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) |
c8b8e32d | 160 | return nfs_file_direct_read(iocb, to); |
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); |
2197e9b0 | 207 | int ret; |
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 TM |
212 | ret = nfs_commit_inode(inode, FLUSH_SYNC); |
213 | if (ret < 0) | |
214 | return ret; | |
215 | return file_check_and_advance_wb_err(file); | |
a5c58892 BS |
216 | } |
217 | ||
4ff79bc7 | 218 | int |
a5c58892 BS |
219 | nfs_file_fsync(struct file *file, loff_t start, loff_t end, int datasync) |
220 | { | |
2197e9b0 | 221 | struct nfs_open_context *ctx = nfs_file_open_context(file); |
496ad9aa | 222 | struct inode *inode = file_inode(file); |
2197e9b0 | 223 | int ret; |
a5c58892 | 224 | |
f4ce1299 TM |
225 | trace_nfs_fsync_enter(inode); |
226 | ||
2197e9b0 | 227 | for (;;) { |
6fbda89b | 228 | ret = file_write_and_wait_range(file, start, end); |
05990d1b TM |
229 | if (ret != 0) |
230 | break; | |
bf4b4905 | 231 | ret = nfs_file_fsync_commit(file, datasync); |
2197e9b0 TM |
232 | if (ret != 0) |
233 | break; | |
234 | ret = pnfs_sync_inode(inode, !!datasync); | |
235 | if (ret != 0) | |
236 | break; | |
237 | if (!test_and_clear_bit(NFS_CONTEXT_RESEND_WRITES, &ctx->flags)) | |
238 | break; | |
dcfc4f25 TM |
239 | /* |
240 | * If nfs_file_fsync_commit detected a server reboot, then | |
241 | * resend all dirty pages that might have been covered by | |
242 | * the NFS_CONTEXT_RESEND_WRITES flag | |
243 | */ | |
244 | start = 0; | |
245 | end = LLONG_MAX; | |
2197e9b0 | 246 | } |
05990d1b | 247 | |
f4ce1299 | 248 | trace_nfs_fsync_exit(inode, ret); |
af7fa165 | 249 | return ret; |
1da177e4 | 250 | } |
4ff79bc7 | 251 | EXPORT_SYMBOL_GPL(nfs_file_fsync); |
1da177e4 | 252 | |
38c73044 PS |
253 | /* |
254 | * Decide whether a read/modify/write cycle may be more efficient | |
255 | * then a modify/write/read cycle when writing to a page in the | |
256 | * page cache. | |
257 | * | |
2cde04e9 KI |
258 | * Some pNFS layout drivers can only read/write at a certain block |
259 | * granularity like all block devices and therefore we must perform | |
260 | * read/modify/write whenever a page hasn't read yet and the data | |
261 | * to be written there is not aligned to a block boundary and/or | |
262 | * smaller than the block size. | |
263 | * | |
38c73044 PS |
264 | * The modify/write/read cycle may occur if a page is read before |
265 | * being completely filled by the writer. In this situation, the | |
266 | * page must be completely written to stable storage on the server | |
267 | * before it can be refilled by reading in the page from the server. | |
268 | * This can lead to expensive, small, FILE_SYNC mode writes being | |
269 | * done. | |
270 | * | |
271 | * It may be more efficient to read the page first if the file is | |
272 | * open for reading in addition to writing, the page is not marked | |
273 | * as Uptodate, it is not dirty or waiting to be committed, | |
274 | * indicating that it was previously allocated and then modified, | |
275 | * that there were valid bytes of data in that range of the file, | |
276 | * and that the new data won't completely replace the old data in | |
277 | * that range of the file. | |
278 | */ | |
2cde04e9 | 279 | static bool nfs_full_page_write(struct page *page, loff_t pos, unsigned int len) |
38c73044 PS |
280 | { |
