1/*
   2*  Copyright (c) 2001 The Regents of the University of Michigan.
   3*  All rights reserved.
   4*
   5*  Kendrick Smith <kmsmith@umich.edu>
   6*  Andy Adamson <kandros@umich.edu>
   7*
   8*  Redistribution and use in source and binary forms, with or without
   9*  modification, are permitted provided that the following conditions
  10*  are met:
  11*
  12*  1. Redistributions of source code must retain the above copyright
  13*     notice, this list of conditions and the following disclaimer.
  14*  2. Redistributions in binary form must reproduce the above copyright
  15*     notice, this list of conditions and the following disclaimer in the
  16*     documentation and/or other materials provided with the distribution.
  17*  3. Neither the name of the University nor the names of its
  18*     contributors may be used to endorse or promote products derived
  19*     from this software without specific prior written permission.
  20*
  21*  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
  22*  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
  23*  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  24*  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  25*  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  26*  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  27*  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
  28*  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
  29*  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
  30*  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
  31*  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  32*
  33*/
  34
  35#include <linux/file.h>
  36#include <linux/fs.h>
  37#include <linux/slab.h>
  38#include <linux/namei.h>
  39#include <linux/swap.h>
  40#include <linux/pagemap.h>
  41#include <linux/ratelimit.h>
  42#include <linux/sunrpc/svcauth_gss.h>
  43#include <linux/sunrpc/addr.h>
  44#include <linux/jhash.h>
  45#include <linux/string_helpers.h>
  46#include <linux/fsnotify.h>
  47#include <linux/rhashtable.h>
  48#include <linux/nfs_ssc.h>
  49
  50#include "xdr4.h"
  51#include "xdr4cb.h"
  52#include "vfs.h"
  53#include "current_stateid.h"
  54
  55#include "netns.h"
  56#include "pnfs.h"
  57#include "filecache.h"
  58#include "trace.h"
  59
  60#define NFSDDBG_FACILITY                NFSDDBG_PROC
  61
  62#define all_ones {{~0,~0},~0}
  63static const stateid_t one_stateid = {
  64	.si_generation = ~0,
  65	.si_opaque = all_ones,
  66};
  67static const stateid_t zero_stateid = {
  68	/* all fields zero */
  69};
  70static const stateid_t currentstateid = {
  71	.si_generation = 1,
  72};
  73static const stateid_t close_stateid = {
  74	.si_generation = 0xffffffffU,
  75};
  76
  77static u64 current_sessionid = 1;
  78
  79#define ZERO_STATEID(stateid) (!memcmp((stateid), &zero_stateid, sizeof(stateid_t)))
  80#define ONE_STATEID(stateid)  (!memcmp((stateid), &one_stateid, sizeof(stateid_t)))
  81#define CURRENT_STATEID(stateid) (!memcmp((stateid), &currentstateid, sizeof(stateid_t)))
  82#define CLOSE_STATEID(stateid)  (!memcmp((stateid), &close_stateid, sizeof(stateid_t)))
  83
  84/* forward declarations */
  85static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
  86static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
  87void nfsd4_end_grace(struct nfsd_net *nn);
  88static void _free_cpntf_state_locked(struct nfsd_net *nn, struct nfs4_cpntf_state *cps);
  89static void nfsd4_file_hash_remove(struct nfs4_file *fi);
  90
  91/* Locking: */
  92
  93/*
  94 * Currently used for the del_recall_lru and file hash table.  In an
  95 * effort to decrease the scope of the client_mutex, this spinlock may
  96 * eventually cover more:
  97 */
  98static DEFINE_SPINLOCK(state_lock);
  99
 100enum nfsd4_st_mutex_lock_subclass {
 101	OPEN_STATEID_MUTEX = 0,
 102	LOCK_STATEID_MUTEX = 1,
 103};
 104
 105/*
 106 * A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
 107 * the refcount on the open stateid to drop.
 108 */
 109static DECLARE_WAIT_QUEUE_HEAD(close_wq);
 110
 111/*
 112 * A waitqueue where a writer to clients/#/ctl destroying a client can
 113 * wait for cl_rpc_users to drop to 0 and then for the client to be
 114 * unhashed.
 115 */
 116static DECLARE_WAIT_QUEUE_HEAD(expiry_wq);
 117
 118static struct kmem_cache *client_slab;
 119static struct kmem_cache *openowner_slab;
 120static struct kmem_cache *lockowner_slab;
 121static struct kmem_cache *file_slab;
 122static struct kmem_cache *stateid_slab;
 123static struct kmem_cache *deleg_slab;
 124static struct kmem_cache *odstate_slab;
 125
 126static void free_session(struct nfsd4_session *);
 127
 128static const struct nfsd4_callback_ops nfsd4_cb_recall_ops;
 129static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops;
 130
 131static struct workqueue_struct *laundry_wq;
 132
 133int nfsd4_create_laundry_wq(void)
 134{
 135	int rc = 0;
 136
 137	laundry_wq = alloc_workqueue("%s", WQ_UNBOUND, 0, "nfsd4");
 138	if (laundry_wq == NULL)
 139		rc = -ENOMEM;
 140	return rc;
 141}
 142
 143void nfsd4_destroy_laundry_wq(void)
 144{
 145	destroy_workqueue(laundry_wq);
 146}
 147
 148static bool is_session_dead(struct nfsd4_session *ses)
 149{
 150	return ses->se_flags & NFS4_SESSION_DEAD;
 151}
 152
 153static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
 154{
 155	if (atomic_read(&ses->se_ref) > ref_held_by_me)
 156		return nfserr_jukebox;
 157	ses->se_flags |= NFS4_SESSION_DEAD;
 158	return nfs_ok;
 159}
 160
 161static bool is_client_expired(struct nfs4_client *clp)
 162{
 163	return clp->cl_time == 0;
 164}
 165
 166static void nfsd4_dec_courtesy_client_count(struct nfsd_net *nn,
 167					struct nfs4_client *clp)
 168{
 169	if (clp->cl_state != NFSD4_ACTIVE)
 170		atomic_add_unless(&nn->nfsd_courtesy_clients, -1, 0);
 171}
 172
 173static __be32 get_client_locked(struct nfs4_client *clp)
 174{
 175	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 176
 177	lockdep_assert_held(&nn->client_lock);
 178
 179	if (is_client_expired(clp))
 180		return nfserr_expired;
 181	atomic_inc(&clp->cl_rpc_users);
 182	nfsd4_dec_courtesy_client_count(nn, clp);
 183	clp->cl_state = NFSD4_ACTIVE;
 184	return nfs_ok;
 185}
 186
 187/* must be called under the client_lock */
 188static inline void
 189renew_client_locked(struct nfs4_client *clp)
 190{
 191	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 192
 193	if (is_client_expired(clp)) {
 194		WARN_ON(1);
 195		printk("%s: client (clientid %08x/%08x) already expired\n",
 196			__func__,
 197			clp->cl_clientid.cl_boot,
 198			clp->cl_clientid.cl_id);
 199		return;
 200	}
 201
 202	list_move_tail(&clp->cl_lru, &nn->client_lru);
 203	clp->cl_time = ktime_get_boottime_seconds();
 204	nfsd4_dec_courtesy_client_count(nn, clp);
 205	clp->cl_state = NFSD4_ACTIVE;
 206}
 207
 208static void put_client_renew_locked(struct nfs4_client *clp)
 209{
 210	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 211
 212	lockdep_assert_held(&nn->client_lock);
 213
 214	if (!atomic_dec_and_test(&clp->cl_rpc_users))
 215		return;
 216	if (!is_client_expired(clp))
 217		renew_client_locked(clp);
 218	else
 219		wake_up_all(&expiry_wq);
 220}
 221
 222static void put_client_renew(struct nfs4_client *clp)
 223{
 224	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 225
 226	if (!atomic_dec_and_lock(&clp->cl_rpc_users, &nn->client_lock))
 227		return;
 228	if (!is_client_expired(clp))
 229		renew_client_locked(clp);
 230	else
 231		wake_up_all(&expiry_wq);
 232	spin_unlock(&nn->client_lock);
 233}
 234
 235static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
 236{
 237	__be32 status;
 238
 239	if (is_session_dead(ses))
 240		return nfserr_badsession;
 241	status = get_client_locked(ses->se_client);
 242	if (status)
 243		return status;
 244	atomic_inc(&ses->se_ref);
 245	return nfs_ok;
 246}
 247
 248static void nfsd4_put_session_locked(struct nfsd4_session *ses)
 249{
 250	struct nfs4_client *clp = ses->se_client;
 251	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 252
 253	lockdep_assert_held(&nn->client_lock);
 254
 255	if (atomic_dec_and_test(&ses->se_ref) && is_session_dead(ses))
 256		free_session(ses);
 257	put_client_renew_locked(clp);
 258}
 259
 260static void nfsd4_put_session(struct nfsd4_session *ses)
 261{
 262	struct nfs4_client *clp = ses->se_client;
 263	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 264
 265	spin_lock(&nn->client_lock);
 266	nfsd4_put_session_locked(ses);
 267	spin_unlock(&nn->client_lock);
 268}
 269
 270static struct nfsd4_blocked_lock *
 271find_blocked_lock(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
 272			struct nfsd_net *nn)
 273{
 274	struct nfsd4_blocked_lock *cur, *found = NULL;
 275
 276	spin_lock(&nn->blocked_locks_lock);
 277	list_for_each_entry(cur, &lo->lo_blocked, nbl_list) {
 278		if (fh_match(fh, &cur->nbl_fh)) {
 279			list_del_init(&cur->nbl_list);
 280			WARN_ON(list_empty(&cur->nbl_lru));
 281			list_del_init(&cur->nbl_lru);
 282			found = cur;
 283			break;
 284		}
 285	}
 286	spin_unlock(&nn->blocked_locks_lock);
 287	if (found)
 288		locks_delete_block(&found->nbl_lock);
 289	return found;
 290}
 291
 292static struct nfsd4_blocked_lock *
 293find_or_allocate_block(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
 294			struct nfsd_net *nn)
 295{
 296	struct nfsd4_blocked_lock *nbl;
 297
 298	nbl = find_blocked_lock(lo, fh, nn);
 299	if (!nbl) {
 300		nbl= kmalloc(sizeof(*nbl), GFP_KERNEL);
 301		if (nbl) {
 302			INIT_LIST_HEAD(&nbl->nbl_list);
 303			INIT_LIST_HEAD(&nbl->nbl_lru);
 304			fh_copy_shallow(&nbl->nbl_fh, fh);
 305			locks_init_lock(&nbl->nbl_lock);
 306			kref_init(&nbl->nbl_kref);
 307			nfsd4_init_cb(&nbl->nbl_cb, lo->lo_owner.so_client,
 308					&nfsd4_cb_notify_lock_ops,
 309					NFSPROC4_CLNT_CB_NOTIFY_LOCK);
 310		}
 311	}
 312	return nbl;
 313}
 314
 315static void
 316free_nbl(struct kref *kref)
 317{
 318	struct nfsd4_blocked_lock *nbl;
 319
 320	nbl = container_of(kref, struct nfsd4_blocked_lock, nbl_kref);
 321	kfree(nbl);
 322}
 323
 324static void
 325free_blocked_lock(struct nfsd4_blocked_lock *nbl)
 326{
 327	locks_delete_block(&nbl->nbl_lock);
 328	locks_release_private(&nbl->nbl_lock);
 329	kref_put(&nbl->nbl_kref, free_nbl);
 330}
 331
 332static void
 333remove_blocked_locks(struct nfs4_lockowner *lo)
 334{
 335	struct nfs4_client *clp = lo->lo_owner.so_client;
 336	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
 337	struct nfsd4_blocked_lock *nbl;
 338	LIST_HEAD(reaplist);
 339
 340	/* Dequeue all blocked locks */
 341	spin_lock(&nn->blocked_locks_lock);
 342	while (!list_empty(&lo->lo_blocked)) {
 343		nbl = list_first_entry(&lo->lo_blocked,
 344					struct nfsd4_blocked_lock,
 345					nbl_list);
 346		list_del_init(&nbl->nbl_list);
 347		WARN_ON(list_empty(&nbl->nbl_lru));
 348		list_move(&nbl->nbl_lru, &reaplist);
 349	}
 350	spin_unlock(&nn->blocked_locks_lock);
 351
 352	/* Now free them */
 353	while (!list_empty(&reaplist)) {
 354		nbl = list_first_entry(&reaplist, struct nfsd4_blocked_lock,
 355					nbl_lru);
 356		list_del_init(&nbl->nbl_lru);
 357		free_blocked_lock(nbl);
 358	}
 359}
 360
 361static void
 362nfsd4_cb_notify_lock_prepare(struct nfsd4_callback *cb)
 363{
 364	struct nfsd4_blocked_lock	*nbl = container_of(cb,
 365						struct nfsd4_blocked_lock, nbl_cb);
 366	locks_delete_block(&nbl->nbl_lock);
 367}
 368
 369static int
 370nfsd4_cb_notify_lock_done(struct nfsd4_callback *cb, struct rpc_task *task)
 371{
 372	trace_nfsd_cb_notify_lock_done(&zero_stateid, task);
 373
 374	/*
 375	 * Since this is just an optimization, we don't try very hard if it
 376	 * turns out not to succeed. We'll requeue it on NFS4ERR_DELAY, and
 377	 * just quit trying on anything else.
 378	 */
 379	switch (task->tk_status) {
 380	case -NFS4ERR_DELAY:
 381		rpc_delay(task, 1 * HZ);
 382		return 0;
 383	default:
 384		return 1;
 385	}
 386}
 387
 388static void
 389nfsd4_cb_notify_lock_release(struct nfsd4_callback *cb)
 390{
 391	struct nfsd4_blocked_lock	*nbl = container_of(cb,
 392						struct nfsd4_blocked_lock, nbl_cb);
 393
 394	free_blocked_lock(nbl);
 395}
 396
 397static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops = {
 398	.prepare	= nfsd4_cb_notify_lock_prepare,
 399	.done		= nfsd4_cb_notify_lock_done,
 400	.release	= nfsd4_cb_notify_lock_release,
 401};
 402
 403/*
 404 * We store the NONE, READ, WRITE, and BOTH bits separately in the
 405 * st_{access,deny}_bmap field of the stateid, in order to track not
 406 * only what share bits are currently in force, but also what
 407 * combinations of share bits previous opens have used.  This allows us
 408 * to enforce the recommendation in
 409 * https://datatracker.ietf.org/doc/html/rfc7530#section-16.19.4 that
 410 * the server return an error if the client attempt to downgrade to a
 411 * combination of share bits not explicable by closing some of its
 412 * previous opens.
 413 *
 414 * This enforcement is arguably incomplete, since we don't keep
 415 * track of access/deny bit combinations; so, e.g., we allow:
 416 *
 417 *	OPEN allow read, deny write
 418 *	OPEN allow both, deny none
 419 *	DOWNGRADE allow read, deny none
 420 *
 421 * which we should reject.
 422 *
 423 * But you could also argue that our current code is already overkill,
 424 * since it only exists to return NFS4ERR_INVAL on incorrect client
 425 * behavior.
 426 */
 427static unsigned int
 428bmap_to_share_mode(unsigned long bmap)
 429{
 430	int i;
 431	unsigned int access = 0;
 432
 433	for (i = 1; i < 4; i++) {
 434		if (test_bit(i, &bmap))
 435			access |= i;
 436	}
 437	return access;
 438}
 439
 440/* set share access for a given stateid */
 441static inline void
 442set_access(u32 access, struct nfs4_ol_stateid *stp)
 443{
 444	unsigned char mask = 1 << access;
 445
 446	WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
 447	stp->st_access_bmap |= mask;
 448}
 449
 450/* clear share access for a given stateid */
 451static inline void
 452clear_access(u32 access, struct nfs4_ol_stateid *stp)
 453{
 454	unsigned char mask = 1 << access;
 455
 456	WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
 457	stp->st_access_bmap &= ~mask;
 458}
 459
 460/* test whether a given stateid has access */
 461static inline bool
 462test_access(u32 access, struct nfs4_ol_stateid *stp)
 463{
 464	unsigned char mask = 1 << access;
 465
 466	return (bool)(stp->st_access_bmap & mask);
 467}
 468
 469/* set share deny for a given stateid */
 470static inline void
 471set_deny(u32 deny, struct nfs4_ol_stateid *stp)
 472{
 473	unsigned char mask = 1 << deny;
 474
 475	WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
 476	stp->st_deny_bmap |= mask;
 477}
 478
 479/* clear share deny for a given stateid */
 480static inline void
 481clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
 482{
 483	unsigned char mask = 1 << deny;
 484
 485	WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
 486	stp->st_deny_bmap &= ~mask;
 487}
 488
 489/* test whether a given stateid is denying specific access */
 490static inline bool
 491test_deny(u32 deny, struct nfs4_ol_stateid *stp)
 492{
 493	unsigned char mask = 1 << deny;
 494
 495	return (bool)(stp->st_deny_bmap & mask);
 496}
 497
 498static int nfs4_access_to_omode(u32 access)
 499{
 500	switch (access & NFS4_SHARE_ACCESS_BOTH) {
 501	case NFS4_SHARE_ACCESS_READ:
 502		return O_RDONLY;
 503	case NFS4_SHARE_ACCESS_WRITE:
 504		return O_WRONLY;
 505	case NFS4_SHARE_ACCESS_BOTH:
 506		return O_RDWR;
 507	}
 508	WARN_ON_ONCE(1);
 509	return O_RDONLY;
 510}
 511
 512static inline int
 513access_permit_read(struct nfs4_ol_stateid *stp)
 514{
 515	return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
 516		test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
 517		test_access(NFS4_SHARE_ACCESS_WRITE, stp);
 518}
 519
 520static inline int
 521access_permit_write(struct nfs4_ol_stateid *stp)
 522{
 523	return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
 524		test_access(NFS4_SHARE_ACCESS_BOTH, stp);
 525}
 526
 527static inline struct nfs4_stateowner *
 528nfs4_get_stateowner(struct nfs4_stateowner *sop)
 529{
 530	atomic_inc(&sop->so_count);
 531	return sop;
 532}
 533
 534static int
 535same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner)
 536{
 537	return (sop->so_owner.len == owner->len) &&
 538		0 == memcmp(sop->so_owner.data, owner->data, owner->len);
 539}
 540
 541static struct nfs4_openowner *
 542find_openstateowner_str_locked(unsigned int hashval, struct nfsd4_open *open,
 543			struct nfs4_client *clp)
 544{
 545	struct nfs4_stateowner *so;
 546
 547	lockdep_assert_held(&clp->cl_lock);
 548
 549	list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[hashval],
 550			    so_strhash) {
 551		if (!so->so_is_open_owner)
 552			continue;
 553		if (same_owner_str(so, &open->op_owner))
 554			return openowner(nfs4_get_stateowner(so));
 555	}
 556	return NULL;
 557}
 558
 559static struct nfs4_openowner *
 560find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
 561			struct nfs4_client *clp)
 562{
 563	struct nfs4_openowner *oo;
 564
 565	spin_lock(&clp->cl_lock);
 566	oo = find_openstateowner_str_locked(hashval, open, clp);
 567	spin_unlock(&clp->cl_lock);
 568	return oo;
 569}
 570
 571static inline u32
 572opaque_hashval(const void *ptr, int nbytes)
 573{
 574	unsigned char *cptr = (unsigned char *) ptr;
 575
 576	u32 x = 0;
 577	while (nbytes--) {
 578		x *= 37;
 579		x += *cptr++;
 580	}
 581	return x;
 582}
 583
 584static void nfsd4_free_file_rcu(struct rcu_head *rcu)
 585{
 586	struct nfs4_file *fp = container_of(rcu, struct nfs4_file, fi_rcu);
 587
 588	kmem_cache_free(file_slab, fp);
 589}
 590
 591void
 592put_nfs4_file(struct nfs4_file *fi)
 593{
 594	if (refcount_dec_and_test(&fi->fi_ref)) {
 595		nfsd4_file_hash_remove(fi);
 
 
 
 596		WARN_ON_ONCE(!list_empty(&fi->fi_clnt_odstate));
 597		WARN_ON_ONCE(!list_empty(&fi->fi_delegations));
 598		call_rcu(&fi->fi_rcu, nfsd4_free_file_rcu);
 599	}
 600}
 601
 602static struct nfsd_file *
 603__nfs4_get_fd(struct nfs4_file *f, int oflag)
 604{
 605	if (f->fi_fds[oflag])
 606		return nfsd_file_get(f->fi_fds[oflag]);
 607	return NULL;
 608}
 609
 610static struct nfsd_file *
 611find_writeable_file_locked(struct nfs4_file *f)
 612{
 613	struct nfsd_file *ret;
 614
 615	lockdep_assert_held(&f->fi_lock);
 616
 617	ret = __nfs4_get_fd(f, O_WRONLY);
 618	if (!ret)
 619		ret = __nfs4_get_fd(f, O_RDWR);
 620	return ret;
 621}
 622
 623static struct nfsd_file *
 624find_writeable_file(struct nfs4_file *f)
 625{
 626	struct nfsd_file *ret;
 627
 628	spin_lock(&f->fi_lock);
 629	ret = find_writeable_file_locked(f);
 630	spin_unlock(&f->fi_lock);
 631
 632	return ret;
 633}
 634
 635static struct nfsd_file *
 636find_readable_file_locked(struct nfs4_file *f)
 637{
 638	struct nfsd_file *ret;
 639
 640	lockdep_assert_held(&f->fi_lock);
 641
 642	ret = __nfs4_get_fd(f, O_RDONLY);
 643	if (!ret)
 644		ret = __nfs4_get_fd(f, O_RDWR);
 645	return ret;
 646}
 647
 648static struct nfsd_file *
 649find_readable_file(struct nfs4_file *f)
 650{
 651	struct nfsd_file *ret;
 652
 653	spin_lock(&f->fi_lock);
 654	ret = find_readable_file_locked(f);
 655	spin_unlock(&f->fi_lock);
 656
 657	return ret;
 658}
 659
 660struct nfsd_file *
 661find_any_file(struct nfs4_file *f)
 662{
 663	struct nfsd_file *ret;
 664
 665	if (!f)
 666		return NULL;
 667	spin_lock(&f->fi_lock);
 668	ret = __nfs4_get_fd(f, O_RDWR);
 669	if (!ret) {
 670		ret = __nfs4_get_fd(f, O_WRONLY);
 671		if (!ret)
 672			ret = __nfs4_get_fd(f, O_RDONLY);
 673	}
 674	spin_unlock(&f->fi_lock);
 675	return ret;
 676}
 677
 678static struct nfsd_file *find_any_file_locked(struct nfs4_file *f)
 679{
 680	lockdep_assert_held(&f->fi_lock);
 681
 682	if (f->fi_fds[O_RDWR])
 683		return f->fi_fds[O_RDWR];
 684	if (f->fi_fds[O_WRONLY])
 685		return f->fi_fds[O_WRONLY];
 686	if (f->fi_fds[O_RDONLY])
 687		return f->fi_fds[O_RDONLY];
 688	return NULL;
 689}
 690
 691static struct nfsd_file *find_deleg_file_locked(struct nfs4_file *f)
 692{
 693	lockdep_assert_held(&f->fi_lock);
 694
 
 695	if (f->fi_deleg_file)
 696		return f->fi_deleg_file;
 697	return NULL;
 
 698}
 699
 700static atomic_long_t num_delegations;
 701unsigned long max_delegations;
 702
 703/*
 704 * Open owner state (share locks)
 705 */
 706
 707/* hash tables for lock and open owners */
 708#define OWNER_HASH_BITS              8
 709#define OWNER_HASH_SIZE             (1 << OWNER_HASH_BITS)
 710#define OWNER_HASH_MASK             (OWNER_HASH_SIZE - 1)
 711
 712static unsigned int ownerstr_hashval(struct xdr_netobj *ownername)
 713{
 714	unsigned int ret;
 715
 716	ret = opaque_hashval(ownername->data, ownername->len);
 717	return ret & OWNER_HASH_MASK;
 718}
 719
 720static struct rhltable nfs4_file_rhltable ____cacheline_aligned_in_smp;
 721
 722static const struct rhashtable_params nfs4_file_rhash_params = {
 723	.key_len		= sizeof_field(struct nfs4_file, fi_inode),
 724	.key_offset		= offsetof(struct nfs4_file, fi_inode),
 725	.head_offset		= offsetof(struct nfs4_file, fi_rlist),
 726
 727	/*
 728	 * Start with a single page hash table to reduce resizing churn
 729	 * on light workloads.
 730	 */
 731	.min_size		= 256,
 732	.automatic_shrinking	= true,
 733};
 734
 735/*
 736 * Check if courtesy clients have conflicting access and resolve it if possible
 737 *
 738 * access:  is op_share_access if share_access is true.
 739 *	    Check if access mode, op_share_access, would conflict with
 740 *	    the current deny mode of the file 'fp'.
 741 * access:  is op_share_deny if share_access is false.
 742 *	    Check if the deny mode, op_share_deny, would conflict with
 743 *	    current access of the file 'fp'.
 744 * stp:     skip checking this entry.
 745 * new_stp: normal open, not open upgrade.
 746 *
 747 * Function returns:
 748 *	false - access/deny mode conflict with normal client.
 749 *	true  - no conflict or conflict with courtesy client(s) is resolved.
 750 */
 751static bool
 752nfs4_resolve_deny_conflicts_locked(struct nfs4_file *fp, bool new_stp,
 753		struct nfs4_ol_stateid *stp, u32 access, bool share_access)
 754{
 755	struct nfs4_ol_stateid *st;
 756	bool resolvable = true;
 757	unsigned char bmap;
 758	struct nfsd_net *nn;
 759	struct nfs4_client *clp;
 760
 761	lockdep_assert_held(&fp->fi_lock);
 762	list_for_each_entry(st, &fp->fi_stateids, st_perfile) {
 763		/* ignore lock stateid */
 764		if (st->st_openstp)
 765			continue;
 766		if (st == stp && new_stp)
 767			continue;
 768		/* check file access against deny mode or vice versa */
 769		bmap = share_access ? st->st_deny_bmap : st->st_access_bmap;
 770		if (!(access & bmap_to_share_mode(bmap)))
 771			continue;
 772		clp = st->st_stid.sc_client;
 773		if (try_to_expire_client(clp))
 774			continue;
 775		resolvable = false;
 776		break;
 777	}
 778	if (resolvable) {
 779		clp = stp->st_stid.sc_client;
 780		nn = net_generic(clp->net, nfsd_net_id);
 781		mod_delayed_work(laundry_wq, &nn->laundromat_work, 0);
 782	}
 783	return resolvable;
 784}
 785
 
 
 786static void
 787__nfs4_file_get_access(struct nfs4_file *fp, u32 access)
 788{
 789	lockdep_assert_held(&fp->fi_lock);
 790
 791	if (access & NFS4_SHARE_ACCESS_WRITE)
 792		atomic_inc(&fp->fi_access[O_WRONLY]);
 793	if (access & NFS4_SHARE_ACCESS_READ)
 794		atomic_inc(&fp->fi_access[O_RDONLY]);
 795}
 796
 797static __be32
 798nfs4_file_get_access(struct nfs4_file *fp, u32 access)
 799{
 800	lockdep_assert_held(&fp->fi_lock);
 801
 802	/* Does this access mode make sense? */
 803	if (access & ~NFS4_SHARE_ACCESS_BOTH)
 804		return nfserr_inval;
 805
 806	/* Does it conflict with a deny mode already set? */
 807	if ((access & fp->fi_share_deny) != 0)
 808		return nfserr_share_denied;
 809
 810	__nfs4_file_get_access(fp, access);
 811	return nfs_ok;
 812}
 813
 814static __be32 nfs4_file_check_deny(struct nfs4_file *fp, u32 deny)
 815{
 816	/* Common case is that there is no deny mode. */
 817	if (deny) {
 818		/* Does this deny mode make sense? */
 819		if (deny & ~NFS4_SHARE_DENY_BOTH)
 820			return nfserr_inval;
 821
 822		if ((deny & NFS4_SHARE_DENY_READ) &&
 823		    atomic_read(&fp->fi_access[O_RDONLY]))
 824			return nfserr_share_denied;
 825
 826		if ((deny & NFS4_SHARE_DENY_WRITE) &&
 827		    atomic_read(&fp->fi_access[O_WRONLY]))
 828			return nfserr_share_denied;
 829	}
 830	return nfs_ok;
 831}
 832
 833static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
 834{
 835	might_lock(&fp->fi_lock);
 836
 837	if (atomic_dec_and_lock(&fp->fi_access[oflag], &fp->fi_lock)) {
 838		struct nfsd_file *f1 = NULL;
 839		struct nfsd_file *f2 = NULL;
 840
 841		swap(f1, fp->fi_fds[oflag]);
 842		if (atomic_read(&fp->fi_access[1 - oflag]) == 0)
 843			swap(f2, fp->fi_fds[O_RDWR]);
 844		spin_unlock(&fp->fi_lock);
 845		if (f1)
 846			nfsd_file_put(f1);
 847		if (f2)
 848			nfsd_file_put(f2);
 849	}
 850}
 851
 852static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
 853{
 854	WARN_ON_ONCE(access & ~NFS4_SHARE_ACCESS_BOTH);
 855
 856	if (access & NFS4_SHARE_ACCESS_WRITE)
 857		__nfs4_file_put_access(fp, O_WRONLY);
 858	if (access & NFS4_SHARE_ACCESS_READ)
 859		__nfs4_file_put_access(fp, O_RDONLY);
 860}
 861
 862/*
 863 * Allocate a new open/delegation state counter. This is needed for
 864 * pNFS for proper return on close semantics.
 865 *
 866 * Note that we only allocate it for pNFS-enabled exports, otherwise
 867 * all pointers to struct nfs4_clnt_odstate are always NULL.
 868 */
 869static struct nfs4_clnt_odstate *
 870alloc_clnt_odstate(struct nfs4_client *clp)
 871{
 872	struct nfs4_clnt_odstate *co;
 873
 874	co = kmem_cache_zalloc(odstate_slab, GFP_KERNEL);
 875	if (co) {
 876		co->co_client = clp;
 877		refcount_set(&co->co_odcount, 1);
 878	}
 879	return co;
 880}
 881
 882static void
 883hash_clnt_odstate_locked(struct nfs4_clnt_odstate *co)
 884{
 885	struct nfs4_file *fp = co->co_file;
 886
 887	lockdep_assert_held(&fp->fi_lock);
 888	list_add(&co->co_perfile, &fp->fi_clnt_odstate);
 889}
 890
 891static inline void
 892get_clnt_odstate(struct nfs4_clnt_odstate *co)
 893{
 894	if (co)
 895		refcount_inc(&co->co_odcount);
 896}
 897
 898static void
 899put_clnt_odstate(struct nfs4_clnt_odstate *co)
 900{
 901	struct nfs4_file *fp;
 902
 903	if (!co)
 904		return;
 905
 906	fp = co->co_file;
 907	if (refcount_dec_and_lock(&co->co_odcount, &fp->fi_lock)) {
 908		list_del(&co->co_perfile);
 909		spin_unlock(&fp->fi_lock);
 910
 911		nfsd4_return_all_file_layouts(co->co_client, fp);
 912		kmem_cache_free(odstate_slab, co);
 913	}
 914}
 915
 916static struct nfs4_clnt_odstate *
 917find_or_hash_clnt_odstate(struct nfs4_file *fp, struct nfs4_clnt_odstate *new)
 918{
 919	struct nfs4_clnt_odstate *co;
 920	struct nfs4_client *cl;
 921
 922	if (!new)
 923		return NULL;
 924
 925	cl = new->co_client;
 926
 927	spin_lock(&fp->fi_lock);
 928	list_for_each_entry(co, &fp->fi_clnt_odstate, co_perfile) {
 929		if (co->co_client == cl) {
 930			get_clnt_odstate(co);
 931			goto out;
 932		}
 933	}
 934	co = new;
 935	co->co_file = fp;
 936	hash_clnt_odstate_locked(new);
 937out:
 938	spin_unlock(&fp->fi_lock);
 939	return co;
 940}
 941
 942struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab,
 943				  void (*sc_free)(struct nfs4_stid *))
 944{
 945	struct nfs4_stid *stid;
 946	int new_id;
 947
 948	stid = kmem_cache_zalloc(slab, GFP_KERNEL);
 949	if (!stid)
 950		return NULL;
 951
 952	idr_preload(GFP_KERNEL);
 953	spin_lock(&cl->cl_lock);
 954	/* Reserving 0 for start of file in nfsdfs "states" file: */
 955	new_id = idr_alloc_cyclic(&cl->cl_stateids, stid, 1, 0, GFP_NOWAIT);
 956	spin_unlock(&cl->cl_lock);
 957	idr_preload_end();
 958	if (new_id < 0)
 959		goto out_free;
 960
 961	stid->sc_free = sc_free;
 962	stid->sc_client = cl;
 963	stid->sc_stateid.si_opaque.so_id = new_id;
 964	stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
 965	/* Will be incremented before return to client: */
 966	refcount_set(&stid->sc_count, 1);
 967	spin_lock_init(&stid->sc_lock);
 968	INIT_LIST_HEAD(&stid->sc_cp_list);
 969
 970	/*
 971	 * It shouldn't be a problem to reuse an opaque stateid value.
 972	 * I don't think it is for 4.1.  But with 4.0 I worry that, for
 973	 * example, a stray write retransmission could be accepted by
 974	 * the server when it should have been rejected.  Therefore,
 975	 * adopt a trick from the sctp code to attempt to maximize the
 976	 * amount of time until an id is reused, by ensuring they always
 977	 * "increase" (mod INT_MAX):
 978	 */
 979	return stid;
 980out_free:
 981	kmem_cache_free(slab, stid);
 982	return NULL;
 983}
 984
 985/*
 986 * Create a unique stateid_t to represent each COPY.
 987 */
 988static int nfs4_init_cp_state(struct nfsd_net *nn, copy_stateid_t *stid,
 989			      unsigned char cs_type)
 990{
 991	int new_id;
 992
 993	stid->cs_stid.si_opaque.so_clid.cl_boot = (u32)nn->boot_time;
 994	stid->cs_stid.si_opaque.so_clid.cl_id = nn->s2s_cp_cl_id;
 995	stid->cs_type = cs_type;
 996
 997	idr_preload(GFP_KERNEL);
 998	spin_lock(&nn->s2s_cp_lock);
 999	new_id = idr_alloc_cyclic(&nn->s2s_cp_stateids, stid, 0, 0, GFP_NOWAIT);
1000	stid->cs_stid.si_opaque.so_id = new_id;
1001	stid->cs_stid.si_generation = 1;
1002	spin_unlock(&nn->s2s_cp_lock);
1003	idr_preload_end();
1004	if (new_id < 0)
1005		return 0;
1006	return 1;
1007}
1008
1009int nfs4_init_copy_state(struct nfsd_net *nn, struct nfsd4_copy *copy)
1010{
1011	return nfs4_init_cp_state(nn, &copy->cp_stateid, NFS4_COPY_STID);
1012}
1013
1014struct nfs4_cpntf_state *nfs4_alloc_init_cpntf_state(struct nfsd_net *nn,
1015						     struct nfs4_stid *p_stid)
1016{
1017	struct nfs4_cpntf_state *cps;
1018
1019	cps = kzalloc(sizeof(struct nfs4_cpntf_state), GFP_KERNEL);
1020	if (!cps)
1021		return NULL;
1022	cps->cpntf_time = ktime_get_boottime_seconds();
1023	refcount_set(&cps->cp_stateid.cs_count, 1);
1024	if (!nfs4_init_cp_state(nn, &cps->cp_stateid, NFS4_COPYNOTIFY_STID))
1025		goto out_free;
1026	spin_lock(&nn->s2s_cp_lock);
1027	list_add(&cps->cp_list, &p_stid->sc_cp_list);
1028	spin_unlock(&nn->s2s_cp_lock);
1029	return cps;
1030out_free:
1031	kfree(cps);
1032	return NULL;
1033}
1034
1035void nfs4_free_copy_state(struct nfsd4_copy *copy)
1036{
1037	struct nfsd_net *nn;
1038
1039	WARN_ON_ONCE(copy->cp_stateid.cs_type != NFS4_COPY_STID);
1040	nn = net_generic(copy->cp_clp->net, nfsd_net_id);
1041	spin_lock(&nn->s2s_cp_lock);
1042	idr_remove(&nn->s2s_cp_stateids,
1043		   copy->cp_stateid.cs_stid.si_opaque.so_id);
1044	spin_unlock(&nn->s2s_cp_lock);
1045}
1046
1047static void nfs4_free_cpntf_statelist(struct net *net, struct nfs4_stid *stid)
1048{
1049	struct nfs4_cpntf_state *cps;
1050	struct nfsd_net *nn;
1051
1052	nn = net_generic(net, nfsd_net_id);
1053	spin_lock(&nn->s2s_cp_lock);
1054	while (!list_empty(&stid->sc_cp_list)) {
1055		cps = list_first_entry(&stid->sc_cp_list,
1056				       struct nfs4_cpntf_state, cp_list);
1057		_free_cpntf_state_locked(nn, cps);
1058	}
1059	spin_unlock(&nn->s2s_cp_lock);
1060}
1061
1062static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
1063{
1064	struct nfs4_stid *stid;
1065
1066	stid = nfs4_alloc_stid(clp, stateid_slab, nfs4_free_ol_stateid);
1067	if (!stid)
1068		return NULL;
1069
1070	return openlockstateid(stid);
1071}
1072
1073static void nfs4_free_deleg(struct nfs4_stid *stid)
1074{
1075	struct nfs4_delegation *dp = delegstateid(stid);
1076
1077	WARN_ON_ONCE(!list_empty(&stid->sc_cp_list));
1078	WARN_ON_ONCE(!list_empty(&dp->dl_perfile));
1079	WARN_ON_ONCE(!list_empty(&dp->dl_perclnt));
1080	WARN_ON_ONCE(!list_empty(&dp->dl_recall_lru));
1081	kmem_cache_free(deleg_slab, stid);
1082	atomic_long_dec(&num_delegations);
1083}
1084
1085/*
1086 * When we recall a delegation, we should be careful not to hand it
1087 * out again straight away.
1088 * To ensure this we keep a pair of bloom filters ('new' and 'old')
1089 * in which the filehandles of recalled delegations are "stored".
1090 * If a filehandle appear in either filter, a delegation is blocked.
1091 * When a delegation is recalled, the filehandle is stored in the "new"
1092 * filter.
1093 * Every 30 seconds we swap the filters and clear the "new" one,
1094 * unless both are empty of course.
1095 *
1096 * Each filter is 256 bits.  We hash the filehandle to 32bit and use the
1097 * low 3 bytes as hash-table indices.
1098 *
1099 * 'blocked_delegations_lock', which is always taken in block_delegations(),
1100 * is used to manage concurrent access.  Testing does not need the lock
1101 * except when swapping the two filters.
1102 */
1103static DEFINE_SPINLOCK(blocked_delegations_lock);
1104static struct bloom_pair {
1105	int	entries, old_entries;
1106	time64_t swap_time;
1107	int	new; /* index into 'set' */
1108	DECLARE_BITMAP(set[2], 256);
1109} blocked_delegations;
1110
1111static int delegation_blocked(struct knfsd_fh *fh)
1112{
1113	u32 hash;
1114	struct bloom_pair *bd = &blocked_delegations;
1115
1116	if (bd->entries == 0)
1117		return 0;
1118	if (ktime_get_seconds() - bd->swap_time > 30) {
1119		spin_lock(&blocked_delegations_lock);
1120		if (ktime_get_seconds() - bd->swap_time > 30) {
1121			bd->entries -= bd->old_entries;
1122			bd->old_entries = bd->entries;
1123			memset(bd->set[bd->new], 0,
1124			       sizeof(bd->set[0]));
1125			bd->new = 1-bd->new;
1126			bd->swap_time = ktime_get_seconds();
1127		}
1128		spin_unlock(&blocked_delegations_lock);
1129	}
1130	hash = jhash(&fh->fh_raw, fh->fh_size, 0);
1131	if (test_bit(hash&255, bd->set[0]) &&
1132	    test_bit((hash>>8)&255, bd->set[0]) &&
1133	    test_bit((hash>>16)&255, bd->set[0]))
1134		return 1;
1135
1136	if (test_bit(hash&255, bd->set[1]) &&
1137	    test_bit((hash>>8)&255, bd->set[1]) &&
1138	    test_bit((hash>>16)&255, bd->set[1]))
1139		return 1;
1140
1141	return 0;
1142}
1143
1144static void block_delegations(struct knfsd_fh *fh)
1145{
1146	u32 hash;
1147	struct bloom_pair *bd = &blocked_delegations;
1148
1149	hash = jhash(&fh->fh_raw, fh->fh_size, 0);
1150
1151	spin_lock(&blocked_delegations_lock);
1152	__set_bit(hash&255, bd->set[bd->new]);
1153	__set_bit((hash>>8)&255, bd->set[bd->new]);
1154	__set_bit((hash>>16)&255, bd->set[bd->new]);
1155	if (bd->entries == 0)
1156		bd->swap_time = ktime_get_seconds();
1157	bd->entries += 1;
1158	spin_unlock(&blocked_delegations_lock);
1159}
1160
1161static struct nfs4_delegation *
1162alloc_init_deleg(struct nfs4_client *clp, struct nfs4_file *fp,
 
