xref: /openbmc/linux/fs/nfs/nfs42proc.c (revision ae213c44)
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/sched.h>
7 #include <linux/nfs.h>
8 #include <linux/nfs3.h>
9 #include <linux/nfs4.h>
10 #include <linux/nfs_xdr.h>
11 #include <linux/nfs_fs.h>
12 #include "nfs4_fs.h"
13 #include "nfs42.h"
14 #include "iostat.h"
15 #include "pnfs.h"
16 #include "nfs4session.h"
17 #include "internal.h"
18 
19 #define NFSDBG_FACILITY NFSDBG_PROC
20 static int nfs42_do_offload_cancel_async(struct file *dst, nfs4_stateid *std);
21 
22 static int _nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep,
23 		struct nfs_lock_context *lock, loff_t offset, loff_t len)
24 {
25 	struct inode *inode = file_inode(filep);
26 	struct nfs_server *server = NFS_SERVER(inode);
27 	struct nfs42_falloc_args args = {
28 		.falloc_fh	= NFS_FH(inode),
29 		.falloc_offset	= offset,
30 		.falloc_length	= len,
31 		.falloc_bitmask	= server->cache_consistency_bitmask,
32 	};
33 	struct nfs42_falloc_res res = {
34 		.falloc_server	= server,
35 	};
36 	int status;
37 
38 	msg->rpc_argp = &args;
39 	msg->rpc_resp = &res;
40 
41 	status = nfs4_set_rw_stateid(&args.falloc_stateid, lock->open_context,
42 			lock, FMODE_WRITE);
43 	if (status)
44 		return status;
45 
46 	res.falloc_fattr = nfs_alloc_fattr();
47 	if (!res.falloc_fattr)
48 		return -ENOMEM;
49 
50 	status = nfs4_call_sync(server->client, server, msg,
51 				&args.seq_args, &res.seq_res, 0);
52 	if (status == 0)
53 		status = nfs_post_op_update_inode(inode, res.falloc_fattr);
54 
55 	kfree(res.falloc_fattr);
56 	return status;
57 }
58 
59 static int nfs42_proc_fallocate(struct rpc_message *msg, struct file *filep,
60 				loff_t offset, loff_t len)
61 {
62 	struct nfs_server *server = NFS_SERVER(file_inode(filep));
63 	struct nfs4_exception exception = { };
64 	struct nfs_lock_context *lock;
65 	int err;
66 
67 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
68 	if (IS_ERR(lock))
69 		return PTR_ERR(lock);
70 
71 	exception.inode = file_inode(filep);
72 	exception.state = lock->open_context->state;
73 
74 	do {
75 		err = _nfs42_proc_fallocate(msg, filep, lock, offset, len);
76 		if (err == -ENOTSUPP) {
77 			err = -EOPNOTSUPP;
78 			break;
79 		}
80 		err = nfs4_handle_exception(server, err, &exception);
81 	} while (exception.retry);
82 
83 	nfs_put_lock_context(lock);
84 	return err;
85 }
86 
87 int nfs42_proc_allocate(struct file *filep, loff_t offset, loff_t len)
88 {
89 	struct rpc_message msg = {
90 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_ALLOCATE],
91 	};
92 	struct inode *inode = file_inode(filep);
93 	int err;
94 
95 	if (!nfs_server_capable(inode, NFS_CAP_ALLOCATE))
96 		return -EOPNOTSUPP;
97 
98 	inode_lock(inode);
99 
100 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
101 	if (err == -EOPNOTSUPP)
102 		NFS_SERVER(inode)->caps &= ~NFS_CAP_ALLOCATE;
103 
104 	inode_unlock(inode);
105 	return err;
106 }
107 
108 int nfs42_proc_deallocate(struct file *filep, loff_t offset, loff_t len)
109 {
110 	struct rpc_message msg = {
111 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_DEALLOCATE],
112 	};
113 	struct inode *inode = file_inode(filep);
114 	int err;
115 
116 	if (!nfs_server_capable(inode, NFS_CAP_DEALLOCATE))
117 		return -EOPNOTSUPP;
118 
119 	inode_lock(inode);
120 	err = nfs_sync_inode(inode);
121 	if (err)
122 		goto out_unlock;
123 
124 	err = nfs42_proc_fallocate(&msg, filep, offset, len);
125 	if (err == 0)
