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}
63	static const stateid_t one_stateid = {
64	.si_generation = ~0,
65	.si_opaque = all_ones,
66	};
67	static const stateid_t zero_stateid = {
68	/* all fields zero */
69	};
70	static const stateid_t currentstateid = {
71	.si_generation = 1,
72	};
73	static const stateid_t close_stateid = {
74	.si_generation = 0xffffffffU,
75	};
76
77	static 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 */
85	static bool check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner);
86	static void nfs4_free_ol_stateid(struct nfs4_stid *stid);
87	static void nfsd4_end_grace(struct nfsd_net *nn);
88	static void _free_cpntf_state_locked(struct nfsd_net *nn, struct nfs4_cpntf_state *cps);
89	static void nfsd4_file_hash_remove(struct nfs4_file *fi);
90	static void deleg_reaper(struct nfsd_net *nn);
91
92	/* Locking: */
93
94	/*
95	* Currently used for the del_recall_lru and file hash table. In an
96	* effort to decrease the scope of the client_mutex, this spinlock may
97	* eventually cover more:
98	*/
99	static DEFINE_SPINLOCK(state_lock);
100
101	enum nfsd4_st_mutex_lock_subclass {
102	OPEN_STATEID_MUTEX = 0,
103	LOCK_STATEID_MUTEX = 1,
104	};
105
106	/*
107	* A waitqueue for all in-progress 4.0 CLOSE operations that are waiting for
108	* the refcount on the open stateid to drop.
109	*/
110	static DECLARE_WAIT_QUEUE_HEAD(close_wq);
111
112	/*
113	* A waitqueue where a writer to clients/#/ctl destroying a client can
114	* wait for cl_rpc_users to drop to 0 and then for the client to be
115	* unhashed.
116	*/
117	static DECLARE_WAIT_QUEUE_HEAD(expiry_wq);
118
119	static struct kmem_cache *client_slab;
120	static struct kmem_cache *openowner_slab;
121	static struct kmem_cache *lockowner_slab;
122	static struct kmem_cache *file_slab;
123	static struct kmem_cache *stateid_slab;
124	static struct kmem_cache *deleg_slab;
125	static struct kmem_cache *odstate_slab;
126
127	static void free_session(struct nfsd4_session *);
128
129	static const struct nfsd4_callback_ops nfsd4_cb_recall_ops;
130	static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops;
131	static const struct nfsd4_callback_ops nfsd4_cb_getattr_ops;
132
133	static struct workqueue_struct *laundry_wq;
134
135	int nfsd4_create_laundry_wq(void)
136	{
137	int rc = 0;
138
139	laundry_wq = alloc_workqueue("%s", WQ_UNBOUND, 0, "nfsd4");
140	if (laundry_wq == NULL)
141	rc = -ENOMEM;
142	return rc;
143	}
144
145	void nfsd4_destroy_laundry_wq(void)
146	{
147	destroy_workqueue(wq: laundry_wq);
148	}
149
150	static bool is_session_dead(struct nfsd4_session *ses)
151	{
152	return ses->se_dead;
153	}
154
155	static __be32 mark_session_dead_locked(struct nfsd4_session *ses, int ref_held_by_me)
156	{
157	if (atomic_read(v: &ses->se_ref) > ref_held_by_me)
158	return nfserr_jukebox;
159	ses->se_dead = true;
160	return nfs_ok;
161	}
162
163	static bool is_client_expired(struct nfs4_client *clp)
164	{
165	return clp->cl_time == 0;
166	}
167
168	static void nfsd4_dec_courtesy_client_count(struct nfsd_net *nn,
169	struct nfs4_client *clp)
170	{
171	if (clp->cl_state != NFSD4_ACTIVE)
172	atomic_add_unless(v: &nn->nfsd_courtesy_clients, a: -1, u: 0);
173	}
174
175	static __be32 get_client_locked(struct nfs4_client *clp)
176	{
177	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
178
179	lockdep_assert_held(&nn->client_lock);
180
181	if (is_client_expired(clp))
182	return nfserr_expired;
183	atomic_inc(v: &clp->cl_rpc_users);
184	nfsd4_dec_courtesy_client_count(nn, clp);
185	clp->cl_state = NFSD4_ACTIVE;
186	return nfs_ok;
187	}
188
189	/* must be called under the client_lock */
190	static inline void
191	renew_client_locked(struct nfs4_client *clp)
192	{
193	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
194
195	if (is_client_expired(clp)) {
196	WARN_ON(1);
197	printk("%s: client (clientid %08x/%08x) already expired\n",
198	__func__,
199	clp->cl_clientid.cl_boot,
200	clp->cl_clientid.cl_id);
201	return;
202	}
203
204	list_move_tail(list: &clp->cl_lru, head: &nn->client_lru);
205	clp->cl_time = ktime_get_boottime_seconds();
206	nfsd4_dec_courtesy_client_count(nn, clp);
207	clp->cl_state = NFSD4_ACTIVE;
208	}
209
210	static void put_client_renew_locked(struct nfs4_client *clp)
211	{
212	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
213
214	lockdep_assert_held(&nn->client_lock);
215
216	if (!atomic_dec_and_test(v: &clp->cl_rpc_users))
217	return;
218	if (!is_client_expired(clp))
219	renew_client_locked(clp);
220	else
221	wake_up_all(&expiry_wq);
222	}
223
224	static void put_client_renew(struct nfs4_client *clp)
225	{
226	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
227
228	if (!atomic_dec_and_lock(&clp->cl_rpc_users, &nn->client_lock))
229	return;
230	if (!is_client_expired(clp))
231	renew_client_locked(clp);
232	else
233	wake_up_all(&expiry_wq);
234	spin_unlock(lock: &nn->client_lock);
235	}
236
237	static __be32 nfsd4_get_session_locked(struct nfsd4_session *ses)
238	{
239	__be32 status;
240
241	if (is_session_dead(ses))
242	return nfserr_badsession;
243	status = get_client_locked(clp: ses->se_client);
244	if (status)
245	return status;
246	atomic_inc(v: &ses->se_ref);
247	return nfs_ok;
248	}
249
250	static void nfsd4_put_session_locked(struct nfsd4_session *ses)
251	{
252	struct nfs4_client *clp = ses->se_client;
253	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
254
255	lockdep_assert_held(&nn->client_lock);
256
257	if (atomic_dec_and_test(v: &ses->se_ref) && is_session_dead(ses))
258	free_session(ses);
259	put_client_renew_locked(clp);
260	}
261
262	static void nfsd4_put_session(struct nfsd4_session *ses)
263	{
264	struct nfs4_client *clp = ses->se_client;
265	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
266
267	spin_lock(lock: &nn->client_lock);
268	nfsd4_put_session_locked(ses);
269	spin_unlock(lock: &nn->client_lock);
270	}
271
272	static struct nfsd4_blocked_lock *
273	find_blocked_lock(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
274	struct nfsd_net *nn)
275	{
276	struct nfsd4_blocked_lock *cur, *found = NULL;
277
278	spin_lock(lock: &nn->blocked_locks_lock);
279	list_for_each_entry(cur, &lo->lo_blocked, nbl_list) {
280	if (fh_match(fh1: fh, fh2: &cur->nbl_fh)) {
281	list_del_init(entry: &cur->nbl_list);
282	WARN_ON(list_empty(&cur->nbl_lru));
283	list_del_init(entry: &cur->nbl_lru);
284	found = cur;
285	break;
286	}
287	}
288	spin_unlock(lock: &nn->blocked_locks_lock);
289	if (found)
290	locks_delete_block(&found->nbl_lock);
291	return found;
292	}
293
294	static struct nfsd4_blocked_lock *
295	find_or_allocate_block(struct nfs4_lockowner *lo, struct knfsd_fh *fh,
296	struct nfsd_net *nn)
297	{
298	struct nfsd4_blocked_lock *nbl;
299
300	nbl = find_blocked_lock(lo, fh, nn);
301	if (!nbl) {
302	nbl = kmalloc(sizeof(*nbl), GFP_KERNEL);
303	if (nbl) {
304	INIT_LIST_HEAD(list: &nbl->nbl_list);
305	INIT_LIST_HEAD(list: &nbl->nbl_lru);
306	fh_copy_shallow(dst: &nbl->nbl_fh, src: fh);
307	locks_init_lock(&nbl->nbl_lock);
308	kref_init(kref: &nbl->nbl_kref);
309	nfsd4_init_cb(cb: &nbl->nbl_cb, clp: lo->lo_owner.so_client,
310	ops: &nfsd4_cb_notify_lock_ops,
311	op: NFSPROC4_CLNT_CB_NOTIFY_LOCK);
312	}
313	}
314	return nbl;
315	}
316
317	static void
318	free_nbl(struct kref *kref)
319	{
320	struct nfsd4_blocked_lock *nbl;
321
322	nbl = container_of(kref, struct nfsd4_blocked_lock, nbl_kref);
323	locks_release_private(&nbl->nbl_lock);
324	kfree(objp: nbl);
325	}
326
327	static void
328	free_blocked_lock(struct nfsd4_blocked_lock *nbl)
329	{
330	locks_delete_block(&nbl->nbl_lock);
331	kref_put(kref: &nbl->nbl_kref, release: free_nbl);
332	}
333
334	static void
335	remove_blocked_locks(struct nfs4_lockowner *lo)
336	{
337	struct nfs4_client *clp = lo->lo_owner.so_client;
338	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
339	struct nfsd4_blocked_lock *nbl;
340	LIST_HEAD(reaplist);
341
342	/* Dequeue all blocked locks */
343	spin_lock(lock: &nn->blocked_locks_lock);
344	while (!list_empty(head: &lo->lo_blocked)) {
345	nbl = list_first_entry(&lo->lo_blocked,
346	struct nfsd4_blocked_lock,
347	nbl_list);
348	list_del_init(entry: &nbl->nbl_list);
349	WARN_ON(list_empty(&nbl->nbl_lru));
350	list_move(list: &nbl->nbl_lru, head: &reaplist);
351	}
352	spin_unlock(lock: &nn->blocked_locks_lock);
353
354	/* Now free them */
355	while (!list_empty(head: &reaplist)) {
356	nbl = list_first_entry(&reaplist, struct nfsd4_blocked_lock,
357	nbl_lru);
358	list_del_init(entry: &nbl->nbl_lru);
359	free_blocked_lock(nbl);
360	}
361	}
362
363	static void
364	nfsd4_cb_notify_lock_prepare(struct nfsd4_callback *cb)
365	{
366	struct nfsd4_blocked_lock *nbl = container_of(cb,
367	struct nfsd4_blocked_lock, nbl_cb);
368	locks_delete_block(&nbl->nbl_lock);
369	}
370
371	static int
372	nfsd4_cb_notify_lock_done(struct nfsd4_callback *cb, struct rpc_task *task)
373	{
374	trace_nfsd_cb_notify_lock_done(stp: &zero_stateid, task);
375
376	/*
377	* Since this is just an optimization, we don't try very hard if it
378	* turns out not to succeed. We'll requeue it on NFS4ERR_DELAY, and
379	* just quit trying on anything else.
380	*/
381	switch (task->tk_status) {
382	case -NFS4ERR_DELAY:
383	rpc_delay(task, 1 * HZ);
384	return 0;
385	default:
386	return 1;
387	}
388	}
389
390	static void
391	nfsd4_cb_notify_lock_release(struct nfsd4_callback *cb)
392	{
393	struct nfsd4_blocked_lock *nbl = container_of(cb,
394	struct nfsd4_blocked_lock, nbl_cb);
395
396	free_blocked_lock(nbl);
397	}
398
399	static const struct nfsd4_callback_ops nfsd4_cb_notify_lock_ops = {
400	.prepare = nfsd4_cb_notify_lock_prepare,
401	.done = nfsd4_cb_notify_lock_done,
402	.release = nfsd4_cb_notify_lock_release,
403	.opcode = OP_CB_NOTIFY_LOCK,
404	};
405
406	/*
407	* We store the NONE, READ, WRITE, and BOTH bits separately in the
408	* st_{access,deny}_bmap field of the stateid, in order to track not
409	* only what share bits are currently in force, but also what
410	* combinations of share bits previous opens have used. This allows us
411	* to enforce the recommendation in
412	* https://datatracker.ietf.org/doc/html/rfc7530#section-16.19.4 that
413	* the server return an error if the client attempt to downgrade to a
414	* combination of share bits not explicable by closing some of its
415	* previous opens.
416	*
417	* This enforcement is arguably incomplete, since we don't keep
418	* track of access/deny bit combinations; so, e.g., we allow:
419	*
420	* OPEN allow read, deny write
421	* OPEN allow both, deny none
422	* DOWNGRADE allow read, deny none
423	*
424	* which we should reject.
425	*
426	* But you could also argue that our current code is already overkill,
427	* since it only exists to return NFS4ERR_INVAL on incorrect client
428	* behavior.
429	*/
430	static unsigned int
431	bmap_to_share_mode(unsigned long bmap)
432	{
433	int i;
434	unsigned int access = 0;
435
436	for (i = 1; i < 4; i++) {
437	if (test_bit(i, &bmap))
438	access |= i;
439	}
440	return access;
441	}
442
443	/* set share access for a given stateid */
444	static inline void
445	set_access(u32 access, struct nfs4_ol_stateid *stp)
446	{
447	unsigned char mask = 1 << access;
448
449	WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
450	stp->st_access_bmap |= mask;
451	}
452
453	/* clear share access for a given stateid */
454	static inline void
455	clear_access(u32 access, struct nfs4_ol_stateid *stp)
456	{
457	unsigned char mask = 1 << access;
458
459	WARN_ON_ONCE(access > NFS4_SHARE_ACCESS_BOTH);
460	stp->st_access_bmap &= ~mask;
461	}
462
463	/* test whether a given stateid has access */
464	static inline bool
465	test_access(u32 access, struct nfs4_ol_stateid *stp)
466	{
467	unsigned char mask = 1 << access;
468
469	return (bool)(stp->st_access_bmap & mask);
470	}
471
472	/* set share deny for a given stateid */
473	static inline void
474	set_deny(u32 deny, struct nfs4_ol_stateid *stp)
475	{
476	unsigned char mask = 1 << deny;
477
478	WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
479	stp->st_deny_bmap |= mask;
480	}
481
482	/* clear share deny for a given stateid */
483	static inline void
484	clear_deny(u32 deny, struct nfs4_ol_stateid *stp)
485	{
486	unsigned char mask = 1 << deny;
487
488	WARN_ON_ONCE(deny > NFS4_SHARE_DENY_BOTH);
489	stp->st_deny_bmap &= ~mask;
490	}
491
492	/* test whether a given stateid is denying specific access */
493	static inline bool
494	test_deny(u32 deny, struct nfs4_ol_stateid *stp)
495	{
496	unsigned char mask = 1 << deny;
497
498	return (bool)(stp->st_deny_bmap & mask);
499	}
500
501	static int nfs4_access_to_omode(u32 access)
502	{
503	switch (access & NFS4_SHARE_ACCESS_BOTH) {
504	case NFS4_SHARE_ACCESS_READ:
505	return O_RDONLY;
506	case NFS4_SHARE_ACCESS_WRITE:
507	return O_WRONLY;
508	case NFS4_SHARE_ACCESS_BOTH:
509	return O_RDWR;
510	}
511	WARN_ON_ONCE(1);
512	return O_RDONLY;
513	}
514
515	static inline int
516	access_permit_read(struct nfs4_ol_stateid *stp)
517	{
518	return test_access(NFS4_SHARE_ACCESS_READ, stp) ||
519	test_access(NFS4_SHARE_ACCESS_BOTH, stp) ||
520	test_access(NFS4_SHARE_ACCESS_WRITE, stp);
521	}
522
523	static inline int
524	access_permit_write(struct nfs4_ol_stateid *stp)
525	{
526	return test_access(NFS4_SHARE_ACCESS_WRITE, stp) ||
527	test_access(NFS4_SHARE_ACCESS_BOTH, stp);
528	}
529
530	static inline struct nfs4_stateowner *
531	nfs4_get_stateowner(struct nfs4_stateowner *sop)
532	{
533	atomic_inc(v: &sop->so_count);
534	return sop;
535	}
536
537	static int
538	same_owner_str(struct nfs4_stateowner *sop, struct xdr_netobj *owner)
539	{
540	return (sop->so_owner.len == owner->len) &&
541	0 == memcmp(p: sop->so_owner.data, q: owner->data, size: owner->len);
542	}
543
544	static struct nfs4_openowner *
545	find_openstateowner_str(unsigned int hashval, struct nfsd4_open *open,
546	struct nfs4_client *clp)
547	{
548	struct nfs4_stateowner *so;
549
550	lockdep_assert_held(&clp->cl_lock);
551
552	list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[hashval],
553	so_strhash) {
554	if (!so->so_is_open_owner)
555	continue;
556	if (same_owner_str(sop: so, owner: &open->op_owner))
557	return openowner(so: nfs4_get_stateowner(sop: so));
558	}
559	return NULL;
560	}
561
562	static inline u32
563	opaque_hashval(const void *ptr, int nbytes)
564	{
565	unsigned char *cptr = (unsigned char *) ptr;
566
567	u32 x = 0;
568	while (nbytes--) {
569	x *= 37;
570	x += *cptr++;
571	}
572	return x;
573	}
574
575	void
576	put_nfs4_file(struct nfs4_file *fi)
577	{
578	if (refcount_dec_and_test(r: &fi->fi_ref)) {
579	nfsd4_file_hash_remove(fi);
580	WARN_ON_ONCE(!list_empty(&fi->fi_clnt_odstate));
581	WARN_ON_ONCE(!list_empty(&fi->fi_delegations));
582	kfree_rcu(fi, fi_rcu);
583	}
584	}
585
586	static struct nfsd_file *
587	find_writeable_file_locked(struct nfs4_file *f)
588	{
589	struct nfsd_file *ret;
590
591	lockdep_assert_held(&f->fi_lock);
592
593	ret = nfsd_file_get(nf: f->fi_fds[O_WRONLY]);
594	if (!ret)
595	ret = nfsd_file_get(nf: f->fi_fds[O_RDWR]);
596	return ret;
597	}
598
599	static struct nfsd_file *
600	find_writeable_file(struct nfs4_file *f)
601	{
602	struct nfsd_file *ret;
603
604	spin_lock(lock: &f->fi_lock);
605	ret = find_writeable_file_locked(f);
606	spin_unlock(lock: &f->fi_lock);
607
608	return ret;
609	}
610
611	static struct nfsd_file *
612	find_readable_file_locked(struct nfs4_file *f)
613	{
614	struct nfsd_file *ret;
615
616	lockdep_assert_held(&f->fi_lock);
617
618	ret = nfsd_file_get(nf: f->fi_fds[O_RDONLY]);
619	if (!ret)
620	ret = nfsd_file_get(nf: f->fi_fds[O_RDWR]);
621	return ret;
622	}
623
624	static struct nfsd_file *
625	find_readable_file(struct nfs4_file *f)
626	{
627	struct nfsd_file *ret;
628
629	spin_lock(lock: &f->fi_lock);
630	ret = find_readable_file_locked(f);
631	spin_unlock(lock: &f->fi_lock);
632
633	return ret;
634	}
635
636	struct nfsd_file *
637	find_any_file(struct nfs4_file *f)
638	{
639	struct nfsd_file *ret;
640
641	if (!f)
642	return NULL;
643	spin_lock(lock: &f->fi_lock);
644	ret = nfsd_file_get(nf: f->fi_fds[O_RDWR]);
645	if (!ret) {
646	ret = nfsd_file_get(nf: f->fi_fds[O_WRONLY]);
647	if (!ret)
648	ret = nfsd_file_get(nf: f->fi_fds[O_RDONLY]);
649	}
650	spin_unlock(lock: &f->fi_lock);
651	return ret;
652	}
653
654	static struct nfsd_file *find_any_file_locked(struct nfs4_file *f)
655	{
656	lockdep_assert_held(&f->fi_lock);
657
658	if (f->fi_fds[O_RDWR])
659	return f->fi_fds[O_RDWR];
660	if (f->fi_fds[O_WRONLY])
661	return f->fi_fds[O_WRONLY];
662	if (f->fi_fds[O_RDONLY])
663	return f->fi_fds[O_RDONLY];
664	return NULL;
665	}
666
667	static atomic_long_t num_delegations;
668	unsigned long max_delegations;
669
670	/*
671	* Open owner state (share locks)
672	*/
673
674	/* hash tables for lock and open owners */
675	#define OWNER_HASH_BITS 8
676	#define OWNER_HASH_SIZE (1 << OWNER_HASH_BITS)
677	#define OWNER_HASH_MASK (OWNER_HASH_SIZE - 1)
678
679	static unsigned int ownerstr_hashval(struct xdr_netobj *ownername)
680	{
681	unsigned int ret;
682
683	ret = opaque_hashval(ptr: ownername->data, nbytes: ownername->len);
684	return ret & OWNER_HASH_MASK;
685	}
686
687	static struct rhltable nfs4_file_rhltable ____cacheline_aligned_in_smp;
688
689	static const struct rhashtable_params nfs4_file_rhash_params = {
690	.key_len = sizeof_field(struct nfs4_file, fi_inode),
691	.key_offset = offsetof(struct nfs4_file, fi_inode),
692	.head_offset = offsetof(struct nfs4_file, fi_rlist),
693
694	/*
695	* Start with a single page hash table to reduce resizing churn
696	* on light workloads.
697	*/
698	.min_size = 256,
699	.automatic_shrinking = true,
700	};
701
702	/*
703	* Check if courtesy clients have conflicting access and resolve it if possible
704	*
705	* access: is op_share_access if share_access is true.
706	* Check if access mode, op_share_access, would conflict with
707	* the current deny mode of the file 'fp'.
708	* access: is op_share_deny if share_access is false.
709	* Check if the deny mode, op_share_deny, would conflict with
710	* current access of the file 'fp'.
711	* stp: skip checking this entry.
712	* new_stp: normal open, not open upgrade.
713	*
714	* Function returns:
715	* false - access/deny mode conflict with normal client.
716	* true - no conflict or conflict with courtesy client(s) is resolved.
717	*/
718	static bool
719	nfs4_resolve_deny_conflicts_locked(struct nfs4_file *fp, bool new_stp,
720	struct nfs4_ol_stateid *stp, u32 access, bool share_access)
721	{
722	struct nfs4_ol_stateid *st;
723	bool resolvable = true;
724	unsigned char bmap;
725	struct nfsd_net *nn;
726	struct nfs4_client *clp;
727
728	lockdep_assert_held(&fp->fi_lock);
729	list_for_each_entry(st, &fp->fi_stateids, st_perfile) {
730	/* ignore lock stateid */
731	if (st->st_openstp)
732	continue;
733	if (st == stp && new_stp)
734	continue;
735	/* check file access against deny mode or vice versa */
736	bmap = share_access ? st->st_deny_bmap : st->st_access_bmap;
737	if (!(access & bmap_to_share_mode(bmap)))
738	continue;
739	clp = st->st_stid.sc_client;
740	if (try_to_expire_client(clp))
741	continue;
742	resolvable = false;
743	break;
744	}
745	if (resolvable) {
746	clp = stp->st_stid.sc_client;
747	nn = net_generic(net: clp->net, id: nfsd_net_id);
748	mod_delayed_work(wq: laundry_wq, dwork: &nn->laundromat_work, delay: 0);
749	}
750	return resolvable;
751	}
752
753	static void
754	__nfs4_file_get_access(struct nfs4_file *fp, u32 access)
755	{
756	lockdep_assert_held(&fp->fi_lock);
757
758	if (access & NFS4_SHARE_ACCESS_WRITE)
759	atomic_inc(v: &fp->fi_access[O_WRONLY]);
760	if (access & NFS4_SHARE_ACCESS_READ)
761	atomic_inc(v: &fp->fi_access[O_RDONLY]);
762	}
763
764	static __be32
765	nfs4_file_get_access(struct nfs4_file *fp, u32 access)
766	{
767	lockdep_assert_held(&fp->fi_lock);
768
769	/* Does this access mode make sense? */
770	if (access & ~NFS4_SHARE_ACCESS_BOTH)
771	return nfserr_inval;
772
773	/* Does it conflict with a deny mode already set? */
774	if ((access & fp->fi_share_deny) != 0)
775	return nfserr_share_denied;
776
777	__nfs4_file_get_access(fp, access);
778	return nfs_ok;
779	}
780
781	static __be32 nfs4_file_check_deny(struct nfs4_file *fp, u32 deny)
782	{
783	/* Common case is that there is no deny mode. */
784	if (deny) {
785	/* Does this deny mode make sense? */
786	if (deny & ~NFS4_SHARE_DENY_BOTH)
787	return nfserr_inval;
788
789	if ((deny & NFS4_SHARE_DENY_READ) &&
790	atomic_read(v: &fp->fi_access[O_RDONLY]))
791	return nfserr_share_denied;
792
793	if ((deny & NFS4_SHARE_DENY_WRITE) &&
794	atomic_read(v: &fp->fi_access[O_WRONLY]))
795	return nfserr_share_denied;
796	}
797	return nfs_ok;
798	}
799
800	static void __nfs4_file_put_access(struct nfs4_file *fp, int oflag)
801	{
802	might_lock(&fp->fi_lock);
803
804	if (atomic_dec_and_lock(&fp->fi_access[oflag], &fp->fi_lock)) {
805	struct nfsd_file *f1 = NULL;
806	struct nfsd_file *f2 = NULL;
807
808	swap(f1, fp->fi_fds[oflag]);
809	if (atomic_read(v: &fp->fi_access[1 - oflag]) == 0)
810	swap(f2, fp->fi_fds[O_RDWR]);
811	spin_unlock(lock: &fp->fi_lock);
812	if (f1)
813	nfsd_file_put(nf: f1);
814	if (f2)
815	nfsd_file_put(nf: f2);
816	}
817	}
818
819	static void nfs4_file_put_access(struct nfs4_file *fp, u32 access)
820	{
821	WARN_ON_ONCE(access & ~NFS4_SHARE_ACCESS_BOTH);
822
823	if (access & NFS4_SHARE_ACCESS_WRITE)
824	__nfs4_file_put_access(fp, O_WRONLY);
825	if (access & NFS4_SHARE_ACCESS_READ)
826	__nfs4_file_put_access(fp, O_RDONLY);
827	}
828
829	/*
830	* Allocate a new open/delegation state counter. This is needed for
831	* pNFS for proper return on close semantics.
832	*
833	* Note that we only allocate it for pNFS-enabled exports, otherwise
834	* all pointers to struct nfs4_clnt_odstate are always NULL.
835	*/
836	static struct nfs4_clnt_odstate *
837	alloc_clnt_odstate(struct nfs4_client *clp)
838	{
839	struct nfs4_clnt_odstate *co;
840
841	co = kmem_cache_zalloc(odstate_slab, GFP_KERNEL);
842	if (co) {
843	co->co_client = clp;
844	refcount_set(r: &co->co_odcount, n: 1);
845	}
846	return co;
847	}
848
849	static void
850	hash_clnt_odstate_locked(struct nfs4_clnt_odstate *co)
851	{
852	struct nfs4_file *fp = co->co_file;
853
854	lockdep_assert_held(&fp->fi_lock);
855	list_add(new: &co->co_perfile, head: &fp->fi_clnt_odstate);
856	}
857
858	static inline void
859	get_clnt_odstate(struct nfs4_clnt_odstate *co)
860	{
861	if (co)
862	refcount_inc(r: &co->co_odcount);
863	}
864
865	static void
866	put_clnt_odstate(struct nfs4_clnt_odstate *co)
867	{
868	struct nfs4_file *fp;
869
870	if (!co)
871	return;
872
873	fp = co->co_file;
874	if (refcount_dec_and_lock(r: &co->co_odcount, lock: &fp->fi_lock)) {
875	list_del(entry: &co->co_perfile);
876	spin_unlock(lock: &fp->fi_lock);
877
878	nfsd4_return_all_file_layouts(clp: co->co_client, fp);
879	kmem_cache_free(s: odstate_slab, objp: co);
880	}
881	}
882
883	static struct nfs4_clnt_odstate *
884	find_or_hash_clnt_odstate(struct nfs4_file *fp, struct nfs4_clnt_odstate *new)
885	{
886	struct nfs4_clnt_odstate *co;
887	struct nfs4_client *cl;
888
889	if (!new)
890	return NULL;
891
892	cl = new->co_client;
893
894	spin_lock(lock: &fp->fi_lock);
895	list_for_each_entry(co, &fp->fi_clnt_odstate, co_perfile) {
896	if (co->co_client == cl) {
897	get_clnt_odstate(co);
898	goto out;
899	}
900	}
901	co = new;
902	co->co_file = fp;
903	hash_clnt_odstate_locked(co: new);
904	out:
905	spin_unlock(lock: &fp->fi_lock);
906	return co;
907	}
908
909	struct nfs4_stid *nfs4_alloc_stid(struct nfs4_client *cl, struct kmem_cache *slab,
910	void (*sc_free)(struct nfs4_stid *))
911	{
912	struct nfs4_stid *stid;
913	int new_id;
914
915	stid = kmem_cache_zalloc(slab, GFP_KERNEL);
916	if (!stid)
917	return NULL;
918
919	idr_preload(GFP_KERNEL);
920	spin_lock(lock: &cl->cl_lock);
921	/* Reserving 0 for start of file in nfsdfs "states" file: */
922	new_id = idr_alloc_cyclic(&cl->cl_stateids, ptr: stid, start: 1, end: 0, GFP_NOWAIT);
923	spin_unlock(lock: &cl->cl_lock);
924	idr_preload_end();
925	if (new_id < 0)
926	goto out_free;
927
928	stid->sc_free = sc_free;
929	stid->sc_client = cl;
930	stid->sc_stateid.si_opaque.so_id = new_id;
931	stid->sc_stateid.si_opaque.so_clid = cl->cl_clientid;
932	/* Will be incremented before return to client: */
933	refcount_set(r: &stid->sc_count, n: 1);
934	spin_lock_init(&stid->sc_lock);
935	INIT_LIST_HEAD(list: &stid->sc_cp_list);
936
937	return stid;
938	out_free:
939	kmem_cache_free(s: slab, objp: stid);
940	return NULL;
941	}
942
943	/*
944	* Create a unique stateid_t to represent each COPY.
945	*/
946	static int nfs4_init_cp_state(struct nfsd_net *nn, copy_stateid_t *stid,
947	unsigned char cs_type)
948	{
949	int new_id;
950
951	stid->cs_stid.si_opaque.so_clid.cl_boot = (u32)nn->boot_time;
952	stid->cs_stid.si_opaque.so_clid.cl_id = nn->s2s_cp_cl_id;
953
954	idr_preload(GFP_KERNEL);
955	spin_lock(lock: &nn->s2s_cp_lock);
956	new_id = idr_alloc_cyclic(&nn->s2s_cp_stateids, ptr: stid, start: 0, end: 0, GFP_NOWAIT);
957	stid->cs_stid.si_opaque.so_id = new_id;
958	stid->cs_stid.si_generation = 1;
959	spin_unlock(lock: &nn->s2s_cp_lock);
960	idr_preload_end();
961	if (new_id < 0)
962	return 0;
963	stid->cs_type = cs_type;
964	return 1;
965	}
966
967	int nfs4_init_copy_state(struct nfsd_net *nn, struct nfsd4_copy *copy)
968	{
969	return nfs4_init_cp_state(nn, stid: &copy->cp_stateid, NFS4_COPY_STID);
970	}
971
972	struct nfs4_cpntf_state *nfs4_alloc_init_cpntf_state(struct nfsd_net *nn,
973	struct nfs4_stid *p_stid)
974	{
975	struct nfs4_cpntf_state *cps;
976
977	cps = kzalloc(sizeof(struct nfs4_cpntf_state), GFP_KERNEL);
978	if (!cps)
979	return NULL;
980	cps->cpntf_time = ktime_get_boottime_seconds();
981	refcount_set(r: &cps->cp_stateid.cs_count, n: 1);
982	if (!nfs4_init_cp_state(nn, stid: &cps->cp_stateid, NFS4_COPYNOTIFY_STID))
983	goto out_free;
984	spin_lock(lock: &nn->s2s_cp_lock);
985	list_add(new: &cps->cp_list, head: &p_stid->sc_cp_list);
986	spin_unlock(lock: &nn->s2s_cp_lock);
987	return cps;
988	out_free:
989	kfree(objp: cps);
990	return NULL;
991	}
992
993	void nfs4_free_copy_state(struct nfsd4_copy *copy)
994	{
995	struct nfsd_net *nn;
996
997	if (copy->cp_stateid.cs_type != NFS4_COPY_STID)
998	return;
999	nn = net_generic(net: copy->cp_clp->net, id: nfsd_net_id);
1000	spin_lock(lock: &nn->s2s_cp_lock);
1001	idr_remove(&nn->s2s_cp_stateids,
1002	id: copy->cp_stateid.cs_stid.si_opaque.so_id);
1003	spin_unlock(lock: &nn->s2s_cp_lock);
1004	}
1005
1006	static void nfs4_free_cpntf_statelist(struct net *net, struct nfs4_stid *stid)
1007	{
1008	struct nfs4_cpntf_state *cps;
1009	struct nfsd_net *nn;
1010
1011	nn = net_generic(net, id: nfsd_net_id);
1012	spin_lock(lock: &nn->s2s_cp_lock);
1013	while (!list_empty(head: &stid->sc_cp_list)) {
1014	cps = list_first_entry(&stid->sc_cp_list,
1015	struct nfs4_cpntf_state, cp_list);
1016	_free_cpntf_state_locked(nn, cps);
1017	}
1018	spin_unlock(lock: &nn->s2s_cp_lock);
1019	}
1020
1021	static struct nfs4_ol_stateid * nfs4_alloc_open_stateid(struct nfs4_client *clp)
1022	{
1023	struct nfs4_stid *stid;
1024
1025	stid = nfs4_alloc_stid(cl: clp, slab: stateid_slab, sc_free: nfs4_free_ol_stateid);
1026	if (!stid)
1027	return NULL;
1028
1029	return openlockstateid(s: stid);
1030	}
1031
1032	/*
1033	* As the sc_free callback of deleg, this may be called by nfs4_put_stid
1034	* in nfsd_break_one_deleg.
1035	* Considering nfsd_break_one_deleg is called with the flc->flc_lock held,
1036	* this function mustn't ever sleep.
1037	*/
1038	static void nfs4_free_deleg(struct nfs4_stid *stid)
1039	{
1040	struct nfs4_delegation *dp = delegstateid(s: stid);
1041
1042	WARN_ON_ONCE(!list_empty(&stid->sc_cp_list));
1043	WARN_ON_ONCE(!list_empty(&dp->dl_perfile));
1044	WARN_ON_ONCE(!list_empty(&dp->dl_perclnt));
1045	WARN_ON_ONCE(!list_empty(&dp->dl_recall_lru));
1046	kmem_cache_free(s: deleg_slab, objp: stid);
1047	atomic_long_dec(v: &num_delegations);
1048	}
1049
1050	/*
1051	* When we recall a delegation, we should be careful not to hand it
1052	* out again straight away.
1053	* To ensure this we keep a pair of bloom filters ('new' and 'old')
1054	* in which the filehandles of recalled delegations are "stored".
1055	* If a filehandle appear in either filter, a delegation is blocked.
1056	* When a delegation is recalled, the filehandle is stored in the "new"
1057	* filter.
1058	* Every 30 seconds we swap the filters and clear the "new" one,
1059	* unless both are empty of course. This results in delegations for a
1060	* given filehandle being blocked for between 30 and 60 seconds.
1061	*
1062	* Each filter is 256 bits. We hash the filehandle to 32bit and use the
1063	* low 3 bytes as hash-table indices.
1064	*
1065	* 'blocked_delegations_lock', which is always taken in block_delegations(),
1066	* is used to manage concurrent access. Testing does not need the lock
1067	* except when swapping the two filters.
1068	*/
1069	static DEFINE_SPINLOCK(blocked_delegations_lock);
1070	static struct bloom_pair {
1071	int entries, old_entries;
1072	time64_t swap_time;
1073	int new; /* index into 'set' */
1074	DECLARE_BITMAP(set[2], 256);
1075	} blocked_delegations;
1076
1077	static int delegation_blocked(struct knfsd_fh *fh)
1078	{
1079	u32 hash;
1080	struct bloom_pair *bd = &blocked_delegations;
1081
1082	if (bd->entries == 0)
1083	return 0;
1084	if (ktime_get_seconds() - bd->swap_time > 30) {
1085	spin_lock(lock: &blocked_delegations_lock);
1086	if (ktime_get_seconds() - bd->swap_time > 30) {
1087	bd->entries -= bd->old_entries;
1088	bd->old_entries = bd->entries;
1089	bd->new = 1-bd->new;
1090	memset(bd->set[bd->new], 0,
1091	sizeof(bd->set[0]));
1092	bd->swap_time = ktime_get_seconds();
1093	}
1094	spin_unlock(lock: &blocked_delegations_lock);
1095	}
1096	hash = jhash(key: &fh->fh_raw, length: fh->fh_size, initval: 0);
1097	if (test_bit(hash&255, bd->set[0]) &&
1098	test_bit((hash>>8)&255, bd->set[0]) &&
1099	test_bit((hash>>16)&255, bd->set[0]))
1100	return 1;
1101
1102	if (test_bit(hash&255, bd->set[1]) &&
1103	test_bit((hash>>8)&255, bd->set[1]) &&
1104	test_bit((hash>>16)&255, bd->set[1]))
1105	return 1;
1106
1107	return 0;
1108	}
1109
1110	static void block_delegations(struct knfsd_fh *fh)
1111	{
1112	u32 hash;
1113	struct bloom_pair *bd = &blocked_delegations;
1114
1115	hash = jhash(key: &fh->fh_raw, length: fh->fh_size, initval: 0);
1116
1117	spin_lock(lock: &blocked_delegations_lock);
1118	__set_bit(hash&255, bd->set[bd->new]);
1119	__set_bit((hash>>8)&255, bd->set[bd->new]);
1120	__set_bit((hash>>16)&255, bd->set[bd->new]);
1121	if (bd->entries == 0)
1122	bd->swap_time = ktime_get_seconds();
1123	bd->entries += 1;
1124	spin_unlock(lock: &blocked_delegations_lock);
1125	}
1126
1127	static struct nfs4_delegation *
1128	alloc_init_deleg(struct nfs4_client *clp, struct nfs4_file *fp,
1129	struct nfs4_clnt_odstate *odstate, u32 dl_type)
1130	{
1131	struct nfs4_delegation *dp;
1132	struct nfs4_stid *stid;
1133	long n;
1134
1135	dprintk("NFSD alloc_init_deleg\n");
1136	n = atomic_long_inc_return(v: &num_delegations);
1137	if (n < 0 || n > max_delegations)
1138	goto out_dec;
1139	if (delegation_blocked(fh: &fp->fi_fhandle))
1140	goto out_dec;
1141	stid = nfs4_alloc_stid(cl: clp, slab: deleg_slab, sc_free: nfs4_free_deleg);
1142	if (stid == NULL)
1143	goto out_dec;
1144	dp = delegstateid(s: stid);
1145
1146	/*
1147	* delegation seqid's are never incremented. The 4.1 special
1148	* meaning of seqid 0 isn't meaningful, really, but let's avoid
1149	* 0 anyway just for consistency and use 1:
1150	*/
1151	dp->dl_stid.sc_stateid.si_generation = 1;
1152	INIT_LIST_HEAD(list: &dp->dl_perfile);
1153	INIT_LIST_HEAD(list: &dp->dl_perclnt);
1154	INIT_LIST_HEAD(list: &dp->dl_recall_lru);
1155	dp->dl_clnt_odstate = odstate;
1156	get_clnt_odstate(co: odstate);
1157	dp->dl_type = dl_type;
1158	dp->dl_retries = 1;
1159	dp->dl_recalled = false;
1160	nfsd4_init_cb(cb: &dp->dl_recall, clp: dp->dl_stid.sc_client,
1161	ops: &nfsd4_cb_recall_ops, op: NFSPROC4_CLNT_CB_RECALL);
1162	nfsd4_init_cb(cb: &dp->dl_cb_fattr.ncf_getattr, clp: dp->dl_stid.sc_client,
1163	ops: &nfsd4_cb_getattr_ops, op: NFSPROC4_CLNT_CB_GETATTR);
1164	dp->dl_cb_fattr.ncf_file_modified = false;
1165	get_nfs4_file(fi: fp);
1166	dp->dl_stid.sc_file = fp;
1167	return dp;
1168	out_dec:
1169	atomic_long_dec(v: &num_delegations);
1170	return NULL;
1171	}
1172
1173	void
1174	nfs4_put_stid(struct nfs4_stid *s)
1175	{
1176	struct nfs4_file *fp = s->sc_file;
1177	struct nfs4_client *clp = s->sc_client;
1178
1179	might_lock(&clp->cl_lock);
1180
1181	if (!refcount_dec_and_lock(r: &s->sc_count, lock: &clp->cl_lock)) {
1182	wake_up_all(&close_wq);
1183	return;
1184	}
1185	idr_remove(&clp->cl_stateids, id: s->sc_stateid.si_opaque.so_id);
1186	if (s->sc_status & SC_STATUS_ADMIN_REVOKED)
1187	atomic_dec(v: &s->sc_client->cl_admin_revoked);
1188	nfs4_free_cpntf_statelist(net: clp->net, stid: s);
1189	spin_unlock(lock: &clp->cl_lock);
1190	s->sc_free(s);
1191	if (fp)
1192	put_nfs4_file(fi: fp);
1193	}
1194
1195	void
1196	nfs4_inc_and_copy_stateid(stateid_t *dst, struct nfs4_stid *stid)
1197	{
1198	stateid_t *src = &stid->sc_stateid;
1199
1200	spin_lock(lock: &stid->sc_lock);
1201	if (unlikely(++src->si_generation == 0))
1202	src->si_generation = 1;
1203	memcpy(dst, src, sizeof(*dst));
1204	spin_unlock(lock: &stid->sc_lock);
1205	}
1206
1207	static void put_deleg_file(struct nfs4_file *fp)
1208	{
1209	struct nfsd_file *rnf = NULL;
1210	struct nfsd_file *nf = NULL;
1211
1212	spin_lock(lock: &fp->fi_lock);
1213	if (--fp->fi_delegees == 0) {
1214	swap(nf, fp->fi_deleg_file);
1215	swap(rnf, fp->fi_rdeleg_file);
1216	}
1217	spin_unlock(lock: &fp->fi_lock);
1218
1219	if (nf)
1220	nfsd_file_put(nf);
1221	if (rnf) {
1222	nfsd_file_put(nf: rnf);
1223	nfs4_file_put_access(fp, NFS4_SHARE_ACCESS_READ);
1224	}
1225	}
1226
1227	static void nfsd4_finalize_deleg_timestamps(struct nfs4_delegation *dp, struct file *f)
1228	{
1229	struct iattr ia = { .ia_valid = ATTR_ATIME | ATTR_CTIME | ATTR_MTIME | ATTR_DELEG };
1230	struct inode *inode = file_inode(f);
1231	int ret;
1232
1233	/* don't do anything if FMODE_NOCMTIME isn't set */
1234	if ((READ_ONCE(f->f_mode) & FMODE_NOCMTIME) == 0)
1235	return;
1236
1237	spin_lock(lock: &f->f_lock);
1238	f->f_mode &= ~FMODE_NOCMTIME;
1239	spin_unlock(lock: &f->f_lock);
1240
1241	/* was it never written? */
1242	if (!dp->dl_written)
1243	return;
1244
1245	/* did it get a setattr for the timestamps at some point? */
1246	if (dp->dl_setattr)
1247	return;
1248
1249	/* Stamp everything to "now" */
1250	inode_lock(inode);
1251	ret = notify_change(&nop_mnt_idmap, f->f_path.dentry, &ia, NULL);
1252	inode_unlock(inode);
1253	if (ret) {
1254	struct inode *inode = file_inode(f);
1255
1256	pr_notice_ratelimited("Unable to update timestamps on inode %02x:%02x:%lu: %d\n",
1257	MAJOR(inode->i_sb->s_dev),
1258	MINOR(inode->i_sb->s_dev),
1259	inode->i_ino, ret);
1260	}
1261	}
1262
1263	static void nfs4_unlock_deleg_lease(struct nfs4_delegation *dp)
1264	{
1265	struct nfs4_file *fp = dp->dl_stid.sc_file;
1266	struct nfsd_file *nf = fp->fi_deleg_file;
1267
1268	WARN_ON_ONCE(!fp->fi_delegees);
1269
1270	nfsd4_finalize_deleg_timestamps(dp, f: nf->nf_file);
1271	kernel_setlease(nf->nf_file, F_UNLCK, NULL, (void **)&dp);
1272	put_deleg_file(fp);
1273	}
1274
1275	static void destroy_unhashed_deleg(struct nfs4_delegation *dp)
1276	{
1277	put_clnt_odstate(co: dp->dl_clnt_odstate);
1278	nfs4_unlock_deleg_lease(dp);
1279	nfs4_put_stid(s: &dp->dl_stid);
1280	}
1281
1282	/**
1283	* nfs4_delegation_exists - Discover if this delegation already exists
1284	* @clp: a pointer to the nfs4_client we're granting a delegation to
1285	* @fp: a pointer to the nfs4_file we're granting a delegation on
1286	*
1287	* Return:
1288	* On success: true iff an existing delegation is found
1289	*/
1290
1291	static bool
1292	nfs4_delegation_exists(struct nfs4_client *clp, struct nfs4_file *fp)
1293	{
1294	struct nfs4_delegation *searchdp = NULL;
1295	struct nfs4_client *searchclp = NULL;
1296
1297	lockdep_assert_held(&state_lock);
1298	lockdep_assert_held(&fp->fi_lock);
1299
1300	list_for_each_entry(searchdp, &fp->fi_delegations, dl_perfile) {
1301	searchclp = searchdp->dl_stid.sc_client;
1302	if (clp == searchclp) {
1303	return true;
1304	}
1305	}
1306	return false;
1307	}
1308
1309	/**
1310	* hash_delegation_locked - Add a delegation to the appropriate lists
1311	* @dp: a pointer to the nfs4_delegation we are adding.
1312	* @fp: a pointer to the nfs4_file we're granting a delegation on
1313	*
1314	* Return:
1315	* On success: NULL if the delegation was successfully hashed.
1316	*
1317	* On error: -EAGAIN if one was previously granted to this
1318	* nfs4_client for this nfs4_file. Delegation is not hashed.