281 | unsigned int pglen = nfs_page_length(page); | |
09cbfeaf | 282 | unsigned int offset = pos & (PAGE_SIZE - 1); |
38c73044 PS |
283 | unsigned int end = offset + len; |
284 | ||
2cde04e9 KI |
285 | return !pglen || (end >= pglen && !offset); |
286 | } | |
612aa983 | 287 | |
2cde04e9 KI |
288 | static bool nfs_want_read_modify_write(struct file *file, struct page *page, |
289 | loff_t pos, unsigned int len) | |
290 | { | |
291 | /* | |
292 | * Up-to-date pages, those with ongoing or full-page write | |
293 | * don't need read/modify/write | |
294 | */ | |
295 | if (PageUptodate(page) || PagePrivate(page) || | |
296 | nfs_full_page_write(page, pos, len)) | |
297 | return false; | |
298 | ||
299 | if (pnfs_ld_read_whole_page(file->f_mapping->host)) | |
300 | return true; | |
301 | /* Open for reading too? */ | |
302 | if (file->f_mode & FMODE_READ) | |
303 | return true; | |
304 | return false; | |
38c73044 PS |
305 | } |
306 | ||
1da177e4 | 307 | /* |
4899f9c8 NP |
308 | * This does the "real" work of the write. We must allocate and lock the |
309 | * page to be sent back to the generic routine, which then copies the | |
310 | * data from user space. | |
1da177e4 LT |
311 | * |
312 | * If the writer ends up delaying the write, the writer needs to | |
313 | * increment the page use counts until he is done with the page. | |
314 | */ | |
4899f9c8 NP |
315 | static int nfs_write_begin(struct file *file, struct address_space *mapping, |
316 | loff_t pos, unsigned len, unsigned flags, | |
317 | struct page **pagep, void **fsdata) | |
1da177e4 | 318 | { |
4899f9c8 | 319 | int ret; |
09cbfeaf | 320 | pgoff_t index = pos >> PAGE_SHIFT; |
4899f9c8 | 321 | struct page *page; |
38c73044 | 322 | int once_thru = 0; |
4899f9c8 | 323 | |
1e8968c5 | 324 | dfprintk(PAGECACHE, "NFS: write_begin(%pD2(%lu), %u@%lld)\n", |
6de1472f | 325 | file, mapping->host->i_ino, len, (long long) pos); |
b7eaefaa | 326 | |
38c73044 | 327 | start: |
54566b2c | 328 | page = grab_cache_page_write_begin(mapping, index, flags); |
4899f9c8 NP |
329 | if (!page) |
330 | return -ENOMEM; | |
331 | *pagep = page; | |
332 | ||
333 | ret = nfs_flush_incompatible(file, page); | |
334 | if (ret) { | |
335 | unlock_page(page); | |
09cbfeaf | 336 | put_page(page); |
38c73044 PS |
337 | } else if (!once_thru && |
338 | nfs_want_read_modify_write(file, page, pos, len)) { | |
339 | once_thru = 1; | |
340 | ret = nfs_readpage(file, page); | |
09cbfeaf | 341 | put_page(page); |
38c73044 PS |
342 | if (!ret) |
343 | goto start; | |
4899f9c8 NP |
344 | } |
345 | return ret; | |
1da177e4 LT |
346 | } |
347 | ||
4899f9c8 NP |
348 | static int nfs_write_end(struct file *file, struct address_space *mapping, |
349 | loff_t pos, unsigned len, unsigned copied, | |
350 | struct page *page, void *fsdata) | |
1da177e4 | 351 | { |
09cbfeaf | 352 | unsigned offset = pos & (PAGE_SIZE - 1); |
dc24826b | 353 | struct nfs_open_context *ctx = nfs_file_open_context(file); |
4899f9c8 | 354 | int status; |
1da177e4 | 355 | |
1e8968c5 | 356 | dfprintk(PAGECACHE, "NFS: write_end(%pD2(%lu), %u@%lld)\n", |
6de1472f | 357 | file, mapping->host->i_ino, len, (long long) pos); |
b7eaefaa | 358 | |
efc91ed0 TM |
359 | /* |
360 | * Zero any uninitialised parts of the page, and then mark the page | |
361 | * as up to date if it turns out that we're extending the file. | |
362 | */ | |
