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); |
1da177e4 LT |
72 | return res; |
73 | } | |
74 | ||
ce4ef7c0 | 75 | int |
1da177e4 LT |
76 | nfs_file_release(struct inode *inode, struct file *filp) |
77 | { | |
6de1472f | 78 | dprintk("NFS: release(%pD2)\n", filp); |
6da24bc9 | 79 | |
91d5b470 | 80 | nfs_inc_stats(inode, NFSIOS_VFSRELEASE); |
aff8d8dc | 81 | nfs_file_clear_open_context(filp); |
a6b5a28e | 82 | nfs_fscache_release_file(inode, filp); |
aff8d8dc | 83 | return 0; |
1da177e4 | 84 | } |
89d77c8f | 85 | EXPORT_SYMBOL_GPL(nfs_file_release); |
1da177e4 | 86 | |
980802e3 | 87 | /** |
37eaeed1 | 88 | * nfs_revalidate_file_size - Revalidate the file size |
302fad7b TM |
89 | * @inode: pointer to inode struct |
90 | * @filp: pointer to struct file | |
980802e3 TM |
91 | * |
92 | * Revalidates the file length. This is basically a wrapper around | |
93 | * nfs_revalidate_inode() that takes into account the fact that we may | |
94 | * have cached writes (in which case we don't care about the server's | |
95 | * idea of what the file length is), or O_DIRECT (in which case we | |
96 | * shouldn't trust the cache). | |
97 | */ | |
98 | static int nfs_revalidate_file_size(struct inode *inode, struct file *filp) | |
99 | { | |
100 | struct nfs_server *server = NFS_SERVER(inode); | |
d7cf8dd0 | 101 | |
980802e3 TM |
102 | if (filp->f_flags & O_DIRECT) |
103 | goto force_reval; | |
13c0b082 | 104 | if (nfs_check_cache_invalid(inode, NFS_INO_INVALID_SIZE)) |
d7cf8dd0 | 105 | goto force_reval; |
d7cf8dd0 | 106 | return 0; |
980802e3 TM |
107 | force_reval: |
108 | return __nfs_revalidate_inode(server, inode); | |
109 | } | |
110 | ||
965c8e59 | 111 | loff_t nfs_file_llseek(struct file *filp, loff_t offset, int whence) |
980802e3 | 112 | { |
6de1472f AV |
113 | dprintk("NFS: llseek file(%pD2, %lld, %d)\n", |
114 | filp, offset, whence); | |
b84e06c5 | 115 | |
06222e49 | 116 | /* |
965c8e59 | 117 | * whence == SEEK_END || SEEK_DATA || SEEK_HOLE => we must revalidate |
06222e49 JB |
118 | * the cached file length |
119 | */ | |
965c8e59 | 120 | if (whence != SEEK_SET && whence != SEEK_CUR) { |
980802e3 | 121 | struct inode *inode = filp->f_mapping->host; |
d5e66348 | 122 | |
980802e3 TM |
123 | int retval = nfs_revalidate_file_size(inode, filp); |
124 | if (retval < 0) | |
125 | return (loff_t)retval; | |
79835a71 | 126 | } |
d5e66348 | 127 | |
965c8e59 | 128 | return generic_file_llseek(filp, offset, whence); |
980802e3 | 129 | } |
89d77c8f | 130 | EXPORT_SYMBOL_GPL(nfs_file_llseek); |
980802e3 | 131 | |
1da177e4 LT |
132 | /* |
133 | * Flush all dirty pages, and check for write errors. | |
1da177e4 | 134 | */ |
5445b1fb | 135 | static int |
75e1fcc0 | 136 | nfs_file_flush(struct file *file, fl_owner_t id) |
1da177e4 | 137 | { |
6de1472f | 138 | struct inode *inode = file_inode(file); |
67dd23f9 | 139 | errseq_t since; |
1da177e4 | 140 | |
6de1472f | 141 | dprintk("NFS: flush(%pD2)\n", file); |
1da177e4 | 142 | |
c2459dc4 | 143 | nfs_inc_stats(inode, NFSIOS_VFSFLUSH); |
1da177e4 LT |
144 | if ((file->f_mode & FMODE_WRITE) == 0) |
145 | return 0; | |
7b159fc1 | 146 | |
7fe5c398 | 147 | /* Flush writes to the server and return any errors */ |
67dd23f9 SM |
148 | since = filemap_sample_wb_err(file->f_mapping); |
149 | nfs_wb_all(inode); | |
150 | return filemap_check_wb_err(file->f_mapping, since); | |
1da177e4 LT |
151 | } |
152 | ||
ce4ef7c0 | 153 | ssize_t |
3aa2d199 | 154 | nfs_file_read(struct kiocb *iocb, struct iov_iter *to) |
1da177e4 | 155 | { |
6de1472f | 156 | struct inode *inode = file_inode(iocb->ki_filp); |
1da177e4 LT |
157 | ssize_t result; |
158 | ||
2ba48ce5 | 159 | if (iocb->ki_flags & IOCB_DIRECT) |
64158668 | 160 | return nfs_file_direct_read(iocb, to, false); |
1da177e4 | 161 | |
619d30b4 | 162 | dprintk("NFS: read(%pD2, %zu@%lu)\n", |
6de1472f | 163 | iocb->ki_filp, |
3aa2d199 | 164 | iov_iter_count(to), (unsigned long) iocb->ki_pos); |
1da177e4 | 165 | |
a5864c99 TM |
166 | nfs_start_io_read(inode); |
