]>
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 | ||
119a394a | 26 | #include <sys/dmu_objset.h> |
1efb473f BB |
27 | #include <sys/zfs_vfsops.h> |
28 | #include <sys/zfs_vnops.h> | |
29 | #include <sys/zfs_znode.h> | |
30 | #include <sys/zpl.h> | |
31 | ||
32 | ||
126400a1 BB |
33 | static int |
34 | zpl_open(struct inode *ip, struct file *filp) | |
35 | { | |
81e97e21 | 36 | cred_t *cr = CRED(); |
126400a1 BB |
37 | int error; |
38 | ||
7dc71949 CC |
39 | error = generic_file_open(ip, filp); |
40 | if (error) | |
41 | return (error); | |
42 | ||
81e97e21 | 43 | crhold(cr); |
126400a1 | 44 | error = -zfs_open(ip, filp->f_mode, filp->f_flags, cr); |
81e97e21 | 45 | crfree(cr); |
126400a1 BB |
46 | ASSERT3S(error, <=, 0); |
47 | ||
7dc71949 | 48 | return (error); |
126400a1 BB |
49 | } |
50 | ||
51 | static int | |
52 | zpl_release(struct inode *ip, struct file *filp) | |
53 | { | |
81e97e21 | 54 | cred_t *cr = CRED(); |
126400a1 BB |
55 | int error; |
56 | ||
78d7a5d7 | 57 | if (ITOZ(ip)->z_atime_dirty) |
1e8db771 | 58 | zfs_mark_inode_dirty(ip); |
78d7a5d7 | 59 | |
81e97e21 | 60 | crhold(cr); |
126400a1 | 61 | error = -zfs_close(ip, filp->f_flags, cr); |
81e97e21 | 62 | crfree(cr); |
126400a1 BB |
63 | ASSERT3S(error, <=, 0); |
64 | ||
65 | return (error); | |
66 | } | |
67 | ||
1efb473f | 68 | static int |
0f37d0c8 | 69 | zpl_iterate(struct file *filp, struct dir_context *ctx) |
1efb473f BB |
70 | { |
71 | struct dentry *dentry = filp->f_path.dentry; | |
81e97e21 | 72 | cred_t *cr = CRED(); |
1efb473f BB |
73 | int error; |
74 | ||
81e97e21 | 75 | crhold(cr); |
0f37d0c8 | 76 | error = -zfs_readdir(dentry->d_inode, ctx, cr); |
81e97e21 | 77 | crfree(cr); |
1efb473f BB |
78 | ASSERT3S(error, <=, 0); |
79 | ||
80 | return (error); | |
81 | } | |
82 | ||
0f37d0c8 RY |
83 | #if !defined(HAVE_VFS_ITERATE) |
84 | static int | |
85 | zpl_readdir(struct file *filp, void *dirent, filldir_t filldir) | |
86 | { | |
87 | struct dir_context ctx = DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos); | |
88 | int error; | |
89 | ||
90 | error = zpl_iterate(filp, &ctx); | |
91 | filp->f_pos = ctx.pos; | |
92 | ||
93 | return (error); | |
94 | } | |
95 | #endif /* HAVE_VFS_ITERATE */ | |
96 | ||
adcd70bd | 97 | #if defined(HAVE_FSYNC_WITH_DENTRY) |
3117dd0b | 98 | /* |
adcd70bd BB |
99 | * Linux 2.6.x - 2.6.34 API, |
100 | * Through 2.6.34 the nfsd kernel server would pass a NULL 'file struct *' | |
101 | * to the fops->fsync() hook. For this reason, we must be careful not to | |
102 | * use filp unconditionally. | |
103 | */ | |
104 | static int | |
105 | zpl_fsync(struct file *filp, struct dentry *dentry, int datasync) | |
106 | { | |
107 | cred_t *cr = CRED(); | |
108 | int error; | |
109 | ||
110 | crhold(cr); | |
111 | error = -zfs_fsync(dentry->d_inode, datasync, cr); | |
112 | crfree(cr); | |
113 | ASSERT3S(error, <=, 0); | |
114 | ||
115 | return (error); | |
116 | } | |
117 | ||
cd3939c5 RY |
118 | static int |
119 | zpl_aio_fsync(struct kiocb *kiocb, int datasync) | |
120 | { | |
121 | struct file *filp = kiocb->ki_filp; | |
122 | return (zpl_fsync(filp, filp->f_path.dentry, datasync)); | |
123 | } | |
adcd70bd BB |
124 | #elif defined(HAVE_FSYNC_WITHOUT_DENTRY) |
125 | /* | |
126 | * Linux 2.6.35 - 3.0 API, | |
127 | * As of 2.6.35 the dentry argument to the fops->fsync() hook was deemed | |
3117dd0b BB |
