Merge tag 'dmaengine-6.4-rc1' of git://git.kernel.org/pub/scm/linux/kernel/git/vkoul...
[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>
5970e15d 34#include <linux/filelock.h>
1da177e4
LT
35
36#include "delegation.h"
94387fb1 37#include "internal.h"
91d5b470 38#include "iostat.h"
545db45f 39#include "fscache.h"
612aa983 40#include "pnfs.h"
1da177e4 41
f4ce1299
TM
42#include "nfstrace.h"
43
1da177e4
LT
44#define NFSDBG_FACILITY NFSDBG_FILE
45
f0f37e2f 46static const struct vm_operations_struct nfs_file_vm_ops;
94387fb1 47
ce4ef7c0 48int nfs_check_flags(int flags)
1da177e4
LT
49{
50 if ((flags & (O_APPEND | O_DIRECT)) == (O_APPEND | O_DIRECT))
51 return -EINVAL;
52
53 return 0;
54}
89d77c8f 55EXPORT_SYMBOL_GPL(nfs_check_flags);
1da177e4
LT
56
57/*
58 * Open file
59 */
60static int
61nfs_file_open(struct inode *inode, struct file *filp)
62{
1da177e4
LT
63 int res;
64
6de1472f 65 dprintk("NFS: open file(%pD2)\n", filp);
cc0dd2d1 66
c2459dc4 67 nfs_inc_stats(inode, NFSIOS_VFSOPEN);
1da177e4
LT
68 res = nfs_check_flags(filp->f_flags);
69 if (res)
70 return res;
71
46cb650c 72 res = nfs_open(inode, filp);
a2ad63da
N
73 if (res == 0)
74 filp->f_mode |= FMODE_CAN_ODIRECT;
1da177e4
LT
75 return res;
76}
77
ce4ef7c0 78int
1da177e4
LT
79nfs_file_release(struct inode *inode, struct file *filp)
80{
6de1472f 81 dprintk("NFS: release(%pD2)\n", filp);
6da24bc9 82
91d5b470 83 nfs_inc_stats(inode, NFSIOS_VFSRELEASE);
aff8d8dc 84 nfs_file_clear_open_context(filp);
a6b5a28e 85 nfs_fscache_release_file(inode, filp);
aff8d8dc 86 return 0;
1da177e4 87}
89d77c8f 88EXPORT_SYMBOL_GPL(nfs_file_release);
1da177e4 89
980802e3 90/**
37eaeed1 91 * nfs_revalidate_file_size - Revalidate the file size
302fad7b
TM
92 * @inode: pointer to inode struct
93 * @filp: pointer to struct file
980802e3
TM
94 *
95 * Revalidates the file length. This is basically a wrapper around
96 * nfs_revalidate_inode() that takes into account the fact that we may
97 * have cached writes (in which case we don't care about the server's
98 * idea of what the file length is), or O_DIRECT (in which case we
99 * shouldn't trust the cache).
100 */
101static int nfs_revalidate_file_size(struct inode *inode, struct file *filp)
102{
103 struct nfs_server *server = NFS_SERVER(inode);
d7cf8dd0 104
980802e3
TM
105 if (filp->f_flags & O_DIRECT)
106 goto force_reval;
13c0b082 107 if (nfs_check_cache_invalid(inode, NFS_INO_INVALID_SIZE))
d7cf8dd0 108 goto force_reval;
d7cf8dd0 109 return 0;
980802e3
TM
110force_reval:
111 return __nfs_revalidate_inode(server, inode);
112}
113
965c8e59 114loff_t nfs_file_llseek(struct file *filp, loff_t offset, int whence)
980802e3 115{
6de1472f
AV
116 dprintk("NFS: llseek file(%pD2, %lld, %d)\n",
117 filp, offset, whence);
b84e06c5 118
06222e49 119 /*
965c8e59 120 * whence == SEEK_END || SEEK_DATA || SEEK_HOLE => we must revalidate
06222e49
JB
121 * the cached file length
122 */
965c8e59 123 if (whence != SEEK_SET && whence != SEEK_CUR) {
980802e3 124 struct inode *inode = filp->f_mapping->host;
d5e66348 125
980802e3
TM
126 int retval = nfs_revalidate_file_size(inode, filp);
127 if (retval < 0)
128 return (loff_t)retval;
79835a71 129 }
d5e66348 130
965c8e59 131 return generic_file_llseek(filp, offset, whence);
980802e3 132}
89d77c8f 133EXPORT_SYMBOL_GPL(nfs_file_llseek);
980802e3 134
1da177e4
LT
135/*
136 * Flush all dirty pages, and check for write errors.