126 		truncate_pagecache_range(inode, offset, (offset + len) -1);
127 	if (err == -EOPNOTSUPP)
128 		NFS_SERVER(inode)->caps &= ~NFS_CAP_DEALLOCATE;
129 out_unlock:
130 	inode_unlock(inode);
131 	return err;
132 }
133 
134 static int handle_async_copy(struct nfs42_copy_res *res,
135 			     struct nfs_server *server,
136 			     struct file *src,
137 			     struct file *dst,
138 			     nfs4_stateid *src_stateid)
139 {
140 	struct nfs4_copy_state *copy, *tmp_copy;
141 	int status = NFS4_OK;
142 	bool found_pending = false;
143 	struct nfs_open_context *ctx = nfs_file_open_context(dst);
144 
145 	copy = kzalloc(sizeof(struct nfs4_copy_state), GFP_NOFS);
146 	if (!copy)
147 		return -ENOMEM;
148 
149 	spin_lock(&server->nfs_client->cl_lock);
150 	list_for_each_entry(tmp_copy, &server->nfs_client->pending_cb_stateids,
151 				copies) {
152 		if (memcmp(&res->write_res.stateid, &tmp_copy->stateid,
153 				NFS4_STATEID_SIZE))
154 			continue;
155 		found_pending = true;
156 		list_del(&tmp_copy->copies);
157 		break;
158 	}
159 	if (found_pending) {
160 		spin_unlock(&server->nfs_client->cl_lock);
161 		kfree(copy);
162 		copy = tmp_copy;
163 		goto out;
164 	}
165 
166 	memcpy(&copy->stateid, &res->write_res.stateid, NFS4_STATEID_SIZE);
167 	init_completion(&copy->completion);
168 	copy->parent_state = ctx->state;
169 
170 	list_add_tail(&copy->copies, &server->ss_copies);
171 	spin_unlock(&server->nfs_client->cl_lock);
172 
173 	status = wait_for_completion_interruptible(&copy->completion);
174 	spin_lock(&server->nfs_client->cl_lock);
175 	list_del_init(&copy->copies);
176 	spin_unlock(&server->nfs_client->cl_lock);
177 	if (status == -ERESTARTSYS) {
178 		goto out_cancel;
179 	} else if (copy->flags) {
180 		status = -EAGAIN;
181 		goto out_cancel;
182 	}
183 out:
184 	res->write_res.count = copy->count;
185 	memcpy(&res->write_res.verifier, &copy->verf, sizeof(copy->verf));
186 	status = -copy->error;
187 
188 	kfree(copy);
189 	return status;
190 out_cancel:
191 	nfs42_do_offload_cancel_async(dst, &copy->stateid);
192 	kfree(copy);
193 	return status;
194 }
195 
196 static int process_copy_commit(struct file *dst, loff_t pos_dst,
197 			       struct nfs42_copy_res *res)
198 {
199 	struct nfs_commitres cres;
200 	int status = -ENOMEM;
201 
202 	cres.verf = kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS);
203 	if (!cres.verf)
204 		goto out;
205 
206 	status = nfs4_proc_commit(dst, pos_dst, res->write_res.count, &cres);
207 	if (status)
208 		goto out_free;
209 	if (nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
210 				    &cres.verf->verifier)) {
211 		dprintk("commit verf differs from copy verf\n");
212 		status = -EAGAIN;
213 	}
214 out_free:
215 	kfree(cres.verf);
216 out:
217 	return status;
218 }
219 
220 static ssize_t _nfs42_proc_copy(struct file *src,
221 				struct nfs_lock_context *src_lock,
222 				struct file *dst,
223 				struct nfs_lock_context *dst_lock,
224 				struct nfs42_copy_args *args,
225 				struct nfs42_copy_res *res)
226 {
227 	struct rpc_message msg = {
228 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_COPY],
229 		.rpc_argp = args,
230 		.rpc_resp = res,
231 	};
232 	struct inode *dst_inode = file_inode(dst);
233 	struct nfs_server *server = NFS_SERVER(dst_inode);
234 	loff_t pos_src = args->src_pos;
235 	loff_t pos_dst = args->dst_pos;
236 	size_t count = args->count;
237 	ssize_t status;
238 
239 	status = nfs4_set_rw_stateid(&args->src_stateid, src_lock->open_context,