363 | if (!PageUptodate(page)) { | |
364 | unsigned pglen = nfs_page_length(page); | |
c0cf3ef5 | 365 | unsigned end = offset + copied; |
efc91ed0 TM |
366 | |
367 | if (pglen == 0) { | |
368 | zero_user_segments(page, 0, offset, | |
09cbfeaf | 369 | end, PAGE_SIZE); |
efc91ed0 TM |
370 | SetPageUptodate(page); |
371 | } else if (end >= pglen) { | |
09cbfeaf | 372 | zero_user_segment(page, end, PAGE_SIZE); |
efc91ed0 TM |
373 | if (offset == 0) |
374 | SetPageUptodate(page); | |
375 | } else | |
09cbfeaf | 376 | zero_user_segment(page, pglen, PAGE_SIZE); |
efc91ed0 TM |
377 | } |
378 | ||
4899f9c8 | 379 | status = nfs_updatepage(file, page, offset, copied); |
4899f9c8 NP |
380 | |
381 | unlock_page(page); | |
09cbfeaf | 382 | put_page(page); |
4899f9c8 | 383 | |
3d509e54 CL |
384 | if (status < 0) |
385 | return status; | |
2701d086 | 386 | NFS_I(mapping->host)->write_io += copied; |
dc24826b | 387 | |
ce52914e | 388 | if (nfs_ctx_key_to_expire(ctx, mapping->host)) { |
dc24826b AA |
389 | status = nfs_wb_all(mapping->host); |
390 | if (status < 0) | |
391 | return status; | |
392 | } | |
393 | ||
3d509e54 | 394 | return copied; |
1da177e4 LT |
395 | } |
396 | ||
6b9b3514 DH |
397 | /* |
398 | * Partially or wholly invalidate a page | |
399 | * - Release the private state associated with a page if undergoing complete | |
400 | * page invalidation | |
545db45f | 401 | * - Called if either PG_private or PG_fscache is set on the page |
6b9b3514 DH |
402 | * - Caller holds page lock |
403 | */ | |
d47992f8 LC |
404 | static void nfs_invalidate_page(struct page *page, unsigned int offset, |
405 | unsigned int length) | |
cd52ed35 | 406 | { |
d47992f8 LC |
407 | dfprintk(PAGECACHE, "NFS: invalidate_page(%p, %u, %u)\n", |
408 | page, offset, length); | |
b7eaefaa | 409 | |
09cbfeaf | 410 | if (offset != 0 || length < PAGE_SIZE) |
1c75950b | 411 | return; |
d2ccddf0 | 412 | /* Cancel any unstarted writes on this page */ |
d56b4ddf | 413 | nfs_wb_page_cancel(page_file_mapping(page)->host, page); |
a6b5a28e | 414 | wait_on_page_fscache(page); |
cd52ed35 TM |
415 | } |
416 | ||
6b9b3514 DH |
417 | /* |
418 | * Attempt to release the private state associated with a page | |
545db45f | 419 | * - Called if either PG_private or PG_fscache is set on the page |
6b9b3514 DH |
420 | * - Caller holds page lock |
421 | * - Return true (may release page) or false (may not) | |
422 | */ | |
cd52ed35 TM |
423 | static int nfs_release_page(struct page *page, gfp_t gfp) |
424 | { | |
b7eaefaa CL |
425 | dfprintk(PAGECACHE, "NFS: release_page(%p)\n", page); |
426 | ||
e3db7691 | 427 | /* If PagePrivate() is set, then the page is not freeable */ |
545db45f DH |
428 | if (PagePrivate(page)) |
429 | return 0; | |
430 | return nfs_fscache_release_page(page, gfp); | |
e3db7691 TM |
431 | } |
432 | ||
f919b196 MG |
433 | static void nfs_check_dirty_writeback(struct page *page, |
434 | bool *dirty, bool *writeback) | |
435 | { | |
436 | struct nfs_inode *nfsi; | |
437 | struct address_space *mapping = page_file_mapping(page); | |
438 | ||
439 | if (!mapping || PageSwapCache(page)) | |
440 | return; | |
441 | ||
442 | /* | |
443 | * Check if an unstable page is currently being committed and | |
444 | * if so, have the VM treat it as if the page is under writeback | |
445 | * so it will not block due to pages that will shortly be freeable. | |
446 | */ | |
