xref: /openbmc/linux/fs/nfs/nfs42proc.c (revision fbb6b31a)
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 = NULL, *iter;
179 	int status = NFS4_OK;
180 	struct nfs_open_context *dst_ctx = nfs_file_open_context(dst);
181 	struct nfs_open_context *src_ctx = nfs_file_open_context(src);
182 
183 	copy = kzalloc(sizeof(struct nfs4_copy_state), GFP_KERNEL);
184 	if (!copy)
185 		return -ENOMEM;
186 
187 	spin_lock(&dst_server->nfs_client->cl_lock);
188 	list_for_each_entry(iter,
189 				&dst_server->nfs_client->pending_cb_stateids,
190 				copies) {
191 		if (memcmp(&res->write_res.stateid, &iter->stateid,
192 				NFS4_STATEID_SIZE))
193 			continue;
194 		tmp_copy = iter;
195 		list_del(&iter->copies);
196 		break;
197 	}
198 	if (tmp_copy) {
199 		spin_unlock(&dst_server->nfs_client->cl_lock);
200 		kfree(copy);
201 		copy = tmp_copy;
202 		goto out;
203 	}
204 
205 	memcpy(&copy->stateid, &res->write_res.stateid, NFS4_STATEID_SIZE);
206 	init_completion(&copy->completion);
207 	copy->parent_dst_state = dst_ctx->state;
208 	copy->parent_src_state = src_ctx->state;
209 
210 	list_add_tail(&copy->copies, &dst_server->ss_copies);
211 	spin_unlock(&dst_server->nfs_client->cl_lock);
212 
213 	if (dst_server != src_server) {
214 		spin_lock(&src_server->nfs_client->cl_lock);
215 		list_add_tail(&copy->src_copies, &src_server->ss_copies);
216 		spin_unlock(&src_server->nfs_client->cl_lock);
217 	}
218 
219 	status = wait_for_completion_interruptible(&copy->completion);
220 	spin_lock(&dst_server->nfs_client->cl_lock);
221 	list_del_init(&copy->copies);
222 	spin_unlock(&dst_server->nfs_client->cl_lock);
223 	if (dst_server != src_server) {
224 		spin_lock(&src_server->nfs_client->cl_lock);
225 		list_del_init(&copy->src_copies);
226 		spin_unlock(&src_server->nfs_client->cl_lock);
227 	}
228 	if (status == -ERESTARTSYS) {
229 		goto out_cancel;
230 	} else if (copy->flags || copy->error == NFS4ERR_PARTNER_NO_AUTH) {
231 		status = -EAGAIN;
232 		*restart = true;
233 		goto out_cancel;
234 	}
235 out:
236 	res->write_res.count = copy->count;
237 	memcpy(&res->write_res.verifier, &copy->verf, sizeof(copy->verf));
238 	status = -copy->error;
239 
240 out_free:
241 	kfree(copy);
242 	return status;
243 out_cancel:
244 	nfs42_do_offload_cancel_async(dst, &copy->stateid);
245 	if (!nfs42_files_from_same_server(src, dst))
246 		nfs42_do_offload_cancel_async(src, src_stateid);
247 	goto out_free;
248 }
249 
250 static int process_copy_commit(struct file *dst, loff_t pos_dst,
251 			       struct nfs42_copy_res *res)
252 {
253 	struct nfs_commitres cres;
254 	int status = -ENOMEM;
255 
256 	cres.verf = kzalloc(sizeof(struct nfs_writeverf), GFP_KERNEL);
257 	if (!cres.verf)
258 		goto out;
259 
260 	status = nfs4_proc_commit(dst, pos_dst, res->write_res.count, &cres);
261 	if (status)
262 		goto out_free;
263 	if (nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
264 				    &cres.verf->verifier)) {
265 		dprintk("commit verf differs from copy verf\n");
266 		status = -EAGAIN;
267 	}
268 out_free:
269 	kfree(cres.verf);
270 out:
271 	return status;
272 }
273 
274 /**
275  * nfs42_copy_dest_done - perform inode cache updates after clone/copy offload
276  * @inode: pointer to destination inode
277  * @pos: destination offset
278  * @len: copy length
279  *
280  * Punch a hole in the inode page cache, so that the NFS client will
281  * know to retrieve new data.
282  * Update the file size if necessary, and then mark the inode as having
283  * invalid cached values for change attribute, ctime, mtime and space used.