128 | * redundant. The dentry is still accessible via filp->f_path.dentry, |
129 | * and we are guaranteed that filp will never be NULL. | |
3117dd0b | 130 | */ |
3117dd0b BB |
131 | static int |
132 | zpl_fsync(struct file *filp, int datasync) | |
133 | { | |
adcd70bd BB |
134 | struct inode *inode = filp->f_mapping->host; |
135 | cred_t *cr = CRED(); | |
136 | int error; | |
137 | ||
138 | crhold(cr); | |
139 | error = -zfs_fsync(inode, datasync, cr); | |
140 | crfree(cr); | |
141 | ASSERT3S(error, <=, 0); | |
142 | ||
143 | return (error); | |
144 | } | |
145 | ||
cd3939c5 RY |
146 | static int |
147 | zpl_aio_fsync(struct kiocb *kiocb, int datasync) | |
148 | { | |
149 | return (zpl_fsync(kiocb->ki_filp, datasync)); | |
150 | } | |
adcd70bd BB |
151 | #elif defined(HAVE_FSYNC_RANGE) |
152 | /* | |
153 | * Linux 3.1 - 3.x API, | |
154 | * As of 3.1 the responsibility to call filemap_write_and_wait_range() has | |
155 | * been pushed down in to the .fsync() vfs hook. Additionally, the i_mutex | |
156 | * lock is no longer held by the caller, for zfs we don't require the lock | |
157 | * to be held so we don't acquire it. | |
158 | */ | |
3117dd0b | 159 | static int |
adcd70bd | 160 | zpl_fsync(struct file *filp, loff_t start, loff_t end, int datasync) |
1efb473f | 161 | { |
adcd70bd | 162 | struct inode *inode = filp->f_mapping->host; |
81e97e21 | 163 | cred_t *cr = CRED(); |
1efb473f BB |
164 | int error; |
165 | ||
adcd70bd BB |
166 | error = filemap_write_and_wait_range(inode->i_mapping, start, end); |
167 | if (error) | |
168 | return (error); | |
169 | ||
81e97e21 | 170 | crhold(cr); |
adcd70bd | 171 | error = -zfs_fsync(inode, datasync, cr); |
81e97e21 | 172 | crfree(cr); |
1efb473f BB |
173 | ASSERT3S(error, <=, 0); |
174 | ||
175 | return (error); | |
176 | } | |
cd3939c5 RY |
177 | |
178 | static int | |
179 | zpl_aio_fsync(struct kiocb *kiocb, int datasync) | |
180 | { | |
181 | return (zpl_fsync(kiocb->ki_filp, kiocb->ki_pos, | |
182 | kiocb->ki_pos + kiocb->ki_nbytes, datasync)); | |
183 | } | |
adcd70bd BB |
184 | #else |
185 | #error "Unsupported fops->fsync() implementation" | |
186 | #endif | |
1efb473f | 187 | |
cd3939c5 RY |
188 | static inline ssize_t |
189 | zpl_read_common_iovec(struct inode *ip, const struct iovec *iovp, size_t count, | |
190 | unsigned long nr_segs, loff_t *ppos, uio_seg_t segment, | |
191 | int flags, cred_t *cr) | |
1efb473f | 192 | { |
e3dc14b8 | 193 | ssize_t read; |
1efb473f | 194 | uio_t uio; |
cd3939c5 | 195 | int error; |
1efb473f | 196 | |
cd3939c5 RY |
197 | uio.uio_iov = (struct iovec *)iovp; |
198 | uio.uio_resid = count; | |
199 | uio.uio_iovcnt = nr_segs; | |
200 | uio.uio_loffset = *ppos; | |
1efb473f BB |
201 | uio.uio_limit = MAXOFFSET_T; |
202 | uio.uio_segflg = segment; | |
203 | ||
204 | error = -zfs_read(ip, &uio, flags, cr); | |
205 | if (error < 0) | |
206 | return (error); | |
207 | ||
cd3939c5 RY |
208 | read = count - uio.uio_resid; |
209 | *ppos += read; | |
e3dc14b8 BB |
210 | task_io_account_read(read); |
211 | ||
212 | return (read); | |
1efb473f BB |
213 | } |
214 | ||
cd3939c5 RY |
215 | inline ssize_t |
216 | zpl_read_common(struct inode *ip, const char *buf, size_t len, loff_t *ppos, | |
217 | uio_seg_t segment, int flags, cred_t *cr) | |
218 | { | |
219 | struct iovec iov; | |
220 | ||
221 | iov.iov_base = (void *)buf; | |
