btrfs: Convert btree_migratepage to migrate_folio
[linux-block.git] / fs / nfs / file.c
CommitLineData
457c8996 1// SPDX-License-Identifier: GPL-2.0-only
1da177e4
LT
2/*
3 * linux/fs/nfs/file.c
4 *
5 * Copyright (C) 1992 Rick Sladkey
6 *
7 * Changes Copyright (C) 1994 by Florian La Roche
8 * - Do not copy data too often around in the kernel.
9 * - In nfs_file_read the return value of kmalloc wasn't checked.
10 * - Put in a better version of read look-ahead buffering. Original idea
11 * and implementation by Wai S Kok elekokws@ee.nus.sg.
12 *
13 * Expire cache on write to a file by Wai S Kok (Oct 1994).
14 *
15 * Total rewrite of read side for new NFS buffer cache.. Linus.
16 *
17 * nfs regular file handling functions
18 */
19
ddda8e0a 20#include <linux/module.h>
1da177e4
LT
21#include <linux/time.h>
22#include <linux/kernel.h>
23#include <linux/errno.h>
24#include <linux/fcntl.h>
25#include <linux/stat.h>
26#include <linux/nfs_fs.h>
27#include <linux/nfs_mount.h>
28#include <linux/mm.h>
1da177e4 29#include <linux/pagemap.h>
5a0e3ad6 30#include <linux/gfp.h>
b608b283 31#include <linux/swap.h>
1da177e4 32
7c0f6ba6 33#include <linux/uaccess.h>
1da177e4
LT
34
35#include "delegation.h"
94387fb1 36#include "internal.h"
91d5b470 37#include "iostat.h"
545db45f 38#include "fscache.h"
612aa983 39#include "pnfs.h"
1da177e4 40
f4ce1299
TM
41#include "nfstrace.h"
42
1da177e4
LT
43#define NFSDBG_FACILITY NFSDBG_FILE
44
f0f37e2f 45static const struct vm_operations_struct nfs_file_vm_ops;
94387fb1 46
ce4ef7c0 47int nfs_check_flags(int flags)
1da177e4
LT
48{
49 if ((flags & (O_APPEND | O_DIRECT)) == (O_APPEND | O_DIRECT))
50 return -EINVAL;
51
52 return 0;
53}
89d77c8f 54EXPORT_SYMBOL_GPL(nfs_check_flags);
1da177e4
LT
55
56/*
57 * Open file
58 */
59static int
60nfs_file_open(struct inode *inode, struct file *filp)
61{
1da177e4
LT
62 int res;
63
6de1472f 64 dprintk("NFS: open file(%pD2)\n", filp);
cc0dd2d1 65
c2459dc4 66 nfs_inc_stats(inode, NFSIOS_VFSOPEN);
1da177e4
LT
67 res = nfs_check_flags(filp->f_flags);
68 if (res)
69 return res;
70
46cb650c 71 res = nfs_open(inode, filp);
a2ad63da
N
72 if (res == 0)
73 filp->f_mode |= FMODE_CAN_ODIRECT;
1da177e4
LT
74 return res;
75}
76
ce4ef7c0 77int
1da177e4
LT
78nfs_file_release(struct inode *inode, struct file *filp)
79{
6de1472f 80 dprintk("NFS: release(%pD2)\n", filp);
6da24bc9 81
91d5b470 82 nfs_inc_stats(inode, NFSIOS_VFSRELEASE);
aff8d8dc 83 nfs_file_clear_open_context(filp);
a6b5a28e 84 nfs_fscache_release_file(inode, filp);
aff8d8dc 85 return 0;
1da177e4 86}
89d77c8f 87EXPORT_SYMBOL_GPL(nfs_file_release);
1da177e4 88
980802e3 89/**
37eaeed1 90 * nfs_revalidate_file_size - Revalidate the file size
302fad7b
TM
91 * @inode: pointer to inode struct
92 * @filp: pointer to struct file
980802e3
TM
93 *
94 * Revalidates the file length. This is basically a wrapper around
95 * nfs_revalidate_inode() that takes into account the fact that we may
96 * have cached writes (in which case we don't care about the server's
97 * idea of what the file length is), or O_DIRECT (in which case we
98 * shouldn't trust the cache).
