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.
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.
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]
22 * Copyright (c) 2011, Lawrence Livermore National Security, LLC.
23 * Copyright (c) 2015 by Chunwei Chen. All rights reserved.
28 #include <linux/compat.h>
31 #include <sys/dmu_objset.h>
32 #include <sys/zfs_vfsops.h>
33 #include <sys/zfs_vnops.h>
34 #include <sys/zfs_znode.h>
35 #include <sys/zfs_project.h>
39 zpl_open(struct inode *ip, struct file *filp)
43 fstrans_cookie_t cookie;
45 error = generic_file_open(ip, filp);
50 cookie = spl_fstrans_mark();
51 error = -zfs_open(ip, filp->f_mode, filp->f_flags, cr);
52 spl_fstrans_unmark(cookie);
54 ASSERT3S(error, <=, 0);
60 zpl_release(struct inode *ip, struct file *filp)
64 fstrans_cookie_t cookie;
66 cookie = spl_fstrans_mark();
67 if (ITOZ(ip)->z_atime_dirty)
68 zfs_mark_inode_dirty(ip);
71 error = -zfs_close(ip, filp->f_flags, cr);
72 spl_fstrans_unmark(cookie);
74 ASSERT3S(error, <=, 0);
80 zpl_iterate(struct file *filp, zpl_dir_context_t *ctx)
84 fstrans_cookie_t cookie;
87 cookie = spl_fstrans_mark();
88 error = -zfs_readdir(file_inode(filp), ctx, cr);
89 spl_fstrans_unmark(cookie);
91 ASSERT3S(error, <=, 0);
96 #if !defined(HAVE_VFS_ITERATE) && !defined(HAVE_VFS_ITERATE_SHARED)
98 zpl_readdir(struct file *filp, void *dirent, filldir_t filldir)
100 zpl_dir_context_t ctx =
101 ZPL_DIR_CONTEXT_INIT(dirent, filldir, filp->f_pos);
104 error = zpl_iterate(filp, &ctx);
105 filp->f_pos = ctx.pos;
109 #endif /* !HAVE_VFS_ITERATE && !HAVE_VFS_ITERATE_SHARED */
111 #if defined(HAVE_FSYNC_WITH_DENTRY)
113 * Linux 2.6.x - 2.6.34 API,
114 * Through 2.6.34 the nfsd kernel server would pass a NULL 'file struct *'
115 * to the fops->fsync() hook. For this reason, we must be careful not to
116 * use filp unconditionally.
119 zpl_fsync(struct file *filp, struct dentry *dentry, int datasync)
123 fstrans_cookie_t cookie;
126 cookie = spl_fstrans_mark();
127 error = -zfs_fsync(dentry->d_inode, datasync, cr);
128 spl_fstrans_unmark(cookie);
130 ASSERT3S(error, <=, 0);
135 #ifdef HAVE_FILE_AIO_FSYNC
137 zpl_aio_fsync(struct kiocb *kiocb, int datasync)
139 struct file *filp = kiocb->ki_filp;
140 return (zpl_fsync(filp, file_dentry(filp), datasync));
144 #elif defined(HAVE_FSYNC_WITHOUT_DENTRY)
146 * Linux 2.6.35 - 3.0 API,
147 * As of 2.6.35 the dentry argument to the fops->fsync() hook was deemed
148 * redundant. The dentry is still accessible via filp->f_path.dentry,
149 * and we are guaranteed that filp will never be NULL.
