xref: /openbmc/linux/fs/nfs/nfs42proc.c (revision 36de991e)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2014 Anna Schumaker <Anna.Schumaker@Netapp.com>
4  */
5 #include <linux/fs.h>
6 #include <linux/sunrpc/addr.h>
7 #include <linux/sunrpc/sched.h>
8 #include <linux/nfs.h>
9 #include <linux/nfs3.h>
10 #include <linux/nfs4.h>
11 #include <linux/nfs_xdr.h>
12 #include <linux/nfs_fs.h>
13 #include "nfs4_fs.h"
14 #include "nfs42.h"
15 #include "iostat.h"
16 #include "pnfs.h"
17 #include "nfs4session.h"
18 #include "internal.h"
19 #include "delegation.h"
20 #include "nfs4trace.h"
21 
22 #define NFSDBG_FACILITY NFSDBG_PROC
23 static int nfs42_do_offload_cancel_async(struct file *dst, nfs4_stateid *std);
24 
25 static void nfs42_set_netaddr(struct file *filep, struct nfs42_netaddr *naddr)
26 {
27 	struct nfs_client *clp = (NFS_SERVER(file_inode(filep)))->nfs_client;
28 	unsigned short port = 2049;
29 
30 	rcu_read_lock();
31 	naddr->netid_len = scnprintf(naddr->netid,
32 					sizeof(naddr->netid), "%s",
33 					rpc_peeraddr2str(clp->cl_rpcclient,
34 					RPC_DISPLAY_NETID));
35 	naddr->addr_len = scnprintf(naddr->addr,
36 					sizeof(naddr->addr),
37 					"%s.%u.%u",
38 					rpc_peeraddr2str(clp->cl_rpcclient,
39 					RPC_DISPLAY_ADDR),
40 					port >> 8, port & 255);
41 	rcu_read_unlock();
42 }
43 
44 static int _nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep,
45 		struct nfs_lock_context *lock, loff_t offset, loff_t len)
46 {
47 	struct inode *inode = file_inode(filep);
48 	struct nfs_server *server = NFS_SERVER(inode);
49 	u32 bitmask[3];
50 	struct nfs42_falloc_args args = {
51 		.falloc_fh	= NFS_FH(inode),
52 		.falloc_offset	= offset,
53 		.falloc_length	= len,
54 		.falloc_bitmask	= bitmask,
55 	};
56 	struct nfs42_falloc_res res = {
57 		.falloc_server	= server,
58 	};
59 	int status;
60 
61 	msg->rpc_argp = &args;
62 	msg->rpc_resp = &res;
63 
64 	status = nfs4_set_rw_stateid(&args.falloc_stateid, lock->open_context,
65 			lock, FMODE_WRITE);
66 	if (status) {
67 		if (status == -EAGAIN)
68 			status = -NFS4ERR_BAD_STATEID;
69 		return status;
70 	}
71 
72 	memcpy(bitmask, server->cache_consistency_bitmask, sizeof(bitmask));
73 	if (server->attr_bitmask[1] & FATTR4_WORD1_SPACE_USED)
74 		bitmask[1] |= FATTR4_WORD1_SPACE_USED;
75 
76 	res.falloc_fattr = nfs_alloc_fattr();
77 	if (!res.falloc_fattr)
78 		return -ENOMEM;
79 
80 	status = nfs4_call_sync(server->client, server, msg,
81 				&args.seq_args, &res.seq_res, 0);
82 	if (status == 0)
83 		status = nfs_post_op_update_inode_force_wcc(inode,
84 							    res.falloc_fattr);
85 
86 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_ALLOCATE])
87 		trace_nfs4_fallocate(inode, &args, status);
88 	else
89 		trace_nfs4_deallocate(inode, &args, status);
90 	kfree(res.falloc_fattr);
91 	return status;
92 }
93 
94 static int nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep,
95 				loff_t offset, loff_t len)
96 {
97 	struct inode *inode = file_inode(filep);
98 	struct nfs_server *server = NFS_SERVER(inode);
99 	struct nfs4_exception exception = { };
100 	struct nfs_lock_context *lock;
101 	int err;
102 
103 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
104 	if (IS_ERR(lock))
105 		return PTR_ERR(lock);
106 
107 	exception.inode = inode;
108 	exception.state = lock->open_context->state;
109 
110 	err = nfs_sync_inode(inode);
111 	if (err)
112 		goto out;
113 
114 	do {
115 		err = _nfs42_proc_fallocate(msg, filep, lock, offset, len);
116 		if (err == -ENOTSUPP) {
117 			err = -EOPNOTSUPP;
118 			break;
119 		}
120 		err = nfs4_handle_exception(server, err, &exception);
121 	} while (exception.retry);
122 out:
123 	nfs_put_lock_context(lock);
124 	return err;
125 }
126 
127 int nfs42_proc_allocate(struct file *filep, loff_t offset, loff_t len)
128 {
129 	struct rpc_message msg = {
130 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ALLOCATE],
131 	};
132 	struct inode *inode = file_inode(filep);
133 	int err;
134 
135 	if (!nfs_server_capable(inode, NFS_CAP_ALLOCATE))
136 		return -EOPNOTSUPP;
137 
138 	inode_lock(inode);
139 
140 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
141 	if (err == -EOPNOTSUPP)
142 		NFS_SERVER(inode)->caps &= ~NFS_CAP_ALLOCATE;
143 
144 	inode_unlock(inode);
145 	return err;
146 }
147 
148 int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len)
149 {
150 	struct rpc_message msg = {
151 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DEALLOCATE],
152 	};
153 	struct inode *inode = file_inode(filep);
154 	int err;
155 
156 	if (!nfs_server_capable(inode, NFS_CAP_DEALLOCATE))
157 		return -EOPNOTSUPP;
158 
159 	inode_lock(inode);
160 
161 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
162 	if (err == 0)
163 		truncate_pagecache_range(inode, offset, (offset + len) -1);
164 	if (err == -EOPNOTSUPP)
165 		NFS_SERVER(inode)->caps &= ~NFS_CAP_DEALLOCATE;
166 
167 	inode_unlock(inode);
168 	return err;
169 }
170 
171 static int handle_async_copy(struct nfs42_copy_res *res,
172 			     struct nfs_server *dst_server,