99 */
100static int nfs_revalidate_file_size(struct inode *inode, struct file *filp)
101{
102 struct nfs_server *server = NFS_SERVER(inode);
d7cf8dd0 103
980802e3
TM
104 if (filp->f_flags & O_DIRECT)
105 goto force_reval;
13c0b082 106 if (nfs_check_cache_invalid(inode, NFS_INO_INVALID_SIZE))
d7cf8dd0 107 goto force_reval;
d7cf8dd0 108 return 0;
980802e3
TM
109force_reval:
110 return __nfs_revalidate_inode(server, inode);
111}
112
965c8e59 113loff_t nfs_file_llseek(struct file *filp, loff_t offset, int whence)
980802e3 114{
6de1472f
AV
115 dprintk("NFS: llseek file(%pD2, %lld, %d)\n",
116 filp, offset, whence);
b84e06c5 117
06222e49 118 /*
965c8e59 119 * whence == SEEK_END || SEEK_DATA || SEEK_HOLE => we must revalidate
06222e49
JB
120 * the cached file length
121 */
965c8e59 122 if (whence != SEEK_SET && whence != SEEK_CUR) {
980802e3 123 struct inode *inode = filp->f_mapping->host;
d5e66348 124
980802e3
TM
125 int retval = nfs_revalidate_file_size(inode, filp);
126 if (retval < 0)
127 return (loff_t)retval;
79835a71 128 }
d5e66348 129
965c8e59 130 return generic_file_llseek(filp, offset, whence);
980802e3 131}
89d77c8f 132EXPORT_SYMBOL_GPL(nfs_file_llseek);
980802e3 133
1da177e4
LT
134/*
135 * Flush all dirty pages, and check for write errors.
1da177e4 136 */
5445b1fb 137static int
75e1fcc0 138nfs_file_flush(struct file *file, fl_owner_t id)
1da177e4 139{
6de1472f 140 struct inode *inode = file_inode(file);
67dd23f9 141 errseq_t since;
1da177e4 142
6de1472f 143 dprintk("NFS: flush(%pD2)\n", file);
1da177e4 144
c2459dc4 145 nfs_inc_stats(inode, NFSIOS_VFSFLUSH);
1da177e4
LT
146 if ((file->f_mode & FMODE_WRITE) == 0)
147 return 0;
7b159fc1 148
7fe5c398 149 /* Flush writes to the server and return any errors */
67dd23f9
SM
150 since = filemap_sample_wb_err(file->f_mapping);
151 nfs_wb_all(inode);
152 return filemap_check_wb_err(file->f_mapping, since);
1da177e4
LT
153}
154
ce4ef7c0 155ssize_t
3aa2d199 156nfs_file_read(struct kiocb *iocb, struct iov_iter *to)
1da177e4 157{
6de1472f 158 struct inode *inode = file_inode(iocb->ki_filp);
1da177e4
LT
159 ssize_t result;
160
2ba48ce5 161 if (iocb->ki_flags & IOCB_DIRECT)
64158668 162 return nfs_file_direct_read(iocb, to, false);
1da177e4 163
619d30b4 164 dprintk("NFS: read(%pD2, %zu@%lu)\n",
6de1472f 165 iocb->ki_filp,
3aa2d199 166 iov_iter_count(to), (unsigned long) iocb->ki_pos);
1da177e4 167
a5864c99
TM
168 nfs_start_io_read(inode);
169 result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping);
4184dcf2 170 if (!result) {
3aa2d199 171 result = generic_file_read_iter(iocb, to);
4184dcf2
CL
172 if (result > 0)
173 nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, result);
174 }
a5864c99 175 nfs_end_io_read(inode);
1da177e4
LT
176 return result;
177}
89d77c8f 178EXPORT_SYMBOL_GPL(nfs_file_read);
1da177e4 179
ce4ef7c0 180int
1da177e4
LT
181nfs_file_mmap(struct file * file, struct vm_area_struct * vma)
182{
6de1472f 183 struct inode *inode = file_inode(file);
1da177e4
LT
184 int status;
185
6de1472f 186 dprintk("NFS: mmap(%pD2)\n", file);
1da177e4 187
e1ebfd33
TM
188 /* Note: generic_file_mmap() returns ENOSYS on nommu systems
189 * so we call that before revalidating the mapping
190 */
191 status = generic_file_mmap(file, vma);
94387fb1
TM
192 if (!status) {
193 vma->vm_ops = &nfs_file_vm_ops;
e1ebfd33 194 status = nfs_revalidate_mapping(inode, file->f_mapping);
94387fb1 195 }
1da177e4
LT
196 return status;
197}
89d77c8f 198EXPORT_SYMBOL_GPL(nfs_file_mmap);
1da177e4
LT
199
200/*
201 * Flush any dirty pages for this process, and check for write errors.
202 * The return status from this call provides a reliable indication of
203 * whether any write errors occurred for this process.