152 zpl_fsync(struct file *filp, int datasync)
154 struct inode *inode = filp->f_mapping->host;
157 fstrans_cookie_t cookie;
160 cookie = spl_fstrans_mark();
161 error = -zfs_fsync(inode, datasync, cr);
162 spl_fstrans_unmark(cookie);
164 ASSERT3S(error, <=, 0);
169 #ifdef HAVE_FILE_AIO_FSYNC
171 zpl_aio_fsync(struct kiocb *kiocb, int datasync)
173 return (zpl_fsync(kiocb->ki_filp, datasync));
177 #elif defined(HAVE_FSYNC_RANGE)
179 * Linux 3.1 - 3.x API,
180 * As of 3.1 the responsibility to call filemap_write_and_wait_range() has
181 * been pushed down in to the .fsync() vfs hook. Additionally, the i_mutex
182 * lock is no longer held by the caller, for zfs we don't require the lock
183 * to be held so we don't acquire it.
186 zpl_fsync(struct file *filp, loff_t start, loff_t end, int datasync)
188 struct inode *inode = filp->f_mapping->host;
191 fstrans_cookie_t cookie;
193 error = filemap_write_and_wait_range(inode->i_mapping, start, end);
198 cookie = spl_fstrans_mark();
199 error = -zfs_fsync(inode, datasync, cr);
200 spl_fstrans_unmark(cookie);
202 ASSERT3S(error, <=, 0);
207 #ifdef HAVE_FILE_AIO_FSYNC
209 zpl_aio_fsync(struct kiocb *kiocb, int datasync)
211 return (zpl_fsync(kiocb->ki_filp, kiocb->ki_pos, -1, datasync));
216 #error "Unsupported fops->fsync() implementation"
220 zfs_io_flags(struct kiocb *kiocb)
224 #if defined(IOCB_DSYNC)
225 if (kiocb->ki_flags & IOCB_DSYNC)
228 #if defined(IOCB_SYNC)
229 if (kiocb->ki_flags & IOCB_SYNC)
232 #if defined(IOCB_APPEND)
233 if (kiocb->ki_flags & IOCB_APPEND)
236 #if defined(IOCB_DIRECT)
237 if (kiocb->ki_flags & IOCB_DIRECT)
244 zpl_read_common_iovec(struct inode *ip, const struct iovec *iovp, size_t count,
245 unsigned long nr_segs, loff_t *ppos, uio_seg_t segment, int flags,
246 cred_t *cr, size_t skip)
249 uio_t uio = { { 0 }, 0 };
251 fstrans_cookie_t cookie;
254 uio.uio_iovcnt = nr_segs;
255 uio.uio_loffset = *ppos;
256 uio.uio_segflg = segment;
257 uio.uio_limit = MAXOFFSET_T;
258 uio.uio_resid = count;
261 cookie = spl_fstrans_mark();
262 error = -zfs_read(ip, &uio, flags, cr);
263 spl_fstrans_unmark(cookie);
267 read = count - uio.uio_resid;
274 zpl_read_common(struct inode *ip, const char *buf, size_t len, loff_t *ppos,
275 uio_seg_t segment, int flags, cred_t *cr)
279 iov.iov_base = (void *)buf;
282 return (zpl_read_common_iovec(ip, &iov, len, 1, ppos, segment,
287 zpl_iter_read_common(struct kiocb *kiocb, const struct iovec *iovp,
288 unsigned long nr_segs, size_t count, uio_seg_t seg, size_t skip)
291 struct file *filp = kiocb->ki_filp;
292 struct inode *ip = filp->f_mapping->host;
293 zfsvfs_t *zfsvfs = ZTOZSB(ITOZ(ip));
295 unsigned int f_flags = filp->f_flags;
297 f_flags |= zfs_io_flags(kiocb);
299 read = zpl_read_common_iovec(filp->f_mapping->host, iovp, count,
300 nr_segs, &kiocb->ki_pos, seg, f_flags, cr, skip);
304 * If relatime is enabled, call file_accessed() only if
305 * zfs_relatime_need_update() is true. This is needed since datasets
306 * with inherited "relatime" property aren't necessarily mounted with
307 * MNT_RELATIME flag (e.g. after `zfs set relatime=...`), which is what
308 * relatime test in VFS by relatime_need_update() is based on.
