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> |
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 | 46 | static const struct vm_operations_struct nfs_file_vm_ops; |
94387fb1 | 47 | |
ce4ef7c0 | 48 | int 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 | 55 | EXPORT_SYMBOL_GPL(nfs_check_flags); |
1da177e4 LT |
56 | |
57 | /* | |
58 | * Open file | |
59 | */ | |
60 | static int | |
61 | nfs_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 | 78 | int |
1da177e4 LT |
79 | nfs_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 | 88 | EXPORT_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 | */ | |
101 | static 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 |
110 | force_reval: |
111 | return __nfs_revalidate_inode(server, inode); | |
112 | } | |
113 | ||
965c8e59 | 114 | loff_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 | 133 | EXPORT_SYMBOL_GPL(nfs_file_llseek); |
980802e3 | 134 | |
1da177e4 LT |
135 | /* |
136 | * Flush all dirty pages, and check for write errors. | |
1da177e4 | 137 | */ |
5445b1fb | 138 | static int |
75e1fcc0 | 139 | nfs_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 | 156 | ssize_t |
3aa2d199 | 157 | nfs_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 | 179 | EXPORT_SYMBOL_GPL(nfs_file_read); |
1da177e4 | 180 | |
a7db5034 DH |
181 | ssize_t |
182 | nfs_file_splice_read(struct file *in, loff_t *ppos, struct pipe_inode_info *pipe, | |
183 | size_t len, unsigned int flags) | |
184 | { | |
185 | struct inode *inode = file_inode(in); | |
186 | ssize_t result; | |
187 | ||
188 | dprintk("NFS: splice_read(%pD2, %zu@%llu)\n", in, len, *ppos); | |
189 | ||
190 | nfs_start_io_read(inode); | |
191 | result = nfs_revalidate_mapping(inode, in->f_mapping); | |
192 | if (!result) { | |
193 | result = filemap_splice_read(in, ppos, pipe, len, flags); | |
194 | if (result > 0) | |
195 | nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, result); | |
196 | } | |
197 | nfs_end_io_read(inode); | |
198 | return result; | |
199 | } | |
200 | EXPORT_SYMBOL_GPL(nfs_file_splice_read); | |
201 | ||
ce4ef7c0 | 202 | int |
a841c9cb | 203 | nfs_file_mmap(struct file *file, struct vm_area_struct *vma) |
1da177e4 | 204 | { |
6de1472f | 205 | struct inode *inode = file_inode(file); |
1da177e4 LT |
206 | int status; |
207 | ||
6de1472f | 208 | dprintk("NFS: mmap(%pD2)\n", file); |
1da177e4 | 209 | |
e1ebfd33 TM |
210 | /* Note: generic_file_mmap() returns ENOSYS on nommu systems |
211 | * so we call that before revalidating the mapping | |
212 | */ | |
213 | status = generic_file_mmap(file, vma); | |
94387fb1 TM |
214 | if (!status) { |
215 | vma->vm_ops = &nfs_file_vm_ops; | |
e1ebfd33 | 216 | status = nfs_revalidate_mapping(inode, file->f_mapping); |
94387fb1 | 217 | } |
1da177e4 LT |
218 | return status; |
219 | } | |
89d77c8f | 220 | EXPORT_SYMBOL_GPL(nfs_file_mmap); |
1da177e4 LT |
221 | |
222 | /* | |
223 | * Flush any dirty pages for this process, and check for write errors. | |
224 | * The return status from this call provides a reliable indication of | |
225 | * whether any write errors occurred for this process. | |
226 | */ | |
4ff79bc7 | 227 | static int |
bf4b4905 | 228 | nfs_file_fsync_commit(struct file *file, int datasync) |
1da177e4 | 229 | { |
6de1472f | 230 | struct inode *inode = file_inode(file); |
9641d9bc | 231 | int ret, ret2; |
af7fa165 | 232 | |
6de1472f | 233 | dprintk("NFS: fsync file(%pD2) datasync %d\n", file, datasync); |
1da177e4 | 234 | |
91d5b470 | 235 | nfs_inc_stats(inode, NFSIOS_VFSFSYNC); |
2197e9b0 | 236 | ret = nfs_commit_inode(inode, FLUSH_SYNC); |
9641d9bc TM |
237 | ret2 = file_check_and_advance_wb_err(file); |
238 | if (ret2 < 0) | |
239 | return ret2; | |
240 | return ret; | |
a5c58892 BS |
241 | } |
242 | ||
4ff79bc7 | 243 | int |
a5c58892 BS |
244 | nfs_file_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
