xref: /linux/fs/nfs/nfs42proc.c (revision ea8a163e02d6925773129e2dd86e419e491b791d)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2014 Anna Schumaker <Anna.Schumaker@Netapp.com>
4  */
5 #include <linux/fs.h>
6 #include <linux/sunrpc/addr.h>
7 #include <linux/sunrpc/sched.h>
8 #include <linux/nfs.h>
9 #include <linux/nfs3.h>
10 #include <linux/nfs4.h>
11 #include <linux/nfs_xdr.h>
12 #include <linux/nfs_fs.h>
13 #include "nfs4_fs.h"
14 #include "nfs42.h"
15 #include "iostat.h"
16 #include "pnfs.h"
17 #include "nfs4session.h"
18 #include "internal.h"
19 #include "delegation.h"
20 #include "nfs4trace.h"
21 
22 #define NFSDBG_FACILITY NFSDBG_PROC
23 static int nfs42_do_offload_cancel_async(struct file *dst, nfs4_stateid *std);
24 
25 static void nfs42_set_netaddr(struct file *filep, struct nfs42_netaddr *naddr)
26 {
27 	struct nfs_client *clp = (NFS_SERVER(file_inode(filep)))->nfs_client;
28 	unsigned short port = 2049;
29 
30 	rcu_read_lock();
31 	naddr->netid_len = scnprintf(naddr->netid,
32 					sizeof(naddr->netid), "%s",
33 					rpc_peeraddr2str(clp->cl_rpcclient,
34 					RPC_DISPLAY_NETID));
35 	naddr->addr_len = scnprintf(naddr->addr,
36 					sizeof(naddr->addr),
37 					"%s.%u.%u",
38 					rpc_peeraddr2str(clp->cl_rpcclient,
39 					RPC_DISPLAY_ADDR),
40 					port >> 8, port & 255);
41 	rcu_read_unlock();
42 }
43 
44 static int _nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep,
45 		struct nfs_lock_context *lock, loff_t offset, loff_t len)
46 {
47 	struct inode *inode = file_inode(filep);
48 	struct nfs_server *server = NFS_SERVER(inode);
49 	u32 bitmask[3];
50 	struct nfs42_falloc_args args = {
51 		.falloc_fh	= NFS_FH(inode),
52 		.falloc_offset	= offset,
53 		.falloc_length	= len,
54 		.falloc_bitmask	= bitmask,
55 	};
56 	struct nfs42_falloc_res res = {
57 		.falloc_server	= server,
58 	};
59 	int status;
60 
61 	msg->rpc_argp = &args;
62 	msg->rpc_resp = &res;
63 
64 	status = nfs4_set_rw_stateid(&args.falloc_stateid, lock->open_context,
65 			lock, FMODE_WRITE);
66 	if (status) {
67 		if (status == -EAGAIN)
68 			status = -NFS4ERR_BAD_STATEID;
69 		return status;
70 	}
71 
72 	memcpy(bitmask, server->cache_consistency_bitmask, sizeof(bitmask));
73 	if (server->attr_bitmask[1] & FATTR4_WORD1_SPACE_USED)
74 		bitmask[1] |= FATTR4_WORD1_SPACE_USED;
75 
76 	res.falloc_fattr = nfs_alloc_fattr();
77 	if (!res.falloc_fattr)
78 		return -ENOMEM;
79 
80 	status = nfs4_call_sync(server->client, server, msg,
81 				&args.seq_args, &res.seq_res, 0);
82 	if (status == 0)
83 		status = nfs_post_op_update_inode_force_wcc(inode,
84 							    res.falloc_fattr);
85 
86 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_ALLOCATE])
87 		trace_nfs4_fallocate(inode, &args, status);
88 	else
89 		trace_nfs4_deallocate(inode, &args, status);
90 	kfree(res.falloc_fattr);
91 	return status;
92 }
93 
94 static int nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep,
95 				loff_t offset, loff_t len)
96 {
97 	struct inode *inode = file_inode(filep);
98 	struct nfs_server *server = NFS_SERVER(inode);
99 	struct nfs4_exception exception = { };
100 	struct nfs_lock_context *lock;
101 	int err;
102 
103 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
104 	if (IS_ERR(lock))
105 		return PTR_ERR(lock);
106 
107 	exception.inode = inode;
108 	exception.state = lock->open_context->state;
109 
110 	err = nfs_sync_inode(inode);
111 	if (err)
112 		goto out;
113 
114 	do {
115 		err = _nfs42_proc_fallocate(msg, filep, lock, offset, len);
116 		if (err == -ENOTSUPP) {
117 			err = -EOPNOTSUPP;
118 			break;
119 		}
120 		err = nfs4_handle_exception(server, err, &exception);
121 	} while (exception.retry);
122 out:
123 	nfs_put_lock_context(lock);
124 	return err;
125 }
126 
127 int nfs42_proc_allocate(struct file *filep, loff_t offset, loff_t len)
128 {
129 	struct rpc_message msg = {
130 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ALLOCATE],
131 	};
132 	struct inode *inode = file_inode(filep);
133 	int err;
134 
135 	if (!nfs_server_capable(inode, NFS_CAP_ALLOCATE))
136 		return -EOPNOTSUPP;
137 
138 	inode_lock(inode);
139 
140 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
141 	if (err == -EOPNOTSUPP)
142 		NFS_SERVER(inode)->caps &= ~NFS_CAP_ALLOCATE;
143 
144 	inode_unlock(inode);
145 	return err;
146 }
147 
148 int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len)
149 {
150 	struct rpc_message msg = {
151 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DEALLOCATE],
152 	};
153 	struct inode *inode = file_inode(filep);
154 	int err;
155 
156 	if (!nfs_server_capable(inode, NFS_CAP_DEALLOCATE))
157 		return -EOPNOTSUPP;
158 
159 	inode_lock(inode);
160 
161 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
162 	if (err == 0)
163 		truncate_pagecache_range(inode, offset, (offset + len) -1);
164 	if (err == -EOPNOTSUPP)
165 		NFS_SERVER(inode)->caps &= ~NFS_CAP_DEALLOCATE;
166 
167 	inode_unlock(inode);
168 	return err;
169 }
170 
171 static int handle_async_copy(struct nfs42_copy_res *res,
172 			     struct nfs_server *dst_server,
173 			     struct nfs_server *src_server,