310 if (!IS_NOATIME(ip) && zfsvfs->z_relatime) {
311 if (zfs_relatime_need_update(ip))
320 #if defined(HAVE_VFS_RW_ITERATE)
322 zpl_iter_read(struct kiocb *kiocb, struct iov_iter *to)
325 uio_seg_t seg = UIO_USERSPACE;
326 if (to->type & ITER_KVEC)
328 if (to->type & ITER_BVEC)
330 ret = zpl_iter_read_common(kiocb, to->iov, to->nr_segs,
331 iov_iter_count(to), seg, to->iov_offset);
333 iov_iter_advance(to, ret);
338 zpl_aio_read(struct kiocb *kiocb, const struct iovec *iovp,
339 unsigned long nr_segs, loff_t pos)
344 ret = generic_segment_checks(iovp, &nr_segs, &count, VERIFY_WRITE);
348 return (zpl_iter_read_common(kiocb, iovp, nr_segs, count,
351 #endif /* HAVE_VFS_RW_ITERATE */
354 zpl_write_common_iovec(struct inode *ip, const struct iovec *iovp, size_t count,
355 unsigned long nr_segs, loff_t *ppos, uio_seg_t segment, int flags,
356 cred_t *cr, size_t skip)
359 uio_t uio = { { 0 }, 0 };
361 fstrans_cookie_t cookie;
363 if (flags & O_APPEND)
364 *ppos = i_size_read(ip);
367 uio.uio_iovcnt = nr_segs;
368 uio.uio_loffset = *ppos;
369 uio.uio_segflg = segment;
370 uio.uio_limit = MAXOFFSET_T;
371 uio.uio_resid = count;
374 cookie = spl_fstrans_mark();
375 error = -zfs_write(ip, &uio, flags, cr);
376 spl_fstrans_unmark(cookie);
380 wrote = count - uio.uio_resid;
387 zpl_write_common(struct inode *ip, const char *buf, size_t len, loff_t *ppos,
388 uio_seg_t segment, int flags, cred_t *cr)
392 iov.iov_base = (void *)buf;
395 return (zpl_write_common_iovec(ip, &iov, len, 1, ppos, segment,
400 zpl_iter_write_common(struct kiocb *kiocb, const struct iovec *iovp,
401 unsigned long nr_segs, size_t count, uio_seg_t seg, size_t skip)
404 struct file *filp = kiocb->ki_filp;
406 unsigned int f_flags = filp->f_flags;
408 f_flags |= zfs_io_flags(kiocb);
410 wrote = zpl_write_common_iovec(filp->f_mapping->host, iovp, count,
411 nr_segs, &kiocb->ki_pos, seg, f_flags, cr, skip);
417 #if defined(HAVE_VFS_RW_ITERATE)
419 zpl_iter_write(struct kiocb *kiocb, struct iov_iter *from)
423 uio_seg_t seg = UIO_USERSPACE;
425 #ifndef HAVE_GENERIC_WRITE_CHECKS_KIOCB
426 struct file *file = kiocb->ki_filp;
427 struct address_space *mapping = file->f_mapping;
428 struct inode *ip = mapping->host;
429 int isblk = S_ISBLK(ip->i_mode);
431 count = iov_iter_count(from);
432 ret = generic_write_checks(file, &kiocb->ki_pos, &count, isblk);
437 * XXX - ideally this check should be in the same lock region with
438 * write operations, so that there's no TOCTTOU race when doing
439 * append and someone else grow the file.
