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 "xdr4.h"
  47#include "xdr4cb.h"
  48#include "vfs.h"
  49#include "current_stateid.h"
  50
  51#include "netns.h"
  52#include "pnfs.h"
  53#include "filecache.h"
  54#include "trace.h"
  55
  56#define NFSDDBG_FACILITY                NFSDDBG_PROC
  57
  58#define all_ones {{~0,~0},~0}
  59static const stateid_t one_stateid = {
  60        .si_generation = ~0,
  61        .si_opaque = all_ones,
  62};
  63static const stateid_t zero_stateid = {
  64        /* all fields zero */
  65};
  66static const stateid_t currentstateid = {
  67        .si_generation = 1,
  68};
  69static const stateid_t close_stateid = {
  70        .si_generation = 0xffffffffU,
  71};
  72
  73static u64 current_sessionid = 1;
  74
  75#define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
  76#define ONE_STATEID(stateid)  (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
  77#define CURRENT_STATEID(stateid) (!memcmp((stateid), &currentstateid, sizeof(stateid_t)))
  78#define CLOSE_STATEID(stateid)  (!memcmp((stateid), &close_stateid, sizeof(stateid_t)))
  79
  80/* forward declarations */
  81static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
  82static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
  83void nfsd4_end_grace(struct nfsd_net *nn);
  84static void _free_cpntf_state_locked(struct nfsd_net *nn, struct nfs4_cpntf_state *cps);
  85
  86/* Locking: */
  87
  88/*
  89 * Currently used for the del_recall_lru and file hash table.  In an
  90 * effort to decrease the scope of the client_mutex, this spinlock may
  91 * eventually cover more:
  92 */
  93static DEFINE_SPINLOCK(state_lock);
  94
  95enum nfsd4_st_mutex_lock_subclass {
  96        OPEN_STATEID_MUTEX = 0,
  97        LOCK_STATEID_MUTEX = 1,
  98};
  99
 100/*
 101 * A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
 102 * the refcount on the open stateid to drop.
 103 */
 104static DECLARE_WAIT_QUEUE_HEAD(close_wq);
 105
 106/*
 107 * A waitqueue where a writer to clients/#/ctl destroying a client can
 108 * wait for cl_rpc_users to drop to 0 and then for the client to be
 109 * unhashed.
 110 */
 111static DECLARE_WAIT_QUEUE_HEAD(expiry_wq);
 112
 113static struct kmem_cache *client_slab;
 114static struct kmem_cache *openowner_slab;
 115static struct kmem_cache *lockowner_slab;
 116static struct kmem_cache *file_slab;
 117static struct kmem_cache *stateid_slab;
 118static struct kmem_cache *deleg_slab;
 119static struct kmem_cache *odstate_slab;
 120
 121static void free_session(struct nfsd4_session *);
 122
 123static const struct nfsd4_callback_ops nfsd4_cb_recall_ops;
 124static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops;
 125
 126static bool is_session_dead(struct nfsd4_session *ses)
 127{
 128        return ses->se_flags & NFS4_SESSION_DEAD;
 129}
 130
 131static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
 132{
 133        if (atomic_read(&ses->se_ref) > ref_held_by_me)
 134                return nfserr_jukebox;
 135        ses->se_flags |= NFS4_SESSION_DEAD;
 136        return nfs_ok;
 137}
 138
 139static bool is_client_expired(struct nfs4_client *clp)
 140{
 141        return clp->cl_time == 0;
 142}
 143
 144static __be32 get_client_locked(struct nfs4_client *clp)
 145{
 146        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 147
 148        lockdep_assert_held(&nn->client_lock);
 149
 150        if (is_client_expired(clp))
 151                return nfserr_expired;
 152        atomic_inc(&clp->cl_rpc_users);
 153        return nfs_ok;
 154}
 155
 156/* must be called under the client_lock */
 157static inline void
 158renew_client_locked(struct nfs4_client *clp)
 159{
 160        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 161
 162        if (is_client_expired(clp)) {
 163                WARN_ON(1);
 164                printk("%s: client (clientid %08x/%08x) already expired\n",
 165                        __func__,
 166                        clp->cl_clientid.cl_boot,
 167                        clp->cl_clientid.cl_id);
 168                return;
 169        }
 170
 171        list_move_tail(&clp->cl_lru, &nn->client_lru);
 172        clp->cl_time = ktime_get_boottime_seconds();
 173}
 174
 175static void put_client_renew_locked(struct nfs4_client *clp)
 176{
 177        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 178
 179        lockdep_assert_held(&nn->client_lock);
 180
 181        if (!atomic_dec_and_test(&clp->cl_rpc_users))
 182                return;
 183        if (!is_client_expired(clp))
 184                renew_client_locked(clp);
 185        else
 186                wake_up_all(&expiry_wq);
 187}
 188
 189static void put_client_renew(struct nfs4_client *clp)
 190{
 191        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 192
 193        if (!atomic_dec_and_lock(&clp->cl_rpc_users, &nn->client_lock))
 194                return;
 195        if (!is_client_expired(clp))
 196                renew_client_locked(clp);
 197        else
 198                wake_up_all(&expiry_wq);
 199        spin_unlock(&nn->client_lock);
 200}
 201
 202static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
 203{
 204        __be32 status;
 205
 206        if (is_session_dead(ses))
 207                return nfserr_badsession;
 208        status = get_client_locked(ses->se_client);
 209        if (status)
 210                return status;
 211        atomic_inc(&ses->se_ref);
 212        return nfs_ok;
 213}
 214
 215static void nfsd4_put_session_locked(struct nfsd4_session *ses)
 216{
 217        struct nfs4_client *clp = ses->se_client;
 218        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 219
 220        lockdep_assert_held(&nn->client_lock);
 221
 222        if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
 223                free_session(ses);
 224        put_client_renew_locked(clp);
 225}
 226
 227static void nfsd4_put_session(struct nfsd4_session *ses)
 228{
 229        struct nfs4_client *clp = ses->se_client;
 230        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 231
 232        spin_lock(&nn->client_lock);
 233        nfsd4_put_session_locked(ses);
 234        spin_unlock(&nn->client_lock);
 235}
 236
 237static struct nfsd4_blocked_lock *
 238find_blocked_lock(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
 239                        struct nfsd_net *nn)
 240{
 241        struct nfsd4_blocked_lock *cur, *found = NULL;
 242
 243        spin_lock(&nn->blocked_locks_lock);
 244        list_for_each_entry(cur, &lo->lo_blocked, nbl_list) {
 245                if (fh_match(fh, &cur->nbl_fh)) {
 246                        list_del_init(&cur->nbl_list);
 247                        list_del_init(&cur->nbl_lru);
 248                        found = cur;
 249                        break;
 250                }
 251        }
 252        spin_unlock(&nn->blocked_locks_lock);
 253        if (found)
 254                locks_delete_block(&found->nbl_lock);
 255        return found;
 256}
 257
 258static struct nfsd4_blocked_lock *
 259find_or_allocate_block(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
 260                        struct nfsd_net *nn)
 261{
 262        struct nfsd4_blocked_lock *nbl;
 263
 264        nbl = find_blocked_lock(lo, fh, nn);
 265        if (!nbl) {
 266                nbl= kmalloc(sizeof(*nbl), GFP_KERNEL);
 267                if (nbl) {
 268                        INIT_LIST_HEAD(&nbl->nbl_list);
 269                        INIT_LIST_HEAD(&nbl->nbl_lru);
 270                        fh_copy_shallow(&nbl->nbl_fh, fh);
 271                        locks_init_lock(&nbl->nbl_lock);
 272                        nfsd4_init_cb(&nbl->nbl_cb, lo->lo_owner.so_client,
 273                                        &nfsd4_cb_notify_lock_ops,
 274                                        NFSPROC4_CLNT_CB_NOTIFY_LOCK);
 275                }
 276        }
 277        return nbl;
 278}
 279
 280static void
 281free_blocked_lock(struct nfsd4_blocked_lock *nbl)
 282{
 283        locks_delete_block(&nbl->nbl_lock);
 284        locks_release_private(&nbl->nbl_lock);
 285        kfree(nbl);
 286}
 287
 288static void
 289remove_blocked_locks(struct nfs4_lockowner *lo)
 290{
 291        struct nfs4_client *clp = lo->lo_owner.so_client;
 292        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 293        struct nfsd4_blocked_lock *nbl;
 294        LIST_HEAD(reaplist);
 295
 296        /* Dequeue all blocked locks */
 297        spin_lock(&nn->blocked_locks_lock);
 298        while (!list_empty(&lo->lo_blocked)) {
 299                nbl = list_first_entry(&lo->lo_blocked,
 300                                        struct nfsd4_blocked_lock,
 301                                        nbl_list);
 302                list_del_init(&nbl->nbl_list);
 303                list_move(&nbl->nbl_lru, &reaplist);
 304        }
 305        spin_unlock(&nn->blocked_locks_lock);
 306
 307        /* Now free them */
 308        while (!list_empty(&reaplist)) {
 309                nbl = list_first_entry(&reaplist, struct nfsd4_blocked_lock,
 310                                        nbl_lru);
 311                list_del_init(&nbl->nbl_lru);
 312                free_blocked_lock(nbl);
 313        }
 314}
 315
 316static void
 317nfsd4_cb_notify_lock_prepare(struct nfsd4_callback *cb)
 318{
 319        struct nfsd4_blocked_lock       *nbl = container_of(cb,
 320                                                struct nfsd4_blocked_lock, nbl_cb);
 321        locks_delete_block(&nbl->nbl_lock);
 322}
 323
 324static int
 325nfsd4_cb_notify_lock_done(struct nfsd4_callback *cb, struct rpc_task *task)
 326{
 327        /*
 328         * Since this is just an optimization, we don't try very hard if it
 329         * turns out not to succeed. We'll requeue it on NFS4ERR_DELAY, and
 330         * just quit trying on anything else.
 331         */
 332        switch (task->tk_status) {
 333        case -NFS4ERR_DELAY:
 334                rpc_delay(task, 1 * HZ);
 335                return 0;
 336        default:
 337                return 1;
 338        }
 339}
 340
 341static void
 342nfsd4_cb_notify_lock_release(struct nfsd4_callback *cb)
 343{
 344        struct nfsd4_blocked_lock       *nbl = container_of(cb,
 345                                                struct nfsd4_blocked_lock, nbl_cb);
 346
 347        free_blocked_lock(nbl);
 348}
 349
 350static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops = {
 351        .prepare        = nfsd4_cb_notify_lock_prepare,
 352        .done           = nfsd4_cb_notify_lock_done,
 353        .release        = nfsd4_cb_notify_lock_release,
 354};
 355
 356static inline struct nfs4_stateowner *
 357nfs4_get_stateowner(struct nfs4_stateowner *sop)
 358{
 359        atomic_inc(&sop->so_count);
 360        return sop;
 361}
 362
 363static int
 364same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner)
 365{
 366        return (sop->so_owner.len == owner->len) &&
 367                0 == memcmp(sop->so_owner.data, owner->data, owner->len);
 368}
 369
 370static struct nfs4_openowner *
 371find_openstateowner_str_locked(unsigned int hashval, struct nfsd4_open *open,
 372                        struct nfs4_client *clp)
 373{
 374        struct nfs4_stateowner *so;
 375
 376        lockdep_assert_held(&clp->cl_lock);
 377
 378        list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[hashval],
 379                            so_strhash) {
 380                if (!so->so_is_open_owner)
 381                        continue;
 382                if (same_owner_str(so, &open->op_owner))
 383                        return openowner(nfs4_get_stateowner(so));
 384        }
 385        return NULL;
 386}
 387
 388static struct nfs4_openowner *
 389find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
 390                        struct nfs4_client *clp)
 391{
 392        struct nfs4_openowner *oo;
 393
 394        spin_lock(&clp->cl_lock);
 395        oo = find_openstateowner_str_locked(hashval, open, clp);
 396        spin_unlock(&clp->cl_lock);
 397        return oo;
 398}
 399
 400static inline u32
 401opaque_hashval(const void *ptr, int nbytes)
 402{
 403        unsigned char *cptr = (unsigned char *) ptr;
 404
 405        u32 x = 0;
 406        while (nbytes--) {
 407                x *= 37;
 408                x += *cptr++;
 409        }
 410        return x;
 411}
 412
 413static void nfsd4_free_file_rcu(struct rcu_head *rcu)
 414{
 415        struct nfs4_file *fp = container_of(rcu, struct nfs4_file, fi_rcu);
 416
 417        kmem_cache_free(file_slab, fp);
 418}
 419
 420void
 421put_nfs4_file(struct nfs4_file *fi)
 422{
 423        might_lock(&state_lock);
 424
 425        if (refcount_dec_and_lock(&fi->fi_ref, &state_lock)) {
 426                hlist_del_rcu(&fi->fi_hash);
 427                spin_unlock(&state_lock);
 428                WARN_ON_ONCE(!list_empty(&fi->fi_clnt_odstate));
 429                WARN_ON_ONCE(!list_empty(&fi->fi_delegations));
 430                call_rcu(&fi->fi_rcu, nfsd4_free_file_rcu);
 431        }
 432}
 433
 434static struct nfsd_file *
 435__nfs4_get_fd(struct nfs4_file *f, int oflag)
 436{
 437        if (f->fi_fds[oflag])
 438                return nfsd_file_get(f->fi_fds[oflag]);
 439        return NULL;
 440}
 441
 442static struct nfsd_file *
 443find_writeable_file_locked(struct nfs4_file *f)
 444{
 445        struct nfsd_file *ret;
 446
 447        lockdep_assert_held(&f->fi_lock);
 448
 449        ret = __nfs4_get_fd(f, O_WRONLY);
 450        if (!ret)
 451                ret = __nfs4_get_fd(f, O_RDWR);
 452        return ret;
 453}
 454
 455static struct nfsd_file *
 456find_writeable_file(struct nfs4_file *f)
 457{
 458        struct nfsd_file *ret;
 459
 460        spin_lock(&f->fi_lock);
 461        ret = find_writeable_file_locked(f);
 462        spin_unlock(&f->fi_lock);
 463
 464        return ret;
 465}
 466
 467static struct nfsd_file *
 468find_readable_file_locked(struct nfs4_file *f)
 469{
 470        struct nfsd_file *ret;
 471
 472        lockdep_assert_held(&f->fi_lock);
 473
 474        ret = __nfs4_get_fd(f, O_RDONLY);
 475        if (!ret)
 476                ret = __nfs4_get_fd(f, O_RDWR);
 477        return ret;
 478}
 479
 480static struct nfsd_file *
 481find_readable_file(struct nfs4_file *f)
 482{
 483        struct nfsd_file *ret;
 484
 485        spin_lock(&f->fi_lock);
 486        ret = find_readable_file_locked(f);
 487        spin_unlock(&f->fi_lock);
 488
 489        return ret;
 490}
 491
 492struct nfsd_file *
 493find_any_file(struct nfs4_file *f)
 494{
 495        struct nfsd_file *ret;
 496
 497        if (!f)
 498                return NULL;
 499        spin_lock(&f->fi_lock);
 500        ret = __nfs4_get_fd(f, O_RDWR);
 501        if (!ret) {
 502                ret = __nfs4_get_fd(f, O_WRONLY);
 503                if (!ret)
 504                        ret = __nfs4_get_fd(f, O_RDONLY);
 505        }
 506        spin_unlock(&f->fi_lock);
 507        return ret;
 508}
 509
 510static struct nfsd_file *find_deleg_file(struct nfs4_file *f)
 511{
 512        struct nfsd_file *ret = NULL;
 513
 514        spin_lock(&f->fi_lock);
 515        if (f->fi_deleg_file)
 516                ret = nfsd_file_get(f->fi_deleg_file);
 517        spin_unlock(&f->fi_lock);
 518        return ret;
 519}
 520
 521static atomic_long_t num_delegations;
 522unsigned long max_delegations;
 523
 524/*
 525 * Open owner state (share locks)
 526 */
 527
 528/* hash tables for lock and open owners */
 529#define OWNER_HASH_BITS              8
 530#define OWNER_HASH_SIZE             (1 << OWNER_HASH_BITS)
 531#define OWNER_HASH_MASK             (OWNER_HASH_SIZE - 1)
 532
 533static unsigned int ownerstr_hashval(struct xdr_netobj *ownername)
 534{
 535        unsigned int ret;
 536
 537        ret = opaque_hashval(ownername->data, ownername->len);
 538        return ret & OWNER_HASH_MASK;
 539}
 540
 541/* hash table for nfs4_file */
 542#define FILE_HASH_BITS                   8
 543#define FILE_HASH_SIZE                  (1 << FILE_HASH_BITS)
 544
 545static unsigned int nfsd_fh_hashval(struct knfsd_fh *fh)
 546{
 547        return jhash2(fh->fh_base.fh_pad, XDR_QUADLEN(fh->fh_size), 0);
 548}
 549
 550static unsigned int file_hashval(struct knfsd_fh *fh)
 551{
 552        return nfsd_fh_hashval(fh) & (FILE_HASH_SIZE - 1);
 553}
 554
 555static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
 556
 557static void
 558__nfs4_file_get_access(struct nfs4_file *fp, u32 access)
 559{
 560        lockdep_assert_held(&fp->fi_lock);
 561
 562        if (access & NFS4_SHARE_ACCESS_WRITE)
 563                atomic_inc(&fp->fi_access[O_WRONLY]);
 564        if (access & NFS4_SHARE_ACCESS_READ)
 565                atomic_inc(&fp->fi_access[O_RDONLY]);
 566}
 567
 568static __be32
 569nfs4_file_get_access(struct nfs4_file *fp, u32 access)
 570{
 571        lockdep_assert_held(&fp->fi_lock);
 572
 573        /* Does this access mode make sense? */
 574        if (access & ~NFS4_SHARE_ACCESS_BOTH)
 575                return nfserr_inval;
 576
 577        /* Does it conflict with a deny mode already set? */
 578        if ((access & fp->fi_share_deny) != 0)
 579                return nfserr_share_denied;
 580
 581        __nfs4_file_get_access(fp, access);
 582        return nfs_ok;
 583}
 584
 585static __be32 nfs4_file_check_deny(struct nfs4_file *fp, u32 deny)
 586{
 587        /* Common case is that there is no deny mode. */
 588        if (deny) {
 589                /* Does this deny mode make sense? */
 590                if (deny & ~NFS4_SHARE_DENY_BOTH)
 591                        return nfserr_inval;
 592
 593                if ((deny & NFS4_SHARE_DENY_READ) &&
 594                    atomic_read(&fp->fi_access[O_RDONLY]))
 595                        return nfserr_share_denied;
 596
 597                if ((deny & NFS4_SHARE_DENY_WRITE) &&
 598                    atomic_read(&fp->fi_access[O_WRONLY]))
 599                        return nfserr_share_denied;
 600        }
 601        return nfs_ok;
 602}
 603
 604static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
 605{
 606        might_lock(&fp->fi_lock);
 607
 608        if (atomic_dec_and_lock(&fp->fi_access[oflag], &fp->fi_lock)) {
 609                struct nfsd_file *f1 = NULL;
 610                struct nfsd_file *f2 = NULL;
 611
 612                swap(f1, fp->fi_fds[oflag]);
 613                if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
 614                        swap(f2, fp->fi_fds[O_RDWR]);
 615                spin_unlock(&fp->fi_lock);
 616                if (f1)
 617                        nfsd_file_put(f1);
 618                if (f2)
 619                        nfsd_file_put(f2);
 620        }
 621}
 622
 623static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
 624{
 625        WARN_ON_ONCE(access & ~NFS4_SHARE_ACCESS_BOTH);
 626
 627        if (access & NFS4_SHARE_ACCESS_WRITE)
 628                __nfs4_file_put_access(fp, O_WRONLY);
 629        if (access & NFS4_SHARE_ACCESS_READ)
 630                __nfs4_file_put_access(fp, O_RDONLY);
 631}
 632
 633/*
 634 * Allocate a new open/delegation state counter. This is needed for
 635 * pNFS for proper return on close semantics.
 636 *
 637 * Note that we only allocate it for pNFS-enabled exports, otherwise
 638 * all pointers to struct nfs4_clnt_odstate are always NULL.
 639 */
 640static struct nfs4_clnt_odstate *
 641alloc_clnt_odstate(struct nfs4_client *clp)
 642{
 643        struct nfs4_clnt_odstate *co;
 644
 645        co = kmem_cache_zalloc(odstate_slab, GFP_KERNEL);
 646        if (co) {
 647                co->co_client = clp;
 648                refcount_set(&co->co_odcount, 1);
 649        }
 650        return co;
 651}
 652
 653static void
 654hash_clnt_odstate_locked(struct nfs4_clnt_odstate *co)
 655{
 656        struct nfs4_file *fp = co->co_file;
 657
 658        lockdep_assert_held(&fp->fi_lock);
 659        list_add(&co->co_perfile, &fp->fi_clnt_odstate);
 660}
 661
 662static inline void
 663get_clnt_odstate(struct nfs4_clnt_odstate *co)
 664{
 665        if (co)
 666                refcount_inc(&co->co_odcount);
 667}
 668
 669static void
 670put_clnt_odstate(struct nfs4_clnt_odstate *co)
 671{
 672        struct nfs4_file *fp;
 673
 674        if (!co)
 675                return;
 676
 677        fp = co->co_file;
 678        if (refcount_dec_and_lock(&co->co_odcount, &fp->fi_lock)) {
 679                list_del(&co->co_perfile);
 680                spin_unlock(&fp->fi_lock);
 681
 682                nfsd4_return_all_file_layouts(co->co_client, fp);
 683                kmem_cache_free(odstate_slab, co);
 684        }
 685}
 686
 687static struct nfs4_clnt_odstate *
 688find_or_hash_clnt_odstate(struct nfs4_file *fp, struct nfs4_clnt_odstate *new)
 689{
 690        struct nfs4_clnt_odstate *co;
 691        struct nfs4_client *cl;
 692
 693        if (!new)
 694                return NULL;
 695
 696        cl = new->co_client;
 697
 698        spin_lock(&fp->fi_lock);
 699        list_for_each_entry(co, &fp->fi_clnt_odstate, co_perfile) {
 700                if (co->co_client == cl) {
 701                        get_clnt_odstate(co);
 702                        goto out;
 703                }
 704        }
 705        co = new;
 706        co->co_file = fp;
 707        hash_clnt_odstate_locked(new);
 708out:
 709        spin_unlock(&fp->fi_lock);
 710        return co;
 711}
 712
 713struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab,
 714                                  void (*sc_free)(struct nfs4_stid *))
 715{
 716        struct nfs4_stid *stid;
 717        int new_id;
 718
 719        stid = kmem_cache_zalloc(slab, GFP_KERNEL);
 720        if (!stid)
 721                return NULL;
 722
 723        idr_preload(GFP_KERNEL);
 724        spin_lock(&cl->cl_lock);
 725        /* Reserving 0 for start of file in nfsdfs "states" file: */
 726        new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 1, 0, GFP_NOWAIT);
 727        spin_unlock(&cl->cl_lock);
 728        idr_preload_end();
 729        if (new_id < 0)
 730                goto out_free;
 731
 732        stid->sc_free = sc_free;
 733        stid->sc_client = cl;
 734        stid->sc_stateid.si_opaque.so_id = new_id;
 735        stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
 736        /* Will be incremented before return to client: */
 737        refcount_set(&stid->sc_count, 1);
 738        spin_lock_init(&stid->sc_lock);
 739        INIT_LIST_HEAD(&stid->sc_cp_list);
 740
 741        /*
 742         * It shouldn't be a problem to reuse an opaque stateid value.
 743         * I don't think it is for 4.1.  But with 4.0 I worry that, for
 744         * example, a stray write retransmission could be accepted by
 745         * the server when it should have been rejected.  Therefore,
 746         * adopt a trick from the sctp code to attempt to maximize the
 747         * amount of time until an id is reused, by ensuring they always
 748         * "increase" (mod INT_MAX):
 749         */
 750        return stid;
 751out_free:
 752        kmem_cache_free(slab, stid);
 753        return NULL;
 754}
 755
 756/*
 757 * Create a unique stateid_t to represent each COPY.
 758 */
 759static int nfs4_init_cp_state(struct nfsd_net *nn, copy_stateid_t *stid,
 760                              unsigned char sc_type)
 761{
 762        int new_id;
 763
 764        stid->stid.si_opaque.so_clid.cl_boot = (u32)nn->boot_time;
 765        stid->stid.si_opaque.so_clid.cl_id = nn->s2s_cp_cl_id;
 766        stid->sc_type = sc_type;
 767
 768        idr_preload(GFP_KERNEL);
 769        spin_lock(&nn->s2s_cp_lock);
 770        new_id = idr_alloc_cyclic(&nn->s2s_cp_stateids, stid, 0, 0, GFP_NOWAIT);
 771        stid->stid.si_opaque.so_id = new_id;
 772        stid->stid.si_generation = 1;
 773        spin_unlock(&nn->s2s_cp_lock);
 774        idr_preload_end();
 775        if (new_id < 0)
 776                return 0;
 777        return 1;
 778}
 779
 780int nfs4_init_copy_state(struct nfsd_net *nn, struct nfsd4_copy *copy)
 781{
 782        return nfs4_init_cp_state(nn, &copy->cp_stateid, NFS4_COPY_STID);
 783}
 784
 785struct nfs4_cpntf_state *nfs4_alloc_init_cpntf_state(struct nfsd_net *nn,
 786                                                     struct nfs4_stid *p_stid)
 787{
 788        struct nfs4_cpntf_state *cps;
 789
 790        cps = kzalloc(sizeof(struct nfs4_cpntf_state), GFP_KERNEL);
 791        if (!cps)
 792                return NULL;
 793        cps->cpntf_time = ktime_get_boottime_seconds();
 794        refcount_set(&cps->cp_stateid.sc_count, 1);
 795        if (!nfs4_init_cp_state(nn, &cps->cp_stateid, NFS4_COPYNOTIFY_STID))
 796                goto out_free;
 797        spin_lock(&nn->s2s_cp_lock);
 798        list_add(&cps->cp_list, &p_stid->sc_cp_list);
 799        spin_unlock(&nn->s2s_cp_lock);
 800        return cps;
 801out_free:
 802        kfree(cps);
 803        return NULL;
 804}
 805
 806void nfs4_free_copy_state(struct nfsd4_copy *copy)
 807{
 808        struct nfsd_net *nn;
 809
 810        WARN_ON_ONCE(copy->cp_stateid.sc_type != NFS4_COPY_STID);
 811        nn = net_generic(copy->cp_clp->net, nfsd_net_id);
 812        spin_lock(&nn->s2s_cp_lock);
 813        idr_remove(&nn->s2s_cp_stateids,
 814                   copy->cp_stateid.stid.si_opaque.so_id);
 815        spin_unlock(&nn->s2s_cp_lock);
 816}
 817
 818static void nfs4_free_cpntf_statelist(struct net *net, struct nfs4_stid *stid)
 819{
 820        struct nfs4_cpntf_state *cps;
 821        struct nfsd_net *nn;
 822
 823        nn = net_generic(net, nfsd_net_id);
 824        spin_lock(&nn->s2s_cp_lock);
 825        while (!list_empty(&stid->sc_cp_list)) {
 826                cps = list_first_entry(&stid->sc_cp_list,
 827                                       struct nfs4_cpntf_state, cp_list);
 828                _free_cpntf_state_locked(nn, cps);
 829        }
 830        spin_unlock(&nn->s2s_cp_lock);
 831}
 832
 833static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
 834{
 835        struct nfs4_stid *stid;
 836
 837        stid = nfs4_alloc_stid(clp, stateid_slab, nfs4_free_ol_stateid);
 838        if (!stid)
 839                return NULL;
 840
 841        return openlockstateid(stid);
 842}
 843
 844static void nfs4_free_deleg(struct nfs4_stid *stid)
 845{
 846        kmem_cache_free(deleg_slab, stid);
 847        atomic_long_dec(&num_delegations);
 848}
 849
 850/*
 851 * When we recall a delegation, we should be careful not to hand it
 852 * out again straight away.
 853 * To ensure this we keep a pair of bloom filters ('new' and 'old')
 854 * in which the filehandles of recalled delegations are "stored".
 855 * If a filehandle appear in either filter, a delegation is blocked.
 856 * When a delegation is recalled, the filehandle is stored in the "new"
 857 * filter.
 858 * Every 30 seconds we swap the filters and clear the "new" one,
 859 * unless both are empty of course.
 860 *
 861 * Each filter is 256 bits.  We hash the filehandle to 32bit and use the
 862 * low 3 bytes as hash-table indices.
 863 *
 864 * 'blocked_delegations_lock', which is always taken in block_delegations(),
 865 * is used to manage concurrent access.  Testing does not need the lock
 866 * except when swapping the two filters.
