xref: /openbmc/linux/fs/nfs/nfs42proc.c (revision e368cd72)
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 	kfree(res.falloc_fattr);
87 	return status;
88 }
89 
90 static int nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep,
91 				loff_t offset, loff_t len)
92 {
93 	struct inode *inode = file_inode(filep);
94 	struct nfs_server *server = NFS_SERVER(inode);
95 	struct nfs4_exception exception = { };
96 	struct nfs_lock_context *lock;
97 	int err;
98 
99 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
100 	if (IS_ERR(lock))
101 		return PTR_ERR(lock);
102 
103 	exception.inode = inode;
104 	exception.state = lock->open_context->state;
105 
106 	err = nfs_sync_inode(inode);
107 	if (err)
108 		goto out;
109 
110 	do {
111 		err = _nfs42_proc_fallocate(msg, filep, lock, offset, len);
112 		if (err == -ENOTSUPP) {
113 			err = -EOPNOTSUPP;
114 			break;
115 		}
116 		err = nfs4_handle_exception(server, err, &exception);
117 	} while (exception.retry);
118 out:
119 	nfs_put_lock_context(lock);
120 	return err;
121 }
122 
123 int nfs42_proc_allocate(struct file *filep, loff_t offset, loff_t len)
124 {
125 	struct rpc_message msg = {
126 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ALLOCATE],
127 	};
128 	struct inode *inode = file_inode(filep);
129 	int err;
130 
131 	if (!nfs_server_capable(inode, NFS_CAP_ALLOCATE))
132 		return -EOPNOTSUPP;
133 
134 	inode_lock(inode);
135 
136 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
137 	if (err == -EOPNOTSUPP)
138 		NFS_SERVER(inode)->caps &= ~NFS_CAP_ALLOCATE;
139 
140 	inode_unlock(inode);
141 	return err;
142 }
143 
144 int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len)
145 {
146 	struct rpc_message msg = {
147 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DEALLOCATE],
148 	};
149 	struct inode *inode = file_inode(filep);
150 	int err;
151 
152 	if (!nfs_server_capable(inode, NFS_CAP_DEALLOCATE))
153 		return -EOPNOTSUPP;
154 
155 	inode_lock(inode);
156 
157 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
158 	if (err == 0)
159 		truncate_pagecache_range(inode, offset, (offset + len) -1);
160 	if (err == -EOPNOTSUPP)
161 		NFS_SERVER(inode)->caps &= ~NFS_CAP_DEALLOCATE;
162 
163 	inode_unlock(inode);
164 	return err;
165 }
166 
167 static int handle_async_copy(struct nfs42_copy_res *res,
168 			     struct nfs_server *dst_server,
169 			     struct nfs_server *src_server,
170 			     struct file *src,
171 			     struct file *dst,
172 			     nfs4_stateid *src_stateid,
173 			     bool *restart)
174 {
175 	struct nfs4_copy_state *copy, *tmp_copy;
176 	int status = NFS4_OK;
177 	bool found_pending = false;
178 	struct nfs_open_context *dst_ctx = nfs_file_open_context(dst);
179 	struct nfs_open_context *src_ctx = nfs_file_open_context(src);
180 
181 	copy = kzalloc(sizeof(struct nfs4_copy_state), GFP_NOFS);
182 	if (!copy)
183 		return -ENOMEM;
184 
185 	spin_lock(&dst_server->nfs_client->cl_lock);
186 	list_for_each_entry(tmp_copy,
187 				&dst_server->nfs_client->pending_cb_stateids,
188 				copies) {
189 		if (memcmp(&res->write_res.stateid, &tmp_copy->stateid,
190 				NFS4_STATEID_SIZE))
191 			continue;
192 		found_pending = true;
193 		list_del(&tmp_copy->copies);
194 		break;
195 	}
196 	if (found_pending) {
197 		spin_unlock(&dst_server->nfs_client->cl_lock);
198 		kfree(copy);
199 		copy = tmp_copy;
200 		goto out;
201 	}
202 
203 	memcpy(&copy->stateid, &res->write_res.stateid, NFS4_STATEID_SIZE);
204 	init_completion(&copy->completion);
205 	copy->parent_dst_state = dst_ctx->state;
206 	copy->parent_src_state = src_ctx->state;
207 
208 	list_add_tail(&copy->copies, &dst_server->ss_copies);
209 	spin_unlock(&dst_server->nfs_client->cl_lock);
210 
211 	if (dst_server != src_server) {
212 		spin_lock(&src_server->nfs_client->cl_lock);
213 		list_add_tail(&copy->src_copies, &src_server->ss_copies);
214 		spin_unlock(&src_server->nfs_client->cl_lock);
215 	}
216 
217 	status = wait_for_completion_interruptible(&copy->completion);
218 	spin_lock(&dst_server->nfs_client->cl_lock);
219 	list_del_init(&copy->copies);
220 	spin_unlock(&dst_server->nfs_client->cl_lock);
221 	if (dst_server != src_server) {
222 		spin_lock(&src_server->nfs_client->cl_lock);
223 		list_del_init(&copy->src_copies);
224 		spin_unlock(&src_server->nfs_client->cl_lock);
225 	}
226 	if (status == -ERESTARTSYS) {
227 		goto out_cancel;
228 	} else if (copy->flags || copy->error == NFS4ERR_PARTNER_NO_AUTH) {
229 		status = -EAGAIN;
230 		*restart = true;
231 		goto out_cancel;
232 	}
233 out:
234 	res->write_res.count = copy->count;
235 	memcpy(&res->write_res.verifier, &copy->verf, sizeof(copy->verf));
236 	status = -copy->error;
237 
238 out_free:
239 	kfree(copy);
240 	return status;
241 out_cancel:
242 	nfs42_do_offload_cancel_async(dst, &copy->stateid);
243 	if (!nfs42_files_from_same_server(src, dst))
244 		nfs42_do_offload_cancel_async(src, src_stateid);
245 	goto out_free;
246 }
247 
248 static int process_copy_commit(struct file *dst, loff_t pos_dst,
249 			       struct nfs42_copy_res *res)
250 {
251 	struct nfs_commitres cres;
252 	int status = -ENOMEM;
253 
254 	cres.verf = kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS);
255 	if (!cres.verf)
256 		goto out;
257 
258 	status = nfs4_proc_commit(dst, pos_dst, res->write_res.count, &cres);
259 	if (status)
260 		goto out_free;
261 	if (nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
262 				    &cres.verf->verifier)) {
263 		dprintk("commit verf differs from copy verf\n");
264 		status = -EAGAIN;
265 	}
266 out_free:
267 	kfree(cres.verf);
268 out:
269 	return status;
270 }
271 
272 /**
273  * nfs42_copy_dest_done - perform inode cache updates after clone/copy offload
274  * @inode: pointer to destination inode
275  * @pos: destination offset
276  * @len: copy length
277  *
278  * Punch a hole in the inode page cache, so that the NFS client will
279  * know to retrieve new data.