1163		 struct nfs4_clnt_odstate *odstate)
1164{
1165	struct nfs4_delegation *dp;
1166	long n;
1167
1168	dprintk("NFSD alloc_init_deleg\n");
1169	n = atomic_long_inc_return(&num_delegations);
1170	if (n < 0 || n > max_delegations)
1171		goto out_dec;
1172	if (delegation_blocked(&fp->fi_fhandle))
1173		goto out_dec;
1174	dp = delegstateid(nfs4_alloc_stid(clp, deleg_slab, nfs4_free_deleg));
1175	if (dp == NULL)
1176		goto out_dec;
1177
1178	/*
1179	 * delegation seqid's are never incremented.  The 4.1 special
1180	 * meaning of seqid 0 isn't meaningful, really, but let's avoid
1181	 * 0 anyway just for consistency and use 1:
1182	 */
1183	dp->dl_stid.sc_stateid.si_generation = 1;
1184	INIT_LIST_HEAD(&dp->dl_perfile);
1185	INIT_LIST_HEAD(&dp->dl_perclnt);
1186	INIT_LIST_HEAD(&dp->dl_recall_lru);
1187	dp->dl_clnt_odstate = odstate;
1188	get_clnt_odstate(odstate);
1189	dp->dl_type = NFS4_OPEN_DELEGATE_READ;
1190	dp->dl_retries = 1;
1191	dp->dl_recalled = false;
1192	nfsd4_init_cb(&dp->dl_recall, dp->dl_stid.sc_client,
1193		      &nfsd4_cb_recall_ops, NFSPROC4_CLNT_CB_RECALL);
1194	get_nfs4_file(fp);
1195	dp->dl_stid.sc_file = fp;
1196	return dp;
1197out_dec:
1198	atomic_long_dec(&num_delegations);
1199	return NULL;
1200}
1201
1202void
1203nfs4_put_stid(struct nfs4_stid *s)
1204{
1205	struct nfs4_file *fp = s->sc_file;
1206	struct nfs4_client *clp = s->sc_client;
1207
1208	might_lock(&clp->cl_lock);
1209
1210	if (!refcount_dec_and_lock(&s->sc_count, &clp->cl_lock)) {
1211		wake_up_all(&close_wq);
1212		return;
1213	}
1214	idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1215	nfs4_free_cpntf_statelist(clp->net, s);
1216	spin_unlock(&clp->cl_lock);
1217	s->sc_free(s);
1218	if (fp)
1219		put_nfs4_file(fp);
1220}
1221
1222void
1223nfs4_inc_and_copy_stateid(stateid_t *dst, struct nfs4_stid *stid)
1224{
1225	stateid_t *src = &stid->sc_stateid;
1226
1227	spin_lock(&stid->sc_lock);
1228	if (unlikely(++src->si_generation == 0))
1229		src->si_generation = 1;
1230	memcpy(dst, src, sizeof(*dst));
1231	spin_unlock(&stid->sc_lock);
1232}
1233
1234static void put_deleg_file(struct nfs4_file *fp)
1235{
1236	struct nfsd_file *nf = NULL;
1237
1238	spin_lock(&fp->fi_lock);
1239	if (--fp->fi_delegees == 0)
1240		swap(nf, fp->fi_deleg_file);
1241	spin_unlock(&fp->fi_lock);
1242
1243	if (nf)
1244		nfsd_file_put(nf);
1245}
1246
1247static void nfs4_unlock_deleg_lease(struct nfs4_delegation *dp)
1248{
1249	struct nfs4_file *fp = dp->dl_stid.sc_file;
1250	struct nfsd_file *nf = fp->fi_deleg_file;
1251
1252	WARN_ON_ONCE(!fp->fi_delegees);
1253
1254	vfs_setlease(nf->nf_file, F_UNLCK, NULL, (void **)&dp);
1255	put_deleg_file(fp);
1256}
1257
1258static void destroy_unhashed_deleg(struct nfs4_delegation *dp)
1259{
1260	put_clnt_odstate(dp->dl_clnt_odstate);
1261	nfs4_unlock_deleg_lease(dp);
1262	nfs4_put_stid(&dp->dl_stid);
1263}
1264
1265void nfs4_unhash_stid(struct nfs4_stid *s)
1266{
1267	s->sc_type = 0;
1268}
1269
1270/**
1271 * nfs4_delegation_exists - Discover if this delegation already exists
1272 * @clp:     a pointer to the nfs4_client we're granting a delegation to
1273 * @fp:      a pointer to the nfs4_file we're granting a delegation on
1274 *
1275 * Return:
1276 *      On success: true iff an existing delegation is found
1277 */
1278
1279static bool
1280nfs4_delegation_exists(struct nfs4_client *clp, struct nfs4_file *fp)
1281{
1282	struct nfs4_delegation *searchdp = NULL;
1283	struct nfs4_client *searchclp = NULL;
1284
1285	lockdep_assert_held(&state_lock);
1286	lockdep_assert_held(&fp->fi_lock);
1287
1288	list_for_each_entry(searchdp, &fp->fi_delegations, dl_perfile) {
1289		searchclp = searchdp->dl_stid.sc_client;
1290		if (clp == searchclp) {
1291			return true;
1292		}
1293	}
1294	return false;
1295}
1296
1297/**
1298 * hash_delegation_locked - Add a delegation to the appropriate lists
1299 * @dp:     a pointer to the nfs4_delegation we are adding.
1300 * @fp:     a pointer to the nfs4_file we're granting a delegation on
1301 *
1302 * Return:
1303 *      On success: NULL if the delegation was successfully hashed.
1304 *
1305 *      On error: -EAGAIN if one was previously granted to this
1306 *                 nfs4_client for this nfs4_file. Delegation is not hashed.
1307 *
1308 */
1309
1310static int
1311hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
1312{
1313	struct nfs4_client *clp = dp->dl_stid.sc_client;
1314
1315	lockdep_assert_held(&state_lock);
1316	lockdep_assert_held(&fp->fi_lock);
1317
1318	if (nfs4_delegation_exists(clp, fp))
1319		return -EAGAIN;
1320	refcount_inc(&dp->dl_stid.sc_count);
1321	dp->dl_stid.sc_type = NFS4_DELEG_STID;
1322	list_add(&dp->dl_perfile, &fp->fi_delegations);
1323	list_add(&dp->dl_perclnt, &clp->cl_delegations);
1324	return 0;
1325}
1326
1327static bool delegation_hashed(struct nfs4_delegation *dp)
1328{
1329	return !(list_empty(&dp->dl_perfile));
1330}
1331
1332static bool
1333unhash_delegation_locked(struct nfs4_delegation *dp)
1334{
1335	struct nfs4_file *fp = dp->dl_stid.sc_file;
1336
1337	lockdep_assert_held(&state_lock);
1338
1339	if (!delegation_hashed(dp))
1340		return false;
1341
1342	dp->dl_stid.sc_type = NFS4_CLOSED_DELEG_STID;
1343	/* Ensure that deleg break won't try to requeue it */
1344	++dp->dl_time;
1345	spin_lock(&fp->fi_lock);
1346	list_del_init(&dp->dl_perclnt);
1347	list_del_init(&dp->dl_recall_lru);
1348	list_del_init(&dp->dl_perfile);
1349	spin_unlock(&fp->fi_lock);
1350	return true;
1351}
1352
1353static void destroy_delegation(struct nfs4_delegation *dp)
1354{
1355	bool unhashed;
1356
1357	spin_lock(&state_lock);
1358	unhashed = unhash_delegation_locked(dp);
1359	spin_unlock(&state_lock);
1360	if (unhashed)
1361		destroy_unhashed_deleg(dp);
1362}
1363
1364static void revoke_delegation(struct nfs4_delegation *dp)
1365{
1366	struct nfs4_client *clp = dp->dl_stid.sc_client;
1367
1368	WARN_ON(!list_empty(&dp->dl_recall_lru));
1369
1370	trace_nfsd_stid_revoke(&dp->dl_stid);
1371
1372	if (clp->cl_minorversion) {
1373		dp->dl_stid.sc_type = NFS4_REVOKED_DELEG_STID;
1374		refcount_inc(&dp->dl_stid.sc_count);
1375		spin_lock(&clp->cl_lock);
1376		list_add(&dp->dl_recall_lru, &clp->cl_revoked);
1377		spin_unlock(&clp->cl_lock);
1378	}
1379	destroy_unhashed_deleg(dp);
1380}
1381
1382/* 
1383 * SETCLIENTID state 
1384 */
1385
1386static unsigned int clientid_hashval(u32 id)
1387{
1388	return id & CLIENT_HASH_MASK;
1389}
1390
1391static unsigned int clientstr_hashval(struct xdr_netobj name)
1392{
1393	return opaque_hashval(name.data, 8) & CLIENT_HASH_MASK;
1394}
1395
1396/*
1397 * A stateid that had a deny mode associated with it is being released
1398 * or downgraded. Recalculate the deny mode on the file.
1399 */
1400static void
1401recalculate_deny_mode(struct nfs4_file *fp)
1402{
1403	struct nfs4_ol_stateid *stp;
1404
1405	spin_lock(&fp->fi_lock);
1406	fp->fi_share_deny = 0;
1407	list_for_each_entry(stp, &fp->fi_stateids, st_perfile)
1408		fp->fi_share_deny |= bmap_to_share_mode(stp->st_deny_bmap);
1409	spin_unlock(&fp->fi_lock);
1410}
1411
1412static void
1413reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
1414{
1415	int i;
1416	bool change = false;
1417
1418	for (i = 1; i < 4; i++) {
1419		if ((i & deny) != i) {
1420			change = true;
1421			clear_deny(i, stp);
1422		}
1423	}
1424
1425	/* Recalculate per-file deny mode if there was a change */
1426	if (change)
1427		recalculate_deny_mode(stp->st_stid.sc_file);
1428}
1429
1430/* release all access and file references for a given stateid */
1431static void
1432release_all_access(struct nfs4_ol_stateid *stp)
1433{
1434	int i;
1435	struct nfs4_file *fp = stp->st_stid.sc_file;
1436
1437	if (fp && stp->st_deny_bmap != 0)
1438		recalculate_deny_mode(fp);
1439
1440	for (i = 1; i < 4; i++) {
1441		if (test_access(i, stp))
1442			nfs4_file_put_access(stp->st_stid.sc_file, i);
1443		clear_access(i, stp);
1444	}
1445}
1446
1447static inline void nfs4_free_stateowner(struct nfs4_stateowner *sop)
1448{
1449	kfree(sop->so_owner.data);
1450	sop->so_ops->so_free(sop);
1451}
1452
1453static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
1454{
1455	struct nfs4_client *clp = sop->so_client;
1456
1457	might_lock(&clp->cl_lock);
1458
1459	if (!atomic_dec_and_lock(&sop->so_count, &clp->cl_lock))
1460		return;
1461	sop->so_ops->so_unhash(sop);
1462	spin_unlock(&clp->cl_lock);
1463	nfs4_free_stateowner(sop);
1464}
1465
1466static bool
1467nfs4_ol_stateid_unhashed(const struct nfs4_ol_stateid *stp)
1468{
1469	return list_empty(&stp->st_perfile);
1470}
1471
1472static bool unhash_ol_stateid(struct nfs4_ol_stateid *stp)
1473{
1474	struct nfs4_file *fp = stp->st_stid.sc_file;
1475
1476	lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
1477
1478	if (list_empty(&stp->st_perfile))
1479		return false;
1480
1481	spin_lock(&fp->fi_lock);
1482	list_del_init(&stp->st_perfile);
1483	spin_unlock(&fp->fi_lock);
1484	list_del(&stp->st_perstateowner);
1485	return true;
1486}
1487
1488static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
1489{
1490	struct nfs4_ol_stateid *stp = openlockstateid(stid);
1491
1492	put_clnt_odstate(stp->st_clnt_odstate);
1493	release_all_access(stp);
1494	if (stp->st_stateowner)
1495		nfs4_put_stateowner(stp->st_stateowner);
1496	WARN_ON(!list_empty(&stid->sc_cp_list));
1497	kmem_cache_free(stateid_slab, stid);
1498}
1499
1500static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
1501{
1502	struct nfs4_ol_stateid *stp = openlockstateid(stid);
1503	struct nfs4_lockowner *lo = lockowner(stp->st_stateowner);
1504	struct nfsd_file *nf;
1505
1506	nf = find_any_file(stp->st_stid.sc_file);
1507	if (nf) {
1508		get_file(nf->nf_file);
1509		filp_close(nf->nf_file, (fl_owner_t)lo);
1510		nfsd_file_put(nf);
1511	}
1512	nfs4_free_ol_stateid(stid);
1513}
1514
1515/*
1516 * Put the persistent reference to an already unhashed generic stateid, while
1517 * holding the cl_lock. If it's the last reference, then put it onto the
1518 * reaplist for later destruction.
1519 */
1520static void put_ol_stateid_locked(struct nfs4_ol_stateid *stp,
1521				       struct list_head *reaplist)
1522{
1523	struct nfs4_stid *s = &stp->st_stid;
1524	struct nfs4_client *clp = s->sc_client;
1525
1526	lockdep_assert_held(&clp->cl_lock);
1527
1528	WARN_ON_ONCE(!list_empty(&stp->st_locks));
1529
1530	if (!refcount_dec_and_test(&s->sc_count)) {
1531		wake_up_all(&close_wq);
1532		return;
1533	}
1534
1535	idr_remove(&clp->cl_stateids, s->sc_stateid.si_opaque.so_id);
1536	list_add(&stp->st_locks, reaplist);
1537}
1538
1539static bool unhash_lock_stateid(struct nfs4_ol_stateid *stp)
1540{
1541	lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1542
1543	if (!unhash_ol_stateid(stp))
1544		return false;
1545	list_del_init(&stp->st_locks);
1546	nfs4_unhash_stid(&stp->st_stid);
1547	return true;
1548}
1549
1550static void release_lock_stateid(struct nfs4_ol_stateid *stp)
1551{
1552	struct nfs4_client *clp = stp->st_stid.sc_client;
1553	bool unhashed;
1554
1555	spin_lock(&clp->cl_lock);
1556	unhashed = unhash_lock_stateid(stp);
1557	spin_unlock(&clp->cl_lock);
1558	if (unhashed)
1559		nfs4_put_stid(&stp->st_stid);
1560}
1561
1562static void unhash_lockowner_locked(struct nfs4_lockowner *lo)
1563{
1564	struct nfs4_client *clp = lo->lo_owner.so_client;
1565
1566	lockdep_assert_held(&clp->cl_lock);
1567
1568	list_del_init(&lo->lo_owner.so_strhash);
1569}
1570
1571/*
1572 * Free a list of generic stateids that were collected earlier after being
1573 * fully unhashed.
1574 */
1575static void
1576free_ol_stateid_reaplist(struct list_head *reaplist)
1577{
1578	struct nfs4_ol_stateid *stp;
1579	struct nfs4_file *fp;
1580
1581	might_sleep();
1582
1583	while (!list_empty(reaplist)) {
1584		stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
1585				       st_locks);
1586		list_del(&stp->st_locks);
1587		fp = stp->st_stid.sc_file;
1588		stp->st_stid.sc_free(&stp->st_stid);
1589		if (fp)
1590			put_nfs4_file(fp);
1591	}
1592}
1593
1594static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp,
1595				       struct list_head *reaplist)
1596{
1597	struct nfs4_ol_stateid *stp;
1598
1599	lockdep_assert_held(&open_stp->st_stid.sc_client->cl_lock);
1600
1601	while (!list_empty(&open_stp->st_locks)) {
1602		stp = list_entry(open_stp->st_locks.next,
1603				struct nfs4_ol_stateid, st_locks);
1604		WARN_ON(!unhash_lock_stateid(stp));
1605		put_ol_stateid_locked(stp, reaplist);
1606	}
1607}
1608
1609static bool unhash_open_stateid(struct nfs4_ol_stateid *stp,
1610				struct list_head *reaplist)
1611{
1612	lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1613
1614	if (!unhash_ol_stateid(stp))
1615		return false;
1616	release_open_stateid_locks(stp, reaplist);
1617	return true;
1618}
1619
1620static void release_open_stateid(struct nfs4_ol_stateid *stp)
1621{
1622	LIST_HEAD(reaplist);
1623
1624	spin_lock(&stp->st_stid.sc_client->cl_lock);
1625	if (unhash_open_stateid(stp, &reaplist))
1626		put_ol_stateid_locked(stp, &reaplist);
1627	spin_unlock(&stp->st_stid.sc_client->cl_lock);
1628	free_ol_stateid_reaplist(&reaplist);
1629}
1630
1631static void unhash_openowner_locked(struct nfs4_openowner *oo)
1632{
1633	struct nfs4_client *clp = oo->oo_owner.so_client;
1634
1635	lockdep_assert_held(&clp->cl_lock);
1636
1637	list_del_init(&oo->oo_owner.so_strhash);
1638	list_del_init(&oo->oo_perclient);
1639}
1640
1641static void release_last_closed_stateid(struct nfs4_openowner *oo)
1642{
1643	struct nfsd_net *nn = net_generic(oo->oo_owner.so_client->net,
1644					  nfsd_net_id);
1645	struct nfs4_ol_stateid *s;
1646
1647	spin_lock(&nn->client_lock);
1648	s = oo->oo_last_closed_stid;
1649	if (s) {
1650		list_del_init(&oo->oo_close_lru);
1651		oo->oo_last_closed_stid = NULL;
1652	}
1653	spin_unlock(&nn->client_lock);
1654	if (s)
1655		nfs4_put_stid(&s->st_stid);
1656}
1657
1658static void release_openowner(struct nfs4_openowner *oo)
1659{
1660	struct nfs4_ol_stateid *stp;
1661	struct nfs4_client *clp = oo->oo_owner.so_client;
1662	struct list_head reaplist;
1663
1664	INIT_LIST_HEAD(&reaplist);
1665
1666	spin_lock(&clp->cl_lock);
1667	unhash_openowner_locked(oo);
1668	while (!list_empty(&oo->oo_owner.so_stateids)) {
1669		stp = list_first_entry(&oo->oo_owner.so_stateids,
1670				struct nfs4_ol_stateid, st_perstateowner);
1671		if (unhash_open_stateid(stp, &reaplist))
1672			put_ol_stateid_locked(stp, &reaplist);
1673	}
1674	spin_unlock(&clp->cl_lock);
1675	free_ol_stateid_reaplist(&reaplist);
1676	release_last_closed_stateid(oo);
1677	nfs4_put_stateowner(&oo->oo_owner);
1678}
1679
1680static inline int
1681hash_sessionid(struct nfs4_sessionid *sessionid)
1682{
1683	struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1684
1685	return sid->sequence % SESSION_HASH_SIZE;
1686}
1687
1688#ifdef CONFIG_SUNRPC_DEBUG
1689static inline void
1690dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1691{
1692	u32 *ptr = (u32 *)(&sessionid->data[0]);
1693	dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1694}
1695#else
1696static inline void
1697dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1698{
1699}
1700#endif
1701
1702/*
1703 * Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1704 * won't be used for replay.
1705 */
1706void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1707{
1708	struct nfs4_stateowner *so = cstate->replay_owner;
1709
1710	if (nfserr == nfserr_replay_me)
1711		return;
1712
1713	if (!seqid_mutating_err(ntohl(nfserr))) {
1714		nfsd4_cstate_clear_replay(cstate);
1715		return;
1716	}
1717	if (!so)
1718		return;
1719	if (so->so_is_open_owner)
1720		release_last_closed_stateid(openowner(so));
1721	so->so_seqid++;
1722	return;
1723}
1724
1725static void
1726gen_sessionid(struct nfsd4_session *ses)
1727{
1728	struct nfs4_client *clp = ses->se_client;
1729	struct nfsd4_sessionid *sid;
1730
1731	sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1732	sid->clientid = clp->cl_clientid;
1733	sid->sequence = current_sessionid++;
1734	sid->reserved = 0;
1735}
1736
1737/*
1738 * The protocol defines ca_maxresponssize_cached to include the size of
1739 * the rpc header, but all we need to cache is the data starting after
1740 * the end of the initial SEQUENCE operation--the rest we regenerate
1741 * each time.  Therefore we can advertise a ca_maxresponssize_cached
1742 * value that is the number of bytes in our cache plus a few additional
1743 * bytes.  In order to stay on the safe side, and not promise more than
1744 * we can cache, those additional bytes must be the minimum possible: 24
1745 * bytes of rpc header (xid through accept state, with AUTH_NULL
1746 * verifier), 12 for the compound header (with zero-length tag), and 44
1747 * for the SEQUENCE op response:
1748 */
1749#define NFSD_MIN_HDR_SEQ_SZ  (24 + 12 + 44)
1750
1751static void
1752free_session_slots(struct nfsd4_session *ses)
1753{
1754	int i;
1755
1756	for (i = 0; i < ses->se_fchannel.maxreqs; i++) {
1757		free_svc_cred(&ses->se_slots[i]->sl_cred);
1758		kfree(ses->se_slots[i]);
1759	}
1760}
1761
1762/*
1763 * We don't actually need to cache the rpc and session headers, so we
1764 * can allocate a little less for each slot:
1765 */
1766static inline u32 slot_bytes(struct nfsd4_channel_attrs *ca)
1767{
1768	u32 size;
1769
1770	if (ca->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ)
1771		size = 0;
1772	else
1773		size = ca->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
1774	return size + sizeof(struct nfsd4_slot);
1775}
1776
1777/*
1778 * XXX: If we run out of reserved DRC memory we could (up to a point)
1779 * re-negotiate active sessions and reduce their slot usage to make
1780 * room for new connections. For now we just fail the create session.
1781 */
1782static u32 nfsd4_get_drc_mem(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
1783{
1784	u32 slotsize = slot_bytes(ca);
1785	u32 num = ca->maxreqs;
1786	unsigned long avail, total_avail;
1787	unsigned int scale_factor;
1788
1789	spin_lock(&nfsd_drc_lock);
1790	if (nfsd_drc_max_mem > nfsd_drc_mem_used)
1791		total_avail = nfsd_drc_max_mem - nfsd_drc_mem_used;
1792	else
1793		/* We have handed out more space than we chose in
1794		 * set_max_drc() to allow.  That isn't really a
1795		 * problem as long as that doesn't make us think we
1796		 * have lots more due to integer overflow.
1797		 */
1798		total_avail = 0;
1799	avail = min((unsigned long)NFSD_MAX_MEM_PER_SESSION, total_avail);
1800	/*
1801	 * Never use more than a fraction of the remaining memory,
1802	 * unless it's the only way to give this client a slot.
1803	 * The chosen fraction is either 1/8 or 1/number of threads,
1804	 * whichever is smaller.  This ensures there are adequate
1805	 * slots to support multiple clients per thread.
1806	 * Give the client one slot even if that would require
1807	 * over-allocation--it is better than failure.
1808	 */
1809	scale_factor = max_t(unsigned int, 8, nn->nfsd_serv->sv_nrthreads);
1810
1811	avail = clamp_t(unsigned long, avail, slotsize,
1812			total_avail/scale_factor);
1813	num = min_t(int, num, avail / slotsize);
1814	num = max_t(int, num, 1);
1815	nfsd_drc_mem_used += num * slotsize;
1816	spin_unlock(&nfsd_drc_lock);
1817
1818	return num;
1819}
1820
1821static void nfsd4_put_drc_mem(struct nfsd4_channel_attrs *ca)
1822{
1823	int slotsize = slot_bytes(ca);
1824
1825	spin_lock(&nfsd_drc_lock);
1826	nfsd_drc_mem_used -= slotsize * ca->maxreqs;
1827	spin_unlock(&nfsd_drc_lock);
1828}
1829
1830static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
1831					   struct nfsd4_channel_attrs *battrs)
1832{
1833	int numslots = fattrs->maxreqs;
1834	int slotsize = slot_bytes(fattrs);
1835	struct nfsd4_session *new;
1836	int i;
1837
1838	BUILD_BUG_ON(struct_size(new, se_slots, NFSD_MAX_SLOTS_PER_SESSION)
1839		     > PAGE_SIZE);
 