 867 */
 868static DEFINE_SPINLOCK(blocked_delegations_lock);
 869static struct bloom_pair {
 870        int     entries, old_entries;
 871        time64_t swap_time;
 872        int     new; /* index into 'set' */
 873        DECLARE_BITMAP(set[2], 256);
 874} blocked_delegations;
 875
 876static int delegation_blocked(struct knfsd_fh *fh)
 877{
 878        u32 hash;
 879        struct bloom_pair *bd = &blocked_delegations;
 880
 881        if (bd->entries == 0)
 882                return 0;
 883        if (ktime_get_seconds() - bd->swap_time > 30) {
 884                spin_lock(&blocked_delegations_lock);
 885                if (ktime_get_seconds() - bd->swap_time > 30) {
 886                        bd->entries -= bd->old_entries;
 887                        bd->old_entries = bd->entries;
 888                        memset(bd->set[bd->new], 0,
 889                               sizeof(bd->set[0]));
 890                        bd->new = 1-bd->new;
 891                        bd->swap_time = ktime_get_seconds();
 892                }
 893                spin_unlock(&blocked_delegations_lock);
 894        }
 895        hash = jhash(&fh->fh_base, fh->fh_size, 0);
 896        if (test_bit(hash&255, bd->set[0]) &&
 897            test_bit((hash>>8)&255, bd->set[0]) &&
 898            test_bit((hash>>16)&255, bd->set[0]))
 899                return 1;
 900
 901        if (test_bit(hash&255, bd->set[1]) &&
 902            test_bit((hash>>8)&255, bd->set[1]) &&
 903            test_bit((hash>>16)&255, bd->set[1]))
 904                return 1;
 905
 906        return 0;
 907}
 908
 909static void block_delegations(struct knfsd_fh *fh)
 910{
 911        u32 hash;
 912        struct bloom_pair *bd = &blocked_delegations;
 913
 914        hash = jhash(&fh->fh_base, fh->fh_size, 0);
 915
 916        spin_lock(&blocked_delegations_lock);
 917        __set_bit(hash&255, bd->set[bd->new]);
 918        __set_bit((hash>>8)&255, bd->set[bd->new]);
 919        __set_bit((hash>>16)&255, bd->set[bd->new]);
 920        if (bd->entries == 0)
 921                bd->swap_time = ktime_get_seconds();
 922        bd->entries += 1;
 923        spin_unlock(&blocked_delegations_lock);
 924}
 925
 926static struct nfs4_delegation *
 927alloc_init_deleg(struct nfs4_client *clp, struct nfs4_file *fp,
 928                 struct svc_fh *current_fh,
 929                 struct nfs4_clnt_odstate *odstate)
 930{
 931        struct nfs4_delegation *dp;
 932        long n;
 933
 934        dprintk("NFSD alloc_init_deleg\n");
 935        n = atomic_long_inc_return(&num_delegations);
 936        if (n < 0 || n > max_delegations)
 937                goto out_dec;
 938        if (delegation_blocked(&current_fh->fh_handle))
 939                goto out_dec;
 940        dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab, nfs4_free_deleg));
 941        if (dp == NULL)
 942                goto out_dec;
 943
 944        /*
 945         * delegation seqid's are never incremented.  The 4.1 special
 946         * meaning of seqid 0 isn't meaningful, really, but let's avoid
 947         * 0 anyway just for consistency and use 1:
 948         */
 949        dp->dl_stid.sc_stateid.si_generation = 1;
 950        INIT_LIST_HEAD(&dp->dl_perfile);
 951        INIT_LIST_HEAD(&dp->dl_perclnt);
 952        INIT_LIST_HEAD(&dp->dl_recall_lru);
 953        dp->dl_clnt_odstate = odstate;
 954        get_clnt_odstate(odstate);
 955        dp->dl_type = NFS4_OPEN_DELEGATE_READ;
 956        dp->dl_retries = 1;
 957        nfsd4_init_cb(&dp->dl_recall, dp->dl_stid.sc_client,
 958                      &nfsd4_cb_recall_ops, NFSPROC4_CLNT_CB_RECALL);
 959        get_nfs4_file(fp);
 960        dp->dl_stid.sc_file = fp;
 961        return dp;
 962out_dec:
 963        atomic_long_dec(&num_delegations);
 964        return NULL;
 965}
 966
 967void
 968nfs4_put_stid(struct nfs4_stid *s)
 969{
 970        struct nfs4_file *fp = s->sc_file;
 971        struct nfs4_client *clp = s->sc_client;
 972
 973        might_lock(&clp->cl_lock);
 974
 975        if (!refcount_dec_and_lock(&s->sc_count, &clp->cl_lock)) {
 976                wake_up_all(&close_wq);
 977                return;
 978        }
 979        idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
 980        nfs4_free_cpntf_statelist(clp->net, s);
 981        spin_unlock(&clp->cl_lock);
 982        s->sc_free(s);
 983        if (fp)
 984                put_nfs4_file(fp);
 985}
 986
 987void
 988nfs4_inc_and_copy_stateid(stateid_t *dst, struct nfs4_stid *stid)
 989{
 990        stateid_t *src = &stid->sc_stateid;
 991
 992        spin_lock(&stid->sc_lock);
 993        if (unlikely(++src->si_generation == 0))
 994                src->si_generation = 1;
 995        memcpy(dst, src, sizeof(*dst));
 996        spin_unlock(&stid->sc_lock);
 997}
 998
 999static void put_deleg_file(struct nfs4_file *fp)
1000{
1001        struct nfsd_file *nf = NULL;
1002
1003        spin_lock(&fp->fi_lock);
1004        if (--fp->fi_delegees == 0)
1005                swap(nf, fp->fi_deleg_file);
1006        spin_unlock(&fp->fi_lock);
1007
1008        if (nf)
1009                nfsd_file_put(nf);
1010}
1011
1012static void nfs4_unlock_deleg_lease(struct nfs4_delegation *dp)
1013{
1014        struct nfs4_file *fp = dp->dl_stid.sc_file;
1015        struct nfsd_file *nf = fp->fi_deleg_file;
1016
1017        WARN_ON_ONCE(!fp->fi_delegees);
1018
1019        vfs_setlease(nf->nf_file, F_UNLCK, NULL, (void **)&dp);
1020        put_deleg_file(fp);
1021}
1022
1023static void destroy_unhashed_deleg(struct nfs4_delegation *dp)
1024{
1025        put_clnt_odstate(dp->dl_clnt_odstate);
1026        nfs4_unlock_deleg_lease(dp);
1027        nfs4_put_stid(&dp->dl_stid);
1028}
1029
1030void nfs4_unhash_stid(struct nfs4_stid *s)
1031{
1032        s->sc_type = 0;
1033}
1034
1035/**
1036 * nfs4_delegation_exists - Discover if this delegation already exists
1037 * @clp:     a pointer to the nfs4_client we're granting a delegation to
1038 * @fp:      a pointer to the nfs4_file we're granting a delegation on
1039 *
1040 * Return:
1041 *      On success: true iff an existing delegation is found
1042 */
1043
1044static bool
1045nfs4_delegation_exists(struct nfs4_client *clp, struct nfs4_file *fp)
1046{
1047        struct nfs4_delegation *searchdp = NULL;
1048        struct nfs4_client *searchclp = NULL;
1049
1050        lockdep_assert_held(&state_lock);
1051        lockdep_assert_held(&fp->fi_lock);
1052
1053        list_for_each_entry(searchdp, &fp->fi_delegations, dl_perfile) {
1054                searchclp = searchdp->dl_stid.sc_client;
1055                if (clp == searchclp) {
1056                        return true;
1057                }
1058        }
1059        return false;
1060}
1061
1062/**
1063 * hash_delegation_locked - Add a delegation to the appropriate lists
1064 * @dp:     a pointer to the nfs4_delegation we are adding.
1065 * @fp:     a pointer to the nfs4_file we're granting a delegation on
1066 *
1067 * Return:
1068 *      On success: NULL if the delegation was successfully hashed.
1069 *
1070 *      On error: -EAGAIN if one was previously granted to this
1071 *                 nfs4_client for this nfs4_file. Delegation is not hashed.
1072 *
1073 */
1074
1075static int
1076hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
1077{
1078        struct nfs4_client *clp = dp->dl_stid.sc_client;
1079
1080        lockdep_assert_held(&state_lock);
1081        lockdep_assert_held(&fp->fi_lock);
1082
1083        if (nfs4_delegation_exists(clp, fp))
1084                return -EAGAIN;
1085        refcount_inc(&dp->dl_stid.sc_count);
1086        dp->dl_stid.sc_type = NFS4_DELEG_STID;
1087        list_add(&dp->dl_perfile, &fp->fi_delegations);
1088        list_add(&dp->dl_perclnt, &clp->cl_delegations);
1089        return 0;
1090}
1091
1092static bool
1093unhash_delegation_locked(struct nfs4_delegation *dp)
1094{
1095        struct nfs4_file *fp = dp->dl_stid.sc_file;
1096
1097        lockdep_assert_held(&state_lock);
1098
1099        if (list_empty(&dp->dl_perfile))
1100                return false;
1101
1102        dp->dl_stid.sc_type = NFS4_CLOSED_DELEG_STID;
1103        /* Ensure that deleg break won't try to requeue it */
1104        ++dp->dl_time;
1105        spin_lock(&fp->fi_lock);
1106        list_del_init(&dp->dl_perclnt);
1107        list_del_init(&dp->dl_recall_lru);
1108        list_del_init(&dp->dl_perfile);
1109        spin_unlock(&fp->fi_lock);
1110        return true;
1111}
1112
1113static void destroy_delegation(struct nfs4_delegation *dp)
1114{
1115        bool unhashed;
1116
1117        spin_lock(&state_lock);
1118        unhashed = unhash_delegation_locked(dp);
1119        spin_unlock(&state_lock);
1120        if (unhashed)
1121                destroy_unhashed_deleg(dp);
1122}
1123
1124static void revoke_delegation(struct nfs4_delegation *dp)
1125{
1126        struct nfs4_client *clp = dp->dl_stid.sc_client;
1127
1128        WARN_ON(!list_empty(&dp->dl_recall_lru));
1129
1130        if (clp->cl_minorversion) {
1131                dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
1132                refcount_inc(&dp->dl_stid.sc_count);
1133                spin_lock(&clp->cl_lock);
1134                list_add(&dp->dl_recall_lru, &clp->cl_revoked);
1135                spin_unlock(&clp->cl_lock);
1136        }
1137        destroy_unhashed_deleg(dp);
1138}
1139
1140/* 
1141 * SETCLIENTID state 
1142 */
1143
1144static unsigned int clientid_hashval(u32 id)
1145{
1146        return id & CLIENT_HASH_MASK;
1147}
1148
1149static unsigned int clientstr_hashval(struct xdr_netobj name)
1150{
1151        return opaque_hashval(name.data, 8) & CLIENT_HASH_MASK;
1152}
1153
1154/*
1155 * We store the NONE, READ, WRITE, and BOTH bits separately in the
1156 * st_{access,deny}_bmap field of the stateid, in order to track not
1157 * only what share bits are currently in force, but also what
1158 * combinations of share bits previous opens have used.  This allows us
1159 * to enforce the recommendation of rfc 3530 14.2.19 that the server
1160 * return an error if the client attempt to downgrade to a combination
1161 * of share bits not explicable by closing some of its previous opens.
1162 *
1163 * XXX: This enforcement is actually incomplete, since we don't keep
1164 * track of access/deny bit combinations; so, e.g., we allow:
1165 *
1166 *      OPEN allow read, deny write
1167 *      OPEN allow both, deny none
1168 *      DOWNGRADE allow read, deny none
1169 *
1170 * which we should reject.
1171 */
1172static unsigned int
1173bmap_to_share_mode(unsigned long bmap) {
1174        int i;
1175        unsigned int access = 0;
1176
1177        for (i = 1; i < 4; i++) {
1178                if (test_bit(i, &bmap))
1179                        access |= i;
1180        }
1181        return access;
1182}
1183
1184/* set share access for a given stateid */
1185static inline void
1186set_access(u32 access, struct nfs4_ol_stateid *stp)
1187{
1188        unsigned char mask = 1 << access;
1189
1190        WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
1191        stp->st_access_bmap |= mask;
1192}
1193
1194/* clear share access for a given stateid */
1195static inline void
1196clear_access(u32 access, struct nfs4_ol_stateid *stp)
1197{
1198        unsigned char mask = 1 << access;
1199
1200        WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
1201        stp->st_access_bmap &= ~mask;
1202}
1203
1204/* test whether a given stateid has access */
1205static inline bool
1206test_access(u32 access, struct nfs4_ol_stateid *stp)
1207{
1208        unsigned char mask = 1 << access;
1209
1210        return (bool)(stp->st_access_bmap & mask);
1211}
1212
1213/* set share deny for a given stateid */
1214static inline void
1215set_deny(u32 deny, struct nfs4_ol_stateid *stp)
1216{
1217        unsigned char mask = 1 << deny;
1218
1219        WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
1220        stp->st_deny_bmap |= mask;
1221}
1222
1223/* clear share deny for a given stateid */
1224static inline void
1225clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
1226{
1227        unsigned char mask = 1 << deny;
1228
1229        WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
1230        stp->st_deny_bmap &= ~mask;
1231}
1232
1233/* test whether a given stateid is denying specific access */
1234static inline bool
1235test_deny(u32 deny, struct nfs4_ol_stateid *stp)
1236{
1237        unsigned char mask = 1 << deny;
1238
1239        return (bool)(stp->st_deny_bmap & mask);
1240}
1241
1242static int nfs4_access_to_omode(u32 access)
1243{
1244        switch (access & NFS4_SHARE_ACCESS_BOTH) {
1245        case NFS4_SHARE_ACCESS_READ:
1246                return O_RDONLY;
1247        case NFS4_SHARE_ACCESS_WRITE:
1248                return O_WRONLY;
1249        case NFS4_SHARE_ACCESS_BOTH:
1250                return O_RDWR;
1251        }
1252        WARN_ON_ONCE(1);
1253        return O_RDONLY;
1254}
1255
1256/*
1257 * A stateid that had a deny mode associated with it is being released
1258 * or downgraded. Recalculate the deny mode on the file.
1259 */
1260static void
1261recalculate_deny_mode(struct nfs4_file *fp)
1262{
1263        struct nfs4_ol_stateid *stp;
1264
1265        spin_lock(&fp->fi_lock);
1266        fp->fi_share_deny = 0;
1267        list_for_each_entry(stp, &fp->fi_stateids, st_perfile)
1268                fp->fi_share_deny |= bmap_to_share_mode(stp->st_deny_bmap);
1269        spin_unlock(&fp->fi_lock);
1270}
1271
1272static void
1273reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
1274{
1275        int i;
1276        bool change = false;
1277
1278        for (i = 1; i < 4; i++) {
1279                if ((i & deny) != i) {
1280                        change = true;
1281                        clear_deny(i, stp);
1282                }
1283        }
1284
1285        /* Recalculate per-file deny mode if there was a change */
1286        if (change)
1287                recalculate_deny_mode(stp->st_stid.sc_file);
1288}
1289
1290/* release all access and file references for a given stateid */
1291static void
1292release_all_access(struct nfs4_ol_stateid *stp)
1293{
1294        int i;
1295        struct nfs4_file *fp = stp->st_stid.sc_file;
1296
1297        if (fp && stp->st_deny_bmap != 0)
1298                recalculate_deny_mode(fp);
1299
1300        for (i = 1; i < 4; i++) {
1301                if (test_access(i, stp))
1302                        nfs4_file_put_access(stp->st_stid.sc_file, i);
1303                clear_access(i, stp);
1304        }
1305}
1306
1307static inline void nfs4_free_stateowner(struct nfs4_stateowner *sop)
1308{
1309        kfree(sop->so_owner.data);
1310        sop->so_ops->so_free(sop);
1311}
1312
1313static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
1314{
1315        struct nfs4_client *clp = sop->so_client;
1316
1317        might_lock(&clp->cl_lock);
1318
1319        if (!atomic_dec_and_lock(&sop->so_count, &clp->cl_lock))
1320                return;
1321        sop->so_ops->so_unhash(sop);
1322        spin_unlock(&clp->cl_lock);
1323        nfs4_free_stateowner(sop);
1324}
1325
1326static bool
1327nfs4_ol_stateid_unhashed(const struct nfs4_ol_stateid *stp)
1328{
1329        return list_empty(&stp->st_perfile);
1330}
1331
1332static bool unhash_ol_stateid(struct nfs4_ol_stateid *stp)
1333{
1334        struct nfs4_file *fp = stp->st_stid.sc_file;
1335
1336        lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
1337
1338        if (list_empty(&stp->st_perfile))
1339                return false;
1340
1341        spin_lock(&fp->fi_lock);
1342        list_del_init(&stp->st_perfile);
1343        spin_unlock(&fp->fi_lock);
1344        list_del(&stp->st_perstateowner);
1345        return true;
1346}
1347
1348static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
1349{
1350        struct nfs4_ol_stateid *stp = openlockstateid(stid);
1351
1352        put_clnt_odstate(stp->st_clnt_odstate);
1353        release_all_access(stp);
1354        if (stp->st_stateowner)
1355                nfs4_put_stateowner(stp->st_stateowner);
1356        kmem_cache_free(stateid_slab, stid);
1357}
1358
1359static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
1360{
1361        struct nfs4_ol_stateid *stp = openlockstateid(stid);
1362        struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
1363        struct nfsd_file *nf;
1364
1365        nf = find_any_file(stp->st_stid.sc_file);
1366        if (nf) {
1367                get_file(nf->nf_file);
1368                filp_close(nf->nf_file, (fl_owner_t)lo);
1369                nfsd_file_put(nf);
1370        }
1371        nfs4_free_ol_stateid(stid);
1372}
1373
1374/*
1375 * Put the persistent reference to an already unhashed generic stateid, while
1376 * holding the cl_lock. If it's the last reference, then put it onto the
1377 * reaplist for later destruction.
1378 */
1379static void put_ol_stateid_locked(struct nfs4_ol_stateid *stp,
1380                                       struct list_head *reaplist)
1381{
1382        struct nfs4_stid *s = &stp->st_stid;
1383        struct nfs4_client *clp = s->sc_client;
1384
1385        lockdep_assert_held(&clp->cl_lock);
1386
1387        WARN_ON_ONCE(!list_empty(&stp->st_locks));
1388
1389        if (!refcount_dec_and_test(&s->sc_count)) {
1390                wake_up_all(&close_wq);
1391                return;
1392        }
1393
1394        idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1395        list_add(&stp->st_locks, reaplist);
1396}
1397
1398static bool unhash_lock_stateid(struct nfs4_ol_stateid *stp)
1399{
1400        lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1401
1402        if (!unhash_ol_stateid(stp))
1403                return false;
1404        list_del_init(&stp->st_locks);
1405        nfs4_unhash_stid(&stp->st_stid);
1406        return true;
1407}
1408
1409static void release_lock_stateid(struct nfs4_ol_stateid *stp)
1410{
1411        struct nfs4_client *clp = stp->st_stid.sc_client;
1412        bool unhashed;
1413
1414        spin_lock(&clp->cl_lock);
1415        unhashed = unhash_lock_stateid(stp);
1416        spin_unlock(&clp->cl_lock);
1417        if (unhashed)
1418                nfs4_put_stid(&stp->st_stid);
1419}
1420
1421static void unhash_lockowner_locked(struct nfs4_lockowner *lo)
1422{
1423        struct nfs4_client *clp = lo->lo_owner.so_client;
1424
1425        lockdep_assert_held(&clp->cl_lock);
1426
1427        list_del_init(&lo->lo_owner.so_strhash);
1428}
1429
1430/*
1431 * Free a list of generic stateids that were collected earlier after being
1432 * fully unhashed.
1433 */
1434static void
1435free_ol_stateid_reaplist(struct list_head *reaplist)
1436{
1437        struct nfs4_ol_stateid *stp;
1438        struct nfs4_file *fp;
1439
1440        might_sleep();
1441
1442        while (!list_empty(reaplist)) {
1443                stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
1444                                       st_locks);
1445                list_del(&stp->st_locks);
1446                fp = stp->st_stid.sc_file;
1447                stp->st_stid.sc_free(&stp->st_stid);
1448                if (fp)
1449                        put_nfs4_file(fp);
1450        }
1451}
1452
1453static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp,
1454                                       struct list_head *reaplist)
1455{
1456        struct nfs4_ol_stateid *stp;
1457
1458        lockdep_assert_held(&open_stp->st_stid.sc_client->cl_lock);
1459
1460        while (!list_empty(&open_stp->st_locks)) {
1461                stp = list_entry(open_stp->st_locks.next,
1462                                struct nfs4_ol_stateid, st_locks);
1463                WARN_ON(!unhash_lock_stateid(stp));
1464                put_ol_stateid_locked(stp, reaplist);
1465        }
1466}
1467
1468static bool unhash_open_stateid(struct nfs4_ol_stateid *stp,
1469                                struct list_head *reaplist)
1470{
1471        lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1472
1473        if (!unhash_ol_stateid(stp))
1474                return false;
1475        release_open_stateid_locks(stp, reaplist);
1476        return true;
1477}
1478
1479static void release_open_stateid(struct nfs4_ol_stateid *stp)
1480{
1481        LIST_HEAD(reaplist);
1482
1483        spin_lock(&stp->st_stid.sc_client->cl_lock);
1484        if (unhash_open_stateid(stp, &reaplist))
1485                put_ol_stateid_locked(stp, &reaplist);
1486        spin_unlock(&stp->st_stid.sc_client->cl_lock);
1487        free_ol_stateid_reaplist(&reaplist);
1488}
1489
1490static void unhash_openowner_locked(struct nfs4_openowner *oo)
1491{
1492        struct nfs4_client *clp = oo->oo_owner.so_client;
1493
1494        lockdep_assert_held(&clp->cl_lock);
1495
1496        list_del_init(&oo->oo_owner.so_strhash);
1497        list_del_init(&oo->oo_perclient);
1498}
1499
1500static void release_last_closed_stateid(struct nfs4_openowner *oo)
1501{
1502        struct nfsd_net *nn = net_generic(oo->oo_owner.so_client->net,
1503                                          nfsd_net_id);
1504        struct nfs4_ol_stateid *s;
1505
1506        spin_lock(&nn->client_lock);
1507        s = oo->oo_last_closed_stid;
1508        if (s) {
1509                list_del_init(&oo->oo_close_lru);
1510                oo->oo_last_closed_stid = NULL;
1511        }
1512        spin_unlock(&nn->client_lock);
1513        if (s)
1514                nfs4_put_stid(&s->st_stid);
1515}
1516
1517static void release_openowner(struct nfs4_openowner *oo)
1518{
1519        struct nfs4_ol_stateid *stp;
1520        struct nfs4_client *clp = oo->oo_owner.so_client;
1521        struct list_head reaplist;
1522
1523        INIT_LIST_HEAD(&reaplist);
1524
1525        spin_lock(&clp->cl_lock);
1526        unhash_openowner_locked(oo);
1527        while (!list_empty(&oo->oo_owner.so_stateids)) {
1528                stp = list_first_entry(&oo->oo_owner.so_stateids,
1529                                struct nfs4_ol_stateid, st_perstateowner);
1530                if (unhash_open_stateid(stp, &reaplist))
1531                        put_ol_stateid_locked(stp, &reaplist);
1532        }
1533        spin_unlock(&clp->cl_lock);
1534        free_ol_stateid_reaplist(&reaplist);
1535        release_last_closed_stateid(oo);
1536        nfs4_put_stateowner(&oo->oo_owner);
1537}
1538
1539static inline int
1540hash_sessionid(struct nfs4_sessionid *sessionid)
1541{
1542        struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1543
1544        return sid->sequence % SESSION_HASH_SIZE;
1545}
1546
1547#ifdef CONFIG_SUNRPC_DEBUG
1548static inline void
1549dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1550{
1551        u32 *ptr = (u32 *)(&sessionid->data[0]);
1552        dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1553}
1554#else
1555static inline void
1556dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1557{
1558}
1559#endif
1560
1561/*
1562 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1563 * won't be used for replay.
1564 */
1565void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1566{
1567        struct nfs4_stateowner *so = cstate->replay_owner;
1568
1569        if (nfserr == nfserr_replay_me)
1570                return;
1571
1572        if (!seqid_mutating_err(ntohl(nfserr))) {
1573                nfsd4_cstate_clear_replay(cstate);
1574                return;
1575        }
1576        if (!so)
1577                return;
1578        if (so->so_is_open_owner)
1579                release_last_closed_stateid(openowner(so));
1580        so->so_seqid++;
1581        return;
1582}
1583
1584static void
1585gen_sessionid(struct nfsd4_session *ses)
1586{
1587        struct nfs4_client *clp = ses->se_client;
1588        struct nfsd4_sessionid *sid;
1589
1590        sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1591        sid->clientid = clp->cl_clientid;
1592        sid->sequence = current_sessionid++;
1593        sid->reserved = 0;
1594}
1595
1596/*
1597 * The protocol defines ca_maxresponssize_cached to include the size of
1598 * the rpc header, but all we need to cache is the data starting after
1599 * the end of the initial SEQUENCE operation--the rest we regenerate
1600 * each time.  Therefore we can advertise a ca_maxresponssize_cached
1601 * value that is the number of bytes in our cache plus a few additional
1602 * bytes.  In order to stay on the safe side, and not promise more than
1603 * we can cache, those additional bytes must be the minimum possible: 24
1604 * bytes of rpc header (xid through accept state, with AUTH_NULL
1605 * verifier), 12 for the compound header (with zero-length tag), and 44
1606 * for the SEQUENCE op response:
1607 */
1608#define NFSD_MIN_HDR_SEQ_SZ  (24 + 12 + 44)
1609
1610static void
1611free_session_slots(struct nfsd4_session *ses)
1612{
1613        int i;
1614
1615        for (i = 0; i < ses->se_fchannel.maxreqs; i++) {
1616                free_svc_cred(&ses->se_slots[i]->sl_cred);
1617                kfree(ses->se_slots[i]);
1618        }
1619}
1620
1621/*
1622 * We don't actually need to cache the rpc and session headers, so we
1623 * can allocate a little less for each slot:
1624 */
1625static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
1626{
1627        u32 size;
1628
1629        if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
1630                size = 0;
1631        else
1632                size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
1633        return size + sizeof(struct nfsd4_slot);
1634}
1635
1636/*
1637 * XXX: If we run out of reserved DRC memory we could (up to a point)
1638 * re-negotiate active sessions and reduce their slot usage to make
1639 * room for new connections. For now we just fail the create session.
1640 */
1641static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
1642{
1643        u32 slotsize = slot_bytes(ca);
1644        u32 num = ca->maxreqs;
1645        unsigned long avail, total_avail;
1646        unsigned int scale_factor;
1647
1648        spin_lock(&nfsd_drc_lock);
1649        if (nfsd_drc_max_mem > nfsd_drc_mem_used)
1650                total_avail = nfsd_drc_max_mem - nfsd_drc_mem_used;
1651        else
1652                /* We have handed out more space than we chose in
1653                 * set_max_drc() to allow.  That isn't really a
1654                 * problem as long as that doesn't make us think we
1655                 * have lots more due to integer overflow.
1656                 */
1657                total_avail = 0;
1658        avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION, total_avail);
1659        /*
1660         * Never use more than a fraction of the remaining memory,
1661         * unless it's the only way to give this client a slot.
1662         * The chosen fraction is either 1/8 or 1/number of threads,
1663         * whichever is smaller.  This ensures there are adequate
1664         * slots to support multiple clients per thread.
1665         * Give the client one slot even if that would require
1666         * over-allocation--it is better than failure.
1667         */
1668        scale_factor = max_t(unsigned int, 8, nn->nfsd_serv->sv_nrthreads);
1669
1670        avail = clamp_t(unsigned long, avail, slotsize,
1671                        total_avail/scale_factor);
1672        num = min_t(int, num, avail / slotsize);
1673        num = max_t(int, num, 1);
1674        nfsd_drc_mem_used += num * slotsize;
1675        spin_unlock(&nfsd_drc_lock);
1676
1677        return num;
1678}
1679
1680static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
1681{
1682        int slotsize = slot_bytes(ca);
1683
1684        spin_lock(&nfsd_drc_lock);
1685        nfsd_drc_mem_used -= slotsize * ca->maxreqs;
1686        spin_unlock(&nfsd_drc_lock);
1687}
1688
1689static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
1690                                           struct nfsd4_channel_attrs *battrs)
1691{
1692        int numslots = fattrs->maxreqs;
1693        int slotsize = slot_bytes(fattrs);
1694        struct nfsd4_session *new;
1695        int mem, i;
1696
1697        BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
1698                        + sizeof(struct nfsd4_session) > PAGE_SIZE);
1699        mem = numslots * sizeof(struct nfsd4_slot *);
1700
1701        new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
1702        if (!new)
1703                return NULL;
1704        /* allocate each struct nfsd4_slot and data cache in one piece */
1705        for (i = 0; i < numslots; i++) {
1706                new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
1707                if (!new->se_slots[i])
1708                        goto out_free;
1709        }
1710
1711        memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
1712        memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
1713
1714        return new;
1715out_free:
1716        while (i--)
1717                kfree(new->se_slots[i]);
1718        kfree(new);
1719        return NULL;
1720}
1721
1722static void free_conn(struct nfsd4_conn *c)
1723{
1724        svc_xprt_put(c->cn_xprt);
1725        kfree(c);
1726}
1727
1728static void nfsd4_conn_lost(struct svc_xpt_user *u)
1729{
1730        struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
1731        struct nfs4_client *clp = c->cn_session->se_client;
1732
1733        spin_lock(&clp->cl_lock);
1734        if (!list_empty(&c->cn_persession)) {
1735                list_del(&c->cn_persession);
1736                free_conn(c);
1737        }
1738        nfsd4_probe_callback(clp);
1739        spin_unlock(&clp->cl_lock);
1740}
1741
1742static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1743{
1744        struct nfsd4_conn *conn;
1745
1746        conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1747        if (!conn)
1748                return NULL;
1749        svc_xprt_get(rqstp->rq_xprt);
1750        conn->cn_xprt = rqstp->rq_xprt;
1751        conn->cn_flags = flags;
1752        INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1753        return conn;
1754}
1755
1756static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1757{
1758        conn->cn_session = ses;
1759        list_add(&conn->cn_persession, &ses->se_conns);
1760}
1761
1762static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1763{
1764        struct nfs4_client *clp = ses->se_client;
1765
1766        spin_lock(&clp->cl_lock);
1767        __nfsd4_hash_conn(conn, ses);
1768        spin_unlock(&clp->cl_lock);
1769}
1770
1771static int nfsd4_register_conn(struct nfsd4_conn *conn)
1772{
1773        conn->cn_xpt_user.callback = nfsd4_conn_lost;
1774        return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1775}
1776
1777static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1778{
1779        int ret;
1780
1781        nfsd4_hash_conn(conn, ses);
1782        ret = nfsd4_register_conn(conn);
1783        if (ret)
1784                /* oops; xprt is already down: */
1785                nfsd4_conn_lost(&conn->cn_xpt_user);
1786        /* We may have gained or lost a callback channel: */
1787        nfsd4_probe_callback_sync(ses->se_client);
1788}
1789
1790static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1791{
1792        u32 dir = NFS4_CDFC4_FORE;
1793
1794        if (cses->flags & SESSION4_BACK_CHAN)
1795                dir |= NFS4_CDFC4_BACK;
1796        return alloc_conn(rqstp, dir);
1797}
1798
1799/* must be called under client_lock */
1800static void nfsd4_del_conns(struct nfsd4_session *s)
1801{
1802        struct nfs4_client *clp = s->se_client;
1803        struct nfsd4_conn *c;
1804
1805        spin_lock(&clp->cl_lock);
1806        while (!list_empty(&s->se_conns)) {
1807                c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1808                list_del_init(&c->cn_persession);
1809                spin_unlock(&clp->cl_lock);
1810
1811                unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1812                free_conn(c);
1813
1814                spin_lock(&clp->cl_lock);
1815        }
1816        spin_unlock(&clp->cl_lock);
1817}
1818
1819static void __free_session(struct nfsd4_session *ses)
1820{
1821        free_session_slots(ses);
1822        kfree(ses);
1823}
1824
1825static void free_session(struct nfsd4_session *ses)
1826{
1827        nfsd4_del_conns(ses);
1828        nfsd4_put_drc_mem(&ses->se_fchannel);
1829        __free_session(ses);
1830}
1831
1832static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1833{
1834        int idx;
1835        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1836
1837        new->se_client = clp;
1838        gen_sessionid(new);
1839
1840        INIT_LIST_HEAD(&new->se_conns);
1841
1842        new->se_cb_seq_nr = 1;
1843        new->se_flags = cses->flags;
1844        new->se_cb_prog = cses->callback_prog;
1845        new->se_cb_sec = cses->cb_sec;
1846        atomic_set(&new->se_ref, 0);
1847        idx = hash_sessionid(&new->se_sessionid);
1848        list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1849        spin_lock(&clp->cl_lock);
1850        list_add(&new->se_perclnt, &clp->cl_sessions);
1851        spin_unlock(&clp->cl_lock);
1852
1853        {
1854                struct sockaddr *sa = svc_addr(rqstp);
1855                /*
1856                 * This is a little silly; with sessions there's no real
1857                 * use for the callback address.  Use the peer address
1858                 * as a reasonable default for now, but consider fixing
1859                 * the rpc client not to require an address in the
1860                 * future:
1861                 */
1862                rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1863                clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1864        }
1865}
1866
1867/* caller must hold client_lock */
1868static struct nfsd4_session *
1869__find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1870{
1871        struct nfsd4_session *elem;
1872        int idx;
1873        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1874
1875        lockdep_assert_held(&nn->client_lock);
1876
1877        dump_sessionid(__func__, sessionid);
1878        idx = hash_sessionid(sessionid);
1879        /* Search in the appropriate list */
1880        list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1881                if (!memcmp(elem->se_sessionid.data, sessionid->data,
1882                            NFS4_MAX_SESSIONID_LEN)) {
1883                        return elem;
1884                }
1885        }
1886
1887        dprintk("%s: session not found\n", __func__);
1888        return NULL;
1889}
1890
1891static struct nfsd4_session *
1892find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
1893                __be32 *ret)
1894{
1895        struct nfsd4_session *session;
1896        __be32 status = nfserr_badsession;
1897
1898        session = __find_in_sessionid_hashtbl(sessionid, net);
1899        if (!session)
1900                goto out;
1901        status = nfsd4_get_session_locked(session);
1902        if (status)
1903                session = NULL;
1904out:
1905        *ret = status;
1906        return session;
1907}
1908
1909/* caller must hold client_lock */
1910static void
1911unhash_session(struct nfsd4_session *ses)
1912{
1913        struct nfs4_client *clp = ses->se_client;
1914        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1915
1916        lockdep_assert_held(&nn->client_lock);
1917
1918        list_del(&ses->se_hash);
1919        spin_lock(&ses->se_client->cl_lock);
1920        list_del(&ses->se_perclnt);
1921        spin_unlock(&ses->se_client->cl_lock);
1922}
1923
1924/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1925static int
1926STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1927{
1928        /*
1929         * We're assuming the clid was not given out from a boot
1930         * precisely 2^32 (about 136 years) before this one.  That seems
1931         * a safe assumption:
1932         */
1933        if (clid->cl_boot == (u32)nn->boot_time)
1934                return 0;
1935        trace_nfsd_clid_stale(clid);
1936        return 1;
1937}
1938
1939/* 
1940 * XXX Should we use a slab cache ?