204 */
4ff79bc7 205static int
bf4b4905 206nfs_file_fsync_commit(struct file *file, int datasync)
1da177e4 207{
6de1472f 208 struct inode *inode = file_inode(file);
9641d9bc 209 int ret, ret2;
af7fa165 210
6de1472f 211 dprintk("NFS: fsync file(%pD2) datasync %d\n", file, datasync);
1da177e4 212
91d5b470 213 nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
2197e9b0 214 ret = nfs_commit_inode(inode, FLUSH_SYNC);
9641d9bc
TM
215 ret2 = file_check_and_advance_wb_err(file);
216 if (ret2 < 0)
217 return ret2;
218 return ret;
a5c58892
BS
219}
220
4ff79bc7 221int
a5c58892
BS
222nfs_file_fsync(struct file *file, loff_t start, loff_t end, int datasync)
223{
2197e9b0 224 struct nfs_open_context *ctx = nfs_file_open_context(file);
496ad9aa 225 struct inode *inode = file_inode(file);
2197e9b0 226 int ret;
a5c58892 227
f4ce1299
TM
228 trace_nfs_fsync_enter(inode);
229
2197e9b0 230 for (;;) {
6fbda89b 231 ret = file_write_and_wait_range(file, start, end);
05990d1b
TM
232 if (ret != 0)
233 break;
bf4b4905 234 ret = nfs_file_fsync_commit(file, datasync);
2197e9b0
TM
235 if (ret != 0)
236 break;
237 ret = pnfs_sync_inode(inode, !!datasync);
238 if (ret != 0)
239 break;
240 if (!test_and_clear_bit(NFS_CONTEXT_RESEND_WRITES, &ctx->flags))
241 break;
dcfc4f25
TM
242 /*
243 * If nfs_file_fsync_commit detected a server reboot, then
244 * resend all dirty pages that might have been covered by
245 * the NFS_CONTEXT_RESEND_WRITES flag
246 */
247 start = 0;
248 end = LLONG_MAX;
2197e9b0 249 }
05990d1b 250
f4ce1299 251 trace_nfs_fsync_exit(inode, ret);
af7fa165 252 return ret;
1da177e4 253}
4ff79bc7 254EXPORT_SYMBOL_GPL(nfs_file_fsync);
1da177e4 255
38c73044
PS
256/*
257 * Decide whether a read/modify/write cycle may be more efficient
258 * then a modify/write/read cycle when writing to a page in the
259 * page cache.
260 *
2cde04e9
KI
261 * Some pNFS layout drivers can only read/write at a certain block
262 * granularity like all block devices and therefore we must perform
263 * read/modify/write whenever a page hasn't read yet and the data
264 * to be written there is not aligned to a block boundary and/or
265 * smaller than the block size.
266 *
38c73044
PS
267 * The modify/write/read cycle may occur if a page is read before
268 * being completely filled by the writer. In this situation, the
269 * page must be completely written to stable storage on the server
270 * before it can be refilled by reading in the page from the server.
271 * This can lead to expensive, small, FILE_SYNC mode writes being
272 * done.
273 *
274 * It may be more efficient to read the page first if the file is
275 * open for reading in addition to writing, the page is not marked
276 * as Uptodate, it is not dirty or waiting to be committed,
277 * indicating that it was previously allocated and then modified,
278 * that there were valid bytes of data in that range of the file,
279 * and that the new data won't completely replace the old data in
280 * that range of the file.
281 */
2cde04e9 282static bool nfs_full_page_write(struct page *page, loff_t pos, unsigned int len)
38c73044
PS
283{
284 unsigned int pglen = nfs_page_length(page);
09cbfeaf 285 unsigned int offset = pos & (PAGE_SIZE - 1);
38c73044
PS
286 unsigned int end = offset + len;
287
2cde04e9
KI
288 return !pglen || (end >= pglen && !offset);
289}
612aa983 290
2cde04e9
KI
291static bool nfs_want_read_modify_write(struct file *file, struct page *page,
292 loff_t pos, unsigned int len)
293{
294 /*
295 * Up-to-date pages, those with ongoing or full-page write
296 * don't need read/modify/write
297 */
298 if (PageUptodate(page) || PagePrivate(page) ||
299 nfs_full_page_write(page, pos, len))
300 return false;
301
302 if (pnfs_ld_read_whole_page(file->f_mapping->host))
303 return true;
304 /* Open for reading too? */
305 if (file->f_mode & FMODE_READ)
306 return true;
307 return false;
38c73044
PS
308}
309
1da177e4 310/*
4899f9c8
NP
311 * This does the "real" work of the write. We must allocate and lock the
312 * page to be sent back to the generic routine, which then copies the
313 * data from user space.
1da177e4
LT
314 *
315 * If the writer ends up delaying the write, the writer needs to
316 * increment the page use counts until he is done with the page.