173 			     struct nfs_server *src_server,
174 			     struct file *src,
175 			     struct file *dst,
176 			     nfs4_stateid *src_stateid,
177 			     bool *restart)
178 {
179 	struct nfs4_copy_state *copy, *tmp_copy;
180 	int status = NFS4_OK;
181 	bool found_pending = false;
182 	struct nfs_open_context *dst_ctx = nfs_file_open_context(dst);
183 	struct nfs_open_context *src_ctx = nfs_file_open_context(src);
184 
185 	copy = kzalloc(sizeof(struct nfs4_copy_state), GFP_NOFS);
186 	if (!copy)
187 		return -ENOMEM;
188 
189 	spin_lock(&dst_server->nfs_client->cl_lock);
190 	list_for_each_entry(tmp_copy,
191 				&dst_server->nfs_client->pending_cb_stateids,
192 				copies) {
193 		if (memcmp(&res->write_res.stateid, &tmp_copy->stateid,
194 				NFS4_STATEID_SIZE))
195 			continue;
196 		found_pending = true;
197 		list_del(&tmp_copy->copies);
198 		break;
199 	}
200 	if (found_pending) {
201 		spin_unlock(&dst_server->nfs_client->cl_lock);
202 		kfree(copy);
203 		copy = tmp_copy;
204 		goto out;
205 	}
206 
207 	memcpy(&copy->stateid, &res->write_res.stateid, NFS4_STATEID_SIZE);
208 	init_completion(&copy->completion);
209 	copy->parent_dst_state = dst_ctx->state;
210 	copy->parent_src_state = src_ctx->state;
211 
212 	list_add_tail(&copy->copies, &dst_server->ss_copies);
213 	spin_unlock(&dst_server->nfs_client->cl_lock);
214 
215 	if (dst_server != src_server) {
216 		spin_lock(&src_server->nfs_client->cl_lock);
217 		list_add_tail(&copy->src_copies, &src_server->ss_copies);
218 		spin_unlock(&src_server->nfs_client->cl_lock);
219 	}
220 
221 	status = wait_for_completion_interruptible(&copy->completion);
222 	spin_lock(&dst_server->nfs_client->cl_lock);
223 	list_del_init(&copy->copies);
224 	spin_unlock(&dst_server->nfs_client->cl_lock);
225 	if (dst_server != src_server) {
226 		spin_lock(&src_server->nfs_client->cl_lock);
227 		list_del_init(&copy->src_copies);
228 		spin_unlock(&src_server->nfs_client->cl_lock);
229 	}
230 	if (status == -ERESTARTSYS) {
231 		goto out_cancel;
232 	} else if (copy->flags || copy->error == NFS4ERR_PARTNER_NO_AUTH) {
233 		status = -EAGAIN;
234 		*restart = true;
235 		goto out_cancel;
236 	}
237 out:
238 	res->write_res.count = copy->count;
239 	memcpy(&res->write_res.verifier, &copy->verf, sizeof(copy->verf));
240 	status = -copy->error;
241 
242 out_free:
243 	kfree(copy);
244 	return status;
245 out_cancel:
246 	nfs42_do_offload_cancel_async(dst, &copy->stateid);
247 	if (!nfs42_files_from_same_server(src, dst))
248 		nfs42_do_offload_cancel_async(src, src_stateid);
249 	goto out_free;
250 }
251 
252 static int process_copy_commit(struct file *dst, loff_t pos_dst,
253 			       struct nfs42_copy_res *res)
254 {
255 	struct nfs_commitres cres;
256 	int status = -ENOMEM;
257 
258 	cres.verf = kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS);
259 	if (!cres.verf)
260 		goto out;
261 
262 	status = nfs4_proc_commit(dst, pos_dst, res->write_res.count, &cres);
263 	if (status)
264 		goto out_free;
265 	if (nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
266 				    &cres.verf->verifier)) {
267 		dprintk("commit verf differs from copy verf\n");
268 		status = -EAGAIN;
269 	}
270 out_free:
271 	kfree(cres.verf);
272 out:
273 	return status;
274 }
275 
276 /**
277  * nfs42_copy_dest_done - perform inode cache updates after clone/copy offload
278  * @inode: pointer to destination inode
279  * @pos: destination offset
280  * @len: copy length
281  *
282  * Punch a hole in the inode page cache, so that the NFS client will
283  * know to retrieve new data.
284  * Update the file size if necessary, and then mark the inode as having
285  * invalid cached values for change attribute, ctime, mtime and space used.
286  */
287 static void nfs42_copy_dest_done(struct inode *inode, loff_t pos, loff_t len)
288 {
289 	loff_t newsize = pos + len;
290 	loff_t end = newsize - 1;
291 
292 	WARN_ON_ONCE(invalidate_inode_pages2_range(inode->i_mapping,
293 				pos >> PAGE_SHIFT, end >> PAGE_SHIFT));
294 
295 	spin_lock(&inode->i_lock);
296 	if (newsize > i_size_read(inode))
297 		i_size_write(inode, newsize);
298 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
299 					     NFS_INO_INVALID_CTIME |
300 					     NFS_INO_INVALID_MTIME |
301 					     NFS_INO_INVALID_BLOCKS);
302 	spin_unlock(&inode->i_lock);
303 }
304 
305 static ssize_t _nfs42_proc_copy(struct file *src,
306 				struct nfs_lock_context *src_lock,
307 				struct file *dst,
308 				struct nfs_lock_context *dst_lock,
309 				struct nfs42_copy_args *args,
310 				struct nfs42_copy_res *res,
311 				struct nl4_server *nss,
312 				nfs4_stateid *cnr_stateid,
313 				bool *restart)
314 {
315 	struct rpc_message msg = {
316 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY],
317 		.rpc_argp = args,
318 		.rpc_resp = res,
319 	};
320 	struct inode *dst_inode = file_inode(dst);
321 	struct inode *src_inode = file_inode(src);
322 	struct nfs_server *dst_server = NFS_SERVER(dst_inode);
323 	struct nfs_server *src_server = NFS_SERVER(src_inode);
