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 "xdr4.h"
  46#include "xdr4cb.h"
  47#include "vfs.h"
  48#include "current_stateid.h"
  49
  50#include "netns.h"
  51#include "pnfs.h"
  52
  53#define NFSDDBG_FACILITY                NFSDDBG_PROC
  54
  55#define all_ones {{~0,~0},~0}
  56static const stateid_t one_stateid = {
  57        .si_generation = ~0,
  58        .si_opaque = all_ones,
  59};
  60static const stateid_t zero_stateid = {
  61        /* all fields zero */
  62};
  63static const stateid_t currentstateid = {
  64        .si_generation = 1,
  65};
  66
  67static u64 current_sessionid = 1;
  68
  69#define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
  70#define ONE_STATEID(stateid)  (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
  71#define CURRENT_STATEID(stateid) (!memcmp((stateid), &currentstateid, sizeof(stateid_t)))
  72
  73/* forward declarations */
  74static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
  75static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
  76
  77/* Locking: */
  78
  79/*
  80 * Currently used for the del_recall_lru and file hash table.  In an
  81 * effort to decrease the scope of the client_mutex, this spinlock may
  82 * eventually cover more:
  83 */
  84static DEFINE_SPINLOCK(state_lock);
  85
  86/*
  87 * A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
  88 * the refcount on the open stateid to drop.
  89 */
  90static DECLARE_WAIT_QUEUE_HEAD(close_wq);
  91
  92static struct kmem_cache *openowner_slab;
  93static struct kmem_cache *lockowner_slab;
  94static struct kmem_cache *file_slab;
  95static struct kmem_cache *stateid_slab;
  96static struct kmem_cache *deleg_slab;
  97static struct kmem_cache *odstate_slab;
  98
  99static void free_session(struct nfsd4_session *);
 100
 101static struct nfsd4_callback_ops nfsd4_cb_recall_ops;
 102
 103static bool is_session_dead(struct nfsd4_session *ses)
 104{
 105        return ses->se_flags & NFS4_SESSION_DEAD;
 106}
 107
 108static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
 109{
 110        if (atomic_read(&ses->se_ref) > ref_held_by_me)
 111                return nfserr_jukebox;
 112        ses->se_flags |= NFS4_SESSION_DEAD;
 113        return nfs_ok;
 114}
 115
 116static bool is_client_expired(struct nfs4_client *clp)
 117{
 118        return clp->cl_time == 0;
 119}
 120
 121static __be32 get_client_locked(struct nfs4_client *clp)
 122{
 123        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 124
 125        lockdep_assert_held(&nn->client_lock);
 126
 127        if (is_client_expired(clp))
 128                return nfserr_expired;
 129        atomic_inc(&clp->cl_refcount);
 130        return nfs_ok;
 131}
 132
 133/* must be called under the client_lock */
 134static inline void
 135renew_client_locked(struct nfs4_client *clp)
 136{
 137        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 138
 139        if (is_client_expired(clp)) {
 140                WARN_ON(1);
 141                printk("%s: client (clientid %08x/%08x) already expired\n",
 142                        __func__,
 143                        clp->cl_clientid.cl_boot,
 144                        clp->cl_clientid.cl_id);
 145                return;
 146        }
 147
 148        dprintk("renewing client (clientid %08x/%08x)\n",
 149                        clp->cl_clientid.cl_boot,
 150                        clp->cl_clientid.cl_id);
 151        list_move_tail(&clp->cl_lru, &nn->client_lru);
 152        clp->cl_time = get_seconds();
 153}
 154
 155static void put_client_renew_locked(struct nfs4_client *clp)
 156{
 157        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 158
 159        lockdep_assert_held(&nn->client_lock);
 160
 161        if (!atomic_dec_and_test(&clp->cl_refcount))
 162                return;
 163        if (!is_client_expired(clp))
 164                renew_client_locked(clp);
 165}
 166
 167static void put_client_renew(struct nfs4_client *clp)
 168{
 169        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 170
 171        if (!atomic_dec_and_lock(&clp->cl_refcount, &nn->client_lock))
 172                return;
 173        if (!is_client_expired(clp))
 174                renew_client_locked(clp);
 175        spin_unlock(&nn->client_lock);
 176}
 177
 178static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
 179{
 180        __be32 status;
 181
 182        if (is_session_dead(ses))
 183                return nfserr_badsession;
 184        status = get_client_locked(ses->se_client);
 185        if (status)
 186                return status;
 187        atomic_inc(&ses->se_ref);
 188        return nfs_ok;
 189}
 190
 191static void nfsd4_put_session_locked(struct nfsd4_session *ses)
 192{
 193        struct nfs4_client *clp = ses->se_client;
 194        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 195
 196        lockdep_assert_held(&nn->client_lock);
 197
 198        if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
 199                free_session(ses);
 200        put_client_renew_locked(clp);
 201}
 202
 203static void nfsd4_put_session(struct nfsd4_session *ses)
 204{
 205        struct nfs4_client *clp = ses->se_client;
 206        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 207
 208        spin_lock(&nn->client_lock);
 209        nfsd4_put_session_locked(ses);
 210        spin_unlock(&nn->client_lock);
 211}
 212
 213static inline struct nfs4_stateowner *
 214nfs4_get_stateowner(struct nfs4_stateowner *sop)
 215{
 216        atomic_inc(&sop->so_count);
 217        return sop;
 218}
 219
 220static int
 221same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner)
 222{
 223        return (sop->so_owner.len == owner->len) &&
 224                0 == memcmp(sop->so_owner.data, owner->data, owner->len);
 225}
 226
 227static struct nfs4_openowner *
 228find_openstateowner_str_locked(unsigned int hashval, struct nfsd4_open *open,
 229                        struct nfs4_client *clp)
 230{
 231        struct nfs4_stateowner *so;
 232
 233        lockdep_assert_held(&clp->cl_lock);
 234
 235        list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[hashval],
 236                            so_strhash) {
 237                if (!so->so_is_open_owner)
 238                        continue;
 239                if (same_owner_str(so, &open->op_owner))
 240                        return openowner(nfs4_get_stateowner(so));
 241        }
 242        return NULL;
 243}
 244
 245static struct nfs4_openowner *
 246find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
 247                        struct nfs4_client *clp)
 248{
 249        struct nfs4_openowner *oo;
 250
 251        spin_lock(&clp->cl_lock);
 252        oo = find_openstateowner_str_locked(hashval, open, clp);
 253        spin_unlock(&clp->cl_lock);
 254        return oo;
 255}
 256
 257static inline u32
 258opaque_hashval(const void *ptr, int nbytes)
 259{
 260        unsigned char *cptr = (unsigned char *) ptr;
 261
 262        u32 x = 0;
 263        while (nbytes--) {
 264                x *= 37;
 265                x += *cptr++;
 266        }
 267        return x;
 268}
 269
 270static void nfsd4_free_file_rcu(struct rcu_head *rcu)
 271{
 272        struct nfs4_file *fp = container_of(rcu, struct nfs4_file, fi_rcu);
 273
 274        kmem_cache_free(file_slab, fp);
 275}
 276
 277void
 278put_nfs4_file(struct nfs4_file *fi)
 279{
 280        might_lock(&state_lock);
 281
 282        if (atomic_dec_and_lock(&fi->fi_ref, &state_lock)) {
 283                hlist_del_rcu(&fi->fi_hash);
 284                spin_unlock(&state_lock);
 285                WARN_ON_ONCE(!list_empty(&fi->fi_clnt_odstate));
 286                WARN_ON_ONCE(!list_empty(&fi->fi_delegations));
 287                call_rcu(&fi->fi_rcu, nfsd4_free_file_rcu);
 288        }
 289}
 290
 291static struct file *
 292__nfs4_get_fd(struct nfs4_file *f, int oflag)
 293{
 294        if (f->fi_fds[oflag])
 295                return get_file(f->fi_fds[oflag]);
 296        return NULL;
 297}
 298
 299static struct file *
 300find_writeable_file_locked(struct nfs4_file *f)
 301{
 302        struct file *ret;
 303
 304        lockdep_assert_held(&f->fi_lock);
 305
 306        ret = __nfs4_get_fd(f, O_WRONLY);
 307        if (!ret)
 308                ret = __nfs4_get_fd(f, O_RDWR);
 309        return ret;
 310}
 311
 312static struct file *
 313find_writeable_file(struct nfs4_file *f)
 314{
 315        struct file *ret;
 316
 317        spin_lock(&f->fi_lock);
 318        ret = find_writeable_file_locked(f);
 319        spin_unlock(&f->fi_lock);
 320
 321        return ret;
 322}
 323
 324static struct file *find_readable_file_locked(struct nfs4_file *f)
 325{
 326        struct file *ret;
 327
 328        lockdep_assert_held(&f->fi_lock);
 329
 330        ret = __nfs4_get_fd(f, O_RDONLY);
 331        if (!ret)
 332                ret = __nfs4_get_fd(f, O_RDWR);
 333        return ret;
 334}
 335
 336static struct file *
 337find_readable_file(struct nfs4_file *f)
 338{
 339        struct file *ret;
 340
 341        spin_lock(&f->fi_lock);
 342        ret = find_readable_file_locked(f);
 343        spin_unlock(&f->fi_lock);
 344
 345        return ret;
 346}
 347
 348struct file *
 349find_any_file(struct nfs4_file *f)
 350{
 351        struct file *ret;
 352
 353        spin_lock(&f->fi_lock);
 354        ret = __nfs4_get_fd(f, O_RDWR);
 355        if (!ret) {
 356                ret = __nfs4_get_fd(f, O_WRONLY);
 357                if (!ret)
 358                        ret = __nfs4_get_fd(f, O_RDONLY);
 359        }
 360        spin_unlock(&f->fi_lock);
 361        return ret;
 362}
 363
 364static atomic_long_t num_delegations;
 365unsigned long max_delegations;
 366
 367/*
 368 * Open owner state (share locks)
 369 */
 370
 371/* hash tables for lock and open owners */
 372#define OWNER_HASH_BITS              8
 373#define OWNER_HASH_SIZE             (1 << OWNER_HASH_BITS)
 374#define OWNER_HASH_MASK             (OWNER_HASH_SIZE - 1)
 375
 376static unsigned int ownerstr_hashval(struct xdr_netobj *ownername)
 377{
 378        unsigned int ret;
 379
 380        ret = opaque_hashval(ownername->data, ownername->len);
 381        return ret & OWNER_HASH_MASK;
 382}
 383
 384/* hash table for nfs4_file */
 385#define FILE_HASH_BITS                   8
 386#define FILE_HASH_SIZE                  (1 << FILE_HASH_BITS)
 387
 388static unsigned int nfsd_fh_hashval(struct knfsd_fh *fh)
 389{
 390        return jhash2(fh->fh_base.fh_pad, XDR_QUADLEN(fh->fh_size), 0);
 391}
 392
 393static unsigned int file_hashval(struct knfsd_fh *fh)
 394{
 395        return nfsd_fh_hashval(fh) & (FILE_HASH_SIZE - 1);
 396}
 397
 398static struct hlist_head file_hashtbl[FILE_HASH_SIZE];
 399
 400static void
 401__nfs4_file_get_access(struct nfs4_file *fp, u32 access)
 402{
 403        lockdep_assert_held(&fp->fi_lock);
 404
 405        if (access & NFS4_SHARE_ACCESS_WRITE)
 406                atomic_inc(&fp->fi_access[O_WRONLY]);
 407        if (access & NFS4_SHARE_ACCESS_READ)
 408                atomic_inc(&fp->fi_access[O_RDONLY]);
 409}
 410
 411static __be32
 412nfs4_file_get_access(struct nfs4_file *fp, u32 access)
 413{
 414        lockdep_assert_held(&fp->fi_lock);
 415
 416        /* Does this access mode make sense? */
 417        if (access & ~NFS4_SHARE_ACCESS_BOTH)
 418                return nfserr_inval;
 419
 420        /* Does it conflict with a deny mode already set? */
 421        if ((access & fp->fi_share_deny) != 0)
 422                return nfserr_share_denied;
 423
 424        __nfs4_file_get_access(fp, access);
 425        return nfs_ok;
 426}
 427
 428static __be32 nfs4_file_check_deny(struct nfs4_file *fp, u32 deny)
 429{
 430        /* Common case is that there is no deny mode. */
 431        if (deny) {
 432                /* Does this deny mode make sense? */
 433                if (deny & ~NFS4_SHARE_DENY_BOTH)
 434                        return nfserr_inval;
 435
 436                if ((deny & NFS4_SHARE_DENY_READ) &&
 437                    atomic_read(&fp->fi_access[O_RDONLY]))
 438                        return nfserr_share_denied;
 439
 440                if ((deny & NFS4_SHARE_DENY_WRITE) &&
 441                    atomic_read(&fp->fi_access[O_WRONLY]))
 442                        return nfserr_share_denied;
 443        }
 444        return nfs_ok;
 445}
 446
 447static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
 448{
 449        might_lock(&fp->fi_lock);
 450
 451        if (atomic_dec_and_lock(&fp->fi_access[oflag], &fp->fi_lock)) {
 452                struct file *f1 = NULL;
 453                struct file *f2 = NULL;
 454
 455                swap(f1, fp->fi_fds[oflag]);
 456                if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
 457                        swap(f2, fp->fi_fds[O_RDWR]);
 458                spin_unlock(&fp->fi_lock);
 459                if (f1)
 460                        fput(f1);
 461                if (f2)
 462                        fput(f2);
 463        }
 464}
 465
 466static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
 467{
 468        WARN_ON_ONCE(access & ~NFS4_SHARE_ACCESS_BOTH);
 469
 470        if (access & NFS4_SHARE_ACCESS_WRITE)
 471                __nfs4_file_put_access(fp, O_WRONLY);
 472        if (access & NFS4_SHARE_ACCESS_READ)
 473                __nfs4_file_put_access(fp, O_RDONLY);
 474}
 475
 476/*
 477 * Allocate a new open/delegation state counter. This is needed for
 478 * pNFS for proper return on close semantics.
 479 *
 480 * Note that we only allocate it for pNFS-enabled exports, otherwise
 481 * all pointers to struct nfs4_clnt_odstate are always NULL.
 482 */
 483static struct nfs4_clnt_odstate *
 484alloc_clnt_odstate(struct nfs4_client *clp)
 485{
 486        struct nfs4_clnt_odstate *co;
 487
 488        co = kmem_cache_zalloc(odstate_slab, GFP_KERNEL);
 489        if (co) {
 490                co->co_client = clp;
 491                atomic_set(&co->co_odcount, 1);
 492        }
 493        return co;
 494}
 495
 496static void
 497hash_clnt_odstate_locked(struct nfs4_clnt_odstate *co)
 498{
 499        struct nfs4_file *fp = co->co_file;
 500
 501        lockdep_assert_held(&fp->fi_lock);
 502        list_add(&co->co_perfile, &fp->fi_clnt_odstate);
 503}
 504
 505static inline void
 506get_clnt_odstate(struct nfs4_clnt_odstate *co)
 507{
 508        if (co)
 509                atomic_inc(&co->co_odcount);
 510}
 511
 512static void
 513put_clnt_odstate(struct nfs4_clnt_odstate *co)
 514{
 515        struct nfs4_file *fp;
 516
 517        if (!co)
 518                return;
 519
 520        fp = co->co_file;
 521        if (atomic_dec_and_lock(&co->co_odcount, &fp->fi_lock)) {
 522                list_del(&co->co_perfile);
 523                spin_unlock(&fp->fi_lock);
 524
 525                nfsd4_return_all_file_layouts(co->co_client, fp);
 526                kmem_cache_free(odstate_slab, co);
 527        }
 528}
 529
 530static struct nfs4_clnt_odstate *
 531find_or_hash_clnt_odstate(struct nfs4_file *fp, struct nfs4_clnt_odstate *new)
 532{
 533        struct nfs4_clnt_odstate *co;
 534        struct nfs4_client *cl;
 535
 536        if (!new)
 537                return NULL;
 538
 539        cl = new->co_client;
 540
 541        spin_lock(&fp->fi_lock);
 542        list_for_each_entry(co, &fp->fi_clnt_odstate, co_perfile) {
 543                if (co->co_client == cl) {
 544                        get_clnt_odstate(co);
 545                        goto out;
 546                }
 547        }
 548        co = new;
 549        co->co_file = fp;
 550        hash_clnt_odstate_locked(new);
 551out:
 552        spin_unlock(&fp->fi_lock);
 553        return co;
 554}
 555
 556struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl,
 557                                         struct kmem_cache *slab)
 558{
 559        struct nfs4_stid *stid;
 560        int new_id;
 561
 562        stid = kmem_cache_zalloc(slab, GFP_KERNEL);
 563        if (!stid)
 564                return NULL;
 565
 566        idr_preload(GFP_KERNEL);
 567        spin_lock(&cl->cl_lock);
 568        new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 0, 0, GFP_NOWAIT);
 569        spin_unlock(&cl->cl_lock);
 570        idr_preload_end();
 571        if (new_id < 0)
 572                goto out_free;
 573        stid->sc_client = cl;
 574        stid->sc_stateid.si_opaque.so_id = new_id;
 575        stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
 576        /* Will be incremented before return to client: */
 577        atomic_set(&stid->sc_count, 1);
 578
 579        /*
 580         * It shouldn't be a problem to reuse an opaque stateid value.
 581         * I don't think it is for 4.1.  But with 4.0 I worry that, for
 582         * example, a stray write retransmission could be accepted by
 583         * the server when it should have been rejected.  Therefore,
 584         * adopt a trick from the sctp code to attempt to maximize the
 585         * amount of time until an id is reused, by ensuring they always
 586         * "increase" (mod INT_MAX):
 587         */
 588        return stid;
 589out_free:
 590        kmem_cache_free(slab, stid);
 591        return NULL;
 592}
 593
 594static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
 595{
 596        struct nfs4_stid *stid;
 597        struct nfs4_ol_stateid *stp;
 598
 599        stid = nfs4_alloc_stid(clp, stateid_slab);
 600        if (!stid)
 601                return NULL;
 602
 603        stp = openlockstateid(stid);
 604        stp->st_stid.sc_free = nfs4_free_ol_stateid;
 605        return stp;
 606}
 607
 608static void nfs4_free_deleg(struct nfs4_stid *stid)
 609{
 610        kmem_cache_free(deleg_slab, stid);
 611        atomic_long_dec(&num_delegations);
 612}
 613
 614/*
 615 * When we recall a delegation, we should be careful not to hand it
 616 * out again straight away.
 617 * To ensure this we keep a pair of bloom filters ('new' and 'old')
 618 * in which the filehandles of recalled delegations are "stored".
 619 * If a filehandle appear in either filter, a delegation is blocked.
 620 * When a delegation is recalled, the filehandle is stored in the "new"
 621 * filter.
 622 * Every 30 seconds we swap the filters and clear the "new" one,
 623 * unless both are empty of course.
 624 *
 625 * Each filter is 256 bits.  We hash the filehandle to 32bit and use the
 626 * low 3 bytes as hash-table indices.
 627 *
 628 * 'blocked_delegations_lock', which is always taken in block_delegations(),
 629 * is used to manage concurrent access.  Testing does not need the lock
 630 * except when swapping the two filters.
 631 */
 632static DEFINE_SPINLOCK(blocked_delegations_lock);
 633static struct bloom_pair {
 634        int     entries, old_entries;
 635        time_t  swap_time;
 636        int     new; /* index into 'set' */
 637        DECLARE_BITMAP(set[2], 256);
 638} blocked_delegations;
 639
 640static int delegation_blocked(struct knfsd_fh *fh)
 641{
 642        u32 hash;
 643        struct bloom_pair *bd = &blocked_delegations;
 644
 645        if (bd->entries == 0)
 646                return 0;
 647        if (seconds_since_boot() - bd->swap_time > 30) {
 648                spin_lock(&blocked_delegations_lock);
 649                if (seconds_since_boot() - bd->swap_time > 30) {
 650                        bd->entries -= bd->old_entries;
 651                        bd->old_entries = bd->entries;
 652                        memset(bd->set[bd->new], 0,
 653                               sizeof(bd->set[0]));
 654                        bd->new = 1-bd->new;
 655                        bd->swap_time = seconds_since_boot();
 656                }
 657                spin_unlock(&blocked_delegations_lock);
 658        }
 659        hash = jhash(&fh->fh_base, fh->fh_size, 0);
 660        if (test_bit(hash&255, bd->set[0]) &&
 661            test_bit((hash>>8)&255, bd->set[0]) &&
 662            test_bit((hash>>16)&255, bd->set[0]))
 663                return 1;
 664
 665        if (test_bit(hash&255, bd->set[1]) &&
 666            test_bit((hash>>8)&255, bd->set[1]) &&
 667            test_bit((hash>>16)&255, bd->set[1]))
 668                return 1;
 669
 670        return 0;
 671}
 672
 673static void block_delegations(struct knfsd_fh *fh)
 674{
 675        u32 hash;
 676        struct bloom_pair *bd = &blocked_delegations;
 677
 678        hash = jhash(&fh->fh_base, fh->fh_size, 0);
 679
 680        spin_lock(&blocked_delegations_lock);
 681        __set_bit(hash&255, bd->set[bd->new]);
 682        __set_bit((hash>>8)&255, bd->set[bd->new]);
 683        __set_bit((hash>>16)&255, bd->set[bd->new]);
 684        if (bd->entries == 0)
 685                bd->swap_time = seconds_since_boot();
 686        bd->entries += 1;
 687        spin_unlock(&blocked_delegations_lock);
 688}
 689
 690static struct nfs4_delegation *
 691alloc_init_deleg(struct nfs4_client *clp, struct svc_fh *current_fh,
 692                 struct nfs4_clnt_odstate *odstate)
 693{
 694        struct nfs4_delegation *dp;
 695        long n;
 696
 697        dprintk("NFSD alloc_init_deleg\n");
 698        n = atomic_long_inc_return(&num_delegations);
 699        if (n < 0 || n > max_delegations)
 700                goto out_dec;
 701        if (delegation_blocked(&current_fh->fh_handle))
 702                goto out_dec;
 703        dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
 704        if (dp == NULL)
 705                goto out_dec;
 706
 707        dp->dl_stid.sc_free = nfs4_free_deleg;
 708        /*
 709         * delegation seqid's are never incremented.  The 4.1 special
 710         * meaning of seqid 0 isn't meaningful, really, but let's avoid
 711         * 0 anyway just for consistency and use 1:
 712         */
 713        dp->dl_stid.sc_stateid.si_generation = 1;
 714        INIT_LIST_HEAD(&dp->dl_perfile);
 715        INIT_LIST_HEAD(&dp->dl_perclnt);
 716        INIT_LIST_HEAD(&dp->dl_recall_lru);
 717        dp->dl_clnt_odstate = odstate;
 718        get_clnt_odstate(odstate);
 719        dp->dl_type = NFS4_OPEN_DELEGATE_READ;
 720        dp->dl_retries = 1;
 721        nfsd4_init_cb(&dp->dl_recall, dp->dl_stid.sc_client,
 722                      &nfsd4_cb_recall_ops, NFSPROC4_CLNT_CB_RECALL);
 723        return dp;
 724out_dec:
 725        atomic_long_dec(&num_delegations);
 726        return NULL;
 727}
 728
 729void
 730nfs4_put_stid(struct nfs4_stid *s)
 731{
 732        struct nfs4_file *fp = s->sc_file;
 733        struct nfs4_client *clp = s->sc_client;
 734
 735        might_lock(&clp->cl_lock);
 736
 737        if (!atomic_dec_and_lock(&s->sc_count, &clp->cl_lock)) {
 738                wake_up_all(&close_wq);
 739                return;
 740        }
 741        idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
 742        spin_unlock(&clp->cl_lock);
 743        s->sc_free(s);
 744        if (fp)
 745                put_nfs4_file(fp);
 746}
 747
 748static void nfs4_put_deleg_lease(struct nfs4_file *fp)
 749{
 750        struct file *filp = NULL;
 751
 752        spin_lock(&fp->fi_lock);
 753        if (fp->fi_deleg_file && --fp->fi_delegees == 0)
 754                swap(filp, fp->fi_deleg_file);
 755        spin_unlock(&fp->fi_lock);
 756
 757        if (filp) {
 758                vfs_setlease(filp, F_UNLCK, NULL, (void **)&fp);
 759                fput(filp);
 760        }
 761}
 762
 763void nfs4_unhash_stid(struct nfs4_stid *s)
 764{
 765        s->sc_type = 0;
 766}
 767
 768static void
 769hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
 770{
 771        lockdep_assert_held(&state_lock);
 772        lockdep_assert_held(&fp->fi_lock);
 773
 774        atomic_inc(&dp->dl_stid.sc_count);
 775        dp->dl_stid.sc_type = NFS4_DELEG_STID;
 776        list_add(&dp->dl_perfile, &fp->fi_delegations);
 777        list_add(&dp->dl_perclnt, &dp->dl_stid.sc_client->cl_delegations);
 778}
 779
 780static void
 781unhash_delegation_locked(struct nfs4_delegation *dp)
 782{
 783        struct nfs4_file *fp = dp->dl_stid.sc_file;
 784
 785        lockdep_assert_held(&state_lock);
 786
 787        dp->dl_stid.sc_type = NFS4_CLOSED_DELEG_STID;
 788        /* Ensure that deleg break won't try to requeue it */
 789        ++dp->dl_time;
 790        spin_lock(&fp->fi_lock);
 791        list_del_init(&dp->dl_perclnt);
 792        list_del_init(&dp->dl_recall_lru);
 793        list_del_init(&dp->dl_perfile);
 794        spin_unlock(&fp->fi_lock);
 795}
 796
 797static void destroy_delegation(struct nfs4_delegation *dp)
 798{
 799        spin_lock(&state_lock);
 800        unhash_delegation_locked(dp);
 801        spin_unlock(&state_lock);
 802        put_clnt_odstate(dp->dl_clnt_odstate);
 803        nfs4_put_deleg_lease(dp->dl_stid.sc_file);
 804        nfs4_put_stid(&dp->dl_stid);
 805}
 806
 807static void revoke_delegation(struct nfs4_delegation *dp)
 808{
 809        struct nfs4_client *clp = dp->dl_stid.sc_client;
 810
 811        WARN_ON(!list_empty(&dp->dl_recall_lru));
 812
 813        put_clnt_odstate(dp->dl_clnt_odstate);
 814        nfs4_put_deleg_lease(dp->dl_stid.sc_file);
 815
 816        if (clp->cl_minorversion == 0)
 817                nfs4_put_stid(&dp->dl_stid);
 818        else {
 819                dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
 820                spin_lock(&clp->cl_lock);
 821                list_add(&dp->dl_recall_lru, &clp->cl_revoked);
 822                spin_unlock(&clp->cl_lock);
 823        }
 824}
 825
 826/* 
 827 * SETCLIENTID state 
 828 */
 829
 830static unsigned int clientid_hashval(u32 id)
 831{
 832        return id & CLIENT_HASH_MASK;
 833}
 834
 835static unsigned int clientstr_hashval(const char *name)
 836{
 837        return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
 838}
 839
 840/*
 841 * We store the NONE, READ, WRITE, and BOTH bits separately in the
 842 * st_{access,deny}_bmap field of the stateid, in order to track not
 843 * only what share bits are currently in force, but also what
 844 * combinations of share bits previous opens have used.  This allows us
 845 * to enforce the recommendation of rfc 3530 14.2.19 that the server
 846 * return an error if the client attempt to downgrade to a combination
 847 * of share bits not explicable by closing some of its previous opens.