441 ret = generic_write_checks(kiocb, from);
447 if (from->type & ITER_KVEC)
449 if (from->type & ITER_BVEC)
452 ret = zpl_iter_write_common(kiocb, from->iov, from->nr_segs,
453 count, seg, from->iov_offset);
455 iov_iter_advance(from, ret);
461 zpl_aio_write(struct kiocb *kiocb, const struct iovec *iovp,
462 unsigned long nr_segs, loff_t pos)
464 struct file *file = kiocb->ki_filp;
465 struct address_space *mapping = file->f_mapping;
466 struct inode *ip = mapping->host;
467 int isblk = S_ISBLK(ip->i_mode);
471 ret = generic_segment_checks(iovp, &nr_segs, &count, VERIFY_READ);
475 ret = generic_write_checks(file, &pos, &count, isblk);
479 return (zpl_iter_write_common(kiocb, iovp, nr_segs, count,
482 #endif /* HAVE_VFS_RW_ITERATE */
484 #if defined(HAVE_VFS_RW_ITERATE)
486 zpl_direct_IO_impl(int rw, struct kiocb *kiocb, struct iov_iter *iter)
489 return (zpl_iter_write(kiocb, iter));
491 return (zpl_iter_read(kiocb, iter));
493 #if defined(HAVE_VFS_DIRECT_IO_ITER)
495 zpl_direct_IO(struct kiocb *kiocb, struct iov_iter *iter)
497 return (zpl_direct_IO_impl(iov_iter_rw(iter), kiocb, iter));
499 #elif defined(HAVE_VFS_DIRECT_IO_ITER_OFFSET)
501 zpl_direct_IO(struct kiocb *kiocb, struct iov_iter *iter, loff_t pos)
503 ASSERT3S(pos, ==, kiocb->ki_pos);
504 return (zpl_direct_IO_impl(iov_iter_rw(iter), kiocb, iter));
506 #elif defined(HAVE_VFS_DIRECT_IO_ITER_RW_OFFSET)
508 zpl_direct_IO(int rw, struct kiocb *kiocb, struct iov_iter *iter, loff_t pos)
510 ASSERT3S(pos, ==, kiocb->ki_pos);
511 return (zpl_direct_IO_impl(rw, kiocb, iter));
514 #error "Unknown direct IO interface"
519 #if defined(HAVE_VFS_DIRECT_IO_IOVEC)
521 zpl_direct_IO(int rw, struct kiocb *kiocb, const struct iovec *iovp,
522 loff_t pos, unsigned long nr_segs)
525 return (zpl_aio_write(kiocb, iovp, nr_segs, pos));
527 return (zpl_aio_read(kiocb, iovp, nr_segs, pos));
530 #error "Unknown direct IO interface"
533 #endif /* HAVE_VFS_RW_ITERATE */
536 zpl_llseek(struct file *filp, loff_t offset, int whence)
538 #if defined(SEEK_HOLE) && defined(SEEK_DATA)
539 fstrans_cookie_t cookie;
541 if (whence == SEEK_DATA || whence == SEEK_HOLE) {
542 struct inode *ip = filp->f_mapping->host;
543 loff_t maxbytes = ip->i_sb->s_maxbytes;
546 spl_inode_lock_shared(ip);
547 cookie = spl_fstrans_mark();
548 error = -zfs_holey(ip, whence, &offset);
549 spl_fstrans_unmark(cookie);
551 error = lseek_execute(filp, ip, offset, maxbytes);
552 spl_inode_unlock_shared(ip);
556 #endif /* SEEK_HOLE && SEEK_DATA */
558 return (generic_file_llseek(filp, offset, whence));
562 * It's worth taking a moment to describe how mmap is implemented
563 * for zfs because it differs considerably from other Linux filesystems.
564 * However, this issue is handled the same way under OpenSolaris.
566 * The issue is that by design zfs bypasses the Linux page cache and
567 * leaves all caching up to the ARC. This has been shown to work
568 * well for the common read(2)/write(2) case. However, mmap(2)
569 * is problem because it relies on being tightly integrated with the
570 * page cache. To handle this we cache mmap'ed files twice, once in
571 * the ARC and a second time in the page cache. The code is careful
572 * to keep both copies synchronized.