324 	loff_t pos_src = args->src_pos;
325 	loff_t pos_dst = args->dst_pos;
326 	size_t count = args->count;
327 	ssize_t status;
328 
329 	if (nss) {
330 		args->cp_src = nss;
331 		nfs4_stateid_copy(&args->src_stateid, cnr_stateid);
332 	} else {
333 		status = nfs4_set_rw_stateid(&args->src_stateid,
334 				src_lock->open_context, src_lock, FMODE_READ);
335 		if (status) {
336 			if (status == -EAGAIN)
337 				status = -NFS4ERR_BAD_STATEID;
338 			return status;
339 		}
340 	}
341 	status = nfs_filemap_write_and_wait_range(file_inode(src)->i_mapping,
342 			pos_src, pos_src + (loff_t)count - 1);
343 	if (status)
344 		return status;
345 
346 	status = nfs4_set_rw_stateid(&args->dst_stateid, dst_lock->open_context,
347 				     dst_lock, FMODE_WRITE);
348 	if (status) {
349 		if (status == -EAGAIN)
350 			status = -NFS4ERR_BAD_STATEID;
351 		return status;
352 	}
353 
354 	status = nfs_sync_inode(dst_inode);
355 	if (status)
356 		return status;
357 
358 	res->commit_res.verf = NULL;
359 	if (args->sync) {
360 		res->commit_res.verf =
361 			kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS);
362 		if (!res->commit_res.verf)
363 			return -ENOMEM;
364 	}
365 	set_bit(NFS_CLNT_SRC_SSC_COPY_STATE,
366 		&src_lock->open_context->state->flags);
367 	set_bit(NFS_CLNT_DST_SSC_COPY_STATE,
368 		&dst_lock->open_context->state->flags);
369 
370 	status = nfs4_call_sync(dst_server->client, dst_server, &msg,
371 				&args->seq_args, &res->seq_res, 0);
372 	trace_nfs4_copy(src_inode, dst_inode, args, res, nss, status);
373 	if (status == -ENOTSUPP)
374 		dst_server->caps &= ~NFS_CAP_COPY;
375 	if (status)
376 		goto out;
377 
378 	if (args->sync &&
379 		nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
380 				    &res->commit_res.verf->verifier)) {
381 		status = -EAGAIN;
382 		goto out;
383 	}
384 
385 	if (!res->synchronous) {
386 		status = handle_async_copy(res, dst_server, src_server, src,
387 				dst, &args->src_stateid, restart);
388 		if (status)
389 			goto out;
390 	}
391 
392 	if ((!res->synchronous || !args->sync) &&
393 			res->write_res.verifier.committed != NFS_FILE_SYNC) {
394 		status = process_copy_commit(dst, pos_dst, res);
395 		if (status)
396 			goto out;
397 	}
398 
399 	nfs42_copy_dest_done(dst_inode, pos_dst, res->write_res.count);
400 	nfs_invalidate_atime(src_inode);
401 	status = res->write_res.count;
402 out:
403 	if (args->sync)
404 		kfree(res->commit_res.verf);
405 	return status;
406 }
407 
408 ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
409 			struct file *dst, loff_t pos_dst, size_t count,
410 			struct nl4_server *nss,
411 			nfs4_stateid *cnr_stateid, bool sync)
412 {
413 	struct nfs_server *server = NFS_SERVER(file_inode(dst));
414 	struct nfs_lock_context *src_lock;
415 	struct nfs_lock_context *dst_lock;
416 	struct nfs42_copy_args args = {
417 		.src_fh		= NFS_FH(file_inode(src)),
418 		.src_pos	= pos_src,
419 		.dst_fh		= NFS_FH(file_inode(dst)),
420 		.dst_pos	= pos_dst,
421 		.count		= count,
422 		.sync		= sync,
423 	};
424 	struct nfs42_copy_res res;
425 	struct nfs4_exception src_exception = {
426 		.inode		= file_inode(src),
427 		.stateid	= &args.src_stateid,
428 	};
429 	struct nfs4_exception dst_exception = {
430 		.inode		= file_inode(dst),
431 		.stateid	= &args.dst_stateid,
432 	};
433 	ssize_t err, err2;
434 	bool restart = false;
435 
436 	src_lock = nfs_get_lock_context(nfs_file_open_context(src));
437 	if (IS_ERR(src_lock))
438 		return PTR_ERR(src_lock);
439 
440 	src_exception.state = src_lock->open_context->state;
441 
442 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst));
443 	if (IS_ERR(dst_lock)) {
444 		err = PTR_ERR(dst_lock);
445 		goto out_put_src_lock;
446 	}
447 
448 	dst_exception.state = dst_lock->open_context->state;
449 
450 	do {
451 		inode_lock(file_inode(dst));
452 		err = _nfs42_proc_copy(src, src_lock,
453 				dst, dst_lock,
454 				&args, &res,
455 				nss, cnr_stateid, &restart);
456 		inode_unlock(file_inode(dst));
457 
458 		if (err >= 0)
459 			break;
460 		if (err == -ENOTSUPP &&
461 				nfs42_files_from_same_server(src, dst)) {
462 			err = -EOPNOTSUPP;
463 			break;
464 		} else if (err == -EAGAIN) {
465 			if (!restart) {
466 				dst_exception.retry = 1;
467 				continue;
468 			}
469 			break;
470 		} else if (err == -NFS4ERR_OFFLOAD_NO_REQS && !args.sync) {
471 			args.sync = true;
472 			dst_exception.retry = 1;
473 			continue;
474 		} else if ((err == -ESTALE ||
475 				err == -NFS4ERR_OFFLOAD_DENIED ||
476 				err == -ENOTSUPP) &&
477 				!nfs42_files_from_same_server(src, dst)) {
478 			nfs42_do_offload_cancel_async(src, &args.src_stateid);
479 			err = -EOPNOTSUPP;
480 			break;
481 		}
482 
483 		err2 = nfs4_handle_exception(server, err, &src_exception);
484 		err  = nfs4_handle_exception(server, err, &dst_exception);
485 		if (!err)
486 			err = err2;
487 	} while (src_exception.retry || dst_exception.retry);
488 
489 	nfs_put_lock_context(dst_lock);
490 out_put_src_lock:
491 	nfs_put_lock_context(src_lock);