280  * Update the file size if necessary, and then mark the inode as having
281  * invalid cached values for change attribute, ctime, mtime and space used.
282  */
283 static void nfs42_copy_dest_done(struct inode *inode, loff_t pos, loff_t len)
284 {
285 	loff_t newsize = pos + len;
286 	loff_t end = newsize - 1;
287 
288 	truncate_pagecache_range(inode, pos, end);
289 	spin_lock(&inode->i_lock);
290 	if (newsize > i_size_read(inode))
291 		i_size_write(inode, newsize);
292 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
293 					     NFS_INO_INVALID_CTIME |
294 					     NFS_INO_INVALID_MTIME |
295 					     NFS_INO_INVALID_BLOCKS);
296 	spin_unlock(&inode->i_lock);
297 }
298 
299 static ssize_t _nfs42_proc_copy(struct file *src,
300 				struct nfs_lock_context *src_lock,
301 				struct file *dst,
302 				struct nfs_lock_context *dst_lock,
303 				struct nfs42_copy_args *args,
304 				struct nfs42_copy_res *res,
305 				struct nl4_server *nss,
306 				nfs4_stateid *cnr_stateid,
307 				bool *restart)
308 {
309 	struct rpc_message msg = {
310 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY],
311 		.rpc_argp = args,
312 		.rpc_resp = res,
313 	};
314 	struct inode *dst_inode = file_inode(dst);
315 	struct inode *src_inode = file_inode(src);
316 	struct nfs_server *dst_server = NFS_SERVER(dst_inode);
317 	struct nfs_server *src_server = NFS_SERVER(src_inode);
318 	loff_t pos_src = args->src_pos;
319 	loff_t pos_dst = args->dst_pos;
320 	size_t count = args->count;
321 	ssize_t status;
322 
323 	if (nss) {
324 		args->cp_src = nss;
325 		nfs4_stateid_copy(&args->src_stateid, cnr_stateid);
326 	} else {
327 		status = nfs4_set_rw_stateid(&args->src_stateid,
328 				src_lock->open_context, src_lock, FMODE_READ);
329 		if (status) {
330 			if (status == -EAGAIN)
331 				status = -NFS4ERR_BAD_STATEID;
332 			return status;
333 		}
334 	}
335 	status = nfs_filemap_write_and_wait_range(file_inode(src)->i_mapping,
336 			pos_src, pos_src + (loff_t)count - 1);
337 	if (status)
338 		return status;
339 
340 	status = nfs4_set_rw_stateid(&args->dst_stateid, dst_lock->open_context,
341 				     dst_lock, FMODE_WRITE);
342 	if (status) {
343 		if (status == -EAGAIN)
344 			status = -NFS4ERR_BAD_STATEID;
345 		return status;
346 	}
347 
348 	status = nfs_sync_inode(dst_inode);
349 	if (status)
350 		return status;
351 
352 	res->commit_res.verf = NULL;
353 	if (args->sync) {
354 		res->commit_res.verf =
355 			kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS);
356 		if (!res->commit_res.verf)
357 			return -ENOMEM;
358 	}
359 	set_bit(NFS_CLNT_SRC_SSC_COPY_STATE,
360 		&src_lock->open_context->state->flags);
361 	set_bit(NFS_CLNT_DST_SSC_COPY_STATE,
362 		&dst_lock->open_context->state->flags);
363 
364 	status = nfs4_call_sync(dst_server->client, dst_server, &msg,
365 				&args->seq_args, &res->seq_res, 0);
366 	if (status == -ENOTSUPP)
367 		dst_server->caps &= ~NFS_CAP_COPY;
368 	if (status)
369 		goto out;
370 
371 	if (args->sync &&
372 		nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
373 				    &res->commit_res.verf->verifier)) {
374 		status = -EAGAIN;
375 		goto out;
376 	}
377 
378 	if (!res->synchronous) {
379 		status = handle_async_copy(res, dst_server, src_server, src,
380 				dst, &args->src_stateid, restart);
381 		if (status)
382 			goto out;
383 	}
384 
385 	if ((!res->synchronous || !args->sync) &&
386 			res->write_res.verifier.committed != NFS_FILE_SYNC) {
387 		status = process_copy_commit(dst, pos_dst, res);
388 		if (status)
389 			goto out;
390 	}
391 
392 	nfs42_copy_dest_done(dst_inode, pos_dst, res->write_res.count);
393 	nfs_invalidate_atime(src_inode);
394 	status = res->write_res.count;
395 out:
396 	if (args->sync)
397 		kfree(res->commit_res.verf);
398 	return status;
399 }
400 
401 ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
402 			struct file *dst, loff_t pos_dst, size_t count,
403 			struct nl4_server *nss,
404 			nfs4_stateid *cnr_stateid, bool sync)
405 {
406 	struct nfs_server *server = NFS_SERVER(file_inode(dst));
407 	struct nfs_lock_context *src_lock;
408 	struct nfs_lock_context *dst_lock;
409 	struct nfs42_copy_args args = {
410 		.src_fh		= NFS_FH(file_inode(src)),
411 		.src_pos	= pos_src,
412 		.dst_fh		= NFS_FH(file_inode(dst)),
413 		.dst_pos	= pos_dst,
414 		.count		= count,
415 		.sync		= sync,
416 	};
417 	struct nfs42_copy_res res;
418 	struct nfs4_exception src_exception = {
419 		.inode		= file_inode(src),
420 		.stateid	= &args.src_stateid,
421 	};
422 	struct nfs4_exception dst_exception = {
423 		.inode		= file_inode(dst),
424 		.stateid	= &args.dst_stateid,
425 	};
426 	ssize_t err, err2;
427 	bool restart = false;
428 
429 	src_lock = nfs_get_lock_context(nfs_file_open_context(src));
430 	if (IS_ERR(src_lock))
431 		return PTR_ERR(src_lock);
432 
433 	src_exception.state = src_lock->open_context->state;
434 
435 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst));
436 	if (IS_ERR(dst_lock)) {
437 		err = PTR_ERR(dst_lock);
438 		goto out_put_src_lock;
439 	}
440 