174 			     struct file *src,
175 			     struct file *dst,
176 			     nfs4_stateid *src_stateid,
177 			     bool *restart)
178 {
179 	struct nfs4_copy_state *copy, *tmp_copy;
180 	int status = NFS4_OK;
181 	bool found_pending = false;
182 	struct nfs_open_context *dst_ctx = nfs_file_open_context(dst);
183 	struct nfs_open_context *src_ctx = nfs_file_open_context(src);
184 
185 	copy = kzalloc(sizeof(struct nfs4_copy_state), GFP_NOFS);
186 	if (!copy)
187 		return -ENOMEM;
188 
189 	spin_lock(&dst_server->nfs_client->cl_lock);
190 	list_for_each_entry(tmp_copy,
191 				&dst_server->nfs_client->pending_cb_stateids,
192 				copies) {
193 		if (memcmp(&res->write_res.stateid, &tmp_copy->stateid,
194 				NFS4_STATEID_SIZE))
195 			continue;
196 		found_pending = true;
197 		list_del(&tmp_copy->copies);
198 		break;
199 	}
200 	if (found_pending) {
201 		spin_unlock(&dst_server->nfs_client->cl_lock);
202 		kfree(copy);
203 		copy = tmp_copy;
204 		goto out;
205 	}
206 
207 	memcpy(&copy->stateid, &res->write_res.stateid, NFS4_STATEID_SIZE);
208 	init_completion(&copy->completion);
209 	copy->parent_dst_state = dst_ctx->state;
210 	copy->parent_src_state = src_ctx->state;
211 
212 	list_add_tail(&copy->copies, &dst_server->ss_copies);
213 	spin_unlock(&dst_server->nfs_client->cl_lock);
214 
215 	if (dst_server != src_server) {
216 		spin_lock(&src_server->nfs_client->cl_lock);
217 		list_add_tail(&copy->src_copies, &src_server->ss_copies);
218 		spin_unlock(&src_server->nfs_client->cl_lock);
219 	}
220 
221 	status = wait_for_completion_interruptible(&copy->completion);
222 	spin_lock(&dst_server->nfs_client->cl_lock);
223 	list_del_init(&copy->copies);
224 	spin_unlock(&dst_server->nfs_client->cl_lock);
225 	if (dst_server != src_server) {
226 		spin_lock(&src_server->nfs_client->cl_lock);
227 		list_del_init(&copy->src_copies);
228 		spin_unlock(&src_server->nfs_client->cl_lock);
229 	}
230 	if (status == -ERESTARTSYS) {
231 		goto out_cancel;
232 	} else if (copy->flags || copy->error == NFS4ERR_PARTNER_NO_AUTH) {
233 		status = -EAGAIN;
234 		*restart = true;
235 		goto out_cancel;
236 	}
237 out:
238 	res->write_res.count = copy->count;
239 	memcpy(&res->write_res.verifier, &copy->verf, sizeof(copy->verf));
240 	status = -copy->error;
241 
242 out_free:
243 	kfree(copy);
244 	return status;
245 out_cancel:
246 	nfs42_do_offload_cancel_async(dst, &copy->stateid);
247 	if (!nfs42_files_from_same_server(src, dst))
248 		nfs42_do_offload_cancel_async(src, src_stateid);
249 	goto out_free;
250 }
251 
252 static int process_copy_commit(struct file *dst, loff_t pos_dst,
253 			       struct nfs42_copy_res *res)
254 {
255 	struct nfs_commitres cres;
256 	int status = -ENOMEM;
257 
258 	cres.verf = kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS);
259 	if (!cres.verf)
260 		goto out;
261 
262 	status = nfs4_proc_commit(dst, pos_dst, res->write_res.count, &cres);
263 	if (status)
264 		goto out_free;
265 	if (nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
266 				    &cres.verf->verifier)) {
267 		dprintk("commit verf differs from copy verf\n");
268 		status = -EAGAIN;
269 	}
270 out_free:
271 	kfree(cres.verf);
272 out:
273 	return status;
274 }
275 
276 /**
277  * nfs42_copy_dest_done - perform inode cache updates after clone/copy offload
278  * @inode: pointer to destination inode
279  * @pos: destination offset
280  * @len: copy length
281  *
282  * Punch a hole in the inode page cache, so that the NFS client will
283  * know to retrieve new data.
284  * Update the file size if necessary, and then mark the inode as having
285  * invalid cached values for change attribute, ctime, mtime and space used.
286  */
287 static void nfs42_copy_dest_done(struct inode *inode, loff_t pos, loff_t len)
288 {
289 	loff_t newsize = pos + len;
290 	loff_t end = newsize - 1;
291 
292 	truncate_pagecache_range(inode, pos, end);
293 	spin_lock(&inode->i_lock);
294 	if (newsize > i_size_read(inode))
295 		i_size_write(inode, newsize);
296 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
297 					     NFS_INO_INVALID_CTIME |
298 					     NFS_INO_INVALID_MTIME |
299 					     NFS_INO_INVALID_BLOCKS);
300 	spin_unlock(&inode->i_lock);
301 }
302 
303 static ssize_t _nfs42_proc_copy(struct file *src,
304 				struct nfs_lock_context *src_lock,
305 				struct file *dst,
306 				struct nfs_lock_context *dst_lock,
307 				struct nfs42_copy_args *args,
308 				struct nfs42_copy_res *res,
309 				struct nl4_server *nss,
310 				nfs4_stateid *cnr_stateid,
311 				bool *restart)
312 {
313 	struct rpc_message msg = {
314 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY],
315 		.rpc_argp = args,
316 		.rpc_resp = res,
317 	};
318 	struct inode *dst_inode = file_inode(dst);
319 	struct inode *src_inode = file_inode(src);
320 	struct nfs_server *dst_server = NFS_SERVER(dst_inode);
321 	struct nfs_server *src_server = NFS_SERVER(src_inode);
322 	loff_t pos_src = args->src_pos;
323 	loff_t pos_dst = args->dst_pos;
324 	size_t count = args->count;
325 	ssize_t status;
326 
327 	if (nss) {
328 		args->cp_src = nss;
329 		nfs4_stateid_copy(&args->src_stateid, cnr_stateid);
