEL);
 926        if (!stid)
 927                return NULL;
 928
 929        idr_preload(GFP_KERNEL);
 930        spin_lock(&cl->cl_lock);
 931        /* Reserving 0 for start of file in nfsdfs "states" file: */
 932        new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 1, 0, GFP_NOWAIT);
 933        spin_unlock(&cl->cl_lock);
 934        idr_preload_end();
 935        if (new_id < 0)
 936                goto out_free;
 937
 938        stid->sc_free = sc_free;
 939        stid->sc_client = cl;
 940        stid->sc_stateid.si_opaque.so_id = new_id;
 941        stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
 942        /* Will be incremented before return to client: */
 943        refcount_set(&stid->sc_count, 1);
 944        spin_lock_init(&stid->sc_lock);
 945        INIT_LIST_HEAD(&stid->sc_cp_list);
 946
 947        /*
 948         * It shouldn't be a problem to reuse an opaque stateid value.
 949         * I don't think it is for 4.1.  But with 4.0 I worry that, for
 950         * example, a stray write retransmission could be accepted by
 951         * the server when it should have been rejected.  Therefore,
 952         * adopt a trick from the sctp code to attempt to maximize the
 953         * amount of time until an id is reused, by ensuring they always
 954         * "increase" (mod INT_MAX):
 955         */
 956        return stid;
 957out_free:
 958        kmem_cache_free(slab, stid);
 959        return NULL;
 960}
 961
 962/*
 963 * Create a unique stateid_t to represent each COPY.
 964 */
 965static int nfs4_init_cp_state(struct nfsd_net *nn, copy_stateid_t *stid,
 966                              unsigned char sc_type)
 967{
 968        int new_id;
 969
 970        stid->stid.si_opaque.so_clid.cl_boot = (u32)nn->boot_time;
 971        stid->stid.si_opaque.so_clid.cl_id = nn->s2s_cp_cl_id;
 972        stid->sc_type = sc_type;
 973
 974        idr_preload(GFP_KERNEL);
 975        spin_lock(&nn->s2s_cp_lock);
 976        new_id = idr_alloc_cyclic(&nn->s2s_cp_stateids, stid, 0, 0, GFP_NOWAIT);
 977        stid->stid.si_opaque.so_id = new_id;
 978        stid->stid.si_generation = 1;
 979        spin_unlock(&nn->s2s_cp_lock);
 980        idr_preload_end();
 981        if (new_id < 0)
 982                return 0;
 983        return 1;
 984}
 985
 986int nfs4_init_copy_state(struct nfsd_net *nn, struct nfsd4_copy *copy)
 987{
 988        return nfs4_init_cp_state(nn, &copy->cp_stateid, NFS4_COPY_STID);
 989}
 990
 991struct nfs4_cpntf_state *nfs4_alloc_init_cpntf_state(struct nfsd_net *nn,
 992                                                     struct nfs4_stid *p_stid)
 993{
 994        struct nfs4_cpntf_state *cps;
 995
 996        cps = kzalloc(sizeof(struct nfs4_cpntf_state), GFP_KERNEL);
 997        if (!cps)
 998                return NULL;
 999        cps->cpntf_time = ktime_get_boottime_seconds();
1000        refcount_set(&cps->cp_stateid.sc_count, 1);
1001        if (!nfs4_init_cp_state(nn, &cps->cp_stateid, NFS4_COPYNOTIFY_STID))
1002                goto out_free;
1003        spin_lock(&nn->s2s_cp_lock);
1004        list_add(&cps->cp_list, &p_stid->sc_cp_list);
1005        spin_unlock(&nn->s2s_cp_lock);
1006        return cps;
1007out_free:
1008        kfree(cps);
1009        return NULL;
1010}
1011
1012void nfs4_free_copy_state(struct nfsd4_copy *copy)
1013{
1014        struct nfsd_net *nn;
1015
1016        WARN_ON_ONCE(copy->cp_stateid.sc_type != NFS4_COPY_STID);
1017        nn = net_generic(copy->cp_clp->net, nfsd_net_id);
1018        spin_lock(&nn->s2s_cp_lock);
1019        idr_remove(&nn->s2s_cp_stateids,
1020                   copy->cp_stateid.stid.si_opaque.so_id);
1021        spin_unlock(&nn->s2s_cp_lock);
1022}
1023
1024static void nfs4_free_cpntf_statelist(struct net *net, struct nfs4_stid *stid)
1025{
1026        struct nfs4_cpntf_state *cps;
1027        struct nfsd_net *nn;
1028
1029        nn = net_generic(net, nfsd_net_id);
1030        spin_lock(&nn->s2s_cp_lock);
1031        while (!list_empty(&stid->sc_cp_list)) {
1032                cps = list_first_entry(&stid->sc_cp_list,
1033                                       struct nfs4_cpntf_state, cp_list);
1034                _free_cpntf_state_locked(nn, cps);
1035        }
1036        spin_unlock(&nn->s2s_cp_lock);
1037}
1038
1039static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
1040{
1041        struct nfs4_stid *stid;
1042
1043        stid = nfs4_alloc_stid(clp, stateid_slab, nfs4_free_ol_stateid);
1044        if (!stid)
1045                return NULL;
1046
1047        return openlockstateid(stid);
1048}
1049
1050static void nfs4_free_deleg(struct nfs4_stid *stid)
1051{
1052        kmem_cache_free(deleg_slab, stid);
1053        atomic_long_dec(&num_delegations);
1054}
1055
1056/*
1057 * When we recall a delegation, we should be careful not to hand it
1058 * out again straight away.
1059 * To ensure this we keep a pair of bloom filters ('new' and 'old')
1060 * in which the filehandles of recalled delegations are "stored".
1061 * If a filehandle appear in either filter, a delegation is blocked.
1062 * When a delegation is recalled, the filehandle is stored in the "new"
1063 * filter.
1064 * Every 30 seconds we swap the filters and clear the "new" one,
1065 * unless both are empty of course.
1066 *
1067 * Each filter is 256 bits.  We hash the filehandle to 32bit and use the
1068 * low 3 bytes as hash-table indices.
1069 *
1070 * 'blocked_delegations_lock', which is always taken in block_delegations(),
1071 * is used to manage concurrent access.  Testing does not need the lock
1072 * except when swapping the two filters.
1073 */
1074static DEFINE_SPINLOCK(blocked_delegations_lock);
1075static struct bloom_pair {
1076        int     entries, old_entries;
1077        time64_t swap_time;
1078        int     new; /* index into 'set' */
1079        DECLARE_BITMAP(set[2], 256);
1080} blocked_delegations;
1081
1082static int delegation_blocked(struct knfsd_fh *fh)
1083{
1084        u32 hash;
1085        struct bloom_pair *bd = &blocked_delegations;
1086
1087        if (bd->entries == 0)
1088                return 0;
1089        if (ktime_get_seconds() - bd->swap_time > 30) {
1090                spin_lock(&blocked_delegations_lock);
1091                if (ktime_get_seconds() - bd->swap_time > 30) {
1092                        bd->entries -= bd->old_entries;
1093                        bd->old_entries = bd->entries;
1094                        memset(bd->set[bd->new], 0,
1095                               sizeof(bd->set[0]));
1096                        bd->new = 1-bd->new;
1097                        bd->swap_time = ktime_get_seconds();
1098                }
1099                spin_unlock(&blocked_delegations_lock);
1100        }
1101        hash = jhash(&fh->fh_raw, fh->fh_size, 0);
1102        if (test_bit(hash&255, bd->set[0]) &&
1103            test_bit((hash>>8)&255, bd->set[0]) &&
1104            test_bit((hash>>16)&255, bd->set[0]))
1105                return 1;
1106
1107        if (test_bit(hash&255, bd->set[1]) &&
1108            test_bit((hash>>8)&255, bd->set[1]) &&
1109            test_bit((hash>>16)&255, bd->set[1]))
1110                return 1;
1111
1112        return 0;
1113}
1114
1115static void block_delegations(struct knfsd_fh *fh)
1116{
1117        u32 hash;
1118        struct bloom_pair *bd = &blocked_delegations;
1119
1120        hash = jhash(&fh->fh_raw, fh->fh_size, 0);
1121
1122        spin_lock(&blocked_delegations_lock);
1123        __set_bit(hash&255, bd->set[bd->new]);
1124        __set_bit((hash>>8)&255, bd->set[bd->new]);
1125        __set_bit((hash>>16)&255, bd->set[bd->new]);
1126        if (bd->entries == 0)
1127                bd->swap_time = ktime_get_seconds();
1128        bd->entries += 1;
1129        spin_unlock(&blocked_delegations_lock);
1130}
1131
1132static struct nfs4_delegation *
1133alloc_init_deleg(struct nfs4_client *clp, struct nfs4_file *fp,
1134                 struct svc_fh *current_fh,
1135                 struct nfs4_clnt_odstate *odstate)
1136{
1137        struct nfs4_delegation *dp;
1138        long n;
1139
1140        dprintk("NFSD alloc_init_deleg\n");
1141        n = atomic_long_inc_return(&num_delegations);
1142        if (n < 0 || n > max_delegations)
1143                goto out_dec;
1144        if (delegation_blocked(&current_fh->fh_handle))
1145                goto out_dec;
1146        dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab, nfs4_free_deleg));
1147        if (dp == NULL)
1148                goto out_dec;
1149
1150        /*
1151         * delegation seqid's are never incremented.  The 4.1 special
1152         * meaning of seqid 0 isn't meaningful, really, but let's avoid
1153         * 0 anyway just for consistency and use 1:
1154         */
1155        dp->dl_stid.sc_stateid.si_generation = 1;
1156        INIT_LIST_HEAD(&dp->dl_perfile);
1157        INIT_LIST_HEAD(&dp->dl_perclnt);
1158        INIT_LIST_HEAD(&dp->dl_recall_lru);
1159        dp->dl_clnt_odstate = odstate;
1160        get_clnt_odstate(odstate);
1161        dp->dl_type = NFS4_OPEN_DELEGATE_READ;
1162        dp->dl_retries = 1;
1163        dp->dl_recalled = false;
1164        nfsd4_init_cb(&dp->dl_recall, dp->dl_stid.sc_client,
1165                      &nfsd4_cb_recall_ops, NFSPROC4_CLNT_CB_RECALL);
1166        get_nfs4_file(fp);
1167        dp->dl_stid.sc_file = fp;
1168        return dp;
1169out_dec:
1170        atomic_long_dec(&num_delegations);
1171        return NULL;
1172}
1173
1174void
1175nfs4_put_stid(struct nfs4_stid *s)
1176{
1177        struct nfs4_file *fp = s->sc_file;
1178        struct nfs4_client *clp = s->sc_client;
1179
1180        might_lock(&clp->cl_lock);
1181
1182        if (!refcount_dec_and_lock(&s->sc_count, &clp->cl_lock)) {
1183                wake_up_all(&close_wq);
1184                return;
1185        }
1186        idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1187        nfs4_free_cpntf_statelist(clp->net, s);
1188        spin_unlock(&clp->cl_lock);
1189        s->sc_free(s);
1190        if (fp)
1191                put_nfs4_file(fp);
1192}
1193
1194void
1195nfs4_inc_and_copy_stateid(stateid_t *dst, struct nfs4_stid *stid)
1196{
1197        stateid_t *src = &stid->sc_stateid;
1198
1199        spin_lock(&stid->sc_lock);
1200        if (unlikely(++src->si_generation == 0))
1201                src->si_generation = 1;
1202        memcpy(dst, src, sizeof(*dst));
1203        spin_unlock(&stid->sc_lock);
1204}
1205
1206static void put_deleg_file(struct nfs4_file *fp)
1207{
1208        struct nfsd_file *nf = NULL;
1209
1210        spin_lock(&fp->fi_lock);
1211        if (--fp->fi_delegees == 0)
1212                swap(nf, fp->fi_deleg_file);
1213        spin_unlock(&fp->fi_lock);
1214
1215        if (nf)
1216                nfsd_file_put(nf);
1217}
1218
1219static void nfs4_unlock_deleg_lease(struct nfs4_delegation *dp)
1220{
1221        struct nfs4_file *fp = dp->dl_stid.sc_file;
1222        struct nfsd_file *nf = fp->fi_deleg_file;
1223
1224        WARN_ON_ONCE(!fp->fi_delegees);
1225
1226        vfs_setlease(nf->nf_file, F_UNLCK, NULL, (void **)&dp);
1227        put_deleg_file(fp);
1228}
1229
1230static void destroy_unhashed_deleg(struct nfs4_delegation *dp)
1231{
1232        put_clnt_odstate(dp->dl_clnt_odstate);
1233        nfs4_unlock_deleg_lease(dp);
1234        nfs4_put_stid(&dp->dl_stid);
1235}
1236
1237void nfs4_unhash_stid(struct nfs4_stid *s)
1238{
1239        s->sc_type = 0;
1240}
1241
1242/**
1243 * nfs4_delegation_exists - Discover if this delegation already exists
1244 * @clp:     a pointer to the nfs4_client we're granting a delegation to
1245 * @fp:      a pointer to the nfs4_file we're granting a delegation on
1246 *
1247 * Return:
1248 *      On success: true iff an existing delegation is found
1249 */
1250
1251static bool
1252nfs4_delegation_exists(struct nfs4_client *clp, struct nfs4_file *fp)
1253{
1254        struct nfs4_delegation *searchdp = NULL;
1255        struct nfs4_client *searchclp = NULL;
1256
1257        lockdep_assert_held(&state_lock);
1258        lockdep_assert_held(&fp->fi_lock);
1259
1260        list_for_each_entry(searchdp, &fp->fi_delegations, dl_perfile) {
1261                searchclp = searchdp->dl_stid.sc_client;
1262                if (clp == searchclp) {
1263                        return true;
1264                }
1265        }
1266        return false;
1267}
1268
1269/**
1270 * hash_delegation_locked - Add a delegation to the appropriate lists
1271 * @dp:     a pointer to the nfs4_delegation we are adding.
1272 * @fp:     a pointer to the nfs4_file we're granting a delegation on
1273 *
1274 * Return:
1275 *      On success: NULL if the delegation was successfully hashed.
1276 *
1277 *      On error: -EAGAIN if one was previously granted to this
1278 *                 nfs4_client for this nfs4_file. Delegation is not hashed.
1279 *
1280 */
1281
1282static int
1283hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
1284{
1285        struct nfs4_client *clp = dp->dl_stid.sc_client;
1286
1287        lockdep_assert_held(&state_lock);
1288        lockdep_assert_held(&fp->fi_lock);
1289
1290        if (nfs4_delegation_exists(clp, fp))
1291                return -EAGAIN;
1292        refcount_inc(&dp->dl_stid.sc_count);
1293        dp->dl_stid.sc_type = NFS4_DELEG_STID;
1294        list_add(&dp->dl_perfile, &fp->fi_delegations);
1295        list_add(&dp->dl_perclnt, &clp->cl_delegations);
1296        return 0;
1297}
1298
1299static bool delegation_hashed(struct nfs4_delegation *dp)
1300{
1301        return !(list_empty(&dp->dl_perfile));
1302}
1303
1304static bool
1305unhash_delegation_locked(struct nfs4_delegation *dp)
1306{
1307        struct nfs4_file *fp = dp->dl_stid.sc_file;
1308
1309        lockdep_assert_held(&state_lock);
1310
1311        if (!delegation_hashed(dp))
1312                return false;
1313
1314        dp->dl_stid.sc_type = NFS4_CLOSED_DELEG_STID;
1315        /* Ensure that deleg break won't try to requeue it */
1316        ++dp->dl_time;
1317        spin_lock(&fp->fi_lock);
1318        list_del_init(&dp->dl_perclnt);
1319        list_del_init(&dp->dl_recall_lru);
1320        list_del_init(&dp->dl_perfile);
1321        spin_unlock(&fp->fi_lock);
1322        return true;
1323}
1324
1325static void destroy_delegation(struct nfs4_delegation *dp)
1326{
1327        bool unhashed;
1328
1329        spin_lock(&state_lock);
1330        unhashed = unhash_delegation_locked(dp);
1331        spin_unlock(&state_lock);
1332        if (unhashed)
1333                destroy_unhashed_deleg(dp);
1334}
1335
1336static void revoke_delegation(struct nfs4_delegation *dp)
1337{
1338        struct nfs4_client *clp = dp->dl_stid.sc_client;
1339
1340        WARN_ON(!list_empty(&dp->dl_recall_lru));
1341
1342        if (clp->cl_minorversion) {
1343                dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
1344                refcount_inc(&dp->dl_stid.sc_count);
1345                spin_lock(&clp->cl_lock);
1346                list_add(&dp->dl_recall_lru, &clp->cl_revoked);
1347                spin_unlock(&clp->cl_lock);
1348        }
1349        destroy_unhashed_deleg(dp);
1350}
1351
1352/* 
1353 * SETCLIENTID state 
1354 */
1355
1356static unsigned int clientid_hashval(u32 id)
1357{
1358        return id & CLIENT_HASH_MASK;
1359}
1360
1361static unsigned int clientstr_hashval(struct xdr_netobj name)
1362{
1363        return opaque_hashval(name.data, 8) & CLIENT_HASH_MASK;
1364}
1365
1366/*
1367 * A stateid that had a deny mode associated with it is being released
1368 * or downgraded. Recalculate the deny mode on the file.
1369 */
1370static void
1371recalculate_deny_mode(struct nfs4_file *fp)
1372{
1373        struct nfs4_ol_stateid *stp;
1374
1375        spin_lock(&fp->fi_lock);
1376        fp->fi_share_deny = 0;
1377        list_for_each_entry(stp, &fp->fi_stateids, st_perfile)
1378                fp->fi_share_deny |= bmap_to_share_mode(stp->st_deny_bmap);
1379        spin_unlock(&fp->fi_lock);
1380}
1381
1382static void
1383reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
1384{
1385        int i;
1386        bool change = false;
1387
1388        for (i = 1; i < 4; i++) {
1389                if ((i & deny) != i) {
1390                        change = true;
1391                        clear_deny(i, stp);
1392                }
1393        }
1394
1395        /* Recalculate per-file deny mode if there was a change */
1396        if (change)
1397                recalculate_deny_mode(stp->st_stid.sc_file);
1398}
1399
1400/* release all access and file references for a given stateid */
1401static void
1402release_all_access(struct nfs4_ol_stateid *stp)
1403{
1404        int i;
1405        struct nfs4_file *fp = stp->st_stid.sc_file;
1406
1407        if (fp && stp->st_deny_bmap != 0)
1408                recalculate_deny_mode(fp);
1409
1410        for (i = 1; i < 4; i++) {
1411                if (test_access(i, stp))
1412                        nfs4_file_put_access(stp->st_stid.sc_file, i);
1413                clear_access(i, stp);
1414        }
1415}
1416
1417static inline void nfs4_free_stateowner(struct nfs4_stateowner *sop)
1418{
1419        kfree(sop->so_owner.data);
1420        sop->so_ops->so_free(sop);
1421}
1422
1423static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
1424{
1425        struct nfs4_client *clp = sop->so_client;
1426
1427        might_lock(&clp->cl_lock);
1428
1429        if (!atomic_dec_and_lock(&sop->so_count, &clp->cl_lock))
1430                return;
1431        sop->so_ops->so_unhash(sop);
1432        spin_unlock(&clp->cl_lock);
1433        nfs4_free_stateowner(sop);
1434}
1435
1436static bool
1437nfs4_ol_stateid_unhashed(const struct nfs4_ol_stateid *stp)
1438{
1439        return list_empty(&stp->st_perfile);
1440}
1441
1442static bool unhash_ol_stateid(struct nfs4_ol_stateid *stp)
1443{
1444        struct nfs4_file *fp = stp->st_stid.sc_file;
1445
1446        lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
1447
1448        if (list_empty(&stp->st_perfile))
1449                return false;
1450
1451        spin_lock(&fp->fi_lock);
1452        list_del_init(&stp->st_perfile);
1453        spin_unlock(&fp->fi_lock);
1454        list_del(&stp->st_perstateowner);
1455        return true;
1456}
1457
1458static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
1459{
1460        struct nfs4_ol_stateid *stp = openlockstateid(stid);
1461
1462        put_clnt_odstate(stp->st_clnt_odstate);
1463        release_all_access(stp);
1464        if (stp->st_stateowner)
1465                nfs4_put_stateowner(stp->st_stateowner);
1466        kmem_cache_free(stateid_slab, stid);
1467}
1468
1469static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
1470{
1471        struct nfs4_ol_stateid *stp = openlockstateid(stid);
1472        struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
1473        struct nfsd_file *nf;
1474
1475        nf = find_any_file(stp->st_stid.sc_file);
1476        if (nf) {
1477                get_file(nf->nf_file);
1478                filp_close(nf->nf_file, (fl_owner_t)lo);
1479                nfsd_file_put(nf);
1480        }
1481        nfs4_free_ol_stateid(stid);
1482}
1483
1484/*
1485 * Put the persistent reference to an already unhashed generic stateid, while
1486 * holding the cl_lock. If it's the last reference, then put it onto the
1487 * reaplist for later destruction.
1488 */
1489static void put_ol_stateid_locked(struct nfs4_ol_stateid *stp,
1490                                       struct list_head *reaplist)
1491{
1492        struct nfs4_stid *s = &stp->st_stid;
1493        struct nfs4_client *clp = s->sc_client;
1494
1495        lockdep_assert_held(&clp->cl_lock);
1496
1497        WARN_ON_ONCE(!list_empty(&stp->st_locks));
1498
1499        if (!refcount_dec_and_test(&s->sc_count)) {
1500                wake_up_all(&close_wq);
1501                return;
1502        }
1503
1504        idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1505        list_add(&stp->st_locks, reaplist);
1506}
1507
1508static bool unhash_lock_stateid(struct nfs4_ol_stateid *stp)
1509{
1510        lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1511
1512        if (!unhash_ol_stateid(stp))
1513                return false;
1514        list_del_init(&stp->st_locks);
1515        nfs4_unhash_stid(&stp->st_stid);
1516        return true;
1517}
1518
1519static void release_lock_stateid(struct nfs4_ol_stateid *stp)
1520{
1521        struct nfs4_client *clp = stp->st_stid.sc_client;
1522        bool unhashed;
1523
1524        spin_lock(&clp->cl_lock);
1525        unhashed = unhash_lock_stateid(stp);
1526        spin_unlock(&clp->cl_lock);
1527        if (unhashed)
1528                nfs4_put_stid(&stp->st_stid);
1529}
1530
1531static void unhash_lockowner_locked(struct nfs4_lockowner *lo)
1532{
1533        struct nfs4_client *clp = lo->lo_owner.so_client;
1534
1535        lockdep_assert_held(&clp->cl_lock);
1536
1537        list_del_init(&lo->lo_owner.so_strhash);
1538}
1539
1540/*
1541 * Free a list of generic stateids that were collected earlier after being
1542 * fully unhashed.
1543 */
1544static void
1545free_ol_stateid_reaplist(struct list_head *reaplist)
1546{
1547        struct nfs4_ol_stateid *stp;
1548        struct nfs4_file *fp;
1549
1550        might_sleep();
1551
1552        while (!list_empty(reaplist)) {
1553                stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
1554                                       st_locks);
1555                list_del(&stp->st_locks);
1556                fp = stp->st_stid.sc_file;
1557                stp->st_stid.sc_free(&stp->st_stid);
1558                if (fp)
1559                        put_nfs4_file(fp);
1560        }
1561}
1562
1563static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp,
1564                                       struct list_head *reaplist)
1565{
1566        struct nfs4_ol_stateid *stp;
1567
1568        lockdep_assert_held(&open_stp->st_stid.sc_client->cl_lock);
1569
1570        while (!list_empty(&open_stp->st_locks)) {
1571                stp = list_entry(open_stp->st_locks.next,
1572                                struct nfs4_ol_stateid, st_locks);
1573                WARN_ON(!unhash_lock_stateid(stp));
1574                put_ol_stateid_locked(stp, reaplist);
1575        }
1576}
1577
1578static bool unhash_open_stateid(struct nfs4_ol_stateid *stp,
1579                                struct list_head *reaplist)
1580{
1581        lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1582
1583        if (!unhash_ol_stateid(stp))
1584                return false;
1585        release_open_stateid_locks(stp, reaplist);
1586        return true;
1587}
1588
1589static void release_open_stateid(struct nfs4_ol_stateid *stp)
1590{
1591        LIST_HEAD(reaplist);
1592
1593        spin_lock(&stp->st_stid.sc_client->cl_lock);
1594        if (unhash_open_stateid(stp, &reaplist))
1595                put_ol_stateid_locked(stp, &reaplist);
1596        spin_unlock(&stp->st_stid.sc_client->cl_lock);
1597        free_ol_stateid_reaplist(&reaplist);
1598}
1599
1600static void unhash_openowner_locked(struct nfs4_openowner *oo)
1601{
1602        struct nfs4_client *clp = oo->oo_owner.so_client;
1603
1604        lockdep_assert_held(&clp->cl_lock);
1605
1606        list_del_init(&oo->oo_owner.so_strhash);
1607        list_del_init(&oo->oo_perclient);
1608}
1609
1610static void release_last_closed_stateid(struct nfs4_openowner *oo)
1611{
1612        struct nfsd_net *nn = net_generic(oo->oo_owner.so_client->net,
1613                                          nfsd_net_id);
1614        struct nfs4_ol_stateid *s;
1615
1616        spin_lock(&nn->client_lock);
1617        s = oo->oo_last_closed_stid;
1618        if (s) {
1619                list_del_init(&oo->oo_close_lru);
1620                oo->oo_last_closed_stid = NULL;
1621        }
1622        spin_unlock(&nn->client_lock);
1623        if (s)
1624                nfs4_put_stid(&s->st_stid);
1625}
1626
1627static void release_openowner(struct nfs4_openowner *oo)
1628{
1629        struct nfs4_ol_stateid *stp;
1630        struct nfs4_client *clp = oo->oo_owner.so_client;
1631        struct list_head reaplist;
1632
1633        INIT_LIST_HEAD(&reaplist);
1634
1635        spin_lock(&clp->cl_lock);
1636        unhash_openowner_locked(oo);
1637        while (!list_empty(&oo->oo_owner.so_stateids)) {
1638                stp = list_first_entry(&oo->oo_owner.so_stateids,
1639                                struct nfs4_ol_stateid, st_perstateowner);
1640                if (unhash_open_stateid(stp, &reaplist))
1641                        put_ol_stateid_locked(stp, &reaplist);
1642        }
1643        spin_unlock(&clp->cl_lock);
1644        free_ol_stateid_reaplist(&reaplist);
1645        release_last_closed_stateid(oo);
1646        nfs4_put_stateowner(&oo->oo_owner);
1647}
1648
1649static inline int
1650hash_sessionid(struct nfs4_sessionid *sessionid)
1651{
1652        struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1653
1654        return sid->sequence % SESSION_HASH_SIZE;
1655}
1656
1657#ifdef CONFIG_SUNRPC_DEBUG
1658static inline void
1659dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1660{
1661        u32 *ptr = (u32 *)(&sessionid->data[0]);
1662        dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1663}
1664#else
1665static inline void
1666dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1667{
1668}
1669#endif
1670
1671/*
1672 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1673 * won't be used for replay.
1674 */
1675void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1676{
1677        struct nfs4_stateowner *so = cstate->replay_owner;
1678
1679        if (nfserr == nfserr_replay_me)
1680                return;
1681
1682        if (!seqid_mutating_err(ntohl(nfserr))) {
1683                nfsd4_cstate_clear_replay(cstate);
1684                return;
1685        }
1686        if (!so)
1687                return;
1688        if (so->so_is_open_owner)
1689                release_last_closed_stateid(openowner(so));
1690        so->so_seqid++;
1691        return;
1692}
1693
1694static void
1695gen_sessionid(struct nfsd4_session *ses)
1696{
1697        struct nfs4_client *clp = ses->se_client;
1698        struct nfsd4_sessionid *sid;
1699
1700        sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1701        sid->clientid = clp->cl_clientid;
1702        sid->sequence = current_sessionid++;
1703        sid->reserved = 0;
1704}
1705
1706/*
1707 * The protocol defines ca_maxresponssize_cached to include the size of
1708 * the rpc header, but all we need to cache is the data starting after
1709 * the end of the initial SEQUENCE operation--the rest we regenerate
1710 * each time.  Therefore we can advertise a ca_maxresponssize_cached
1711 * value that is the number of bytes in our cache plus a few additional
1712 * bytes.  In order to stay on the safe side, and not promise more than
1713 * we can cache, those additional bytes must be the minimum possible: 24
1714 * bytes of rpc header (xid through accept state, with AUTH_NULL
1715 * verifier), 12 for the compound header (with zero-length tag), and 44
1716 * for the SEQUENCE op response:
1717 */
1718#define NFSD_MIN_HDR_SEQ_SZ  (24 + 12 + 44)
1719
1720static void
1721free_session_slots(struct nfsd4_session *ses)
1722{
1723        int i;
1724
1725        for (i = 0; i < ses->se_fchannel.maxreqs; i++) {
1726                free_svc_cred(&ses->se_slots[i]->sl_cred);
1727                kfree(ses->se_slots[i]);
1728        }
1729}
1730
1731/*
1732 * We don't actually need to cache the rpc and session headers, so we
1733 * can allocate a little less for each slot:
1734 */
1735static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
1736{
1737        u32 size;
1738
1739        if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
1740                size = 0;
1741        else
1742                size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
1743        return size + sizeof(struct nfsd4_slot);
1744}
1745
1746/*
1747 * XXX: If we run out of reserved DRC memory we could (up to a point)
1748 * re-negotiate active sessions and reduce their slot usage to make
1749 * room for new connections. For now we just fail the create session.
1750 */
1751static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
1752{
1753        u32 slotsize = slot_bytes(ca);
1754        u32 num = ca->maxreqs;
1755        unsigned long avail, total_avail;
1756        unsigned int scale_factor;
1757
1758        spin_lock(&nfsd_drc_lock);
1759        if (nfsd_drc_max_mem > nfsd_drc_mem_used)
1760                total_avail = nfsd_drc_max_mem - nfsd_drc_mem_used;
1761        else
1762                /* We have handed out more space than we chose in
1763                 * set_max_drc() to allow.  That isn't really a
1764                 * problem as long as that doesn't make us think we
1765                 * have lots more due to integer overflow.
1766                 */
1767                total_avail = 0;
1768        avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION, total_avail);
1769        /*
1770         * Never use more than a fraction of the remaining memory,
1771         * unless it's the only way to give this client a slot.
