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_readdir(struct file
*filp
, void *dirent
, filldir_t filldir
)
35 struct dentry
*dentry
= filp
->f_path
.dentry
;
39 cr
= (cred_t
*)get_current_cred();
40 error
= -zfs_readdir(dentry
->d_inode
, dirent
, filldir
,
43 ASSERT3S(error
, <=, 0);
48 ZPL_FSYNC_PROTO(zpl_fsync
, filp
, unused_dentry
, datasync
)
53 cr
= (cred_t
*)get_current_cred();
54 error
= -zfs_fsync(filp
->f_path
.dentry
->d_inode
, datasync
, cr
);
56 ASSERT3S(error
, <=, 0);
62 zpl_read_common(struct inode
*ip
, const char *buf
, size_t len
, loff_t pos
,
63 uio_seg_t segment
, int flags
, cred_t
*cr
)
69 iov
.iov_base
= (void *)buf
;
75 uio
.uio_loffset
= pos
;
76 uio
.uio_limit
= MAXOFFSET_T
;
77 uio
.uio_segflg
= segment
;
79 error
= -zfs_read(ip
, &uio
, flags
, cr
);
83 return (len
- uio
.uio_resid
);
87 zpl_read(struct file
*filp
, char __user
*buf
, size_t len
, loff_t
*ppos
)
92 cr
= (cred_t
*)get_current_cred();
93 read
= zpl_read_common(filp
->f_mapping
->host
, buf
, len
, *ppos
,
94 UIO_USERSPACE
, filp
->f_flags
, cr
);
105 zpl_write_common(struct inode
*ip
, const char *buf
, size_t len
, loff_t pos
,
106 uio_seg_t segment
, int flags
, cred_t
*cr
)
112 iov
.iov_base
= (void *)buf
;
118 uio
.uio_loffset
= pos
;
119 uio
.uio_limit
= MAXOFFSET_T
;
120 uio
.uio_segflg
= segment
;
122 error
= -zfs_write(ip
, &uio
, flags
, cr
);
126 return (len
- uio
.uio_resid
);
130 zpl_write(struct file
*filp
, const char __user
*buf
, size_t len
, loff_t
*ppos
)
135 cr
= (cred_t
*)get_current_cred();
136 wrote
= zpl_write_common(filp
->f_mapping
->host
, buf
, len
, *ppos
,
137 UIO_USERSPACE
, filp
->f_flags
, cr
);
148 * It's worth taking a moment to describe how mmap is implemented
149 * for zfs because it differs considerably from other Linux filesystems.
150 * However, this issue is handled the same way under OpenSolaris.
152 * The issue is that by design zfs bypasses the Linux page cache and
153 * leaves all caching up to the ARC. This has been shown to work
154 * well for the common read(2)/write(2) case. However, mmap(2)
155 * is problem because it relies on being tightly integrated with the
156 * page cache. To handle this we cache mmap'ed files twice, once in
157 * the ARC and a second time in the page cache. The code is careful
158 * to keep both copies synchronized.
160 * When a file with an mmap'ed region is written to using write(2)
161 * both the data in the ARC and existing pages in the page cache
162 * are updated. For a read(2) data will be read first from the page
163 * cache then the ARC if needed. Neither a write(2) or read(2) will
164 * will ever result in new pages being added to the page cache.
166 * New pages are added to the page cache only via .readpage() which
167 * is called when the vfs needs to read a page off disk to back the
168 * virtual memory region. These pages may be modified without
169 * notifying the ARC and will be written out periodically via
170 * .writepage(). This will occur due to either a sync or the usual
171 * page aging behavior. Note because a read(2) of a mmap'ed file
172 * will always check the page cache first even when the ARC is out
173 * of date correct data will still be returned.
175 * While this implementation ensures correct behavior it does have
176 * have some drawbacks. The most obvious of which is that it
177 * increases the required memory footprint when access mmap'ed
178 * files. It also adds additional complexity to the code keeping
179 * both caches synchronized.
181 * Longer term it may be possible to cleanly resolve this wart by
182 * mapping page cache pages directly on to the ARC buffers. The
183 * Linux address space operations are flexible enough to allow
184 * selection of which pages back a particular index. The trick
185 * would be working out the details of which subsystem is in
186 * charge, the ARC, the page cache, or both. It may also prove
187 * helpful to move the ARC buffers to a scatter-gather lists
188 * rather than a vmalloc'ed region.
191 zpl_mmap(struct file
*filp
, struct vm_area_struct
*vma
)
193 znode_t
*zp
= ITOZ(filp
->f_mapping
->host
);
196 error
= generic_file_mmap(filp
, vma
);
200 mutex_enter(&zp
->z_lock
);
202 mutex_exit(&zp
->z_lock
);
208 * Populate a page with data for the Linux page cache. This function is
209 * only used to support mmap(2). There will be an identical copy of the
210 * data in the ARC which is kept up to date via .write() and .writepage().
212 * Current this function relies on zpl_read_common() and the O_DIRECT
213 * flag to read in a page. This works but the more correct way is to
214 * update zfs_fillpage() to be Linux friendly and use that interface.
217 zpl_readpage(struct file
*filp
, struct page
*pp
)
226 ASSERT(PageLocked(pp
));
227 ip
= pp
->mapping
->host
;
228 off
= page_offset(pp
);
229 i_size
= i_size_read(ip
);
230 ASSERT3S(off
, <, i_size
);
232 cr
= (cred_t
*)get_current_cred();
233 len
= MIN(PAGE_CACHE_SIZE
, i_size
- off
);
237 /* O_DIRECT is passed to bypass the page cache and avoid deadlock. */
238 wrote
= zpl_read_common(ip
, pb
, len
, off
, UIO_SYSSPACE
, O_DIRECT
, cr
);
242 if (!error
&& (len
< PAGE_CACHE_SIZE
))
243 memset(pb
+ len
, 0, PAGE_CACHE_SIZE
- len
);
250 ClearPageUptodate(pp
);
254 flush_dcache_page(pp
);
263 * Write out dirty pages to the ARC, this function is only required to
264 * support mmap(2). Mapped pages may be dirtied by memory operations
265 * which never call .write(). These dirty pages are kept in sync with
266 * the ARC buffers via this hook.
268 * Currently this function relies on zpl_write_common() and the O_DIRECT
269 * flag to push out the page. This works but the more correct way is
270 * to update zfs_putapage() to be Linux friendly and use that interface.
273 zpl_writepage(struct page
*pp
, struct writeback_control
*wbc
)
282 ASSERT(PageLocked(pp
));
283 ip
= pp
->mapping
->host
;
284 off
= page_offset(pp
);
285 i_size
= i_size_read(ip
);
287 cr
= (cred_t
*)get_current_cred();
288 len
= MIN(PAGE_CACHE_SIZE
, i_size
- off
);
292 /* O_DIRECT is passed to bypass the page cache and avoid deadlock. */
293 read
= zpl_write_common(ip
, pb
, len
, off
, UIO_SYSSPACE
, O_DIRECT
, cr
);
302 ClearPageUptodate(pp
);
313 const struct address_space_operations zpl_address_space_operations
= {
314 .readpage
= zpl_readpage
,
315 .writepage
= zpl_writepage
,
318 const struct file_operations zpl_file_operations
= {
319 .open
= generic_file_open
,
320 .llseek
= generic_file_llseek
,
323 .readdir
= zpl_readdir
,
328 const struct file_operations zpl_dir_file_operations
= {
329 .llseek
= generic_file_llseek
,
330 .read
= generic_read_dir
,
331 .readdir
= zpl_readdir
,