240 				     src_lock, FMODE_READ);
241 	if (status)
242 		return status;
243 
244 	status = nfs_filemap_write_and_wait_range(file_inode(src)->i_mapping,
245 			pos_src, pos_src + (loff_t)count - 1);
246 	if (status)
247 		return status;
248 
249 	status = nfs4_set_rw_stateid(&args->dst_stateid, dst_lock->open_context,
250 				     dst_lock, FMODE_WRITE);
251 	if (status)
252 		return status;
253 
254 	status = nfs_sync_inode(dst_inode);
255 	if (status)
256 		return status;
257 
258 	res->commit_res.verf = NULL;
259 	if (args->sync) {
260 		res->commit_res.verf =
261 			kzalloc(sizeof(struct nfs_writeverf), GFP_NOFS);
262 		if (!res->commit_res.verf)
263 			return -ENOMEM;
264 	}
265 	set_bit(NFS_CLNT_DST_SSC_COPY_STATE,
266 		&dst_lock->open_context->state->flags);
267 
268 	status = nfs4_call_sync(server->client, server, &msg,
269 				&args->seq_args, &res->seq_res, 0);
270 	if (status == -ENOTSUPP)
271 		server->caps &= ~NFS_CAP_COPY;
272 	if (status)
273 		goto out;
274 
275 	if (args->sync &&
276 		nfs_write_verifier_cmp(&res->write_res.verifier.verifier,
277 				    &res->commit_res.verf->verifier)) {
278 		status = -EAGAIN;
279 		goto out;
280 	}
281 
282 	if (!res->synchronous) {
283 		status = handle_async_copy(res, server, src, dst,
284 				&args->src_stateid);
285 		if (status)
286 			return status;
287 	}
288 
289 	if ((!res->synchronous || !args->sync) &&
290 			res->write_res.verifier.committed != NFS_FILE_SYNC) {
291 		status = process_copy_commit(dst, pos_dst, res);
292 		if (status)
293 			return status;
294 	}
295 
296 	truncate_pagecache_range(dst_inode, pos_dst,
297 				 pos_dst + res->write_res.count);
298 
299 	status = res->write_res.count;
300 out:
301 	if (args->sync)
302 		kfree(res->commit_res.verf);
303 	return status;
304 }
305 
306 ssize_t nfs42_proc_copy(struct file *src, loff_t pos_src,
307 			struct file *dst, loff_t pos_dst,
308 			size_t count)
309 {
310 	struct nfs_server *server = NFS_SERVER(file_inode(dst));
311 	struct nfs_lock_context *src_lock;
312 	struct nfs_lock_context *dst_lock;
313 	struct nfs42_copy_args args = {
314 		.src_fh		= NFS_FH(file_inode(src)),
315 		.src_pos	= pos_src,
316 		.dst_fh		= NFS_FH(file_inode(dst)),
317 		.dst_pos	= pos_dst,
318 		.count		= count,
319 		.sync		= false,
320 	};
321 	struct nfs42_copy_res res;
322 	struct nfs4_exception src_exception = {
323 		.inode		= file_inode(src),
324 		.stateid	= &args.src_stateid,
325 	};
326 	struct nfs4_exception dst_exception = {
327 		.inode		= file_inode(dst),
328 		.stateid	= &args.dst_stateid,
329 	};
330 	ssize_t err, err2;
331 
332 	src_lock = nfs_get_lock_context(nfs_file_open_context(src));
333 	if (IS_ERR(src_lock))
334 		return PTR_ERR(src_lock);
335 
336 	src_exception.state = src_lock->open_context->state;
337 
338 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst));
339 	if (IS_ERR(dst_lock)) {
340 		err = PTR_ERR(dst_lock);
341 		goto out_put_src_lock;
342 	}
343 
344 	dst_exception.state = dst_lock->open_context->state;
345 
346 	do {
347 		inode_lock(file_inode(dst));
348 		err = _nfs42_proc_copy(src, src_lock,
349 				dst, dst_lock,
350 				&args, &res);
351 		inode_unlock(file_inode(dst));
352 
353 		if (err >= 0)
354 			break;
355 		if (err == -ENOTSUPP) {
356 			err = -EOPNOTSUPP;
357 			break;
358 		} else if (err == -EAGAIN) {
359 			dst_exception.retry = 1;
360 			continue;
361 		} else if (err == -NFS4ERR_OFFLOAD_NO_REQS && !args.sync) {
362 			args.sync = true;
