1/*
2 *  fs/nfs/nfs4proc.c
3 *
4 *  Client-side procedure declarations for NFSv4.
5 *
6 *  Copyright (c) 2002 The Regents of the University of Michigan.
7 *  All rights reserved.
8 *
9 *  Kendrick Smith <kmsmith@umich.edu>
10 *  Andy Adamson   <andros@umich.edu>
11 *
12 *  Redistribution and use in source and binary forms, with or without
13 *  modification, are permitted provided that the following conditions
14 *  are met:
15 *
16 *  1. Redistributions of source code must retain the above copyright
17 *     notice, this list of conditions and the following disclaimer.
18 *  2. Redistributions in binary form must reproduce the above copyright
19 *     notice, this list of conditions and the following disclaimer in the
20 *     documentation and/or other materials provided with the distribution.
21 *  3. Neither the name of the University nor the names of its
22 *     contributors may be used to endorse or promote products derived
23 *     from this software without specific prior written permission.
24 *
25 *  THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED
26 *  WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
27 *  MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
28 *  DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 *  FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 *  CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 *  SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR
32 *  BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
33 *  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
34 *  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
35 *  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
36 */
37
38#include <linux/mm.h>
39#include <linux/delay.h>
40#include <linux/errno.h>
41#include <linux/string.h>
42#include <linux/ratelimit.h>
43#include <linux/printk.h>
44#include <linux/slab.h>
45#include <linux/sunrpc/clnt.h>
46#include <linux/nfs.h>
47#include <linux/nfs4.h>
48#include <linux/nfs_fs.h>
49#include <linux/nfs_page.h>
50#include <linux/nfs_mount.h>
51#include <linux/namei.h>
52#include <linux/mount.h>
53#include <linux/module.h>
54#include <linux/xattr.h>
55#include <linux/utsname.h>
56#include <linux/freezer.h>
57#include <linux/iversion.h>
58
59#include "nfs4_fs.h"
60#include "delegation.h"
61#include "internal.h"
62#include "iostat.h"
63#include "callback.h"
64#include "pnfs.h"
65#include "netns.h"
66#include "sysfs.h"
67#include "nfs4idmap.h"
68#include "nfs4session.h"
69#include "fscache.h"
70#include "nfs42.h"
71
72#include "nfs4trace.h"
73
74#define NFSDBG_FACILITY		NFSDBG_PROC
75
76#define NFS4_BITMASK_SZ		3
77
78#define NFS4_POLL_RETRY_MIN	(HZ/10)
79#define NFS4_POLL_RETRY_MAX	(15*HZ)
80
81/* file attributes which can be mapped to nfs attributes */
82#define NFS4_VALID_ATTRS (ATTR_MODE \
83	| ATTR_UID \
84	| ATTR_GID \
85	| ATTR_SIZE \
86	| ATTR_ATIME \
87	| ATTR_MTIME \
88	| ATTR_CTIME \
89	| ATTR_ATIME_SET \
90	| ATTR_MTIME_SET)
91
92struct nfs4_opendata;
93static int _nfs4_recover_proc_open(struct nfs4_opendata *data);
94static int nfs4_do_fsinfo(struct nfs_server *, struct nfs_fh *, struct nfs_fsinfo *);
95static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr);
96static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
97			      struct nfs_fattr *fattr, struct inode *inode);
98static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
99			    struct nfs_fattr *fattr, struct iattr *sattr,
100			    struct nfs_open_context *ctx, struct nfs4_label *ilabel);
101#ifdef CONFIG_NFS_V4_1
102static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
103		const struct cred *cred,
104		struct nfs4_slot *slot,
105		bool is_privileged);
106static int nfs41_test_stateid(struct nfs_server *, nfs4_stateid *,
107		const struct cred *);
108static int nfs41_free_stateid(struct nfs_server *, const nfs4_stateid *,
109		const struct cred *, bool);
110#endif
111
112#ifdef CONFIG_NFS_V4_SECURITY_LABEL
113static inline struct nfs4_label *
114nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
115	struct iattr *sattr, struct nfs4_label *label)
116{
117	int err;
118
119	if (label == NULL)
120		return NULL;
121
122	if (nfs_server_capable(dir, NFS_CAP_SECURITY_LABEL) == 0)
123		return NULL;
124
125	label->lfs = 0;
126	label->pi = 0;
127	label->len = 0;
128	label->label = NULL;
129
130	err = security_dentry_init_security(dentry, sattr->ia_mode,
131				&dentry->d_name, NULL,
132				(void **)&label->label, &label->len);
133	if (err == 0)
134		return label;
135
136	return NULL;
137}
138static inline void
139nfs4_label_release_security(struct nfs4_label *label)
140{
141	if (label)
142		security_release_secctx(label->label, label->len);
143}
144static inline u32 *nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
145{
146	if (label)
147		return server->attr_bitmask;
148
149	return server->attr_bitmask_nl;
150}
151#else
152static inline struct nfs4_label *
153nfs4_label_init_security(struct inode *dir, struct dentry *dentry,
154	struct iattr *sattr, struct nfs4_label *l)
155{ return NULL; }
156static inline void
157nfs4_label_release_security(struct nfs4_label *label)
158{ return; }
159static inline u32 *
160nfs4_bitmask(struct nfs_server *server, struct nfs4_label *label)
161{ return server->attr_bitmask; }
162#endif
163
164/* Prevent leaks of NFSv4 errors into userland */
165static int nfs4_map_errors(int err)
166{
167	if (err >= -1000)
168		return err;
169	switch (err) {
170	case -NFS4ERR_RESOURCE:
171	case -NFS4ERR_LAYOUTTRYLATER:
172	case -NFS4ERR_RECALLCONFLICT:
173	case -NFS4ERR_RETURNCONFLICT:
174		return -EREMOTEIO;
175	case -NFS4ERR_WRONGSEC:
176	case -NFS4ERR_WRONG_CRED:
177		return -EPERM;
178	case -NFS4ERR_BADOWNER:
179	case -NFS4ERR_BADNAME:
180		return -EINVAL;
181	case -NFS4ERR_SHARE_DENIED:
182		return -EACCES;
183	case -NFS4ERR_MINOR_VERS_MISMATCH:
184		return -EPROTONOSUPPORT;
185	case -NFS4ERR_FILE_OPEN:
186		return -EBUSY;
187	case -NFS4ERR_NOT_SAME:
188		return -ENOTSYNC;
189	default:
190		dprintk("%s could not handle NFSv4 error %d\n",
191				__func__, -err);
192		break;
193	}
194	return -EIO;
195}
196
197/*
198 * This is our standard bitmap for GETATTR requests.
199 */
200const u32 nfs4_fattr_bitmap[3] = {
201	FATTR4_WORD0_TYPE
202	| FATTR4_WORD0_CHANGE
203	| FATTR4_WORD0_SIZE
204	| FATTR4_WORD0_FSID
205	| FATTR4_WORD0_FILEID,
206	FATTR4_WORD1_MODE
207	| FATTR4_WORD1_NUMLINKS
208	| FATTR4_WORD1_OWNER
209	| FATTR4_WORD1_OWNER_GROUP
210	| FATTR4_WORD1_RAWDEV
211	| FATTR4_WORD1_SPACE_USED
212	| FATTR4_WORD1_TIME_ACCESS
213	| FATTR4_WORD1_TIME_METADATA
214	| FATTR4_WORD1_TIME_MODIFY
215	| FATTR4_WORD1_MOUNTED_ON_FILEID,
216#ifdef CONFIG_NFS_V4_SECURITY_LABEL
217	FATTR4_WORD2_SECURITY_LABEL
218#endif
219};
220
221static const u32 nfs4_pnfs_open_bitmap[3] = {
222	FATTR4_WORD0_TYPE
223	| FATTR4_WORD0_CHANGE
224	| FATTR4_WORD0_SIZE
225	| FATTR4_WORD0_FSID
226	| FATTR4_WORD0_FILEID,
227	FATTR4_WORD1_MODE
228	| FATTR4_WORD1_NUMLINKS
229	| FATTR4_WORD1_OWNER
230	| FATTR4_WORD1_OWNER_GROUP
231	| FATTR4_WORD1_RAWDEV
232	| FATTR4_WORD1_SPACE_USED
233	| FATTR4_WORD1_TIME_ACCESS
234	| FATTR4_WORD1_TIME_METADATA
235	| FATTR4_WORD1_TIME_MODIFY,
236	FATTR4_WORD2_MDSTHRESHOLD
237#ifdef CONFIG_NFS_V4_SECURITY_LABEL
238	| FATTR4_WORD2_SECURITY_LABEL
239#endif
240};
241
242static const u32 nfs4_open_noattr_bitmap[3] = {
243	FATTR4_WORD0_TYPE
244	| FATTR4_WORD0_FILEID,
245};
246
247const u32 nfs4_statfs_bitmap[3] = {
248	FATTR4_WORD0_FILES_AVAIL
249	| FATTR4_WORD0_FILES_FREE
250	| FATTR4_WORD0_FILES_TOTAL,
251	FATTR4_WORD1_SPACE_AVAIL
252	| FATTR4_WORD1_SPACE_FREE
253	| FATTR4_WORD1_SPACE_TOTAL
254};
255
256const u32 nfs4_pathconf_bitmap[3] = {
257	FATTR4_WORD0_MAXLINK
258	| FATTR4_WORD0_MAXNAME,
259	0
260};
261
262const u32 nfs4_fsinfo_bitmap[3] = { FATTR4_WORD0_MAXFILESIZE
263			| FATTR4_WORD0_MAXREAD
264			| FATTR4_WORD0_MAXWRITE
265			| FATTR4_WORD0_LEASE_TIME,
266			FATTR4_WORD1_TIME_DELTA
267			| FATTR4_WORD1_FS_LAYOUT_TYPES,
268			FATTR4_WORD2_LAYOUT_BLKSIZE
269			| FATTR4_WORD2_CLONE_BLKSIZE
270			| FATTR4_WORD2_CHANGE_ATTR_TYPE
271			| FATTR4_WORD2_XATTR_SUPPORT
272};
273
274const u32 nfs4_fs_locations_bitmap[3] = {
275	FATTR4_WORD0_CHANGE
276	| FATTR4_WORD0_SIZE
277	| FATTR4_WORD0_FSID
278	| FATTR4_WORD0_FILEID
279	| FATTR4_WORD0_FS_LOCATIONS,
280	FATTR4_WORD1_OWNER
281	| FATTR4_WORD1_OWNER_GROUP
282	| FATTR4_WORD1_RAWDEV
283	| FATTR4_WORD1_SPACE_USED
284	| FATTR4_WORD1_TIME_ACCESS
285	| FATTR4_WORD1_TIME_METADATA
286	| FATTR4_WORD1_TIME_MODIFY
287	| FATTR4_WORD1_MOUNTED_ON_FILEID,
288};
289
290static void nfs4_bitmap_copy_adjust(__u32 *dst, const __u32 *src,
291				    struct inode *inode, unsigned long flags)
292{
293	unsigned long cache_validity;
294
295	memcpy(dst, src, NFS4_BITMASK_SZ*sizeof(*dst));
296	if (!inode || !nfs4_have_delegation(inode, FMODE_READ))
297		return;
298
299	cache_validity = READ_ONCE(NFS_I(inode)->cache_validity) | flags;
300
301	/* Remove the attributes over which we have full control */
302	dst[1] &= ~FATTR4_WORD1_RAWDEV;
303	if (!(cache_validity & NFS_INO_INVALID_SIZE))
304		dst[0] &= ~FATTR4_WORD0_SIZE;
305
306	if (!(cache_validity & NFS_INO_INVALID_CHANGE))
307		dst[0] &= ~FATTR4_WORD0_CHANGE;
308
309	if (!(cache_validity & NFS_INO_INVALID_MODE))
310		dst[1] &= ~FATTR4_WORD1_MODE;
311	if (!(cache_validity & NFS_INO_INVALID_OTHER))
312		dst[1] &= ~(FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP);
313}
314
315static void nfs4_setup_readdir(u64 cookie, __be32 *verifier, struct dentry *dentry,
316		struct nfs4_readdir_arg *readdir)
317{
318	unsigned int attrs = FATTR4_WORD0_FILEID | FATTR4_WORD0_TYPE;
319	__be32 *start, *p;
320
321	if (cookie > 2) {
322		readdir->cookie = cookie;
323		memcpy(&readdir->verifier, verifier, sizeof(readdir->verifier));
324		return;
325	}
326
327	readdir->cookie = 0;
328	memset(&readdir->verifier, 0, sizeof(readdir->verifier));
329	if (cookie == 2)
330		return;
331
332	/*
333	 * NFSv4 servers do not return entries for '.' and '..'
334	 * Therefore, we fake these entries here.  We let '.'
335	 * have cookie 0 and '..' have cookie 1.  Note that
336	 * when talking to the server, we always send cookie 0
337	 * instead of 1 or 2.
338	 */
339	start = p = kmap_atomic(*readdir->pages);
340
341	if (cookie == 0) {
342		*p++ = xdr_one;                                  /* next */
343		*p++ = xdr_zero;                   /* cookie, first word */
344		*p++ = xdr_one;                   /* cookie, second word */
345		*p++ = xdr_one;                             /* entry len */
346		memcpy(p, ".\0\0\0", 4);                        /* entry */
347		p++;
348		*p++ = xdr_one;                         /* bitmap length */
349		*p++ = htonl(attrs);                           /* bitmap */
350		*p++ = htonl(12);             /* attribute buffer length */
351		*p++ = htonl(NF4DIR);
352		p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry)));
353	}
354
355	*p++ = xdr_one;                                  /* next */
356	*p++ = xdr_zero;                   /* cookie, first word */
357	*p++ = xdr_two;                   /* cookie, second word */
358	*p++ = xdr_two;                             /* entry len */
359	memcpy(p, "..\0\0", 4);                         /* entry */
360	p++;
361	*p++ = xdr_one;                         /* bitmap length */
362	*p++ = htonl(attrs);                           /* bitmap */
363	*p++ = htonl(12);             /* attribute buffer length */
364	*p++ = htonl(NF4DIR);
365	p = xdr_encode_hyper(p, NFS_FILEID(d_inode(dentry->d_parent)));
366
367	readdir->pgbase = (char *)p - (char *)start;
368	readdir->count -= readdir->pgbase;
369	kunmap_atomic(start);
370}
371
372static void nfs4_fattr_set_prechange(struct nfs_fattr *fattr, u64 version)
373{
374	if (!(fattr->valid & NFS_ATTR_FATTR_PRECHANGE)) {
375		fattr->pre_change_attr = version;
376		fattr->valid |= NFS_ATTR_FATTR_PRECHANGE;
377	}
378}
379
380static void nfs4_test_and_free_stateid(struct nfs_server *server,
381		nfs4_stateid *stateid,
382		const struct cred *cred)
383{
384	const struct nfs4_minor_version_ops *ops = server->nfs_client->cl_mvops;
385
386	ops->test_and_free_expired(server, stateid, cred);
387}
388
389static void __nfs4_free_revoked_stateid(struct nfs_server *server,
390		nfs4_stateid *stateid,
391		const struct cred *cred)
392{
393	stateid->type = NFS4_REVOKED_STATEID_TYPE;
394	nfs4_test_and_free_stateid(server, stateid, cred);
395}
396
397static void nfs4_free_revoked_stateid(struct nfs_server *server,
398		const nfs4_stateid *stateid,
399		const struct cred *cred)
400{
401	nfs4_stateid tmp;
402
403	nfs4_stateid_copy(&tmp, stateid);
404	__nfs4_free_revoked_stateid(server, &tmp, cred);
405}
406
407static long nfs4_update_delay(long *timeout)
408{
409	long ret;
410	if (!timeout)
411		return NFS4_POLL_RETRY_MAX;
412	if (*timeout <= 0)
413		*timeout = NFS4_POLL_RETRY_MIN;
414	if (*timeout > NFS4_POLL_RETRY_MAX)
415		*timeout = NFS4_POLL_RETRY_MAX;
416	ret = *timeout;
417	*timeout <<= 1;
418	return ret;
419}
420
421static int nfs4_delay_killable(long *timeout)
422{
423	might_sleep();
424
425	__set_current_state(TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
426	schedule_timeout(nfs4_update_delay(timeout));
427	if (!__fatal_signal_pending(current))
428		return 0;
429	return -EINTR;
430}
431
432static int nfs4_delay_interruptible(long *timeout)
433{
434	might_sleep();
435
436	__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE_UNSAFE);
437	schedule_timeout(nfs4_update_delay(timeout));
438	if (!signal_pending(current))
439		return 0;
440	return __fatal_signal_pending(current) ? -EINTR :-ERESTARTSYS;
441}
442
443static int nfs4_delay(long *timeout, bool interruptible)
444{
445	if (interruptible)
446		return nfs4_delay_interruptible(timeout);
447	return nfs4_delay_killable(timeout);
448}
449
450static const nfs4_stateid *
451nfs4_recoverable_stateid(const nfs4_stateid *stateid)
452{
453	if (!stateid)
454		return NULL;
455	switch (stateid->type) {
456	case NFS4_OPEN_STATEID_TYPE:
457	case NFS4_LOCK_STATEID_TYPE:
458	case NFS4_DELEGATION_STATEID_TYPE:
459		return stateid;
460	default:
461		break;
462	}
463	return NULL;
464}
465
466/* This is the error handling routine for processes that are allowed
467 * to sleep.
468 */
469static int nfs4_do_handle_exception(struct nfs_server *server,
470		int errorcode, struct nfs4_exception *exception)
471{
472	struct nfs_client *clp = server->nfs_client;
473	struct nfs4_state *state = exception->state;
474	const nfs4_stateid *stateid;
475	struct inode *inode = exception->inode;
476	int ret = errorcode;
477
478	exception->delay = 0;
479	exception->recovering = 0;
480	exception->retry = 0;
481
482	stateid = nfs4_recoverable_stateid(exception->stateid);
483	if (stateid == NULL && state != NULL)
484		stateid = nfs4_recoverable_stateid(&state->stateid);
485
486	switch(errorcode) {
487		case 0:
488			return 0;
489		case -NFS4ERR_BADHANDLE:
490		case -ESTALE:
491			if (inode != NULL && S_ISREG(inode->i_mode))
492				pnfs_destroy_layout(NFS_I(inode));
493			break;
494		case -NFS4ERR_DELEG_REVOKED:
495		case -NFS4ERR_ADMIN_REVOKED:
496		case -NFS4ERR_EXPIRED:
497		case -NFS4ERR_BAD_STATEID:
498		case -NFS4ERR_PARTNER_NO_AUTH:
499			if (inode != NULL && stateid != NULL) {
500				nfs_inode_find_state_and_recover(inode,
501						stateid);
502				goto wait_on_recovery;
503			}
504			fallthrough;
505		case -NFS4ERR_OPENMODE:
506			if (inode) {
507				int err;
508
509				err = nfs_async_inode_return_delegation(inode,
510						stateid);
511				if (err == 0)
512					goto wait_on_recovery;
513				if (stateid != NULL && stateid->type == NFS4_DELEGATION_STATEID_TYPE) {
514					exception->retry = 1;
515					break;
516				}
517			}
518			if (state == NULL)
519				break;
520			ret = nfs4_schedule_stateid_recovery(server, state);
521			if (ret < 0)
522				break;
523			goto wait_on_recovery;
524		case -NFS4ERR_STALE_STATEID:
525		case -NFS4ERR_STALE_CLIENTID:
526			nfs4_schedule_lease_recovery(clp);
527			goto wait_on_recovery;
528		case -NFS4ERR_MOVED:
529			ret = nfs4_schedule_migration_recovery(server);
530			if (ret < 0)
531				break;
532			goto wait_on_recovery;
533		case -NFS4ERR_LEASE_MOVED:
534			nfs4_schedule_lease_moved_recovery(clp);
535			goto wait_on_recovery;
536#if defined(CONFIG_NFS_V4_1)
537		case -NFS4ERR_BADSESSION:
538		case -NFS4ERR_BADSLOT:
539		case -NFS4ERR_BAD_HIGH_SLOT:
540		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
541		case -NFS4ERR_DEADSESSION:
542		case -NFS4ERR_SEQ_FALSE_RETRY:
543		case -NFS4ERR_SEQ_MISORDERED:
544			/* Handled in nfs41_sequence_process() */
545			goto wait_on_recovery;
546#endif /* defined(CONFIG_NFS_V4_1) */
547		case -NFS4ERR_FILE_OPEN:
548			if (exception->timeout > HZ) {
549				/* We have retried a decent amount, time to
550				 * fail
551				 */
552				ret = -EBUSY;
553				break;
554			}
555			fallthrough;
556		case -NFS4ERR_DELAY:
557			nfs_inc_server_stats(server, NFSIOS_DELAY);
558			fallthrough;
559		case -NFS4ERR_GRACE:
560		case -NFS4ERR_LAYOUTTRYLATER:
561		case -NFS4ERR_RECALLCONFLICT:
562		case -NFS4ERR_RETURNCONFLICT:
563			exception->delay = 1;
564			return 0;
565
566		case -NFS4ERR_RETRY_UNCACHED_REP:
567		case -NFS4ERR_OLD_STATEID:
568			exception->retry = 1;
569			break;
570		case -NFS4ERR_BADOWNER:
571			/* The following works around a Linux server bug! */
572		case -NFS4ERR_BADNAME:
573			if (server->caps & NFS_CAP_UIDGID_NOMAP) {
574				server->caps &= ~NFS_CAP_UIDGID_NOMAP;
575				exception->retry = 1;
576				printk(KERN_WARNING "NFS: v4 server %s "
577						"does not accept raw "
578						"uid/gids. "
579						"Reenabling the idmapper.\n",
580						server->nfs_client->cl_hostname);
581			}
582	}
583	/* We failed to handle the error */
584	return nfs4_map_errors(ret);
585wait_on_recovery:
586	exception->recovering = 1;
587	return 0;
588}
589
590/*
591 * Track the number of NFS4ERR_DELAY related retransmissions and return
592 * EAGAIN if the 'softerr' mount option is set, and we've exceeded the limit
593 * set by 'nfs_delay_retrans'.
594 */
595static int nfs4_exception_should_retrans(const struct nfs_server *server,
596					 struct nfs4_exception *exception)
597{
598	if (server->flags & NFS_MOUNT_SOFTERR && nfs_delay_retrans >= 0) {
599		if (exception->retrans++ >= (unsigned short)nfs_delay_retrans)
600			return -EAGAIN;
601	}
602	return 0;
603}
604
605/* This is the error handling routine for processes that are allowed
606 * to sleep.
607 */
608int nfs4_handle_exception(struct nfs_server *server, int errorcode, struct nfs4_exception *exception)
609{
610	struct nfs_client *clp = server->nfs_client;
611	int ret;
612
613	ret = nfs4_do_handle_exception(server, errorcode, exception);
614	if (exception->delay) {
615		int ret2 = nfs4_exception_should_retrans(server, exception);
616		if (ret2 < 0) {
617			exception->retry = 0;
618			return ret2;
619		}
620		ret = nfs4_delay(&exception->timeout,
621				exception->interruptible);
622		goto out_retry;
623	}
624	if (exception->recovering) {
625		if (exception->task_is_privileged)
626			return -EDEADLOCK;
627		ret = nfs4_wait_clnt_recover(clp);
628		if (test_bit(NFS_MIG_FAILED, &server->mig_status))
629			return -EIO;
630		goto out_retry;
631	}
632	return ret;
633out_retry:
634	if (ret == 0)
635		exception->retry = 1;
636	return ret;
637}
638
639static int
640nfs4_async_handle_exception(struct rpc_task *task, struct nfs_server *server,
641		int errorcode, struct nfs4_exception *exception)
642{
643	struct nfs_client *clp = server->nfs_client;
644	int ret;
645
646	ret = nfs4_do_handle_exception(server, errorcode, exception);
647	if (exception->delay) {
648		int ret2 = nfs4_exception_should_retrans(server, exception);
649		if (ret2 < 0) {
650			exception->retry = 0;
651			return ret2;
652		}
653		rpc_delay(task, nfs4_update_delay(&exception->timeout));
654		goto out_retry;
655	}
656	if (exception->recovering) {
657		if (exception->task_is_privileged)
658			return -EDEADLOCK;
659		rpc_sleep_on(&clp->cl_rpcwaitq, task, NULL);
660		if (test_bit(NFS4CLNT_MANAGER_RUNNING, &clp->cl_state) == 0)
661			rpc_wake_up_queued_task(&clp->cl_rpcwaitq, task);
662		goto out_retry;
663	}
664	if (test_bit(NFS_MIG_FAILED, &server->mig_status))
665		ret = -EIO;
666	return ret;
667out_retry:
668	if (ret == 0) {
669		exception->retry = 1;
670		/*
671		 * For NFS4ERR_MOVED, the client transport will need to
672		 * be recomputed after migration recovery has completed.
673		 */
674		if (errorcode == -NFS4ERR_MOVED)
675			rpc_task_release_transport(task);
676	}
677	return ret;
678}
679
680int
681nfs4_async_handle_error(struct rpc_task *task, struct nfs_server *server,
682			struct nfs4_state *state, long *timeout)
683{
684	struct nfs4_exception exception = {
685		.state = state,
686	};
687
688	if (task->tk_status >= 0)
689		return 0;
690	if (timeout)
691		exception.timeout = *timeout;
692	task->tk_status = nfs4_async_handle_exception(task, server,
693			task->tk_status,
694			&exception);
695	if (exception.delay && timeout)
696		*timeout = exception.timeout;
697	if (exception.retry)
698		return -EAGAIN;
699	return 0;
700}
701
702/*
703 * Return 'true' if 'clp' is using an rpc_client that is integrity protected
704 * or 'false' otherwise.
705 */
706static bool _nfs4_is_integrity_protected(struct nfs_client *clp)
707{
708	rpc_authflavor_t flavor = clp->cl_rpcclient->cl_auth->au_flavor;
709	return (flavor == RPC_AUTH_GSS_KRB5I) || (flavor == RPC_AUTH_GSS_KRB5P);
710}
711
712static void do_renew_lease(struct nfs_client *clp, unsigned long timestamp)
713{
714	spin_lock(&clp->cl_lock);
715	if (time_before(clp->cl_last_renewal,timestamp))
716		clp->cl_last_renewal = timestamp;
717	spin_unlock(&clp->cl_lock);
718}
719
720static void renew_lease(const struct nfs_server *server, unsigned long timestamp)
721{
722	struct nfs_client *clp = server->nfs_client;
723
724	if (!nfs4_has_session(clp))
725		do_renew_lease(clp, timestamp);
726}
727
728struct nfs4_call_sync_data {
729	const struct nfs_server *seq_server;
730	struct nfs4_sequence_args *seq_args;
731	struct nfs4_sequence_res *seq_res;
732};
733
734void nfs4_init_sequence(struct nfs4_sequence_args *args,
735			struct nfs4_sequence_res *res, int cache_reply,
736			int privileged)
737{
738	args->sa_slot = NULL;
739	args->sa_cache_this = cache_reply;
740	args->sa_privileged = privileged;
741
742	res->sr_slot = NULL;
743}
744
745static void nfs40_sequence_free_slot(struct nfs4_sequence_res *res)
746{
747	struct nfs4_slot *slot = res->sr_slot;
748	struct nfs4_slot_table *tbl;
749
750	tbl = slot->table;
751	spin_lock(&tbl->slot_tbl_lock);
752	if (!nfs41_wake_and_assign_slot(tbl, slot))
753		nfs4_free_slot(tbl, slot);
754	spin_unlock(&tbl->slot_tbl_lock);
755
756	res->sr_slot = NULL;
757}
758
759static int nfs40_sequence_done(struct rpc_task *task,
760			       struct nfs4_sequence_res *res)
761{
762	if (res->sr_slot != NULL)
763		nfs40_sequence_free_slot(res);
764	return 1;
765}
766
767#if defined(CONFIG_NFS_V4_1)
768
769static void nfs41_release_slot(struct nfs4_slot *slot)
770{
771	struct nfs4_session *session;
772	struct nfs4_slot_table *tbl;
773	bool send_new_highest_used_slotid = false;
774
775	if (!slot)
776		return;
777	tbl = slot->table;
778	session = tbl->session;
779
780	/* Bump the slot sequence number */
781	if (slot->seq_done)
782		slot->seq_nr++;
783	slot->seq_done = 0;
784
785	spin_lock(&tbl->slot_tbl_lock);
786	/* Be nice to the server: try to ensure that the last transmitted
787	 * value for highest_user_slotid <= target_highest_slotid
788	 */
789	if (tbl->highest_used_slotid > tbl->target_highest_slotid)
790		send_new_highest_used_slotid = true;
791
792	if (nfs41_wake_and_assign_slot(tbl, slot)) {
793		send_new_highest_used_slotid = false;
794		goto out_unlock;
795	}
796	nfs4_free_slot(tbl, slot);
797
798	if (tbl->highest_used_slotid != NFS4_NO_SLOT)
799		send_new_highest_used_slotid = false;
800out_unlock:
801	spin_unlock(&tbl->slot_tbl_lock);
802	if (send_new_highest_used_slotid)
803		nfs41_notify_server(session->clp);
804	if (waitqueue_active(&tbl->slot_waitq))
805		wake_up_all(&tbl->slot_waitq);
806}
807
808static void nfs41_sequence_free_slot(struct nfs4_sequence_res *res)
809{
810	nfs41_release_slot(res->sr_slot);
811	res->sr_slot = NULL;
812}
813
814static void nfs4_slot_sequence_record_sent(struct nfs4_slot *slot,
815		u32 seqnr)
816{
817	if ((s32)(seqnr - slot->seq_nr_highest_sent) > 0)
818		slot->seq_nr_highest_sent = seqnr;
819}
820static void nfs4_slot_sequence_acked(struct nfs4_slot *slot, u32 seqnr)
821{
822	nfs4_slot_sequence_record_sent(slot, seqnr);
823	slot->seq_nr_last_acked = seqnr;
824}
825
826static void nfs4_probe_sequence(struct nfs_client *client, const struct cred *cred,
827				struct nfs4_slot *slot)
828{
829	struct rpc_task *task = _nfs41_proc_sequence(client, cred, slot, true);
830	if (!IS_ERR(task))
831		rpc_put_task_async(task);
832}
833
834static int nfs41_sequence_process(struct rpc_task *task,
835		struct nfs4_sequence_res *res)
836{
837	struct nfs4_session *session;
838	struct nfs4_slot *slot = res->sr_slot;
839	struct nfs_client *clp;
840	int status;
841	int ret = 1;
842
843	if (slot == NULL)
844		goto out_noaction;
845	/* don't increment the sequence number if the task wasn't sent */
846	if (!RPC_WAS_SENT(task) || slot->seq_done)
847		goto out;
848
849	session = slot->table->session;
850	clp = session->clp;
851
852	trace_nfs4_sequence_done(session, res);
853
854	status = res->sr_status;
855	if (task->tk_status == -NFS4ERR_DEADSESSION)
856		status = -NFS4ERR_DEADSESSION;
857
858	/* Check the SEQUENCE operation status */
859	switch (status) {
860	case 0:
861		/* Mark this sequence number as having been acked */
862		nfs4_slot_sequence_acked(slot, slot->seq_nr);
863		/* Update the slot's sequence and clientid lease timer */
864		slot->seq_done = 1;
865		do_renew_lease(clp, res->sr_timestamp);
866		/* Check sequence flags */
867		nfs41_handle_sequence_flag_errors(clp, res->sr_status_flags,
868				!!slot->privileged);
869		nfs41_update_target_slotid(slot->table, slot, res);
870		break;
871	case 1:
872		/*
873		 * sr_status remains 1 if an RPC level error occurred.
874		 * The server may or may not have processed the sequence
875		 * operation..
876		 */
877		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
878		slot->seq_done = 1;
879		goto out;
880	case -NFS4ERR_DELAY:
881		/* The server detected a resend of the RPC call and
882		 * returned NFS4ERR_DELAY as per Section 2.10.6.2
883		 * of RFC5661.
884		 */
885		dprintk("%s: slot=%u seq=%u: Operation in progress\n",
886			__func__,
887			slot->slot_nr,
888			slot->seq_nr);
889		goto out_retry;
890	case -NFS4ERR_RETRY_UNCACHED_REP:
891	case -NFS4ERR_SEQ_FALSE_RETRY:
892		/*
893		 * The server thinks we tried to replay a request.
894		 * Retry the call after bumping the sequence ID.
895		 */
896		nfs4_slot_sequence_acked(slot, slot->seq_nr);
897		goto retry_new_seq;
898	case -NFS4ERR_BADSLOT:
899		/*
900		 * The slot id we used was probably retired. Try again
901		 * using a different slot id.
902		 */
903		if (slot->slot_nr < slot->table->target_highest_slotid)
904			goto session_recover;
905		goto retry_nowait;
906	case -NFS4ERR_SEQ_MISORDERED:
907		nfs4_slot_sequence_record_sent(slot, slot->seq_nr);
908		/*
909		 * Were one or more calls using this slot interrupted?
910		 * If the server never received the request, then our
911		 * transmitted slot sequence number may be too high. However,
912		 * if the server did receive the request then it might
913		 * accidentally give us a reply with a mismatched operation.
914		 * We can sort this out by sending a lone sequence operation
915		 * to the server on the same slot.
916		 */
917		if ((s32)(slot->seq_nr - slot->seq_nr_last_acked) > 1) {
918			slot->seq_nr--;
919			if (task->tk_msg.rpc_proc != &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE]) {
920				nfs4_probe_sequence(clp, task->tk_msg.rpc_cred, slot);
921				res->sr_slot = NULL;
922			}
923			goto retry_nowait;
924		}
925		/*
926		 * RFC5661:
927		 * A retry might be sent while the original request is
928		 * still in progress on the replier. The replier SHOULD
929		 * deal with the issue by returning NFS4ERR_DELAY as the
930		 * reply to SEQUENCE or CB_SEQUENCE operation, but
931		 * implementations MAY return NFS4ERR_SEQ_MISORDERED.
932		 *
933		 * Restart the search after a delay.
934		 */
935		slot->seq_nr = slot->seq_nr_highest_sent;
936		goto out_retry;
937	case -NFS4ERR_BADSESSION:
938	case -NFS4ERR_DEADSESSION:
939	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
940		goto session_recover;
941	default:
942		/* Just update the slot sequence no. */
943		slot->seq_done = 1;
944	}
945out:
946	/* The session may be reset by one of the error handlers. */
947	dprintk("%s: Error %d free the slot \n", __func__, res->sr_status);
948out_noaction:
949	return ret;
950session_recover:
951	set_bit(NFS4_SLOT_TBL_DRAINING, &session->fc_slot_table.slot_tbl_state);
952	nfs4_schedule_session_recovery(session, status);
953	dprintk("%s ERROR: %d Reset session\n", __func__, status);
954	nfs41_sequence_free_slot(res);
955	goto out;
956retry_new_seq:
957	++slot->seq_nr;
958retry_nowait:
959	if (rpc_restart_call_prepare(task)) {
960		nfs41_sequence_free_slot(res);
961		task->tk_status = 0;
962		ret = 0;
963	}
964	goto out;
965out_retry:
966	if (!rpc_restart_call(task))
967		goto out;
968	rpc_delay(task, NFS4_POLL_RETRY_MAX);
969	return 0;
970}
971
972int nfs41_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
973{
974	if (!nfs41_sequence_process(task, res))
975		return 0;
976	if (res->sr_slot != NULL)
977		nfs41_sequence_free_slot(res);
978	return 1;
979
980}
981EXPORT_SYMBOL_GPL(nfs41_sequence_done);
982
983static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
984{
985	if (res->sr_slot == NULL)
986		return 1;
987	if (res->sr_slot->table->session != NULL)
988		return nfs41_sequence_process(task, res);
989	return nfs40_sequence_done(task, res);
990}
991
992static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
993{
994	if (res->sr_slot != NULL) {
995		if (res->sr_slot->table->session != NULL)
996			nfs41_sequence_free_slot(res);
997		else
998			nfs40_sequence_free_slot(res);
999	}
1000}
1001
1002int nfs4_sequence_done(struct rpc_task *task, struct nfs4_sequence_res *res)
1003{
1004	if (res->sr_slot == NULL)
1005		return 1;
1006	if (!res->sr_slot->table->session)
1007		return nfs40_sequence_done(task, res);
1008	return nfs41_sequence_done(task, res);
1009}
1010EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1011
1012static void nfs41_call_sync_prepare(struct rpc_task *task, void *calldata)
1013{
1014	struct nfs4_call_sync_data *data = calldata;
1015
1016	dprintk("--> %s data->seq_server %p\n", __func__, data->seq_server);
1017
1018	nfs4_setup_sequence(data->seq_server->nfs_client,
1019			    data->seq_args, data->seq_res, task);
1020}
1021
1022static void nfs41_call_sync_done(struct rpc_task *task, void *calldata)
1023{
1024	struct nfs4_call_sync_data *data = calldata;
1025
1026	nfs41_sequence_done(task, data->seq_res);
1027}
1028
1029static const struct rpc_call_ops nfs41_call_sync_ops = {
1030	.rpc_call_prepare = nfs41_call_sync_prepare,
1031	.rpc_call_done = nfs41_call_sync_done,
1032};
1033
1034#else	/* !CONFIG_NFS_V4_1 */
1035
1036static int nfs4_sequence_process(struct rpc_task *task, struct nfs4_sequence_res *res)
1037{
1038	return nfs40_sequence_done(task, res);
1039}
1040
1041static void nfs4_sequence_free_slot(struct nfs4_sequence_res *res)
1042{
1043	if (res->sr_slot != NULL)
1044		nfs40_sequence_free_slot(res);
1045}
1046
1047int nfs4_sequence_done(struct rpc_task *task,
1048		       struct nfs4_sequence_res *res)
1049{
1050	return nfs40_sequence_done(task, res);
1051}
1052EXPORT_SYMBOL_GPL(nfs4_sequence_done);
1053
1054#endif	/* !CONFIG_NFS_V4_1 */
1055
1056static void nfs41_sequence_res_init(struct nfs4_sequence_res *res)
1057{
1058	res->sr_timestamp = jiffies;
1059	res->sr_status_flags = 0;
1060	res->sr_status = 1;
1061}
1062
1063static
1064void nfs4_sequence_attach_slot(struct nfs4_sequence_args *args,
1065		struct nfs4_sequence_res *res,
1066		struct nfs4_slot *slot)
1067{
1068	if (!slot)
1069		return;
1070	slot->privileged = args->sa_privileged ? 1 : 0;
1071	args->sa_slot = slot;
1072
1073	res->sr_slot = slot;
1074}
1075
1076int nfs4_setup_sequence(struct nfs_client *client,
1077			struct nfs4_sequence_args *args,
1078			struct nfs4_sequence_res *res,
1079			struct rpc_task *task)
1080{
1081	struct nfs4_session *session = nfs4_get_session(client);
1082	struct nfs4_slot_table *tbl  = client->cl_slot_tbl;
1083	struct nfs4_slot *slot;
1084
1085	/* slot already allocated? */
1086	if (res->sr_slot != NULL)
1087		goto out_start;
1088
1089	if (session)
1090		tbl = &session->fc_slot_table;
1091
1092	spin_lock(&tbl->slot_tbl_lock);
1093	/* The state manager will wait until the slot table is empty */
1094	if (nfs4_slot_tbl_draining(tbl) && !args->sa_privileged)
1095		goto out_sleep;
1096
1097	slot = nfs4_alloc_slot(tbl);
1098	if (IS_ERR(slot)) {
1099		if (slot == ERR_PTR(-ENOMEM))
1100			goto out_sleep_timeout;
1101		goto out_sleep;
1102	}
1103	spin_unlock(&tbl->slot_tbl_lock);
1104
1105	nfs4_sequence_attach_slot(args, res, slot);
1106
1107	trace_nfs4_setup_sequence(session, args);
1108out_start:
1109	nfs41_sequence_res_init(res);
1110	rpc_call_start(task);
1111	return 0;
1112out_sleep_timeout:
1113	/* Try again in 1/4 second */
1114	if (args->sa_privileged)
1115		rpc_sleep_on_priority_timeout(&tbl->slot_tbl_waitq, task,
1116				jiffies + (HZ >> 2), RPC_PRIORITY_PRIVILEGED);
1117	else
1118		rpc_sleep_on_timeout(&tbl->slot_tbl_waitq, task,
1119				NULL, jiffies + (HZ >> 2));
1120	spin_unlock(&tbl->slot_tbl_lock);
1121	return -EAGAIN;
1122out_sleep:
1123	if (args->sa_privileged)
1124		rpc_sleep_on_priority(&tbl->slot_tbl_waitq, task,
1125				RPC_PRIORITY_PRIVILEGED);
1126	else
1127		rpc_sleep_on(&tbl->slot_tbl_waitq, task, NULL);
1128	spin_unlock(&tbl->slot_tbl_lock);
1129	return -EAGAIN;
1130}
1131EXPORT_SYMBOL_GPL(nfs4_setup_sequence);
1132
1133static void nfs40_call_sync_prepare(struct rpc_task *task, void *calldata)
1134{
1135	struct nfs4_call_sync_data *data = calldata;
1136	nfs4_setup_sequence(data->seq_server->nfs_client,
1137				data->seq_args, data->seq_res, task);
1138}
1139
1140static void nfs40_call_sync_done(struct rpc_task *task, void *calldata)
1141{
1142	struct nfs4_call_sync_data *data = calldata;
1143	nfs4_sequence_done(task, data->seq_res);
1144}
1145
1146static const struct rpc_call_ops nfs40_call_sync_ops = {
1147	.rpc_call_prepare = nfs40_call_sync_prepare,
1148	.rpc_call_done = nfs40_call_sync_done,
1149};
1150
1151static int nfs4_call_sync_custom(struct rpc_task_setup *task_setup)
1152{
1153	int ret;
1154	struct rpc_task *task;
1155
1156	task = rpc_run_task(task_setup);
1157	if (IS_ERR(task))
1158		return PTR_ERR(task);
1159
1160	ret = task->tk_status;
1161	rpc_put_task(task);
1162	return ret;
1163}
1164
1165static int nfs4_do_call_sync(struct rpc_clnt *clnt,
1166			     struct nfs_server *server,
1167			     struct rpc_message *msg,
1168			     struct nfs4_sequence_args *args,
1169			     struct nfs4_sequence_res *res,
1170			     unsigned short task_flags)
1171{
1172	struct nfs_client *clp = server->nfs_client;
1173	struct nfs4_call_sync_data data = {
1174		.seq_server = server,
1175		.seq_args = args,
1176		.seq_res = res,
1177	};
1178	struct rpc_task_setup task_setup = {
1179		.rpc_client = clnt,
1180		.rpc_message = msg,
1181		.callback_ops = clp->cl_mvops->call_sync_ops,
1182		.callback_data = &data,
1183		.flags = task_flags,
1184	};
1185
1186	return nfs4_call_sync_custom(&task_setup);
1187}
1188
1189static int nfs4_call_sync_sequence(struct rpc_clnt *clnt,
1190				   struct nfs_server *server,
1191				   struct rpc_message *msg,
1192				   struct nfs4_sequence_args *args,
1193				   struct nfs4_sequence_res *res)
1194{
1195	unsigned short task_flags = 0;
1196
1197	if (server->caps & NFS_CAP_MOVEABLE)
1198		task_flags = RPC_TASK_MOVEABLE;
1199	return nfs4_do_call_sync(clnt, server, msg, args, res, task_flags);
1200}
1201
1202
1203int nfs4_call_sync(struct rpc_clnt *clnt,
1204		   struct nfs_server *server,
1205		   struct rpc_message *msg,
1206		   struct nfs4_sequence_args *args,
1207		   struct nfs4_sequence_res *res,
1208		   int cache_reply)
1209{
1210	nfs4_init_sequence(args, res, cache_reply, 0);
1211	return nfs4_call_sync_sequence(clnt, server, msg, args, res);
1212}
1213
1214static void
1215nfs4_inc_nlink_locked(struct inode *inode)
1216{
1217	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1218					     NFS_INO_INVALID_CTIME |
1219					     NFS_INO_INVALID_NLINK);
1220	inc_nlink(inode);
1221}
1222
1223static void
1224nfs4_inc_nlink(struct inode *inode)
1225{
1226	spin_lock(&inode->i_lock);
1227	nfs4_inc_nlink_locked(inode);
1228	spin_unlock(&inode->i_lock);
1229}
1230
1231static void
1232nfs4_dec_nlink_locked(struct inode *inode)
1233{
1234	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
1235					     NFS_INO_INVALID_CTIME |
1236					     NFS_INO_INVALID_NLINK);
1237	drop_nlink(inode);
1238}
1239
1240static void
1241nfs4_update_changeattr_locked(struct inode *inode,
1242		struct nfs4_change_info *cinfo,
1243		unsigned long timestamp, unsigned long cache_validity)
1244{
1245	struct nfs_inode *nfsi = NFS_I(inode);
1246	u64 change_attr = inode_peek_iversion_raw(inode);
1247
1248	cache_validity |= NFS_INO_INVALID_CTIME | NFS_INO_INVALID_MTIME;
1249	if (S_ISDIR(inode->i_mode))
1250		cache_validity |= NFS_INO_INVALID_DATA;
1251
1252	switch (NFS_SERVER(inode)->change_attr_type) {
1253	case NFS4_CHANGE_TYPE_IS_UNDEFINED:
1254		if (cinfo->after == change_attr)
1255			goto out;
1256		break;
1257	default:
1258		if ((s64)(change_attr - cinfo->after) >= 0)
1259			goto out;
1260	}
1261
1262	inode_set_iversion_raw(inode, cinfo->after);
1263	if (!cinfo->atomic || cinfo->before != change_attr) {
1264		if (S_ISDIR(inode->i_mode))
1265			nfs_force_lookup_revalidate(inode);
1266
1267		if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
1268			cache_validity |=
1269				NFS_INO_INVALID_ACCESS | NFS_INO_INVALID_ACL |
1270				NFS_INO_INVALID_SIZE | NFS_INO_INVALID_OTHER |
1271				NFS_INO_INVALID_BLOCKS | NFS_INO_INVALID_NLINK |
1272				NFS_INO_INVALID_MODE | NFS_INO_INVALID_XATTR;
1273		nfsi->attrtimeo = NFS_MINATTRTIMEO(inode);
1274	}
1275	nfsi->attrtimeo_timestamp = jiffies;
1276	nfsi->read_cache_jiffies = timestamp;
1277	nfsi->attr_gencount = nfs_inc_attr_generation_counter();
1278	nfsi->cache_validity &= ~NFS_INO_INVALID_CHANGE;
1279out:
1280	nfs_set_cache_invalid(inode, cache_validity);
1281}
1282
1283void
1284nfs4_update_changeattr(struct inode *dir, struct nfs4_change_info *cinfo,
1285		unsigned long timestamp, unsigned long cache_validity)
1286{
1287	spin_lock(&dir->i_lock);
1288	nfs4_update_changeattr_locked(dir, cinfo, timestamp, cache_validity);
1289	spin_unlock(&dir->i_lock);
1290}
1291
1292struct nfs4_open_createattrs {
1293	struct nfs4_label *label;
1294	struct iattr *sattr;
1295	const __u32 verf[2];
1296};
1297
1298static bool nfs4_clear_cap_atomic_open_v1(struct nfs_server *server,
1299		int err, struct nfs4_exception *exception)
1300{
1301	if (err != -EINVAL)
1302		return false;
1303	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1304		return false;
1305	server->caps &= ~NFS_CAP_ATOMIC_OPEN_V1;
1306	exception->retry = 1;
1307	return true;
1308}
1309
1310static fmode_t _nfs4_ctx_to_accessmode(const struct nfs_open_context *ctx)
1311{
1312	 return ctx->mode & (FMODE_READ|FMODE_WRITE|FMODE_EXEC);
1313}
1314
1315static fmode_t _nfs4_ctx_to_openmode(const struct nfs_open_context *ctx)
1316{
1317	fmode_t ret = ctx->mode & (FMODE_READ|FMODE_WRITE);
1318
1319	return (ctx->mode & FMODE_EXEC) ? FMODE_READ | ret : ret;
1320}
1321
1322static u32
1323nfs4_map_atomic_open_share(struct nfs_server *server,
1324		fmode_t fmode, int openflags)
1325{
1326	u32 res = 0;
1327
1328	switch (fmode & (FMODE_READ | FMODE_WRITE)) {
1329	case FMODE_READ:
1330		res = NFS4_SHARE_ACCESS_READ;
1331		break;
1332	case FMODE_WRITE:
1333		res = NFS4_SHARE_ACCESS_WRITE;
1334		break;
1335	case FMODE_READ|FMODE_WRITE:
1336		res = NFS4_SHARE_ACCESS_BOTH;
1337	}
1338	if (!(server->caps & NFS_CAP_ATOMIC_OPEN_V1))
1339		goto out;
1340	/* Want no delegation if we're using O_DIRECT */
1341	if (openflags & O_DIRECT)
1342		res |= NFS4_SHARE_WANT_NO_DELEG;
1343out:
1344	return res;
1345}
1346
1347static enum open_claim_type4
1348nfs4_map_atomic_open_claim(struct nfs_server *server,
1349		enum open_claim_type4 claim)
1350{
1351	if (server->caps & NFS_CAP_ATOMIC_OPEN_V1)
1352		return claim;
1353	switch (claim) {
1354	default:
1355		return claim;
1356	case NFS4_OPEN_CLAIM_FH:
1357		return NFS4_OPEN_CLAIM_NULL;
1358	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1359		return NFS4_OPEN_CLAIM_DELEGATE_CUR;
1360	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1361		return NFS4_OPEN_CLAIM_DELEGATE_PREV;
1362	}
1363}
1364
1365static void nfs4_init_opendata_res(struct nfs4_opendata *p)
1366{
1367	p->o_res.f_attr = &p->f_attr;
1368	p->o_res.seqid = p->o_arg.seqid;
1369	p->c_res.seqid = p->c_arg.seqid;
1370	p->o_res.server = p->o_arg.server;
1371	p->o_res.access_request = p->o_arg.access;
1372	nfs_fattr_init(&p->f_attr);
1373	nfs_fattr_init_names(&p->f_attr, &p->owner_name, &p->group_name);
1374}
1375
1376static struct nfs4_opendata *nfs4_opendata_alloc(struct dentry *dentry,
1377		struct nfs4_state_owner *sp, fmode_t fmode, int flags,
1378		const struct nfs4_open_createattrs *c,
1379		enum open_claim_type4 claim,
1380		gfp_t gfp_mask)
1381{
1382	struct dentry *parent = dget_parent(dentry);
1383	struct inode *dir = d_inode(parent);
1384	struct nfs_server *server = NFS_SERVER(dir);
1385	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
1386	struct nfs4_label *label = (c != NULL) ? c->label : NULL;
1387	struct nfs4_opendata *p;
1388
1389	p = kzalloc(sizeof(*p), gfp_mask);
1390	if (p == NULL)
1391		goto err;
1392
1393	p->f_attr.label = nfs4_label_alloc(server, gfp_mask);
1394	if (IS_ERR(p->f_attr.label))
1395		goto err_free_p;
1396
1397	p->a_label = nfs4_label_alloc(server, gfp_mask);
1398	if (IS_ERR(p->a_label))
1399		goto err_free_f;
1400
1401	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
1402	p->o_arg.seqid = alloc_seqid(&sp->so_seqid, gfp_mask);
1403	if (IS_ERR(p->o_arg.seqid))
1404		goto err_free_label;
1405	nfs_sb_active(dentry->d_sb);
1406	p->dentry = dget(dentry);
1407	p->dir = parent;
1408	p->owner = sp;
1409	atomic_inc(&sp->so_count);
1410	p->o_arg.open_flags = flags;
1411	p->o_arg.fmode = fmode & (FMODE_READ|FMODE_WRITE);
1412	p->o_arg.claim = nfs4_map_atomic_open_claim(server, claim);
1413	p->o_arg.share_access = nfs4_map_atomic_open_share(server,
1414			fmode, flags);
1415	if (flags & O_CREAT) {
1416		p->o_arg.umask = current_umask();
1417		p->o_arg.label = nfs4_label_copy(p->a_label, label);
1418		if (c->sattr != NULL && c->sattr->ia_valid != 0) {
1419			p->o_arg.u.attrs = &p->attrs;
1420			memcpy(&p->attrs, c->sattr, sizeof(p->attrs));
1421
1422			memcpy(p->o_arg.u.verifier.data, c->verf,
1423					sizeof(p->o_arg.u.verifier.data));
1424		}
1425	}
1426	/* ask server to check for all possible rights as results
1427	 * are cached */
1428	switch (p->o_arg.claim) {
1429	default:
1430		break;
1431	case NFS4_OPEN_CLAIM_NULL:
1432	case NFS4_OPEN_CLAIM_FH:
1433		p->o_arg.access = NFS4_ACCESS_READ | NFS4_ACCESS_MODIFY |
1434				  NFS4_ACCESS_EXTEND | NFS4_ACCESS_DELETE |
1435				  NFS4_ACCESS_EXECUTE |
1436				  nfs_access_xattr_mask(server);
1437	}
1438	p->o_arg.clientid = server->nfs_client->cl_clientid;
1439	p->o_arg.id.create_time = ktime_to_ns(sp->so_seqid.create_time);
1440	p->o_arg.id.uniquifier = sp->so_seqid.owner_id;
1441	p->o_arg.name = &dentry->d_name;
1442	p->o_arg.server = server;
1443	p->o_arg.bitmask = nfs4_bitmask(server, label);
1444	p->o_arg.open_bitmap = &nfs4_fattr_bitmap[0];
1445	switch (p->o_arg.claim) {
1446	case NFS4_OPEN_CLAIM_NULL:
1447	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1448	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
1449		p->o_arg.fh = NFS_FH(dir);
1450		break;
1451	case NFS4_OPEN_CLAIM_PREVIOUS:
1452	case NFS4_OPEN_CLAIM_FH:
1453	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1454	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
1455		p->o_arg.fh = NFS_FH(d_inode(dentry));
1456	}
1457	p->c_arg.fh = &p->o_res.fh;
1458	p->c_arg.stateid = &p->o_res.stateid;
1459	p->c_arg.seqid = p->o_arg.seqid;
1460	nfs4_init_opendata_res(p);
1461	kref_init(&p->kref);
1462	return p;
1463
1464err_free_label:
1465	nfs4_label_free(p->a_label);
1466err_free_f:
1467	nfs4_label_free(p->f_attr.label);
1468err_free_p:
1469	kfree(p);
1470err:
1471	dput(parent);
1472	return NULL;
1473}
1474
1475static void nfs4_opendata_free(struct kref *kref)
1476{
1477	struct nfs4_opendata *p = container_of(kref,
1478			struct nfs4_opendata, kref);
1479	struct super_block *sb = p->dentry->d_sb;
1480
1481	nfs4_lgopen_release(p->lgp);
1482	nfs_free_seqid(p->o_arg.seqid);
1483	nfs4_sequence_free_slot(&p->o_res.seq_res);
1484	if (p->state != NULL)
1485		nfs4_put_open_state(p->state);
1486	nfs4_put_state_owner(p->owner);
1487
1488	nfs4_label_free(p->a_label);
1489	nfs4_label_free(p->f_attr.label);
1490
1491	dput(p->dir);
1492	dput(p->dentry);
1493	nfs_sb_deactive(sb);
1494	nfs_fattr_free_names(&p->f_attr);
1495	kfree(p->f_attr.mdsthreshold);
1496	kfree(p);
1497}
1498
1499static void nfs4_opendata_put(struct nfs4_opendata *p)
1500{
1501	if (p != NULL)
1502		kref_put(&p->kref, nfs4_opendata_free);
1503}
1504
1505static bool nfs4_mode_match_open_stateid(struct nfs4_state *state,
1506		fmode_t fmode)
1507{
1508	switch(fmode & (FMODE_READ|FMODE_WRITE)) {
1509	case FMODE_READ|FMODE_WRITE:
1510		return state->n_rdwr != 0;
1511	case FMODE_WRITE:
1512		return state->n_wronly != 0;
1513	case FMODE_READ:
1514		return state->n_rdonly != 0;
1515	}
1516	WARN_ON_ONCE(1);
1517	return false;
1518}
1519
1520static int can_open_cached(struct nfs4_state *state, fmode_t mode,
1521		int open_mode, enum open_claim_type4 claim)
1522{
1523	int ret = 0;
1524
1525	if (open_mode & (O_EXCL|O_TRUNC))
1526		goto out;
1527	switch (claim) {
1528	case NFS4_OPEN_CLAIM_NULL:
1529	case NFS4_OPEN_CLAIM_FH:
1530		goto out;
1531	default:
1532		break;
1533	}
1534	switch (mode & (FMODE_READ|FMODE_WRITE)) {
1535		case FMODE_READ:
1536			ret |= test_bit(NFS_O_RDONLY_STATE, &state->flags) != 0
1537				&& state->n_rdonly != 0;
1538			break;
1539		case FMODE_WRITE:
1540			ret |= test_bit(NFS_O_WRONLY_STATE, &state->flags) != 0
1541				&& state->n_wronly != 0;
1542			break;
1543		case FMODE_READ|FMODE_WRITE:
1544			ret |= test_bit(NFS_O_RDWR_STATE, &state->flags) != 0
1545				&& state->n_rdwr != 0;
1546	}
1547out:
1548	return ret;
1549}
1550
1551static int can_open_delegated(struct nfs_delegation *delegation, fmode_t fmode,
1552		enum open_claim_type4 claim)
1553{
1554	if (delegation == NULL)
1555		return 0;
1556	if ((delegation->type & fmode) != fmode)
1557		return 0;
1558	switch (claim) {
1559	case NFS4_OPEN_CLAIM_NULL:
1560	case NFS4_OPEN_CLAIM_FH:
1561		break;
1562	case NFS4_OPEN_CLAIM_PREVIOUS:
1563		if (!test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags))
1564			break;
1565		fallthrough;
1566	default:
1567		return 0;
1568	}
1569	nfs_mark_delegation_referenced(delegation);
1570	return 1;
1571}
1572
1573static void update_open_stateflags(struct nfs4_state *state, fmode_t fmode)
1574{
1575	switch (fmode) {
1576		case FMODE_WRITE:
1577			state->n_wronly++;
1578			break;
1579		case FMODE_READ:
1580			state->n_rdonly++;
1581			break;
1582		case FMODE_READ|FMODE_WRITE:
1583			state->n_rdwr++;
1584	}
1585	nfs4_state_set_mode_locked(state, state->state | fmode);
1586}
1587
1588#ifdef CONFIG_NFS_V4_1
1589static bool nfs_open_stateid_recover_openmode(struct nfs4_state *state)
1590{
1591	if (state->n_rdonly && !test_bit(NFS_O_RDONLY_STATE, &state->flags))
1592		return true;
1593	if (state->n_wronly && !test_bit(NFS_O_WRONLY_STATE, &state->flags))
1594		return true;
1595	if (state->n_rdwr && !test_bit(NFS_O_RDWR_STATE, &state->flags))
1596		return true;
1597	return false;
1598}
1599#endif /* CONFIG_NFS_V4_1 */
1600
1601static void nfs_state_log_update_open_stateid(struct nfs4_state *state)
1602{
1603	if (test_and_clear_bit(NFS_STATE_CHANGE_WAIT, &state->flags))
1604		wake_up_all(&state->waitq);
1605}
1606
1607static void nfs_test_and_clear_all_open_stateid(struct nfs4_state *state)
1608{
1609	struct nfs_client *clp = state->owner->so_server->nfs_client;
1610	bool need_recover = false;
1611
1612	if (test_and_clear_bit(NFS_O_RDONLY_STATE, &state->flags) && state->n_rdonly)
1613		need_recover = true;
1614	if (test_and_clear_bit(NFS_O_WRONLY_STATE, &state->flags) && state->n_wronly)
1615		need_recover = true;
1616	if (test_and_clear_bit(NFS_O_RDWR_STATE, &state->flags) && state->n_rdwr)
1617		need_recover = true;
1618	if (need_recover)
1619		nfs4_state_mark_reclaim_nograce(clp, state);
1620}
1621
1622/*
1623 * Check for whether or not the caller may update the open stateid
1624 * to the value passed in by stateid.
