1 // SPDX-License-Identifier: CDDL-1.0
2 /*
3 * CDDL HEADER START
4 *
5 * The contents of this file are subject to the terms of the
6 * Common Development and Distribution License (the "License").
7 * You may not use this file except in compliance with the License.
8 *
9 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
10 * or https://opensource.org/licenses/CDDL-1.0.
11 * See the License for the specific language governing permissions
12 * and limitations under the License.
13 *
14 * When distributing Covered Code, include this CDDL HEADER in each
15 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
16 * If applicable, add the following below this CDDL HEADER, with the
17 * fields enclosed by brackets "[]" replaced with your own identifying
18 * information: Portions Copyright [yyyy] [name of copyright owner]
19 *
20 * CDDL HEADER END
21 */
22 /*
23 * Copyright (c) 2011, Lawrence Livermore National Security, LLC.
24 * Copyright (c) 2015 by Chunwei Chen. All rights reserved.
25 * Copyright (c) 2025, Klara, Inc.
26 * Copyright (c) 2025, Rob Norris <robn@despairlabs.com>
27 */
28
29
30 #ifdef CONFIG_COMPAT
31 #include <linux/compat.h>
32 #endif
33 #include <linux/fs.h>
34 #include <linux/migrate.h>
35 #include <sys/file.h>
36 #include <sys/dmu_objset.h>
37 #include <sys/zfs_znode.h>
38 #include <sys/zfs_vfsops.h>
39 #include <sys/zfs_vnops.h>
40 #include <sys/zfs_project.h>
41 #include <linux/pagemap_compat.h>
42 #include <linux/fadvise.h>
43 #ifdef HAVE_VFS_FILEMAP_DIRTY_FOLIO
44 #include <linux/writeback.h>
45 #endif
46 #ifdef HAVE_FILELOCK_HEADER
47 #include <linux/filelock.h>
48 #endif
49
50 /*
51 * When using fallocate(2) to preallocate space, inflate the requested
52 * capacity check by 10% to account for the required metadata blocks.
53 */
54 static unsigned int zfs_fallocate_reserve_percent = 110;
55
56 static int
zpl_open(struct inode * ip,struct file * filp)57 zpl_open(struct inode *ip, struct file *filp)
58 {
59 cred_t *cr = CRED();
60 int error;
61 fstrans_cookie_t cookie;
62
63 error = generic_file_open(ip, filp);
64 if (error)
65 return (error);
66
67 crhold(cr);
68 cookie = spl_fstrans_mark();
69 error = -zfs_open(ip, filp->f_mode, filp->f_flags, cr);
70 spl_fstrans_unmark(cookie);
71 crfree(cr);
72 ASSERT3S(error, <=, 0);
73
74 return (error);
75 }
76
77 static int
zpl_release(struct inode * ip,struct file * filp)78 zpl_release(struct inode *ip, struct file *filp)
79 {
80 cred_t *cr = CRED();
81 int error;
82 fstrans_cookie_t cookie;
83
84 cookie = spl_fstrans_mark();
85 if (ITOZ(ip)->z_atime_dirty)
86 zfs_mark_inode_dirty(ip);
87
88 crhold(cr);
89 error = -zfs_close(ip, filp->f_flags, cr);
90 spl_fstrans_unmark(cookie);
91 crfree(cr);
92 ASSERT3S(error, <=, 0);
93
94 return (error);
95 }
96
97 static int
zpl_iterate(struct file * filp,struct dir_context * ctx)98 zpl_iterate(struct file *filp, struct dir_context *ctx)
99 {
100 cred_t *cr = CRED();
101 int error;
102 fstrans_cookie_t cookie;
103
104 crhold(cr);
105 cookie = spl_fstrans_mark();
106 error = -zfs_readdir(file_inode(filp), ctx, cr);
107 spl_fstrans_unmark(cookie);
108 crfree(cr);
109 ASSERT3S(error, <=, 0);
110
111 return (error);
112 }
113
114 static inline int
115 zpl_write_cache_pages(struct address_space *mapping,
116 struct writeback_control *wbc, void *data);
117
118 static int
zpl_fsync(struct file * filp,loff_t start,loff_t end,int datasync)119 zpl_fsync(struct file *filp, loff_t start, loff_t end, int datasync)
120 {
121 struct inode *inode = filp->f_mapping->host;
122 znode_t *zp = ITOZ(inode);
123 cred_t *cr = CRED();
124 int error;
125 fstrans_cookie_t cookie;
126
127 /*
128 * Force dirty pages in the range out to the DMU and the log, ready
129 * for zil_commit() to write down.
130 *
131 * We call write_cache_pages() directly to ensure that zpl_putpage() is
132 * called with the flags we need. We need WB_SYNC_NONE to avoid a call
133 * to zil_commit() (since we're doing this as a kind of pre-sync); but
134 * we do need for_sync so that the pages remain in writeback until
135 * they're on disk, and so that we get an error if the DMU write fails.
136 */
137 if (filemap_range_has_page(inode->i_mapping, start, end)) {
138 int for_sync = 1;
139 struct writeback_control wbc = {
140 .sync_mode = WB_SYNC_NONE,
141 .nr_to_write = LONG_MAX,
142 .range_start = start,
143 .range_end = end,
144 };
145 error =
146 zpl_write_cache_pages(inode->i_mapping, &wbc, &for_sync);
147 if (error != 0) {
148 /*
149 * Unclear what state things are in. zfs_putpage() will
150 * ensure the pages remain dirty if they haven't been
151 * written down to the DMU, but because there may be
152 * nothing logged, we can't assume that zfs_sync() ->
153 * zil_commit() will give us a useful error. It's
154 * safest if we just error out here.
155 */
156 return (error);
157 }
158 }
159
160 crhold(cr);
161 cookie = spl_fstrans_mark();
162 error = -zfs_fsync(zp, datasync, cr);
163 spl_fstrans_unmark(cookie);
164 crfree(cr);
165 ASSERT3S(error, <=, 0);
166
167 return (error);
168 }
169
170 static inline int
zfs_io_flags(struct kiocb * kiocb)171 zfs_io_flags(struct kiocb *kiocb)
172 {
173 int flags = 0;
174
175 #if defined(IOCB_DSYNC)
176 if (kiocb->ki_flags & IOCB_DSYNC)
177 flags |= O_DSYNC;
178 #endif
179 #if defined(IOCB_SYNC)
180 if (kiocb->ki_flags & IOCB_SYNC)
181 flags |= O_SYNC;
182 #endif
183 #if defined(IOCB_APPEND)
184 if (kiocb->ki_flags & IOCB_APPEND)
185 flags |= O_APPEND;
186 #endif
187 #if defined(IOCB_DIRECT)
188 if (kiocb->ki_flags & IOCB_DIRECT)
189 flags |= O_DIRECT;
190 #endif
191 return (flags);
192 }
193
194 /*
195 * If relatime is enabled, call file_accessed() if zfs_relatime_need_update()
196 * is true. This is needed since datasets with inherited "relatime" property
197 * aren't necessarily mounted with the MNT_RELATIME flag (e.g. after
198 * `zfs set relatime=...`), which is what relatime test in VFS by
199 * relatime_need_update() is based on.