245 | { | |
496ad9aa | 246 | struct inode *inode = file_inode(file); |
67f4b5dc TM |
247 | struct nfs_inode *nfsi = NFS_I(inode); |
248 | long save_nredirtied = atomic_long_read(&nfsi->redirtied_pages); | |
249 | long nredirtied; | |
2197e9b0 | 250 | int ret; |
a5c58892 | 251 | |
f4ce1299 TM |
252 | trace_nfs_fsync_enter(inode); |
253 | ||
2197e9b0 | 254 | for (;;) { |
6fbda89b | 255 | ret = file_write_and_wait_range(file, start, end); |
05990d1b TM |
256 | if (ret != 0) |
257 | break; | |
bf4b4905 | 258 | ret = nfs_file_fsync_commit(file, datasync); |
2197e9b0 TM |
259 | if (ret != 0) |
260 | break; | |
261 | ret = pnfs_sync_inode(inode, !!datasync); | |
262 | if (ret != 0) | |
263 | break; | |
67f4b5dc TM |
264 | nredirtied = atomic_long_read(&nfsi->redirtied_pages); |
265 | if (nredirtied == save_nredirtied) | |
2197e9b0 | 266 | break; |
67f4b5dc | 267 | save_nredirtied = nredirtied; |
2197e9b0 | 268 | } |
05990d1b | 269 | |
f4ce1299 | 270 | trace_nfs_fsync_exit(inode, ret); |
af7fa165 | 271 | return ret; |
1da177e4 | 272 | } |
4ff79bc7 | 273 | EXPORT_SYMBOL_GPL(nfs_file_fsync); |
1da177e4 | 274 | |
38c73044 PS |
275 | /* |
276 | * Decide whether a read/modify/write cycle may be more efficient | |
277 | * then a modify/write/read cycle when writing to a page in the | |
278 | * page cache. | |
279 | * | |
2cde04e9 KI |
280 | * Some pNFS layout drivers can only read/write at a certain block |
281 | * granularity like all block devices and therefore we must perform | |
282 | * read/modify/write whenever a page hasn't read yet and the data | |
283 | * to be written there is not aligned to a block boundary and/or | |
284 | * smaller than the block size. | |
285 | * | |
38c73044 PS |
286 | * The modify/write/read cycle may occur if a page is read before |
287 | * being completely filled by the writer. In this situation, the | |
288 | * page must be completely written to stable storage on the server | |
289 | * before it can be refilled by reading in the page from the server. | |
290 | * This can lead to expensive, small, FILE_SYNC mode writes being | |
291 | * done. | |
292 | * | |
293 | * It may be more efficient to read the page first if the file is | |
294 | * open for reading in addition to writing, the page is not marked | |
295 | * as Uptodate, it is not dirty or waiting to be committed, | |
296 | * indicating that it was previously allocated and then modified, | |
297 | * that there were valid bytes of data in that range of the file, | |
298 | * and that the new data won't completely replace the old data in | |
299 | * that range of the file. | |
300 | */ | |
54d99381 TM |
301 | static bool nfs_folio_is_full_write(struct folio *folio, loff_t pos, |
302 | unsigned int len) | |
38c73044 | 303 | { |
54d99381 TM |
304 | unsigned int pglen = nfs_folio_length(folio); |
305 | unsigned int offset = offset_in_folio(folio, pos); | |
38c73044 PS |
306 | unsigned int end = offset + len; |
307 | ||
2cde04e9 KI |
308 | return !pglen || (end >= pglen && !offset); |
309 | } | |
612aa983 | 310 | |
54d99381 TM |
311 | static bool nfs_want_read_modify_write(struct file *file, struct folio *folio, |
312 | loff_t pos, unsigned int len) | |
2cde04e9 KI |
313 | { |
314 | /* | |
315 | * Up-to-date pages, those with ongoing or full-page write | |
316 | * don't need read/modify/write | |
317 | */ | |
54d99381 TM |
318 | if (folio_test_uptodate(folio) || folio_test_private(folio) || |
319 | nfs_folio_is_full_write(folio, pos, len)) | |
2cde04e9 KI |
320 | return false; |
321 | ||
54d99381 | 322 | if (pnfs_ld_read_whole_page(file_inode(file))) |
2cde04e9 KI |
323 | return true; |
324 | /* Open for reading too? */ | |
325 | if (file->f_mode & FMODE_READ) | |
326 | return true; | |
327 | return false; | |
38c73044 PS |
328 | } |
329 | ||
1da177e4 | 330 | /* |
4899f9c8 NP |