330 	} else {
331 		status = nfs4_set_rw_stateid(&args->src_stateid,
332 				src_lock->open_context, src_lock, FMODE_READ);
333 		if (status) {
334 			if (status == -EAGAIN)
335 				status = -NFS4ERR_BAD_STATEID;
336 			return status;
337 		}
338 	}
339 	status = nfs_filemap_write_and_wait_range(file_inode(src)->i_mapping,
340 			pos_src, pos_src + (loff_t)count - 1);
341 	if (status)
342 		return status;
343 
344 	status = nfs4_set_rw_stateid(&args->dst_stateid, dst_lock->open_context,
345 				     dst_lock, FMODE_WRITE);
346 	if (status) {
347 		if (status == -EAGAIN)
348 			status = -NFS4ERR_BAD_STATEID;
349 		return status;
350 	}
351 
352 	status = nfs_sync_inode(dst_inode);
353 	if (status)
354 		return status;
355 
356 	res->commit_res.verf = NULL;
357 	if (args->sync) {
358 		res->commit_res.verf =
359 			kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS);
360 		if (!res->commit_res.verf)
361 			return -ENOMEM;
362 	}
363 	set_bit(NFS_CLNT_SRC_SSC_COPY_STATE,
364 		&src_lock->open_context->state->flags);
365 	set_bit(NFS_CLNT_DST_SSC_COPY_STATE,
366 		&dst_lock->open_context->state->flags);
367 
368 	status = nfs4_call_sync(dst_server->client, dst_server, &msg,
369 				&args->seq_args, &res->seq_res, 0);
370 	trace_nfs4_copy(src_inode, dst_inode, args, res, nss, status);
371 	if (status == -ENOTSUPP)
372 		dst_server->caps &= ~NFS_CAP_COPY;
373 	if (status)
374 		goto out;
375 
376 	if (args->sync &&
377 		nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
378 				    &res->commit_res.verf->verifier)) {
379 		status = -EAGAIN;
380 		goto out;
381 	}
382 
383 	if (!res->synchronous) {
384 		status = handle_async_copy(res, dst_server, src_server, src,
385 				dst, &args->src_stateid, restart);
386 		if (status)
387 			goto out;
388 	}
389 
390 	if ((!res->synchronous || !args->sync) &&
391 			res->write_res.verifier.committed != NFS_FILE_SYNC) {
392 		status = process_copy_commit(dst, pos_dst, res);
393 		if (status)
394 			goto out;
395 	}
396 
397 	nfs42_copy_dest_done(dst_inode, pos_dst, res->write_res.count);
398 	nfs_invalidate_atime(src_inode);
399 	status = res->write_res.count;
400 out:
401 	if (args->sync)
402 		kfree(res->commit_res.verf);
403 	return status;
404 }
405 
406 ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
407 			struct file *dst, loff_t pos_dst, size_t count,
408 			struct nl4_server *nss,
409 			nfs4_stateid *cnr_stateid, bool sync)
410 {
411 	struct nfs_server *server = NFS_SERVER(file_inode(dst));
412 	struct nfs_lock_context *src_lock;
413 	struct nfs_lock_context *dst_lock;
414 	struct nfs42_copy_args args = {
415 		.src_fh		= NFS_FH(file_inode(src)),
416 		.src_pos	= pos_src,
417 		.dst_fh		= NFS_FH(file_inode(dst)),
418 		.dst_pos	= pos_dst,
419 		.count		= count,
420 		.sync		= sync,
421 	};
422 	struct nfs42_copy_res res;
423 	struct nfs4_exception src_exception = {
424 		.inode		= file_inode(src),
425 		.stateid	= &args.src_stateid,
426 	};
427 	struct nfs4_exception dst_exception = {
428 		.inode		= file_inode(dst),
429 		.stateid	= &args.dst_stateid,
430 	};
431 	ssize_t err, err2;
432 	bool restart = false;
433 
434 	src_lock = nfs_get_lock_context(nfs_file_open_context(src));
435 	if (IS_ERR(src_lock))
436 		return PTR_ERR(src_lock);
437 
438 	src_exception.state = src_lock->open_context->state;
439 
440 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst));
441 	if (IS_ERR(dst_lock)) {
442 		err = PTR_ERR(dst_lock);
443 		goto out_put_src_lock;
444 	}
445 
446 	dst_exception.state = dst_lock->open_context->state;
447 
448 	do {
449 		inode_lock(file_inode(dst));
450 		err = _nfs42_proc_copy(src, src_lock,
451 				dst, dst_lock,
452 				&args, &res,
453 				nss, cnr_stateid, &restart);
454 		inode_unlock(file_inode(dst));
455 
456 		if (err >= 0)
457 			break;
458 		if (err == -ENOTSUPP &&
459 				nfs42_files_from_same_server(src, dst)) {
460 			err = -EOPNOTSUPP;
461 			break;
462 		} else if (err == -EAGAIN) {
463 			if (!restart) {
464 				dst_exception.retry = 1;
465 				continue;
466 			}
467 			break;
468 		} else if (err == -NFS4ERR_OFFLOAD_NO_REQS && !args.sync) {
469 			args.sync = true;
470 			dst_exception.retry = 1;
471 			continue;
472 		} else if ((err == -ESTALE ||
473 				err == -NFS4ERR_OFFLOAD_DENIED ||
474 				err == -ENOTSUPP) &&
475 				!nfs42_files_from_same_server(src, dst)) {
476 			nfs42_do_offload_cancel_async(src, &args.src_stateid);
477 			err = -EOPNOTSUPP;
478 			break;
479 		}
480 
481 		err2 = nfs4_handle_exception(server, err, &src_exception);
482 		err  = nfs4_handle_exception(server, err, &dst_exception);
483 		if (!err)
484 			err = err2;
485 	} while (src_exception.retry || dst_exception.retry);
486 
487 	nfs_put_lock_context(dst_lock);
488 out_put_src_lock:
489 	nfs_put_lock_context(src_lock);
490 	return err;
491 }
492 
493 struct nfs42_offloadcancel_data {
494 	struct nfs_server *seq_server;
495 	struct nfs42_offload_status_args args;
496 	struct nfs42_offload_status_res res;
497 };
498 
499 static void nfs42_offload_cancel_prepare(struct rpc_task *task, void *calldata)
500 {