 848 *
 849 * XXX: This enforcement is actually incomplete, since we don't keep
 850 * track of access/deny bit combinations; so, e.g., we allow:
 851 *
 852 *      OPEN allow read, deny write
 853 *      OPEN allow both, deny none
 854 *      DOWNGRADE allow read, deny none
 855 *
 856 * which we should reject.
 857 */
 858static unsigned int
 859bmap_to_share_mode(unsigned long bmap) {
 860        int i;
 861        unsigned int access = 0;
 862
 863        for (i = 1; i < 4; i++) {
 864                if (test_bit(i, &bmap))
 865                        access |= i;
 866        }
 867        return access;
 868}
 869
 870/* set share access for a given stateid */
 871static inline void
 872set_access(u32 access, struct nfs4_ol_stateid *stp)
 873{
 874        unsigned char mask = 1 << access;
 875
 876        WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
 877        stp->st_access_bmap |= mask;
 878}
 879
 880/* clear share access for a given stateid */
 881static inline void
 882clear_access(u32 access, struct nfs4_ol_stateid *stp)
 883{
 884        unsigned char mask = 1 << access;
 885
 886        WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
 887        stp->st_access_bmap &= ~mask;
 888}
 889
 890/* test whether a given stateid has access */
 891static inline bool
 892test_access(u32 access, struct nfs4_ol_stateid *stp)
 893{
 894        unsigned char mask = 1 << access;
 895
 896        return (bool)(stp->st_access_bmap & mask);
 897}
 898
 899/* set share deny for a given stateid */
 900static inline void
 901set_deny(u32 deny, struct nfs4_ol_stateid *stp)
 902{
 903        unsigned char mask = 1 << deny;
 904
 905        WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
 906        stp->st_deny_bmap |= mask;
 907}
 908
 909/* clear share deny for a given stateid */
 910static inline void
 911clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
 912{
 913        unsigned char mask = 1 << deny;
 914
 915        WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
 916        stp->st_deny_bmap &= ~mask;
 917}
 918
 919/* test whether a given stateid is denying specific access */
 920static inline bool
 921test_deny(u32 deny, struct nfs4_ol_stateid *stp)
 922{
 923        unsigned char mask = 1 << deny;
 924
 925        return (bool)(stp->st_deny_bmap & mask);
 926}
 927
 928static int nfs4_access_to_omode(u32 access)
 929{
 930        switch (access & NFS4_SHARE_ACCESS_BOTH) {
 931        case NFS4_SHARE_ACCESS_READ:
 932                return O_RDONLY;
 933        case NFS4_SHARE_ACCESS_WRITE:
 934                return O_WRONLY;
 935        case NFS4_SHARE_ACCESS_BOTH:
 936                return O_RDWR;
 937        }
 938        WARN_ON_ONCE(1);
 939        return O_RDONLY;
 940}
 941
 942/*
 943 * A stateid that had a deny mode associated with it is being released
 944 * or downgraded. Recalculate the deny mode on the file.
 945 */
 946static void
 947recalculate_deny_mode(struct nfs4_file *fp)
 948{
 949        struct nfs4_ol_stateid *stp;
 950
 951        spin_lock(&fp->fi_lock);
 952        fp->fi_share_deny = 0;
 953        list_for_each_entry(stp, &fp->fi_stateids, st_perfile)
 954                fp->fi_share_deny |= bmap_to_share_mode(stp->st_deny_bmap);
 955        spin_unlock(&fp->fi_lock);
 956}
 957
 958static void
 959reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
 960{
 961        int i;
 962        bool change = false;
 963
 964        for (i = 1; i < 4; i++) {
 965                if ((i & deny) != i) {
 966                        change = true;
 967                        clear_deny(i, stp);
 968                }
 969        }
 970
 971        /* Recalculate per-file deny mode if there was a change */
 972        if (change)
 973                recalculate_deny_mode(stp->st_stid.sc_file);
 974}
 975
 976/* release all access and file references for a given stateid */
 977static void
 978release_all_access(struct nfs4_ol_stateid *stp)
 979{
 980        int i;
 981        struct nfs4_file *fp = stp->st_stid.sc_file;
 982
 983        if (fp && stp->st_deny_bmap != 0)
 984                recalculate_deny_mode(fp);
 985
 986        for (i = 1; i < 4; i++) {
 987                if (test_access(i, stp))
 988                        nfs4_file_put_access(stp->st_stid.sc_file, i);
 989                clear_access(i, stp);
 990        }
 991}
 992
 993static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
 994{
 995        struct nfs4_client *clp = sop->so_client;
 996
 997        might_lock(&clp->cl_lock);
 998
 999        if (!atomic_dec_and_lock(&sop->so_count, &clp->cl_lock))
1000                return;
1001        sop->so_ops->so_unhash(sop);
1002        spin_unlock(&clp->cl_lock);
1003        kfree(sop->so_owner.data);
1004        sop->so_ops->so_free(sop);
1005}
1006
1007static void unhash_ol_stateid(struct nfs4_ol_stateid *stp)
1008{
1009        struct nfs4_file *fp = stp->st_stid.sc_file;
1010
1011        lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
1012
1013        spin_lock(&fp->fi_lock);
1014        list_del(&stp->st_perfile);
1015        spin_unlock(&fp->fi_lock);
1016        list_del(&stp->st_perstateowner);
1017}
1018
1019static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
1020{
1021        struct nfs4_ol_stateid *stp = openlockstateid(stid);
1022
1023        put_clnt_odstate(stp->st_clnt_odstate);
1024        release_all_access(stp);
1025        if (stp->st_stateowner)
1026                nfs4_put_stateowner(stp->st_stateowner);
1027        kmem_cache_free(stateid_slab, stid);
1028}
1029
1030static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
1031{
1032        struct nfs4_ol_stateid *stp = openlockstateid(stid);
1033        struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
1034        struct file *file;
1035
1036        file = find_any_file(stp->st_stid.sc_file);
1037        if (file)
1038                filp_close(file, (fl_owner_t)lo);
1039        nfs4_free_ol_stateid(stid);
1040}
1041
1042/*
1043 * Put the persistent reference to an already unhashed generic stateid, while
1044 * holding the cl_lock. If it's the last reference, then put it onto the
1045 * reaplist for later destruction.
1046 */
1047static void put_ol_stateid_locked(struct nfs4_ol_stateid *stp,
1048                                       struct list_head *reaplist)
1049{
1050        struct nfs4_stid *s = &stp->st_stid;
1051        struct nfs4_client *clp = s->sc_client;
1052
1053        lockdep_assert_held(&clp->cl_lock);
1054
1055        WARN_ON_ONCE(!list_empty(&stp->st_locks));
1056
1057        if (!atomic_dec_and_test(&s->sc_count)) {
1058                wake_up_all(&close_wq);
1059                return;
1060        }
1061
1062        idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1063        list_add(&stp->st_locks, reaplist);
1064}
1065
1066static void unhash_lock_stateid(struct nfs4_ol_stateid *stp)
1067{
1068        struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
1069
1070        lockdep_assert_held(&oo->oo_owner.so_client->cl_lock);
1071
1072        list_del_init(&stp->st_locks);
1073        unhash_ol_stateid(stp);
1074        nfs4_unhash_stid(&stp->st_stid);
1075}
1076
1077static void release_lock_stateid(struct nfs4_ol_stateid *stp)
1078{
1079        struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
1080
1081        spin_lock(&oo->oo_owner.so_client->cl_lock);
1082        unhash_lock_stateid(stp);
1083        spin_unlock(&oo->oo_owner.so_client->cl_lock);
1084        nfs4_put_stid(&stp->st_stid);
1085}
1086
1087static void unhash_lockowner_locked(struct nfs4_lockowner *lo)
1088{
1089        struct nfs4_client *clp = lo->lo_owner.so_client;
1090
1091        lockdep_assert_held(&clp->cl_lock);
1092
1093        list_del_init(&lo->lo_owner.so_strhash);
1094}
1095
1096/*
1097 * Free a list of generic stateids that were collected earlier after being
1098 * fully unhashed.
1099 */
1100static void
1101free_ol_stateid_reaplist(struct list_head *reaplist)
1102{
1103        struct nfs4_ol_stateid *stp;
1104        struct nfs4_file *fp;
1105
1106        might_sleep();
1107
1108        while (!list_empty(reaplist)) {
1109                stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
1110                                       st_locks);
1111                list_del(&stp->st_locks);
1112                fp = stp->st_stid.sc_file;
1113                stp->st_stid.sc_free(&stp->st_stid);
1114                if (fp)
1115                        put_nfs4_file(fp);
1116        }
1117}
1118
1119static void release_lockowner(struct nfs4_lockowner *lo)
1120{
1121        struct nfs4_client *clp = lo->lo_owner.so_client;
1122        struct nfs4_ol_stateid *stp;
1123        struct list_head reaplist;
1124
1125        INIT_LIST_HEAD(&reaplist);
1126
1127        spin_lock(&clp->cl_lock);
1128        unhash_lockowner_locked(lo);
1129        while (!list_empty(&lo->lo_owner.so_stateids)) {
1130                stp = list_first_entry(&lo->lo_owner.so_stateids,
1131                                struct nfs4_ol_stateid, st_perstateowner);
1132                unhash_lock_stateid(stp);
1133                put_ol_stateid_locked(stp, &reaplist);
1134        }
1135        spin_unlock(&clp->cl_lock);
1136        free_ol_stateid_reaplist(&reaplist);
1137        nfs4_put_stateowner(&lo->lo_owner);
1138}
1139
1140static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp,
1141                                       struct list_head *reaplist)
1142{
1143        struct nfs4_ol_stateid *stp;
1144
1145        while (!list_empty(&open_stp->st_locks)) {
1146                stp = list_entry(open_stp->st_locks.next,
1147                                struct nfs4_ol_stateid, st_locks);
1148                unhash_lock_stateid(stp);
1149                put_ol_stateid_locked(stp, reaplist);
1150        }
1151}
1152
1153static void unhash_open_stateid(struct nfs4_ol_stateid *stp,
1154                                struct list_head *reaplist)
1155{
1156        lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1157
1158        unhash_ol_stateid(stp);
1159        release_open_stateid_locks(stp, reaplist);
1160}
1161
1162static void release_open_stateid(struct nfs4_ol_stateid *stp)
1163{
1164        LIST_HEAD(reaplist);
1165
1166        spin_lock(&stp->st_stid.sc_client->cl_lock);
1167        unhash_open_stateid(stp, &reaplist);
1168        put_ol_stateid_locked(stp, &reaplist);
1169        spin_unlock(&stp->st_stid.sc_client->cl_lock);
1170        free_ol_stateid_reaplist(&reaplist);
1171}
1172
1173static void unhash_openowner_locked(struct nfs4_openowner *oo)
1174{
1175        struct nfs4_client *clp = oo->oo_owner.so_client;
1176
1177        lockdep_assert_held(&clp->cl_lock);
1178
1179        list_del_init(&oo->oo_owner.so_strhash);
1180        list_del_init(&oo->oo_perclient);
1181}
1182
1183static void release_last_closed_stateid(struct nfs4_openowner *oo)
1184{
1185        struct nfsd_net *nn = net_generic(oo->oo_owner.so_client->net,
1186                                          nfsd_net_id);
1187        struct nfs4_ol_stateid *s;
1188
1189        spin_lock(&nn->client_lock);
1190        s = oo->oo_last_closed_stid;
1191        if (s) {
1192                list_del_init(&oo->oo_close_lru);
1193                oo->oo_last_closed_stid = NULL;
1194        }
1195        spin_unlock(&nn->client_lock);
1196        if (s)
1197                nfs4_put_stid(&s->st_stid);
1198}
1199
1200static void release_openowner(struct nfs4_openowner *oo)
1201{
1202        struct nfs4_ol_stateid *stp;
1203        struct nfs4_client *clp = oo->oo_owner.so_client;
1204        struct list_head reaplist;
1205
1206        INIT_LIST_HEAD(&reaplist);
1207
1208        spin_lock(&clp->cl_lock);
1209        unhash_openowner_locked(oo);
1210        while (!list_empty(&oo->oo_owner.so_stateids)) {
1211                stp = list_first_entry(&oo->oo_owner.so_stateids,
1212                                struct nfs4_ol_stateid, st_perstateowner);
1213                unhash_open_stateid(stp, &reaplist);
1214                put_ol_stateid_locked(stp, &reaplist);
1215        }
1216        spin_unlock(&clp->cl_lock);
1217        free_ol_stateid_reaplist(&reaplist);
1218        release_last_closed_stateid(oo);
1219        nfs4_put_stateowner(&oo->oo_owner);
1220}
1221
1222static inline int
1223hash_sessionid(struct nfs4_sessionid *sessionid)
1224{
1225        struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1226
1227        return sid->sequence % SESSION_HASH_SIZE;
1228}
1229
1230#ifdef CONFIG_SUNRPC_DEBUG
1231static inline void
1232dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1233{
1234        u32 *ptr = (u32 *)(&sessionid->data[0]);
1235        dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1236}
1237#else
1238static inline void
1239dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1240{
1241}
1242#endif
1243
1244/*
1245 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1246 * won't be used for replay.
1247 */
1248void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1249{
1250        struct nfs4_stateowner *so = cstate->replay_owner;
1251
1252        if (nfserr == nfserr_replay_me)
1253                return;
1254
1255        if (!seqid_mutating_err(ntohl(nfserr))) {
1256                nfsd4_cstate_clear_replay(cstate);
1257                return;
1258        }
1259        if (!so)
1260                return;
1261        if (so->so_is_open_owner)
1262                release_last_closed_stateid(openowner(so));
1263        so->so_seqid++;
1264        return;
1265}
1266
1267static void
1268gen_sessionid(struct nfsd4_session *ses)
1269{
1270        struct nfs4_client *clp = ses->se_client;
1271        struct nfsd4_sessionid *sid;
1272
1273        sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1274        sid->clientid = clp->cl_clientid;
1275        sid->sequence = current_sessionid++;
1276        sid->reserved = 0;
1277}
1278
1279/*
1280 * The protocol defines ca_maxresponssize_cached to include the size of
1281 * the rpc header, but all we need to cache is the data starting after
1282 * the end of the initial SEQUENCE operation--the rest we regenerate
1283 * each time.  Therefore we can advertise a ca_maxresponssize_cached
1284 * value that is the number of bytes in our cache plus a few additional
1285 * bytes.  In order to stay on the safe side, and not promise more than
1286 * we can cache, those additional bytes must be the minimum possible: 24
1287 * bytes of rpc header (xid through accept state, with AUTH_NULL
1288 * verifier), 12 for the compound header (with zero-length tag), and 44
1289 * for the SEQUENCE op response:
1290 */
1291#define NFSD_MIN_HDR_SEQ_SZ  (24 + 12 + 44)
1292
1293static void
1294free_session_slots(struct nfsd4_session *ses)
1295{
1296        int i;
1297
1298        for (i = 0; i < ses->se_fchannel.maxreqs; i++)
1299                kfree(ses->se_slots[i]);
1300}
1301
1302/*
1303 * We don't actually need to cache the rpc and session headers, so we
1304 * can allocate a little less for each slot:
1305 */
1306static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
1307{
1308        u32 size;
1309
1310        if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
1311                size = 0;
1312        else
1313                size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
1314        return size + sizeof(struct nfsd4_slot);
1315}
1316
1317/*
1318 * XXX: If we run out of reserved DRC memory we could (up to a point)
1319 * re-negotiate active sessions and reduce their slot usage to make
1320 * room for new connections. For now we just fail the create session.
1321 */
1322static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
1323{
1324        u32 slotsize = slot_bytes(ca);
1325        u32 num = ca->maxreqs;
1326        int avail;
1327
1328        spin_lock(&nfsd_drc_lock);
1329        avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
1330                    nfsd_drc_max_mem - nfsd_drc_mem_used);
1331        num = min_t(int, num, avail / slotsize);
1332        nfsd_drc_mem_used += num * slotsize;
1333        spin_unlock(&nfsd_drc_lock);
1334
1335        return num;
1336}
1337
1338static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
1339{
1340        int slotsize = slot_bytes(ca);
1341
1342        spin_lock(&nfsd_drc_lock);
1343        nfsd_drc_mem_used -= slotsize * ca->maxreqs;
1344        spin_unlock(&nfsd_drc_lock);
1345}
1346
1347static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
1348                                           struct nfsd4_channel_attrs *battrs)
1349{
1350        int numslots = fattrs->maxreqs;
1351        int slotsize = slot_bytes(fattrs);
1352        struct nfsd4_session *new;
1353        int mem, i;
1354
1355        BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
1356                        + sizeof(struct nfsd4_session) > PAGE_SIZE);
1357        mem = numslots * sizeof(struct nfsd4_slot *);
1358
1359        new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
1360        if (!new)
1361                return NULL;
1362        /* allocate each struct nfsd4_slot and data cache in one piece */
1363        for (i = 0; i < numslots; i++) {
1364                new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
1365                if (!new->se_slots[i])
1366                        goto out_free;
1367        }
1368
1369        memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
1370        memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
1371
1372        return new;
1373out_free:
1374        while (i--)
1375                kfree(new->se_slots[i]);
1376        kfree(new);
1377        return NULL;
1378}
1379
1380static void free_conn(struct nfsd4_conn *c)
1381{
1382        svc_xprt_put(c->cn_xprt);
1383        kfree(c);
1384}
1385
1386static void nfsd4_conn_lost(struct svc_xpt_user *u)
1387{
1388        struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
1389        struct nfs4_client *clp = c->cn_session->se_client;
1390
1391        spin_lock(&clp->cl_lock);
1392        if (!list_empty(&c->cn_persession)) {
1393                list_del(&c->cn_persession);
1394                free_conn(c);
1395        }
1396        nfsd4_probe_callback(clp);
1397        spin_unlock(&clp->cl_lock);
1398}
1399
1400static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1401{
1402        struct nfsd4_conn *conn;
1403
1404        conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1405        if (!conn)
1406                return NULL;
1407        svc_xprt_get(rqstp->rq_xprt);
1408        conn->cn_xprt = rqstp->rq_xprt;
1409        conn->cn_flags = flags;
1410        INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1411        return conn;
1412}
1413
1414static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1415{
1416        conn->cn_session = ses;
1417        list_add(&conn->cn_persession, &ses->se_conns);
1418}
1419
1420static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1421{
1422        struct nfs4_client *clp = ses->se_client;
1423
1424        spin_lock(&clp->cl_lock);
1425        __nfsd4_hash_conn(conn, ses);
1426        spin_unlock(&clp->cl_lock);
1427}
1428
1429static int nfsd4_register_conn(struct nfsd4_conn *conn)
1430{
1431        conn->cn_xpt_user.callback = nfsd4_conn_lost;
1432        return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1433}
1434
1435static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1436{
1437        int ret;
1438
1439        nfsd4_hash_conn(conn, ses);
1440        ret = nfsd4_register_conn(conn);
1441        if (ret)
1442                /* oops; xprt is already down: */
1443                nfsd4_conn_lost(&conn->cn_xpt_user);
1444        /* We may have gained or lost a callback channel: */
1445        nfsd4_probe_callback_sync(ses->se_client);
1446}
1447
1448static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1449{
1450        u32 dir = NFS4_CDFC4_FORE;
1451
1452        if (cses->flags & SESSION4_BACK_CHAN)
1453                dir |= NFS4_CDFC4_BACK;
1454        return alloc_conn(rqstp, dir);
1455}
1456
1457/* must be called under client_lock */
1458static void nfsd4_del_conns(struct nfsd4_session *s)
1459{
1460        struct nfs4_client *clp = s->se_client;
1461        struct nfsd4_conn *c;
1462
1463        spin_lock(&clp->cl_lock);
1464        while (!list_empty(&s->se_conns)) {
1465                c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1466                list_del_init(&c->cn_persession);
1467                spin_unlock(&clp->cl_lock);
1468
1469                unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1470                free_conn(c);
1471
1472                spin_lock(&clp->cl_lock);
1473        }
1474        spin_unlock(&clp->cl_lock);
1475}
1476
1477static void __free_session(struct nfsd4_session *ses)
1478{
1479        free_session_slots(ses);
1480        kfree(ses);
1481}
1482
1483static void free_session(struct nfsd4_session *ses)
1484{
1485        nfsd4_del_conns(ses);
1486        nfsd4_put_drc_mem(&ses->se_fchannel);
1487        __free_session(ses);
1488}
1489
1490static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1491{
1492        int idx;
1493        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1494
1495        new->se_client = clp;
1496        gen_sessionid(new);
1497
1498        INIT_LIST_HEAD(&new->se_conns);
1499
1500        new->se_cb_seq_nr = 1;
1501        new->se_flags = cses->flags;
1502        new->se_cb_prog = cses->callback_prog;
1503        new->se_cb_sec = cses->cb_sec;
1504        atomic_set(&new->se_ref, 0);
1505        idx = hash_sessionid(&new->se_sessionid);
1506        list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1507        spin_lock(&clp->cl_lock);
1508        list_add(&new->se_perclnt, &clp->cl_sessions);
1509        spin_unlock(&clp->cl_lock);
1510
1511        {
1512                struct sockaddr *sa = svc_addr(rqstp);
1513                /*
1514                 * This is a little silly; with sessions there's no real
1515                 * use for the callback address.  Use the peer address
1516                 * as a reasonable default for now, but consider fixing
1517                 * the rpc client not to require an address in the
1518                 * future:
1519                 */
1520                rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1521                clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1522        }
1523}
1524
1525/* caller must hold client_lock */
1526static struct nfsd4_session *
1527__find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1528{
1529        struct nfsd4_session *elem;
1530        int idx;
1531        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1532
1533        lockdep_assert_held(&nn->client_lock);
1534
1535        dump_sessionid(__func__, sessionid);
1536        idx = hash_sessionid(sessionid);
1537        /* Search in the appropriate list */
1538        list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1539                if (!memcmp(elem->se_sessionid.data, sessionid->data,
1540                            NFS4_MAX_SESSIONID_LEN)) {
1541                        return elem;
1542                }
1543        }
1544
1545        dprintk("%s: session not found\n", __func__);
1546        return NULL;
1547}
1548
1549static struct nfsd4_session *
1550find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
1551                __be32 *ret)
1552{
1553        struct nfsd4_session *session;
1554        __be32 status = nfserr_badsession;
1555
1556        session = __find_in_sessionid_hashtbl(sessionid, net);
1557        if (!session)
1558                goto out;
1559        status = nfsd4_get_session_locked(session);
1560        if (status)
1561                session = NULL;
1562out:
1563        *ret = status;
1564        return session;
1565}
1566
1567/* caller must hold client_lock */
1568static void
1569unhash_session(struct nfsd4_session *ses)
1570{
1571        struct nfs4_client *clp = ses->se_client;
1572        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1573
1574        lockdep_assert_held(&nn->client_lock);
1575
1576        list_del(&ses->se_hash);
1577        spin_lock(&ses->se_client->cl_lock);
1578        list_del(&ses->se_perclnt);
1579        spin_unlock(&ses->se_client->cl_lock);
1580}
1581
1582/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1583static int
1584STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1585{
1586        /*
1587         * We're assuming the clid was not given out from a boot
1588         * precisely 2^32 (about 136 years) before this one.  That seems
1589         * a safe assumption:
1590         */
1591        if (clid->cl_boot == (u32)nn->boot_time)
1592                return 0;
1593        dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1594                clid->cl_boot, clid->cl_id, nn->boot_time);
1595        return 1;
1596}
1597
1598/* 
1599 * XXX Should we use a slab cache ?
1600 * This type of memory management is somewhat inefficient, but we use it
1601 * anyway since SETCLIENTID is not a common operation.
