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 const 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        spin_lock_init(&stid->sc_lock);
 579
 580        /*
 581         * It shouldn't be a problem to reuse an opaque stateid value.
 582         * I don't think it is for 4.1.  But with 4.0 I worry that, for
 583         * example, a stray write retransmission could be accepted by
 584         * the server when it should have been rejected.  Therefore,
 585         * adopt a trick from the sctp code to attempt to maximize the
 586         * amount of time until an id is reused, by ensuring they always
 587         * "increase" (mod INT_MAX):
 588         */
 589        return stid;
 590out_free:
 591        kmem_cache_free(slab, stid);
 592        return NULL;
 593}
 594
 595static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
 596{
 597        struct nfs4_stid *stid;
 598        struct nfs4_ol_stateid *stp;
 599
 600        stid = nfs4_alloc_stid(clp, stateid_slab);
 601        if (!stid)
 602                return NULL;
 603
 604        stp = openlockstateid(stid);
 605        stp->st_stid.sc_free = nfs4_free_ol_stateid;
 606        return stp;
 607}
 608
 609static void nfs4_free_deleg(struct nfs4_stid *stid)
 610{
 611        kmem_cache_free(deleg_slab, stid);
 612        atomic_long_dec(&num_delegations);
 613}
 614
 615/*
 616 * When we recall a delegation, we should be careful not to hand it
 617 * out again straight away.
 618 * To ensure this we keep a pair of bloom filters ('new' and 'old')
 619 * in which the filehandles of recalled delegations are "stored".
 620 * If a filehandle appear in either filter, a delegation is blocked.
 621 * When a delegation is recalled, the filehandle is stored in the "new"
 622 * filter.
 623 * Every 30 seconds we swap the filters and clear the "new" one,
 624 * unless both are empty of course.
 625 *
 626 * Each filter is 256 bits.  We hash the filehandle to 32bit and use the
 627 * low 3 bytes as hash-table indices.
 628 *
 629 * 'blocked_delegations_lock', which is always taken in block_delegations(),
 630 * is used to manage concurrent access.  Testing does not need the lock
 631 * except when swapping the two filters.
 632 */
 633static DEFINE_SPINLOCK(blocked_delegations_lock);
 634static struct bloom_pair {
 635        int     entries, old_entries;
 636        time_t  swap_time;
 637        int     new; /* index into 'set' */
 638        DECLARE_BITMAP(set[2], 256);
 639} blocked_delegations;
 640
 641static int delegation_blocked(struct knfsd_fh *fh)
 642{
 643        u32 hash;
 644        struct bloom_pair *bd = &blocked_delegations;
 645
 646        if (bd->entries == 0)
 647                return 0;
 648        if (seconds_since_boot() - bd->swap_time > 30) {
 649                spin_lock(&blocked_delegations_lock);
 650                if (seconds_since_boot() - bd->swap_time > 30) {
 651                        bd->entries -= bd->old_entries;
 652                        bd->old_entries = bd->entries;
 653                        memset(bd->set[bd->new], 0,
 654                               sizeof(bd->set[0]));
 655                        bd->new = 1-bd->new;
 656                        bd->swap_time = seconds_since_boot();
 657                }
 658                spin_unlock(&blocked_delegations_lock);
 659        }
 660        hash = jhash(&fh->fh_base, fh->fh_size, 0);
 661        if (test_bit(hash&255, bd->set[0]) &&
 662            test_bit((hash>>8)&255, bd->set[0]) &&
 663            test_bit((hash>>16)&255, bd->set[0]))
 664                return 1;
 665
 666        if (test_bit(hash&255, bd->set[1]) &&
 667            test_bit((hash>>8)&255, bd->set[1]) &&
 668            test_bit((hash>>16)&255, bd->set[1]))
 669                return 1;
 670
 671        return 0;
 672}
 673
 674static void block_delegations(struct knfsd_fh *fh)
 675{
 676        u32 hash;
 677        struct bloom_pair *bd = &blocked_delegations;
 678
 679        hash = jhash(&fh->fh_base, fh->fh_size, 0);
 680
 681        spin_lock(&blocked_delegations_lock);
 682        __set_bit(hash&255, bd->set[bd->new]);
 683        __set_bit((hash>>8)&255, bd->set[bd->new]);
 684        __set_bit((hash>>16)&255, bd->set[bd->new]);
 685        if (bd->entries == 0)
 686                bd->swap_time = seconds_since_boot();
 687        bd->entries += 1;
 688        spin_unlock(&blocked_delegations_lock);
 689}
 690
 691static struct nfs4_delegation *
 692alloc_init_deleg(struct nfs4_client *clp, struct svc_fh *current_fh,
 693                 struct nfs4_clnt_odstate *odstate)
 694{
 695        struct nfs4_delegation *dp;
 696        long n;
 697
 698        dprintk("NFSD alloc_init_deleg\n");
 699        n = atomic_long_inc_return(&num_delegations);
 700        if (n < 0 || n > max_delegations)
 701                goto out_dec;
 702        if (delegation_blocked(&current_fh->fh_handle))
 703                goto out_dec;
 704        dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab));
 705        if (dp == NULL)
 706                goto out_dec;
 707
 708        dp->dl_stid.sc_free = nfs4_free_deleg;
 709        /*
 710         * delegation seqid's are never incremented.  The 4.1 special
 711         * meaning of seqid 0 isn't meaningful, really, but let's avoid
 712         * 0 anyway just for consistency and use 1:
 713         */
 714        dp->dl_stid.sc_stateid.si_generation = 1;
 715        INIT_LIST_HEAD(&dp->dl_perfile);
 716        INIT_LIST_HEAD(&dp->dl_perclnt);
 717        INIT_LIST_HEAD(&dp->dl_recall_lru);
 718        dp->dl_clnt_odstate = odstate;
 719        get_clnt_odstate(odstate);
 720        dp->dl_type = NFS4_OPEN_DELEGATE_READ;
 721        dp->dl_retries = 1;
 722        nfsd4_init_cb(&dp->dl_recall, dp->dl_stid.sc_client,
 723                      &nfsd4_cb_recall_ops, NFSPROC4_CLNT_CB_RECALL);
 724        return dp;
 725out_dec:
 726        atomic_long_dec(&num_delegations);
 727        return NULL;
 728}
 729
 730void
 731nfs4_put_stid(struct nfs4_stid *s)
 732{
 733        struct nfs4_file *fp = s->sc_file;
 734        struct nfs4_client *clp = s->sc_client;
 735
 736        might_lock(&clp->cl_lock);
 737
 738        if (!atomic_dec_and_lock(&s->sc_count, &clp->cl_lock)) {
 739                wake_up_all(&close_wq);
 740                return;
 741        }
 742        idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
 743        spin_unlock(&clp->cl_lock);
 744        s->sc_free(s);
 745        if (fp)
 746                put_nfs4_file(fp);
 747}
 748
 749void
 750nfs4_inc_and_copy_stateid(stateid_t *dst, struct nfs4_stid *stid)
 751{
 752        stateid_t *src = &stid->sc_stateid;
 753
 754        spin_lock(&stid->sc_lock);
 755        if (unlikely(++src->si_generation == 0))
 756                src->si_generation = 1;
 757        memcpy(dst, src, sizeof(*dst));
 758        spin_unlock(&stid->sc_lock);
 759}
 760
 761static void nfs4_put_deleg_lease(struct nfs4_file *fp)
 762{
 763        struct file *filp = NULL;
 764
 765        spin_lock(&fp->fi_lock);
 766        if (fp->fi_deleg_file && --fp->fi_delegees == 0)
 767                swap(filp, fp->fi_deleg_file);
 768        spin_unlock(&fp->fi_lock);
 769
 770        if (filp) {
 771                vfs_setlease(filp, F_UNLCK, NULL, (void **)&fp);
 772                fput(filp);
 773        }
 774}
 775
 776void nfs4_unhash_stid(struct nfs4_stid *s)
 777{
 778        s->sc_type = 0;
 779}
 780
 781/**
 782 * nfs4_get_existing_delegation - Discover if this delegation already exists
 783 * @clp:     a pointer to the nfs4_client we're granting a delegation to
 784 * @fp:      a pointer to the nfs4_file we're granting a delegation on
 785 *
 786 * Return:
 787 *      On success: NULL if an existing delegation was not found.
 788 *
 789 *      On error: -EAGAIN if one was previously granted to this nfs4_client
 790 *                 for this nfs4_file.
 791 *
 792 */
 793
 794static int
 795nfs4_get_existing_delegation(struct nfs4_client *clp, struct nfs4_file *fp)
 796{
 797        struct nfs4_delegation *searchdp = NULL;
 798        struct nfs4_client *searchclp = NULL;
 799
 800        lockdep_assert_held(&state_lock);
 801        lockdep_assert_held(&fp->fi_lock);
 802
 803        list_for_each_entry(searchdp, &fp->fi_delegations, dl_perfile) {
 804                searchclp = searchdp->dl_stid.sc_client;
 805                if (clp == searchclp) {
 806                        return -EAGAIN;
 807                }
 808        }
 809        return 0;
 810}
 811
 812/**
 813 * hash_delegation_locked - Add a delegation to the appropriate lists
 814 * @dp:     a pointer to the nfs4_delegation we are adding.
 815 * @fp:     a pointer to the nfs4_file we're granting a delegation on
 816 *
 817 * Return:
 818 *      On success: NULL if the delegation was successfully hashed.
 819 *
 820 *      On error: -EAGAIN if one was previously granted to this
 821 *                 nfs4_client for this nfs4_file. Delegation is not hashed.
 822 *
 823 */
 824
 825static int
 826hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
 827{
 828        int status;
 829        struct nfs4_client *clp = dp->dl_stid.sc_client;
 830
 831        lockdep_assert_held(&state_lock);
 832        lockdep_assert_held(&fp->fi_lock);
 833
 834        status = nfs4_get_existing_delegation(clp, fp);
 835        if (status)
 836                return status;
 837        ++fp->fi_delegees;
 838        atomic_inc(&dp->dl_stid.sc_count);
 839        dp->dl_stid.sc_type = NFS4_DELEG_STID;
 840        list_add(&dp->dl_perfile, &fp->fi_delegations);
 841        list_add(&dp->dl_perclnt, &clp->cl_delegations);
 842        return 0;
 843}
 844
 845static bool
 846unhash_delegation_locked(struct nfs4_delegation *dp)
 847{
 848        struct nfs4_file *fp = dp->dl_stid.sc_file;
 849
 850        lockdep_assert_held(&state_lock);
 851
 852        if (list_empty(&dp->dl_perfile))
 853                return false;
 854
 855        dp->dl_stid.sc_type = NFS4_CLOSED_DELEG_STID;
 856        /* Ensure that deleg break won't try to requeue it */
 857        ++dp->dl_time;
 858        spin_lock(&fp->fi_lock);
 859        list_del_init(&dp->dl_perclnt);
 860        list_del_init(&dp->dl_recall_lru);
 861        list_del_init(&dp->dl_perfile);
 862        spin_unlock(&fp->fi_lock);
 863        return true;
 864}
 865
 866static void destroy_delegation(struct nfs4_delegation *dp)
 867{
 868        bool unhashed;
 869
 870        spin_lock(&state_lock);
 871        unhashed = unhash_delegation_locked(dp);
 872        spin_unlock(&state_lock);
 873        if (unhashed) {
 874                put_clnt_odstate(dp->dl_clnt_odstate);
 875                nfs4_put_deleg_lease(dp->dl_stid.sc_file);
 876                nfs4_put_stid(&dp->dl_stid);
 877        }
 878}
 879
 880static void revoke_delegation(struct nfs4_delegation *dp)
 881{
 882        struct nfs4_client *clp = dp->dl_stid.sc_client;
 883
 884        WARN_ON(!list_empty(&dp->dl_recall_lru));
 885
 886        put_clnt_odstate(dp->dl_clnt_odstate);
 887        nfs4_put_deleg_lease(dp->dl_stid.sc_file);
 888
 889        if (clp->cl_minorversion == 0)
 890                nfs4_put_stid(&dp->dl_stid);
 891        else {
 892                dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
 893                spin_lock(&clp->cl_lock);
 894                list_add(&dp->dl_recall_lru, &clp->cl_revoked);
 895                spin_unlock(&clp->cl_lock);
 896        }
 897}
 898
 899/* 
 900 * SETCLIENTID state 
 901 */
 902
 903static unsigned int clientid_hashval(u32 id)
 904{
 905        return id & CLIENT_HASH_MASK;
 906}
 907
 908static unsigned int clientstr_hashval(const char *name)
 909{
 910        return opaque_hashval(name, 8) & CLIENT_HASH_MASK;
 911}
 912
 913/*
 914 * We store the NONE, READ, WRITE, and BOTH bits separately in the
 915 * st_{access,deny}_bmap field of the stateid, in order to track not
 916 * only what share bits are currently in force, but also what
 917 * combinations of share bits previous opens have used.  This allows us
 918 * to enforce the recommendation of rfc 3530 14.2.19 that the server
 919 * return an error if the client attempt to downgrade to a combination
 920 * of share bits not explicable by closing some of its previous opens.
 921 *
 922 * XXX: This enforcement is actually incomplete, since we don't keep
 923 * track of access/deny bit combinations; so, e.g., we allow:
 924 *
 925 *      OPEN allow read, deny write
 926 *      OPEN allow both, deny none
 927 *      DOWNGRADE allow read, deny none
 928 *
 929 * which we should reject.
 930 */
 931static unsigned int
 932bmap_to_share_mode(unsigned long bmap) {
 933        int i;
 934        unsigned int access = 0;
 935
 936        for (i = 1; i < 4; i++) {
 937                if (test_bit(i, &bmap))
 938                        access |= i;
 939        }
 940        return access;
 941}
 942
 943/* set share access for a given stateid */
 944static inline void
 945set_access(u32 access, struct nfs4_ol_stateid *stp)
 946{
 947        unsigned char mask = 1 << access;
 948
 949        WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
 950        stp->st_access_bmap |= mask;
 951}
 952
 953/* clear share access for a given stateid */
 954static inline void
 955clear_access(u32 access, struct nfs4_ol_stateid *stp)
 956{
 957        unsigned char mask = 1 << access;
 958
 959        WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
 960        stp->st_access_bmap &= ~mask;
 961}
 962
 963/* test whether a given stateid has access */
 964static inline bool
 965test_access(u32 access, struct nfs4_ol_stateid *stp)
 966{
 967        unsigned char mask = 1 << access;
 968
 969        return (bool)(stp->st_access_bmap & mask);
 970}
 971
 972/* set share deny for a given stateid */
 973static inline void
 974set_deny(u32 deny, struct nfs4_ol_stateid *stp)
 975{
 976        unsigned char mask = 1 << deny;
 977
 978        WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
 979        stp->st_deny_bmap |= mask;
 980}
 981
 982/* clear share deny for a given stateid */
 983static inline void
 984clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
 985{
 986        unsigned char mask = 1 << deny;
 987
 988        WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
 989        stp->st_deny_bmap &= ~mask;
 990}
 991
 992/* test whether a given stateid is denying specific access */
 993static inline bool
 994test_deny(u32 deny, struct nfs4_ol_stateid *stp)
 995{
 996        unsigned char mask = 1 << deny;
 997
 998        return (bool)(stp->st_deny_bmap & mask);
 999}
1000
1001static int nfs4_access_to_omode(u32 access)
1002{
1003        switch (access & NFS4_SHARE_ACCESS_BOTH) {
1004        case NFS4_SHARE_ACCESS_READ:
1005                return O_RDONLY;
1006        case NFS4_SHARE_ACCESS_WRITE:
1007                return O_WRONLY;
1008        case NFS4_SHARE_ACCESS_BOTH:
1009                return O_RDWR;
1010        }
1011        WARN_ON_ONCE(1);
1012        return O_RDONLY;
1013}
1014
1015/*
1016 * A stateid that had a deny mode associated with it is being released
1017 * or downgraded. Recalculate the deny mode on the file.
1018 */
1019static void
1020recalculate_deny_mode(struct nfs4_file *fp)
1021{
1022        struct nfs4_ol_stateid *stp;
1023
1024        spin_lock(&fp->fi_lock);
1025        fp->fi_share_deny = 0;
1026        list_for_each_entry(stp, &fp->fi_stateids, st_perfile)
1027                fp->fi_share_deny |= bmap_to_share_mode(stp->st_deny_bmap);
1028        spin_unlock(&fp->fi_lock);
1029}
1030
1031static void
1032reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
1033{
1034        int i;
1035        bool change = false;
1036
1037        for (i = 1; i < 4; i++) {
1038                if ((i & deny) != i) {
1039                        change = true;
1040                        clear_deny(i, stp);
1041                }
1042        }
1043
1044        /* Recalculate per-file deny mode if there was a change */
1045        if (change)
1046                recalculate_deny_mode(stp->st_stid.sc_file);
1047}
1048
1049/* release all access and file references for a given stateid */
1050static void
1051release_all_access(struct nfs4_ol_stateid *stp)
1052{
1053        int i;
1054        struct nfs4_file *fp = stp->st_stid.sc_file;
1055
1056        if (fp && stp->st_deny_bmap != 0)
1057                recalculate_deny_mode(fp);
1058
1059        for (i = 1; i < 4; i++) {
1060                if (test_access(i, stp))
1061                        nfs4_file_put_access(stp->st_stid.sc_file, i);
1062                clear_access(i, stp);
1063        }
1064}
1065
1066static inline void nfs4_free_stateowner(struct nfs4_stateowner *sop)
1067{
1068        kfree(sop->so_owner.data);
1069        sop->so_ops->so_free(sop);
1070}
1071
1072static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
1073{
1074        struct nfs4_client *clp = sop->so_client;
1075
1076        might_lock(&clp->cl_lock);
1077
1078        if (!atomic_dec_and_lock(&sop->so_count, &clp->cl_lock))
1079                return;
1080        sop->so_ops->so_unhash(sop);
1081        spin_unlock(&clp->cl_lock);
1082        nfs4_free_stateowner(sop);
1083}
1084
1085static bool unhash_ol_stateid(struct nfs4_ol_stateid *stp)
1086{
1087        struct nfs4_file *fp = stp->st_stid.sc_file;
1088
1089        lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
1090
1091        if (list_empty(&stp->st_perfile))
1092                return false;
1093
1094        spin_lock(&fp->fi_lock);
1095        list_del_init(&stp->st_perfile);
1096        spin_unlock(&fp->fi_lock);
1097        list_del(&stp->st_perstateowner);
1098        return true;
1099}
1100
1101static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
1102{
1103        struct nfs4_ol_stateid *stp = openlockstateid(stid);
1104
1105        put_clnt_odstate(stp->st_clnt_odstate);
1106        release_all_access(stp);
1107        if (stp->st_stateowner)
1108                nfs4_put_stateowner(stp->st_stateowner);
1109        kmem_cache_free(stateid_slab, stid);
1110}
1111
1112static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
1113{
1114        struct nfs4_ol_stateid *stp = openlockstateid(stid);
1115        struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
1116        struct file *file;
1117
1118        file = find_any_file(stp->st_stid.sc_file);
1119        if (file)
1120                filp_close(file, (fl_owner_t)lo);
1121        nfs4_free_ol_stateid(stid);
1122}
1123
1124/*
1125 * Put the persistent reference to an already unhashed generic stateid, while
1126 * holding the cl_lock. If it's the last reference, then put it onto the
1127 * reaplist for later destruction.
1128 */
1129static void put_ol_stateid_locked(struct nfs4_ol_stateid *stp,
1130                                       struct list_head *reaplist)
1131{
1132        struct nfs4_stid *s = &stp->st_stid;
1133        struct nfs4_client *clp = s->sc_client;
1134
1135        lockdep_assert_held(&clp->cl_lock);
1136
1137        WARN_ON_ONCE(!list_empty(&stp->st_locks));
1138
1139        if (!atomic_dec_and_test(&s->sc_count)) {
1140                wake_up_all(&close_wq);
1141                return;
1142        }
1143
1144        idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1145        list_add(&stp->st_locks, reaplist);
1146}
1147
1148static bool unhash_lock_stateid(struct nfs4_ol_stateid *stp)
1149{
1150        struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
1151
1152        lockdep_assert_held(&oo->oo_owner.so_client->cl_lock);
1153
1154        list_del_init(&stp->st_locks);
1155        nfs4_unhash_stid(&stp->st_stid);
1156        return unhash_ol_stateid(stp);
1157}
1158
1159static void release_lock_stateid(struct nfs4_ol_stateid *stp)
1160{
1161        struct nfs4_openowner *oo = openowner(stp->st_openstp->st_stateowner);
1162        bool unhashed;
1163
1164        spin_lock(&oo->oo_owner.so_client->cl_lock);
1165        unhashed = unhash_lock_stateid(stp);
1166        spin_unlock(&oo->oo_owner.so_client->cl_lock);
1167        if (unhashed)
1168                nfs4_put_stid(&stp->st_stid);
1169}
1170
1171static void unhash_lockowner_locked(struct nfs4_lockowner *lo)
1172{
1173        struct nfs4_client *clp = lo->lo_owner.so_client;
1174
1175        lockdep_assert_held(&clp->cl_lock);
1176
1177        list_del_init(&lo->lo_owner.so_strhash);
1178}
1179
1180/*
1181 * Free a list of generic stateids that were collected earlier after being
1182 * fully unhashed.
1183 */
1184static void
1185free_ol_stateid_reaplist(struct list_head *reaplist)
1186{
1187        struct nfs4_ol_stateid *stp;
1188        struct nfs4_file *fp;
1189
1190        might_sleep();
1191
1192        while (!list_empty(reaplist)) {
1193                stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
1194                                       st_locks);
1195                list_del(&stp->st_locks);
1196                fp = stp->st_stid.sc_file;
1197                stp->st_stid.sc_free(&stp->st_stid);
1198                if (fp)
1199                        put_nfs4_file(fp);
1200        }
1201}
1202
1203static void release_lockowner(struct nfs4_lockowner *lo)
1204{
1205        struct nfs4_client *clp = lo->lo_owner.so_client;
1206        struct nfs4_ol_stateid *stp;
1207        struct list_head reaplist;
1208
1209        INIT_LIST_HEAD(&reaplist);
1210
1211        spin_lock(&clp->cl_lock);
1212        unhash_lockowner_locked(lo);
1213        while (!list_empty(&lo->lo_owner.so_stateids)) {
1214                stp = list_first_entry(&lo->lo_owner.so_stateids,
1215                                struct nfs4_ol_stateid, st_perstateowner);
1216                WARN_ON(!unhash_lock_stateid(stp));
1217                put_ol_stateid_locked(stp, &reaplist);
1218        }
1219        spin_unlock(&clp->cl_lock);
1220        free_ol_stateid_reaplist(&reaplist);
1221        nfs4_put_stateowner(&lo->lo_owner);
1222}
1223
1224static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp,
1225                                       struct list_head *reaplist)
1226{
1227        struct nfs4_ol_stateid *stp;
1228
1229        lockdep_assert_held(&open_stp->st_stid.sc_client->cl_lock);
1230
1231        while (!list_empty(&open_stp->st_locks)) {
1232                stp = list_entry(open_stp->st_locks.next,
1233                                struct nfs4_ol_stateid, st_locks);
1234                WARN_ON(!unhash_lock_stateid(stp));
1235                put_ol_stateid_locked(stp, reaplist);
1236        }
1237}
1238
1239static bool unhash_open_stateid(struct nfs4_ol_stateid *stp,
1240                                struct list_head *reaplist)
1241{
1242        bool unhashed;
1243
1244        lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1245
1246        unhashed = unhash_ol_stateid(stp);
1247        release_open_stateid_locks(stp, reaplist);
1248        return unhashed;
1249}
1250
1251static void release_open_stateid(struct nfs4_ol_stateid *stp)
1252{
1253        LIST_HEAD(reaplist);
1254
1255        spin_lock(&stp->st_stid.sc_client->cl_lock);
1256        if (unhash_open_stateid(stp, &reaplist))
1257                put_ol_stateid_locked(stp, &reaplist);
1258        spin_unlock(&stp->st_stid.sc_client->cl_lock);
1259        free_ol_stateid_reaplist(&reaplist);
1260}
1261
1262static void unhash_openowner_locked(struct nfs4_openowner *oo)
1263{
1264        struct nfs4_client *clp = oo->oo_owner.so_client;
1265
1266        lockdep_assert_held(&clp->cl_lock);
1267
1268        list_del_init(&oo->oo_owner.so_strhash);
1269        list_del_init(&oo->oo_perclient);
1270}
1271
1272static void release_last_closed_stateid(struct nfs4_openowner *oo)
1273{
1274        struct nfsd_net *nn = net_generic(oo->oo_owner.so_client->net,
1275                                          nfsd_net_id);
1276        struct nfs4_ol_stateid *s;
1277
1278        spin_lock(&nn->client_lock);
1279        s = oo->oo_last_closed_stid;
1280        if (s) {
1281                list_del_init(&oo->oo_close_lru);
1282                oo->oo_last_closed_stid = NULL;
1283        }
1284        spin_unlock(&nn->client_lock);
1285        if (s)
1286                nfs4_put_stid(&s->st_stid);
1287}
1288
1289static void release_openowner(struct nfs4_openowner *oo)
1290{
1291        struct nfs4_ol_stateid *stp;
1292        struct nfs4_client *clp = oo->oo_owner.so_client;
1293        struct list_head reaplist;
1294
1295        INIT_LIST_HEAD(&reaplist);
1296
1297        spin_lock(&clp->cl_lock);
1298        unhash_openowner_locked(oo);
1299        while (!list_empty(&oo->oo_owner.so_stateids)) {
1300                stp = list_first_entry(&oo->oo_owner.so_stateids,
1301                                struct nfs4_ol_stateid, st_perstateowner);
1302                if (unhash_open_stateid(stp, &reaplist))
1303                        put_ol_stateid_locked(stp, &reaplist);
1304        }
1305        spin_unlock(&clp->cl_lock);
1306        free_ol_stateid_reaplist(&reaplist);
1307        release_last_closed_stateid(oo);
1308        nfs4_put_stateowner(&oo->oo_owner);
1309}
1310
1311static inline int
1312hash_sessionid(struct nfs4_sessionid *sessionid)
1313{
1314        struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1315
1316        return sid->sequence % SESSION_HASH_SIZE;
1317}
1318
1319#ifdef CONFIG_SUNRPC_DEBUG
1320static inline void
1321dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1322{
1323        u32 *ptr = (u32 *)(&sessionid->data[0]);
1324        dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1325}
1326#else
1327static inline void
1328dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1329{
1330}
1331#endif
1332
1333/*
1334 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1335 * won't be used for replay.