574 * When a file with an mmap'ed region is written to using write(2)
575 * both the data in the ARC and existing pages in the page cache
576 * are updated. For a read(2) data will be read first from the page
577 * cache then the ARC if needed. Neither a write(2) or read(2) will
578 * will ever result in new pages being added to the page cache.
580 * New pages are added to the page cache only via .readpage() which
581 * is called when the vfs needs to read a page off disk to back the
582 * virtual memory region. These pages may be modified without
583 * notifying the ARC and will be written out periodically via
584 * .writepage(). This will occur due to either a sync or the usual
585 * page aging behavior. Note because a read(2) of a mmap'ed file
586 * will always check the page cache first even when the ARC is out
587 * of date correct data will still be returned.
589 * While this implementation ensures correct behavior it does have
590 * have some drawbacks. The most obvious of which is that it
591 * increases the required memory footprint when access mmap'ed
592 * files. It also adds additional complexity to the code keeping
593 * both caches synchronized.
595 * Longer term it may be possible to cleanly resolve this wart by
596 * mapping page cache pages directly on to the ARC buffers. The
597 * Linux address space operations are flexible enough to allow
598 * selection of which pages back a particular index. The trick
599 * would be working out the details of which subsystem is in
600 * charge, the ARC, the page cache, or both. It may also prove
601 * helpful to move the ARC buffers to a scatter-gather lists
602 * rather than a vmalloc'ed region.
605 zpl_mmap(struct file *filp, struct vm_area_struct *vma)
607 struct inode *ip = filp->f_mapping->host;
608 znode_t *zp = ITOZ(ip);
610 fstrans_cookie_t cookie;
612 cookie = spl_fstrans_mark();
613 error = -zfs_map(ip, vma->vm_pgoff, (caddr_t *)vma->vm_start,
614 (size_t)(vma->vm_end - vma->vm_start), vma->vm_flags);
615 spl_fstrans_unmark(cookie);
619 error = generic_file_mmap(filp, vma);
623 mutex_enter(&zp->z_lock);
624 zp->z_is_mapped = B_TRUE;
625 mutex_exit(&zp->z_lock);
631 * Populate a page with data for the Linux page cache. This function is
632 * only used to support mmap(2). There will be an identical copy of the
633 * data in the ARC which is kept up to date via .write() and .writepage().
635 * Current this function relies on zpl_read_common() and the O_DIRECT
636 * flag to read in a page. This works but the more correct way is to
637 * update zfs_fillpage() to be Linux friendly and use that interface.
640 zpl_readpage(struct file *filp, struct page *pp)
645 fstrans_cookie_t cookie;
647 ASSERT(PageLocked(pp));
648 ip = pp->mapping->host;
651 cookie = spl_fstrans_mark();
652 error = -zfs_getpage(ip, pl, 1);
653 spl_fstrans_unmark(cookie);
657 ClearPageUptodate(pp);
661 flush_dcache_page(pp);
669 * Populate a set of pages with data for the Linux page cache. This
670 * function will only be called for read ahead and never for demand
671 * paging. For simplicity, the code relies on read_cache_pages() to
672 * correctly lock each page for IO and call zpl_readpage().
675 zpl_readpages(struct file *filp, struct address_space *mapping,
676 struct list_head *pages, unsigned nr_pages)
678 return (read_cache_pages(mapping, pages,
679 (filler_t *)zpl_readpage, filp));
683 zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data)
685 struct address_space *mapping = data;
686 fstrans_cookie_t cookie;
688 ASSERT(PageLocked(pp));
689 ASSERT(!PageWriteback(pp));
691 cookie = spl_fstrans_mark();
692 (void) zfs_putpage(mapping->host, pp, wbc);
693 spl_fstrans_unmark(cookie);
699 zpl_writepages(struct address_space *mapping, struct writeback_control *wbc)
701 znode_t *zp = ITOZ(mapping->host);
702 zfsvfs_t *zfsvfs = ITOZSB(mapping->host);
703 enum writeback_sync_modes sync_mode;
707 if (zfsvfs->z_os->os_sync == ZFS_SYNC_ALWAYS)
708 wbc->sync_mode = WB_SYNC_ALL;
710 sync_mode = wbc->sync_mode;
713 * We don't want to run write_cache_pages() in SYNC mode here, because
714 * that would make putpage() wait for a single page to be committed to
715 * disk every single time, resulting in atrocious performance. Instead
716 * we run it once in non-SYNC mode so that the ZIL gets all the data,
717 * and then we commit it all in one go.
