xref: /linux/fs/nfs/nfs42proc.c (revision 4b660dbd9ee2059850fd30e0df420ca7a38a1856)
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 		if (nfs_should_remove_suid(inode)) {
83 			spin_lock(&inode->i_lock);
84 			nfs_set_cache_invalid(inode,
85 				NFS_INO_REVAL_FORCED | NFS_INO_INVALID_MODE);
86 			spin_unlock(&inode->i_lock);
87 		}
88 		status = nfs_post_op_update_inode_force_wcc(inode,
89 							    res.falloc_fattr);
90 	}
91 	if (msg->rpc_proc == &nfs4_procedures[NFSPROC4_CLNT_ALLOCATE])
92 		trace_nfs4_fallocate(inode, &args, status);
93 	else
94 		trace_nfs4_deallocate(inode, &args, status);
95 	kfree(res.falloc_fattr);
96 	return status;
97 }
98 
99 static int nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep,
100 				loff_t offset, loff_t len)
101 {
102 	struct inode *inode = file_inode(filep);
103 	struct nfs_server *server = NFS_SERVER(inode);
104 	struct nfs4_exception exception = { };
105 	struct nfs_lock_context *lock;
106 	int err;
107 
108 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
109 	if (IS_ERR(lock))
110 		return PTR_ERR(lock);
111 
112 	exception.inode = inode;
113 	exception.state = lock->open_context->state;
114 
115 	err = nfs_sync_inode(inode);
116 	if (err)
117 		goto out;
118 
119 	do {
120 		err = _nfs42_proc_fallocate(msg, filep, lock, offset, len);
121 		if (err == -ENOTSUPP) {
122 			err = -EOPNOTSUPP;
123 			break;
124 		}
125 		err = nfs4_handle_exception(server, err, &exception);
126 	} while (exception.retry);
127 out:
128 	nfs_put_lock_context(lock);
129 	return err;
130 }
131 
132 int nfs42_proc_allocate(struct file *filep, loff_t offset, loff_t len)
133 {
134 	struct rpc_message msg = {
135 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ALLOCATE],
136 	};
137 	struct inode *inode = file_inode(filep);
138 	int err;
139 
140 	if (!nfs_server_capable(inode, NFS_CAP_ALLOCATE))
141 		return -EOPNOTSUPP;
142 
143 	inode_lock(inode);
144 
145 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
146 	if (err == -EOPNOTSUPP)
147 		NFS_SERVER(inode)->caps &= ~NFS_CAP_ALLOCATE;
148 
149 	inode_unlock(inode);
150 	return err;
151 }
152 
153 int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len)
154 {
155 	struct rpc_message msg = {
156 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DEALLOCATE],
157 	};
158 	struct inode *inode = file_inode(filep);
159 	int err;
160 
161 	if (!nfs_server_capable(inode, NFS_CAP_DEALLOCATE))
162 		return -EOPNOTSUPP;
163 
164 	inode_lock(inode);
165 
166 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
167 	if (err == 0)
168 		truncate_pagecache_range(inode, offset, (offset + len) -1);
169 	if (err == -EOPNOTSUPP)
170 		NFS_SERVER(inode)->caps &= ~NFS_CAP_DEALLOCATE;
171 
172 	inode_unlock(inode);
173 	return err;
174 }
175 
176 static int handle_async_copy(struct nfs42_copy_res *res,
177 			     struct nfs_server *dst_server,
178 			     struct nfs_server *src_server,
179 			     struct file *src,
180 			     struct file *dst,
181 			     nfs4_stateid *src_stateid,
182 			     bool *restart)
183 {
184 	struct nfs4_copy_state *copy, *tmp_copy = NULL, *iter;
185 	int status = NFS4_OK;
186 	struct nfs_open_context *dst_ctx = nfs_file_open_context(dst);
187 	struct nfs_open_context *src_ctx = nfs_file_open_context(src);
188 
189 	copy = kzalloc(sizeof(struct nfs4_copy_state), GFP_KERNEL);
190 	if (!copy)
191 		return -ENOMEM;
192 
193 	spin_lock(&dst_server->nfs_client->cl_lock);
194 	list_for_each_entry(iter,
195 				&dst_server->nfs_client->pending_cb_stateids,
196 				copies) {
197 		if (memcmp(&res->write_res.stateid, &iter->stateid,
198 				NFS4_STATEID_SIZE))
199 			continue;
200 		tmp_copy = iter;
201 		list_del(&iter->copies);
202 		break;
203 	}
204 	if (tmp_copy) {
205 		spin_unlock(&dst_server->nfs_client->cl_lock);
206 		kfree(copy);
207 		copy = tmp_copy;
208 		goto out;
209 	}
210 
211 	memcpy(&copy->stateid, &res->write_res.stateid, NFS4_STATEID_SIZE);
212 	init_completion(&copy->completion);
213 	copy->parent_dst_state = dst_ctx->state;
214 	copy->parent_src_state = src_ctx->state;
215 
216 	list_add_tail(&copy->copies, &dst_server->ss_copies);
217 	spin_unlock(&dst_server->nfs_client->cl_lock);
218 
219 	if (dst_server != src_server) {
220 		spin_lock(&src_server->nfs_client->cl_lock);
221 		list_add_tail(&copy->src_copies, &src_server->ss_copies);
222 		spin_unlock(&src_server->nfs_client->cl_lock);
223 	}
224 
225 	status = wait_for_completion_interruptible(&copy->completion);
226 	spin_lock(&dst_server->nfs_client->cl_lock);
227 	list_del_init(&copy->copies);
228 	spin_unlock(&dst_server->nfs_client->cl_lock);
229 	if (dst_server != src_server) {
230 		spin_lock(&src_server->nfs_client->cl_lock);
231 		list_del_init(&copy->src_copies);
232 		spin_unlock(&src_server->nfs_client->cl_lock);
233 	}
234 	if (status == -ERESTARTSYS) {
235 		goto out_cancel;
236 	} else if (copy->flags || copy->error == NFS4ERR_PARTNER_NO_AUTH) {
237 		status = -EAGAIN;
238 		*restart = true;
239 		goto out_cancel;
240 	}
241 out:
242 	res->write_res.count = copy->count;
243 	memcpy(&res->write_res.verifier, &copy->verf, sizeof(copy->verf));
244 	status = -copy->error;
245 
246 out_free:
247 	kfree(copy);
248 	return status;
249 out_cancel:
250 	nfs42_do_offload_cancel_async(dst, &copy->stateid);
251 	if (!nfs42_files_from_same_server(src, dst))
252 		nfs42_do_offload_cancel_async(src, src_stateid);
253 	goto out_free;
254 }
255 
256 static int process_copy_commit(struct file *dst, loff_t pos_dst,
257 			       struct nfs42_copy_res *res)
258 {
259 	struct nfs_commitres cres;
260 	int status = -ENOMEM;
261 
262 	cres.verf = kzalloc(sizeof(struct nfs_writeverf), GFP_KERNEL);
263 	if (!cres.verf)
264 		goto out;
265 
266 	status = nfs4_proc_commit(dst, pos_dst, res->write_res.count, &cres);
267 	if (status)
268 		goto out_free;
269 	if (nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
270 				    &cres.verf->verifier)) {
271 		dprintk("commit verf differs from copy verf\n");
272 		status = -EAGAIN;
273 	}
274 out_free:
275 	kfree(cres.verf);
276 out:
277 	return status;
278 }
279 
280 /**
281  * nfs42_copy_dest_done - perform inode cache updates after clone/copy offload
282  * @inode: pointer to destination inode
283  * @pos: destination offset
284  * @len: copy length
285  *
286  * Punch a hole in the inode page cache, so that the NFS client will
287  * know to retrieve new data.