492 	return err;
493 }
494 
495 struct nfs42_offloadcancel_data {
496 	struct nfs_server *seq_server;
497 	struct nfs42_offload_status_args args;
498 	struct nfs42_offload_status_res res;
499 };
500 
501 static void nfs42_offload_cancel_prepare(struct rpc_task *task, void *calldata)
502 {
503 	struct nfs42_offloadcancel_data *data = calldata;
504 
505 	nfs4_setup_sequence(data->seq_server->nfs_client,
506 				&data->args.osa_seq_args,
507 				&data->res.osr_seq_res, task);
508 }
509 
510 static void nfs42_offload_cancel_done(struct rpc_task *task, void *calldata)
511 {
512 	struct nfs42_offloadcancel_data *data = calldata;
513 
514 	trace_nfs4_offload_cancel(&data->args, task->tk_status);
515 	nfs41_sequence_done(task, &data->res.osr_seq_res);
516 	if (task->tk_status &&
517 		nfs4_async_handle_error(task, data->seq_server, NULL,
518 			NULL) == -EAGAIN)
519 		rpc_restart_call_prepare(task);
520 }
521 
522 static void nfs42_free_offloadcancel_data(void *data)
523 {
524 	kfree(data);
525 }
526 
527 static const struct rpc_call_ops nfs42_offload_cancel_ops = {
528 	.rpc_call_prepare = nfs42_offload_cancel_prepare,
529 	.rpc_call_done = nfs42_offload_cancel_done,
530 	.rpc_release = nfs42_free_offloadcancel_data,
531 };
532 
533 static int nfs42_do_offload_cancel_async(struct file *dst,
534 					 nfs4_stateid *stateid)
535 {
536 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
537 	struct nfs42_offloadcancel_data *data = NULL;
538 	struct nfs_open_context *ctx = nfs_file_open_context(dst);
539 	struct rpc_task *task;
540 	struct rpc_message msg = {
541 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OFFLOAD_CANCEL],
542 		.rpc_cred = ctx->cred,
543 	};
544 	struct rpc_task_setup task_setup_data = {
545 		.rpc_client = dst_server->client,
546 		.rpc_message = &msg,
547 		.callback_ops = &nfs42_offload_cancel_ops,
548 		.workqueue = nfsiod_workqueue,
549 		.flags = RPC_TASK_ASYNC,
550 	};
551 	int status;
552 
553 	if (!(dst_server->caps & NFS_CAP_OFFLOAD_CANCEL))
554 		return -EOPNOTSUPP;
555 
556 	data = kzalloc(sizeof(struct nfs42_offloadcancel_data), GFP_NOFS);
557 	if (data == NULL)
558 		return -ENOMEM;
559 
560 	data->seq_server = dst_server;
561 	data->args.osa_src_fh = NFS_FH(file_inode(dst));
562 	memcpy(&data->args.osa_stateid, stateid,
563 		sizeof(data->args.osa_stateid));
564 	msg.rpc_argp = &data->args;
565 	msg.rpc_resp = &data->res;
566 	task_setup_data.callback_data = data;
567 	nfs4_init_sequence(&data->args.osa_seq_args, &data->res.osr_seq_res,
568 			   1, 0);
569 	task = rpc_run_task(&task_setup_data);
570 	if (IS_ERR(task))
571 		return PTR_ERR(task);
572 	status = rpc_wait_for_completion_task(task);
573 	if (status == -ENOTSUPP)
574 		dst_server->caps &= ~NFS_CAP_OFFLOAD_CANCEL;
575 	rpc_put_task(task);
576 	return status;
577 }
578 
579 static int _nfs42_proc_copy_notify(struct file *src, struct file *dst,
580 				   struct nfs42_copy_notify_args *args,
581 				   struct nfs42_copy_notify_res *res)
582 {
583 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
584 	struct rpc_message msg = {
585 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY_NOTIFY],
586 		.rpc_argp = args,
587 		.rpc_resp = res,
588 	};
589 	int status;
590 	struct nfs_open_context *ctx;
591 	struct nfs_lock_context *l_ctx;
592 
593 	ctx = get_nfs_open_context(nfs_file_open_context(src));
594 	l_ctx = nfs_get_lock_context(ctx);
595 	if (IS_ERR(l_ctx))
596 		return PTR_ERR(l_ctx);
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 		return status;
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 	put_nfs_open_context(nfs_file_open_context(src));
614 	return status;
615 }
616 
617 int nfs42_proc_copy_notify(struct file *src, struct file *dst,
618 				struct nfs42_copy_notify_res *res)
619 {
620 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
621 	struct nfs42_copy_notify_args *args;
622 	struct nfs4_exception exception = {
623 		.inode = file_inode(src),
624 	};
625 	int status;
626 
627 	if (!(src_server->caps & NFS_CAP_COPY_NOTIFY))
628 		return -EOPNOTSUPP;
629 
630 	args = kzalloc(sizeof(struct nfs42_copy_notify_args), GFP_NOFS);
631 	if (args == NULL)
632 		return -ENOMEM;
633 
634 	args->cna_src_fh  = NFS_FH(file_inode(src)),
635 	args->cna_dst.nl4_type = NL4_NETADDR;
636 	nfs42_set_netaddr(dst, &args->cna_dst.u.nl4_addr);
637 	exception.stateid = &args->cna_src_stateid;
638 
639 	do {
640 		status = _nfs42_proc_copy_notify(src, dst, args, res);
641 		if (status == -ENOTSUPP) {
642 			status = -EOPNOTSUPP;
643 			goto out;
644 		}
645 		status = nfs4_handle_exception(src_server, status, &exception);
646 	} while (exception.retry);
647 
648 out:
649 	kfree(args);
650 	return status;
651 }
652 
653 static loff_t _nfs42_proc_llseek(struct file *filep,
654 		struct nfs_lock_context *lock, loff_t offset, int whence)
655 {
656 	struct inode *inode = file_inode(filep);
657 	struct nfs42_seek_args args = {
658 		.sa_fh		= NFS_FH(inode),
659 		.sa_offset	= offset,
660 		.sa_what	= (whence == SEEK_HOLE) ?