363 			dst_exception.retry = 1;
364 			continue;
365 		}
366 
367 		err2 = nfs4_handle_exception(server, err, &src_exception);
368 		err  = nfs4_handle_exception(server, err, &dst_exception);
369 		if (!err)
370 			err = err2;
371 	} while (src_exception.retry || dst_exception.retry);
372 
373 	nfs_put_lock_context(dst_lock);
374 out_put_src_lock:
375 	nfs_put_lock_context(src_lock);
376 	return err;
377 }
378 
379 struct nfs42_offloadcancel_data {
380 	struct nfs_server *seq_server;
381 	struct nfs42_offload_status_args args;
382 	struct nfs42_offload_status_res res;
383 };
384 
385 static void nfs42_offload_cancel_prepare(struct rpc_task *task, void *calldata)
386 {
387 	struct nfs42_offloadcancel_data *data = calldata;
388 
389 	nfs4_setup_sequence(data->seq_server->nfs_client,
390 				&data->args.osa_seq_args,
391 				&data->res.osr_seq_res, task);
392 }
393 
394 static void nfs42_offload_cancel_done(struct rpc_task *task, void *calldata)
395 {
396 	struct nfs42_offloadcancel_data *data = calldata;
397 
398 	nfs41_sequence_done(task, &data->res.osr_seq_res);
399 	if (task->tk_status &&
400 		nfs4_async_handle_error(task, data->seq_server, NULL,
401 			NULL) == -EAGAIN)
402 		rpc_restart_call_prepare(task);
403 }
404 
405 static void nfs42_free_offloadcancel_data(void *data)
406 {
407 	kfree(data);
408 }
409 
410 static const struct rpc_call_ops nfs42_offload_cancel_ops = {
411 	.rpc_call_prepare = nfs42_offload_cancel_prepare,
412 	.rpc_call_done = nfs42_offload_cancel_done,
413 	.rpc_release = nfs42_free_offloadcancel_data,
414 };
415 
416 static int nfs42_do_offload_cancel_async(struct file *dst,
417 					 nfs4_stateid *stateid)
418 {
419 	struct nfs_server *dst_server = NFS_SERVER(file_inode(dst));
420 	struct nfs42_offloadcancel_data *data = NULL;
421 	struct nfs_open_context *ctx = nfs_file_open_context(dst);
422 	struct rpc_task *task;
423 	struct rpc_message msg = {
424 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_OFFLOAD_CANCEL],
425 		.rpc_cred = ctx->cred,
426 	};
427 	struct rpc_task_setup task_setup_data = {
428 		.rpc_client = dst_server->client,
429 		.rpc_message = &msg,
430 		.callback_ops = &nfs42_offload_cancel_ops,
431 		.workqueue = nfsiod_workqueue,
432 		.flags = RPC_TASK_ASYNC,
433 	};
434 	int status;
435 
436 	if (!(dst_server->caps & NFS_CAP_OFFLOAD_CANCEL))
437 		return -EOPNOTSUPP;
438 
439 	data = kzalloc(sizeof(struct nfs42_offloadcancel_data), GFP_NOFS);
440 	if (data == NULL)
441 		return -ENOMEM;
442 
443 	data->seq_server = dst_server;
444 	data->args.osa_src_fh = NFS_FH(file_inode(dst));
445 	memcpy(&data->args.osa_stateid, stateid,
446 		sizeof(data->args.osa_stateid));
447 	msg.rpc_argp = &data->args;
448 	msg.rpc_resp = &data->res;
449 	task_setup_data.callback_data = data;
450 	nfs4_init_sequence(&data->args.osa_seq_args, &data->res.osr_seq_res,
451 			   1, 0);
452 	task = rpc_run_task(&task_setup_data);
453 	if (IS_ERR(task))
454 		return PTR_ERR(task);
455 	status = rpc_wait_for_completion_task(task);
456 	if (status == -ENOTSUPP)
457 		dst_server->caps &= ~NFS_CAP_OFFLOAD_CANCEL;
458 	rpc_put_task(task);
459 	return status;
460 }
461 
462 static loff_t _nfs42_proc_llseek(struct file *filep,
463 		struct nfs_lock_context *lock, loff_t offset, int whence)
464 {
465 	struct inode *inode = file_inode(filep);
466 	struct nfs42_seek_args args = {
467 		.sa_fh		= NFS_FH(inode),
468 		.sa_offset	= offset,
469 		.sa_what	= (whence == SEEK_HOLE) ?
