Commit | Line | Data |
---|---|---|
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 | 45 | static const struct vm_operations_struct nfs_file_vm_ops; |
94387fb1 | 46 | |
ce4ef7c0 | 47 | int 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 | 54 | EXPORT_SYMBOL_GPL(nfs_check_flags); |
1da177e4 LT |
55 | |
56 | /* | |
57 | * Open file | |
58 | */ | |
59 | static int | |
60 | nfs_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 | 77 | int |
1da177e4 LT |
78 | nfs_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 | 87 | EXPORT_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 | */ | |
100 | static 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 |
109 | force_reval: |
110 | return __nfs_revalidate_inode(server, inode); | |
111 | } | |
112 | ||
965c8e59 | 113 | loff_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 | 132 | EXPORT_SYMBOL_GPL(nfs_file_llseek); |
980802e3 | 133 | |
1da177e4 LT |
134 | /* |
135 | * Flush all dirty pages, and check for write errors. | |
1da177e4 | 136 | */ |
5445b1fb | 137 | static int |
75e1fcc0 | 138 | nfs_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 | 155 | ssize_t |
3aa2d199 | 156 | nfs_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 | 178 | EXPORT_SYMBOL_GPL(nfs_file_read); |
1da177e4 | 179 | |
ce4ef7c0 | 180 | int |
1da177e4 LT |
181 | nfs_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 | 198 | EXPORT_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 | 205 | static int |
bf4b4905 | 206 | nfs_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 | 221 | int |
a5c58892 BS |
222 | nfs_file_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
223 | { | |
496ad9aa | 224 | struct inode *inode = file_inode(file); |
67f4b5dc TM |
225 | struct nfs_inode *nfsi = NFS_I(inode); |
226 | long save_nredirtied = atomic_long_read(&nfsi->redirtied_pages); | |
227 | long nredirtied; | |
2197e9b0 | 228 | int ret; |
a5c58892 | 229 | |
f4ce1299 TM |
230 | trace_nfs_fsync_enter(inode); |
231 | ||
2197e9b0 | 232 | for (;;) { |
6fbda89b | 233 | ret = file_write_and_wait_range(file, start, end); |
05990d1b TM |
234 | if (ret != 0) |
235 | break; | |
bf4b4905 | 236 | ret = nfs_file_fsync_commit(file, datasync); |
2197e9b0 TM |
237 | if (ret != 0) |
238 | break; | |
239 | ret = pnfs_sync_inode(inode, !!datasync); | |
240 | if (ret != 0) | |
241 | break; | |
67f4b5dc TM |
242 | nredirtied = atomic_long_read(&nfsi->redirtied_pages); |
243 | if (nredirtied == save_nredirtied) | |
2197e9b0 | 244 | break; |
67f4b5dc | 245 | save_nredirtied = nredirtied; |
2197e9b0 | 246 | } |
05990d1b | 247 | |
f4ce1299 | 248 | trace_nfs_fsync_exit(inode, ret); |
af7fa165 | 249 | return ret; |
1da177e4 | 250 | } |
4ff79bc7 | 251 | EXPORT_SYMBOL_GPL(nfs_file_fsync); |
1da177e4 | 252 | |
38c73044 PS |
253 | /* |
254 | * Decide whether a read/modify/write cycle may be more efficient | |
255 | * then a modify/write/read cycle when writing to a page in the | |
256 | * page cache. | |
257 | * | |
2cde04e9 KI |
258 | * Some pNFS layout drivers can only read/write at a certain block |
259 | * granularity like all block devices and therefore we must perform | |
260 | * read/modify/write whenever a page hasn't read yet and the data | |
261 | * to be written there is not aligned to a block boundary and/or | |
262 | * smaller than the block size. | |
263 | * | |
38c73044 PS |
264 | * The modify/write/read cycle may occur if a page is read before |
265 | * being completely filled by the writer. In this situation, the | |