288  * Update the file size if necessary, and then mark the inode as having
289  * invalid cached values for change attribute, ctime, mtime and space used.
290  */
291 static void nfs42_copy_dest_done(struct inode *inode, loff_t pos, loff_t len)
292 {
293 	loff_t newsize = pos + len;
294 	loff_t end = newsize - 1;
295 
296 	WARN_ON_ONCE(invalidate_inode_pages2_range(inode->i_mapping,
297 				pos >> PAGE_SHIFT, end >> PAGE_SHIFT));
298 
299 	spin_lock(&inode->i_lock);
300 	if (newsize > i_size_read(inode))
301 		i_size_write(inode, newsize);
302 	nfs_set_cache_invalid(inode, NFS_INO_INVALID_CHANGE |
303 					     NFS_INO_INVALID_CTIME |
304 					     NFS_INO_INVALID_MTIME |
305 					     NFS_INO_INVALID_BLOCKS);
306 	spin_unlock(&inode->i_lock);
307 }
308 
309 static ssize_t _nfs42_proc_copy(struct file *src,
310 				struct nfs_lock_context *src_lock,
311 				struct file *dst,
312 				struct nfs_lock_context *dst_lock,
313 				struct nfs42_copy_args *args,
314 				struct nfs42_copy_res *res,
315 				struct nl4_server *nss,
316 				nfs4_stateid *cnr_stateid,
317 				bool *restart)
318 {
319 	struct rpc_message msg = {
320 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY],
321 		.rpc_argp = args,
322 		.rpc_resp = res,
323 	};
324 	struct inode *dst_inode = file_inode(dst);
325 	struct inode *src_inode = file_inode(src);
326 	struct nfs_server *dst_server = NFS_SERVER(dst_inode);
327 	struct nfs_server *src_server = NFS_SERVER(src_inode);
328 	loff_t pos_src = args->src_pos;
329 	loff_t pos_dst = args->dst_pos;
330 	size_t count = args->count;
331 	ssize_t status;
332 
333 	if (nss) {
334 		args->cp_src = nss;
335 		nfs4_stateid_copy(&args->src_stateid, cnr_stateid);
336 	} else {
337 		status = nfs4_set_rw_stateid(&args->src_stateid,
338 				src_lock->open_context, src_lock, FMODE_READ);
339 		if (status) {
340 			if (status == -EAGAIN)
341 				status = -NFS4ERR_BAD_STATEID;
342 			return status;
343 		}
344 	}
345 	status = nfs_filemap_write_and_wait_range(src->f_mapping,
346 			pos_src, pos_src + (loff_t)count - 1);
347 	if (status)
348 		return status;
349 
350 	status = nfs4_set_rw_stateid(&args->dst_stateid, dst_lock->open_context,
351 				     dst_lock, FMODE_WRITE);
352 	if (status) {
353 		if (status == -EAGAIN)
354 			status = -NFS4ERR_BAD_STATEID;
355 		return status;
356 	}
357 
358 	status = nfs_sync_inode(dst_inode);
359 	if (status)
360 		return status;
361 
362 	res->commit_res.verf = NULL;
363 	if (args->sync) {
364 		res->commit_res.verf =
365 			kzalloc(sizeof(struct nfs_writeverf), GFP_KERNEL);
366 		if (!res->commit_res.verf)
367 			return -ENOMEM;
368 	}
369 	set_bit(NFS_CLNT_SRC_SSC_COPY_STATE,
370 		&src_lock->open_context->state->flags);
371 	set_bit(NFS_CLNT_DST_SSC_COPY_STATE,
372 		&dst_lock->open_context->state->flags);
373 
374 	status = nfs4_call_sync(dst_server->client, dst_server, &msg,
375 				&args->seq_args, &res->seq_res, 0);
376 	trace_nfs4_copy(src_inode, dst_inode, args, res, nss, status);
377 	if (status == -ENOTSUPP)
378 		dst_server->caps &= ~NFS_CAP_COPY;
379 	if (status)
380 		goto out;
381 
382 	if (args->sync &&
383 		nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
384 				    &res->commit_res.verf->verifier)) {
385 		status = -EAGAIN;
386 		goto out;
387 	}
388 
389 	if (!res->synchronous) {
390 		status = handle_async_copy(res, dst_server, src_server, src,
391 				dst, &args->src_stateid, restart);
392 		if (status)
393 			goto out;
394 	}
395 
396 	if ((!res->synchronous || !args->sync) &&
397 			res->write_res.verifier.committed != NFS_FILE_SYNC) {
398 		status = process_copy_commit(dst, pos_dst, res);
399 		if (status)
400 			goto out;
401 	}
402 
403 	nfs42_copy_dest_done(dst_inode, pos_dst, res->write_res.count);
404 	nfs_invalidate_atime(src_inode);
405 	status = res->write_res.count;
406 out:
407 	if (args->sync)
408 		kfree(res->commit_res.verf);
409 	return status;
410 }
411 
412 ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
413 			struct file *dst, loff_t pos_dst, size_t count,
414 			struct nl4_server *nss,
415 			nfs4_stateid *cnr_stateid, bool sync)
416 {
417 	struct nfs_server *server = NFS_SERVER(file_inode(dst));
418 	struct nfs_lock_context *src_lock;
419 	struct nfs_lock_context *dst_lock;
420 	struct nfs42_copy_args args = {
421 		.src_fh		= NFS_FH(file_inode(src)),
422 		.src_pos	= pos_src,
423 		.dst_fh		= NFS_FH(file_inode(dst)),
424 		.dst_pos	= pos_dst,
425 		.count		= count,
426 		.sync		= sync,
427 	};
428 	struct nfs42_copy_res res;
429 	struct nfs4_exception src_exception = {
430 		.inode		= file_inode(src),
431 		.stateid	= &args.src_stateid,
432 	};
433 	struct nfs4_exception dst_exception = {
434 		.inode		= file_inode(dst),
435 		.stateid	= &args.dst_stateid,
436 	};
437 	ssize_t err, err2;
438 	bool restart = false;
439 
440 	src_lock = nfs_get_lock_context(nfs_file_open_context(src));
441 	if (IS_ERR(src_lock))
442 		return PTR_ERR(src_lock);
443 
444 	src_exception.state = src_lock->open_context->state;
445 
446 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst));
447 	if (IS_ERR(dst_lock)) {
448 		err = PTR_ERR(dst_lock);
449 		goto out_put_src_lock;
450 	}
451 
