]>
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 | ||
26 | #include <sys/zfs_vfsops.h> | |
27 | #include <sys/zfs_vnops.h> | |
28 | #include <sys/zfs_znode.h> | |
29 | #include <sys/zpl.h> | |
30 | ||
31 | ||
126400a1 BB |
32 | static int |
33 | zpl_open(struct inode *ip, struct file *filp) | |
34 | { | |
81e97e21 | 35 | cred_t *cr = CRED(); |
126400a1 BB |
36 | int error; |
37 | ||
81e97e21 | 38 | crhold(cr); |
126400a1 | 39 | error = -zfs_open(ip, filp->f_mode, filp->f_flags, cr); |
81e97e21 | 40 | crfree(cr); |
126400a1 BB |
41 | ASSERT3S(error, <=, 0); |
42 | ||
43 | if (error) | |
44 | return (error); | |
45 | ||
46 | return generic_file_open(ip, filp); | |
47 | } | |
48 | ||
49 | static int | |
50 | zpl_release(struct inode *ip, struct file *filp) | |
51 | { | |
81e97e21 | 52 | cred_t *cr = CRED(); |
126400a1 BB |
53 | int error; |
54 | ||
81e97e21 | 55 | crhold(cr); |
126400a1 | 56 | error = -zfs_close(ip, filp->f_flags, cr); |
81e97e21 | 57 | crfree(cr); |
126400a1 BB |
58 | ASSERT3S(error, <=, 0); |
59 | ||
60 | return (error); | |
61 | } | |
62 | ||
1efb473f BB |
63 | static int |
64 | zpl_readdir(struct file *filp, void *dirent, filldir_t filldir) | |
65 | { | |
66 | struct dentry *dentry = filp->f_path.dentry; | |
81e97e21 | 67 | cred_t *cr = CRED(); |
1efb473f BB |
68 | int error; |
69 | ||
81e97e21 | 70 | crhold(cr); |
1efb473f BB |
71 | error = -zfs_readdir(dentry->d_inode, dirent, filldir, |
72 | &filp->f_pos, cr); | |
81e97e21 | 73 | crfree(cr); |
1efb473f BB |
74 | ASSERT3S(error, <=, 0); |
75 | ||
76 | return (error); | |
77 | } | |
78 | ||
3117dd0b BB |
79 | /* |
80 | * 2.6.35 API change, | |
81 | * As of 2.6.35 the dentry argument to the .fsync() vfs hook was deemed | |
82 | * redundant. The dentry is still accessible via filp->f_path.dentry, | |
83 | * and we are guaranteed that filp will never be NULL. | |
84 | * | |
85 | * 2.6.34 API change, | |
86 | * Prior to 2.6.34 the nfsd kernel server would pass a NULL file struct * | |
87 | * to the .fsync() hook. For this reason, we must be careful not to use | |
88 | * filp unconditionally in the 3 argument case. | |
89 | */ | |
90 | #ifdef HAVE_2ARGS_FSYNC | |
91 | static int | |
92 | zpl_fsync(struct file *filp, int datasync) | |
93 | { | |
94 | struct dentry *dentry = filp->f_path.dentry; | |
95 | #else | |
96 | static int | |
97 | zpl_fsync(struct file *filp, struct dentry *dentry, int datasync) | |
1efb473f | 98 | { |
3117dd0b | 99 | #endif /* HAVE_2ARGS_FSYNC */ |
81e97e21 | 100 | cred_t *cr = CRED(); |
1efb473f BB |
101 | int error; |
102 | ||
81e97e21 | 103 | crhold(cr); |
3117dd0b | 104 | error = -zfs_fsync(dentry->d_inode, datasync, cr); |
81e97e21 | 105 | crfree(cr); |
1efb473f BB |
106 | ASSERT3S(error, <=, 0); |
107 | ||
108 | return (error); | |
109 | } | |
110 | ||
111 | ssize_t | |
112 | zpl_read_common(struct inode *ip, const char *buf, size_t len, loff_t pos, | |
113 | uio_seg_t segment, int flags, cred_t *cr) | |
114 | { | |
115 | int error; | |
116 | struct iovec iov; | |
117 | uio_t uio; | |
118 | ||
119 | iov.iov_base = (void *)buf; | |
120 | iov.iov_len = len; | |
121 | ||
122 | uio.uio_iov = &iov; | |
123 | uio.uio_resid = len; | |
124 | uio.uio_iovcnt = 1; | |
125 | uio.uio_loffset = pos; | |
126 | uio.uio_limit = MAXOFFSET_T; | |
127 | uio.uio_segflg = segment; | |