1336 */
1337void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1338{
1339        struct nfs4_stateowner *so = cstate->replay_owner;
1340
1341        if (nfserr == nfserr_replay_me)
1342                return;
1343
1344        if (!seqid_mutating_err(ntohl(nfserr))) {
1345                nfsd4_cstate_clear_replay(cstate);
1346                return;
1347        }
1348        if (!so)
1349                return;
1350        if (so->so_is_open_owner)
1351                release_last_closed_stateid(openowner(so));
1352        so->so_seqid++;
1353        return;
1354}
1355
1356static void
1357gen_sessionid(struct nfsd4_session *ses)
1358{
1359        struct nfs4_client *clp = ses->se_client;
1360        struct nfsd4_sessionid *sid;
1361
1362        sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1363        sid->clientid = clp->cl_clientid;
1364        sid->sequence = current_sessionid++;
1365        sid->reserved = 0;
1366}
1367
1368/*
1369 * The protocol defines ca_maxresponssize_cached to include the size of
1370 * the rpc header, but all we need to cache is the data starting after
1371 * the end of the initial SEQUENCE operation--the rest we regenerate
1372 * each time.  Therefore we can advertise a ca_maxresponssize_cached
1373 * value that is the number of bytes in our cache plus a few additional
1374 * bytes.  In order to stay on the safe side, and not promise more than
1375 * we can cache, those additional bytes must be the minimum possible: 24
1376 * bytes of rpc header (xid through accept state, with AUTH_NULL
1377 * verifier), 12 for the compound header (with zero-length tag), and 44
1378 * for the SEQUENCE op response:
1379 */
1380#define NFSD_MIN_HDR_SEQ_SZ  (24 + 12 + 44)
1381
1382static void
1383free_session_slots(struct nfsd4_session *ses)
1384{
1385        int i;
1386
1387        for (i = 0; i < ses->se_fchannel.maxreqs; i++)
1388                kfree(ses->se_slots[i]);
1389}
1390
1391/*
1392 * We don't actually need to cache the rpc and session headers, so we
1393 * can allocate a little less for each slot:
1394 */
1395static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
1396{
1397        u32 size;
1398
1399        if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
1400                size = 0;
1401        else
1402                size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
1403        return size + sizeof(struct nfsd4_slot);
1404}
1405
1406/*
1407 * XXX: If we run out of reserved DRC memory we could (up to a point)
1408 * re-negotiate active sessions and reduce their slot usage to make
1409 * room for new connections. For now we just fail the create session.
1410 */
1411static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca)
1412{
1413        u32 slotsize = slot_bytes(ca);
1414        u32 num = ca->maxreqs;
1415        int avail;
1416
1417        spin_lock(&nfsd_drc_lock);
1418        avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION,
1419                    nfsd_drc_max_mem - nfsd_drc_mem_used);
1420        num = min_t(int, num, avail / slotsize);
1421        nfsd_drc_mem_used += num * slotsize;
1422        spin_unlock(&nfsd_drc_lock);
1423
1424        return num;
1425}
1426
1427static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
1428{
1429        int slotsize = slot_bytes(ca);
1430
1431        spin_lock(&nfsd_drc_lock);
1432        nfsd_drc_mem_used -= slotsize * ca->maxreqs;
1433        spin_unlock(&nfsd_drc_lock);
1434}
1435
1436static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
1437                                           struct nfsd4_channel_attrs *battrs)
1438{
1439        int numslots = fattrs->maxreqs;
1440        int slotsize = slot_bytes(fattrs);
1441        struct nfsd4_session *new;
1442        int mem, i;
1443
1444        BUILD_BUG_ON(NFSD_MAX_SLOTS_PER_SESSION * sizeof(struct nfsd4_slot *)
1445                        + sizeof(struct nfsd4_session) > PAGE_SIZE);
1446        mem = numslots * sizeof(struct nfsd4_slot *);
1447
1448        new = kzalloc(sizeof(*new) + mem, GFP_KERNEL);
1449        if (!new)
1450                return NULL;
1451        /* allocate each struct nfsd4_slot and data cache in one piece */
1452        for (i = 0; i < numslots; i++) {
1453                new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
1454                if (!new->se_slots[i])
1455                        goto out_free;
1456        }
1457
1458        memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
1459        memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
1460
1461        return new;
1462out_free:
1463        while (i--)
1464                kfree(new->se_slots[i]);
1465        kfree(new);
1466        return NULL;
1467}
1468
1469static void free_conn(struct nfsd4_conn *c)
1470{
1471        svc_xprt_put(c->cn_xprt);
1472        kfree(c);
1473}
1474
1475static void nfsd4_conn_lost(struct svc_xpt_user *u)
1476{
1477        struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
1478        struct nfs4_client *clp = c->cn_session->se_client;
1479
1480        spin_lock(&clp->cl_lock);
1481        if (!list_empty(&c->cn_persession)) {
1482                list_del(&c->cn_persession);
1483                free_conn(c);
1484        }
1485        nfsd4_probe_callback(clp);
1486        spin_unlock(&clp->cl_lock);
1487}
1488
1489static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1490{
1491        struct nfsd4_conn *conn;
1492
1493        conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1494        if (!conn)
1495                return NULL;
1496        svc_xprt_get(rqstp->rq_xprt);
1497        conn->cn_xprt = rqstp->rq_xprt;
1498        conn->cn_flags = flags;
1499        INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1500        return conn;
1501}
1502
1503static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1504{
1505        conn->cn_session = ses;
1506        list_add(&conn->cn_persession, &ses->se_conns);
1507}
1508
1509static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1510{
1511        struct nfs4_client *clp = ses->se_client;
1512
1513        spin_lock(&clp->cl_lock);
1514        __nfsd4_hash_conn(conn, ses);
1515        spin_unlock(&clp->cl_lock);
1516}
1517
1518static int nfsd4_register_conn(struct nfsd4_conn *conn)
1519{
1520        conn->cn_xpt_user.callback = nfsd4_conn_lost;
1521        return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1522}
1523
1524static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1525{
1526        int ret;
1527
1528        nfsd4_hash_conn(conn, ses);
1529        ret = nfsd4_register_conn(conn);
1530        if (ret)
1531                /* oops; xprt is already down: */
1532                nfsd4_conn_lost(&conn->cn_xpt_user);
1533        /* We may have gained or lost a callback channel: */
1534        nfsd4_probe_callback_sync(ses->se_client);
1535}
1536
1537static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1538{
1539        u32 dir = NFS4_CDFC4_FORE;
1540
1541        if (cses->flags & SESSION4_BACK_CHAN)
1542                dir |= NFS4_CDFC4_BACK;
1543        return alloc_conn(rqstp, dir);
1544}
1545
1546/* must be called under client_lock */
1547static void nfsd4_del_conns(struct nfsd4_session *s)
1548{
1549        struct nfs4_client *clp = s->se_client;
1550        struct nfsd4_conn *c;
1551
1552        spin_lock(&clp->cl_lock);
1553        while (!list_empty(&s->se_conns)) {
1554                c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1555                list_del_init(&c->cn_persession);
1556                spin_unlock(&clp->cl_lock);
1557
1558                unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1559                free_conn(c);
1560
1561                spin_lock(&clp->cl_lock);
1562        }
1563        spin_unlock(&clp->cl_lock);
1564}
1565
1566static void __free_session(struct nfsd4_session *ses)
1567{
1568        free_session_slots(ses);
1569        kfree(ses);
1570}
1571
1572static void free_session(struct nfsd4_session *ses)
1573{
1574        nfsd4_del_conns(ses);
1575        nfsd4_put_drc_mem(&ses->se_fchannel);
1576        __free_session(ses);
1577}
1578
1579static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1580{
1581        int idx;
1582        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1583
1584        new->se_client = clp;
1585        gen_sessionid(new);
1586
1587        INIT_LIST_HEAD(&new->se_conns);
1588
1589        new->se_cb_seq_nr = 1;
1590        new->se_flags = cses->flags;
1591        new->se_cb_prog = cses->callback_prog;
1592        new->se_cb_sec = cses->cb_sec;
1593        atomic_set(&new->se_ref, 0);
1594        idx = hash_sessionid(&new->se_sessionid);
1595        list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1596        spin_lock(&clp->cl_lock);
1597        list_add(&new->se_perclnt, &clp->cl_sessions);
1598        spin_unlock(&clp->cl_lock);
1599
1600        {
1601                struct sockaddr *sa = svc_addr(rqstp);
1602                /*
1603                 * This is a little silly; with sessions there's no real
1604                 * use for the callback address.  Use the peer address
1605                 * as a reasonable default for now, but consider fixing
1606                 * the rpc client not to require an address in the
1607                 * future:
1608                 */
1609                rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
1610                clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
1611        }
1612}
1613
1614/* caller must hold client_lock */
1615static struct nfsd4_session *
1616__find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
1617{
1618        struct nfsd4_session *elem;
1619        int idx;
1620        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
1621
1622        lockdep_assert_held(&nn->client_lock);
1623
1624        dump_sessionid(__func__, sessionid);
1625        idx = hash_sessionid(sessionid);
1626        /* Search in the appropriate list */
1627        list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
1628                if (!memcmp(elem->se_sessionid.data, sessionid->data,
1629                            NFS4_MAX_SESSIONID_LEN)) {
1630                        return elem;
1631                }
1632        }
1633
1634        dprintk("%s: session not found\n", __func__);
1635        return NULL;
1636}
1637
1638static struct nfsd4_session *
1639find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
1640                __be32 *ret)
1641{
1642        struct nfsd4_session *session;
1643        __be32 status = nfserr_badsession;
1644
1645        session = __find_in_sessionid_hashtbl(sessionid, net);
1646        if (!session)
1647                goto out;
1648        status = nfsd4_get_session_locked(session);
1649        if (status)
1650                session = NULL;
1651out:
1652        *ret = status;
1653        return session;
1654}
1655
1656/* caller must hold client_lock */
1657static void
1658unhash_session(struct nfsd4_session *ses)
1659{
1660        struct nfs4_client *clp = ses->se_client;
1661        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1662
1663        lockdep_assert_held(&nn->client_lock);
1664
1665        list_del(&ses->se_hash);
1666        spin_lock(&ses->se_client->cl_lock);
1667        list_del(&ses->se_perclnt);
1668        spin_unlock(&ses->se_client->cl_lock);
1669}
1670
1671/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
1672static int
1673STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
1674{
1675        /*
1676         * We're assuming the clid was not given out from a boot
1677         * precisely 2^32 (about 136 years) before this one.  That seems
1678         * a safe assumption:
1679         */
1680        if (clid->cl_boot == (u32)nn->boot_time)
1681                return 0;
1682        dprintk("NFSD stale clientid (%08x/%08x) boot_time %08lx\n",
1683                clid->cl_boot, clid->cl_id, nn->boot_time);
1684        return 1;
1685}
1686
1687/* 
1688 * XXX Should we use a slab cache ?
1689 * This type of memory management is somewhat inefficient, but we use it
1690 * anyway since SETCLIENTID is not a common operation.
1691 */
1692static struct nfs4_client *alloc_client(struct xdr_netobj name)
1693{
1694        struct nfs4_client *clp;
1695        int i;
1696
1697        clp = kzalloc(sizeof(struct nfs4_client), GFP_KERNEL);
1698        if (clp == NULL)
1699                return NULL;
1700        clp->cl_name.data = kmemdup(name.data, name.len, GFP_KERNEL);
1701        if (clp->cl_name.data == NULL)
1702                goto err_no_name;
1703        clp->cl_ownerstr_hashtbl = kmalloc(sizeof(struct list_head) *
1704                        OWNER_HASH_SIZE, GFP_KERNEL);
1705        if (!clp->cl_ownerstr_hashtbl)
1706                goto err_no_hashtbl;
1707        for (i = 0; i < OWNER_HASH_SIZE; i++)
1708                INIT_LIST_HEAD(&clp->cl_ownerstr_hashtbl[i]);
1709        clp->cl_name.len = name.len;
1710        INIT_LIST_HEAD(&clp->cl_sessions);
1711        idr_init(&clp->cl_stateids);
1712        atomic_set(&clp->cl_refcount, 0);
1713        clp->cl_cb_state = NFSD4_CB_UNKNOWN;
1714        INIT_LIST_HEAD(&clp->cl_idhash);
1715        INIT_LIST_HEAD(&clp->cl_openowners);
1716        INIT_LIST_HEAD(&clp->cl_delegations);
1717        INIT_LIST_HEAD(&clp->cl_lru);
1718        INIT_LIST_HEAD(&clp->cl_revoked);
1719#ifdef CONFIG_NFSD_PNFS
1720        INIT_LIST_HEAD(&clp->cl_lo_states);
1721#endif
1722        spin_lock_init(&clp->cl_lock);
1723        rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
1724        return clp;
1725err_no_hashtbl:
1726        kfree(clp->cl_name.data);
1727err_no_name:
1728        kfree(clp);
1729        return NULL;
1730}
1731
1732static void
1733free_client(struct nfs4_client *clp)
1734{
1735        while (!list_empty(&clp->cl_sessions)) {
1736                struct nfsd4_session *ses;
1737                ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
1738                                se_perclnt);
1739                list_del(&ses->se_perclnt);
1740                WARN_ON_ONCE(atomic_read(&ses->se_ref));
1741                free_session(ses);
1742        }
1743        rpc_destroy_wait_queue(&clp->cl_cb_waitq);
1744        free_svc_cred(&clp->cl_cred);
1745        kfree(clp->cl_ownerstr_hashtbl);
1746        kfree(clp->cl_name.data);
1747        idr_destroy(&clp->cl_stateids);
1748        kfree(clp);
1749}
1750
1751/* must be called under the client_lock */
1752static void
1753unhash_client_locked(struct nfs4_client *clp)
1754{
1755        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1756        struct nfsd4_session *ses;
1757
1758        lockdep_assert_held(&nn->client_lock);
1759
1760        /* Mark the client as expired! */
1761        clp->cl_time = 0;
1762        /* Make it invisible */
1763        if (!list_empty(&clp->cl_idhash)) {
1764                list_del_init(&clp->cl_idhash);
1765                if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
1766                        rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
1767                else
1768                        rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
1769        }
1770        list_del_init(&clp->cl_lru);
1771        spin_lock(&clp->cl_lock);
1772        list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
1773                list_del_init(&ses->se_hash);
1774        spin_unlock(&clp->cl_lock);
1775}
1776
1777static void
1778unhash_client(struct nfs4_client *clp)
1779{
1780        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
1781
1782        spin_lock(&nn->client_lock);
1783        unhash_client_locked(clp);
1784        spin_unlock(&nn->client_lock);
1785}
1786
1787static __be32 mark_client_expired_locked(struct nfs4_client *clp)
1788{
1789        if (atomic_read(&clp->cl_refcount))
1790                return nfserr_jukebox;
1791        unhash_client_locked(clp);
1792        return nfs_ok;
1793}
1794
1795static void
1796__destroy_client(struct nfs4_client *clp)
1797{
1798        struct nfs4_openowner *oo;
1799        struct nfs4_delegation *dp;
1800        struct list_head reaplist;
1801
1802        INIT_LIST_HEAD(&reaplist);
1803        spin_lock(&state_lock);
1804        while (!list_empty(&clp->cl_delegations)) {
1805                dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
1806                WARN_ON(!unhash_delegation_locked(dp));
1807                list_add(&dp->dl_recall_lru, &reaplist);
1808        }
1809        spin_unlock(&state_lock);
1810        while (!list_empty(&reaplist)) {
1811                dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
1812                list_del_init(&dp->dl_recall_lru);
1813                put_clnt_odstate(dp->dl_clnt_odstate);
1814                nfs4_put_deleg_lease(dp->dl_stid.sc_file);
1815                nfs4_put_stid(&dp->dl_stid);
1816        }
1817        while (!list_empty(&clp->cl_revoked)) {
1818                dp = list_entry(clp->cl_revoked.next, struct nfs4_delegation, dl_recall_lru);
1819                list_del_init(&dp->dl_recall_lru);
1820                nfs4_put_stid(&dp->dl_stid);
1821        }
1822        while (!list_empty(&clp->cl_openowners)) {
1823                oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
1824                nfs4_get_stateowner(&oo->oo_owner);
1825                release_openowner(oo);
1826        }
1827        nfsd4_return_all_client_layouts(clp);
1828        nfsd4_shutdown_callback(clp);
1829        if (clp->cl_cb_conn.cb_xprt)
1830                svc_xprt_put(clp->cl_cb_conn.cb_xprt);
1831        free_client(clp);
1832}
1833
1834static void
1835destroy_client(struct nfs4_client *clp)
1836{
1837        unhash_client(clp);
1838        __destroy_client(clp);
1839}
1840
1841static void expire_client(struct nfs4_client *clp)
1842{
1843        unhash_client(clp);
1844        nfsd4_client_record_remove(clp);
1845        __destroy_client(clp);
1846}
1847
1848static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
1849{
1850        memcpy(target->cl_verifier.data, source->data,
1851                        sizeof(target->cl_verifier.data));
1852}
1853
1854static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
1855{
1856        target->cl_clientid.cl_boot = source->cl_clientid.cl_boot; 
1857        target->cl_clientid.cl_id = source->cl_clientid.cl_id; 
1858}
1859
1860int strdup_if_nonnull(char **target, char *source)
1861{
1862        if (source) {
1863                *target = kstrdup(source, GFP_KERNEL);
1864                if (!*target)
1865                        return -ENOMEM;
1866        } else
1867                *target = NULL;
1868        return 0;
1869}
1870
1871static int copy_cred(struct svc_cred *target, struct svc_cred *source)
1872{
1873        int ret;
1874
1875        ret = strdup_if_nonnull(&target->cr_principal, source->cr_principal);
1876        if (ret)
1877                return ret;
1878        ret = strdup_if_nonnull(&target->cr_raw_principal,
1879                                        source->cr_raw_principal);
1880        if (ret)
1881                return ret;
1882        target->cr_flavor = source->cr_flavor;
1883        target->cr_uid = source->cr_uid;
1884        target->cr_gid = source->cr_gid;
1885        target->cr_group_info = source->cr_group_info;
1886        get_group_info(target->cr_group_info);
1887        target->cr_gss_mech = source->cr_gss_mech;
1888        if (source->cr_gss_mech)
1889                gss_mech_get(source->cr_gss_mech);
1890        return 0;
1891}
1892
1893static int
1894compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
1895{
1896        if (o1->len < o2->len)
1897                return -1;
1898        if (o1->len > o2->len)
1899                return 1;
1900        return memcmp(o1->data, o2->data, o1->len);
1901}
1902
1903static int same_name(const char *n1, const char *n2)
1904{
1905        return 0 == memcmp(n1, n2, HEXDIR_LEN);
1906}
1907
1908static int
1909same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
1910{
1911        return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
1912}
1913
1914static int
1915same_clid(clientid_t *cl1, clientid_t *cl2)
1916{
1917        return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
1918}
1919
1920static bool groups_equal(struct group_info *g1, struct group_info *g2)
1921{
1922        int i;
1923
1924        if (g1->ngroups != g2->ngroups)
1925                return false;
1926        for (i=0; i<g1->ngroups; i++)
1927                if (!gid_eq(GROUP_AT(g1, i), GROUP_AT(g2, i)))
1928                        return false;
1929        return true;
1930}
1931
1932/*
1933 * RFC 3530 language requires clid_inuse be returned when the
1934 * "principal" associated with a requests differs from that previously
1935 * used.  We use uid, gid's, and gss principal string as our best
1936 * approximation.  We also don't want to allow non-gss use of a client
1937 * established using gss: in theory cr_principal should catch that
1938 * change, but in practice cr_principal can be null even in the gss case
1939 * since gssd doesn't always pass down a principal string.
1940 */
1941static bool is_gss_cred(struct svc_cred *cr)
1942{
1943        /* Is cr_flavor one of the gss "pseudoflavors"?: */
1944        return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
1945}
1946
1947
1948static bool
1949same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
1950{
1951        if ((is_gss_cred(cr1) != is_gss_cred(cr2))
1952                || (!uid_eq(cr1->cr_uid, cr2->cr_uid))
1953                || (!gid_eq(cr1->cr_gid, cr2->cr_gid))
1954                || !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
1955                return false;
1956        if (cr1->cr_principal == cr2->cr_principal)
1957                return true;
1958        if (!cr1->cr_principal || !cr2->cr_principal)
1959                return false;
1960        return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
1961}
1962
1963static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
1964{
1965        struct svc_cred *cr = &rqstp->rq_cred;
1966        u32 service;
1967
1968        if (!cr->cr_gss_mech)
1969                return false;
1970        service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
1971        return service == RPC_GSS_SVC_INTEGRITY ||
1972               service == RPC_GSS_SVC_PRIVACY;
1973}
1974
1975static bool mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
1976{
1977        struct svc_cred *cr = &rqstp->rq_cred;
1978
1979        if (!cl->cl_mach_cred)
1980                return true;
1981        if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
1982                return false;
1983        if (!svc_rqst_integrity_protected(rqstp))
1984                return false;
1985        if (cl->cl_cred.cr_raw_principal)
1986                return 0 == strcmp(cl->cl_cred.cr_raw_principal,
1987                                                cr->cr_raw_principal);
1988        if (!cr->cr_principal)
1989                return false;
1990        return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
1991}
1992
1993static void gen_confirm(struct nfs4_client *clp, struct nfsd_net *nn)
1994{
1995        __be32 verf[2];
1996
1997        /*
1998         * This is opaque to client, so no need to byte-swap. Use
1999         * __force to keep sparse happy
2000         */
2001        verf[0] = (__force __be32)get_seconds();
2002        verf[1] = (__force __be32)nn->clverifier_counter++;
2003        memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
2004}
2005
2006static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
2007{
2008        clp->cl_clientid.cl_boot = nn->boot_time;
2009        clp->cl_clientid.cl_id = nn->clientid_counter++;
2010        gen_confirm(clp, nn);
2011}
2012
2013static struct nfs4_stid *
2014find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
2015{
2016        struct nfs4_stid *ret;
2017
2018        ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
2019        if (!ret || !ret->sc_type)
2020                return NULL;
2021        return ret;
2022}
2023
2024static struct nfs4_stid *
2025find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
2026{
2027        struct nfs4_stid *s;
2028
2029        spin_lock(&cl->cl_lock);
2030        s = find_stateid_locked(cl, t);
2031        if (s != NULL) {
2032                if (typemask & s->sc_type)
2033                        atomic_inc(&s->sc_count);
2034                else
2035                        s = NULL;
2036        }
2037        spin_unlock(&cl->cl_lock);
2038        return s;
2039}
2040
2041static struct nfs4_client *create_client(struct xdr_netobj name,
2042                struct svc_rqst *rqstp, nfs4_verifier *verf)
2043{
2044        struct nfs4_client *clp;
2045        struct sockaddr *sa = svc_addr(rqstp);
2046        int ret;
2047        struct net *net = SVC_NET(rqstp);
2048
2049        clp = alloc_client(name);
2050        if (clp == NULL)
2051                return NULL;
2052
2053        ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
2054        if (ret) {
2055                free_client(clp);
2056                return NULL;
2057        }
2058        nfsd4_init_cb(&clp->cl_cb_null, clp, NULL, NFSPROC4_CLNT_CB_NULL);
2059        clp->cl_time = get_seconds();
2060        clear_bit(0, &clp->cl_cb_slot_busy);
2061        copy_verf(clp, verf);
2062        rpc_copy_addr((struct sockaddr *) &clp->cl_addr, sa);
2063        clp->cl_cb_session = NULL;
2064        clp->net = net;
2065        return clp;
2066}
2067
2068static void
2069add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
2070{
2071        struct rb_node **new = &(root->rb_node), *parent = NULL;
2072        struct nfs4_client *clp;
2073
2074        while (*new) {
2075                clp = rb_entry(*new, struct nfs4_client, cl_namenode);
2076                parent = *new;
2077
2078                if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
2079                        new = &((*new)->rb_left);
2080                else
2081                        new = &((*new)->rb_right);
2082        }
2083
2084        rb_link_node(&new_clp->cl_namenode, parent, new);
2085        rb_insert_color(&new_clp->cl_namenode, root);
2086}
2087
2088static struct nfs4_client *
2089find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
2090{
2091        int cmp;
2092        struct rb_node *node = root->rb_node;
2093        struct nfs4_client *clp;
2094
2095        while (node) {
2096                clp = rb_entry(node, struct nfs4_client, cl_namenode);
2097                cmp = compare_blob(&clp->cl_name, name);
2098                if (cmp > 0)
2099                        node = node->rb_left;
2100                else if (cmp < 0)
2101                        node = node->rb_right;
2102                else
2103                        return clp;
2104        }
2105        return NULL;
2106}
2107
2108static void
2109add_to_unconfirmed(struct nfs4_client *clp)
2110{
2111        unsigned int idhashval;
2112        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2113
2114        lockdep_assert_held(&nn->client_lock);
2115
2116        clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2117        add_clp_to_name_tree(clp, &nn->unconf_name_tree);
2118        idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2119        list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
2120        renew_client_locked(clp);
2121}
2122
2123static void
2124move_to_confirmed(struct nfs4_client *clp)
2125{
2126        unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
2127        struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2128
2129        lockdep_assert_held(&nn->client_lock);
2130
2131        dprintk("NFSD: move_to_confirm nfs4_client %p\n", clp);
2132        list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
2133        rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2134        add_clp_to_name_tree(clp, &nn->conf_name_tree);
2135        set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
2136        renew_client_locked(clp);
2137}
2138
2139static struct nfs4_client *
2140find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
2141{
2142        struct nfs4_client *clp;
2143        unsigned int idhashval = clientid_hashval(clid->cl_id);
2144
2145        list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
2146                if (same_clid(&clp->cl_clientid, clid)) {
2147                        if ((bool)clp->cl_minorversion != sessions)
2148                                return NULL;
2149                        renew_client_locked(clp);
2150                        return clp;
2151                }
2152        }
2153        return NULL;
2154}
2155
2156static struct nfs4_client *
2157find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2158{
2159        struct list_head *tbl = nn->conf_id_hashtbl;
2160
2161        lockdep_assert_held(&nn->client_lock);
2162        return find_client_in_id_table(tbl, clid, sessions);
2163}
2164
2165static struct nfs4_client *
2166find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
2167{
2168        struct list_head *tbl = nn->unconf_id_hashtbl;
2169
2170        lockdep_assert_held(&nn->client_lock);
2171        return find_client_in_id_table(tbl, clid, sessions);
2172}
2173
2174static bool clp_used_exchangeid(struct nfs4_client *clp)
2175{
2176        return clp->cl_exchange_flags != 0;
2177} 
2178
2179static struct nfs4_client *
2180find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2181{
2182        lockdep_assert_held(&nn->client_lock);
2183        return find_clp_in_name_tree(name, &nn->conf_name_tree);
2184}
2185
2186static struct nfs4_client *
2187find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
2188{
2189        lockdep_assert_held(&nn->client_lock);
2190        return find_clp_in_name_tree(name, &nn->unconf_name_tree);
2191}
2192
2193static void
2194gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
2195{
2196        struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
2197        struct sockaddr *sa = svc_addr(rqstp);
2198        u32 scopeid = rpc_get_scope_id(sa);
2199        unsigned short expected_family;
2200
2201        /* Currently, we only support tcp and tcp6 for the callback channel */
2202        if (se->se_callback_netid_len == 3 &&
2203            !memcmp(se->se_callback_netid_val, "tcp", 3))
2204                expected_family = AF_INET;
2205        else if (se->se_callback_netid_len == 4 &&
2206                 !memcmp(se->se_callback_netid_val, "tcp6", 4))
2207                expected_family = AF_INET6;
2208        else
2209                goto out_err;
2210
2211        conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
2212                                            se->se_callback_addr_len,
2213                                            (struct sockaddr *)&conn->cb_addr,
2214                                            sizeof(conn->cb_addr));
2215
2216        if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
2217                goto out_err;
2218
2219        if (conn->cb_addr.ss_family == AF_INET6)
2220                ((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
2221
2222        conn->cb_prog = se->se_callback_prog;
2223        conn->cb_ident = se->se_callback_ident;
2224        memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
2225        return;
2226out_err:
2227        conn->cb_addr.ss_family = AF_UNSPEC;
2228        conn->cb_addrlen = 0;
2229        dprintk(KERN_INFO "NFSD: this client (clientid %08x/%08x) "
2230                "will not receive delegations\n",
2231                clp->cl_clientid.cl_boot, clp->cl_clientid.cl_id);
2232
2233        return;
2234}
2235
2236/*
2237 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
2238 */
2239static void
2240nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
2241{
2242        struct xdr_buf *buf = resp->xdr.buf;
2243        struct nfsd4_slot *slot = resp->cstate.slot;
2244        unsigned int base;
2245
2246        dprintk("--> %s slot %p\n", __func__, slot);
2247
2248        slot->sl_opcnt = resp->opcnt;
2249        slot->sl_status = resp->cstate.status;
2250
2251        slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
2252        if (nfsd4_not_cached(resp)) {
2253                slot->sl_datalen = 0;
2254                return;
2255        }
2256        base = resp->cstate.data_offset;
2257        slot->sl_datalen = buf->len - base;
2258        if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
2259                WARN(1, "%s: sessions DRC could not cache compound\n",
2260                     __func__);
2261        return;
2262}
2263
2264/*
2265 * Encode the replay sequence operation from the slot values.