501 	struct nfs42_offloadcancel_data *data = calldata;
502 
503 	nfs4_setup_sequence(data->seq_server->nfs_client,
504 				&data->args.osa_seq_args,
505 				&data->res.osr_seq_res, task);
506 }
507 
508 static void nfs42_offload_cancel_done(struct rpc_task *task, void *calldata)
509 {
510 	struct nfs42_offloadcancel_data *data = calldata;
511 
512 	trace_nfs4_offload_cancel(&data->args, task->tk_status);
513 	nfs41_sequence_done(task, &data->res.osr_seq_res);
514 	if (task->tk_status &&
515 		nfs4_async_handle_error(task, data->seq_server, NULL,
516 			NULL) == -EAGAIN)
517 		rpc_restart_call_prepare(task);
518 }
519 
520 static void nfs42_free_offloadcancel_data(void *data)
521 {
522 	kfree(data);
523 }
524 
525 static const struct rpc_call_ops nfs42_offload_cancel_ops = {
526 	.rpc_call_prepare = nfs42_offload_cancel_prepare,
527 	.rpc_call_done = nfs42_offload_cancel_done,
528 	.rpc_release = nfs42_free_offloadcancel_data,
529 };
530 
531 static int nfs42_do_offload_cancel_async(struct file *dst,
532 					 nfs4_stateid *stateid)
533 {
534 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
535 	struct nfs42_offloadcancel_data *data = NULL;
536 	struct nfs_open_context *ctx = nfs_file_open_context(dst);
537 	struct rpc_task *task;
538 	struct rpc_message msg = {
539 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OFFLOAD_CANCEL],
540 		.rpc_cred = ctx->cred,
541 	};
542 	struct rpc_task_setup task_setup_data = {
543 		.rpc_client = dst_server->client,
544 		.rpc_message = &msg,
545 		.callback_ops = &nfs42_offload_cancel_ops,
546 		.workqueue = nfsiod_workqueue,
547 		.flags = RPC_TASK_ASYNC,
548 	};
549 	int status;
550 
551 	if (!(dst_server->caps & NFS_CAP_OFFLOAD_CANCEL))
552 		return -EOPNOTSUPP;
553 
554 	data = kzalloc(sizeof(struct nfs42_offloadcancel_data), GFP_NOFS);
555 	if (data == NULL)
556 		return -ENOMEM;
557 
558 	data->seq_server = dst_server;
559 	data->args.osa_src_fh = NFS_FH(file_inode(dst));
560 	memcpy(&data->args.osa_stateid, stateid,
561 		sizeof(data->args.osa_stateid));
562 	msg.rpc_argp = &data->args;
563 	msg.rpc_resp = &data->res;
564 	task_setup_data.callback_data = data;
565 	nfs4_init_sequence(&data->args.osa_seq_args, &data->res.osr_seq_res,
566 			   1, 0);
567 	task = rpc_run_task(&task_setup_data);
568 	if (IS_ERR(task))
569 		return PTR_ERR(task);
570 	status = rpc_wait_for_completion_task(task);
571 	if (status == -ENOTSUPP)
572 		dst_server->caps &= ~NFS_CAP_OFFLOAD_CANCEL;
573 	rpc_put_task(task);
574 	return status;
575 }
576 
577 static int _nfs42_proc_copy_notify(struct file *src, struct file *dst,
578 				   struct nfs42_copy_notify_args *args,
579 				   struct nfs42_copy_notify_res *res)
580 {
581 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
582 	struct rpc_message msg = {
583 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY_NOTIFY],
584 		.rpc_argp = args,
585 		.rpc_resp = res,
586 	};
587 	int status;
588 	struct nfs_open_context *ctx;
589 	struct nfs_lock_context *l_ctx;
590 
591 	ctx = get_nfs_open_context(nfs_file_open_context(src));
592 	l_ctx = nfs_get_lock_context(ctx);
593 	if (IS_ERR(l_ctx))
594 		return PTR_ERR(l_ctx);
595 
596 	status = nfs4_set_rw_stateid(&args->cna_src_stateid, ctx, l_ctx,
597 				     FMODE_READ);
598 	nfs_put_lock_context(l_ctx);
599 	if (status) {
600 		if (status == -EAGAIN)
601 			status = -NFS4ERR_BAD_STATEID;
602 		return status;
603 	}
604 
605 	status = nfs4_call_sync(src_server->client, src_server, &msg,
606 				&args->cna_seq_args, &res->cnr_seq_res, 0);
607 	trace_nfs4_copy_notify(file_inode(src), args, res, status);
608 	if (status == -ENOTSUPP)
609 		src_server->caps &= ~NFS_CAP_COPY_NOTIFY;
610 
611 	put_nfs_open_context(nfs_file_open_context(src));
612 	return status;
613 }
614 
615 int nfs42_proc_copy_notify(struct file *src, struct file *dst,
616 				struct nfs42_copy_notify_res *res)
617 {
618 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
619 	struct nfs42_copy_notify_args *args;
620 	struct nfs4_exception exception = {
621 		.inode = file_inode(src),
622 	};
623 	int status;
624 
625 	if (!(src_server->caps & NFS_CAP_COPY_NOTIFY))
626 		return -EOPNOTSUPP;
627 
628 	args = kzalloc(sizeof(struct nfs42_copy_notify_args), GFP_NOFS);
629 	if (args == NULL)
630 		return -ENOMEM;
631 
632 	args->cna_src_fh  = NFS_FH(file_inode(src)),
633 	args->cna_dst.nl4_type = NL4_NETADDR;
634 	nfs42_set_netaddr(dst, &args->cna_dst.u.nl4_addr);
635 	exception.stateid = &args->cna_src_stateid;
636 
637 	do {
638 		status = _nfs42_proc_copy_notify(src, dst, args, res);
639 		if (status == -ENOTSUPP) {
640 			status = -EOPNOTSUPP;
641 			goto out;
642 		}
643 		status = nfs4_handle_exception(src_server, status, &exception);
644 	} while (exception.retry);
645 
646 out:
647 	kfree(args);
648 	return status;
649 }
650 
651 static loff_t _nfs42_proc_llseek(struct file *filep,
652 		struct nfs_lock_context *lock, loff_t offset, int whence)
653 {
654 	struct inode *inode = file_inode(filep);
655 	struct nfs42_seek_args args = {
656 		.sa_fh		= NFS_FH(inode),
657 		.sa_offset	= offset,
658 		.sa_what	= (whence == SEEK_HOLE) ?