1625 *
1626 * Note: This function relies heavily on the server implementing
1627 * RFC7530 Section 9.1.4.2, and RFC5661 Section 8.2.2
1628 * correctly.
1629 * i.e. The stateid seqids have to be initialised to 1, and
1630 * are then incremented on every state transition.
1631 */
1632static bool nfs_stateid_is_sequential(struct nfs4_state *state,
1633		const nfs4_stateid *stateid)
1634{
1635	if (test_bit(NFS_OPEN_STATE, &state->flags)) {
1636		/* The common case - we're updating to a new sequence number */
1637		if (nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1638			if (nfs4_stateid_is_next(&state->open_stateid, stateid))
1639				return true;
1640			return false;
1641		}
1642		/* The server returned a new stateid */
1643	}
1644	/* This is the first OPEN in this generation */
1645	if (stateid->seqid == cpu_to_be32(1))
1646		return true;
1647	return false;
1648}
1649
1650static void nfs_resync_open_stateid_locked(struct nfs4_state *state)
1651{
1652	if (!(state->n_wronly || state->n_rdonly || state->n_rdwr))
1653		return;
1654	if (state->n_wronly)
1655		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1656	if (state->n_rdonly)
1657		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1658	if (state->n_rdwr)
1659		set_bit(NFS_O_RDWR_STATE, &state->flags);
1660	set_bit(NFS_OPEN_STATE, &state->flags);
1661}
1662
1663static void nfs_clear_open_stateid_locked(struct nfs4_state *state,
1664		nfs4_stateid *stateid, fmode_t fmode)
1665{
1666	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1667	switch (fmode & (FMODE_READ|FMODE_WRITE)) {
1668	case FMODE_WRITE:
1669		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1670		break;
1671	case FMODE_READ:
1672		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1673		break;
1674	case 0:
1675		clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1676		clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1677		clear_bit(NFS_OPEN_STATE, &state->flags);
1678	}
1679	if (stateid == NULL)
1680		return;
1681	/* Handle OPEN+OPEN_DOWNGRADE races */
1682	if (nfs4_stateid_match_other(stateid, &state->open_stateid) &&
1683	    !nfs4_stateid_is_newer(stateid, &state->open_stateid)) {
1684		nfs_resync_open_stateid_locked(state);
1685		goto out;
1686	}
1687	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1688		nfs4_stateid_copy(&state->stateid, stateid);
1689	nfs4_stateid_copy(&state->open_stateid, stateid);
1690	trace_nfs4_open_stateid_update(state->inode, stateid, 0);
1691out:
1692	nfs_state_log_update_open_stateid(state);
1693}
1694
1695static void nfs_clear_open_stateid(struct nfs4_state *state,
1696	nfs4_stateid *arg_stateid,
1697	nfs4_stateid *stateid, fmode_t fmode)
1698{
1699	write_seqlock(&state->seqlock);
1700	/* Ignore, if the CLOSE argment doesn't match the current stateid */
1701	if (nfs4_state_match_open_stateid_other(state, arg_stateid))
1702		nfs_clear_open_stateid_locked(state, stateid, fmode);
1703	write_sequnlock(&state->seqlock);
1704	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1705		nfs4_schedule_state_manager(state->owner->so_server->nfs_client);
1706}
1707
1708static void nfs_set_open_stateid_locked(struct nfs4_state *state,
1709		const nfs4_stateid *stateid, nfs4_stateid *freeme)
1710	__must_hold(&state->owner->so_lock)
1711	__must_hold(&state->seqlock)
1712	__must_hold(RCU)
1713
1714{
1715	DEFINE_WAIT(wait);
1716	int status = 0;
1717	for (;;) {
1718
1719		if (nfs_stateid_is_sequential(state, stateid))
1720			break;
1721
1722		if (status)
1723			break;
1724		/* Rely on seqids for serialisation with NFSv4.0 */
1725		if (!nfs4_has_session(NFS_SERVER(state->inode)->nfs_client))
1726			break;
1727
1728		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
1729		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
1730		/*
1731		 * Ensure we process the state changes in the same order
1732		 * in which the server processed them by delaying the
1733		 * update of the stateid until we are in sequence.
1734		 */
1735		write_sequnlock(&state->seqlock);
1736		spin_unlock(&state->owner->so_lock);
1737		rcu_read_unlock();
1738		trace_nfs4_open_stateid_update_wait(state->inode, stateid, 0);
1739
1740		if (!fatal_signal_pending(current)) {
1741			if (schedule_timeout(5*HZ) == 0)
1742				status = -EAGAIN;
1743			else
1744				status = 0;
1745		} else
1746			status = -EINTR;
1747		finish_wait(&state->waitq, &wait);
1748		rcu_read_lock();
1749		spin_lock(&state->owner->so_lock);
1750		write_seqlock(&state->seqlock);
1751	}
1752
1753	if (test_bit(NFS_OPEN_STATE, &state->flags) &&
1754	    !nfs4_stateid_match_other(stateid, &state->open_stateid)) {
1755		nfs4_stateid_copy(freeme, &state->open_stateid);
1756		nfs_test_and_clear_all_open_stateid(state);
1757	}
1758
1759	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
1760		nfs4_stateid_copy(&state->stateid, stateid);
1761	nfs4_stateid_copy(&state->open_stateid, stateid);
1762	trace_nfs4_open_stateid_update(state->inode, stateid, status);
1763	nfs_state_log_update_open_stateid(state);
1764}
1765
1766static void nfs_state_set_open_stateid(struct nfs4_state *state,
1767		const nfs4_stateid *open_stateid,
1768		fmode_t fmode,
1769		nfs4_stateid *freeme)
1770{
1771	/*
1772	 * Protect the call to nfs4_state_set_mode_locked and
1773	 * serialise the stateid update
1774	 */
1775	write_seqlock(&state->seqlock);
1776	nfs_set_open_stateid_locked(state, open_stateid, freeme);
1777	switch (fmode) {
1778	case FMODE_READ:
1779		set_bit(NFS_O_RDONLY_STATE, &state->flags);
1780		break;
1781	case FMODE_WRITE:
1782		set_bit(NFS_O_WRONLY_STATE, &state->flags);
1783		break;
1784	case FMODE_READ|FMODE_WRITE:
1785		set_bit(NFS_O_RDWR_STATE, &state->flags);
1786	}
1787	set_bit(NFS_OPEN_STATE, &state->flags);
1788	write_sequnlock(&state->seqlock);
1789}
1790
1791static void nfs_state_clear_open_state_flags(struct nfs4_state *state)
1792{
1793	clear_bit(NFS_O_RDWR_STATE, &state->flags);
1794	clear_bit(NFS_O_WRONLY_STATE, &state->flags);
1795	clear_bit(NFS_O_RDONLY_STATE, &state->flags);
1796	clear_bit(NFS_OPEN_STATE, &state->flags);
1797}
1798
1799static void nfs_state_set_delegation(struct nfs4_state *state,
1800		const nfs4_stateid *deleg_stateid,
1801		fmode_t fmode)
1802{
1803	/*
1804	 * Protect the call to nfs4_state_set_mode_locked and
1805	 * serialise the stateid update
1806	 */
1807	write_seqlock(&state->seqlock);
1808	nfs4_stateid_copy(&state->stateid, deleg_stateid);
1809	set_bit(NFS_DELEGATED_STATE, &state->flags);
1810	write_sequnlock(&state->seqlock);
1811}
1812
1813static void nfs_state_clear_delegation(struct nfs4_state *state)
1814{
1815	write_seqlock(&state->seqlock);
1816	nfs4_stateid_copy(&state->stateid, &state->open_stateid);
1817	clear_bit(NFS_DELEGATED_STATE, &state->flags);
1818	write_sequnlock(&state->seqlock);
1819}
1820
1821int update_open_stateid(struct nfs4_state *state,
1822		const nfs4_stateid *open_stateid,
1823		const nfs4_stateid *delegation,
1824		fmode_t fmode)
1825{
1826	struct nfs_server *server = NFS_SERVER(state->inode);
1827	struct nfs_client *clp = server->nfs_client;
1828	struct nfs_inode *nfsi = NFS_I(state->inode);
1829	struct nfs_delegation *deleg_cur;
1830	nfs4_stateid freeme = { };
1831	int ret = 0;
1832
1833	fmode &= (FMODE_READ|FMODE_WRITE);
1834
1835	rcu_read_lock();
1836	spin_lock(&state->owner->so_lock);
1837	if (open_stateid != NULL) {
1838		nfs_state_set_open_stateid(state, open_stateid, fmode, &freeme);
1839		ret = 1;
1840	}
1841
1842	deleg_cur = nfs4_get_valid_delegation(state->inode);
1843	if (deleg_cur == NULL)
1844		goto no_delegation;
1845
1846	spin_lock(&deleg_cur->lock);
1847	if (rcu_dereference(nfsi->delegation) != deleg_cur ||
1848	   test_bit(NFS_DELEGATION_RETURNING, &deleg_cur->flags) ||
1849	    (deleg_cur->type & fmode) != fmode)
1850		goto no_delegation_unlock;
1851
1852	if (delegation == NULL)
1853		delegation = &deleg_cur->stateid;
1854	else if (!nfs4_stateid_match_other(&deleg_cur->stateid, delegation))
1855		goto no_delegation_unlock;
1856
1857	nfs_mark_delegation_referenced(deleg_cur);
1858	nfs_state_set_delegation(state, &deleg_cur->stateid, fmode);
1859	ret = 1;
1860no_delegation_unlock:
1861	spin_unlock(&deleg_cur->lock);
1862no_delegation:
1863	if (ret)
1864		update_open_stateflags(state, fmode);
1865	spin_unlock(&state->owner->so_lock);
1866	rcu_read_unlock();
1867
1868	if (test_bit(NFS_STATE_RECLAIM_NOGRACE, &state->flags))
1869		nfs4_schedule_state_manager(clp);
1870	if (freeme.type != 0)
1871		nfs4_test_and_free_stateid(server, &freeme,
1872				state->owner->so_cred);
1873
1874	return ret;
1875}
1876
1877static bool nfs4_update_lock_stateid(struct nfs4_lock_state *lsp,
1878		const nfs4_stateid *stateid)
1879{
1880	struct nfs4_state *state = lsp->ls_state;
1881	bool ret = false;
1882
1883	spin_lock(&state->state_lock);
1884	if (!nfs4_stateid_match_other(stateid, &lsp->ls_stateid))
1885		goto out_noupdate;
1886	if (!nfs4_stateid_is_newer(stateid, &lsp->ls_stateid))
1887		goto out_noupdate;
1888	nfs4_stateid_copy(&lsp->ls_stateid, stateid);
1889	ret = true;
1890out_noupdate:
1891	spin_unlock(&state->state_lock);
1892	return ret;
1893}
1894
1895static void nfs4_return_incompatible_delegation(struct inode *inode, fmode_t fmode)
1896{
1897	struct nfs_delegation *delegation;
1898
1899	fmode &= FMODE_READ|FMODE_WRITE;
1900	rcu_read_lock();
1901	delegation = nfs4_get_valid_delegation(inode);
1902	if (delegation == NULL || (delegation->type & fmode) == fmode) {
1903		rcu_read_unlock();
1904		return;
1905	}
1906	rcu_read_unlock();
1907	nfs4_inode_return_delegation(inode);
1908}
1909
1910static struct nfs4_state *nfs4_try_open_cached(struct nfs4_opendata *opendata)
1911{
1912	struct nfs4_state *state = opendata->state;
1913	struct nfs_delegation *delegation;
1914	int open_mode = opendata->o_arg.open_flags;
1915	fmode_t fmode = opendata->o_arg.fmode;
1916	enum open_claim_type4 claim = opendata->o_arg.claim;
1917	nfs4_stateid stateid;
1918	int ret = -EAGAIN;
1919
1920	for (;;) {
1921		spin_lock(&state->owner->so_lock);
1922		if (can_open_cached(state, fmode, open_mode, claim)) {
1923			update_open_stateflags(state, fmode);
1924			spin_unlock(&state->owner->so_lock);
1925			goto out_return_state;
1926		}
1927		spin_unlock(&state->owner->so_lock);
1928		rcu_read_lock();
1929		delegation = nfs4_get_valid_delegation(state->inode);
1930		if (!can_open_delegated(delegation, fmode, claim)) {
1931			rcu_read_unlock();
1932			break;
1933		}
1934		/* Save the delegation */
1935		nfs4_stateid_copy(&stateid, &delegation->stateid);
1936		rcu_read_unlock();
1937		nfs_release_seqid(opendata->o_arg.seqid);
1938		if (!opendata->is_recover) {
1939			ret = nfs_may_open(state->inode, state->owner->so_cred, open_mode);
1940			if (ret != 0)
1941				goto out;
1942		}
1943		ret = -EAGAIN;
1944
1945		/* Try to update the stateid using the delegation */
1946		if (update_open_stateid(state, NULL, &stateid, fmode))
1947			goto out_return_state;
1948	}
1949out:
1950	return ERR_PTR(ret);
1951out_return_state:
1952	refcount_inc(&state->count);
1953	return state;
1954}
1955
1956static void
1957nfs4_opendata_check_deleg(struct nfs4_opendata *data, struct nfs4_state *state)
1958{
1959	struct nfs_client *clp = NFS_SERVER(state->inode)->nfs_client;
1960	struct nfs_delegation *delegation;
1961	int delegation_flags = 0;
1962
1963	rcu_read_lock();
1964	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
1965	if (delegation)
1966		delegation_flags = delegation->flags;
1967	rcu_read_unlock();
1968	switch (data->o_arg.claim) {
1969	default:
1970		break;
1971	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
1972	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
1973		pr_err_ratelimited("NFS: Broken NFSv4 server %s is "
1974				   "returning a delegation for "
1975				   "OPEN(CLAIM_DELEGATE_CUR)\n",
1976				   clp->cl_hostname);
1977		return;
1978	}
1979	if ((delegation_flags & 1UL<<NFS_DELEGATION_NEED_RECLAIM) == 0)
1980		nfs_inode_set_delegation(state->inode,
1981				data->owner->so_cred,
1982				data->o_res.delegation_type,
1983				&data->o_res.delegation,
1984				data->o_res.pagemod_limit);
1985	else
1986		nfs_inode_reclaim_delegation(state->inode,
1987				data->owner->so_cred,
1988				data->o_res.delegation_type,
1989				&data->o_res.delegation,
1990				data->o_res.pagemod_limit);
1991
1992	if (data->o_res.do_recall)
1993		nfs_async_inode_return_delegation(state->inode,
1994						  &data->o_res.delegation);
1995}
1996
1997/*
1998 * Check the inode attributes against the CLAIM_PREVIOUS returned attributes
1999 * and update the nfs4_state.
2000 */
2001static struct nfs4_state *
2002_nfs4_opendata_reclaim_to_nfs4_state(struct nfs4_opendata *data)
2003{
2004	struct inode *inode = data->state->inode;
2005	struct nfs4_state *state = data->state;
2006	int ret;
2007
2008	if (!data->rpc_done) {
2009		if (data->rpc_status)
2010			return ERR_PTR(data->rpc_status);
2011		return nfs4_try_open_cached(data);
2012	}
2013
2014	ret = nfs_refresh_inode(inode, &data->f_attr);
2015	if (ret)
2016		return ERR_PTR(ret);
2017
2018	if (data->o_res.delegation_type != 0)
2019		nfs4_opendata_check_deleg(data, state);
2020
2021	if (!update_open_stateid(state, &data->o_res.stateid,
2022				NULL, data->o_arg.fmode))
2023		return ERR_PTR(-EAGAIN);
2024	refcount_inc(&state->count);
2025
2026	return state;
2027}
2028
2029static struct inode *
2030nfs4_opendata_get_inode(struct nfs4_opendata *data)
2031{
2032	struct inode *inode;
2033
2034	switch (data->o_arg.claim) {
2035	case NFS4_OPEN_CLAIM_NULL:
2036	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
2037	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
2038		if (!(data->f_attr.valid & NFS_ATTR_FATTR))
2039			return ERR_PTR(-EAGAIN);
2040		inode = nfs_fhget(data->dir->d_sb, &data->o_res.fh,
2041				&data->f_attr);
2042		break;
2043	default:
2044		inode = d_inode(data->dentry);
2045		ihold(inode);
2046		nfs_refresh_inode(inode, &data->f_attr);
2047	}
2048	return inode;
2049}
2050
2051static struct nfs4_state *
2052nfs4_opendata_find_nfs4_state(struct nfs4_opendata *data)
2053{
2054	struct nfs4_state *state;
2055	struct inode *inode;
2056
2057	inode = nfs4_opendata_get_inode(data);
2058	if (IS_ERR(inode))
2059		return ERR_CAST(inode);
2060	if (data->state != NULL && data->state->inode == inode) {
2061		state = data->state;
2062		refcount_inc(&state->count);
2063	} else
2064		state = nfs4_get_open_state(inode, data->owner);
2065	iput(inode);
2066	if (state == NULL)
2067		state = ERR_PTR(-ENOMEM);
2068	return state;
2069}
2070
2071static struct nfs4_state *
2072_nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2073{
2074	struct nfs4_state *state;
2075
2076	if (!data->rpc_done) {
2077		state = nfs4_try_open_cached(data);
2078		trace_nfs4_cached_open(data->state);
2079		goto out;
2080	}
2081
2082	state = nfs4_opendata_find_nfs4_state(data);
2083	if (IS_ERR(state))
2084		goto out;
2085
2086	if (data->o_res.delegation_type != 0)
2087		nfs4_opendata_check_deleg(data, state);
2088	if (!update_open_stateid(state, &data->o_res.stateid,
2089				NULL, data->o_arg.fmode)) {
2090		nfs4_put_open_state(state);
2091		state = ERR_PTR(-EAGAIN);
2092	}
2093out:
2094	nfs_release_seqid(data->o_arg.seqid);
2095	return state;
2096}
2097
2098static struct nfs4_state *
2099nfs4_opendata_to_nfs4_state(struct nfs4_opendata *data)
2100{
2101	struct nfs4_state *ret;
2102
2103	if (data->o_arg.claim == NFS4_OPEN_CLAIM_PREVIOUS)
2104		ret =_nfs4_opendata_reclaim_to_nfs4_state(data);
2105	else
2106		ret = _nfs4_opendata_to_nfs4_state(data);
2107	nfs4_sequence_free_slot(&data->o_res.seq_res);
2108	return ret;
2109}
2110
2111static struct nfs_open_context *
2112nfs4_state_find_open_context_mode(struct nfs4_state *state, fmode_t mode)
2113{
2114	struct nfs_inode *nfsi = NFS_I(state->inode);
2115	struct nfs_open_context *ctx;
2116
2117	rcu_read_lock();
2118	list_for_each_entry_rcu(ctx, &nfsi->open_files, list) {
2119		if (ctx->state != state)
2120			continue;
2121		if ((ctx->mode & mode) != mode)
2122			continue;
2123		if (!get_nfs_open_context(ctx))
2124			continue;
2125		rcu_read_unlock();
2126		return ctx;
2127	}
2128	rcu_read_unlock();
2129	return ERR_PTR(-ENOENT);
2130}
2131
2132static struct nfs_open_context *
2133nfs4_state_find_open_context(struct nfs4_state *state)
2134{
2135	struct nfs_open_context *ctx;
2136
2137	ctx = nfs4_state_find_open_context_mode(state, FMODE_READ|FMODE_WRITE);
2138	if (!IS_ERR(ctx))
2139		return ctx;
2140	ctx = nfs4_state_find_open_context_mode(state, FMODE_WRITE);
2141	if (!IS_ERR(ctx))
2142		return ctx;
2143	return nfs4_state_find_open_context_mode(state, FMODE_READ);
2144}
2145
2146static struct nfs4_opendata *nfs4_open_recoverdata_alloc(struct nfs_open_context *ctx,
2147		struct nfs4_state *state, enum open_claim_type4 claim)
2148{
2149	struct nfs4_opendata *opendata;
2150
2151	opendata = nfs4_opendata_alloc(ctx->dentry, state->owner, 0, 0,
2152			NULL, claim, GFP_NOFS);
2153	if (opendata == NULL)
2154		return ERR_PTR(-ENOMEM);
2155	opendata->state = state;
2156	refcount_inc(&state->count);
2157	return opendata;
2158}
2159
2160static int nfs4_open_recover_helper(struct nfs4_opendata *opendata,
2161				    fmode_t fmode)
2162{
2163	struct nfs4_state *newstate;
2164	struct nfs_server *server = NFS_SB(opendata->dentry->d_sb);
2165	int openflags = opendata->o_arg.open_flags;
2166	int ret;
2167
2168	if (!nfs4_mode_match_open_stateid(opendata->state, fmode))
2169		return 0;
2170	opendata->o_arg.fmode = fmode;
2171	opendata->o_arg.share_access =
2172		nfs4_map_atomic_open_share(server, fmode, openflags);
2173	memset(&opendata->o_res, 0, sizeof(opendata->o_res));
2174	memset(&opendata->c_res, 0, sizeof(opendata->c_res));
2175	nfs4_init_opendata_res(opendata);
2176	ret = _nfs4_recover_proc_open(opendata);
2177	if (ret != 0)
2178		return ret;
2179	newstate = nfs4_opendata_to_nfs4_state(opendata);
2180	if (IS_ERR(newstate))
2181		return PTR_ERR(newstate);
2182	if (newstate != opendata->state)
2183		ret = -ESTALE;
2184	nfs4_close_state(newstate, fmode);
2185	return ret;
2186}
2187
2188static int nfs4_open_recover(struct nfs4_opendata *opendata, struct nfs4_state *state)
2189{
2190	int ret;
2191
2192	/* memory barrier prior to reading state->n_* */
2193	smp_rmb();
2194	ret = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2195	if (ret != 0)
2196		return ret;
2197	ret = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2198	if (ret != 0)
2199		return ret;
2200	ret = nfs4_open_recover_helper(opendata, FMODE_READ);
2201	if (ret != 0)
2202		return ret;
2203	/*
2204	 * We may have performed cached opens for all three recoveries.
2205	 * Check if we need to update the current stateid.
2206	 */
2207	if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0 &&
2208	    !nfs4_stateid_match(&state->stateid, &state->open_stateid)) {
2209		write_seqlock(&state->seqlock);
2210		if (test_bit(NFS_DELEGATED_STATE, &state->flags) == 0)
2211			nfs4_stateid_copy(&state->stateid, &state->open_stateid);
2212		write_sequnlock(&state->seqlock);
2213	}
2214	return 0;
2215}
2216
2217/*
2218 * OPEN_RECLAIM:
2219 * 	reclaim state on the server after a reboot.
2220 */
2221static int _nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2222{
2223	struct nfs_delegation *delegation;
2224	struct nfs4_opendata *opendata;
2225	fmode_t delegation_type = 0;
2226	int status;
2227
2228	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2229			NFS4_OPEN_CLAIM_PREVIOUS);
2230	if (IS_ERR(opendata))
2231		return PTR_ERR(opendata);
2232	rcu_read_lock();
2233	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2234	if (delegation != NULL && test_bit(NFS_DELEGATION_NEED_RECLAIM, &delegation->flags) != 0)
2235		delegation_type = delegation->type;
2236	rcu_read_unlock();
2237	opendata->o_arg.u.delegation_type = delegation_type;
2238	status = nfs4_open_recover(opendata, state);
2239	nfs4_opendata_put(opendata);
2240	return status;
2241}
2242
2243static int nfs4_do_open_reclaim(struct nfs_open_context *ctx, struct nfs4_state *state)
2244{
2245	struct nfs_server *server = NFS_SERVER(state->inode);
2246	struct nfs4_exception exception = { };
2247	int err;
2248	do {
2249		err = _nfs4_do_open_reclaim(ctx, state);
2250		trace_nfs4_open_reclaim(ctx, 0, err);
2251		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2252			continue;
2253		if (err != -NFS4ERR_DELAY)
2254			break;
2255		nfs4_handle_exception(server, err, &exception);
2256	} while (exception.retry);
2257	return err;
2258}
2259
2260static int nfs4_open_reclaim(struct nfs4_state_owner *sp, struct nfs4_state *state)
2261{
2262	struct nfs_open_context *ctx;
2263	int ret;
2264
2265	ctx = nfs4_state_find_open_context(state);
2266	if (IS_ERR(ctx))
2267		return -EAGAIN;
2268	clear_bit(NFS_DELEGATED_STATE, &state->flags);
2269	nfs_state_clear_open_state_flags(state);
2270	ret = nfs4_do_open_reclaim(ctx, state);
2271	put_nfs_open_context(ctx);
2272	return ret;
2273}
2274
2275static int nfs4_handle_delegation_recall_error(struct nfs_server *server, struct nfs4_state *state, const nfs4_stateid *stateid, struct file_lock *fl, int err)
2276{
2277	switch (err) {
2278		default:
2279			printk(KERN_ERR "NFS: %s: unhandled error "
2280					"%d.\n", __func__, err);
2281			fallthrough;
2282		case 0:
2283		case -ENOENT:
2284		case -EAGAIN:
2285		case -ESTALE:
2286		case -ETIMEDOUT:
2287			break;
2288		case -NFS4ERR_BADSESSION:
2289		case -NFS4ERR_BADSLOT:
2290		case -NFS4ERR_BAD_HIGH_SLOT:
2291		case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
2292		case -NFS4ERR_DEADSESSION:
2293			return -EAGAIN;
2294		case -NFS4ERR_STALE_CLIENTID:
2295		case -NFS4ERR_STALE_STATEID:
2296			/* Don't recall a delegation if it was lost */
2297			nfs4_schedule_lease_recovery(server->nfs_client);
2298			return -EAGAIN;
2299		case -NFS4ERR_MOVED:
2300			nfs4_schedule_migration_recovery(server);
2301			return -EAGAIN;
2302		case -NFS4ERR_LEASE_MOVED:
2303			nfs4_schedule_lease_moved_recovery(server->nfs_client);
2304			return -EAGAIN;
2305		case -NFS4ERR_DELEG_REVOKED:
2306		case -NFS4ERR_ADMIN_REVOKED:
2307		case -NFS4ERR_EXPIRED:
2308		case -NFS4ERR_BAD_STATEID:
2309		case -NFS4ERR_OPENMODE:
2310			nfs_inode_find_state_and_recover(state->inode,
2311					stateid);
2312			nfs4_schedule_stateid_recovery(server, state);
2313			return -EAGAIN;
2314		case -NFS4ERR_DELAY:
2315		case -NFS4ERR_GRACE:
2316			ssleep(1);
2317			return -EAGAIN;
2318		case -ENOMEM:
2319		case -NFS4ERR_DENIED:
2320			if (fl) {
2321				struct nfs4_lock_state *lsp = fl->fl_u.nfs4_fl.owner;
2322				if (lsp)
2323					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2324			}
2325			return 0;
2326	}
2327	return err;
2328}
2329
2330int nfs4_open_delegation_recall(struct nfs_open_context *ctx,
2331		struct nfs4_state *state, const nfs4_stateid *stateid)
2332{
2333	struct nfs_server *server = NFS_SERVER(state->inode);
2334	struct nfs4_opendata *opendata;
2335	int err = 0;
2336
2337	opendata = nfs4_open_recoverdata_alloc(ctx, state,
2338			NFS4_OPEN_CLAIM_DELEG_CUR_FH);
2339	if (IS_ERR(opendata))
2340		return PTR_ERR(opendata);
2341	nfs4_stateid_copy(&opendata->o_arg.u.delegation, stateid);
2342	if (!test_bit(NFS_O_RDWR_STATE, &state->flags)) {
2343		err = nfs4_open_recover_helper(opendata, FMODE_READ|FMODE_WRITE);
2344		if (err)
2345			goto out;
2346	}
2347	if (!test_bit(NFS_O_WRONLY_STATE, &state->flags)) {
2348		err = nfs4_open_recover_helper(opendata, FMODE_WRITE);
2349		if (err)
2350			goto out;
2351	}
2352	if (!test_bit(NFS_O_RDONLY_STATE, &state->flags)) {
2353		err = nfs4_open_recover_helper(opendata, FMODE_READ);
2354		if (err)
2355			goto out;
2356	}
2357	nfs_state_clear_delegation(state);
2358out:
2359	nfs4_opendata_put(opendata);
2360	return nfs4_handle_delegation_recall_error(server, state, stateid, NULL, err);
2361}
2362
2363static void nfs4_open_confirm_prepare(struct rpc_task *task, void *calldata)
2364{
2365	struct nfs4_opendata *data = calldata;
2366
2367	nfs4_setup_sequence(data->o_arg.server->nfs_client,
2368			   &data->c_arg.seq_args, &data->c_res.seq_res, task);
2369}
2370
2371static void nfs4_open_confirm_done(struct rpc_task *task, void *calldata)
2372{
2373	struct nfs4_opendata *data = calldata;
2374
2375	nfs40_sequence_done(task, &data->c_res.seq_res);
2376
2377	data->rpc_status = task->tk_status;
2378	if (data->rpc_status == 0) {
2379		nfs4_stateid_copy(&data->o_res.stateid, &data->c_res.stateid);
2380		nfs_confirm_seqid(&data->owner->so_seqid, 0);
2381		renew_lease(data->o_res.server, data->timestamp);
2382		data->rpc_done = true;
2383	}
2384}
2385
2386static void nfs4_open_confirm_release(void *calldata)
2387{
2388	struct nfs4_opendata *data = calldata;
2389	struct nfs4_state *state = NULL;
2390
2391	/* If this request hasn't been cancelled, do nothing */
2392	if (!data->cancelled)
2393		goto out_free;
2394	/* In case of error, no cleanup! */
2395	if (!data->rpc_done)
2396		goto out_free;
2397	state = nfs4_opendata_to_nfs4_state(data);
2398	if (!IS_ERR(state))
2399		nfs4_close_state(state, data->o_arg.fmode);
2400out_free:
2401	nfs4_opendata_put(data);
2402}
2403
2404static const struct rpc_call_ops nfs4_open_confirm_ops = {
2405	.rpc_call_prepare = nfs4_open_confirm_prepare,
2406	.rpc_call_done = nfs4_open_confirm_done,
2407	.rpc_release = nfs4_open_confirm_release,
2408};
2409
2410/*
2411 * Note: On error, nfs4_proc_open_confirm will free the struct nfs4_opendata
2412 */
2413static int _nfs4_proc_open_confirm(struct nfs4_opendata *data)
2414{
2415	struct nfs_server *server = NFS_SERVER(d_inode(data->dir));
2416	struct rpc_task *task;
2417	struct  rpc_message msg = {
2418		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_CONFIRM],
2419		.rpc_argp = &data->c_arg,
2420		.rpc_resp = &data->c_res,
2421		.rpc_cred = data->owner->so_cred,
2422	};
2423	struct rpc_task_setup task_setup_data = {
2424		.rpc_client = server->client,
2425		.rpc_message = &msg,
2426		.callback_ops = &nfs4_open_confirm_ops,
2427		.callback_data = data,
2428		.workqueue = nfsiod_workqueue,
2429		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2430	};
2431	int status;
2432
2433	nfs4_init_sequence(&data->c_arg.seq_args, &data->c_res.seq_res, 1,
2434				data->is_recover);
2435	kref_get(&data->kref);
2436	data->rpc_done = false;
2437	data->rpc_status = 0;
2438	data->timestamp = jiffies;
2439	task = rpc_run_task(&task_setup_data);
2440	if (IS_ERR(task))
2441		return PTR_ERR(task);
2442	status = rpc_wait_for_completion_task(task);
2443	if (status != 0) {
2444		data->cancelled = true;
2445		smp_wmb();
2446	} else
2447		status = data->rpc_status;
2448	rpc_put_task(task);
2449	return status;
2450}
2451
2452static void nfs4_open_prepare(struct rpc_task *task, void *calldata)
2453{
2454	struct nfs4_opendata *data = calldata;
2455	struct nfs4_state_owner *sp = data->owner;
2456	struct nfs_client *clp = sp->so_server->nfs_client;
2457	enum open_claim_type4 claim = data->o_arg.claim;
2458
2459	if (nfs_wait_on_sequence(data->o_arg.seqid, task) != 0)
2460		goto out_wait;
2461	/*
2462	 * Check if we still need to send an OPEN call, or if we can use
2463	 * a delegation instead.
2464	 */
2465	if (data->state != NULL) {
2466		struct nfs_delegation *delegation;
2467
2468		if (can_open_cached(data->state, data->o_arg.fmode,
2469					data->o_arg.open_flags, claim))
2470			goto out_no_action;
2471		rcu_read_lock();
2472		delegation = nfs4_get_valid_delegation(data->state->inode);
2473		if (can_open_delegated(delegation, data->o_arg.fmode, claim))
2474			goto unlock_no_action;
2475		rcu_read_unlock();
2476	}
2477	/* Update client id. */
2478	data->o_arg.clientid = clp->cl_clientid;
2479	switch (claim) {
2480	default:
2481		break;
2482	case NFS4_OPEN_CLAIM_PREVIOUS:
2483	case NFS4_OPEN_CLAIM_DELEG_CUR_FH:
2484	case NFS4_OPEN_CLAIM_DELEG_PREV_FH:
2485		data->o_arg.open_bitmap = &nfs4_open_noattr_bitmap[0];
2486		fallthrough;
2487	case NFS4_OPEN_CLAIM_FH:
2488		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_NOATTR];
2489	}
2490	data->timestamp = jiffies;
2491	if (nfs4_setup_sequence(data->o_arg.server->nfs_client,
2492				&data->o_arg.seq_args,
2493				&data->o_res.seq_res,
2494				task) != 0)
2495		nfs_release_seqid(data->o_arg.seqid);
2496
2497	/* Set the create mode (note dependency on the session type) */
2498	data->o_arg.createmode = NFS4_CREATE_UNCHECKED;
2499	if (data->o_arg.open_flags & O_EXCL) {
2500		data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE4_1;
2501		if (clp->cl_mvops->minor_version == 0) {
2502			data->o_arg.createmode = NFS4_CREATE_EXCLUSIVE;
2503			/* don't put an ACCESS op in OPEN compound if O_EXCL,
2504			 * because ACCESS will return permission denied for
2505			 * all bits until close */
2506			data->o_res.access_request = data->o_arg.access = 0;
2507		} else if (nfs4_has_persistent_session(clp))
2508			data->o_arg.createmode = NFS4_CREATE_GUARDED;
2509	}
2510	return;
2511unlock_no_action:
2512	trace_nfs4_cached_open(data->state);
2513	rcu_read_unlock();
2514out_no_action:
2515	task->tk_action = NULL;
2516out_wait:
2517	nfs4_sequence_done(task, &data->o_res.seq_res);
2518}
2519
2520static void nfs4_open_done(struct rpc_task *task, void *calldata)
2521{
2522	struct nfs4_opendata *data = calldata;
2523
2524	data->rpc_status = task->tk_status;
2525
2526	if (!nfs4_sequence_process(task, &data->o_res.seq_res))
2527		return;
2528
2529	if (task->tk_status == 0) {
2530		if (data->o_res.f_attr->valid & NFS_ATTR_FATTR_TYPE) {
2531			switch (data->o_res.f_attr->mode & S_IFMT) {
2532			case S_IFREG:
2533				break;
2534			case S_IFLNK:
2535				data->rpc_status = -ELOOP;
2536				break;
2537			case S_IFDIR:
2538				data->rpc_status = -EISDIR;
2539				break;
2540			default:
2541				data->rpc_status = -ENOTDIR;
2542			}
2543		}
2544		renew_lease(data->o_res.server, data->timestamp);
2545		if (!(data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM))
2546			nfs_confirm_seqid(&data->owner->so_seqid, 0);
2547	}
2548	data->rpc_done = true;
2549}
2550
2551static void nfs4_open_release(void *calldata)
2552{
2553	struct nfs4_opendata *data = calldata;
2554	struct nfs4_state *state = NULL;
2555
2556	/* If this request hasn't been cancelled, do nothing */
2557	if (!data->cancelled)
2558		goto out_free;
2559	/* In case of error, no cleanup! */
2560	if (data->rpc_status != 0 || !data->rpc_done)
2561		goto out_free;
2562	/* In case we need an open_confirm, no cleanup! */
2563	if (data->o_res.rflags & NFS4_OPEN_RESULT_CONFIRM)
2564		goto out_free;
2565	state = nfs4_opendata_to_nfs4_state(data);
2566	if (!IS_ERR(state))
2567		nfs4_close_state(state, data->o_arg.fmode);
2568out_free:
2569	nfs4_opendata_put(data);
2570}
2571
2572static const struct rpc_call_ops nfs4_open_ops = {
2573	.rpc_call_prepare = nfs4_open_prepare,
2574	.rpc_call_done = nfs4_open_done,
2575	.rpc_release = nfs4_open_release,
2576};
2577
2578static int nfs4_run_open_task(struct nfs4_opendata *data,
2579			      struct nfs_open_context *ctx)
2580{
2581	struct inode *dir = d_inode(data->dir);
2582	struct nfs_server *server = NFS_SERVER(dir);
2583	struct nfs_openargs *o_arg = &data->o_arg;
2584	struct nfs_openres *o_res = &data->o_res;
2585	struct rpc_task *task;
2586	struct rpc_message msg = {
2587		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN],
2588		.rpc_argp = o_arg,
2589		.rpc_resp = o_res,
2590		.rpc_cred = data->owner->so_cred,
2591	};
2592	struct rpc_task_setup task_setup_data = {
2593		.rpc_client = server->client,
2594		.rpc_message = &msg,
2595		.callback_ops = &nfs4_open_ops,
2596		.callback_data = data,
2597		.workqueue = nfsiod_workqueue,
2598		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
2599	};
2600	int status;
2601
2602	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
2603		task_setup_data.flags |= RPC_TASK_MOVEABLE;
2604
2605	kref_get(&data->kref);
2606	data->rpc_done = false;
2607	data->rpc_status = 0;
2608	data->cancelled = false;
2609	data->is_recover = false;
2610	if (!ctx) {
2611		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 1);
2612		data->is_recover = true;
2613		task_setup_data.flags |= RPC_TASK_TIMEOUT;
2614	} else {
2615		nfs4_init_sequence(&o_arg->seq_args, &o_res->seq_res, 1, 0);
2616		pnfs_lgopen_prepare(data, ctx);
2617	}
2618	task = rpc_run_task(&task_setup_data);
2619	if (IS_ERR(task))
2620		return PTR_ERR(task);
2621	status = rpc_wait_for_completion_task(task);
2622	if (status != 0) {
2623		data->cancelled = true;
2624		smp_wmb();
2625	} else
2626		status = data->rpc_status;
2627	rpc_put_task(task);
2628
2629	return status;
2630}
2631
2632static int _nfs4_recover_proc_open(struct nfs4_opendata *data)
2633{
2634	struct inode *dir = d_inode(data->dir);
2635	struct nfs_openres *o_res = &data->o_res;
2636	int status;
2637
2638	status = nfs4_run_open_task(data, NULL);
2639	if (status != 0 || !data->rpc_done)
2640		return status;
2641
2642	nfs_fattr_map_and_free_names(NFS_SERVER(dir), &data->f_attr);
2643
2644	if (o_res->rflags & NFS4_OPEN_RESULT_CONFIRM)
2645		status = _nfs4_proc_open_confirm(data);
2646
2647	return status;
2648}
2649
2650/*
2651 * Additional permission checks in order to distinguish between an
2652 * open for read, and an open for execute. This works around the
2653 * fact that NFSv4 OPEN treats read and execute permissions as being
2654 * the same.
2655 * Note that in the non-execute case, we want to turn off permission
2656 * checking if we just created a new file (POSIX open() semantics).
2657 */
2658static int nfs4_opendata_access(const struct cred *cred,
2659				struct nfs4_opendata *opendata,
2660				struct nfs4_state *state, fmode_t fmode)
2661{
2662	struct nfs_access_entry cache;
2663	u32 mask, flags;
2664
2665	/* access call failed or for some reason the server doesn't
2666	 * support any access modes -- defer access call until later */
2667	if (opendata->o_res.access_supported == 0)
2668		return 0;
2669
2670	mask = 0;
2671	if (fmode & FMODE_EXEC) {
2672		/* ONLY check for exec rights */
2673		if (S_ISDIR(state->inode->i_mode))
2674			mask = NFS4_ACCESS_LOOKUP;
2675		else
2676			mask = NFS4_ACCESS_EXECUTE;
2677	} else if ((fmode & FMODE_READ) && !opendata->file_created)
2678		mask = NFS4_ACCESS_READ;
2679
2680	nfs_access_set_mask(&cache, opendata->o_res.access_result);
2681	nfs_access_add_cache(state->inode, &cache, cred);
2682
2683	flags = NFS4_ACCESS_READ | NFS4_ACCESS_EXECUTE | NFS4_ACCESS_LOOKUP;
2684	if ((mask & ~cache.mask & flags) == 0)
2685		return 0;
2686
2687	return -EACCES;
2688}
2689
2690/*
2691 * Note: On error, nfs4_proc_open will free the struct nfs4_opendata
2692 */
2693static int _nfs4_proc_open(struct nfs4_opendata *data,
2694			   struct nfs_open_context *ctx)
2695{
2696	struct inode *dir = d_inode(data->dir);
2697	struct nfs_server *server = NFS_SERVER(dir);
2698	struct nfs_openargs *o_arg = &data->o_arg;
2699	struct nfs_openres *o_res = &data->o_res;
2700	int status;
2701
2702	status = nfs4_run_open_task(data, ctx);
2703	if (!data->rpc_done)
2704		return status;
2705	if (status != 0) {
2706		if (status == -NFS4ERR_BADNAME &&
2707				!(o_arg->open_flags & O_CREAT))
2708			return -ENOENT;
2709		return status;
2710	}
2711
2712	nfs_fattr_map_and_free_names(server, &data->f_attr);
2713
2714	if (o_arg->open_flags & O_CREAT) {
2715		if (o_arg->open_flags & O_EXCL)
2716			data->file_created = true;
2717		else if (o_res->cinfo.before != o_res->cinfo.after)
2718			data->file_created = true;
2719		if (data->file_created ||
2720		    inode_peek_iversion_raw(dir) != o_res->cinfo.after)
2721			nfs4_update_changeattr(dir, &o_res->cinfo,
2722					o_res->f_attr->time_start,
2723					NFS_INO_INVALID_DATA);
2724	}
2725	if ((o_res->rflags & NFS4_OPEN_RESULT_LOCKTYPE_POSIX) == 0)
2726		server->caps &= ~NFS_CAP_POSIX_LOCK;
2727	if(o_res->rflags & NFS4_OPEN_RESULT_CONFIRM) {
2728		status = _nfs4_proc_open_confirm(data);
2729		if (status != 0)
2730			return status;
2731	}
2732	if (!(o_res->f_attr->valid & NFS_ATTR_FATTR)) {
2733		struct nfs_fh *fh = &o_res->fh;
2734
2735		nfs4_sequence_free_slot(&o_res->seq_res);
2736		if (o_arg->claim == NFS4_OPEN_CLAIM_FH)
2737			fh = NFS_FH(d_inode(data->dentry));
2738		nfs4_proc_getattr(server, fh, o_res->f_attr, NULL);
2739	}
2740	return 0;
2741}
2742
2743/*
2744 * OPEN_EXPIRED:
2745 * 	reclaim state on the server after a network partition.