452 	dst_exception.state = dst_lock->open_context->state;
453 
454 	do {
455 		inode_lock(file_inode(dst));
456 		err = _nfs42_proc_copy(src, src_lock,
457 				dst, dst_lock,
458 				&args, &res,
459 				nss, cnr_stateid, &restart);
460 		inode_unlock(file_inode(dst));
461 
462 		if (err >= 0)
463 			break;
464 		if ((err == -ENOTSUPP ||
465 				err == -NFS4ERR_OFFLOAD_DENIED) &&
466 				nfs42_files_from_same_server(src, dst)) {
467 			err = -EOPNOTSUPP;
468 			break;
469 		} else if (err == -EAGAIN) {
470 			if (!restart) {
471 				dst_exception.retry = 1;
472 				continue;
473 			}
474 			break;
475 		} else if (err == -NFS4ERR_OFFLOAD_NO_REQS &&
476 				args.sync != res.synchronous) {
477 			args.sync = res.synchronous;
478 			dst_exception.retry = 1;
479 			continue;
480 		} else if ((err == -ESTALE ||
481 				err == -NFS4ERR_OFFLOAD_DENIED ||
482 				err == -ENOTSUPP) &&
483 				!nfs42_files_from_same_server(src, dst)) {
484 			nfs42_do_offload_cancel_async(src, &args.src_stateid);
485 			err = -EOPNOTSUPP;
486 			break;
487 		}
488 
489 		err2 = nfs4_handle_exception(server, err, &src_exception);
490 		err  = nfs4_handle_exception(server, err, &dst_exception);
491 		if (!err)
492 			err = err2;
493 	} while (src_exception.retry || dst_exception.retry);
494 
495 	nfs_put_lock_context(dst_lock);
496 out_put_src_lock:
497 	nfs_put_lock_context(src_lock);
498 	return err;
499 }
500 
501 struct nfs42_offloadcancel_data {
502 	struct nfs_server *seq_server;
503 	struct nfs42_offload_status_args args;
504 	struct nfs42_offload_status_res res;
505 };
506 
507 static void nfs42_offload_cancel_prepare(struct rpc_task *task, void *calldata)
508 {
509 	struct nfs42_offloadcancel_data *data = calldata;
510 
511 	nfs4_setup_sequence(data->seq_server->nfs_client,
512 				&data->args.osa_seq_args,
513 				&data->res.osr_seq_res, task);
514 }
515 
516 static void nfs42_offload_cancel_done(struct rpc_task *task, void *calldata)
517 {
518 	struct nfs42_offloadcancel_data *data = calldata;
519 
520 	trace_nfs4_offload_cancel(&data->args, task->tk_status);
521 	nfs41_sequence_done(task, &data->res.osr_seq_res);
522 	if (task->tk_status &&
523 		nfs4_async_handle_error(task, data->seq_server, NULL,
524 			NULL) == -EAGAIN)
525 		rpc_restart_call_prepare(task);
526 }
527 
528 static void nfs42_free_offloadcancel_data(void *data)
529 {
530 	kfree(data);
531 }
532 
533 static const struct rpc_call_ops nfs42_offload_cancel_ops = {
534 	.rpc_call_prepare = nfs42_offload_cancel_prepare,
535 	.rpc_call_done = nfs42_offload_cancel_done,
536 	.rpc_release = nfs42_free_offloadcancel_data,
537 };
538 
539 static int nfs42_do_offload_cancel_async(struct file *dst,
540 					 nfs4_stateid *stateid)
541 {
542 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
543 	struct nfs42_offloadcancel_data *data = NULL;
544 	struct nfs_open_context *ctx = nfs_file_open_context(dst);
545 	struct rpc_task *task;
546 	struct rpc_message msg = {
547 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OFFLOAD_CANCEL],
548 		.rpc_cred = ctx->cred,
549 	};
550 	struct rpc_task_setup task_setup_data = {
551 		.rpc_client = dst_server->client,
552 		.rpc_message = &msg,
553 		.callback_ops = &nfs42_offload_cancel_ops,
554 		.workqueue = nfsiod_workqueue,
555 		.flags = RPC_TASK_ASYNC,
556 	};
557 	int status;
558 
559 	if (!(dst_server->caps & NFS_CAP_OFFLOAD_CANCEL))
560 		return -EOPNOTSUPP;
561 
562 	data = kzalloc(sizeof(struct nfs42_offloadcancel_data), GFP_KERNEL);
563 	if (data == NULL)
564 		return -ENOMEM;
565 
566 	data->seq_server = dst_server;
567 	data->args.osa_src_fh = NFS_FH(file_inode(dst));
568 	memcpy(&data->args.osa_stateid, stateid,
569 		sizeof(data->args.osa_stateid));
570 	msg.rpc_argp = &data->args;
571 	msg.rpc_resp = &data->res;
572 	task_setup_data.callback_data = data;
573 	nfs4_init_sequence(&data->args.osa_seq_args, &data->res.osr_seq_res,
574 			   1, 0);
575 	task = rpc_run_task(&task_setup_data);
576 	if (IS_ERR(task))
577 		return PTR_ERR(task);
578 	status = rpc_wait_for_completion_task(task);
579 	if (status == -ENOTSUPP)
580 		dst_server->caps &= ~NFS_CAP_OFFLOAD_CANCEL;
581 	rpc_put_task(task);
582 	return status;
583 }
584 
585 static int _nfs42_proc_copy_notify(struct file *src, struct file *dst,
586 				   struct nfs42_copy_notify_args *args,
587 				   struct nfs42_copy_notify_res *res)
588 {
589 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
590 	struct rpc_message msg = {
591 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY_NOTIFY],
592 		.rpc_argp = args,
593 		.rpc_resp = res,
594 	};
595 	int status;
596 	struct nfs_open_context *ctx;
597 	struct nfs_lock_context *l_ctx;
598 
599 	ctx = get_nfs_open_context(nfs_file_open_context(src));
600 	l_ctx = nfs_get_lock_context(ctx);
601 	if (IS_ERR(l_ctx)) {
602 		status = PTR_ERR(l_ctx);
603 		goto out;
604 	}
605 
606 	status = nfs4_set_rw_stateid(&args->cna_src_stateid, ctx, l_ctx,
607 				     FMODE_READ);
608 	nfs_put_lock_context(l_ctx);
609 	if (status) {
610 		if (status == -EAGAIN)
611 			status = -NFS4ERR_BAD_STATEID;
612 		goto out;
613 	}
614 
615 	status = nfs4_call_sync(src_server->client, src_server, &msg,
616 				&args->cna_seq_args, &res->cnr_seq_res, 0);