1602 */
1603static struct nfs4_client *alloc_client(struct xdr_netobj name)
1604{
1605        struct nfs4_client *clp;
1606        int i;
1607
1608        clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1609        if (clp == NULL)
1610                return NULL;
1611        clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1612        if (clp->cl_name.data == NULL)
1613                goto err_no_name;
1614        clp->cl_ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
1615                        OWNER_HASH_SIZE, GFP_KERNEL);
1616        if (!clp->cl_ownerstr_hashtbl)
1617                goto err_no_hashtbl;
1618        for (i = 0; i < OWNER_HASH_SIZE; i++)
1619                INIT_LIST_HEAD(&clp->cl_ownerstr_hashtbl[i]);
1620        clp->cl_name.len = name.len;
1621        INIT_LIST_HEAD(&clp->cl_sessions);
1622        idr_init(&clp->cl_stateids);
1623        atomic_set(&clp->cl_refcount, 0);
1624        clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1625        INIT_LIST_HEAD(&clp->cl_idhash);
1626        INIT_LIST_HEAD(&clp->cl_openowners);
1627        INIT_LIST_HEAD(&clp->cl_delegations);
1628        INIT_LIST_HEAD(&clp->cl_lru);
1629        INIT_LIST_HEAD(&clp->cl_revoked);
1630#ifdef CONFIG_NFSD_PNFS
1631        INIT_LIST_HEAD(&clp->cl_lo_states);
1632#endif
1633        spin_lock_init(&clp->cl_lock);
1634        rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1635        return clp;
1636err_no_hashtbl:
1637        kfree(clp->cl_name.data);
1638err_no_name:
1639        kfree(clp);
1640        return NULL;
1641}
1642
1643static void
1644free_client(struct nfs4_client *clp)
1645{
1646        while (!list_empty(&clp->cl_sessions)) {
1647                struct nfsd4_session *ses;
1648                ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1649                                se_perclnt);
1650                list_del(&ses->se_perclnt);
1651                WARN_ON_ONCE(atomic_read(&ses->se_ref));
1652                free_session(ses);
1653        }
1654        rpc_destroy_wait_queue(&clp->cl_cb_waitq);
1655        free_svc_cred(&clp->cl_cred);
1656        kfree(clp->cl_ownerstr_hashtbl);
1657        kfree(clp->cl_name.data);
1658        idr_destroy(&clp->cl_stateids);
1659        kfree(clp);
1660}
1661
1662/* must be called under the client_lock */
1663static void
1664unhash_client_locked(struct nfs4_client *clp)
1665{
1666        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1667        struct nfsd4_session *ses;
1668
1669        lockdep_assert_held(&nn->client_lock);
1670
1671        /* Mark the client as expired! */
1672        clp->cl_time = 0;
1673        /* Make it invisible */
1674        if (!list_empty(&clp->cl_idhash)) {
1675                list_del_init(&clp->cl_idhash);
1676                if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1677                        rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1678                else
1679                        rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1680        }
1681        list_del_init(&clp->cl_lru);
1682        spin_lock(&clp->cl_lock);
1683        list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1684                list_del_init(&ses->se_hash);
1685        spin_unlock(&clp->cl_lock);
1686}
1687
1688static void
1689unhash_client(struct nfs4_client *clp)
1690{
1691        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1692
1693        spin_lock(&nn->client_lock);
1694        unhash_client_locked(clp);
1695        spin_unlock(&nn->client_lock);
1696}
1697
1698static __be32 mark_client_expired_locked(struct nfs4_client *clp)
1699{
1700        if (atomic_read(&clp->cl_refcount))
1701                return nfserr_jukebox;
1702        unhash_client_locked(clp);
1703        return nfs_ok;
1704}
1705
1706static void
1707__destroy_client(struct nfs4_client *clp)
1708{
1709        struct nfs4_openowner *oo;
1710        struct nfs4_delegation *dp;
1711        struct list_head reaplist;
1712
1713        INIT_LIST_HEAD(&reaplist);
1714        spin_lock(&state_lock);
1715        while (!list_empty(&clp->cl_delegations)) {
1716                dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1717                unhash_delegation_locked(dp);
1718                list_add(&dp->dl_recall_lru, &reaplist);
1719        }
1720        spin_unlock(&state_lock);
1721        while (!list_empty(&reaplist)) {
1722                dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1723                list_del_init(&dp->dl_recall_lru);
1724                put_clnt_odstate(dp->dl_clnt_odstate);
1725                nfs4_put_deleg_lease(dp->dl_stid.sc_file);
1726                nfs4_put_stid(&dp->dl_stid);
1727        }
1728        while (!list_empty(&clp->cl_revoked)) {
1729                dp = list_entry(clp->cl_revoked.next, struct nfs4_delegation, dl_recall_lru);
1730                list_del_init(&dp->dl_recall_lru);
1731                nfs4_put_stid(&dp->dl_stid);
1732        }
1733        while (!list_empty(&clp->cl_openowners)) {
1734                oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1735                nfs4_get_stateowner(&oo->oo_owner);
1736                release_openowner(oo);
1737        }
1738        nfsd4_return_all_client_layouts(clp);
1739        nfsd4_shutdown_callback(clp);
1740        if (clp->cl_cb_conn.cb_xprt)
1741                svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1742        free_client(clp);
1743}
1744
1745static void
1746destroy_client(struct nfs4_client *clp)
1747{
1748        unhash_client(clp);
1749        __destroy_client(clp);
1750}
1751
1752static void expire_client(struct nfs4_client *clp)
1753{
1754        unhash_client(clp);
1755        nfsd4_client_record_remove(clp);
1756        __destroy_client(clp);
1757}
1758
1759static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1760{
1761        memcpy(target->cl_verifier.data, source->data,
1762                        sizeof(target->cl_verifier.data));
1763}
1764
1765static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1766{
1767        target->cl_clientid.cl_boot = source->cl_clientid.cl_boot; 
1768        target->cl_clientid.cl_id = source->cl_clientid.cl_id; 
1769}
1770
1771static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1772{
1773        if (source->cr_principal) {
1774                target->cr_principal =
1775                                kstrdup(source->cr_principal, GFP_KERNEL);
1776                if (target->cr_principal == NULL)
1777                        return -ENOMEM;
1778        } else
1779                target->cr_principal = NULL;
1780        target->cr_flavor = source->cr_flavor;
1781        target->cr_uid = source->cr_uid;
1782        target->cr_gid = source->cr_gid;
1783        target->cr_group_info = source->cr_group_info;
1784        get_group_info(target->cr_group_info);
1785        target->cr_gss_mech = source->cr_gss_mech;
1786        if (source->cr_gss_mech)
1787                gss_mech_get(source->cr_gss_mech);
1788        return 0;
1789}
1790
1791static int
1792compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1793{
1794        if (o1->len < o2->len)
1795                return -1;
1796        if (o1->len > o2->len)
1797                return 1;
1798        return memcmp(o1->data, o2->data, o1->len);
1799}
1800
1801static int same_name(const char *n1, const char *n2)
1802{
1803        return 0 == memcmp(n1, n2, HEXDIR_LEN);
1804}
1805
1806static int
1807same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1808{
1809        return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1810}
1811
1812static int
1813same_clid(clientid_t *cl1, clientid_t *cl2)
1814{
1815        return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1816}
1817
1818static bool groups_equal(struct group_info *g1, struct group_info *g2)
1819{
1820        int i;
1821
1822        if (g1->ngroups != g2->ngroups)
1823                return false;
1824        for (i=0; i<g1->ngroups; i++)
1825                if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1826                        return false;
1827        return true;
1828}
1829
1830/*
1831 * RFC 3530 language requires clid_inuse be returned when the
1832 * "principal" associated with a requests differs from that previously
1833 * used.  We use uid, gid's, and gss principal string as our best
1834 * approximation.  We also don't want to allow non-gss use of a client
1835 * established using gss: in theory cr_principal should catch that
1836 * change, but in practice cr_principal can be null even in the gss case
1837 * since gssd doesn't always pass down a principal string.
1838 */
1839static bool is_gss_cred(struct svc_cred *cr)
1840{
1841        /* Is cr_flavor one of the gss "pseudoflavors"?: */
1842        return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1843}
1844
1845
1846static bool
1847same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1848{
1849        if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1850                || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1851                || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1852                || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1853                return false;
1854        if (cr1->cr_principal == cr2->cr_principal)
1855                return true;
1856        if (!cr1->cr_principal || !cr2->cr_principal)
1857                return false;
1858        return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1859}
1860
1861static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
1862{
1863        struct svc_cred *cr = &rqstp->rq_cred;
1864        u32 service;
1865
1866        if (!cr->cr_gss_mech)
1867                return false;
1868        service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
1869        return service == RPC_GSS_SVC_INTEGRITY ||
1870               service == RPC_GSS_SVC_PRIVACY;
1871}
1872
1873static bool mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
1874{
1875        struct svc_cred *cr = &rqstp->rq_cred;
1876
1877        if (!cl->cl_mach_cred)
1878                return true;
1879        if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
1880                return false;
1881        if (!svc_rqst_integrity_protected(rqstp))
1882                return false;
1883        if (!cr->cr_principal)
1884                return false;
1885        return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
1886}
1887
1888static void gen_confirm(struct nfs4_client *clp, struct nfsd_net *nn)
1889{
1890        __be32 verf[2];
1891
1892        /*
1893         * This is opaque to client, so no need to byte-swap. Use
1894         * __force to keep sparse happy
1895         */
1896        verf[0] = (__force __be32)get_seconds();
1897        verf[1] = (__force __be32)nn->clientid_counter;
1898        memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
1899}
1900
1901static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
1902{
1903        clp->cl_clientid.cl_boot = nn->boot_time;
1904        clp->cl_clientid.cl_id = nn->clientid_counter++;
1905        gen_confirm(clp, nn);
1906}
1907
1908static struct nfs4_stid *
1909find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
1910{
1911        struct nfs4_stid *ret;
1912
1913        ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
1914        if (!ret || !ret->sc_type)
1915                return NULL;
1916        return ret;
1917}
1918
1919static struct nfs4_stid *
1920find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
1921{
1922        struct nfs4_stid *s;
1923
1924        spin_lock(&cl->cl_lock);
1925        s = find_stateid_locked(cl, t);
1926        if (s != NULL) {
1927                if (typemask & s->sc_type)
1928                        atomic_inc(&s->sc_count);
1929                else
1930                        s = NULL;
1931        }
1932        spin_unlock(&cl->cl_lock);
1933        return s;
1934}
1935
1936static struct nfs4_client *create_client(struct xdr_netobj name,
1937                struct svc_rqst *rqstp, nfs4_verifier *verf)
1938{
1939        struct nfs4_client *clp;
1940        struct sockaddr *sa = svc_addr(rqstp);
1941        int ret;
1942        struct net *net = SVC_NET(rqstp);
1943
1944        clp = alloc_client(name);
1945        if (clp == NULL)
1946                return NULL;
1947
1948        ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
1949        if (ret) {
1950                free_client(clp);
1951                return NULL;
1952        }
1953        nfsd4_init_cb(&clp->cl_cb_null, clp, NULL, NFSPROC4_CLNT_CB_NULL);
1954        clp->cl_time = get_seconds();
1955        clear_bit(0, &clp->cl_cb_slot_busy);
1956        copy_verf(clp, verf);
1957        rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
1958        clp->cl_cb_session = NULL;
1959        clp->net = net;
1960        return clp;
1961}
1962
1963static void
1964add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
1965{
1966        struct rb_node **new = &(root->rb_node), *parent = NULL;
1967        struct nfs4_client *clp;
1968
1969        while (*new) {
1970                clp = rb_entry(*new, struct nfs4_client, cl_namenode);
1971                parent = *new;
1972
1973                if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
1974                        new = &((*new)->rb_left);
1975                else
1976                        new = &((*new)->rb_right);
1977        }
1978
1979        rb_link_node(&new_clp->cl_namenode, parent, new);
1980        rb_insert_color(&new_clp->cl_namenode, root);
1981}
1982
1983static struct nfs4_client *
1984find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
1985{
1986        int cmp;
1987        struct rb_node *node = root->rb_node;
1988        struct nfs4_client *clp;
1989
1990        while (node) {
1991                clp = rb_entry(node, struct nfs4_client, cl_namenode);
1992                cmp = compare_blob(&clp->cl_name, name);
1993                if (cmp > 0)
1994                        node = node->rb_left;
1995                else if (cmp < 0)
1996                        node = node->rb_right;
1997                else
1998                        return clp;
1999        }
2000        return NULL;
2001}
2002
2003static void
2004add_to_unconfirmed(struct nfs4_client *clp)
2005{
2006        unsigned int idhashval;
2007        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2008
2009        lockdep_assert_held(&nn->client_lock);
2010
2011        clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2012        add_clp_to_name_tree(clp, &nn->unconf_name_tree);
2013        idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2014        list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
2015        renew_client_locked(clp);
2016}
2017
2018static void
2019move_to_confirmed(struct nfs4_client *clp)
2020{
2021        unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2022        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2023
2024        lockdep_assert_held(&nn->client_lock);
2025
2026        dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
2027        list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
2028        rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2029        add_clp_to_name_tree(clp, &nn->conf_name_tree);
2030        set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2031        renew_client_locked(clp);
2032}
2033
2034static struct nfs4_client *
2035find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
2036{
2037        struct nfs4_client *clp;
2038        unsigned int idhashval = clientid_hashval(clid->cl_id);
2039
2040        list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
2041                if (same_clid(&clp->cl_clientid, clid)) {
2042                        if ((bool)clp->cl_minorversion != sessions)
2043                                return NULL;
2044                        renew_client_locked(clp);
2045                        return clp;
2046                }
2047        }
2048        return NULL;
2049}
2050
2051static struct nfs4_client *
2052find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2053{
2054        struct list_head *tbl = nn->conf_id_hashtbl;
2055
2056        lockdep_assert_held(&nn->client_lock);
2057        return find_client_in_id_table(tbl, clid, sessions);
2058}
2059
2060static struct nfs4_client *
2061find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2062{
2063        struct list_head *tbl = nn->unconf_id_hashtbl;
2064
2065        lockdep_assert_held(&nn->client_lock);
2066        return find_client_in_id_table(tbl, clid, sessions);
2067}
2068
2069static bool clp_used_exchangeid(struct nfs4_client *clp)
2070{
2071        return clp->cl_exchange_flags != 0;
2072} 
2073
2074static struct nfs4_client *
2075find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2076{
2077        lockdep_assert_held(&nn->client_lock);
2078        return find_clp_in_name_tree(name, &nn->conf_name_tree);
2079}
2080
2081static struct nfs4_client *
2082find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2083{
2084        lockdep_assert_held(&nn->client_lock);
2085        return find_clp_in_name_tree(name, &nn->unconf_name_tree);
2086}
2087
2088static void
2089gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
2090{
2091        struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
2092        struct sockaddr *sa = svc_addr(rqstp);
2093        u32 scopeid = rpc_get_scope_id(sa);
2094        unsigned short expected_family;
2095
2096        /* Currently, we only support tcp and tcp6 for the callback channel */
2097        if (se->se_callback_netid_len == 3 &&
2098            !memcmp(se->se_callback_netid_val, "tcp", 3))
2099                expected_family = AF_INET;
2100        else if (se->se_callback_netid_len == 4 &&
2101                 !memcmp(se->se_callback_netid_val, "tcp6", 4))
2102                expected_family = AF_INET6;
2103        else
2104                goto out_err;
2105
2106        conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
2107                                            se->se_callback_addr_len,
2108                                            (struct sockaddr *)&conn->cb_addr,
2109                                            sizeof(conn->cb_addr));
2110
2111        if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
2112                goto out_err;
2113
2114        if (conn->cb_addr.ss_family == AF_INET6)
2115                ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
2116
2117        conn->cb_prog = se->se_callback_prog;
2118        conn->cb_ident = se->se_callback_ident;
2119        memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
2120        return;
2121out_err:
2122        conn->cb_addr.ss_family = AF_UNSPEC;
2123        conn->cb_addrlen = 0;
2124        dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
2125                "will not receive delegations\n",
2126                clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
2127
2128        return;
2129}
2130
2131/*
2132 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
2133 */
2134static void
2135nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
2136{
2137        struct xdr_buf *buf = resp->xdr.buf;
2138        struct nfsd4_slot *slot = resp->cstate.slot;
2139        unsigned int base;
2140
2141        dprintk("--> %s slot %p\n", __func__, slot);
2142
2143        slot->sl_opcnt = resp->opcnt;
2144        slot->sl_status = resp->cstate.status;
2145
2146        slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
2147        if (nfsd4_not_cached(resp)) {
2148                slot->sl_datalen = 0;
2149                return;
2150        }
2151        base = resp->cstate.data_offset;
2152        slot->sl_datalen = buf->len - base;
2153        if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
2154                WARN("%s: sessions DRC could not cache compound\n", __func__);
2155        return;
2156}
2157
2158/*
2159 * Encode the replay sequence operation from the slot values.
2160 * If cachethis is FALSE encode the uncached rep error on the next
2161 * operation which sets resp->p and increments resp->opcnt for
2162 * nfs4svc_encode_compoundres.
2163 *
2164 */
2165static __be32
2166nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
2167                          struct nfsd4_compoundres *resp)
2168{
2169        struct nfsd4_op *op;
2170        struct nfsd4_slot *slot = resp->cstate.slot;
2171
2172        /* Encode the replayed sequence operation */
2173        op = &args->ops[resp->opcnt - 1];
2174        nfsd4_encode_operation(resp, op);
2175
2176        /* Return nfserr_retry_uncached_rep in next operation. */
2177        if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
2178                op = &args->ops[resp->opcnt++];
2179                op->status = nfserr_retry_uncached_rep;
2180                nfsd4_encode_operation(resp, op);
2181        }
2182        return op->status;
2183}
2184
2185/*
2186 * The sequence operation is not cached because we can use the slot and
2187 * session values.
2188 */
2189static __be32
2190nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
2191                         struct nfsd4_sequence *seq)
2192{
2193        struct nfsd4_slot *slot = resp->cstate.slot;
2194        struct xdr_stream *xdr = &resp->xdr;
2195        __be32 *p;
2196        __be32 status;
2197
2198        dprintk("--> %s slot %p\n", __func__, slot);
2199
2200        status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
2201        if (status)
2202                return status;
2203
2204        p = xdr_reserve_space(xdr, slot->sl_datalen);
2205        if (!p) {
2206                WARN_ON_ONCE(1);
2207                return nfserr_serverfault;
2208        }
2209        xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
2210        xdr_commit_encode(xdr);
2211
2212        resp->opcnt = slot->sl_opcnt;
2213        return slot->sl_status;
2214}
2215
2216/*
2217 * Set the exchange_id flags returned by the server.
2218 */
2219static void
2220nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
2221{
2222#ifdef CONFIG_NFSD_PNFS
2223        new->cl_exchange_flags |= EXCHGID4_FLAG_USE_PNFS_MDS;
2224#else
2225        new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
2226#endif
2227
2228        /* Referrals are supported, Migration is not. */
2229        new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
2230
2231        /* set the wire flags to return to client. */
2232        clid->flags = new->cl_exchange_flags;
2233}
2234
2235static bool client_has_state(struct nfs4_client *clp)
2236{
2237        /*
2238         * Note clp->cl_openowners check isn't quite right: there's no
2239         * need to count owners without stateid's.
2240         *
2241         * Also note we should probably be using this in 4.0 case too.
2242         */
2243        return !list_empty(&clp->cl_openowners)
2244                || !list_empty(&clp->cl_delegations)
2245                || !list_empty(&clp->cl_sessions);
2246}
2247
2248__be32
2249nfsd4_exchange_id(struct svc_rqst *rqstp,
2250                  struct nfsd4_compound_state *cstate,
2251                  struct nfsd4_exchange_id *exid)
2252{
2253        struct nfs4_client *conf, *new;
2254        struct nfs4_client *unconf = NULL;
2255        __be32 status;
2256        char                    addr_str[INET6_ADDRSTRLEN];
2257        nfs4_verifier           verf = exid->verifier;
2258        struct sockaddr         *sa = svc_addr(rqstp);
2259        bool    update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
2260        struct nfsd_net         *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2261
2262        rpc_ntop(sa, addr_str, sizeof(addr_str));
2263        dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
2264                "ip_addr=%s flags %x, spa_how %d\n",
2265                __func__, rqstp, exid, exid->clname.len, exid->clname.data,
2266                addr_str, exid->flags, exid->spa_how);
2267
2268        if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
2269                return nfserr_inval;
2270
2271        switch (exid->spa_how) {
2272        case SP4_MACH_CRED:
2273                if (!svc_rqst_integrity_protected(rqstp))
2274                        return nfserr_inval;
2275        case SP4_NONE:
2276                break;
2277        default:                                /* checked by xdr code */
2278                WARN_ON_ONCE(1);
2279        case SP4_SSV:
2280                return nfserr_encr_alg_unsupp;
2281        }
2282
2283        new = create_client(exid->clname, rqstp, &verf);
2284        if (new == NULL)
2285                return nfserr_jukebox;
2286
2287        /* Cases below refer to rfc 5661 section 18.35.4: */
2288        spin_lock(&nn->client_lock);
2289        conf = find_confirmed_client_by_name(&exid->clname, nn);
2290        if (conf) {
2291                bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
2292                bool verfs_match = same_verf(&verf, &conf->cl_verifier);
2293
2294                if (update) {
2295                        if (!clp_used_exchangeid(conf)) { /* buggy client */
2296                                status = nfserr_inval;
2297                                goto out;
2298                        }
2299                        if (!mach_creds_match(conf, rqstp)) {
2300                                status = nfserr_wrong_cred;
2301                                goto out;
2302                        }
2303                        if (!creds_match) { /* case 9 */
2304                                status = nfserr_perm;
2305                                goto out;
2306                        }
2307                        if (!verfs_match) { /* case 8 */
2308                                status = nfserr_not_same;
2309                                goto out;
2310                        }
2311                        /* case 6 */
2312                        exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
2313                        goto out_copy;
2314                }
2315                if (!creds_match) { /* case 3 */
2316                        if (client_has_state(conf)) {
2317                                status = nfserr_clid_inuse;
2318                                goto out;
2319                        }
2320                        goto out_new;
2321                }
2322                if (verfs_match) { /* case 2 */
2323                        conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
2324                        goto out_copy;
2325                }
2326                /* case 5, client reboot */
2327                conf = NULL;
2328                goto out_new;
2329        }
2330
2331        if (update) { /* case 7 */
2332                status = nfserr_noent;
2333                goto out;
2334        }
2335
2336        unconf  = find_unconfirmed_client_by_name(&exid->clname, nn);
2337        if (unconf) /* case 4, possible retry or client restart */
2338                unhash_client_locked(unconf);
2339
2340        /* case 1 (normal case) */
2341out_new:
2342        if (conf) {
2343                status = mark_client_expired_locked(conf);
2344                if (status)
2345                        goto out;
2346        }
2347        new->cl_minorversion = cstate->minorversion;
2348        new->cl_mach_cred = (exid->spa_how == SP4_MACH_CRED);
2349
2350        gen_clid(new, nn);
2351        add_to_unconfirmed(new);
2352        swap(new, conf);
2353out_copy:
2354        exid->clientid.cl_boot = conf->cl_clientid.cl_boot;
2355        exid->clientid.cl_id = conf->cl_clientid.cl_id;
2356
2357        exid->seqid = conf->cl_cs_slot.sl_seqid + 1;
2358        nfsd4_set_ex_flags(conf, exid);
2359
2360        dprintk("nfsd4_exchange_id seqid %d flags %x\n",
2361                conf->cl_cs_slot.sl_seqid, conf->cl_exchange_flags);
2362        status = nfs_ok;
2363
2364out:
2365        spin_unlock(&nn->client_lock);
2366        if (new)
2367                expire_client(new);
2368        if (unconf)
2369                expire_client(unconf);
2370        return status;
2371}
2372
2373static __be32
2374check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
2375{
2376        dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
2377                slot_seqid);
2378
2379        /* The slot is in use, and no response has been sent. */
2380        if (slot_inuse) {
2381                if (seqid == slot_seqid)
2382                        return nfserr_jukebox;
2383                else
2384                        return nfserr_seq_misordered;
2385        }
2386        /* Note unsigned 32-bit arithmetic handles wraparound: */
2387        if (likely(seqid == slot_seqid + 1))
2388                return nfs_ok;
2389        if (seqid == slot_seqid)
2390                return nfserr_replay_cache;
2391        return nfserr_seq_misordered;
2392}
2393
2394/*
2395 * Cache the create session result into the create session single DRC
2396 * slot cache by saving the xdr structure. sl_seqid has been set.
2397 * Do this for solo or embedded create session operations.
2398 */
2399static void
2400nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
2401                           struct nfsd4_clid_slot *slot, __be32 nfserr)
2402{
2403        slot->sl_status = nfserr;
2404        memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
2405}
2406
2407static __be32
2408nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
2409                            struct nfsd4_clid_slot *slot)
2410{
2411        memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
2412        return slot->sl_status;
2413}
2414
2415#define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
2416                        2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
2417                        1 +     /* MIN tag is length with zero, only length */ \
2418                        3 +     /* version, opcount, opcode */ \
2419                        XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2420                                /* seqid, slotID, slotID, cache */ \
2421                        4 ) * sizeof(__be32))
2422
2423#define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
2424                        2 +     /* verifier: AUTH_NULL, length 0 */\
2425                        1 +     /* status */ \
2426                        1 +     /* MIN tag is length with zero, only length */ \
2427                        3 +     /* opcount, opcode, opstatus*/ \
2428                        XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2429                                /* seqid, slotID, slotID, slotID, status */ \
2430                        5 ) * sizeof(__be32))
2431
2432static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
2433{
2434        u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
2435
2436        if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
2437                return nfserr_toosmall;
2438        if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
2439                return nfserr_toosmall;
2440        ca->headerpadsz = 0;
2441        ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
2442        ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
2443        ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
2444        ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
2445                        NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
2446        ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
2447        /*
2448         * Note decreasing slot size below client's request may make it
2449         * difficult for client to function correctly, whereas
2450         * decreasing the number of slots will (just?) affect
2451         * performance.  When short on memory we therefore prefer to
2452         * decrease number of slots instead of their size.  Clients that
2453         * request larger slots than they need will get poor results:
2454         */
2455        ca->maxreqs = nfsd4_get_drc_mem(ca);
2456        if (!ca->maxreqs)
2457                return nfserr_jukebox;
2458
2459        return nfs_ok;
2460}
2461
2462#define NFSD_CB_MAX_REQ_SZ      ((NFS4_enc_cb_recall_sz + \
2463                                 RPC_MAX_HEADER_WITH_AUTH) * sizeof(__be32))
2464#define NFSD_CB_MAX_RESP_SZ     ((NFS4_dec_cb_recall_sz + \
2465                                 RPC_MAX_REPHEADER_WITH_AUTH) * sizeof(__be32))
2466
2467static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
2468{
2469        ca->headerpadsz = 0;
2470
2471        /*
2472         * These RPC_MAX_HEADER macros are overkill, especially since we
2473         * don't even do gss on the backchannel yet.  But this is still
2474         * less than 1k.  Tighten up this estimate in the unlikely event
2475         * it turns out to be a problem for some client:
2476         */
2477        if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
2478                return nfserr_toosmall;
2479        if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
2480                return nfserr_toosmall;
2481        ca->maxresp_cached = 0;
2482        if (ca->maxops < 2)
2483                return nfserr_toosmall;
2484
2485        return nfs_ok;
2486}
2487
2488static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
2489{
2490        switch (cbs->flavor) {
2491        case RPC_AUTH_NULL:
2492        case RPC_AUTH_UNIX:
2493                return nfs_ok;
2494        default:
2495                /*
2496                 * GSS case: the spec doesn't allow us to return this
2497                 * error.  But it also doesn't allow us not to support
2498                 * GSS.