167 | result = nfs_revalidate_mapping(inode, iocb->ki_filp->f_mapping); | |
4184dcf2 | 168 | if (!result) { |
3aa2d199 | 169 | result = generic_file_read_iter(iocb, to); |
4184dcf2 CL |
170 | if (result > 0) |
171 | nfs_add_stats(inode, NFSIOS_NORMALREADBYTES, result); | |
172 | } | |
a5864c99 | 173 | nfs_end_io_read(inode); |
1da177e4 LT |
174 | return result; |
175 | } | |
89d77c8f | 176 | EXPORT_SYMBOL_GPL(nfs_file_read); |
1da177e4 | 177 | |
ce4ef7c0 | 178 | int |
1da177e4 LT |
179 | nfs_file_mmap(struct file * file, struct vm_area_struct * vma) |
180 | { | |
6de1472f | 181 | struct inode *inode = file_inode(file); |
1da177e4 LT |
182 | int status; |
183 | ||
6de1472f | 184 | dprintk("NFS: mmap(%pD2)\n", file); |
1da177e4 | 185 | |
e1ebfd33 TM |
186 | /* Note: generic_file_mmap() returns ENOSYS on nommu systems |
187 | * so we call that before revalidating the mapping | |
188 | */ | |
189 | status = generic_file_mmap(file, vma); | |
94387fb1 TM |
190 | if (!status) { |
191 | vma->vm_ops = &nfs_file_vm_ops; | |
e1ebfd33 | 192 | status = nfs_revalidate_mapping(inode, file->f_mapping); |
94387fb1 | 193 | } |
1da177e4 LT |
194 | return status; |
195 | } | |
89d77c8f | 196 | EXPORT_SYMBOL_GPL(nfs_file_mmap); |
1da177e4 LT |
197 | |
198 | /* | |
199 | * Flush any dirty pages for this process, and check for write errors. | |
200 | * The return status from this call provides a reliable indication of | |
201 | * whether any write errors occurred for this process. | |
202 | */ | |
4ff79bc7 | 203 | static int |
bf4b4905 | 204 | nfs_file_fsync_commit(struct file *file, int datasync) |
1da177e4 | 205 | { |
6de1472f | 206 | struct inode *inode = file_inode(file); |
9641d9bc | 207 | int ret, ret2; |
af7fa165 | 208 | |
6de1472f | 209 | dprintk("NFS: fsync file(%pD2) datasync %d\n", file, datasync); |
1da177e4 | 210 | |
91d5b470 | 211 | nfs_inc_stats(inode, NFSIOS_VFSFSYNC); |
2197e9b0 | 212 | ret = nfs_commit_inode(inode, FLUSH_SYNC); |
9641d9bc TM |
213 | ret2 = file_check_and_advance_wb_err(file); |
214 | if (ret2 < 0) | |
215 | return ret2; | |
216 | return ret; | |
a5c58892 BS |
217 | } |
218 | ||
4ff79bc7 | 219 | int |
a5c58892 BS |
220 | nfs_file_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
221 | { | |
2197e9b0 | 222 | struct nfs_open_context *ctx = nfs_file_open_context(file); |
496ad9aa | 223 | struct inode *inode = file_inode(file); |
2197e9b0 | 224 | int ret; |
a5c58892 | 225 | |
f4ce1299 TM |
226 | trace_nfs_fsync_enter(inode); |
227 | ||
2197e9b0 | 228 | for (;;) { |
6fbda89b | 229 | ret = file_write_and_wait_range(file, start, end); |
05990d1b TM |
230 | if (ret != 0) |
231 | break; | |
bf4b4905 | 232 | ret = nfs_file_fsync_commit(file, datasync); |
2197e9b0 TM |
233 | if (ret != 0) |
234 | break; | |
235 | ret = pnfs_sync_inode(inode, !!datasync); | |
236 | if (ret != 0) | |
237 | break; | |
238 | if (!test_and_clear_bit(NFS_CONTEXT_RESEND_WRITES, &ctx->flags)) | |
239 | break; | |
dcfc4f25 TM |
240 | /* |
241 | * If nfs_file_fsync_commit detected a server reboot, then | |
242 | * resend all dirty pages that might have been covered by | |
243 | * the NFS_CONTEXT_RESEND_WRITES flag | |
244 | */ | |
245 | start = 0; | |
246 | end = LLONG_MAX; | |
2197e9b0 | 247 | } |
05990d1b | 248 | |
f4ce1299 | 249 | trace_nfs_fsync_exit(inode, ret); |
af7fa165 | 250 | return ret; |
1da177e4 | 251 | } |
4ff79bc7 | 252 | EXPORT_SYMBOL_GPL(nfs_file_fsync); |
1da177e4 | 253 | |
38c73044 PS |
254 | /* |
255 | * Decide whether a read/modify/write cycle may be more efficient | |
256 | * then a modify/write/read cycle when writing to a page in the | |
257 | * page cache. | |
258 | * | |
2cde04e9 KI |
259 | * Some pNFS layout drivers can only read/write at a certain block |
260 | * granularity like all block devices and therefore we must perform | |
261 | * read/modify/write whenever a page hasn't read yet and the data | |
262 | * to be written there is not aligned to a block boundary and/or | |
263 | * smaller than the block size. | |
264 | * | |
38c73044 PS |