719 wbc->sync_mode = WB_SYNC_NONE;
720 result = write_cache_pages(mapping, wbc, zpl_putpage, mapping);
721 if (sync_mode != wbc->sync_mode) {
724 if (zfsvfs->z_log != NULL)
725 zil_commit(zfsvfs->z_log, zp->z_id);
729 * We need to call write_cache_pages() again (we can't just
730 * return after the commit) because the previous call in
731 * non-SYNC mode does not guarantee that we got all the dirty
732 * pages (see the implementation of write_cache_pages() for
733 * details). That being said, this is a no-op in most cases.
735 wbc->sync_mode = sync_mode;
736 result = write_cache_pages(mapping, wbc, zpl_putpage, mapping);
742 * Write out dirty pages to the ARC, this function is only required to
743 * support mmap(2). Mapped pages may be dirtied by memory operations
744 * which never call .write(). These dirty pages are kept in sync with
745 * the ARC buffers via this hook.
748 zpl_writepage(struct page *pp, struct writeback_control *wbc)
750 if (ITOZSB(pp->mapping->host)->z_os->os_sync == ZFS_SYNC_ALWAYS)
751 wbc->sync_mode = WB_SYNC_ALL;
753 return (zpl_putpage(pp, wbc, pp->mapping));
757 * The only flag combination which matches the behavior of zfs_space()
758 * is FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE. The FALLOC_FL_PUNCH_HOLE
759 * flag was introduced in the 2.6.38 kernel.
761 #if defined(HAVE_FILE_FALLOCATE) || defined(HAVE_INODE_FALLOCATE)
763 zpl_fallocate_common(struct inode *ip, int mode, loff_t offset, loff_t len)
765 int error = -EOPNOTSUPP;
767 #if defined(FALLOC_FL_PUNCH_HOLE) && defined(FALLOC_FL_KEEP_SIZE)
771 fstrans_cookie_t cookie;
773 if (mode != (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))
776 if (offset < 0 || len <= 0)
780 olen = i_size_read(ip);
783 spl_inode_unlock(ip);
786 if (offset + len > olen)
789 bf.l_whence = SEEK_SET;
795 cookie = spl_fstrans_mark();
796 error = -zfs_space(ip, F_FREESP, &bf, FWRITE, offset, cr);
797 spl_fstrans_unmark(cookie);
798 spl_inode_unlock(ip);
801 #endif /* defined(FALLOC_FL_PUNCH_HOLE) && defined(FALLOC_FL_KEEP_SIZE) */
803 ASSERT3S(error, <=, 0);
806 #endif /* defined(HAVE_FILE_FALLOCATE) || defined(HAVE_INODE_FALLOCATE) */
808 #ifdef HAVE_FILE_FALLOCATE
810 zpl_fallocate(struct file *filp, int mode, loff_t offset, loff_t len)
812 return zpl_fallocate_common(file_inode(filp),
815 #endif /* HAVE_FILE_FALLOCATE */
817 #define ZFS_FL_USER_VISIBLE (FS_FL_USER_VISIBLE | ZFS_PROJINHERIT_FL)
818 #define ZFS_FL_USER_MODIFIABLE (FS_FL_USER_MODIFIABLE | ZFS_PROJINHERIT_FL)
821 __zpl_ioctl_getflags(struct inode *ip)
823 uint64_t zfs_flags = ITOZ(ip)->z_pflags;
824 uint32_t ioctl_flags = 0;
826 if (zfs_flags & ZFS_IMMUTABLE)
827 ioctl_flags |= FS_IMMUTABLE_FL;
829 if (zfs_flags & ZFS_APPENDONLY)
830 ioctl_flags |= FS_APPEND_FL;
832 if (zfs_flags & ZFS_NODUMP)
833 ioctl_flags |= FS_NODUMP_FL;
835 if (zfs_flags & ZFS_PROJINHERIT)
836 ioctl_flags |= ZFS_PROJINHERIT_FL;
838 return (ioctl_flags & ZFS_FL_USER_VISIBLE);
842 * Map zfs file z_pflags (xvattr_t) to linux file attributes. Only file
843 * attributes common to both Linux and Solaris are mapped.
