xref: /openbmc/linux/fs/nfs/nfs42proc.c (revision dd21bfa4)
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[NFS_BITMASK_SZ];
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 	nfs4_bitmask_set(bitmask, server->cache_consistency_bitmask, inode,
73 			 NFS_INO_INVALID_BLOCKS);
74 
75 	res.falloc_fattr = nfs_alloc_fattr();
76 	if (!res.falloc_fattr)
77 		return -ENOMEM;
78 
79 	status = nfs4_call_sync(server->client, server, msg,
80 				&args.seq_args, &res.seq_res, 0);
81 	if (status == 0)
82 		status = nfs_post_op_update_inode_force_wcc(inode,
83 							    res.falloc_fattr);
84 
85 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_ALLOCATE])
86 		trace_nfs4_fallocate(inode, &args, status);
87 	else
88 		trace_nfs4_deallocate(inode, &args, status);
89 	kfree(res.falloc_fattr);
90 	return status;
91 }
92 
93 static int nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep,
94 				loff_t offset, loff_t len)
95 {
96 	struct inode *inode = file_inode(filep);
97 	struct nfs_server *server = NFS_SERVER(inode);
98 	struct nfs4_exception exception = { };
99 	struct nfs_lock_context *lock;
100 	int err;
101 
102 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
103 	if (IS_ERR(lock))
104 		return PTR_ERR(lock);
105 
106 	exception.inode = inode;
107 	exception.state = lock->open_context->state;
108 
109 	err = nfs_sync_inode(inode);
110 	if (err)
111 		goto out;
112 
113 	do {
114 		err = _nfs42_proc_fallocate(msg, filep, lock, offset, len);
115 		if (err == -ENOTSUPP) {
116 			err = -EOPNOTSUPP;
117 			break;
118 		}
119 		err = nfs4_handle_exception(server, err, &exception);
120 	} while (exception.retry);
121 out:
122 	nfs_put_lock_context(lock);
123 	return err;
124 }
125 
126 int nfs42_proc_allocate(struct file *filep, loff_t offset, loff_t len)
127 {
128 	struct rpc_message msg = {
129 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ALLOCATE],
130 	};
131 	struct inode *inode = file_inode(filep);
132 	int err;
133 
134 	if (!nfs_server_capable(inode, NFS_CAP_ALLOCATE))
135 		return -EOPNOTSUPP;
136 
137 	inode_lock(inode);
138 
139 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
140 	if (err == -EOPNOTSUPP)
141 		NFS_SERVER(inode)->caps &= ~NFS_CAP_ALLOCATE;
142 
143 	inode_unlock(inode);
144 	return err;
145 }
146 
147 int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len)
148 {
149 	struct rpc_message msg = {
150 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DEALLOCATE],
151 	};
152 	struct inode *inode = file_inode(filep);
153 	int err;
154 
155 	if (!nfs_server_capable(inode, NFS_CAP_DEALLOCATE))
156 		return -EOPNOTSUPP;
157 
158 	inode_lock(inode);
159 
160 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
161 	if (err == 0)
162 		truncate_pagecache_range(inode, offset, (offset + len) -1);
163 	if (err == -EOPNOTSUPP)
164 		NFS_SERVER(inode)->caps &= ~NFS_CAP_DEALLOCATE;
165 
166 	inode_unlock(inode);
167 	return err;
168 }
169 
170 static int handle_async_copy(struct nfs42_copy_res *res,
171 			     struct nfs_server *dst_server,
172 			     struct nfs_server *src_server,
173 			     struct file *src,
174 			     struct file *dst,
175 			     nfs4_stateid *src_stateid,
176 			     bool *restart)
177 {
178 	struct nfs4_copy_state *copy, *tmp_copy;
179 	int status = NFS4_OK;
180 	bool found_pending = false;
181 	struct nfs_open_context *dst_ctx = nfs_file_open_context(dst);
182 	struct nfs_open_context *src_ctx = nfs_file_open_context(src);
183 
184 	copy = kzalloc(sizeof(struct nfs4_copy_state), GFP_NOFS);
185 	if (!copy)
186 		return -ENOMEM;
187 
188 	spin_lock(&dst_server->nfs_client->cl_lock);
189 	list_for_each_entry(tmp_copy,
190 				&dst_server->nfs_client->pending_cb_stateids,
191 				copies) {
192 		if (memcmp(&res->write_res.stateid, &tmp_copy->stateid,
193 				NFS4_STATEID_SIZE))
194 			continue;
195 		found_pending = true;
196 		list_del(&tmp_copy->copies);
197 		break;
198 	}
199 	if (found_pending) {
200 		spin_unlock(&dst_server->nfs_client->cl_lock);
201 		kfree(copy);
202 		copy = tmp_copy;
203 		goto out;
204 	}
205 
206 	memcpy(&copy->stateid, &res->write_res.stateid, NFS4_STATEID_SIZE);
207 	init_completion(&copy->completion);
208 	copy->parent_dst_state = dst_ctx->state;
209 	copy->parent_src_state = src_ctx->state;
210 
211 	list_add_tail(&copy->copies, &dst_server->ss_copies);
212 	spin_unlock(&dst_server->nfs_client->cl_lock);
213 
214 	if (dst_server != src_server) {
215 		spin_lock(&src_server->nfs_client->cl_lock);
216 		list_add_tail(&copy->src_copies, &src_server->ss_copies);
217 		spin_unlock(&src_server->nfs_client->cl_lock);
218 	}
219 
220 	status = wait_for_completion_interruptible(&copy->completion);
221 	spin_lock(&dst_server->nfs_client->cl_lock);
222 	list_del_init(&copy->copies);
223 	spin_unlock(&dst_server->nfs_client->cl_lock);
224 	if (dst_server != src_server) {
225 		spin_lock(&src_server->nfs_client->cl_lock);
226 		list_del_init(&copy->src_copies);
227 		spin_unlock(&src_server->nfs_client->cl_lock);
228 	}
229 	if (status == -ERESTARTSYS) {
230 		goto out_cancel;
231 	} else if (copy->flags || copy->error == NFS4ERR_PARTNER_NO_AUTH) {
232 		status = -EAGAIN;
233 		*restart = true;
234 		goto out_cancel;
235 	}
236 out:
237 	res->write_res.count = copy->count;
238 	memcpy(&res->write_res.verifier, &copy->verf, sizeof(copy->verf));
239 	status = -copy->error;
240 
241 out_free:
242 	kfree(copy);
243 	return status;
244 out_cancel:
245 	nfs42_do_offload_cancel_async(dst, &copy->stateid);
246 	if (!nfs42_files_from_same_server(src, dst))
247 		nfs42_do_offload_cancel_async(src, src_stateid);
248 	goto out_free;
249 }
250 
251 static int process_copy_commit(struct file *dst, loff_t pos_dst,
252 			       struct nfs42_copy_res *res)
253 {
254 	struct nfs_commitres cres;
255 	int status = -ENOMEM;
256 
257 	cres.verf = kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS);
258 	if (!cres.verf)
259 		goto out;
260 
261 	status = nfs4_proc_commit(dst, pos_dst, res->write_res.count, &cres);
262 	if (status)
263 		goto out_free;
264 	if (nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
265 				    &cres.verf->verifier)) {
266 		dprintk("commit verf differs from copy verf\n");
267 		status = -EAGAIN;
268 	}
269 out_free:
270 	kfree(cres.verf);
271 out:
272 	return status;
273 }
274 
275 /**
276  * nfs42_copy_dest_done - perform inode cache updates after clone/copy offload
277  * @inode: pointer to destination inode
278  * @pos: destination offset
279  * @len: copy length
280  *
281  * Punch a hole in the inode page cache, so that the NFS client will
282  * know to retrieve new data.