265 | * The modify/write/read cycle may occur if a page is read before |
266 | * being completely filled by the writer. In this situation, the | |
267 | * page must be completely written to stable storage on the server | |
268 | * before it can be refilled by reading in the page from the server. | |
269 | * This can lead to expensive, small, FILE_SYNC mode writes being | |
270 | * done. | |
271 | * | |
272 | * It may be more efficient to read the page first if the file is | |
273 | * open for reading in addition to writing, the page is not marked | |
274 | * as Uptodate, it is not dirty or waiting to be committed, | |
275 | * indicating that it was previously allocated and then modified, | |
276 | * that there were valid bytes of data in that range of the file, | |
277 | * and that the new data won't completely replace the old data in | |
278 | * that range of the file. | |
279 | */ | |
2cde04e9 | 280 | static bool nfs_full_page_write(struct page *page, loff_t pos, unsigned int len) |
38c73044 PS |
281 | { |
282 | unsigned int pglen = nfs_page_length(page); | |
09cbfeaf | 283 | unsigned int offset = pos & (PAGE_SIZE - 1); |
38c73044 PS |
284 | unsigned int end = offset + len; |
285 | ||
2cde04e9 KI |
286 | return !pglen || (end >= pglen && !offset); |
287 | } | |
612aa983 | 288 | |
2cde04e9 KI |
289 | static bool nfs_want_read_modify_write(struct file *file, struct page *page, |
290 | loff_t pos, unsigned int len) | |
291 | { | |
292 | /* | |
293 | * Up-to-date pages, those with ongoing or full-page write | |
294 | * don't need read/modify/write | |
295 | */ | |
296 | if (PageUptodate(page) || PagePrivate(page) || | |
297 | nfs_full_page_write(page, pos, len)) | |
298 | return false; | |
299 | ||
300 | if (pnfs_ld_read_whole_page(file->f_mapping->host)) | |
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 | ||
1da177e4 | 308 | /* |
4899f9c8 NP |
309 | * This does the "real" work of the write. We must allocate and lock the |
310 | * page to be sent back to the generic routine, which then copies the | |
311 | * data from user space. | |
1da177e4 LT |
312 | * |
313 | * If the writer ends up delaying the write, the writer needs to | |
314 | * increment the page use counts until he is done with the page. | |
315 | */ | |
4899f9c8 NP |
316 | static int nfs_write_begin(struct file *file, struct address_space *mapping, |
317 | loff_t pos, unsigned len, unsigned flags, | |
318 | struct page **pagep, void **fsdata) | |
1da177e4 | 319 | { |
4899f9c8 | 320 | int ret; |
09cbfeaf | 321 | pgoff_t index = pos >> PAGE_SHIFT; |
4899f9c8 | 322 | struct page *page; |
38c73044 | 323 | int once_thru = 0; |
4899f9c8 | 324 | |
1e8968c5 | 325 | dfprintk(PAGECACHE, "NFS: write_begin(%pD2(%lu), %u@%lld)\n", |
6de1472f | 326 | file, mapping->host->i_ino, len, (long long) pos); |
b7eaefaa | 327 | |
38c73044 | 328 | start: |
54566b2c | 329 | page = grab_cache_page_write_begin(mapping, index, flags); |
4899f9c8 NP |
330 | if (!page) |
331 | return -ENOMEM; | |
332 | *pagep = page; | |
333 | ||
334 | ret = nfs_flush_incompatible(file, page); | |
335 | if (ret) { | |
336 | unlock_page(page); | |
09cbfeaf | 337 | put_page(page); |
38c73044 PS |
338 | } else if (!once_thru && |
339 | nfs_want_read_modify_write(file, page, pos, len)) { | |
340 | once_thru = 1; | |
341 | ret = nfs_readpage(file, page); | |
09cbfeaf | 342 | put_page(page); |
38c73044 PS |
343 | if (!ret) |
344 | goto start; | |
4899f9c8 NP |
345 | } |
346 | return ret; | |
1da177e4 LT |
347 | } |
348 | ||
4899f9c8 NP |
349 | static int nfs_write_end(struct file *file, struct address_space *mapping, |
350 | loff_t pos, unsigned len, unsigned copied, | |
351 | struct page *page, void *fsdata) | |
1da177e4 | 352 | { |
09cbfeaf | 353 | unsigned offset = pos & (PAGE_SIZE - 1); |
dc24826b | 354 | struct nfs_open_context *ctx = nfs_file_open_context(file); |
4899f9c8 | 355 | int status; |
1da177e4 | 356 | |
1e8968c5 | 357 | dfprintk(PAGECACHE, "NFS: write_end(%pD2(%lu), %u@%lld)\n", |
6de1472f | 358 | file, mapping->host->i_ino, len, (long long) pos); |