222 | iov.iov_len = len; | |
223 | ||
224 | return (zpl_read_common_iovec(ip, &iov, len, 1, ppos, segment, | |
225 | flags, cr)); | |
226 | } | |
227 | ||
1efb473f BB |
228 | static ssize_t |
229 | zpl_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos) | |
230 | { | |
81e97e21 | 231 | cred_t *cr = CRED(); |
1efb473f BB |
232 | ssize_t read; |
233 | ||
81e97e21 | 234 | crhold(cr); |
cd3939c5 | 235 | read = zpl_read_common(filp->f_mapping->host, buf, len, ppos, |
1efb473f | 236 | UIO_USERSPACE, filp->f_flags, cr); |
81e97e21 | 237 | crfree(cr); |
1efb473f | 238 | |
cd3939c5 RY |
239 | return (read); |
240 | } | |
241 | ||
242 | static ssize_t | |
243 | zpl_aio_read(struct kiocb *kiocb, const struct iovec *iovp, | |
244 | unsigned long nr_segs, loff_t pos) | |
245 | { | |
246 | cred_t *cr = CRED(); | |
247 | struct file *filp = kiocb->ki_filp; | |
248 | size_t count = kiocb->ki_nbytes; | |
249 | ssize_t read; | |
250 | size_t alloc_size = sizeof (struct iovec) * nr_segs; | |
251 | struct iovec *iov_tmp = kmem_alloc(alloc_size, KM_SLEEP); | |
252 | bcopy(iovp, iov_tmp, alloc_size); | |
253 | ||
254 | ASSERT(iovp); | |
255 | ||
256 | crhold(cr); | |
257 | read = zpl_read_common_iovec(filp->f_mapping->host, iov_tmp, count, | |
258 | nr_segs, &kiocb->ki_pos, UIO_USERSPACE, filp->f_flags, cr); | |
259 | crfree(cr); | |
260 | ||
261 | kmem_free(iov_tmp, alloc_size); | |
1efb473f | 262 | |
1efb473f BB |
263 | return (read); |
264 | } | |
265 | ||
cd3939c5 RY |
266 | static inline ssize_t |
267 | zpl_write_common_iovec(struct inode *ip, const struct iovec *iovp, size_t count, | |
268 | unsigned long nr_segs, loff_t *ppos, uio_seg_t segment, | |
269 | int flags, cred_t *cr) | |
1efb473f | 270 | { |
e3dc14b8 | 271 | ssize_t wrote; |
1efb473f | 272 | uio_t uio; |
cd3939c5 | 273 | int error; |
1efb473f | 274 | |
cd3939c5 RY |
275 | uio.uio_iov = (struct iovec *)iovp; |
276 | uio.uio_resid = count; | |
277 | uio.uio_iovcnt = nr_segs; | |
278 | uio.uio_loffset = *ppos; | |
1efb473f BB |
279 | uio.uio_limit = MAXOFFSET_T; |
280 | uio.uio_segflg = segment; | |
281 | ||
282 | error = -zfs_write(ip, &uio, flags, cr); | |
283 | if (error < 0) | |
284 | return (error); | |
285 | ||
cd3939c5 RY |
286 | wrote = count - uio.uio_resid; |
287 | *ppos += wrote; | |
e3dc14b8 BB |
288 | task_io_account_write(wrote); |
289 | ||
290 | return (wrote); | |
1efb473f | 291 | } |
cd3939c5 RY |
292 | inline ssize_t |
293 | zpl_write_common(struct inode *ip, const char *buf, size_t len, loff_t *ppos, | |
294 | uio_seg_t segment, int flags, cred_t *cr) | |
295 | { | |
296 | struct iovec iov; | |
297 | ||
298 | iov.iov_base = (void *)buf; | |
299 | iov.iov_len = len; | |
300 | ||
301 | return (zpl_write_common_iovec(ip, &iov, len, 1, ppos, segment, | |
302 | flags, cr)); | |
303 | } | |
1efb473f BB |
304 | |
305 | static ssize_t | |
306 | zpl_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos) | |
307 | { | |
81e97e21 | 308 | cred_t *cr = CRED(); |
1efb473f BB |
309 | ssize_t wrote; |
310 | ||
81e97e21 | 311 | crhold(cr); |
cd3939c5 | 312 | wrote = zpl_write_common(filp->f_mapping->host, buf, len, ppos, |
1efb473f | 313 | UIO_USERSPACE, filp->f_flags, cr); |
81e97e21 | 314 | crfree(cr); |
1efb473f | 315 | |
cd3939c5 RY |
316 | return (wrote); |
317 | } | |
318 | ||
319 | static ssize_t | |
320 | zpl_aio_write(struct kiocb *kiocb, const struct iovec *iovp, | |