1840
1841	new = kzalloc(struct_size(new, se_slots, numslots), GFP_KERNEL);
1842	if (!new)
1843		return NULL;
1844	/* allocate each struct nfsd4_slot and data cache in one piece */
1845	for (i = 0; i < numslots; i++) {
1846		new->se_slots[i] = kzalloc(slotsize, GFP_KERNEL);
1847		if (!new->se_slots[i])
1848			goto out_free;
1849	}
1850
1851	memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
1852	memcpy(&new->se_bchannel, battrs, sizeof(struct nfsd4_channel_attrs));
1853
1854	return new;
1855out_free:
1856	while (i--)
1857		kfree(new->se_slots[i]);
1858	kfree(new);
1859	return NULL;
1860}
1861
1862static void free_conn(struct nfsd4_conn *c)
1863{
1864	svc_xprt_put(c->cn_xprt);
1865	kfree(c);
1866}
1867
1868static void nfsd4_conn_lost(struct svc_xpt_user *u)
1869{
1870	struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
1871	struct nfs4_client *clp = c->cn_session->se_client;
1872
1873	trace_nfsd_cb_lost(clp);
1874
1875	spin_lock(&clp->cl_lock);
1876	if (!list_empty(&c->cn_persession)) {
1877		list_del(&c->cn_persession);
1878		free_conn(c);
1879	}
1880	nfsd4_probe_callback(clp);
1881	spin_unlock(&clp->cl_lock);
1882}
1883
1884static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
1885{
1886	struct nfsd4_conn *conn;
1887
1888	conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
1889	if (!conn)
1890		return NULL;
1891	svc_xprt_get(rqstp->rq_xprt);
1892	conn->cn_xprt = rqstp->rq_xprt;
1893	conn->cn_flags = flags;
1894	INIT_LIST_HEAD(&conn->cn_xpt_user.list);
1895	return conn;
1896}
1897
1898static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1899{
1900	conn->cn_session = ses;
1901	list_add(&conn->cn_persession, &ses->se_conns);
1902}
1903
1904static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
1905{
1906	struct nfs4_client *clp = ses->se_client;
1907
1908	spin_lock(&clp->cl_lock);
1909	__nfsd4_hash_conn(conn, ses);
1910	spin_unlock(&clp->cl_lock);
1911}
1912
1913static int nfsd4_register_conn(struct nfsd4_conn *conn)
1914{
1915	conn->cn_xpt_user.callback = nfsd4_conn_lost;
1916	return register_xpt_user(conn->cn_xprt, &conn->cn_xpt_user);
1917}
1918
1919static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
1920{
1921	int ret;
1922
1923	nfsd4_hash_conn(conn, ses);
1924	ret = nfsd4_register_conn(conn);
1925	if (ret)
1926		/* oops; xprt is already down: */
1927		nfsd4_conn_lost(&conn->cn_xpt_user);
1928	/* We may have gained or lost a callback channel: */
1929	nfsd4_probe_callback_sync(ses->se_client);
1930}
1931
1932static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
1933{
1934	u32 dir = NFS4_CDFC4_FORE;
1935
1936	if (cses->flags & SESSION4_BACK_CHAN)
1937		dir |= NFS4_CDFC4_BACK;
1938	return alloc_conn(rqstp, dir);
1939}
1940
1941/* must be called under client_lock */
1942static void nfsd4_del_conns(struct nfsd4_session *s)
1943{
1944	struct nfs4_client *clp = s->se_client;
1945	struct nfsd4_conn *c;
1946
1947	spin_lock(&clp->cl_lock);
1948	while (!list_empty(&s->se_conns)) {
1949		c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
1950		list_del_init(&c->cn_persession);
1951		spin_unlock(&clp->cl_lock);
1952
1953		unregister_xpt_user(c->cn_xprt, &c->cn_xpt_user);
1954		free_conn(c);
1955
1956		spin_lock(&clp->cl_lock);
1957	}
1958	spin_unlock(&clp->cl_lock);
1959}
1960
1961static void __free_session(struct nfsd4_session *ses)
1962{
1963	free_session_slots(ses);
1964	kfree(ses);
1965}
1966
1967static void free_session(struct nfsd4_session *ses)
1968{
1969	nfsd4_del_conns(ses);
1970	nfsd4_put_drc_mem(&ses->se_fchannel);
1971	__free_session(ses);
1972}
1973
1974static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
1975{
1976	int idx;
1977	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
1978
1979	new->se_client = clp;
1980	gen_sessionid(new);
1981
1982	INIT_LIST_HEAD(&new->se_conns);
1983
1984	new->se_cb_seq_nr = 1;
1985	new->se_flags = cses->flags;
1986	new->se_cb_prog = cses->callback_prog;
1987	new->se_cb_sec = cses->cb_sec;
1988	atomic_set(&new->se_ref, 0);
1989	idx = hash_sessionid(&new->se_sessionid);
1990	list_add(&new->se_hash, &nn->sessionid_hashtbl[idx]);
1991	spin_lock(&clp->cl_lock);
1992	list_add(&new->se_perclnt, &clp->cl_sessions);
1993	spin_unlock(&clp->cl_lock);
1994
1995	{
1996		struct sockaddr *sa = svc_addr(rqstp);
1997		/*
1998		 * This is a little silly; with sessions there's no real
1999		 * use for the callback address.  Use the peer address
2000		 * as a reasonable default for now, but consider fixing
2001		 * the rpc client not to require an address in the
2002		 * future:
2003		 */
2004		rpc_copy_addr((struct sockaddr *)&clp->cl_cb_conn.cb_addr, sa);
2005		clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
2006	}
2007}
2008
2009/* caller must hold client_lock */
2010static struct nfsd4_session *
2011__find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
2012{
2013	struct nfsd4_session *elem;
2014	int idx;
2015	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2016
2017	lockdep_assert_held(&nn->client_lock);
2018
2019	dump_sessionid(__func__, sessionid);
2020	idx = hash_sessionid(sessionid);
2021	/* Search in the appropriate list */
2022	list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
2023		if (!memcmp(elem->se_sessionid.data, sessionid->data,
2024			    NFS4_MAX_SESSIONID_LEN)) {
2025			return elem;
2026		}
2027	}
2028
2029	dprintk("%s: session not found\n", __func__);
2030	return NULL;
2031}
2032
2033static struct nfsd4_session *
2034find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
2035		__be32 *ret)
2036{
2037	struct nfsd4_session *session;
2038	__be32 status = nfserr_badsession;
2039
2040	session = __find_in_sessionid_hashtbl(sessionid, net);
2041	if (!session)
2042		goto out;
2043	status = nfsd4_get_session_locked(session);
2044	if (status)
2045		session = NULL;
2046out:
2047	*ret = status;
2048	return session;
2049}
2050
2051/* caller must hold client_lock */
2052static void
2053unhash_session(struct nfsd4_session *ses)
2054{
2055	struct nfs4_client *clp = ses->se_client;
2056	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2057
2058	lockdep_assert_held(&nn->client_lock);
2059
2060	list_del(&ses->se_hash);
2061	spin_lock(&ses->se_client->cl_lock);
2062	list_del(&ses->se_perclnt);
2063	spin_unlock(&ses->se_client->cl_lock);
2064}
2065
2066/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
2067static int
2068STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
2069{
2070	/*
2071	 * We're assuming the clid was not given out from a boot
2072	 * precisely 2^32 (about 136 years) before this one.  That seems
2073	 * a safe assumption:
2074	 */
2075	if (clid->cl_boot == (u32)nn->boot_time)
2076		return 0;
2077	trace_nfsd_clid_stale(clid);
2078	return 1;
2079}
2080
2081/* 
2082 * XXX Should we use a slab cache ?
2083 * This type of memory management is somewhat inefficient, but we use it
2084 * anyway since SETCLIENTID is not a common operation.
2085 */
2086static struct nfs4_client *alloc_client(struct xdr_netobj name,
2087				struct nfsd_net *nn)
2088{
2089	struct nfs4_client *clp;
2090	int i;
2091
2092	if (atomic_read(&nn->nfs4_client_count) >= nn->nfs4_max_clients) {
2093		mod_delayed_work(laundry_wq, &nn->laundromat_work, 0);
2094		return NULL;
2095	}
2096	clp = kmem_cache_zalloc(client_slab, GFP_KERNEL);
2097	if (clp == NULL)
2098		return NULL;
2099	xdr_netobj_dup(&clp->cl_name, &name, GFP_KERNEL);
2100	if (clp->cl_name.data == NULL)
2101		goto err_no_name;
2102	clp->cl_ownerstr_hashtbl = kmalloc_array(OWNER_HASH_SIZE,
2103						 sizeof(struct list_head),
2104						 GFP_KERNEL);
2105	if (!clp->cl_ownerstr_hashtbl)
2106		goto err_no_hashtbl;
2107	for (i = 0; i < OWNER_HASH_SIZE; i++)
2108		INIT_LIST_HEAD(&clp->cl_ownerstr_hashtbl[i]);
2109	INIT_LIST_HEAD(&clp->cl_sessions);
2110	idr_init(&clp->cl_stateids);
2111	atomic_set(&clp->cl_rpc_users, 0);
2112	clp->cl_cb_state = NFSD4_CB_UNKNOWN;
2113	clp->cl_state = NFSD4_ACTIVE;
2114	atomic_inc(&nn->nfs4_client_count);
2115	atomic_set(&clp->cl_delegs_in_recall, 0);
2116	INIT_LIST_HEAD(&clp->cl_idhash);
2117	INIT_LIST_HEAD(&clp->cl_openowners);
2118	INIT_LIST_HEAD(&clp->cl_delegations);
2119	INIT_LIST_HEAD(&clp->cl_lru);
2120	INIT_LIST_HEAD(&clp->cl_revoked);
2121#ifdef CONFIG_NFSD_PNFS
2122	INIT_LIST_HEAD(&clp->cl_lo_states);
2123#endif
2124	INIT_LIST_HEAD(&clp->async_copies);
2125	spin_lock_init(&clp->async_lock);
2126	spin_lock_init(&clp->cl_lock);
2127	rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
2128	return clp;
2129err_no_hashtbl:
2130	kfree(clp->cl_name.data);
2131err_no_name:
2132	kmem_cache_free(client_slab, clp);
2133	return NULL;
2134}
2135
2136static void __free_client(struct kref *k)
2137{
2138	struct nfsdfs_client *c = container_of(k, struct nfsdfs_client, cl_ref);
2139	struct nfs4_client *clp = container_of(c, struct nfs4_client, cl_nfsdfs);
2140
2141	free_svc_cred(&clp->cl_cred);
2142	kfree(clp->cl_ownerstr_hashtbl);
2143	kfree(clp->cl_name.data);
2144	kfree(clp->cl_nii_domain.data);
2145	kfree(clp->cl_nii_name.data);
2146	idr_destroy(&clp->cl_stateids);
2147	kfree(clp->cl_ra);
2148	kmem_cache_free(client_slab, clp);
2149}
2150
2151static void drop_client(struct nfs4_client *clp)
2152{
2153	kref_put(&clp->cl_nfsdfs.cl_ref, __free_client);
2154}
2155
2156static void
2157free_client(struct nfs4_client *clp)
2158{
2159	while (!list_empty(&clp->cl_sessions)) {
2160		struct nfsd4_session *ses;
2161		ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
2162				se_perclnt);
2163		list_del(&ses->se_perclnt);
2164		WARN_ON_ONCE(atomic_read(&ses->se_ref));
2165		free_session(ses);
2166	}
2167	rpc_destroy_wait_queue(&clp->cl_cb_waitq);
2168	if (clp->cl_nfsd_dentry) {
2169		nfsd_client_rmdir(clp->cl_nfsd_dentry);
2170		clp->cl_nfsd_dentry = NULL;
2171		wake_up_all(&expiry_wq);
2172	}
2173	drop_client(clp);
2174}
2175
2176/* must be called under the client_lock */
2177static void
2178unhash_client_locked(struct nfs4_client *clp)
2179{
2180	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2181	struct nfsd4_session *ses;
2182
2183	lockdep_assert_held(&nn->client_lock);
2184
2185	/* Mark the client as expired! */
2186	clp->cl_time = 0;
2187	/* Make it invisible */
2188	if (!list_empty(&clp->cl_idhash)) {
2189		list_del_init(&clp->cl_idhash);
2190		if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
2191			rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
2192		else
2193			rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2194	}
2195	list_del_init(&clp->cl_lru);
2196	spin_lock(&clp->cl_lock);
2197	list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
2198		list_del_init(&ses->se_hash);
2199	spin_unlock(&clp->cl_lock);
2200}
2201
2202static void
2203unhash_client(struct nfs4_client *clp)
2204{
2205	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2206
2207	spin_lock(&nn->client_lock);
2208	unhash_client_locked(clp);
2209	spin_unlock(&nn->client_lock);
2210}
2211
2212static __be32 mark_client_expired_locked(struct nfs4_client *clp)
2213{
2214	if (atomic_read(&clp->cl_rpc_users))
2215		return nfserr_jukebox;
2216	unhash_client_locked(clp);
2217	return nfs_ok;
2218}
2219
2220static void
2221__destroy_client(struct nfs4_client *clp)
2222{
2223	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2224	int i;
2225	struct nfs4_openowner *oo;
2226	struct nfs4_delegation *dp;
2227	struct list_head reaplist;
2228
2229	INIT_LIST_HEAD(&reaplist);
2230	spin_lock(&state_lock);
2231	while (!list_empty(&clp->cl_delegations)) {
2232		dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
2233		WARN_ON(!unhash_delegation_locked(dp));
2234		list_add(&dp->dl_recall_lru, &reaplist);
2235	}
2236	spin_unlock(&state_lock);
2237	while (!list_empty(&reaplist)) {
2238		dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
2239		list_del_init(&dp->dl_recall_lru);
2240		destroy_unhashed_deleg(dp);
2241	}
2242	while (!list_empty(&clp->cl_revoked)) {
2243		dp = list_entry(clp->cl_revoked.next, struct nfs4_delegation, dl_recall_lru);
2244		list_del_init(&dp->dl_recall_lru);
2245		nfs4_put_stid(&dp->dl_stid);
2246	}
2247	while (!list_empty(&clp->cl_openowners)) {
2248		oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
2249		nfs4_get_stateowner(&oo->oo_owner);
2250		release_openowner(oo);
2251	}
2252	for (i = 0; i < OWNER_HASH_SIZE; i++) {
2253		struct nfs4_stateowner *so, *tmp;
2254
2255		list_for_each_entry_safe(so, tmp, &clp->cl_ownerstr_hashtbl[i],
2256					 so_strhash) {
2257			/* Should be no openowners at this point */
2258			WARN_ON_ONCE(so->so_is_open_owner);
2259			remove_blocked_locks(lockowner(so));
2260		}
2261	}
2262	nfsd4_return_all_client_layouts(clp);
2263	nfsd4_shutdown_copy(clp);
2264	nfsd4_shutdown_callback(clp);
2265	if (clp->cl_cb_conn.cb_xprt)
2266		svc_xprt_put(clp->cl_cb_conn.cb_xprt);
2267	atomic_add_unless(&nn->nfs4_client_count, -1, 0);
2268	nfsd4_dec_courtesy_client_count(nn, clp);
2269	free_client(clp);
2270	wake_up_all(&expiry_wq);
2271}
2272
2273static void
2274destroy_client(struct nfs4_client *clp)
2275{
2276	unhash_client(clp);
2277	__destroy_client(clp);
2278}
2279
2280static void inc_reclaim_complete(struct nfs4_client *clp)
2281{
2282	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2283
2284	if (!nn->track_reclaim_completes)
2285		return;
2286	if (!nfsd4_find_reclaim_client(clp->cl_name, nn))
2287		return;
2288	if (atomic_inc_return(&nn->nr_reclaim_complete) ==
2289			nn->reclaim_str_hashtbl_size) {
2290		printk(KERN_INFO "NFSD: all clients done reclaiming, ending NFSv4 grace period (net %x)\n",
2291				clp->net->ns.inum);
2292		nfsd4_end_grace(nn);
2293	}
2294}
2295
2296static void expire_client(struct nfs4_client *clp)
2297{
2298	unhash_client(clp);
2299	nfsd4_client_record_remove(clp);
2300	__destroy_client(clp);
2301}
2302
2303static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
2304{
2305	memcpy(target->cl_verifier.data, source->data,
2306			sizeof(target->cl_verifier.data));
2307}
2308
2309static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
2310{
2311	target->cl_clientid.cl_boot = source->cl_clientid.cl_boot; 
2312	target->cl_clientid.cl_id = source->cl_clientid.cl_id; 
2313}
2314
2315static int copy_cred(struct svc_cred *target, struct svc_cred *source)
2316{
2317	target->cr_principal = kstrdup(source->cr_principal, GFP_KERNEL);
2318	target->cr_raw_principal = kstrdup(source->cr_raw_principal,
2319								GFP_KERNEL);
2320	target->cr_targ_princ = kstrdup(source->cr_targ_princ, GFP_KERNEL);
2321	if ((source->cr_principal && !target->cr_principal) ||
2322	    (source->cr_raw_principal && !target->cr_raw_principal) ||
2323	    (source->cr_targ_princ && !target->cr_targ_princ))
2324		return -ENOMEM;
2325
2326	target->cr_flavor = source->cr_flavor;
2327	target->cr_uid = source->cr_uid;
2328	target->cr_gid = source->cr_gid;
2329	target->cr_group_info = source->cr_group_info;
2330	get_group_info(target->cr_group_info);
2331	target->cr_gss_mech = source->cr_gss_mech;
2332	if (source->cr_gss_mech)
2333		gss_mech_get(source->cr_gss_mech);
2334	return 0;
2335}
2336
2337static int
2338compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
2339{
2340	if (o1->len < o2->len)
2341		return -1;
2342	if (o1->len > o2->len)
2343		return 1;
2344	return memcmp(o1->data, o2->data, o1->len);
2345}
2346
2347static int
2348same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
2349{
2350	return 0 == memcmp(v1->data, v2->data, sizeof(v1->data));
2351}
2352
2353static int
2354same_clid(clientid_t *cl1, clientid_t *cl2)
2355{
2356	return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
2357}
2358
2359static bool groups_equal(struct group_info *g1, struct group_info *g2)
2360{
2361	int i;
2362
2363	if (g1->ngroups != g2->ngroups)
2364		return false;
2365	for (i=0; i<g1->ngroups; i++)
2366		if (!gid_eq(g1->gid[i], g2->gid[i]))
2367			return false;
2368	return true;
2369}
2370
2371/*
2372 * RFC 3530 language requires clid_inuse be returned when the
2373 * "principal" associated with a requests differs from that previously
2374 * used.  We use uid, gid's, and gss principal string as our best
2375 * approximation.  We also don't want to allow non-gss use of a client
2376 * established using gss: in theory cr_principal should catch that
2377 * change, but in practice cr_principal can be null even in the gss case
2378 * since gssd doesn't always pass down a principal string.
2379 */
2380static bool is_gss_cred(struct svc_cred *cr)
2381{
2382	/* Is cr_flavor one of the gss "pseudoflavors"?: */
2383	return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
2384}
2385
2386
2387static bool
2388same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
2389{
2390	if ((is_gss_cred(cr1) != is_gss_cred(cr2))
2391		|| (!uid_eq(cr1->cr_uid, cr2->cr_uid))
2392		|| (!gid_eq(cr1->cr_gid, cr2->cr_gid))
2393		|| !groups_equal(cr1->cr_group_info, cr2->cr_group_info))
2394		return false;
2395	/* XXX: check that cr_targ_princ fields match ? */
2396	if (cr1->cr_principal == cr2->cr_principal)
2397		return true;
2398	if (!cr1->cr_principal || !cr2->cr_principal)
2399		return false;
2400	return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
2401}
2402
2403static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
2404{
2405	struct svc_cred *cr = &rqstp->rq_cred;
2406	u32 service;
2407
2408	if (!cr->cr_gss_mech)
2409		return false;
2410	service = gss_pseudoflavor_to_service(cr->cr_gss_mech, cr->cr_flavor);
2411	return service == RPC_GSS_SVC_INTEGRITY ||
2412	       service == RPC_GSS_SVC_PRIVACY;
2413}
2414
2415bool nfsd4_mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
2416{
2417	struct svc_cred *cr = &rqstp->rq_cred;
2418
2419	if (!cl->cl_mach_cred)
2420		return true;
2421	if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
2422		return false;
2423	if (!svc_rqst_integrity_protected(rqstp))
2424		return false;
2425	if (cl->cl_cred.cr_raw_principal)
2426		return 0 == strcmp(cl->cl_cred.cr_raw_principal,
2427						cr->cr_raw_principal);
2428	if (!cr->cr_principal)
2429		return false;
2430	return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
2431}
2432
2433static void gen_confirm(struct nfs4_client *clp, struct nfsd_net *nn)
2434{
2435	__be32 verf[2];
2436
2437	/*
2438	 * This is opaque to client, so no need to byte-swap. Use
2439	 * __force to keep sparse happy
2440	 */
2441	verf[0] = (__force __be32)(u32)ktime_get_real_seconds();
2442	verf[1] = (__force __be32)nn->clverifier_counter++;
2443	memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
2444}
2445
2446static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
2447{
2448	clp->cl_clientid.cl_boot = (u32)nn->boot_time;
2449	clp->cl_clientid.cl_id = nn->clientid_counter++;
2450	gen_confirm(clp, nn);
2451}
2452
2453static struct nfs4_stid *
2454find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
2455{
2456	struct nfs4_stid *ret;
2457
2458	ret = idr_find(&cl->cl_stateids, t->si_opaque.so_id);
2459	if (!ret || !ret->sc_type)
2460		return NULL;
2461	return ret;
2462}
2463
2464static struct nfs4_stid *
2465find_stateid_by_type(struct nfs4_client *cl, stateid_t *t, char typemask)
2466{
2467	struct nfs4_stid *s;
2468
2469	spin_lock(&cl->cl_lock);
2470	s = find_stateid_locked(cl, t);
2471	if (s != NULL) {
2472		if (typemask & s->sc_type)
2473			refcount_inc(&s->sc_count);
2474		else
2475			s = NULL;
2476	}
2477	spin_unlock(&cl->cl_lock);
2478	return s;
2479}
2480
2481static struct nfs4_client *get_nfsdfs_clp(struct inode *inode)
2482{
2483	struct nfsdfs_client *nc;
2484	nc = get_nfsdfs_client(inode);
2485	if (!nc)
2486		return NULL;
2487	return container_of(nc, struct nfs4_client, cl_nfsdfs);
2488}
2489
2490static void seq_quote_mem(struct seq_file *m, char *data, int len)
2491{
2492	seq_printf(m, "\"");
2493	seq_escape_mem(m, data, len, ESCAPE_HEX | ESCAPE_NAP | ESCAPE_APPEND, "\"\\");
2494	seq_printf(m, "\"");
2495}
2496
2497static const char *cb_state2str(int state)
2498{
2499	switch (state) {
2500	case NFSD4_CB_UP:
2501		return "UP";
2502	case NFSD4_CB_UNKNOWN:
2503		return "UNKNOWN";
2504	case NFSD4_CB_DOWN:
2505		return "DOWN";
2506	case NFSD4_CB_FAULT:
2507		return "FAULT";
2508	}
2509	return "UNDEFINED";
2510}
2511
2512static int client_info_show(struct seq_file *m, void *v)
2513{
2514	struct inode *inode = file_inode(m->file);
2515	struct nfs4_client *clp;
2516	u64 clid;
2517
2518	clp = get_nfsdfs_clp(inode);
2519	if (!clp)
2520		return -ENXIO;
2521	memcpy(&clid, &clp->cl_clientid, sizeof(clid));
2522	seq_printf(m, "clientid: 0x%llx\n", clid);
2523	seq_printf(m, "address: \"%pISpc\"\n", (struct sockaddr *)&clp->cl_addr);
2524
2525	if (clp->cl_state == NFSD4_COURTESY)
2526		seq_puts(m, "status: courtesy\n");
2527	else if (clp->cl_state == NFSD4_EXPIRABLE)
2528		seq_puts(m, "status: expirable\n");
2529	else if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
2530		seq_puts(m, "status: confirmed\n");
2531	else
2532		seq_puts(m, "status: unconfirmed\n");
2533	seq_printf(m, "seconds from last renew: %lld\n",
2534		ktime_get_boottime_seconds() - clp->cl_time);
2535	seq_printf(m, "name: ");
2536	seq_quote_mem(m, clp->cl_name.data, clp->cl_name.len);
2537	seq_printf(m, "\nminor version: %d\n", clp->cl_minorversion);
2538	if (clp->cl_nii_domain.data) {
2539		seq_printf(m, "Implementation domain: ");
2540		seq_quote_mem(m, clp->cl_nii_domain.data,
2541					clp->cl_nii_domain.len);
2542		seq_printf(m, "\nImplementation name: ");
2543		seq_quote_mem(m, clp->cl_nii_name.data, clp->cl_nii_name.len);
2544		seq_printf(m, "\nImplementation time: [%lld, %ld]\n",
2545			clp->cl_nii_time.tv_sec, clp->cl_nii_time.tv_nsec);
2546	}
2547	seq_printf(m, "callback state: %s\n", cb_state2str(clp->cl_cb_state));
2548	seq_printf(m, "callback address: %pISpc\n", &clp->cl_cb_conn.cb_addr);
2549	drop_client(clp);
2550
2551	return 0;
2552}
2553
2554DEFINE_SHOW_ATTRIBUTE(client_info);
 
 
 
 
 
 
 
 
 