1319	*
1320	*/
1321
1322	static int
1323	hash_delegation_locked(struct nfs4_delegation *dp, struct nfs4_file *fp)
1324	{
1325	struct nfs4_client *clp = dp->dl_stid.sc_client;
1326
1327	lockdep_assert_held(&state_lock);
1328	lockdep_assert_held(&fp->fi_lock);
1329	lockdep_assert_held(&clp->cl_lock);
1330
1331	if (nfs4_delegation_exists(clp, fp))
1332	return -EAGAIN;
1333	refcount_inc(r: &dp->dl_stid.sc_count);
1334	dp->dl_stid.sc_type = SC_TYPE_DELEG;
1335	list_add(new: &dp->dl_perfile, head: &fp->fi_delegations);
1336	list_add(new: &dp->dl_perclnt, head: &clp->cl_delegations);
1337	return 0;
1338	}
1339
1340	static bool delegation_hashed(struct nfs4_delegation *dp)
1341	{
1342	return !(list_empty(head: &dp->dl_perfile));
1343	}
1344
1345	static bool
1346	unhash_delegation_locked(struct nfs4_delegation *dp, unsigned short statusmask)
1347	{
1348	struct nfs4_file *fp = dp->dl_stid.sc_file;
1349
1350	lockdep_assert_held(&state_lock);
1351
1352	if (!delegation_hashed(dp))
1353	return false;
1354
1355	if (statusmask == SC_STATUS_REVOKED &&
1356	dp->dl_stid.sc_client->cl_minorversion == 0)
1357	statusmask = SC_STATUS_CLOSED;
1358	dp->dl_stid.sc_status |= statusmask;
1359	if (statusmask & SC_STATUS_ADMIN_REVOKED)
1360	atomic_inc(v: &dp->dl_stid.sc_client->cl_admin_revoked);
1361
1362	/* Ensure that deleg break won't try to requeue it */
1363	++dp->dl_time;
1364	spin_lock(lock: &fp->fi_lock);
1365	list_del_init(entry: &dp->dl_perclnt);
1366	list_del_init(entry: &dp->dl_recall_lru);
1367	list_del_init(entry: &dp->dl_perfile);
1368	spin_unlock(lock: &fp->fi_lock);
1369	return true;
1370	}
1371
1372	static void destroy_delegation(struct nfs4_delegation *dp)
1373	{
1374	bool unhashed;
1375
1376	spin_lock(lock: &state_lock);
1377	unhashed = unhash_delegation_locked(dp, SC_STATUS_CLOSED);
1378	spin_unlock(lock: &state_lock);
1379	if (unhashed)
1380	destroy_unhashed_deleg(dp);
1381	}
1382
1383	/**
1384	* revoke_delegation - perform nfs4 delegation structure cleanup
1385	* @dp: pointer to the delegation
1386	*
1387	* This function assumes that it's called either from the administrative
1388	* interface (nfsd4_revoke_states()) that's revoking a specific delegation
1389	* stateid or it's called from a laundromat thread (nfsd4_landromat()) that
1390	* determined that this specific state has expired and needs to be revoked
1391	* (both mark state with the appropriate stid sc_status mode). It is also
1392	* assumed that a reference was taken on the @dp state.
1393	*
1394	* If this function finds that the @dp state is SC_STATUS_FREED it means
1395	* that a FREE_STATEID operation for this stateid has been processed and
1396	* we can proceed to removing it from recalled list. However, if @dp state
1397	* isn't marked SC_STATUS_FREED, it means we need place it on the cl_revoked
1398	* list and wait for the FREE_STATEID to arrive from the client. At the same
1399	* time, we need to mark it as SC_STATUS_FREEABLE to indicate to the
1400	* nfsd4_free_stateid() function that this stateid has already been added
1401	* to the cl_revoked list and that nfsd4_free_stateid() is now responsible
1402	* for removing it from the list. Inspection of where the delegation state
1403	* in the revocation process is protected by the clp->cl_lock.
1404	*/
1405	static void revoke_delegation(struct nfs4_delegation *dp)
1406	{
1407	struct nfs4_client *clp = dp->dl_stid.sc_client;
1408
1409	WARN_ON(!list_empty(&dp->dl_recall_lru));
1410	WARN_ON_ONCE(dp->dl_stid.sc_client->cl_minorversion > 0 &&
1411	!(dp->dl_stid.sc_status &
1412	(SC_STATUS_REVOKED | SC_STATUS_ADMIN_REVOKED)));
1413
1414	trace_nfsd_stid_revoke(stid: &dp->dl_stid);
1415
1416	spin_lock(lock: &clp->cl_lock);
1417	if (dp->dl_stid.sc_status & SC_STATUS_FREED) {
1418	list_del_init(entry: &dp->dl_recall_lru);
1419	goto out;
1420	}
1421	list_add(new: &dp->dl_recall_lru, head: &clp->cl_revoked);
1422	dp->dl_stid.sc_status |= SC_STATUS_FREEABLE;
1423	out:
1424	spin_unlock(lock: &clp->cl_lock);
1425	destroy_unhashed_deleg(dp);
1426	}
1427
1428	/*
1429	* SETCLIENTID state
1430	*/
1431
1432	static unsigned int clientid_hashval(u32 id)
1433	{
1434	return id & CLIENT_HASH_MASK;
1435	}
1436
1437	static unsigned int clientstr_hashval(struct xdr_netobj name)
1438	{
1439	return opaque_hashval(ptr: name.data, nbytes: 8) & CLIENT_HASH_MASK;
1440	}
1441
1442	/*
1443	* A stateid that had a deny mode associated with it is being released
1444	* or downgraded. Recalculate the deny mode on the file.
1445	*/
1446	static void
1447	recalculate_deny_mode(struct nfs4_file *fp)
1448	{
1449	struct nfs4_ol_stateid *stp;
1450	u32 old_deny;
1451
1452	spin_lock(lock: &fp->fi_lock);
1453	old_deny = fp->fi_share_deny;
1454	fp->fi_share_deny = 0;
1455	list_for_each_entry(stp, &fp->fi_stateids, st_perfile) {
1456	fp->fi_share_deny |= bmap_to_share_mode(bmap: stp->st_deny_bmap);
1457	if (fp->fi_share_deny == old_deny)
1458	break;
1459	}
1460	spin_unlock(lock: &fp->fi_lock);
1461	}
1462
1463	static void
1464	reset_union_bmap_deny(u32 deny, struct nfs4_ol_stateid *stp)
1465	{
1466	int i;
1467	bool change = false;
1468
1469	for (i = 1; i < 4; i++) {
1470	if ((i & deny) != i) {
1471	change = true;
1472	clear_deny(deny: i, stp);
1473	}
1474	}
1475
1476	/* Recalculate per-file deny mode if there was a change */
1477	if (change)
1478	recalculate_deny_mode(fp: stp->st_stid.sc_file);
1479	}
1480
1481	/* release all access and file references for a given stateid */
1482	static void
1483	release_all_access(struct nfs4_ol_stateid *stp)
1484	{
1485	int i;
1486	struct nfs4_file *fp = stp->st_stid.sc_file;
1487
1488	if (fp && stp->st_deny_bmap != 0)
1489	recalculate_deny_mode(fp);
1490
1491	for (i = 1; i < 4; i++) {
1492	if (test_access(access: i, stp))
1493	nfs4_file_put_access(fp: stp->st_stid.sc_file, access: i);
1494	clear_access(access: i, stp);
1495	}
1496	}
1497
1498	static inline void nfs4_free_stateowner(struct nfs4_stateowner *sop)
1499	{
1500	kfree(objp: sop->so_owner.data);
1501	sop->so_ops->so_free(sop);
1502	}
1503
1504	static void nfs4_put_stateowner(struct nfs4_stateowner *sop)
1505	{
1506	struct nfs4_client *clp = sop->so_client;
1507
1508	might_lock(&clp->cl_lock);
1509
1510	if (!atomic_dec_and_lock(&sop->so_count, &clp->cl_lock))
1511	return;
1512	sop->so_ops->so_unhash(sop);
1513	spin_unlock(lock: &clp->cl_lock);
1514	nfs4_free_stateowner(sop);
1515	}
1516
1517	static bool
1518	nfs4_ol_stateid_unhashed(const struct nfs4_ol_stateid *stp)
1519	{
1520	return list_empty(head: &stp->st_perfile);
1521	}
1522
1523	static bool unhash_ol_stateid(struct nfs4_ol_stateid *stp)
1524	{
1525	struct nfs4_file *fp = stp->st_stid.sc_file;
1526
1527	lockdep_assert_held(&stp->st_stateowner->so_client->cl_lock);
1528
1529	if (list_empty(head: &stp->st_perfile))
1530	return false;
1531
1532	spin_lock(lock: &fp->fi_lock);
1533	list_del_init(entry: &stp->st_perfile);
1534	spin_unlock(lock: &fp->fi_lock);
1535	list_del(entry: &stp->st_perstateowner);
1536	return true;
1537	}
1538
1539	static void nfs4_free_ol_stateid(struct nfs4_stid *stid)
1540	{
1541	struct nfs4_ol_stateid *stp = openlockstateid(s: stid);
1542
1543	put_clnt_odstate(co: stp->st_clnt_odstate);
1544	release_all_access(stp);
1545	if (stp->st_stateowner)
1546	nfs4_put_stateowner(sop: stp->st_stateowner);
1547	if (!list_empty(head: &stid->sc_cp_list))
1548	nfs4_free_cpntf_statelist(net: stid->sc_client->net, stid);
1549	kmem_cache_free(s: stateid_slab, objp: stid);
1550	}
1551
1552	static void nfs4_free_lock_stateid(struct nfs4_stid *stid)
1553	{
1554	struct nfs4_ol_stateid *stp = openlockstateid(s: stid);
1555	struct nfs4_lockowner *lo = lockowner(so: stp->st_stateowner);
1556	struct nfsd_file *nf;
1557
1558	nf = find_any_file(f: stp->st_stid.sc_file);
1559	if (nf) {
1560	get_file(f: nf->nf_file);
1561	filp_close(nf->nf_file, id: (fl_owner_t)lo);
1562	nfsd_file_put(nf);
1563	}
1564	nfs4_free_ol_stateid(stid);
1565	}
1566
1567	/*
1568	* Put the persistent reference to an already unhashed generic stateid, while
1569	* holding the cl_lock. If it's the last reference, then put it onto the
1570	* reaplist for later destruction.
1571	*/
1572	static void put_ol_stateid_locked(struct nfs4_ol_stateid *stp,
1573	struct list_head *reaplist)
1574	{
1575	struct nfs4_stid *s = &stp->st_stid;
1576	struct nfs4_client *clp = s->sc_client;
1577
1578	lockdep_assert_held(&clp->cl_lock);
1579
1580	WARN_ON_ONCE(!list_empty(&stp->st_locks));
1581
1582	if (!refcount_dec_and_test(r: &s->sc_count)) {
1583	wake_up_all(&close_wq);
1584	return;
1585	}
1586
1587	idr_remove(&clp->cl_stateids, id: s->sc_stateid.si_opaque.so_id);
1588	if (s->sc_status & SC_STATUS_ADMIN_REVOKED)
1589	atomic_dec(v: &s->sc_client->cl_admin_revoked);
1590	list_add(new: &stp->st_locks, head: reaplist);
1591	}
1592
1593	static bool unhash_lock_stateid(struct nfs4_ol_stateid *stp)
1594	{
1595	lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1596
1597	if (!unhash_ol_stateid(stp))
1598	return false;
1599	list_del_init(entry: &stp->st_locks);
1600	stp->st_stid.sc_status |= SC_STATUS_CLOSED;
1601	return true;
1602	}
1603
1604	static void release_lock_stateid(struct nfs4_ol_stateid *stp)
1605	{
1606	struct nfs4_client *clp = stp->st_stid.sc_client;
1607	bool unhashed;
1608
1609	spin_lock(lock: &clp->cl_lock);
1610	unhashed = unhash_lock_stateid(stp);
1611	spin_unlock(lock: &clp->cl_lock);
1612	if (unhashed)
1613	nfs4_put_stid(s: &stp->st_stid);
1614	}
1615
1616	static void unhash_lockowner_locked(struct nfs4_lockowner *lo)
1617	{
1618	struct nfs4_client *clp = lo->lo_owner.so_client;
1619
1620	lockdep_assert_held(&clp->cl_lock);
1621
1622	list_del_init(entry: &lo->lo_owner.so_strhash);
1623	}
1624
1625	/*
1626	* Free a list of generic stateids that were collected earlier after being
1627	* fully unhashed.
1628	*/
1629	static void
1630	free_ol_stateid_reaplist(struct list_head *reaplist)
1631	{
1632	struct nfs4_ol_stateid *stp;
1633	struct nfs4_file *fp;
1634
1635	might_sleep();
1636
1637	while (!list_empty(head: reaplist)) {
1638	stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
1639	st_locks);
1640	list_del(entry: &stp->st_locks);
1641	fp = stp->st_stid.sc_file;
1642	stp->st_stid.sc_free(&stp->st_stid);
1643	if (fp)
1644	put_nfs4_file(fi: fp);
1645	}
1646	}
1647
1648	static void release_open_stateid_locks(struct nfs4_ol_stateid *open_stp,
1649	struct list_head *reaplist)
1650	{
1651	struct nfs4_ol_stateid *stp;
1652
1653	lockdep_assert_held(&open_stp->st_stid.sc_client->cl_lock);
1654
1655	while (!list_empty(head: &open_stp->st_locks)) {
1656	stp = list_entry(open_stp->st_locks.next,
1657	struct nfs4_ol_stateid, st_locks);
1658	unhash_lock_stateid(stp);
1659	put_ol_stateid_locked(stp, reaplist);
1660	}
1661	}
1662
1663	static bool unhash_open_stateid(struct nfs4_ol_stateid *stp,
1664	struct list_head *reaplist)
1665	{
1666	lockdep_assert_held(&stp->st_stid.sc_client->cl_lock);
1667
1668	if (!unhash_ol_stateid(stp))
1669	return false;
1670	release_open_stateid_locks(open_stp: stp, reaplist);
1671	return true;
1672	}
1673
1674	static void release_open_stateid(struct nfs4_ol_stateid *stp)
1675	{
1676	LIST_HEAD(reaplist);
1677
1678	spin_lock(lock: &stp->st_stid.sc_client->cl_lock);
1679	stp->st_stid.sc_status |= SC_STATUS_CLOSED;
1680	if (unhash_open_stateid(stp, reaplist: &reaplist))
1681	put_ol_stateid_locked(stp, reaplist: &reaplist);
1682	spin_unlock(lock: &stp->st_stid.sc_client->cl_lock);
1683	free_ol_stateid_reaplist(reaplist: &reaplist);
1684	}
1685
1686	static bool nfs4_openowner_unhashed(struct nfs4_openowner *oo)
1687	{
1688	lockdep_assert_held(&oo->oo_owner.so_client->cl_lock);
1689
1690	return list_empty(head: &oo->oo_owner.so_strhash) &&
1691	list_empty(head: &oo->oo_perclient);
1692	}
1693
1694	static void unhash_openowner_locked(struct nfs4_openowner *oo)
1695	{
1696	struct nfs4_client *clp = oo->oo_owner.so_client;
1697
1698	lockdep_assert_held(&clp->cl_lock);
1699
1700	list_del_init(entry: &oo->oo_owner.so_strhash);
1701	list_del_init(entry: &oo->oo_perclient);
1702	}
1703
1704	static void release_last_closed_stateid(struct nfs4_openowner *oo)
1705	{
1706	struct nfsd_net *nn = net_generic(net: oo->oo_owner.so_client->net,
1707	id: nfsd_net_id);
1708	struct nfs4_ol_stateid *s;
1709
1710	spin_lock(lock: &nn->client_lock);
1711	s = oo->oo_last_closed_stid;
1712	if (s) {
1713	list_del_init(entry: &oo->oo_close_lru);
1714	oo->oo_last_closed_stid = NULL;
1715	}
1716	spin_unlock(lock: &nn->client_lock);
1717	if (s)
1718	nfs4_put_stid(s: &s->st_stid);
1719	}
1720
1721	static void release_openowner(struct nfs4_openowner *oo)
1722	{
1723	struct nfs4_ol_stateid *stp;
1724	struct nfs4_client *clp = oo->oo_owner.so_client;
1725	LIST_HEAD(reaplist);
1726
1727	spin_lock(lock: &clp->cl_lock);
1728	unhash_openowner_locked(oo);
1729	while (!list_empty(head: &oo->oo_owner.so_stateids)) {
1730	stp = list_first_entry(&oo->oo_owner.so_stateids,
1731	struct nfs4_ol_stateid, st_perstateowner);
1732	if (unhash_open_stateid(stp, reaplist: &reaplist))
1733	put_ol_stateid_locked(stp, reaplist: &reaplist);
1734	}
1735	spin_unlock(lock: &clp->cl_lock);
1736	free_ol_stateid_reaplist(reaplist: &reaplist);
1737	release_last_closed_stateid(oo);
1738	nfs4_put_stateowner(sop: &oo->oo_owner);
1739	}
1740
1741	static struct nfs4_stid *find_one_sb_stid(struct nfs4_client *clp,
1742	struct super_block *sb,
1743	unsigned int sc_types)
1744	{
1745	unsigned long id, tmp;
1746	struct nfs4_stid *stid;
1747
1748	spin_lock(lock: &clp->cl_lock);
1749	idr_for_each_entry_ul(&clp->cl_stateids, stid, tmp, id)
1750	if ((stid->sc_type & sc_types) &&
1751	stid->sc_status == 0 &&
1752	stid->sc_file->fi_inode->i_sb == sb) {
1753	refcount_inc(r: &stid->sc_count);
1754	break;
1755	}
1756	spin_unlock(lock: &clp->cl_lock);
1757	return stid;
1758	}
1759
1760	/**
1761	* nfsd4_revoke_states - revoke all nfsv4 states associated with given filesystem
1762	* @nn: used to identify instance of nfsd (there is one per net namespace)
1763	* @sb: super_block used to identify target filesystem
1764	*
1765	* All nfs4 states (open, lock, delegation, layout) held by the server instance
1766	* and associated with a file on the given filesystem will be revoked resulting
1767	* in any files being closed and so all references from nfsd to the filesystem
1768	* being released. Thus nfsd will no longer prevent the filesystem from being
1769	* unmounted.
1770	*
1771	* The clients which own the states will subsequently being notified that the
1772	* states have been "admin-revoked".
1773	*/
1774	void nfsd4_revoke_states(struct nfsd_net *nn, struct super_block *sb)
1775	{
1776	unsigned int idhashval;
1777	unsigned int sc_types;
1778
1779	sc_types = SC_TYPE_OPEN | SC_TYPE_LOCK | SC_TYPE_DELEG | SC_TYPE_LAYOUT;
1780
1781	spin_lock(lock: &nn->client_lock);
1782	for (idhashval = 0; idhashval < CLIENT_HASH_SIZE; idhashval++) {
1783	struct list_head *head = &nn->conf_id_hashtbl[idhashval];
1784	struct nfs4_client *clp;
1785	retry:
1786	list_for_each_entry(clp, head, cl_idhash) {
1787	struct nfs4_stid *stid = find_one_sb_stid(clp, sb,
1788	sc_types);
1789	if (stid) {
1790	struct nfs4_ol_stateid *stp;
1791	struct nfs4_delegation *dp;
1792	struct nfs4_layout_stateid *ls;
1793
1794	spin_unlock(lock: &nn->client_lock);
1795	switch (stid->sc_type) {
1796	case SC_TYPE_OPEN:
1797	stp = openlockstateid(s: stid);
1798	mutex_lock_nested(lock: &stp->st_mutex,
1799	subclass: OPEN_STATEID_MUTEX);
1800
1801	spin_lock(lock: &clp->cl_lock);
1802	if (stid->sc_status == 0) {
1803	stid->sc_status |=
1804	SC_STATUS_ADMIN_REVOKED;
1805	atomic_inc(v: &clp->cl_admin_revoked);
1806	spin_unlock(lock: &clp->cl_lock);
1807	release_all_access(stp);
1808	} else
1809	spin_unlock(lock: &clp->cl_lock);
1810	mutex_unlock(lock: &stp->st_mutex);
1811	break;
1812	case SC_TYPE_LOCK:
1813	stp = openlockstateid(s: stid);
1814	mutex_lock_nested(lock: &stp->st_mutex,
1815	subclass: LOCK_STATEID_MUTEX);
1816	spin_lock(lock: &clp->cl_lock);
1817	if (stid->sc_status == 0) {
1818	struct nfs4_lockowner *lo =
1819	lockowner(so: stp->st_stateowner);
1820	struct nfsd_file *nf;
1821
1822	stid->sc_status |=
1823	SC_STATUS_ADMIN_REVOKED;
1824	atomic_inc(v: &clp->cl_admin_revoked);
1825	spin_unlock(lock: &clp->cl_lock);
1826	nf = find_any_file(f: stp->st_stid.sc_file);
1827	if (nf) {
1828	get_file(f: nf->nf_file);
1829	filp_close(nf->nf_file,
1830	id: (fl_owner_t)lo);
1831	nfsd_file_put(nf);
1832	}
1833	release_all_access(stp);
1834	} else
1835	spin_unlock(lock: &clp->cl_lock);
1836	mutex_unlock(lock: &stp->st_mutex);
1837	break;
1838	case SC_TYPE_DELEG:
1839	refcount_inc(r: &stid->sc_count);
1840	dp = delegstateid(s: stid);
1841	spin_lock(lock: &state_lock);
1842	if (!unhash_delegation_locked(
1843	dp, SC_STATUS_ADMIN_REVOKED))
1844	dp = NULL;
1845	spin_unlock(lock: &state_lock);
1846	if (dp)
1847	revoke_delegation(dp);
1848	break;
1849	case SC_TYPE_LAYOUT:
1850	ls = layoutstateid(s: stid);
1851	nfsd4_close_layout(ls);
1852	break;
1853	}
1854	nfs4_put_stid(s: stid);
1855	spin_lock(lock: &nn->client_lock);
1856	if (clp->cl_minorversion == 0)
1857	/* Allow cleanup after a lease period.
1858	* store_release ensures cleanup will
1859	* see any newly revoked states if it
1860	* sees the time updated.
1861	*/
1862	nn->nfs40_last_revoke =
1863	ktime_get_boottime_seconds();
1864	goto retry;
1865	}
1866	}
1867	}
1868	spin_unlock(lock: &nn->client_lock);
1869	}
1870
1871	static inline int
1872	hash_sessionid(struct nfs4_sessionid *sessionid)
1873	{
1874	struct nfsd4_sessionid *sid = (struct nfsd4_sessionid *)sessionid;
1875
1876	return sid->sequence % SESSION_HASH_SIZE;
1877	}
1878
1879	#ifdef CONFIG_SUNRPC_DEBUG
1880	static inline void
1881	dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1882	{
1883	u32 *ptr = (u32 *)(&sessionid->data[0]);
1884	dprintk("%s: %u:%u:%u:%u\n", fn, ptr[0], ptr[1], ptr[2], ptr[3]);
1885	}
1886	#else
1887	static inline void
1888	dump_sessionid(const char *fn, struct nfs4_sessionid *sessionid)
1889	{
1890	}
1891	#endif
1892
1893	/*
1894	* Bump the seqid on cstate->replay_owner, and clear replay_owner if it
1895	* won't be used for replay.
1896	*/
1897	void nfsd4_bump_seqid(struct nfsd4_compound_state *cstate, __be32 nfserr)
1898	{
1899	struct nfs4_stateowner *so = cstate->replay_owner;
1900
1901	if (nfserr == nfserr_replay_me)
1902	return;
1903
1904	if (!seqid_mutating_err(ntohl(nfserr))) {
1905	nfsd4_cstate_clear_replay(cstate);
1906	return;
1907	}
1908	if (!so)
1909	return;
1910	if (so->so_is_open_owner)
1911	release_last_closed_stateid(oo: openowner(so));
1912	so->so_seqid++;
1913	return;
1914	}
1915
1916	static void
1917	gen_sessionid(struct nfsd4_session *ses)
1918	{
1919	struct nfs4_client *clp = ses->se_client;
1920	struct nfsd4_sessionid *sid;
1921
1922	sid = (struct nfsd4_sessionid *)ses->se_sessionid.data;
1923	sid->clientid = clp->cl_clientid;
1924	sid->sequence = current_sessionid++;
1925	sid->reserved = 0;
1926	}
1927
1928	/*
1929	* The protocol defines ca_maxresponssize_cached to include the size of
1930	* the rpc header, but all we need to cache is the data starting after
1931	* the end of the initial SEQUENCE operation--the rest we regenerate
1932	* each time. Therefore we can advertise a ca_maxresponssize_cached
1933	* value that is the number of bytes in our cache plus a few additional
1934	* bytes. In order to stay on the safe side, and not promise more than
1935	* we can cache, those additional bytes must be the minimum possible: 24
1936	* bytes of rpc header (xid through accept state, with AUTH_NULL
1937	* verifier), 12 for the compound header (with zero-length tag), and 44
1938	* for the SEQUENCE op response:
1939	*/
1940	#define NFSD_MIN_HDR_SEQ_SZ (24 + 12 + 44)
1941
1942	static struct shrinker *nfsd_slot_shrinker;
1943	static DEFINE_SPINLOCK(nfsd_session_list_lock);
1944	static LIST_HEAD(nfsd_session_list);
1945	/* The sum of "target_slots-1" on every session. The shrinker can push this
1946	* down, though it can take a little while for the memory to actually
1947	* be freed. The "-1" is because we can never free slot 0 while the
1948	* session is active.
1949	*/
1950	static atomic_t nfsd_total_target_slots = ATOMIC_INIT(0);
1951
1952	static void
1953	free_session_slots(struct nfsd4_session *ses, int from)
1954	{
1955	int i;
1956
1957	if (from >= ses->se_fchannel.maxreqs)
1958	return;
1959
1960	for (i = from; i < ses->se_fchannel.maxreqs; i++) {
1961	struct nfsd4_slot *slot = xa_load(&ses->se_slots, index: i);
1962
1963	/*
1964	* Save the seqid in case we reactivate this slot.
1965	* This will never require a memory allocation so GFP
1966	* flag is irrelevant
1967	*/
1968	xa_store(&ses->se_slots, index: i, entry: xa_mk_value(v: slot->sl_seqid), 0);
1969	free_svc_cred(cred: &slot->sl_cred);
1970	kfree(objp: slot);
1971	}
1972	ses->se_fchannel.maxreqs = from;
1973	if (ses->se_target_maxslots > from) {
1974	int new_target = from ?: 1;
1975	atomic_sub(i: ses->se_target_maxslots - new_target, v: &nfsd_total_target_slots);
1976	ses->se_target_maxslots = new_target;
1977	}
1978	}
1979
1980	/**
1981	* reduce_session_slots - reduce the target max-slots of a session if possible
1982	* @ses: The session to affect
1983	* @dec: how much to decrease the target by
1984	*
1985	* This interface can be used by a shrinker to reduce the target max-slots
1986	* for a session so that some slots can eventually be freed.
1987	* It uses spin_trylock() as it may be called in a context where another
1988	* spinlock is held that has a dependency on client_lock. As shrinkers are
1989	* best-effort, skiping a session is client_lock is already held has no
1990	* great coast
1991	*
1992	* Return value:
1993	* The number of slots that the target was reduced by.
1994	*/
1995	static int
1996	reduce_session_slots(struct nfsd4_session *ses, int dec)
1997	{
1998	struct nfsd_net *nn = net_generic(net: ses->se_client->net,
1999	id: nfsd_net_id);
2000	int ret = 0;
2001
2002	if (ses->se_target_maxslots <= 1)
2003	return ret;
2004	if (!spin_trylock(lock: &nn->client_lock))
2005	return ret;
2006	ret = min(dec, ses->se_target_maxslots-1);
2007	ses->se_target_maxslots -= ret;
2008	atomic_sub(i: ret, v: &nfsd_total_target_slots);
2009	ses->se_slot_gen += 1;
2010	if (ses->se_slot_gen == 0) {
2011	int i;
2012	ses->se_slot_gen = 1;
2013	for (i = 0; i < ses->se_fchannel.maxreqs; i++) {
2014	struct nfsd4_slot *slot = xa_load(&ses->se_slots, index: i);
2015	slot->sl_generation = 0;
2016	}
2017	}
2018	spin_unlock(lock: &nn->client_lock);
2019	return ret;
2020	}
2021
2022	static struct nfsd4_slot *nfsd4_alloc_slot(struct nfsd4_channel_attrs *fattrs,
2023	int index, gfp_t gfp)
2024	{
2025	struct nfsd4_slot *slot;
2026	size_t size;
2027
2028	/*
2029	* The RPC and NFS session headers are never saved in
2030	* the slot reply cache buffer.
2031	*/
2032	size = fattrs->maxresp_cached < NFSD_MIN_HDR_SEQ_SZ ?
2033	0 : fattrs->maxresp_cached - NFSD_MIN_HDR_SEQ_SZ;
2034
2035	slot = kzalloc(struct_size(slot, sl_data, size), gfp);
2036	if (!slot)
2037	return NULL;
2038	slot->sl_index = index;
2039	return slot;
2040	}
2041
2042	static struct nfsd4_session *alloc_session(struct nfsd4_channel_attrs *fattrs,
2043	struct nfsd4_channel_attrs *battrs)
2044	{
2045	int numslots = fattrs->maxreqs;
2046	struct nfsd4_session *new;
2047	struct nfsd4_slot *slot;
2048	int i;
2049
2050	new = kzalloc(sizeof(*new), GFP_KERNEL);
2051	if (!new)
2052	return NULL;
2053	xa_init(xa: &new->se_slots);
2054
2055	slot = nfsd4_alloc_slot(fattrs, index: 0, GFP_KERNEL);
2056	if (!slot || xa_is_err(entry: xa_store(&new->se_slots, index: 0, entry: slot, GFP_KERNEL)))
2057	goto out_free;
2058
2059	for (i = 1; i < numslots; i++) {
2060	const gfp_t gfp = GFP_KERNEL | __GFP_NORETRY | __GFP_NOWARN;
2061	slot = nfsd4_alloc_slot(fattrs, index: i, gfp);
2062	if (!slot)
2063	break;
2064	if (xa_is_err(entry: xa_store(&new->se_slots, index: i, entry: slot, gfp))) {
2065	kfree(objp: slot);
2066	break;
2067	}
2068	}
2069	fattrs->maxreqs = i;
2070	memcpy(&new->se_fchannel, fattrs, sizeof(struct nfsd4_channel_attrs));
2071	new->se_target_maxslots = i;
2072	atomic_add(i: i - 1, v: &nfsd_total_target_slots);
2073	new->se_cb_slot_avail = ~0U;
2074	new->se_cb_highest_slot = min(battrs->maxreqs - 1,
2075	NFSD_BC_SLOT_TABLE_SIZE - 1);
2076	spin_lock_init(&new->se_lock);
2077	return new;
2078	out_free:
2079	kfree(objp: slot);
2080	xa_destroy(&new->se_slots);
2081	kfree(objp: new);
2082	return NULL;
2083	}
2084
2085	static void free_conn(struct nfsd4_conn *c)
2086	{
2087	svc_xprt_put(xprt: c->cn_xprt);
2088	kfree(objp: c);
2089	}
2090
2091	static void nfsd4_conn_lost(struct svc_xpt_user *u)
2092	{
2093	struct nfsd4_conn *c = container_of(u, struct nfsd4_conn, cn_xpt_user);
2094	struct nfs4_client *clp = c->cn_session->se_client;
2095
2096	trace_nfsd_cb_lost(clp);
2097
2098	spin_lock(lock: &clp->cl_lock);
2099	if (!list_empty(head: &c->cn_persession)) {
2100	list_del(entry: &c->cn_persession);
2101	free_conn(c);
2102	}
2103	nfsd4_probe_callback(clp);
2104	spin_unlock(lock: &clp->cl_lock);
2105	}
2106
2107	static struct nfsd4_conn *alloc_conn(struct svc_rqst *rqstp, u32 flags)
2108	{
2109	struct nfsd4_conn *conn;
2110
2111	conn = kmalloc(sizeof(struct nfsd4_conn), GFP_KERNEL);
2112	if (!conn)
2113	return NULL;
2114	svc_xprt_get(xprt: rqstp->rq_xprt);
2115	conn->cn_xprt = rqstp->rq_xprt;
2116	conn->cn_flags = flags;
2117	INIT_LIST_HEAD(list: &conn->cn_xpt_user.list);
2118	return conn;
2119	}
2120
2121	static void __nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
2122	{
2123	conn->cn_session = ses;
2124	list_add(new: &conn->cn_persession, head: &ses->se_conns);
2125	}
2126
2127	static void nfsd4_hash_conn(struct nfsd4_conn *conn, struct nfsd4_session *ses)
2128	{
2129	struct nfs4_client *clp = ses->se_client;
2130
2131	spin_lock(lock: &clp->cl_lock);
2132	__nfsd4_hash_conn(conn, ses);
2133	spin_unlock(lock: &clp->cl_lock);
2134	}
2135
2136	static int nfsd4_register_conn(struct nfsd4_conn *conn)
2137	{
2138	conn->cn_xpt_user.callback = nfsd4_conn_lost;
2139	return register_xpt_user(xpt: conn->cn_xprt, u: &conn->cn_xpt_user);
2140	}
2141
2142	static void nfsd4_init_conn(struct svc_rqst *rqstp, struct nfsd4_conn *conn, struct nfsd4_session *ses)
2143	{
2144	int ret;
2145
2146	nfsd4_hash_conn(conn, ses);
2147	ret = nfsd4_register_conn(conn);
2148	if (ret)
2149	/* oops; xprt is already down: */
2150	nfsd4_conn_lost(u: &conn->cn_xpt_user);
2151	/* We may have gained or lost a callback channel: */
2152	nfsd4_probe_callback_sync(clp: ses->se_client);
2153	}
2154
2155	static struct nfsd4_conn *alloc_conn_from_crses(struct svc_rqst *rqstp, struct nfsd4_create_session *cses)
2156	{
2157	u32 dir = NFS4_CDFC4_FORE;
2158
2159	if (cses->flags & SESSION4_BACK_CHAN)
2160	dir |= NFS4_CDFC4_BACK;
2161	return alloc_conn(rqstp, flags: dir);
2162	}
2163
2164	/* must be called under client_lock */
2165	static void nfsd4_del_conns(struct nfsd4_session *s)
2166	{
2167	struct nfs4_client *clp = s->se_client;
2168	struct nfsd4_conn *c;
2169
2170	spin_lock(lock: &clp->cl_lock);
2171	while (!list_empty(head: &s->se_conns)) {
2172	c = list_first_entry(&s->se_conns, struct nfsd4_conn, cn_persession);
2173	list_del_init(entry: &c->cn_persession);
2174	spin_unlock(lock: &clp->cl_lock);
2175
2176	unregister_xpt_user(xpt: c->cn_xprt, u: &c->cn_xpt_user);
2177	free_conn(c);
2178
2179	spin_lock(lock: &clp->cl_lock);
2180	}
2181	spin_unlock(lock: &clp->cl_lock);
2182	}
2183
2184	static void __free_session(struct nfsd4_session *ses)
2185	{
2186	free_session_slots(ses, from: 0);
2187	xa_destroy(&ses->se_slots);
2188	kfree(objp: ses);
2189	}
2190
2191	static void free_session(struct nfsd4_session *ses)
2192	{
2193	nfsd4_del_conns(s: ses);
2194	__free_session(ses);
2195	}
2196
2197	static unsigned long
2198	nfsd_slot_count(struct shrinker *s, struct shrink_control *sc)
2199	{
2200	unsigned long cnt = atomic_read(v: &nfsd_total_target_slots);
2201
2202	return cnt ? cnt : SHRINK_EMPTY;
2203	}
2204
2205	static unsigned long
2206	nfsd_slot_scan(struct shrinker *s, struct shrink_control *sc)
2207	{
2208	struct nfsd4_session *ses;
2209	unsigned long scanned = 0;
2210	unsigned long freed = 0;
2211
2212	spin_lock(lock: &nfsd_session_list_lock);
2213	list_for_each_entry(ses, &nfsd_session_list, se_all_sessions) {
2214	freed += reduce_session_slots(ses, dec: 1);
2215	scanned += 1;
2216	if (scanned >= sc->nr_to_scan) {
2217	/* Move starting point for next scan */
2218	list_move(list: &nfsd_session_list, head: &ses->se_all_sessions);
2219	break;
2220	}
2221	}
2222	spin_unlock(lock: &nfsd_session_list_lock);
2223	sc->nr_scanned = scanned;
2224	return freed;
2225	}
2226
2227	static void init_session(struct svc_rqst *rqstp, struct nfsd4_session *new, struct nfs4_client *clp, struct nfsd4_create_session *cses)
2228	{
2229	int idx;
2230	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
2231
2232	new->se_client = clp;
2233	gen_sessionid(ses: new);
2234
2235	INIT_LIST_HEAD(list: &new->se_conns);
2236
2237	atomic_set(v: &new->se_ref, i: 0);
2238	new->se_dead = false;
2239	new->se_cb_prog = cses->callback_prog;
2240	new->se_cb_sec = cses->cb_sec;
2241
2242	for (idx = 0; idx < NFSD_BC_SLOT_TABLE_SIZE; ++idx)
2243	new->se_cb_seq_nr[idx] = 1;
2244
2245	idx = hash_sessionid(sessionid: &new->se_sessionid);
2246	list_add(new: &new->se_hash, head: &nn->sessionid_hashtbl[idx]);
2247	spin_lock(lock: &clp->cl_lock);
2248	list_add(new: &new->se_perclnt, head: &clp->cl_sessions);
2249	spin_unlock(lock: &clp->cl_lock);
2250
2251	spin_lock(lock: &nfsd_session_list_lock);
2252	list_add_tail(new: &new->se_all_sessions, head: &nfsd_session_list);
2253	spin_unlock(lock: &nfsd_session_list_lock);
2254
2255	{
2256	struct sockaddr *sa = svc_addr(rqst: rqstp);
2257	/*
2258	* This is a little silly; with sessions there's no real
2259	* use for the callback address. Use the peer address
2260	* as a reasonable default for now, but consider fixing
2261	* the rpc client not to require an address in the
2262	* future:
2263	*/
2264	rpc_copy_addr(dst: (struct sockaddr *)&clp->cl_cb_conn.cb_addr, src: sa);
2265	clp->cl_cb_conn.cb_addrlen = svc_addr_len(sa);
2266	}
2267	}
2268
2269	/* caller must hold client_lock */
2270	static struct nfsd4_session *
2271	__find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net)
2272	{
2273	struct nfsd4_session *elem;
2274	int idx;
2275	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
2276
2277	lockdep_assert_held(&nn->client_lock);
2278
2279	dump_sessionid(fn: __func__, sessionid);
2280	idx = hash_sessionid(sessionid);
2281	/* Search in the appropriate list */
2282	list_for_each_entry(elem, &nn->sessionid_hashtbl[idx], se_hash) {
2283	if (!memcmp(p: elem->se_sessionid.data, q: sessionid->data,
2284	NFS4_MAX_SESSIONID_LEN)) {
2285	return elem;
2286	}
2287	}
2288
2289	dprintk("%s: session not found\n", __func__);
2290	return NULL;
2291	}
2292
2293	static struct nfsd4_session *
2294	find_in_sessionid_hashtbl(struct nfs4_sessionid *sessionid, struct net *net,
2295	__be32 *ret)
2296	{
2297	struct nfsd4_session *session;
2298	__be32 status = nfserr_badsession;
2299
2300	session = __find_in_sessionid_hashtbl(sessionid, net);
2301	if (!session)
2302	goto out;
2303	status = nfsd4_get_session_locked(ses: session);
2304	if (status)
2305	session = NULL;
2306	out:
2307	*ret = status;
2308	return session;
2309	}
2310
2311	/* caller must hold client_lock */
2312	static void
2313	unhash_session(struct nfsd4_session *ses)
2314	{
2315	struct nfs4_client *clp = ses->se_client;
2316	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
2317
2318	lockdep_assert_held(&nn->client_lock);
2319
2320	list_del(entry: &ses->se_hash);
2321	spin_lock(lock: &ses->se_client->cl_lock);
2322	list_del(entry: &ses->se_perclnt);
2323	spin_unlock(lock: &ses->se_client->cl_lock);
2324	spin_lock(lock: &nfsd_session_list_lock);
2325	list_del(entry: &ses->se_all_sessions);
2326	spin_unlock(lock: &nfsd_session_list_lock);
2327	}
2328
2329	/* SETCLIENTID and SETCLIENTID_CONFIRM Helper functions */
2330	static int
2331	STALE_CLIENTID(clientid_t *clid, struct nfsd_net *nn)
2332	{
2333	/*
2334	* We're assuming the clid was not given out from a boot
2335	* precisely 2^32 (about 136 years) before this one. That seems
2336	* a safe assumption:
2337	*/
2338	if (clid->cl_boot == (u32)nn->boot_time)
2339	return 0;
2340	trace_nfsd_clid_stale(clid);
2341	return 1;
2342	}
2343
2344	static struct nfs4_client *alloc_client(struct xdr_netobj name,
2345	struct nfsd_net *nn)
2346	{
2347	struct nfs4_client *clp;
2348	int i;
2349
2350	if (atomic_read(v: &nn->nfs4_client_count) >= nn->nfs4_max_clients &&
2351	atomic_read(v: &nn->nfsd_courtesy_clients) > 0)
2352	mod_delayed_work(wq: laundry_wq, dwork: &nn->laundromat_work, delay: 0);
2353
2354	clp = kmem_cache_zalloc(client_slab, GFP_KERNEL);
2355	if (clp == NULL)
2356	return NULL;
2357	xdr_netobj_dup(dst: &clp->cl_name, src: &name, GFP_KERNEL);
2358	if (clp->cl_name.data == NULL)
2359	goto err_no_name;
2360	clp->cl_ownerstr_hashtbl = kmalloc_array(OWNER_HASH_SIZE,
2361	sizeof(struct list_head),
2362	GFP_KERNEL);
2363	if (!clp->cl_ownerstr_hashtbl)
2364	goto err_no_hashtbl;
2365	clp->cl_callback_wq = alloc_ordered_workqueue("nfsd4_callbacks", 0);
2366	if (!clp->cl_callback_wq)
2367	goto err_no_callback_wq;
2368
2369	for (i = 0; i < OWNER_HASH_SIZE; i++)
2370	INIT_LIST_HEAD(list: &clp->cl_ownerstr_hashtbl[i]);
2371	INIT_LIST_HEAD(list: &clp->cl_sessions);
2372	idr_init(idr: &clp->cl_stateids);
2373	atomic_set(v: &clp->cl_rpc_users, i: 0);
2374	clp->cl_cb_state = NFSD4_CB_UNKNOWN;
2375	clp->cl_state = NFSD4_ACTIVE;
2376	atomic_inc(v: &nn->nfs4_client_count);
2377	atomic_set(v: &clp->cl_delegs_in_recall, i: 0);
2378	INIT_LIST_HEAD(list: &clp->cl_idhash);
2379	INIT_LIST_HEAD(list: &clp->cl_openowners);
2380	INIT_LIST_HEAD(list: &clp->cl_delegations);
2381	INIT_LIST_HEAD(list: &clp->cl_lru);
2382	INIT_LIST_HEAD(list: &clp->cl_revoked);
2383	#ifdef CONFIG_NFSD_PNFS
2384	INIT_LIST_HEAD(list: &clp->cl_lo_states);
2385	#endif
2386	INIT_LIST_HEAD(list: &clp->async_copies);
2387	spin_lock_init(&clp->async_lock);
2388	spin_lock_init(&clp->cl_lock);
2389	rpc_init_wait_queue(&clp->cl_cb_waitq, "Backchannel slot table");
2390	return clp;
2391	err_no_callback_wq:
2392	kfree(objp: clp->cl_ownerstr_hashtbl);
2393	err_no_hashtbl:
2394	kfree(objp: clp->cl_name.data);
2395	err_no_name:
2396	kmem_cache_free(s: client_slab, objp: clp);
2397	return NULL;
2398	}
2399
2400	static void __free_client(struct kref *k)
2401	{
2402	struct nfsdfs_client *c = container_of(k, struct nfsdfs_client, cl_ref);
2403	struct nfs4_client *clp = container_of(c, struct nfs4_client, cl_nfsdfs);
2404
2405	free_svc_cred(cred: &clp->cl_cred);
2406	destroy_workqueue(wq: clp->cl_callback_wq);
2407	kfree(objp: clp->cl_ownerstr_hashtbl);
2408	kfree(objp: clp->cl_name.data);
2409	kfree(objp: clp->cl_nii_domain.data);
2410	kfree(objp: clp->cl_nii_name.data);
2411	idr_destroy(&clp->cl_stateids);
2412	kfree(objp: clp->cl_ra);
2413	kmem_cache_free(s: client_slab, objp: clp);
2414	}
2415
2416	static void drop_client(struct nfs4_client *clp)
2417	{
2418	kref_put(kref: &clp->cl_nfsdfs.cl_ref, release: __free_client);
2419	}
2420
2421	static void
2422	free_client(struct nfs4_client *clp)
2423	{
2424	while (!list_empty(head: &clp->cl_sessions)) {
2425	struct nfsd4_session *ses;
2426	ses = list_entry(clp->cl_sessions.next, struct nfsd4_session,
2427	se_perclnt);
2428	list_del(entry: &ses->se_perclnt);
2429	WARN_ON_ONCE(atomic_read(&ses->se_ref));
2430	free_session(ses);
2431	}
2432	rpc_destroy_wait_queue(&clp->cl_cb_waitq);
2433	if (clp->cl_nfsd_dentry) {
2434	nfsd_client_rmdir(dentry: clp->cl_nfsd_dentry);
2435	clp->cl_nfsd_dentry = NULL;
2436	wake_up_all(&expiry_wq);
2437	}
2438	drop_client(clp);
2439	}
2440
2441	/* must be called under the client_lock */
2442	static void
2443	unhash_client_locked(struct nfs4_client *clp)
2444	{
2445	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
2446	struct nfsd4_session *ses;
2447
2448	lockdep_assert_held(&nn->client_lock);
2449
2450	/* Mark the client as expired! */
2451	clp->cl_time = 0;
2452	/* Make it invisible */
2453	if (!list_empty(head: &clp->cl_idhash)) {
2454	list_del_init(entry: &clp->cl_idhash);
2455	if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
2456	rb_erase(&clp->cl_namenode, &nn->conf_name_tree);
2457	else
2458	rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
2459	}
2460	list_del_init(entry: &clp->cl_lru);
2461	spin_lock(lock: &clp->cl_lock);
2462	spin_lock(lock: &nfsd_session_list_lock);
2463	list_for_each_entry(ses, &clp->cl_sessions, se_perclnt) {
2464	list_del_init(entry: &ses->se_hash);
2465	list_del_init(entry: &ses->se_all_sessions);
2466	}
2467	spin_unlock(lock: &nfsd_session_list_lock);
2468	spin_unlock(lock: &clp->cl_lock);
2469	}
2470
2471	static void
2472	unhash_client(struct nfs4_client *clp)
2473	{
2474	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
2475
2476	spin_lock(lock: &nn->client_lock);
2477	unhash_client_locked(clp);
2478	spin_unlock(lock: &nn->client_lock);
2479	}
2480
2481	static __be32 mark_client_expired_locked(struct nfs4_client *clp)
2482	{
2483	int users = atomic_read(v: &clp->cl_rpc_users);
2484
2485	trace_nfsd_mark_client_expired(clp, cl_rpc_users: users);
2486
2487	if (users)
2488	return nfserr_jukebox;
2489	unhash_client_locked(clp);
2490	return nfs_ok;
2491	}
2492
2493	static void
2494	__destroy_client(struct nfs4_client *clp)
2495	{
2496	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
2497	int i;
2498	struct nfs4_openowner *oo;
2499	struct nfs4_delegation *dp;
2500	LIST_HEAD(reaplist);
2501
2502	spin_lock(lock: &state_lock);
2503	while (!list_empty(head: &clp->cl_delegations)) {
2504	dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt);
2505	unhash_delegation_locked(dp, SC_STATUS_CLOSED);
2506	list_add(new: &dp->dl_recall_lru, head: &reaplist);
2507	}
2508	spin_unlock(lock: &state_lock);
2509	while (!list_empty(head: &reaplist)) {
2510	dp = list_entry(reaplist.next, struct nfs4_delegation, dl_recall_lru);
2511	list_del_init(entry: &dp->dl_recall_lru);
2512	destroy_unhashed_deleg(dp);
2513	}
2514	while (!list_empty(head: &clp->cl_revoked)) {
2515	dp = list_entry(clp->cl_revoked.next, struct nfs4_delegation, dl_recall_lru);
2516	list_del_init(entry: &dp->dl_recall_lru);
2517	nfs4_put_stid(s: &dp->dl_stid);
2518	}
2519	while (!list_empty(head: &clp->cl_openowners)) {
2520	oo = list_entry(clp->cl_openowners.next, struct nfs4_openowner, oo_perclient);
2521	nfs4_get_stateowner(sop: &oo->oo_owner);
2522	release_openowner(oo);
2523	}
2524	for (i = 0; i < OWNER_HASH_SIZE; i++) {
2525	struct nfs4_stateowner *so, *tmp;
2526
2527	list_for_each_entry_safe(so, tmp, &clp->cl_ownerstr_hashtbl[i],
2528	so_strhash) {
2529	/* Should be no openowners at this point */
2530	WARN_ON_ONCE(so->so_is_open_owner);
2531	remove_blocked_locks(lo: lockowner(so));
2532	}
2533	}
2534	nfsd4_return_all_client_layouts(clp);
2535	nfsd4_shutdown_copy(clp);
2536	nfsd4_shutdown_callback(clp);
2537	if (clp->cl_cb_conn.cb_xprt)
2538	svc_xprt_put(xprt: clp->cl_cb_conn.cb_xprt);
2539	atomic_add_unless(v: &nn->nfs4_client_count, a: -1, u: 0);
2540	nfsd4_dec_courtesy_client_count(nn, clp);
2541	free_client(clp);
2542	wake_up_all(&expiry_wq);
2543	}
2544
2545	static void
2546	destroy_client(struct nfs4_client *clp)
2547	{
2548	unhash_client(clp);
2549	__destroy_client(clp);
2550	}
2551
2552	static void inc_reclaim_complete(struct nfs4_client *clp)
2553	{
2554	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
2555
2556	if (!nn->track_reclaim_completes)
2557	return;
2558	if (!nfsd4_find_reclaim_client(name: clp->cl_name, nn))
2559	return;
2560	if (atomic_inc_return(v: &nn->nr_reclaim_complete) ==
2561	nn->reclaim_str_hashtbl_size) {
2562	printk(KERN_INFO "NFSD: all clients done reclaiming, ending NFSv4 grace period (net %x)\n",
2563	clp->net->ns.inum);
2564	nfsd4_end_grace(nn);
2565	}
2566	}
2567
2568	static void expire_client(struct nfs4_client *clp)
2569	{
2570	unhash_client(clp);
2571	nfsd4_client_record_remove(clp);
2572	__destroy_client(clp);
2573	}
2574
2575	static void copy_verf(struct nfs4_client *target, nfs4_verifier *source)
2576	{
2577	memcpy(target->cl_verifier.data, source->data,
2578	sizeof(target->cl_verifier.data));
2579	}
2580
2581	static void copy_clid(struct nfs4_client *target, struct nfs4_client *source)
2582	{
2583	target->cl_clientid.cl_boot = source->cl_clientid.cl_boot;
2584	target->cl_clientid.cl_id = source->cl_clientid.cl_id;
2585	}
2586
2587	static int copy_cred(struct svc_cred *target, struct svc_cred *source)
2588	{
2589	target->cr_principal = kstrdup(s: source->cr_principal, GFP_KERNEL);
2590	target->cr_raw_principal = kstrdup(s: source->cr_raw_principal,
2591	GFP_KERNEL);
2592	target->cr_targ_princ = kstrdup(s: source->cr_targ_princ, GFP_KERNEL);
2593	if ((source->cr_principal && !target->cr_principal) ||
2594	(source->cr_raw_principal && !target->cr_raw_principal) ||
2595	(source->cr_targ_princ && !target->cr_targ_princ))
2596	return -ENOMEM;
2597
2598	target->cr_flavor = source->cr_flavor;
2599	target->cr_uid = source->cr_uid;
2600	target->cr_gid = source->cr_gid;
2601	target->cr_group_info = source->cr_group_info;
2602	get_group_info(gi: target->cr_group_info);
2603	target->cr_gss_mech = source->cr_gss_mech;
2604	if (source->cr_gss_mech)
2605	gss_mech_get(source->cr_gss_mech);
2606	return 0;
2607	}
2608
2609	static int
2610	compare_blob(const struct xdr_netobj *o1, const struct xdr_netobj *o2)
2611	{
2612	if (o1->len < o2->len)
2613	return -1;
2614	if (o1->len > o2->len)
2615	return 1;
2616	return memcmp(p: o1->data, q: o2->data, size: o1->len);
2617	}
2618
2619	static int
2620	same_verf(nfs4_verifier *v1, nfs4_verifier *v2)
2621	{
2622	return 0 == memcmp(p: v1->data, q: v2->data, size: sizeof(v1->data));
2623	}
2624
2625	static int
2626	same_clid(clientid_t *cl1, clientid_t *cl2)
2627	{
2628	return (cl1->cl_boot == cl2->cl_boot) && (cl1->cl_id == cl2->cl_id);
2629	}
2630
2631	static bool groups_equal(struct group_info *g1, struct group_info *g2)
2632	{
2633	int i;
2634
2635	if (g1->ngroups != g2->ngroups)
2636	return false;
2637	for (i=0; i<g1->ngroups; i++)
2638	if (!gid_eq(left: g1->gid[i], right: g2->gid[i]))
2639	return false;
2640	return true;
2641	}
2642
2643	/*
2644	* RFC 3530 language requires clid_inuse be returned when the
2645	* "principal" associated with a requests differs from that previously
2646	* used. We use uid, gid's, and gss principal string as our best
2647	* approximation. We also don't want to allow non-gss use of a client
2648	* established using gss: in theory cr_principal should catch that
2649	* change, but in practice cr_principal can be null even in the gss case
2650	* since gssd doesn't always pass down a principal string.