1da177e4 137 */
5445b1fb 138static int
75e1fcc0 139nfs_file_flush(struct file *file, fl_owner_t id)
1da177e4 140{
6de1472f 141 struct inode *inode = file_inode(file);
67dd23f9 142 errseq_t since;
1da177e4 143
6de1472f 144 dprintk("NFS: flush(%pD2)\n", file);
1da177e4 145
c2459dc4 146 nfs_inc_stats(inode, NFSIOS_VFSFLUSH);
1da177e4
LT
147 if ((file->f_mode & FMODE_WRITE) == 0)
148 return 0;
7b159fc1 149
7fe5c398 150 /* Flush writes to the server and return any errors */
67dd23f9
SM
151 since = filemap_sample_wb_err(file->f_mapping);
152 nfs_wb_all(inode);
153 return filemap_check_wb_err(file->f_mapping, since);
1da177e4
LT
154}
155
ce4ef7c0 156ssize_t
3aa2d199 157nfs_file_read(struct kiocb *iocb, struct iov_iter *to)
1da177e4 158{
6de1472f 159 struct inode *inode = file_inode(iocb->ki_filp);
1da177e4
LT
160 ssize_t result;
161
2ba48ce5 162 if (iocb->ki_flags & IOCB_DIRECT)
64158668 163 return nfs_file_direct_read(iocb, to, false);
1da177e4 164
619d30b4 165 dprintk("NFS: read(%pD2, %zu@%lu)\n",
6de1472f 166 iocb->ki_filp,
3aa2d199 167 iov_iter_count(to), (unsigned long) iocb->ki_pos);
1da177e4 168
a5864c99
TM
169 nfs_start_io_read(inode);
170 result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping);
4184dcf2 171 if (!result) {
3aa2d199 172 result = generic_file_read_iter(iocb, to);
4184dcf2
CL
173 if (result > 0)
174 nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, result);
175 }
a5864c99 176 nfs_end_io_read(inode);
1da177e4
LT
177 return result;
178}
89d77c8f 179EXPORT_SYMBOL_GPL(nfs_file_read);
1da177e4 180
ce4ef7c0 181int
1da177e4
LT
182nfs_file_mmap(struct file * file, struct vm_area_struct * vma)
183{
6de1472f 184 struct inode *inode = file_inode(file);
1da177e4
LT
185 int status;
186
6de1472f 187 dprintk("NFS: mmap(%pD2)\n", file);
1da177e4 188
e1ebfd33
TM
189 /* Note: generic_file_mmap() returns ENOSYS on nommu systems
190 * so we call that before revalidating the mapping
191 */
192 status = generic_file_mmap(file, vma);
94387fb1
TM
193 if (!status) {
194 vma->vm_ops = &nfs_file_vm_ops;
e1ebfd33 195 status = nfs_revalidate_mapping(inode, file->f_mapping);
94387fb1 196 }
1da177e4
LT
197 return status;
198}
89d77c8f 199EXPORT_SYMBOL_GPL(nfs_file_mmap);
1da177e4
LT
200
201/*
202 * Flush any dirty pages for this process, and check for write errors.
203 * The return status from this call provides a reliable indication of
204 * whether any write errors occurred for this process.
205 */
4ff79bc7 206static int
bf4b4905 207nfs_file_fsync_commit(struct file *file, int datasync)
1da177e4 208{
6de1472f 209 struct inode *inode = file_inode(file);
9641d9bc 210 int ret, ret2;
af7fa165 211
6de1472f 212 dprintk("NFS: fsync file(%pD2) datasync %d\n", file, datasync);
1da177e4 213
91d5b470 214 nfs_inc_stats(inode, NFSIOS_VFSFSYNC);
2197e9b0 215 ret = nfs_commit_inode(inode, FLUSH_SYNC);
9641d9bc
TM
216 ret2 = file_check_and_advance_wb_err(file);
217 if (ret2 < 0)
218 return ret2;
219 return ret;
a5c58892
BS
220}
221
4ff79bc7 222int
a5c58892
BS
223nfs_file_fsync(struct file *file, loff_t start, loff_t end, int datasync)
224{
496ad9aa 225 struct inode *inode = file_inode(file);
67f4b5dc
TM
226 struct nfs_inode *nfsi = NFS_I(inode);
227 long save_nredirtied = atomic_long_read(&nfsi->redirtied_pages);
228 long nredirtied;
2197e9b0 229 int ret;
a5c58892 230
f4ce1299
TM
231 trace_nfs_fsync_enter(inode);
232
2197e9b0 233 for (;;) {
6fbda89b 234 ret = file_write_and_wait_range(file, start, end);
05990d1b
TM
235 if (ret != 0)
236 break;
bf4b4905 237 ret = nfs_file_fsync_commit(file, datasync);
2197e9b0
TM
238 if (ret != 0)
239 break;
240 ret = pnfs_sync_inode(inode, !!datasync);
241 if (ret != 0)
242 break;
67f4b5dc
TM
243 nredirtied = atomic_long_read(&nfsi->redirtied_pages);
244 if (nredirtied == save_nredirtied)
2197e9b0 245 break;
67f4b5dc 246 save_nredirtied = nredirtied;
2197e9b0 247 }
05990d1b 248
f4ce1299 249 trace_nfs_fsync_exit(inode, ret);
af7fa165 250 return ret;
1da177e4 251}
4ff79bc7 252EXPORT_SYMBOL_GPL(nfs_file_fsync);
1da177e4 253
38c73044
PS
254/*
255 * Decide whether a read/modify/write cycle may be more efficient
256 * then a modify/write/read cycle when writing to a page in the
257 * page cache.