661 					NFS4_CONTENT_HOLE : NFS4_CONTENT_DATA,
662 	};
663 	struct nfs42_seek_res res;
664 	struct rpc_message msg = {
665 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEEK],
666 		.rpc_argp = &args,
667 		.rpc_resp = &res,
668 	};
669 	struct nfs_server *server = NFS_SERVER(inode);
670 	int status;
671 
672 	if (!nfs_server_capable(inode, NFS_CAP_SEEK))
673 		return -ENOTSUPP;
674 
675 	status = nfs4_set_rw_stateid(&args.sa_stateid, lock->open_context,
676 			lock, FMODE_READ);
677 	if (status) {
678 		if (status == -EAGAIN)
679 			status = -NFS4ERR_BAD_STATEID;
680 		return status;
681 	}
682 
683 	status = nfs_filemap_write_and_wait_range(inode->i_mapping,
684 			offset, LLONG_MAX);
685 	if (status)
686 		return status;
687 
688 	status = nfs4_call_sync(server->client, server, &msg,
689 				&args.seq_args, &res.seq_res, 0);
690 	trace_nfs4_llseek(inode, &args, &res, status);
691 	if (status == -ENOTSUPP)
692 		server->caps &= ~NFS_CAP_SEEK;
693 	if (status)
694 		return status;
695 
696 	if (whence == SEEK_DATA && res.sr_eof)
697 		return -NFS4ERR_NXIO;
698 	else
699 		return vfs_setpos(filep, res.sr_offset, inode->i_sb->s_maxbytes);
700 }
701 
702 loff_t nfs42_proc_llseek(struct file *filep, loff_t offset, int whence)
703 {
704 	struct nfs_server *server = NFS_SERVER(file_inode(filep));
705 	struct nfs4_exception exception = { };
706 	struct nfs_lock_context *lock;
707 	loff_t err;
708 
709 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
710 	if (IS_ERR(lock))
711 		return PTR_ERR(lock);
712 
713 	exception.inode = file_inode(filep);
714 	exception.state = lock->open_context->state;
715 
716 	do {
717 		err = _nfs42_proc_llseek(filep, lock, offset, whence);
718 		if (err >= 0)
719 			break;
720 		if (err == -ENOTSUPP) {
721 			err = -EOPNOTSUPP;
722 			break;
723 		}
724 		err = nfs4_handle_exception(server, err, &exception);
725 	} while (exception.retry);
726 
727 	nfs_put_lock_context(lock);
728 	return err;
729 }
730 
731 
732 static void
733 nfs42_layoutstat_prepare(struct rpc_task *task, void *calldata)
734 {
735 	struct nfs42_layoutstat_data *data = calldata;
736 	struct inode *inode = data->inode;
737 	struct nfs_server *server = NFS_SERVER(inode);
738 	struct pnfs_layout_hdr *lo;
739 
740 	spin_lock(&inode->i_lock);
741 	lo = NFS_I(inode)->layout;
742 	if (!pnfs_layout_is_valid(lo)) {
743 		spin_unlock(&inode->i_lock);
744 		rpc_exit(task, 0);
745 		return;
746 	}
747 	nfs4_stateid_copy(&data->args.stateid, &lo->plh_stateid);
748 	spin_unlock(&inode->i_lock);
749 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
750 			    &data->res.seq_res, task);
751 }
752 
753 static void
754 nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
755 {
756 	struct nfs42_layoutstat_data *data = calldata;
757 	struct inode *inode = data->inode;
758 	struct pnfs_layout_hdr *lo;
759 
760 	if (!nfs4_sequence_done(task, &data->res.seq_res))
761 		return;
762 
763 	switch (task->tk_status) {
764 	case 0:
765 		return;
766 	case -NFS4ERR_BADHANDLE:
767 	case -ESTALE:
768 		pnfs_destroy_layout(NFS_I(inode));
769 		break;
770 	case -NFS4ERR_EXPIRED:
771 	case -NFS4ERR_ADMIN_REVOKED:
772 	case -NFS4ERR_DELEG_REVOKED:
773 	case -NFS4ERR_STALE_STATEID:
774 	case -NFS4ERR_BAD_STATEID:
775 		spin_lock(&inode->i_lock);
776 		lo = NFS_I(inode)->layout;
777 		if (pnfs_layout_is_valid(lo) &&
778 		    nfs4_stateid_match(&data->args.stateid,
779 					     &lo->plh_stateid)) {
780 			LIST_HEAD(head);
781 
782 			/*
783 			 * Mark the bad layout state as invalid, then retry
784 			 * with the current stateid.
785 			 */
786 			pnfs_mark_layout_stateid_invalid(lo, &head);
787 			spin_unlock(&inode->i_lock);
788 			pnfs_free_lseg_list(&head);
789 			nfs_commit_inode(inode, 0);
790 		} else
791 			spin_unlock(&inode->i_lock);
792 		break;
793 	case -NFS4ERR_OLD_STATEID:
794 		spin_lock(&inode->i_lock);
795 		lo = NFS_I(inode)->layout;
796 		if (pnfs_layout_is_valid(lo) &&
797 		    nfs4_stateid_match_other(&data->args.stateid,
798 					&lo->plh_stateid)) {
799 			/* Do we need to delay before resending? */
800 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
801 						&data->args.stateid))
802 				rpc_delay(task, HZ);
803 			rpc_restart_call_prepare(task);
804 		}
805 		spin_unlock(&inode->i_lock);
806 		break;
807 	case -ENOTSUPP:
808 	case -EOPNOTSUPP:
809 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTSTATS;
810 	}
811 
812 	trace_nfs4_layoutstats(inode, &data->args.stateid, task->tk_status);
813 }
814 
815 static void
816 nfs42_layoutstat_release(void *calldata)
817 {
818 	struct nfs42_layoutstat_data *data = calldata;
819 	struct nfs42_layoutstat_devinfo *devinfo = data->args.devinfo;
820 	int i;
821 
822 	for (i = 0; i < data->args.num_dev; i++) {
823 		if (devinfo[i].ld_private.ops && devinfo[i].ld_private.ops->free)
824 			devinfo[i].ld_private.ops->free(&devinfo[i].ld_private);
825 	}
826 
827 	pnfs_put_layout_hdr(NFS_I(data->args.inode)->layout);
828 	smp_mb__before_atomic();
829 	clear_bit(NFS_INO_LAYOUTSTATS, &NFS_I(data->args.inode)->flags);
830 	smp_mb__after_atomic();
831 	nfs_iput_and_deactive(data->inode);
832 	kfree(data->args.devinfo);
833 	kfree(data);
834 }
835 
836 static const struct rpc_call_ops nfs42_layoutstat_ops = {
837 	.rpc_call_prepare = nfs42_layoutstat_prepare,
838 	.rpc_call_done = nfs42_layoutstat_done,
839 	.rpc_release = nfs42_layoutstat_release,
840 };
841 
842 int nfs42_proc_layoutstats_generic(struct nfs_server *server,
843 				   struct nfs42_layoutstat_data *data)
844 {
845 	struct rpc_message msg = {
846 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTSTATS],
847 		.rpc_argp = &data->args,
848 		.rpc_resp = &data->res,
849 	};
850 	struct rpc_task_setup task_setup = {
851 		.rpc_client = server->client,