2651	*/
2652	static bool is_gss_cred(struct svc_cred *cr)
2653	{
2654	/* Is cr_flavor one of the gss "pseudoflavors"?: */
2655	return (cr->cr_flavor > RPC_AUTH_MAXFLAVOR);
2656	}
2657
2658
2659	static bool
2660	same_creds(struct svc_cred *cr1, struct svc_cred *cr2)
2661	{
2662	if ((is_gss_cred(cr: cr1) != is_gss_cred(cr: cr2))
2663	|| (!uid_eq(left: cr1->cr_uid, right: cr2->cr_uid))
2664	|| (!gid_eq(left: cr1->cr_gid, right: cr2->cr_gid))
2665	|| !groups_equal(g1: cr1->cr_group_info, g2: cr2->cr_group_info))
2666	return false;
2667	/* XXX: check that cr_targ_princ fields match ? */
2668	if (cr1->cr_principal == cr2->cr_principal)
2669	return true;
2670	if (!cr1->cr_principal || !cr2->cr_principal)
2671	return false;
2672	return 0 == strcmp(cr1->cr_principal, cr2->cr_principal);
2673	}
2674
2675	static bool svc_rqst_integrity_protected(struct svc_rqst *rqstp)
2676	{
2677	struct svc_cred *cr = &rqstp->rq_cred;
2678	u32 service;
2679
2680	if (!cr->cr_gss_mech)
2681	return false;
2682	service = gss_pseudoflavor_to_service(cr->cr_gss_mech, pseudoflavor: cr->cr_flavor);
2683	return service == RPC_GSS_SVC_INTEGRITY ||
2684	service == RPC_GSS_SVC_PRIVACY;
2685	}
2686
2687	bool nfsd4_mach_creds_match(struct nfs4_client *cl, struct svc_rqst *rqstp)
2688	{
2689	struct svc_cred *cr = &rqstp->rq_cred;
2690
2691	if (!cl->cl_mach_cred)
2692	return true;
2693	if (cl->cl_cred.cr_gss_mech != cr->cr_gss_mech)
2694	return false;
2695	if (!svc_rqst_integrity_protected(rqstp))
2696	return false;
2697	if (cl->cl_cred.cr_raw_principal)
2698	return 0 == strcmp(cl->cl_cred.cr_raw_principal,
2699	cr->cr_raw_principal);
2700	if (!cr->cr_principal)
2701	return false;
2702	return 0 == strcmp(cl->cl_cred.cr_principal, cr->cr_principal);
2703	}
2704
2705	static void gen_confirm(struct nfs4_client *clp, struct nfsd_net *nn)
2706	{
2707	__be32 verf[2];
2708
2709	/*
2710	* This is opaque to client, so no need to byte-swap. Use
2711	* __force to keep sparse happy
2712	*/
2713	verf[0] = (__force __be32)(u32)ktime_get_real_seconds();
2714	verf[1] = (__force __be32)nn->clverifier_counter++;
2715	memcpy(clp->cl_confirm.data, verf, sizeof(clp->cl_confirm.data));
2716	}
2717
2718	static void gen_clid(struct nfs4_client *clp, struct nfsd_net *nn)
2719	{
2720	clp->cl_clientid.cl_boot = (u32)nn->boot_time;
2721	clp->cl_clientid.cl_id = nn->clientid_counter++;
2722	gen_confirm(clp, nn);
2723	}
2724
2725	static struct nfs4_stid *
2726	find_stateid_locked(struct nfs4_client *cl, stateid_t *t)
2727	{
2728	struct nfs4_stid *ret;
2729
2730	ret = idr_find(&cl->cl_stateids, id: t->si_opaque.so_id);
2731	if (!ret || !ret->sc_type)
2732	return NULL;
2733	return ret;
2734	}
2735
2736	static struct nfs4_stid *
2737	find_stateid_by_type(struct nfs4_client *cl, stateid_t *t,
2738	unsigned short typemask, unsigned short ok_states)
2739	{
2740	struct nfs4_stid *s;
2741
2742	spin_lock(lock: &cl->cl_lock);
2743	s = find_stateid_locked(cl, t);
2744	if (s != NULL) {
2745	if ((s->sc_status & ~ok_states) == 0 &&
2746	(typemask & s->sc_type))
2747	refcount_inc(r: &s->sc_count);
2748	else
2749	s = NULL;
2750	}
2751	spin_unlock(lock: &cl->cl_lock);
2752	return s;
2753	}
2754
2755	static struct nfs4_client *get_nfsdfs_clp(struct inode *inode)
2756	{
2757	struct nfsdfs_client *nc;
2758	nc = get_nfsdfs_client(inode);
2759	if (!nc)
2760	return NULL;
2761	return container_of(nc, struct nfs4_client, cl_nfsdfs);
2762	}
2763
2764	static void seq_quote_mem(struct seq_file *m, char *data, int len)
2765	{
2766	seq_puts(m, s: "\"");
2767	seq_escape_mem(m, src: data, len, ESCAPE_HEX | ESCAPE_NAP | ESCAPE_APPEND, esc: "\"\\");
2768	seq_puts(m, s: "\"");
2769	}
2770
2771	static const char *cb_state2str(int state)
2772	{
2773	switch (state) {
2774	case NFSD4_CB_UP:
2775	return "UP";
2776	case NFSD4_CB_UNKNOWN:
2777	return "UNKNOWN";
2778	case NFSD4_CB_DOWN:
2779	return "DOWN";
2780	case NFSD4_CB_FAULT:
2781	return "FAULT";
2782	}
2783	return "UNDEFINED";
2784	}
2785
2786	static int client_info_show(struct seq_file *m, void *v)
2787	{
2788	struct inode *inode = file_inode(f: m->file);
2789	struct nfsd4_session *ses;
2790	struct nfs4_client *clp;
2791	u64 clid;
2792
2793	clp = get_nfsdfs_clp(inode);
2794	if (!clp)
2795	return -ENXIO;
2796	memcpy(&clid, &clp->cl_clientid, sizeof(clid));
2797	seq_printf(m, fmt: "clientid: 0x%llx\n", clid);
2798	seq_printf(m, fmt: "address: \"%pISpc\"\n", (struct sockaddr *)&clp->cl_addr);
2799
2800	if (clp->cl_state == NFSD4_COURTESY)
2801	seq_puts(m, s: "status: courtesy\n");
2802	else if (clp->cl_state == NFSD4_EXPIRABLE)
2803	seq_puts(m, s: "status: expirable\n");
2804	else if (test_bit(NFSD4_CLIENT_CONFIRMED, &clp->cl_flags))
2805	seq_puts(m, s: "status: confirmed\n");
2806	else
2807	seq_puts(m, s: "status: unconfirmed\n");
2808	seq_printf(m, fmt: "seconds from last renew: %lld\n",
2809	ktime_get_boottime_seconds() - clp->cl_time);
2810	seq_puts(m, s: "name: ");
2811	seq_quote_mem(m, data: clp->cl_name.data, len: clp->cl_name.len);
2812	seq_printf(m, fmt: "\nminor version: %d\n", clp->cl_minorversion);
2813	if (clp->cl_nii_domain.data) {
2814	seq_puts(m, s: "Implementation domain: ");
2815	seq_quote_mem(m, data: clp->cl_nii_domain.data,
2816	len: clp->cl_nii_domain.len);
2817	seq_puts(m, s: "\nImplementation name: ");
2818	seq_quote_mem(m, data: clp->cl_nii_name.data, len: clp->cl_nii_name.len);
2819	seq_printf(m, fmt: "\nImplementation time: [%lld, %ld]\n",
2820	clp->cl_nii_time.tv_sec, clp->cl_nii_time.tv_nsec);
2821	}
2822	seq_printf(m, fmt: "callback state: %s\n", cb_state2str(state: clp->cl_cb_state));
2823	seq_printf(m, fmt: "callback address: \"%pISpc\"\n", &clp->cl_cb_conn.cb_addr);
2824	seq_printf(m, fmt: "admin-revoked states: %d\n",
2825	atomic_read(v: &clp->cl_admin_revoked));
2826	spin_lock(lock: &clp->cl_lock);
2827	seq_printf(m, fmt: "session slots:");
2828	list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
2829	seq_printf(m, fmt: " %u", ses->se_fchannel.maxreqs);
2830	seq_printf(m, fmt: "\nsession target slots:");
2831	list_for_each_entry(ses, &clp->cl_sessions, se_perclnt)
2832	seq_printf(m, fmt: " %u", ses->se_target_maxslots);
2833	spin_unlock(lock: &clp->cl_lock);
2834	seq_puts(m, s: "\n");
2835
2836	drop_client(clp);
2837
2838	return 0;
2839	}
2840
2841	DEFINE_SHOW_ATTRIBUTE(client_info);
2842
2843	static void *states_start(struct seq_file *s, loff_t *pos)
2844	__acquires(&clp->cl_lock)
2845	{
2846	struct nfs4_client *clp = s->private;
2847	unsigned long id = *pos;
2848	void *ret;
2849
2850	spin_lock(lock: &clp->cl_lock);
2851	ret = idr_get_next_ul(&clp->cl_stateids, nextid: &id);
2852	*pos = id;
2853	return ret;
2854	}
2855
2856	static void *states_next(struct seq_file *s, void *v, loff_t *pos)
2857	{
2858	struct nfs4_client *clp = s->private;
2859	unsigned long id = *pos;
2860	void *ret;
2861
2862	id = *pos;
2863	id++;
2864	ret = idr_get_next_ul(&clp->cl_stateids, nextid: &id);
2865	*pos = id;
2866	return ret;
2867	}
2868
2869	static void states_stop(struct seq_file *s, void *v)
2870	__releases(&clp->cl_lock)
2871	{
2872	struct nfs4_client *clp = s->private;
2873
2874	spin_unlock(lock: &clp->cl_lock);
2875	}
2876
2877	static void nfs4_show_fname(struct seq_file *s, struct nfsd_file *f)
2878	{
2879	seq_printf(m: s, fmt: "filename: \"%pD2\"", f->nf_file);
2880	}
2881
2882	static void nfs4_show_superblock(struct seq_file *s, struct nfsd_file *f)
2883	{
2884	struct inode *inode = file_inode(f: f->nf_file);
2885
2886	seq_printf(m: s, fmt: "superblock: \"%02x:%02x:%ld\"",
2887	MAJOR(inode->i_sb->s_dev),
2888	MINOR(inode->i_sb->s_dev),
2889	inode->i_ino);
2890	}
2891
2892	static void nfs4_show_owner(struct seq_file *s, struct nfs4_stateowner *oo)
2893	{
2894	seq_puts(m: s, s: "owner: ");
2895	seq_quote_mem(m: s, data: oo->so_owner.data, len: oo->so_owner.len);
2896	}
2897
2898	static void nfs4_show_stateid(struct seq_file *s, stateid_t *stid)
2899	{
2900	seq_printf(m: s, fmt: "0x%.8x", stid->si_generation);
2901	seq_printf(m: s, fmt: "%12phN", &stid->si_opaque);
2902	}
2903
2904	static int nfs4_show_open(struct seq_file *s, struct nfs4_stid *st)
2905	{
2906	struct nfs4_ol_stateid *ols;
2907	struct nfs4_file *nf;
2908	struct nfsd_file *file;
2909	struct nfs4_stateowner *oo;
2910	unsigned int access, deny;
2911
2912	ols = openlockstateid(s: st);
2913	oo = ols->st_stateowner;
2914	nf = st->sc_file;
2915
2916	seq_puts(m: s, s: "- ");
2917	nfs4_show_stateid(s, stid: &st->sc_stateid);
2918	seq_puts(m: s, s: ": { type: open, ");
2919
2920	access = bmap_to_share_mode(bmap: ols->st_access_bmap);
2921	deny = bmap_to_share_mode(bmap: ols->st_deny_bmap);
2922
2923	seq_printf(m: s, fmt: "access: %s%s, ",
2924	access & NFS4_SHARE_ACCESS_READ ? "r" : "-",
2925	access & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
2926	seq_printf(m: s, fmt: "deny: %s%s, ",
2927	deny & NFS4_SHARE_ACCESS_READ ? "r" : "-",
2928	deny & NFS4_SHARE_ACCESS_WRITE ? "w" : "-");
2929
2930	if (nf) {
2931	spin_lock(lock: &nf->fi_lock);
2932	file = find_any_file_locked(f: nf);
2933	if (file) {
2934	nfs4_show_superblock(s, f: file);
2935	seq_puts(m: s, s: ", ");
2936	nfs4_show_fname(s, f: file);
2937	seq_puts(m: s, s: ", ");
2938	}
2939	spin_unlock(lock: &nf->fi_lock);
2940	} else
2941	seq_puts(m: s, s: "closed, ");
2942	nfs4_show_owner(s, oo);
2943	if (st->sc_status & SC_STATUS_ADMIN_REVOKED)
2944	seq_puts(m: s, s: ", admin-revoked");
2945	seq_puts(m: s, s: " }\n");
2946	return 0;
2947	}
2948
2949	static int nfs4_show_lock(struct seq_file *s, struct nfs4_stid *st)
2950	{
2951	struct nfs4_ol_stateid *ols;
2952	struct nfs4_file *nf;
2953	struct nfsd_file *file;
2954	struct nfs4_stateowner *oo;
2955
2956	ols = openlockstateid(s: st);
2957	oo = ols->st_stateowner;
2958	nf = st->sc_file;
2959
2960	seq_puts(m: s, s: "- ");
2961	nfs4_show_stateid(s, stid: &st->sc_stateid);
2962	seq_puts(m: s, s: ": { type: lock, ");
2963
2964	spin_lock(lock: &nf->fi_lock);
2965	file = find_any_file_locked(f: nf);
2966	if (file) {
2967	/*
2968	* Note: a lock stateid isn't really the same thing as a lock,
2969	* it's the locking state held by one owner on a file, and there
2970	* may be multiple (or no) lock ranges associated with it.
2971	* (Same for the matter is true of open stateids.)
2972	*/
2973
2974	nfs4_show_superblock(s, f: file);
2975	/* XXX: open stateid? */
2976	seq_puts(m: s, s: ", ");
2977	nfs4_show_fname(s, f: file);
2978	seq_puts(m: s, s: ", ");
2979	}
2980	nfs4_show_owner(s, oo);
2981	if (st->sc_status & SC_STATUS_ADMIN_REVOKED)
2982	seq_puts(m: s, s: ", admin-revoked");
2983	seq_puts(m: s, s: " }\n");
2984	spin_unlock(lock: &nf->fi_lock);
2985	return 0;
2986	}
2987
2988	static char *nfs4_show_deleg_type(u32 dl_type)
2989	{
2990	switch (dl_type) {
2991	case OPEN_DELEGATE_READ:
2992	return "r";
2993	case OPEN_DELEGATE_WRITE:
2994	return "w";
2995	case OPEN_DELEGATE_READ_ATTRS_DELEG:
2996	return "ra";
2997	case OPEN_DELEGATE_WRITE_ATTRS_DELEG:
2998	return "wa";
2999	}
3000	return "?";
3001	}
3002
3003	static int nfs4_show_deleg(struct seq_file *s, struct nfs4_stid *st)
3004	{
3005	struct nfs4_delegation *ds;
3006	struct nfs4_file *nf;
3007	struct nfsd_file *file;
3008
3009	ds = delegstateid(s: st);
3010	nf = st->sc_file;
3011
3012	seq_puts(m: s, s: "- ");
3013	nfs4_show_stateid(s, stid: &st->sc_stateid);
3014	seq_puts(m: s, s: ": { type: deleg, ");
3015
3016	seq_printf(m: s, fmt: "access: %s", nfs4_show_deleg_type(dl_type: ds->dl_type));
3017
3018	/* XXX: lease time, whether it's being recalled. */
3019
3020	spin_lock(lock: &nf->fi_lock);
3021	file = nf->fi_deleg_file;
3022	if (file) {
3023	seq_puts(m: s, s: ", ");
3024	nfs4_show_superblock(s, f: file);
3025	seq_puts(m: s, s: ", ");
3026	nfs4_show_fname(s, f: file);
3027	}
3028	spin_unlock(lock: &nf->fi_lock);
3029	if (st->sc_status & SC_STATUS_ADMIN_REVOKED)
3030	seq_puts(m: s, s: ", admin-revoked");
3031	seq_puts(m: s, s: " }\n");
3032	return 0;
3033	}
3034
3035	static int nfs4_show_layout(struct seq_file *s, struct nfs4_stid *st)
3036	{
3037	struct nfs4_layout_stateid *ls;
3038	struct nfsd_file *file;
3039
3040	ls = container_of(st, struct nfs4_layout_stateid, ls_stid);
3041
3042	seq_puts(m: s, s: "- ");
3043	nfs4_show_stateid(s, stid: &st->sc_stateid);
3044	seq_puts(m: s, s: ": { type: layout");
3045
3046	/* XXX: What else would be useful? */
3047
3048	spin_lock(lock: &ls->ls_stid.sc_file->fi_lock);
3049	file = ls->ls_file;
3050	if (file) {
3051	seq_puts(m: s, s: ", ");
3052	nfs4_show_superblock(s, f: file);
3053	seq_puts(m: s, s: ", ");
3054	nfs4_show_fname(s, f: file);
3055	}
3056	spin_unlock(lock: &ls->ls_stid.sc_file->fi_lock);
3057	if (st->sc_status & SC_STATUS_ADMIN_REVOKED)
3058	seq_puts(m: s, s: ", admin-revoked");
3059	seq_puts(m: s, s: " }\n");
3060
3061	return 0;
3062	}
3063
3064	static int states_show(struct seq_file *s, void *v)
3065	{
3066	struct nfs4_stid *st = v;
3067
3068	switch (st->sc_type) {
3069	case SC_TYPE_OPEN:
3070	return nfs4_show_open(s, st);
3071	case SC_TYPE_LOCK:
3072	return nfs4_show_lock(s, st);
3073	case SC_TYPE_DELEG:
3074	return nfs4_show_deleg(s, st);
3075	case SC_TYPE_LAYOUT:
3076	return nfs4_show_layout(s, st);
3077	default:
3078	return 0; /* XXX: or SEQ_SKIP? */
3079	}
3080	/* XXX: copy stateids? */
3081	}
3082
3083	static struct seq_operations states_seq_ops = {
3084	.start = states_start,
3085	.next = states_next,
3086	.stop = states_stop,
3087	.show = states_show
3088	};
3089
3090	static int client_states_open(struct inode *inode, struct file *file)
3091	{
3092	struct seq_file *s;
3093	struct nfs4_client *clp;
3094	int ret;
3095
3096	clp = get_nfsdfs_clp(inode);
3097	if (!clp)
3098	return -ENXIO;
3099
3100	ret = seq_open(file, &states_seq_ops);
3101	if (ret) {
3102	drop_client(clp);
3103	return ret;
3104	}
3105	s = file->private_data;
3106	s->private = clp;
3107	return 0;
3108	}
3109
3110	static int client_opens_release(struct inode *inode, struct file *file)
3111	{
3112	struct seq_file *m = file->private_data;
3113	struct nfs4_client *clp = m->private;
3114
3115	/* XXX: alternatively, we could get/drop in seq start/stop */
3116	drop_client(clp);
3117	return seq_release(inode, file);
3118	}
3119
3120	static const struct file_operations client_states_fops = {
3121	.open = client_states_open,
3122	.read = seq_read,
3123	.llseek = seq_lseek,
3124	.release = client_opens_release,
3125	};
3126
3127	/*
3128	* Normally we refuse to destroy clients that are in use, but here the
3129	* administrator is telling us to just do it. We also want to wait
3130	* so the caller has a guarantee that the client's locks are gone by
3131	* the time the write returns:
3132	*/
3133	static void force_expire_client(struct nfs4_client *clp)
3134	{
3135	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
3136	bool already_expired;
3137
3138	trace_nfsd_clid_admin_expired(clid: &clp->cl_clientid);
3139
3140	spin_lock(lock: &nn->client_lock);
3141	clp->cl_time = 0;
3142	spin_unlock(lock: &nn->client_lock);
3143
3144	wait_event(expiry_wq, atomic_read(&clp->cl_rpc_users) == 0);
3145	spin_lock(lock: &nn->client_lock);
3146	already_expired = list_empty(head: &clp->cl_lru);
3147	if (!already_expired)
3148	unhash_client_locked(clp);
3149	spin_unlock(lock: &nn->client_lock);
3150
3151	if (!already_expired)
3152	expire_client(clp);
3153	else
3154	wait_event(expiry_wq, clp->cl_nfsd_dentry == NULL);
3155	}
3156
3157	static ssize_t client_ctl_write(struct file *file, const char __user *buf,
3158	size_t size, loff_t *pos)
3159	{
3160	char *data;
3161	struct nfs4_client *clp;
3162
3163	data = simple_transaction_get(file, buf, size);
3164	if (IS_ERR(ptr: data))
3165	return PTR_ERR(ptr: data);
3166	if (size != 7 || 0 != memcmp(p: data, q: "expire\n", size: 7))
3167	return -EINVAL;
3168	clp = get_nfsdfs_clp(inode: file_inode(f: file));
3169	if (!clp)
3170	return -ENXIO;
3171	force_expire_client(clp);
3172	drop_client(clp);
3173	return 7;
3174	}
3175
3176	static const struct file_operations client_ctl_fops = {
3177	.write = client_ctl_write,
3178	.release = simple_transaction_release,
3179	};
3180
3181	static const struct tree_descr client_files[] = {
3182	[0] = {"info", &client_info_fops, S_IRUSR},
3183	[1] = {"states", &client_states_fops, S_IRUSR},
3184	[2] = {"ctl", &client_ctl_fops, S_IWUSR},
3185	[3] = {""},
3186	};
3187
3188	static int
3189	nfsd4_cb_recall_any_done(struct nfsd4_callback *cb,
3190	struct rpc_task *task)
3191	{
3192	trace_nfsd_cb_recall_any_done(cb, task);
3193	switch (task->tk_status) {
3194	case -NFS4ERR_DELAY:
3195	rpc_delay(task, 2 * HZ);
3196	return 0;
3197	default:
3198	return 1;
3199	}
3200	}
3201
3202	static void
3203	nfsd4_cb_recall_any_release(struct nfsd4_callback *cb)
3204	{
3205	struct nfs4_client *clp = cb->cb_clp;
3206
3207	drop_client(clp);
3208	}
3209
3210	static int
3211	nfsd4_cb_getattr_done(struct nfsd4_callback *cb, struct rpc_task *task)
3212	{
3213	struct nfs4_cb_fattr *ncf =
3214	container_of(cb, struct nfs4_cb_fattr, ncf_getattr);
3215	struct nfs4_delegation *dp =
3216	container_of(ncf, struct nfs4_delegation, dl_cb_fattr);
3217
3218	trace_nfsd_cb_getattr_done(stp: &dp->dl_stid.sc_stateid, task);
3219	ncf->ncf_cb_status = task->tk_status;
3220	switch (task->tk_status) {
3221	case -NFS4ERR_DELAY:
3222	rpc_delay(task, 2 * HZ);
3223	return 0;
3224	default:
3225	return 1;
3226	}
3227	}
3228
3229	static void
3230	nfsd4_cb_getattr_release(struct nfsd4_callback *cb)
3231	{
3232	struct nfs4_cb_fattr *ncf =
3233	container_of(cb, struct nfs4_cb_fattr, ncf_getattr);
3234	struct nfs4_delegation *dp =
3235	container_of(ncf, struct nfs4_delegation, dl_cb_fattr);
3236
3237	nfs4_put_stid(s: &dp->dl_stid);
3238	}
3239
3240	static const struct nfsd4_callback_ops nfsd4_cb_recall_any_ops = {
3241	.done = nfsd4_cb_recall_any_done,
3242	.release = nfsd4_cb_recall_any_release,
3243	.opcode = OP_CB_RECALL_ANY,
3244	};
3245
3246	static const struct nfsd4_callback_ops nfsd4_cb_getattr_ops = {
3247	.done = nfsd4_cb_getattr_done,
3248	.release = nfsd4_cb_getattr_release,
3249	.opcode = OP_CB_GETATTR,
3250	};
3251
3252	static void nfs4_cb_getattr(struct nfs4_cb_fattr *ncf)
3253	{
3254	struct nfs4_delegation *dp =
3255	container_of(ncf, struct nfs4_delegation, dl_cb_fattr);
3256
3257	if (test_and_set_bit(NFSD4_CALLBACK_RUNNING, addr: &ncf->ncf_getattr.cb_flags))
3258	return;
3259
3260	/* set to proper status when nfsd4_cb_getattr_done runs */
3261	ncf->ncf_cb_status = NFS4ERR_IO;
3262
3263	/* ensure that wake_bit is done when RUNNING is cleared */
3264	set_bit(NFSD4_CALLBACK_WAKE, addr: &ncf->ncf_getattr.cb_flags);
3265
3266	refcount_inc(r: &dp->dl_stid.sc_count);
3267	nfsd4_run_cb(cb: &ncf->ncf_getattr);
3268	}
3269
3270	static struct nfs4_client *create_client(struct xdr_netobj name,
3271	struct svc_rqst *rqstp, nfs4_verifier *verf)
3272	{
3273	struct nfs4_client *clp;
3274	struct sockaddr *sa = svc_addr(rqst: rqstp);
3275	int ret;
3276	struct net *net = SVC_NET(rqstp);
3277	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
3278	struct dentry *dentries[ARRAY_SIZE(client_files)];
3279
3280	clp = alloc_client(name, nn);
3281	if (clp == NULL)
3282	return NULL;
3283
3284	ret = copy_cred(target: &clp->cl_cred, source: &rqstp->rq_cred);
3285	if (ret) {
3286	free_client(clp);
3287	return NULL;
3288	}
3289	gen_clid(clp, nn);
3290	kref_init(kref: &clp->cl_nfsdfs.cl_ref);
3291	nfsd4_init_cb(cb: &clp->cl_cb_null, clp, NULL, op: NFSPROC4_CLNT_CB_NULL);
3292	clp->cl_time = ktime_get_boottime_seconds();
3293	copy_verf(target: clp, source: verf);
3294	memcpy(&clp->cl_addr, sa, sizeof(struct sockaddr_storage));
3295	clp->cl_cb_session = NULL;
3296	clp->net = net;
3297	clp->cl_nfsd_dentry = nfsd_client_mkdir(
3298	nn, ncl: &clp->cl_nfsdfs,
3299	id: clp->cl_clientid.cl_id - nn->clientid_base,
3300	client_files, fdentries: dentries);
3301	clp->cl_nfsd_info_dentry = dentries[0];
3302	if (!clp->cl_nfsd_dentry) {
3303	free_client(clp);
3304	return NULL;
3305	}
3306	clp->cl_ra = kzalloc(sizeof(*clp->cl_ra), GFP_KERNEL);
3307	if (!clp->cl_ra) {
3308	free_client(clp);
3309	return NULL;
3310	}
3311	clp->cl_ra_time = 0;
3312	nfsd4_init_cb(cb: &clp->cl_ra->ra_cb, clp, ops: &nfsd4_cb_recall_any_ops,
3313	op: NFSPROC4_CLNT_CB_RECALL_ANY);
3314	return clp;
3315	}
3316
3317	static void
3318	add_clp_to_name_tree(struct nfs4_client *new_clp, struct rb_root *root)
3319	{
3320	struct rb_node **new = &(root->rb_node), *parent = NULL;
3321	struct nfs4_client *clp;
3322
3323	while (*new) {
3324	clp = rb_entry(*new, struct nfs4_client, cl_namenode);
3325	parent = *new;
3326
3327	if (compare_blob(o1: &clp->cl_name, o2: &new_clp->cl_name) > 0)
3328	new = &((*new)->rb_left);
3329	else
3330	new = &((*new)->rb_right);
3331	}
3332
3333	rb_link_node(node: &new_clp->cl_namenode, parent, rb_link: new);
3334	rb_insert_color(&new_clp->cl_namenode, root);
3335	}
3336
3337	static struct nfs4_client *
3338	find_clp_in_name_tree(struct xdr_netobj *name, struct rb_root *root)
3339	{
3340	int cmp;
3341	struct rb_node *node = root->rb_node;
3342	struct nfs4_client *clp;
3343
3344	while (node) {
3345	clp = rb_entry(node, struct nfs4_client, cl_namenode);
3346	cmp = compare_blob(o1: &clp->cl_name, o2: name);
3347	if (cmp > 0)
3348	node = node->rb_left;
3349	else if (cmp < 0)
3350	node = node->rb_right;
3351	else
3352	return clp;
3353	}
3354	return NULL;
3355	}
3356
3357	static void
3358	add_to_unconfirmed(struct nfs4_client *clp)
3359	{
3360	unsigned int idhashval;
3361	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
3362
3363	lockdep_assert_held(&nn->client_lock);
3364
3365	clear_bit(NFSD4_CLIENT_CONFIRMED, addr: &clp->cl_flags);
3366	add_clp_to_name_tree(new_clp: clp, root: &nn->unconf_name_tree);
3367	idhashval = clientid_hashval(id: clp->cl_clientid.cl_id);
3368	list_add(new: &clp->cl_idhash, head: &nn->unconf_id_hashtbl[idhashval]);
3369	renew_client_locked(clp);
3370	}
3371
3372	static void
3373	move_to_confirmed(struct nfs4_client *clp)
3374	{
3375	unsigned int idhashval = clientid_hashval(id: clp->cl_clientid.cl_id);
3376	struct nfsd_net *nn = net_generic(net: clp->net, id: nfsd_net_id);
3377
3378	lockdep_assert_held(&nn->client_lock);
3379
3380	list_move(list: &clp->cl_idhash, head: &nn->conf_id_hashtbl[idhashval]);
3381	rb_erase(&clp->cl_namenode, &nn->unconf_name_tree);
3382	add_clp_to_name_tree(new_clp: clp, root: &nn->conf_name_tree);
3383	set_bit(NFSD4_CLIENT_CONFIRMED, addr: &clp->cl_flags);
3384	trace_nfsd_clid_confirmed(clid: &clp->cl_clientid);
3385	renew_client_locked(clp);
3386	}
3387
3388	static struct nfs4_client *
3389	find_client_in_id_table(struct list_head *tbl, clientid_t *clid, bool sessions)
3390	{
3391	struct nfs4_client *clp;
3392	unsigned int idhashval = clientid_hashval(id: clid->cl_id);
3393
3394	list_for_each_entry(clp, &tbl[idhashval], cl_idhash) {
3395	if (same_clid(cl1: &clp->cl_clientid, cl2: clid)) {
3396	if ((bool)clp->cl_minorversion != sessions)
3397	return NULL;
3398	renew_client_locked(clp);
3399	return clp;
3400	}
3401	}
3402	return NULL;
3403	}
3404
3405	static struct nfs4_client *
3406	find_confirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
3407	{
3408	struct list_head *tbl = nn->conf_id_hashtbl;
3409
3410	lockdep_assert_held(&nn->client_lock);
3411	return find_client_in_id_table(tbl, clid, sessions);
3412	}
3413
3414	static struct nfs4_client *
3415	find_unconfirmed_client(clientid_t *clid, bool sessions, struct nfsd_net *nn)
3416	{
3417	struct list_head *tbl = nn->unconf_id_hashtbl;
3418
3419	lockdep_assert_held(&nn->client_lock);
3420	return find_client_in_id_table(tbl, clid, sessions);
3421	}
3422
3423	static bool clp_used_exchangeid(struct nfs4_client *clp)
3424	{
3425	return clp->cl_exchange_flags != 0;
3426	}
3427
3428	static struct nfs4_client *
3429	find_confirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
3430	{
3431	lockdep_assert_held(&nn->client_lock);
3432	return find_clp_in_name_tree(name, root: &nn->conf_name_tree);
3433	}
3434
3435	static struct nfs4_client *
3436	find_unconfirmed_client_by_name(struct xdr_netobj *name, struct nfsd_net *nn)
3437	{
3438	lockdep_assert_held(&nn->client_lock);
3439	return find_clp_in_name_tree(name, root: &nn->unconf_name_tree);
3440	}
3441
3442	static void
3443	gen_callback(struct nfs4_client *clp, struct nfsd4_setclientid *se, struct svc_rqst *rqstp)
3444	{
3445	struct nfs4_cb_conn *conn = &clp->cl_cb_conn;
3446	struct sockaddr *sa = svc_addr(rqst: rqstp);
3447	u32 scopeid = rpc_get_scope_id(sa);
3448	unsigned short expected_family;
3449
3450	/* Currently, we only support tcp and tcp6 for the callback channel */
3451	if (se->se_callback_netid_len == 3 &&
3452	!memcmp(p: se->se_callback_netid_val, q: "tcp", size: 3))
3453	expected_family = AF_INET;
3454	else if (se->se_callback_netid_len == 4 &&
3455	!memcmp(p: se->se_callback_netid_val, q: "tcp6", size: 4))
3456	expected_family = AF_INET6;
3457	else
3458	goto out_err;
3459
3460	conn->cb_addrlen = rpc_uaddr2sockaddr(clp->net, se->se_callback_addr_val,
3461	se->se_callback_addr_len,
3462	(struct sockaddr *)&conn->cb_addr,
3463	sizeof(conn->cb_addr));
3464
3465	if (!conn->cb_addrlen || conn->cb_addr.ss_family != expected_family)
3466	goto out_err;
3467
3468	if (conn->cb_addr.ss_family == AF_INET6)
3469	((struct sockaddr_in6 *)&conn->cb_addr)->sin6_scope_id = scopeid;
3470
3471	conn->cb_prog = se->se_callback_prog;
3472	conn->cb_ident = se->se_callback_ident;
3473	memcpy(&conn->cb_saddr, &rqstp->rq_daddr, rqstp->rq_daddrlen);
3474	trace_nfsd_cb_args(clp, conn);
3475	return;
3476	out_err:
3477	conn->cb_addr.ss_family = AF_UNSPEC;
3478	conn->cb_addrlen = 0;
3479	trace_nfsd_cb_nodelegs(clp);
3480	return;
3481	}
3482
3483	/*
3484	* Cache a reply. nfsd4_check_resp_size() has bounded the cache size.
3485	*/
3486	static void
3487	nfsd4_store_cache_entry(struct nfsd4_compoundres *resp)
3488	{
3489	struct xdr_buf *buf = resp->xdr->buf;
3490	struct nfsd4_slot *slot = resp->cstate.slot;
3491	unsigned int base;
3492
3493	/*
3494	* RFC 5661 Section 2.10.6.1.2:
3495	*
3496	* Any time SEQUENCE ... returns an error ... [t]he replier MUST NOT
3497	* modify the reply cache entry for the slot whenever an error is
3498	* returned from SEQUENCE ...
3499	*
3500	* Because nfsd4_store_cache_entry is called only by
3501	* nfsd4_sequence_done(), nfsd4_store_cache_entry() is called only
3502	* when a SEQUENCE operation was part of the COMPOUND.
3503	* nfs41_check_op_ordering() ensures SEQUENCE is the first op.
3504	*/
3505	if (resp->opcnt == 1 && resp->cstate.status != nfs_ok)
3506	return;
3507
3508	slot->sl_flags |= NFSD4_SLOT_INITIALIZED;
3509	slot->sl_opcnt = resp->opcnt;
3510	slot->sl_status = resp->cstate.status;
3511	free_svc_cred(cred: &slot->sl_cred);
3512	copy_cred(target: &slot->sl_cred, source: &resp->rqstp->rq_cred);
3513
3514	if (!(resp->cstate.slot->sl_flags & NFSD4_SLOT_CACHETHIS)) {
3515	slot->sl_flags &= ~NFSD4_SLOT_CACHED;
3516	return;
3517	}
3518	slot->sl_flags |= NFSD4_SLOT_CACHED;
3519
3520	base = resp->cstate.data_offset;
3521	slot->sl_datalen = buf->len - base;
3522	if (read_bytes_from_xdr_buf(buf, base, slot->sl_data, slot->sl_datalen))
3523	WARN(1, "%s: sessions DRC could not cache compound\n",
3524	__func__);
3525	return;
3526	}
3527
3528	/*
3529	* The sequence operation is not cached because we can use the slot and
3530	* session values.
3531	*/
3532	static __be32
3533	nfsd4_replay_cache_entry(struct nfsd4_compoundres *resp,
3534	struct nfsd4_sequence *seq)
3535	{
3536	struct nfsd4_compoundargs *args = resp->rqstp->rq_argp;
3537	struct nfsd4_slot *slot = resp->cstate.slot;
3538	struct xdr_stream *xdr = resp->xdr;
3539	__be32 *p;
3540
3541	dprintk("--> %s slot %p\n", __func__, slot);
3542
3543	/* Always encode the SEQUENCE response. */
3544	nfsd4_encode_operation(resp, &args->ops[0]);
3545	if (args->opcnt == 1)
3546	/* A solo SEQUENCE - nothing was cached */
3547	return args->ops[0].status;
3548
3549	if (!(slot->sl_flags & NFSD4_SLOT_CACHED)) {
3550	/* We weren't asked to cache this. */
3551	struct nfsd4_op *op;
3552
3553	op = &args->ops[resp->opcnt++];
3554	op->status = nfserr_retry_uncached_rep;
3555	nfsd4_encode_operation(resp, op);
3556	return op->status;
3557	}
3558
3559	/* return reply from cache */
3560	p = xdr_reserve_space(xdr, nbytes: slot->sl_datalen);
3561	if (!p) {
3562	WARN_ON_ONCE(1);
3563	return nfserr_serverfault;
3564	}
3565	xdr_encode_opaque_fixed(p, ptr: slot->sl_data, len: slot->sl_datalen);
3566	xdr_commit_encode(xdr);
3567
3568	resp->opcnt = slot->sl_opcnt;
3569	return slot->sl_status;
3570	}
3571
3572	/*
3573	* Set the exchange_id flags returned by the server.
3574	*/
3575	static void
3576	nfsd4_set_ex_flags(struct nfs4_client *new, struct nfsd4_exchange_id *clid)
3577	{
3578	#ifdef CONFIG_NFSD_PNFS
3579	new->cl_exchange_flags |= EXCHGID4_FLAG_USE_PNFS_MDS;
3580	#else
3581	new->cl_exchange_flags |= EXCHGID4_FLAG_USE_NON_PNFS;
3582	#endif
3583
3584	/* Referrals are supported, Migration is not. */
3585	new->cl_exchange_flags |= EXCHGID4_FLAG_SUPP_MOVED_REFER;
3586
3587	/* set the wire flags to return to client. */
3588	clid->flags = new->cl_exchange_flags;
3589	}
3590
3591	static bool client_has_openowners(struct nfs4_client *clp)
3592	{
3593	struct nfs4_openowner *oo;
3594
3595	list_for_each_entry(oo, &clp->cl_openowners, oo_perclient) {
3596	if (!list_empty(head: &oo->oo_owner.so_stateids))
3597	return true;
3598	}
3599	return false;
3600	}
3601
3602	static bool client_has_state(struct nfs4_client *clp)
3603	{
3604	return client_has_openowners(clp)
3605	#ifdef CONFIG_NFSD_PNFS
3606	|| !list_empty(head: &clp->cl_lo_states)
3607	#endif
3608	|| !list_empty(head: &clp->cl_delegations)
3609	|| !list_empty(head: &clp->cl_sessions)
3610	|| nfsd4_has_active_async_copies(clp);
3611	}
3612
3613	static __be32 copy_impl_id(struct nfs4_client *clp,
3614	struct nfsd4_exchange_id *exid)
3615	{
3616	if (!exid->nii_domain.data)
3617	return 0;
3618	xdr_netobj_dup(dst: &clp->cl_nii_domain, src: &exid->nii_domain, GFP_KERNEL);
3619	if (!clp->cl_nii_domain.data)
3620	return nfserr_jukebox;
3621	xdr_netobj_dup(dst: &clp->cl_nii_name, src: &exid->nii_name, GFP_KERNEL);
3622	if (!clp->cl_nii_name.data)
3623	return nfserr_jukebox;
3624	clp->cl_nii_time = exid->nii_time;
3625	return 0;
3626	}
3627
3628	__be32
3629	nfsd4_exchange_id(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
3630	union nfsd4_op_u *u)
3631	{
3632	struct nfsd4_exchange_id *exid = &u->exchange_id;
3633	struct nfs4_client *conf, *new;
3634	struct nfs4_client *unconf = NULL;
3635	__be32 status;
3636	char addr_str[INET6_ADDRSTRLEN];
3637	nfs4_verifier verf = exid->verifier;
3638	struct sockaddr *sa = svc_addr(rqst: rqstp);
3639	bool update = exid->flags & EXCHGID4_FLAG_UPD_CONFIRMED_REC_A;
3640	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
3641
3642	rpc_ntop(sa, addr_str, sizeof(addr_str));
3643	dprintk("%s rqstp=%p exid=%p clname.len=%u clname.data=%p "
3644	"ip_addr=%s flags %x, spa_how %u\n",
3645	__func__, rqstp, exid, exid->clname.len, exid->clname.data,
3646	addr_str, exid->flags, exid->spa_how);
3647
3648	exid->server_impl_name = kasprintf(GFP_KERNEL, fmt: "%s %s %s %s",
3649	utsname()->sysname, utsname()->release,
3650	utsname()->version, utsname()->machine);
3651	if (!exid->server_impl_name)
3652	return nfserr_jukebox;
3653
3654	if (exid->flags & ~EXCHGID4_FLAG_MASK_A)
3655	return nfserr_inval;
3656
3657	new = create_client(name: exid->clname, rqstp, verf: &verf);
3658	if (new == NULL)
3659	return nfserr_jukebox;
3660	status = copy_impl_id(clp: new, exid);
3661	if (status)
3662	goto out_nolock;
3663
3664	switch (exid->spa_how) {
3665	case SP4_MACH_CRED:
3666	exid->spo_must_enforce[0] = 0;
3667	exid->spo_must_enforce[1] = (
3668	1 << (OP_BIND_CONN_TO_SESSION - 32) |
3669	1 << (OP_EXCHANGE_ID - 32) |
3670	1 << (OP_CREATE_SESSION - 32) |
3671	1 << (OP_DESTROY_SESSION - 32) |
3672	1 << (OP_DESTROY_CLIENTID - 32));
3673
3674	exid->spo_must_allow[0] &= (1 << (OP_CLOSE) |
3675	1 << (OP_OPEN_DOWNGRADE) |
3676	1 << (OP_LOCKU) |
3677	1 << (OP_DELEGRETURN));
3678
3679	exid->spo_must_allow[1] &= (
3680	1 << (OP_TEST_STATEID - 32) |
3681	1 << (OP_FREE_STATEID - 32));
3682	if (!svc_rqst_integrity_protected(rqstp)) {
3683	status = nfserr_inval;
3684	goto out_nolock;
3685	}
3686	/*
3687	* Sometimes userspace doesn't give us a principal.