441 	dst_exception.state = dst_lock->open_context->state;
442 
443 	do {
444 		inode_lock(file_inode(dst));
445 		err = _nfs42_proc_copy(src, src_lock,
446 				dst, dst_lock,
447 				&args, &res,
448 				nss, cnr_stateid, &restart);
449 		inode_unlock(file_inode(dst));
450 
451 		if (err >= 0)
452 			break;
453 		if (err == -ENOTSUPP &&
454 				nfs42_files_from_same_server(src, dst)) {
455 			err = -EOPNOTSUPP;
456 			break;
457 		} else if (err == -EAGAIN) {
458 			if (!restart) {
459 				dst_exception.retry = 1;
460 				continue;
461 			}
462 			break;
463 		} else if (err == -NFS4ERR_OFFLOAD_NO_REQS && !args.sync) {
464 			args.sync = true;
465 			dst_exception.retry = 1;
466 			continue;
467 		} else if ((err == -ESTALE ||
468 				err == -NFS4ERR_OFFLOAD_DENIED ||
469 				err == -ENOTSUPP) &&
470 				!nfs42_files_from_same_server(src, dst)) {
471 			nfs42_do_offload_cancel_async(src, &args.src_stateid);
472 			err = -EOPNOTSUPP;
473 			break;
474 		}
475 
476 		err2 = nfs4_handle_exception(server, err, &src_exception);
477 		err  = nfs4_handle_exception(server, err, &dst_exception);
478 		if (!err)
479 			err = err2;
480 	} while (src_exception.retry || dst_exception.retry);
481 
482 	nfs_put_lock_context(dst_lock);
483 out_put_src_lock:
484 	nfs_put_lock_context(src_lock);
485 	return err;
486 }
487 
488 struct nfs42_offloadcancel_data {
489 	struct nfs_server *seq_server;
490 	struct nfs42_offload_status_args args;
491 	struct nfs42_offload_status_res res;
492 };
493 
494 static void nfs42_offload_cancel_prepare(struct rpc_task *task, void *calldata)
495 {
496 	struct nfs42_offloadcancel_data *data = calldata;
497 
498 	nfs4_setup_sequence(data->seq_server->nfs_client,
499 				&data->args.osa_seq_args,
500 				&data->res.osr_seq_res, task);
501 }
502 
503 static void nfs42_offload_cancel_done(struct rpc_task *task, void *calldata)
504 {
505 	struct nfs42_offloadcancel_data *data = calldata;
506 
507 	nfs41_sequence_done(task, &data->res.osr_seq_res);
508 	if (task->tk_status &&
509 		nfs4_async_handle_error(task, data->seq_server, NULL,
510 			NULL) == -EAGAIN)
511 		rpc_restart_call_prepare(task);
512 }
513 
514 static void nfs42_free_offloadcancel_data(void *data)
515 {
516 	kfree(data);
517 }
518 
519 static const struct rpc_call_ops nfs42_offload_cancel_ops = {
520 	.rpc_call_prepare = nfs42_offload_cancel_prepare,
521 	.rpc_call_done = nfs42_offload_cancel_done,
522 	.rpc_release = nfs42_free_offloadcancel_data,
523 };
524 
525 static int nfs42_do_offload_cancel_async(struct file *dst,
526 					 nfs4_stateid *stateid)
527 {
528 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
529 	struct nfs42_offloadcancel_data *data = NULL;
530 	struct nfs_open_context *ctx = nfs_file_open_context(dst);
531 	struct rpc_task *task;
532 	struct rpc_message msg = {
533 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OFFLOAD_CANCEL],
534 		.rpc_cred = ctx->cred,
535 	};
536 	struct rpc_task_setup task_setup_data = {
537 		.rpc_client = dst_server->client,
538 		.rpc_message = &msg,
539 		.callback_ops = &nfs42_offload_cancel_ops,
540 		.workqueue = nfsiod_workqueue,
541 		.flags = RPC_TASK_ASYNC,
542 	};
543 	int status;
544 
545 	if (!(dst_server->caps & NFS_CAP_OFFLOAD_CANCEL))
546 		return -EOPNOTSUPP;
547 
548 	data = kzalloc(sizeof(struct nfs42_offloadcancel_data), GFP_NOFS);
549 	if (data == NULL)
550 		return -ENOMEM;
551 
552 	data->seq_server = dst_server;
553 	data->args.osa_src_fh = NFS_FH(file_inode(dst));
554 	memcpy(&data->args.osa_stateid, stateid,
555 		sizeof(data->args.osa_stateid));
556 	msg.rpc_argp = &data->args;
557 	msg.rpc_resp = &data->res;
558 	task_setup_data.callback_data = data;
559 	nfs4_init_sequence(&data->args.osa_seq_args, &data->res.osr_seq_res,
560 			   1, 0);
561 	task = rpc_run_task(&task_setup_data);
562 	if (IS_ERR(task))
563 		return PTR_ERR(task);
564 	status = rpc_wait_for_completion_task(task);
565 	if (status == -ENOTSUPP)
566 		dst_server->caps &= ~NFS_CAP_OFFLOAD_CANCEL;
567 	rpc_put_task(task);
568 	return status;
569 }
570 
571 static int _nfs42_proc_copy_notify(struct file *src, struct file *dst,
572 				   struct nfs42_copy_notify_args *args,
573 				   struct nfs42_copy_notify_res *res)
574 {
575 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
576 	struct rpc_message msg = {
577 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY_NOTIFY],
578 		.rpc_argp = args,
579 		.rpc_resp = res,
580 	};
581 	int status;
582 	struct nfs_open_context *ctx;
583 	struct nfs_lock_context *l_ctx;
584 
585 	ctx = get_nfs_open_context(nfs_file_open_context(src));
586 	l_ctx = nfs_get_lock_context(ctx);
587 	if (IS_ERR(l_ctx))
588 		return PTR_ERR(l_ctx);
589 
590 	status = nfs4_set_rw_stateid(&args->cna_src_stateid, ctx, l_ctx,
591 				     FMODE_READ);
592 	nfs_put_lock_context(l_ctx);
593 	if (status) {
594 		if (status == -EAGAIN)
595 			status = -NFS4ERR_BAD_STATEID;
596 		return status;
597 	}
598 
599 	status = nfs4_call_sync(src_server->client, src_server, &msg,
600 				&args->cna_seq_args, &res->cnr_seq_res, 0);