2746 * 	Assumes caller holds the appropriate lock
2747 */
2748static int _nfs4_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2749{
2750	struct nfs4_opendata *opendata;
2751	int ret;
2752
2753	opendata = nfs4_open_recoverdata_alloc(ctx, state, NFS4_OPEN_CLAIM_FH);
2754	if (IS_ERR(opendata))
2755		return PTR_ERR(opendata);
2756	/*
2757	 * We're not recovering a delegation, so ask for no delegation.
2758	 * Otherwise the recovery thread could deadlock with an outstanding
2759	 * delegation return.
2760	 */
2761	opendata->o_arg.open_flags = O_DIRECT;
2762	ret = nfs4_open_recover(opendata, state);
2763	if (ret == -ESTALE)
2764		d_drop(ctx->dentry);
2765	nfs4_opendata_put(opendata);
2766	return ret;
2767}
2768
2769static int nfs4_do_open_expired(struct nfs_open_context *ctx, struct nfs4_state *state)
2770{
2771	struct nfs_server *server = NFS_SERVER(state->inode);
2772	struct nfs4_exception exception = { };
2773	int err;
2774
2775	do {
2776		err = _nfs4_open_expired(ctx, state);
2777		trace_nfs4_open_expired(ctx, 0, err);
2778		if (nfs4_clear_cap_atomic_open_v1(server, err, &exception))
2779			continue;
2780		switch (err) {
2781		default:
2782			goto out;
2783		case -NFS4ERR_GRACE:
2784		case -NFS4ERR_DELAY:
2785			nfs4_handle_exception(server, err, &exception);
2786			err = 0;
2787		}
2788	} while (exception.retry);
2789out:
2790	return err;
2791}
2792
2793static int nfs4_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2794{
2795	struct nfs_open_context *ctx;
2796	int ret;
2797
2798	ctx = nfs4_state_find_open_context(state);
2799	if (IS_ERR(ctx))
2800		return -EAGAIN;
2801	ret = nfs4_do_open_expired(ctx, state);
2802	put_nfs_open_context(ctx);
2803	return ret;
2804}
2805
2806static void nfs_finish_clear_delegation_stateid(struct nfs4_state *state,
2807		const nfs4_stateid *stateid)
2808{
2809	nfs_remove_bad_delegation(state->inode, stateid);
2810	nfs_state_clear_delegation(state);
2811}
2812
2813static void nfs40_clear_delegation_stateid(struct nfs4_state *state)
2814{
2815	if (rcu_access_pointer(NFS_I(state->inode)->delegation) != NULL)
2816		nfs_finish_clear_delegation_stateid(state, NULL);
2817}
2818
2819static int nfs40_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
2820{
2821	/* NFSv4.0 doesn't allow for delegation recovery on open expire */
2822	nfs40_clear_delegation_stateid(state);
2823	nfs_state_clear_open_state_flags(state);
2824	return nfs4_open_expired(sp, state);
2825}
2826
2827static int nfs40_test_and_free_expired_stateid(struct nfs_server *server,
2828		nfs4_stateid *stateid,
2829		const struct cred *cred)
2830{
2831	return -NFS4ERR_BAD_STATEID;
2832}
2833
2834#if defined(CONFIG_NFS_V4_1)
2835static int nfs41_test_and_free_expired_stateid(struct nfs_server *server,
2836		nfs4_stateid *stateid,
2837		const struct cred *cred)
2838{
2839	int status;
2840
2841	switch (stateid->type) {
2842	default:
2843		break;
2844	case NFS4_INVALID_STATEID_TYPE:
2845	case NFS4_SPECIAL_STATEID_TYPE:
2846		return -NFS4ERR_BAD_STATEID;
2847	case NFS4_REVOKED_STATEID_TYPE:
2848		goto out_free;
2849	}
2850
2851	status = nfs41_test_stateid(server, stateid, cred);
2852	switch (status) {
2853	case -NFS4ERR_EXPIRED:
2854	case -NFS4ERR_ADMIN_REVOKED:
2855	case -NFS4ERR_DELEG_REVOKED:
2856		break;
2857	default:
2858		return status;
2859	}
2860out_free:
2861	/* Ack the revoked state to the server */
2862	nfs41_free_stateid(server, stateid, cred, true);
2863	return -NFS4ERR_EXPIRED;
2864}
2865
2866static int nfs41_check_delegation_stateid(struct nfs4_state *state)
2867{
2868	struct nfs_server *server = NFS_SERVER(state->inode);
2869	nfs4_stateid stateid;
2870	struct nfs_delegation *delegation;
2871	const struct cred *cred = NULL;
2872	int status, ret = NFS_OK;
2873
2874	/* Get the delegation credential for use by test/free_stateid */
2875	rcu_read_lock();
2876	delegation = rcu_dereference(NFS_I(state->inode)->delegation);
2877	if (delegation == NULL) {
2878		rcu_read_unlock();
2879		nfs_state_clear_delegation(state);
2880		return NFS_OK;
2881	}
2882
2883	spin_lock(&delegation->lock);
2884	nfs4_stateid_copy(&stateid, &delegation->stateid);
2885
2886	if (!test_and_clear_bit(NFS_DELEGATION_TEST_EXPIRED,
2887				&delegation->flags)) {
2888		spin_unlock(&delegation->lock);
2889		rcu_read_unlock();
2890		return NFS_OK;
2891	}
2892
2893	if (delegation->cred)
2894		cred = get_cred(delegation->cred);
2895	spin_unlock(&delegation->lock);
2896	rcu_read_unlock();
2897	status = nfs41_test_and_free_expired_stateid(server, &stateid, cred);
2898	trace_nfs4_test_delegation_stateid(state, NULL, status);
2899	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID)
2900		nfs_finish_clear_delegation_stateid(state, &stateid);
2901	else
2902		ret = status;
2903
2904	put_cred(cred);
2905	return ret;
2906}
2907
2908static void nfs41_delegation_recover_stateid(struct nfs4_state *state)
2909{
2910	nfs4_stateid tmp;
2911
2912	if (test_bit(NFS_DELEGATED_STATE, &state->flags) &&
2913	    nfs4_copy_delegation_stateid(state->inode, state->state,
2914				&tmp, NULL) &&
2915	    nfs4_stateid_match_other(&state->stateid, &tmp))
2916		nfs_state_set_delegation(state, &tmp, state->state);
2917	else
2918		nfs_state_clear_delegation(state);
2919}
2920
2921/**
2922 * nfs41_check_expired_locks - possibly free a lock stateid
2923 *
2924 * @state: NFSv4 state for an inode
2925 *
2926 * Returns NFS_OK if recovery for this stateid is now finished.
2927 * Otherwise a negative NFS4ERR value is returned.
2928 */
2929static int nfs41_check_expired_locks(struct nfs4_state *state)
2930{
2931	int status, ret = NFS_OK;
2932	struct nfs4_lock_state *lsp, *prev = NULL;
2933	struct nfs_server *server = NFS_SERVER(state->inode);
2934
2935	if (!test_bit(LK_STATE_IN_USE, &state->flags))
2936		goto out;
2937
2938	spin_lock(&state->state_lock);
2939	list_for_each_entry(lsp, &state->lock_states, ls_locks) {
2940		if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags)) {
2941			const struct cred *cred = lsp->ls_state->owner->so_cred;
2942
2943			refcount_inc(&lsp->ls_count);
2944			spin_unlock(&state->state_lock);
2945
2946			nfs4_put_lock_state(prev);
2947			prev = lsp;
2948
2949			status = nfs41_test_and_free_expired_stateid(server,
2950					&lsp->ls_stateid,
2951					cred);
2952			trace_nfs4_test_lock_stateid(state, lsp, status);
2953			if (status == -NFS4ERR_EXPIRED ||
2954			    status == -NFS4ERR_BAD_STATEID) {
2955				clear_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
2956				lsp->ls_stateid.type = NFS4_INVALID_STATEID_TYPE;
2957				if (!recover_lost_locks)
2958					set_bit(NFS_LOCK_LOST, &lsp->ls_flags);
2959			} else if (status != NFS_OK) {
2960				ret = status;
2961				nfs4_put_lock_state(prev);
2962				goto out;
2963			}
2964			spin_lock(&state->state_lock);
2965		}
2966	}
2967	spin_unlock(&state->state_lock);
2968	nfs4_put_lock_state(prev);
2969out:
2970	return ret;
2971}
2972
2973/**
2974 * nfs41_check_open_stateid - possibly free an open stateid
2975 *
2976 * @state: NFSv4 state for an inode
2977 *
2978 * Returns NFS_OK if recovery for this stateid is now finished.
2979 * Otherwise a negative NFS4ERR value is returned.
2980 */
2981static int nfs41_check_open_stateid(struct nfs4_state *state)
2982{
2983	struct nfs_server *server = NFS_SERVER(state->inode);
2984	nfs4_stateid *stateid = &state->open_stateid;
2985	const struct cred *cred = state->owner->so_cred;
2986	int status;
2987
2988	if (test_bit(NFS_OPEN_STATE, &state->flags) == 0)
2989		return -NFS4ERR_BAD_STATEID;
2990	status = nfs41_test_and_free_expired_stateid(server, stateid, cred);
2991	trace_nfs4_test_open_stateid(state, NULL, status);
2992	if (status == -NFS4ERR_EXPIRED || status == -NFS4ERR_BAD_STATEID) {
2993		nfs_state_clear_open_state_flags(state);
2994		stateid->type = NFS4_INVALID_STATEID_TYPE;
2995		return status;
2996	}
2997	if (nfs_open_stateid_recover_openmode(state))
2998		return -NFS4ERR_OPENMODE;
2999	return NFS_OK;
3000}
3001
3002static int nfs41_open_expired(struct nfs4_state_owner *sp, struct nfs4_state *state)
3003{
3004	int status;
3005
3006	status = nfs41_check_delegation_stateid(state);
3007	if (status != NFS_OK)
3008		return status;
3009	nfs41_delegation_recover_stateid(state);
3010
3011	status = nfs41_check_expired_locks(state);
3012	if (status != NFS_OK)
3013		return status;
3014	status = nfs41_check_open_stateid(state);
3015	if (status != NFS_OK)
3016		status = nfs4_open_expired(sp, state);
3017	return status;
3018}
3019#endif
3020
3021/*
3022 * on an EXCLUSIVE create, the server should send back a bitmask with FATTR4-*
3023 * fields corresponding to attributes that were used to store the verifier.
3024 * Make sure we clobber those fields in the later setattr call
3025 */
3026static unsigned nfs4_exclusive_attrset(struct nfs4_opendata *opendata,
3027				struct iattr *sattr, struct nfs4_label **label)
3028{
3029	const __u32 *bitmask = opendata->o_arg.server->exclcreat_bitmask;
3030	__u32 attrset[3];
3031	unsigned ret;
3032	unsigned i;
3033
3034	for (i = 0; i < ARRAY_SIZE(attrset); i++) {
3035		attrset[i] = opendata->o_res.attrset[i];
3036		if (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE4_1)
3037			attrset[i] &= ~bitmask[i];
3038	}
3039
3040	ret = (opendata->o_arg.createmode == NFS4_CREATE_EXCLUSIVE) ?
3041		sattr->ia_valid : 0;
3042
3043	if ((attrset[1] & (FATTR4_WORD1_TIME_ACCESS|FATTR4_WORD1_TIME_ACCESS_SET))) {
3044		if (sattr->ia_valid & ATTR_ATIME_SET)
3045			ret |= ATTR_ATIME_SET;
3046		else
3047			ret |= ATTR_ATIME;
3048	}
3049
3050	if ((attrset[1] & (FATTR4_WORD1_TIME_MODIFY|FATTR4_WORD1_TIME_MODIFY_SET))) {
3051		if (sattr->ia_valid & ATTR_MTIME_SET)
3052			ret |= ATTR_MTIME_SET;
3053		else
3054			ret |= ATTR_MTIME;
3055	}
3056
3057	if (!(attrset[2] & FATTR4_WORD2_SECURITY_LABEL))
3058		*label = NULL;
3059	return ret;
3060}
3061
3062static int _nfs4_open_and_get_state(struct nfs4_opendata *opendata,
3063		struct nfs_open_context *ctx)
3064{
3065	struct nfs4_state_owner *sp = opendata->owner;
3066	struct nfs_server *server = sp->so_server;
3067	struct dentry *dentry;
3068	struct nfs4_state *state;
3069	fmode_t acc_mode = _nfs4_ctx_to_accessmode(ctx);
3070	struct inode *dir = d_inode(opendata->dir);
3071	unsigned long dir_verifier;
3072	int ret;
3073
3074	dir_verifier = nfs_save_change_attribute(dir);
3075
3076	ret = _nfs4_proc_open(opendata, ctx);
3077	if (ret != 0)
3078		goto out;
3079
3080	state = _nfs4_opendata_to_nfs4_state(opendata);
3081	ret = PTR_ERR(state);
3082	if (IS_ERR(state))
3083		goto out;
3084	ctx->state = state;
3085	if (server->caps & NFS_CAP_POSIX_LOCK)
3086		set_bit(NFS_STATE_POSIX_LOCKS, &state->flags);
3087	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_MAY_NOTIFY_LOCK)
3088		set_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags);
3089	if (opendata->o_res.rflags & NFS4_OPEN_RESULT_PRESERVE_UNLINKED)
3090		set_bit(NFS_INO_PRESERVE_UNLINKED, &NFS_I(state->inode)->flags);
3091
3092	dentry = opendata->dentry;
3093	if (d_really_is_negative(dentry)) {
3094		struct dentry *alias;
3095		d_drop(dentry);
3096		alias = d_exact_alias(dentry, state->inode);
3097		if (!alias)
3098			alias = d_splice_alias(igrab(state->inode), dentry);
3099		/* d_splice_alias() can't fail here - it's a non-directory */
3100		if (alias) {
3101			dput(ctx->dentry);
3102			ctx->dentry = dentry = alias;
3103		}
3104	}
3105
3106	switch(opendata->o_arg.claim) {
3107	default:
3108		break;
3109	case NFS4_OPEN_CLAIM_NULL:
3110	case NFS4_OPEN_CLAIM_DELEGATE_CUR:
3111	case NFS4_OPEN_CLAIM_DELEGATE_PREV:
3112		if (!opendata->rpc_done)
3113			break;
3114		if (opendata->o_res.delegation_type != 0)
3115			dir_verifier = nfs_save_change_attribute(dir);
3116		nfs_set_verifier(dentry, dir_verifier);
3117	}
3118
3119	/* Parse layoutget results before we check for access */
3120	pnfs_parse_lgopen(state->inode, opendata->lgp, ctx);
3121
3122	ret = nfs4_opendata_access(sp->so_cred, opendata, state, acc_mode);
3123	if (ret != 0)
3124		goto out;
3125
3126	if (d_inode(dentry) == state->inode)
3127		nfs_inode_attach_open_context(ctx);
3128
3129out:
3130	if (!opendata->cancelled) {
3131		if (opendata->lgp) {
3132			nfs4_lgopen_release(opendata->lgp);
3133			opendata->lgp = NULL;
3134		}
3135		nfs4_sequence_free_slot(&opendata->o_res.seq_res);
3136	}
3137	return ret;
3138}
3139
3140/*
3141 * Returns a referenced nfs4_state
3142 */
3143static int _nfs4_do_open(struct inode *dir,
3144			struct nfs_open_context *ctx,
3145			int flags,
3146			const struct nfs4_open_createattrs *c,
3147			int *opened)
3148{
3149	struct nfs4_state_owner  *sp;
3150	struct nfs4_state     *state = NULL;
3151	struct nfs_server       *server = NFS_SERVER(dir);
3152	struct nfs4_opendata *opendata;
3153	struct dentry *dentry = ctx->dentry;
3154	const struct cred *cred = ctx->cred;
3155	struct nfs4_threshold **ctx_th = &ctx->mdsthreshold;
3156	fmode_t fmode = _nfs4_ctx_to_openmode(ctx);
3157	enum open_claim_type4 claim = NFS4_OPEN_CLAIM_NULL;
3158	struct iattr *sattr = c->sattr;
3159	struct nfs4_label *label = c->label;
3160	int status;
3161
3162	/* Protect against reboot recovery conflicts */
3163	status = -ENOMEM;
3164	sp = nfs4_get_state_owner(server, cred, GFP_KERNEL);
3165	if (sp == NULL) {
3166		dprintk("nfs4_do_open: nfs4_get_state_owner failed!\n");
3167		goto out_err;
3168	}
3169	status = nfs4_client_recover_expired_lease(server->nfs_client);
3170	if (status != 0)
3171		goto err_put_state_owner;
3172	if (d_really_is_positive(dentry))
3173		nfs4_return_incompatible_delegation(d_inode(dentry), fmode);
3174	status = -ENOMEM;
3175	if (d_really_is_positive(dentry))
3176		claim = NFS4_OPEN_CLAIM_FH;
3177	opendata = nfs4_opendata_alloc(dentry, sp, fmode, flags,
3178			c, claim, GFP_KERNEL);
3179	if (opendata == NULL)
3180		goto err_put_state_owner;
3181
3182	if (server->attr_bitmask[2] & FATTR4_WORD2_MDSTHRESHOLD) {
3183		if (!opendata->f_attr.mdsthreshold) {
3184			opendata->f_attr.mdsthreshold = pnfs_mdsthreshold_alloc();
3185			if (!opendata->f_attr.mdsthreshold)
3186				goto err_opendata_put;
3187		}
3188		opendata->o_arg.open_bitmap = &nfs4_pnfs_open_bitmap[0];
3189	}
3190	if (d_really_is_positive(dentry))
3191		opendata->state = nfs4_get_open_state(d_inode(dentry), sp);
3192
3193	status = _nfs4_open_and_get_state(opendata, ctx);
3194	if (status != 0)
3195		goto err_opendata_put;
3196	state = ctx->state;
3197
3198	if ((opendata->o_arg.open_flags & (O_CREAT|O_EXCL)) == (O_CREAT|O_EXCL) &&
3199	    (opendata->o_arg.createmode != NFS4_CREATE_GUARDED)) {
3200		unsigned attrs = nfs4_exclusive_attrset(opendata, sattr, &label);
3201		/*
3202		 * send create attributes which was not set by open
3203		 * with an extra setattr.
3204		 */
3205		if (attrs || label) {
3206			unsigned ia_old = sattr->ia_valid;
3207
3208			sattr->ia_valid = attrs;
3209			nfs_fattr_init(opendata->o_res.f_attr);
3210			status = nfs4_do_setattr(state->inode, cred,
3211					opendata->o_res.f_attr, sattr,
3212					ctx, label);
3213			if (status == 0) {
3214				nfs_setattr_update_inode(state->inode, sattr,
3215						opendata->o_res.f_attr);
3216				nfs_setsecurity(state->inode, opendata->o_res.f_attr);
3217			}
3218			sattr->ia_valid = ia_old;
3219		}
3220	}
3221	if (opened && opendata->file_created)
3222		*opened = 1;
3223
3224	if (pnfs_use_threshold(ctx_th, opendata->f_attr.mdsthreshold, server)) {
3225		*ctx_th = opendata->f_attr.mdsthreshold;
3226		opendata->f_attr.mdsthreshold = NULL;
3227	}
3228
3229	nfs4_opendata_put(opendata);
3230	nfs4_put_state_owner(sp);
3231	return 0;
3232err_opendata_put:
3233	nfs4_opendata_put(opendata);
3234err_put_state_owner:
3235	nfs4_put_state_owner(sp);
3236out_err:
3237	return status;
3238}
3239
3240
3241static struct nfs4_state *nfs4_do_open(struct inode *dir,
3242					struct nfs_open_context *ctx,
3243					int flags,
3244					struct iattr *sattr,
3245					struct nfs4_label *label,
3246					int *opened)
3247{
3248	struct nfs_server *server = NFS_SERVER(dir);
3249	struct nfs4_exception exception = {
3250		.interruptible = true,
3251	};
3252	struct nfs4_state *res;
3253	struct nfs4_open_createattrs c = {
3254		.label = label,
3255		.sattr = sattr,
3256		.verf = {
3257			[0] = (__u32)jiffies,
3258			[1] = (__u32)current->pid,
3259		},
3260	};
3261	int status;
3262
3263	do {
3264		status = _nfs4_do_open(dir, ctx, flags, &c, opened);
3265		res = ctx->state;
3266		trace_nfs4_open_file(ctx, flags, status);
3267		if (status == 0)
3268			break;
3269		/* NOTE: BAD_SEQID means the server and client disagree about the
3270		 * book-keeping w.r.t. state-changing operations
3271		 * (OPEN/CLOSE/LOCK/LOCKU...)
3272		 * It is actually a sign of a bug on the client or on the server.
3273		 *
3274		 * If we receive a BAD_SEQID error in the particular case of
3275		 * doing an OPEN, we assume that nfs_increment_open_seqid() will
3276		 * have unhashed the old state_owner for us, and that we can
3277		 * therefore safely retry using a new one. We should still warn
3278		 * the user though...
3279		 */
3280		if (status == -NFS4ERR_BAD_SEQID) {
3281			pr_warn_ratelimited("NFS: v4 server %s "
3282					" returned a bad sequence-id error!\n",
3283					NFS_SERVER(dir)->nfs_client->cl_hostname);
3284			exception.retry = 1;
3285			continue;
3286		}
3287		/*
3288		 * BAD_STATEID on OPEN means that the server cancelled our
3289		 * state before it received the OPEN_CONFIRM.
3290		 * Recover by retrying the request as per the discussion
3291		 * on Page 181 of RFC3530.
3292		 */
3293		if (status == -NFS4ERR_BAD_STATEID) {
3294			exception.retry = 1;
3295			continue;
3296		}
3297		if (status == -NFS4ERR_EXPIRED) {
3298			nfs4_schedule_lease_recovery(server->nfs_client);
3299			exception.retry = 1;
3300			continue;
3301		}
3302		if (status == -EAGAIN) {
3303			/* We must have found a delegation */
3304			exception.retry = 1;
3305			continue;
3306		}
3307		if (nfs4_clear_cap_atomic_open_v1(server, status, &exception))
3308			continue;
3309		res = ERR_PTR(nfs4_handle_exception(server,
3310					status, &exception));
3311	} while (exception.retry);
3312	return res;
3313}
3314
3315static int _nfs4_do_setattr(struct inode *inode,
3316			    struct nfs_setattrargs *arg,
3317			    struct nfs_setattrres *res,
3318			    const struct cred *cred,
3319			    struct nfs_open_context *ctx)
3320{
3321	struct nfs_server *server = NFS_SERVER(inode);
3322	struct rpc_message msg = {
3323		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
3324		.rpc_argp	= arg,
3325		.rpc_resp	= res,
3326		.rpc_cred	= cred,
3327	};
3328	const struct cred *delegation_cred = NULL;
3329	unsigned long timestamp = jiffies;
3330	bool truncate;
3331	int status;
3332
3333	nfs_fattr_init(res->fattr);
3334
3335	/* Servers should only apply open mode checks for file size changes */
3336	truncate = (arg->iap->ia_valid & ATTR_SIZE) ? true : false;
3337	if (!truncate) {
3338		nfs4_inode_make_writeable(inode);
3339		goto zero_stateid;
3340	}
3341
3342	if (nfs4_copy_delegation_stateid(inode, FMODE_WRITE, &arg->stateid, &delegation_cred)) {
3343		/* Use that stateid */
3344	} else if (ctx != NULL && ctx->state) {
3345		struct nfs_lock_context *l_ctx;
3346		if (!nfs4_valid_open_stateid(ctx->state))
3347			return -EBADF;
3348		l_ctx = nfs_get_lock_context(ctx);
3349		if (IS_ERR(l_ctx))
3350			return PTR_ERR(l_ctx);
3351		status = nfs4_select_rw_stateid(ctx->state, FMODE_WRITE, l_ctx,
3352						&arg->stateid, &delegation_cred);
3353		nfs_put_lock_context(l_ctx);
3354		if (status == -EIO)
3355			return -EBADF;
3356		else if (status == -EAGAIN)
3357			goto zero_stateid;
3358	} else {
3359zero_stateid:
3360		nfs4_stateid_copy(&arg->stateid, &zero_stateid);
3361	}
3362	if (delegation_cred)
3363		msg.rpc_cred = delegation_cred;
3364
3365	status = nfs4_call_sync(server->client, server, &msg, &arg->seq_args, &res->seq_res, 1);
3366
3367	put_cred(delegation_cred);
3368	if (status == 0 && ctx != NULL)
3369		renew_lease(server, timestamp);
3370	trace_nfs4_setattr(inode, &arg->stateid, status);
3371	return status;
3372}
3373
3374static int nfs4_do_setattr(struct inode *inode, const struct cred *cred,
3375			   struct nfs_fattr *fattr, struct iattr *sattr,
3376			   struct nfs_open_context *ctx, struct nfs4_label *ilabel)
3377{
3378	struct nfs_server *server = NFS_SERVER(inode);
3379	__u32 bitmask[NFS4_BITMASK_SZ];
3380	struct nfs4_state *state = ctx ? ctx->state : NULL;
3381	struct nfs_setattrargs	arg = {
3382		.fh		= NFS_FH(inode),
3383		.iap		= sattr,
3384		.server		= server,
3385		.bitmask = bitmask,
3386		.label		= ilabel,
3387	};
3388	struct nfs_setattrres  res = {
3389		.fattr		= fattr,
3390		.server		= server,
3391	};
3392	struct nfs4_exception exception = {
3393		.state = state,
3394		.inode = inode,
3395		.stateid = &arg.stateid,
3396	};
3397	unsigned long adjust_flags = NFS_INO_INVALID_CHANGE;
3398	int err;
3399
3400	if (sattr->ia_valid & (ATTR_MODE | ATTR_KILL_SUID | ATTR_KILL_SGID))
3401		adjust_flags |= NFS_INO_INVALID_MODE;
3402	if (sattr->ia_valid & (ATTR_UID | ATTR_GID))
3403		adjust_flags |= NFS_INO_INVALID_OTHER;
3404
3405	do {
3406		nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label),
3407					inode, adjust_flags);
3408
3409		err = _nfs4_do_setattr(inode, &arg, &res, cred, ctx);
3410		switch (err) {
3411		case -NFS4ERR_OPENMODE:
3412			if (!(sattr->ia_valid & ATTR_SIZE)) {
3413				pr_warn_once("NFSv4: server %s is incorrectly "
3414						"applying open mode checks to "
3415						"a SETATTR that is not "
3416						"changing file size.\n",
3417						server->nfs_client->cl_hostname);
3418			}
3419			if (state && !(state->state & FMODE_WRITE)) {
3420				err = -EBADF;
3421				if (sattr->ia_valid & ATTR_OPEN)
3422					err = -EACCES;
3423				goto out;
3424			}
3425		}
3426		err = nfs4_handle_exception(server, err, &exception);
3427	} while (exception.retry);
3428out:
3429	return err;
3430}
3431
3432static bool
3433nfs4_wait_on_layoutreturn(struct inode *inode, struct rpc_task *task)
3434{
3435	if (inode == NULL || !nfs_have_layout(inode))
3436		return false;
3437
3438	return pnfs_wait_on_layoutreturn(inode, task);
3439}
3440
3441/*
3442 * Update the seqid of an open stateid
3443 */
3444static void nfs4_sync_open_stateid(nfs4_stateid *dst,
3445		struct nfs4_state *state)
3446{
3447	__be32 seqid_open;
3448	u32 dst_seqid;
3449	int seq;
3450
3451	for (;;) {
3452		if (!nfs4_valid_open_stateid(state))
3453			break;
3454		seq = read_seqbegin(&state->seqlock);
3455		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3456			nfs4_stateid_copy(dst, &state->open_stateid);
3457			if (read_seqretry(&state->seqlock, seq))
3458				continue;
3459			break;
3460		}
3461		seqid_open = state->open_stateid.seqid;
3462		if (read_seqretry(&state->seqlock, seq))
3463			continue;
3464
3465		dst_seqid = be32_to_cpu(dst->seqid);
3466		if ((s32)(dst_seqid - be32_to_cpu(seqid_open)) < 0)
3467			dst->seqid = seqid_open;
3468		break;
3469	}
3470}
3471
3472/*
3473 * Update the seqid of an open stateid after receiving
3474 * NFS4ERR_OLD_STATEID
3475 */
3476static bool nfs4_refresh_open_old_stateid(nfs4_stateid *dst,
3477		struct nfs4_state *state)
3478{
3479	__be32 seqid_open;
3480	u32 dst_seqid;
3481	bool ret;
3482	int seq, status = -EAGAIN;
3483	DEFINE_WAIT(wait);
3484
3485	for (;;) {
3486		ret = false;
3487		if (!nfs4_valid_open_stateid(state))
3488			break;
3489		seq = read_seqbegin(&state->seqlock);
3490		if (!nfs4_state_match_open_stateid_other(state, dst)) {
3491			if (read_seqretry(&state->seqlock, seq))
3492				continue;
3493			break;
3494		}
3495
3496		write_seqlock(&state->seqlock);
3497		seqid_open = state->open_stateid.seqid;
3498
3499		dst_seqid = be32_to_cpu(dst->seqid);
3500
3501		/* Did another OPEN bump the state's seqid?  try again: */
3502		if ((s32)(be32_to_cpu(seqid_open) - dst_seqid) > 0) {
3503			dst->seqid = seqid_open;
3504			write_sequnlock(&state->seqlock);
3505			ret = true;
3506			break;
3507		}
3508
3509		/* server says we're behind but we haven't seen the update yet */
3510		set_bit(NFS_STATE_CHANGE_WAIT, &state->flags);
3511		prepare_to_wait(&state->waitq, &wait, TASK_KILLABLE);
3512		write_sequnlock(&state->seqlock);
3513		trace_nfs4_close_stateid_update_wait(state->inode, dst, 0);
3514
3515		if (fatal_signal_pending(current))
3516			status = -EINTR;
3517		else
3518			if (schedule_timeout(5*HZ) != 0)
3519				status = 0;
3520
3521		finish_wait(&state->waitq, &wait);
3522
3523		if (!status)
3524			continue;
3525		if (status == -EINTR)
3526			break;
3527
3528		/* we slept the whole 5 seconds, we must have lost a seqid */
3529		dst->seqid = cpu_to_be32(dst_seqid + 1);
3530		ret = true;
3531		break;
3532	}
3533
3534	return ret;
3535}
3536
3537struct nfs4_closedata {
3538	struct inode *inode;
3539	struct nfs4_state *state;
3540	struct nfs_closeargs arg;
3541	struct nfs_closeres res;
3542	struct {
3543		struct nfs4_layoutreturn_args arg;
3544		struct nfs4_layoutreturn_res res;
3545		struct nfs4_xdr_opaque_data ld_private;
3546		u32 roc_barrier;
3547		bool roc;
3548	} lr;
3549	struct nfs_fattr fattr;
3550	unsigned long timestamp;
3551};
3552
3553static void nfs4_free_closedata(void *data)
3554{
3555	struct nfs4_closedata *calldata = data;
3556	struct nfs4_state_owner *sp = calldata->state->owner;
3557	struct super_block *sb = calldata->state->inode->i_sb;
3558
3559	if (calldata->lr.roc)
3560		pnfs_roc_release(&calldata->lr.arg, &calldata->lr.res,
3561				calldata->res.lr_ret);
3562	nfs4_put_open_state(calldata->state);
3563	nfs_free_seqid(calldata->arg.seqid);
3564	nfs4_put_state_owner(sp);
3565	nfs_sb_deactive(sb);
3566	kfree(calldata);
3567}
3568
3569static void nfs4_close_done(struct rpc_task *task, void *data)
3570{
3571	struct nfs4_closedata *calldata = data;
3572	struct nfs4_state *state = calldata->state;
3573	struct nfs_server *server = NFS_SERVER(calldata->inode);
3574	nfs4_stateid *res_stateid = NULL;
3575	struct nfs4_exception exception = {
3576		.state = state,
3577		.inode = calldata->inode,
3578		.stateid = &calldata->arg.stateid,
3579	};
3580
3581	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
3582		return;
3583	trace_nfs4_close(state, &calldata->arg, &calldata->res, task->tk_status);
3584
3585	/* Handle Layoutreturn errors */
3586	if (pnfs_roc_done(task, &calldata->arg.lr_args, &calldata->res.lr_res,
3587			  &calldata->res.lr_ret) == -EAGAIN)
3588		goto out_restart;
3589
3590	/* hmm. we are done with the inode, and in the process of freeing
3591	 * the state_owner. we keep this around to process errors
3592	 */
3593	switch (task->tk_status) {
3594		case 0:
3595			res_stateid = &calldata->res.stateid;
3596			renew_lease(server, calldata->timestamp);
3597			break;
3598		case -NFS4ERR_ACCESS:
3599			if (calldata->arg.bitmask != NULL) {
3600				calldata->arg.bitmask = NULL;
3601				calldata->res.fattr = NULL;
3602				goto out_restart;
3603
3604			}
3605			break;
3606		case -NFS4ERR_OLD_STATEID:
3607			/* Did we race with OPEN? */
3608			if (nfs4_refresh_open_old_stateid(&calldata->arg.stateid,
3609						state))
3610				goto out_restart;
3611			goto out_release;
3612		case -NFS4ERR_ADMIN_REVOKED:
3613		case -NFS4ERR_STALE_STATEID:
3614		case -NFS4ERR_EXPIRED:
3615			nfs4_free_revoked_stateid(server,
3616					&calldata->arg.stateid,
3617					task->tk_msg.rpc_cred);
3618			fallthrough;
3619		case -NFS4ERR_BAD_STATEID:
3620			if (calldata->arg.fmode == 0)
3621				break;
3622			fallthrough;
3623		default:
3624			task->tk_status = nfs4_async_handle_exception(task,
3625					server, task->tk_status, &exception);
3626			if (exception.retry)
3627				goto out_restart;
3628	}
3629	nfs_clear_open_stateid(state, &calldata->arg.stateid,
3630			res_stateid, calldata->arg.fmode);
3631out_release:
3632	task->tk_status = 0;
3633	nfs_release_seqid(calldata->arg.seqid);
3634	nfs_refresh_inode(calldata->inode, &calldata->fattr);
3635	dprintk("%s: ret = %d\n", __func__, task->tk_status);
3636	return;
3637out_restart:
3638	task->tk_status = 0;
3639	rpc_restart_call_prepare(task);
3640	goto out_release;
3641}
3642
3643static void nfs4_close_prepare(struct rpc_task *task, void *data)
3644{
3645	struct nfs4_closedata *calldata = data;
3646	struct nfs4_state *state = calldata->state;
3647	struct inode *inode = calldata->inode;
3648	struct nfs_server *server = NFS_SERVER(inode);
3649	struct pnfs_layout_hdr *lo;
3650	bool is_rdonly, is_wronly, is_rdwr;
3651	int call_close = 0;
3652
3653	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
3654		goto out_wait;
3655
3656	task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OPEN_DOWNGRADE];
3657	spin_lock(&state->owner->so_lock);
3658	is_rdwr = test_bit(NFS_O_RDWR_STATE, &state->flags);
3659	is_rdonly = test_bit(NFS_O_RDONLY_STATE, &state->flags);
3660	is_wronly = test_bit(NFS_O_WRONLY_STATE, &state->flags);
3661	/* Calculate the change in open mode */
3662	calldata->arg.fmode = 0;
3663	if (state->n_rdwr == 0) {
3664		if (state->n_rdonly == 0)
3665			call_close |= is_rdonly;
3666		else if (is_rdonly)
3667			calldata->arg.fmode |= FMODE_READ;
3668		if (state->n_wronly == 0)
3669			call_close |= is_wronly;
3670		else if (is_wronly)
3671			calldata->arg.fmode |= FMODE_WRITE;
3672		if (calldata->arg.fmode != (FMODE_READ|FMODE_WRITE))
3673			call_close |= is_rdwr;
3674	} else if (is_rdwr)
3675		calldata->arg.fmode |= FMODE_READ|FMODE_WRITE;
3676
3677	nfs4_sync_open_stateid(&calldata->arg.stateid, state);
3678	if (!nfs4_valid_open_stateid(state))
3679		call_close = 0;
3680	spin_unlock(&state->owner->so_lock);
3681
3682	if (!call_close) {
3683		/* Note: exit _without_ calling nfs4_close_done */
3684		goto out_no_action;
3685	}
3686
3687	if (!calldata->lr.roc && nfs4_wait_on_layoutreturn(inode, task)) {
3688		nfs_release_seqid(calldata->arg.seqid);
3689		goto out_wait;
3690	}
3691
3692	lo = calldata->arg.lr_args ? calldata->arg.lr_args->layout : NULL;
3693	if (lo && !pnfs_layout_is_valid(lo)) {
3694		calldata->arg.lr_args = NULL;
3695		calldata->res.lr_res = NULL;
3696	}
3697
3698	if (calldata->arg.fmode == 0)
3699		task->tk_msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE];
3700
3701	if (calldata->arg.fmode == 0 || calldata->arg.fmode == FMODE_READ) {
3702		/* Close-to-open cache consistency revalidation */
3703		if (!nfs4_have_delegation(inode, FMODE_READ)) {
3704			nfs4_bitmask_set(calldata->arg.bitmask_store,
3705					 server->cache_consistency_bitmask,
3706					 inode, 0);
3707			calldata->arg.bitmask = calldata->arg.bitmask_store;
3708		} else
3709			calldata->arg.bitmask = NULL;
3710	}
3711
3712	calldata->arg.share_access =
3713		nfs4_map_atomic_open_share(NFS_SERVER(inode),
3714				calldata->arg.fmode, 0);
3715
3716	if (calldata->res.fattr == NULL)
3717		calldata->arg.bitmask = NULL;
3718	else if (calldata->arg.bitmask == NULL)
3719		calldata->res.fattr = NULL;
3720	calldata->timestamp = jiffies;
3721	if (nfs4_setup_sequence(NFS_SERVER(inode)->nfs_client,
3722				&calldata->arg.seq_args,
3723				&calldata->res.seq_res,
3724				task) != 0)
3725		nfs_release_seqid(calldata->arg.seqid);
3726	return;
3727out_no_action:
3728	task->tk_action = NULL;
3729out_wait:
3730	nfs4_sequence_done(task, &calldata->res.seq_res);
3731}
3732
3733static const struct rpc_call_ops nfs4_close_ops = {
3734	.rpc_call_prepare = nfs4_close_prepare,
3735	.rpc_call_done = nfs4_close_done,
3736	.rpc_release = nfs4_free_closedata,
3737};
3738
3739/*
3740 * It is possible for data to be read/written from a mem-mapped file
3741 * after the sys_close call (which hits the vfs layer as a flush).
3742 * This means that we can't safely call nfsv4 close on a file until
3743 * the inode is cleared. This in turn means that we are not good
3744 * NFSv4 citizens - we do not indicate to the server to update the file's
3745 * share state even when we are done with one of the three share
3746 * stateid's in the inode.
3747 *
3748 * NOTE: Caller must be holding the sp->so_owner semaphore!
3749 */
3750int nfs4_do_close(struct nfs4_state *state, gfp_t gfp_mask, int wait)
3751{
3752	struct nfs_server *server = NFS_SERVER(state->inode);
3753	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
3754	struct nfs4_closedata *calldata;
3755	struct nfs4_state_owner *sp = state->owner;
3756	struct rpc_task *task;
3757	struct rpc_message msg = {
3758		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLOSE],
3759		.rpc_cred = state->owner->so_cred,
3760	};
3761	struct rpc_task_setup task_setup_data = {
3762		.rpc_client = server->client,
3763		.rpc_message = &msg,
3764		.callback_ops = &nfs4_close_ops,
3765		.workqueue = nfsiod_workqueue,
3766		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
3767	};
3768	int status = -ENOMEM;
3769
3770	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
3771		task_setup_data.flags |= RPC_TASK_MOVEABLE;
3772
3773	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_CLEANUP,
3774		&task_setup_data.rpc_client, &msg);
3775
3776	calldata = kzalloc(sizeof(*calldata), gfp_mask);
3777	if (calldata == NULL)
3778		goto out;
3779	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 1, 0);
3780	calldata->inode = state->inode;
3781	calldata->state = state;
3782	calldata->arg.fh = NFS_FH(state->inode);
3783	if (!nfs4_copy_open_stateid(&calldata->arg.stateid, state))
3784		goto out_free_calldata;
3785	/* Serialization for the sequence id */
3786	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
3787	calldata->arg.seqid = alloc_seqid(&state->owner->so_seqid, gfp_mask);
3788	if (IS_ERR(calldata->arg.seqid))
3789		goto out_free_calldata;
3790	nfs_fattr_init(&calldata->fattr);
3791	calldata->arg.fmode = 0;
3792	calldata->lr.arg.ld_private = &calldata->lr.ld_private;
3793	calldata->res.fattr = &calldata->fattr;
3794	calldata->res.seqid = calldata->arg.seqid;
3795	calldata->res.server = server;
3796	calldata->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
3797	calldata->lr.roc = pnfs_roc(state->inode,
3798			&calldata->lr.arg, &calldata->lr.res, msg.rpc_cred);
3799	if (calldata->lr.roc) {
3800		calldata->arg.lr_args = &calldata->lr.arg;
3801		calldata->res.lr_res = &calldata->lr.res;
3802	}
3803	nfs_sb_active(calldata->inode->i_sb);
3804
3805	msg.rpc_argp = &calldata->arg;
3806	msg.rpc_resp = &calldata->res;
3807	task_setup_data.callback_data = calldata;
3808	task = rpc_run_task(&task_setup_data);
3809	if (IS_ERR(task))
3810		return PTR_ERR(task);
3811	status = 0;
3812	if (wait)
3813		status = rpc_wait_for_completion_task(task);
3814	rpc_put_task(task);
3815	return status;
3816out_free_calldata:
3817	kfree(calldata);
3818out:
3819	nfs4_put_open_state(state);
3820	nfs4_put_state_owner(sp);
3821	return status;
3822}
3823
3824static struct inode *
3825nfs4_atomic_open(struct inode *dir, struct nfs_open_context *ctx,
3826		int open_flags, struct iattr *attr, int *opened)
3827{
3828	struct nfs4_state *state;
3829	struct nfs4_label l, *label;
3830
3831	label = nfs4_label_init_security(dir, ctx->dentry, attr, &l);
3832
3833	/* Protect against concurrent sillydeletes */
3834	state = nfs4_do_open(dir, ctx, open_flags, attr, label, opened);
3835
3836	nfs4_label_release_security(label);
3837
3838	if (IS_ERR(state))
3839		return ERR_CAST(state);
3840	return state->inode;
3841}
3842
3843static void nfs4_close_context(struct nfs_open_context *ctx, int is_sync)
3844{
3845	if (ctx->state == NULL)
3846		return;
3847	if (is_sync)
3848		nfs4_close_sync(ctx->state, _nfs4_ctx_to_openmode(ctx));
3849	else
3850		nfs4_close_state(ctx->state, _nfs4_ctx_to_openmode(ctx));
3851}
3852
3853#define FATTR4_WORD1_NFS40_MASK (2*FATTR4_WORD1_MOUNTED_ON_FILEID - 1UL)
3854#define FATTR4_WORD2_NFS41_MASK (2*FATTR4_WORD2_SUPPATTR_EXCLCREAT - 1UL)
3855#define FATTR4_WORD2_NFS42_MASK (2*FATTR4_WORD2_XATTR_SUPPORT - 1UL)
3856
3857static int _nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3858{
3859	u32 bitmask[3] = {}, minorversion = server->nfs_client->cl_minorversion;
3860	struct nfs4_server_caps_arg args = {
3861		.fhandle = fhandle,
3862		.bitmask = bitmask,
3863	};
3864	struct nfs4_server_caps_res res = {};
3865	struct rpc_message msg = {
3866		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SERVER_CAPS],
3867		.rpc_argp = &args,
3868		.rpc_resp = &res,
3869	};
3870	int status;
3871	int i;
3872
3873	bitmask[0] = FATTR4_WORD0_SUPPORTED_ATTRS |
3874		     FATTR4_WORD0_FH_EXPIRE_TYPE |
3875		     FATTR4_WORD0_LINK_SUPPORT |
3876		     FATTR4_WORD0_SYMLINK_SUPPORT |
3877		     FATTR4_WORD0_ACLSUPPORT |
3878		     FATTR4_WORD0_CASE_INSENSITIVE |
3879		     FATTR4_WORD0_CASE_PRESERVING;
3880	if (minorversion)
3881		bitmask[2] = FATTR4_WORD2_SUPPATTR_EXCLCREAT;
3882
3883	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
3884	if (status == 0) {
3885		/* Sanity check the server answers */
3886		switch (minorversion) {
3887		case 0:
3888			res.attr_bitmask[1] &= FATTR4_WORD1_NFS40_MASK;
3889			res.attr_bitmask[2] = 0;
3890			break;
3891		case 1:
3892			res.attr_bitmask[2] &= FATTR4_WORD2_NFS41_MASK;
3893			break;
3894		case 2:
3895			res.attr_bitmask[2] &= FATTR4_WORD2_NFS42_MASK;
3896		}
3897		memcpy(server->attr_bitmask, res.attr_bitmask, sizeof(server->attr_bitmask));
3898		server->caps &= ~(NFS_CAP_ACLS | NFS_CAP_HARDLINKS |
3899				  NFS_CAP_SYMLINKS| NFS_CAP_SECURITY_LABEL);
3900		server->fattr_valid = NFS_ATTR_FATTR_V4;
3901		if (res.attr_bitmask[0] & FATTR4_WORD0_ACL &&
3902				res.acl_bitmask & ACL4_SUPPORT_ALLOW_ACL)
3903			server->caps |= NFS_CAP_ACLS;
3904		if (res.has_links != 0)
3905			server->caps |= NFS_CAP_HARDLINKS;
3906		if (res.has_symlinks != 0)
3907			server->caps |= NFS_CAP_SYMLINKS;
3908		if (res.case_insensitive)
3909			server->caps |= NFS_CAP_CASE_INSENSITIVE;
3910		if (res.case_preserving)
3911			server->caps |= NFS_CAP_CASE_PRESERVING;
3912#ifdef CONFIG_NFS_V4_SECURITY_LABEL
3913		if (res.attr_bitmask[2] & FATTR4_WORD2_SECURITY_LABEL)
3914			server->caps |= NFS_CAP_SECURITY_LABEL;
3915#endif
3916		if (res.attr_bitmask[0] & FATTR4_WORD0_FS_LOCATIONS)
3917			server->caps |= NFS_CAP_FS_LOCATIONS;
3918		if (!(res.attr_bitmask[0] & FATTR4_WORD0_FILEID))
3919			server->fattr_valid &= ~NFS_ATTR_FATTR_FILEID;
3920		if (!(res.attr_bitmask[1] & FATTR4_WORD1_MODE))
3921			server->fattr_valid &= ~NFS_ATTR_FATTR_MODE;
3922		if (!(res.attr_bitmask[1] & FATTR4_WORD1_NUMLINKS))
3923			server->fattr_valid &= ~NFS_ATTR_FATTR_NLINK;
3924		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER))
3925			server->fattr_valid &= ~(NFS_ATTR_FATTR_OWNER |
3926				NFS_ATTR_FATTR_OWNER_NAME);
3927		if (!(res.attr_bitmask[1] & FATTR4_WORD1_OWNER_GROUP))
3928			server->fattr_valid &= ~(NFS_ATTR_FATTR_GROUP |
3929				NFS_ATTR_FATTR_GROUP_NAME);
3930		if (!(res.attr_bitmask[1] & FATTR4_WORD1_SPACE_USED))
3931			server->fattr_valid &= ~NFS_ATTR_FATTR_SPACE_USED;
3932		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_ACCESS))
3933			server->fattr_valid &= ~NFS_ATTR_FATTR_ATIME;
3934		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_METADATA))
3935			server->fattr_valid &= ~NFS_ATTR_FATTR_CTIME;
3936		if (!(res.attr_bitmask[1] & FATTR4_WORD1_TIME_MODIFY))
3937			server->fattr_valid &= ~NFS_ATTR_FATTR_MTIME;
3938		memcpy(server->attr_bitmask_nl, res.attr_bitmask,
3939				sizeof(server->attr_bitmask));
3940		server->attr_bitmask_nl[2] &= ~FATTR4_WORD2_SECURITY_LABEL;
3941
3942		memcpy(server->cache_consistency_bitmask, res.attr_bitmask, sizeof(server->cache_consistency_bitmask));
3943		server->cache_consistency_bitmask[0] &= FATTR4_WORD0_CHANGE|FATTR4_WORD0_SIZE;
3944		server->cache_consistency_bitmask[1] &= FATTR4_WORD1_TIME_METADATA|FATTR4_WORD1_TIME_MODIFY;
3945		server->cache_consistency_bitmask[2] = 0;
3946
3947		/* Avoid a regression due to buggy server */
3948		for (i = 0; i < ARRAY_SIZE(res.exclcreat_bitmask); i++)
3949			res.exclcreat_bitmask[i] &= res.attr_bitmask[i];
3950		memcpy(server->exclcreat_bitmask, res.exclcreat_bitmask,
3951			sizeof(server->exclcreat_bitmask));
3952
3953		server->acl_bitmask = res.acl_bitmask;
3954		server->fh_expire_type = res.fh_expire_type;
3955	}
3956
3957	return status;
3958}
3959
3960int nfs4_server_capabilities(struct nfs_server *server, struct nfs_fh *fhandle)
3961{
3962	struct nfs4_exception exception = {
3963		.interruptible = true,
3964	};
3965	int err;
3966
3967	nfs4_server_set_init_caps(server);
3968	do {
3969		err = nfs4_handle_exception(server,
3970				_nfs4_server_capabilities(server, fhandle),
3971				&exception);
3972	} while (exception.retry);
3973	return err;
3974}
3975
3976static void test_fs_location_for_trunking(struct nfs4_fs_location *location,
3977					  struct nfs_client *clp,
3978					  struct nfs_server *server)
3979{
3980	int i;
3981
3982	for (i = 0; i < location->nservers; i++) {
3983		struct nfs4_string *srv_loc = &location->servers[i];
3984		struct sockaddr_storage addr;
3985		size_t addrlen;
3986		struct xprt_create xprt_args = {
3987			.ident = 0,
3988			.net = clp->cl_net,
3989		};
3990		struct nfs4_add_xprt_data xprtdata = {
3991			.clp = clp,
3992		};
3993		struct rpc_add_xprt_test rpcdata = {
3994			.add_xprt_test = clp->cl_mvops->session_trunk,
3995			.data = &xprtdata,
3996		};
3997		char *servername = NULL;
3998
3999		if (!srv_loc->len)
4000			continue;
4001
4002		addrlen = nfs_parse_server_name(srv_loc->data, srv_loc->len,
4003						&addr, sizeof(addr),
4004						clp->cl_net, server->port);
4005		if (!addrlen)
4006			return;
4007		xprt_args.dstaddr = (struct sockaddr *)&addr;
4008		xprt_args.addrlen = addrlen;
4009		servername = kmalloc(srv_loc->len + 1, GFP_KERNEL);
4010		if (!servername)
4011			return;
4012		memcpy(servername, srv_loc->data, srv_loc->len);
4013		servername[srv_loc->len] = '\0';
4014		xprt_args.servername = servername;
4015
4016		xprtdata.cred = nfs4_get_clid_cred(clp);
4017		rpc_clnt_add_xprt(clp->cl_rpcclient, &xprt_args,
4018				  rpc_clnt_setup_test_and_add_xprt,
4019				  &rpcdata);
4020		if (xprtdata.cred)
4021			put_cred(xprtdata.cred);
4022		kfree(servername);
4023	}
4024}
4025
4026static int _nfs4_discover_trunking(struct nfs_server *server,
4027				   struct nfs_fh *fhandle)
4028{
4029	struct nfs4_fs_locations *locations = NULL;
4030	struct page *page;
4031	const struct cred *cred;
4032	struct nfs_client *clp = server->nfs_client;
4033	const struct nfs4_state_maintenance_ops *ops =
4034		clp->cl_mvops->state_renewal_ops;
4035	int status = -ENOMEM, i;
4036
4037	cred = ops->get_state_renewal_cred(clp);
4038	if (cred == NULL) {
4039		cred = nfs4_get_clid_cred(clp);
4040		if (cred == NULL)
4041			return -ENOKEY;
4042	}
4043
4044	page = alloc_page(GFP_KERNEL);
4045	if (!page)
4046		goto out_put_cred;
4047	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4048	if (!locations)
4049		goto out_free;
4050	locations->fattr = nfs_alloc_fattr();
4051	if (!locations->fattr)
4052		goto out_free_2;
4053
4054	status = nfs4_proc_get_locations(server, fhandle, locations, page,
4055					 cred);
4056	if (status)
4057		goto out_free_3;
4058
4059	for (i = 0; i < locations->nlocations; i++)
4060		test_fs_location_for_trunking(&locations->locations[i], clp,
4061					      server);
4062out_free_3:
4063	kfree(locations->fattr);
4064out_free_2:
4065	kfree(locations);
4066out_free:
4067	__free_page(page);
4068out_put_cred:
4069	put_cred(cred);
4070	return status;
4071}
4072
4073static int nfs4_discover_trunking(struct nfs_server *server,
4074				  struct nfs_fh *fhandle)
4075{
4076	struct nfs4_exception exception = {
4077		.interruptible = true,
4078	};
4079	struct nfs_client *clp = server->nfs_client;
4080	int err = 0;
4081
4082	if (!nfs4_has_session(clp))
4083		goto out;
4084	do {
4085		err = nfs4_handle_exception(server,
4086				_nfs4_discover_trunking(server, fhandle),
4087				&exception);
4088	} while (exception.retry);
4089out:
4090	return err;
4091}
4092
4093static int _nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4094		struct nfs_fsinfo *info)
4095{
4096	u32 bitmask[3];
4097	struct nfs4_lookup_root_arg args = {
4098		.bitmask = bitmask,
4099	};
4100	struct nfs4_lookup_res res = {
4101		.server = server,
4102		.fattr = info->fattr,
4103		.fh = fhandle,
4104	};
4105	struct rpc_message msg = {
4106		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP_ROOT],
4107		.rpc_argp = &args,
4108		.rpc_resp = &res,
4109	};
4110
4111	bitmask[0] = nfs4_fattr_bitmap[0];
4112	bitmask[1] = nfs4_fattr_bitmap[1];
4113	/*
4114	 * Process the label in the upcoming getfattr
4115	 */
4116	bitmask[2] = nfs4_fattr_bitmap[2] & ~FATTR4_WORD2_SECURITY_LABEL;
4117
4118	nfs_fattr_init(info->fattr);
4119	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4120}
4121
4122static int nfs4_lookup_root(struct nfs_server *server, struct nfs_fh *fhandle,
4123		struct nfs_fsinfo *info)
4124{
4125	struct nfs4_exception exception = {
4126		.interruptible = true,
4127	};
4128	int err;
4129	do {
4130		err = _nfs4_lookup_root(server, fhandle, info);
4131		trace_nfs4_lookup_root(server, fhandle, info->fattr, err);
4132		switch (err) {
4133		case 0:
4134		case -NFS4ERR_WRONGSEC:
4135			goto out;
4136		default:
4137			err = nfs4_handle_exception(server, err, &exception);
4138		}
4139	} while (exception.retry);
4140out:
4141	return err;
4142}
4143
4144static int nfs4_lookup_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4145				struct nfs_fsinfo *info, rpc_authflavor_t flavor)
4146{
4147	struct rpc_auth_create_args auth_args = {
4148		.pseudoflavor = flavor,
4149	};
4150	struct rpc_auth *auth;
4151
4152	auth = rpcauth_create(&auth_args, server->client);
4153	if (IS_ERR(auth))
4154		return -EACCES;
4155	return nfs4_lookup_root(server, fhandle, info);
4156}
4157
4158/*
4159 * Retry pseudoroot lookup with various security flavors.  We do this when:
4160 *
4161 *   NFSv4.0: the PUTROOTFH operation returns NFS4ERR_WRONGSEC
4162 *   NFSv4.1: the server does not support the SECINFO_NO_NAME operation
4163 *
4164 * Returns zero on success, or a negative NFS4ERR value, or a
4165 * negative errno value.