2266 * If cachethis is FALSE encode the uncached rep error on the next
2267 * operation which sets resp->p and increments resp->opcnt for
2268 * nfs4svc_encode_compoundres.
2269 *
2270 */
2271static __be32
2272nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
2273                          struct nfsd4_compoundres *resp)
2274{
2275        struct nfsd4_op *op;
2276        struct nfsd4_slot *slot = resp->cstate.slot;
2277
2278        /* Encode the replayed sequence operation */
2279        op = &args->ops[resp->opcnt - 1];
2280        nfsd4_encode_operation(resp, op);
2281
2282        /* Return nfserr_retry_uncached_rep in next operation. */
2283        if (args->opcnt > 1 && !(slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
2284                op = &args->ops[resp->opcnt++];
2285                op->status = nfserr_retry_uncached_rep;
2286                nfsd4_encode_operation(resp, op);
2287        }
2288        return op->status;
2289}
2290
2291/*
2292 * The sequence operation is not cached because we can use the slot and
2293 * session values.
2294 */
2295static __be32
2296nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
2297                         struct nfsd4_sequence *seq)
2298{
2299        struct nfsd4_slot *slot = resp->cstate.slot;
2300        struct xdr_stream *xdr = &resp->xdr;
2301        __be32 *p;
2302        __be32 status;
2303
2304        dprintk("--> %s slot %p\n", __func__, slot);
2305
2306        status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
2307        if (status)
2308                return status;
2309
2310        p = xdr_reserve_space(xdr, slot->sl_datalen);
2311        if (!p) {
2312                WARN_ON_ONCE(1);
2313                return nfserr_serverfault;
2314        }
2315        xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
2316        xdr_commit_encode(xdr);
2317
2318        resp->opcnt = slot->sl_opcnt;
2319        return slot->sl_status;
2320}
2321
2322/*
2323 * Set the exchange_id flags returned by the server.
2324 */
2325static void
2326nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
2327{
2328#ifdef CONFIG_NFSD_PNFS
2329        new->cl_exchange_flags |= EXCHGID4_FLAG_USE_PNFS_MDS;
2330#else
2331        new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
2332#endif
2333
2334        /* Referrals are supported, Migration is not. */
2335        new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
2336
2337        /* set the wire flags to return to client. */
2338        clid->flags = new->cl_exchange_flags;
2339}
2340
2341static bool client_has_openowners(struct nfs4_client *clp)
2342{
2343        struct nfs4_openowner *oo;
2344
2345        list_for_each_entry(oo, &clp->cl_openowners, oo_perclient) {
2346                if (!list_empty(&oo->oo_owner.so_stateids))
2347                        return true;
2348        }
2349        return false;
2350}
2351
2352static bool client_has_state(struct nfs4_client *clp)
2353{
2354        return client_has_openowners(clp)
2355#ifdef CONFIG_NFSD_PNFS
2356                || !list_empty(&clp->cl_lo_states)
2357#endif
2358                || !list_empty(&clp->cl_delegations)
2359                || !list_empty(&clp->cl_sessions);
2360}
2361
2362__be32
2363nfsd4_exchange_id(struct svc_rqst *rqstp,
2364                  struct nfsd4_compound_state *cstate,
2365                  struct nfsd4_exchange_id *exid)
2366{
2367        struct nfs4_client *conf, *new;
2368        struct nfs4_client *unconf = NULL;
2369        __be32 status;
2370        char                    addr_str[INET6_ADDRSTRLEN];
2371        nfs4_verifier           verf = exid->verifier;
2372        struct sockaddr         *sa = svc_addr(rqstp);
2373        bool    update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
2374        struct nfsd_net         *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2375
2376        rpc_ntop(sa, addr_str, sizeof(addr_str));
2377        dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
2378                "ip_addr=%s flags %x, spa_how %d\n",
2379                __func__, rqstp, exid, exid->clname.len, exid->clname.data,
2380                addr_str, exid->flags, exid->spa_how);
2381
2382        if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
2383                return nfserr_inval;
2384
2385        new = create_client(exid->clname, rqstp, &verf);
2386        if (new == NULL)
2387                return nfserr_jukebox;
2388
2389        switch (exid->spa_how) {
2390        case SP4_MACH_CRED:
2391                if (!svc_rqst_integrity_protected(rqstp)) {
2392                        status = nfserr_inval;
2393                        goto out_nolock;
2394                }
2395                /*
2396                 * Sometimes userspace doesn't give us a principal.
2397                 * Which is a bug, really.  Anyway, we can't enforce
2398                 * MACH_CRED in that case, better to give up now:
2399                 */
2400                if (!new->cl_cred.cr_principal &&
2401                                        !new->cl_cred.cr_raw_principal) {
2402                        status = nfserr_serverfault;
2403                        goto out_nolock;
2404                }
2405                new->cl_mach_cred = true;
2406        case SP4_NONE:
2407                break;
2408        default:                                /* checked by xdr code */
2409                WARN_ON_ONCE(1);
2410        case SP4_SSV:
2411                status = nfserr_encr_alg_unsupp;
2412                goto out_nolock;
2413        }
2414
2415        /* Cases below refer to rfc 5661 section 18.35.4: */
2416        spin_lock(&nn->client_lock);
2417        conf = find_confirmed_client_by_name(&exid->clname, nn);
2418        if (conf) {
2419                bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
2420                bool verfs_match = same_verf(&verf, &conf->cl_verifier);
2421
2422                if (update) {
2423                        if (!clp_used_exchangeid(conf)) { /* buggy client */
2424                                status = nfserr_inval;
2425                                goto out;
2426                        }
2427                        if (!mach_creds_match(conf, rqstp)) {
2428                                status = nfserr_wrong_cred;
2429                                goto out;
2430                        }
2431                        if (!creds_match) { /* case 9 */
2432                                status = nfserr_perm;
2433                                goto out;
2434                        }
2435                        if (!verfs_match) { /* case 8 */
2436                                status = nfserr_not_same;
2437                                goto out;
2438                        }
2439                        /* case 6 */
2440                        exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
2441                        goto out_copy;
2442                }
2443                if (!creds_match) { /* case 3 */
2444                        if (client_has_state(conf)) {
2445                                status = nfserr_clid_inuse;
2446                                goto out;
2447                        }
2448                        goto out_new;
2449                }
2450                if (verfs_match) { /* case 2 */
2451                        conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
2452                        goto out_copy;
2453                }
2454                /* case 5, client reboot */
2455                conf = NULL;
2456                goto out_new;
2457        }
2458
2459        if (update) { /* case 7 */
2460                status = nfserr_noent;
2461                goto out;
2462        }
2463
2464        unconf  = find_unconfirmed_client_by_name(&exid->clname, nn);
2465        if (unconf) /* case 4, possible retry or client restart */
2466                unhash_client_locked(unconf);
2467
2468        /* case 1 (normal case) */
2469out_new:
2470        if (conf) {
2471                status = mark_client_expired_locked(conf);
2472                if (status)
2473                        goto out;
2474        }
2475        new->cl_minorversion = cstate->minorversion;
2476
2477        gen_clid(new, nn);
2478        add_to_unconfirmed(new);
2479        swap(new, conf);
2480out_copy:
2481        exid->clientid.cl_boot = conf->cl_clientid.cl_boot;
2482        exid->clientid.cl_id = conf->cl_clientid.cl_id;
2483
2484        exid->seqid = conf->cl_cs_slot.sl_seqid + 1;
2485        nfsd4_set_ex_flags(conf, exid);
2486
2487        dprintk("nfsd4_exchange_id seqid %d flags %x\n",
2488                conf->cl_cs_slot.sl_seqid, conf->cl_exchange_flags);
2489        status = nfs_ok;
2490
2491out:
2492        spin_unlock(&nn->client_lock);
2493out_nolock:
2494        if (new)
2495                expire_client(new);
2496        if (unconf)
2497                expire_client(unconf);
2498        return status;
2499}
2500
2501static __be32
2502check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
2503{
2504        dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
2505                slot_seqid);
2506
2507        /* The slot is in use, and no response has been sent. */
2508        if (slot_inuse) {
2509                if (seqid == slot_seqid)
2510                        return nfserr_jukebox;
2511                else
2512                        return nfserr_seq_misordered;
2513        }
2514        /* Note unsigned 32-bit arithmetic handles wraparound: */
2515        if (likely(seqid == slot_seqid + 1))
2516                return nfs_ok;
2517        if (seqid == slot_seqid)
2518                return nfserr_replay_cache;
2519        return nfserr_seq_misordered;
2520}
2521
2522/*
2523 * Cache the create session result into the create session single DRC
2524 * slot cache by saving the xdr structure. sl_seqid has been set.
2525 * Do this for solo or embedded create session operations.
2526 */
2527static void
2528nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
2529                           struct nfsd4_clid_slot *slot, __be32 nfserr)
2530{
2531        slot->sl_status = nfserr;
2532        memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
2533}
2534
2535static __be32
2536nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
2537                            struct nfsd4_clid_slot *slot)
2538{
2539        memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
2540        return slot->sl_status;
2541}
2542
2543#define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
2544                        2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
2545                        1 +     /* MIN tag is length with zero, only length */ \
2546                        3 +     /* version, opcount, opcode */ \
2547                        XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2548                                /* seqid, slotID, slotID, cache */ \
2549                        4 ) * sizeof(__be32))
2550
2551#define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
2552                        2 +     /* verifier: AUTH_NULL, length 0 */\
2553                        1 +     /* status */ \
2554                        1 +     /* MIN tag is length with zero, only length */ \
2555                        3 +     /* opcount, opcode, opstatus*/ \
2556                        XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
2557                                /* seqid, slotID, slotID, slotID, status */ \
2558                        5 ) * sizeof(__be32))
2559
2560static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
2561{
2562        u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
2563
2564        if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
2565                return nfserr_toosmall;
2566        if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
2567                return nfserr_toosmall;
2568        ca->headerpadsz = 0;
2569        ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
2570        ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
2571        ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
2572        ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
2573                        NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
2574        ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
2575        /*
2576         * Note decreasing slot size below client's request may make it
2577         * difficult for client to function correctly, whereas
2578         * decreasing the number of slots will (just?) affect
2579         * performance.  When short on memory we therefore prefer to
2580         * decrease number of slots instead of their size.  Clients that
2581         * request larger slots than they need will get poor results:
2582         */
2583        ca->maxreqs = nfsd4_get_drc_mem(ca);
2584        if (!ca->maxreqs)
2585                return nfserr_jukebox;
2586
2587        return nfs_ok;
2588}
2589
2590/*
2591 * Server's NFSv4.1 backchannel support is AUTH_SYS-only for now.
2592 * These are based on similar macros in linux/sunrpc/msg_prot.h .
2593 */
2594#define RPC_MAX_HEADER_WITH_AUTH_SYS \
2595        (RPC_CALLHDRSIZE + 2 * (2 + UNX_CALLSLACK))
2596
2597#define RPC_MAX_REPHEADER_WITH_AUTH_SYS \
2598        (RPC_REPHDRSIZE + (2 + NUL_REPLYSLACK))
2599
2600#define NFSD_CB_MAX_REQ_SZ      ((NFS4_enc_cb_recall_sz + \
2601                                 RPC_MAX_HEADER_WITH_AUTH_SYS) * sizeof(__be32))
2602#define NFSD_CB_MAX_RESP_SZ     ((NFS4_dec_cb_recall_sz + \
2603                                 RPC_MAX_REPHEADER_WITH_AUTH_SYS) * \
2604                                 sizeof(__be32))
2605
2606static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
2607{
2608        ca->headerpadsz = 0;
2609
2610        if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
2611                return nfserr_toosmall;
2612        if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
2613                return nfserr_toosmall;
2614        ca->maxresp_cached = 0;
2615        if (ca->maxops < 2)
2616                return nfserr_toosmall;
2617
2618        return nfs_ok;
2619}
2620
2621static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
2622{
2623        switch (cbs->flavor) {
2624        case RPC_AUTH_NULL:
2625        case RPC_AUTH_UNIX:
2626                return nfs_ok;
2627        default:
2628                /*
2629                 * GSS case: the spec doesn't allow us to return this
2630                 * error.  But it also doesn't allow us not to support
2631                 * GSS.
2632                 * I'd rather this fail hard than return some error the
2633                 * client might think it can already handle:
2634                 */
2635                return nfserr_encr_alg_unsupp;
2636        }
2637}
2638
2639__be32
2640nfsd4_create_session(struct svc_rqst *rqstp,
2641                     struct nfsd4_compound_state *cstate,
2642                     struct nfsd4_create_session *cr_ses)
2643{
2644        struct sockaddr *sa = svc_addr(rqstp);
2645        struct nfs4_client *conf, *unconf;
2646        struct nfs4_client *old = NULL;
2647        struct nfsd4_session *new;
2648        struct nfsd4_conn *conn;
2649        struct nfsd4_clid_slot *cs_slot = NULL;
2650        __be32 status = 0;
2651        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2652
2653        if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
2654                return nfserr_inval;
2655        status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
2656        if (status)
2657                return status;
2658        status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
2659        if (status)
2660                return status;
2661        status = check_backchannel_attrs(&cr_ses->back_channel);
2662        if (status)
2663                goto out_release_drc_mem;
2664        status = nfserr_jukebox;
2665        new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
2666        if (!new)
2667                goto out_release_drc_mem;
2668        conn = alloc_conn_from_crses(rqstp, cr_ses);
2669        if (!conn)
2670                goto out_free_session;
2671
2672        spin_lock(&nn->client_lock);
2673        unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
2674        conf = find_confirmed_client(&cr_ses->clientid, true, nn);
2675        WARN_ON_ONCE(conf && unconf);
2676
2677        if (conf) {
2678                status = nfserr_wrong_cred;
2679                if (!mach_creds_match(conf, rqstp))
2680                        goto out_free_conn;
2681                cs_slot = &conf->cl_cs_slot;
2682                status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2683                if (status) {
2684                        if (status == nfserr_replay_cache)
2685                                status = nfsd4_replay_create_session(cr_ses, cs_slot);
2686                        goto out_free_conn;
2687                }
2688        } else if (unconf) {
2689                if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
2690                    !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
2691                        status = nfserr_clid_inuse;
2692                        goto out_free_conn;
2693                }
2694                status = nfserr_wrong_cred;
2695                if (!mach_creds_match(unconf, rqstp))
2696                        goto out_free_conn;
2697                cs_slot = &unconf->cl_cs_slot;
2698                status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
2699                if (status) {
2700                        /* an unconfirmed replay returns misordered */
2701                        status = nfserr_seq_misordered;
2702                        goto out_free_conn;
2703                }
2704                old = find_confirmed_client_by_name(&unconf->cl_name, nn);
2705                if (old) {
2706                        status = mark_client_expired_locked(old);
2707                        if (status) {
2708                                old = NULL;
2709                                goto out_free_conn;
2710                        }
2711                }
2712                move_to_confirmed(unconf);
2713                conf = unconf;
2714        } else {
2715                status = nfserr_stale_clientid;
2716                goto out_free_conn;
2717        }
2718        status = nfs_ok;
2719        /* Persistent sessions are not supported */
2720        cr_ses->flags &= ~SESSION4_PERSIST;
2721        /* Upshifting from TCP to RDMA is not supported */
2722        cr_ses->flags &= ~SESSION4_RDMA;
2723
2724        init_session(rqstp, new, conf, cr_ses);
2725        nfsd4_get_session_locked(new);
2726
2727        memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
2728               NFS4_MAX_SESSIONID_LEN);
2729        cs_slot->sl_seqid++;
2730        cr_ses->seqid = cs_slot->sl_seqid;
2731
2732        /* cache solo and embedded create sessions under the client_lock */
2733        nfsd4_cache_create_session(cr_ses, cs_slot, status);
2734        spin_unlock(&nn->client_lock);
2735        /* init connection and backchannel */
2736        nfsd4_init_conn(rqstp, conn, new);
2737        nfsd4_put_session(new);
2738        if (old)
2739                expire_client(old);
2740        return status;
2741out_free_conn:
2742        spin_unlock(&nn->client_lock);
2743        free_conn(conn);
2744        if (old)
2745                expire_client(old);
2746out_free_session:
2747        __free_session(new);
2748out_release_drc_mem:
2749        nfsd4_put_drc_mem(&cr_ses->fore_channel);
2750        return status;
2751}
2752
2753static __be32 nfsd4_map_bcts_dir(u32 *dir)
2754{
2755        switch (*dir) {
2756        case NFS4_CDFC4_FORE:
2757        case NFS4_CDFC4_BACK:
2758                return nfs_ok;
2759        case NFS4_CDFC4_FORE_OR_BOTH:
2760        case NFS4_CDFC4_BACK_OR_BOTH:
2761                *dir = NFS4_CDFC4_BOTH;
2762                return nfs_ok;
2763        };
2764        return nfserr_inval;
2765}
2766
2767__be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_backchannel_ctl *bc)
2768{
2769        struct nfsd4_session *session = cstate->session;
2770        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
2771        __be32 status;
2772
2773        status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
2774        if (status)
2775                return status;
2776        spin_lock(&nn->client_lock);
2777        session->se_cb_prog = bc->bc_cb_program;
2778        session->se_cb_sec = bc->bc_cb_sec;
2779        spin_unlock(&nn->client_lock);
2780
2781        nfsd4_probe_callback(session->se_client);
2782
2783        return nfs_ok;
2784}
2785
2786__be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
2787                     struct nfsd4_compound_state *cstate,
2788                     struct nfsd4_bind_conn_to_session *bcts)
2789{
2790        __be32 status;
2791        struct nfsd4_conn *conn;
2792        struct nfsd4_session *session;
2793        struct net *net = SVC_NET(rqstp);
2794        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2795
2796        if (!nfsd4_last_compound_op(rqstp))
2797                return nfserr_not_only_op;
2798        spin_lock(&nn->client_lock);
2799        session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
2800        spin_unlock(&nn->client_lock);
2801        if (!session)
2802                goto out_no_session;
2803        status = nfserr_wrong_cred;
2804        if (!mach_creds_match(session->se_client, rqstp))
2805                goto out;
2806        status = nfsd4_map_bcts_dir(&bcts->dir);
2807        if (status)
2808                goto out;
2809        conn = alloc_conn(rqstp, bcts->dir);
2810        status = nfserr_jukebox;
2811        if (!conn)
2812                goto out;
2813        nfsd4_init_conn(rqstp, conn, session);
2814        status = nfs_ok;
2815out:
2816        nfsd4_put_session(session);
2817out_no_session:
2818        return status;
2819}
2820
2821static bool nfsd4_compound_in_session(struct nfsd4_session *session, struct nfs4_sessionid *sid)
2822{
2823        if (!session)
2824                return 0;
2825        return !memcmp(sid, &session->se_sessionid, sizeof(*sid));
2826}
2827
2828__be32
2829nfsd4_destroy_session(struct svc_rqst *r,
2830                      struct nfsd4_compound_state *cstate,
2831                      struct nfsd4_destroy_session *sessionid)
2832{
2833        struct nfsd4_session *ses;
2834        __be32 status;
2835        int ref_held_by_me = 0;
2836        struct net *net = SVC_NET(r);
2837        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2838
2839        status = nfserr_not_only_op;
2840        if (nfsd4_compound_in_session(cstate->session, &sessionid->sessionid)) {
2841                if (!nfsd4_last_compound_op(r))
2842                        goto out;
2843                ref_held_by_me++;
2844        }
2845        dump_sessionid(__func__, &sessionid->sessionid);
2846        spin_lock(&nn->client_lock);
2847        ses = find_in_sessionid_hashtbl(&sessionid->sessionid, net, &status);
2848        if (!ses)
2849                goto out_client_lock;
2850        status = nfserr_wrong_cred;
2851        if (!mach_creds_match(ses->se_client, r))
2852                goto out_put_session;
2853        status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
2854        if (status)
2855                goto out_put_session;
2856        unhash_session(ses);
2857        spin_unlock(&nn->client_lock);
2858
2859        nfsd4_probe_callback_sync(ses->se_client);
2860
2861        spin_lock(&nn->client_lock);
2862        status = nfs_ok;
2863out_put_session:
2864        nfsd4_put_session_locked(ses);
2865out_client_lock:
2866        spin_unlock(&nn->client_lock);
2867out:
2868        return status;
2869}
2870
2871static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
2872{
2873        struct nfsd4_conn *c;
2874
2875        list_for_each_entry(c, &s->se_conns, cn_persession) {
2876                if (c->cn_xprt == xpt) {
2877                        return c;
2878                }
2879        }
2880        return NULL;
2881}
2882
2883static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
2884{
2885        struct nfs4_client *clp = ses->se_client;
2886        struct nfsd4_conn *c;
2887        __be32 status = nfs_ok;
2888        int ret;
2889
2890        spin_lock(&clp->cl_lock);
2891        c = __nfsd4_find_conn(new->cn_xprt, ses);
2892        if (c)
2893                goto out_free;
2894        status = nfserr_conn_not_bound_to_session;
2895        if (clp->cl_mach_cred)
2896                goto out_free;
2897        __nfsd4_hash_conn(new, ses);
2898        spin_unlock(&clp->cl_lock);
2899        ret = nfsd4_register_conn(new);
2900        if (ret)
2901                /* oops; xprt is already down: */
2902                nfsd4_conn_lost(&new->cn_xpt_user);
2903        return nfs_ok;
2904out_free:
2905        spin_unlock(&clp->cl_lock);
2906        free_conn(new);
2907        return status;
2908}
2909
2910static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
2911{
2912        struct nfsd4_compoundargs *args = rqstp->rq_argp;
2913
2914        return args->opcnt > session->se_fchannel.maxops;
2915}
2916
2917static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
2918                                  struct nfsd4_session *session)
2919{
2920        struct xdr_buf *xb = &rqstp->rq_arg;
2921
2922        return xb->len > session->se_fchannel.maxreq_sz;
2923}
2924
2925__be32
2926nfsd4_sequence(struct svc_rqst *rqstp,
2927               struct nfsd4_compound_state *cstate,
2928               struct nfsd4_sequence *seq)
2929{
2930        struct nfsd4_compoundres *resp = rqstp->rq_resp;
2931        struct xdr_stream *xdr = &resp->xdr;
2932        struct nfsd4_session *session;
2933        struct nfs4_client *clp;
2934        struct nfsd4_slot *slot;
2935        struct nfsd4_conn *conn;
2936        __be32 status;
2937        int buflen;
2938        struct net *net = SVC_NET(rqstp);
2939        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2940
2941        if (resp->opcnt != 1)
2942                return nfserr_sequence_pos;
2943
2944        /*
2945         * Will be either used or freed by nfsd4_sequence_check_conn
2946         * below.