317 */
4899f9c8 318static int nfs_write_begin(struct file *file, struct address_space *mapping,
9d6b0cd7 319 loff_t pos, unsigned len,
4899f9c8 320 struct page **pagep, void **fsdata)
1da177e4 321{
4899f9c8 322 int ret;
09cbfeaf 323 pgoff_t index = pos >> PAGE_SHIFT;
4899f9c8 324 struct page *page;
38c73044 325 int once_thru = 0;
4899f9c8 326
1e8968c5 327 dfprintk(PAGECACHE, "NFS: write_begin(%pD2(%lu), %u@%lld)\n",
6de1472f 328 file, mapping->host->i_ino, len, (long long) pos);
b7eaefaa 329
38c73044 330start:
b7446e7c 331 page = grab_cache_page_write_begin(mapping, index);
4899f9c8
NP
332 if (!page)
333 return -ENOMEM;
334 *pagep = page;
335
336 ret = nfs_flush_incompatible(file, page);
337 if (ret) {
338 unlock_page(page);
09cbfeaf 339 put_page(page);
38c73044
PS
340 } else if (!once_thru &&
341 nfs_want_read_modify_write(file, page, pos, len)) {
342 once_thru = 1;
65d023af 343 ret = nfs_read_folio(file, page_folio(page));
09cbfeaf 344 put_page(page);
38c73044
PS
345 if (!ret)
346 goto start;
4899f9c8
NP
347 }
348 return ret;
1da177e4
LT
349}
350
4899f9c8
NP
351static int nfs_write_end(struct file *file, struct address_space *mapping,
352 loff_t pos, unsigned len, unsigned copied,
353 struct page *page, void *fsdata)
1da177e4 354{
09cbfeaf 355 unsigned offset = pos & (PAGE_SIZE - 1);
dc24826b 356 struct nfs_open_context *ctx = nfs_file_open_context(file);
4899f9c8 357 int status;
1da177e4 358
1e8968c5 359 dfprintk(PAGECACHE, "NFS: write_end(%pD2(%lu), %u@%lld)\n",
6de1472f 360 file, mapping->host->i_ino, len, (long long) pos);
b7eaefaa 361
efc91ed0
TM
362 /*
363 * Zero any uninitialised parts of the page, and then mark the page
364 * as up to date if it turns out that we're extending the file.
365 */
366 if (!PageUptodate(page)) {
367 unsigned pglen = nfs_page_length(page);
c0cf3ef5 368 unsigned end = offset + copied;
efc91ed0
TM
369
370 if (pglen == 0) {
371 zero_user_segments(page, 0, offset,
09cbfeaf 372 end, PAGE_SIZE);
efc91ed0
TM
373 SetPageUptodate(page);
374 } else if (end >= pglen) {
09cbfeaf 375 zero_user_segment(page, end, PAGE_SIZE);
efc91ed0
TM
376 if (offset == 0)
377 SetPageUptodate(page);
378 } else
09cbfeaf 379 zero_user_segment(page, pglen, PAGE_SIZE);
efc91ed0
TM
380 }
381
4899f9c8 382 status = nfs_updatepage(file, page, offset, copied);
4899f9c8
NP
383
384 unlock_page(page);
09cbfeaf 385 put_page(page);
4899f9c8 386
3d509e54
CL
387 if (status < 0)
388 return status;
2701d086 389 NFS_I(mapping->host)->write_io += copied;
dc24826b 390
d95b2665
TM
391 if (nfs_ctx_key_to_expire(ctx, mapping->host))
392 nfs_wb_all(mapping->host);
dc24826b 393
3d509e54 394 return copied;
1da177e4
LT
395}
396
6b9b3514
DH
397/*
398 * Partially or wholly invalidate a page
399 * - Release the private state associated with a page if undergoing complete
400 * page invalidation
545db45f 401 * - Called if either PG_private or PG_fscache is set on the page
6b9b3514
DH
402 * - Caller holds page lock
403 */
6d740c76
MWO
404static void nfs_invalidate_folio(struct folio *folio, size_t offset,
405 size_t length)
cd52ed35 406{
6d740c76
MWO
407 dfprintk(PAGECACHE, "NFS: invalidate_folio(%lu, %zu, %zu)\n",
408 folio->index, offset, length);
b7eaefaa 409
6d740c76 410 if (offset != 0 || length < folio_size(folio))
1c75950b 411 return;
d2ccddf0 412 /* Cancel any unstarted writes on this page */
6d740c76
MWO
413 nfs_wb_folio_cancel(folio->mapping->host, folio);
414 folio_wait_fscache(folio);
cd52ed35
TM
415}
416
6b9b3514 417/*
3577da4a
MWO
418 * Attempt to release the private state associated with a folio
419 * - Called if either private or fscache flags are set on the folio
420 * - Caller holds folio lock
421 * - Return true (may release folio) or false (may not)
6b9b3514 422 */
3577da4a 423static bool nfs_release_folio(struct folio *folio, gfp_t gfp)
cd52ed35 424{
3577da4a 425 dfprintk(PAGECACHE, "NFS: release_folio(%p)\n", folio);
b7eaefaa 426
3577da4a
MWO
427 /* If the private flag is set, then the folio is not freeable */
428 if (folio_test_private(folio))
429 return false;
430 return nfs_fscache_release_folio(folio, gfp);
e3db7691
TM
431}
432
520f301c 433static void nfs_check_dirty_writeback(struct folio *folio,
f919b196
MG
434 bool *dirty, bool *writeback)
435{
436 struct nfs_inode *nfsi;
520f301c 437 struct address_space *mapping = folio->mapping;
f919b196
MG
438
439 /*
520f301c
MWO
440 * Check if an unstable folio is currently being committed and
441 * if so, have the VM treat it as if the folio is under writeback
442 * so it will not block due to folios that will shortly be freeable.