321 | unsigned long nr_segs, loff_t pos) | |
322 | { | |
323 | cred_t *cr = CRED(); | |
324 | struct file *filp = kiocb->ki_filp; | |
325 | size_t count = kiocb->ki_nbytes; | |
326 | ssize_t wrote; | |
327 | size_t alloc_size = sizeof (struct iovec) * nr_segs; | |
328 | struct iovec *iov_tmp = kmem_alloc(alloc_size, KM_SLEEP); | |
329 | bcopy(iovp, iov_tmp, alloc_size); | |
330 | ||
331 | ASSERT(iovp); | |
332 | ||
333 | crhold(cr); | |
334 | wrote = zpl_write_common_iovec(filp->f_mapping->host, iov_tmp, count, | |
335 | nr_segs, &kiocb->ki_pos, UIO_USERSPACE, filp->f_flags, cr); | |
336 | crfree(cr); | |
337 | ||
338 | kmem_free(iov_tmp, alloc_size); | |
1efb473f | 339 | |
1efb473f BB |
340 | return (wrote); |
341 | } | |
342 | ||
802e7b5f LD |
343 | static loff_t |
344 | zpl_llseek(struct file *filp, loff_t offset, int whence) | |
345 | { | |
346 | #if defined(SEEK_HOLE) && defined(SEEK_DATA) | |
347 | if (whence == SEEK_DATA || whence == SEEK_HOLE) { | |
348 | struct inode *ip = filp->f_mapping->host; | |
349 | loff_t maxbytes = ip->i_sb->s_maxbytes; | |
350 | loff_t error; | |
351 | ||
352 | spl_inode_lock(ip); | |
353 | error = -zfs_holey(ip, whence, &offset); | |
354 | if (error == 0) | |
355 | error = lseek_execute(filp, ip, offset, maxbytes); | |
356 | spl_inode_unlock(ip); | |
357 | ||
358 | return (error); | |
359 | } | |
360 | #endif /* SEEK_HOLE && SEEK_DATA */ | |
361 | ||
d1d7e268 | 362 | return (generic_file_llseek(filp, offset, whence)); |
802e7b5f LD |
363 | } |
364 | ||
c0d35759 BB |
365 | /* |
366 | * It's worth taking a moment to describe how mmap is implemented | |
367 | * for zfs because it differs considerably from other Linux filesystems. | |
368 | * However, this issue is handled the same way under OpenSolaris. | |
369 | * | |
370 | * The issue is that by design zfs bypasses the Linux page cache and | |
371 | * leaves all caching up to the ARC. This has been shown to work | |
372 | * well for the common read(2)/write(2) case. However, mmap(2) | |
373 | * is problem because it relies on being tightly integrated with the | |
374 | * page cache. To handle this we cache mmap'ed files twice, once in | |
375 | * the ARC and a second time in the page cache. The code is careful | |
376 | * to keep both copies synchronized. | |
377 | * | |
378 | * When a file with an mmap'ed region is written to using write(2) | |
379 | * both the data in the ARC and existing pages in the page cache | |
380 | * are updated. For a read(2) data will be read first from the page | |
381 | * cache then the ARC if needed. Neither a write(2) or read(2) will | |
382 | * will ever result in new pages being added to the page cache. | |
383 | * | |
384 | * New pages are added to the page cache only via .readpage() which | |
385 | * is called when the vfs needs to read a page off disk to back the | |
386 | * virtual memory region. These pages may be modified without | |
387 | * notifying the ARC and will be written out periodically via | |
388 | * .writepage(). This will occur due to either a sync or the usual | |
389 | * page aging behavior. Note because a read(2) of a mmap'ed file | |
390 | * will always check the page cache first even when the ARC is out | |
391 | * of date correct data will still be returned. | |
392 | * | |
393 | * While this implementation ensures correct behavior it does have | |