200 */
201 static inline void
zpl_file_accessed(struct file * filp)202 zpl_file_accessed(struct file *filp)
203 {
204 struct inode *ip = filp->f_mapping->host;
205
206 if (!IS_NOATIME(ip) && ITOZSB(ip)->z_relatime) {
207 if (zfs_relatime_need_update(ip))
208 file_accessed(filp);
209 } else {
210 file_accessed(filp);
211 }
212 }
213
214 static ssize_t
zpl_iter_read(struct kiocb * kiocb,struct iov_iter * to)215 zpl_iter_read(struct kiocb *kiocb, struct iov_iter *to)
216 {
217 cred_t *cr = CRED();
218 fstrans_cookie_t cookie;
219 struct file *filp = kiocb->ki_filp;
220 ssize_t count = iov_iter_count(to);
221 zfs_uio_t uio;
222
223 zfs_uio_iov_iter_init(&uio, to, kiocb->ki_pos, count);
224
225 crhold(cr);
226 cookie = spl_fstrans_mark();
227
228 ssize_t ret = -zfs_read(ITOZ(filp->f_mapping->host), &uio,
229 filp->f_flags | zfs_io_flags(kiocb), cr);
230
231 spl_fstrans_unmark(cookie);
232 crfree(cr);
233
234 if (ret < 0)
235 return (ret);
236
237 ssize_t read = count - uio.uio_resid;
238 kiocb->ki_pos += read;
239
240 zpl_file_accessed(filp);
241
242 return (read);
243 }
244
245 static inline ssize_t
zpl_generic_write_checks(struct kiocb * kiocb,struct iov_iter * from,size_t * countp)246 zpl_generic_write_checks(struct kiocb *kiocb, struct iov_iter *from,
247 size_t *countp)
248 {
249 ssize_t ret = generic_write_checks(kiocb, from);
250 if (ret <= 0)
251 return (ret);
252
253 *countp = ret;
254
255 return (0);
256 }
257
258 static ssize_t
zpl_iter_write(struct kiocb * kiocb,struct iov_iter * from)259 zpl_iter_write(struct kiocb *kiocb, struct iov_iter *from)
260 {
261 cred_t *cr = CRED();
262 fstrans_cookie_t cookie;
263 struct file *filp = kiocb->ki_filp;
264 struct inode *ip = filp->f_mapping->host;
265 zfs_uio_t uio;
266 size_t count = 0;
267 ssize_t ret;
268
269 ret = zpl_generic_write_checks(kiocb, from, &count);
270 if (ret)
271 return (ret);
272
273 zfs_uio_iov_iter_init(&uio, from, kiocb->ki_pos, count);
274
275 crhold(cr);
276 cookie = spl_fstrans_mark();
277
278 ret = -zfs_write(ITOZ(ip), &uio,
279 filp->f_flags | zfs_io_flags(kiocb), cr);
280
281 spl_fstrans_unmark(cookie);
282 crfree(cr);
283
284 if (ret < 0)
285 return (ret);
286
287 ssize_t wrote = count - uio.uio_resid;
288 kiocb->ki_pos += wrote;
289
290 return (wrote);
291 }
292
293 static ssize_t
zpl_direct_IO(struct kiocb * kiocb,struct iov_iter * iter)294 zpl_direct_IO(struct kiocb *kiocb, struct iov_iter *iter)
295 {
296 /*
297 * All O_DIRECT requests should be handled by
298 * zpl_iter_write/read}(). There is no way kernel generic code should
299 * call the direct_IO address_space_operations function. We set this
300 * code path to be fatal if it is executed.
301 */
302 PANIC(0);
303 return (0);
304 }
305
306 static loff_t
zpl_llseek(struct file * filp,loff_t offset,int whence)307 zpl_llseek(struct file *filp, loff_t offset, int whence)
308 {
309 #if defined(SEEK_HOLE) && defined(SEEK_DATA)
310 fstrans_cookie_t cookie;
311
312 if (whence == SEEK_DATA || whence == SEEK_HOLE) {
313 struct inode *ip = filp->f_mapping->host;
314 loff_t maxbytes = ip->i_sb->s_maxbytes;
315 loff_t error;
316
317 spl_inode_lock_shared(ip);
318 cookie = spl_fstrans_mark();
319 error = -zfs_holey(ITOZ(ip), whence, &offset);
320 spl_fstrans_unmark(cookie);
321 if (error == 0)
322 error = lseek_execute(filp, ip, offset, maxbytes);
323 spl_inode_unlock_shared(ip);
324
325 return (error);
326 }
327 #endif /* SEEK_HOLE && SEEK_DATA */
328
329 return (generic_file_llseek(filp, offset, whence));
330 }
331
332 /*
333 * It's worth taking a moment to describe how mmap is implemented
334 * for zfs because it differs considerably from other Linux filesystems.
335 * However, this issue is handled the same way under OpenSolaris.
336 *
337 * The issue is that by design zfs bypasses the Linux page cache and
338 * leaves all caching up to the ARC. This has been shown to work
339 * well for the common read(2)/write(2) case. However, mmap(2)
340 * is problem because it relies on being tightly integrated with the
341 * page cache. To handle this we cache mmap'ed files twice, once in
342 * the ARC and a second time in the page cache. The code is careful
343 * to keep both copies synchronized.
344 *
345 * When a file with an mmap'ed region is written to using write(2)
346 * both the data in the ARC and existing pages in the page cache
347 * are updated. For a read(2) data will be read first from the page
348 * cache then the ARC if needed. Neither a write(2) or read(2) will
349 * will ever result in new pages being added to the page cache.