f919b196
MG
443 */
444 nfsi = NFS_I(mapping->host);
af7cf057 445 if (atomic_read(&nfsi->commit_info.rpcs_out)) {
f919b196
MG
446 *writeback = true;
447 return;
448 }
449
450 /*
520f301c
MWO
451 * If the private flag is set, then the folio is not freeable
452 * and as the inode is not being committed, it's not going to
453 * be cleaned in the near future so treat it as dirty
f919b196 454 */
520f301c 455 if (folio_test_private(folio))
f919b196
MG
456 *dirty = true;
457}
458
6b9b3514
DH
459/*
460 * Attempt to clear the private state associated with a page when an error
461 * occurs that requires the cached contents of an inode to be written back or
462 * destroyed
545db45f 463 * - Called if either PG_private or fscache is set on the page
6b9b3514
DH
464 * - Caller holds page lock
465 * - Return 0 if successful, -error otherwise
466 */
15a30ab2 467static int nfs_launder_folio(struct folio *folio)
e3db7691 468{
15a30ab2 469 struct inode *inode = folio->mapping->host;
b7eaefaa 470
15a30ab2
MWO
471 dfprintk(PAGECACHE, "NFS: launder_folio(%ld, %llu)\n",
472 inode->i_ino, folio_pos(folio));
b7eaefaa 473
15a30ab2
MWO
474 folio_wait_fscache(folio);
475 return nfs_wb_page(inode, &folio->page);
cd52ed35
TM
476}
477
a564b8f0
MG
478static int nfs_swap_activate(struct swap_info_struct *sis, struct file *file,
479 sector_t *span)
480{
bd89bc67
MZ
481 unsigned long blocks;
482 long long isize;
4b60c0ff 483 int ret;
4dc73c67
N
484 struct inode *inode = file_inode(file);
485 struct rpc_clnt *clnt = NFS_CLIENT(inode);
486 struct nfs_client *cl = NFS_SERVER(inode)->nfs_client;
bd89bc67
MZ
487
488 spin_lock(&inode->i_lock);
489 blocks = inode->i_blocks;
490 isize = inode->i_size;
491 spin_unlock(&inode->i_lock);
492 if (blocks*512 < isize) {
493 pr_warn("swap activate: swapfile has holes\n");
494 return -EINVAL;
495 }
dad2b015 496
4b60c0ff
N
497 ret = rpc_clnt_swap_activate(clnt);
498 if (ret)
499 return ret;
500 ret = add_swap_extent(sis, 0, sis->max, 0);
501 if (ret < 0) {
502 rpc_clnt_swap_deactivate(clnt);
503 return ret;
504 }
dad2b015 505
4b60c0ff 506 *span = sis->pages;
4dc73c67
N
507
508 if (cl->rpc_ops->enable_swap)
509 cl->rpc_ops->enable_swap(inode);
510
4b60c0ff
N
511 sis->flags |= SWP_FS_OPS;
512 return ret;
a564b8f0
MG
513}
514
515static void nfs_swap_deactivate(struct file *file)
516{
4dc73c67
N
517 struct inode *inode = file_inode(file);
518 struct rpc_clnt *clnt = NFS_CLIENT(inode);
519 struct nfs_client *cl = NFS_SERVER(inode)->nfs_client;
dad2b015 520
3c87ef6e 521 rpc_clnt_swap_deactivate(clnt);
4dc73c67
N
522 if (cl->rpc_ops->disable_swap)
523 cl->rpc_ops->disable_swap(file_inode(file));
a564b8f0 524}
a564b8f0 525
f5e54d6e 526const struct address_space_operations nfs_file_aops = {
65d023af 527 .read_folio = nfs_read_folio,
8786fde8 528 .readahead = nfs_readahead,
187c82cb 529 .dirty_folio = filemap_dirty_folio,
1da177e4
LT
530 .writepage = nfs_writepage,
531 .writepages = nfs_writepages,
4899f9c8
NP
532 .write_begin = nfs_write_begin,
533 .write_end = nfs_write_end,
6d740c76 534 .invalidate_folio = nfs_invalidate_folio,
3577da4a 535 .release_folio = nfs_release_folio,
f844cd0d 536#ifdef CONFIG_MIGRATION
074cc1de 537 .migratepage = nfs_migrate_page,
f844cd0d 538#endif