617 	trace_nfs4_copy_notify(file_inode(src), args, res, status);
618 	if (status == -ENOTSUPP)
619 		src_server->caps &= ~NFS_CAP_COPY_NOTIFY;
620 
621 out:
622 	put_nfs_open_context(nfs_file_open_context(src));
623 	return status;
624 }
625 
626 int nfs42_proc_copy_notify(struct file *src, struct file *dst,
627 				struct nfs42_copy_notify_res *res)
628 {
629 	struct nfs_server *src_server = NFS_SERVER(file_inode(src));
630 	struct nfs42_copy_notify_args *args;
631 	struct nfs4_exception exception = {
632 		.inode = file_inode(src),
633 	};
634 	int status;
635 
636 	if (!(src_server->caps & NFS_CAP_COPY_NOTIFY))
637 		return -EOPNOTSUPP;
638 
639 	args = kzalloc(sizeof(struct nfs42_copy_notify_args), GFP_KERNEL);
640 	if (args == NULL)
641 		return -ENOMEM;
642 
643 	args->cna_src_fh  = NFS_FH(file_inode(src)),
644 	args->cna_dst.nl4_type = NL4_NETADDR;
645 	nfs42_set_netaddr(dst, &args->cna_dst.u.nl4_addr);
646 	exception.stateid = &args->cna_src_stateid;
647 
648 	do {
649 		status = _nfs42_proc_copy_notify(src, dst, args, res);
650 		if (status == -ENOTSUPP) {
651 			status = -EOPNOTSUPP;
652 			goto out;
653 		}
654 		status = nfs4_handle_exception(src_server, status, &exception);
655 	} while (exception.retry);
656 
657 out:
658 	kfree(args);
659 	return status;
660 }
661 
662 static loff_t _nfs42_proc_llseek(struct file *filep,
663 		struct nfs_lock_context *lock, loff_t offset, int whence)
664 {
665 	struct inode *inode = file_inode(filep);
666 	struct nfs42_seek_args args = {
667 		.sa_fh		= NFS_FH(inode),
668 		.sa_offset	= offset,
669 		.sa_what	= (whence == SEEK_HOLE) ?
670 					NFS4_CONTENT_HOLE : NFS4_CONTENT_DATA,
671 	};
672 	struct nfs42_seek_res res;
673 	struct rpc_message msg = {
674 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEEK],
675 		.rpc_argp = &args,
676 		.rpc_resp = &res,
677 	};
678 	struct nfs_server *server = NFS_SERVER(inode);
679 	int status;
680 
681 	if (!nfs_server_capable(inode, NFS_CAP_SEEK))
682 		return -ENOTSUPP;
683 
684 	status = nfs4_set_rw_stateid(&args.sa_stateid, lock->open_context,
685 			lock, FMODE_READ);
686 	if (status) {
687 		if (status == -EAGAIN)
688 			status = -NFS4ERR_BAD_STATEID;
689 		return status;
690 	}
691 
692 	status = nfs_filemap_write_and_wait_range(inode->i_mapping,
693 			offset, LLONG_MAX);
694 	if (status)
695 		return status;
696 
697 	status = nfs4_call_sync(server->client, server, &msg,
698 				&args.seq_args, &res.seq_res, 0);
699 	trace_nfs4_llseek(inode, &args, &res, status);
700 	if (status == -ENOTSUPP)
701 		server->caps &= ~NFS_CAP_SEEK;
702 	if (status)
703 		return status;
704 
705 	if (whence == SEEK_DATA && res.sr_eof)
706 		return -NFS4ERR_NXIO;
707 	else
708 		return vfs_setpos(filep, res.sr_offset, inode->i_sb->s_maxbytes);
709 }
710 
711 loff_t nfs42_proc_llseek(struct file *filep, loff_t offset, int whence)
712 {
713 	struct nfs_server *server = NFS_SERVER(file_inode(filep));
714 	struct nfs4_exception exception = { };
715 	struct nfs_lock_context *lock;
716 	loff_t err;
717 
718 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
719 	if (IS_ERR(lock))
720 		return PTR_ERR(lock);
721 
722 	exception.inode = file_inode(filep);
723 	exception.state = lock->open_context->state;
724 
725 	do {
726 		err = _nfs42_proc_llseek(filep, lock, offset, whence);
727 		if (err >= 0)
728 			break;
729 		if (err == -ENOTSUPP) {
730 			err = -EOPNOTSUPP;
731 			break;
732 		}
733 		err = nfs4_handle_exception(server, err, &exception);
734 	} while (exception.retry);
735 
736 	nfs_put_lock_context(lock);
737 	return err;
738 }
739 
740 
741 static void
742 nfs42_layoutstat_prepare(struct rpc_task *task, void *calldata)
743 {
744 	struct nfs42_layoutstat_data *data = calldata;
745 	struct inode *inode = data->inode;
746 	struct nfs_server *server = NFS_SERVER(inode);
747 	struct pnfs_layout_hdr *lo;
748 
749 	spin_lock(&inode->i_lock);
750 	lo = NFS_I(inode)->layout;
751 	if (!pnfs_layout_is_valid(lo)) {
752 		spin_unlock(&inode->i_lock);
753 		rpc_exit(task, 0);
754 		return;
755 	}
756 	nfs4_stateid_copy(&data->args.stateid, &lo->plh_stateid);
757 	spin_unlock(&inode->i_lock);
758 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
759 			    &data->res.seq_res, task);
760 }
761 
762 static void
763 nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
764 {
765 	struct nfs42_layoutstat_data *data = calldata;
766 	struct inode *inode = data->inode;
767 	struct pnfs_layout_hdr *lo;
768 
769 	if (!nfs4_sequence_done(task, &data->res.seq_res))
770 		return;
771 
772 	switch (task->tk_status) {
773 	case 0:
774 		return;
775 	case -NFS4ERR_BADHANDLE:
776 	case -ESTALE:
777 		pnfs_destroy_layout(NFS_I(inode));
778 		break;
779 	case -NFS4ERR_EXPIRED:
780 	case -NFS4ERR_ADMIN_REVOKED:
781 	case -NFS4ERR_DELEG_REVOKED:
782 	case -NFS4ERR_STALE_STATEID:
783 	case -NFS4ERR_BAD_STATEID:
784 		spin_lock(&inode->i_lock);
785 		lo = NFS_I(inode)->layout;
786 		if (pnfs_layout_is_valid(lo) &&
787 		    nfs4_stateid_match(&data->args.stateid,
788 					     &lo->plh_stateid)) {
789 			LIST_HEAD(head);
790 
791 			/*
792 			 * Mark the bad layout state as invalid, then retry
793 			 * with the current stateid.