 
2555
2556static void *states_start(struct seq_file *s, loff_t *pos)
2557	__acquires(&clp->cl_lock)
2558{
2559	struct nfs4_client *clp = s->private;
2560	unsigned long id = *pos;
2561	void *ret;
2562
2563	spin_lock(&clp->cl_lock);
2564	ret = idr_get_next_ul(&clp->cl_stateids, &id);
2565	*pos = id;
2566	return ret;
2567}
2568
2569static void *states_next(struct seq_file *s, void *v, loff_t *pos)
2570{
2571	struct nfs4_client *clp = s->private;
2572	unsigned long id = *pos;
2573	void *ret;
2574
2575	id = *pos;
2576	id++;
2577	ret = idr_get_next_ul(&clp->cl_stateids, &id);
2578	*pos = id;
2579	return ret;
2580}
2581
2582static void states_stop(struct seq_file *s, void *v)
2583	__releases(&clp->cl_lock)
2584{
2585	struct nfs4_client *clp = s->private;
2586
2587	spin_unlock(&clp->cl_lock);
2588}
2589
2590static void nfs4_show_fname(struct seq_file *s, struct nfsd_file *f)
2591{
2592         seq_printf(s, "filename: \"%pD2\"", f->nf_file);
2593}
2594
2595static void nfs4_show_superblock(struct seq_file *s, struct nfsd_file *f)
2596{
2597	struct inode *inode = file_inode(f->nf_file);
2598
2599	seq_printf(s, "superblock: \"%02x:%02x:%ld\"",
2600					MAJOR(inode->i_sb->s_dev),
2601					 MINOR(inode->i_sb->s_dev),
2602					 inode->i_ino);
2603}
2604
2605static void nfs4_show_owner(struct seq_file *s, struct nfs4_stateowner *oo)
2606{
2607	seq_printf(s, "owner: ");
2608	seq_quote_mem(s, oo->so_owner.data, oo->so_owner.len);
2609}
2610
2611static void nfs4_show_stateid(struct seq_file *s, stateid_t *stid)
2612{
2613	seq_printf(s, "0x%.8x", stid->si_generation);
2614	seq_printf(s, "%12phN", &stid->si_opaque);
2615}
2616
2617static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
2618{
2619	struct nfs4_ol_stateid *ols;
2620	struct nfs4_file *nf;
2621	struct nfsd_file *file;
2622	struct nfs4_stateowner *oo;
2623	unsigned int access, deny;
2624
2625	if (st->sc_type != NFS4_OPEN_STID && st->sc_type != NFS4_LOCK_STID)
2626		return 0; /* XXX: or SEQ_SKIP? */
2627	ols = openlockstateid(st);
2628	oo = ols->st_stateowner;
2629	nf = st->sc_file;
2630
2631	spin_lock(&nf->fi_lock);
2632	file = find_any_file_locked(nf);
2633	if (!file)
2634		goto out;
2635
2636	seq_printf(s, "- ");
2637	nfs4_show_stateid(s, &st->sc_stateid);
2638	seq_printf(s, ": { type: open, ");
2639
2640	access = bmap_to_share_mode(ols->st_access_bmap);
2641	deny   = bmap_to_share_mode(ols->st_deny_bmap);
2642
2643	seq_printf(s, "access: %s%s, ",
2644		access & NFS4_SHARE_ACCESS_READ ? "r" : "-",
2645		access & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
2646	seq_printf(s, "deny: %s%s, ",
2647		deny & NFS4_SHARE_ACCESS_READ ? "r" : "-",
2648		deny & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
2649
2650	nfs4_show_superblock(s, file);
2651	seq_printf(s, ", ");
2652	nfs4_show_fname(s, file);
2653	seq_printf(s, ", ");
2654	nfs4_show_owner(s, oo);
2655	seq_printf(s, " }\n");
2656out:
2657	spin_unlock(&nf->fi_lock);
2658	return 0;
2659}
2660
2661static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
2662{
2663	struct nfs4_ol_stateid *ols;
2664	struct nfs4_file *nf;
2665	struct nfsd_file *file;
2666	struct nfs4_stateowner *oo;
2667
2668	ols = openlockstateid(st);
2669	oo = ols->st_stateowner;
2670	nf = st->sc_file;
2671	spin_lock(&nf->fi_lock);
2672	file = find_any_file_locked(nf);
2673	if (!file)
2674		goto out;
2675
2676	seq_printf(s, "- ");
2677	nfs4_show_stateid(s, &st->sc_stateid);
2678	seq_printf(s, ": { type: lock, ");
2679
2680	/*
2681	 * Note: a lock stateid isn't really the same thing as a lock,
2682	 * it's the locking state held by one owner on a file, and there
2683	 * may be multiple (or no) lock ranges associated with it.
2684	 * (Same for the matter is true of open stateids.)
2685	 */
2686
2687	nfs4_show_superblock(s, file);
2688	/* XXX: open stateid? */
2689	seq_printf(s, ", ");
2690	nfs4_show_fname(s, file);
2691	seq_printf(s, ", ");
2692	nfs4_show_owner(s, oo);
2693	seq_printf(s, " }\n");
2694out:
2695	spin_unlock(&nf->fi_lock);
2696	return 0;
2697}
2698
2699static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
2700{
2701	struct nfs4_delegation *ds;
2702	struct nfs4_file *nf;
2703	struct nfsd_file *file;
2704
2705	ds = delegstateid(st);
2706	nf = st->sc_file;
2707	spin_lock(&nf->fi_lock);
2708	file = find_deleg_file_locked(nf);
2709	if (!file)
2710		goto out;
2711
2712	seq_printf(s, "- ");
2713	nfs4_show_stateid(s, &st->sc_stateid);
2714	seq_printf(s, ": { type: deleg, ");
2715
2716	/* Kinda dead code as long as we only support read delegs: */
2717	seq_printf(s, "access: %s, ",
2718		ds->dl_type == NFS4_OPEN_DELEGATE_READ ? "r" : "w");
2719
2720	/* XXX: lease time, whether it's being recalled. */
2721
2722	nfs4_show_superblock(s, file);
2723	seq_printf(s, ", ");
2724	nfs4_show_fname(s, file);
2725	seq_printf(s, " }\n");
2726out:
2727	spin_unlock(&nf->fi_lock);
2728	return 0;
2729}
2730
2731static int nfs4_show_layout(struct seq_file *s, struct nfs4_stid *st)
2732{
2733	struct nfs4_layout_stateid *ls;
2734	struct nfsd_file *file;
2735
2736	ls = container_of(st, struct nfs4_layout_stateid, ls_stid);
2737	file = ls->ls_file;
2738
2739	seq_printf(s, "- ");
2740	nfs4_show_stateid(s, &st->sc_stateid);
2741	seq_printf(s, ": { type: layout, ");
2742
2743	/* XXX: What else would be useful? */
2744
2745	nfs4_show_superblock(s, file);
2746	seq_printf(s, ", ");
2747	nfs4_show_fname(s, file);
2748	seq_printf(s, " }\n");
2749
2750	return 0;
2751}
2752
2753static int states_show(struct seq_file *s, void *v)
2754{
2755	struct nfs4_stid *st = v;
2756
2757	switch (st->sc_type) {
2758	case NFS4_OPEN_STID:
2759		return nfs4_show_open(s, st);
2760	case NFS4_LOCK_STID:
2761		return nfs4_show_lock(s, st);
2762	case NFS4_DELEG_STID:
2763		return nfs4_show_deleg(s, st);
2764	case NFS4_LAYOUT_STID:
2765		return nfs4_show_layout(s, st);
2766	default:
2767		return 0; /* XXX: or SEQ_SKIP? */
2768	}
2769	/* XXX: copy stateids? */
2770}
2771
2772static struct seq_operations states_seq_ops = {
2773	.start = states_start,
2774	.next = states_next,
2775	.stop = states_stop,
2776	.show = states_show
2777};
2778
2779static int client_states_open(struct inode *inode, struct file *file)
2780{
2781	struct seq_file *s;
2782	struct nfs4_client *clp;
2783	int ret;
2784
2785	clp = get_nfsdfs_clp(inode);
2786	if (!clp)
2787		return -ENXIO;
2788
2789	ret = seq_open(file, &states_seq_ops);
2790	if (ret)
2791		return ret;
2792	s = file->private_data;
2793	s->private = clp;
2794	return 0;
2795}
2796
2797static int client_opens_release(struct inode *inode, struct file *file)
2798{
2799	struct seq_file *m = file->private_data;
2800	struct nfs4_client *clp = m->private;
2801
2802	/* XXX: alternatively, we could get/drop in seq start/stop */
2803	drop_client(clp);
2804	return 0;
2805}
2806
2807static const struct file_operations client_states_fops = {
2808	.open		= client_states_open,
2809	.read		= seq_read,
2810	.llseek		= seq_lseek,
2811	.release	= client_opens_release,
2812};
2813
2814/*
2815 * Normally we refuse to destroy clients that are in use, but here the
2816 * administrator is telling us to just do it.  We also want to wait
2817 * so the caller has a guarantee that the client's locks are gone by
2818 * the time the write returns:
2819 */
2820static void force_expire_client(struct nfs4_client *clp)
2821{
2822	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2823	bool already_expired;
2824
2825	trace_nfsd_clid_admin_expired(&clp->cl_clientid);
2826
2827	spin_lock(&nn->client_lock);
2828	clp->cl_time = 0;
2829	spin_unlock(&nn->client_lock);
2830
2831	wait_event(expiry_wq, atomic_read(&clp->cl_rpc_users) == 0);
2832	spin_lock(&nn->client_lock);
2833	already_expired = list_empty(&clp->cl_lru);
2834	if (!already_expired)
2835		unhash_client_locked(clp);
2836	spin_unlock(&nn->client_lock);
2837
2838	if (!already_expired)
2839		expire_client(clp);
2840	else
2841		wait_event(expiry_wq, clp->cl_nfsd_dentry == NULL);
2842}
2843
2844static ssize_t client_ctl_write(struct file *file, const char __user *buf,
2845				   size_t size, loff_t *pos)
2846{
2847	char *data;
2848	struct nfs4_client *clp;
2849
2850	data = simple_transaction_get(file, buf, size);
2851	if (IS_ERR(data))
2852		return PTR_ERR(data);
2853	if (size != 7 || 0 != memcmp(data, "expire\n", 7))
2854		return -EINVAL;
2855	clp = get_nfsdfs_clp(file_inode(file));
2856	if (!clp)
2857		return -ENXIO;
2858	force_expire_client(clp);
2859	drop_client(clp);
2860	return 7;
2861}
2862
2863static const struct file_operations client_ctl_fops = {
2864	.write		= client_ctl_write,
2865	.release	= simple_transaction_release,
2866};
2867
2868static const struct tree_descr client_files[] = {
2869	[0] = {"info", &client_info_fops, S_IRUSR},
2870	[1] = {"states", &client_states_fops, S_IRUSR},
2871	[2] = {"ctl", &client_ctl_fops, S_IWUSR},
2872	[3] = {""},
2873};
2874
2875static int
2876nfsd4_cb_recall_any_done(struct nfsd4_callback *cb,
2877				struct rpc_task *task)
2878{
2879	trace_nfsd_cb_recall_any_done(cb, task);
2880	switch (task->tk_status) {
2881	case -NFS4ERR_DELAY:
2882		rpc_delay(task, 2 * HZ);
2883		return 0;
2884	default:
2885		return 1;
2886	}
2887}
2888
2889static void
2890nfsd4_cb_recall_any_release(struct nfsd4_callback *cb)
2891{
2892	struct nfs4_client *clp = cb->cb_clp;
2893	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2894
2895	spin_lock(&nn->client_lock);
2896	clear_bit(NFSD4_CLIENT_CB_RECALL_ANY, &clp->cl_flags);
2897	put_client_renew_locked(clp);
2898	spin_unlock(&nn->client_lock);
2899}
2900
2901static const struct nfsd4_callback_ops nfsd4_cb_recall_any_ops = {
2902	.done		= nfsd4_cb_recall_any_done,
2903	.release	= nfsd4_cb_recall_any_release,
2904};
2905
2906static struct nfs4_client *create_client(struct xdr_netobj name,
2907		struct svc_rqst *rqstp, nfs4_verifier *verf)
2908{
2909	struct nfs4_client *clp;
2910	struct sockaddr *sa = svc_addr(rqstp);
2911	int ret;
2912	struct net *net = SVC_NET(rqstp);
2913	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
2914	struct dentry *dentries[ARRAY_SIZE(client_files)];
2915
2916	clp = alloc_client(name, nn);
2917	if (clp == NULL)
2918		return NULL;
2919
2920	ret = copy_cred(&clp->cl_cred, &rqstp->rq_cred);
2921	if (ret) {
2922		free_client(clp);
2923		return NULL;
2924	}
2925	gen_clid(clp, nn);
2926	kref_init(&clp->cl_nfsdfs.cl_ref);
2927	nfsd4_init_cb(&clp->cl_cb_null, clp, NULL, NFSPROC4_CLNT_CB_NULL);
2928	clp->cl_time = ktime_get_boottime_seconds();
2929	clear_bit(0, &clp->cl_cb_slot_busy);
2930	copy_verf(clp, verf);
2931	memcpy(&clp->cl_addr, sa, sizeof(struct sockaddr_storage));
2932	clp->cl_cb_session = NULL;
2933	clp->net = net;
2934	clp->cl_nfsd_dentry = nfsd_client_mkdir(
2935		nn, &clp->cl_nfsdfs,
2936		clp->cl_clientid.cl_id - nn->clientid_base,
2937		client_files, dentries);
2938	clp->cl_nfsd_info_dentry = dentries[0];
2939	if (!clp->cl_nfsd_dentry) {
2940		free_client(clp);
2941		return NULL;
2942	}
2943	clp->cl_ra = kzalloc(sizeof(*clp->cl_ra), GFP_KERNEL);
2944	if (!clp->cl_ra) {
2945		free_client(clp);
2946		return NULL;
2947	}
2948	clp->cl_ra_time = 0;
2949	nfsd4_init_cb(&clp->cl_ra->ra_cb, clp, &nfsd4_cb_recall_any_ops,
2950			NFSPROC4_CLNT_CB_RECALL_ANY);
2951	return clp;
2952}
2953
2954static void
2955add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
2956{
2957	struct rb_node **new = &(root->rb_node), *parent = NULL;
2958	struct nfs4_client *clp;
2959
2960	while (*new) {
2961		clp = rb_entry(*new, struct nfs4_client, cl_namenode);
2962		parent = *new;
2963
2964		if (compare_blob(&clp->cl_name, &new_clp->cl_name) > 0)
2965			new = &((*new)->rb_left);
2966		else
2967			new = &((*new)->rb_right);
2968	}
2969
2970	rb_link_node(&new_clp->cl_namenode, parent, new);
2971	rb_insert_color(&new_clp->cl_namenode, root);
2972}
2973
2974static struct nfs4_client *
2975find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
2976{
2977	int cmp;
2978	struct rb_node *node = root->rb_node;
2979	struct nfs4_client *clp;
2980
2981	while (node) {
2982		clp = rb_entry(node, struct nfs4_client, cl_namenode);
2983		cmp = compare_blob(&clp->cl_name, name);
2984		if (cmp > 0)
2985			node = node->rb_left;
2986		else if (cmp < 0)
2987			node = node->rb_right;
2988		else
2989			return clp;
2990	}
2991	return NULL;
2992}
2993
2994static void
2995add_to_unconfirmed(struct nfs4_client *clp)
2996{
2997	unsigned int idhashval;
2998	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
2999
3000	lockdep_assert_held(&nn->client_lock);
3001
3002	clear_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
3003	add_clp_to_name_tree(clp, &nn->unconf_name_tree);
3004	idhashval = clientid_hashval(clp->cl_clientid.cl_id);
3005	list_add(&clp->cl_idhash, &nn->unconf_id_hashtbl[idhashval]);
3006	renew_client_locked(clp);
3007}
3008
3009static void
3010move_to_confirmed(struct nfs4_client *clp)
3011{
3012	unsigned int idhashval = clientid_hashval(clp->cl_clientid.cl_id);
3013	struct nfsd_net *nn = net_generic(clp->net, nfsd_net_id);
3014
3015	lockdep_assert_held(&nn->client_lock);
3016
3017	list_move(&clp->cl_idhash, &nn->conf_id_hashtbl[idhashval]);
3018	rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
3019	add_clp_to_name_tree(clp, &nn->conf_name_tree);
3020	set_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags);
3021	trace_nfsd_clid_confirmed(&clp->cl_clientid);
3022	renew_client_locked(clp);
3023}
3024
3025static struct nfs4_client *
3026find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
3027{
3028	struct nfs4_client *clp;
3029	unsigned int idhashval = clientid_hashval(clid->cl_id);
3030
3031	list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
3032		if (same_clid(&clp->cl_clientid, clid)) {
3033			if ((bool)clp->cl_minorversion != sessions)
3034				return NULL;
3035			renew_client_locked(clp);
3036			return clp;
3037		}
3038	}
3039	return NULL;
3040}
3041
3042static struct nfs4_client *
3043find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
3044{
3045	struct list_head *tbl = nn->conf_id_hashtbl;
3046
3047	lockdep_assert_held(&nn->client_lock);
3048	return find_client_in_id_table(tbl, clid, sessions);
3049}
3050
3051static struct nfs4_client *
3052find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
3053{
3054	struct list_head *tbl = nn->unconf_id_hashtbl;
3055
3056	lockdep_assert_held(&nn->client_lock);
3057	return find_client_in_id_table(tbl, clid, sessions);
3058}
3059
3060static bool clp_used_exchangeid(struct nfs4_client *clp)
3061{
3062	return clp->cl_exchange_flags != 0;
3063} 
3064
3065static struct nfs4_client *
3066find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
3067{
3068	lockdep_assert_held(&nn->client_lock);
3069	return find_clp_in_name_tree(name, &nn->conf_name_tree);
3070}
3071
3072static struct nfs4_client *
3073find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
3074{
3075	lockdep_assert_held(&nn->client_lock);
3076	return find_clp_in_name_tree(name, &nn->unconf_name_tree);
3077}
3078
3079static void
3080gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
3081{
3082	struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
3083	struct sockaddr	*sa = svc_addr(rqstp);
3084	u32 scopeid = rpc_get_scope_id(sa);
3085	unsigned short expected_family;
3086
3087	/* Currently, we only support tcp and tcp6 for the callback channel */
3088	if (se->se_callback_netid_len == 3 &&
3089	    !memcmp(se->se_callback_netid_val, "tcp", 3))
3090		expected_family = AF_INET;
3091	else if (se->se_callback_netid_len == 4 &&
3092		 !memcmp(se->se_callback_netid_val, "tcp6", 4))
3093		expected_family = AF_INET6;
3094	else
3095		goto out_err;
3096
3097	conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
3098					    se->se_callback_addr_len,
3099					    (struct sockaddr *)&conn->cb_addr,
3100					    sizeof(conn->cb_addr));
3101
3102	if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
3103		goto out_err;
3104
3105	if (conn->cb_addr.ss_family == AF_INET6)
3106		((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
3107
3108	conn->cb_prog = se->se_callback_prog;
3109	conn->cb_ident = se->se_callback_ident;
3110	memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
3111	trace_nfsd_cb_args(clp, conn);
3112	return;
3113out_err:
3114	conn->cb_addr.ss_family = AF_UNSPEC;
3115	conn->cb_addrlen = 0;
3116	trace_nfsd_cb_nodelegs(clp);
3117	return;
3118}
3119
3120/*
3121 * Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
3122 */
3123static void
3124nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
3125{
3126	struct xdr_buf *buf = resp->xdr->buf;
3127	struct nfsd4_slot *slot = resp->cstate.slot;
3128	unsigned int base;
3129
3130	dprintk("--> %s slot %p\n", __func__, slot);
3131
3132	slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
3133	slot->sl_opcnt = resp->opcnt;
3134	slot->sl_status = resp->cstate.status;
3135	free_svc_cred(&slot->sl_cred);
3136	copy_cred(&slot->sl_cred, &resp->rqstp->rq_cred);
3137
3138	if (!nfsd4_cache_this(resp)) {
3139		slot->sl_flags &= ~NFSD4_SLOT_CACHED;
3140		return;
3141	}
3142	slot->sl_flags |= NFSD4_SLOT_CACHED;
3143
3144	base = resp->cstate.data_offset;
3145	slot->sl_datalen = buf->len - base;
3146	if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
3147		WARN(1, "%s: sessions DRC could not cache compound\n",
3148		     __func__);
3149	return;
3150}
3151
3152/*
3153 * Encode the replay sequence operation from the slot values.
3154 * If cachethis is FALSE encode the uncached rep error on the next
3155 * operation which sets resp->p and increments resp->opcnt for
3156 * nfs4svc_encode_compoundres.
3157 *
3158 */
3159static __be32
3160nfsd4_enc_sequence_replay(struct nfsd4_compoundargs *args,
3161			  struct nfsd4_compoundres *resp)
3162{
3163	struct nfsd4_op *op;
3164	struct nfsd4_slot *slot = resp->cstate.slot;
3165
3166	/* Encode the replayed sequence operation */
3167	op = &args->ops[resp->opcnt - 1];
3168	nfsd4_encode_operation(resp, op);
3169
3170	if (slot->sl_flags & NFSD4_SLOT_CACHED)
3171		return op->status;
3172	if (args->opcnt == 1) {
3173		/*
3174		 * The original operation wasn't a solo sequence--we
3175		 * always cache those--so this retry must not match the
3176		 * original:
3177		 */
3178		op->status = nfserr_seq_false_retry;
3179	} else {
3180		op = &args->ops[resp->opcnt++];
3181		op->status = nfserr_retry_uncached_rep;
3182		nfsd4_encode_operation(resp, op);
3183	}
3184	return op->status;
3185}
3186
3187/*
3188 * The sequence operation is not cached because we can use the slot and
3189 * session values.
3190 */
3191static __be32
3192nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
3193			 struct nfsd4_sequence *seq)
3194{
3195	struct nfsd4_slot *slot = resp->cstate.slot;
3196	struct xdr_stream *xdr = resp->xdr;
3197	__be32 *p;
3198	__be32 status;
3199
3200	dprintk("--> %s slot %p\n", __func__, slot);
3201
3202	status = nfsd4_enc_sequence_replay(resp->rqstp->rq_argp, resp);
3203	if (status)
3204		return status;
3205
3206	p = xdr_reserve_space(xdr, slot->sl_datalen);
3207	if (!p) {
3208		WARN_ON_ONCE(1);
3209		return nfserr_serverfault;
3210	}
3211	xdr_encode_opaque_fixed(p, slot->sl_data, slot->sl_datalen);
3212	xdr_commit_encode(xdr);
3213
3214	resp->opcnt = slot->sl_opcnt;
3215	return slot->sl_status;
3216}
3217
3218/*
3219 * Set the exchange_id flags returned by the server.
3220 */
3221static void
3222nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
3223{
3224#ifdef CONFIG_NFSD_PNFS
3225	new->cl_exchange_flags |= EXCHGID4_FLAG_USE_PNFS_MDS;
3226#else
3227	new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
3228#endif
3229
3230	/* Referrals are supported, Migration is not. */
3231	new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
3232
3233	/* set the wire flags to return to client. */
3234	clid->flags = new->cl_exchange_flags;
3235}
3236
3237static bool client_has_openowners(struct nfs4_client *clp)
3238{
3239	struct nfs4_openowner *oo;
3240
3241	list_for_each_entry(oo, &clp->cl_openowners, oo_perclient) {
3242		if (!list_empty(&oo->oo_owner.so_stateids))
3243			return true;
3244	}
3245	return false;
3246}
3247
3248static bool client_has_state(struct nfs4_client *clp)
3249{
3250	return client_has_openowners(clp)
3251#ifdef CONFIG_NFSD_PNFS
3252		|| !list_empty(&clp->cl_lo_states)
3253#endif
3254		|| !list_empty(&clp->cl_delegations)
3255		|| !list_empty(&clp->cl_sessions)
3256		|| !list_empty(&clp->async_copies);
3257}
3258
3259static __be32 copy_impl_id(struct nfs4_client *clp,
3260				struct nfsd4_exchange_id *exid)
3261{
3262	if (!exid->nii_domain.data)
3263		return 0;
3264	xdr_netobj_dup(&clp->cl_nii_domain, &exid->nii_domain, GFP_KERNEL);
3265	if (!clp->cl_nii_domain.data)
3266		return nfserr_jukebox;
3267	xdr_netobj_dup(&clp->cl_nii_name, &exid->nii_name, GFP_KERNEL);
3268	if (!clp->cl_nii_name.data)
3269		return nfserr_jukebox;
3270	clp->cl_nii_time = exid->nii_time;
3271	return 0;
3272}
3273
3274__be32
3275nfsd4_exchange_id(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3276		union nfsd4_op_u *u)
3277{
3278	struct nfsd4_exchange_id *exid = &u->exchange_id;
3279	struct nfs4_client *conf, *new;
3280	struct nfs4_client *unconf = NULL;
3281	__be32 status;
3282	char			addr_str[INET6_ADDRSTRLEN];
3283	nfs4_verifier		verf = exid->verifier;
3284	struct sockaddr		*sa = svc_addr(rqstp);
3285	bool	update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
3286	struct nfsd_net		*nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3287
3288	rpc_ntop(sa, addr_str, sizeof(addr_str));
3289	dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
3290		"ip_addr=%s flags %x, spa_how %u\n",
3291		__func__, rqstp, exid, exid->clname.len, exid->clname.data,
3292		addr_str, exid->flags, exid->spa_how);
3293
3294	if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
3295		return nfserr_inval;
3296
3297	new = create_client(exid->clname, rqstp, &verf);
3298	if (new == NULL)
3299		return nfserr_jukebox;
3300	status = copy_impl_id(new, exid);
3301	if (status)
3302		goto out_nolock;
3303
3304	switch (exid->spa_how) {
3305	case SP4_MACH_CRED:
3306		exid->spo_must_enforce[0] = 0;
3307		exid->spo_must_enforce[1] = (
3308			1 << (OP_BIND_CONN_TO_SESSION - 32) |
3309			1 << (OP_EXCHANGE_ID - 32) |
3310			1 << (OP_CREATE_SESSION - 32) |
3311			1 << (OP_DESTROY_SESSION - 32) |
3312			1 << (OP_DESTROY_CLIENTID - 32));
3313
3314		exid->spo_must_allow[0] &= (1 << (OP_CLOSE) |
3315					1 << (OP_OPEN_DOWNGRADE) |
3316					1 << (OP_LOCKU) |
3317					1 << (OP_DELEGRETURN));
3318
3319		exid->spo_must_allow[1] &= (
3320					1 << (OP_TEST_STATEID - 32) |
3321					1 << (OP_FREE_STATEID - 32));
3322		if (!svc_rqst_integrity_protected(rqstp)) {
3323			status = nfserr_inval;
3324			goto out_nolock;
3325		}
3326		/*
3327		 * Sometimes userspace doesn't give us a principal.
3328		 * Which is a bug, really.  Anyway, we can't enforce
3329		 * MACH_CRED in that case, better to give up now:
3330		 */
3331		if (!new->cl_cred.cr_principal &&
3332					!new->cl_cred.cr_raw_principal) {
3333			status = nfserr_serverfault;
3334			goto out_nolock;
3335		}
3336		new->cl_mach_cred = true;
3337		break;
3338	case SP4_NONE:
3339		break;
3340	default:				/* checked by xdr code */
3341		WARN_ON_ONCE(1);
3342		fallthrough;
3343	case SP4_SSV:
3344		status = nfserr_encr_alg_unsupp;
3345		goto out_nolock;
3346	}
3347
3348	/* Cases below refer to rfc 5661 section 18.35.4: */
3349	spin_lock(&nn->client_lock);
3350	conf = find_confirmed_client_by_name(&exid->clname, nn);
3351	if (conf) {
3352		bool creds_match = same_creds(&conf->cl_cred, &rqstp->rq_cred);
3353		bool verfs_match = same_verf(&verf, &conf->cl_verifier);
3354
3355		if (update) {
3356			if (!clp_used_exchangeid(conf)) { /* buggy client */
3357				status = nfserr_inval;
3358				goto out;
3359			}
3360			if (!nfsd4_mach_creds_match(conf, rqstp)) {
3361				status = nfserr_wrong_cred;
3362				goto out;
3363			}
3364			if (!creds_match) { /* case 9 */
3365				status = nfserr_perm;
3366				goto out;
3367			}
3368			if (!verfs_match) { /* case 8 */
3369				status = nfserr_not_same;
3370				goto out;
3371			}
3372			/* case 6 */
3373			exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
3374			trace_nfsd_clid_confirmed_r(conf);
3375			goto out_copy;
3376		}
3377		if (!creds_match) { /* case 3 */
3378			if (client_has_state(conf)) {
3379				status = nfserr_clid_inuse;
3380				trace_nfsd_clid_cred_mismatch(conf, rqstp);
3381				goto out;
3382			}
3383			goto out_new;
3384		}
3385		if (verfs_match) { /* case 2 */
3386			conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
3387			trace_nfsd_clid_confirmed_r(conf);
3388			goto out_copy;
3389		}
3390		/* case 5, client reboot */
3391		trace_nfsd_clid_verf_mismatch(conf, rqstp, &verf);
3392		conf = NULL;
3393		goto out_new;
3394	}
3395
3396	if (update) { /* case 7 */
3397		status = nfserr_noent;
3398		goto out;
3399	}
3400
3401	unconf = find_unconfirmed_client_by_name(&exid->clname, nn);
3402	if (unconf) /* case 4, possible retry or client restart */
3403		unhash_client_locked(unconf);
3404
3405	/* case 1, new owner ID */
3406	trace_nfsd_clid_fresh(new);
3407
3408out_new:
3409	if (conf) {
3410		status = mark_client_expired_locked(conf);
3411		if (status)
3412			goto out;
3413		trace_nfsd_clid_replaced(&conf->cl_clientid);
3414	}
3415	new->cl_minorversion = cstate->minorversion;
3416	new->cl_spo_must_allow.u.words[0] = exid->spo_must_allow[0];
3417	new->cl_spo_must_allow.u.words[1] = exid->spo_must_allow[1];
3418
3419	add_to_unconfirmed(new);
3420	swap(new, conf);
3421out_copy:
3422	exid->clientid.cl_boot = conf->cl_clientid.cl_boot;
3423	exid->clientid.cl_id = conf->cl_clientid.cl_id;
3424
3425	exid->seqid = conf->cl_cs_slot.sl_seqid + 1;
3426	nfsd4_set_ex_flags(conf, exid);
3427
3428	dprintk("nfsd4_exchange_id seqid %d flags %x\n",
3429		conf->cl_cs_slot.sl_seqid, conf->cl_exchange_flags);
3430	status = nfs_ok;
3431
3432out:
3433	spin_unlock(&nn->client_lock);
3434out_nolock:
3435	if (new)
3436		expire_client(new);
3437	if (unconf) {
3438		trace_nfsd_clid_expire_unconf(&unconf->cl_clientid);
3439		expire_client(unconf);
3440	}
3441	return status;
3442}
3443
3444static __be32
3445check_slot_seqid(u32 seqid, u32 slot_seqid, int slot_inuse)
3446{
3447	dprintk("%s enter. seqid %d slot_seqid %d\n", __func__, seqid,
3448		slot_seqid);
3449
3450	/* The slot is in use, and no response has been sent. */
3451	if (slot_inuse) {
3452		if (seqid == slot_seqid)
3453			return nfserr_jukebox;
3454		else
3455			return nfserr_seq_misordered;
3456	}
3457	/* Note unsigned 32-bit arithmetic handles wraparound: */
3458	if (likely(seqid == slot_seqid + 1))
3459		return nfs_ok;
3460	if (seqid == slot_seqid)
3461		return nfserr_replay_cache;
3462	return nfserr_seq_misordered;
3463}
3464
3465/*
3466 * Cache the create session result into the create session single DRC
3467 * slot cache by saving the xdr structure. sl_seqid has been set.
3468 * Do this for solo or embedded create session operations.
3469 */
3470static void
3471nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
3472			   struct nfsd4_clid_slot *slot, __be32 nfserr)
3473{
3474	slot->sl_status = nfserr;
3475	memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
3476}
3477
3478static __be32
3479nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
3480			    struct nfsd4_clid_slot *slot)
3481{
3482	memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
3483	return slot->sl_status;
3484}
3485
3486#define NFSD_MIN_REQ_HDR_SEQ_SZ	((\
3487			2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
3488			1 +	/* MIN tag is length with zero, only length */ \
3489			3 +	/* version, opcount, opcode */ \
3490			XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
3491				/* seqid, slotID, slotID, cache */ \
3492			4 ) * sizeof(__be32))
3493
3494#define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
3495			2 +	/* verifier: AUTH_NULL, length 0 */\
3496			1 +	/* status */ \
3497			1 +	/* MIN tag is length with zero, only length */ \
3498			3 +	/* opcount, opcode, opstatus*/ \
3499			XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
3500				/* seqid, slotID, slotID, slotID, status */ \
3501			5 ) * sizeof(__be32))
3502
3503static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
3504{
3505	u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
3506
3507	if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
3508		return nfserr_toosmall;
3509	if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
3510		return nfserr_toosmall;
3511	ca->headerpadsz = 0;
3512	ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
3513	ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
3514	ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
3515	ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
3516			NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
3517	ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
3518	/*
3519	 * Note decreasing slot size below client's request may make it
3520	 * difficult for client to function correctly, whereas
3521	 * decreasing the number of slots will (just?) affect
3522	 * performance.  When short on memory we therefore prefer to
3523	 * decrease number of slots instead of their size.  Clients that
3524	 * request larger slots than they need will get poor results:
3525	 * Note that we always allow at least one slot, because our
3526	 * accounting is soft and provides no guarantees either way.
3527	 */
3528	ca->maxreqs = nfsd4_get_drc_mem(ca, nn);
3529
3530	return nfs_ok;
3531}
3532
3533/*
3534 * Server's NFSv4.1 backchannel support is AUTH_SYS-only for now.
3535 * These are based on similar macros in linux/sunrpc/msg_prot.h .
3536 */
3537#define RPC_MAX_HEADER_WITH_AUTH_SYS \
3538	(RPC_CALLHDRSIZE + 2 * (2 + UNX_CALLSLACK))
3539
3540#define RPC_MAX_REPHEADER_WITH_AUTH_SYS \
3541	(RPC_REPHDRSIZE + (2 + NUL_REPLYSLACK))
3542
3543#define NFSD_CB_MAX_REQ_SZ	((NFS4_enc_cb_recall_sz + \
3544				 RPC_MAX_HEADER_WITH_AUTH_SYS) * sizeof(__be32))
3545#define NFSD_CB_MAX_RESP_SZ	((NFS4_dec_cb_recall_sz + \
3546				 RPC_MAX_REPHEADER_WITH_AUTH_SYS) * \
3547				 sizeof(__be32))
3548
3549static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
3550{
3551	ca->headerpadsz = 0;
3552
3553	if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
3554		return nfserr_toosmall;
3555	if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
3556		return nfserr_toosmall;
3557	ca->maxresp_cached = 0;
3558	if (ca->maxops < 2)
3559		return nfserr_toosmall;
3560
3561	return nfs_ok;
3562}
3563
3564static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
3565{
3566	switch (cbs->flavor) {
3567	case RPC_AUTH_NULL:
3568	case RPC_AUTH_UNIX:
3569		return nfs_ok;
3570	default:
3571		/*
3572		 * GSS case: the spec doesn't allow us to return this
3573		 * error.  But it also doesn't allow us not to support
3574		 * GSS.
3575		 * I'd rather this fail hard than return some error the
3576		 * client might think it can already handle:
3577		 */
3578		return nfserr_encr_alg_unsupp;
3579	}
3580}
3581
3582__be32
3583nfsd4_create_session(struct svc_rqst *rqstp,
3584		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
3585{
3586	struct nfsd4_create_session *cr_ses = &u->create_session;
3587	struct sockaddr *sa = svc_addr(rqstp);
3588	struct nfs4_client *conf, *unconf;
3589	struct nfs4_client *old = NULL;
3590	struct nfsd4_session *new;
3591	struct nfsd4_conn *conn;
3592	struct nfsd4_clid_slot *cs_slot = NULL;
3593	__be32 status = 0;
3594	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3595
3596	if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
3597		return nfserr_inval;
3598	status = nfsd4_check_cb_sec(&cr_ses->cb_sec);
3599	if (status)
3600		return status;
3601	status = check_forechannel_attrs(&cr_ses->fore_channel, nn);
3602	if (status)
3603		return status;
3604	status = check_backchannel_attrs(&cr_ses->back_channel);
3605	if (status)
3606		goto out_release_drc_mem;
3607	status = nfserr_jukebox;
3608	new = alloc_session(&cr_ses->fore_channel, &cr_ses->back_channel);
3609	if (!new)
3610		goto out_release_drc_mem;
3611	conn = alloc_conn_from_crses(rqstp, cr_ses);
3612	if (!conn)
3613		goto out_free_session;
3614
3615	spin_lock(&nn->client_lock);
3616	unconf = find_unconfirmed_client(&cr_ses->clientid, true, nn);
3617	conf = find_confirmed_client(&cr_ses->clientid, true, nn);
3618	WARN_ON_ONCE(conf && unconf);
3619
3620	if (conf) {
3621		status = nfserr_wrong_cred;
3622		if (!nfsd4_mach_creds_match(conf, rqstp))
3623			goto out_free_conn;
3624		cs_slot = &conf->cl_cs_slot;
3625		status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
3626		if (status) {
3627			if (status == nfserr_replay_cache)
3628				status = nfsd4_replay_create_session(cr_ses, cs_slot);
3629			goto out_free_conn;
3630		}
3631	} else if (unconf) {
3632		status = nfserr_clid_inuse;
3633		if (!same_creds(&unconf->cl_cred, &rqstp->rq_cred) ||
3634		    !rpc_cmp_addr(sa, (struct sockaddr *) &unconf->cl_addr)) {
3635			trace_nfsd_clid_cred_mismatch(unconf, rqstp);
3636			goto out_free_conn;
3637		}
3638		status = nfserr_wrong_cred;
3639		if (!nfsd4_mach_creds_match(unconf, rqstp))
3640			goto out_free_conn;
3641		cs_slot = &unconf->cl_cs_slot;
3642		status = check_slot_seqid(cr_ses->seqid, cs_slot->sl_seqid, 0);
3643		if (status) {
3644			/* an unconfirmed replay returns misordered */
3645			status = nfserr_seq_misordered;
3646			goto out_free_conn;
3647		}
3648		old = find_confirmed_client_by_name(&unconf->cl_name, nn);
3649		if (old) {
3650			status = mark_client_expired_locked(old);
3651			if (status) {
3652				old = NULL;
3653				goto out_free_conn;
3654			}
3655			trace_nfsd_clid_replaced(&old->cl_clientid);
3656		}
3657		move_to_confirmed(unconf);
3658		conf = unconf;
3659	} else {
3660		status = nfserr_stale_clientid;
3661		goto out_free_conn;
3662	}
3663	status = nfs_ok;
3664	/* Persistent sessions are not supported */
3665	cr_ses->flags &= ~SESSION4_PERSIST;
3666	/* Upshifting from TCP to RDMA is not supported */
3667	cr_ses->flags &= ~SESSION4_RDMA;
3668
3669	init_session(rqstp, new, conf, cr_ses);
3670	nfsd4_get_session_locked(new);
3671
3672	memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
3673	       NFS4_MAX_SESSIONID_LEN);
3674	cs_slot->sl_seqid++;
3675	cr_ses->seqid = cs_slot->sl_seqid;
3676
3677	/* cache solo and embedded create sessions under the client_lock */
3678	nfsd4_cache_create_session(cr_ses, cs_slot, status);
3679	spin_unlock(&nn->client_lock);
3680	if (conf == unconf)
3681		fsnotify_dentry(conf->cl_nfsd_info_dentry, FS_MODIFY);
3682	/* init connection and backchannel */
3683	nfsd4_init_conn(rqstp, conn, new);
3684	nfsd4_put_session(new);
3685	if (old)
3686		expire_client(old);
3687	return status;
3688out_free_conn:
3689	spin_unlock(&nn->client_lock);
3690	free_conn(conn);
3691	if (old)
3692		expire_client(old);
3693out_free_session:
3694	__free_session(new);
3695out_release_drc_mem:
3696	nfsd4_put_drc_mem(&cr_ses->fore_channel);
3697	return status;
3698}
3699
3700static __be32 nfsd4_map_bcts_dir(u32 *dir)
3701{
3702	switch (*dir) {
3703	case NFS4_CDFC4_FORE:
3704	case NFS4_CDFC4_BACK:
3705		return nfs_ok;
3706	case NFS4_CDFC4_FORE_OR_BOTH:
3707	case NFS4_CDFC4_BACK_OR_BOTH:
3708		*dir = NFS4_CDFC4_BOTH;
3709		return nfs_ok;
3710	}
3711	return nfserr_inval;
3712}
3713
3714__be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp,
3715		struct nfsd4_compound_state *cstate,
3716		union nfsd4_op_u *u)
3717{
3718	struct nfsd4_backchannel_ctl *bc = &u->backchannel_ctl;
3719	struct nfsd4_session *session = cstate->session;
3720	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
3721	__be32 status;
3722
3723	status = nfsd4_check_cb_sec(&bc->bc_cb_sec);
3724	if (status)
3725		return status;
3726	spin_lock(&nn->client_lock);
3727	session->se_cb_prog = bc->bc_cb_program;
3728	session->se_cb_sec = bc->bc_cb_sec;
3729	spin_unlock(&nn->client_lock);
3730
3731	nfsd4_probe_callback(session->se_client);
3732
3733	return nfs_ok;
3734}
3735
3736static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
3737{
3738	struct nfsd4_conn *c;
3739
3740	list_for_each_entry(c, &s->se_conns, cn_persession) {
3741		if (c->cn_xprt == xpt) {
3742			return c;
3743		}
3744	}
3745	return NULL;
3746}
3747
3748static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst,
3749		struct nfsd4_session *session, u32 req, struct nfsd4_conn **conn)
3750{
3751	struct nfs4_client *clp = session->se_client;
3752	struct svc_xprt *xpt = rqst->rq_xprt;
3753	struct nfsd4_conn *c;
3754	__be32 status;
3755
3756	/* Following the last paragraph of RFC 5661 Section 18.34.3: */
3757	spin_lock(&clp->cl_lock);
3758	c = __nfsd4_find_conn(xpt, session);
3759	if (!c)
3760		status = nfserr_noent;
3761	else if (req == c->cn_flags)
3762		status = nfs_ok;
3763	else if (req == NFS4_CDFC4_FORE_OR_BOTH &&
3764				c->cn_flags != NFS4_CDFC4_BACK)
3765		status = nfs_ok;
3766	else if (req == NFS4_CDFC4_BACK_OR_BOTH &&
3767				c->cn_flags != NFS4_CDFC4_FORE)
3768		status = nfs_ok;
3769	else
3770		status = nfserr_inval;
3771	spin_unlock(&clp->cl_lock);
3772	if (status == nfs_ok && conn)
3773		*conn = c;
3774	return status;
3775}
3776
3777__be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
3778		     struct nfsd4_compound_state *cstate,
3779		     union nfsd4_op_u *u)
3780{
3781	struct nfsd4_bind_conn_to_session *bcts = &u->bind_conn_to_session;
3782	__be32 status;
3783	struct nfsd4_conn *conn;
3784	struct nfsd4_session *session;
3785	struct net *net = SVC_NET(rqstp);
3786	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3787
3788	if (!nfsd4_last_compound_op(rqstp))
3789		return nfserr_not_only_op;
3790	spin_lock(&nn->client_lock);
3791	session = find_in_sessionid_hashtbl(&bcts->sessionid, net, &status);
3792	spin_unlock(&nn->client_lock);
3793	if (!session)
3794		goto out_no_session;
3795	status = nfserr_wrong_cred;
3796	if (!nfsd4_mach_creds_match(session->se_client, rqstp))
3797		goto out;
3798	status = nfsd4_match_existing_connection(rqstp, session,
3799			bcts->dir, &conn);
3800	if (status == nfs_ok) {
3801		if (bcts->dir == NFS4_CDFC4_FORE_OR_BOTH ||
3802				bcts->dir == NFS4_CDFC4_BACK)
3803			conn->cn_flags |= NFS4_CDFC4_BACK;
3804		nfsd4_probe_callback(session->se_client);
3805		goto out;
3806	}
3807	if (status == nfserr_inval)
3808		goto out;
3809	status = nfsd4_map_bcts_dir(&bcts->dir);
3810	if (status)
3811		goto out;
3812	conn = alloc_conn(rqstp, bcts->dir);
3813	status = nfserr_jukebox;
3814	if (!conn)
3815		goto out;
3816	nfsd4_init_conn(rqstp, conn, session);
3817	status = nfs_ok;
3818out:
3819	nfsd4_put_session(session);
3820out_no_session:
3821	return status;
3822}
3823
3824static bool nfsd4_compound_in_session(struct nfsd4_compound_state *cstate, struct nfs4_sessionid *sid)
3825{
3826	if (!cstate->session)
3827		return false;
3828	return !memcmp(sid, &cstate->session->se_sessionid, sizeof(*sid));
3829}
3830
3831__be32
3832nfsd4_destroy_session(struct svc_rqst *r, struct nfsd4_compound_state *cstate,
3833		union nfsd4_op_u *u)
3834{
3835	struct nfs4_sessionid *sessionid = &u->destroy_session.sessionid;
3836	struct nfsd4_session *ses;
3837	__be32 status;
3838	int ref_held_by_me = 0;
3839	struct net *net = SVC_NET(r);
3840	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3841
3842	status = nfserr_not_only_op;
3843	if (nfsd4_compound_in_session(cstate, sessionid)) {
3844		if (!nfsd4_last_compound_op(r))
3845			goto out;
3846		ref_held_by_me++;
3847	}
3848	dump_sessionid(__func__, sessionid);
3849	spin_lock(&nn->client_lock);
3850	ses = find_in_sessionid_hashtbl(sessionid, net, &status);
3851	if (!ses)
3852		goto out_client_lock;
3853	status = nfserr_wrong_cred;
3854	if (!nfsd4_mach_creds_match(ses->se_client, r))
3855		goto out_put_session;
3856	status = mark_session_dead_locked(ses, 1 + ref_held_by_me);
3857	if (status)
3858		goto out_put_session;
3859	unhash_session(ses);
3860	spin_unlock(&nn->client_lock);
3861
3862	nfsd4_probe_callback_sync(ses->se_client);
3863
3864	spin_lock(&nn->client_lock);
3865	status = nfs_ok;
3866out_put_session:
3867	nfsd4_put_session_locked(ses);
3868out_client_lock:
3869	spin_unlock(&nn->client_lock);
3870out:
3871	return status;
3872}
3873
3874static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
3875{
3876	struct nfs4_client *clp = ses->se_client;
3877	struct nfsd4_conn *c;
3878	__be32 status = nfs_ok;
3879	int ret;
3880
3881	spin_lock(&clp->cl_lock);
3882	c = __nfsd4_find_conn(new->cn_xprt, ses);
3883	if (c)
3884		goto out_free;
3885	status = nfserr_conn_not_bound_to_session;
3886	if (clp->cl_mach_cred)
3887		goto out_free;
3888	__nfsd4_hash_conn(new, ses);
3889	spin_unlock(&clp->cl_lock);
3890	ret = nfsd4_register_conn(new);
3891	if (ret)
3892		/* oops; xprt is already down: */
3893		nfsd4_conn_lost(&new->cn_xpt_user);
3894	return nfs_ok;
3895out_free:
3896	spin_unlock(&clp->cl_lock);
3897	free_conn(new);
3898	return status;
3899}
3900
3901static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
3902{
3903	struct nfsd4_compoundargs *args = rqstp->rq_argp;
3904
3905	return args->opcnt > session->se_fchannel.maxops;
3906}
3907
3908static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
3909				  struct nfsd4_session *session)
3910{
3911	struct xdr_buf *xb = &rqstp->rq_arg;
3912
3913	return xb->len > session->se_fchannel.maxreq_sz;
3914}
3915
3916static bool replay_matches_cache(struct svc_rqst *rqstp,
3917		 struct nfsd4_sequence *seq, struct nfsd4_slot *slot)
3918{
3919	struct nfsd4_compoundargs *argp = rqstp->rq_argp;
3920
3921	if ((bool)(slot->sl_flags & NFSD4_SLOT_CACHETHIS) !=
3922	    (bool)seq->cachethis)
3923		return false;
3924	/*
3925	 * If there's an error then the reply can have fewer ops than
3926	 * the call.
3927	 */
3928	if (slot->sl_opcnt < argp->opcnt && !slot->sl_status)
3929		return false;
3930	/*
3931	 * But if we cached a reply with *more* ops than the call you're
3932	 * sending us now, then this new call is clearly not really a
3933	 * replay of the old one:
3934	 */
3935	if (slot->sl_opcnt > argp->opcnt)
3936		return false;
3937	/* This is the only check explicitly called by spec: */
3938	if (!same_creds(&rqstp->rq_cred, &slot->sl_cred))
3939		return false;
3940	/*
3941	 * There may be more comparisons we could actually do, but the
3942	 * spec doesn't require us to catch every case where the calls
3943	 * don't match (that would require caching the call as well as
3944	 * the reply), so we don't bother.
3945	 */
3946	return true;
3947}
3948
3949__be32
3950nfsd4_sequence(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3951		union nfsd4_op_u *u)
3952{
3953	struct nfsd4_sequence *seq = &u->sequence;
3954	struct nfsd4_compoundres *resp = rqstp->rq_resp;
3955	struct xdr_stream *xdr = resp->xdr;
3956	struct nfsd4_session *session;
3957	struct nfs4_client *clp;
3958	struct nfsd4_slot *slot;
3959	struct nfsd4_conn *conn;
3960	__be32 status;
3961	int buflen;
3962	struct net *net = SVC_NET(rqstp);
3963	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
3964
3965	if (resp->opcnt != 1)
3966		return nfserr_sequence_pos;
3967
3968	/*
3969	 * Will be either used or freed by nfsd4_sequence_check_conn
3970	 * below.
3971	 */
3972	conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
3973	if (!conn)
3974		return nfserr_jukebox;
3975
3976	spin_lock(&nn->client_lock);
3977	session = find_in_sessionid_hashtbl(&seq->sessionid, net, &status);
3978	if (!session)
3979		goto out_no_session;
3980	clp = session->se_client;
3981
3982	status = nfserr_too_many_ops;
3983	if (nfsd4_session_too_many_ops(rqstp, session))
3984		goto out_put_session;
3985
3986	status = nfserr_req_too_big;
3987	if (nfsd4_request_too_big(rqstp, session))
3988		goto out_put_session;
3989
3990	status = nfserr_badslot;
3991	if (seq->slotid >= session->se_fchannel.maxreqs)
3992		goto out_put_session;
3993
3994	slot = session->se_slots[seq->slotid];
3995	dprintk("%s: slotid %d\n", __func__, seq->slotid);
3996
3997	/* We do not negotiate the number of slots yet, so set the
3998	 * maxslots to the session maxreqs which is used to encode
3999	 * sr_highest_slotid and the sr_target_slot id to maxslots */
4000	seq->maxslots = session->se_fchannel.maxreqs;
4001
4002	status = check_slot_seqid(seq->seqid, slot->sl_seqid,
4003					slot->sl_flags & NFSD4_SLOT_INUSE);
4004	if (status == nfserr_replay_cache) {
4005		status = nfserr_seq_misordered;
4006		if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
4007			goto out_put_session;
4008		status = nfserr_seq_false_retry;
4009		if (!replay_matches_cache(rqstp, seq, slot))
4010			goto out_put_session;
4011		cstate->slot = slot;
4012		cstate->session = session;
4013		cstate->clp = clp;
4014		/* Return the cached reply status and set cstate->status
4015		 * for nfsd4_proc_compound processing */
4016		status = nfsd4_replay_cache_entry(resp, seq);
4017		cstate->status = nfserr_replay_cache;
4018		goto out;
4019	}
4020	if (status)
4021		goto out_put_session;
4022
4023	status = nfsd4_sequence_check_conn(conn, session);
4024	conn = NULL;
4025	if (status)
4026		goto out_put_session;
4027
4028	buflen = (seq->cachethis) ?
4029			session->se_fchannel.maxresp_cached :
4030			session->se_fchannel.maxresp_sz;
4031	status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
4032				    nfserr_rep_too_big;
4033	if (xdr_restrict_buflen(xdr, buflen - rqstp->rq_auth_slack))
4034		goto out_put_session;
4035	svc_reserve(rqstp, buflen);
4036
4037	status = nfs_ok;
4038	/* Success! bump slot seqid */
4039	slot->sl_seqid = seq->seqid;
4040	slot->sl_flags |= NFSD4_SLOT_INUSE;
4041	if (seq->cachethis)
4042		slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
4043	else
4044		slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
4045
4046	cstate->slot = slot;
4047	cstate->session = session;
4048	cstate->clp = clp;
4049
4050out:
4051	switch (clp->cl_cb_state) {
4052	case NFSD4_CB_DOWN:
4053		seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
4054		break;
4055	case NFSD4_CB_FAULT:
4056		seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
4057		break;
4058	default:
4059		seq->status_flags = 0;
4060	}
4061	if (!list_empty(&clp->cl_revoked))
4062		seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
4063out_no_session:
4064	if (conn)
4065		free_conn(conn);
4066	spin_unlock(&nn->client_lock);
4067	return status;
4068out_put_session:
4069	nfsd4_put_session_locked(session);
4070	goto out_no_session;
4071}
4072
4073void
4074nfsd4_sequence_done(struct nfsd4_compoundres *resp)
4075{
4076	struct nfsd4_compound_state *cs = &resp->cstate;
4077
4078	if (nfsd4_has_session(cs)) {
4079		if (cs->status != nfserr_replay_cache) {
4080			nfsd4_store_cache_entry(resp);
4081			cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
4082		}
4083		/* Drop session reference that was taken in nfsd4_sequence() */
4084		nfsd4_put_session(cs->session);
4085	} else if (cs->clp)
4086		put_client_renew(cs->clp);
4087}
4088
4089__be32
4090nfsd4_destroy_clientid(struct svc_rqst *rqstp,
4091		struct nfsd4_compound_state *cstate,
4092		union nfsd4_op_u *u)
4093{
4094	struct nfsd4_destroy_clientid *dc = &u->destroy_clientid;
4095	struct nfs4_client *conf, *unconf;
4096	struct nfs4_client *clp = NULL;
4097	__be32 status = 0;
4098	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4099
4100	spin_lock(&nn->client_lock);
4101	unconf = find_unconfirmed_client(&dc->clientid, true, nn);
4102	conf = find_confirmed_client(&dc->clientid, true, nn);
4103	WARN_ON_ONCE(conf && unconf);
4104
4105	if (conf) {
4106		if (client_has_state(conf)) {
4107			status = nfserr_clientid_busy;
4108			goto out;
4109		}
4110		status = mark_client_expired_locked(conf);
4111		if (status)
4112			goto out;
4113		clp = conf;
4114	} else if (unconf)
4115		clp = unconf;
4116	else {
4117		status = nfserr_stale_clientid;
4118		goto out;
4119	}
4120	if (!nfsd4_mach_creds_match(clp, rqstp)) {
4121		clp = NULL;
4122		status = nfserr_wrong_cred;
4123		goto out;
4124	}
4125	trace_nfsd_clid_destroyed(&clp->cl_clientid);
4126	unhash_client_locked(clp);
4127out:
4128	spin_unlock(&nn->client_lock);
4129	if (clp)
4130		expire_client(clp);
4131	return status;
4132}
4133
4134__be32
4135nfsd4_reclaim_complete(struct svc_rqst *rqstp,
4136		struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
4137{
4138	struct nfsd4_reclaim_complete *rc = &u->reclaim_complete;
4139	struct nfs4_client *clp = cstate->clp;
4140	__be32 status = 0;
4141
4142	if (rc->rca_one_fs) {
4143		if (!cstate->current_fh.fh_dentry)
4144			return nfserr_nofilehandle;
4145		/*
4146		 * We don't take advantage of the rca_one_fs case.
4147		 * That's OK, it's optional, we can safely ignore it.
4148		 */
4149		return nfs_ok;
4150	}
4151
4152	status = nfserr_complete_already;
4153	if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &clp->cl_flags))
4154		goto out;
4155
4156	status = nfserr_stale_clientid;
4157	if (is_client_expired(clp))
4158		/*
4159		 * The following error isn't really legal.
4160		 * But we only get here if the client just explicitly
4161		 * destroyed the client.  Surely it no longer cares what
4162		 * error it gets back on an operation for the dead
4163		 * client.
4164		 */
4165		goto out;
4166
4167	status = nfs_ok;
4168	trace_nfsd_clid_reclaim_complete(&clp->cl_clientid);
4169	nfsd4_client_record_create(clp);
4170	inc_reclaim_complete(clp);
4171out:
4172	return status;
4173}
4174
4175__be32
4176nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4177		  union nfsd4_op_u *u)
4178{
4179	struct nfsd4_setclientid *setclid = &u->setclientid;
4180	struct xdr_netobj 	clname = setclid->se_name;
4181	nfs4_verifier		clverifier = setclid->se_verf;
4182	struct nfs4_client	*conf, *new;
4183	struct nfs4_client	*unconf = NULL;
4184	__be32 			status;
4185	struct nfsd_net		*nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4186
4187	new = create_client(clname, rqstp, &clverifier);
4188	if (new == NULL)
4189		return nfserr_jukebox;
4190	spin_lock(&nn->client_lock);
4191	conf = find_confirmed_client_by_name(&clname, nn);
4192	if (conf && client_has_state(conf)) {
4193		status = nfserr_clid_inuse;
4194		if (clp_used_exchangeid(conf))
4195			goto out;
4196		if (!same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
4197			trace_nfsd_clid_cred_mismatch(conf, rqstp);
4198			goto out;
4199		}
4200	}
4201	unconf = find_unconfirmed_client_by_name(&clname, nn);
4202	if (unconf)
4203		unhash_client_locked(unconf);
4204	if (conf) {
4205		if (same_verf(&conf->cl_verifier, &clverifier)) {
4206			copy_clid(new, conf);
4207			gen_confirm(new, nn);
4208		} else
4209			trace_nfsd_clid_verf_mismatch(conf, rqstp,
4210						      &clverifier);
4211	} else
4212		trace_nfsd_clid_fresh(new);
4213	new->cl_minorversion = 0;
4214	gen_callback(new, setclid, rqstp);
4215	add_to_unconfirmed(new);
4216	setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
4217	setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
4218	memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
4219	new = NULL;
4220	status = nfs_ok;
4221out:
4222	spin_unlock(&nn->client_lock);
4223	if (new)
4224		free_client(new);
4225	if (unconf) {
4226		trace_nfsd_clid_expire_unconf(&unconf->cl_clientid);
4227		expire_client(unconf);
4228	}
4229	return status;
4230}
4231
4232__be32
4233nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
4234			struct nfsd4_compound_state *cstate,
4235			union nfsd4_op_u *u)
4236{
4237	struct nfsd4_setclientid_confirm *setclientid_confirm =
4238			&u->setclientid_confirm;
4239	struct nfs4_client *conf, *unconf;
4240	struct nfs4_client *old = NULL;
4241	nfs4_verifier confirm = setclientid_confirm->sc_confirm; 
4242	clientid_t * clid = &setclientid_confirm->sc_clientid;
4243	__be32 status;
4244	struct nfsd_net	*nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
4245
4246	if (STALE_CLIENTID(clid, nn))
4247		return nfserr_stale_clientid;
4248
4249	spin_lock(&nn->client_lock);
4250	conf = find_confirmed_client(clid, false, nn);
4251	unconf = find_unconfirmed_client(clid, false, nn);
4252	/*
4253	 * We try hard to give out unique clientid's, so if we get an
4254	 * attempt to confirm the same clientid with a different cred,
4255	 * the client may be buggy; this should never happen.
4256	 *
4257	 * Nevertheless, RFC 7530 recommends INUSE for this case:
4258	 */
4259	status = nfserr_clid_inuse;
4260	if (unconf && !same_creds(&unconf->cl_cred, &rqstp->rq_cred)) {
4261		trace_nfsd_clid_cred_mismatch(unconf, rqstp);
4262		goto out;
4263	}
4264	if (conf && !same_creds(&conf->cl_cred, &rqstp->rq_cred)) {
4265		trace_nfsd_clid_cred_mismatch(conf, rqstp);
4266		goto out;
4267	}
4268	if (!unconf || !same_verf(&confirm, &unconf->cl_confirm)) {
4269		if (conf && same_verf(&confirm, &conf->cl_confirm)) {
4270			status = nfs_ok;
4271		} else
4272			status = nfserr_stale_clientid;
4273		goto out;
4274	}
4275	status = nfs_ok;
4276	if (conf) {
4277		old = unconf;
4278		unhash_client_locked(old);
4279		nfsd4_change_callback(conf, &unconf->cl_cb_conn);
4280	} else {
4281		old = find_confirmed_client_by_name(&unconf->cl_name, nn);
4282		if (old) {
4283			status = nfserr_clid_inuse;
4284			if (client_has_state(old)
4285					&& !same_creds(&unconf->cl_cred,
4286							&old->cl_cred)) {
4287				old = NULL;
4288				goto out;
4289			}
4290			status = mark_client_expired_locked(old);
4291			if (status) {
4292				old = NULL;
4293				goto out;
4294			}
4295			trace_nfsd_clid_replaced(&old->cl_clientid);
4296		}
4297		move_to_confirmed(unconf);
4298		conf = unconf;
4299	}
4300	get_client_locked(conf);
4301	spin_unlock(&nn->client_lock);
4302	if (conf == unconf)
4303		fsnotify_dentry(conf->cl_nfsd_info_dentry, FS_MODIFY);
4304	nfsd4_probe_callback(conf);
4305	spin_lock(&nn->client_lock);
4306	put_client_renew_locked(conf);
4307out:
4308	spin_unlock(&nn->client_lock);
4309	if (old)
4310		expire_client(old);
4311	return status;
4312}
4313
4314static struct nfs4_file *nfsd4_alloc_file(void)
4315{
4316	return kmem_cache_alloc(file_slab, GFP_KERNEL);
4317}
4318
4319/* OPEN Share state helper functions */
4320
4321static void nfsd4_file_init(const struct svc_fh *fh, struct nfs4_file *fp)
4322{
 