283  * Update the file size if necessary, and then mark the inode as having
284  * invalid cached values for change attribute, ctime, mtime and space used.
285  */
286 static void nfs42_copy_dest_done(struct inode *inode, loff_t pos, loff_t len)
287 {
288 	loff_t newsize = pos + len;
289 	loff_t end = newsize - 1;
290 
291 	WARN_ON_ONCE(invalidate_inode_pages2_range(inode->i_mapping,
292 				pos >> PAGE_SHIFT, end >> PAGE_SHIFT));
293 
294 	spin_lock(&inode->i_lock);
295 	if (newsize > i_size_read(inode))
296 		i_size_write(inode, newsize);
297 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
298 					     NFS_INO_INVALID_CTIME |
299 					     NFS_INO_INVALID_MTIME |
300 					     NFS_INO_INVALID_BLOCKS);
301 	spin_unlock(&inode->i_lock);
302 }
303 
304 static ssize_t _nfs42_proc_copy(struct file *src,
305 				struct nfs_lock_context *src_lock,
306 				struct file *dst,
307 				struct nfs_lock_context *dst_lock,
308 				struct nfs42_copy_args *args,
309 				struct nfs42_copy_res *res,
310 				struct nl4_server *nss,
311 				nfs4_stateid *cnr_stateid,
312 				bool *restart)
313 {
314 	struct rpc_message msg = {
315 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY],
316 		.rpc_argp = args,
317 		.rpc_resp = res,
318 	};
319 	struct inode *dst_inode = file_inode(dst);
320 	struct inode *src_inode = file_inode(src);
321 	struct nfs_server *dst_server = NFS_SERVER(dst_inode);
322 	struct nfs_server *src_server = NFS_SERVER(src_inode);
323 	loff_t pos_src = args->src_pos;
324 	loff_t pos_dst = args->dst_pos;
325 	size_t count = args->count;
326 	ssize_t status;
327 
328 	if (nss) {
329 		args->cp_src = nss;
330 		nfs4_stateid_copy(&args->src_stateid, cnr_stateid);
331 	} else {
332 		status = nfs4_set_rw_stateid(&args->src_stateid,
333 				src_lock->open_context, src_lock, FMODE_READ);
334 		if (status) {
335 			if (status == -EAGAIN)
336 				status = -NFS4ERR_BAD_STATEID;
337 			return status;
338 		}
339 	}
340 	status = nfs_filemap_write_and_wait_range(file_inode(src)->i_mapping,
341 			pos_src, pos_src + (loff_t)count - 1);
342 	if (status)
343 		return status;
344 
345 	status = nfs4_set_rw_stateid(&args->dst_stateid, dst_lock->open_context,
346 				     dst_lock, FMODE_WRITE);
347 	if (status) {
348 		if (status == -EAGAIN)
349 			status = -NFS4ERR_BAD_STATEID;
350 		return status;
351 	}
352 
353 	status = nfs_sync_inode(dst_inode);
354 	if (status)
355 		return status;
356 
357 	res->commit_res.verf = NULL;
358 	if (args->sync) {
359 		res->commit_res.verf =
360 			kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS);
361 		if (!res->commit_res.verf)
362 			return -ENOMEM;
363 	}
364 	set_bit(NFS_CLNT_SRC_SSC_COPY_STATE,
365 		&src_lock->open_context->state->flags);
366 	set_bit(NFS_CLNT_DST_SSC_COPY_STATE,
367 		&dst_lock->open_context->state->flags);
368 
369 	status = nfs4_call_sync(dst_server->client, dst_server, &msg,
370 				&args->seq_args, &res->seq_res, 0);
371 	trace_nfs4_copy(src_inode, dst_inode, args, res, nss, status);
372 	if (status == -ENOTSUPP)
373 		dst_server->caps &= ~NFS_CAP_COPY;
374 	if (status)
375 		goto out;
376 
377 	if (args->sync &&
378 		nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
379 				    &res->commit_res.verf->verifier)) {
380 		status = -EAGAIN;
381 		goto out;
382 	}
383 
384 	if (!res->synchronous) {
385 		status = handle_async_copy(res, dst_server, src_server, src,
386 				dst, &args->src_stateid, restart);
387 		if (status)
388 			goto out;
389 	}
390 
391 	if ((!res->synchronous || !args->sync) &&
392 			res->write_res.verifier.committed != NFS_FILE_SYNC) {
393 		status = process_copy_commit(dst, pos_dst, res);
394 		if (status)
395 			goto out;
396 	}
397 
398 	nfs42_copy_dest_done(dst_inode, pos_dst, res->write_res.count);
399 	nfs_invalidate_atime(src_inode);
400 	status = res->write_res.count;
401 out:
402 	if (args->sync)
403 		kfree(res->commit_res.verf);
404 	return status;
405 }
406 
407 ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
408 			struct file *dst, loff_t pos_dst, size_t count,
409 			struct nl4_server *nss,
410 			nfs4_stateid *cnr_stateid, bool sync)
411 {
412 	struct nfs_server *server = NFS_SERVER(file_inode(dst));
413 	struct nfs_lock_context *src_lock;
414 	struct nfs_lock_context *dst_lock;
415 	struct nfs42_copy_args args = {
416 		.src_fh		= NFS_FH(file_inode(src)),
417 		.src_pos	= pos_src,
418 		.dst_fh		= NFS_FH(file_inode(dst)),
419 		.dst_pos	= pos_dst,
420 		.count		= count,
421 		.sync		= sync,
422 	};
423 	struct nfs42_copy_res res;
424 	struct nfs4_exception src_exception = {
425 		.inode		= file_inode(src),
426 		.stateid	= &args.src_stateid,
427 	};
428 	struct nfs4_exception dst_exception = {
429 		.inode		= file_inode(dst),
430 		.stateid	= &args.dst_stateid,
431 	};
432 	ssize_t err, err2;
433 	bool restart = false;
434 
435 	src_lock = nfs_get_lock_context(nfs_file_open_context(src));
436 	if (IS_ERR(src_lock))
437 		return PTR_ERR(src_lock);
438 
439 	src_exception.state = src_lock->open_context->state;
440 
441 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst));
442 	if (IS_ERR(dst_lock)) {
443 		err = PTR_ERR(dst_lock);