852 		.rpc_message = &msg,
853 		.callback_ops = &nfs42_layoutstat_ops,
854 		.callback_data = data,
855 		.flags = RPC_TASK_ASYNC,
856 	};
857 	struct rpc_task *task;
858 
859 	data->inode = nfs_igrab_and_active(data->args.inode);
860 	if (!data->inode) {
861 		nfs42_layoutstat_release(data);
862 		return -EAGAIN;
863 	}
864 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
865 	task = rpc_run_task(&task_setup);
866 	if (IS_ERR(task))
867 		return PTR_ERR(task);
868 	rpc_put_task(task);
869 	return 0;
870 }
871 
872 static struct nfs42_layouterror_data *
873 nfs42_alloc_layouterror_data(struct pnfs_layout_segment *lseg, gfp_t gfp_flags)
874 {
875 	struct nfs42_layouterror_data *data;
876 	struct inode *inode = lseg->pls_layout->plh_inode;
877 
878 	data = kzalloc(sizeof(*data), gfp_flags);
879 	if (data) {
880 		data->args.inode = data->inode = nfs_igrab_and_active(inode);
881 		if (data->inode) {
882 			data->lseg = pnfs_get_lseg(lseg);
883 			if (data->lseg)
884 				return data;
885 			nfs_iput_and_deactive(data->inode);
886 		}
887 		kfree(data);
888 	}
889 	return NULL;
890 }
891 
892 static void
893 nfs42_free_layouterror_data(struct nfs42_layouterror_data *data)
894 {
895 	pnfs_put_lseg(data->lseg);
896 	nfs_iput_and_deactive(data->inode);
897 	kfree(data);
898 }
899 
900 static void
901 nfs42_layouterror_prepare(struct rpc_task *task, void *calldata)
902 {
903 	struct nfs42_layouterror_data *data = calldata;
904 	struct inode *inode = data->inode;
905 	struct nfs_server *server = NFS_SERVER(inode);
906 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
907 	unsigned i;
908 
909 	spin_lock(&inode->i_lock);
910 	if (!pnfs_layout_is_valid(lo)) {
911 		spin_unlock(&inode->i_lock);
912 		rpc_exit(task, 0);
913 		return;
914 	}
915 	for (i = 0; i < data->args.num_errors; i++)
916 		nfs4_stateid_copy(&data->args.errors[i].stateid,
917 				&lo->plh_stateid);
918 	spin_unlock(&inode->i_lock);
919 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
920 			    &data->res.seq_res, task);
921 }
922 
923 static void
924 nfs42_layouterror_done(struct rpc_task *task, void *calldata)
925 {
926 	struct nfs42_layouterror_data *data = calldata;
927 	struct inode *inode = data->inode;
928 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
929 
930 	if (!nfs4_sequence_done(task, &data->res.seq_res))
931 		return;
932 
933 	switch (task->tk_status) {
934 	case 0:
935 		return;
936 	case -NFS4ERR_BADHANDLE:
937 	case -ESTALE:
938 		pnfs_destroy_layout(NFS_I(inode));
939 		break;
940 	case -NFS4ERR_EXPIRED:
941 	case -NFS4ERR_ADMIN_REVOKED:
942 	case -NFS4ERR_DELEG_REVOKED:
943 	case -NFS4ERR_STALE_STATEID:
944 	case -NFS4ERR_BAD_STATEID:
945 		spin_lock(&inode->i_lock);
946 		if (pnfs_layout_is_valid(lo) &&
947 		    nfs4_stateid_match(&data->args.errors[0].stateid,
948 					     &lo->plh_stateid)) {
949 			LIST_HEAD(head);
950 
951 			/*
952 			 * Mark the bad layout state as invalid, then retry
953 			 * with the current stateid.
954 			 */
955 			pnfs_mark_layout_stateid_invalid(lo, &head);
956 			spin_unlock(&inode->i_lock);
957 			pnfs_free_lseg_list(&head);
958 			nfs_commit_inode(inode, 0);
959 		} else
960 			spin_unlock(&inode->i_lock);
961 		break;
962 	case -NFS4ERR_OLD_STATEID:
963 		spin_lock(&inode->i_lock);
964 		if (pnfs_layout_is_valid(lo) &&
965 		    nfs4_stateid_match_other(&data->args.errors[0].stateid,
966 					&lo->plh_stateid)) {
967 			/* Do we need to delay before resending? */
968 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
969 						&data->args.errors[0].stateid))
970 				rpc_delay(task, HZ);
971 			rpc_restart_call_prepare(task);
972 		}
973 		spin_unlock(&inode->i_lock);
974 		break;
975 	case -ENOTSUPP:
976 	case -EOPNOTSUPP:
977 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTERROR;
978 	}
979 
980 	trace_nfs4_layouterror(inode, &data->args.errors[0].stateid,
981 			       task->tk_status);
982 }
983 
984 static void
985 nfs42_layouterror_release(void *calldata)
986 {
987 	struct nfs42_layouterror_data *data = calldata;
988 
989 	nfs42_free_layouterror_data(data);
990 }
991 
992 static const struct rpc_call_ops nfs42_layouterror_ops = {
993 	.rpc_call_prepare = nfs42_layouterror_prepare,
994 	.rpc_call_done = nfs42_layouterror_done,
995 	.rpc_release = nfs42_layouterror_release,
996 };
997 
998 int nfs42_proc_layouterror(struct pnfs_layout_segment *lseg,
999 		const struct nfs42_layout_error *errors, size_t n)
1000 {
1001 	struct inode *inode = lseg->pls_layout->plh_inode;
1002 	struct nfs42_layouterror_data *data;
1003 	struct rpc_task *task;
1004 	struct rpc_message msg = {
1005 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTERROR],
1006 	};
1007 	struct rpc_task_setup task_setup = {
1008 		.rpc_message = &msg,
1009 		.callback_ops = &nfs42_layouterror_ops,
1010 		.flags = RPC_TASK_ASYNC,
1011 	};
1012 	unsigned int i;
1013 
1014 	if (!nfs_server_capable(inode, NFS_CAP_LAYOUTERROR))
1015 		return -EOPNOTSUPP;
1016 	if (n > NFS42_LAYOUTERROR_MAX)
1017 		return -EINVAL;
1018 	data = nfs42_alloc_layouterror_data(lseg, GFP_NOFS);
1019 	if (!data)
1020 		return -ENOMEM;
1021 	for (i = 0; i < n; i++) {
1022 		data->args.errors[i] = errors[i];
1023 		data->args.num_errors++;
1024 		data->res.num_errors++;
1025 	}
1026 	msg.rpc_argp = &data->args;
1027 	msg.rpc_resp = &data->res;
1028 	task_setup.callback_data = data;
1029 	task_setup.rpc_client = NFS_SERVER(inode)->client;
1030 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
1031 	task = rpc_run_task(&task_setup);
1032 	if (IS_ERR(task))
1033 		return PTR_ERR(task);
1034 	rpc_put_task(task);
1035 	return 0;
1036 }
1037 EXPORT_SYMBOL_GPL(nfs42_proc_layouterror);
1038 
1039 static int _nfs42_proc_clone(struct rpc_message *msg, struct file *src_f,