794 			 */
795 			pnfs_mark_layout_stateid_invalid(lo, &head);
796 			spin_unlock(&inode->i_lock);
797 			pnfs_free_lseg_list(&head);
798 			nfs_commit_inode(inode, 0);
799 		} else
800 			spin_unlock(&inode->i_lock);
801 		break;
802 	case -NFS4ERR_OLD_STATEID:
803 		spin_lock(&inode->i_lock);
804 		lo = NFS_I(inode)->layout;
805 		if (pnfs_layout_is_valid(lo) &&
806 		    nfs4_stateid_match_other(&data->args.stateid,
807 					&lo->plh_stateid)) {
808 			/* Do we need to delay before resending? */
809 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
810 						&data->args.stateid))
811 				rpc_delay(task, HZ);
812 			rpc_restart_call_prepare(task);
813 		}
814 		spin_unlock(&inode->i_lock);
815 		break;
816 	case -ENOTSUPP:
817 	case -EOPNOTSUPP:
818 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTSTATS;
819 	}
820 
821 	trace_nfs4_layoutstats(inode, &data->args.stateid, task->tk_status);
822 }
823 
824 static void
825 nfs42_layoutstat_release(void *calldata)
826 {
827 	struct nfs42_layoutstat_data *data = calldata;
828 	struct nfs42_layoutstat_devinfo *devinfo = data->args.devinfo;
829 	int i;
830 
831 	for (i = 0; i < data->args.num_dev; i++) {
832 		if (devinfo[i].ld_private.ops && devinfo[i].ld_private.ops->free)
833 			devinfo[i].ld_private.ops->free(&devinfo[i].ld_private);
834 	}
835 
836 	pnfs_put_layout_hdr(NFS_I(data->args.inode)->layout);
837 	smp_mb__before_atomic();
838 	clear_bit(NFS_INO_LAYOUTSTATS, &NFS_I(data->args.inode)->flags);
839 	smp_mb__after_atomic();
840 	nfs_iput_and_deactive(data->inode);
841 	kfree(data->args.devinfo);
842 	kfree(data);
843 }
844 
845 static const struct rpc_call_ops nfs42_layoutstat_ops = {
846 	.rpc_call_prepare = nfs42_layoutstat_prepare,
847 	.rpc_call_done = nfs42_layoutstat_done,
848 	.rpc_release = nfs42_layoutstat_release,
849 };
850 
851 int nfs42_proc_layoutstats_generic(struct nfs_server *server,
852 				   struct nfs42_layoutstat_data *data)
853 {
854 	struct rpc_message msg = {
855 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTSTATS],
856 		.rpc_argp = &data->args,
857 		.rpc_resp = &data->res,
858 	};
859 	struct rpc_task_setup task_setup = {
860 		.rpc_client = server->client,
861 		.rpc_message = &msg,
862 		.callback_ops = &nfs42_layoutstat_ops,
863 		.callback_data = data,
864 		.flags = RPC_TASK_ASYNC,
865 	};
866 	struct rpc_task *task;
867 
868 	data->inode = nfs_igrab_and_active(data->args.inode);
869 	if (!data->inode) {
870 		nfs42_layoutstat_release(data);
871 		return -EAGAIN;
872 	}
873 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
874 	task = rpc_run_task(&task_setup);
875 	if (IS_ERR(task))
876 		return PTR_ERR(task);
877 	rpc_put_task(task);
878 	return 0;
879 }
880 
881 static struct nfs42_layouterror_data *
882 nfs42_alloc_layouterror_data(struct pnfs_layout_segment *lseg, gfp_t gfp_flags)
883 {
884 	struct nfs42_layouterror_data *data;
885 	struct inode *inode = lseg->pls_layout->plh_inode;
886 
887 	data = kzalloc(sizeof(*data), gfp_flags);
888 	if (data) {
889 		data->args.inode = data->inode = nfs_igrab_and_active(inode);
890 		if (data->inode) {
891 			data->lseg = pnfs_get_lseg(lseg);
892 			if (data->lseg)
893 				return data;
894 			nfs_iput_and_deactive(data->inode);
895 		}
896 		kfree(data);
897 	}
898 	return NULL;
899 }
900 
901 static void
902 nfs42_free_layouterror_data(struct nfs42_layouterror_data *data)
903 {
904 	pnfs_put_lseg(data->lseg);
905 	nfs_iput_and_deactive(data->inode);
906 	kfree(data);
907 }
908 
909 static void
910 nfs42_layouterror_prepare(struct rpc_task *task, void *calldata)
911 {
912 	struct nfs42_layouterror_data *data = calldata;
913 	struct inode *inode = data->inode;
914 	struct nfs_server *server = NFS_SERVER(inode);
915 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
916 	unsigned i;
917 
918 	spin_lock(&inode->i_lock);
919 	if (!pnfs_layout_is_valid(lo)) {
920 		spin_unlock(&inode->i_lock);
921 		rpc_exit(task, 0);
922 		return;
923 	}
924 	for (i = 0; i < data->args.num_errors; i++)
925 		nfs4_stateid_copy(&data->args.errors[i].stateid,
926 				&lo->plh_stateid);
927 	spin_unlock(&inode->i_lock);
928 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
929 			    &data->res.seq_res, task);
930 }
931 
932 static void
933 nfs42_layouterror_done(struct rpc_task *task, void *calldata)
934 {
935 	struct nfs42_layouterror_data *data = calldata;
936 	struct inode *inode = data->inode;
937 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
938 
939 	if (!nfs4_sequence_done(task, &data->res.seq_res))
940 		return;
941 
942 	switch (task->tk_status) {
943 	case 0:
944 		return;
945 	case -NFS4ERR_BADHANDLE:
946 	case -ESTALE:
947 		pnfs_destroy_layout(NFS_I(inode));
948 		break;
949 	case -NFS4ERR_EXPIRED:
950 	case -NFS4ERR_ADMIN_REVOKED:
951 	case -NFS4ERR_DELEG_REVOKED:
952 	case -NFS4ERR_STALE_STATEID:
953 	case -NFS4ERR_BAD_STATEID:
954 		spin_lock(&inode->i_lock);
955 		if (pnfs_layout_is_valid(lo) &&
956 		    nfs4_stateid_match(&data->args.errors[0].stateid,
957 					     &lo->plh_stateid)) {
958 			LIST_HEAD(head);
959 
960 			/*
961 			 * Mark the bad layout state as invalid, then retry
962 			 * with the current stateid.