b7eaefaa | 359 | |
efc91ed0 TM |
360 | /* |
361 | * Zero any uninitialised parts of the page, and then mark the page | |
362 | * as up to date if it turns out that we're extending the file. | |
363 | */ | |
364 | if (!PageUptodate(page)) { | |
365 | unsigned pglen = nfs_page_length(page); | |
c0cf3ef5 | 366 | unsigned end = offset + copied; |
efc91ed0 TM |
367 | |
368 | if (pglen == 0) { | |
369 | zero_user_segments(page, 0, offset, | |
09cbfeaf | 370 | end, PAGE_SIZE); |
efc91ed0 TM |
371 | SetPageUptodate(page); |
372 | } else if (end >= pglen) { | |
09cbfeaf | 373 | zero_user_segment(page, end, PAGE_SIZE); |
efc91ed0 TM |
374 | if (offset == 0) |
375 | SetPageUptodate(page); | |
376 | } else | |
09cbfeaf | 377 | zero_user_segment(page, pglen, PAGE_SIZE); |
efc91ed0 TM |
378 | } |
379 | ||
4899f9c8 | 380 | status = nfs_updatepage(file, page, offset, copied); |
4899f9c8 NP |
381 | |
382 | unlock_page(page); | |
09cbfeaf | 383 | put_page(page); |
4899f9c8 | 384 | |
3d509e54 CL |
385 | if (status < 0) |
386 | return status; | |
2701d086 | 387 | NFS_I(mapping->host)->write_io += copied; |
dc24826b | 388 | |
d95b2665 TM |
389 | if (nfs_ctx_key_to_expire(ctx, mapping->host)) |
390 | nfs_wb_all(mapping->host); | |
dc24826b | 391 | |
3d509e54 | 392 | return copied; |
1da177e4 LT |
393 | } |
394 | ||
6b9b3514 DH |
395 | /* |
396 | * Partially or wholly invalidate a page | |
397 | * - Release the private state associated with a page if undergoing complete | |
398 | * page invalidation | |
545db45f | 399 | * - Called if either PG_private or PG_fscache is set on the page |
6b9b3514 DH |
400 | * - Caller holds page lock |
401 | */ | |
6d740c76 MWO |
402 | static void nfs_invalidate_folio(struct folio *folio, size_t offset, |
403 | size_t length) | |
cd52ed35 | 404 | { |
6d740c76 MWO |
405 | dfprintk(PAGECACHE, "NFS: invalidate_folio(%lu, %zu, %zu)\n", |
406 | folio->index, offset, length); | |
b7eaefaa | 407 | |
6d740c76 | 408 | if (offset != 0 || length < folio_size(folio)) |
1c75950b | 409 | return; |
d2ccddf0 | 410 | /* Cancel any unstarted writes on this page */ |
6d740c76 MWO |
411 | nfs_wb_folio_cancel(folio->mapping->host, folio); |
412 | folio_wait_fscache(folio); | |
cd52ed35 TM |
413 | } |
414 | ||
6b9b3514 DH |
415 | /* |
416 | * Attempt to release the private state associated with a page | |
545db45f | 417 | * - Called if either PG_private or PG_fscache is set on the page |
6b9b3514 DH |
418 | * - Caller holds page lock |
419 | * - Return true (may release page) or false (may not) | |
420 | */ | |
cd52ed35 TM |
421 | static int nfs_release_page(struct page *page, gfp_t gfp) |
422 | { | |
b7eaefaa CL |
423 | dfprintk(PAGECACHE, "NFS: release_page(%p)\n", page); |
424 | ||
e3db7691 | 425 | /* If PagePrivate() is set, then the page is not freeable */ |
545db45f DH |
426 | if (PagePrivate(page)) |
427 | return 0; | |
428 | return nfs_fscache_release_page(page, gfp); | |
e3db7691 TM |
429 | } |
430 | ||
f919b196 MG |
431 | static void nfs_check_dirty_writeback(struct page *page, |
432 | bool *dirty, bool *writeback) | |
433 | { | |
434 | struct nfs_inode *nfsi; | |
435 | struct address_space *mapping = page_file_mapping(page); | |
436 | ||
437 | if (!mapping || PageSwapCache(page)) | |
438 | return; | |
439 | ||
440 | /* | |
441 | * Check if an unstable page is currently being committed and | |
442 | * if so, have the VM treat it as if the page is under writeback | |
443 | * so it will not block due to pages that will shortly be freeable. | |
444 | */ | |
445 | nfsi = NFS_I(mapping->host); | |
af7cf057 | 446 | if (atomic_read(&nfsi->commit_info.rpcs_out)) { |
f919b196 MG |
447 | *writeback = true; |
448 | return; | |
449 | } | |
450 | ||
451 | /* | |
452 | * If PagePrivate() is set, then the page is not freeable and as the | |
453 | * inode is not being committed, it's not going to be cleaned in the | |
454 | * near future so treat it as dirty | |
455 | */ | |
456 | if (PagePrivate(page)) | |
457 | *dirty = true; | |
458 | } | |
459 | ||
6b9b3514 DH |