394 | * have some drawbacks. The most obvious of which is that it | |
395 | * increases the required memory footprint when access mmap'ed | |
396 | * files. It also adds additional complexity to the code keeping | |
397 | * both caches synchronized. | |
398 | * | |
399 | * Longer term it may be possible to cleanly resolve this wart by | |
400 | * mapping page cache pages directly on to the ARC buffers. The | |
401 | * Linux address space operations are flexible enough to allow | |
402 | * selection of which pages back a particular index. The trick | |
403 | * would be working out the details of which subsystem is in | |
404 | * charge, the ARC, the page cache, or both. It may also prove | |
405 | * helpful to move the ARC buffers to a scatter-gather lists | |
406 | * rather than a vmalloc'ed region. | |
407 | */ | |
408 | static int | |
409 | zpl_mmap(struct file *filp, struct vm_area_struct *vma) | |
410 | { | |
e2e7aa2d BB |
411 | struct inode *ip = filp->f_mapping->host; |
412 | znode_t *zp = ITOZ(ip); | |
c0d35759 BB |
413 | int error; |
414 | ||
e2e7aa2d BB |
415 | error = -zfs_map(ip, vma->vm_pgoff, (caddr_t *)vma->vm_start, |
416 | (size_t)(vma->vm_end - vma->vm_start), vma->vm_flags); | |
417 | if (error) | |
418 | return (error); | |
419 | ||
c0d35759 BB |
420 | error = generic_file_mmap(filp, vma); |
421 | if (error) | |
422 | return (error); | |
423 | ||
424 | mutex_enter(&zp->z_lock); | |
425 | zp->z_is_mapped = 1; | |
426 | mutex_exit(&zp->z_lock); | |
427 | ||
428 | return (error); | |
429 | } | |
430 | ||
431 | /* | |
432 | * Populate a page with data for the Linux page cache. This function is | |
433 | * only used to support mmap(2). There will be an identical copy of the | |
434 | * data in the ARC which is kept up to date via .write() and .writepage(). | |
435 | * | |
436 | * Current this function relies on zpl_read_common() and the O_DIRECT | |
437 | * flag to read in a page. This works but the more correct way is to | |
438 | * update zfs_fillpage() to be Linux friendly and use that interface. | |
439 | */ | |
440 | static int | |
441 | zpl_readpage(struct file *filp, struct page *pp) | |
442 | { | |
443 | struct inode *ip; | |
dde471ef | 444 | struct page *pl[1]; |
c0d35759 BB |
445 | int error = 0; |
446 | ||
447 | ASSERT(PageLocked(pp)); | |
448 | ip = pp->mapping->host; | |
dde471ef | 449 | pl[0] = pp; |
c0d35759 | 450 | |
dde471ef | 451 | error = -zfs_getpage(ip, pl, 1); |
c0d35759 | 452 | |
dde471ef PJ |
453 | if (error) { |
454 | SetPageError(pp); | |
455 | ClearPageUptodate(pp); | |
456 | } else { | |
457 | ClearPageError(pp); | |
458 | SetPageUptodate(pp); | |
459 | flush_dcache_page(pp); | |
460 | } | |
c0d35759 | 461 | |
dde471ef | 462 | unlock_page(pp); |
d1d7e268 | 463 | return (error); |
dde471ef | 464 | } |
c0d35759 | 465 | |
f3ab88d6 BB |
466 | /* |
467 | * Populate a set of pages with data for the Linux page cache. This | |
468 | * function will only be called for read ahead and never for demand | |
469 | * paging. For simplicity, the code relies on read_cache_pages() to | |
470 | * correctly lock each page for IO and call zpl_readpage(). | |
471 | */ | |
472 | static int | |
473 | zpl_readpages(struct file *filp, struct address_space *mapping, | |
474 | struct list_head *pages, unsigned nr_pages) | |
475 | { | |
95d9fd02 BB |
476 | return (read_cache_pages(mapping, pages, |