15a30ab2 539 .launder_folio = nfs_launder_folio,
f919b196 540 .is_dirty_writeback = nfs_check_dirty_writeback,
f590f333 541 .error_remove_page = generic_error_remove_page,
a564b8f0
MG
542 .swap_activate = nfs_swap_activate,
543 .swap_deactivate = nfs_swap_deactivate,
eb79f3af 544 .swap_rw = nfs_swap_rw,
1da177e4
LT
545};
546
6b9b3514
DH
547/*
548 * Notification that a PTE pointing to an NFS page is about to be made
549 * writable, implying that someone is about to modify the page through a
550 * shared-writable mapping
551 */
01a36844 552static vm_fault_t nfs_vm_page_mkwrite(struct vm_fault *vmf)
94387fb1 553{
c2ec175c 554 struct page *page = vmf->page;
11bac800 555 struct file *filp = vmf->vma->vm_file;
6de1472f 556 struct inode *inode = file_inode(filp);
94387fb1 557 unsigned pagelen;
01a36844 558 vm_fault_t ret = VM_FAULT_NOPAGE;
4899f9c8 559 struct address_space *mapping;
94387fb1 560
1e8968c5 561 dfprintk(PAGECACHE, "NFS: vm_page_mkwrite(%pD2(%lu), offset %lld)\n",
6de1472f 562 filp, filp->f_mapping->host->i_ino,
b7eaefaa
CL
563 (long long)page_offset(page));
564
9a773e7c
TM
565 sb_start_pagefault(inode->i_sb);
566
545db45f 567 /* make sure the cache has finished storing the page */
a6b5a28e
DW
568 if (PageFsCache(page) &&
569 wait_on_page_fscache_killable(vmf->page) < 0) {
570 ret = VM_FAULT_RETRY;
571 goto out;
572 }
545db45f 573
ef070dcb
TM
574 wait_on_bit_action(&NFS_I(inode)->flags, NFS_INO_INVALIDATING,
575 nfs_wait_bit_killable, TASK_KILLABLE);
576
94387fb1 577 lock_page(page);
d56b4ddf 578 mapping = page_file_mapping(page);
6de1472f 579 if (mapping != inode->i_mapping)
8b1f9ee5
TM
580 goto out_unlock;
581
2aeb98f4
TM
582 wait_on_page_writeback(page);
583
94387fb1 584 pagelen = nfs_page_length(page);
8b1f9ee5
TM
585 if (pagelen == 0)
586 goto out_unlock;
4899f9c8 587
bc4866b6
TM
588 ret = VM_FAULT_LOCKED;
589 if (nfs_flush_incompatible(filp, page) == 0 &&
590 nfs_updatepage(filp, page, 0, pagelen) == 0)
591 goto out;
8b1f9ee5 592
bc4866b6 593 ret = VM_FAULT_SIGBUS;
8b1f9ee5
TM
594out_unlock:
595 unlock_page(page);
bc4866b6 596out:
9a773e7c 597 sb_end_pagefault(inode->i_sb);
bc4866b6 598 return ret;
94387fb1
TM
599}
600
f0f37e2f 601static const struct vm_operations_struct nfs_file_vm_ops = {
94387fb1 602 .fault = filemap_fault,
f1820361 603 .map_pages = filemap_map_pages,
94387fb1
TM
604 .page_mkwrite = nfs_vm_page_mkwrite,
605};
606
edaf4369 607ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from)
1da177e4 608{
6de1472f
AV
609 struct file *file = iocb->ki_filp;
610 struct inode *inode = file_inode(file);
ed7bcdb3
TM
611 unsigned int mntflags = NFS_SERVER(inode)->flags;
612 ssize_t result, written;
ce368536
SM
613 errseq_t since;
614 int error;
1da177e4 615
6de1472f 616 result = nfs_key_timeout_notify(file, inode);
dc24826b
AA
617 if (result)
618 return result;
619
89698b24 620 if (iocb->ki_flags & IOCB_DIRECT)
64158668 621 return nfs_file_direct_write(iocb, from, false);
1da177e4 622
619d30b4 623 dprintk("NFS: write(%pD2, %zu@%Ld)\n",
18290650 624 file, iov_iter_count(from), (long long) iocb->ki_pos);
1da177e4 625
1da177e4
LT
626 if (IS_SWAPFILE(inode))
627 goto out_swapfile;
7d52e862
TM
628 /*
629 * O_APPEND implies that we must revalidate the file length.