2499                 * I'd rather this fail hard than return some error the
2500                 * client might think it can already handle:
2501                 */
2502                return nfserr_encr_alg_unsupp;
2503        }
2504}
2505
2506__be32
2507nfsd4_create_session(struct svc_rqst *rqstp,
2508                     struct nfsd4_compound_state *cstate,
2509                     struct nfsd4_create_session *cr_ses)
2510{
2511        struct sockaddr *sa = svc_addr(rqstp);
2512        struct nfs4_client *conf, *unconf;
2513        struct nfs4_client *old = NULL;
2514        struct nfsd4_session *new;
2515        struct nfsd4_conn *conn;
2516        struct nfsd4_clid_slot *cs_slot = NULL;
2517        __be32 status = 0;
2518        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2519
2520        if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
2521                return nfserr_inval;
2522        status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
2523        if (status)
2524                return status;
2525        status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
2526        if (status)
2527                return status;
2528        status = check_backchannel_attrs(&cr_ses->back_channel);
2529        if (status)
2530                goto out_release_drc_mem;
2531        status = nfserr_jukebox;
2532        new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
2533        if (!new)
2534                goto out_release_drc_mem;
2535        conn = alloc_conn_from_crses(rqstp, cr_ses);
2536        if (!conn)
2537                goto out_free_session;
2538
2539        spin_lock(&nn->client_lock);
2540        unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
2541        conf = find_confirmed_client(&cr_ses->clientid, true, nn);
2542        WARN_ON_ONCE(conf && unconf);
2543
2544        if (conf) {
2545                status = nfserr_wrong_cred;
2546                if (!mach_creds_match(conf, rqstp))
2547                        goto out_free_conn;
2548                cs_slot = &conf->cl_cs_slot;
2549                status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2550                if (status == nfserr_replay_cache) {
2551                        status = nfsd4_replay_create_session(cr_ses, cs_slot);
2552                        goto out_free_conn;
2553                } else if (cr_ses->seqid != cs_slot->sl_seqid + 1) {
2554                        status = nfserr_seq_misordered;
2555                        goto out_free_conn;
2556                }
2557        } else if (unconf) {
2558                if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
2559                    !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
2560                        status = nfserr_clid_inuse;
2561                        goto out_free_conn;
2562                }
2563                status = nfserr_wrong_cred;
2564                if (!mach_creds_match(unconf, rqstp))
2565                        goto out_free_conn;
2566                cs_slot = &unconf->cl_cs_slot;
2567                status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2568                if (status) {
2569                        /* an unconfirmed replay returns misordered */
2570                        status = nfserr_seq_misordered;
2571                        goto out_free_conn;
2572                }
2573                old = find_confirmed_client_by_name(&unconf->cl_name, nn);
2574                if (old) {
2575                        status = mark_client_expired_locked(old);
2576                        if (status) {
2577                                old = NULL;
2578                                goto out_free_conn;
2579                        }
2580                }
2581                move_to_confirmed(unconf);
2582                conf = unconf;
2583        } else {
2584                status = nfserr_stale_clientid;
2585                goto out_free_conn;
2586        }
2587        status = nfs_ok;
2588        /*
2589         * We do not support RDMA or persistent sessions
2590         */
2591        cr_ses->flags &= ~SESSION4_PERSIST;
2592        cr_ses->flags &= ~SESSION4_RDMA;
2593
2594        init_session(rqstp, new, conf, cr_ses);
2595        nfsd4_get_session_locked(new);
2596
2597        memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
2598               NFS4_MAX_SESSIONID_LEN);
2599        cs_slot->sl_seqid++;
2600        cr_ses->seqid = cs_slot->sl_seqid;
2601
2602        /* cache solo and embedded create sessions under the client_lock */
2603        nfsd4_cache_create_session(cr_ses, cs_slot, status);
2604        spin_unlock(&nn->client_lock);
2605        /* init connection and backchannel */
2606        nfsd4_init_conn(rqstp, conn, new);
2607        nfsd4_put_session(new);
2608        if (old)
2609                expire_client(old);
2610        return status;
2611out_free_conn:
2612        spin_unlock(&nn->client_lock);
2613        free_conn(conn);
2614        if (old)
2615                expire_client(old);
2616out_free_session:
2617        __free_session(new);
2618out_release_drc_mem:
2619        nfsd4_put_drc_mem(&cr_ses->fore_channel);
2620        return status;
2621}
2622
2623static __be32 nfsd4_map_bcts_dir(u32 *dir)
2624{
2625        switch (*dir) {
2626        case NFS4_CDFC4_FORE:
2627        case NFS4_CDFC4_BACK:
2628                return nfs_ok;
2629        case NFS4_CDFC4_FORE_OR_BOTH:
2630        case NFS4_CDFC4_BACK_OR_BOTH:
2631                *dir = NFS4_CDFC4_BOTH;
2632                return nfs_ok;
2633        };
2634        return nfserr_inval;
2635}
2636
2637__be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
2638{
2639        struct nfsd4_session *session = cstate->session;
2640        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2641        __be32 status;
2642
2643        status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
2644        if (status)
2645                return status;
2646        spin_lock(&nn->client_lock);
2647        session->se_cb_prog = bc->bc_cb_program;
2648        session->se_cb_sec = bc->bc_cb_sec;
2649        spin_unlock(&nn->client_lock);
2650
2651        nfsd4_probe_callback(session->se_client);
2652
2653        return nfs_ok;
2654}
2655
2656__be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
2657                     struct nfsd4_compound_state *cstate,
2658                     struct nfsd4_bind_conn_to_session *bcts)
2659{
2660        __be32 status;
2661        struct nfsd4_conn *conn;
2662        struct nfsd4_session *session;
2663        struct net *net = SVC_NET(rqstp);
2664        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2665
2666        if (!nfsd4_last_compound_op(rqstp))
2667                return nfserr_not_only_op;
2668        spin_lock(&nn->client_lock);
2669        session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
2670        spin_unlock(&nn->client_lock);
2671        if (!session)
2672                goto out_no_session;
2673        status = nfserr_wrong_cred;
2674        if (!mach_creds_match(session->se_client, rqstp))
2675                goto out;
2676        status = nfsd4_map_bcts_dir(&bcts->dir);
2677        if (status)
2678                goto out;
2679        conn = alloc_conn(rqstp, bcts->dir);
2680        status = nfserr_jukebox;
2681        if (!conn)
2682                goto out;
2683        nfsd4_init_conn(rqstp, conn, session);
2684        status = nfs_ok;
2685out:
2686        nfsd4_put_session(session);
2687out_no_session:
2688        return status;
2689}
2690
2691static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
2692{
2693        if (!session)
2694                return 0;
2695        return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
2696}
2697
2698__be32
2699nfsd4_destroy_session(struct svc_rqst *r,
2700                      struct nfsd4_compound_state *cstate,
2701                      struct nfsd4_destroy_session *sessionid)
2702{
2703        struct nfsd4_session *ses;
2704        __be32 status;
2705        int ref_held_by_me = 0;
2706        struct net *net = SVC_NET(r);
2707        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2708
2709        status = nfserr_not_only_op;
2710        if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
2711                if (!nfsd4_last_compound_op(r))
2712                        goto out;
2713                ref_held_by_me++;
2714        }
2715        dump_sessionid(__func__, &sessionid->sessionid);
2716        spin_lock(&nn->client_lock);
2717        ses = find_in_sessionid_hashtbl(&sessionid->sessionid, net, &status);
2718        if (!ses)
2719                goto out_client_lock;
2720        status = nfserr_wrong_cred;
2721        if (!mach_creds_match(ses->se_client, r))
2722                goto out_put_session;
2723        status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
2724        if (status)
2725                goto out_put_session;
2726        unhash_session(ses);
2727        spin_unlock(&nn->client_lock);
2728
2729        nfsd4_probe_callback_sync(ses->se_client);
2730
2731        spin_lock(&nn->client_lock);
2732        status = nfs_ok;
2733out_put_session:
2734        nfsd4_put_session_locked(ses);
2735out_client_lock:
2736        spin_unlock(&nn->client_lock);
2737out:
2738        return status;
2739}
2740
2741static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2742{
2743        struct nfsd4_conn *c;
2744
2745        list_for_each_entry(c, &s->se_conns, cn_persession) {
2746                if (c->cn_xprt == xpt) {
2747                        return c;
2748                }
2749        }
2750        return NULL;
2751}
2752
2753static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2754{
2755        struct nfs4_client *clp = ses->se_client;
2756        struct nfsd4_conn *c;
2757        __be32 status = nfs_ok;
2758        int ret;
2759
2760        spin_lock(&clp->cl_lock);
2761        c = __nfsd4_find_conn(new->cn_xprt, ses);
2762        if (c)
2763                goto out_free;
2764        status = nfserr_conn_not_bound_to_session;
2765        if (clp->cl_mach_cred)
2766                goto out_free;
2767        __nfsd4_hash_conn(new, ses);
2768        spin_unlock(&clp->cl_lock);
2769        ret = nfsd4_register_conn(new);
2770        if (ret)
2771                /* oops; xprt is already down: */
2772                nfsd4_conn_lost(&new->cn_xpt_user);
2773        return nfs_ok;
2774out_free:
2775        spin_unlock(&clp->cl_lock);
2776        free_conn(new);
2777        return status;
2778}
2779
2780static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2781{
2782        struct nfsd4_compoundargs *args = rqstp->rq_argp;
2783
2784        return args->opcnt > session->se_fchannel.maxops;
2785}
2786
2787static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2788                                  struct nfsd4_session *session)
2789{
2790        struct xdr_buf *xb = &rqstp->rq_arg;
2791
2792        return xb->len > session->se_fchannel.maxreq_sz;
2793}
2794
2795__be32
2796nfsd4_sequence(struct svc_rqst *rqstp,
2797               struct nfsd4_compound_state *cstate,
2798               struct nfsd4_sequence *seq)
2799{
2800        struct nfsd4_compoundres *resp = rqstp->rq_resp;
2801        struct xdr_stream *xdr = &resp->xdr;
2802        struct nfsd4_session *session;
2803        struct nfs4_client *clp;
2804        struct nfsd4_slot *slot;
2805        struct nfsd4_conn *conn;
2806        __be32 status;
2807        int buflen;
2808        struct net *net = SVC_NET(rqstp);
2809        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2810
2811        if (resp->opcnt != 1)
2812                return nfserr_sequence_pos;
2813
2814        /*
2815         * Will be either used or freed by nfsd4_sequence_check_conn
2816         * below.
2817         */
2818        conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2819        if (!conn)
2820                return nfserr_jukebox;
2821
2822        spin_lock(&nn->client_lock);
2823        session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
2824        if (!session)
2825                goto out_no_session;
2826        clp = session->se_client;
2827
2828        status = nfserr_too_many_ops;
2829        if (nfsd4_session_too_many_ops(rqstp, session))
2830                goto out_put_session;
2831
2832        status = nfserr_req_too_big;
2833        if (nfsd4_request_too_big(rqstp, session))
2834                goto out_put_session;
2835
2836        status = nfserr_badslot;
2837        if (seq->slotid >= session->se_fchannel.maxreqs)
2838                goto out_put_session;
2839
2840        slot = session->se_slots[seq->slotid];
2841        dprintk("%s: slotid %d\n", __func__, seq->slotid);
2842
2843        /* We do not negotiate the number of slots yet, so set the
2844         * maxslots to the session maxreqs which is used to encode
2845         * sr_highest_slotid and the sr_target_slot id to maxslots */
2846        seq->maxslots = session->se_fchannel.maxreqs;
2847
2848        status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2849                                        slot->sl_flags & NFSD4_SLOT_INUSE);
2850        if (status == nfserr_replay_cache) {
2851                status = nfserr_seq_misordered;
2852                if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2853                        goto out_put_session;
2854                cstate->slot = slot;
2855                cstate->session = session;
2856                cstate->clp = clp;
2857                /* Return the cached reply status and set cstate->status
2858                 * for nfsd4_proc_compound processing */
2859                status = nfsd4_replay_cache_entry(resp, seq);
2860                cstate->status = nfserr_replay_cache;
2861                goto out;
2862        }
2863        if (status)
2864                goto out_put_session;
2865
2866        status = nfsd4_sequence_check_conn(conn, session);
2867        conn = NULL;
2868        if (status)
2869                goto out_put_session;
2870
2871        buflen = (seq->cachethis) ?
2872                        session->se_fchannel.maxresp_cached :
2873                        session->se_fchannel.maxresp_sz;
2874        status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
2875                                    nfserr_rep_too_big;
2876        if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
2877                goto out_put_session;
2878        svc_reserve(rqstp, buflen);
2879
2880        status = nfs_ok;
2881        /* Success! bump slot seqid */
2882        slot->sl_seqid = seq->seqid;
2883        slot->sl_flags |= NFSD4_SLOT_INUSE;
2884        if (seq->cachethis)
2885                slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
2886        else
2887                slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
2888
2889        cstate->slot = slot;
2890        cstate->session = session;
2891        cstate->clp = clp;
2892
2893out:
2894        switch (clp->cl_cb_state) {
2895        case NFSD4_CB_DOWN:
2896                seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
2897                break;
2898        case NFSD4_CB_FAULT:
2899                seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
2900                break;
2901        default:
2902                seq->status_flags = 0;
2903        }
2904        if (!list_empty(&clp->cl_revoked))
2905                seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
2906out_no_session:
2907        if (conn)
2908                free_conn(conn);
2909        spin_unlock(&nn->client_lock);
2910        return status;
2911out_put_session:
2912        nfsd4_put_session_locked(session);
2913        goto out_no_session;
2914}
2915
2916void
2917nfsd4_sequence_done(struct nfsd4_compoundres *resp)
2918{
2919        struct nfsd4_compound_state *cs = &resp->cstate;
2920
2921        if (nfsd4_has_session(cs)) {
2922                if (cs->status != nfserr_replay_cache) {
2923                        nfsd4_store_cache_entry(resp);
2924                        cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
2925                }
2926                /* Drop session reference that was taken in nfsd4_sequence() */
2927                nfsd4_put_session(cs->session);
2928        } else if (cs->clp)
2929                put_client_renew(cs->clp);
2930}
2931
2932__be32
2933nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
2934{
2935        struct nfs4_client *conf, *unconf;
2936        struct nfs4_client *clp = NULL;
2937        __be32 status = 0;
2938        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2939
2940        spin_lock(&nn->client_lock);
2941        unconf = find_unconfirmed_client(&dc->clientid, true, nn);
2942        conf = find_confirmed_client(&dc->clientid, true, nn);
2943        WARN_ON_ONCE(conf && unconf);
2944
2945        if (conf) {
2946                if (client_has_state(conf)) {
2947                        status = nfserr_clientid_busy;
2948                        goto out;
2949                }
2950                status = mark_client_expired_locked(conf);
2951                if (status)
2952                        goto out;
2953                clp = conf;
2954        } else if (unconf)
2955                clp = unconf;
2956        else {
2957                status = nfserr_stale_clientid;
2958                goto out;
2959        }
2960        if (!mach_creds_match(clp, rqstp)) {
2961                clp = NULL;
2962                status = nfserr_wrong_cred;
2963                goto out;
2964        }
2965        unhash_client_locked(clp);
2966out:
2967        spin_unlock(&nn->client_lock);
2968        if (clp)
2969                expire_client(clp);
2970        return status;
2971}
2972
2973__be32
2974nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
2975{
2976        __be32 status = 0;
2977
2978        if (rc->rca_one_fs) {
2979                if (!cstate->current_fh.fh_dentry)
2980                        return nfserr_nofilehandle;
2981                /*
2982                 * We don't take advantage of the rca_one_fs case.
2983                 * That's OK, it's optional, we can safely ignore it.
2984                 */
2985                 return nfs_ok;
2986        }
2987
2988        status = nfserr_complete_already;
2989        if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
2990                             &cstate->session->se_client->cl_flags))
2991                goto out;
2992
2993        status = nfserr_stale_clientid;
2994        if (is_client_expired(cstate->session->se_client))
2995                /*
2996                 * The following error isn't really legal.
2997                 * But we only get here if the client just explicitly
2998                 * destroyed the client.  Surely it no longer cares what
2999                 * error it gets back on an operation for the dead
3000                 * client.
3001                 */
3002                goto out;
3003
3004        status = nfs_ok;
3005        nfsd4_client_record_create(cstate->session->se_client);
3006out:
3007        return status;
3008}
3009
3010__be32
3011nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3012                  struct nfsd4_setclientid *setclid)
3013{
3014        struct xdr_netobj       clname = setclid->se_name;
3015        nfs4_verifier           clverifier = setclid->se_verf;
3016        struct nfs4_client      *conf, *new;
3017        struct nfs4_client      *unconf = NULL;
3018        __be32                  status;
3019        struct nfsd_net         *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3020
3021        new = create_client(clname, rqstp, &clverifier);
3022        if (new == NULL)
3023                return nfserr_jukebox;
3024        /* Cases below refer to rfc 3530 section 14.2.33: */
3025        spin_lock(&nn->client_lock);
3026        conf = find_confirmed_client_by_name(&clname, nn);
3027        if (conf) {
3028                /* case 0: */
3029                status = nfserr_clid_inuse;
3030                if (clp_used_exchangeid(conf))
3031                        goto out;
3032                if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
3033                        char addr_str[INET6_ADDRSTRLEN];
3034                        rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
3035                                 sizeof(addr_str));
3036                        dprintk("NFSD: setclientid: string in use by client "
3037                                "at %s\n", addr_str);
3038                        goto out;
3039                }
3040        }
3041        unconf = find_unconfirmed_client_by_name(&clname, nn);
3042        if (unconf)
3043                unhash_client_locked(unconf);
3044        if (conf && same_verf(&conf->cl_verifier, &clverifier))
3045                /* case 1: probable callback update */
3046                copy_clid(new, conf);
3047        else /* case 4 (new client) or cases 2, 3 (client reboot): */
3048                gen_clid(new, nn);
3049        new->cl_minorversion = 0;
3050        gen_callback(new, setclid, rqstp);
3051        add_to_unconfirmed(new);
3052        setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
3053        setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
3054        memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
3055        new = NULL;
3056        status = nfs_ok;
3057out:
3058        spin_unlock(&nn->client_lock);
3059        if (new)
3060                free_client(new);
3061        if (unconf)
3062                expire_client(unconf);
3063        return status;
3064}
3065
3066
3067__be32
3068nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
3069                         struct nfsd4_compound_state *cstate,
3070                         struct nfsd4_setclientid_confirm *setclientid_confirm)
3071{
3072        struct nfs4_client *conf, *unconf;
3073        struct nfs4_client *old = NULL;
3074        nfs4_verifier confirm = setclientid_confirm->sc_confirm; 
3075        clientid_t * clid = &setclientid_confirm->sc_clientid;
3076        __be32 status;
3077        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3078
3079        if (STALE_CLIENTID(clid, nn))
3080                return nfserr_stale_clientid;
3081
3082        spin_lock(&nn->client_lock);
3083        conf = find_confirmed_client(clid, false, nn);
3084        unconf = find_unconfirmed_client(clid, false, nn);
3085        /*
3086         * We try hard to give out unique clientid's, so if we get an
3087         * attempt to confirm the same clientid with a different cred,
3088         * there's a bug somewhere.  Let's charitably assume it's our
3089         * bug.
3090         */
3091        status = nfserr_serverfault;
3092        if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
3093                goto out;
3094        if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
3095                goto out;
3096        /* cases below refer to rfc 3530 section 14.2.34: */
3097        if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
3098                if (conf && !unconf) /* case 2: probable retransmit */
3099                        status = nfs_ok;
3100                else /* case 4: client hasn't noticed we rebooted yet? */
3101                        status = nfserr_stale_clientid;
3102                goto out;
3103        }
3104        status = nfs_ok;
3105        if (conf) { /* case 1: callback update */
3106                old = unconf;
3107                unhash_client_locked(old);
3108                nfsd4_change_callback(conf, &unconf->cl_cb_conn);
3109        } else { /* case 3: normal case; new or rebooted client */
3110                old = find_confirmed_client_by_name(&unconf->cl_name, nn);
3111                if (old) {
3112                        status = mark_client_expired_locked(old);
3113                        if (status) {
3114                                old = NULL;
3115                                goto out;
3116                        }
3117                }
3118                move_to_confirmed(unconf);
3119                conf = unconf;
3120        }
3121        get_client_locked(conf);
3122        spin_unlock(&nn->client_lock);
3123        nfsd4_probe_callback(conf);
3124        spin_lock(&nn->client_lock);
3125        put_client_renew_locked(conf);
3126out:
3127        spin_unlock(&nn->client_lock);
3128        if (old)
3129                expire_client(old);
3130        return status;
3131}
3132
3133static struct nfs4_file *nfsd4_alloc_file(void)
3134{
3135        return kmem_cache_alloc(file_slab, GFP_KERNEL);
3136}
3137
3138/* OPEN Share state helper functions */
3139static void nfsd4_init_file(struct knfsd_fh *fh, unsigned int hashval,
3140                                struct nfs4_file *fp)
3141{
3142        lockdep_assert_held(&state_lock);
3143
3144        atomic_set(&fp->fi_ref, 1);
3145        spin_lock_init(&fp->fi_lock);
3146        INIT_LIST_HEAD(&fp->fi_stateids);
3147        INIT_LIST_HEAD(&fp->fi_delegations);
3148        INIT_LIST_HEAD(&fp->fi_clnt_odstate);
3149        fh_copy_shallow(&fp->fi_fhandle, fh);
3150        fp->fi_deleg_file = NULL;
3151        fp->fi_had_conflict = false;
3152        fp->fi_share_deny = 0;
3153        memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
3154        memset(fp->fi_access, 0, sizeof(fp->fi_access));
3155#ifdef CONFIG_NFSD_PNFS
3156        INIT_LIST_HEAD(&fp->fi_lo_states);
3157        atomic_set(&fp->fi_lo_recalls, 0);
3158#endif
3159        hlist_add_head_rcu(&fp->fi_hash, &file_hashtbl[hashval]);
3160}
3161
3162void
3163nfsd4_free_slabs(void)
3164{
3165        kmem_cache_destroy(odstate_slab);
3166        kmem_cache_destroy(openowner_slab);
3167        kmem_cache_destroy(lockowner_slab);
3168        kmem_cache_destroy(file_slab);
3169        kmem_cache_destroy(stateid_slab);
3170        kmem_cache_destroy(deleg_slab);
3171}
3172
3173int
3174nfsd4_init_slabs(void)
3175{
3176        openowner_slab = kmem_cache_create("nfsd4_openowners",
3177                        sizeof(struct nfs4_openowner), 0, 0, NULL);
3178        if (openowner_slab == NULL)
3179                goto out;
3180        lockowner_slab = kmem_cache_create("nfsd4_lockowners",
3181                        sizeof(struct nfs4_lockowner), 0, 0, NULL);
3182        if (lockowner_slab == NULL)
3183                goto out_free_openowner_slab;
3184        file_slab = kmem_cache_create("nfsd4_files",
3185                        sizeof(struct nfs4_file), 0, 0, NULL);
3186        if (file_slab == NULL)
3187                goto out_free_lockowner_slab;
3188        stateid_slab = kmem_cache_create("nfsd4_stateids",
3189                        sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
3190        if (stateid_slab == NULL)
3191                goto out_free_file_slab;
3192        deleg_slab = kmem_cache_create("nfsd4_delegations",
3193                        sizeof(struct nfs4_delegation), 0, 0, NULL);
3194        if (deleg_slab == NULL)
3195                goto out_free_stateid_slab;
3196        odstate_slab = kmem_cache_create("nfsd4_odstate",
3197                        sizeof(struct nfs4_clnt_odstate), 0, 0, NULL);
3198        if (odstate_slab == NULL)
3199                goto out_free_deleg_slab;
3200        return 0;
3201
3202out_free_deleg_slab:
3203        kmem_cache_destroy(deleg_slab);
3204out_free_stateid_slab:
3205        kmem_cache_destroy(stateid_slab);
3206out_free_file_slab:
3207        kmem_cache_destroy(file_slab);
3208out_free_lockowner_slab:
3209        kmem_cache_destroy(lockowner_slab);
3210out_free_openowner_slab:
3211        kmem_cache_destroy(openowner_slab);
3212out:
3213        dprintk("nfsd4: out of memory while initializing nfsv4\n");
3214        return -ENOMEM;
3215}
3216
3217static void init_nfs4_replay(struct nfs4_replay *rp)
3218{
3219        rp->rp_status = nfserr_serverfault;
3220        rp->rp_buflen = 0;
3221        rp->rp_buf = rp->rp_ibuf;
3222        mutex_init(&rp->rp_mutex);
3223}
3224
3225static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
3226                struct nfs4_stateowner *so)
3227{
3228        if (!nfsd4_has_session(cstate)) {
3229                mutex_lock(&so->so_replay.rp_mutex);
3230                cstate->replay_owner = nfs4_get_stateowner(so);
3231        }
3232}
3233
3234void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
3235{
3236        struct nfs4_stateowner *so = cstate->replay_owner;
3237
3238        if (so != NULL) {
3239                cstate->replay_owner = NULL;
3240                mutex_unlock(&so->so_replay.rp_mutex);
3241                nfs4_put_stateowner(so);
3242        }
3243}
3244
3245static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
3246{
3247        struct nfs4_stateowner *sop;
3248
3249        sop = kmem_cache_alloc(slab, GFP_KERNEL);
3250        if (!sop)
3251                return NULL;
3252
3253        sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
3254        if (!sop->so_owner.data) {
3255                kmem_cache_free(slab, sop);
3256                return NULL;
3257        }
3258        sop->so_owner.len = owner->len;
3259
3260        INIT_LIST_HEAD(&sop->so_stateids);
3261        sop->so_client = clp;
3262        init_nfs4_replay(&sop->so_replay);
3263        atomic_set(&sop->so_count, 1);
3264        return sop;
3265}
3266
3267static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
3268{
3269        lockdep_assert_held(&clp->cl_lock);
3270
3271        list_add(&oo->oo_owner.so_strhash,
3272                 &clp->cl_ownerstr_hashtbl[strhashval]);
3273        list_add(&oo->oo_perclient, &clp->cl_openowners);
3274}
3275
3276static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
3277{
3278        unhash_openowner_locked(openowner(so));
3279}
3280
3281static void nfs4_free_openowner(struct nfs4_stateowner *so)
3282{
3283        struct nfs4_openowner *oo = openowner(so);
3284
3285        kmem_cache_free(openowner_slab, oo);
3286}
3287
3288static const struct nfs4_stateowner_operations openowner_ops = {
3289        .so_unhash =    nfs4_unhash_openowner,
3290        .so_free =      nfs4_free_openowner,
3291};
3292
3293static struct nfs4_openowner *
3294alloc_init_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
3295                           struct nfsd4_compound_state *cstate)
3296{
3297        struct nfs4_client *clp = cstate->clp;
3298        struct nfs4_openowner *oo, *ret;
3299
3300        oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
3301        if (!oo)
3302                return NULL;
3303        oo->oo_owner.so_ops = &openowner_ops;
3304        oo->oo_owner.so_is_open_owner = 1;
3305        oo->oo_owner.so_seqid = open->op_seqid;
3306        oo->oo_flags = 0;
3307        if (nfsd4_has_session(cstate))
3308                oo->oo_flags |= NFS4_OO_CONFIRMED;
3309        oo->oo_time = 0;
3310        oo->oo_last_closed_stid = NULL;
3311        INIT_LIST_HEAD(&oo->oo_close_lru);
3312        spin_lock(&clp->cl_lock);
3313        ret = find_openstateowner_str_locked(strhashval, open, clp);
3314        if (ret == NULL) {
3315                hash_openowner(oo, clp, strhashval);
3316                ret = oo;
3317        } else
3318                nfs4_free_openowner(&oo->oo_owner);
3319        spin_unlock(&clp->cl_lock);
3320        return ret;
3321}
3322
3323static void init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp, struct nfsd4_open *open) {
3324        struct nfs4_openowner *oo = open->op_openowner;
3325
3326        atomic_inc(&stp->st_stid.sc_count);
3327        stp->st_stid.sc_type = NFS4_OPEN_STID;
3328        INIT_LIST_HEAD(&stp->st_locks);
3329        stp->st_stateowner = nfs4_get_stateowner(&oo->oo_owner);
3330        get_nfs4_file(fp);
3331        stp->st_stid.sc_file = fp;
3332        stp->st_access_bmap = 0;
3333        stp->st_deny_bmap = 0;
3334        stp->st_openstp = NULL;
3335        spin_lock(&oo->oo_owner.so_client->cl_lock);
3336        list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
3337        spin_lock(&fp->fi_lock);
3338        list_add(&stp->st_perfile, &fp->fi_stateids);
3339        spin_unlock(&fp->fi_lock);
3340        spin_unlock(&oo->oo_owner.so_client->cl_lock);
3341}
3342
3343/*
3344 * In the 4.0 case we need to keep the owners around a little while to handle
3345 * CLOSE replay. We still do need to release any file access that is held by
3346 * them before returning however.