 
4323	refcount_set(&fp->fi_ref, 1);
4324	spin_lock_init(&fp->fi_lock);
4325	INIT_LIST_HEAD(&fp->fi_stateids);
4326	INIT_LIST_HEAD(&fp->fi_delegations);
4327	INIT_LIST_HEAD(&fp->fi_clnt_odstate);
4328	fh_copy_shallow(&fp->fi_fhandle, &fh->fh_handle);
4329	fp->fi_deleg_file = NULL;
4330	fp->fi_had_conflict = false;
4331	fp->fi_share_deny = 0;
4332	memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
4333	memset(fp->fi_access, 0, sizeof(fp->fi_access));
4334	fp->fi_aliased = false;
4335	fp->fi_inode = d_inode(fh->fh_dentry);
4336#ifdef CONFIG_NFSD_PNFS
4337	INIT_LIST_HEAD(&fp->fi_lo_states);
4338	atomic_set(&fp->fi_lo_recalls, 0);
4339#endif
 
4340}
4341
4342void
4343nfsd4_free_slabs(void)
4344{
4345	kmem_cache_destroy(client_slab);
4346	kmem_cache_destroy(openowner_slab);
4347	kmem_cache_destroy(lockowner_slab);
4348	kmem_cache_destroy(file_slab);
4349	kmem_cache_destroy(stateid_slab);
4350	kmem_cache_destroy(deleg_slab);
4351	kmem_cache_destroy(odstate_slab);
4352}
4353
4354int
4355nfsd4_init_slabs(void)
4356{
4357	client_slab = kmem_cache_create("nfsd4_clients",
4358			sizeof(struct nfs4_client), 0, 0, NULL);
4359	if (client_slab == NULL)
4360		goto out;
4361	openowner_slab = kmem_cache_create("nfsd4_openowners",
4362			sizeof(struct nfs4_openowner), 0, 0, NULL);
4363	if (openowner_slab == NULL)
4364		goto out_free_client_slab;
4365	lockowner_slab = kmem_cache_create("nfsd4_lockowners",
4366			sizeof(struct nfs4_lockowner), 0, 0, NULL);
4367	if (lockowner_slab == NULL)
4368		goto out_free_openowner_slab;
4369	file_slab = kmem_cache_create("nfsd4_files",
4370			sizeof(struct nfs4_file), 0, 0, NULL);
4371	if (file_slab == NULL)
4372		goto out_free_lockowner_slab;
4373	stateid_slab = kmem_cache_create("nfsd4_stateids",
4374			sizeof(struct nfs4_ol_stateid), 0, 0, NULL);
4375	if (stateid_slab == NULL)
4376		goto out_free_file_slab;
4377	deleg_slab = kmem_cache_create("nfsd4_delegations",
4378			sizeof(struct nfs4_delegation), 0, 0, NULL);
4379	if (deleg_slab == NULL)
4380		goto out_free_stateid_slab;
4381	odstate_slab = kmem_cache_create("nfsd4_odstate",
4382			sizeof(struct nfs4_clnt_odstate), 0, 0, NULL);
4383	if (odstate_slab == NULL)
4384		goto out_free_deleg_slab;
4385	return 0;
4386
4387out_free_deleg_slab:
4388	kmem_cache_destroy(deleg_slab);
4389out_free_stateid_slab:
4390	kmem_cache_destroy(stateid_slab);
4391out_free_file_slab:
4392	kmem_cache_destroy(file_slab);
4393out_free_lockowner_slab:
4394	kmem_cache_destroy(lockowner_slab);
4395out_free_openowner_slab:
4396	kmem_cache_destroy(openowner_slab);
4397out_free_client_slab:
4398	kmem_cache_destroy(client_slab);
4399out:
4400	return -ENOMEM;
4401}
4402
4403static unsigned long
4404nfsd4_state_shrinker_count(struct shrinker *shrink, struct shrink_control *sc)
4405{
4406	int count;
4407	struct nfsd_net *nn = container_of(shrink,
4408			struct nfsd_net, nfsd_client_shrinker);
4409
4410	count = atomic_read(&nn->nfsd_courtesy_clients);
4411	if (!count)
4412		count = atomic_long_read(&num_delegations);
4413	if (count)
4414		queue_work(laundry_wq, &nn->nfsd_shrinker_work);
4415	return (unsigned long)count;
4416}
4417
4418static unsigned long
4419nfsd4_state_shrinker_scan(struct shrinker *shrink, struct shrink_control *sc)
4420{
4421	return SHRINK_STOP;
4422}
4423
4424void
4425nfsd4_init_leases_net(struct nfsd_net *nn)
4426{
4427	struct sysinfo si;
4428	u64 max_clients;
4429
4430	nn->nfsd4_lease = 90;	/* default lease time */
4431	nn->nfsd4_grace = 90;
4432	nn->somebody_reclaimed = false;
4433	nn->track_reclaim_completes = false;
4434	nn->clverifier_counter = get_random_u32();
4435	nn->clientid_base = get_random_u32();
4436	nn->clientid_counter = nn->clientid_base + 1;
4437	nn->s2s_cp_cl_id = nn->clientid_counter++;
4438
4439	atomic_set(&nn->nfs4_client_count, 0);
4440	si_meminfo(&si);
4441	max_clients = (u64)si.totalram * si.mem_unit / (1024 * 1024 * 1024);
4442	max_clients *= NFS4_CLIENTS_PER_GB;
4443	nn->nfs4_max_clients = max_t(int, max_clients, NFS4_CLIENTS_PER_GB);
4444
4445	atomic_set(&nn->nfsd_courtesy_clients, 0);
4446}
4447
4448static void init_nfs4_replay(struct nfs4_replay *rp)
4449{
4450	rp->rp_status = nfserr_serverfault;
4451	rp->rp_buflen = 0;
4452	rp->rp_buf = rp->rp_ibuf;
4453	mutex_init(&rp->rp_mutex);
4454}
4455
4456static void nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
4457		struct nfs4_stateowner *so)
4458{
4459	if (!nfsd4_has_session(cstate)) {
4460		mutex_lock(&so->so_replay.rp_mutex);
4461		cstate->replay_owner = nfs4_get_stateowner(so);
4462	}
4463}
4464
4465void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
4466{
4467	struct nfs4_stateowner *so = cstate->replay_owner;
4468
4469	if (so != NULL) {
4470		cstate->replay_owner = NULL;
4471		mutex_unlock(&so->so_replay.rp_mutex);
4472		nfs4_put_stateowner(so);
4473	}
4474}
4475
4476static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
4477{
4478	struct nfs4_stateowner *sop;
4479
4480	sop = kmem_cache_alloc(slab, GFP_KERNEL);
4481	if (!sop)
4482		return NULL;
4483
4484	xdr_netobj_dup(&sop->so_owner, owner, GFP_KERNEL);
4485	if (!sop->so_owner.data) {
4486		kmem_cache_free(slab, sop);
4487		return NULL;
4488	}
4489
4490	INIT_LIST_HEAD(&sop->so_stateids);
4491	sop->so_client = clp;
4492	init_nfs4_replay(&sop->so_replay);
4493	atomic_set(&sop->so_count, 1);
4494	return sop;
4495}
4496
4497static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
4498{
4499	lockdep_assert_held(&clp->cl_lock);
4500
4501	list_add(&oo->oo_owner.so_strhash,
4502		 &clp->cl_ownerstr_hashtbl[strhashval]);
4503	list_add(&oo->oo_perclient, &clp->cl_openowners);
4504}
4505
4506static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
4507{
4508	unhash_openowner_locked(openowner(so));
4509}
4510
4511static void nfs4_free_openowner(struct nfs4_stateowner *so)
4512{
4513	struct nfs4_openowner *oo = openowner(so);
4514
4515	kmem_cache_free(openowner_slab, oo);
4516}
4517
4518static const struct nfs4_stateowner_operations openowner_ops = {
4519	.so_unhash =	nfs4_unhash_openowner,
4520	.so_free =	nfs4_free_openowner,
4521};
4522
4523static struct nfs4_ol_stateid *
4524nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
4525{
4526	struct nfs4_ol_stateid *local, *ret = NULL;
4527	struct nfs4_openowner *oo = open->op_openowner;
4528
4529	lockdep_assert_held(&fp->fi_lock);
4530
4531	list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
4532		/* ignore lock owners */
4533		if (local->st_stateowner->so_is_open_owner == 0)
4534			continue;
4535		if (local->st_stateowner != &oo->oo_owner)
4536			continue;
4537		if (local->st_stid.sc_type == NFS4_OPEN_STID) {
4538			ret = local;
4539			refcount_inc(&ret->st_stid.sc_count);
4540			break;
4541		}
4542	}
4543	return ret;
4544}
4545
4546static __be32
4547nfsd4_verify_open_stid(struct nfs4_stid *s)
4548{
4549	__be32 ret = nfs_ok;
4550
4551	switch (s->sc_type) {
4552	default:
4553		break;
4554	case 0:
4555	case NFS4_CLOSED_STID:
4556	case NFS4_CLOSED_DELEG_STID:
4557		ret = nfserr_bad_stateid;
4558		break;
4559	case NFS4_REVOKED_DELEG_STID:
4560		ret = nfserr_deleg_revoked;
4561	}
4562	return ret;
4563}
4564
4565/* Lock the stateid st_mutex, and deal with races with CLOSE */
4566static __be32
4567nfsd4_lock_ol_stateid(struct nfs4_ol_stateid *stp)
4568{
4569	__be32 ret;
4570
4571	mutex_lock_nested(&stp->st_mutex, LOCK_STATEID_MUTEX);
4572	ret = nfsd4_verify_open_stid(&stp->st_stid);
4573	if (ret != nfs_ok)
4574		mutex_unlock(&stp->st_mutex);
4575	return ret;
4576}
4577
4578static struct nfs4_ol_stateid *
4579nfsd4_find_and_lock_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
4580{
4581	struct nfs4_ol_stateid *stp;
4582	for (;;) {
4583		spin_lock(&fp->fi_lock);
4584		stp = nfsd4_find_existing_open(fp, open);
4585		spin_unlock(&fp->fi_lock);
4586		if (!stp || nfsd4_lock_ol_stateid(stp) == nfs_ok)
4587			break;
4588		nfs4_put_stid(&stp->st_stid);
4589	}
4590	return stp;
4591}
4592
4593static struct nfs4_openowner *
4594alloc_init_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
4595			   struct nfsd4_compound_state *cstate)
4596{
4597	struct nfs4_client *clp = cstate->clp;
4598	struct nfs4_openowner *oo, *ret;
4599
4600	oo = alloc_stateowner(openowner_slab, &open->op_owner, clp);
4601	if (!oo)
4602		return NULL;
4603	oo->oo_owner.so_ops = &openowner_ops;
4604	oo->oo_owner.so_is_open_owner = 1;
4605	oo->oo_owner.so_seqid = open->op_seqid;
4606	oo->oo_flags = 0;
4607	if (nfsd4_has_session(cstate))
4608		oo->oo_flags |= NFS4_OO_CONFIRMED;
4609	oo->oo_time = 0;
4610	oo->oo_last_closed_stid = NULL;
4611	INIT_LIST_HEAD(&oo->oo_close_lru);
4612	spin_lock(&clp->cl_lock);
4613	ret = find_openstateowner_str_locked(strhashval, open, clp);
4614	if (ret == NULL) {
4615		hash_openowner(oo, clp, strhashval);
4616		ret = oo;
4617	} else
4618		nfs4_free_stateowner(&oo->oo_owner);
4619
4620	spin_unlock(&clp->cl_lock);
4621	return ret;
4622}
4623
4624static struct nfs4_ol_stateid *
4625init_open_stateid(struct nfs4_file *fp, struct nfsd4_open *open)
4626{
4627
4628	struct nfs4_openowner *oo = open->op_openowner;
4629	struct nfs4_ol_stateid *retstp = NULL;
4630	struct nfs4_ol_stateid *stp;
4631
4632	stp = open->op_stp;
4633	/* We are moving these outside of the spinlocks to avoid the warnings */
4634	mutex_init(&stp->st_mutex);
4635	mutex_lock_nested(&stp->st_mutex, OPEN_STATEID_MUTEX);
4636
4637retry:
4638	spin_lock(&oo->oo_owner.so_client->cl_lock);
4639	spin_lock(&fp->fi_lock);
4640
4641	retstp = nfsd4_find_existing_open(fp, open);
4642	if (retstp)
4643		goto out_unlock;
4644
4645	open->op_stp = NULL;
4646	refcount_inc(&stp->st_stid.sc_count);
4647	stp->st_stid.sc_type = NFS4_OPEN_STID;
4648	INIT_LIST_HEAD(&stp->st_locks);
4649	stp->st_stateowner = nfs4_get_stateowner(&oo->oo_owner);
4650	get_nfs4_file(fp);
4651	stp->st_stid.sc_file = fp;
4652	stp->st_access_bmap = 0;
4653	stp->st_deny_bmap = 0;
4654	stp->st_openstp = NULL;
4655	list_add(&stp->st_perstateowner, &oo->oo_owner.so_stateids);
4656	list_add(&stp->st_perfile, &fp->fi_stateids);
4657
4658out_unlock:
4659	spin_unlock(&fp->fi_lock);
4660	spin_unlock(&oo->oo_owner.so_client->cl_lock);
4661	if (retstp) {
4662		/* Handle races with CLOSE */
4663		if (nfsd4_lock_ol_stateid(retstp) != nfs_ok) {
4664			nfs4_put_stid(&retstp->st_stid);
4665			goto retry;
4666		}
4667		/* To keep mutex tracking happy */
4668		mutex_unlock(&stp->st_mutex);
4669		stp = retstp;
4670	}
4671	return stp;
4672}
4673
4674/*
4675 * In the 4.0 case we need to keep the owners around a little while to handle
4676 * CLOSE replay. We still do need to release any file access that is held by
4677 * them before returning however.
4678 */
4679static void
4680move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
4681{
4682	struct nfs4_ol_stateid *last;
4683	struct nfs4_openowner *oo = openowner(s->st_stateowner);
4684	struct nfsd_net *nn = net_generic(s->st_stid.sc_client->net,
4685						nfsd_net_id);
4686
4687	dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
4688
4689	/*
4690	 * We know that we hold one reference via nfsd4_close, and another
4691	 * "persistent" reference for the client. If the refcount is higher
4692	 * than 2, then there are still calls in progress that are using this
4693	 * stateid. We can't put the sc_file reference until they are finished.
4694	 * Wait for the refcount to drop to 2. Since it has been unhashed,
4695	 * there should be no danger of the refcount going back up again at
4696	 * this point.
4697	 */
4698	wait_event(close_wq, refcount_read(&s->st_stid.sc_count) == 2);
4699
4700	release_all_access(s);
4701	if (s->st_stid.sc_file) {
4702		put_nfs4_file(s->st_stid.sc_file);
4703		s->st_stid.sc_file = NULL;
4704	}
4705
4706	spin_lock(&nn->client_lock);
4707	last = oo->oo_last_closed_stid;
4708	oo->oo_last_closed_stid = s;
4709	list_move_tail(&oo->oo_close_lru, &nn->close_lru);
4710	oo->oo_time = ktime_get_boottime_seconds();
4711	spin_unlock(&nn->client_lock);
4712	if (last)
4713		nfs4_put_stid(&last->st_stid);
4714}
4715
4716static noinline_for_stack struct nfs4_file *
4717nfsd4_file_hash_lookup(const struct svc_fh *fhp)
 
4718{
4719	struct inode *inode = d_inode(fhp->fh_dentry);
4720	struct rhlist_head *tmp, *list;
4721	struct nfs4_file *fi;
4722
4723	rcu_read_lock();
4724	list = rhltable_lookup(&nfs4_file_rhltable, &inode,
4725			       nfs4_file_rhash_params);
4726	rhl_for_each_entry_rcu(fi, tmp, list, fi_rlist) {
4727		if (fh_match(&fi->fi_fhandle, &fhp->fh_handle)) {
4728			if (refcount_inc_not_zero(&fi->fi_ref)) {
4729				rcu_read_unlock();
4730				return fi;
4731			}
4732		}
4733	}
4734	rcu_read_unlock();
4735	return NULL;
4736}
4737
4738/*
4739 * On hash insertion, identify entries with the same inode but
4740 * distinct filehandles. They will all be on the list returned
4741 * by rhltable_lookup().
4742 *
4743 * inode->i_lock prevents racing insertions from adding an entry
4744 * for the same inode/fhp pair twice.
4745 */
4746static noinline_for_stack struct nfs4_file *
4747nfsd4_file_hash_insert(struct nfs4_file *new, const struct svc_fh *fhp)
4748{
4749	struct inode *inode = d_inode(fhp->fh_dentry);
4750	struct rhlist_head *tmp, *list;
4751	struct nfs4_file *ret = NULL;
4752	bool alias_found = false;
4753	struct nfs4_file *fi;
4754	int err;
4755
4756	rcu_read_lock();
4757	spin_lock(&inode->i_lock);
4758
4759	list = rhltable_lookup(&nfs4_file_rhltable, &inode,
4760			       nfs4_file_rhash_params);
4761	rhl_for_each_entry_rcu(fi, tmp, list, fi_rlist) {
4762		if (fh_match(&fi->fi_fhandle, &fhp->fh_handle)) {
4763			if (refcount_inc_not_zero(&fi->fi_ref))
4764				ret = fi;
4765		} else
4766			fi->fi_aliased = alias_found = true;
 
 
4767	}
4768	if (ret)
4769		goto out_unlock;
4770
4771	nfsd4_file_init(fhp, new);
4772	err = rhltable_insert(&nfs4_file_rhltable, &new->fi_rlist,
4773			      nfs4_file_rhash_params);
4774	if (err)
4775		goto out_unlock;
4776
4777	new->fi_aliased = alias_found;
4778	ret = new;
 
 
4779
4780out_unlock:
4781	spin_unlock(&inode->i_lock);
4782	rcu_read_unlock();
4783	return ret;
4784}
4785
4786static noinline_for_stack void nfsd4_file_hash_remove(struct nfs4_file *fi)
 
4787{
4788	rhltable_remove(&nfs4_file_rhltable, &fi->fi_rlist,
4789			nfs4_file_rhash_params);
 
 
 
 
 
 
 
 
4790}
4791
4792/*
4793 * Called to check deny when READ with all zero stateid or
4794 * WRITE with all zero or all one stateid
4795 */
4796static __be32
4797nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
4798{
4799	struct nfs4_file *fp;
4800	__be32 ret = nfs_ok;
4801
4802	fp = nfsd4_file_hash_lookup(current_fh);
4803	if (!fp)
4804		return ret;
4805
4806	/* Check for conflicting share reservations */
4807	spin_lock(&fp->fi_lock);
4808	if (fp->fi_share_deny & deny_type)
4809		ret = nfserr_locked;
4810	spin_unlock(&fp->fi_lock);
4811	put_nfs4_file(fp);
4812	return ret;
4813}
4814
4815static bool nfsd4_deleg_present(const struct inode *inode)
4816{
4817	struct file_lock_context *ctx = locks_inode_context(inode);
4818
4819	return ctx && !list_empty_careful(&ctx->flc_lease);
4820}
4821
4822/**
4823 * nfsd_wait_for_delegreturn - wait for delegations to be returned
4824 * @rqstp: the RPC transaction being executed
4825 * @inode: in-core inode of the file being waited for
4826 *
4827 * The timeout prevents deadlock if all nfsd threads happen to be
4828 * tied up waiting for returning delegations.
4829 *
4830 * Return values:
4831 *   %true: delegation was returned
4832 *   %false: timed out waiting for delegreturn
4833 */
4834bool nfsd_wait_for_delegreturn(struct svc_rqst *rqstp, struct inode *inode)
4835{
4836	long __maybe_unused timeo;
4837
4838	timeo = wait_var_event_timeout(inode, !nfsd4_deleg_present(inode),
4839				       NFSD_DELEGRETURN_TIMEOUT);
4840	trace_nfsd_delegret_wakeup(rqstp, inode, timeo);
4841	return timeo > 0;
4842}
4843
4844static void nfsd4_cb_recall_prepare(struct nfsd4_callback *cb)
4845{
4846	struct nfs4_delegation *dp = cb_to_delegation(cb);
4847	struct nfsd_net *nn = net_generic(dp->dl_stid.sc_client->net,
4848					  nfsd_net_id);
4849
4850	block_delegations(&dp->dl_stid.sc_file->fi_fhandle);
4851
4852	/*
4853	 * We can't do this in nfsd_break_deleg_cb because it is
4854	 * already holding inode->i_lock.
4855	 *
4856	 * If the dl_time != 0, then we know that it has already been
4857	 * queued for a lease break. Don't queue it again.
4858	 */
4859	spin_lock(&state_lock);
4860	if (delegation_hashed(dp) && dp->dl_time == 0) {
4861		dp->dl_time = ktime_get_boottime_seconds();
4862		list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru);
4863	}
4864	spin_unlock(&state_lock);
4865}
4866
4867static int nfsd4_cb_recall_done(struct nfsd4_callback *cb,
4868		struct rpc_task *task)
4869{
4870	struct nfs4_delegation *dp = cb_to_delegation(cb);
4871
4872	trace_nfsd_cb_recall_done(&dp->dl_stid.sc_stateid, task);
4873
4874	if (dp->dl_stid.sc_type == NFS4_CLOSED_DELEG_STID ||
4875	    dp->dl_stid.sc_type == NFS4_REVOKED_DELEG_STID)
4876	        return 1;
4877
4878	switch (task->tk_status) {
4879	case 0:
4880		return 1;
4881	case -NFS4ERR_DELAY:
4882		rpc_delay(task, 2 * HZ);
4883		return 0;
4884	case -EBADHANDLE:
4885	case -NFS4ERR_BAD_STATEID:
4886		/*
4887		 * Race: client probably got cb_recall before open reply
4888		 * granting delegation.
4889		 */
4890		if (dp->dl_retries--) {
4891			rpc_delay(task, 2 * HZ);
4892			return 0;
4893		}
4894		fallthrough;
4895	default:
4896		return 1;
4897	}
4898}
4899
4900static void nfsd4_cb_recall_release(struct nfsd4_callback *cb)
4901{
4902	struct nfs4_delegation *dp = cb_to_delegation(cb);
4903
4904	nfs4_put_stid(&dp->dl_stid);
4905}
4906
4907static const struct nfsd4_callback_ops nfsd4_cb_recall_ops = {
4908	.prepare	= nfsd4_cb_recall_prepare,
4909	.done		= nfsd4_cb_recall_done,
4910	.release	= nfsd4_cb_recall_release,
4911};
4912
4913static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
4914{
4915	/*
4916	 * We're assuming the state code never drops its reference
4917	 * without first removing the lease.  Since we're in this lease
4918	 * callback (and since the lease code is serialized by the
4919	 * flc_lock) we know the server hasn't removed the lease yet, and
4920	 * we know it's safe to take a reference.
4921	 */
4922	refcount_inc(&dp->dl_stid.sc_count);
4923	WARN_ON_ONCE(!nfsd4_run_cb(&dp->dl_recall));
4924}
4925
4926/* Called from break_lease() with flc_lock held. */
4927static bool
4928nfsd_break_deleg_cb(struct file_lock *fl)
4929{
 