601 	if (status == -ENOTSUPP)
602 		src_server->caps &= ~NFS_CAP_COPY_NOTIFY;
603 
604 	put_nfs_open_context(nfs_file_open_context(src));
605 	return status;
606 }
607 
608 int nfs42_proc_copy_notify(struct file *src, struct file *dst,
609 				struct nfs42_copy_notify_res *res)
610 {
611 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
612 	struct nfs42_copy_notify_args *args;
613 	struct nfs4_exception exception = {
614 		.inode = file_inode(src),
615 	};
616 	int status;
617 
618 	if (!(src_server->caps & NFS_CAP_COPY_NOTIFY))
619 		return -EOPNOTSUPP;
620 
621 	args = kzalloc(sizeof(struct nfs42_copy_notify_args), GFP_NOFS);
622 	if (args == NULL)
623 		return -ENOMEM;
624 
625 	args->cna_src_fh  = NFS_FH(file_inode(src)),
626 	args->cna_dst.nl4_type = NL4_NETADDR;
627 	nfs42_set_netaddr(dst, &args->cna_dst.u.nl4_addr);
628 	exception.stateid = &args->cna_src_stateid;
629 
630 	do {
631 		status = _nfs42_proc_copy_notify(src, dst, args, res);
632 		if (status == -ENOTSUPP) {
633 			status = -EOPNOTSUPP;
634 			goto out;
635 		}
636 		status = nfs4_handle_exception(src_server, status, &exception);
637 	} while (exception.retry);
638 
639 out:
640 	kfree(args);
641 	return status;
642 }
643 
644 static loff_t _nfs42_proc_llseek(struct file *filep,
645 		struct nfs_lock_context *lock, loff_t offset, int whence)
646 {
647 	struct inode *inode = file_inode(filep);
648 	struct nfs42_seek_args args = {
649 		.sa_fh		= NFS_FH(inode),
650 		.sa_offset	= offset,
651 		.sa_what	= (whence == SEEK_HOLE) ?
652 					NFS4_CONTENT_HOLE : NFS4_CONTENT_DATA,
653 	};
654 	struct nfs42_seek_res res;
655 	struct rpc_message msg = {
656 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEEK],
657 		.rpc_argp = &args,
658 		.rpc_resp = &res,
659 	};
660 	struct nfs_server *server = NFS_SERVER(inode);
661 	int status;
662 
663 	if (!nfs_server_capable(inode, NFS_CAP_SEEK))
664 		return -ENOTSUPP;
665 
666 	status = nfs4_set_rw_stateid(&args.sa_stateid, lock->open_context,
667 			lock, FMODE_READ);
668 	if (status) {
669 		if (status == -EAGAIN)
670 			status = -NFS4ERR_BAD_STATEID;
671 		return status;
672 	}
673 
674 	status = nfs_filemap_write_and_wait_range(inode->i_mapping,
675 			offset, LLONG_MAX);
676 	if (status)
677 		return status;
678 
679 	status = nfs4_call_sync(server->client, server, &msg,
680 				&args.seq_args, &res.seq_res, 0);
681 	if (status == -ENOTSUPP)
682 		server->caps &= ~NFS_CAP_SEEK;
683 	if (status)
684 		return status;
685 
686 	if (whence == SEEK_DATA && res.sr_eof)
687 		return -NFS4ERR_NXIO;
688 	else
689 		return vfs_setpos(filep, res.sr_offset, inode->i_sb->s_maxbytes);
690 }
691 
692 loff_t nfs42_proc_llseek(struct file *filep, loff_t offset, int whence)
693 {
694 	struct nfs_server *server = NFS_SERVER(file_inode(filep));
695 	struct nfs4_exception exception = { };
696 	struct nfs_lock_context *lock;
697 	loff_t err;
698 
699 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
700 	if (IS_ERR(lock))
701 		return PTR_ERR(lock);
702 
703 	exception.inode = file_inode(filep);
704 	exception.state = lock->open_context->state;
705 
706 	do {
707 		err = _nfs42_proc_llseek(filep, lock, offset, whence);
708 		if (err >= 0)
709 			break;
710 		if (err == -ENOTSUPP) {
711 			err = -EOPNOTSUPP;
712 			break;
713 		}
714 		err = nfs4_handle_exception(server, err, &exception);
715 	} while (exception.retry);
716 
717 	nfs_put_lock_context(lock);
718 	return err;
719 }
720 
721 
722 static void
723 nfs42_layoutstat_prepare(struct rpc_task *task, void *calldata)
724 {
725 	struct nfs42_layoutstat_data *data = calldata;
726 	struct inode *inode = data->inode;
727 	struct nfs_server *server = NFS_SERVER(inode);
728 	struct pnfs_layout_hdr *lo;
729 
730 	spin_lock(&inode->i_lock);
731 	lo = NFS_I(inode)->layout;
732 	if (!pnfs_layout_is_valid(lo)) {
733 		spin_unlock(&inode->i_lock);
734 		rpc_exit(task, 0);
735 		return;
736 	}
737 	nfs4_stateid_copy(&data->args.stateid, &lo->plh_stateid);
738 	spin_unlock(&inode->i_lock);
739 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
740 			    &data->res.seq_res, task);
741 }
742 
743 static void
744 nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
745 {
746 	struct nfs42_layoutstat_data *data = calldata;
747 	struct inode *inode = data->inode;
748 	struct pnfs_layout_hdr *lo;
749 
750 	if (!nfs4_sequence_done(task, &data->res.seq_res))
751 		return;
752 
753 	switch (task->tk_status) {
754 	case 0:
755 		return;
756 	case -NFS4ERR_BADHANDLE:
757 	case -ESTALE:
758 		pnfs_destroy_layout(NFS_I(inode));
759 		break;
760 	case -NFS4ERR_EXPIRED:
761 	case -NFS4ERR_ADMIN_REVOKED:
762 	case -NFS4ERR_DELEG_REVOKED:
763 	case -NFS4ERR_STALE_STATEID:
764 	case -NFS4ERR_BAD_STATEID:
765 		spin_lock(&inode->i_lock);
766 		lo = NFS_I(inode)->layout;
767 		if (pnfs_layout_is_valid(lo) &&
768 		    nfs4_stateid_match(&data->args.stateid,
769 					     &lo->plh_stateid)) {
770 			LIST_HEAD(head);
771 
772 			/*
773 			 * Mark the bad layout state as invalid, then retry
774 			 * with the current stateid.