477 | (filler_t *)zpl_readpage, filp)); | |
f3ab88d6 BB |
478 | } |
479 | ||
dde471ef PJ |
480 | int |
481 | zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data) | |
482 | { | |
3c0e5c0f BB |
483 | struct address_space *mapping = data; |
484 | ||
485 | ASSERT(PageLocked(pp)); | |
486 | ASSERT(!PageWriteback(pp)); | |
8630650a BB |
487 | ASSERT(!(current->flags & PF_NOFS)); |
488 | ||
489 | /* | |
490 | * Annotate this call path with a flag that indicates that it is | |
491 | * unsafe to use KM_SLEEP during memory allocations due to the | |
492 | * potential for a deadlock. KM_PUSHPAGE should be used instead. | |
493 | */ | |
494 | current->flags |= PF_NOFS; | |
62c4165a | 495 | (void) zfs_putpage(mapping->host, pp, wbc); |
8630650a | 496 | current->flags &= ~PF_NOFS; |
c0d35759 | 497 | |
3c0e5c0f | 498 | return (0); |
dde471ef | 499 | } |
c0d35759 | 500 | |
dde471ef PJ |
501 | static int |
502 | zpl_writepages(struct address_space *mapping, struct writeback_control *wbc) | |
503 | { | |
119a394a ED |
504 | znode_t *zp = ITOZ(mapping->host); |
505 | zfs_sb_t *zsb = ITOZSB(mapping->host); | |
506 | enum writeback_sync_modes sync_mode; | |
507 | int result; | |
508 | ||
509 | ZFS_ENTER(zsb); | |
510 | if (zsb->z_os->os_sync == ZFS_SYNC_ALWAYS) | |
511 | wbc->sync_mode = WB_SYNC_ALL; | |
512 | ZFS_EXIT(zsb); | |
513 | sync_mode = wbc->sync_mode; | |
514 | ||
515 | /* | |
516 | * We don't want to run write_cache_pages() in SYNC mode here, because | |
517 | * that would make putpage() wait for a single page to be committed to | |
518 | * disk every single time, resulting in atrocious performance. Instead | |
519 | * we run it once in non-SYNC mode so that the ZIL gets all the data, | |
520 | * and then we commit it all in one go. | |
521 | */ | |
522 | wbc->sync_mode = WB_SYNC_NONE; | |
523 | result = write_cache_pages(mapping, wbc, zpl_putpage, mapping); | |
524 | if (sync_mode != wbc->sync_mode) { | |
525 | ZFS_ENTER(zsb); | |
526 | ZFS_VERIFY_ZP(zp); | |
1e8db771 BB |
527 | if (zsb->z_log != NULL) |
528 | zil_commit(zsb->z_log, zp->z_id); | |
119a394a ED |
529 | ZFS_EXIT(zsb); |
530 | ||
531 | /* | |
532 | * We need to call write_cache_pages() again (we can't just | |
533 | * return after the commit) because the previous call in | |
534 | * non-SYNC mode does not guarantee that we got all the dirty | |
535 | * pages (see the implementation of write_cache_pages() for | |
536 | * details). That being said, this is a no-op in most cases. | |
537 | */ | |
538 | wbc->sync_mode = sync_mode; | |
539 | result = write_cache_pages(mapping, wbc, zpl_putpage, mapping); | |
540 | } | |
541 | return (result); | |
c0d35759 BB |
542 | } |
543 | ||
544 | /* | |
545 | * Write out dirty pages to the ARC, this function is only required to | |
546 | * support mmap(2). Mapped pages may be dirtied by memory operations | |
547 | * which never call .write(). These dirty pages are kept in sync with | |
548 | * the ARC buffers via this hook. | |
c0d35759 BB |
549 | */ |
550 | static int | |
551 | zpl_writepage(struct page *pp, struct writeback_control *wbc) | |
552 | { | |
119a394a ED |
553 | if (ITOZSB(pp->mapping->host)->z_os->os_sync == ZFS_SYNC_ALWAYS) |
554 | wbc->sync_mode = WB_SYNC_ALL; | |
555 | ||
556 | return (zpl_putpage(pp, wbc, pp->mapping)); | |
c0d35759 BB |
557 | } |
558 | ||
cb2d1901 ED |
559 | /* |