4166 */
4167static int nfs4_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
4168			      struct nfs_fsinfo *info)
4169{
4170	/* Per 3530bis 15.33.5 */
4171	static const rpc_authflavor_t flav_array[] = {
4172		RPC_AUTH_GSS_KRB5P,
4173		RPC_AUTH_GSS_KRB5I,
4174		RPC_AUTH_GSS_KRB5,
4175		RPC_AUTH_UNIX,			/* courtesy */
4176		RPC_AUTH_NULL,
4177	};
4178	int status = -EPERM;
4179	size_t i;
4180
4181	if (server->auth_info.flavor_len > 0) {
4182		/* try each flavor specified by user */
4183		for (i = 0; i < server->auth_info.flavor_len; i++) {
4184			status = nfs4_lookup_root_sec(server, fhandle, info,
4185						server->auth_info.flavors[i]);
4186			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4187				continue;
4188			break;
4189		}
4190	} else {
4191		/* no flavors specified by user, try default list */
4192		for (i = 0; i < ARRAY_SIZE(flav_array); i++) {
4193			status = nfs4_lookup_root_sec(server, fhandle, info,
4194						      flav_array[i]);
4195			if (status == -NFS4ERR_WRONGSEC || status == -EACCES)
4196				continue;
4197			break;
4198		}
4199	}
4200
4201	/*
4202	 * -EACCES could mean that the user doesn't have correct permissions
4203	 * to access the mount.  It could also mean that we tried to mount
4204	 * with a gss auth flavor, but rpc.gssd isn't running.  Either way,
4205	 * existing mount programs don't handle -EACCES very well so it should
4206	 * be mapped to -EPERM instead.
4207	 */
4208	if (status == -EACCES)
4209		status = -EPERM;
4210	return status;
4211}
4212
4213/**
4214 * nfs4_proc_get_rootfh - get file handle for server's pseudoroot
4215 * @server: initialized nfs_server handle
4216 * @fhandle: we fill in the pseudo-fs root file handle
4217 * @info: we fill in an FSINFO struct
4218 * @auth_probe: probe the auth flavours
4219 *
4220 * Returns zero on success, or a negative errno.
4221 */
4222int nfs4_proc_get_rootfh(struct nfs_server *server, struct nfs_fh *fhandle,
4223			 struct nfs_fsinfo *info,
4224			 bool auth_probe)
4225{
4226	int status = 0;
4227
4228	if (!auth_probe)
4229		status = nfs4_lookup_root(server, fhandle, info);
4230
4231	if (auth_probe || status == NFS4ERR_WRONGSEC)
4232		status = server->nfs_client->cl_mvops->find_root_sec(server,
4233				fhandle, info);
4234
4235	if (status == 0)
4236		status = nfs4_server_capabilities(server, fhandle);
4237	if (status == 0)
4238		status = nfs4_do_fsinfo(server, fhandle, info);
4239
4240	return nfs4_map_errors(status);
4241}
4242
4243static int nfs4_proc_get_root(struct nfs_server *server, struct nfs_fh *mntfh,
4244			      struct nfs_fsinfo *info)
4245{
4246	int error;
4247	struct nfs_fattr *fattr = info->fattr;
4248
4249	error = nfs4_server_capabilities(server, mntfh);
4250	if (error < 0) {
4251		dprintk("nfs4_get_root: getcaps error = %d\n", -error);
4252		return error;
4253	}
4254
4255	error = nfs4_proc_getattr(server, mntfh, fattr, NULL);
4256	if (error < 0) {
4257		dprintk("nfs4_get_root: getattr error = %d\n", -error);
4258		goto out;
4259	}
4260
4261	if (fattr->valid & NFS_ATTR_FATTR_FSID &&
4262	    !nfs_fsid_equal(&server->fsid, &fattr->fsid))
4263		memcpy(&server->fsid, &fattr->fsid, sizeof(server->fsid));
4264
4265out:
4266	return error;
4267}
4268
4269/*
4270 * Get locations and (maybe) other attributes of a referral.
4271 * Note that we'll actually follow the referral later when
4272 * we detect fsid mismatch in inode revalidation
4273 */
4274static int nfs4_get_referral(struct rpc_clnt *client, struct inode *dir,
4275			     const struct qstr *name, struct nfs_fattr *fattr,
4276			     struct nfs_fh *fhandle)
4277{
4278	int status = -ENOMEM;
4279	struct page *page = NULL;
4280	struct nfs4_fs_locations *locations = NULL;
4281
4282	page = alloc_page(GFP_KERNEL);
4283	if (page == NULL)
4284		goto out;
4285	locations = kmalloc(sizeof(struct nfs4_fs_locations), GFP_KERNEL);
4286	if (locations == NULL)
4287		goto out;
4288
4289	locations->fattr = fattr;
4290
4291	status = nfs4_proc_fs_locations(client, dir, name, locations, page);
4292	if (status != 0)
4293		goto out;
4294
4295	/*
4296	 * If the fsid didn't change, this is a migration event, not a
4297	 * referral.  Cause us to drop into the exception handler, which
4298	 * will kick off migration recovery.
4299	 */
4300	if (nfs_fsid_equal(&NFS_SERVER(dir)->fsid, &fattr->fsid)) {
4301		dprintk("%s: server did not return a different fsid for"
4302			" a referral at %s\n", __func__, name->name);
4303		status = -NFS4ERR_MOVED;
4304		goto out;
4305	}
4306	/* Fixup attributes for the nfs_lookup() call to nfs_fhget() */
4307	nfs_fixup_referral_attributes(fattr);
4308	memset(fhandle, 0, sizeof(struct nfs_fh));
4309out:
4310	if (page)
4311		__free_page(page);
4312	kfree(locations);
4313	return status;
4314}
4315
4316static int _nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4317				struct nfs_fattr *fattr, struct inode *inode)
4318{
4319	__u32 bitmask[NFS4_BITMASK_SZ];
4320	struct nfs4_getattr_arg args = {
4321		.fh = fhandle,
4322		.bitmask = bitmask,
4323	};
4324	struct nfs4_getattr_res res = {
4325		.fattr = fattr,
4326		.server = server,
4327	};
4328	struct rpc_message msg = {
4329		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
4330		.rpc_argp = &args,
4331		.rpc_resp = &res,
4332	};
4333	unsigned short task_flags = 0;
4334
4335	if (nfs4_has_session(server->nfs_client))
4336		task_flags = RPC_TASK_MOVEABLE;
4337
4338	/* Is this is an attribute revalidation, subject to softreval? */
4339	if (inode && (server->flags & NFS_MOUNT_SOFTREVAL))
4340		task_flags |= RPC_TASK_TIMEOUT;
4341
4342	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, fattr->label), inode, 0);
4343	nfs_fattr_init(fattr);
4344	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4345	return nfs4_do_call_sync(server->client, server, &msg,
4346			&args.seq_args, &res.seq_res, task_flags);
4347}
4348
4349int nfs4_proc_getattr(struct nfs_server *server, struct nfs_fh *fhandle,
4350				struct nfs_fattr *fattr, struct inode *inode)
4351{
4352	struct nfs4_exception exception = {
4353		.interruptible = true,
4354	};
4355	int err;
4356	do {
4357		err = _nfs4_proc_getattr(server, fhandle, fattr, inode);
4358		trace_nfs4_getattr(server, fhandle, fattr, err);
4359		err = nfs4_handle_exception(server, err,
4360				&exception);
4361	} while (exception.retry);
4362	return err;
4363}
4364
4365/*
4366 * The file is not closed if it is opened due to the a request to change
4367 * the size of the file. The open call will not be needed once the
4368 * VFS layer lookup-intents are implemented.
4369 *
4370 * Close is called when the inode is destroyed.
4371 * If we haven't opened the file for O_WRONLY, we
4372 * need to in the size_change case to obtain a stateid.
4373 *
4374 * Got race?
4375 * Because OPEN is always done by name in nfsv4, it is
4376 * possible that we opened a different file by the same
4377 * name.  We can recognize this race condition, but we
4378 * can't do anything about it besides returning an error.
4379 *
4380 * This will be fixed with VFS changes (lookup-intent).
4381 */
4382static int
4383nfs4_proc_setattr(struct dentry *dentry, struct nfs_fattr *fattr,
4384		  struct iattr *sattr)
4385{
4386	struct inode *inode = d_inode(dentry);
4387	const struct cred *cred = NULL;
4388	struct nfs_open_context *ctx = NULL;
4389	int status;
4390
4391	if (pnfs_ld_layoutret_on_setattr(inode) &&
4392	    sattr->ia_valid & ATTR_SIZE &&
4393	    sattr->ia_size < i_size_read(inode))
4394		pnfs_commit_and_return_layout(inode);
4395
4396	nfs_fattr_init(fattr);
4397
4398	/* Deal with open(O_TRUNC) */
4399	if (sattr->ia_valid & ATTR_OPEN)
4400		sattr->ia_valid &= ~(ATTR_MTIME|ATTR_CTIME);
4401
4402	/* Optimization: if the end result is no change, don't RPC */
4403	if ((sattr->ia_valid & ~(ATTR_FILE|ATTR_OPEN)) == 0)
4404		return 0;
4405
4406	/* Search for an existing open(O_WRITE) file */
4407	if (sattr->ia_valid & ATTR_FILE) {
4408
4409		ctx = nfs_file_open_context(sattr->ia_file);
4410		if (ctx)
4411			cred = ctx->cred;
4412	}
4413
4414	/* Return any delegations if we're going to change ACLs */
4415	if ((sattr->ia_valid & (ATTR_MODE|ATTR_UID|ATTR_GID)) != 0)
4416		nfs4_inode_make_writeable(inode);
4417
4418	status = nfs4_do_setattr(inode, cred, fattr, sattr, ctx, NULL);
4419	if (status == 0) {
4420		nfs_setattr_update_inode(inode, sattr, fattr);
4421		nfs_setsecurity(inode, fattr);
4422	}
4423	return status;
4424}
4425
4426static int _nfs4_proc_lookup(struct rpc_clnt *clnt, struct inode *dir,
4427		struct dentry *dentry, struct nfs_fh *fhandle,
4428		struct nfs_fattr *fattr)
4429{
4430	struct nfs_server *server = NFS_SERVER(dir);
4431	int		       status;
4432	struct nfs4_lookup_arg args = {
4433		.bitmask = server->attr_bitmask,
4434		.dir_fh = NFS_FH(dir),
4435		.name = &dentry->d_name,
4436	};
4437	struct nfs4_lookup_res res = {
4438		.server = server,
4439		.fattr = fattr,
4440		.fh = fhandle,
4441	};
4442	struct rpc_message msg = {
4443		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUP],
4444		.rpc_argp = &args,
4445		.rpc_resp = &res,
4446	};
4447	unsigned short task_flags = 0;
4448
4449	if (nfs_server_capable(dir, NFS_CAP_MOVEABLE))
4450		task_flags = RPC_TASK_MOVEABLE;
4451
4452	/* Is this is an attribute revalidation, subject to softreval? */
4453	if (nfs_lookup_is_soft_revalidate(dentry))
4454		task_flags |= RPC_TASK_TIMEOUT;
4455
4456	args.bitmask = nfs4_bitmask(server, fattr->label);
4457
4458	nfs_fattr_init(fattr);
4459
4460	dprintk("NFS call  lookup %pd2\n", dentry);
4461	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
4462	status = nfs4_do_call_sync(clnt, server, &msg,
4463			&args.seq_args, &res.seq_res, task_flags);
4464	dprintk("NFS reply lookup: %d\n", status);
4465	return status;
4466}
4467
4468static void nfs_fixup_secinfo_attributes(struct nfs_fattr *fattr)
4469{
4470	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
4471		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_MOUNTPOINT;
4472	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
4473	fattr->nlink = 2;
4474}
4475
4476static int nfs4_proc_lookup_common(struct rpc_clnt **clnt, struct inode *dir,
4477				   struct dentry *dentry, struct nfs_fh *fhandle,
4478				   struct nfs_fattr *fattr)
4479{
4480	struct nfs4_exception exception = {
4481		.interruptible = true,
4482	};
4483	struct rpc_clnt *client = *clnt;
4484	const struct qstr *name = &dentry->d_name;
4485	int err;
4486	do {
4487		err = _nfs4_proc_lookup(client, dir, dentry, fhandle, fattr);
4488		trace_nfs4_lookup(dir, name, err);
4489		switch (err) {
4490		case -NFS4ERR_BADNAME:
4491			err = -ENOENT;
4492			goto out;
4493		case -NFS4ERR_MOVED:
4494			err = nfs4_get_referral(client, dir, name, fattr, fhandle);
4495			if (err == -NFS4ERR_MOVED)
4496				err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4497			goto out;
4498		case -NFS4ERR_WRONGSEC:
4499			err = -EPERM;
4500			if (client != *clnt)
4501				goto out;
4502			client = nfs4_negotiate_security(client, dir, name);
4503			if (IS_ERR(client))
4504				return PTR_ERR(client);
4505
4506			exception.retry = 1;
4507			break;
4508		default:
4509			err = nfs4_handle_exception(NFS_SERVER(dir), err, &exception);
4510		}
4511	} while (exception.retry);
4512
4513out:
4514	if (err == 0)
4515		*clnt = client;
4516	else if (client != *clnt)
4517		rpc_shutdown_client(client);
4518
4519	return err;
4520}
4521
4522static int nfs4_proc_lookup(struct inode *dir, struct dentry *dentry,
4523			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4524{
4525	int status;
4526	struct rpc_clnt *client = NFS_CLIENT(dir);
4527
4528	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4529	if (client != NFS_CLIENT(dir)) {
4530		rpc_shutdown_client(client);
4531		nfs_fixup_secinfo_attributes(fattr);
4532	}
4533	return status;
4534}
4535
4536struct rpc_clnt *
4537nfs4_proc_lookup_mountpoint(struct inode *dir, struct dentry *dentry,
4538			    struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4539{
4540	struct rpc_clnt *client = NFS_CLIENT(dir);
4541	int status;
4542
4543	status = nfs4_proc_lookup_common(&client, dir, dentry, fhandle, fattr);
4544	if (status < 0)
4545		return ERR_PTR(status);
4546	return (client == NFS_CLIENT(dir)) ? rpc_clone_client(client) : client;
4547}
4548
4549static int _nfs4_proc_lookupp(struct inode *inode,
4550		struct nfs_fh *fhandle, struct nfs_fattr *fattr)
4551{
4552	struct rpc_clnt *clnt = NFS_CLIENT(inode);
4553	struct nfs_server *server = NFS_SERVER(inode);
4554	int		       status;
4555	struct nfs4_lookupp_arg args = {
4556		.bitmask = server->attr_bitmask,
4557		.fh = NFS_FH(inode),
4558	};
4559	struct nfs4_lookupp_res res = {
4560		.server = server,
4561		.fattr = fattr,
4562		.fh = fhandle,
4563	};
4564	struct rpc_message msg = {
4565		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOOKUPP],
4566		.rpc_argp = &args,
4567		.rpc_resp = &res,
4568	};
4569	unsigned short task_flags = 0;
4570
4571	if (NFS_SERVER(inode)->flags & NFS_MOUNT_SOFTREVAL)
4572		task_flags |= RPC_TASK_TIMEOUT;
4573
4574	args.bitmask = nfs4_bitmask(server, fattr->label);
4575
4576	nfs_fattr_init(fattr);
4577
4578	dprintk("NFS call  lookupp ino=0x%lx\n", inode->i_ino);
4579	status = nfs4_call_sync(clnt, server, &msg, &args.seq_args,
4580				&res.seq_res, task_flags);
4581	dprintk("NFS reply lookupp: %d\n", status);
4582	return status;
4583}
4584
4585static int nfs4_proc_lookupp(struct inode *inode, struct nfs_fh *fhandle,
4586			     struct nfs_fattr *fattr)
4587{
4588	struct nfs4_exception exception = {
4589		.interruptible = true,
4590	};
4591	int err;
4592	do {
4593		err = _nfs4_proc_lookupp(inode, fhandle, fattr);
4594		trace_nfs4_lookupp(inode, err);
4595		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4596				&exception);
4597	} while (exception.retry);
4598	return err;
4599}
4600
4601static int _nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4602			     const struct cred *cred)
4603{
4604	struct nfs_server *server = NFS_SERVER(inode);
4605	struct nfs4_accessargs args = {
4606		.fh = NFS_FH(inode),
4607		.access = entry->mask,
4608	};
4609	struct nfs4_accessres res = {
4610		.server = server,
4611	};
4612	struct rpc_message msg = {
4613		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ACCESS],
4614		.rpc_argp = &args,
4615		.rpc_resp = &res,
4616		.rpc_cred = cred,
4617	};
4618	int status = 0;
4619
4620	if (!nfs4_have_delegation(inode, FMODE_READ)) {
4621		res.fattr = nfs_alloc_fattr();
4622		if (res.fattr == NULL)
4623			return -ENOMEM;
4624		args.bitmask = server->cache_consistency_bitmask;
4625	}
4626	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
4627	if (!status) {
4628		nfs_access_set_mask(entry, res.access);
4629		if (res.fattr)
4630			nfs_refresh_inode(inode, res.fattr);
4631	}
4632	nfs_free_fattr(res.fattr);
4633	return status;
4634}
4635
4636static int nfs4_proc_access(struct inode *inode, struct nfs_access_entry *entry,
4637			    const struct cred *cred)
4638{
4639	struct nfs4_exception exception = {
4640		.interruptible = true,
4641	};
4642	int err;
4643	do {
4644		err = _nfs4_proc_access(inode, entry, cred);
4645		trace_nfs4_access(inode, err);
4646		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4647				&exception);
4648	} while (exception.retry);
4649	return err;
4650}
4651
4652/*
4653 * TODO: For the time being, we don't try to get any attributes
4654 * along with any of the zero-copy operations READ, READDIR,
4655 * READLINK, WRITE.
4656 *
4657 * In the case of the first three, we want to put the GETATTR
4658 * after the read-type operation -- this is because it is hard
4659 * to predict the length of a GETATTR response in v4, and thus
4660 * align the READ data correctly.  This means that the GETATTR
4661 * may end up partially falling into the page cache, and we should
4662 * shift it into the 'tail' of the xdr_buf before processing.
4663 * To do this efficiently, we need to know the total length
4664 * of data received, which doesn't seem to be available outside
4665 * of the RPC layer.
4666 *
4667 * In the case of WRITE, we also want to put the GETATTR after
4668 * the operation -- in this case because we want to make sure
4669 * we get the post-operation mtime and size.
4670 *
4671 * Both of these changes to the XDR layer would in fact be quite
4672 * minor, but I decided to leave them for a subsequent patch.
4673 */
4674static int _nfs4_proc_readlink(struct inode *inode, struct page *page,
4675		unsigned int pgbase, unsigned int pglen)
4676{
4677	struct nfs4_readlink args = {
4678		.fh       = NFS_FH(inode),
4679		.pgbase	  = pgbase,
4680		.pglen    = pglen,
4681		.pages    = &page,
4682	};
4683	struct nfs4_readlink_res res;
4684	struct rpc_message msg = {
4685		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READLINK],
4686		.rpc_argp = &args,
4687		.rpc_resp = &res,
4688	};
4689
4690	return nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode), &msg, &args.seq_args, &res.seq_res, 0);
4691}
4692
4693static int nfs4_proc_readlink(struct inode *inode, struct page *page,
4694		unsigned int pgbase, unsigned int pglen)
4695{
4696	struct nfs4_exception exception = {
4697		.interruptible = true,
4698	};
4699	int err;
4700	do {
4701		err = _nfs4_proc_readlink(inode, page, pgbase, pglen);
4702		trace_nfs4_readlink(inode, err);
4703		err = nfs4_handle_exception(NFS_SERVER(inode), err,
4704				&exception);
4705	} while (exception.retry);
4706	return err;
4707}
4708
4709/*
4710 * This is just for mknod.  open(O_CREAT) will always do ->open_context().
4711 */
4712static int
4713nfs4_proc_create(struct inode *dir, struct dentry *dentry, struct iattr *sattr,
4714		 int flags)
4715{
4716	struct nfs_server *server = NFS_SERVER(dir);
4717	struct nfs4_label l, *ilabel;
4718	struct nfs_open_context *ctx;
4719	struct nfs4_state *state;
4720	int status = 0;
4721
4722	ctx = alloc_nfs_open_context(dentry, FMODE_READ, NULL);
4723	if (IS_ERR(ctx))
4724		return PTR_ERR(ctx);
4725
4726	ilabel = nfs4_label_init_security(dir, dentry, sattr, &l);
4727
4728	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
4729		sattr->ia_mode &= ~current_umask();
4730	state = nfs4_do_open(dir, ctx, flags, sattr, ilabel, NULL);
4731	if (IS_ERR(state)) {
4732		status = PTR_ERR(state);
4733		goto out;
4734	}
4735out:
4736	nfs4_label_release_security(ilabel);
4737	put_nfs_open_context(ctx);
4738	return status;
4739}
4740
4741static int
4742_nfs4_proc_remove(struct inode *dir, const struct qstr *name, u32 ftype)
4743{
4744	struct nfs_server *server = NFS_SERVER(dir);
4745	struct nfs_removeargs args = {
4746		.fh = NFS_FH(dir),
4747		.name = *name,
4748	};
4749	struct nfs_removeres res = {
4750		.server = server,
4751	};
4752	struct rpc_message msg = {
4753		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE],
4754		.rpc_argp = &args,
4755		.rpc_resp = &res,
4756	};
4757	unsigned long timestamp = jiffies;
4758	int status;
4759
4760	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 1);
4761	if (status == 0) {
4762		spin_lock(&dir->i_lock);
4763		/* Removing a directory decrements nlink in the parent */
4764		if (ftype == NF4DIR && dir->i_nlink > 2)
4765			nfs4_dec_nlink_locked(dir);
4766		nfs4_update_changeattr_locked(dir, &res.cinfo, timestamp,
4767					      NFS_INO_INVALID_DATA);
4768		spin_unlock(&dir->i_lock);
4769	}
4770	return status;
4771}
4772
4773static int nfs4_proc_remove(struct inode *dir, struct dentry *dentry)
4774{
4775	struct nfs4_exception exception = {
4776		.interruptible = true,
4777	};
4778	struct inode *inode = d_inode(dentry);
4779	int err;
4780
4781	if (inode) {
4782		if (inode->i_nlink == 1)
4783			nfs4_inode_return_delegation(inode);
4784		else
4785			nfs4_inode_make_writeable(inode);
4786	}
4787	do {
4788		err = _nfs4_proc_remove(dir, &dentry->d_name, NF4REG);
4789		trace_nfs4_remove(dir, &dentry->d_name, err);
4790		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4791				&exception);
4792	} while (exception.retry);
4793	return err;
4794}
4795
4796static int nfs4_proc_rmdir(struct inode *dir, const struct qstr *name)
4797{
4798	struct nfs4_exception exception = {
4799		.interruptible = true,
4800	};
4801	int err;
4802
4803	do {
4804		err = _nfs4_proc_remove(dir, name, NF4DIR);
4805		trace_nfs4_remove(dir, name, err);
4806		err = nfs4_handle_exception(NFS_SERVER(dir), err,
4807				&exception);
4808	} while (exception.retry);
4809	return err;
4810}
4811
4812static void nfs4_proc_unlink_setup(struct rpc_message *msg,
4813		struct dentry *dentry,
4814		struct inode *inode)
4815{
4816	struct nfs_removeargs *args = msg->rpc_argp;
4817	struct nfs_removeres *res = msg->rpc_resp;
4818
4819	res->server = NFS_SB(dentry->d_sb);
4820	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVE];
4821	nfs4_init_sequence(&args->seq_args, &res->seq_res, 1, 0);
4822
4823	nfs_fattr_init(res->dir_attr);
4824
4825	if (inode) {
4826		nfs4_inode_return_delegation(inode);
4827		nfs_d_prune_case_insensitive_aliases(inode);
4828	}
4829}
4830
4831static void nfs4_proc_unlink_rpc_prepare(struct rpc_task *task, struct nfs_unlinkdata *data)
4832{
4833	nfs4_setup_sequence(NFS_SB(data->dentry->d_sb)->nfs_client,
4834			&data->args.seq_args,
4835			&data->res.seq_res,
4836			task);
4837}
4838
4839static int nfs4_proc_unlink_done(struct rpc_task *task, struct inode *dir)
4840{
4841	struct nfs_unlinkdata *data = task->tk_calldata;
4842	struct nfs_removeres *res = &data->res;
4843
4844	if (!nfs4_sequence_done(task, &res->seq_res))
4845		return 0;
4846	if (nfs4_async_handle_error(task, res->server, NULL,
4847				    &data->timeout) == -EAGAIN)
4848		return 0;
4849	if (task->tk_status == 0)
4850		nfs4_update_changeattr(dir, &res->cinfo,
4851				res->dir_attr->time_start,
4852				NFS_INO_INVALID_DATA);
4853	return 1;
4854}
4855
4856static void nfs4_proc_rename_setup(struct rpc_message *msg,
4857		struct dentry *old_dentry,
4858		struct dentry *new_dentry)
4859{
4860	struct nfs_renameargs *arg = msg->rpc_argp;
4861	struct nfs_renameres *res = msg->rpc_resp;
4862	struct inode *old_inode = d_inode(old_dentry);
4863	struct inode *new_inode = d_inode(new_dentry);
4864
4865	if (old_inode)
4866		nfs4_inode_make_writeable(old_inode);
4867	if (new_inode)
4868		nfs4_inode_return_delegation(new_inode);
4869	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RENAME];
4870	res->server = NFS_SB(old_dentry->d_sb);
4871	nfs4_init_sequence(&arg->seq_args, &res->seq_res, 1, 0);
4872}
4873
4874static void nfs4_proc_rename_rpc_prepare(struct rpc_task *task, struct nfs_renamedata *data)
4875{
4876	nfs4_setup_sequence(NFS_SERVER(data->old_dir)->nfs_client,
4877			&data->args.seq_args,
4878			&data->res.seq_res,
4879			task);
4880}
4881
4882static int nfs4_proc_rename_done(struct rpc_task *task, struct inode *old_dir,
4883				 struct inode *new_dir)
4884{
4885	struct nfs_renamedata *data = task->tk_calldata;
4886	struct nfs_renameres *res = &data->res;
4887
4888	if (!nfs4_sequence_done(task, &res->seq_res))
4889		return 0;
4890	if (nfs4_async_handle_error(task, res->server, NULL, &data->timeout) == -EAGAIN)
4891		return 0;
4892
4893	if (task->tk_status == 0) {
4894		nfs_d_prune_case_insensitive_aliases(d_inode(data->old_dentry));
4895		if (new_dir != old_dir) {
4896			/* Note: If we moved a directory, nlink will change */
4897			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4898					res->old_fattr->time_start,
4899					NFS_INO_INVALID_NLINK |
4900					    NFS_INO_INVALID_DATA);
4901			nfs4_update_changeattr(new_dir, &res->new_cinfo,
4902					res->new_fattr->time_start,
4903					NFS_INO_INVALID_NLINK |
4904					    NFS_INO_INVALID_DATA);
4905		} else
4906			nfs4_update_changeattr(old_dir, &res->old_cinfo,
4907					res->old_fattr->time_start,
4908					NFS_INO_INVALID_DATA);
4909	}
4910	return 1;
4911}
4912
4913static int _nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4914{
4915	struct nfs_server *server = NFS_SERVER(inode);
4916	__u32 bitmask[NFS4_BITMASK_SZ];
4917	struct nfs4_link_arg arg = {
4918		.fh     = NFS_FH(inode),
4919		.dir_fh = NFS_FH(dir),
4920		.name   = name,
4921		.bitmask = bitmask,
4922	};
4923	struct nfs4_link_res res = {
4924		.server = server,
4925	};
4926	struct rpc_message msg = {
4927		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LINK],
4928		.rpc_argp = &arg,
4929		.rpc_resp = &res,
4930	};
4931	int status = -ENOMEM;
4932
4933	res.fattr = nfs_alloc_fattr_with_label(server);
4934	if (res.fattr == NULL)
4935		goto out;
4936
4937	nfs4_inode_make_writeable(inode);
4938	nfs4_bitmap_copy_adjust(bitmask, nfs4_bitmask(server, res.fattr->label), inode,
4939				NFS_INO_INVALID_CHANGE);
4940	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
4941	if (!status) {
4942		nfs4_update_changeattr(dir, &res.cinfo, res.fattr->time_start,
4943				       NFS_INO_INVALID_DATA);
4944		nfs4_inc_nlink(inode);
4945		status = nfs_post_op_update_inode(inode, res.fattr);
4946		if (!status)
4947			nfs_setsecurity(inode, res.fattr);
4948	}
4949
4950out:
4951	nfs_free_fattr(res.fattr);
4952	return status;
4953}
4954
4955static int nfs4_proc_link(struct inode *inode, struct inode *dir, const struct qstr *name)
4956{
4957	struct nfs4_exception exception = {
4958		.interruptible = true,
4959	};
4960	int err;
4961	do {
4962		err = nfs4_handle_exception(NFS_SERVER(inode),
4963				_nfs4_proc_link(inode, dir, name),
4964				&exception);
4965	} while (exception.retry);
4966	return err;
4967}
4968
4969struct nfs4_createdata {
4970	struct rpc_message msg;
4971	struct nfs4_create_arg arg;
4972	struct nfs4_create_res res;
4973	struct nfs_fh fh;
4974	struct nfs_fattr fattr;
4975};
4976
4977static struct nfs4_createdata *nfs4_alloc_createdata(struct inode *dir,
4978		const struct qstr *name, struct iattr *sattr, u32 ftype)
4979{
4980	struct nfs4_createdata *data;
4981
4982	data = kzalloc(sizeof(*data), GFP_KERNEL);
4983	if (data != NULL) {
4984		struct nfs_server *server = NFS_SERVER(dir);
4985
4986		data->fattr.label = nfs4_label_alloc(server, GFP_KERNEL);
4987		if (IS_ERR(data->fattr.label))
4988			goto out_free;
4989
4990		data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE];
4991		data->msg.rpc_argp = &data->arg;
4992		data->msg.rpc_resp = &data->res;
4993		data->arg.dir_fh = NFS_FH(dir);
4994		data->arg.server = server;
4995		data->arg.name = name;
4996		data->arg.attrs = sattr;
4997		data->arg.ftype = ftype;
4998		data->arg.bitmask = nfs4_bitmask(server, data->fattr.label);
4999		data->arg.umask = current_umask();
5000		data->res.server = server;
5001		data->res.fh = &data->fh;
5002		data->res.fattr = &data->fattr;
5003		nfs_fattr_init(data->res.fattr);
5004	}
5005	return data;
5006out_free:
5007	kfree(data);
5008	return NULL;
5009}
5010
5011static int nfs4_do_create(struct inode *dir, struct dentry *dentry, struct nfs4_createdata *data)
5012{
5013	int status = nfs4_call_sync(NFS_SERVER(dir)->client, NFS_SERVER(dir), &data->msg,
5014				    &data->arg.seq_args, &data->res.seq_res, 1);
5015	if (status == 0) {
5016		spin_lock(&dir->i_lock);
5017		/* Creating a directory bumps nlink in the parent */
5018		if (data->arg.ftype == NF4DIR)
5019			nfs4_inc_nlink_locked(dir);
5020		nfs4_update_changeattr_locked(dir, &data->res.dir_cinfo,
5021					      data->res.fattr->time_start,
5022					      NFS_INO_INVALID_DATA);
5023		spin_unlock(&dir->i_lock);
5024		status = nfs_instantiate(dentry, data->res.fh, data->res.fattr);
5025	}
5026	return status;
5027}
5028
5029static void nfs4_free_createdata(struct nfs4_createdata *data)
5030{
5031	nfs4_label_free(data->fattr.label);
5032	kfree(data);
5033}
5034
5035static int _nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5036		struct folio *folio, unsigned int len, struct iattr *sattr,
5037		struct nfs4_label *label)
5038{
5039	struct page *page = &folio->page;
5040	struct nfs4_createdata *data;
5041	int status = -ENAMETOOLONG;
5042
5043	if (len > NFS4_MAXPATHLEN)
5044		goto out;
5045
5046	status = -ENOMEM;
5047	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4LNK);
5048	if (data == NULL)
5049		goto out;
5050
5051	data->msg.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SYMLINK];
5052	data->arg.u.symlink.pages = &page;
5053	data->arg.u.symlink.len = len;
5054	data->arg.label = label;
5055
5056	status = nfs4_do_create(dir, dentry, data);
5057
5058	nfs4_free_createdata(data);
5059out:
5060	return status;
5061}
5062
5063static int nfs4_proc_symlink(struct inode *dir, struct dentry *dentry,
5064		struct folio *folio, unsigned int len, struct iattr *sattr)
5065{
5066	struct nfs4_exception exception = {
5067		.interruptible = true,
5068	};
5069	struct nfs4_label l, *label;
5070	int err;
5071
5072	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5073
5074	do {
5075		err = _nfs4_proc_symlink(dir, dentry, folio, len, sattr, label);
5076		trace_nfs4_symlink(dir, &dentry->d_name, err);
5077		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5078				&exception);
5079	} while (exception.retry);
5080
5081	nfs4_label_release_security(label);
5082	return err;
5083}
5084
5085static int _nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5086		struct iattr *sattr, struct nfs4_label *label)
5087{
5088	struct nfs4_createdata *data;
5089	int status = -ENOMEM;
5090
5091	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4DIR);
5092	if (data == NULL)
5093		goto out;
5094
5095	data->arg.label = label;
5096	status = nfs4_do_create(dir, dentry, data);
5097
5098	nfs4_free_createdata(data);
5099out:
5100	return status;
5101}
5102
5103static int nfs4_proc_mkdir(struct inode *dir, struct dentry *dentry,
5104		struct iattr *sattr)
5105{
5106	struct nfs_server *server = NFS_SERVER(dir);
5107	struct nfs4_exception exception = {
5108		.interruptible = true,
5109	};
5110	struct nfs4_label l, *label;
5111	int err;
5112
5113	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5114
5115	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5116		sattr->ia_mode &= ~current_umask();
5117	do {
5118		err = _nfs4_proc_mkdir(dir, dentry, sattr, label);
5119		trace_nfs4_mkdir(dir, &dentry->d_name, err);
5120		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5121				&exception);
5122	} while (exception.retry);
5123	nfs4_label_release_security(label);
5124
5125	return err;
5126}
5127
5128static int _nfs4_proc_readdir(struct nfs_readdir_arg *nr_arg,
5129			      struct nfs_readdir_res *nr_res)
5130{
5131	struct inode		*dir = d_inode(nr_arg->dentry);
5132	struct nfs_server	*server = NFS_SERVER(dir);
5133	struct nfs4_readdir_arg args = {
5134		.fh = NFS_FH(dir),
5135		.pages = nr_arg->pages,
5136		.pgbase = 0,
5137		.count = nr_arg->page_len,
5138		.plus = nr_arg->plus,
5139	};
5140	struct nfs4_readdir_res res;
5141	struct rpc_message msg = {
5142		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READDIR],
5143		.rpc_argp = &args,
5144		.rpc_resp = &res,
5145		.rpc_cred = nr_arg->cred,
5146	};
5147	int			status;
5148
5149	dprintk("%s: dentry = %pd2, cookie = %llu\n", __func__,
5150		nr_arg->dentry, (unsigned long long)nr_arg->cookie);
5151	if (!(server->caps & NFS_CAP_SECURITY_LABEL))
5152		args.bitmask = server->attr_bitmask_nl;
5153	else
5154		args.bitmask = server->attr_bitmask;
5155
5156	nfs4_setup_readdir(nr_arg->cookie, nr_arg->verf, nr_arg->dentry, &args);
5157	res.pgbase = args.pgbase;
5158	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
5159			&res.seq_res, 0);
5160	if (status >= 0) {
5161		memcpy(nr_res->verf, res.verifier.data, NFS4_VERIFIER_SIZE);
5162		status += args.pgbase;
5163	}
5164
5165	nfs_invalidate_atime(dir);
5166
5167	dprintk("%s: returns %d\n", __func__, status);
5168	return status;
5169}
5170
5171static int nfs4_proc_readdir(struct nfs_readdir_arg *arg,
5172			     struct nfs_readdir_res *res)
5173{
5174	struct nfs4_exception exception = {
5175		.interruptible = true,
5176	};
5177	int err;
5178	do {
5179		err = _nfs4_proc_readdir(arg, res);
5180		trace_nfs4_readdir(d_inode(arg->dentry), err);
5181		err = nfs4_handle_exception(NFS_SERVER(d_inode(arg->dentry)),
5182					    err, &exception);
5183	} while (exception.retry);
5184	return err;
5185}
5186
5187static int _nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5188		struct iattr *sattr, struct nfs4_label *label, dev_t rdev)
5189{
5190	struct nfs4_createdata *data;
5191	int mode = sattr->ia_mode;
5192	int status = -ENOMEM;
5193
5194	data = nfs4_alloc_createdata(dir, &dentry->d_name, sattr, NF4SOCK);
5195	if (data == NULL)
5196		goto out;
5197
5198	if (S_ISFIFO(mode))
5199		data->arg.ftype = NF4FIFO;
5200	else if (S_ISBLK(mode)) {
5201		data->arg.ftype = NF4BLK;
5202		data->arg.u.device.specdata1 = MAJOR(rdev);
5203		data->arg.u.device.specdata2 = MINOR(rdev);
5204	}
5205	else if (S_ISCHR(mode)) {
5206		data->arg.ftype = NF4CHR;
5207		data->arg.u.device.specdata1 = MAJOR(rdev);
5208		data->arg.u.device.specdata2 = MINOR(rdev);
5209	} else if (!S_ISSOCK(mode)) {
5210		status = -EINVAL;
5211		goto out_free;
5212	}
5213
5214	data->arg.label = label;
5215	status = nfs4_do_create(dir, dentry, data);
5216out_free:
5217	nfs4_free_createdata(data);
5218out:
5219	return status;
5220}
5221
5222static int nfs4_proc_mknod(struct inode *dir, struct dentry *dentry,
5223		struct iattr *sattr, dev_t rdev)
5224{
5225	struct nfs_server *server = NFS_SERVER(dir);
5226	struct nfs4_exception exception = {
5227		.interruptible = true,
5228	};
5229	struct nfs4_label l, *label;
5230	int err;
5231
5232	label = nfs4_label_init_security(dir, dentry, sattr, &l);
5233
5234	if (!(server->attr_bitmask[2] & FATTR4_WORD2_MODE_UMASK))
5235		sattr->ia_mode &= ~current_umask();
5236	do {
5237		err = _nfs4_proc_mknod(dir, dentry, sattr, label, rdev);
5238		trace_nfs4_mknod(dir, &dentry->d_name, err);
5239		err = nfs4_handle_exception(NFS_SERVER(dir), err,
5240				&exception);
5241	} while (exception.retry);
5242
5243	nfs4_label_release_security(label);
5244
5245	return err;
5246}
5247
5248static int _nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle,
5249		 struct nfs_fsstat *fsstat)
5250{
5251	struct nfs4_statfs_arg args = {
5252		.fh = fhandle,
5253		.bitmask = server->attr_bitmask,
5254	};
5255	struct nfs4_statfs_res res = {
5256		.fsstat = fsstat,
5257	};
5258	struct rpc_message msg = {
5259		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_STATFS],
5260		.rpc_argp = &args,
5261		.rpc_resp = &res,
5262	};
5263
5264	nfs_fattr_init(fsstat->fattr);
5265	return  nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5266}
5267
5268static int nfs4_proc_statfs(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsstat *fsstat)
5269{
5270	struct nfs4_exception exception = {
5271		.interruptible = true,
5272	};
5273	int err;
5274	do {
5275		err = nfs4_handle_exception(server,
5276				_nfs4_proc_statfs(server, fhandle, fsstat),
5277				&exception);
5278	} while (exception.retry);
5279	return err;
5280}
5281
5282static int _nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle,
5283		struct nfs_fsinfo *fsinfo)
5284{
5285	struct nfs4_fsinfo_arg args = {
5286		.fh = fhandle,
5287		.bitmask = server->attr_bitmask,
5288	};
5289	struct nfs4_fsinfo_res res = {
5290		.fsinfo = fsinfo,
5291	};
5292	struct rpc_message msg = {
5293		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FSINFO],
5294		.rpc_argp = &args,
5295		.rpc_resp = &res,
5296	};
5297
5298	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5299}
5300
5301static int nfs4_do_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5302{
5303	struct nfs4_exception exception = {
5304		.interruptible = true,
5305	};
5306	int err;
5307
5308	do {
5309		err = _nfs4_do_fsinfo(server, fhandle, fsinfo);
5310		trace_nfs4_fsinfo(server, fhandle, fsinfo->fattr, err);
5311		if (err == 0) {
5312			nfs4_set_lease_period(server->nfs_client, fsinfo->lease_time * HZ);
5313			break;
5314		}
5315		err = nfs4_handle_exception(server, err, &exception);
5316	} while (exception.retry);
5317	return err;
5318}
5319
5320static int nfs4_proc_fsinfo(struct nfs_server *server, struct nfs_fh *fhandle, struct nfs_fsinfo *fsinfo)
5321{
5322	int error;
5323
5324	nfs_fattr_init(fsinfo->fattr);
5325	error = nfs4_do_fsinfo(server, fhandle, fsinfo);
5326	if (error == 0) {
5327		/* block layout checks this! */
5328		server->pnfs_blksize = fsinfo->blksize;
5329		set_pnfs_layoutdriver(server, fhandle, fsinfo);
5330	}
5331
5332	return error;
5333}
5334
5335static int _nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5336		struct nfs_pathconf *pathconf)
5337{
5338	struct nfs4_pathconf_arg args = {
5339		.fh = fhandle,
5340		.bitmask = server->attr_bitmask,
5341	};
5342	struct nfs4_pathconf_res res = {
5343		.pathconf = pathconf,
5344	};
5345	struct rpc_message msg = {
5346		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_PATHCONF],
5347		.rpc_argp = &args,
5348		.rpc_resp = &res,
5349	};
5350
5351	/* None of the pathconf attributes are mandatory to implement */
5352	if ((args.bitmask[0] & nfs4_pathconf_bitmap[0]) == 0) {
5353		memset(pathconf, 0, sizeof(*pathconf));
5354		return 0;
5355	}
5356
5357	nfs_fattr_init(pathconf->fattr);
5358	return nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
5359}
5360
5361static int nfs4_proc_pathconf(struct nfs_server *server, struct nfs_fh *fhandle,
5362		struct nfs_pathconf *pathconf)
5363{
5364	struct nfs4_exception exception = {
5365		.interruptible = true,
5366	};
5367	int err;
5368
5369	do {
5370		err = nfs4_handle_exception(server,
5371				_nfs4_proc_pathconf(server, fhandle, pathconf),
5372				&exception);
5373	} while (exception.retry);
5374	return err;
5375}
5376
5377int nfs4_set_rw_stateid(nfs4_stateid *stateid,
5378		const struct nfs_open_context *ctx,
5379		const struct nfs_lock_context *l_ctx,
5380		fmode_t fmode)
5381{
5382	return nfs4_select_rw_stateid(ctx->state, fmode, l_ctx, stateid, NULL);
5383}
5384EXPORT_SYMBOL_GPL(nfs4_set_rw_stateid);
5385
5386static bool nfs4_stateid_is_current(nfs4_stateid *stateid,
5387		const struct nfs_open_context *ctx,
5388		const struct nfs_lock_context *l_ctx,
5389		fmode_t fmode)
5390{
5391	nfs4_stateid _current_stateid;
5392
5393	/* If the current stateid represents a lost lock, then exit */
5394	if (nfs4_set_rw_stateid(&_current_stateid, ctx, l_ctx, fmode) == -EIO)
5395		return true;
5396	return nfs4_stateid_match(stateid, &_current_stateid);
5397}
5398
5399static bool nfs4_error_stateid_expired(int err)
5400{
5401	switch (err) {
5402	case -NFS4ERR_DELEG_REVOKED:
5403	case -NFS4ERR_ADMIN_REVOKED:
5404	case -NFS4ERR_BAD_STATEID:
5405	case -NFS4ERR_STALE_STATEID:
5406	case -NFS4ERR_OLD_STATEID:
5407	case -NFS4ERR_OPENMODE:
5408	case -NFS4ERR_EXPIRED:
5409		return true;
5410	}
5411	return false;
5412}
5413
5414static int nfs4_read_done_cb(struct rpc_task *task, struct nfs_pgio_header *hdr)
5415{
5416	struct nfs_server *server = NFS_SERVER(hdr->inode);
5417
5418	trace_nfs4_read(hdr, task->tk_status);
5419	if (task->tk_status < 0) {
5420		struct nfs4_exception exception = {
5421			.inode = hdr->inode,
5422			.state = hdr->args.context->state,
5423			.stateid = &hdr->args.stateid,
5424		};
5425		task->tk_status = nfs4_async_handle_exception(task,
5426				server, task->tk_status, &exception);
5427		if (exception.retry) {
5428			rpc_restart_call_prepare(task);
5429			return -EAGAIN;
5430		}
5431	}
5432
5433	if (task->tk_status > 0)
5434		renew_lease(server, hdr->timestamp);
5435	return 0;
5436}
5437
5438static bool nfs4_read_stateid_changed(struct rpc_task *task,
5439		struct nfs_pgio_args *args)
5440{
5441
5442	if (!nfs4_error_stateid_expired(task->tk_status) ||
5443		nfs4_stateid_is_current(&args->stateid,
5444				args->context,
5445				args->lock_context,
5446				FMODE_READ))
5447		return false;
5448	rpc_restart_call_prepare(task);
5449	return true;
5450}
5451
5452static bool nfs4_read_plus_not_supported(struct rpc_task *task,
5453					 struct nfs_pgio_header *hdr)
5454{
5455	struct nfs_server *server = NFS_SERVER(hdr->inode);
5456	struct rpc_message *msg = &task->tk_msg;
5457
5458	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS] &&
5459	    server->caps & NFS_CAP_READ_PLUS && task->tk_status == -ENOTSUPP) {
5460		server->caps &= ~NFS_CAP_READ_PLUS;
5461		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5462		rpc_restart_call_prepare(task);
5463		return true;
5464	}
5465	return false;
5466}
5467
5468static int nfs4_read_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5469{
5470	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5471		return -EAGAIN;
5472	if (nfs4_read_stateid_changed(task, &hdr->args))
5473		return -EAGAIN;
5474	if (nfs4_read_plus_not_supported(task, hdr))
5475		return -EAGAIN;
5476	if (task->tk_status > 0)
5477		nfs_invalidate_atime(hdr->inode);
5478	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5479				    nfs4_read_done_cb(task, hdr);
5480}
5481
5482#if defined CONFIG_NFS_V4_2 && defined CONFIG_NFS_V4_2_READ_PLUS
5483static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5484				    struct rpc_message *msg)
5485{
5486	/* Note: We don't use READ_PLUS with pNFS yet */
5487	if (nfs_server_capable(hdr->inode, NFS_CAP_READ_PLUS) && !hdr->ds_clp) {
5488		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ_PLUS];
5489		return nfs_read_alloc_scratch(hdr, READ_PLUS_SCRATCH_SIZE);
5490	}
5491	return false;
5492}
5493#else
5494static bool nfs42_read_plus_support(struct nfs_pgio_header *hdr,
5495				    struct rpc_message *msg)
5496{
5497	return false;
5498}
5499#endif /* CONFIG_NFS_V4_2 */
5500
5501static void nfs4_proc_read_setup(struct nfs_pgio_header *hdr,
5502				 struct rpc_message *msg)
5503{
5504	hdr->timestamp   = jiffies;
5505	if (!hdr->pgio_done_cb)
5506		hdr->pgio_done_cb = nfs4_read_done_cb;
5507	if (!nfs42_read_plus_support(hdr, msg))
5508		msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_READ];
5509	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5510}
5511
5512static int nfs4_proc_pgio_rpc_prepare(struct rpc_task *task,
5513				      struct nfs_pgio_header *hdr)
5514{
5515	if (nfs4_setup_sequence(NFS_SERVER(hdr->inode)->nfs_client,
5516			&hdr->args.seq_args,
5517			&hdr->res.seq_res,
5518			task))
5519		return 0;
5520	if (nfs4_set_rw_stateid(&hdr->args.stateid, hdr->args.context,
5521				hdr->args.lock_context,
5522				hdr->rw_mode) == -EIO)
5523		return -EIO;
5524	if (unlikely(test_bit(NFS_CONTEXT_BAD, &hdr->args.context->flags)))
5525		return -EIO;
5526	return 0;
5527}
5528
5529static int nfs4_write_done_cb(struct rpc_task *task,
5530			      struct nfs_pgio_header *hdr)
5531{
5532	struct inode *inode = hdr->inode;
5533
5534	trace_nfs4_write(hdr, task->tk_status);
5535	if (task->tk_status < 0) {
5536		struct nfs4_exception exception = {
5537			.inode = hdr->inode,
5538			.state = hdr->args.context->state,
5539			.stateid = &hdr->args.stateid,
5540		};
5541		task->tk_status = nfs4_async_handle_exception(task,
5542				NFS_SERVER(inode), task->tk_status,
5543				&exception);
5544		if (exception.retry) {
5545			rpc_restart_call_prepare(task);
5546			return -EAGAIN;
5547		}
5548	}
5549	if (task->tk_status >= 0) {
5550		renew_lease(NFS_SERVER(inode), hdr->timestamp);
5551		nfs_writeback_update_inode(hdr);
5552	}
5553	return 0;
5554}
5555
5556static bool nfs4_write_stateid_changed(struct rpc_task *task,
5557		struct nfs_pgio_args *args)
5558{
5559
5560	if (!nfs4_error_stateid_expired(task->tk_status) ||
5561		nfs4_stateid_is_current(&args->stateid,
5562				args->context,
5563				args->lock_context,
5564				FMODE_WRITE))
5565		return false;
5566	rpc_restart_call_prepare(task);
5567	return true;
5568}
5569
5570static int nfs4_write_done(struct rpc_task *task, struct nfs_pgio_header *hdr)
5571{
5572	if (!nfs4_sequence_done(task, &hdr->res.seq_res))
5573		return -EAGAIN;
5574	if (nfs4_write_stateid_changed(task, &hdr->args))
5575		return -EAGAIN;
5576	return hdr->pgio_done_cb ? hdr->pgio_done_cb(task, hdr) :
5577		nfs4_write_done_cb(task, hdr);
5578}
5579
5580static
5581bool nfs4_write_need_cache_consistency_data(struct nfs_pgio_header *hdr)
5582{
5583	/* Don't request attributes for pNFS or O_DIRECT writes */
5584	if (hdr->ds_clp != NULL || hdr->dreq != NULL)
5585		return false;
5586	/* Otherwise, request attributes if and only if we don't hold
5587	 * a delegation
5588	 */
5589	return nfs4_have_delegation(hdr->inode, FMODE_READ) == 0;
5590}
5591
5592void nfs4_bitmask_set(__u32 bitmask[], const __u32 src[],
5593		      struct inode *inode, unsigned long cache_validity)
5594{
5595	struct nfs_server *server = NFS_SERVER(inode);
5596	unsigned int i;
5597
5598	memcpy(bitmask, src, sizeof(*bitmask) * NFS4_BITMASK_SZ);
5599	cache_validity |= READ_ONCE(NFS_I(inode)->cache_validity);
5600
5601	if (cache_validity & NFS_INO_INVALID_CHANGE)
5602		bitmask[0] |= FATTR4_WORD0_CHANGE;
5603	if (cache_validity & NFS_INO_INVALID_ATIME)
5604		bitmask[1] |= FATTR4_WORD1_TIME_ACCESS;
5605	if (cache_validity & NFS_INO_INVALID_MODE)
5606		bitmask[1] |= FATTR4_WORD1_MODE;
5607	if (cache_validity & NFS_INO_INVALID_OTHER)
5608		bitmask[1] |= FATTR4_WORD1_OWNER | FATTR4_WORD1_OWNER_GROUP;
5609	if (cache_validity & NFS_INO_INVALID_NLINK)
5610		bitmask[1] |= FATTR4_WORD1_NUMLINKS;
5611	if (cache_validity & NFS_INO_INVALID_CTIME)
5612		bitmask[1] |= FATTR4_WORD1_TIME_METADATA;
5613	if (cache_validity & NFS_INO_INVALID_MTIME)
5614		bitmask[1] |= FATTR4_WORD1_TIME_MODIFY;
5615	if (cache_validity & NFS_INO_INVALID_BLOCKS)
5616		bitmask[1] |= FATTR4_WORD1_SPACE_USED;
5617
5618	if (cache_validity & NFS_INO_INVALID_SIZE)
5619		bitmask[0] |= FATTR4_WORD0_SIZE;
5620
5621	for (i = 0; i < NFS4_BITMASK_SZ; i++)
5622		bitmask[i] &= server->attr_bitmask[i];
5623}
5624
5625static void nfs4_proc_write_setup(struct nfs_pgio_header *hdr,
5626				  struct rpc_message *msg,
5627				  struct rpc_clnt **clnt)
5628{
5629	struct nfs_server *server = NFS_SERVER(hdr->inode);
5630
5631	if (!nfs4_write_need_cache_consistency_data(hdr)) {
5632		hdr->args.bitmask = NULL;
5633		hdr->res.fattr = NULL;
5634	} else {
5635		nfs4_bitmask_set(hdr->args.bitmask_store,
5636				 server->cache_consistency_bitmask,
5637				 hdr->inode, NFS_INO_INVALID_BLOCKS);
5638		hdr->args.bitmask = hdr->args.bitmask_store;
5639	}
5640
5641	if (!hdr->pgio_done_cb)
5642		hdr->pgio_done_cb = nfs4_write_done_cb;
5643	hdr->res.server = server;
5644	hdr->timestamp   = jiffies;
5645
5646	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_WRITE];
5647	nfs4_init_sequence(&hdr->args.seq_args, &hdr->res.seq_res, 0, 0);
5648	nfs4_state_protect_write(hdr->ds_clp ? hdr->ds_clp : server->nfs_client, clnt, msg, hdr);
5649}
5650
5651static void nfs4_proc_commit_rpc_prepare(struct rpc_task *task, struct nfs_commit_data *data)
5652{
5653	nfs4_setup_sequence(NFS_SERVER(data->inode)->nfs_client,
5654			&data->args.seq_args,
5655			&data->res.seq_res,
5656			task);
5657}
5658
5659static int nfs4_commit_done_cb(struct rpc_task *task, struct nfs_commit_data *data)
5660{
5661	struct inode *inode = data->inode;
5662
5663	trace_nfs4_commit(data, task->tk_status);
5664	if (nfs4_async_handle_error(task, NFS_SERVER(inode),
5665				    NULL, NULL) == -EAGAIN) {
5666		rpc_restart_call_prepare(task);
5667		return -EAGAIN;
5668	}
5669	return 0;
5670}
5671
5672static int nfs4_commit_done(struct rpc_task *task, struct nfs_commit_data *data)
5673{
5674	if (!nfs4_sequence_done(task, &data->res.seq_res))
5675		return -EAGAIN;
5676	return data->commit_done_cb(task, data);
5677}
5678
5679static void nfs4_proc_commit_setup(struct nfs_commit_data *data, struct rpc_message *msg,
5680				   struct rpc_clnt **clnt)
5681{
5682	struct nfs_server *server = NFS_SERVER(data->inode);
5683
5684	if (data->commit_done_cb == NULL)
5685		data->commit_done_cb = nfs4_commit_done_cb;
5686	data->res.server = server;
5687	msg->rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT];
5688	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
5689	nfs4_state_protect(data->ds_clp ? data->ds_clp : server->nfs_client,
5690			NFS_SP4_MACH_CRED_COMMIT, clnt, msg);
5691}
5692
5693static int _nfs4_proc_commit(struct file *dst, struct nfs_commitargs *args,
5694				struct nfs_commitres *res)
5695{
5696	struct inode *dst_inode = file_inode(dst);
5697	struct nfs_server *server = NFS_SERVER(dst_inode);
5698	struct rpc_message msg = {
5699		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COMMIT],
5700		.rpc_argp = args,
5701		.rpc_resp = res,
5702	};
5703
5704	args->fh = NFS_FH(dst_inode);
5705	return nfs4_call_sync(server->client, server, &msg,
5706			&args->seq_args, &res->seq_res, 1);
5707}
5708
5709int nfs4_proc_commit(struct file *dst, __u64 offset, __u32 count, struct nfs_commitres *res)
5710{
5711	struct nfs_commitargs args = {
5712		.offset = offset,
5713		.count = count,
5714	};
5715	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
5716	struct nfs4_exception exception = { };
5717	int status;
5718
5719	do {
5720		status = _nfs4_proc_commit(dst, &args, res);
5721		status = nfs4_handle_exception(dst_server, status, &exception);
5722	} while (exception.retry);
5723
5724	return status;
5725}
5726
5727struct nfs4_renewdata {
5728	struct nfs_client	*client;
5729	unsigned long		timestamp;
5730};
5731
5732/*
5733 * nfs4_proc_async_renew(): This is not one of the nfs_rpc_ops; it is a special
5734 * standalone procedure for queueing an asynchronous RENEW.