630 */
fc9dc401 631 if (iocb->ki_flags & IOCB_APPEND || iocb->ki_pos > i_size_read(inode)) {
6de1472f 632 result = nfs_revalidate_file_size(inode, file);
7d52e862 633 if (result)
e6005436 634 return result;
fe51beec 635 }
1da177e4 636
28aa2f9e
TM
637 nfs_clear_invalid_mapping(file->f_mapping);
638
ce368536 639 since = filemap_sample_wb_err(file->f_mapping);
a5864c99 640 nfs_start_io_write(inode);
18290650
TM
641 result = generic_write_checks(iocb, from);
642 if (result > 0) {
643 current->backing_dev_info = inode_to_bdi(inode);
800ba295 644 result = generic_perform_write(iocb, from);
18290650
TM
645 current->backing_dev_info = NULL;
646 }
a5864c99 647 nfs_end_io_write(inode);
18290650 648 if (result <= 0)
1da177e4
LT
649 goto out;
650
c49edecd 651 written = result;
18290650 652 iocb->ki_pos += written;
e6005436 653 nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written);
ed7bcdb3
TM
654
655 if (mntflags & NFS_MOUNT_WRITE_EAGER) {
656 result = filemap_fdatawrite_range(file->f_mapping,
657 iocb->ki_pos - written,
658 iocb->ki_pos - 1);
659 if (result < 0)
660 goto out;
661 }
662 if (mntflags & NFS_MOUNT_WRITE_WAIT) {
663 result = filemap_fdatawait_range(file->f_mapping,
664 iocb->ki_pos - written,
665 iocb->ki_pos - 1);
666 if (result < 0)
667 goto out;
668 }
e973b1a5 669 result = generic_write_sync(iocb, written);
670 if (result < 0)
e6005436 671 return result;
7e381172 672
e6005436 673out:
7e94d6c4 674 /* Return error values */
ce368536 675 error = filemap_check_wb_err(file->f_mapping, since);
e6005436
TM
676 switch (error) {
677 default:
678 break;
679 case -EDQUOT:
680 case -EFBIG:
681 case -ENOSPC:
682 nfs_wb_all(inode);
683 error = file_check_and_advance_wb_err(file);
684 if (error < 0)
685 result = error;
200baa21 686 }
1da177e4
LT
687 return result;
688
689out_swapfile:
690 printk(KERN_INFO "NFS: attempt to write to active swap file!\n");
89658c4d 691 return -ETXTBSY;
1da177e4 692}
89d77c8f 693EXPORT_SYMBOL_GPL(nfs_file_write);
1da177e4 694
5eebde23
SJ
695static int
696do_getlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
1da177e4
LT
697{
698 struct inode *inode = filp->f_mapping->host;
699 int status = 0;
21ac19d4 700 unsigned int saved_type = fl->fl_type;
1da177e4 701
039c4d7a 702 /* Try local locking first */
6d34ac19
BF
703 posix_test_lock(filp, fl);
704 if (fl->fl_type != F_UNLCK) {
705 /* found a conflict */
039c4d7a 706 goto out;
1da177e4 707 }
21ac19d4 708 fl->fl_type = saved_type;
039c4d7a 709
011e2a7f 710 if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
039c4d7a
TM
711 goto out_noconflict;
712
5eebde23 713 if (is_local)
039c4d7a
TM
714 goto out_noconflict;
715
716 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
717out:
1da177e4 718 return status;
039c4d7a
TM
719out_noconflict:
720 fl->fl_type = F_UNLCK;
721 goto out;
1da177e4
LT
722}
723
5eebde23
SJ
724static int
725do_unlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
1da177e4
LT
726{
727 struct inode *inode = filp->f_mapping->host;
7a8203d8 728 struct nfs_lock_context *l_ctx;
1da177e4
LT
729 int status;
730
1da177e4
LT
731 /*
732 * Flush all pending writes before doing anything
733 * with locks..
734 */
aded8d7b 735 nfs_wb_all(inode);
1da177e4 736
7a8203d8
TM
737 l_ctx = nfs_get_lock_context(nfs_file_open_context(filp));
738 if (!IS_ERR(l_ctx)) {
210c7c17 739 status = nfs_iocounter_wait(l_ctx);
7a8203d8 740 nfs_put_lock_context(l_ctx);
f30cb757
BC
741 /* NOTE: special case
742 * If we're signalled while cleaning up locks on process exit, we
743 * still need to complete the unlock.
744 */
745 if (status < 0 && !(fl->fl_flags & FL_CLOSE))
7a8203d8
TM
746 return status;
747 }
748
5eebde23
SJ
749 /*
750 * Use local locking if mounted with "-onolock" or with appropriate
751 * "-olocal_lock="
752 */
753 if (!is_local)
1da177e4
LT
754 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
755 else
75575ddf 756 status = locks_lock_file_wait(filp, fl);
1da177e4
LT
757 return status;
758}
759
5eebde23
SJ
760static int
761do_setlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
1da177e4
LT
762{
763 struct inode *inode = filp->f_mapping->host;
1da177e4
LT
764 int status;
765
1da177e4
LT
766 /*
767 * Flush all pending writes before doing anything
768 * with locks..