1941 * This type of memory management is somewhat inefficient, but we use it
1942 * anyway since SETCLIENTID is not a common operation.
1943 */
1944static struct nfs4_client *alloc_client(struct xdr_netobj name)
1945{
1946        struct nfs4_client *clp;
1947        int i;
1948
1949        clp = kmem_cache_zalloc(client_slab, GFP_KERNEL);
1950        if (clp == NULL)
1951                return NULL;
1952        xdr_netobj_dup(&clp->cl_name, &name, GFP_KERNEL);
1953        if (clp->cl_name.data == NULL)
1954                goto err_no_name;
1955        clp->cl_ownerstr_hashtbl = kmalloc_array(OWNER_HASH_SIZE,
1956                                                 sizeof(struct list_head),
1957                                                 GFP_KERNEL);
1958        if (!clp->cl_ownerstr_hashtbl)
1959                goto err_no_hashtbl;
1960        for (i = 0; i < OWNER_HASH_SIZE; i++)
1961                INIT_LIST_HEAD(&clp->cl_ownerstr_hashtbl[i]);
1962        INIT_LIST_HEAD(&clp->cl_sessions);
1963        idr_init(&clp->cl_stateids);
1964        atomic_set(&clp->cl_rpc_users, 0);
1965        clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1966        INIT_LIST_HEAD(&clp->cl_idhash);
1967        INIT_LIST_HEAD(&clp->cl_openowners);
1968        INIT_LIST_HEAD(&clp->cl_delegations);
1969        INIT_LIST_HEAD(&clp->cl_lru);
1970        INIT_LIST_HEAD(&clp->cl_revoked);
1971#ifdef CONFIG_NFSD_PNFS
1972        INIT_LIST_HEAD(&clp->cl_lo_states);
1973#endif
1974        INIT_LIST_HEAD(&clp->async_copies);
1975        spin_lock_init(&clp->async_lock);
1976        spin_lock_init(&clp->cl_lock);
1977        rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1978        return clp;
1979err_no_hashtbl:
1980        kfree(clp->cl_name.data);
1981err_no_name:
1982        kmem_cache_free(client_slab, clp);
1983        return NULL;
1984}
1985
1986static void __free_client(struct kref *k)
1987{
1988        struct nfsdfs_client *c = container_of(k, struct nfsdfs_client, cl_ref);
1989        struct nfs4_client *clp = container_of(c, struct nfs4_client, cl_nfsdfs);
1990
1991        free_svc_cred(&clp->cl_cred);
1992        kfree(clp->cl_ownerstr_hashtbl);
1993        kfree(clp->cl_name.data);
1994        kfree(clp->cl_nii_domain.data);
1995        kfree(clp->cl_nii_name.data);
1996        idr_destroy(&clp->cl_stateids);
1997        kmem_cache_free(client_slab, clp);
1998}
1999
2000static void drop_client(struct nfs4_client *clp)
2001{
2002        kref_put(&clp->cl_nfsdfs.cl_ref, __free_client);
2003}
2004
2005static void
2006free_client(struct nfs4_client *clp)
2007{
2008        while (!list_empty(&clp->cl_sessions)) {
2009                struct nfsd4_session *ses;
2010                ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
2011                                se_perclnt);
2012                list_del(&ses->se_perclnt);
2013                WARN_ON_ONCE(atomic_read(&ses->se_ref));
2014                free_session(ses);
2015        }
2016        rpc_destroy_wait_queue(&clp->cl_cb_waitq);
2017        if (clp->cl_nfsd_dentry) {
2018                nfsd_client_rmdir(clp->cl_nfsd_dentry);
2019                clp->cl_nfsd_dentry = NULL;
2020                wake_up_all(&expiry_wq);
2021        }
2022        drop_client(clp);
2023}
2024
2025/* must be called under the client_lock */
2026static void
2027unhash_client_locked(struct nfs4_client *clp)
2028{
2029        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2030        struct nfsd4_session *ses;
2031
2032        lockdep_assert_held(&nn->client_lock);
2033
2034        /* Mark the client as expired! */
2035        clp->cl_time = 0;
2036        /* Make it invisible */
2037        if (!list_empty(&clp->cl_idhash)) {
2038                list_del_init(&clp->cl_idhash);
2039                if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
2040                        rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
2041                else
2042                        rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2043        }
2044        list_del_init(&clp->cl_lru);
2045        spin_lock(&clp->cl_lock);
2046        list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
2047                list_del_init(&ses->se_hash);
2048        spin_unlock(&clp->cl_lock);
2049}
2050
2051static void
2052unhash_client(struct nfs4_client *clp)
2053{
2054        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2055
2056        spin_lock(&nn->client_lock);
2057        unhash_client_locked(clp);
2058        spin_unlock(&nn->client_lock);
2059}
2060
2061static __be32 mark_client_expired_locked(struct nfs4_client *clp)
2062{
2063        if (atomic_read(&clp->cl_rpc_users))
2064                return nfserr_jukebox;
2065        unhash_client_locked(clp);
2066        return nfs_ok;
2067}
2068
2069static void
2070__destroy_client(struct nfs4_client *clp)
2071{
2072        int i;
2073        struct nfs4_openowner *oo;
2074        struct nfs4_delegation *dp;
2075        struct list_head reaplist;
2076
2077        INIT_LIST_HEAD(&reaplist);
2078        spin_lock(&state_lock);
2079        while (!list_empty(&clp->cl_delegations)) {
2080                dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
2081                WARN_ON(!unhash_delegation_locked(dp));
2082                list_add(&dp->dl_recall_lru, &reaplist);
2083        }
2084        spin_unlock(&state_lock);
2085        while (!list_empty(&reaplist)) {
2086                dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
2087                list_del_init(&dp->dl_recall_lru);
2088                destroy_unhashed_deleg(dp);
2089        }
2090        while (!list_empty(&clp->cl_revoked)) {
2091                dp = list_entry(clp->cl_revoked.next, struct nfs4_delegation, dl_recall_lru);
2092                list_del_init(&dp->dl_recall_lru);
2093                nfs4_put_stid(&dp->dl_stid);
2094        }
2095        while (!list_empty(&clp->cl_openowners)) {
2096                oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
2097                nfs4_get_stateowner(&oo->oo_owner);
2098                release_openowner(oo);
2099        }
2100        for (i = 0; i < OWNER_HASH_SIZE; i++) {
2101                struct nfs4_stateowner *so, *tmp;
2102
2103                list_for_each_entry_safe(so, tmp, &clp->cl_ownerstr_hashtbl[i],
2104                                         so_strhash) {
2105                        /* Should be no openowners at this point */
2106                        WARN_ON_ONCE(so->so_is_open_owner);
2107                        remove_blocked_locks(lockowner(so));
2108                }
2109        }
2110        nfsd4_return_all_client_layouts(clp);
2111        nfsd4_shutdown_copy(clp);
2112        nfsd4_shutdown_callback(clp);
2113        if (clp->cl_cb_conn.cb_xprt)
2114                svc_xprt_put(clp->cl_cb_conn.cb_xprt);
2115        free_client(clp);
2116        wake_up_all(&expiry_wq);
2117}
2118
2119static void
2120destroy_client(struct nfs4_client *clp)
2121{
2122        unhash_client(clp);
2123        __destroy_client(clp);
2124}
2125
2126static void inc_reclaim_complete(struct nfs4_client *clp)
2127{
2128        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2129
2130        if (!nn->track_reclaim_completes)
2131                return;
2132        if (!nfsd4_find_reclaim_client(clp->cl_name, nn))
2133                return;
2134        if (atomic_inc_return(&nn->nr_reclaim_complete) ==
2135                        nn->reclaim_str_hashtbl_size) {
2136                printk(KERN_INFO "NFSD: all clients done reclaiming, ending NFSv4 grace period (net %x)\n",
2137                                clp->net->ns.inum);
2138                nfsd4_end_grace(nn);
2139        }
2140}
2141
2142static void expire_client(struct nfs4_client *clp)
2143{
2144        unhash_client(clp);
2145        nfsd4_client_record_remove(clp);
2146        __destroy_client(clp);
2147}
2148
2149static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
2150{
2151        memcpy(target->cl_verifier.data, source->data,
2152                        sizeof(target->cl_verifier.data));
2153}
2154
2155static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
2156{
2157        target->cl_clientid.cl_boot = source->cl_clientid.cl_boot; 
2158        target->cl_clientid.cl_id = source->cl_clientid.cl_id; 
2159}
2160
2161static int copy_cred(struct svc_cred *target, struct svc_cred *source)
2162{
2163        target->cr_principal = kstrdup(source->cr_principal, GFP_KERNEL);
2164        target->cr_raw_principal = kstrdup(source->cr_raw_principal,
2165                                                                GFP_KERNEL);
2166        target->cr_targ_princ = kstrdup(source->cr_targ_princ, GFP_KERNEL);
2167        if ((source->cr_principal && !target->cr_principal) ||
2168            (source->cr_raw_principal && !target->cr_raw_principal) ||
2169            (source->cr_targ_princ && !target->cr_targ_princ))
2170                return -ENOMEM;
2171
2172        target->cr_flavor = source->cr_flavor;
2173        target->cr_uid = source->cr_uid;
2174        target->cr_gid = source->cr_gid;
2175        target->cr_group_info = source->cr_group_info;
2176        get_group_info(target->cr_group_info);
2177        target->cr_gss_mech = source->cr_gss_mech;
2178        if (source->cr_gss_mech)
2179                gss_mech_get(source->cr_gss_mech);
2180        return 0;
2181}
2182
2183static int
2184compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
2185{
2186        if (o1->len < o2->len)
2187                return -1;
2188        if (o1->len > o2->len)
2189                return 1;
2190        return memcmp(o1->data, o2->data, o1->len);
2191}
2192
2193static int
2194same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
2195{
2196        return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
2197}
2198
2199static int
2200same_clid(clientid_t *cl1, clientid_t *cl2)
2201{
2202        return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
2203}
2204
2205static bool groups_equal(struct group_info *g1, struct group_info *g2)
2206{
2207        int i;
2208
2209        if (g1->ngroups != g2->ngroups)
2210                return false;
2211        for (i=0; i<g1->ngroups; i++)
2212                if (!gid_eq(g1->gid[i], g2->gid[i]))
2213                        return false;
2214        return true;
2215}
2216
2217/*
2218 * RFC 3530 language requires clid_inuse be returned when the
2219 * "principal" associated with a requests differs from that previously
2220 * used.  We use uid, gid's, and gss principal string as our best
2221 * approximation.  We also don't want to allow non-gss use of a client
2222 * established using gss: in theory cr_principal should catch that
2223 * change, but in practice cr_principal can be null even in the gss case
2224 * since gssd doesn't always pass down a principal string.
2225 */
2226static bool is_gss_cred(struct svc_cred *cr)
2227{
2228        /* Is cr_flavor one of the gss "pseudoflavors"?: */
2229        return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
2230}
2231
2232
2233static bool
2234same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
2235{
2236        if ((is_gss_cred(cr1) != is_gss_cred(cr2))
2237                || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
2238                || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
2239                || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
2240                return false;
2241        /* XXX: check that cr_targ_princ fields match ? */
2242        if (cr1->cr_principal == cr2->cr_principal)
2243                return true;
2244        if (!cr1->cr_principal || !cr2->cr_principal)
2245                return false;
2246        return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
2247}
2248
2249static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
2250{
2251        struct svc_cred *cr = &rqstp->rq_cred;
2252        u32 service;
2253
2254        if (!cr->cr_gss_mech)
2255                return false;
2256        service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
2257        return service == RPC_GSS_SVC_INTEGRITY ||
2258               service == RPC_GSS_SVC_PRIVACY;
2259}
2260
2261bool nfsd4_mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
2262{
2263        struct svc_cred *cr = &rqstp->rq_cred;
2264
2265        if (!cl->cl_mach_cred)
2266                return true;
2267        if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
2268                return false;
2269        if (!svc_rqst_integrity_protected(rqstp))
2270                return false;
2271        if (cl->cl_cred.cr_raw_principal)
2272                return 0 == strcmp(cl->cl_cred.cr_raw_principal,
2273                                                cr->cr_raw_principal);
2274        if (!cr->cr_principal)
2275                return false;
2276        return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
2277}
2278
2279static void gen_confirm(struct nfs4_client *clp, struct nfsd_net *nn)
2280{
2281        __be32 verf[2];
2282
2283        /*
2284         * This is opaque to client, so no need to byte-swap. Use
2285         * __force to keep sparse happy
2286         */
2287        verf[0] = (__force __be32)(u32)ktime_get_real_seconds();
2288        verf[1] = (__force __be32)nn->clverifier_counter++;
2289        memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
2290}
2291
2292static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
2293{
2294        clp->cl_clientid.cl_boot = (u32)nn->boot_time;
2295        clp->cl_clientid.cl_id = nn->clientid_counter++;
2296        gen_confirm(clp, nn);
2297}
2298
2299static struct nfs4_stid *
2300find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
2301{
2302        struct nfs4_stid *ret;
2303
2304        ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
2305        if (!ret || !ret->sc_type)
2306                return NULL;
2307        return ret;
2308}
2309
2310static struct nfs4_stid *
2311find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
2312{
2313        struct nfs4_stid *s;
2314
2315        spin_lock(&cl->cl_lock);
2316        s = find_stateid_locked(cl, t);
2317        if (s != NULL) {
2318                if (typemask & s->sc_type)
2319                        refcount_inc(&s->sc_count);
2320                else
2321                        s = NULL;
2322        }
2323        spin_unlock(&cl->cl_lock);
2324        return s;
2325}
2326
2327static struct nfs4_client *get_nfsdfs_clp(struct inode *inode)
2328{
2329        struct nfsdfs_client *nc;
2330        nc = get_nfsdfs_client(inode);
2331        if (!nc)
2332                return NULL;
2333        return container_of(nc, struct nfs4_client, cl_nfsdfs);
2334}
2335
2336static void seq_quote_mem(struct seq_file *m, char *data, int len)
2337{
2338        seq_printf(m, "\"");
2339        seq_escape_mem_ascii(m, data, len);
2340        seq_printf(m, "\"");
2341}
2342
2343static int client_info_show(struct seq_file *m, void *v)
2344{
2345        struct inode *inode = m->private;
2346        struct nfs4_client *clp;
2347        u64 clid;
2348
2349        clp = get_nfsdfs_clp(inode);
2350        if (!clp)
2351                return -ENXIO;
2352        memcpy(&clid, &clp->cl_clientid, sizeof(clid));
2353        seq_printf(m, "clientid: 0x%llx\n", clid);
2354        seq_printf(m, "address: \"%pISpc\"\n", (struct sockaddr *)&clp->cl_addr);
2355        seq_printf(m, "name: ");
2356        seq_quote_mem(m, clp->cl_name.data, clp->cl_name.len);
2357        seq_printf(m, "\nminor version: %d\n", clp->cl_minorversion);
2358        if (clp->cl_nii_domain.data) {
2359                seq_printf(m, "Implementation domain: ");
2360                seq_quote_mem(m, clp->cl_nii_domain.data,
2361                                        clp->cl_nii_domain.len);
2362                seq_printf(m, "\nImplementation name: ");
2363                seq_quote_mem(m, clp->cl_nii_name.data, clp->cl_nii_name.len);
2364                seq_printf(m, "\nImplementation time: [%lld, %ld]\n",
2365                        clp->cl_nii_time.tv_sec, clp->cl_nii_time.tv_nsec);
2366        }
2367        drop_client(clp);
2368
2369        return 0;
2370}
2371
2372static int client_info_open(struct inode *inode, struct file *file)
2373{
2374        return single_open(file, client_info_show, inode);
2375}
2376
2377static const struct file_operations client_info_fops = {
2378        .open           = client_info_open,
2379        .read           = seq_read,
2380        .llseek         = seq_lseek,
2381        .release        = single_release,
2382};
2383
2384static void *states_start(struct seq_file *s, loff_t *pos)
2385        __acquires(&clp->cl_lock)
2386{
2387        struct nfs4_client *clp = s->private;
2388        unsigned long id = *pos;
2389        void *ret;
2390
2391        spin_lock(&clp->cl_lock);
2392        ret = idr_get_next_ul(&clp->cl_stateids, &id);
2393        *pos = id;
2394        return ret;
2395}
2396
2397static void *states_next(struct seq_file *s, void *v, loff_t *pos)
2398{
2399        struct nfs4_client *clp = s->private;
2400        unsigned long id = *pos;
2401        void *ret;
2402
2403        id = *pos;
2404        id++;
2405        ret = idr_get_next_ul(&clp->cl_stateids, &id);
2406        *pos = id;
2407        return ret;
2408}
2409
2410static void states_stop(struct seq_file *s, void *v)
2411        __releases(&clp->cl_lock)
2412{
2413        struct nfs4_client *clp = s->private;
2414
2415        spin_unlock(&clp->cl_lock);
2416}
2417
2418static void nfs4_show_fname(struct seq_file *s, struct nfsd_file *f)
2419{
2420         seq_printf(s, "filename: \"%pD2\"", f->nf_file);
2421}
2422
2423static void nfs4_show_superblock(struct seq_file *s, struct nfsd_file *f)
2424{
2425        struct inode *inode = f->nf_inode;
2426
2427        seq_printf(s, "superblock: \"%02x:%02x:%ld\"",
2428                                        MAJOR(inode->i_sb->s_dev),
2429                                         MINOR(inode->i_sb->s_dev),
2430                                         inode->i_ino);
2431}
2432
2433static void nfs4_show_owner(struct seq_file *s, struct nfs4_stateowner *oo)
2434{
2435        seq_printf(s, "owner: ");
2436        seq_quote_mem(s, oo->so_owner.data, oo->so_owner.len);
2437}
2438
2439static void nfs4_show_stateid(struct seq_file *s, stateid_t *stid)
2440{
2441        seq_printf(s, "0x%.8x", stid->si_generation);
2442        seq_printf(s, "%12phN", &stid->si_opaque);
2443}
2444
2445static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
2446{
2447        struct nfs4_ol_stateid *ols;
2448        struct nfs4_file *nf;
2449        struct nfsd_file *file;
2450        struct nfs4_stateowner *oo;
2451        unsigned int access, deny;
2452
2453        if (st->sc_type != NFS4_OPEN_STID && st->sc_type != NFS4_LOCK_STID)
2454                return 0; /* XXX: or SEQ_SKIP? */
2455        ols = openlockstateid(st);
2456        oo = ols->st_stateowner;
2457        nf = st->sc_file;
2458        file = find_any_file(nf);
2459        if (!file)
2460                return 0;
2461
2462        seq_printf(s, "- ");
2463        nfs4_show_stateid(s, &st->sc_stateid);
2464        seq_printf(s, ": { type: open, ");
2465
2466        access = bmap_to_share_mode(ols->st_access_bmap);
2467        deny   = bmap_to_share_mode(ols->st_deny_bmap);
2468
2469        seq_printf(s, "access: %s%s, ",
2470                access & NFS4_SHARE_ACCESS_READ ? "r" : "-",
2471                access & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
2472        seq_printf(s, "deny: %s%s, ",
2473                deny & NFS4_SHARE_ACCESS_READ ? "r" : "-",
2474                deny & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
2475
2476        nfs4_show_superblock(s, file);
2477        seq_printf(s, ", ");
2478        nfs4_show_fname(s, file);
2479        seq_printf(s, ", ");
2480        nfs4_show_owner(s, oo);
2481        seq_printf(s, " }\n");
2482        nfsd_file_put(file);
2483
2484        return 0;
2485}
2486
2487static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
2488{
2489        struct nfs4_ol_stateid *ols;
2490        struct nfs4_file *nf;
2491        struct nfsd_file *file;
2492        struct nfs4_stateowner *oo;
2493
2494        ols = openlockstateid(st);
2495        oo = ols->st_stateowner;
2496        nf = st->sc_file;
2497        file = find_any_file(nf);
2498        if (!file)
2499                return 0;
2500
2501        seq_printf(s, "- ");
2502        nfs4_show_stateid(s, &st->sc_stateid);
2503        seq_printf(s, ": { type: lock, ");
2504
2505        /*
2506         * Note: a lock stateid isn't really the same thing as a lock,
2507         * it's the locking state held by one owner on a file, and there
2508         * may be multiple (or no) lock ranges associated with it.
2509         * (Same for the matter is true of open stateids.)
2510         */
2511
2512        nfs4_show_superblock(s, file);
2513        /* XXX: open stateid? */
2514        seq_printf(s, ", ");
2515        nfs4_show_fname(s, file);
2516        seq_printf(s, ", ");
2517        nfs4_show_owner(s, oo);
2518        seq_printf(s, " }\n");
2519        nfsd_file_put(file);
2520
2521        return 0;
2522}
2523
2524static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
2525{
2526        struct nfs4_delegation *ds;
2527        struct nfs4_file *nf;
2528        struct nfsd_file *file;
2529
2530        ds = delegstateid(st);
2531        nf = st->sc_file;
2532        file = find_deleg_file(nf);
2533        if (!file)
2534                return 0;
2535
2536        seq_printf(s, "- ");
2537        nfs4_show_stateid(s, &st->sc_stateid);
2538        seq_printf(s, ": { type: deleg, ");
2539
2540        /* Kinda dead code as long as we only support read delegs: */
2541        seq_printf(s, "access: %s, ",
2542                ds->dl_type == NFS4_OPEN_DELEGATE_READ ? "r" : "w");
2543
2544        /* XXX: lease time, whether it's being recalled. */
2545
2546        nfs4_show_superblock(s, file);
2547        seq_printf(s, ", ");
2548        nfs4_show_fname(s, file);
2549        seq_printf(s, " }\n");
2550        nfsd_file_put(file);
2551
2552        return 0;
2553}
2554
2555static int nfs4_show_layout(struct seq_file *s, struct nfs4_stid *st)
2556{
2557        struct nfs4_layout_stateid *ls;
2558        struct nfsd_file *file;
2559
2560        ls = container_of(st, struct nfs4_layout_stateid, ls_stid);
2561        file = ls->ls_file;
2562
2563        seq_printf(s, "- ");
2564        nfs4_show_stateid(s, &st->sc_stateid);
2565        seq_printf(s, ": { type: layout, ");
2566
2567        /* XXX: What else would be useful? */
2568
2569        nfs4_show_superblock(s, file);
2570        seq_printf(s, ", ");
2571        nfs4_show_fname(s, file);
2572        seq_printf(s, " }\n");
2573
2574        return 0;
2575}
2576
2577static int states_show(struct seq_file *s, void *v)
2578{
2579        struct nfs4_stid *st = v;
2580
2581        switch (st->sc_type) {
2582        case NFS4_OPEN_STID:
2583                return nfs4_show_open(s, st);
2584        case NFS4_LOCK_STID:
2585                return nfs4_show_lock(s, st);
2586        case NFS4_DELEG_STID:
2587                return nfs4_show_deleg(s, st);
2588        case NFS4_LAYOUT_STID:
2589                return nfs4_show_layout(s, st);
2590        default:
2591                return 0; /* XXX: or SEQ_SKIP? */
2592        }
2593        /* XXX: copy stateids? */
2594}
2595
2596static struct seq_operations states_seq_ops = {
2597        .start = states_start,
2598        .next = states_next,
2599        .stop = states_stop,
2600        .show = states_show
2601};
2602
2603static int client_states_open(struct inode *inode, struct file *file)
2604{
2605        struct seq_file *s;
2606        struct nfs4_client *clp;
2607        int ret;
2608
2609        clp = get_nfsdfs_clp(inode);
2610        if (!clp)
2611                return -ENXIO;
2612
2613        ret = seq_open(file, &states_seq_ops);
2614        if (ret)
2615                return ret;
2616        s = file->private_data;
2617        s->private = clp;
2618        return 0;
2619}
2620
2621static int client_opens_release(struct inode *inode, struct file *file)
2622{
2623        struct seq_file *m = file->private_data;
2624        struct nfs4_client *clp = m->private;
2625
2626        /* XXX: alternatively, we could get/drop in seq start/stop */
2627        drop_client(clp);
2628        return 0;
2629}
2630
2631static const struct file_operations client_states_fops = {
2632        .open           = client_states_open,
2633        .read           = seq_read,
2634        .llseek         = seq_lseek,
2635        .release        = client_opens_release,
2636};
2637
2638/*
2639 * Normally we refuse to destroy clients that are in use, but here the
2640 * administrator is telling us to just do it.  We also want to wait
2641 * so the caller has a guarantee that the client's locks are gone by
2642 * the time the write returns:
2643 */
2644static void force_expire_client(struct nfs4_client *clp)
2645{
2646        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2647        bool already_expired;
2648
2649        spin_lock(&clp->cl_lock);
2650        clp->cl_time = 0;
2651        spin_unlock(&clp->cl_lock);
2652
2653        wait_event(expiry_wq, atomic_read(&clp->cl_rpc_users) == 0);
2654        spin_lock(&nn->client_lock);
2655        already_expired = list_empty(&clp->cl_lru);
2656        if (!already_expired)
2657                unhash_client_locked(clp);
2658        spin_unlock(&nn->client_lock);
2659
2660        if (!already_expired)
2661                expire_client(clp);
2662        else
2663                wait_event(expiry_wq, clp->cl_nfsd_dentry == NULL);
2664}
2665
2666static ssize_t client_ctl_write(struct file *file, const char __user *buf,
2667                                   size_t size, loff_t *pos)
2668{
2669        char *data;
2670        struct nfs4_client *clp;
2671
2672        data = simple_transaction_get(file, buf, size);
2673        if (IS_ERR(data))
2674                return PTR_ERR(data);
2675        if (size != 7 || 0 != memcmp(data, "expire\n", 7))
2676                return -EINVAL;
2677        clp = get_nfsdfs_clp(file_inode(file));
2678        if (!clp)
2679                return -ENXIO;
2680        force_expire_client(clp);
2681        drop_client(clp);
2682        return 7;
2683}
2684
2685static const struct file_operations client_ctl_fops = {
2686        .write          = client_ctl_write,
2687        .release        = simple_transaction_release,
2688};
2689
2690static const struct tree_descr client_files[] = {
2691        [0] = {"info", &client_info_fops, S_IRUSR},
2692        [1] = {"states", &client_states_fops, S_IRUSR},
2693        [2] = {"ctl", &client_ctl_fops, S_IWUSR},
2694        [3] = {""},
2695};
2696
2697static struct nfs4_client *create_client(struct xdr_netobj name,
2698                struct svc_rqst *rqstp, nfs4_verifier *verf)
2699{
2700        struct nfs4_client *clp;
2701        struct sockaddr *sa = svc_addr(rqstp);
2702        int ret;
2703        struct net *net = SVC_NET(rqstp);
2704        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2705
2706        clp = alloc_client(name);
2707        if (clp == NULL)
2708                return NULL;
2709
2710        ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
2711        if (ret) {
2712                free_client(clp);
2713                return NULL;
2714        }
2715        gen_clid(clp, nn);
2716        kref_init(&clp->cl_nfsdfs.cl_ref);
2717        nfsd4_init_cb(&clp->cl_cb_null, clp, NULL, NFSPROC4_CLNT_CB_NULL);
2718        clp->cl_time = ktime_get_boottime_seconds();
2719        clear_bit(0, &clp->cl_cb_slot_busy);
2720        copy_verf(clp, verf);
2721        memcpy(&clp->cl_addr, sa, sizeof(struct sockaddr_storage));
2722        clp->cl_cb_session = NULL;
2723        clp->net = net;
2724        clp->cl_nfsd_dentry = nfsd_client_mkdir(nn, &clp->cl_nfsdfs,
2725                        clp->cl_clientid.cl_id - nn->clientid_base,
2726                        client_files);
2727        if (!clp->cl_nfsd_dentry) {
2728                free_client(clp);
2729                return NULL;
2730        }
2731        return clp;
2732}
2733
2734static void
2735add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
2736{
2737        struct rb_node **new = &(root->rb_node), *parent = NULL;
2738        struct nfs4_client *clp;
2739
2740        while (*new) {
2741                clp = rb_entry(*new, struct nfs4_client, cl_namenode);
2742                parent = *new;
2743
2744                if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
2745                        new = &((*new)->rb_left);
2746                else
2747                        new = &((*new)->rb_right);
2748        }
2749
2750        rb_link_node(&new_clp->cl_namenode, parent, new);
2751        rb_insert_color(&new_clp->cl_namenode, root);
2752}
2753
2754static struct nfs4_client *
2755find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
2756{
2757        int cmp;
2758        struct rb_node *node = root->rb_node;
2759        struct nfs4_client *clp;
2760
2761        while (node) {
2762                clp = rb_entry(node, struct nfs4_client, cl_namenode);
2763                cmp = compare_blob(&clp->cl_name, name);
2764                if (cmp > 0)
2765                        node = node->rb_left;
2766                else if (cmp < 0)
2767                        node = node->rb_right;
2768                else
2769                        return clp;
2770        }
2771        return NULL;
2772}
2773
2774static void
2775add_to_unconfirmed(struct nfs4_client *clp)
2776{
2777        unsigned int idhashval;
2778        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2779
2780        lockdep_assert_held(&nn->client_lock);
2781
2782        clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2783        add_clp_to_name_tree(clp, &nn->unconf_name_tree);
2784        idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2785        list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
2786        renew_client_locked(clp);
2787}
2788
2789static void
2790move_to_confirmed(struct nfs4_client *clp)
2791{
2792        unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2793        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2794
2795        lockdep_assert_held(&nn->client_lock);
2796
2797        dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
2798        list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
2799        rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2800        add_clp_to_name_tree(clp, &nn->conf_name_tree);
2801        set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2802        renew_client_locked(clp);
2803}
2804
2805static struct nfs4_client *
2806find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
2807{
2808        struct nfs4_client *clp;
2809        unsigned int idhashval = clientid_hashval(clid->cl_id);
2810
2811        list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
2812                if (same_clid(&clp->cl_clientid, clid)) {
2813                        if ((bool)clp->cl_minorversion != sessions)
2814                                return NULL;
2815                        renew_client_locked(clp);
2816                        return clp;
2817                }
2818        }
2819        return NULL;
2820}
2821
2822static struct nfs4_client *
2823find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2824{
2825        struct list_head *tbl = nn->conf_id_hashtbl;
2826
2827        lockdep_assert_held(&nn->client_lock);
2828        return find_client_in_id_table(tbl, clid, sessions);
2829}
2830
2831static struct nfs4_client *
2832find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2833{
2834        struct list_head *tbl = nn->unconf_id_hashtbl;
2835
2836        lockdep_assert_held(&nn->client_lock);
2837        return find_client_in_id_table(tbl, clid, sessions);
2838}
2839
2840static bool clp_used_exchangeid(struct nfs4_client *clp)
2841{
2842        return clp->cl_exchange_flags != 0;
2843} 
2844
2845static struct nfs4_client *
2846find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2847{
2848        lockdep_assert_held(&nn->client_lock);
2849        return find_clp_in_name_tree(name, &nn->conf_name_tree);
2850}
2851
2852static struct nfs4_client *
2853find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2854{
2855        lockdep_assert_held(&nn->client_lock);
2856        return find_clp_in_name_tree(name, &nn->unconf_name_tree);
2857}
2858
2859static void
2860gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
2861{
2862        struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
2863        struct sockaddr *sa = svc_addr(rqstp);
2864        u32 scopeid = rpc_get_scope_id(sa);
2865        unsigned short expected_family;
2866
2867        /* Currently, we only support tcp and tcp6 for the callback channel */
2868        if (se->se_callback_netid_len == 3 &&
2869            !memcmp(se->se_callback_netid_val, "tcp", 3))
2870                expected_family = AF_INET;
2871        else if (se->se_callback_netid_len == 4 &&
2872                 !memcmp(se->se_callback_netid_val, "tcp6", 4))
2873                expected_family = AF_INET6;
2874        else
2875                goto out_err;
2876
2877        conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
2878                                            se->se_callback_addr_len,
2879                                            (struct sockaddr *)&conn->cb_addr,
2880                                            sizeof(conn->cb_addr));
2881
2882        if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
2883                goto out_err;
2884
2885        if (conn->cb_addr.ss_family == AF_INET6)
2886                ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
2887
2888        conn->cb_prog = se->se_callback_prog;
2889        conn->cb_ident = se->se_callback_ident;
2890        memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
2891        trace_nfsd_cb_args(clp, conn);
2892        return;
2893out_err:
2894        conn->cb_addr.ss_family = AF_UNSPEC;
2895        conn->cb_addrlen = 0;
2896        trace_nfsd_cb_nodelegs(clp);
2897        return;
2898}
2899
2900/*
2901 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
2902 */
2903static void
2904nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
2905{
2906        struct xdr_buf *buf = resp->xdr.buf;
2907        struct nfsd4_slot *slot = resp->cstate.slot;
2908        unsigned int base;
2909
2910        dprintk("--> %s slot %p\n", __func__, slot);
2911
2912        slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
2913        slot->sl_opcnt = resp->opcnt;
2914        slot->sl_status = resp->cstate.status;
2915        free_svc_cred(&slot->sl_cred);
2916        copy_cred(&slot->sl_cred, &resp->rqstp->rq_cred);
2917
2918        if (!nfsd4_cache_this(resp)) {
2919                slot->sl_flags &= ~NFSD4_SLOT_CACHED;
2920                return;
2921        }
2922        slot->sl_flags |= NFSD4_SLOT_CACHED;
2923
2924        base = resp->cstate.data_offset;
2925        slot->sl_datalen = buf->len - base;
2926        if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
2927                WARN(1, "%s: sessions DRC could not cache compound\n",
2928                     __func__);
2929        return;
2930}
2931
2932/*
2933 * Encode the replay sequence operation from the slot values.
