]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
a805bad5 | 2 | * Copyright (c) 2000-2006 Silicon Graphics, Inc. |
7b718769 | 3 | * All Rights Reserved. |
1da177e4 | 4 | * |
7b718769 NS |
5 | * This program is free software; you can redistribute it and/or |
6 | * modify it under the terms of the GNU General Public License as | |
1da177e4 LT |
7 | * published by the Free Software Foundation. |
8 | * | |
7b718769 NS |
9 | * This program is distributed in the hope that it would be useful, |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
1da177e4 | 13 | * |
7b718769 NS |
14 | * You should have received a copy of the GNU General Public License |
15 | * along with this program; if not, write the Free Software Foundation, | |
16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
1da177e4 | 17 | */ |
1da177e4 | 18 | #include "xfs.h" |
a844f451 | 19 | #include "xfs_bit.h" |
1da177e4 LT |
20 | #include "xfs_log.h" |
21 | #include "xfs_clnt.h" | |
a844f451 | 22 | #include "xfs_inum.h" |
1da177e4 LT |
23 | #include "xfs_trans.h" |
24 | #include "xfs_sb.h" | |
a844f451 | 25 | #include "xfs_ag.h" |
1da177e4 LT |
26 | #include "xfs_dir2.h" |
27 | #include "xfs_alloc.h" | |
28 | #include "xfs_dmapi.h" | |
29 | #include "xfs_quota.h" | |
30 | #include "xfs_mount.h" | |
1da177e4 | 31 | #include "xfs_bmap_btree.h" |
a844f451 | 32 | #include "xfs_alloc_btree.h" |
1da177e4 | 33 | #include "xfs_ialloc_btree.h" |
1da177e4 | 34 | #include "xfs_dir2_sf.h" |
a844f451 | 35 | #include "xfs_attr_sf.h" |
1da177e4 LT |
36 | #include "xfs_dinode.h" |
37 | #include "xfs_inode.h" | |
a844f451 NS |
38 | #include "xfs_btree.h" |
39 | #include "xfs_ialloc.h" | |
1da177e4 | 40 | #include "xfs_bmap.h" |
1da177e4 LT |
41 | #include "xfs_rtalloc.h" |
42 | #include "xfs_error.h" | |
43 | #include "xfs_itable.h" | |
44 | #include "xfs_rw.h" | |
45 | #include "xfs_acl.h" | |
1da177e4 LT |
46 | #include "xfs_attr.h" |
47 | #include "xfs_buf_item.h" | |
48 | #include "xfs_utils.h" | |
739bfb2a | 49 | #include "xfs_vnodeops.h" |
745f6919 | 50 | #include "xfs_vfsops.h" |
1da177e4 | 51 | #include "xfs_version.h" |
1da177e4 LT |
52 | |
53 | #include <linux/namei.h> | |
54 | #include <linux/init.h> | |
55 | #include <linux/mount.h> | |
0829c360 | 56 | #include <linux/mempool.h> |
1da177e4 | 57 | #include <linux/writeback.h> |
4df08c52 | 58 | #include <linux/kthread.h> |
7dfb7103 | 59 | #include <linux/freezer.h> |
1da177e4 | 60 | |
7989cb8e DC |
61 | static struct quotactl_ops xfs_quotactl_operations; |
62 | static struct super_operations xfs_super_operations; | |
63 | static kmem_zone_t *xfs_vnode_zone; | |
64 | static kmem_zone_t *xfs_ioend_zone; | |
0829c360 | 65 | mempool_t *xfs_ioend_pool; |
1da177e4 LT |
66 | |
67 | STATIC struct xfs_mount_args * | |
68 | xfs_args_allocate( | |
764d1f89 NS |
69 | struct super_block *sb, |
70 | int silent) | |
1da177e4 LT |
71 | { |
72 | struct xfs_mount_args *args; | |
73 | ||
74 | args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP); | |
75 | args->logbufs = args->logbufsize = -1; | |
76 | strncpy(args->fsname, sb->s_id, MAXNAMELEN); | |
77 | ||
78 | /* Copy the already-parsed mount(2) flags we're interested in */ | |
1da177e4 LT |
79 | if (sb->s_flags & MS_DIRSYNC) |
80 | args->flags |= XFSMNT_DIRSYNC; | |
81 | if (sb->s_flags & MS_SYNCHRONOUS) | |
82 | args->flags |= XFSMNT_WSYNC; | |
764d1f89 NS |
83 | if (silent) |
84 | args->flags |= XFSMNT_QUIET; | |
1da177e4 LT |
85 | args->flags |= XFSMNT_32BITINODES; |
86 | ||
87 | return args; | |
88 | } | |
89 | ||
90 | __uint64_t | |
91 | xfs_max_file_offset( | |
92 | unsigned int blockshift) | |