350 *
351 * New pages are added to the page cache only via .readpage() which
352 * is called when the vfs needs to read a page off disk to back the
353 * virtual memory region. These pages may be modified without
354 * notifying the ARC and will be written out periodically via
355 * .writepage(). This will occur due to either a sync or the usual
356 * page aging behavior. Note because a read(2) of a mmap'ed file
357 * will always check the page cache first even when the ARC is out
358 * of date correct data will still be returned.
359 *
360 * While this implementation ensures correct behavior it does have
361 * have some drawbacks. The most obvious of which is that it
362 * increases the required memory footprint when access mmap'ed
363 * files. It also adds additional complexity to the code keeping
364 * both caches synchronized.
365 *
366 * Longer term it may be possible to cleanly resolve this wart by
367 * mapping page cache pages directly on to the ARC buffers. The
368 * Linux address space operations are flexible enough to allow
369 * selection of which pages back a particular index. The trick
370 * would be working out the details of which subsystem is in
371 * charge, the ARC, the page cache, or both. It may also prove
372 * helpful to move the ARC buffers to a scatter-gather lists
373 * rather than a vmalloc'ed region.
374 */
375 static int
zpl_mmap(struct file * filp,struct vm_area_struct * vma)376 zpl_mmap(struct file *filp, struct vm_area_struct *vma)
377 {
378 struct inode *ip = filp->f_mapping->host;
379 int error;
380 fstrans_cookie_t cookie;
381
382 cookie = spl_fstrans_mark();
383 error = -zfs_map(ip, vma->vm_pgoff, (caddr_t *)vma->vm_start,
384 (size_t)(vma->vm_end - vma->vm_start), vma->vm_flags);
385 spl_fstrans_unmark(cookie);
386
387 if (error)
388 return (error);
389
390 error = generic_file_mmap(filp, vma);
391 if (error)
392 return (error);
393
394 return (error);
395 }
396
397 /*
398 * Populate a page with data for the Linux page cache. This function is
399 * only used to support mmap(2). There will be an identical copy of the
400 * data in the ARC which is kept up to date via .write() and .writepage().
401 */
402 static inline int
zpl_readpage_common(struct page * pp)403 zpl_readpage_common(struct page *pp)
404 {
405 fstrans_cookie_t cookie;
406
407 ASSERT(PageLocked(pp));
408
409 cookie = spl_fstrans_mark();
410 int error = -zfs_getpage(pp->mapping->host, pp);
411 spl_fstrans_unmark(cookie);
412
413 unlock_page(pp);
414
415 return (error);
416 }
417
418 #ifdef HAVE_VFS_READ_FOLIO
419 static int
zpl_read_folio(struct file * filp,struct folio * folio)420 zpl_read_folio(struct file *filp, struct folio *folio)
421 {
422 return (zpl_readpage_common(&folio->page));
423 }
424 #else
425 static int
zpl_readpage(struct file * filp,struct page * pp)426 zpl_readpage(struct file *filp, struct page *pp)
427 {
428 return (zpl_readpage_common(pp));
429 }
430 #endif
431
432 static int
zpl_readpage_filler(void * data,struct page * pp)433 zpl_readpage_filler(void *data, struct page *pp)
434 {
435 return (zpl_readpage_common(pp));
436 }
437
438 /*
439 * Populate a set of pages with data for the Linux page cache. This
440 * function will only be called for read ahead and never for demand
441 * paging. For simplicity, the code relies on read_cache_pages() to
442 * correctly lock each page for IO and call zpl_readpage().
443 */
444 #ifdef HAVE_VFS_READPAGES
445 static int
zpl_readpages(struct file * filp,struct address_space * mapping,struct list_head * pages,unsigned nr_pages)446 zpl_readpages(struct file *filp, struct address_space *mapping,
447 struct list_head *pages, unsigned nr_pages)
448 {
449 return (read_cache_pages(mapping, pages, zpl_readpage_filler, NULL));
450 }
451 #else
452 static void
zpl_readahead(struct readahead_control * ractl)453 zpl_readahead(struct readahead_control *ractl)
454 {
455 struct page *page;
456
457 while ((page = readahead_page(ractl)) != NULL) {
458 int ret;
459
460 ret = zpl_readpage_filler(NULL, page);
461 put_page(page);
462 if (ret)
463 break;
464 }
465 }
466 #endif
467
468 static int
zpl_putpage(struct page * pp,struct writeback_control * wbc,void * data)469 zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data)
470 {
471 boolean_t *for_sync = data;
472 fstrans_cookie_t cookie;
473 int ret;
474
475 ASSERT(PageLocked(pp));
476 ASSERT(!PageWriteback(pp));
477
478 cookie = spl_fstrans_mark();
479 ret = zfs_putpage(pp->mapping->host, pp, wbc, *for_sync);
480 spl_fstrans_unmark(cookie);
481
482 return (ret);
483 }
484
485 #ifdef HAVE_WRITE_CACHE_PAGES
486 #ifdef HAVE_WRITEPAGE_T_FOLIO
487 static int
zpl_putfolio(struct folio * pp,struct writeback_control * wbc,void * data)488 zpl_putfolio(struct folio *pp, struct writeback_control *wbc, void *data)
489 {
490 return (zpl_putpage(&pp->page, wbc, data));
491 }
492 #endif
493
494 static inline int
zpl_write_cache_pages(struct address_space * mapping,struct writeback_control * wbc,void * data)495 zpl_write_cache_pages(struct address_space *mapping,
496 struct writeback_control *wbc, void *data)
497 {
498 int result;
499
500 #ifdef HAVE_WRITEPAGE_T_FOLIO
501 result = write_cache_pages(mapping, wbc, zpl_putfolio, data);
502 #else
503 result = write_cache_pages(mapping, wbc, zpl_putpage, data);
504 #endif
505 return (result);
506 }
507 #else
508 static inline int
zpl_write_cache_pages(struct address_space * mapping,struct writeback_control * wbc,void * data)509 zpl_write_cache_pages(struct address_space *mapping,
510 struct writeback_control *wbc, void *data)
511 {
512 pgoff_t start = wbc->range_start >> PAGE_SHIFT;
513 pgoff_t end = wbc->range_end >> PAGE_SHIFT;
514
515 struct folio_batch fbatch;
516 folio_batch_init(&fbatch);
517
518 /*
519 * This atomically (-ish) tags all DIRTY pages in the range with
520 * TOWRITE, allowing users to continue dirtying or undirtying pages
521 * while we get on with writeback, without us treading on each other.
522 */
523 tag_pages_for_writeback(mapping, start, end);
524
525 int err = 0;
526 unsigned int npages;
527
528 /*
529 * Grab references to the TOWRITE pages just flagged. This may not get
530 * all of them, so we do it in a loop until there are none left.