284  */
285 static void nfs42_copy_dest_done(struct inode *inode, loff_t pos, loff_t len)
286 {
287 	loff_t newsize = pos + len;
288 	loff_t end = newsize - 1;
289 
290 	WARN_ON_ONCE(invalidate_inode_pages2_range(inode->i_mapping,
291 				pos >> PAGE_SHIFT, end >> PAGE_SHIFT));
292 
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_KERNEL);
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_KERNEL);
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 		status = PTR_ERR(l_ctx);
595 		goto out;
596 	}
597 
598 	status = nfs4_set_rw_stateid(&args->cna_src_stateid, ctx, l_ctx,
599 				     FMODE_READ);
600 	nfs_put_lock_context(l_ctx);
601 	if (status) {
602 		if (status == -EAGAIN)
603 			status = -NFS4ERR_BAD_STATEID;
604 		goto out;
605 	}
606 
607 	status = nfs4_call_sync(src_server->client, src_server, &msg,
608 				&args->cna_seq_args, &res->cnr_seq_res, 0);
609 	trace_nfs4_copy_notify(file_inode(src), args, res, status);
610 	if (status == -ENOTSUPP)
611 		src_server->caps &= ~NFS_CAP_COPY_NOTIFY;
612 
613 out:
614 	put_nfs_open_context(nfs_file_open_context(src));
615 	return status;
616 }
617 
618 int nfs42_proc_copy_notify(struct file *src, struct file *dst,
619 				struct nfs42_copy_notify_res *res)
620 {
621 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
622 	struct nfs42_copy_notify_args *args;
623 	struct nfs4_exception exception = {
624 		.inode = file_inode(src),
625 	};
626 	int status;
627 
628 	if (!(src_server->caps & NFS_CAP_COPY_NOTIFY))
629 		return -EOPNOTSUPP;
630 
631 	args = kzalloc(sizeof(struct nfs42_copy_notify_args), GFP_KERNEL);
632 	if (args == NULL)
633 		return -ENOMEM;
634 
635 	args->cna_src_fh  = NFS_FH(file_inode(src)),
636 	args->cna_dst.nl4_type = NL4_NETADDR;
637 	nfs42_set_netaddr(dst, &args->cna_dst.u.nl4_addr);
638 	exception.stateid = &args->cna_src_stateid;
639 
640 	do {
641 		status = _nfs42_proc_copy_notify(src, dst, args, res);
642 		if (status == -ENOTSUPP) {
643 			status = -EOPNOTSUPP;
644 			goto out;
645 		}
646 		status = nfs4_handle_exception(src_server, status, &exception);
647 	} while (exception.retry);
648 
649 out:
650 	kfree(args);
651 	return status;
652 }
653 
654 static loff_t _nfs42_proc_llseek(struct file *filep,
655 		struct nfs_lock_context *lock, loff_t offset, int whence)
656 {
657 	struct inode *inode = file_inode(filep);
658 	struct nfs42_seek_args args = {
659 		.sa_fh		= NFS_FH(inode),
660 		.sa_offset	= offset,
661 		.sa_what	= (whence == SEEK_HOLE) ?
662 					NFS4_CONTENT_HOLE : NFS4_CONTENT_DATA,
663 	};
664 	struct nfs42_seek_res res;
665 	struct rpc_message msg = {
666 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEEK],
667 		.rpc_argp = &args,
668 		.rpc_resp = &res,
669 	};
670 	struct nfs_server *server = NFS_SERVER(inode);
671 	int status;
672 
673 	if (!nfs_server_capable(inode, NFS_CAP_SEEK))
674 		return -ENOTSUPP;
675 
676 	status = nfs4_set_rw_stateid(&args.sa_stateid, lock->open_context,
677 			lock, FMODE_READ);
678 	if (status) {
679 		if (status == -EAGAIN)
680 			status = -NFS4ERR_BAD_STATEID;
681 		return status;
682 	}
683 
684 	status = nfs_filemap_write_and_wait_range(inode->i_mapping,
685 			offset, LLONG_MAX);
686 	if (status)
687 		return status;
688 
689 	status = nfs4_call_sync(server->client, server, &msg,
690 				&args.seq_args, &res.seq_res, 0);
691 	trace_nfs4_llseek(inode, &args, &res, status);
692 	if (status == -ENOTSUPP)
693 		server->caps &= ~NFS_CAP_SEEK;
694 	if (status)
695 		return status;
696 
697 	if (whence == SEEK_DATA && res.sr_eof)
698 		return -NFS4ERR_NXIO;
699 	else
700 		return vfs_setpos(filep, res.sr_offset, inode->i_sb->s_maxbytes);
701 }
702 
703 loff_t nfs42_proc_llseek(struct file *filep, loff_t offset, int whence)
704 {