775 			 */
776 			pnfs_mark_layout_stateid_invalid(lo, &head);
777 			spin_unlock(&inode->i_lock);
778 			pnfs_free_lseg_list(&head);
779 			nfs_commit_inode(inode, 0);
780 		} else
781 			spin_unlock(&inode->i_lock);
782 		break;
783 	case -NFS4ERR_OLD_STATEID:
784 		spin_lock(&inode->i_lock);
785 		lo = NFS_I(inode)->layout;
786 		if (pnfs_layout_is_valid(lo) &&
787 		    nfs4_stateid_match_other(&data->args.stateid,
788 					&lo->plh_stateid)) {
789 			/* Do we need to delay before resending? */
790 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
791 						&data->args.stateid))
792 				rpc_delay(task, HZ);
793 			rpc_restart_call_prepare(task);
794 		}
795 		spin_unlock(&inode->i_lock);
796 		break;
797 	case -ENOTSUPP:
798 	case -EOPNOTSUPP:
799 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTSTATS;
800 	}
801 
802 	trace_nfs4_layoutstats(inode, &data->args.stateid, task->tk_status);
803 }
804 
805 static void
806 nfs42_layoutstat_release(void *calldata)
807 {
808 	struct nfs42_layoutstat_data *data = calldata;
809 	struct nfs42_layoutstat_devinfo *devinfo = data->args.devinfo;
810 	int i;
811 
812 	for (i = 0; i < data->args.num_dev; i++) {
813 		if (devinfo[i].ld_private.ops && devinfo[i].ld_private.ops->free)
814 			devinfo[i].ld_private.ops->free(&devinfo[i].ld_private);
815 	}
816 
817 	pnfs_put_layout_hdr(NFS_I(data->args.inode)->layout);
818 	smp_mb__before_atomic();
819 	clear_bit(NFS_INO_LAYOUTSTATS, &NFS_I(data->args.inode)->flags);
820 	smp_mb__after_atomic();
821 	nfs_iput_and_deactive(data->inode);
822 	kfree(data->args.devinfo);
823 	kfree(data);
824 }
825 
826 static const struct rpc_call_ops nfs42_layoutstat_ops = {
827 	.rpc_call_prepare = nfs42_layoutstat_prepare,
828 	.rpc_call_done = nfs42_layoutstat_done,
829 	.rpc_release = nfs42_layoutstat_release,
830 };
831 
832 int nfs42_proc_layoutstats_generic(struct nfs_server *server,
833 				   struct nfs42_layoutstat_data *data)
834 {
835 	struct rpc_message msg = {
836 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTSTATS],
837 		.rpc_argp = &data->args,
838 		.rpc_resp = &data->res,
839 	};
840 	struct rpc_task_setup task_setup = {
841 		.rpc_client = server->client,
842 		.rpc_message = &msg,
843 		.callback_ops = &nfs42_layoutstat_ops,
844 		.callback_data = data,
845 		.flags = RPC_TASK_ASYNC,
846 	};
847 	struct rpc_task *task;
848 
849 	data->inode = nfs_igrab_and_active(data->args.inode);
850 	if (!data->inode) {
851 		nfs42_layoutstat_release(data);
852 		return -EAGAIN;
853 	}
854 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
855 	task = rpc_run_task(&task_setup);
856 	if (IS_ERR(task))
857 		return PTR_ERR(task);
858 	rpc_put_task(task);
859 	return 0;
860 }
861 
862 static struct nfs42_layouterror_data *
863 nfs42_alloc_layouterror_data(struct pnfs_layout_segment *lseg, gfp_t gfp_flags)
864 {
865 	struct nfs42_layouterror_data *data;
866 	struct inode *inode = lseg->pls_layout->plh_inode;
867 
868 	data = kzalloc(sizeof(*data), gfp_flags);
869 	if (data) {
870 		data->args.inode = data->inode = nfs_igrab_and_active(inode);
871 		if (data->inode) {
872 			data->lseg = pnfs_get_lseg(lseg);
873 			if (data->lseg)
874 				return data;
875 			nfs_iput_and_deactive(data->inode);
876 		}
877 		kfree(data);
878 	}
879 	return NULL;
880 }
881 
882 static void
883 nfs42_free_layouterror_data(struct nfs42_layouterror_data *data)
884 {
885 	pnfs_put_lseg(data->lseg);
886 	nfs_iput_and_deactive(data->inode);
887 	kfree(data);
888 }
889 
890 static void
891 nfs42_layouterror_prepare(struct rpc_task *task, void *calldata)
892 {
893 	struct nfs42_layouterror_data *data = calldata;
894 	struct inode *inode = data->inode;
895 	struct nfs_server *server = NFS_SERVER(inode);
896 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
897 	unsigned i;
898 
899 	spin_lock(&inode->i_lock);
900 	if (!pnfs_layout_is_valid(lo)) {
901 		spin_unlock(&inode->i_lock);
902 		rpc_exit(task, 0);
903 		return;
904 	}
905 	for (i = 0; i < data->args.num_errors; i++)
906 		nfs4_stateid_copy(&data->args.errors[i].stateid,
907 				&lo->plh_stateid);
908 	spin_unlock(&inode->i_lock);
909 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
910 			    &data->res.seq_res, task);
911 }
912 
913 static void
914 nfs42_layouterror_done(struct rpc_task *task, void *calldata)
915 {
916 	struct nfs42_layouterror_data *data = calldata;
917 	struct inode *inode = data->inode;
918 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
919 
920 	if (!nfs4_sequence_done(task, &data->res.seq_res))
921 		return;
922 
923 	switch (task->tk_status) {
924 	case 0:
925 		return;
926 	case -NFS4ERR_BADHANDLE:
927 	case -ESTALE:
928 		pnfs_destroy_layout(NFS_I(inode));
929 		break;
930 	case -NFS4ERR_EXPIRED:
931 	case -NFS4ERR_ADMIN_REVOKED:
932 	case -NFS4ERR_DELEG_REVOKED:
933 	case -NFS4ERR_STALE_STATEID:
934 	case -NFS4ERR_BAD_STATEID:
935 		spin_lock(&inode->i_lock);
936 		if (pnfs_layout_is_valid(lo) &&
937 		    nfs4_stateid_match(&data->args.errors[0].stateid,
938 					     &lo->plh_stateid)) {
939 			LIST_HEAD(head);
940 
941 			/*
942 			 * Mark the bad layout state as invalid, then retry
943 			 * with the current stateid.