266 | * page must be completely written to stable storage on the server | |
267 | * before it can be refilled by reading in the page from the server. | |
268 | * This can lead to expensive, small, FILE_SYNC mode writes being | |
269 | * done. | |
270 | * | |
271 | * It may be more efficient to read the page first if the file is | |
272 | * open for reading in addition to writing, the page is not marked | |
273 | * as Uptodate, it is not dirty or waiting to be committed, | |
274 | * indicating that it was previously allocated and then modified, | |
275 | * that there were valid bytes of data in that range of the file, | |
276 | * and that the new data won't completely replace the old data in | |
277 | * that range of the file. | |
278 | */ | |
2cde04e9 | 279 | static bool nfs_full_page_write(struct page *page, loff_t pos, unsigned int len) |
38c73044 PS |
280 | { |
281 | unsigned int pglen = nfs_page_length(page); | |
09cbfeaf | 282 | unsigned int offset = pos & (PAGE_SIZE - 1); |
38c73044 PS |
283 | unsigned int end = offset + len; |
284 | ||
2cde04e9 KI |
285 | return !pglen || (end >= pglen && !offset); |
286 | } | |
612aa983 | 287 | |
2cde04e9 KI |
288 | static bool nfs_want_read_modify_write(struct file *file, struct page *page, |
289 | loff_t pos, unsigned int len) | |
290 | { | |
291 | /* | |
292 | * Up-to-date pages, those with ongoing or full-page write | |
293 | * don't need read/modify/write | |
294 | */ | |
295 | if (PageUptodate(page) || PagePrivate(page) || | |
296 | nfs_full_page_write(page, pos, len)) | |
297 | return false; | |
298 | ||
299 | if (pnfs_ld_read_whole_page(file->f_mapping->host)) | |
300 | return true; | |
301 | /* Open for reading too? */ | |
302 | if (file->f_mode & FMODE_READ) | |
303 | return true; | |
304 | return false; | |
38c73044 PS |
305 | } |
306 | ||
1da177e4 | 307 | /* |
4899f9c8 NP |
308 | * This does the "real" work of the write. We must allocate and lock the |
309 | * page to be sent back to the generic routine, which then copies the | |
310 | * data from user space. | |
1da177e4 LT |
311 | * |
312 | * If the writer ends up delaying the write, the writer needs to | |
313 | * increment the page use counts until he is done with the page. | |
314 | */ | |
4899f9c8 | 315 | static int nfs_write_begin(struct file *file, struct address_space *mapping, |
9d6b0cd7 | 316 | loff_t pos, unsigned len, |
4899f9c8 | 317 | struct page **pagep, void **fsdata) |
1da177e4 | 318 | { |
4899f9c8 | 319 | int ret; |
09cbfeaf | 320 | pgoff_t index = pos >> PAGE_SHIFT; |
4899f9c8 | 321 | struct page *page; |
38c73044 | 322 | int once_thru = 0; |
4899f9c8 | 323 | |
1e8968c5 | 324 | dfprintk(PAGECACHE, "NFS: write_begin(%pD2(%lu), %u@%lld)\n", |
6de1472f | 325 | file, mapping->host->i_ino, len, (long long) pos); |
b7eaefaa | 326 | |
38c73044 | 327 | start: |
b7446e7c | 328 | page = grab_cache_page_write_begin(mapping, index); |
4899f9c8 NP |
329 | if (!page) |
330 | return -ENOMEM; | |
331 | *pagep = page; | |
332 | ||
333 | ret = nfs_flush_incompatible(file, page); | |
334 | if (ret) { | |
335 | unlock_page(page); | |
09cbfeaf | 336 | put_page(page); |
38c73044 PS |
337 | } else if (!once_thru && |
338 | nfs_want_read_modify_write(file, page, pos, len)) { | |
339 | once_thru = 1; | |
65d023af | 340 | ret = nfs_read_folio(file, page_folio(page)); |
09cbfeaf | 341 | put_page(page); |
38c73044 PS |
342 | if (!ret) |
343 | goto start; | |
4899f9c8 NP |
344 | } |
345 | return ret; | |
1da177e4 LT |
346 | } |
347 | ||