3688	* Which is a bug, really. Anyway, we can't enforce
3689	* MACH_CRED in that case, better to give up now:
3690	*/
3691	if (!new->cl_cred.cr_principal &&
3692	!new->cl_cred.cr_raw_principal) {
3693	status = nfserr_serverfault;
3694	goto out_nolock;
3695	}
3696	new->cl_mach_cred = true;
3697	break;
3698	case SP4_NONE:
3699	break;
3700	default: /* checked by xdr code */
3701	WARN_ON_ONCE(1);
3702	fallthrough;
3703	case SP4_SSV:
3704	status = nfserr_encr_alg_unsupp;
3705	goto out_nolock;
3706	}
3707
3708	/* Cases below refer to rfc 5661 section 18.35.4: */
3709	spin_lock(lock: &nn->client_lock);
3710	conf = find_confirmed_client_by_name(name: &exid->clname, nn);
3711	if (conf) {
3712	bool creds_match = same_creds(cr1: &conf->cl_cred, cr2: &rqstp->rq_cred);
3713	bool verfs_match = same_verf(v1: &verf, v2: &conf->cl_verifier);
3714
3715	if (update) {
3716	if (!clp_used_exchangeid(clp: conf)) { /* buggy client */
3717	status = nfserr_inval;
3718	goto out;
3719	}
3720	if (!nfsd4_mach_creds_match(cl: conf, rqstp)) {
3721	status = nfserr_wrong_cred;
3722	goto out;
3723	}
3724	if (!creds_match) { /* case 9 */
3725	status = nfserr_perm;
3726	goto out;
3727	}
3728	if (!verfs_match) { /* case 8 */
3729	status = nfserr_not_same;
3730	goto out;
3731	}
3732	/* case 6 */
3733	exid->flags |= EXCHGID4_FLAG_CONFIRMED_R;
3734	trace_nfsd_clid_confirmed_r(clp: conf);
3735	goto out_copy;
3736	}
3737	if (!creds_match) { /* case 3 */
3738	if (client_has_state(clp: conf)) {
3739	status = nfserr_clid_inuse;
3740	trace_nfsd_clid_cred_mismatch(clp: conf, rqstp);
3741	goto out;
3742	}
3743	goto out_new;
3744	}
3745	if (verfs_match) { /* case 2 */
3746	conf->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
3747	trace_nfsd_clid_confirmed_r(clp: conf);
3748	goto out_copy;
3749	}
3750	/* case 5, client reboot */
3751	trace_nfsd_clid_verf_mismatch(clp: conf, rqstp, verf: &verf);
3752	conf = NULL;
3753	goto out_new;
3754	}
3755
3756	if (update) { /* case 7 */
3757	status = nfserr_noent;
3758	goto out;
3759	}
3760
3761	unconf = find_unconfirmed_client_by_name(name: &exid->clname, nn);
3762	if (unconf) /* case 4, possible retry or client restart */
3763	unhash_client_locked(clp: unconf);
3764
3765	/* case 1, new owner ID */
3766	trace_nfsd_clid_fresh(clp: new);
3767
3768	out_new:
3769	if (conf) {
3770	status = mark_client_expired_locked(clp: conf);
3771	if (status)
3772	goto out;
3773	trace_nfsd_clid_replaced(clid: &conf->cl_clientid);
3774	}
3775	new->cl_minorversion = cstate->minorversion;
3776	new->cl_spo_must_allow.u.words[0] = exid->spo_must_allow[0];
3777	new->cl_spo_must_allow.u.words[1] = exid->spo_must_allow[1];
3778
3779	/* Contrived initial CREATE_SESSION response */
3780	new->cl_cs_slot.sl_status = nfserr_seq_misordered;
3781
3782	add_to_unconfirmed(clp: new);
3783	swap(new, conf);
3784	out_copy:
3785	exid->clientid.cl_boot = conf->cl_clientid.cl_boot;
3786	exid->clientid.cl_id = conf->cl_clientid.cl_id;
3787
3788	exid->seqid = conf->cl_cs_slot.sl_seqid + 1;
3789	nfsd4_set_ex_flags(new: conf, clid: exid);
3790
3791	exid->nii_domain.len = sizeof("kernel.org") - 1;
3792	exid->nii_domain.data = "kernel.org";
3793
3794	/*
3795	* Note that RFC 8881 places no length limit on
3796	* nii_name, but this implementation permits no
3797	* more than NFS4_OPAQUE_LIMIT bytes.
3798	*/
3799	exid->nii_name.len = strlen(exid->server_impl_name);
3800	if (exid->nii_name.len > NFS4_OPAQUE_LIMIT)
3801	exid->nii_name.len = NFS4_OPAQUE_LIMIT;
3802	exid->nii_name.data = exid->server_impl_name;
3803
3804	/* just send zeros - the date is in nii_name */
3805	exid->nii_time.tv_sec = 0;
3806	exid->nii_time.tv_nsec = 0;
3807
3808	dprintk("nfsd4_exchange_id seqid %d flags %x\n",
3809	conf->cl_cs_slot.sl_seqid, conf->cl_exchange_flags);
3810	status = nfs_ok;
3811
3812	out:
3813	spin_unlock(lock: &nn->client_lock);
3814	out_nolock:
3815	if (new)
3816	expire_client(clp: new);
3817	if (unconf) {
3818	trace_nfsd_clid_expire_unconf(clid: &unconf->cl_clientid);
3819	expire_client(clp: unconf);
3820	}
3821	return status;
3822	}
3823
3824	void
3825	nfsd4_exchange_id_release(union nfsd4_op_u *u)
3826	{
3827	struct nfsd4_exchange_id *exid = &u->exchange_id;
3828
3829	kfree(objp: exid->server_impl_name);
3830	}
3831
3832	static __be32 check_slot_seqid(u32 seqid, u32 slot_seqid, u8 flags)
3833	{
3834	/* The slot is in use, and no response has been sent. */
3835	if (flags & NFSD4_SLOT_INUSE) {
3836	if (seqid == slot_seqid)
3837	return nfserr_jukebox;
3838	else
3839	return nfserr_seq_misordered;
3840	}
3841	/* Note unsigned 32-bit arithmetic handles wraparound: */
3842	if (likely(seqid == slot_seqid + 1))
3843	return nfs_ok;
3844	if ((flags & NFSD4_SLOT_REUSED) && seqid == 1)
3845	return nfs_ok;
3846	if (seqid == slot_seqid)
3847	return nfserr_replay_cache;
3848	return nfserr_seq_misordered;
3849	}
3850
3851	/*
3852	* Cache the create session result into the create session single DRC
3853	* slot cache by saving the xdr structure. sl_seqid has been set.
3854	* Do this for solo or embedded create session operations.
3855	*/
3856	static void
3857	nfsd4_cache_create_session(struct nfsd4_create_session *cr_ses,
3858	struct nfsd4_clid_slot *slot, __be32 nfserr)
3859	{
3860	slot->sl_status = nfserr;
3861	memcpy(&slot->sl_cr_ses, cr_ses, sizeof(*cr_ses));
3862	}
3863
3864	static __be32
3865	nfsd4_replay_create_session(struct nfsd4_create_session *cr_ses,
3866	struct nfsd4_clid_slot *slot)
3867	{
3868	memcpy(cr_ses, &slot->sl_cr_ses, sizeof(*cr_ses));
3869	return slot->sl_status;
3870	}
3871
3872	#define NFSD_MIN_REQ_HDR_SEQ_SZ ((\
3873	2 * 2 + /* credential,verifier: AUTH_NULL, length 0 */ \
3874	1 + /* MIN tag is length with zero, only length */ \
3875	3 + /* version, opcount, opcode */ \
3876	XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
3877	/* seqid, slotID, slotID, cache */ \
3878	4 ) * sizeof(__be32))
3879
3880	#define NFSD_MIN_RESP_HDR_SEQ_SZ ((\
3881	2 + /* verifier: AUTH_NULL, length 0 */\
3882	1 + /* status */ \
3883	1 + /* MIN tag is length with zero, only length */ \
3884	3 + /* opcount, opcode, opstatus*/ \
3885	XDR_QUADLEN(NFS4_MAX_SESSIONID_LEN) + \
3886	/* seqid, slotID, slotID, slotID, status */ \
3887	5 ) * sizeof(__be32))
3888
3889	static __be32 check_forechannel_attrs(struct nfsd4_channel_attrs *ca, struct nfsd_net *nn)
3890	{
3891	u32 maxrpc = nn->nfsd_serv->sv_max_mesg;
3892
3893	if (ca->maxreq_sz < NFSD_MIN_REQ_HDR_SEQ_SZ)
3894	return nfserr_toosmall;
3895	if (ca->maxresp_sz < NFSD_MIN_RESP_HDR_SEQ_SZ)
3896	return nfserr_toosmall;
3897	ca->headerpadsz = 0;
3898	ca->maxreq_sz = min_t(u32, ca->maxreq_sz, maxrpc);
3899	ca->maxresp_sz = min_t(u32, ca->maxresp_sz, maxrpc);
3900	ca->maxops = min_t(u32, ca->maxops, NFSD_MAX_OPS_PER_COMPOUND);
3901	ca->maxresp_cached = min_t(u32, ca->maxresp_cached,
3902	NFSD_SLOT_CACHE_SIZE + NFSD_MIN_HDR_SEQ_SZ);
3903	ca->maxreqs = min_t(u32, ca->maxreqs, NFSD_MAX_SLOTS_PER_SESSION);
3904
3905	return nfs_ok;
3906	}
3907
3908	/*
3909	* Server's NFSv4.1 backchannel support is AUTH_SYS-only for now.
3910	* These are based on similar macros in linux/sunrpc/msg_prot.h .
3911	*/
3912	#define RPC_MAX_HEADER_WITH_AUTH_SYS \
3913	(RPC_CALLHDRSIZE + 2 * (2 + UNX_CALLSLACK))
3914
3915	#define RPC_MAX_REPHEADER_WITH_AUTH_SYS \
3916	(RPC_REPHDRSIZE + (2 + NUL_REPLYSLACK))
3917
3918	#define NFSD_CB_MAX_REQ_SZ ((NFS4_enc_cb_recall_sz + \
3919	RPC_MAX_HEADER_WITH_AUTH_SYS) * sizeof(__be32))
3920	#define NFSD_CB_MAX_RESP_SZ ((NFS4_dec_cb_recall_sz + \
3921	RPC_MAX_REPHEADER_WITH_AUTH_SYS) * \
3922	sizeof(__be32))
3923
3924	static __be32 check_backchannel_attrs(struct nfsd4_channel_attrs *ca)
3925	{
3926	ca->headerpadsz = 0;
3927
3928	if (ca->maxreq_sz < NFSD_CB_MAX_REQ_SZ)
3929	return nfserr_toosmall;
3930	if (ca->maxresp_sz < NFSD_CB_MAX_RESP_SZ)
3931	return nfserr_toosmall;
3932	ca->maxresp_cached = 0;
3933	if (ca->maxops < 2)
3934	return nfserr_toosmall;
3935
3936	return nfs_ok;
3937	}
3938
3939	static __be32 nfsd4_check_cb_sec(struct nfsd4_cb_sec *cbs)
3940	{
3941	switch (cbs->flavor) {
3942	case RPC_AUTH_NULL:
3943	case RPC_AUTH_UNIX:
3944	return nfs_ok;
3945	default:
3946	/*
3947	* GSS case: the spec doesn't allow us to return this
3948	* error. But it also doesn't allow us not to support
3949	* GSS.
3950	* I'd rather this fail hard than return some error the
3951	* client might think it can already handle:
3952	*/
3953	return nfserr_encr_alg_unsupp;
3954	}
3955	}
3956
3957	__be32
3958	nfsd4_create_session(struct svc_rqst *rqstp,
3959	struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
3960	{
3961	struct nfsd4_create_session *cr_ses = &u->create_session;
3962	struct sockaddr *sa = svc_addr(rqst: rqstp);
3963	struct nfs4_client *conf, *unconf;
3964	struct nfsd4_clid_slot *cs_slot;
3965	struct nfs4_client *old = NULL;
3966	struct nfsd4_session *new;
3967	struct nfsd4_conn *conn;
3968	__be32 status = 0;
3969	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
3970
3971	if (cr_ses->flags & ~SESSION4_FLAG_MASK_A)
3972	return nfserr_inval;
3973	status = nfsd4_check_cb_sec(cbs: &cr_ses->cb_sec);
3974	if (status)
3975	return status;
3976	status = check_forechannel_attrs(ca: &cr_ses->fore_channel, nn);
3977	if (status)
3978	return status;
3979	status = check_backchannel_attrs(ca: &cr_ses->back_channel);
3980	if (status)
3981	goto out_err;
3982	status = nfserr_jukebox;
3983	new = alloc_session(fattrs: &cr_ses->fore_channel, battrs: &cr_ses->back_channel);
3984	if (!new)
3985	goto out_err;
3986	conn = alloc_conn_from_crses(rqstp, cses: cr_ses);
3987	if (!conn)
3988	goto out_free_session;
3989
3990	spin_lock(lock: &nn->client_lock);
3991
3992	/* RFC 8881 Section 18.36.4 Phase 1: Client record look-up. */
3993	unconf = find_unconfirmed_client(clid: &cr_ses->clientid, sessions: true, nn);
3994	conf = find_confirmed_client(clid: &cr_ses->clientid, sessions: true, nn);
3995	if (!conf && !unconf) {
3996	status = nfserr_stale_clientid;
3997	goto out_free_conn;
3998	}
3999
4000	/* RFC 8881 Section 18.36.4 Phase 2: Sequence ID processing. */
4001	if (conf) {
4002	cs_slot = &conf->cl_cs_slot;
4003	trace_nfsd_slot_seqid_conf(clp: conf, cs: cr_ses);
4004	} else {
4005	cs_slot = &unconf->cl_cs_slot;
4006	trace_nfsd_slot_seqid_unconf(clp: unconf, cs: cr_ses);
4007	}
4008	status = check_slot_seqid(seqid: cr_ses->seqid, slot_seqid: cs_slot->sl_seqid, flags: 0);
4009	switch (status) {
4010	case nfs_ok:
4011	cs_slot->sl_seqid++;
4012	cr_ses->seqid = cs_slot->sl_seqid;
4013	break;
4014	case nfserr_replay_cache:
4015	status = nfsd4_replay_create_session(cr_ses, slot: cs_slot);
4016	fallthrough;
4017	case nfserr_jukebox:
4018	/* The server MUST NOT cache NFS4ERR_DELAY */
4019	goto out_free_conn;
4020	default:
4021	goto out_cache_error;
4022	}
4023
4024	/* RFC 8881 Section 18.36.4 Phase 3: Client ID confirmation. */
4025	if (conf) {
4026	status = nfserr_wrong_cred;
4027	if (!nfsd4_mach_creds_match(cl: conf, rqstp))
4028	goto out_cache_error;
4029	} else {
4030	status = nfserr_clid_inuse;
4031	if (!same_creds(cr1: &unconf->cl_cred, cr2: &rqstp->rq_cred) ||
4032	!rpc_cmp_addr(sap1: sa, sap2: (struct sockaddr *) &unconf->cl_addr)) {
4033	trace_nfsd_clid_cred_mismatch(clp: unconf, rqstp);
4034	goto out_cache_error;
4035	}
4036	status = nfserr_wrong_cred;
4037	if (!nfsd4_mach_creds_match(cl: unconf, rqstp))
4038	goto out_cache_error;
4039	old = find_confirmed_client_by_name(name: &unconf->cl_name, nn);
4040	if (old) {
4041	status = mark_client_expired_locked(clp: old);
4042	if (status)
4043	goto out_expired_error;
4044	trace_nfsd_clid_replaced(clid: &old->cl_clientid);
4045	}
4046	move_to_confirmed(clp: unconf);
4047	conf = unconf;
4048	}
4049
4050	/* RFC 8881 Section 18.36.4 Phase 4: Session creation. */
4051	status = nfs_ok;
4052	/* Persistent sessions are not supported */
4053	cr_ses->flags &= ~SESSION4_PERSIST;
4054	/* Upshifting from TCP to RDMA is not supported */
4055	cr_ses->flags &= ~SESSION4_RDMA;
4056	/* Report the correct number of backchannel slots */
4057	cr_ses->back_channel.maxreqs = new->se_cb_highest_slot + 1;
4058
4059	init_session(rqstp, new, clp: conf, cses: cr_ses);
4060	nfsd4_get_session_locked(ses: new);
4061
4062	memcpy(cr_ses->sessionid.data, new->se_sessionid.data,
4063	NFS4_MAX_SESSIONID_LEN);
4064
4065	/* cache solo and embedded create sessions under the client_lock */
4066	nfsd4_cache_create_session(cr_ses, slot: cs_slot, nfserr: status);
4067	spin_unlock(lock: &nn->client_lock);
4068	if (conf == unconf)
4069	fsnotify_dentry(dentry: conf->cl_nfsd_info_dentry, FS_MODIFY);
4070	/* init connection and backchannel */
4071	nfsd4_init_conn(rqstp, conn, ses: new);
4072	nfsd4_put_session(ses: new);
4073	if (old)
4074	expire_client(clp: old);
4075	return status;
4076
4077	out_expired_error:
4078	/*
4079	* Revert the slot seq_nr change so the server will process
4080	* the client's resend instead of returning a cached response.
4081	*/
4082	if (status == nfserr_jukebox) {
4083	cs_slot->sl_seqid--;
4084	cr_ses->seqid = cs_slot->sl_seqid;
4085	goto out_free_conn;
4086	}
4087	out_cache_error:
4088	nfsd4_cache_create_session(cr_ses, slot: cs_slot, nfserr: status);
4089	out_free_conn:
4090	spin_unlock(lock: &nn->client_lock);
4091	free_conn(c: conn);
4092	out_free_session:
4093	__free_session(ses: new);
4094	out_err:
4095	return status;
4096	}
4097
4098	static __be32 nfsd4_map_bcts_dir(u32 *dir)
4099	{
4100	switch (*dir) {
4101	case NFS4_CDFC4_FORE:
4102	case NFS4_CDFC4_BACK:
4103	return nfs_ok;
4104	case NFS4_CDFC4_FORE_OR_BOTH:
4105	case NFS4_CDFC4_BACK_OR_BOTH:
4106	*dir = NFS4_CDFC4_BOTH;
4107	return nfs_ok;
4108	}
4109	return nfserr_inval;
4110	}
4111
4112	__be32 nfsd4_backchannel_ctl(struct svc_rqst *rqstp,
4113	struct nfsd4_compound_state *cstate,
4114	union nfsd4_op_u *u)
4115	{
4116	struct nfsd4_backchannel_ctl *bc = &u->backchannel_ctl;
4117	struct nfsd4_session *session = cstate->session;
4118	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
4119	__be32 status;
4120
4121	status = nfsd4_check_cb_sec(cbs: &bc->bc_cb_sec);
4122	if (status)
4123	return status;
4124	spin_lock(lock: &nn->client_lock);
4125	session->se_cb_prog = bc->bc_cb_program;
4126	session->se_cb_sec = bc->bc_cb_sec;
4127	spin_unlock(lock: &nn->client_lock);
4128
4129	nfsd4_probe_callback(clp: session->se_client);
4130
4131	return nfs_ok;
4132	}
4133
4134	static struct nfsd4_conn *__nfsd4_find_conn(struct svc_xprt *xpt, struct nfsd4_session *s)
4135	{
4136	struct nfsd4_conn *c;
4137
4138	list_for_each_entry(c, &s->se_conns, cn_persession) {
4139	if (c->cn_xprt == xpt) {
4140	return c;
4141	}
4142	}
4143	return NULL;
4144	}
4145
4146	static __be32 nfsd4_match_existing_connection(struct svc_rqst *rqst,
4147	struct nfsd4_session *session, u32 req, struct nfsd4_conn **conn)
4148	{
4149	struct nfs4_client *clp = session->se_client;
4150	struct svc_xprt *xpt = rqst->rq_xprt;
4151	struct nfsd4_conn *c;
4152	__be32 status;
4153
4154	/* Following the last paragraph of RFC 5661 Section 18.34.3: */
4155	spin_lock(lock: &clp->cl_lock);
4156	c = __nfsd4_find_conn(xpt, s: session);
4157	if (!c)
4158	status = nfserr_noent;
4159	else if (req == c->cn_flags)
4160	status = nfs_ok;
4161	else if (req == NFS4_CDFC4_FORE_OR_BOTH &&
4162	c->cn_flags != NFS4_CDFC4_BACK)
4163	status = nfs_ok;
4164	else if (req == NFS4_CDFC4_BACK_OR_BOTH &&
4165	c->cn_flags != NFS4_CDFC4_FORE)
4166	status = nfs_ok;
4167	else
4168	status = nfserr_inval;
4169	spin_unlock(lock: &clp->cl_lock);
4170	if (status == nfs_ok && conn)
4171	*conn = c;
4172	return status;
4173	}
4174
4175	__be32 nfsd4_bind_conn_to_session(struct svc_rqst *rqstp,
4176	struct nfsd4_compound_state *cstate,
4177	union nfsd4_op_u *u)
4178	{
4179	struct nfsd4_bind_conn_to_session *bcts = &u->bind_conn_to_session;
4180	__be32 status;
4181	struct nfsd4_conn *conn;
4182	struct nfsd4_session *session;
4183	struct net *net = SVC_NET(rqstp);
4184	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
4185
4186	if (!nfsd4_last_compound_op(rqstp))
4187	return nfserr_not_only_op;
4188	spin_lock(lock: &nn->client_lock);
4189	session = find_in_sessionid_hashtbl(sessionid: &bcts->sessionid, net, ret: &status);
4190	spin_unlock(lock: &nn->client_lock);
4191	if (!session)
4192	goto out_no_session;
4193	status = nfserr_wrong_cred;
4194	if (!nfsd4_mach_creds_match(cl: session->se_client, rqstp))
4195	goto out;
4196	status = nfsd4_match_existing_connection(rqst: rqstp, session,
4197	req: bcts->dir, conn: &conn);
4198	if (status == nfs_ok) {
4199	if (bcts->dir == NFS4_CDFC4_FORE_OR_BOTH ||
4200	bcts->dir == NFS4_CDFC4_BACK)
4201	conn->cn_flags |= NFS4_CDFC4_BACK;
4202	nfsd4_probe_callback(clp: session->se_client);
4203	goto out;
4204	}
4205	if (status == nfserr_inval)
4206	goto out;
4207	status = nfsd4_map_bcts_dir(dir: &bcts->dir);
4208	if (status)
4209	goto out;
4210	conn = alloc_conn(rqstp, flags: bcts->dir);
4211	status = nfserr_jukebox;
4212	if (!conn)
4213	goto out;
4214	nfsd4_init_conn(rqstp, conn, ses: session);
4215	status = nfs_ok;
4216	out:
4217	nfsd4_put_session(ses: session);
4218	out_no_session:
4219	return status;
4220	}
4221
4222	static bool nfsd4_compound_in_session(struct nfsd4_compound_state *cstate, struct nfs4_sessionid *sid)
4223	{
4224	if (!cstate->session)
4225	return false;
4226	return !memcmp(p: sid, q: &cstate->session->se_sessionid, size: sizeof(*sid));
4227	}
4228
4229	__be32
4230	nfsd4_destroy_session(struct svc_rqst *r, struct nfsd4_compound_state *cstate,
4231	union nfsd4_op_u *u)
4232	{
4233	struct nfs4_sessionid *sessionid = &u->destroy_session.sessionid;
4234	struct nfsd4_session *ses;
4235	__be32 status;
4236	int ref_held_by_me = 0;
4237	struct net *net = SVC_NET(r);
4238	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
4239
4240	status = nfserr_not_only_op;
4241	if (nfsd4_compound_in_session(cstate, sid: sessionid)) {
4242	if (!nfsd4_last_compound_op(rqstp: r))
4243	goto out;
4244	ref_held_by_me++;
4245	}
4246	dump_sessionid(fn: __func__, sessionid);
4247	spin_lock(lock: &nn->client_lock);
4248	ses = find_in_sessionid_hashtbl(sessionid, net, ret: &status);
4249	if (!ses)
4250	goto out_client_lock;
4251	status = nfserr_wrong_cred;
4252	if (!nfsd4_mach_creds_match(cl: ses->se_client, rqstp: r))
4253	goto out_put_session;
4254	status = mark_session_dead_locked(ses, ref_held_by_me: 1 + ref_held_by_me);
4255	if (status)
4256	goto out_put_session;
4257	unhash_session(ses);
4258	spin_unlock(lock: &nn->client_lock);
4259
4260	nfsd4_probe_callback_sync(clp: ses->se_client);
4261
4262	spin_lock(lock: &nn->client_lock);
4263	status = nfs_ok;
4264	out_put_session:
4265	nfsd4_put_session_locked(ses);
4266	out_client_lock:
4267	spin_unlock(lock: &nn->client_lock);
4268	out:
4269	return status;
4270	}
4271
4272	static __be32 nfsd4_sequence_check_conn(struct nfsd4_conn *new, struct nfsd4_session *ses)
4273	{
4274	struct nfs4_client *clp = ses->se_client;
4275	struct nfsd4_conn *c;
4276	__be32 status = nfs_ok;
4277	int ret;
4278
4279	spin_lock(lock: &clp->cl_lock);
4280	c = __nfsd4_find_conn(xpt: new->cn_xprt, s: ses);
4281	if (c)
4282	goto out_free;
4283	status = nfserr_conn_not_bound_to_session;
4284	if (clp->cl_mach_cred)
4285	goto out_free;
4286	__nfsd4_hash_conn(conn: new, ses);
4287	spin_unlock(lock: &clp->cl_lock);
4288	ret = nfsd4_register_conn(conn: new);
4289	if (ret)
4290	/* oops; xprt is already down: */
4291	nfsd4_conn_lost(u: &new->cn_xpt_user);
4292	return nfs_ok;
4293	out_free:
4294	spin_unlock(lock: &clp->cl_lock);
4295	free_conn(c: new);
4296	return status;
4297	}
4298
4299	static bool nfsd4_session_too_many_ops(struct svc_rqst *rqstp, struct nfsd4_session *session)
4300	{
4301	struct nfsd4_compoundargs *args = rqstp->rq_argp;
4302
4303	return args->opcnt > session->se_fchannel.maxops;
4304	}
4305
4306	static bool nfsd4_request_too_big(struct svc_rqst *rqstp,
4307	struct nfsd4_session *session)
4308	{
4309	struct xdr_buf *xb = &rqstp->rq_arg;
4310
4311	return xb->len > session->se_fchannel.maxreq_sz;
4312	}
4313
4314	static bool replay_matches_cache(struct svc_rqst *rqstp,
4315	struct nfsd4_sequence *seq, struct nfsd4_slot *slot)
4316	{
4317	struct nfsd4_compoundargs *argp = rqstp->rq_argp;
4318
4319	if ((bool)(slot->sl_flags & NFSD4_SLOT_CACHETHIS) !=
4320	(bool)seq->cachethis)
4321	return false;
4322	/*
4323	* If there's an error then the reply can have fewer ops than
4324	* the call.
4325	*/
4326	if (slot->sl_opcnt < argp->opcnt && !slot->sl_status)
4327	return false;
4328	/*
4329	* But if we cached a reply with *more* ops than the call you're
4330	* sending us now, then this new call is clearly not really a
4331	* replay of the old one:
4332	*/
4333	if (slot->sl_opcnt > argp->opcnt)
4334	return false;
4335	/* This is the only check explicitly called by spec: */
4336	if (!same_creds(cr1: &rqstp->rq_cred, cr2: &slot->sl_cred))
4337	return false;
4338	/*
4339	* There may be more comparisons we could actually do, but the
4340	* spec doesn't require us to catch every case where the calls
4341	* don't match (that would require caching the call as well as
4342	* the reply), so we don't bother.
4343	*/
4344	return true;
4345	}
4346
4347	/*
4348	* Note that the response is constructed here both for the case
4349	* of a new SEQUENCE request and for a replayed SEQUENCE request.
4350	* We do not cache SEQUENCE responses as SEQUENCE is idempotent.
4351	*/
4352	static void nfsd4_construct_sequence_response(struct nfsd4_session *session,
4353	struct nfsd4_sequence *seq)
4354	{
4355	struct nfs4_client *clp = session->se_client;
4356
4357	seq->maxslots_response = max(session->se_target_maxslots,
4358	seq->maxslots);
4359	seq->target_maxslots = session->se_target_maxslots;
4360
4361	switch (clp->cl_cb_state) {
4362	case NFSD4_CB_DOWN:
4363	seq->status_flags = SEQ4_STATUS_CB_PATH_DOWN;
4364	break;
4365	case NFSD4_CB_FAULT:
4366	seq->status_flags = SEQ4_STATUS_BACKCHANNEL_FAULT;
4367	break;
4368	default:
4369	seq->status_flags = 0;
4370	}
4371	if (!list_empty(head: &clp->cl_revoked))
4372	seq->status_flags |= SEQ4_STATUS_RECALLABLE_STATE_REVOKED;
4373	if (atomic_read(v: &clp->cl_admin_revoked))
4374	seq->status_flags |= SEQ4_STATUS_ADMIN_STATE_REVOKED;
4375	}
4376
4377	__be32
4378	nfsd4_sequence(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4379	union nfsd4_op_u *u)
4380	{
4381	struct nfsd4_sequence *seq = &u->sequence;
4382	struct nfsd4_compoundres *resp = rqstp->rq_resp;
4383	struct xdr_stream *xdr = resp->xdr;
4384	struct nfsd4_session *session;
4385	struct nfs4_client *clp;
4386	struct nfsd4_slot *slot;
4387	struct nfsd4_conn *conn;
4388	__be32 status;
4389	int buflen;
4390	struct net *net = SVC_NET(rqstp);
4391	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
4392
4393	if (resp->opcnt != 1)
4394	return nfserr_sequence_pos;
4395
4396	/*
4397	* Will be either used or freed by nfsd4_sequence_check_conn
4398	* below.
4399	*/
4400	conn = alloc_conn(rqstp, NFS4_CDFC4_FORE);
4401	if (!conn)
4402	return nfserr_jukebox;
4403
4404	spin_lock(lock: &nn->client_lock);
4405	session = find_in_sessionid_hashtbl(sessionid: &seq->sessionid, net, ret: &status);
4406	if (!session)
4407	goto out_no_session;
4408	clp = session->se_client;
4409
4410	status = nfserr_too_many_ops;
4411	if (nfsd4_session_too_many_ops(rqstp, session))
4412	goto out_put_session;
4413
4414	status = nfserr_req_too_big;
4415	if (nfsd4_request_too_big(rqstp, session))
4416	goto out_put_session;
4417
4418	status = nfserr_badslot;
4419	if (seq->slotid >= session->se_fchannel.maxreqs)
4420	goto out_put_session;
4421
4422	slot = xa_load(&session->se_slots, index: seq->slotid);
4423	dprintk("%s: slotid %d\n", __func__, seq->slotid);
4424
4425	trace_nfsd_slot_seqid_sequence(clp, seq, slot);
4426
4427	nfsd4_construct_sequence_response(session, seq);
4428
4429	status = check_slot_seqid(seqid: seq->seqid, slot_seqid: slot->sl_seqid, flags: slot->sl_flags);
4430	if (status == nfserr_replay_cache) {
4431	status = nfserr_seq_misordered;
4432	if (!(slot->sl_flags & NFSD4_SLOT_INITIALIZED))
4433	goto out_put_session;
4434	status = nfserr_seq_false_retry;
4435	if (!replay_matches_cache(rqstp, seq, slot))
4436	goto out_put_session;
4437	cstate->slot = slot;
4438	cstate->session = session;
4439	cstate->clp = clp;
4440	/* Return the cached reply status and set cstate->status
4441	* for nfsd4_proc_compound processing */
4442	status = nfsd4_replay_cache_entry(resp, seq);
4443	cstate->status = nfserr_replay_cache;
4444	goto out;
4445	}
4446	if (status)
4447	goto out_put_session;
4448
4449	status = nfsd4_sequence_check_conn(new: conn, ses: session);
4450	conn = NULL;
4451	if (status)
4452	goto out_put_session;
4453
4454	if (session->se_target_maxslots < session->se_fchannel.maxreqs &&
4455	slot->sl_generation == session->se_slot_gen &&
4456	seq->maxslots <= session->se_target_maxslots)
4457	/* Client acknowledged our reduce maxreqs */
4458	free_session_slots(ses: session, from: session->se_target_maxslots);
4459
4460	buflen = (seq->cachethis) ?
4461	session->se_fchannel.maxresp_cached :
4462	session->se_fchannel.maxresp_sz;
4463	status = (seq->cachethis) ? nfserr_rep_too_big_to_cache :
4464	nfserr_rep_too_big;
4465	if (xdr_restrict_buflen(xdr, newbuflen: buflen - rqstp->rq_auth_slack))
4466	goto out_put_session;
4467	svc_reserve_auth(rqstp, space: buflen);
4468
4469	status = nfs_ok;
4470	/* Success! accept new slot seqid */
4471	slot->sl_seqid = seq->seqid;
4472	slot->sl_flags &= ~NFSD4_SLOT_REUSED;
4473	slot->sl_flags |= NFSD4_SLOT_INUSE;
4474	slot->sl_generation = session->se_slot_gen;
4475	if (seq->cachethis)
4476	slot->sl_flags |= NFSD4_SLOT_CACHETHIS;
4477	else
4478	slot->sl_flags &= ~NFSD4_SLOT_CACHETHIS;
4479
4480	cstate->slot = slot;
4481	cstate->session = session;
4482	cstate->clp = clp;
4483
4484	/*
4485	* If the client ever uses the highest available slot,
4486	* gently try to allocate another 20%. This allows
4487	* fairly quick growth without grossly over-shooting what
4488	* the client might use.
4489	*/
4490	if (seq->slotid == session->se_fchannel.maxreqs - 1 &&
4491	session->se_target_maxslots >= session->se_fchannel.maxreqs &&
4492	session->se_fchannel.maxreqs < NFSD_MAX_SLOTS_PER_SESSION) {
4493	int s = session->se_fchannel.maxreqs;
4494	int cnt = DIV_ROUND_UP(s, 5);
4495	void *prev_slot;
4496
4497	do {
4498	/*
4499	* GFP_NOWAIT both allows allocation under a
4500	* spinlock, and only succeeds if there is
4501	* plenty of memory.
4502	*/
4503	slot = nfsd4_alloc_slot(fattrs: &session->se_fchannel, index: s,
4504	GFP_NOWAIT);
4505	prev_slot = xa_load(&session->se_slots, index: s);
4506	if (xa_is_value(entry: prev_slot) && slot) {
4507	slot->sl_seqid = xa_to_value(entry: prev_slot);
4508	slot->sl_flags |= NFSD4_SLOT_REUSED;
4509	}
4510	if (slot &&
4511	!xa_is_err(entry: xa_store(&session->se_slots, index: s, entry: slot,
4512	GFP_NOWAIT))) {
4513	s += 1;
4514	session->se_fchannel.maxreqs = s;
4515	atomic_add(i: s - session->se_target_maxslots,
4516	v: &nfsd_total_target_slots);
4517	session->se_target_maxslots = s;
4518	} else {
4519	kfree(objp: slot);
4520	slot = NULL;
4521	}
4522	} while (slot && --cnt > 0);
4523	}
4524
4525	out:
4526	trace_nfsd_seq4_status(rqstp, sequence: seq);
4527	out_no_session:
4528	if (conn)
4529	free_conn(c: conn);
4530	spin_unlock(lock: &nn->client_lock);
4531	return status;
4532	out_put_session:
4533	nfsd4_put_session_locked(ses: session);
4534	goto out_no_session;
4535	}
4536
4537	void
4538	nfsd4_sequence_done(struct nfsd4_compoundres *resp)
4539	{
4540	struct nfsd4_compound_state *cs = &resp->cstate;
4541
4542	if (nfsd4_has_session(cs)) {
4543	if (cs->status != nfserr_replay_cache) {
4544	nfsd4_store_cache_entry(resp);
4545	cs->slot->sl_flags &= ~NFSD4_SLOT_INUSE;
4546	}
4547	/* Drop session reference that was taken in nfsd4_sequence() */
4548	nfsd4_put_session(ses: cs->session);
4549	} else if (cs->clp)
4550	put_client_renew(clp: cs->clp);
4551	}
4552
4553	__be32
4554	nfsd4_destroy_clientid(struct svc_rqst *rqstp,
4555	struct nfsd4_compound_state *cstate,
4556	union nfsd4_op_u *u)
4557	{
4558	struct nfsd4_destroy_clientid *dc = &u->destroy_clientid;
4559	struct nfs4_client *conf, *unconf;
4560	struct nfs4_client *clp = NULL;
4561	__be32 status = 0;
4562	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
4563
4564	spin_lock(lock: &nn->client_lock);
4565	unconf = find_unconfirmed_client(clid: &dc->clientid, sessions: true, nn);
4566	conf = find_confirmed_client(clid: &dc->clientid, sessions: true, nn);
4567	WARN_ON_ONCE(conf && unconf);
4568
4569	if (conf) {
4570	if (client_has_state(clp: conf)) {
4571	status = nfserr_clientid_busy;
4572	goto out;
4573	}
4574	status = mark_client_expired_locked(clp: conf);
4575	if (status)
4576	goto out;
4577	clp = conf;
4578	} else if (unconf)
4579	clp = unconf;
4580	else {
4581	status = nfserr_stale_clientid;
4582	goto out;
4583	}
4584	if (!nfsd4_mach_creds_match(cl: clp, rqstp)) {
4585	clp = NULL;
4586	status = nfserr_wrong_cred;
4587	goto out;
4588	}
4589	trace_nfsd_clid_destroyed(clid: &clp->cl_clientid);
4590	unhash_client_locked(clp);
4591	out:
4592	spin_unlock(lock: &nn->client_lock);
4593	if (clp)
4594	expire_client(clp);
4595	return status;
4596	}
4597
4598	__be32
4599	nfsd4_reclaim_complete(struct svc_rqst *rqstp,
4600	struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
4601	{
4602	struct nfsd4_reclaim_complete *rc = &u->reclaim_complete;
4603	struct nfs4_client *clp = cstate->clp;
4604	__be32 status = 0;
4605
4606	if (rc->rca_one_fs) {
4607	if (!cstate->current_fh.fh_dentry)
4608	return nfserr_nofilehandle;
4609	/*
4610	* We don't take advantage of the rca_one_fs case.
4611	* That's OK, it's optional, we can safely ignore it.
4612	*/
4613	return nfs_ok;
4614	}
4615
4616	status = nfserr_complete_already;
4617	if (test_and_set_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, addr: &clp->cl_flags))
4618	goto out;
4619
4620	status = nfserr_stale_clientid;
4621	if (is_client_expired(clp))
4622	/*
4623	* The following error isn't really legal.
4624	* But we only get here if the client just explicitly
4625	* destroyed the client. Surely it no longer cares what
4626	* error it gets back on an operation for the dead
4627	* client.
4628	*/
4629	goto out;
4630
4631	status = nfs_ok;
4632	trace_nfsd_clid_reclaim_complete(clid: &clp->cl_clientid);
4633	nfsd4_client_record_create(clp);
4634	inc_reclaim_complete(clp);
4635	out:
4636	return status;
4637	}
4638
4639	__be32
4640	nfsd4_setclientid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
4641	union nfsd4_op_u *u)
4642	{
4643	struct nfsd4_setclientid *setclid = &u->setclientid;
4644	struct xdr_netobj clname = setclid->se_name;
4645	nfs4_verifier clverifier = setclid->se_verf;
4646	struct nfs4_client *conf, *new;
4647	struct nfs4_client *unconf = NULL;
4648	__be32 status;
4649	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
4650
4651	new = create_client(name: clname, rqstp, verf: &clverifier);
4652	if (new == NULL)
4653	return nfserr_jukebox;
4654	spin_lock(lock: &nn->client_lock);
4655	conf = find_confirmed_client_by_name(name: &clname, nn);
4656	if (conf && client_has_state(clp: conf)) {
4657	status = nfserr_clid_inuse;
4658	if (clp_used_exchangeid(clp: conf))
4659	goto out;
4660	if (!same_creds(cr1: &conf->cl_cred, cr2: &rqstp->rq_cred)) {
4661	trace_nfsd_clid_cred_mismatch(clp: conf, rqstp);
4662	goto out;
4663	}
4664	}
4665	unconf = find_unconfirmed_client_by_name(name: &clname, nn);
4666	if (unconf)
4667	unhash_client_locked(clp: unconf);
4668	if (conf) {
4669	if (same_verf(v1: &conf->cl_verifier, v2: &clverifier)) {
4670	copy_clid(target: new, source: conf);
4671	gen_confirm(clp: new, nn);
4672	} else
4673	trace_nfsd_clid_verf_mismatch(clp: conf, rqstp,
4674	verf: &clverifier);
4675	} else
4676	trace_nfsd_clid_fresh(clp: new);
4677	new->cl_minorversion = 0;
4678	gen_callback(clp: new, se: setclid, rqstp);
4679	add_to_unconfirmed(clp: new);
4680	setclid->se_clientid.cl_boot = new->cl_clientid.cl_boot;
4681	setclid->se_clientid.cl_id = new->cl_clientid.cl_id;
4682	memcpy(setclid->se_confirm.data, new->cl_confirm.data, sizeof(setclid->se_confirm.data));
4683	new = NULL;
4684	status = nfs_ok;
4685	out:
4686	spin_unlock(lock: &nn->client_lock);
4687	if (new)
4688	free_client(clp: new);
4689	if (unconf) {
4690	trace_nfsd_clid_expire_unconf(clid: &unconf->cl_clientid);
4691	expire_client(clp: unconf);
4692	}
4693	return status;
4694	}
4695
4696	__be32
4697	nfsd4_setclientid_confirm(struct svc_rqst *rqstp,
4698	struct nfsd4_compound_state *cstate,
4699	union nfsd4_op_u *u)
4700	{
4701	struct nfsd4_setclientid_confirm *setclientid_confirm =
4702	&u->setclientid_confirm;
4703	struct nfs4_client *conf, *unconf;
4704	struct nfs4_client *old = NULL;
4705	nfs4_verifier confirm = setclientid_confirm->sc_confirm;
4706	clientid_t * clid = &setclientid_confirm->sc_clientid;
4707	__be32 status;
4708	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
4709
4710	if (STALE_CLIENTID(clid, nn))
4711	return nfserr_stale_clientid;
4712
4713	spin_lock(lock: &nn->client_lock);
4714	conf = find_confirmed_client(clid, sessions: false, nn);
4715	unconf = find_unconfirmed_client(clid, sessions: false, nn);
4716	/*
4717	* We try hard to give out unique clientid's, so if we get an
4718	* attempt to confirm the same clientid with a different cred,
4719	* the client may be buggy; this should never happen.