2934 * If cachethis is FALSE encode the uncached rep error on the next
2935 * operation which sets resp->p and increments resp->opcnt for
2936 * nfs4svc_encode_compoundres.
2937 *
2938 */
2939static __be32
2940nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
2941                          struct nfsd4_compoundres *resp)
2942{
2943        struct nfsd4_op *op;
2944        struct nfsd4_slot *slot = resp->cstate.slot;
2945
2946        /* Encode the replayed sequence operation */
2947        op = &args->ops[resp->opcnt - 1];
2948        nfsd4_encode_operation(resp, op);
2949
2950        if (slot->sl_flags & NFSD4_SLOT_CACHED)
2951                return op->status;
2952        if (args->opcnt == 1) {
2953                /*
2954                 * The original operation wasn't a solo sequence--we
2955                 * always cache those--so this retry must not match the
2956                 * original:
2957                 */
2958                op->status = nfserr_seq_false_retry;
2959        } else {
2960                op = &args->ops[resp->opcnt++];
2961                op->status = nfserr_retry_uncached_rep;
2962                nfsd4_encode_operation(resp, op);
2963        }
2964        return op->status;
2965}
2966
2967/*
2968 * The sequence operation is not cached because we can use the slot and
2969 * session values.
2970 */
2971static __be32
2972nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
2973                         struct nfsd4_sequence *seq)
2974{
2975        struct nfsd4_slot *slot = resp->cstate.slot;
2976        struct xdr_stream *xdr = &resp->xdr;
2977        __be32 *p;
2978        __be32 status;
2979
2980        dprintk("--> %s slot %p\n", __func__, slot);
2981
2982        status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
2983        if (status)
2984                return status;
2985
2986        p = xdr_reserve_space(xdr, slot->sl_datalen);
2987        if (!p) {
2988                WARN_ON_ONCE(1);
2989                return nfserr_serverfault;
2990        }
2991        xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
2992        xdr_commit_encode(xdr);
2993
2994        resp->opcnt = slot->sl_opcnt;
2995        return slot->sl_status;
2996}
2997
2998/*
2999 * Set the exchange_id flags returned by the server.
3000 */
3001static void
3002nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
3003{
3004#ifdef CONFIG_NFSD_PNFS
3005        new->cl_exchange_flags |= EXCHGID4_FLAG_USE_PNFS_MDS;
3006#else
3007        new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
3008#endif
3009
3010        /* Referrals are supported, Migration is not. */
3011        new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
3012
3013        /* set the wire flags to return to client. */
3014        clid->flags = new->cl_exchange_flags;
3015}
3016
3017static bool client_has_openowners(struct nfs4_client *clp)
3018{
3019        struct nfs4_openowner *oo;
3020
3021        list_for_each_entry(oo, &clp->cl_openowners, oo_perclient) {
3022                if (!list_empty(&oo->oo_owner.so_stateids))
3023                        return true;
3024        }
3025        return false;
3026}
3027
3028static bool client_has_state(struct nfs4_client *clp)
3029{
3030        return client_has_openowners(clp)
3031#ifdef CONFIG_NFSD_PNFS
3032                || !list_empty(&clp->cl_lo_states)
3033#endif
3034                || !list_empty(&clp->cl_delegations)
3035                || !list_empty(&clp->cl_sessions)
3036                || !list_empty(&clp->async_copies);
3037}
3038
3039static __be32 copy_impl_id(struct nfs4_client *clp,
3040                                struct nfsd4_exchange_id *exid)
3041{
3042        if (!exid->nii_domain.data)
3043                return 0;
3044        xdr_netobj_dup(&clp->cl_nii_domain, &exid->nii_domain, GFP_KERNEL);
3045        if (!clp->cl_nii_domain.data)
3046                return nfserr_jukebox;
3047        xdr_netobj_dup(&clp->cl_nii_name, &exid->nii_name, GFP_KERNEL);
3048        if (!clp->cl_nii_name.data)
3049                return nfserr_jukebox;
3050        clp->cl_nii_time = exid->nii_time;
3051        return 0;
3052}
3053
3054__be32
3055nfsd4_exchange_id(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3056                union nfsd4_op_u *u)
3057{
3058        struct nfsd4_exchange_id *exid = &u->exchange_id;
3059        struct nfs4_client *conf, *new;
3060        struct nfs4_client *unconf = NULL;
3061        __be32 status;
3062        char                    addr_str[INET6_ADDRSTRLEN];
3063        nfs4_verifier           verf = exid->verifier;
3064        struct sockaddr         *sa = svc_addr(rqstp);
3065        bool    update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
3066        struct nfsd_net         *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3067
3068        rpc_ntop(sa, addr_str, sizeof(addr_str));
3069        dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
3070                "ip_addr=%s flags %x, spa_how %u\n",
3071                __func__, rqstp, exid, exid->clname.len, exid->clname.data,
3072                addr_str, exid->flags, exid->spa_how);
3073
3074        if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
3075                return nfserr_inval;
3076
3077        new = create_client(exid->clname, rqstp, &verf);
3078        if (new == NULL)
3079                return nfserr_jukebox;
3080        status = copy_impl_id(new, exid);
3081        if (status)
3082                goto out_nolock;
3083
3084        switch (exid->spa_how) {
3085        case SP4_MACH_CRED:
3086                exid->spo_must_enforce[0] = 0;
3087                exid->spo_must_enforce[1] = (
3088                        1 << (OP_BIND_CONN_TO_SESSION - 32) |
3089                        1 << (OP_EXCHANGE_ID - 32) |
3090                        1 << (OP_CREATE_SESSION - 32) |
3091                        1 << (OP_DESTROY_SESSION - 32) |
3092                        1 << (OP_DESTROY_CLIENTID - 32));
3093
3094                exid->spo_must_allow[0] &= (1 << (OP_CLOSE) |
3095                                        1 << (OP_OPEN_DOWNGRADE) |
3096                                        1 << (OP_LOCKU) |
3097                                        1 << (OP_DELEGRETURN));
3098
3099                exid->spo_must_allow[1] &= (
3100                                        1 << (OP_TEST_STATEID - 32) |
3101                                        1 << (OP_FREE_STATEID - 32));
3102                if (!svc_rqst_integrity_protected(rqstp)) {
3103                        status = nfserr_inval;
3104                        goto out_nolock;
3105                }
3106                /*
3107                 * Sometimes userspace doesn't give us a principal.
3108                 * Which is a bug, really.  Anyway, we can't enforce
3109                 * MACH_CRED in that case, better to give up now:
3110                 */
3111                if (!new->cl_cred.cr_principal &&
3112                                        !new->cl_cred.cr_raw_principal) {
3113                        status = nfserr_serverfault;
3114                        goto out_nolock;
3115                }
3116                new->cl_mach_cred = true;
3117        case SP4_NONE:
3118                break;
3119        default:                                /* checked by xdr code */
3120                WARN_ON_ONCE(1);
3121                fallthrough;
3122        case SP4_SSV:
3123                status = nfserr_encr_alg_unsupp;
3124                goto out_nolock;
3125        }
3126
3127        /* Cases below refer to rfc 5661 section 18.35.4: */
3128        spin_lock(&nn->client_lock);
3129        conf = find_confirmed_client_by_name(&exid->clname, nn);
3130        if (conf) {
3131                bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
3132                bool verfs_match = same_verf(&verf, &conf->cl_verifier);
3133
3134                if (update) {
3135                        if (!clp_used_exchangeid(conf)) { /* buggy client */
3136                                status = nfserr_inval;
3137                                goto out;
3138                        }
3139                        if (!nfsd4_mach_creds_match(conf, rqstp)) {
3140                                status = nfserr_wrong_cred;
3141                                goto out;
3142                        }
3143                        if (!creds_match) { /* case 9 */
3144                                status = nfserr_perm;
3145                                goto out;
3146                        }
3147                        if (!verfs_match) { /* case 8 */
3148                                status = nfserr_not_same;
3149                                goto out;
3150                        }
3151                        /* case 6 */
3152                        exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
3153                        goto out_copy;
3154                }
3155                if (!creds_match) { /* case 3 */
3156                        if (client_has_state(conf)) {
3157                                status = nfserr_clid_inuse;
3158                                goto out;
3159                        }
3160                        goto out_new;
3161                }
3162                if (verfs_match) { /* case 2 */
3163                        conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
3164                        goto out_copy;
3165                }
3166                /* case 5, client reboot */
3167                conf = NULL;
3168                goto out_new;
3169        }
3170
3171        if (update) { /* case 7 */
3172                status = nfserr_noent;
3173                goto out;
3174        }
3175
3176        unconf  = find_unconfirmed_client_by_name(&exid->clname, nn);
3177        if (unconf) /* case 4, possible retry or client restart */
3178                unhash_client_locked(unconf);
3179
3180        /* case 1 (normal case) */
3181out_new:
3182        if (conf) {
3183                status = mark_client_expired_locked(conf);
3184                if (status)
3185                        goto out;
3186        }
3187        new->cl_minorversion = cstate->minorversion;
3188        new->cl_spo_must_allow.u.words[0] = exid->spo_must_allow[0];
3189        new->cl_spo_must_allow.u.words[1] = exid->spo_must_allow[1];
3190
3191        add_to_unconfirmed(new);
3192        swap(new, conf);
3193out_copy:
3194        exid->clientid.cl_boot = conf->cl_clientid.cl_boot;
3195        exid->clientid.cl_id = conf->cl_clientid.cl_id;
3196
3197        exid->seqid = conf->cl_cs_slot.sl_seqid + 1;
3198        nfsd4_set_ex_flags(conf, exid);
3199
3200        dprintk("nfsd4_exchange_id seqid %d flags %x\n",
3201                conf->cl_cs_slot.sl_seqid, conf->cl_exchange_flags);
3202        status = nfs_ok;
3203
3204out:
3205        spin_unlock(&nn->client_lock);
3206out_nolock:
3207        if (new)
3208                expire_client(new);
3209        if (unconf)
3210                expire_client(unconf);
3211        return status;
3212}
3213
3214static __be32
3215check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
3216{
3217        dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
3218                slot_seqid);
3219
3220        /* The slot is in use, and no response has been sent. */
3221        if (slot_inuse) {
3222                if (seqid == slot_seqid)
3223                        return nfserr_jukebox;
3224                else
3225                        return nfserr_seq_misordered;
3226        }
3227        /* Note unsigned 32-bit arithmetic handles wraparound: */
3228        if (likely(seqid == slot_seqid + 1))
3229                return nfs_ok;
3230        if (seqid == slot_seqid)
3231                return nfserr_replay_cache;
3232        return nfserr_seq_misordered;
3233}
3234
3235/*
3236 * Cache the create session result into the create session single DRC
3237 * slot cache by saving the xdr structure. sl_seqid has been set.
3238 * Do this for solo or embedded create session operations.
3239 */
3240static void
3241nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
3242                           struct nfsd4_clid_slot *slot, __be32 nfserr)
3243{
3244        slot->sl_status = nfserr;
3245        memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
3246}
3247
3248static __be32
3249nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
3250                            struct nfsd4_clid_slot *slot)
3251{
3252        memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
3253        return slot->sl_status;
3254}
3255
3256#define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
3257                        2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
3258                        1 +     /* MIN tag is length with zero, only length */ \
3259                        3 +     /* version, opcount, opcode */ \
3260                        XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
3261                                /* seqid, slotID, slotID, cache */ \
3262                        4 ) * sizeof(__be32))
3263
3264#define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
3265                        2 +     /* verifier: AUTH_NULL, length 0 */\
3266                        1 +     /* status */ \
3267                        1 +     /* MIN tag is length with zero, only length */ \
3268                        3 +     /* opcount, opcode, opstatus*/ \
3269                        XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
3270                                /* seqid, slotID, slotID, slotID, status */ \
3271                        5 ) * sizeof(__be32))
3272
3273static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
3274{
3275        u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
3276
3277        if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
3278                return nfserr_toosmall;
3279        if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
3280                return nfserr_toosmall;
3281        ca->headerpadsz = 0;
3282        ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
3283        ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
3284        ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
3285        ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
3286                        NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
3287        ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
3288        /*
3289         * Note decreasing slot size below client's request may make it
3290         * difficult for client to function correctly, whereas
3291         * decreasing the number of slots will (just?) affect
3292         * performance.  When short on memory we therefore prefer to
3293         * decrease number of slots instead of their size.  Clients that
3294         * request larger slots than they need will get poor results:
3295         * Note that we always allow at least one slot, because our
3296         * accounting is soft and provides no guarantees either way.
3297         */
3298        ca->maxreqs = nfsd4_get_drc_mem(ca, nn);
3299
3300        return nfs_ok;
3301}
3302
3303/*
3304 * Server's NFSv4.1 backchannel support is AUTH_SYS-only for now.
3305 * These are based on similar macros in linux/sunrpc/msg_prot.h .
3306 */
3307#define RPC_MAX_HEADER_WITH_AUTH_SYS \
3308        (RPC_CALLHDRSIZE + 2 * (2 + UNX_CALLSLACK))
3309
3310#define RPC_MAX_REPHEADER_WITH_AUTH_SYS \
3311        (RPC_REPHDRSIZE + (2 + NUL_REPLYSLACK))
3312
3313#define NFSD_CB_MAX_REQ_SZ      ((NFS4_enc_cb_recall_sz + \
3314                                 RPC_MAX_HEADER_WITH_AUTH_SYS) * sizeof(__be32))
3315#define NFSD_CB_MAX_RESP_SZ     ((NFS4_dec_cb_recall_sz + \
3316                                 RPC_MAX_REPHEADER_WITH_AUTH_SYS) * \
3317                                 sizeof(__be32))
3318
3319static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
3320{
3321        ca->headerpadsz = 0;
3322
3323        if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
3324                return nfserr_toosmall;
3325        if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
3326                return nfserr_toosmall;
3327        ca->maxresp_cached = 0;
3328        if (ca->maxops < 2)
3329                return nfserr_toosmall;
3330
3331        return nfs_ok;
3332}
3333
3334static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
3335{
3336        switch (cbs->flavor) {
3337        case RPC_AUTH_NULL:
3338        case RPC_AUTH_UNIX:
3339                return nfs_ok;
3340        default:
3341                /*
3342                 * GSS case: the spec doesn't allow us to return this
3343                 * error.  But it also doesn't allow us not to support
3344                 * GSS.
3345                 * I'd rather this fail hard than return some error the
3346                 * client might think it can already handle:
3347                 */
3348                return nfserr_encr_alg_unsupp;
3349        }
3350}
3351
3352__be32
3353nfsd4_create_session(struct svc_rqst *rqstp,
3354                struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
3355{
3356        struct nfsd4_create_session *cr_ses = &u->create_session;
3357        struct sockaddr *sa = svc_addr(rqstp);
3358        struct nfs4_client *conf, *unconf;
3359        struct nfs4_client *old = NULL;
3360        struct nfsd4_session *new;
3361        struct nfsd4_conn *conn;
3362        struct nfsd4_clid_slot *cs_slot = NULL;
3363        __be32 status = 0;
3364        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3365
3366        if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
3367                return nfserr_inval;
3368        status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
3369        if (status)
3370                return status;
3371        status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
3372        if (status)
3373                return status;
3374        status = check_backchannel_attrs(&cr_ses->back_channel);
3375        if (status)
3376                goto out_release_drc_mem;
3377        status = nfserr_jukebox;
3378        new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
3379        if (!new)
3380                goto out_release_drc_mem;
3381        conn = alloc_conn_from_crses(rqstp, cr_ses);
3382        if (!conn)
3383                goto out_free_session;
3384
3385        spin_lock(&nn->client_lock);
3386        unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
3387        conf = find_confirmed_client(&cr_ses->clientid, true, nn);
3388        WARN_ON_ONCE(conf && unconf);
3389
3390        if (conf) {
3391                status = nfserr_wrong_cred;
3392                if (!nfsd4_mach_creds_match(conf, rqstp))
3393                        goto out_free_conn;
3394                cs_slot = &conf->cl_cs_slot;
3395                status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
3396                if (status) {
3397                        if (status == nfserr_replay_cache)
3398                                status = nfsd4_replay_create_session(cr_ses, cs_slot);
3399                        goto out_free_conn;
3400                }
3401        } else if (unconf) {
3402                if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
3403                    !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
3404                        status = nfserr_clid_inuse;
3405                        goto out_free_conn;
3406                }
3407                status = nfserr_wrong_cred;
3408                if (!nfsd4_mach_creds_match(unconf, rqstp))
3409                        goto out_free_conn;
3410                cs_slot = &unconf->cl_cs_slot;
3411                status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
3412                if (status) {
3413                        /* an unconfirmed replay returns misordered */
3414                        status = nfserr_seq_misordered;
3415                        goto out_free_conn;
3416                }
3417                old = find_confirmed_client_by_name(&unconf->cl_name, nn);
3418                if (old) {
3419                        status = mark_client_expired_locked(old);
3420                        if (status) {
3421                                old = NULL;
3422                                goto out_free_conn;
3423                        }
3424                }
3425                move_to_confirmed(unconf);
3426                conf = unconf;
3427        } else {
3428                status = nfserr_stale_clientid;
3429                goto out_free_conn;
3430        }
3431        status = nfs_ok;
3432        /* Persistent sessions are not supported */
3433        cr_ses->flags &= ~SESSION4_PERSIST;
3434        /* Upshifting from TCP to RDMA is not supported */
3435        cr_ses->flags &= ~SESSION4_RDMA;
3436
3437        init_session(rqstp, new, conf, cr_ses);
3438        nfsd4_get_session_locked(new);
3439
3440        memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
3441               NFS4_MAX_SESSIONID_LEN);
3442        cs_slot->sl_seqid++;
3443        cr_ses->seqid = cs_slot->sl_seqid;
3444
3445        /* cache solo and embedded create sessions under the client_lock */
3446        nfsd4_cache_create_session(cr_ses, cs_slot, status);
3447        spin_unlock(&nn->client_lock);
3448        /* init connection and backchannel */
3449        nfsd4_init_conn(rqstp, conn, new);
3450        nfsd4_put_session(new);
3451        if (old)
3452                expire_client(old);
3453        return status;
3454out_free_conn:
3455        spin_unlock(&nn->client_lock);
3456        free_conn(conn);
3457        if (old)
3458                expire_client(old);
3459out_free_session:
3460        __free_session(new);
3461out_release_drc_mem:
3462        nfsd4_put_drc_mem(&cr_ses->fore_channel);
3463        return status;
3464}
3465
3466static __be32 nfsd4_map_bcts_dir(u32 *dir)
3467{
3468        switch (*dir) {
3469        case NFS4_CDFC4_FORE:
3470        case NFS4_CDFC4_BACK:
3471                return nfs_ok;
3472        case NFS4_CDFC4_FORE_OR_BOTH:
3473        case NFS4_CDFC4_BACK_OR_BOTH:
3474                *dir = NFS4_CDFC4_BOTH;
3475                return nfs_ok;
3476        }
3477        return nfserr_inval;
3478}
3479
3480__be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp,
3481                struct nfsd4_compound_state *cstate,
3482                union nfsd4_op_u *u)
3483{
3484        struct nfsd4_backchannel_ctl *bc = &u->backchannel_ctl;
3485        struct nfsd4_session *session = cstate->session;
3486        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3487        __be32 status;
3488
3489        status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
3490        if (status)
3491                return status;
3492        spin_lock(&nn->client_lock);
3493        session->se_cb_prog = bc->bc_cb_program;
3494        session->se_cb_sec = bc->bc_cb_sec;
3495        spin_unlock(&nn->client_lock);
3496
3497        nfsd4_probe_callback(session->se_client);
3498
3499        return nfs_ok;
3500}
3501
3502static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
3503{
3504        struct nfsd4_conn *c;
3505
3506        list_for_each_entry(c, &s->se_conns, cn_persession) {
3507                if (c->cn_xprt == xpt) {
3508                        return c;
3509                }
3510        }
3511        return NULL;
3512}
3513
3514static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst,
3515                                struct nfsd4_session *session, u32 req)
3516{
3517        struct nfs4_client *clp = session->se_client;
3518        struct svc_xprt *xpt = rqst->rq_xprt;
3519        struct nfsd4_conn *c;
3520        __be32 status;
3521
3522        /* Following the last paragraph of RFC 5661 Section 18.34.3: */
3523        spin_lock(&clp->cl_lock);
3524        c = __nfsd4_find_conn(xpt, session);
3525        if (!c)
3526                status = nfserr_noent;
3527        else if (req == c->cn_flags)
3528                status = nfs_ok;
3529        else if (req == NFS4_CDFC4_FORE_OR_BOTH &&
3530                                c->cn_flags != NFS4_CDFC4_BACK)
3531                status = nfs_ok;
3532        else if (req == NFS4_CDFC4_BACK_OR_BOTH &&
3533                                c->cn_flags != NFS4_CDFC4_FORE)
3534                status = nfs_ok;
3535        else
3536                status = nfserr_inval;
3537        spin_unlock(&clp->cl_lock);
3538        return status;
3539}
3540
3541__be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
3542                     struct nfsd4_compound_state *cstate,
3543                     union nfsd4_op_u *u)
3544{
3545        struct nfsd4_bind_conn_to_session *bcts = &u->bind_conn_to_session;
3546        __be32 status;
3547        struct nfsd4_conn *conn;
3548        struct nfsd4_session *session;
3549        struct net *net = SVC_NET(rqstp);
3550        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3551
3552        if (!nfsd4_last_compound_op(rqstp))
3553                return nfserr_not_only_op;
3554        spin_lock(&nn->client_lock);
3555        session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
3556        spin_unlock(&nn->client_lock);
3557        if (!session)
3558                goto out_no_session;
3559        status = nfserr_wrong_cred;
3560        if (!nfsd4_mach_creds_match(session->se_client, rqstp))
3561                goto out;
3562        status = nfsd4_match_existing_connection(rqstp, session, bcts->dir);
3563        if (status == nfs_ok || status == nfserr_inval)
3564                goto out;
3565        status = nfsd4_map_bcts_dir(&bcts->dir);
3566        if (status)
3567                goto out;
3568        conn = alloc_conn(rqstp, bcts->dir);
3569        status = nfserr_jukebox;
3570        if (!conn)
3571                goto out;
3572        nfsd4_init_conn(rqstp, conn, session);
3573        status = nfs_ok;
3574out:
3575        nfsd4_put_session(session);
3576out_no_session:
3577        return status;
3578}
3579
3580static bool nfsd4_compound_in_session(struct nfsd4_compound_state *cstate, struct nfs4_sessionid *sid)
3581{
3582        if (!cstate->session)
3583                return false;
3584        return !memcmp(sid, &cstate->session->se_sessionid, sizeof(*sid));
3585}
3586
3587__be32
3588nfsd4_destroy_session(struct svc_rqst *r, struct nfsd4_compound_state *cstate,
3589                union nfsd4_op_u *u)
3590{
3591        struct nfs4_sessionid *sessionid = &u->destroy_session.sessionid;
3592        struct nfsd4_session *ses;
3593        __be32 status;
3594        int ref_held_by_me = 0;
3595        struct net *net = SVC_NET(r);
3596        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3597
3598        status = nfserr_not_only_op;
3599        if (nfsd4_compound_in_session(cstate, sessionid)) {
3600                if (!nfsd4_last_compound_op(r))
3601                        goto out;
3602                ref_held_by_me++;
3603        }
3604        dump_sessionid(__func__, sessionid);
3605        spin_lock(&nn->client_lock);
3606        ses = find_in_sessionid_hashtbl(sessionid, net, &status);
3607        if (!ses)
3608                goto out_client_lock;
3609        status = nfserr_wrong_cred;
3610        if (!nfsd4_mach_creds_match(ses->se_client, r))
3611                goto out_put_session;
3612        status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
3613        if (status)
3614                goto out_put_session;
3615        unhash_session(ses);
3616        spin_unlock(&nn->client_lock);
3617
3618        nfsd4_probe_callback_sync(ses->se_client);
3619
3620        spin_lock(&nn->client_lock);
3621        status = nfs_ok;
3622out_put_session:
3623        nfsd4_put_session_locked(ses);
3624out_client_lock:
3625        spin_unlock(&nn->client_lock);
3626out:
3627        return status;
3628}
3629
3630static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
3631{
3632        struct nfs4_client *clp = ses->se_client;
3633        struct nfsd4_conn *c;
3634        __be32 status = nfs_ok;
3635        int ret;
3636
3637        spin_lock(&clp->cl_lock);
3638        c = __nfsd4_find_conn(new->cn_xprt, ses);
3639        if (c)
3640                goto out_free;
3641        status = nfserr_conn_not_bound_to_session;
3642        if (clp->cl_mach_cred)
3643                goto out_free;
3644        __nfsd4_hash_conn(new, ses);
3645        spin_unlock(&clp->cl_lock);
3646        ret = nfsd4_register_conn(new);
3647        if (ret)
3648                /* oops; xprt is already down: */
3649                nfsd4_conn_lost(&new->cn_xpt_user);
3650        return nfs_ok;
3651out_free:
3652        spin_unlock(&clp->cl_lock);
3653        free_conn(new);
3654        return status;
3655}
3656
3657static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
3658{
3659        struct nfsd4_compoundargs *args = rqstp->rq_argp;
3660
3661        return args->opcnt > session->se_fchannel.maxops;
3662}
3663
3664static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
3665                                  struct nfsd4_session *session)
3666{
3667        struct xdr_buf *xb = &rqstp->rq_arg;
3668
3669        return xb->len > session->se_fchannel.maxreq_sz;
3670}
3671
3672static bool replay_matches_cache(struct svc_rqst *rqstp,
3673                 struct nfsd4_sequence *seq, struct nfsd4_slot *slot)
3674{
3675        struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3676
3677        if ((bool)(slot->sl_flags & NFSD4_SLOT_CACHETHIS) !=
3678            (bool)seq->cachethis)
3679                return false;
3680        /*
3681         * If there's an error then the reply can have fewer ops than
3682         * the call.
3683         */
3684        if (slot->sl_opcnt < argp->opcnt && !slot->sl_status)
3685                return false;
3686        /*
3687         * But if we cached a reply with *more* ops than the call you're
3688         * sending us now, then this new call is clearly not really a
3689         * replay of the old one:
3690         */
3691        if (slot->sl_opcnt > argp->opcnt)
3692                return false;
3693        /* This is the only check explicitly called by spec: */
3694        if (!same_creds(&rqstp->rq_cred, &slot->sl_cred))
3695                return false;
3696        /*
3697         * There may be more comparisons we could actually do, but the
3698         * spec doesn't require us to catch every case where the calls
3699         * don't match (that would require caching the call as well as
3700         * the reply), so we don't bother.
3701         */
3702        return true;
3703}
3704
3705__be32
3706nfsd4_sequence(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3707                union nfsd4_op_u *u)
3708{
3709        struct nfsd4_sequence *seq = &u->sequence;
3710        struct nfsd4_compoundres *resp = rqstp->rq_resp;
3711        struct xdr_stream *xdr = &resp->xdr;
3712        struct nfsd4_session *session;
3713        struct nfs4_client *clp;
3714        struct nfsd4_slot *slot;
3715        struct nfsd4_conn *conn;
3716        __be32 status;
3717        int buflen;
3718        struct net *net = SVC_NET(rqstp);
3719        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3720
3721        if (resp->opcnt != 1)
3722                return nfserr_sequence_pos;
3723
3724        /*
3725         * Will be either used or freed by nfsd4_sequence_check_conn
3726         * below.
3727         */
3728        conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
3729        if (!conn)
3730                return nfserr_jukebox;
3731
3732        spin_lock(&nn->client_lock);
3733        session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
3734        if (!session)
3735                goto out_no_session;
3736        clp = session->se_client;
3737
3738        status = nfserr_too_many_ops;
3739        if (nfsd4_session_too_many_ops(rqstp, session))
3740                goto out_put_session;
3741
3742        status = nfserr_req_too_big;
3743        if (nfsd4_request_too_big(rqstp, session))
3744                goto out_put_session;
3745
3746        status = nfserr_badslot;
3747        if (seq->slotid >= session->se_fchannel.maxreqs)
3748                goto out_put_session;
3749
3750        slot = session->se_slots[seq->slotid];
3751        dprintk("%s: slotid %d\n", __func__, seq->slotid);
3752
3753        /* We do not negotiate the number of slots yet, so set the
3754         * maxslots to the session maxreqs which is used to encode
3755         * sr_highest_slotid and the sr_target_slot id to maxslots */
3756        seq->maxslots = session->se_fchannel.maxreqs;
3757
3758        status = check_slot_seqid(seq->seqid, slot->sl_seqid,
3759                                        slot->sl_flags & NFSD4_SLOT_INUSE);
3760        if (status == nfserr_replay_cache) {
3761                status = nfserr_seq_misordered;
3762                if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
3763                        goto out_put_session;
3764                status = nfserr_seq_false_retry;
3765                if (!replay_matches_cache(rqstp, seq, slot))
3766                        goto out_put_session;
3767                cstate->slot = slot;
3768                cstate->session = session;
3769                cstate->clp = clp;
3770                /* Return the cached reply status and set cstate->status
3771                 * for nfsd4_proc_compound processing */
3772                status = nfsd4_replay_cache_entry(resp, seq);
3773                cstate->status = nfserr_replay_cache;
3774                goto out;
3775        }
3776        if (status)
3777                goto out_put_session;
3778
3779        status = nfsd4_sequence_check_conn(conn, session);
3780        conn = NULL;
3781        if (status)
3782                goto out_put_session;
3783
3784        buflen = (seq->cachethis) ?