705 	struct nfs_server *server = NFS_SERVER(file_inode(filep));
706 	struct nfs4_exception exception = { };
707 	struct nfs_lock_context *lock;
708 	loff_t err;
709 
710 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
711 	if (IS_ERR(lock))
712 		return PTR_ERR(lock);
713 
714 	exception.inode = file_inode(filep);
715 	exception.state = lock->open_context->state;
716 
717 	do {
718 		err = _nfs42_proc_llseek(filep, lock, offset, whence);
719 		if (err >= 0)
720 			break;
721 		if (err == -ENOTSUPP) {
722 			err = -EOPNOTSUPP;
723 			break;
724 		}
725 		err = nfs4_handle_exception(server, err, &exception);
726 	} while (exception.retry);
727 
728 	nfs_put_lock_context(lock);
729 	return err;
730 }
731 
732 
733 static void
734 nfs42_layoutstat_prepare(struct rpc_task *task, void *calldata)
735 {
736 	struct nfs42_layoutstat_data *data = calldata;
737 	struct inode *inode = data->inode;
738 	struct nfs_server *server = NFS_SERVER(inode);
739 	struct pnfs_layout_hdr *lo;
740 
741 	spin_lock(&inode->i_lock);
742 	lo = NFS_I(inode)->layout;
743 	if (!pnfs_layout_is_valid(lo)) {
744 		spin_unlock(&inode->i_lock);
745 		rpc_exit(task, 0);
746 		return;
747 	}
748 	nfs4_stateid_copy(&data->args.stateid, &lo->plh_stateid);
749 	spin_unlock(&inode->i_lock);
750 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
751 			    &data->res.seq_res, task);
752 }
753 
754 static void
755 nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
756 {
757 	struct nfs42_layoutstat_data *data = calldata;
758 	struct inode *inode = data->inode;
759 	struct pnfs_layout_hdr *lo;
760 
761 	if (!nfs4_sequence_done(task, &data->res.seq_res))
762 		return;
763 
764 	switch (task->tk_status) {
765 	case 0:
766 		return;
767 	case -NFS4ERR_BADHANDLE:
768 	case -ESTALE:
769 		pnfs_destroy_layout(NFS_I(inode));
770 		break;
771 	case -NFS4ERR_EXPIRED:
772 	case -NFS4ERR_ADMIN_REVOKED:
773 	case -NFS4ERR_DELEG_REVOKED:
774 	case -NFS4ERR_STALE_STATEID:
775 	case -NFS4ERR_BAD_STATEID:
776 		spin_lock(&inode->i_lock);
777 		lo = NFS_I(inode)->layout;
778 		if (pnfs_layout_is_valid(lo) &&
779 		    nfs4_stateid_match(&data->args.stateid,
780 					     &lo->plh_stateid)) {
781 			LIST_HEAD(head);
782 
783 			/*
784 			 * Mark the bad layout state as invalid, then retry
785 			 * with the current stateid.
786 			 */
787 			pnfs_mark_layout_stateid_invalid(lo, &head);
788 			spin_unlock(&inode->i_lock);
789 			pnfs_free_lseg_list(&head);
790 			nfs_commit_inode(inode, 0);
791 		} else
792 			spin_unlock(&inode->i_lock);
793 		break;
794 	case -NFS4ERR_OLD_STATEID:
795 		spin_lock(&inode->i_lock);
796 		lo = NFS_I(inode)->layout;
797 		if (pnfs_layout_is_valid(lo) &&
798 		    nfs4_stateid_match_other(&data->args.stateid,
799 					&lo->plh_stateid)) {
800 			/* Do we need to delay before resending? */
801 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
802 						&data->args.stateid))
803 				rpc_delay(task, HZ);
804 			rpc_restart_call_prepare(task);
805 		}
806 		spin_unlock(&inode->i_lock);
807 		break;
808 	case -ENOTSUPP:
809 	case -EOPNOTSUPP:
810 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTSTATS;
811 	}
812 
813 	trace_nfs4_layoutstats(inode, &data->args.stateid, task->tk_status);
814 }
815 
816 static void
817 nfs42_layoutstat_release(void *calldata)
818 {
819 	struct nfs42_layoutstat_data *data = calldata;
820 	struct nfs42_layoutstat_devinfo *devinfo = data->args.devinfo;
821 	int i;
822 
823 	for (i = 0; i < data->args.num_dev; i++) {
824 		if (devinfo[i].ld_private.ops && devinfo[i].ld_private.ops->free)
825 			devinfo[i].ld_private.ops->free(&devinfo[i].ld_private);
826 	}
827 
828 	pnfs_put_layout_hdr(NFS_I(data->args.inode)->layout);
829 	smp_mb__before_atomic();