963 			 */
964 			pnfs_mark_layout_stateid_invalid(lo, &head);
965 			spin_unlock(&inode->i_lock);
966 			pnfs_free_lseg_list(&head);
967 			nfs_commit_inode(inode, 0);
968 		} else
969 			spin_unlock(&inode->i_lock);
970 		break;
971 	case -NFS4ERR_OLD_STATEID:
972 		spin_lock(&inode->i_lock);
973 		if (pnfs_layout_is_valid(lo) &&
974 		    nfs4_stateid_match_other(&data->args.errors[0].stateid,
975 					&lo->plh_stateid)) {
976 			/* Do we need to delay before resending? */
977 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
978 						&data->args.errors[0].stateid))
979 				rpc_delay(task, HZ);
980 			rpc_restart_call_prepare(task);
981 		}
982 		spin_unlock(&inode->i_lock);
983 		break;
984 	case -ENOTSUPP:
985 	case -EOPNOTSUPP:
986 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTERROR;
987 	}
988 
989 	trace_nfs4_layouterror(inode, &data->args.errors[0].stateid,
990 			       task->tk_status);
991 }
992 
993 static void
994 nfs42_layouterror_release(void *calldata)
995 {
996 	struct nfs42_layouterror_data *data = calldata;
997 
998 	nfs42_free_layouterror_data(data);
999 }
1000 
1001 static const struct rpc_call_ops nfs42_layouterror_ops = {
1002 	.rpc_call_prepare = nfs42_layouterror_prepare,
1003 	.rpc_call_done = nfs42_layouterror_done,
1004 	.rpc_release = nfs42_layouterror_release,
1005 };
1006 
1007 int nfs42_proc_layouterror(struct pnfs_layout_segment *lseg,
1008 		const struct nfs42_layout_error *errors, size_t n)
1009 {
1010 	struct inode *inode = lseg->pls_layout->plh_inode;
1011 	struct nfs42_layouterror_data *data;
1012 	struct rpc_task *task;
1013 	struct rpc_message msg = {
1014 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTERROR],
1015 	};
1016 	struct rpc_task_setup task_setup = {
1017 		.rpc_message = &msg,
1018 		.callback_ops = &nfs42_layouterror_ops,
1019 		.flags = RPC_TASK_ASYNC,
1020 	};
1021 	unsigned int i;
1022 
1023 	if (!nfs_server_capable(inode, NFS_CAP_LAYOUTERROR))
1024 		return -EOPNOTSUPP;
1025 	if (n > NFS42_LAYOUTERROR_MAX)
1026 		return -EINVAL;
1027 	data = nfs42_alloc_layouterror_data(lseg, nfs_io_gfp_mask());
1028 	if (!data)
1029 		return -ENOMEM;
1030 	for (i = 0; i < n; i++) {
1031 		data->args.errors[i] = errors[i];
1032 		data->args.num_errors++;
1033 		data->res.num_errors++;
1034 	}
1035 	msg.rpc_argp = &data->args;
1036 	msg.rpc_resp = &data->res;
1037 	task_setup.callback_data = data;
1038 	task_setup.rpc_client = NFS_SERVER(inode)->client;
1039 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
1040 	task = rpc_run_task(&task_setup);
1041 	if (IS_ERR(task))
1042 		return PTR_ERR(task);
1043 	rpc_put_task(task);
1044 	return 0;
1045 }
1046 EXPORT_SYMBOL_GPL(nfs42_proc_layouterror);
1047 
1048 static int _nfs42_proc_clone(struct rpc_message *msg, struct file *src_f,
1049 		struct file *dst_f, struct nfs_lock_context *src_lock,
1050 		struct nfs_lock_context *dst_lock, loff_t src_offset,
1051 		loff_t dst_offset, loff_t count)
1052 {
1053 	struct inode *src_inode = file_inode(src_f);
1054 	struct inode *dst_inode = file_inode(dst_f);
1055 	struct nfs_server *server = NFS_SERVER(dst_inode);
1056 	__u32 dst_bitmask[NFS_BITMASK_SZ];
1057 	struct nfs42_clone_args args = {
1058 		.src_fh = NFS_FH(src_inode),
1059 		.dst_fh = NFS_FH(dst_inode),
1060 		.src_offset = src_offset,
1061 		.dst_offset = dst_offset,
1062 		.count = count,
1063 		.dst_bitmask = dst_bitmask,
1064 	};
1065 	struct nfs42_clone_res res = {
1066 		.server	= server,
1067 	};
1068 	int status;
1069 
1070 	msg->rpc_argp = &args;
1071 	msg->rpc_resp = &res;
1072 
1073 	status = nfs4_set_rw_stateid(&args.src_stateid, src_lock->open_context,
1074 			src_lock, FMODE_READ);
1075 	if (status) {
1076 		if (status == -EAGAIN)
1077 			status = -NFS4ERR_BAD_STATEID;
1078 		return status;
1079 	}
1080 	status = nfs4_set_rw_stateid(&args.dst_stateid, dst_lock->open_context,
1081 			dst_lock, FMODE_WRITE);
1082 	if (status) {
1083 		if (status == -EAGAIN)
1084 			status = -NFS4ERR_BAD_STATEID;
1085 		return status;
1086 	}
1087 
1088 	res.dst_fattr = nfs_alloc_fattr();
1089 	if (!res.dst_fattr)
1090 		return -ENOMEM;
1091 
1092 	nfs4_bitmask_set(dst_bitmask, server->cache_consistency_bitmask,
1093 			 dst_inode, NFS_INO_INVALID_BLOCKS);
1094 
1095 	status = nfs4_call_sync(server->client, server, msg,
1096 				&args.seq_args, &res.seq_res, 0);
1097 	trace_nfs4_clone(src_inode, dst_inode, &args, status);
1098 	if (status == 0) {
1099 		/* a zero-length count means clone to EOF in src */
1100 		if (count == 0 && res.dst_fattr->valid & NFS_ATTR_FATTR_SIZE)
1101 			count = nfs_size_to_loff_t(res.dst_fattr->size) - dst_offset;
1102 		nfs42_copy_dest_done(dst_inode, dst_offset, count);
1103 		status = nfs_post_op_update_inode(dst_inode, res.dst_fattr);
1104 	}
1105 
1106 	kfree(res.dst_fattr);
1107 	return status;
1108 }
1109 
1110 int nfs42_proc_clone(struct file *src_f, struct file *dst_f,
1111 		     loff_t src_offset, loff_t dst_offset, loff_t count)
1112 {
1113 	struct rpc_message msg = {
1114 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLONE],
1115 	};
1116 	struct inode *inode = file_inode(src_f);