769 */
29884df0
TM
770 status = nfs_sync_mapping(filp->f_mapping);
771 if (status != 0)
1da177e4
LT
772 goto out;
773
5eebde23
SJ
774 /*
775 * Use local locking if mounted with "-onolock" or with appropriate
776 * "-olocal_lock="
777 */
778 if (!is_local)
1da177e4 779 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
c4d7c402 780 else
75575ddf 781 status = locks_lock_file_wait(filp, fl);
1da177e4
LT
782 if (status < 0)
783 goto out;
6b96724e 784
1da177e4 785 /*
779eafab
N
786 * Invalidate cache to prevent missing any changes. If
787 * the file is mapped, clear the page cache as well so
788 * those mappings will be loaded.
6b96724e 789 *
1da177e4
LT
790 * This makes locking act as a cache coherency point.
791 */
29884df0 792 nfs_sync_mapping(filp->f_mapping);
779eafab 793 if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) {
442ce049 794 nfs_zap_caches(inode);
779eafab
N
795 if (mapping_mapped(filp->f_mapping))
796 nfs_revalidate_mapping(inode, filp->f_mapping);
797 }
1da177e4 798out:
1da177e4
LT
799 return status;
800}
801
802/*
803 * Lock a (portion of) a file
804 */
ce4ef7c0 805int nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
1da177e4 806{
6da24bc9 807 struct inode *inode = filp->f_mapping->host;
2116271a 808 int ret = -ENOLCK;
5eebde23 809 int is_local = 0;
1da177e4 810
6de1472f
AV
811 dprintk("NFS: lock(%pD2, t=%x, fl=%x, r=%lld:%lld)\n",
812 filp, fl->fl_type, fl->fl_flags,
1da177e4 813 (long long)fl->fl_start, (long long)fl->fl_end);
6da24bc9 814
91d5b470 815 nfs_inc_stats(inode, NFSIOS_VFSLOCK);
1da177e4 816
bb0a55bb
BF
817 if (fl->fl_flags & FL_RECLAIM)
818 return -ENOGRACE;
819
5eebde23
SJ
820 if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FCNTL)
821 is_local = 1;
822
2116271a
TM
823 if (NFS_PROTO(inode)->lock_check_bounds != NULL) {
824 ret = NFS_PROTO(inode)->lock_check_bounds(fl);
825 if (ret < 0)
826 goto out_err;
827 }
1da177e4
LT
828
829 if (IS_GETLK(cmd))
5eebde23 830 ret = do_getlk(filp, cmd, fl, is_local);
2116271a 831 else if (fl->fl_type == F_UNLCK)
5eebde23 832 ret = do_unlk(filp, cmd, fl, is_local);
2116271a 833 else
5eebde23 834 ret = do_setlk(filp, cmd, fl, is_local);
2116271a
TM
835out_err:
836 return ret;
1da177e4 837}
89d77c8f 838EXPORT_SYMBOL_GPL(nfs_lock);
1da177e4
LT
839
840/*
841 * Lock a (portion of) a file
842 */
ce4ef7c0 843int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
1da177e4 844{
5eebde23
SJ
845 struct inode *inode = filp->f_mapping->host;
846 int is_local = 0;
847
6de1472f
AV
848 dprintk("NFS: flock(%pD2, t=%x, fl=%x)\n",
849 filp, fl->fl_type, fl->fl_flags);
1da177e4 850
1da177e4
LT
851 if (!(fl->fl_flags & FL_FLOCK))
852 return -ENOLCK;
853
5eebde23
SJ
854 if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK)
855 is_local = 1;
856
fcfa4470
BC
857 /* We're simulating flock() locks using posix locks on the server */
858 if (fl->fl_type == F_UNLCK)
5eebde23
SJ
859 return do_unlk(filp, cmd, fl, is_local);
860 return do_setlk(filp, cmd, fl, is_local);
1da177e4 861}
89d77c8f 862EXPORT_SYMBOL_GPL(nfs_flock);
370f6599 863
0486958f
JL
864const struct file_operations nfs_file_operations = {
865 .llseek = nfs_file_llseek,
3aa2d199 866 .read_iter = nfs_file_read,
edaf4369 867 .write_iter = nfs_file_write,
0486958f
JL
868 .mmap = nfs_file_mmap,
869 .open = nfs_file_open,
870 .flush = nfs_file_flush,
871 .release = nfs_file_release,
872 .fsync = nfs_file_fsync,
873 .lock = nfs_lock,
874 .flock = nfs_flock,
82c156f8 875 .splice_read = generic_file_splice_read,
4da54c21 876 .splice_write = iter_file_splice_write,
0486958f 877 .check_flags = nfs_check_flags,
1c994a09 878 .setlease = simple_nosetlease,
0486958f 879};
ddda8e0a 880EXPORT_SYMBOL_GPL(nfs_file_operations);