444 		goto out_put_src_lock;
445 	}
446 
447 	dst_exception.state = dst_lock->open_context->state;
448 
449 	do {
450 		inode_lock(file_inode(dst));
451 		err = _nfs42_proc_copy(src, src_lock,
452 				dst, dst_lock,
453 				&args, &res,
454 				nss, cnr_stateid, &restart);
455 		inode_unlock(file_inode(dst));
456 
457 		if (err >= 0)
458 			break;
459 		if (err == -ENOTSUPP &&
460 				nfs42_files_from_same_server(src, dst)) {
461 			err = -EOPNOTSUPP;
462 			break;
463 		} else if (err == -EAGAIN) {
464 			if (!restart) {
465 				dst_exception.retry = 1;
466 				continue;
467 			}
468 			break;
469 		} else if (err == -NFS4ERR_OFFLOAD_NO_REQS && !args.sync) {
470 			args.sync = true;
471 			dst_exception.retry = 1;
472 			continue;
473 		} else if ((err == -ESTALE ||
474 				err == -NFS4ERR_OFFLOAD_DENIED ||
475 				err == -ENOTSUPP) &&
476 				!nfs42_files_from_same_server(src, dst)) {
477 			nfs42_do_offload_cancel_async(src, &args.src_stateid);
478 			err = -EOPNOTSUPP;
479 			break;
480 		}
481 
482 		err2 = nfs4_handle_exception(server, err, &src_exception);
483 		err  = nfs4_handle_exception(server, err, &dst_exception);
484 		if (!err)
485 			err = err2;
486 	} while (src_exception.retry || dst_exception.retry);
487 
488 	nfs_put_lock_context(dst_lock);
489 out_put_src_lock:
490 	nfs_put_lock_context(src_lock);
491 	return err;
492 }
493 
494 struct nfs42_offloadcancel_data {
495 	struct nfs_server *seq_server;
496 	struct nfs42_offload_status_args args;
497 	struct nfs42_offload_status_res res;
498 };
499 
500 static void nfs42_offload_cancel_prepare(struct rpc_task *task, void *calldata)
501 {
502 	struct nfs42_offloadcancel_data *data = calldata;
503 
504 	nfs4_setup_sequence(data->seq_server->nfs_client,
505 				&data->args.osa_seq_args,
506 				&data->res.osr_seq_res, task);
507 }
508 
509 static void nfs42_offload_cancel_done(struct rpc_task *task, void *calldata)
510 {
511 	struct nfs42_offloadcancel_data *data = calldata;
512 
513 	trace_nfs4_offload_cancel(&data->args, task->tk_status);
514 	nfs41_sequence_done(task, &data->res.osr_seq_res);
515 	if (task->tk_status &&
516 		nfs4_async_handle_error(task, data->seq_server, NULL,
517 			NULL) == -EAGAIN)
518 		rpc_restart_call_prepare(task);
519 }
520 
521 static void nfs42_free_offloadcancel_data(void *data)
522 {
523 	kfree(data);
524 }
525 
526 static const struct rpc_call_ops nfs42_offload_cancel_ops = {
527 	.rpc_call_prepare = nfs42_offload_cancel_prepare,
528 	.rpc_call_done = nfs42_offload_cancel_done,
529 	.rpc_release = nfs42_free_offloadcancel_data,
530 };
531 
532 static int nfs42_do_offload_cancel_async(struct file *dst,
533 					 nfs4_stateid *stateid)
534 {
535 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
536 	struct nfs42_offloadcancel_data *data = NULL;
537 	struct nfs_open_context *ctx = nfs_file_open_context(dst);
538 	struct rpc_task *task;
539 	struct rpc_message msg = {
540 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OFFLOAD_CANCEL],
541 		.rpc_cred = ctx->cred,
542 	};
543 	struct rpc_task_setup task_setup_data = {
544 		.rpc_client = dst_server->client,
545 		.rpc_message = &msg,
546 		.callback_ops = &nfs42_offload_cancel_ops,
547 		.workqueue = nfsiod_workqueue,
548 		.flags = RPC_TASK_ASYNC,
549 	};
550 	int status;
551 
552 	if (!(dst_server->caps & NFS_CAP_OFFLOAD_CANCEL))
553 		return -EOPNOTSUPP;
554 
555 	data = kzalloc(sizeof(struct nfs42_offloadcancel_data), GFP_NOFS);
556 	if (data == NULL)
557 		return -ENOMEM;
558 
559 	data->seq_server = dst_server;
560 	data->args.osa_src_fh = NFS_FH(file_inode(dst));
561 	memcpy(&data->args.osa_stateid, stateid,
562 		sizeof(data->args.osa_stateid));
563 	msg.rpc_argp = &data->args;
564 	msg.rpc_resp = &data->res;
565 	task_setup_data.callback_data = data;
566 	nfs4_init_sequence(&data->args.osa_seq_args, &data->res.osr_seq_res,
567 			   1, 0);
568 	task = rpc_run_task(&task_setup_data);
569 	if (IS_ERR(task))
570 		return PTR_ERR(task);
571 	status = rpc_wait_for_completion_task(task);
572 	if (status == -ENOTSUPP)
573 		dst_server->caps &= ~NFS_CAP_OFFLOAD_CANCEL;
574 	rpc_put_task(task);
575 	return status;
576 }
577 
578 static int _nfs42_proc_copy_notify(struct file *src, struct file *dst,
579 				   struct nfs42_copy_notify_args *args,
580 				   struct nfs42_copy_notify_res *res)
581 {
582 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
583 	struct rpc_message msg = {
584 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY_NOTIFY],
585 		.rpc_argp = args,
586 		.rpc_resp = res,
587 	};
588 	int status;
589 	struct nfs_open_context *ctx;
590 	struct nfs_lock_context *l_ctx;
591 
592 	ctx = get_nfs_open_context(nfs_file_open_context(src));
593 	l_ctx = nfs_get_lock_context(ctx);
594 	if (IS_ERR(l_ctx))
595 		return PTR_ERR(l_ctx);
596 
597 	status = nfs4_set_rw_stateid(&args->cna_src_stateid, ctx, l_ctx,
598 				     FMODE_READ);
599 	nfs_put_lock_context(l_ctx);
600 	if (status) {
601 		if (status == -EAGAIN)
602 			status = -NFS4ERR_BAD_STATEID;
603 		return status;
604 	}
605 
606 	status = nfs4_call_sync(src_server->client, src_server, &msg,
607 				&args->cna_seq_args, &res->cnr_seq_res, 0);