944 			 */
945 			pnfs_mark_layout_stateid_invalid(lo, &head);
946 			spin_unlock(&inode->i_lock);
947 			pnfs_free_lseg_list(&head);
948 			nfs_commit_inode(inode, 0);
949 		} else
950 			spin_unlock(&inode->i_lock);
951 		break;
952 	case -NFS4ERR_OLD_STATEID:
953 		spin_lock(&inode->i_lock);
954 		if (pnfs_layout_is_valid(lo) &&
955 		    nfs4_stateid_match_other(&data->args.errors[0].stateid,
956 					&lo->plh_stateid)) {
957 			/* Do we need to delay before resending? */
958 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
959 						&data->args.errors[0].stateid))
960 				rpc_delay(task, HZ);
961 			rpc_restart_call_prepare(task);
962 		}
963 		spin_unlock(&inode->i_lock);
964 		break;
965 	case -ENOTSUPP:
966 	case -EOPNOTSUPP:
967 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTERROR;
968 	}
969 
970 	trace_nfs4_layouterror(inode, &data->args.errors[0].stateid,
971 			       task->tk_status);
972 }
973 
974 static void
975 nfs42_layouterror_release(void *calldata)
976 {
977 	struct nfs42_layouterror_data *data = calldata;
978 
979 	nfs42_free_layouterror_data(data);
980 }
981 
982 static const struct rpc_call_ops nfs42_layouterror_ops = {
983 	.rpc_call_prepare = nfs42_layouterror_prepare,
984 	.rpc_call_done = nfs42_layouterror_done,
985 	.rpc_release = nfs42_layouterror_release,
986 };
987 
988 int nfs42_proc_layouterror(struct pnfs_layout_segment *lseg,
989 		const struct nfs42_layout_error *errors, size_t n)
990 {
991 	struct inode *inode = lseg->pls_layout->plh_inode;
992 	struct nfs42_layouterror_data *data;
993 	struct rpc_task *task;
994 	struct rpc_message msg = {
995 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTERROR],
996 	};
997 	struct rpc_task_setup task_setup = {
998 		.rpc_message = &msg,
999 		.callback_ops = &nfs42_layouterror_ops,
1000 		.flags = RPC_TASK_ASYNC,
1001 	};
1002 	unsigned int i;
1003 
1004 	if (!nfs_server_capable(inode, NFS_CAP_LAYOUTERROR))
1005 		return -EOPNOTSUPP;
1006 	if (n > NFS42_LAYOUTERROR_MAX)
1007 		return -EINVAL;
1008 	data = nfs42_alloc_layouterror_data(lseg, GFP_NOFS);
1009 	if (!data)
1010 		return -ENOMEM;
1011 	for (i = 0; i < n; i++) {
1012 		data->args.errors[i] = errors[i];
1013 		data->args.num_errors++;
1014 		data->res.num_errors++;
1015 	}
1016 	msg.rpc_argp = &data->args;
1017 	msg.rpc_resp = &data->res;
1018 	task_setup.callback_data = data;
1019 	task_setup.rpc_client = NFS_SERVER(inode)->client;
1020 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
1021 	task = rpc_run_task(&task_setup);
1022 	if (IS_ERR(task))
1023 		return PTR_ERR(task);
1024 	rpc_put_task(task);
1025 	return 0;
1026 }
1027 EXPORT_SYMBOL_GPL(nfs42_proc_layouterror);
1028 
1029 static int _nfs42_proc_clone(struct rpc_message *msg, struct file *src_f,
1030 		struct file *dst_f, struct nfs_lock_context *src_lock,
1031 		struct nfs_lock_context *dst_lock, loff_t src_offset,
1032 		loff_t dst_offset, loff_t count)
1033 {
1034 	struct inode *src_inode = file_inode(src_f);
1035 	struct inode *dst_inode = file_inode(dst_f);
1036 	struct nfs_server *server = NFS_SERVER(dst_inode);
1037 	struct nfs42_clone_args args = {
1038 		.src_fh = NFS_FH(src_inode),
1039 		.dst_fh = NFS_FH(dst_inode),
1040 		.src_offset = src_offset,
1041 		.dst_offset = dst_offset,
1042 		.count = count,
1043 		.dst_bitmask = server->cache_consistency_bitmask,
1044 	};
1045 	struct nfs42_clone_res res = {
1046 		.server	= server,
1047 	};
1048 	int status;
1049 
1050 	msg->rpc_argp = &args;
1051 	msg->rpc_resp = &res;
1052 
1053 	status = nfs4_set_rw_stateid(&args.src_stateid, src_lock->open_context,
1054 			src_lock, FMODE_READ);
1055 	if (status) {
1056 		if (status == -EAGAIN)
1057 			status = -NFS4ERR_BAD_STATEID;
1058 		return status;
1059 	}
1060 	status = nfs4_set_rw_stateid(&args.dst_stateid, dst_lock->open_context,
1061 			dst_lock, FMODE_WRITE);
1062 	if (status) {
1063 		if (status == -EAGAIN)
1064 			status = -NFS4ERR_BAD_STATEID;
1065 		return status;
1066 	}
1067 
1068 	res.dst_fattr = nfs_alloc_fattr();
1069 	if (!res.dst_fattr)
1070 		return -ENOMEM;
1071 
1072 	status = nfs4_call_sync(server->client, server, msg,
1073 				&args.seq_args, &res.seq_res, 0);
1074 	if (status == 0) {
1075 		nfs42_copy_dest_done(dst_inode, dst_offset, count);
1076 		status = nfs_post_op_update_inode(dst_inode, res.dst_fattr);
1077 	}
1078 
1079 	kfree(res.dst_fattr);
1080 	return status;
1081 }
1082 
1083 int nfs42_proc_clone(struct file *src_f, struct file *dst_f,
1084 		     loff_t src_offset, loff_t dst_offset, loff_t count)
1085 {
1086 	struct rpc_message msg = {
1087 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLONE],
1088 	};