2947         */
2948        conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
2949        if (!conn)
2950                return nfserr_jukebox;
2951
2952        spin_lock(&nn->client_lock);
2953        session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
2954        if (!session)
2955                goto out_no_session;
2956        clp = session->se_client;
2957
2958        status = nfserr_too_many_ops;
2959        if (nfsd4_session_too_many_ops(rqstp, session))
2960                goto out_put_session;
2961
2962        status = nfserr_req_too_big;
2963        if (nfsd4_request_too_big(rqstp, session))
2964                goto out_put_session;
2965
2966        status = nfserr_badslot;
2967        if (seq->slotid >= session->se_fchannel.maxreqs)
2968                goto out_put_session;
2969
2970        slot = session->se_slots[seq->slotid];
2971        dprintk("%s: slotid %d\n", __func__, seq->slotid);
2972
2973        /* We do not negotiate the number of slots yet, so set the
2974         * maxslots to the session maxreqs which is used to encode
2975         * sr_highest_slotid and the sr_target_slot id to maxslots */
2976        seq->maxslots = session->se_fchannel.maxreqs;
2977
2978        status = check_slot_seqid(seq->seqid, slot->sl_seqid,
2979                                        slot->sl_flags & NFSD4_SLOT_INUSE);
2980        if (status == nfserr_replay_cache) {
2981                status = nfserr_seq_misordered;
2982                if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
2983                        goto out_put_session;
2984                cstate->slot = slot;
2985                cstate->session = session;
2986                cstate->clp = clp;
2987                /* Return the cached reply status and set cstate->status
2988                 * for nfsd4_proc_compound processing */
2989                status = nfsd4_replay_cache_entry(resp, seq);
2990                cstate->status = nfserr_replay_cache;
2991                goto out;
2992        }
2993        if (status)
2994                goto out_put_session;
2995
2996        status = nfsd4_sequence_check_conn(conn, session);
2997        conn = NULL;
2998        if (status)
2999                goto out_put_session;
3000
3001        buflen = (seq->cachethis) ?
3002                        session->se_fchannel.maxresp_cached :
3003                        session->se_fchannel.maxresp_sz;
3004        status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
3005                                    nfserr_rep_too_big;
3006        if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
3007                goto out_put_session;
3008        svc_reserve(rqstp, buflen);
3009
3010        status = nfs_ok;
3011        /* Success! bump slot seqid */
3012        slot->sl_seqid = seq->seqid;
3013        slot->sl_flags |= NFSD4_SLOT_INUSE;
3014        if (seq->cachethis)
3015                slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
3016        else
3017                slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
3018
3019        cstate->slot = slot;
3020        cstate->session = session;
3021        cstate->clp = clp;
3022
3023out:
3024        switch (clp->cl_cb_state) {
3025        case NFSD4_CB_DOWN:
3026                seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
3027                break;
3028        case NFSD4_CB_FAULT:
3029                seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
3030                break;
3031        default:
3032                seq->status_flags = 0;
3033        }
3034        if (!list_empty(&clp->cl_revoked))
3035                seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
3036out_no_session:
3037        if (conn)
3038                free_conn(conn);
3039        spin_unlock(&nn->client_lock);
3040        return status;
3041out_put_session:
3042        nfsd4_put_session_locked(session);
3043        goto out_no_session;
3044}
3045
3046void
3047nfsd4_sequence_done(struct nfsd4_compoundres *resp)
3048{
3049        struct nfsd4_compound_state *cs = &resp->cstate;
3050
3051        if (nfsd4_has_session(cs)) {
3052                if (cs->status != nfserr_replay_cache) {
3053                        nfsd4_store_cache_entry(resp);
3054                        cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
3055                }
3056                /* Drop session reference that was taken in nfsd4_sequence() */
3057                nfsd4_put_session(cs->session);
3058        } else if (cs->clp)
3059                put_client_renew(cs->clp);
3060}
3061
3062__be32
3063nfsd4_destroy_clientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_destroy_clientid *dc)
3064{
3065        struct nfs4_client *conf, *unconf;
3066        struct nfs4_client *clp = NULL;
3067        __be32 status = 0;
3068        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3069
3070        spin_lock(&nn->client_lock);
3071        unconf = find_unconfirmed_client(&dc->clientid, true, nn);
3072        conf = find_confirmed_client(&dc->clientid, true, nn);
3073        WARN_ON_ONCE(conf && unconf);
3074
3075        if (conf) {
3076                if (client_has_state(conf)) {
3077                        status = nfserr_clientid_busy;
3078                        goto out;
3079                }
3080                status = mark_client_expired_locked(conf);
3081                if (status)
3082                        goto out;
3083                clp = conf;
3084        } else if (unconf)
3085                clp = unconf;
3086        else {
3087                status = nfserr_stale_clientid;
3088                goto out;
3089        }
3090        if (!mach_creds_match(clp, rqstp)) {
3091                clp = NULL;
3092                status = nfserr_wrong_cred;
3093                goto out;
3094        }
3095        unhash_client_locked(clp);
3096out:
3097        spin_unlock(&nn->client_lock);
3098        if (clp)
3099                expire_client(clp);
3100        return status;
3101}
3102
3103__be32
3104nfsd4_reclaim_complete(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate, struct nfsd4_reclaim_complete *rc)
3105{
3106        __be32 status = 0;
3107
3108        if (rc->rca_one_fs) {
3109                if (!cstate->current_fh.fh_dentry)
3110                        return nfserr_nofilehandle;
3111                /*
3112                 * We don't take advantage of the rca_one_fs case.
3113                 * That's OK, it's optional, we can safely ignore it.
3114                 */
3115                 return nfs_ok;
3116        }
3117
3118        status = nfserr_complete_already;
3119        if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE,
3120                             &cstate->session->se_client->cl_flags))
3121                goto out;
3122
3123        status = nfserr_stale_clientid;
3124        if (is_client_expired(cstate->session->se_client))
3125                /*
3126                 * The following error isn't really legal.
3127                 * But we only get here if the client just explicitly
3128                 * destroyed the client.  Surely it no longer cares what
3129                 * error it gets back on an operation for the dead
3130                 * client.
3131                 */
3132                goto out;
3133
3134        status = nfs_ok;
3135        nfsd4_client_record_create(cstate->session->se_client);
3136out:
3137        return status;
3138}
3139
3140__be32
3141nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3142                  struct nfsd4_setclientid *setclid)
3143{
3144        struct xdr_netobj       clname = setclid->se_name;
3145        nfs4_verifier           clverifier = setclid->se_verf;
3146        struct nfs4_client      *conf, *new;
3147        struct nfs4_client      *unconf = NULL;
3148        __be32                  status;
3149        struct nfsd_net         *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3150
3151        new = create_client(clname, rqstp, &clverifier);
3152        if (new == NULL)
3153                return nfserr_jukebox;
3154        /* Cases below refer to rfc 3530 section 14.2.33: */
3155        spin_lock(&nn->client_lock);
3156        conf = find_confirmed_client_by_name(&clname, nn);
3157        if (conf && client_has_state(conf)) {
3158                /* case 0: */
3159                status = nfserr_clid_inuse;
3160                if (clp_used_exchangeid(conf))
3161                        goto out;
3162                if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
3163                        char addr_str[INET6_ADDRSTRLEN];
3164                        rpc_ntop((struct sockaddr *) &conf->cl_addr, addr_str,
3165                                 sizeof(addr_str));
3166                        dprintk("NFSD: setclientid: string in use by client "
3167                                "at %s\n", addr_str);
3168                        goto out;
3169                }
3170        }
3171        unconf = find_unconfirmed_client_by_name(&clname, nn);
3172        if (unconf)
3173                unhash_client_locked(unconf);
3174        if (conf && same_verf(&conf->cl_verifier, &clverifier)) {
3175                /* case 1: probable callback update */
3176                copy_clid(new, conf);
3177                gen_confirm(new, nn);
3178        } else /* case 4 (new client) or cases 2, 3 (client reboot): */
3179                gen_clid(new, nn);
3180        new->cl_minorversion = 0;
3181        gen_callback(new, setclid, rqstp);
3182        add_to_unconfirmed(new);
3183        setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
3184        setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
3185        memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
3186        new = NULL;
3187        status = nfs_ok;
3188out:
3189        spin_unlock(&nn->client_lock);
3190        if (new)
3191                free_client(new);
3192        if (unconf)
3193                expire_client(unconf);
3194        return status;
3195}
3196
3197
3198__be32
3199nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
3200                         struct nfsd4_compound_state *cstate,
3201                         struct nfsd4_setclientid_confirm *setclientid_confirm)
3202{
3203        struct nfs4_client *conf, *unconf;
3204        struct nfs4_client *old = NULL;
3205        nfs4_verifier confirm = setclientid_confirm->sc_confirm; 
3206        clientid_t * clid = &setclientid_confirm->sc_clientid;
3207        __be32 status;
3208        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3209
3210        if (STALE_CLIENTID(clid, nn))
3211                return nfserr_stale_clientid;
3212
3213        spin_lock(&nn->client_lock);
3214        conf = find_confirmed_client(clid, false, nn);
3215        unconf = find_unconfirmed_client(clid, false, nn);
3216        /*
3217         * We try hard to give out unique clientid's, so if we get an
3218         * attempt to confirm the same clientid with a different cred,
3219         * the client may be buggy; this should never happen.
3220         *
3221         * Nevertheless, RFC 7530 recommends INUSE for this case:
3222         */
3223        status = nfserr_clid_inuse;
3224        if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred))
3225                goto out;
3226        if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred))
3227                goto out;
3228        /* cases below refer to rfc 3530 section 14.2.34: */
3229        if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
3230                if (conf && !unconf) /* case 2: probable retransmit */
3231                        status = nfs_ok;
3232                else /* case 4: client hasn't noticed we rebooted yet? */
3233                        status = nfserr_stale_clientid;
3234                goto out;
3235        }
3236        status = nfs_ok;
3237        if (conf) { /* case 1: callback update */
3238                old = unconf;
3239                unhash_client_locked(old);
3240                nfsd4_change_callback(conf, &unconf->cl_cb_conn);
3241        } else { /* case 3: normal case; new or rebooted client */
3242                old = find_confirmed_client_by_name(&unconf->cl_name, nn);
3243                if (old) {
3244                        status = nfserr_clid_inuse;
3245                        if (client_has_state(old)
3246                                        && !same_creds(&unconf->cl_cred,
3247                                                        &old->cl_cred))
3248                                goto out;
3249                        status = mark_client_expired_locked(old);
3250                        if (status) {
3251                                old = NULL;
3252                                goto out;
3253                        }
3254                }
3255                move_to_confirmed(unconf);
3256                conf = unconf;
3257        }
3258        get_client_locked(conf);
3259        spin_unlock(&nn->client_lock);
3260        nfsd4_probe_callback(conf);
3261        spin_lock(&nn->client_lock);
3262        put_client_renew_locked(conf);
3263out:
3264        spin_unlock(&nn->client_lock);
3265        if (old)
3266                expire_client(old);
3267        return status;
3268}
3269
3270static struct nfs4_file *nfsd4_alloc_file(void)
3271{
3272        return kmem_cache_alloc(file_slab, GFP_KERNEL);
3273}
3274
3275/* OPEN Share state helper functions */
3276static void nfsd4_init_file(struct knfsd_fh *fh, unsigned int hashval,
3277                                struct nfs4_file *fp)
3278{
3279        lockdep_assert_held(&state_lock);
3280
3281        atomic_set(&fp->fi_ref, 1);
3282        spin_lock_init(&fp->fi_lock);
3283        INIT_LIST_HEAD(&fp->fi_stateids);
3284        INIT_LIST_HEAD(&fp->fi_delegations);
3285        INIT_LIST_HEAD(&fp->fi_clnt_odstate);
3286        fh_copy_shallow(&fp->fi_fhandle, fh);
3287        fp->fi_deleg_file = NULL;
3288        fp->fi_had_conflict = false;
3289        fp->fi_share_deny = 0;
3290        memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
3291        memset(fp->fi_access, 0, sizeof(fp->fi_access));
3292#ifdef CONFIG_NFSD_PNFS
3293        INIT_LIST_HEAD(&fp->fi_lo_states);
3294        atomic_set(&fp->fi_lo_recalls, 0);
3295#endif
3296        hlist_add_head_rcu(&fp->fi_hash, &file_hashtbl[hashval]);
3297}
3298
3299void
3300nfsd4_free_slabs(void)
3301{
3302        kmem_cache_destroy(odstate_slab);
3303        kmem_cache_destroy(openowner_slab);
3304        kmem_cache_destroy(lockowner_slab);
3305        kmem_cache_destroy(file_slab);
3306        kmem_cache_destroy(stateid_slab);
3307        kmem_cache_destroy(deleg_slab);
3308}
3309
3310int
3311nfsd4_init_slabs(void)
3312{
3313        openowner_slab = kmem_cache_create("nfsd4_openowners",
3314                        sizeof(struct nfs4_openowner), 0, 0, NULL);
3315        if (openowner_slab == NULL)
3316                goto out;
3317        lockowner_slab = kmem_cache_create("nfsd4_lockowners",
3318                        sizeof(struct nfs4_lockowner), 0, 0, NULL);
3319        if (lockowner_slab == NULL)
3320                goto out_free_openowner_slab;
3321        file_slab = kmem_cache_create("nfsd4_files",
3322                        sizeof(struct nfs4_file), 0, 0, NULL);
3323        if (file_slab == NULL)
3324                goto out_free_lockowner_slab;
3325        stateid_slab = kmem_cache_create("nfsd4_stateids",
3326                        sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
3327        if (stateid_slab == NULL)
3328                goto out_free_file_slab;
3329        deleg_slab = kmem_cache_create("nfsd4_delegations",
3330                        sizeof(struct nfs4_delegation), 0, 0, NULL);
3331        if (deleg_slab == NULL)
3332                goto out_free_stateid_slab;
3333        odstate_slab = kmem_cache_create("nfsd4_odstate",
3334                        sizeof(struct nfs4_clnt_odstate), 0, 0, NULL);
3335        if (odstate_slab == NULL)
3336                goto out_free_deleg_slab;
3337        return 0;
3338
3339out_free_deleg_slab:
3340        kmem_cache_destroy(deleg_slab);
3341out_free_stateid_slab:
3342        kmem_cache_destroy(stateid_slab);
3343out_free_file_slab:
3344        kmem_cache_destroy(file_slab);
3345out_free_lockowner_slab:
3346        kmem_cache_destroy(lockowner_slab);
3347out_free_openowner_slab:
3348        kmem_cache_destroy(openowner_slab);
3349out:
3350        dprintk("nfsd4: out of memory while initializing nfsv4\n");
3351        return -ENOMEM;
3352}
3353
3354static void init_nfs4_replay(struct nfs4_replay *rp)
3355{
3356        rp->rp_status = nfserr_serverfault;
3357        rp->rp_buflen = 0;
3358        rp->rp_buf = rp->rp_ibuf;
3359        mutex_init(&rp->rp_mutex);
3360}
3361
3362static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
3363                struct nfs4_stateowner *so)
3364{
3365        if (!nfsd4_has_session(cstate)) {
3366                mutex_lock(&so->so_replay.rp_mutex);
3367                cstate->replay_owner = nfs4_get_stateowner(so);
3368        }
3369}
3370
3371void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
3372{
3373        struct nfs4_stateowner *so = cstate->replay_owner;
3374
3375        if (so != NULL) {
3376                cstate->replay_owner = NULL;
3377                mutex_unlock(&so->so_replay.rp_mutex);
3378                nfs4_put_stateowner(so);
3379        }
3380}
3381
3382static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
3383{
3384        struct nfs4_stateowner *sop;
3385
3386        sop = kmem_cache_alloc(slab, GFP_KERNEL);
3387        if (!sop)
3388                return NULL;
3389
3390        sop->so_owner.data = kmemdup(owner->data, owner->len, GFP_KERNEL);
3391        if (!sop->so_owner.data) {
3392                kmem_cache_free(slab, sop);
3393                return NULL;
3394        }
3395        sop->so_owner.len = owner->len;
3396
3397        INIT_LIST_HEAD(&sop->so_stateids);
3398        sop->so_client = clp;
3399        init_nfs4_replay(&sop->so_replay);
3400        atomic_set(&sop->so_count, 1);
3401        return sop;
3402}
3403
3404static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
3405{
3406        lockdep_assert_held(&clp->cl_lock);
3407
3408        list_add(&oo->oo_owner.so_strhash,
3409                 &clp->cl_ownerstr_hashtbl[strhashval]);
3410        list_add(&oo->oo_perclient, &clp->cl_openowners);
3411}
3412
3413static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
3414{
3415        unhash_openowner_locked(openowner(so));
3416}
3417
3418static void nfs4_free_openowner(struct nfs4_stateowner *so)
3419{
3420        struct nfs4_openowner *oo = openowner(so);
3421
3422        kmem_cache_free(openowner_slab, oo);
3423}
3424
3425static const struct nfs4_stateowner_operations openowner_ops = {
3426        .so_unhash =    nfs4_unhash_openowner,
3427        .so_free =      nfs4_free_openowner,
3428};
3429
3430static struct nfs4_ol_stateid *
3431nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
3432{
3433        struct nfs4_ol_stateid *local, *ret = NULL;
3434        struct nfs4_openowner *oo = open->op_openowner;
3435
3436        lockdep_assert_held(&fp->fi_lock);
3437
3438        list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
3439                /* ignore lock owners */
3440                if (local->st_stateowner->so_is_open_owner == 0)
3441                        continue;
3442                if (local->st_stateowner == &oo->oo_owner) {
3443                        ret = local;
3444                        atomic_inc(&ret->st_stid.sc_count);
3445                        break;
3446                }
3447        }
3448        return ret;
3449}
3450
3451static struct nfs4_openowner *
3452alloc_init_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
3453                           struct nfsd4_compound_state *cstate)
3454{
3455        struct nfs4_client *clp = cstate->clp;
3456        struct nfs4_openowner *oo, *ret;
3457
3458        oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
3459        if (!oo)
3460                return NULL;
3461        oo->oo_owner.so_ops = &openowner_ops;
3462        oo->oo_owner.so_is_open_owner = 1;
3463        oo->oo_owner.so_seqid = open->op_seqid;
3464        oo->oo_flags = 0;
3465        if (nfsd4_has_session(cstate))
3466                oo->oo_flags |= NFS4_OO_CONFIRMED;
3467        oo->oo_time = 0;
3468        oo->oo_last_closed_stid = NULL;
3469        INIT_LIST_HEAD(&oo->oo_close_lru);
3470        spin_lock(&clp->cl_lock);
3471        ret = find_openstateowner_str_locked(strhashval, open, clp);
3472        if (ret == NULL) {
3473                hash_openowner(oo, clp, strhashval);
3474                ret = oo;
3475        } else
3476                nfs4_free_stateowner(&oo->oo_owner);
3477
3478        spin_unlock(&clp->cl_lock);
3479        return ret;
3480}
3481
3482static struct nfs4_ol_stateid *
3483init_open_stateid(struct nfs4_ol_stateid *stp, struct nfs4_file *fp,
3484                struct nfsd4_open *open)
3485{
3486
3487        struct nfs4_openowner *oo = open->op_openowner;
3488        struct nfs4_ol_stateid *retstp = NULL;
3489
3490        spin_lock(&oo->oo_owner.so_client->cl_lock);
3491        spin_lock(&fp->fi_lock);
3492
3493        retstp = nfsd4_find_existing_open(fp, open);
3494        if (retstp)
3495                goto out_unlock;
3496        atomic_inc(&stp->st_stid.sc_count);
3497        stp->st_stid.sc_type = NFS4_OPEN_STID;
3498        INIT_LIST_HEAD(&stp->st_locks);
3499        stp->st_stateowner = nfs4_get_stateowner(&oo->oo_owner);
3500        get_nfs4_file(fp);
3501        stp->st_stid.sc_file = fp;
3502        stp->st_access_bmap = 0;
3503        stp->st_deny_bmap = 0;
3504        stp->st_openstp = NULL;
3505        init_rwsem(&stp->st_rwsem);
3506        list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
3507        list_add(&stp->st_perfile, &fp->fi_stateids);
3508
3509out_unlock:
3510        spin_unlock(&fp->fi_lock);
3511        spin_unlock(&oo->oo_owner.so_client->cl_lock);
3512        return retstp;
3513}
3514
3515/*
3516 * In the 4.0 case we need to keep the owners around a little while to handle
3517 * CLOSE replay. We still do need to release any file access that is held by
3518 * them before returning however.
3519 */
3520static void
3521move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
3522{
3523        struct nfs4_ol_stateid *last;
3524        struct nfs4_openowner *oo = openowner(s->st_stateowner);
3525        struct nfsd_net *nn = net_generic(s->st_stid.sc_client->net,
3526                                                nfsd_net_id);
3527
3528        dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
3529
3530        /*
3531         * We know that we hold one reference via nfsd4_close, and another
3532         * "persistent" reference for the client. If the refcount is higher
3533         * than 2, then there are still calls in progress that are using this
3534         * stateid. We can't put the sc_file reference until they are finished.
3535         * Wait for the refcount to drop to 2. Since it has been unhashed,
3536         * there should be no danger of the refcount going back up again at
3537         * this point.
3538         */
3539        wait_event(close_wq, atomic_read(&s->st_stid.sc_count) == 2);
3540
3541        release_all_access(s);
3542        if (s->st_stid.sc_file) {
3543                put_nfs4_file(s->st_stid.sc_file);
3544                s->st_stid.sc_file = NULL;
3545        }
3546
3547        spin_lock(&nn->client_lock);
3548        last = oo->oo_last_closed_stid;
3549        oo->oo_last_closed_stid = s;
3550        list_move_tail(&oo->oo_close_lru, &nn->close_lru);
3551        oo->oo_time = get_seconds();
3552        spin_unlock(&nn->client_lock);
3553        if (last)
3554                nfs4_put_stid(&last->st_stid);
3555}
3556
3557/* search file_hashtbl[] for file */
3558static struct nfs4_file *
3559find_file_locked(struct knfsd_fh *fh, unsigned int hashval)
3560{
3561        struct nfs4_file *fp;
3562
3563        hlist_for_each_entry_rcu(fp, &file_hashtbl[hashval], fi_hash) {
3564                if (fh_match(&fp->fi_fhandle, fh)) {
3565                        if (atomic_inc_not_zero(&fp->fi_ref))
3566                                return fp;
3567                }
3568        }
3569        return NULL;
3570}
3571
3572struct nfs4_file *
3573find_file(struct knfsd_fh *fh)
3574{
3575        struct nfs4_file *fp;
3576        unsigned int hashval = file_hashval(fh);
3577
3578        rcu_read_lock();
3579        fp = find_file_locked(fh, hashval);
3580        rcu_read_unlock();
3581        return fp;
3582}
3583
3584static struct nfs4_file *
3585find_or_add_file(struct nfs4_file *new, struct knfsd_fh *fh)
3586{
3587        struct nfs4_file *fp;
3588        unsigned int hashval = file_hashval(fh);
3589
3590        rcu_read_lock();
3591        fp = find_file_locked(fh, hashval);
3592        rcu_read_unlock();
3593        if (fp)
3594                return fp;
3595
3596        spin_lock(&state_lock);
3597        fp = find_file_locked(fh, hashval);
3598        if (likely(fp == NULL)) {
3599                nfsd4_init_file(fh, hashval, new);
3600                fp = new;
3601        }
3602        spin_unlock(&state_lock);
3603
3604        return fp;
3605}
3606
3607/*
3608 * Called to check deny when READ with all zero stateid or
3609 * WRITE with all zero or all one stateid
3610 */
3611static __be32
3612nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
3613{
3614        struct nfs4_file *fp;
3615        __be32 ret = nfs_ok;
3616
3617        fp = find_file(&current_fh->fh_handle);
3618        if (!fp)
3619                return ret;
3620        /* Check for conflicting share reservations */
3621        spin_lock(&fp->fi_lock);
3622        if (fp->fi_share_deny & deny_type)
3623                ret = nfserr_locked;
3624        spin_unlock(&fp->fi_lock);
3625        put_nfs4_file(fp);
3626        return ret;
3627}
3628
3629static void nfsd4_cb_recall_prepare(struct nfsd4_callback *cb)
3630{
3631        struct nfs4_delegation *dp = cb_to_delegation(cb);
3632        struct nfsd_net *nn = net_generic(dp->dl_stid.sc_client->net,
3633                                          nfsd_net_id);
3634
3635        block_delegations(&dp->dl_stid.sc_file->fi_fhandle);
3636
3637        /*
3638         * We can't do this in nfsd_break_deleg_cb because it is
3639         * already holding inode->i_lock.
3640         *
3641         * If the dl_time != 0, then we know that it has already been
3642         * queued for a lease break. Don't queue it again.
3643         */
3644        spin_lock(&state_lock);
3645        if (dp->dl_time == 0) {
3646                dp->dl_time = get_seconds();
3647                list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
3648        }
3649        spin_unlock(&state_lock);
3650}
3651
3652static int nfsd4_cb_recall_done(struct nfsd4_callback *cb,
3653                struct rpc_task *task)
3654{
3655        struct nfs4_delegation *dp = cb_to_delegation(cb);
3656
3657        if (dp->dl_stid.sc_type == NFS4_CLOSED_DELEG_STID)
3658                return 1;
3659
3660        switch (task->tk_status) {
3661        case 0:
3662                return 1;
3663        case -EBADHANDLE:
3664        case -NFS4ERR_BAD_STATEID:
3665                /*
3666                 * Race: client probably got cb_recall before open reply
3667                 * granting delegation.
3668                 */
3669                if (dp->dl_retries--) {
3670                        rpc_delay(task, 2 * HZ);
3671                        return 0;
3672                }
3673                /*FALLTHRU*/
3674        default:
3675                return -1;
3676        }
3677}
3678
3679static void nfsd4_cb_recall_release(struct nfsd4_callback *cb)
3680{
3681        struct nfs4_delegation *dp = cb_to_delegation(cb);
3682
3683        nfs4_put_stid(&dp->dl_stid);
3684}
3685
3686static const struct nfsd4_callback_ops nfsd4_cb_recall_ops = {
3687        .prepare        = nfsd4_cb_recall_prepare,
3688        .done           = nfsd4_cb_recall_done,
3689        .release        = nfsd4_cb_recall_release,
3690};
3691
3692static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
3693{
3694        /*
3695         * We're assuming the state code never drops its reference
3696         * without first removing the lease.  Since we're in this lease
3697         * callback (and since the lease code is serialized by the kernel
3698         * lock) we know the server hasn't removed the lease yet, we know
3699         * it's safe to take a reference.
3700         */
3701        atomic_inc(&dp->dl_stid.sc_count);
3702        nfsd4_run_cb(&dp->dl_recall);
3703}
3704
3705/* Called from break_lease() with i_lock held. */
3706static bool
3707nfsd_break_deleg_cb(struct file_lock *fl)
3708{
3709        bool ret = false;
3710        struct nfs4_file *fp = (struct nfs4_file *)fl->fl_owner;
3711        struct nfs4_delegation *dp;
3712
3713        if (!fp) {
3714                WARN(1, "(%p)->fl_owner NULL\n", fl);
3715                return ret;
3716        }
3717        if (fp->fi_had_conflict) {
3718                WARN(1, "duplicate break on %p\n", fp);
3719                return ret;
3720        }
3721        /*
3722         * We don't want the locks code to timeout the lease for us;
3723         * we'll remove it ourself if a delegation isn't returned
3724         * in time:
3725         */
3726        fl->fl_break_time = 0;
3727
3728        spin_lock(&fp->fi_lock);
3729        fp->fi_had_conflict = true;
3730        /*
3731         * If there are no delegations on the list, then return true
3732         * so that the lease code will go ahead and delete it.