4899f9c8 NP |
348 | static int nfs_write_end(struct file *file, struct address_space *mapping, |
349 | loff_t pos, unsigned len, unsigned copied, | |
350 | struct page *page, void *fsdata) | |
1da177e4 | 351 | { |
09cbfeaf | 352 | unsigned offset = pos & (PAGE_SIZE - 1); |
dc24826b | 353 | struct nfs_open_context *ctx = nfs_file_open_context(file); |
4899f9c8 | 354 | int status; |
1da177e4 | 355 | |
1e8968c5 | 356 | dfprintk(PAGECACHE, "NFS: write_end(%pD2(%lu), %u@%lld)\n", |
6de1472f | 357 | file, mapping->host->i_ino, len, (long long) pos); |
b7eaefaa | 358 | |
efc91ed0 TM |
359 | /* |
360 | * Zero any uninitialised parts of the page, and then mark the page | |
361 | * as up to date if it turns out that we're extending the file. | |
362 | */ | |
363 | if (!PageUptodate(page)) { | |
364 | unsigned pglen = nfs_page_length(page); | |
c0cf3ef5 | 365 | unsigned end = offset + copied; |
efc91ed0 TM |
366 | |
367 | if (pglen == 0) { | |
368 | zero_user_segments(page, 0, offset, | |
09cbfeaf | 369 | end, PAGE_SIZE); |
efc91ed0 TM |
370 | SetPageUptodate(page); |
371 | } else if (end >= pglen) { | |
09cbfeaf | 372 | zero_user_segment(page, end, PAGE_SIZE); |
efc91ed0 TM |
373 | if (offset == 0) |
374 | SetPageUptodate(page); | |
375 | } else | |
09cbfeaf | 376 | zero_user_segment(page, pglen, PAGE_SIZE); |
efc91ed0 TM |
377 | } |
378 | ||
4899f9c8 | 379 | status = nfs_updatepage(file, page, offset, copied); |
4899f9c8 NP |
380 | |
381 | unlock_page(page); | |
09cbfeaf | 382 | put_page(page); |
4899f9c8 | 383 | |
3d509e54 CL |
384 | if (status < 0) |
385 | return status; | |
2701d086 | 386 | NFS_I(mapping->host)->write_io += copied; |
dc24826b | 387 | |
d95b2665 TM |
388 | if (nfs_ctx_key_to_expire(ctx, mapping->host)) |
389 | nfs_wb_all(mapping->host); | |
dc24826b | 390 | |
3d509e54 | 391 | return copied; |
1da177e4 LT |
392 | } |
393 | ||
6b9b3514 DH |
394 | /* |
395 | * Partially or wholly invalidate a page | |
396 | * - Release the private state associated with a page if undergoing complete | |
397 | * page invalidation | |
545db45f | 398 | * - Called if either PG_private or PG_fscache is set on the page |
6b9b3514 DH |
399 | * - Caller holds page lock |
400 | */ | |
6d740c76 MWO |
401 | static void nfs_invalidate_folio(struct folio *folio, size_t offset, |
402 | size_t length) | |
cd52ed35 | 403 | { |
6d740c76 MWO |
404 | dfprintk(PAGECACHE, "NFS: invalidate_folio(%lu, %zu, %zu)\n", |
405 | folio->index, offset, length); | |
b7eaefaa | 406 | |
6d740c76 | 407 | if (offset != 0 || length < folio_size(folio)) |
1c75950b | 408 | return; |
d2ccddf0 | 409 | /* Cancel any unstarted writes on this page */ |
6d740c76 MWO |
410 | nfs_wb_folio_cancel(folio->mapping->host, folio); |
411 | folio_wait_fscache(folio); | |
cd52ed35 TM |
412 | } |
413 | ||
6b9b3514 | 414 | /* |
3577da4a MWO |
415 | * Attempt to release the private state associated with a folio |
416 | * - Called if either private or fscache flags are set on the folio | |
417 | * - Caller holds folio lock | |
418 | * - Return true (may release folio) or false (may not) | |
6b9b3514 | 419 | */ |
3577da4a | 420 | static bool nfs_release_folio(struct folio *folio, gfp_t gfp) |
cd52ed35 | 421 | { |
3577da4a | 422 | dfprintk(PAGECACHE, "NFS: release_folio(%p)\n", folio); |
b7eaefaa | 423 | |
3577da4a MWO |
424 | /* If the private flag is set, then the folio is not freeable */ |
425 | if (folio_test_private(folio)) | |
426 | return false; | |
427 | return nfs_fscache_release_folio(folio, gfp); | |