4930	struct nfs4_delegation *dp = (struct nfs4_delegation *)fl->fl_owner;
4931	struct nfs4_file *fp = dp->dl_stid.sc_file;
4932	struct nfs4_client *clp = dp->dl_stid.sc_client;
4933	struct nfsd_net *nn;
4934
4935	trace_nfsd_cb_recall(&dp->dl_stid);
4936
4937	dp->dl_recalled = true;
4938	atomic_inc(&clp->cl_delegs_in_recall);
4939	if (try_to_expire_client(clp)) {
4940		nn = net_generic(clp->net, nfsd_net_id);
4941		mod_delayed_work(laundry_wq, &nn->laundromat_work, 0);
4942	}
4943
4944	/*
4945	 * We don't want the locks code to timeout the lease for us;
4946	 * we'll remove it ourself if a delegation isn't returned
4947	 * in time:
4948	 */
4949	fl->fl_break_time = 0;
4950
4951	spin_lock(&fp->fi_lock);
4952	fp->fi_had_conflict = true;
4953	nfsd_break_one_deleg(dp);
4954	spin_unlock(&fp->fi_lock);
4955	return false;
4956}
4957
4958/**
4959 * nfsd_breaker_owns_lease - Check if lease conflict was resolved
4960 * @fl: Lock state to check
4961 *
4962 * Return values:
4963 *   %true: Lease conflict was resolved
4964 *   %false: Lease conflict was not resolved.
4965 */
4966static bool nfsd_breaker_owns_lease(struct file_lock *fl)
4967{
4968	struct nfs4_delegation *dl = fl->fl_owner;
4969	struct svc_rqst *rqst;
4970	struct nfs4_client *clp;
4971
4972	if (!i_am_nfsd())
4973		return false;
4974	rqst = kthread_data(current);
4975	/* Note rq_prog == NFS_ACL_PROGRAM is also possible: */
4976	if (rqst->rq_prog != NFS_PROGRAM || rqst->rq_vers < 4)
4977		return false;
4978	clp = *(rqst->rq_lease_breaker);
4979	return dl->dl_stid.sc_client == clp;
4980}
4981
4982static int
4983nfsd_change_deleg_cb(struct file_lock *onlist, int arg,
4984		     struct list_head *dispose)
4985{
4986	struct nfs4_delegation *dp = (struct nfs4_delegation *)onlist->fl_owner;
4987	struct nfs4_client *clp = dp->dl_stid.sc_client;
4988
4989	if (arg & F_UNLCK) {
4990		if (dp->dl_recalled)
4991			atomic_dec(&clp->cl_delegs_in_recall);
4992		return lease_modify(onlist, arg, dispose);
4993	} else
4994		return -EAGAIN;
4995}
4996
4997static const struct lock_manager_operations nfsd_lease_mng_ops = {
4998	.lm_breaker_owns_lease = nfsd_breaker_owns_lease,
4999	.lm_break = nfsd_break_deleg_cb,
5000	.lm_change = nfsd_change_deleg_cb,
5001};
5002
5003static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
5004{
5005	if (nfsd4_has_session(cstate))
5006		return nfs_ok;
5007	if (seqid == so->so_seqid - 1)
5008		return nfserr_replay_me;
5009	if (seqid == so->so_seqid)
5010		return nfs_ok;
5011	return nfserr_bad_seqid;
5012}
5013
5014static struct nfs4_client *lookup_clientid(clientid_t *clid, bool sessions,
5015						struct nfsd_net *nn)
5016{
5017	struct nfs4_client *found;
5018
5019	spin_lock(&nn->client_lock);
5020	found = find_confirmed_client(clid, sessions, nn);
5021	if (found)
5022		atomic_inc(&found->cl_rpc_users);
5023	spin_unlock(&nn->client_lock);
5024	return found;
5025}
5026
5027static __be32 set_client(clientid_t *clid,
5028		struct nfsd4_compound_state *cstate,
5029		struct nfsd_net *nn)
5030{
5031	if (cstate->clp) {
5032		if (!same_clid(&cstate->clp->cl_clientid, clid))
5033			return nfserr_stale_clientid;
5034		return nfs_ok;
5035	}
5036	if (STALE_CLIENTID(clid, nn))
5037		return nfserr_stale_clientid;
5038	/*
5039	 * We're in the 4.0 case (otherwise the SEQUENCE op would have
5040	 * set cstate->clp), so session = false:
5041	 */
5042	cstate->clp = lookup_clientid(clid, false, nn);
5043	if (!cstate->clp)
5044		return nfserr_expired;
5045	return nfs_ok;
5046}
5047
5048__be32
5049nfsd4_process_open1(struct nfsd4_compound_state *cstate,
5050		    struct nfsd4_open *open, struct nfsd_net *nn)
5051{
5052	clientid_t *clientid = &open->op_clientid;
5053	struct nfs4_client *clp = NULL;
5054	unsigned int strhashval;
5055	struct nfs4_openowner *oo = NULL;
5056	__be32 status;
5057
5058	/*
5059	 * In case we need it later, after we've already created the
5060	 * file and don't want to risk a further failure:
5061	 */
5062	open->op_file = nfsd4_alloc_file();
5063	if (open->op_file == NULL)
5064		return nfserr_jukebox;
5065
5066	status = set_client(clientid, cstate, nn);
5067	if (status)
5068		return status;
5069	clp = cstate->clp;
5070
5071	strhashval = ownerstr_hashval(&open->op_owner);
5072	oo = find_openstateowner_str(strhashval, open, clp);
5073	open->op_openowner = oo;
5074	if (!oo) {
5075		goto new_owner;
5076	}
5077	if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
5078		/* Replace unconfirmed owners without checking for replay. */
5079		release_openowner(oo);
5080		open->op_openowner = NULL;
5081		goto new_owner;
5082	}
5083	status = nfsd4_check_seqid(cstate, &oo->oo_owner, open->op_seqid);
5084	if (status)
5085		return status;
5086	goto alloc_stateid;
5087new_owner:
5088	oo = alloc_init_open_stateowner(strhashval, open, cstate);
5089	if (oo == NULL)
5090		return nfserr_jukebox;
5091	open->op_openowner = oo;
5092alloc_stateid:
5093	open->op_stp = nfs4_alloc_open_stateid(clp);
5094	if (!open->op_stp)
5095		return nfserr_jukebox;
5096
5097	if (nfsd4_has_session(cstate) &&
5098	    (cstate->current_fh.fh_export->ex_flags & NFSEXP_PNFS)) {
5099		open->op_odstate = alloc_clnt_odstate(clp);
5100		if (!open->op_odstate)
5101			return nfserr_jukebox;
5102	}
5103
5104	return nfs_ok;
5105}
5106
5107static inline __be32
5108nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
5109{
5110	if ((flags & WR_STATE) && (dp->dl_type == NFS4_OPEN_DELEGATE_READ))
5111		return nfserr_openmode;
5112	else
5113		return nfs_ok;
5114}
5115
5116static int share_access_to_flags(u32 share_access)
5117{
5118	return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
5119}
5120
5121static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl, stateid_t *s)
5122{
5123	struct nfs4_stid *ret;
5124
5125	ret = find_stateid_by_type(cl, s,
5126				NFS4_DELEG_STID|NFS4_REVOKED_DELEG_STID);
5127	if (!ret)
5128		return NULL;
5129	return delegstateid(ret);
5130}
5131
5132static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
5133{
5134	return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
5135	       open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
5136}
5137
5138static __be32
5139nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
5140		struct nfs4_delegation **dp)
5141{
5142	int flags;
5143	__be32 status = nfserr_bad_stateid;
5144	struct nfs4_delegation *deleg;
5145
5146	deleg = find_deleg_stateid(cl, &open->op_delegate_stateid);
5147	if (deleg == NULL)
5148		goto out;
5149	if (deleg->dl_stid.sc_type == NFS4_REVOKED_DELEG_STID) {
5150		nfs4_put_stid(&deleg->dl_stid);
5151		if (cl->cl_minorversion)
5152			status = nfserr_deleg_revoked;
5153		goto out;
5154	}
5155	flags = share_access_to_flags(open->op_share_access);
5156	status = nfs4_check_delegmode(deleg, flags);
5157	if (status) {
5158		nfs4_put_stid(&deleg->dl_stid);
5159		goto out;
5160	}
5161	*dp = deleg;
5162out:
5163	if (!nfsd4_is_deleg_cur(open))
5164		return nfs_ok;
5165	if (status)
5166		return status;
5167	open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
5168	return nfs_ok;
5169}
5170
5171static inline int nfs4_access_to_access(u32 nfs4_access)
5172{
5173	int flags = 0;
5174
5175	if (nfs4_access & NFS4_SHARE_ACCESS_READ)
5176		flags |= NFSD_MAY_READ;
5177	if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
5178		flags |= NFSD_MAY_WRITE;
5179	return flags;
5180}
5181
5182static inline __be32
5183nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
5184		struct nfsd4_open *open)
5185{
5186	struct iattr iattr = {
5187		.ia_valid = ATTR_SIZE,
5188		.ia_size = 0,
5189	};
5190	struct nfsd_attrs attrs = {
5191		.na_iattr	= &iattr,
5192	};
5193	if (!open->op_truncate)
5194		return 0;
5195	if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
5196		return nfserr_inval;
5197	return nfsd_setattr(rqstp, fh, &attrs, 0, (time64_t)0);
5198}
5199
5200static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
5201		struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
5202		struct nfsd4_open *open, bool new_stp)
5203{
5204	struct nfsd_file *nf = NULL;
5205	__be32 status;
5206	int oflag = nfs4_access_to_omode(open->op_share_access);
5207	int access = nfs4_access_to_access(open->op_share_access);
5208	unsigned char old_access_bmap, old_deny_bmap;
5209
5210	spin_lock(&fp->fi_lock);
5211
5212	/*
5213	 * Are we trying to set a deny mode that would conflict with
5214	 * current access?
5215	 */
5216	status = nfs4_file_check_deny(fp, open->op_share_deny);
5217	if (status != nfs_ok) {
5218		if (status != nfserr_share_denied) {
5219			spin_unlock(&fp->fi_lock);
5220			goto out;
5221		}
5222		if (nfs4_resolve_deny_conflicts_locked(fp, new_stp,
5223				stp, open->op_share_deny, false))
5224			status = nfserr_jukebox;
5225		spin_unlock(&fp->fi_lock);
5226		goto out;
5227	}
5228
5229	/* set access to the file */
5230	status = nfs4_file_get_access(fp, open->op_share_access);
5231	if (status != nfs_ok) {
5232		if (status != nfserr_share_denied) {
5233			spin_unlock(&fp->fi_lock);
5234			goto out;
5235		}
5236		if (nfs4_resolve_deny_conflicts_locked(fp, new_stp,
5237				stp, open->op_share_access, true))
5238			status = nfserr_jukebox;
5239		spin_unlock(&fp->fi_lock);
5240		goto out;
5241	}
5242
5243	/* Set access bits in stateid */
5244	old_access_bmap = stp->st_access_bmap;
5245	set_access(open->op_share_access, stp);
5246
5247	/* Set new deny mask */
5248	old_deny_bmap = stp->st_deny_bmap;
5249	set_deny(open->op_share_deny, stp);
5250	fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
5251
5252	if (!fp->fi_fds[oflag]) {
5253		spin_unlock(&fp->fi_lock);
5254
5255		status = nfsd_file_acquire_opened(rqstp, cur_fh, access,
5256						  open->op_filp, &nf);
5257		if (status != nfs_ok)
5258			goto out_put_access;
5259
5260		spin_lock(&fp->fi_lock);
5261		if (!fp->fi_fds[oflag]) {
5262			fp->fi_fds[oflag] = nf;
5263			nf = NULL;
5264		}
5265	}
5266	spin_unlock(&fp->fi_lock);
5267	if (nf)
5268		nfsd_file_put(nf);
5269
5270	status = nfserrno(nfsd_open_break_lease(cur_fh->fh_dentry->d_inode,
5271								access));
5272	if (status)
5273		goto out_put_access;
5274
5275	status = nfsd4_truncate(rqstp, cur_fh, open);
5276	if (status)
5277		goto out_put_access;
5278out:
5279	return status;
5280out_put_access:
5281	stp->st_access_bmap = old_access_bmap;
5282	nfs4_file_put_access(fp, open->op_share_access);
5283	reset_union_bmap_deny(bmap_to_share_mode(old_deny_bmap), stp);
5284	goto out;
5285}
5286
5287static __be32
5288nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp,
5289		struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
5290		struct nfsd4_open *open)
5291{
5292	__be32 status;
5293	unsigned char old_deny_bmap = stp->st_deny_bmap;
5294
5295	if (!test_access(open->op_share_access, stp))
5296		return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open, false);
5297
5298	/* test and set deny mode */
5299	spin_lock(&fp->fi_lock);
5300	status = nfs4_file_check_deny(fp, open->op_share_deny);
5301	if (status == nfs_ok) {
5302		if (status != nfserr_share_denied) {
5303			set_deny(open->op_share_deny, stp);
5304			fp->fi_share_deny |=
5305				(open->op_share_deny & NFS4_SHARE_DENY_BOTH);
5306		} else {
5307			if (nfs4_resolve_deny_conflicts_locked(fp, false,
5308					stp, open->op_share_deny, false))
5309				status = nfserr_jukebox;
5310		}
5311	}
5312	spin_unlock(&fp->fi_lock);
5313
5314	if (status != nfs_ok)
5315		return status;
5316
5317	status = nfsd4_truncate(rqstp, cur_fh, open);
5318	if (status != nfs_ok)
5319		reset_union_bmap_deny(old_deny_bmap, stp);
5320	return status;
5321}
5322
5323/* Should we give out recallable state?: */
5324static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
5325{
5326	if (clp->cl_cb_state == NFSD4_CB_UP)
5327		return true;
5328	/*
5329	 * In the sessions case, since we don't have to establish a
5330	 * separate connection for callbacks, we assume it's OK
5331	 * until we hear otherwise:
5332	 */
5333	return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
5334}
5335
5336static struct file_lock *nfs4_alloc_init_lease(struct nfs4_delegation *dp,
5337						int flag)
5338{
5339	struct file_lock *fl;
5340
5341	fl = locks_alloc_lock();
5342	if (!fl)
5343		return NULL;
5344	fl->fl_lmops = &nfsd_lease_mng_ops;
5345	fl->fl_flags = FL_DELEG;
5346	fl->fl_type = flag == NFS4_OPEN_DELEGATE_READ? F_RDLCK: F_WRLCK;
5347	fl->fl_end = OFFSET_MAX;
5348	fl->fl_owner = (fl_owner_t)dp;
5349	fl->fl_pid = current->tgid;
5350	fl->fl_file = dp->dl_stid.sc_file->fi_deleg_file->nf_file;
5351	return fl;
5352}
5353
5354static int nfsd4_check_conflicting_opens(struct nfs4_client *clp,
5355					 struct nfs4_file *fp)
5356{
5357	struct nfs4_ol_stateid *st;
5358	struct file *f = fp->fi_deleg_file->nf_file;
5359	struct inode *ino = locks_inode(f);
5360	int writes;
5361
5362	writes = atomic_read(&ino->i_writecount);
5363	if (!writes)
5364		return 0;
5365	/*
5366	 * There could be multiple filehandles (hence multiple
5367	 * nfs4_files) referencing this file, but that's not too
5368	 * common; let's just give up in that case rather than
5369	 * trying to go look up all the clients using that other
5370	 * nfs4_file as well:
5371	 */
5372	if (fp->fi_aliased)
5373		return -EAGAIN;
5374	/*
5375	 * If there's a close in progress, make sure that we see it
5376	 * clear any fi_fds[] entries before we see it decrement
5377	 * i_writecount:
5378	 */
5379	smp_mb__after_atomic();
5380
5381	if (fp->fi_fds[O_WRONLY])
5382		writes--;
5383	if (fp->fi_fds[O_RDWR])
5384		writes--;
5385	if (writes > 0)
5386		return -EAGAIN; /* There may be non-NFSv4 writers */
5387	/*
5388	 * It's possible there are non-NFSv4 write opens in progress,
5389	 * but if they haven't incremented i_writecount yet then they
5390	 * also haven't called break lease yet; so, they'll break this
5391	 * lease soon enough.  So, all that's left to check for is NFSv4
5392	 * opens:
5393	 */
5394	spin_lock(&fp->fi_lock);
5395	list_for_each_entry(st, &fp->fi_stateids, st_perfile) {
5396		if (st->st_openstp == NULL /* it's an open */ &&
5397		    access_permit_write(st) &&
5398		    st->st_stid.sc_client != clp) {
5399			spin_unlock(&fp->fi_lock);
5400			return -EAGAIN;
5401		}
5402	}
5403	spin_unlock(&fp->fi_lock);
5404	/*
5405	 * There's a small chance that we could be racing with another
5406	 * NFSv4 open.  However, any open that hasn't added itself to
5407	 * the fi_stateids list also hasn't called break_lease yet; so,
5408	 * they'll break this lease soon enough.
5409	 */
5410	return 0;
5411}
5412
5413/*
5414 * It's possible that between opening the dentry and setting the delegation,
5415 * that it has been renamed or unlinked. Redo the lookup to verify that this
5416 * hasn't happened.
5417 */
5418static int
5419nfsd4_verify_deleg_dentry(struct nfsd4_open *open, struct nfs4_file *fp,
5420			  struct svc_fh *parent)
5421{
5422	struct svc_export *exp;
5423	struct dentry *child;
5424	__be32 err;
5425
5426	err = nfsd_lookup_dentry(open->op_rqstp, parent,
5427				 open->op_fname, open->op_fnamelen,
5428				 &exp, &child);
5429
5430	if (err)
5431		return -EAGAIN;
5432
5433	exp_put(exp);
5434	dput(child);
5435	if (child != file_dentry(fp->fi_deleg_file->nf_file))
5436		return -EAGAIN;
5437
5438	return 0;
5439}
5440
5441static struct nfs4_delegation *
5442nfs4_set_delegation(struct nfsd4_open *open, struct nfs4_ol_stateid *stp,
5443		    struct svc_fh *parent)
5444{
5445	int status = 0;
5446	struct nfs4_client *clp = stp->st_stid.sc_client;
5447	struct nfs4_file *fp = stp->st_stid.sc_file;
5448	struct nfs4_clnt_odstate *odstate = stp->st_clnt_odstate;
5449	struct nfs4_delegation *dp;
5450	struct nfsd_file *nf;
5451	struct file_lock *fl;
5452
5453	/*
5454	 * The fi_had_conflict and nfs_get_existing_delegation checks
5455	 * here are just optimizations; we'll need to recheck them at
5456	 * the end:
5457	 */
5458	if (fp->fi_had_conflict)
5459		return ERR_PTR(-EAGAIN);
5460
5461	nf = find_readable_file(fp);
5462	if (!nf) {
5463		/*
5464		 * We probably could attempt another open and get a read
5465		 * delegation, but for now, don't bother until the
5466		 * client actually sends us one.
5467		 */
5468		return ERR_PTR(-EAGAIN);
5469	}
5470	spin_lock(&state_lock);
5471	spin_lock(&fp->fi_lock);
5472	if (nfs4_delegation_exists(clp, fp))
5473		status = -EAGAIN;
5474	else if (!fp->fi_deleg_file) {
5475		fp->fi_deleg_file = nf;
5476		/* increment early to prevent fi_deleg_file from being
5477		 * cleared */
5478		fp->fi_delegees = 1;
5479		nf = NULL;
5480	} else
5481		fp->fi_delegees++;
5482	spin_unlock(&fp->fi_lock);
5483	spin_unlock(&state_lock);
5484	if (nf)
5485		nfsd_file_put(nf);
5486	if (status)
5487		return ERR_PTR(status);
5488
5489	status = -ENOMEM;
5490	dp = alloc_init_deleg(clp, fp, odstate);
5491	if (!dp)
5492		goto out_delegees;
5493
5494	fl = nfs4_alloc_init_lease(dp, NFS4_OPEN_DELEGATE_READ);
5495	if (!fl)
5496		goto out_clnt_odstate;
5497
5498	status = vfs_setlease(fp->fi_deleg_file->nf_file, fl->fl_type, &fl, NULL);
5499	if (fl)
5500		locks_free_lock(fl);
5501	if (status)
5502		goto out_clnt_odstate;
5503
5504	if (parent) {
5505		status = nfsd4_verify_deleg_dentry(open, fp, parent);
5506		if (status)
5507			goto out_unlock;
5508	}
5509
5510	status = nfsd4_check_conflicting_opens(clp, fp);
5511	if (status)
5512		goto out_unlock;
5513
5514	spin_lock(&state_lock);
5515	spin_lock(&fp->fi_lock);
5516	if (fp->fi_had_conflict)
5517		status = -EAGAIN;
5518	else
5519		status = hash_delegation_locked(dp, fp);
5520	spin_unlock(&fp->fi_lock);
5521	spin_unlock(&state_lock);
5522
5523	if (status)
5524		goto out_unlock;
5525
5526	return dp;
5527out_unlock:
5528	vfs_setlease(fp->fi_deleg_file->nf_file, F_UNLCK, NULL, (void **)&dp);
5529out_clnt_odstate:
5530	put_clnt_odstate(dp->dl_clnt_odstate);
5531	nfs4_put_stid(&dp->dl_stid);
5532out_delegees:
5533	put_deleg_file(fp);
5534	return ERR_PTR(status);
5535}
5536
5537static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
5538{
5539	open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5540	if (status == -EAGAIN)
5541		open->op_why_no_deleg = WND4_CONTENTION;
5542	else {
5543		open->op_why_no_deleg = WND4_RESOURCE;
5544		switch (open->op_deleg_want) {
5545		case NFS4_SHARE_WANT_READ_DELEG:
5546		case NFS4_SHARE_WANT_WRITE_DELEG:
5547		case NFS4_SHARE_WANT_ANY_DELEG:
5548			break;
5549		case NFS4_SHARE_WANT_CANCEL:
5550			open->op_why_no_deleg = WND4_CANCELLED;
5551			break;
5552		case NFS4_SHARE_WANT_NO_DELEG:
5553			WARN_ON_ONCE(1);
5554		}
5555	}
5556}
5557
5558/*
5559 * Attempt to hand out a delegation.
5560 *
5561 * Note we don't support write delegations, and won't until the vfs has
5562 * proper support for them.
5563 */
5564static void
5565nfs4_open_delegation(struct nfsd4_open *open, struct nfs4_ol_stateid *stp,
5566		     struct svc_fh *currentfh)
5567{
5568	struct nfs4_delegation *dp;
5569	struct nfs4_openowner *oo = openowner(stp->st_stateowner);
5570	struct nfs4_client *clp = stp->st_stid.sc_client;
5571	struct svc_fh *parent = NULL;
5572	int cb_up;
5573	int status = 0;
5574
5575	cb_up = nfsd4_cb_channel_good(oo->oo_owner.so_client);
5576	open->op_recall = 0;
5577	switch (open->op_claim_type) {
5578		case NFS4_OPEN_CLAIM_PREVIOUS:
5579			if (!cb_up)
5580				open->op_recall = 1;
5581			if (open->op_delegate_type != NFS4_OPEN_DELEGATE_READ)
5582				goto out_no_deleg;
5583			break;
5584		case NFS4_OPEN_CLAIM_NULL:
5585			parent = currentfh;
5586			fallthrough;
5587		case NFS4_OPEN_CLAIM_FH:
5588			/*
5589			 * Let's not give out any delegations till everyone's
5590			 * had the chance to reclaim theirs, *and* until
5591			 * NLM locks have all been reclaimed:
5592			 */
5593			if (locks_in_grace(clp->net))
5594				goto out_no_deleg;
5595			if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
5596				goto out_no_deleg;
5597			break;
5598		default:
5599			goto out_no_deleg;
5600	}
5601	dp = nfs4_set_delegation(open, stp, parent);
5602	if (IS_ERR(dp))
5603		goto out_no_deleg;
5604
5605	memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
5606
5607	trace_nfsd_deleg_read(&dp->dl_stid.sc_stateid);
5608	open->op_delegate_type = NFS4_OPEN_DELEGATE_READ;
5609	nfs4_put_stid(&dp->dl_stid);
5610	return;
5611out_no_deleg:
5612	open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE;
5613	if (open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS &&
5614	    open->op_delegate_type != NFS4_OPEN_DELEGATE_NONE) {
5615		dprintk("NFSD: WARNING: refusing delegation reclaim\n");
5616		open->op_recall = 1;
5617	}
5618
5619	/* 4.1 client asking for a delegation? */
5620	if (open->op_deleg_want)
5621		nfsd4_open_deleg_none_ext(open, status);
5622	return;
5623}
5624
5625static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
5626					struct nfs4_delegation *dp)
5627{
5628	if (open->op_deleg_want == NFS4_SHARE_WANT_READ_DELEG &&
5629	    dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
5630		open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5631		open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
5632	} else if (open->op_deleg_want == NFS4_SHARE_WANT_WRITE_DELEG &&
5633		   dp->dl_type == NFS4_OPEN_DELEGATE_WRITE) {
5634		open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5635		open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
5636	}
5637	/* Otherwise the client must be confused wanting a delegation
5638	 * it already has, therefore we don't return
5639	 * NFS4_OPEN_DELEGATE_NONE_EXT and reason.
5640	 */
5641}
5642
5643/**
5644 * nfsd4_process_open2 - finish open processing
5645 * @rqstp: the RPC transaction being executed
5646 * @current_fh: NFSv4 COMPOUND's current filehandle
5647 * @open: OPEN arguments
5648 *
5649 * If successful, (1) truncate the file if open->op_truncate was
5650 * set, (2) set open->op_stateid, (3) set open->op_delegation.
5651 *
5652 * Returns %nfs_ok on success; otherwise an nfs4stat value in
5653 * network byte order is returned.
5654 */
5655__be32
5656nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
5657{
5658	struct nfsd4_compoundres *resp = rqstp->rq_resp;
5659	struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
5660	struct nfs4_file *fp = NULL;
5661	struct nfs4_ol_stateid *stp = NULL;
5662	struct nfs4_delegation *dp = NULL;
5663	__be32 status;
5664	bool new_stp = false;
5665
5666	/*
5667	 * Lookup file; if found, lookup stateid and check open request,
5668	 * and check for delegations in the process of being recalled.
5669	 * If not found, create the nfs4_file struct
5670	 */
5671	fp = nfsd4_file_hash_insert(open->op_file, current_fh);
5672	if (unlikely(!fp))
5673		return nfserr_jukebox;
5674	if (fp != open->op_file) {
5675		status = nfs4_check_deleg(cl, open, &dp);
5676		if (status)
5677			goto out;
5678		stp = nfsd4_find_and_lock_existing_open(fp, open);
5679	} else {
5680		open->op_file = NULL;
5681		status = nfserr_bad_stateid;
5682		if (nfsd4_is_deleg_cur(open))
5683			goto out;
5684	}
5685
5686	if (!stp) {
5687		stp = init_open_stateid(fp, open);
5688		if (!open->op_stp)
5689			new_stp = true;
5690	}
5691
5692	/*
5693	 * OPEN the file, or upgrade an existing OPEN.
5694	 * If truncate fails, the OPEN fails.
5695	 *
5696	 * stp is already locked.
5697	 */
5698	if (!new_stp) {
5699		/* Stateid was found, this is an OPEN upgrade */
5700		status = nfs4_upgrade_open(rqstp, fp, current_fh, stp, open);
5701		if (status) {
5702			mutex_unlock(&stp->st_mutex);
5703			goto out;
5704		}
5705	} else {
5706		status = nfs4_get_vfs_file(rqstp, fp, current_fh, stp, open, true);
5707		if (status) {
5708			stp->st_stid.sc_type = NFS4_CLOSED_STID;
5709			release_open_stateid(stp);
5710			mutex_unlock(&stp->st_mutex);
5711			goto out;
5712		}
5713
5714		stp->st_clnt_odstate = find_or_hash_clnt_odstate(fp,
5715							open->op_odstate);
5716		if (stp->st_clnt_odstate == open->op_odstate)
5717			open->op_odstate = NULL;
5718	}
5719
5720	nfs4_inc_and_copy_stateid(&open->op_stateid, &stp->st_stid);
5721	mutex_unlock(&stp->st_mutex);
5722
5723	if (nfsd4_has_session(&resp->cstate)) {
5724		if (open->op_deleg_want & NFS4_SHARE_WANT_NO_DELEG) {
5725			open->op_delegate_type = NFS4_OPEN_DELEGATE_NONE_EXT;
5726			open->op_why_no_deleg = WND4_NOT_WANTED;
5727			goto nodeleg;
5728		}
5729	}
5730
5731	/*
5732	* Attempt to hand out a delegation. No error return, because the
5733	* OPEN succeeds even if we fail.
5734	*/
5735	nfs4_open_delegation(open, stp, &resp->cstate.current_fh);
5736nodeleg:
5737	status = nfs_ok;
5738	trace_nfsd_open(&stp->st_stid.sc_stateid);
5739out:
5740	/* 4.1 client trying to upgrade/downgrade delegation? */
5741	if (open->op_delegate_type == NFS4_OPEN_DELEGATE_NONE && dp &&
5742	    open->op_deleg_want)
5743		nfsd4_deleg_xgrade_none_ext(open, dp);
5744
5745	if (fp)
5746		put_nfs4_file(fp);
5747	if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
5748		open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
5749	/*
5750	* To finish the open response, we just need to set the rflags.
5751	*/
5752	open->op_rflags = NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
5753	if (nfsd4_has_session(&resp->cstate))
5754		open->op_rflags |= NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK;
5755	else if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED))
5756		open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
5757
5758	if (dp)
5759		nfs4_put_stid(&dp->dl_stid);
5760	if (stp)
5761		nfs4_put_stid(&stp->st_stid);
5762
5763	return status;
5764}
5765
5766void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
5767			      struct nfsd4_open *open)
5768{
5769	if (open->op_openowner) {
5770		struct nfs4_stateowner *so = &open->op_openowner->oo_owner;
5771
5772		nfsd4_cstate_assign_replay(cstate, so);
5773		nfs4_put_stateowner(so);
5774	}
5775	if (open->op_file)
5776		kmem_cache_free(file_slab, open->op_file);
5777	if (open->op_stp)
5778		nfs4_put_stid(&open->op_stp->st_stid);
5779	if (open->op_odstate)
5780		kmem_cache_free(odstate_slab, open->op_odstate);
5781}
5782
5783__be32
5784nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
5785	    union nfsd4_op_u *u)
5786{
5787	clientid_t *clid = &u->renew;
5788	struct nfs4_client *clp;
5789	__be32 status;
5790	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
5791
5792	trace_nfsd_clid_renew(clid);
5793	status = set_client(clid, cstate, nn);
5794	if (status)
5795		return status;
5796	clp = cstate->clp;
5797	if (!list_empty(&clp->cl_delegations)
5798			&& clp->cl_cb_state != NFSD4_CB_UP)
5799		return nfserr_cb_path_down;
5800	return nfs_ok;
5801}
5802
5803void
5804nfsd4_end_grace(struct nfsd_net *nn)
5805{
5806	/* do nothing if grace period already ended */
5807	if (nn->grace_ended)
5808		return;
5809
5810	trace_nfsd_grace_complete(nn);
5811	nn->grace_ended = true;
5812	/*
5813	 * If the server goes down again right now, an NFSv4
5814	 * client will still be allowed to reclaim after it comes back up,
5815	 * even if it hasn't yet had a chance to reclaim state this time.
5816	 *
5817	 */
5818	nfsd4_record_grace_done(nn);
5819	/*
5820	 * At this point, NFSv4 clients can still reclaim.  But if the
5821	 * server crashes, any that have not yet reclaimed will be out
5822	 * of luck on the next boot.
5823	 *
5824	 * (NFSv4.1+ clients are considered to have reclaimed once they
5825	 * call RECLAIM_COMPLETE.  NFSv4.0 clients are considered to
5826	 * have reclaimed after their first OPEN.)
5827	 */
5828	locks_end_grace(&nn->nfsd4_manager);
5829	/*
5830	 * At this point, and once lockd and/or any other containers
5831	 * exit their grace period, further reclaims will fail and
5832	 * regular locking can resume.
5833	 */
5834}
5835
5836/*
5837 * If we've waited a lease period but there are still clients trying to
5838 * reclaim, wait a little longer to give them a chance to finish.
5839 */
5840static bool clients_still_reclaiming(struct nfsd_net *nn)
5841{
5842	time64_t double_grace_period_end = nn->boot_time +
5843					   2 * nn->nfsd4_lease;
5844
5845	if (nn->track_reclaim_completes &&
5846			atomic_read(&nn->nr_reclaim_complete) ==
5847			nn->reclaim_str_hashtbl_size)
5848		return false;
5849	if (!nn->somebody_reclaimed)
5850		return false;
5851	nn->somebody_reclaimed = false;
5852	/*
5853	 * If we've given them *two* lease times to reclaim, and they're
5854	 * still not done, give up:
5855	 */
5856	if (ktime_get_boottime_seconds() > double_grace_period_end)
5857		return false;
5858	return true;
5859}
5860
5861struct laundry_time {
5862	time64_t cutoff;
5863	time64_t new_timeo;
5864};
5865
5866static bool state_expired(struct laundry_time *lt, time64_t last_refresh)
5867{
5868	time64_t time_remaining;
5869
5870	if (last_refresh < lt->cutoff)
5871		return true;
5872	time_remaining = last_refresh - lt->cutoff;
5873	lt->new_timeo = min(lt->new_timeo, time_remaining);
5874	return false;
5875}
5876
5877#ifdef CONFIG_NFSD_V4_2_INTER_SSC
5878void nfsd4_ssc_init_umount_work(struct nfsd_net *nn)
5879{
5880	spin_lock_init(&nn->nfsd_ssc_lock);
5881	INIT_LIST_HEAD(&nn->nfsd_ssc_mount_list);
5882	init_waitqueue_head(&nn->nfsd_ssc_waitq);
5883}
5884EXPORT_SYMBOL_GPL(nfsd4_ssc_init_umount_work);
5885
5886/*
5887 * This is called when nfsd is being shutdown, after all inter_ssc
5888 * cleanup were done, to destroy the ssc delayed unmount list.
5889 */
5890static void nfsd4_ssc_shutdown_umount(struct nfsd_net *nn)
5891{
5892	struct nfsd4_ssc_umount_item *ni = NULL;
5893	struct nfsd4_ssc_umount_item *tmp;
5894
5895	spin_lock(&nn->nfsd_ssc_lock);
5896	list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
5897		list_del(&ni->nsui_list);
5898		spin_unlock(&nn->nfsd_ssc_lock);
5899		mntput(ni->nsui_vfsmount);
5900		kfree(ni);
5901		spin_lock(&nn->nfsd_ssc_lock);
5902	}
5903	spin_unlock(&nn->nfsd_ssc_lock);
5904}
5905
5906static void nfsd4_ssc_expire_umount(struct nfsd_net *nn)
5907{
5908	bool do_wakeup = false;
5909	struct nfsd4_ssc_umount_item *ni = NULL;
5910	struct nfsd4_ssc_umount_item *tmp;
5911
5912	spin_lock(&nn->nfsd_ssc_lock);
5913	list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
5914		if (time_after(jiffies, ni->nsui_expire)) {
5915			if (refcount_read(&ni->nsui_refcnt) > 1)
5916				continue;
5917
5918			/* mark being unmount */
5919			ni->nsui_busy = true;
5920			spin_unlock(&nn->nfsd_ssc_lock);
5921			mntput(ni->nsui_vfsmount);
5922			spin_lock(&nn->nfsd_ssc_lock);
5923
5924			/* waiters need to start from begin of list */
5925			list_del(&ni->nsui_list);
5926			kfree(ni);
5927
5928			/* wakeup ssc_connect waiters */
5929			do_wakeup = true;
5930			continue;
5931		}
5932		break;
5933	}
5934	if (do_wakeup)
5935		wake_up_all(&nn->nfsd_ssc_waitq);
5936	spin_unlock(&nn->nfsd_ssc_lock);
5937}
5938#endif
5939
5940/* Check if any lock belonging to this lockowner has any blockers */
5941static bool
5942nfs4_lockowner_has_blockers(struct nfs4_lockowner *lo)
5943{
5944	struct file_lock_context *ctx;
5945	struct nfs4_ol_stateid *stp;
5946	struct nfs4_file *nf;
5947
5948	list_for_each_entry(stp, &lo->lo_owner.so_stateids, st_perstateowner) {
5949		nf = stp->st_stid.sc_file;
5950		ctx = locks_inode_context(nf->fi_inode);
5951		if (!ctx)
5952			continue;
5953		if (locks_owner_has_blockers(ctx, lo))
5954			return true;
5955	}
5956	return false;
5957}
5958
5959static bool
5960nfs4_anylock_blockers(struct nfs4_client *clp)
5961{
5962	int i;
5963	struct nfs4_stateowner *so;
5964	struct nfs4_lockowner *lo;
5965
5966	if (atomic_read(&clp->cl_delegs_in_recall))
5967		return true;
5968	spin_lock(&clp->cl_lock);
5969	for (i = 0; i < OWNER_HASH_SIZE; i++) {
5970		list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[i],
5971				so_strhash) {
5972			if (so->so_is_open_owner)
5973				continue;
5974			lo = lockowner(so);
5975			if (nfs4_lockowner_has_blockers(lo)) {
5976				spin_unlock(&clp->cl_lock);
5977				return true;
5978			}
5979		}
5980	}
5981	spin_unlock(&clp->cl_lock);
5982	return false;
5983}
5984
5985static void
5986nfs4_get_client_reaplist(struct nfsd_net *nn, struct list_head *reaplist,
5987				struct laundry_time *lt)
5988{
5989	unsigned int maxreap, reapcnt = 0;
5990	struct list_head *pos, *next;
5991	struct nfs4_client *clp;
5992
5993	maxreap = (atomic_read(&nn->nfs4_client_count) >= nn->nfs4_max_clients) ?
5994			NFSD_CLIENT_MAX_TRIM_PER_RUN : 0;
5995	INIT_LIST_HEAD(reaplist);
5996	spin_lock(&nn->client_lock);
5997	list_for_each_safe(pos, next, &nn->client_lru) {
5998		clp = list_entry(pos, struct nfs4_client, cl_lru);
5999		if (clp->cl_state == NFSD4_EXPIRABLE)
6000			goto exp_client;
6001		if (!state_expired(lt, clp->cl_time))
6002			break;
6003		if (!atomic_read(&clp->cl_rpc_users)) {
6004			if (clp->cl_state == NFSD4_ACTIVE)
6005				atomic_inc(&nn->nfsd_courtesy_clients);
6006			clp->cl_state = NFSD4_COURTESY;
6007		}
6008		if (!client_has_state(clp))
6009			goto exp_client;
6010		if (!nfs4_anylock_blockers(clp))
6011			if (reapcnt >= maxreap)
6012				continue;
6013exp_client:
6014		if (!mark_client_expired_locked(clp)) {
6015			list_add(&clp->cl_lru, reaplist);
6016			reapcnt++;
6017		}
6018	}
6019	spin_unlock(&nn->client_lock);
6020}
6021
6022static void
6023nfs4_get_courtesy_client_reaplist(struct nfsd_net *nn,
6024				struct list_head *reaplist)
6025{
6026	unsigned int maxreap = 0, reapcnt = 0;
6027	struct list_head *pos, *next;
6028	struct nfs4_client *clp;
6029
6030	maxreap = NFSD_CLIENT_MAX_TRIM_PER_RUN;
6031	INIT_LIST_HEAD(reaplist);
6032
6033	spin_lock(&nn->client_lock);
6034	list_for_each_safe(pos, next, &nn->client_lru) {
6035		clp = list_entry(pos, struct nfs4_client, cl_lru);
6036		if (clp->cl_state == NFSD4_ACTIVE)
6037			break;
6038		if (reapcnt >= maxreap)
6039			break;
6040		if (!mark_client_expired_locked(clp)) {
6041			list_add(&clp->cl_lru, reaplist);
6042			reapcnt++;
6043		}
6044	}
6045	spin_unlock(&nn->client_lock);
6046}
6047
6048static void
6049nfs4_process_client_reaplist(struct list_head *reaplist)
6050{
6051	struct list_head *pos, *next;
6052	struct nfs4_client *clp;
6053
6054	list_for_each_safe(pos, next, reaplist) {
6055		clp = list_entry(pos, struct nfs4_client, cl_lru);
6056		trace_nfsd_clid_purged(&clp->cl_clientid);
6057		list_del_init(&clp->cl_lru);
6058		expire_client(clp);
6059	}
6060}
6061
6062static time64_t
6063nfs4_laundromat(struct nfsd_net *nn)
6064{
 
6065	struct nfs4_openowner *oo;
6066	struct nfs4_delegation *dp;
6067	struct nfs4_ol_stateid *stp;
6068	struct nfsd4_blocked_lock *nbl;
6069	struct list_head *pos, *next, reaplist;
6070	struct laundry_time lt = {
6071		.cutoff = ktime_get_boottime_seconds() - nn->nfsd4_lease,
6072		.new_timeo = nn->nfsd4_lease
6073	};
6074	struct nfs4_cpntf_state *cps;
6075	copy_stateid_t *cps_t;
6076	int i;
6077
6078	if (clients_still_reclaiming(nn)) {
6079		lt.new_timeo = 0;
6080		goto out;
6081	}
6082	nfsd4_end_grace(nn);
 
6083
6084	spin_lock(&nn->s2s_cp_lock);
6085	idr_for_each_entry(&nn->s2s_cp_stateids, cps_t, i) {
6086		cps = container_of(cps_t, struct nfs4_cpntf_state, cp_stateid);
6087		if (cps->cp_stateid.cs_type == NFS4_COPYNOTIFY_STID &&
6088				state_expired(&lt, cps->cpntf_time))
6089			_free_cpntf_state_locked(nn, cps);
6090	}
6091	spin_unlock(&nn->s2s_cp_lock);
6092	nfs4_get_client_reaplist(nn, &reaplist, &lt);
6093	nfs4_process_client_reaplist(&reaplist);
6094
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6095	spin_lock(&state_lock);
6096	list_for_each_safe(pos, next, &nn->del_recall_lru) {
6097		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
6098		if (!state_expired(&lt, dp->dl_time))
6099			break;
6100		WARN_ON(!unhash_delegation_locked(dp));
6101		list_add(&dp->dl_recall_lru, &reaplist);
6102	}
6103	spin_unlock(&state_lock);
6104	while (!list_empty(&reaplist)) {
6105		dp = list_first_entry(&reaplist, struct nfs4_delegation,
6106					dl_recall_lru);
6107		list_del_init(&dp->dl_recall_lru);
6108		revoke_delegation(dp);
6109	}
6110
6111	spin_lock(&nn->client_lock);
6112	while (!list_empty(&nn->close_lru)) {
6113		oo = list_first_entry(&nn->close_lru, struct nfs4_openowner,
6114					oo_close_lru);
6115		if (!state_expired(&lt, oo->oo_time))
6116			break;
6117		list_del_init(&oo->oo_close_lru);
6118		stp = oo->oo_last_closed_stid;
6119		oo->oo_last_closed_stid = NULL;
6120		spin_unlock(&nn->client_lock);
6121		nfs4_put_stid(&stp->st_stid);
6122		spin_lock(&nn->client_lock);
6123	}
6124	spin_unlock(&nn->client_lock);
6125
6126	/*
6127	 * It's possible for a client to try and acquire an already held lock
6128	 * that is being held for a long time, and then lose interest in it.
6129	 * So, we clean out any un-revisited request after a lease period
6130	 * under the assumption that the client is no longer interested.
6131	 *
6132	 * RFC5661, sec. 9.6 states that the client must not rely on getting
6133	 * notifications and must continue to poll for locks, even when the
6134	 * server supports them. Thus this shouldn't lead to clients blocking
6135	 * indefinitely once the lock does become free.
6136	 */
6137	BUG_ON(!list_empty(&reaplist));
6138	spin_lock(&nn->blocked_locks_lock);
6139	while (!list_empty(&nn->blocked_locks_lru)) {
6140		nbl = list_first_entry(&nn->blocked_locks_lru,
6141					struct nfsd4_blocked_lock, nbl_lru);
6142		if (!state_expired(&lt, nbl->nbl_time))
6143			break;
6144		list_move(&nbl->nbl_lru, &reaplist);
6145		list_del_init(&nbl->nbl_list);
6146	}
6147	spin_unlock(&nn->blocked_locks_lock);
6148
6149	while (!list_empty(&reaplist)) {
6150		nbl = list_first_entry(&reaplist,
6151					struct nfsd4_blocked_lock, nbl_lru);
6152		list_del_init(&nbl->nbl_lru);
6153		free_blocked_lock(nbl);
6154	}
6155#ifdef CONFIG_NFSD_V4_2_INTER_SSC
6156	/* service the server-to-server copy delayed unmount list */
6157	nfsd4_ssc_expire_umount(nn);
6158#endif
6159out:
6160	return max_t(time64_t, lt.new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
6161}
6162
 