3733         */
3734        if (list_empty(&fp->fi_delegations))
3735                ret = true;
3736        else
3737                list_for_each_entry(dp, &fp->fi_delegations, dl_perfile)
3738                        nfsd_break_one_deleg(dp);
3739        spin_unlock(&fp->fi_lock);
3740        return ret;
3741}
3742
3743static int
3744nfsd_change_deleg_cb(struct file_lock *onlist, int arg,
3745                     struct list_head *dispose)
3746{
3747        if (arg & F_UNLCK)
3748                return lease_modify(onlist, arg, dispose);
3749        else
3750                return -EAGAIN;
3751}
3752
3753static const struct lock_manager_operations nfsd_lease_mng_ops = {
3754        .lm_break = nfsd_break_deleg_cb,
3755        .lm_change = nfsd_change_deleg_cb,
3756};
3757
3758static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
3759{
3760        if (nfsd4_has_session(cstate))
3761                return nfs_ok;
3762        if (seqid == so->so_seqid - 1)
3763                return nfserr_replay_me;
3764        if (seqid == so->so_seqid)
3765                return nfs_ok;
3766        return nfserr_bad_seqid;
3767}
3768
3769static __be32 lookup_clientid(clientid_t *clid,
3770                struct nfsd4_compound_state *cstate,
3771                struct nfsd_net *nn)
3772{
3773        struct nfs4_client *found;
3774
3775        if (cstate->clp) {
3776                found = cstate->clp;
3777                if (!same_clid(&found->cl_clientid, clid))
3778                        return nfserr_stale_clientid;
3779                return nfs_ok;
3780        }
3781
3782        if (STALE_CLIENTID(clid, nn))
3783                return nfserr_stale_clientid;
3784
3785        /*
3786         * For v4.1+ we get the client in the SEQUENCE op. If we don't have one
3787         * cached already then we know this is for is for v4.0 and "sessions"
3788         * will be false.
3789         */
3790        WARN_ON_ONCE(cstate->session);
3791        spin_lock(&nn->client_lock);
3792        found = find_confirmed_client(clid, false, nn);
3793        if (!found) {
3794                spin_unlock(&nn->client_lock);
3795                return nfserr_expired;
3796        }
3797        atomic_inc(&found->cl_refcount);
3798        spin_unlock(&nn->client_lock);
3799
3800        /* Cache the nfs4_client in cstate! */
3801        cstate->clp = found;
3802        return nfs_ok;
3803}
3804
3805__be32
3806nfsd4_process_open1(struct nfsd4_compound_state *cstate,
3807                    struct nfsd4_open *open, struct nfsd_net *nn)
3808{
3809        clientid_t *clientid = &open->op_clientid;
3810        struct nfs4_client *clp = NULL;
3811        unsigned int strhashval;
3812        struct nfs4_openowner *oo = NULL;
3813        __be32 status;
3814
3815        if (STALE_CLIENTID(&open->op_clientid, nn))
3816                return nfserr_stale_clientid;
3817        /*
3818         * In case we need it later, after we've already created the
3819         * file and don't want to risk a further failure:
3820         */
3821        open->op_file = nfsd4_alloc_file();
3822        if (open->op_file == NULL)
3823                return nfserr_jukebox;
3824
3825        status = lookup_clientid(clientid, cstate, nn);
3826        if (status)
3827                return status;
3828        clp = cstate->clp;
3829
3830        strhashval = ownerstr_hashval(&open->op_owner);
3831        oo = find_openstateowner_str(strhashval, open, clp);
3832        open->op_openowner = oo;
3833        if (!oo) {
3834                goto new_owner;
3835        }
3836        if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
3837                /* Replace unconfirmed owners without checking for replay. */
3838                release_openowner(oo);
3839                open->op_openowner = NULL;
3840                goto new_owner;
3841        }
3842        status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
3843        if (status)
3844                return status;
3845        goto alloc_stateid;
3846new_owner:
3847        oo = alloc_init_open_stateowner(strhashval, open, cstate);
3848        if (oo == NULL)
3849                return nfserr_jukebox;
3850        open->op_openowner = oo;
3851alloc_stateid:
3852        open->op_stp = nfs4_alloc_open_stateid(clp);
3853        if (!open->op_stp)
3854                return nfserr_jukebox;
3855
3856        if (nfsd4_has_session(cstate) &&
3857            (cstate->current_fh.fh_export->ex_flags & NFSEXP_PNFS)) {
3858                open->op_odstate = alloc_clnt_odstate(clp);
3859                if (!open->op_odstate)
3860                        return nfserr_jukebox;
3861        }
3862
3863        return nfs_ok;
3864}
3865
3866static inline __be32
3867nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
3868{
3869        if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
3870                return nfserr_openmode;
3871        else
3872                return nfs_ok;
3873}
3874
3875static int share_access_to_flags(u32 share_access)
3876{
3877        return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
3878}
3879
3880static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
3881{
3882        struct nfs4_stid *ret;
3883
3884        ret = find_stateid_by_type(cl, s, NFS4_DELEG_STID);
3885        if (!ret)
3886                return NULL;
3887        return delegstateid(ret);
3888}
3889
3890static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
3891{
3892        return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
3893               open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
3894}
3895
3896static __be32
3897nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
3898                struct nfs4_delegation **dp)
3899{
3900        int flags;
3901        __be32 status = nfserr_bad_stateid;
3902        struct nfs4_delegation *deleg;
3903
3904        deleg = find_deleg_stateid(cl, &open->op_delegate_stateid);
3905        if (deleg == NULL)
3906                goto out;
3907        flags = share_access_to_flags(open->op_share_access);
3908        status = nfs4_check_delegmode(deleg, flags);
3909        if (status) {
3910                nfs4_put_stid(&deleg->dl_stid);
3911                goto out;
3912        }
3913        *dp = deleg;
3914out:
3915        if (!nfsd4_is_deleg_cur(open))
3916                return nfs_ok;
3917        if (status)
3918                return status;
3919        open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
3920        return nfs_ok;
3921}
3922
3923static inline int nfs4_access_to_access(u32 nfs4_access)
3924{
3925        int flags = 0;
3926
3927        if (nfs4_access & NFS4_SHARE_ACCESS_READ)
3928                flags |= NFSD_MAY_READ;
3929        if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
3930                flags |= NFSD_MAY_WRITE;
3931        return flags;
3932}
3933
3934static inline __be32
3935nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
3936                struct nfsd4_open *open)
3937{
3938        struct iattr iattr = {
3939                .ia_valid = ATTR_SIZE,
3940                .ia_size = 0,
3941        };
3942        if (!open->op_truncate)
3943                return 0;
3944        if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
3945                return nfserr_inval;
3946        return nfsd_setattr(rqstp, fh, &iattr, 0, (time_t)0);
3947}
3948
3949static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
3950                struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
3951                struct nfsd4_open *open)
3952{
3953        struct file *filp = NULL;
3954        __be32 status;
3955        int oflag = nfs4_access_to_omode(open->op_share_access);
3956        int access = nfs4_access_to_access(open->op_share_access);
3957        unsigned char old_access_bmap, old_deny_bmap;
3958
3959        spin_lock(&fp->fi_lock);
3960
3961        /*
3962         * Are we trying to set a deny mode that would conflict with
3963         * current access?
3964         */
3965        status = nfs4_file_check_deny(fp, open->op_share_deny);
3966        if (status != nfs_ok) {
3967                spin_unlock(&fp->fi_lock);
3968                goto out;
3969        }
3970
3971        /* set access to the file */
3972        status = nfs4_file_get_access(fp, open->op_share_access);
3973        if (status != nfs_ok) {
3974                spin_unlock(&fp->fi_lock);
3975                goto out;
3976        }
3977
3978        /* Set access bits in stateid */
3979        old_access_bmap = stp->st_access_bmap;
3980        set_access(open->op_share_access, stp);
3981
3982        /* Set new deny mask */
3983        old_deny_bmap = stp->st_deny_bmap;
3984        set_deny(open->op_share_deny, stp);
3985        fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
3986
3987        if (!fp->fi_fds[oflag]) {
3988                spin_unlock(&fp->fi_lock);
3989                status = nfsd_open(rqstp, cur_fh, S_IFREG, access, &filp);
3990                if (status)
3991                        goto out_put_access;
3992                spin_lock(&fp->fi_lock);
3993                if (!fp->fi_fds[oflag]) {
3994                        fp->fi_fds[oflag] = filp;
3995                        filp = NULL;
3996                }
3997        }
3998        spin_unlock(&fp->fi_lock);
3999        if (filp)
4000                fput(filp);
4001
4002        status = nfsd4_truncate(rqstp, cur_fh, open);
4003        if (status)
4004                goto out_put_access;
4005out:
4006        return status;
4007out_put_access:
4008        stp->st_access_bmap = old_access_bmap;
4009        nfs4_file_put_access(fp, open->op_share_access);
4010        reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap), stp);
4011        goto out;
4012}
4013
4014static __be32
4015nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp, struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp, struct nfsd4_open *open)
4016{
4017        __be32 status;
4018        unsigned char old_deny_bmap = stp->st_deny_bmap;
4019
4020        if (!test_access(open->op_share_access, stp))
4021                return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open);
4022
4023        /* test and set deny mode */
4024        spin_lock(&fp->fi_lock);
4025        status = nfs4_file_check_deny(fp, open->op_share_deny);
4026        if (status == nfs_ok) {
4027                set_deny(open->op_share_deny, stp);
4028                fp->fi_share_deny |=
4029                                (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
4030        }
4031        spin_unlock(&fp->fi_lock);
4032
4033        if (status != nfs_ok)
4034                return status;
4035
4036        status = nfsd4_truncate(rqstp, cur_fh, open);
4037        if (status != nfs_ok)
4038                reset_union_bmap_deny(old_deny_bmap, stp);
4039        return status;
4040}
4041
4042/* Should we give out recallable state?: */
4043static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
4044{
4045        if (clp->cl_cb_state == NFSD4_CB_UP)
4046                return true;
4047        /*
4048         * In the sessions case, since we don't have to establish a
4049         * separate connection for callbacks, we assume it's OK
4050         * until we hear otherwise:
4051         */
4052        return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
4053}
4054
4055static struct file_lock *nfs4_alloc_init_lease(struct nfs4_file *fp, int flag)
4056{
4057        struct file_lock *fl;
4058
4059        fl = locks_alloc_lock();
4060        if (!fl)
4061                return NULL;
4062        fl->fl_lmops = &nfsd_lease_mng_ops;
4063        fl->fl_flags = FL_DELEG;
4064        fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
4065        fl->fl_end = OFFSET_MAX;
4066        fl->fl_owner = (fl_owner_t)fp;
4067        fl->fl_pid = current->tgid;
4068        return fl;
4069}
4070
4071/**
4072 * nfs4_setlease - Obtain a delegation by requesting lease from vfs layer
4073 * @dp:   a pointer to the nfs4_delegation we're adding.
4074 *
4075 * Return:
4076 *      On success: Return code will be 0 on success.
4077 *
4078 *      On error: -EAGAIN if there was an existing delegation.
4079 *                 nonzero if there is an error in other cases.
4080 *
4081 */
4082
4083static int nfs4_setlease(struct nfs4_delegation *dp)
4084{
4085        struct nfs4_file *fp = dp->dl_stid.sc_file;
4086        struct file_lock *fl;
4087        struct file *filp;
4088        int status = 0;
4089
4090        fl = nfs4_alloc_init_lease(fp, NFS4_OPEN_DELEGATE_READ);
4091        if (!fl)
4092                return -ENOMEM;
4093        filp = find_readable_file(fp);
4094        if (!filp) {
4095                /* We should always have a readable file here */
4096                WARN_ON_ONCE(1);
4097                locks_free_lock(fl);
4098                return -EBADF;
4099        }
4100        fl->fl_file = filp;
4101        status = vfs_setlease(filp, fl->fl_type, &fl, NULL);
4102        if (fl)
4103                locks_free_lock(fl);
4104        if (status)
4105                goto out_fput;
4106        spin_lock(&state_lock);
4107        spin_lock(&fp->fi_lock);
4108        /* Did the lease get broken before we took the lock? */
4109        status = -EAGAIN;
4110        if (fp->fi_had_conflict)
4111                goto out_unlock;
4112        /* Race breaker */
4113        if (fp->fi_deleg_file) {
4114                status = hash_delegation_locked(dp, fp);
4115                goto out_unlock;
4116        }
4117        fp->fi_deleg_file = filp;
4118        fp->fi_delegees = 0;
4119        status = hash_delegation_locked(dp, fp);
4120        spin_unlock(&fp->fi_lock);
4121        spin_unlock(&state_lock);
4122        if (status) {
4123                /* Should never happen, this is a new fi_deleg_file  */
4124                WARN_ON_ONCE(1);
4125                goto out_fput;
4126        }
4127        return 0;
4128out_unlock:
4129        spin_unlock(&fp->fi_lock);
4130        spin_unlock(&state_lock);
4131out_fput:
4132        fput(filp);
4133        return status;
4134}
4135
4136static struct nfs4_delegation *
4137nfs4_set_delegation(struct nfs4_client *clp, struct svc_fh *fh,
4138                    struct nfs4_file *fp, struct nfs4_clnt_odstate *odstate)
4139{
4140        int status;
4141        struct nfs4_delegation *dp;
4142
4143        if (fp->fi_had_conflict)
4144                return ERR_PTR(-EAGAIN);
4145
4146        spin_lock(&state_lock);
4147        spin_lock(&fp->fi_lock);
4148        status = nfs4_get_existing_delegation(clp, fp);
4149        spin_unlock(&fp->fi_lock);
4150        spin_unlock(&state_lock);
4151
4152        if (status)
4153                return ERR_PTR(status);
4154
4155        dp = alloc_init_deleg(clp, fh, odstate);
4156        if (!dp)
4157                return ERR_PTR(-ENOMEM);
4158
4159        get_nfs4_file(fp);
4160        spin_lock(&state_lock);
4161        spin_lock(&fp->fi_lock);
4162        dp->dl_stid.sc_file = fp;
4163        if (!fp->fi_deleg_file) {
4164                spin_unlock(&fp->fi_lock);
4165                spin_unlock(&state_lock);
4166                status = nfs4_setlease(dp);
4167                goto out;
4168        }
4169        if (fp->fi_had_conflict) {
4170                status = -EAGAIN;
4171                goto out_unlock;
4172        }
4173        status = hash_delegation_locked(dp, fp);
4174out_unlock:
4175        spin_unlock(&fp->fi_lock);
4176        spin_unlock(&state_lock);
4177out:
4178        if (status) {
4179                put_clnt_odstate(dp->dl_clnt_odstate);
4180                nfs4_put_stid(&dp->dl_stid);
4181                return ERR_PTR(status);
4182        }
4183        return dp;
4184}
4185
4186static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
4187{
4188        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4189        if (status == -EAGAIN)
4190                open->op_why_no_deleg = WND4_CONTENTION;
4191        else {
4192                open->op_why_no_deleg = WND4_RESOURCE;
4193                switch (open->op_deleg_want) {
4194                case NFS4_SHARE_WANT_READ_DELEG:
4195                case NFS4_SHARE_WANT_WRITE_DELEG:
4196                case NFS4_SHARE_WANT_ANY_DELEG:
4197                        break;
4198                case NFS4_SHARE_WANT_CANCEL:
4199                        open->op_why_no_deleg = WND4_CANCELLED;
4200                        break;
4201                case NFS4_SHARE_WANT_NO_DELEG:
4202                        WARN_ON_ONCE(1);
4203                }
4204        }
4205}
4206
4207/*
4208 * Attempt to hand out a delegation.
4209 *
4210 * Note we don't support write delegations, and won't until the vfs has
4211 * proper support for them.
4212 */
4213static void
4214nfs4_open_delegation(struct svc_fh *fh, struct nfsd4_open *open,
4215                        struct nfs4_ol_stateid *stp)
4216{
4217        struct nfs4_delegation *dp;
4218        struct nfs4_openowner *oo = openowner(stp->st_stateowner);
4219        struct nfs4_client *clp = stp->st_stid.sc_client;
4220        int cb_up;
4221        int status = 0;
4222
4223        cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
4224        open->op_recall = 0;
4225        switch (open->op_claim_type) {
4226                case NFS4_OPEN_CLAIM_PREVIOUS:
4227                        if (!cb_up)
4228                                open->op_recall = 1;
4229                        if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
4230                                goto out_no_deleg;
4231                        break;
4232                case NFS4_OPEN_CLAIM_NULL:
4233                case NFS4_OPEN_CLAIM_FH:
4234                        /*
4235                         * Let's not give out any delegations till everyone's
4236                         * had the chance to reclaim theirs, *and* until
4237                         * NLM locks have all been reclaimed:
4238                         */
4239                        if (locks_in_grace(clp->net))
4240                                goto out_no_deleg;
4241                        if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
4242                                goto out_no_deleg;
4243                        /*
4244                         * Also, if the file was opened for write or
4245                         * create, there's a good chance the client's
4246                         * about to write to it, resulting in an
4247                         * immediate recall (since we don't support
4248                         * write delegations):
4249                         */
4250                        if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE)
4251                                goto out_no_deleg;
4252                        if (open->op_create == NFS4_OPEN_CREATE)
4253                                goto out_no_deleg;
4254                        break;
4255                default:
4256                        goto out_no_deleg;
4257        }
4258        dp = nfs4_set_delegation(clp, fh, stp->st_stid.sc_file, stp->st_clnt_odstate);
4259        if (IS_ERR(dp))
4260                goto out_no_deleg;
4261
4262        memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
4263
4264        dprintk("NFSD: delegation stateid=" STATEID_FMT "\n",
4265                STATEID_VAL(&dp->dl_stid.sc_stateid));
4266        open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
4267        nfs4_put_stid(&dp->dl_stid);
4268        return;
4269out_no_deleg:
4270        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
4271        if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
4272            open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
4273                dprintk("NFSD: WARNING: refusing delegation reclaim\n");
4274                open->op_recall = 1;
4275        }
4276
4277        /* 4.1 client asking for a delegation? */
4278        if (open->op_deleg_want)
4279                nfsd4_open_deleg_none_ext(open, status);
4280        return;
4281}
4282
4283static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
4284                                        struct nfs4_delegation *dp)
4285{
4286        if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
4287            dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
4288                open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4289                open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
4290        } else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
4291                   dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
4292                open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4293                open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
4294        }
4295        /* Otherwise the client must be confused wanting a delegation
4296         * it already has, therefore we don't return
4297         * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
4298         */
4299}
4300
4301__be32
4302nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
4303{
4304        struct nfsd4_compoundres *resp = rqstp->rq_resp;
4305        struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
4306        struct nfs4_file *fp = NULL;
4307        struct nfs4_ol_stateid *stp = NULL;
4308        struct nfs4_ol_stateid *swapstp = NULL;
4309        struct nfs4_delegation *dp = NULL;
4310        __be32 status;
4311
4312        /*
4313         * Lookup file; if found, lookup stateid and check open request,
4314         * and check for delegations in the process of being recalled.
4315         * If not found, create the nfs4_file struct
4316         */
4317        fp = find_or_add_file(open->op_file, &current_fh->fh_handle);
4318        if (fp != open->op_file) {
4319                status = nfs4_check_deleg(cl, open, &dp);
4320                if (status)
4321                        goto out;
4322                spin_lock(&fp->fi_lock);
4323                stp = nfsd4_find_existing_open(fp, open);
4324                spin_unlock(&fp->fi_lock);
4325        } else {
4326                open->op_file = NULL;
4327                status = nfserr_bad_stateid;
4328                if (nfsd4_is_deleg_cur(open))
4329                        goto out;
4330        }
4331
4332        /*
4333         * OPEN the file, or upgrade an existing OPEN.
4334         * If truncate fails, the OPEN fails.
4335         */
4336        if (stp) {
4337                /* Stateid was found, this is an OPEN upgrade */
4338                down_read(&stp->st_rwsem);
4339                status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
4340                if (status) {
4341                        up_read(&stp->st_rwsem);
4342                        goto out;
4343                }
4344        } else {
4345                stp = open->op_stp;
4346                open->op_stp = NULL;
4347                swapstp = init_open_stateid(stp, fp, open);
4348                if (swapstp) {
4349                        nfs4_put_stid(&stp->st_stid);
4350                        stp = swapstp;
4351                        down_read(&stp->st_rwsem);
4352                        status = nfs4_upgrade_open(rqstp, fp, current_fh,
4353                                                stp, open);
4354                        if (status) {
4355                                up_read(&stp->st_rwsem);
4356                                goto out;
4357                        }
4358                        goto upgrade_out;
4359                }
4360                down_read(&stp->st_rwsem);
4361                status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open);
4362                if (status) {
4363                        up_read(&stp->st_rwsem);
4364                        release_open_stateid(stp);
4365                        goto out;
4366                }
4367
4368                stp->st_clnt_odstate = find_or_hash_clnt_odstate(fp,
4369                                                        open->op_odstate);
4370                if (stp->st_clnt_odstate == open->op_odstate)
4371                        open->op_odstate = NULL;
4372        }
4373upgrade_out:
4374        nfs4_inc_and_copy_stateid(&open->op_stateid, &stp->st_stid);
4375        up_read(&stp->st_rwsem);
4376
4377        if (nfsd4_has_session(&resp->cstate)) {
4378                if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
4379                        open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
4380                        open->op_why_no_deleg = WND4_NOT_WANTED;
4381                        goto nodeleg;
4382                }
4383        }
4384
4385        /*
4386        * Attempt to hand out a delegation. No error return, because the
4387        * OPEN succeeds even if we fail.
4388        */
4389        nfs4_open_delegation(current_fh, open, stp);
4390nodeleg:
4391        status = nfs_ok;
4392
4393        dprintk("%s: stateid=" STATEID_FMT "\n", __func__,
4394                STATEID_VAL(&stp->st_stid.sc_stateid));
4395out:
4396        /* 4.1 client trying to upgrade/downgrade delegation? */
4397        if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
4398            open->op_deleg_want)
4399                nfsd4_deleg_xgrade_none_ext(open, dp);
4400
4401        if (fp)
4402                put_nfs4_file(fp);
4403        if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
4404                open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
4405        /*
4406        * To finish the open response, we just need to set the rflags.
4407        */
4408        open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
4409        if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED) &&
4410            !nfsd4_has_session(&resp->cstate))
4411                open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
4412        if (dp)
4413                nfs4_put_stid(&dp->dl_stid);
4414        if (stp)
4415                nfs4_put_stid(&stp->st_stid);
4416
4417        return status;
4418}
4419
4420void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
4421                              struct nfsd4_open *open)
4422{
4423        if (open->op_openowner) {
4424                struct nfs4_stateowner *so = &open->op_openowner->oo_owner;
4425
4426                nfsd4_cstate_assign_replay(cstate, so);
4427                nfs4_put_stateowner(so);
4428        }
4429        if (open->op_file)
4430                kmem_cache_free(file_slab, open->op_file);
4431        if (open->op_stp)
4432                nfs4_put_stid(&open->op_stp->st_stid);
4433        if (open->op_odstate)
4434                kmem_cache_free(odstate_slab, open->op_odstate);
4435}
4436
4437__be32
4438nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4439            clientid_t *clid)
4440{
4441        struct nfs4_client *clp;
4442        __be32 status;
4443        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4444
4445        dprintk("process_renew(%08x/%08x): starting\n", 
4446                        clid->cl_boot, clid->cl_id);
4447        status = lookup_clientid(clid, cstate, nn);
4448        if (status)
4449                goto out;
4450        clp = cstate->clp;
4451        status = nfserr_cb_path_down;
4452        if (!list_empty(&clp->cl_delegations)
4453                        && clp->cl_cb_state != NFSD4_CB_UP)
4454                goto out;
4455        status = nfs_ok;
4456out:
4457        return status;
4458}
4459
4460void
4461nfsd4_end_grace(struct nfsd_net *nn)
4462{
4463        /* do nothing if grace period already ended */
4464        if (nn->grace_ended)
4465                return;
4466
4467        dprintk("NFSD: end of grace period\n");
4468        nn->grace_ended = true;
4469        /*
4470         * If the server goes down again right now, an NFSv4
4471         * client will still be allowed to reclaim after it comes back up,
4472         * even if it hasn't yet had a chance to reclaim state this time.
4473         *
4474         */
4475        nfsd4_record_grace_done(nn);
4476        /*
4477         * At this point, NFSv4 clients can still reclaim.  But if the
4478         * server crashes, any that have not yet reclaimed will be out
4479         * of luck on the next boot.
4480         *
4481         * (NFSv4.1+ clients are considered to have reclaimed once they
4482         * call RECLAIM_COMPLETE.  NFSv4.0 clients are considered to
4483         * have reclaimed after their first OPEN.)
4484         */
4485        locks_end_grace(&nn->nfsd4_manager);
4486        /*
4487         * At this point, and once lockd and/or any other containers
4488         * exit their grace period, further reclaims will fail and
4489         * regular locking can resume.