e3db7691 TM |
428 | } |
429 | ||
520f301c | 430 | static void nfs_check_dirty_writeback(struct folio *folio, |
f919b196 MG |
431 | bool *dirty, bool *writeback) |
432 | { | |
433 | struct nfs_inode *nfsi; | |
520f301c | 434 | struct address_space *mapping = folio->mapping; |
f919b196 MG |
435 | |
436 | /* | |
520f301c MWO |
437 | * Check if an unstable folio is currently being committed and |
438 | * if so, have the VM treat it as if the folio is under writeback | |
439 | * so it will not block due to folios that will shortly be freeable. | |
f919b196 MG |
440 | */ |
441 | nfsi = NFS_I(mapping->host); | |
af7cf057 | 442 | if (atomic_read(&nfsi->commit_info.rpcs_out)) { |
f919b196 MG |
443 | *writeback = true; |
444 | return; | |
445 | } | |
446 | ||
447 | /* | |
520f301c MWO |
448 | * If the private flag is set, then the folio is not freeable |
449 | * and as the inode is not being committed, it's not going to | |
450 | * be cleaned in the near future so treat it as dirty | |
f919b196 | 451 | */ |
520f301c | 452 | if (folio_test_private(folio)) |
f919b196 MG |
453 | *dirty = true; |
454 | } | |
455 | ||
6b9b3514 DH |
456 | /* |
457 | * Attempt to clear the private state associated with a page when an error | |
458 | * occurs that requires the cached contents of an inode to be written back or | |
459 | * destroyed | |
545db45f | 460 | * - Called if either PG_private or fscache is set on the page |
6b9b3514 DH |
461 | * - Caller holds page lock |
462 | * - Return 0 if successful, -error otherwise | |
463 | */ | |
15a30ab2 | 464 | static int nfs_launder_folio(struct folio *folio) |
e3db7691 | 465 | { |
15a30ab2 | 466 | struct inode *inode = folio->mapping->host; |
b7eaefaa | 467 | |
15a30ab2 MWO |
468 | dfprintk(PAGECACHE, "NFS: launder_folio(%ld, %llu)\n", |
469 | inode->i_ino, folio_pos(folio)); | |
b7eaefaa | 470 | |
15a30ab2 MWO |
471 | folio_wait_fscache(folio); |
472 | return nfs_wb_page(inode, &folio->page); | |
cd52ed35 TM |
473 | } |
474 | ||
a564b8f0 MG |
475 | static int nfs_swap_activate(struct swap_info_struct *sis, struct file *file, |
476 | sector_t *span) | |
477 | { | |
bd89bc67 MZ |
478 | unsigned long blocks; |
479 | long long isize; | |
4b60c0ff | 480 | int ret; |
4dc73c67 N |
481 | struct inode *inode = file_inode(file); |
482 | struct rpc_clnt *clnt = NFS_CLIENT(inode); | |
483 | struct nfs_client *cl = NFS_SERVER(inode)->nfs_client; | |
bd89bc67 MZ |
484 | |
485 | spin_lock(&inode->i_lock); | |
486 | blocks = inode->i_blocks; | |
487 | isize = inode->i_size; | |
488 | spin_unlock(&inode->i_lock); | |
489 | if (blocks*512 < isize) { | |
490 | pr_warn("swap activate: swapfile has holes\n"); | |
491 | return -EINVAL; | |
492 | } | |
dad2b015 | 493 | |
4b60c0ff N |
494 | ret = rpc_clnt_swap_activate(clnt); |
495 | if (ret) | |
496 | return ret; | |
497 | ret = add_swap_extent(sis, 0, sis->max, 0); | |
498 | if (ret < 0) { | |
499 | rpc_clnt_swap_deactivate(clnt); | |
500 | return ret; | |
501 | } | |
dad2b015 | 502 | |
4b60c0ff | 503 | *span = sis->pages; |
4dc73c67 N |
504 | |
505 | if (cl->rpc_ops->enable_swap) | |
506 | cl->rpc_ops->enable_swap(inode); | |
507 | ||
4b60c0ff N |
508 | sis->flags |= SWP_FS_OPS; |
509 | return ret; | |
a564b8f0 MG |
510 | } |
511 | ||
512 | static void nfs_swap_deactivate(struct file *file) | |
513 | { | |
4dc73c67 N |
514 | struct inode *inode = file_inode(file); |
515 | struct rpc_clnt *clnt = NFS_CLIENT(inode); | |
516 | struct nfs_client *cl = NFS_SERVER(inode)->nfs_client; | |