4720	*
4721	* Nevertheless, RFC 7530 recommends INUSE for this case:
4722	*/
4723	status = nfserr_clid_inuse;
4724	if (unconf && !same_creds(cr1: &unconf->cl_cred, cr2: &rqstp->rq_cred)) {
4725	trace_nfsd_clid_cred_mismatch(clp: unconf, rqstp);
4726	goto out;
4727	}
4728	if (conf && !same_creds(cr1: &conf->cl_cred, cr2: &rqstp->rq_cred)) {
4729	trace_nfsd_clid_cred_mismatch(clp: conf, rqstp);
4730	goto out;
4731	}
4732	if (!unconf || !same_verf(v1: &confirm, v2: &unconf->cl_confirm)) {
4733	if (conf && same_verf(v1: &confirm, v2: &conf->cl_confirm)) {
4734	status = nfs_ok;
4735	} else
4736	status = nfserr_stale_clientid;
4737	goto out;
4738	}
4739	status = nfs_ok;
4740	if (conf) {
4741	if (get_client_locked(clp: conf) == nfs_ok) {
4742	old = unconf;
4743	unhash_client_locked(clp: old);
4744	nfsd4_change_callback(clp: conf, &unconf->cl_cb_conn);
4745	} else {
4746	conf = NULL;
4747	}
4748	}
4749
4750	if (!conf) {
4751	old = find_confirmed_client_by_name(name: &unconf->cl_name, nn);
4752	if (old) {
4753	status = nfserr_clid_inuse;
4754	if (client_has_state(clp: old)
4755	&& !same_creds(cr1: &unconf->cl_cred,
4756	cr2: &old->cl_cred)) {
4757	old = NULL;
4758	goto out;
4759	}
4760	status = mark_client_expired_locked(clp: old);
4761	if (status) {
4762	old = NULL;
4763	goto out;
4764	}
4765	trace_nfsd_clid_replaced(clid: &old->cl_clientid);
4766	}
4767	status = get_client_locked(clp: unconf);
4768	if (status != nfs_ok) {
4769	old = NULL;
4770	goto out;
4771	}
4772	move_to_confirmed(clp: unconf);
4773	conf = unconf;
4774	}
4775	spin_unlock(lock: &nn->client_lock);
4776	if (conf == unconf)
4777	fsnotify_dentry(dentry: conf->cl_nfsd_info_dentry, FS_MODIFY);
4778	nfsd4_probe_callback(clp: conf);
4779	spin_lock(lock: &nn->client_lock);
4780	put_client_renew_locked(clp: conf);
4781	out:
4782	spin_unlock(lock: &nn->client_lock);
4783	if (old)
4784	expire_client(clp: old);
4785	return status;
4786	}
4787
4788	static struct nfs4_file *nfsd4_alloc_file(void)
4789	{
4790	return kmem_cache_alloc(file_slab, GFP_KERNEL);
4791	}
4792
4793	/* OPEN Share state helper functions */
4794
4795	static void nfsd4_file_init(const struct svc_fh *fh, struct nfs4_file *fp)
4796	{
4797	refcount_set(r: &fp->fi_ref, n: 1);
4798	spin_lock_init(&fp->fi_lock);
4799	INIT_LIST_HEAD(list: &fp->fi_stateids);
4800	INIT_LIST_HEAD(list: &fp->fi_delegations);
4801	INIT_LIST_HEAD(list: &fp->fi_clnt_odstate);
4802	fh_copy_shallow(dst: &fp->fi_fhandle, src: &fh->fh_handle);
4803	fp->fi_deleg_file = NULL;
4804	fp->fi_rdeleg_file = NULL;
4805	fp->fi_had_conflict = false;
4806	fp->fi_share_deny = 0;
4807	memset(fp->fi_fds, 0, sizeof(fp->fi_fds));
4808	memset(fp->fi_access, 0, sizeof(fp->fi_access));
4809	fp->fi_aliased = false;
4810	fp->fi_inode = d_inode(dentry: fh->fh_dentry);
4811	#ifdef CONFIG_NFSD_PNFS
4812	INIT_LIST_HEAD(list: &fp->fi_lo_states);
4813	atomic_set(v: &fp->fi_lo_recalls, i: 0);
4814	#endif
4815	}
4816
4817	void
4818	nfsd4_free_slabs(void)
4819	{
4820	kmem_cache_destroy(s: client_slab);
4821	kmem_cache_destroy(s: openowner_slab);
4822	kmem_cache_destroy(s: lockowner_slab);
4823	kmem_cache_destroy(s: file_slab);
4824	kmem_cache_destroy(s: stateid_slab);
4825	kmem_cache_destroy(s: deleg_slab);
4826	kmem_cache_destroy(s: odstate_slab);
4827	}
4828
4829	int
4830	nfsd4_init_slabs(void)
4831	{
4832	client_slab = KMEM_CACHE(nfs4_client, 0);
4833	if (client_slab == NULL)
4834	goto out;
4835	openowner_slab = KMEM_CACHE(nfs4_openowner, 0);
4836	if (openowner_slab == NULL)
4837	goto out_free_client_slab;
4838	lockowner_slab = KMEM_CACHE(nfs4_lockowner, 0);
4839	if (lockowner_slab == NULL)
4840	goto out_free_openowner_slab;
4841	file_slab = KMEM_CACHE(nfs4_file, 0);
4842	if (file_slab == NULL)
4843	goto out_free_lockowner_slab;
4844	stateid_slab = KMEM_CACHE(nfs4_ol_stateid, 0);
4845	if (stateid_slab == NULL)
4846	goto out_free_file_slab;
4847	deleg_slab = KMEM_CACHE(nfs4_delegation, 0);
4848	if (deleg_slab == NULL)
4849	goto out_free_stateid_slab;
4850	odstate_slab = KMEM_CACHE(nfs4_clnt_odstate, 0);
4851	if (odstate_slab == NULL)
4852	goto out_free_deleg_slab;
4853	return 0;
4854
4855	out_free_deleg_slab:
4856	kmem_cache_destroy(s: deleg_slab);
4857	out_free_stateid_slab:
4858	kmem_cache_destroy(s: stateid_slab);
4859	out_free_file_slab:
4860	kmem_cache_destroy(s: file_slab);
4861	out_free_lockowner_slab:
4862	kmem_cache_destroy(s: lockowner_slab);
4863	out_free_openowner_slab:
4864	kmem_cache_destroy(s: openowner_slab);
4865	out_free_client_slab:
4866	kmem_cache_destroy(s: client_slab);
4867	out:
4868	return -ENOMEM;
4869	}
4870
4871	static unsigned long
4872	nfsd4_state_shrinker_count(struct shrinker *shrink, struct shrink_control *sc)
4873	{
4874	struct nfsd_net *nn = shrink->private_data;
4875	long count;
4876
4877	count = atomic_read(v: &nn->nfsd_courtesy_clients);
4878	if (!count)
4879	count = atomic_long_read(v: &num_delegations);
4880	if (count)
4881	queue_work(wq: laundry_wq, work: &nn->nfsd_shrinker_work);
4882	return (unsigned long)count;
4883	}
4884
4885	static unsigned long
4886	nfsd4_state_shrinker_scan(struct shrinker *shrink, struct shrink_control *sc)
4887	{
4888	return SHRINK_STOP;
4889	}
4890
4891	void
4892	nfsd4_init_leases_net(struct nfsd_net *nn)
4893	{
4894	struct sysinfo si;
4895	u64 max_clients;
4896
4897	nn->nfsd4_lease = 90; /* default lease time */
4898	nn->nfsd4_grace = 90;
4899	nn->somebody_reclaimed = false;
4900	nn->track_reclaim_completes = false;
4901	nn->clverifier_counter = get_random_u32();
4902	nn->clientid_base = get_random_u32();
4903	nn->clientid_counter = nn->clientid_base + 1;
4904	nn->s2s_cp_cl_id = nn->clientid_counter++;
4905
4906	atomic_set(v: &nn->nfs4_client_count, i: 0);
4907	si_meminfo(val: &si);
4908	max_clients = (u64)si.totalram * si.mem_unit / (1024 * 1024 * 1024);
4909	max_clients *= NFS4_CLIENTS_PER_GB;
4910	nn->nfs4_max_clients = max_t(int, max_clients, NFS4_CLIENTS_PER_GB);
4911
4912	atomic_set(v: &nn->nfsd_courtesy_clients, i: 0);
4913	}
4914
4915	enum rp_lock {
4916	RP_UNLOCKED,
4917	RP_LOCKED,
4918	RP_UNHASHED,
4919	};
4920
4921	static void init_nfs4_replay(struct nfs4_replay *rp)
4922	{
4923	rp->rp_status = nfserr_serverfault;
4924	rp->rp_buflen = 0;
4925	rp->rp_buf = rp->rp_ibuf;
4926	rp->rp_locked = RP_UNLOCKED;
4927	}
4928
4929	static int nfsd4_cstate_assign_replay(struct nfsd4_compound_state *cstate,
4930	struct nfs4_stateowner *so)
4931	{
4932	if (!nfsd4_has_session(cs: cstate)) {
4933	wait_var_event(&so->so_replay.rp_locked,
4934	cmpxchg(&so->so_replay.rp_locked,
4935	RP_UNLOCKED, RP_LOCKED) != RP_LOCKED);
4936	if (so->so_replay.rp_locked == RP_UNHASHED)
4937	return -EAGAIN;
4938	cstate->replay_owner = nfs4_get_stateowner(sop: so);
4939	}
4940	return 0;
4941	}
4942
4943	void nfsd4_cstate_clear_replay(struct nfsd4_compound_state *cstate)
4944	{
4945	struct nfs4_stateowner *so = cstate->replay_owner;
4946
4947	if (so != NULL) {
4948	cstate->replay_owner = NULL;
4949	store_release_wake_up(&so->so_replay.rp_locked, RP_UNLOCKED);
4950	nfs4_put_stateowner(sop: so);
4951	}
4952	}
4953
4954	static inline void *alloc_stateowner(struct kmem_cache *slab, struct xdr_netobj *owner, struct nfs4_client *clp)
4955	{
4956	struct nfs4_stateowner *sop;
4957
4958	sop = kmem_cache_alloc(slab, GFP_KERNEL);
4959	if (!sop)
4960	return NULL;
4961
4962	xdr_netobj_dup(dst: &sop->so_owner, src: owner, GFP_KERNEL);
4963	if (!sop->so_owner.data) {
4964	kmem_cache_free(s: slab, objp: sop);
4965	return NULL;
4966	}
4967
4968	INIT_LIST_HEAD(list: &sop->so_stateids);
4969	sop->so_client = clp;
4970	init_nfs4_replay(rp: &sop->so_replay);
4971	atomic_set(v: &sop->so_count, i: 1);
4972	return sop;
4973	}
4974
4975	static void hash_openowner(struct nfs4_openowner *oo, struct nfs4_client *clp, unsigned int strhashval)
4976	{
4977	lockdep_assert_held(&clp->cl_lock);
4978
4979	list_add(new: &oo->oo_owner.so_strhash,
4980	head: &clp->cl_ownerstr_hashtbl[strhashval]);
4981	list_add(new: &oo->oo_perclient, head: &clp->cl_openowners);
4982	}
4983
4984	static void nfs4_unhash_openowner(struct nfs4_stateowner *so)
4985	{
4986	unhash_openowner_locked(oo: openowner(so));
4987	}
4988
4989	static void nfs4_free_openowner(struct nfs4_stateowner *so)
4990	{
4991	struct nfs4_openowner *oo = openowner(so);
4992
4993	kmem_cache_free(s: openowner_slab, objp: oo);
4994	}
4995
4996	static const struct nfs4_stateowner_operations openowner_ops = {
4997	.so_unhash = nfs4_unhash_openowner,
4998	.so_free = nfs4_free_openowner,
4999	};
5000
5001	static struct nfs4_ol_stateid *
5002	nfsd4_find_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
5003	{
5004	struct nfs4_ol_stateid *local, *ret = NULL;
5005	struct nfs4_openowner *oo = open->op_openowner;
5006
5007	lockdep_assert_held(&fp->fi_lock);
5008
5009	list_for_each_entry(local, &fp->fi_stateids, st_perfile) {
5010	/* ignore lock owners */
5011	if (local->st_stateowner->so_is_open_owner == 0)
5012	continue;
5013	if (local->st_stateowner != &oo->oo_owner)
5014	continue;
5015	if (local->st_stid.sc_type == SC_TYPE_OPEN &&
5016	!local->st_stid.sc_status) {
5017	ret = local;
5018	refcount_inc(r: &ret->st_stid.sc_count);
5019	break;
5020	}
5021	}
5022	return ret;
5023	}
5024
5025	static void nfsd4_drop_revoked_stid(struct nfs4_stid *s)
5026	__releases(&s->sc_client->cl_lock)
5027	{
5028	struct nfs4_client *cl = s->sc_client;
5029	LIST_HEAD(reaplist);
5030	struct nfs4_ol_stateid *stp;
5031	struct nfs4_delegation *dp;
5032	bool unhashed;
5033
5034	switch (s->sc_type) {
5035	case SC_TYPE_OPEN:
5036	stp = openlockstateid(s);
5037	if (unhash_open_stateid(stp, reaplist: &reaplist))
5038	put_ol_stateid_locked(stp, reaplist: &reaplist);
5039	spin_unlock(lock: &cl->cl_lock);
5040	free_ol_stateid_reaplist(reaplist: &reaplist);
5041	break;
5042	case SC_TYPE_LOCK:
5043	stp = openlockstateid(s);
5044	unhashed = unhash_lock_stateid(stp);
5045	spin_unlock(lock: &cl->cl_lock);
5046	if (unhashed)
5047	nfs4_put_stid(s);
5048	break;
5049	case SC_TYPE_DELEG:
5050	dp = delegstateid(s);
5051	list_del_init(entry: &dp->dl_recall_lru);
5052	spin_unlock(lock: &cl->cl_lock);
5053	nfs4_put_stid(s);
5054	break;
5055	default:
5056	spin_unlock(lock: &cl->cl_lock);
5057	}
5058	}
5059
5060	static void nfsd40_drop_revoked_stid(struct nfs4_client *cl,
5061	stateid_t *stid)
5062	{
5063	/* NFSv4.0 has no way for the client to tell the server
5064	* that it can forget an admin-revoked stateid.
5065	* So we keep it around until the first time that the
5066	* client uses it, and drop it the first time
5067	* nfserr_admin_revoked is returned.
5068	* For v4.1 and later we wait until explicitly told
5069	* to free the stateid.
5070	*/
5071	if (cl->cl_minorversion == 0) {
5072	struct nfs4_stid *st;
5073
5074	spin_lock(lock: &cl->cl_lock);
5075	st = find_stateid_locked(cl, t: stid);
5076	if (st)
5077	nfsd4_drop_revoked_stid(s: st);
5078	else
5079	spin_unlock(lock: &cl->cl_lock);
5080	}
5081	}
5082
5083	static __be32
5084	nfsd4_verify_open_stid(struct nfs4_stid *s)
5085	{
5086	__be32 ret = nfs_ok;
5087
5088	if (s->sc_status & SC_STATUS_ADMIN_REVOKED)
5089	ret = nfserr_admin_revoked;
5090	else if (s->sc_status & SC_STATUS_REVOKED)
5091	ret = nfserr_deleg_revoked;
5092	else if (s->sc_status & SC_STATUS_CLOSED)
5093	ret = nfserr_bad_stateid;
5094	return ret;
5095	}
5096
5097	/* Lock the stateid st_mutex, and deal with races with CLOSE */
5098	static __be32
5099	nfsd4_lock_ol_stateid(struct nfs4_ol_stateid *stp)
5100	{
5101	__be32 ret;
5102
5103	mutex_lock_nested(lock: &stp->st_mutex, subclass: LOCK_STATEID_MUTEX);
5104	ret = nfsd4_verify_open_stid(s: &stp->st_stid);
5105	if (ret == nfserr_admin_revoked)
5106	nfsd40_drop_revoked_stid(cl: stp->st_stid.sc_client,
5107	stid: &stp->st_stid.sc_stateid);
5108
5109	if (ret != nfs_ok)
5110	mutex_unlock(lock: &stp->st_mutex);
5111	return ret;
5112	}
5113
5114	static struct nfs4_ol_stateid *
5115	nfsd4_find_and_lock_existing_open(struct nfs4_file *fp, struct nfsd4_open *open)
5116	{
5117	struct nfs4_ol_stateid *stp;
5118	for (;;) {
5119	spin_lock(lock: &fp->fi_lock);
5120	stp = nfsd4_find_existing_open(fp, open);
5121	spin_unlock(lock: &fp->fi_lock);
5122	if (!stp || nfsd4_lock_ol_stateid(stp) == nfs_ok)
5123	break;
5124	nfs4_put_stid(s: &stp->st_stid);
5125	}
5126	return stp;
5127	}
5128
5129	static struct nfs4_openowner *
5130	find_or_alloc_open_stateowner(unsigned int strhashval, struct nfsd4_open *open,
5131	struct nfsd4_compound_state *cstate)
5132	{
5133	struct nfs4_client *clp = cstate->clp;
5134	struct nfs4_openowner *oo, *new = NULL;
5135
5136	retry:
5137	spin_lock(lock: &clp->cl_lock);
5138	oo = find_openstateowner_str(hashval: strhashval, open, clp);
5139	if (!oo && new) {
5140	hash_openowner(oo: new, clp, strhashval);
5141	spin_unlock(lock: &clp->cl_lock);
5142	return new;
5143	}
5144	spin_unlock(lock: &clp->cl_lock);
5145
5146	if (oo && !(oo->oo_flags & NFS4_OO_CONFIRMED)) {
5147	/* Replace unconfirmed owners without checking for replay. */
5148	release_openowner(oo);
5149	oo = NULL;
5150	}
5151	if (oo) {
5152	if (new)
5153	nfs4_free_stateowner(sop: &new->oo_owner);
5154	return oo;
5155	}
5156
5157	new = alloc_stateowner(slab: openowner_slab, owner: &open->op_owner, clp);
5158	if (!new)
5159	return NULL;
5160	new->oo_owner.so_ops = &openowner_ops;
5161	new->oo_owner.so_is_open_owner = 1;
5162	new->oo_owner.so_seqid = open->op_seqid;
5163	new->oo_flags = 0;
5164	if (nfsd4_has_session(cs: cstate))
5165	new->oo_flags |= NFS4_OO_CONFIRMED;
5166	new->oo_time = 0;
5167	new->oo_last_closed_stid = NULL;
5168	INIT_LIST_HEAD(list: &new->oo_close_lru);
5169	goto retry;
5170	}
5171
5172	static struct nfs4_ol_stateid *
5173	init_open_stateid(struct nfs4_file *fp, struct nfsd4_open *open)
5174	{
5175
5176	struct nfs4_openowner *oo = open->op_openowner;
5177	struct nfs4_ol_stateid *retstp = NULL;
5178	struct nfs4_ol_stateid *stp;
5179
5180	stp = open->op_stp;
5181	/* We are moving these outside of the spinlocks to avoid the warnings */
5182	mutex_init(&stp->st_mutex);
5183	mutex_lock_nested(lock: &stp->st_mutex, subclass: OPEN_STATEID_MUTEX);
5184
5185	retry:
5186	spin_lock(lock: &oo->oo_owner.so_client->cl_lock);
5187	spin_lock(lock: &fp->fi_lock);
5188
5189	if (nfs4_openowner_unhashed(oo)) {
5190	mutex_unlock(lock: &stp->st_mutex);
5191	stp = NULL;
5192	goto out_unlock;
5193	}
5194
5195	retstp = nfsd4_find_existing_open(fp, open);
5196	if (retstp)
5197	goto out_unlock;
5198
5199	open->op_stp = NULL;
5200	refcount_inc(r: &stp->st_stid.sc_count);
5201	stp->st_stid.sc_type = SC_TYPE_OPEN;
5202	INIT_LIST_HEAD(list: &stp->st_locks);
5203	stp->st_stateowner = nfs4_get_stateowner(sop: &oo->oo_owner);
5204	get_nfs4_file(fi: fp);
5205	stp->st_stid.sc_file = fp;
5206	stp->st_access_bmap = 0;
5207	stp->st_deny_bmap = 0;
5208	stp->st_openstp = NULL;
5209	list_add(new: &stp->st_perstateowner, head: &oo->oo_owner.so_stateids);
5210	list_add(new: &stp->st_perfile, head: &fp->fi_stateids);
5211
5212	out_unlock:
5213	spin_unlock(lock: &fp->fi_lock);
5214	spin_unlock(lock: &oo->oo_owner.so_client->cl_lock);
5215	if (retstp) {
5216	/* Handle races with CLOSE */
5217	if (nfsd4_lock_ol_stateid(stp: retstp) != nfs_ok) {
5218	nfs4_put_stid(s: &retstp->st_stid);
5219	goto retry;
5220	}
5221	/* To keep mutex tracking happy */
5222	mutex_unlock(lock: &stp->st_mutex);
5223	stp = retstp;
5224	}
5225	return stp;
5226	}
5227
5228	/*
5229	* In the 4.0 case we need to keep the owners around a little while to handle
5230	* CLOSE replay. We still do need to release any file access that is held by
5231	* them before returning however.
5232	*/
5233	static void
5234	move_to_close_lru(struct nfs4_ol_stateid *s, struct net *net)
5235	{
5236	struct nfs4_ol_stateid *last;
5237	struct nfs4_openowner *oo = openowner(so: s->st_stateowner);
5238	struct nfsd_net *nn = net_generic(net: s->st_stid.sc_client->net,
5239	id: nfsd_net_id);
5240
5241	dprintk("NFSD: move_to_close_lru nfs4_openowner %p\n", oo);
5242
5243	/*
5244	* We know that we hold one reference via nfsd4_close, and another
5245	* "persistent" reference for the client. If the refcount is higher
5246	* than 2, then there are still calls in progress that are using this
5247	* stateid. We can't put the sc_file reference until they are finished.
5248	* Wait for the refcount to drop to 2. Since it has been unhashed,
5249	* there should be no danger of the refcount going back up again at
5250	* this point.
5251	* Some threads with a reference might be waiting for rp_locked,
5252	* so tell them to stop waiting.
5253	*/
5254	store_release_wake_up(&oo->oo_owner.so_replay.rp_locked, RP_UNHASHED);
5255	wait_event(close_wq, refcount_read(&s->st_stid.sc_count) == 2);
5256
5257	release_all_access(stp: s);
5258	if (s->st_stid.sc_file) {
5259	put_nfs4_file(fi: s->st_stid.sc_file);
5260	s->st_stid.sc_file = NULL;
5261	}
5262
5263	spin_lock(lock: &nn->client_lock);
5264	last = oo->oo_last_closed_stid;
5265	oo->oo_last_closed_stid = s;
5266	list_move_tail(list: &oo->oo_close_lru, head: &nn->close_lru);
5267	oo->oo_time = ktime_get_boottime_seconds();
5268	spin_unlock(lock: &nn->client_lock);
5269	if (last)
5270	nfs4_put_stid(s: &last->st_stid);
5271	}
5272
5273	static noinline_for_stack struct nfs4_file *
5274	nfsd4_file_hash_lookup(const struct svc_fh *fhp)
5275	{
5276	struct inode *inode = d_inode(dentry: fhp->fh_dentry);
5277	struct rhlist_head *tmp, *list;
5278	struct nfs4_file *fi;
5279
5280	rcu_read_lock();
5281	list = rhltable_lookup(hlt: &nfs4_file_rhltable, key: &inode,
5282	params: nfs4_file_rhash_params);
5283	rhl_for_each_entry_rcu(fi, tmp, list, fi_rlist) {
5284	if (fh_match(fh1: &fi->fi_fhandle, fh2: &fhp->fh_handle)) {
5285	if (refcount_inc_not_zero(r: &fi->fi_ref)) {
5286	rcu_read_unlock();
5287	return fi;
5288	}
5289	}
5290	}
5291	rcu_read_unlock();
5292	return NULL;
5293	}
5294
5295	/*
5296	* On hash insertion, identify entries with the same inode but
5297	* distinct filehandles. They will all be on the list returned
5298	* by rhltable_lookup().
5299	*
5300	* inode->i_lock prevents racing insertions from adding an entry
5301	* for the same inode/fhp pair twice.
5302	*/
5303	static noinline_for_stack struct nfs4_file *
5304	nfsd4_file_hash_insert(struct nfs4_file *new, const struct svc_fh *fhp)
5305	{
5306	struct inode *inode = d_inode(dentry: fhp->fh_dentry);
5307	struct rhlist_head *tmp, *list;
5308	struct nfs4_file *ret = NULL;
5309	bool alias_found = false;
5310	struct nfs4_file *fi;
5311	int err;
5312
5313	rcu_read_lock();
5314	spin_lock(lock: &inode->i_lock);
5315
5316	list = rhltable_lookup(hlt: &nfs4_file_rhltable, key: &inode,
5317	params: nfs4_file_rhash_params);
5318	rhl_for_each_entry_rcu(fi, tmp, list, fi_rlist) {
5319	if (fh_match(fh1: &fi->fi_fhandle, fh2: &fhp->fh_handle)) {
5320	if (refcount_inc_not_zero(r: &fi->fi_ref))
5321	ret = fi;
5322	} else
5323	fi->fi_aliased = alias_found = true;
5324	}
5325	if (ret)
5326	goto out_unlock;
5327
5328	nfsd4_file_init(fh: fhp, fp: new);
5329	err = rhltable_insert(hlt: &nfs4_file_rhltable, list: &new->fi_rlist,
5330	params: nfs4_file_rhash_params);
5331	if (err)
5332	goto out_unlock;
5333
5334	new->fi_aliased = alias_found;
5335	ret = new;
5336
5337	out_unlock:
5338	spin_unlock(lock: &inode->i_lock);
5339	rcu_read_unlock();
5340	return ret;
5341	}
5342
5343	static noinline_for_stack void nfsd4_file_hash_remove(struct nfs4_file *fi)
5344	{
5345	rhltable_remove(hlt: &nfs4_file_rhltable, list: &fi->fi_rlist,
5346	params: nfs4_file_rhash_params);
5347	}
5348
5349	/*
5350	* Called to check deny when READ with all zero stateid or
5351	* WRITE with all zero or all one stateid
5352	*/
5353	static __be32
5354	nfs4_share_conflict(struct svc_fh *current_fh, unsigned int deny_type)
5355	{
5356	struct nfs4_file *fp;
5357	__be32 ret = nfs_ok;
5358
5359	fp = nfsd4_file_hash_lookup(fhp: current_fh);
5360	if (!fp)
5361	return ret;
5362
5363	/* Check for conflicting share reservations */
5364	spin_lock(lock: &fp->fi_lock);
5365	if (fp->fi_share_deny & deny_type)
5366	ret = nfserr_locked;
5367	spin_unlock(lock: &fp->fi_lock);
5368	put_nfs4_file(fi: fp);
5369	return ret;
5370	}
5371
5372	static bool nfsd4_deleg_present(const struct inode *inode)
5373	{
5374	struct file_lock_context *ctx = locks_inode_context(inode);
5375
5376	return ctx && !list_empty_careful(head: &ctx->flc_lease);
5377	}
5378
5379	/**
5380	* nfsd_wait_for_delegreturn - wait for delegations to be returned
5381	* @rqstp: the RPC transaction being executed
5382	* @inode: in-core inode of the file being waited for
5383	*
5384	* The timeout prevents deadlock if all nfsd threads happen to be
5385	* tied up waiting for returning delegations.
5386	*
5387	* Return values:
5388	* %true: delegation was returned
5389	* %false: timed out waiting for delegreturn
5390	*/
5391	bool nfsd_wait_for_delegreturn(struct svc_rqst *rqstp, struct inode *inode)
5392	{
5393	long __maybe_unused timeo;
5394
5395	timeo = wait_var_event_timeout(inode, !nfsd4_deleg_present(inode),
5396	NFSD_DELEGRETURN_TIMEOUT);
5397	trace_nfsd_delegret_wakeup(rqstp, inode, timeo);
5398	return timeo > 0;
5399	}
5400
5401	static void nfsd4_cb_recall_prepare(struct nfsd4_callback *cb)
5402	{
5403	struct nfs4_delegation *dp = cb_to_delegation(cb);
5404	struct nfsd_net *nn = net_generic(net: dp->dl_stid.sc_client->net,
5405	id: nfsd_net_id);
5406
5407	block_delegations(fh: &dp->dl_stid.sc_file->fi_fhandle);
5408
5409	/*
5410	* We can't do this in nfsd_break_deleg_cb because it is
5411	* already holding inode->i_lock.
5412	*
5413	* If the dl_time != 0, then we know that it has already been
5414	* queued for a lease break. Don't queue it again.
5415	*/
5416	spin_lock(lock: &state_lock);
5417	if (delegation_hashed(dp) && dp->dl_time == 0) {
5418	dp->dl_time = ktime_get_boottime_seconds();
5419	list_add_tail(new: &dp->dl_recall_lru, head: &nn->del_recall_lru);
5420	}
5421	spin_unlock(lock: &state_lock);
5422	}
5423
5424	static int nfsd4_cb_recall_done(struct nfsd4_callback *cb,
5425	struct rpc_task *task)
5426	{
5427	struct nfs4_delegation *dp = cb_to_delegation(cb);
5428
5429	trace_nfsd_cb_recall_done(stp: &dp->dl_stid.sc_stateid, task);
5430
5431	if (dp->dl_stid.sc_status)
5432	/* CLOSED or REVOKED */
5433	return 1;
5434
5435	switch (task->tk_status) {
5436	case 0:
5437	return 1;
5438	case -NFS4ERR_DELAY:
5439	rpc_delay(task, 2 * HZ);
5440	return 0;
5441	case -EBADHANDLE:
5442	case -NFS4ERR_BAD_STATEID:
5443	/*
5444	* Race: client probably got cb_recall before open reply
5445	* granting delegation.
5446	*/
5447	if (dp->dl_retries--) {
5448	rpc_delay(task, 2 * HZ);
5449	return 0;
5450	}
5451	fallthrough;
5452	default:
5453	return 1;
5454	}
5455	}
5456
5457	static void nfsd4_cb_recall_release(struct nfsd4_callback *cb)
5458	{
5459	struct nfs4_delegation *dp = cb_to_delegation(cb);
5460
5461	nfs4_put_stid(s: &dp->dl_stid);
5462	}
5463
5464	static const struct nfsd4_callback_ops nfsd4_cb_recall_ops = {
5465	.prepare = nfsd4_cb_recall_prepare,
5466	.done = nfsd4_cb_recall_done,
5467	.release = nfsd4_cb_recall_release,
5468	.opcode = OP_CB_RECALL,
5469	};
5470
5471	static void nfsd_break_one_deleg(struct nfs4_delegation *dp)
5472	{
5473	bool queued;
5474
5475	if (test_and_set_bit(NFSD4_CALLBACK_RUNNING, addr: &dp->dl_recall.cb_flags))
5476	return;
5477
5478	/*
5479	* We're assuming the state code never drops its reference
5480	* without first removing the lease. Since we're in this lease
5481	* callback (and since the lease code is serialized by the
5482	* flc_lock) we know the server hasn't removed the lease yet, and
5483	* we know it's safe to take a reference.
5484	*/
5485	refcount_inc(r: &dp->dl_stid.sc_count);
5486	queued = nfsd4_run_cb(cb: &dp->dl_recall);
5487	WARN_ON_ONCE(!queued);
5488	if (!queued)
5489	refcount_dec(r: &dp->dl_stid.sc_count);
5490	}
5491
5492	/* Called from break_lease() with flc_lock held. */
5493	static bool
5494	nfsd_break_deleg_cb(struct file_lease *fl)
5495	{
5496	struct nfs4_delegation *dp = (struct nfs4_delegation *) fl->c.flc_owner;
5497	struct nfs4_file *fp = dp->dl_stid.sc_file;
5498	struct nfs4_client *clp = dp->dl_stid.sc_client;
5499	struct nfsd_net *nn;
5500
5501	trace_nfsd_cb_recall(stid: &dp->dl_stid);
5502
5503	dp->dl_recalled = true;
5504	atomic_inc(v: &clp->cl_delegs_in_recall);
5505	if (try_to_expire_client(clp)) {
5506	nn = net_generic(net: clp->net, id: nfsd_net_id);
5507	mod_delayed_work(wq: laundry_wq, dwork: &nn->laundromat_work, delay: 0);
5508	}
5509
5510	/*
5511	* We don't want the locks code to timeout the lease for us;
5512	* we'll remove it ourself if a delegation isn't returned
5513	* in time:
5514	*/
5515	fl->fl_break_time = 0;
5516
5517	fp->fi_had_conflict = true;
5518	nfsd_break_one_deleg(dp);
5519	return false;
5520	}
5521
5522	/**
5523	* nfsd_breaker_owns_lease - Check if lease conflict was resolved
5524	* @fl: Lock state to check
5525	*
5526	* Return values:
5527	* %true: Lease conflict was resolved
5528	* %false: Lease conflict was not resolved.
5529	*/
5530	static bool nfsd_breaker_owns_lease(struct file_lease *fl)
5531	{
5532	struct nfs4_delegation *dl = fl->c.flc_owner;
5533	struct svc_rqst *rqst;
5534	struct nfs4_client *clp;
5535
5536	rqst = nfsd_current_rqst();
5537	if (!nfsd_v4client(rq: rqst))
5538	return false;
5539	clp = *(rqst->rq_lease_breaker);
5540	return dl->dl_stid.sc_client == clp;
5541	}
5542
5543	static int
5544	nfsd_change_deleg_cb(struct file_lease *onlist, int arg,
5545	struct list_head *dispose)
5546	{
5547	struct nfs4_delegation *dp = (struct nfs4_delegation *) onlist->c.flc_owner;
5548	struct nfs4_client *clp = dp->dl_stid.sc_client;
5549
5550	if (arg & F_UNLCK) {
5551	if (dp->dl_recalled)
5552	atomic_dec(v: &clp->cl_delegs_in_recall);
5553	return lease_modify(onlist, arg, dispose);
5554	} else
5555	return -EAGAIN;
5556	}
5557
5558	/**
5559	* nfsd4_deleg_lm_open_conflict - see if the given file points to an inode that has
5560	* an existing open that would conflict with the
5561	* desired lease.
5562	* @filp: file to check
5563	* @arg: type of lease that we're trying to acquire
5564	*
5565	* The kernel will call into this operation to determine whether there
5566	* are conflicting opens that may prevent the deleg from being granted.
5567	* For nfsd, that check is done at a higher level, so this trivially
5568	* returns 0.
5569	*/
5570	static int
5571	nfsd4_deleg_lm_open_conflict(struct file *filp, int arg)
5572	{
5573	return 0;
5574	}
5575
5576	static const struct lease_manager_operations nfsd_lease_mng_ops = {
5577	.lm_breaker_owns_lease = nfsd_breaker_owns_lease,
5578	.lm_break = nfsd_break_deleg_cb,
5579	.lm_change = nfsd_change_deleg_cb,
5580	.lm_open_conflict = nfsd4_deleg_lm_open_conflict,
5581	};
5582
5583	static __be32 nfsd4_check_seqid(struct nfsd4_compound_state *cstate, struct nfs4_stateowner *so, u32 seqid)
5584	{
5585	if (nfsd4_has_session(cs: cstate))
5586	return nfs_ok;
5587	if (seqid == so->so_seqid - 1)
5588	return nfserr_replay_me;
5589	if (seqid == so->so_seqid)
5590	return nfs_ok;
5591	return nfserr_bad_seqid;
5592	}
5593
5594	static struct nfs4_client *lookup_clientid(clientid_t *clid, bool sessions,
5595	struct nfsd_net *nn)
5596	{
5597	struct nfs4_client *found;
5598
5599	spin_lock(lock: &nn->client_lock);
5600	found = find_confirmed_client(clid, sessions, nn);
5601	if (found)
5602	atomic_inc(v: &found->cl_rpc_users);
5603	spin_unlock(lock: &nn->client_lock);
5604	return found;
5605	}
5606
5607	static __be32 set_client(clientid_t *clid,
5608	struct nfsd4_compound_state *cstate,
5609	struct nfsd_net *nn)
5610	{
5611	if (cstate->clp) {
5612	if (!same_clid(cl1: &cstate->clp->cl_clientid, cl2: clid))
5613	return nfserr_stale_clientid;
5614	return nfs_ok;
5615	}
5616	if (STALE_CLIENTID(clid, nn))
5617	return nfserr_stale_clientid;
5618	/*
5619	* We're in the 4.0 case (otherwise the SEQUENCE op would have
5620	* set cstate->clp), so session = false:
5621	*/
5622	cstate->clp = lookup_clientid(clid, sessions: false, nn);
5623	if (!cstate->clp)
5624	return nfserr_expired;
5625	return nfs_ok;
5626	}
5627
5628	__be32
5629	nfsd4_process_open1(struct nfsd4_compound_state *cstate,
5630	struct nfsd4_open *open, struct nfsd_net *nn)
5631	{
5632	clientid_t *clientid = &open->op_clientid;
5633	struct nfs4_client *clp = NULL;
5634	unsigned int strhashval;
5635	struct nfs4_openowner *oo = NULL;
5636	__be32 status;
5637
5638	/*
5639	* In case we need it later, after we've already created the
5640	* file and don't want to risk a further failure:
5641	*/
5642	open->op_file = nfsd4_alloc_file();
5643	if (open->op_file == NULL)
5644	return nfserr_jukebox;
5645
5646	status = set_client(clid: clientid, cstate, nn);
5647	if (status)
5648	return status;
5649	clp = cstate->clp;
5650
5651	strhashval = ownerstr_hashval(ownername: &open->op_owner);
5652	retry:
5653	oo = find_or_alloc_open_stateowner(strhashval, open, cstate);
5654	open->op_openowner = oo;
5655	if (!oo)
5656	return nfserr_jukebox;
5657	if (nfsd4_cstate_assign_replay(cstate, so: &oo->oo_owner) == -EAGAIN) {
5658	nfs4_put_stateowner(sop: &oo->oo_owner);
5659	goto retry;
5660	}
5661	status = nfsd4_check_seqid(cstate, so: &oo->oo_owner, seqid: open->op_seqid);
5662	if (status)
5663	return status;
5664
5665	open->op_stp = nfs4_alloc_open_stateid(clp);
5666	if (!open->op_stp)
5667	return nfserr_jukebox;
5668
5669	if (nfsd4_has_session(cs: cstate) &&
5670	(cstate->current_fh.fh_export->ex_flags & NFSEXP_PNFS)) {
5671	open->op_odstate = alloc_clnt_odstate(clp);
5672	if (!open->op_odstate)
5673	return nfserr_jukebox;
5674	}
5675
5676	return nfs_ok;
5677	}
5678
5679	static inline __be32
5680	nfs4_check_delegmode(struct nfs4_delegation *dp, int flags)
5681	{
5682	if (!(flags & RD_STATE) && deleg_is_read(dl_type: dp->dl_type))
5683	return nfserr_openmode;
5684	else
5685	return nfs_ok;
5686	}
5687
5688	static int share_access_to_flags(u32 share_access)
5689	{
5690	return share_access == NFS4_SHARE_ACCESS_READ ? RD_STATE : WR_STATE;
5691	}
5692
5693	static struct nfs4_delegation *find_deleg_stateid(struct nfs4_client *cl,
5694	stateid_t *s)
5695	{
5696	struct nfs4_stid *ret;
5697
5698	ret = find_stateid_by_type(cl, t: s, SC_TYPE_DELEG, SC_STATUS_REVOKED);
5699	if (!ret)
5700	return NULL;
5701	return delegstateid(s: ret);
5702	}
5703
5704	static bool nfsd4_is_deleg_cur(struct nfsd4_open *open)
5705	{
5706	return open->op_claim_type == NFS4_OPEN_CLAIM_DELEGATE_CUR ||
5707	open->op_claim_type == NFS4_OPEN_CLAIM_DELEG_CUR_FH;
5708	}
5709
5710	static __be32
5711	nfs4_check_deleg(struct nfs4_client *cl, struct nfsd4_open *open,
5712	struct nfs4_delegation **dp)
5713	{
5714	int flags;
5715	__be32 status = nfserr_bad_stateid;
5716	struct nfs4_delegation *deleg;
5717
5718	deleg = find_deleg_stateid(cl, s: &open->op_delegate_stateid);
5719	if (deleg == NULL)
5720	goto out;
5721	if (deleg->dl_stid.sc_status & SC_STATUS_ADMIN_REVOKED) {
5722	nfs4_put_stid(s: &deleg->dl_stid);
5723	status = nfserr_admin_revoked;
5724	goto out;
5725	}
5726	if (deleg->dl_stid.sc_status & SC_STATUS_REVOKED) {
5727	nfs4_put_stid(s: &deleg->dl_stid);
5728	nfsd40_drop_revoked_stid(cl, stid: &open->op_delegate_stateid);
5729	status = nfserr_deleg_revoked;
5730	goto out;
5731	}
5732	flags = share_access_to_flags(share_access: open->op_share_access);
5733	status = nfs4_check_delegmode(dp: deleg, flags);
5734	if (status) {
5735	nfs4_put_stid(s: &deleg->dl_stid);
5736	goto out;
5737	}
5738	*dp = deleg;
5739	out:
5740	if (!nfsd4_is_deleg_cur(open))
5741	return nfs_ok;
5742	if (status)
5743	return status;
5744	open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
5745	return nfs_ok;
5746	}
5747
5748	static inline int nfs4_access_to_access(u32 nfs4_access)
5749	{
5750	int flags = 0;
5751
5752	if (nfs4_access & NFS4_SHARE_ACCESS_READ)
5753	flags |= NFSD_MAY_READ;
5754	if (nfs4_access & NFS4_SHARE_ACCESS_WRITE)
5755	flags |= NFSD_MAY_WRITE;
5756	return flags;
5757	}
5758
5759	static inline __be32
5760	nfsd4_truncate(struct svc_rqst *rqstp, struct svc_fh *fh,
5761	struct nfsd4_open *open)
5762	{
5763	struct iattr iattr = {
5764	.ia_valid = ATTR_SIZE,
5765	.ia_size = 0,
5766	};
5767	struct nfsd_attrs attrs = {
5768	.na_iattr = &iattr,
5769	};
5770	if (!open->op_truncate)
5771	return 0;
5772	if (!(open->op_share_access & NFS4_SHARE_ACCESS_WRITE))
5773	return nfserr_inval;
5774	return nfsd_setattr(rqstp, fh, &attrs, NULL);
5775	}
5776
5777	static __be32 nfs4_get_vfs_file(struct svc_rqst *rqstp, struct nfs4_file *fp,
5778	struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
5779	struct nfsd4_open *open, bool new_stp)
5780	{
5781	struct nfsd_file *nf = NULL;
5782	__be32 status;
5783	int oflag = nfs4_access_to_omode(access: open->op_share_access);
5784	int access = nfs4_access_to_access(nfs4_access: open->op_share_access);
5785	unsigned char old_access_bmap, old_deny_bmap;
5786
5787	spin_lock(lock: &fp->fi_lock);
5788
5789	/*
5790	* Are we trying to set a deny mode that would conflict with
5791	* current access?
5792	*/
5793	status = nfs4_file_check_deny(fp, deny: open->op_share_deny);
5794	if (status != nfs_ok) {
5795	if (status != nfserr_share_denied) {
5796	spin_unlock(lock: &fp->fi_lock);
5797	goto out;
5798	}
5799	if (nfs4_resolve_deny_conflicts_locked(fp, new_stp,
5800	stp, access: open->op_share_deny, share_access: false))
5801	status = nfserr_jukebox;
5802	spin_unlock(lock: &fp->fi_lock);
5803	goto out;
5804	}
5805
5806	/* set access to the file */
5807	status = nfs4_file_get_access(fp, access: open->op_share_access);
5808	if (status != nfs_ok) {
5809	if (status != nfserr_share_denied) {
5810	spin_unlock(lock: &fp->fi_lock);
5811	goto out;
5812	}
5813	if (nfs4_resolve_deny_conflicts_locked(fp, new_stp,
5814	stp, access: open->op_share_access, share_access: true))
5815	status = nfserr_jukebox;
5816	spin_unlock(lock: &fp->fi_lock);
5817	goto out;
5818	}
5819
5820	/* Set access bits in stateid */
5821	old_access_bmap = stp->st_access_bmap;
5822	set_access(access: open->op_share_access, stp);
5823
5824	/* Set new deny mask */
5825	old_deny_bmap = stp->st_deny_bmap;
5826	set_deny(deny: open->op_share_deny, stp);
5827	fp->fi_share_deny |= (open->op_share_deny & NFS4_SHARE_DENY_BOTH);
5828
5829	if (!fp->fi_fds[oflag]) {
5830	spin_unlock(lock: &fp->fi_lock);
5831
5832	status = nfsd_file_acquire_opened(rqstp, fhp: cur_fh, may_flags: access,
5833	file: open->op_filp, nfp: &nf);
5834	if (status != nfs_ok)
5835	goto out_put_access;
5836
5837	spin_lock(lock: &fp->fi_lock);
5838	if (!fp->fi_fds[oflag]) {
5839	fp->fi_fds[oflag] = nf;
5840	nf = NULL;
5841	}
5842	}
5843	spin_unlock(lock: &fp->fi_lock);
5844	if (nf)
5845	nfsd_file_put(nf);
5846
5847	status = nfserrno(errno: nfsd_open_break_lease(cur_fh->fh_dentry->d_inode,
5848	access));
5849	if (status)
5850	goto out_put_access;
5851
5852	status = nfsd4_truncate(rqstp, fh: cur_fh, open);
5853	if (status)
5854	goto out_put_access;
5855	out:
5856	return status;
5857	out_put_access:
5858	stp->st_access_bmap = old_access_bmap;
5859	nfs4_file_put_access(fp, access: open->op_share_access);
5860	reset_union_bmap_deny(deny: bmap_to_share_mode(bmap: old_deny_bmap), stp);
5861	goto out;
5862	}
5863
5864	static __be32
5865	nfs4_upgrade_open(struct svc_rqst *rqstp, struct nfs4_file *fp,
5866	struct svc_fh *cur_fh, struct nfs4_ol_stateid *stp,
5867	struct nfsd4_open *open)
5868	{
5869	__be32 status;
5870	unsigned char old_deny_bmap = stp->st_deny_bmap;
5871
5872	if (!test_access(access: open->op_share_access, stp))
5873	return nfs4_get_vfs_file(rqstp, fp, cur_fh, stp, open, new_stp: false);
5874
5875	/* test and set deny mode */
5876	spin_lock(lock: &fp->fi_lock);
5877	status = nfs4_file_check_deny(fp, deny: open->op_share_deny);
5878	switch (status) {
5879	case nfs_ok:
5880	set_deny(deny: open->op_share_deny, stp);
5881	fp->fi_share_deny |=
5882	(open->op_share_deny & NFS4_SHARE_DENY_BOTH);
5883	break;
5884	case nfserr_share_denied:
5885	if (nfs4_resolve_deny_conflicts_locked(fp, new_stp: false,
5886	stp, access: open->op_share_deny, share_access: false))
5887	status = nfserr_jukebox;
5888	break;
5889	}
5890	spin_unlock(lock: &fp->fi_lock);
5891
5892	if (status != nfs_ok)
5893	return status;
5894
5895	status = nfsd4_truncate(rqstp, fh: cur_fh, open);
5896	if (status != nfs_ok)
5897	reset_union_bmap_deny(deny: old_deny_bmap, stp);
5898	return status;
5899	}
5900
5901	/* Should we give out recallable state?: */
5902	static bool nfsd4_cb_channel_good(struct nfs4_client *clp)
5903	{
5904	if (clp->cl_cb_state == NFSD4_CB_UP)
5905	return true;
5906	/*
5907	* In the sessions case, since we don't have to establish a
5908	* separate connection for callbacks, we assume it's OK
5909	* until we hear otherwise:
5910	*/
5911	return clp->cl_minorversion && clp->cl_cb_state == NFSD4_CB_UNKNOWN;
5912	}
5913
5914	static struct file_lease *nfs4_alloc_init_lease(struct nfs4_delegation *dp)
5915	{
5916	struct file_lease *fl;
5917
5918	fl = locks_alloc_lease();
5919	if (!fl)
5920	return NULL;
5921	fl->fl_lmops = &nfsd_lease_mng_ops;
5922	fl->c.flc_flags = FL_DELEG;
5923	fl->c.flc_type = deleg_is_read(dl_type: dp->dl_type) ? F_RDLCK : F_WRLCK;
5924	fl->c.flc_owner = (fl_owner_t)dp;
5925	fl->c.flc_pid = current->tgid;
5926	fl->c.flc_file = dp->dl_stid.sc_file->fi_deleg_file->nf_file;
5927	return fl;
5928	}
5929
5930	static int nfsd4_check_conflicting_opens(struct nfs4_client *clp,
5931	struct nfs4_file *fp)
5932	{
5933	struct nfs4_ol_stateid *st;
5934	struct file *f = fp->fi_deleg_file->nf_file;
5935	struct inode *ino = file_inode(f);
5936	int writes;
5937
5938	writes = atomic_read(v: &ino->i_writecount);
5939	if (!writes)
5940	return 0;
5941	/*
5942	* There could be multiple filehandles (hence multiple
5943	* nfs4_files) referencing this file, but that's not too
5944	* common; let's just give up in that case rather than
5945	* trying to go look up all the clients using that other
5946	* nfs4_file as well:
5947	*/
5948	if (fp->fi_aliased)
5949	return -EAGAIN;
5950	/*
5951	* If there's a close in progress, make sure that we see it
5952	* clear any fi_fds[] entries before we see it decrement
5953	* i_writecount:
5954	*/
5955	smp_mb__after_atomic();
5956
5957	if (fp->fi_fds[O_WRONLY])
5958	writes--;
5959	if (fp->fi_fds[O_RDWR])
5960	writes--;
5961	if (writes > 0)
5962	return -EAGAIN; /* There may be non-NFSv4 writers */
5963	/*
5964	* It's possible there are non-NFSv4 write opens in progress,
5965	* but if they haven't incremented i_writecount yet then they
5966	* also haven't called break lease yet; so, they'll break this
5967	* lease soon enough. So, all that's left to check for is NFSv4
5968	* opens:
5969	*/
5970	spin_lock(lock: &fp->fi_lock);
5971	list_for_each_entry(st, &fp->fi_stateids, st_perfile) {
5972	if (st->st_openstp == NULL /* it's an open */ &&
5973	access_permit_write(stp: st) &&
5974	st->st_stid.sc_client != clp) {
5975	spin_unlock(lock: &fp->fi_lock);
5976	return -EAGAIN;
5977	}
5978	}
5979	spin_unlock(lock: &fp->fi_lock);
5980	/*
5981	* There's a small chance that we could be racing with another
5982	* NFSv4 open. However, any open that hasn't added itself to
5983	* the fi_stateids list also hasn't called break_lease yet; so,
5984	* they'll break this lease soon enough.
5985	*/
5986	return 0;
5987	}
5988
5989	/*
5990	* It's possible that between opening the dentry and setting the delegation,
5991	* that it has been renamed or unlinked. Redo the lookup to verify that this
5992	* hasn't happened.
5993	*/
5994	static int
5995	nfsd4_verify_deleg_dentry(struct nfsd4_open *open, struct nfs4_file *fp,
5996	struct svc_fh *parent)
5997	{
5998	struct svc_export *exp;
5999	struct dentry *child;
6000	__be32 err;
6001
6002	err = nfsd_lookup_dentry(open->op_rqstp, parent,
6003	open->op_fname, open->op_fnamelen,
6004	&exp, &child);
6005
6006	if (err)
6007	return -EAGAIN;
6008
6009	exp_put(exp);
6010	dput(child);
6011	if (child != file_dentry(file: fp->fi_deleg_file->nf_file))
6012	return -EAGAIN;
6013
6014	return 0;
6015	}
6016
6017	/*
6018	* We avoid breaking delegations held by a client due to its own activity, but
6019	* clearing setuid/setgid bits on a write is an implicit activity and the client
6020	* may not notice and continue using the old mode. Avoid giving out a delegation
6021	* on setuid/setgid files when the client is requesting an open for write.