659 					NFS4_CONTENT_HOLE : NFS4_CONTENT_DATA,
660 	};
661 	struct nfs42_seek_res res;
662 	struct rpc_message msg = {
663 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEEK],
664 		.rpc_argp = &args,
665 		.rpc_resp = &res,
666 	};
667 	struct nfs_server *server = NFS_SERVER(inode);
668 	int status;
669 
670 	if (!nfs_server_capable(inode, NFS_CAP_SEEK))
671 		return -ENOTSUPP;
672 
673 	status = nfs4_set_rw_stateid(&args.sa_stateid, lock->open_context,
674 			lock, FMODE_READ);
675 	if (status) {
676 		if (status == -EAGAIN)
677 			status = -NFS4ERR_BAD_STATEID;
678 		return status;
679 	}
680 
681 	status = nfs_filemap_write_and_wait_range(inode->i_mapping,
682 			offset, LLONG_MAX);
683 	if (status)
684 		return status;
685 
686 	status = nfs4_call_sync(server->client, server, &msg,
687 				&args.seq_args, &res.seq_res, 0);
688 	trace_nfs4_llseek(inode, &args, &res, status);
689 	if (status == -ENOTSUPP)
690 		server->caps &= ~NFS_CAP_SEEK;
691 	if (status)
692 		return status;
693 
694 	if (whence == SEEK_DATA && res.sr_eof)
695 		return -NFS4ERR_NXIO;
696 	else
697 		return vfs_setpos(filep, res.sr_offset, inode->i_sb->s_maxbytes);
698 }
699 
700 loff_t nfs42_proc_llseek(struct file *filep, loff_t offset, int whence)
701 {
702 	struct nfs_server *server = NFS_SERVER(file_inode(filep));
703 	struct nfs4_exception exception = { };
704 	struct nfs_lock_context *lock;
705 	loff_t err;
706 
707 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
708 	if (IS_ERR(lock))
709 		return PTR_ERR(lock);
710 
711 	exception.inode = file_inode(filep);
712 	exception.state = lock->open_context->state;
713 
714 	do {
715 		err = _nfs42_proc_llseek(filep, lock, offset, whence);
716 		if (err >= 0)
717 			break;
718 		if (err == -ENOTSUPP) {
719 			err = -EOPNOTSUPP;
720 			break;
721 		}
722 		err = nfs4_handle_exception(server, err, &exception);
723 	} while (exception.retry);
724 
725 	nfs_put_lock_context(lock);
726 	return err;
727 }
728 
729 
730 static void
731 nfs42_layoutstat_prepare(struct rpc_task *task, void *calldata)
732 {
733 	struct nfs42_layoutstat_data *data = calldata;
734 	struct inode *inode = data->inode;
735 	struct nfs_server *server = NFS_SERVER(inode);
736 	struct pnfs_layout_hdr *lo;
737 
738 	spin_lock(&inode->i_lock);
739 	lo = NFS_I(inode)->layout;
740 	if (!pnfs_layout_is_valid(lo)) {
741 		spin_unlock(&inode->i_lock);
742 		rpc_exit(task, 0);
743 		return;
744 	}
745 	nfs4_stateid_copy(&data->args.stateid, &lo->plh_stateid);
746 	spin_unlock(&inode->i_lock);
747 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
748 			    &data->res.seq_res, task);
749 }
750 
751 static void
752 nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
753 {
754 	struct nfs42_layoutstat_data *data = calldata;
755 	struct inode *inode = data->inode;
756 	struct pnfs_layout_hdr *lo;
757 
758 	if (!nfs4_sequence_done(task, &data->res.seq_res))
759 		return;
760 
761 	switch (task->tk_status) {
762 	case 0:
763 		return;
764 	case -NFS4ERR_BADHANDLE:
765 	case -ESTALE:
766 		pnfs_destroy_layout(NFS_I(inode));
767 		break;
768 	case -NFS4ERR_EXPIRED:
769 	case -NFS4ERR_ADMIN_REVOKED:
770 	case -NFS4ERR_DELEG_REVOKED:
771 	case -NFS4ERR_STALE_STATEID:
772 	case -NFS4ERR_BAD_STATEID:
773 		spin_lock(&inode->i_lock);
774 		lo = NFS_I(inode)->layout;
775 		if (pnfs_layout_is_valid(lo) &&
776 		    nfs4_stateid_match(&data->args.stateid,
777 					     &lo->plh_stateid)) {
778 			LIST_HEAD(head);
779 
780 			/*
781 			 * Mark the bad layout state as invalid, then retry
782 			 * with the current stateid.
783 			 */
784 			pnfs_mark_layout_stateid_invalid(lo, &head);
785 			spin_unlock(&inode->i_lock);
786 			pnfs_free_lseg_list(&head);
787 			nfs_commit_inode(inode, 0);
788 		} else
789 			spin_unlock(&inode->i_lock);
790 		break;
791 	case -NFS4ERR_OLD_STATEID:
792 		spin_lock(&inode->i_lock);
793 		lo = NFS_I(inode)->layout;
794 		if (pnfs_layout_is_valid(lo) &&
795 		    nfs4_stateid_match_other(&data->args.stateid,
796 					&lo->plh_stateid)) {
797 			/* Do we need to delay before resending? */
798 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
799 						&data->args.stateid))
800 				rpc_delay(task, HZ);
801 			rpc_restart_call_prepare(task);
802 		}
803 		spin_unlock(&inode->i_lock);
804 		break;
805 	case -ENOTSUPP:
806 	case -EOPNOTSUPP:
807 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTSTATS;
808 	}
809 
810 	trace_nfs4_layoutstats(inode, &data->args.stateid, task->tk_status);
811 }
812 
813 static void
814 nfs42_layoutstat_release(void *calldata)
815 {
816 	struct nfs42_layoutstat_data *data = calldata;
817 	struct nfs42_layoutstat_devinfo *devinfo = data->args.devinfo;
818 	int i;
819 
820 	for (i = 0; i < data->args.num_dev; i++) {
821 		if (devinfo[i].ld_private.ops && devinfo[i].ld_private.ops->free)
822 			devinfo[i].ld_private.ops->free(&devinfo[i].ld_private);
823 	}
824 
825 	pnfs_put_layout_hdr(NFS_I(data->args.inode)->layout);
826 	smp_mb__before_atomic();
827 	clear_bit(NFS_INO_LAYOUTSTATS, &NFS_I(data->args.inode)->flags);
828 	smp_mb__after_atomic();
829 	nfs_iput_and_deactive(data->inode);
830 	kfree(data->args.devinfo);
831 	kfree(data);
832 }
833 
834 static const struct rpc_call_ops nfs42_layoutstat_ops = {
835 	.rpc_call_prepare = nfs42_layoutstat_prepare,
836 	.rpc_call_done = nfs42_layoutstat_done,
837 	.rpc_release = nfs42_layoutstat_release,
838 };
839 
840 int nfs42_proc_layoutstats_generic(struct nfs_server *server,
841 				   struct nfs42_layoutstat_data *data)
842 {
843 	struct rpc_message msg = {
844 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTSTATS],
845 		.rpc_argp = &data->args,
846 		.rpc_resp = &data->res,
847 	};
848 	struct rpc_task_setup task_setup = {
849 		.rpc_client = server->client,
850 		.rpc_message = &msg,
851 		.callback_ops = &nfs42_layoutstat_ops,
852 		.callback_data = data,
853 		.flags = RPC_TASK_ASYNC,
854 	};
855 	struct rpc_task *task;
856 
857 	data->inode = nfs_igrab_and_active(data->args.inode);
858 	if (!data->inode) {