460 | /* |
461 | * Attempt to clear the private state associated with a page when an error | |
462 | * occurs that requires the cached contents of an inode to be written back or | |
463 | * destroyed | |
545db45f | 464 | * - Called if either PG_private or fscache is set on the page |
6b9b3514 DH |
465 | * - Caller holds page lock |
466 | * - Return 0 if successful, -error otherwise | |
467 | */ | |
15a30ab2 | 468 | static int nfs_launder_folio(struct folio *folio) |
e3db7691 | 469 | { |
15a30ab2 | 470 | struct inode *inode = folio->mapping->host; |
b7eaefaa | 471 | |
15a30ab2 MWO |
472 | dfprintk(PAGECACHE, "NFS: launder_folio(%ld, %llu)\n", |
473 | inode->i_ino, folio_pos(folio)); | |
b7eaefaa | 474 | |
15a30ab2 MWO |
475 | folio_wait_fscache(folio); |
476 | return nfs_wb_page(inode, &folio->page); | |
cd52ed35 TM |
477 | } |
478 | ||
a564b8f0 MG |
479 | static int nfs_swap_activate(struct swap_info_struct *sis, struct file *file, |
480 | sector_t *span) | |
481 | { | |
bd89bc67 MZ |
482 | unsigned long blocks; |
483 | long long isize; | |
4dc73c67 N |
484 | struct inode *inode = file_inode(file); |
485 | struct rpc_clnt *clnt = NFS_CLIENT(inode); | |
486 | struct nfs_client *cl = NFS_SERVER(inode)->nfs_client; | |
bd89bc67 MZ |
487 | |
488 | spin_lock(&inode->i_lock); | |
489 | blocks = inode->i_blocks; | |
490 | isize = inode->i_size; | |
491 | spin_unlock(&inode->i_lock); | |
492 | if (blocks*512 < isize) { | |
493 | pr_warn("swap activate: swapfile has holes\n"); | |
494 | return -EINVAL; | |
495 | } | |
dad2b015 | 496 | |
a564b8f0 | 497 | *span = sis->pages; |
dad2b015 | 498 | |
4dc73c67 N |
499 | |
500 | if (cl->rpc_ops->enable_swap) | |
501 | cl->rpc_ops->enable_swap(inode); | |
502 | ||
3c87ef6e | 503 | return rpc_clnt_swap_activate(clnt); |
a564b8f0 MG |
504 | } |
505 | ||
506 | static void nfs_swap_deactivate(struct file *file) | |
507 | { | |
4dc73c67 N |
508 | struct inode *inode = file_inode(file); |
509 | struct rpc_clnt *clnt = NFS_CLIENT(inode); | |
510 | struct nfs_client *cl = NFS_SERVER(inode)->nfs_client; | |
dad2b015 | 511 | |
3c87ef6e | 512 | rpc_clnt_swap_deactivate(clnt); |
4dc73c67 N |
513 | if (cl->rpc_ops->disable_swap) |
514 | cl->rpc_ops->disable_swap(file_inode(file)); | |
a564b8f0 | 515 | } |
a564b8f0 | 516 | |
f5e54d6e | 517 | const struct address_space_operations nfs_file_aops = { |
1da177e4 | 518 | .readpage = nfs_readpage, |
8786fde8 | 519 | .readahead = nfs_readahead, |
187c82cb | 520 | .dirty_folio = filemap_dirty_folio, |
1da177e4 LT |
521 | .writepage = nfs_writepage, |
522 | .writepages = nfs_writepages, | |
4899f9c8 NP |
523 | .write_begin = nfs_write_begin, |
524 | .write_end = nfs_write_end, | |
6d740c76 | 525 | .invalidate_folio = nfs_invalidate_folio, |
cd52ed35 | 526 | .releasepage = nfs_release_page, |
1da177e4 | 527 | .direct_IO = nfs_direct_IO, |
f844cd0d | 528 | #ifdef CONFIG_MIGRATION |
074cc1de | 529 | .migratepage = nfs_migrate_page, |
f844cd0d | 530 | #endif |
15a30ab2 | 531 | .launder_folio = nfs_launder_folio, |
f919b196 | 532 | .is_dirty_writeback = nfs_check_dirty_writeback, |
f590f333 | 533 | .error_remove_page = generic_error_remove_page, |
a564b8f0 MG |
534 | .swap_activate = nfs_swap_activate, |
535 | .swap_deactivate = nfs_swap_deactivate, | |
1da177e4 LT |
536 | }; |
537 | ||
6b9b3514 DH |
538 | /* |
539 | * Notification that a PTE pointing to an NFS page is about to be made | |
540 | * writable, implying that someone is about to modify the page through a | |
541 | * shared-writable mapping | |
542 | */ | |
01a36844 | 543 | static vm_fault_t nfs_vm_page_mkwrite(struct vm_fault *vmf) |
94387fb1 | 544 | { |
c2ec175c | 545 | struct page *page = vmf->page; |
11bac800 | 546 | struct file *filp = vmf->vma->vm_file; |
6de1472f | 547 | struct inode *inode = file_inode(filp); |
94387fb1 | 548 | unsigned pagelen; |
01a36844 | 549 | vm_fault_t ret = VM_FAULT_NOPAGE; |
4899f9c8 | 550 | struct address_space *mapping; |
94387fb1 | 551 | |
1e8968c5 | 552 | dfprintk(PAGECACHE, "NFS: vm_page_mkwrite(%pD2(%lu), offset %lld)\n", |