6022	*/
6023	static int
6024	nfsd4_verify_setuid_write(struct nfsd4_open *open, struct nfsd_file *nf)
6025	{
6026	struct inode *inode = file_inode(f: nf->nf_file);
6027
6028	if ((open->op_share_access & NFS4_SHARE_ACCESS_WRITE) &&
6029	(inode->i_mode & (S_ISUID|S_ISGID)))
6030	return -EAGAIN;
6031	return 0;
6032	}
6033
6034	#ifdef CONFIG_NFSD_V4_DELEG_TIMESTAMPS
6035	static bool nfsd4_want_deleg_timestamps(const struct nfsd4_open *open)
6036	{
6037	return open->op_deleg_want & OPEN4_SHARE_ACCESS_WANT_DELEG_TIMESTAMPS;
6038	}
6039	#else /* CONFIG_NFSD_V4_DELEG_TIMESTAMPS */
6040	static bool nfsd4_want_deleg_timestamps(const struct nfsd4_open *open)
6041	{
6042	return false;
6043	}
6044	#endif /* CONFIG NFSD_V4_DELEG_TIMESTAMPS */
6045
6046	static struct nfs4_delegation *
6047	nfs4_set_delegation(struct nfsd4_open *open, struct nfs4_ol_stateid *stp,
6048	struct svc_fh *parent)
6049	{
6050	bool deleg_ts = nfsd4_want_deleg_timestamps(open);
6051	struct nfs4_client *clp = stp->st_stid.sc_client;
6052	struct nfs4_file *fp = stp->st_stid.sc_file;
6053	struct nfs4_clnt_odstate *odstate = stp->st_clnt_odstate;
6054	struct nfs4_delegation *dp;
6055	struct nfsd_file *nf = NULL;
6056	struct file_lease *fl;
6057	int status = 0;
6058	u32 dl_type;
6059
6060	/*
6061	* The fi_had_conflict and nfs_get_existing_delegation checks
6062	* here are just optimizations; we'll need to recheck them at
6063	* the end:
6064	*/
6065	if (fp->fi_had_conflict)
6066	return ERR_PTR(error: -EAGAIN);
6067
6068	/*
6069	* Try for a write delegation first. RFC8881 section 10.4 says:
6070	*
6071	* "An OPEN_DELEGATE_WRITE delegation allows the client to handle,
6072	* on its own, all opens."
6073	*
6074	* Furthermore, section 9.1.2 says:
6075	*
6076	* "In the case of READ, the server may perform the corresponding
6077	* check on the access mode, or it may choose to allow READ for
6078	* OPEN4_SHARE_ACCESS_WRITE, to accommodate clients whose WRITE
6079	* implementation may unavoidably do reads (e.g., due to buffer
6080	* cache constraints)."
6081	*
6082	* We choose to offer a write delegation for OPEN with the
6083	* OPEN4_SHARE_ACCESS_WRITE access mode to accommodate such clients.
6084	*/
6085	if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE) {
6086	nf = find_writeable_file(f: fp);
6087	dl_type = deleg_ts ? OPEN_DELEGATE_WRITE_ATTRS_DELEG : OPEN_DELEGATE_WRITE;
6088	}
6089
6090	/*
6091	* If the file is being opened O_RDONLY or we couldn't get a O_RDWR
6092	* file for some reason, then try for a read delegation instead.
6093	*/
6094	if (!nf && (open->op_share_access & NFS4_SHARE_ACCESS_READ)) {
6095	nf = find_readable_file(f: fp);
6096	dl_type = deleg_ts ? OPEN_DELEGATE_READ_ATTRS_DELEG : OPEN_DELEGATE_READ;
6097	}
6098
6099	if (!nf)
6100	return ERR_PTR(error: -EAGAIN);
6101
6102	/*
6103	* File delegations and associated locks cannot be recovered if the
6104	* export is from an NFS proxy server.
6105	*/
6106	if (exportfs_cannot_lock(export_ops: nf->nf_file->f_path.mnt->mnt_sb->s_export_op)) {
6107	nfsd_file_put(nf);
6108	return ERR_PTR(error: -EOPNOTSUPP);
6109	}
6110
6111	spin_lock(lock: &state_lock);
6112	spin_lock(lock: &fp->fi_lock);
6113	if (nfs4_delegation_exists(clp, fp))
6114	status = -EAGAIN;
6115	else if (nfsd4_verify_setuid_write(open, nf))
6116	status = -EAGAIN;
6117	else if (!fp->fi_deleg_file) {
6118	fp->fi_deleg_file = nf;
6119	/* increment early to prevent fi_deleg_file from being
6120	* cleared */
6121	fp->fi_delegees = 1;
6122	nf = NULL;
6123	} else
6124	fp->fi_delegees++;
6125	spin_unlock(lock: &fp->fi_lock);
6126	spin_unlock(lock: &state_lock);
6127	if (nf)
6128	nfsd_file_put(nf);
6129	if (status)
6130	return ERR_PTR(error: status);
6131
6132	status = -ENOMEM;
6133	dp = alloc_init_deleg(clp, fp, odstate, dl_type);
6134	if (!dp)
6135	goto out_delegees;
6136
6137	fl = nfs4_alloc_init_lease(dp);
6138	if (!fl)
6139	goto out_clnt_odstate;
6140
6141	status = kernel_setlease(fp->fi_deleg_file->nf_file,
6142	fl->c.flc_type, &fl, NULL);
6143	if (fl)
6144	locks_free_lease(fl);
6145	if (status)
6146	goto out_clnt_odstate;
6147
6148	if (parent) {
6149	status = nfsd4_verify_deleg_dentry(open, fp, parent);
6150	if (status)
6151	goto out_unlock;
6152	}
6153
6154	status = nfsd4_check_conflicting_opens(clp, fp);
6155	if (status)
6156	goto out_unlock;
6157
6158	/*
6159	* Now that the deleg is set, check again to ensure that nothing
6160	* raced in and changed the mode while we weren't looking.
6161	*/
6162	status = nfsd4_verify_setuid_write(open, nf: fp->fi_deleg_file);
6163	if (status)
6164	goto out_unlock;
6165
6166	status = -EAGAIN;
6167	if (fp->fi_had_conflict)
6168	goto out_unlock;
6169
6170	spin_lock(lock: &state_lock);
6171	spin_lock(lock: &clp->cl_lock);
6172	spin_lock(lock: &fp->fi_lock);
6173	status = hash_delegation_locked(dp, fp);
6174	spin_unlock(lock: &fp->fi_lock);
6175	spin_unlock(lock: &clp->cl_lock);
6176	spin_unlock(lock: &state_lock);
6177
6178	if (status)
6179	goto out_unlock;
6180
6181	return dp;
6182	out_unlock:
6183	kernel_setlease(fp->fi_deleg_file->nf_file, F_UNLCK, NULL, (void **)&dp);
6184	out_clnt_odstate:
6185	put_clnt_odstate(co: dp->dl_clnt_odstate);
6186	nfs4_put_stid(s: &dp->dl_stid);
6187	out_delegees:
6188	put_deleg_file(fp);
6189	return ERR_PTR(error: status);
6190	}
6191
6192	static void nfsd4_open_deleg_none_ext(struct nfsd4_open *open, int status)
6193	{
6194	open->op_delegate_type = OPEN_DELEGATE_NONE_EXT;
6195	if (status == -EAGAIN)
6196	open->op_why_no_deleg = WND4_CONTENTION;
6197	else {
6198	open->op_why_no_deleg = WND4_RESOURCE;
6199	switch (open->op_deleg_want) {
6200	case OPEN4_SHARE_ACCESS_WANT_READ_DELEG:
6201	case OPEN4_SHARE_ACCESS_WANT_WRITE_DELEG:
6202	case OPEN4_SHARE_ACCESS_WANT_ANY_DELEG:
6203	break;
6204	case OPEN4_SHARE_ACCESS_WANT_CANCEL:
6205	open->op_why_no_deleg = WND4_CANCELLED;
6206	break;
6207	case OPEN4_SHARE_ACCESS_WANT_NO_DELEG:
6208	WARN_ON_ONCE(1);
6209	}
6210	}
6211	}
6212
6213	static bool
6214	nfs4_delegation_stat(struct nfs4_delegation *dp, struct svc_fh *currentfh,
6215	struct kstat *stat)
6216	{
6217	struct nfsd_file *nf = find_writeable_file(f: dp->dl_stid.sc_file);
6218	struct path path;
6219	int rc;
6220
6221	if (!nf)
6222	return false;
6223
6224	path.mnt = currentfh->fh_export->ex_path.mnt;
6225	path.dentry = file_dentry(file: nf->nf_file);
6226
6227	rc = vfs_getattr(&path, stat,
6228	STATX_MODE | STATX_SIZE | STATX_ATIME |
6229	STATX_MTIME | STATX_CTIME | STATX_CHANGE_COOKIE,
6230	AT_STATX_SYNC_AS_STAT);
6231
6232	nfsd_file_put(nf);
6233	return rc == 0;
6234	}
6235
6236	/*
6237	* Add NFS4_SHARE_ACCESS_READ to the write delegation granted on OPEN
6238	* with NFS4_SHARE_ACCESS_WRITE by allocating separate nfsd_file and
6239	* struct file to be used for read with delegation stateid.
6240	*
6241	*/
6242	static bool
6243	nfsd4_add_rdaccess_to_wrdeleg(struct svc_rqst *rqstp, struct nfsd4_open *open,
6244	struct svc_fh *fh, struct nfs4_ol_stateid *stp)
6245	{
6246	struct nfs4_file *fp;
6247	struct nfsd_file *nf = NULL;
6248
6249	if ((open->op_share_access & NFS4_SHARE_ACCESS_BOTH) ==
6250	NFS4_SHARE_ACCESS_WRITE) {
6251	if (nfsd_file_acquire_opened(rqstp, fhp: fh, NFSD_MAY_READ, NULL, nfp: &nf))
6252	return (false);
6253	fp = stp->st_stid.sc_file;
6254	spin_lock(lock: &fp->fi_lock);
6255	__nfs4_file_get_access(fp, NFS4_SHARE_ACCESS_READ);
6256	if (!fp->fi_fds[O_RDONLY]) {
6257	fp->fi_fds[O_RDONLY] = nf;
6258	nf = NULL;
6259	}
6260	fp->fi_rdeleg_file = nfsd_file_get(nf: fp->fi_fds[O_RDONLY]);
6261	spin_unlock(lock: &fp->fi_lock);
6262	if (nf)
6263	nfsd_file_put(nf);
6264	}
6265	return true;
6266	}
6267
6268	/*
6269	* The Linux NFS server does not offer write delegations to NFSv4.0
6270	* clients in order to avoid conflicts between write delegations and
6271	* GETATTRs requesting CHANGE or SIZE attributes.
6272	*
6273	* With NFSv4.1 and later minorversions, the SEQUENCE operation that
6274	* begins each COMPOUND contains a client ID. Delegation recall can
6275	* be avoided when the server recognizes the client sending a
6276	* GETATTR also holds write delegation it conflicts with.
6277	*
6278	* However, the NFSv4.0 protocol does not enable a server to
6279	* determine that a GETATTR originated from the client holding the
6280	* conflicting delegation versus coming from some other client. Per
6281	* RFC 7530 Section 16.7.5, the server must recall or send a
6282	* CB_GETATTR even when the GETATTR originates from the client that
6283	* holds the conflicting delegation.
6284	*
6285	* An NFSv4.0 client can trigger a pathological situation if it
6286	* always sends a DELEGRETURN preceded by a conflicting GETATTR in
6287	* the same COMPOUND. COMPOUND execution will always stop at the
6288	* GETATTR and the DELEGRETURN will never get executed. The server
6289	* eventually revokes the delegation, which can result in loss of
6290	* open or lock state.
6291	*/
6292	static void
6293	nfs4_open_delegation(struct svc_rqst *rqstp, struct nfsd4_open *open,
6294	struct nfs4_ol_stateid *stp, struct svc_fh *currentfh,
6295	struct svc_fh *fh)
6296	{
6297	struct nfs4_openowner *oo = openowner(so: stp->st_stateowner);
6298	bool deleg_ts = nfsd4_want_deleg_timestamps(open);
6299	struct nfs4_client *clp = stp->st_stid.sc_client;
6300	struct svc_fh *parent = NULL;
6301	struct nfs4_delegation *dp;
6302	struct kstat stat;
6303	int status = 0;
6304	int cb_up;
6305
6306	cb_up = nfsd4_cb_channel_good(clp: oo->oo_owner.so_client);
6307	open->op_recall = false;
6308	switch (open->op_claim_type) {
6309	case NFS4_OPEN_CLAIM_PREVIOUS:
6310	if (!cb_up)
6311	open->op_recall = true;
6312	break;
6313	case NFS4_OPEN_CLAIM_NULL:
6314	parent = currentfh;
6315	fallthrough;
6316	case NFS4_OPEN_CLAIM_FH:
6317	/*
6318	* Let's not give out any delegations till everyone's
6319	* had the chance to reclaim theirs, *and* until
6320	* NLM locks have all been reclaimed:
6321	*/
6322	if (locks_in_grace(clp->net))
6323	goto out_no_deleg;
6324	if (!cb_up || !(oo->oo_flags & NFS4_OO_CONFIRMED))
6325	goto out_no_deleg;
6326	if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE &&
6327	!clp->cl_minorversion)
6328	goto out_no_deleg;
6329	break;
6330	default:
6331	goto out_no_deleg;
6332	}
6333	dp = nfs4_set_delegation(open, stp, parent);
6334	if (IS_ERR(ptr: dp))
6335	goto out_no_deleg;
6336
6337	memcpy(&open->op_delegate_stateid, &dp->dl_stid.sc_stateid, sizeof(dp->dl_stid.sc_stateid));
6338
6339	if (open->op_share_access & NFS4_SHARE_ACCESS_WRITE) {
6340	struct file *f = dp->dl_stid.sc_file->fi_deleg_file->nf_file;
6341
6342	if (!nfsd4_add_rdaccess_to_wrdeleg(rqstp, open, fh, stp) ||
6343	!nfs4_delegation_stat(dp, currentfh, stat: &stat)) {
6344	nfs4_put_stid(s: &dp->dl_stid);
6345	destroy_delegation(dp);
6346	goto out_no_deleg;
6347	}
6348	open->op_delegate_type = deleg_ts ? OPEN_DELEGATE_WRITE_ATTRS_DELEG :
6349	OPEN_DELEGATE_WRITE;
6350	dp->dl_cb_fattr.ncf_cur_fsize = stat.size;
6351	dp->dl_cb_fattr.ncf_initial_cinfo = nfsd4_change_attribute(stat: &stat);
6352	dp->dl_atime = stat.atime;
6353	dp->dl_ctime = stat.ctime;
6354	dp->dl_mtime = stat.mtime;
6355	spin_lock(lock: &f->f_lock);
6356	f->f_mode |= FMODE_NOCMTIME;
6357	spin_unlock(lock: &f->f_lock);
6358	trace_nfsd_deleg_write(stp: &dp->dl_stid.sc_stateid);
6359	} else {
6360	open->op_delegate_type = deleg_ts && nfs4_delegation_stat(dp, currentfh, stat: &stat) ?
6361	OPEN_DELEGATE_READ_ATTRS_DELEG : OPEN_DELEGATE_READ;
6362	dp->dl_atime = stat.atime;
6363	trace_nfsd_deleg_read(stp: &dp->dl_stid.sc_stateid);
6364	}
6365	nfs4_put_stid(s: &dp->dl_stid);
6366	return;
6367	out_no_deleg:
6368	open->op_delegate_type = OPEN_DELEGATE_NONE;
6369
6370	/* 4.1 client asking for a delegation? */
6371	if (open->op_deleg_want)
6372	nfsd4_open_deleg_none_ext(open, status);
6373	return;
6374	}
6375
6376	static void nfsd4_deleg_xgrade_none_ext(struct nfsd4_open *open,
6377	struct nfs4_delegation *dp)
6378	{
6379	if (deleg_is_write(dl_type: dp->dl_type)) {
6380	if (open->op_deleg_want & OPEN4_SHARE_ACCESS_WANT_READ_DELEG) {
6381	open->op_delegate_type = OPEN_DELEGATE_NONE_EXT;
6382	open->op_why_no_deleg = WND4_NOT_SUPP_DOWNGRADE;
6383	} else if (open->op_deleg_want & OPEN4_SHARE_ACCESS_WANT_WRITE_DELEG) {
6384	open->op_delegate_type = OPEN_DELEGATE_NONE_EXT;
6385	open->op_why_no_deleg = WND4_NOT_SUPP_UPGRADE;
6386	}
6387	}
6388	/* Otherwise the client must be confused wanting a delegation
6389	* it already has, therefore we don't return
6390	* OPEN_DELEGATE_NONE_EXT and reason.
6391	*/
6392	}
6393
6394	/* Are we returning only a delegation stateid? */
6395	static bool open_xor_delegation(struct nfsd4_open *open)
6396	{
6397	if (!(open->op_deleg_want & OPEN4_SHARE_ACCESS_WANT_OPEN_XOR_DELEGATION))
6398	return false;
6399	/* Did we actually get a delegation? */
6400	if (!deleg_is_read(dl_type: open->op_delegate_type) && !deleg_is_write(dl_type: open->op_delegate_type))
6401	return false;
6402	return true;
6403	}
6404
6405	/**
6406	* nfsd4_process_open2 - finish open processing
6407	* @rqstp: the RPC transaction being executed
6408	* @current_fh: NFSv4 COMPOUND's current filehandle
6409	* @open: OPEN arguments
6410	*
6411	* If successful, (1) truncate the file if open->op_truncate was
6412	* set, (2) set open->op_stateid, (3) set open->op_delegation.
6413	*
6414	* Returns %nfs_ok on success; otherwise an nfs4stat value in
6415	* network byte order is returned.
6416	*/
6417	__be32
6418	nfsd4_process_open2(struct svc_rqst *rqstp, struct svc_fh *current_fh, struct nfsd4_open *open)
6419	{
6420	struct nfsd4_compoundres *resp = rqstp->rq_resp;
6421	struct nfs4_client *cl = open->op_openowner->oo_owner.so_client;
6422	struct nfs4_file *fp = NULL;
6423	struct nfs4_ol_stateid *stp = NULL;
6424	struct nfs4_delegation *dp = NULL;
6425	__be32 status;
6426	bool new_stp = false;
6427
6428	/*
6429	* Lookup file; if found, lookup stateid and check open request,
6430	* and check for delegations in the process of being recalled.
6431	* If not found, create the nfs4_file struct
6432	*/
6433	fp = nfsd4_file_hash_insert(new: open->op_file, fhp: current_fh);
6434	if (unlikely(!fp))
6435	return nfserr_jukebox;
6436	if (fp != open->op_file) {
6437	status = nfs4_check_deleg(cl, open, dp: &dp);
6438	if (status)
6439	goto out;
6440	if (dp && nfsd4_is_deleg_cur(open) &&
6441	(dp->dl_stid.sc_file != fp)) {
6442	/*
6443	* RFC8881 section 8.2.4 mandates the server to return
6444	* NFS4ERR_BAD_STATEID if the selected table entry does
6445	* not match the current filehandle. However returning
6446	* NFS4ERR_BAD_STATEID in the OPEN can cause the client
6447	* to repeatedly retry the operation with the same
6448	* stateid, since the stateid itself is valid. To avoid
6449	* this situation NFSD returns NFS4ERR_INVAL instead.
6450	*/
6451	status = nfserr_inval;
6452	goto out;
6453	}
6454	stp = nfsd4_find_and_lock_existing_open(fp, open);
6455	} else {
6456	open->op_file = NULL;
6457	status = nfserr_bad_stateid;
6458	if (nfsd4_is_deleg_cur(open))
6459	goto out;
6460	}
6461
6462	if (!stp) {
6463	stp = init_open_stateid(fp, open);
6464	if (!stp) {
6465	status = nfserr_jukebox;
6466	goto out;
6467	}
6468
6469	if (!open->op_stp)
6470	new_stp = true;
6471	}
6472
6473	/*
6474	* OPEN the file, or upgrade an existing OPEN.
6475	* If truncate fails, the OPEN fails.
6476	*
6477	* stp is already locked.
6478	*/
6479	if (!new_stp) {
6480	/* Stateid was found, this is an OPEN upgrade */
6481	status = nfs4_upgrade_open(rqstp, fp, cur_fh: current_fh, stp, open);
6482	if (status) {
6483	mutex_unlock(lock: &stp->st_mutex);
6484	goto out;
6485	}
6486	} else {
6487	status = nfs4_get_vfs_file(rqstp, fp, cur_fh: current_fh, stp, open, new_stp: true);
6488	if (status) {
6489	release_open_stateid(stp);
6490	mutex_unlock(lock: &stp->st_mutex);
6491	goto out;
6492	}
6493
6494	stp->st_clnt_odstate = find_or_hash_clnt_odstate(fp,
6495	new: open->op_odstate);
6496	if (stp->st_clnt_odstate == open->op_odstate)
6497	open->op_odstate = NULL;
6498	}
6499
6500	nfs4_inc_and_copy_stateid(dst: &open->op_stateid, stid: &stp->st_stid);
6501	mutex_unlock(lock: &stp->st_mutex);
6502
6503	if (nfsd4_has_session(cs: &resp->cstate)) {
6504	if (open->op_deleg_want & OPEN4_SHARE_ACCESS_WANT_NO_DELEG) {
6505	open->op_delegate_type = OPEN_DELEGATE_NONE_EXT;
6506	open->op_why_no_deleg = WND4_NOT_WANTED;
6507	goto nodeleg;
6508	}
6509	}
6510
6511	/*
6512	* Attempt to hand out a delegation. No error return, because the
6513	* OPEN succeeds even if we fail.
6514	*/
6515	nfs4_open_delegation(rqstp, open, stp,
6516	currentfh: &resp->cstate.current_fh, fh: current_fh);
6517
6518	/*
6519	* If there is an existing open stateid, it must be updated and
6520	* returned. Only respect WANT_OPEN_XOR_DELEGATION when a new
6521	* open stateid would have to be created.
6522	*/
6523	if (new_stp && open_xor_delegation(open)) {
6524	memcpy(&open->op_stateid, &zero_stateid, sizeof(open->op_stateid));
6525	open->op_rflags |= OPEN4_RESULT_NO_OPEN_STATEID;
6526	release_open_stateid(stp);
6527	}
6528	nodeleg:
6529	status = nfs_ok;
6530	trace_nfsd_open(stp: &stp->st_stid.sc_stateid);
6531	out:
6532	/* 4.1 client trying to upgrade/downgrade delegation? */
6533	if (open->op_delegate_type == OPEN_DELEGATE_NONE && dp &&
6534	open->op_deleg_want)
6535	nfsd4_deleg_xgrade_none_ext(open, dp);
6536
6537	if (fp)
6538	put_nfs4_file(fi: fp);
6539	if (status == 0 && open->op_claim_type == NFS4_OPEN_CLAIM_PREVIOUS)
6540	open->op_openowner->oo_flags |= NFS4_OO_CONFIRMED;
6541	/*
6542	* To finish the open response, we just need to set the rflags.
6543	*/
6544	open->op_rflags |= NFS4_OPEN_RESULT_LOCKTYPE_POSIX;
6545	if (nfsd4_has_session(cs: &resp->cstate))
6546	open->op_rflags |= NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK;
6547	else if (!(open->op_openowner->oo_flags & NFS4_OO_CONFIRMED))
6548	open->op_rflags |= NFS4_OPEN_RESULT_CONFIRM;
6549
6550	if (dp)
6551	nfs4_put_stid(s: &dp->dl_stid);
6552	if (stp)
6553	nfs4_put_stid(s: &stp->st_stid);
6554
6555	return status;
6556	}
6557
6558	void nfsd4_cleanup_open_state(struct nfsd4_compound_state *cstate,
6559	struct nfsd4_open *open)
6560	{
6561	if (open->op_openowner)
6562	nfs4_put_stateowner(sop: &open->op_openowner->oo_owner);
6563	if (open->op_file)
6564	kmem_cache_free(s: file_slab, objp: open->op_file);
6565	if (open->op_stp)
6566	nfs4_put_stid(s: &open->op_stp->st_stid);
6567	if (open->op_odstate)
6568	kmem_cache_free(s: odstate_slab, objp: open->op_odstate);
6569	}
6570
6571	__be32
6572	nfsd4_renew(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
6573	union nfsd4_op_u *u)
6574	{
6575	clientid_t *clid = &u->renew;
6576	struct nfs4_client *clp;
6577	__be32 status;
6578	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
6579
6580	trace_nfsd_clid_renew(clid);
6581	status = set_client(clid, cstate, nn);
6582	if (status)
6583	return status;
6584	clp = cstate->clp;
6585	if (!list_empty(head: &clp->cl_delegations)
6586	&& clp->cl_cb_state != NFSD4_CB_UP)
6587	return nfserr_cb_path_down;
6588	return nfs_ok;
6589	}
6590
6591	static void
6592	nfsd4_end_grace(struct nfsd_net *nn)
6593	{
6594	/* do nothing if grace period already ended */
6595	if (nn->grace_ended)
6596	return;
6597
6598	trace_nfsd_grace_complete(nn);
6599	nn->grace_ended = true;
6600	/*
6601	* If the server goes down again right now, an NFSv4
6602	* client will still be allowed to reclaim after it comes back up,
6603	* even if it hasn't yet had a chance to reclaim state this time.
6604	*
6605	*/
6606	nfsd4_record_grace_done(nn);
6607	/*
6608	* At this point, NFSv4 clients can still reclaim. But if the
6609	* server crashes, any that have not yet reclaimed will be out
6610	* of luck on the next boot.
6611	*
6612	* (NFSv4.1+ clients are considered to have reclaimed once they
6613	* call RECLAIM_COMPLETE. NFSv4.0 clients are considered to
6614	* have reclaimed after their first OPEN.)
6615	*/
6616	locks_end_grace(&nn->nfsd4_manager);
6617	/*
6618	* At this point, and once lockd and/or any other containers
6619	* exit their grace period, further reclaims will fail and
6620	* regular locking can resume.
6621	*/
6622	}
6623
6624	/**
6625	* nfsd4_force_end_grace - forcibly end the NFSv4 grace period
6626	* @nn: network namespace for the server instance to be updated
6627	*
6628	* Forces bypass of normal grace period completion, then schedules
6629	* the laundromat to end the grace period immediately. Does not wait
6630	* for the grace period to fully terminate before returning.
6631	*
6632	* Return values:
6633	* %true: Grace termination schedule
6634	* %false: No action was taken
6635	*/
6636	bool nfsd4_force_end_grace(struct nfsd_net *nn)
6637	{
6638	if (!nn->client_tracking_ops)
6639	return false;
6640	spin_lock(lock: &nn->client_lock);
6641	if (nn->grace_ended || !nn->client_tracking_active) {
6642	spin_unlock(lock: &nn->client_lock);
6643	return false;
6644	}
6645	WRITE_ONCE(nn->grace_end_forced, true);
6646	mod_delayed_work(wq: laundry_wq, dwork: &nn->laundromat_work, delay: 0);
6647	spin_unlock(lock: &nn->client_lock);
6648	return true;
6649	}
6650
6651	/*
6652	* If we've waited a lease period but there are still clients trying to
6653	* reclaim, wait a little longer to give them a chance to finish.
6654	*/
6655	static bool clients_still_reclaiming(struct nfsd_net *nn)
6656	{
6657	time64_t double_grace_period_end = nn->boot_time +
6658	2 * nn->nfsd4_lease;
6659
6660	if (READ_ONCE(nn->grace_end_forced))
6661	return false;
6662	if (nn->track_reclaim_completes &&
6663	atomic_read(v: &nn->nr_reclaim_complete) ==
6664	nn->reclaim_str_hashtbl_size)
6665	return false;
6666	if (!nn->somebody_reclaimed)
6667	return false;
6668	nn->somebody_reclaimed = false;
6669	/*
6670	* If we've given them *two* lease times to reclaim, and they're
6671	* still not done, give up:
6672	*/
6673	if (ktime_get_boottime_seconds() > double_grace_period_end)
6674	return false;
6675	return true;
6676	}
6677
6678	struct laundry_time {
6679	time64_t cutoff;
6680	time64_t new_timeo;
6681	};
6682
6683	static bool state_expired(struct laundry_time *lt, time64_t last_refresh)
6684	{
6685	time64_t time_remaining;
6686
6687	if (last_refresh < lt->cutoff)
6688	return true;
6689	time_remaining = last_refresh - lt->cutoff;
6690	lt->new_timeo = min(lt->new_timeo, time_remaining);
6691	return false;
6692	}
6693
6694	#ifdef CONFIG_NFSD_V4_2_INTER_SSC
6695	void nfsd4_ssc_init_umount_work(struct nfsd_net *nn)
6696	{
6697	spin_lock_init(&nn->nfsd_ssc_lock);
6698	INIT_LIST_HEAD(list: &nn->nfsd_ssc_mount_list);
6699	init_waitqueue_head(&nn->nfsd_ssc_waitq);
6700	}
6701
6702	/*
6703	* This is called when nfsd is being shutdown, after all inter_ssc
6704	* cleanup were done, to destroy the ssc delayed unmount list.
6705	*/
6706	static void nfsd4_ssc_shutdown_umount(struct nfsd_net *nn)
6707	{
6708	struct nfsd4_ssc_umount_item *ni = NULL;
6709	struct nfsd4_ssc_umount_item *tmp;
6710
6711	spin_lock(lock: &nn->nfsd_ssc_lock);
6712	list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
6713	list_del(entry: &ni->nsui_list);
6714	spin_unlock(lock: &nn->nfsd_ssc_lock);
6715	mntput(mnt: ni->nsui_vfsmount);
6716	kfree(objp: ni);
6717	spin_lock(lock: &nn->nfsd_ssc_lock);
6718	}
6719	spin_unlock(lock: &nn->nfsd_ssc_lock);
6720	}
6721
6722	static void nfsd4_ssc_expire_umount(struct nfsd_net *nn)
6723	{
6724	bool do_wakeup = false;
6725	struct nfsd4_ssc_umount_item *ni = NULL;
6726	struct nfsd4_ssc_umount_item *tmp;
6727
6728	spin_lock(lock: &nn->nfsd_ssc_lock);
6729	list_for_each_entry_safe(ni, tmp, &nn->nfsd_ssc_mount_list, nsui_list) {
6730	if (time_after(jiffies, ni->nsui_expire)) {
6731	if (refcount_read(r: &ni->nsui_refcnt) > 1)
6732	continue;
6733
6734	/* mark being unmount */
6735	ni->nsui_busy = true;
6736	spin_unlock(lock: &nn->nfsd_ssc_lock);
6737	mntput(mnt: ni->nsui_vfsmount);
6738	spin_lock(lock: &nn->nfsd_ssc_lock);
6739
6740	/* waiters need to start from begin of list */
6741	list_del(entry: &ni->nsui_list);
6742	kfree(objp: ni);
6743
6744	/* wakeup ssc_connect waiters */
6745	do_wakeup = true;
6746	continue;
6747	}
6748	break;
6749	}
6750	if (do_wakeup)
6751	wake_up_all(&nn->nfsd_ssc_waitq);
6752	spin_unlock(lock: &nn->nfsd_ssc_lock);
6753	}
6754	#endif
6755
6756	/* Check if any lock belonging to this lockowner has any blockers */
6757	static bool
6758	nfs4_lockowner_has_blockers(struct nfs4_lockowner *lo)
6759	{
6760	struct file_lock_context *ctx;
6761	struct nfs4_ol_stateid *stp;
6762	struct nfs4_file *nf;
6763
6764	list_for_each_entry(stp, &lo->lo_owner.so_stateids, st_perstateowner) {
6765	nf = stp->st_stid.sc_file;
6766	ctx = locks_inode_context(inode: nf->fi_inode);
6767	if (!ctx)
6768	continue;
6769	if (locks_owner_has_blockers(flctx: ctx, owner: lo))
6770	return true;
6771	}
6772	return false;
6773	}
6774
6775	static bool
6776	nfs4_anylock_blockers(struct nfs4_client *clp)
6777	{
6778	int i;
6779	struct nfs4_stateowner *so;
6780	struct nfs4_lockowner *lo;
6781
6782	if (atomic_read(v: &clp->cl_delegs_in_recall))
6783	return true;
6784	spin_lock(lock: &clp->cl_lock);
6785	for (i = 0; i < OWNER_HASH_SIZE; i++) {
6786	list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[i],
6787	so_strhash) {
6788	if (so->so_is_open_owner)
6789	continue;
6790	lo = lockowner(so);
6791	if (nfs4_lockowner_has_blockers(lo)) {
6792	spin_unlock(lock: &clp->cl_lock);
6793	return true;
6794	}
6795	}
6796	}
6797	spin_unlock(lock: &clp->cl_lock);
6798	return false;
6799	}
6800
6801	static void
6802	nfs4_get_client_reaplist(struct nfsd_net *nn, struct list_head *reaplist,
6803	struct laundry_time *lt)
6804	{
6805	unsigned int maxreap, reapcnt = 0;
6806	struct list_head *pos, *next;
6807	struct nfs4_client *clp;
6808
6809	maxreap = (atomic_read(v: &nn->nfs4_client_count) >= nn->nfs4_max_clients) ?
6810	NFSD_CLIENT_MAX_TRIM_PER_RUN : 0;
6811	INIT_LIST_HEAD(list: reaplist);
6812	spin_lock(lock: &nn->client_lock);
6813	list_for_each_safe(pos, next, &nn->client_lru) {
6814	clp = list_entry(pos, struct nfs4_client, cl_lru);
6815	if (clp->cl_state == NFSD4_EXPIRABLE)
6816	goto exp_client;
6817	if (!state_expired(lt, last_refresh: clp->cl_time))
6818	break;
6819	if (!atomic_read(v: &clp->cl_rpc_users)) {
6820	if (clp->cl_state == NFSD4_ACTIVE)
6821	atomic_inc(v: &nn->nfsd_courtesy_clients);
6822	clp->cl_state = NFSD4_COURTESY;
6823	}
6824	if (!client_has_state(clp))
6825	goto exp_client;
6826	if (!nfs4_anylock_blockers(clp))
6827	if (reapcnt >= maxreap)
6828	continue;
6829	exp_client:
6830	if (!mark_client_expired_locked(clp)) {
6831	list_add(new: &clp->cl_lru, head: reaplist);
6832	reapcnt++;
6833	}
6834	}
6835	spin_unlock(lock: &nn->client_lock);
6836	}
6837
6838	static void
6839	nfs4_get_courtesy_client_reaplist(struct nfsd_net *nn,
6840	struct list_head *reaplist)
6841	{
6842	unsigned int maxreap = 0, reapcnt = 0;
6843	struct list_head *pos, *next;
6844	struct nfs4_client *clp;
6845
6846	maxreap = NFSD_CLIENT_MAX_TRIM_PER_RUN;
6847	INIT_LIST_HEAD(list: reaplist);
6848
6849	spin_lock(lock: &nn->client_lock);
6850	list_for_each_safe(pos, next, &nn->client_lru) {
6851	clp = list_entry(pos, struct nfs4_client, cl_lru);
6852	if (clp->cl_state == NFSD4_ACTIVE)
6853	break;
6854	if (reapcnt >= maxreap)
6855	break;
6856	if (!mark_client_expired_locked(clp)) {
6857	list_add(new: &clp->cl_lru, head: reaplist);
6858	reapcnt++;
6859	}
6860	}
6861	spin_unlock(lock: &nn->client_lock);
6862	}
6863
6864	static void
6865	nfs4_process_client_reaplist(struct list_head *reaplist)
6866	{
6867	struct list_head *pos, *next;
6868	struct nfs4_client *clp;
6869
6870	list_for_each_safe(pos, next, reaplist) {
6871	clp = list_entry(pos, struct nfs4_client, cl_lru);
6872	trace_nfsd_clid_purged(clid: &clp->cl_clientid);
6873	list_del_init(entry: &clp->cl_lru);
6874	expire_client(clp);
6875	}
6876	}
6877
6878	static void nfs40_clean_admin_revoked(struct nfsd_net *nn,
6879	struct laundry_time *lt)
6880	{
6881	struct nfs4_client *clp;
6882
6883	spin_lock(lock: &nn->client_lock);
6884	if (nn->nfs40_last_revoke == 0 ||
6885	nn->nfs40_last_revoke > lt->cutoff) {
6886	spin_unlock(lock: &nn->client_lock);
6887	return;
6888	}
6889	nn->nfs40_last_revoke = 0;
6890
6891	retry:
6892	list_for_each_entry(clp, &nn->client_lru, cl_lru) {
6893	unsigned long id, tmp;
6894	struct nfs4_stid *stid;
6895
6896	if (atomic_read(v: &clp->cl_admin_revoked) == 0)
6897	continue;
6898
6899	spin_lock(lock: &clp->cl_lock);
6900	idr_for_each_entry_ul(&clp->cl_stateids, stid, tmp, id)
6901	if (stid->sc_status & SC_STATUS_ADMIN_REVOKED) {
6902	refcount_inc(r: &stid->sc_count);
6903	spin_unlock(lock: &nn->client_lock);
6904	/* this function drops ->cl_lock */
6905	nfsd4_drop_revoked_stid(s: stid);
6906	nfs4_put_stid(s: stid);
6907	spin_lock(lock: &nn->client_lock);
6908	goto retry;
6909	}
6910	spin_unlock(lock: &clp->cl_lock);
6911	}
6912	spin_unlock(lock: &nn->client_lock);
6913	}
6914
6915	static time64_t
6916	nfs4_laundromat(struct nfsd_net *nn)
6917	{
6918	struct nfs4_openowner *oo;
6919	struct nfs4_delegation *dp;
6920	struct nfs4_ol_stateid *stp;
6921	struct nfsd4_blocked_lock *nbl;
6922	struct list_head *pos, *next, reaplist;
6923	struct laundry_time lt = {
6924	.cutoff = ktime_get_boottime_seconds() - nn->nfsd4_lease,
6925	.new_timeo = nn->nfsd4_lease
6926	};
6927	struct nfs4_cpntf_state *cps;
6928	copy_stateid_t *cps_t;
6929	int i;
6930
6931	if (clients_still_reclaiming(nn)) {
6932	lt.new_timeo = 0;
6933	goto out;
6934	}
6935	nfsd4_end_grace(nn);
6936
6937	spin_lock(lock: &nn->s2s_cp_lock);
6938	idr_for_each_entry(&nn->s2s_cp_stateids, cps_t, i) {
6939	cps = container_of(cps_t, struct nfs4_cpntf_state, cp_stateid);
6940	if (cps->cp_stateid.cs_type == NFS4_COPYNOTIFY_STID &&
6941	state_expired(lt: &lt, last_refresh: cps->cpntf_time))
6942	_free_cpntf_state_locked(nn, cps);
6943	}
6944	spin_unlock(lock: &nn->s2s_cp_lock);
6945	nfsd4_async_copy_reaper(nn);
6946	nfs4_get_client_reaplist(nn, reaplist: &reaplist, lt: &lt);
6947	nfs4_process_client_reaplist(reaplist: &reaplist);
6948
6949	nfs40_clean_admin_revoked(nn, lt: &lt);
6950
6951	spin_lock(lock: &state_lock);
6952	list_for_each_safe(pos, next, &nn->del_recall_lru) {
6953	dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
6954	if (!state_expired(lt: &lt, last_refresh: dp->dl_time))
6955	break;
6956	refcount_inc(r: &dp->dl_stid.sc_count);
6957	unhash_delegation_locked(dp, SC_STATUS_REVOKED);
6958	list_add(new: &dp->dl_recall_lru, head: &reaplist);
6959	}
6960	spin_unlock(lock: &state_lock);
6961	while (!list_empty(head: &reaplist)) {
6962	dp = list_first_entry(&reaplist, struct nfs4_delegation,
6963	dl_recall_lru);
6964	list_del_init(entry: &dp->dl_recall_lru);
6965	revoke_delegation(dp);
6966	}
6967
6968	spin_lock(lock: &nn->client_lock);
6969	while (!list_empty(head: &nn->close_lru)) {
6970	oo = list_first_entry(&nn->close_lru, struct nfs4_openowner,
6971	oo_close_lru);
6972	if (!state_expired(lt: &lt, last_refresh: oo->oo_time))
6973	break;
6974	list_del_init(entry: &oo->oo_close_lru);
6975	stp = oo->oo_last_closed_stid;
6976	oo->oo_last_closed_stid = NULL;
6977	spin_unlock(lock: &nn->client_lock);
6978	nfs4_put_stid(s: &stp->st_stid);
6979	spin_lock(lock: &nn->client_lock);
6980	}
6981	spin_unlock(lock: &nn->client_lock);
6982
6983	/*
6984	* It's possible for a client to try and acquire an already held lock
6985	* that is being held for a long time, and then lose interest in it.
6986	* So, we clean out any un-revisited request after a lease period
6987	* under the assumption that the client is no longer interested.
6988	*
6989	* RFC5661, sec. 9.6 states that the client must not rely on getting
6990	* notifications and must continue to poll for locks, even when the
6991	* server supports them. Thus this shouldn't lead to clients blocking
6992	* indefinitely once the lock does become free.
6993	*/
6994	BUG_ON(!list_empty(&reaplist));
6995	spin_lock(lock: &nn->blocked_locks_lock);
6996	while (!list_empty(head: &nn->blocked_locks_lru)) {
6997	nbl = list_first_entry(&nn->blocked_locks_lru,
6998	struct nfsd4_blocked_lock, nbl_lru);
6999	if (!state_expired(lt: &lt, last_refresh: nbl->nbl_time))
7000	break;
7001	list_move(list: &nbl->nbl_lru, head: &reaplist);
7002	list_del_init(entry: &nbl->nbl_list);
7003	}
7004	spin_unlock(lock: &nn->blocked_locks_lock);
7005
7006	while (!list_empty(head: &reaplist)) {
7007	nbl = list_first_entry(&reaplist,
7008	struct nfsd4_blocked_lock, nbl_lru);
7009	list_del_init(entry: &nbl->nbl_lru);
7010	free_blocked_lock(nbl);
7011	}
7012	#ifdef CONFIG_NFSD_V4_2_INTER_SSC
7013	/* service the server-to-server copy delayed unmount list */
7014	nfsd4_ssc_expire_umount(nn);
7015	#endif
7016	if (atomic_long_read(v: &num_delegations) >= max_delegations)
7017	deleg_reaper(nn);
7018	out:
7019	return max_t(time64_t, lt.new_timeo, NFSD_LAUNDROMAT_MINTIMEOUT);
7020	}
7021
7022	static void laundromat_main(struct work_struct *);
7023
7024	static void
7025	laundromat_main(struct work_struct *laundry)
7026	{
7027	time64_t t;
7028	struct delayed_work *dwork = to_delayed_work(work: laundry);
7029	struct nfsd_net *nn = container_of(dwork, struct nfsd_net,
7030	laundromat_work);
7031
7032	t = nfs4_laundromat(nn);
7033	queue_delayed_work(wq: laundry_wq, dwork: &nn->laundromat_work, delay: t*HZ);
7034	}
7035
7036	static void
7037	courtesy_client_reaper(struct nfsd_net *nn)
7038	{
7039	struct list_head reaplist;
7040
7041	nfs4_get_courtesy_client_reaplist(nn, reaplist: &reaplist);
7042	nfs4_process_client_reaplist(reaplist: &reaplist);
7043	}
7044
7045	static void
7046	deleg_reaper(struct nfsd_net *nn)
7047	{
7048	struct list_head *pos, *next;
7049	struct nfs4_client *clp;
7050
7051	spin_lock(lock: &nn->client_lock);
7052	list_for_each_safe(pos, next, &nn->client_lru) {
7053	clp = list_entry(pos, struct nfs4_client, cl_lru);
7054
7055	if (clp->cl_state != NFSD4_ACTIVE)
7056	continue;
7057	if (list_empty(head: &clp->cl_delegations))
7058	continue;
7059	if (atomic_read(v: &clp->cl_delegs_in_recall))
7060	continue;
7061	if (test_and_set_bit(NFSD4_CALLBACK_RUNNING, addr: &clp->cl_ra->ra_cb.cb_flags))
7062	continue;
7063	if (ktime_get_boottime_seconds() - clp->cl_ra_time < 5)
7064	continue;
7065	if (clp->cl_cb_state != NFSD4_CB_UP)
7066	continue;
7067
7068	/* release in nfsd4_cb_recall_any_release */
7069	kref_get(kref: &clp->cl_nfsdfs.cl_ref);
7070	clp->cl_ra_time = ktime_get_boottime_seconds();
7071	clp->cl_ra->ra_keep = 0;
7072	clp->cl_ra->ra_bmval[0] = BIT(RCA4_TYPE_MASK_RDATA_DLG) |
7073	BIT(RCA4_TYPE_MASK_WDATA_DLG);
7074	trace_nfsd_cb_recall_any(ra: clp->cl_ra);
7075	nfsd4_run_cb(cb: &clp->cl_ra->ra_cb);
7076	}
7077	spin_unlock(lock: &nn->client_lock);
7078	}
7079
7080	static void
7081	nfsd4_state_shrinker_worker(struct work_struct *work)
7082	{
7083	struct nfsd_net *nn = container_of(work, struct nfsd_net,
7084	nfsd_shrinker_work);
7085
7086	courtesy_client_reaper(nn);
7087	deleg_reaper(nn);
7088	}
7089
7090	static inline __be32 nfs4_check_fh(struct svc_fh *fhp, struct nfs4_stid *stp)
7091	{
7092	if (!fh_match(fh1: &fhp->fh_handle, fh2: &stp->sc_file->fi_fhandle))
7093	return nfserr_bad_stateid;
7094	return nfs_ok;
7095	}
7096
7097	static
7098	__be32 nfs4_check_openmode(struct nfs4_ol_stateid *stp, int flags)
7099	{
7100	__be32 status = nfserr_openmode;
7101
7102	/* For lock stateid's, we test the parent open, not the lock: */
7103	if (stp->st_openstp)
7104	stp = stp->st_openstp;
7105	if ((flags & WR_STATE) && !access_permit_write(stp))
7106	goto out;
7107	if ((flags & RD_STATE) && !access_permit_read(stp))
7108	goto out;
7109	status = nfs_ok;
7110	out:
7111	return status;
7112	}
7113
7114	static inline __be32
7115	check_special_stateids(struct net *net, svc_fh *current_fh, stateid_t *stateid, int flags)
7116	{
7117	if (ONE_STATEID(stateid) && (flags & RD_STATE))
7118	return nfs_ok;
7119	else if (opens_in_grace(net)) {
7120	/* Answer in remaining cases depends on existence of
7121	* conflicting state; so we must wait out the grace period. */
7122	return nfserr_grace;
7123	} else if (flags & WR_STATE)
7124	return nfs4_share_conflict(current_fh,
7125	NFS4_SHARE_DENY_WRITE);
7126	else /* (flags & RD_STATE) && ZERO_STATEID(stateid) */
7127	return nfs4_share_conflict(current_fh,
7128	NFS4_SHARE_DENY_READ);
7129	}
7130
7131	static __be32 check_stateid_generation(stateid_t *in, stateid_t *ref, bool has_session)
7132	{
7133	/*
7134	* When sessions are used the stateid generation number is ignored
7135	* when it is zero.