dad2b015 | 517 | |
3c87ef6e | 518 | rpc_clnt_swap_deactivate(clnt); |
4dc73c67 N |
519 | if (cl->rpc_ops->disable_swap) |
520 | cl->rpc_ops->disable_swap(file_inode(file)); | |
a564b8f0 | 521 | } |
a564b8f0 | 522 | |
f5e54d6e | 523 | const struct address_space_operations nfs_file_aops = { |
65d023af | 524 | .read_folio = nfs_read_folio, |
8786fde8 | 525 | .readahead = nfs_readahead, |
187c82cb | 526 | .dirty_folio = filemap_dirty_folio, |
1da177e4 LT |
527 | .writepage = nfs_writepage, |
528 | .writepages = nfs_writepages, | |
4899f9c8 NP |
529 | .write_begin = nfs_write_begin, |
530 | .write_end = nfs_write_end, | |
6d740c76 | 531 | .invalidate_folio = nfs_invalidate_folio, |
3577da4a | 532 | .release_folio = nfs_release_folio, |
f844cd0d | 533 | #ifdef CONFIG_MIGRATION |
074cc1de | 534 | .migratepage = nfs_migrate_page, |
f844cd0d | 535 | #endif |
15a30ab2 | 536 | .launder_folio = nfs_launder_folio, |
f919b196 | 537 | .is_dirty_writeback = nfs_check_dirty_writeback, |
f590f333 | 538 | .error_remove_page = generic_error_remove_page, |
a564b8f0 MG |
539 | .swap_activate = nfs_swap_activate, |
540 | .swap_deactivate = nfs_swap_deactivate, | |
eb79f3af | 541 | .swap_rw = nfs_swap_rw, |
1da177e4 LT |
542 | }; |
543 | ||
6b9b3514 DH |
544 | /* |
545 | * Notification that a PTE pointing to an NFS page is about to be made | |
546 | * writable, implying that someone is about to modify the page through a | |
547 | * shared-writable mapping | |
548 | */ | |
01a36844 | 549 | static vm_fault_t nfs_vm_page_mkwrite(struct vm_fault *vmf) |
94387fb1 | 550 | { |
c2ec175c | 551 | struct page *page = vmf->page; |
11bac800 | 552 | struct file *filp = vmf->vma->vm_file; |
6de1472f | 553 | struct inode *inode = file_inode(filp); |
94387fb1 | 554 | unsigned pagelen; |
01a36844 | 555 | vm_fault_t ret = VM_FAULT_NOPAGE; |
4899f9c8 | 556 | struct address_space *mapping; |
94387fb1 | 557 | |
1e8968c5 | 558 | dfprintk(PAGECACHE, "NFS: vm_page_mkwrite(%pD2(%lu), offset %lld)\n", |
6de1472f | 559 | filp, filp->f_mapping->host->i_ino, |
b7eaefaa CL |
560 | (long long)page_offset(page)); |
561 | ||
9a773e7c TM |
562 | sb_start_pagefault(inode->i_sb); |
563 | ||
545db45f | 564 | /* make sure the cache has finished storing the page */ |
a6b5a28e DW |
565 | if (PageFsCache(page) && |
566 | wait_on_page_fscache_killable(vmf->page) < 0) { | |
567 | ret = VM_FAULT_RETRY; | |
568 | goto out; | |
569 | } | |
545db45f | 570 | |
ef070dcb TM |
571 | wait_on_bit_action(&NFS_I(inode)->flags, NFS_INO_INVALIDATING, |
572 | nfs_wait_bit_killable, TASK_KILLABLE); | |
573 | ||
94387fb1 | 574 | lock_page(page); |
d56b4ddf | 575 | mapping = page_file_mapping(page); |
6de1472f | 576 | if (mapping != inode->i_mapping) |
8b1f9ee5 TM |
577 | goto out_unlock; |
578 | ||
2aeb98f4 TM |
579 | wait_on_page_writeback(page); |
580 | ||
94387fb1 | 581 | pagelen = nfs_page_length(page); |
8b1f9ee5 TM |
582 | if (pagelen == 0) |
583 | goto out_unlock; | |
4899f9c8 | 584 | |
bc4866b6 TM |
585 | ret = VM_FAULT_LOCKED; |
586 | if (nfs_flush_incompatible(filp, page) == 0 && | |
587 | nfs_updatepage(filp, page, 0, pagelen) == 0) | |
588 | goto out; | |
8b1f9ee5 | 589 | |
bc4866b6 | 590 | ret = VM_FAULT_SIGBUS; |
8b1f9ee5 TM |
591 | out_unlock: |
592 | unlock_page(page); | |
bc4866b6 | 593 | out: |
9a773e7c | 594 | sb_end_pagefault(inode->i_sb); |
bc4866b6 | 595 | return ret; |
94387fb1 TM |
596 | } |
597 | ||
f0f37e2f | 598 | static const struct vm_operations_struct nfs_file_vm_ops = { |