3347 */
3348static void
3349move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
3350{
3351        struct nfs4_ol_stateid *last;
3352        struct nfs4_openowner *oo = openowner(s->st_stateowner);
3353        struct nfsd_net *nn = net_generic(s->st_stid.sc_client->net,
3354                                                nfsd_net_id);
3355
3356        dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
3357
3358        /*
3359         * We know that we hold one reference via nfsd4_close, and another
3360         * "persistent" reference for the client. If the refcount is higher
3361         * than 2, then there are still calls in progress that are using this
3362         * stateid. We can't put the sc_file reference until they are finished.
3363         * Wait for the refcount to drop to 2. Since it has been unhashed,
3364         * there should be no danger of the refcount going back up again at
3365         * this point.
3366         */
3367        wait_event(close_wq, atomic_read(&s->st_stid.sc_count) == 2);
3368
3369        release_all_access(s);
3370        if (s->st_stid.sc_file) {
3371                put_nfs4_file(s->st_stid.sc_file);
3372                s->st_stid.sc_file = NULL;
3373        }
3374
3375        spin_lock(&nn->client_lock);
3376        last = oo->oo_last_closed_stid;
3377        oo->oo_last_closed_stid = s;
3378        list_move_tail(&oo->oo_close_lru, &nn->close_lru);
3379        oo->oo_time = get_seconds();
3380        spin_unlock(&nn->client_lock);
3381        if (last)
3382                nfs4_put_stid(&last->st_stid);
3383}
3384
3385/* search file_hashtbl[] for file */
3386static struct nfs4_file *
3387find_file_locked(struct knfsd_fh *fh, unsigned int hashval)
3388{
3389        struct nfs4_file *fp;
3390
3391        hlist_for_each_entry_rcu(fp, &file_hashtbl[hashval], fi_hash) {
3392                if (fh_match(&fp->fi_fhandle, fh)) {
3393                        if (atomic_inc_not_zero(&fp->fi_ref))
3394                                return fp;
3395                }
3396        }
3397        return NULL;
3398}
3399
3400struct nfs4_file *
3401find_file(struct knfsd_fh *fh)
3402{
3403        struct nfs4_file *fp;
3404        unsigned int hashval = file_hashval(fh);
3405
3406        rcu_read_lock();
3407        fp = find_file_locked(fh, hashval);
3408        rcu_read_unlock();
3409        return fp;
3410}
3411
3412static struct nfs4_file *
3413find_or_add_file(struct nfs4_file *new, struct knfsd_fh *fh)
3414{
3415        struct nfs4_file *fp;
3416        unsigned int hashval = file_hashval(fh);
3417
3418        rcu_read_lock();
3419        fp = find_file_locked(fh, hashval);
3420        rcu_read_unlock();
3421        if (fp)
3422                return fp;
3423
3424        spin_lock(&state_lock);
3425        fp = find_file_locked(fh, hashval);
3426        if (likely(fp == NULL)) {
3427                nfsd4_init_file(fh, hashval, new);
3428                fp = new;
3429        }
3430        spin_unlock(&state_lock);
3431
3432        return fp;
3433}
3434
3435/*
3436 * Called to check deny when READ with all zero stateid or
3437 * WRITE with all zero or all one stateid
3438 */
3439static __be32
3440nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
3441{
3442        struct nfs4_file *fp;
3443        __be32 ret = nfs_ok;
3444
3445        fp = find_file(&current_fh->fh_handle);
3446        if (!fp)
3447                return ret;
3448        /* Check for conflicting share reservations */
3449        spin_lock(&fp->fi_lock);
3450        if (fp->fi_share_deny & deny_type)
3451                ret = nfserr_locked;
3452        spin_unlock(&fp->fi_lock);
3453        put_nfs4_file(fp);
3454        return ret;
3455}
3456
3457static void nfsd4_cb_recall_prepare(struct nfsd4_callback *cb)
3458{
3459        struct nfs4_delegation *dp = cb_to_delegation(cb);
3460        struct nfsd_net *nn = net_generic(dp->dl_stid.sc_client->net,
3461                                          nfsd_net_id);
3462
3463        block_delegations(&dp->dl_stid.sc_file->fi_fhandle);
3464
3465        /*
3466         * We can't do this in nfsd_break_deleg_cb because it is
3467         * already holding inode->i_lock.
3468         *
3469         * If the dl_time != 0, then we know that it has already been
3470         * queued for a lease break. Don't queue it again.
3471         */
3472        spin_lock(&state_lock);
3473        if (dp->dl_time == 0) {
3474                dp->dl_time = get_seconds();
3475                list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
3476        }
3477        spin_unlock(&state_lock);
3478}
3479
3480static int nfsd4_cb_recall_done(struct nfsd4_callback *cb,
3481                struct rpc_task *task)
3482{
3483        struct nfs4_delegation *dp = cb_to_delegation(cb);
3484
3485        switch (task->tk_status) {
3486        case 0:
3487                return 1;
3488        case -EBADHANDLE:
3489        case -NFS4ERR_BAD_STATEID:
3490                /*
3491                 * Race: client probably got cb_recall before open reply
3492                 * granting delegation.
3493                 */
3494                if (dp->dl_retries--) {
3495                        rpc_delay(task, 2 * HZ);
3496                        return 0;
3497                }
3498                /*FALLTHRU*/
3499        default:
3500                return -1;
3501        }
3502}
3503
3504static void nfsd4_cb_recall_release(struct nfsd4_callback *cb)
3505{
3506        struct nfs4_delegation *dp = cb_to_delegation(cb);
3507
3508        nfs4_put_stid(&dp->dl_stid);
3509}
3510
3511static struct nfsd4_callback_ops nfsd4_cb_recall_ops = {
3512        .prepare        = nfsd4_cb_recall_prepare,
3513        .done           = nfsd4_cb_recall_done,
3514        .release        = nfsd4_cb_recall_release,
3515};
3516
3517static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
3518{
3519        /*
3520         * We're assuming the state code never drops its reference
3521         * without first removing the lease.  Since we're in this lease
3522         * callback (and since the lease code is serialized by the kernel
3523         * lock) we know the server hasn't removed the lease yet, we know
3524         * it's safe to take a reference.
3525         */
3526        atomic_inc(&dp->dl_stid.sc_count);
3527        nfsd4_run_cb(&dp->dl_recall);
3528}
3529
3530/* Called from break_lease() with i_lock held. */
3531static bool
3532nfsd_break_deleg_cb(struct file_lock *fl)
3533{
3534        bool ret = false;
3535        struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
3536        struct nfs4_delegation *dp;
3537
3538        if (!fp) {
3539                WARN(1, "(%p)->fl_owner NULL\n", fl);
3540                return ret;
3541        }
3542        if (fp->fi_had_conflict) {
3543                WARN(1, "duplicate break on %p\n", fp);
3544                return ret;
3545        }
3546        /*
3547         * We don't want the locks code to timeout the lease for us;
3548         * we'll remove it ourself if a delegation isn't returned
3549         * in time:
3550         */
3551        fl->fl_break_time = 0;
3552
3553        spin_lock(&fp->fi_lock);
3554        fp->fi_had_conflict = true;
3555        /*
3556         * If there are no delegations on the list, then return true
3557         * so that the lease code will go ahead and delete it.
3558         */
3559        if (list_empty(&fp->fi_delegations))
3560                ret = true;
3561        else
3562                list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
3563                        nfsd_break_one_deleg(dp);
3564        spin_unlock(&fp->fi_lock);
3565        return ret;
3566}
3567
3568static int
3569nfsd_change_deleg_cb(struct file_lock *onlist, int arg,
3570                     struct list_head *dispose)
3571{
3572        if (arg & F_UNLCK)
3573                return lease_modify(onlist, arg, dispose);
3574        else
3575                return -EAGAIN;
3576}
3577
3578static const struct lock_manager_operations nfsd_lease_mng_ops = {
3579        .lm_break = nfsd_break_deleg_cb,
3580        .lm_change = nfsd_change_deleg_cb,
3581};
3582
3583static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
3584{
3585        if (nfsd4_has_session(cstate))
3586                return nfs_ok;
3587        if (seqid == so->so_seqid - 1)
3588                return nfserr_replay_me;
3589        if (seqid == so->so_seqid)
3590                return nfs_ok;
3591        return nfserr_bad_seqid;
3592}
3593
3594static __be32 lookup_clientid(clientid_t *clid,
3595                struct nfsd4_compound_state *cstate,
3596                struct nfsd_net *nn)
3597{
3598        struct nfs4_client *found;
3599
3600        if (cstate->clp) {
3601                found = cstate->clp;
3602                if (!same_clid(&found->cl_clientid, clid))
3603                        return nfserr_stale_clientid;
3604                return nfs_ok;
3605        }
3606
3607        if (STALE_CLIENTID(clid, nn))
3608                return nfserr_stale_clientid;
3609
3610        /*
3611         * For v4.1+ we get the client in the SEQUENCE op. If we don't have one
3612         * cached already then we know this is for is for v4.0 and "sessions"
3613         * will be false.
3614         */
3615        WARN_ON_ONCE(cstate->session);
3616        spin_lock(&nn->client_lock);
3617        found = find_confirmed_client(clid, false, nn);
3618        if (!found) {
3619                spin_unlock(&nn->client_lock);
3620                return nfserr_expired;
3621        }
3622        atomic_inc(&found->cl_refcount);
3623        spin_unlock(&nn->client_lock);
3624
3625        /* Cache the nfs4_client in cstate! */
3626        cstate->clp = found;
3627        return nfs_ok;
3628}
3629
3630__be32
3631nfsd4_process_open1(struct nfsd4_compound_state *cstate,
3632                    struct nfsd4_open *open, struct nfsd_net *nn)
3633{
3634        clientid_t *clientid = &open->op_clientid;
3635        struct nfs4_client *clp = NULL;
3636        unsigned int strhashval;
3637        struct nfs4_openowner *oo = NULL;
3638        __be32 status;
3639
3640        if (STALE_CLIENTID(&open->op_clientid, nn))
3641                return nfserr_stale_clientid;
3642        /*
3643         * In case we need it later, after we've already created the
3644         * file and don't want to risk a further failure:
3645         */
3646        open->op_file = nfsd4_alloc_file();
3647        if (open->op_file == NULL)
3648                return nfserr_jukebox;
3649
3650        status = lookup_clientid(clientid, cstate, nn);
3651        if (status)
3652                return status;
3653        clp = cstate->clp;
3654
3655        strhashval = ownerstr_hashval(&open->op_owner);
3656        oo = find_openstateowner_str(strhashval, open, clp);
3657        open->op_openowner = oo;
3658        if (!oo) {
3659                goto new_owner;
3660        }
3661        if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
3662                /* Replace unconfirmed owners without checking for replay. */
3663                release_openowner(oo);
3664                open->op_openowner = NULL;
3665                goto new_owner;
3666        }
3667        status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
3668        if (status)
3669                return status;
3670        goto alloc_stateid;
3671new_owner:
3672        oo = alloc_init_open_stateowner(strhashval, open, cstate);
3673        if (oo == NULL)
3674                return nfserr_jukebox;
3675        open->op_openowner = oo;
3676alloc_stateid:
3677        open->op_stp = nfs4_alloc_open_stateid(clp);
3678        if (!open->op_stp)
3679                return nfserr_jukebox;
3680
3681        if (nfsd4_has_session(cstate) &&
3682            (cstate->current_fh.fh_export->ex_flags & NFSEXP_PNFS)) {
3683                open->op_odstate = alloc_clnt_odstate(clp);
3684                if (!open->op_odstate)
3685                        return nfserr_jukebox;
3686        }
3687
3688        return nfs_ok;
3689}
3690
3691static inline __be32
3692nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
3693{
3694        if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
3695                return nfserr_openmode;
3696        else
3697                return nfs_ok;
3698}
3699
3700static int share_access_to_flags(u32 share_access)
3701{
3702        return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
3703}
3704
3705static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
3706{
3707        struct nfs4_stid *ret;
3708
3709        ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
3710        if (!ret)
3711                return NULL;
3712        return delegstateid(ret);
3713}
3714
3715static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
3716{
3717        return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
3718               open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
3719}
3720
3721static __be32
3722nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
3723                struct nfs4_delegation **dp)
3724{
3725        int flags;
3726        __be32 status = nfserr_bad_stateid;
3727        struct nfs4_delegation *deleg;
3728
3729        deleg = find_deleg_stateid(cl, &open->op_delegate_stateid);
3730        if (deleg == NULL)
3731                goto out;
3732        flags = share_access_to_flags(open->op_share_access);
3733        status = nfs4_check_delegmode(deleg, flags);
3734        if (status) {
3735                nfs4_put_stid(&deleg->dl_stid);
3736                goto out;
3737        }
3738        *dp = deleg;
3739out:
3740        if (!nfsd4_is_deleg_cur(open))
3741                return nfs_ok;
3742        if (status)
3743                return status;
3744        open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3745        return nfs_ok;
3746}
3747
3748static struct nfs4_ol_stateid *
3749nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
3750{
3751        struct nfs4_ol_stateid *local, *ret = NULL;
3752        struct nfs4_openowner *oo = open->op_openowner;
3753
3754        spin_lock(&fp->fi_lock);
3755        list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
3756                /* ignore lock owners */
3757                if (local->st_stateowner->so_is_open_owner == 0)
3758                        continue;
3759                if (local->st_stateowner == &oo->oo_owner) {
3760                        ret = local;
3761                        atomic_inc(&ret->st_stid.sc_count);
3762                        break;
3763                }
3764        }
3765        spin_unlock(&fp->fi_lock);
3766        return ret;
3767}
3768
3769static inline int nfs4_access_to_access(u32 nfs4_access)
3770{
3771        int flags = 0;
3772
3773        if (nfs4_access & NFS4_SHARE_ACCESS_READ)
3774                flags |= NFSD_MAY_READ;
3775        if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
3776                flags |= NFSD_MAY_WRITE;
3777        return flags;
3778}
3779
3780static inline __be32
3781nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
3782                struct nfsd4_open *open)
3783{
3784        struct iattr iattr = {
3785                .ia_valid = ATTR_SIZE,
3786                .ia_size = 0,
3787        };
3788        if (!open->op_truncate)
3789                return 0;
3790        if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
3791                return nfserr_inval;
3792        return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
3793}
3794
3795static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
3796                struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
3797                struct nfsd4_open *open)
3798{
3799        struct file *filp = NULL;
3800        __be32 status;
3801        int oflag = nfs4_access_to_omode(open->op_share_access);
3802        int access = nfs4_access_to_access(open->op_share_access);
3803        unsigned char old_access_bmap, old_deny_bmap;
3804
3805        spin_lock(&fp->fi_lock);
3806
3807        /*
3808         * Are we trying to set a deny mode that would conflict with
3809         * current access?
3810         */
3811        status = nfs4_file_check_deny(fp, open->op_share_deny);
3812        if (status != nfs_ok) {
3813                spin_unlock(&fp->fi_lock);
3814                goto out;
3815        }
3816
3817        /* set access to the file */
3818        status = nfs4_file_get_access(fp, open->op_share_access);
3819        if (status != nfs_ok) {
3820                spin_unlock(&fp->fi_lock);
3821                goto out;
3822        }
3823
3824        /* Set access bits in stateid */
3825        old_access_bmap = stp->st_access_bmap;
3826        set_access(open->op_share_access, stp);
3827
3828        /* Set new deny mask */
3829        old_deny_bmap = stp->st_deny_bmap;
3830        set_deny(open->op_share_deny, stp);
3831        fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3832
3833        if (!fp->fi_fds[oflag]) {
3834                spin_unlock(&fp->fi_lock);
3835                status = nfsd_open(rqstp, cur_fh, S_IFREG, access, &filp);
3836                if (status)
3837                        goto out_put_access;
3838                spin_lock(&fp->fi_lock);
3839                if (!fp->fi_fds[oflag]) {
3840                        fp->fi_fds[oflag] = filp;
3841                        filp = NULL;
3842                }
3843        }
3844        spin_unlock(&fp->fi_lock);
3845        if (filp)
3846                fput(filp);
3847
3848        status = nfsd4_truncate(rqstp, cur_fh, open);
3849        if (status)
3850                goto out_put_access;
3851out:
3852        return status;
3853out_put_access:
3854        stp->st_access_bmap = old_access_bmap;
3855        nfs4_file_put_access(fp, open->op_share_access);
3856        reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap), stp);
3857        goto out;
3858}
3859
3860static __be32
3861nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp, struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp, struct nfsd4_open *open)
3862{
3863        __be32 status;
3864        unsigned char old_deny_bmap;
3865
3866        if (!test_access(open->op_share_access, stp))
3867                return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open);
3868
3869        /* test and set deny mode */
3870        spin_lock(&fp->fi_lock);
3871        status = nfs4_file_check_deny(fp, open->op_share_deny);
3872        if (status == nfs_ok) {
3873                old_deny_bmap = stp->st_deny_bmap;
3874                set_deny(open->op_share_deny, stp);
3875                fp->fi_share_deny |=
3876                                (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3877        }
3878        spin_unlock(&fp->fi_lock);
3879
3880        if (status != nfs_ok)
3881                return status;
3882
3883        status = nfsd4_truncate(rqstp, cur_fh, open);
3884        if (status != nfs_ok)
3885                reset_union_bmap_deny(old_deny_bmap, stp);
3886        return status;
3887}
3888
3889static void
3890nfs4_set_claim_prev(struct nfsd4_open *open, bool has_session)
3891{
3892        open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3893}
3894
3895/* Should we give out recallable state?: */
3896static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
3897{
3898        if (clp->cl_cb_state == NFSD4_CB_UP)
3899                return true;
3900        /*
3901         * In the sessions case, since we don't have to establish a
3902         * separate connection for callbacks, we assume it's OK
3903         * until we hear otherwise:
3904         */
3905        return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
3906}
3907
3908static struct file_lock *nfs4_alloc_init_lease(struct nfs4_file *fp, int flag)
3909{
3910        struct file_lock *fl;
3911
3912        fl = locks_alloc_lock();
3913        if (!fl)
3914                return NULL;
3915        fl->fl_lmops = &nfsd_lease_mng_ops;
3916        fl->fl_flags = FL_DELEG;
3917        fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
3918        fl->fl_end = OFFSET_MAX;
3919        fl->fl_owner = (fl_owner_t)fp;
3920        fl->fl_pid = current->tgid;
3921        return fl;
3922}
3923
3924static int nfs4_setlease(struct nfs4_delegation *dp)
3925{
3926        struct nfs4_file *fp = dp->dl_stid.sc_file;
3927        struct file_lock *fl, *ret;
3928        struct file *filp;
3929        int status = 0;
3930
3931        fl = nfs4_alloc_init_lease(fp, NFS4_OPEN_DELEGATE_READ);
3932        if (!fl)
3933                return -ENOMEM;
3934        filp = find_readable_file(fp);
3935        if (!filp) {
3936                /* We should always have a readable file here */
3937                WARN_ON_ONCE(1);
3938                return -EBADF;
3939        }
3940        fl->fl_file = filp;
3941        ret = fl;
3942        status = vfs_setlease(filp, fl->fl_type, &fl, NULL);
3943        if (fl)
3944                locks_free_lock(fl);
3945        if (status)
3946                goto out_fput;
3947        spin_lock(&state_lock);
3948        spin_lock(&fp->fi_lock);
3949        /* Did the lease get broken before we took the lock? */
3950        status = -EAGAIN;
3951        if (fp->fi_had_conflict)
3952                goto out_unlock;
3953        /* Race breaker */
3954        if (fp->fi_deleg_file) {
3955                status = 0;
3956                ++fp->fi_delegees;
3957                hash_delegation_locked(dp, fp);
3958                goto out_unlock;
3959        }
3960        fp->fi_deleg_file = filp;
3961        fp->fi_delegees = 1;
3962        hash_delegation_locked(dp, fp);
3963        spin_unlock(&fp->fi_lock);
3964        spin_unlock(&state_lock);
3965        return 0;
3966out_unlock:
3967        spin_unlock(&fp->fi_lock);
3968        spin_unlock(&state_lock);
3969out_fput:
3970        fput(filp);
3971        return status;
3972}
3973
3974static struct nfs4_delegation *
3975nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh,
3976                    struct nfs4_file *fp, struct nfs4_clnt_odstate *odstate)
3977{
3978        int status;
3979        struct nfs4_delegation *dp;
3980
3981        if (fp->fi_had_conflict)
3982                return ERR_PTR(-EAGAIN);
3983
3984        dp = alloc_init_deleg(clp, fh, odstate);
3985        if (!dp)
3986                return ERR_PTR(-ENOMEM);
3987
3988        get_nfs4_file(fp);
3989        spin_lock(&state_lock);
3990        spin_lock(&fp->fi_lock);
3991        dp->dl_stid.sc_file = fp;
3992        if (!fp->fi_deleg_file) {
3993                spin_unlock(&fp->fi_lock);
3994                spin_unlock(&state_lock);
3995                status = nfs4_setlease(dp);
3996                goto out;
3997        }
3998        if (fp->fi_had_conflict) {
3999                status = -EAGAIN;
4000                goto out_unlock;
4001        }
4002        ++fp->fi_delegees;
4003        hash_delegation_locked(dp, fp);
4004        status = 0;
4005out_unlock:
4006        spin_unlock(&fp->fi_lock);
4007        spin_unlock(&state_lock);
4008out:
4009        if (status) {
4010                put_clnt_odstate(dp->dl_clnt_odstate);
4011                nfs4_put_stid(&dp->dl_stid);
4012                return ERR_PTR(status);
4013        }
4014        return dp;
4015}
4016
4017static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
4018{
4019        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4020        if (status == -EAGAIN)
4021                open->op_why_no_deleg = WND4_CONTENTION;
4022        else {
4023                open->op_why_no_deleg = WND4_RESOURCE;
4024                switch (open->op_deleg_want) {
4025                case NFS4_SHARE_WANT_READ_DELEG:
4026                case NFS4_SHARE_WANT_WRITE_DELEG:
4027                case NFS4_SHARE_WANT_ANY_DELEG:
4028                        break;
4029                case NFS4_SHARE_WANT_CANCEL:
4030                        open->op_why_no_deleg = WND4_CANCELLED;
4031                        break;
4032                case NFS4_SHARE_WANT_NO_DELEG:
4033                        WARN_ON_ONCE(1);
4034                }
4035        }
4036}
4037
4038/*
4039 * Attempt to hand out a delegation.
4040 *
4041 * Note we don't support write delegations, and won't until the vfs has
4042 * proper support for them.
4043 */
4044static void
4045nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open,
4046                        struct nfs4_ol_stateid *stp)
4047{
4048        struct nfs4_delegation *dp;
4049        struct nfs4_openowner *oo = openowner(stp->st_stateowner);
4050        struct nfs4_client *clp = stp->st_stid.sc_client;
4051        int cb_up;
4052        int status = 0;
4053
4054        cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
4055        open->op_recall = 0;
4056        switch (open->op_claim_type) {
4057                case NFS4_OPEN_CLAIM_PREVIOUS:
4058                        if (!cb_up)
4059                                open->op_recall = 1;
4060                        if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
4061                                goto out_no_deleg;
4062                        break;
4063                case NFS4_OPEN_CLAIM_NULL:
4064                case NFS4_OPEN_CLAIM_FH:
4065                        /*
4066                         * Let's not give out any delegations till everyone's
4067                         * had the chance to reclaim theirs....
4068                         */
4069                        if (locks_in_grace(clp->net))
4070                                goto out_no_deleg;
4071                        if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
4072                                goto out_no_deleg;
4073                        /*
4074                         * Also, if the file was opened for write or
4075                         * create, there's a good chance the client's
4076                         * about to write to it, resulting in an
4077                         * immediate recall (since we don't support
4078                         * write delegations):
4079                         */
4080                        if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
4081                                goto out_no_deleg;
4082                        if (open->op_create == NFS4_OPEN_CREATE)
4083                                goto out_no_deleg;
4084                        break;
4085                default:
4086                        goto out_no_deleg;
4087        }
4088        dp = nfs4_set_delegation(clp, fh, stp->st_stid.sc_file, stp->st_clnt_odstate);
4089        if (IS_ERR(dp))
4090                goto out_no_deleg;
4091
4092        memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
4093
4094        dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
4095                STATEID_VAL(&dp->dl_stid.sc_stateid));
4096        open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
4097        nfs4_put_stid(&dp->dl_stid);
4098        return;
4099out_no_deleg:
4100        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
4101        if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
4102            open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
4103                dprintk("NFSD: WARNING: refusing delegation reclaim\n");
4104                open->op_recall = 1;
4105        }
4106
4107        /* 4.1 client asking for a delegation? */
4108        if (open->op_deleg_want)
4109                nfsd4_open_deleg_none_ext(open, status);
4110        return;
4111}
4112
4113static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
4114                                        struct nfs4_delegation *dp)
4115{
4116        if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
4117            dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
4118                open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4119                open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
4120        } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
4121                   dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
4122                open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4123                open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
4124        }
4125        /* Otherwise the client must be confused wanting a delegation
4126         * it already has, therefore we don't return
4127         * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
4128         */
4129}
4130
4131__be32
4132nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
4133{
4134        struct nfsd4_compoundres *resp = rqstp->rq_resp;
4135        struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
4136        struct nfs4_file *fp = NULL;
4137        struct nfs4_ol_stateid *stp = NULL;
4138        struct nfs4_delegation *dp = NULL;
4139        __be32 status;
4140
4141        /*
4142         * Lookup file; if found, lookup stateid and check open request,
4143         * and check for delegations in the process of being recalled.