531 */
532 while ((npages = filemap_get_folios_tag(mapping, &start, end,
533 PAGECACHE_TAG_TOWRITE, &fbatch)) != 0) {
534
535 /* Loop over each page and write it out. */
536 struct folio *folio;
537 while ((folio = folio_batch_next(&fbatch)) != NULL) {
538 folio_lock(folio);
539
540 /*
541 * If the folio has been remapped, or is no longer
542 * dirty, then there's nothing to do.
543 */
544 if (folio->mapping != mapping ||
545 !folio_test_dirty(folio)) {
546 folio_unlock(folio);
547 continue;
548 }
549
550 /*
551 * If writeback is already in progress, wait for it to
552 * finish. We continue after this even if the page
553 * ends up clean; zfs_putpage() will skip it if no
554 * further work is required.
555 */
556 while (folio_test_writeback(folio))
557 folio_wait_bit(folio, PG_writeback);
558
559 /*
560 * Write it out and collect any error. zfs_putpage()
561 * will clear the TOWRITE and DIRTY flags, and return
562 * with the page unlocked.
563 */
564 int ferr = zpl_putpage(&folio->page, wbc, data);
565 if (err == 0 && ferr != 0)
566 err = ferr;
567
568 /* Housekeeping for the caller. */
569 wbc->nr_to_write -= folio_nr_pages(folio);
570 }
571
572 /* Release any remaining references on the batch. */
573 folio_batch_release(&fbatch);
574 }
575
576 return (err);
577 }
578 #endif
579
580 static int
zpl_writepages(struct address_space * mapping,struct writeback_control * wbc)581 zpl_writepages(struct address_space *mapping, struct writeback_control *wbc)
582 {
583 znode_t *zp = ITOZ(mapping->host);
584 zfsvfs_t *zfsvfs = ITOZSB(mapping->host);
585 enum writeback_sync_modes sync_mode;
586 int result;
587
588 if ((result = zpl_enter(zfsvfs, FTAG)) != 0)
589 return (result);
590 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
591 wbc->sync_mode = WB_SYNC_ALL;
592 zpl_exit(zfsvfs, FTAG);
593 sync_mode = wbc->sync_mode;
594
595 /*
596 * We don't want to run write_cache_pages() in SYNC mode here, because
597 * that would make putpage() wait for a single page to be committed to
598 * disk every single time, resulting in atrocious performance. Instead
599 * we run it once in non-SYNC mode so that the ZIL gets all the data,
600 * and then we commit it all in one go.
601 */
602 boolean_t for_sync = (sync_mode == WB_SYNC_ALL);
603 wbc->sync_mode = WB_SYNC_NONE;
604 result = zpl_write_cache_pages(mapping, wbc, &for_sync);
605 if (sync_mode != wbc->sync_mode) {
606 if ((result = zpl_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
607 return (result);
608
609 if (zfsvfs->z_log != NULL) {
610 /*
611 * We don't want to block here if the pool suspends,
612 * because this is not a syncing op by itself, but
613 * might be part of one that the caller will
614 * coordinate.
615 */
616 result = -zil_commit_flags(zfsvfs->z_log, zp->z_id,
617 ZIL_COMMIT_NOW);
618 }
619
620 zpl_exit(zfsvfs, FTAG);
621
622 /*
623 * If zil_commit_flags() failed, it's unclear what state things
624 * are currently in. putpage() has written back out what it can
625 * to the DMU, but it may not be on disk. We have little choice
626 * but to escape.
627 */
628 if (result != 0)
629 return (result);
630
631 /*
632 * We need to call write_cache_pages() again (we can't just
633 * return after the commit) because the previous call in
634 * non-SYNC mode does not guarantee that we got all the dirty
635 * pages (see the implementation of write_cache_pages() for
636 * details). That being said, this is a no-op in most cases.
637 */
638 wbc->sync_mode = sync_mode;
639 result = zpl_write_cache_pages(mapping, wbc, &for_sync);
640 }
641 return (result);
642 }
643
644 #ifdef HAVE_VFS_WRITEPAGE
645 /*
646 * Write out dirty pages to the ARC, this function is only required to
647 * support mmap(2). Mapped pages may be dirtied by memory operations
648 * which never call .write(). These dirty pages are kept in sync with
649 * the ARC buffers via this hook.
650 */
651 static int
zpl_writepage(struct page * pp,struct writeback_control * wbc)652 zpl_writepage(struct page *pp, struct writeback_control *wbc)
653 {
654 if (ITOZSB(pp->mapping->host)->z_os->os_sync == ZFS_SYNC_ALWAYS)
655 wbc->sync_mode = WB_SYNC_ALL;
656
657 boolean_t for_sync = (wbc->sync_mode == WB_SYNC_ALL);
658
659 return (zpl_putpage(pp, wbc, &for_sync));
660 }
661 #endif
662
663 /*
664 * The flag combination which matches the behavior of zfs_space() is
665 * FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE. The FALLOC_FL_PUNCH_HOLE
666 * flag was introduced in the 2.6.38 kernel.
667 *
668 * The original mode=0 (allocate space) behavior can be reasonably emulated
669 * by checking if enough space exists and creating a sparse file, as real
670 * persistent space reservation is not possible due to COW, snapshots, etc.
671 */
672 static long
zpl_fallocate_common(struct inode * ip,int mode,loff_t offset,loff_t len)673 zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len)
674 {
675 cred_t *cr = CRED();
676 znode_t *zp = ITOZ(ip);
677 zfsvfs_t *zfsvfs = ITOZSB(ip);
678 loff_t olen;
679 fstrans_cookie_t cookie;
680 int error = 0;
681
682 int test_mode = FALLOC_FL_PUNCH_HOLE | FALLOC_FL_ZERO_RANGE;
683
684 if ((mode & ~(FALLOC_FL_KEEP_SIZE | test_mode)) != 0)
685 return (-EOPNOTSUPP);
686
687 if (offset < 0 || len <= 0)
688 return (-EINVAL);
689
690 spl_inode_lock(ip);
691 olen = i_size_read(ip);
692
693 crhold(cr);
694 cookie = spl_fstrans_mark();
695 if (mode & (test_mode)) {
696 flock64_t bf;
697
698 if (mode & FALLOC_FL_KEEP_SIZE) {
699 if (offset > olen)
700 goto out_unmark;
701
702 if (offset + len > olen)
703 len = olen - offset;
704 }
705 bf.l_type = F_WRLCK;
706 bf.l_whence = SEEK_SET;
707 bf.l_start = offset;
708 bf.l_len = len;
709 bf.l_pid = 0;
710
711 error = -zfs_space(zp, F_FREESP, &bf, O_RDWR, offset, cr);
712 } else if ((mode & ~FALLOC_FL_KEEP_SIZE) == 0) {
713 unsigned int percent = zfs_fallocate_reserve_percent;
714 struct kstatfs statfs;
715
716 /* Legacy mode, disable fallocate compatibility. */
717 if (percent == 0) {
718 error = -EOPNOTSUPP;
719 goto out_unmark;
720 }
721
722 /*
723 * Use zfs_statvfs() instead of dmu_objset_space() since it
724 * also checks project quota limits, which are relevant here.