258 *
2cde04e9
KI
259 * Some pNFS layout drivers can only read/write at a certain block
260 * granularity like all block devices and therefore we must perform
261 * read/modify/write whenever a page hasn't read yet and the data
262 * to be written there is not aligned to a block boundary and/or
263 * smaller than the block size.
264 *
38c73044
PS
265 * The modify/write/read cycle may occur if a page is read before
266 * being completely filled by the writer. In this situation, the
267 * page must be completely written to stable storage on the server
268 * before it can be refilled by reading in the page from the server.
269 * This can lead to expensive, small, FILE_SYNC mode writes being
270 * done.
271 *
272 * It may be more efficient to read the page first if the file is
273 * open for reading in addition to writing, the page is not marked
274 * as Uptodate, it is not dirty or waiting to be committed,
275 * indicating that it was previously allocated and then modified,
276 * that there were valid bytes of data in that range of the file,
277 * and that the new data won't completely replace the old data in
278 * that range of the file.
279 */
54d99381
TM
280static bool nfs_folio_is_full_write(struct folio *folio, loff_t pos,
281 unsigned int len)
38c73044 282{
54d99381
TM
283 unsigned int pglen = nfs_folio_length(folio);
284 unsigned int offset = offset_in_folio(folio, pos);
38c73044
PS
285 unsigned int end = offset + len;
286
2cde04e9
KI
287 return !pglen || (end >= pglen && !offset);
288}
612aa983 289
54d99381
TM
290static bool nfs_want_read_modify_write(struct file *file, struct folio *folio,
291 loff_t pos, unsigned int len)
2cde04e9
KI
292{
293 /*
294 * Up-to-date pages, those with ongoing or full-page write
295 * don't need read/modify/write
296 */
54d99381
TM
297 if (folio_test_uptodate(folio) || folio_test_private(folio) ||
298 nfs_folio_is_full_write(folio, pos, len))
2cde04e9
KI
299 return false;
300
54d99381 301 if (pnfs_ld_read_whole_page(file_inode(file)))
2cde04e9
KI
302 return true;
303 /* Open for reading too? */
304 if (file->f_mode & FMODE_READ)
305 return true;
306 return false;
38c73044
PS
307}
308
1da177e4 309/*
4899f9c8
NP
310 * This does the "real" work of the write. We must allocate and lock the
311 * page to be sent back to the generic routine, which then copies the
312 * data from user space.
1da177e4
LT
313 *
314 * If the writer ends up delaying the write, the writer needs to
315 * increment the page use counts until he is done with the page.