470 					NFS4_CONTENT_HOLE : NFS4_CONTENT_DATA,
471 	};
472 	struct nfs42_seek_res res;
473 	struct rpc_message msg = {
474 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_SEEK],
475 		.rpc_argp = &args,
476 		.rpc_resp = &res,
477 	};
478 	struct nfs_server *server = NFS_SERVER(inode);
479 	int status;
480 
481 	if (!nfs_server_capable(inode, NFS_CAP_SEEK))
482 		return -ENOTSUPP;
483 
484 	status = nfs4_set_rw_stateid(&args.sa_stateid, lock->open_context,
485 			lock, FMODE_READ);
486 	if (status)
487 		return status;
488 
489 	status = nfs_filemap_write_and_wait_range(inode->i_mapping,
490 			offset, LLONG_MAX);
491 	if (status)
492 		return status;
493 
494 	status = nfs4_call_sync(server->client, server, &msg,
495 				&args.seq_args, &res.seq_res, 0);
496 	if (status == -ENOTSUPP)
497 		server->caps &= ~NFS_CAP_SEEK;
498 	if (status)
499 		return status;
500 
501 	return vfs_setpos(filep, res.sr_offset, inode->i_sb->s_maxbytes);
502 }
503 
504 loff_t nfs42_proc_llseek(struct file *filep, loff_t offset, int whence)
505 {
506 	struct nfs_server *server = NFS_SERVER(file_inode(filep));
507 	struct nfs4_exception exception = { };
508 	struct nfs_lock_context *lock;
509 	loff_t err;
510 
511 	lock = nfs_get_lock_context(nfs_file_open_context(filep));
512 	if (IS_ERR(lock))
513 		return PTR_ERR(lock);
514 
515 	exception.inode = file_inode(filep);
516 	exception.state = lock->open_context->state;
517 
518 	do {
519 		err = _nfs42_proc_llseek(filep, lock, offset, whence);
520 		if (err >= 0)
521 			break;
522 		if (err == -ENOTSUPP) {
523 			err = -EOPNOTSUPP;
524 			break;
525 		}
526 		err = nfs4_handle_exception(server, err, &exception);
527 	} while (exception.retry);
528 
529 	nfs_put_lock_context(lock);
530 	return err;
531 }
532 
533 
534 static void
535 nfs42_layoutstat_prepare(struct rpc_task *task, void *calldata)
536 {
537 	struct nfs42_layoutstat_data *data = calldata;
538 	struct inode *inode = data->inode;
539 	struct nfs_server *server = NFS_SERVER(inode);
540 	struct pnfs_layout_hdr *lo;
541 
542 	spin_lock(&inode->i_lock);
543 	lo = NFS_I(inode)->layout;
544 	if (!pnfs_layout_is_valid(lo)) {
545 		spin_unlock(&inode->i_lock);
546 		rpc_exit(task, 0);
547 		return;
548 	}
549 	nfs4_stateid_copy(&data->args.stateid, &lo->plh_stateid);
550 	spin_unlock(&inode->i_lock);
551 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
552 			    &data->res.seq_res, task);
553 }
554 
555 static void
556 nfs42_layoutstat_done(struct rpc_task *task, void *calldata)
557 {
558 	struct nfs42_layoutstat_data *data = calldata;
559 	struct inode *inode = data->inode;
560 	struct pnfs_layout_hdr *lo;
561 
562 	if (!nfs4_sequence_done(task, &data->res.seq_res))
563 		return;
564 
565 	switch (task->tk_status) {
566 	case 0:
567 		break;
568 	case -NFS4ERR_BADHANDLE:
569 	case -ESTALE:
570 		pnfs_destroy_layout(NFS_I(inode));
571 		break;
572 	case -NFS4ERR_EXPIRED:
573 	case -NFS4ERR_ADMIN_REVOKED:
574 	case -NFS4ERR_DELEG_REVOKED:
575 	case -NFS4ERR_STALE_STATEID:
576 	case -NFS4ERR_BAD_STATEID:
577 		spin_lock(&inode->i_lock);
578 		lo = NFS_I(inode)->layout;
579 		if (pnfs_layout_is_valid(lo) &&
580 		    nfs4_stateid_match(&data->args.stateid,
581 					     &lo->plh_stateid)) {
582 			LIST_HEAD(head);
583 
584 			/*
585 			 * Mark the bad layout state as invalid, then retry
586 			 * with the current stateid.