4490         */
4491}
4492
4493static time_t
4494nfs4_laundromat(struct nfsd_net *nn)
4495{
4496        struct nfs4_client *clp;
4497        struct nfs4_openowner *oo;
4498        struct nfs4_delegation *dp;
4499        struct nfs4_ol_stateid *stp;
4500        struct list_head *pos, *next, reaplist;
4501        time_t cutoff = get_seconds() - nn->nfsd4_lease;
4502        time_t t, new_timeo = nn->nfsd4_lease;
4503
4504        dprintk("NFSD: laundromat service - starting\n");
4505        nfsd4_end_grace(nn);
4506        INIT_LIST_HEAD(&reaplist);
4507        spin_lock(&nn->client_lock);
4508        list_for_each_safe(pos, next, &nn->client_lru) {
4509                clp = list_entry(pos, struct nfs4_client, cl_lru);
4510                if (time_after((unsigned long)clp->cl_time, (unsigned long)cutoff)) {
4511                        t = clp->cl_time - cutoff;
4512                        new_timeo = min(new_timeo, t);
4513                        break;
4514                }
4515                if (mark_client_expired_locked(clp)) {
4516                        dprintk("NFSD: client in use (clientid %08x)\n",
4517                                clp->cl_clientid.cl_id);
4518                        continue;
4519                }
4520                list_add(&clp->cl_lru, &reaplist);
4521        }
4522        spin_unlock(&nn->client_lock);
4523        list_for_each_safe(pos, next, &reaplist) {
4524                clp = list_entry(pos, struct nfs4_client, cl_lru);
4525                dprintk("NFSD: purging unused client (clientid %08x)\n",
4526                        clp->cl_clientid.cl_id);
4527                list_del_init(&clp->cl_lru);
4528                expire_client(clp);
4529        }
4530        spin_lock(&state_lock);
4531        list_for_each_safe(pos, next, &nn->del_recall_lru) {
4532                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
4533                if (time_after((unsigned long)dp->dl_time, (unsigned long)cutoff)) {
4534                        t = dp->dl_time - cutoff;
4535                        new_timeo = min(new_timeo, t);
4536                        break;
4537                }
4538                WARN_ON(!unhash_delegation_locked(dp));
4539                list_add(&dp->dl_recall_lru, &reaplist);
4540        }
4541        spin_unlock(&state_lock);
4542        while (!list_empty(&reaplist)) {
4543                dp = list_first_entry(&reaplist, struct nfs4_delegation,
4544                                        dl_recall_lru);
4545                list_del_init(&dp->dl_recall_lru);
4546                revoke_delegation(dp);
4547        }
4548
4549        spin_lock(&nn->client_lock);
4550        while (!list_empty(&nn->close_lru)) {
4551                oo = list_first_entry(&nn->close_lru, struct nfs4_openowner,
4552                                        oo_close_lru);
4553                if (time_after((unsigned long)oo->oo_time,
4554                               (unsigned long)cutoff)) {
4555                        t = oo->oo_time - cutoff;
4556                        new_timeo = min(new_timeo, t);
4557                        break;
4558                }
4559                list_del_init(&oo->oo_close_lru);
4560                stp = oo->oo_last_closed_stid;
4561                oo->oo_last_closed_stid = NULL;
4562                spin_unlock(&nn->client_lock);
4563                nfs4_put_stid(&stp->st_stid);
4564                spin_lock(&nn->client_lock);
4565        }
4566        spin_unlock(&nn->client_lock);
4567
4568        new_timeo = max_t(time_t, new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
4569        return new_timeo;
4570}
4571
4572static struct workqueue_struct *laundry_wq;
4573static void laundromat_main(struct work_struct *);
4574
4575static void
4576laundromat_main(struct work_struct *laundry)
4577{
4578        time_t t;
4579        struct delayed_work *dwork = to_delayed_work(laundry);
4580        struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
4581                                           laundromat_work);
4582
4583        t = nfs4_laundromat(nn);
4584        dprintk("NFSD: laundromat_main - sleeping for %ld seconds\n", t);
4585        queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
4586}
4587
4588static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stid *stp)
4589{
4590        if (!fh_match(&fhp->fh_handle, &stp->sc_file->fi_fhandle))
4591                return nfserr_bad_stateid;
4592        return nfs_ok;
4593}
4594
4595static inline int
4596access_permit_read(struct nfs4_ol_stateid *stp)
4597{
4598        return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
4599                test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
4600                test_access(NFS4_SHARE_ACCESS_WRITE, stp);
4601}
4602
4603static inline int
4604access_permit_write(struct nfs4_ol_stateid *stp)
4605{
4606        return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
4607                test_access(NFS4_SHARE_ACCESS_BOTH, stp);
4608}
4609
4610static
4611__be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
4612{
4613        __be32 status = nfserr_openmode;
4614
4615        /* For lock stateid's, we test the parent open, not the lock: */
4616        if (stp->st_openstp)
4617                stp = stp->st_openstp;
4618        if ((flags & WR_STATE) && !access_permit_write(stp))
4619                goto out;
4620        if ((flags & RD_STATE) && !access_permit_read(stp))
4621                goto out;
4622        status = nfs_ok;
4623out:
4624        return status;
4625}
4626
4627static inline __be32
4628check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
4629{
4630        if (ONE_STATEID(stateid) && (flags & RD_STATE))
4631                return nfs_ok;
4632        else if (opens_in_grace(net)) {
4633                /* Answer in remaining cases depends on existence of
4634                 * conflicting state; so we must wait out the grace period. */
4635                return nfserr_grace;
4636        } else if (flags & WR_STATE)
4637                return nfs4_share_conflict(current_fh,
4638                                NFS4_SHARE_DENY_WRITE);
4639        else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
4640                return nfs4_share_conflict(current_fh,
4641                                NFS4_SHARE_DENY_READ);
4642}
4643
4644/*
4645 * Allow READ/WRITE during grace period on recovered state only for files
4646 * that are not able to provide mandatory locking.
4647 */
4648static inline int
4649grace_disallows_io(struct net *net, struct inode *inode)
4650{
4651        return opens_in_grace(net) && mandatory_lock(inode);
4652}
4653
4654/* Returns true iff a is later than b: */
4655static bool stateid_generation_after(stateid_t *a, stateid_t *b)
4656{
4657        return (s32)(a->si_generation - b->si_generation) > 0;
4658}
4659
4660static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
4661{
4662        /*
4663         * When sessions are used the stateid generation number is ignored
4664         * when it is zero.
4665         */
4666        if (has_session && in->si_generation == 0)
4667                return nfs_ok;
4668
4669        if (in->si_generation == ref->si_generation)
4670                return nfs_ok;
4671
4672        /* If the client sends us a stateid from the future, it's buggy: */
4673        if (stateid_generation_after(in, ref))
4674                return nfserr_bad_stateid;
4675        /*
4676         * However, we could see a stateid from the past, even from a
4677         * non-buggy client.  For example, if the client sends a lock
4678         * while some IO is outstanding, the lock may bump si_generation
4679         * while the IO is still in flight.  The client could avoid that
4680         * situation by waiting for responses on all the IO requests,
4681         * but better performance may result in retrying IO that
4682         * receives an old_stateid error if requests are rarely
4683         * reordered in flight:
4684         */
4685        return nfserr_old_stateid;
4686}
4687
4688static __be32 nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid *ols)
4689{
4690        if (ols->st_stateowner->so_is_open_owner &&
4691            !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
4692                return nfserr_bad_stateid;
4693        return nfs_ok;
4694}
4695
4696static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
4697{
4698        struct nfs4_stid *s;
4699        __be32 status = nfserr_bad_stateid;
4700
4701        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4702                return status;
4703        /* Client debugging aid. */
4704        if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid)) {
4705                char addr_str[INET6_ADDRSTRLEN];
4706                rpc_ntop((struct sockaddr *)&cl->cl_addr, addr_str,
4707                                 sizeof(addr_str));
4708                pr_warn_ratelimited("NFSD: client %s testing state ID "
4709                                        "with incorrect client ID\n", addr_str);
4710                return status;
4711        }
4712        spin_lock(&cl->cl_lock);
4713        s = find_stateid_locked(cl, stateid);
4714        if (!s)
4715                goto out_unlock;
4716        status = check_stateid_generation(stateid, &s->sc_stateid, 1);
4717        if (status)
4718                goto out_unlock;
4719        switch (s->sc_type) {
4720        case NFS4_DELEG_STID:
4721                status = nfs_ok;
4722                break;
4723        case NFS4_REVOKED_DELEG_STID:
4724                status = nfserr_deleg_revoked;
4725                break;
4726        case NFS4_OPEN_STID:
4727        case NFS4_LOCK_STID:
4728                status = nfsd4_check_openowner_confirmed(openlockstateid(s));
4729                break;
4730        default:
4731                printk("unknown stateid type %x\n", s->sc_type);
4732                /* Fallthrough */
4733        case NFS4_CLOSED_STID:
4734        case NFS4_CLOSED_DELEG_STID:
4735                status = nfserr_bad_stateid;
4736        }
4737out_unlock:
4738        spin_unlock(&cl->cl_lock);
4739        return status;
4740}
4741
4742__be32
4743nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
4744                     stateid_t *stateid, unsigned char typemask,
4745                     struct nfs4_stid **s, struct nfsd_net *nn)
4746{
4747        __be32 status;
4748
4749        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid))
4750                return nfserr_bad_stateid;
4751        status = lookup_clientid(&stateid->si_opaque.so_clid, cstate, nn);
4752        if (status == nfserr_stale_clientid) {
4753                if (cstate->session)
4754                        return nfserr_bad_stateid;
4755                return nfserr_stale_stateid;
4756        }
4757        if (status)
4758                return status;
4759        *s = find_stateid_by_type(cstate->clp, stateid, typemask);
4760        if (!*s)
4761                return nfserr_bad_stateid;
4762        return nfs_ok;
4763}
4764
4765static struct file *
4766nfs4_find_file(struct nfs4_stid *s, int flags)
4767{
4768        if (!s)
4769                return NULL;
4770
4771        switch (s->sc_type) {
4772        case NFS4_DELEG_STID:
4773                if (WARN_ON_ONCE(!s->sc_file->fi_deleg_file))
4774                        return NULL;
4775                return get_file(s->sc_file->fi_deleg_file);
4776        case NFS4_OPEN_STID:
4777        case NFS4_LOCK_STID:
4778                if (flags & RD_STATE)
4779                        return find_readable_file(s->sc_file);
4780                else
4781                        return find_writeable_file(s->sc_file);
4782                break;
4783        }
4784
4785        return NULL;
4786}
4787
4788static __be32
4789nfs4_check_olstateid(struct svc_fh *fhp, struct nfs4_ol_stateid *ols, int flags)
4790{
4791        __be32 status;
4792
4793        status = nfsd4_check_openowner_confirmed(ols);
4794        if (status)
4795                return status;
4796        return nfs4_check_openmode(ols, flags);
4797}
4798
4799static __be32
4800nfs4_check_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct nfs4_stid *s,
4801                struct file **filpp, bool *tmp_file, int flags)
4802{
4803        int acc = (flags & RD_STATE) ? NFSD_MAY_READ : NFSD_MAY_WRITE;
4804        struct file *file;
4805        __be32 status;
4806
4807        file = nfs4_find_file(s, flags);
4808        if (file) {
4809                status = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
4810                                acc | NFSD_MAY_OWNER_OVERRIDE);
4811                if (status) {
4812                        fput(file);
4813                        return status;
4814                }
4815
4816                *filpp = file;
4817        } else {
4818                status = nfsd_open(rqstp, fhp, S_IFREG, acc, filpp);
4819                if (status)
4820                        return status;
4821
4822                if (tmp_file)
4823                        *tmp_file = true;
4824        }
4825
4826        return 0;
4827}
4828
4829/*
4830 * Checks for stateid operations
4831 */
4832__be32
4833nfs4_preprocess_stateid_op(struct svc_rqst *rqstp,
4834                struct nfsd4_compound_state *cstate, struct svc_fh *fhp,
4835                stateid_t *stateid, int flags, struct file **filpp, bool *tmp_file)
4836{
4837        struct inode *ino = d_inode(fhp->fh_dentry);
4838        struct net *net = SVC_NET(rqstp);
4839        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
4840        struct nfs4_stid *s = NULL;
4841        __be32 status;
4842
4843        if (filpp)
4844                *filpp = NULL;
4845        if (tmp_file)
4846                *tmp_file = false;
4847
4848        if (grace_disallows_io(net, ino))
4849                return nfserr_grace;
4850
4851        if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
4852                status = check_special_stateids(net, fhp, stateid, flags);
4853                goto done;
4854        }
4855
4856        status = nfsd4_lookup_stateid(cstate, stateid,
4857                                NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
4858                                &s, nn);
4859        if (status)
4860                return status;
4861        status = check_stateid_generation(stateid, &s->sc_stateid,
4862                        nfsd4_has_session(cstate));
4863        if (status)
4864                goto out;
4865
4866        switch (s->sc_type) {
4867        case NFS4_DELEG_STID:
4868                status = nfs4_check_delegmode(delegstateid(s), flags);
4869                break;
4870        case NFS4_OPEN_STID:
4871        case NFS4_LOCK_STID:
4872                status = nfs4_check_olstateid(fhp, openlockstateid(s), flags);
4873                break;
4874        default:
4875                status = nfserr_bad_stateid;
4876                break;
4877        }
4878        if (status)
4879                goto out;
4880        status = nfs4_check_fh(fhp, s);
4881
4882done:
4883        if (!status && filpp)
4884                status = nfs4_check_file(rqstp, fhp, s, filpp, tmp_file, flags);
4885out:
4886        if (s)
4887                nfs4_put_stid(s);
4888        return status;
4889}
4890
4891/*
4892 * Test if the stateid is valid
4893 */
4894__be32
4895nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4896                   struct nfsd4_test_stateid *test_stateid)
4897{
4898        struct nfsd4_test_stateid_id *stateid;
4899        struct nfs4_client *cl = cstate->session->se_client;
4900
4901        list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
4902                stateid->ts_id_status =
4903                        nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
4904
4905        return nfs_ok;
4906}
4907
4908__be32
4909nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4910                   struct nfsd4_free_stateid *free_stateid)
4911{
4912        stateid_t *stateid = &free_stateid->fr_stateid;
4913        struct nfs4_stid *s;
4914        struct nfs4_delegation *dp;
4915        struct nfs4_ol_stateid *stp;
4916        struct nfs4_client *cl = cstate->session->se_client;
4917        __be32 ret = nfserr_bad_stateid;
4918
4919        spin_lock(&cl->cl_lock);
4920        s = find_stateid_locked(cl, stateid);
4921        if (!s)
4922                goto out_unlock;
4923        switch (s->sc_type) {
4924        case NFS4_DELEG_STID:
4925                ret = nfserr_locks_held;
4926                break;
4927        case NFS4_OPEN_STID:
4928                ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4929                if (ret)
4930                        break;
4931                ret = nfserr_locks_held;
4932                break;
4933        case NFS4_LOCK_STID:
4934                ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
4935                if (ret)
4936                        break;
4937                stp = openlockstateid(s);
4938                ret = nfserr_locks_held;
4939                if (check_for_locks(stp->st_stid.sc_file,
4940                                    lockowner(stp->st_stateowner)))
4941                        break;
4942                WARN_ON(!unhash_lock_stateid(stp));
4943                spin_unlock(&cl->cl_lock);
4944                nfs4_put_stid(s);
4945                ret = nfs_ok;
4946                goto out;
4947        case NFS4_REVOKED_DELEG_STID:
4948                dp = delegstateid(s);
4949                list_del_init(&dp->dl_recall_lru);
4950                spin_unlock(&cl->cl_lock);
4951                nfs4_put_stid(s);
4952                ret = nfs_ok;
4953                goto out;
4954        /* Default falls through and returns nfserr_bad_stateid */
4955        }
4956out_unlock:
4957        spin_unlock(&cl->cl_lock);
4958out:
4959        return ret;
4960}
4961
4962static inline int
4963setlkflg (int type)
4964{
4965        return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
4966                RD_STATE : WR_STATE;
4967}
4968
4969static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
4970{
4971        struct svc_fh *current_fh = &cstate->current_fh;
4972        struct nfs4_stateowner *sop = stp->st_stateowner;
4973        __be32 status;
4974
4975        status = nfsd4_check_seqid(cstate, sop, seqid);
4976        if (status)
4977                return status;
4978        if (stp->st_stid.sc_type == NFS4_CLOSED_STID
4979                || stp->st_stid.sc_type == NFS4_REVOKED_DELEG_STID)
4980                /*
4981                 * "Closed" stateid's exist *only* to return
4982                 * nfserr_replay_me from the previous step, and
4983                 * revoked delegations are kept only for free_stateid.
4984                 */
4985                return nfserr_bad_stateid;
4986        down_write(&stp->st_rwsem);
4987        status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
4988        if (status == nfs_ok)
4989                status = nfs4_check_fh(current_fh, &stp->st_stid);
4990        if (status != nfs_ok)
4991                up_write(&stp->st_rwsem);
4992        return status;
4993}
4994
4995/* 
4996 * Checks for sequence id mutating operations. 
4997 */
4998static __be32
4999nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
5000                         stateid_t *stateid, char typemask,
5001                         struct nfs4_ol_stateid **stpp,
5002                         struct nfsd_net *nn)
5003{
5004        __be32 status;
5005        struct nfs4_stid *s;
5006        struct nfs4_ol_stateid *stp = NULL;
5007
5008        dprintk("NFSD: %s: seqid=%d stateid = " STATEID_FMT "\n", __func__,
5009                seqid, STATEID_VAL(stateid));
5010
5011        *stpp = NULL;
5012        status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
5013        if (status)
5014                return status;
5015        stp = openlockstateid(s);
5016        nfsd4_cstate_assign_replay(cstate, stp->st_stateowner);
5017
5018        status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
5019        if (!status)
5020                *stpp = stp;
5021        else
5022                nfs4_put_stid(&stp->st_stid);
5023        return status;
5024}
5025
5026static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
5027                                                 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
5028{
5029        __be32 status;
5030        struct nfs4_openowner *oo;
5031        struct nfs4_ol_stateid *stp;
5032
5033        status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
5034                                                NFS4_OPEN_STID, &stp, nn);
5035        if (status)
5036                return status;
5037        oo = openowner(stp->st_stateowner);
5038        if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
5039                up_write(&stp->st_rwsem);
5040                nfs4_put_stid(&stp->st_stid);
5041                return nfserr_bad_stateid;
5042        }
5043        *stpp = stp;
5044        return nfs_ok;
5045}
5046
5047__be32
5048nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5049                   struct nfsd4_open_confirm *oc)
5050{
5051        __be32 status;
5052        struct nfs4_openowner *oo;
5053        struct nfs4_ol_stateid *stp;
5054        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5055
5056        dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
5057                        cstate->current_fh.fh_dentry);
5058
5059        status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
5060        if (status)
5061                return status;
5062
5063        status = nfs4_preprocess_seqid_op(cstate,
5064                                        oc->oc_seqid, &oc->oc_req_stateid,
5065                                        NFS4_OPEN_STID, &stp, nn);
5066        if (status)
5067                goto out;
5068        oo = openowner(stp->st_stateowner);
5069        status = nfserr_bad_stateid;
5070        if (oo->oo_flags & NFS4_OO_CONFIRMED) {
5071                up_write(&stp->st_rwsem);
5072                goto put_stateid;
5073        }
5074        oo->oo_flags |= NFS4_OO_CONFIRMED;
5075        nfs4_inc_and_copy_stateid(&oc->oc_resp_stateid, &stp->st_stid);
5076        up_write(&stp->st_rwsem);
5077        dprintk("NFSD: %s: success, seqid=%d stateid=" STATEID_FMT "\n",
5078                __func__, oc->oc_seqid, STATEID_VAL(&stp->st_stid.sc_stateid));
5079
5080        nfsd4_client_record_create(oo->oo_owner.so_client);
5081        status = nfs_ok;
5082put_stateid:
5083        nfs4_put_stid(&stp->st_stid);
5084out:
5085        nfsd4_bump_seqid(cstate, status);
5086        return status;
5087}
5088
5089static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
5090{
5091        if (!test_access(access, stp))
5092                return;
5093        nfs4_file_put_access(stp->st_stid.sc_file, access);
5094        clear_access(access, stp);
5095}
5096
5097static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
5098{
5099        switch (to_access) {
5100        case NFS4_SHARE_ACCESS_READ:
5101                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
5102                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
5103                break;
5104        case NFS4_SHARE_ACCESS_WRITE:
5105                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
5106                nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
5107                break;
5108        case NFS4_SHARE_ACCESS_BOTH:
5109                break;
5110        default:
5111                WARN_ON_ONCE(1);
5112        }
5113}
5114
5115__be32
5116nfsd4_open_downgrade(struct svc_rqst *rqstp,
5117                     struct nfsd4_compound_state *cstate,
5118                     struct nfsd4_open_downgrade *od)
5119{
5120        __be32 status;
5121        struct nfs4_ol_stateid *stp;
5122        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5123
5124        dprintk("NFSD: nfsd4_open_downgrade on file %pd\n", 
5125                        cstate->current_fh.fh_dentry);
5126
5127        /* We don't yet support WANT bits: */
5128        if (od->od_deleg_want)
5129                dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
5130                        od->od_deleg_want);
5131
5132        status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
5133                                        &od->od_stateid, &stp, nn);
5134        if (status)
5135                goto out; 
5136        status = nfserr_inval;
5137        if (!test_access(od->od_share_access, stp)) {
5138                dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
5139                        stp->st_access_bmap, od->od_share_access);
5140                goto put_stateid;
5141        }
5142        if (!test_deny(od->od_share_deny, stp)) {
5143                dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
5144                        stp->st_deny_bmap, od->od_share_deny);
5145                goto put_stateid;
5146        }
5147        nfs4_stateid_downgrade(stp, od->od_share_access);
5148        reset_union_bmap_deny(od->od_share_deny, stp);
5149        nfs4_inc_and_copy_stateid(&od->od_stateid, &stp->st_stid);
5150        status = nfs_ok;
5151put_stateid:
5152        up_write(&stp->st_rwsem);
5153        nfs4_put_stid(&stp->st_stid);
5154out:
5155        nfsd4_bump_seqid(cstate, status);
5156        return status;
5157}
5158
5159static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
5160{
5161        struct nfs4_client *clp = s->st_stid.sc_client;
5162        bool unhashed;
5163        LIST_HEAD(reaplist);
5164
5165        s->st_stid.sc_type = NFS4_CLOSED_STID;
5166        spin_lock(&clp->cl_lock);
5167        unhashed = unhash_open_stateid(s, &reaplist);
5168
5169        if (clp->cl_minorversion) {
5170                if (unhashed)
5171                        put_ol_stateid_locked(s, &reaplist);
5172                spin_unlock(&clp->cl_lock);
5173                free_ol_stateid_reaplist(&reaplist);
5174        } else {
5175                spin_unlock(&clp->cl_lock);
5176                free_ol_stateid_reaplist(&reaplist);
5177                if (unhashed)
5178                        move_to_close_lru(s, clp->net);
5179        }
5180}
5181
5182/*
5183 * nfs4_unlock_state() called after encode
5184 */
5185__be32
5186nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5187            struct nfsd4_close *close)
5188{
5189        __be32 status;
5190        struct nfs4_ol_stateid *stp;
5191        struct net *net = SVC_NET(rqstp);
5192        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5193
5194        dprintk("NFSD: nfsd4_close on file %pd\n", 
5195                        cstate->current_fh.fh_dentry);
5196
5197        status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
5198                                        &close->cl_stateid,
5199                                        NFS4_OPEN_STID|NFS4_CLOSED_STID,
5200                                        &stp, nn);
5201        nfsd4_bump_seqid(cstate, status);
5202        if (status)
5203                goto out; 
5204        nfs4_inc_and_copy_stateid(&close->cl_stateid, &stp->st_stid);
5205        up_write(&stp->st_rwsem);
5206
5207        nfsd4_close_open_stateid(stp);
5208
5209        /* put reference from nfs4_preprocess_seqid_op */
5210        nfs4_put_stid(&stp->st_stid);
5211out:
5212        return status;
5213}
5214
5215__be32
5216nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5217                  struct nfsd4_delegreturn *dr)
5218{
5219        struct nfs4_delegation *dp;
5220        stateid_t *stateid = &dr->dr_stateid;
5221        struct nfs4_stid *s;
5222        __be32 status;
5223        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5224
5225        if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
5226                return status;
5227
5228        status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
5229        if (status)
5230                goto out;
5231        dp = delegstateid(s);
5232        status = check_stateid_generation(stateid, &dp->dl_stid.sc_stateid, nfsd4_has_session(cstate));
5233        if (status)
5234                goto put_stateid;
5235
5236        destroy_delegation(dp);
5237put_stateid:
5238        nfs4_put_stid(&dp->dl_stid);
5239out:
5240        return status;
5241}
5242
5243static inline u64
5244end_offset(u64 start, u64 len)
5245{
5246        u64 end;
5247
5248        end = start + len;
5249        return end >= start ? end: NFS4_MAX_UINT64;
5250}
5251
5252/* last octet in a range */
5253static inline u64
5254last_byte_offset(u64 start, u64 len)
5255{
5256        u64 end;
5257
5258        WARN_ON_ONCE(!len);
5259        end = start + len;
5260        return end > start ? end - 1: NFS4_MAX_UINT64;
5261}
5262
5263/*
5264 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
5265 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
5266 * byte, because of sign extension problems.  Since NFSv4 calls for 64-bit
5267 * locking, this prevents us from being completely protocol-compliant.  The
5268 * real solution to this problem is to start using unsigned file offsets in
5269 * the VFS, but this is a very deep change!