316 */
4899f9c8 317static int nfs_write_begin(struct file *file, struct address_space *mapping,
54d99381
TM
318 loff_t pos, unsigned len, struct page **pagep,
319 void **fsdata)
1da177e4 320{
0c493b5c 321 struct folio *folio;
38c73044 322 int once_thru = 0;
54d99381 323 int ret;
4899f9c8 324
1e8968c5 325 dfprintk(PAGECACHE, "NFS: write_begin(%pD2(%lu), %u@%lld)\n",
6de1472f 326 file, mapping->host->i_ino, len, (long long) pos);
b7eaefaa 327
38c73044 328start:
e999a5c5
MW
329 folio = __filemap_get_folio(mapping, pos >> PAGE_SHIFT, FGP_WRITEBEGIN,
330 mapping_gfp_mask(mapping));
66dabbb6
CH
331 if (IS_ERR(folio))
332 return PTR_ERR(folio);
54d99381 333 *pagep = &folio->page;
4899f9c8 334
0c493b5c 335 ret = nfs_flush_incompatible(file, folio);
4899f9c8 336 if (ret) {
54d99381
TM
337 folio_unlock(folio);
338 folio_put(folio);
38c73044 339 } else if (!once_thru &&
54d99381 340 nfs_want_read_modify_write(file, folio, pos, len)) {
38c73044 341 once_thru = 1;
0c493b5c 342 ret = nfs_read_folio(file, folio);
54d99381 343 folio_put(folio);
38c73044
PS
344 if (!ret)
345 goto start;
4899f9c8
NP
346 }
347 return ret;
1da177e4
LT
348}
349
4899f9c8 350static int nfs_write_end(struct file *file, struct address_space *mapping,
54d99381
TM
351 loff_t pos, unsigned len, unsigned copied,
352 struct page *page, void *fsdata)
1da177e4 353{
dc24826b 354 struct nfs_open_context *ctx = nfs_file_open_context(file);
0c493b5c 355 struct folio *folio = page_folio(page);
54d99381 356 unsigned offset = offset_in_folio(folio, pos);
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 */
54d99381
TM
366 if (!folio_test_uptodate(folio)) {
367 size_t fsize = folio_size(folio);
368 unsigned pglen = nfs_folio_length(folio);
c0cf3ef5 369 unsigned end = offset + copied;
efc91ed0
TM
370
371 if (pglen == 0) {
54d99381
TM
372 folio_zero_segments(folio, 0, offset, end, fsize);
373 folio_mark_uptodate(folio);
efc91ed0 374 } else if (end >= pglen) {
54d99381 375 folio_zero_segment(folio, end, fsize);
efc91ed0 376 if (offset == 0)
54d99381 377 folio_mark_uptodate(folio);
efc91ed0 378 } else
54d99381 379 folio_zero_segment(folio, pglen, fsize);
efc91ed0
TM
380 }
381
0c493b5c 382 status = nfs_update_folio(file, folio, offset, copied);
4899f9c8 383
54d99381
TM
384 folio_unlock(folio);
385 folio_put(folio);
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{
eb5654b3 407 struct inode *inode = folio_file_mapping(folio)->host;
6d740c76
MWO
408 dfprintk(PAGECACHE, "NFS: invalidate_folio(%lu, %zu, %zu)\n",
409 folio->index, offset, length);
b7eaefaa 410
6d740c76 411 if (offset != 0 || length < folio_size(folio))
1c75950b 412 return;
d2ccddf0 413 /* Cancel any unstarted writes on this page */
eb5654b3 414 nfs_wb_folio_cancel(inode, folio);
6d740c76 415 folio_wait_fscache(folio);
eb5654b3 416 trace_nfs_invalidate_folio(inode, folio);
cd52ed35
TM
417}
418
6b9b3514 419/*
3577da4a
MWO
420 * Attempt to release the private state associated with a folio
421 * - Called if either private or fscache flags are set on the folio
422 * - Caller holds folio lock
423 * - Return true (may release folio) or false (may not)
6b9b3514 424 */
3577da4a 425static bool nfs_release_folio(struct folio *folio, gfp_t gfp)
cd52ed35 426{
3577da4a 427 dfprintk(PAGECACHE, "NFS: release_folio(%p)\n", folio);
b7eaefaa 428
3577da4a 429 /* If the private flag is set, then the folio is not freeable */
96780ca5
TM
430 if (folio_test_private(folio)) {
431 if ((current_gfp_context(gfp) & GFP_KERNEL) != GFP_KERNEL ||
432 current_is_kswapd())
433 return false;
434 if (nfs_wb_folio(folio_file_mapping(folio)->host, folio) < 0)
435 return false;
436 }
3577da4a 437 return nfs_fscache_release_folio(folio, gfp);
e3db7691
TM
438}
439
520f301c 440static void nfs_check_dirty_writeback(struct folio *folio,
f919b196
MG
441 bool *dirty, bool *writeback)
442{
443 struct nfs_inode *nfsi;
520f301c 444 struct address_space *mapping = folio->mapping;
f919b196
MG
445
446 /*
520f301c
MWO
447 * Check if an unstable folio is currently being committed and
448 * if so, have the VM treat it as if the folio is under writeback
449 * so it will not block due to folios that will shortly be freeable.