1772         * The chosen fraction is either 1/8 or 1/number of threads,
1773         * whichever is smaller.  This ensures there are adequate
1774         * slots to support multiple clients per thread.
1775         * Give the client one slot even if that would require
1776         * over-allocation--it is better than failure.
1777         */
1778        scale_factor = max_t(unsigned int, 8, nn->nfsd_serv->sv_nrthreads);
1779
1780        avail = clamp_t(unsigned long, avail, slotsize,
1781                        total_avail/scale_factor);
1782        num = min_t(int, num, avail / slotsize);
1783        num = max_t(int, num, 1);
1784        nfsd_drc_mem_used += num * slotsize;
1785        spin_unlock(&nfsd_drc_lock);
1786
1787        return num;
1788}
1789
1790static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
1791{
1792        int slotsize = slot_bytes(ca);
1793
1794        spin_lock(&nfsd_drc_lock);
1795        nfsd_drc_mem_used -= slotsize * ca->maxreqs;
1796        spin_unlock(&nfsd_drc_lock);
1797}
1798
1799static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
1800                                           struct nfsd4_channel_attrs *battrs)
1801{
1802        int numslots = fattrs->maxreqs;
1803        int slotsize = slot_bytes(fattrs);
1804        struct nfsd4_session *new;
1805        int mem, i;
1806
1807        BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
1808                        + sizeof(struct nfsd4_session) > PAGE_SIZE);
1809        mem = numslots * sizeof(struct nfsd4_slot *);
1810
1811        new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
1812        if (!new)
1813                return NULL;
1814        /* allocate each struct nfsd4_slot and data cache in one piece */
1815        for (i = 0; i < numslots; i++) {
1816                new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
1817                if (!new->se_slots[i])
1818                        goto out_free;
1819        }
1820
1821        memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
1822        memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
1823
1824        return new;
1825out_free:
1826        while (i--)
1827                kfree(new->se_slots[i]);
1828        kfree(new);
1829        return NULL;
1830}
1831
1832static void free_conn(struct nfsd4_conn *c)
1833{
1834        svc_xprt_put(c->cn_xprt);
1835        kfree(c);
1836}
1837
1838static void nfsd4_conn_lost(struct svc_xpt_user *u)
1839{
1840        struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
1841        struct nfs4_client *clp = c->cn_session->se_client;
1842
1843        trace_nfsd_cb_lost(clp);
1844
1845        spin_lock(&clp->cl_lock);
1846        if (!list_empty(&c->cn_persession)) {
1847                list_del(&c->cn_persession);
1848                free_conn(c);
1849        }
1850        nfsd4_probe_callback(clp);
1851        spin_unlock(&clp->cl_lock);
1852}
1853
1854static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1855{
1856        struct nfsd4_conn *conn;
1857
1858        conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1859        if (!conn)
1860                return NULL;
1861        svc_xprt_get(rqstp->rq_xprt);
1862        conn->cn_xprt = rqstp->rq_xprt;
1863        conn->cn_flags = flags;
1864        INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1865        return conn;
1866}
1867
1868static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1869{
1870        conn->cn_session = ses;
1871        list_add(&conn->cn_persession, &ses->se_conns);
1872}
1873
1874static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1875{
1876        struct nfs4_client *clp = ses->se_client;
1877
1878        spin_lock(&clp->cl_lock);
1879        __nfsd4_hash_conn(conn, ses);
1880        spin_unlock(&clp->cl_lock);
1881}
1882
1883static int nfsd4_register_conn(struct nfsd4_conn *conn)
1884{
1885        conn->cn_xpt_user.callback = nfsd4_conn_lost;
1886        return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1887}
1888
1889static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1890{
1891        int ret;
1892
1893        nfsd4_hash_conn(conn, ses);
1894        ret = nfsd4_register_conn(conn);
1895        if (ret)
1896                /* oops; xprt is already down: */
1897                nfsd4_conn_lost(&conn->cn_xpt_user);
1898        /* We may have gained or lost a callback channel: */
1899        nfsd4_probe_callback_sync(ses->se_client);
1900}
1901
1902static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1903{
1904        u32 dir = NFS4_CDFC4_FORE;
1905
1906        if (cses->flags & SESSION4_BACK_CHAN)
1907                dir |= NFS4_CDFC4_BACK;
1908        return alloc_conn(rqstp, dir);
1909}
1910
1911/* must be called under client_lock */
1912static void nfsd4_del_conns(struct nfsd4_session *s)
1913{
1914        struct nfs4_client *clp = s->se_client;
1915        struct nfsd4_conn *c;
1916
1917        spin_lock(&clp->cl_lock);
1918        while (!list_empty(&s->se_conns)) {
1919                c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1920                list_del_init(&c->cn_persession);
1921                spin_unlock(&clp->cl_lock);
1922
1923                unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1924                free_conn(c);
1925
1926                spin_lock(&clp->cl_lock);
1927        }
1928        spin_unlock(&clp->cl_lock);
1929}
1930
1931static void __free_session(struct nfsd4_session *ses)
1932{
1933        free_session_slots(ses);
1934        kfree(ses);
1935}
1936
1937static void free_session(struct nfsd4_session *ses)
1938{
1939        nfsd4_del_conns(ses);
1940        nfsd4_put_drc_mem(&ses->se_fchannel);
1941        __free_session(ses);
1942}
1943
1944static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1945{
1946        int idx;
1947        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1948
1949        new->se_client = clp;
1950        gen_sessionid(new);
1951
1952        INIT_LIST_HEAD(&new->se_conns);
1953
1954        new->se_cb_seq_nr = 1;
1955        new->se_flags = cses->flags;
1956        new->se_cb_prog = cses->callback_prog;
1957        new->se_cb_sec = cses->cb_sec;
1958        atomic_set(&new->se_ref, 0);
1959        idx = hash_sessionid(&new->se_sessionid);
1960        list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1961        spin_lock(&clp->cl_lock);
1962        list_add(&new->se_perclnt, &clp->cl_sessions);
1963        spin_unlock(&clp->cl_lock);
1964
1965        {
1966                struct sockaddr *sa = svc_addr(rqstp);
1967                /*
1968                 * This is a little silly; with sessions there's no real
1969                 * use for the callback address.  Use the peer address
1970                 * as a reasonable default for now, but consider fixing
1971                 * the rpc client not to require an address in the
1972                 * future:
1973                 */
1974                rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1975                clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1976        }
1977}
1978
1979/* caller must hold client_lock */
1980static struct nfsd4_session *
1981__find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1982{
1983        struct nfsd4_session *elem;
1984        int idx;
1985        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1986
1987        lockdep_assert_held(&nn->client_lock);
1988
1989        dump_sessionid(__func__, sessionid);
1990        idx = hash_sessionid(sessionid);
1991        /* Search in the appropriate list */
1992        list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1993                if (!memcmp(elem->se_sessionid.data, sessionid->data,
1994                            NFS4_MAX_SESSIONID_LEN)) {
1995                        return elem;
1996                }
1997        }
1998
1999        dprintk("%s: session not found\n", __func__);
2000        return NULL;
2001}
2002
2003static struct nfsd4_session *
2004find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
2005                __be32 *ret)
2006{
2007        struct nfsd4_session *session;
2008        __be32 status = nfserr_badsession;
2009
2010        session = __find_in_sessionid_hashtbl(sessionid, net);
2011        if (!session)
2012                goto out;
2013        status = nfsd4_get_session_locked(session);
2014        if (status)
2015                session = NULL;
2016out:
2017        *ret = status;
2018        return session;
2019}
2020
2021/* caller must hold client_lock */
2022static void
2023unhash_session(struct nfsd4_session *ses)
2024{
2025        struct nfs4_client *clp = ses->se_client;
2026        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2027
2028        lockdep_assert_held(&nn->client_lock);
2029
2030        list_del(&ses->se_hash);
2031        spin_lock(&ses->se_client->cl_lock);
2032        list_del(&ses->se_perclnt);
2033        spin_unlock(&ses->se_client->cl_lock);
2034}
2035
2036/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
2037static int
2038STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
2039{
2040        /*
2041         * We're assuming the clid was not given out from a boot
2042         * precisely 2^32 (about 136 years) before this one.  That seems
2043         * a safe assumption:
2044         */
2045        if (clid->cl_boot == (u32)nn->boot_time)
2046                return 0;
2047        trace_nfsd_clid_stale(clid);
2048        return 1;
2049}
2050
2051/* 
2052 * XXX Should we use a slab cache ?
2053 * This type of memory management is somewhat inefficient, but we use it
2054 * anyway since SETCLIENTID is not a common operation.
2055 */
2056static struct nfs4_client *alloc_client(struct xdr_netobj name)
2057{
2058        struct nfs4_client *clp;
2059        int i;
2060
2061        clp = kmem_cache_zalloc(client_slab, GFP_KERNEL);
2062        if (clp == NULL)
2063                return NULL;
2064        xdr_netobj_dup(&clp->cl_name, &name, GFP_KERNEL);
2065        if (clp->cl_name.data == NULL)
2066                goto err_no_name;
2067        clp->cl_ownerstr_hashtbl = kmalloc_array(OWNER_HASH_SIZE,
2068                                                 sizeof(struct list_head),
2069                                                 GFP_KERNEL);
2070        if (!clp->cl_ownerstr_hashtbl)
2071                goto err_no_hashtbl;
2072        for (i = 0; i < OWNER_HASH_SIZE; i++)
2073                INIT_LIST_HEAD(&clp->cl_ownerstr_hashtbl[i]);
2074        INIT_LIST_HEAD(&clp->cl_sessions);
2075        idr_init(&clp->cl_stateids);
2076        atomic_set(&clp->cl_rpc_users, 0);
2077        clp->cl_cb_state = NFSD4_CB_UNKNOWN;
2078        clp->cl_state = NFSD4_ACTIVE;
2079        atomic_set(&clp->cl_delegs_in_recall, 0);
2080        INIT_LIST_HEAD(&clp->cl_idhash);
2081        INIT_LIST_HEAD(&clp->cl_openowners);
2082        INIT_LIST_HEAD(&clp->cl_delegations);
2083        INIT_LIST_HEAD(&clp->cl_lru);
2084        INIT_LIST_HEAD(&clp->cl_revoked);
2085#ifdef CONFIG_NFSD_PNFS
2086        INIT_LIST_HEAD(&clp->cl_lo_states);
2087#endif
2088        INIT_LIST_HEAD(&clp->async_copies);
2089        spin_lock_init(&clp->async_lock);
2090        spin_lock_init(&clp->cl_lock);
2091        rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
2092        return clp;
2093err_no_hashtbl:
2094        kfree(clp->cl_name.data);
2095err_no_name:
2096        kmem_cache_free(client_slab, clp);
2097        return NULL;
2098}
2099
2100static void __free_client(struct kref *k)
2101{
2102        struct nfsdfs_client *c = container_of(k, struct nfsdfs_client, cl_ref);
2103        struct nfs4_client *clp = container_of(c, struct nfs4_client, cl_nfsdfs);
2104
2105        free_svc_cred(&clp->cl_cred);
2106        kfree(clp->cl_ownerstr_hashtbl);
2107        kfree(clp->cl_name.data);
2108        kfree(clp->cl_nii_domain.data);
2109        kfree(clp->cl_nii_name.data);
2110        idr_destroy(&clp->cl_stateids);
2111        kmem_cache_free(client_slab, clp);
2112}
2113
2114static void drop_client(struct nfs4_client *clp)
2115{
2116        kref_put(&clp->cl_nfsdfs.cl_ref, __free_client);
2117}
2118
2119static void
2120free_client(struct nfs4_client *clp)
2121{
2122        while (!list_empty(&clp->cl_sessions)) {
2123                struct nfsd4_session *ses;
2124                ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
2125                                se_perclnt);
2126                list_del(&ses->se_perclnt);
2127                WARN_ON_ONCE(atomic_read(&ses->se_ref));
2128                free_session(ses);
2129        }
2130        rpc_destroy_wait_queue(&clp->cl_cb_waitq);
2131        if (clp->cl_nfsd_dentry) {
2132                nfsd_client_rmdir(clp->cl_nfsd_dentry);
2133                clp->cl_nfsd_dentry = NULL;
2134                wake_up_all(&expiry_wq);
2135        }
2136        drop_client(clp);
2137}
2138
2139/* must be called under the client_lock */
2140static void
2141unhash_client_locked(struct nfs4_client *clp)
2142{
2143        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2144        struct nfsd4_session *ses;
2145
2146        lockdep_assert_held(&nn->client_lock);
2147
2148        /* Mark the client as expired! */
2149        clp->cl_time = 0;
2150        /* Make it invisible */
2151        if (!list_empty(&clp->cl_idhash)) {
2152                list_del_init(&clp->cl_idhash);
2153                if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
2154                        rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
2155                else
2156                        rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2157        }
2158        list_del_init(&clp->cl_lru);
2159        spin_lock(&clp->cl_lock);
2160        list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
2161                list_del_init(&ses->se_hash);
2162        spin_unlock(&clp->cl_lock);
2163}
2164
2165static void
2166unhash_client(struct nfs4_client *clp)
2167{
2168        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2169
2170        spin_lock(&nn->client_lock);
2171        unhash_client_locked(clp);
2172        spin_unlock(&nn->client_lock);
2173}
2174
2175static __be32 mark_client_expired_locked(struct nfs4_client *clp)
2176{
2177        if (atomic_read(&clp->cl_rpc_users))
2178                return nfserr_jukebox;
2179        unhash_client_locked(clp);
2180        return nfs_ok;
2181}
2182
2183static void
2184__destroy_client(struct nfs4_client *clp)
2185{
2186        int i;
2187        struct nfs4_openowner *oo;
2188        struct nfs4_delegation *dp;
2189        struct list_head reaplist;
2190
2191        INIT_LIST_HEAD(&reaplist);
2192        spin_lock(&state_lock);
2193        while (!list_empty(&clp->cl_delegations)) {
2194                dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
2195                WARN_ON(!unhash_delegation_locked(dp));
2196                list_add(&dp->dl_recall_lru, &reaplist);
2197        }
2198        spin_unlock(&state_lock);
2199        while (!list_empty(&reaplist)) {
2200                dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
2201                list_del_init(&dp->dl_recall_lru);
2202                destroy_unhashed_deleg(dp);
2203        }
2204        while (!list_empty(&clp->cl_revoked)) {
2205                dp = list_entry(clp->cl_revoked.next, struct nfs4_delegation, dl_recall_lru);
2206                list_del_init(&dp->dl_recall_lru);
2207                nfs4_put_stid(&dp->dl_stid);
2208        }
2209        while (!list_empty(&clp->cl_openowners)) {
2210                oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
2211                nfs4_get_stateowner(&oo->oo_owner);
2212                release_openowner(oo);
2213        }
2214        for (i = 0; i < OWNER_HASH_SIZE; i++) {
2215                struct nfs4_stateowner *so, *tmp;
2216
2217                list_for_each_entry_safe(so, tmp, &clp->cl_ownerstr_hashtbl[i],
2218                                         so_strhash) {
2219                        /* Should be no openowners at this point */
2220                        WARN_ON_ONCE(so->so_is_open_owner);
2221                        remove_blocked_locks(lockowner(so));
2222                }
2223        }
2224        nfsd4_return_all_client_layouts(clp);
2225        nfsd4_shutdown_copy(clp);
2226        nfsd4_shutdown_callback(clp);
2227        if (clp->cl_cb_conn.cb_xprt)
2228                svc_xprt_put(clp->cl_cb_conn.cb_xprt);
2229        free_client(clp);
2230        wake_up_all(&expiry_wq);
2231}
2232
2233static void
2234destroy_client(struct nfs4_client *clp)
2235{
2236        unhash_client(clp);
2237        __destroy_client(clp);
2238}
2239
2240static void inc_reclaim_complete(struct nfs4_client *clp)
2241{
2242        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2243
2244        if (!nn->track_reclaim_completes)
2245                return;
2246        if (!nfsd4_find_reclaim_client(clp->cl_name, nn))
2247                return;
2248        if (atomic_inc_return(&nn->nr_reclaim_complete) ==
2249                        nn->reclaim_str_hashtbl_size) {
2250                printk(KERN_INFO "NFSD: all clients done reclaiming, ending NFSv4 grace period (net %x)\n",
2251                                clp->net->ns.inum);
2252                nfsd4_end_grace(nn);
2253        }
2254}
2255
2256static void expire_client(struct nfs4_client *clp)
2257{
2258        unhash_client(clp);
2259        nfsd4_client_record_remove(clp);
2260        __destroy_client(clp);
2261}
2262
2263static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
2264{
2265        memcpy(target->cl_verifier.data, source->data,
2266                        sizeof(target->cl_verifier.data));
2267}
2268
2269static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
2270{
2271        target->cl_clientid.cl_boot = source->cl_clientid.cl_boot; 
2272        target->cl_clientid.cl_id = source->cl_clientid.cl_id; 
2273}
2274
2275static int copy_cred(struct svc_cred *target, struct svc_cred *source)
2276{
2277        target->cr_principal = kstrdup(source->cr_principal, GFP_KERNEL);
2278        target->cr_raw_principal = kstrdup(source->cr_raw_principal,
2279                                                                GFP_KERNEL);
2280        target->cr_targ_princ = kstrdup(source->cr_targ_princ, GFP_KERNEL);
2281        if ((source->cr_principal && !target->cr_principal) ||
2282            (source->cr_raw_principal && !target->cr_raw_principal) ||
2283            (source->cr_targ_princ && !target->cr_targ_princ))
2284                return -ENOMEM;
2285
2286        target->cr_flavor = source->cr_flavor;
2287        target->cr_uid = source->cr_uid;
2288        target->cr_gid = source->cr_gid;
2289        target->cr_group_info = source->cr_group_info;
2290        get_group_info(target->cr_group_info);
2291        target->cr_gss_mech = source->cr_gss_mech;
2292        if (source->cr_gss_mech)
2293                gss_mech_get(source->cr_gss_mech);
2294        return 0;
2295}
2296
2297static int
2298compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
2299{
2300        if (o1->len < o2->len)
2301                return -1;
2302        if (o1->len > o2->len)
2303                return 1;
2304        return memcmp(o1->data, o2->data, o1->len);
2305}
2306
2307static int
2308same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
2309{
2310        return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
2311}
2312
2313static int
2314same_clid(clientid_t *cl1, clientid_t *cl2)
2315{
2316        return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
2317}
2318
2319static bool groups_equal(struct group_info *g1, struct group_info *g2)
2320{
2321        int i;
2322
2323        if (g1->ngroups != g2->ngroups)
2324                return false;
2325        for (i=0; i<g1->ngroups; i++)
2326                if (!gid_eq(g1->gid[i], g2->gid[i]))
2327                        return false;
2328        return true;
2329}
2330
2331/*
2332 * RFC 3530 language requires clid_inuse be returned when the
2333 * "principal" associated with a requests differs from that previously
2334 * used.  We use uid, gid's, and gss principal string as our best
2335 * approximation.  We also don't want to allow non-gss use of a client
2336 * established using gss: in theory cr_principal should catch that
2337 * change, but in practice cr_principal can be null even in the gss case
2338 * since gssd doesn't always pass down a principal string.
2339 */
2340static bool is_gss_cred(struct svc_cred *cr)
2341{
2342        /* Is cr_flavor one of the gss "pseudoflavors"?: */
2343        return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
2344}
2345
2346
2347static bool
2348same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
2349{
2350        if ((is_gss_cred(cr1) != is_gss_cred(cr2))
2351                || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
2352                || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
2353                || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
2354                return false;
2355        /* XXX: check that cr_targ_princ fields match ? */
2356        if (cr1->cr_principal == cr2->cr_principal)
2357                return true;
2358        if (!cr1->cr_principal || !cr2->cr_principal)
2359                return false;
2360        return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
2361}
2362
2363static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
2364{
2365        struct svc_cred *cr = &rqstp->rq_cred;
2366        u32 service;
2367
2368        if (!cr->cr_gss_mech)
2369                return false;
2370        service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
2371        return service == RPC_GSS_SVC_INTEGRITY ||
2372               service == RPC_GSS_SVC_PRIVACY;
2373}
2374
2375bool nfsd4_mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
2376{
2377        struct svc_cred *cr = &rqstp->rq_cred;
2378
2379        if (!cl->cl_mach_cred)
2380                return true;
2381        if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
2382                return false;
2383        if (!svc_rqst_integrity_protected(rqstp))
2384                return false;
2385        if (cl->cl_cred.cr_raw_principal)
2386                return 0 == strcmp(cl->cl_cred.cr_raw_principal,
2387                                                cr->cr_raw_principal);
2388        if (!cr->cr_principal)
2389                return false;
2390        return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
2391}
2392
2393static void gen_confirm(struct nfs4_client *clp, struct nfsd_net *nn)
2394{
2395        __be32 verf[2];
2396
2397        /*
2398         * This is opaque to client, so no need to byte-swap. Use
2399         * __force to keep sparse happy
2400         */
2401        verf[0] = (__force __be32)(u32)ktime_get_real_seconds();
2402        verf[1] = (__force __be32)nn->clverifier_counter++;
2403        memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
2404}
2405
2406static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
2407{
2408        clp->cl_clientid.cl_boot = (u32)nn->boot_time;
2409        clp->cl_clientid.cl_id = nn->clientid_counter++;
2410        gen_confirm(clp, nn);
2411}
2412
2413static struct nfs4_stid *
2414find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
2415{
2416        struct nfs4_stid *ret;
2417
2418        ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
2419        if (!ret || !ret->sc_type)
2420                return NULL;
2421        return ret;
2422}
2423
2424static struct nfs4_stid *
2425find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
2426{
2427        struct nfs4_stid *s;
2428
2429        spin_lock(&cl->cl_lock);
2430        s = find_stateid_locked(cl, t);
2431        if (s != NULL) {
2432                if (typemask & s->sc_type)
2433                        refcount_inc(&s->sc_count);
2434                else
2435                        s = NULL;
2436        }
2437        spin_unlock(&cl->cl_lock);
2438        return s;
2439}
2440
2441static struct nfs4_client *get_nfsdfs_clp(struct inode *inode)
2442{
2443        struct nfsdfs_client *nc;
2444        nc = get_nfsdfs_client(inode);
2445        if (!nc)
2446                return NULL;
2447        return container_of(nc, struct nfs4_client, cl_nfsdfs);
2448}
2449
2450static void seq_quote_mem(struct seq_file *m, char *data, int len)
2451{
2452        seq_printf(m, "\"");
2453        seq_escape_mem(m, data, len, ESCAPE_HEX | ESCAPE_NAP | ESCAPE_APPEND, "\"\\");
2454        seq_printf(m, "\"");
2455}
2456
2457static const char *cb_state2str(int state)
2458{
2459        switch (state) {
2460        case NFSD4_CB_UP:
2461                return "UP";
2462        case NFSD4_CB_UNKNOWN:
2463                return "UNKNOWN";
2464        case NFSD4_CB_DOWN:
2465                return "DOWN";
2466        case NFSD4_CB_FAULT:
2467                return "FAULT";
2468        }
2469        return "UNDEFINED";
2470}
2471
2472static int client_info_show(struct seq_file *m, void *v)
2473{
2474        struct inode *inode = m->private;
2475        struct nfs4_client *clp;
2476        u64 clid;
2477
2478        clp = get_nfsdfs_clp(inode);
2479        if (!clp)
2480                return -ENXIO;
2481        memcpy(&clid, &clp->cl_clientid, sizeof(clid));
2482        seq_printf(m, "clientid: 0x%llx\n", clid);
2483        seq_printf(m, "address: \"%pISpc\"\n", (struct sockaddr *)&clp->cl_addr);
2484
2485        if (clp->cl_state == NFSD4_COURTESY)
2486                seq_puts(m, "status: courtesy\n");
2487        else if (clp->cl_state == NFSD4_EXPIRABLE)
2488                seq_puts(m, "status: expirable\n");
2489        else if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
2490                seq_puts(m, "status: confirmed\n");
2491        else
2492                seq_puts(m, "status: unconfirmed\n");
2493        seq_printf(m, "seconds from last renew: %lld\n",
2494                ktime_get_boottime_seconds() - clp->cl_time);
2495        seq_printf(m, "name: ");
2496        seq_quote_mem(m, clp->cl_name.data, clp->cl_name.len);
2497        seq_printf(m, "\nminor version: %d\n", clp->cl_minorversion);
2498        if (clp->cl_nii_domain.data) {
2499                seq_printf(m, "Implementation domain: ");
2500                seq_quote_mem(m, clp->cl_nii_domain.data,
2501                                        clp->cl_nii_domain.len);
2502                seq_printf(m, "\nImplementation name: ");
2503                seq_quote_mem(m, clp->cl_nii_name.data, clp->cl_nii_name.len);
2504                seq_printf(m, "\nImplementation time: [%lld, %ld]\n",
2505                        clp->cl_nii_time.tv_sec, clp->cl_nii_time.tv_nsec);
2506        }
2507        seq_printf(m, "callback state: %s\n", cb_state2str(clp->cl_cb_state));
2508        seq_printf(m, "callback address: %pISpc\n", &clp->cl_cb_conn.cb_addr);
2509        drop_client(clp);
2510
2511        return 0;
2512}
2513
2514static int client_info_open(struct inode *inode, struct file *file)
2515{
2516        return single_open(file, client_info_show, inode);
2517}
2518
2519static const struct file_operations client_info_fops = {
2520        .open           = client_info_open,
2521        .read           = seq_read,
2522        .llseek         = seq_lseek,
2523        .release        = single_release,
2524};
2525
2526static void *states_start(struct seq_file *s, loff_t *pos)
2527        __acquires(&clp->cl_lock)
2528{
2529        struct nfs4_client *clp = s->private;
2530        unsigned long id = *pos;
2531        void *ret;
2532
2533        spin_lock(&clp->cl_lock);
2534        ret = idr_get_next_ul(&clp->cl_stateids, &id);
2535        *pos = id;
2536        return ret;
2537}
2538
2539static void *states_next(struct seq_file *s, void *v, loff_t *pos)
2540{
2541        struct nfs4_client *clp = s->private;
2542        unsigned long id = *pos;
2543        void *ret;
2544
2545        id = *pos;
2546        id++;
2547        ret = idr_get_next_ul(&clp->cl_stateids, &id);
2548        *pos = id;
2549        return ret;
2550}
2551
2552static void states_stop(struct seq_file *s, void *v)
2553        __releases(&clp->cl_lock)
2554{
2555        struct nfs4_client *clp = s->private;
2556
2557        spin_unlock(&clp->cl_lock);
2558}
2559
2560static void nfs4_show_fname(struct seq_file *s, struct nfsd_file *f)
2561{
2562         seq_printf(s, "filename: \"%pD2\"", f->nf_file);
2563}
2564
2565static void nfs4_show_superblock(struct seq_file *s, struct nfsd_file *f)
2566{
2567        struct inode *inode = f->nf_inode;
2568
2569        seq_printf(s, "superblock: \"%02x:%02x:%ld\"",
2570                                        MAJOR(inode->i_sb->s_dev),
2571                                         MINOR(inode->i_sb->s_dev),
2572                                         inode->i_ino);
2573}
2574
2575static void nfs4_show_owner(struct seq_file *s, struct nfs4_stateowner *oo)
2576{
2577        seq_printf(s, "owner: ");
2578        seq_quote_mem(s, oo->so_owner.data, oo->so_owner.len);
2579}
2580
2581static void nfs4_show_stateid(struct seq_file *s, stateid_t *stid)
2582{
2583        seq_printf(s, "0x%.8x", stid->si_generation);
2584        seq_printf(s, "%12phN", &stid->si_opaque);
2585}
2586
2587static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
2588{
2589        struct nfs4_ol_stateid *ols;
2590        struct nfs4_file *nf;
2591        struct nfsd_file *file;
2592        struct nfs4_stateowner *oo;
2593        unsigned int access, deny;
2594
2595        if (st->sc_type != NFS4_OPEN_STID && st->sc_type != NFS4_LOCK_STID)
2596                return 0; /* XXX: or SEQ_SKIP? */
2597        ols = openlockstateid(st);
2598        oo = ols->st_stateowner;
2599        nf = st->sc_file;
2600        file = find_any_file(nf);
2601        if (!file)
2602                return 0;
2603
2604        seq_printf(s, "- ");
2605        nfs4_show_stateid(s, &st->sc_stateid);
2606        seq_printf(s, ": { type: open, ");
2607
2608        access = bmap_to_share_mode(ols->st_access_bmap);
2609        deny   = bmap_to_share_mode(ols->st_deny_bmap);
2610
2611        seq_printf(s, "access: %s%s, ",
2612                access & NFS4_SHARE_ACCESS_READ ? "r" : "-",
2613                access & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
2614        seq_printf(s, "deny: %s%s, ",
2615                deny & NFS4_SHARE_ACCESS_READ ? "r" : "-",
2616                deny & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
2617
2618        nfs4_show_superblock(s, file);
2619        seq_printf(s, ", ");
2620        nfs4_show_fname(s, file);
2621        seq_printf(s, ", ");
2622        nfs4_show_owner(s, oo);
2623        seq_printf(s, " }\n");
2624        nfsd_file_put(file);
2625
2626        return 0;
2627}
2628
2629static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
2630{
2631        struct nfs4_ol_stateid *ols;
2632        struct nfs4_file *nf;
2633        struct nfsd_file *file;
2634        struct nfs4_stateowner *oo;
2635
2636        ols = openlockstateid(st);
2637        oo = ols->st_stateowner;
2638        nf = st->sc_file;
2639        file = find_any_file(nf);
2640        if (!file)
2641                return 0;
2642
2643        seq_printf(s, "- ");
2644        nfs4_show_stateid(s, &st->sc_stateid);
2645        seq_printf(s, ": { type: lock, ");
2646
2647        /*
2648         * Note: a lock stateid isn't really the same thing as a lock,
2649         * it's the locking state held by one owner on a file, and there
2650         * may be multiple (or no) lock ranges associated with it.
2651         * (Same for the matter is true of open stateids.)
