]>
Commit | Line | Data |
---|---|---|
1efb473f BB |
1 | /* |
2 | * CDDL HEADER START | |
3 | * | |
4 | * The contents of this file are subject to the terms of the | |
5 | * Common Development and Distribution License (the "License"). | |
6 | * You may not use this file except in compliance with the License. | |
7 | * | |
8 | * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE | |
9 | * or http://www.opensolaris.org/os/licensing. | |
10 | * See the License for the specific language governing permissions | |
11 | * and limitations under the License. | |
12 | * | |
13 | * When distributing Covered Code, include this CDDL HEADER in each | |
14 | * file and include the License file at usr/src/OPENSOLARIS.LICENSE. | |
15 | * If applicable, add the following below this CDDL HEADER, with the | |
16 | * fields enclosed by brackets "[]" replaced with your own identifying | |
17 | * information: Portions Copyright [yyyy] [name of copyright owner] | |
18 | * | |
19 | * CDDL HEADER END | |
20 | */ | |
21 | /* | |
22 | * Copyright (c) 2011, Lawrence Livermore National Security, LLC. | |
23 | */ | |
24 | ||
25 | ||
26 | #include <sys/zfs_vfsops.h> | |
27 | #include <sys/zfs_vnops.h> | |
28 | #include <sys/zfs_znode.h> | |
29 | #include <sys/zpl.h> | |
30 | ||
31 | ||
126400a1 BB |
32 | static int |
33 | zpl_open(struct inode *ip, struct file *filp) | |
34 | { | |
81e97e21 | 35 | cred_t *cr = CRED(); |
126400a1 BB |
36 | int error; |
37 | ||
81e97e21 | 38 | crhold(cr); |
126400a1 | 39 | error = -zfs_open(ip, filp->f_mode, filp->f_flags, cr); |
81e97e21 | 40 | crfree(cr); |
126400a1 BB |
41 | ASSERT3S(error, <=, 0); |
42 | ||
43 | if (error) | |
44 | return (error); | |
45 | ||
46 | return generic_file_open(ip, filp); | |
47 | } | |
48 | ||
49 | static int | |
50 | zpl_release(struct inode *ip, struct file *filp) | |
51 | { | |
81e97e21 | 52 | cred_t *cr = CRED(); |
126400a1 BB |
53 | int error; |
54 | ||
78d7a5d7 BB |
55 | if (ITOZ(ip)->z_atime_dirty) |
56 | mark_inode_dirty(ip); | |
57 | ||
81e97e21 | 58 | crhold(cr); |
126400a1 | 59 | error = -zfs_close(ip, filp->f_flags, cr); |
81e97e21 | 60 | crfree(cr); |
126400a1 BB |
61 | ASSERT3S(error, <=, 0); |
62 | ||
63 | return (error); | |
64 | } | |
65 | ||
1efb473f | 66 | static int |
0f37d0c8 | 67 | zpl_iterate(struct file *filp, struct dir_context *ctx) |
1efb473f BB |
68 | { |
69 | struct dentry *dentry = filp->f_path.dentry; | |
81e97e21 | 70 | cred_t *cr = CRED(); |
1efb473f BB |
71 | int error; |
72 | ||
81e97e21 | 73 | crhold(cr); |
0f37d0c8 | 74 | error = -zfs_readdir(dentry->d_inode, ctx, cr); |
81e97e21 | 75 | crfree(cr); |
1efb473f BB |
76 | ASSERT3S(error, <=, 0); |
77 | ||
78 | return (error); | |
79 | } | |
80 | ||
0f37d0c8 RY |
81 | #if !defined(HAVE_VFS_ITERATE) |
82 | static int | |
83 | zpl_readdir(struct file *filp, void *dirent, filldir_t filldir) | |
84 | { | |
85 | struct dir_context ctx = DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos); | |
86 | int error; | |
87 | ||
88 | error = zpl_iterate(filp, &ctx); | |
89 | filp->f_pos = ctx.pos; | |
90 | ||
91 | return (error); | |
92 | } | |
93 | #endif /* HAVE_VFS_ITERATE */ | |
94 | ||
adcd70bd | 95 | #if defined(HAVE_FSYNC_WITH_DENTRY) |