f919b196
MG
450 */
451 nfsi = NFS_I(mapping->host);
af7cf057 452 if (atomic_read(&nfsi->commit_info.rpcs_out)) {
f919b196
MG
453 *writeback = true;
454 return;
455 }
456
457 /*
520f301c
MWO
458 * If the private flag is set, then the folio is not freeable
459 * and as the inode is not being committed, it's not going to
460 * be cleaned in the near future so treat it as dirty
f919b196 461 */
520f301c 462 if (folio_test_private(folio))
f919b196
MG
463 *dirty = true;
464}
465
6b9b3514
DH
466/*
467 * Attempt to clear the private state associated with a page when an error
468 * occurs that requires the cached contents of an inode to be written back or
469 * destroyed
545db45f 470 * - Called if either PG_private or fscache is set on the page
6b9b3514
DH
471 * - Caller holds page lock
472 * - Return 0 if successful, -error otherwise
473 */
15a30ab2 474static int nfs_launder_folio(struct folio *folio)
e3db7691 475{
15a30ab2 476 struct inode *inode = folio->mapping->host;
eb5654b3 477 int ret;
b7eaefaa 478
15a30ab2
MWO
479 dfprintk(PAGECACHE, "NFS: launder_folio(%ld, %llu)\n",
480 inode->i_ino, folio_pos(folio));
b7eaefaa 481
15a30ab2 482 folio_wait_fscache(folio);
eb5654b3
TM
483 ret = nfs_wb_folio(inode, folio);
484 trace_nfs_launder_folio_done(inode, folio, ret);
485 return ret;
cd52ed35
TM
486}
487
a564b8f0
MG
488static int nfs_swap_activate(struct swap_info_struct *sis, struct file *file,
489 sector_t *span)
490{
bd89bc67
MZ
491 unsigned long blocks;
492 long long isize;
4b60c0ff 493 int ret;
4dc73c67
N
494 struct inode *inode = file_inode(file);
495 struct rpc_clnt *clnt = NFS_CLIENT(inode);
496 struct nfs_client *cl = NFS_SERVER(inode)->nfs_client;
bd89bc67
MZ
497
498 spin_lock(&inode->i_lock);
499 blocks = inode->i_blocks;
500 isize = inode->i_size;
501 spin_unlock(&inode->i_lock);
502 if (blocks*512 < isize) {
503 pr_warn("swap activate: swapfile has holes\n");
504 return -EINVAL;
505 }
dad2b015 506
4b60c0ff
N
507 ret = rpc_clnt_swap_activate(clnt);
508 if (ret)
509 return ret;
510 ret = add_swap_extent(sis, 0, sis->max, 0);
511 if (ret < 0) {
512 rpc_clnt_swap_deactivate(clnt);
513 return ret;
514 }
dad2b015 515
4b60c0ff 516 *span = sis->pages;
4dc73c67
N
517
518 if (cl->rpc_ops->enable_swap)
519 cl->rpc_ops->enable_swap(inode);
520
4b60c0ff
N
521 sis->flags |= SWP_FS_OPS;
522 return ret;
a564b8f0
MG
523}
524
525static void nfs_swap_deactivate(struct file *file)
526{
4dc73c67
N
527 struct inode *inode = file_inode(file);
528 struct rpc_clnt *clnt = NFS_CLIENT(inode);
529 struct nfs_client *cl = NFS_SERVER(inode)->nfs_client;
dad2b015 530
3c87ef6e 531 rpc_clnt_swap_deactivate(clnt);
4dc73c67
N
532 if (cl->rpc_ops->disable_swap)
533 cl->rpc_ops->disable_swap(file_inode(file));
a564b8f0 534}
a564b8f0 535
f5e54d6e 536const struct address_space_operations nfs_file_aops = {
65d023af 537 .read_folio = nfs_read_folio,
8786fde8 538 .readahead = nfs_readahead,
187c82cb 539 .dirty_folio = filemap_dirty_folio,
1da177e4
LT
540 .writepage = nfs_writepage,
541 .writepages = nfs_writepages,
4899f9c8
NP
542 .write_begin = nfs_write_begin,
543 .write_end = nfs_write_end,
6d740c76 544 .invalidate_folio = nfs_invalidate_folio,
3577da4a 545 .release_folio = nfs_release_folio,
4ae84a80 546 .migrate_folio = nfs_migrate_folio,
15a30ab2 547 .launder_folio = nfs_launder_folio,