846 zpl_ioctl_getflags(struct file *filp, void __user *arg)
851 flags = __zpl_ioctl_getflags(file_inode(filp));
852 err = copy_to_user(arg, &flags, sizeof (flags));
858 * fchange() is a helper macro to detect if we have been asked to change a
859 * flag. This is ugly, but the requirement that we do this is a consequence of
860 * how the Linux file attribute interface was designed. Another consequence is
861 * that concurrent modification of files suffers from a TOCTOU race. Neither
862 * are things we can fix without modifying the kernel-userland interface, which
863 * is outside of our jurisdiction.
866 #define fchange(f0, f1, b0, b1) (!((f0) & (b0)) != !((f1) & (b1)))
869 __zpl_ioctl_setflags(struct inode *ip, uint32_t ioctl_flags, xvattr_t *xva)
871 uint64_t zfs_flags = ITOZ(ip)->z_pflags;
874 if (ioctl_flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | FS_NODUMP_FL |
876 return (-EOPNOTSUPP);
878 if (ioctl_flags & ~ZFS_FL_USER_MODIFIABLE)
881 if ((fchange(ioctl_flags, zfs_flags, FS_IMMUTABLE_FL, ZFS_IMMUTABLE) ||
882 fchange(ioctl_flags, zfs_flags, FS_APPEND_FL, ZFS_APPENDONLY)) &&
883 !capable(CAP_LINUX_IMMUTABLE))
886 if (!zpl_inode_owner_or_capable(ip))
890 xoap = xva_getxoptattr(xva);
892 XVA_SET_REQ(xva, XAT_IMMUTABLE);
893 if (ioctl_flags & FS_IMMUTABLE_FL)
894 xoap->xoa_immutable = B_TRUE;
896 XVA_SET_REQ(xva, XAT_APPENDONLY);
897 if (ioctl_flags & FS_APPEND_FL)
898 xoap->xoa_appendonly = B_TRUE;
900 XVA_SET_REQ(xva, XAT_NODUMP);
901 if (ioctl_flags & FS_NODUMP_FL)
902 xoap->xoa_nodump = B_TRUE;
904 XVA_SET_REQ(xva, XAT_PROJINHERIT);
905 if (ioctl_flags & ZFS_PROJINHERIT_FL)
906 xoap->xoa_projinherit = B_TRUE;
912 zpl_ioctl_setflags(struct file *filp, void __user *arg)
914 struct inode *ip = file_inode(filp);
919 fstrans_cookie_t cookie;
921 if (copy_from_user(&flags, arg, sizeof (flags)))
924 err = __zpl_ioctl_setflags(ip, flags, &xva);
929 cookie = spl_fstrans_mark();
930 err = -zfs_setattr(ip, (vattr_t *)&xva, 0, cr);
931 spl_fstrans_unmark(cookie);
938 zpl_ioctl_getxattr(struct file *filp, void __user *arg)
940 zfsxattr_t fsx = { 0 };
941 struct inode *ip = file_inode(filp);
944 fsx.fsx_xflags = __zpl_ioctl_getflags(ip);
945 fsx.fsx_projid = ITOZ(ip)->z_projid;
946 err = copy_to_user(arg, &fsx, sizeof (fsx));
952 zpl_ioctl_setxattr(struct file *filp, void __user *arg)
954 struct inode *ip = file_inode(filp);
960 fstrans_cookie_t cookie;