830 	clear_bit(NFS_INO_LAYOUTSTATS, &NFS_I(data->args.inode)->flags);
831 	smp_mb__after_atomic();
832 	nfs_iput_and_deactive(data->inode);
833 	kfree(data->args.devinfo);
834 	kfree(data);
835 }
836 
837 static const struct rpc_call_ops nfs42_layoutstat_ops = {
838 	.rpc_call_prepare = nfs42_layoutstat_prepare,
839 	.rpc_call_done = nfs42_layoutstat_done,
840 	.rpc_release = nfs42_layoutstat_release,
841 };
842 
843 int nfs42_proc_layoutstats_generic(struct nfs_server *server,
844 				   struct nfs42_layoutstat_data *data)
845 {
846 	struct rpc_message msg = {
847 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTSTATS],
848 		.rpc_argp = &data->args,
849 		.rpc_resp = &data->res,
850 	};
851 	struct rpc_task_setup task_setup = {
852 		.rpc_client = server->client,
853 		.rpc_message = &msg,
854 		.callback_ops = &nfs42_layoutstat_ops,
855 		.callback_data = data,
856 		.flags = RPC_TASK_ASYNC,
857 	};
858 	struct rpc_task *task;
859 
860 	data->inode = nfs_igrab_and_active(data->args.inode);
861 	if (!data->inode) {
862 		nfs42_layoutstat_release(data);
863 		return -EAGAIN;
864 	}
865 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
866 	task = rpc_run_task(&task_setup);
867 	if (IS_ERR(task))
868 		return PTR_ERR(task);
869 	rpc_put_task(task);
870 	return 0;
871 }
872 
873 static struct nfs42_layouterror_data *
874 nfs42_alloc_layouterror_data(struct pnfs_layout_segment *lseg, gfp_t gfp_flags)
875 {
876 	struct nfs42_layouterror_data *data;
877 	struct inode *inode = lseg->pls_layout->plh_inode;
878 
879 	data = kzalloc(sizeof(*data), gfp_flags);
880 	if (data) {
881 		data->args.inode = data->inode = nfs_igrab_and_active(inode);
882 		if (data->inode) {
883 			data->lseg = pnfs_get_lseg(lseg);
884 			if (data->lseg)
885 				return data;
886 			nfs_iput_and_deactive(data->inode);
887 		}
888 		kfree(data);
889 	}
890 	return NULL;
891 }
892 
893 static void
894 nfs42_free_layouterror_data(struct nfs42_layouterror_data *data)
895 {
896 	pnfs_put_lseg(data->lseg);
897 	nfs_iput_and_deactive(data->inode);
898 	kfree(data);
899 }
900 
901 static void
902 nfs42_layouterror_prepare(struct rpc_task *task, void *calldata)
903 {
904 	struct nfs42_layouterror_data *data = calldata;
905 	struct inode *inode = data->inode;
906 	struct nfs_server *server = NFS_SERVER(inode);
907 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
908 	unsigned i;
909 
910 	spin_lock(&inode->i_lock);
911 	if (!pnfs_layout_is_valid(lo)) {
912 		spin_unlock(&inode->i_lock);
913 		rpc_exit(task, 0);
914 		return;
915 	}
916 	for (i = 0; i < data->args.num_errors; i++)
917 		nfs4_stateid_copy(&data->args.errors[i].stateid,
918 				&lo->plh_stateid);
919 	spin_unlock(&inode->i_lock);
920 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
921 			    &data->res.seq_res, task);
922 }
923 
924 static void
925 nfs42_layouterror_done(struct rpc_task *task, void *calldata)
926 {
927 	struct nfs42_layouterror_data *data = calldata;
928 	struct inode *inode = data->inode;
929 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
930 
931 	if (!nfs4_sequence_done(task, &data->res.seq_res))
932 		return;
933 
934 	switch (task->tk_status) {
935 	case 0:
936 		return;
937 	case -NFS4ERR_BADHANDLE:
938 	case -ESTALE:
939 		pnfs_destroy_layout(NFS_I(inode));
940 		break;
941 	case -NFS4ERR_EXPIRED:
942 	case -NFS4ERR_ADMIN_REVOKED:
943 	case -NFS4ERR_DELEG_REVOKED:
944 	case -NFS4ERR_STALE_STATEID:
945 	case -NFS4ERR_BAD_STATEID:
946 		spin_lock(&inode->i_lock);
947 		if (pnfs_layout_is_valid(lo) &&
948 		    nfs4_stateid_match(&data->args.errors[0].stateid,
949 					     &lo->plh_stateid)) {
950 			LIST_HEAD(head);
951 
952 			/*
953 			 * Mark the bad layout state as invalid, then retry
954 			 * with the current stateid.