3117dd0b | 96 | /* |
adcd70bd BB |
97 | * Linux 2.6.x - 2.6.34 API, |
98 | * Through 2.6.34 the nfsd kernel server would pass a NULL 'file struct *' | |
99 | * to the fops->fsync() hook. For this reason, we must be careful not to | |
100 | * use filp unconditionally. | |
101 | */ | |
102 | static int | |
103 | zpl_fsync(struct file *filp, struct dentry *dentry, int datasync) | |
104 | { | |
105 | cred_t *cr = CRED(); | |
106 | int error; | |
107 | ||
108 | crhold(cr); | |
109 | error = -zfs_fsync(dentry->d_inode, datasync, cr); | |
110 | crfree(cr); | |
111 | ASSERT3S(error, <=, 0); | |
112 | ||
113 | return (error); | |
114 | } | |
115 | ||
116 | #elif defined(HAVE_FSYNC_WITHOUT_DENTRY) | |
117 | /* | |
118 | * Linux 2.6.35 - 3.0 API, | |
119 | * As of 2.6.35 the dentry argument to the fops->fsync() hook was deemed | |
3117dd0b BB |
120 | * redundant. The dentry is still accessible via filp->f_path.dentry, |
121 | * and we are guaranteed that filp will never be NULL. | |
3117dd0b | 122 | */ |
3117dd0b BB |
123 | static int |
124 | zpl_fsync(struct file *filp, int datasync) | |
125 | { | |
adcd70bd BB |
126 | struct inode *inode = filp->f_mapping->host; |
127 | cred_t *cr = CRED(); | |
128 | int error; | |
129 | ||
130 | crhold(cr); | |
131 | error = -zfs_fsync(inode, datasync, cr); | |
132 | crfree(cr); | |
133 | ASSERT3S(error, <=, 0); | |
134 | ||
135 | return (error); | |
136 | } | |
137 | ||
138 | #elif defined(HAVE_FSYNC_RANGE) | |
139 | /* | |
140 | * Linux 3.1 - 3.x API, | |
141 | * As of 3.1 the responsibility to call filemap_write_and_wait_range() has | |
142 | * been pushed down in to the .fsync() vfs hook. Additionally, the i_mutex | |
143 | * lock is no longer held by the caller, for zfs we don't require the lock | |
144 | * to be held so we don't acquire it. | |
145 | */ | |
3117dd0b | 146 | static int |
adcd70bd | 147 | zpl_fsync(struct file *filp, loff_t start, loff_t end, int datasync) |
1efb473f | 148 | { |
adcd70bd | 149 | struct inode *inode = filp->f_mapping->host; |
81e97e21 | 150 | cred_t *cr = CRED(); |
1efb473f BB |
151 | int error; |
152 | ||
adcd70bd BB |
153 | error = filemap_write_and_wait_range(inode->i_mapping, start, end); |
154 | if (error) | |
155 | return (error); | |
156 | ||
81e97e21 | 157 | crhold(cr); |
adcd70bd | 158 | error = -zfs_fsync(inode, datasync, cr); |
81e97e21 | 159 | crfree(cr); |
1efb473f BB |
160 | ASSERT3S(error, <=, 0); |
161 | ||
162 | return (error); | |
163 | } | |
adcd70bd BB |
164 | #else |
165 | #error "Unsupported fops->fsync() implementation" | |
166 | #endif | |
1efb473f BB |
167 | |
168 | ssize_t | |
169 | zpl_read_common(struct inode *ip, const char *buf, size_t len, loff_t pos, | |
170 | uio_seg_t segment, int flags, cred_t *cr) | |
171 | { | |
172 | int error; | |
173 | struct iovec iov; | |
174 | uio_t uio; | |
175 | ||
176 | iov.iov_base = (void *)buf; | |
177 | iov.iov_len = len; | |
178 | ||
179 | uio.uio_iov = &iov; | |
180 | uio.uio_resid = len; | |
181 | uio.uio_iovcnt = 1; | |
182 | uio.uio_loffset = pos; | |
183 | uio.uio_limit = MAXOFFSET_T; | |
184 | uio.uio_segflg = segment; | |
185 | ||
186 | error = -zfs_read(ip, &uio, flags, cr); | |