587 			 */
588 			pnfs_mark_layout_stateid_invalid(lo, &head);
589 			spin_unlock(&inode->i_lock);
590 			pnfs_free_lseg_list(&head);
591 			nfs_commit_inode(inode, 0);
592 		} else
593 			spin_unlock(&inode->i_lock);
594 		break;
595 	case -NFS4ERR_OLD_STATEID:
596 		spin_lock(&inode->i_lock);
597 		lo = NFS_I(inode)->layout;
598 		if (pnfs_layout_is_valid(lo) &&
599 		    nfs4_stateid_match_other(&data->args.stateid,
600 					&lo->plh_stateid)) {
601 			/* Do we need to delay before resending? */
602 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
603 						&data->args.stateid))
604 				rpc_delay(task, HZ);
605 			rpc_restart_call_prepare(task);
606 		}
607 		spin_unlock(&inode->i_lock);
608 		break;
609 	case -ENOTSUPP:
610 	case -EOPNOTSUPP:
611 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTSTATS;
612 	}
613 }
614 
615 static void
616 nfs42_layoutstat_release(void *calldata)
617 {
618 	struct nfs42_layoutstat_data *data = calldata;
619 	struct nfs42_layoutstat_devinfo *devinfo = data->args.devinfo;
620 	int i;
621 
622 	for (i = 0; i < data->args.num_dev; i++) {
623 		if (devinfo[i].ld_private.ops && devinfo[i].ld_private.ops->free)
624 			devinfo[i].ld_private.ops->free(&devinfo[i].ld_private);
625 	}
626 
627 	pnfs_put_layout_hdr(NFS_I(data->args.inode)->layout);
628 	smp_mb__before_atomic();
629 	clear_bit(NFS_INO_LAYOUTSTATS, &NFS_I(data->args.inode)->flags);
630 	smp_mb__after_atomic();
631 	nfs_iput_and_deactive(data->inode);
632 	kfree(data->args.devinfo);
633 	kfree(data);
634 }
635 
636 static const struct rpc_call_ops nfs42_layoutstat_ops = {
637 	.rpc_call_prepare = nfs42_layoutstat_prepare,
638 	.rpc_call_done = nfs42_layoutstat_done,
639 	.rpc_release = nfs42_layoutstat_release,
640 };
641 
642 int nfs42_proc_layoutstats_generic(struct nfs_server *server,
643 				   struct nfs42_layoutstat_data *data)
644 {
645 	struct rpc_message msg = {
646 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTSTATS],
647 		.rpc_argp = &data->args,
648 		.rpc_resp = &data->res,
649 	};
650 	struct rpc_task_setup task_setup = {
651 		.rpc_client = server->client,
652 		.rpc_message = &msg,
653 		.callback_ops = &nfs42_layoutstat_ops,
654 		.callback_data = data,
655 		.flags = RPC_TASK_ASYNC,
656 	};
657 	struct rpc_task *task;
658 
659 	data->inode = nfs_igrab_and_active(data->args.inode);
660 	if (!data->inode) {
661 		nfs42_layoutstat_release(data);
662 		return -EAGAIN;
663 	}
664 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
665 	task = rpc_run_task(&task_setup);
666 	if (IS_ERR(task))
667 		return PTR_ERR(task);
668 	rpc_put_task(task);
669 	return 0;
670 }
671 
672 static struct nfs42_layouterror_data *
673 nfs42_alloc_layouterror_data(struct pnfs_layout_segment *lseg, gfp_t gfp_flags)
674 {
675 	struct nfs42_layouterror_data *data;
676 	struct inode *inode = lseg->pls_layout->plh_inode;
677 
678 	data = kzalloc(sizeof(*data), gfp_flags);
679 	if (data) {
680 		data->args.inode = data->inode = nfs_igrab_and_active(inode);
681 		if (data->inode) {
682 			data->lseg = pnfs_get_lseg(lseg);
683 			if (data->lseg)
684 				return data;
685 			nfs_iput_and_deactive(data->inode);
686 		}
687 		kfree(data);
688 	}
689 	return NULL;
690 }
691 
692 static void
693 nfs42_free_layouterror_data(struct nfs42_layouterror_data *data)
694 {
695 	pnfs_put_lseg(data->lseg);
696 	nfs_iput_and_deactive(data->inode);
697 	kfree(data);
698 }
699 
700 static void
701 nfs42_layouterror_prepare(struct rpc_task *task, void *calldata)
702 {
703 	struct nfs42_layouterror_data *data = calldata;
704 	struct inode *inode = data->inode;
705 	struct nfs_server *server = NFS_SERVER(inode);