608 	trace_nfs4_copy_notify(file_inode(src), args, res, status);
609 	if (status == -ENOTSUPP)
610 		src_server->caps &= ~NFS_CAP_COPY_NOTIFY;
611 
612 	put_nfs_open_context(nfs_file_open_context(src));
613 	return status;
614 }
615 
616 int nfs42_proc_copy_notify(struct file *src, struct file *dst,
617 				struct nfs42_copy_notify_res *res)
618 {
619 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
620 	struct nfs42_copy_notify_args *args;
621 	struct nfs4_exception exception = {
622 		.inode = file_inode(src),
623 	};
624 	int status;
625 
626 	if (!(src_server->caps & NFS_CAP_COPY_NOTIFY))
627 		return -EOPNOTSUPP;
628 
629 	args = kzalloc(sizeof(struct nfs42_copy_notify_args), GFP_NOFS);
630 	if (args == NULL)
631 		return -ENOMEM;
632 
633 	args->cna_src_fh  = NFS_FH(file_inode(src)),
634 	args->cna_dst.nl4_type = NL4_NETADDR;
635 	nfs42_set_netaddr(dst, &args->cna_dst.u.nl4_addr);
636 	exception.stateid = &args->cna_src_stateid;
637 
638 	do {
639 		status = _nfs42_proc_copy_notify(src, dst, args, res);
640 		if (status == -ENOTSUPP) {
641 			status = -EOPNOTSUPP;
642 			goto out;
643 		}
644 		status = nfs4_handle_exception(src_server, status, &exception);
645 	} while (exception.retry);
646 
647 out:
648 	kfree(args);
649 	return status;
650 }
651 
652 static loff_t _nfs42_proc_llseek(struct file *filep,
653 		struct nfs_lock_context *lock, loff_t offset, int whence)
654 {
655 	struct inode *inode = file_inode(filep);
656 	struct nfs42_seek_args args = {
657 		.sa_fh		= NFS_FH(inode),
658 		.sa_offset	= offset,
659 		.sa_what	= (whence == SEEK_HOLE) ?
660 					NFS4_CONTENT_HOLE : NFS4_CONTENT_DATA,
661 	};
662 	struct nfs42_seek_res res;
663 	struct rpc_message msg = {
664 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEEK],
665 		.rpc_argp = &args,
666 		.rpc_resp = &res,
667 	};
668 	struct nfs_server *server = NFS_SERVER(inode);
669 	int status;
670 
671 	if (!nfs_server_capable(inode, NFS_CAP_SEEK))
672 		return -ENOTSUPP;
673 
674 	status = nfs4_set_rw_stateid(&args.sa_stateid, lock->open_context,
675 			lock, FMODE_READ);
676 	if (status) {
677 		if (status == -EAGAIN)
678 			status = -NFS4ERR_BAD_STATEID;
679 		return status;
680 	}
681 
682 	status = nfs_filemap_write_and_wait_range(inode->i_mapping,
683 			offset, LLONG_MAX);
684 	if (status)
685 		return status;
686 
687 	status = nfs4_call_sync(server->client, server, &msg,
688 				&args.seq_args, &res.seq_res, 0);
689 	trace_nfs4_llseek(inode, &args, &res, status);
690 	if (status == -ENOTSUPP)
691 		server->caps &= ~NFS_CAP_SEEK;
692 	if (status)
693 		return status;
694 
695 	if (whence == SEEK_DATA && res.sr_eof)
696 		return -NFS4ERR_NXIO;
697 	else
698 		return vfs_setpos(filep, res.sr_offset, inode->i_sb->s_maxbytes);
699 }
700 
701 loff_t nfs42_proc_llseek(struct file *filep, loff_t offset, int whence)
702 {
703 	struct nfs_server *server = NFS_SERVER(file_inode(filep));
704 	struct nfs4_exception exception = { };
705 	struct nfs_lock_context *lock;
706 	loff_t err;
707 
708 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
709 	if (IS_ERR(lock))
710 		return PTR_ERR(lock);
711 
712 	exception.inode = file_inode(filep);
713 	exception.state = lock->open_context->state;
714 
715 	do {
716 		err = _nfs42_proc_llseek(filep, lock, offset, whence);
717 		if (err >= 0)
718 			break;
719 		if (err == -ENOTSUPP) {
720 			err = -EOPNOTSUPP;
721 			break;
722 		}
723 		err = nfs4_handle_exception(server, err, &exception);
724 	} while (exception.retry);
725 
726 	nfs_put_lock_context(lock);
727 	return err;
728 }
729 
730 
731 static void
732 nfs42_layoutstat_prepare(struct rpc_task *task, void *calldata)
733 {
734 	struct nfs42_layoutstat_data *data = calldata;
735 	struct inode *inode = data->inode;
736 	struct nfs_server *server = NFS_SERVER(inode);
737 	struct pnfs_layout_hdr *lo;
738 
739 	spin_lock(&inode->i_lock);
740 	lo = NFS_I(inode)->layout;
741 	if (!pnfs_layout_is_valid(lo)) {
742 		spin_unlock(&inode->i_lock);
743 		rpc_exit(task, 0);
744 		return;
745 	}
746 	nfs4_stateid_copy(&data->args.stateid, &lo->plh_stateid);
747 	spin_unlock(&inode->i_lock);
748 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
749 			    &data->res.seq_res, task);
750 }
751 
752 static void
753 nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
754 {
755 	struct nfs42_layoutstat_data *data = calldata;
756 	struct inode *inode = data->inode;
757 	struct pnfs_layout_hdr *lo;
758 
759 	if (!nfs4_sequence_done(task, &data->res.seq_res))
760 		return;
761 
762 	switch (task->tk_status) {
763 	case 0:
764 		return;
765 	case -NFS4ERR_BADHANDLE:
766 	case -ESTALE:
767 		pnfs_destroy_layout(NFS_I(inode));
768 		break;
769 	case -NFS4ERR_EXPIRED:
770 	case -NFS4ERR_ADMIN_REVOKED:
771 	case -NFS4ERR_DELEG_REVOKED:
772 	case -NFS4ERR_STALE_STATEID:
773 	case -NFS4ERR_BAD_STATEID:
774 		spin_lock(&inode->i_lock);
775 		lo = NFS_I(inode)->layout;
776 		if (pnfs_layout_is_valid(lo) &&
777 		    nfs4_stateid_match(&data->args.stateid,
778 					     &lo->plh_stateid)) {
779 			LIST_HEAD(head);
780 
781 			/*
782 			 * Mark the bad layout state as invalid, then retry
783 			 * with the current stateid.