93 | { | |
94 | unsigned int pagefactor = 1; | |
95 | unsigned int bitshift = BITS_PER_LONG - 1; | |
96 | ||
97 | /* Figure out maximum filesize, on Linux this can depend on | |
98 | * the filesystem blocksize (on 32 bit platforms). | |
99 | * __block_prepare_write does this in an [unsigned] long... | |
100 | * page->index << (PAGE_CACHE_SHIFT - bbits) | |
101 | * So, for page sized blocks (4K on 32 bit platforms), | |
102 | * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is | |
103 | * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1) | |
104 | * but for smaller blocksizes it is less (bbits = log2 bsize). | |
105 | * Note1: get_block_t takes a long (implicit cast from above) | |
106 | * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch | |
107 | * can optionally convert the [unsigned] long from above into | |
108 | * an [unsigned] long long. | |
109 | */ | |
110 | ||
111 | #if BITS_PER_LONG == 32 | |
112 | # if defined(CONFIG_LBD) | |
113 | ASSERT(sizeof(sector_t) == 8); | |
114 | pagefactor = PAGE_CACHE_SIZE; | |
115 | bitshift = BITS_PER_LONG; | |
116 | # else | |
117 | pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift); | |
118 | # endif | |
119 | #endif | |
120 | ||
121 | return (((__uint64_t)pagefactor) << bitshift) - 1; | |
122 | } | |
123 | ||
7989cb8e | 124 | STATIC_INLINE void |
1da177e4 LT |
125 | xfs_set_inodeops( |
126 | struct inode *inode) | |
127 | { | |
0432dab2 CH |
128 | switch (inode->i_mode & S_IFMT) { |
129 | case S_IFREG: | |
416c6d5b | 130 | inode->i_op = &xfs_inode_operations; |
3562fd45 | 131 | inode->i_fop = &xfs_file_operations; |
e4c573bb | 132 | inode->i_mapping->a_ops = &xfs_address_space_operations; |
0432dab2 CH |
133 | break; |
134 | case S_IFDIR: | |
416c6d5b | 135 | inode->i_op = &xfs_dir_inode_operations; |
3562fd45 | 136 | inode->i_fop = &xfs_dir_file_operations; |
0432dab2 CH |
137 | break; |
138 | case S_IFLNK: | |
416c6d5b | 139 | inode->i_op = &xfs_symlink_inode_operations; |
1da177e4 | 140 | if (inode->i_blocks) |
e4c573bb | 141 | inode->i_mapping->a_ops = &xfs_address_space_operations; |
0432dab2 CH |
142 | break; |
143 | default: | |
416c6d5b | 144 | inode->i_op = &xfs_inode_operations; |
1da177e4 | 145 | init_special_inode(inode, inode->i_mode, inode->i_rdev); |
0432dab2 | 146 | break; |
1da177e4 LT |
147 | } |
148 | } | |
149 | ||
7989cb8e | 150 | STATIC_INLINE void |
1da177e4 LT |
151 | xfs_revalidate_inode( |
152 | xfs_mount_t *mp, | |
67fcaa73 | 153 | bhv_vnode_t *vp, |
1da177e4 LT |
154 | xfs_inode_t *ip) |
155 | { | |
ec86dc02 | 156 | struct inode *inode = vn_to_inode(vp); |
1da177e4 | 157 | |
0432dab2 | 158 | inode->i_mode = ip->i_d.di_mode; |
1da177e4 LT |
159 | inode->i_nlink = ip->i_d.di_nlink; |
160 | inode->i_uid = ip->i_d.di_uid; | |
161 | inode->i_gid = ip->i_d.di_gid; | |
0432dab2 CH |
162 | |
163 | switch (inode->i_mode & S_IFMT) { | |
164 | case S_IFBLK: | |
165 | case S_IFCHR: | |
166 | inode->i_rdev = | |
167 | MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff, | |
168 | sysv_minor(ip->i_df.if_u2.if_rdev)); | |
169 | break; | |
170 | default: | |
1da177e4 | 171 | inode->i_rdev = 0; |
0432dab2 | 172 | break; |
1da177e4 | 173 | } |
0432dab2 | 174 | |
1da177e4 LT |
175 | inode->i_generation = ip->i_d.di_gen; |
176 | i_size_write(inode, ip->i_d.di_size); | |
177 | inode->i_blocks = | |
178 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
179 | inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec; | |
180 | inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec; | |
181 | inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec; | |