6163static void laundromat_main(struct work_struct *);
6164
6165static void
6166laundromat_main(struct work_struct *laundry)
6167{
6168	time64_t t;
6169	struct delayed_work *dwork = to_delayed_work(laundry);
6170	struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
6171					   laundromat_work);
6172
6173	t = nfs4_laundromat(nn);
6174	queue_delayed_work(laundry_wq, &nn->laundromat_work, t*HZ);
6175}
6176
6177static void
6178courtesy_client_reaper(struct nfsd_net *nn)
6179{
6180	struct list_head reaplist;
6181
6182	nfs4_get_courtesy_client_reaplist(nn, &reaplist);
6183	nfs4_process_client_reaplist(&reaplist);
6184}
6185
6186static void
6187deleg_reaper(struct nfsd_net *nn)
6188{
6189	struct list_head *pos, *next;
6190	struct nfs4_client *clp;
6191	struct list_head cblist;
6192
6193	INIT_LIST_HEAD(&cblist);
6194	spin_lock(&nn->client_lock);
6195	list_for_each_safe(pos, next, &nn->client_lru) {
6196		clp = list_entry(pos, struct nfs4_client, cl_lru);
6197		if (clp->cl_state != NFSD4_ACTIVE ||
6198			list_empty(&clp->cl_delegations) ||
6199			atomic_read(&clp->cl_delegs_in_recall) ||
6200			test_bit(NFSD4_CLIENT_CB_RECALL_ANY, &clp->cl_flags) ||
6201			(ktime_get_boottime_seconds() -
6202				clp->cl_ra_time < 5)) {
6203			continue;
6204		}
6205		list_add(&clp->cl_ra_cblist, &cblist);
6206
6207		/* release in nfsd4_cb_recall_any_release */
6208		atomic_inc(&clp->cl_rpc_users);
6209		set_bit(NFSD4_CLIENT_CB_RECALL_ANY, &clp->cl_flags);
6210		clp->cl_ra_time = ktime_get_boottime_seconds();
6211	}
6212	spin_unlock(&nn->client_lock);
6213
6214	while (!list_empty(&cblist)) {
6215		clp = list_first_entry(&cblist, struct nfs4_client,
6216					cl_ra_cblist);
6217		list_del_init(&clp->cl_ra_cblist);
6218		clp->cl_ra->ra_keep = 0;
6219		clp->cl_ra->ra_bmval[0] = BIT(RCA4_TYPE_MASK_RDATA_DLG);
6220		trace_nfsd_cb_recall_any(clp->cl_ra);
6221		nfsd4_run_cb(&clp->cl_ra->ra_cb);
6222	}
6223}
6224
6225static void
6226nfsd4_state_shrinker_worker(struct work_struct *work)
6227{
6228	struct nfsd_net *nn = container_of(work, struct nfsd_net,
6229				nfsd_shrinker_work);
6230
6231	courtesy_client_reaper(nn);
6232	deleg_reaper(nn);
6233}
6234
6235static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stid *stp)
6236{
6237	if (!fh_match(&fhp->fh_handle, &stp->sc_file->fi_fhandle))
6238		return nfserr_bad_stateid;
6239	return nfs_ok;
6240}
6241
6242static
6243__be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
6244{
6245        __be32 status = nfserr_openmode;
6246
6247	/* For lock stateid's, we test the parent open, not the lock: */
6248	if (stp->st_openstp)
6249		stp = stp->st_openstp;
6250	if ((flags & WR_STATE) && !access_permit_write(stp))
6251                goto out;
6252	if ((flags & RD_STATE) && !access_permit_read(stp))
6253                goto out;
6254	status = nfs_ok;
6255out:
6256	return status;
6257}
6258
6259static inline __be32
6260check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
6261{
6262	if (ONE_STATEID(stateid) && (flags & RD_STATE))
6263		return nfs_ok;
6264	else if (opens_in_grace(net)) {
6265		/* Answer in remaining cases depends on existence of
6266		 * conflicting state; so we must wait out the grace period. */
6267		return nfserr_grace;
6268	} else if (flags & WR_STATE)
6269		return nfs4_share_conflict(current_fh,
6270				NFS4_SHARE_DENY_WRITE);
6271	else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
6272		return nfs4_share_conflict(current_fh,
6273				NFS4_SHARE_DENY_READ);
6274}
6275
 
 
 
 
 
 
 
 
 
 
6276static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
6277{
6278	/*
6279	 * When sessions are used the stateid generation number is ignored
6280	 * when it is zero.
6281	 */
6282	if (has_session && in->si_generation == 0)
6283		return nfs_ok;
6284
6285	if (in->si_generation == ref->si_generation)
6286		return nfs_ok;
6287
6288	/* If the client sends us a stateid from the future, it's buggy: */
6289	if (nfsd4_stateid_generation_after(in, ref))
6290		return nfserr_bad_stateid;
6291	/*
6292	 * However, we could see a stateid from the past, even from a
6293	 * non-buggy client.  For example, if the client sends a lock
6294	 * while some IO is outstanding, the lock may bump si_generation
6295	 * while the IO is still in flight.  The client could avoid that
6296	 * situation by waiting for responses on all the IO requests,
6297	 * but better performance may result in retrying IO that
6298	 * receives an old_stateid error if requests are rarely
6299	 * reordered in flight:
6300	 */
6301	return nfserr_old_stateid;
6302}
6303
6304static __be32 nfsd4_stid_check_stateid_generation(stateid_t *in, struct nfs4_stid *s, bool has_session)
6305{
6306	__be32 ret;
6307
6308	spin_lock(&s->sc_lock);
6309	ret = nfsd4_verify_open_stid(s);
6310	if (ret == nfs_ok)
6311		ret = check_stateid_generation(in, &s->sc_stateid, has_session);
6312	spin_unlock(&s->sc_lock);
6313	return ret;
6314}
6315
6316static __be32 nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid *ols)
6317{
6318	if (ols->st_stateowner->so_is_open_owner &&
6319	    !(openowner(ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
6320		return nfserr_bad_stateid;
6321	return nfs_ok;
6322}
6323
6324static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
6325{
6326	struct nfs4_stid *s;
6327	__be32 status = nfserr_bad_stateid;
6328
6329	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
6330		CLOSE_STATEID(stateid))
6331		return status;
6332	if (!same_clid(&stateid->si_opaque.so_clid, &cl->cl_clientid))
6333		return status;
6334	spin_lock(&cl->cl_lock);
6335	s = find_stateid_locked(cl, stateid);
6336	if (!s)
6337		goto out_unlock;
6338	status = nfsd4_stid_check_stateid_generation(stateid, s, 1);
6339	if (status)
6340		goto out_unlock;
6341	switch (s->sc_type) {
6342	case NFS4_DELEG_STID:
6343		status = nfs_ok;
6344		break;
6345	case NFS4_REVOKED_DELEG_STID:
6346		status = nfserr_deleg_revoked;
6347		break;
6348	case NFS4_OPEN_STID:
6349	case NFS4_LOCK_STID:
6350		status = nfsd4_check_openowner_confirmed(openlockstateid(s));
6351		break;
6352	default:
6353		printk("unknown stateid type %x\n", s->sc_type);
6354		fallthrough;
6355	case NFS4_CLOSED_STID:
6356	case NFS4_CLOSED_DELEG_STID:
6357		status = nfserr_bad_stateid;
6358	}
6359out_unlock:
6360	spin_unlock(&cl->cl_lock);
6361	return status;
6362}
6363
6364__be32
6365nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
6366		     stateid_t *stateid, unsigned char typemask,
6367		     struct nfs4_stid **s, struct nfsd_net *nn)
6368{
6369	__be32 status;
6370	struct nfs4_stid *stid;
6371	bool return_revoked = false;
6372
6373	/*
6374	 *  only return revoked delegations if explicitly asked.
6375	 *  otherwise we report revoked or bad_stateid status.
6376	 */
6377	if (typemask & NFS4_REVOKED_DELEG_STID)
6378		return_revoked = true;
6379	else if (typemask & NFS4_DELEG_STID)
6380		typemask |= NFS4_REVOKED_DELEG_STID;
6381
6382	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
6383		CLOSE_STATEID(stateid))
6384		return nfserr_bad_stateid;
6385	status = set_client(&stateid->si_opaque.so_clid, cstate, nn);
6386	if (status == nfserr_stale_clientid) {
6387		if (cstate->session)
6388			return nfserr_bad_stateid;
6389		return nfserr_stale_stateid;
6390	}
6391	if (status)
6392		return status;
6393	stid = find_stateid_by_type(cstate->clp, stateid, typemask);
6394	if (!stid)
6395		return nfserr_bad_stateid;
6396	if ((stid->sc_type == NFS4_REVOKED_DELEG_STID) && !return_revoked) {
6397		nfs4_put_stid(stid);
6398		if (cstate->minorversion)
6399			return nfserr_deleg_revoked;
6400		return nfserr_bad_stateid;
6401	}
6402	*s = stid;
6403	return nfs_ok;
6404}
6405
6406static struct nfsd_file *
6407nfs4_find_file(struct nfs4_stid *s, int flags)
6408{
6409	if (!s)
6410		return NULL;
6411
6412	switch (s->sc_type) {
6413	case NFS4_DELEG_STID:
6414		if (WARN_ON_ONCE(!s->sc_file->fi_deleg_file))
6415			return NULL;
6416		return nfsd_file_get(s->sc_file->fi_deleg_file);
6417	case NFS4_OPEN_STID:
6418	case NFS4_LOCK_STID:
6419		if (flags & RD_STATE)
6420			return find_readable_file(s->sc_file);
6421		else
6422			return find_writeable_file(s->sc_file);
6423	}
6424
6425	return NULL;
6426}
6427
6428static __be32
6429nfs4_check_olstateid(struct nfs4_ol_stateid *ols, int flags)
6430{
6431	__be32 status;
6432
6433	status = nfsd4_check_openowner_confirmed(ols);
6434	if (status)
6435		return status;
6436	return nfs4_check_openmode(ols, flags);
6437}
6438
6439static __be32
6440nfs4_check_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct nfs4_stid *s,
6441		struct nfsd_file **nfp, int flags)
6442{
6443	int acc = (flags & RD_STATE) ? NFSD_MAY_READ : NFSD_MAY_WRITE;
6444	struct nfsd_file *nf;
6445	__be32 status;
6446
6447	nf = nfs4_find_file(s, flags);
6448	if (nf) {
6449		status = nfsd_permission(rqstp, fhp->fh_export, fhp->fh_dentry,
6450				acc | NFSD_MAY_OWNER_OVERRIDE);
6451		if (status) {
6452			nfsd_file_put(nf);
6453			goto out;
6454		}
6455	} else {
6456		status = nfsd_file_acquire(rqstp, fhp, acc, &nf);
6457		if (status)
6458			return status;
6459	}
6460	*nfp = nf;
6461out:
6462	return status;
6463}
6464static void
6465_free_cpntf_state_locked(struct nfsd_net *nn, struct nfs4_cpntf_state *cps)
6466{
6467	WARN_ON_ONCE(cps->cp_stateid.cs_type != NFS4_COPYNOTIFY_STID);
6468	if (!refcount_dec_and_test(&cps->cp_stateid.cs_count))
6469		return;
6470	list_del(&cps->cp_list);
6471	idr_remove(&nn->s2s_cp_stateids,
6472		   cps->cp_stateid.cs_stid.si_opaque.so_id);
6473	kfree(cps);
6474}
6475/*
6476 * A READ from an inter server to server COPY will have a
6477 * copy stateid. Look up the copy notify stateid from the
6478 * idr structure and take a reference on it.
6479 */
6480__be32 manage_cpntf_state(struct nfsd_net *nn, stateid_t *st,
6481			  struct nfs4_client *clp,
6482			  struct nfs4_cpntf_state **cps)
6483{
6484	copy_stateid_t *cps_t;
6485	struct nfs4_cpntf_state *state = NULL;
6486
6487	if (st->si_opaque.so_clid.cl_id != nn->s2s_cp_cl_id)
6488		return nfserr_bad_stateid;
6489	spin_lock(&nn->s2s_cp_lock);
6490	cps_t = idr_find(&nn->s2s_cp_stateids, st->si_opaque.so_id);
6491	if (cps_t) {
6492		state = container_of(cps_t, struct nfs4_cpntf_state,
6493				     cp_stateid);
6494		if (state->cp_stateid.cs_type != NFS4_COPYNOTIFY_STID) {
6495			state = NULL;
6496			goto unlock;
6497		}
6498		if (!clp)
6499			refcount_inc(&state->cp_stateid.cs_count);
6500		else
6501			_free_cpntf_state_locked(nn, state);
6502	}
6503unlock:
6504	spin_unlock(&nn->s2s_cp_lock);
6505	if (!state)
6506		return nfserr_bad_stateid;
6507	if (!clp && state)
6508		*cps = state;
6509	return 0;
6510}
6511
6512static __be32 find_cpntf_state(struct nfsd_net *nn, stateid_t *st,
6513			       struct nfs4_stid **stid)
6514{
6515	__be32 status;
6516	struct nfs4_cpntf_state *cps = NULL;
6517	struct nfs4_client *found;
6518
6519	status = manage_cpntf_state(nn, st, NULL, &cps);
6520	if (status)
6521		return status;
6522
6523	cps->cpntf_time = ktime_get_boottime_seconds();
6524
6525	status = nfserr_expired;
6526	found = lookup_clientid(&cps->cp_p_clid, true, nn);
6527	if (!found)
6528		goto out;
6529
6530	*stid = find_stateid_by_type(found, &cps->cp_p_stateid,
6531			NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID);
6532	if (*stid)
6533		status = nfs_ok;
6534	else
6535		status = nfserr_bad_stateid;
6536
6537	put_client_renew(found);
6538out:
6539	nfs4_put_cpntf_state(nn, cps);
6540	return status;
6541}
6542
6543void nfs4_put_cpntf_state(struct nfsd_net *nn, struct nfs4_cpntf_state *cps)
6544{
6545	spin_lock(&nn->s2s_cp_lock);
6546	_free_cpntf_state_locked(nn, cps);
6547	spin_unlock(&nn->s2s_cp_lock);
6548}
6549
6550/*
6551 * Checks for stateid operations
6552 */
6553__be32
6554nfs4_preprocess_stateid_op(struct svc_rqst *rqstp,
6555		struct nfsd4_compound_state *cstate, struct svc_fh *fhp,
6556		stateid_t *stateid, int flags, struct nfsd_file **nfp,
6557		struct nfs4_stid **cstid)
6558{
 
6559	struct net *net = SVC_NET(rqstp);
6560	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6561	struct nfs4_stid *s = NULL;
6562	__be32 status;
6563
6564	if (nfp)
6565		*nfp = NULL;
6566
 
 
 
6567	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
6568		if (cstid)
6569			status = nfserr_bad_stateid;
6570		else
6571			status = check_special_stateids(net, fhp, stateid,
6572									flags);
6573		goto done;
6574	}
6575
6576	status = nfsd4_lookup_stateid(cstate, stateid,
6577				NFS4_DELEG_STID|NFS4_OPEN_STID|NFS4_LOCK_STID,
6578				&s, nn);
6579	if (status == nfserr_bad_stateid)
6580		status = find_cpntf_state(nn, stateid, &s);
6581	if (status)
6582		return status;
6583	status = nfsd4_stid_check_stateid_generation(stateid, s,
6584			nfsd4_has_session(cstate));
6585	if (status)
6586		goto out;
6587
6588	switch (s->sc_type) {
6589	case NFS4_DELEG_STID:
6590		status = nfs4_check_delegmode(delegstateid(s), flags);
6591		break;
6592	case NFS4_OPEN_STID:
6593	case NFS4_LOCK_STID:
6594		status = nfs4_check_olstateid(openlockstateid(s), flags);
6595		break;
6596	default:
6597		status = nfserr_bad_stateid;
6598		break;
6599	}
6600	if (status)
6601		goto out;
6602	status = nfs4_check_fh(fhp, s);
6603
6604done:
6605	if (status == nfs_ok && nfp)
6606		status = nfs4_check_file(rqstp, fhp, s, nfp, flags);
6607out:
6608	if (s) {
6609		if (!status && cstid)
6610			*cstid = s;
6611		else
6612			nfs4_put_stid(s);
6613	}
6614	return status;
6615}
6616
6617/*
6618 * Test if the stateid is valid
6619 */
6620__be32
6621nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6622		   union nfsd4_op_u *u)
6623{
6624	struct nfsd4_test_stateid *test_stateid = &u->test_stateid;
6625	struct nfsd4_test_stateid_id *stateid;
6626	struct nfs4_client *cl = cstate->clp;
6627
6628	list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
6629		stateid->ts_id_status =
6630			nfsd4_validate_stateid(cl, &stateid->ts_id_stateid);
6631
6632	return nfs_ok;
6633}
6634
6635static __be32
6636nfsd4_free_lock_stateid(stateid_t *stateid, struct nfs4_stid *s)
6637{
6638	struct nfs4_ol_stateid *stp = openlockstateid(s);
6639	__be32 ret;
6640
6641	ret = nfsd4_lock_ol_stateid(stp);
6642	if (ret)
6643		goto out_put_stid;
6644
6645	ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
6646	if (ret)
6647		goto out;
6648
6649	ret = nfserr_locks_held;
6650	if (check_for_locks(stp->st_stid.sc_file,
6651			    lockowner(stp->st_stateowner)))
6652		goto out;
6653
6654	release_lock_stateid(stp);
6655	ret = nfs_ok;
6656
6657out:
6658	mutex_unlock(&stp->st_mutex);
6659out_put_stid:
6660	nfs4_put_stid(s);
6661	return ret;
6662}
6663
6664__be32
6665nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6666		   union nfsd4_op_u *u)
6667{
6668	struct nfsd4_free_stateid *free_stateid = &u->free_stateid;
6669	stateid_t *stateid = &free_stateid->fr_stateid;
6670	struct nfs4_stid *s;
6671	struct nfs4_delegation *dp;
6672	struct nfs4_client *cl = cstate->clp;
6673	__be32 ret = nfserr_bad_stateid;
6674
6675	spin_lock(&cl->cl_lock);
6676	s = find_stateid_locked(cl, stateid);
6677	if (!s)
6678		goto out_unlock;
6679	spin_lock(&s->sc_lock);
6680	switch (s->sc_type) {
6681	case NFS4_DELEG_STID:
6682		ret = nfserr_locks_held;
6683		break;
6684	case NFS4_OPEN_STID:
6685		ret = check_stateid_generation(stateid, &s->sc_stateid, 1);
6686		if (ret)
6687			break;
6688		ret = nfserr_locks_held;
6689		break;
6690	case NFS4_LOCK_STID:
6691		spin_unlock(&s->sc_lock);
6692		refcount_inc(&s->sc_count);
6693		spin_unlock(&cl->cl_lock);
6694		ret = nfsd4_free_lock_stateid(stateid, s);
6695		goto out;
6696	case NFS4_REVOKED_DELEG_STID:
6697		spin_unlock(&s->sc_lock);
6698		dp = delegstateid(s);
6699		list_del_init(&dp->dl_recall_lru);
6700		spin_unlock(&cl->cl_lock);
6701		nfs4_put_stid(s);
6702		ret = nfs_ok;
6703		goto out;
6704	/* Default falls through and returns nfserr_bad_stateid */
6705	}
6706	spin_unlock(&s->sc_lock);
6707out_unlock:
6708	spin_unlock(&cl->cl_lock);
6709out:
6710	return ret;
6711}
6712
6713static inline int
6714setlkflg (int type)
6715{
6716	return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
6717		RD_STATE : WR_STATE;
6718}
6719
6720static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
6721{
6722	struct svc_fh *current_fh = &cstate->current_fh;
6723	struct nfs4_stateowner *sop = stp->st_stateowner;
6724	__be32 status;
6725
6726	status = nfsd4_check_seqid(cstate, sop, seqid);
6727	if (status)
6728		return status;
6729	status = nfsd4_lock_ol_stateid(stp);
6730	if (status != nfs_ok)
6731		return status;
6732	status = check_stateid_generation(stateid, &stp->st_stid.sc_stateid, nfsd4_has_session(cstate));
6733	if (status == nfs_ok)
6734		status = nfs4_check_fh(current_fh, &stp->st_stid);
6735	if (status != nfs_ok)
6736		mutex_unlock(&stp->st_mutex);
6737	return status;
6738}
6739
6740/* 
6741 * Checks for sequence id mutating operations. 
6742 */
6743static __be32
6744nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
6745			 stateid_t *stateid, char typemask,
6746			 struct nfs4_ol_stateid **stpp,
6747			 struct nfsd_net *nn)
6748{
6749	__be32 status;
6750	struct nfs4_stid *s;
6751	struct nfs4_ol_stateid *stp = NULL;
6752
6753	trace_nfsd_preprocess(seqid, stateid);
6754
6755	*stpp = NULL;
6756	status = nfsd4_lookup_stateid(cstate, stateid, typemask, &s, nn);
6757	if (status)
6758		return status;
6759	stp = openlockstateid(s);
6760	nfsd4_cstate_assign_replay(cstate, stp->st_stateowner);
6761
6762	status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
6763	if (!status)
6764		*stpp = stp;
6765	else
6766		nfs4_put_stid(&stp->st_stid);
6767	return status;
6768}
6769
6770static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
6771						 stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
6772{
6773	__be32 status;
6774	struct nfs4_openowner *oo;
6775	struct nfs4_ol_stateid *stp;
6776
6777	status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
6778						NFS4_OPEN_STID, &stp, nn);
6779	if (status)
6780		return status;
6781	oo = openowner(stp->st_stateowner);
6782	if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
6783		mutex_unlock(&stp->st_mutex);
6784		nfs4_put_stid(&stp->st_stid);
6785		return nfserr_bad_stateid;
6786	}
6787	*stpp = stp;
6788	return nfs_ok;
6789}
6790
6791__be32
6792nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6793		   union nfsd4_op_u *u)
6794{
6795	struct nfsd4_open_confirm *oc = &u->open_confirm;
6796	__be32 status;
6797	struct nfs4_openowner *oo;
6798	struct nfs4_ol_stateid *stp;
6799	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6800
6801	dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
6802			cstate->current_fh.fh_dentry);
6803
6804	status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
6805	if (status)
6806		return status;
6807
6808	status = nfs4_preprocess_seqid_op(cstate,
6809					oc->oc_seqid, &oc->oc_req_stateid,
6810					NFS4_OPEN_STID, &stp, nn);
6811	if (status)
6812		goto out;
6813	oo = openowner(stp->st_stateowner);
6814	status = nfserr_bad_stateid;
6815	if (oo->oo_flags & NFS4_OO_CONFIRMED) {
6816		mutex_unlock(&stp->st_mutex);
6817		goto put_stateid;
6818	}
6819	oo->oo_flags |= NFS4_OO_CONFIRMED;
6820	nfs4_inc_and_copy_stateid(&oc->oc_resp_stateid, &stp->st_stid);
6821	mutex_unlock(&stp->st_mutex);
6822	trace_nfsd_open_confirm(oc->oc_seqid, &stp->st_stid.sc_stateid);
6823	nfsd4_client_record_create(oo->oo_owner.so_client);
6824	status = nfs_ok;
6825put_stateid:
6826	nfs4_put_stid(&stp->st_stid);
6827out:
6828	nfsd4_bump_seqid(cstate, status);
6829	return status;
6830}
6831
6832static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
6833{
6834	if (!test_access(access, stp))
6835		return;
6836	nfs4_file_put_access(stp->st_stid.sc_file, access);
6837	clear_access(access, stp);
6838}
6839
6840static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
6841{
6842	switch (to_access) {
6843	case NFS4_SHARE_ACCESS_READ:
6844		nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
6845		nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
6846		break;
6847	case NFS4_SHARE_ACCESS_WRITE:
6848		nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
6849		nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
6850		break;
6851	case NFS4_SHARE_ACCESS_BOTH:
6852		break;
6853	default:
6854		WARN_ON_ONCE(1);
6855	}
6856}
6857
6858__be32
6859nfsd4_open_downgrade(struct svc_rqst *rqstp,
6860		     struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
6861{
6862	struct nfsd4_open_downgrade *od = &u->open_downgrade;
6863	__be32 status;
6864	struct nfs4_ol_stateid *stp;
6865	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6866
6867	dprintk("NFSD: nfsd4_open_downgrade on file %pd\n", 
6868			cstate->current_fh.fh_dentry);
6869
6870	/* We don't yet support WANT bits: */
6871	if (od->od_deleg_want)
6872		dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
6873			od->od_deleg_want);
6874
6875	status = nfs4_preprocess_confirmed_seqid_op(cstate, od->od_seqid,
6876					&od->od_stateid, &stp, nn);
6877	if (status)
6878		goto out; 
6879	status = nfserr_inval;
6880	if (!test_access(od->od_share_access, stp)) {
6881		dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
6882			stp->st_access_bmap, od->od_share_access);
6883		goto put_stateid;
6884	}
6885	if (!test_deny(od->od_share_deny, stp)) {
6886		dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
6887			stp->st_deny_bmap, od->od_share_deny);
6888		goto put_stateid;
6889	}
6890	nfs4_stateid_downgrade(stp, od->od_share_access);
6891	reset_union_bmap_deny(od->od_share_deny, stp);
6892	nfs4_inc_and_copy_stateid(&od->od_stateid, &stp->st_stid);
6893	status = nfs_ok;
6894put_stateid:
6895	mutex_unlock(&stp->st_mutex);
6896	nfs4_put_stid(&stp->st_stid);
6897out:
6898	nfsd4_bump_seqid(cstate, status);
6899	return status;
6900}
6901
6902static void nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
6903{
6904	struct nfs4_client *clp = s->st_stid.sc_client;
6905	bool unhashed;
6906	LIST_HEAD(reaplist);
6907	struct nfs4_ol_stateid *stp;
6908
6909	spin_lock(&clp->cl_lock);
6910	unhashed = unhash_open_stateid(s, &reaplist);
6911
6912	if (clp->cl_minorversion) {
6913		if (unhashed)
6914			put_ol_stateid_locked(s, &reaplist);
6915		spin_unlock(&clp->cl_lock);
6916		list_for_each_entry(stp, &reaplist, st_locks)
6917			nfs4_free_cpntf_statelist(clp->net, &stp->st_stid);
6918		free_ol_stateid_reaplist(&reaplist);
6919	} else {
6920		spin_unlock(&clp->cl_lock);
6921		free_ol_stateid_reaplist(&reaplist);
6922		if (unhashed)
6923			move_to_close_lru(s, clp->net);
6924	}
6925}
6926
6927/*
6928 * nfs4_unlock_state() called after encode
6929 */
6930__be32
6931nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6932		union nfsd4_op_u *u)
6933{
6934	struct nfsd4_close *close = &u->close;
6935	__be32 status;
6936	struct nfs4_ol_stateid *stp;
6937	struct net *net = SVC_NET(rqstp);
6938	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
6939
6940	dprintk("NFSD: nfsd4_close on file %pd\n", 
6941			cstate->current_fh.fh_dentry);
6942
6943	status = nfs4_preprocess_seqid_op(cstate, close->cl_seqid,
6944					&close->cl_stateid,
6945					NFS4_OPEN_STID|NFS4_CLOSED_STID,
6946					&stp, nn);
6947	nfsd4_bump_seqid(cstate, status);
6948	if (status)
6949		goto out; 
6950
6951	stp->st_stid.sc_type = NFS4_CLOSED_STID;
6952
6953	/*
6954	 * Technically we don't _really_ have to increment or copy it, since
6955	 * it should just be gone after this operation and we clobber the
6956	 * copied value below, but we continue to do so here just to ensure
6957	 * that racing ops see that there was a state change.
6958	 */
6959	nfs4_inc_and_copy_stateid(&close->cl_stateid, &stp->st_stid);
6960
6961	nfsd4_close_open_stateid(stp);
6962	mutex_unlock(&stp->st_mutex);
6963
6964	/* v4.1+ suggests that we send a special stateid in here, since the
6965	 * clients should just ignore this anyway. Since this is not useful
6966	 * for v4.0 clients either, we set it to the special close_stateid
6967	 * universally.
6968	 *
6969	 * See RFC5661 section 18.2.4, and RFC7530 section 16.2.5
6970	 */
6971	memcpy(&close->cl_stateid, &close_stateid, sizeof(close->cl_stateid));
6972
6973	/* put reference from nfs4_preprocess_seqid_op */
6974	nfs4_put_stid(&stp->st_stid);
6975out:
6976	return status;
6977}
6978
6979__be32
6980nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6981		  union nfsd4_op_u *u)
6982{
6983	struct nfsd4_delegreturn *dr = &u->delegreturn;
6984	struct nfs4_delegation *dp;
6985	stateid_t *stateid = &dr->dr_stateid;
6986	struct nfs4_stid *s;
6987	__be32 status;
6988	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
6989
6990	if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
6991		return status;
6992
6993	status = nfsd4_lookup_stateid(cstate, stateid, NFS4_DELEG_STID, &s, nn);
6994	if (status)
6995		goto out;
6996	dp = delegstateid(s);
6997	status = nfsd4_stid_check_stateid_generation(stateid, &dp->dl_stid, nfsd4_has_session(cstate));
6998	if (status)
6999		goto put_stateid;
7000
7001	trace_nfsd_deleg_return(stateid);
7002	wake_up_var(d_inode(cstate->current_fh.fh_dentry));
7003	destroy_delegation(dp);
7004put_stateid:
7005	nfs4_put_stid(&dp->dl_stid);
7006out:
7007	return status;
7008}
7009
7010/* last octet in a range */
7011static inline u64
7012last_byte_offset(u64 start, u64 len)
7013{
7014	u64 end;
7015
7016	WARN_ON_ONCE(!len);
7017	end = start + len;
7018	return end > start ? end - 1: NFS4_MAX_UINT64;
7019}
7020
7021/*
7022 * TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
7023 * we can't properly handle lock requests that go beyond the (2^63 - 1)-th
7024 * byte, because of sign extension problems.  Since NFSv4 calls for 64-bit
7025 * locking, this prevents us from being completely protocol-compliant.  The
7026 * real solution to this problem is to start using unsigned file offsets in
7027 * the VFS, but this is a very deep change!
7028 */
7029static inline void
7030nfs4_transform_lock_offset(struct file_lock *lock)
7031{
7032	if (lock->fl_start < 0)
7033		lock->fl_start = OFFSET_MAX;
7034	if (lock->fl_end < 0)
7035		lock->fl_end = OFFSET_MAX;
7036}
7037
7038static fl_owner_t
7039nfsd4_lm_get_owner(fl_owner_t owner)
7040{
7041	struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
7042
7043	nfs4_get_stateowner(&lo->lo_owner);
7044	return owner;
7045}
7046
7047static void
7048nfsd4_lm_put_owner(fl_owner_t owner)
7049{
7050	struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
7051
7052	if (lo)
7053		nfs4_put_stateowner(&lo->lo_owner);
7054}
7055
7056/* return pointer to struct nfs4_client if client is expirable */
7057static bool
7058nfsd4_lm_lock_expirable(struct file_lock *cfl)
7059{
7060	struct nfs4_lockowner *lo = (struct nfs4_lockowner *)cfl->fl_owner;
7061	struct nfs4_client *clp = lo->lo_owner.so_client;
7062	struct nfsd_net *nn;
7063
7064	if (try_to_expire_client(clp)) {
7065		nn = net_generic(clp->net, nfsd_net_id);
7066		mod_delayed_work(laundry_wq, &nn->laundromat_work, 0);
7067		return true;
7068	}
7069	return false;
7070}
7071
7072/* schedule laundromat to run immediately and wait for it to complete */
7073static void
7074nfsd4_lm_expire_lock(void)
7075{
7076	flush_workqueue(laundry_wq);
7077}
7078
7079static void
7080nfsd4_lm_notify(struct file_lock *fl)
7081{
7082	struct nfs4_lockowner		*lo = (struct nfs4_lockowner *)fl->fl_owner;
7083	struct net			*net = lo->lo_owner.so_client->net;
7084	struct nfsd_net			*nn = net_generic(net, nfsd_net_id);
7085	struct nfsd4_blocked_lock	*nbl = container_of(fl,
7086						struct nfsd4_blocked_lock, nbl_lock);
7087	bool queue = false;
7088
7089	/* An empty list means that something else is going to be using it */
7090	spin_lock(&nn->blocked_locks_lock);
7091	if (!list_empty(&nbl->nbl_list)) {
7092		list_del_init(&nbl->nbl_list);
7093		list_del_init(&nbl->nbl_lru);
7094		queue = true;
7095	}
7096	spin_unlock(&nn->blocked_locks_lock);
7097
7098	if (queue) {
7099		trace_nfsd_cb_notify_lock(lo, nbl);
7100		nfsd4_run_cb(&nbl->nbl_cb);
7101	}
7102}
7103
7104static const struct lock_manager_operations nfsd_posix_mng_ops  = {
7105	.lm_mod_owner = THIS_MODULE,
7106	.lm_notify = nfsd4_lm_notify,
7107	.lm_get_owner = nfsd4_lm_get_owner,
7108	.lm_put_owner = nfsd4_lm_put_owner,
7109	.lm_lock_expirable = nfsd4_lm_lock_expirable,
7110	.lm_expire_lock = nfsd4_lm_expire_lock,
7111};
7112
7113static inline void
7114nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
7115{
7116	struct nfs4_lockowner *lo;
7117
7118	if (fl->fl_lmops == &nfsd_posix_mng_ops) {
7119		lo = (struct nfs4_lockowner *) fl->fl_owner;
7120		xdr_netobj_dup(&deny->ld_owner, &lo->lo_owner.so_owner,
7121						GFP_KERNEL);
7122		if (!deny->ld_owner.data)
7123			/* We just don't care that much */
7124			goto nevermind;
7125		deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
7126	} else {
7127nevermind:
7128		deny->ld_owner.len = 0;
7129		deny->ld_owner.data = NULL;
7130		deny->ld_clientid.cl_boot = 0;
7131		deny->ld_clientid.cl_id = 0;
7132	}
7133	deny->ld_start = fl->fl_start;
7134	deny->ld_length = NFS4_MAX_UINT64;
7135	if (fl->fl_end != NFS4_MAX_UINT64)
7136		deny->ld_length = fl->fl_end - fl->fl_start + 1;        
7137	deny->ld_type = NFS4_READ_LT;
7138	if (fl->fl_type != F_RDLCK)
7139		deny->ld_type = NFS4_WRITE_LT;
7140}
7141
7142static struct nfs4_lockowner *
7143find_lockowner_str_locked(struct nfs4_client *clp, struct xdr_netobj *owner)
7144{
7145	unsigned int strhashval = ownerstr_hashval(owner);
7146	struct nfs4_stateowner *so;
7147
7148	lockdep_assert_held(&clp->cl_lock);
7149
7150	list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[strhashval],
7151			    so_strhash) {
7152		if (so->so_is_open_owner)
7153			continue;
7154		if (same_owner_str(so, owner))
7155			return lockowner(nfs4_get_stateowner(so));
7156	}
7157	return NULL;
7158}
7159
7160static struct nfs4_lockowner *
7161find_lockowner_str(struct nfs4_client *clp, struct xdr_netobj *owner)
7162{
7163	struct nfs4_lockowner *lo;
7164
7165	spin_lock(&clp->cl_lock);
7166	lo = find_lockowner_str_locked(clp, owner);
7167	spin_unlock(&clp->cl_lock);
7168	return lo;
7169}
7170
7171static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
7172{
7173	unhash_lockowner_locked(lockowner(sop));
7174}
7175
7176static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
7177{
7178	struct nfs4_lockowner *lo = lockowner(sop);
7179
7180	kmem_cache_free(lockowner_slab, lo);
7181}
7182
7183static const struct nfs4_stateowner_operations lockowner_ops = {
7184	.so_unhash =	nfs4_unhash_lockowner,
7185	.so_free =	nfs4_free_lockowner,
7186};
7187
7188/*
7189 * Alloc a lock owner structure.
7190 * Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has 
7191 * occurred. 
7192 *
7193 * strhashval = ownerstr_hashval
7194 */
7195static struct nfs4_lockowner *
7196alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp,
7197			   struct nfs4_ol_stateid *open_stp,
7198			   struct nfsd4_lock *lock)
7199{
7200	struct nfs4_lockowner *lo, *ret;
7201
7202	lo = alloc_stateowner(lockowner_slab, &lock->lk_new_owner, clp);
7203	if (!lo)
7204		return NULL;
7205	INIT_LIST_HEAD(&lo->lo_blocked);
7206	INIT_LIST_HEAD(&lo->lo_owner.so_stateids);
7207	lo->lo_owner.so_is_open_owner = 0;
7208	lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
7209	lo->lo_owner.so_ops = &lockowner_ops;
7210	spin_lock(&clp->cl_lock);
7211	ret = find_lockowner_str_locked(clp, &lock->lk_new_owner);
7212	if (ret == NULL) {
7213		list_add(&lo->lo_owner.so_strhash,
7214			 &clp->cl_ownerstr_hashtbl[strhashval]);
7215		ret = lo;
7216	} else
7217		nfs4_free_stateowner(&lo->lo_owner);
7218
7219	spin_unlock(&clp->cl_lock);
7220	return ret;
7221}
7222
7223static struct nfs4_ol_stateid *
7224find_lock_stateid(const struct nfs4_lockowner *lo,
7225		  const struct nfs4_ol_stateid *ost)
7226{
7227	struct nfs4_ol_stateid *lst;
7228
7229	lockdep_assert_held(&ost->st_stid.sc_client->cl_lock);
7230
7231	/* If ost is not hashed, ost->st_locks will not be valid */
7232	if (!nfs4_ol_stateid_unhashed(ost))
7233		list_for_each_entry(lst, &ost->st_locks, st_locks) {
7234			if (lst->st_stateowner == &lo->lo_owner) {
7235				refcount_inc(&lst->st_stid.sc_count);
7236				return lst;
7237			}
7238		}
7239	return NULL;
7240}
7241
7242static struct nfs4_ol_stateid *
7243init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
7244		  struct nfs4_file *fp, struct inode *inode,
7245		  struct nfs4_ol_stateid *open_stp)
7246{
7247	struct nfs4_client *clp = lo->lo_owner.so_client;
7248	struct nfs4_ol_stateid *retstp;
7249
7250	mutex_init(&stp->st_mutex);
7251	mutex_lock_nested(&stp->st_mutex, OPEN_STATEID_MUTEX);
7252retry:
7253	spin_lock(&clp->cl_lock);
7254	if (nfs4_ol_stateid_unhashed(open_stp))
7255		goto out_close;
7256	retstp = find_lock_stateid(lo, open_stp);
7257	if (retstp)
7258		goto out_found;
7259	refcount_inc(&stp->st_stid.sc_count);
7260	stp->st_stid.sc_type = NFS4_LOCK_STID;
7261	stp->st_stateowner = nfs4_get_stateowner(&lo->lo_owner);
7262	get_nfs4_file(fp);
7263	stp->st_stid.sc_file = fp;
7264	stp->st_access_bmap = 0;
7265	stp->st_deny_bmap = open_stp->st_deny_bmap;
7266	stp->st_openstp = open_stp;
7267	spin_lock(&fp->fi_lock);
7268	list_add(&stp->st_locks, &open_stp->st_locks);
7269	list_add(&stp->st_perstateowner, &lo->lo_owner.so_stateids);
7270	list_add(&stp->st_perfile, &fp->fi_stateids);
7271	spin_unlock(&fp->fi_lock);
7272	spin_unlock(&clp->cl_lock);
7273	return stp;
7274out_found:
7275	spin_unlock(&clp->cl_lock);
7276	if (nfsd4_lock_ol_stateid(retstp) != nfs_ok) {
7277		nfs4_put_stid(&retstp->st_stid);
7278		goto retry;
7279	}
7280	/* To keep mutex tracking happy */
7281	mutex_unlock(&stp->st_mutex);
7282	return retstp;
7283out_close:
7284	spin_unlock(&clp->cl_lock);
7285	mutex_unlock(&stp->st_mutex);
7286	return NULL;
7287}
7288
7289static struct nfs4_ol_stateid *
7290find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
7291			    struct inode *inode, struct nfs4_ol_stateid *ost,
7292			    bool *new)
7293{
7294	struct nfs4_stid *ns = NULL;
7295	struct nfs4_ol_stateid *lst;
7296	struct nfs4_openowner *oo = openowner(ost->st_stateowner);
7297	struct nfs4_client *clp = oo->oo_owner.so_client;
7298
7299	*new = false;
7300	spin_lock(&clp->cl_lock);
7301	lst = find_lock_stateid(lo, ost);
7302	spin_unlock(&clp->cl_lock);
7303	if (lst != NULL) {
7304		if (nfsd4_lock_ol_stateid(lst) == nfs_ok)
7305			goto out;
7306		nfs4_put_stid(&lst->st_stid);
7307	}
7308	ns = nfs4_alloc_stid(clp, stateid_slab, nfs4_free_lock_stateid);
7309	if (ns == NULL)
7310		return NULL;
7311
7312	lst = init_lock_stateid(openlockstateid(ns), lo, fi, inode, ost);
7313	if (lst == openlockstateid(ns))
7314		*new = true;
7315	else
7316		nfs4_put_stid(ns);
7317out:
7318	return lst;
7319}
7320
7321static int
7322check_lock_length(u64 offset, u64 length)
7323{
7324	return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
7325		(length > ~offset)));
7326}
7327
7328static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
7329{
7330	struct nfs4_file *fp = lock_stp->st_stid.sc_file;
7331
7332	lockdep_assert_held(&fp->fi_lock);
7333
7334	if (test_access(access, lock_stp))
7335		return;
7336	__nfs4_file_get_access(fp, access);
7337	set_access(access, lock_stp);
7338}
7339
7340static __be32
7341lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
7342			    struct nfs4_ol_stateid *ost,
7343			    struct nfsd4_lock *lock,
7344			    struct nfs4_ol_stateid **plst, bool *new)
7345{
7346	__be32 status;
7347	struct nfs4_file *fi = ost->st_stid.sc_file;
7348	struct nfs4_openowner *oo = openowner(ost->st_stateowner);
7349	struct nfs4_client *cl = oo->oo_owner.so_client;
7350	struct inode *inode = d_inode(cstate->current_fh.fh_dentry);
7351	struct nfs4_lockowner *lo;
7352	struct nfs4_ol_stateid *lst;
7353	unsigned int strhashval;
7354
7355	lo = find_lockowner_str(cl, &lock->lk_new_owner);
7356	if (!lo) {
7357		strhashval = ownerstr_hashval(&lock->lk_new_owner);
7358		lo = alloc_init_lock_stateowner(strhashval, cl, ost, lock);
7359		if (lo == NULL)
7360			return nfserr_jukebox;
7361	} else {
7362		/* with an existing lockowner, seqids must be the same */
7363		status = nfserr_bad_seqid;
7364		if (!cstate->minorversion &&
7365		    lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
7366			goto out;
7367	}
7368
7369	lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
7370	if (lst == NULL) {
7371		status = nfserr_jukebox;
7372		goto out;
7373	}
7374
7375	status = nfs_ok;
7376	*plst = lst;
7377out:
7378	nfs4_put_stateowner(&lo->lo_owner);
7379	return status;
7380}
7381
7382/*
7383 *  LOCK operation 
7384 */
7385__be32
7386nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7387	   union nfsd4_op_u *u)
7388{
7389	struct nfsd4_lock *lock = &u->lock;
7390	struct nfs4_openowner *open_sop = NULL;
7391	struct nfs4_lockowner *lock_sop = NULL;
7392	struct nfs4_ol_stateid *lock_stp = NULL;
7393	struct nfs4_ol_stateid *open_stp = NULL;
7394	struct nfs4_file *fp;
7395	struct nfsd_file *nf = NULL;
7396	struct nfsd4_blocked_lock *nbl = NULL;
7397	struct file_lock *file_lock = NULL;
7398	struct file_lock *conflock = NULL;
 