560 | * The only flag combination which matches the behavior of zfs_space() | |
561 | * is FALLOC_FL_PUNCH_HOLE. This flag was introduced in the 2.6.38 kernel. | |
562 | */ | |
563 | long | |
564 | zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len) | |
565 | { | |
566 | cred_t *cr = CRED(); | |
567 | int error = -EOPNOTSUPP; | |
568 | ||
569 | if (mode & FALLOC_FL_KEEP_SIZE) | |
570 | return (-EOPNOTSUPP); | |
571 | ||
572 | crhold(cr); | |
573 | ||
574 | #ifdef FALLOC_FL_PUNCH_HOLE | |
575 | if (mode & FALLOC_FL_PUNCH_HOLE) { | |
576 | flock64_t bf; | |
577 | ||
578 | bf.l_type = F_WRLCK; | |
579 | bf.l_whence = 0; | |
580 | bf.l_start = offset; | |
581 | bf.l_len = len; | |
582 | bf.l_pid = 0; | |
583 | ||
584 | error = -zfs_space(ip, F_FREESP, &bf, FWRITE, offset, cr); | |
585 | } | |
586 | #endif /* FALLOC_FL_PUNCH_HOLE */ | |
587 | ||
588 | crfree(cr); | |
589 | ||
590 | ASSERT3S(error, <=, 0); | |
591 | return (error); | |
592 | } | |
593 | ||
594 | #ifdef HAVE_FILE_FALLOCATE | |
595 | static long | |
596 | zpl_fallocate(struct file *filp, int mode, loff_t offset, loff_t len) | |
597 | { | |
598 | return zpl_fallocate_common(filp->f_path.dentry->d_inode, | |
599 | mode, offset, len); | |
600 | } | |
601 | #endif /* HAVE_FILE_FALLOCATE */ | |
602 | ||
9d317793 RY |
603 | /* |
604 | * Map zfs file z_pflags (xvattr_t) to linux file attributes. Only file | |
605 | * attributes common to both Linux and Solaris are mapped. | |
606 | */ | |
607 | static int | |
608 | zpl_ioctl_getflags(struct file *filp, void __user *arg) | |
609 | { | |
610 | struct inode *ip = filp->f_dentry->d_inode; | |
611 | unsigned int ioctl_flags = 0; | |
612 | uint64_t zfs_flags = ITOZ(ip)->z_pflags; | |
613 | int error; | |
614 | ||
615 | if (zfs_flags & ZFS_IMMUTABLE) | |
616 | ioctl_flags |= FS_IMMUTABLE_FL; | |
617 | ||
618 | if (zfs_flags & ZFS_APPENDONLY) | |
619 | ioctl_flags |= FS_APPEND_FL; | |
620 | ||
621 | if (zfs_flags & ZFS_NODUMP) | |
622 | ioctl_flags |= FS_NODUMP_FL; | |
623 | ||
624 | ioctl_flags &= FS_FL_USER_VISIBLE; | |
625 | ||
626 | error = copy_to_user(arg, &ioctl_flags, sizeof (ioctl_flags)); | |
627 | ||
628 | return (error); | |
629 | } | |
630 | ||
631 | /* | |
632 | * fchange() is a helper macro to detect if we have been asked to change a | |
633 | * flag. This is ugly, but the requirement that we do this is a consequence of | |
634 | * how the Linux file attribute interface was designed. Another consequence is | |
635 | * that concurrent modification of files suffers from a TOCTOU race. Neither | |
636 | * are things we can fix without modifying the kernel-userland interface, which | |
637 | * is outside of our jurisdiction. | |
638 | */ | |
639 | ||
640 | #define fchange(f0, f1, b0, b1) ((((f0) & (b0)) == (b0)) != \ | |
641 | (((b1) & (f1)) == (f1))) | |
642 | ||
643 | static int | |
644 | zpl_ioctl_setflags(struct file *filp, void __user *arg) | |
645 | { | |
646 | struct inode *ip = filp->f_dentry->d_inode; | |
647 | uint64_t zfs_flags = ITOZ(ip)->z_pflags; | |
648 | unsigned int ioctl_flags; | |
649 | cred_t *cr = CRED(); | |
650 | xvattr_t xva; | |
651 | xoptattr_t *xoap; | |
652 | int error; | |
653 | ||
654 | if (copy_from_user(&ioctl_flags, arg, sizeof (ioctl_flags))) | |
655 | return (-EFAULT); | |
656 | ||
657 | if ((ioctl_flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | FS_NODUMP_FL))) | |
658 | return (-EOPNOTSUPP); | |