2652         */
2653
2654        nfs4_show_superblock(s, file);
2655        /* XXX: open stateid? */
2656        seq_printf(s, ", ");
2657        nfs4_show_fname(s, file);
2658        seq_printf(s, ", ");
2659        nfs4_show_owner(s, oo);
2660        seq_printf(s, " }\n");
2661        nfsd_file_put(file);
2662
2663        return 0;
2664}
2665
2666static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
2667{
2668        struct nfs4_delegation *ds;
2669        struct nfs4_file *nf;
2670        struct nfsd_file *file;
2671
2672        ds = delegstateid(st);
2673        nf = st->sc_file;
2674        file = find_deleg_file(nf);
2675        if (!file)
2676                return 0;
2677
2678        seq_printf(s, "- ");
2679        nfs4_show_stateid(s, &st->sc_stateid);
2680        seq_printf(s, ": { type: deleg, ");
2681
2682        /* Kinda dead code as long as we only support read delegs: */
2683        seq_printf(s, "access: %s, ",
2684                ds->dl_type == NFS4_OPEN_DELEGATE_READ ? "r" : "w");
2685
2686        /* XXX: lease time, whether it's being recalled. */
2687
2688        nfs4_show_superblock(s, file);
2689        seq_printf(s, ", ");
2690        nfs4_show_fname(s, file);
2691        seq_printf(s, " }\n");
2692        nfsd_file_put(file);
2693
2694        return 0;
2695}
2696
2697static int nfs4_show_layout(struct seq_file *s, struct nfs4_stid *st)
2698{
2699        struct nfs4_layout_stateid *ls;
2700        struct nfsd_file *file;
2701
2702        ls = container_of(st, struct nfs4_layout_stateid, ls_stid);
2703        file = ls->ls_file;
2704
2705        seq_printf(s, "- ");
2706        nfs4_show_stateid(s, &st->sc_stateid);
2707        seq_printf(s, ": { type: layout, ");
2708
2709        /* XXX: What else would be useful? */
2710
2711        nfs4_show_superblock(s, file);
2712        seq_printf(s, ", ");
2713        nfs4_show_fname(s, file);
2714        seq_printf(s, " }\n");
2715
2716        return 0;
2717}
2718
2719static int states_show(struct seq_file *s, void *v)
2720{
2721        struct nfs4_stid *st = v;
2722
2723        switch (st->sc_type) {
2724        case NFS4_OPEN_STID:
2725                return nfs4_show_open(s, st);
2726        case NFS4_LOCK_STID:
2727                return nfs4_show_lock(s, st);
2728        case NFS4_DELEG_STID:
2729                return nfs4_show_deleg(s, st);
2730        case NFS4_LAYOUT_STID:
2731                return nfs4_show_layout(s, st);
2732        default:
2733                return 0; /* XXX: or SEQ_SKIP? */
2734        }
2735        /* XXX: copy stateids? */
2736}
2737
2738static struct seq_operations states_seq_ops = {
2739        .start = states_start,
2740        .next = states_next,
2741        .stop = states_stop,
2742        .show = states_show
2743};
2744
2745static int client_states_open(struct inode *inode, struct file *file)
2746{
2747        struct seq_file *s;
2748        struct nfs4_client *clp;
2749        int ret;
2750
2751        clp = get_nfsdfs_clp(inode);
2752        if (!clp)
2753                return -ENXIO;
2754
2755        ret = seq_open(file, &states_seq_ops);
2756        if (ret)
2757                return ret;
2758        s = file->private_data;
2759        s->private = clp;
2760        return 0;
2761}
2762
2763static int client_opens_release(struct inode *inode, struct file *file)
2764{
2765        struct seq_file *m = file->private_data;
2766        struct nfs4_client *clp = m->private;
2767
2768        /* XXX: alternatively, we could get/drop in seq start/stop */
2769        drop_client(clp);
2770        return 0;
2771}
2772
2773static const struct file_operations client_states_fops = {
2774        .open           = client_states_open,
2775        .read           = seq_read,
2776        .llseek         = seq_lseek,
2777        .release        = client_opens_release,
2778};
2779
2780/*
2781 * Normally we refuse to destroy clients that are in use, but here the
2782 * administrator is telling us to just do it.  We also want to wait
2783 * so the caller has a guarantee that the client's locks are gone by
2784 * the time the write returns:
2785 */
2786static void force_expire_client(struct nfs4_client *clp)
2787{
2788        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2789        bool already_expired;
2790
2791        trace_nfsd_clid_admin_expired(&clp->cl_clientid);
2792
2793        spin_lock(&nn->client_lock);
2794        clp->cl_time = 0;
2795        spin_unlock(&nn->client_lock);
2796
2797        wait_event(expiry_wq, atomic_read(&clp->cl_rpc_users) == 0);
2798        spin_lock(&nn->client_lock);
2799        already_expired = list_empty(&clp->cl_lru);
2800        if (!already_expired)
2801                unhash_client_locked(clp);
2802        spin_unlock(&nn->client_lock);
2803
2804        if (!already_expired)
2805                expire_client(clp);
2806        else
2807                wait_event(expiry_wq, clp->cl_nfsd_dentry == NULL);
2808}
2809
2810static ssize_t client_ctl_write(struct file *file, const char __user *buf,
2811                                   size_t size, loff_t *pos)
2812{
2813        char *data;
2814        struct nfs4_client *clp;
2815
2816        data = simple_transaction_get(file, buf, size);
2817        if (IS_ERR(data))
2818                return PTR_ERR(data);
2819        if (size != 7 || 0 != memcmp(data, "expire\n", 7))
2820                return -EINVAL;
2821        clp = get_nfsdfs_clp(file_inode(file));
2822        if (!clp)
2823                return -ENXIO;
2824        force_expire_client(clp);
2825        drop_client(clp);
2826        return 7;
2827}
2828
2829static const struct file_operations client_ctl_fops = {
2830        .write          = client_ctl_write,
2831        .release        = simple_transaction_release,
2832};
2833
2834static const struct tree_descr client_files[] = {
2835        [0] = {"info", &client_info_fops, S_IRUSR},
2836        [1] = {"states", &client_states_fops, S_IRUSR},
2837        [2] = {"ctl", &client_ctl_fops, S_IWUSR},
2838        [3] = {""},
2839};
2840
2841static struct nfs4_client *create_client(struct xdr_netobj name,
2842                struct svc_rqst *rqstp, nfs4_verifier *verf)
2843{
2844        struct nfs4_client *clp;
2845        struct sockaddr *sa = svc_addr(rqstp);
2846        int ret;
2847        struct net *net = SVC_NET(rqstp);
2848        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2849        struct dentry *dentries[ARRAY_SIZE(client_files)];
2850
2851        clp = alloc_client(name);
2852        if (clp == NULL)
2853                return NULL;
2854
2855        ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
2856        if (ret) {
2857                free_client(clp);
2858                return NULL;
2859        }
2860        gen_clid(clp, nn);
2861        kref_init(&clp->cl_nfsdfs.cl_ref);
2862        nfsd4_init_cb(&clp->cl_cb_null, clp, NULL, NFSPROC4_CLNT_CB_NULL);
2863        clp->cl_time = ktime_get_boottime_seconds();
2864        clear_bit(0, &clp->cl_cb_slot_busy);
2865        copy_verf(clp, verf);
2866        memcpy(&clp->cl_addr, sa, sizeof(struct sockaddr_storage));
2867        clp->cl_cb_session = NULL;
2868        clp->net = net;
2869        clp->cl_nfsd_dentry = nfsd_client_mkdir(
2870                nn, &clp->cl_nfsdfs,
2871                clp->cl_clientid.cl_id - nn->clientid_base,
2872                client_files, dentries);
2873        clp->cl_nfsd_info_dentry = dentries[0];
2874        if (!clp->cl_nfsd_dentry) {
2875                free_client(clp);
2876                return NULL;
2877        }
2878        return clp;
2879}
2880
2881static void
2882add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
2883{
2884        struct rb_node **new = &(root->rb_node), *parent = NULL;
2885        struct nfs4_client *clp;
2886
2887        while (*new) {
2888                clp = rb_entry(*new, struct nfs4_client, cl_namenode);
2889                parent = *new;
2890
2891                if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
2892                        new = &((*new)->rb_left);
2893                else
2894                        new = &((*new)->rb_right);
2895        }
2896
2897        rb_link_node(&new_clp->cl_namenode, parent, new);
2898        rb_insert_color(&new_clp->cl_namenode, root);
2899}
2900
2901static struct nfs4_client *
2902find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
2903{
2904        int cmp;
2905        struct rb_node *node = root->rb_node;
2906        struct nfs4_client *clp;
2907
2908        while (node) {
2909                clp = rb_entry(node, struct nfs4_client, cl_namenode);
2910                cmp = compare_blob(&clp->cl_name, name);
2911                if (cmp > 0)
2912                        node = node->rb_left;
2913                else if (cmp < 0)
2914                        node = node->rb_right;
2915                else
2916                        return clp;
2917        }
2918        return NULL;
2919}
2920
2921static void
2922add_to_unconfirmed(struct nfs4_client *clp)
2923{
2924        unsigned int idhashval;
2925        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2926
2927        lockdep_assert_held(&nn->client_lock);
2928
2929        clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2930        add_clp_to_name_tree(clp, &nn->unconf_name_tree);
2931        idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2932        list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
2933        renew_client_locked(clp);
2934}
2935
2936static void
2937move_to_confirmed(struct nfs4_client *clp)
2938{
2939        unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2940        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2941
2942        lockdep_assert_held(&nn->client_lock);
2943
2944        list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
2945        rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2946        add_clp_to_name_tree(clp, &nn->conf_name_tree);
2947        set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2948        trace_nfsd_clid_confirmed(&clp->cl_clientid);
2949        renew_client_locked(clp);
2950}
2951
2952static struct nfs4_client *
2953find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
2954{
2955        struct nfs4_client *clp;
2956        unsigned int idhashval = clientid_hashval(clid->cl_id);
2957
2958        list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
2959                if (same_clid(&clp->cl_clientid, clid)) {
2960                        if ((bool)clp->cl_minorversion != sessions)
2961                                return NULL;
2962                        renew_client_locked(clp);
2963                        return clp;
2964                }
2965        }
2966        return NULL;
2967}
2968
2969static struct nfs4_client *
2970find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2971{
2972        struct list_head *tbl = nn->conf_id_hashtbl;
2973
2974        lockdep_assert_held(&nn->client_lock);
2975        return find_client_in_id_table(tbl, clid, sessions);
2976}
2977
2978static struct nfs4_client *
2979find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2980{
2981        struct list_head *tbl = nn->unconf_id_hashtbl;
2982
2983        lockdep_assert_held(&nn->client_lock);
2984        return find_client_in_id_table(tbl, clid, sessions);
2985}
2986
2987static bool clp_used_exchangeid(struct nfs4_client *clp)
2988{
2989        return clp->cl_exchange_flags != 0;
2990} 
2991
2992static struct nfs4_client *
2993find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2994{
2995        lockdep_assert_held(&nn->client_lock);
2996        return find_clp_in_name_tree(name, &nn->conf_name_tree);
2997}
2998
2999static struct nfs4_client *
3000find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
3001{
3002        lockdep_assert_held(&nn->client_lock);
3003        return find_clp_in_name_tree(name, &nn->unconf_name_tree);
3004}
3005
3006static void
3007gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
3008{
3009        struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
3010        struct sockaddr *sa = svc_addr(rqstp);
3011        u32 scopeid = rpc_get_scope_id(sa);
3012        unsigned short expected_family;
3013
3014        /* Currently, we only support tcp and tcp6 for the callback channel */
3015        if (se->se_callback_netid_len == 3 &&
3016            !memcmp(se->se_callback_netid_val, "tcp", 3))
3017                expected_family = AF_INET;
3018        else if (se->se_callback_netid_len == 4 &&
3019                 !memcmp(se->se_callback_netid_val, "tcp6", 4))
3020                expected_family = AF_INET6;
3021        else
3022                goto out_err;
3023
3024        conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
3025                                            se->se_callback_addr_len,
3026                                            (struct sockaddr *)&conn->cb_addr,
3027                                            sizeof(conn->cb_addr));
3028
3029        if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
3030                goto out_err;
3031
3032        if (conn->cb_addr.ss_family == AF_INET6)
3033                ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
3034
3035        conn->cb_prog = se->se_callback_prog;
3036        conn->cb_ident = se->se_callback_ident;
3037        memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
3038        trace_nfsd_cb_args(clp, conn);
3039        return;
3040out_err:
3041        conn->cb_addr.ss_family = AF_UNSPEC;
3042        conn->cb_addrlen = 0;
3043        trace_nfsd_cb_nodelegs(clp);
3044        return;
3045}
3046
3047/*
3048 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
3049 */
3050static void
3051nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
3052{
3053        struct xdr_buf *buf = resp->xdr->buf;
3054        struct nfsd4_slot *slot = resp->cstate.slot;
3055        unsigned int base;
3056
3057        dprintk("--> %s slot %p\n", __func__, slot);
3058
3059        slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
3060        slot->sl_opcnt = resp->opcnt;
3061        slot->sl_status = resp->cstate.status;
3062        free_svc_cred(&slot->sl_cred);
3063        copy_cred(&slot->sl_cred, &resp->rqstp->rq_cred);
3064
3065        if (!nfsd4_cache_this(resp)) {
3066                slot->sl_flags &= ~NFSD4_SLOT_CACHED;
3067                return;
3068        }
3069        slot->sl_flags |= NFSD4_SLOT_CACHED;
3070
3071        base = resp->cstate.data_offset;
3072        slot->sl_datalen = buf->len - base;
3073        if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
3074                WARN(1, "%s: sessions DRC could not cache compound\n",
3075                     __func__);
3076        return;
3077}
3078
3079/*
3080 * Encode the replay sequence operation from the slot values.
3081 * If cachethis is FALSE encode the uncached rep error on the next
3082 * operation which sets resp->p and increments resp->opcnt for
3083 * nfs4svc_encode_compoundres.
3084 *
3085 */
3086static __be32
3087nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
3088                          struct nfsd4_compoundres *resp)
3089{
3090        struct nfsd4_op *op;
3091        struct nfsd4_slot *slot = resp->cstate.slot;
3092
3093        /* Encode the replayed sequence operation */
3094        op = &args->ops[resp->opcnt - 1];
3095        nfsd4_encode_operation(resp, op);
3096
3097        if (slot->sl_flags & NFSD4_SLOT_CACHED)
3098                return op->status;
3099        if (args->opcnt == 1) {
3100                /*
3101                 * The original operation wasn't a solo sequence--we
3102                 * always cache those--so this retry must not match the
3103                 * original:
3104                 */
3105                op->status = nfserr_seq_false_retry;
3106        } else {
3107                op = &args->ops[resp->opcnt++];
3108                op->status = nfserr_retry_uncached_rep;
3109                nfsd4_encode_operation(resp, op);
3110        }
3111        return op->status;
3112}
3113
3114/*
3115 * The sequence operation is not cached because we can use the slot and
3116 * session values.
3117 */
3118static __be32
3119nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
3120                         struct nfsd4_sequence *seq)
3121{
3122        struct nfsd4_slot *slot = resp->cstate.slot;
3123        struct xdr_stream *xdr = resp->xdr;
3124        __be32 *p;
3125        __be32 status;
3126
3127        dprintk("--> %s slot %p\n", __func__, slot);
3128
3129        status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
3130        if (status)
3131                return status;
3132
3133        p = xdr_reserve_space(xdr, slot->sl_datalen);
3134        if (!p) {
3135                WARN_ON_ONCE(1);
3136                return nfserr_serverfault;
3137        }
3138        xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
3139        xdr_commit_encode(xdr);
3140
3141        resp->opcnt = slot->sl_opcnt;
3142        return slot->sl_status;
3143}
3144
3145/*
3146 * Set the exchange_id flags returned by the server.
3147 */
3148static void
3149nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
3150{
3151#ifdef CONFIG_NFSD_PNFS
3152        new->cl_exchange_flags |= EXCHGID4_FLAG_USE_PNFS_MDS;
3153#else
3154        new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
3155#endif
3156
3157        /* Referrals are supported, Migration is not. */
3158        new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
3159
3160        /* set the wire flags to return to client. */
3161        clid->flags = new->cl_exchange_flags;
3162}
3163
3164static bool client_has_openowners(struct nfs4_client *clp)
3165{
3166        struct nfs4_openowner *oo;
3167
3168        list_for_each_entry(oo, &clp->cl_openowners, oo_perclient) {
3169                if (!list_empty(&oo->oo_owner.so_stateids))
3170                        return true;
3171        }
3172        return false;
3173}
3174
3175static bool client_has_state(struct nfs4_client *clp)
3176{
3177        return client_has_openowners(clp)
3178#ifdef CONFIG_NFSD_PNFS
3179                || !list_empty(&clp->cl_lo_states)
3180#endif
3181                || !list_empty(&clp->cl_delegations)
3182                || !list_empty(&clp->cl_sessions)
3183                || !list_empty(&clp->async_copies);
3184}
3185
3186static __be32 copy_impl_id(struct nfs4_client *clp,
3187                                struct nfsd4_exchange_id *exid)
3188{
3189        if (!exid->nii_domain.data)
3190                return 0;
3191        xdr_netobj_dup(&clp->cl_nii_domain, &exid->nii_domain, GFP_KERNEL);
3192        if (!clp->cl_nii_domain.data)
3193                return nfserr_jukebox;
3194        xdr_netobj_dup(&clp->cl_nii_name, &exid->nii_name, GFP_KERNEL);
3195        if (!clp->cl_nii_name.data)
3196                return nfserr_jukebox;
3197        clp->cl_nii_time = exid->nii_time;
3198        return 0;
3199}
3200
3201__be32
3202nfsd4_exchange_id(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3203                union nfsd4_op_u *u)
3204{
3205        struct nfsd4_exchange_id *exid = &u->exchange_id;
3206        struct nfs4_client *conf, *new;
3207        struct nfs4_client *unconf = NULL;
3208        __be32 status;
3209        char                    addr_str[INET6_ADDRSTRLEN];
3210        nfs4_verifier           verf = exid->verifier;
3211        struct sockaddr         *sa = svc_addr(rqstp);
3212        bool    update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
3213        struct nfsd_net         *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3214
3215        rpc_ntop(sa, addr_str, sizeof(addr_str));
3216        dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
3217                "ip_addr=%s flags %x, spa_how %u\n",
3218                __func__, rqstp, exid, exid->clname.len, exid->clname.data,
3219                addr_str, exid->flags, exid->spa_how);
3220
3221        if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
3222                return nfserr_inval;
3223
3224        new = create_client(exid->clname, rqstp, &verf);
3225        if (new == NULL)
3226                return nfserr_jukebox;
3227        status = copy_impl_id(new, exid);
3228        if (status)
3229                goto out_nolock;
3230
3231        switch (exid->spa_how) {
3232        case SP4_MACH_CRED:
3233                exid->spo_must_enforce[0] = 0;
3234                exid->spo_must_enforce[1] = (
3235                        1 << (OP_BIND_CONN_TO_SESSION - 32) |
3236                        1 << (OP_EXCHANGE_ID - 32) |
3237                        1 << (OP_CREATE_SESSION - 32) |
3238                        1 << (OP_DESTROY_SESSION - 32) |
3239                        1 << (OP_DESTROY_CLIENTID - 32));
3240
3241                exid->spo_must_allow[0] &= (1 << (OP_CLOSE) |
3242                                        1 << (OP_OPEN_DOWNGRADE) |
3243                                        1 << (OP_LOCKU) |
3244                                        1 << (OP_DELEGRETURN));
3245
3246                exid->spo_must_allow[1] &= (
3247                                        1 << (OP_TEST_STATEID - 32) |
3248                                        1 << (OP_FREE_STATEID - 32));
3249                if (!svc_rqst_integrity_protected(rqstp)) {
3250                        status = nfserr_inval;
3251                        goto out_nolock;
3252                }
3253                /*
3254                 * Sometimes userspace doesn't give us a principal.
3255                 * Which is a bug, really.  Anyway, we can't enforce
3256                 * MACH_CRED in that case, better to give up now:
3257                 */
3258                if (!new->cl_cred.cr_principal &&
3259                                        !new->cl_cred.cr_raw_principal) {
3260                        status = nfserr_serverfault;
3261                        goto out_nolock;
3262                }
3263                new->cl_mach_cred = true;
3264                break;
3265        case SP4_NONE:
3266                break;
3267        default:                                /* checked by xdr code */
3268                WARN_ON_ONCE(1);
3269                fallthrough;
3270        case SP4_SSV:
3271                status = nfserr_encr_alg_unsupp;
3272                goto out_nolock;
3273        }
3274
3275        /* Cases below refer to rfc 5661 section 18.35.4: */
3276        spin_lock(&nn->client_lock);
3277        conf = find_confirmed_client_by_name(&exid->clname, nn);
3278        if (conf) {
3279                bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
3280                bool verfs_match = same_verf(&verf, &conf->cl_verifier);
3281
3282                if (update) {
3283                        if (!clp_used_exchangeid(conf)) { /* buggy client */
3284                                status = nfserr_inval;
3285                                goto out;
3286                        }
3287                        if (!nfsd4_mach_creds_match(conf, rqstp)) {
3288                                status = nfserr_wrong_cred;
3289                                goto out;
3290                        }
3291                        if (!creds_match) { /* case 9 */
3292                                status = nfserr_perm;
3293                                goto out;
3294                        }
3295                        if (!verfs_match) { /* case 8 */
3296                                status = nfserr_not_same;
3297                                goto out;
3298                        }
3299                        /* case 6 */
3300                        exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
3301                        trace_nfsd_clid_confirmed_r(conf);
3302                        goto out_copy;
3303                }
3304                if (!creds_match) { /* case 3 */
3305                        if (client_has_state(conf)) {
3306                                status = nfserr_clid_inuse;
3307                                trace_nfsd_clid_cred_mismatch(conf, rqstp);
3308                                goto out;
3309                        }
3310                        goto out_new;
3311                }
3312                if (verfs_match) { /* case 2 */
3313                        conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
3314                        trace_nfsd_clid_confirmed_r(conf);
3315                        goto out_copy;
3316                }
3317                /* case 5, client reboot */
3318                trace_nfsd_clid_verf_mismatch(conf, rqstp, &verf);
3319                conf = NULL;
3320                goto out_new;
3321        }
3322
3323        if (update) { /* case 7 */
3324                status = nfserr_noent;
3325                goto out;
3326        }
3327
3328        unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
3329        if (unconf) /* case 4, possible retry or client restart */
3330                unhash_client_locked(unconf);
3331
3332        /* case 1, new owner ID */
3333        trace_nfsd_clid_fresh(new);
3334
3335out_new:
3336        if (conf) {
3337                status = mark_client_expired_locked(conf);
3338                if (status)
3339                        goto out;
3340                trace_nfsd_clid_replaced(&conf->cl_clientid);
3341        }
3342        new->cl_minorversion = cstate->minorversion;
3343        new->cl_spo_must_allow.u.words[0] = exid->spo_must_allow[0];
3344        new->cl_spo_must_allow.u.words[1] = exid->spo_must_allow[1];
3345
3346        add_to_unconfirmed(new);
3347        swap(new, conf);
3348out_copy:
3349        exid->clientid.cl_boot = conf->cl_clientid.cl_boot;
3350        exid->clientid.cl_id = conf->cl_clientid.cl_id;
3351
3352        exid->seqid = conf->cl_cs_slot.sl_seqid + 1;
3353        nfsd4_set_ex_flags(conf, exid);
3354
3355        dprintk("nfsd4_exchange_id seqid %d flags %x\n",
3356                conf->cl_cs_slot.sl_seqid, conf->cl_exchange_flags);
3357        status = nfs_ok;
3358
3359out:
3360        spin_unlock(&nn->client_lock);
3361out_nolock:
3362        if (new)
3363                expire_client(new);
3364        if (unconf) {
3365                trace_nfsd_clid_expire_unconf(&unconf->cl_clientid);
3366                expire_client(unconf);
3367        }
3368        return status;
3369}
3370
3371static __be32
3372check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
3373{
3374        dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
3375                slot_seqid);
3376
3377        /* The slot is in use, and no response has been sent. */
3378        if (slot_inuse) {
3379                if (seqid == slot_seqid)
3380                        return nfserr_jukebox;
3381                else
3382                        return nfserr_seq_misordered;
3383        }
3384        /* Note unsigned 32-bit arithmetic handles wraparound: */
3385        if (likely(seqid == slot_seqid + 1))
3386                return nfs_ok;
3387        if (seqid == slot_seqid)
3388                return nfserr_replay_cache;
3389        return nfserr_seq_misordered;
3390}
3391
3392/*
3393 * Cache the create session result into the create session single DRC
3394 * slot cache by saving the xdr structure. sl_seqid has been set.
3395 * Do this for solo or embedded create session operations.
3396 */
3397static void
3398nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
3399                           struct nfsd4_clid_slot *slot, __be32 nfserr)
3400{
3401        slot->sl_status = nfserr;
3402        memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
3403}
3404
3405static __be32
3406nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
3407                            struct nfsd4_clid_slot *slot)
3408{
3409        memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
3410        return slot->sl_status;
3411}
3412
3413#define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
3414                        2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
3415                        1 +     /* MIN tag is length with zero, only length */ \
3416                        3 +     /* version, opcount, opcode */ \
3417                        XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
3418                                /* seqid, slotID, slotID, cache */ \
3419                        4 ) * sizeof(__be32))
3420
3421#define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
3422                        2 +     /* verifier: AUTH_NULL, length 0 */\
3423                        1 +     /* status */ \
3424                        1 +     /* MIN tag is length with zero, only length */ \
3425                        3 +     /* opcount, opcode, opstatus*/ \
3426                        XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
3427                                /* seqid, slotID, slotID, slotID, status */ \
3428                        5 ) * sizeof(__be32))
3429
3430static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
3431{
3432        u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
3433
3434        if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
3435                return nfserr_toosmall;
3436        if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
3437                return nfserr_toosmall;
3438        ca->headerpadsz = 0;
3439        ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
3440        ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
3441        ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
3442        ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
3443                        NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
3444        ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
3445        /*
3446         * Note decreasing slot size below client's request may make it
3447         * difficult for client to function correctly, whereas
3448         * decreasing the number of slots will (just?) affect
3449         * performance.  When short on memory we therefore prefer to
3450         * decrease number of slots instead of their size.  Clients that
3451         * request larger slots than they need will get poor results:
3452         * Note that we always allow at least one slot, because our
3453         * accounting is soft and provides no guarantees either way.
3454         */
3455        ca->maxreqs = nfsd4_get_drc_mem(ca, nn);
3456
3457        return nfs_ok;
3458}
3459
3460/*
3461 * Server's NFSv4.1 backchannel support is AUTH_SYS-only for now.
3462 * These are based on similar macros in linux/sunrpc/msg_prot.h .
3463 */
3464#define RPC_MAX_HEADER_WITH_AUTH_SYS \
3465        (RPC_CALLHDRSIZE + 2 * (2 + UNX_CALLSLACK))
3466
3467#define RPC_MAX_REPHEADER_WITH_AUTH_SYS \
3468        (RPC_REPHDRSIZE + (2 + NUL_REPLYSLACK))
3469
3470#define NFSD_CB_MAX_REQ_SZ      ((NFS4_enc_cb_recall_sz + \
3471                                 RPC_MAX_HEADER_WITH_AUTH_SYS) * sizeof(__be32))
3472#define NFSD_CB_MAX_RESP_SZ     ((NFS4_dec_cb_recall_sz + \
3473                                 RPC_MAX_REPHEADER_WITH_AUTH_SYS) * \
3474                                 sizeof(__be32))
3475
3476static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
3477{
3478        ca->headerpadsz = 0;
3479
3480        if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
3481                return nfserr_toosmall;
3482        if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
3483                return nfserr_toosmall;
3484        ca->maxresp_cached = 0;
3485        if (ca->maxops < 2)
3486                return nfserr_toosmall;
3487
3488        return nfs_ok;
3489}
3490
3491static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
3492{
3493        switch (cbs->flavor) {
3494        case RPC_AUTH_NULL:
3495        case RPC_AUTH_UNIX:
3496                return nfs_ok;
3497        default:
3498                /*
3499                 * GSS case: the spec doesn't allow us to return this
3500                 * error.  But it also doesn't allow us not to support
3501                 * GSS.