5735 */
5736static void nfs4_renew_release(void *calldata)
5737{
5738	struct nfs4_renewdata *data = calldata;
5739	struct nfs_client *clp = data->client;
5740
5741	if (refcount_read(&clp->cl_count) > 1)
5742		nfs4_schedule_state_renewal(clp);
5743	nfs_put_client(clp);
5744	kfree(data);
5745}
5746
5747static void nfs4_renew_done(struct rpc_task *task, void *calldata)
5748{
5749	struct nfs4_renewdata *data = calldata;
5750	struct nfs_client *clp = data->client;
5751	unsigned long timestamp = data->timestamp;
5752
5753	trace_nfs4_renew_async(clp, task->tk_status);
5754	switch (task->tk_status) {
5755	case 0:
5756		break;
5757	case -NFS4ERR_LEASE_MOVED:
5758		nfs4_schedule_lease_moved_recovery(clp);
5759		break;
5760	default:
5761		/* Unless we're shutting down, schedule state recovery! */
5762		if (test_bit(NFS_CS_RENEWD, &clp->cl_res_state) == 0)
5763			return;
5764		if (task->tk_status != NFS4ERR_CB_PATH_DOWN) {
5765			nfs4_schedule_lease_recovery(clp);
5766			return;
5767		}
5768		nfs4_schedule_path_down_recovery(clp);
5769	}
5770	do_renew_lease(clp, timestamp);
5771}
5772
5773static const struct rpc_call_ops nfs4_renew_ops = {
5774	.rpc_call_done = nfs4_renew_done,
5775	.rpc_release = nfs4_renew_release,
5776};
5777
5778static int nfs4_proc_async_renew(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
5779{
5780	struct rpc_message msg = {
5781		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5782		.rpc_argp	= clp,
5783		.rpc_cred	= cred,
5784	};
5785	struct nfs4_renewdata *data;
5786
5787	if (renew_flags == 0)
5788		return 0;
5789	if (!refcount_inc_not_zero(&clp->cl_count))
5790		return -EIO;
5791	data = kmalloc(sizeof(*data), GFP_NOFS);
5792	if (data == NULL) {
5793		nfs_put_client(clp);
5794		return -ENOMEM;
5795	}
5796	data->client = clp;
5797	data->timestamp = jiffies;
5798	return rpc_call_async(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT,
5799			&nfs4_renew_ops, data);
5800}
5801
5802static int nfs4_proc_renew(struct nfs_client *clp, const struct cred *cred)
5803{
5804	struct rpc_message msg = {
5805		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_RENEW],
5806		.rpc_argp	= clp,
5807		.rpc_cred	= cred,
5808	};
5809	unsigned long now = jiffies;
5810	int status;
5811
5812	status = rpc_call_sync(clp->cl_rpcclient, &msg, RPC_TASK_TIMEOUT);
5813	if (status < 0)
5814		return status;
5815	do_renew_lease(clp, now);
5816	return 0;
5817}
5818
5819static bool nfs4_server_supports_acls(const struct nfs_server *server,
5820				      enum nfs4_acl_type type)
5821{
5822	switch (type) {
5823	default:
5824		return server->attr_bitmask[0] & FATTR4_WORD0_ACL;
5825	case NFS4ACL_DACL:
5826		return server->attr_bitmask[1] & FATTR4_WORD1_DACL;
5827	case NFS4ACL_SACL:
5828		return server->attr_bitmask[1] & FATTR4_WORD1_SACL;
5829	}
5830}
5831
5832/* Assuming that XATTR_SIZE_MAX is a multiple of PAGE_SIZE, and that
5833 * it's OK to put sizeof(void) * (XATTR_SIZE_MAX/PAGE_SIZE) bytes on
5834 * the stack.
5835 */
5836#define NFS4ACL_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
5837
5838int nfs4_buf_to_pages_noslab(const void *buf, size_t buflen,
5839		struct page **pages)
5840{
5841	struct page *newpage, **spages;
5842	int rc = 0;
5843	size_t len;
5844	spages = pages;
5845
5846	do {
5847		len = min_t(size_t, PAGE_SIZE, buflen);
5848		newpage = alloc_page(GFP_KERNEL);
5849
5850		if (newpage == NULL)
5851			goto unwind;
5852		memcpy(page_address(newpage), buf, len);
5853		buf += len;
5854		buflen -= len;
5855		*pages++ = newpage;
5856		rc++;
5857	} while (buflen != 0);
5858
5859	return rc;
5860
5861unwind:
5862	for(; rc > 0; rc--)
5863		__free_page(spages[rc-1]);
5864	return -ENOMEM;
5865}
5866
5867struct nfs4_cached_acl {
5868	enum nfs4_acl_type type;
5869	int cached;
5870	size_t len;
5871	char data[];
5872};
5873
5874static void nfs4_set_cached_acl(struct inode *inode, struct nfs4_cached_acl *acl)
5875{
5876	struct nfs_inode *nfsi = NFS_I(inode);
5877
5878	spin_lock(&inode->i_lock);
5879	kfree(nfsi->nfs4_acl);
5880	nfsi->nfs4_acl = acl;
5881	spin_unlock(&inode->i_lock);
5882}
5883
5884static void nfs4_zap_acl_attr(struct inode *inode)
5885{
5886	nfs4_set_cached_acl(inode, NULL);
5887}
5888
5889static ssize_t nfs4_read_cached_acl(struct inode *inode, char *buf,
5890				    size_t buflen, enum nfs4_acl_type type)
5891{
5892	struct nfs_inode *nfsi = NFS_I(inode);
5893	struct nfs4_cached_acl *acl;
5894	int ret = -ENOENT;
5895
5896	spin_lock(&inode->i_lock);
5897	acl = nfsi->nfs4_acl;
5898	if (acl == NULL)
5899		goto out;
5900	if (acl->type != type)
5901		goto out;
5902	if (buf == NULL) /* user is just asking for length */
5903		goto out_len;
5904	if (acl->cached == 0)
5905		goto out;
5906	ret = -ERANGE; /* see getxattr(2) man page */
5907	if (acl->len > buflen)
5908		goto out;
5909	memcpy(buf, acl->data, acl->len);
5910out_len:
5911	ret = acl->len;
5912out:
5913	spin_unlock(&inode->i_lock);
5914	return ret;
5915}
5916
5917static void nfs4_write_cached_acl(struct inode *inode, struct page **pages,
5918				  size_t pgbase, size_t acl_len,
5919				  enum nfs4_acl_type type)
5920{
5921	struct nfs4_cached_acl *acl;
5922	size_t buflen = sizeof(*acl) + acl_len;
5923
5924	if (buflen <= PAGE_SIZE) {
5925		acl = kmalloc(buflen, GFP_KERNEL);
5926		if (acl == NULL)
5927			goto out;
5928		acl->cached = 1;
5929		_copy_from_pages(acl->data, pages, pgbase, acl_len);
5930	} else {
5931		acl = kmalloc(sizeof(*acl), GFP_KERNEL);
5932		if (acl == NULL)
5933			goto out;
5934		acl->cached = 0;
5935	}
5936	acl->type = type;
5937	acl->len = acl_len;
5938out:
5939	nfs4_set_cached_acl(inode, acl);
5940}
5941
5942/*
5943 * The getxattr API returns the required buffer length when called with a
5944 * NULL buf. The NFSv4 acl tool then calls getxattr again after allocating
5945 * the required buf.  On a NULL buf, we send a page of data to the server
5946 * guessing that the ACL request can be serviced by a page. If so, we cache
5947 * up to the page of ACL data, and the 2nd call to getxattr is serviced by
5948 * the cache. If not so, we throw away the page, and cache the required
5949 * length. The next getxattr call will then produce another round trip to
5950 * the server, this time with the input buf of the required size.
5951 */
5952static ssize_t __nfs4_get_acl_uncached(struct inode *inode, void *buf,
5953				       size_t buflen, enum nfs4_acl_type type)
5954{
5955	struct page **pages;
5956	struct nfs_getaclargs args = {
5957		.fh = NFS_FH(inode),
5958		.acl_type = type,
5959		.acl_len = buflen,
5960	};
5961	struct nfs_getaclres res = {
5962		.acl_type = type,
5963		.acl_len = buflen,
5964	};
5965	struct rpc_message msg = {
5966		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETACL],
5967		.rpc_argp = &args,
5968		.rpc_resp = &res,
5969	};
5970	unsigned int npages;
5971	int ret = -ENOMEM, i;
5972	struct nfs_server *server = NFS_SERVER(inode);
5973
5974	if (buflen == 0)
5975		buflen = server->rsize;
5976
5977	npages = DIV_ROUND_UP(buflen, PAGE_SIZE) + 1;
5978	pages = kmalloc_array(npages, sizeof(struct page *), GFP_KERNEL);
5979	if (!pages)
5980		return -ENOMEM;
5981
5982	args.acl_pages = pages;
5983
5984	for (i = 0; i < npages; i++) {
5985		pages[i] = alloc_page(GFP_KERNEL);
5986		if (!pages[i])
5987			goto out_free;
5988	}
5989
5990	/* for decoding across pages */
5991	res.acl_scratch = alloc_page(GFP_KERNEL);
5992	if (!res.acl_scratch)
5993		goto out_free;
5994
5995	args.acl_len = npages * PAGE_SIZE;
5996
5997	dprintk("%s  buf %p buflen %zu npages %d args.acl_len %zu\n",
5998		__func__, buf, buflen, npages, args.acl_len);
5999	ret = nfs4_call_sync(NFS_SERVER(inode)->client, NFS_SERVER(inode),
6000			     &msg, &args.seq_args, &res.seq_res, 0);
6001	if (ret)
6002		goto out_free;
6003
6004	/* Handle the case where the passed-in buffer is too short */
6005	if (res.acl_flags & NFS4_ACL_TRUNC) {
6006		/* Did the user only issue a request for the acl length? */
6007		if (buf == NULL)
6008			goto out_ok;
6009		ret = -ERANGE;
6010		goto out_free;
6011	}
6012	nfs4_write_cached_acl(inode, pages, res.acl_data_offset, res.acl_len,
6013			      type);
6014	if (buf) {
6015		if (res.acl_len > buflen) {
6016			ret = -ERANGE;
6017			goto out_free;
6018		}
6019		_copy_from_pages(buf, pages, res.acl_data_offset, res.acl_len);
6020	}
6021out_ok:
6022	ret = res.acl_len;
6023out_free:
6024	while (--i >= 0)
6025		__free_page(pages[i]);
6026	if (res.acl_scratch)
6027		__free_page(res.acl_scratch);
6028	kfree(pages);
6029	return ret;
6030}
6031
6032static ssize_t nfs4_get_acl_uncached(struct inode *inode, void *buf,
6033				     size_t buflen, enum nfs4_acl_type type)
6034{
6035	struct nfs4_exception exception = {
6036		.interruptible = true,
6037	};
6038	ssize_t ret;
6039	do {
6040		ret = __nfs4_get_acl_uncached(inode, buf, buflen, type);
6041		trace_nfs4_get_acl(inode, ret);
6042		if (ret >= 0)
6043			break;
6044		ret = nfs4_handle_exception(NFS_SERVER(inode), ret, &exception);
6045	} while (exception.retry);
6046	return ret;
6047}
6048
6049static ssize_t nfs4_proc_get_acl(struct inode *inode, void *buf, size_t buflen,
6050				 enum nfs4_acl_type type)
6051{
6052	struct nfs_server *server = NFS_SERVER(inode);
6053	int ret;
6054
6055	if (!nfs4_server_supports_acls(server, type))
6056		return -EOPNOTSUPP;
6057	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
6058	if (ret < 0)
6059		return ret;
6060	if (NFS_I(inode)->cache_validity & NFS_INO_INVALID_ACL)
6061		nfs_zap_acl_cache(inode);
6062	ret = nfs4_read_cached_acl(inode, buf, buflen, type);
6063	if (ret != -ENOENT)
6064		/* -ENOENT is returned if there is no ACL or if there is an ACL
6065		 * but no cached acl data, just the acl length */
6066		return ret;
6067	return nfs4_get_acl_uncached(inode, buf, buflen, type);
6068}
6069
6070static int __nfs4_proc_set_acl(struct inode *inode, const void *buf,
6071			       size_t buflen, enum nfs4_acl_type type)
6072{
6073	struct nfs_server *server = NFS_SERVER(inode);
6074	struct page *pages[NFS4ACL_MAXPAGES];
6075	struct nfs_setaclargs arg = {
6076		.fh = NFS_FH(inode),
6077		.acl_type = type,
6078		.acl_len = buflen,
6079		.acl_pages = pages,
6080	};
6081	struct nfs_setaclres res;
6082	struct rpc_message msg = {
6083		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETACL],
6084		.rpc_argp	= &arg,
6085		.rpc_resp	= &res,
6086	};
6087	unsigned int npages = DIV_ROUND_UP(buflen, PAGE_SIZE);
6088	int ret, i;
6089
6090	/* You can't remove system.nfs4_acl: */
6091	if (buflen == 0)
6092		return -EINVAL;
6093	if (!nfs4_server_supports_acls(server, type))
6094		return -EOPNOTSUPP;
6095	if (npages > ARRAY_SIZE(pages))
6096		return -ERANGE;
6097	i = nfs4_buf_to_pages_noslab(buf, buflen, arg.acl_pages);
6098	if (i < 0)
6099		return i;
6100	nfs4_inode_make_writeable(inode);
6101	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6102
6103	/*
6104	 * Free each page after tx, so the only ref left is
6105	 * held by the network stack
6106	 */
6107	for (; i > 0; i--)
6108		put_page(pages[i-1]);
6109
6110	/*
6111	 * Acl update can result in inode attribute update.
6112	 * so mark the attribute cache invalid.
6113	 */
6114	spin_lock(&inode->i_lock);
6115	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
6116					     NFS_INO_INVALID_CTIME |
6117					     NFS_INO_REVAL_FORCED);
6118	spin_unlock(&inode->i_lock);
6119	nfs_access_zap_cache(inode);
6120	nfs_zap_acl_cache(inode);
6121	return ret;
6122}
6123
6124static int nfs4_proc_set_acl(struct inode *inode, const void *buf,
6125			     size_t buflen, enum nfs4_acl_type type)
6126{
6127	struct nfs4_exception exception = { };
6128	int err;
6129	do {
6130		err = __nfs4_proc_set_acl(inode, buf, buflen, type);
6131		trace_nfs4_set_acl(inode, err);
6132		if (err == -NFS4ERR_BADOWNER || err == -NFS4ERR_BADNAME) {
6133			/*
6134			 * no need to retry since the kernel
6135			 * isn't involved in encoding the ACEs.
6136			 */
6137			err = -EINVAL;
6138			break;
6139		}
6140		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6141				&exception);
6142	} while (exception.retry);
6143	return err;
6144}
6145
6146#ifdef CONFIG_NFS_V4_SECURITY_LABEL
6147static int _nfs4_get_security_label(struct inode *inode, void *buf,
6148					size_t buflen)
6149{
6150	struct nfs_server *server = NFS_SERVER(inode);
6151	struct nfs4_label label = {0, 0, buflen, buf};
6152
6153	u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6154	struct nfs_fattr fattr = {
6155		.label = &label,
6156	};
6157	struct nfs4_getattr_arg arg = {
6158		.fh		= NFS_FH(inode),
6159		.bitmask	= bitmask,
6160	};
6161	struct nfs4_getattr_res res = {
6162		.fattr		= &fattr,
6163		.server		= server,
6164	};
6165	struct rpc_message msg = {
6166		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETATTR],
6167		.rpc_argp	= &arg,
6168		.rpc_resp	= &res,
6169	};
6170	int ret;
6171
6172	nfs_fattr_init(&fattr);
6173
6174	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 0);
6175	if (ret)
6176		return ret;
6177	if (!(fattr.valid & NFS_ATTR_FATTR_V4_SECURITY_LABEL))
6178		return -ENOENT;
6179	return label.len;
6180}
6181
6182static int nfs4_get_security_label(struct inode *inode, void *buf,
6183					size_t buflen)
6184{
6185	struct nfs4_exception exception = {
6186		.interruptible = true,
6187	};
6188	int err;
6189
6190	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6191		return -EOPNOTSUPP;
6192
6193	do {
6194		err = _nfs4_get_security_label(inode, buf, buflen);
6195		trace_nfs4_get_security_label(inode, err);
6196		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6197				&exception);
6198	} while (exception.retry);
6199	return err;
6200}
6201
6202static int _nfs4_do_set_security_label(struct inode *inode,
6203		struct nfs4_label *ilabel,
6204		struct nfs_fattr *fattr)
6205{
6206
6207	struct iattr sattr = {0};
6208	struct nfs_server *server = NFS_SERVER(inode);
6209	const u32 bitmask[3] = { 0, 0, FATTR4_WORD2_SECURITY_LABEL };
6210	struct nfs_setattrargs arg = {
6211		.fh		= NFS_FH(inode),
6212		.iap		= &sattr,
6213		.server		= server,
6214		.bitmask	= bitmask,
6215		.label		= ilabel,
6216	};
6217	struct nfs_setattrres res = {
6218		.fattr		= fattr,
6219		.server		= server,
6220	};
6221	struct rpc_message msg = {
6222		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETATTR],
6223		.rpc_argp	= &arg,
6224		.rpc_resp	= &res,
6225	};
6226	int status;
6227
6228	nfs4_stateid_copy(&arg.stateid, &zero_stateid);
6229
6230	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6231	if (status)
6232		dprintk("%s failed: %d\n", __func__, status);
6233
6234	return status;
6235}
6236
6237static int nfs4_do_set_security_label(struct inode *inode,
6238		struct nfs4_label *ilabel,
6239		struct nfs_fattr *fattr)
6240{
6241	struct nfs4_exception exception = { };
6242	int err;
6243
6244	do {
6245		err = _nfs4_do_set_security_label(inode, ilabel, fattr);
6246		trace_nfs4_set_security_label(inode, err);
6247		err = nfs4_handle_exception(NFS_SERVER(inode), err,
6248				&exception);
6249	} while (exception.retry);
6250	return err;
6251}
6252
6253static int
6254nfs4_set_security_label(struct inode *inode, const void *buf, size_t buflen)
6255{
6256	struct nfs4_label ilabel = {0, 0, buflen, (char *)buf };
6257	struct nfs_fattr *fattr;
6258	int status;
6259
6260	if (!nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL))
6261		return -EOPNOTSUPP;
6262
6263	fattr = nfs_alloc_fattr_with_label(NFS_SERVER(inode));
6264	if (fattr == NULL)
6265		return -ENOMEM;
6266
6267	status = nfs4_do_set_security_label(inode, &ilabel, fattr);
6268	if (status == 0)
6269		nfs_setsecurity(inode, fattr);
6270
6271	return status;
6272}
6273#endif	/* CONFIG_NFS_V4_SECURITY_LABEL */
6274
6275
6276static void nfs4_init_boot_verifier(const struct nfs_client *clp,
6277				    nfs4_verifier *bootverf)
6278{
6279	__be32 verf[2];
6280
6281	if (test_bit(NFS4CLNT_PURGE_STATE, &clp->cl_state)) {
6282		/* An impossible timestamp guarantees this value
6283		 * will never match a generated boot time. */
6284		verf[0] = cpu_to_be32(U32_MAX);
6285		verf[1] = cpu_to_be32(U32_MAX);
6286	} else {
6287		struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6288		u64 ns = ktime_to_ns(nn->boot_time);
6289
6290		verf[0] = cpu_to_be32(ns >> 32);
6291		verf[1] = cpu_to_be32(ns);
6292	}
6293	memcpy(bootverf->data, verf, sizeof(bootverf->data));
6294}
6295
6296static size_t
6297nfs4_get_uniquifier(struct nfs_client *clp, char *buf, size_t buflen)
6298{
6299	struct nfs_net *nn = net_generic(clp->cl_net, nfs_net_id);
6300	struct nfs_netns_client *nn_clp = nn->nfs_client;
6301	const char *id;
6302
6303	buf[0] = '\0';
6304
6305	if (nn_clp) {
6306		rcu_read_lock();
6307		id = rcu_dereference(nn_clp->identifier);
6308		if (id)
6309			strscpy(buf, id, buflen);
6310		rcu_read_unlock();
6311	}
6312
6313	if (nfs4_client_id_uniquifier[0] != '\0' && buf[0] == '\0')
6314		strscpy(buf, nfs4_client_id_uniquifier, buflen);
6315
6316	return strlen(buf);
6317}
6318
6319static int
6320nfs4_init_nonuniform_client_string(struct nfs_client *clp)
6321{
6322	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6323	size_t buflen;
6324	size_t len;
6325	char *str;
6326
6327	if (clp->cl_owner_id != NULL)
6328		return 0;
6329
6330	rcu_read_lock();
6331	len = 14 +
6332		strlen(clp->cl_rpcclient->cl_nodename) +
6333		1 +
6334		strlen(rpc_peeraddr2str(clp->cl_rpcclient, RPC_DISPLAY_ADDR)) +
6335		1;
6336	rcu_read_unlock();
6337
6338	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6339	if (buflen)
6340		len += buflen + 1;
6341
6342	if (len > NFS4_OPAQUE_LIMIT + 1)
6343		return -EINVAL;
6344
6345	/*
6346	 * Since this string is allocated at mount time, and held until the
6347	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6348	 * about a memory-reclaim deadlock.
6349	 */
6350	str = kmalloc(len, GFP_KERNEL);
6351	if (!str)
6352		return -ENOMEM;
6353
6354	rcu_read_lock();
6355	if (buflen)
6356		scnprintf(str, len, "Linux NFSv4.0 %s/%s/%s",
6357			  clp->cl_rpcclient->cl_nodename, buf,
6358			  rpc_peeraddr2str(clp->cl_rpcclient,
6359					   RPC_DISPLAY_ADDR));
6360	else
6361		scnprintf(str, len, "Linux NFSv4.0 %s/%s",
6362			  clp->cl_rpcclient->cl_nodename,
6363			  rpc_peeraddr2str(clp->cl_rpcclient,
6364					   RPC_DISPLAY_ADDR));
6365	rcu_read_unlock();
6366
6367	clp->cl_owner_id = str;
6368	return 0;
6369}
6370
6371static int
6372nfs4_init_uniform_client_string(struct nfs_client *clp)
6373{
6374	char buf[NFS4_CLIENT_ID_UNIQ_LEN];
6375	size_t buflen;
6376	size_t len;
6377	char *str;
6378
6379	if (clp->cl_owner_id != NULL)
6380		return 0;
6381
6382	len = 10 + 10 + 1 + 10 + 1 +
6383		strlen(clp->cl_rpcclient->cl_nodename) + 1;
6384
6385	buflen = nfs4_get_uniquifier(clp, buf, sizeof(buf));
6386	if (buflen)
6387		len += buflen + 1;
6388
6389	if (len > NFS4_OPAQUE_LIMIT + 1)
6390		return -EINVAL;
6391
6392	/*
6393	 * Since this string is allocated at mount time, and held until the
6394	 * nfs_client is destroyed, we can use GFP_KERNEL here w/o worrying
6395	 * about a memory-reclaim deadlock.
6396	 */
6397	str = kmalloc(len, GFP_KERNEL);
6398	if (!str)
6399		return -ENOMEM;
6400
6401	if (buflen)
6402		scnprintf(str, len, "Linux NFSv%u.%u %s/%s",
6403			  clp->rpc_ops->version, clp->cl_minorversion,
6404			  buf, clp->cl_rpcclient->cl_nodename);
6405	else
6406		scnprintf(str, len, "Linux NFSv%u.%u %s",
6407			  clp->rpc_ops->version, clp->cl_minorversion,
6408			  clp->cl_rpcclient->cl_nodename);
6409	clp->cl_owner_id = str;
6410	return 0;
6411}
6412
6413/*
6414 * nfs4_callback_up_net() starts only "tcp" and "tcp6" callback
6415 * services.  Advertise one based on the address family of the
6416 * clientaddr.
6417 */
6418static unsigned int
6419nfs4_init_callback_netid(const struct nfs_client *clp, char *buf, size_t len)
6420{
6421	if (strchr(clp->cl_ipaddr, ':') != NULL)
6422		return scnprintf(buf, len, "tcp6");
6423	else
6424		return scnprintf(buf, len, "tcp");
6425}
6426
6427static void nfs4_setclientid_done(struct rpc_task *task, void *calldata)
6428{
6429	struct nfs4_setclientid *sc = calldata;
6430
6431	if (task->tk_status == 0)
6432		sc->sc_cred = get_rpccred(task->tk_rqstp->rq_cred);
6433}
6434
6435static const struct rpc_call_ops nfs4_setclientid_ops = {
6436	.rpc_call_done = nfs4_setclientid_done,
6437};
6438
6439/**
6440 * nfs4_proc_setclientid - Negotiate client ID
6441 * @clp: state data structure
6442 * @program: RPC program for NFSv4 callback service
6443 * @port: IP port number for NFS4 callback service
6444 * @cred: credential to use for this call
6445 * @res: where to place the result
6446 *
6447 * Returns zero, a negative errno, or a negative NFS4ERR status code.
6448 */
6449int nfs4_proc_setclientid(struct nfs_client *clp, u32 program,
6450		unsigned short port, const struct cred *cred,
6451		struct nfs4_setclientid_res *res)
6452{
6453	nfs4_verifier sc_verifier;
6454	struct nfs4_setclientid setclientid = {
6455		.sc_verifier = &sc_verifier,
6456		.sc_prog = program,
6457		.sc_clnt = clp,
6458	};
6459	struct rpc_message msg = {
6460		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID],
6461		.rpc_argp = &setclientid,
6462		.rpc_resp = res,
6463		.rpc_cred = cred,
6464	};
6465	struct rpc_task_setup task_setup_data = {
6466		.rpc_client = clp->cl_rpcclient,
6467		.rpc_message = &msg,
6468		.callback_ops = &nfs4_setclientid_ops,
6469		.callback_data = &setclientid,
6470		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
6471	};
6472	unsigned long now = jiffies;
6473	int status;
6474
6475	/* nfs_client_id4 */
6476	nfs4_init_boot_verifier(clp, &sc_verifier);
6477
6478	if (test_bit(NFS_CS_MIGRATION, &clp->cl_flags))
6479		status = nfs4_init_uniform_client_string(clp);
6480	else
6481		status = nfs4_init_nonuniform_client_string(clp);
6482
6483	if (status)
6484		goto out;
6485
6486	/* cb_client4 */
6487	setclientid.sc_netid_len =
6488				nfs4_init_callback_netid(clp,
6489						setclientid.sc_netid,
6490						sizeof(setclientid.sc_netid));
6491	setclientid.sc_uaddr_len = scnprintf(setclientid.sc_uaddr,
6492				sizeof(setclientid.sc_uaddr), "%s.%u.%u",
6493				clp->cl_ipaddr, port >> 8, port & 255);
6494
6495	dprintk("NFS call  setclientid auth=%s, '%s'\n",
6496		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6497		clp->cl_owner_id);
6498
6499	status = nfs4_call_sync_custom(&task_setup_data);
6500	if (setclientid.sc_cred) {
6501		kfree(clp->cl_acceptor);
6502		clp->cl_acceptor = rpcauth_stringify_acceptor(setclientid.sc_cred);
6503		put_rpccred(setclientid.sc_cred);
6504	}
6505
6506	if (status == 0)
6507		do_renew_lease(clp, now);
6508out:
6509	trace_nfs4_setclientid(clp, status);
6510	dprintk("NFS reply setclientid: %d\n", status);
6511	return status;
6512}
6513
6514/**
6515 * nfs4_proc_setclientid_confirm - Confirm client ID
6516 * @clp: state data structure
6517 * @arg: result of a previous SETCLIENTID
6518 * @cred: credential to use for this call
6519 *
6520 * Returns zero, a negative errno, or a negative NFS4ERR status code.
6521 */
6522int nfs4_proc_setclientid_confirm(struct nfs_client *clp,
6523		struct nfs4_setclientid_res *arg,
6524		const struct cred *cred)
6525{
6526	struct rpc_message msg = {
6527		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SETCLIENTID_CONFIRM],
6528		.rpc_argp = arg,
6529		.rpc_cred = cred,
6530	};
6531	int status;
6532
6533	dprintk("NFS call  setclientid_confirm auth=%s, (client ID %llx)\n",
6534		clp->cl_rpcclient->cl_auth->au_ops->au_name,
6535		clp->cl_clientid);
6536	status = rpc_call_sync(clp->cl_rpcclient, &msg,
6537			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
6538	trace_nfs4_setclientid_confirm(clp, status);
6539	dprintk("NFS reply setclientid_confirm: %d\n", status);
6540	return status;
6541}
6542
6543struct nfs4_delegreturndata {
6544	struct nfs4_delegreturnargs args;
6545	struct nfs4_delegreturnres res;
6546	struct nfs_fh fh;
6547	nfs4_stateid stateid;
6548	unsigned long timestamp;
6549	struct {
6550		struct nfs4_layoutreturn_args arg;
6551		struct nfs4_layoutreturn_res res;
6552		struct nfs4_xdr_opaque_data ld_private;
6553		u32 roc_barrier;
6554		bool roc;
6555	} lr;
6556	struct nfs_fattr fattr;
6557	int rpc_status;
6558	struct inode *inode;
6559};
6560
6561static void nfs4_delegreturn_done(struct rpc_task *task, void *calldata)
6562{
6563	struct nfs4_delegreturndata *data = calldata;
6564	struct nfs4_exception exception = {
6565		.inode = data->inode,
6566		.stateid = &data->stateid,
6567		.task_is_privileged = data->args.seq_args.sa_privileged,
6568	};
6569
6570	if (!nfs4_sequence_done(task, &data->res.seq_res))
6571		return;
6572
6573	trace_nfs4_delegreturn_exit(&data->args, &data->res, task->tk_status);
6574
6575	/* Handle Layoutreturn errors */
6576	if (pnfs_roc_done(task, &data->args.lr_args, &data->res.lr_res,
6577			  &data->res.lr_ret) == -EAGAIN)
6578		goto out_restart;
6579
6580	switch (task->tk_status) {
6581	case 0:
6582		renew_lease(data->res.server, data->timestamp);
6583		break;
6584	case -NFS4ERR_ADMIN_REVOKED:
6585	case -NFS4ERR_DELEG_REVOKED:
6586	case -NFS4ERR_EXPIRED:
6587		nfs4_free_revoked_stateid(data->res.server,
6588				data->args.stateid,
6589				task->tk_msg.rpc_cred);
6590		fallthrough;
6591	case -NFS4ERR_BAD_STATEID:
6592	case -NFS4ERR_STALE_STATEID:
6593	case -ETIMEDOUT:
6594		task->tk_status = 0;
6595		break;
6596	case -NFS4ERR_OLD_STATEID:
6597		if (!nfs4_refresh_delegation_stateid(&data->stateid, data->inode))
6598			nfs4_stateid_seqid_inc(&data->stateid);
6599		if (data->args.bitmask) {
6600			data->args.bitmask = NULL;
6601			data->res.fattr = NULL;
6602		}
6603		goto out_restart;
6604	case -NFS4ERR_ACCESS:
6605		if (data->args.bitmask) {
6606			data->args.bitmask = NULL;
6607			data->res.fattr = NULL;
6608			goto out_restart;
6609		}
6610		fallthrough;
6611	default:
6612		task->tk_status = nfs4_async_handle_exception(task,
6613				data->res.server, task->tk_status,
6614				&exception);
6615		if (exception.retry)
6616			goto out_restart;
6617	}
6618	nfs_delegation_mark_returned(data->inode, data->args.stateid);
6619	data->rpc_status = task->tk_status;
6620	return;
6621out_restart:
6622	task->tk_status = 0;
6623	rpc_restart_call_prepare(task);
6624}
6625
6626static void nfs4_delegreturn_release(void *calldata)
6627{
6628	struct nfs4_delegreturndata *data = calldata;
6629	struct inode *inode = data->inode;
6630
6631	if (data->lr.roc)
6632		pnfs_roc_release(&data->lr.arg, &data->lr.res,
6633				 data->res.lr_ret);
6634	if (inode) {
6635		nfs4_fattr_set_prechange(&data->fattr,
6636					 inode_peek_iversion_raw(inode));
6637		nfs_refresh_inode(inode, &data->fattr);
6638		nfs_iput_and_deactive(inode);
6639	}
6640	kfree(calldata);
6641}
6642
6643static void nfs4_delegreturn_prepare(struct rpc_task *task, void *data)
6644{
6645	struct nfs4_delegreturndata *d_data;
6646	struct pnfs_layout_hdr *lo;
6647
6648	d_data = data;
6649
6650	if (!d_data->lr.roc && nfs4_wait_on_layoutreturn(d_data->inode, task)) {
6651		nfs4_sequence_done(task, &d_data->res.seq_res);
6652		return;
6653	}
6654
6655	lo = d_data->args.lr_args ? d_data->args.lr_args->layout : NULL;
6656	if (lo && !pnfs_layout_is_valid(lo)) {
6657		d_data->args.lr_args = NULL;
6658		d_data->res.lr_res = NULL;
6659	}
6660
6661	nfs4_setup_sequence(d_data->res.server->nfs_client,
6662			&d_data->args.seq_args,
6663			&d_data->res.seq_res,
6664			task);
6665}
6666
6667static const struct rpc_call_ops nfs4_delegreturn_ops = {
6668	.rpc_call_prepare = nfs4_delegreturn_prepare,
6669	.rpc_call_done = nfs4_delegreturn_done,
6670	.rpc_release = nfs4_delegreturn_release,
6671};
6672
6673static int _nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6674{
6675	struct nfs4_delegreturndata *data;
6676	struct nfs_server *server = NFS_SERVER(inode);
6677	struct rpc_task *task;
6678	struct rpc_message msg = {
6679		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DELEGRETURN],
6680		.rpc_cred = cred,
6681	};
6682	struct rpc_task_setup task_setup_data = {
6683		.rpc_client = server->client,
6684		.rpc_message = &msg,
6685		.callback_ops = &nfs4_delegreturn_ops,
6686		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT,
6687	};
6688	int status = 0;
6689
6690	if (nfs_server_capable(inode, NFS_CAP_MOVEABLE))
6691		task_setup_data.flags |= RPC_TASK_MOVEABLE;
6692
6693	data = kzalloc(sizeof(*data), GFP_KERNEL);
6694	if (data == NULL)
6695		return -ENOMEM;
6696
6697	nfs4_state_protect(server->nfs_client,
6698			NFS_SP4_MACH_CRED_CLEANUP,
6699			&task_setup_data.rpc_client, &msg);
6700
6701	data->args.fhandle = &data->fh;
6702	data->args.stateid = &data->stateid;
6703	nfs4_bitmask_set(data->args.bitmask_store,
6704			 server->cache_consistency_bitmask, inode, 0);
6705	data->args.bitmask = data->args.bitmask_store;
6706	nfs_copy_fh(&data->fh, NFS_FH(inode));
6707	nfs4_stateid_copy(&data->stateid, stateid);
6708	data->res.fattr = &data->fattr;
6709	data->res.server = server;
6710	data->res.lr_ret = -NFS4ERR_NOMATCHING_LAYOUT;
6711	data->lr.arg.ld_private = &data->lr.ld_private;
6712	nfs_fattr_init(data->res.fattr);
6713	data->timestamp = jiffies;
6714	data->rpc_status = 0;
6715	data->inode = nfs_igrab_and_active(inode);
6716	if (data->inode || issync) {
6717		data->lr.roc = pnfs_roc(inode, &data->lr.arg, &data->lr.res,
6718					cred);
6719		if (data->lr.roc) {
6720			data->args.lr_args = &data->lr.arg;
6721			data->res.lr_res = &data->lr.res;
6722		}
6723	}
6724
6725	if (!data->inode)
6726		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6727				   1);
6728	else
6729		nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1,
6730				   0);
6731	task_setup_data.callback_data = data;
6732	msg.rpc_argp = &data->args;
6733	msg.rpc_resp = &data->res;
6734	task = rpc_run_task(&task_setup_data);
6735	if (IS_ERR(task))
6736		return PTR_ERR(task);
6737	if (!issync)
6738		goto out;
6739	status = rpc_wait_for_completion_task(task);
6740	if (status != 0)
6741		goto out;
6742	status = data->rpc_status;
6743out:
6744	rpc_put_task(task);
6745	return status;
6746}
6747
6748int nfs4_proc_delegreturn(struct inode *inode, const struct cred *cred, const nfs4_stateid *stateid, int issync)
6749{
6750	struct nfs_server *server = NFS_SERVER(inode);
6751	struct nfs4_exception exception = { };
6752	int err;
6753	do {
6754		err = _nfs4_proc_delegreturn(inode, cred, stateid, issync);
6755		trace_nfs4_delegreturn(inode, stateid, err);
6756		switch (err) {
6757			case -NFS4ERR_STALE_STATEID:
6758			case -NFS4ERR_EXPIRED:
6759			case 0:
6760				return 0;
6761		}
6762		err = nfs4_handle_exception(server, err, &exception);
6763	} while (exception.retry);
6764	return err;
6765}
6766
6767static int _nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6768{
6769	struct inode *inode = state->inode;
6770	struct nfs_server *server = NFS_SERVER(inode);
6771	struct nfs_client *clp = server->nfs_client;
6772	struct nfs_lockt_args arg = {
6773		.fh = NFS_FH(inode),
6774		.fl = request,
6775	};
6776	struct nfs_lockt_res res = {
6777		.denied = request,
6778	};
6779	struct rpc_message msg = {
6780		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LOCKT],
6781		.rpc_argp	= &arg,
6782		.rpc_resp	= &res,
6783		.rpc_cred	= state->owner->so_cred,
6784	};
6785	struct nfs4_lock_state *lsp;
6786	int status;
6787
6788	arg.lock_owner.clientid = clp->cl_clientid;
6789	status = nfs4_set_lock_state(state, request);
6790	if (status != 0)
6791		goto out;
6792	lsp = request->fl_u.nfs4_fl.owner;
6793	arg.lock_owner.id = lsp->ls_seqid.owner_id;
6794	arg.lock_owner.s_dev = server->s_dev;
6795	status = nfs4_call_sync(server->client, server, &msg, &arg.seq_args, &res.seq_res, 1);
6796	switch (status) {
6797		case 0:
6798			request->c.flc_type = F_UNLCK;
6799			break;
6800		case -NFS4ERR_DENIED:
6801			status = 0;
6802	}
6803	request->fl_ops->fl_release_private(request);
6804	request->fl_ops = NULL;
6805out:
6806	return status;
6807}
6808
6809static int nfs4_proc_getlk(struct nfs4_state *state, int cmd, struct file_lock *request)
6810{
6811	struct nfs4_exception exception = {
6812		.interruptible = true,
6813	};
6814	int err;
6815
6816	do {
6817		err = _nfs4_proc_getlk(state, cmd, request);
6818		trace_nfs4_get_lock(request, state, cmd, err);
6819		err = nfs4_handle_exception(NFS_SERVER(state->inode), err,
6820				&exception);
6821	} while (exception.retry);
6822	return err;
6823}
6824
6825/*
6826 * Update the seqid of a lock stateid after receiving
6827 * NFS4ERR_OLD_STATEID
6828 */
6829static bool nfs4_refresh_lock_old_stateid(nfs4_stateid *dst,
6830		struct nfs4_lock_state *lsp)
6831{
6832	struct nfs4_state *state = lsp->ls_state;
6833	bool ret = false;
6834
6835	spin_lock(&state->state_lock);
6836	if (!nfs4_stateid_match_other(dst, &lsp->ls_stateid))
6837		goto out;
6838	if (!nfs4_stateid_is_newer(&lsp->ls_stateid, dst))
6839		nfs4_stateid_seqid_inc(dst);
6840	else
6841		dst->seqid = lsp->ls_stateid.seqid;
6842	ret = true;
6843out:
6844	spin_unlock(&state->state_lock);
6845	return ret;
6846}
6847
6848static bool nfs4_sync_lock_stateid(nfs4_stateid *dst,
6849		struct nfs4_lock_state *lsp)
6850{
6851	struct nfs4_state *state = lsp->ls_state;
6852	bool ret;
6853
6854	spin_lock(&state->state_lock);
6855	ret = !nfs4_stateid_match_other(dst, &lsp->ls_stateid);
6856	nfs4_stateid_copy(dst, &lsp->ls_stateid);
6857	spin_unlock(&state->state_lock);
6858	return ret;
6859}
6860
6861struct nfs4_unlockdata {
6862	struct nfs_locku_args arg;
6863	struct nfs_locku_res res;
6864	struct nfs4_lock_state *lsp;
6865	struct nfs_open_context *ctx;
6866	struct nfs_lock_context *l_ctx;
6867	struct file_lock fl;
6868	struct nfs_server *server;
6869	unsigned long timestamp;
6870};
6871
6872static struct nfs4_unlockdata *nfs4_alloc_unlockdata(struct file_lock *fl,
6873		struct nfs_open_context *ctx,
6874		struct nfs4_lock_state *lsp,
6875		struct nfs_seqid *seqid)
6876{
6877	struct nfs4_unlockdata *p;
6878	struct nfs4_state *state = lsp->ls_state;
6879	struct inode *inode = state->inode;
6880
6881	p = kzalloc(sizeof(*p), GFP_KERNEL);
6882	if (p == NULL)
6883		return NULL;
6884	p->arg.fh = NFS_FH(inode);
6885	p->arg.fl = &p->fl;
6886	p->arg.seqid = seqid;
6887	p->res.seqid = seqid;
6888	p->lsp = lsp;
6889	/* Ensure we don't close file until we're done freeing locks! */
6890	p->ctx = get_nfs_open_context(ctx);
6891	p->l_ctx = nfs_get_lock_context(ctx);
6892	locks_init_lock(&p->fl);
6893	locks_copy_lock(&p->fl, fl);
6894	p->server = NFS_SERVER(inode);
6895	spin_lock(&state->state_lock);
6896	nfs4_stateid_copy(&p->arg.stateid, &lsp->ls_stateid);
6897	spin_unlock(&state->state_lock);
6898	return p;
6899}
6900
6901static void nfs4_locku_release_calldata(void *data)
6902{
6903	struct nfs4_unlockdata *calldata = data;
6904	nfs_free_seqid(calldata->arg.seqid);
6905	nfs4_put_lock_state(calldata->lsp);
6906	nfs_put_lock_context(calldata->l_ctx);
6907	put_nfs_open_context(calldata->ctx);
6908	kfree(calldata);
6909}
6910
6911static void nfs4_locku_done(struct rpc_task *task, void *data)
6912{
6913	struct nfs4_unlockdata *calldata = data;
6914	struct nfs4_exception exception = {
6915		.inode = calldata->lsp->ls_state->inode,
6916		.stateid = &calldata->arg.stateid,
6917	};
6918
6919	if (!nfs4_sequence_done(task, &calldata->res.seq_res))
6920		return;
6921	switch (task->tk_status) {
6922		case 0:
6923			renew_lease(calldata->server, calldata->timestamp);
6924			locks_lock_inode_wait(calldata->lsp->ls_state->inode, &calldata->fl);
6925			if (nfs4_update_lock_stateid(calldata->lsp,
6926					&calldata->res.stateid))
6927				break;
6928			fallthrough;
6929		case -NFS4ERR_ADMIN_REVOKED:
6930		case -NFS4ERR_EXPIRED:
6931			nfs4_free_revoked_stateid(calldata->server,
6932					&calldata->arg.stateid,
6933					task->tk_msg.rpc_cred);
6934			fallthrough;
6935		case -NFS4ERR_BAD_STATEID:
6936		case -NFS4ERR_STALE_STATEID:
6937			if (nfs4_sync_lock_stateid(&calldata->arg.stateid,
6938						calldata->lsp))
6939				rpc_restart_call_prepare(task);
6940			break;
6941		case -NFS4ERR_OLD_STATEID:
6942			if (nfs4_refresh_lock_old_stateid(&calldata->arg.stateid,
6943						calldata->lsp))
6944				rpc_restart_call_prepare(task);
6945			break;
6946		default:
6947			task->tk_status = nfs4_async_handle_exception(task,
6948					calldata->server, task->tk_status,
6949					&exception);
6950			if (exception.retry)
6951				rpc_restart_call_prepare(task);
6952	}
6953	nfs_release_seqid(calldata->arg.seqid);
6954}
6955
6956static void nfs4_locku_prepare(struct rpc_task *task, void *data)
6957{
6958	struct nfs4_unlockdata *calldata = data;
6959
6960	if (test_bit(NFS_CONTEXT_UNLOCK, &calldata->l_ctx->open_context->flags) &&
6961		nfs_async_iocounter_wait(task, calldata->l_ctx))
6962		return;
6963
6964	if (nfs_wait_on_sequence(calldata->arg.seqid, task) != 0)
6965		goto out_wait;
6966	if (test_bit(NFS_LOCK_INITIALIZED, &calldata->lsp->ls_flags) == 0) {
6967		/* Note: exit _without_ running nfs4_locku_done */
6968		goto out_no_action;
6969	}
6970	calldata->timestamp = jiffies;
6971	if (nfs4_setup_sequence(calldata->server->nfs_client,
6972				&calldata->arg.seq_args,
6973				&calldata->res.seq_res,
6974				task) != 0)
6975		nfs_release_seqid(calldata->arg.seqid);
6976	return;
6977out_no_action:
6978	task->tk_action = NULL;
6979out_wait:
6980	nfs4_sequence_done(task, &calldata->res.seq_res);
6981}
6982
6983static const struct rpc_call_ops nfs4_locku_ops = {
6984	.rpc_call_prepare = nfs4_locku_prepare,
6985	.rpc_call_done = nfs4_locku_done,
6986	.rpc_release = nfs4_locku_release_calldata,
6987};
6988
6989static struct rpc_task *nfs4_do_unlck(struct file_lock *fl,
6990		struct nfs_open_context *ctx,
6991		struct nfs4_lock_state *lsp,
6992		struct nfs_seqid *seqid)
6993{
6994	struct nfs4_unlockdata *data;
6995	struct rpc_message msg = {
6996		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCKU],
6997		.rpc_cred = ctx->cred,
6998	};
6999	struct rpc_task_setup task_setup_data = {
7000		.rpc_client = NFS_CLIENT(lsp->ls_state->inode),
7001		.rpc_message = &msg,
7002		.callback_ops = &nfs4_locku_ops,
7003		.workqueue = nfsiod_workqueue,
7004		.flags = RPC_TASK_ASYNC,
7005	};
7006
7007	if (nfs_server_capable(lsp->ls_state->inode, NFS_CAP_MOVEABLE))
7008		task_setup_data.flags |= RPC_TASK_MOVEABLE;
7009
7010	nfs4_state_protect(NFS_SERVER(lsp->ls_state->inode)->nfs_client,
7011		NFS_SP4_MACH_CRED_CLEANUP, &task_setup_data.rpc_client, &msg);
7012
7013	/* Ensure this is an unlock - when canceling a lock, the
7014	 * canceled lock is passed in, and it won't be an unlock.