1040 		struct file *dst_f, struct nfs_lock_context *src_lock,
1041 		struct nfs_lock_context *dst_lock, loff_t src_offset,
1042 		loff_t dst_offset, loff_t count)
1043 {
1044 	struct inode *src_inode = file_inode(src_f);
1045 	struct inode *dst_inode = file_inode(dst_f);
1046 	struct nfs_server *server = NFS_SERVER(dst_inode);
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 = server->cache_consistency_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 	status = nfs4_call_sync(server->client, server, msg,
1083 				&args.seq_args, &res.seq_res, 0);
1084 	trace_nfs4_clone(src_inode, dst_inode, &args, status);
1085 	if (status == 0) {
1086 		nfs42_copy_dest_done(dst_inode, dst_offset, count);
1087 		status = nfs_post_op_update_inode(dst_inode, res.dst_fattr);
1088 	}
1089 
1090 	kfree(res.dst_fattr);
1091 	return status;
1092 }
1093 
1094 int nfs42_proc_clone(struct file *src_f, struct file *dst_f,
1095 		     loff_t src_offset, loff_t dst_offset, loff_t count)
1096 {
1097 	struct rpc_message msg = {
1098 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLONE],
1099 	};
1100 	struct inode *inode = file_inode(src_f);
1101 	struct nfs_server *server = NFS_SERVER(file_inode(src_f));
1102 	struct nfs_lock_context *src_lock;
1103 	struct nfs_lock_context *dst_lock;
1104 	struct nfs4_exception src_exception = { };
1105 	struct nfs4_exception dst_exception = { };
1106 	int err, err2;
1107 
1108 	if (!nfs_server_capable(inode, NFS_CAP_CLONE))
1109 		return -EOPNOTSUPP;
1110 
1111 	src_lock = nfs_get_lock_context(nfs_file_open_context(src_f));
1112 	if (IS_ERR(src_lock))
1113 		return PTR_ERR(src_lock);
1114 
1115 	src_exception.inode = file_inode(src_f);
1116 	src_exception.state = src_lock->open_context->state;
1117 
1118 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst_f));
1119 	if (IS_ERR(dst_lock)) {
1120 		err = PTR_ERR(dst_lock);
1121 		goto out_put_src_lock;
1122 	}
1123 
1124 	dst_exception.inode = file_inode(dst_f);
1125 	dst_exception.state = dst_lock->open_context->state;
1126 
1127 	do {
1128 		err = _nfs42_proc_clone(&msg, src_f, dst_f, src_lock, dst_lock,
1129 					src_offset, dst_offset, count);
1130 		if (err == -ENOTSUPP || err == -EOPNOTSUPP) {
1131 			NFS_SERVER(inode)->caps &= ~NFS_CAP_CLONE;
1132 			err = -EOPNOTSUPP;
1133 			break;
1134 		}
1135 
1136 		err2 = nfs4_handle_exception(server, err, &src_exception);
1137 		err = nfs4_handle_exception(server, err, &dst_exception);
1138 		if (!err)
1139 			err = err2;
1140 	} while (src_exception.retry || dst_exception.retry);
1141 
1142 	nfs_put_lock_context(dst_lock);
1143 out_put_src_lock:
1144 	nfs_put_lock_context(src_lock);
1145 	return err;
1146 }
1147 
1148 #define NFS4XATTR_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
1149 
1150 static int _nfs42_proc_removexattr(struct inode *inode, const char *name)
1151 {
1152 	struct nfs_server *server = NFS_SERVER(inode);
1153 	struct nfs42_removexattrargs args = {
1154 		.fh = NFS_FH(inode),
1155 		.xattr_name = name,
1156 	};
1157 	struct nfs42_removexattrres res;
1158 	struct rpc_message msg = {
1159 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVEXATTR],
1160 		.rpc_argp = &args,
1161 		.rpc_resp = &res,
1162 	};
1163 	int ret;
1164 	unsigned long timestamp = jiffies;
1165 
1166 	ret = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
1167 	    &res.seq_res, 1);
1168 	if (!ret)
1169 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1170 
1171 	return ret;
1172 }
1173 
1174 static int _nfs42_proc_setxattr(struct inode *inode, const char *name,
1175 				const void *buf, size_t buflen, int flags)
1176 {
1177 	struct nfs_server *server = NFS_SERVER(inode);
1178 	struct page *pages[NFS4XATTR_MAXPAGES];
1179 	struct nfs42_setxattrargs arg = {
1180 		.fh		= NFS_FH(inode),
1181 		.xattr_pages	= pages,
1182 		.xattr_len	= buflen,
1183 		.xattr_name	= name,
1184 		.xattr_flags	= flags,
1185 	};
1186 	struct nfs42_setxattrres res;
1187 	struct rpc_message msg = {
1188 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETXATTR],
1189 		.rpc_argp	= &arg,
1190 		.rpc_resp	= &res,
1191 	};
1192 	int ret, np;
1193 	unsigned long timestamp = jiffies;
1194 
1195 	if (buflen > server->sxasize)
1196 		return -ERANGE;
1197 
1198 	if (buflen > 0) {
1199 		np = nfs4_buf_to_pages_noslab(buf, buflen, arg.xattr_pages);
1200 		if (np < 0)
1201 			return np;
1202 	} else
1203 		np = 0;
1204 
1205 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1206 	    &res.seq_res, 1);
1207 
1208 	for (; np > 0; np--)
1209 		put_page(pages[np - 1]);
1210 
1211 	if (!ret)
1212 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1213 
1214 	return ret;
1215 }
1216 
1217 static ssize_t _nfs42_proc_getxattr(struct inode *inode, const char *name,
1218 				void *buf, size_t buflen, struct page **pages,
1219 				size_t plen)
1220 {
1221 	struct nfs_server *server = NFS_SERVER(inode);
1222 	struct nfs42_getxattrargs arg = {
1223 		.fh		= NFS_FH(inode),
1224 		.xattr_name	= name,
1225 	};
1226 	struct nfs42_getxattrres res;
1227 	struct rpc_message msg = {
1228 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETXATTR],
1229 		.rpc_argp	= &arg,
1230 		.rpc_resp	= &res,
1231 	};
1232 	ssize_t ret;
1233 
1234 	arg.xattr_len = plen;
1235 	arg.xattr_pages = pages;
1236 
1237 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1238 	    &res.seq_res, 0);
1239 	if (ret < 0)
1240 		return ret;
1241 
1242 	/*
1243 	 * Normally, the caching is done one layer up, but for successful
1244 	 * RPCS, always cache the result here, even if the caller was
1245 	 * just querying the length, or if the reply was too big for
1246 	 * the caller. This avoids a second RPC in the case of the
1247 	 * common query-alloc-retrieve cycle for xattrs.