3785                        session->se_fchannel.maxresp_cached :
3786                        session->se_fchannel.maxresp_sz;
3787        status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
3788                                    nfserr_rep_too_big;
3789        if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
3790                goto out_put_session;
3791        svc_reserve(rqstp, buflen);
3792
3793        status = nfs_ok;
3794        /* Success! bump slot seqid */
3795        slot->sl_seqid = seq->seqid;
3796        slot->sl_flags |= NFSD4_SLOT_INUSE;
3797        if (seq->cachethis)
3798                slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
3799        else
3800                slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
3801
3802        cstate->slot = slot;
3803        cstate->session = session;
3804        cstate->clp = clp;
3805
3806out:
3807        switch (clp->cl_cb_state) {
3808        case NFSD4_CB_DOWN:
3809                seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
3810                break;
3811        case NFSD4_CB_FAULT:
3812                seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
3813                break;
3814        default:
3815                seq->status_flags = 0;
3816        }
3817        if (!list_empty(&clp->cl_revoked))
3818                seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
3819out_no_session:
3820        if (conn)
3821                free_conn(conn);
3822        spin_unlock(&nn->client_lock);
3823        return status;
3824out_put_session:
3825        nfsd4_put_session_locked(session);
3826        goto out_no_session;
3827}
3828
3829void
3830nfsd4_sequence_done(struct nfsd4_compoundres *resp)
3831{
3832        struct nfsd4_compound_state *cs = &resp->cstate;
3833
3834        if (nfsd4_has_session(cs)) {
3835                if (cs->status != nfserr_replay_cache) {
3836                        nfsd4_store_cache_entry(resp);
3837                        cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
3838                }
3839                /* Drop session reference that was taken in nfsd4_sequence() */
3840                nfsd4_put_session(cs->session);
3841        } else if (cs->clp)
3842                put_client_renew(cs->clp);
3843}
3844
3845__be32
3846nfsd4_destroy_clientid(struct svc_rqst *rqstp,
3847                struct nfsd4_compound_state *cstate,
3848                union nfsd4_op_u *u)
3849{
3850        struct nfsd4_destroy_clientid *dc = &u->destroy_clientid;
3851        struct nfs4_client *conf, *unconf;
3852        struct nfs4_client *clp = NULL;
3853        __be32 status = 0;
3854        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3855
3856        spin_lock(&nn->client_lock);
3857        unconf = find_unconfirmed_client(&dc->clientid, true, nn);
3858        conf = find_confirmed_client(&dc->clientid, true, nn);
3859        WARN_ON_ONCE(conf && unconf);
3860
3861        if (conf) {
3862                if (client_has_state(conf)) {
3863                        status = nfserr_clientid_busy;
3864                        goto out;
3865                }
3866                status = mark_client_expired_locked(conf);
3867                if (status)
3868                        goto out;
3869                clp = conf;
3870        } else if (unconf)
3871                clp = unconf;
3872        else {
3873                status = nfserr_stale_clientid;
3874                goto out;
3875        }
3876        if (!nfsd4_mach_creds_match(clp, rqstp)) {
3877                clp = NULL;
3878                status = nfserr_wrong_cred;
3879                goto out;
3880        }
3881        unhash_client_locked(clp);
3882out:
3883        spin_unlock(&nn->client_lock);
3884        if (clp)
3885                expire_client(clp);
3886        return status;
3887}
3888
3889__be32
3890nfsd4_reclaim_complete(struct svc_rqst *rqstp,
3891                struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
3892{
3893        struct nfsd4_reclaim_complete *rc = &u->reclaim_complete;
3894        struct nfs4_client *clp = cstate->clp;
3895        __be32 status = 0;
3896
3897        if (rc->rca_one_fs) {
3898                if (!cstate->current_fh.fh_dentry)
3899                        return nfserr_nofilehandle;
3900                /*
3901                 * We don't take advantage of the rca_one_fs case.
3902                 * That's OK, it's optional, we can safely ignore it.
3903                 */
3904                return nfs_ok;
3905        }
3906
3907        status = nfserr_complete_already;
3908        if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &clp->cl_flags))
3909                goto out;
3910
3911        status = nfserr_stale_clientid;
3912        if (is_client_expired(clp))
3913                /*
3914                 * The following error isn't really legal.
3915                 * But we only get here if the client just explicitly
3916                 * destroyed the client.  Surely it no longer cares what
3917                 * error it gets back on an operation for the dead
3918                 * client.
3919                 */
3920                goto out;
3921
3922        status = nfs_ok;
3923        nfsd4_client_record_create(clp);
3924        inc_reclaim_complete(clp);
3925out:
3926        return status;
3927}
3928
3929__be32
3930nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3931                  union nfsd4_op_u *u)
3932{
3933        struct nfsd4_setclientid *setclid = &u->setclientid;
3934        struct xdr_netobj       clname = setclid->se_name;
3935        nfs4_verifier           clverifier = setclid->se_verf;
3936        struct nfs4_client      *conf, *new;
3937        struct nfs4_client      *unconf = NULL;
3938        __be32                  status;
3939        struct nfsd_net         *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3940
3941        new = create_client(clname, rqstp, &clverifier);
3942        if (new == NULL)
3943                return nfserr_jukebox;
3944        /* Cases below refer to rfc 3530 section 14.2.33: */
3945        spin_lock(&nn->client_lock);
3946        conf = find_confirmed_client_by_name(&clname, nn);
3947        if (conf && client_has_state(conf)) {
3948                /* case 0: */
3949                status = nfserr_clid_inuse;
3950                if (clp_used_exchangeid(conf))
3951                        goto out;
3952                if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
3953                        trace_nfsd_clid_inuse_err(conf);
3954                        goto out;
3955                }
3956        }
3957        unconf = find_unconfirmed_client_by_name(&clname, nn);
3958        if (unconf)
3959                unhash_client_locked(unconf);
3960        /* We need to handle only case 1: probable callback update */
3961        if (conf && same_verf(&conf->cl_verifier, &clverifier)) {
3962                copy_clid(new, conf);
3963                gen_confirm(new, nn);
3964        }
3965        new->cl_minorversion = 0;
3966        gen_callback(new, setclid, rqstp);
3967        add_to_unconfirmed(new);
3968        setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
3969        setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
3970        memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
3971        new = NULL;
3972        status = nfs_ok;
3973out:
3974        spin_unlock(&nn->client_lock);
3975        if (new)
3976                free_client(new);
3977        if (unconf)
3978                expire_client(unconf);
3979        return status;
3980}
3981
3982
3983__be32
3984nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
3985                        struct nfsd4_compound_state *cstate,
3986                        union nfsd4_op_u *u)
3987{
3988        struct nfsd4_setclientid_confirm *setclientid_confirm =
3989                        &u->setclientid_confirm;
3990        struct nfs4_client *conf, *unconf;
3991        struct nfs4_client *old = NULL;
3992        nfs4_verifier confirm = setclientid_confirm->sc_confirm; 
3993        clientid_t * clid = &setclientid_confirm->sc_clientid;
3994        __be32 status;
3995        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3996
3997        if (STALE_CLIENTID(clid, nn))
3998                return nfserr_stale_clientid;
3999
4000        spin_lock(&nn->client_lock);
4001        conf = find_confirmed_client(clid, false, nn);
4002        unconf = find_unconfirmed_client(clid, false, nn);
4003        /*
4004         * We try hard to give out unique clientid's, so if we get an
4005         * attempt to confirm the same clientid with a different cred,
4006         * the client may be buggy; this should never happen.
4007         *
4008         * Nevertheless, RFC 7530 recommends INUSE for this case:
4009         */
4010        status = nfserr_clid_inuse;
4011        if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
4012                goto out;
4013        if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
4014                goto out;
4015        /* cases below refer to rfc 3530 section 14.2.34: */
4016        if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
4017                if (conf && same_verf(&confirm, &conf->cl_confirm)) {
4018                        /* case 2: probable retransmit */
4019                        status = nfs_ok;
4020                } else /* case 4: client hasn't noticed we rebooted yet? */
4021                        status = nfserr_stale_clientid;
4022                goto out;
4023        }
4024        status = nfs_ok;
4025        if (conf) { /* case 1: callback update */
4026                old = unconf;
4027                unhash_client_locked(old);
4028                nfsd4_change_callback(conf, &unconf->cl_cb_conn);
4029        } else { /* case 3: normal case; new or rebooted client */
4030                old = find_confirmed_client_by_name(&unconf->cl_name, nn);
4031                if (old) {
4032                        status = nfserr_clid_inuse;
4033                        if (client_has_state(old)
4034                                        && !same_creds(&unconf->cl_cred,
4035                                                        &old->cl_cred))
4036                                goto out;
4037                        status = mark_client_expired_locked(old);
4038                        if (status) {
4039                                old = NULL;
4040                                goto out;
4041                        }
4042                }
4043                move_to_confirmed(unconf);
4044                conf = unconf;
4045        }
4046        get_client_locked(conf);
4047        spin_unlock(&nn->client_lock);
4048        nfsd4_probe_callback(conf);
4049        spin_lock(&nn->client_lock);
4050        put_client_renew_locked(conf);
4051out:
4052        spin_unlock(&nn->client_lock);
4053        if (old)
4054                expire_client(old);
4055        return status;
4056}
4057
4058static struct nfs4_file *nfsd4_alloc_file(void)
4059{
4060        return kmem_cache_alloc(file_slab, GFP_KERNEL);
4061}
4062
4063/* OPEN Share state helper functions */
4064static void nfsd4_init_file(struct knfsd_fh *fh, unsigned int hashval,
4065                                struct nfs4_file *fp)
4066{
4067        lockdep_assert_held(&state_lock);
4068
4069        refcount_set(&fp->fi_ref, 1);
4070        spin_lock_init(&fp->fi_lock);
4071        INIT_LIST_HEAD(&fp->fi_stateids);
4072        INIT_LIST_HEAD(&fp->fi_delegations);
4073        INIT_LIST_HEAD(&fp->fi_clnt_odstate);
4074        fh_copy_shallow(&fp->fi_fhandle, fh);
4075        fp->fi_deleg_file = NULL;
4076        fp->fi_had_conflict = false;
4077        fp->fi_share_deny = 0;
4078        memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
4079        memset(fp->fi_access, 0, sizeof(fp->fi_access));
4080#ifdef CONFIG_NFSD_PNFS
4081        INIT_LIST_HEAD(&fp->fi_lo_states);
4082        atomic_set(&fp->fi_lo_recalls, 0);
4083#endif
4084        hlist_add_head_rcu(&fp->fi_hash, &file_hashtbl[hashval]);
4085}
4086
4087void
4088nfsd4_free_slabs(void)
4089{
4090        kmem_cache_destroy(client_slab);
4091        kmem_cache_destroy(openowner_slab);
4092        kmem_cache_destroy(lockowner_slab);
4093        kmem_cache_destroy(file_slab);
4094        kmem_cache_destroy(stateid_slab);
4095        kmem_cache_destroy(deleg_slab);
4096        kmem_cache_destroy(odstate_slab);
4097}
4098
4099int
4100nfsd4_init_slabs(void)
4101{
4102        client_slab = kmem_cache_create("nfsd4_clients",
4103                        sizeof(struct nfs4_client), 0, 0, NULL);
4104        if (client_slab == NULL)
4105                goto out;
4106        openowner_slab = kmem_cache_create("nfsd4_openowners",
4107                        sizeof(struct nfs4_openowner), 0, 0, NULL);
4108        if (openowner_slab == NULL)
4109                goto out_free_client_slab;
4110        lockowner_slab = kmem_cache_create("nfsd4_lockowners",
4111                        sizeof(struct nfs4_lockowner), 0, 0, NULL);
4112        if (lockowner_slab == NULL)
4113                goto out_free_openowner_slab;
4114        file_slab = kmem_cache_create("nfsd4_files",
4115                        sizeof(struct nfs4_file), 0, 0, NULL);
4116        if (file_slab == NULL)
4117                goto out_free_lockowner_slab;
4118        stateid_slab = kmem_cache_create("nfsd4_stateids",
4119                        sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
4120        if (stateid_slab == NULL)
4121                goto out_free_file_slab;
4122        deleg_slab = kmem_cache_create("nfsd4_delegations",
4123                        sizeof(struct nfs4_delegation), 0, 0, NULL);
4124        if (deleg_slab == NULL)
4125                goto out_free_stateid_slab;
4126        odstate_slab = kmem_cache_create("nfsd4_odstate",
4127                        sizeof(struct nfs4_clnt_odstate), 0, 0, NULL);
4128        if (odstate_slab == NULL)
4129                goto out_free_deleg_slab;
4130        return 0;
4131
4132out_free_deleg_slab:
4133        kmem_cache_destroy(deleg_slab);
4134out_free_stateid_slab:
4135        kmem_cache_destroy(stateid_slab);
4136out_free_file_slab:
4137        kmem_cache_destroy(file_slab);
4138out_free_lockowner_slab:
4139        kmem_cache_destroy(lockowner_slab);
4140out_free_openowner_slab:
4141        kmem_cache_destroy(openowner_slab);
4142out_free_client_slab:
4143        kmem_cache_destroy(client_slab);
4144out:
4145        return -ENOMEM;
4146}
4147
4148static void init_nfs4_replay(struct nfs4_replay *rp)
4149{
4150        rp->rp_status = nfserr_serverfault;
4151        rp->rp_buflen = 0;
4152        rp->rp_buf = rp->rp_ibuf;
4153        mutex_init(&rp->rp_mutex);
4154}
4155
4156static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
4157                struct nfs4_stateowner *so)
4158{
4159        if (!nfsd4_has_session(cstate)) {
4160                mutex_lock(&so->so_replay.rp_mutex);
4161                cstate->replay_owner = nfs4_get_stateowner(so);
4162        }
4163}
4164
4165void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
4166{
4167        struct nfs4_stateowner *so = cstate->replay_owner;
4168
4169        if (so != NULL) {
4170                cstate->replay_owner = NULL;
4171                mutex_unlock(&so->so_replay.rp_mutex);
4172                nfs4_put_stateowner(so);
4173        }
4174}
4175
4176static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
4177{
4178        struct nfs4_stateowner *sop;
4179
4180        sop = kmem_cache_alloc(slab, GFP_KERNEL);
4181        if (!sop)
4182                return NULL;
4183
4184        xdr_netobj_dup(&sop->so_owner, owner, GFP_KERNEL);
4185        if (!sop->so_owner.data) {
4186                kmem_cache_free(slab, sop);
4187                return NULL;
4188        }
4189
4190        INIT_LIST_HEAD(&sop->so_stateids);
4191        sop->so_client = clp;
4192        init_nfs4_replay(&sop->so_replay);
4193        atomic_set(&sop->so_count, 1);
4194        return sop;
4195}
4196
4197static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
4198{
4199        lockdep_assert_held(&clp->cl_lock);
4200
4201        list_add(&oo->oo_owner.so_strhash,
4202                 &clp->cl_ownerstr_hashtbl[strhashval]);
4203        list_add(&oo->oo_perclient, &clp->cl_openowners);
4204}
4205
4206static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
4207{
4208        unhash_openowner_locked(openowner(so));
4209}
4210
4211static void nfs4_free_openowner(struct nfs4_stateowner *so)
4212{
4213        struct nfs4_openowner *oo = openowner(so);
4214
4215        kmem_cache_free(openowner_slab, oo);
4216}
4217
4218static const struct nfs4_stateowner_operations openowner_ops = {
4219        .so_unhash =    nfs4_unhash_openowner,
4220        .so_free =      nfs4_free_openowner,
4221};
4222
4223static struct nfs4_ol_stateid *
4224nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
4225{
4226        struct nfs4_ol_stateid *local, *ret = NULL;
4227        struct nfs4_openowner *oo = open->op_openowner;
4228
4229        lockdep_assert_held(&fp->fi_lock);
4230
4231        list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
4232                /* ignore lock owners */
4233                if (local->st_stateowner->so_is_open_owner == 0)
4234                        continue;
4235                if (local->st_stateowner != &oo->oo_owner)
4236                        continue;
4237                if (local->st_stid.sc_type == NFS4_OPEN_STID) {
4238                        ret = local;
4239                        refcount_inc(&ret->st_stid.sc_count);
4240                        break;
4241                }
4242        }
4243        return ret;
4244}
4245
4246static __be32
4247nfsd4_verify_open_stid(struct nfs4_stid *s)
4248{
4249        __be32 ret = nfs_ok;
4250
4251        switch (s->sc_type) {
4252        default:
4253                break;
4254        case 0:
4255        case NFS4_CLOSED_STID:
4256        case NFS4_CLOSED_DELEG_STID:
4257                ret = nfserr_bad_stateid;
4258                break;
4259        case NFS4_REVOKED_DELEG_STID:
4260                ret = nfserr_deleg_revoked;
4261        }
4262        return ret;
4263}
4264
4265/* Lock the stateid st_mutex, and deal with races with CLOSE */
4266static __be32
4267nfsd4_lock_ol_stateid(struct nfs4_ol_stateid *stp)
4268{
4269        __be32 ret;
4270
4271        mutex_lock_nested(&stp->st_mutex, LOCK_STATEID_MUTEX);
4272        ret = nfsd4_verify_open_stid(&stp->st_stid);
4273        if (ret != nfs_ok)
4274                mutex_unlock(&stp->st_mutex);
4275        return ret;
4276}
4277
4278static struct nfs4_ol_stateid *
4279nfsd4_find_and_lock_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
4280{
4281        struct nfs4_ol_stateid *stp;
4282        for (;;) {
4283                spin_lock(&fp->fi_lock);
4284                stp = nfsd4_find_existing_open(fp, open);
4285                spin_unlock(&fp->fi_lock);
4286                if (!stp || nfsd4_lock_ol_stateid(stp) == nfs_ok)
4287                        break;
4288                nfs4_put_stid(&stp->st_stid);
4289        }
4290        return stp;
4291}
4292
4293static struct nfs4_openowner *
4294alloc_init_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
4295                           struct nfsd4_compound_state *cstate)
4296{
4297        struct nfs4_client *clp = cstate->clp;
4298        struct nfs4_openowner *oo, *ret;
4299
4300        oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
4301        if (!oo)
4302                return NULL;
4303        oo->oo_owner.so_ops = &openowner_ops;
4304        oo->oo_owner.so_is_open_owner = 1;
4305        oo->oo_owner.so_seqid = open->op_seqid;
4306        oo->oo_flags = 0;
4307        if (nfsd4_has_session(cstate))
4308                oo->oo_flags |= NFS4_OO_CONFIRMED;
4309        oo->oo_time = 0;
4310        oo->oo_last_closed_stid = NULL;
4311        INIT_LIST_HEAD(&oo->oo_close_lru);
4312        spin_lock(&clp->cl_lock);
4313        ret = find_openstateowner_str_locked(strhashval, open, clp);
4314        if (ret == NULL) {
4315                hash_openowner(oo, clp, strhashval);
4316                ret = oo;
4317        } else
4318                nfs4_free_stateowner(&oo->oo_owner);
4319
4320        spin_unlock(&clp->cl_lock);
4321        return ret;
4322}
4323
4324static struct nfs4_ol_stateid *
4325init_open_stateid(struct nfs4_file *fp, struct nfsd4_open *open)
4326{
4327
4328        struct nfs4_openowner *oo = open->op_openowner;
4329        struct nfs4_ol_stateid *retstp = NULL;
4330        struct nfs4_ol_stateid *stp;
4331
4332        stp = open->op_stp;
4333        /* We are moving these outside of the spinlocks to avoid the warnings */
4334        mutex_init(&stp->st_mutex);
4335        mutex_lock_nested(&stp->st_mutex, OPEN_STATEID_MUTEX);
4336
4337retry:
4338        spin_lock(&oo->oo_owner.so_client->cl_lock);
4339        spin_lock(&fp->fi_lock);
4340
4341        retstp = nfsd4_find_existing_open(fp, open);
4342        if (retstp)
4343                goto out_unlock;
4344
4345        open->op_stp = NULL;
4346        refcount_inc(&stp->st_stid.sc_count);
4347        stp->st_stid.sc_type = NFS4_OPEN_STID;
4348        INIT_LIST_HEAD(&stp->st_locks);
4349        stp->st_stateowner = nfs4_get_stateowner(&oo->oo_owner);
4350        get_nfs4_file(fp);
4351        stp->st_stid.sc_file = fp;
4352        stp->st_access_bmap = 0;
4353        stp->st_deny_bmap = 0;
4354        stp->st_openstp = NULL;
4355        list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
4356        list_add(&stp->st_perfile, &fp->fi_stateids);
4357
4358out_unlock:
4359        spin_unlock(&fp->fi_lock);
4360        spin_unlock(&oo->oo_owner.so_client->cl_lock);
4361        if (retstp) {
4362                /* Handle races with CLOSE */
4363                if (nfsd4_lock_ol_stateid(retstp) != nfs_ok) {
4364                        nfs4_put_stid(&retstp->st_stid);
4365                        goto retry;
4366                }
4367                /* To keep mutex tracking happy */
4368                mutex_unlock(&stp->st_mutex);
4369                stp = retstp;
4370        }
4371        return stp;
4372}
4373
4374/*
4375 * In the 4.0 case we need to keep the owners around a little while to handle
4376 * CLOSE replay. We still do need to release any file access that is held by
4377 * them before returning however.
4378 */
4379static void
4380move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
4381{
4382        struct nfs4_ol_stateid *last;
4383        struct nfs4_openowner *oo = openowner(s->st_stateowner);
4384        struct nfsd_net *nn = net_generic(s->st_stid.sc_client->net,
4385                                                nfsd_net_id);
4386
4387        dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
4388
4389        /*
4390         * We know that we hold one reference via nfsd4_close, and another
4391         * "persistent" reference for the client. If the refcount is higher
4392         * than 2, then there are still calls in progress that are using this
4393         * stateid. We can't put the sc_file reference until they are finished.
4394         * Wait for the refcount to drop to 2. Since it has been unhashed,
4395         * there should be no danger of the refcount going back up again at
4396         * this point.
4397         */
4398        wait_event(close_wq, refcount_read(&s->st_stid.sc_count) == 2);
4399
4400        release_all_access(s);
4401        if (s->st_stid.sc_file) {
4402                put_nfs4_file(s->st_stid.sc_file);
4403                s->st_stid.sc_file = NULL;
4404        }
4405
4406        spin_lock(&nn->client_lock);
4407        last = oo->oo_last_closed_stid;
4408        oo->oo_last_closed_stid = s;
4409        list_move_tail(&oo->oo_close_lru, &nn->close_lru);
4410        oo->oo_time = ktime_get_boottime_seconds();
4411        spin_unlock(&nn->client_lock);
4412        if (last)
4413                nfs4_put_stid(&last->st_stid);
4414}
4415
4416/* search file_hashtbl[] for file */
4417static struct nfs4_file *
4418find_file_locked(struct knfsd_fh *fh, unsigned int hashval)
4419{
4420        struct nfs4_file *fp;
4421
4422        hlist_for_each_entry_rcu(fp, &file_hashtbl[hashval], fi_hash,
4423                                lockdep_is_held(&state_lock)) {
4424                if (fh_match(&fp->fi_fhandle, fh)) {
4425                        if (refcount_inc_not_zero(&fp->fi_ref))
4426                                return fp;
4427                }
4428        }
4429        return NULL;
4430}
4431
4432struct nfs4_file *
4433find_file(struct knfsd_fh *fh)
4434{
4435        struct nfs4_file *fp;
4436        unsigned int hashval = file_hashval(fh);
4437
4438        rcu_read_lock();
4439        fp = find_file_locked(fh, hashval);
4440        rcu_read_unlock();
4441        return fp;
4442}
4443
4444static struct nfs4_file *
4445find_or_add_file(struct nfs4_file *new, struct knfsd_fh *fh)
4446{
4447        struct nfs4_file *fp;
4448        unsigned int hashval = file_hashval(fh);
4449
4450        rcu_read_lock();
4451        fp = find_file_locked(fh, hashval);
4452        rcu_read_unlock();
4453        if (fp)
4454                return fp;
4455
4456        spin_lock(&state_lock);
4457        fp = find_file_locked(fh, hashval);
4458        if (likely(fp == NULL)) {
4459                nfsd4_init_file(fh, hashval, new);
4460                fp = new;
4461        }
4462        spin_unlock(&state_lock);
4463
4464        return fp;
4465}
4466
4467/*
4468 * Called to check deny when READ with all zero stateid or
4469 * WRITE with all zero or all one stateid
4470 */
4471static __be32
4472nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
4473{
4474        struct nfs4_file *fp;
4475        __be32 ret = nfs_ok;
4476
4477        fp = find_file(&current_fh->fh_handle);
4478        if (!fp)
4479                return ret;
4480        /* Check for conflicting share reservations */
4481        spin_lock(&fp->fi_lock);
4482        if (fp->fi_share_deny & deny_type)
4483                ret = nfserr_locked;
4484        spin_unlock(&fp->fi_lock);
4485        put_nfs4_file(fp);
4486        return ret;
4487}
4488
4489static void nfsd4_cb_recall_prepare(struct nfsd4_callback *cb)
4490{
4491        struct nfs4_delegation *dp = cb_to_delegation(cb);
4492        struct nfsd_net *nn = net_generic(dp->dl_stid.sc_client->net,
4493                                          nfsd_net_id);
4494
4495        block_delegations(&dp->dl_stid.sc_file->fi_fhandle);
4496
4497        /*
4498         * We can't do this in nfsd_break_deleg_cb because it is
4499         * already holding inode->i_lock.
4500         *
4501         * If the dl_time != 0, then we know that it has already been
4502         * queued for a lease break. Don't queue it again.
4503         */
4504        spin_lock(&state_lock);
4505        if (dp->dl_time == 0) {
4506                dp->dl_time = ktime_get_boottime_seconds();
4507                list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
4508        }
4509        spin_unlock(&state_lock);
4510}
4511
4512static int nfsd4_cb_recall_done(struct nfsd4_callback *cb,
4513                struct rpc_task *task)
4514{
4515        struct nfs4_delegation *dp = cb_to_delegation(cb);
4516
4517        if (dp->dl_stid.sc_type == NFS4_CLOSED_DELEG_STID ||
4518            dp->dl_stid.sc_type == NFS4_REVOKED_DELEG_STID)
4519                return 1;
4520
4521        switch (task->tk_status) {
4522        case 0:
4523                return 1;
4524        case -NFS4ERR_DELAY:
4525                rpc_delay(task, 2 * HZ);
4526                return 0;
4527        case -EBADHANDLE:
4528        case -NFS4ERR_BAD_STATEID:
4529                /*
4530                 * Race: client probably got cb_recall before open reply
4531                 * granting delegation.
4532                 */
4533                if (dp->dl_retries--) {
4534                        rpc_delay(task, 2 * HZ);
4535                        return 0;
4536                }
4537                fallthrough;
4538        default:
4539                return 1;
4540        }
4541}
4542
4543static void nfsd4_cb_recall_release(struct nfsd4_callback *cb)
4544{
4545        struct nfs4_delegation *dp = cb_to_delegation(cb);
4546
4547        nfs4_put_stid(&dp->dl_stid);
4548}
4549
4550static const struct nfsd4_callback_ops nfsd4_cb_recall_ops = {
4551        .prepare        = nfsd4_cb_recall_prepare,
4552        .done           = nfsd4_cb_recall_done,
4553        .release        = nfsd4_cb_recall_release,
4554};
4555
4556static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
4557{
4558        /*
4559         * We're assuming the state code never drops its reference
4560         * without first removing the lease.  Since we're in this lease
4561         * callback (and since the lease code is serialized by the
4562         * i_lock) we know the server hasn't removed the lease yet, and
4563         * we know it's safe to take a reference.