725 */
726 error = -zfs_statvfs(ip, &statfs);
727 if (error)
728 goto out_unmark;
729
730 /*
731 * Shrink available space a bit to account for overhead/races.
732 * We know the product previously fit into availbytes from
733 * dmu_objset_space(), so the smaller product will also fit.
734 */
735 if (len > statfs.f_bavail * (statfs.f_bsize * 100 / percent)) {
736 error = -ENOSPC;
737 goto out_unmark;
738 }
739 if (!(mode & FALLOC_FL_KEEP_SIZE) && offset + len > olen) {
740 error = zpl_enter_verify_zp(zfsvfs, zp, FTAG);
741 if (error)
742 goto out_unmark;
743
744 error = -zfs_freesp(zp, offset + len, 0, 0, FALSE);
745 zfs_exit(zfsvfs, FTAG);
746 }
747 }
748 out_unmark:
749 spl_fstrans_unmark(cookie);
750 spl_inode_unlock(ip);
751
752 crfree(cr);
753
754 return (error);
755 }
756
757 static long
zpl_fallocate(struct file * filp,int mode,loff_t offset,loff_t len)758 zpl_fallocate(struct file *filp, int mode, loff_t offset, loff_t len)
759 {
760 return zpl_fallocate_common(file_inode(filp),
761 mode, offset, len);
762 }
763
764 static int
zpl_ioctl_getversion(struct file * filp,void __user * arg)765 zpl_ioctl_getversion(struct file *filp, void __user *arg)
766 {
767 uint32_t generation = file_inode(filp)->i_generation;
768
769 return (copy_to_user(arg, &generation, sizeof (generation)));
770 }
771
772 static int
zpl_fadvise(struct file * filp,loff_t offset,loff_t len,int advice)773 zpl_fadvise(struct file *filp, loff_t offset, loff_t len, int advice)
774 {
775 struct inode *ip = file_inode(filp);
776 znode_t *zp = ITOZ(ip);
777 zfsvfs_t *zfsvfs = ITOZSB(ip);
778 objset_t *os = zfsvfs->z_os;
779 int error = 0;
780
781 if (S_ISFIFO(ip->i_mode))
782 return (-ESPIPE);
783
784 if (offset < 0 || len < 0)
785 return (-EINVAL);
786
787 if ((error = zpl_enter_verify_zp(zfsvfs, zp, FTAG)) != 0)
788 return (error);
789
790 if (advice == POSIX_FADV_WILLNEED) {
791 loff_t rlen = len ? len : i_size_read(ip) - offset;
792 dmu_prefetch(os, zp->z_id, 0, offset, rlen,
793 ZIO_PRIORITY_ASYNC_READ);
794 if (!zn_has_cached_data(zp, offset, offset + rlen - 1)) {
795 zfs_exit(zfsvfs, FTAG);
796 return (error);
797 }
798 }
799
800 #ifdef HAVE_GENERIC_FADVISE
801 error = generic_fadvise(filp, offset, len, advice);
802 #endif
803
804 if (error == 0 && advice == POSIX_FADV_DONTNEED) {
805 loff_t rlen = len ? len : i_size_read(ip) - offset;
806 dmu_evict_range(os, zp->z_id, offset, rlen);
807 }
808
809 zfs_exit(zfsvfs, FTAG);
810
811 return (error);
812 }
813
814 #define ZFS_FL_USER_VISIBLE (FS_FL_USER_VISIBLE | FS_PROJINHERIT_FL)
815 #define ZFS_FL_USER_MODIFIABLE (FS_FL_USER_MODIFIABLE | FS_PROJINHERIT_FL)
816
817
818 static struct {
819 uint64_t zfs_flag;
820 uint32_t fs_flag;
821 uint32_t xflag;
822 } flags_lookup[] = {
823 {ZFS_IMMUTABLE, FS_IMMUTABLE_FL, FS_XFLAG_IMMUTABLE},
824 {ZFS_APPENDONLY, FS_APPEND_FL, FS_XFLAG_APPEND},
825 {ZFS_NODUMP, FS_NODUMP_FL, FS_XFLAG_NODUMP},
826 {ZFS_PROJINHERIT, FS_PROJINHERIT_FL, FS_XFLAG_PROJINHERIT}
827 };
828
829 static uint32_t
__zpl_ioctl_getflags(struct inode * ip)830 __zpl_ioctl_getflags(struct inode *ip)
831 {
832 uint64_t zfs_flags = ITOZ(ip)->z_pflags;
833 uint32_t ioctl_flags = 0;
834 for (int i = 0; i < ARRAY_SIZE(flags_lookup); i++)
835 if (zfs_flags & flags_lookup[i].zfs_flag)
836 ioctl_flags |= flags_lookup[i].fs_flag;
837
838 return (ioctl_flags);
839 }
840
841 static uint32_t
__zpl_ioctl_getxflags(struct inode * ip)842 __zpl_ioctl_getxflags(struct inode *ip)
843 {
844 uint64_t zfs_flags = ITOZ(ip)->z_pflags;
845 uint32_t ioctl_flags = 0;
846
847 for (int i = 0; i < ARRAY_SIZE(flags_lookup); i++)
848 if (zfs_flags & flags_lookup[i].zfs_flag)
849 ioctl_flags |= flags_lookup[i].xflag;
850
851 return (ioctl_flags);
852 }
853
854 /*
855 * Map zfs file z_pflags (xvattr_t) to linux file attributes. Only file
856 * attributes common to both Linux and Solaris are mapped.