1248 	 *
1249 	 * Note that xattr_len is always capped to XATTR_SIZE_MAX.
1250 	 */
1251 
1252 	nfs4_xattr_cache_add(inode, name, NULL, pages, res.xattr_len);
1253 
1254 	if (buflen) {
1255 		if (res.xattr_len > buflen)
1256 			return -ERANGE;
1257 		_copy_from_pages(buf, pages, 0, res.xattr_len);
1258 	}
1259 
1260 	return res.xattr_len;
1261 }
1262 
1263 static ssize_t _nfs42_proc_listxattrs(struct inode *inode, void *buf,
1264 				 size_t buflen, u64 *cookiep, bool *eofp)
1265 {
1266 	struct nfs_server *server = NFS_SERVER(inode);
1267 	struct page **pages;
1268 	struct nfs42_listxattrsargs arg = {
1269 		.fh		= NFS_FH(inode),
1270 		.cookie		= *cookiep,
1271 	};
1272 	struct nfs42_listxattrsres res = {
1273 		.eof = false,
1274 		.xattr_buf = buf,
1275 		.xattr_len = buflen,
1276 	};
1277 	struct rpc_message msg = {
1278 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LISTXATTRS],
1279 		.rpc_argp	= &arg,
1280 		.rpc_resp	= &res,
1281 	};
1282 	u32 xdrlen;
1283 	int ret, np, i;
1284 
1285 
1286 	ret = -ENOMEM;
1287 	res.scratch = alloc_page(GFP_KERNEL);
1288 	if (!res.scratch)
1289 		goto out;
1290 
1291 	xdrlen = nfs42_listxattr_xdrsize(buflen);
1292 	if (xdrlen > server->lxasize)
1293 		xdrlen = server->lxasize;
1294 	np = xdrlen / PAGE_SIZE + 1;
1295 
1296 	pages = kcalloc(np, sizeof(struct page *), GFP_KERNEL);
1297 	if (!pages)
1298 		goto out_free_scratch;
1299 	for (i = 0; i < np; i++) {
1300 		pages[i] = alloc_page(GFP_KERNEL);
1301 		if (!pages[i])
1302 			goto out_free_pages;
1303 	}
1304 
1305 	arg.xattr_pages = pages;
1306 	arg.count = xdrlen;
1307 
1308 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1309 	    &res.seq_res, 0);
1310 
1311 	if (ret >= 0) {
1312 		ret = res.copied;
1313 		*cookiep = res.cookie;
1314 		*eofp = res.eof;
1315 	}
1316 
1317 out_free_pages:
1318 	while (--np >= 0) {
1319 		if (pages[np])
1320 			__free_page(pages[np]);
1321 	}
1322 	kfree(pages);
1323 out_free_scratch:
1324 	__free_page(res.scratch);
1325 out:
1326 	return ret;
1327 
1328 }
1329 
1330 ssize_t nfs42_proc_getxattr(struct inode *inode, const char *name,
1331 			      void *buf, size_t buflen)
1332 {
1333 	struct nfs4_exception exception = { };
1334 	ssize_t err, np, i;
1335 	struct page **pages;
1336 
1337 	np = nfs_page_array_len(0, buflen ?: XATTR_SIZE_MAX);
1338 	pages = kmalloc_array(np, sizeof(*pages), GFP_KERNEL);
1339 	if (!pages)
1340 		return -ENOMEM;
1341 
1342 	for (i = 0; i < np; i++) {
1343 		pages[i] = alloc_page(GFP_KERNEL);
1344 		if (!pages[i]) {
1345 			np = i + 1;
1346 			err = -ENOMEM;
1347 			goto out;
1348 		}
1349 	}
1350 
1351 	/*
1352 	 * The GETXATTR op has no length field in the call, and the
1353 	 * xattr data is at the end of the reply.
1354 	 *
1355 	 * There is no downside in using the page-aligned length. It will
1356 	 * allow receiving and caching xattrs that are too large for the
1357 	 * caller but still fit in the page-rounded value.
1358 	 */
1359 	do {
1360 		err = _nfs42_proc_getxattr(inode, name, buf, buflen,
1361 			pages, np * PAGE_SIZE);
1362 		if (err >= 0)
1363 			break;
1364 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1365 				&exception);
1366 	} while (exception.retry);
1367 
1368 out:
1369 	while (--np >= 0)
1370 		__free_page(pages[np]);
1371 	kfree(pages);
1372 
1373 	return err;
1374 }
1375 
1376 int nfs42_proc_setxattr(struct inode *inode, const char *name,
1377 			      const void *buf, size_t buflen, int flags)
1378 {
1379 	struct nfs4_exception exception = { };
1380 	int err;
1381 
1382 	do {
1383 		err = _nfs42_proc_setxattr(inode, name, buf, buflen, flags);
1384 		if (!err)
1385 			break;
1386 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1387 				&exception);
1388 	} while (exception.retry);
1389 
1390 	return err;
1391 }
1392 
1393 ssize_t nfs42_proc_listxattrs(struct inode *inode, void *buf,
1394 			      size_t buflen, u64 *cookiep, bool *eofp)
1395 {
1396 	struct nfs4_exception exception = { };
1397 	ssize_t err;
1398 
1399 	do {
1400 		err = _nfs42_proc_listxattrs(inode, buf, buflen,
1401 		    cookiep, eofp);
1402 		if (err >= 0)
1403 			break;
1404 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1405 				&exception);
1406 	} while (exception.retry);
1407 
1408 	return err;
1409 }
1410 
1411 int nfs42_proc_removexattr(struct inode *inode, const char *name)
1412 {
1413 	struct nfs4_exception exception = { };
1414 	int err;
1415 
1416 	do {
1417 		err = _nfs42_proc_removexattr(inode, name);
1418 		if (!err)
1419 			break;
1420 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1421 				&exception);
1422 	} while (exception.retry);
1423 
1424 	return err;
1425 }
1426