331 | * This does the "real" work of the write. We must allocate and lock the |
332 | * page to be sent back to the generic routine, which then copies the | |
333 | * data from user space. | |
1da177e4 LT |
334 | * |
335 | * If the writer ends up delaying the write, the writer needs to | |
336 | * increment the page use counts until he is done with the page. | |
337 | */ | |
4899f9c8 | 338 | static int nfs_write_begin(struct file *file, struct address_space *mapping, |
54d99381 TM |
339 | loff_t pos, unsigned len, struct page **pagep, |
340 | void **fsdata) | |
1da177e4 | 341 | { |
0c493b5c | 342 | struct folio *folio; |
38c73044 | 343 | int once_thru = 0; |
54d99381 | 344 | int ret; |
4899f9c8 | 345 | |
1e8968c5 | 346 | dfprintk(PAGECACHE, "NFS: write_begin(%pD2(%lu), %u@%lld)\n", |
6de1472f | 347 | file, mapping->host->i_ino, len, (long long) pos); |
b7eaefaa | 348 | |
38c73044 | 349 | start: |
e999a5c5 MW |
350 | folio = __filemap_get_folio(mapping, pos >> PAGE_SHIFT, FGP_WRITEBEGIN, |
351 | mapping_gfp_mask(mapping)); | |
66dabbb6 CH |
352 | if (IS_ERR(folio)) |
353 | return PTR_ERR(folio); | |
54d99381 | 354 | *pagep = &folio->page; |
4899f9c8 | 355 | |
0c493b5c | 356 | ret = nfs_flush_incompatible(file, folio); |
4899f9c8 | 357 | if (ret) { |
54d99381 TM |
358 | folio_unlock(folio); |
359 | folio_put(folio); | |
38c73044 | 360 | } else if (!once_thru && |
54d99381 | 361 | nfs_want_read_modify_write(file, folio, pos, len)) { |
38c73044 | 362 | once_thru = 1; |
0c493b5c | 363 | ret = nfs_read_folio(file, folio); |
54d99381 | 364 | folio_put(folio); |
38c73044 PS |
365 | if (!ret) |
366 | goto start; | |
4899f9c8 NP |
367 | } |
368 | return ret; | |
1da177e4 LT |
369 | } |
370 | ||
4899f9c8 | 371 | static int nfs_write_end(struct file *file, struct address_space *mapping, |
54d99381 TM |
372 | loff_t pos, unsigned len, unsigned copied, |
373 | struct page *page, void *fsdata) | |
1da177e4 | 374 | { |
dc24826b | 375 | struct nfs_open_context *ctx = nfs_file_open_context(file); |
0c493b5c | 376 | struct folio *folio = page_folio(page); |
54d99381 | 377 | unsigned offset = offset_in_folio(folio, pos); |
4899f9c8 | 378 | int status; |
1da177e4 | 379 | |
1e8968c5 | 380 | dfprintk(PAGECACHE, "NFS: write_end(%pD2(%lu), %u@%lld)\n", |
6de1472f | 381 | file, mapping->host->i_ino, len, (long long) pos); |
b7eaefaa | 382 | |
efc91ed0 TM |
383 | /* |
384 | * Zero any uninitialised parts of the page, and then mark the page | |
385 | * as up to date if it turns out that we're extending the file. | |
386 | */ | |
54d99381 TM |
387 | if (!folio_test_uptodate(folio)) { |
388 | size_t fsize = folio_size(folio); | |
389 | unsigned pglen = nfs_folio_length(folio); | |
c0cf3ef5 | 390 | unsigned end = offset + copied; |
efc91ed0 TM |
391 | |
392 | if (pglen == 0) { | |
54d99381 TM |
393 | folio_zero_segments(folio, 0, offset, end, fsize); |
394 | folio_mark_uptodate(folio); | |
efc91ed0 | 395 | } else if (end >= pglen) { |
54d99381 | 396 | folio_zero_segment(folio, end, fsize); |
efc91ed0 | 397 | if (offset == 0) |
54d99381 | 398 | folio_mark_uptodate(folio); |
efc91ed0 | 399 | } else |
54d99381 | 400 | folio_zero_segment(folio, pglen, fsize); |
efc91ed0 TM |
401 | } |
402 | ||
0c493b5c | 403 | status = nfs_update_folio(file, folio, offset, copied); |
4899f9c8 | 404 | |
54d99381 TM |
405 | folio_unlock(folio); |
406 | folio_put(folio); | |
4899f9c8 | 407 | |
3d509e54 CL |
408 | if (status < 0) |
409 | return status; | |
2701d086 | 410 | NFS_I(mapping->host)->write_io += copied; |
dc24826b | 411 | |
d95b2665 TM |
412 | if (nfs_ctx_key_to_expire(ctx, mapping->host)) |
413 | nfs_wb_all(mapping->host); | |
dc24826b | 414 | |
3d509e54 | 415 | return copied; |
1da177e4 LT |
416 | } |
417 | ||
6b9b3514 DH |
418 | /* |
419 | * Partially or wholly invalidate a page | |
420 | * - Release the private state associated with a page if undergoing complete | |