7136	*/
7137	if (has_session && in->si_generation == 0)
7138	return nfs_ok;
7139
7140	if (in->si_generation == ref->si_generation)
7141	return nfs_ok;
7142
7143	/* If the client sends us a stateid from the future, it's buggy: */
7144	if (nfsd4_stateid_generation_after(a: in, b: ref))
7145	return nfserr_bad_stateid;
7146	/*
7147	* However, we could see a stateid from the past, even from a
7148	* non-buggy client. For example, if the client sends a lock
7149	* while some IO is outstanding, the lock may bump si_generation
7150	* while the IO is still in flight. The client could avoid that
7151	* situation by waiting for responses on all the IO requests,
7152	* but better performance may result in retrying IO that
7153	* receives an old_stateid error if requests are rarely
7154	* reordered in flight:
7155	*/
7156	return nfserr_old_stateid;
7157	}
7158
7159	static __be32 nfsd4_stid_check_stateid_generation(stateid_t *in, struct nfs4_stid *s, bool has_session)
7160	{
7161	__be32 ret;
7162
7163	spin_lock(lock: &s->sc_lock);
7164	ret = nfsd4_verify_open_stid(s);
7165	if (ret == nfs_ok)
7166	ret = check_stateid_generation(in, ref: &s->sc_stateid, has_session);
7167	spin_unlock(lock: &s->sc_lock);
7168	if (ret == nfserr_admin_revoked)
7169	nfsd40_drop_revoked_stid(cl: s->sc_client,
7170	stid: &s->sc_stateid);
7171	return ret;
7172	}
7173
7174	static __be32 nfsd4_check_openowner_confirmed(struct nfs4_ol_stateid *ols)
7175	{
7176	if (ols->st_stateowner->so_is_open_owner &&
7177	!(openowner(so: ols->st_stateowner)->oo_flags & NFS4_OO_CONFIRMED))
7178	return nfserr_bad_stateid;
7179	return nfs_ok;
7180	}
7181
7182	static __be32 nfsd4_validate_stateid(struct nfs4_client *cl, stateid_t *stateid)
7183	{
7184	struct nfs4_stid *s;
7185	__be32 status = nfserr_bad_stateid;
7186
7187	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
7188	CLOSE_STATEID(stateid))
7189	return status;
7190	spin_lock(lock: &cl->cl_lock);
7191	s = find_stateid_locked(cl, t: stateid);
7192	if (!s)
7193	goto out_unlock;
7194	status = nfsd4_stid_check_stateid_generation(in: stateid, s, has_session: 1);
7195	if (status)
7196	goto out_unlock;
7197	status = nfsd4_verify_open_stid(s);
7198	if (status)
7199	goto out_unlock;
7200
7201	switch (s->sc_type) {
7202	case SC_TYPE_DELEG:
7203	status = nfs_ok;
7204	break;
7205	case SC_TYPE_OPEN:
7206	case SC_TYPE_LOCK:
7207	status = nfsd4_check_openowner_confirmed(ols: openlockstateid(s));
7208	break;
7209	default:
7210	printk("unknown stateid type %x\n", s->sc_type);
7211	status = nfserr_bad_stateid;
7212	}
7213	out_unlock:
7214	spin_unlock(lock: &cl->cl_lock);
7215	if (status == nfserr_admin_revoked)
7216	nfsd40_drop_revoked_stid(cl, stid: stateid);
7217	return status;
7218	}
7219
7220	__be32
7221	nfsd4_lookup_stateid(struct nfsd4_compound_state *cstate,
7222	stateid_t *stateid,
7223	unsigned short typemask, unsigned short statusmask,
7224	struct nfs4_stid **s, struct nfsd_net *nn)
7225	{
7226	__be32 status;
7227	struct nfs4_stid *stid;
7228	bool return_revoked = false;
7229
7230	/*
7231	* only return revoked delegations if explicitly asked.
7232	* otherwise we report revoked or bad_stateid status.
7233	*/
7234	if (statusmask & SC_STATUS_REVOKED)
7235	return_revoked = true;
7236	if (typemask & SC_TYPE_DELEG)
7237	/* Always allow REVOKED for DELEG so we can
7238	* return the appropriate error.
7239	*/
7240	statusmask |= SC_STATUS_REVOKED;
7241
7242	statusmask |= SC_STATUS_ADMIN_REVOKED | SC_STATUS_FREEABLE;
7243
7244	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid) ||
7245	CLOSE_STATEID(stateid))
7246	return nfserr_bad_stateid;
7247	status = set_client(clid: &stateid->si_opaque.so_clid, cstate, nn);
7248	if (status == nfserr_stale_clientid) {
7249	if (cstate->session)
7250	return nfserr_bad_stateid;
7251	return nfserr_stale_stateid;
7252	}
7253	if (status)
7254	return status;
7255	stid = find_stateid_by_type(cl: cstate->clp, t: stateid, typemask, ok_states: statusmask);
7256	if (!stid)
7257	return nfserr_bad_stateid;
7258	if ((stid->sc_status & SC_STATUS_REVOKED) && !return_revoked) {
7259	nfs4_put_stid(s: stid);
7260	return nfserr_deleg_revoked;
7261	}
7262	if (stid->sc_status & SC_STATUS_ADMIN_REVOKED) {
7263	nfsd40_drop_revoked_stid(cl: cstate->clp, stid: stateid);
7264	nfs4_put_stid(s: stid);
7265	return nfserr_admin_revoked;
7266	}
7267	*s = stid;
7268	return nfs_ok;
7269	}
7270
7271	static struct nfsd_file *
7272	nfs4_find_file(struct nfs4_stid *s, int flags)
7273	{
7274	struct nfsd_file *ret = NULL;
7275
7276	if (!s || s->sc_status)
7277	return NULL;
7278
7279	switch (s->sc_type) {
7280	case SC_TYPE_DELEG:
7281	case SC_TYPE_OPEN:
7282	case SC_TYPE_LOCK:
7283	if (flags & RD_STATE)
7284	ret = find_readable_file(f: s->sc_file);
7285	else
7286	ret = find_writeable_file(f: s->sc_file);
7287	}
7288
7289	return ret;
7290	}
7291
7292	static __be32
7293	nfs4_check_olstateid(struct nfs4_ol_stateid *ols, int flags)
7294	{
7295	__be32 status;
7296
7297	status = nfsd4_check_openowner_confirmed(ols);
7298	if (status)
7299	return status;
7300	return nfs4_check_openmode(stp: ols, flags);
7301	}
7302
7303	static __be32
7304	nfs4_check_file(struct svc_rqst *rqstp, struct svc_fh *fhp, struct nfs4_stid *s,
7305	struct nfsd_file **nfp, int flags)
7306	{
7307	int acc = (flags & RD_STATE) ? NFSD_MAY_READ : NFSD_MAY_WRITE;
7308	struct nfsd_file *nf;
7309	__be32 status;
7310
7311	nf = nfs4_find_file(s, flags);
7312	if (nf) {
7313	status = nfsd_permission(cred: &rqstp->rq_cred,
7314	exp: fhp->fh_export, dentry: fhp->fh_dentry,
7315	acc: acc | NFSD_MAY_OWNER_OVERRIDE);
7316	if (status) {
7317	nfsd_file_put(nf);
7318	goto out;
7319	}
7320	} else {
7321	status = nfsd_file_acquire(rqstp, fhp, may_flags: acc, nfp: &nf);
7322	if (status)
7323	return status;
7324	}
7325	*nfp = nf;
7326	out:
7327	return status;
7328	}
7329	static void
7330	_free_cpntf_state_locked(struct nfsd_net *nn, struct nfs4_cpntf_state *cps)
7331	{
7332	WARN_ON_ONCE(cps->cp_stateid.cs_type != NFS4_COPYNOTIFY_STID);
7333	if (!refcount_dec_and_test(r: &cps->cp_stateid.cs_count))
7334	return;
7335	list_del(entry: &cps->cp_list);
7336	idr_remove(&nn->s2s_cp_stateids,
7337	id: cps->cp_stateid.cs_stid.si_opaque.so_id);
7338	kfree(objp: cps);
7339	}
7340	/*
7341	* A READ from an inter server to server COPY will have a
7342	* copy stateid. Look up the copy notify stateid from the
7343	* idr structure and take a reference on it.
7344	*/
7345	__be32 manage_cpntf_state(struct nfsd_net *nn, stateid_t *st,
7346	struct nfs4_client *clp,
7347	struct nfs4_cpntf_state **cps)
7348	{
7349	copy_stateid_t *cps_t;
7350	struct nfs4_cpntf_state *state = NULL;
7351
7352	if (st->si_opaque.so_clid.cl_id != nn->s2s_cp_cl_id)
7353	return nfserr_bad_stateid;
7354	spin_lock(lock: &nn->s2s_cp_lock);
7355	cps_t = idr_find(&nn->s2s_cp_stateids, id: st->si_opaque.so_id);
7356	if (cps_t) {
7357	state = container_of(cps_t, struct nfs4_cpntf_state,
7358	cp_stateid);
7359	if (state->cp_stateid.cs_type != NFS4_COPYNOTIFY_STID) {
7360	state = NULL;
7361	goto unlock;
7362	}
7363	if (!clp)
7364	refcount_inc(r: &state->cp_stateid.cs_count);
7365	else
7366	_free_cpntf_state_locked(nn, cps: state);
7367	}
7368	unlock:
7369	spin_unlock(lock: &nn->s2s_cp_lock);
7370	if (!state)
7371	return nfserr_bad_stateid;
7372	if (!clp)
7373	*cps = state;
7374	return 0;
7375	}
7376
7377	static __be32 find_cpntf_state(struct nfsd_net *nn, stateid_t *st,
7378	struct nfs4_stid **stid)
7379	{
7380	__be32 status;
7381	struct nfs4_cpntf_state *cps = NULL;
7382	struct nfs4_client *found;
7383
7384	status = manage_cpntf_state(nn, st, NULL, cps: &cps);
7385	if (status)
7386	return status;
7387
7388	cps->cpntf_time = ktime_get_boottime_seconds();
7389
7390	status = nfserr_expired;
7391	found = lookup_clientid(clid: &cps->cp_p_clid, sessions: true, nn);
7392	if (!found)
7393	goto out;
7394
7395	*stid = find_stateid_by_type(cl: found, t: &cps->cp_p_stateid,
7396	SC_TYPE_DELEG|SC_TYPE_OPEN|SC_TYPE_LOCK,
7397	ok_states: 0);
7398	if (*stid)
7399	status = nfs_ok;
7400	else
7401	status = nfserr_bad_stateid;
7402
7403	put_client_renew(clp: found);
7404	out:
7405	nfs4_put_cpntf_state(nn, cps);
7406	return status;
7407	}
7408
7409	void nfs4_put_cpntf_state(struct nfsd_net *nn, struct nfs4_cpntf_state *cps)
7410	{
7411	spin_lock(lock: &nn->s2s_cp_lock);
7412	_free_cpntf_state_locked(nn, cps);
7413	spin_unlock(lock: &nn->s2s_cp_lock);
7414	}
7415
7416	/**
7417	* nfs4_preprocess_stateid_op - find and prep stateid for an operation
7418	* @rqstp: incoming request from client
7419	* @cstate: current compound state
7420	* @fhp: filehandle associated with requested stateid
7421	* @stateid: stateid (provided by client)
7422	* @flags: flags describing type of operation to be done
7423	* @nfp: optional nfsd_file return pointer (may be NULL)
7424	* @cstid: optional returned nfs4_stid pointer (may be NULL)
7425	*
7426	* Given info from the client, look up a nfs4_stid for the operation. On
7427	* success, it returns a reference to the nfs4_stid and/or the nfsd_file
7428	* associated with it.
7429	*/
7430	__be32
7431	nfs4_preprocess_stateid_op(struct svc_rqst *rqstp,
7432	struct nfsd4_compound_state *cstate, struct svc_fh *fhp,
7433	stateid_t *stateid, int flags, struct nfsd_file **nfp,
7434	struct nfs4_stid **cstid)
7435	{
7436	struct net *net = SVC_NET(rqstp);
7437	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
7438	struct nfs4_stid *s = NULL;
7439	__be32 status;
7440
7441	if (nfp)
7442	*nfp = NULL;
7443
7444	if (ZERO_STATEID(stateid) || ONE_STATEID(stateid)) {
7445	status = check_special_stateids(net, current_fh: fhp, stateid, flags);
7446	goto done;
7447	}
7448
7449	status = nfsd4_lookup_stateid(cstate, stateid,
7450	SC_TYPE_DELEG|SC_TYPE_OPEN|SC_TYPE_LOCK,
7451	statusmask: 0, s: &s, nn);
7452	if (status == nfserr_bad_stateid)
7453	status = find_cpntf_state(nn, st: stateid, stid: &s);
7454	if (status)
7455	return status;
7456	status = nfsd4_stid_check_stateid_generation(in: stateid, s,
7457	has_session: nfsd4_has_session(cs: cstate));
7458	if (status)
7459	goto out;
7460
7461	switch (s->sc_type) {
7462	case SC_TYPE_DELEG:
7463	status = nfs4_check_delegmode(dp: delegstateid(s), flags);
7464	break;
7465	case SC_TYPE_OPEN:
7466	case SC_TYPE_LOCK:
7467	status = nfs4_check_olstateid(ols: openlockstateid(s), flags);
7468	break;
7469	}
7470	if (status)
7471	goto out;
7472	status = nfs4_check_fh(fhp, stp: s);
7473
7474	done:
7475	if (status == nfs_ok && nfp)
7476	status = nfs4_check_file(rqstp, fhp, s, nfp, flags);
7477	out:
7478	if (s) {
7479	if (!status && cstid)
7480	*cstid = s;
7481	else
7482	nfs4_put_stid(s);
7483	}
7484	return status;
7485	}
7486
7487	/*
7488	* Test if the stateid is valid
7489	*/
7490	__be32
7491	nfsd4_test_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7492	union nfsd4_op_u *u)
7493	{
7494	struct nfsd4_test_stateid *test_stateid = &u->test_stateid;
7495	struct nfsd4_test_stateid_id *stateid;
7496	struct nfs4_client *cl = cstate->clp;
7497
7498	list_for_each_entry(stateid, &test_stateid->ts_stateid_list, ts_id_list)
7499	stateid->ts_id_status =
7500	nfsd4_validate_stateid(cl, stateid: &stateid->ts_id_stateid);
7501
7502	return nfs_ok;
7503	}
7504
7505	static __be32
7506	nfsd4_free_lock_stateid(stateid_t *stateid, struct nfs4_stid *s)
7507	{
7508	struct nfs4_ol_stateid *stp = openlockstateid(s);
7509	__be32 ret;
7510
7511	ret = nfsd4_lock_ol_stateid(stp);
7512	if (ret)
7513	goto out_put_stid;
7514
7515	ret = check_stateid_generation(in: stateid, ref: &s->sc_stateid, has_session: 1);
7516	if (ret)
7517	goto out;
7518
7519	ret = nfserr_locks_held;
7520	if (check_for_locks(fp: stp->st_stid.sc_file,
7521	lowner: lockowner(so: stp->st_stateowner)))
7522	goto out;
7523
7524	release_lock_stateid(stp);
7525	ret = nfs_ok;
7526
7527	out:
7528	mutex_unlock(lock: &stp->st_mutex);
7529	out_put_stid:
7530	nfs4_put_stid(s);
7531	return ret;
7532	}
7533
7534	__be32
7535	nfsd4_free_stateid(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7536	union nfsd4_op_u *u)
7537	{
7538	struct nfsd4_free_stateid *free_stateid = &u->free_stateid;
7539	stateid_t *stateid = &free_stateid->fr_stateid;
7540	struct nfs4_stid *s;
7541	struct nfs4_delegation *dp;
7542	struct nfs4_client *cl = cstate->clp;
7543	__be32 ret = nfserr_bad_stateid;
7544
7545	spin_lock(lock: &cl->cl_lock);
7546	s = find_stateid_locked(cl, t: stateid);
7547	if (!s || s->sc_status & SC_STATUS_CLOSED)
7548	goto out_unlock;
7549	if (s->sc_status & SC_STATUS_ADMIN_REVOKED) {
7550	nfsd4_drop_revoked_stid(s);
7551	ret = nfs_ok;
7552	goto out;
7553	}
7554	spin_lock(lock: &s->sc_lock);
7555	switch (s->sc_type) {
7556	case SC_TYPE_DELEG:
7557	if (s->sc_status & SC_STATUS_REVOKED) {
7558	s->sc_status |= SC_STATUS_CLOSED;
7559	spin_unlock(lock: &s->sc_lock);
7560	dp = delegstateid(s);
7561	if (s->sc_status & SC_STATUS_FREEABLE)
7562	list_del_init(entry: &dp->dl_recall_lru);
7563	s->sc_status |= SC_STATUS_FREED;
7564	spin_unlock(lock: &cl->cl_lock);
7565	nfs4_put_stid(s);
7566	ret = nfs_ok;
7567	goto out;
7568	}
7569	ret = nfserr_locks_held;
7570	break;
7571	case SC_TYPE_OPEN:
7572	ret = check_stateid_generation(in: stateid, ref: &s->sc_stateid, has_session: 1);
7573	if (ret)
7574	break;
7575	ret = nfserr_locks_held;
7576	break;
7577	case SC_TYPE_LOCK:
7578	spin_unlock(lock: &s->sc_lock);
7579	refcount_inc(r: &s->sc_count);
7580	spin_unlock(lock: &cl->cl_lock);
7581	ret = nfsd4_free_lock_stateid(stateid, s);
7582	goto out;
7583	}
7584	spin_unlock(lock: &s->sc_lock);
7585	out_unlock:
7586	spin_unlock(lock: &cl->cl_lock);
7587	out:
7588	return ret;
7589	}
7590
7591	static inline int
7592	setlkflg (int type)
7593	{
7594	return (type == NFS4_READW_LT || type == NFS4_READ_LT) ?
7595	RD_STATE : WR_STATE;
7596	}
7597
7598	static __be32 nfs4_seqid_op_checks(struct nfsd4_compound_state *cstate, stateid_t *stateid, u32 seqid, struct nfs4_ol_stateid *stp)
7599	{
7600	struct svc_fh *current_fh = &cstate->current_fh;
7601	struct nfs4_stateowner *sop = stp->st_stateowner;
7602	__be32 status;
7603
7604	status = nfsd4_check_seqid(cstate, so: sop, seqid);
7605	if (status)
7606	return status;
7607	status = nfsd4_lock_ol_stateid(stp);
7608	if (status != nfs_ok)
7609	return status;
7610	status = check_stateid_generation(in: stateid, ref: &stp->st_stid.sc_stateid, has_session: nfsd4_has_session(cs: cstate));
7611	if (status == nfs_ok)
7612	status = nfs4_check_fh(fhp: current_fh, stp: &stp->st_stid);
7613	if (status != nfs_ok)
7614	mutex_unlock(lock: &stp->st_mutex);
7615	return status;
7616	}
7617
7618	/**
7619	* nfs4_preprocess_seqid_op - find and prep an ol_stateid for a seqid-morphing op
7620	* @cstate: compund state
7621	* @seqid: seqid (provided by client)
7622	* @stateid: stateid (provided by client)
7623	* @typemask: mask of allowable types for this operation
7624	* @statusmask: mask of allowed states: 0 or STID_CLOSED
7625	* @stpp: return pointer for the stateid found
7626	* @nn: net namespace for request
7627	*
7628	* Given a stateid+seqid from a client, look up an nfs4_ol_stateid and
7629	* return it in @stpp. On a nfs_ok return, the returned stateid will
7630	* have its st_mutex locked.
7631	*/
7632	static __be32
7633	nfs4_preprocess_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
7634	stateid_t *stateid,
7635	unsigned short typemask, unsigned short statusmask,
7636	struct nfs4_ol_stateid **stpp,
7637	struct nfsd_net *nn)
7638	{
7639	__be32 status;
7640	struct nfs4_stid *s;
7641	struct nfs4_ol_stateid *stp = NULL;
7642
7643	trace_nfsd_preprocess(seqid, stp: stateid);
7644
7645	*stpp = NULL;
7646	retry:
7647	status = nfsd4_lookup_stateid(cstate, stateid,
7648	typemask, statusmask, s: &s, nn);
7649	if (status)
7650	return status;
7651	stp = openlockstateid(s);
7652	if (nfsd4_cstate_assign_replay(cstate, so: stp->st_stateowner) == -EAGAIN) {
7653	nfs4_put_stateowner(sop: stp->st_stateowner);
7654	goto retry;
7655	}
7656
7657	status = nfs4_seqid_op_checks(cstate, stateid, seqid, stp);
7658	if (!status)
7659	*stpp = stp;
7660	else
7661	nfs4_put_stid(s: &stp->st_stid);
7662	return status;
7663	}
7664
7665	static __be32 nfs4_preprocess_confirmed_seqid_op(struct nfsd4_compound_state *cstate, u32 seqid,
7666	stateid_t *stateid, struct nfs4_ol_stateid **stpp, struct nfsd_net *nn)
7667	{
7668	__be32 status;
7669	struct nfs4_openowner *oo;
7670	struct nfs4_ol_stateid *stp;
7671
7672	status = nfs4_preprocess_seqid_op(cstate, seqid, stateid,
7673	SC_TYPE_OPEN, statusmask: 0, stpp: &stp, nn);
7674	if (status)
7675	return status;
7676	oo = openowner(so: stp->st_stateowner);
7677	if (!(oo->oo_flags & NFS4_OO_CONFIRMED)) {
7678	mutex_unlock(lock: &stp->st_mutex);
7679	nfs4_put_stid(s: &stp->st_stid);
7680	return nfserr_bad_stateid;
7681	}
7682	*stpp = stp;
7683	return nfs_ok;
7684	}
7685
7686	__be32
7687	nfsd4_open_confirm(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7688	union nfsd4_op_u *u)
7689	{
7690	struct nfsd4_open_confirm *oc = &u->open_confirm;
7691	__be32 status;
7692	struct nfs4_openowner *oo;
7693	struct nfs4_ol_stateid *stp;
7694	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
7695
7696	dprintk("NFSD: nfsd4_open_confirm on file %pd\n",
7697	cstate->current_fh.fh_dentry);
7698
7699	status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
7700	if (status)
7701	return status;
7702
7703	status = nfs4_preprocess_seqid_op(cstate,
7704	seqid: oc->oc_seqid, stateid: &oc->oc_req_stateid,
7705	SC_TYPE_OPEN, statusmask: 0, stpp: &stp, nn);
7706	if (status)
7707	goto out;
7708	oo = openowner(so: stp->st_stateowner);
7709	status = nfserr_bad_stateid;
7710	if (oo->oo_flags & NFS4_OO_CONFIRMED) {
7711	mutex_unlock(lock: &stp->st_mutex);
7712	goto put_stateid;
7713	}
7714	oo->oo_flags |= NFS4_OO_CONFIRMED;
7715	nfs4_inc_and_copy_stateid(dst: &oc->oc_resp_stateid, stid: &stp->st_stid);
7716	mutex_unlock(lock: &stp->st_mutex);
7717	trace_nfsd_open_confirm(seqid: oc->oc_seqid, stp: &stp->st_stid.sc_stateid);
7718	nfsd4_client_record_create(clp: oo->oo_owner.so_client);
7719	status = nfs_ok;
7720	put_stateid:
7721	nfs4_put_stid(s: &stp->st_stid);
7722	out:
7723	nfsd4_bump_seqid(cstate, nfserr: status);
7724	return status;
7725	}
7726
7727	static inline void nfs4_stateid_downgrade_bit(struct nfs4_ol_stateid *stp, u32 access)
7728	{
7729	if (!test_access(access, stp))
7730	return;
7731	nfs4_file_put_access(fp: stp->st_stid.sc_file, access);
7732	clear_access(access, stp);
7733	}
7734
7735	static inline void nfs4_stateid_downgrade(struct nfs4_ol_stateid *stp, u32 to_access)
7736	{
7737	switch (to_access) {
7738	case NFS4_SHARE_ACCESS_READ:
7739	nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_WRITE);
7740	nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
7741	break;
7742	case NFS4_SHARE_ACCESS_WRITE:
7743	nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_READ);
7744	nfs4_stateid_downgrade_bit(stp, NFS4_SHARE_ACCESS_BOTH);
7745	break;
7746	case NFS4_SHARE_ACCESS_BOTH:
7747	break;
7748	default:
7749	WARN_ON_ONCE(1);
7750	}
7751	}
7752
7753	__be32
7754	nfsd4_open_downgrade(struct svc_rqst *rqstp,
7755	struct nfsd4_compound_state *cstate, union nfsd4_op_u *u)
7756	{
7757	struct nfsd4_open_downgrade *od = &u->open_downgrade;
7758	__be32 status;
7759	struct nfs4_ol_stateid *stp;
7760	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
7761
7762	dprintk("NFSD: nfsd4_open_downgrade on file %pd\n",
7763	cstate->current_fh.fh_dentry);
7764
7765	/* We don't yet support WANT bits: */
7766	if (od->od_deleg_want)
7767	dprintk("NFSD: %s: od_deleg_want=0x%x ignored\n", __func__,
7768	od->od_deleg_want);
7769
7770	status = nfs4_preprocess_confirmed_seqid_op(cstate, seqid: od->od_seqid,
7771	stateid: &od->od_stateid, stpp: &stp, nn);
7772	if (status)
7773	goto out;
7774	status = nfserr_inval;
7775	if (!test_access(access: od->od_share_access, stp)) {
7776	dprintk("NFSD: access not a subset of current bitmap: 0x%hhx, input access=%08x\n",
7777	stp->st_access_bmap, od->od_share_access);
7778	goto put_stateid;
7779	}
7780	if (!test_deny(deny: od->od_share_deny, stp)) {
7781	dprintk("NFSD: deny not a subset of current bitmap: 0x%hhx, input deny=%08x\n",
7782	stp->st_deny_bmap, od->od_share_deny);
7783	goto put_stateid;
7784	}
7785	nfs4_stateid_downgrade(stp, to_access: od->od_share_access);
7786	reset_union_bmap_deny(deny: od->od_share_deny, stp);
7787	nfs4_inc_and_copy_stateid(dst: &od->od_stateid, stid: &stp->st_stid);
7788	status = nfs_ok;
7789	put_stateid:
7790	mutex_unlock(lock: &stp->st_mutex);
7791	nfs4_put_stid(s: &stp->st_stid);
7792	out:
7793	nfsd4_bump_seqid(cstate, nfserr: status);
7794	return status;
7795	}
7796
7797	static bool nfsd4_close_open_stateid(struct nfs4_ol_stateid *s)
7798	{
7799	struct nfs4_client *clp = s->st_stid.sc_client;
7800	bool unhashed;
7801	LIST_HEAD(reaplist);
7802	struct nfs4_ol_stateid *stp;
7803
7804	spin_lock(lock: &clp->cl_lock);
7805	unhashed = unhash_open_stateid(stp: s, reaplist: &reaplist);
7806
7807	if (clp->cl_minorversion) {
7808	if (unhashed)
7809	put_ol_stateid_locked(stp: s, reaplist: &reaplist);
7810	spin_unlock(lock: &clp->cl_lock);
7811	list_for_each_entry(stp, &reaplist, st_locks)
7812	nfs4_free_cpntf_statelist(net: clp->net, stid: &stp->st_stid);
7813	free_ol_stateid_reaplist(reaplist: &reaplist);
7814	return false;
7815	} else {
7816	spin_unlock(lock: &clp->cl_lock);
7817	free_ol_stateid_reaplist(reaplist: &reaplist);
7818	return unhashed;
7819	}
7820	}
7821
7822	/*
7823	* nfs4_unlock_state() called after encode
7824	*/
7825	__be32
7826	nfsd4_close(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7827	union nfsd4_op_u *u)
7828	{
7829	struct nfsd4_close *close = &u->close;
7830	__be32 status;
7831	struct nfs4_ol_stateid *stp;
7832	struct net *net = SVC_NET(rqstp);
7833	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
7834	bool need_move_to_close_list;
7835
7836	dprintk("NFSD: nfsd4_close on file %pd\n",
7837	cstate->current_fh.fh_dentry);
7838
7839	status = nfs4_preprocess_seqid_op(cstate, seqid: close->cl_seqid,
7840	stateid: &close->cl_stateid,
7841	SC_TYPE_OPEN, SC_STATUS_CLOSED,
7842	stpp: &stp, nn);
7843	nfsd4_bump_seqid(cstate, nfserr: status);
7844	if (status)
7845	goto out;
7846
7847	spin_lock(lock: &stp->st_stid.sc_client->cl_lock);
7848	stp->st_stid.sc_status |= SC_STATUS_CLOSED;
7849	spin_unlock(lock: &stp->st_stid.sc_client->cl_lock);
7850
7851	/*
7852	* Technically we don't _really_ have to increment or copy it, since
7853	* it should just be gone after this operation and we clobber the
7854	* copied value below, but we continue to do so here just to ensure
7855	* that racing ops see that there was a state change.
7856	*/
7857	nfs4_inc_and_copy_stateid(dst: &close->cl_stateid, stid: &stp->st_stid);
7858
7859	need_move_to_close_list = nfsd4_close_open_stateid(s: stp);
7860	mutex_unlock(lock: &stp->st_mutex);
7861	if (need_move_to_close_list)
7862	move_to_close_lru(s: stp, net);
7863
7864	/* v4.1+ suggests that we send a special stateid in here, since the
7865	* clients should just ignore this anyway. Since this is not useful
7866	* for v4.0 clients either, we set it to the special close_stateid
7867	* universally.
7868	*
7869	* See RFC5661 section 18.2.4, and RFC7530 section 16.2.5
7870	*/
7871	memcpy(&close->cl_stateid, &close_stateid, sizeof(close->cl_stateid));
7872
7873	/* put reference from nfs4_preprocess_seqid_op */
7874	nfs4_put_stid(s: &stp->st_stid);
7875	out:
7876	return status;
7877	}
7878
7879	__be32
7880	nfsd4_delegreturn(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
7881	union nfsd4_op_u *u)
7882	{
7883	struct nfsd4_delegreturn *dr = &u->delegreturn;
7884	struct nfs4_delegation *dp;
7885	stateid_t *stateid = &dr->dr_stateid;
7886	struct nfs4_stid *s;
7887	__be32 status;
7888	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
7889
7890	status = fh_verify(rqstp, &cstate->current_fh, 0, 0);
7891	if (status)
7892	return status;
7893
7894	status = nfsd4_lookup_stateid(cstate, stateid, SC_TYPE_DELEG, SC_STATUS_REVOKED, s: &s, nn);
7895	if (status)
7896	goto out;
7897	dp = delegstateid(s);
7898	status = nfsd4_stid_check_stateid_generation(in: stateid, s: &dp->dl_stid, has_session: nfsd4_has_session(cs: cstate));
7899	if (status)
7900	goto put_stateid;
7901
7902	trace_nfsd_deleg_return(stp: stateid);
7903	destroy_delegation(dp);
7904	smp_mb__after_atomic();
7905	wake_up_var(var: d_inode(dentry: cstate->current_fh.fh_dentry));
7906	put_stateid:
7907	nfs4_put_stid(s: &dp->dl_stid);
7908	out:
7909	return status;
7910	}
7911
7912	/* last octet in a range */
7913	static inline u64
7914	last_byte_offset(u64 start, u64 len)
7915	{
7916	u64 end;
7917
7918	WARN_ON_ONCE(!len);
7919	end = start + len;
7920	return end > start ? end - 1: NFS4_MAX_UINT64;
7921	}
7922
7923	/*
7924	* TODO: Linux file offsets are _signed_ 64-bit quantities, which means that
7925	* we can't properly handle lock requests that go beyond the (2^63 - 1)-th
7926	* byte, because of sign extension problems. Since NFSv4 calls for 64-bit
7927	* locking, this prevents us from being completely protocol-compliant. The
7928	* real solution to this problem is to start using unsigned file offsets in
7929	* the VFS, but this is a very deep change!
7930	*/
7931	static inline void
7932	nfs4_transform_lock_offset(struct file_lock *lock)
7933	{
7934	if (lock->fl_start < 0)
7935	lock->fl_start = OFFSET_MAX;
7936	if (lock->fl_end < 0)
7937	lock->fl_end = OFFSET_MAX;
7938	}
7939
7940	static fl_owner_t
7941	nfsd4_lm_get_owner(fl_owner_t owner)
7942	{
7943	struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
7944
7945	nfs4_get_stateowner(sop: &lo->lo_owner);
7946	return owner;
7947	}
7948
7949	static void
7950	nfsd4_lm_put_owner(fl_owner_t owner)
7951	{
7952	struct nfs4_lockowner *lo = (struct nfs4_lockowner *)owner;
7953
7954	if (lo)
7955	nfs4_put_stateowner(sop: &lo->lo_owner);
7956	}
7957
7958	/* return pointer to struct nfs4_client if client is expirable */
7959	static bool
7960	nfsd4_lm_lock_expirable(struct file_lock *cfl)
7961	{
7962	struct nfs4_lockowner *lo = (struct nfs4_lockowner *) cfl->c.flc_owner;
7963	struct nfs4_client *clp = lo->lo_owner.so_client;
7964	struct nfsd_net *nn;
7965
7966	if (try_to_expire_client(clp)) {
7967	nn = net_generic(net: clp->net, id: nfsd_net_id);
7968	mod_delayed_work(wq: laundry_wq, dwork: &nn->laundromat_work, delay: 0);
7969	return true;
7970	}
7971	return false;
7972	}
7973
7974	/* schedule laundromat to run immediately and wait for it to complete */
7975	static void
7976	nfsd4_lm_expire_lock(void)
7977	{
7978	flush_workqueue(laundry_wq);
7979	}
7980
7981	static void
7982	nfsd4_lm_notify(struct file_lock *fl)
7983	{
7984	struct nfs4_lockowner *lo = (struct nfs4_lockowner *) fl->c.flc_owner;
7985	struct net *net = lo->lo_owner.so_client->net;
7986	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
7987	struct nfsd4_blocked_lock *nbl = container_of(fl,
7988	struct nfsd4_blocked_lock, nbl_lock);
7989	bool queue = false;
7990
7991	/* An empty list means that something else is going to be using it */
7992	spin_lock(lock: &nn->blocked_locks_lock);
7993	if (!list_empty(head: &nbl->nbl_list)) {
7994	list_del_init(entry: &nbl->nbl_list);
7995	list_del_init(entry: &nbl->nbl_lru);
7996	queue = true;
7997	}
7998	spin_unlock(lock: &nn->blocked_locks_lock);
7999
8000	if (queue) {
8001	trace_nfsd_cb_notify_lock(lo, nbl);
8002	nfsd4_try_run_cb(cb: &nbl->nbl_cb);
8003	}
8004	}
8005
8006	static const struct lock_manager_operations nfsd_posix_mng_ops = {
8007	.lm_mod_owner = THIS_MODULE,
8008	.lm_notify = nfsd4_lm_notify,
8009	.lm_get_owner = nfsd4_lm_get_owner,
8010	.lm_put_owner = nfsd4_lm_put_owner,
8011	.lm_lock_expirable = nfsd4_lm_lock_expirable,
8012	.lm_expire_lock = nfsd4_lm_expire_lock,
8013	};
8014
8015	static inline void
8016	nfs4_set_lock_denied(struct file_lock *fl, struct nfsd4_lock_denied *deny)
8017	{
8018	struct nfs4_lockowner *lo;
8019
8020	if (fl->fl_lmops == &nfsd_posix_mng_ops) {
8021	lo = (struct nfs4_lockowner *) fl->c.flc_owner;
8022	xdr_netobj_dup(dst: &deny->ld_owner, src: &lo->lo_owner.so_owner,
8023	GFP_KERNEL);
8024	if (!deny->ld_owner.data)
8025	/* We just don't care that much */
8026	goto nevermind;
8027	deny->ld_clientid = lo->lo_owner.so_client->cl_clientid;
8028	} else {
8029	nevermind:
8030	deny->ld_owner.len = 0;
8031	deny->ld_owner.data = NULL;
8032	deny->ld_clientid.cl_boot = 0;
8033	deny->ld_clientid.cl_id = 0;
8034	}
8035	deny->ld_start = fl->fl_start;
8036	deny->ld_length = NFS4_MAX_UINT64;
8037	if (fl->fl_end != NFS4_MAX_UINT64)
8038	deny->ld_length = fl->fl_end - fl->fl_start + 1;
8039	deny->ld_type = NFS4_READ_LT;
8040	if (fl->c.flc_type != F_RDLCK)
8041	deny->ld_type = NFS4_WRITE_LT;
8042	}
8043
8044	static struct nfs4_lockowner *
8045	find_lockowner_str_locked(struct nfs4_client *clp, struct xdr_netobj *owner)
8046	{
8047	unsigned int strhashval = ownerstr_hashval(ownername: owner);
8048	struct nfs4_stateowner *so;
8049
8050	lockdep_assert_held(&clp->cl_lock);
8051
8052	list_for_each_entry(so, &clp->cl_ownerstr_hashtbl[strhashval],
8053	so_strhash) {
8054	if (so->so_is_open_owner)
8055	continue;
8056	if (same_owner_str(sop: so, owner))
8057	return lockowner(so: nfs4_get_stateowner(sop: so));
8058	}
8059	return NULL;
8060	}
8061
8062	static struct nfs4_lockowner *
8063	find_lockowner_str(struct nfs4_client *clp, struct xdr_netobj *owner)
8064	{
8065	struct nfs4_lockowner *lo;
8066
8067	spin_lock(lock: &clp->cl_lock);
8068	lo = find_lockowner_str_locked(clp, owner);
8069	spin_unlock(lock: &clp->cl_lock);
8070	return lo;
8071	}
8072
8073	static void nfs4_unhash_lockowner(struct nfs4_stateowner *sop)
8074	{
8075	unhash_lockowner_locked(lo: lockowner(so: sop));
8076	}
8077
8078	static void nfs4_free_lockowner(struct nfs4_stateowner *sop)
8079	{
8080	struct nfs4_lockowner *lo = lockowner(so: sop);
8081
8082	kmem_cache_free(s: lockowner_slab, objp: lo);
8083	}
8084
8085	static const struct nfs4_stateowner_operations lockowner_ops = {
8086	.so_unhash = nfs4_unhash_lockowner,
8087	.so_free = nfs4_free_lockowner,
8088	};
8089
8090	/*
8091	* Alloc a lock owner structure.
8092	* Called in nfsd4_lock - therefore, OPEN and OPEN_CONFIRM (if needed) has
8093	* occurred.
8094	*
8095	* strhashval = ownerstr_hashval
8096	*/
8097	static struct nfs4_lockowner *
8098	alloc_init_lock_stateowner(unsigned int strhashval, struct nfs4_client *clp,
8099	struct nfs4_ol_stateid *open_stp,
8100	struct nfsd4_lock *lock)
8101	{
8102	struct nfs4_lockowner *lo, *ret;
8103
8104	lo = alloc_stateowner(slab: lockowner_slab, owner: &lock->lk_new_owner, clp);
8105	if (!lo)
8106	return NULL;
8107	INIT_LIST_HEAD(list: &lo->lo_blocked);
8108	INIT_LIST_HEAD(list: &lo->lo_owner.so_stateids);
8109	lo->lo_owner.so_is_open_owner = 0;
8110	lo->lo_owner.so_seqid = lock->lk_new_lock_seqid;
8111	lo->lo_owner.so_ops = &lockowner_ops;
8112	spin_lock(lock: &clp->cl_lock);
8113	ret = find_lockowner_str_locked(clp, owner: &lock->lk_new_owner);
8114	if (ret == NULL) {
8115	list_add(new: &lo->lo_owner.so_strhash,
8116	head: &clp->cl_ownerstr_hashtbl[strhashval]);
8117	ret = lo;
8118	} else
8119	nfs4_free_stateowner(sop: &lo->lo_owner);
8120
8121	spin_unlock(lock: &clp->cl_lock);
8122	return ret;
8123	}
8124
8125	static struct nfs4_ol_stateid *
8126	find_lock_stateid(const struct nfs4_lockowner *lo,
8127	const struct nfs4_ol_stateid *ost)
8128	{
8129	struct nfs4_ol_stateid *lst;
8130
8131	lockdep_assert_held(&ost->st_stid.sc_client->cl_lock);
8132
8133	/* If ost is not hashed, ost->st_locks will not be valid */
8134	if (!nfs4_ol_stateid_unhashed(stp: ost))
8135	list_for_each_entry(lst, &ost->st_locks, st_locks) {
8136	if (lst->st_stateowner == &lo->lo_owner) {
8137	refcount_inc(r: &lst->st_stid.sc_count);
8138	return lst;
8139	}
8140	}
8141	return NULL;
8142	}
8143
8144	static struct nfs4_ol_stateid *
8145	init_lock_stateid(struct nfs4_ol_stateid *stp, struct nfs4_lockowner *lo,
8146	struct nfs4_file *fp, struct inode *inode,
8147	struct nfs4_ol_stateid *open_stp)
8148	{
8149	struct nfs4_client *clp = lo->lo_owner.so_client;
8150	struct nfs4_ol_stateid *retstp;
8151
8152	mutex_init(&stp->st_mutex);
8153	mutex_lock_nested(lock: &stp->st_mutex, subclass: OPEN_STATEID_MUTEX);
8154	retry:
8155	spin_lock(lock: &clp->cl_lock);
8156	if (nfs4_ol_stateid_unhashed(stp: open_stp))
8157	goto out_close;
8158	retstp = find_lock_stateid(lo, ost: open_stp);
8159	if (retstp)
8160	goto out_found;
8161	refcount_inc(r: &stp->st_stid.sc_count);
8162	stp->st_stid.sc_type = SC_TYPE_LOCK;
8163	stp->st_stateowner = nfs4_get_stateowner(sop: &lo->lo_owner);
8164	get_nfs4_file(fi: fp);
8165	stp->st_stid.sc_file = fp;
8166	stp->st_access_bmap = 0;
8167	stp->st_deny_bmap = open_stp->st_deny_bmap;
8168	stp->st_openstp = open_stp;
8169	spin_lock(lock: &fp->fi_lock);
8170	list_add(new: &stp->st_locks, head: &open_stp->st_locks);
8171	list_add(new: &stp->st_perstateowner, head: &lo->lo_owner.so_stateids);
8172	list_add(new: &stp->st_perfile, head: &fp->fi_stateids);
8173	spin_unlock(lock: &fp->fi_lock);
8174	spin_unlock(lock: &clp->cl_lock);
8175	return stp;
8176	out_found:
8177	spin_unlock(lock: &clp->cl_lock);
8178	if (nfsd4_lock_ol_stateid(stp: retstp) != nfs_ok) {
8179	nfs4_put_stid(s: &retstp->st_stid);
8180	goto retry;
8181	}
8182	/* To keep mutex tracking happy */
8183	mutex_unlock(lock: &stp->st_mutex);
8184	return retstp;
8185	out_close:
8186	spin_unlock(lock: &clp->cl_lock);
8187	mutex_unlock(lock: &stp->st_mutex);
8188	return NULL;
8189	}
8190
8191	static struct nfs4_ol_stateid *
8192	find_or_create_lock_stateid(struct nfs4_lockowner *lo, struct nfs4_file *fi,
8193	struct inode *inode, struct nfs4_ol_stateid *ost,
8194	bool *new)
8195	{
8196	struct nfs4_stid *ns = NULL;
8197	struct nfs4_ol_stateid *lst;
8198	struct nfs4_openowner *oo = openowner(so: ost->st_stateowner);
8199	struct nfs4_client *clp = oo->oo_owner.so_client;
8200
8201	*new = false;
8202	spin_lock(lock: &clp->cl_lock);
8203	lst = find_lock_stateid(lo, ost);
8204	spin_unlock(lock: &clp->cl_lock);
8205	if (lst != NULL) {
8206	if (nfsd4_lock_ol_stateid(stp: lst) == nfs_ok)
8207	goto out;
8208	nfs4_put_stid(s: &lst->st_stid);
8209	}
8210	ns = nfs4_alloc_stid(cl: clp, slab: stateid_slab, sc_free: nfs4_free_lock_stateid);
8211	if (ns == NULL)
8212	return NULL;
8213
8214	lst = init_lock_stateid(stp: openlockstateid(s: ns), lo, fp: fi, inode, open_stp: ost);
8215	if (lst == openlockstateid(s: ns))
8216	*new = true;
8217	else
8218	nfs4_put_stid(s: ns);
8219	out:
8220	return lst;
8221	}
8222
8223	static int
8224	check_lock_length(u64 offset, u64 length)
8225	{
8226	return ((length == 0) || ((length != NFS4_MAX_UINT64) &&
8227	(length > ~offset)));
8228	}
8229
8230	static void get_lock_access(struct nfs4_ol_stateid *lock_stp, u32 access)
8231	{
8232	struct nfs4_file *fp = lock_stp->st_stid.sc_file;
8233
8234	lockdep_assert_held(&fp->fi_lock);
8235
8236	if (test_access(access, stp: lock_stp))
8237	return;
8238	__nfs4_file_get_access(fp, access);
8239	set_access(access, stp: lock_stp);
8240	}
8241
8242	static __be32
8243	lookup_or_create_lock_state(struct nfsd4_compound_state *cstate,
8244	struct nfs4_ol_stateid *ost,
8245	struct nfsd4_lock *lock,
8246	struct nfs4_ol_stateid **plst, bool *new)
8247	{
8248	__be32 status;
8249	struct nfs4_file *fi = ost->st_stid.sc_file;
8250	struct nfs4_openowner *oo = openowner(so: ost->st_stateowner);
8251	struct nfs4_client *cl = oo->oo_owner.so_client;
8252	struct inode *inode = d_inode(dentry: cstate->current_fh.fh_dentry);
8253	struct nfs4_lockowner *lo;
8254	struct nfs4_ol_stateid *lst;
8255	unsigned int strhashval;
8256
8257	lo = find_lockowner_str(clp: cl, owner: &lock->lk_new_owner);
8258	if (!lo) {
8259	strhashval = ownerstr_hashval(ownername: &lock->lk_new_owner);
8260	lo = alloc_init_lock_stateowner(strhashval, clp: cl, open_stp: ost, lock);
8261	if (lo == NULL)
8262	return nfserr_jukebox;
8263	} else {
8264	/* with an existing lockowner, seqids must be the same */
8265	status = nfserr_bad_seqid;
8266	if (!cstate->minorversion &&
8267	lock->lk_new_lock_seqid != lo->lo_owner.so_seqid)
8268	goto out;
8269	}
8270
8271	lst = find_or_create_lock_stateid(lo, fi, inode, ost, new);
8272	if (lst == NULL) {
8273	status = nfserr_jukebox;
8274	goto out;
8275	}
8276
8277	status = nfs_ok;
8278	*plst = lst;
8279	out:
8280	nfs4_put_stateowner(sop: &lo->lo_owner);
8281	return status;
8282	}
8283
8284	/*
8285	* LOCK operation
8286	*/
8287	__be32
8288	nfsd4_lock(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
8289	union nfsd4_op_u *u)
8290	{
8291	struct nfsd4_lock *lock = &u->lock;
8292	struct nfs4_openowner *open_sop = NULL;
8293	struct nfs4_lockowner *lock_sop = NULL;
8294	struct nfs4_ol_stateid *lock_stp = NULL;
8295	struct nfs4_ol_stateid *open_stp = NULL;
8296	struct nfs4_file *fp;
8297	struct nfsd_file *nf = NULL;
8298	struct nfsd4_blocked_lock *nbl = NULL;
8299	struct file_lock *file_lock = NULL;
8300	struct file_lock *conflock = NULL;
8301	__be32 status = 0;
8302	int lkflg;
8303	int err;
8304	bool new = false;
8305	unsigned char type;
8306	unsigned int flags = FL_POSIX;
8307	struct net *net = SVC_NET(rqstp);
8308	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
8309
8310	dprintk("NFSD: nfsd4_lock: start=%Ld length=%Ld\n",
8311	(long long) lock->lk_offset,
8312	(long long) lock->lk_length);
8313
8314	if (check_lock_length(offset: lock->lk_offset, length: lock->lk_length))
8315	return nfserr_inval;
8316
8317	status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0);
8318	if (status != nfs_ok)
8319	return status;
8320	if (exportfs_cannot_lock(export_ops: cstate->current_fh.fh_dentry->d_sb->s_export_op)) {
8321	status = nfserr_notsupp;
8322	goto out;
8323	}
8324
8325	if (lock->lk_is_new) {
8326	if (nfsd4_has_session(cs: cstate))
8327	/* See rfc 5661 18.10.3: given clientid is ignored: */
8328	memcpy(&lock->lk_new_clientid,
8329	&cstate->clp->cl_clientid,
8330	sizeof(clientid_t));
8331
8332	/* validate and update open stateid and open seqid */
8333	status = nfs4_preprocess_confirmed_seqid_op(cstate,
8334	seqid: lock->lk_new_open_seqid,
8335	stateid: &lock->lk_new_open_stateid,
8336	stpp: &open_stp, nn);
8337	if (status)
8338	goto out;
8339	mutex_unlock(lock: &open_stp->st_mutex);
8340	open_sop = openowner(so: open_stp->st_stateowner);
8341	status = nfserr_bad_stateid;
8342	if (!same_clid(cl1: &open_sop->oo_owner.so_client->cl_clientid,
8343	cl2: &lock->lk_new_clientid))
8344	goto out;
8345	status = lookup_or_create_lock_state(cstate, ost: open_stp, lock,
8346	plst: &lock_stp, new: &new);
8347	} else {
8348	status = nfs4_preprocess_seqid_op(cstate,
8349	seqid: lock->lk_old_lock_seqid,
8350	stateid: &lock->lk_old_lock_stateid,
8351	SC_TYPE_LOCK, statusmask: 0, stpp: &lock_stp,
8352	nn);
8353	}
8354	if (status)
8355	goto out;
8356	lock_sop = lockowner(so: lock_stp->st_stateowner);
8357
8358	lkflg = setlkflg(lock->lk_type);
8359	status = nfs4_check_openmode(stp: lock_stp, flags: lkflg);
8360	if (status)
8361	goto out;
8362
8363	status = nfserr_grace;
8364	if (locks_in_grace(net) && !lock->lk_reclaim)
8365	goto out;
8366	status = nfserr_no_grace;
8367	if (!locks_in_grace(net) && lock->lk_reclaim)
8368	goto out;
8369
8370	if (lock->lk_reclaim)
8371	flags |= FL_RECLAIM;
8372
8373	fp = lock_stp->st_stid.sc_file;
8374	switch (lock->lk_type) {
8375	case NFS4_READW_LT:
8376	fallthrough;
8377	case NFS4_READ_LT:
8378	spin_lock(lock: &fp->fi_lock);
8379	nf = find_readable_file_locked(f: fp);
8380	if (nf)
8381	get_lock_access(lock_stp, NFS4_SHARE_ACCESS_READ);
8382	spin_unlock(lock: &fp->fi_lock);
8383	type = F_RDLCK;
8384	break;
8385	case NFS4_WRITEW_LT:
8386	fallthrough;
8387	case NFS4_WRITE_LT:
8388	spin_lock(lock: &fp->fi_lock);
8389	nf = find_writeable_file_locked(f: fp);
8390	if (nf)
8391	get_lock_access(lock_stp, NFS4_SHARE_ACCESS_WRITE);
8392	spin_unlock(lock: &fp->fi_lock);
8393	type = F_WRLCK;
8394	break;
8395	default:
8396	status = nfserr_inval;
8397	goto out;
8398	}
8399
8400	if (!nf) {
8401	status = nfserr_openmode;
8402	goto out;
8403	}
8404
8405	if (lock->lk_type & (NFS4_READW_LT | NFS4_WRITEW_LT) &&
8406	nfsd4_has_session(cs: cstate) &&
8407	locks_can_async_lock(fops: nf->nf_file->f_op))
8408	flags |= FL_SLEEP;
8409
8410	nbl = find_or_allocate_block(lo: lock_sop, fh: &fp->fi_fhandle, nn);
8411	if (!nbl) {
8412	dprintk("NFSD: %s: unable to allocate block!\n", __func__);
8413	status = nfserr_jukebox;
8414	goto out;
8415	}
8416
8417	file_lock = &nbl->nbl_lock;
8418	file_lock->c.flc_type = type;
8419	file_lock->c.flc_owner = (fl_owner_t)lockowner(so: nfs4_get_stateowner(sop: &lock_sop->lo_owner));
8420	file_lock->c.flc_pid = current->tgid;
8421	file_lock->c.flc_file = nf->nf_file;
8422	file_lock->c.flc_flags = flags;
8423	file_lock->fl_lmops = &nfsd_posix_mng_ops;
8424	file_lock->fl_start = lock->lk_offset;
8425	file_lock->fl_end = last_byte_offset(start: lock->lk_offset, len: lock->lk_length);
8426	nfs4_transform_lock_offset(lock: file_lock);
8427
8428	conflock = locks_alloc_lock();
8429	if (!conflock) {
8430	dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
8431	status = nfserr_jukebox;
8432	goto out;
8433	}
8434
8435	if (flags & FL_SLEEP) {
8436	nbl->nbl_time = ktime_get_boottime_seconds();
8437	spin_lock(lock: &nn->blocked_locks_lock);
8438	list_add_tail(new: &nbl->nbl_list, head: &lock_sop->lo_blocked);
8439	list_add_tail(new: &nbl->nbl_lru, head: &nn->blocked_locks_lru);
8440	kref_get(kref: &nbl->nbl_kref);
8441	spin_unlock(lock: &nn->blocked_locks_lock);
8442	}
8443
8444	err = vfs_lock_file(nf->nf_file, F_SETLK, file_lock, conflock);
8445	switch (err) {
8446	case 0: /* success! */
8447	nfs4_inc_and_copy_stateid(dst: &lock->lk_resp_stateid, stid: &lock_stp->st_stid);
8448	status = 0;
8449	if (lock->lk_reclaim)
8450	nn->somebody_reclaimed = true;
8451	break;
8452	case FILE_LOCK_DEFERRED:
8453	kref_put(kref: &nbl->nbl_kref, release: free_nbl);
8454	nbl = NULL;
8455	fallthrough;
8456	case -EAGAIN: /* conflock holds conflicting lock */
8457	status = nfserr_denied;
8458	dprintk("NFSD: nfsd4_lock: conflicting lock found!\n");
8459	nfs4_set_lock_denied(fl: conflock, deny: &lock->lk_denied);
8460	break;
8461	case -EDEADLK:
8462	status = nfserr_deadlock;
8463	break;
8464	default:
8465	dprintk("NFSD: nfsd4_lock: vfs_lock_file() failed! status %d\n",err);
8466	status = nfserrno(errno: err);
8467	break;
8468	}
8469	out:
8470	if (nbl) {
8471	/* dequeue it if we queued it before */
8472	if (flags & FL_SLEEP) {
8473	spin_lock(lock: &nn->blocked_locks_lock);
8474	if (!list_empty(head: &nbl->nbl_list) &&
8475	!list_empty(head: &nbl->nbl_lru)) {
8476	list_del_init(entry: &nbl->nbl_list);
8477	list_del_init(entry: &nbl->nbl_lru);
8478	kref_put(kref: &nbl->nbl_kref, release: free_nbl);
8479	}
8480	/* nbl can use one of lists to be linked to reaplist */
8481	spin_unlock(lock: &nn->blocked_locks_lock);
8482	}
8483	free_blocked_lock(nbl);
8484	}
8485	if (nf)
8486	nfsd_file_put(nf);
8487	if (lock_stp) {
8488	/* Bump seqid manually if the 4.0 replay owner is openowner */
8489	if (cstate->replay_owner &&
8490	cstate->replay_owner != &lock_sop->lo_owner &&
8491	seqid_mutating_err(ntohl(status)))
8492	lock_sop->lo_owner.so_seqid++;
8493
8494	/*
8495	* If this is a new, never-before-used stateid, and we are
8496	* returning an error, then just go ahead and release it.
