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 | */ | |
54d99381 TM |
279 | static bool nfs_folio_is_full_write(struct folio *folio, loff_t pos, |
280 | unsigned int len) | |
38c73044 | 281 | { |
54d99381 TM |
282 | unsigned int pglen = nfs_folio_length(folio); |
283 | unsigned int offset = offset_in_folio(folio, pos); | |
38c73044 PS |
284 | unsigned int end = offset + len; |
285 | ||
2cde04e9 KI |
286 | return !pglen || (end >= pglen && !offset); |
287 | } | |
612aa983 | 288 | |
54d99381 TM |
289 | static bool nfs_want_read_modify_write(struct file *file, struct folio *folio, |
290 | loff_t pos, unsigned int len) | |
2cde04e9 KI |
291 | { |
292 | /* | |
293 | * Up-to-date pages, those with ongoing or full-page write | |
294 | * don't need read/modify/write | |
295 | */ | |
54d99381 TM |
296 | if (folio_test_uptodate(folio) || folio_test_private(folio) || |
297 | nfs_folio_is_full_write(folio, pos, len)) | |
2cde04e9 KI |
298 | return false; |
299 | ||
54d99381 | 300 | if (pnfs_ld_read_whole_page(file_inode(file))) |
2cde04e9 KI |
301 | return true; |
302 | /* Open for reading too? */ | |
303 | if (file->f_mode & FMODE_READ) | |
304 | return true; | |
305 | return false; | |
38c73044 PS |
306 | } |
307 | ||
54d99381 TM |
308 | static struct folio * |
309 | nfs_folio_grab_cache_write_begin(struct address_space *mapping, pgoff_t index) | |
310 | { | |
311 | unsigned fgp_flags = FGP_LOCK | FGP_WRITE | FGP_CREAT | FGP_STABLE; | |
312 | ||
313 | return __filemap_get_folio(mapping, index, fgp_flags, | |
314 | mapping_gfp_mask(mapping)); | |
315 | } | |
316 | ||
1da177e4 | 317 | /* |
4899f9c8 NP |
318 | * This does the "real" work of the write. We must allocate and lock the |
319 | * page to be sent back to the generic routine, which then copies the | |
320 | * data from user space. | |
1da177e4 LT |
321 | * |
322 | * If the writer ends up delaying the write, the writer needs to | |
323 | * increment the page use counts until he is done with the page. | |
324 | */ | |
4899f9c8 | 325 | static int nfs_write_begin(struct file *file, struct address_space *mapping, |
54d99381 TM |
326 | loff_t pos, unsigned len, struct page **pagep, |
327 | void **fsdata) | |
1da177e4 | 328 | { |
0c493b5c | 329 | struct folio *folio; |
38c73044 | 330 | int once_thru = 0; |
54d99381 | 331 | int ret; |
4899f9c8 | 332 | |
1e8968c5 | 333 | dfprintk(PAGECACHE, "NFS: write_begin(%pD2(%lu), %u@%lld)\n", |
6de1472f | 334 | file, mapping->host->i_ino, len, (long long) pos); |
b7eaefaa | 335 | |
38c73044 | 336 | start: |
54d99381 TM |
337 | folio = nfs_folio_grab_cache_write_begin(mapping, pos >> PAGE_SHIFT); |
338 | if (!folio) | |
4899f9c8 | 339 | return -ENOMEM; |
54d99381 | 340 | *pagep = &folio->page; |
4899f9c8 | 341 | |
0c493b5c | 342 | ret = nfs_flush_incompatible(file, folio); |
4899f9c8 | 343 | if (ret) { |
54d99381 TM |
344 | folio_unlock(folio); |
345 | folio_put(folio); | |
38c73044 | 346 | } else if (!once_thru && |
54d99381 | 347 | nfs_want_read_modify_write(file, folio, pos, len)) { |
38c73044 | 348 | once_thru = 1; |
0c493b5c | 349 | ret = nfs_read_folio(file, folio); |
54d99381 | 350 | folio_put(folio); |
38c73044 PS |
351 | if (!ret) |
352 | goto start; | |
4899f9c8 NP |
353 | } |
354 | return ret; | |
1da177e4 LT |
355 | } |
356 | ||