1089 	struct inode *inode = file_inode(src_f);
1090 	struct nfs_server *server = NFS_SERVER(file_inode(src_f));
1091 	struct nfs_lock_context *src_lock;
1092 	struct nfs_lock_context *dst_lock;
1093 	struct nfs4_exception src_exception = { };
1094 	struct nfs4_exception dst_exception = { };
1095 	int err, err2;
1096 
1097 	if (!nfs_server_capable(inode, NFS_CAP_CLONE))
1098 		return -EOPNOTSUPP;
1099 
1100 	src_lock = nfs_get_lock_context(nfs_file_open_context(src_f));
1101 	if (IS_ERR(src_lock))
1102 		return PTR_ERR(src_lock);
1103 
1104 	src_exception.inode = file_inode(src_f);
1105 	src_exception.state = src_lock->open_context->state;
1106 
1107 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst_f));
1108 	if (IS_ERR(dst_lock)) {
1109 		err = PTR_ERR(dst_lock);
1110 		goto out_put_src_lock;
1111 	}
1112 
1113 	dst_exception.inode = file_inode(dst_f);
1114 	dst_exception.state = dst_lock->open_context->state;
1115 
1116 	do {
1117 		err = _nfs42_proc_clone(&msg, src_f, dst_f, src_lock, dst_lock,
1118 					src_offset, dst_offset, count);
1119 		if (err == -ENOTSUPP || err == -EOPNOTSUPP) {
1120 			NFS_SERVER(inode)->caps &= ~NFS_CAP_CLONE;
1121 			err = -EOPNOTSUPP;
1122 			break;
1123 		}
1124 
1125 		err2 = nfs4_handle_exception(server, err, &src_exception);
1126 		err = nfs4_handle_exception(server, err, &dst_exception);
1127 		if (!err)
1128 			err = err2;
1129 	} while (src_exception.retry || dst_exception.retry);
1130 
1131 	nfs_put_lock_context(dst_lock);
1132 out_put_src_lock:
1133 	nfs_put_lock_context(src_lock);
1134 	return err;
1135 }
1136 
1137 #define NFS4XATTR_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
1138 
1139 static int _nfs42_proc_removexattr(struct inode *inode, const char *name)
1140 {
1141 	struct nfs_server *server = NFS_SERVER(inode);
1142 	struct nfs42_removexattrargs args = {
1143 		.fh = NFS_FH(inode),
1144 		.xattr_name = name,
1145 	};
1146 	struct nfs42_removexattrres res;
1147 	struct rpc_message msg = {
1148 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVEXATTR],
1149 		.rpc_argp = &args,
1150 		.rpc_resp = &res,
1151 	};
1152 	int ret;
1153 	unsigned long timestamp = jiffies;
1154 
1155 	ret = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
1156 	    &res.seq_res, 1);
1157 	if (!ret)
1158 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1159 
1160 	return ret;
1161 }
1162 
1163 static int _nfs42_proc_setxattr(struct inode *inode, const char *name,
1164 				const void *buf, size_t buflen, int flags)
1165 {
1166 	struct nfs_server *server = NFS_SERVER(inode);
1167 	struct page *pages[NFS4XATTR_MAXPAGES];
1168 	struct nfs42_setxattrargs arg = {
1169 		.fh		= NFS_FH(inode),
1170 		.xattr_pages	= pages,
1171 		.xattr_len	= buflen,
1172 		.xattr_name	= name,
1173 		.xattr_flags	= flags,
1174 	};
1175 	struct nfs42_setxattrres res;
1176 	struct rpc_message msg = {
1177 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETXATTR],
1178 		.rpc_argp	= &arg,
1179 		.rpc_resp	= &res,
1180 	};
1181 	int ret, np;
1182 	unsigned long timestamp = jiffies;
1183 
1184 	if (buflen > server->sxasize)
1185 		return -ERANGE;
1186 
1187 	if (buflen > 0) {
1188 		np = nfs4_buf_to_pages_noslab(buf, buflen, arg.xattr_pages);
1189 		if (np < 0)
1190 			return np;
1191 	} else
1192 		np = 0;
1193 
1194 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1195 	    &res.seq_res, 1);
1196 
1197 	for (; np > 0; np--)
1198 		put_page(pages[np - 1]);
1199 
1200 	if (!ret)
1201 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1202 
1203 	return ret;
1204 }
1205 
1206 static ssize_t _nfs42_proc_getxattr(struct inode *inode, const char *name,
1207 				void *buf, size_t buflen, struct page **pages,
1208 				size_t plen)
1209 {
1210 	struct nfs_server *server = NFS_SERVER(inode);
1211 	struct nfs42_getxattrargs arg = {
1212 		.fh		= NFS_FH(inode),
1213 		.xattr_name	= name,
1214 	};
1215 	struct nfs42_getxattrres res;
1216 	struct rpc_message msg = {
1217 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETXATTR],
1218 		.rpc_argp	= &arg,
1219 		.rpc_resp	= &res,
1220 	};
1221 	ssize_t ret;
1222 
1223 	arg.xattr_len = plen;
1224 	arg.xattr_pages = pages;
1225 
1226 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1227 	    &res.seq_res, 0);
1228 	if (ret < 0)
1229 		return ret;
1230 
1231 	/*
1232 	 * Normally, the caching is done one layer up, but for successful
1233 	 * RPCS, always cache the result here, even if the caller was
1234 	 * just querying the length, or if the reply was too big for
1235 	 * the caller. This avoids a second RPC in the case of the
1236 	 * common query-alloc-retrieve cycle for xattrs.