857 */
858 static int
zpl_ioctl_getflags(struct file * filp,void __user * arg)859 zpl_ioctl_getflags(struct file *filp, void __user *arg)
860 {
861 uint32_t flags;
862 int err;
863
864 flags = __zpl_ioctl_getflags(file_inode(filp));
865 flags = flags & ZFS_FL_USER_VISIBLE;
866 err = copy_to_user(arg, &flags, sizeof (flags));
867
868 return (err);
869 }
870
871 /*
872 * fchange() is a helper macro to detect if we have been asked to change a
873 * flag. This is ugly, but the requirement that we do this is a consequence of
874 * how the Linux file attribute interface was designed. Another consequence is
875 * that concurrent modification of files suffers from a TOCTOU race. Neither
876 * are things we can fix without modifying the kernel-userland interface, which
877 * is outside of our jurisdiction.
878 */
879
880 #define fchange(f0, f1, b0, b1) (!((f0) & (b0)) != !((f1) & (b1)))
881
882 static int
__zpl_ioctl_setflags(struct inode * ip,uint32_t ioctl_flags,xvattr_t * xva)883 __zpl_ioctl_setflags(struct inode *ip, uint32_t ioctl_flags, xvattr_t *xva)
884 {
885 uint64_t zfs_flags = ITOZ(ip)->z_pflags;
886 xoptattr_t *xoap;
887
888 if (ioctl_flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | FS_NODUMP_FL |
889 FS_PROJINHERIT_FL))
890 return (-EOPNOTSUPP);
891
892 if (ioctl_flags & ~ZFS_FL_USER_MODIFIABLE)
893 return (-EACCES);
894
895 if ((fchange(ioctl_flags, zfs_flags, FS_IMMUTABLE_FL, ZFS_IMMUTABLE) ||
896 fchange(ioctl_flags, zfs_flags, FS_APPEND_FL, ZFS_APPENDONLY)) &&
897 !capable(CAP_LINUX_IMMUTABLE))
898 return (-EPERM);
899
900 if (!zpl_inode_owner_or_capable(zfs_init_idmap, ip))
901 return (-EACCES);
902
903 xva_init(xva);
904 xoap = xva_getxoptattr(xva);
905
906 #define FLAG_CHANGE(iflag, zflag, xflag, xfield) do { \
907 if (((ioctl_flags & (iflag)) && !(zfs_flags & (zflag))) || \
908 ((zfs_flags & (zflag)) && !(ioctl_flags & (iflag)))) { \
909 XVA_SET_REQ(xva, (xflag)); \
910 (xfield) = ((ioctl_flags & (iflag)) != 0); \
911 } \
912 } while (0)
913
914 FLAG_CHANGE(FS_IMMUTABLE_FL, ZFS_IMMUTABLE, XAT_IMMUTABLE,
915 xoap->xoa_immutable);
916 FLAG_CHANGE(FS_APPEND_FL, ZFS_APPENDONLY, XAT_APPENDONLY,
917 xoap->xoa_appendonly);
918 FLAG_CHANGE(FS_NODUMP_FL, ZFS_NODUMP, XAT_NODUMP,
919 xoap->xoa_nodump);
920 FLAG_CHANGE(FS_PROJINHERIT_FL, ZFS_PROJINHERIT, XAT_PROJINHERIT,
921 xoap->xoa_projinherit);
922
923 #undef FLAG_CHANGE
924
925 return (0);
926 }
927
928 static int
__zpl_ioctl_setxflags(struct inode * ip,uint32_t ioctl_flags,xvattr_t * xva)929 __zpl_ioctl_setxflags(struct inode *ip, uint32_t ioctl_flags, xvattr_t *xva)
930 {
931 uint64_t zfs_flags = ITOZ(ip)->z_pflags;
932 xoptattr_t *xoap;
933
934 if (ioctl_flags & ~(FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND |
935 FS_XFLAG_NODUMP | FS_XFLAG_PROJINHERIT))
936 return (-EOPNOTSUPP);
937
938 if ((fchange(ioctl_flags, zfs_flags, FS_XFLAG_IMMUTABLE,
939 ZFS_IMMUTABLE) ||
940 fchange(ioctl_flags, zfs_flags, FS_XFLAG_APPEND, ZFS_APPENDONLY)) &&
941 !capable(CAP_LINUX_IMMUTABLE))
942 return (-EPERM);
943
944 if (!zpl_inode_owner_or_capable(zfs_init_idmap, ip))
945 return (-EACCES);
946
947 xva_init(xva);
948 xoap = xva_getxoptattr(xva);
949
950 #define FLAG_CHANGE(iflag, zflag, xflag, xfield) do { \
951 if (((ioctl_flags & (iflag)) && !(zfs_flags & (zflag))) || \
952 ((zfs_flags & (zflag)) && !(ioctl_flags & (iflag)))) { \
953 XVA_SET_REQ(xva, (xflag)); \
954 (xfield) = ((ioctl_flags & (iflag)) != 0); \
955 } \
956 } while (0)
957
958 FLAG_CHANGE(FS_XFLAG_IMMUTABLE, ZFS_IMMUTABLE, XAT_IMMUTABLE,
959 xoap->xoa_immutable);
960 FLAG_CHANGE(FS_XFLAG_APPEND, ZFS_APPENDONLY, XAT_APPENDONLY,
961 xoap->xoa_appendonly);
962 FLAG_CHANGE(FS_XFLAG_NODUMP, ZFS_NODUMP, XAT_NODUMP,
963 xoap->xoa_nodump);
964 FLAG_CHANGE(FS_XFLAG_PROJINHERIT, ZFS_PROJINHERIT, XAT_PROJINHERIT,
965 xoap->xoa_projinherit);
966
967 #undef FLAG_CHANGE
968
969 return (0);
970 }
971
972 static int
zpl_ioctl_setflags(struct file * filp,void __user * arg)973 zpl_ioctl_setflags(struct file *filp, void __user *arg)
974 {
975 struct inode *ip = file_inode(filp);