4899f9c8 | 357 | static int nfs_write_end(struct file *file, struct address_space *mapping, |
54d99381 TM |
358 | loff_t pos, unsigned len, unsigned copied, |
359 | struct page *page, void *fsdata) | |
1da177e4 | 360 | { |
dc24826b | 361 | struct nfs_open_context *ctx = nfs_file_open_context(file); |
0c493b5c | 362 | struct folio *folio = page_folio(page); |
54d99381 | 363 | unsigned offset = offset_in_folio(folio, pos); |
4899f9c8 | 364 | int status; |
1da177e4 | 365 | |
1e8968c5 | 366 | dfprintk(PAGECACHE, "NFS: write_end(%pD2(%lu), %u@%lld)\n", |
6de1472f | 367 | file, mapping->host->i_ino, len, (long long) pos); |
b7eaefaa | 368 | |
efc91ed0 TM |
369 | /* |
370 | * Zero any uninitialised parts of the page, and then mark the page | |
371 | * as up to date if it turns out that we're extending the file. | |
372 | */ | |
54d99381 TM |
373 | if (!folio_test_uptodate(folio)) { |
374 | size_t fsize = folio_size(folio); | |
375 | unsigned pglen = nfs_folio_length(folio); | |
c0cf3ef5 | 376 | unsigned end = offset + copied; |
efc91ed0 TM |
377 | |
378 | if (pglen == 0) { | |
54d99381 TM |
379 | folio_zero_segments(folio, 0, offset, end, fsize); |
380 | folio_mark_uptodate(folio); | |
efc91ed0 | 381 | } else if (end >= pglen) { |
54d99381 | 382 | folio_zero_segment(folio, end, fsize); |
efc91ed0 | 383 | if (offset == 0) |
54d99381 | 384 | folio_mark_uptodate(folio); |
efc91ed0 | 385 | } else |
54d99381 | 386 | folio_zero_segment(folio, pglen, fsize); |
efc91ed0 TM |
387 | } |
388 | ||
0c493b5c | 389 | status = nfs_update_folio(file, folio, offset, copied); |
4899f9c8 | 390 | |
54d99381 TM |
391 | folio_unlock(folio); |
392 | folio_put(folio); | |
4899f9c8 | 393 | |
3d509e54 CL |
394 | if (status < 0) |
395 | return status; | |
2701d086 | 396 | NFS_I(mapping->host)->write_io += copied; |
dc24826b | 397 | |
d95b2665 TM |
398 | if (nfs_ctx_key_to_expire(ctx, mapping->host)) |
399 | nfs_wb_all(mapping->host); | |
dc24826b | 400 | |
3d509e54 | 401 | return copied; |
1da177e4 LT |
402 | } |
403 | ||
6b9b3514 DH |
404 | /* |
405 | * Partially or wholly invalidate a page | |
406 | * - Release the private state associated with a page if undergoing complete | |
407 | * page invalidation | |
545db45f | 408 | * - Called if either PG_private or PG_fscache is set on the page |
6b9b3514 DH |
409 | * - Caller holds page lock |
410 | */ | |
6d740c76 MWO |
411 | static void nfs_invalidate_folio(struct folio *folio, size_t offset, |
412 | size_t length) | |
cd52ed35 | 413 | { |
6d740c76 MWO |
414 | dfprintk(PAGECACHE, "NFS: invalidate_folio(%lu, %zu, %zu)\n", |
415 | folio->index, offset, length); | |
b7eaefaa | 416 | |
6d740c76 | 417 | if (offset != 0 || length < folio_size(folio)) |
1c75950b | 418 | return; |
d2ccddf0 | 419 | /* Cancel any unstarted writes on this page */ |
6d740c76 MWO |
420 | nfs_wb_folio_cancel(folio->mapping->host, folio); |
421 | folio_wait_fscache(folio); | |
cd52ed35 TM |
422 | } |
423 | ||
6b9b3514 | 424 | /* |
3577da4a MWO |
425 | * Attempt to release the private state associated with a folio |
426 | * - Called if either private or fscache flags are set on the folio | |
427 | * - Caller holds folio lock | |
428 | * - Return true (may release folio) or false (may not) | |
6b9b3514 | 429 | */ |
3577da4a | 430 | static bool nfs_release_folio(struct folio *folio, gfp_t gfp) |