3502                 * I'd rather this fail hard than return some error the
3503                 * client might think it can already handle:
3504                 */
3505                return nfserr_encr_alg_unsupp;
3506        }
3507}
3508
3509__be32
3510nfsd4_create_session(struct svc_rqst *rqstp,
3511                struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
3512{
3513        struct nfsd4_create_session *cr_ses = &u->create_session;
3514        struct sockaddr *sa = svc_addr(rqstp);
3515        struct nfs4_client *conf, *unconf;
3516        struct nfs4_client *old = NULL;
3517        struct nfsd4_session *new;
3518        struct nfsd4_conn *conn;
3519        struct nfsd4_clid_slot *cs_slot = NULL;
3520        __be32 status = 0;
3521        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3522
3523        if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
3524                return nfserr_inval;
3525        status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
3526        if (status)
3527                return status;
3528        status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
3529        if (status)
3530                return status;
3531        status = check_backchannel_attrs(&cr_ses->back_channel);
3532        if (status)
3533                goto out_release_drc_mem;
3534        status = nfserr_jukebox;
3535        new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
3536        if (!new)
3537                goto out_release_drc_mem;
3538        conn = alloc_conn_from_crses(rqstp, cr_ses);
3539        if (!conn)
3540                goto out_free_session;
3541
3542        spin_lock(&nn->client_lock);
3543        unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
3544        conf = find_confirmed_client(&cr_ses->clientid, true, nn);
3545        WARN_ON_ONCE(conf && unconf);
3546
3547        if (conf) {
3548                status = nfserr_wrong_cred;
3549                if (!nfsd4_mach_creds_match(conf, rqstp))
3550                        goto out_free_conn;
3551                cs_slot = &conf->cl_cs_slot;
3552                status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
3553                if (status) {
3554                        if (status == nfserr_replay_cache)
3555                                status = nfsd4_replay_create_session(cr_ses, cs_slot);
3556                        goto out_free_conn;
3557                }
3558        } else if (unconf) {
3559                status = nfserr_clid_inuse;
3560                if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
3561                    !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
3562                        trace_nfsd_clid_cred_mismatch(unconf, rqstp);
3563                        goto out_free_conn;
3564                }
3565                status = nfserr_wrong_cred;
3566                if (!nfsd4_mach_creds_match(unconf, rqstp))
3567                        goto out_free_conn;
3568                cs_slot = &unconf->cl_cs_slot;
3569                status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
3570                if (status) {
3571                        /* an unconfirmed replay returns misordered */
3572                        status = nfserr_seq_misordered;
3573                        goto out_free_conn;
3574                }
3575                old = find_confirmed_client_by_name(&unconf->cl_name, nn);
3576                if (old) {
3577                        status = mark_client_expired_locked(old);
3578                        if (status) {
3579                                old = NULL;
3580                                goto out_free_conn;
3581                        }
3582                        trace_nfsd_clid_replaced(&old->cl_clientid);
3583                }
3584                move_to_confirmed(unconf);
3585                conf = unconf;
3586        } else {
3587                status = nfserr_stale_clientid;
3588                goto out_free_conn;
3589        }
3590        status = nfs_ok;
3591        /* Persistent sessions are not supported */
3592        cr_ses->flags &= ~SESSION4_PERSIST;
3593        /* Upshifting from TCP to RDMA is not supported */
3594        cr_ses->flags &= ~SESSION4_RDMA;
3595
3596        init_session(rqstp, new, conf, cr_ses);
3597        nfsd4_get_session_locked(new);
3598
3599        memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
3600               NFS4_MAX_SESSIONID_LEN);
3601        cs_slot->sl_seqid++;
3602        cr_ses->seqid = cs_slot->sl_seqid;
3603
3604        /* cache solo and embedded create sessions under the client_lock */
3605        nfsd4_cache_create_session(cr_ses, cs_slot, status);
3606        spin_unlock(&nn->client_lock);
3607        if (conf == unconf)
3608                fsnotify_dentry(conf->cl_nfsd_info_dentry, FS_MODIFY);
3609        /* init connection and backchannel */
3610        nfsd4_init_conn(rqstp, conn, new);
3611        nfsd4_put_session(new);
3612        if (old)
3613                expire_client(old);
3614        return status;
3615out_free_conn:
3616        spin_unlock(&nn->client_lock);
3617        free_conn(conn);
3618        if (old)
3619                expire_client(old);
3620out_free_session:
3621        __free_session(new);
3622out_release_drc_mem:
3623        nfsd4_put_drc_mem(&cr_ses->fore_channel);
3624        return status;
3625}
3626
3627static __be32 nfsd4_map_bcts_dir(u32 *dir)
3628{
3629        switch (*dir) {
3630        case NFS4_CDFC4_FORE:
3631        case NFS4_CDFC4_BACK:
3632                return nfs_ok;
3633        case NFS4_CDFC4_FORE_OR_BOTH:
3634        case NFS4_CDFC4_BACK_OR_BOTH:
3635                *dir = NFS4_CDFC4_BOTH;
3636                return nfs_ok;
3637        }
3638        return nfserr_inval;
3639}
3640
3641__be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp,
3642                struct nfsd4_compound_state *cstate,
3643                union nfsd4_op_u *u)
3644{
3645        struct nfsd4_backchannel_ctl *bc = &u->backchannel_ctl;
3646        struct nfsd4_session *session = cstate->session;
3647        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3648        __be32 status;
3649
3650        status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
3651        if (status)
3652                return status;
3653        spin_lock(&nn->client_lock);
3654        session->se_cb_prog = bc->bc_cb_program;
3655        session->se_cb_sec = bc->bc_cb_sec;
3656        spin_unlock(&nn->client_lock);
3657
3658        nfsd4_probe_callback(session->se_client);
3659
3660        return nfs_ok;
3661}
3662
3663static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
3664{
3665        struct nfsd4_conn *c;
3666
3667        list_for_each_entry(c, &s->se_conns, cn_persession) {
3668                if (c->cn_xprt == xpt) {
3669                        return c;
3670                }
3671        }
3672        return NULL;
3673}
3674
3675static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst,
3676                struct nfsd4_session *session, u32 req, struct nfsd4_conn **conn)
3677{
3678        struct nfs4_client *clp = session->se_client;
3679        struct svc_xprt *xpt = rqst->rq_xprt;
3680        struct nfsd4_conn *c;
3681        __be32 status;
3682
3683        /* Following the last paragraph of RFC 5661 Section 18.34.3: */
3684        spin_lock(&clp->cl_lock);
3685        c = __nfsd4_find_conn(xpt, session);
3686        if (!c)
3687                status = nfserr_noent;
3688        else if (req == c->cn_flags)
3689                status = nfs_ok;
3690        else if (req == NFS4_CDFC4_FORE_OR_BOTH &&
3691                                c->cn_flags != NFS4_CDFC4_BACK)
3692                status = nfs_ok;
3693        else if (req == NFS4_CDFC4_BACK_OR_BOTH &&
3694                                c->cn_flags != NFS4_CDFC4_FORE)
3695                status = nfs_ok;
3696        else
3697                status = nfserr_inval;
3698        spin_unlock(&clp->cl_lock);
3699        if (status == nfs_ok && conn)
3700                *conn = c;
3701        return status;
3702}
3703
3704__be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
3705                     struct nfsd4_compound_state *cstate,
3706                     union nfsd4_op_u *u)
3707{
3708        struct nfsd4_bind_conn_to_session *bcts = &u->bind_conn_to_session;
3709        __be32 status;
3710        struct nfsd4_conn *conn;
3711        struct nfsd4_session *session;
3712        struct net *net = SVC_NET(rqstp);
3713        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3714
3715        if (!nfsd4_last_compound_op(rqstp))
3716                return nfserr_not_only_op;
3717        spin_lock(&nn->client_lock);
3718        session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
3719        spin_unlock(&nn->client_lock);
3720        if (!session)
3721                goto out_no_session;
3722        status = nfserr_wrong_cred;
3723        if (!nfsd4_mach_creds_match(session->se_client, rqstp))
3724                goto out;
3725        status = nfsd4_match_existing_connection(rqstp, session,
3726                        bcts->dir, &conn);
3727        if (status == nfs_ok) {
3728                if (bcts->dir == NFS4_CDFC4_FORE_OR_BOTH ||
3729                                bcts->dir == NFS4_CDFC4_BACK)
3730                        conn->cn_flags |= NFS4_CDFC4_BACK;
3731                nfsd4_probe_callback(session->se_client);
3732                goto out;
3733        }
3734        if (status == nfserr_inval)
3735                goto out;
3736        status = nfsd4_map_bcts_dir(&bcts->dir);
3737        if (status)
3738                goto out;
3739        conn = alloc_conn(rqstp, bcts->dir);
3740        status = nfserr_jukebox;
3741        if (!conn)
3742                goto out;
3743        nfsd4_init_conn(rqstp, conn, session);
3744        status = nfs_ok;
3745out:
3746        nfsd4_put_session(session);
3747out_no_session:
3748        return status;
3749}
3750
3751static bool nfsd4_compound_in_session(struct nfsd4_compound_state *cstate, struct nfs4_sessionid *sid)
3752{
3753        if (!cstate->session)
3754                return false;
3755        return !memcmp(sid, &cstate->session->se_sessionid, sizeof(*sid));
3756}
3757
3758__be32
3759nfsd4_destroy_session(struct svc_rqst *r, struct nfsd4_compound_state *cstate,
3760                union nfsd4_op_u *u)
3761{
3762        struct nfs4_sessionid *sessionid = &u->destroy_session.sessionid;
3763        struct nfsd4_session *ses;
3764        __be32 status;
3765        int ref_held_by_me = 0;
3766        struct net *net = SVC_NET(r);
3767        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3768
3769        status = nfserr_not_only_op;
3770        if (nfsd4_compound_in_session(cstate, sessionid)) {
3771                if (!nfsd4_last_compound_op(r))
3772                        goto out;
3773                ref_held_by_me++;
3774        }
3775        dump_sessionid(__func__, sessionid);
3776        spin_lock(&nn->client_lock);
3777        ses = find_in_sessionid_hashtbl(sessionid, net, &status);
3778        if (!ses)
3779                goto out_client_lock;
3780        status = nfserr_wrong_cred;
3781        if (!nfsd4_mach_creds_match(ses->se_client, r))
3782                goto out_put_session;
3783        status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
3784        if (status)
3785                goto out_put_session;
3786        unhash_session(ses);
3787        spin_unlock(&nn->client_lock);
3788
3789        nfsd4_probe_callback_sync(ses->se_client);
3790
3791        spin_lock(&nn->client_lock);
3792        status = nfs_ok;
3793out_put_session:
3794        nfsd4_put_session_locked(ses);
3795out_client_lock:
3796        spin_unlock(&nn->client_lock);
3797out:
3798        return status;
3799}
3800
3801static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
3802{
3803        struct nfs4_client *clp = ses->se_client;
3804        struct nfsd4_conn *c;
3805        __be32 status = nfs_ok;
3806        int ret;
3807
3808        spin_lock(&clp->cl_lock);
3809        c = __nfsd4_find_conn(new->cn_xprt, ses);
3810        if (c)
3811                goto out_free;
3812        status = nfserr_conn_not_bound_to_session;
3813        if (clp->cl_mach_cred)
3814                goto out_free;
3815        __nfsd4_hash_conn(new, ses);
3816        spin_unlock(&clp->cl_lock);
3817        ret = nfsd4_register_conn(new);
3818        if (ret)
3819                /* oops; xprt is already down: */
3820                nfsd4_conn_lost(&new->cn_xpt_user);
3821        return nfs_ok;
3822out_free:
3823        spin_unlock(&clp->cl_lock);
3824        free_conn(new);
3825        return status;
3826}
3827
3828static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
3829{
3830        struct nfsd4_compoundargs *args = rqstp->rq_argp;
3831
3832        return args->opcnt > session->se_fchannel.maxops;
3833}
3834
3835static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
3836                                  struct nfsd4_session *session)
3837{
3838        struct xdr_buf *xb = &rqstp->rq_arg;
3839
3840        return xb->len > session->se_fchannel.maxreq_sz;
3841}
3842
3843static bool replay_matches_cache(struct svc_rqst *rqstp,
3844                 struct nfsd4_sequence *seq, struct nfsd4_slot *slot)
3845{
3846        struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3847
3848        if ((bool)(slot->sl_flags & NFSD4_SLOT_CACHETHIS) !=
3849            (bool)seq->cachethis)
3850                return false;
3851        /*
3852         * If there's an error then the reply can have fewer ops than
3853         * the call.
3854         */
3855        if (slot->sl_opcnt < argp->opcnt && !slot->sl_status)
3856                return false;
3857        /*
3858         * But if we cached a reply with *more* ops than the call you're
3859         * sending us now, then this new call is clearly not really a
3860         * replay of the old one:
3861         */
3862        if (slot->sl_opcnt > argp->opcnt)
3863                return false;
3864        /* This is the only check explicitly called by spec: */
3865        if (!same_creds(&rqstp->rq_cred, &slot->sl_cred))
3866                return false;
3867        /*
3868         * There may be more comparisons we could actually do, but the
3869         * spec doesn't require us to catch every case where the calls
3870         * don't match (that would require caching the call as well as
3871         * the reply), so we don't bother.
3872         */
3873        return true;
3874}
3875
3876__be32
3877nfsd4_sequence(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3878                union nfsd4_op_u *u)
3879{
3880        struct nfsd4_sequence *seq = &u->sequence;
3881        struct nfsd4_compoundres *resp = rqstp->rq_resp;
3882        struct xdr_stream *xdr = resp->xdr;
3883        struct nfsd4_session *session;
3884        struct nfs4_client *clp;
3885        struct nfsd4_slot *slot;
3886        struct nfsd4_conn *conn;
3887        __be32 status;
3888        int buflen;
3889        struct net *net = SVC_NET(rqstp);
3890        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3891
3892        if (resp->opcnt != 1)
3893                return nfserr_sequence_pos;
3894
3895        /*
3896         * Will be either used or freed by nfsd4_sequence_check_conn
3897         * below.
3898         */
3899        conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
3900        if (!conn)
3901                return nfserr_jukebox;
3902
3903        spin_lock(&nn->client_lock);
3904        session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
3905        if (!session)
3906                goto out_no_session;
3907        clp = session->se_client;
3908
3909        status = nfserr_too_many_ops;
3910        if (nfsd4_session_too_many_ops(rqstp, session))
3911                goto out_put_session;
3912
3913        status = nfserr_req_too_big;
3914        if (nfsd4_request_too_big(rqstp, session))
3915                goto out_put_session;
3916
3917        status = nfserr_badslot;
3918        if (seq->slotid >= session->se_fchannel.maxreqs)
3919                goto out_put_session;
3920
3921        slot = session->se_slots[seq->slotid];
3922        dprintk("%s: slotid %d\n", __func__, seq->slotid);
3923
3924        /* We do not negotiate the number of slots yet, so set the
3925         * maxslots to the session maxreqs which is used to encode
3926         * sr_highest_slotid and the sr_target_slot id to maxslots */
3927        seq->maxslots = session->se_fchannel.maxreqs;
3928
3929        status = check_slot_seqid(seq->seqid, slot->sl_seqid,
3930                                        slot->sl_flags & NFSD4_SLOT_INUSE);
3931        if (status == nfserr_replay_cache) {
3932                status = nfserr_seq_misordered;
3933                if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
3934                        goto out_put_session;
3935                status = nfserr_seq_false_retry;
3936                if (!replay_matches_cache(rqstp, seq, slot))
3937                        goto out_put_session;
3938                cstate->slot = slot;
3939                cstate->session = session;
3940                cstate->clp = clp;
3941                /* Return the cached reply status and set cstate->status
3942                 * for nfsd4_proc_compound processing */
3943                status = nfsd4_replay_cache_entry(resp, seq);
3944                cstate->status = nfserr_replay_cache;
3945                goto out;
3946        }
3947        if (status)
3948                goto out_put_session;
3949
3950        status = nfsd4_sequence_check_conn(conn, session);
3951        conn = NULL;
3952        if (status)
3953                goto out_put_session;
3954
3955        buflen = (seq->cachethis) ?
3956                        session->se_fchannel.maxresp_cached :
3957                        session->se_fchannel.maxresp_sz;
3958        status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
3959                                    nfserr_rep_too_big;
3960        if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
3961                goto out_put_session;
3962        svc_reserve(rqstp, buflen);
3963
3964        status = nfs_ok;
3965        /* Success! bump slot seqid */
3966        slot->sl_seqid = seq->seqid;
3967        slot->sl_flags |= NFSD4_SLOT_INUSE;
3968        if (seq->cachethis)
3969                slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
3970        else
3971                slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
3972
3973        cstate->slot = slot;
3974        cstate->session = session;
3975        cstate->clp = clp;
3976
3977out:
3978        switch (clp->cl_cb_state) {
3979        case NFSD4_CB_DOWN:
3980                seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
3981                break;
3982        case NFSD4_CB_FAULT:
3983                seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
3984                break;
3985        default:
3986                seq->status_flags = 0;
3987        }
3988        if (!list_empty(&clp->cl_revoked))
3989                seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
3990out_no_session:
3991        if (conn)
3992                free_conn(conn);
3993        spin_unlock(&nn->client_lock);
3994        return status;
3995out_put_session:
3996        nfsd4_put_session_locked(session);
3997        goto out_no_session;
3998}
3999
4000void
4001nfsd4_sequence_done(struct nfsd4_compoundres *resp)
4002{
4003        struct nfsd4_compound_state *cs = &resp->cstate;
4004
4005        if (nfsd4_has_session(cs)) {
4006                if (cs->status != nfserr_replay_cache) {
4007                        nfsd4_store_cache_entry(resp);
4008                        cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
4009                }
4010                /* Drop session reference that was taken in nfsd4_sequence() */
4011                nfsd4_put_session(cs->session);
4012        } else if (cs->clp)
4013                put_client_renew(cs->clp);
4014}
4015
4016__be32
4017nfsd4_destroy_clientid(struct svc_rqst *rqstp,
4018                struct nfsd4_compound_state *cstate,
4019                union nfsd4_op_u *u)
4020{
4021        struct nfsd4_destroy_clientid *dc = &u->destroy_clientid;
4022        struct nfs4_client *conf, *unconf;
4023        struct nfs4_client *clp = NULL;
4024        __be32 status = 0;
4025        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4026
4027        spin_lock(&nn->client_lock);
4028        unconf = find_unconfirmed_client(&dc->clientid, true, nn);
4029        conf = find_confirmed_client(&dc->clientid, true, nn);
4030        WARN_ON_ONCE(conf && unconf);
4031
4032        if (conf) {
4033                if (client_has_state(conf)) {
4034                        status = nfserr_clientid_busy;
4035                        goto out;
4036                }
4037                status = mark_client_expired_locked(conf);
4038                if (status)
4039                        goto out;
4040                clp = conf;
4041        } else if (unconf)
4042                clp = unconf;
4043        else {
4044                status = nfserr_stale_clientid;
4045                goto out;
4046        }
4047        if (!nfsd4_mach_creds_match(clp, rqstp)) {
4048                clp = NULL;
4049                status = nfserr_wrong_cred;
4050                goto out;
4051        }
4052        trace_nfsd_clid_destroyed(&clp->cl_clientid);
4053        unhash_client_locked(clp);
4054out:
4055        spin_unlock(&nn->client_lock);
4056        if (clp)
4057                expire_client(clp);
4058        return status;
4059}
4060
4061__be32
4062nfsd4_reclaim_complete(struct svc_rqst *rqstp,
4063                struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
4064{
4065        struct nfsd4_reclaim_complete *rc = &u->reclaim_complete;
4066        struct nfs4_client *clp = cstate->clp;
4067        __be32 status = 0;
4068
4069        if (rc->rca_one_fs) {
4070                if (!cstate->current_fh.fh_dentry)
4071                        return nfserr_nofilehandle;
4072                /*
4073                 * We don't take advantage of the rca_one_fs case.
4074                 * That's OK, it's optional, we can safely ignore it.
4075                 */
4076                return nfs_ok;
4077        }
4078
4079        status = nfserr_complete_already;
4080        if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &clp->cl_flags))
4081                goto out;
4082
4083        status = nfserr_stale_clientid;
4084        if (is_client_expired(clp))
4085                /*
4086                 * The following error isn't really legal.
4087                 * But we only get here if the client just explicitly
4088                 * destroyed the client.  Surely it no longer cares what
4089                 * error it gets back on an operation for the dead
4090                 * client.
4091                 */
4092                goto out;
4093
4094        status = nfs_ok;
4095        trace_nfsd_clid_reclaim_complete(&clp->cl_clientid);
4096        nfsd4_client_record_create(clp);
4097        inc_reclaim_complete(clp);
4098out:
4099        return status;
4100}
4101
4102__be32
4103nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4104                  union nfsd4_op_u *u)
4105{
4106        struct nfsd4_setclientid *setclid = &u->setclientid;
4107        struct xdr_netobj       clname = setclid->se_name;
4108        nfs4_verifier           clverifier = setclid->se_verf;
4109        struct nfs4_client      *conf, *new;
4110        struct nfs4_client      *unconf = NULL;
4111        __be32                  status;
4112        struct nfsd_net         *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4113
4114        new = create_client(clname, rqstp, &clverifier);
4115        if (new == NULL)
4116                return nfserr_jukebox;
4117        spin_lock(&nn->client_lock);
4118        conf = find_confirmed_client_by_name(&clname, nn);
4119        if (conf && client_has_state(conf)) {
4120                status = nfserr_clid_inuse;
4121                if (clp_used_exchangeid(conf))
4122                        goto out;
4123                if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
4124                        trace_nfsd_clid_cred_mismatch(conf, rqstp);
4125                        goto out;
4126                }
4127        }
4128        unconf = find_unconfirmed_client_by_name(&clname, nn);
4129        if (unconf)
4130                unhash_client_locked(unconf);
4131        if (conf) {
4132                if (same_verf(&conf->cl_verifier, &clverifier)) {
4133                        copy_clid(new, conf);
4134                        gen_confirm(new, nn);
4135                } else
4136                        trace_nfsd_clid_verf_mismatch(conf, rqstp,
4137                                                      &clverifier);
4138        } else
4139                trace_nfsd_clid_fresh(new);
4140        new->cl_minorversion = 0;
4141        gen_callback(new, setclid, rqstp);
4142        add_to_unconfirmed(new);
4143        setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
4144        setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
4145        memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
4146        new = NULL;
4147        status = nfs_ok;
4148out:
4149        spin_unlock(&nn->client_lock);
4150        if (new)
4151                free_client(new);
4152        if (unconf) {
4153                trace_nfsd_clid_expire_unconf(&unconf->cl_clientid);
4154                expire_client(unconf);
4155        }
4156        return status;
4157}
4158
4159__be32
4160nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
4161                        struct nfsd4_compound_state *cstate,
4162                        union nfsd4_op_u *u)
4163{
4164        struct nfsd4_setclientid_confirm *setclientid_confirm =
4165                        &u->setclientid_confirm;
4166        struct nfs4_client *conf, *unconf;
4167        struct nfs4_client *old = NULL;
4168        nfs4_verifier confirm = setclientid_confirm->sc_confirm; 
4169        clientid_t * clid = &setclientid_confirm->sc_clientid;
4170        __be32 status;
4171        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4172
4173        if (STALE_CLIENTID(clid, nn))
4174                return nfserr_stale_clientid;
4175
4176        spin_lock(&nn->client_lock);
4177        conf = find_confirmed_client(clid, false, nn);
4178        unconf = find_unconfirmed_client(clid, false, nn);
4179        /*
4180         * We try hard to give out unique clientid's, so if we get an
4181         * attempt to confirm the same clientid with a different cred,
4182         * the client may be buggy; this should never happen.
4183         *
4184         * Nevertheless, RFC 7530 recommends INUSE for this case:
4185         */
4186        status = nfserr_clid_inuse;
4187        if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
4188                trace_nfsd_clid_cred_mismatch(unconf, rqstp);
4189                goto out;
4190        }
4191        if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
4192                trace_nfsd_clid_cred_mismatch(conf, rqstp);
4193                goto out;
4194        }
4195        if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
4196                if (conf && same_verf(&confirm, &conf->cl_confirm)) {
4197                        status = nfs_ok;
4198                } else
4199                        status = nfserr_stale_clientid;
4200                goto out;
4201        }
4202        status = nfs_ok;
4203        if (conf) {
4204                old = unconf;
4205                unhash_client_locked(old);
4206                nfsd4_change_callback(conf, &unconf->cl_cb_conn);
4207        } else {
4208                old = find_confirmed_client_by_name(&unconf->cl_name, nn);
4209                if (old) {
4210                        status = nfserr_clid_inuse;
4211                        if (client_has_state(old)
4212                                        && !same_creds(&unconf->cl_cred,
4213                                                        &old->cl_cred)) {
4214                                old = NULL;
4215                                goto out;
4216                        }
4217                        status = mark_client_expired_locked(old);
4218                        if (status) {
4219                                old = NULL;
4220                                goto out;
4221                        }
4222                        trace_nfsd_clid_replaced(&old->cl_clientid);
4223                }
4224                move_to_confirmed(unconf);
4225                conf = unconf;
4226        }
4227        get_client_locked(conf);
4228        spin_unlock(&nn->client_lock);
4229        if (conf == unconf)
4230                fsnotify_dentry(conf->cl_nfsd_info_dentry, FS_MODIFY);
4231        nfsd4_probe_callback(conf);
4232        spin_lock(&nn->client_lock);
4233        put_client_renew_locked(conf);
4234out:
4235        spin_unlock(&nn->client_lock);
4236        if (old)
4237                expire_client(old);
4238        return status;
4239}
4240
4241static struct nfs4_file *nfsd4_alloc_file(void)
4242{
4243        return kmem_cache_alloc(file_slab, GFP_KERNEL);
4244}
4245
4246/* OPEN Share state helper functions */
4247static void nfsd4_init_file(struct svc_fh *fh, unsigned int hashval,
4248                                struct nfs4_file *fp)
4249{
4250        lockdep_assert_held(&state_lock);
4251
4252        refcount_set(&fp->fi_ref, 1);
4253        spin_lock_init(&fp->fi_lock);
4254        INIT_LIST_HEAD(&fp->fi_stateids);
4255        INIT_LIST_HEAD(&fp->fi_delegations);
4256        INIT_LIST_HEAD(&fp->fi_clnt_odstate);
4257        fh_copy_shallow(&fp->fi_fhandle, &fh->fh_handle);
4258        fp->fi_deleg_file = NULL;
4259        fp->fi_had_conflict = false;
4260        fp->fi_share_deny = 0;
4261        memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
4262        memset(fp->fi_access, 0, sizeof(fp->fi_access));
4263        fp->fi_aliased = false;
4264        fp->fi_inode = d_inode(fh->fh_dentry);
4265#ifdef CONFIG_NFSD_PNFS
4266        INIT_LIST_HEAD(&fp->fi_lo_states);
4267        atomic_set(&fp->fi_lo_recalls, 0);
4268#endif
4269        hlist_add_head_rcu(&fp->fi_hash, &file_hashtbl[hashval]);
4270}
4271
4272void
4273nfsd4_free_slabs(void)
4274{
4275        kmem_cache_destroy(client_slab);
4276        kmem_cache_destroy(openowner_slab);
4277        kmem_cache_destroy(lockowner_slab);
4278        kmem_cache_destroy(file_slab);
4279        kmem_cache_destroy(stateid_slab);
4280        kmem_cache_destroy(deleg_slab);
4281        kmem_cache_destroy(odstate_slab);
4282}
4283
4284int
4285nfsd4_init_slabs(void)
4286{
4287        client_slab = kmem_cache_create("nfsd4_clients",
4288                        sizeof(struct nfs4_client), 0, 0, NULL);
4289        if (client_slab == NULL)
4290                goto out;
4291        openowner_slab = kmem_cache_create("nfsd4_openowners",
4292                        sizeof(struct nfs4_openowner), 0, 0, NULL);
4293        if (openowner_slab == NULL)
4294                goto out_free_client_slab;
4295        lockowner_slab = kmem_cache_create("nfsd4_lockowners",
4296                        sizeof(struct nfs4_lockowner), 0, 0, NULL);
4297        if (lockowner_slab == NULL)
4298                goto out_free_openowner_slab;
4299        file_slab = kmem_cache_create("nfsd4_files",
4300                        sizeof(struct nfs4_file), 0, 0, NULL);
4301        if (file_slab == NULL)
4302                goto out_free_lockowner_slab;
4303        stateid_slab = kmem_cache_create("nfsd4_stateids",
4304                        sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
4305        if (stateid_slab == NULL)
4306                goto out_free_file_slab;
4307        deleg_slab = kmem_cache_create("nfsd4_delegations",
4308                        sizeof(struct nfs4_delegation), 0, 0, NULL);
4309        if (deleg_slab == NULL)
4310                goto out_free_stateid_slab;
4311        odstate_slab = kmem_cache_create("nfsd4_odstate",
4312                        sizeof(struct nfs4_clnt_odstate), 0, 0, NULL);
4313        if (odstate_slab == NULL)
4314                goto out_free_deleg_slab;
4315        return 0;
4316
4317out_free_deleg_slab:
4318        kmem_cache_destroy(deleg_slab);
4319out_free_stateid_slab:
4320        kmem_cache_destroy(stateid_slab);
4321out_free_file_slab:
4322        kmem_cache_destroy(file_slab);
4323out_free_lockowner_slab:
4324        kmem_cache_destroy(lockowner_slab);
4325out_free_openowner_slab:
4326        kmem_cache_destroy(openowner_slab);
4327out_free_client_slab:
4328        kmem_cache_destroy(client_slab);
4329out:
4330        return -ENOMEM;
4331}
4332
4333static void init_nfs4_replay(struct nfs4_replay *rp)
4334{
4335        rp->rp_status = nfserr_serverfault;
4336        rp->rp_buflen = 0;
4337        rp->rp_buf = rp->rp_ibuf;
4338        mutex_init(&rp->rp_mutex);
4339}
4340
4341static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
4342                struct nfs4_stateowner *so)
4343{
4344        if (!nfsd4_has_session(cstate)) {
4345                mutex_lock(&so->so_replay.rp_mutex);
4346                cstate->replay_owner = nfs4_get_stateowner(so);
4347        }
4348}
4349
4350void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
4351{
4352        struct nfs4_stateowner *so = cstate->replay_owner;
4353
4354        if (so != NULL) {
4355                cstate->replay_owner = NULL;
4356                mutex_unlock(&so->so_replay.rp_mutex);
4357                nfs4_put_stateowner(so);
4358        }
4359}
4360
4361static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
4362{
4363        struct nfs4_stateowner *sop;
4364
4365        sop = kmem_cache_alloc(slab, GFP_KERNEL);
4366        if (!sop)
4367                return NULL;
4368
4369        xdr_netobj_dup(&sop->so_owner, owner, GFP_KERNEL);
4370        if (!sop->so_owner.data) {
4371                kmem_cache_free(slab, sop);
4372                return NULL;
4373        }
4374
4375        INIT_LIST_HEAD(&sop->so_stateids);
4376        sop->so_client = clp;
4377        init_nfs4_replay(&sop->so_replay);
4378        atomic_set(&sop->so_count, 1);
4379        return sop;
4380}
4381
4382static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
4383{
4384        lockdep_assert_held(&clp->cl_lock);
4385
4386        list_add(&oo->oo_owner.so_strhash,
4387                 &clp->cl_ownerstr_hashtbl[strhashval]);
4388        list_add(&oo->oo_perclient, &clp->cl_openowners);
4389}
4390
4391static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
4392{
4393        unhash_openowner_locked(openowner(so));
4394}
4395
4396static void nfs4_free_openowner(struct nfs4_stateowner *so)
4397{
4398        struct nfs4_openowner *oo = openowner(so);
4399
4400        kmem_cache_free(openowner_slab, oo);
4401}
4402
4403static const struct nfs4_stateowner_operations openowner_ops = {
4404        .so_unhash =    nfs4_unhash_openowner,
4405        .so_free =      nfs4_free_openowner,
4406};
4407
4408static struct nfs4_ol_stateid *
4409nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
4410{
4411        struct nfs4_ol_stateid *local, *ret = NULL;
4412        struct nfs4_openowner *oo = open->op_openowner;
4413
4414        lockdep_assert_held(&fp->fi_lock);
4415
4416        list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
4417                /* ignore lock owners */
4418                if (local->st_stateowner->so_is_open_owner == 0)
4419                        continue;
4420                if (local->st_stateowner != &oo->oo_owner)
4421                        continue;
4422                if (local->st_stid.sc_type == NFS4_OPEN_STID) {
4423                        ret = local;
4424                        refcount_inc(&ret->st_stid.sc_count);
4425                        break;
4426                }
4427        }
4428        return ret;
4429}
4430
4431static __be32
4432nfsd4_verify_open_stid(struct nfs4_stid *s)
4433{
4434        __be32 ret = nfs_ok;
4435
4436        switch (s->sc_type) {
4437        default:
4438                break;
4439        case 0:
4440        case NFS4_CLOSED_STID:
4441        case NFS4_CLOSED_DELEG_STID:
4442                ret = nfserr_bad_stateid;
4443                break;
4444        case NFS4_REVOKED_DELEG_STID:
4445                ret = nfserr_deleg_revoked;
4446        }
4447        return ret;
4448}
4449
4450/* Lock the stateid st_mutex, and deal with races with CLOSE */
4451static __be32
4452nfsd4_lock_ol_stateid(struct nfs4_ol_stateid *stp)
4453{
4454        __be32 ret;
4455
4456        mutex_lock_nested(&stp->st_mutex, LOCK_STATEID_MUTEX);
4457        ret = nfsd4_verify_open_stid(&stp->st_stid);
4458        if (ret != nfs_ok)
4459                mutex_unlock(&stp->st_mutex);
4460        return ret;
4461}
4462
4463static struct nfs4_ol_stateid *
4464nfsd4_find_and_lock_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
4465{
4466        struct nfs4_ol_stateid *stp;
4467        for (;;) {
4468                spin_lock(&fp->fi_lock);
4469                stp = nfsd4_find_existing_open(fp, open);
4470                spin_unlock(&fp->fi_lock);
4471                if (!stp || nfsd4_lock_ol_stateid(stp) == nfs_ok)
4472                        break;
4473                nfs4_put_stid(&stp->st_stid);
4474        }
4475        return stp;
4476}
4477
4478static struct nfs4_openowner *
4479alloc_init_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
4480                           struct nfsd4_compound_state *cstate)
4481{
4482        struct nfs4_client *clp = cstate->clp;
4483        struct nfs4_openowner *oo, *ret;
4484
4485        oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
4486        if (!oo)
4487                return NULL;
4488        oo->oo_owner.so_ops = &openowner_ops;
4489        oo->oo_owner.so_is_open_owner = 1;
4490        oo->oo_owner.so_seqid = open->op_seqid;
4491        oo->oo_flags = 0;
4492        if (nfsd4_has_session(cstate))
4493                oo->oo_flags |= NFS4_OO_CONFIRMED;
4494        oo->oo_time = 0;
4495        oo->oo_last_closed_stid = NULL;
4496        INIT_LIST_HEAD(&oo->oo_close_lru);
4497        spin_lock(&clp->cl_lock);
4498        ret = find_openstateowner_str_locked(strhashval, open, clp);
4499        if (ret == NULL) {
4500                hash_openowner(oo, clp, strhashval);
4501                ret = oo;
4502        } else
4503                nfs4_free_stateowner(&oo->oo_owner);
4504
4505        spin_unlock(&clp->cl_lock);
4506        return ret;
4507}
4508
4509static struct nfs4_ol_stateid *
4510init_open_stateid(struct nfs4_file *fp, struct nfsd4_open *open)
4511{
4512
4513        struct nfs4_openowner *oo = open->op_openowner;
4514        struct nfs4_ol_stateid *retstp = NULL;
4515        struct nfs4_ol_stateid *stp;
4516
4517        stp = open->op_stp;
4518        /* We are moving these outside of the spinlocks to avoid the warnings */
4519        mutex_init(&stp->st_mutex);
4520        mutex_lock_nested(&stp->st_mutex, OPEN_STATEID_MUTEX);
4521
4522retry:
4523        spin_lock(&oo->oo_owner.so_client->cl_lock);
4524        spin_lock(&fp->fi_lock);
4525
4526        retstp = nfsd4_find_existing_open(fp, open);
4527        if (retstp)
4528                goto out_unlock;
4529
4530        open->op_stp = NULL;
4531        refcount_inc(&stp->st_stid.sc_count);
4532        stp->st_stid.sc_type = NFS4_OPEN_STID;
4533        INIT_LIST_HEAD(&stp->st_locks);
4534        stp->st_stateowner = nfs4_get_stateowner(&oo->oo_owner);
4535        get_nfs4_file(fp);
4536        stp->st_stid.sc_file = fp;
4537        stp->st_access_bmap = 0;
4538        stp->st_deny_bmap = 0;
4539        stp->st_openstp = NULL;
4540        list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
4541        list_add(&stp->st_perfile, &fp->fi_stateids);
4542
4543out_unlock:
4544        spin_unlock(&fp->fi_lock);
4545        spin_unlock(&oo->oo_owner.so_client->cl_lock);
4546        if (retstp) {
4547                /* Handle races with CLOSE */
4548                if (nfsd4_lock_ol_stateid(retstp) != nfs_ok) {
4549                        nfs4_put_stid(&retstp->st_stid);
4550                        goto retry;
4551                }
4552                /* To keep mutex tracking happy */
4553                mutex_unlock(&stp->st_mutex);
4554                stp = retstp;
4555        }
4556        return stp;
4557}
4558
4559/*
4560 * In the 4.0 case we need to keep the owners around a little while to handle
4561 * CLOSE replay. We still do need to release any file access that is held by
4562 * them before returning however.