5270 */
5271static inline void
5272nfs4_transform_lock_offset(struct file_lock *lock)
5273{
5274        if (lock->fl_start < 0)
5275                lock->fl_start = OFFSET_MAX;
5276        if (lock->fl_end < 0)
5277                lock->fl_end = OFFSET_MAX;
5278}
5279
5280static fl_owner_t
5281nfsd4_fl_get_owner(fl_owner_t owner)
5282{
5283        struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
5284
5285        nfs4_get_stateowner(&lo->lo_owner);
5286        return owner;
5287}
5288
5289static void
5290nfsd4_fl_put_owner(fl_owner_t owner)
5291{
5292        struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
5293
5294        if (lo)
5295                nfs4_put_stateowner(&lo->lo_owner);
5296}
5297
5298static const struct lock_manager_operations nfsd_posix_mng_ops  = {
5299        .lm_get_owner = nfsd4_fl_get_owner,
5300        .lm_put_owner = nfsd4_fl_put_owner,
5301};
5302
5303static inline void
5304nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
5305{
5306        struct nfs4_lockowner *lo;
5307
5308        if (fl->fl_lmops == &nfsd_posix_mng_ops) {
5309                lo = (struct nfs4_lockowner *) fl->fl_owner;
5310                deny->ld_owner.data = kmemdup(lo->lo_owner.so_owner.data,
5311                                        lo->lo_owner.so_owner.len, GFP_KERNEL);
5312                if (!deny->ld_owner.data)
5313                        /* We just don't care that much */
5314                        goto nevermind;
5315                deny->ld_owner.len = lo->lo_owner.so_owner.len;
5316                deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
5317        } else {
5318nevermind:
5319                deny->ld_owner.len = 0;
5320                deny->ld_owner.data = NULL;
5321                deny->ld_clientid.cl_boot = 0;
5322                deny->ld_clientid.cl_id = 0;
5323        }
5324        deny->ld_start = fl->fl_start;
5325        deny->ld_length = NFS4_MAX_UINT64;
5326        if (fl->fl_end != NFS4_MAX_UINT64)
5327                deny->ld_length = fl->fl_end - fl->fl_start + 1;        
5328        deny->ld_type = NFS4_READ_LT;
5329        if (fl->fl_type != F_RDLCK)
5330                deny->ld_type = NFS4_WRITE_LT;
5331}
5332
5333static struct nfs4_lockowner *
5334find_lockowner_str_locked(struct nfs4_client *clp, struct xdr_netobj *owner)
5335{
5336        unsigned int strhashval = ownerstr_hashval(owner);
5337        struct nfs4_stateowner *so;
5338
5339        lockdep_assert_held(&clp->cl_lock);
5340
5341        list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[strhashval],
5342                            so_strhash) {
5343                if (so->so_is_open_owner)
5344                        continue;
5345                if (same_owner_str(so, owner))
5346                        return lockowner(nfs4_get_stateowner(so));
5347        }
5348        return NULL;
5349}
5350
5351static struct nfs4_lockowner *
5352find_lockowner_str(struct nfs4_client *clp, struct xdr_netobj *owner)
5353{
5354        struct nfs4_lockowner *lo;
5355
5356        spin_lock(&clp->cl_lock);
5357        lo = find_lockowner_str_locked(clp, owner);
5358        spin_unlock(&clp->cl_lock);
5359        return lo;
5360}
5361
5362static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
5363{
5364        unhash_lockowner_locked(lockowner(sop));
5365}
5366
5367static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
5368{
5369        struct nfs4_lockowner *lo = lockowner(sop);
5370
5371        kmem_cache_free(lockowner_slab, lo);
5372}
5373
5374static const struct nfs4_stateowner_operations lockowner_ops = {
5375        .so_unhash =    nfs4_unhash_lockowner,
5376        .so_free =      nfs4_free_lockowner,
5377};
5378
5379/*
5380 * Alloc a lock owner structure.
5381 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has 
5382 * occurred. 
5383 *
5384 * strhashval = ownerstr_hashval
5385 */
5386static struct nfs4_lockowner *
5387alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp,
5388                           struct nfs4_ol_stateid *open_stp,
5389                           struct nfsd4_lock *lock)
5390{
5391        struct nfs4_lockowner *lo, *ret;
5392
5393        lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
5394        if (!lo)
5395                return NULL;
5396        INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
5397        lo->lo_owner.so_is_open_owner = 0;
5398        lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
5399        lo->lo_owner.so_ops = &lockowner_ops;
5400        spin_lock(&clp->cl_lock);
5401        ret = find_lockowner_str_locked(clp, &lock->lk_new_owner);
5402        if (ret == NULL) {
5403                list_add(&lo->lo_owner.so_strhash,
5404                         &clp->cl_ownerstr_hashtbl[strhashval]);
5405                ret = lo;
5406        } else
5407                nfs4_free_stateowner(&lo->lo_owner);
5408
5409        spin_unlock(&clp->cl_lock);
5410        return ret;
5411}
5412
5413static void
5414init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
5415                  struct nfs4_file *fp, struct inode *inode,
5416                  struct nfs4_ol_stateid *open_stp)
5417{
5418        struct nfs4_client *clp = lo->lo_owner.so_client;
5419
5420        lockdep_assert_held(&clp->cl_lock);
5421
5422        atomic_inc(&stp->st_stid.sc_count);
5423        stp->st_stid.sc_type = NFS4_LOCK_STID;
5424        stp->st_stateowner = nfs4_get_stateowner(&lo->lo_owner);
5425        get_nfs4_file(fp);
5426        stp->st_stid.sc_file = fp;
5427        stp->st_stid.sc_free = nfs4_free_lock_stateid;
5428        stp->st_access_bmap = 0;
5429        stp->st_deny_bmap = open_stp->st_deny_bmap;
5430        stp->st_openstp = open_stp;
5431        init_rwsem(&stp->st_rwsem);
5432        list_add(&stp->st_locks, &open_stp->st_locks);
5433        list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
5434        spin_lock(&fp->fi_lock);
5435        list_add(&stp->st_perfile, &fp->fi_stateids);
5436        spin_unlock(&fp->fi_lock);
5437}
5438
5439static struct nfs4_ol_stateid *
5440find_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fp)
5441{
5442        struct nfs4_ol_stateid *lst;
5443        struct nfs4_client *clp = lo->lo_owner.so_client;
5444
5445        lockdep_assert_held(&clp->cl_lock);
5446
5447        list_for_each_entry(lst, &lo->lo_owner.so_stateids, st_perstateowner) {
5448                if (lst->st_stid.sc_file == fp) {
5449                        atomic_inc(&lst->st_stid.sc_count);
5450                        return lst;
5451                }
5452        }
5453        return NULL;
5454}
5455
5456static struct nfs4_ol_stateid *
5457find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
5458                            struct inode *inode, struct nfs4_ol_stateid *ost,
5459                            bool *new)
5460{
5461        struct nfs4_stid *ns = NULL;
5462        struct nfs4_ol_stateid *lst;
5463        struct nfs4_openowner *oo = openowner(ost->st_stateowner);
5464        struct nfs4_client *clp = oo->oo_owner.so_client;
5465
5466        spin_lock(&clp->cl_lock);
5467        lst = find_lock_stateid(lo, fi);
5468        if (lst == NULL) {
5469                spin_unlock(&clp->cl_lock);
5470                ns = nfs4_alloc_stid(clp, stateid_slab);
5471                if (ns == NULL)
5472                        return NULL;
5473
5474                spin_lock(&clp->cl_lock);
5475                lst = find_lock_stateid(lo, fi);
5476                if (likely(!lst)) {
5477                        lst = openlockstateid(ns);
5478                        init_lock_stateid(lst, lo, fi, inode, ost);
5479                        ns = NULL;
5480                        *new = true;
5481                }
5482        }
5483        spin_unlock(&clp->cl_lock);
5484        if (ns)
5485                nfs4_put_stid(ns);
5486        return lst;
5487}
5488
5489static int
5490check_lock_length(u64 offset, u64 length)
5491{
5492        return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
5493                (length > ~offset)));
5494}
5495
5496static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
5497{
5498        struct nfs4_file *fp = lock_stp->st_stid.sc_file;
5499
5500        lockdep_assert_held(&fp->fi_lock);
5501
5502        if (test_access(access, lock_stp))
5503                return;
5504        __nfs4_file_get_access(fp, access);
5505        set_access(access, lock_stp);
5506}
5507
5508static __be32
5509lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
5510                            struct nfs4_ol_stateid *ost,
5511                            struct nfsd4_lock *lock,
5512                            struct nfs4_ol_stateid **lst, bool *new)
5513{
5514        __be32 status;
5515        struct nfs4_file *fi = ost->st_stid.sc_file;
5516        struct nfs4_openowner *oo = openowner(ost->st_stateowner);
5517        struct nfs4_client *cl = oo->oo_owner.so_client;
5518        struct inode *inode = d_inode(cstate->current_fh.fh_dentry);
5519        struct nfs4_lockowner *lo;
5520        unsigned int strhashval;
5521
5522        lo = find_lockowner_str(cl, &lock->lk_new_owner);
5523        if (!lo) {
5524                strhashval = ownerstr_hashval(&lock->lk_new_owner);
5525                lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
5526                if (lo == NULL)
5527                        return nfserr_jukebox;
5528        } else {
5529                /* with an existing lockowner, seqids must be the same */
5530                status = nfserr_bad_seqid;
5531                if (!cstate->minorversion &&
5532                    lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
5533                        goto out;
5534        }
5535
5536        *lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
5537        if (*lst == NULL) {
5538                status = nfserr_jukebox;
5539                goto out;
5540        }
5541        status = nfs_ok;
5542out:
5543        nfs4_put_stateowner(&lo->lo_owner);
5544        return status;
5545}
5546
5547/*
5548 *  LOCK operation 
5549 */
5550__be32
5551nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5552           struct nfsd4_lock *lock)
5553{
5554        struct nfs4_openowner *open_sop = NULL;
5555        struct nfs4_lockowner *lock_sop = NULL;
5556        struct nfs4_ol_stateid *lock_stp = NULL;
5557        struct nfs4_ol_stateid *open_stp = NULL;
5558        struct nfs4_file *fp;
5559        struct file *filp = NULL;
5560        struct file_lock *file_lock = NULL;
5561        struct file_lock *conflock = NULL;
5562        __be32 status = 0;
5563        int lkflg;
5564        int err;
5565        bool new = false;
5566        struct net *net = SVC_NET(rqstp);
5567        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
5568
5569        dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
5570                (long long) lock->lk_offset,
5571                (long long) lock->lk_length);
5572
5573        if (check_lock_length(lock->lk_offset, lock->lk_length))
5574                 return nfserr_inval;
5575
5576        if ((status = fh_verify(rqstp, &cstate->current_fh,
5577                                S_IFREG, NFSD_MAY_LOCK))) {
5578                dprintk("NFSD: nfsd4_lock: permission denied!\n");
5579                return status;
5580        }
5581
5582        if (lock->lk_is_new) {
5583                if (nfsd4_has_session(cstate))
5584                        /* See rfc 5661 18.10.3: given clientid is ignored: */
5585                        memcpy(&lock->lk_new_clientid,
5586                                &cstate->session->se_client->cl_clientid,
5587                                sizeof(clientid_t));
5588
5589                status = nfserr_stale_clientid;
5590                if (STALE_CLIENTID(&lock->lk_new_clientid, nn))
5591                        goto out;
5592
5593                /* validate and update open stateid and open seqid */
5594                status = nfs4_preprocess_confirmed_seqid_op(cstate,
5595                                        lock->lk_new_open_seqid,
5596                                        &lock->lk_new_open_stateid,
5597                                        &open_stp, nn);
5598                if (status)
5599                        goto out;
5600                up_write(&open_stp->st_rwsem);
5601                open_sop = openowner(open_stp->st_stateowner);
5602                status = nfserr_bad_stateid;
5603                if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
5604                                                &lock->lk_new_clientid))
5605                        goto out;
5606                status = lookup_or_create_lock_state(cstate, open_stp, lock,
5607                                                        &lock_stp, &new);
5608                if (status == nfs_ok)
5609                        down_write(&lock_stp->st_rwsem);
5610        } else {
5611                status = nfs4_preprocess_seqid_op(cstate,
5612                                       lock->lk_old_lock_seqid,
5613                                       &lock->lk_old_lock_stateid,
5614                                       NFS4_LOCK_STID, &lock_stp, nn);
5615        }
5616        if (status)
5617                goto out;
5618        lock_sop = lockowner(lock_stp->st_stateowner);
5619
5620        lkflg = setlkflg(lock->lk_type);
5621        status = nfs4_check_openmode(lock_stp, lkflg);
5622        if (status)
5623                goto out;
5624
5625        status = nfserr_grace;
5626        if (locks_in_grace(net) && !lock->lk_reclaim)
5627                goto out;
5628        status = nfserr_no_grace;
5629        if (!locks_in_grace(net) && lock->lk_reclaim)
5630                goto out;
5631
5632        file_lock = locks_alloc_lock();
5633        if (!file_lock) {
5634                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5635                status = nfserr_jukebox;
5636                goto out;
5637        }
5638
5639        fp = lock_stp->st_stid.sc_file;
5640        switch (lock->lk_type) {
5641                case NFS4_READ_LT:
5642                case NFS4_READW_LT:
5643                        spin_lock(&fp->fi_lock);
5644                        filp = find_readable_file_locked(fp);
5645                        if (filp)
5646                                get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
5647                        spin_unlock(&fp->fi_lock);
5648                        file_lock->fl_type = F_RDLCK;
5649                        break;
5650                case NFS4_WRITE_LT:
5651                case NFS4_WRITEW_LT:
5652                        spin_lock(&fp->fi_lock);
5653                        filp = find_writeable_file_locked(fp);
5654                        if (filp)
5655                                get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
5656                        spin_unlock(&fp->fi_lock);
5657                        file_lock->fl_type = F_WRLCK;
5658                        break;
5659                default:
5660                        status = nfserr_inval;
5661                goto out;
5662        }
5663        if (!filp) {
5664                status = nfserr_openmode;
5665                goto out;
5666        }
5667
5668        file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(&lock_sop->lo_owner));
5669        file_lock->fl_pid = current->tgid;
5670        file_lock->fl_file = filp;
5671        file_lock->fl_flags = FL_POSIX;
5672        file_lock->fl_lmops = &nfsd_posix_mng_ops;
5673        file_lock->fl_start = lock->lk_offset;
5674        file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
5675        nfs4_transform_lock_offset(file_lock);
5676
5677        conflock = locks_alloc_lock();
5678        if (!conflock) {
5679                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5680                status = nfserr_jukebox;
5681                goto out;
5682        }
5683
5684        err = vfs_lock_file(filp, F_SETLK, file_lock, conflock);
5685        switch (-err) {
5686        case 0: /* success! */
5687                nfs4_inc_and_copy_stateid(&lock->lk_resp_stateid, &lock_stp->st_stid);
5688                status = 0;
5689                break;
5690        case (EAGAIN):          /* conflock holds conflicting lock */
5691                status = nfserr_denied;
5692                dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
5693                nfs4_set_lock_denied(conflock, &lock->lk_denied);
5694                break;
5695        case (EDEADLK):
5696                status = nfserr_deadlock;
5697                break;
5698        default:
5699                dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
5700                status = nfserrno(err);
5701                break;
5702        }
5703out:
5704        if (filp)
5705                fput(filp);
5706        if (lock_stp) {
5707                /* Bump seqid manually if the 4.0 replay owner is openowner */
5708                if (cstate->replay_owner &&
5709                    cstate->replay_owner != &lock_sop->lo_owner &&
5710                    seqid_mutating_err(ntohl(status)))
5711                        lock_sop->lo_owner.so_seqid++;
5712
5713                up_write(&lock_stp->st_rwsem);
5714
5715                /*
5716                 * If this is a new, never-before-used stateid, and we are
5717                 * returning an error, then just go ahead and release it.
5718                 */
5719                if (status && new)
5720                        release_lock_stateid(lock_stp);
5721
5722                nfs4_put_stid(&lock_stp->st_stid);
5723        }
5724        if (open_stp)
5725                nfs4_put_stid(&open_stp->st_stid);
5726        nfsd4_bump_seqid(cstate, status);
5727        if (file_lock)
5728                locks_free_lock(file_lock);
5729        if (conflock)
5730                locks_free_lock(conflock);
5731        return status;
5732}
5733
5734/*
5735 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
5736 * so we do a temporary open here just to get an open file to pass to
5737 * vfs_test_lock.  (Arguably perhaps test_lock should be done with an
5738 * inode operation.)
5739 */
5740static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
5741{
5742        struct file *file;
5743        __be32 err = nfsd_open(rqstp, fhp, S_IFREG, NFSD_MAY_READ, &file);
5744        if (!err) {
5745                err = nfserrno(vfs_test_lock(file, lock));
5746                fput(file);
5747        }
5748        return err;
5749}
5750
5751/*
5752 * LOCKT operation
5753 */
5754__be32
5755nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5756            struct nfsd4_lockt *lockt)
5757{
5758        struct file_lock *file_lock = NULL;
5759        struct nfs4_lockowner *lo = NULL;
5760        __be32 status;
5761        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5762
5763        if (locks_in_grace(SVC_NET(rqstp)))
5764                return nfserr_grace;
5765
5766        if (check_lock_length(lockt->lt_offset, lockt->lt_length))
5767                 return nfserr_inval;
5768
5769        if (!nfsd4_has_session(cstate)) {
5770                status = lookup_clientid(&lockt->lt_clientid, cstate, nn);
5771                if (status)
5772                        goto out;
5773        }
5774
5775        if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
5776                goto out;
5777
5778        file_lock = locks_alloc_lock();
5779        if (!file_lock) {
5780                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5781                status = nfserr_jukebox;
5782                goto out;
5783        }
5784
5785        switch (lockt->lt_type) {
5786                case NFS4_READ_LT:
5787                case NFS4_READW_LT:
5788                        file_lock->fl_type = F_RDLCK;
5789                break;
5790                case NFS4_WRITE_LT:
5791                case NFS4_WRITEW_LT:
5792                        file_lock->fl_type = F_WRLCK;
5793                break;
5794                default:
5795                        dprintk("NFSD: nfs4_lockt: bad lock type!\n");
5796                        status = nfserr_inval;
5797                goto out;
5798        }
5799
5800        lo = find_lockowner_str(cstate->clp, &lockt->lt_owner);
5801        if (lo)
5802                file_lock->fl_owner = (fl_owner_t)lo;
5803        file_lock->fl_pid = current->tgid;
5804        file_lock->fl_flags = FL_POSIX;
5805
5806        file_lock->fl_start = lockt->lt_offset;
5807        file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
5808
5809        nfs4_transform_lock_offset(file_lock);
5810
5811        status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
5812        if (status)
5813                goto out;
5814
5815        if (file_lock->fl_type != F_UNLCK) {
5816                status = nfserr_denied;
5817                nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
5818        }
5819out:
5820        if (lo)
5821                nfs4_put_stateowner(&lo->lo_owner);
5822        if (file_lock)
5823                locks_free_lock(file_lock);
5824        return status;
5825}
5826
5827__be32
5828nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5829            struct nfsd4_locku *locku)
5830{
5831        struct nfs4_ol_stateid *stp;
5832        struct file *filp = NULL;
5833        struct file_lock *file_lock = NULL;
5834        __be32 status;
5835        int err;
5836        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5837
5838        dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
5839                (long long) locku->lu_offset,
5840                (long long) locku->lu_length);
5841
5842        if (check_lock_length(locku->lu_offset, locku->lu_length))
5843                 return nfserr_inval;
5844
5845        status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
5846                                        &locku->lu_stateid, NFS4_LOCK_STID,
5847                                        &stp, nn);
5848        if (status)
5849                goto out;
5850        filp = find_any_file(stp->st_stid.sc_file);
5851        if (!filp) {
5852                status = nfserr_lock_range;
5853                goto put_stateid;
5854        }
5855        file_lock = locks_alloc_lock();
5856        if (!file_lock) {
5857                dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
5858                status = nfserr_jukebox;
5859                goto fput;
5860        }
5861
5862        file_lock->fl_type = F_UNLCK;
5863        file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(stp->st_stateowner));
5864        file_lock->fl_pid = current->tgid;
5865        file_lock->fl_file = filp;
5866        file_lock->fl_flags = FL_POSIX;
5867        file_lock->fl_lmops = &nfsd_posix_mng_ops;
5868        file_lock->fl_start = locku->lu_offset;
5869
5870        file_lock->fl_end = last_byte_offset(locku->lu_offset,
5871                                                locku->lu_length);
5872        nfs4_transform_lock_offset(file_lock);
5873
5874        err = vfs_lock_file(filp, F_SETLK, file_lock, NULL);
5875        if (err) {
5876                dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
5877                goto out_nfserr;
5878        }
5879        nfs4_inc_and_copy_stateid(&locku->lu_stateid, &stp->st_stid);
5880fput:
5881        fput(filp);
5882put_stateid:
5883        up_write(&stp->st_rwsem);
5884        nfs4_put_stid(&stp->st_stid);
5885out:
5886        nfsd4_bump_seqid(cstate, status);
5887        if (file_lock)
5888                locks_free_lock(file_lock);
5889        return status;
5890
5891out_nfserr:
5892        status = nfserrno(err);
5893        goto fput;
5894}
5895
5896/*
5897 * returns
5898 *      true:  locks held by lockowner
5899 *      false: no locks held by lockowner
5900 */
5901static bool
5902check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
5903{
5904        struct file_lock *fl;
5905        int status = false;
5906        struct file *filp = find_any_file(fp);
5907        struct inode *inode;
5908        struct file_lock_context *flctx;
5909
5910        if (!filp) {
5911                /* Any valid lock stateid should have some sort of access */
5912                WARN_ON_ONCE(1);
5913                return status;
5914        }
5915
5916        inode = file_inode(filp);
5917        flctx = inode->i_flctx;
5918
5919        if (flctx && !list_empty_careful(&flctx->flc_posix)) {
5920                spin_lock(&flctx->flc_lock);
5921                list_for_each_entry(fl, &flctx->flc_posix, fl_list) {
5922                        if (fl->fl_owner == (fl_owner_t)lowner) {
5923                                status = true;
5924                                break;
5925                        }
5926                }
5927                spin_unlock(&flctx->flc_lock);
5928        }
5929        fput(filp);
5930        return status;
5931}
5932
5933__be32
5934nfsd4_release_lockowner(struct svc_rqst *rqstp,
5935                        struct nfsd4_compound_state *cstate,
5936                        struct nfsd4_release_lockowner *rlockowner)
5937{
5938        clientid_t *clid = &rlockowner->rl_clientid;
5939        struct nfs4_stateowner *sop;
5940        struct nfs4_lockowner *lo = NULL;
5941        struct nfs4_ol_stateid *stp;
5942        struct xdr_netobj *owner = &rlockowner->rl_owner;
5943        unsigned int hashval = ownerstr_hashval(owner);
5944        __be32 status;
5945        struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5946        struct nfs4_client *clp;
5947
5948        dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
5949                clid->cl_boot, clid->cl_id);
5950
5951        status = lookup_clientid(clid, cstate, nn);
5952        if (status)
5953                return status;
5954
5955        clp = cstate->clp;
5956        /* Find the matching lock stateowner */
5957        spin_lock(&clp->cl_lock);
5958        list_for_each_entry(sop, &clp->cl_ownerstr_hashtbl[hashval],
5959                            so_strhash) {
5960
5961                if (sop->so_is_open_owner || !same_owner_str(sop, owner))
5962                        continue;
5963
5964                /* see if there are still any locks associated with it */
5965                lo = lockowner(sop);
5966                list_for_each_entry(stp, &sop->so_stateids, st_perstateowner) {
5967                        if (check_for_locks(stp->st_stid.sc_file, lo)) {
5968                                status = nfserr_locks_held;
5969                                spin_unlock(&clp->cl_lock);
5970                                return status;
5971                        }
5972                }
5973
5974                nfs4_get_stateowner(sop);
5975                break;
5976        }
5977        spin_unlock(&clp->cl_lock);
5978        if (lo)
5979                release_lockowner(lo);
5980        return status;
5981}
5982
5983static inline struct nfs4_client_reclaim *
5984alloc_reclaim(void)
5985{
5986        return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
5987}
5988
5989bool
5990nfs4_has_reclaimed_state(const char *name, struct nfsd_net *nn)
5991{
5992        struct nfs4_client_reclaim *crp;
5993
5994        crp = nfsd4_find_reclaim_client(name, nn);
5995        return (crp && crp->cr_clp);
5996}
5997
5998/*
5999 * failure => all reset bets are off, nfserr_no_grace...
6000 */
6001struct nfs4_client_reclaim *
6002nfs4_client_to_reclaim(const char *name, struct nfsd_net *nn)
6003{
6004        unsigned int strhashval;
6005        struct nfs4_client_reclaim *crp;
6006
6007        dprintk("NFSD nfs4_client_to_reclaim NAME: %.*s\n", HEXDIR_LEN, name);
6008        crp = alloc_reclaim();
6009        if (crp) {
6010                strhashval = clientstr_hashval(name);
6011                INIT_LIST_HEAD(&crp->cr_strhash);
6012                list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
6013                memcpy(crp->cr_recdir, name, HEXDIR_LEN);
6014                crp->cr_clp = NULL;
6015                nn->reclaim_str_hashtbl_size++;
6016        }
6017        return crp;
6018}
6019
6020void
6021nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
6022{
6023        list_del(&crp->cr_strhash);
6024        kfree(crp);
6025        nn->reclaim_str_hashtbl_size--;
6026}
6027
6028void
6029nfs4_release_reclaim(struct nfsd_net *nn)
6030{
6031        struct nfs4_client_reclaim *crp = NULL;
6032        int i;
6033
6034        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6035                while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
6036                        crp = list_entry(nn->reclaim_str_hashtbl[i].next,
6037                                        struct nfs4_client_reclaim, cr_strhash);
6038                        nfs4_remove_reclaim_record(crp, nn);
6039                }
6040        }
6041        WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
6042}
6043
6044/*
6045 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
6046struct nfs4_client_reclaim *
6047nfsd4_find_reclaim_client(const char *recdir, struct nfsd_net *nn)
6048{
6049        unsigned int strhashval;
6050        struct nfs4_client_reclaim *crp = NULL;
6051
6052        dprintk("NFSD: nfs4_find_reclaim_client for recdir %s\n", recdir);
6053
6054        strhashval = clientstr_hashval(recdir);
6055        list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
6056                if (same_name(crp->cr_recdir, recdir)) {
6057                        return crp;
6058                }
6059        }
6060        return NULL;
6061}
6062
6063/*
6064* Called from OPEN. Look for clientid in reclaim list.