7015	 */
7016	fl->c.flc_type = F_UNLCK;
7017	if (fl->c.flc_flags & FL_CLOSE)
7018		set_bit(NFS_CONTEXT_UNLOCK, &ctx->flags);
7019
7020	data = nfs4_alloc_unlockdata(fl, ctx, lsp, seqid);
7021	if (data == NULL) {
7022		nfs_free_seqid(seqid);
7023		return ERR_PTR(-ENOMEM);
7024	}
7025
7026	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1, 0);
7027	msg.rpc_argp = &data->arg;
7028	msg.rpc_resp = &data->res;
7029	task_setup_data.callback_data = data;
7030	return rpc_run_task(&task_setup_data);
7031}
7032
7033static int nfs4_proc_unlck(struct nfs4_state *state, int cmd, struct file_lock *request)
7034{
7035	struct inode *inode = state->inode;
7036	struct nfs4_state_owner *sp = state->owner;
7037	struct nfs_inode *nfsi = NFS_I(inode);
7038	struct nfs_seqid *seqid;
7039	struct nfs4_lock_state *lsp;
7040	struct rpc_task *task;
7041	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7042	int status = 0;
7043	unsigned char saved_flags = request->c.flc_flags;
7044
7045	status = nfs4_set_lock_state(state, request);
7046	/* Unlock _before_ we do the RPC call */
7047	request->c.flc_flags |= FL_EXISTS;
7048	/* Exclude nfs_delegation_claim_locks() */
7049	mutex_lock(&sp->so_delegreturn_mutex);
7050	/* Exclude nfs4_reclaim_open_stateid() - note nesting! */
7051	down_read(&nfsi->rwsem);
7052	if (locks_lock_inode_wait(inode, request) == -ENOENT) {
7053		up_read(&nfsi->rwsem);
7054		mutex_unlock(&sp->so_delegreturn_mutex);
7055		goto out;
7056	}
7057	lsp = request->fl_u.nfs4_fl.owner;
7058	set_bit(NFS_LOCK_UNLOCKING, &lsp->ls_flags);
7059	up_read(&nfsi->rwsem);
7060	mutex_unlock(&sp->so_delegreturn_mutex);
7061	if (status != 0)
7062		goto out;
7063	/* Is this a delegated lock? */
7064	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) == 0)
7065		goto out;
7066	alloc_seqid = NFS_SERVER(inode)->nfs_client->cl_mvops->alloc_seqid;
7067	seqid = alloc_seqid(&lsp->ls_seqid, GFP_KERNEL);
7068	status = -ENOMEM;
7069	if (IS_ERR(seqid))
7070		goto out;
7071	task = nfs4_do_unlck(request,
7072			     nfs_file_open_context(request->c.flc_file),
7073			     lsp, seqid);
7074	status = PTR_ERR(task);
7075	if (IS_ERR(task))
7076		goto out;
7077	status = rpc_wait_for_completion_task(task);
7078	rpc_put_task(task);
7079out:
7080	request->c.flc_flags = saved_flags;
7081	trace_nfs4_unlock(request, state, F_SETLK, status);
7082	return status;
7083}
7084
7085struct nfs4_lockdata {
7086	struct nfs_lock_args arg;
7087	struct nfs_lock_res res;
7088	struct nfs4_lock_state *lsp;
7089	struct nfs_open_context *ctx;
7090	struct file_lock fl;
7091	unsigned long timestamp;
7092	int rpc_status;
7093	int cancelled;
7094	struct nfs_server *server;
7095};
7096
7097static struct nfs4_lockdata *nfs4_alloc_lockdata(struct file_lock *fl,
7098		struct nfs_open_context *ctx, struct nfs4_lock_state *lsp,
7099		gfp_t gfp_mask)
7100{
7101	struct nfs4_lockdata *p;
7102	struct inode *inode = lsp->ls_state->inode;
7103	struct nfs_server *server = NFS_SERVER(inode);
7104	struct nfs_seqid *(*alloc_seqid)(struct nfs_seqid_counter *, gfp_t);
7105
7106	p = kzalloc(sizeof(*p), gfp_mask);
7107	if (p == NULL)
7108		return NULL;
7109
7110	p->arg.fh = NFS_FH(inode);
7111	p->arg.fl = &p->fl;
7112	p->arg.open_seqid = nfs_alloc_seqid(&lsp->ls_state->owner->so_seqid, gfp_mask);
7113	if (IS_ERR(p->arg.open_seqid))
7114		goto out_free;
7115	alloc_seqid = server->nfs_client->cl_mvops->alloc_seqid;
7116	p->arg.lock_seqid = alloc_seqid(&lsp->ls_seqid, gfp_mask);
7117	if (IS_ERR(p->arg.lock_seqid))
7118		goto out_free_seqid;
7119	p->arg.lock_owner.clientid = server->nfs_client->cl_clientid;
7120	p->arg.lock_owner.id = lsp->ls_seqid.owner_id;
7121	p->arg.lock_owner.s_dev = server->s_dev;
7122	p->res.lock_seqid = p->arg.lock_seqid;
7123	p->lsp = lsp;
7124	p->server = server;
7125	p->ctx = get_nfs_open_context(ctx);
7126	locks_init_lock(&p->fl);
7127	locks_copy_lock(&p->fl, fl);
7128	return p;
7129out_free_seqid:
7130	nfs_free_seqid(p->arg.open_seqid);
7131out_free:
7132	kfree(p);
7133	return NULL;
7134}
7135
7136static void nfs4_lock_prepare(struct rpc_task *task, void *calldata)
7137{
7138	struct nfs4_lockdata *data = calldata;
7139	struct nfs4_state *state = data->lsp->ls_state;
7140
7141	if (nfs_wait_on_sequence(data->arg.lock_seqid, task) != 0)
7142		goto out_wait;
7143	/* Do we need to do an open_to_lock_owner? */
7144	if (!test_bit(NFS_LOCK_INITIALIZED, &data->lsp->ls_flags)) {
7145		if (nfs_wait_on_sequence(data->arg.open_seqid, task) != 0) {
7146			goto out_release_lock_seqid;
7147		}
7148		nfs4_stateid_copy(&data->arg.open_stateid,
7149				&state->open_stateid);
7150		data->arg.new_lock_owner = 1;
7151		data->res.open_seqid = data->arg.open_seqid;
7152	} else {
7153		data->arg.new_lock_owner = 0;
7154		nfs4_stateid_copy(&data->arg.lock_stateid,
7155				&data->lsp->ls_stateid);
7156	}
7157	if (!nfs4_valid_open_stateid(state)) {
7158		data->rpc_status = -EBADF;
7159		task->tk_action = NULL;
7160		goto out_release_open_seqid;
7161	}
7162	data->timestamp = jiffies;
7163	if (nfs4_setup_sequence(data->server->nfs_client,
7164				&data->arg.seq_args,
7165				&data->res.seq_res,
7166				task) == 0)
7167		return;
7168out_release_open_seqid:
7169	nfs_release_seqid(data->arg.open_seqid);
7170out_release_lock_seqid:
7171	nfs_release_seqid(data->arg.lock_seqid);
7172out_wait:
7173	nfs4_sequence_done(task, &data->res.seq_res);
7174	dprintk("%s: ret = %d\n", __func__, data->rpc_status);
7175}
7176
7177static void nfs4_lock_done(struct rpc_task *task, void *calldata)
7178{
7179	struct nfs4_lockdata *data = calldata;
7180	struct nfs4_lock_state *lsp = data->lsp;
7181
7182	if (!nfs4_sequence_done(task, &data->res.seq_res))
7183		return;
7184
7185	data->rpc_status = task->tk_status;
7186	switch (task->tk_status) {
7187	case 0:
7188		renew_lease(NFS_SERVER(d_inode(data->ctx->dentry)),
7189				data->timestamp);
7190		if (data->arg.new_lock && !data->cancelled) {
7191			data->fl.c.flc_flags &= ~(FL_SLEEP | FL_ACCESS);
7192			if (locks_lock_inode_wait(lsp->ls_state->inode, &data->fl) < 0)
7193				goto out_restart;
7194		}
7195		if (data->arg.new_lock_owner != 0) {
7196			nfs_confirm_seqid(&lsp->ls_seqid, 0);
7197			nfs4_stateid_copy(&lsp->ls_stateid, &data->res.stateid);
7198			set_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags);
7199		} else if (!nfs4_update_lock_stateid(lsp, &data->res.stateid))
7200			goto out_restart;
7201		break;
7202	case -NFS4ERR_OLD_STATEID:
7203		if (data->arg.new_lock_owner != 0 &&
7204			nfs4_refresh_open_old_stateid(&data->arg.open_stateid,
7205					lsp->ls_state))
7206			goto out_restart;
7207		if (nfs4_refresh_lock_old_stateid(&data->arg.lock_stateid, lsp))
7208			goto out_restart;
7209		fallthrough;
7210	case -NFS4ERR_BAD_STATEID:
7211	case -NFS4ERR_STALE_STATEID:
7212	case -NFS4ERR_EXPIRED:
7213		if (data->arg.new_lock_owner != 0) {
7214			if (!nfs4_stateid_match(&data->arg.open_stateid,
7215						&lsp->ls_state->open_stateid))
7216				goto out_restart;
7217		} else if (!nfs4_stateid_match(&data->arg.lock_stateid,
7218						&lsp->ls_stateid))
7219				goto out_restart;
7220	}
7221out_done:
7222	dprintk("%s: ret = %d!\n", __func__, data->rpc_status);
7223	return;
7224out_restart:
7225	if (!data->cancelled)
7226		rpc_restart_call_prepare(task);
7227	goto out_done;
7228}
7229
7230static void nfs4_lock_release(void *calldata)
7231{
7232	struct nfs4_lockdata *data = calldata;
7233
7234	nfs_free_seqid(data->arg.open_seqid);
7235	if (data->cancelled && data->rpc_status == 0) {
7236		struct rpc_task *task;
7237		task = nfs4_do_unlck(&data->fl, data->ctx, data->lsp,
7238				data->arg.lock_seqid);
7239		if (!IS_ERR(task))
7240			rpc_put_task_async(task);
7241		dprintk("%s: cancelling lock!\n", __func__);
7242	} else
7243		nfs_free_seqid(data->arg.lock_seqid);
7244	nfs4_put_lock_state(data->lsp);
7245	put_nfs_open_context(data->ctx);
7246	kfree(data);
7247}
7248
7249static const struct rpc_call_ops nfs4_lock_ops = {
7250	.rpc_call_prepare = nfs4_lock_prepare,
7251	.rpc_call_done = nfs4_lock_done,
7252	.rpc_release = nfs4_lock_release,
7253};
7254
7255static void nfs4_handle_setlk_error(struct nfs_server *server, struct nfs4_lock_state *lsp, int new_lock_owner, int error)
7256{
7257	switch (error) {
7258	case -NFS4ERR_ADMIN_REVOKED:
7259	case -NFS4ERR_EXPIRED:
7260	case -NFS4ERR_BAD_STATEID:
7261		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7262		if (new_lock_owner != 0 ||
7263		   test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) != 0)
7264			nfs4_schedule_stateid_recovery(server, lsp->ls_state);
7265		break;
7266	case -NFS4ERR_STALE_STATEID:
7267		lsp->ls_seqid.flags &= ~NFS_SEQID_CONFIRMED;
7268		nfs4_schedule_lease_recovery(server->nfs_client);
7269	}
7270}
7271
7272static int _nfs4_do_setlk(struct nfs4_state *state, int cmd, struct file_lock *fl, int recovery_type)
7273{
7274	struct nfs4_lockdata *data;
7275	struct rpc_task *task;
7276	struct rpc_message msg = {
7277		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LOCK],
7278		.rpc_cred = state->owner->so_cred,
7279	};
7280	struct rpc_task_setup task_setup_data = {
7281		.rpc_client = NFS_CLIENT(state->inode),
7282		.rpc_message = &msg,
7283		.callback_ops = &nfs4_lock_ops,
7284		.workqueue = nfsiod_workqueue,
7285		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF,
7286	};
7287	int ret;
7288
7289	if (nfs_server_capable(state->inode, NFS_CAP_MOVEABLE))
7290		task_setup_data.flags |= RPC_TASK_MOVEABLE;
7291
7292	data = nfs4_alloc_lockdata(fl,
7293				   nfs_file_open_context(fl->c.flc_file),
7294				   fl->fl_u.nfs4_fl.owner, GFP_KERNEL);
7295	if (data == NULL)
7296		return -ENOMEM;
7297	if (IS_SETLKW(cmd))
7298		data->arg.block = 1;
7299	nfs4_init_sequence(&data->arg.seq_args, &data->res.seq_res, 1,
7300				recovery_type > NFS_LOCK_NEW);
7301	msg.rpc_argp = &data->arg;
7302	msg.rpc_resp = &data->res;
7303	task_setup_data.callback_data = data;
7304	if (recovery_type > NFS_LOCK_NEW) {
7305		if (recovery_type == NFS_LOCK_RECLAIM)
7306			data->arg.reclaim = NFS_LOCK_RECLAIM;
7307	} else
7308		data->arg.new_lock = 1;
7309	task = rpc_run_task(&task_setup_data);
7310	if (IS_ERR(task))
7311		return PTR_ERR(task);
7312	ret = rpc_wait_for_completion_task(task);
7313	if (ret == 0) {
7314		ret = data->rpc_status;
7315		if (ret)
7316			nfs4_handle_setlk_error(data->server, data->lsp,
7317					data->arg.new_lock_owner, ret);
7318	} else
7319		data->cancelled = true;
7320	trace_nfs4_set_lock(fl, state, &data->res.stateid, cmd, ret);
7321	rpc_put_task(task);
7322	dprintk("%s: ret = %d\n", __func__, ret);
7323	return ret;
7324}
7325
7326static int nfs4_lock_reclaim(struct nfs4_state *state, struct file_lock *request)
7327{
7328	struct nfs_server *server = NFS_SERVER(state->inode);
7329	struct nfs4_exception exception = {
7330		.inode = state->inode,
7331	};
7332	int err;
7333
7334	do {
7335		/* Cache the lock if possible... */
7336		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7337			return 0;
7338		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_RECLAIM);
7339		if (err != -NFS4ERR_DELAY)
7340			break;
7341		nfs4_handle_exception(server, err, &exception);
7342	} while (exception.retry);
7343	return err;
7344}
7345
7346static int nfs4_lock_expired(struct nfs4_state *state, struct file_lock *request)
7347{
7348	struct nfs_server *server = NFS_SERVER(state->inode);
7349	struct nfs4_exception exception = {
7350		.inode = state->inode,
7351	};
7352	int err;
7353
7354	err = nfs4_set_lock_state(state, request);
7355	if (err != 0)
7356		return err;
7357	if (!recover_lost_locks) {
7358		set_bit(NFS_LOCK_LOST, &request->fl_u.nfs4_fl.owner->ls_flags);
7359		return 0;
7360	}
7361	do {
7362		if (test_bit(NFS_DELEGATED_STATE, &state->flags) != 0)
7363			return 0;
7364		err = _nfs4_do_setlk(state, F_SETLK, request, NFS_LOCK_EXPIRED);
7365		switch (err) {
7366		default:
7367			goto out;
7368		case -NFS4ERR_GRACE:
7369		case -NFS4ERR_DELAY:
7370			nfs4_handle_exception(server, err, &exception);
7371			err = 0;
7372		}
7373	} while (exception.retry);
7374out:
7375	return err;
7376}
7377
7378#if defined(CONFIG_NFS_V4_1)
7379static int nfs41_lock_expired(struct nfs4_state *state, struct file_lock *request)
7380{
7381	struct nfs4_lock_state *lsp;
7382	int status;
7383
7384	status = nfs4_set_lock_state(state, request);
7385	if (status != 0)
7386		return status;
7387	lsp = request->fl_u.nfs4_fl.owner;
7388	if (test_bit(NFS_LOCK_INITIALIZED, &lsp->ls_flags) ||
7389	    test_bit(NFS_LOCK_LOST, &lsp->ls_flags))
7390		return 0;
7391	return nfs4_lock_expired(state, request);
7392}
7393#endif
7394
7395static int _nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7396{
7397	struct nfs_inode *nfsi = NFS_I(state->inode);
7398	struct nfs4_state_owner *sp = state->owner;
7399	unsigned char flags = request->c.flc_flags;
7400	int status;
7401
7402	request->c.flc_flags |= FL_ACCESS;
7403	status = locks_lock_inode_wait(state->inode, request);
7404	if (status < 0)
7405		goto out;
7406	mutex_lock(&sp->so_delegreturn_mutex);
7407	down_read(&nfsi->rwsem);
7408	if (test_bit(NFS_DELEGATED_STATE, &state->flags)) {
7409		/* Yes: cache locks! */
7410		/* ...but avoid races with delegation recall... */
7411		request->c.flc_flags = flags & ~FL_SLEEP;
7412		status = locks_lock_inode_wait(state->inode, request);
7413		up_read(&nfsi->rwsem);
7414		mutex_unlock(&sp->so_delegreturn_mutex);
7415		goto out;
7416	}
7417	up_read(&nfsi->rwsem);
7418	mutex_unlock(&sp->so_delegreturn_mutex);
7419	status = _nfs4_do_setlk(state, cmd, request, NFS_LOCK_NEW);
7420out:
7421	request->c.flc_flags = flags;
7422	return status;
7423}
7424
7425static int nfs4_proc_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7426{
7427	struct nfs4_exception exception = {
7428		.state = state,
7429		.inode = state->inode,
7430		.interruptible = true,
7431	};
7432	int err;
7433
7434	do {
7435		err = _nfs4_proc_setlk(state, cmd, request);
7436		if (err == -NFS4ERR_DENIED)
7437			err = -EAGAIN;
7438		err = nfs4_handle_exception(NFS_SERVER(state->inode),
7439				err, &exception);
7440	} while (exception.retry);
7441	return err;
7442}
7443
7444#define NFS4_LOCK_MINTIMEOUT (1 * HZ)
7445#define NFS4_LOCK_MAXTIMEOUT (30 * HZ)
7446
7447static int
7448nfs4_retry_setlk_simple(struct nfs4_state *state, int cmd,
7449			struct file_lock *request)
7450{
7451	int		status = -ERESTARTSYS;
7452	unsigned long	timeout = NFS4_LOCK_MINTIMEOUT;
7453
7454	while(!signalled()) {
7455		status = nfs4_proc_setlk(state, cmd, request);
7456		if ((status != -EAGAIN) || IS_SETLK(cmd))
7457			break;
7458		__set_current_state(TASK_INTERRUPTIBLE|TASK_FREEZABLE);
7459		schedule_timeout(timeout);
7460		timeout *= 2;
7461		timeout = min_t(unsigned long, NFS4_LOCK_MAXTIMEOUT, timeout);
7462		status = -ERESTARTSYS;
7463	}
7464	return status;
7465}
7466
7467#ifdef CONFIG_NFS_V4_1
7468struct nfs4_lock_waiter {
7469	struct inode		*inode;
7470	struct nfs_lowner	owner;
7471	wait_queue_entry_t	wait;
7472};
7473
7474static int
7475nfs4_wake_lock_waiter(wait_queue_entry_t *wait, unsigned int mode, int flags, void *key)
7476{
7477	struct nfs4_lock_waiter	*waiter	=
7478		container_of(wait, struct nfs4_lock_waiter, wait);
7479
7480	/* NULL key means to wake up everyone */
7481	if (key) {
7482		struct cb_notify_lock_args	*cbnl = key;
7483		struct nfs_lowner		*lowner = &cbnl->cbnl_owner,
7484						*wowner = &waiter->owner;
7485
7486		/* Only wake if the callback was for the same owner. */
7487		if (lowner->id != wowner->id || lowner->s_dev != wowner->s_dev)
7488			return 0;
7489
7490		/* Make sure it's for the right inode */
7491		if (nfs_compare_fh(NFS_FH(waiter->inode), &cbnl->cbnl_fh))
7492			return 0;
7493	}
7494
7495	return woken_wake_function(wait, mode, flags, key);
7496}
7497
7498static int
7499nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7500{
7501	struct nfs4_lock_state *lsp = request->fl_u.nfs4_fl.owner;
7502	struct nfs_server *server = NFS_SERVER(state->inode);
7503	struct nfs_client *clp = server->nfs_client;
7504	wait_queue_head_t *q = &clp->cl_lock_waitq;
7505	struct nfs4_lock_waiter waiter = {
7506		.inode = state->inode,
7507		.owner = { .clientid = clp->cl_clientid,
7508			   .id = lsp->ls_seqid.owner_id,
7509			   .s_dev = server->s_dev },
7510	};
7511	int status;
7512
7513	/* Don't bother with waitqueue if we don't expect a callback */
7514	if (!test_bit(NFS_STATE_MAY_NOTIFY_LOCK, &state->flags))
7515		return nfs4_retry_setlk_simple(state, cmd, request);
7516
7517	init_wait(&waiter.wait);
7518	waiter.wait.func = nfs4_wake_lock_waiter;
7519	add_wait_queue(q, &waiter.wait);
7520
7521	do {
7522		status = nfs4_proc_setlk(state, cmd, request);
7523		if (status != -EAGAIN || IS_SETLK(cmd))
7524			break;
7525
7526		status = -ERESTARTSYS;
7527		wait_woken(&waiter.wait, TASK_INTERRUPTIBLE|TASK_FREEZABLE,
7528			   NFS4_LOCK_MAXTIMEOUT);
7529	} while (!signalled());
7530
7531	remove_wait_queue(q, &waiter.wait);
7532
7533	return status;
7534}
7535#else /* !CONFIG_NFS_V4_1 */
7536static inline int
7537nfs4_retry_setlk(struct nfs4_state *state, int cmd, struct file_lock *request)
7538{
7539	return nfs4_retry_setlk_simple(state, cmd, request);
7540}
7541#endif
7542
7543static int
7544nfs4_proc_lock(struct file *filp, int cmd, struct file_lock *request)
7545{
7546	struct nfs_open_context *ctx;
7547	struct nfs4_state *state;
7548	int status;
7549
7550	/* verify open state */
7551	ctx = nfs_file_open_context(filp);
7552	state = ctx->state;
7553
7554	if (IS_GETLK(cmd)) {
7555		if (state != NULL)
7556			return nfs4_proc_getlk(state, F_GETLK, request);
7557		return 0;
7558	}
7559
7560	if (!(IS_SETLK(cmd) || IS_SETLKW(cmd)))
7561		return -EINVAL;
7562
7563	if (lock_is_unlock(request)) {
7564		if (state != NULL)
7565			return nfs4_proc_unlck(state, cmd, request);
7566		return 0;
7567	}
7568
7569	if (state == NULL)
7570		return -ENOLCK;
7571
7572	if ((request->c.flc_flags & FL_POSIX) &&
7573	    !test_bit(NFS_STATE_POSIX_LOCKS, &state->flags))
7574		return -ENOLCK;
7575
7576	/*
7577	 * Don't rely on the VFS having checked the file open mode,
7578	 * since it won't do this for flock() locks.
7579	 */
7580	switch (request->c.flc_type) {
7581	case F_RDLCK:
7582		if (!(filp->f_mode & FMODE_READ))
7583			return -EBADF;
7584		break;
7585	case F_WRLCK:
7586		if (!(filp->f_mode & FMODE_WRITE))
7587			return -EBADF;
7588	}
7589
7590	status = nfs4_set_lock_state(state, request);
7591	if (status != 0)
7592		return status;
7593
7594	return nfs4_retry_setlk(state, cmd, request);
7595}
7596
7597static int nfs4_delete_lease(struct file *file, void **priv)
7598{
7599	return generic_setlease(file, F_UNLCK, NULL, priv);
7600}
7601
7602static int nfs4_add_lease(struct file *file, int arg, struct file_lease **lease,
7603			  void **priv)
7604{
7605	struct inode *inode = file_inode(file);
7606	fmode_t type = arg == F_RDLCK ? FMODE_READ : FMODE_WRITE;
7607	int ret;
7608
7609	/* No delegation, no lease */
7610	if (!nfs4_have_delegation(inode, type))
7611		return -EAGAIN;
7612	ret = generic_setlease(file, arg, lease, priv);
7613	if (ret || nfs4_have_delegation(inode, type))
7614		return ret;
7615	/* We raced with a delegation return */
7616	nfs4_delete_lease(file, priv);
7617	return -EAGAIN;
7618}
7619
7620int nfs4_proc_setlease(struct file *file, int arg, struct file_lease **lease,
7621		       void **priv)
7622{
7623	switch (arg) {
7624	case F_RDLCK:
7625	case F_WRLCK:
7626		return nfs4_add_lease(file, arg, lease, priv);
7627	case F_UNLCK:
7628		return nfs4_delete_lease(file, priv);
7629	default:
7630		return -EINVAL;
7631	}
7632}
7633
7634int nfs4_lock_delegation_recall(struct file_lock *fl, struct nfs4_state *state, const nfs4_stateid *stateid)
7635{
7636	struct nfs_server *server = NFS_SERVER(state->inode);
7637	int err;
7638
7639	err = nfs4_set_lock_state(state, fl);
7640	if (err != 0)
7641		return err;
7642	do {
7643		err = _nfs4_do_setlk(state, F_SETLK, fl, NFS_LOCK_NEW);
7644		if (err != -NFS4ERR_DELAY)
7645			break;
7646		ssleep(1);
7647	} while (err == -NFS4ERR_DELAY);
7648	return nfs4_handle_delegation_recall_error(server, state, stateid, fl, err);
7649}
7650
7651struct nfs_release_lockowner_data {
7652	struct nfs4_lock_state *lsp;
7653	struct nfs_server *server;
7654	struct nfs_release_lockowner_args args;
7655	struct nfs_release_lockowner_res res;
7656	unsigned long timestamp;
7657};
7658
7659static void nfs4_release_lockowner_prepare(struct rpc_task *task, void *calldata)
7660{
7661	struct nfs_release_lockowner_data *data = calldata;
7662	struct nfs_server *server = data->server;
7663	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
7664			   &data->res.seq_res, task);
7665	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7666	data->timestamp = jiffies;
7667}
7668
7669static void nfs4_release_lockowner_done(struct rpc_task *task, void *calldata)
7670{
7671	struct nfs_release_lockowner_data *data = calldata;
7672	struct nfs_server *server = data->server;
7673
7674	nfs40_sequence_done(task, &data->res.seq_res);
7675
7676	switch (task->tk_status) {
7677	case 0:
7678		renew_lease(server, data->timestamp);
7679		break;
7680	case -NFS4ERR_STALE_CLIENTID:
7681	case -NFS4ERR_EXPIRED:
7682		nfs4_schedule_lease_recovery(server->nfs_client);
7683		break;
7684	case -NFS4ERR_LEASE_MOVED:
7685	case -NFS4ERR_DELAY:
7686		if (nfs4_async_handle_error(task, server,
7687					    NULL, NULL) == -EAGAIN)
7688			rpc_restart_call_prepare(task);
7689	}
7690}
7691
7692static void nfs4_release_lockowner_release(void *calldata)
7693{
7694	struct nfs_release_lockowner_data *data = calldata;
7695	nfs4_free_lock_state(data->server, data->lsp);
7696	kfree(calldata);
7697}
7698
7699static const struct rpc_call_ops nfs4_release_lockowner_ops = {
7700	.rpc_call_prepare = nfs4_release_lockowner_prepare,
7701	.rpc_call_done = nfs4_release_lockowner_done,
7702	.rpc_release = nfs4_release_lockowner_release,
7703};
7704
7705static void
7706nfs4_release_lockowner(struct nfs_server *server, struct nfs4_lock_state *lsp)
7707{
7708	struct nfs_release_lockowner_data *data;
7709	struct rpc_message msg = {
7710		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RELEASE_LOCKOWNER],
7711	};
7712
7713	if (server->nfs_client->cl_mvops->minor_version != 0)
7714		return;
7715
7716	data = kmalloc(sizeof(*data), GFP_KERNEL);
7717	if (!data)
7718		return;
7719	data->lsp = lsp;
7720	data->server = server;
7721	data->args.lock_owner.clientid = server->nfs_client->cl_clientid;
7722	data->args.lock_owner.id = lsp->ls_seqid.owner_id;
7723	data->args.lock_owner.s_dev = server->s_dev;
7724
7725	msg.rpc_argp = &data->args;
7726	msg.rpc_resp = &data->res;
7727	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
7728	rpc_call_async(server->client, &msg, 0, &nfs4_release_lockowner_ops, data);
7729}
7730
7731#define XATTR_NAME_NFSV4_ACL "system.nfs4_acl"
7732
7733static int nfs4_xattr_set_nfs4_acl(const struct xattr_handler *handler,
7734				   struct mnt_idmap *idmap,
7735				   struct dentry *unused, struct inode *inode,
7736				   const char *key, const void *buf,
7737				   size_t buflen, int flags)
7738{
7739	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_ACL);
7740}
7741
7742static int nfs4_xattr_get_nfs4_acl(const struct xattr_handler *handler,
7743				   struct dentry *unused, struct inode *inode,
7744				   const char *key, void *buf, size_t buflen)
7745{
7746	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_ACL);
7747}
7748
7749static bool nfs4_xattr_list_nfs4_acl(struct dentry *dentry)
7750{
7751	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_ACL);
7752}
7753
7754#if defined(CONFIG_NFS_V4_1)
7755#define XATTR_NAME_NFSV4_DACL "system.nfs4_dacl"
7756
7757static int nfs4_xattr_set_nfs4_dacl(const struct xattr_handler *handler,
7758				    struct mnt_idmap *idmap,
7759				    struct dentry *unused, struct inode *inode,
7760				    const char *key, const void *buf,
7761				    size_t buflen, int flags)
7762{
7763	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_DACL);
7764}
7765
7766static int nfs4_xattr_get_nfs4_dacl(const struct xattr_handler *handler,
7767				    struct dentry *unused, struct inode *inode,
7768				    const char *key, void *buf, size_t buflen)
7769{
7770	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_DACL);
7771}
7772
7773static bool nfs4_xattr_list_nfs4_dacl(struct dentry *dentry)
7774{
7775	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_DACL);
7776}
7777
7778#define XATTR_NAME_NFSV4_SACL "system.nfs4_sacl"
7779
7780static int nfs4_xattr_set_nfs4_sacl(const struct xattr_handler *handler,
7781				    struct mnt_idmap *idmap,
7782				    struct dentry *unused, struct inode *inode,
7783				    const char *key, const void *buf,
7784				    size_t buflen, int flags)
7785{
7786	return nfs4_proc_set_acl(inode, buf, buflen, NFS4ACL_SACL);
7787}
7788
7789static int nfs4_xattr_get_nfs4_sacl(const struct xattr_handler *handler,
7790				    struct dentry *unused, struct inode *inode,
7791				    const char *key, void *buf, size_t buflen)
7792{
7793	return nfs4_proc_get_acl(inode, buf, buflen, NFS4ACL_SACL);
7794}
7795
7796static bool nfs4_xattr_list_nfs4_sacl(struct dentry *dentry)
7797{
7798	return nfs4_server_supports_acls(NFS_SB(dentry->d_sb), NFS4ACL_SACL);
7799}
7800
7801#endif
7802
7803#ifdef CONFIG_NFS_V4_SECURITY_LABEL
7804
7805static int nfs4_xattr_set_nfs4_label(const struct xattr_handler *handler,
7806				     struct mnt_idmap *idmap,
7807				     struct dentry *unused, struct inode *inode,
7808				     const char *key, const void *buf,
7809				     size_t buflen, int flags)
7810{
7811	if (security_ismaclabel(key))
7812		return nfs4_set_security_label(inode, buf, buflen);
7813
7814	return -EOPNOTSUPP;
7815}
7816
7817static int nfs4_xattr_get_nfs4_label(const struct xattr_handler *handler,
7818				     struct dentry *unused, struct inode *inode,
7819				     const char *key, void *buf, size_t buflen)
7820{
7821	if (security_ismaclabel(key))
7822		return nfs4_get_security_label(inode, buf, buflen);
7823	return -EOPNOTSUPP;
7824}
7825
7826static ssize_t
7827nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7828{
7829	int len = 0;
7830
7831	if (nfs_server_capable(inode, NFS_CAP_SECURITY_LABEL)) {
7832		len = security_inode_listsecurity(inode, list, list_len);
7833		if (len >= 0 && list_len && len > list_len)
7834			return -ERANGE;
7835	}
7836	return len;
7837}
7838
7839static const struct xattr_handler nfs4_xattr_nfs4_label_handler = {
7840	.prefix = XATTR_SECURITY_PREFIX,
7841	.get	= nfs4_xattr_get_nfs4_label,
7842	.set	= nfs4_xattr_set_nfs4_label,
7843};
7844
7845#else
7846
7847static ssize_t
7848nfs4_listxattr_nfs4_label(struct inode *inode, char *list, size_t list_len)
7849{
7850	return 0;
7851}
7852
7853#endif
7854
7855#ifdef CONFIG_NFS_V4_2
7856static int nfs4_xattr_set_nfs4_user(const struct xattr_handler *handler,
7857				    struct mnt_idmap *idmap,
7858				    struct dentry *unused, struct inode *inode,
7859				    const char *key, const void *buf,
7860				    size_t buflen, int flags)
7861{
7862	u32 mask;
7863	int ret;
7864
7865	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7866		return -EOPNOTSUPP;
7867
7868	/*
7869	 * There is no mapping from the MAY_* flags to the NFS_ACCESS_XA*
7870	 * flags right now. Handling of xattr operations use the normal
7871	 * file read/write permissions.
7872	 *
7873	 * Just in case the server has other ideas (which RFC 8276 allows),
7874	 * do a cached access check for the XA* flags to possibly avoid
7875	 * doing an RPC and getting EACCES back.
7876	 */
7877	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7878		if (!(mask & NFS_ACCESS_XAWRITE))
7879			return -EACCES;
7880	}
7881
7882	if (buf == NULL) {
7883		ret = nfs42_proc_removexattr(inode, key);
7884		if (!ret)
7885			nfs4_xattr_cache_remove(inode, key);
7886	} else {
7887		ret = nfs42_proc_setxattr(inode, key, buf, buflen, flags);
7888		if (!ret)
7889			nfs4_xattr_cache_add(inode, key, buf, NULL, buflen);
7890	}
7891
7892	return ret;
7893}
7894
7895static int nfs4_xattr_get_nfs4_user(const struct xattr_handler *handler,
7896				    struct dentry *unused, struct inode *inode,
7897				    const char *key, void *buf, size_t buflen)
7898{
7899	u32 mask;
7900	ssize_t ret;
7901
7902	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7903		return -EOPNOTSUPP;
7904
7905	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7906		if (!(mask & NFS_ACCESS_XAREAD))
7907			return -EACCES;
7908	}
7909
7910	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7911	if (ret)
7912		return ret;
7913
7914	ret = nfs4_xattr_cache_get(inode, key, buf, buflen);
7915	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7916		return ret;
7917
7918	ret = nfs42_proc_getxattr(inode, key, buf, buflen);
7919
7920	return ret;
7921}
7922
7923static ssize_t
7924nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7925{
7926	u64 cookie;
7927	bool eof;
7928	ssize_t ret, size;
7929	char *buf;
7930	size_t buflen;
7931	u32 mask;
7932
7933	if (!nfs_server_capable(inode, NFS_CAP_XATTR))
7934		return 0;
7935
7936	if (!nfs_access_get_cached(inode, current_cred(), &mask, true)) {
7937		if (!(mask & NFS_ACCESS_XALIST))
7938			return 0;
7939	}
7940
7941	ret = nfs_revalidate_inode(inode, NFS_INO_INVALID_CHANGE);
7942	if (ret)
7943		return ret;
7944
7945	ret = nfs4_xattr_cache_list(inode, list, list_len);
7946	if (ret >= 0 || (ret < 0 && ret != -ENOENT))
7947		return ret;
7948
7949	cookie = 0;
7950	eof = false;
7951	buflen = list_len ? list_len : XATTR_LIST_MAX;
7952	buf = list_len ? list : NULL;
7953	size = 0;
7954
7955	while (!eof) {
7956		ret = nfs42_proc_listxattrs(inode, buf, buflen,
7957		    &cookie, &eof);
7958		if (ret < 0)
7959			return ret;
7960
7961		if (list_len) {
7962			buf += ret;
7963			buflen -= ret;
7964		}
7965		size += ret;
7966	}
7967
7968	if (list_len)
7969		nfs4_xattr_cache_set_list(inode, list, size);
7970
7971	return size;
7972}
7973
7974#else
7975
7976static ssize_t
7977nfs4_listxattr_nfs4_user(struct inode *inode, char *list, size_t list_len)
7978{
7979	return 0;
7980}
7981#endif /* CONFIG_NFS_V4_2 */
7982
7983/*
7984 * nfs_fhget will use either the mounted_on_fileid or the fileid
7985 */
7986static void nfs_fixup_referral_attributes(struct nfs_fattr *fattr)
7987{
7988	if (!(((fattr->valid & NFS_ATTR_FATTR_MOUNTED_ON_FILEID) ||
7989	       (fattr->valid & NFS_ATTR_FATTR_FILEID)) &&
7990	      (fattr->valid & NFS_ATTR_FATTR_FSID) &&
7991	      (fattr->valid & NFS_ATTR_FATTR_V4_LOCATIONS)))
7992		return;
7993
7994	fattr->valid |= NFS_ATTR_FATTR_TYPE | NFS_ATTR_FATTR_MODE |
7995		NFS_ATTR_FATTR_NLINK | NFS_ATTR_FATTR_V4_REFERRAL;
7996	fattr->mode = S_IFDIR | S_IRUGO | S_IXUGO;
7997	fattr->nlink = 2;
7998}
7999
8000static int _nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8001				   const struct qstr *name,
8002				   struct nfs4_fs_locations *fs_locations,
8003				   struct page *page)
8004{
8005	struct nfs_server *server = NFS_SERVER(dir);
8006	u32 bitmask[3];
8007	struct nfs4_fs_locations_arg args = {
8008		.dir_fh = NFS_FH(dir),
8009		.name = name,
8010		.page = page,
8011		.bitmask = bitmask,
8012	};
8013	struct nfs4_fs_locations_res res = {
8014		.fs_locations = fs_locations,
8015	};
8016	struct rpc_message msg = {
8017		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8018		.rpc_argp = &args,
8019		.rpc_resp = &res,
8020	};
8021	int status;
8022
8023	dprintk("%s: start\n", __func__);
8024
8025	bitmask[0] = nfs4_fattr_bitmap[0] | FATTR4_WORD0_FS_LOCATIONS;
8026	bitmask[1] = nfs4_fattr_bitmap[1];
8027
8028	/* Ask for the fileid of the absent filesystem if mounted_on_fileid
8029	 * is not supported */
8030	if (NFS_SERVER(dir)->attr_bitmask[1] & FATTR4_WORD1_MOUNTED_ON_FILEID)
8031		bitmask[0] &= ~FATTR4_WORD0_FILEID;
8032	else
8033		bitmask[1] &= ~FATTR4_WORD1_MOUNTED_ON_FILEID;
8034
8035	nfs_fattr_init(fs_locations->fattr);
8036	fs_locations->server = server;
8037	fs_locations->nlocations = 0;
8038	status = nfs4_call_sync(client, server, &msg, &args.seq_args, &res.seq_res, 0);
8039	dprintk("%s: returned status = %d\n", __func__, status);
8040	return status;
8041}
8042
8043int nfs4_proc_fs_locations(struct rpc_clnt *client, struct inode *dir,
8044			   const struct qstr *name,
8045			   struct nfs4_fs_locations *fs_locations,
8046			   struct page *page)
8047{
8048	struct nfs4_exception exception = {
8049		.interruptible = true,
8050	};
8051	int err;
8052	do {
8053		err = _nfs4_proc_fs_locations(client, dir, name,
8054				fs_locations, page);
8055		trace_nfs4_get_fs_locations(dir, name, err);
8056		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8057				&exception);
8058	} while (exception.retry);
8059	return err;
8060}
8061
8062/*
8063 * This operation also signals the server that this client is
8064 * performing migration recovery.  The server can stop returning
8065 * NFS4ERR_LEASE_MOVED to this client.  A RENEW operation is
8066 * appended to this compound to identify the client ID which is
8067 * performing recovery.
8068 */
8069static int _nfs40_proc_get_locations(struct nfs_server *server,
8070				     struct nfs_fh *fhandle,
8071				     struct nfs4_fs_locations *locations,
8072				     struct page *page, const struct cred *cred)
8073{
8074	struct rpc_clnt *clnt = server->client;
8075	u32 bitmask[2] = {
8076		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8077	};
8078	struct nfs4_fs_locations_arg args = {
8079		.clientid	= server->nfs_client->cl_clientid,
8080		.fh		= fhandle,
8081		.page		= page,
8082		.bitmask	= bitmask,
8083		.migration	= 1,		/* skip LOOKUP */
8084		.renew		= 1,		/* append RENEW */
8085	};
8086	struct nfs4_fs_locations_res res = {
8087		.fs_locations	= locations,
8088		.migration	= 1,
8089		.renew		= 1,
8090	};
8091	struct rpc_message msg = {
8092		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8093		.rpc_argp	= &args,
8094		.rpc_resp	= &res,
8095		.rpc_cred	= cred,
8096	};
8097	unsigned long now = jiffies;
8098	int status;
8099
8100	nfs_fattr_init(locations->fattr);
8101	locations->server = server;
8102	locations->nlocations = 0;
8103
8104	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8105	status = nfs4_call_sync_sequence(clnt, server, &msg,
8106					&args.seq_args, &res.seq_res);
8107	if (status)
8108		return status;
8109
8110	renew_lease(server, now);
8111	return 0;
8112}
8113
8114#ifdef CONFIG_NFS_V4_1
8115
8116/*
8117 * This operation also signals the server that this client is
8118 * performing migration recovery.  The server can stop asserting
8119 * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID
8120 * performing this operation is identified in the SEQUENCE
8121 * operation in this compound.
8122 *
8123 * When the client supports GETATTR(fs_locations_info), it can
8124 * be plumbed in here.
8125 */
8126static int _nfs41_proc_get_locations(struct nfs_server *server,
8127				     struct nfs_fh *fhandle,
8128				     struct nfs4_fs_locations *locations,
8129				     struct page *page, const struct cred *cred)
8130{
8131	struct rpc_clnt *clnt = server->client;
8132	u32 bitmask[2] = {
8133		[0] = FATTR4_WORD0_FSID | FATTR4_WORD0_FS_LOCATIONS,
8134	};
8135	struct nfs4_fs_locations_arg args = {
8136		.fh		= fhandle,
8137		.page		= page,
8138		.bitmask	= bitmask,
8139		.migration	= 1,		/* skip LOOKUP */
8140	};
8141	struct nfs4_fs_locations_res res = {
8142		.fs_locations	= locations,
8143		.migration	= 1,
8144	};
8145	struct rpc_message msg = {
8146		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FS_LOCATIONS],
8147		.rpc_argp	= &args,
8148		.rpc_resp	= &res,
8149		.rpc_cred	= cred,
8150	};
8151	struct nfs4_call_sync_data data = {
8152		.seq_server = server,
8153		.seq_args = &args.seq_args,
8154		.seq_res = &res.seq_res,
8155	};
8156	struct rpc_task_setup task_setup_data = {
8157		.rpc_client = clnt,
8158		.rpc_message = &msg,
8159		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
8160		.callback_data = &data,
8161		.flags = RPC_TASK_NO_ROUND_ROBIN,
8162	};
8163	int status;
8164
8165	nfs_fattr_init(locations->fattr);
8166	locations->server = server;
8167	locations->nlocations = 0;
8168
8169	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8170	status = nfs4_call_sync_custom(&task_setup_data);
8171	if (status == NFS4_OK &&
8172	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8173		status = -NFS4ERR_LEASE_MOVED;
8174	return status;
8175}
8176
8177#endif	/* CONFIG_NFS_V4_1 */
8178
8179/**
8180 * nfs4_proc_get_locations - discover locations for a migrated FSID
8181 * @server: pointer to nfs_server to process
8182 * @fhandle: pointer to the kernel NFS client file handle
8183 * @locations: result of query
8184 * @page: buffer
8185 * @cred: credential to use for this operation
8186 *
8187 * Returns NFS4_OK on success, a negative NFS4ERR status code if the
8188 * operation failed, or a negative errno if a local error occurred.
8189 *
8190 * On success, "locations" is filled in, but if the server has
8191 * no locations information, NFS_ATTR_FATTR_V4_LOCATIONS is not
8192 * asserted.
8193 *
8194 * -NFS4ERR_LEASE_MOVED is returned if the server still has leases
8195 * from this client that require migration recovery.