659 | ||
660 | if ((ioctl_flags & ~(FS_FL_USER_MODIFIABLE))) | |
661 | return (-EACCES); | |
662 | ||
663 | if ((fchange(ioctl_flags, zfs_flags, FS_IMMUTABLE_FL, ZFS_IMMUTABLE) || | |
664 | fchange(ioctl_flags, zfs_flags, FS_APPEND_FL, ZFS_APPENDONLY)) && | |
665 | !capable(CAP_LINUX_IMMUTABLE)) | |
666 | return (-EACCES); | |
667 | ||
668 | if (!zpl_inode_owner_or_capable(ip)) | |
669 | return (-EACCES); | |
670 | ||
671 | xva_init(&xva); | |
672 | xoap = xva_getxoptattr(&xva); | |
673 | ||
674 | XVA_SET_REQ(&xva, XAT_IMMUTABLE); | |
675 | if (ioctl_flags & FS_IMMUTABLE_FL) | |
676 | xoap->xoa_immutable = B_TRUE; | |
677 | ||
678 | XVA_SET_REQ(&xva, XAT_APPENDONLY); | |
679 | if (ioctl_flags & FS_APPEND_FL) | |
680 | xoap->xoa_appendonly = B_TRUE; | |
681 | ||
682 | XVA_SET_REQ(&xva, XAT_NODUMP); | |
683 | if (ioctl_flags & FS_NODUMP_FL) | |
684 | xoap->xoa_nodump = B_TRUE; | |
685 | ||
686 | crhold(cr); | |
687 | error = -zfs_setattr(ip, (vattr_t *)&xva, 0, cr); | |
688 | crfree(cr); | |
689 | ||
690 | return (error); | |
691 | } | |
692 | ||
88c28395 BB |
693 | static long |
694 | zpl_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) | |
695 | { | |
696 | switch (cmd) { | |
9d317793 RY |
697 | case FS_IOC_GETFLAGS: |
698 | return (zpl_ioctl_getflags(filp, (void *)arg)); | |
699 | case FS_IOC_SETFLAGS: | |
700 | return (zpl_ioctl_setflags(filp, (void *)arg)); | |
88c28395 BB |
701 | default: |
702 | return (-ENOTTY); | |
703 | } | |
704 | } | |
705 | ||
706 | #ifdef CONFIG_COMPAT | |
707 | static long | |
708 | zpl_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) | |
709 | { | |
d1d7e268 | 710 | return (zpl_ioctl(filp, cmd, arg)); |
88c28395 BB |
711 | } |
712 | #endif /* CONFIG_COMPAT */ | |
713 | ||
714 | ||
1efb473f | 715 | const struct address_space_operations zpl_address_space_operations = { |
dde471ef | 716 | .readpages = zpl_readpages, |
1efb473f BB |
717 | .readpage = zpl_readpage, |
718 | .writepage = zpl_writepage, | |
d1d7e268 | 719 | .writepages = zpl_writepages, |
1efb473f BB |
720 | }; |
721 | ||
722 | const struct file_operations zpl_file_operations = { | |
126400a1 BB |
723 | .open = zpl_open, |
724 | .release = zpl_release, | |
802e7b5f | 725 | .llseek = zpl_llseek, |
c0d35759 BB |
726 | .read = zpl_read, |
727 | .write = zpl_write, | |
cd3939c5 RY |
728 | .aio_read = zpl_aio_read, |
729 | .aio_write = zpl_aio_write, | |
c0d35759 | 730 | .mmap = zpl_mmap, |
1efb473f | 731 | .fsync = zpl_fsync, |
cd3939c5 | 732 | .aio_fsync = zpl_aio_fsync, |
cb2d1901 | 733 | #ifdef HAVE_FILE_FALLOCATE |
d1d7e268 | 734 | .fallocate = zpl_fallocate, |
cb2d1901 | 735 | #endif /* HAVE_FILE_FALLOCATE */ |
d1d7e268 | 736 | .unlocked_ioctl = zpl_ioctl, |
88c28395 | 737 | #ifdef CONFIG_COMPAT |
d1d7e268 | 738 | .compat_ioctl = zpl_compat_ioctl, |
88c28395 | 739 | #endif |
1efb473f BB |
740 | }; |
741 | ||
742 | const struct file_operations zpl_dir_file_operations = { | |
743 | .llseek = generic_file_llseek, | |
744 | .read = generic_read_dir, | |
0f37d0c8 RY |
745 | #ifdef HAVE_VFS_ITERATE |
746 | .iterate = zpl_iterate, | |
747 | #else | |
1efb473f | 748 | .readdir = zpl_readdir, |
0f37d0c8 | 749 | #endif |
1efb473f | 750 | .fsync = zpl_fsync, |
88c28395 BB |
751 | .unlocked_ioctl = zpl_ioctl, |
752 | #ifdef CONFIG_COMPAT | |
753 | .compat_ioctl = zpl_compat_ioctl, | |
754 | #endif | |
1efb473f | 755 | }; |