187 | if (error < 0) | |
188 | return (error); | |
189 | ||
190 | return (len - uio.uio_resid); | |
191 | } | |
192 | ||
193 | static ssize_t | |
194 | zpl_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos) | |
195 | { | |
81e97e21 | 196 | cred_t *cr = CRED(); |
1efb473f BB |
197 | ssize_t read; |
198 | ||
81e97e21 | 199 | crhold(cr); |
1efb473f BB |
200 | read = zpl_read_common(filp->f_mapping->host, buf, len, *ppos, |
201 | UIO_USERSPACE, filp->f_flags, cr); | |
81e97e21 | 202 | crfree(cr); |
1efb473f BB |
203 | |
204 | if (read < 0) | |
205 | return (read); | |
206 | ||
207 | *ppos += read; | |
208 | return (read); | |
209 | } | |
210 | ||
211 | ssize_t | |
212 | zpl_write_common(struct inode *ip, const char *buf, size_t len, loff_t pos, | |
213 | uio_seg_t segment, int flags, cred_t *cr) | |
214 | { | |
215 | int error; | |
216 | struct iovec iov; | |
217 | uio_t uio; | |
218 | ||
219 | iov.iov_base = (void *)buf; | |
220 | iov.iov_len = len; | |
221 | ||
222 | uio.uio_iov = &iov; | |
223 | uio.uio_resid = len, | |
224 | uio.uio_iovcnt = 1; | |
225 | uio.uio_loffset = pos; | |
226 | uio.uio_limit = MAXOFFSET_T; | |
227 | uio.uio_segflg = segment; | |
228 | ||
229 | error = -zfs_write(ip, &uio, flags, cr); | |
230 | if (error < 0) | |
231 | return (error); | |
232 | ||
233 | return (len - uio.uio_resid); | |
234 | } | |
235 | ||
236 | static ssize_t | |
237 | zpl_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos) | |
238 | { | |
81e97e21 | 239 | cred_t *cr = CRED(); |
1efb473f BB |
240 | ssize_t wrote; |
241 | ||
81e97e21 | 242 | crhold(cr); |
1efb473f BB |
243 | wrote = zpl_write_common(filp->f_mapping->host, buf, len, *ppos, |
244 | UIO_USERSPACE, filp->f_flags, cr); | |
81e97e21 | 245 | crfree(cr); |
1efb473f BB |
246 | |
247 | if (wrote < 0) | |
248 | return (wrote); | |
249 | ||
250 | *ppos += wrote; | |
251 | return (wrote); | |
252 | } | |
253 | ||
802e7b5f LD |
254 | static loff_t |
255 | zpl_llseek(struct file *filp, loff_t offset, int whence) | |
256 | { | |
257 | #if defined(SEEK_HOLE) && defined(SEEK_DATA) | |
258 | if (whence == SEEK_DATA || whence == SEEK_HOLE) { | |
259 | struct inode *ip = filp->f_mapping->host; | |
260 | loff_t maxbytes = ip->i_sb->s_maxbytes; | |
261 | loff_t error; | |
262 | ||
263 | spl_inode_lock(ip); | |
264 | error = -zfs_holey(ip, whence, &offset); | |
265 | if (error == 0) | |
266 | error = lseek_execute(filp, ip, offset, maxbytes); | |
267 | spl_inode_unlock(ip); | |
268 | ||
269 | return (error); | |
270 | } | |
271 | #endif /* SEEK_HOLE && SEEK_DATA */ | |
272 | ||
273 | return generic_file_llseek(filp, offset, whence); | |
274 | } | |
275 | ||
c0d35759 BB |
276 | /* |
277 | * It's worth taking a moment to describe how mmap is implemented | |
278 | * for zfs because it differs considerably from other Linux filesystems. | |
279 | * However, this issue is handled the same way under OpenSolaris. | |
280 | * | |
281 | * The issue is that by design zfs bypasses the Linux page cache and | |
282 | * leaves all caching up to the ARC. This has been shown to work | |
283 | * well for the common read(2)/write(2) case. However, mmap(2) | |
284 | * is problem because it relies on being tightly integrated with the | |