128 | ||
129 | error = -zfs_read(ip, &uio, flags, cr); | |
130 | if (error < 0) | |
131 | return (error); | |
132 | ||
133 | return (len - uio.uio_resid); | |
134 | } | |
135 | ||
136 | static ssize_t | |
137 | zpl_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos) | |
138 | { | |
81e97e21 | 139 | cred_t *cr = CRED(); |
1efb473f BB |
140 | ssize_t read; |
141 | ||
81e97e21 | 142 | crhold(cr); |
1efb473f BB |
143 | read = zpl_read_common(filp->f_mapping->host, buf, len, *ppos, |
144 | UIO_USERSPACE, filp->f_flags, cr); | |
81e97e21 | 145 | crfree(cr); |
1efb473f BB |
146 | |
147 | if (read < 0) | |
148 | return (read); | |
149 | ||
150 | *ppos += read; | |
151 | return (read); | |
152 | } | |
153 | ||
154 | ssize_t | |
155 | zpl_write_common(struct inode *ip, const char *buf, size_t len, loff_t pos, | |
156 | uio_seg_t segment, int flags, cred_t *cr) | |
157 | { | |
158 | int error; | |
159 | struct iovec iov; | |
160 | uio_t uio; | |
161 | ||
162 | iov.iov_base = (void *)buf; | |
163 | iov.iov_len = len; | |
164 | ||
165 | uio.uio_iov = &iov; | |
166 | uio.uio_resid = len, | |
167 | uio.uio_iovcnt = 1; | |
168 | uio.uio_loffset = pos; | |
169 | uio.uio_limit = MAXOFFSET_T; | |
170 | uio.uio_segflg = segment; | |
171 | ||
172 | error = -zfs_write(ip, &uio, flags, cr); | |
173 | if (error < 0) | |
174 | return (error); | |
175 | ||
176 | return (len - uio.uio_resid); | |
177 | } | |
178 | ||
179 | static ssize_t | |
180 | zpl_write(struct file *filp, const char __user *buf, size_t len, loff_t *ppos) | |
181 | { | |
81e97e21 | 182 | cred_t *cr = CRED(); |
1efb473f BB |
183 | ssize_t wrote; |
184 | ||
81e97e21 | 185 | crhold(cr); |
1efb473f BB |
186 | wrote = zpl_write_common(filp->f_mapping->host, buf, len, *ppos, |
187 | UIO_USERSPACE, filp->f_flags, cr); | |
81e97e21 | 188 | crfree(cr); |
1efb473f BB |
189 | |
190 | if (wrote < 0) | |
191 | return (wrote); | |
192 | ||
193 | *ppos += wrote; | |
194 | return (wrote); | |
195 | } | |
196 | ||
c0d35759 BB |
197 | /* |
198 | * It's worth taking a moment to describe how mmap is implemented | |
199 | * for zfs because it differs considerably from other Linux filesystems. | |
200 | * However, this issue is handled the same way under OpenSolaris. | |
201 | * | |
202 | * The issue is that by design zfs bypasses the Linux page cache and | |
203 | * leaves all caching up to the ARC. This has been shown to work | |
204 | * well for the common read(2)/write(2) case. However, mmap(2) | |
205 | * is problem because it relies on being tightly integrated with the | |
206 | * page cache. To handle this we cache mmap'ed files twice, once in | |
207 | * the ARC and a second time in the page cache. The code is careful | |
208 | * to keep both copies synchronized. | |
209 | * | |
210 | * When a file with an mmap'ed region is written to using write(2) | |
211 | * both the data in the ARC and existing pages in the page cache | |
212 | * are updated. For a read(2) data will be read first from the page | |
213 | * cache then the ARC if needed. Neither a write(2) or read(2) will | |
214 | * will ever result in new pages being added to the page cache. | |
215 | * | |
216 | * New pages are added to the page cache only via .readpage() which | |
217 | * is called when the vfs needs to read a page off disk to back the | |