94387fb1 | 599 | .fault = filemap_fault, |
f1820361 | 600 | .map_pages = filemap_map_pages, |
94387fb1 TM |
601 | .page_mkwrite = nfs_vm_page_mkwrite, |
602 | }; | |
603 | ||
edaf4369 | 604 | ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 605 | { |
6de1472f AV |
606 | struct file *file = iocb->ki_filp; |
607 | struct inode *inode = file_inode(file); | |
ed7bcdb3 TM |
608 | unsigned int mntflags = NFS_SERVER(inode)->flags; |
609 | ssize_t result, written; | |
ce368536 SM |
610 | errseq_t since; |
611 | int error; | |
1da177e4 | 612 | |
6de1472f | 613 | result = nfs_key_timeout_notify(file, inode); |
dc24826b AA |
614 | if (result) |
615 | return result; | |
616 | ||
89698b24 | 617 | if (iocb->ki_flags & IOCB_DIRECT) |
64158668 | 618 | return nfs_file_direct_write(iocb, from, false); |
1da177e4 | 619 | |
619d30b4 | 620 | dprintk("NFS: write(%pD2, %zu@%Ld)\n", |
18290650 | 621 | file, iov_iter_count(from), (long long) iocb->ki_pos); |
1da177e4 | 622 | |
1da177e4 LT |
623 | if (IS_SWAPFILE(inode)) |
624 | goto out_swapfile; | |
7d52e862 TM |
625 | /* |
626 | * O_APPEND implies that we must revalidate the file length. | |
627 | */ | |
fc9dc401 | 628 | if (iocb->ki_flags & IOCB_APPEND || iocb->ki_pos > i_size_read(inode)) { |
6de1472f | 629 | result = nfs_revalidate_file_size(inode, file); |
7d52e862 | 630 | if (result) |
e6005436 | 631 | return result; |
fe51beec | 632 | } |
1da177e4 | 633 | |
28aa2f9e TM |
634 | nfs_clear_invalid_mapping(file->f_mapping); |
635 | ||
ce368536 | 636 | since = filemap_sample_wb_err(file->f_mapping); |
a5864c99 | 637 | nfs_start_io_write(inode); |
18290650 TM |
638 | result = generic_write_checks(iocb, from); |
639 | if (result > 0) { | |
640 | current->backing_dev_info = inode_to_bdi(inode); | |
800ba295 | 641 | result = generic_perform_write(iocb, from); |
18290650 TM |
642 | current->backing_dev_info = NULL; |
643 | } | |
a5864c99 | 644 | nfs_end_io_write(inode); |
18290650 | 645 | if (result <= 0) |
1da177e4 LT |
646 | goto out; |
647 | ||
c49edecd | 648 | written = result; |
18290650 | 649 | iocb->ki_pos += written; |
e6005436 | 650 | nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written); |
ed7bcdb3 TM |
651 | |
652 | if (mntflags & NFS_MOUNT_WRITE_EAGER) { | |
653 | result = filemap_fdatawrite_range(file->f_mapping, | |
654 | iocb->ki_pos - written, | |
655 | iocb->ki_pos - 1); | |
656 | if (result < 0) | |
657 | goto out; | |
658 | } | |
659 | if (mntflags & NFS_MOUNT_WRITE_WAIT) { | |
660 | result = filemap_fdatawait_range(file->f_mapping, | |
661 | iocb->ki_pos - written, | |
662 | iocb->ki_pos - 1); | |
ed7bcdb3 | 663 | } |
e973b1a5 | 664 | result = generic_write_sync(iocb, written); |
665 | if (result < 0) | |
e6005436 | 666 | return result; |
7e381172 | 667 | |
e6005436 | 668 | out: |
7e94d6c4 | 669 | /* Return error values */ |
ce368536 | 670 | error = filemap_check_wb_err(file->f_mapping, since); |
e6005436 TM |
671 | switch (error) { |
672 | default: | |
673 | break; | |
674 | case -EDQUOT: | |
675 | case -EFBIG: | |
676 | case -ENOSPC: | |
677 | nfs_wb_all(inode); | |
678 | error = file_check_and_advance_wb_err(file); | |
679 | if (error < 0) | |
680 | result = error; | |
200baa21 | 681 | } |
1da177e4 LT |
682 | return result; |
683 | ||
684 | out_swapfile: | |
685 | printk(KERN_INFO "NFS: attempt to write to active swap file!\n"); | |
89658c4d | 686 | return -ETXTBSY; |
1da177e4 | 687 | } |
89d77c8f | 688 | EXPORT_SYMBOL_GPL(nfs_file_write); |
1da177e4 | 689 | |
5eebde23 SJ |
690 | static int |