706 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
707 	unsigned i;
708 
709 	spin_lock(&inode->i_lock);
710 	if (!pnfs_layout_is_valid(lo)) {
711 		spin_unlock(&inode->i_lock);
712 		rpc_exit(task, 0);
713 		return;
714 	}
715 	for (i = 0; i < data->args.num_errors; i++)
716 		nfs4_stateid_copy(&data->args.errors[i].stateid,
717 				&lo->plh_stateid);
718 	spin_unlock(&inode->i_lock);
719 	nfs4_setup_sequence(server->nfs_client, &data->args.seq_args,
720 			    &data->res.seq_res, task);
721 }
722 
723 static void
724 nfs42_layouterror_done(struct rpc_task *task, void *calldata)
725 {
726 	struct nfs42_layouterror_data *data = calldata;
727 	struct inode *inode = data->inode;
728 	struct pnfs_layout_hdr *lo = data->lseg->pls_layout;
729 
730 	if (!nfs4_sequence_done(task, &data->res.seq_res))
731 		return;
732 
733 	switch (task->tk_status) {
734 	case 0:
735 		break;
736 	case -NFS4ERR_BADHANDLE:
737 	case -ESTALE:
738 		pnfs_destroy_layout(NFS_I(inode));
739 		break;
740 	case -NFS4ERR_EXPIRED:
741 	case -NFS4ERR_ADMIN_REVOKED:
742 	case -NFS4ERR_DELEG_REVOKED:
743 	case -NFS4ERR_STALE_STATEID:
744 	case -NFS4ERR_BAD_STATEID:
745 		spin_lock(&inode->i_lock);
746 		if (pnfs_layout_is_valid(lo) &&
747 		    nfs4_stateid_match(&data->args.errors[0].stateid,
748 					     &lo->plh_stateid)) {
749 			LIST_HEAD(head);
750 
751 			/*
752 			 * Mark the bad layout state as invalid, then retry
753 			 * with the current stateid.
754 			 */
755 			pnfs_mark_layout_stateid_invalid(lo, &head);
756 			spin_unlock(&inode->i_lock);
757 			pnfs_free_lseg_list(&head);
758 			nfs_commit_inode(inode, 0);
759 		} else
760 			spin_unlock(&inode->i_lock);
761 		break;
762 	case -NFS4ERR_OLD_STATEID:
763 		spin_lock(&inode->i_lock);
764 		if (pnfs_layout_is_valid(lo) &&
765 		    nfs4_stateid_match_other(&data->args.errors[0].stateid,
766 					&lo->plh_stateid)) {
767 			/* Do we need to delay before resending? */
768 			if (!nfs4_stateid_is_newer(&lo->plh_stateid,
769 						&data->args.errors[0].stateid))
770 				rpc_delay(task, HZ);
771 			rpc_restart_call_prepare(task);
772 		}
773 		spin_unlock(&inode->i_lock);
774 		break;
775 	case -ENOTSUPP:
776 	case -EOPNOTSUPP:
777 		NFS_SERVER(inode)->caps &= ~NFS_CAP_LAYOUTERROR;
778 	}
779 }
780 
781 static void
782 nfs42_layouterror_release(void *calldata)
783 {
784 	struct nfs42_layouterror_data *data = calldata;
785 
786 	nfs42_free_layouterror_data(data);
787 }
788 
789 static const struct rpc_call_ops nfs42_layouterror_ops = {
790 	.rpc_call_prepare = nfs42_layouterror_prepare,
791 	.rpc_call_done = nfs42_layouterror_done,
792 	.rpc_release = nfs42_layouterror_release,
793 };
794 
795 int nfs42_proc_layouterror(struct pnfs_layout_segment *lseg,
796 		const struct nfs42_layout_error *errors, size_t n)
797 {
798 	struct inode *inode = lseg->pls_layout->plh_inode;
799 	struct nfs42_layouterror_data *data;
800 	struct rpc_task *task;
801 	struct rpc_message msg = {
802 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_LAYOUTERROR],
803 	};
804 	struct rpc_task_setup task_setup = {
805 		.rpc_message = &msg,
806 		.callback_ops = &nfs42_layouterror_ops,
807 		.flags = RPC_TASK_ASYNC,
808 	};
809 	unsigned int i;
810 
811 	if (!nfs_server_capable(inode, NFS_CAP_LAYOUTERROR))
812 		return -EOPNOTSUPP;
813 	if (n > NFS42_LAYOUTERROR_MAX)
814 		return -EINVAL;
815 	data = nfs42_alloc_layouterror_data(lseg, GFP_NOFS);
816 	if (!data)
817 		return -ENOMEM;
818 	for (i = 0; i < n; i++) {
819 		data->args.errors[i] = errors[i];
820 		data->args.num_errors++;
821 		data->res.num_errors++;
822 	}