285 | * page cache. To handle this we cache mmap'ed files twice, once in | |
286 | * the ARC and a second time in the page cache. The code is careful | |
287 | * to keep both copies synchronized. | |
288 | * | |
289 | * When a file with an mmap'ed region is written to using write(2) | |
290 | * both the data in the ARC and existing pages in the page cache | |
291 | * are updated. For a read(2) data will be read first from the page | |
292 | * cache then the ARC if needed. Neither a write(2) or read(2) will | |
293 | * will ever result in new pages being added to the page cache. | |
294 | * | |
295 | * New pages are added to the page cache only via .readpage() which | |
296 | * is called when the vfs needs to read a page off disk to back the | |
297 | * virtual memory region. These pages may be modified without | |
298 | * notifying the ARC and will be written out periodically via | |
299 | * .writepage(). This will occur due to either a sync or the usual | |
300 | * page aging behavior. Note because a read(2) of a mmap'ed file | |
301 | * will always check the page cache first even when the ARC is out | |
302 | * of date correct data will still be returned. | |
303 | * | |
304 | * While this implementation ensures correct behavior it does have | |
305 | * have some drawbacks. The most obvious of which is that it | |
306 | * increases the required memory footprint when access mmap'ed | |
307 | * files. It also adds additional complexity to the code keeping | |
308 | * both caches synchronized. | |
309 | * | |
310 | * Longer term it may be possible to cleanly resolve this wart by | |
311 | * mapping page cache pages directly on to the ARC buffers. The | |
312 | * Linux address space operations are flexible enough to allow | |
313 | * selection of which pages back a particular index. The trick | |
314 | * would be working out the details of which subsystem is in | |
315 | * charge, the ARC, the page cache, or both. It may also prove | |
316 | * helpful to move the ARC buffers to a scatter-gather lists | |
317 | * rather than a vmalloc'ed region. | |
318 | */ | |
319 | static int | |
320 | zpl_mmap(struct file *filp, struct vm_area_struct *vma) | |
321 | { | |
e2e7aa2d BB |
322 | struct inode *ip = filp->f_mapping->host; |
323 | znode_t *zp = ITOZ(ip); | |
c0d35759 BB |
324 | int error; |
325 | ||
e2e7aa2d BB |
326 | error = -zfs_map(ip, vma->vm_pgoff, (caddr_t *)vma->vm_start, |
327 | (size_t)(vma->vm_end - vma->vm_start), vma->vm_flags); | |
328 | if (error) | |
329 | return (error); | |
330 | ||
c0d35759 BB |
331 | error = generic_file_mmap(filp, vma); |
332 | if (error) | |
333 | return (error); | |
334 | ||
335 | mutex_enter(&zp->z_lock); | |
336 | zp->z_is_mapped = 1; | |
337 | mutex_exit(&zp->z_lock); | |
338 | ||
339 | return (error); | |
340 | } | |
341 | ||
342 | /* | |
343 | * Populate a page with data for the Linux page cache. This function is | |
344 | * only used to support mmap(2). There will be an identical copy of the | |
345 | * data in the ARC which is kept up to date via .write() and .writepage(). | |
346 | * | |
347 | * Current this function relies on zpl_read_common() and the O_DIRECT | |