8196 */
8197int nfs4_proc_get_locations(struct nfs_server *server,
8198			    struct nfs_fh *fhandle,
8199			    struct nfs4_fs_locations *locations,
8200			    struct page *page, const struct cred *cred)
8201{
8202	struct nfs_client *clp = server->nfs_client;
8203	const struct nfs4_mig_recovery_ops *ops =
8204					clp->cl_mvops->mig_recovery_ops;
8205	struct nfs4_exception exception = {
8206		.interruptible = true,
8207	};
8208	int status;
8209
8210	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8211		(unsigned long long)server->fsid.major,
8212		(unsigned long long)server->fsid.minor,
8213		clp->cl_hostname);
8214	nfs_display_fhandle(fhandle, __func__);
8215
8216	do {
8217		status = ops->get_locations(server, fhandle, locations, page,
8218					    cred);
8219		if (status != -NFS4ERR_DELAY)
8220			break;
8221		nfs4_handle_exception(server, status, &exception);
8222	} while (exception.retry);
8223	return status;
8224}
8225
8226/*
8227 * This operation also signals the server that this client is
8228 * performing "lease moved" recovery.  The server can stop
8229 * returning NFS4ERR_LEASE_MOVED to this client.  A RENEW operation
8230 * is appended to this compound to identify the client ID which is
8231 * performing recovery.
8232 */
8233static int _nfs40_proc_fsid_present(struct inode *inode, const struct cred *cred)
8234{
8235	struct nfs_server *server = NFS_SERVER(inode);
8236	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
8237	struct rpc_clnt *clnt = server->client;
8238	struct nfs4_fsid_present_arg args = {
8239		.fh		= NFS_FH(inode),
8240		.clientid	= clp->cl_clientid,
8241		.renew		= 1,		/* append RENEW */
8242	};
8243	struct nfs4_fsid_present_res res = {
8244		.renew		= 1,
8245	};
8246	struct rpc_message msg = {
8247		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8248		.rpc_argp	= &args,
8249		.rpc_resp	= &res,
8250		.rpc_cred	= cred,
8251	};
8252	unsigned long now = jiffies;
8253	int status;
8254
8255	res.fh = nfs_alloc_fhandle();
8256	if (res.fh == NULL)
8257		return -ENOMEM;
8258
8259	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8260	status = nfs4_call_sync_sequence(clnt, server, &msg,
8261						&args.seq_args, &res.seq_res);
8262	nfs_free_fhandle(res.fh);
8263	if (status)
8264		return status;
8265
8266	do_renew_lease(clp, now);
8267	return 0;
8268}
8269
8270#ifdef CONFIG_NFS_V4_1
8271
8272/*
8273 * This operation also signals the server that this client is
8274 * performing "lease moved" recovery.  The server can stop asserting
8275 * SEQ4_STATUS_LEASE_MOVED for this client.  The client ID performing
8276 * this operation is identified in the SEQUENCE operation in this
8277 * compound.
8278 */
8279static int _nfs41_proc_fsid_present(struct inode *inode, const struct cred *cred)
8280{
8281	struct nfs_server *server = NFS_SERVER(inode);
8282	struct rpc_clnt *clnt = server->client;
8283	struct nfs4_fsid_present_arg args = {
8284		.fh		= NFS_FH(inode),
8285	};
8286	struct nfs4_fsid_present_res res = {
8287	};
8288	struct rpc_message msg = {
8289		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_FSID_PRESENT],
8290		.rpc_argp	= &args,
8291		.rpc_resp	= &res,
8292		.rpc_cred	= cred,
8293	};
8294	int status;
8295
8296	res.fh = nfs_alloc_fhandle();
8297	if (res.fh == NULL)
8298		return -ENOMEM;
8299
8300	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
8301	status = nfs4_call_sync_sequence(clnt, server, &msg,
8302						&args.seq_args, &res.seq_res);
8303	nfs_free_fhandle(res.fh);
8304	if (status == NFS4_OK &&
8305	    res.seq_res.sr_status_flags & SEQ4_STATUS_LEASE_MOVED)
8306		status = -NFS4ERR_LEASE_MOVED;
8307	return status;
8308}
8309
8310#endif	/* CONFIG_NFS_V4_1 */
8311
8312/**
8313 * nfs4_proc_fsid_present - Is this FSID present or absent on server?
8314 * @inode: inode on FSID to check
8315 * @cred: credential to use for this operation
8316 *
8317 * Server indicates whether the FSID is present, moved, or not
8318 * recognized.  This operation is necessary to clear a LEASE_MOVED
8319 * condition for this client ID.
8320 *
8321 * Returns NFS4_OK if the FSID is present on this server,
8322 * -NFS4ERR_MOVED if the FSID is no longer present, a negative
8323 *  NFS4ERR code if some error occurred on the server, or a
8324 *  negative errno if a local failure occurred.
8325 */
8326int nfs4_proc_fsid_present(struct inode *inode, const struct cred *cred)
8327{
8328	struct nfs_server *server = NFS_SERVER(inode);
8329	struct nfs_client *clp = server->nfs_client;
8330	const struct nfs4_mig_recovery_ops *ops =
8331					clp->cl_mvops->mig_recovery_ops;
8332	struct nfs4_exception exception = {
8333		.interruptible = true,
8334	};
8335	int status;
8336
8337	dprintk("%s: FSID %llx:%llx on \"%s\"\n", __func__,
8338		(unsigned long long)server->fsid.major,
8339		(unsigned long long)server->fsid.minor,
8340		clp->cl_hostname);
8341	nfs_display_fhandle(NFS_FH(inode), __func__);
8342
8343	do {
8344		status = ops->fsid_present(inode, cred);
8345		if (status != -NFS4ERR_DELAY)
8346			break;
8347		nfs4_handle_exception(server, status, &exception);
8348	} while (exception.retry);
8349	return status;
8350}
8351
8352/*
8353 * If 'use_integrity' is true and the state managment nfs_client
8354 * cl_rpcclient is using krb5i/p, use the integrity protected cl_rpcclient
8355 * and the machine credential as per RFC3530bis and RFC5661 Security
8356 * Considerations sections. Otherwise, just use the user cred with the
8357 * filesystem's rpc_client.
8358 */
8359static int _nfs4_proc_secinfo(struct inode *dir, const struct qstr *name, struct nfs4_secinfo_flavors *flavors, bool use_integrity)
8360{
8361	int status;
8362	struct rpc_clnt *clnt = NFS_SERVER(dir)->client;
8363	struct nfs_client *clp = NFS_SERVER(dir)->nfs_client;
8364	struct nfs4_secinfo_arg args = {
8365		.dir_fh = NFS_FH(dir),
8366		.name   = name,
8367	};
8368	struct nfs4_secinfo_res res = {
8369		.flavors     = flavors,
8370	};
8371	struct rpc_message msg = {
8372		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO],
8373		.rpc_argp = &args,
8374		.rpc_resp = &res,
8375	};
8376	struct nfs4_call_sync_data data = {
8377		.seq_server = NFS_SERVER(dir),
8378		.seq_args = &args.seq_args,
8379		.seq_res = &res.seq_res,
8380	};
8381	struct rpc_task_setup task_setup = {
8382		.rpc_client = clnt,
8383		.rpc_message = &msg,
8384		.callback_ops = clp->cl_mvops->call_sync_ops,
8385		.callback_data = &data,
8386		.flags = RPC_TASK_NO_ROUND_ROBIN,
8387	};
8388	const struct cred *cred = NULL;
8389
8390	if (use_integrity) {
8391		clnt = clp->cl_rpcclient;
8392		task_setup.rpc_client = clnt;
8393
8394		cred = nfs4_get_clid_cred(clp);
8395		msg.rpc_cred = cred;
8396	}
8397
8398	dprintk("NFS call  secinfo %s\n", name->name);
8399
8400	nfs4_state_protect(clp, NFS_SP4_MACH_CRED_SECINFO, &clnt, &msg);
8401	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
8402	status = nfs4_call_sync_custom(&task_setup);
8403
8404	dprintk("NFS reply  secinfo: %d\n", status);
8405
8406	put_cred(cred);
8407	return status;
8408}
8409
8410int nfs4_proc_secinfo(struct inode *dir, const struct qstr *name,
8411		      struct nfs4_secinfo_flavors *flavors)
8412{
8413	struct nfs4_exception exception = {
8414		.interruptible = true,
8415	};
8416	int err;
8417	do {
8418		err = -NFS4ERR_WRONGSEC;
8419
8420		/* try to use integrity protection with machine cred */
8421		if (_nfs4_is_integrity_protected(NFS_SERVER(dir)->nfs_client))
8422			err = _nfs4_proc_secinfo(dir, name, flavors, true);
8423
8424		/*
8425		 * if unable to use integrity protection, or SECINFO with
8426		 * integrity protection returns NFS4ERR_WRONGSEC (which is
8427		 * disallowed by spec, but exists in deployed servers) use
8428		 * the current filesystem's rpc_client and the user cred.
8429		 */
8430		if (err == -NFS4ERR_WRONGSEC)
8431			err = _nfs4_proc_secinfo(dir, name, flavors, false);
8432
8433		trace_nfs4_secinfo(dir, name, err);
8434		err = nfs4_handle_exception(NFS_SERVER(dir), err,
8435				&exception);
8436	} while (exception.retry);
8437	return err;
8438}
8439
8440#ifdef CONFIG_NFS_V4_1
8441/*
8442 * Check the exchange flags returned by the server for invalid flags, having
8443 * both PNFS and NON_PNFS flags set, and not having one of NON_PNFS, PNFS, or
8444 * DS flags set.
8445 */
8446static int nfs4_check_cl_exchange_flags(u32 flags, u32 version)
8447{
8448	if (version >= 2 && (flags & ~EXCHGID4_2_FLAG_MASK_R))
8449		goto out_inval;
8450	else if (version < 2 && (flags & ~EXCHGID4_FLAG_MASK_R))
8451		goto out_inval;
8452	if ((flags & EXCHGID4_FLAG_USE_PNFS_MDS) &&
8453	    (flags & EXCHGID4_FLAG_USE_NON_PNFS))
8454		goto out_inval;
8455	if (!(flags & (EXCHGID4_FLAG_MASK_PNFS)))
8456		goto out_inval;
8457	return NFS_OK;
8458out_inval:
8459	return -NFS4ERR_INVAL;
8460}
8461
8462static bool
8463nfs41_same_server_scope(struct nfs41_server_scope *a,
8464			struct nfs41_server_scope *b)
8465{
8466	if (a->server_scope_sz != b->server_scope_sz)
8467		return false;
8468	return memcmp(a->server_scope, b->server_scope, a->server_scope_sz) == 0;
8469}
8470
8471static void
8472nfs4_bind_one_conn_to_session_done(struct rpc_task *task, void *calldata)
8473{
8474	struct nfs41_bind_conn_to_session_args *args = task->tk_msg.rpc_argp;
8475	struct nfs41_bind_conn_to_session_res *res = task->tk_msg.rpc_resp;
8476	struct nfs_client *clp = args->client;
8477
8478	switch (task->tk_status) {
8479	case -NFS4ERR_BADSESSION:
8480	case -NFS4ERR_DEADSESSION:
8481		nfs4_schedule_session_recovery(clp->cl_session,
8482				task->tk_status);
8483		return;
8484	}
8485	if (args->dir == NFS4_CDFC4_FORE_OR_BOTH &&
8486			res->dir != NFS4_CDFS4_BOTH) {
8487		rpc_task_close_connection(task);
8488		if (args->retries++ < MAX_BIND_CONN_TO_SESSION_RETRIES)
8489			rpc_restart_call(task);
8490	}
8491}
8492
8493static const struct rpc_call_ops nfs4_bind_one_conn_to_session_ops = {
8494	.rpc_call_done =  nfs4_bind_one_conn_to_session_done,
8495};
8496
8497/*
8498 * nfs4_proc_bind_one_conn_to_session()
8499 *
8500 * The 4.1 client currently uses the same TCP connection for the
8501 * fore and backchannel.
8502 */
8503static
8504int nfs4_proc_bind_one_conn_to_session(struct rpc_clnt *clnt,
8505		struct rpc_xprt *xprt,
8506		struct nfs_client *clp,
8507		const struct cred *cred)
8508{
8509	int status;
8510	struct nfs41_bind_conn_to_session_args args = {
8511		.client = clp,
8512		.dir = NFS4_CDFC4_FORE_OR_BOTH,
8513		.retries = 0,
8514	};
8515	struct nfs41_bind_conn_to_session_res res;
8516	struct rpc_message msg = {
8517		.rpc_proc =
8518			&nfs4_procedures[NFSPROC4_CLNT_BIND_CONN_TO_SESSION],
8519		.rpc_argp = &args,
8520		.rpc_resp = &res,
8521		.rpc_cred = cred,
8522	};
8523	struct rpc_task_setup task_setup_data = {
8524		.rpc_client = clnt,
8525		.rpc_xprt = xprt,
8526		.callback_ops = &nfs4_bind_one_conn_to_session_ops,
8527		.rpc_message = &msg,
8528		.flags = RPC_TASK_TIMEOUT,
8529	};
8530	struct rpc_task *task;
8531
8532	nfs4_copy_sessionid(&args.sessionid, &clp->cl_session->sess_id);
8533	if (!(clp->cl_session->flags & SESSION4_BACK_CHAN))
8534		args.dir = NFS4_CDFC4_FORE;
8535
8536	/* Do not set the backchannel flag unless this is clnt->cl_xprt */
8537	if (xprt != rcu_access_pointer(clnt->cl_xprt))
8538		args.dir = NFS4_CDFC4_FORE;
8539
8540	task = rpc_run_task(&task_setup_data);
8541	if (!IS_ERR(task)) {
8542		status = task->tk_status;
8543		rpc_put_task(task);
8544	} else
8545		status = PTR_ERR(task);
8546	trace_nfs4_bind_conn_to_session(clp, status);
8547	if (status == 0) {
8548		if (memcmp(res.sessionid.data,
8549		    clp->cl_session->sess_id.data, NFS4_MAX_SESSIONID_LEN)) {
8550			dprintk("NFS: %s: Session ID mismatch\n", __func__);
8551			return -EIO;
8552		}
8553		if ((res.dir & args.dir) != res.dir || res.dir == 0) {
8554			dprintk("NFS: %s: Unexpected direction from server\n",
8555				__func__);
8556			return -EIO;
8557		}
8558		if (res.use_conn_in_rdma_mode != args.use_conn_in_rdma_mode) {
8559			dprintk("NFS: %s: Server returned RDMA mode = true\n",
8560				__func__);
8561			return -EIO;
8562		}
8563	}
8564
8565	return status;
8566}
8567
8568struct rpc_bind_conn_calldata {
8569	struct nfs_client *clp;
8570	const struct cred *cred;
8571};
8572
8573static int
8574nfs4_proc_bind_conn_to_session_callback(struct rpc_clnt *clnt,
8575		struct rpc_xprt *xprt,
8576		void *calldata)
8577{
8578	struct rpc_bind_conn_calldata *p = calldata;
8579
8580	return nfs4_proc_bind_one_conn_to_session(clnt, xprt, p->clp, p->cred);
8581}
8582
8583int nfs4_proc_bind_conn_to_session(struct nfs_client *clp, const struct cred *cred)
8584{
8585	struct rpc_bind_conn_calldata data = {
8586		.clp = clp,
8587		.cred = cred,
8588	};
8589	return rpc_clnt_iterate_for_each_xprt(clp->cl_rpcclient,
8590			nfs4_proc_bind_conn_to_session_callback, &data);
8591}
8592
8593/*
8594 * Minimum set of SP4_MACH_CRED operations from RFC 5661 in the enforce map
8595 * and operations we'd like to see to enable certain features in the allow map
8596 */
8597static const struct nfs41_state_protection nfs4_sp4_mach_cred_request = {
8598	.how = SP4_MACH_CRED,
8599	.enforce.u.words = {
8600		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8601		      1 << (OP_EXCHANGE_ID - 32) |
8602		      1 << (OP_CREATE_SESSION - 32) |
8603		      1 << (OP_DESTROY_SESSION - 32) |
8604		      1 << (OP_DESTROY_CLIENTID - 32)
8605	},
8606	.allow.u.words = {
8607		[0] = 1 << (OP_CLOSE) |
8608		      1 << (OP_OPEN_DOWNGRADE) |
8609		      1 << (OP_LOCKU) |
8610		      1 << (OP_DELEGRETURN) |
8611		      1 << (OP_COMMIT),
8612		[1] = 1 << (OP_SECINFO - 32) |
8613		      1 << (OP_SECINFO_NO_NAME - 32) |
8614		      1 << (OP_LAYOUTRETURN - 32) |
8615		      1 << (OP_TEST_STATEID - 32) |
8616		      1 << (OP_FREE_STATEID - 32) |
8617		      1 << (OP_WRITE - 32)
8618	}
8619};
8620
8621/*
8622 * Select the state protection mode for client `clp' given the server results
8623 * from exchange_id in `sp'.
8624 *
8625 * Returns 0 on success, negative errno otherwise.
8626 */
8627static int nfs4_sp4_select_mode(struct nfs_client *clp,
8628				 struct nfs41_state_protection *sp)
8629{
8630	static const u32 supported_enforce[NFS4_OP_MAP_NUM_WORDS] = {
8631		[1] = 1 << (OP_BIND_CONN_TO_SESSION - 32) |
8632		      1 << (OP_EXCHANGE_ID - 32) |
8633		      1 << (OP_CREATE_SESSION - 32) |
8634		      1 << (OP_DESTROY_SESSION - 32) |
8635		      1 << (OP_DESTROY_CLIENTID - 32)
8636	};
8637	unsigned long flags = 0;
8638	unsigned int i;
8639	int ret = 0;
8640
8641	if (sp->how == SP4_MACH_CRED) {
8642		/* Print state protect result */
8643		dfprintk(MOUNT, "Server SP4_MACH_CRED support:\n");
8644		for (i = 0; i <= LAST_NFS4_OP; i++) {
8645			if (test_bit(i, sp->enforce.u.longs))
8646				dfprintk(MOUNT, "  enforce op %d\n", i);
8647			if (test_bit(i, sp->allow.u.longs))
8648				dfprintk(MOUNT, "  allow op %d\n", i);
8649		}
8650
8651		/* make sure nothing is on enforce list that isn't supported */
8652		for (i = 0; i < NFS4_OP_MAP_NUM_WORDS; i++) {
8653			if (sp->enforce.u.words[i] & ~supported_enforce[i]) {
8654				dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8655				ret = -EINVAL;
8656				goto out;
8657			}
8658		}
8659
8660		/*
8661		 * Minimal mode - state operations are allowed to use machine
8662		 * credential.  Note this already happens by default, so the
8663		 * client doesn't have to do anything more than the negotiation.
8664		 *
8665		 * NOTE: we don't care if EXCHANGE_ID is in the list -
8666		 *       we're already using the machine cred for exchange_id
8667		 *       and will never use a different cred.
8668		 */
8669		if (test_bit(OP_BIND_CONN_TO_SESSION, sp->enforce.u.longs) &&
8670		    test_bit(OP_CREATE_SESSION, sp->enforce.u.longs) &&
8671		    test_bit(OP_DESTROY_SESSION, sp->enforce.u.longs) &&
8672		    test_bit(OP_DESTROY_CLIENTID, sp->enforce.u.longs)) {
8673			dfprintk(MOUNT, "sp4_mach_cred:\n");
8674			dfprintk(MOUNT, "  minimal mode enabled\n");
8675			__set_bit(NFS_SP4_MACH_CRED_MINIMAL, &flags);
8676		} else {
8677			dfprintk(MOUNT, "sp4_mach_cred: disabled\n");
8678			ret = -EINVAL;
8679			goto out;
8680		}
8681
8682		if (test_bit(OP_CLOSE, sp->allow.u.longs) &&
8683		    test_bit(OP_OPEN_DOWNGRADE, sp->allow.u.longs) &&
8684		    test_bit(OP_DELEGRETURN, sp->allow.u.longs) &&
8685		    test_bit(OP_LOCKU, sp->allow.u.longs)) {
8686			dfprintk(MOUNT, "  cleanup mode enabled\n");
8687			__set_bit(NFS_SP4_MACH_CRED_CLEANUP, &flags);
8688		}
8689
8690		if (test_bit(OP_LAYOUTRETURN, sp->allow.u.longs)) {
8691			dfprintk(MOUNT, "  pnfs cleanup mode enabled\n");
8692			__set_bit(NFS_SP4_MACH_CRED_PNFS_CLEANUP, &flags);
8693		}
8694
8695		if (test_bit(OP_SECINFO, sp->allow.u.longs) &&
8696		    test_bit(OP_SECINFO_NO_NAME, sp->allow.u.longs)) {
8697			dfprintk(MOUNT, "  secinfo mode enabled\n");
8698			__set_bit(NFS_SP4_MACH_CRED_SECINFO, &flags);
8699		}
8700
8701		if (test_bit(OP_TEST_STATEID, sp->allow.u.longs) &&
8702		    test_bit(OP_FREE_STATEID, sp->allow.u.longs)) {
8703			dfprintk(MOUNT, "  stateid mode enabled\n");
8704			__set_bit(NFS_SP4_MACH_CRED_STATEID, &flags);
8705		}
8706
8707		if (test_bit(OP_WRITE, sp->allow.u.longs)) {
8708			dfprintk(MOUNT, "  write mode enabled\n");
8709			__set_bit(NFS_SP4_MACH_CRED_WRITE, &flags);
8710		}
8711
8712		if (test_bit(OP_COMMIT, sp->allow.u.longs)) {
8713			dfprintk(MOUNT, "  commit mode enabled\n");
8714			__set_bit(NFS_SP4_MACH_CRED_COMMIT, &flags);
8715		}
8716	}
8717out:
8718	clp->cl_sp4_flags = flags;
8719	return ret;
8720}
8721
8722struct nfs41_exchange_id_data {
8723	struct nfs41_exchange_id_res res;
8724	struct nfs41_exchange_id_args args;
8725};
8726
8727static void nfs4_exchange_id_release(void *data)
8728{
8729	struct nfs41_exchange_id_data *cdata =
8730					(struct nfs41_exchange_id_data *)data;
8731
8732	nfs_put_client(cdata->args.client);
8733	kfree(cdata->res.impl_id);
8734	kfree(cdata->res.server_scope);
8735	kfree(cdata->res.server_owner);
8736	kfree(cdata);
8737}
8738
8739static const struct rpc_call_ops nfs4_exchange_id_call_ops = {
8740	.rpc_release = nfs4_exchange_id_release,
8741};
8742
8743/*
8744 * _nfs4_proc_exchange_id()
8745 *
8746 * Wrapper for EXCHANGE_ID operation.
8747 */
8748static struct rpc_task *
8749nfs4_run_exchange_id(struct nfs_client *clp, const struct cred *cred,
8750			u32 sp4_how, struct rpc_xprt *xprt)
8751{
8752	struct rpc_message msg = {
8753		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_EXCHANGE_ID],
8754		.rpc_cred = cred,
8755	};
8756	struct rpc_task_setup task_setup_data = {
8757		.rpc_client = clp->cl_rpcclient,
8758		.callback_ops = &nfs4_exchange_id_call_ops,
8759		.rpc_message = &msg,
8760		.flags = RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN,
8761	};
8762	struct nfs41_exchange_id_data *calldata;
8763	int status;
8764
8765	if (!refcount_inc_not_zero(&clp->cl_count))
8766		return ERR_PTR(-EIO);
8767
8768	status = -ENOMEM;
8769	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
8770	if (!calldata)
8771		goto out;
8772
8773	nfs4_init_boot_verifier(clp, &calldata->args.verifier);
8774
8775	status = nfs4_init_uniform_client_string(clp);
8776	if (status)
8777		goto out_calldata;
8778
8779	calldata->res.server_owner = kzalloc(sizeof(struct nfs41_server_owner),
8780						GFP_NOFS);
8781	status = -ENOMEM;
8782	if (unlikely(calldata->res.server_owner == NULL))
8783		goto out_calldata;
8784
8785	calldata->res.server_scope = kzalloc(sizeof(struct nfs41_server_scope),
8786					GFP_NOFS);
8787	if (unlikely(calldata->res.server_scope == NULL))
8788		goto out_server_owner;
8789
8790	calldata->res.impl_id = kzalloc(sizeof(struct nfs41_impl_id), GFP_NOFS);
8791	if (unlikely(calldata->res.impl_id == NULL))
8792		goto out_server_scope;
8793
8794	switch (sp4_how) {
8795	case SP4_NONE:
8796		calldata->args.state_protect.how = SP4_NONE;
8797		break;
8798
8799	case SP4_MACH_CRED:
8800		calldata->args.state_protect = nfs4_sp4_mach_cred_request;
8801		break;
8802
8803	default:
8804		/* unsupported! */
8805		WARN_ON_ONCE(1);
8806		status = -EINVAL;
8807		goto out_impl_id;
8808	}
8809	if (xprt) {
8810		task_setup_data.rpc_xprt = xprt;
8811		task_setup_data.flags |= RPC_TASK_SOFTCONN;
8812		memcpy(calldata->args.verifier.data, clp->cl_confirm.data,
8813				sizeof(calldata->args.verifier.data));
8814	}
8815	calldata->args.client = clp;
8816	calldata->args.flags = EXCHGID4_FLAG_SUPP_MOVED_REFER |
8817	EXCHGID4_FLAG_BIND_PRINC_STATEID;
8818#ifdef CONFIG_NFS_V4_1_MIGRATION
8819	calldata->args.flags |= EXCHGID4_FLAG_SUPP_MOVED_MIGR;
8820#endif
8821	if (test_bit(NFS_CS_DS, &clp->cl_flags))
8822		calldata->args.flags |= EXCHGID4_FLAG_USE_PNFS_DS;
8823	msg.rpc_argp = &calldata->args;
8824	msg.rpc_resp = &calldata->res;
8825	task_setup_data.callback_data = calldata;
8826
8827	return rpc_run_task(&task_setup_data);
8828
8829out_impl_id:
8830	kfree(calldata->res.impl_id);
8831out_server_scope:
8832	kfree(calldata->res.server_scope);
8833out_server_owner:
8834	kfree(calldata->res.server_owner);
8835out_calldata:
8836	kfree(calldata);
8837out:
8838	nfs_put_client(clp);
8839	return ERR_PTR(status);
8840}
8841
8842/*
8843 * _nfs4_proc_exchange_id()
8844 *
8845 * Wrapper for EXCHANGE_ID operation.
8846 */
8847static int _nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred,
8848			u32 sp4_how)
8849{
8850	struct rpc_task *task;
8851	struct nfs41_exchange_id_args *argp;
8852	struct nfs41_exchange_id_res *resp;
8853	unsigned long now = jiffies;
8854	int status;
8855
8856	task = nfs4_run_exchange_id(clp, cred, sp4_how, NULL);
8857	if (IS_ERR(task))
8858		return PTR_ERR(task);
8859
8860	argp = task->tk_msg.rpc_argp;
8861	resp = task->tk_msg.rpc_resp;
8862	status = task->tk_status;
8863	if (status  != 0)
8864		goto out;
8865
8866	status = nfs4_check_cl_exchange_flags(resp->flags,
8867			clp->cl_mvops->minor_version);
8868	if (status  != 0)
8869		goto out;
8870
8871	status = nfs4_sp4_select_mode(clp, &resp->state_protect);
8872	if (status != 0)
8873		goto out;
8874
8875	do_renew_lease(clp, now);
8876
8877	clp->cl_clientid = resp->clientid;
8878	clp->cl_exchange_flags = resp->flags;
8879	clp->cl_seqid = resp->seqid;
8880	/* Client ID is not confirmed */
8881	if (!(resp->flags & EXCHGID4_FLAG_CONFIRMED_R))
8882		clear_bit(NFS4_SESSION_ESTABLISHED,
8883			  &clp->cl_session->session_state);
8884
8885	if (clp->cl_serverscope != NULL &&
8886	    !nfs41_same_server_scope(clp->cl_serverscope,
8887				resp->server_scope)) {
8888		dprintk("%s: server_scope mismatch detected\n",
8889			__func__);
8890		set_bit(NFS4CLNT_SERVER_SCOPE_MISMATCH, &clp->cl_state);
8891	}
8892
8893	swap(clp->cl_serverowner, resp->server_owner);
8894	swap(clp->cl_serverscope, resp->server_scope);
8895	swap(clp->cl_implid, resp->impl_id);
8896
8897	/* Save the EXCHANGE_ID verifier session trunk tests */
8898	memcpy(clp->cl_confirm.data, argp->verifier.data,
8899	       sizeof(clp->cl_confirm.data));
8900out:
8901	trace_nfs4_exchange_id(clp, status);
8902	rpc_put_task(task);
8903	return status;
8904}
8905
8906/*
8907 * nfs4_proc_exchange_id()
8908 *
8909 * Returns zero, a negative errno, or a negative NFS4ERR status code.
8910 *
8911 * Since the clientid has expired, all compounds using sessions
8912 * associated with the stale clientid will be returning
8913 * NFS4ERR_BADSESSION in the sequence operation, and will therefore
8914 * be in some phase of session reset.
8915 *
8916 * Will attempt to negotiate SP4_MACH_CRED if krb5i / krb5p auth is used.
8917 */
8918int nfs4_proc_exchange_id(struct nfs_client *clp, const struct cred *cred)
8919{
8920	rpc_authflavor_t authflavor = clp->cl_rpcclient->cl_auth->au_flavor;
8921	int status;
8922
8923	/* try SP4_MACH_CRED if krb5i/p	*/
8924	if (authflavor == RPC_AUTH_GSS_KRB5I ||
8925	    authflavor == RPC_AUTH_GSS_KRB5P) {
8926		status = _nfs4_proc_exchange_id(clp, cred, SP4_MACH_CRED);
8927		if (!status)
8928			return 0;
8929	}
8930
8931	/* try SP4_NONE */
8932	return _nfs4_proc_exchange_id(clp, cred, SP4_NONE);
8933}
8934
8935/**
8936 * nfs4_test_session_trunk
8937 *
8938 * This is an add_xprt_test() test function called from
8939 * rpc_clnt_setup_test_and_add_xprt.
8940 *
8941 * The rpc_xprt_switch is referrenced by rpc_clnt_setup_test_and_add_xprt
8942 * and is dereferrenced in nfs4_exchange_id_release
8943 *
8944 * Upon success, add the new transport to the rpc_clnt
8945 *
8946 * @clnt: struct rpc_clnt to get new transport
8947 * @xprt: the rpc_xprt to test
8948 * @data: call data for _nfs4_proc_exchange_id.
8949 */
8950void nfs4_test_session_trunk(struct rpc_clnt *clnt, struct rpc_xprt *xprt,
8951			    void *data)
8952{
8953	struct nfs4_add_xprt_data *adata = data;
8954	struct rpc_task *task;
8955	int status;
8956
8957	u32 sp4_how;
8958
8959	dprintk("--> %s try %s\n", __func__,
8960		xprt->address_strings[RPC_DISPLAY_ADDR]);
8961
8962	sp4_how = (adata->clp->cl_sp4_flags == 0 ? SP4_NONE : SP4_MACH_CRED);
8963
8964try_again:
8965	/* Test connection for session trunking. Async exchange_id call */
8966	task = nfs4_run_exchange_id(adata->clp, adata->cred, sp4_how, xprt);
8967	if (IS_ERR(task))
8968		return;
8969
8970	status = task->tk_status;
8971	if (status == 0) {
8972		status = nfs4_detect_session_trunking(adata->clp,
8973				task->tk_msg.rpc_resp, xprt);
8974		trace_nfs4_trunked_exchange_id(adata->clp,
8975			xprt->address_strings[RPC_DISPLAY_ADDR], status);
8976	}
8977	if (status == 0)
8978		rpc_clnt_xprt_switch_add_xprt(clnt, xprt);
8979	else if (status != -NFS4ERR_DELAY && rpc_clnt_xprt_switch_has_addr(clnt,
8980				(struct sockaddr *)&xprt->addr))
8981		rpc_clnt_xprt_switch_remove_xprt(clnt, xprt);
8982
8983	rpc_put_task(task);
8984	if (status == -NFS4ERR_DELAY) {
8985		ssleep(1);
8986		goto try_again;
8987	}
8988}
8989EXPORT_SYMBOL_GPL(nfs4_test_session_trunk);
8990
8991static int _nfs4_proc_destroy_clientid(struct nfs_client *clp,
8992		const struct cred *cred)
8993{
8994	struct rpc_message msg = {
8995		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_CLIENTID],
8996		.rpc_argp = clp,
8997		.rpc_cred = cred,
8998	};
8999	int status;
9000
9001	status = rpc_call_sync(clp->cl_rpcclient, &msg,
9002			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9003	trace_nfs4_destroy_clientid(clp, status);
9004	if (status)
9005		dprintk("NFS: Got error %d from the server %s on "
9006			"DESTROY_CLIENTID.", status, clp->cl_hostname);
9007	return status;
9008}
9009
9010static int nfs4_proc_destroy_clientid(struct nfs_client *clp,
9011		const struct cred *cred)
9012{
9013	unsigned int loop;
9014	int ret;
9015
9016	for (loop = NFS4_MAX_LOOP_ON_RECOVER; loop != 0; loop--) {
9017		ret = _nfs4_proc_destroy_clientid(clp, cred);
9018		switch (ret) {
9019		case -NFS4ERR_DELAY:
9020		case -NFS4ERR_CLIENTID_BUSY:
9021			ssleep(1);
9022			break;
9023		default:
9024			return ret;
9025		}
9026	}
9027	return 0;
9028}
9029
9030int nfs4_destroy_clientid(struct nfs_client *clp)
9031{
9032	const struct cred *cred;
9033	int ret = 0;
9034
9035	if (clp->cl_mvops->minor_version < 1)
9036		goto out;
9037	if (clp->cl_exchange_flags == 0)
9038		goto out;
9039	if (clp->cl_preserve_clid)
9040		goto out;
9041	cred = nfs4_get_clid_cred(clp);
9042	ret = nfs4_proc_destroy_clientid(clp, cred);
9043	put_cred(cred);
9044	switch (ret) {
9045	case 0:
9046	case -NFS4ERR_STALE_CLIENTID:
9047		clp->cl_exchange_flags = 0;
9048	}
9049out:
9050	return ret;
9051}
9052
9053#endif /* CONFIG_NFS_V4_1 */
9054
9055struct nfs4_get_lease_time_data {
9056	struct nfs4_get_lease_time_args *args;
9057	struct nfs4_get_lease_time_res *res;
9058	struct nfs_client *clp;
9059};
9060
9061static void nfs4_get_lease_time_prepare(struct rpc_task *task,
9062					void *calldata)
9063{
9064	struct nfs4_get_lease_time_data *data =
9065			(struct nfs4_get_lease_time_data *)calldata;
9066
9067	/* just setup sequence, do not trigger session recovery
9068	   since we're invoked within one */
9069	nfs4_setup_sequence(data->clp,
9070			&data->args->la_seq_args,
9071			&data->res->lr_seq_res,
9072			task);
9073}
9074
9075/*
9076 * Called from nfs4_state_manager thread for session setup, so don't recover
9077 * from sequence operation or clientid errors.
9078 */
9079static void nfs4_get_lease_time_done(struct rpc_task *task, void *calldata)
9080{
9081	struct nfs4_get_lease_time_data *data =
9082			(struct nfs4_get_lease_time_data *)calldata;
9083
9084	if (!nfs4_sequence_done(task, &data->res->lr_seq_res))
9085		return;
9086	switch (task->tk_status) {
9087	case -NFS4ERR_DELAY:
9088	case -NFS4ERR_GRACE:
9089		rpc_delay(task, NFS4_POLL_RETRY_MIN);
9090		task->tk_status = 0;
9091		fallthrough;
9092	case -NFS4ERR_RETRY_UNCACHED_REP:
9093		rpc_restart_call_prepare(task);
9094		return;
9095	}
9096}
9097
9098static const struct rpc_call_ops nfs4_get_lease_time_ops = {
9099	.rpc_call_prepare = nfs4_get_lease_time_prepare,
9100	.rpc_call_done = nfs4_get_lease_time_done,
9101};
9102
9103int nfs4_proc_get_lease_time(struct nfs_client *clp, struct nfs_fsinfo *fsinfo)
9104{
9105	struct nfs4_get_lease_time_args args;
9106	struct nfs4_get_lease_time_res res = {
9107		.lr_fsinfo = fsinfo,
9108	};
9109	struct nfs4_get_lease_time_data data = {
9110		.args = &args,
9111		.res = &res,
9112		.clp = clp,
9113	};
9114	struct rpc_message msg = {
9115		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GET_LEASE_TIME],
9116		.rpc_argp = &args,
9117		.rpc_resp = &res,
9118	};
9119	struct rpc_task_setup task_setup = {
9120		.rpc_client = clp->cl_rpcclient,
9121		.rpc_message = &msg,
9122		.callback_ops = &nfs4_get_lease_time_ops,
9123		.callback_data = &data,
9124		.flags = RPC_TASK_TIMEOUT,
9125	};
9126
9127	nfs4_init_sequence(&args.la_seq_args, &res.lr_seq_res, 0, 1);
9128	return nfs4_call_sync_custom(&task_setup);
9129}
9130
9131#ifdef CONFIG_NFS_V4_1
9132
9133/*
9134 * Initialize the values to be used by the client in CREATE_SESSION
9135 * If nfs4_init_session set the fore channel request and response sizes,
9136 * use them.
9137 *
9138 * Set the back channel max_resp_sz_cached to zero to force the client to
9139 * always set csa_cachethis to FALSE because the current implementation
9140 * of the back channel DRC only supports caching the CB_SEQUENCE operation.
9141 */
9142static void nfs4_init_channel_attrs(struct nfs41_create_session_args *args,
9143				    struct rpc_clnt *clnt)
9144{
9145	unsigned int max_rqst_sz, max_resp_sz;
9146	unsigned int max_bc_payload = rpc_max_bc_payload(clnt);
9147	unsigned int max_bc_slots = rpc_num_bc_slots(clnt);
9148
9149	max_rqst_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxwrite_overhead;
9150	max_resp_sz = NFS_MAX_FILE_IO_SIZE + nfs41_maxread_overhead;
9151
9152	/* Fore channel attributes */
9153	args->fc_attrs.max_rqst_sz = max_rqst_sz;
9154	args->fc_attrs.max_resp_sz = max_resp_sz;
9155	args->fc_attrs.max_ops = NFS4_MAX_OPS;
9156	args->fc_attrs.max_reqs = max_session_slots;
9157
9158	dprintk("%s: Fore Channel : max_rqst_sz=%u max_resp_sz=%u "
9159		"max_ops=%u max_reqs=%u\n",
9160		__func__,
9161		args->fc_attrs.max_rqst_sz, args->fc_attrs.max_resp_sz,
9162		args->fc_attrs.max_ops, args->fc_attrs.max_reqs);
9163
9164	/* Back channel attributes */
9165	args->bc_attrs.max_rqst_sz = max_bc_payload;
9166	args->bc_attrs.max_resp_sz = max_bc_payload;
9167	args->bc_attrs.max_resp_sz_cached = 0;
9168	args->bc_attrs.max_ops = NFS4_MAX_BACK_CHANNEL_OPS;
9169	args->bc_attrs.max_reqs = max_t(unsigned short, max_session_cb_slots, 1);
9170	if (args->bc_attrs.max_reqs > max_bc_slots)
9171		args->bc_attrs.max_reqs = max_bc_slots;
9172
9173	dprintk("%s: Back Channel : max_rqst_sz=%u max_resp_sz=%u "
9174		"max_resp_sz_cached=%u max_ops=%u max_reqs=%u\n",
9175		__func__,
9176		args->bc_attrs.max_rqst_sz, args->bc_attrs.max_resp_sz,
9177		args->bc_attrs.max_resp_sz_cached, args->bc_attrs.max_ops,
9178		args->bc_attrs.max_reqs);
9179}
9180
9181static int nfs4_verify_fore_channel_attrs(struct nfs41_create_session_args *args,
9182		struct nfs41_create_session_res *res)
9183{
9184	struct nfs4_channel_attrs *sent = &args->fc_attrs;
9185	struct nfs4_channel_attrs *rcvd = &res->fc_attrs;
9186
9187	if (rcvd->max_resp_sz > sent->max_resp_sz)
9188		return -EINVAL;
9189	/*
9190	 * Our requested max_ops is the minimum we need; we're not
9191	 * prepared to break up compounds into smaller pieces than that.
9192	 * So, no point even trying to continue if the server won't
9193	 * cooperate:
9194	 */
9195	if (rcvd->max_ops < sent->max_ops)
9196		return -EINVAL;
9197	if (rcvd->max_reqs == 0)
9198		return -EINVAL;
9199	if (rcvd->max_reqs > NFS4_MAX_SLOT_TABLE)
9200		rcvd->max_reqs = NFS4_MAX_SLOT_TABLE;
9201	return 0;
9202}
9203
9204static int nfs4_verify_back_channel_attrs(struct nfs41_create_session_args *args,
9205		struct nfs41_create_session_res *res)
9206{
9207	struct nfs4_channel_attrs *sent = &args->bc_attrs;
9208	struct nfs4_channel_attrs *rcvd = &res->bc_attrs;
9209
9210	if (!(res->flags & SESSION4_BACK_CHAN))
9211		goto out;
9212	if (rcvd->max_rqst_sz > sent->max_rqst_sz)
9213		return -EINVAL;
9214	if (rcvd->max_resp_sz < sent->max_resp_sz)
9215		return -EINVAL;
9216	if (rcvd->max_resp_sz_cached > sent->max_resp_sz_cached)
9217		return -EINVAL;
9218	if (rcvd->max_ops > sent->max_ops)
9219		return -EINVAL;
9220	if (rcvd->max_reqs > sent->max_reqs)
9221		return -EINVAL;
9222out:
9223	return 0;
9224}
9225
9226static int nfs4_verify_channel_attrs(struct nfs41_create_session_args *args,
9227				     struct nfs41_create_session_res *res)
9228{
9229	int ret;
9230
9231	ret = nfs4_verify_fore_channel_attrs(args, res);
9232	if (ret)
9233		return ret;
9234	return nfs4_verify_back_channel_attrs(args, res);
9235}
9236
9237static void nfs4_update_session(struct nfs4_session *session,
9238		struct nfs41_create_session_res *res)
9239{
9240	nfs4_copy_sessionid(&session->sess_id, &res->sessionid);
9241	/* Mark client id and session as being confirmed */
9242	session->clp->cl_exchange_flags |= EXCHGID4_FLAG_CONFIRMED_R;
9243	set_bit(NFS4_SESSION_ESTABLISHED, &session->session_state);
9244	session->flags = res->flags;
9245	memcpy(&session->fc_attrs, &res->fc_attrs, sizeof(session->fc_attrs));
9246	if (res->flags & SESSION4_BACK_CHAN)
9247		memcpy(&session->bc_attrs, &res->bc_attrs,
9248				sizeof(session->bc_attrs));
9249}
9250
9251static int _nfs4_proc_create_session(struct nfs_client *clp,
9252		const struct cred *cred)
9253{
9254	struct nfs4_session *session = clp->cl_session;
9255	struct nfs41_create_session_args args = {
9256		.client = clp,
9257		.clientid = clp->cl_clientid,
9258		.seqid = clp->cl_seqid,
9259		.cb_program = NFS4_CALLBACK,
9260	};
9261	struct nfs41_create_session_res res;
9262
9263	struct rpc_message msg = {
9264		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CREATE_SESSION],
9265		.rpc_argp = &args,
9266		.rpc_resp = &res,
9267		.rpc_cred = cred,
9268	};
9269	int status;
9270
9271	nfs4_init_channel_attrs(&args, clp->cl_rpcclient);
9272	args.flags = (SESSION4_PERSIST | SESSION4_BACK_CHAN);
9273
9274	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9275			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9276	trace_nfs4_create_session(clp, status);
9277
9278	switch (status) {
9279	case -NFS4ERR_STALE_CLIENTID:
9280	case -NFS4ERR_DELAY:
9281	case -ETIMEDOUT:
9282	case -EACCES:
9283	case -EAGAIN:
9284		goto out;
9285	}
9286
9287	clp->cl_seqid++;
9288	if (!status) {
9289		/* Verify the session's negotiated channel_attrs values */
9290		status = nfs4_verify_channel_attrs(&args, &res);
9291		/* Increment the clientid slot sequence id */
9292		if (status)
9293			goto out;
9294		nfs4_update_session(session, &res);
9295	}
9296out:
9297	return status;
9298}
9299
9300/*
9301 * Issues a CREATE_SESSION operation to the server.
9302 * It is the responsibility of the caller to verify the session is
9303 * expired before calling this routine.
9304 */
9305int nfs4_proc_create_session(struct nfs_client *clp, const struct cred *cred)
9306{
9307	int status;
9308	unsigned *ptr;
9309	struct nfs4_session *session = clp->cl_session;
9310	struct nfs4_add_xprt_data xprtdata = {
9311		.clp = clp,
9312	};
9313	struct rpc_add_xprt_test rpcdata = {
9314		.add_xprt_test = clp->cl_mvops->session_trunk,
9315		.data = &xprtdata,
9316	};
9317
9318	dprintk("--> %s clp=%p session=%p\n", __func__, clp, session);
9319
9320	status = _nfs4_proc_create_session(clp, cred);
9321	if (status)
9322		goto out;
9323
9324	/* Init or reset the session slot tables */
9325	status = nfs4_setup_session_slot_tables(session);
9326	dprintk("slot table setup returned %d\n", status);
9327	if (status)
9328		goto out;
9329
9330	ptr = (unsigned *)&session->sess_id.data[0];
9331	dprintk("%s client>seqid %d sessionid %u:%u:%u:%u\n", __func__,
9332		clp->cl_seqid, ptr[0], ptr[1], ptr[2], ptr[3]);
9333	rpc_clnt_probe_trunked_xprts(clp->cl_rpcclient, &rpcdata);
9334out:
9335	return status;
9336}
9337
9338/*
9339 * Issue the over-the-wire RPC DESTROY_SESSION.
9340 * The caller must serialize access to this routine.
9341 */
9342int nfs4_proc_destroy_session(struct nfs4_session *session,
9343		const struct cred *cred)
9344{
9345	struct rpc_message msg = {
9346		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DESTROY_SESSION],
9347		.rpc_argp = session,
9348		.rpc_cred = cred,
9349	};
9350	int status = 0;
9351
9352	/* session is still being setup */
9353	if (!test_and_clear_bit(NFS4_SESSION_ESTABLISHED, &session->session_state))
9354		return 0;
9355
9356	status = rpc_call_sync(session->clp->cl_rpcclient, &msg,
9357			       RPC_TASK_TIMEOUT | RPC_TASK_NO_ROUND_ROBIN);
9358	trace_nfs4_destroy_session(session->clp, status);
9359
9360	if (status)
9361		dprintk("NFS: Got error %d from the server on DESTROY_SESSION. "
9362			"Session has been destroyed regardless...\n", status);
9363	rpc_clnt_manage_trunked_xprts(session->clp->cl_rpcclient);
9364	return status;
9365}
9366
9367/*
9368 * Renew the cl_session lease.
9369 */
9370struct nfs4_sequence_data {
9371	struct nfs_client *clp;
9372	struct nfs4_sequence_args args;
9373	struct nfs4_sequence_res res;
9374};
9375
9376static void nfs41_sequence_release(void *data)
9377{
9378	struct nfs4_sequence_data *calldata = data;
9379	struct nfs_client *clp = calldata->clp;
9380
9381	if (refcount_read(&clp->cl_count) > 1)
9382		nfs4_schedule_state_renewal(clp);
9383	nfs_put_client(clp);
9384	kfree(calldata);
9385}
9386
9387static int nfs41_sequence_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9388{
9389	switch(task->tk_status) {
9390	case -NFS4ERR_DELAY:
9391		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9392		return -EAGAIN;
9393	default:
9394		nfs4_schedule_lease_recovery(clp);
9395	}
9396	return 0;
9397}
9398
9399static void nfs41_sequence_call_done(struct rpc_task *task, void *data)
9400{
9401	struct nfs4_sequence_data *calldata = data;
9402	struct nfs_client *clp = calldata->clp;
9403
9404	if (!nfs41_sequence_done(task, task->tk_msg.rpc_resp))
9405		return;
9406
9407	trace_nfs4_sequence(clp, task->tk_status);
9408	if (task->tk_status < 0 && !task->tk_client->cl_shutdown) {
9409		dprintk("%s ERROR %d\n", __func__, task->tk_status);
9410		if (refcount_read(&clp->cl_count) == 1)
9411			return;
9412
9413		if (nfs41_sequence_handle_errors(task, clp) == -EAGAIN) {
9414			rpc_restart_call_prepare(task);
9415			return;
9416		}
9417	}
9418	dprintk("%s rpc_cred %p\n", __func__, task->tk_msg.rpc_cred);
9419}
9420
9421static void nfs41_sequence_prepare(struct rpc_task *task, void *data)
9422{
9423	struct nfs4_sequence_data *calldata = data;
9424	struct nfs_client *clp = calldata->clp;
9425	struct nfs4_sequence_args *args;
9426	struct nfs4_sequence_res *res;
9427
9428	args = task->tk_msg.rpc_argp;
9429	res = task->tk_msg.rpc_resp;
9430
9431	nfs4_setup_sequence(clp, args, res, task);
9432}
9433
9434static const struct rpc_call_ops nfs41_sequence_ops = {
9435	.rpc_call_done = nfs41_sequence_call_done,
9436	.rpc_call_prepare = nfs41_sequence_prepare,
9437	.rpc_release = nfs41_sequence_release,
9438};
9439
9440static struct rpc_task *_nfs41_proc_sequence(struct nfs_client *clp,
9441		const struct cred *cred,
9442		struct nfs4_slot *slot,
9443		bool is_privileged)
9444{
9445	struct nfs4_sequence_data *calldata;
9446	struct rpc_message msg = {
9447		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEQUENCE],
9448		.rpc_cred = cred,
9449	};
9450	struct rpc_task_setup task_setup_data = {
9451		.rpc_client = clp->cl_rpcclient,
9452		.rpc_message = &msg,
9453		.callback_ops = &nfs41_sequence_ops,
9454		.flags = RPC_TASK_ASYNC | RPC_TASK_TIMEOUT | RPC_TASK_MOVEABLE,
9455	};
9456	struct rpc_task *ret;
9457
9458	ret = ERR_PTR(-EIO);
9459	if (!refcount_inc_not_zero(&clp->cl_count))
9460		goto out_err;
9461
9462	ret = ERR_PTR(-ENOMEM);
9463	calldata = kzalloc(sizeof(*calldata), GFP_KERNEL);
9464	if (calldata == NULL)
9465		goto out_put_clp;
9466	nfs4_init_sequence(&calldata->args, &calldata->res, 0, is_privileged);
9467	nfs4_sequence_attach_slot(&calldata->args, &calldata->res, slot);
9468	msg.rpc_argp = &calldata->args;
9469	msg.rpc_resp = &calldata->res;
9470	calldata->clp = clp;
9471	task_setup_data.callback_data = calldata;
9472
9473	ret = rpc_run_task(&task_setup_data);
9474	if (IS_ERR(ret))
9475		goto out_err;
9476	return ret;
9477out_put_clp:
9478	nfs_put_client(clp);
9479out_err:
9480	nfs41_release_slot(slot);
9481	return ret;
9482}
9483
9484static int nfs41_proc_async_sequence(struct nfs_client *clp, const struct cred *cred, unsigned renew_flags)
9485{
9486	struct rpc_task *task;
9487	int ret = 0;
9488
9489	if ((renew_flags & NFS4_RENEW_TIMEOUT) == 0)
9490		return -EAGAIN;
9491	task = _nfs41_proc_sequence(clp, cred, NULL, false);
9492	if (IS_ERR(task))
9493		ret = PTR_ERR(task);
9494	else
9495		rpc_put_task_async(task);
9496	dprintk("<-- %s status=%d\n", __func__, ret);
9497	return ret;
9498}
9499
9500static int nfs4_proc_sequence(struct nfs_client *clp, const struct cred *cred)
9501{
9502	struct rpc_task *task;
9503	int ret;
9504
9505	task = _nfs41_proc_sequence(clp, cred, NULL, true);
9506	if (IS_ERR(task)) {
9507		ret = PTR_ERR(task);
9508		goto out;
9509	}
9510	ret = rpc_wait_for_completion_task(task);
9511	if (!ret)
9512		ret = task->tk_status;
9513	rpc_put_task(task);
9514out:
9515	dprintk("<-- %s status=%d\n", __func__, ret);
9516	return ret;
9517}
9518
9519struct nfs4_reclaim_complete_data {
9520	struct nfs_client *clp;
9521	struct nfs41_reclaim_complete_args arg;
9522	struct nfs41_reclaim_complete_res res;
9523};
9524
9525static void nfs4_reclaim_complete_prepare(struct rpc_task *task, void *data)
9526{
9527	struct nfs4_reclaim_complete_data *calldata = data;
9528
9529	nfs4_setup_sequence(calldata->clp,
9530			&calldata->arg.seq_args,
9531			&calldata->res.seq_res,
9532			task);
9533}
9534
9535static int nfs41_reclaim_complete_handle_errors(struct rpc_task *task, struct nfs_client *clp)
9536{
9537	switch(task->tk_status) {
9538	case 0:
9539		wake_up_all(&clp->cl_lock_waitq);
9540		fallthrough;
9541	case -NFS4ERR_COMPLETE_ALREADY:
9542	case -NFS4ERR_WRONG_CRED: /* What to do here? */
9543		break;
9544	case -NFS4ERR_DELAY:
9545		rpc_delay(task, NFS4_POLL_RETRY_MAX);
9546		fallthrough;
9547	case -NFS4ERR_RETRY_UNCACHED_REP:
9548	case -EACCES:
9549		dprintk("%s: failed to reclaim complete error %d for server %s, retrying\n",
9550			__func__, task->tk_status, clp->cl_hostname);
9551		return -EAGAIN;
9552	case -NFS4ERR_BADSESSION:
9553	case -NFS4ERR_DEADSESSION:
9554	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
9555		break;
9556	default:
9557		nfs4_schedule_lease_recovery(clp);
9558	}
9559	return 0;
9560}
9561
9562static void nfs4_reclaim_complete_done(struct rpc_task *task, void *data)
9563{
9564	struct nfs4_reclaim_complete_data *calldata = data;
9565	struct nfs_client *clp = calldata->clp;
9566	struct nfs4_sequence_res *res = &calldata->res.seq_res;
9567
9568	if (!nfs41_sequence_done(task, res))
9569		return;
9570
9571	trace_nfs4_reclaim_complete(clp, task->tk_status);
9572	if (nfs41_reclaim_complete_handle_errors(task, clp) == -EAGAIN) {
9573		rpc_restart_call_prepare(task);
9574		return;
9575	}
9576}
9577
9578static void nfs4_free_reclaim_complete_data(void *data)
9579{
9580	struct nfs4_reclaim_complete_data *calldata = data;
9581
9582	kfree(calldata);
9583}
9584
9585static const struct rpc_call_ops nfs4_reclaim_complete_call_ops = {
9586	.rpc_call_prepare = nfs4_reclaim_complete_prepare,
9587	.rpc_call_done = nfs4_reclaim_complete_done,
9588	.rpc_release = nfs4_free_reclaim_complete_data,
9589};
9590
9591/*
9592 * Issue a global reclaim complete.