f919b196 548 .is_dirty_writeback = nfs_check_dirty_writeback,
f590f333 549 .error_remove_page = generic_error_remove_page,
a564b8f0
MG
550 .swap_activate = nfs_swap_activate,
551 .swap_deactivate = nfs_swap_deactivate,
eb79f3af 552 .swap_rw = nfs_swap_rw,
1da177e4
LT
553};
554
6b9b3514
DH
555/*
556 * Notification that a PTE pointing to an NFS page is about to be made
557 * writable, implying that someone is about to modify the page through a
558 * shared-writable mapping
559 */
01a36844 560static vm_fault_t nfs_vm_page_mkwrite(struct vm_fault *vmf)
94387fb1 561{
11bac800 562 struct file *filp = vmf->vma->vm_file;
6de1472f 563 struct inode *inode = file_inode(filp);
94387fb1 564 unsigned pagelen;
01a36844 565 vm_fault_t ret = VM_FAULT_NOPAGE;
4899f9c8 566 struct address_space *mapping;
4fa7a717 567 struct folio *folio = page_folio(vmf->page);
94387fb1 568
1e8968c5 569 dfprintk(PAGECACHE, "NFS: vm_page_mkwrite(%pD2(%lu), offset %lld)\n",
4fa7a717
TM
570 filp, filp->f_mapping->host->i_ino,
571 (long long)folio_file_pos(folio));
b7eaefaa 572
9a773e7c
TM
573 sb_start_pagefault(inode->i_sb);
574
545db45f 575 /* make sure the cache has finished storing the page */
4fa7a717
TM
576 if (folio_test_fscache(folio) &&
577 folio_wait_fscache_killable(folio) < 0) {
a6b5a28e
DW
578 ret = VM_FAULT_RETRY;
579 goto out;
580 }
545db45f 581
ef070dcb 582 wait_on_bit_action(&NFS_I(inode)->flags, NFS_INO_INVALIDATING,
f5d39b02
PZ
583 nfs_wait_bit_killable,
584 TASK_KILLABLE|TASK_FREEZABLE_UNSAFE);
ef070dcb 585
4fa7a717
TM
586 folio_lock(folio);
587 mapping = folio_file_mapping(folio);
6de1472f 588 if (mapping != inode->i_mapping)
8b1f9ee5
TM
589 goto out_unlock;
590
4fa7a717 591 folio_wait_writeback(folio);
2aeb98f4 592
4fa7a717 593 pagelen = nfs_folio_length(folio);
8b1f9ee5
TM
594 if (pagelen == 0)
595 goto out_unlock;
4899f9c8 596
bc4866b6 597 ret = VM_FAULT_LOCKED;
0c493b5c
TM
598 if (nfs_flush_incompatible(filp, folio) == 0 &&
599 nfs_update_folio(filp, folio, 0, pagelen) == 0)
bc4866b6 600 goto out;
8b1f9ee5 601
bc4866b6 602 ret = VM_FAULT_SIGBUS;
8b1f9ee5 603out_unlock:
4fa7a717 604 folio_unlock(folio);
bc4866b6 605out:
9a773e7c 606 sb_end_pagefault(inode->i_sb);
bc4866b6 607 return ret;
94387fb1
TM
608}
609
f0f37e2f 610static const struct vm_operations_struct nfs_file_vm_ops = {
94387fb1 611 .fault = filemap_fault,
f1820361 612 .map_pages = filemap_map_pages,
94387fb1
TM
613 .page_mkwrite = nfs_vm_page_mkwrite,
614};
615
edaf4369 616ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from)
1da177e4 617{
6de1472f
AV
618 struct file *file = iocb->ki_filp;
619 struct inode *inode = file_inode(file);
ed7bcdb3
TM
620 unsigned int mntflags = NFS_SERVER(inode)->flags;
621 ssize_t result, written;
ce368536
SM
622 errseq_t since;
623 int error;
1da177e4 624
6de1472f 625 result = nfs_key_timeout_notify(file, inode);
dc24826b
AA
626 if (result)
627 return result;
628
89698b24 629 if (iocb->ki_flags & IOCB_DIRECT)
64158668 630 return nfs_file_direct_write(iocb, from, false);
1da177e4 631
619d30b4 632 dprintk("NFS: write(%pD2, %zu@%Ld)\n",
18290650 633 file, iov_iter_count(from), (long long) iocb->ki_pos);
1da177e4 634
1da177e4
LT
635 if (IS_SWAPFILE(inode))
636 goto out_swapfile;
7d52e862
TM
637 /*
638 * O_APPEND implies that we must revalidate the file length.