447 | nfsi = NFS_I(mapping->host); | |
af7cf057 | 448 | if (atomic_read(&nfsi->commit_info.rpcs_out)) { |
f919b196 MG |
449 | *writeback = true; |
450 | return; | |
451 | } | |
452 | ||
453 | /* | |
454 | * If PagePrivate() is set, then the page is not freeable and as the | |
455 | * inode is not being committed, it's not going to be cleaned in the | |
456 | * near future so treat it as dirty | |
457 | */ | |
458 | if (PagePrivate(page)) | |
459 | *dirty = true; | |
460 | } | |
461 | ||
6b9b3514 DH |
462 | /* |
463 | * Attempt to clear the private state associated with a page when an error | |
464 | * occurs that requires the cached contents of an inode to be written back or | |
465 | * destroyed | |
545db45f | 466 | * - Called if either PG_private or fscache is set on the page |
6b9b3514 DH |
467 | * - Caller holds page lock |
468 | * - Return 0 if successful, -error otherwise | |
469 | */ | |
e3db7691 TM |
470 | static int nfs_launder_page(struct page *page) |
471 | { | |
d56b4ddf | 472 | struct inode *inode = page_file_mapping(page)->host; |
b7eaefaa CL |
473 | |
474 | dfprintk(PAGECACHE, "NFS: launder_page(%ld, %llu)\n", | |
475 | inode->i_ino, (long long)page_offset(page)); | |
476 | ||
a6b5a28e | 477 | wait_on_page_fscache(page); |
c373fff7 | 478 | return nfs_wb_page(inode, page); |
cd52ed35 TM |
479 | } |
480 | ||
a564b8f0 MG |
481 | static int nfs_swap_activate(struct swap_info_struct *sis, struct file *file, |
482 | sector_t *span) | |
483 | { | |
bd89bc67 MZ |
484 | unsigned long blocks; |
485 | long long isize; | |
4dc73c67 N |
486 | struct inode *inode = file_inode(file); |
487 | struct rpc_clnt *clnt = NFS_CLIENT(inode); | |
488 | struct nfs_client *cl = NFS_SERVER(inode)->nfs_client; | |
bd89bc67 MZ |
489 | |
490 | spin_lock(&inode->i_lock); | |
491 | blocks = inode->i_blocks; | |
492 | isize = inode->i_size; | |
493 | spin_unlock(&inode->i_lock); | |
494 | if (blocks*512 < isize) { | |
495 | pr_warn("swap activate: swapfile has holes\n"); | |
496 | return -EINVAL; | |
497 | } | |
dad2b015 | 498 | |
a564b8f0 | 499 | *span = sis->pages; |
dad2b015 | 500 | |
4dc73c67 N |
501 | |
502 | if (cl->rpc_ops->enable_swap) | |
503 | cl->rpc_ops->enable_swap(inode); | |
504 | ||
3c87ef6e | 505 | return rpc_clnt_swap_activate(clnt); |
a564b8f0 MG |
506 | } |
507 | ||
508 | static void nfs_swap_deactivate(struct file *file) | |
509 | { | |
4dc73c67 N |
510 | struct inode *inode = file_inode(file); |
511 | struct rpc_clnt *clnt = NFS_CLIENT(inode); | |
512 | struct nfs_client *cl = NFS_SERVER(inode)->nfs_client; | |
dad2b015 | 513 | |
3c87ef6e | 514 | rpc_clnt_swap_deactivate(clnt); |
4dc73c67 N |
515 | if (cl->rpc_ops->disable_swap) |
516 | cl->rpc_ops->disable_swap(file_inode(file)); | |
a564b8f0 | 517 | } |
a564b8f0 | 518 | |
f5e54d6e | 519 | const struct address_space_operations nfs_file_aops = { |
1da177e4 | 520 | .readpage = nfs_readpage, |
8786fde8 | 521 | .readahead = nfs_readahead, |
9cccef95 | 522 | .set_page_dirty = __set_page_dirty_nobuffers, |
1da177e4 LT |
523 | .writepage = nfs_writepage, |
524 | .writepages = nfs_writepages, | |
4899f9c8 NP |
525 | .write_begin = nfs_write_begin, |
526 | .write_end = nfs_write_end, | |
cd52ed35 TM |
527 | .invalidatepage = nfs_invalidate_page, |
528 | .releasepage = nfs_release_page, | |
1da177e4 | 529 | .direct_IO = nfs_direct_IO, |
f844cd0d | 530 | #ifdef CONFIG_MIGRATION |