cd52ed35 | 431 | { |
3577da4a | 432 | dfprintk(PAGECACHE, "NFS: release_folio(%p)\n", folio); |
b7eaefaa | 433 | |
3577da4a MWO |
434 | /* If the private flag is set, then the folio is not freeable */ |
435 | if (folio_test_private(folio)) | |
436 | return false; | |
437 | return nfs_fscache_release_folio(folio, gfp); | |
e3db7691 TM |
438 | } |
439 | ||
520f301c | 440 | static 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 | 474 | static int nfs_launder_folio(struct folio *folio) |
e3db7691 | 475 | { |
15a30ab2 | 476 | struct inode *inode = folio->mapping->host; |
b7eaefaa | 477 | |
15a30ab2 MWO |
478 | dfprintk(PAGECACHE, "NFS: launder_folio(%ld, %llu)\n", |
479 | inode->i_ino, folio_pos(folio)); | |
b7eaefaa | 480 | |
15a30ab2 | 481 | folio_wait_fscache(folio); |
4b27232a | 482 | return nfs_wb_folio(inode, folio); |
cd52ed35 TM |
483 | } |
484 | ||
a564b8f0 MG |
485 | static int nfs_swap_activate(struct swap_info_struct *sis, struct file *file, |
486 | sector_t *span) | |
487 | { | |
bd89bc67 MZ |
488 | unsigned long blocks; |
489 | long long isize; | |
4b60c0ff | 490 | int ret; |
4dc73c67 N |
491 | struct inode *inode = file_inode(file); |
492 | struct rpc_clnt *clnt = NFS_CLIENT(inode); | |
493 | struct nfs_client *cl = NFS_SERVER(inode)->nfs_client; | |
bd89bc67 MZ |
494 | |
495 | spin_lock(&inode->i_lock); | |
496 | blocks = inode->i_blocks; | |
497 | isize = inode->i_size; | |
498 | spin_unlock(&inode->i_lock); | |
499 | if (blocks*512 < isize) { | |
500 | pr_warn("swap activate: swapfile has holes\n"); | |
501 | return -EINVAL; | |
502 | } | |
dad2b015 | 503 | |
4b60c0ff N |
504 | ret = rpc_clnt_swap_activate(clnt); |
505 | if (ret) | |
506 | return ret; | |
507 | ret = add_swap_extent(sis, 0, sis->max, 0); | |
508 | if (ret < 0) { | |
509 | rpc_clnt_swap_deactivate(clnt); | |
510 | return ret; | |
511 | } | |
dad2b015 | 512 | |
4b60c0ff | 513 | *span = sis->pages; |
4dc73c67 N |
514 | |
515 | if (cl->rpc_ops->enable_swap) | |
516 | cl->rpc_ops->enable_swap(inode); | |
517 | ||
4b60c0ff N |
518 | sis->flags |= SWP_FS_OPS; |
519 | return ret; | |
a564b8f0 MG |
520 | } |
521 | ||
522 | static void nfs_swap_deactivate(struct file *file) | |
523 | { | |
4dc73c67 N |
524 | struct inode *inode = file_inode(file); |
525 | struct rpc_clnt *clnt = NFS_CLIENT(inode); | |
526 | struct nfs_client *cl = NFS_SERVER(inode)->nfs_client; | |
dad2b015 | 527 | |
3c87ef6e | 528 | rpc_clnt_swap_deactivate(clnt); |
4dc73c67 N |
529 | if (cl->rpc_ops->disable_swap) |
530 | cl->rpc_ops->disable_swap(file_inode(file)); | |
a564b8f0 | 531 | } |
a564b8f0 | 532 | |
f5e54d6e | 533 | const struct address_space_operations nfs_file_aops = { |
65d023af | 534 | .read_folio = nfs_read_folio, |
8786fde8 | 535 | .readahead = nfs_readahead, |
187c82cb | 536 | .dirty_folio = filemap_dirty_folio, |
1da177e4 LT |
537 | .writepage = nfs_writepage, |
538 | .writepages = nfs_writepages, | |
4899f9c8 NP |
539 | .write_begin = nfs_write_begin, |
540 | .write_end = nfs_write_end, | |
6d740c76 | 541 | .invalidate_folio = nfs_invalidate_folio, |
3577da4a | 542 | .release_folio = nfs_release_folio, |
4ae84a80 | 543 | .migrate_folio = nfs_migrate_folio, |
15a30ab2 | 544 | .launder_folio = nfs_launder_folio, |