955 			 */
956 			pnfs_mark_layout_stateid_invalid(lo, &head);
957 			spin_unlock(&inode->i_lock);
958 			pnfs_free_lseg_list(&head);
959 			nfs_commit_inode(inode, 0);
960 		} else
961 			spin_unlock(&inode->i_lock);
962 		break;
963 	case -NFS4ERR_OLD_STATEID:
964 		spin_lock(&inode->i_lock);
965 		if (pnfs_layout_is_valid(lo) &&
966 		    nfs4_stateid_match_other(&data->args.errors[0].stateid,
967 					&lo->plh_stateid)) {
968 			/* Do we need to delay before resending? */
969 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
970 						&data->args.errors[0].stateid))
971 				rpc_delay(task, HZ);
972 			rpc_restart_call_prepare(task);
973 		}
974 		spin_unlock(&inode->i_lock);
975 		break;
976 	case -ENOTSUPP:
977 	case -EOPNOTSUPP:
978 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTERROR;
979 	}
980 
981 	trace_nfs4_layouterror(inode, &data->args.errors[0].stateid,
982 			       task->tk_status);
983 }
984 
985 static void
986 nfs42_layouterror_release(void *calldata)
987 {
988 	struct nfs42_layouterror_data *data = calldata;
989 
990 	nfs42_free_layouterror_data(data);
991 }
992 
993 static const struct rpc_call_ops nfs42_layouterror_ops = {
994 	.rpc_call_prepare = nfs42_layouterror_prepare,
995 	.rpc_call_done = nfs42_layouterror_done,
996 	.rpc_release = nfs42_layouterror_release,
997 };
998 
999 int nfs42_proc_layouterror(struct pnfs_layout_segment *lseg,
1000 		const struct nfs42_layout_error *errors, size_t n)
1001 {
1002 	struct inode *inode = lseg->pls_layout->plh_inode;
1003 	struct nfs42_layouterror_data *data;
1004 	struct rpc_task *task;
1005 	struct rpc_message msg = {
1006 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTERROR],
1007 	};
1008 	struct rpc_task_setup task_setup = {
1009 		.rpc_message = &msg,
1010 		.callback_ops = &nfs42_layouterror_ops,
1011 		.flags = RPC_TASK_ASYNC,
1012 	};
1013 	unsigned int i;
1014 
1015 	if (!nfs_server_capable(inode, NFS_CAP_LAYOUTERROR))
1016 		return -EOPNOTSUPP;
1017 	if (n > NFS42_LAYOUTERROR_MAX)
1018 		return -EINVAL;
1019 	data = nfs42_alloc_layouterror_data(lseg, nfs_io_gfp_mask());
1020 	if (!data)
1021 		return -ENOMEM;
1022 	for (i = 0; i < n; i++) {
1023 		data->args.errors[i] = errors[i];
1024 		data->args.num_errors++;
1025 		data->res.num_errors++;
1026 	}
1027 	msg.rpc_argp = &data->args;
1028 	msg.rpc_resp = &data->res;
1029 	task_setup.callback_data = data;
1030 	task_setup.rpc_client = NFS_SERVER(inode)->client;
1031 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
1032 	task = rpc_run_task(&task_setup);
1033 	if (IS_ERR(task))
1034 		return PTR_ERR(task);
1035 	rpc_put_task(task);
1036 	return 0;
1037 }
1038 EXPORT_SYMBOL_GPL(nfs42_proc_layouterror);
1039 
1040 static int _nfs42_proc_clone(struct rpc_message *msg, struct file *src_f,
1041 		struct file *dst_f, struct nfs_lock_context *src_lock,
1042 		struct nfs_lock_context *dst_lock, loff_t src_offset,
1043 		loff_t dst_offset, loff_t count)
1044 {
1045 	struct inode *src_inode = file_inode(src_f);
1046 	struct inode *dst_inode = file_inode(dst_f);
1047 	struct nfs_server *server = NFS_SERVER(dst_inode);
1048 	__u32 dst_bitmask[NFS_BITMASK_SZ];
1049 	struct nfs42_clone_args args = {