859 		nfs42_layoutstat_release(data);
860 		return -EAGAIN;
861 	}
862 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
863 	task = rpc_run_task(&task_setup);
864 	if (IS_ERR(task))
865 		return PTR_ERR(task);
866 	rpc_put_task(task);
867 	return 0;
868 }
869 
870 static struct nfs42_layouterror_data *
871 nfs42_alloc_layouterror_data(struct pnfs_layout_segment *lseg, gfp_t gfp_flags)
872 {
873 	struct nfs42_layouterror_data *data;
874 	struct inode *inode = lseg->pls_layout->plh_inode;
875 
876 	data = kzalloc(sizeof(*data), gfp_flags);
877 	if (data) {
878 		data->args.inode = data->inode = nfs_igrab_and_active(inode);
879 		if (data->inode) {
880 			data->lseg = pnfs_get_lseg(lseg);
881 			if (data->lseg)
882 				return data;
883 			nfs_iput_and_deactive(data->inode);
884 		}
885 		kfree(data);
886 	}
887 	return NULL;
888 }
889 
890 static void
891 nfs42_free_layouterror_data(struct nfs42_layouterror_data *data)
892 {
893 	pnfs_put_lseg(data->lseg);
894 	nfs_iput_and_deactive(data->inode);
895 	kfree(data);
896 }
897 
898 static void
899 nfs42_layouterror_prepare(struct rpc_task *task, void *calldata)
900 {
901 	struct nfs42_layouterror_data *data = calldata;
902 	struct inode *inode = data->inode;
903 	struct nfs_server *server = NFS_SERVER(inode);
904 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
905 	unsigned i;
906 
907 	spin_lock(&inode->i_lock);
908 	if (!pnfs_layout_is_valid(lo)) {
909 		spin_unlock(&inode->i_lock);
910 		rpc_exit(task, 0);
911 		return;
912 	}
913 	for (i = 0; i < data->args.num_errors; i++)
914 		nfs4_stateid_copy(&data->args.errors[i].stateid,
915 				&lo->plh_stateid);
916 	spin_unlock(&inode->i_lock);
917 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
918 			    &data->res.seq_res, task);
919 }
920 
921 static void
922 nfs42_layouterror_done(struct rpc_task *task, void *calldata)
923 {
924 	struct nfs42_layouterror_data *data = calldata;
925 	struct inode *inode = data->inode;
926 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
927 
928 	if (!nfs4_sequence_done(task, &data->res.seq_res))
929 		return;
930 
931 	switch (task->tk_status) {
932 	case 0:
933 		return;
934 	case -NFS4ERR_BADHANDLE:
935 	case -ESTALE:
936 		pnfs_destroy_layout(NFS_I(inode));
937 		break;
938 	case -NFS4ERR_EXPIRED:
939 	case -NFS4ERR_ADMIN_REVOKED:
940 	case -NFS4ERR_DELEG_REVOKED:
941 	case -NFS4ERR_STALE_STATEID:
942 	case -NFS4ERR_BAD_STATEID:
943 		spin_lock(&inode->i_lock);
944 		if (pnfs_layout_is_valid(lo) &&
945 		    nfs4_stateid_match(&data->args.errors[0].stateid,
946 					     &lo->plh_stateid)) {
947 			LIST_HEAD(head);
948 
949 			/*
950 			 * Mark the bad layout state as invalid, then retry
951 			 * with the current stateid.
952 			 */
953 			pnfs_mark_layout_stateid_invalid(lo, &head);
954 			spin_unlock(&inode->i_lock);
955 			pnfs_free_lseg_list(&head);
956 			nfs_commit_inode(inode, 0);
957 		} else
958 			spin_unlock(&inode->i_lock);
959 		break;
960 	case -NFS4ERR_OLD_STATEID:
961 		spin_lock(&inode->i_lock);
962 		if (pnfs_layout_is_valid(lo) &&
963 		    nfs4_stateid_match_other(&data->args.errors[0].stateid,
964 					&lo->plh_stateid)) {
965 			/* Do we need to delay before resending? */
966 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
967 						&data->args.errors[0].stateid))
968 				rpc_delay(task, HZ);
969 			rpc_restart_call_prepare(task);
970 		}
971 		spin_unlock(&inode->i_lock);
972 		break;
973 	case -ENOTSUPP:
974 	case -EOPNOTSUPP:
975 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTERROR;
976 	}
977 
978 	trace_nfs4_layouterror(inode, &data->args.errors[0].stateid,
979 			       task->tk_status);
980 }
981 
982 static void
983 nfs42_layouterror_release(void *calldata)
984 {
985 	struct nfs42_layouterror_data *data = calldata;
986 
987 	nfs42_free_layouterror_data(data);
988 }
989 
990 static const struct rpc_call_ops nfs42_layouterror_ops = {
991 	.rpc_call_prepare = nfs42_layouterror_prepare,
992 	.rpc_call_done = nfs42_layouterror_done,
993 	.rpc_release = nfs42_layouterror_release,
994 };
995 
996 int nfs42_proc_layouterror(struct pnfs_layout_segment *lseg,
997 		const struct nfs42_layout_error *errors, size_t n)
998 {
999 	struct inode *inode = lseg->pls_layout->plh_inode;
1000 	struct nfs42_layouterror_data *data;
1001 	struct rpc_task *task;
1002 	struct rpc_message msg = {
1003 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTERROR],
1004 	};
1005 	struct rpc_task_setup task_setup = {
1006 		.rpc_message = &msg,
1007 		.callback_ops = &nfs42_layouterror_ops,
1008 		.flags = RPC_TASK_ASYNC,
1009 	};
1010 	unsigned int i;
1011 
1012 	if (!nfs_server_capable(inode, NFS_CAP_LAYOUTERROR))
1013 		return -EOPNOTSUPP;
1014 	if (n > NFS42_LAYOUTERROR_MAX)
1015 		return -EINVAL;
1016 	data = nfs42_alloc_layouterror_data(lseg, GFP_NOFS);
1017 	if (!data)
1018 		return -ENOMEM;
1019 	for (i = 0; i < n; i++) {
1020 		data->args.errors[i] = errors[i];
1021 		data->args.num_errors++;
1022 		data->res.num_errors++;
1023 	}
1024 	msg.rpc_argp = &data->args;
1025 	msg.rpc_resp = &data->res;
1026 	task_setup.callback_data = data;
1027 	task_setup.rpc_client = NFS_SERVER(inode)->client;
1028 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
1029 	task = rpc_run_task(&task_setup);
1030 	if (IS_ERR(task))
1031 		return PTR_ERR(task);
1032 	rpc_put_task(task);
1033 	return 0;
1034 }
1035 EXPORT_SYMBOL_GPL(nfs42_proc_layouterror);
1036 
1037 static int _nfs42_proc_clone(struct rpc_message *msg, struct file *src_f,
1038 		struct file *dst_f, struct nfs_lock_context *src_lock,
1039 		struct nfs_lock_context *dst_lock, loff_t src_offset,
1040 		loff_t dst_offset, loff_t count)
1041 {
1042 	struct inode *src_inode = file_inode(src_f);
1043 	struct inode *dst_inode = file_inode(dst_f);
1044 	struct nfs_server *server = NFS_SERVER(dst_inode);
1045 	struct nfs42_clone_args args = {
1046 		.src_fh = NFS_FH(src_inode),
1047 		.dst_fh = NFS_FH(dst_inode),
1048 		.src_offset = src_offset,
1049 		.dst_offset = dst_offset,
1050 		.count = count,