4563 */
4564static void
4565move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
4566{
4567        struct nfs4_ol_stateid *last;
4568        struct nfs4_openowner *oo = openowner(s->st_stateowner);
4569        struct nfsd_net *nn = net_generic(s->st_stid.sc_client->net,
4570                                                nfsd_net_id);
4571
4572        dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
4573
4574        /*
4575         * We know that we hold one reference via nfsd4_close, and another
4576         * "persistent" reference for the client. If the refcount is higher
4577         * than 2, then there are still calls in progress that are using this
4578         * stateid. We can't put the sc_file reference until they are finished.
4579         * Wait for the refcount to drop to 2. Since it has been unhashed,
4580         * there should be no danger of the refcount going back up again at
4581         * this point.
4582         */
4583        wait_event(close_wq, refcount_read(&s->st_stid.sc_count) == 2);
4584
4585        release_all_access(s);
4586        if (s->st_stid.sc_file) {
4587                put_nfs4_file(s->st_stid.sc_file);
4588                s->st_stid.sc_file = NULL;
4589        }
4590
4591        spin_lock(&nn->client_lock);
4592        last = oo->oo_last_closed_stid;
4593        oo->oo_last_closed_stid = s;
4594        list_move_tail(&oo->oo_close_lru, &nn->close_lru);
4595        oo->oo_time = ktime_get_boottime_seconds();
4596        spin_unlock(&nn->client_lock);
4597        if (last)
4598                nfs4_put_stid(&last->st_stid);
4599}
4600
4601/* search file_hashtbl[] for file */
4602static struct nfs4_file *
4603find_file_locked(struct svc_fh *fh, unsigned int hashval)
4604{
4605        struct nfs4_file *fp;
4606
4607        hlist_for_each_entry_rcu(fp, &file_hashtbl[hashval], fi_hash,
4608                                lockdep_is_held(&state_lock)) {
4609                if (fh_match(&fp->fi_fhandle, &fh->fh_handle)) {
4610                        if (refcount_inc_not_zero(&fp->fi_ref))
4611                                return fp;
4612                }
4613        }
4614        return NULL;
4615}
4616
4617static struct nfs4_file *insert_file(struct nfs4_file *new, struct svc_fh *fh,
4618                                     unsigned int hashval)
4619{
4620        struct nfs4_file *fp;
4621        struct nfs4_file *ret = NULL;
4622        bool alias_found = false;
4623
4624        spin_lock(&state_lock);
4625        hlist_for_each_entry_rcu(fp, &file_hashtbl[hashval], fi_hash,
4626                                 lockdep_is_held(&state_lock)) {
4627                if (fh_match(&fp->fi_fhandle, &fh->fh_handle)) {
4628                        if (refcount_inc_not_zero(&fp->fi_ref))
4629                                ret = fp;
4630                } else if (d_inode(fh->fh_dentry) == fp->fi_inode)
4631                        fp->fi_aliased = alias_found = true;
4632        }
4633        if (likely(ret == NULL)) {
4634                nfsd4_init_file(fh, hashval, new);
4635                new->fi_aliased = alias_found;
4636                ret = new;
4637        }
4638        spin_unlock(&state_lock);
4639        return ret;
4640}
4641
4642static struct nfs4_file * find_file(struct svc_fh *fh)
4643{
4644        struct nfs4_file *fp;
4645        unsigned int hashval = file_hashval(fh);
4646
4647        rcu_read_lock();
4648        fp = find_file_locked(fh, hashval);
4649        rcu_read_unlock();
4650        return fp;
4651}
4652
4653static struct nfs4_file *
4654find_or_add_file(struct nfs4_file *new, struct svc_fh *fh)
4655{
4656        struct nfs4_file *fp;
4657        unsigned int hashval = file_hashval(fh);
4658
4659        rcu_read_lock();
4660        fp = find_file_locked(fh, hashval);
4661        rcu_read_unlock();
4662        if (fp)
4663                return fp;
4664
4665        return insert_file(new, fh, hashval);
4666}
4667
4668/*
4669 * Called to check deny when READ with all zero stateid or
4670 * WRITE with all zero or all one stateid
4671 */
4672static __be32
4673nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
4674{
4675        struct nfs4_file *fp;
4676        __be32 ret = nfs_ok;
4677
4678        fp = find_file(current_fh);
4679        if (!fp)
4680                return ret;
4681        /* Check for conflicting share reservations */
4682        spin_lock(&fp->fi_lock);
4683        if (fp->fi_share_deny & deny_type)
4684                ret = nfserr_locked;
4685        spin_unlock(&fp->fi_lock);
4686        put_nfs4_file(fp);
4687        return ret;
4688}
4689
4690static void nfsd4_cb_recall_prepare(struct nfsd4_callback *cb)
4691{
4692        struct nfs4_delegation *dp = cb_to_delegation(cb);
4693        struct nfsd_net *nn = net_generic(dp->dl_stid.sc_client->net,
4694                                          nfsd_net_id);
4695
4696        block_delegations(&dp->dl_stid.sc_file->fi_fhandle);
4697
4698        /*
4699         * We can't do this in nfsd_break_deleg_cb because it is
4700         * already holding inode->i_lock.
4701         *
4702         * If the dl_time != 0, then we know that it has already been
4703         * queued for a lease break. Don't queue it again.
4704         */
4705        spin_lock(&state_lock);
4706        if (delegation_hashed(dp) && dp->dl_time == 0) {
4707                dp->dl_time = ktime_get_boottime_seconds();
4708                list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
4709        }
4710        spin_unlock(&state_lock);
4711}
4712
4713static int nfsd4_cb_recall_done(struct nfsd4_callback *cb,
4714                struct rpc_task *task)
4715{
4716        struct nfs4_delegation *dp = cb_to_delegation(cb);
4717
4718        if (dp->dl_stid.sc_type == NFS4_CLOSED_DELEG_STID ||
4719            dp->dl_stid.sc_type == NFS4_REVOKED_DELEG_STID)
4720                return 1;
4721
4722        switch (task->tk_status) {
4723        case 0:
4724                return 1;
4725        case -NFS4ERR_DELAY:
4726                rpc_delay(task, 2 * HZ);
4727                return 0;
4728        case -EBADHANDLE:
4729        case -NFS4ERR_BAD_STATEID:
4730                /*
4731                 * Race: client probably got cb_recall before open reply
4732                 * granting delegation.
4733                 */
4734                if (dp->dl_retries--) {
4735                        rpc_delay(task, 2 * HZ);
4736                        return 0;
4737                }
4738                fallthrough;
4739        default:
4740                return 1;
4741        }
4742}
4743
4744static void nfsd4_cb_recall_release(struct nfsd4_callback *cb)
4745{
4746        struct nfs4_delegation *dp = cb_to_delegation(cb);
4747
4748        nfs4_put_stid(&dp->dl_stid);
4749}
4750
4751static const struct nfsd4_callback_ops nfsd4_cb_recall_ops = {
4752        .prepare        = nfsd4_cb_recall_prepare,
4753        .done           = nfsd4_cb_recall_done,
4754        .release        = nfsd4_cb_recall_release,
4755};
4756
4757static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
4758{
4759        /*
4760         * We're assuming the state code never drops its reference
4761         * without first removing the lease.  Since we're in this lease
4762         * callback (and since the lease code is serialized by the
4763         * i_lock) we know the server hasn't removed the lease yet, and
4764         * we know it's safe to take a reference.
4765         */
4766        refcount_inc(&dp->dl_stid.sc_count);
4767        nfsd4_run_cb(&dp->dl_recall);
4768}
4769
4770/* Called from break_lease() with i_lock held. */
4771static bool
4772nfsd_break_deleg_cb(struct file_lock *fl)
4773{
4774        bool ret = false;
4775        struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
4776        struct nfs4_file *fp = dp->dl_stid.sc_file;
4777        struct nfs4_client *clp = dp->dl_stid.sc_client;
4778        struct nfsd_net *nn;
4779
4780        trace_nfsd_cb_recall(&dp->dl_stid);
4781
4782        dp->dl_recalled = true;
4783        atomic_inc(&clp->cl_delegs_in_recall);
4784        if (try_to_expire_client(clp)) {
4785                nn = net_generic(clp->net, nfsd_net_id);
4786                mod_delayed_work(laundry_wq, &nn->laundromat_work, 0);
4787        }
4788
4789        /*
4790         * We don't want the locks code to timeout the lease for us;
4791         * we'll remove it ourself if a delegation isn't returned
4792         * in time:
4793         */
4794        fl->fl_break_time = 0;
4795
4796        spin_lock(&fp->fi_lock);
4797        fp->fi_had_conflict = true;
4798        nfsd_break_one_deleg(dp);
4799        spin_unlock(&fp->fi_lock);
4800        return ret;
4801}
4802
4803/**
4804 * nfsd_breaker_owns_lease - Check if lease conflict was resolved
4805 * @fl: Lock state to check
4806 *
4807 * Return values:
4808 *   %true: Lease conflict was resolved
4809 *   %false: Lease conflict was not resolved.
4810 */
4811static bool nfsd_breaker_owns_lease(struct file_lock *fl)
4812{
4813        struct nfs4_delegation *dl = fl->fl_owner;
4814        struct svc_rqst *rqst;
4815        struct nfs4_client *clp;
4816
4817        if (!i_am_nfsd())
4818                return false;
4819        rqst = kthread_data(current);
4820        /* Note rq_prog == NFS_ACL_PROGRAM is also possible: */
4821        if (rqst->rq_prog != NFS_PROGRAM || rqst->rq_vers < 4)
4822                return false;
4823        clp = *(rqst->rq_lease_breaker);
4824        return dl->dl_stid.sc_client == clp;
4825}
4826
4827static int
4828nfsd_change_deleg_cb(struct file_lock *onlist, int arg,
4829                     struct list_head *dispose)
4830{
4831        struct nfs4_delegation *dp = (struct nfs4_delegation *)onlist->fl_owner;
4832        struct nfs4_client *clp = dp->dl_stid.sc_client;
4833
4834        if (arg & F_UNLCK) {
4835                if (dp->dl_recalled)
4836                        atomic_dec(&clp->cl_delegs_in_recall);
4837                return lease_modify(onlist, arg, dispose);
4838        } else
4839                return -EAGAIN;
4840}
4841
4842static const struct lock_manager_operations nfsd_lease_mng_ops = {
4843        .lm_breaker_owns_lease = nfsd_breaker_owns_lease,
4844        .lm_break = nfsd_break_deleg_cb,
4845        .lm_change = nfsd_change_deleg_cb,
4846};
4847
4848static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
4849{
4850        if (nfsd4_has_session(cstate))
4851                return nfs_ok;
4852        if (seqid == so->so_seqid - 1)
4853                return nfserr_replay_me;
4854        if (seqid == so->so_seqid)
4855                return nfs_ok;
4856        return nfserr_bad_seqid;
4857}
4858
4859static struct nfs4_client *lookup_clientid(clientid_t *clid, bool sessions,
4860                                                struct nfsd_net *nn)
4861{
4862        struct nfs4_client *found;
4863
4864        spin_lock(&nn->client_lock);
4865        found = find_confirmed_client(clid, sessions, nn);
4866        if (found)
4867                atomic_inc(&found->cl_rpc_users);
4868        spin_unlock(&nn->client_lock);
4869        return found;
4870}
4871
4872static __be32 set_client(clientid_t *clid,
4873                struct nfsd4_compound_state *cstate,
4874                struct nfsd_net *nn)
4875{
4876        if (cstate->clp) {
4877                if (!same_clid(&cstate->clp->cl_clientid, clid))
4878                        return nfserr_stale_clientid;
4879                return nfs_ok;
4880        }
4881        if (STALE_CLIENTID(clid, nn))
4882                return nfserr_stale_clientid;
4883        /*
4884         * We're in the 4.0 case (otherwise the SEQUENCE op would have
4885         * set cstate->clp), so session = false:
4886         */
4887        cstate->clp = lookup_clientid(clid, false, nn);
4888        if (!cstate->clp)
4889                return nfserr_expired;
4890        return nfs_ok;
4891}
4892
4893__be32
4894nfsd4_process_open1(struct nfsd4_compound_state *cstate,
4895                    struct nfsd4_open *open, struct nfsd_net *nn)
4896{
4897        clientid_t *clientid = &open->op_clientid;
4898        struct nfs4_client *clp = NULL;
4899        unsigned int strhashval;
4900        struct nfs4_openowner *oo = NULL;
4901        __be32 status;
4902
4903        /*
4904         * In case we need it later, after we've already created the
4905         * file and don't want to risk a further failure:
4906         */
4907        open->op_file = nfsd4_alloc_file();
4908        if (open->op_file == NULL)
4909                return nfserr_jukebox;
4910
4911        status = set_client(clientid, cstate, nn);
4912        if (status)
4913                return status;
4914        clp = cstate->clp;
4915
4916        strhashval = ownerstr_hashval(&open->op_owner);
4917        oo = find_openstateowner_str(strhashval, open, clp);
4918        open->op_openowner = oo;
4919        if (!oo) {
4920                goto new_owner;
4921        }
4922        if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
4923                /* Replace unconfirmed owners without checking for replay. */
4924                release_openowner(oo);
4925                open->op_openowner = NULL;
4926                goto new_owner;
4927        }
4928        status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
4929        if (status)
4930                return status;
4931        goto alloc_stateid;
4932new_owner:
4933        oo = alloc_init_open_stateowner(strhashval, open, cstate);
4934        if (oo == NULL)
4935                return nfserr_jukebox;
4936        open->op_openowner = oo;
4937alloc_stateid:
4938        open->op_stp = nfs4_alloc_open_stateid(clp);
4939        if (!open->op_stp)
4940                return nfserr_jukebox;
4941
4942        if (nfsd4_has_session(cstate) &&
4943            (cstate->current_fh.fh_export->ex_flags & NFSEXP_PNFS)) {
4944                open->op_odstate = alloc_clnt_odstate(clp);
4945                if (!open->op_odstate)
4946                        return nfserr_jukebox;
4947        }
4948
4949        return nfs_ok;
4950}
4951
4952static inline __be32
4953nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
4954{
4955        if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
4956                return nfserr_openmode;
4957        else
4958                return nfs_ok;
4959}
4960
4961static int share_access_to_flags(u32 share_access)
4962{
4963        return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
4964}
4965
4966static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
4967{
4968        struct nfs4_stid *ret;
4969
4970        ret = find_stateid_by_type(cl, s,
4971                                NFS4_DELEG_STID|NFS4_REVOKED_DELEG_STID);
4972        if (!ret)
4973                return NULL;
4974        return delegstateid(ret);
4975}
4976
4977static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
4978{
4979        return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
4980               open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
4981}
4982
4983static __be32
4984nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
4985                struct nfs4_delegation **dp)
4986{
4987        int flags;
4988        __be32 status = nfserr_bad_stateid;
4989        struct nfs4_delegation *deleg;
4990
4991        deleg = find_deleg_stateid(cl, &open->op_delegate_stateid);
4992        if (deleg == NULL)
4993                goto out;
4994        if (deleg->dl_stid.sc_type == NFS4_REVOKED_DELEG_STID) {
4995                nfs4_put_stid(&deleg->dl_stid);
4996                if (cl->cl_minorversion)
4997                        status = nfserr_deleg_revoked;
4998                goto out;
4999        }
5000        flags = share_access_to_flags(open->op_share_access);
5001        status = nfs4_check_delegmode(deleg, flags);
5002        if (status) {
5003                nfs4_put_stid(&deleg->dl_stid);
5004                goto out;
5005        }
5006        *dp = deleg;
5007out:
5008        if (!nfsd4_is_deleg_cur(open))
5009                return nfs_ok;
5010        if (status)
5011                return status;
5012        open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
5013        return nfs_ok;
5014}
5015
5016static inline int nfs4_access_to_access(u32 nfs4_access)
5017{
5018        int flags = 0;
5019
5020        if (nfs4_access & NFS4_SHARE_ACCESS_READ)
5021                flags |= NFSD_MAY_READ;
5022        if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
5023                flags |= NFSD_MAY_WRITE;
5024        return flags;
5025}
5026
5027static inline __be32
5028nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
5029                struct nfsd4_open *open)
5030{
5031        struct iattr iattr = {
5032                .ia_valid = ATTR_SIZE,
5033                .ia_size = 0,
5034        };
5035        if (!open->op_truncate)
5036                return 0;
5037        if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
5038                return nfserr_inval;
5039        return nfsd_setattr(rqstp, fh, &iattr, 0, (time64_t)0);
5040}
5041
5042static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
5043                struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
5044                struct nfsd4_open *open, bool new_stp)
5045{
5046        struct nfsd_file *nf = NULL;
5047        __be32 status;
5048        int oflag = nfs4_access_to_omode(open->op_share_access);
5049        int access = nfs4_access_to_access(open->op_share_access);
5050        unsigned char old_access_bmap, old_deny_bmap;
5051
5052        spin_lock(&fp->fi_lock);
5053
5054        /*
5055         * Are we trying to set a deny mode that would conflict with
5056         * current access?
5057         */
5058        status = nfs4_file_check_deny(fp, open->op_share_deny);
5059        if (status != nfs_ok) {
5060                if (status != nfserr_share_denied) {
5061                        spin_unlock(&fp->fi_lock);
5062                        goto out;
5063                }
5064                if (nfs4_resolve_deny_conflicts_locked(fp, new_stp,
5065                                stp, open->op_share_deny, false))
5066                        status = nfserr_jukebox;
5067                spin_unlock(&fp->fi_lock);
5068                goto out;
5069        }
5070
5071        /* set access to the file */
5072        status = nfs4_file_get_access(fp, open->op_share_access);
5073        if (status != nfs_ok) {
5074                if (status != nfserr_share_denied) {
5075                        spin_unlock(&fp->fi_lock);
5076                        goto out;
5077                }
5078                if (nfs4_resolve_deny_conflicts_locked(fp, new_stp,
5079                                stp, open->op_share_access, true))
5080                        status = nfserr_jukebox;
5081                spin_unlock(&fp->fi_lock);
5082                goto out;
5083        }
5084
5085        /* Set access bits in stateid */
5086        old_access_bmap = stp->st_access_bmap;
5087        set_access(open->op_share_access, stp);
5088
5089        /* Set new deny mask */
5090        old_deny_bmap = stp->st_deny_bmap;
5091        set_deny(open->op_share_deny, stp);
5092        fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
5093
5094        if (!fp->fi_fds[oflag]) {
5095                spin_unlock(&fp->fi_lock);
5096
5097                if (!open->op_filp) {
5098                        status = nfsd_file_acquire(rqstp, cur_fh, access, &nf);
5099                        if (status != nfs_ok)
5100                                goto out_put_access;
5101                } else {
5102                        status = nfsd_file_create(rqstp, cur_fh, access, &nf);
5103                        if (status != nfs_ok)
5104                                goto out_put_access;
5105                        nf->nf_file = open->op_filp;
5106                        open->op_filp = NULL;
5107                }
5108
5109                spin_lock(&fp->fi_lock);
5110                if (!fp->fi_fds[oflag]) {
5111                        fp->fi_fds[oflag] = nf;
5112                        nf = NULL;
5113                }
5114        }
5115        spin_unlock(&fp->fi_lock);
5116        if (nf)
5117                nfsd_file_put(nf);
5118
5119        status = nfserrno(nfsd_open_break_lease(cur_fh->fh_dentry->d_inode,
5120                                                                access));
5121        if (status)
5122                goto out_put_access;
5123
5124        status = nfsd4_truncate(rqstp, cur_fh, open);
5125        if (status)
5126                goto out_put_access;
5127out:
5128        return status;
5129out_put_access:
5130        stp->st_access_bmap = old_access_bmap;
5131        nfs4_file_put_access(fp, open->op_share_access);
5132        reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap), stp);
5133        goto out;
5134}
5135
5136static __be32
5137nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp,
5138                struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
5139                struct nfsd4_open *open)
5140{
5141        __be32 status;
5142        unsigned char old_deny_bmap = stp->st_deny_bmap;
5143
5144        if (!test_access(open->op_share_access, stp))
5145                return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open, false);
5146
5147        /* test and set deny mode */
5148        spin_lock(&fp->fi_lock);
5149        status = nfs4_file_check_deny(fp, open->op_share_deny);
5150        if (status == nfs_ok) {
5151                if (status != nfserr_share_denied) {
5152                        set_deny(open->op_share_deny, stp);
5153                        fp->fi_share_deny |=
5154                                (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
5155                } else {
5156                        if (nfs4_resolve_deny_conflicts_locked(fp, false,
5157                                        stp, open->op_share_deny, false))
5158                                status = nfserr_jukebox;
5159                }
5160        }
5161        spin_unlock(&fp->fi_lock);
5162
5163        if (status != nfs_ok)
5164                return status;
5165
5166        status = nfsd4_truncate(rqstp, cur_fh, open);
5167        if (status != nfs_ok)
5168                reset_union_bmap_deny(old_deny_bmap, stp);
5169        return status;
5170}
5171
5172/* Should we give out recallable state?: */
5173static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
5174{
5175        if (clp->cl_cb_state == NFSD4_CB_UP)
5176                return true;
5177        /*
5178         * In the sessions case, since we don't have to establish a
5179         * separate connection for callbacks, we assume it's OK
5180         * until we hear otherwise:
5181         */
5182        return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
5183}
5184
5185static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp,
5186                                                int flag)
5187{
5188        struct file_lock *fl;
5189
5190        fl = locks_alloc_lock();
5191        if (!fl)
5192                return NULL;
5193        fl->fl_lmops = &nfsd_lease_mng_ops;
5194        fl->fl_flags = FL_DELEG;
5195        fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
5196        fl->fl_end = OFFSET_MAX;
5197        fl->fl_owner = (fl_owner_t)dp;
5198        fl->fl_pid = current->tgid;
5199        fl->fl_file = dp->dl_stid.sc_file->fi_deleg_file->nf_file;
5200        return fl;
5201}
5202
5203static int nfsd4_check_conflicting_opens(struct nfs4_client *clp,
5204                                         struct nfs4_file *fp)
5205{
5206        struct nfs4_ol_stateid *st;
5207        struct file *f = fp->fi_deleg_file->nf_file;
5208        struct inode *ino = locks_inode(f);
5209        int writes;
5210
5211        writes = atomic_read(&ino->i_writecount);
5212        if (!writes)
5213                return 0;
5214        /*
5215         * There could be multiple filehandles (hence multiple
5216         * nfs4_files) referencing this file, but that's not too
5217         * common; let's just give up in that case rather than
5218         * trying to go look up all the clients using that other
5219         * nfs4_file as well:
5220         */
5221        if (fp->fi_aliased)
5222                return -EAGAIN;
5223        /*
5224         * If there's a close in progress, make sure that we see it
5225         * clear any fi_fds[] entries before we see it decrement
5226         * i_writecount:
5227         */
5228        smp_mb__after_atomic();
5229
5230        if (fp->fi_fds[O_WRONLY])
5231                writes--;
5232        if (fp->fi_fds[O_RDWR])
5233                writes--;
5234        if (writes > 0)
5235                return -EAGAIN; /* There may be non-NFSv4 writers */
5236        /*
5237         * It's possible there are non-NFSv4 write opens in progress,
5238         * but if they haven't incremented i_writecount yet then they
5239         * also haven't called break lease yet; so, they'll break this
5240         * lease soon enough.  So, all that's left to check for is NFSv4
5241         * opens:
5242         */
5243        spin_lock(&fp->fi_lock);
5244        list_for_each_entry(st, &fp->fi_stateids, st_perfile) {
5245                if (st->st_openstp == NULL /* it's an open */ &&
5246                    access_permit_write(st) &&
5247                    st->st_stid.sc_client != clp) {
5248                        spin_unlock(&fp->fi_lock);
5249                        return -EAGAIN;
5250                }
5251        }
5252        spin_unlock(&fp->fi_lock);
5253        /*
5254         * There's a small chance that we could be racing with another
5255         * NFSv4 open.  However, any open that hasn't added itself to
5256         * the fi_stateids list also hasn't called break_lease yet; so,
5257         * they'll break this lease soon enough.
5258         */
5259        return 0;
5260}
5261
5262static struct nfs4_delegation *
5263nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh,
5264                    struct nfs4_file *fp, struct nfs4_clnt_odstate *odstate)
5265{
5266        int status = 0;
5267        struct nfs4_delegation *dp;
5268        struct nfsd_file *nf;
5269        struct file_lock *fl;
5270
5271        /*
5272         * The fi_had_conflict and nfs_get_existing_delegation checks
5273         * here are just optimizations; we'll need to recheck them at
5274         * the end:
5275         */
5276        if (fp->fi_had_conflict)
5277                return ERR_PTR(-EAGAIN);
5278
5279        nf = find_readable_file(fp);
5280        if (!nf) {
5281                /*
5282                 * We probably could attempt another open and get a read
5283                 * delegation, but for now, don't bother until the
5284                 * client actually sends us one.