6de1472f | 553 | filp, filp->f_mapping->host->i_ino, |
b7eaefaa CL |
554 | (long long)page_offset(page)); |
555 | ||
9a773e7c TM |
556 | sb_start_pagefault(inode->i_sb); |
557 | ||
545db45f | 558 | /* make sure the cache has finished storing the page */ |
a6b5a28e DW |
559 | if (PageFsCache(page) && |
560 | wait_on_page_fscache_killable(vmf->page) < 0) { | |
561 | ret = VM_FAULT_RETRY; | |
562 | goto out; | |
563 | } | |
545db45f | 564 | |
ef070dcb TM |
565 | wait_on_bit_action(&NFS_I(inode)->flags, NFS_INO_INVALIDATING, |
566 | nfs_wait_bit_killable, TASK_KILLABLE); | |
567 | ||
94387fb1 | 568 | lock_page(page); |
d56b4ddf | 569 | mapping = page_file_mapping(page); |
6de1472f | 570 | if (mapping != inode->i_mapping) |
8b1f9ee5 TM |
571 | goto out_unlock; |
572 | ||
2aeb98f4 TM |
573 | wait_on_page_writeback(page); |
574 | ||
94387fb1 | 575 | pagelen = nfs_page_length(page); |
8b1f9ee5 TM |
576 | if (pagelen == 0) |
577 | goto out_unlock; | |
4899f9c8 | 578 | |
bc4866b6 TM |
579 | ret = VM_FAULT_LOCKED; |
580 | if (nfs_flush_incompatible(filp, page) == 0 && | |
581 | nfs_updatepage(filp, page, 0, pagelen) == 0) | |
582 | goto out; | |
8b1f9ee5 | 583 | |
bc4866b6 | 584 | ret = VM_FAULT_SIGBUS; |
8b1f9ee5 TM |
585 | out_unlock: |
586 | unlock_page(page); | |
bc4866b6 | 587 | out: |
9a773e7c | 588 | sb_end_pagefault(inode->i_sb); |
bc4866b6 | 589 | return ret; |
94387fb1 TM |
590 | } |
591 | ||
f0f37e2f | 592 | static const struct vm_operations_struct nfs_file_vm_ops = { |
94387fb1 | 593 | .fault = filemap_fault, |
f1820361 | 594 | .map_pages = filemap_map_pages, |
94387fb1 TM |
595 | .page_mkwrite = nfs_vm_page_mkwrite, |
596 | }; | |
597 | ||
edaf4369 | 598 | ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 599 | { |
6de1472f AV |
600 | struct file *file = iocb->ki_filp; |
601 | struct inode *inode = file_inode(file); | |
ed7bcdb3 TM |
602 | unsigned int mntflags = NFS_SERVER(inode)->flags; |
603 | ssize_t result, written; | |
ce368536 SM |
604 | errseq_t since; |
605 | int error; | |
1da177e4 | 606 | |
6de1472f | 607 | result = nfs_key_timeout_notify(file, inode); |
dc24826b AA |
608 | if (result) |
609 | return result; | |
610 | ||
89698b24 | 611 | if (iocb->ki_flags & IOCB_DIRECT) |
64158668 | 612 | return nfs_file_direct_write(iocb, from, false); |
1da177e4 | 613 | |
619d30b4 | 614 | dprintk("NFS: write(%pD2, %zu@%Ld)\n", |
18290650 | 615 | file, iov_iter_count(from), (long long) iocb->ki_pos); |
1da177e4 | 616 | |
1da177e4 LT |
617 | if (IS_SWAPFILE(inode)) |
618 | goto out_swapfile; | |
7d52e862 TM |
619 | /* |
620 | * O_APPEND implies that we must revalidate the file length. | |
621 | */ | |
fc9dc401 | 622 | if (iocb->ki_flags & IOCB_APPEND || iocb->ki_pos > i_size_read(inode)) { |
6de1472f | 623 | result = nfs_revalidate_file_size(inode, file); |
7d52e862 | 624 | if (result) |
e6005436 | 625 | return result; |
fe51beec | 626 | } |
1da177e4 | 627 | |
28aa2f9e TM |
628 | nfs_clear_invalid_mapping(file->f_mapping); |
629 | ||
ce368536 | 630 | since = filemap_sample_wb_err(file->f_mapping); |
a5864c99 | 631 | nfs_start_io_write(inode); |
18290650 TM |
632 | result = generic_write_checks(iocb, from); |
633 | if (result > 0) { | |
634 | current->backing_dev_info = inode_to_bdi(inode); | |
800ba295 | 635 | result = generic_perform_write(iocb, from); |
18290650 TM |
636 | current->backing_dev_info = NULL; |
637 | } | |
a5864c99 | 638 | nfs_end_io_write(inode); |
18290650 | 639 | if (result <= 0) |
1da177e4 LT |
640 | goto out; |
641 | ||
c49edecd | 642 | written = result; |
18290650 | 643 | iocb->ki_pos += written; |
e6005436 | 644 | nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written); |
ed7bcdb3 TM |
645 | |
646 | if (mntflags & NFS_MOUNT_WRITE_EAGER) { | |
647 | result = filemap_fdatawrite_range(file->f_mapping, | |
648 | iocb->ki_pos - written, | |
649 | iocb->ki_pos - 1); | |
650 | if (result < 0) | |
651 | goto out; | |
652 | } | |
653 | if (mntflags & NFS_MOUNT_WRITE_WAIT) { | |