f919b196 | 545 | .is_dirty_writeback = nfs_check_dirty_writeback, |
f590f333 | 546 | .error_remove_page = generic_error_remove_page, |
a564b8f0 MG |
547 | .swap_activate = nfs_swap_activate, |
548 | .swap_deactivate = nfs_swap_deactivate, | |
eb79f3af | 549 | .swap_rw = nfs_swap_rw, |
1da177e4 LT |
550 | }; |
551 | ||
6b9b3514 DH |
552 | /* |
553 | * Notification that a PTE pointing to an NFS page is about to be made | |
554 | * writable, implying that someone is about to modify the page through a | |
555 | * shared-writable mapping | |
556 | */ | |
01a36844 | 557 | static vm_fault_t nfs_vm_page_mkwrite(struct vm_fault *vmf) |
94387fb1 | 558 | { |
11bac800 | 559 | struct file *filp = vmf->vma->vm_file; |
6de1472f | 560 | struct inode *inode = file_inode(filp); |
94387fb1 | 561 | unsigned pagelen; |
01a36844 | 562 | vm_fault_t ret = VM_FAULT_NOPAGE; |
4899f9c8 | 563 | struct address_space *mapping; |
4fa7a717 | 564 | struct folio *folio = page_folio(vmf->page); |
94387fb1 | 565 | |
1e8968c5 | 566 | dfprintk(PAGECACHE, "NFS: vm_page_mkwrite(%pD2(%lu), offset %lld)\n", |
4fa7a717 TM |
567 | filp, filp->f_mapping->host->i_ino, |
568 | (long long)folio_file_pos(folio)); | |
b7eaefaa | 569 | |
9a773e7c TM |
570 | sb_start_pagefault(inode->i_sb); |
571 | ||
545db45f | 572 | /* make sure the cache has finished storing the page */ |
4fa7a717 TM |
573 | if (folio_test_fscache(folio) && |
574 | folio_wait_fscache_killable(folio) < 0) { | |
a6b5a28e DW |
575 | ret = VM_FAULT_RETRY; |
576 | goto out; | |
577 | } | |
545db45f | 578 | |
ef070dcb | 579 | wait_on_bit_action(&NFS_I(inode)->flags, NFS_INO_INVALIDATING, |
f5d39b02 PZ |
580 | nfs_wait_bit_killable, |
581 | TASK_KILLABLE|TASK_FREEZABLE_UNSAFE); | |
ef070dcb | 582 | |
4fa7a717 TM |
583 | folio_lock(folio); |
584 | mapping = folio_file_mapping(folio); | |
6de1472f | 585 | if (mapping != inode->i_mapping) |
8b1f9ee5 TM |
586 | goto out_unlock; |
587 | ||
4fa7a717 | 588 | folio_wait_writeback(folio); |
2aeb98f4 | 589 | |
4fa7a717 | 590 | pagelen = nfs_folio_length(folio); |
8b1f9ee5 TM |
591 | if (pagelen == 0) |
592 | goto out_unlock; | |
4899f9c8 | 593 | |
bc4866b6 | 594 | ret = VM_FAULT_LOCKED; |
0c493b5c TM |
595 | if (nfs_flush_incompatible(filp, folio) == 0 && |
596 | nfs_update_folio(filp, folio, 0, pagelen) == 0) | |
bc4866b6 | 597 | goto out; |
8b1f9ee5 | 598 | |
bc4866b6 | 599 | ret = VM_FAULT_SIGBUS; |
8b1f9ee5 | 600 | out_unlock: |
4fa7a717 | 601 | folio_unlock(folio); |
bc4866b6 | 602 | out: |
9a773e7c | 603 | sb_end_pagefault(inode->i_sb); |
bc4866b6 | 604 | return ret; |
94387fb1 TM |
605 | } |
606 | ||
f0f37e2f | 607 | static const struct vm_operations_struct nfs_file_vm_ops = { |
94387fb1 | 608 | .fault = filemap_fault, |
f1820361 | 609 | .map_pages = filemap_map_pages, |
94387fb1 TM |
610 | .page_mkwrite = nfs_vm_page_mkwrite, |
611 | }; | |
612 | ||
edaf4369 | 613 | ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 614 | { |
6de1472f AV |
615 | struct file *file = iocb->ki_filp; |
616 | struct inode *inode = file_inode(file); | |
ed7bcdb3 TM |
617 | unsigned int mntflags = NFS_SERVER(inode)->flags; |
618 | ssize_t result, written; | |
ce368536 SM |
619 | errseq_t since; |
620 | int error; | |
1da177e4 | 621 | |
6de1472f | 622 | result = nfs_key_timeout_notify(file, inode); |
dc24826b AA |
623 | if (result) |