9593 */
9594static int nfs41_proc_reclaim_complete(struct nfs_client *clp,
9595		const struct cred *cred)
9596{
9597	struct nfs4_reclaim_complete_data *calldata;
9598	struct rpc_message msg = {
9599		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_RECLAIM_COMPLETE],
9600		.rpc_cred = cred,
9601	};
9602	struct rpc_task_setup task_setup_data = {
9603		.rpc_client = clp->cl_rpcclient,
9604		.rpc_message = &msg,
9605		.callback_ops = &nfs4_reclaim_complete_call_ops,
9606		.flags = RPC_TASK_NO_ROUND_ROBIN,
9607	};
9608	int status = -ENOMEM;
9609
9610	calldata = kzalloc(sizeof(*calldata), GFP_NOFS);
9611	if (calldata == NULL)
9612		goto out;
9613	calldata->clp = clp;
9614	calldata->arg.one_fs = 0;
9615
9616	nfs4_init_sequence(&calldata->arg.seq_args, &calldata->res.seq_res, 0, 1);
9617	msg.rpc_argp = &calldata->arg;
9618	msg.rpc_resp = &calldata->res;
9619	task_setup_data.callback_data = calldata;
9620	status = nfs4_call_sync_custom(&task_setup_data);
9621out:
9622	dprintk("<-- %s status=%d\n", __func__, status);
9623	return status;
9624}
9625
9626static void
9627nfs4_layoutget_prepare(struct rpc_task *task, void *calldata)
9628{
9629	struct nfs4_layoutget *lgp = calldata;
9630	struct nfs_server *server = NFS_SERVER(lgp->args.inode);
9631
9632	nfs4_setup_sequence(server->nfs_client, &lgp->args.seq_args,
9633				&lgp->res.seq_res, task);
9634}
9635
9636static void nfs4_layoutget_done(struct rpc_task *task, void *calldata)
9637{
9638	struct nfs4_layoutget *lgp = calldata;
9639
9640	nfs41_sequence_process(task, &lgp->res.seq_res);
9641}
9642
9643static int
9644nfs4_layoutget_handle_exception(struct rpc_task *task,
9645		struct nfs4_layoutget *lgp, struct nfs4_exception *exception)
9646{
9647	struct inode *inode = lgp->args.inode;
9648	struct nfs_server *server = NFS_SERVER(inode);
9649	struct pnfs_layout_hdr *lo = lgp->lo;
9650	int nfs4err = task->tk_status;
9651	int err, status = 0;
9652	LIST_HEAD(head);
9653
9654	dprintk("--> %s tk_status => %d\n", __func__, -task->tk_status);
9655
9656	nfs4_sequence_free_slot(&lgp->res.seq_res);
9657
9658	exception->state = NULL;
9659	exception->stateid = NULL;
9660
9661	switch (nfs4err) {
9662	case 0:
9663		goto out;
9664
9665	/*
9666	 * NFS4ERR_LAYOUTUNAVAILABLE means we are not supposed to use pnfs
9667	 * on the file. set tk_status to -ENODATA to tell upper layer to
9668	 * retry go inband.
9669	 */
9670	case -NFS4ERR_LAYOUTUNAVAILABLE:
9671		status = -ENODATA;
9672		goto out;
9673	/*
9674	 * NFS4ERR_BADLAYOUT means the MDS cannot return a layout of
9675	 * length lgp->args.minlength != 0 (see RFC5661 section 18.43.3).
9676	 */
9677	case -NFS4ERR_BADLAYOUT:
9678		status = -EOVERFLOW;
9679		goto out;
9680	/*
9681	 * NFS4ERR_LAYOUTTRYLATER is a conflict with another client
9682	 * (or clients) writing to the same RAID stripe except when
9683	 * the minlength argument is 0 (see RFC5661 section 18.43.3).
9684	 *
9685	 * Treat it like we would RECALLCONFLICT -- we retry for a little
9686	 * while, and then eventually give up.
9687	 */
9688	case -NFS4ERR_LAYOUTTRYLATER:
9689		if (lgp->args.minlength == 0) {
9690			status = -EOVERFLOW;
9691			goto out;
9692		}
9693		status = -EBUSY;
9694		break;
9695	case -NFS4ERR_RECALLCONFLICT:
9696	case -NFS4ERR_RETURNCONFLICT:
9697		status = -ERECALLCONFLICT;
9698		break;
9699	case -NFS4ERR_DELEG_REVOKED:
9700	case -NFS4ERR_ADMIN_REVOKED:
9701	case -NFS4ERR_EXPIRED:
9702	case -NFS4ERR_BAD_STATEID:
9703		exception->timeout = 0;
9704		spin_lock(&inode->i_lock);
9705		/* If the open stateid was bad, then recover it. */
9706		if (!lo || test_bit(NFS_LAYOUT_INVALID_STID, &lo->plh_flags) ||
9707		    !nfs4_stateid_match_other(&lgp->args.stateid, &lo->plh_stateid)) {
9708			spin_unlock(&inode->i_lock);
9709			exception->state = lgp->args.ctx->state;
9710			exception->stateid = &lgp->args.stateid;
9711			break;
9712		}
9713
9714		/*
9715		 * Mark the bad layout state as invalid, then retry
9716		 */
9717		pnfs_mark_layout_stateid_invalid(lo, &head);
9718		spin_unlock(&inode->i_lock);
9719		nfs_commit_inode(inode, 0);
9720		pnfs_free_lseg_list(&head);
9721		status = -EAGAIN;
9722		goto out;
9723	}
9724
9725	err = nfs4_handle_exception(server, nfs4err, exception);
9726	if (!status) {
9727		if (exception->retry)
9728			status = -EAGAIN;
9729		else
9730			status = err;
9731	}
9732out:
9733	return status;
9734}
9735
9736size_t max_response_pages(struct nfs_server *server)
9737{
9738	u32 max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz;
9739	return nfs_page_array_len(0, max_resp_sz);
9740}
9741
9742static void nfs4_layoutget_release(void *calldata)
9743{
9744	struct nfs4_layoutget *lgp = calldata;
9745
9746	nfs4_sequence_free_slot(&lgp->res.seq_res);
9747	pnfs_layoutget_free(lgp);
9748}
9749
9750static const struct rpc_call_ops nfs4_layoutget_call_ops = {
9751	.rpc_call_prepare = nfs4_layoutget_prepare,
9752	.rpc_call_done = nfs4_layoutget_done,
9753	.rpc_release = nfs4_layoutget_release,
9754};
9755
9756struct pnfs_layout_segment *
9757nfs4_proc_layoutget(struct nfs4_layoutget *lgp,
9758		    struct nfs4_exception *exception)
9759{
9760	struct inode *inode = lgp->args.inode;
9761	struct nfs_server *server = NFS_SERVER(inode);
9762	struct rpc_task *task;
9763	struct rpc_message msg = {
9764		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTGET],
9765		.rpc_argp = &lgp->args,
9766		.rpc_resp = &lgp->res,
9767		.rpc_cred = lgp->cred,
9768	};
9769	struct rpc_task_setup task_setup_data = {
9770		.rpc_client = server->client,
9771		.rpc_message = &msg,
9772		.callback_ops = &nfs4_layoutget_call_ops,
9773		.callback_data = lgp,
9774		.flags = RPC_TASK_ASYNC | RPC_TASK_CRED_NOREF |
9775			 RPC_TASK_MOVEABLE,
9776	};
9777	struct pnfs_layout_segment *lseg = NULL;
9778	int status = 0;
9779
9780	nfs4_init_sequence(&lgp->args.seq_args, &lgp->res.seq_res, 0, 0);
9781	exception->retry = 0;
9782
9783	task = rpc_run_task(&task_setup_data);
9784	if (IS_ERR(task))
9785		return ERR_CAST(task);
9786
9787	status = rpc_wait_for_completion_task(task);
9788	if (status != 0)
9789		goto out;
9790
9791	if (task->tk_status < 0) {
9792		exception->retry = 1;
9793		status = nfs4_layoutget_handle_exception(task, lgp, exception);
9794	} else if (lgp->res.layoutp->len == 0) {
9795		exception->retry = 1;
9796		status = -EAGAIN;
9797		nfs4_update_delay(&exception->timeout);
9798	} else
9799		lseg = pnfs_layout_process(lgp);
9800out:
9801	trace_nfs4_layoutget(lgp->args.ctx,
9802			&lgp->args.range,
9803			&lgp->res.range,
9804			&lgp->res.stateid,
9805			status);
9806
9807	rpc_put_task(task);
9808	dprintk("<-- %s status=%d\n", __func__, status);
9809	if (status)
9810		return ERR_PTR(status);
9811	return lseg;
9812}
9813
9814static void
9815nfs4_layoutreturn_prepare(struct rpc_task *task, void *calldata)
9816{
9817	struct nfs4_layoutreturn *lrp = calldata;
9818
9819	nfs4_setup_sequence(lrp->clp,
9820			&lrp->args.seq_args,
9821			&lrp->res.seq_res,
9822			task);
9823	if (!pnfs_layout_is_valid(lrp->args.layout))
9824		rpc_exit(task, 0);
9825}
9826
9827static void nfs4_layoutreturn_done(struct rpc_task *task, void *calldata)
9828{
9829	struct nfs4_layoutreturn *lrp = calldata;
9830	struct nfs_server *server;
9831
9832	if (!nfs41_sequence_process(task, &lrp->res.seq_res))
9833		return;
9834
9835	/*
9836	 * Was there an RPC level error? Assume the call succeeded,
9837	 * and that we need to release the layout
9838	 */
9839	if (task->tk_rpc_status != 0 && RPC_WAS_SENT(task)) {
9840		lrp->res.lrs_present = 0;
9841		return;
9842	}
9843
9844	server = NFS_SERVER(lrp->args.inode);
9845	switch (task->tk_status) {
9846	case -NFS4ERR_OLD_STATEID:
9847		if (nfs4_layout_refresh_old_stateid(&lrp->args.stateid,
9848					&lrp->args.range,
9849					lrp->args.inode))
9850			goto out_restart;
9851		fallthrough;
9852	default:
9853		task->tk_status = 0;
9854		fallthrough;
9855	case 0:
9856		break;
9857	case -NFS4ERR_DELAY:
9858		if (nfs4_async_handle_error(task, server, NULL, NULL) != -EAGAIN)
9859			break;
9860		goto out_restart;
9861	}
9862	return;
9863out_restart:
9864	task->tk_status = 0;
9865	nfs4_sequence_free_slot(&lrp->res.seq_res);
9866	rpc_restart_call_prepare(task);
9867}
9868
9869static void nfs4_layoutreturn_release(void *calldata)
9870{
9871	struct nfs4_layoutreturn *lrp = calldata;
9872	struct pnfs_layout_hdr *lo = lrp->args.layout;
9873
9874	pnfs_layoutreturn_free_lsegs(lo, &lrp->args.stateid, &lrp->args.range,
9875			lrp->res.lrs_present ? &lrp->res.stateid : NULL);
9876	nfs4_sequence_free_slot(&lrp->res.seq_res);
9877	if (lrp->ld_private.ops && lrp->ld_private.ops->free)
9878		lrp->ld_private.ops->free(&lrp->ld_private);
9879	pnfs_put_layout_hdr(lrp->args.layout);
9880	nfs_iput_and_deactive(lrp->inode);
9881	put_cred(lrp->cred);
9882	kfree(calldata);
9883}
9884
9885static const struct rpc_call_ops nfs4_layoutreturn_call_ops = {
9886	.rpc_call_prepare = nfs4_layoutreturn_prepare,
9887	.rpc_call_done = nfs4_layoutreturn_done,
9888	.rpc_release = nfs4_layoutreturn_release,
9889};
9890
9891int nfs4_proc_layoutreturn(struct nfs4_layoutreturn *lrp, bool sync)
9892{
9893	struct rpc_task *task;
9894	struct rpc_message msg = {
9895		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTRETURN],
9896		.rpc_argp = &lrp->args,
9897		.rpc_resp = &lrp->res,
9898		.rpc_cred = lrp->cred,
9899	};
9900	struct rpc_task_setup task_setup_data = {
9901		.rpc_client = NFS_SERVER(lrp->args.inode)->client,
9902		.rpc_message = &msg,
9903		.callback_ops = &nfs4_layoutreturn_call_ops,
9904		.callback_data = lrp,
9905		.flags = RPC_TASK_MOVEABLE,
9906	};
9907	int status = 0;
9908
9909	nfs4_state_protect(NFS_SERVER(lrp->args.inode)->nfs_client,
9910			NFS_SP4_MACH_CRED_PNFS_CLEANUP,
9911			&task_setup_data.rpc_client, &msg);
9912
9913	lrp->inode = nfs_igrab_and_active(lrp->args.inode);
9914	if (!sync) {
9915		if (!lrp->inode) {
9916			nfs4_layoutreturn_release(lrp);
9917			return -EAGAIN;
9918		}
9919		task_setup_data.flags |= RPC_TASK_ASYNC;
9920	}
9921	if (!lrp->inode)
9922		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9923				   1);
9924	else
9925		nfs4_init_sequence(&lrp->args.seq_args, &lrp->res.seq_res, 1,
9926				   0);
9927	task = rpc_run_task(&task_setup_data);
9928	if (IS_ERR(task))
9929		return PTR_ERR(task);
9930	if (sync)
9931		status = task->tk_status;
9932	trace_nfs4_layoutreturn(lrp->args.inode, &lrp->args.stateid, status);
9933	dprintk("<-- %s status=%d\n", __func__, status);
9934	rpc_put_task(task);
9935	return status;
9936}
9937
9938static int
9939_nfs4_proc_getdeviceinfo(struct nfs_server *server,
9940		struct pnfs_device *pdev,
9941		const struct cred *cred)
9942{
9943	struct nfs4_getdeviceinfo_args args = {
9944		.pdev = pdev,
9945		.notify_types = NOTIFY_DEVICEID4_CHANGE |
9946			NOTIFY_DEVICEID4_DELETE,
9947	};
9948	struct nfs4_getdeviceinfo_res res = {
9949		.pdev = pdev,
9950	};
9951	struct rpc_message msg = {
9952		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_GETDEVICEINFO],
9953		.rpc_argp = &args,
9954		.rpc_resp = &res,
9955		.rpc_cred = cred,
9956	};
9957	int status;
9958
9959	status = nfs4_call_sync(server->client, server, &msg, &args.seq_args, &res.seq_res, 0);
9960	if (res.notification & ~args.notify_types)
9961		dprintk("%s: unsupported notification\n", __func__);
9962	if (res.notification != args.notify_types)
9963		pdev->nocache = 1;
9964
9965	trace_nfs4_getdeviceinfo(server, &pdev->dev_id, status);
9966
9967	dprintk("<-- %s status=%d\n", __func__, status);
9968
9969	return status;
9970}
9971
9972int nfs4_proc_getdeviceinfo(struct nfs_server *server,
9973		struct pnfs_device *pdev,
9974		const struct cred *cred)
9975{
9976	struct nfs4_exception exception = { };
9977	int err;
9978
9979	do {
9980		err = nfs4_handle_exception(server,
9981					_nfs4_proc_getdeviceinfo(server, pdev, cred),
9982					&exception);
9983	} while (exception.retry);
9984	return err;
9985}
9986EXPORT_SYMBOL_GPL(nfs4_proc_getdeviceinfo);
9987
9988static void nfs4_layoutcommit_prepare(struct rpc_task *task, void *calldata)
9989{
9990	struct nfs4_layoutcommit_data *data = calldata;
9991	struct nfs_server *server = NFS_SERVER(data->args.inode);
9992
9993	nfs4_setup_sequence(server->nfs_client,
9994			&data->args.seq_args,
9995			&data->res.seq_res,
9996			task);
9997}
9998
9999static void
10000nfs4_layoutcommit_done(struct rpc_task *task, void *calldata)
10001{
10002	struct nfs4_layoutcommit_data *data = calldata;
10003	struct nfs_server *server = NFS_SERVER(data->args.inode);
10004
10005	if (!nfs41_sequence_done(task, &data->res.seq_res))
10006		return;
10007
10008	switch (task->tk_status) { /* Just ignore these failures */
10009	case -NFS4ERR_DELEG_REVOKED: /* layout was recalled */
10010	case -NFS4ERR_BADIOMODE:     /* no IOMODE_RW layout for range */
10011	case -NFS4ERR_BADLAYOUT:     /* no layout */
10012	case -NFS4ERR_GRACE:	    /* loca_recalim always false */
10013		task->tk_status = 0;
10014		break;
10015	case 0:
10016		break;
10017	default:
10018		if (nfs4_async_handle_error(task, server, NULL, NULL) == -EAGAIN) {
10019			rpc_restart_call_prepare(task);
10020			return;
10021		}
10022	}
10023}
10024
10025static void nfs4_layoutcommit_release(void *calldata)
10026{
10027	struct nfs4_layoutcommit_data *data = calldata;
10028
10029	pnfs_cleanup_layoutcommit(data);
10030	nfs_post_op_update_inode_force_wcc(data->args.inode,
10031					   data->res.fattr);
10032	put_cred(data->cred);
10033	nfs_iput_and_deactive(data->inode);
10034	kfree(data);
10035}
10036
10037static const struct rpc_call_ops nfs4_layoutcommit_ops = {
10038	.rpc_call_prepare = nfs4_layoutcommit_prepare,
10039	.rpc_call_done = nfs4_layoutcommit_done,
10040	.rpc_release = nfs4_layoutcommit_release,
10041};
10042
10043int
10044nfs4_proc_layoutcommit(struct nfs4_layoutcommit_data *data, bool sync)
10045{
10046	struct rpc_message msg = {
10047		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTCOMMIT],
10048		.rpc_argp = &data->args,
10049		.rpc_resp = &data->res,
10050		.rpc_cred = data->cred,
10051	};
10052	struct rpc_task_setup task_setup_data = {
10053		.task = &data->task,
10054		.rpc_client = NFS_CLIENT(data->args.inode),
10055		.rpc_message = &msg,
10056		.callback_ops = &nfs4_layoutcommit_ops,
10057		.callback_data = data,
10058		.flags = RPC_TASK_MOVEABLE,
10059	};
10060	struct rpc_task *task;
10061	int status = 0;
10062
10063	dprintk("NFS: initiating layoutcommit call. sync %d "
10064		"lbw: %llu inode %lu\n", sync,
10065		data->args.lastbytewritten,
10066		data->args.inode->i_ino);
10067
10068	if (!sync) {
10069		data->inode = nfs_igrab_and_active(data->args.inode);
10070		if (data->inode == NULL) {
10071			nfs4_layoutcommit_release(data);
10072			return -EAGAIN;
10073		}
10074		task_setup_data.flags = RPC_TASK_ASYNC;
10075	}
10076	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, 0);
10077	task = rpc_run_task(&task_setup_data);
10078	if (IS_ERR(task))
10079		return PTR_ERR(task);
10080	if (sync)
10081		status = task->tk_status;
10082	trace_nfs4_layoutcommit(data->args.inode, &data->args.stateid, status);
10083	dprintk("%s: status %d\n", __func__, status);
10084	rpc_put_task(task);
10085	return status;
10086}
10087
10088/*
10089 * Use the state managment nfs_client cl_rpcclient, which uses krb5i (if
10090 * possible) as per RFC3530bis and RFC5661 Security Considerations sections
10091 */
10092static int
10093_nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10094		    struct nfs_fsinfo *info,
10095		    struct nfs4_secinfo_flavors *flavors, bool use_integrity)
10096{
10097	struct nfs41_secinfo_no_name_args args = {
10098		.style = SECINFO_STYLE_CURRENT_FH,
10099	};
10100	struct nfs4_secinfo_res res = {
10101		.flavors = flavors,
10102	};
10103	struct rpc_message msg = {
10104		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SECINFO_NO_NAME],
10105		.rpc_argp = &args,
10106		.rpc_resp = &res,
10107	};
10108	struct nfs4_call_sync_data data = {
10109		.seq_server = server,
10110		.seq_args = &args.seq_args,
10111		.seq_res = &res.seq_res,
10112	};
10113	struct rpc_task_setup task_setup = {
10114		.rpc_client = server->client,
10115		.rpc_message = &msg,
10116		.callback_ops = server->nfs_client->cl_mvops->call_sync_ops,
10117		.callback_data = &data,
10118		.flags = RPC_TASK_NO_ROUND_ROBIN,
10119	};
10120	const struct cred *cred = NULL;
10121	int status;
10122
10123	if (use_integrity) {
10124		task_setup.rpc_client = server->nfs_client->cl_rpcclient;
10125
10126		cred = nfs4_get_clid_cred(server->nfs_client);
10127		msg.rpc_cred = cred;
10128	}
10129
10130	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 0);
10131	status = nfs4_call_sync_custom(&task_setup);
10132	dprintk("<-- %s status=%d\n", __func__, status);
10133
10134	put_cred(cred);
10135
10136	return status;
10137}
10138
10139static int
10140nfs41_proc_secinfo_no_name(struct nfs_server *server, struct nfs_fh *fhandle,
10141			   struct nfs_fsinfo *info, struct nfs4_secinfo_flavors *flavors)
10142{
10143	struct nfs4_exception exception = {
10144		.interruptible = true,
10145	};
10146	int err;
10147	do {
10148		/* first try using integrity protection */
10149		err = -NFS4ERR_WRONGSEC;
10150
10151		/* try to use integrity protection with machine cred */
10152		if (_nfs4_is_integrity_protected(server->nfs_client))
10153			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10154							  flavors, true);
10155
10156		/*
10157		 * if unable to use integrity protection, or SECINFO with
10158		 * integrity protection returns NFS4ERR_WRONGSEC (which is
10159		 * disallowed by spec, but exists in deployed servers) use
10160		 * the current filesystem's rpc_client and the user cred.
10161		 */
10162		if (err == -NFS4ERR_WRONGSEC)
10163			err = _nfs41_proc_secinfo_no_name(server, fhandle, info,
10164							  flavors, false);
10165
10166		switch (err) {
10167		case 0:
10168		case -NFS4ERR_WRONGSEC:
10169		case -ENOTSUPP:
10170			goto out;
10171		default:
10172			err = nfs4_handle_exception(server, err, &exception);
10173		}
10174	} while (exception.retry);
10175out:
10176	return err;
10177}
10178
10179static int
10180nfs41_find_root_sec(struct nfs_server *server, struct nfs_fh *fhandle,
10181		    struct nfs_fsinfo *info)
10182{
10183	int err;
10184	struct page *page;
10185	rpc_authflavor_t flavor = RPC_AUTH_MAXFLAVOR;
10186	struct nfs4_secinfo_flavors *flavors;
10187	struct nfs4_secinfo4 *secinfo;
10188	int i;
10189
10190	page = alloc_page(GFP_KERNEL);
10191	if (!page) {
10192		err = -ENOMEM;
10193		goto out;
10194	}
10195
10196	flavors = page_address(page);
10197	err = nfs41_proc_secinfo_no_name(server, fhandle, info, flavors);
10198
10199	/*
10200	 * Fall back on "guess and check" method if
10201	 * the server doesn't support SECINFO_NO_NAME
10202	 */
10203	if (err == -NFS4ERR_WRONGSEC || err == -ENOTSUPP) {
10204		err = nfs4_find_root_sec(server, fhandle, info);
10205		goto out_freepage;
10206	}
10207	if (err)
10208		goto out_freepage;
10209
10210	for (i = 0; i < flavors->num_flavors; i++) {
10211		secinfo = &flavors->flavors[i];
10212
10213		switch (secinfo->flavor) {
10214		case RPC_AUTH_NULL:
10215		case RPC_AUTH_UNIX:
10216		case RPC_AUTH_GSS:
10217			flavor = rpcauth_get_pseudoflavor(secinfo->flavor,
10218					&secinfo->flavor_info);
10219			break;
10220		default:
10221			flavor = RPC_AUTH_MAXFLAVOR;
10222			break;
10223		}
10224
10225		if (!nfs_auth_info_match(&server->auth_info, flavor))
10226			flavor = RPC_AUTH_MAXFLAVOR;
10227
10228		if (flavor != RPC_AUTH_MAXFLAVOR) {
10229			err = nfs4_lookup_root_sec(server, fhandle,
10230						   info, flavor);
10231			if (!err)
10232				break;
10233		}
10234	}
10235
10236	if (flavor == RPC_AUTH_MAXFLAVOR)
10237		err = -EPERM;
10238
10239out_freepage:
10240	put_page(page);
10241	if (err == -EACCES)
10242		return -EPERM;
10243out:
10244	return err;
10245}
10246
10247static int _nfs41_test_stateid(struct nfs_server *server,
10248		nfs4_stateid *stateid,
10249		const struct cred *cred)
10250{
10251	int status;
10252	struct nfs41_test_stateid_args args = {
10253		.stateid = stateid,
10254	};
10255	struct nfs41_test_stateid_res res;
10256	struct rpc_message msg = {
10257		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_TEST_STATEID],
10258		.rpc_argp = &args,
10259		.rpc_resp = &res,
10260		.rpc_cred = cred,
10261	};
10262	struct rpc_clnt *rpc_client = server->client;
10263
10264	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10265		&rpc_client, &msg);
10266
10267	dprintk("NFS call  test_stateid %p\n", stateid);
10268	nfs4_init_sequence(&args.seq_args, &res.seq_res, 0, 1);
10269	status = nfs4_call_sync_sequence(rpc_client, server, &msg,
10270			&args.seq_args, &res.seq_res);
10271	if (status != NFS_OK) {
10272		dprintk("NFS reply test_stateid: failed, %d\n", status);
10273		return status;
10274	}
10275	dprintk("NFS reply test_stateid: succeeded, %d\n", -res.status);
10276	return -res.status;
10277}
10278
10279static void nfs4_handle_delay_or_session_error(struct nfs_server *server,
10280		int err, struct nfs4_exception *exception)
10281{
10282	exception->retry = 0;
10283	switch(err) {
10284	case -NFS4ERR_DELAY:
10285	case -NFS4ERR_RETRY_UNCACHED_REP:
10286		nfs4_handle_exception(server, err, exception);
10287		break;
10288	case -NFS4ERR_BADSESSION:
10289	case -NFS4ERR_BADSLOT:
10290	case -NFS4ERR_BAD_HIGH_SLOT:
10291	case -NFS4ERR_CONN_NOT_BOUND_TO_SESSION:
10292	case -NFS4ERR_DEADSESSION:
10293		nfs4_do_handle_exception(server, err, exception);
10294	}
10295}
10296
10297/**
10298 * nfs41_test_stateid - perform a TEST_STATEID operation
10299 *
10300 * @server: server / transport on which to perform the operation
10301 * @stateid: state ID to test
10302 * @cred: credential
10303 *
10304 * Returns NFS_OK if the server recognizes that "stateid" is valid.
10305 * Otherwise a negative NFS4ERR value is returned if the operation
10306 * failed or the state ID is not currently valid.
10307 */
10308static int nfs41_test_stateid(struct nfs_server *server,
10309		nfs4_stateid *stateid,
10310		const struct cred *cred)
10311{
10312	struct nfs4_exception exception = {
10313		.interruptible = true,
10314	};
10315	int err;
10316	do {
10317		err = _nfs41_test_stateid(server, stateid, cred);
10318		nfs4_handle_delay_or_session_error(server, err, &exception);
10319	} while (exception.retry);
10320	return err;
10321}
10322
10323struct nfs_free_stateid_data {
10324	struct nfs_server *server;
10325	struct nfs41_free_stateid_args args;
10326	struct nfs41_free_stateid_res res;
10327};
10328
10329static void nfs41_free_stateid_prepare(struct rpc_task *task, void *calldata)
10330{
10331	struct nfs_free_stateid_data *data = calldata;
10332	nfs4_setup_sequence(data->server->nfs_client,
10333			&data->args.seq_args,
10334			&data->res.seq_res,
10335			task);
10336}
10337
10338static void nfs41_free_stateid_done(struct rpc_task *task, void *calldata)
10339{
10340	struct nfs_free_stateid_data *data = calldata;
10341
10342	nfs41_sequence_done(task, &data->res.seq_res);
10343
10344	switch (task->tk_status) {
10345	case -NFS4ERR_DELAY:
10346		if (nfs4_async_handle_error(task, data->server, NULL, NULL) == -EAGAIN)
10347			rpc_restart_call_prepare(task);
10348	}
10349}
10350
10351static void nfs41_free_stateid_release(void *calldata)
10352{
10353	struct nfs_free_stateid_data *data = calldata;
10354	struct nfs_client *clp = data->server->nfs_client;
10355
10356	nfs_put_client(clp);
10357	kfree(calldata);
10358}
10359
10360static const struct rpc_call_ops nfs41_free_stateid_ops = {
10361	.rpc_call_prepare = nfs41_free_stateid_prepare,
10362	.rpc_call_done = nfs41_free_stateid_done,
10363	.rpc_release = nfs41_free_stateid_release,
10364};
10365
10366/**
10367 * nfs41_free_stateid - perform a FREE_STATEID operation
10368 *
10369 * @server: server / transport on which to perform the operation
10370 * @stateid: state ID to release
10371 * @cred: credential
10372 * @privileged: set to true if this call needs to be privileged
10373 *
10374 * Note: this function is always asynchronous.
10375 */
10376static int nfs41_free_stateid(struct nfs_server *server,
10377		const nfs4_stateid *stateid,
10378		const struct cred *cred,
10379		bool privileged)
10380{
10381	struct rpc_message msg = {
10382		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_FREE_STATEID],
10383		.rpc_cred = cred,
10384	};
10385	struct rpc_task_setup task_setup = {
10386		.rpc_client = server->client,
10387		.rpc_message = &msg,
10388		.callback_ops = &nfs41_free_stateid_ops,
10389		.flags = RPC_TASK_ASYNC | RPC_TASK_MOVEABLE,
10390	};
10391	struct nfs_free_stateid_data *data;
10392	struct rpc_task *task;
10393	struct nfs_client *clp = server->nfs_client;
10394
10395	if (!refcount_inc_not_zero(&clp->cl_count))
10396		return -EIO;
10397
10398	nfs4_state_protect(server->nfs_client, NFS_SP4_MACH_CRED_STATEID,
10399		&task_setup.rpc_client, &msg);
10400
10401	dprintk("NFS call  free_stateid %p\n", stateid);
10402	data = kmalloc(sizeof(*data), GFP_KERNEL);
10403	if (!data)
10404		return -ENOMEM;
10405	data->server = server;
10406	nfs4_stateid_copy(&data->args.stateid, stateid);
10407
10408	task_setup.callback_data = data;
10409
10410	msg.rpc_argp = &data->args;
10411	msg.rpc_resp = &data->res;
10412	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 1, privileged);
10413	task = rpc_run_task(&task_setup);
10414	if (IS_ERR(task))
10415		return PTR_ERR(task);
10416	rpc_put_task(task);
10417	return 0;
10418}
10419
10420static void
10421nfs41_free_lock_state(struct nfs_server *server, struct nfs4_lock_state *lsp)
10422{
10423	const struct cred *cred = lsp->ls_state->owner->so_cred;
10424
10425	nfs41_free_stateid(server, &lsp->ls_stateid, cred, false);
10426	nfs4_free_lock_state(server, lsp);
10427}
10428
10429static bool nfs41_match_stateid(const nfs4_stateid *s1,
10430		const nfs4_stateid *s2)
10431{
10432	if (s1->type != s2->type)
10433		return false;
10434
10435	if (memcmp(s1->other, s2->other, sizeof(s1->other)) != 0)
10436		return false;
10437
10438	if (s1->seqid == s2->seqid)
10439		return true;
10440
10441	return s1->seqid == 0 || s2->seqid == 0;
10442}
10443
10444#endif /* CONFIG_NFS_V4_1 */
10445
10446static bool nfs4_match_stateid(const nfs4_stateid *s1,
10447		const nfs4_stateid *s2)
10448{
10449	return nfs4_stateid_match(s1, s2);
10450}
10451
10452
10453static const struct nfs4_state_recovery_ops nfs40_reboot_recovery_ops = {
10454	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10455	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10456	.recover_open	= nfs4_open_reclaim,
10457	.recover_lock	= nfs4_lock_reclaim,
10458	.establish_clid = nfs4_init_clientid,
10459	.detect_trunking = nfs40_discover_server_trunking,
10460};
10461
10462#if defined(CONFIG_NFS_V4_1)
10463static const struct nfs4_state_recovery_ops nfs41_reboot_recovery_ops = {
10464	.owner_flag_bit = NFS_OWNER_RECLAIM_REBOOT,
10465	.state_flag_bit	= NFS_STATE_RECLAIM_REBOOT,
10466	.recover_open	= nfs4_open_reclaim,
10467	.recover_lock	= nfs4_lock_reclaim,
10468	.establish_clid = nfs41_init_clientid,
10469	.reclaim_complete = nfs41_proc_reclaim_complete,
10470	.detect_trunking = nfs41_discover_server_trunking,
10471};
10472#endif /* CONFIG_NFS_V4_1 */
10473
10474static const struct nfs4_state_recovery_ops nfs40_nograce_recovery_ops = {
10475	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10476	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10477	.recover_open	= nfs40_open_expired,
10478	.recover_lock	= nfs4_lock_expired,
10479	.establish_clid = nfs4_init_clientid,
10480};
10481
10482#if defined(CONFIG_NFS_V4_1)
10483static const struct nfs4_state_recovery_ops nfs41_nograce_recovery_ops = {
10484	.owner_flag_bit = NFS_OWNER_RECLAIM_NOGRACE,
10485	.state_flag_bit	= NFS_STATE_RECLAIM_NOGRACE,
10486	.recover_open	= nfs41_open_expired,
10487	.recover_lock	= nfs41_lock_expired,
10488	.establish_clid = nfs41_init_clientid,
10489};
10490#endif /* CONFIG_NFS_V4_1 */
10491
10492static const struct nfs4_state_maintenance_ops nfs40_state_renewal_ops = {
10493	.sched_state_renewal = nfs4_proc_async_renew,
10494	.get_state_renewal_cred = nfs4_get_renew_cred,
10495	.renew_lease = nfs4_proc_renew,
10496};
10497
10498#if defined(CONFIG_NFS_V4_1)
10499static const struct nfs4_state_maintenance_ops nfs41_state_renewal_ops = {
10500	.sched_state_renewal = nfs41_proc_async_sequence,
10501	.get_state_renewal_cred = nfs4_get_machine_cred,
10502	.renew_lease = nfs4_proc_sequence,
10503};
10504#endif
10505
10506static const struct nfs4_mig_recovery_ops nfs40_mig_recovery_ops = {
10507	.get_locations = _nfs40_proc_get_locations,
10508	.fsid_present = _nfs40_proc_fsid_present,
10509};
10510
10511#if defined(CONFIG_NFS_V4_1)
10512static const struct nfs4_mig_recovery_ops nfs41_mig_recovery_ops = {
10513	.get_locations = _nfs41_proc_get_locations,
10514	.fsid_present = _nfs41_proc_fsid_present,
10515};
10516#endif	/* CONFIG_NFS_V4_1 */
10517
10518static const struct nfs4_minor_version_ops nfs_v4_0_minor_ops = {
10519	.minor_version = 0,
10520	.init_caps = NFS_CAP_READDIRPLUS
10521		| NFS_CAP_ATOMIC_OPEN
10522		| NFS_CAP_POSIX_LOCK,
10523	.init_client = nfs40_init_client,
10524	.shutdown_client = nfs40_shutdown_client,
10525	.match_stateid = nfs4_match_stateid,
10526	.find_root_sec = nfs4_find_root_sec,
10527	.free_lock_state = nfs4_release_lockowner,
10528	.test_and_free_expired = nfs40_test_and_free_expired_stateid,
10529	.alloc_seqid = nfs_alloc_seqid,
10530	.call_sync_ops = &nfs40_call_sync_ops,
10531	.reboot_recovery_ops = &nfs40_reboot_recovery_ops,
10532	.nograce_recovery_ops = &nfs40_nograce_recovery_ops,
10533	.state_renewal_ops = &nfs40_state_renewal_ops,
10534	.mig_recovery_ops = &nfs40_mig_recovery_ops,
10535};
10536
10537#if defined(CONFIG_NFS_V4_1)
10538static struct nfs_seqid *
10539nfs_alloc_no_seqid(struct nfs_seqid_counter *arg1, gfp_t arg2)
10540{
10541	return NULL;
10542}
10543
10544static const struct nfs4_minor_version_ops nfs_v4_1_minor_ops = {
10545	.minor_version = 1,
10546	.init_caps = NFS_CAP_READDIRPLUS
10547		| NFS_CAP_ATOMIC_OPEN
10548		| NFS_CAP_POSIX_LOCK
10549		| NFS_CAP_STATEID_NFSV41
10550		| NFS_CAP_ATOMIC_OPEN_V1
10551		| NFS_CAP_LGOPEN
10552		| NFS_CAP_MOVEABLE,
10553	.init_client = nfs41_init_client,
10554	.shutdown_client = nfs41_shutdown_client,
10555	.match_stateid = nfs41_match_stateid,
10556	.find_root_sec = nfs41_find_root_sec,
10557	.free_lock_state = nfs41_free_lock_state,
10558	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10559	.alloc_seqid = nfs_alloc_no_seqid,
10560	.session_trunk = nfs4_test_session_trunk,
10561	.call_sync_ops = &nfs41_call_sync_ops,
10562	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10563	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10564	.state_renewal_ops = &nfs41_state_renewal_ops,
10565	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10566};
10567#endif
10568
10569#if defined(CONFIG_NFS_V4_2)
10570static const struct nfs4_minor_version_ops nfs_v4_2_minor_ops = {
10571	.minor_version = 2,
10572	.init_caps = NFS_CAP_READDIRPLUS
10573		| NFS_CAP_ATOMIC_OPEN
10574		| NFS_CAP_POSIX_LOCK
10575		| NFS_CAP_STATEID_NFSV41
10576		| NFS_CAP_ATOMIC_OPEN_V1
10577		| NFS_CAP_LGOPEN
10578		| NFS_CAP_ALLOCATE
10579		| NFS_CAP_COPY
10580		| NFS_CAP_OFFLOAD_CANCEL
10581		| NFS_CAP_COPY_NOTIFY
10582		| NFS_CAP_DEALLOCATE
10583		| NFS_CAP_SEEK
10584		| NFS_CAP_LAYOUTSTATS
10585		| NFS_CAP_CLONE
10586		| NFS_CAP_LAYOUTERROR
10587		| NFS_CAP_READ_PLUS
10588		| NFS_CAP_MOVEABLE,
10589	.init_client = nfs41_init_client,
10590	.shutdown_client = nfs41_shutdown_client,
10591	.match_stateid = nfs41_match_stateid,
10592	.find_root_sec = nfs41_find_root_sec,
10593	.free_lock_state = nfs41_free_lock_state,
10594	.call_sync_ops = &nfs41_call_sync_ops,
10595	.test_and_free_expired = nfs41_test_and_free_expired_stateid,
10596	.alloc_seqid = nfs_alloc_no_seqid,
10597	.session_trunk = nfs4_test_session_trunk,
10598	.reboot_recovery_ops = &nfs41_reboot_recovery_ops,
10599	.nograce_recovery_ops = &nfs41_nograce_recovery_ops,
10600	.state_renewal_ops = &nfs41_state_renewal_ops,
10601	.mig_recovery_ops = &nfs41_mig_recovery_ops,
10602};
10603#endif
10604
10605const struct nfs4_minor_version_ops *nfs_v4_minor_ops[] = {
10606	[0] = &nfs_v4_0_minor_ops,
10607#if defined(CONFIG_NFS_V4_1)
10608	[1] = &nfs_v4_1_minor_ops,
10609#endif
10610#if defined(CONFIG_NFS_V4_2)
10611	[2] = &nfs_v4_2_minor_ops,
10612#endif
10613};
10614
10615static ssize_t nfs4_listxattr(struct dentry *dentry, char *list, size_t size)
10616{
10617	ssize_t error, error2, error3;
10618	size_t left = size;
10619
10620	error = generic_listxattr(dentry, list, left);
10621	if (error < 0)
10622		return error;
10623	if (list) {
10624		list += error;
10625		left -= error;
10626	}
10627
10628	error2 = nfs4_listxattr_nfs4_label(d_inode(dentry), list, left);
10629	if (error2 < 0)
10630		return error2;
10631
10632	if (list) {
10633		list += error2;
10634		left -= error2;
10635	}
10636
10637	error3 = nfs4_listxattr_nfs4_user(d_inode(dentry), list, left);
10638	if (error3 < 0)
10639		return error3;
10640
10641	error += error2 + error3;
10642	if (size && error > size)
10643		return -ERANGE;
10644	return error;
10645}
10646
10647static void nfs4_enable_swap(struct inode *inode)
10648{
10649	/* The state manager thread must always be running.
10650	 * It will notice the client is a swapper, and stay put.
10651	 */
10652	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10653
10654	nfs4_schedule_state_manager(clp);
10655}
10656
10657static void nfs4_disable_swap(struct inode *inode)
10658{
10659	/* The state manager thread will now exit once it is
10660	 * woken.
10661	 */
10662	struct nfs_client *clp = NFS_SERVER(inode)->nfs_client;
10663
10664	set_bit(NFS4CLNT_RUN_MANAGER, &clp->cl_state);
10665	clear_bit(NFS4CLNT_MANAGER_AVAILABLE, &clp->cl_state);
10666	wake_up_var(&clp->cl_state);
10667}
10668
10669static const struct inode_operations nfs4_dir_inode_operations = {
10670	.create		= nfs_create,
10671	.lookup		= nfs_lookup,
10672	.atomic_open	= nfs_atomic_open,
10673	.link		= nfs_link,
10674	.unlink		= nfs_unlink,
10675	.symlink	= nfs_symlink,
10676	.mkdir		= nfs_mkdir,
10677	.rmdir		= nfs_rmdir,
10678	.mknod		= nfs_mknod,
10679	.rename		= nfs_rename,
10680	.permission	= nfs_permission,
10681	.getattr	= nfs_getattr,
10682	.setattr	= nfs_setattr,
10683	.listxattr	= nfs4_listxattr,
10684};
10685
10686static const struct inode_operations nfs4_file_inode_operations = {
10687	.permission	= nfs_permission,
10688	.getattr	= nfs_getattr,
10689	.setattr	= nfs_setattr,
10690	.listxattr	= nfs4_listxattr,
10691};
10692
10693const struct nfs_rpc_ops nfs_v4_clientops = {
10694	.version	= 4,			/* protocol version */
10695	.dentry_ops	= &nfs4_dentry_operations,
10696	.dir_inode_ops	= &nfs4_dir_inode_operations,
10697	.file_inode_ops	= &nfs4_file_inode_operations,
10698	.file_ops	= &nfs4_file_operations,
10699	.getroot	= nfs4_proc_get_root,
10700	.submount	= nfs4_submount,
10701	.try_get_tree	= nfs4_try_get_tree,
10702	.getattr	= nfs4_proc_getattr,
10703	.setattr	= nfs4_proc_setattr,
10704	.lookup		= nfs4_proc_lookup,
10705	.lookupp	= nfs4_proc_lookupp,
10706	.access		= nfs4_proc_access,
10707	.readlink	= nfs4_proc_readlink,
10708	.create		= nfs4_proc_create,
10709	.remove		= nfs4_proc_remove,
10710	.unlink_setup	= nfs4_proc_unlink_setup,
10711	.unlink_rpc_prepare = nfs4_proc_unlink_rpc_prepare,
10712	.unlink_done	= nfs4_proc_unlink_done,
10713	.rename_setup	= nfs4_proc_rename_setup,
10714	.rename_rpc_prepare = nfs4_proc_rename_rpc_prepare,
10715	.rename_done	= nfs4_proc_rename_done,
10716	.link		= nfs4_proc_link,
10717	.symlink	= nfs4_proc_symlink,
10718	.mkdir		= nfs4_proc_mkdir,
10719	.rmdir		= nfs4_proc_rmdir,
10720	.readdir	= nfs4_proc_readdir,
10721	.mknod		= nfs4_proc_mknod,
10722	.statfs		= nfs4_proc_statfs,
10723	.fsinfo		= nfs4_proc_fsinfo,
10724	.pathconf	= nfs4_proc_pathconf,
10725	.set_capabilities = nfs4_server_capabilities,
10726	.decode_dirent	= nfs4_decode_dirent,
10727	.pgio_rpc_prepare = nfs4_proc_pgio_rpc_prepare,
10728	.read_setup	= nfs4_proc_read_setup,
10729	.read_done	= nfs4_read_done,
10730	.write_setup	= nfs4_proc_write_setup,
10731	.write_done	= nfs4_write_done,
10732	.commit_setup	= nfs4_proc_commit_setup,
10733	.commit_rpc_prepare = nfs4_proc_commit_rpc_prepare,
10734	.commit_done	= nfs4_commit_done,
10735	.lock		= nfs4_proc_lock,
10736	.clear_acl_cache = nfs4_zap_acl_attr,
10737	.close_context  = nfs4_close_context,
10738	.open_context	= nfs4_atomic_open,
10739	.have_delegation = nfs4_have_delegation,
10740	.alloc_client	= nfs4_alloc_client,
10741	.init_client	= nfs4_init_client,
10742	.free_client	= nfs4_free_client,
10743	.create_server	= nfs4_create_server,
10744	.clone_server	= nfs_clone_server,
10745	.discover_trunking = nfs4_discover_trunking,
10746	.enable_swap	= nfs4_enable_swap,
10747	.disable_swap	= nfs4_disable_swap,
10748};
10749
10750static const struct xattr_handler nfs4_xattr_nfs4_acl_handler = {
10751	.name	= XATTR_NAME_NFSV4_ACL,
10752	.list	= nfs4_xattr_list_nfs4_acl,
10753	.get	= nfs4_xattr_get_nfs4_acl,
10754	.set	= nfs4_xattr_set_nfs4_acl,
10755};
10756
10757#if defined(CONFIG_NFS_V4_1)
10758static const struct xattr_handler nfs4_xattr_nfs4_dacl_handler = {
10759	.name	= XATTR_NAME_NFSV4_DACL,
10760	.list	= nfs4_xattr_list_nfs4_dacl,
10761	.get	= nfs4_xattr_get_nfs4_dacl,
10762	.set	= nfs4_xattr_set_nfs4_dacl,
10763};
10764
10765static const struct xattr_handler nfs4_xattr_nfs4_sacl_handler = {
10766	.name	= XATTR_NAME_NFSV4_SACL,
10767	.list	= nfs4_xattr_list_nfs4_sacl,
10768	.get	= nfs4_xattr_get_nfs4_sacl,
10769	.set	= nfs4_xattr_set_nfs4_sacl,
10770};
10771#endif
10772
10773#ifdef CONFIG_NFS_V4_2
10774static const struct xattr_handler nfs4_xattr_nfs4_user_handler = {
10775	.prefix	= XATTR_USER_PREFIX,
10776	.get	= nfs4_xattr_get_nfs4_user,
10777	.set	= nfs4_xattr_set_nfs4_user,
10778};
10779#endif
10780
10781const struct xattr_handler * const nfs4_xattr_handlers[] = {
10782	&nfs4_xattr_nfs4_acl_handler,
10783#if defined(CONFIG_NFS_V4_1)
10784	&nfs4_xattr_nfs4_dacl_handler,
10785	&nfs4_xattr_nfs4_sacl_handler,
10786#endif
10787#ifdef CONFIG_NFS_V4_SECURITY_LABEL
10788	&nfs4_xattr_nfs4_label_handler,
10789#endif
10790#ifdef CONFIG_NFS_V4_2
10791	&nfs4_xattr_nfs4_user_handler,
10792#endif
10793	NULL
10794};
10795