348 | * flag to read in a page. This works but the more correct way is to | |
349 | * update zfs_fillpage() to be Linux friendly and use that interface. | |
350 | */ | |
351 | static int | |
352 | zpl_readpage(struct file *filp, struct page *pp) | |
353 | { | |
354 | struct inode *ip; | |
dde471ef | 355 | struct page *pl[1]; |
c0d35759 BB |
356 | int error = 0; |
357 | ||
358 | ASSERT(PageLocked(pp)); | |
359 | ip = pp->mapping->host; | |
dde471ef | 360 | pl[0] = pp; |
c0d35759 | 361 | |
dde471ef | 362 | error = -zfs_getpage(ip, pl, 1); |
c0d35759 | 363 | |
dde471ef PJ |
364 | if (error) { |
365 | SetPageError(pp); | |
366 | ClearPageUptodate(pp); | |
367 | } else { | |
368 | ClearPageError(pp); | |
369 | SetPageUptodate(pp); | |
370 | flush_dcache_page(pp); | |
371 | } | |
c0d35759 | 372 | |
dde471ef PJ |
373 | unlock_page(pp); |
374 | return error; | |
375 | } | |
c0d35759 | 376 | |
f3ab88d6 BB |
377 | /* |
378 | * Populate a set of pages with data for the Linux page cache. This | |
379 | * function will only be called for read ahead and never for demand | |
380 | * paging. For simplicity, the code relies on read_cache_pages() to | |
381 | * correctly lock each page for IO and call zpl_readpage(). | |
382 | */ | |
383 | static int | |
384 | zpl_readpages(struct file *filp, struct address_space *mapping, | |
385 | struct list_head *pages, unsigned nr_pages) | |
386 | { | |
95d9fd02 BB |
387 | return (read_cache_pages(mapping, pages, |
388 | (filler_t *)zpl_readpage, filp)); | |
f3ab88d6 BB |
389 | } |
390 | ||
dde471ef PJ |
391 | int |
392 | zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data) | |
393 | { | |
3c0e5c0f BB |
394 | struct address_space *mapping = data; |
395 | ||
396 | ASSERT(PageLocked(pp)); | |
397 | ASSERT(!PageWriteback(pp)); | |
8630650a BB |
398 | ASSERT(!(current->flags & PF_NOFS)); |
399 | ||
400 | /* | |
401 | * Annotate this call path with a flag that indicates that it is | |
402 | * unsafe to use KM_SLEEP during memory allocations due to the | |
403 | * potential for a deadlock. KM_PUSHPAGE should be used instead. | |
404 | */ | |
405 | current->flags |= PF_NOFS; | |
62c4165a | 406 | (void) zfs_putpage(mapping->host, pp, wbc); |
8630650a | 407 | current->flags &= ~PF_NOFS; |
c0d35759 | 408 | |
3c0e5c0f | 409 | return (0); |
dde471ef | 410 | } |
c0d35759 | 411 | |
dde471ef PJ |
412 | static int |
413 | zpl_writepages(struct address_space *mapping, struct writeback_control *wbc) | |
414 | { | |
415 | return write_cache_pages(mapping, wbc, zpl_putpage, mapping); | |
c0d35759 BB |
416 | } |
417 | ||
418 | /* | |
419 | * Write out dirty pages to the ARC, this function is only required to | |
420 | * support mmap(2). Mapped pages may be dirtied by memory operations | |
421 | * which never call .write(). These dirty pages are kept in sync with | |
422 | * the ARC buffers via this hook. | |
c0d35759 BB |
423 | */ |
424 | static int | |
425 | zpl_writepage(struct page *pp, struct writeback_control *wbc) | |
426 | { | |
dde471ef | 427 | return zpl_putpage(pp, wbc, pp->mapping); |
c0d35759 BB |
428 | } |
429 | ||
cb2d1901 ED |
430 | /* |
431 | * The only flag combination which matches the behavior of zfs_space() | |