784 			 */
785 			pnfs_mark_layout_stateid_invalid(lo, &head);
786 			spin_unlock(&inode->i_lock);
787 			pnfs_free_lseg_list(&head);
788 			nfs_commit_inode(inode, 0);
789 		} else
790 			spin_unlock(&inode->i_lock);
791 		break;
792 	case -NFS4ERR_OLD_STATEID:
793 		spin_lock(&inode->i_lock);
794 		lo = NFS_I(inode)->layout;
795 		if (pnfs_layout_is_valid(lo) &&
796 		    nfs4_stateid_match_other(&data->args.stateid,
797 					&lo->plh_stateid)) {
798 			/* Do we need to delay before resending? */
799 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
800 						&data->args.stateid))
801 				rpc_delay(task, HZ);
802 			rpc_restart_call_prepare(task);
803 		}
804 		spin_unlock(&inode->i_lock);
805 		break;
806 	case -ENOTSUPP:
807 	case -EOPNOTSUPP:
808 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTSTATS;
809 	}
810 
811 	trace_nfs4_layoutstats(inode, &data->args.stateid, task->tk_status);
812 }
813 
814 static void
815 nfs42_layoutstat_release(void *calldata)
816 {
817 	struct nfs42_layoutstat_data *data = calldata;
818 	struct nfs42_layoutstat_devinfo *devinfo = data->args.devinfo;
819 	int i;
820 
821 	for (i = 0; i < data->args.num_dev; i++) {
822 		if (devinfo[i].ld_private.ops && devinfo[i].ld_private.ops->free)
823 			devinfo[i].ld_private.ops->free(&devinfo[i].ld_private);
824 	}
825 
826 	pnfs_put_layout_hdr(NFS_I(data->args.inode)->layout);
827 	smp_mb__before_atomic();
828 	clear_bit(NFS_INO_LAYOUTSTATS, &NFS_I(data->args.inode)->flags);
829 	smp_mb__after_atomic();
830 	nfs_iput_and_deactive(data->inode);
831 	kfree(data->args.devinfo);
832 	kfree(data);
833 }
834 
835 static const struct rpc_call_ops nfs42_layoutstat_ops = {
836 	.rpc_call_prepare = nfs42_layoutstat_prepare,
837 	.rpc_call_done = nfs42_layoutstat_done,
838 	.rpc_release = nfs42_layoutstat_release,
839 };
840 
841 int nfs42_proc_layoutstats_generic(struct nfs_server *server,
842 				   struct nfs42_layoutstat_data *data)
843 {
844 	struct rpc_message msg = {
845 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTSTATS],
846 		.rpc_argp = &data->args,
847 		.rpc_resp = &data->res,
848 	};
849 	struct rpc_task_setup task_setup = {
850 		.rpc_client = server->client,
851 		.rpc_message = &msg,
852 		.callback_ops = &nfs42_layoutstat_ops,
853 		.callback_data = data,
854 		.flags = RPC_TASK_ASYNC,
855 	};
856 	struct rpc_task *task;
857 
858 	data->inode = nfs_igrab_and_active(data->args.inode);
859 	if (!data->inode) {
860 		nfs42_layoutstat_release(data);
861 		return -EAGAIN;
862 	}
863 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
864 	task = rpc_run_task(&task_setup);
865 	if (IS_ERR(task))
866 		return PTR_ERR(task);
867 	rpc_put_task(task);
868 	return 0;
869 }
870 
871 static struct nfs42_layouterror_data *
872 nfs42_alloc_layouterror_data(struct pnfs_layout_segment *lseg, gfp_t gfp_flags)
873 {
874 	struct nfs42_layouterror_data *data;
875 	struct inode *inode = lseg->pls_layout->plh_inode;
876 
877 	data = kzalloc(sizeof(*data), gfp_flags);
878 	if (data) {
879 		data->args.inode = data->inode = nfs_igrab_and_active(inode);
880 		if (data->inode) {
881 			data->lseg = pnfs_get_lseg(lseg);
882 			if (data->lseg)
883 				return data;
884 			nfs_iput_and_deactive(data->inode);
885 		}
886 		kfree(data);
887 	}
888 	return NULL;
889 }
890 
891 static void
892 nfs42_free_layouterror_data(struct nfs42_layouterror_data *data)
893 {
894 	pnfs_put_lseg(data->lseg);
895 	nfs_iput_and_deactive(data->inode);
896 	kfree(data);
897 }
898 
899 static void
900 nfs42_layouterror_prepare(struct rpc_task *task, void *calldata)
901 {
902 	struct nfs42_layouterror_data *data = calldata;
903 	struct inode *inode = data->inode;
904 	struct nfs_server *server = NFS_SERVER(inode);
905 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
906 	unsigned i;
907 
908 	spin_lock(&inode->i_lock);
909 	if (!pnfs_layout_is_valid(lo)) {
910 		spin_unlock(&inode->i_lock);
911 		rpc_exit(task, 0);
912 		return;
913 	}
914 	for (i = 0; i < data->args.num_errors; i++)
915 		nfs4_stateid_copy(&data->args.errors[i].stateid,
916 				&lo->plh_stateid);
917 	spin_unlock(&inode->i_lock);
918 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
919 			    &data->res.seq_res, task);
920 }
921 
922 static void
923 nfs42_layouterror_done(struct rpc_task *task, void *calldata)
924 {
925 	struct nfs42_layouterror_data *data = calldata;
926 	struct inode *inode = data->inode;
927 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
928 
929 	if (!nfs4_sequence_done(task, &data->res.seq_res))
930 		return;
931 
932 	switch (task->tk_status) {
933 	case 0:
934 		return;
935 	case -NFS4ERR_BADHANDLE:
936 	case -ESTALE:
937 		pnfs_destroy_layout(NFS_I(inode));
938 		break;
939 	case -NFS4ERR_EXPIRED:
940 	case -NFS4ERR_ADMIN_REVOKED:
941 	case -NFS4ERR_DELEG_REVOKED:
942 	case -NFS4ERR_STALE_STATEID:
943 	case -NFS4ERR_BAD_STATEID:
944 		spin_lock(&inode->i_lock);
945 		if (pnfs_layout_is_valid(lo) &&
946 		    nfs4_stateid_match(&data->args.errors[0].stateid,
947 					     &lo->plh_stateid)) {
948 			LIST_HEAD(head);
949 
950 			/*
951 			 * Mark the bad layout state as invalid, then retry
952 			 * with the current stateid.