421 | * page invalidation | |
545db45f | 422 | * - Called if either PG_private or PG_fscache is set on the page |
6b9b3514 DH |
423 | * - Caller holds page lock |
424 | */ | |
6d740c76 MWO |
425 | static void nfs_invalidate_folio(struct folio *folio, size_t offset, |
426 | size_t length) | |
cd52ed35 | 427 | { |
eb5654b3 | 428 | struct inode *inode = folio_file_mapping(folio)->host; |
6d740c76 MWO |
429 | dfprintk(PAGECACHE, "NFS: invalidate_folio(%lu, %zu, %zu)\n", |
430 | folio->index, offset, length); | |
b7eaefaa | 431 | |
6d740c76 | 432 | if (offset != 0 || length < folio_size(folio)) |
1c75950b | 433 | return; |
d2ccddf0 | 434 | /* Cancel any unstarted writes on this page */ |
eb5654b3 | 435 | nfs_wb_folio_cancel(inode, folio); |
6d740c76 | 436 | folio_wait_fscache(folio); |
eb5654b3 | 437 | trace_nfs_invalidate_folio(inode, folio); |
cd52ed35 TM |
438 | } |
439 | ||
6b9b3514 | 440 | /* |
3577da4a MWO |
441 | * Attempt to release the private state associated with a folio |
442 | * - Called if either private or fscache flags are set on the folio | |
443 | * - Caller holds folio lock | |
444 | * - Return true (may release folio) or false (may not) | |
6b9b3514 | 445 | */ |
3577da4a | 446 | static bool nfs_release_folio(struct folio *folio, gfp_t gfp) |
cd52ed35 | 447 | { |
3577da4a | 448 | dfprintk(PAGECACHE, "NFS: release_folio(%p)\n", folio); |
b7eaefaa | 449 | |
3577da4a | 450 | /* If the private flag is set, then the folio is not freeable */ |
96780ca5 TM |
451 | if (folio_test_private(folio)) { |
452 | if ((current_gfp_context(gfp) & GFP_KERNEL) != GFP_KERNEL || | |
453 | current_is_kswapd()) | |
454 | return false; | |
455 | if (nfs_wb_folio(folio_file_mapping(folio)->host, folio) < 0) | |
456 | return false; | |
457 | } | |
3577da4a | 458 | return nfs_fscache_release_folio(folio, gfp); |
e3db7691 TM |
459 | } |
460 | ||
520f301c | 461 | static void nfs_check_dirty_writeback(struct folio *folio, |
f919b196 MG |
462 | bool *dirty, bool *writeback) |
463 | { | |
464 | struct nfs_inode *nfsi; | |
520f301c | 465 | struct address_space *mapping = folio->mapping; |
f919b196 MG |
466 | |
467 | /* | |
520f301c MWO |
468 | * Check if an unstable folio is currently being committed and |
469 | * if so, have the VM treat it as if the folio is under writeback | |
470 | * so it will not block due to folios that will shortly be freeable. | |
f919b196 MG |
471 | */ |
472 | nfsi = NFS_I(mapping->host); | |
af7cf057 | 473 | if (atomic_read(&nfsi->commit_info.rpcs_out)) { |
f919b196 MG |
474 | *writeback = true; |
475 | return; | |
476 | } | |
477 | ||
478 | /* | |
520f301c MWO |
479 | * If the private flag is set, then the folio is not freeable |
480 | * and as the inode is not being committed, it's not going to | |
481 | * be cleaned in the near future so treat it as dirty | |
f919b196 | 482 | */ |
520f301c | 483 | if (folio_test_private(folio)) |
f919b196 MG |
484 | *dirty = true; |
485 | } | |
486 | ||
6b9b3514 DH |
487 | /* |
488 | * Attempt to clear the private state associated with a page when an error | |
489 | * occurs that requires the cached contents of an inode to be written back or | |
490 | * destroyed | |
545db45f | 491 | * - Called if either PG_private or fscache is set on the page |
6b9b3514 DH |
492 | * - Caller holds page lock |
493 | * - Return 0 if successful, -error otherwise | |
494 | */ | |
15a30ab2 | 495 | static int nfs_launder_folio(struct folio *folio) |
e3db7691 | 496 | { |
15a30ab2 | 497 | struct inode *inode = folio->mapping->host; |
eb5654b3 | 498 | int ret; |
b7eaefaa | 499 | |
15a30ab2 MWO |
500 | dfprintk(PAGECACHE, "NFS: launder_folio(%ld, %llu)\n", |
501 | inode->i_ino, folio_pos(folio)); | |
b7eaefaa | 502 | |
15a30ab2 | 503 | folio_wait_fscache(folio); |
eb5654b3 TM |
504 | ret = nfs_wb_folio(inode, folio); |
505 | trace_nfs_launder_folio_done(inode, folio, ret); | |