976 uint32_t flags;
977 cred_t *cr = CRED();
978 xvattr_t xva;
979 int err;
980 fstrans_cookie_t cookie;
981
982 if (copy_from_user(&flags, arg, sizeof (flags)))
983 return (-EFAULT);
984
985 err = __zpl_ioctl_setflags(ip, flags, &xva);
986 if (err)
987 return (err);
988
989 crhold(cr);
990 cookie = spl_fstrans_mark();
991 err = -zfs_setattr(ITOZ(ip), (vattr_t *)&xva, 0, cr, zfs_init_idmap);
992 spl_fstrans_unmark(cookie);
993 crfree(cr);
994
995 return (err);
996 }
997
998 static int
zpl_ioctl_getxattr(struct file * filp,void __user * arg)999 zpl_ioctl_getxattr(struct file *filp, void __user *arg)
1000 {
1001 zfsxattr_t fsx = { 0 };
1002 struct inode *ip = file_inode(filp);
1003 int err;
1004
1005 fsx.fsx_xflags = __zpl_ioctl_getxflags(ip);
1006 fsx.fsx_projid = ITOZ(ip)->z_projid;
1007 err = copy_to_user(arg, &fsx, sizeof (fsx));
1008
1009 return (err);
1010 }
1011
1012 static int
zpl_ioctl_setxattr(struct file * filp,void __user * arg)1013 zpl_ioctl_setxattr(struct file *filp, void __user *arg)
1014 {
1015 struct inode *ip = file_inode(filp);
1016 zfsxattr_t fsx;
1017 cred_t *cr = CRED();
1018 xvattr_t xva;
1019 xoptattr_t *xoap;
1020 int err;
1021 fstrans_cookie_t cookie;
1022
1023 if (copy_from_user(&fsx, arg, sizeof (fsx)))
1024 return (-EFAULT);
1025
1026 if (!zpl_is_valid_projid(fsx.fsx_projid))
1027 return (-EINVAL);
1028
1029 err = __zpl_ioctl_setxflags(ip, fsx.fsx_xflags, &xva);
1030 if (err)
1031 return (err);
1032
1033 xoap = xva_getxoptattr(&xva);
1034 XVA_SET_REQ(&xva, XAT_PROJID);
1035 xoap->xoa_projid = fsx.fsx_projid;
1036
1037 crhold(cr);
1038 cookie = spl_fstrans_mark();
1039 err = -zfs_setattr(ITOZ(ip), (vattr_t *)&xva, 0, cr, zfs_init_idmap);
1040 spl_fstrans_unmark(cookie);
1041 crfree(cr);
1042
1043 return (err);
1044 }
1045
1046 /*
1047 * Expose Additional File Level Attributes of ZFS.
1048 */
1049 static int
zpl_ioctl_getdosflags(struct file * filp,void __user * arg)1050 zpl_ioctl_getdosflags(struct file *filp, void __user *arg)
1051 {
1052 struct inode *ip = file_inode(filp);
1053 uint64_t dosflags = ITOZ(ip)->z_pflags;
1054 dosflags &= ZFS_DOS_FL_USER_VISIBLE;
1055 int err = copy_to_user(arg, &dosflags, sizeof (dosflags));
1056
1057 return (err);
1058 }
1059
1060 static int
__zpl_ioctl_setdosflags(struct inode * ip,uint64_t ioctl_flags,xvattr_t * xva)1061 __zpl_ioctl_setdosflags(struct inode *ip, uint64_t ioctl_flags, xvattr_t *xva)
1062 {
1063 uint64_t zfs_flags = ITOZ(ip)->z_pflags;
1064 xoptattr_t *xoap;
1065
1066 if (ioctl_flags & (~ZFS_DOS_FL_USER_VISIBLE))
1067 return (-EOPNOTSUPP);
1068
1069 if ((fchange(ioctl_flags, zfs_flags, ZFS_IMMUTABLE, ZFS_IMMUTABLE) ||
1070 fchange(ioctl_flags, zfs_flags, ZFS_APPENDONLY, ZFS_APPENDONLY)) &&
1071 !capable(CAP_LINUX_IMMUTABLE))
1072 return (-EPERM);
1073
1074 if (!zpl_inode_owner_or_capable(zfs_init_idmap, ip))
1075 return (-EACCES);
1076
1077 xva_init(xva);
1078 xoap = xva_getxoptattr(xva);
1079
1080 #define FLAG_CHANGE(iflag, xflag, xfield) do { \
1081 if (((ioctl_flags & (iflag)) && !(zfs_flags & (iflag))) || \
1082 ((zfs_flags & (iflag)) && !(ioctl_flags & (iflag)))) { \
1083 XVA_SET_REQ(xva, (xflag)); \
1084 (xfield) = ((ioctl_flags & (iflag)) != 0); \
1085 } \
1086 } while (0)
1087
1088 FLAG_CHANGE(ZFS_IMMUTABLE, XAT_IMMUTABLE, xoap->xoa_immutable);
1089 FLAG_CHANGE(ZFS_APPENDONLY, XAT_APPENDONLY, xoap->xoa_appendonly);
1090 FLAG_CHANGE(ZFS_NODUMP, XAT_NODUMP, xoap->xoa_nodump);
1091 FLAG_CHANGE(ZFS_READONLY, XAT_READONLY, xoap->xoa_readonly);
1092 FLAG_CHANGE(ZFS_HIDDEN, XAT_HIDDEN, xoap->xoa_hidden);
1093 FLAG_CHANGE(ZFS_SYSTEM, XAT_SYSTEM, xoap->xoa_system);
1094 FLAG_CHANGE(ZFS_ARCHIVE, XAT_ARCHIVE, xoap->xoa_archive);
1095 FLAG_CHANGE(ZFS_NOUNLINK, XAT_NOUNLINK, xoap->xoa_nounlink);
1096 FLAG_CHANGE(ZFS_REPARSE, XAT_REPARSE, xoap->xoa_reparse);
1097 FLAG_CHANGE(ZFS_OFFLINE, XAT_OFFLINE, xoap->xoa_offline);
1098 FLAG_CHANGE(ZFS_SPARSE, XAT_SPARSE, xoap->xoa_sparse);
1099
1100 #undef FLAG_CHANGE
1101
1102 return (0);
1103 }
1104
1105 /*
1106 * Set Additional File Level Attributes of ZFS.