953 			 */
954 			pnfs_mark_layout_stateid_invalid(lo, &head);
955 			spin_unlock(&inode->i_lock);
956 			pnfs_free_lseg_list(&head);
957 			nfs_commit_inode(inode, 0);
958 		} else
959 			spin_unlock(&inode->i_lock);
960 		break;
961 	case -NFS4ERR_OLD_STATEID:
962 		spin_lock(&inode->i_lock);
963 		if (pnfs_layout_is_valid(lo) &&
964 		    nfs4_stateid_match_other(&data->args.errors[0].stateid,
965 					&lo->plh_stateid)) {
966 			/* Do we need to delay before resending? */
967 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
968 						&data->args.errors[0].stateid))
969 				rpc_delay(task, HZ);
970 			rpc_restart_call_prepare(task);
971 		}
972 		spin_unlock(&inode->i_lock);
973 		break;
974 	case -ENOTSUPP:
975 	case -EOPNOTSUPP:
976 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTERROR;
977 	}
978 
979 	trace_nfs4_layouterror(inode, &data->args.errors[0].stateid,
980 			       task->tk_status);
981 }
982 
983 static void
984 nfs42_layouterror_release(void *calldata)
985 {
986 	struct nfs42_layouterror_data *data = calldata;
987 
988 	nfs42_free_layouterror_data(data);
989 }
990 
991 static const struct rpc_call_ops nfs42_layouterror_ops = {
992 	.rpc_call_prepare = nfs42_layouterror_prepare,
993 	.rpc_call_done = nfs42_layouterror_done,
994 	.rpc_release = nfs42_layouterror_release,
995 };
996 
997 int nfs42_proc_layouterror(struct pnfs_layout_segment *lseg,
998 		const struct nfs42_layout_error *errors, size_t n)
999 {
1000 	struct inode *inode = lseg->pls_layout->plh_inode;
1001 	struct nfs42_layouterror_data *data;
1002 	struct rpc_task *task;
1003 	struct rpc_message msg = {
1004 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTERROR],
1005 	};
1006 	struct rpc_task_setup task_setup = {
1007 		.rpc_message = &msg,
1008 		.callback_ops = &nfs42_layouterror_ops,
1009 		.flags = RPC_TASK_ASYNC,
1010 	};
1011 	unsigned int i;
1012 
1013 	if (!nfs_server_capable(inode, NFS_CAP_LAYOUTERROR))
1014 		return -EOPNOTSUPP;
1015 	if (n > NFS42_LAYOUTERROR_MAX)
1016 		return -EINVAL;
1017 	data = nfs42_alloc_layouterror_data(lseg, GFP_NOFS);
1018 	if (!data)
1019 		return -ENOMEM;
1020 	for (i = 0; i < n; i++) {
1021 		data->args.errors[i] = errors[i];
1022 		data->args.num_errors++;
1023 		data->res.num_errors++;
1024 	}
1025 	msg.rpc_argp = &data->args;
1026 	msg.rpc_resp = &data->res;
1027 	task_setup.callback_data = data;
1028 	task_setup.rpc_client = NFS_SERVER(inode)->client;
1029 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
1030 	task = rpc_run_task(&task_setup);
1031 	if (IS_ERR(task))
1032 		return PTR_ERR(task);
1033 	rpc_put_task(task);
1034 	return 0;
1035 }
1036 EXPORT_SYMBOL_GPL(nfs42_proc_layouterror);
1037 
1038 static int _nfs42_proc_clone(struct rpc_message *msg, struct file *src_f,
1039 		struct file *dst_f, struct nfs_lock_context *src_lock,
1040 		struct nfs_lock_context *dst_lock, loff_t src_offset,
1041 		loff_t dst_offset, loff_t count)
1042 {
1043 	struct inode *src_inode = file_inode(src_f);
1044 	struct inode *dst_inode = file_inode(dst_f);
1045 	struct nfs_server *server = NFS_SERVER(dst_inode);
1046 	__u32 dst_bitmask[NFS_BITMASK_SZ];
1047 	struct nfs42_clone_args args = {
1048 		.src_fh = NFS_FH(src_inode),
1049 		.dst_fh = NFS_FH(dst_inode),
1050 		.src_offset = src_offset,
1051 		.dst_offset = dst_offset,
1052 		.count = count,
1053 		.dst_bitmask = dst_bitmask,
1054 	};
1055 	struct nfs42_clone_res res = {
1056 		.server	= server,
1057 	};
1058 	int status;
1059 
1060 	msg->rpc_argp = &args;
1061 	msg->rpc_resp = &res;
1062 
1063 	status = nfs4_set_rw_stateid(&args.src_stateid, src_lock->open_context,
1064 			src_lock, FMODE_READ);
1065 	if (status) {
1066 		if (status == -EAGAIN)
1067 			status = -NFS4ERR_BAD_STATEID;
1068 		return status;
1069 	}
1070 	status = nfs4_set_rw_stateid(&args.dst_stateid, dst_lock->open_context,
1071 			dst_lock, FMODE_WRITE);
1072 	if (status) {
1073 		if (status == -EAGAIN)
1074 			status = -NFS4ERR_BAD_STATEID;
1075 		return status;
1076 	}
1077 
1078 	res.dst_fattr = nfs_alloc_fattr();
1079 	if (!res.dst_fattr)
1080 		return -ENOMEM;
1081 
1082 	nfs4_bitmask_set(dst_bitmask, server->cache_consistency_bitmask,
1083 			 dst_inode, NFS_INO_INVALID_BLOCKS);
1084 
1085 	status = nfs4_call_sync(server->client, server, msg,
1086 				&args.seq_args, &res.seq_res, 0);
1087 	trace_nfs4_clone(src_inode, dst_inode, &args, status);
1088 	if (status == 0) {
1089 		nfs42_copy_dest_done(dst_inode, dst_offset, count);
1090 		status = nfs_post_op_update_inode(dst_inode, res.dst_fattr);
1091 	}
1092 
1093 	kfree(res.dst_fattr);
1094 	return status;
1095 }
1096 
1097 int nfs42_proc_clone(struct file *src_f, struct file *dst_f,
1098 		     loff_t src_offset, loff_t dst_offset, loff_t count)
1099 {