962 if (copy_from_user(&fsx, arg, sizeof (fsx)))
965 if (!zpl_is_valid_projid(fsx.fsx_projid))
968 err = __zpl_ioctl_setflags(ip, fsx.fsx_xflags, &xva);
972 xoap = xva_getxoptattr(&xva);
973 XVA_SET_REQ(&xva, XAT_PROJID);
974 xoap->xoa_projid = fsx.fsx_projid;
977 cookie = spl_fstrans_mark();
978 err = -zfs_setattr(ip, (vattr_t *)&xva, 0, cr);
979 spl_fstrans_unmark(cookie);
986 zpl_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
989 case FS_IOC_GETFLAGS:
990 return (zpl_ioctl_getflags(filp, (void *)arg));
991 case FS_IOC_SETFLAGS:
992 return (zpl_ioctl_setflags(filp, (void *)arg));
993 case ZFS_IOC_FSGETXATTR:
994 return (zpl_ioctl_getxattr(filp, (void *)arg));
995 case ZFS_IOC_FSSETXATTR:
996 return (zpl_ioctl_setxattr(filp, (void *)arg));
1002 #ifdef CONFIG_COMPAT
1004 zpl_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
1007 case FS_IOC32_GETFLAGS:
1008 cmd = FS_IOC_GETFLAGS;
1010 case FS_IOC32_SETFLAGS:
1011 cmd = FS_IOC_SETFLAGS;
1016 return (zpl_ioctl(filp, cmd, (unsigned long)compat_ptr(arg)));
1018 #endif /* CONFIG_COMPAT */
1021 const struct address_space_operations zpl_address_space_operations = {
1022 .readpages = zpl_readpages,
1023 .readpage = zpl_readpage,
1024 .writepage = zpl_writepage,
1025 .writepages = zpl_writepages,
1026 .direct_IO = zpl_direct_IO,
1029 const struct file_operations zpl_file_operations = {
1031 .release = zpl_release,
1032 .llseek = zpl_llseek,
1033 #ifdef HAVE_VFS_RW_ITERATE
1034 #ifdef HAVE_NEW_SYNC_READ
1035 .read = new_sync_read,
1036 .write = new_sync_write,
1038 .read_iter = zpl_iter_read,
1039 .write_iter = zpl_iter_write,
1041 .read = do_sync_read,
1042 .write = do_sync_write,
1043 .aio_read = zpl_aio_read,
1044 .aio_write = zpl_aio_write,
1048 #ifdef HAVE_FILE_AIO_FSYNC
1049 .aio_fsync = zpl_aio_fsync,
1051 #ifdef HAVE_FILE_FALLOCATE
1052 .fallocate = zpl_fallocate,
1053 #endif /* HAVE_FILE_FALLOCATE */
1054 .unlocked_ioctl = zpl_ioctl,
1055 #ifdef CONFIG_COMPAT
1056 .compat_ioctl = zpl_compat_ioctl,
1060 const struct file_operations zpl_dir_file_operations = {
1061 .llseek = generic_file_llseek,
1062 .read = generic_read_dir,
1063 #if defined(HAVE_VFS_ITERATE_SHARED)
1064 .iterate_shared = zpl_iterate,
1065 #elif defined(HAVE_VFS_ITERATE)
1066 .iterate = zpl_iterate,
1068 .readdir = zpl_readdir,
1071 .unlocked_ioctl = zpl_ioctl,
1072 #ifdef CONFIG_COMPAT
1073 .compat_ioctl = zpl_compat_ioctl,