691 | do_getlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
692 | { |
693 | struct inode *inode = filp->f_mapping->host; | |
694 | int status = 0; | |
21ac19d4 | 695 | unsigned int saved_type = fl->fl_type; |
1da177e4 | 696 | |
039c4d7a | 697 | /* Try local locking first */ |
6d34ac19 BF |
698 | posix_test_lock(filp, fl); |
699 | if (fl->fl_type != F_UNLCK) { | |
700 | /* found a conflict */ | |
039c4d7a | 701 | goto out; |
1da177e4 | 702 | } |
21ac19d4 | 703 | fl->fl_type = saved_type; |
039c4d7a | 704 | |
011e2a7f | 705 | if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) |
039c4d7a TM |
706 | goto out_noconflict; |
707 | ||
5eebde23 | 708 | if (is_local) |
039c4d7a TM |
709 | goto out_noconflict; |
710 | ||
711 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); | |
712 | out: | |
1da177e4 | 713 | return status; |
039c4d7a TM |
714 | out_noconflict: |
715 | fl->fl_type = F_UNLCK; | |
716 | goto out; | |
1da177e4 LT |
717 | } |
718 | ||
5eebde23 SJ |
719 | static int |
720 | do_unlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
721 | { |
722 | struct inode *inode = filp->f_mapping->host; | |
7a8203d8 | 723 | struct nfs_lock_context *l_ctx; |
1da177e4 LT |
724 | int status; |
725 | ||
1da177e4 LT |
726 | /* |
727 | * Flush all pending writes before doing anything | |
728 | * with locks.. | |
729 | */ | |
aded8d7b | 730 | nfs_wb_all(inode); |
1da177e4 | 731 | |
7a8203d8 TM |
732 | l_ctx = nfs_get_lock_context(nfs_file_open_context(filp)); |
733 | if (!IS_ERR(l_ctx)) { | |
210c7c17 | 734 | status = nfs_iocounter_wait(l_ctx); |
7a8203d8 | 735 | nfs_put_lock_context(l_ctx); |
f30cb757 BC |
736 | /* NOTE: special case |
737 | * If we're signalled while cleaning up locks on process exit, we | |
738 | * still need to complete the unlock. | |
739 | */ | |
740 | if (status < 0 && !(fl->fl_flags & FL_CLOSE)) | |
7a8203d8 TM |
741 | return status; |
742 | } | |
743 | ||
5eebde23 SJ |
744 | /* |
745 | * Use local locking if mounted with "-onolock" or with appropriate | |
746 | * "-olocal_lock=" | |
747 | */ | |
748 | if (!is_local) | |
1da177e4 LT |
749 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); |
750 | else | |
75575ddf | 751 | status = locks_lock_file_wait(filp, fl); |
1da177e4 LT |
752 | return status; |
753 | } | |
754 | ||
5eebde23 SJ |
755 | static int |
756 | do_setlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
757 | { |
758 | struct inode *inode = filp->f_mapping->host; | |
1da177e4 LT |
759 | int status; |
760 | ||
1da177e4 LT |
761 | /* |
762 | * Flush all pending writes before doing anything | |
763 | * with locks.. | |
764 | */ | |
29884df0 TM |
765 | status = nfs_sync_mapping(filp->f_mapping); |
766 | if (status != 0) | |
1da177e4 LT |
767 | goto out; |
768 | ||
5eebde23 SJ |
769 | /* |
770 | * Use local locking if mounted with "-onolock" or with appropriate | |
771 | * "-olocal_lock=" | |
772 | */ | |
773 | if (!is_local) | |
1da177e4 | 774 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); |
c4d7c402 | 775 | else |
75575ddf | 776 | status = locks_lock_file_wait(filp, fl); |
1da177e4 LT |
777 | if (status < 0) |
778 | goto out; | |
6b96724e | 779 | |
1da177e4 | 780 | /* |
779eafab N |
781 | * Invalidate cache to prevent missing any changes. If |
782 | * the file is mapped, clear the page cache as well so | |
783 | * those mappings will be loaded. | |
6b96724e | 784 | * |
1da177e4 LT |
785 | * This makes locking act as a cache coherency point. |
786 | */ | |
29884df0 | 787 | nfs_sync_mapping(filp->f_mapping); |