1100 	struct rpc_message msg = {
1101 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLONE],
1102 	};
1103 	struct inode *inode = file_inode(src_f);
1104 	struct nfs_server *server = NFS_SERVER(file_inode(src_f));
1105 	struct nfs_lock_context *src_lock;
1106 	struct nfs_lock_context *dst_lock;
1107 	struct nfs4_exception src_exception = { };
1108 	struct nfs4_exception dst_exception = { };
1109 	int err, err2;
1110 
1111 	if (!nfs_server_capable(inode, NFS_CAP_CLONE))
1112 		return -EOPNOTSUPP;
1113 
1114 	src_lock = nfs_get_lock_context(nfs_file_open_context(src_f));
1115 	if (IS_ERR(src_lock))
1116 		return PTR_ERR(src_lock);
1117 
1118 	src_exception.inode = file_inode(src_f);
1119 	src_exception.state = src_lock->open_context->state;
1120 
1121 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst_f));
1122 	if (IS_ERR(dst_lock)) {
1123 		err = PTR_ERR(dst_lock);
1124 		goto out_put_src_lock;
1125 	}
1126 
1127 	dst_exception.inode = file_inode(dst_f);
1128 	dst_exception.state = dst_lock->open_context->state;
1129 
1130 	do {
1131 		err = _nfs42_proc_clone(&msg, src_f, dst_f, src_lock, dst_lock,
1132 					src_offset, dst_offset, count);
1133 		if (err == -ENOTSUPP || err == -EOPNOTSUPP) {
1134 			NFS_SERVER(inode)->caps &= ~NFS_CAP_CLONE;
1135 			err = -EOPNOTSUPP;
1136 			break;
1137 		}
1138 
1139 		err2 = nfs4_handle_exception(server, err, &src_exception);
1140 		err = nfs4_handle_exception(server, err, &dst_exception);
1141 		if (!err)
1142 			err = err2;
1143 	} while (src_exception.retry || dst_exception.retry);
1144 
1145 	nfs_put_lock_context(dst_lock);
1146 out_put_src_lock:
1147 	nfs_put_lock_context(src_lock);
1148 	return err;
1149 }
1150 
1151 #define NFS4XATTR_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
1152 
1153 static int _nfs42_proc_removexattr(struct inode *inode, const char *name)
1154 {
1155 	struct nfs_server *server = NFS_SERVER(inode);
1156 	struct nfs42_removexattrargs args = {
1157 		.fh = NFS_FH(inode),
1158 		.xattr_name = name,
1159 	};
1160 	struct nfs42_removexattrres res;
1161 	struct rpc_message msg = {
1162 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVEXATTR],
1163 		.rpc_argp = &args,
1164 		.rpc_resp = &res,
1165 	};
1166 	int ret;
1167 	unsigned long timestamp = jiffies;
1168 
1169 	ret = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
1170 	    &res.seq_res, 1);
1171 	if (!ret)
1172 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1173 
1174 	return ret;
1175 }
1176 
1177 static int _nfs42_proc_setxattr(struct inode *inode, const char *name,
1178 				const void *buf, size_t buflen, int flags)
1179 {
1180 	struct nfs_server *server = NFS_SERVER(inode);
1181 	struct page *pages[NFS4XATTR_MAXPAGES];
1182 	struct nfs42_setxattrargs arg = {
1183 		.fh		= NFS_FH(inode),
1184 		.xattr_pages	= pages,
1185 		.xattr_len	= buflen,
1186 		.xattr_name	= name,
1187 		.xattr_flags	= flags,
1188 	};
1189 	struct nfs42_setxattrres res;
1190 	struct rpc_message msg = {
1191 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETXATTR],
1192 		.rpc_argp	= &arg,
1193 		.rpc_resp	= &res,
1194 	};
1195 	int ret, np;
1196 	unsigned long timestamp = jiffies;
1197 
1198 	if (buflen > server->sxasize)
1199 		return -ERANGE;
1200 
1201 	if (buflen > 0) {
1202 		np = nfs4_buf_to_pages_noslab(buf, buflen, arg.xattr_pages);
1203 		if (np < 0)
1204 			return np;
1205 	} else
1206 		np = 0;
1207 
1208 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1209 	    &res.seq_res, 1);
1210 
1211 	for (; np > 0; np--)
1212 		put_page(pages[np - 1]);
1213 
1214 	if (!ret)
1215 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1216 
1217 	return ret;
1218 }
1219 
1220 static ssize_t _nfs42_proc_getxattr(struct inode *inode, const char *name,
1221 				void *buf, size_t buflen, struct page **pages,
1222 				size_t plen)
1223 {
1224 	struct nfs_server *server = NFS_SERVER(inode);
1225 	struct nfs42_getxattrargs arg = {
1226 		.fh		= NFS_FH(inode),
1227 		.xattr_name	= name,
1228 	};
1229 	struct nfs42_getxattrres res;
1230 	struct rpc_message msg = {
1231 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETXATTR],
1232 		.rpc_argp	= &arg,
1233 		.rpc_resp	= &res,
1234 	};
1235 	ssize_t ret;
1236 
1237 	arg.xattr_len = plen;
1238 	arg.xattr_pages = pages;
1239 
1240 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1241 	    &res.seq_res, 0);
1242 	if (ret < 0)
1243 		return ret;
1244 
1245 	/*
1246 	 * Normally, the caching is done one layer up, but for successful
1247 	 * RPCS, always cache the result here, even if the caller was
1248 	 * just querying the length, or if the reply was too big for
1249 	 * the caller. This avoids a second RPC in the case of the
1250 	 * common query-alloc-retrieve cycle for xattrs.