1237 	 *
1238 	 * Note that xattr_len is always capped to XATTR_SIZE_MAX.
1239 	 */
1240 
1241 	nfs4_xattr_cache_add(inode, name, NULL, pages, res.xattr_len);
1242 
1243 	if (buflen) {
1244 		if (res.xattr_len > buflen)
1245 			return -ERANGE;
1246 		_copy_from_pages(buf, pages, 0, res.xattr_len);
1247 	}
1248 
1249 	return res.xattr_len;
1250 }
1251 
1252 static ssize_t _nfs42_proc_listxattrs(struct inode *inode, void *buf,
1253 				 size_t buflen, u64 *cookiep, bool *eofp)
1254 {
1255 	struct nfs_server *server = NFS_SERVER(inode);
1256 	struct page **pages;
1257 	struct nfs42_listxattrsargs arg = {
1258 		.fh		= NFS_FH(inode),
1259 		.cookie		= *cookiep,
1260 	};
1261 	struct nfs42_listxattrsres res = {
1262 		.eof = false,
1263 		.xattr_buf = buf,
1264 		.xattr_len = buflen,
1265 	};
1266 	struct rpc_message msg = {
1267 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LISTXATTRS],
1268 		.rpc_argp	= &arg,
1269 		.rpc_resp	= &res,
1270 	};
1271 	u32 xdrlen;
1272 	int ret, np, i;
1273 
1274 
1275 	ret = -ENOMEM;
1276 	res.scratch = alloc_page(GFP_KERNEL);
1277 	if (!res.scratch)
1278 		goto out;
1279 
1280 	xdrlen = nfs42_listxattr_xdrsize(buflen);
1281 	if (xdrlen > server->lxasize)
1282 		xdrlen = server->lxasize;
1283 	np = xdrlen / PAGE_SIZE + 1;
1284 
1285 	pages = kcalloc(np, sizeof(struct page *), GFP_KERNEL);
1286 	if (!pages)
1287 		goto out_free_scratch;
1288 	for (i = 0; i < np; i++) {
1289 		pages[i] = alloc_page(GFP_KERNEL);
1290 		if (!pages[i])
1291 			goto out_free_pages;
1292 	}
1293 
1294 	arg.xattr_pages = pages;
1295 	arg.count = xdrlen;
1296 
1297 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1298 	    &res.seq_res, 0);
1299 
1300 	if (ret >= 0) {
1301 		ret = res.copied;
1302 		*cookiep = res.cookie;
1303 		*eofp = res.eof;
1304 	}
1305 
1306 out_free_pages:
1307 	while (--np >= 0) {
1308 		if (pages[np])
1309 			__free_page(pages[np]);
1310 	}
1311 	kfree(pages);
1312 out_free_scratch:
1313 	__free_page(res.scratch);
1314 out:
1315 	return ret;
1316 
1317 }
1318 
1319 ssize_t nfs42_proc_getxattr(struct inode *inode, const char *name,
1320 			      void *buf, size_t buflen)
1321 {
1322 	struct nfs4_exception exception = { };
1323 	ssize_t err, np, i;
1324 	struct page **pages;
1325 
1326 	np = nfs_page_array_len(0, buflen ?: XATTR_SIZE_MAX);
1327 	pages = kmalloc_array(np, sizeof(*pages), GFP_KERNEL);
1328 	if (!pages)
1329 		return -ENOMEM;
1330 
1331 	for (i = 0; i < np; i++) {
1332 		pages[i] = alloc_page(GFP_KERNEL);
1333 		if (!pages[i]) {
1334 			np = i + 1;
1335 			err = -ENOMEM;
1336 			goto out;
1337 		}
1338 	}
1339 
1340 	/*
1341 	 * The GETXATTR op has no length field in the call, and the
1342 	 * xattr data is at the end of the reply.
1343 	 *
1344 	 * There is no downside in using the page-aligned length. It will
1345 	 * allow receiving and caching xattrs that are too large for the
1346 	 * caller but still fit in the page-rounded value.
1347 	 */
1348 	do {
1349 		err = _nfs42_proc_getxattr(inode, name, buf, buflen,
1350 			pages, np * PAGE_SIZE);
1351 		if (err >= 0)
1352 			break;
1353 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1354 				&exception);
1355 	} while (exception.retry);
1356 
1357 out:
1358 	while (--np >= 0)
1359 		__free_page(pages[np]);
1360 	kfree(pages);
1361 
1362 	return err;
1363 }
1364 
1365 int nfs42_proc_setxattr(struct inode *inode, const char *name,
1366 			      const void *buf, size_t buflen, int flags)
1367 {
1368 	struct nfs4_exception exception = { };
1369 	int err;
1370 
1371 	do {
1372 		err = _nfs42_proc_setxattr(inode, name, buf, buflen, flags);
1373 		if (!err)
1374 			break;
1375 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1376 				&exception);
1377 	} while (exception.retry);
1378 
1379 	return err;
1380 }
1381 
1382 ssize_t nfs42_proc_listxattrs(struct inode *inode, void *buf,
1383 			      size_t buflen, u64 *cookiep, bool *eofp)
1384 {
1385 	struct nfs4_exception exception = { };
1386 	ssize_t err;
1387 
1388 	do {
1389 		err = _nfs42_proc_listxattrs(inode, buf, buflen,
1390 		    cookiep, eofp);
1391 		if (err >= 0)
1392 			break;
1393 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1394 				&exception);
1395 	} while (exception.retry);
1396 
1397 	return err;
1398 }
1399 
1400 int nfs42_proc_removexattr(struct inode *inode, const char *name)
1401 {
1402 	struct nfs4_exception exception = { };
1403 	int err;
1404 
1405 	do {
1406 		err = _nfs42_proc_removexattr(inode, name);
1407 		if (!err)
1408 			break;
1409 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1410 				&exception);
1411 	} while (exception.retry);
1412 
1413 	return err;
1414 }
1415