8497	*/
8498	if (status && new)
8499	release_lock_stateid(stp: lock_stp);
8500
8501	mutex_unlock(lock: &lock_stp->st_mutex);
8502
8503	nfs4_put_stid(s: &lock_stp->st_stid);
8504	}
8505	if (open_stp)
8506	nfs4_put_stid(s: &open_stp->st_stid);
8507	nfsd4_bump_seqid(cstate, nfserr: status);
8508	if (conflock)
8509	locks_free_lock(fl: conflock);
8510	return status;
8511	}
8512
8513	void nfsd4_lock_release(union nfsd4_op_u *u)
8514	{
8515	struct nfsd4_lock *lock = &u->lock;
8516	struct nfsd4_lock_denied *deny = &lock->lk_denied;
8517
8518	kfree(objp: deny->ld_owner.data);
8519	}
8520
8521	/*
8522	* The NFSv4 spec allows a client to do a LOCKT without holding an OPEN,
8523	* so we do a temporary open here just to get an open file to pass to
8524	* vfs_test_lock.
8525	*/
8526	static __be32 nfsd_test_lock(struct svc_rqst *rqstp, struct svc_fh *fhp, struct file_lock *lock)
8527	{
8528	struct nfsd_file *nf;
8529	struct inode *inode;
8530	__be32 err;
8531
8532	err = nfsd_file_acquire(rqstp, fhp, NFSD_MAY_READ, nfp: &nf);
8533	if (err)
8534	return err;
8535	inode = fhp->fh_dentry->d_inode;
8536	inode_lock(inode); /* to block new leases till after test_lock: */
8537	err = nfserrno(errno: nfsd_open_break_lease(inode, NFSD_MAY_READ));
8538	if (err)
8539	goto out;
8540	lock->c.flc_file = nf->nf_file;
8541	err = nfserrno(errno: vfs_test_lock(nf->nf_file, lock));
8542	lock->c.flc_file = NULL;
8543	out:
8544	inode_unlock(inode);
8545	nfsd_file_put(nf);
8546	return err;
8547	}
8548
8549	/*
8550	* LOCKT operation
8551	*/
8552	__be32
8553	nfsd4_lockt(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
8554	union nfsd4_op_u *u)
8555	{
8556	struct nfsd4_lockt *lockt = &u->lockt;
8557	struct file_lock *file_lock = NULL;
8558	struct nfs4_lockowner *lo = NULL;
8559	__be32 status;
8560	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
8561
8562	if (locks_in_grace(SVC_NET(rqstp)))
8563	return nfserr_grace;
8564
8565	if (check_lock_length(offset: lockt->lt_offset, length: lockt->lt_length))
8566	return nfserr_inval;
8567
8568	if (!nfsd4_has_session(cs: cstate)) {
8569	status = set_client(clid: &lockt->lt_clientid, cstate, nn);
8570	if (status)
8571	goto out;
8572	}
8573
8574	if ((status = fh_verify(rqstp, &cstate->current_fh, S_IFREG, 0)))
8575	goto out;
8576
8577	file_lock = locks_alloc_lock();
8578	if (!file_lock) {
8579	dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
8580	status = nfserr_jukebox;
8581	goto out;
8582	}
8583
8584	switch (lockt->lt_type) {
8585	case NFS4_READ_LT:
8586	case NFS4_READW_LT:
8587	file_lock->c.flc_type = F_RDLCK;
8588	break;
8589	case NFS4_WRITE_LT:
8590	case NFS4_WRITEW_LT:
8591	file_lock->c.flc_type = F_WRLCK;
8592	break;
8593	default:
8594	dprintk("NFSD: nfs4_lockt: bad lock type!\n");
8595	status = nfserr_inval;
8596	goto out;
8597	}
8598
8599	lo = find_lockowner_str(clp: cstate->clp, owner: &lockt->lt_owner);
8600	if (lo)
8601	file_lock->c.flc_owner = (fl_owner_t)lo;
8602	file_lock->c.flc_pid = current->tgid;
8603	file_lock->c.flc_flags = FL_POSIX;
8604
8605	file_lock->fl_start = lockt->lt_offset;
8606	file_lock->fl_end = last_byte_offset(start: lockt->lt_offset, len: lockt->lt_length);
8607
8608	nfs4_transform_lock_offset(lock: file_lock);
8609
8610	status = nfsd_test_lock(rqstp, fhp: &cstate->current_fh, lock: file_lock);
8611	if (status)
8612	goto out;
8613
8614	if (file_lock->c.flc_type != F_UNLCK) {
8615	status = nfserr_denied;
8616	nfs4_set_lock_denied(fl: file_lock, deny: &lockt->lt_denied);
8617	}
8618	out:
8619	if (lo)
8620	nfs4_put_stateowner(sop: &lo->lo_owner);
8621	if (file_lock)
8622	locks_free_lock(fl: file_lock);
8623	return status;
8624	}
8625
8626	void nfsd4_lockt_release(union nfsd4_op_u *u)
8627	{
8628	struct nfsd4_lockt *lockt = &u->lockt;
8629	struct nfsd4_lock_denied *deny = &lockt->lt_denied;
8630
8631	kfree(objp: deny->ld_owner.data);
8632	}
8633
8634	__be32
8635	nfsd4_locku(struct svc_rqst *rqstp, struct nfsd4_compound_state *cstate,
8636	union nfsd4_op_u *u)
8637	{
8638	struct nfsd4_locku *locku = &u->locku;
8639	struct nfs4_ol_stateid *stp;
8640	struct nfsd_file *nf = NULL;
8641	struct file_lock *file_lock = NULL;
8642	__be32 status;
8643	int err;
8644	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
8645
8646	dprintk("NFSD: nfsd4_locku: start=%Ld length=%Ld\n",
8647	(long long) locku->lu_offset,
8648	(long long) locku->lu_length);
8649
8650	if (check_lock_length(offset: locku->lu_offset, length: locku->lu_length))
8651	return nfserr_inval;
8652
8653	status = nfs4_preprocess_seqid_op(cstate, seqid: locku->lu_seqid,
8654	stateid: &locku->lu_stateid, SC_TYPE_LOCK, statusmask: 0,
8655	stpp: &stp, nn);
8656	if (status)
8657	goto out;
8658	nf = find_any_file(f: stp->st_stid.sc_file);
8659	if (!nf) {
8660	status = nfserr_lock_range;
8661	goto put_stateid;
8662	}
8663	if (exportfs_cannot_lock(export_ops: nf->nf_file->f_path.mnt->mnt_sb->s_export_op)) {
8664	status = nfserr_notsupp;
8665	goto put_file;
8666	}
8667
8668	file_lock = locks_alloc_lock();
8669	if (!file_lock) {
8670	dprintk("NFSD: %s: unable to allocate lock!\n", __func__);
8671	status = nfserr_jukebox;
8672	goto put_file;
8673	}
8674
8675	file_lock->c.flc_type = F_UNLCK;
8676	file_lock->c.flc_owner = (fl_owner_t)lockowner(so: nfs4_get_stateowner(sop: stp->st_stateowner));
8677	file_lock->c.flc_pid = current->tgid;
8678	file_lock->c.flc_file = nf->nf_file;
8679	file_lock->c.flc_flags = FL_POSIX;
8680	file_lock->fl_lmops = &nfsd_posix_mng_ops;
8681	file_lock->fl_start = locku->lu_offset;
8682
8683	file_lock->fl_end = last_byte_offset(start: locku->lu_offset,
8684	len: locku->lu_length);
8685	nfs4_transform_lock_offset(lock: file_lock);
8686
8687	err = vfs_lock_file(nf->nf_file, F_SETLK, file_lock, NULL);
8688	if (err) {
8689	dprintk("NFSD: nfs4_locku: vfs_lock_file failed!\n");
8690	goto out_nfserr;
8691	}
8692	nfs4_inc_and_copy_stateid(dst: &locku->lu_stateid, stid: &stp->st_stid);
8693	put_file:
8694	nfsd_file_put(nf);
8695	put_stateid:
8696	mutex_unlock(lock: &stp->st_mutex);
8697	nfs4_put_stid(s: &stp->st_stid);
8698	out:
8699	nfsd4_bump_seqid(cstate, nfserr: status);
8700	if (file_lock)
8701	locks_free_lock(fl: file_lock);
8702	return status;
8703
8704	out_nfserr:
8705	status = nfserrno(errno: err);
8706	goto put_file;
8707	}
8708
8709	/*
8710	* returns
8711	* true: locks held by lockowner
8712	* false: no locks held by lockowner
8713	*/
8714	static bool
8715	check_for_locks(struct nfs4_file *fp, struct nfs4_lockowner *lowner)
8716	{
8717	struct file_lock *fl;
8718	int status = false;
8719	struct nfsd_file *nf;
8720	struct inode *inode;
8721	struct file_lock_context *flctx;
8722
8723	spin_lock(lock: &fp->fi_lock);
8724	nf = find_any_file_locked(f: fp);
8725	if (!nf) {
8726	/* Any valid lock stateid should have some sort of access */
8727	WARN_ON_ONCE(1);
8728	goto out;
8729	}
8730
8731	inode = file_inode(f: nf->nf_file);
8732	flctx = locks_inode_context(inode);
8733
8734	if (flctx && !list_empty_careful(head: &flctx->flc_posix)) {
8735	spin_lock(lock: &flctx->flc_lock);
8736	for_each_file_lock(fl, &flctx->flc_posix) {
8737	if (fl->c.flc_owner == (fl_owner_t)lowner) {
8738	status = true;
8739	break;
8740	}
8741	}
8742	spin_unlock(lock: &flctx->flc_lock);
8743	}
8744	out:
8745	spin_unlock(lock: &fp->fi_lock);
8746	return status;
8747	}
8748
8749	/**
8750	* nfsd4_release_lockowner - process NFSv4.0 RELEASE_LOCKOWNER operations
8751	* @rqstp: RPC transaction
8752	* @cstate: NFSv4 COMPOUND state
8753	* @u: RELEASE_LOCKOWNER arguments
8754	*
8755	* Check if there are any locks still held and if not, free the lockowner
8756	* and any lock state that is owned.
8757	*
8758	* Return values:
8759	* %nfs_ok: lockowner released or not found
8760	* %nfserr_locks_held: lockowner still in use
8761	* %nfserr_stale_clientid: clientid no longer active
8762	* %nfserr_expired: clientid not recognized
8763	*/
8764	__be32
8765	nfsd4_release_lockowner(struct svc_rqst *rqstp,
8766	struct nfsd4_compound_state *cstate,
8767	union nfsd4_op_u *u)
8768	{
8769	struct nfsd4_release_lockowner *rlockowner = &u->release_lockowner;
8770	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
8771	clientid_t *clid = &rlockowner->rl_clientid;
8772	struct nfs4_ol_stateid *stp;
8773	struct nfs4_lockowner *lo;
8774	struct nfs4_client *clp;
8775	LIST_HEAD(reaplist);
8776	__be32 status;
8777
8778	dprintk("nfsd4_release_lockowner clientid: (%08x/%08x):\n",
8779	clid->cl_boot, clid->cl_id);
8780
8781	status = set_client(clid, cstate, nn);
8782	if (status)
8783	return status;
8784	clp = cstate->clp;
8785
8786	spin_lock(lock: &clp->cl_lock);
8787	lo = find_lockowner_str_locked(clp, owner: &rlockowner->rl_owner);
8788	if (!lo) {
8789	spin_unlock(lock: &clp->cl_lock);
8790	return nfs_ok;
8791	}
8792
8793	list_for_each_entry(stp, &lo->lo_owner.so_stateids, st_perstateowner) {
8794	if (check_for_locks(fp: stp->st_stid.sc_file, lowner: lo)) {
8795	spin_unlock(lock: &clp->cl_lock);
8796	nfs4_put_stateowner(sop: &lo->lo_owner);
8797	return nfserr_locks_held;
8798	}
8799	}
8800	unhash_lockowner_locked(lo);
8801	while (!list_empty(head: &lo->lo_owner.so_stateids)) {
8802	stp = list_first_entry(&lo->lo_owner.so_stateids,
8803	struct nfs4_ol_stateid,
8804	st_perstateowner);
8805	unhash_lock_stateid(stp);
8806	put_ol_stateid_locked(stp, reaplist: &reaplist);
8807	}
8808	spin_unlock(lock: &clp->cl_lock);
8809
8810	free_ol_stateid_reaplist(reaplist: &reaplist);
8811	remove_blocked_locks(lo);
8812	nfs4_put_stateowner(sop: &lo->lo_owner);
8813	return nfs_ok;
8814	}
8815
8816	static inline struct nfs4_client_reclaim *
8817	alloc_reclaim(void)
8818	{
8819	return kmalloc(sizeof(struct nfs4_client_reclaim), GFP_KERNEL);
8820	}
8821
8822	bool
8823	nfs4_has_reclaimed_state(struct xdr_netobj name, struct nfsd_net *nn)
8824	{
8825	struct nfs4_client_reclaim *crp;
8826
8827	crp = nfsd4_find_reclaim_client(name, nn);
8828	return (crp && crp->cr_clp);
8829	}
8830
8831	/*
8832	* failure => all reset bets are off, nfserr_no_grace...
8833	*/
8834	struct nfs4_client_reclaim *
8835	nfs4_client_to_reclaim(struct xdr_netobj name, struct xdr_netobj princhash,
8836	struct nfsd_net *nn)
8837	{
8838	unsigned int strhashval;
8839	struct nfs4_client_reclaim *crp;
8840
8841	name.data = kmemdup(name.data, name.len, GFP_KERNEL);
8842	if (!name.data) {
8843	dprintk("%s: failed to allocate memory for name.data!\n",
8844	__func__);
8845	return NULL;
8846	}
8847	if (princhash.len) {
8848	princhash.data = kmemdup(princhash.data, princhash.len, GFP_KERNEL);
8849	if (!princhash.data) {
8850	dprintk("%s: failed to allocate memory for princhash.data!\n",
8851	__func__);
8852	kfree(objp: name.data);
8853	return NULL;
8854	}
8855	} else
8856	princhash.data = NULL;
8857	crp = alloc_reclaim();
8858	if (crp) {
8859	strhashval = clientstr_hashval(name);
8860	INIT_LIST_HEAD(list: &crp->cr_strhash);
8861	list_add(new: &crp->cr_strhash, head: &nn->reclaim_str_hashtbl[strhashval]);
8862	crp->cr_name.data = name.data;
8863	crp->cr_name.len = name.len;
8864	crp->cr_princhash.data = princhash.data;
8865	crp->cr_princhash.len = princhash.len;
8866	crp->cr_clp = NULL;
8867	nn->reclaim_str_hashtbl_size++;
8868	} else {
8869	kfree(objp: name.data);
8870	kfree(objp: princhash.data);
8871	}
8872	return crp;
8873	}
8874
8875	void
8876	nfs4_remove_reclaim_record(struct nfs4_client_reclaim *crp, struct nfsd_net *nn)
8877	{
8878	list_del(entry: &crp->cr_strhash);
8879	kfree(objp: crp->cr_name.data);
8880	kfree(objp: crp->cr_princhash.data);
8881	kfree(objp: crp);
8882	nn->reclaim_str_hashtbl_size--;
8883	}
8884
8885	void
8886	nfs4_release_reclaim(struct nfsd_net *nn)
8887	{
8888	struct nfs4_client_reclaim *crp = NULL;
8889	int i;
8890
8891	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
8892	while (!list_empty(head: &nn->reclaim_str_hashtbl[i])) {
8893	crp = list_entry(nn->reclaim_str_hashtbl[i].next,
8894	struct nfs4_client_reclaim, cr_strhash);
8895	nfs4_remove_reclaim_record(crp, nn);
8896	}
8897	}
8898	WARN_ON_ONCE(nn->reclaim_str_hashtbl_size);
8899	}
8900
8901	/*
8902	* called from OPEN, CLAIM_PREVIOUS with a new clientid. */
8903	struct nfs4_client_reclaim *
8904	nfsd4_find_reclaim_client(struct xdr_netobj name, struct nfsd_net *nn)
8905	{
8906	unsigned int strhashval;
8907	struct nfs4_client_reclaim *crp = NULL;
8908
8909	strhashval = clientstr_hashval(name);
8910	list_for_each_entry(crp, &nn->reclaim_str_hashtbl[strhashval], cr_strhash) {
8911	if (compare_blob(o1: &crp->cr_name, o2: &name) == 0) {
8912	return crp;
8913	}
8914	}
8915	return NULL;
8916	}
8917
8918	__be32
8919	nfs4_check_open_reclaim(struct nfs4_client *clp)
8920	{
8921	if (test_bit(NFSD4_CLIENT_RECLAIM_COMPLETE, &clp->cl_flags))
8922	return nfserr_no_grace;
8923
8924	if (nfsd4_client_record_check(clp))
8925	return nfserr_reclaim_bad;
8926
8927	return nfs_ok;
8928	}
8929
8930	/*
8931	* Since the lifetime of a delegation isn't limited to that of an open, a
8932	* client may quite reasonably hang on to a delegation as long as it has
8933	* the inode cached. This becomes an obvious problem the first time a
8934	* client's inode cache approaches the size of the server's total memory.
8935	*
8936	* For now we avoid this problem by imposing a hard limit on the number
8937	* of delegations, which varies according to the server's memory size.
8938	*/
8939	static void
8940	set_max_delegations(void)
8941	{
8942	/*
8943	* Allow at most 4 delegations per megabyte of RAM. Quick
8944	* estimates suggest that in the worst case (where every delegation
8945	* is for a different inode), a delegation could take about 1.5K,
8946	* giving a worst case usage of about 6% of memory.
8947	*/
8948	max_delegations = nr_free_buffer_pages() >> (20 - 2 - PAGE_SHIFT);
8949	}
8950
8951	static int nfs4_state_create_net(struct net *net)
8952	{
8953	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
8954	int i;
8955
8956	nn->conf_id_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
8957	sizeof(struct list_head),
8958	GFP_KERNEL);
8959	if (!nn->conf_id_hashtbl)
8960	goto err;
8961	nn->unconf_id_hashtbl = kmalloc_array(CLIENT_HASH_SIZE,
8962	sizeof(struct list_head),
8963	GFP_KERNEL);
8964	if (!nn->unconf_id_hashtbl)
8965	goto err_unconf_id;
8966	nn->sessionid_hashtbl = kmalloc_array(SESSION_HASH_SIZE,
8967	sizeof(struct list_head),
8968	GFP_KERNEL);
8969	if (!nn->sessionid_hashtbl)
8970	goto err_sessionid;
8971
8972	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
8973	INIT_LIST_HEAD(list: &nn->conf_id_hashtbl[i]);
8974	INIT_LIST_HEAD(list: &nn->unconf_id_hashtbl[i]);
8975	}
8976	for (i = 0; i < SESSION_HASH_SIZE; i++)
8977	INIT_LIST_HEAD(list: &nn->sessionid_hashtbl[i]);
8978	nn->conf_name_tree = RB_ROOT;
8979	nn->unconf_name_tree = RB_ROOT;
8980	nn->boot_time = ktime_get_real_seconds();
8981	nn->grace_ended = false;
8982	nn->grace_end_forced = false;
8983	nn->client_tracking_active = false;
8984	nn->nfsd4_manager.block_opens = true;
8985	INIT_LIST_HEAD(list: &nn->nfsd4_manager.list);
8986	INIT_LIST_HEAD(list: &nn->client_lru);
8987	INIT_LIST_HEAD(list: &nn->close_lru);
8988	INIT_LIST_HEAD(list: &nn->del_recall_lru);
8989	spin_lock_init(&nn->client_lock);
8990	spin_lock_init(&nn->s2s_cp_lock);
8991	idr_init(idr: &nn->s2s_cp_stateids);
8992	atomic_set(v: &nn->pending_async_copies, i: 0);
8993
8994	spin_lock_init(&nn->blocked_locks_lock);
8995	INIT_LIST_HEAD(list: &nn->blocked_locks_lru);
8996
8997	INIT_DELAYED_WORK(&nn->laundromat_work, laundromat_main);
8998	INIT_WORK(&nn->nfsd_shrinker_work, nfsd4_state_shrinker_worker);
8999	get_net(net);
9000
9001	nn->nfsd_client_shrinker = shrinker_alloc(flags: 0, fmt: "nfsd-client");
9002	if (!nn->nfsd_client_shrinker)
9003	goto err_shrinker;
9004
9005	nn->nfsd_client_shrinker->scan_objects = nfsd4_state_shrinker_scan;
9006	nn->nfsd_client_shrinker->count_objects = nfsd4_state_shrinker_count;
9007	nn->nfsd_client_shrinker->private_data = nn;
9008
9009	shrinker_register(shrinker: nn->nfsd_client_shrinker);
9010
9011	return 0;
9012
9013	err_shrinker:
9014	put_net(net);
9015	kfree(objp: nn->sessionid_hashtbl);
9016	err_sessionid:
9017	kfree(objp: nn->unconf_id_hashtbl);
9018	err_unconf_id:
9019	kfree(objp: nn->conf_id_hashtbl);
9020	err:
9021	return -ENOMEM;
9022	}
9023
9024	static void
9025	nfs4_state_destroy_net(struct net *net)
9026	{
9027	int i;
9028	struct nfs4_client *clp = NULL;
9029	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
9030
9031	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
9032	while (!list_empty(head: &nn->conf_id_hashtbl[i])) {
9033	clp = list_entry(nn->conf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
9034	destroy_client(clp);
9035	}
9036	}
9037
9038	WARN_ON(!list_empty(&nn->blocked_locks_lru));
9039
9040	for (i = 0; i < CLIENT_HASH_SIZE; i++) {
9041	while (!list_empty(head: &nn->unconf_id_hashtbl[i])) {
9042	clp = list_entry(nn->unconf_id_hashtbl[i].next, struct nfs4_client, cl_idhash);
9043	destroy_client(clp);
9044	}
9045	}
9046
9047	kfree(objp: nn->sessionid_hashtbl);
9048	kfree(objp: nn->unconf_id_hashtbl);
9049	kfree(objp: nn->conf_id_hashtbl);
9050	put_net(net);
9051	}
9052
9053	int
9054	nfs4_state_start_net(struct net *net)
9055	{
9056	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
9057	int ret;
9058
9059	ret = nfs4_state_create_net(net);
9060	if (ret)
9061	return ret;
9062	locks_start_grace(net, &nn->nfsd4_manager);
9063	nfsd4_client_tracking_init(net);
9064	/* safe for laundromat to run now */
9065	spin_lock(lock: &nn->client_lock);
9066	nn->client_tracking_active = true;
9067	spin_unlock(lock: &nn->client_lock);
9068	if (nn->track_reclaim_completes && nn->reclaim_str_hashtbl_size == 0)
9069	goto skip_grace;
9070	printk(KERN_INFO "NFSD: starting %lld-second grace period (net %x)\n",
9071	nn->nfsd4_grace, net->ns.inum);
9072	trace_nfsd_grace_start(nn);
9073	queue_delayed_work(wq: laundry_wq, dwork: &nn->laundromat_work, delay: nn->nfsd4_grace * HZ);
9074	return 0;
9075
9076	skip_grace:
9077	printk(KERN_INFO "NFSD: no clients to reclaim, skipping NFSv4 grace period (net %x)\n",
9078	net->ns.inum);
9079	queue_delayed_work(wq: laundry_wq, dwork: &nn->laundromat_work, delay: nn->nfsd4_lease * HZ);
9080	nfsd4_end_grace(nn);
9081	return 0;
9082	}
9083
9084	/* initialization to perform when the nfsd service is started: */
9085	int
9086	nfs4_state_start(void)
9087	{
9088	int ret;
9089
9090	ret = rhltable_init(&nfs4_file_rhltable, &nfs4_file_rhash_params);
9091	if (ret)
9092	return ret;
9093
9094	nfsd_slot_shrinker = shrinker_alloc(flags: 0, fmt: "nfsd-DRC-slot");
9095	if (!nfsd_slot_shrinker) {
9096	rhltable_destroy(hlt: &nfs4_file_rhltable);
9097	return -ENOMEM;
9098	}
9099	nfsd_slot_shrinker->count_objects = nfsd_slot_count;
9100	nfsd_slot_shrinker->scan_objects = nfsd_slot_scan;
9101	shrinker_register(shrinker: nfsd_slot_shrinker);
9102
9103	set_max_delegations();
9104	return 0;
9105	}
9106
9107	void
9108	nfs4_state_shutdown_net(struct net *net)
9109	{
9110	struct nfs4_delegation *dp = NULL;
9111	struct list_head *pos, *next, reaplist;
9112	struct nfsd_net *nn = net_generic(net, id: nfsd_net_id);
9113
9114	shrinker_free(shrinker: nn->nfsd_client_shrinker);
9115	cancel_work_sync(work: &nn->nfsd_shrinker_work);
9116	spin_lock(lock: &nn->client_lock);
9117	nn->client_tracking_active = false;
9118	spin_unlock(lock: &nn->client_lock);
9119	cancel_delayed_work_sync(dwork: &nn->laundromat_work);
9120	locks_end_grace(&nn->nfsd4_manager);
9121
9122	INIT_LIST_HEAD(list: &reaplist);
9123	spin_lock(lock: &state_lock);
9124	list_for_each_safe(pos, next, &nn->del_recall_lru) {
9125	dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
9126	unhash_delegation_locked(dp, SC_STATUS_CLOSED);
9127	list_add(new: &dp->dl_recall_lru, head: &reaplist);
9128	}
9129	spin_unlock(lock: &state_lock);
9130	list_for_each_safe(pos, next, &reaplist) {
9131	dp = list_entry (pos, struct nfs4_delegation, dl_recall_lru);
9132	list_del_init(entry: &dp->dl_recall_lru);
9133	destroy_unhashed_deleg(dp);
9134	}
9135
9136	nfsd4_client_tracking_exit(net);
9137	nfs4_state_destroy_net(net);
9138	#ifdef CONFIG_NFSD_V4_2_INTER_SSC
9139	nfsd4_ssc_shutdown_umount(nn);
9140	#endif
9141	}
9142
9143	void
9144	nfs4_state_shutdown(void)
9145	{
9146	rhltable_destroy(hlt: &nfs4_file_rhltable);
9147	shrinker_free(shrinker: nfsd_slot_shrinker);
9148	}
9149
9150	static void
9151	get_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
9152	{
9153	if (HAS_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG) &&
9154	CURRENT_STATEID(stateid))
9155	memcpy(stateid, &cstate->current_stateid, sizeof(stateid_t));
9156	}
9157
9158	static void
9159	put_stateid(struct nfsd4_compound_state *cstate, stateid_t *stateid)
9160	{
9161	if (cstate->minorversion) {
9162	memcpy(&cstate->current_stateid, stateid, sizeof(stateid_t));
9163	SET_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
9164	}
9165	}
9166
9167	void
9168	clear_current_stateid(struct nfsd4_compound_state *cstate)
9169	{
9170	CLEAR_CSTATE_FLAG(cstate, CURRENT_STATE_ID_FLAG);
9171	}
9172
9173	/*
9174	* functions to set current state id
9175	*/
9176	void
9177	nfsd4_set_opendowngradestateid(struct nfsd4_compound_state *cstate,
9178	union nfsd4_op_u *u)
9179	{
9180	put_stateid(cstate, stateid: &u->open_downgrade.od_stateid);
9181	}
9182
9183	void
9184	nfsd4_set_openstateid(struct nfsd4_compound_state *cstate,
9185	union nfsd4_op_u *u)
9186	{
9187	put_stateid(cstate, stateid: &u->open.op_stateid);
9188	}
9189
9190	void
9191	nfsd4_set_closestateid(struct nfsd4_compound_state *cstate,
9192	union nfsd4_op_u *u)
9193	{
9194	put_stateid(cstate, stateid: &u->close.cl_stateid);
9195	}
9196
9197	void
9198	nfsd4_set_lockstateid(struct nfsd4_compound_state *cstate,
9199	union nfsd4_op_u *u)
9200	{
9201	put_stateid(cstate, stateid: &u->lock.lk_resp_stateid);
9202	}
9203
9204	/*
9205	* functions to consume current state id
9206	*/
9207
9208	void
9209	nfsd4_get_opendowngradestateid(struct nfsd4_compound_state *cstate,
9210	union nfsd4_op_u *u)
9211	{
9212	get_stateid(cstate, stateid: &u->open_downgrade.od_stateid);
9213	}
9214
9215	void
9216	nfsd4_get_delegreturnstateid(struct nfsd4_compound_state *cstate,
9217	union nfsd4_op_u *u)
9218	{
9219	get_stateid(cstate, stateid: &u->delegreturn.dr_stateid);
9220	}
9221
9222	void
9223	nfsd4_get_freestateid(struct nfsd4_compound_state *cstate,
9224	union nfsd4_op_u *u)
9225	{
9226	get_stateid(cstate, stateid: &u->free_stateid.fr_stateid);
9227	}
9228
9229	void
9230	nfsd4_get_setattrstateid(struct nfsd4_compound_state *cstate,
9231	union nfsd4_op_u *u)
9232	{
9233	get_stateid(cstate, stateid: &u->setattr.sa_stateid);
9234	}
9235
9236	void
9237	nfsd4_get_closestateid(struct nfsd4_compound_state *cstate,
9238	union nfsd4_op_u *u)
9239	{
9240	get_stateid(cstate, stateid: &u->close.cl_stateid);
9241	}
9242
9243	void
9244	nfsd4_get_lockustateid(struct nfsd4_compound_state *cstate,
9245	union nfsd4_op_u *u)
9246	{
9247	get_stateid(cstate, stateid: &u->locku.lu_stateid);
9248	}
9249
9250	void
9251	nfsd4_get_readstateid(struct nfsd4_compound_state *cstate,
9252	union nfsd4_op_u *u)
9253	{
9254	get_stateid(cstate, stateid: &u->read.rd_stateid);
9255	}
9256
9257	void
9258	nfsd4_get_writestateid(struct nfsd4_compound_state *cstate,
9259	union nfsd4_op_u *u)
9260	{
9261	get_stateid(cstate, stateid: &u->write.wr_stateid);
9262	}
9263
9264	/**
9265	* nfsd4_vet_deleg_time - vet and set the timespec for a delegated timestamp update
9266	* @req: timestamp from the client
9267	* @orig: original timestamp in the inode
9268	* @now: current time
9269	*
9270	* Given a timestamp from the client response, check it against the
9271	* current timestamp in the inode and the current time. Returns true
9272	* if the inode's timestamp needs to be updated, and false otherwise.
9273	* @req may also be changed if the timestamp needs to be clamped.
9274	*/
9275	bool nfsd4_vet_deleg_time(struct timespec64 *req, const struct timespec64 *orig,
9276	const struct timespec64 *now)
9277	{
9278
9279	/*
9280	* "When the time presented is before the original time, then the
9281	* update is ignored." Also no need to update if there is no change.
9282	*/
9283	if (timespec64_compare(lhs: req, rhs: orig) <= 0)
9284	return false;
9285
9286	/*
9287	* "When the time presented is in the future, the server can either
9288	* clamp the new time to the current time, or it may
9289	* return NFS4ERR_DELAY to the client, allowing it to retry."
9290	*/
9291	if (timespec64_compare(lhs: req, rhs: now) > 0)
9292	*req = *now;
9293
9294	return true;
9295	}
9296
9297	static int cb_getattr_update_times(struct dentry *dentry, struct nfs4_delegation *dp)
9298	{
9299	struct inode *inode = d_inode(dentry);
9300	struct nfs4_cb_fattr *ncf = &dp->dl_cb_fattr;
9301	struct iattr attrs = { };
9302	int ret;
9303
9304	if (deleg_attrs_deleg(dl_type: dp->dl_type)) {
9305	struct timespec64 now = current_time(inode);
9306
9307	attrs.ia_atime = ncf->ncf_cb_atime;
9308	attrs.ia_mtime = ncf->ncf_cb_mtime;
9309
9310	if (nfsd4_vet_deleg_time(req: &attrs.ia_atime, orig: &dp->dl_atime, now: &now))
9311	attrs.ia_valid |= ATTR_ATIME | ATTR_ATIME_SET;
9312
9313	if (nfsd4_vet_deleg_time(req: &attrs.ia_mtime, orig: &dp->dl_mtime, now: &now)) {
9314	attrs.ia_valid |= ATTR_MTIME | ATTR_MTIME_SET;
9315	attrs.ia_ctime = attrs.ia_mtime;
9316	if (nfsd4_vet_deleg_time(req: &attrs.ia_ctime, orig: &dp->dl_ctime, now: &now))
9317	attrs.ia_valid |= ATTR_CTIME | ATTR_CTIME_SET;
9318	}
9319	} else {
9320	attrs.ia_valid |= ATTR_MTIME | ATTR_CTIME;
9321	}
9322
9323	if (!attrs.ia_valid)
9324	return 0;
9325
9326	attrs.ia_valid |= ATTR_DELEG;
9327	inode_lock(inode);
9328	ret = notify_change(&nop_mnt_idmap, dentry, &attrs, NULL);
9329	inode_unlock(inode);
9330	return ret;
9331	}
9332
9333	/**
9334	* nfsd4_deleg_getattr_conflict - Recall if GETATTR causes conflict
9335	* @rqstp: RPC transaction context
9336	* @dentry: dentry of inode to be checked for a conflict
9337	* @pdp: returned WRITE delegation, if one was found
9338	*
9339	* This function is called when there is a conflict between a write
9340	* delegation and a change/size GETATTR from another client. The server
9341	* must either use the CB_GETATTR to get the current values of the
9342	* attributes from the client that holds the delegation or recall the
9343	* delegation before replying to the GETATTR. See RFC 8881 section
9344	* 18.7.4.
9345	*
9346	* Returns 0 if there is no conflict; otherwise an nfs_stat
9347	* code is returned. If @pdp is set to a non-NULL value, then the
9348	* caller must put the reference.
9349	*/
9350	__be32
9351	nfsd4_deleg_getattr_conflict(struct svc_rqst *rqstp, struct dentry *dentry,
9352	struct nfs4_delegation **pdp)
9353	{
9354	__be32 status;
9355	struct nfsd_net *nn = net_generic(SVC_NET(rqstp), id: nfsd_net_id);
9356	struct file_lock_context *ctx;
9357	struct nfs4_delegation *dp = NULL;
9358	struct file_lease *fl;
9359	struct nfs4_cb_fattr *ncf;
9360	struct inode *inode = d_inode(dentry);
9361
9362	ctx = locks_inode_context(inode);
9363	if (!ctx)
9364	return nfs_ok;
9365
9366	#define NON_NFSD_LEASE ((void *)1)
9367
9368	spin_lock(lock: &ctx->flc_lock);
9369	for_each_file_lock(fl, &ctx->flc_lease) {
9370	if (fl->c.flc_flags == FL_LAYOUT)
9371	continue;
9372	if (fl->c.flc_type == F_WRLCK) {
9373	if (fl->fl_lmops == &nfsd_lease_mng_ops)
9374	dp = fl->c.flc_owner;
9375	else
9376	dp = NON_NFSD_LEASE;
9377	}
9378	break;
9379	}
9380	if (dp == NULL || dp == NON_NFSD_LEASE ||
9381	dp->dl_recall.cb_clp == *(rqstp->rq_lease_breaker)) {
9382	spin_unlock(lock: &ctx->flc_lock);
9383	if (dp == NON_NFSD_LEASE) {
9384	status = nfserrno(errno: nfsd_open_break_lease(inode,
9385	NFSD_MAY_READ));
9386	if (status != nfserr_jukebox ||
9387	!nfsd_wait_for_delegreturn(rqstp, inode))
9388	return status;
9389	}
9390	return 0;
9391	}
9392
9393	nfsd_stats_wdeleg_getattr_inc(nn);
9394	refcount_inc(r: &dp->dl_stid.sc_count);
9395	ncf = &dp->dl_cb_fattr;
9396	nfs4_cb_getattr(ncf: &dp->dl_cb_fattr);
9397	spin_unlock(lock: &ctx->flc_lock);
9398
9399	wait_on_bit_timeout(word: &ncf->ncf_getattr.cb_flags, NFSD4_CALLBACK_RUNNING,
9400	TASK_UNINTERRUPTIBLE, NFSD_CB_GETATTR_TIMEOUT);
9401	if (ncf->ncf_cb_status) {
9402	/* Recall delegation only if client didn't respond */
9403	status = nfserrno(errno: nfsd_open_break_lease(inode, NFSD_MAY_READ));
9404	if (status != nfserr_jukebox ||
9405	!nfsd_wait_for_delegreturn(rqstp, inode))
9406	goto out_status;
9407	}
9408	if (!ncf->ncf_file_modified &&
9409	(ncf->ncf_initial_cinfo != ncf->ncf_cb_change ||
9410	ncf->ncf_cur_fsize != ncf->ncf_cb_fsize))
9411	ncf->ncf_file_modified = true;
9412	if (ncf->ncf_file_modified) {
9413	int err;
9414
9415	/*
9416	* Per section 10.4.3 of RFC 8881, the server would
9417	* not update the file's metadata with the client's
9418	* modified size
9419	*/
9420	err = cb_getattr_update_times(dentry, dp);
9421	if (err) {
9422	status = nfserrno(errno: err);
9423	goto out_status;
9424	}
9425	ncf->ncf_cur_fsize = ncf->ncf_cb_fsize;
9426	*pdp = dp;
9427	return nfs_ok;
9428	}
9429	status = nfs_ok;
9430	out_status:
9431	nfs4_put_stid(s: &dp->dl_stid);
9432	return status;
9433	}
9434
9435	/**
9436	* nfsd_get_dir_deleg - attempt to get a directory delegation
9437	* @cstate: compound state
9438	* @gdd: GET_DIR_DELEGATION arg/resp structure
9439	* @nf: nfsd_file opened on the directory
9440	*
9441	* Given a GET_DIR_DELEGATION request @gdd, attempt to acquire a delegation
9442	* on the directory to which @nf refers. Note that this does not set up any
9443	* sort of async notifications for the delegation.
9444	*/
9445	struct nfs4_delegation *
9446	nfsd_get_dir_deleg(struct nfsd4_compound_state *cstate,
9447	struct nfsd4_get_dir_delegation *gdd,
9448	struct nfsd_file *nf)
9449	{
9450	struct nfs4_client *clp = cstate->clp;
9451	struct nfs4_delegation *dp;
9452	struct file_lease *fl;
9453	struct nfs4_file *fp, *rfp;
9454	int status = 0;
9455
9456	fp = nfsd4_alloc_file();
9457	if (!fp)
9458	return ERR_PTR(error: -ENOMEM);
9459
9460	nfsd4_file_init(fh: &cstate->current_fh, fp);
9461
9462	rfp = nfsd4_file_hash_insert(new: fp, fhp: &cstate->current_fh);
9463	if (unlikely(!rfp)) {
9464	put_nfs4_file(fi: fp);
9465	return ERR_PTR(error: -ENOMEM);
9466	}
9467
9468	if (rfp != fp) {
9469	put_nfs4_file(fi: fp);
9470	fp = rfp;
9471	}
9472
9473	/* if this client already has one, return that it's unavailable */
9474	spin_lock(lock: &state_lock);
9475	spin_lock(lock: &fp->fi_lock);
9476	/* existing delegation? */
9477	if (nfs4_delegation_exists(clp, fp)) {
9478	status = -EAGAIN;
9479	} else if (!fp->fi_deleg_file) {
9480	fp->fi_deleg_file = nfsd_file_get(nf);
9481	fp->fi_delegees = 1;
9482	} else {
9483	++fp->fi_delegees;
9484	}
9485	spin_unlock(lock: &fp->fi_lock);
9486	spin_unlock(lock: &state_lock);
9487
9488	if (status) {
9489	put_nfs4_file(fi: fp);
9490	return ERR_PTR(error: status);
9491	}
9492
9493	/* Try to set up the lease */
9494	status = -ENOMEM;
9495	dp = alloc_init_deleg(clp, fp, NULL, dl_type: NFS4_OPEN_DELEGATE_READ);
9496	if (!dp)
9497	goto out_delegees;
9498
9499	fl = nfs4_alloc_init_lease(dp);
9500	if (!fl)
9501	goto out_put_stid;
9502
9503	status = kernel_setlease(nf->nf_file,
9504	fl->c.flc_type, &fl, NULL);
9505	if (fl)
9506	locks_free_lease(fl);
9507	if (status)
9508	goto out_put_stid;
9509
9510	/*
9511	* Now, try to hash it. This can fail if we race another nfsd task
9512	* trying to set a delegation on the same file. If that happens,
9513	* then just say UNAVAIL.
9514	*/
9515	spin_lock(lock: &state_lock);
9516	spin_lock(lock: &clp->cl_lock);
9517	spin_lock(lock: &fp->fi_lock);
9518	status = hash_delegation_locked(dp, fp);
9519	spin_unlock(lock: &fp->fi_lock);
9520	spin_unlock(lock: &clp->cl_lock);
9521	spin_unlock(lock: &state_lock);
9522
9523	if (!status)
9524	return dp;
9525
9526	/* Something failed. Drop the lease and clean up the stid */
9527	kernel_setlease(fp->fi_deleg_file->nf_file, F_UNLCK, NULL, (void **)&dp);
9528	out_put_stid:
9529	nfs4_put_stid(s: &dp->dl_stid);
9530	out_delegees:
9531	put_deleg_file(fp);
9532	return ERR_PTR(error: status);
9533	}
9534