624 | return result; | |
625 | ||
89698b24 | 626 | if (iocb->ki_flags & IOCB_DIRECT) |
64158668 | 627 | return nfs_file_direct_write(iocb, from, false); |
1da177e4 | 628 | |
619d30b4 | 629 | dprintk("NFS: write(%pD2, %zu@%Ld)\n", |
18290650 | 630 | file, iov_iter_count(from), (long long) iocb->ki_pos); |
1da177e4 | 631 | |
1da177e4 LT |
632 | if (IS_SWAPFILE(inode)) |
633 | goto out_swapfile; | |
7d52e862 TM |
634 | /* |
635 | * O_APPEND implies that we must revalidate the file length. | |
636 | */ | |
fc9dc401 | 637 | if (iocb->ki_flags & IOCB_APPEND || iocb->ki_pos > i_size_read(inode)) { |
6de1472f | 638 | result = nfs_revalidate_file_size(inode, file); |
7d52e862 | 639 | if (result) |
e6005436 | 640 | return result; |
fe51beec | 641 | } |
1da177e4 | 642 | |
28aa2f9e TM |
643 | nfs_clear_invalid_mapping(file->f_mapping); |
644 | ||
ce368536 | 645 | since = filemap_sample_wb_err(file->f_mapping); |
a5864c99 | 646 | nfs_start_io_write(inode); |
18290650 TM |
647 | result = generic_write_checks(iocb, from); |
648 | if (result > 0) { | |
649 | current->backing_dev_info = inode_to_bdi(inode); | |
800ba295 | 650 | result = generic_perform_write(iocb, from); |
18290650 TM |
651 | current->backing_dev_info = NULL; |
652 | } | |
a5864c99 | 653 | nfs_end_io_write(inode); |
18290650 | 654 | if (result <= 0) |
1da177e4 LT |
655 | goto out; |
656 | ||
c49edecd | 657 | written = result; |
18290650 | 658 | iocb->ki_pos += written; |
e6005436 | 659 | nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written); |
ed7bcdb3 TM |
660 | |
661 | if (mntflags & NFS_MOUNT_WRITE_EAGER) { | |
662 | result = filemap_fdatawrite_range(file->f_mapping, | |
663 | iocb->ki_pos - written, | |
664 | iocb->ki_pos - 1); | |
665 | if (result < 0) | |
666 | goto out; | |
667 | } | |
668 | if (mntflags & NFS_MOUNT_WRITE_WAIT) { | |
384edeb4 LB |
669 | filemap_fdatawait_range(file->f_mapping, |
670 | iocb->ki_pos - written, | |
671 | iocb->ki_pos - 1); | |
ed7bcdb3 | 672 | } |
e973b1a5 | 673 | result = generic_write_sync(iocb, written); |
674 | if (result < 0) | |
e6005436 | 675 | return result; |
7e381172 | 676 | |
e6005436 | 677 | out: |
7e94d6c4 | 678 | /* Return error values */ |
ce368536 | 679 | error = filemap_check_wb_err(file->f_mapping, since); |
e6005436 TM |
680 | switch (error) { |
681 | default: | |
682 | break; | |
683 | case -EDQUOT: | |
684 | case -EFBIG: | |
685 | case -ENOSPC: | |
686 | nfs_wb_all(inode); | |
687 | error = file_check_and_advance_wb_err(file); | |
688 | if (error < 0) | |
689 | result = error; | |
200baa21 | 690 | } |
1da177e4 LT |
691 | return result; |
692 | ||
693 | out_swapfile: | |
694 | printk(KERN_INFO "NFS: attempt to write to active swap file!\n"); | |
89658c4d | 695 | return -ETXTBSY; |
1da177e4 | 696 | } |
89d77c8f | 697 | EXPORT_SYMBOL_GPL(nfs_file_write); |
1da177e4 | 698 | |
5eebde23 SJ |
699 | static int |
700 | do_getlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
701 | { |
702 | struct inode *inode = filp->f_mapping->host; | |
703 | int status = 0; | |
21ac19d4 | 704 | unsigned int saved_type = fl->fl_type; |
1da177e4 | 705 | |
039c4d7a | 706 | /* Try local locking first */ |
6d34ac19 BF |
707 | posix_test_lock(filp, fl); |
708 | if (fl->fl_type != F_UNLCK) { | |
709 | /* found a conflict */ | |