432 | * is FALLOC_FL_PUNCH_HOLE. This flag was introduced in the 2.6.38 kernel. | |
433 | */ | |
434 | long | |
435 | zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len) | |
436 | { | |
437 | cred_t *cr = CRED(); | |
438 | int error = -EOPNOTSUPP; | |
439 | ||
440 | if (mode & FALLOC_FL_KEEP_SIZE) | |
441 | return (-EOPNOTSUPP); | |
442 | ||
443 | crhold(cr); | |
444 | ||
445 | #ifdef FALLOC_FL_PUNCH_HOLE | |
446 | if (mode & FALLOC_FL_PUNCH_HOLE) { | |
447 | flock64_t bf; | |
448 | ||
449 | bf.l_type = F_WRLCK; | |
450 | bf.l_whence = 0; | |
451 | bf.l_start = offset; | |
452 | bf.l_len = len; | |
453 | bf.l_pid = 0; | |
454 | ||
455 | error = -zfs_space(ip, F_FREESP, &bf, FWRITE, offset, cr); | |
456 | } | |
457 | #endif /* FALLOC_FL_PUNCH_HOLE */ | |
458 | ||
459 | crfree(cr); | |
460 | ||
461 | ASSERT3S(error, <=, 0); | |
462 | return (error); | |
463 | } | |
464 | ||
465 | #ifdef HAVE_FILE_FALLOCATE | |
466 | static long | |
467 | zpl_fallocate(struct file *filp, int mode, loff_t offset, loff_t len) | |
468 | { | |
469 | return zpl_fallocate_common(filp->f_path.dentry->d_inode, | |
470 | mode, offset, len); | |
471 | } | |
472 | #endif /* HAVE_FILE_FALLOCATE */ | |
473 | ||
88c28395 BB |
474 | static long |
475 | zpl_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) | |
476 | { | |
477 | switch (cmd) { | |
478 | case ZFS_IOC_GETFLAGS: | |
479 | case ZFS_IOC_SETFLAGS: | |
480 | return (-EOPNOTSUPP); | |
481 | default: | |
482 | return (-ENOTTY); | |
483 | } | |
484 | } | |
485 | ||
486 | #ifdef CONFIG_COMPAT | |
487 | static long | |
488 | zpl_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) | |
489 | { | |
490 | return zpl_ioctl(filp, cmd, arg); | |
491 | } | |
492 | #endif /* CONFIG_COMPAT */ | |
493 | ||
494 | ||
1efb473f | 495 | const struct address_space_operations zpl_address_space_operations = { |
dde471ef | 496 | .readpages = zpl_readpages, |
1efb473f BB |
497 | .readpage = zpl_readpage, |
498 | .writepage = zpl_writepage, | |
dde471ef | 499 | .writepages = zpl_writepages, |
1efb473f BB |
500 | }; |
501 | ||
502 | const struct file_operations zpl_file_operations = { | |
126400a1 BB |
503 | .open = zpl_open, |
504 | .release = zpl_release, | |
802e7b5f | 505 | .llseek = zpl_llseek, |
c0d35759 BB |
506 | .read = zpl_read, |
507 | .write = zpl_write, | |
c0d35759 | 508 | .mmap = zpl_mmap, |
1efb473f | 509 | .fsync = zpl_fsync, |
cb2d1901 ED |
510 | #ifdef HAVE_FILE_FALLOCATE |
511 | .fallocate = zpl_fallocate, | |
512 | #endif /* HAVE_FILE_FALLOCATE */ | |
88c28395 BB |
513 | .unlocked_ioctl = zpl_ioctl, |
514 | #ifdef CONFIG_COMPAT | |
515 | .compat_ioctl = zpl_compat_ioctl, | |
516 | #endif | |
1efb473f BB |
517 | }; |
518 | ||
519 | const struct file_operations zpl_dir_file_operations = { | |
520 | .llseek = generic_file_llseek, | |
521 | .read = generic_read_dir, | |
0f37d0c8 RY |
522 | #ifdef HAVE_VFS_ITERATE |
523 | .iterate = zpl_iterate, | |
524 | #else | |
1efb473f | 525 | .readdir = zpl_readdir, |
0f37d0c8 | 526 | #endif |
1efb473f | 527 | .fsync = zpl_fsync, |
88c28395 BB |
528 | .unlocked_ioctl = zpl_ioctl, |
529 | #ifdef CONFIG_COMPAT | |
530 | .compat_ioctl = zpl_compat_ioctl, | |
531 | #endif | |
1efb473f | 532 | }; |