7399	__be32 status = 0;
7400	int lkflg;
7401	int err;
7402	bool new = false;
7403	unsigned char fl_type;
7404	unsigned int fl_flags = FL_POSIX;
7405	struct net *net = SVC_NET(rqstp);
7406	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
7407
7408	dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
7409		(long long) lock->lk_offset,
7410		(long long) lock->lk_length);
7411
7412	if (check_lock_length(lock->lk_offset, lock->lk_length))
7413		 return nfserr_inval;
7414
7415	if ((status = fh_verify(rqstp, &cstate->current_fh,
7416				S_IFREG, NFSD_MAY_LOCK))) {
7417		dprintk("NFSD: nfsd4_lock: permission denied!\n");
7418		return status;
7419	}
 
7420
7421	if (lock->lk_is_new) {
7422		if (nfsd4_has_session(cstate))
7423			/* See rfc 5661 18.10.3: given clientid is ignored: */
7424			memcpy(&lock->lk_new_clientid,
7425				&cstate->clp->cl_clientid,
7426				sizeof(clientid_t));
7427
7428		/* validate and update open stateid and open seqid */
7429		status = nfs4_preprocess_confirmed_seqid_op(cstate,
7430				        lock->lk_new_open_seqid,
7431		                        &lock->lk_new_open_stateid,
7432					&open_stp, nn);
7433		if (status)
7434			goto out;
7435		mutex_unlock(&open_stp->st_mutex);
7436		open_sop = openowner(open_stp->st_stateowner);
7437		status = nfserr_bad_stateid;
7438		if (!same_clid(&open_sop->oo_owner.so_client->cl_clientid,
7439						&lock->lk_new_clientid))
7440			goto out;
7441		status = lookup_or_create_lock_state(cstate, open_stp, lock,
7442							&lock_stp, &new);
7443	} else {
7444		status = nfs4_preprocess_seqid_op(cstate,
7445				       lock->lk_old_lock_seqid,
7446				       &lock->lk_old_lock_stateid,
7447				       NFS4_LOCK_STID, &lock_stp, nn);
7448	}
7449	if (status)
7450		goto out;
7451	lock_sop = lockowner(lock_stp->st_stateowner);
7452
7453	lkflg = setlkflg(lock->lk_type);
7454	status = nfs4_check_openmode(lock_stp, lkflg);
7455	if (status)
7456		goto out;
7457
7458	status = nfserr_grace;
7459	if (locks_in_grace(net) && !lock->lk_reclaim)
7460		goto out;
7461	status = nfserr_no_grace;
7462	if (!locks_in_grace(net) && lock->lk_reclaim)
7463		goto out;
7464
7465	if (lock->lk_reclaim)
7466		fl_flags |= FL_RECLAIM;
7467
7468	fp = lock_stp->st_stid.sc_file;
7469	switch (lock->lk_type) {
7470		case NFS4_READW_LT:
7471			if (nfsd4_has_session(cstate))
 
7472				fl_flags |= FL_SLEEP;
7473			fallthrough;
7474		case NFS4_READ_LT:
7475			spin_lock(&fp->fi_lock);
7476			nf = find_readable_file_locked(fp);
7477			if (nf)
7478				get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
7479			spin_unlock(&fp->fi_lock);
7480			fl_type = F_RDLCK;
7481			break;
7482		case NFS4_WRITEW_LT:
7483			if (nfsd4_has_session(cstate))
 
7484				fl_flags |= FL_SLEEP;
7485			fallthrough;
7486		case NFS4_WRITE_LT:
7487			spin_lock(&fp->fi_lock);
7488			nf = find_writeable_file_locked(fp);
7489			if (nf)
7490				get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
7491			spin_unlock(&fp->fi_lock);
7492			fl_type = F_WRLCK;
7493			break;
7494		default:
7495			status = nfserr_inval;
7496		goto out;
7497	}
7498
7499	if (!nf) {
7500		status = nfserr_openmode;
7501		goto out;
7502	}
7503
7504	/*
7505	 * Most filesystems with their own ->lock operations will block
7506	 * the nfsd thread waiting to acquire the lock.  That leads to
7507	 * deadlocks (we don't want every nfsd thread tied up waiting
7508	 * for file locks), so don't attempt blocking lock notifications
7509	 * on those filesystems:
7510	 */
7511	if (nf->nf_file->f_op->lock)
7512		fl_flags &= ~FL_SLEEP;
7513
7514	nbl = find_or_allocate_block(lock_sop, &fp->fi_fhandle, nn);
7515	if (!nbl) {
7516		dprintk("NFSD: %s: unable to allocate block!\n", __func__);
7517		status = nfserr_jukebox;
7518		goto out;
7519	}
7520
7521	file_lock = &nbl->nbl_lock;
7522	file_lock->fl_type = fl_type;
7523	file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(&lock_sop->lo_owner));
7524	file_lock->fl_pid = current->tgid;
7525	file_lock->fl_file = nf->nf_file;
7526	file_lock->fl_flags = fl_flags;
7527	file_lock->fl_lmops = &nfsd_posix_mng_ops;
7528	file_lock->fl_start = lock->lk_offset;
7529	file_lock->fl_end = last_byte_offset(lock->lk_offset, lock->lk_length);
7530	nfs4_transform_lock_offset(file_lock);
7531
7532	conflock = locks_alloc_lock();
7533	if (!conflock) {
7534		dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
7535		status = nfserr_jukebox;
7536		goto out;
7537	}
7538
7539	if (fl_flags & FL_SLEEP) {
7540		nbl->nbl_time = ktime_get_boottime_seconds();
7541		spin_lock(&nn->blocked_locks_lock);
7542		list_add_tail(&nbl->nbl_list, &lock_sop->lo_blocked);
7543		list_add_tail(&nbl->nbl_lru, &nn->blocked_locks_lru);
7544		kref_get(&nbl->nbl_kref);
7545		spin_unlock(&nn->blocked_locks_lock);
7546	}
7547
7548	err = vfs_lock_file(nf->nf_file, F_SETLK, file_lock, conflock);
7549	switch (err) {
7550	case 0: /* success! */
7551		nfs4_inc_and_copy_stateid(&lock->lk_resp_stateid, &lock_stp->st_stid);
7552		status = 0;
7553		if (lock->lk_reclaim)
7554			nn->somebody_reclaimed = true;
7555		break;
7556	case FILE_LOCK_DEFERRED:
7557		kref_put(&nbl->nbl_kref, free_nbl);
7558		nbl = NULL;
7559		fallthrough;
7560	case -EAGAIN:		/* conflock holds conflicting lock */
7561		status = nfserr_denied;
7562		dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
7563		nfs4_set_lock_denied(conflock, &lock->lk_denied);
7564		break;
7565	case -EDEADLK:
7566		status = nfserr_deadlock;
7567		break;
7568	default:
7569		dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
7570		status = nfserrno(err);
7571		break;
7572	}
7573out:
7574	if (nbl) {
7575		/* dequeue it if we queued it before */
7576		if (fl_flags & FL_SLEEP) {
7577			spin_lock(&nn->blocked_locks_lock);
7578			if (!list_empty(&nbl->nbl_list) &&
7579			    !list_empty(&nbl->nbl_lru)) {
7580				list_del_init(&nbl->nbl_list);
7581				list_del_init(&nbl->nbl_lru);
7582				kref_put(&nbl->nbl_kref, free_nbl);
7583			}
7584			/* nbl can use one of lists to be linked to reaplist */
7585			spin_unlock(&nn->blocked_locks_lock);
7586		}
7587		free_blocked_lock(nbl);
7588	}
7589	if (nf)
7590		nfsd_file_put(nf);
7591	if (lock_stp) {
7592		/* Bump seqid manually if the 4.0 replay owner is openowner */
7593		if (cstate->replay_owner &&
7594		    cstate->replay_owner != &lock_sop->lo_owner &&
7595		    seqid_mutating_err(ntohl(status)))
7596			lock_sop->lo_owner.so_seqid++;
7597
7598		/*
7599		 * If this is a new, never-before-used stateid, and we are
7600		 * returning an error, then just go ahead and release it.
7601		 */
7602		if (status && new)
7603			release_lock_stateid(lock_stp);
7604
7605		mutex_unlock(&lock_stp->st_mutex);
7606
7607		nfs4_put_stid(&lock_stp->st_stid);
7608	}
7609	if (open_stp)
7610		nfs4_put_stid(&open_stp->st_stid);
7611	nfsd4_bump_seqid(cstate, status);
7612	if (conflock)
7613		locks_free_lock(conflock);
7614	return status;
7615}
7616
7617/*
7618 * The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
7619 * so we do a temporary open here just to get an open file to pass to
7620 * vfs_test_lock.
7621 */
7622static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
7623{
7624	struct nfsd_file *nf;
7625	struct inode *inode;
7626	__be32 err;
7627
7628	err = nfsd_file_acquire(rqstp, fhp, NFSD_MAY_READ, &nf);
7629	if (err)
7630		return err;
7631	inode = fhp->fh_dentry->d_inode;
7632	inode_lock(inode); /* to block new leases till after test_lock: */
7633	err = nfserrno(nfsd_open_break_lease(inode, NFSD_MAY_READ));
7634	if (err)
7635		goto out;
7636	lock->fl_file = nf->nf_file;
7637	err = nfserrno(vfs_test_lock(nf->nf_file, lock));
7638	lock->fl_file = NULL;
7639out:
7640	inode_unlock(inode);
7641	nfsd_file_put(nf);
7642	return err;
7643}
7644
7645/*
7646 * LOCKT operation
7647 */
7648__be32
7649nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7650	    union nfsd4_op_u *u)
7651{
7652	struct nfsd4_lockt *lockt = &u->lockt;
7653	struct file_lock *file_lock = NULL;
7654	struct nfs4_lockowner *lo = NULL;
7655	__be32 status;
7656	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
7657
7658	if (locks_in_grace(SVC_NET(rqstp)))
7659		return nfserr_grace;
7660
7661	if (check_lock_length(lockt->lt_offset, lockt->lt_length))
7662		 return nfserr_inval;
7663
7664	if (!nfsd4_has_session(cstate)) {
7665		status = set_client(&lockt->lt_clientid, cstate, nn);
7666		if (status)
7667			goto out;
7668	}
7669
7670	if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
7671		goto out;
7672
7673	file_lock = locks_alloc_lock();
7674	if (!file_lock) {
7675		dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
7676		status = nfserr_jukebox;
7677		goto out;
7678	}
7679
7680	switch (lockt->lt_type) {
7681		case NFS4_READ_LT:
7682		case NFS4_READW_LT:
7683			file_lock->fl_type = F_RDLCK;
7684			break;
7685		case NFS4_WRITE_LT:
7686		case NFS4_WRITEW_LT:
7687			file_lock->fl_type = F_WRLCK;
7688			break;
7689		default:
7690			dprintk("NFSD: nfs4_lockt: bad lock type!\n");
7691			status = nfserr_inval;
7692			goto out;
7693	}
7694
7695	lo = find_lockowner_str(cstate->clp, &lockt->lt_owner);
7696	if (lo)
7697		file_lock->fl_owner = (fl_owner_t)lo;
7698	file_lock->fl_pid = current->tgid;
7699	file_lock->fl_flags = FL_POSIX;
7700
7701	file_lock->fl_start = lockt->lt_offset;
7702	file_lock->fl_end = last_byte_offset(lockt->lt_offset, lockt->lt_length);
7703
7704	nfs4_transform_lock_offset(file_lock);
7705
7706	status = nfsd_test_lock(rqstp, &cstate->current_fh, file_lock);
7707	if (status)
7708		goto out;
7709
7710	if (file_lock->fl_type != F_UNLCK) {
7711		status = nfserr_denied;
7712		nfs4_set_lock_denied(file_lock, &lockt->lt_denied);
7713	}
7714out:
7715	if (lo)
7716		nfs4_put_stateowner(&lo->lo_owner);
7717	if (file_lock)
7718		locks_free_lock(file_lock);
7719	return status;
7720}
7721
7722__be32
7723nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7724	    union nfsd4_op_u *u)
7725{
7726	struct nfsd4_locku *locku = &u->locku;
7727	struct nfs4_ol_stateid *stp;
7728	struct nfsd_file *nf = NULL;
7729	struct file_lock *file_lock = NULL;
7730	__be32 status;
7731	int err;
7732	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
7733
7734	dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
7735		(long long) locku->lu_offset,
7736		(long long) locku->lu_length);
7737
7738	if (check_lock_length(locku->lu_offset, locku->lu_length))
7739		 return nfserr_inval;
7740
7741	status = nfs4_preprocess_seqid_op(cstate, locku->lu_seqid,
7742					&locku->lu_stateid, NFS4_LOCK_STID,
7743					&stp, nn);
7744	if (status)
7745		goto out;
7746	nf = find_any_file(stp->st_stid.sc_file);
7747	if (!nf) {
7748		status = nfserr_lock_range;
7749		goto put_stateid;
7750	}
7751	file_lock = locks_alloc_lock();
7752	if (!file_lock) {
7753		dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
7754		status = nfserr_jukebox;
7755		goto put_file;
7756	}
7757
7758	file_lock->fl_type = F_UNLCK;
7759	file_lock->fl_owner = (fl_owner_t)lockowner(nfs4_get_stateowner(stp->st_stateowner));
7760	file_lock->fl_pid = current->tgid;
7761	file_lock->fl_file = nf->nf_file;
7762	file_lock->fl_flags = FL_POSIX;
7763	file_lock->fl_lmops = &nfsd_posix_mng_ops;
7764	file_lock->fl_start = locku->lu_offset;
7765
7766	file_lock->fl_end = last_byte_offset(locku->lu_offset,
7767						locku->lu_length);
7768	nfs4_transform_lock_offset(file_lock);
7769
7770	err = vfs_lock_file(nf->nf_file, F_SETLK, file_lock, NULL);
7771	if (err) {
7772		dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
7773		goto out_nfserr;
7774	}
7775	nfs4_inc_and_copy_stateid(&locku->lu_stateid, &stp->st_stid);
7776put_file:
7777	nfsd_file_put(nf);
7778put_stateid:
7779	mutex_unlock(&stp->st_mutex);
7780	nfs4_put_stid(&stp->st_stid);
7781out:
7782	nfsd4_bump_seqid(cstate, status);
7783	if (file_lock)
7784		locks_free_lock(file_lock);
7785	return status;
7786
7787out_nfserr:
7788	status = nfserrno(err);
7789	goto put_file;
7790}
7791
7792/*
7793 * returns
7794 * 	true:  locks held by lockowner
7795 * 	false: no locks held by lockowner
7796 */
7797static bool
7798check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
7799{
7800	struct file_lock *fl;
7801	int status = false;
7802	struct nfsd_file *nf = find_any_file(fp);
7803	struct inode *inode;
7804	struct file_lock_context *flctx;
7805
7806	if (!nf) {
7807		/* Any valid lock stateid should have some sort of access */
7808		WARN_ON_ONCE(1);
7809		return status;
7810	}
7811
7812	inode = locks_inode(nf->nf_file);
7813	flctx = locks_inode_context(inode);
7814
7815	if (flctx && !list_empty_careful(&flctx->flc_posix)) {
7816		spin_lock(&flctx->flc_lock);
7817		list_for_each_entry(fl, &flctx->flc_posix, fl_list) {
7818			if (fl->fl_owner == (fl_owner_t)lowner) {
7819				status = true;
7820				break;
7821			}
7822		}
7823		spin_unlock(&flctx->flc_lock);
7824	}
7825	nfsd_file_put(nf);
7826	return status;
7827}
7828
7829/**
7830 * nfsd4_release_lockowner - process NFSv4.0 RELEASE_LOCKOWNER operations
7831 * @rqstp: RPC transaction
7832 * @cstate: NFSv4 COMPOUND state
7833 * @u: RELEASE_LOCKOWNER arguments
7834 *
7835 * The lockowner's so_count is bumped when a lock record is added
7836 * or when copying a conflicting lock. The latter case is brief,
7837 * but can lead to fleeting false positives when looking for
7838 * locks-in-use.
7839 *
7840 * Return values:
7841 *   %nfs_ok: lockowner released or not found
7842 *   %nfserr_locks_held: lockowner still in use
7843 *   %nfserr_stale_clientid: clientid no longer active
7844 *   %nfserr_expired: clientid not recognized
7845 */
7846__be32
7847nfsd4_release_lockowner(struct svc_rqst *rqstp,
7848			struct nfsd4_compound_state *cstate,
7849			union nfsd4_op_u *u)
7850{
7851	struct nfsd4_release_lockowner *rlockowner = &u->release_lockowner;
7852	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), nfsd_net_id);
7853	clientid_t *clid = &rlockowner->rl_clientid;
 
 
7854	struct nfs4_ol_stateid *stp;
7855	struct nfs4_lockowner *lo;
7856	struct nfs4_client *clp;
7857	LIST_HEAD(reaplist);
7858	__be32 status;
 
 
 
7859
7860	dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
7861		clid->cl_boot, clid->cl_id);
7862
7863	status = set_client(clid, cstate, nn);
7864	if (status)
7865		return status;
7866	clp = cstate->clp;
7867
 
 
7868	spin_lock(&clp->cl_lock);
7869	lo = find_lockowner_str_locked(clp, &rlockowner->rl_owner);
7870	if (!lo) {
7871		spin_unlock(&clp->cl_lock);
7872		return nfs_ok;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7873	}
7874	if (atomic_read(&lo->lo_owner.so_count) != 2) {
7875		spin_unlock(&clp->cl_lock);
7876		nfs4_put_stateowner(&lo->lo_owner);
7877		return nfserr_locks_held;
7878	}
 
7879	unhash_lockowner_locked(lo);
7880	while (!list_empty(&lo->lo_owner.so_stateids)) {
7881		stp = list_first_entry(&lo->lo_owner.so_stateids,
7882				       struct nfs4_ol_stateid,
7883				       st_perstateowner);
7884		WARN_ON(!unhash_lock_stateid(stp));
7885		put_ol_stateid_locked(stp, &reaplist);
7886	}
7887	spin_unlock(&clp->cl_lock);
7888
7889	free_ol_stateid_reaplist(&reaplist);
7890	remove_blocked_locks(lo);
7891	nfs4_put_stateowner(&lo->lo_owner);
7892	return nfs_ok;
 
7893}
7894
7895static inline struct nfs4_client_reclaim *
7896alloc_reclaim(void)
7897{
7898	return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
7899}
7900
7901bool
7902nfs4_has_reclaimed_state(struct xdr_netobj name, struct nfsd_net *nn)
7903{
7904	struct nfs4_client_reclaim *crp;
7905
7906	crp = nfsd4_find_reclaim_client(name, nn);
7907	return (crp && crp->cr_clp);
7908}
7909
7910/*
7911 * failure => all reset bets are off, nfserr_no_grace...
7912 *
7913 * The caller is responsible for freeing name.data if NULL is returned (it
7914 * will be freed in nfs4_remove_reclaim_record in the normal case).
7915 */
7916struct nfs4_client_reclaim *
7917nfs4_client_to_reclaim(struct xdr_netobj name, struct xdr_netobj princhash,
7918		struct nfsd_net *nn)
7919{
7920	unsigned int strhashval;
7921	struct nfs4_client_reclaim *crp;
7922
7923	crp = alloc_reclaim();
7924	if (crp) {
7925		strhashval = clientstr_hashval(name);
7926		INIT_LIST_HEAD(&crp->cr_strhash);
7927		list_add(&crp->cr_strhash, &nn->reclaim_str_hashtbl[strhashval]);
7928		crp->cr_name.data = name.data;
7929		crp->cr_name.len = name.len;
7930		crp->cr_princhash.data = princhash.data;
7931		crp->cr_princhash.len = princhash.len;
7932		crp->cr_clp = NULL;
7933		nn->reclaim_str_hashtbl_size++;
7934	}
7935	return crp;
7936}
7937
7938void
7939nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
7940{
7941	list_del(&crp->cr_strhash);
7942	kfree(crp->cr_name.data);
7943	kfree(crp->cr_princhash.data);
7944	kfree(crp);
7945	nn->reclaim_str_hashtbl_size--;
7946}
7947
7948void
7949nfs4_release_reclaim(struct nfsd_net *nn)
7950{
7951	struct nfs4_client_reclaim *crp = NULL;
7952	int i;
7953
7954	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
7955		while (!list_empty(&nn->reclaim_str_hashtbl[i])) {
7956			crp = list_entry(nn->reclaim_str_hashtbl[i].next,
7957			                struct nfs4_client_reclaim, cr_strhash);
7958			nfs4_remove_reclaim_record(crp, nn);
7959		}
7960	}
7961	WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
7962}
7963
7964/*
7965 * called from OPEN, CLAIM_PREVIOUS with a new clientid. */
7966struct nfs4_client_reclaim *
7967nfsd4_find_reclaim_client(struct xdr_netobj name, struct nfsd_net *nn)
7968{
7969	unsigned int strhashval;
7970	struct nfs4_client_reclaim *crp = NULL;
7971
7972	strhashval = clientstr_hashval(name);
7973	list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
7974		if (compare_blob(&crp->cr_name, &name) == 0) {
7975			return crp;
7976		}
7977	}
7978	return NULL;
7979}
7980
7981__be32
7982nfs4_check_open_reclaim(struct nfs4_client *clp)
7983{
7984	if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &clp->cl_flags))
7985		return nfserr_no_grace;
7986
7987	if (nfsd4_client_record_check(clp))
7988		return nfserr_reclaim_bad;
7989
7990	return nfs_ok;
7991}
7992
7993/*
7994 * Since the lifetime of a delegation isn't limited to that of an open, a
7995 * client may quite reasonably hang on to a delegation as long as it has
7996 * the inode cached.  This becomes an obvious problem the first time a
7997 * client's inode cache approaches the size of the server's total memory.
7998 *
7999 * For now we avoid this problem by imposing a hard limit on the number
8000 * of delegations, which varies according to the server's memory size.
8001 */
8002static void
8003set_max_delegations(void)
8004{
8005	/*
8006	 * Allow at most 4 delegations per megabyte of RAM.  Quick
8007	 * estimates suggest that in the worst case (where every delegation
8008	 * is for a different inode), a delegation could take about 1.5K,
8009	 * giving a worst case usage of about 6% of memory.
8010	 */
8011	max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
8012}
8013
8014static int nfs4_state_create_net(struct net *net)
8015{
8016	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
8017	int i;
8018
8019	nn->conf_id_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
8020					    sizeof(struct list_head),
8021					    GFP_KERNEL);
8022	if (!nn->conf_id_hashtbl)
8023		goto err;
8024	nn->unconf_id_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
8025					      sizeof(struct list_head),
8026					      GFP_KERNEL);
8027	if (!nn->unconf_id_hashtbl)
8028		goto err_unconf_id;
8029	nn->sessionid_hashtbl = kmalloc_array(SESSION_HASH_SIZE,
8030					      sizeof(struct list_head),
8031					      GFP_KERNEL);
8032	if (!nn->sessionid_hashtbl)
8033		goto err_sessionid;
8034
8035	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
8036		INIT_LIST_HEAD(&nn->conf_id_hashtbl[i]);
8037		INIT_LIST_HEAD(&nn->unconf_id_hashtbl[i]);
8038	}
8039	for (i = 0; i < SESSION_HASH_SIZE; i++)
8040		INIT_LIST_HEAD(&nn->sessionid_hashtbl[i]);
8041	nn->conf_name_tree = RB_ROOT;
8042	nn->unconf_name_tree = RB_ROOT;
8043	nn->boot_time = ktime_get_real_seconds();
8044	nn->grace_ended = false;
8045	nn->nfsd4_manager.block_opens = true;
8046	INIT_LIST_HEAD(&nn->nfsd4_manager.list);
8047	INIT_LIST_HEAD(&nn->client_lru);
8048	INIT_LIST_HEAD(&nn->close_lru);
8049	INIT_LIST_HEAD(&nn->del_recall_lru);
8050	spin_lock_init(&nn->client_lock);
8051	spin_lock_init(&nn->s2s_cp_lock);
8052	idr_init(&nn->s2s_cp_stateids);
8053
8054	spin_lock_init(&nn->blocked_locks_lock);
8055	INIT_LIST_HEAD(&nn->blocked_locks_lru);
8056
8057	INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
8058	INIT_WORK(&nn->nfsd_shrinker_work, nfsd4_state_shrinker_worker);
8059	get_net(net);
8060
8061	nn->nfsd_client_shrinker.scan_objects = nfsd4_state_shrinker_scan;
8062	nn->nfsd_client_shrinker.count_objects = nfsd4_state_shrinker_count;
8063	nn->nfsd_client_shrinker.seeks = DEFAULT_SEEKS;
8064
8065	if (register_shrinker(&nn->nfsd_client_shrinker, "nfsd-client"))
8066		goto err_shrinker;
8067	return 0;
8068
8069err_shrinker:
8070	put_net(net);
8071	kfree(nn->sessionid_hashtbl);
8072err_sessionid:
8073	kfree(nn->unconf_id_hashtbl);
8074err_unconf_id:
8075	kfree(nn->conf_id_hashtbl);
8076err:
8077	return -ENOMEM;
8078}
8079
8080static void
8081nfs4_state_destroy_net(struct net *net)
8082{
8083	int i;
8084	struct nfs4_client *clp = NULL;
8085	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
8086
8087	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
8088		while (!list_empty(&nn->conf_id_hashtbl[i])) {
8089			clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
8090			destroy_client(clp);
8091		}
8092	}
8093
8094	WARN_ON(!list_empty(&nn->blocked_locks_lru));
8095
8096	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
8097		while (!list_empty(&nn->unconf_id_hashtbl[i])) {
8098			clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
8099			destroy_client(clp);
8100		}
8101	}
8102
8103	kfree(nn->sessionid_hashtbl);
8104	kfree(nn->unconf_id_hashtbl);
8105	kfree(nn->conf_id_hashtbl);
8106	put_net(net);
8107}
8108
8109int
8110nfs4_state_start_net(struct net *net)
8111{
8112	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
8113	int ret;
8114
8115	ret = nfs4_state_create_net(net);
8116	if (ret)
8117		return ret;
8118	locks_start_grace(net, &nn->nfsd4_manager);
8119	nfsd4_client_tracking_init(net);
8120	if (nn->track_reclaim_completes && nn->reclaim_str_hashtbl_size == 0)
8121		goto skip_grace;
8122	printk(KERN_INFO "NFSD: starting %lld-second grace period (net %x)\n",
8123	       nn->nfsd4_grace, net->ns.inum);
8124	trace_nfsd_grace_start(nn);
8125	queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_grace * HZ);
8126	return 0;
8127
8128skip_grace:
8129	printk(KERN_INFO "NFSD: no clients to reclaim, skipping NFSv4 grace period (net %x)\n",
8130			net->ns.inum);
8131	queue_delayed_work(laundry_wq, &nn->laundromat_work, nn->nfsd4_lease * HZ);
8132	nfsd4_end_grace(nn);
8133	return 0;
8134}
8135
8136/* initialization to perform when the nfsd service is started: */
8137
8138int
8139nfs4_state_start(void)
8140{
8141	int ret;
8142
8143	ret = rhltable_init(&nfs4_file_rhltable, &nfs4_file_rhash_params);
8144	if (ret)
8145		return ret;
8146
8147	ret = nfsd4_create_callback_queue();
8148	if (ret) {
8149		rhltable_destroy(&nfs4_file_rhltable);
8150		return ret;
8151	}
 
 
 
8152
8153	set_max_delegations();
8154	return 0;
 
 
 
 
 
8155}
8156
8157void
8158nfs4_state_shutdown_net(struct net *net)
8159{
8160	struct nfs4_delegation *dp = NULL;
8161	struct list_head *pos, *next, reaplist;
8162	struct nfsd_net *nn = net_generic(net, nfsd_net_id);
8163
8164	unregister_shrinker(&nn->nfsd_client_shrinker);
8165	cancel_work(&nn->nfsd_shrinker_work);
8166	cancel_delayed_work_sync(&nn->laundromat_work);
8167	locks_end_grace(&nn->nfsd4_manager);
8168
8169	INIT_LIST_HEAD(&reaplist);
8170	spin_lock(&state_lock);
8171	list_for_each_safe(pos, next, &nn->del_recall_lru) {
8172		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
8173		WARN_ON(!unhash_delegation_locked(dp));
8174		list_add(&dp->dl_recall_lru, &reaplist);
8175	}
8176	spin_unlock(&state_lock);
8177	list_for_each_safe(pos, next, &reaplist) {
8178		dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
8179		list_del_init(&dp->dl_recall_lru);
8180		destroy_unhashed_deleg(dp);
8181	}
8182
8183	nfsd4_client_tracking_exit(net);
8184	nfs4_state_destroy_net(net);
8185#ifdef CONFIG_NFSD_V4_2_INTER_SSC
8186	nfsd4_ssc_shutdown_umount(nn);
8187#endif
8188}
8189
8190void
8191nfs4_state_shutdown(void)
8192{
 
8193	nfsd4_destroy_callback_queue();
8194	rhltable_destroy(&nfs4_file_rhltable);
8195}
8196
8197static void
8198get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
8199{
8200	if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG) &&
8201	    CURRENT_STATEID(stateid))
8202		memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
8203}
8204
8205static void
8206put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
8207{
8208	if (cstate->minorversion) {
8209		memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
8210		SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
8211	}
8212}
8213
8214void
8215clear_current_stateid(struct nfsd4_compound_state *cstate)
8216{
8217	CLEAR_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
8218}
8219
8220/*
8221 * functions to set current state id
8222 */
8223void
8224nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate,
8225		union nfsd4_op_u *u)
8226{
8227	put_stateid(cstate, &u->open_downgrade.od_stateid);
8228}
8229
8230void
8231nfsd4_set_openstateid(struct nfsd4_compound_state *cstate,
8232		union nfsd4_op_u *u)
8233{
8234	put_stateid(cstate, &u->open.op_stateid);
8235}
8236
8237void
8238nfsd4_set_closestateid(struct nfsd4_compound_state *cstate,
8239		union nfsd4_op_u *u)
8240{
8241	put_stateid(cstate, &u->close.cl_stateid);
8242}
8243
8244void
8245nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate,
8246		union nfsd4_op_u *u)
8247{
8248	put_stateid(cstate, &u->lock.lk_resp_stateid);
8249}
8250
8251/*
8252 * functions to consume current state id
8253 */
8254
8255void
8256nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate,
8257		union nfsd4_op_u *u)
8258{
8259	get_stateid(cstate, &u->open_downgrade.od_stateid);
8260}
8261
8262void
8263nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate,
8264		union nfsd4_op_u *u)
8265{
8266	get_stateid(cstate, &u->delegreturn.dr_stateid);
8267}
8268
8269void
8270nfsd4_get_freestateid(struct nfsd4_compound_state *cstate,
8271		union nfsd4_op_u *u)
8272{
8273	get_stateid(cstate, &u->free_stateid.fr_stateid);
8274}
8275
8276void
8277nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate,
8278		union nfsd4_op_u *u)
8279{
8280	get_stateid(cstate, &u->setattr.sa_stateid);
8281}
8282
8283void
8284nfsd4_get_closestateid(struct nfsd4_compound_state *cstate,
8285		union nfsd4_op_u *u)
8286{
8287	get_stateid(cstate, &u->close.cl_stateid);
8288}
8289
8290void
8291nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate,
8292		union nfsd4_op_u *u)
8293{
8294	get_stateid(cstate, &u->locku.lu_stateid);
8295}
8296
8297void
8298nfsd4_get_readstateid(struct nfsd4_compound_state *cstate,
8299		union nfsd4_op_u *u)
8300{
8301	get_stateid(cstate, &u->read.rd_stateid);
8302}
8303
8304void
8305nfsd4_get_writestateid(struct nfsd4_compound_state *cstate,
8306		union nfsd4_op_u *u)
8307{
8308	get_stateid(cstate, &u->write.wr_stateid);
8309}