218 | * virtual memory region. These pages may be modified without | |
219 | * notifying the ARC and will be written out periodically via | |
220 | * .writepage(). This will occur due to either a sync or the usual | |
221 | * page aging behavior. Note because a read(2) of a mmap'ed file | |
222 | * will always check the page cache first even when the ARC is out | |
223 | * of date correct data will still be returned. | |
224 | * | |
225 | * While this implementation ensures correct behavior it does have | |
226 | * have some drawbacks. The most obvious of which is that it | |
227 | * increases the required memory footprint when access mmap'ed | |
228 | * files. It also adds additional complexity to the code keeping | |
229 | * both caches synchronized. | |
230 | * | |
231 | * Longer term it may be possible to cleanly resolve this wart by | |
232 | * mapping page cache pages directly on to the ARC buffers. The | |
233 | * Linux address space operations are flexible enough to allow | |
234 | * selection of which pages back a particular index. The trick | |
235 | * would be working out the details of which subsystem is in | |
236 | * charge, the ARC, the page cache, or both. It may also prove | |
237 | * helpful to move the ARC buffers to a scatter-gather lists | |
238 | * rather than a vmalloc'ed region. | |
239 | */ | |
240 | static int | |
241 | zpl_mmap(struct file *filp, struct vm_area_struct *vma) | |
242 | { | |
e2e7aa2d BB |
243 | struct inode *ip = filp->f_mapping->host; |
244 | znode_t *zp = ITOZ(ip); | |
c0d35759 BB |
245 | int error; |
246 | ||
e2e7aa2d BB |
247 | error = -zfs_map(ip, vma->vm_pgoff, (caddr_t *)vma->vm_start, |
248 | (size_t)(vma->vm_end - vma->vm_start), vma->vm_flags); | |
249 | if (error) | |
250 | return (error); | |
251 | ||
c0d35759 BB |
252 | error = generic_file_mmap(filp, vma); |
253 | if (error) | |
254 | return (error); | |
255 | ||
256 | mutex_enter(&zp->z_lock); | |
257 | zp->z_is_mapped = 1; | |
258 | mutex_exit(&zp->z_lock); | |
259 | ||
260 | return (error); | |
261 | } | |
262 | ||
263 | /* | |
264 | * Populate a page with data for the Linux page cache. This function is | |
265 | * only used to support mmap(2). There will be an identical copy of the | |
266 | * data in the ARC which is kept up to date via .write() and .writepage(). | |
267 | * | |
268 | * Current this function relies on zpl_read_common() and the O_DIRECT | |
269 | * flag to read in a page. This works but the more correct way is to | |
270 | * update zfs_fillpage() to be Linux friendly and use that interface. | |
271 | */ | |
272 | static int | |
273 | zpl_readpage(struct file *filp, struct page *pp) | |
274 | { | |
275 | struct inode *ip; | |
dde471ef | 276 | struct page *pl[1]; |
c0d35759 BB |
277 | int error = 0; |
278 | ||
279 | ASSERT(PageLocked(pp)); | |
280 | ip = pp->mapping->host; | |
dde471ef | 281 | pl[0] = pp; |
c0d35759 | 282 | |
dde471ef | 283 | error = -zfs_getpage(ip, pl, 1); |
c0d35759 | 284 | |
dde471ef PJ |
285 | if (error) { |
286 | SetPageError(pp); | |
287 | ClearPageUptodate(pp); | |
288 | } else { | |
289 | ClearPageError(pp); | |
290 | SetPageUptodate(pp); | |
291 | flush_dcache_page(pp); | |
292 | } | |
c0d35759 | 293 | |
dde471ef PJ |
294 | unlock_page(pp); |
295 | return error; | |
296 | } | |
c0d35759 | 297 | |
f3ab88d6 BB |
