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.
26 #include <sys/zfs_vfsops.h>
27 #include <sys/zfs_vnops.h>
28 #include <sys/zfs_znode.h>
33 zpl_open(struct inode
*ip
, struct file
*filp
)
39 error
= -zfs_open(ip
, filp
->f_mode
, filp
->f_flags
, cr
);
41 ASSERT3S(error
, <=, 0);
46 return generic_file_open(ip
, filp
);
50 zpl_release(struct inode
*ip
, struct file
*filp
)
56 error
= -zfs_close(ip
, filp
->f_flags
, cr
);
58 ASSERT3S(error
, <=, 0);
64 zpl_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
66 struct dentry
*dentry
= filp
->f_path
.dentry
;
71 error
= -zfs_readdir(dentry
->d_inode
, dirent
, filldir
,
74 ASSERT3S(error
, <=, 0);
79 #if defined(HAVE_FSYNC_WITH_DENTRY)
81 * Linux 2.6.x - 2.6.34 API,
82 * Through 2.6.34 the nfsd kernel server would pass a NULL 'file struct *'
83 * to the fops->fsync() hook. For this reason, we must be careful not to
84 * use filp unconditionally.
87 zpl_fsync(struct file
*filp
, struct dentry
*dentry
, int datasync
)
93 error
= -zfs_fsync(dentry
->d_inode
, datasync
, cr
);
95 ASSERT3S(error
, <=, 0);
100 #elif defined(HAVE_FSYNC_WITHOUT_DENTRY)
102 * Linux 2.6.35 - 3.0 API,
103 * As of 2.6.35 the dentry argument to the fops->fsync() hook was deemed
104 * redundant. The dentry is still accessible via filp->f_path.dentry,
105 * and we are guaranteed that filp will never be NULL.
108 zpl_fsync(struct file
*filp
, int datasync
)
110 struct inode
*inode
= filp
->f_mapping
->host
;
115 error
= -zfs_fsync(inode
, datasync
, cr
);
117 ASSERT3S(error
, <=, 0);
122 #elif defined(HAVE_FSYNC_RANGE)
124 * Linux 3.1 - 3.x API,
125 * As of 3.1 the responsibility to call filemap_write_and_wait_range() has
126 * been pushed down in to the .fsync() vfs hook. Additionally, the i_mutex
127 * lock is no longer held by the caller, for zfs we don't require the lock
128 * to be held so we don't acquire it.
131 zpl_fsync(struct file
*filp
, loff_t start
, loff_t end
, int datasync
)
133 struct inode
*inode
= filp
->f_mapping
->host
;
137 error
= filemap_write_and_wait_range(inode
->i_mapping
, start
, end
);
142 error
= -zfs_fsync(inode
, datasync
, cr
);
144 ASSERT3S(error
, <=, 0);
149 #error "Unsupported fops->fsync() implementation"
153 zpl_read_common(struct inode
*ip
, const char *buf
, size_t len
, loff_t pos
,
154 uio_seg_t segment
, int flags
, cred_t
*cr
)
160 iov
.iov_base
= (void *)buf
;
166 uio
.uio_loffset
= pos
;
167 uio
.uio_limit
= MAXOFFSET_T
;
168 uio
.uio_segflg
= segment
;
170 error
= -zfs_read(ip
, &uio
, flags
, cr
);
174 return (len
- uio
.uio_resid
);
178 zpl_read(struct file
*filp
, char __user
*buf
, size_t len
, loff_t
*ppos
)
184 read
= zpl_read_common(filp
->f_mapping
->host
, buf
, len
, *ppos
,
185 UIO_USERSPACE
, filp
->f_flags
, cr
);
196 zpl_write_common(struct inode
*ip
, const char *buf
, size_t len
, loff_t pos
,
197 uio_seg_t segment
, int flags
, cred_t
*cr
)
203 iov
.iov_base
= (void *)buf
;
209 uio
.uio_loffset
= pos
;
210 uio
.uio_limit
= MAXOFFSET_T
;
211 uio
.uio_segflg
= segment
;
213 error
= -zfs_write(ip
, &uio
, flags
, cr
);
217 return (len
- uio
.uio_resid
);
221 zpl_write(struct file
*filp
, const char __user
*buf
, size_t len
, loff_t
*ppos
)
227 wrote
= zpl_write_common(filp
->f_mapping
->host
, buf
, len
, *ppos
,
228 UIO_USERSPACE
, filp
->f_flags
, cr
);
239 * It's worth taking a moment to describe how mmap is implemented
240 * for zfs because it differs considerably from other Linux filesystems.
241 * However, this issue is handled the same way under OpenSolaris.
243 * The issue is that by design zfs bypasses the Linux page cache and
244 * leaves all caching up to the ARC. This has been shown to work
245 * well for the common read(2)/write(2) case. However, mmap(2)
246 * is problem because it relies on being tightly integrated with the
247 * page cache. To handle this we cache mmap'ed files twice, once in
248 * the ARC and a second time in the page cache. The code is careful
249 * to keep both copies synchronized.
251 * When a file with an mmap'ed region is written to using write(2)
252 * both the data in the ARC and existing pages in the page cache
253 * are updated. For a read(2) data will be read first from the page
254 * cache then the ARC if needed. Neither a write(2) or read(2) will
255 * will ever result in new pages being added to the page cache.