074cc1de | 531 | .migratepage = nfs_migrate_page, |
f844cd0d | 532 | #endif |
e3db7691 | 533 | .launder_page = nfs_launder_page, |
f919b196 | 534 | .is_dirty_writeback = nfs_check_dirty_writeback, |
f590f333 | 535 | .error_remove_page = generic_error_remove_page, |
a564b8f0 MG |
536 | .swap_activate = nfs_swap_activate, |
537 | .swap_deactivate = nfs_swap_deactivate, | |
1da177e4 LT |
538 | }; |
539 | ||
6b9b3514 DH |
540 | /* |
541 | * Notification that a PTE pointing to an NFS page is about to be made | |
542 | * writable, implying that someone is about to modify the page through a | |
543 | * shared-writable mapping | |
544 | */ | |
01a36844 | 545 | static vm_fault_t nfs_vm_page_mkwrite(struct vm_fault *vmf) |
94387fb1 | 546 | { |
c2ec175c | 547 | struct page *page = vmf->page; |
11bac800 | 548 | struct file *filp = vmf->vma->vm_file; |
6de1472f | 549 | struct inode *inode = file_inode(filp); |
94387fb1 | 550 | unsigned pagelen; |
01a36844 | 551 | vm_fault_t ret = VM_FAULT_NOPAGE; |
4899f9c8 | 552 | struct address_space *mapping; |
94387fb1 | 553 | |
1e8968c5 | 554 | dfprintk(PAGECACHE, "NFS: vm_page_mkwrite(%pD2(%lu), offset %lld)\n", |
6de1472f | 555 | filp, filp->f_mapping->host->i_ino, |
b7eaefaa CL |
556 | (long long)page_offset(page)); |
557 | ||
9a773e7c TM |
558 | sb_start_pagefault(inode->i_sb); |
559 | ||
545db45f | 560 | /* make sure the cache has finished storing the page */ |
a6b5a28e DW |
561 | if (PageFsCache(page) && |
562 | wait_on_page_fscache_killable(vmf->page) < 0) { | |
563 | ret = VM_FAULT_RETRY; | |
564 | goto out; | |
565 | } | |
545db45f | 566 | |
ef070dcb TM |
567 | wait_on_bit_action(&NFS_I(inode)->flags, NFS_INO_INVALIDATING, |
568 | nfs_wait_bit_killable, TASK_KILLABLE); | |
569 | ||
94387fb1 | 570 | lock_page(page); |
d56b4ddf | 571 | mapping = page_file_mapping(page); |
6de1472f | 572 | if (mapping != inode->i_mapping) |
8b1f9ee5 TM |
573 | goto out_unlock; |
574 | ||
2aeb98f4 TM |
575 | wait_on_page_writeback(page); |
576 | ||
94387fb1 | 577 | pagelen = nfs_page_length(page); |
8b1f9ee5 TM |
578 | if (pagelen == 0) |
579 | goto out_unlock; | |
4899f9c8 | 580 | |
bc4866b6 TM |
581 | ret = VM_FAULT_LOCKED; |
582 | if (nfs_flush_incompatible(filp, page) == 0 && | |
583 | nfs_updatepage(filp, page, 0, pagelen) == 0) | |
584 | goto out; | |
8b1f9ee5 | 585 | |
bc4866b6 | 586 | ret = VM_FAULT_SIGBUS; |
8b1f9ee5 TM |
587 | out_unlock: |
588 | unlock_page(page); | |
bc4866b6 | 589 | out: |
9a773e7c | 590 | sb_end_pagefault(inode->i_sb); |
bc4866b6 | 591 | return ret; |
94387fb1 TM |
592 | } |
593 | ||
f0f37e2f | 594 | static const struct vm_operations_struct nfs_file_vm_ops = { |
94387fb1 | 595 | .fault = filemap_fault, |
f1820361 | 596 | .map_pages = filemap_map_pages, |
94387fb1 TM |
597 | .page_mkwrite = nfs_vm_page_mkwrite, |
598 | }; | |
599 | ||
ce368536 SM |
600 | static int nfs_need_check_write(struct file *filp, struct inode *inode, |
601 | int error) | |
7b159fc1 TM |
602 | { |
603 | struct nfs_open_context *ctx; | |
604 | ||
cd3758e3 | 605 | ctx = nfs_file_open_context(filp); |
ce368536 SM |
606 | if (nfs_error_is_fatal_on_server(error) || |
607 | nfs_ctx_key_to_expire(ctx, inode)) | |
7b159fc1 TM |
608 | return 1; |
609 | return 0; | |
610 | } | |
611 | ||
edaf4369 | 612 | ssize_t nfs_file_write(struct kiocb *iocb, struct iov_iter *from) |
1da177e4 | 613 | { |
6de1472f AV |