506 | return ret; | |
cd52ed35 TM |
507 | } |
508 | ||
a564b8f0 MG |
509 | static int nfs_swap_activate(struct swap_info_struct *sis, struct file *file, |
510 | sector_t *span) | |
511 | { | |
bd89bc67 MZ |
512 | unsigned long blocks; |
513 | long long isize; | |
4b60c0ff | 514 | int ret; |
4dc73c67 N |
515 | struct inode *inode = file_inode(file); |
516 | struct rpc_clnt *clnt = NFS_CLIENT(inode); | |
517 | struct nfs_client *cl = NFS_SERVER(inode)->nfs_client; | |
bd89bc67 MZ |
518 | |
519 | spin_lock(&inode->i_lock); | |
520 | blocks = inode->i_blocks; | |
521 | isize = inode->i_size; | |
522 | spin_unlock(&inode->i_lock); | |
523 | if (blocks*512 < isize) { | |
524 | pr_warn("swap activate: swapfile has holes\n"); | |
525 | return -EINVAL; | |
526 | } | |
dad2b015 | 527 | |
4b60c0ff N |
528 | ret = rpc_clnt_swap_activate(clnt); |
529 | if (ret) | |
530 | return ret; | |
531 | ret = add_swap_extent(sis, 0, sis->max, 0); | |
532 | if (ret < 0) { | |
533 | rpc_clnt_swap_deactivate(clnt); | |
534 | return ret; | |
535 | } | |
dad2b015 | 536 | |
4b60c0ff | 537 | *span = sis->pages; |
4dc73c67 N |
538 | |
539 | if (cl->rpc_ops->enable_swap) | |
540 | cl->rpc_ops->enable_swap(inode); | |
541 | ||
4b60c0ff N |
542 | sis->flags |= SWP_FS_OPS; |
543 | return ret; | |
a564b8f0 MG |
544 | } |
545 | ||
546 | static void nfs_swap_deactivate(struct file *file) | |
547 | { | |
4dc73c67 N |
548 | struct inode *inode = file_inode(file); |
549 | struct rpc_clnt *clnt = NFS_CLIENT(inode); | |
550 | struct nfs_client *cl = NFS_SERVER(inode)->nfs_client; | |
dad2b015 | 551 | |
3c87ef6e | 552 | rpc_clnt_swap_deactivate(clnt); |
4dc73c67 N |
553 | if (cl->rpc_ops->disable_swap) |
554 | cl->rpc_ops->disable_swap(file_inode(file)); | |
a564b8f0 | 555 | } |
a564b8f0 | 556 | |
f5e54d6e | 557 | const struct address_space_operations nfs_file_aops = { |
65d023af | 558 | .read_folio = nfs_read_folio, |
8786fde8 | 559 | .readahead = nfs_readahead, |
187c82cb | 560 | .dirty_folio = filemap_dirty_folio, |
1da177e4 LT |
561 | .writepage = nfs_writepage, |
562 | .writepages = nfs_writepages, | |
4899f9c8 NP |
563 | .write_begin = nfs_write_begin, |
564 | .write_end = nfs_write_end, | |
6d740c76 | 565 | .invalidate_folio = nfs_invalidate_folio, |
3577da4a | 566 | .release_folio = nfs_release_folio, |
4ae84a80 | 567 | .migrate_folio = nfs_migrate_folio, |
15a30ab2 | 568 | .launder_folio = nfs_launder_folio, |
f919b196 | 569 | .is_dirty_writeback = nfs_check_dirty_writeback, |
f590f333 | 570 | .error_remove_page = generic_error_remove_page, |
a564b8f0 MG |
571 | .swap_activate = nfs_swap_activate, |
572 | .swap_deactivate = nfs_swap_deactivate, | |
eb79f3af | 573 | .swap_rw = nfs_swap_rw, |
1da177e4 LT |
574 | }; |
575 | ||
6b9b3514 DH |
576 | /* |
577 | * Notification that a PTE pointing to an NFS page is about to be made | |
578 | * writable, implying that someone is about to modify the page through a | |
579 | * shared-writable mapping | |
580 | */ | |
01a36844 | 581 | static vm_fault_t nfs_vm_page_mkwrite(struct vm_fault *vmf) |
94387fb1 | 582 | { |
11bac800 | 583 | struct file *filp = vmf->vma->vm_file; |
6de1472f | 584 | struct inode *inode = file_inode(filp); |
94387fb1 | 585 | unsigned pagelen; |
01a36844 | 586 | vm_fault_t ret = VM_FAULT_NOPAGE; |
4899f9c8 | 587 | struct address_space *mapping; |
4fa7a717 | 588 | struct folio *folio = page_folio(vmf->page); |
94387fb1 | 589 | |
1e8968c5 | 590 | dfprintk(PAGECACHE, "NFS: vm_page_mkwrite(%pD2(%lu), offset %lld)\n", |
4fa7a717 TM |
591 | filp, filp->f_mapping->host->i_ino, |
592 | (long long)folio_file_pos(folio)); | |
b7eaefaa | 593 | |
9a773e7c TM |
594 | sb_start_pagefault(inode->i_sb); |
595 | ||
545db45f | 596 | /* make sure the cache has finished storing the page */ |
4fa7a717 TM |
597 | if (folio_test_fscache(folio) && |
598 | folio_wait_fscache_killable(folio) < 0) { | |