1051 		.dst_bitmask = server->cache_consistency_bitmask,
1052 	};
1053 	struct nfs42_clone_res res = {
1054 		.server	= server,
1055 	};
1056 	int status;
1057 
1058 	msg->rpc_argp = &args;
1059 	msg->rpc_resp = &res;
1060 
1061 	status = nfs4_set_rw_stateid(&args.src_stateid, src_lock->open_context,
1062 			src_lock, FMODE_READ);
1063 	if (status) {
1064 		if (status == -EAGAIN)
1065 			status = -NFS4ERR_BAD_STATEID;
1066 		return status;
1067 	}
1068 	status = nfs4_set_rw_stateid(&args.dst_stateid, dst_lock->open_context,
1069 			dst_lock, FMODE_WRITE);
1070 	if (status) {
1071 		if (status == -EAGAIN)
1072 			status = -NFS4ERR_BAD_STATEID;
1073 		return status;
1074 	}
1075 
1076 	res.dst_fattr = nfs_alloc_fattr();
1077 	if (!res.dst_fattr)
1078 		return -ENOMEM;
1079 
1080 	status = nfs4_call_sync(server->client, server, msg,
1081 				&args.seq_args, &res.seq_res, 0);
1082 	trace_nfs4_clone(src_inode, dst_inode, &args, status);
1083 	if (status == 0) {
1084 		nfs42_copy_dest_done(dst_inode, dst_offset, count);
1085 		status = nfs_post_op_update_inode(dst_inode, res.dst_fattr);
1086 	}
1087 
1088 	kfree(res.dst_fattr);
1089 	return status;
1090 }
1091 
1092 int nfs42_proc_clone(struct file *src_f, struct file *dst_f,
1093 		     loff_t src_offset, loff_t dst_offset, loff_t count)
1094 {
1095 	struct rpc_message msg = {
1096 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLONE],
1097 	};
1098 	struct inode *inode = file_inode(src_f);
1099 	struct nfs_server *server = NFS_SERVER(file_inode(src_f));
1100 	struct nfs_lock_context *src_lock;
1101 	struct nfs_lock_context *dst_lock;
1102 	struct nfs4_exception src_exception = { };
1103 	struct nfs4_exception dst_exception = { };
1104 	int err, err2;
1105 
1106 	if (!nfs_server_capable(inode, NFS_CAP_CLONE))
1107 		return -EOPNOTSUPP;
1108 
1109 	src_lock = nfs_get_lock_context(nfs_file_open_context(src_f));
1110 	if (IS_ERR(src_lock))
1111 		return PTR_ERR(src_lock);
1112 
1113 	src_exception.inode = file_inode(src_f);
1114 	src_exception.state = src_lock->open_context->state;
1115 
1116 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst_f));
1117 	if (IS_ERR(dst_lock)) {
1118 		err = PTR_ERR(dst_lock);
1119 		goto out_put_src_lock;
1120 	}
1121 
1122 	dst_exception.inode = file_inode(dst_f);
1123 	dst_exception.state = dst_lock->open_context->state;
1124 
1125 	do {
1126 		err = _nfs42_proc_clone(&msg, src_f, dst_f, src_lock, dst_lock,
1127 					src_offset, dst_offset, count);
1128 		if (err == -ENOTSUPP || err == -EOPNOTSUPP) {
1129 			NFS_SERVER(inode)->caps &= ~NFS_CAP_CLONE;
1130 			err = -EOPNOTSUPP;
1131 			break;
1132 		}
1133 
1134 		err2 = nfs4_handle_exception(server, err, &src_exception);
1135 		err = nfs4_handle_exception(server, err, &dst_exception);
1136 		if (!err)
1137 			err = err2;
1138 	} while (src_exception.retry || dst_exception.retry);
1139 
1140 	nfs_put_lock_context(dst_lock);
1141 out_put_src_lock:
1142 	nfs_put_lock_context(src_lock);
1143 	return err;
1144 }
1145 
1146 #define NFS4XATTR_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
1147 
1148 static int _nfs42_proc_removexattr(struct inode *inode, const char *name)
1149 {
1150 	struct nfs_server *server = NFS_SERVER(inode);
1151 	struct nfs42_removexattrargs args = {
1152 		.fh = NFS_FH(inode),
1153 		.xattr_name = name,
1154 	};
1155 	struct nfs42_removexattrres res;
1156 	struct rpc_message msg = {
1157 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVEXATTR],
1158 		.rpc_argp = &args,
1159 		.rpc_resp = &res,
1160 	};
1161 	int ret;
1162 	unsigned long timestamp = jiffies;
1163 
1164 	ret = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
1165 	    &res.seq_res, 1);
1166 	if (!ret)
1167 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1168 
1169 	return ret;
1170 }
1171 
1172 static int _nfs42_proc_setxattr(struct inode *inode, const char *name,
1173 				const void *buf, size_t buflen, int flags)
1174 {
1175 	struct nfs_server *server = NFS_SERVER(inode);
1176 	struct page *pages[NFS4XATTR_MAXPAGES];
1177 	struct nfs42_setxattrargs arg = {
1178 		.fh		= NFS_FH(inode),
1179 		.xattr_pages	= pages,
1180 		.xattr_len	= buflen,
1181 		.xattr_name	= name,
1182 		.xattr_flags	= flags,
1183 	};
1184 	struct nfs42_setxattrres res;
1185 	struct rpc_message msg = {
1186 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETXATTR],
1187 		.rpc_argp	= &arg,
1188 		.rpc_resp	= &res,
1189 	};
1190 	int ret, np;
1191 	unsigned long timestamp = jiffies;
1192 
1193 	if (buflen > server->sxasize)
1194 		return -ERANGE;
1195 
1196 	if (buflen > 0) {
1197 		np = nfs4_buf_to_pages_noslab(buf, buflen, arg.xattr_pages);
1198 		if (np < 0)
1199 			return np;
1200 	} else
1201 		np = 0;
1202 
1203 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1204 	    &res.seq_res, 1);
1205 
1206 	for (; np > 0; np--)
1207 		put_page(pages[np - 1]);
1208 
1209 	if (!ret)
1210 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1211 
1212 	return ret;
1213 }
1214 
1215 static ssize_t _nfs42_proc_getxattr(struct inode *inode, const char *name,
1216 				void *buf, size_t buflen, struct page **pages,
1217 				size_t plen)
1218 {
1219 	struct nfs_server *server = NFS_SERVER(inode);
1220 	struct nfs42_getxattrargs arg = {
1221 		.fh		= NFS_FH(inode),
1222 		.xattr_name	= name,
1223 	};
1224 	struct nfs42_getxattrres res;
1225 	struct rpc_message msg = {
1226 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETXATTR],
1227 		.rpc_argp	= &arg,
1228 		.rpc_resp	= &res,
1229 	};
1230 	ssize_t ret;
1231 
1232 	arg.xattr_len = plen;
1233 	arg.xattr_pages = pages;
1234 
1235 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1236 	    &res.seq_res, 0);
1237 	if (ret < 0)
1238 		return ret;
1239 
1240 	/*
1241 	 * Normally, the caching is done one layer up, but for successful
1242 	 * RPCS, always cache the result here, even if the caller was
1243 	 * just querying the length, or if the reply was too big for
1244 	 * the caller. This avoids a second RPC in the case of the
1245 	 * common query-alloc-retrieve cycle for xattrs.