4144         * If not found, create the nfs4_file struct
4145         */
4146        fp = find_or_add_file(open->op_file, &current_fh->fh_handle);
4147        if (fp != open->op_file) {
4148                status = nfs4_check_deleg(cl, open, &dp);
4149                if (status)
4150                        goto out;
4151                stp = nfsd4_find_existing_open(fp, open);
4152        } else {
4153                open->op_file = NULL;
4154                status = nfserr_bad_stateid;
4155                if (nfsd4_is_deleg_cur(open))
4156                        goto out;
4157        }
4158
4159        /*
4160         * OPEN the file, or upgrade an existing OPEN.
4161         * If truncate fails, the OPEN fails.
4162         */
4163        if (stp) {
4164                /* Stateid was found, this is an OPEN upgrade */
4165                status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
4166                if (status)
4167                        goto out;
4168        } else {
4169                stp = open->op_stp;
4170                open->op_stp = NULL;
4171                init_open_stateid(stp, fp, open);
4172                status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open);
4173                if (status) {
4174                        release_open_stateid(stp);
4175                        goto out;
4176                }
4177
4178                stp->st_clnt_odstate = find_or_hash_clnt_odstate(fp,
4179                                                        open->op_odstate);
4180                if (stp->st_clnt_odstate == open->op_odstate)
4181                        open->op_odstate = NULL;
4182        }
4183        update_stateid(&stp->st_stid.sc_stateid);
4184        memcpy(&open->op_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4185
4186        if (nfsd4_has_session(&resp->cstate)) {
4187                if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
4188                        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4189                        open->op_why_no_deleg = WND4_NOT_WANTED;
4190                        goto nodeleg;
4191                }
4192        }
4193
4194        /*
4195        * Attempt to hand out a delegation. No error return, because the
4196        * OPEN succeeds even if we fail.
4197        */
4198        nfs4_open_delegation(current_fh, open, stp);
4199nodeleg:
4200        status = nfs_ok;
4201
4202        dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
4203                STATEID_VAL(&stp->st_stid.sc_stateid));
4204out:
4205        /* 4.1 client trying to upgrade/downgrade delegation? */
4206        if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
4207            open->op_deleg_want)
4208                nfsd4_deleg_xgrade_none_ext(open, dp);
4209
4210        if (fp)
4211                put_nfs4_file(fp);
4212        if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
4213                nfs4_set_claim_prev(open, nfsd4_has_session(&resp->cstate));
4214        /*
4215        * To finish the open response, we just need to set the rflags.
4216        */
4217        open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
4218        if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
4219            !nfsd4_has_session(&resp->cstate))
4220                open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
4221        if (dp)
4222                nfs4_put_stid(&dp->dl_stid);
4223        if (stp)
4224                nfs4_put_stid(&stp->st_stid);
4225
4226        return status;
4227}
4228
4229void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
4230                              struct nfsd4_open *open)
4231{
4232        if (open->op_openowner) {
4233                struct nfs4_stateowner *so = &open->op_openowner->oo_owner;
4234
4235                nfsd4_cstate_assign_replay(cstate, so);
4236                nfs4_put_stateowner(so);
4237        }
4238        if (open->op_file)
4239                kmem_cache_free(file_slab, open->op_file);
4240        if (open->op_stp)
4241                nfs4_put_stid(&open->op_stp->st_stid);
4242        if (open->op_odstate)
4243                kmem_cache_free(odstate_slab, open->op_odstate);
4244}
4245
4246__be32
4247nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4248            clientid_t *clid)
4249{
4250        struct nfs4_client *clp;
4251        __be32 status;
4252        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4253
4254        dprintk("process_renew(%08x/%08x): starting\n", 
4255                        clid->cl_boot, clid->cl_id);
4256        status = lookup_clientid(clid, cstate, nn);
4257        if (status)
4258                goto out;
4259        clp = cstate->clp;
4260        status = nfserr_cb_path_down;
4261        if (!list_empty(&clp->cl_delegations)
4262                        && clp->cl_cb_state != NFSD4_CB_UP)
4263                goto out;
4264        status = nfs_ok;
4265out:
4266        return status;
4267}
4268
4269void
4270nfsd4_end_grace(struct nfsd_net *nn)
4271{
4272        /* do nothing if grace period already ended */
4273        if (nn->grace_ended)
4274                return;
4275
4276        dprintk("NFSD: end of grace period\n");
4277        nn->grace_ended = true;
4278        /*
4279         * If the server goes down again right now, an NFSv4
4280         * client will still be allowed to reclaim after it comes back up,
4281         * even if it hasn't yet had a chance to reclaim state this time.
4282         *
4283         */
4284        nfsd4_record_grace_done(nn);
4285        /*
4286         * At this point, NFSv4 clients can still reclaim.  But if the
4287         * server crashes, any that have not yet reclaimed will be out
4288         * of luck on the next boot.
4289         *
4290         * (NFSv4.1+ clients are considered to have reclaimed once they
4291         * call RECLAIM_COMPLETE.  NFSv4.0 clients are considered to
4292         * have reclaimed after their first OPEN.)
4293         */
4294        locks_end_grace(&nn->nfsd4_manager);
4295        /*
4296         * At this point, and once lockd and/or any other containers
4297         * exit their grace period, further reclaims will fail and
4298         * regular locking can resume.
4299         */
4300}
4301
4302static time_t
4303nfs4_laundromat(struct nfsd_net *nn)
4304{
4305        struct nfs4_client *clp;
4306        struct nfs4_openowner *oo;
4307        struct nfs4_delegation *dp;
4308        struct nfs4_ol_stateid *stp;
4309        struct list_head *pos, *next, reaplist;
4310        time_t cutoff = get_seconds() - nn->nfsd4_lease;
4311        time_t t, new_timeo = nn->nfsd4_lease;
4312
4313        dprintk("NFSD: laundromat service - starting\n");
4314        nfsd4_end_grace(nn);
4315        INIT_LIST_HEAD(&reaplist);
4316        spin_lock(&nn->client_lock);
4317        list_for_each_safe(pos, next, &nn->client_lru) {
4318                clp = list_entry(pos, struct nfs4_client, cl_lru);
4319                if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
4320                        t = clp->cl_time - cutoff;
4321                        new_timeo = min(new_timeo, t);
4322                        break;
4323                }
4324                if (mark_client_expired_locked(clp)) {
4325                        dprintk("NFSD: client in use (clientid %08x)\n",
4326                                clp->cl_clientid.cl_id);
4327                        continue;
4328                }
4329                list_add(&clp->cl_lru, &reaplist);
4330        }
4331        spin_unlock(&nn->client_lock);
4332        list_for_each_safe(pos, next, &reaplist) {
4333                clp = list_entry(pos, struct nfs4_client, cl_lru);
4334                dprintk("NFSD: purging unused client (clientid %08x)\n",
4335                        clp->cl_clientid.cl_id);
4336                list_del_init(&clp->cl_lru);
4337                expire_client(clp);
4338        }
4339        spin_lock(&state_lock);
4340        list_for_each_safe(pos, next, &nn->del_recall_lru) {
4341                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4342                if (net_generic(dp->dl_stid.sc_client->net, nfsd_net_id) != nn)
4343                        continue;
4344                if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
4345                        t = dp->dl_time - cutoff;
4346                        new_timeo = min(new_timeo, t);
4347                        break;
4348                }
4349                unhash_delegation_locked(dp);
4350                list_add(&dp->dl_recall_lru, &reaplist);
4351        }
4352        spin_unlock(&state_lock);
4353        while (!list_empty(&reaplist)) {
4354                dp = list_first_entry(&reaplist, struct nfs4_delegation,
4355                                        dl_recall_lru);
4356                list_del_init(&dp->dl_recall_lru);
4357                revoke_delegation(dp);
4358        }
4359
4360        spin_lock(&nn->client_lock);
4361        while (!list_empty(&nn->close_lru)) {
4362                oo = list_first_entry(&nn->close_lru, struct nfs4_openowner,
4363                                        oo_close_lru);
4364                if (time_after((unsigned long)oo->oo_time,
4365                               (unsigned long)cutoff)) {
4366                        t = oo->oo_time - cutoff;
4367                        new_timeo = min(new_timeo, t);
4368                        break;
4369                }
4370                list_del_init(&oo->oo_close_lru);
4371                stp = oo->oo_last_closed_stid;
4372                oo->oo_last_closed_stid = NULL;
4373                spin_unlock(&nn->client_lock);
4374                nfs4_put_stid(&stp->st_stid);
4375                spin_lock(&nn->client_lock);
4376        }
4377        spin_unlock(&nn->client_lock);
4378
4379        new_timeo = max_t(time_t, new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
4380        return new_timeo;
4381}
4382
4383static struct workqueue_struct *laundry_wq;
4384static void laundromat_main(struct work_struct *);
4385
4386static void
4387laundromat_main(struct work_struct *laundry)
4388{
4389        time_t t;
4390        struct delayed_work *dwork = container_of(laundry, struct delayed_work,
4391                                                  work);
4392        struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
4393                                           laundromat_work);
4394
4395        t = nfs4_laundromat(nn);
4396        dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
4397        queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
4398}
4399
4400static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_ol_stateid *stp)
4401{
4402        if (!fh_match(&fhp->fh_handle, &stp->st_stid.sc_file->fi_fhandle))
4403                return nfserr_bad_stateid;
4404        return nfs_ok;
4405}
4406
4407static inline int
4408access_permit_read(struct nfs4_ol_stateid *stp)
4409{
4410        return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
4411                test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
4412                test_access(NFS4_SHARE_ACCESS_WRITE, stp);
4413}
4414
4415static inline int
4416access_permit_write(struct nfs4_ol_stateid *stp)
4417{
4418        return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
4419                test_access(NFS4_SHARE_ACCESS_BOTH, stp);
4420}
4421
4422static
4423__be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
4424{
4425        __be32 status = nfserr_openmode;
4426
4427        /* For lock stateid's, we test the parent open, not the lock: */
4428        if (stp->st_openstp)
4429                stp = stp->st_openstp;
4430        if ((flags & WR_STATE) && !access_permit_write(stp))
4431                goto out;
4432        if ((flags & RD_STATE) && !access_permit_read(stp))
4433                goto out;
4434        status = nfs_ok;
4435out:
4436        return status;
4437}
4438
4439static inline __be32
4440check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
4441{
4442        if (ONE_STATEID(stateid) && (flags & RD_STATE))
4443                return nfs_ok;
4444        else if (locks_in_grace(net)) {
4445                /* Answer in remaining cases depends on existence of
4446                 * conflicting state; so we must wait out the grace period. */
4447                return nfserr_grace;
4448        } else if (flags & WR_STATE)
4449                return nfs4_share_conflict(current_fh,
4450                                NFS4_SHARE_DENY_WRITE);
4451        else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
4452                return nfs4_share_conflict(current_fh,
4453                                NFS4_SHARE_DENY_READ);
4454}
4455
4456/*
4457 * Allow READ/WRITE during grace period on recovered state only for files
4458 * that are not able to provide mandatory locking.
4459 */
4460static inline int
4461grace_disallows_io(struct net *net, struct inode *inode)
4462{
4463        return locks_in_grace(net) && mandatory_lock(inode);
4464}
4465
4466/* Returns true iff a is later than b: */
4467static bool stateid_generation_after(stateid_t *a, stateid_t *b)
4468{
4469        return (s32)(a->si_generation - b->si_generation) > 0;
4470}
4471
4472static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
4473{
4474        /*
4475         * When sessions are used the stateid generation number is ignored
4476         * when it is zero.
4477         */
4478        if (has_session && in->si_generation == 0)
4479                return nfs_ok;
4480
4481        if (in->si_generation == ref->si_generation)
4482                return nfs_ok;
4483
4484        /* If the client sends us a stateid from the future, it's buggy: */
4485        if (stateid_generation_after(in, ref))
4486                return nfserr_bad_stateid;
4487        /*
4488         * However, we could see a stateid from the past, even from a
4489         * non-buggy client.  For example, if the client sends a lock
4490         * while some IO is outstanding, the lock may bump si_generation
4491         * while the IO is still in flight.  The client could avoid that
4492         * situation by waiting for responses on all the IO requests,
4493         * but better performance may result in retrying IO that
4494         * receives an old_stateid error if requests are rarely
4495         * reordered in flight:
4496         */
4497        return nfserr_old_stateid;
4498}
4499
4500static __be32 nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid *ols)
4501{
4502        if (ols->st_stateowner->so_is_open_owner &&
4503            !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
4504                return nfserr_bad_stateid;
4505        return nfs_ok;
4506}
4507
4508static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
4509{
4510        struct nfs4_stid *s;
4511        __be32 status = nfserr_bad_stateid;
4512
4513        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4514                return status;
4515        /* Client debugging aid. */
4516        if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
4517                char addr_str[INET6_ADDRSTRLEN];
4518                rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
4519                                 sizeof(addr_str));
4520                pr_warn_ratelimited("NFSD: client %s testing state ID "
4521                                        "with incorrect client ID\n", addr_str);
4522                return status;
4523        }
4524        spin_lock(&cl->cl_lock);
4525        s = find_stateid_locked(cl, stateid);
4526        if (!s)
4527                goto out_unlock;
4528        status = check_stateid_generation(stateid, &s->sc_stateid, 1);
4529        if (status)
4530                goto out_unlock;
4531        switch (s->sc_type) {
4532        case NFS4_DELEG_STID:
4533                status = nfs_ok;
4534                break;
4535        case NFS4_REVOKED_DELEG_STID:
4536                status = nfserr_deleg_revoked;
4537                break;
4538        case NFS4_OPEN_STID:
4539        case NFS4_LOCK_STID:
4540                status = nfsd4_check_openowner_confirmed(openlockstateid(s));
4541                break;
4542        default:
4543                printk("unknown stateid type %x\n", s->sc_type);
4544                /* Fallthrough */
4545        case NFS4_CLOSED_STID:
4546        case NFS4_CLOSED_DELEG_STID:
4547                status = nfserr_bad_stateid;
4548        }
4549out_unlock:
4550        spin_unlock(&cl->cl_lock);
4551        return status;
4552}
4553
4554__be32
4555nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
4556                     stateid_t *stateid, unsigned char typemask,
4557                     struct nfs4_stid **s, struct nfsd_net *nn)
4558{
4559        __be32 status;
4560
4561        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4562                return nfserr_bad_stateid;
4563        status = lookup_clientid(&stateid->si_opaque.so_clid, cstate, nn);
4564        if (status == nfserr_stale_clientid) {
4565                if (cstate->session)
4566                        return nfserr_bad_stateid;
4567                return nfserr_stale_stateid;
4568        }
4569        if (status)
4570                return status;
4571        *s = find_stateid_by_type(cstate->clp, stateid, typemask);
4572        if (!*s)
4573                return nfserr_bad_stateid;
4574        return nfs_ok;
4575}
4576
4577/*
4578* Checks for stateid operations
4579*/
4580__be32
4581nfs4_preprocess_stateid_op(struct net *net, struct nfsd4_compound_state *cstate,
4582                           stateid_t *stateid, int flags, struct file **filpp)
4583{
4584        struct nfs4_stid *s;
4585        struct nfs4_ol_stateid *stp = NULL;
4586        struct nfs4_delegation *dp = NULL;
4587        struct svc_fh *current_fh = &cstate->current_fh;
4588        struct inode *ino = d_inode(current_fh->fh_dentry);
4589        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4590        struct file *file = NULL;
4591        __be32 status;
4592
4593        if (filpp)
4594                *filpp = NULL;
4595
4596        if (grace_disallows_io(net, ino))
4597                return nfserr_grace;
4598
4599        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4600                return check_special_stateids(net, current_fh, stateid, flags);
4601
4602        status = nfsd4_lookup_stateid(cstate, stateid,
4603                                NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
4604                                &s, nn);
4605        if (status)
4606                return status;
4607        status = check_stateid_generation(stateid, &s->sc_stateid, nfsd4_has_session(cstate));
4608        if (status)
4609                goto out;
4610        switch (s->sc_type) {
4611        case NFS4_DELEG_STID:
4612                dp = delegstateid(s);
4613                status = nfs4_check_delegmode(dp, flags);
4614                if (status)
4615                        goto out;
4616                if (filpp) {
4617                        file = dp->dl_stid.sc_file->fi_deleg_file;
4618                        if (!file) {
4619                                WARN_ON_ONCE(1);
4620                                status = nfserr_serverfault;
4621                                goto out;
4622                        }
4623                        get_file(file);
4624                }
4625                break;
4626        case NFS4_OPEN_STID:
4627        case NFS4_LOCK_STID:
4628                stp = openlockstateid(s);
4629                status = nfs4_check_fh(current_fh, stp);
4630                if (status)
4631                        goto out;
4632                status = nfsd4_check_openowner_confirmed(stp);
4633                if (status)
4634                        goto out;
4635                status = nfs4_check_openmode(stp, flags);
4636                if (status)
4637                        goto out;
4638                if (filpp) {
4639                        struct nfs4_file *fp = stp->st_stid.sc_file;
4640
4641                        if (flags & RD_STATE)
4642                                file = find_readable_file(fp);
4643                        else
4644                                file = find_writeable_file(fp);
4645                }
4646                break;
4647        default:
4648                status = nfserr_bad_stateid;
4649                goto out;
4650        }
4651        status = nfs_ok;
4652        if (file)
4653                *filpp = file;
4654out:
4655        nfs4_put_stid(s);
4656        return status;
4657}
4658
4659/*
4660 * Test if the stateid is valid
4661 */
4662__be32
4663nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4664                   struct nfsd4_test_stateid *test_stateid)
4665{
4666        struct nfsd4_test_stateid_id *stateid;
4667        struct nfs4_client *cl = cstate->session->se_client;
4668
4669        list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
4670                stateid->ts_id_status =
4671                        nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
4672
4673        return nfs_ok;
4674}
4675
4676__be32
4677nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4678                   struct nfsd4_free_stateid *free_stateid)
4679{
4680        stateid_t *stateid = &free_stateid->fr_stateid;
4681        struct nfs4_stid *s;
4682        struct nfs4_delegation *dp;
4683        struct nfs4_ol_stateid *stp;
4684        struct nfs4_client *cl = cstate->session->se_client;
4685        __be32 ret = nfserr_bad_stateid;
4686
4687        spin_lock(&cl->cl_lock);
4688        s = find_stateid_locked(cl, stateid);
4689        if (!s)
4690                goto out_unlock;
4691        switch (s->sc_type) {
4692        case NFS4_DELEG_STID:
4693                ret = nfserr_locks_held;
4694                break;
4695        case NFS4_OPEN_STID:
4696                ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4697                if (ret)
4698                        break;
4699                ret = nfserr_locks_held;
4700                break;
4701        case NFS4_LOCK_STID:
4702                ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4703                if (ret)
4704                        break;
4705                stp = openlockstateid(s);
4706                ret = nfserr_locks_held;
4707                if (check_for_locks(stp->st_stid.sc_file,
4708                                    lockowner(stp->st_stateowner)))
4709                        break;
4710                unhash_lock_stateid(stp);
4711                spin_unlock(&cl->cl_lock);
4712                nfs4_put_stid(s);
4713                ret = nfs_ok;
4714                goto out;
4715        case NFS4_REVOKED_DELEG_STID:
4716                dp = delegstateid(s);
4717                list_del_init(&dp->dl_recall_lru);
4718                spin_unlock(&cl->cl_lock);
4719                nfs4_put_stid(s);
4720                ret = nfs_ok;
4721                goto out;
4722        /* Default falls through and returns nfserr_bad_stateid */
4723        }
4724out_unlock:
4725        spin_unlock(&cl->cl_lock);
4726out:
4727        return ret;
4728}
4729
4730static inline int
4731setlkflg (int type)
4732{
4733        return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
4734                RD_STATE : WR_STATE;
4735}
4736
4737static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
4738{
4739        struct svc_fh *current_fh = &cstate->current_fh;
4740        struct nfs4_stateowner *sop = stp->st_stateowner;
4741        __be32 status;
4742
4743        status = nfsd4_check_seqid(cstate, sop, seqid);
4744        if (status)
4745                return status;
4746        if (stp->st_stid.sc_type == NFS4_CLOSED_STID
4747                || stp->st_stid.sc_type == NFS4_REVOKED_DELEG_STID)
4748                /*
4749                 * "Closed" stateid's exist *only* to return
4750                 * nfserr_replay_me from the previous step, and
4751                 * revoked delegations are kept only for free_stateid.
4752                 */
4753                return nfserr_bad_stateid;
4754        status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
4755        if (status)
4756                return status;
4757        return nfs4_check_fh(current_fh, stp);
4758}
4759
4760/* 
4761 * Checks for sequence id mutating operations. 
4762 */
4763static __be32
4764nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4765                         stateid_t *stateid, char typemask,
4766                         struct nfs4_ol_stateid **stpp,
4767                         struct nfsd_net *nn)
4768{
4769        __be32 status;
4770        struct nfs4_stid *s;
4771        struct nfs4_ol_stateid *stp = NULL;
4772
4773        dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
4774                seqid, STATEID_VAL(stateid));
4775
4776        *stpp = NULL;
4777        status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
4778        if (status)
4779                return status;
4780        stp = openlockstateid(s);
4781        nfsd4_cstate_assign_replay(cstate, stp->st_stateowner);
4782
4783        status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
4784        if (!status)
4785                *stpp = stp;
4786        else
4787                nfs4_put_stid(&stp->st_stid);
4788        return status;
4789}
4790
4791static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
4792                                                 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
4793{
4794        __be32 status;
4795        struct nfs4_openowner *oo;
4796        struct nfs4_ol_stateid *stp;
4797
4798        status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
4799                                                NFS4_OPEN_STID, &stp, nn);
4800        if (status)
4801                return status;
4802        oo = openowner(stp->st_stateowner);
4803        if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
4804                nfs4_put_stid(&stp->st_stid);
4805                return nfserr_bad_stateid;
4806        }
4807        *stpp = stp;
4808        return nfs_ok;
4809}
4810
4811__be32
4812nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4813                   struct nfsd4_open_confirm *oc)
4814{
4815        __be32 status;
4816        struct nfs4_openowner *oo;
4817        struct nfs4_ol_stateid *stp;
4818        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4819
4820        dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
4821                        cstate->current_fh.fh_dentry);
4822
4823        status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
4824        if (status)
4825                return status;
4826
4827        status = nfs4_preprocess_seqid_op(cstate,
4828                                        oc->oc_seqid, &oc->oc_req_stateid,
4829                                        NFS4_OPEN_STID, &stp, nn);
4830        if (status)
4831                goto out;
4832        oo = openowner(stp->st_stateowner);
4833        status = nfserr_bad_stateid;
4834        if (oo->oo_flags & NFS4_OO_CONFIRMED)
4835                goto put_stateid;
4836        oo->oo_flags |= NFS4_OO_CONFIRMED;
4837        update_stateid(&stp->st_stid.sc_stateid);
4838        memcpy(&oc->oc_resp_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4839        dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
4840                __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
4841
4842        nfsd4_client_record_create(oo->oo_owner.so_client);
4843        status = nfs_ok;
4844put_stateid:
4845        nfs4_put_stid(&stp->st_stid);
4846out:
4847        nfsd4_bump_seqid(cstate, status);
4848        return status;
4849}
4850
4851static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
4852{
4853        if (!test_access(access, stp))
4854                return;
4855        nfs4_file_put_access(stp->st_stid.sc_file, access);
4856        clear_access(access, stp);
4857}
4858
4859static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
4860{
4861        switch (to_access) {
4862        case NFS4_SHARE_ACCESS_READ:
4863                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
4864                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4865                break;
4866        case NFS4_SHARE_ACCESS_WRITE:
4867                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
4868                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
4869                break;
4870        case NFS4_SHARE_ACCESS_BOTH:
4871                break;
4872        default:
4873                WARN_ON_ONCE(1);
4874        }
4875}
4876
4877__be32
4878nfsd4_open_downgrade(struct svc_rqst *rqstp,
4879                     struct nfsd4_compound_state *cstate,
4880                     struct nfsd4_open_downgrade *od)
4881{
4882        __be32 status;
4883        struct nfs4_ol_stateid *stp;
4884        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4885
4886        dprintk("NFSD: nfsd4_open_downgrade on file %pd\n", 
4887                        cstate->current_fh.fh_dentry);
4888
4889        /* We don't yet support WANT bits: */
4890        if (od->od_deleg_want)
4891                dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
4892                        od->od_deleg_want);
4893
4894        status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
4895                                        &od->od_stateid, &stp, nn);
4896        if (status)
4897                goto out; 
4898        status = nfserr_inval;
4899        if (!test_access(od->od_share_access, stp)) {
4900                dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
4901                        stp->st_access_bmap, od->od_share_access);
4902                goto put_stateid;
4903        }
4904        if (!test_deny(od->od_share_deny, stp)) {
4905                dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
4906                        stp->st_deny_bmap, od->od_share_deny);
4907                goto put_stateid;
4908        }
4909        nfs4_stateid_downgrade(stp, od->od_share_access);
4910
4911        reset_union_bmap_deny(od->od_share_deny, stp);
4912
4913        update_stateid(&stp->st_stid.sc_stateid);
4914        memcpy(&od->od_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4915        status = nfs_ok;
4916put_stateid:
4917        nfs4_put_stid(&stp->st_stid);
4918out:
4919        nfsd4_bump_seqid(cstate, status);
4920        return status;
4921}
4922
4923static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
4924{
4925        struct nfs4_client *clp = s->st_stid.sc_client;
4926        LIST_HEAD(reaplist);
4927
4928        s->st_stid.sc_type = NFS4_CLOSED_STID;
4929        spin_lock(&clp->cl_lock);
4930        unhash_open_stateid(s, &reaplist);
4931
4932        if (clp->cl_minorversion) {
4933                put_ol_stateid_locked(s, &reaplist);
4934                spin_unlock(&clp->cl_lock);
4935                free_ol_stateid_reaplist(&reaplist);
4936        } else {
4937                spin_unlock(&clp->cl_lock);
4938                free_ol_stateid_reaplist(&reaplist);
4939                move_to_close_lru(s, clp->net);
4940        }
4941}
4942
4943/*
4944 * nfs4_unlock_state() called after encode
4945 */
4946__be32
4947nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4948            struct nfsd4_close *close)
4949{
4950        __be32 status;
4951        struct nfs4_ol_stateid *stp;
4952        struct net *net = SVC_NET(rqstp);
4953        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4954
4955        dprintk("NFSD: nfsd4_close on file %pd\n", 
4956                        cstate->current_fh.fh_dentry);
4957
4958        status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
4959                                        &close->cl_stateid,
4960                                        NFS4_OPEN_STID|NFS4_CLOSED_STID,
4961                                        &stp, nn);
4962        nfsd4_bump_seqid(cstate, status);
4963        if (status)
4964                goto out; 
4965        update_stateid(&stp->st_stid.sc_stateid);
4966        memcpy(&close->cl_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
4967
4968        nfsd4_close_open_stateid(stp);
4969
4970        /* put reference from nfs4_preprocess_seqid_op */
4971        nfs4_put_stid(&stp->st_stid);
4972out:
4973        return status;
4974}
4975
4976__be32
4977nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4978                  struct nfsd4_delegreturn *dr)
4979{
4980        struct nfs4_delegation *dp;
4981        stateid_t *stateid = &dr->dr_stateid;
4982        struct nfs4_stid *s;
4983        __be32 status;
4984        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4985
4986        if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
4987                return status;
4988
4989        status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
4990        if (status)
4991                goto out;
4992        dp = delegstateid(s);
4993        status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
4994        if (status)
4995                goto put_stateid;
4996
4997        destroy_delegation(dp);
4998put_stateid:
4999        nfs4_put_stid(&dp->dl_stid);
5000out:
5001        return status;
5002}
5003
5004
5005#define LOFF_OVERFLOW(start, len)      ((u64)(len) > ~(u64)(start))
5006
5007static inline u64
5008end_offset(u64 start, u64 len)
5009{
5010        u64 end;
5011
5012        end = start + len;
5013        return end >= start ? end: NFS4_MAX_UINT64;
5014}
5015
5016/* last octet in a range */
5017static inline u64
5018last_byte_offset(u64 start, u64 len)
5019{
5020        u64 end;
5021
5022        WARN_ON_ONCE(!len);
5023        end = start + len;
5024        return end > start ? end - 1: NFS4_MAX_UINT64;
5025}
5026
5027/*
5028 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
5029 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
5030 * byte, because of sign extension problems.  Since NFSv4 calls for 64-bit
5031 * locking, this prevents us from being completely protocol-compliant.  The
5032 * real solution to this problem is to start using unsigned file offsets in
5033 * the VFS, but this is a very deep change!