a6b5a28e DW |
599 | ret = VM_FAULT_RETRY; |
600 | goto out; | |
601 | } | |
545db45f | 602 | |
ef070dcb | 603 | wait_on_bit_action(&NFS_I(inode)->flags, NFS_INO_INVALIDATING, |
f5d39b02 PZ |
604 | nfs_wait_bit_killable, |
605 | TASK_KILLABLE|TASK_FREEZABLE_UNSAFE); | |
ef070dcb | 606 | |
4fa7a717 TM |
607 | folio_lock(folio); |
608 | mapping = folio_file_mapping(folio); | |
6de1472f | 609 | if (mapping != inode->i_mapping) |
8b1f9ee5 TM |
610 | goto out_unlock; |
611 | ||
4fa7a717 | 612 | folio_wait_writeback(folio); |
2aeb98f4 | 613 | |
4fa7a717 | 614 | pagelen = nfs_folio_length(folio); |
8b1f9ee5 TM |
615 | if (pagelen == 0) |
616 | goto out_unlock; | |
4899f9c8 | 617 | |
bc4866b6 | 618 | ret = VM_FAULT_LOCKED; |
0c493b5c TM |
619 | if (nfs_flush_incompatible(filp, folio) == 0 && |
620 | nfs_update_folio(filp, folio, 0, pagelen) == 0) | |
bc4866b6 | 621 | goto out; |
8b1f9ee5 | 622 | |
bc4866b6 | 623 | ret = VM_FAULT_SIGBUS; |
8b1f9ee5 | 624 | out_unlock: |
4fa7a717 | 625 | folio_unlock(folio); |
bc4866b6 | 626 | out: |
9a773e7c | 627 | sb_end_pagefault(inode->i_sb); |
bc4866b6 | 628 | return ret; |
94387fb1 TM |
629 | } |
630 | ||
f0f37e2f | 631 | static const struct vm_operations_struct nfs_file_vm_ops = { |
94387fb1 | 632 | .fault = filemap_fault, |
f1820361 | 633 | .map_pages = filemap_map_pages, |
94387fb1 TM |
634 | .page_mkwrite = nfs_vm_page_mkwrite, |
635 | }; | |
636 | ||
edaf4369 | 637 | ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 638 | { |
6de1472f AV |
639 | struct file *file = iocb->ki_filp; |
640 | struct inode *inode = file_inode(file); | |
ed7bcdb3 TM |
641 | unsigned int mntflags = NFS_SERVER(inode)->flags; |
642 | ssize_t result, written; | |
ce368536 SM |
643 | errseq_t since; |
644 | int error; | |
1da177e4 | 645 | |
6de1472f | 646 | result = nfs_key_timeout_notify(file, inode); |
dc24826b AA |
647 | if (result) |
648 | return result; | |
649 | ||
89698b24 | 650 | if (iocb->ki_flags & IOCB_DIRECT) |
64158668 | 651 | return nfs_file_direct_write(iocb, from, false); |
1da177e4 | 652 | |
619d30b4 | 653 | dprintk("NFS: write(%pD2, %zu@%Ld)\n", |
18290650 | 654 | file, iov_iter_count(from), (long long) iocb->ki_pos); |
1da177e4 | 655 | |
1da177e4 LT |
656 | if (IS_SWAPFILE(inode)) |
657 | goto out_swapfile; | |
7d52e862 TM |
658 | /* |
659 | * O_APPEND implies that we must revalidate the file length. | |
660 | */ | |
fc9dc401 | 661 | if (iocb->ki_flags & IOCB_APPEND || iocb->ki_pos > i_size_read(inode)) { |
6de1472f | 662 | result = nfs_revalidate_file_size(inode, file); |
7d52e862 | 663 | if (result) |
e6005436 | 664 | return result; |
fe51beec | 665 | } |
1da177e4 | 666 | |
28aa2f9e TM |
667 | nfs_clear_invalid_mapping(file->f_mapping); |
668 | ||
ce368536 | 669 | since = filemap_sample_wb_err(file->f_mapping); |
a5864c99 | 670 | nfs_start_io_write(inode); |
18290650 | 671 | result = generic_write_checks(iocb, from); |
0d625446 | 672 | if (result > 0) |
800ba295 | 673 | result = generic_perform_write(iocb, from); |
a5864c99 | 674 | nfs_end_io_write(inode); |
18290650 | 675 | if (result <= 0) |
1da177e4 LT |
676 | goto out; |
677 | ||
c49edecd | 678 | written = result; |
e6005436 | 679 | nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written); |
ed7bcdb3 TM |
680 | |
681 | if (mntflags & NFS_MOUNT_WRITE_EAGER) { | |
682 | result = filemap_fdatawrite_range(file->f_mapping, | |
683 | iocb->ki_pos - written, | |
684 | iocb->ki_pos - 1); | |
685 | if (result < 0) | |
686 | goto out; | |
687 | } | |
688 | if (mntflags & NFS_MOUNT_WRITE_WAIT) { | |
384edeb4 LB |
689 | filemap_fdatawait_range(file->f_mapping, |
690 | iocb->ki_pos - written, | |
691 | iocb->ki_pos - 1); | |
ed7bcdb3 | 692 | } |
e973b1a5 | 693 | result = generic_write_sync(iocb, written); |
694 | if (result < 0) | |
e6005436 | 695 | return result; |
7e381172 | 696 | |
e6005436 | 697 | out: |