1117 	struct nfs_server *server = NFS_SERVER(file_inode(src_f));
1118 	struct nfs_lock_context *src_lock;
1119 	struct nfs_lock_context *dst_lock;
1120 	struct nfs4_exception src_exception = { };
1121 	struct nfs4_exception dst_exception = { };
1122 	int err, err2;
1123 
1124 	if (!nfs_server_capable(inode, NFS_CAP_CLONE))
1125 		return -EOPNOTSUPP;
1126 
1127 	src_lock = nfs_get_lock_context(nfs_file_open_context(src_f));
1128 	if (IS_ERR(src_lock))
1129 		return PTR_ERR(src_lock);
1130 
1131 	src_exception.inode = file_inode(src_f);
1132 	src_exception.state = src_lock->open_context->state;
1133 
1134 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst_f));
1135 	if (IS_ERR(dst_lock)) {
1136 		err = PTR_ERR(dst_lock);
1137 		goto out_put_src_lock;
1138 	}
1139 
1140 	dst_exception.inode = file_inode(dst_f);
1141 	dst_exception.state = dst_lock->open_context->state;
1142 
1143 	do {
1144 		err = _nfs42_proc_clone(&msg, src_f, dst_f, src_lock, dst_lock,
1145 					src_offset, dst_offset, count);
1146 		if (err == -ENOTSUPP || err == -EOPNOTSUPP) {
1147 			NFS_SERVER(inode)->caps &= ~NFS_CAP_CLONE;
1148 			err = -EOPNOTSUPP;
1149 			break;
1150 		}
1151 
1152 		err2 = nfs4_handle_exception(server, err, &src_exception);
1153 		err = nfs4_handle_exception(server, err, &dst_exception);
1154 		if (!err)
1155 			err = err2;
1156 	} while (src_exception.retry || dst_exception.retry);
1157 
1158 	nfs_put_lock_context(dst_lock);
1159 out_put_src_lock:
1160 	nfs_put_lock_context(src_lock);
1161 	return err;
1162 }
1163 
1164 #define NFS4XATTR_MAXPAGES DIV_ROUND_UP(XATTR_SIZE_MAX, PAGE_SIZE)
1165 
1166 static int _nfs42_proc_removexattr(struct inode *inode, const char *name)
1167 {
1168 	struct nfs_server *server = NFS_SERVER(inode);
1169 	struct nfs42_removexattrargs args = {
1170 		.fh = NFS_FH(inode),
1171 		.xattr_name = name,
1172 	};
1173 	struct nfs42_removexattrres res;
1174 	struct rpc_message msg = {
1175 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_REMOVEXATTR],
1176 		.rpc_argp = &args,
1177 		.rpc_resp = &res,
1178 	};
1179 	int ret;
1180 	unsigned long timestamp = jiffies;
1181 
1182 	ret = nfs4_call_sync(server->client, server, &msg, &args.seq_args,
1183 	    &res.seq_res, 1);
1184 	trace_nfs4_removexattr(inode, name, ret);
1185 	if (!ret)
1186 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1187 
1188 	return ret;
1189 }
1190 
1191 static int _nfs42_proc_setxattr(struct inode *inode, const char *name,
1192 				const void *buf, size_t buflen, int flags)
1193 {
1194 	struct nfs_server *server = NFS_SERVER(inode);
1195 	__u32 bitmask[NFS_BITMASK_SZ];
1196 	struct page *pages[NFS4XATTR_MAXPAGES];
1197 	struct nfs42_setxattrargs arg = {
1198 		.fh		= NFS_FH(inode),
1199 		.bitmask	= bitmask,
1200 		.xattr_pages	= pages,
1201 		.xattr_len	= buflen,
1202 		.xattr_name	= name,
1203 		.xattr_flags	= flags,
1204 	};
1205 	struct nfs42_setxattrres res = {
1206 		.server		= server,
1207 	};
1208 	struct rpc_message msg = {
1209 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_SETXATTR],
1210 		.rpc_argp	= &arg,
1211 		.rpc_resp	= &res,
1212 	};
1213 	int ret, np;
1214 	unsigned long timestamp = jiffies;
1215 
1216 	if (buflen > server->sxasize)
1217 		return -ERANGE;
1218 
1219 	res.fattr = nfs_alloc_fattr();
1220 	if (!res.fattr)
1221 		return -ENOMEM;
1222 
1223 	if (buflen > 0) {
1224 		np = nfs4_buf_to_pages_noslab(buf, buflen, arg.xattr_pages);
1225 		if (np < 0) {
1226 			ret = np;
1227 			goto out;
1228 		}
1229 	} else
1230 		np = 0;
1231 
1232 	nfs4_bitmask_set(bitmask, server->cache_consistency_bitmask,
1233 			 inode, NFS_INO_INVALID_CHANGE);
1234 
1235 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1236 	    &res.seq_res, 1);
1237 	trace_nfs4_setxattr(inode, name, ret);
1238 
1239 	for (; np > 0; np--)
1240 		put_page(pages[np - 1]);
1241 
1242 	if (!ret) {
1243 		nfs4_update_changeattr(inode, &res.cinfo, timestamp, 0);
1244 		ret = nfs_post_op_update_inode(inode, res.fattr);
1245 	}
1246 
1247 out:
1248 	kfree(res.fattr);
1249 	return ret;
1250 }
1251 
1252 static ssize_t _nfs42_proc_getxattr(struct inode *inode, const char *name,
1253 				void *buf, size_t buflen, struct page **pages,
1254 				size_t plen)
1255 {
1256 	struct nfs_server *server = NFS_SERVER(inode);
1257 	struct nfs42_getxattrargs arg = {
1258 		.fh		= NFS_FH(inode),
1259 		.xattr_name	= name,
1260 	};
1261 	struct nfs42_getxattrres res;
1262 	struct rpc_message msg = {
1263 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_GETXATTR],
1264 		.rpc_argp	= &arg,
1265 		.rpc_resp	= &res,
1266 	};
1267 	ssize_t ret;
1268 
1269 	arg.xattr_len = plen;
1270 	arg.xattr_pages = pages;
1271 
1272 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1273 	    &res.seq_res, 0);
1274 	trace_nfs4_getxattr(inode, name, ret);
1275 	if (ret < 0)
1276 		return ret;
1277 
1278 	/*
1279 	 * Normally, the caching is done one layer up, but for successful
1280 	 * RPCS, always cache the result here, even if the caller was
1281 	 * just querying the length, or if the reply was too big for
1282 	 * the caller. This avoids a second RPC in the case of the
1283 	 * common query-alloc-retrieve cycle for xattrs.