5034 */
5035static inline void
5036nfs4_transform_lock_offset(struct file_lock *lock)
5037{
5038        if (lock->fl_start < 0)
5039                lock->fl_start = OFFSET_MAX;
5040        if (lock->fl_end < 0)
5041                lock->fl_end = OFFSET_MAX;
5042}
5043
5044static fl_owner_t
5045nfsd4_fl_get_owner(fl_owner_t owner)
5046{
5047        struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
5048
5049        nfs4_get_stateowner(&lo->lo_owner);
5050        return owner;
5051}
5052
5053static void
5054nfsd4_fl_put_owner(fl_owner_t owner)
5055{
5056        struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
5057
5058        if (lo)
5059                nfs4_put_stateowner(&lo->lo_owner);
5060}
5061
5062static const struct lock_manager_operations nfsd_posix_mng_ops  = {
5063        .lm_get_owner = nfsd4_fl_get_owner,
5064        .lm_put_owner = nfsd4_fl_put_owner,
5065};
5066
5067static inline void
5068nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
5069{
5070        struct nfs4_lockowner *lo;
5071
5072        if (fl->fl_lmops == &nfsd_posix_mng_ops) {
5073                lo = (struct nfs4_lockowner *) fl->fl_owner;
5074                deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
5075                                        lo->lo_owner.so_owner.len, GFP_KERNEL);
5076                if (!deny->ld_owner.data)
5077                        /* We just don't care that much */
5078                        goto nevermind;
5079                deny->ld_owner.len = lo->lo_owner.so_owner.len;
5080                deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
5081        } else {
5082nevermind:
5083                deny->ld_owner.len = 0;
5084                deny->ld_owner.data = NULL;
5085                deny->ld_clientid.cl_boot = 0;
5086                deny->ld_clientid.cl_id = 0;
5087        }
5088        deny->ld_start = fl->fl_start;
5089        deny->ld_length = NFS4_MAX_UINT64;
5090        if (fl->fl_end != NFS4_MAX_UINT64)
5091                deny->ld_length = fl->fl_end - fl->fl_start + 1;        
5092        deny->ld_type = NFS4_READ_LT;
5093        if (fl->fl_type != F_RDLCK)
5094                deny->ld_type = NFS4_WRITE_LT;
5095}
5096
5097static struct nfs4_lockowner *
5098find_lockowner_str_locked(clientid_t *clid, struct xdr_netobj *owner,
5099                struct nfs4_client *clp)
5100{
5101        unsigned int strhashval = ownerstr_hashval(owner);
5102        struct nfs4_stateowner *so;
5103
5104        lockdep_assert_held(&clp->cl_lock);
5105
5106        list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[strhashval],
5107                            so_strhash) {
5108                if (so->so_is_open_owner)
5109                        continue;
5110                if (same_owner_str(so, owner))
5111                        return lockowner(nfs4_get_stateowner(so));
5112        }
5113        return NULL;
5114}
5115
5116static struct nfs4_lockowner *
5117find_lockowner_str(clientid_t *clid, struct xdr_netobj *owner,
5118                struct nfs4_client *clp)
5119{
5120        struct nfs4_lockowner *lo;
5121
5122        spin_lock(&clp->cl_lock);
5123        lo = find_lockowner_str_locked(clid, owner, clp);
5124        spin_unlock(&clp->cl_lock);
5125        return lo;
5126}
5127
5128static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
5129{
5130        unhash_lockowner_locked(lockowner(sop));
5131}
5132
5133static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
5134{
5135        struct nfs4_lockowner *lo = lockowner(sop);
5136
5137        kmem_cache_free(lockowner_slab, lo);
5138}
5139
5140static const struct nfs4_stateowner_operations lockowner_ops = {
5141        .so_unhash =    nfs4_unhash_lockowner,
5142        .so_free =      nfs4_free_lockowner,
5143};
5144
5145/*
5146 * Alloc a lock owner structure.
5147 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has 
5148 * occurred. 
5149 *
5150 * strhashval = ownerstr_hashval
5151 */
5152static struct nfs4_lockowner *
5153alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp,
5154                           struct nfs4_ol_stateid *open_stp,
5155                           struct nfsd4_lock *lock)
5156{
5157        struct nfs4_lockowner *lo, *ret;
5158
5159        lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
5160        if (!lo)
5161                return NULL;
5162        INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
5163        lo->lo_owner.so_is_open_owner = 0;
5164        lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
5165        lo->lo_owner.so_ops = &lockowner_ops;
5166        spin_lock(&clp->cl_lock);
5167        ret = find_lockowner_str_locked(&clp->cl_clientid,
5168                        &lock->lk_new_owner, clp);
5169        if (ret == NULL) {
5170                list_add(&lo->lo_owner.so_strhash,
5171                         &clp->cl_ownerstr_hashtbl[strhashval]);
5172                ret = lo;
5173        } else
5174                nfs4_free_lockowner(&lo->lo_owner);
5175        spin_unlock(&clp->cl_lock);
5176        return ret;
5177}
5178
5179static void
5180init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
5181                  struct nfs4_file *fp, struct inode *inode,
5182                  struct nfs4_ol_stateid *open_stp)
5183{
5184        struct nfs4_client *clp = lo->lo_owner.so_client;
5185
5186        lockdep_assert_held(&clp->cl_lock);
5187
5188        atomic_inc(&stp->st_stid.sc_count);
5189        stp->st_stid.sc_type = NFS4_LOCK_STID;
5190        stp->st_stateowner = nfs4_get_stateowner(&lo->lo_owner);
5191        get_nfs4_file(fp);
5192        stp->st_stid.sc_file = fp;
5193        stp->st_stid.sc_free = nfs4_free_lock_stateid;
5194        stp->st_access_bmap = 0;
5195        stp->st_deny_bmap = open_stp->st_deny_bmap;
5196        stp->st_openstp = open_stp;
5197        list_add(&stp->st_locks, &open_stp->st_locks);
5198        list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
5199        spin_lock(&fp->fi_lock);
5200        list_add(&stp->st_perfile, &fp->fi_stateids);
5201        spin_unlock(&fp->fi_lock);
5202}
5203
5204static struct nfs4_ol_stateid *
5205find_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp)
5206{
5207        struct nfs4_ol_stateid *lst;
5208        struct nfs4_client *clp = lo->lo_owner.so_client;
5209
5210        lockdep_assert_held(&clp->cl_lock);
5211
5212        list_for_each_entry(lst, &lo->lo_owner.so_stateids, st_perstateowner) {
5213                if (lst->st_stid.sc_file == fp) {
5214                        atomic_inc(&lst->st_stid.sc_count);
5215                        return lst;
5216                }
5217        }
5218        return NULL;
5219}
5220
5221static struct nfs4_ol_stateid *
5222find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
5223                            struct inode *inode, struct nfs4_ol_stateid *ost,
5224                            bool *new)
5225{
5226        struct nfs4_stid *ns = NULL;
5227        struct nfs4_ol_stateid *lst;
5228        struct nfs4_openowner *oo = openowner(ost->st_stateowner);
5229        struct nfs4_client *clp = oo->oo_owner.so_client;
5230
5231        spin_lock(&clp->cl_lock);
5232        lst = find_lock_stateid(lo, fi);
5233        if (lst == NULL) {
5234                spin_unlock(&clp->cl_lock);
5235                ns = nfs4_alloc_stid(clp, stateid_slab);
5236                if (ns == NULL)
5237                        return NULL;
5238
5239                spin_lock(&clp->cl_lock);
5240                lst = find_lock_stateid(lo, fi);
5241                if (likely(!lst)) {
5242                        lst = openlockstateid(ns);
5243                        init_lock_stateid(lst, lo, fi, inode, ost);
5244                        ns = NULL;
5245                        *new = true;
5246                }
5247        }
5248        spin_unlock(&clp->cl_lock);
5249        if (ns)
5250                nfs4_put_stid(ns);
5251        return lst;
5252}
5253
5254static int
5255check_lock_length(u64 offset, u64 length)
5256{
5257        return ((length == 0)  || ((length != NFS4_MAX_UINT64) &&
5258             LOFF_OVERFLOW(offset, length)));
5259}
5260
5261static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
5262{
5263        struct nfs4_file *fp = lock_stp->st_stid.sc_file;
5264
5265        lockdep_assert_held(&fp->fi_lock);
5266
5267        if (test_access(access, lock_stp))
5268                return;
5269        __nfs4_file_get_access(fp, access);
5270        set_access(access, lock_stp);
5271}
5272
5273static __be32
5274lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
5275                            struct nfs4_ol_stateid *ost,
5276                            struct nfsd4_lock *lock,
5277                            struct nfs4_ol_stateid **lst, bool *new)
5278{
5279        __be32 status;
5280        struct nfs4_file *fi = ost->st_stid.sc_file;
5281        struct nfs4_openowner *oo = openowner(ost->st_stateowner);
5282        struct nfs4_client *cl = oo->oo_owner.so_client;
5283        struct inode *inode = d_inode(cstate->current_fh.fh_dentry);
5284        struct nfs4_lockowner *lo;
5285        unsigned int strhashval;
5286
5287        lo = find_lockowner_str(&cl->cl_clientid, &lock->v.new.owner, cl);
5288        if (!lo) {
5289                strhashval = ownerstr_hashval(&lock->v.new.owner);
5290                lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
5291                if (lo == NULL)
5292                        return nfserr_jukebox;
5293        } else {
5294                /* with an existing lockowner, seqids must be the same */
5295                status = nfserr_bad_seqid;
5296                if (!cstate->minorversion &&
5297                    lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
5298                        goto out;
5299        }
5300
5301        *lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
5302        if (*lst == NULL) {
5303                status = nfserr_jukebox;
5304                goto out;
5305        }
5306        status = nfs_ok;
5307out:
5308        nfs4_put_stateowner(&lo->lo_owner);
5309        return status;
5310}
5311
5312/*
5313 *  LOCK operation 
5314 */
5315__be32
5316nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5317           struct nfsd4_lock *lock)
5318{
5319        struct nfs4_openowner *open_sop = NULL;
5320        struct nfs4_lockowner *lock_sop = NULL;
5321        struct nfs4_ol_stateid *lock_stp = NULL;
5322        struct nfs4_ol_stateid *open_stp = NULL;
5323        struct nfs4_file *fp;
5324        struct file *filp = NULL;
5325        struct file_lock *file_lock = NULL;
5326        struct file_lock *conflock = NULL;
5327        __be32 status = 0;
5328        int lkflg;
5329        int err;
5330        bool new = false;
5331        struct net *net = SVC_NET(rqstp);
5332        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5333
5334        dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
5335                (long long) lock->lk_offset,
5336                (long long) lock->lk_length);
5337
5338        if (check_lock_length(lock->lk_offset, lock->lk_length))
5339                 return nfserr_inval;
5340
5341        if ((status = fh_verify(rqstp, &cstate->current_fh,
5342                                S_IFREG, NFSD_MAY_LOCK))) {
5343                dprintk("NFSD: nfsd4_lock: permission denied!\n");
5344                return status;
5345        }
5346
5347        if (lock->lk_is_new) {
5348                if (nfsd4_has_session(cstate))
5349                        /* See rfc 5661 18.10.3: given clientid is ignored: */
5350                        memcpy(&lock->v.new.clientid,
5351                                &cstate->session->se_client->cl_clientid,
5352                                sizeof(clientid_t));
5353
5354                status = nfserr_stale_clientid;
5355                if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
5356                        goto out;
5357
5358                /* validate and update open stateid and open seqid */
5359                status = nfs4_preprocess_confirmed_seqid_op(cstate,
5360                                        lock->lk_new_open_seqid,
5361                                        &lock->lk_new_open_stateid,
5362                                        &open_stp, nn);
5363                if (status)
5364                        goto out;
5365                open_sop = openowner(open_stp->st_stateowner);
5366                status = nfserr_bad_stateid;
5367                if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
5368                                                &lock->v.new.clientid))
5369                        goto out;
5370                status = lookup_or_create_lock_state(cstate, open_stp, lock,
5371                                                        &lock_stp, &new);
5372        } else {
5373                status = nfs4_preprocess_seqid_op(cstate,
5374                                       lock->lk_old_lock_seqid,
5375                                       &lock->lk_old_lock_stateid,
5376                                       NFS4_LOCK_STID, &lock_stp, nn);
5377        }
5378        if (status)
5379                goto out;
5380        lock_sop = lockowner(lock_stp->st_stateowner);
5381
5382        lkflg = setlkflg(lock->lk_type);
5383        status = nfs4_check_openmode(lock_stp, lkflg);
5384        if (status)
5385                goto out;
5386
5387        status = nfserr_grace;
5388        if (locks_in_grace(net) && !lock->lk_reclaim)
5389                goto out;
5390        status = nfserr_no_grace;
5391        if (!locks_in_grace(net) && lock->lk_reclaim)
5392                goto out;
5393
5394        file_lock = locks_alloc_lock();
5395        if (!file_lock) {
5396                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5397                status = nfserr_jukebox;
5398                goto out;
5399        }
5400
5401        fp = lock_stp->st_stid.sc_file;
5402        switch (lock->lk_type) {
5403                case NFS4_READ_LT:
5404                case NFS4_READW_LT:
5405                        spin_lock(&fp->fi_lock);
5406                        filp = find_readable_file_locked(fp);
5407                        if (filp)
5408                                get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
5409                        spin_unlock(&fp->fi_lock);
5410                        file_lock->fl_type = F_RDLCK;
5411                        break;
5412                case NFS4_WRITE_LT:
5413                case NFS4_WRITEW_LT:
5414                        spin_lock(&fp->fi_lock);
5415                        filp = find_writeable_file_locked(fp);
5416                        if (filp)
5417                                get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
5418                        spin_unlock(&fp->fi_lock);
5419                        file_lock->fl_type = F_WRLCK;
5420                        break;
5421                default:
5422                        status = nfserr_inval;
5423                goto out;
5424        }
5425        if (!filp) {
5426                status = nfserr_openmode;
5427                goto out;
5428        }
5429
5430        file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(&lock_sop->lo_owner));
5431        file_lock->fl_pid = current->tgid;
5432        file_lock->fl_file = filp;
5433        file_lock->fl_flags = FL_POSIX;
5434        file_lock->fl_lmops = &nfsd_posix_mng_ops;
5435        file_lock->fl_start = lock->lk_offset;
5436        file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
5437        nfs4_transform_lock_offset(file_lock);
5438
5439        conflock = locks_alloc_lock();
5440        if (!conflock) {
5441                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5442                status = nfserr_jukebox;
5443                goto out;
5444        }
5445
5446        err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
5447        switch (-err) {
5448        case 0: /* success! */
5449                update_stateid(&lock_stp->st_stid.sc_stateid);
5450                memcpy(&lock->lk_resp_stateid, &lock_stp->st_stid.sc_stateid, 
5451                                sizeof(stateid_t));
5452                status = 0;
5453                break;
5454        case (EAGAIN):          /* conflock holds conflicting lock */
5455                status = nfserr_denied;
5456                dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
5457                nfs4_set_lock_denied(conflock, &lock->lk_denied);
5458                break;
5459        case (EDEADLK):
5460                status = nfserr_deadlock;
5461                break;
5462        default:
5463                dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
5464                status = nfserrno(err);
5465                break;
5466        }
5467out:
5468        if (filp)
5469                fput(filp);
5470        if (lock_stp) {
5471                /* Bump seqid manually if the 4.0 replay owner is openowner */
5472                if (cstate->replay_owner &&
5473                    cstate->replay_owner != &lock_sop->lo_owner &&
5474                    seqid_mutating_err(ntohl(status)))
5475                        lock_sop->lo_owner.so_seqid++;
5476
5477                /*
5478                 * If this is a new, never-before-used stateid, and we are
5479                 * returning an error, then just go ahead and release it.
5480                 */
5481                if (status && new)
5482                        release_lock_stateid(lock_stp);
5483
5484                nfs4_put_stid(&lock_stp->st_stid);
5485        }
5486        if (open_stp)
5487                nfs4_put_stid(&open_stp->st_stid);
5488        nfsd4_bump_seqid(cstate, status);
5489        if (file_lock)
5490                locks_free_lock(file_lock);
5491        if (conflock)
5492                locks_free_lock(conflock);
5493        return status;
5494}
5495
5496/*
5497 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
5498 * so we do a temporary open here just to get an open file to pass to
5499 * vfs_test_lock.  (Arguably perhaps test_lock should be done with an
5500 * inode operation.)
5501 */
5502static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
5503{
5504        struct file *file;
5505        __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
5506        if (!err) {
5507                err = nfserrno(vfs_test_lock(file, lock));
5508                nfsd_close(file);
5509        }
5510        return err;
5511}
5512
5513/*
5514 * LOCKT operation
5515 */
5516__be32
5517nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5518            struct nfsd4_lockt *lockt)
5519{
5520        struct file_lock *file_lock = NULL;
5521        struct nfs4_lockowner *lo = NULL;
5522        __be32 status;
5523        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5524
5525        if (locks_in_grace(SVC_NET(rqstp)))
5526                return nfserr_grace;
5527
5528        if (check_lock_length(lockt->lt_offset, lockt->lt_length))
5529                 return nfserr_inval;
5530
5531        if (!nfsd4_has_session(cstate)) {
5532                status = lookup_clientid(&lockt->lt_clientid, cstate, nn);
5533                if (status)
5534                        goto out;
5535        }
5536
5537        if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
5538                goto out;
5539
5540        file_lock = locks_alloc_lock();
5541        if (!file_lock) {
5542                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5543                status = nfserr_jukebox;
5544                goto out;
5545        }
5546
5547        switch (lockt->lt_type) {
5548                case NFS4_READ_LT:
5549                case NFS4_READW_LT:
5550                        file_lock->fl_type = F_RDLCK;
5551                break;
5552                case NFS4_WRITE_LT:
5553                case NFS4_WRITEW_LT:
5554                        file_lock->fl_type = F_WRLCK;
5555                break;
5556                default:
5557                        dprintk("NFSD: nfs4_lockt: bad lock type!\n");
5558                        status = nfserr_inval;
5559                goto out;
5560        }
5561
5562        lo = find_lockowner_str(&lockt->lt_clientid, &lockt->lt_owner,
5563                                cstate->clp);
5564        if (lo)
5565                file_lock->fl_owner = (fl_owner_t)lo;
5566        file_lock->fl_pid = current->tgid;
5567        file_lock->fl_flags = FL_POSIX;
5568
5569        file_lock->fl_start = lockt->lt_offset;
5570        file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
5571
5572        nfs4_transform_lock_offset(file_lock);
5573
5574        status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
5575        if (status)
5576                goto out;
5577
5578        if (file_lock->fl_type != F_UNLCK) {
5579                status = nfserr_denied;
5580                nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
5581        }
5582out:
5583        if (lo)
5584                nfs4_put_stateowner(&lo->lo_owner);
5585        if (file_lock)
5586                locks_free_lock(file_lock);
5587        return status;
5588}
5589
5590__be32
5591nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5592            struct nfsd4_locku *locku)
5593{
5594        struct nfs4_ol_stateid *stp;
5595        struct file *filp = NULL;
5596        struct file_lock *file_lock = NULL;
5597        __be32 status;
5598        int err;
5599        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5600
5601        dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
5602                (long long) locku->lu_offset,
5603                (long long) locku->lu_length);
5604
5605        if (check_lock_length(locku->lu_offset, locku->lu_length))
5606                 return nfserr_inval;
5607
5608        status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
5609                                        &locku->lu_stateid, NFS4_LOCK_STID,
5610                                        &stp, nn);
5611        if (status)
5612                goto out;
5613        filp = find_any_file(stp->st_stid.sc_file);
5614        if (!filp) {
5615                status = nfserr_lock_range;
5616                goto put_stateid;
5617        }
5618        file_lock = locks_alloc_lock();
5619        if (!file_lock) {
5620                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5621                status = nfserr_jukebox;
5622                goto fput;
5623        }
5624
5625        file_lock->fl_type = F_UNLCK;
5626        file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(stp->st_stateowner));
5627        file_lock->fl_pid = current->tgid;
5628        file_lock->fl_file = filp;
5629        file_lock->fl_flags = FL_POSIX;
5630        file_lock->fl_lmops = &nfsd_posix_mng_ops;
5631        file_lock->fl_start = locku->lu_offset;
5632
5633        file_lock->fl_end = last_byte_offset(locku->lu_offset,
5634                                                locku->lu_length);
5635        nfs4_transform_lock_offset(file_lock);
5636
5637        err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
5638        if (err) {
5639                dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
5640                goto out_nfserr;
5641        }
5642        update_stateid(&stp->st_stid.sc_stateid);
5643        memcpy(&locku->lu_stateid, &stp->st_stid.sc_stateid, sizeof(stateid_t));
5644fput:
5645        fput(filp);
5646put_stateid:
5647        nfs4_put_stid(&stp->st_stid);
5648out:
5649        nfsd4_bump_seqid(cstate, status);
5650        if (file_lock)
5651                locks_free_lock(file_lock);
5652        return status;
5653
5654out_nfserr:
5655        status = nfserrno(err);
5656        goto fput;
5657}
5658
5659/*
5660 * returns
5661 *      true:  locks held by lockowner
5662 *      false: no locks held by lockowner
5663 */
5664static bool
5665check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
5666{
5667        struct file_lock *fl;
5668        int status = false;
5669        struct file *filp = find_any_file(fp);
5670        struct inode *inode;
5671        struct file_lock_context *flctx;
5672
5673        if (!filp) {
5674                /* Any valid lock stateid should have some sort of access */
5675                WARN_ON_ONCE(1);
5676                return status;
5677        }
5678
5679        inode = file_inode(filp);
5680        flctx = inode->i_flctx;
5681
5682        if (flctx && !list_empty_careful(&flctx->flc_posix)) {
5683                spin_lock(&flctx->flc_lock);
5684                list_for_each_entry(fl, &flctx->flc_posix, fl_list) {
5685                        if (fl->fl_owner == (fl_owner_t)lowner) {
5686                                status = true;
5687                                break;
5688                        }
5689                }
5690                spin_unlock(&flctx->flc_lock);
5691        }
5692        fput(filp);
5693        return status;
5694}
5695
5696__be32
5697nfsd4_release_lockowner(struct svc_rqst *rqstp,
5698                        struct nfsd4_compound_state *cstate,
5699                        struct nfsd4_release_lockowner *rlockowner)
5700{
5701        clientid_t *clid = &rlockowner->rl_clientid;
5702        struct nfs4_stateowner *sop;
5703        struct nfs4_lockowner *lo = NULL;
5704        struct nfs4_ol_stateid *stp;
5705        struct xdr_netobj *owner = &rlockowner->rl_owner;
5706        unsigned int hashval = ownerstr_hashval(owner);
5707        __be32 status;
5708        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5709        struct nfs4_client *clp;
5710
5711        dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
5712                clid->cl_boot, clid->cl_id);
5713
5714        status = lookup_clientid(clid, cstate, nn);
5715        if (status)
5716                return status;
5717
5718        clp = cstate->clp;
5719        /* Find the matching lock stateowner */
5720        spin_lock(&clp->cl_lock);
5721        list_for_each_entry(sop, &clp->cl_ownerstr_hashtbl[hashval],
5722                            so_strhash) {
5723
5724                if (sop->so_is_open_owner || !same_owner_str(sop, owner))
5725                        continue;
5726
5727                /* see if there are still any locks associated with it */
5728                lo = lockowner(sop);
5729                list_for_each_entry(stp, &sop->so_stateids, st_perstateowner) {
5730                        if (check_for_locks(stp->st_stid.sc_file, lo)) {
5731                                status = nfserr_locks_held;
5732                                spin_unlock(&clp->cl_lock);
5733                                return status;
5734                        }
5735                }
5736
5737                nfs4_get_stateowner(sop);
5738                break;
5739        }
5740        spin_unlock(&clp->cl_lock);
5741        if (lo)
5742                release_lockowner(lo);
5743        return status;
5744}
5745
5746static inline struct nfs4_client_reclaim *
5747alloc_reclaim(void)
5748{
5749        return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
5750}
5751
5752bool
5753nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
5754{
5755        struct nfs4_client_reclaim *crp;
5756
5757        crp = nfsd4_find_reclaim_client(name, nn);
5758        return (crp && crp->cr_clp);
5759}
5760
5761/*
5762 * failure => all reset bets are off, nfserr_no_grace...