1246 	 *
1247 	 * Note that xattr_len is always capped to XATTR_SIZE_MAX.
1248 	 */
1249 
1250 	nfs4_xattr_cache_add(inode, name, NULL, pages, res.xattr_len);
1251 
1252 	if (buflen) {
1253 		if (res.xattr_len > buflen)
1254 			return -ERANGE;
1255 		_copy_from_pages(buf, pages, 0, res.xattr_len);
1256 	}
1257 
1258 	return res.xattr_len;
1259 }
1260 
1261 static ssize_t _nfs42_proc_listxattrs(struct inode *inode, void *buf,
1262 				 size_t buflen, u64 *cookiep, bool *eofp)
1263 {
1264 	struct nfs_server *server = NFS_SERVER(inode);
1265 	struct page **pages;
1266 	struct nfs42_listxattrsargs arg = {
1267 		.fh		= NFS_FH(inode),
1268 		.cookie		= *cookiep,
1269 	};
1270 	struct nfs42_listxattrsres res = {
1271 		.eof = false,
1272 		.xattr_buf = buf,
1273 		.xattr_len = buflen,
1274 	};
1275 	struct rpc_message msg = {
1276 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LISTXATTRS],
1277 		.rpc_argp	= &arg,
1278 		.rpc_resp	= &res,
1279 	};
1280 	u32 xdrlen;
1281 	int ret, np, i;
1282 
1283 
1284 	ret = -ENOMEM;
1285 	res.scratch = alloc_page(GFP_KERNEL);
1286 	if (!res.scratch)
1287 		goto out;
1288 
1289 	xdrlen = nfs42_listxattr_xdrsize(buflen);
1290 	if (xdrlen > server->lxasize)
1291 		xdrlen = server->lxasize;
1292 	np = xdrlen / PAGE_SIZE + 1;
1293 
1294 	pages = kcalloc(np, sizeof(struct page *), GFP_KERNEL);
1295 	if (!pages)
1296 		goto out_free_scratch;
1297 	for (i = 0; i < np; i++) {
1298 		pages[i] = alloc_page(GFP_KERNEL);
1299 		if (!pages[i])
1300 			goto out_free_pages;
1301 	}
1302 
1303 	arg.xattr_pages = pages;
1304 	arg.count = xdrlen;
1305 
1306 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1307 	    &res.seq_res, 0);
1308 
1309 	if (ret >= 0) {
1310 		ret = res.copied;
1311 		*cookiep = res.cookie;
1312 		*eofp = res.eof;
1313 	}
1314 
1315 out_free_pages:
1316 	while (--np >= 0) {
1317 		if (pages[np])
1318 			__free_page(pages[np]);
1319 	}
1320 	kfree(pages);
1321 out_free_scratch:
1322 	__free_page(res.scratch);
1323 out:
1324 	return ret;
1325 
1326 }
1327 
1328 ssize_t nfs42_proc_getxattr(struct inode *inode, const char *name,
1329 			      void *buf, size_t buflen)
1330 {
1331 	struct nfs4_exception exception = { };
1332 	ssize_t err, np, i;
1333 	struct page **pages;
1334 
1335 	np = nfs_page_array_len(0, buflen ?: XATTR_SIZE_MAX);
1336 	pages = kmalloc_array(np, sizeof(*pages), GFP_KERNEL);
1337 	if (!pages)
1338 		return -ENOMEM;
1339 
1340 	for (i = 0; i < np; i++) {
1341 		pages[i] = alloc_page(GFP_KERNEL);
1342 		if (!pages[i]) {
1343 			np = i + 1;
1344 			err = -ENOMEM;
1345 			goto out;
1346 		}
1347 	}
1348 
1349 	/*
1350 	 * The GETXATTR op has no length field in the call, and the
1351 	 * xattr data is at the end of the reply.
1352 	 *
1353 	 * There is no downside in using the page-aligned length. It will
1354 	 * allow receiving and caching xattrs that are too large for the
1355 	 * caller but still fit in the page-rounded value.
1356 	 */
1357 	do {
1358 		err = _nfs42_proc_getxattr(inode, name, buf, buflen,
1359 			pages, np * PAGE_SIZE);
1360 		if (err >= 0)
1361 			break;
1362 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1363 				&exception);
1364 	} while (exception.retry);
1365 
1366 out:
1367 	while (--np >= 0)
1368 		__free_page(pages[np]);
1369 	kfree(pages);
1370 
1371 	return err;
1372 }
1373 
1374 int nfs42_proc_setxattr(struct inode *inode, const char *name,
1375 			      const void *buf, size_t buflen, int flags)
1376 {
1377 	struct nfs4_exception exception = { };
1378 	int err;
1379 
1380 	do {
1381 		err = _nfs42_proc_setxattr(inode, name, buf, buflen, flags);
1382 		if (!err)
1383 			break;
1384 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1385 				&exception);
1386 	} while (exception.retry);
1387 
1388 	return err;
1389 }
1390 
1391 ssize_t nfs42_proc_listxattrs(struct inode *inode, void *buf,
1392 			      size_t buflen, u64 *cookiep, bool *eofp)
1393 {
1394 	struct nfs4_exception exception = { };
1395 	ssize_t err;
1396 
1397 	do {
1398 		err = _nfs42_proc_listxattrs(inode, buf, buflen,
1399 		    cookiep, eofp);
1400 		if (err >= 0)
1401 			break;
1402 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1403 				&exception);
1404 	} while (exception.retry);
1405 
1406 	return err;
1407 }
1408 
1409 int nfs42_proc_removexattr(struct inode *inode, const char *name)
1410 {
1411 	struct nfs4_exception exception = { };
1412 	int err;
1413 
1414 	do {
1415 		err = _nfs42_proc_removexattr(inode, name);
1416 		if (!err)
1417 			break;
1418 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1419 				&exception);
1420 	} while (exception.retry);
1421 
1422 	return err;
1423 }
1424