1050 		.src_fh = NFS_FH(src_inode),
1051 		.dst_fh = NFS_FH(dst_inode),
1052 		.src_offset = src_offset,
1053 		.dst_offset = dst_offset,
1054 		.count = count,
1055 		.dst_bitmask = dst_bitmask,
1056 	};
1057 	struct nfs42_clone_res res = {
1058 		.server	= server,
1059 	};
1060 	int status;
1061 
1062 	msg->rpc_argp = &args;
1063 	msg->rpc_resp = &res;
1064 
1065 	status = nfs4_set_rw_stateid(&args.src_stateid, src_lock->open_context,
1066 			src_lock, FMODE_READ);
1067 	if (status) {
1068 		if (status == -EAGAIN)
1069 			status = -NFS4ERR_BAD_STATEID;
1070 		return status;
1071 	}
1072 	status = nfs4_set_rw_stateid(&args.dst_stateid, dst_lock->open_context,
1073 			dst_lock, FMODE_WRITE);
1074 	if (status) {
1075 		if (status == -EAGAIN)
1076 			status = -NFS4ERR_BAD_STATEID;
1077 		return status;
1078 	}
1079 
1080 	res.dst_fattr = nfs_alloc_fattr();
1081 	if (!res.dst_fattr)
1082 		return -ENOMEM;
1083 
1084 	nfs4_bitmask_set(dst_bitmask, server->cache_consistency_bitmask,
1085 			 dst_inode, NFS_INO_INVALID_BLOCKS);
1086 
1087 	status = nfs4_call_sync(server->client, server, msg,
1088 				&args.seq_args, &res.seq_res, 0);
1089 	trace_nfs4_clone(src_inode, dst_inode, &args, status);
1090 	if (status == 0) {
1091 		nfs42_copy_dest_done(dst_inode, dst_offset, count);
1092 		status = nfs_post_op_update_inode(dst_inode, res.dst_fattr);
1093 	}
1094 
1095 	kfree(res.dst_fattr);
1096 	return status;
1097 }
1098 
1099 int nfs42_proc_clone(struct file *src_f, struct file *dst_f,
1100 		     loff_t src_offset, loff_t dst_offset, loff_t count)
1101 {
1102 	struct rpc_message msg = {
1103 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLONE],
1104 	};
1105 	struct inode *inode = file_inode(src_f);
1106 	struct nfs_server *server = NFS_SERVER(file_inode(src_f));
1107 	struct nfs_lock_context *src_lock;
1108 	struct nfs_lock_context *dst_lock;
1109 	struct nfs4_exception src_exception = { };
1110 	struct nfs4_exception dst_exception = { };
1111 	int err, err2;
1112 
1113 	if (!nfs_server_capable(inode, NFS_CAP_CLONE))
1114 		return -EOPNOTSUPP;
1115 
1116 	src_lock = nfs_get_lock_context(nfs_file_open_context(src_f));
1117 	if (IS_ERR(src_lock))
1118 		return PTR_ERR(src_lock);
1119 
1120 	src_exception.inode = file_inode(src_f);
1121 	src_exception.state = src_lock->open_context->state;
1122 
1123 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst_f));
1124 	if (IS_ERR(dst_lock)) {
1125 		err = PTR_ERR(dst_lock);
1126 		goto out_put_src_lock;
1127 	}
1128 
1129 	dst_exception.inode = file_inode(dst_f);
1130 	dst_exception.state = dst_lock->open_context->state;
1131 
1132 	do {
1133 		err = _nfs42_proc_clone(&msg, src_f, dst_f, src_lock, dst_lock,
1134 					src_offset, dst_offset, count);
1135 		if (err == -ENOTSUPP || err == -EOPNOTSUPP) {
1136 			NFS_SERVER(inode)->caps &= ~NFS_CAP_CLONE;
1137 			err = -EOPNOTSUPP;
1138 			break;
1139 		}
1140 
1141 		err2 = nfs4_handle_exception(server, err, &src_exception);
1142 		err = nfs4_handle_exception(server, err, &dst_exception);
1143 		if (!err)
1144 			err = err2;
1145 	} while (src_exception.retry || dst_exception.retry);
1146 
1147 	nfs_put_lock_context(dst_lock);
1148 out_put_src_lock:
1149 	nfs_put_lock_context(src_lock);
1150 	return err;
1151 }
1152 
1153 #define NFS4XATTR_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
1154 
1155 static int _nfs42_proc_removexattr(struct inode *inode, const char *name)
1156 {
1157 	struct nfs_server *server = NFS_SERVER(inode);
1158 	struct nfs42_removexattrargs args = {
1159 		.fh = NFS_FH(inode),
1160 		.xattr_name = name,
1161 	};
1162 	struct nfs42_removexattrres res;
1163 	struct rpc_message msg = {
1164 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVEXATTR],
1165 		.rpc_argp = &args,
1166 		.rpc_resp = &res,
1167 	};
1168 	int ret;
1169 	unsigned long timestamp = jiffies;
1170 
1171 	ret = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
1172 	    &res.seq_res, 1);
1173 	if (!ret)
1174 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1175 
1176 	return ret;
1177 }
1178 
1179 static int _nfs42_proc_setxattr(struct inode *inode, const char *name,
1180 				const void *buf, size_t buflen, int flags)
1181 {
1182 	struct nfs_server *server = NFS_SERVER(inode);
1183 	struct page *pages[NFS4XATTR_MAXPAGES];
1184 	struct nfs42_setxattrargs arg = {
1185 		.fh		= NFS_FH(inode),
1186 		.xattr_pages	= pages,
1187 		.xattr_len	= buflen,
1188 		.xattr_name	= name,
1189 		.xattr_flags	= flags,
1190 	};
1191 	struct nfs42_setxattrres res;
1192 	struct rpc_message msg = {
1193 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETXATTR],
1194 		.rpc_argp	= &arg,
1195 		.rpc_resp	= &res,
1196 	};
1197 	int ret, np;
1198 	unsigned long timestamp = jiffies;
1199 
1200 	if (buflen > server->sxasize)
1201 		return -ERANGE;
1202 
1203 	if (buflen > 0) {
1204 		np = nfs4_buf_to_pages_noslab(buf, buflen, arg.xattr_pages);
1205 		if (np < 0)
1206 			return np;
1207 	} else
1208 		np = 0;
1209 
1210 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1211 	    &res.seq_res, 1);
1212 
1213 	for (; np > 0; np--)
1214 		put_page(pages[np - 1]);
1215 
1216 	if (!ret)
1217 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1218 
1219 	return ret;
1220 }
1221 
1222 static ssize_t _nfs42_proc_getxattr(struct inode *inode, const char *name,
1223 				void *buf, size_t buflen, struct page **pages,
1224 				size_t plen)
1225 {
1226 	struct nfs_server *server = NFS_SERVER(inode);
1227 	struct nfs42_getxattrargs arg = {
1228 		.fh		= NFS_FH(inode),
1229 		.xattr_name	= name,
1230 	};
1231 	struct nfs42_getxattrres res;
1232 	struct rpc_message msg = {
1233 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETXATTR],
1234 		.rpc_argp	= &arg,
1235 		.rpc_resp	= &res,
1236 	};
1237 	ssize_t ret;
1238 
1239 	arg.xattr_len = plen;
1240 	arg.xattr_pages = pages;
1241 
1242 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1243 	    &res.seq_res, 0);
1244 	if (ret < 0)
1245 		return ret;
1246 
1247 	/*
1248 	 * Normally, the caching is done one layer up, but for successful
1249 	 * RPCS, always cache the result here, even if the caller was
1250 	 * just querying the length, or if the reply was too big for
1251 	 * the caller. This avoids a second RPC in the case of the
1252 	 * common query-alloc-retrieve cycle for xattrs.
1253 	 *
1254 	 * Note that xattr_len is always capped to XATTR_SIZE_MAX.