7e94d6c4 | 698 | /* Return error values */ |
ce368536 | 699 | error = filemap_check_wb_err(file->f_mapping, since); |
e6005436 TM |
700 | switch (error) { |
701 | default: | |
702 | break; | |
703 | case -EDQUOT: | |
704 | case -EFBIG: | |
705 | case -ENOSPC: | |
706 | nfs_wb_all(inode); | |
707 | error = file_check_and_advance_wb_err(file); | |
708 | if (error < 0) | |
709 | result = error; | |
200baa21 | 710 | } |
1da177e4 LT |
711 | return result; |
712 | ||
713 | out_swapfile: | |
714 | printk(KERN_INFO "NFS: attempt to write to active swap file!\n"); | |
89658c4d | 715 | return -ETXTBSY; |
1da177e4 | 716 | } |
89d77c8f | 717 | EXPORT_SYMBOL_GPL(nfs_file_write); |
1da177e4 | 718 | |
5eebde23 SJ |
719 | static int |
720 | do_getlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
721 | { |
722 | struct inode *inode = filp->f_mapping->host; | |
723 | int status = 0; | |
21ac19d4 | 724 | unsigned int saved_type = fl->fl_type; |
1da177e4 | 725 | |
039c4d7a | 726 | /* Try local locking first */ |
6d34ac19 BF |
727 | posix_test_lock(filp, fl); |
728 | if (fl->fl_type != F_UNLCK) { | |
729 | /* found a conflict */ | |
039c4d7a | 730 | goto out; |
1da177e4 | 731 | } |
21ac19d4 | 732 | fl->fl_type = saved_type; |
039c4d7a | 733 | |
011e2a7f | 734 | if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) |
039c4d7a TM |
735 | goto out_noconflict; |
736 | ||
5eebde23 | 737 | if (is_local) |
039c4d7a TM |
738 | goto out_noconflict; |
739 | ||
740 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); | |
741 | out: | |
1da177e4 | 742 | return status; |
039c4d7a TM |
743 | out_noconflict: |
744 | fl->fl_type = F_UNLCK; | |
745 | goto out; | |
1da177e4 LT |
746 | } |
747 | ||
5eebde23 SJ |
748 | static int |
749 | do_unlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
750 | { |
751 | struct inode *inode = filp->f_mapping->host; | |
7a8203d8 | 752 | struct nfs_lock_context *l_ctx; |
1da177e4 LT |
753 | int status; |
754 | ||
1da177e4 LT |
755 | /* |
756 | * Flush all pending writes before doing anything | |
757 | * with locks.. | |
758 | */ | |
aded8d7b | 759 | nfs_wb_all(inode); |
1da177e4 | 760 | |
7a8203d8 TM |
761 | l_ctx = nfs_get_lock_context(nfs_file_open_context(filp)); |
762 | if (!IS_ERR(l_ctx)) { | |
210c7c17 | 763 | status = nfs_iocounter_wait(l_ctx); |
7a8203d8 | 764 | nfs_put_lock_context(l_ctx); |
f30cb757 BC |
765 | /* NOTE: special case |
766 | * If we're signalled while cleaning up locks on process exit, we | |
767 | * still need to complete the unlock. | |
768 | */ | |
769 | if (status < 0 && !(fl->fl_flags & FL_CLOSE)) | |
7a8203d8 TM |
770 | return status; |
771 | } | |
772 | ||
5eebde23 SJ |
773 | /* |
774 | * Use local locking if mounted with "-onolock" or with appropriate | |
775 | * "-olocal_lock=" | |
776 | */ | |
777 | if (!is_local) | |
1da177e4 LT |
778 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); |
779 | else | |
75575ddf | 780 | status = locks_lock_file_wait(filp, fl); |
1da177e4 LT |
781 | return status; |
782 | } | |
783 | ||
5eebde23 SJ |
784 | static int |
785 | do_setlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
786 | { |
787 | struct inode *inode = filp->f_mapping->host; | |
1da177e4 LT |
788 | int status; |
789 | ||
1da177e4 LT |
790 | /* |
791 | * Flush all pending writes before doing anything | |
792 | * with locks.. | |
793 | */ | |
29884df0 TM |
794 | status = nfs_sync_mapping(filp->f_mapping); |
795 | if (status != 0) | |
1da177e4 LT |
796 | goto out; |
797 | ||
5eebde23 SJ |
798 | /* |
799 | * Use local locking if mounted with "-onolock" or with appropriate | |
800 | * "-olocal_lock=" | |
801 | */ | |
802 | if (!is_local) | |
1da177e4 | 803 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); |
c4d7c402 | 804 | else |
75575ddf | 805 | status = locks_lock_file_wait(filp, fl); |
1da177e4 LT |
806 | if (status < 0) |
807 | goto out; | |