4564         */
4565        refcount_inc(&dp->dl_stid.sc_count);
4566        nfsd4_run_cb(&dp->dl_recall);
4567}
4568
4569/* Called from break_lease() with i_lock held. */
4570static bool
4571nfsd_break_deleg_cb(struct file_lock *fl)
4572{
4573        bool ret = false;
4574        struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
4575        struct nfs4_file *fp = dp->dl_stid.sc_file;
4576
4577        trace_nfsd_deleg_break(&dp->dl_stid.sc_stateid);
4578
4579        /*
4580         * We don't want the locks code to timeout the lease for us;
4581         * we'll remove it ourself if a delegation isn't returned
4582         * in time:
4583         */
4584        fl->fl_break_time = 0;
4585
4586        spin_lock(&fp->fi_lock);
4587        fp->fi_had_conflict = true;
4588        nfsd_break_one_deleg(dp);
4589        spin_unlock(&fp->fi_lock);
4590        return ret;
4591}
4592
4593static bool nfsd_breaker_owns_lease(struct file_lock *fl)
4594{
4595        struct nfs4_delegation *dl = fl->fl_owner;
4596        struct svc_rqst *rqst;
4597        struct nfs4_client *clp;
4598
4599        if (!i_am_nfsd())
4600                return NULL;
4601        rqst = kthread_data(current);
4602        /* Note rq_prog == NFS_ACL_PROGRAM is also possible: */
4603        if (rqst->rq_prog != NFS_PROGRAM || rqst->rq_vers < 4)
4604                return NULL;
4605        clp = *(rqst->rq_lease_breaker);
4606        return dl->dl_stid.sc_client == clp;
4607}
4608
4609static int
4610nfsd_change_deleg_cb(struct file_lock *onlist, int arg,
4611                     struct list_head *dispose)
4612{
4613        if (arg & F_UNLCK)
4614                return lease_modify(onlist, arg, dispose);
4615        else
4616                return -EAGAIN;
4617}
4618
4619static const struct lock_manager_operations nfsd_lease_mng_ops = {
4620        .lm_breaker_owns_lease = nfsd_breaker_owns_lease,
4621        .lm_break = nfsd_break_deleg_cb,
4622        .lm_change = nfsd_change_deleg_cb,
4623};
4624
4625static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
4626{
4627        if (nfsd4_has_session(cstate))
4628                return nfs_ok;
4629        if (seqid == so->so_seqid - 1)
4630                return nfserr_replay_me;
4631        if (seqid == so->so_seqid)
4632                return nfs_ok;
4633        return nfserr_bad_seqid;
4634}
4635
4636static struct nfs4_client *lookup_clientid(clientid_t *clid, bool sessions,
4637                                                struct nfsd_net *nn)
4638{
4639        struct nfs4_client *found;
4640
4641        spin_lock(&nn->client_lock);
4642        found = find_confirmed_client(clid, sessions, nn);
4643        if (found)
4644                atomic_inc(&found->cl_rpc_users);
4645        spin_unlock(&nn->client_lock);
4646        return found;
4647}
4648
4649static __be32 set_client(clientid_t *clid,
4650                struct nfsd4_compound_state *cstate,
4651                struct nfsd_net *nn)
4652{
4653        if (cstate->clp) {
4654                if (!same_clid(&cstate->clp->cl_clientid, clid))
4655                        return nfserr_stale_clientid;
4656                return nfs_ok;
4657        }
4658        if (STALE_CLIENTID(clid, nn))
4659                return nfserr_stale_clientid;
4660        /*
4661         * We're in the 4.0 case (otherwise the SEQUENCE op would have
4662         * set cstate->clp), so session = false:
4663         */
4664        cstate->clp = lookup_clientid(clid, false, nn);
4665        if (!cstate->clp)
4666                return nfserr_expired;
4667        return nfs_ok;
4668}
4669
4670__be32
4671nfsd4_process_open1(struct nfsd4_compound_state *cstate,
4672                    struct nfsd4_open *open, struct nfsd_net *nn)
4673{
4674        clientid_t *clientid = &open->op_clientid;
4675        struct nfs4_client *clp = NULL;
4676        unsigned int strhashval;
4677        struct nfs4_openowner *oo = NULL;
4678        __be32 status;
4679
4680        /*
4681         * In case we need it later, after we've already created the
4682         * file and don't want to risk a further failure:
4683         */
4684        open->op_file = nfsd4_alloc_file();
4685        if (open->op_file == NULL)
4686                return nfserr_jukebox;
4687
4688        status = set_client(clientid, cstate, nn);
4689        if (status)
4690                return status;
4691        clp = cstate->clp;
4692
4693        strhashval = ownerstr_hashval(&open->op_owner);
4694        oo = find_openstateowner_str(strhashval, open, clp);
4695        open->op_openowner = oo;
4696        if (!oo) {
4697                goto new_owner;
4698        }
4699        if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
4700                /* Replace unconfirmed owners without checking for replay. */
4701                release_openowner(oo);
4702                open->op_openowner = NULL;
4703                goto new_owner;
4704        }
4705        status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
4706        if (status)
4707                return status;
4708        goto alloc_stateid;
4709new_owner:
4710        oo = alloc_init_open_stateowner(strhashval, open, cstate);
4711        if (oo == NULL)
4712                return nfserr_jukebox;
4713        open->op_openowner = oo;
4714alloc_stateid:
4715        open->op_stp = nfs4_alloc_open_stateid(clp);
4716        if (!open->op_stp)
4717                return nfserr_jukebox;
4718
4719        if (nfsd4_has_session(cstate) &&
4720            (cstate->current_fh.fh_export->ex_flags & NFSEXP_PNFS)) {
4721                open->op_odstate = alloc_clnt_odstate(clp);
4722                if (!open->op_odstate)
4723                        return nfserr_jukebox;
4724        }
4725
4726        return nfs_ok;
4727}
4728
4729static inline __be32
4730nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
4731{
4732        if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
4733                return nfserr_openmode;
4734        else
4735                return nfs_ok;
4736}
4737
4738static int share_access_to_flags(u32 share_access)
4739{
4740        return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
4741}
4742
4743static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
4744{
4745        struct nfs4_stid *ret;
4746
4747        ret = find_stateid_by_type(cl, s,
4748                                NFS4_DELEG_STID|NFS4_REVOKED_DELEG_STID);
4749        if (!ret)
4750                return NULL;
4751        return delegstateid(ret);
4752}
4753
4754static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
4755{
4756        return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
4757               open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
4758}
4759
4760static __be32
4761nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
4762                struct nfs4_delegation **dp)
4763{
4764        int flags;
4765        __be32 status = nfserr_bad_stateid;
4766        struct nfs4_delegation *deleg;
4767
4768        deleg = find_deleg_stateid(cl, &open->op_delegate_stateid);
4769        if (deleg == NULL)
4770                goto out;
4771        if (deleg->dl_stid.sc_type == NFS4_REVOKED_DELEG_STID) {
4772                nfs4_put_stid(&deleg->dl_stid);
4773                if (cl->cl_minorversion)
4774                        status = nfserr_deleg_revoked;
4775                goto out;
4776        }
4777        flags = share_access_to_flags(open->op_share_access);
4778        status = nfs4_check_delegmode(deleg, flags);
4779        if (status) {
4780                nfs4_put_stid(&deleg->dl_stid);
4781                goto out;
4782        }
4783        *dp = deleg;
4784out:
4785        if (!nfsd4_is_deleg_cur(open))
4786                return nfs_ok;
4787        if (status)
4788                return status;
4789        open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
4790        return nfs_ok;
4791}
4792
4793static inline int nfs4_access_to_access(u32 nfs4_access)
4794{
4795        int flags = 0;
4796
4797        if (nfs4_access & NFS4_SHARE_ACCESS_READ)
4798                flags |= NFSD_MAY_READ;
4799        if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
4800                flags |= NFSD_MAY_WRITE;
4801        return flags;
4802}
4803
4804static inline __be32
4805nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
4806                struct nfsd4_open *open)
4807{
4808        struct iattr iattr = {
4809                .ia_valid = ATTR_SIZE,
4810                .ia_size = 0,
4811        };
4812        if (!open->op_truncate)
4813                return 0;
4814        if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
4815                return nfserr_inval;
4816        return nfsd_setattr(rqstp, fh, &iattr, 0, (time64_t)0);
4817}
4818
4819static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
4820                struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
4821                struct nfsd4_open *open)
4822{
4823        struct nfsd_file *nf = NULL;
4824        __be32 status;
4825        int oflag = nfs4_access_to_omode(open->op_share_access);
4826        int access = nfs4_access_to_access(open->op_share_access);
4827        unsigned char old_access_bmap, old_deny_bmap;
4828
4829        spin_lock(&fp->fi_lock);
4830
4831        /*
4832         * Are we trying to set a deny mode that would conflict with
4833         * current access?
4834         */
4835        status = nfs4_file_check_deny(fp, open->op_share_deny);
4836        if (status != nfs_ok) {
4837                spin_unlock(&fp->fi_lock);
4838                goto out;
4839        }
4840
4841        /* set access to the file */
4842        status = nfs4_file_get_access(fp, open->op_share_access);
4843        if (status != nfs_ok) {
4844                spin_unlock(&fp->fi_lock);
4845                goto out;
4846        }
4847
4848        /* Set access bits in stateid */
4849        old_access_bmap = stp->st_access_bmap;
4850        set_access(open->op_share_access, stp);
4851
4852        /* Set new deny mask */
4853        old_deny_bmap = stp->st_deny_bmap;
4854        set_deny(open->op_share_deny, stp);
4855        fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
4856
4857        if (!fp->fi_fds[oflag]) {
4858                spin_unlock(&fp->fi_lock);
4859                status = nfsd_file_acquire(rqstp, cur_fh, access, &nf);
4860                if (status)
4861                        goto out_put_access;
4862                spin_lock(&fp->fi_lock);
4863                if (!fp->fi_fds[oflag]) {
4864                        fp->fi_fds[oflag] = nf;
4865                        nf = NULL;
4866                }
4867        }
4868        spin_unlock(&fp->fi_lock);
4869        if (nf)
4870                nfsd_file_put(nf);
4871
4872        status = nfsd4_truncate(rqstp, cur_fh, open);
4873        if (status)
4874                goto out_put_access;
4875out:
4876        return status;
4877out_put_access:
4878        stp->st_access_bmap = old_access_bmap;
4879        nfs4_file_put_access(fp, open->op_share_access);
4880        reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap), stp);
4881        goto out;
4882}
4883
4884static __be32
4885nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp, struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp, struct nfsd4_open *open)
4886{
4887        __be32 status;
4888        unsigned char old_deny_bmap = stp->st_deny_bmap;
4889
4890        if (!test_access(open->op_share_access, stp))
4891                return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open);
4892
4893        /* test and set deny mode */
4894        spin_lock(&fp->fi_lock);
4895        status = nfs4_file_check_deny(fp, open->op_share_deny);
4896        if (status == nfs_ok) {
4897                set_deny(open->op_share_deny, stp);
4898                fp->fi_share_deny |=
4899                                (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
4900        }
4901        spin_unlock(&fp->fi_lock);
4902
4903        if (status != nfs_ok)
4904                return status;
4905
4906        status = nfsd4_truncate(rqstp, cur_fh, open);
4907        if (status != nfs_ok)
4908                reset_union_bmap_deny(old_deny_bmap, stp);
4909        return status;
4910}
4911
4912/* Should we give out recallable state?: */
4913static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
4914{
4915        if (clp->cl_cb_state == NFSD4_CB_UP)
4916                return true;
4917        /*
4918         * In the sessions case, since we don't have to establish a
4919         * separate connection for callbacks, we assume it's OK
4920         * until we hear otherwise:
4921         */
4922        return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
4923}
4924
4925static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp,
4926                                                int flag)
4927{
4928        struct file_lock *fl;
4929
4930        fl = locks_alloc_lock();
4931        if (!fl)
4932                return NULL;
4933        fl->fl_lmops = &nfsd_lease_mng_ops;
4934        fl->fl_flags = FL_DELEG;
4935        fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
4936        fl->fl_end = OFFSET_MAX;
4937        fl->fl_owner = (fl_owner_t)dp;
4938        fl->fl_pid = current->tgid;
4939        fl->fl_file = dp->dl_stid.sc_file->fi_deleg_file->nf_file;
4940        return fl;
4941}
4942
4943static struct nfs4_delegation *
4944nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh,
4945                    struct nfs4_file *fp, struct nfs4_clnt_odstate *odstate)
4946{
4947        int status = 0;
4948        struct nfs4_delegation *dp;
4949        struct nfsd_file *nf;
4950        struct file_lock *fl;
4951
4952        /*
4953         * The fi_had_conflict and nfs_get_existing_delegation checks
4954         * here are just optimizations; we'll need to recheck them at
4955         * the end:
4956         */
4957        if (fp->fi_had_conflict)
4958                return ERR_PTR(-EAGAIN);
4959
4960        nf = find_readable_file(fp);
4961        if (!nf) {
4962                /* We should always have a readable file here */
4963                WARN_ON_ONCE(1);
4964                return ERR_PTR(-EBADF);
4965        }
4966        spin_lock(&state_lock);
4967        spin_lock(&fp->fi_lock);
4968        if (nfs4_delegation_exists(clp, fp))
4969                status = -EAGAIN;
4970        else if (!fp->fi_deleg_file) {
4971                fp->fi_deleg_file = nf;
4972                /* increment early to prevent fi_deleg_file from being
4973                 * cleared */
4974                fp->fi_delegees = 1;
4975                nf = NULL;
4976        } else
4977                fp->fi_delegees++;
4978        spin_unlock(&fp->fi_lock);
4979        spin_unlock(&state_lock);
4980        if (nf)
4981                nfsd_file_put(nf);
4982        if (status)
4983                return ERR_PTR(status);
4984
4985        status = -ENOMEM;
4986        dp = alloc_init_deleg(clp, fp, fh, odstate);
4987        if (!dp)
4988                goto out_delegees;
4989
4990        fl = nfs4_alloc_init_lease(dp, NFS4_OPEN_DELEGATE_READ);
4991        if (!fl)
4992                goto out_clnt_odstate;
4993
4994        status = vfs_setlease(fp->fi_deleg_file->nf_file, fl->fl_type, &fl, NULL);
4995        if (fl)
4996                locks_free_lock(fl);
4997        if (status)
4998                goto out_clnt_odstate;
4999
5000        spin_lock(&state_lock);
5001        spin_lock(&fp->fi_lock);
5002        if (fp->fi_had_conflict)
5003                status = -EAGAIN;
5004        else
5005                status = hash_delegation_locked(dp, fp);
5006        spin_unlock(&fp->fi_lock);
5007        spin_unlock(&state_lock);
5008
5009        if (status)
5010                goto out_unlock;
5011
5012        return dp;
5013out_unlock:
5014        vfs_setlease(fp->fi_deleg_file->nf_file, F_UNLCK, NULL, (void **)&dp);
5015out_clnt_odstate:
5016        put_clnt_odstate(dp->dl_clnt_odstate);
5017        nfs4_put_stid(&dp->dl_stid);
5018out_delegees:
5019        put_deleg_file(fp);
5020        return ERR_PTR(status);
5021}
5022
5023static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
5024{
5025        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5026        if (status == -EAGAIN)
5027                open->op_why_no_deleg = WND4_CONTENTION;
5028        else {
5029                open->op_why_no_deleg = WND4_RESOURCE;
5030                switch (open->op_deleg_want) {
5031                case NFS4_SHARE_WANT_READ_DELEG:
5032                case NFS4_SHARE_WANT_WRITE_DELEG:
5033                case NFS4_SHARE_WANT_ANY_DELEG:
5034                        break;
5035                case NFS4_SHARE_WANT_CANCEL:
5036                        open->op_why_no_deleg = WND4_CANCELLED;
5037                        break;
5038                case NFS4_SHARE_WANT_NO_DELEG:
5039                        WARN_ON_ONCE(1);
5040                }
5041        }
5042}
5043
5044/*
5045 * Attempt to hand out a delegation.
5046 *
5047 * Note we don't support write delegations, and won't until the vfs has
5048 * proper support for them.
5049 */
5050static void
5051nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open,
5052                        struct nfs4_ol_stateid *stp)
5053{
5054        struct nfs4_delegation *dp;
5055        struct nfs4_openowner *oo = openowner(stp->st_stateowner);
5056        struct nfs4_client *clp = stp->st_stid.sc_client;
5057        int cb_up;
5058        int status = 0;
5059
5060        cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
5061        open->op_recall = 0;
5062        switch (open->op_claim_type) {
5063                case NFS4_OPEN_CLAIM_PREVIOUS:
5064                        if (!cb_up)
5065                                open->op_recall = 1;
5066                        if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
5067                                goto out_no_deleg;
5068                        break;
5069                case NFS4_OPEN_CLAIM_NULL:
5070                case NFS4_OPEN_CLAIM_FH:
5071                        /*
5072                         * Let's not give out any delegations till everyone's
5073                         * had the chance to reclaim theirs, *and* until
5074                         * NLM locks have all been reclaimed:
5075                         */
5076                        if (locks_in_grace(clp->net))
5077                                goto out_no_deleg;
5078                        if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
5079                                goto out_no_deleg;
5080                        /*
5081                         * Also, if the file was opened for write or
5082                         * create, there's a good chance the client's
5083                         * about to write to it, resulting in an
5084                         * immediate recall (since we don't support
5085                         * write delegations):
5086                         */
5087                        if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
5088                                goto out_no_deleg;
5089                        if (open->op_create == NFS4_OPEN_CREATE)
5090                                goto out_no_deleg;
5091                        break;
5092                default:
5093                        goto out_no_deleg;
5094        }
5095        dp = nfs4_set_delegation(clp, fh, stp->st_stid.sc_file, stp->st_clnt_odstate);
5096        if (IS_ERR(dp))
5097                goto out_no_deleg;
5098
5099        memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
5100
5101        trace_nfsd_deleg_read(&dp->dl_stid.sc_stateid);
5102        open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
5103        nfs4_put_stid(&dp->dl_stid);
5104        return;
5105out_no_deleg:
5106        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
5107        if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
5108            open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
5109                dprintk("NFSD: WARNING: refusing delegation reclaim\n");
5110                open->op_recall = 1;
5111        }
5112
5113        /* 4.1 client asking for a delegation? */
5114        if (open->op_deleg_want)
5115                nfsd4_open_deleg_none_ext(open, status);
5116        return;
5117}
5118
5119static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
5120                                        struct nfs4_delegation *dp)
5121{
5122        if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
5123            dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
5124                open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5125                open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
5126        } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
5127                   dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
5128                open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5129                open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
5130        }
5131        /* Otherwise the client must be confused wanting a delegation
5132         * it already has, therefore we don't return
5133         * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
5134         */
5135}
5136
5137__be32
5138nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
5139{
5140        struct nfsd4_compoundres *resp = rqstp->rq_resp;
5141        struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
5142        struct nfs4_file *fp = NULL;
5143        struct nfs4_ol_stateid *stp = NULL;
5144        struct nfs4_delegation *dp = NULL;
5145        __be32 status;
5146        bool new_stp = false;
5147
5148        /*
5149         * Lookup file; if found, lookup stateid and check open request,
5150         * and check for delegations in the process of being recalled.
5151         * If not found, create the nfs4_file struct
5152         */
5153        fp = find_or_add_file(open->op_file, &current_fh->fh_handle);
5154        if (fp != open->op_file) {
5155                status = nfs4_check_deleg(cl, open, &dp);
5156                if (status)
5157                        goto out;
5158                stp = nfsd4_find_and_lock_existing_open(fp, open);
5159        } else {
5160                open->op_file = NULL;
5161                status = nfserr_bad_stateid;
5162                if (nfsd4_is_deleg_cur(open))
5163                        goto out;
5164        }
5165
5166        if (!stp) {
5167                stp = init_open_stateid(fp, open);
5168                if (!open->op_stp)
5169                        new_stp = true;
5170        }
5171
5172        /*
5173         * OPEN the file, or upgrade an existing OPEN.
5174         * If truncate fails, the OPEN fails.
5175         *
5176         * stp is already locked.
5177         */
5178        if (!new_stp) {
5179                /* Stateid was found, this is an OPEN upgrade */
5180                status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
5181                if (status) {
5182                        mutex_unlock(&stp->st_mutex);
5183                        goto out;
5184                }
5185        } else {
5186                status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open);
5187                if (status) {
5188                        stp->st_stid.sc_type = NFS4_CLOSED_STID;
5189                        release_open_stateid(stp);
5190                        mutex_unlock(&stp->st_mutex);
5191                        goto out;
5192                }
5193
5194                stp->st_clnt_odstate = find_or_hash_clnt_odstate(fp,
5195                                                        open->op_odstate);
5196                if (stp->st_clnt_odstate == open->op_odstate)
5197                        open->op_odstate = NULL;
5198        }
5199
5200        nfs4_inc_and_copy_stateid(&open->op_stateid, &stp->st_stid);
5201        mutex_unlock(&stp->st_mutex);
5202
5203        if (nfsd4_has_session(&resp->cstate)) {
5204                if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
5205                        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5206                        open->op_why_no_deleg = WND4_NOT_WANTED;
5207                        goto nodeleg;
5208                }
5209        }
5210
5211        /*
5212        * Attempt to hand out a delegation. No error return, because the
5213        * OPEN succeeds even if we fail.
5214        */
5215        nfs4_open_delegation(current_fh, open, stp);
5216nodeleg:
5217        status = nfs_ok;
5218        trace_nfsd_open(&stp->st_stid.sc_stateid);
5219out:
5220        /* 4.1 client trying to upgrade/downgrade delegation? */
5221        if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
5222            open->op_deleg_want)
5223                nfsd4_deleg_xgrade_none_ext(open, dp);
5224
5225        if (fp)
5226                put_nfs4_file(fp);
5227        if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
5228                open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
5229        /*
5230        * To finish the open response, we just need to set the rflags.
5231        */
5232        open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
5233        if (nfsd4_has_session(&resp->cstate))
5234                open->op_rflags |= NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK;
5235        else if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED))
5236                open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
5237
5238        if (dp)
5239                nfs4_put_stid(&dp->dl_stid);
5240        if (stp)
5241                nfs4_put_stid(&stp->st_stid);
5242
5243        return status;
5244}
5245
5246void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
5247                              struct nfsd4_open *open)
5248{
5249        if (open->op_openowner) {
5250                struct nfs4_stateowner *so = &open->op_openowner->oo_owner;
5251
5252                nfsd4_cstate_assign_replay(cstate, so);
5253                nfs4_put_stateowner(so);
5254        }
5255        if (open->op_file)
5256                kmem_cache_free(file_slab, open->op_file);
5257        if (open->op_stp)
5258                nfs4_put_stid(&open->op_stp->st_stid);
5259        if (open->op_odstate)
5260                kmem_cache_free(odstate_slab, open->op_odstate);
5261}
5262
5263__be32
5264nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5265            union nfsd4_op_u *u)
5266{
5267        clientid_t *clid = &u->renew;
5268        struct nfs4_client *clp;
5269        __be32 status;
5270        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5271
5272        trace_nfsd_clid_renew(clid);
5273        status = set_client(clid, cstate, nn);
5274        if (status)
5275                return status;
5276        clp = cstate->clp;
5277        if (!list_empty(&clp->cl_delegations)
5278                        && clp->cl_cb_state != NFSD4_CB_UP)
5279                return nfserr_cb_path_down;
5280        return nfs_ok;
5281}
5282
5283void
5284nfsd4_end_grace(struct nfsd_net *nn)
5285{
5286        /* do nothing if grace period already ended */
5287        if (nn->grace_ended)
5288                return;
5289
5290        trace_nfsd_grace_complete(nn);
5291        nn->grace_ended = true;
5292        /*
5293         * If the server goes down again right now, an NFSv4
5294         * client will still be allowed to reclaim after it comes back up,
5295         * even if it hasn't yet had a chance to reclaim state this time.
5296         *
5297         */
5298        nfsd4_record_grace_done(nn);
5299        /*
5300         * At this point, NFSv4 clients can still reclaim.  But if the
5301         * server crashes, any that have not yet reclaimed will be out
5302         * of luck on the next boot.
5303         *
5304         * (NFSv4.1+ clients are considered to have reclaimed once they
5305         * call RECLAIM_COMPLETE.  NFSv4.0 clients are considered to
5306         * have reclaimed after their first OPEN.)
5307         */
5308        locks_end_grace(&nn->nfsd4_manager);
5309        /*
5310         * At this point, and once lockd and/or any other containers
5311         * exit their grace period, further reclaims will fail and
5312         * regular locking can resume.
5313         */
5314}
5315
5316/*
5317 * If we've waited a lease period but there are still clients trying to
5318 * reclaim, wait a little longer to give them a chance to finish.
5319 */
5320static bool clients_still_reclaiming(struct nfsd_net *nn)
5321{
5322        time64_t double_grace_period_end = nn->boot_time +
5323                                           2 * nn->nfsd4_lease;
5324
5325        if (nn->track_reclaim_completes &&
5326                        atomic_read(&nn->nr_reclaim_complete) ==
5327                        nn->reclaim_str_hashtbl_size)
5328                return false;
5329        if (!nn->somebody_reclaimed)
5330                return false;
5331        nn->somebody_reclaimed = false;
5332        /*
5333         * If we've given them *two* lease times to reclaim, and they're
5334         * still not done, give up:
5335         */
5336        if (ktime_get_boottime_seconds() > double_grace_period_end)
5337                return false;
5338        return true;
5339}
5340
5341static time64_t
5342nfs4_laundromat(struct nfsd_net *nn)
5343{
5344        struct nfs4_client *clp;
5345        struct nfs4_openowner *oo;
5346        struct nfs4_delegation *dp;
5347        struct nfs4_ol_stateid *stp;
5348        struct nfsd4_blocked_lock *nbl;
5349        struct list_head *pos, *next, reaplist;
5350        time64_t cutoff = ktime_get_boottime_seconds() - nn->nfsd4_lease;
5351        time64_t t, new_timeo = nn->nfsd4_lease;
5352        struct nfs4_cpntf_state *cps;
5353        copy_stateid_t *cps_t;
5354        int i;
5355
5356        if (clients_still_reclaiming(nn)) {
5357                new_timeo = 0;
5358                goto out;
5359        }
5360        nfsd4_end_grace(nn);
5361        INIT_LIST_HEAD(&reaplist);
5362
5363        spin_lock(&nn->s2s_cp_lock);
5364        idr_for_each_entry(&nn->s2s_cp_stateids, cps_t, i) {
5365                cps = container_of(cps_t, struct nfs4_cpntf_state, cp_stateid);
5366                if (cps->cp_stateid.sc_type == NFS4_COPYNOTIFY_STID &&
5367                                cps->cpntf_time < cutoff)
5368                        _free_cpntf_state_locked(nn, cps);
5369        }
5370        spin_unlock(&nn->s2s_cp_lock);
5371
5372        spin_lock(&nn->client_lock);
5373        list_for_each_safe(pos, next, &nn->client_lru) {
5374                clp = list_entry(pos, struct nfs4_client, cl_lru);
5375                if (clp->cl_time > cutoff) {
5376                        t = clp->cl_time - cutoff;
5377                        new_timeo = min(new_timeo, t);
5378                        break;
5379                }
5380                if (mark_client_expired_locked(clp)) {
5381                        trace_nfsd_clid_expired(&clp->cl_clientid);
5382                        continue;
5383                }
5384                list_add(&clp->cl_lru, &reaplist);
5385        }
5386        spin_unlock(&nn->client_lock);
5387        list_for_each_safe(pos, next, &reaplist) {
5388                clp = list_entry(pos, struct nfs4_client, cl_lru);
5389                trace_nfsd_clid_purged(&clp->cl_clientid);
5390                list_del_init(&clp->cl_lru);
5391                expire_client(clp);
5392        }
5393        spin_lock(&state_lock);
5394        list_for_each_safe(pos, next, &nn->del_recall_lru) {
5395                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
5396                if (dp->dl_time > cutoff) {
5397                        t = dp->dl_time - cutoff;
5398                        new_timeo = min(new_timeo, t);
5399                        break;
5400                }
5401                WARN_ON(!unhash_delegation_locked(dp));
5402                list_add(&dp->dl_recall_lru, &reaplist);
5403        }
5404        spin_unlock(&state_lock);
5405        while (!list_empty(&reaplist)) {
5406                dp = list_first_entry(&reaplist, struct nfs4_delegation,
5407                                        dl_recall_lru);
5408                list_del_init(&dp->dl_recall_lru);
5409                revoke_delegation(dp);
5410        }
5411
5412        spin_lock(&nn->client_lock);
5413        while (!list_empty(&nn->close_lru)) {
5414                oo = list_first_entry(&nn->close_lru, struct nfs4_openowner,
5415                                        oo_close_lru);
5416                if (oo->oo_time > cutoff) {
5417                        t = oo->oo_time - cutoff;
5418                        new_timeo = min(new_timeo, t);
5419                        break;
5420                }
5421                list_del_init(&oo->oo_close_lru);
5422                stp = oo->oo_last_closed_stid;
5423                oo->oo_last_closed_stid = NULL;
5424                spin_unlock(&nn->client_lock);
5425                nfs4_put_stid(&stp->st_stid);
5426                spin_lock(&nn->client_lock);
5427        }
5428        spin_unlock(&nn->client_lock);
5429
5430        /*
5431         * It's possible for a client to try and acquire an already held lock
5432         * that is being held for a long time, and then lose interest in it.
5433         * So, we clean out any un-revisited request after a lease period
5434         * under the assumption that the client is no longer interested.
5435         *
5436         * RFC5661, sec. 9.6 states that the client must not rely on getting
5437         * notifications and must continue to poll for locks, even when the
5438         * server supports them. Thus this shouldn't lead to clients blocking
5439         * indefinitely once the lock does become free.
5440         */
5441        BUG_ON(!list_empty(&reaplist));
5442        spin_lock(&nn->blocked_locks_lock);
5443        while (!list_empty(&nn->blocked_locks_lru)) {
5444                nbl = list_first_entry(&nn->blocked_locks_lru,
5445                                        struct nfsd4_blocked_lock, nbl_lru);
5446                if (nbl->nbl_time > cutoff) {
5447                        t = nbl->nbl_time - cutoff;
5448                        new_timeo = min(new_timeo, t);
5449                        break;
5450                }
5451                list_move(&nbl->nbl_lru, &reaplist);
5452                list_del_init(&nbl->nbl_list);
5453        }
5454        spin_unlock(&nn->blocked_locks_lock);
5455
5456        while (!list_empty(&reaplist)) {
5457                nbl = list_first_entry(&reaplist,
5458                                        struct nfsd4_blocked_lock, nbl_lru);
5459                list_del_init(&nbl->nbl_lru);
5460                free_blocked_lock(nbl);
5461        }
5462out:
5463        new_timeo = max_t(time64_t, new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
5464        return new_timeo;
5465}
5466
5467static struct workqueue_struct *laundry_wq;
5468static void laundromat_main(struct work_struct *);
5469
5470static void
5471laundromat_main(struct work_struct *laundry)
5472{
5473        time64_t t;
5474        struct delayed_work *dwork = to_delayed_work(laundry);
5475        struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
5476                                           laundromat_work);
5477
5478        t = nfs4_laundromat(nn);
5479        queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
5480}
5481
5482static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stid *stp)
5483{
5484        if (!fh_match(&fhp->fh_handle, &stp->sc_file->fi_fhandle))
5485                return nfserr_bad_stateid;
5486        return nfs_ok;
5487}
5488
5489static inline int
5490access_permit_read(struct nfs4_ol_stateid *stp)
5491{
5492        return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
5493                test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
5494                test_access(NFS4_SHARE_ACCESS_WRITE, stp);
5495}
5496
5497static inline int
5498access_permit_write(struct nfs4_ol_stateid *stp)
5499{
5500        return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
5501                test_access(NFS4_SHARE_ACCESS_BOTH, stp);
5502}
5503
5504static
5505__be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
5506{
5507        __be32 status = nfserr_openmode;
5508
5509        /* For lock stateid's, we test the parent open, not the lock: */
5510        if (stp->st_openstp)
5511                stp = stp->st_openstp;
5512        if ((flags & WR_STATE) && !access_permit_write(stp))
5513                goto out;
5514        if ((flags & RD_STATE) && !access_permit_read(stp))
5515                goto out;
5516        status = nfs_ok;
5517out:
5518        return status;
5519}
5520
5521static inline __be32
5522check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
5523{
5524        if (ONE_STATEID(stateid) && (flags & RD_STATE))
5525                return nfs_ok;
5526        else if (opens_in_grace(net)) {
5527                /* Answer in remaining cases depends on existence of
5528                 * conflicting state; so we must wait out the grace period. */
5529                return nfserr_grace;
5530        } else if (flags & WR_STATE)
5531                return nfs4_share_conflict(current_fh,
5532                                NFS4_SHARE_DENY_WRITE);
5533        else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
5534                return nfs4_share_conflict(current_fh,
5535                                NFS4_SHARE_DENY_READ);
5536}
5537
5538/*
5539 * Allow READ/WRITE during grace period on recovered state only for files
5540 * that are not able to provide mandatory locking.
5541 */
5542static inline int
5543grace_disallows_io(struct net *net, struct inode *inode)
5544{
5545        return opens_in_grace(net) && mandatory_lock(inode);
5546}
5547
5548static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
5549{
5550        /*
5551         * When sessions are used the stateid generation number is ignored
5552         * when it is zero.
5553         */
5554        if (has_session && in->si_generation == 0)
5555                return nfs_ok;
5556
5557        if (in->si_generation == ref->si_generation)
5558                return nfs_ok;
5559
5560        /* If the client sends us a stateid from the future, it's buggy: */
5561        if (nfsd4_stateid_generation_after(in, ref))
5562                return nfserr_bad_stateid;
5563        /*
5564         * However, we could see a stateid from the past, even from a
5565         * non-buggy client.  For example, if the client sends a lock
5566         * while some IO is outstanding, the lock may bump si_generation
5567         * while the IO is still in flight.  The client could avoid that
5568         * situation by waiting for responses on all the IO requests,
5569         * but better performance may result in retrying IO that
5570         * receives an old_stateid error if requests are rarely
5571         * reordered in flight:
5572         */
5573        return nfserr_old_stateid;
5574}
5575
5576static __be32 nfsd4_stid_check_stateid_generation(stateid_t *in, struct nfs4_stid *s, bool has_session)
5577{
5578        __be32 ret;
5579
5580        spin_lock(&s->sc_lock);
5581        ret = nfsd4_verify_open_stid(s);
5582        if (ret == nfs_ok)
5583                ret = check_stateid_generation(in, &s->sc_stateid, has_session);
5584        spin_unlock(&s->sc_lock);
5585        return ret;
5586}
5587
5588static __be32 nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid *ols)
5589{
5590        if (ols->st_stateowner->so_is_open_owner &&
5591            !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
5592                return nfserr_bad_stateid;
5593        return nfs_ok;
5594}
5595
5596static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
5597{
5598        struct nfs4_stid *s;
5599        __be32 status = nfserr_bad_stateid;
5600
5601        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
5602                CLOSE_STATEID(stateid))
5603                return status;
5604        if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid))
5605                return status;
5606        spin_lock(&cl->cl_lock);
5607        s = find_stateid_locked(cl, stateid);
5608        if (!s)
5609                goto out_unlock;
5610        status = nfsd4_stid_check_stateid_generation(stateid, s, 1);
5611        if (status)
5612                goto out_unlock;
5613        switch (s->sc_type) {
5614        case NFS4_DELEG_STID:
5615                status = nfs_ok;
5616                break;
5617        case NFS4_REVOKED_DELEG_STID:
5618                status = nfserr_deleg_revoked;
5619                break;
5620        case NFS4_OPEN_STID:
5621        case NFS4_LOCK_STID:
5622                status = nfsd4_check_openowner_confirmed(openlockstateid(s));
5623                break;
5624        default:
5625                printk("unknown stateid type %x\n", s->sc_type);
5626                fallthrough;
5627        case NFS4_CLOSED_STID:
5628        case NFS4_CLOSED_DELEG_STID:
5629                status = nfserr_bad_stateid;
5630        }
5631out_unlock:
5632        spin_unlock(&cl->cl_lock);
5633        return status;
5634}
5635
5636__be32
5637nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
5638                     stateid_t *stateid, unsigned char typemask,
5639                     struct nfs4_stid **s, struct nfsd_net *nn)
5640{
5641        __be32 status;
5642        bool return_revoked = false;
5643
5644        /*
5645         *  only return revoked delegations if explicitly asked.