639 */
fc9dc401 640 if (iocb->ki_flags & IOCB_APPEND || iocb->ki_pos > i_size_read(inode)) {
6de1472f 641 result = nfs_revalidate_file_size(inode, file);
7d52e862 642 if (result)
e6005436 643 return result;
fe51beec 644 }
1da177e4 645
28aa2f9e
TM
646 nfs_clear_invalid_mapping(file->f_mapping);
647
ce368536 648 since = filemap_sample_wb_err(file->f_mapping);
a5864c99 649 nfs_start_io_write(inode);
18290650
TM
650 result = generic_write_checks(iocb, from);
651 if (result > 0) {
652 current->backing_dev_info = inode_to_bdi(inode);
800ba295 653 result = generic_perform_write(iocb, from);
18290650
TM
654 current->backing_dev_info = NULL;
655 }
a5864c99 656 nfs_end_io_write(inode);
18290650 657 if (result <= 0)
1da177e4
LT
658 goto out;
659
c49edecd 660 written = result;
18290650 661 iocb->ki_pos += written;
e6005436 662 nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written);
ed7bcdb3
TM
663
664 if (mntflags & NFS_MOUNT_WRITE_EAGER) {
665 result = filemap_fdatawrite_range(file->f_mapping,
666 iocb->ki_pos - written,
667 iocb->ki_pos - 1);
668 if (result < 0)
669 goto out;
670 }
671 if (mntflags & NFS_MOUNT_WRITE_WAIT) {
384edeb4
LB
672 filemap_fdatawait_range(file->f_mapping,
673 iocb->ki_pos - written,
674 iocb->ki_pos - 1);
ed7bcdb3 675 }
e973b1a5 676 result = generic_write_sync(iocb, written);
677 if (result < 0)
e6005436 678 return result;
7e381172 679
e6005436 680out:
7e94d6c4 681 /* Return error values */
ce368536 682 error = filemap_check_wb_err(file->f_mapping, since);
e6005436
TM
683 switch (error) {
684 default:
685 break;
686 case -EDQUOT:
687 case -EFBIG:
688 case -ENOSPC:
689 nfs_wb_all(inode);
690 error = file_check_and_advance_wb_err(file);
691 if (error < 0)
692 result = error;
200baa21 693 }
1da177e4
LT
694 return result;
695
696out_swapfile:
697 printk(KERN_INFO "NFS: attempt to write to active swap file!\n");
89658c4d 698 return -ETXTBSY;
1da177e4 699}
89d77c8f 700EXPORT_SYMBOL_GPL(nfs_file_write);
1da177e4 701
5eebde23
SJ
702static int
703do_getlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
1da177e4
LT
704{
705 struct inode *inode = filp->f_mapping->host;
706 int status = 0;
21ac19d4 707 unsigned int saved_type = fl->fl_type;
1da177e4 708
039c4d7a 709 /* Try local locking first */
6d34ac19
BF
710 posix_test_lock(filp, fl);
711 if (fl->fl_type != F_UNLCK) {
712 /* found a conflict */
039c4d7a 713 goto out;
1da177e4 714 }
21ac19d4 715 fl->fl_type = saved_type;
039c4d7a 716
011e2a7f 717 if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ))
039c4d7a
TM
718 goto out_noconflict;
719
5eebde23 720 if (is_local)
039c4d7a
TM
721 goto out_noconflict;
722
723 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
724out:
1da177e4 725 return status;
039c4d7a
TM
726out_noconflict:
727 fl->fl_type = F_UNLCK;
728 goto out;
1da177e4
LT
729}
730
5eebde23
SJ
731static int
732do_unlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
1da177e4
LT
733{
734 struct inode *inode = filp->f_mapping->host;
7a8203d8 735 struct nfs_lock_context *l_ctx;
1da177e4
LT
736 int status;
737
1da177e4
LT
738 /*
739 * Flush all pending writes before doing anything
740 * with locks..
741 */
aded8d7b 742 nfs_wb_all(inode);
1da177e4 743
7a8203d8
TM
744 l_ctx = nfs_get_lock_context(nfs_file_open_context(filp));
745 if (!IS_ERR(l_ctx)) {
210c7c17 746 status = nfs_iocounter_wait(l_ctx);
7a8203d8 747 nfs_put_lock_context(l_ctx);
f30cb757
BC
748 /* NOTE: special case
749 * If we're signalled while cleaning up locks on process exit, we
750 * still need to complete the unlock.
751 */
752 if (status < 0 && !(fl->fl_flags & FL_CLOSE))
7a8203d8
TM
753 return status;
754 }
755
5eebde23
SJ
756 /*
757 * Use local locking if mounted with "-onolock" or with appropriate
758 * "-olocal_lock="
759 */
760 if (!is_local)
1da177e4
LT
761 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
762 else
75575ddf 763 status = locks_lock_file_wait(filp, fl);
1da177e4
LT
764 return status;
765}
766
5eebde23
SJ
767static int
768do_setlk(struct file *filp, int cmd, struct file_lock *fl, int is_local)
1da177e4
LT
769{
770 struct inode *inode = filp->f_mapping->host;
1da177e4
LT
771 int status;
772
1da177e4
LT
773 /*
774 * Flush all pending writes before doing anything
775 * with locks..