298 | /* |
299 | * Populate a set of pages with data for the Linux page cache. This | |
300 | * function will only be called for read ahead and never for demand | |
301 | * paging. For simplicity, the code relies on read_cache_pages() to | |
302 | * correctly lock each page for IO and call zpl_readpage(). | |
303 | */ | |
304 | static int | |
305 | zpl_readpages(struct file *filp, struct address_space *mapping, | |
306 | struct list_head *pages, unsigned nr_pages) | |
307 | { | |
95d9fd02 BB |
308 | return (read_cache_pages(mapping, pages, |
309 | (filler_t *)zpl_readpage, filp)); | |
f3ab88d6 BB |
310 | } |
311 | ||
dde471ef PJ |
312 | int |
313 | zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data) | |
314 | { | |
3c0e5c0f BB |
315 | struct address_space *mapping = data; |
316 | ||
317 | ASSERT(PageLocked(pp)); | |
318 | ASSERT(!PageWriteback(pp)); | |
c0d35759 | 319 | |
cfc9a5c8 BB |
320 | /* |
321 | * Disable the normal reclaim path for zpl_putpage(). This | |
322 | * ensures that all memory allocations under this call path | |
323 | * will never enter direct reclaim. If this were to happen | |
324 | * the VM might try to write out additional pages by calling | |
325 | * zpl_putpage() again resulting in a deadlock. | |
326 | */ | |
6a95d0b7 BB |
327 | if (current->flags & PF_MEMALLOC) { |
328 | (void) zfs_putpage(mapping->host, pp, wbc); | |
329 | } else { | |
330 | current->flags |= PF_MEMALLOC; | |
331 | (void) zfs_putpage(mapping->host, pp, wbc); | |
332 | current->flags &= ~PF_MEMALLOC; | |
333 | } | |
c0d35759 | 334 | |
3c0e5c0f | 335 | return (0); |
dde471ef | 336 | } |
c0d35759 | 337 | |
dde471ef PJ |
338 | static int |
339 | zpl_writepages(struct address_space *mapping, struct writeback_control *wbc) | |
340 | { | |
341 | return write_cache_pages(mapping, wbc, zpl_putpage, mapping); | |
c0d35759 BB |
342 | } |
343 | ||
344 | /* | |
345 | * Write out dirty pages to the ARC, this function is only required to | |
346 | * support mmap(2). Mapped pages may be dirtied by memory operations | |
347 | * which never call .write(). These dirty pages are kept in sync with | |
348 | * the ARC buffers via this hook. | |
c0d35759 BB |
349 | */ |
350 | static int | |
351 | zpl_writepage(struct page *pp, struct writeback_control *wbc) | |
352 | { | |
dde471ef | 353 | return zpl_putpage(pp, wbc, pp->mapping); |
c0d35759 BB |
354 | } |
355 | ||
1efb473f | 356 | const struct address_space_operations zpl_address_space_operations = { |
dde471ef | 357 | .readpages = zpl_readpages, |
1efb473f BB |
358 | .readpage = zpl_readpage, |
359 | .writepage = zpl_writepage, | |
dde471ef | 360 | .writepages = zpl_writepages, |
1efb473f BB |
361 | }; |
362 | ||
363 | const struct file_operations zpl_file_operations = { | |
126400a1 BB |
364 | .open = zpl_open, |
365 | .release = zpl_release, | |
1efb473f | 366 | .llseek = generic_file_llseek, |
c0d35759 BB |
367 | .read = zpl_read, |
368 | .write = zpl_write, | |
1efb473f | 369 | .readdir = zpl_readdir, |
c0d35759 | 370 | .mmap = zpl_mmap, |
1efb473f | 371 | .fsync = zpl_fsync, |
1efb473f BB |
372 | }; |
373 | ||
374 | const struct file_operations zpl_dir_file_operations = { | |
375 | .llseek = generic_file_llseek, | |
376 | .read = generic_read_dir, | |
377 | .readdir = zpl_readdir, | |
378 | .fsync = zpl_fsync, | |
379 | }; |