779eafab | 788 | if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) { |
442ce049 | 789 | nfs_zap_caches(inode); |
779eafab N |
790 | if (mapping_mapped(filp->f_mapping)) |
791 | nfs_revalidate_mapping(inode, filp->f_mapping); | |
792 | } | |
1da177e4 | 793 | out: |
1da177e4 LT |
794 | return status; |
795 | } | |
796 | ||
797 | /* | |
798 | * Lock a (portion of) a file | |
799 | */ | |
ce4ef7c0 | 800 | int nfs_lock(struct file *filp, int cmd, struct file_lock *fl) |
1da177e4 | 801 | { |
6da24bc9 | 802 | struct inode *inode = filp->f_mapping->host; |
2116271a | 803 | int ret = -ENOLCK; |
5eebde23 | 804 | int is_local = 0; |
1da177e4 | 805 | |
6de1472f AV |
806 | dprintk("NFS: lock(%pD2, t=%x, fl=%x, r=%lld:%lld)\n", |
807 | filp, fl->fl_type, fl->fl_flags, | |
1da177e4 | 808 | (long long)fl->fl_start, (long long)fl->fl_end); |
6da24bc9 | 809 | |
91d5b470 | 810 | nfs_inc_stats(inode, NFSIOS_VFSLOCK); |
1da177e4 | 811 | |
bb0a55bb BF |
812 | if (fl->fl_flags & FL_RECLAIM) |
813 | return -ENOGRACE; | |
814 | ||
5eebde23 SJ |
815 | if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FCNTL) |
816 | is_local = 1; | |
817 | ||
2116271a TM |
818 | if (NFS_PROTO(inode)->lock_check_bounds != NULL) { |
819 | ret = NFS_PROTO(inode)->lock_check_bounds(fl); | |
820 | if (ret < 0) | |
821 | goto out_err; | |
822 | } | |
1da177e4 LT |
823 | |
824 | if (IS_GETLK(cmd)) | |
5eebde23 | 825 | ret = do_getlk(filp, cmd, fl, is_local); |
2116271a | 826 | else if (fl->fl_type == F_UNLCK) |
5eebde23 | 827 | ret = do_unlk(filp, cmd, fl, is_local); |
2116271a | 828 | else |
5eebde23 | 829 | ret = do_setlk(filp, cmd, fl, is_local); |
2116271a TM |
830 | out_err: |
831 | return ret; | |
1da177e4 | 832 | } |
89d77c8f | 833 | EXPORT_SYMBOL_GPL(nfs_lock); |
1da177e4 LT |
834 | |
835 | /* | |
836 | * Lock a (portion of) a file | |
837 | */ | |
ce4ef7c0 | 838 | int nfs_flock(struct file *filp, int cmd, struct file_lock *fl) |
1da177e4 | 839 | { |
5eebde23 SJ |
840 | struct inode *inode = filp->f_mapping->host; |
841 | int is_local = 0; | |
842 | ||
6de1472f AV |
843 | dprintk("NFS: flock(%pD2, t=%x, fl=%x)\n", |
844 | filp, fl->fl_type, fl->fl_flags); | |
1da177e4 | 845 | |
1da177e4 LT |
846 | if (!(fl->fl_flags & FL_FLOCK)) |
847 | return -ENOLCK; | |
848 | ||
5eebde23 SJ |
849 | if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK) |
850 | is_local = 1; | |
851 | ||
fcfa4470 BC |
852 | /* We're simulating flock() locks using posix locks on the server */ |
853 | if (fl->fl_type == F_UNLCK) | |
5eebde23 SJ |
854 | return do_unlk(filp, cmd, fl, is_local); |
855 | return do_setlk(filp, cmd, fl, is_local); | |
1da177e4 | 856 | } |
89d77c8f | 857 | EXPORT_SYMBOL_GPL(nfs_flock); |
370f6599 | 858 | |
0486958f JL |
859 | const struct file_operations nfs_file_operations = { |
860 | .llseek = nfs_file_llseek, | |
3aa2d199 | 861 | .read_iter = nfs_file_read, |
edaf4369 | 862 | .write_iter = nfs_file_write, |
0486958f JL |
863 | .mmap = nfs_file_mmap, |
864 | .open = nfs_file_open, | |
865 | .flush = nfs_file_flush, | |
866 | .release = nfs_file_release, | |
867 | .fsync = nfs_file_fsync, | |
868 | .lock = nfs_lock, | |
869 | .flock = nfs_flock, | |
82c156f8 | 870 | .splice_read = generic_file_splice_read, |
4da54c21 | 871 | .splice_write = iter_file_splice_write, |
0486958f | 872 | .check_flags = nfs_check_flags, |
1c994a09 | 873 | .setlease = simple_nosetlease, |
0486958f | 874 | }; |
ddda8e0a | 875 | EXPORT_SYMBOL_GPL(nfs_file_operations); |