5763 */
5764struct nfs4_client_reclaim *
5765nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
5766{
5767        unsigned int strhashval;
5768        struct nfs4_client_reclaim *crp;
5769
5770        dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
5771        crp = alloc_reclaim();
5772        if (crp) {
5773                strhashval = clientstr_hashval(name);
5774                INIT_LIST_HEAD(&crp->cr_strhash);
5775                list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
5776                memcpy(crp->cr_recdir, name, HEXDIR_LEN);
5777                crp->cr_clp = NULL;
5778                nn->reclaim_str_hashtbl_size++;
5779        }
5780        return crp;
5781}
5782
5783void
5784nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
5785{
5786        list_del(&crp->cr_strhash);
5787        kfree(crp);
5788        nn->reclaim_str_hashtbl_size--;
5789}
5790
5791void
5792nfs4_release_reclaim(struct nfsd_net *nn)
5793{
5794        struct nfs4_client_reclaim *crp = NULL;
5795        int i;
5796
5797        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
5798                while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
5799                        crp = list_entry(nn->reclaim_str_hashtbl[i].next,
5800                                        struct nfs4_client_reclaim, cr_strhash);
5801                        nfs4_remove_reclaim_record(crp, nn);
5802                }
5803        }
5804        WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
5805}
5806
5807/*
5808 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
5809struct nfs4_client_reclaim *
5810nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
5811{
5812        unsigned int strhashval;
5813        struct nfs4_client_reclaim *crp = NULL;
5814
5815        dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
5816
5817        strhashval = clientstr_hashval(recdir);
5818        list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
5819                if (same_name(crp->cr_recdir, recdir)) {
5820                        return crp;
5821                }
5822        }
5823        return NULL;
5824}
5825
5826/*
5827* Called from OPEN. Look for clientid in reclaim list.
5828*/
5829__be32
5830nfs4_check_open_reclaim(clientid_t *clid,
5831                struct nfsd4_compound_state *cstate,
5832                struct nfsd_net *nn)
5833{
5834        __be32 status;
5835
5836        /* find clientid in conf_id_hashtbl */
5837        status = lookup_clientid(clid, cstate, nn);
5838        if (status)
5839                return nfserr_reclaim_bad;
5840
5841        if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags))
5842                return nfserr_no_grace;
5843
5844        if (nfsd4_client_record_check(cstate->clp))
5845                return nfserr_reclaim_bad;
5846
5847        return nfs_ok;
5848}
5849
5850#ifdef CONFIG_NFSD_FAULT_INJECTION
5851static inline void
5852put_client(struct nfs4_client *clp)
5853{
5854        atomic_dec(&clp->cl_refcount);
5855}
5856
5857static struct nfs4_client *
5858nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
5859{
5860        struct nfs4_client *clp;
5861        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5862                                          nfsd_net_id);
5863
5864        if (!nfsd_netns_ready(nn))
5865                return NULL;
5866
5867        list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5868                if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
5869                        return clp;
5870        }
5871        return NULL;
5872}
5873
5874u64
5875nfsd_inject_print_clients(void)
5876{
5877        struct nfs4_client *clp;
5878        u64 count = 0;
5879        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5880                                          nfsd_net_id);
5881        char buf[INET6_ADDRSTRLEN];
5882
5883        if (!nfsd_netns_ready(nn))
5884                return 0;
5885
5886        spin_lock(&nn->client_lock);
5887        list_for_each_entry(clp, &nn->client_lru, cl_lru) {
5888                rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
5889                pr_info("NFS Client: %s\n", buf);
5890                ++count;
5891        }
5892        spin_unlock(&nn->client_lock);
5893
5894        return count;
5895}
5896
5897u64
5898nfsd_inject_forget_client(struct sockaddr_storage *addr, size_t addr_size)
5899{
5900        u64 count = 0;
5901        struct nfs4_client *clp;
5902        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5903                                          nfsd_net_id);
5904
5905        if (!nfsd_netns_ready(nn))
5906                return count;
5907
5908        spin_lock(&nn->client_lock);
5909        clp = nfsd_find_client(addr, addr_size);
5910        if (clp) {
5911                if (mark_client_expired_locked(clp) == nfs_ok)
5912                        ++count;
5913                else
5914                        clp = NULL;
5915        }
5916        spin_unlock(&nn->client_lock);
5917
5918        if (clp)
5919                expire_client(clp);
5920
5921        return count;
5922}
5923
5924u64
5925nfsd_inject_forget_clients(u64 max)
5926{
5927        u64 count = 0;
5928        struct nfs4_client *clp, *next;
5929        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5930                                                nfsd_net_id);
5931        LIST_HEAD(reaplist);
5932
5933        if (!nfsd_netns_ready(nn))
5934                return count;
5935
5936        spin_lock(&nn->client_lock);
5937        list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
5938                if (mark_client_expired_locked(clp) == nfs_ok) {
5939                        list_add(&clp->cl_lru, &reaplist);
5940                        if (max != 0 && ++count >= max)
5941                                break;
5942                }
5943        }
5944        spin_unlock(&nn->client_lock);
5945
5946        list_for_each_entry_safe(clp, next, &reaplist, cl_lru)
5947                expire_client(clp);
5948
5949        return count;
5950}
5951
5952static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
5953                             const char *type)
5954{
5955        char buf[INET6_ADDRSTRLEN];
5956        rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
5957        printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
5958}
5959
5960static void
5961nfsd_inject_add_lock_to_list(struct nfs4_ol_stateid *lst,
5962                             struct list_head *collect)
5963{
5964        struct nfs4_client *clp = lst->st_stid.sc_client;
5965        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
5966                                          nfsd_net_id);
5967
5968        if (!collect)
5969                return;
5970
5971        lockdep_assert_held(&nn->client_lock);
5972        atomic_inc(&clp->cl_refcount);
5973        list_add(&lst->st_locks, collect);
5974}
5975
5976static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max,
5977                                    struct list_head *collect,
5978                                    void (*func)(struct nfs4_ol_stateid *))
5979{
5980        struct nfs4_openowner *oop;
5981        struct nfs4_ol_stateid *stp, *st_next;
5982        struct nfs4_ol_stateid *lst, *lst_next;
5983        u64 count = 0;
5984
5985        spin_lock(&clp->cl_lock);
5986        list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
5987                list_for_each_entry_safe(stp, st_next,
5988                                &oop->oo_owner.so_stateids, st_perstateowner) {
5989                        list_for_each_entry_safe(lst, lst_next,
5990                                        &stp->st_locks, st_locks) {
5991                                if (func) {
5992                                        func(lst);
5993                                        nfsd_inject_add_lock_to_list(lst,
5994                                                                collect);
5995                                }
5996                                ++count;
5997                                /*
5998                                 * Despite the fact that these functions deal
5999                                 * with 64-bit integers for "count", we must
6000                                 * ensure that it doesn't blow up the
6001                                 * clp->cl_refcount. Throw a warning if we
6002                                 * start to approach INT_MAX here.
6003                                 */
6004                                WARN_ON_ONCE(count == (INT_MAX / 2));
6005                                if (count == max)
6006                                        goto out;
6007                        }
6008                }
6009        }
6010out:
6011        spin_unlock(&clp->cl_lock);
6012
6013        return count;
6014}
6015
6016static u64
6017nfsd_collect_client_locks(struct nfs4_client *clp, struct list_head *collect,
6018                          u64 max)
6019{
6020        return nfsd_foreach_client_lock(clp, max, collect, unhash_lock_stateid);
6021}
6022
6023static u64
6024nfsd_print_client_locks(struct nfs4_client *clp)
6025{
6026        u64 count = nfsd_foreach_client_lock(clp, 0, NULL, NULL);
6027        nfsd_print_count(clp, count, "locked files");
6028        return count;
6029}
6030
6031u64
6032nfsd_inject_print_locks(void)
6033{
6034        struct nfs4_client *clp;
6035        u64 count = 0;
6036        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6037                                                nfsd_net_id);
6038
6039        if (!nfsd_netns_ready(nn))
6040                return 0;
6041
6042        spin_lock(&nn->client_lock);
6043        list_for_each_entry(clp, &nn->client_lru, cl_lru)
6044                count += nfsd_print_client_locks(clp);
6045        spin_unlock(&nn->client_lock);
6046
6047        return count;
6048}
6049
6050static void
6051nfsd_reap_locks(struct list_head *reaplist)
6052{
6053        struct nfs4_client *clp;
6054        struct nfs4_ol_stateid *stp, *next;
6055
6056        list_for_each_entry_safe(stp, next, reaplist, st_locks) {
6057                list_del_init(&stp->st_locks);
6058                clp = stp->st_stid.sc_client;
6059                nfs4_put_stid(&stp->st_stid);
6060                put_client(clp);
6061        }
6062}
6063
6064u64
6065nfsd_inject_forget_client_locks(struct sockaddr_storage *addr, size_t addr_size)
6066{
6067        unsigned int count = 0;
6068        struct nfs4_client *clp;
6069        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6070                                                nfsd_net_id);
6071        LIST_HEAD(reaplist);
6072
6073        if (!nfsd_netns_ready(nn))
6074                return count;
6075
6076        spin_lock(&nn->client_lock);
6077        clp = nfsd_find_client(addr, addr_size);
6078        if (clp)
6079                count = nfsd_collect_client_locks(clp, &reaplist, 0);
6080        spin_unlock(&nn->client_lock);
6081        nfsd_reap_locks(&reaplist);
6082        return count;
6083}
6084
6085u64
6086nfsd_inject_forget_locks(u64 max)
6087{
6088        u64 count = 0;
6089        struct nfs4_client *clp;
6090        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6091                                                nfsd_net_id);
6092        LIST_HEAD(reaplist);
6093
6094        if (!nfsd_netns_ready(nn))
6095                return count;
6096
6097        spin_lock(&nn->client_lock);
6098        list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6099                count += nfsd_collect_client_locks(clp, &reaplist, max - count);
6100                if (max != 0 && count >= max)
6101                        break;
6102        }
6103        spin_unlock(&nn->client_lock);
6104        nfsd_reap_locks(&reaplist);
6105        return count;
6106}
6107
6108static u64
6109nfsd_foreach_client_openowner(struct nfs4_client *clp, u64 max,
6110                              struct list_head *collect,
6111                              void (*func)(struct nfs4_openowner *))
6112{
6113        struct nfs4_openowner *oop, *next;
6114        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6115                                                nfsd_net_id);
6116        u64 count = 0;
6117
6118        lockdep_assert_held(&nn->client_lock);
6119
6120        spin_lock(&clp->cl_lock);
6121        list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
6122                if (func) {
6123                        func(oop);
6124                        if (collect) {
6125                                atomic_inc(&clp->cl_refcount);
6126                                list_add(&oop->oo_perclient, collect);
6127                        }
6128                }
6129                ++count;
6130                /*
6131                 * Despite the fact that these functions deal with
6132                 * 64-bit integers for "count", we must ensure that
6133                 * it doesn't blow up the clp->cl_refcount. Throw a
6134                 * warning if we start to approach INT_MAX here.
6135                 */
6136                WARN_ON_ONCE(count == (INT_MAX / 2));
6137                if (count == max)
6138                        break;
6139        }
6140        spin_unlock(&clp->cl_lock);
6141
6142        return count;
6143}
6144
6145static u64
6146nfsd_print_client_openowners(struct nfs4_client *clp)
6147{
6148        u64 count = nfsd_foreach_client_openowner(clp, 0, NULL, NULL);
6149
6150        nfsd_print_count(clp, count, "openowners");
6151        return count;
6152}
6153
6154static u64
6155nfsd_collect_client_openowners(struct nfs4_client *clp,
6156                               struct list_head *collect, u64 max)
6157{
6158        return nfsd_foreach_client_openowner(clp, max, collect,
6159                                                unhash_openowner_locked);
6160}
6161
6162u64
6163nfsd_inject_print_openowners(void)
6164{
6165        struct nfs4_client *clp;
6166        u64 count = 0;
6167        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6168                                                nfsd_net_id);
6169
6170        if (!nfsd_netns_ready(nn))
6171                return 0;
6172
6173        spin_lock(&nn->client_lock);
6174        list_for_each_entry(clp, &nn->client_lru, cl_lru)
6175                count += nfsd_print_client_openowners(clp);
6176        spin_unlock(&nn->client_lock);
6177
6178        return count;
6179}
6180
6181static void
6182nfsd_reap_openowners(struct list_head *reaplist)
6183{
6184        struct nfs4_client *clp;
6185        struct nfs4_openowner *oop, *next;
6186
6187        list_for_each_entry_safe(oop, next, reaplist, oo_perclient) {
6188                list_del_init(&oop->oo_perclient);
6189                clp = oop->oo_owner.so_client;
6190                release_openowner(oop);
6191                put_client(clp);
6192        }
6193}
6194
6195u64
6196nfsd_inject_forget_client_openowners(struct sockaddr_storage *addr,
6197                                     size_t addr_size)
6198{
6199        unsigned int count = 0;
6200        struct nfs4_client *clp;
6201        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6202                                                nfsd_net_id);
6203        LIST_HEAD(reaplist);
6204
6205        if (!nfsd_netns_ready(nn))
6206                return count;
6207
6208        spin_lock(&nn->client_lock);
6209        clp = nfsd_find_client(addr, addr_size);
6210        if (clp)
6211                count = nfsd_collect_client_openowners(clp, &reaplist, 0);
6212        spin_unlock(&nn->client_lock);
6213        nfsd_reap_openowners(&reaplist);
6214        return count;
6215}
6216
6217u64
6218nfsd_inject_forget_openowners(u64 max)
6219{
6220        u64 count = 0;
6221        struct nfs4_client *clp;
6222        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6223                                                nfsd_net_id);
6224        LIST_HEAD(reaplist);
6225
6226        if (!nfsd_netns_ready(nn))
6227                return count;
6228
6229        spin_lock(&nn->client_lock);
6230        list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6231                count += nfsd_collect_client_openowners(clp, &reaplist,
6232                                                        max - count);
6233                if (max != 0 && count >= max)
6234                        break;
6235        }
6236        spin_unlock(&nn->client_lock);
6237        nfsd_reap_openowners(&reaplist);
6238        return count;
6239}
6240
6241static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
6242                                     struct list_head *victims)
6243{
6244        struct nfs4_delegation *dp, *next;
6245        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6246                                                nfsd_net_id);
6247        u64 count = 0;
6248
6249        lockdep_assert_held(&nn->client_lock);
6250
6251        spin_lock(&state_lock);
6252        list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
6253                if (victims) {
6254                        /*
6255                         * It's not safe to mess with delegations that have a
6256                         * non-zero dl_time. They might have already been broken
6257                         * and could be processed by the laundromat outside of
6258                         * the state_lock. Just leave them be.
6259                         */
6260                        if (dp->dl_time != 0)
6261                                continue;
6262
6263                        atomic_inc(&clp->cl_refcount);
6264                        unhash_delegation_locked(dp);
6265                        list_add(&dp->dl_recall_lru, victims);
6266                }
6267                ++count;
6268                /*
6269                 * Despite the fact that these functions deal with
6270                 * 64-bit integers for "count", we must ensure that
6271                 * it doesn't blow up the clp->cl_refcount. Throw a
6272                 * warning if we start to approach INT_MAX here.
6273                 */
6274                WARN_ON_ONCE(count == (INT_MAX / 2));
6275                if (count == max)
6276                        break;
6277        }
6278        spin_unlock(&state_lock);
6279        return count;
6280}
6281
6282static u64
6283nfsd_print_client_delegations(struct nfs4_client *clp)
6284{
6285        u64 count = nfsd_find_all_delegations(clp, 0, NULL);
6286
6287        nfsd_print_count(clp, count, "delegations");
6288        return count;
6289}
6290
6291u64
6292nfsd_inject_print_delegations(void)
6293{
6294        struct nfs4_client *clp;
6295        u64 count = 0;
6296        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6297                                                nfsd_net_id);
6298
6299        if (!nfsd_netns_ready(nn))
6300                return 0;
6301
6302        spin_lock(&nn->client_lock);
6303        list_for_each_entry(clp, &nn->client_lru, cl_lru)
6304                count += nfsd_print_client_delegations(clp);
6305        spin_unlock(&nn->client_lock);
6306
6307        return count;
6308}
6309
6310static void
6311nfsd_forget_delegations(struct list_head *reaplist)
6312{
6313        struct nfs4_client *clp;
6314        struct nfs4_delegation *dp, *next;
6315
6316        list_for_each_entry_safe(dp, next, reaplist, dl_recall_lru) {
6317                list_del_init(&dp->dl_recall_lru);
6318                clp = dp->dl_stid.sc_client;
6319                revoke_delegation(dp);
6320                put_client(clp);
6321        }
6322}
6323
6324u64
6325nfsd_inject_forget_client_delegations(struct sockaddr_storage *addr,
6326                                      size_t addr_size)
6327{
6328        u64 count = 0;
6329        struct nfs4_client *clp;
6330        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6331                                                nfsd_net_id);
6332        LIST_HEAD(reaplist);
6333
6334        if (!nfsd_netns_ready(nn))
6335                return count;
6336
6337        spin_lock(&nn->client_lock);
6338        clp = nfsd_find_client(addr, addr_size);
6339        if (clp)
6340                count = nfsd_find_all_delegations(clp, 0, &reaplist);
6341        spin_unlock(&nn->client_lock);
6342
6343        nfsd_forget_delegations(&reaplist);
6344        return count;
6345}
6346
6347u64
6348nfsd_inject_forget_delegations(u64 max)
6349{
6350        u64 count = 0;
6351        struct nfs4_client *clp;
6352        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6353                                                nfsd_net_id);
6354        LIST_HEAD(reaplist);
6355
6356        if (!nfsd_netns_ready(nn))
6357                return count;
6358
6359        spin_lock(&nn->client_lock);
6360        list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6361                count += nfsd_find_all_delegations(clp, max - count, &reaplist);
6362                if (max != 0 && count >= max)
6363                        break;
6364        }
6365        spin_unlock(&nn->client_lock);
6366        nfsd_forget_delegations(&reaplist);
6367        return count;
6368}
6369
6370static void
6371nfsd_recall_delegations(struct list_head *reaplist)
6372{
6373        struct nfs4_client *clp;
6374        struct nfs4_delegation *dp, *next;
6375
6376        list_for_each_entry_safe(dp, next, reaplist, dl_recall_lru) {
6377                list_del_init(&dp->dl_recall_lru);
6378                clp = dp->dl_stid.sc_client;
6379                /*
6380                 * We skipped all entries that had a zero dl_time before,
6381                 * so we can now reset the dl_time back to 0. If a delegation
6382                 * break comes in now, then it won't make any difference since
6383                 * we're recalling it either way.
6384                 */
6385                spin_lock(&state_lock);
6386                dp->dl_time = 0;
6387                spin_unlock(&state_lock);
6388                nfsd_break_one_deleg(dp);
6389                put_client(clp);
6390        }
6391}
6392
6393u64
6394nfsd_inject_recall_client_delegations(struct sockaddr_storage *addr,
6395                                      size_t addr_size)
6396{
6397        u64 count = 0;
6398        struct nfs4_client *clp;
6399        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6400                                                nfsd_net_id);
6401        LIST_HEAD(reaplist);
6402
6403        if (!nfsd_netns_ready(nn))
6404                return count;
6405
6406        spin_lock(&nn->client_lock);
6407        clp = nfsd_find_client(addr, addr_size);
6408        if (clp)
6409                count = nfsd_find_all_delegations(clp, 0, &reaplist);
6410        spin_unlock(&nn->client_lock);
6411
6412        nfsd_recall_delegations(&reaplist);
6413        return count;
6414}
6415
6416u64
6417nfsd_inject_recall_delegations(u64 max)
6418{
6419        u64 count = 0;
6420        struct nfs4_client *clp, *next;
6421        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6422                                                nfsd_net_id);
6423        LIST_HEAD(reaplist);
6424
6425        if (!nfsd_netns_ready(nn))
6426                return count;
6427
6428        spin_lock(&nn->client_lock);
6429        list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
6430                count += nfsd_find_all_delegations(clp, max - count, &reaplist);
6431                if (max != 0 && ++count >= max)
6432                        break;
6433        }
6434        spin_unlock(&nn->client_lock);
6435        nfsd_recall_delegations(&reaplist);
6436        return count;
6437}
6438#endif /* CONFIG_NFSD_FAULT_INJECTION */
6439
6440/*
6441 * Since the lifetime of a delegation isn't limited to that of an open, a
6442 * client may quite reasonably hang on to a delegation as long as it has
6443 * the inode cached.  This becomes an obvious problem the first time a
6444 * client's inode cache approaches the size of the server's total memory.
6445 *
6446 * For now we avoid this problem by imposing a hard limit on the number
6447 * of delegations, which varies according to the server's memory size.
6448 */
6449static void
6450set_max_delegations(void)
6451{
6452        /*
6453         * Allow at most 4 delegations per megabyte of RAM.  Quick
6454         * estimates suggest that in the worst case (where every delegation
6455         * is for a different inode), a delegation could take about 1.5K,
6456         * giving a worst case usage of about 6% of memory.
6457         */
6458        max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
6459}
6460
6461static int nfs4_state_create_net(struct net *net)
6462{
6463        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6464        int i;
6465
6466        nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
6467                        CLIENT_HASH_SIZE, GFP_KERNEL);
6468        if (!nn->conf_id_hashtbl)
6469                goto err;
6470        nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
6471                        CLIENT_HASH_SIZE, GFP_KERNEL);
6472        if (!nn->unconf_id_hashtbl)
6473                goto err_unconf_id;
6474        nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
6475                        SESSION_HASH_SIZE, GFP_KERNEL);
6476        if (!nn->sessionid_hashtbl)
6477                goto err_sessionid;
6478
6479        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6480                INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
6481                INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
6482        }
6483        for (i = 0; i < SESSION_HASH_SIZE; i++)
6484                INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
6485        nn->conf_name_tree = RB_ROOT;
6486        nn->unconf_name_tree = RB_ROOT;
6487        INIT_LIST_HEAD(&nn->client_lru);
6488        INIT_LIST_HEAD(&nn->close_lru);
6489        INIT_LIST_HEAD(&nn->del_recall_lru);
6490        spin_lock_init(&nn->client_lock);
6491
6492        INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
6493        get_net(net);
6494
6495        return 0;
6496
6497err_sessionid:
6498        kfree(nn->unconf_id_hashtbl);
6499err_unconf_id:
6500        kfree(nn->conf_id_hashtbl);
6501err:
6502        return -ENOMEM;
6503}
6504
6505static void
6506nfs4_state_destroy_net(struct net *net)
6507{
6508        int i;
6509        struct nfs4_client *clp = NULL;
6510        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6511
6512        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6513                while (!list_empty(&nn->conf_id_hashtbl[i])) {
6514                        clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
6515                        destroy_client(clp);
6516                }
6517        }
6518
6519        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6520                while (!list_empty(&nn->unconf_id_hashtbl[i])) {
6521                        clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
6522                        destroy_client(clp);
6523                }
6524        }
6525
6526        kfree(nn->sessionid_hashtbl);
6527        kfree(nn->unconf_id_hashtbl);
6528        kfree(nn->conf_id_hashtbl);
6529        put_net(net);
6530}
6531
6532int
6533nfs4_state_start_net(struct net *net)
6534{
6535        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6536        int ret;
6537
6538        ret = nfs4_state_create_net(net);
6539        if (ret)
6540                return ret;
6541        nn->boot_time = get_seconds();
6542        nn->grace_ended = false;
6543        locks_start_grace(net, &nn->nfsd4_manager);
6544        nfsd4_client_tracking_init(net);
6545        printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
6546               nn->nfsd4_grace, net);
6547        queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
6548        return 0;
6549}
6550
6551/* initialization to perform when the nfsd service is started: */
6552
6553int
6554nfs4_state_start(void)
6555{
6556        int ret;
6557
6558        ret = set_callback_cred();
6559        if (ret)
6560                return -ENOMEM;
6561        laundry_wq = create_singlethread_workqueue("nfsd4");
6562        if (laundry_wq == NULL) {
6563                ret = -ENOMEM;
6564                goto out_recovery;
6565        }
6566        ret = nfsd4_create_callback_queue();
6567        if (ret)
6568                goto out_free_laundry;
6569
6570        set_max_delegations();
6571
6572        return 0;
6573
6574out_free_laundry:
6575        destroy_workqueue(laundry_wq);
6576out_recovery:
6577        return ret;
6578}
6579
6580void
6581nfs4_state_shutdown_net(struct net *net)
6582{
6583        struct nfs4_delegation *dp = NULL;
6584        struct list_head *pos, *next, reaplist;
6585        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6586
6587        cancel_delayed_work_sync(&nn->laundromat_work);
6588        locks_end_grace(&nn->nfsd4_manager);
6589
6590        INIT_LIST_HEAD(&reaplist);
6591        spin_lock(&state_lock);
6592        list_for_each_safe(pos, next, &nn->del_recall_lru) {
6593                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
6594                unhash_delegation_locked(dp);
6595                list_add(&dp->dl_recall_lru, &reaplist);
6596        }
6597        spin_unlock(&state_lock);
6598        list_for_each_safe(pos, next, &reaplist) {
6599                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
6600                list_del_init(&dp->dl_recall_lru);
6601                put_clnt_odstate(dp->dl_clnt_odstate);
6602                nfs4_put_deleg_lease(dp->dl_stid.sc_file);
6603                nfs4_put_stid(&dp->dl_stid);
6604        }
6605
6606        nfsd4_client_tracking_exit(net);
6607        nfs4_state_destroy_net(net);
6608}
6609
6610void
6611nfs4_state_shutdown(void)
6612{
6613        destroy_workqueue(laundry_wq);
6614        nfsd4_destroy_callback_queue();
6615}
6616
6617static void
6618get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
6619{
6620        if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
6621                memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
6622}
6623
6624static void
6625put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
6626{
6627        if (cstate->minorversion) {
6628                memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
6629                SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
6630        }
6631}
6632
6633void
6634clear_current_stateid(struct nfsd4_compound_state *cstate)
6635{
6636        CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
6637}
6638
6639/*
6640 * functions to set current state id
6641 */
6642void
6643nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
6644{
6645        put_stateid(cstate, &odp->od_stateid);
6646}
6647
6648void
6649nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
6650{
6651        put_stateid(cstate, &open->op_stateid);
6652}
6653
6654void
6655nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
6656{
6657        put_stateid(cstate, &close->cl_stateid);
6658}
6659
6660void
6661nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
6662{
6663        put_stateid(cstate, &lock->lk_resp_stateid);
6664}
6665
6666/*
6667 * functions to consume current state id
6668 */
6669
6670void
6671nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
6672{
6673        get_stateid(cstate, &odp->od_stateid);
6674}
6675
6676void
6677nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
6678{
6679        get_stateid(cstate, &drp->dr_stateid);
6680}
6681
6682void
6683nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
6684{
6685        get_stateid(cstate, &fsp->fr_stateid);
6686}
6687
6688void
6689nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
6690{
6691        get_stateid(cstate, &setattr->sa_stateid);
6692}
6693
6694void
6695nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
6696{
6697        get_stateid(cstate, &close->cl_stateid);
6698}
6699
6700void
6701nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
6702{
6703        get_stateid(cstate, &locku->lu_stateid);
6704}
6705
6706void
6707nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
6708{
6709        get_stateid(cstate, &read->rd_stateid);
6710}
6711
6712void
6713nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
6714{
6715        get_stateid(cstate, &write->wr_stateid);
6716}
6717