614 | struct file *file = iocb->ki_filp; |
615 | struct inode *inode = file_inode(file); | |
ed7bcdb3 TM |
616 | unsigned int mntflags = NFS_SERVER(inode)->flags; |
617 | ssize_t result, written; | |
ce368536 SM |
618 | errseq_t since; |
619 | int error; | |
1da177e4 | 620 | |
6de1472f | 621 | result = nfs_key_timeout_notify(file, inode); |
dc24826b AA |
622 | if (result) |
623 | return result; | |
624 | ||
89698b24 | 625 | if (iocb->ki_flags & IOCB_DIRECT) |
65a4a1ca | 626 | return nfs_file_direct_write(iocb, from); |
1da177e4 | 627 | |
619d30b4 | 628 | dprintk("NFS: write(%pD2, %zu@%Ld)\n", |
18290650 | 629 | file, iov_iter_count(from), (long long) iocb->ki_pos); |
1da177e4 | 630 | |
1da177e4 LT |
631 | if (IS_SWAPFILE(inode)) |
632 | goto out_swapfile; | |
7d52e862 TM |
633 | /* |
634 | * O_APPEND implies that we must revalidate the file length. | |
635 | */ | |
fc9dc401 | 636 | if (iocb->ki_flags & IOCB_APPEND || iocb->ki_pos > i_size_read(inode)) { |
6de1472f | 637 | result = nfs_revalidate_file_size(inode, file); |
7d52e862 TM |
638 | if (result) |
639 | goto out; | |
fe51beec | 640 | } |
1da177e4 | 641 | |
28aa2f9e TM |
642 | nfs_clear_invalid_mapping(file->f_mapping); |
643 | ||
ce368536 | 644 | since = filemap_sample_wb_err(file->f_mapping); |
a5864c99 | 645 | nfs_start_io_write(inode); |
18290650 TM |
646 | result = generic_write_checks(iocb, from); |
647 | if (result > 0) { | |
648 | current->backing_dev_info = inode_to_bdi(inode); | |
649 | result = generic_perform_write(file, from, iocb->ki_pos); | |
650 | current->backing_dev_info = NULL; | |
651 | } | |
a5864c99 | 652 | nfs_end_io_write(inode); |
18290650 | 653 | if (result <= 0) |
1da177e4 LT |
654 | goto out; |
655 | ||
c49edecd | 656 | written = result; |
18290650 | 657 | iocb->ki_pos += written; |
ed7bcdb3 TM |
658 | |
659 | if (mntflags & NFS_MOUNT_WRITE_EAGER) { | |
660 | result = filemap_fdatawrite_range(file->f_mapping, | |
661 | iocb->ki_pos - written, | |
662 | iocb->ki_pos - 1); | |
663 | if (result < 0) | |
664 | goto out; | |
665 | } | |
666 | if (mntflags & NFS_MOUNT_WRITE_WAIT) { | |
667 | result = filemap_fdatawait_range(file->f_mapping, | |
668 | iocb->ki_pos - written, | |
669 | iocb->ki_pos - 1); | |
670 | if (result < 0) | |
671 | goto out; | |
672 | } | |
e973b1a5 | 673 | result = generic_write_sync(iocb, written); |
674 | if (result < 0) | |
675 | goto out; | |
7e381172 | 676 | |
7e94d6c4 | 677 | /* Return error values */ |
ce368536 SM |
678 | error = filemap_check_wb_err(file->f_mapping, since); |
679 | if (nfs_need_check_write(file, inode, error)) { | |
aded8d7b | 680 | int err = nfs_wb_all(inode); |
200baa21 TM |
681 | if (err < 0) |
682 | result = err; | |
683 | } | |
18290650 | 684 | nfs_add_stats(inode, NFSIOS_NORMALWRITTENBYTES, written); |
1da177e4 LT |
685 | out: |
686 | return result; | |
687 | ||
688 | out_swapfile: | |
689 | printk(KERN_INFO "NFS: attempt to write to active swap file!\n"); | |
89658c4d | 690 | return -ETXTBSY; |
1da177e4 | 691 | } |
89d77c8f | 692 | EXPORT_SYMBOL_GPL(nfs_file_write); |
1da177e4 | 693 | |
5eebde23 SJ |
694 | static int |
695 | do_getlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
696 | { |
697 | struct inode *inode = filp->f_mapping->host; | |
698 | int status = 0; | |
21ac19d4 | 699 | unsigned int saved_type = fl->fl_type; |
1da177e4 | 700 | |
039c4d7a | 701 | /* Try local locking first */ |