5285                 */
5286                return ERR_PTR(-EAGAIN);
5287        }
5288        spin_lock(&state_lock);
5289        spin_lock(&fp->fi_lock);
5290        if (nfs4_delegation_exists(clp, fp))
5291                status = -EAGAIN;
5292        else if (!fp->fi_deleg_file) {
5293                fp->fi_deleg_file = nf;
5294                /* increment early to prevent fi_deleg_file from being
5295                 * cleared */
5296                fp->fi_delegees = 1;
5297                nf = NULL;
5298        } else
5299                fp->fi_delegees++;
5300        spin_unlock(&fp->fi_lock);
5301        spin_unlock(&state_lock);
5302        if (nf)
5303                nfsd_file_put(nf);
5304        if (status)
5305                return ERR_PTR(status);
5306
5307        status = -ENOMEM;
5308        dp = alloc_init_deleg(clp, fp, fh, odstate);
5309        if (!dp)
5310                goto out_delegees;
5311
5312        fl = nfs4_alloc_init_lease(dp, NFS4_OPEN_DELEGATE_READ);
5313        if (!fl)
5314                goto out_clnt_odstate;
5315
5316        status = vfs_setlease(fp->fi_deleg_file->nf_file, fl->fl_type, &fl, NULL);
5317        if (fl)
5318                locks_free_lock(fl);
5319        if (status)
5320                goto out_clnt_odstate;
5321        status = nfsd4_check_conflicting_opens(clp, fp);
5322        if (status)
5323                goto out_unlock;
5324
5325        spin_lock(&state_lock);
5326        spin_lock(&fp->fi_lock);
5327        if (fp->fi_had_conflict)
5328                status = -EAGAIN;
5329        else
5330                status = hash_delegation_locked(dp, fp);
5331        spin_unlock(&fp->fi_lock);
5332        spin_unlock(&state_lock);
5333
5334        if (status)
5335                goto out_unlock;
5336
5337        return dp;
5338out_unlock:
5339        vfs_setlease(fp->fi_deleg_file->nf_file, F_UNLCK, NULL, (void **)&dp);
5340out_clnt_odstate:
5341        put_clnt_odstate(dp->dl_clnt_odstate);
5342        nfs4_put_stid(&dp->dl_stid);
5343out_delegees:
5344        put_deleg_file(fp);
5345        return ERR_PTR(status);
5346}
5347
5348static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
5349{
5350        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5351        if (status == -EAGAIN)
5352                open->op_why_no_deleg = WND4_CONTENTION;
5353        else {
5354                open->op_why_no_deleg = WND4_RESOURCE;
5355                switch (open->op_deleg_want) {
5356                case NFS4_SHARE_WANT_READ_DELEG:
5357                case NFS4_SHARE_WANT_WRITE_DELEG:
5358                case NFS4_SHARE_WANT_ANY_DELEG:
5359                        break;
5360                case NFS4_SHARE_WANT_CANCEL:
5361                        open->op_why_no_deleg = WND4_CANCELLED;
5362                        break;
5363                case NFS4_SHARE_WANT_NO_DELEG:
5364                        WARN_ON_ONCE(1);
5365                }
5366        }
5367}
5368
5369/*
5370 * Attempt to hand out a delegation.
5371 *
5372 * Note we don't support write delegations, and won't until the vfs has
5373 * proper support for them.
5374 */
5375static void
5376nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open,
5377                        struct nfs4_ol_stateid *stp)
5378{
5379        struct nfs4_delegation *dp;
5380        struct nfs4_openowner *oo = openowner(stp->st_stateowner);
5381        struct nfs4_client *clp = stp->st_stid.sc_client;
5382        int cb_up;
5383        int status = 0;
5384
5385        cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
5386        open->op_recall = 0;
5387        switch (open->op_claim_type) {
5388                case NFS4_OPEN_CLAIM_PREVIOUS:
5389                        if (!cb_up)
5390                                open->op_recall = 1;
5391                        if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
5392                                goto out_no_deleg;
5393                        break;
5394                case NFS4_OPEN_CLAIM_NULL:
5395                case NFS4_OPEN_CLAIM_FH:
5396                        /*
5397                         * Let's not give out any delegations till everyone's
5398                         * had the chance to reclaim theirs, *and* until
5399                         * NLM locks have all been reclaimed:
5400                         */
5401                        if (locks_in_grace(clp->net))
5402                                goto out_no_deleg;
5403                        if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
5404                                goto out_no_deleg;
5405                        break;
5406                default:
5407                        goto out_no_deleg;
5408        }
5409        dp = nfs4_set_delegation(clp, fh, stp->st_stid.sc_file, stp->st_clnt_odstate);
5410        if (IS_ERR(dp))
5411                goto out_no_deleg;
5412
5413        memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
5414
5415        trace_nfsd_deleg_read(&dp->dl_stid.sc_stateid);
5416        open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
5417        nfs4_put_stid(&dp->dl_stid);
5418        return;
5419out_no_deleg:
5420        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
5421        if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
5422            open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
5423                dprintk("NFSD: WARNING: refusing delegation reclaim\n");
5424                open->op_recall = 1;
5425        }
5426
5427        /* 4.1 client asking for a delegation? */
5428        if (open->op_deleg_want)
5429                nfsd4_open_deleg_none_ext(open, status);
5430        return;
5431}
5432
5433static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
5434                                        struct nfs4_delegation *dp)
5435{
5436        if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
5437            dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
5438                open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5439                open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
5440        } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
5441                   dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
5442                open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5443                open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
5444        }
5445        /* Otherwise the client must be confused wanting a delegation
5446         * it already has, therefore we don't return
5447         * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
5448         */
5449}
5450
5451/**
5452 * nfsd4_process_open2 - finish open processing
5453 * @rqstp: the RPC transaction being executed
5454 * @current_fh: NFSv4 COMPOUND's current filehandle
5455 * @open: OPEN arguments
5456 *
5457 * If successful, (1) truncate the file if open->op_truncate was
5458 * set, (2) set open->op_stateid, (3) set open->op_delegation.
5459 *
5460 * Returns %nfs_ok on success; otherwise an nfs4stat value in
5461 * network byte order is returned.
5462 */
5463__be32
5464nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
5465{
5466        struct nfsd4_compoundres *resp = rqstp->rq_resp;
5467        struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
5468        struct nfs4_file *fp = NULL;
5469        struct nfs4_ol_stateid *stp = NULL;
5470        struct nfs4_delegation *dp = NULL;
5471        __be32 status;
5472        bool new_stp = false;
5473
5474        /*
5475         * Lookup file; if found, lookup stateid and check open request,
5476         * and check for delegations in the process of being recalled.
5477         * If not found, create the nfs4_file struct
5478         */
5479        fp = find_or_add_file(open->op_file, current_fh);
5480        if (fp != open->op_file) {
5481                status = nfs4_check_deleg(cl, open, &dp);
5482                if (status)
5483                        goto out;
5484                stp = nfsd4_find_and_lock_existing_open(fp, open);
5485        } else {
5486                open->op_file = NULL;
5487                status = nfserr_bad_stateid;
5488                if (nfsd4_is_deleg_cur(open))
5489                        goto out;
5490        }
5491
5492        if (!stp) {
5493                stp = init_open_stateid(fp, open);
5494                if (!open->op_stp)
5495                        new_stp = true;
5496        }
5497
5498        /*
5499         * OPEN the file, or upgrade an existing OPEN.
5500         * If truncate fails, the OPEN fails.
5501         *
5502         * stp is already locked.
5503         */
5504        if (!new_stp) {
5505                /* Stateid was found, this is an OPEN upgrade */
5506                status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
5507                if (status) {
5508                        mutex_unlock(&stp->st_mutex);
5509                        goto out;
5510                }
5511        } else {
5512                status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open, true);
5513                if (status) {
5514                        stp->st_stid.sc_type = NFS4_CLOSED_STID;
5515                        release_open_stateid(stp);
5516                        mutex_unlock(&stp->st_mutex);
5517                        goto out;
5518                }
5519
5520                stp->st_clnt_odstate = find_or_hash_clnt_odstate(fp,
5521                                                        open->op_odstate);
5522                if (stp->st_clnt_odstate == open->op_odstate)
5523                        open->op_odstate = NULL;
5524        }
5525
5526        nfs4_inc_and_copy_stateid(&open->op_stateid, &stp->st_stid);
5527        mutex_unlock(&stp->st_mutex);
5528
5529        if (nfsd4_has_session(&resp->cstate)) {
5530                if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
5531                        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5532                        open->op_why_no_deleg = WND4_NOT_WANTED;
5533                        goto nodeleg;
5534                }
5535        }
5536
5537        /*
5538        * Attempt to hand out a delegation. No error return, because the
5539        * OPEN succeeds even if we fail.
5540        */
5541        nfs4_open_delegation(current_fh, open, stp);
5542nodeleg:
5543        status = nfs_ok;
5544        trace_nfsd_open(&stp->st_stid.sc_stateid);
5545out:
5546        /* 4.1 client trying to upgrade/downgrade delegation? */
5547        if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
5548            open->op_deleg_want)
5549                nfsd4_deleg_xgrade_none_ext(open, dp);
5550
5551        if (fp)
5552                put_nfs4_file(fp);
5553        if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
5554                open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
5555        /*
5556        * To finish the open response, we just need to set the rflags.
5557        */
5558        open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
5559        if (nfsd4_has_session(&resp->cstate))
5560                open->op_rflags |= NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK;
5561        else if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED))
5562                open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
5563
5564        if (dp)
5565                nfs4_put_stid(&dp->dl_stid);
5566        if (stp)
5567                nfs4_put_stid(&stp->st_stid);
5568
5569        return status;
5570}
5571
5572void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
5573                              struct nfsd4_open *open)
5574{
5575        if (open->op_openowner) {
5576                struct nfs4_stateowner *so = &open->op_openowner->oo_owner;
5577
5578                nfsd4_cstate_assign_replay(cstate, so);
5579                nfs4_put_stateowner(so);
5580        }
5581        if (open->op_file)
5582                kmem_cache_free(file_slab, open->op_file);
5583        if (open->op_stp)
5584                nfs4_put_stid(&open->op_stp->st_stid);
5585        if (open->op_odstate)
5586                kmem_cache_free(odstate_slab, open->op_odstate);
5587}
5588
5589__be32
5590nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5591            union nfsd4_op_u *u)
5592{
5593        clientid_t *clid = &u->renew;
5594        struct nfs4_client *clp;
5595        __be32 status;
5596        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5597
5598        trace_nfsd_clid_renew(clid);
5599        status = set_client(clid, cstate, nn);
5600        if (status)
5601                return status;
5602        clp = cstate->clp;
5603        if (!list_empty(&clp->cl_delegations)
5604                        && clp->cl_cb_state != NFSD4_CB_UP)
5605                return nfserr_cb_path_down;
5606        return nfs_ok;
5607}
5608
5609void
5610nfsd4_end_grace(struct nfsd_net *nn)
5611{
5612        /* do nothing if grace period already ended */
5613        if (nn->grace_ended)
5614                return;
5615
5616        trace_nfsd_grace_complete(nn);
5617        nn->grace_ended = true;
5618        /*
5619         * If the server goes down again right now, an NFSv4
5620         * client will still be allowed to reclaim after it comes back up,
5621         * even if it hasn't yet had a chance to reclaim state this time.
5622         *
5623         */
5624        nfsd4_record_grace_done(nn);
5625        /*
5626         * At this point, NFSv4 clients can still reclaim.  But if the
5627         * server crashes, any that have not yet reclaimed will be out
5628         * of luck on the next boot.
5629         *
5630         * (NFSv4.1+ clients are considered to have reclaimed once they
5631         * call RECLAIM_COMPLETE.  NFSv4.0 clients are considered to
5632         * have reclaimed after their first OPEN.)
5633         */
5634        locks_end_grace(&nn->nfsd4_manager);
5635        /*
5636         * At this point, and once lockd and/or any other containers
5637         * exit their grace period, further reclaims will fail and
5638         * regular locking can resume.
5639         */
5640}
5641
5642/*
5643 * If we've waited a lease period but there are still clients trying to
5644 * reclaim, wait a little longer to give them a chance to finish.
5645 */
5646static bool clients_still_reclaiming(struct nfsd_net *nn)
5647{
5648        time64_t double_grace_period_end = nn->boot_time +
5649                                           2 * nn->nfsd4_lease;
5650
5651        if (nn->track_reclaim_completes &&
5652                        atomic_read(&nn->nr_reclaim_complete) ==
5653                        nn->reclaim_str_hashtbl_size)
5654                return false;
5655        if (!nn->somebody_reclaimed)
5656                return false;
5657        nn->somebody_reclaimed = false;
5658        /*
5659         * If we've given them *two* lease times to reclaim, and they're
5660         * still not done, give up:
5661         */
5662        if (ktime_get_boottime_seconds() > double_grace_period_end)
5663                return false;
5664        return true;
5665}
5666
5667struct laundry_time {
5668        time64_t cutoff;
5669        time64_t new_timeo;
5670};
5671
5672static bool state_expired(struct laundry_time *lt, time64_t last_refresh)
5673{
5674        time64_t time_remaining;
5675
5676        if (last_refresh < lt->cutoff)
5677                return true;
5678        time_remaining = last_refresh - lt->cutoff;
5679        lt->new_timeo = min(lt->new_timeo, time_remaining);
5680        return false;
5681}
5682
5683#ifdef CONFIG_NFSD_V4_2_INTER_SSC
5684void nfsd4_ssc_init_umount_work(struct nfsd_net *nn)
5685{
5686        spin_lock_init(&nn->nfsd_ssc_lock);
5687        INIT_LIST_HEAD(&nn->nfsd_ssc_mount_list);
5688        init_waitqueue_head(&nn->nfsd_ssc_waitq);
5689}
5690EXPORT_SYMBOL_GPL(nfsd4_ssc_init_umount_work);
5691
5692/*
5693 * This is called when nfsd is being shutdown, after all inter_ssc
5694 * cleanup were done, to destroy the ssc delayed unmount list.
5695 */
5696static void nfsd4_ssc_shutdown_umount(struct nfsd_net *nn)
5697{
5698        struct nfsd4_ssc_umount_item *ni = NULL;
5699        struct nfsd4_ssc_umount_item *tmp;
5700
5701        spin_lock(&nn->nfsd_ssc_lock);
5702        list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
5703                list_del(&ni->nsui_list);
5704                spin_unlock(&nn->nfsd_ssc_lock);
5705                mntput(ni->nsui_vfsmount);
5706                kfree(ni);
5707                spin_lock(&nn->nfsd_ssc_lock);
5708        }
5709        spin_unlock(&nn->nfsd_ssc_lock);
5710}
5711
5712static void nfsd4_ssc_expire_umount(struct nfsd_net *nn)
5713{
5714        bool do_wakeup = false;
5715        struct nfsd4_ssc_umount_item *ni = NULL;
5716        struct nfsd4_ssc_umount_item *tmp;
5717
5718        spin_lock(&nn->nfsd_ssc_lock);
5719        list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
5720                if (time_after(jiffies, ni->nsui_expire)) {
5721                        if (refcount_read(&ni->nsui_refcnt) > 1)
5722                                continue;
5723
5724                        /* mark being unmount */
5725                        ni->nsui_busy = true;
5726                        spin_unlock(&nn->nfsd_ssc_lock);
5727                        mntput(ni->nsui_vfsmount);
5728                        spin_lock(&nn->nfsd_ssc_lock);
5729
5730                        /* waiters need to start from begin of list */
5731                        list_del(&ni->nsui_list);
5732                        kfree(ni);
5733
5734                        /* wakeup ssc_connect waiters */
5735                        do_wakeup = true;
5736                        continue;
5737                }
5738                break;
5739        }
5740        if (do_wakeup)
5741                wake_up_all(&nn->nfsd_ssc_waitq);
5742        spin_unlock(&nn->nfsd_ssc_lock);
5743}
5744#endif
5745
5746/* Check if any lock belonging to this lockowner has any blockers */
5747static bool
5748nfs4_lockowner_has_blockers(struct nfs4_lockowner *lo)
5749{
5750        struct file_lock_context *ctx;
5751        struct nfs4_ol_stateid *stp;
5752        struct nfs4_file *nf;
5753
5754        list_for_each_entry(stp, &lo->lo_owner.so_stateids, st_perstateowner) {
5755                nf = stp->st_stid.sc_file;
5756                ctx = nf->fi_inode->i_flctx;
5757                if (!ctx)
5758                        continue;
5759                if (locks_owner_has_blockers(ctx, lo))
5760                        return true;
5761        }
5762        return false;
5763}
5764
5765static bool
5766nfs4_anylock_blockers(struct nfs4_client *clp)
5767{
5768        int i;
5769        struct nfs4_stateowner *so;
5770        struct nfs4_lockowner *lo;
5771
5772        if (atomic_read(&clp->cl_delegs_in_recall))
5773                return true;
5774        spin_lock(&clp->cl_lock);
5775        for (i = 0; i < OWNER_HASH_SIZE; i++) {
5776                list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[i],
5777                                so_strhash) {
5778                        if (so->so_is_open_owner)
5779                                continue;
5780                        lo = lockowner(so);
5781                        if (nfs4_lockowner_has_blockers(lo)) {
5782                                spin_unlock(&clp->cl_lock);
5783                                return true;
5784                        }
5785                }
5786        }
5787        spin_unlock(&clp->cl_lock);
5788        return false;
5789}
5790
5791static void
5792nfs4_get_client_reaplist(struct nfsd_net *nn, struct list_head *reaplist,
5793                                struct laundry_time *lt)
5794{
5795        struct list_head *pos, *next;
5796        struct nfs4_client *clp;
5797
5798        INIT_LIST_HEAD(reaplist);
5799        spin_lock(&nn->client_lock);
5800        list_for_each_safe(pos, next, &nn->client_lru) {
5801                clp = list_entry(pos, struct nfs4_client, cl_lru);
5802                if (clp->cl_state == NFSD4_EXPIRABLE)
5803                        goto exp_client;
5804                if (!state_expired(lt, clp->cl_time))
5805                        break;
5806                if (!atomic_read(&clp->cl_rpc_users))
5807                        clp->cl_state = NFSD4_COURTESY;
5808                if (!client_has_state(clp) ||
5809                                ktime_get_boottime_seconds() >=
5810                                (clp->cl_time + NFSD_COURTESY_CLIENT_TIMEOUT))
5811                        goto exp_client;
5812                if (nfs4_anylock_blockers(clp)) {
5813exp_client:
5814                        if (!mark_client_expired_locked(clp))
5815                                list_add(&clp->cl_lru, reaplist);
5816                }
5817        }
5818        spin_unlock(&nn->client_lock);
5819}
5820
5821static time64_t
5822nfs4_laundromat(struct nfsd_net *nn)
5823{
5824        struct nfs4_client *clp;
5825        struct nfs4_openowner *oo;
5826        struct nfs4_delegation *dp;
5827        struct nfs4_ol_stateid *stp;
5828        struct nfsd4_blocked_lock *nbl;
5829        struct list_head *pos, *next, reaplist;
5830        struct laundry_time lt = {
5831                .cutoff = ktime_get_boottime_seconds() - nn->nfsd4_lease,
5832                .new_timeo = nn->nfsd4_lease
5833        };
5834        struct nfs4_cpntf_state *cps;
5835        copy_stateid_t *cps_t;
5836        int i;
5837
5838        if (clients_still_reclaiming(nn)) {
5839                lt.new_timeo = 0;
5840                goto out;
5841        }
5842        nfsd4_end_grace(nn);
5843
5844        spin_lock(&nn->s2s_cp_lock);
5845        idr_for_each_entry(&nn->s2s_cp_stateids, cps_t, i) {
5846                cps = container_of(cps_t, struct nfs4_cpntf_state, cp_stateid);
5847                if (cps->cp_stateid.sc_type == NFS4_COPYNOTIFY_STID &&
5848                                state_expired(&lt, cps->cpntf_time))
5849                        _free_cpntf_state_locked(nn, cps);
5850        }
5851        spin_unlock(&nn->s2s_cp_lock);
5852        nfs4_get_client_reaplist(nn, &reaplist, &lt);
5853        list_for_each_safe(pos, next, &reaplist) {
5854                clp = list_entry(pos, struct nfs4_client, cl_lru);
5855                trace_nfsd_clid_purged(&clp->cl_clientid);
5856                list_del_init(&clp->cl_lru);
5857                expire_client(clp);
5858        }
5859        spin_lock(&state_lock);
5860        list_for_each_safe(pos, next, &nn->del_recall_lru) {
5861                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5862                if (!state_expired(&lt, dp->dl_time))
5863                        break;
5864                WARN_ON(!unhash_delegation_locked(dp));
5865                list_add(&dp->dl_recall_lru, &reaplist);
5866        }
5867        spin_unlock(&state_lock);
5868        while (!list_empty(&reaplist)) {
5869                dp = list_first_entry(&reaplist, struct nfs4_delegation,
5870                                        dl_recall_lru);
5871                list_del_init(&dp->dl_recall_lru);
5872                revoke_delegation(dp);
5873        }
5874
5875        spin_lock(&nn->client_lock);
5876        while (!list_empty(&nn->close_lru)) {
5877                oo = list_first_entry(&nn->close_lru, struct nfs4_openowner,
5878                                        oo_close_lru);
5879                if (!state_expired(&lt, oo->oo_time))
5880                        break;
5881                list_del_init(&oo->oo_close_lru);
5882                stp = oo->oo_last_closed_stid;
5883                oo->oo_last_closed_stid = NULL;
5884                spin_unlock(&nn->client_lock);
5885                nfs4_put_stid(&stp->st_stid);
5886                spin_lock(&nn->client_lock);
5887        }
5888        spin_unlock(&nn->client_lock);
5889
5890        /*
5891         * It's possible for a client to try and acquire an already held lock
5892         * that is being held for a long time, and then lose interest in it.
5893         * So, we clean out any un-revisited request after a lease period
5894         * under the assumption that the client is no longer interested.
5895         *
5896         * RFC5661, sec. 9.6 states that the client must not rely on getting
5897         * notifications and must continue to poll for locks, even when the
5898         * server supports them. Thus this shouldn't lead to clients blocking
5899         * indefinitely once the lock does become free.
5900         */
5901        BUG_ON(!list_empty(&reaplist));
5902        spin_lock(&nn->blocked_locks_lock);
5903        while (!list_empty(&nn->blocked_locks_lru)) {
5904                nbl = list_first_entry(&nn->blocked_locks_lru,
5905                                        struct nfsd4_blocked_lock, nbl_lru);
5906                if (!state_expired(&lt, nbl->nbl_time))
5907                        break;
5908                list_move(&nbl->nbl_lru, &reaplist);
5909                list_del_init(&nbl->nbl_list);
5910        }
5911        spin_unlock(&nn->blocked_locks_lock);
5912
5913        while (!list_empty(&reaplist)) {
5914                nbl = list_first_entry(&reaplist,
5915                                        struct nfsd4_blocked_lock, nbl_lru);
5916                list_del_init(&nbl->nbl_lru);
5917                free_blocked_lock(nbl);
5918        }
5919#ifdef CONFIG_NFSD_V4_2_INTER_SSC
5920        /* service the server-to-server copy delayed unmount list */
5921        nfsd4_ssc_expire_umount(nn);
5922#endif
5923out:
5924        return max_t(time64_t, lt.new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
5925}
5926
5927static void laundromat_main(struct work_struct *);
5928
5929static void
5930laundromat_main(struct work_struct *laundry)
5931{
5932        time64_t t;
5933        struct delayed_work *dwork = to_delayed_work(laundry);
5934        struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
5935                                           laundromat_work);
5936
5937        t = nfs4_laundromat(nn);
5938        queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
5939}
5940
5941static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stid *stp)
5942{
5943        if (!fh_match(&fhp->fh_handle, &stp->sc_file->fi_fhandle))
5944                return nfserr_bad_stateid;
5945        return nfs_ok;
5946}
5947
5948static
5949__be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
5950{
5951        __be32 status = nfserr_openmode;
5952
5953        /* For lock stateid's, we test the parent open, not the lock: */
5954        if (stp->st_openstp)
5955                stp = stp->st_openstp;
5956        if ((flags & WR_STATE) && !access_permit_write(stp))
5957                goto out;
5958        if ((flags & RD_STATE) && !access_permit_read(stp))
5959                goto out;
5960        status = nfs_ok;
5961out:
5962        return status;
5963}
5964
5965static inline __be32
5966check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
5967{
5968        if (ONE_STATEID(stateid) && (flags & RD_STATE))
5969                return nfs_ok;
5970        else if (opens_in_grace(net)) {
5971                /* Answer in remaining cases depends on existence of
5972                 * conflicting state; so we must wait out the grace period. */
5973                return nfserr_grace;
5974        } else if (flags & WR_STATE)
5975                return nfs4_share_conflict(current_fh,
5976                                NFS4_SHARE_DENY_WRITE);
5977        else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
5978                return nfs4_share_conflict(current_fh,
5979                                NFS4_SHARE_DENY_READ);
5980}
5981
5982static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
5983{
5984        /*
5985         * When sessions are used the stateid generation number is ignored
5986         * when it is zero.
5987         */
5988        if (has_session && in->si_generation == 0)
5989                return nfs_ok;
5990
5991        if (in->si_generation == ref->si_generation)
5992                return nfs_ok;
5993
5994        /* If the client sends us a stateid from the future, it's buggy: */
5995        if (nfsd4_stateid_generation_after(in, ref))
5996                return nfserr_bad_stateid;
5997        /*
5998         * However, we could see a stateid from the past, even from a
5999         * non-buggy client.  For example, if the client sends a lock
6000         * while some IO is outstanding, the lock may bump si_generation
6001         * while the IO is still in flight.  The client could avoid that
6002         * situation by waiting for responses on all the IO requests,
6003         * but better performance may result in retrying IO that
6004         * receives an old_stateid error if requests are rarely
6005         * reordered in flight:
6006         */
6007        return nfserr_old_stateid;
6008}
6009
6010static __be32 nfsd4_stid_check_stateid_generation(stateid_t *in, struct nfs4_stid *s, bool has_session)
6011{
6012        __be32 ret;
6013
6014        spin_lock(&s->sc_lock);
6015        ret = nfsd4_verify_open_stid(s);
6016        if (ret == nfs_ok)
6017                ret = check_stateid_generation(in, &s->sc_stateid, has_session);
6018        spin_unlock(&s->sc_lock);
6019        return ret;
6020}
6021
6022static __be32 nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid *ols)
6023{
6024        if (ols->st_stateowner->so_is_open_owner &&
6025            !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
6026                return nfserr_bad_stateid;
6027        return nfs_ok;
6028}
6029
6030static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
6031{
6032        struct nfs4_stid *s;
6033        __be32 status = nfserr_bad_stateid;
6034
6035        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
6036                CLOSE_STATEID(stateid))
6037                return status;
6038        if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid))
6039                return status;
6040        spin_lock(&cl->cl_lock);
6041        s = find_stateid_locked(cl, stateid);
6042        if (!s)
6043                goto out_unlock;
6044        status = nfsd4_stid_check_stateid_generation(stateid, s, 1);
6045        if (status)
6046                goto out_unlock;
6047        switch (s->sc_type) {
6048        case NFS4_DELEG_STID:
6049                status = nfs_ok;
6050                break;
6051        case NFS4_REVOKED_DELEG_STID:
6052                status = nfserr_deleg_revoked;
6053                break;
6054        case NFS4_OPEN_STID:
6055        case NFS4_LOCK_STID:
6056                status = nfsd4_check_openowner_confirmed(openlockstateid(s));
6057                break;
6058        default:
6059                printk("unknown stateid type %x\n", s->sc_type);
6060                fallthrough;
6061        case NFS4_CLOSED_STID:
6062        case NFS4_CLOSED_DELEG_STID:
6063                status = nfserr_bad_stateid;
6064        }
6065out_unlock:
6066        spin_unlock(&cl->cl_lock);
6067        return status;
6068}
6069
6070__be32
6071nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
6072                     stateid_t *stateid, unsigned char typemask,
6073                     struct nfs4_stid **s, struct nfsd_net *nn)
6074{
6075        __be32 status;
6076        bool return_revoked = false;
6077
6078        /*
6079         *  only return revoked delegations if explicitly asked.
6080         *  otherwise we report revoked or bad_stateid status.
6081         */
6082        if (typemask & NFS4_REVOKED_DELEG_STID)
6083                return_revoked = true;
6084        else if (typemask & NFS4_DELEG_STID)
6085                typemask |= NFS4_REVOKED_DELEG_STID;
6086
6087        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
6088                CLOSE_STATEID(stateid))
6089                return nfserr_bad_stateid;
6090        status = set_client(&stateid->si_opaque.so_clid, cstate, nn);
6091        if (status == nfserr_stale_clientid) {
6092                if (cstate->session)
6093                        return nfserr_bad_stateid;
6094                return nfserr_stale_stateid;
6095        }
6096        if (status)
6097                return status;
6098        *s = find_stateid_by_type(cstate->clp, stateid, typemask);
6099        if (!*s)
6100                return nfserr_bad_stateid;
6101        if (((*s)->sc_type == NFS4_REVOKED_DELEG_STID) && !return_revoked) {
6102                nfs4_put_stid(*s);
6103                if (cstate->minorversion)
6104                        return nfserr_deleg_revoked;
6105                return nfserr_bad_stateid;
6106        }
6107        return nfs_ok;
6108}
6109
6110static struct nfsd_file *
6111nfs4_find_file(struct nfs4_stid *s, int flags)
6112{
6113        if (!s)
6114                return NULL;
6115
6116        switch (s->sc_type) {
6117        case NFS4_DELEG_STID:
6118                if (WARN_ON_ONCE(!s->sc_file->fi_deleg_file))
6119                        return NULL;
6120                return nfsd_file_get(s->sc_file->fi_deleg_file);
6121        case NFS4_OPEN_STID:
6122        case NFS4_LOCK_STID:
6123                if (flags & RD_STATE)
6124                        return find_readable_file(s->sc_file);
6125                else
6126                        return find_writeable_file(s->sc_file);
6127        }
6128
6129        return NULL;
6130}
6131
6132static __be32
6133nfs4_check_olstateid(struct nfs4_ol_stateid *ols, int flags)
6134{
6135        __be32 status;
6136
6137        status = nfsd4_check_openowner_confirmed(ols);
6138        if (status)
6139                return status;
6140        return nfs4_check_openmode(ols, flags);
6141}
6142
6143static __be32
6144nfs4_check_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct nfs4_stid *s,
6145                struct nfsd_file **nfp, int flags)
6146{
6147        int acc = (flags & RD_STATE) ? NFSD_MAY_READ : NFSD_MAY_WRITE;
6148        struct nfsd_file *nf;
6149        __be32 status;
6150
6151        nf = nfs4_find_file(s, flags);
6152        if (nf) {
6153                status = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
6154                                acc | NFSD_MAY_OWNER_OVERRIDE);
6155                if (status) {
6156                        nfsd_file_put(nf);
6157                        goto out;
6158                }
6159        } else {
6160                status = nfsd_file_acquire(rqstp, fhp, acc, &nf);
6161                if (status)
6162                        return status;
6163        }
6164        *nfp = nf;
6165out:
6166        return status;
6167}
6168static void
6169_free_cpntf_state_locked(struct nfsd_net *nn, struct nfs4_cpntf_state *cps)
6170{
6171        WARN_ON_ONCE(cps->cp_stateid.sc_type != NFS4_COPYNOTIFY_STID);
6172        if (!refcount_dec_and_test(&cps->cp_stateid.sc_count))
6173                return;
6174        list_del(&cps->cp_list);
6175        idr_remove(&nn->s2s_cp_stateids,
6176                   cps->cp_stateid.stid.si_opaque.so_id);
6177        kfree(cps);
6178}
6179/*
6180 * A READ from an inter server to server COPY will have a
6181 * copy stateid. Look up the copy notify stateid from the
6182 * idr structure and take a reference on it.