1284 	 *
1285 	 * Note that xattr_len is always capped to XATTR_SIZE_MAX.
1286 	 */
1287 
1288 	nfs4_xattr_cache_add(inode, name, NULL, pages, res.xattr_len);
1289 
1290 	if (buflen) {
1291 		if (res.xattr_len > buflen)
1292 			return -ERANGE;
1293 		_copy_from_pages(buf, pages, 0, res.xattr_len);
1294 	}
1295 
1296 	return res.xattr_len;
1297 }
1298 
1299 static ssize_t _nfs42_proc_listxattrs(struct inode *inode, void *buf,
1300 				 size_t buflen, u64 *cookiep, bool *eofp)
1301 {
1302 	struct nfs_server *server = NFS_SERVER(inode);
1303 	struct page **pages;
1304 	struct nfs42_listxattrsargs arg = {
1305 		.fh		= NFS_FH(inode),
1306 		.cookie		= *cookiep,
1307 	};
1308 	struct nfs42_listxattrsres res = {
1309 		.eof = false,
1310 		.xattr_buf = buf,
1311 		.xattr_len = buflen,
1312 	};
1313 	struct rpc_message msg = {
1314 		.rpc_proc	= &nfs4_procedures[NFSPROC4_CLNT_LISTXATTRS],
1315 		.rpc_argp	= &arg,
1316 		.rpc_resp	= &res,
1317 	};
1318 	u32 xdrlen;
1319 	int ret, np, i;
1320 
1321 
1322 	ret = -ENOMEM;
1323 	res.scratch = alloc_page(GFP_KERNEL);
1324 	if (!res.scratch)
1325 		goto out;
1326 
1327 	xdrlen = nfs42_listxattr_xdrsize(buflen);
1328 	if (xdrlen > server->lxasize)
1329 		xdrlen = server->lxasize;
1330 	np = xdrlen / PAGE_SIZE + 1;
1331 
1332 	pages = kcalloc(np, sizeof(struct page *), GFP_KERNEL);
1333 	if (!pages)
1334 		goto out_free_scratch;
1335 	for (i = 0; i < np; i++) {
1336 		pages[i] = alloc_page(GFP_KERNEL);
1337 		if (!pages[i])
1338 			goto out_free_pages;
1339 	}
1340 
1341 	arg.xattr_pages = pages;
1342 	arg.count = xdrlen;
1343 
1344 	ret = nfs4_call_sync(server->client, server, &msg, &arg.seq_args,
1345 	    &res.seq_res, 0);
1346 	trace_nfs4_listxattr(inode, ret);
1347 
1348 	if (ret >= 0) {
1349 		ret = res.copied;
1350 		*cookiep = res.cookie;
1351 		*eofp = res.eof;
1352 	}
1353 
1354 out_free_pages:
1355 	while (--np >= 0) {
1356 		if (pages[np])
1357 			__free_page(pages[np]);
1358 	}
1359 	kfree(pages);
1360 out_free_scratch:
1361 	__free_page(res.scratch);
1362 out:
1363 	return ret;
1364 
1365 }
1366 
1367 ssize_t nfs42_proc_getxattr(struct inode *inode, const char *name,
1368 			      void *buf, size_t buflen)
1369 {
1370 	struct nfs4_exception exception = { };
1371 	ssize_t err, np, i;
1372 	struct page **pages;
1373 
1374 	np = nfs_page_array_len(0, buflen ?: XATTR_SIZE_MAX);
1375 	pages = kmalloc_array(np, sizeof(*pages), GFP_KERNEL);
1376 	if (!pages)
1377 		return -ENOMEM;
1378 
1379 	for (i = 0; i < np; i++) {
1380 		pages[i] = alloc_page(GFP_KERNEL);
1381 		if (!pages[i]) {
1382 			err = -ENOMEM;
1383 			goto out;
1384 		}
1385 	}
1386 
1387 	/*
1388 	 * The GETXATTR op has no length field in the call, and the
1389 	 * xattr data is at the end of the reply.
1390 	 *
1391 	 * There is no downside in using the page-aligned length. It will
1392 	 * allow receiving and caching xattrs that are too large for the
1393 	 * caller but still fit in the page-rounded value.
1394 	 */
1395 	do {
1396 		err = _nfs42_proc_getxattr(inode, name, buf, buflen,
1397 			pages, np * PAGE_SIZE);
1398 		if (err >= 0)
1399 			break;
1400 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1401 				&exception);
1402 	} while (exception.retry);
1403 
1404 out:
1405 	while (--i >= 0)
1406 		__free_page(pages[i]);
1407 	kfree(pages);
1408 
1409 	return err;
1410 }
1411 
1412 int nfs42_proc_setxattr(struct inode *inode, const char *name,
1413 			      const void *buf, size_t buflen, int flags)
1414 {
1415 	struct nfs4_exception exception = { };
1416 	int err;
1417 
1418 	do {
1419 		err = _nfs42_proc_setxattr(inode, name, buf, buflen, flags);
1420 		if (!err)
1421 			break;
1422 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1423 				&exception);
1424 	} while (exception.retry);
1425 
1426 	return err;
1427 }
1428 
1429 ssize_t nfs42_proc_listxattrs(struct inode *inode, void *buf,
1430 			      size_t buflen, u64 *cookiep, bool *eofp)
1431 {
1432 	struct nfs4_exception exception = { };
1433 	ssize_t err;
1434 
1435 	do {
1436 		err = _nfs42_proc_listxattrs(inode, buf, buflen,
1437 		    cookiep, eofp);
1438 		if (err >= 0)
1439 			break;
1440 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1441 				&exception);
1442 	} while (exception.retry);
1443 
1444 	return err;
1445 }
1446 
1447 int nfs42_proc_removexattr(struct inode *inode, const char *name)
1448 {
1449 	struct nfs4_exception exception = { };
1450 	int err;
1451 
1452 	do {
1453 		err = _nfs42_proc_removexattr(inode, name);
1454 		if (!err)
1455 			break;
1456 		err = nfs4_handle_exception(NFS_SERVER(inode), err,
1457 				&exception);
1458 	} while (exception.retry);
1459 
1460 	return err;
1461 }
1462