039c4d7a | 710 | goto out; |
1da177e4 | 711 | } |
21ac19d4 | 712 | fl->fl_type = saved_type; |
039c4d7a | 713 | |
011e2a7f | 714 | if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) |
039c4d7a TM |
715 | goto out_noconflict; |
716 | ||
5eebde23 | 717 | if (is_local) |
039c4d7a TM |
718 | goto out_noconflict; |
719 | ||
720 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); | |
721 | out: | |
1da177e4 | 722 | return status; |
039c4d7a TM |
723 | out_noconflict: |
724 | fl->fl_type = F_UNLCK; | |
725 | goto out; | |
1da177e4 LT |
726 | } |
727 | ||
5eebde23 SJ |
728 | static int |
729 | do_unlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
730 | { |
731 | struct inode *inode = filp->f_mapping->host; | |
7a8203d8 | 732 | struct nfs_lock_context *l_ctx; |
1da177e4 LT |
733 | int status; |
734 | ||
1da177e4 LT |
735 | /* |
736 | * Flush all pending writes before doing anything | |
737 | * with locks.. | |
738 | */ | |
aded8d7b | 739 | nfs_wb_all(inode); |
1da177e4 | 740 | |
7a8203d8 TM |
741 | l_ctx = nfs_get_lock_context(nfs_file_open_context(filp)); |
742 | if (!IS_ERR(l_ctx)) { | |
210c7c17 | 743 | status = nfs_iocounter_wait(l_ctx); |
7a8203d8 | 744 | nfs_put_lock_context(l_ctx); |
f30cb757 BC |
745 | /* NOTE: special case |
746 | * If we're signalled while cleaning up locks on process exit, we | |
747 | * still need to complete the unlock. | |
748 | */ | |
749 | if (status < 0 && !(fl->fl_flags & FL_CLOSE)) | |
7a8203d8 TM |
750 | return status; |
751 | } | |
752 | ||
5eebde23 SJ |
753 | /* |
754 | * Use local locking if mounted with "-onolock" or with appropriate | |
755 | * "-olocal_lock=" | |
756 | */ | |
757 | if (!is_local) | |
1da177e4 LT |
758 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); |
759 | else | |
75575ddf | 760 | status = locks_lock_file_wait(filp, fl); |
1da177e4 LT |
761 | return status; |
762 | } | |
763 | ||
5eebde23 SJ |
764 | static int |
765 | do_setlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
766 | { |
767 | struct inode *inode = filp->f_mapping->host; | |
1da177e4 LT |
768 | int status; |
769 | ||
1da177e4 LT |
770 | /* |
771 | * Flush all pending writes before doing anything | |
772 | * with locks.. | |
773 | */ | |
29884df0 TM |
774 | status = nfs_sync_mapping(filp->f_mapping); |
775 | if (status != 0) | |
1da177e4 LT |
776 | goto out; |
777 | ||
5eebde23 SJ |
778 | /* |
779 | * Use local locking if mounted with "-onolock" or with appropriate | |
780 | * "-olocal_lock=" | |
781 | */ | |
782 | if (!is_local) | |
1da177e4 | 783 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); |
c4d7c402 | 784 | else |
75575ddf | 785 | status = locks_lock_file_wait(filp, fl); |
1da177e4 LT |
786 | if (status < 0) |
787 | goto out; | |
6b96724e | 788 | |
1da177e4 | 789 | /* |
779eafab N |
790 | * Invalidate cache to prevent missing any changes. If |
791 | * the file is mapped, clear the page cache as well so | |
792 | * those mappings will be loaded. | |
6b96724e | 793 | * |
1da177e4 LT |
794 | * This makes locking act as a cache coherency point. |
795 | */ | |
29884df0 | 796 | nfs_sync_mapping(filp->f_mapping); |
779eafab | 797 | if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) { |
442ce049 | 798 | nfs_zap_caches(inode); |
779eafab N |
799 | if (mapping_mapped(filp->f_mapping)) |