823 	msg.rpc_argp = &data->args;
824 	msg.rpc_resp = &data->res;
825 	task_setup.callback_data = data;
826 	task_setup.rpc_client = NFS_SERVER(inode)->client;
827 	nfs4_init_sequence(&data->args.seq_args, &data->res.seq_res, 0, 0);
828 	task = rpc_run_task(&task_setup);
829 	if (IS_ERR(task))
830 		return PTR_ERR(task);
831 	rpc_put_task(task);
832 	return 0;
833 }
834 EXPORT_SYMBOL_GPL(nfs42_proc_layouterror);
835 
836 static int _nfs42_proc_clone(struct rpc_message *msg, struct file *src_f,
837 		struct file *dst_f, struct nfs_lock_context *src_lock,
838 		struct nfs_lock_context *dst_lock, loff_t src_offset,
839 		loff_t dst_offset, loff_t count)
840 {
841 	struct inode *src_inode = file_inode(src_f);
842 	struct inode *dst_inode = file_inode(dst_f);
843 	struct nfs_server *server = NFS_SERVER(dst_inode);
844 	struct nfs42_clone_args args = {
845 		.src_fh = NFS_FH(src_inode),
846 		.dst_fh = NFS_FH(dst_inode),
847 		.src_offset = src_offset,
848 		.dst_offset = dst_offset,
849 		.count = count,
850 		.dst_bitmask = server->cache_consistency_bitmask,
851 	};
852 	struct nfs42_clone_res res = {
853 		.server	= server,
854 	};
855 	int status;
856 
857 	msg->rpc_argp = &args;
858 	msg->rpc_resp = &res;
859 
860 	status = nfs4_set_rw_stateid(&args.src_stateid, src_lock->open_context,
861 			src_lock, FMODE_READ);
862 	if (status)
863 		return status;
864 
865 	status = nfs4_set_rw_stateid(&args.dst_stateid, dst_lock->open_context,
866 			dst_lock, FMODE_WRITE);
867 	if (status)
868 		return status;
869 
870 	res.dst_fattr = nfs_alloc_fattr();
871 	if (!res.dst_fattr)
872 		return -ENOMEM;
873 
874 	status = nfs4_call_sync(server->client, server, msg,
875 				&args.seq_args, &res.seq_res, 0);
876 	if (status == 0)
877 		status = nfs_post_op_update_inode(dst_inode, res.dst_fattr);
878 
879 	kfree(res.dst_fattr);
880 	return status;
881 }
882 
883 int nfs42_proc_clone(struct file *src_f, struct file *dst_f,
884 		     loff_t src_offset, loff_t dst_offset, loff_t count)
885 {
886 	struct rpc_message msg = {
887 		.rpc_proc = &nfs4_procedures[NFSPROC4_CLNT_CLONE],
888 	};
889 	struct inode *inode = file_inode(src_f);
890 	struct nfs_server *server = NFS_SERVER(file_inode(src_f));
891 	struct nfs_lock_context *src_lock;
892 	struct nfs_lock_context *dst_lock;
893 	struct nfs4_exception src_exception = { };
894 	struct nfs4_exception dst_exception = { };
895 	int err, err2;
896 
897 	if (!nfs_server_capable(inode, NFS_CAP_CLONE))
898 		return -EOPNOTSUPP;
899 
900 	src_lock = nfs_get_lock_context(nfs_file_open_context(src_f));
901 	if (IS_ERR(src_lock))
902 		return PTR_ERR(src_lock);
903 
904 	src_exception.inode = file_inode(src_f);
905 	src_exception.state = src_lock->open_context->state;
906 
907 	dst_lock = nfs_get_lock_context(nfs_file_open_context(dst_f));
908 	if (IS_ERR(dst_lock)) {
909 		err = PTR_ERR(dst_lock);
910 		goto out_put_src_lock;
911 	}
912 
913 	dst_exception.inode = file_inode(dst_f);
914 	dst_exception.state = dst_lock->open_context->state;
915 
916 	do {
917 		err = _nfs42_proc_clone(&msg, src_f, dst_f, src_lock, dst_lock,
918 					src_offset, dst_offset, count);
919 		if (err == -ENOTSUPP || err == -EOPNOTSUPP) {
920 			NFS_SERVER(inode)->caps &= ~NFS_CAP_CLONE;
921 			err = -EOPNOTSUPP;
922 			break;
923 		}
924 
925 		err2 = nfs4_handle_exception(server, err, &src_exception);
926 		err = nfs4_handle_exception(server, err, &dst_exception);
927 		if (!err)
928 			err = err2;
929 	} while (src_exception.retry || dst_exception.retry);
930 
931 	nfs_put_lock_context(dst_lock);
932 out_put_src_lock:
933 	nfs_put_lock_context(src_lock);
934 	return err;
935 }
936