1107 */
1108 static int
zpl_ioctl_setdosflags(struct file * filp,void __user * arg)1109 zpl_ioctl_setdosflags(struct file *filp, void __user *arg)
1110 {
1111 struct inode *ip = file_inode(filp);
1112 uint64_t dosflags;
1113 cred_t *cr = CRED();
1114 xvattr_t xva;
1115 int err;
1116 fstrans_cookie_t cookie;
1117
1118 if (copy_from_user(&dosflags, arg, sizeof (dosflags)))
1119 return (-EFAULT);
1120
1121 err = __zpl_ioctl_setdosflags(ip, dosflags, &xva);
1122 if (err)
1123 return (err);
1124
1125 crhold(cr);
1126 cookie = spl_fstrans_mark();
1127 err = -zfs_setattr(ITOZ(ip), (vattr_t *)&xva, 0, cr, zfs_init_idmap);
1128 spl_fstrans_unmark(cookie);
1129 crfree(cr);
1130
1131 return (err);
1132 }
1133
1134 static int
zpl_ioctl_rewrite(struct file * filp,void __user * arg)1135 zpl_ioctl_rewrite(struct file *filp, void __user *arg)
1136 {
1137 struct inode *ip = file_inode(filp);
1138 zfs_rewrite_args_t args;
1139 fstrans_cookie_t cookie;
1140 int err;
1141
1142 if (copy_from_user(&args, arg, sizeof (args)))
1143 return (-EFAULT);
1144
1145 if (unlikely(!(filp->f_mode & FMODE_WRITE)))
1146 return (-EBADF);
1147
1148 cookie = spl_fstrans_mark();
1149 err = -zfs_rewrite(ITOZ(ip), args.off, args.len, args.flags, args.arg);
1150 spl_fstrans_unmark(cookie);
1151
1152 return (err);
1153 }
1154
1155 static long
zpl_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)1156 zpl_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1157 {
1158 switch (cmd) {
1159 case FS_IOC_GETVERSION:
1160 return (zpl_ioctl_getversion(filp, (void *)arg));
1161 case FS_IOC_GETFLAGS:
1162 return (zpl_ioctl_getflags(filp, (void *)arg));
1163 case FS_IOC_SETFLAGS:
1164 return (zpl_ioctl_setflags(filp, (void *)arg));
1165 case ZFS_IOC_FSGETXATTR:
1166 return (zpl_ioctl_getxattr(filp, (void *)arg));
1167 case ZFS_IOC_FSSETXATTR:
1168 return (zpl_ioctl_setxattr(filp, (void *)arg));
1169 case ZFS_IOC_GETDOSFLAGS:
1170 return (zpl_ioctl_getdosflags(filp, (void *)arg));
1171 case ZFS_IOC_SETDOSFLAGS:
1172 return (zpl_ioctl_setdosflags(filp, (void *)arg));
1173 case ZFS_IOC_REWRITE:
1174 return (zpl_ioctl_rewrite(filp, (void *)arg));
1175 default:
1176 return (-ENOTTY);
1177 }
1178 }
1179
1180 #ifdef CONFIG_COMPAT
1181 static long
zpl_compat_ioctl(struct file * filp,unsigned int cmd,unsigned long arg)1182 zpl_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1183 {
1184 switch (cmd) {
1185 case FS_IOC32_GETVERSION:
1186 cmd = FS_IOC_GETVERSION;
1187 break;
1188 case FS_IOC32_GETFLAGS:
1189 cmd = FS_IOC_GETFLAGS;
1190 break;
1191 case FS_IOC32_SETFLAGS:
1192 cmd = FS_IOC_SETFLAGS;
1193 break;
1194 default:
1195 return (-ENOTTY);
1196 }
1197 return (zpl_ioctl(filp, cmd, (unsigned long)compat_ptr(arg)));
1198 }
1199 #endif /* CONFIG_COMPAT */
1200
1201 const struct address_space_operations zpl_address_space_operations = {
1202 #ifdef HAVE_VFS_READPAGES
1203 .readpages = zpl_readpages,
1204 #else
1205 .readahead = zpl_readahead,
1206 #endif
1207 #ifdef HAVE_VFS_READ_FOLIO
1208 .read_folio = zpl_read_folio,
1209 #else
1210 .readpage = zpl_readpage,
1211 #endif
1212 #ifdef HAVE_VFS_WRITEPAGE
1213 .writepage = zpl_writepage,
1214 #endif
1215 .writepages = zpl_writepages,
1216 .direct_IO = zpl_direct_IO,
1217 #ifdef HAVE_VFS_SET_PAGE_DIRTY_NOBUFFERS
1218 .set_page_dirty = __set_page_dirty_nobuffers,
1219 #endif
1220 #ifdef HAVE_VFS_FILEMAP_DIRTY_FOLIO
1221 .dirty_folio = filemap_dirty_folio,
1222 #endif
1223 #ifdef HAVE_VFS_MIGRATE_FOLIO
1224 .migrate_folio = migrate_folio,
1225 #elif defined(HAVE_VFS_MIGRATEPAGE)
1226 .migratepage = migrate_page,
1227 #endif
1228 };
1229
1230 const struct file_operations zpl_file_operations = {
1231 .open = zpl_open,
1232 .release = zpl_release,
1233 .llseek = zpl_llseek,
1234 .read_iter = zpl_iter_read,
1235 .write_iter = zpl_iter_write,
1236 #ifdef HAVE_COPY_SPLICE_READ
1237 .splice_read = copy_splice_read,
1238 #else
1239 .splice_read = generic_file_splice_read,
1240 #endif
1241 .splice_write = iter_file_splice_write,
1242 .mmap = zpl_mmap,
1243 .fsync = zpl_fsync,
1244 .fallocate = zpl_fallocate,
1245 .setlease = generic_setlease,
1246 .copy_file_range = zpl_copy_file_range,
1247 #ifdef HAVE_VFS_CLONE_FILE_RANGE
1248 .clone_file_range = zpl_clone_file_range,
1249 #endif
1250 #ifdef HAVE_VFS_REMAP_FILE_RANGE
1251 .remap_file_range = zpl_remap_file_range,
1252 #endif
1253 #ifdef HAVE_VFS_DEDUPE_FILE_RANGE
1254 .dedupe_file_range = zpl_dedupe_file_range,
1255 #endif
1256 .fadvise = zpl_fadvise,
1257 .unlocked_ioctl = zpl_ioctl,
1258 #ifdef CONFIG_COMPAT
1259 .compat_ioctl = zpl_compat_ioctl,
1260 #endif
1261 };
1262
1263 const struct file_operations zpl_dir_file_operations = {
1264 .llseek = generic_file_llseek,
1265 .read = generic_read_dir,
1266 .iterate_shared = zpl_iterate,
1267 .fsync = zpl_fsync,
1268 .setlease = generic_setlease,
1269 .unlocked_ioctl = zpl_ioctl,
1270 #ifdef CONFIG_COMPAT
1271 .compat_ioctl = zpl_compat_ioctl,
1272 #endif
1273 };
1274
1275 module_param(zfs_fallocate_reserve_percent, uint, 0644);
1276 MODULE_PARM_DESC(zfs_fallocate_reserve_percent,
1277 "Percentage of length to use for the available capacity check");
1278