257 * New pages are added to the page cache only via .readpage() which
258 * is called when the vfs needs to read a page off disk to back the
259 * virtual memory region. These pages may be modified without
260 * notifying the ARC and will be written out periodically via
261 * .writepage(). This will occur due to either a sync or the usual
262 * page aging behavior. Note because a read(2) of a mmap'ed file
263 * will always check the page cache first even when the ARC is out
264 * of date correct data will still be returned.
266 * While this implementation ensures correct behavior it does have
267 * have some drawbacks. The most obvious of which is that it
268 * increases the required memory footprint when access mmap'ed
269 * files. It also adds additional complexity to the code keeping
270 * both caches synchronized.
272 * Longer term it may be possible to cleanly resolve this wart by
273 * mapping page cache pages directly on to the ARC buffers. The
274 * Linux address space operations are flexible enough to allow
275 * selection of which pages back a particular index. The trick
276 * would be working out the details of which subsystem is in
277 * charge, the ARC, the page cache, or both. It may also prove
278 * helpful to move the ARC buffers to a scatter-gather lists
279 * rather than a vmalloc'ed region.
282 zpl_mmap(struct file
*filp
, struct vm_area_struct
*vma
)
284 struct inode
*ip
= filp
->f_mapping
->host
;
285 znode_t
*zp
= ITOZ(ip
);
288 error
= -zfs_map(ip
, vma
->vm_pgoff
, (caddr_t
*)vma
->vm_start
,
289 (size_t)(vma
->vm_end
- vma
->vm_start
), vma
->vm_flags
);
293 error
= generic_file_mmap(filp
, vma
);
297 mutex_enter(&zp
->z_lock
);
299 mutex_exit(&zp
->z_lock
);
305 * Populate a page with data for the Linux page cache. This function is
306 * only used to support mmap(2). There will be an identical copy of the
307 * data in the ARC which is kept up to date via .write() and .writepage().
309 * Current this function relies on zpl_read_common() and the O_DIRECT
310 * flag to read in a page. This works but the more correct way is to
311 * update zfs_fillpage() to be Linux friendly and use that interface.
314 zpl_readpage(struct file
*filp
, struct page
*pp
)
320 ASSERT(PageLocked(pp
));
321 ip
= pp
->mapping
->host
;
324 error
= -zfs_getpage(ip
, pl
, 1);
328 ClearPageUptodate(pp
);
332 flush_dcache_page(pp
);
340 * Populate a set of pages with data for the Linux page cache. This
341 * function will only be called for read ahead and never for demand
342 * paging. For simplicity, the code relies on read_cache_pages() to
343 * correctly lock each page for IO and call zpl_readpage().
346 zpl_readpages(struct file
*filp
, struct address_space
*mapping
,
347 struct list_head
*pages
, unsigned nr_pages
)
349 return (read_cache_pages(mapping
, pages
,
350 (filler_t
*)zpl_readpage
, filp
));
354 zpl_putpage(struct page
*pp
, struct writeback_control
*wbc
, void *data
)
356 struct address_space
*mapping
= data
;
358 ASSERT(PageLocked(pp
));
359 ASSERT(!PageWriteback(pp
));
361 (void) zfs_putpage(mapping
->host
, pp
, wbc
);
367 zpl_writepages(struct address_space
*mapping
, struct writeback_control
*wbc
)
369 return write_cache_pages(mapping
, wbc
, zpl_putpage
, mapping
);
373 * Write out dirty pages to the ARC, this function is only required to
374 * support mmap(2). Mapped pages may be dirtied by memory operations
375 * which never call .write(). These dirty pages are kept in sync with
376 * the ARC buffers via this hook.
379 zpl_writepage(struct page
*pp
, struct writeback_control
*wbc
)
381 return zpl_putpage(pp
, wbc
, pp
->mapping
);
385 * The only flag combination which matches the behavior of zfs_space()
386 * is FALLOC_FL_PUNCH_HOLE. This flag was introduced in the 2.6.38 kernel.
389 zpl_fallocate_common(struct inode
*ip
, int mode
, loff_t offset
, loff_t len
)
392 int error
= -EOPNOTSUPP
;
394 if (mode
& FALLOC_FL_KEEP_SIZE
)
395 return (-EOPNOTSUPP
);
399 #ifdef FALLOC_FL_PUNCH_HOLE
400 if (mode
& FALLOC_FL_PUNCH_HOLE
) {
409 error
= -zfs_space(ip
, F_FREESP
, &bf
, FWRITE
, offset
, cr
);
411 #endif /* FALLOC_FL_PUNCH_HOLE */
415 ASSERT3S(error
, <=, 0);
419 #ifdef HAVE_FILE_FALLOCATE
421 zpl_fallocate(struct file
*filp
, int mode
, loff_t offset
, loff_t len
)
423 return zpl_fallocate_common(filp
->f_path
.dentry
->d_inode
,
426 #endif /* HAVE_FILE_FALLOCATE */
428 const struct address_space_operations zpl_address_space_operations
= {
429 .readpages
= zpl_readpages
,
430 .readpage
= zpl_readpage
,
431 .writepage
= zpl_writepage
,
432 .writepages
= zpl_writepages
,
435 const struct file_operations zpl_file_operations
= {
437 .release
= zpl_release
,
438 .llseek
= generic_file_llseek
,
441 .readdir
= zpl_readdir
,
444 #ifdef HAVE_FILE_FALLOCATE
445 .fallocate
= zpl_fallocate
,
446 #endif /* HAVE_FILE_FALLOCATE */
449 const struct file_operations zpl_dir_file_operations
= {
450 .llseek
= generic_file_llseek
,
451 .read
= generic_read_dir
,
452 .readdir
= zpl_readdir
,