6065*/
6066__be32
6067nfs4_check_open_reclaim(clientid_t *clid,
6068                struct nfsd4_compound_state *cstate,
6069                struct nfsd_net *nn)
6070{
6071        __be32 status;
6072
6073        /* find clientid in conf_id_hashtbl */
6074        status = lookup_clientid(clid, cstate, nn);
6075        if (status)
6076                return nfserr_reclaim_bad;
6077
6078        if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &cstate->clp->cl_flags))
6079                return nfserr_no_grace;
6080
6081        if (nfsd4_client_record_check(cstate->clp))
6082                return nfserr_reclaim_bad;
6083
6084        return nfs_ok;
6085}
6086
6087#ifdef CONFIG_NFSD_FAULT_INJECTION
6088static inline void
6089put_client(struct nfs4_client *clp)
6090{
6091        atomic_dec(&clp->cl_refcount);
6092}
6093
6094static struct nfs4_client *
6095nfsd_find_client(struct sockaddr_storage *addr, size_t addr_size)
6096{
6097        struct nfs4_client *clp;
6098        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6099                                          nfsd_net_id);
6100
6101        if (!nfsd_netns_ready(nn))
6102                return NULL;
6103
6104        list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6105                if (memcmp(&clp->cl_addr, addr, addr_size) == 0)
6106                        return clp;
6107        }
6108        return NULL;
6109}
6110
6111u64
6112nfsd_inject_print_clients(void)
6113{
6114        struct nfs4_client *clp;
6115        u64 count = 0;
6116        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6117                                          nfsd_net_id);
6118        char buf[INET6_ADDRSTRLEN];
6119
6120        if (!nfsd_netns_ready(nn))
6121                return 0;
6122
6123        spin_lock(&nn->client_lock);
6124        list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6125                rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
6126                pr_info("NFS Client: %s\n", buf);
6127                ++count;
6128        }
6129        spin_unlock(&nn->client_lock);
6130
6131        return count;
6132}
6133
6134u64
6135nfsd_inject_forget_client(struct sockaddr_storage *addr, size_t addr_size)
6136{
6137        u64 count = 0;
6138        struct nfs4_client *clp;
6139        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6140                                          nfsd_net_id);
6141
6142        if (!nfsd_netns_ready(nn))
6143                return count;
6144
6145        spin_lock(&nn->client_lock);
6146        clp = nfsd_find_client(addr, addr_size);
6147        if (clp) {
6148                if (mark_client_expired_locked(clp) == nfs_ok)
6149                        ++count;
6150                else
6151                        clp = NULL;
6152        }
6153        spin_unlock(&nn->client_lock);
6154
6155        if (clp)
6156                expire_client(clp);
6157
6158        return count;
6159}
6160
6161u64
6162nfsd_inject_forget_clients(u64 max)
6163{
6164        u64 count = 0;
6165        struct nfs4_client *clp, *next;
6166        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6167                                                nfsd_net_id);
6168        LIST_HEAD(reaplist);
6169
6170        if (!nfsd_netns_ready(nn))
6171                return count;
6172
6173        spin_lock(&nn->client_lock);
6174        list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
6175                if (mark_client_expired_locked(clp) == nfs_ok) {
6176                        list_add(&clp->cl_lru, &reaplist);
6177                        if (max != 0 && ++count >= max)
6178                                break;
6179                }
6180        }
6181        spin_unlock(&nn->client_lock);
6182
6183        list_for_each_entry_safe(clp, next, &reaplist, cl_lru)
6184                expire_client(clp);
6185
6186        return count;
6187}
6188
6189static void nfsd_print_count(struct nfs4_client *clp, unsigned int count,
6190                             const char *type)
6191{
6192        char buf[INET6_ADDRSTRLEN];
6193        rpc_ntop((struct sockaddr *)&clp->cl_addr, buf, sizeof(buf));
6194        printk(KERN_INFO "NFS Client: %s has %u %s\n", buf, count, type);
6195}
6196
6197static void
6198nfsd_inject_add_lock_to_list(struct nfs4_ol_stateid *lst,
6199                             struct list_head *collect)
6200{
6201        struct nfs4_client *clp = lst->st_stid.sc_client;
6202        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6203                                          nfsd_net_id);
6204
6205        if (!collect)
6206                return;
6207
6208        lockdep_assert_held(&nn->client_lock);
6209        atomic_inc(&clp->cl_refcount);
6210        list_add(&lst->st_locks, collect);
6211}
6212
6213static u64 nfsd_foreach_client_lock(struct nfs4_client *clp, u64 max,
6214                                    struct list_head *collect,
6215                                    bool (*func)(struct nfs4_ol_stateid *))
6216{
6217        struct nfs4_openowner *oop;
6218        struct nfs4_ol_stateid *stp, *st_next;
6219        struct nfs4_ol_stateid *lst, *lst_next;
6220        u64 count = 0;
6221
6222        spin_lock(&clp->cl_lock);
6223        list_for_each_entry(oop, &clp->cl_openowners, oo_perclient) {
6224                list_for_each_entry_safe(stp, st_next,
6225                                &oop->oo_owner.so_stateids, st_perstateowner) {
6226                        list_for_each_entry_safe(lst, lst_next,
6227                                        &stp->st_locks, st_locks) {
6228                                if (func) {
6229                                        if (func(lst))
6230                                                nfsd_inject_add_lock_to_list(lst,
6231                                                                        collect);
6232                                }
6233                                ++count;
6234                                /*
6235                                 * Despite the fact that these functions deal
6236                                 * with 64-bit integers for "count", we must
6237                                 * ensure that it doesn't blow up the
6238                                 * clp->cl_refcount. Throw a warning if we
6239                                 * start to approach INT_MAX here.
6240                                 */
6241                                WARN_ON_ONCE(count == (INT_MAX / 2));
6242                                if (count == max)
6243                                        goto out;
6244                        }
6245                }
6246        }
6247out:
6248        spin_unlock(&clp->cl_lock);
6249
6250        return count;
6251}
6252
6253static u64
6254nfsd_collect_client_locks(struct nfs4_client *clp, struct list_head *collect,
6255                          u64 max)
6256{
6257        return nfsd_foreach_client_lock(clp, max, collect, unhash_lock_stateid);
6258}
6259
6260static u64
6261nfsd_print_client_locks(struct nfs4_client *clp)
6262{
6263        u64 count = nfsd_foreach_client_lock(clp, 0, NULL, NULL);
6264        nfsd_print_count(clp, count, "locked files");
6265        return count;
6266}
6267
6268u64
6269nfsd_inject_print_locks(void)
6270{
6271        struct nfs4_client *clp;
6272        u64 count = 0;
6273        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6274                                                nfsd_net_id);
6275
6276        if (!nfsd_netns_ready(nn))
6277                return 0;
6278
6279        spin_lock(&nn->client_lock);
6280        list_for_each_entry(clp, &nn->client_lru, cl_lru)
6281                count += nfsd_print_client_locks(clp);
6282        spin_unlock(&nn->client_lock);
6283
6284        return count;
6285}
6286
6287static void
6288nfsd_reap_locks(struct list_head *reaplist)
6289{
6290        struct nfs4_client *clp;
6291        struct nfs4_ol_stateid *stp, *next;
6292
6293        list_for_each_entry_safe(stp, next, reaplist, st_locks) {
6294                list_del_init(&stp->st_locks);
6295                clp = stp->st_stid.sc_client;
6296                nfs4_put_stid(&stp->st_stid);
6297                put_client(clp);
6298        }
6299}
6300
6301u64
6302nfsd_inject_forget_client_locks(struct sockaddr_storage *addr, size_t addr_size)
6303{
6304        unsigned int count = 0;
6305        struct nfs4_client *clp;
6306        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6307                                                nfsd_net_id);
6308        LIST_HEAD(reaplist);
6309
6310        if (!nfsd_netns_ready(nn))
6311                return count;
6312
6313        spin_lock(&nn->client_lock);
6314        clp = nfsd_find_client(addr, addr_size);
6315        if (clp)
6316                count = nfsd_collect_client_locks(clp, &reaplist, 0);
6317        spin_unlock(&nn->client_lock);
6318        nfsd_reap_locks(&reaplist);
6319        return count;
6320}
6321
6322u64
6323nfsd_inject_forget_locks(u64 max)
6324{
6325        u64 count = 0;
6326        struct nfs4_client *clp;
6327        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6328                                                nfsd_net_id);
6329        LIST_HEAD(reaplist);
6330
6331        if (!nfsd_netns_ready(nn))
6332                return count;
6333
6334        spin_lock(&nn->client_lock);
6335        list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6336                count += nfsd_collect_client_locks(clp, &reaplist, max - count);
6337                if (max != 0 && count >= max)
6338                        break;
6339        }
6340        spin_unlock(&nn->client_lock);
6341        nfsd_reap_locks(&reaplist);
6342        return count;
6343}
6344
6345static u64
6346nfsd_foreach_client_openowner(struct nfs4_client *clp, u64 max,
6347                              struct list_head *collect,
6348                              void (*func)(struct nfs4_openowner *))
6349{
6350        struct nfs4_openowner *oop, *next;
6351        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6352                                                nfsd_net_id);
6353        u64 count = 0;
6354
6355        lockdep_assert_held(&nn->client_lock);
6356
6357        spin_lock(&clp->cl_lock);
6358        list_for_each_entry_safe(oop, next, &clp->cl_openowners, oo_perclient) {
6359                if (func) {
6360                        func(oop);
6361                        if (collect) {
6362                                atomic_inc(&clp->cl_refcount);
6363                                list_add(&oop->oo_perclient, collect);
6364                        }
6365                }
6366                ++count;
6367                /*
6368                 * Despite the fact that these functions deal with
6369                 * 64-bit integers for "count", we must ensure that
6370                 * it doesn't blow up the clp->cl_refcount. Throw a
6371                 * warning if we start to approach INT_MAX here.
6372                 */
6373                WARN_ON_ONCE(count == (INT_MAX / 2));
6374                if (count == max)
6375                        break;
6376        }
6377        spin_unlock(&clp->cl_lock);
6378
6379        return count;
6380}
6381
6382static u64
6383nfsd_print_client_openowners(struct nfs4_client *clp)
6384{
6385        u64 count = nfsd_foreach_client_openowner(clp, 0, NULL, NULL);
6386
6387        nfsd_print_count(clp, count, "openowners");
6388        return count;
6389}
6390
6391static u64
6392nfsd_collect_client_openowners(struct nfs4_client *clp,
6393                               struct list_head *collect, u64 max)
6394{
6395        return nfsd_foreach_client_openowner(clp, max, collect,
6396                                                unhash_openowner_locked);
6397}
6398
6399u64
6400nfsd_inject_print_openowners(void)
6401{
6402        struct nfs4_client *clp;
6403        u64 count = 0;
6404        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6405                                                nfsd_net_id);
6406
6407        if (!nfsd_netns_ready(nn))
6408                return 0;
6409
6410        spin_lock(&nn->client_lock);
6411        list_for_each_entry(clp, &nn->client_lru, cl_lru)
6412                count += nfsd_print_client_openowners(clp);
6413        spin_unlock(&nn->client_lock);
6414
6415        return count;
6416}
6417
6418static void
6419nfsd_reap_openowners(struct list_head *reaplist)
6420{
6421        struct nfs4_client *clp;
6422        struct nfs4_openowner *oop, *next;
6423
6424        list_for_each_entry_safe(oop, next, reaplist, oo_perclient) {
6425                list_del_init(&oop->oo_perclient);
6426                clp = oop->oo_owner.so_client;
6427                release_openowner(oop);
6428                put_client(clp);
6429        }
6430}
6431
6432u64
6433nfsd_inject_forget_client_openowners(struct sockaddr_storage *addr,
6434                                     size_t addr_size)
6435{
6436        unsigned int count = 0;
6437        struct nfs4_client *clp;
6438        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6439                                                nfsd_net_id);
6440        LIST_HEAD(reaplist);
6441
6442        if (!nfsd_netns_ready(nn))
6443                return count;
6444
6445        spin_lock(&nn->client_lock);
6446        clp = nfsd_find_client(addr, addr_size);
6447        if (clp)
6448                count = nfsd_collect_client_openowners(clp, &reaplist, 0);
6449        spin_unlock(&nn->client_lock);
6450        nfsd_reap_openowners(&reaplist);
6451        return count;
6452}
6453
6454u64
6455nfsd_inject_forget_openowners(u64 max)
6456{
6457        u64 count = 0;
6458        struct nfs4_client *clp;
6459        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6460                                                nfsd_net_id);
6461        LIST_HEAD(reaplist);
6462
6463        if (!nfsd_netns_ready(nn))
6464                return count;
6465
6466        spin_lock(&nn->client_lock);
6467        list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6468                count += nfsd_collect_client_openowners(clp, &reaplist,
6469                                                        max - count);
6470                if (max != 0 && count >= max)
6471                        break;
6472        }
6473        spin_unlock(&nn->client_lock);
6474        nfsd_reap_openowners(&reaplist);
6475        return count;
6476}
6477
6478static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max,
6479                                     struct list_head *victims)
6480{
6481        struct nfs4_delegation *dp, *next;
6482        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6483                                                nfsd_net_id);
6484        u64 count = 0;
6485
6486        lockdep_assert_held(&nn->client_lock);
6487
6488        spin_lock(&state_lock);
6489        list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) {
6490                if (victims) {
6491                        /*
6492                         * It's not safe to mess with delegations that have a
6493                         * non-zero dl_time. They might have already been broken
6494                         * and could be processed by the laundromat outside of
6495                         * the state_lock. Just leave them be.
6496                         */
6497                        if (dp->dl_time != 0)
6498                                continue;
6499
6500                        atomic_inc(&clp->cl_refcount);
6501                        WARN_ON(!unhash_delegation_locked(dp));
6502                        list_add(&dp->dl_recall_lru, victims);
6503                }
6504                ++count;
6505                /*
6506                 * Despite the fact that these functions deal with
6507                 * 64-bit integers for "count", we must ensure that
6508                 * it doesn't blow up the clp->cl_refcount. Throw a
6509                 * warning if we start to approach INT_MAX here.
6510                 */
6511                WARN_ON_ONCE(count == (INT_MAX / 2));
6512                if (count == max)
6513                        break;
6514        }
6515        spin_unlock(&state_lock);
6516        return count;
6517}
6518
6519static u64
6520nfsd_print_client_delegations(struct nfs4_client *clp)
6521{
6522        u64 count = nfsd_find_all_delegations(clp, 0, NULL);
6523
6524        nfsd_print_count(clp, count, "delegations");
6525        return count;
6526}
6527
6528u64
6529nfsd_inject_print_delegations(void)
6530{
6531        struct nfs4_client *clp;
6532        u64 count = 0;
6533        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6534                                                nfsd_net_id);
6535
6536        if (!nfsd_netns_ready(nn))
6537                return 0;
6538
6539        spin_lock(&nn->client_lock);
6540        list_for_each_entry(clp, &nn->client_lru, cl_lru)
6541                count += nfsd_print_client_delegations(clp);
6542        spin_unlock(&nn->client_lock);
6543
6544        return count;
6545}
6546
6547static void
6548nfsd_forget_delegations(struct list_head *reaplist)
6549{
6550        struct nfs4_client *clp;
6551        struct nfs4_delegation *dp, *next;
6552
6553        list_for_each_entry_safe(dp, next, reaplist, dl_recall_lru) {
6554                list_del_init(&dp->dl_recall_lru);
6555                clp = dp->dl_stid.sc_client;
6556                revoke_delegation(dp);
6557                put_client(clp);
6558        }
6559}
6560
6561u64
6562nfsd_inject_forget_client_delegations(struct sockaddr_storage *addr,
6563                                      size_t addr_size)
6564{
6565        u64 count = 0;
6566        struct nfs4_client *clp;
6567        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6568                                                nfsd_net_id);
6569        LIST_HEAD(reaplist);
6570
6571        if (!nfsd_netns_ready(nn))
6572                return count;
6573
6574        spin_lock(&nn->client_lock);
6575        clp = nfsd_find_client(addr, addr_size);
6576        if (clp)
6577                count = nfsd_find_all_delegations(clp, 0, &reaplist);
6578        spin_unlock(&nn->client_lock);
6579
6580        nfsd_forget_delegations(&reaplist);
6581        return count;
6582}
6583
6584u64
6585nfsd_inject_forget_delegations(u64 max)
6586{
6587        u64 count = 0;
6588        struct nfs4_client *clp;
6589        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6590                                                nfsd_net_id);
6591        LIST_HEAD(reaplist);
6592
6593        if (!nfsd_netns_ready(nn))
6594                return count;
6595
6596        spin_lock(&nn->client_lock);
6597        list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6598                count += nfsd_find_all_delegations(clp, max - count, &reaplist);
6599                if (max != 0 && count >= max)
6600                        break;
6601        }
6602        spin_unlock(&nn->client_lock);
6603        nfsd_forget_delegations(&reaplist);
6604        return count;
6605}
6606
6607static void
6608nfsd_recall_delegations(struct list_head *reaplist)
6609{
6610        struct nfs4_client *clp;
6611        struct nfs4_delegation *dp, *next;
6612
6613        list_for_each_entry_safe(dp, next, reaplist, dl_recall_lru) {
6614                list_del_init(&dp->dl_recall_lru);
6615                clp = dp->dl_stid.sc_client;
6616                /*
6617                 * We skipped all entries that had a zero dl_time before,
6618                 * so we can now reset the dl_time back to 0. If a delegation
6619                 * break comes in now, then it won't make any difference since
6620                 * we're recalling it either way.
6621                 */
6622                spin_lock(&state_lock);
6623                dp->dl_time = 0;
6624                spin_unlock(&state_lock);
6625                nfsd_break_one_deleg(dp);
6626                put_client(clp);
6627        }
6628}
6629
6630u64
6631nfsd_inject_recall_client_delegations(struct sockaddr_storage *addr,
6632                                      size_t addr_size)
6633{
6634        u64 count = 0;
6635        struct nfs4_client *clp;
6636        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6637                                                nfsd_net_id);
6638        LIST_HEAD(reaplist);
6639
6640        if (!nfsd_netns_ready(nn))
6641                return count;
6642
6643        spin_lock(&nn->client_lock);
6644        clp = nfsd_find_client(addr, addr_size);
6645        if (clp)
6646                count = nfsd_find_all_delegations(clp, 0, &reaplist);
6647        spin_unlock(&nn->client_lock);
6648
6649        nfsd_recall_delegations(&reaplist);
6650        return count;
6651}
6652
6653u64
6654nfsd_inject_recall_delegations(u64 max)
6655{
6656        u64 count = 0;
6657        struct nfs4_client *clp, *next;
6658        struct nfsd_net *nn = net_generic(current->nsproxy->net_ns,
6659                                                nfsd_net_id);
6660        LIST_HEAD(reaplist);
6661
6662        if (!nfsd_netns_ready(nn))
6663                return count;
6664
6665        spin_lock(&nn->client_lock);
6666        list_for_each_entry_safe(clp, next, &nn->client_lru, cl_lru) {
6667                count += nfsd_find_all_delegations(clp, max - count, &reaplist);
6668                if (max != 0 && ++count >= max)
6669                        break;
6670        }
6671        spin_unlock(&nn->client_lock);
6672        nfsd_recall_delegations(&reaplist);
6673        return count;
6674}
6675#endif /* CONFIG_NFSD_FAULT_INJECTION */
6676
6677/*
6678 * Since the lifetime of a delegation isn't limited to that of an open, a
6679 * client may quite reasonably hang on to a delegation as long as it has
6680 * the inode cached.  This becomes an obvious problem the first time a
6681 * client's inode cache approaches the size of the server's total memory.
6682 *
6683 * For now we avoid this problem by imposing a hard limit on the number
6684 * of delegations, which varies according to the server's memory size.
6685 */
6686static void
6687set_max_delegations(void)
6688{
6689        /*
6690         * Allow at most 4 delegations per megabyte of RAM.  Quick
6691         * estimates suggest that in the worst case (where every delegation
6692         * is for a different inode), a delegation could take about 1.5K,
6693         * giving a worst case usage of about 6% of memory.
6694         */
6695        max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
6696}
6697
6698static int nfs4_state_create_net(struct net *net)
6699{
6700        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6701        int i;
6702
6703        nn->conf_id_hashtbl = kmalloc(sizeof(struct list_head) *
6704                        CLIENT_HASH_SIZE, GFP_KERNEL);
6705        if (!nn->conf_id_hashtbl)
6706                goto err;
6707        nn->unconf_id_hashtbl = kmalloc(sizeof(struct list_head) *
6708                        CLIENT_HASH_SIZE, GFP_KERNEL);
6709        if (!nn->unconf_id_hashtbl)
6710                goto err_unconf_id;
6711        nn->sessionid_hashtbl = kmalloc(sizeof(struct list_head) *
6712                        SESSION_HASH_SIZE, GFP_KERNEL);
6713        if (!nn->sessionid_hashtbl)
6714                goto err_sessionid;
6715
6716        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6717                INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
6718                INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
6719        }
6720        for (i = 0; i < SESSION_HASH_SIZE; i++)
6721                INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
6722        nn->conf_name_tree = RB_ROOT;
6723        nn->unconf_name_tree = RB_ROOT;
6724        INIT_LIST_HEAD(&nn->client_lru);
6725        INIT_LIST_HEAD(&nn->close_lru);
6726        INIT_LIST_HEAD(&nn->del_recall_lru);
6727        spin_lock_init(&nn->client_lock);
6728
6729        INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
6730        get_net(net);
6731
6732        return 0;
6733
6734err_sessionid:
6735        kfree(nn->unconf_id_hashtbl);
6736err_unconf_id:
6737        kfree(nn->conf_id_hashtbl);
6738err:
6739        return -ENOMEM;
6740}
6741
6742static void
6743nfs4_state_destroy_net(struct net *net)
6744{
6745        int i;
6746        struct nfs4_client *clp = NULL;
6747        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6748
6749        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6750                while (!list_empty(&nn->conf_id_hashtbl[i])) {
6751                        clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
6752                        destroy_client(clp);
6753                }
6754        }
6755
6756        for (i = 0; i < CLIENT_HASH_SIZE; i++) {
6757                while (!list_empty(&nn->unconf_id_hashtbl[i])) {
6758                        clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
6759                        destroy_client(clp);
6760                }
6761        }
6762
6763        kfree(nn->sessionid_hashtbl);
6764        kfree(nn->unconf_id_hashtbl);
6765        kfree(nn->conf_id_hashtbl);
6766        put_net(net);
6767}
6768
6769int
6770nfs4_state_start_net(struct net *net)
6771{
6772        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6773        int ret;
6774
6775        ret = nfs4_state_create_net(net);
6776        if (ret)
6777                return ret;
6778        nn->boot_time = get_seconds();
6779        nn->grace_ended = false;
6780        nn->nfsd4_manager.block_opens = true;
6781        locks_start_grace(net, &nn->nfsd4_manager);
6782        nfsd4_client_tracking_init(net);
6783        printk(KERN_INFO "NFSD: starting %ld-second grace period (net %p)\n",
6784               nn->nfsd4_grace, net);
6785        queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
6786        return 0;
6787}
6788
6789/* initialization to perform when the nfsd service is started: */
6790
6791int
6792nfs4_state_start(void)
6793{
6794        int ret;
6795
6796        ret = set_callback_cred();
6797        if (ret)
6798                return -ENOMEM;
6799        laundry_wq = alloc_workqueue("%s", WQ_UNBOUND, 0, "nfsd4");
6800        if (laundry_wq == NULL) {
6801                ret = -ENOMEM;
6802                goto out_recovery;
6803        }
6804        ret = nfsd4_create_callback_queue();
6805        if (ret)
6806                goto out_free_laundry;
6807
6808        set_max_delegations();
6809
6810        return 0;
6811
6812out_free_laundry:
6813        destroy_workqueue(laundry_wq);
6814out_recovery:
6815        return ret;
6816}
6817
6818void
6819nfs4_state_shutdown_net(struct net *net)
6820{
6821        struct nfs4_delegation *dp = NULL;
6822        struct list_head *pos, *next, reaplist;
6823        struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6824
6825        cancel_delayed_work_sync(&nn->laundromat_work);
6826        locks_end_grace(&nn->nfsd4_manager);
6827
6828        INIT_LIST_HEAD(&reaplist);
6829        spin_lock(&state_lock);
6830        list_for_each_safe(pos, next, &nn->del_recall_lru) {
6831                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
6832                WARN_ON(!unhash_delegation_locked(dp));
6833                list_add(&dp->dl_recall_lru, &reaplist);
6834        }
6835        spin_unlock(&state_lock);
6836        list_for_each_safe(pos, next, &reaplist) {
6837                dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
6838                list_del_init(&dp->dl_recall_lru);
6839                put_clnt_odstate(dp->dl_clnt_odstate);
6840                nfs4_put_deleg_lease(dp->dl_stid.sc_file);
6841                nfs4_put_stid(&dp->dl_stid);
6842        }
6843
6844        nfsd4_client_tracking_exit(net);
6845        nfs4_state_destroy_net(net);
6846}
6847
6848void
6849nfs4_state_shutdown(void)
6850{
6851        destroy_workqueue(laundry_wq);
6852        nfsd4_destroy_callback_queue();
6853}
6854
6855static void
6856get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
6857{
6858        if (HAS_STATE_ID(cstate, CURRENT_STATE_ID_FLAG) && CURRENT_STATEID(stateid))
6859                memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
6860}
6861
6862static void
6863put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
6864{
6865        if (cstate->minorversion) {
6866                memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
6867                SET_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
6868        }
6869}
6870
6871void
6872clear_current_stateid(struct nfsd4_compound_state *cstate)
6873{
6874        CLEAR_STATE_ID(cstate, CURRENT_STATE_ID_FLAG);
6875}
6876
6877/*
6878 * functions to set current state id
6879 */
6880void
6881nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
6882{
6883        put_stateid(cstate, &odp->od_stateid);
6884}
6885
6886void
6887nfsd4_set_openstateid(struct nfsd4_compound_state *cstate, struct nfsd4_open *open)
6888{
6889        put_stateid(cstate, &open->op_stateid);
6890}
6891
6892void
6893nfsd4_set_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
6894{
6895        put_stateid(cstate, &close->cl_stateid);
6896}
6897
6898void
6899nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate, struct nfsd4_lock *lock)
6900{
6901        put_stateid(cstate, &lock->lk_resp_stateid);
6902}
6903
6904/*
6905 * functions to consume current state id
6906 */
6907
6908void
6909nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate, struct nfsd4_open_downgrade *odp)
6910{
6911        get_stateid(cstate, &odp->od_stateid);
6912}
6913
6914void
6915nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate, struct nfsd4_delegreturn *drp)
6916{
6917        get_stateid(cstate, &drp->dr_stateid);
6918}
6919
6920void
6921nfsd4_get_freestateid(struct nfsd4_compound_state *cstate, struct nfsd4_free_stateid *fsp)
6922{
6923        get_stateid(cstate, &fsp->fr_stateid);
6924}
6925
6926void
6927nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate, struct nfsd4_setattr *setattr)
6928{
6929        get_stateid(cstate, &setattr->sa_stateid);
6930}
6931
6932void
6933nfsd4_get_closestateid(struct nfsd4_compound_state *cstate, struct nfsd4_close *close)
6934{
6935        get_stateid(cstate, &close->cl_stateid);
6936}
6937
6938void
6939nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate, struct nfsd4_locku *locku)
6940{
6941        get_stateid(cstate, &locku->lu_stateid);
6942}
6943
6944void
6945nfsd4_get_readstateid(struct nfsd4_compound_state *cstate, struct nfsd4_read *read)
6946{
6947        get_stateid(cstate, &read->rd_stateid);
6948}
6949
6950void
6951nfsd4_get_writestateid(struct nfsd4_compound_state *cstate, struct nfsd4_write *write)
6952{
6953        get_stateid(cstate, &write->wr_stateid);
6954}
6955