5646         *  otherwise we report revoked or bad_stateid status.
5647         */
5648        if (typemask & NFS4_REVOKED_DELEG_STID)
5649                return_revoked = true;
5650        else if (typemask & NFS4_DELEG_STID)
5651                typemask |= NFS4_REVOKED_DELEG_STID;
5652
5653        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
5654                CLOSE_STATEID(stateid))
5655                return nfserr_bad_stateid;
5656        status = set_client(&stateid->si_opaque.so_clid, cstate, nn);
5657        if (status == nfserr_stale_clientid) {
5658                if (cstate->session)
5659                        return nfserr_bad_stateid;
5660                return nfserr_stale_stateid;
5661        }
5662        if (status)
5663                return status;
5664        *s = find_stateid_by_type(cstate->clp, stateid, typemask);
5665        if (!*s)
5666                return nfserr_bad_stateid;
5667        if (((*s)->sc_type == NFS4_REVOKED_DELEG_STID) && !return_revoked) {
5668                nfs4_put_stid(*s);
5669                if (cstate->minorversion)
5670                        return nfserr_deleg_revoked;
5671                return nfserr_bad_stateid;
5672        }
5673        return nfs_ok;
5674}
5675
5676static struct nfsd_file *
5677nfs4_find_file(struct nfs4_stid *s, int flags)
5678{
5679        if (!s)
5680                return NULL;
5681
5682        switch (s->sc_type) {
5683        case NFS4_DELEG_STID:
5684                if (WARN_ON_ONCE(!s->sc_file->fi_deleg_file))
5685                        return NULL;
5686                return nfsd_file_get(s->sc_file->fi_deleg_file);
5687        case NFS4_OPEN_STID:
5688        case NFS4_LOCK_STID:
5689                if (flags & RD_STATE)
5690                        return find_readable_file(s->sc_file);
5691                else
5692                        return find_writeable_file(s->sc_file);
5693        }
5694
5695        return NULL;
5696}
5697
5698static __be32
5699nfs4_check_olstateid(struct nfs4_ol_stateid *ols, int flags)
5700{
5701        __be32 status;
5702
5703        status = nfsd4_check_openowner_confirmed(ols);
5704        if (status)
5705                return status;
5706        return nfs4_check_openmode(ols, flags);
5707}
5708
5709static __be32
5710nfs4_check_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct nfs4_stid *s,
5711                struct nfsd_file **nfp, int flags)
5712{
5713        int acc = (flags & RD_STATE) ? NFSD_MAY_READ : NFSD_MAY_WRITE;
5714        struct nfsd_file *nf;
5715        __be32 status;
5716
5717        nf = nfs4_find_file(s, flags);
5718        if (nf) {
5719                status = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
5720                                acc | NFSD_MAY_OWNER_OVERRIDE);
5721                if (status) {
5722                        nfsd_file_put(nf);
5723                        goto out;
5724                }
5725        } else {
5726                status = nfsd_file_acquire(rqstp, fhp, acc, &nf);
5727                if (status)
5728                        return status;
5729        }
5730        *nfp = nf;
5731out:
5732        return status;
5733}
5734static void
5735_free_cpntf_state_locked(struct nfsd_net *nn, struct nfs4_cpntf_state *cps)
5736{
5737        WARN_ON_ONCE(cps->cp_stateid.sc_type != NFS4_COPYNOTIFY_STID);
5738        if (!refcount_dec_and_test(&cps->cp_stateid.sc_count))
5739                return;
5740        list_del(&cps->cp_list);
5741        idr_remove(&nn->s2s_cp_stateids,
5742                   cps->cp_stateid.stid.si_opaque.so_id);
5743        kfree(cps);
5744}
5745/*
5746 * A READ from an inter server to server COPY will have a
5747 * copy stateid. Look up the copy notify stateid from the
5748 * idr structure and take a reference on it.
5749 */
5750__be32 manage_cpntf_state(struct nfsd_net *nn, stateid_t *st,
5751                          struct nfs4_client *clp,
5752                          struct nfs4_cpntf_state **cps)
5753{
5754        copy_stateid_t *cps_t;
5755        struct nfs4_cpntf_state *state = NULL;
5756
5757        if (st->si_opaque.so_clid.cl_id != nn->s2s_cp_cl_id)
5758                return nfserr_bad_stateid;
5759        spin_lock(&nn->s2s_cp_lock);
5760        cps_t = idr_find(&nn->s2s_cp_stateids, st->si_opaque.so_id);
5761        if (cps_t) {
5762                state = container_of(cps_t, struct nfs4_cpntf_state,
5763                                     cp_stateid);
5764                if (state->cp_stateid.sc_type != NFS4_COPYNOTIFY_STID) {
5765                        state = NULL;
5766                        goto unlock;
5767                }
5768                if (!clp)
5769                        refcount_inc(&state->cp_stateid.sc_count);
5770                else
5771                        _free_cpntf_state_locked(nn, state);
5772        }
5773unlock:
5774        spin_unlock(&nn->s2s_cp_lock);
5775        if (!state)
5776                return nfserr_bad_stateid;
5777        if (!clp && state)
5778                *cps = state;
5779        return 0;
5780}
5781
5782static __be32 find_cpntf_state(struct nfsd_net *nn, stateid_t *st,
5783                               struct nfs4_stid **stid)
5784{
5785        __be32 status;
5786        struct nfs4_cpntf_state *cps = NULL;
5787        struct nfs4_client *found;
5788
5789        status = manage_cpntf_state(nn, st, NULL, &cps);
5790        if (status)
5791                return status;
5792
5793        cps->cpntf_time = ktime_get_boottime_seconds();
5794
5795        status = nfserr_expired;
5796        found = lookup_clientid(&cps->cp_p_clid, true, nn);
5797        if (!found)
5798                goto out;
5799
5800        *stid = find_stateid_by_type(found, &cps->cp_p_stateid,
5801                        NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID);
5802        if (*stid)
5803                status = nfs_ok;
5804        else
5805                status = nfserr_bad_stateid;
5806
5807        put_client_renew(found);
5808out:
5809        nfs4_put_cpntf_state(nn, cps);
5810        return status;
5811}
5812
5813void nfs4_put_cpntf_state(struct nfsd_net *nn, struct nfs4_cpntf_state *cps)
5814{
5815        spin_lock(&nn->s2s_cp_lock);
5816        _free_cpntf_state_locked(nn, cps);
5817        spin_unlock(&nn->s2s_cp_lock);
5818}
5819
5820/*
5821 * Checks for stateid operations
5822 */
5823__be32
5824nfs4_preprocess_stateid_op(struct svc_rqst *rqstp,
5825                struct nfsd4_compound_state *cstate, struct svc_fh *fhp,
5826                stateid_t *stateid, int flags, struct nfsd_file **nfp,
5827                struct nfs4_stid **cstid)
5828{
5829        struct inode *ino = d_inode(fhp->fh_dentry);
5830        struct net *net = SVC_NET(rqstp);
5831        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5832        struct nfs4_stid *s = NULL;
5833        __be32 status;
5834
5835        if (nfp)
5836                *nfp = NULL;
5837
5838        if (grace_disallows_io(net, ino))
5839                return nfserr_grace;
5840
5841        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
5842                status = check_special_stateids(net, fhp, stateid, flags);
5843                goto done;
5844        }
5845
5846        status = nfsd4_lookup_stateid(cstate, stateid,
5847                                NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
5848                                &s, nn);
5849        if (status == nfserr_bad_stateid)
5850                status = find_cpntf_state(nn, stateid, &s);
5851        if (status)
5852                return status;
5853        status = nfsd4_stid_check_stateid_generation(stateid, s,
5854                        nfsd4_has_session(cstate));
5855        if (status)
5856                goto out;
5857
5858        switch (s->sc_type) {
5859        case NFS4_DELEG_STID:
5860                status = nfs4_check_delegmode(delegstateid(s), flags);
5861                break;
5862        case NFS4_OPEN_STID:
5863        case NFS4_LOCK_STID:
5864                status = nfs4_check_olstateid(openlockstateid(s), flags);
5865                break;
5866        default:
5867                status = nfserr_bad_stateid;
5868                break;
5869        }
5870        if (status)
5871                goto out;
5872        status = nfs4_check_fh(fhp, s);
5873
5874done:
5875        if (status == nfs_ok && nfp)
5876                status = nfs4_check_file(rqstp, fhp, s, nfp, flags);
5877out:
5878        if (s) {
5879                if (!status && cstid)
5880                        *cstid = s;
5881                else
5882                        nfs4_put_stid(s);
5883        }
5884        return status;
5885}
5886
5887/*
5888 * Test if the stateid is valid
5889 */
5890__be32
5891nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5892                   union nfsd4_op_u *u)
5893{
5894        struct nfsd4_test_stateid *test_stateid = &u->test_stateid;
5895        struct nfsd4_test_stateid_id *stateid;
5896        struct nfs4_client *cl = cstate->clp;
5897
5898        list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
5899                stateid->ts_id_status =
5900                        nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
5901
5902        return nfs_ok;
5903}
5904
5905static __be32
5906nfsd4_free_lock_stateid(stateid_t *stateid, struct nfs4_stid *s)
5907{
5908        struct nfs4_ol_stateid *stp = openlockstateid(s);
5909        __be32 ret;
5910
5911        ret = nfsd4_lock_ol_stateid(stp);
5912        if (ret)
5913                goto out_put_stid;
5914
5915        ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
5916        if (ret)
5917                goto out;
5918
5919        ret = nfserr_locks_held;
5920        if (check_for_locks(stp->st_stid.sc_file,
5921                            lockowner(stp->st_stateowner)))
5922                goto out;
5923
5924        release_lock_stateid(stp);
5925        ret = nfs_ok;
5926
5927out:
5928        mutex_unlock(&stp->st_mutex);
5929out_put_stid:
5930        nfs4_put_stid(s);
5931        return ret;
5932}
5933
5934__be32
5935nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5936                   union nfsd4_op_u *u)
5937{
5938        struct nfsd4_free_stateid *free_stateid = &u->free_stateid;
5939        stateid_t *stateid = &free_stateid->fr_stateid;
5940        struct nfs4_stid *s;
5941        struct nfs4_delegation *dp;
5942        struct nfs4_client *cl = cstate->clp;
5943        __be32 ret = nfserr_bad_stateid;
5944
5945        spin_lock(&cl->cl_lock);
5946        s = find_stateid_locked(cl, stateid);
5947        if (!s)
5948                goto out_unlock;
5949        spin_lock(&s->sc_lock);
5950        switch (s->sc_type) {
5951        case NFS4_DELEG_STID:
5952                ret = nfserr_locks_held;
5953                break;
5954        case NFS4_OPEN_STID:
5955                ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
5956                if (ret)
5957                        break;
5958                ret = nfserr_locks_held;
5959                break;
5960        case NFS4_LOCK_STID:
5961                spin_unlock(&s->sc_lock);
5962                refcount_inc(&s->sc_count);
5963                spin_unlock(&cl->cl_lock);
5964                ret = nfsd4_free_lock_stateid(stateid, s);
5965                goto out;
5966        case NFS4_REVOKED_DELEG_STID:
5967                spin_unlock(&s->sc_lock);
5968                dp = delegstateid(s);
5969                list_del_init(&dp->dl_recall_lru);
5970                spin_unlock(&cl->cl_lock);
5971                nfs4_put_stid(s);
5972                ret = nfs_ok;
5973                goto out;
5974        /* Default falls through and returns nfserr_bad_stateid */
5975        }
5976        spin_unlock(&s->sc_lock);
5977out_unlock:
5978        spin_unlock(&cl->cl_lock);
5979out:
5980        return ret;
5981}
5982
5983static inline int
5984setlkflg (int type)
5985{
5986        return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
5987                RD_STATE : WR_STATE;
5988}
5989
5990static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
5991{
5992        struct svc_fh *current_fh = &cstate->current_fh;
5993        struct nfs4_stateowner *sop = stp->st_stateowner;
5994        __be32 status;
5995
5996        status = nfsd4_check_seqid(cstate, sop, seqid);
5997        if (status)
5998                return status;
5999        status = nfsd4_lock_ol_stateid(stp);
6000        if (status != nfs_ok)
6001                return status;
6002        status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
6003        if (status == nfs_ok)
6004                status = nfs4_check_fh(current_fh, &stp->st_stid);
6005        if (status != nfs_ok)
6006                mutex_unlock(&stp->st_mutex);
6007        return status;
6008}
6009
6010/* 
6011 * Checks for sequence id mutating operations. 
6012 */
6013static __be32
6014nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
6015                         stateid_t *stateid, char typemask,
6016                         struct nfs4_ol_stateid **stpp,
6017                         struct nfsd_net *nn)
6018{
6019        __be32 status;
6020        struct nfs4_stid *s;
6021        struct nfs4_ol_stateid *stp = NULL;
6022
6023        trace_nfsd_preprocess(seqid, stateid);
6024
6025        *stpp = NULL;
6026        status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
6027        if (status)
6028                return status;
6029        stp = openlockstateid(s);
6030        nfsd4_cstate_assign_replay(cstate, stp->st_stateowner);
6031
6032        status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
6033        if (!status)
6034                *stpp = stp;
6035        else
6036                nfs4_put_stid(&stp->st_stid);
6037        return status;
6038}
6039
6040static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
6041                                                 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
6042{
6043        __be32 status;
6044        struct nfs4_openowner *oo;
6045        struct nfs4_ol_stateid *stp;
6046
6047        status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
6048                                                NFS4_OPEN_STID, &stp, nn);
6049        if (status)
6050                return status;
6051        oo = openowner(stp->st_stateowner);
6052        if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
6053                mutex_unlock(&stp->st_mutex);
6054                nfs4_put_stid(&stp->st_stid);
6055                return nfserr_bad_stateid;
6056        }
6057        *stpp = stp;
6058        return nfs_ok;
6059}
6060
6061__be32
6062nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6063                   union nfsd4_op_u *u)
6064{
6065        struct nfsd4_open_confirm *oc = &u->open_confirm;
6066        __be32 status;
6067        struct nfs4_openowner *oo;
6068        struct nfs4_ol_stateid *stp;
6069        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6070
6071        dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
6072                        cstate->current_fh.fh_dentry);
6073
6074        status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
6075        if (status)
6076                return status;
6077
6078        status = nfs4_preprocess_seqid_op(cstate,
6079                                        oc->oc_seqid, &oc->oc_req_stateid,
6080                                        NFS4_OPEN_STID, &stp, nn);
6081        if (status)
6082                goto out;
6083        oo = openowner(stp->st_stateowner);
6084        status = nfserr_bad_stateid;
6085        if (oo->oo_flags & NFS4_OO_CONFIRMED) {
6086                mutex_unlock(&stp->st_mutex);
6087                goto put_stateid;
6088        }
6089        oo->oo_flags |= NFS4_OO_CONFIRMED;
6090        nfs4_inc_and_copy_stateid(&oc->oc_resp_stateid, &stp->st_stid);
6091        mutex_unlock(&stp->st_mutex);
6092        trace_nfsd_open_confirm(oc->oc_seqid, &stp->st_stid.sc_stateid);
6093        nfsd4_client_record_create(oo->oo_owner.so_client);
6094        status = nfs_ok;
6095put_stateid:
6096        nfs4_put_stid(&stp->st_stid);
6097out:
6098        nfsd4_bump_seqid(cstate, status);
6099        return status;
6100}
6101
6102static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
6103{
6104        if (!test_access(access, stp))
6105                return;
6106        nfs4_file_put_access(stp->st_stid.sc_file, access);
6107        clear_access(access, stp);
6108}
6109
6110static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
6111{
6112        switch (to_access) {
6113        case NFS4_SHARE_ACCESS_READ:
6114                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
6115                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
6116                break;
6117        case NFS4_SHARE_ACCESS_WRITE:
6118                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
6119                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
6120                break;
6121        case NFS4_SHARE_ACCESS_BOTH:
6122                break;
6123        default:
6124                WARN_ON_ONCE(1);
6125        }
6126}
6127
6128__be32
6129nfsd4_open_downgrade(struct svc_rqst *rqstp,
6130                     struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
6131{
6132        struct nfsd4_open_downgrade *od = &u->open_downgrade;
6133        __be32 status;
6134        struct nfs4_ol_stateid *stp;
6135        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6136
6137        dprintk("NFSD: nfsd4_open_downgrade on file %pd\n", 
6138                        cstate->current_fh.fh_dentry);
6139
6140        /* We don't yet support WANT bits: */
6141        if (od->od_deleg_want)
6142                dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
6143                        od->od_deleg_want);
6144
6145        status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
6146                                        &od->od_stateid, &stp, nn);
6147        if (status)
6148                goto out; 
6149        status = nfserr_inval;
6150        if (!test_access(od->od_share_access, stp)) {
6151                dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
6152                        stp->st_access_bmap, od->od_share_access);
6153                goto put_stateid;
6154        }
6155        if (!test_deny(od->od_share_deny, stp)) {
6156                dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
6157                        stp->st_deny_bmap, od->od_share_deny);
6158                goto put_stateid;
6159        }
6160        nfs4_stateid_downgrade(stp, od->od_share_access);
6161        reset_union_bmap_deny(od->od_share_deny, stp);
6162        nfs4_inc_and_copy_stateid(&od->od_stateid, &stp->st_stid);
6163        status = nfs_ok;
6164put_stateid:
6165        mutex_unlock(&stp->st_mutex);
6166        nfs4_put_stid(&stp->st_stid);
6167out:
6168        nfsd4_bump_seqid(cstate, status);
6169        return status;
6170}
6171
6172static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
6173{
6174        struct nfs4_client *clp = s->st_stid.sc_client;
6175        bool unhashed;
6176        LIST_HEAD(reaplist);
6177
6178        spin_lock(&clp->cl_lock);
6179        unhashed = unhash_open_stateid(s, &reaplist);
6180
6181        if (clp->cl_minorversion) {
6182                if (unhashed)
6183                        put_ol_stateid_locked(s, &reaplist);
6184                spin_unlock(&clp->cl_lock);
6185                free_ol_stateid_reaplist(&reaplist);
6186        } else {
6187                spin_unlock(&clp->cl_lock);
6188                free_ol_stateid_reaplist(&reaplist);
6189                if (unhashed)
6190                        move_to_close_lru(s, clp->net);
6191        }
6192}
6193
6194/*
6195 * nfs4_unlock_state() called after encode
6196 */
6197__be32
6198nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6199                union nfsd4_op_u *u)
6200{
6201        struct nfsd4_close *close = &u->close;
6202        __be32 status;
6203        struct nfs4_ol_stateid *stp;
6204        struct net *net = SVC_NET(rqstp);
6205        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6206
6207        dprintk("NFSD: nfsd4_close on file %pd\n", 
6208                        cstate->current_fh.fh_dentry);
6209
6210        status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
6211                                        &close->cl_stateid,
6212                                        NFS4_OPEN_STID|NFS4_CLOSED_STID,
6213                                        &stp, nn);
6214        nfsd4_bump_seqid(cstate, status);
6215        if (status)
6216                goto out; 
6217
6218        stp->st_stid.sc_type = NFS4_CLOSED_STID;
6219
6220        /*
6221         * Technically we don't _really_ have to increment or copy it, since
6222         * it should just be gone after this operation and we clobber the
6223         * copied value below, but we continue to do so here just to ensure
6224         * that racing ops see that there was a state change.
6225         */
6226        nfs4_inc_and_copy_stateid(&close->cl_stateid, &stp->st_stid);
6227
6228        nfsd4_close_open_stateid(stp);
6229        mutex_unlock(&stp->st_mutex);
6230
6231        /* v4.1+ suggests that we send a special stateid in here, since the
6232         * clients should just ignore this anyway. Since this is not useful
6233         * for v4.0 clients either, we set it to the special close_stateid
6234         * universally.
6235         *
6236         * See RFC5661 section 18.2.4, and RFC7530 section 16.2.5
6237         */
6238        memcpy(&close->cl_stateid, &close_stateid, sizeof(close->cl_stateid));
6239
6240        /* put reference from nfs4_preprocess_seqid_op */
6241        nfs4_put_stid(&stp->st_stid);
6242out:
6243        return status;
6244}
6245
6246__be32
6247nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6248                  union nfsd4_op_u *u)
6249{
6250        struct nfsd4_delegreturn *dr = &u->delegreturn;
6251        struct nfs4_delegation *dp;
6252        stateid_t *stateid = &dr->dr_stateid;
6253        struct nfs4_stid *s;
6254        __be32 status;
6255        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6256
6257        if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
6258                return status;
6259
6260        status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
6261        if (status)
6262                goto out;
6263        dp = delegstateid(s);
6264        status = nfsd4_stid_check_stateid_generation(stateid, &dp->dl_stid, nfsd4_has_session(cstate));
6265        if (status)
6266                goto put_stateid;
6267
6268        destroy_delegation(dp);
6269put_stateid:
6270        nfs4_put_stid(&dp->dl_stid);
6271out:
6272        return status;
6273}
6274
6275static inline u64
6276end_offset(u64 start, u64 len)
6277{
6278        u64 end;
6279
6280        end = start + len;
6281        return end >= start ? end: NFS4_MAX_UINT64;
6282}
6283
6284/* last octet in a range */
6285static inline u64
6286last_byte_offset(u64 start, u64 len)
6287{
6288        u64 end;
6289
6290        WARN_ON_ONCE(!len);
6291        end = start + len;
6292        return end > start ? end - 1: NFS4_MAX_UINT64;
6293}
6294
6295/*
6296 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
6297 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
6298 * byte, because of sign extension problems.  Since NFSv4 calls for 64-bit
6299 * locking, this prevents us from being completely protocol-compliant.  The
6300 * real solution to this problem is to start using unsigned file offsets in
6301 * the VFS, but this is a very deep change!
6302 */
6303static inline void
6304nfs4_transform_lock_offset(struct file_lock *lock)
6305{
6306        if (lock->fl_start < 0)
6307                lock->fl_start = OFFSET_MAX;
6308        if (lock->fl_end < 0)
6309                lock->fl_end = OFFSET_MAX;
6310}
6311
6312static fl_owner_t
6313nfsd4_fl_get_owner(fl_owner_t owner)
6314{
6315        struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
6316
6317        nfs4_get_stateowner(&lo->lo_owner);
6318        return owner;
6319}
6320
6321static void
6322nfsd4_fl_put_owner(fl_owner_t owner)
6323{
6324        struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
6325
6326        if (lo)
6327                nfs4_put_stateowner(&lo->lo_owner);
6328}
6329
6330static void
6331nfsd4_lm_notify(struct file_lock *fl)
6332{
6333        struct nfs4_lockowner           *lo = (struct nfs4_lockowner *)fl->fl_owner;
6334        struct net                      *net = lo->lo_owner.so_client->net;
6335        struct nfsd_net                 *nn = net_generic(net, nfsd_net_id);
6336        struct nfsd4_blocked_lock       *nbl = container_of(fl,
6337                                                struct nfsd4_blocked_lock, nbl_lock);
6338        bool queue = false;
6339
6340        /* An empty list means that something else is going to be using it */
6341        spin_lock(&nn->blocked_locks_lock);
6342        if (!list_empty(&nbl->nbl_list)) {
6343                list_del_init(&nbl->nbl_list);
6344                list_del_init(&nbl->nbl_lru);
6345                queue = true;
6346        }
6347        spin_unlock(&nn->blocked_locks_lock);
6348
6349        if (queue)
6350                nfsd4_run_cb(&nbl->nbl_cb);
6351}
6352
6353static const struct lock_manager_operations nfsd_posix_mng_ops  = {
6354        .lm_notify = nfsd4_lm_notify,
6355        .lm_get_owner = nfsd4_fl_get_owner,
6356        .lm_put_owner = nfsd4_fl_put_owner,
6357};
6358
6359static inline void
6360nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
6361{
6362        struct nfs4_lockowner *lo;
6363
6364        if (fl->fl_lmops == &nfsd_posix_mng_ops) {
6365                lo = (struct nfs4_lockowner *) fl->fl_owner;
6366                xdr_netobj_dup(&deny->ld_owner, &lo->lo_owner.so_owner,
6367                                                GFP_KERNEL);
6368                if (!deny->ld_owner.data)
6369                        /* We just don't care that much */
6370                        goto nevermind;
6371                deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
6372        } else {
6373nevermind:
6374                deny->ld_owner.len = 0;
6375                deny->ld_owner.data = NULL;
6376                deny->ld_clientid.cl_boot = 0;
6377                deny->ld_clientid.cl_id = 0;
6378        }
6379        deny->ld_start = fl->fl_start;
6380        deny->ld_length = NFS4_MAX_UINT64;
6381        if (fl->fl_end != NFS4_MAX_UINT64)
6382                deny->ld_length = fl->fl_end - fl->fl_start + 1;        
6383        deny->ld_type = NFS4_READ_LT;
6384        if (fl->fl_type != F_RDLCK)
6385                deny->ld_type = NFS4_WRITE_LT;
6386}
6387
6388static struct nfs4_lockowner *
6389find_lockowner_str_locked(struct nfs4_client *clp, struct xdr_netobj *owner)
6390{
6391        unsigned int strhashval = ownerstr_hashval(owner);
6392        struct nfs4_stateowner *so;
6393
6394        lockdep_assert_held(&clp->cl_lock);
6395
6396        list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[strhashval],
6397                            so_strhash) {
6398                if (so->so_is_open_owner)
6399                        continue;
6400                if (same_owner_str(so, owner))
6401                        return lockowner(nfs4_get_stateowner(so));
6402        }
6403        return NULL;
6404}
6405
6406static struct nfs4_lockowner *
6407find_lockowner_str(struct nfs4_client *clp, struct xdr_netobj *owner)
6408{
6409        struct nfs4_lockowner *lo;
6410
6411        spin_lock(&clp->cl_lock);
6412        lo = find_lockowner_str_locked(clp, owner);
6413        spin_unlock(&clp->cl_lock);
6414        return lo;
6415}
6416
6417static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
6418{
6419        unhash_lockowner_locked(lockowner(sop));
6420}
6421
6422static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
6423{
6424        struct nfs4_lockowner *lo = lockowner(sop);
6425
6426        kmem_cache_free(lockowner_slab, lo);
6427}
6428
6429static const struct nfs4_stateowner_operations lockowner_ops = {
6430        .so_unhash =    nfs4_unhash_lockowner,
6431        .so_free =      nfs4_free_lockowner,
6432};
6433
6434/*
6435 * Alloc a lock owner structure.
6436 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has 
6437 * occurred. 