800 | nfs_revalidate_mapping(inode, filp->f_mapping); | |
801 | } | |
1da177e4 | 802 | out: |
1da177e4 LT |
803 | return status; |
804 | } | |
805 | ||
806 | /* | |
807 | * Lock a (portion of) a file | |
808 | */ | |
ce4ef7c0 | 809 | int nfs_lock(struct file *filp, int cmd, struct file_lock *fl) |
1da177e4 | 810 | { |
6da24bc9 | 811 | struct inode *inode = filp->f_mapping->host; |
2116271a | 812 | int ret = -ENOLCK; |
5eebde23 | 813 | int is_local = 0; |
1da177e4 | 814 | |
6de1472f AV |
815 | dprintk("NFS: lock(%pD2, t=%x, fl=%x, r=%lld:%lld)\n", |
816 | filp, fl->fl_type, fl->fl_flags, | |
1da177e4 | 817 | (long long)fl->fl_start, (long long)fl->fl_end); |
6da24bc9 | 818 | |
91d5b470 | 819 | nfs_inc_stats(inode, NFSIOS_VFSLOCK); |
1da177e4 | 820 | |
bb0a55bb BF |
821 | if (fl->fl_flags & FL_RECLAIM) |
822 | return -ENOGRACE; | |
823 | ||
5eebde23 SJ |
824 | if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FCNTL) |
825 | is_local = 1; | |
826 | ||
2116271a TM |
827 | if (NFS_PROTO(inode)->lock_check_bounds != NULL) { |
828 | ret = NFS_PROTO(inode)->lock_check_bounds(fl); | |
829 | if (ret < 0) | |
830 | goto out_err; | |
831 | } | |
1da177e4 LT |
832 | |
833 | if (IS_GETLK(cmd)) | |
5eebde23 | 834 | ret = do_getlk(filp, cmd, fl, is_local); |
2116271a | 835 | else if (fl->fl_type == F_UNLCK) |
5eebde23 | 836 | ret = do_unlk(filp, cmd, fl, is_local); |
2116271a | 837 | else |
5eebde23 | 838 | ret = do_setlk(filp, cmd, fl, is_local); |
2116271a TM |
839 | out_err: |
840 | return ret; | |
1da177e4 | 841 | } |
89d77c8f | 842 | EXPORT_SYMBOL_GPL(nfs_lock); |
1da177e4 LT |
843 | |
844 | /* | |
845 | * Lock a (portion of) a file | |
846 | */ | |
ce4ef7c0 | 847 | int nfs_flock(struct file *filp, int cmd, struct file_lock *fl) |
1da177e4 | 848 | { |
5eebde23 SJ |
849 | struct inode *inode = filp->f_mapping->host; |
850 | int is_local = 0; | |
851 | ||
6de1472f AV |
852 | dprintk("NFS: flock(%pD2, t=%x, fl=%x)\n", |
853 | filp, fl->fl_type, fl->fl_flags); | |
1da177e4 | 854 | |
1da177e4 LT |
855 | if (!(fl->fl_flags & FL_FLOCK)) |
856 | return -ENOLCK; | |
857 | ||
5eebde23 SJ |
858 | if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK) |
859 | is_local = 1; | |
860 | ||
fcfa4470 BC |
861 | /* We're simulating flock() locks using posix locks on the server */ |
862 | if (fl->fl_type == F_UNLCK) | |
5eebde23 SJ |
863 | return do_unlk(filp, cmd, fl, is_local); |
864 | return do_setlk(filp, cmd, fl, is_local); | |
1da177e4 | 865 | } |
89d77c8f | 866 | EXPORT_SYMBOL_GPL(nfs_flock); |
370f6599 | 867 | |
0486958f JL |
868 | const struct file_operations nfs_file_operations = { |
869 | .llseek = nfs_file_llseek, | |
3aa2d199 | 870 | .read_iter = nfs_file_read, |
edaf4369 | 871 | .write_iter = nfs_file_write, |
0486958f JL |
872 | .mmap = nfs_file_mmap, |
873 | .open = nfs_file_open, | |
874 | .flush = nfs_file_flush, | |
875 | .release = nfs_file_release, | |
876 | .fsync = nfs_file_fsync, | |
877 | .lock = nfs_lock, | |
878 | .flock = nfs_flock, | |
82c156f8 | 879 | .splice_read = generic_file_splice_read, |
4da54c21 | 880 | .splice_write = iter_file_splice_write, |
0486958f | 881 | .check_flags = nfs_check_flags, |
1c994a09 | 882 | .setlease = simple_nosetlease, |
0486958f | 883 | }; |
ddda8e0a | 884 | EXPORT_SYMBOL_GPL(nfs_file_operations); |