1251 	 *
1252 	 * Note that xattr_len is always capped to XATTR_SIZE_MAX.
1253 	 */
1254 
1255 	nfs4_xattr_cache_add(inode, name, NULL, pages, res.xattr_len);
1256 
1257 	if (buflen) {
1258 		if (res.xattr_len > buflen)
1259 			return -ERANGE;
1260 		_copy_from_pages(buf, pages, 0, res.xattr_len);
1261 	}
1262 
1263 	return res.xattr_len;
1264 }
1265 
1266 static ssize_t _nfs42_proc_listxattrs(struct inode *inode, void *buf,
1267 				 size_t buflen, u64 *cookiep, bool *eofp)
1268 {
1269 	struct nfs_server *server = NFS_SERVER(inode);
1270 	struct page **pages;
1271 	struct nfs42_listxattrsargs arg = {
1272 		.fh		= NFS_FH(inode),
1273 		.cookie		= *cookiep,
1274 	};
1275 	struct nfs42_listxattrsres res = {
1276 		.eof = false,
1277 		.xattr_buf = buf,
1278 		.xattr_len = buflen,
1279 	};
1280 	struct rpc_message msg = {
1281 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LISTXATTRS],
1282 		.rpc_argp	= &arg,
1283 		.rpc_resp	= &res,
1284 	};
1285 	u32 xdrlen;
1286 	int ret, np, i;
1287 
1288 
1289 	ret = -ENOMEM;
1290 	res.scratch = alloc_page(GFP_KERNEL);
1291 	if (!res.scratch)
1292 		goto out;
1293 
1294 	xdrlen = nfs42_listxattr_xdrsize(buflen);
1295 	if (xdrlen > server->lxasize)
1296 		xdrlen = server->lxasize;
1297 	np = xdrlen / PAGE_SIZE + 1;
1298 
1299 	pages = kcalloc(np, sizeof(struct page *), GFP_KERNEL);
1300 	if (!pages)
1301 		goto out_free_scratch;
1302 	for (i = 0; i < np; i++) {
1303 		pages[i] = alloc_page(GFP_KERNEL);
1304 		if (!pages[i])
1305 			goto out_free_pages;
1306 	}
1307 
1308 	arg.xattr_pages = pages;
1309 	arg.count = xdrlen;
1310 
1311 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1312 	    &res.seq_res, 0);
1313 
1314 	if (ret >= 0) {
1315 		ret = res.copied;
1316 		*cookiep = res.cookie;
1317 		*eofp = res.eof;
1318 	}
1319 
1320 out_free_pages:
1321 	while (--np >= 0) {
1322 		if (pages[np])
1323 			__free_page(pages[np]);
1324 	}
1325 	kfree(pages);
1326 out_free_scratch:
1327 	__free_page(res.scratch);
1328 out:
1329 	return ret;
1330 
1331 }
1332 
1333 ssize_t nfs42_proc_getxattr(struct inode *inode, const char *name,
1334 			      void *buf, size_t buflen)
1335 {
1336 	struct nfs4_exception exception = { };
1337 	ssize_t err, np, i;
1338 	struct page **pages;
1339 
1340 	np = nfs_page_array_len(0, buflen ?: XATTR_SIZE_MAX);
1341 	pages = kmalloc_array(np, sizeof(*pages), GFP_KERNEL);
1342 	if (!pages)
1343 		return -ENOMEM;
1344 
1345 	for (i = 0; i < np; i++) {
1346 		pages[i] = alloc_page(GFP_KERNEL);
1347 		if (!pages[i]) {
1348 			np = i + 1;
1349 			err = -ENOMEM;
1350 			goto out;
1351 		}
1352 	}
1353 
1354 	/*
1355 	 * The GETXATTR op has no length field in the call, and the
1356 	 * xattr data is at the end of the reply.
1357 	 *
1358 	 * There is no downside in using the page-aligned length. It will
1359 	 * allow receiving and caching xattrs that are too large for the
1360 	 * caller but still fit in the page-rounded value.
1361 	 */
1362 	do {
1363 		err = _nfs42_proc_getxattr(inode, name, buf, buflen,
1364 			pages, np * PAGE_SIZE);
1365 		if (err >= 0)
1366 			break;
1367 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1368 				&exception);
1369 	} while (exception.retry);
1370 
1371 out:
1372 	while (--np >= 0)
1373 		__free_page(pages[np]);
1374 	kfree(pages);
1375 
1376 	return err;
1377 }
1378 
1379 int nfs42_proc_setxattr(struct inode *inode, const char *name,
1380 			      const void *buf, size_t buflen, int flags)
1381 {
1382 	struct nfs4_exception exception = { };
1383 	int err;
1384 
1385 	do {
1386 		err = _nfs42_proc_setxattr(inode, name, buf, buflen, flags);
1387 		if (!err)
1388 			break;
1389 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1390 				&exception);
1391 	} while (exception.retry);
1392 
1393 	return err;
1394 }
1395 
1396 ssize_t nfs42_proc_listxattrs(struct inode *inode, void *buf,
1397 			      size_t buflen, u64 *cookiep, bool *eofp)
1398 {
1399 	struct nfs4_exception exception = { };
1400 	ssize_t err;
1401 
1402 	do {
1403 		err = _nfs42_proc_listxattrs(inode, buf, buflen,
1404 		    cookiep, eofp);
1405 		if (err >= 0)
1406 			break;
1407 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1408 				&exception);
1409 	} while (exception.retry);
1410 
1411 	return err;
1412 }
1413 
1414 int nfs42_proc_removexattr(struct inode *inode, const char *name)
1415 {
1416 	struct nfs4_exception exception = { };
1417 	int err;
1418 
1419 	do {
1420 		err = _nfs42_proc_removexattr(inode, name);
1421 		if (!err)
1422 			break;
1423 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1424 				&exception);
1425 	} while (exception.retry);
1426 
1427 	return err;
1428 }
1429