6438 *
6439 * strhashval = ownerstr_hashval
6440 */
6441static struct nfs4_lockowner *
6442alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp,
6443                           struct nfs4_ol_stateid *open_stp,
6444                           struct nfsd4_lock *lock)
6445{
6446        struct nfs4_lockowner *lo, *ret;
6447
6448        lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
6449        if (!lo)
6450                return NULL;
6451        INIT_LIST_HEAD(&lo->lo_blocked);
6452        INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
6453        lo->lo_owner.so_is_open_owner = 0;
6454        lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
6455        lo->lo_owner.so_ops = &lockowner_ops;
6456        spin_lock(&clp->cl_lock);
6457        ret = find_lockowner_str_locked(clp, &lock->lk_new_owner);
6458        if (ret == NULL) {
6459                list_add(&lo->lo_owner.so_strhash,
6460                         &clp->cl_ownerstr_hashtbl[strhashval]);
6461                ret = lo;
6462        } else
6463                nfs4_free_stateowner(&lo->lo_owner);
6464
6465        spin_unlock(&clp->cl_lock);
6466        return ret;
6467}
6468
6469static struct nfs4_ol_stateid *
6470find_lock_stateid(const struct nfs4_lockowner *lo,
6471                  const struct nfs4_ol_stateid *ost)
6472{
6473        struct nfs4_ol_stateid *lst;
6474
6475        lockdep_assert_held(&ost->st_stid.sc_client->cl_lock);
6476
6477        /* If ost is not hashed, ost->st_locks will not be valid */
6478        if (!nfs4_ol_stateid_unhashed(ost))
6479                list_for_each_entry(lst, &ost->st_locks, st_locks) {
6480                        if (lst->st_stateowner == &lo->lo_owner) {
6481                                refcount_inc(&lst->st_stid.sc_count);
6482                                return lst;
6483                        }
6484                }
6485        return NULL;
6486}
6487
6488static struct nfs4_ol_stateid *
6489init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
6490                  struct nfs4_file *fp, struct inode *inode,
6491                  struct nfs4_ol_stateid *open_stp)
6492{
6493        struct nfs4_client *clp = lo->lo_owner.so_client;
6494        struct nfs4_ol_stateid *retstp;
6495
6496        mutex_init(&stp->st_mutex);
6497        mutex_lock_nested(&stp->st_mutex, OPEN_STATEID_MUTEX);
6498retry:
6499        spin_lock(&clp->cl_lock);
6500        if (nfs4_ol_stateid_unhashed(open_stp))
6501                goto out_close;
6502        retstp = find_lock_stateid(lo, open_stp);
6503        if (retstp)
6504                goto out_found;
6505        refcount_inc(&stp->st_stid.sc_count);
6506        stp->st_stid.sc_type = NFS4_LOCK_STID;
6507        stp->st_stateowner = nfs4_get_stateowner(&lo->lo_owner);
6508        get_nfs4_file(fp);
6509        stp->st_stid.sc_file = fp;
6510        stp->st_access_bmap = 0;
6511        stp->st_deny_bmap = open_stp->st_deny_bmap;
6512        stp->st_openstp = open_stp;
6513        spin_lock(&fp->fi_lock);
6514        list_add(&stp->st_locks, &open_stp->st_locks);
6515        list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
6516        list_add(&stp->st_perfile, &fp->fi_stateids);
6517        spin_unlock(&fp->fi_lock);
6518        spin_unlock(&clp->cl_lock);
6519        return stp;
6520out_found:
6521        spin_unlock(&clp->cl_lock);
6522        if (nfsd4_lock_ol_stateid(retstp) != nfs_ok) {
6523                nfs4_put_stid(&retstp->st_stid);
6524                goto retry;
6525        }
6526        /* To keep mutex tracking happy */
6527        mutex_unlock(&stp->st_mutex);
6528        return retstp;
6529out_close:
6530        spin_unlock(&clp->cl_lock);
6531        mutex_unlock(&stp->st_mutex);
6532        return NULL;
6533}
6534
6535static struct nfs4_ol_stateid *
6536find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
6537                            struct inode *inode, struct nfs4_ol_stateid *ost,
6538                            bool *new)
6539{
6540        struct nfs4_stid *ns = NULL;
6541        struct nfs4_ol_stateid *lst;
6542        struct nfs4_openowner *oo = openowner(ost->st_stateowner);
6543        struct nfs4_client *clp = oo->oo_owner.so_client;
6544
6545        *new = false;
6546        spin_lock(&clp->cl_lock);
6547        lst = find_lock_stateid(lo, ost);
6548        spin_unlock(&clp->cl_lock);
6549        if (lst != NULL) {
6550                if (nfsd4_lock_ol_stateid(lst) == nfs_ok)
6551                        goto out;
6552                nfs4_put_stid(&lst->st_stid);
6553        }
6554        ns = nfs4_alloc_stid(clp, stateid_slab, nfs4_free_lock_stateid);
6555        if (ns == NULL)
6556                return NULL;
6557
6558        lst = init_lock_stateid(openlockstateid(ns), lo, fi, inode, ost);
6559        if (lst == openlockstateid(ns))
6560                *new = true;
6561        else
6562                nfs4_put_stid(ns);
6563out:
6564        return lst;
6565}
6566
6567static int
6568check_lock_length(u64 offset, u64 length)
6569{
6570        return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
6571                (length > ~offset)));
6572}
6573
6574static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
6575{
6576        struct nfs4_file *fp = lock_stp->st_stid.sc_file;
6577
6578        lockdep_assert_held(&fp->fi_lock);
6579
6580        if (test_access(access, lock_stp))
6581                return;
6582        __nfs4_file_get_access(fp, access);
6583        set_access(access, lock_stp);
6584}
6585
6586static __be32
6587lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
6588                            struct nfs4_ol_stateid *ost,
6589                            struct nfsd4_lock *lock,
6590                            struct nfs4_ol_stateid **plst, bool *new)
6591{
6592        __be32 status;
6593        struct nfs4_file *fi = ost->st_stid.sc_file;
6594        struct nfs4_openowner *oo = openowner(ost->st_stateowner);
6595        struct nfs4_client *cl = oo->oo_owner.so_client;
6596        struct inode *inode = d_inode(cstate->current_fh.fh_dentry);
6597        struct nfs4_lockowner *lo;
6598        struct nfs4_ol_stateid *lst;
6599        unsigned int strhashval;
6600
6601        lo = find_lockowner_str(cl, &lock->lk_new_owner);
6602        if (!lo) {
6603                strhashval = ownerstr_hashval(&lock->lk_new_owner);
6604                lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
6605                if (lo == NULL)
6606                        return nfserr_jukebox;
6607        } else {
6608                /* with an existing lockowner, seqids must be the same */
6609                status = nfserr_bad_seqid;
6610                if (!cstate->minorversion &&
6611                    lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
6612                        goto out;
6613        }
6614
6615        lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
6616        if (lst == NULL) {
6617                status = nfserr_jukebox;
6618                goto out;
6619        }
6620
6621        status = nfs_ok;
6622        *plst = lst;
6623out:
6624        nfs4_put_stateowner(&lo->lo_owner);
6625        return status;
6626}
6627
6628/*
6629 *  LOCK operation 
6630 */
6631__be32
6632nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6633           union nfsd4_op_u *u)
6634{
6635        struct nfsd4_lock *lock = &u->lock;
6636        struct nfs4_openowner *open_sop = NULL;
6637        struct nfs4_lockowner *lock_sop = NULL;
6638        struct nfs4_ol_stateid *lock_stp = NULL;
6639        struct nfs4_ol_stateid *open_stp = NULL;
6640        struct nfs4_file *fp;
6641        struct nfsd_file *nf = NULL;
6642        struct nfsd4_blocked_lock *nbl = NULL;
6643        struct file_lock *file_lock = NULL;
6644        struct file_lock *conflock = NULL;
6645        __be32 status = 0;
6646        int lkflg;
6647        int err;
6648        bool new = false;
6649        unsigned char fl_type;
6650        unsigned int fl_flags = FL_POSIX;
6651        struct net *net = SVC_NET(rqstp);
6652        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6653
6654        dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
6655                (long long) lock->lk_offset,
6656                (long long) lock->lk_length);
6657
6658        if (check_lock_length(lock->lk_offset, lock->lk_length))
6659                 return nfserr_inval;
6660
6661        if ((status = fh_verify(rqstp, &cstate->current_fh,
6662                                S_IFREG, NFSD_MAY_LOCK))) {
6663                dprintk("NFSD: nfsd4_lock: permission denied!\n");
6664                return status;
6665        }
6666
6667        if (lock->lk_is_new) {
6668                if (nfsd4_has_session(cstate))
6669                        /* See rfc 5661 18.10.3: given clientid is ignored: */
6670                        memcpy(&lock->lk_new_clientid,
6671                                &cstate->clp->cl_clientid,
6672                                sizeof(clientid_t));
6673
6674                /* validate and update open stateid and open seqid */
6675                status = nfs4_preprocess_confirmed_seqid_op(cstate,
6676                                        lock->lk_new_open_seqid,
6677                                        &lock->lk_new_open_stateid,
6678                                        &open_stp, nn);
6679                if (status)
6680                        goto out;
6681                mutex_unlock(&open_stp->st_mutex);
6682                open_sop = openowner(open_stp->st_stateowner);
6683                status = nfserr_bad_stateid;
6684                if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
6685                                                &lock->lk_new_clientid))
6686                        goto out;
6687                status = lookup_or_create_lock_state(cstate, open_stp, lock,
6688                                                        &lock_stp, &new);
6689        } else {
6690                status = nfs4_preprocess_seqid_op(cstate,
6691                                       lock->lk_old_lock_seqid,
6692                                       &lock->lk_old_lock_stateid,
6693                                       NFS4_LOCK_STID, &lock_stp, nn);
6694        }
6695        if (status)
6696                goto out;
6697        lock_sop = lockowner(lock_stp->st_stateowner);
6698
6699        lkflg = setlkflg(lock->lk_type);
6700        status = nfs4_check_openmode(lock_stp, lkflg);
6701        if (status)
6702                goto out;
6703
6704        status = nfserr_grace;
6705        if (locks_in_grace(net) && !lock->lk_reclaim)
6706                goto out;
6707        status = nfserr_no_grace;
6708        if (!locks_in_grace(net) && lock->lk_reclaim)
6709                goto out;
6710
6711        fp = lock_stp->st_stid.sc_file;
6712        switch (lock->lk_type) {
6713                case NFS4_READW_LT:
6714                        if (nfsd4_has_session(cstate))
6715                                fl_flags |= FL_SLEEP;
6716                        fallthrough;
6717                case NFS4_READ_LT:
6718                        spin_lock(&fp->fi_lock);
6719                        nf = find_readable_file_locked(fp);
6720                        if (nf)
6721                                get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
6722                        spin_unlock(&fp->fi_lock);
6723                        fl_type = F_RDLCK;
6724                        break;
6725                case NFS4_WRITEW_LT:
6726                        if (nfsd4_has_session(cstate))
6727                                fl_flags |= FL_SLEEP;
6728                        fallthrough;
6729                case NFS4_WRITE_LT:
6730                        spin_lock(&fp->fi_lock);
6731                        nf = find_writeable_file_locked(fp);
6732                        if (nf)
6733                                get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
6734                        spin_unlock(&fp->fi_lock);
6735                        fl_type = F_WRLCK;
6736                        break;
6737                default:
6738                        status = nfserr_inval;
6739                goto out;
6740        }
6741
6742        if (!nf) {
6743                status = nfserr_openmode;
6744                goto out;
6745        }
6746
6747        nbl = find_or_allocate_block(lock_sop, &fp->fi_fhandle, nn);
6748        if (!nbl) {
6749                dprintk("NFSD: %s: unable to allocate block!\n", __func__);
6750                status = nfserr_jukebox;
6751                goto out;
6752        }
6753
6754        file_lock = &nbl->nbl_lock;
6755        file_lock->fl_type = fl_type;
6756        file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(&lock_sop->lo_owner));
6757        file_lock->fl_pid = current->tgid;
6758        file_lock->fl_file = nf->nf_file;
6759        file_lock->fl_flags = fl_flags;
6760        file_lock->fl_lmops = &nfsd_posix_mng_ops;
6761        file_lock->fl_start = lock->lk_offset;
6762        file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
6763        nfs4_transform_lock_offset(file_lock);
6764
6765        conflock = locks_alloc_lock();
6766        if (!conflock) {
6767                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
6768                status = nfserr_jukebox;
6769                goto out;
6770        }
6771
6772        if (fl_flags & FL_SLEEP) {
6773                nbl->nbl_time = ktime_get_boottime_seconds();
6774                spin_lock(&nn->blocked_locks_lock);
6775                list_add_tail(&nbl->nbl_list, &lock_sop->lo_blocked);
6776                list_add_tail(&nbl->nbl_lru, &nn->blocked_locks_lru);
6777                spin_unlock(&nn->blocked_locks_lock);
6778        }
6779
6780        err = vfs_lock_file(nf->nf_file, F_SETLK, file_lock, conflock);
6781        switch (err) {
6782        case 0: /* success! */
6783                nfs4_inc_and_copy_stateid(&lock->lk_resp_stateid, &lock_stp->st_stid);
6784                status = 0;
6785                if (lock->lk_reclaim)
6786                        nn->somebody_reclaimed = true;
6787                break;
6788        case FILE_LOCK_DEFERRED:
6789                nbl = NULL;
6790                fallthrough;
6791        case -EAGAIN:           /* conflock holds conflicting lock */
6792                status = nfserr_denied;
6793                dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
6794                nfs4_set_lock_denied(conflock, &lock->lk_denied);
6795                break;
6796        case -EDEADLK:
6797                status = nfserr_deadlock;
6798                break;
6799        default:
6800                dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
6801                status = nfserrno(err);
6802                break;
6803        }
6804out:
6805        if (nbl) {
6806                /* dequeue it if we queued it before */
6807                if (fl_flags & FL_SLEEP) {
6808                        spin_lock(&nn->blocked_locks_lock);
6809                        list_del_init(&nbl->nbl_list);
6810                        list_del_init(&nbl->nbl_lru);
6811                        spin_unlock(&nn->blocked_locks_lock);
6812                }
6813                free_blocked_lock(nbl);
6814        }
6815        if (nf)
6816                nfsd_file_put(nf);
6817        if (lock_stp) {
6818                /* Bump seqid manually if the 4.0 replay owner is openowner */
6819                if (cstate->replay_owner &&
6820                    cstate->replay_owner != &lock_sop->lo_owner &&
6821                    seqid_mutating_err(ntohl(status)))
6822                        lock_sop->lo_owner.so_seqid++;
6823
6824                /*
6825                 * If this is a new, never-before-used stateid, and we are
6826                 * returning an error, then just go ahead and release it.
6827                 */
6828                if (status && new)
6829                        release_lock_stateid(lock_stp);
6830
6831                mutex_unlock(&lock_stp->st_mutex);
6832
6833                nfs4_put_stid(&lock_stp->st_stid);
6834        }
6835        if (open_stp)
6836                nfs4_put_stid(&open_stp->st_stid);
6837        nfsd4_bump_seqid(cstate, status);
6838        if (conflock)
6839                locks_free_lock(conflock);
6840        return status;
6841}
6842
6843/*
6844 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
6845 * so we do a temporary open here just to get an open file to pass to
6846 * vfs_test_lock.  (Arguably perhaps test_lock should be done with an
6847 * inode operation.)
6848 */
6849static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
6850{
6851        struct nfsd_file *nf;
6852        __be32 err = nfsd_file_acquire(rqstp, fhp, NFSD_MAY_READ, &nf);
6853        if (!err) {
6854                err = nfserrno(vfs_test_lock(nf->nf_file, lock));
6855                nfsd_file_put(nf);
6856        }
6857        return err;
6858}
6859
6860/*
6861 * LOCKT operation
6862 */
6863__be32
6864nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6865            union nfsd4_op_u *u)
6866{
6867        struct nfsd4_lockt *lockt = &u->lockt;
6868        struct file_lock *file_lock = NULL;
6869        struct nfs4_lockowner *lo = NULL;
6870        __be32 status;
6871        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6872
6873        if (locks_in_grace(SVC_NET(rqstp)))
6874                return nfserr_grace;
6875
6876        if (check_lock_length(lockt->lt_offset, lockt->lt_length))
6877                 return nfserr_inval;
6878
6879        if (!nfsd4_has_session(cstate)) {
6880                status = set_client(&lockt->lt_clientid, cstate, nn);
6881                if (status)
6882                        goto out;
6883        }
6884
6885        if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
6886                goto out;
6887
6888        file_lock = locks_alloc_lock();
6889        if (!file_lock) {
6890                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
6891                status = nfserr_jukebox;
6892                goto out;
6893        }
6894
6895        switch (lockt->lt_type) {
6896                case NFS4_READ_LT:
6897                case NFS4_READW_LT:
6898                        file_lock->fl_type = F_RDLCK;
6899                        break;
6900                case NFS4_WRITE_LT:
6901                case NFS4_WRITEW_LT:
6902                        file_lock->fl_type = F_WRLCK;
6903                        break;
6904                default:
6905                        dprintk("NFSD: nfs4_lockt: bad lock type!\n");
6906                        status = nfserr_inval;
6907                        goto out;
6908        }
6909
6910        lo = find_lockowner_str(cstate->clp, &lockt->lt_owner);
6911        if (lo)
6912                file_lock->fl_owner = (fl_owner_t)lo;
6913        file_lock->fl_pid = current->tgid;
6914        file_lock->fl_flags = FL_POSIX;
6915
6916        file_lock->fl_start = lockt->lt_offset;
6917        file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
6918
6919        nfs4_transform_lock_offset(file_lock);
6920
6921        status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
6922        if (status)
6923                goto out;
6924
6925        if (file_lock->fl_type != F_UNLCK) {
6926                status = nfserr_denied;
6927                nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
6928        }
6929out:
6930        if (lo)
6931                nfs4_put_stateowner(&lo->lo_owner);
6932        if (file_lock)
6933                locks_free_lock(file_lock);
6934        return status;
6935}
6936
6937__be32
6938nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6939            union nfsd4_op_u *u)
6940{
6941        struct nfsd4_locku *locku = &u->locku;
6942        struct nfs4_ol_stateid *stp;
6943        struct nfsd_file *nf = NULL;
6944        struct file_lock *file_lock = NULL;
6945        __be32 status;
6946        int err;
6947        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6948
6949        dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
6950                (long long) locku->lu_offset,
6951                (long long) locku->lu_length);
6952
6953        if (check_lock_length(locku->lu_offset, locku->lu_length))
6954                 return nfserr_inval;
6955
6956        status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
6957                                        &locku->lu_stateid, NFS4_LOCK_STID,
6958                                        &stp, nn);
6959        if (status)
6960                goto out;
6961        nf = find_any_file(stp->st_stid.sc_file);
6962        if (!nf) {
6963                status = nfserr_lock_range;
6964                goto put_stateid;
6965        }
6966        file_lock = locks_alloc_lock();
6967        if (!file_lock) {
6968                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
6969                status = nfserr_jukebox;
6970                goto put_file;
6971        }
6972
6973        file_lock->fl_type = F_UNLCK;
6974        file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(stp->st_stateowner));
6975        file_lock->fl_pid = current->tgid;
6976        file_lock->fl_file = nf->nf_file;
6977        file_lock->fl_flags = FL_POSIX;
6978        file_lock->fl_lmops = &nfsd_posix_mng_ops;
6979        file_lock->fl_start = locku->lu_offset;
6980
6981        file_lock->fl_end = last_byte_offset(locku->lu_offset,
6982                                                locku->lu_length);
6983        nfs4_transform_lock_offset(file_lock);
6984
6985        err = vfs_lock_file(nf->nf_file, F_SETLK, file_lock, NULL);
6986        if (err) {
6987                dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
6988                goto out_nfserr;
6989        }
6990        nfs4_inc_and_copy_stateid(&locku->lu_stateid, &stp->st_stid);
6991put_file:
6992        nfsd_file_put(nf);
6993put_stateid:
6994        mutex_unlock(&stp->st_mutex);
6995        nfs4_put_stid(&stp->st_stid);
6996out:
6997        nfsd4_bump_seqid(cstate, status);
6998        if (file_lock)
6999                locks_free_lock(file_lock);
7000        return status;
7001
7002out_nfserr:
7003        status = nfserrno(err);
7004        goto put_file;
7005}
7006
7007/*
7008 * returns
7009 *      true:  locks held by lockowner
7010 *      false: no locks held by lockowner
7011 */
7012static bool
7013check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
7014{
7015        struct file_lock *fl;
7016        int status = false;
7017        struct nfsd_file *nf = find_any_file(fp);
7018        struct inode *inode;
7019        struct file_lock_context *flctx;
7020
7021        if (!nf) {
7022                /* Any valid lock stateid should have some sort of access */
7023                WARN_ON_ONCE(1);
7024                return status;
7025        }
7026
7027        inode = locks_inode(nf->nf_file);
7028        flctx = inode->i_flctx;
7029
7030        if (flctx && !list_empty_careful(&flctx->flc_posix)) {
7031                spin_lock(&flctx->flc_lock);
7032                list_for_each_entry(fl, &flctx->flc_posix, fl_list) {
7033                        if (fl->fl_owner == (fl_owner_t)lowner) {
7034                                status = true;
7035                                break;
7036                        }
7037                }
7038                spin_unlock(&flctx->flc_lock);
7039        }
7040        nfsd_file_put(nf);
7041        return status;
7042}
7043
7044__be32
7045nfsd4_release_lockowner(struct svc_rqst *rqstp,
7046                        struct nfsd4_compound_state *cstate,
7047                        union nfsd4_op_u *u)
7048{
7049        struct nfsd4_release_lockowner *rlockowner = &u->release_lockowner;
7050        clientid_t *clid = &rlockowner->rl_clientid;
7051        struct nfs4_stateowner *sop;
7052        struct nfs4_lockowner *lo = NULL;
7053        struct nfs4_ol_stateid *stp;
7054        struct xdr_netobj *owner = &rlockowner->rl_owner;
7055        unsigned int hashval = ownerstr_hashval(owner);
7056        __be32 status;
7057        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
7058        struct nfs4_client *clp;
7059        LIST_HEAD (reaplist);
7060
7061        dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
7062                clid->cl_boot, clid->cl_id);
7063
7064        status = set_client(clid, cstate, nn);
7065        if (status)
7066                return status;
7067
7068        clp = cstate->clp;
7069        /* Find the matching lock stateowner */
7070        spin_lock(&clp->cl_lock);
7071        list_for_each_entry(sop, &clp->cl_ownerstr_hashtbl[hashval],
7072                            so_strhash) {
7073
7074                if (sop->so_is_open_owner || !same_owner_str(sop, owner))
7075                        continue;
7076
7077                /* see if there are still any locks associated with it */
7078                lo = lockowner(sop);
7079                list_for_each_entry(stp, &sop->so_stateids, st_perstateowner) {
7080                        if (check_for_locks(stp->st_stid.sc_file, lo)) {
7081                                status = nfserr_locks_held;
7082                                spin_unlock(&clp->cl_lock);
7083                                return status;
7084                        }
7085                }
7086
7087                nfs4_get_stateowner(sop);
7088                break;
7089        }
7090        if (!lo) {
7091                spin_unlock(&clp->cl_lock);
7092                return status;
7093        }
7094
7095        unhash_lockowner_locked(lo);
7096        while (!list_empty(&lo->lo_owner.so_stateids)) {
7097                stp = list_first_entry(&lo->lo_owner.so_stateids,
7098                                       struct nfs4_ol_stateid,
7099                                       st_perstateowner);
7100                WARN_ON(!unhash_lock_stateid(stp));
7101                put_ol_stateid_locked(stp, &reaplist);
7102        }
7103        spin_unlock(&clp->cl_lock);
7104        free_ol_stateid_reaplist(&reaplist);
7105        remove_blocked_locks(lo);
7106        nfs4_put_stateowner(&lo->lo_owner);
7107
7108        return status;
7109}
7110
7111static inline struct nfs4_client_reclaim *
7112alloc_reclaim(void)
7113{
7114        return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
7115}
7116
7117bool
7118nfs4_has_reclaimed_state(struct xdr_netobj name, struct nfsd_net *nn)
7119{
7120        struct nfs4_client_reclaim *crp;
7121
7122        crp = nfsd4_find_reclaim_client(name, nn);
7123        return (crp && crp->cr_clp);
7124}
7125
7126/*
7127 * failure => all reset bets are off, nfserr_no_grace...
7128 *
7129 * The caller is responsible for freeing name.data if NULL is returned (it
7130 * will be freed in nfs4_remove_reclaim_record in the normal case).
7131 */
7132struct nfs4_client_reclaim *
7133nfs4_client_to_reclaim(struct xdr_netobj name, struct xdr_netobj princhash,
7134                struct nfsd_net *nn)
7135{
7136        unsigned int strhashval;
7137        struct nfs4_client_reclaim *crp;
7138
7139        trace_nfsd_clid_reclaim(nn, name.len, name.data);
7140        crp = alloc_reclaim();
7141        if (crp) {
7142                strhashval = clientstr_hashval(name);
7143                INIT_LIST_HEAD(&crp->cr_strhash);
7144                list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
7145                crp->cr_name.data = name.data;
7146                crp->cr_name.len = name.len;
7147                crp->cr_princhash.data = princhash.data;
7148                crp->cr_princhash.len = princhash.len;
7149                crp->cr_clp = NULL;
7150                nn->reclaim_str_hashtbl_size++;
7151        }
7152        return crp;
7153}
7154
7155void
7156nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
7157{
7158        list_del(&crp->cr_strhash);
7159        kfree(crp->cr_name.data);
7160        kfree(crp->cr_princhash.data);
7161        kfree(crp);
7162        nn->reclaim_str_hashtbl_size--;
7163}
7164
7165void
7166nfs4_release_reclaim(struct nfsd_net *nn)
7167{
7168        struct nfs4_client_reclaim *crp = NULL;
7169        int i;
7170
7171        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7172                while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
7173                        crp = list_entry(nn->reclaim_str_hashtbl[i].next,
7174                                        struct nfs4_client_reclaim, cr_strhash);
7175                        nfs4_remove_reclaim_record(crp, nn);
7176                }
7177        }
7178        WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
7179}
7180
7181/*
7182 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
7183struct nfs4_client_reclaim *
7184nfsd4_find_reclaim_client(struct xdr_netobj name, struct nfsd_net *nn)
7185{
7186        unsigned int strhashval;
7187        struct nfs4_client_reclaim *crp = NULL;
7188
7189        trace_nfsd_clid_find(nn, name.len, name.data);
7190
7191        strhashval = clientstr_hashval(name);
7192        list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
7193                if (compare_blob(&crp->cr_name, &name) == 0) {
7194                        return crp;
7195                }
7196        }
7197        return NULL;
7198}
7199
7200__be32
7201nfs4_check_open_reclaim(struct nfs4_client *clp)
7202{
7203        if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &clp->cl_flags))
7204                return nfserr_no_grace;
7205
7206        if (nfsd4_client_record_check(clp))
7207                return nfserr_reclaim_bad;
7208
7209        return nfs_ok;
7210}
7211
7212/*
7213 * Since the lifetime of a delegation isn't limited to that of an open, a
7214 * client may quite reasonably hang on to a delegation as long as it has
7215 * the inode cached.  This becomes an obvious problem the first time a
7216 * client's inode cache approaches the size of the server's total memory.
7217 *
7218 * For now we avoid this problem by imposing a hard limit on the number
7219 * of delegations, which varies according to the server's memory size.
7220 */
7221static void
7222set_max_delegations(void)
7223{
7224        /*
7225         * Allow at most 4 delegations per megabyte of RAM.  Quick
7226         * estimates suggest that in the worst case (where every delegation
7227         * is for a different inode), a delegation could take about 1.5K,
7228         * giving a worst case usage of about 6% of memory.
7229         */
7230        max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
7231}
7232
7233static int nfs4_state_create_net(struct net *net)
7234{
7235        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7236        int i;
7237
7238        nn->conf_id_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
7239                                            sizeof(struct list_head),
7240                                            GFP_KERNEL);
7241        if (!nn->conf_id_hashtbl)
7242                goto err;
7243        nn->unconf_id_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
7244                                              sizeof(struct list_head),
7245                                              GFP_KERNEL);
7246        if (!nn->unconf_id_hashtbl)
7247                goto err_unconf_id;
7248        nn->sessionid_hashtbl = kmalloc_array(SESSION_HASH_SIZE,
7249                                              sizeof(struct list_head),
7250                                              GFP_KERNEL);
7251        if (!nn->sessionid_hashtbl)
7252                goto err_sessionid;
7253
7254        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7255                INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
7256                INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
7257        }
7258        for (i = 0; i < SESSION_HASH_SIZE; i++)
7259                INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
7260        nn->conf_name_tree = RB_ROOT;
7261        nn->unconf_name_tree = RB_ROOT;
7262        nn->boot_time = ktime_get_real_seconds();
7263        nn->grace_ended = false;
7264        nn->nfsd4_manager.block_opens = true;
7265        INIT_LIST_HEAD(&nn->nfsd4_manager.list);
7266        INIT_LIST_HEAD(&nn->client_lru);
7267        INIT_LIST_HEAD(&nn->close_lru);
7268        INIT_LIST_HEAD(&nn->del_recall_lru);
7269        spin_lock_init(&nn->client_lock);
7270        spin_lock_init(&nn->s2s_cp_lock);
7271        idr_init(&nn->s2s_cp_stateids);
7272
7273        spin_lock_init(&nn->blocked_locks_lock);
7274        INIT_LIST_HEAD(&nn->blocked_locks_lru);
7275
7276        INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
7277        get_net(net);
7278
7279        return 0;
7280
7281err_sessionid:
7282        kfree(nn->unconf_id_hashtbl);
7283err_unconf_id:
7284        kfree(nn->conf_id_hashtbl);
7285err:
7286        return -ENOMEM;
7287}
7288
7289static void
7290nfs4_state_destroy_net(struct net *net)
7291{
7292        int i;
7293        struct nfs4_client *clp = NULL;
7294        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7295
7296        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7297                while (!list_empty(&nn->conf_id_hashtbl[i])) {
7298                        clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
7299                        destroy_client(clp);
7300                }
7301        }
7302
7303        WARN_ON(!list_empty(&nn->blocked_locks_lru));
7304
7305        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7306                while (!list_empty(&nn->unconf_id_hashtbl[i])) {
7307                        clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
7308                        destroy_client(clp);
7309                }
7310        }
7311
7312        kfree(nn->sessionid_hashtbl);
7313        kfree(nn->unconf_id_hashtbl);
7314        kfree(nn->conf_id_hashtbl);
7315        put_net(net);
7316}
7317
7318int
7319nfs4_state_start_net(struct net *net)
7320{
7321        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7322        int ret;
7323
7324        ret = get_nfsdfs(net);
7325        if (ret)
7326                return ret;
7327        ret = nfs4_state_create_net(net);
7328        if (ret) {
7329                mntput(nn->nfsd_mnt);
7330                return ret;
7331        }
7332        locks_start_grace(net, &nn->nfsd4_manager);
7333        nfsd4_client_tracking_init(net);
7334        if (nn->track_reclaim_completes && nn->reclaim_str_hashtbl_size == 0)
7335                goto skip_grace;
7336        printk(KERN_INFO "NFSD: starting %lld-second grace period (net %x)\n",
7337               nn->nfsd4_grace, net->ns.inum);
7338        trace_nfsd_grace_start(nn);
7339        queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
7340        return 0;
7341
7342skip_grace:
7343        printk(KERN_INFO "NFSD: no clients to reclaim, skipping NFSv4 grace period (net %x)\n",
7344                        net->ns.inum);
7345        queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_lease * HZ);
7346        nfsd4_end_grace(nn);
7347        return 0;
7348}
7349
7350/* initialization to perform when the nfsd service is started: */
7351
7352int
7353nfs4_state_start(void)
7354{
7355        int ret;
7356
7357        laundry_wq = alloc_workqueue("%s", WQ_UNBOUND, 0, "nfsd4");
7358        if (laundry_wq == NULL) {
7359                ret = -ENOMEM;
7360                goto out;
7361        }
7362        ret = nfsd4_create_callback_queue();
7363        if (ret)
7364                goto out_free_laundry;
7365
7366        set_max_delegations();
7367        return 0;
7368
7369out_free_laundry:
7370        destroy_workqueue(laundry_wq);
7371out:
7372        return ret;
7373}
7374
7375void
7376nfs4_state_shutdown_net(struct net *net)
7377{
7378        struct nfs4_delegation *dp = NULL;
7379        struct list_head *pos, *next, reaplist;
7380        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7381
7382        cancel_delayed_work_sync(&nn->laundromat_work);
7383        locks_end_grace(&nn->nfsd4_manager);
7384
7385        INIT_LIST_HEAD(&reaplist);
7386        spin_lock(&state_lock);
7387        list_for_each_safe(pos, next, &nn->del_recall_lru) {
7388                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
7389                WARN_ON(!unhash_delegation_locked(dp));
7390                list_add(&dp->dl_recall_lru, &reaplist);
7391        }
7392        spin_unlock(&state_lock);
7393        list_for_each_safe(pos, next, &reaplist) {
7394                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
7395                list_del_init(&dp->dl_recall_lru);
7396                destroy_unhashed_deleg(dp);
7397        }
7398
7399        nfsd4_client_tracking_exit(net);
7400        nfs4_state_destroy_net(net);
7401        mntput(nn->nfsd_mnt);
7402}
7403
7404void
7405nfs4_state_shutdown(void)
7406{
7407        destroy_workqueue(laundry_wq);
7408        nfsd4_destroy_callback_queue();
7409}
7410
7411static void
7412get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
7413{
7414        if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG) &&
7415            CURRENT_STATEID(stateid))
7416                memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
7417}
7418
7419static void
7420put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
7421{
7422        if (cstate->minorversion) {
7423                memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
7424                SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
7425        }
7426}
7427
7428void
7429clear_current_stateid(struct nfsd4_compound_state *cstate)
7430{
7431        CLEAR_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
7432}
7433
7434/*
7435 * functions to set current state id
7436 */
7437void
7438nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate,
7439                union nfsd4_op_u *u)
7440{
7441        put_stateid(cstate, &u->open_downgrade.od_stateid);
7442}
7443
7444void
7445nfsd4_set_openstateid(struct nfsd4_compound_state *cstate,
7446                union nfsd4_op_u *u)
7447{
7448        put_stateid(cstate, &u->open.op_stateid);
7449}
7450
7451void
7452nfsd4_set_closestateid(struct nfsd4_compound_state *cstate,
7453                union nfsd4_op_u *u)
7454{
7455        put_stateid(cstate, &u->close.cl_stateid);
7456}
7457
7458void
7459nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate,
7460                union nfsd4_op_u *u)
7461{
7462        put_stateid(cstate, &u->lock.lk_resp_stateid);
7463}
7464
7465/*
7466 * functions to consume current state id
7467 */
7468
7469void
7470nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate,
7471                union nfsd4_op_u *u)
7472{
7473        get_stateid(cstate, &u->open_downgrade.od_stateid);
7474}
7475
7476void
7477nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate,
7478                union nfsd4_op_u *u)
7479{
7480        get_stateid(cstate, &u->delegreturn.dr_stateid);
7481}
7482
7483void
7484nfsd4_get_freestateid(struct nfsd4_compound_state *cstate,
7485                union nfsd4_op_u *u)
7486{
7487        get_stateid(cstate, &u->free_stateid.fr_stateid);
7488}
7489
7490void
7491nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate,
7492                union nfsd4_op_u *u)
7493{
7494        get_stateid(cstate, &u->setattr.sa_stateid);
7495}
7496
7497void
7498nfsd4_get_closestateid(struct nfsd4_compound_state *cstate,
7499                union nfsd4_op_u *u)
7500{
7501        get_stateid(cstate, &u->close.cl_stateid);
7502}
7503
7504void
7505nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate,
7506                union nfsd4_op_u *u)
7507{
7508        get_stateid(cstate, &u->locku.lu_stateid);
7509}
7510
7511void
7512nfsd4_get_readstateid(struct nfsd4_compound_state *cstate,
7513                union nfsd4_op_u *u)
7514{
7515        get_stateid(cstate, &u->read.rd_stateid);
7516}
7517
7518void
7519nfsd4_get_writestateid(struct nfsd4_compound_state *cstate,
7520                union nfsd4_op_u *u)
7521{
7522        get_stateid(cstate, &u->write.wr_stateid);
7523}
7524