1255 	 */
1256 
1257 	nfs4_xattr_cache_add(inode, name, NULL, pages, res.xattr_len);
1258 
1259 	if (buflen) {
1260 		if (res.xattr_len > buflen)
1261 			return -ERANGE;
1262 		_copy_from_pages(buf, pages, 0, res.xattr_len);
1263 	}
1264 
1265 	return res.xattr_len;
1266 }
1267 
1268 static ssize_t _nfs42_proc_listxattrs(struct inode *inode, void *buf,
1269 				 size_t buflen, u64 *cookiep, bool *eofp)
1270 {
1271 	struct nfs_server *server = NFS_SERVER(inode);
1272 	struct page **pages;
1273 	struct nfs42_listxattrsargs arg = {
1274 		.fh		= NFS_FH(inode),
1275 		.cookie		= *cookiep,
1276 	};
1277 	struct nfs42_listxattrsres res = {
1278 		.eof = false,
1279 		.xattr_buf = buf,
1280 		.xattr_len = buflen,
1281 	};
1282 	struct rpc_message msg = {
1283 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LISTXATTRS],
1284 		.rpc_argp	= &arg,
1285 		.rpc_resp	= &res,
1286 	};
1287 	u32 xdrlen;
1288 	int ret, np, i;
1289 
1290 
1291 	ret = -ENOMEM;
1292 	res.scratch = alloc_page(GFP_KERNEL);
1293 	if (!res.scratch)
1294 		goto out;
1295 
1296 	xdrlen = nfs42_listxattr_xdrsize(buflen);
1297 	if (xdrlen > server->lxasize)
1298 		xdrlen = server->lxasize;
1299 	np = xdrlen / PAGE_SIZE + 1;
1300 
1301 	pages = kcalloc(np, sizeof(struct page *), GFP_KERNEL);
1302 	if (!pages)
1303 		goto out_free_scratch;
1304 	for (i = 0; i < np; i++) {
1305 		pages[i] = alloc_page(GFP_KERNEL);
1306 		if (!pages[i])
1307 			goto out_free_pages;
1308 	}
1309 
1310 	arg.xattr_pages = pages;
1311 	arg.count = xdrlen;
1312 
1313 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1314 	    &res.seq_res, 0);
1315 
1316 	if (ret >= 0) {
1317 		ret = res.copied;
1318 		*cookiep = res.cookie;
1319 		*eofp = res.eof;
1320 	}
1321 
1322 out_free_pages:
1323 	while (--np >= 0) {
1324 		if (pages[np])
1325 			__free_page(pages[np]);
1326 	}
1327 	kfree(pages);
1328 out_free_scratch:
1329 	__free_page(res.scratch);
1330 out:
1331 	return ret;
1332 
1333 }
1334 
1335 ssize_t nfs42_proc_getxattr(struct inode *inode, const char *name,
1336 			      void *buf, size_t buflen)
1337 {
1338 	struct nfs4_exception exception = { };
1339 	ssize_t err, np, i;
1340 	struct page **pages;
1341 
1342 	np = nfs_page_array_len(0, buflen ?: XATTR_SIZE_MAX);
1343 	pages = kmalloc_array(np, sizeof(*pages), GFP_KERNEL);
1344 	if (!pages)
1345 		return -ENOMEM;
1346 
1347 	for (i = 0; i < np; i++) {
1348 		pages[i] = alloc_page(GFP_KERNEL);
1349 		if (!pages[i]) {
1350 			np = i + 1;
1351 			err = -ENOMEM;
1352 			goto out;
1353 		}
1354 	}
1355 
1356 	/*
1357 	 * The GETXATTR op has no length field in the call, and the
1358 	 * xattr data is at the end of the reply.
1359 	 *
1360 	 * There is no downside in using the page-aligned length. It will
1361 	 * allow receiving and caching xattrs that are too large for the
1362 	 * caller but still fit in the page-rounded value.
1363 	 */
1364 	do {
1365 		err = _nfs42_proc_getxattr(inode, name, buf, buflen,
1366 			pages, np * PAGE_SIZE);
1367 		if (err >= 0)
1368 			break;
1369 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1370 				&exception);
1371 	} while (exception.retry);
1372 
1373 out:
1374 	while (--np >= 0)
1375 		__free_page(pages[np]);
1376 	kfree(pages);
1377 
1378 	return err;
1379 }
1380 
1381 int nfs42_proc_setxattr(struct inode *inode, const char *name,
1382 			      const void *buf, size_t buflen, int flags)
1383 {
1384 	struct nfs4_exception exception = { };
1385 	int err;
1386 
1387 	do {
1388 		err = _nfs42_proc_setxattr(inode, name, buf, buflen, flags);
1389 		if (!err)
1390 			break;
1391 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1392 				&exception);
1393 	} while (exception.retry);
1394 
1395 	return err;
1396 }
1397 
1398 ssize_t nfs42_proc_listxattrs(struct inode *inode, void *buf,
1399 			      size_t buflen, u64 *cookiep, bool *eofp)
1400 {
1401 	struct nfs4_exception exception = { };
1402 	ssize_t err;
1403 
1404 	do {
1405 		err = _nfs42_proc_listxattrs(inode, buf, buflen,
1406 		    cookiep, eofp);
1407 		if (err >= 0)
1408 			break;
1409 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1410 				&exception);
1411 	} while (exception.retry);
1412 
1413 	return err;
1414 }
1415 
1416 int nfs42_proc_removexattr(struct inode *inode, const char *name)
1417 {
1418 	struct nfs4_exception exception = { };
1419 	int err;
1420 
1421 	do {
1422 		err = _nfs42_proc_removexattr(inode, name);
1423 		if (!err)
1424 			break;
1425 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1426 				&exception);
1427 	} while (exception.retry);
1428 
1429 	return err;
1430 }
1431