6b96724e | 808 | |
1da177e4 | 809 | /* |
779eafab N |
810 | * Invalidate cache to prevent missing any changes. If |
811 | * the file is mapped, clear the page cache as well so | |
812 | * those mappings will be loaded. | |
6b96724e | 813 | * |
1da177e4 LT |
814 | * This makes locking act as a cache coherency point. |
815 | */ | |
29884df0 | 816 | nfs_sync_mapping(filp->f_mapping); |
779eafab | 817 | if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) { |
442ce049 | 818 | nfs_zap_caches(inode); |
779eafab N |
819 | if (mapping_mapped(filp->f_mapping)) |
820 | nfs_revalidate_mapping(inode, filp->f_mapping); | |
821 | } | |
1da177e4 | 822 | out: |
1da177e4 LT |
823 | return status; |
824 | } | |
825 | ||
826 | /* | |
827 | * Lock a (portion of) a file | |
828 | */ | |
ce4ef7c0 | 829 | int nfs_lock(struct file *filp, int cmd, struct file_lock *fl) |
1da177e4 | 830 | { |
6da24bc9 | 831 | struct inode *inode = filp->f_mapping->host; |
2116271a | 832 | int ret = -ENOLCK; |
5eebde23 | 833 | int is_local = 0; |
1da177e4 | 834 | |
6de1472f AV |
835 | dprintk("NFS: lock(%pD2, t=%x, fl=%x, r=%lld:%lld)\n", |
836 | filp, fl->fl_type, fl->fl_flags, | |
1da177e4 | 837 | (long long)fl->fl_start, (long long)fl->fl_end); |
6da24bc9 | 838 | |
91d5b470 | 839 | nfs_inc_stats(inode, NFSIOS_VFSLOCK); |
1da177e4 | 840 | |
bb0a55bb BF |
841 | if (fl->fl_flags & FL_RECLAIM) |
842 | return -ENOGRACE; | |
843 | ||
5eebde23 SJ |
844 | if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FCNTL) |
845 | is_local = 1; | |
846 | ||
2116271a TM |
847 | if (NFS_PROTO(inode)->lock_check_bounds != NULL) { |
848 | ret = NFS_PROTO(inode)->lock_check_bounds(fl); | |
849 | if (ret < 0) | |
850 | goto out_err; | |
851 | } | |
1da177e4 LT |
852 | |
853 | if (IS_GETLK(cmd)) | |
5eebde23 | 854 | ret = do_getlk(filp, cmd, fl, is_local); |
2116271a | 855 | else if (fl->fl_type == F_UNLCK) |
5eebde23 | 856 | ret = do_unlk(filp, cmd, fl, is_local); |
2116271a | 857 | else |
5eebde23 | 858 | ret = do_setlk(filp, cmd, fl, is_local); |
2116271a TM |
859 | out_err: |
860 | return ret; | |
1da177e4 | 861 | } |
89d77c8f | 862 | EXPORT_SYMBOL_GPL(nfs_lock); |
1da177e4 LT |
863 | |
864 | /* | |
865 | * Lock a (portion of) a file | |
866 | */ | |
ce4ef7c0 | 867 | int nfs_flock(struct file *filp, int cmd, struct file_lock *fl) |
1da177e4 | 868 | { |
5eebde23 SJ |
869 | struct inode *inode = filp->f_mapping->host; |
870 | int is_local = 0; | |
871 | ||
6de1472f AV |
872 | dprintk("NFS: flock(%pD2, t=%x, fl=%x)\n", |
873 | filp, fl->fl_type, fl->fl_flags); | |
1da177e4 | 874 | |
1da177e4 LT |
875 | if (!(fl->fl_flags & FL_FLOCK)) |
876 | return -ENOLCK; | |
877 | ||
5eebde23 SJ |
878 | if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK) |
879 | is_local = 1; | |
880 | ||
fcfa4470 BC |
881 | /* We're simulating flock() locks using posix locks on the server */ |
882 | if (fl->fl_type == F_UNLCK) | |
5eebde23 SJ |
883 | return do_unlk(filp, cmd, fl, is_local); |
884 | return do_setlk(filp, cmd, fl, is_local); | |
1da177e4 | 885 | } |
89d77c8f | 886 | EXPORT_SYMBOL_GPL(nfs_flock); |
370f6599 | 887 | |
0486958f JL |
888 | const struct file_operations nfs_file_operations = { |
889 | .llseek = nfs_file_llseek, | |
3aa2d199 | 890 | .read_iter = nfs_file_read, |
edaf4369 | 891 | .write_iter = nfs_file_write, |
0486958f JL |
892 | .mmap = nfs_file_mmap, |
893 | .open = nfs_file_open, | |
894 | .flush = nfs_file_flush, | |
895 | .release = nfs_file_release, | |
896 | .fsync = nfs_file_fsync, | |
897 | .lock = nfs_lock, | |
898 | .flock = nfs_flock, | |
a7db5034 | 899 | .splice_read = nfs_file_splice_read, |
4da54c21 | 900 | .splice_write = iter_file_splice_write, |
0486958f | 901 | .check_flags = nfs_check_flags, |
1c994a09 | 902 | .setlease = simple_nosetlease, |
0486958f | 903 | }; |
ddda8e0a | 904 | EXPORT_SYMBOL_GPL(nfs_file_operations); |