6d34ac19 BF |
702 | posix_test_lock(filp, fl); |
703 | if (fl->fl_type != F_UNLCK) { | |
704 | /* found a conflict */ | |
039c4d7a | 705 | goto out; |
1da177e4 | 706 | } |
21ac19d4 | 707 | fl->fl_type = saved_type; |
039c4d7a | 708 | |
011e2a7f | 709 | if (NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) |
039c4d7a TM |
710 | goto out_noconflict; |
711 | ||
5eebde23 | 712 | if (is_local) |
039c4d7a TM |
713 | goto out_noconflict; |
714 | ||
715 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); | |
716 | out: | |
1da177e4 | 717 | return status; |
039c4d7a TM |
718 | out_noconflict: |
719 | fl->fl_type = F_UNLCK; | |
720 | goto out; | |
1da177e4 LT |
721 | } |
722 | ||
5eebde23 SJ |
723 | static int |
724 | do_unlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
725 | { |
726 | struct inode *inode = filp->f_mapping->host; | |
7a8203d8 | 727 | struct nfs_lock_context *l_ctx; |
1da177e4 LT |
728 | int status; |
729 | ||
1da177e4 LT |
730 | /* |
731 | * Flush all pending writes before doing anything | |
732 | * with locks.. | |
733 | */ | |
aded8d7b | 734 | nfs_wb_all(inode); |
1da177e4 | 735 | |
7a8203d8 TM |
736 | l_ctx = nfs_get_lock_context(nfs_file_open_context(filp)); |
737 | if (!IS_ERR(l_ctx)) { | |
210c7c17 | 738 | status = nfs_iocounter_wait(l_ctx); |
7a8203d8 | 739 | nfs_put_lock_context(l_ctx); |
f30cb757 BC |
740 | /* NOTE: special case |
741 | * If we're signalled while cleaning up locks on process exit, we | |
742 | * still need to complete the unlock. | |
743 | */ | |
744 | if (status < 0 && !(fl->fl_flags & FL_CLOSE)) | |
7a8203d8 TM |
745 | return status; |
746 | } | |
747 | ||
5eebde23 SJ |
748 | /* |
749 | * Use local locking if mounted with "-onolock" or with appropriate | |
750 | * "-olocal_lock=" | |
751 | */ | |
752 | if (!is_local) | |
1da177e4 LT |
753 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); |
754 | else | |
75575ddf | 755 | status = locks_lock_file_wait(filp, fl); |
1da177e4 LT |
756 | return status; |
757 | } | |
758 | ||
5eebde23 SJ |
759 | static int |
760 | do_setlk(struct file *filp, int cmd, struct file_lock *fl, int is_local) | |
1da177e4 LT |
761 | { |
762 | struct inode *inode = filp->f_mapping->host; | |
1da177e4 LT |
763 | int status; |
764 | ||
1da177e4 LT |
765 | /* |
766 | * Flush all pending writes before doing anything | |
767 | * with locks.. | |
768 | */ | |
29884df0 TM |
769 | status = nfs_sync_mapping(filp->f_mapping); |
770 | if (status != 0) | |
1da177e4 LT |
771 | goto out; |
772 | ||
5eebde23 SJ |
773 | /* |
774 | * Use local locking if mounted with "-onolock" or with appropriate | |
775 | * "-olocal_lock=" | |
776 | */ | |
777 | if (!is_local) | |
1da177e4 | 778 | status = NFS_PROTO(inode)->lock(filp, cmd, fl); |
c4d7c402 | 779 | else |
75575ddf | 780 | status = locks_lock_file_wait(filp, fl); |
1da177e4 LT |
781 | if (status < 0) |
782 | goto out; | |
6b96724e | 783 | |
1da177e4 | 784 | /* |
779eafab N |
785 | * Invalidate cache to prevent missing any changes. If |
786 | * the file is mapped, clear the page cache as well so | |
787 | * those mappings will be loaded. | |
6b96724e | 788 | * |
1da177e4 LT |
789 | * This makes locking act as a cache coherency point. |
790 | */ | |
29884df0 | 791 | nfs_sync_mapping(filp->f_mapping); |
779eafab | 792 | if (!NFS_PROTO(inode)->have_delegation(inode, FMODE_READ)) { |