6183 */
6184__be32 manage_cpntf_state(struct nfsd_net *nn, stateid_t *st,
6185                          struct nfs4_client *clp,
6186                          struct nfs4_cpntf_state **cps)
6187{
6188        copy_stateid_t *cps_t;
6189        struct nfs4_cpntf_state *state = NULL;
6190
6191        if (st->si_opaque.so_clid.cl_id != nn->s2s_cp_cl_id)
6192                return nfserr_bad_stateid;
6193        spin_lock(&nn->s2s_cp_lock);
6194        cps_t = idr_find(&nn->s2s_cp_stateids, st->si_opaque.so_id);
6195        if (cps_t) {
6196                state = container_of(cps_t, struct nfs4_cpntf_state,
6197                                     cp_stateid);
6198                if (state->cp_stateid.sc_type != NFS4_COPYNOTIFY_STID) {
6199                        state = NULL;
6200                        goto unlock;
6201                }
6202                if (!clp)
6203                        refcount_inc(&state->cp_stateid.sc_count);
6204                else
6205                        _free_cpntf_state_locked(nn, state);
6206        }
6207unlock:
6208        spin_unlock(&nn->s2s_cp_lock);
6209        if (!state)
6210                return nfserr_bad_stateid;
6211        if (!clp && state)
6212                *cps = state;
6213        return 0;
6214}
6215
6216static __be32 find_cpntf_state(struct nfsd_net *nn, stateid_t *st,
6217                               struct nfs4_stid **stid)
6218{
6219        __be32 status;
6220        struct nfs4_cpntf_state *cps = NULL;
6221        struct nfs4_client *found;
6222
6223        status = manage_cpntf_state(nn, st, NULL, &cps);
6224        if (status)
6225                return status;
6226
6227        cps->cpntf_time = ktime_get_boottime_seconds();
6228
6229        status = nfserr_expired;
6230        found = lookup_clientid(&cps->cp_p_clid, true, nn);
6231        if (!found)
6232                goto out;
6233
6234        *stid = find_stateid_by_type(found, &cps->cp_p_stateid,
6235                        NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID);
6236        if (*stid)
6237                status = nfs_ok;
6238        else
6239                status = nfserr_bad_stateid;
6240
6241        put_client_renew(found);
6242out:
6243        nfs4_put_cpntf_state(nn, cps);
6244        return status;
6245}
6246
6247void nfs4_put_cpntf_state(struct nfsd_net *nn, struct nfs4_cpntf_state *cps)
6248{
6249        spin_lock(&nn->s2s_cp_lock);
6250        _free_cpntf_state_locked(nn, cps);
6251        spin_unlock(&nn->s2s_cp_lock);
6252}
6253
6254/*
6255 * Checks for stateid operations
6256 */
6257__be32
6258nfs4_preprocess_stateid_op(struct svc_rqst *rqstp,
6259                struct nfsd4_compound_state *cstate, struct svc_fh *fhp,
6260                stateid_t *stateid, int flags, struct nfsd_file **nfp,
6261                struct nfs4_stid **cstid)
6262{
6263        struct net *net = SVC_NET(rqstp);
6264        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6265        struct nfs4_stid *s = NULL;
6266        __be32 status;
6267
6268        if (nfp)
6269                *nfp = NULL;
6270
6271        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
6272                if (cstid)
6273                        status = nfserr_bad_stateid;
6274                else
6275                        status = check_special_stateids(net, fhp, stateid,
6276                                                                        flags);
6277                goto done;
6278        }
6279
6280        status = nfsd4_lookup_stateid(cstate, stateid,
6281                                NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
6282                                &s, nn);
6283        if (status == nfserr_bad_stateid)
6284                status = find_cpntf_state(nn, stateid, &s);
6285        if (status)
6286                return status;
6287        status = nfsd4_stid_check_stateid_generation(stateid, s,
6288                        nfsd4_has_session(cstate));
6289        if (status)
6290                goto out;
6291
6292        switch (s->sc_type) {
6293        case NFS4_DELEG_STID:
6294                status = nfs4_check_delegmode(delegstateid(s), flags);
6295                break;
6296        case NFS4_OPEN_STID:
6297        case NFS4_LOCK_STID:
6298                status = nfs4_check_olstateid(openlockstateid(s), flags);
6299                break;
6300        default:
6301                status = nfserr_bad_stateid;
6302                break;
6303        }
6304        if (status)
6305                goto out;
6306        status = nfs4_check_fh(fhp, s);
6307
6308done:
6309        if (status == nfs_ok && nfp)
6310                status = nfs4_check_file(rqstp, fhp, s, nfp, flags);
6311out:
6312        if (s) {
6313                if (!status && cstid)
6314                        *cstid = s;
6315                else
6316                        nfs4_put_stid(s);
6317        }
6318        return status;
6319}
6320
6321/*
6322 * Test if the stateid is valid
6323 */
6324__be32
6325nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6326                   union nfsd4_op_u *u)
6327{
6328        struct nfsd4_test_stateid *test_stateid = &u->test_stateid;
6329        struct nfsd4_test_stateid_id *stateid;
6330        struct nfs4_client *cl = cstate->clp;
6331
6332        list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
6333                stateid->ts_id_status =
6334                        nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
6335
6336        return nfs_ok;
6337}
6338
6339static __be32
6340nfsd4_free_lock_stateid(stateid_t *stateid, struct nfs4_stid *s)
6341{
6342        struct nfs4_ol_stateid *stp = openlockstateid(s);
6343        __be32 ret;
6344
6345        ret = nfsd4_lock_ol_stateid(stp);
6346        if (ret)
6347                goto out_put_stid;
6348
6349        ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
6350        if (ret)
6351                goto out;
6352
6353        ret = nfserr_locks_held;
6354        if (check_for_locks(stp->st_stid.sc_file,
6355                            lockowner(stp->st_stateowner)))
6356                goto out;
6357
6358        release_lock_stateid(stp);
6359        ret = nfs_ok;
6360
6361out:
6362        mutex_unlock(&stp->st_mutex);
6363out_put_stid:
6364        nfs4_put_stid(s);
6365        return ret;
6366}
6367
6368__be32
6369nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6370                   union nfsd4_op_u *u)
6371{
6372        struct nfsd4_free_stateid *free_stateid = &u->free_stateid;
6373        stateid_t *stateid = &free_stateid->fr_stateid;
6374        struct nfs4_stid *s;
6375        struct nfs4_delegation *dp;
6376        struct nfs4_client *cl = cstate->clp;
6377        __be32 ret = nfserr_bad_stateid;
6378
6379        spin_lock(&cl->cl_lock);
6380        s = find_stateid_locked(cl, stateid);
6381        if (!s)
6382                goto out_unlock;
6383        spin_lock(&s->sc_lock);
6384        switch (s->sc_type) {
6385        case NFS4_DELEG_STID:
6386                ret = nfserr_locks_held;
6387                break;
6388        case NFS4_OPEN_STID:
6389                ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
6390                if (ret)
6391                        break;
6392                ret = nfserr_locks_held;
6393                break;
6394        case NFS4_LOCK_STID:
6395                spin_unlock(&s->sc_lock);
6396                refcount_inc(&s->sc_count);
6397                spin_unlock(&cl->cl_lock);
6398                ret = nfsd4_free_lock_stateid(stateid, s);
6399                goto out;
6400        case NFS4_REVOKED_DELEG_STID:
6401                spin_unlock(&s->sc_lock);
6402                dp = delegstateid(s);
6403                list_del_init(&dp->dl_recall_lru);
6404                spin_unlock(&cl->cl_lock);
6405                nfs4_put_stid(s);
6406                ret = nfs_ok;
6407                goto out;
6408        /* Default falls through and returns nfserr_bad_stateid */
6409        }
6410        spin_unlock(&s->sc_lock);
6411out_unlock:
6412        spin_unlock(&cl->cl_lock);
6413out:
6414        return ret;
6415}
6416
6417static inline int
6418setlkflg (int type)
6419{
6420        return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
6421                RD_STATE : WR_STATE;
6422}
6423
6424static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
6425{
6426        struct svc_fh *current_fh = &cstate->current_fh;
6427        struct nfs4_stateowner *sop = stp->st_stateowner;
6428        __be32 status;
6429
6430        status = nfsd4_check_seqid(cstate, sop, seqid);
6431        if (status)
6432                return status;
6433        status = nfsd4_lock_ol_stateid(stp);
6434        if (status != nfs_ok)
6435                return status;
6436        status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
6437        if (status == nfs_ok)
6438                status = nfs4_check_fh(current_fh, &stp->st_stid);
6439        if (status != nfs_ok)
6440                mutex_unlock(&stp->st_mutex);
6441        return status;
6442}
6443
6444/* 
6445 * Checks for sequence id mutating operations. 
6446 */
6447static __be32
6448nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
6449                         stateid_t *stateid, char typemask,
6450                         struct nfs4_ol_stateid **stpp,
6451                         struct nfsd_net *nn)
6452{
6453        __be32 status;
6454        struct nfs4_stid *s;
6455        struct nfs4_ol_stateid *stp = NULL;
6456
6457        trace_nfsd_preprocess(seqid, stateid);
6458
6459        *stpp = NULL;
6460        status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
6461        if (status)
6462                return status;
6463        stp = openlockstateid(s);
6464        nfsd4_cstate_assign_replay(cstate, stp->st_stateowner);
6465
6466        status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
6467        if (!status)
6468                *stpp = stp;
6469        else
6470                nfs4_put_stid(&stp->st_stid);
6471        return status;
6472}
6473
6474static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
6475                                                 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
6476{
6477        __be32 status;
6478        struct nfs4_openowner *oo;
6479        struct nfs4_ol_stateid *stp;
6480
6481        status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
6482                                                NFS4_OPEN_STID, &stp, nn);
6483        if (status)
6484                return status;
6485        oo = openowner(stp->st_stateowner);
6486        if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
6487                mutex_unlock(&stp->st_mutex);
6488                nfs4_put_stid(&stp->st_stid);
6489                return nfserr_bad_stateid;
6490        }
6491        *stpp = stp;
6492        return nfs_ok;
6493}
6494
6495__be32
6496nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6497                   union nfsd4_op_u *u)
6498{
6499        struct nfsd4_open_confirm *oc = &u->open_confirm;
6500        __be32 status;
6501        struct nfs4_openowner *oo;
6502        struct nfs4_ol_stateid *stp;
6503        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6504
6505        dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
6506                        cstate->current_fh.fh_dentry);
6507
6508        status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
6509        if (status)
6510                return status;
6511
6512        status = nfs4_preprocess_seqid_op(cstate,
6513                                        oc->oc_seqid, &oc->oc_req_stateid,
6514                                        NFS4_OPEN_STID, &stp, nn);
6515        if (status)
6516                goto out;
6517        oo = openowner(stp->st_stateowner);
6518        status = nfserr_bad_stateid;
6519        if (oo->oo_flags & NFS4_OO_CONFIRMED) {
6520                mutex_unlock(&stp->st_mutex);
6521                goto put_stateid;
6522        }
6523        oo->oo_flags |= NFS4_OO_CONFIRMED;
6524        nfs4_inc_and_copy_stateid(&oc->oc_resp_stateid, &stp->st_stid);
6525        mutex_unlock(&stp->st_mutex);
6526        trace_nfsd_open_confirm(oc->oc_seqid, &stp->st_stid.sc_stateid);
6527        nfsd4_client_record_create(oo->oo_owner.so_client);
6528        status = nfs_ok;
6529put_stateid:
6530        nfs4_put_stid(&stp->st_stid);
6531out:
6532        nfsd4_bump_seqid(cstate, status);
6533        return status;
6534}
6535
6536static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
6537{
6538        if (!test_access(access, stp))
6539                return;
6540        nfs4_file_put_access(stp->st_stid.sc_file, access);
6541        clear_access(access, stp);
6542}
6543
6544static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
6545{
6546        switch (to_access) {
6547        case NFS4_SHARE_ACCESS_READ:
6548                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
6549                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
6550                break;
6551        case NFS4_SHARE_ACCESS_WRITE:
6552                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
6553                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
6554                break;
6555        case NFS4_SHARE_ACCESS_BOTH:
6556                break;
6557        default:
6558                WARN_ON_ONCE(1);
6559        }
6560}
6561
6562__be32
6563nfsd4_open_downgrade(struct svc_rqst *rqstp,
6564                     struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
6565{
6566        struct nfsd4_open_downgrade *od = &u->open_downgrade;
6567        __be32 status;
6568        struct nfs4_ol_stateid *stp;
6569        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6570
6571        dprintk("NFSD: nfsd4_open_downgrade on file %pd\n", 
6572                        cstate->current_fh.fh_dentry);
6573
6574        /* We don't yet support WANT bits: */
6575        if (od->od_deleg_want)
6576                dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
6577                        od->od_deleg_want);
6578
6579        status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
6580                                        &od->od_stateid, &stp, nn);
6581        if (status)
6582                goto out; 
6583        status = nfserr_inval;
6584        if (!test_access(od->od_share_access, stp)) {
6585                dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
6586                        stp->st_access_bmap, od->od_share_access);
6587                goto put_stateid;
6588        }
6589        if (!test_deny(od->od_share_deny, stp)) {
6590                dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
6591                        stp->st_deny_bmap, od->od_share_deny);
6592                goto put_stateid;
6593        }
6594        nfs4_stateid_downgrade(stp, od->od_share_access);
6595        reset_union_bmap_deny(od->od_share_deny, stp);
6596        nfs4_inc_and_copy_stateid(&od->od_stateid, &stp->st_stid);
6597        status = nfs_ok;
6598put_stateid:
6599        mutex_unlock(&stp->st_mutex);
6600        nfs4_put_stid(&stp->st_stid);
6601out:
6602        nfsd4_bump_seqid(cstate, status);
6603        return status;
6604}
6605
6606static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
6607{
6608        struct nfs4_client *clp = s->st_stid.sc_client;
6609        bool unhashed;
6610        LIST_HEAD(reaplist);
6611
6612        spin_lock(&clp->cl_lock);
6613        unhashed = unhash_open_stateid(s, &reaplist);
6614
6615        if (clp->cl_minorversion) {
6616                if (unhashed)
6617                        put_ol_stateid_locked(s, &reaplist);
6618                spin_unlock(&clp->cl_lock);
6619                free_ol_stateid_reaplist(&reaplist);
6620        } else {
6621                spin_unlock(&clp->cl_lock);
6622                free_ol_stateid_reaplist(&reaplist);
6623                if (unhashed)
6624                        move_to_close_lru(s, clp->net);
6625        }
6626}
6627
6628/*
6629 * nfs4_unlock_state() called after encode
6630 */
6631__be32
6632nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6633                union nfsd4_op_u *u)
6634{
6635        struct nfsd4_close *close = &u->close;
6636        __be32 status;
6637        struct nfs4_ol_stateid *stp;
6638        struct net *net = SVC_NET(rqstp);
6639        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6640
6641        dprintk("NFSD: nfsd4_close on file %pd\n", 
6642                        cstate->current_fh.fh_dentry);
6643
6644        status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
6645                                        &close->cl_stateid,
6646                                        NFS4_OPEN_STID|NFS4_CLOSED_STID,
6647                                        &stp, nn);
6648        nfsd4_bump_seqid(cstate, status);
6649        if (status)
6650                goto out; 
6651
6652        stp->st_stid.sc_type = NFS4_CLOSED_STID;
6653
6654        /*
6655         * Technically we don't _really_ have to increment or copy it, since
6656         * it should just be gone after this operation and we clobber the
6657         * copied value below, but we continue to do so here just to ensure
6658         * that racing ops see that there was a state change.
6659         */
6660        nfs4_inc_and_copy_stateid(&close->cl_stateid, &stp->st_stid);
6661
6662        nfsd4_close_open_stateid(stp);
6663        mutex_unlock(&stp->st_mutex);
6664
6665        /* v4.1+ suggests that we send a special stateid in here, since the
6666         * clients should just ignore this anyway. Since this is not useful
6667         * for v4.0 clients either, we set it to the special close_stateid
6668         * universally.
6669         *
6670         * See RFC5661 section 18.2.4, and RFC7530 section 16.2.5
6671         */
6672        memcpy(&close->cl_stateid, &close_stateid, sizeof(close->cl_stateid));
6673
6674        /* put reference from nfs4_preprocess_seqid_op */
6675        nfs4_put_stid(&stp->st_stid);
6676out:
6677        return status;
6678}
6679
6680__be32
6681nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6682                  union nfsd4_op_u *u)
6683{
6684        struct nfsd4_delegreturn *dr = &u->delegreturn;
6685        struct nfs4_delegation *dp;
6686        stateid_t *stateid = &dr->dr_stateid;
6687        struct nfs4_stid *s;
6688        __be32 status;
6689        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6690
6691        if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
6692                return status;
6693
6694        status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
6695        if (status)
6696                goto out;
6697        dp = delegstateid(s);
6698        status = nfsd4_stid_check_stateid_generation(stateid, &dp->dl_stid, nfsd4_has_session(cstate));
6699        if (status)
6700                goto put_stateid;
6701
6702        destroy_delegation(dp);
6703put_stateid:
6704        nfs4_put_stid(&dp->dl_stid);
6705out:
6706        return status;
6707}
6708
6709/* last octet in a range */
6710static inline u64
6711last_byte_offset(u64 start, u64 len)
6712{
6713        u64 end;
6714
6715        WARN_ON_ONCE(!len);
6716        end = start + len;
6717        return end > start ? end - 1: NFS4_MAX_UINT64;
6718}
6719
6720/*
6721 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
6722 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
6723 * byte, because of sign extension problems.  Since NFSv4 calls for 64-bit
6724 * locking, this prevents us from being completely protocol-compliant.  The
6725 * real solution to this problem is to start using unsigned file offsets in
6726 * the VFS, but this is a very deep change!
6727 */
6728static inline void
6729nfs4_transform_lock_offset(struct file_lock *lock)
6730{
6731        if (lock->fl_start < 0)
6732                lock->fl_start = OFFSET_MAX;
6733        if (lock->fl_end < 0)
6734                lock->fl_end = OFFSET_MAX;
6735}
6736
6737static fl_owner_t
6738nfsd4_lm_get_owner(fl_owner_t owner)
6739{
6740        struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
6741
6742        nfs4_get_stateowner(&lo->lo_owner);
6743        return owner;
6744}
6745
6746static void
6747nfsd4_lm_put_owner(fl_owner_t owner)
6748{
6749        struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
6750
6751        if (lo)
6752                nfs4_put_stateowner(&lo->lo_owner);
6753}
6754
6755/* return pointer to struct nfs4_client if client is expirable */
6756static bool
6757nfsd4_lm_lock_expirable(struct file_lock *cfl)
6758{
6759        struct nfs4_lockowner *lo = (struct nfs4_lockowner *)cfl->fl_owner;
6760        struct nfs4_client *clp = lo->lo_owner.so_client;
6761        struct nfsd_net *nn;
6762
6763        if (try_to_expire_client(clp)) {
6764                nn = net_generic(clp->net, nfsd_net_id);
6765                mod_delayed_work(laundry_wq, &nn->laundromat_work, 0);
6766                return true;
6767        }
6768        return false;
6769}
6770
6771/* schedule laundromat to run immediately and wait for it to complete */
6772static void
6773nfsd4_lm_expire_lock(void)
6774{
6775        flush_workqueue(laundry_wq);
6776}
6777
6778static void
6779nfsd4_lm_notify(struct file_lock *fl)
6780{
6781        struct nfs4_lockowner           *lo = (struct nfs4_lockowner *)fl->fl_owner;
6782        struct net                      *net = lo->lo_owner.so_client->net;
6783        struct nfsd_net                 *nn = net_generic(net, nfsd_net_id);
6784        struct nfsd4_blocked_lock       *nbl = container_of(fl,
6785                                                struct nfsd4_blocked_lock, nbl_lock);
6786        bool queue = false;
6787
6788        /* An empty list means that something else is going to be using it */
6789        spin_lock(&nn->blocked_locks_lock);
6790        if (!list_empty(&nbl->nbl_list)) {
6791                list_del_init(&nbl->nbl_list);
6792                list_del_init(&nbl->nbl_lru);
6793                queue = true;
6794        }
6795        spin_unlock(&nn->blocked_locks_lock);
6796
6797        if (queue) {
6798                trace_nfsd_cb_notify_lock(lo, nbl);
6799                nfsd4_run_cb(&nbl->nbl_cb);
6800        }
6801}
6802
6803static const struct lock_manager_operations nfsd_posix_mng_ops  = {
6804        .lm_mod_owner = THIS_MODULE,
6805        .lm_notify = nfsd4_lm_notify,
6806        .lm_get_owner = nfsd4_lm_get_owner,
6807        .lm_put_owner = nfsd4_lm_put_owner,
6808        .lm_lock_expirable = nfsd4_lm_lock_expirable,
6809        .lm_expire_lock = nfsd4_lm_expire_lock,
6810};
6811
6812static inline void
6813nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
6814{
6815        struct nfs4_lockowner *lo;
6816
6817        if (fl->fl_lmops == &nfsd_posix_mng_ops) {
6818                lo = (struct nfs4_lockowner *) fl->fl_owner;
6819                xdr_netobj_dup(&deny->ld_owner, &lo->lo_owner.so_owner,
6820                                                GFP_KERNEL);
6821                if (!deny->ld_owner.data)
6822                        /* We just don't care that much */
6823                        goto nevermind;
6824                deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
6825        } else {
6826nevermind:
6827                deny->ld_owner.len = 0;
6828                deny->ld_owner.data = NULL;
6829                deny->ld_clientid.cl_boot = 0;
6830                deny->ld_clientid.cl_id = 0;
6831        }
6832        deny->ld_start = fl->fl_start;
6833        deny->ld_length = NFS4_MAX_UINT64;
6834        if (fl->fl_end != NFS4_MAX_UINT64)
6835                deny->ld_length = fl->fl_end - fl->fl_start + 1;        
6836        deny->ld_type = NFS4_READ_LT;
6837        if (fl->fl_type != F_RDLCK)
6838                deny->ld_type = NFS4_WRITE_LT;
6839}
6840
6841static struct nfs4_lockowner *
6842find_lockowner_str_locked(struct nfs4_client *clp, struct xdr_netobj *owner)
6843{
6844        unsigned int strhashval = ownerstr_hashval(owner);
6845        struct nfs4_stateowner *so;
6846
6847        lockdep_assert_held(&clp->cl_lock);
6848
6849        list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[strhashval],
6850                            so_strhash) {
6851                if (so->so_is_open_owner)
6852                        continue;
6853                if (same_owner_str(so, owner))
6854                        return lockowner(nfs4_get_stateowner(so));
6855        }
6856        return NULL;
6857}
6858
6859static struct nfs4_lockowner *
6860find_lockowner_str(struct nfs4_client *clp, struct xdr_netobj *owner)
6861{
6862        struct nfs4_lockowner *lo;
6863
6864        spin_lock(&clp->cl_lock);
6865        lo = find_lockowner_str_locked(clp, owner);
6866        spin_unlock(&clp->cl_lock);
6867        return lo;
6868}
6869
6870static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
6871{
6872        unhash_lockowner_locked(lockowner(sop));
6873}
6874
6875static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
6876{
6877        struct nfs4_lockowner *lo = lockowner(sop);
6878
6879        kmem_cache_free(lockowner_slab, lo);
6880}
6881
6882static const struct nfs4_stateowner_operations lockowner_ops = {
6883        .so_unhash =    nfs4_unhash_lockowner,
6884        .so_free =      nfs4_free_lockowner,
6885};
6886
6887/*
6888 * Alloc a lock owner structure.
6889 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has 
6890 * occurred. 