654 | result = filemap_fdatawait_range(file->f_mapping, | |
655 | iocb->ki_pos - written, | |
656 | iocb->ki_pos - 1); | |
657 | if (result < 0) | |
658 | goto out; | |
659 | } | |
e973b1a5 | 660 | result = generic_write_sync(iocb, written); |
661 | if (result < 0) | |
e6005436 | 662 | return result; |
7e381172 | 663 | |
e6005436 | 664 | out: |
7e94d6c4 | 665 | /* Return error values */ |
ce368536 | 666 | error = filemap_check_wb_err(file->f_mapping, since); |
e6005436 TM |
667 | switch (error) { |
668 | default: | |
669 | break; | |
670 | case -EDQUOT: | |
671 | case -EFBIG: | |
672 | case -ENOSPC: | |
673 | nfs_wb_all(inode); | |
674 | error = file_check_and_advance_wb_err(file); | |
675 | if (error < 0) | |
676 | result = error; | |
200baa21 | 677 | } |
1da177e4 LT |
678 | return result; |
679 | ||
680 | out_swapfile: | |
681 | printk(KERN_INFO "NFS: attempt to write to active swap file!\n"); | |
89658c4d | 682 | return -ETXTBSY; |
1da177e4 | 683 | } |
89d77c8f | 684 | EXPORT_SYMBOL_GPL(nfs_file_write); |
1da177e4 | 685 | |
5eebde23 SJ |
686 | static int |
687 | do_getlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
688 | { |
689 | struct inode *inode = filp->f_mapping->host; | |
690 | int status = 0; | |
21ac19d4 | 691 | unsigned int saved_type = fl->fl_type; |
1da177e4 | 692 | |
039c4d7a | 693 | /* Try local locking first */ |
6d34ac19 BF |
694 | posix_test_lock(filp, fl); |
695 | if (fl->fl_type != F_UNLCK) { | |
696 | /* found a conflict */ | |
039c4d7a | 697 | goto out; |
1da177e4 | 698 | } |
21ac19d4 | 699 | fl->fl_type = saved_type; |
039c4d7a | 700 | |
011e2a7f | 701 | if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) |
039c4d7a TM |
702 | goto out_noconflict; |
703 | ||
5eebde23 | 704 | if (is_local) |
039c4d7a TM |
705 | goto out_noconflict; |
706 | ||
707 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); | |
708 | out: | |
1da177e4 | 709 | return status; |
039c4d7a TM |
710 | out_noconflict: |
711 | fl->fl_type = F_UNLCK; | |
712 | goto out; | |
1da177e4 LT |
713 | } |
714 | ||
5eebde23 SJ |
715 | static int |
716 | do_unlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
717 | { |
718 | struct inode *inode = filp->f_mapping->host; | |
7a8203d8 | 719 | struct nfs_lock_context *l_ctx; |
1da177e4 LT |
720 | int status; |
721 | ||
1da177e4 LT |
722 | /* |
723 | * Flush all pending writes before doing anything | |
724 | * with locks.. | |
725 | */ | |
aded8d7b | 726 | nfs_wb_all(inode); |
1da177e4 | 727 | |
7a8203d8 TM |
728 | l_ctx = nfs_get_lock_context(nfs_file_open_context(filp)); |
729 | if (!IS_ERR(l_ctx)) { | |
210c7c17 | 730 | status = nfs_iocounter_wait(l_ctx); |
7a8203d8 | 731 | nfs_put_lock_context(l_ctx); |
f30cb757 BC |
732 | /* NOTE: special case |
733 | * If we're signalled while cleaning up locks on process exit, we | |
734 | * still need to complete the unlock. | |
735 | */ | |
736 | if (status < 0 && !(fl->fl_flags & FL_CLOSE)) | |
7a8203d8 TM |
737 | return status; |
738 | } | |
739 | ||
5eebde23 SJ |
740 | /* |
741 | * Use local locking if mounted with "-onolock" or with appropriate | |
742 | * "-olocal_lock=" | |
743 | */ | |
744 | if (!is_local) | |
1da177e4 LT |
745 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); |
746 | else | |
75575ddf | 747 | status = locks_lock_file_wait(filp, fl); |
1da177e4 LT |
748 | return status; |
749 | } | |
750 | ||
5eebde23 SJ |
751 | static int |
752 | do_setlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
753 | { |
754 | struct inode *inode = filp->f_mapping->host; | |
1da177e4 LT |
755 | int status; |
756 | ||
1da177e4 LT |
757 | /* |
758 | * Flush all pending writes before doing anything | |
759 | * with locks.. | |
760 | */ | |
29884df0 TM |
761 | status = nfs_sync_mapping(filp->f_mapping); |
762 | if (status != 0) | |
1da177e4 LT |
763 | goto out; |
764 | ||
5eebde23 SJ |
765 | /* |
766 | * Use local locking if mounted with "-onolock" or with appropriate | |
767 | * "-olocal_lock=" | |
768 | */ | |
769 | if (!is_local) | |
1da177e4 | 770 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); |