442ce049 | 793 | nfs_zap_caches(inode); |
779eafab N |
794 | if (mapping_mapped(filp->f_mapping)) |
795 | nfs_revalidate_mapping(inode, filp->f_mapping); | |
796 | } | |
1da177e4 | 797 | out: |
1da177e4 LT |
798 | return status; |
799 | } | |
800 | ||
801 | /* | |
802 | * Lock a (portion of) a file | |
803 | */ | |
ce4ef7c0 | 804 | int nfs_lock(struct file *filp, int cmd, struct file_lock *fl) |
1da177e4 | 805 | { |
6da24bc9 | 806 | struct inode *inode = filp->f_mapping->host; |
2116271a | 807 | int ret = -ENOLCK; |
5eebde23 | 808 | int is_local = 0; |
1da177e4 | 809 | |
6de1472f AV |
810 | dprintk("NFS: lock(%pD2, t=%x, fl=%x, r=%lld:%lld)\n", |
811 | filp, fl->fl_type, fl->fl_flags, | |
1da177e4 | 812 | (long long)fl->fl_start, (long long)fl->fl_end); |
6da24bc9 | 813 | |
91d5b470 | 814 | nfs_inc_stats(inode, NFSIOS_VFSLOCK); |
1da177e4 | 815 | |
bb0a55bb BF |
816 | if (fl->fl_flags & FL_RECLAIM) |
817 | return -ENOGRACE; | |
818 | ||
5eebde23 SJ |
819 | if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FCNTL) |
820 | is_local = 1; | |
821 | ||
2116271a TM |
822 | if (NFS_PROTO(inode)->lock_check_bounds != NULL) { |
823 | ret = NFS_PROTO(inode)->lock_check_bounds(fl); | |
824 | if (ret < 0) | |
825 | goto out_err; | |
826 | } | |
1da177e4 LT |
827 | |
828 | if (IS_GETLK(cmd)) | |
5eebde23 | 829 | ret = do_getlk(filp, cmd, fl, is_local); |
2116271a | 830 | else if (fl->fl_type == F_UNLCK) |
5eebde23 | 831 | ret = do_unlk(filp, cmd, fl, is_local); |
2116271a | 832 | else |
5eebde23 | 833 | ret = do_setlk(filp, cmd, fl, is_local); |
2116271a TM |
834 | out_err: |
835 | return ret; | |
1da177e4 | 836 | } |
89d77c8f | 837 | EXPORT_SYMBOL_GPL(nfs_lock); |
1da177e4 LT |
838 | |
839 | /* | |
840 | * Lock a (portion of) a file | |
841 | */ | |
ce4ef7c0 | 842 | int nfs_flock(struct file *filp, int cmd, struct file_lock *fl) |
1da177e4 | 843 | { |
5eebde23 SJ |
844 | struct inode *inode = filp->f_mapping->host; |
845 | int is_local = 0; | |
846 | ||
6de1472f AV |
847 | dprintk("NFS: flock(%pD2, t=%x, fl=%x)\n", |
848 | filp, fl->fl_type, fl->fl_flags); | |
1da177e4 | 849 | |
1da177e4 LT |
850 | if (!(fl->fl_flags & FL_FLOCK)) |
851 | return -ENOLCK; | |
852 | ||
5eebde23 SJ |
853 | if (NFS_SERVER(inode)->flags & NFS_MOUNT_LOCAL_FLOCK) |
854 | is_local = 1; | |
855 | ||
fcfa4470 BC |
856 | /* We're simulating flock() locks using posix locks on the server */ |
857 | if (fl->fl_type == F_UNLCK) | |
5eebde23 SJ |
858 | return do_unlk(filp, cmd, fl, is_local); |
859 | return do_setlk(filp, cmd, fl, is_local); | |
1da177e4 | 860 | } |
89d77c8f | 861 | EXPORT_SYMBOL_GPL(nfs_flock); |
370f6599 | 862 | |
0486958f JL |
863 | const struct file_operations nfs_file_operations = { |
864 | .llseek = nfs_file_llseek, | |
3aa2d199 | 865 | .read_iter = nfs_file_read, |
edaf4369 | 866 | .write_iter = nfs_file_write, |
0486958f JL |
867 | .mmap = nfs_file_mmap, |
868 | .open = nfs_file_open, | |
869 | .flush = nfs_file_flush, | |
870 | .release = nfs_file_release, | |
871 | .fsync = nfs_file_fsync, | |
872 | .lock = nfs_lock, | |
873 | .flock = nfs_flock, | |
82c156f8 | 874 | .splice_read = generic_file_splice_read, |
4da54c21 | 875 | .splice_write = iter_file_splice_write, |
0486958f | 876 | .check_flags = nfs_check_flags, |
1c994a09 | 877 | .setlease = simple_nosetlease, |
0486958f | 878 | }; |
ddda8e0a | 879 | EXPORT_SYMBOL_GPL(nfs_file_operations); |