6891 *
6892 * strhashval = ownerstr_hashval
6893 */
6894static struct nfs4_lockowner *
6895alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp,
6896                           struct nfs4_ol_stateid *open_stp,
6897                           struct nfsd4_lock *lock)
6898{
6899        struct nfs4_lockowner *lo, *ret;
6900
6901        lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
6902        if (!lo)
6903                return NULL;
6904        INIT_LIST_HEAD(&lo->lo_blocked);
6905        INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
6906        lo->lo_owner.so_is_open_owner = 0;
6907        lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
6908        lo->lo_owner.so_ops = &lockowner_ops;
6909        spin_lock(&clp->cl_lock);
6910        ret = find_lockowner_str_locked(clp, &lock->lk_new_owner);
6911        if (ret == NULL) {
6912                list_add(&lo->lo_owner.so_strhash,
6913                         &clp->cl_ownerstr_hashtbl[strhashval]);
6914                ret = lo;
6915        } else
6916                nfs4_free_stateowner(&lo->lo_owner);
6917
6918        spin_unlock(&clp->cl_lock);
6919        return ret;
6920}
6921
6922static struct nfs4_ol_stateid *
6923find_lock_stateid(const struct nfs4_lockowner *lo,
6924                  const struct nfs4_ol_stateid *ost)
6925{
6926        struct nfs4_ol_stateid *lst;
6927
6928        lockdep_assert_held(&ost->st_stid.sc_client->cl_lock);
6929
6930        /* If ost is not hashed, ost->st_locks will not be valid */
6931        if (!nfs4_ol_stateid_unhashed(ost))
6932                list_for_each_entry(lst, &ost->st_locks, st_locks) {
6933                        if (lst->st_stateowner == &lo->lo_owner) {
6934                                refcount_inc(&lst->st_stid.sc_count);
6935                                return lst;
6936                        }
6937                }
6938        return NULL;
6939}
6940
6941static struct nfs4_ol_stateid *
6942init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
6943                  struct nfs4_file *fp, struct inode *inode,
6944                  struct nfs4_ol_stateid *open_stp)
6945{
6946        struct nfs4_client *clp = lo->lo_owner.so_client;
6947        struct nfs4_ol_stateid *retstp;
6948
6949        mutex_init(&stp->st_mutex);
6950        mutex_lock_nested(&stp->st_mutex, OPEN_STATEID_MUTEX);
6951retry:
6952        spin_lock(&clp->cl_lock);
6953        if (nfs4_ol_stateid_unhashed(open_stp))
6954                goto out_close;
6955        retstp = find_lock_stateid(lo, open_stp);
6956        if (retstp)
6957                goto out_found;
6958        refcount_inc(&stp->st_stid.sc_count);
6959        stp->st_stid.sc_type = NFS4_LOCK_STID;
6960        stp->st_stateowner = nfs4_get_stateowner(&lo->lo_owner);
6961        get_nfs4_file(fp);
6962        stp->st_stid.sc_file = fp;
6963        stp->st_access_bmap = 0;
6964        stp->st_deny_bmap = open_stp->st_deny_bmap;
6965        stp->st_openstp = open_stp;
6966        spin_lock(&fp->fi_lock);
6967        list_add(&stp->st_locks, &open_stp->st_locks);
6968        list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
6969        list_add(&stp->st_perfile, &fp->fi_stateids);
6970        spin_unlock(&fp->fi_lock);
6971        spin_unlock(&clp->cl_lock);
6972        return stp;
6973out_found:
6974        spin_unlock(&clp->cl_lock);
6975        if (nfsd4_lock_ol_stateid(retstp) != nfs_ok) {
6976                nfs4_put_stid(&retstp->st_stid);
6977                goto retry;
6978        }
6979        /* To keep mutex tracking happy */
6980        mutex_unlock(&stp->st_mutex);
6981        return retstp;
6982out_close:
6983        spin_unlock(&clp->cl_lock);
6984        mutex_unlock(&stp->st_mutex);
6985        return NULL;
6986}
6987
6988static struct nfs4_ol_stateid *
6989find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
6990                            struct inode *inode, struct nfs4_ol_stateid *ost,
6991                            bool *new)
6992{
6993        struct nfs4_stid *ns = NULL;
6994        struct nfs4_ol_stateid *lst;
6995        struct nfs4_openowner *oo = openowner(ost->st_stateowner);
6996        struct nfs4_client *clp = oo->oo_owner.so_client;
6997
6998        *new = false;
6999        spin_lock(&clp->cl_lock);
7000        lst = find_lock_stateid(lo, ost);
7001        spin_unlock(&clp->cl_lock);
7002        if (lst != NULL) {
7003                if (nfsd4_lock_ol_stateid(lst) == nfs_ok)
7004                        goto out;
7005                nfs4_put_stid(&lst->st_stid);
7006        }
7007        ns = nfs4_alloc_stid(clp, stateid_slab, nfs4_free_lock_stateid);
7008        if (ns == NULL)
7009                return NULL;
7010
7011        lst = init_lock_stateid(openlockstateid(ns), lo, fi, inode, ost);
7012        if (lst == openlockstateid(ns))
7013                *new = true;
7014        else
7015                nfs4_put_stid(ns);
7016out:
7017        return lst;
7018}
7019
7020static int
7021check_lock_length(u64 offset, u64 length)
7022{
7023        return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
7024                (length > ~offset)));
7025}
7026
7027static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
7028{
7029        struct nfs4_file *fp = lock_stp->st_stid.sc_file;
7030
7031        lockdep_assert_held(&fp->fi_lock);
7032
7033        if (test_access(access, lock_stp))
7034                return;
7035        __nfs4_file_get_access(fp, access);
7036        set_access(access, lock_stp);
7037}
7038
7039static __be32
7040lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
7041                            struct nfs4_ol_stateid *ost,
7042                            struct nfsd4_lock *lock,
7043                            struct nfs4_ol_stateid **plst, bool *new)
7044{
7045        __be32 status;
7046        struct nfs4_file *fi = ost->st_stid.sc_file;
7047        struct nfs4_openowner *oo = openowner(ost->st_stateowner);
7048        struct nfs4_client *cl = oo->oo_owner.so_client;
7049        struct inode *inode = d_inode(cstate->current_fh.fh_dentry);
7050        struct nfs4_lockowner *lo;
7051        struct nfs4_ol_stateid *lst;
7052        unsigned int strhashval;
7053
7054        lo = find_lockowner_str(cl, &lock->lk_new_owner);
7055        if (!lo) {
7056                strhashval = ownerstr_hashval(&lock->lk_new_owner);
7057                lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
7058                if (lo == NULL)
7059                        return nfserr_jukebox;
7060        } else {
7061                /* with an existing lockowner, seqids must be the same */
7062                status = nfserr_bad_seqid;
7063                if (!cstate->minorversion &&
7064                    lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
7065                        goto out;
7066        }
7067
7068        lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
7069        if (lst == NULL) {
7070                status = nfserr_jukebox;
7071                goto out;
7072        }
7073
7074        status = nfs_ok;
7075        *plst = lst;
7076out:
7077        nfs4_put_stateowner(&lo->lo_owner);
7078        return status;
7079}
7080
7081/*
7082 *  LOCK operation 
7083 */
7084__be32
7085nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7086           union nfsd4_op_u *u)
7087{
7088        struct nfsd4_lock *lock = &u->lock;
7089        struct nfs4_openowner *open_sop = NULL;
7090        struct nfs4_lockowner *lock_sop = NULL;
7091        struct nfs4_ol_stateid *lock_stp = NULL;
7092        struct nfs4_ol_stateid *open_stp = NULL;
7093        struct nfs4_file *fp;
7094        struct nfsd_file *nf = NULL;
7095        struct nfsd4_blocked_lock *nbl = NULL;
7096        struct file_lock *file_lock = NULL;
7097        struct file_lock *conflock = NULL;
7098        __be32 status = 0;
7099        int lkflg;
7100        int err;
7101        bool new = false;
7102        unsigned char fl_type;
7103        unsigned int fl_flags = FL_POSIX;
7104        struct net *net = SVC_NET(rqstp);
7105        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7106
7107        dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
7108                (long long) lock->lk_offset,
7109                (long long) lock->lk_length);
7110
7111        if (check_lock_length(lock->lk_offset, lock->lk_length))
7112                 return nfserr_inval;
7113
7114        if ((status = fh_verify(rqstp, &cstate->current_fh,
7115                                S_IFREG, NFSD_MAY_LOCK))) {
7116                dprintk("NFSD: nfsd4_lock: permission denied!\n");
7117                return status;
7118        }
7119
7120        if (lock->lk_is_new) {
7121                if (nfsd4_has_session(cstate))
7122                        /* See rfc 5661 18.10.3: given clientid is ignored: */
7123                        memcpy(&lock->lk_new_clientid,
7124                                &cstate->clp->cl_clientid,
7125                                sizeof(clientid_t));
7126
7127                /* validate and update open stateid and open seqid */
7128                status = nfs4_preprocess_confirmed_seqid_op(cstate,
7129                                        lock->lk_new_open_seqid,
7130                                        &lock->lk_new_open_stateid,
7131                                        &open_stp, nn);
7132                if (status)
7133                        goto out;
7134                mutex_unlock(&open_stp->st_mutex);
7135                open_sop = openowner(open_stp->st_stateowner);
7136                status = nfserr_bad_stateid;
7137                if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
7138                                                &lock->lk_new_clientid))
7139                        goto out;
7140                status = lookup_or_create_lock_state(cstate, open_stp, lock,
7141                                                        &lock_stp, &new);
7142        } else {
7143                status = nfs4_preprocess_seqid_op(cstate,
7144                                       lock->lk_old_lock_seqid,
7145                                       &lock->lk_old_lock_stateid,
7146                                       NFS4_LOCK_STID, &lock_stp, nn);
7147        }
7148        if (status)
7149                goto out;
7150        lock_sop = lockowner(lock_stp->st_stateowner);
7151
7152        lkflg = setlkflg(lock->lk_type);
7153        status = nfs4_check_openmode(lock_stp, lkflg);
7154        if (status)
7155                goto out;
7156
7157        status = nfserr_grace;
7158        if (locks_in_grace(net) && !lock->lk_reclaim)
7159                goto out;
7160        status = nfserr_no_grace;
7161        if (!locks_in_grace(net) && lock->lk_reclaim)
7162                goto out;
7163
7164        if (lock->lk_reclaim)
7165                fl_flags |= FL_RECLAIM;
7166
7167        fp = lock_stp->st_stid.sc_file;
7168        switch (lock->lk_type) {
7169                case NFS4_READW_LT:
7170                        if (nfsd4_has_session(cstate))
7171                                fl_flags |= FL_SLEEP;
7172                        fallthrough;
7173                case NFS4_READ_LT:
7174                        spin_lock(&fp->fi_lock);
7175                        nf = find_readable_file_locked(fp);
7176                        if (nf)
7177                                get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
7178                        spin_unlock(&fp->fi_lock);
7179                        fl_type = F_RDLCK;
7180                        break;
7181                case NFS4_WRITEW_LT:
7182                        if (nfsd4_has_session(cstate))
7183                                fl_flags |= FL_SLEEP;
7184                        fallthrough;
7185                case NFS4_WRITE_LT:
7186                        spin_lock(&fp->fi_lock);
7187                        nf = find_writeable_file_locked(fp);
7188                        if (nf)
7189                                get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
7190                        spin_unlock(&fp->fi_lock);
7191                        fl_type = F_WRLCK;
7192                        break;
7193                default:
7194                        status = nfserr_inval;
7195                goto out;
7196        }
7197
7198        if (!nf) {
7199                status = nfserr_openmode;
7200                goto out;
7201        }
7202
7203        /*
7204         * Most filesystems with their own ->lock operations will block
7205         * the nfsd thread waiting to acquire the lock.  That leads to
7206         * deadlocks (we don't want every nfsd thread tied up waiting
7207         * for file locks), so don't attempt blocking lock notifications
7208         * on those filesystems:
7209         */
7210        if (nf->nf_file->f_op->lock)
7211                fl_flags &= ~FL_SLEEP;
7212
7213        nbl = find_or_allocate_block(lock_sop, &fp->fi_fhandle, nn);
7214        if (!nbl) {
7215                dprintk("NFSD: %s: unable to allocate block!\n", __func__);
7216                status = nfserr_jukebox;
7217                goto out;
7218        }
7219
7220        file_lock = &nbl->nbl_lock;
7221        file_lock->fl_type = fl_type;
7222        file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(&lock_sop->lo_owner));
7223        file_lock->fl_pid = current->tgid;
7224        file_lock->fl_file = nf->nf_file;
7225        file_lock->fl_flags = fl_flags;
7226        file_lock->fl_lmops = &nfsd_posix_mng_ops;
7227        file_lock->fl_start = lock->lk_offset;
7228        file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
7229        nfs4_transform_lock_offset(file_lock);
7230
7231        conflock = locks_alloc_lock();
7232        if (!conflock) {
7233                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
7234                status = nfserr_jukebox;
7235                goto out;
7236        }
7237
7238        if (fl_flags & FL_SLEEP) {
7239                nbl->nbl_time = ktime_get_boottime_seconds();
7240                spin_lock(&nn->blocked_locks_lock);
7241                list_add_tail(&nbl->nbl_list, &lock_sop->lo_blocked);
7242                list_add_tail(&nbl->nbl_lru, &nn->blocked_locks_lru);
7243                kref_get(&nbl->nbl_kref);
7244                spin_unlock(&nn->blocked_locks_lock);
7245        }
7246
7247        err = vfs_lock_file(nf->nf_file, F_SETLK, file_lock, conflock);
7248        switch (err) {
7249        case 0: /* success! */
7250                nfs4_inc_and_copy_stateid(&lock->lk_resp_stateid, &lock_stp->st_stid);
7251                status = 0;
7252                if (lock->lk_reclaim)
7253                        nn->somebody_reclaimed = true;
7254                break;
7255        case FILE_LOCK_DEFERRED:
7256                kref_put(&nbl->nbl_kref, free_nbl);
7257                nbl = NULL;
7258                fallthrough;
7259        case -EAGAIN:           /* conflock holds conflicting lock */
7260                status = nfserr_denied;
7261                dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
7262                nfs4_set_lock_denied(conflock, &lock->lk_denied);
7263                break;
7264        case -EDEADLK:
7265                status = nfserr_deadlock;
7266                break;
7267        default:
7268                dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
7269                status = nfserrno(err);
7270                break;
7271        }
7272out:
7273        if (nbl) {
7274                /* dequeue it if we queued it before */
7275                if (fl_flags & FL_SLEEP) {
7276                        spin_lock(&nn->blocked_locks_lock);
7277                        if (!list_empty(&nbl->nbl_list) &&
7278                            !list_empty(&nbl->nbl_lru)) {
7279                                list_del_init(&nbl->nbl_list);
7280                                list_del_init(&nbl->nbl_lru);
7281                                kref_put(&nbl->nbl_kref, free_nbl);
7282                        }
7283                        /* nbl can use one of lists to be linked to reaplist */
7284                        spin_unlock(&nn->blocked_locks_lock);
7285                }
7286                free_blocked_lock(nbl);
7287        }
7288        if (nf)
7289                nfsd_file_put(nf);
7290        if (lock_stp) {
7291                /* Bump seqid manually if the 4.0 replay owner is openowner */
7292                if (cstate->replay_owner &&
7293                    cstate->replay_owner != &lock_sop->lo_owner &&
7294                    seqid_mutating_err(ntohl(status)))
7295                        lock_sop->lo_owner.so_seqid++;
7296
7297                /*
7298                 * If this is a new, never-before-used stateid, and we are
7299                 * returning an error, then just go ahead and release it.
7300                 */
7301                if (status && new)
7302                        release_lock_stateid(lock_stp);
7303
7304                mutex_unlock(&lock_stp->st_mutex);
7305
7306                nfs4_put_stid(&lock_stp->st_stid);
7307        }
7308        if (open_stp)
7309                nfs4_put_stid(&open_stp->st_stid);
7310        nfsd4_bump_seqid(cstate, status);
7311        if (conflock)
7312                locks_free_lock(conflock);
7313        return status;
7314}
7315
7316/*
7317 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
7318 * so we do a temporary open here just to get an open file to pass to
7319 * vfs_test_lock.
7320 */
7321static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
7322{
7323        struct nfsd_file *nf;
7324        __be32 err;
7325
7326        err = nfsd_file_acquire(rqstp, fhp, NFSD_MAY_READ, &nf);
7327        if (err)
7328                return err;
7329        fh_lock(fhp); /* to block new leases till after test_lock: */
7330        err = nfserrno(nfsd_open_break_lease(fhp->fh_dentry->d_inode,
7331                                                        NFSD_MAY_READ));
7332        if (err)
7333                goto out;
7334        lock->fl_file = nf->nf_file;
7335        err = nfserrno(vfs_test_lock(nf->nf_file, lock));
7336        lock->fl_file = NULL;
7337out:
7338        fh_unlock(fhp);
7339        nfsd_file_put(nf);
7340        return err;
7341}
7342
7343/*
7344 * LOCKT operation
7345 */
7346__be32
7347nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7348            union nfsd4_op_u *u)
7349{
7350        struct nfsd4_lockt *lockt = &u->lockt;
7351        struct file_lock *file_lock = NULL;
7352        struct nfs4_lockowner *lo = NULL;
7353        __be32 status;
7354        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
7355
7356        if (locks_in_grace(SVC_NET(rqstp)))
7357                return nfserr_grace;
7358
7359        if (check_lock_length(lockt->lt_offset, lockt->lt_length))
7360                 return nfserr_inval;
7361
7362        if (!nfsd4_has_session(cstate)) {
7363                status = set_client(&lockt->lt_clientid, cstate, nn);
7364                if (status)
7365                        goto out;
7366        }
7367
7368        if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
7369                goto out;
7370
7371        file_lock = locks_alloc_lock();
7372        if (!file_lock) {
7373                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
7374                status = nfserr_jukebox;
7375                goto out;
7376        }
7377
7378        switch (lockt->lt_type) {
7379                case NFS4_READ_LT:
7380                case NFS4_READW_LT:
7381                        file_lock->fl_type = F_RDLCK;
7382                        break;
7383                case NFS4_WRITE_LT:
7384                case NFS4_WRITEW_LT:
7385                        file_lock->fl_type = F_WRLCK;
7386                        break;
7387                default:
7388                        dprintk("NFSD: nfs4_lockt: bad lock type!\n");
7389                        status = nfserr_inval;
7390                        goto out;
7391        }
7392
7393        lo = find_lockowner_str(cstate->clp, &lockt->lt_owner);
7394        if (lo)
7395                file_lock->fl_owner = (fl_owner_t)lo;
7396        file_lock->fl_pid = current->tgid;
7397        file_lock->fl_flags = FL_POSIX;
7398
7399        file_lock->fl_start = lockt->lt_offset;
7400        file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
7401
7402        nfs4_transform_lock_offset(file_lock);
7403
7404        status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
7405        if (status)
7406                goto out;
7407
7408        if (file_lock->fl_type != F_UNLCK) {
7409                status = nfserr_denied;
7410                nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
7411        }
7412out:
7413        if (lo)
7414                nfs4_put_stateowner(&lo->lo_owner);
7415        if (file_lock)
7416                locks_free_lock(file_lock);
7417        return status;
7418}
7419
7420__be32
7421nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7422            union nfsd4_op_u *u)
7423{
7424        struct nfsd4_locku *locku = &u->locku;
7425        struct nfs4_ol_stateid *stp;
7426        struct nfsd_file *nf = NULL;
7427        struct file_lock *file_lock = NULL;
7428        __be32 status;
7429        int err;
7430        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
7431
7432        dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
7433                (long long) locku->lu_offset,
7434                (long long) locku->lu_length);
7435
7436        if (check_lock_length(locku->lu_offset, locku->lu_length))
7437                 return nfserr_inval;
7438
7439        status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
7440                                        &locku->lu_stateid, NFS4_LOCK_STID,
7441                                        &stp, nn);
7442        if (status)
7443                goto out;
7444        nf = find_any_file(stp->st_stid.sc_file);
7445        if (!nf) {
7446                status = nfserr_lock_range;
7447                goto put_stateid;
7448        }
7449        file_lock = locks_alloc_lock();
7450        if (!file_lock) {
7451                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
7452                status = nfserr_jukebox;
7453                goto put_file;
7454        }
7455
7456        file_lock->fl_type = F_UNLCK;
7457        file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(stp->st_stateowner));
7458        file_lock->fl_pid = current->tgid;
7459        file_lock->fl_file = nf->nf_file;
7460        file_lock->fl_flags = FL_POSIX;
7461        file_lock->fl_lmops = &nfsd_posix_mng_ops;
7462        file_lock->fl_start = locku->lu_offset;
7463
7464        file_lock->fl_end = last_byte_offset(locku->lu_offset,
7465                                                locku->lu_length);
7466        nfs4_transform_lock_offset(file_lock);
7467
7468        err = vfs_lock_file(nf->nf_file, F_SETLK, file_lock, NULL);
7469        if (err) {
7470                dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
7471                goto out_nfserr;
7472        }
7473        nfs4_inc_and_copy_stateid(&locku->lu_stateid, &stp->st_stid);
7474put_file:
7475        nfsd_file_put(nf);
7476put_stateid:
7477        mutex_unlock(&stp->st_mutex);
7478        nfs4_put_stid(&stp->st_stid);
7479out:
7480        nfsd4_bump_seqid(cstate, status);
7481        if (file_lock)
7482                locks_free_lock(file_lock);
7483        return status;
7484
7485out_nfserr:
7486        status = nfserrno(err);
7487        goto put_file;
7488}
7489
7490/*
7491 * returns
7492 *      true:  locks held by lockowner
7493 *      false: no locks held by lockowner
7494 */
7495static bool
7496check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
7497{
7498        struct file_lock *fl;
7499        int status = false;
7500        struct nfsd_file *nf = find_any_file(fp);
7501        struct inode *inode;
7502        struct file_lock_context *flctx;
7503
7504        if (!nf) {
7505                /* Any valid lock stateid should have some sort of access */
7506                WARN_ON_ONCE(1);
7507                return status;
7508        }
7509
7510        inode = locks_inode(nf->nf_file);
7511        flctx = inode->i_flctx;
7512
7513        if (flctx && !list_empty_careful(&flctx->flc_posix)) {
7514                spin_lock(&flctx->flc_lock);
7515                list_for_each_entry(fl, &flctx->flc_posix, fl_list) {
7516                        if (fl->fl_owner == (fl_owner_t)lowner) {
7517                                status = true;
7518                                break;
7519                        }
7520                }
7521                spin_unlock(&flctx->flc_lock);
7522        }
7523        nfsd_file_put(nf);
7524        return status;
7525}
7526
7527/**
7528 * nfsd4_release_lockowner - process NFSv4.0 RELEASE_LOCKOWNER operations
7529 * @rqstp: RPC transaction
7530 * @cstate: NFSv4 COMPOUND state
7531 * @u: RELEASE_LOCKOWNER arguments
7532 *
7533 * The lockowner's so_count is bumped when a lock record is added
7534 * or when copying a conflicting lock. The latter case is brief,
7535 * but can lead to fleeting false positives when looking for
7536 * locks-in-use.
7537 *
7538 * Return values:
7539 *   %nfs_ok: lockowner released or not found
7540 *   %nfserr_locks_held: lockowner still in use
7541 *   %nfserr_stale_clientid: clientid no longer active
7542 *   %nfserr_expired: clientid not recognized
7543 */
7544__be32
7545nfsd4_release_lockowner(struct svc_rqst *rqstp,
7546                        struct nfsd4_compound_state *cstate,
7547                        union nfsd4_op_u *u)
7548{
7549        struct nfsd4_release_lockowner *rlockowner = &u->release_lockowner;
7550        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
7551        clientid_t *clid = &rlockowner->rl_clientid;
7552        struct nfs4_ol_stateid *stp;
7553        struct nfs4_lockowner *lo;
7554        struct nfs4_client *clp;
7555        LIST_HEAD(reaplist);
7556        __be32 status;
7557
7558        dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
7559                clid->cl_boot, clid->cl_id);
7560
7561        status = set_client(clid, cstate, nn);
7562        if (status)
7563                return status;
7564        clp = cstate->clp;
7565
7566        spin_lock(&clp->cl_lock);
7567        lo = find_lockowner_str_locked(clp, &rlockowner->rl_owner);
7568        if (!lo) {
7569                spin_unlock(&clp->cl_lock);
7570                return nfs_ok;
7571        }
7572        if (atomic_read(&lo->lo_owner.so_count) != 2) {
7573                spin_unlock(&clp->cl_lock);
7574                nfs4_put_stateowner(&lo->lo_owner);
7575                return nfserr_locks_held;
7576        }
7577        unhash_lockowner_locked(lo);
7578        while (!list_empty(&lo->lo_owner.so_stateids)) {
7579                stp = list_first_entry(&lo->lo_owner.so_stateids,
7580                                       struct nfs4_ol_stateid,
7581                                       st_perstateowner);
7582                WARN_ON(!unhash_lock_stateid(stp));
7583                put_ol_stateid_locked(stp, &reaplist);
7584        }
7585        spin_unlock(&clp->cl_lock);
7586
7587        free_ol_stateid_reaplist(&reaplist);
7588        remove_blocked_locks(lo);
7589        nfs4_put_stateowner(&lo->lo_owner);
7590        return nfs_ok;
7591}
7592
7593static inline struct nfs4_client_reclaim *
7594alloc_reclaim(void)
7595{
7596        return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
7597}
7598
7599bool
7600nfs4_has_reclaimed_state(struct xdr_netobj name, struct nfsd_net *nn)
7601{
7602        struct nfs4_client_reclaim *crp;
7603
7604        crp = nfsd4_find_reclaim_client(name, nn);
7605        return (crp && crp->cr_clp);
7606}
7607
7608/*
7609 * failure => all reset bets are off, nfserr_no_grace...
7610 *
7611 * The caller is responsible for freeing name.data if NULL is returned (it
7612 * will be freed in nfs4_remove_reclaim_record in the normal case).
7613 */
7614struct nfs4_client_reclaim *
7615nfs4_client_to_reclaim(struct xdr_netobj name, struct xdr_netobj princhash,
7616                struct nfsd_net *nn)
7617{
7618        unsigned int strhashval;
7619        struct nfs4_client_reclaim *crp;
7620
7621        crp = alloc_reclaim();
7622        if (crp) {
7623                strhashval = clientstr_hashval(name);
7624                INIT_LIST_HEAD(&crp->cr_strhash);
7625                list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
7626                crp->cr_name.data = name.data;
7627                crp->cr_name.len = name.len;
7628                crp->cr_princhash.data = princhash.data;
7629                crp->cr_princhash.len = princhash.len;
7630                crp->cr_clp = NULL;
7631                nn->reclaim_str_hashtbl_size++;
7632        }
7633        return crp;
7634}
7635
7636void
7637nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
7638{
7639        list_del(&crp->cr_strhash);
7640        kfree(crp->cr_name.data);
7641        kfree(crp->cr_princhash.data);
7642        kfree(crp);
7643        nn->reclaim_str_hashtbl_size--;
7644}
7645
7646void
7647nfs4_release_reclaim(struct nfsd_net *nn)
7648{
7649        struct nfs4_client_reclaim *crp = NULL;
7650        int i;
7651
7652        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7653                while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
7654                        crp = list_entry(nn->reclaim_str_hashtbl[i].next,
7655                                        struct nfs4_client_reclaim, cr_strhash);
7656                        nfs4_remove_reclaim_record(crp, nn);
7657                }
7658        }
7659        WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
7660}
7661
7662/*
7663 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
7664struct nfs4_client_reclaim *
7665nfsd4_find_reclaim_client(struct xdr_netobj name, struct nfsd_net *nn)
7666{
7667        unsigned int strhashval;
7668        struct nfs4_client_reclaim *crp = NULL;
7669
7670        strhashval = clientstr_hashval(name);
7671        list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
7672                if (compare_blob(&crp->cr_name, &name) == 0) {
7673                        return crp;
7674                }
7675        }
7676        return NULL;
7677}
7678
7679__be32
7680nfs4_check_open_reclaim(struct nfs4_client *clp)
7681{
7682        if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &clp->cl_flags))
7683                return nfserr_no_grace;
7684
7685        if (nfsd4_client_record_check(clp))
7686                return nfserr_reclaim_bad;
7687
7688        return nfs_ok;
7689}
7690
7691/*
7692 * Since the lifetime of a delegation isn't limited to that of an open, a
7693 * client may quite reasonably hang on to a delegation as long as it has
7694 * the inode cached.  This becomes an obvious problem the first time a
7695 * client's inode cache approaches the size of the server's total memory.
7696 *
7697 * For now we avoid this problem by imposing a hard limit on the number
7698 * of delegations, which varies according to the server's memory size.
7699 */
7700static void
7701set_max_delegations(void)
7702{
7703        /*
7704         * Allow at most 4 delegations per megabyte of RAM.  Quick
7705         * estimates suggest that in the worst case (where every delegation
7706         * is for a different inode), a delegation could take about 1.5K,
7707         * giving a worst case usage of about 6% of memory.
7708         */
7709        max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
7710}
7711
7712static int nfs4_state_create_net(struct net *net)
7713{
7714        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7715        int i;
7716
7717        nn->conf_id_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
7718                                            sizeof(struct list_head),
7719                                            GFP_KERNEL);
7720        if (!nn->conf_id_hashtbl)
7721                goto err;
7722        nn->unconf_id_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
7723                                              sizeof(struct list_head),
7724                                              GFP_KERNEL);
7725        if (!nn->unconf_id_hashtbl)
7726                goto err_unconf_id;
7727        nn->sessionid_hashtbl = kmalloc_array(SESSION_HASH_SIZE,
7728                                              sizeof(struct list_head),
7729                                              GFP_KERNEL);
7730        if (!nn->sessionid_hashtbl)
7731                goto err_sessionid;
7732
7733        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7734                INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
7735                INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
7736        }
7737        for (i = 0; i < SESSION_HASH_SIZE; i++)
7738                INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
7739        nn->conf_name_tree = RB_ROOT;
7740        nn->unconf_name_tree = RB_ROOT;
7741        nn->boot_time = ktime_get_real_seconds();
7742        nn->grace_ended = false;
7743        nn->nfsd4_manager.block_opens = true;
7744        INIT_LIST_HEAD(&nn->nfsd4_manager.list);
7745        INIT_LIST_HEAD(&nn->client_lru);
7746        INIT_LIST_HEAD(&nn->close_lru);
7747        INIT_LIST_HEAD(&nn->del_recall_lru);
7748        spin_lock_init(&nn->client_lock);
7749        spin_lock_init(&nn->s2s_cp_lock);
7750        idr_init(&nn->s2s_cp_stateids);
7751
7752        spin_lock_init(&nn->blocked_locks_lock);
7753        INIT_LIST_HEAD(&nn->blocked_locks_lru);
7754
7755        INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
7756        get_net(net);
7757
7758        return 0;
7759
7760err_sessionid:
7761        kfree(nn->unconf_id_hashtbl);
7762err_unconf_id:
7763        kfree(nn->conf_id_hashtbl);
7764err:
7765        return -ENOMEM;
7766}
7767
7768static void
7769nfs4_state_destroy_net(struct net *net)
7770{
7771        int i;
7772        struct nfs4_client *clp = NULL;
7773        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7774
7775        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7776                while (!list_empty(&nn->conf_id_hashtbl[i])) {
7777                        clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
7778                        destroy_client(clp);
7779                }
7780        }
7781
7782        WARN_ON(!list_empty(&nn->blocked_locks_lru));
7783
7784        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7785                while (!list_empty(&nn->unconf_id_hashtbl[i])) {
7786                        clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
7787                        destroy_client(clp);
7788                }
7789        }
7790
7791        kfree(nn->sessionid_hashtbl);
7792        kfree(nn->unconf_id_hashtbl);
7793        kfree(nn->conf_id_hashtbl);
7794        put_net(net);
7795}
7796
7797int
7798nfs4_state_start_net(struct net *net)
7799{
7800        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7801        int ret;
7802
7803        ret = nfs4_state_create_net(net);
7804        if (ret)
7805                return ret;
7806        locks_start_grace(net, &nn->nfsd4_manager);
7807        nfsd4_client_tracking_init(net);
7808        if (nn->track_reclaim_completes && nn->reclaim_str_hashtbl_size == 0)
7809                goto skip_grace;
7810        printk(KERN_INFO "NFSD: starting %lld-second grace period (net %x)\n",
7811               nn->nfsd4_grace, net->ns.inum);
7812        trace_nfsd_grace_start(nn);
7813        queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
7814        return 0;
7815
7816skip_grace:
7817        printk(KERN_INFO "NFSD: no clients to reclaim, skipping NFSv4 grace period (net %x)\n",
7818                        net->ns.inum);
7819        queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_lease * HZ);
7820        nfsd4_end_grace(nn);
7821        return 0;
7822}
7823
7824/* initialization to perform when the nfsd service is started: */
7825
7826int
7827nfs4_state_start(void)
7828{
7829        int ret;
7830
7831        ret = nfsd4_create_callback_queue();
7832        if (ret)
7833                return ret;
7834
7835        set_max_delegations();
7836        return 0;
7837}
7838
7839void
7840nfs4_state_shutdown_net(struct net *net)
7841{
7842        struct nfs4_delegation *dp = NULL;
7843        struct list_head *pos, *next, reaplist;
7844        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7845
7846        cancel_delayed_work_sync(&nn->laundromat_work);
7847        locks_end_grace(&nn->nfsd4_manager);
7848
7849        INIT_LIST_HEAD(&reaplist);
7850        spin_lock(&state_lock);
7851        list_for_each_safe(pos, next, &nn->del_recall_lru) {
7852                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
7853                WARN_ON(!unhash_delegation_locked(dp));
7854                list_add(&dp->dl_recall_lru, &reaplist);
7855        }
7856        spin_unlock(&state_lock);
7857        list_for_each_safe(pos, next, &reaplist) {
7858                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
7859                list_del_init(&dp->dl_recall_lru);
7860                destroy_unhashed_deleg(dp);
7861        }
7862
7863        nfsd4_client_tracking_exit(net);
7864        nfs4_state_destroy_net(net);
7865#ifdef CONFIG_NFSD_V4_2_INTER_SSC
7866        nfsd4_ssc_shutdown_umount(nn);
7867#endif
7868}
7869
7870void
7871nfs4_state_shutdown(void)
7872{
7873        nfsd4_destroy_callback_queue();
7874}
7875
7876static void
7877get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
7878{
7879        if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG) &&
7880            CURRENT_STATEID(stateid))
7881                memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
7882}
7883
7884static void
7885put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
7886{
7887        if (cstate->minorversion) {
7888                memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
7889                SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
7890        }
7891}
7892
7893void
7894clear_current_stateid(struct nfsd4_compound_state *cstate)
7895{
7896        CLEAR_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
7897}
7898
7899/*
7900 * functions to set current state id
7901 */
7902void
7903nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate,
7904                union nfsd4_op_u *u)
7905{
7906        put_stateid(cstate, &u->open_downgrade.od_stateid);
7907}
7908
7909void
7910nfsd4_set_openstateid(struct nfsd4_compound_state *cstate,
7911                union nfsd4_op_u *u)
7912{
7913        put_stateid(cstate, &u->open.op_stateid);
7914}
7915
7916void
7917nfsd4_set_closestateid(struct nfsd4_compound_state *cstate,
7918                union nfsd4_op_u *u)
7919{
7920        put_stateid(cstate, &u->close.cl_stateid);
7921}
7922
7923void
7924nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate,
7925                union nfsd4_op_u *u)
7926{
7927        put_stateid(cstate, &u->lock.lk_resp_stateid);
7928}
7929
7930/*
7931 * functions to consume current state id
7932 */
7933
7934void
7935nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate,
7936                union nfsd4_op_u *u)
7937{
7938        get_stateid(cstate, &u->open_downgrade.od_stateid);
7939}
7940
7941void
7942nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate,
7943                union nfsd4_op_u *u)
7944{
7945        get_stateid(cstate, &u->delegreturn.dr_stateid);
7946}
7947
7948void
7949nfsd4_get_freestateid(struct nfsd4_compound_state *cstate,
7950                union nfsd4_op_u *u)
7951{
7952        get_stateid(cstate, &u->free_stateid.fr_stateid);
7953}
7954
7955void
7956nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate,
7957                union nfsd4_op_u *u)
7958{
7959        get_stateid(cstate, &u->setattr.sa_stateid);
7960}
7961
7962void
7963nfsd4_get_closestateid(struct nfsd4_compound_state *cstate,
7964                union nfsd4_op_u *u)
7965{
7966        get_stateid(cstate, &u->close.cl_stateid);
7967}
7968
7969void
7970nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate,
7971                union nfsd4_op_u *u)
7972{
7973        get_stateid(cstate, &u->locku.lu_stateid);
7974}
7975
7976void
7977nfsd4_get_readstateid(struct nfsd4_compound_state *cstate,
7978                union nfsd4_op_u *u)
7979{
7980        get_stateid(cstate, &u->read.rd_stateid);
7981}
7982
7983void
7984nfsd4_get_writestateid(struct nfsd4_compound_state *cstate,
7985                union nfsd4_op_u *u)
7986{
7987        get_stateid(cstate, &u->write.wr_stateid);
7988}
7989