776 */
29884df0
TM
777 status = nfs_sync_mapping(filp->f_mapping);
778 if (status != 0)
1da177e4
LT
779 goto out;
780
5eebde23
SJ
781 /*
782 * Use local locking if mounted with "-onolock" or with appropriate
783 * "-olocal_lock="
784 */
785 if (!is_local)
1da177e4 786 status = NFS_PROTO(inode)->lock(filp, cmd, fl);
c4d7c402 787 else
75575ddf 788 status = locks_lock_file_wait(filp, fl);
1da177e4
LT
789 if (status < 0)
790 goto out;
6b96724e 791
1da177e4 792 /*
779eafab
N
793 * Invalidate cache to prevent missing any changes. If
794 * the file is mapped, clear the page cache as well so
795 * those mappings will be loaded.
6b96724e 796 *
1da177e4
LT
797 * This makes locking act as a cache coherency point.
798 */
29884df0 799 nfs_sync_mapping(filp->f_mapping);
779eafab 800 if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) {
442ce049 801 nfs_zap_caches(inode);
779eafab
N
802 if (mapping_mapped(filp->f_mapping))
803 nfs_revalidate_mapping(inode, filp->f_mapping);
804 }
1da177e4 805out:
1da177e4
LT
806 return status;
807}
808
809/*
810 * Lock a (portion of) a file
811 */
ce4ef7c0 812int nfs_lock(struct file *filp, int cmd, struct file_lock *fl)
1da177e4 813{
6da24bc9 814 struct inode *inode = filp->f_mapping->host;
2116271a 815 int ret = -ENOLCK;
5eebde23 816 int is_local = 0;
1da177e4 817
6de1472f
AV
818 dprintk("NFS: lock(%pD2, t=%x, fl=%x, r=%lld:%lld)\n",
819 filp, fl->fl_type, fl->fl_flags,
1da177e4 820 (long long)fl->fl_start, (long long)fl->fl_end);
6da24bc9 821
91d5b470 822 nfs_inc_stats(inode, NFSIOS_VFSLOCK);
1da177e4 823
bb0a55bb
BF
824 if (fl->fl_flags & FL_RECLAIM)
825 return -ENOGRACE;
826
5eebde23
SJ
827 if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FCNTL)
828 is_local = 1;
829
2116271a
TM
830 if (NFS_PROTO(inode)->lock_check_bounds != NULL) {
831 ret = NFS_PROTO(inode)->lock_check_bounds(fl);
832 if (ret < 0)
833 goto out_err;
834 }
1da177e4
LT
835
836 if (IS_GETLK(cmd))
5eebde23 837 ret = do_getlk(filp, cmd, fl, is_local);
2116271a 838 else if (fl->fl_type == F_UNLCK)
5eebde23 839 ret = do_unlk(filp, cmd, fl, is_local);
2116271a 840 else
5eebde23 841 ret = do_setlk(filp, cmd, fl, is_local);
2116271a
TM
842out_err:
843 return ret;
1da177e4 844}
89d77c8f 845EXPORT_SYMBOL_GPL(nfs_lock);
1da177e4
LT
846
847/*
848 * Lock a (portion of) a file
849 */
ce4ef7c0 850int nfs_flock(struct file *filp, int cmd, struct file_lock *fl)
1da177e4 851{
5eebde23
SJ
852 struct inode *inode = filp->f_mapping->host;
853 int is_local = 0;
854
6de1472f
AV
855 dprintk("NFS: flock(%pD2, t=%x, fl=%x)\n",
856 filp, fl->fl_type, fl->fl_flags);
1da177e4 857
1da177e4
LT
858 if (!(fl->fl_flags & FL_FLOCK))
859 return -ENOLCK;
860
5eebde23
SJ
861 if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK)
862 is_local = 1;
863
fcfa4470
BC
864 /* We're simulating flock() locks using posix locks on the server */
865 if (fl->fl_type == F_UNLCK)
5eebde23
SJ
866 return do_unlk(filp, cmd, fl, is_local);
867 return do_setlk(filp, cmd, fl, is_local);
1da177e4 868}
89d77c8f 869EXPORT_SYMBOL_GPL(nfs_flock);
370f6599 870
0486958f
JL
871const struct file_operations nfs_file_operations = {
872 .llseek = nfs_file_llseek,
3aa2d199 873 .read_iter = nfs_file_read,
edaf4369 874 .write_iter = nfs_file_write,
0486958f
JL
875 .mmap = nfs_file_mmap,
876 .open = nfs_file_open,
877 .flush = nfs_file_flush,
878 .release = nfs_file_release,
879 .fsync = nfs_file_fsync,
880 .lock = nfs_lock,
881 .flock = nfs_flock,
82c156f8 882 .splice_read = generic_file_splice_read,
4da54c21 883 .splice_write = iter_file_splice_write,
0486958f 884 .check_flags = nfs_check_flags,
1c994a09 885 .setlease = simple_nosetlease,
0486958f 886};
ddda8e0a 887EXPORT_SYMBOL_GPL(nfs_file_operations);