c4d7c402 | 771 | else |
75575ddf | 772 | status = locks_lock_file_wait(filp, fl); |
1da177e4 LT |
773 | if (status < 0) |
774 | goto out; | |
6b96724e | 775 | |
1da177e4 | 776 | /* |
779eafab N |
777 | * Invalidate cache to prevent missing any changes. If |
778 | * the file is mapped, clear the page cache as well so | |
779 | * those mappings will be loaded. | |
6b96724e | 780 | * |
1da177e4 LT |
781 | * This makes locking act as a cache coherency point. |
782 | */ | |
29884df0 | 783 | nfs_sync_mapping(filp->f_mapping); |
779eafab | 784 | if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) { |
442ce049 | 785 | nfs_zap_caches(inode); |
779eafab N |
786 | if (mapping_mapped(filp->f_mapping)) |
787 | nfs_revalidate_mapping(inode, filp->f_mapping); | |
788 | } | |
1da177e4 | 789 | out: |
1da177e4 LT |
790 | return status; |
791 | } | |
792 | ||
793 | /* | |
794 | * Lock a (portion of) a file | |
795 | */ | |
ce4ef7c0 | 796 | int nfs_lock(struct file *filp, int cmd, struct file_lock *fl) |
1da177e4 | 797 | { |
6da24bc9 | 798 | struct inode *inode = filp->f_mapping->host; |
2116271a | 799 | int ret = -ENOLCK; |
5eebde23 | 800 | int is_local = 0; |
1da177e4 | 801 | |
6de1472f AV |
802 | dprintk("NFS: lock(%pD2, t=%x, fl=%x, r=%lld:%lld)\n", |
803 | filp, fl->fl_type, fl->fl_flags, | |
1da177e4 | 804 | (long long)fl->fl_start, (long long)fl->fl_end); |
6da24bc9 | 805 | |
91d5b470 | 806 | nfs_inc_stats(inode, NFSIOS_VFSLOCK); |
1da177e4 | 807 | |
bb0a55bb BF |
808 | if (fl->fl_flags & FL_RECLAIM) |
809 | return -ENOGRACE; | |
810 | ||
5eebde23 SJ |
811 | if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FCNTL) |
812 | is_local = 1; | |
813 | ||
2116271a TM |
814 | if (NFS_PROTO(inode)->lock_check_bounds != NULL) { |
815 | ret = NFS_PROTO(inode)->lock_check_bounds(fl); | |
816 | if (ret < 0) | |
817 | goto out_err; | |
818 | } | |
1da177e4 LT |
819 | |
820 | if (IS_GETLK(cmd)) | |
5eebde23 | 821 | ret = do_getlk(filp, cmd, fl, is_local); |
2116271a | 822 | else if (fl->fl_type == F_UNLCK) |
5eebde23 | 823 | ret = do_unlk(filp, cmd, fl, is_local); |
2116271a | 824 | else |
5eebde23 | 825 | ret = do_setlk(filp, cmd, fl, is_local); |
2116271a TM |
826 | out_err: |
827 | return ret; | |
1da177e4 | 828 | } |
89d77c8f | 829 | EXPORT_SYMBOL_GPL(nfs_lock); |
1da177e4 LT |
830 | |
831 | /* | |
832 | * Lock a (portion of) a file | |
833 | */ | |
ce4ef7c0 | 834 | int nfs_flock(struct file *filp, int cmd, struct file_lock *fl) |
1da177e4 | 835 | { |
5eebde23 SJ |
836 | struct inode *inode = filp->f_mapping->host; |
837 | int is_local = 0; | |
838 | ||
6de1472f AV |
839 | dprintk("NFS: flock(%pD2, t=%x, fl=%x)\n", |
840 | filp, fl->fl_type, fl->fl_flags); | |
1da177e4 | 841 | |
1da177e4 LT |
842 | if (!(fl->fl_flags & FL_FLOCK)) |
843 | return -ENOLCK; | |
844 | ||
5eebde23 SJ |
845 | if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK) |
846 | is_local = 1; | |
847 | ||
fcfa4470 BC |
848 | /* We're simulating flock() locks using posix locks on the server */ |
849 | if (fl->fl_type == F_UNLCK) | |
5eebde23 SJ |
850 | return do_unlk(filp, cmd, fl, is_local); |
851 | return do_setlk(filp, cmd, fl, is_local); | |
1da177e4 | 852 | } |
89d77c8f | 853 | EXPORT_SYMBOL_GPL(nfs_flock); |
370f6599 | 854 | |
0486958f JL |
855 | const struct file_operations nfs_file_operations = { |
856 | .llseek = nfs_file_llseek, | |
3aa2d199 | 857 | .read_iter = nfs_file_read, |
edaf4369 | 858 | .write_iter = nfs_file_write, |
0486958f JL |
859 | .mmap = nfs_file_mmap, |
860 | .open = nfs_file_open, | |
861 | .flush = nfs_file_flush, | |
862 | .release = nfs_file_release, | |
863 | .fsync = nfs_file_fsync, | |
864 | .lock = nfs_lock, | |
865 | .flock = nfs_flock, | |
82c156f8 | 866 | .splice_read = generic_file_splice_read, |
4da54c21 | 867 | .splice_write = iter_file_splice_write, |
0486958f | 868 | .check_flags = nfs_check_flags, |
1c994a09 | 869 | .setlease = simple_nosetlease, |
0486958f | 870 | }; |
ddda8e0a | 871 | EXPORT_SYMBOL_GPL(nfs_file_operations); |