182 | inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec; | |
183 | inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec; | |
184 | inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec; | |
185 | if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE) | |
186 | inode->i_flags |= S_IMMUTABLE; | |
187 | else | |
188 | inode->i_flags &= ~S_IMMUTABLE; | |
189 | if (ip->i_d.di_flags & XFS_DIFLAG_APPEND) | |
190 | inode->i_flags |= S_APPEND; | |
191 | else | |
192 | inode->i_flags &= ~S_APPEND; | |
193 | if (ip->i_d.di_flags & XFS_DIFLAG_SYNC) | |
194 | inode->i_flags |= S_SYNC; | |
195 | else | |
196 | inode->i_flags &= ~S_SYNC; | |
197 | if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME) | |
198 | inode->i_flags |= S_NOATIME; | |
199 | else | |
200 | inode->i_flags &= ~S_NOATIME; | |
b3aea4ed | 201 | xfs_iflags_clear(ip, XFS_IMODIFIED); |
1da177e4 LT |
202 | } |
203 | ||
204 | void | |
205 | xfs_initialize_vnode( | |
48c872a9 | 206 | struct xfs_mount *mp, |
67fcaa73 | 207 | bhv_vnode_t *vp, |
745f6919 | 208 | struct xfs_inode *ip) |
1da177e4 | 209 | { |
ec86dc02 | 210 | struct inode *inode = vn_to_inode(vp); |
1da177e4 | 211 | |
739bfb2a | 212 | if (!ip->i_vnode) { |
739bfb2a CH |
213 | ip->i_vnode = vp; |
214 | inode->i_private = ip; | |
1da177e4 LT |
215 | } |
216 | ||
217 | /* | |
218 | * We need to set the ops vectors, and unlock the inode, but if | |
219 | * we have been called during the new inode create process, it is | |
220 | * too early to fill in the Linux inode. We will get called a | |
221 | * second time once the inode is properly set up, and then we can | |
222 | * finish our work. | |
223 | */ | |
745f6919 | 224 | if (ip->i_d.di_mode != 0 && (inode->i_state & I_NEW)) { |
48c872a9 | 225 | xfs_revalidate_inode(mp, vp, ip); |
1da177e4 | 226 | xfs_set_inodeops(inode); |
ec86dc02 | 227 | |
7a18c386 | 228 | xfs_iflags_clear(ip, XFS_INEW); |
1da177e4 LT |
229 | barrier(); |
230 | ||
231 | unlock_new_inode(inode); | |
232 | } | |
233 | } | |
234 | ||
235 | int | |
236 | xfs_blkdev_get( | |
237 | xfs_mount_t *mp, | |
238 | const char *name, | |
239 | struct block_device **bdevp) | |
240 | { | |
241 | int error = 0; | |
242 | ||
243 | *bdevp = open_bdev_excl(name, 0, mp); | |
244 | if (IS_ERR(*bdevp)) { | |
245 | error = PTR_ERR(*bdevp); | |
246 | printk("XFS: Invalid device [%s], error=%d\n", name, error); | |
247 | } | |
248 | ||
249 | return -error; | |
250 | } | |
251 | ||
252 | void | |
253 | xfs_blkdev_put( | |
254 | struct block_device *bdev) | |
255 | { | |
256 | if (bdev) | |
257 | close_bdev_excl(bdev); | |
258 | } | |
259 | ||
f538d4da CH |
260 | /* |
261 | * Try to write out the superblock using barriers. | |
262 | */ | |
263 | STATIC int | |
264 | xfs_barrier_test( | |
265 | xfs_mount_t *mp) | |
266 | { | |
267 | xfs_buf_t *sbp = xfs_getsb(mp, 0); | |
268 | int error; | |
269 | ||
270 | XFS_BUF_UNDONE(sbp); | |
271 | XFS_BUF_UNREAD(sbp); | |
272 | XFS_BUF_UNDELAYWRITE(sbp); | |
273 | XFS_BUF_WRITE(sbp); | |
274 | XFS_BUF_UNASYNC(sbp); | |
275 | XFS_BUF_ORDERED(sbp); | |
276 | ||
277 | xfsbdstrat(mp, sbp); | |
278 | error = xfs_iowait(sbp); | |
279 | ||
280 | /* | |
281 | * Clear all the flags we set and possible error state in the | |
282 | * buffer. We only did the write to try out whether barriers | |
283 | * worked and shouldn't leave any traces in the superblock | |
284 | * buffer. | |
285 | */ | |
286 | XFS_BUF_DONE(sbp); | |
287 | XFS_BUF_ERROR(sbp, 0); | |
288 | XFS_BUF_UNORDERED(sbp); | |
289 | ||
290 | xfs_buf_relse(sbp); | |
291 | return error; | |
292 | } | |
293 | ||
294 | void | |
295 | xfs_mountfs_check_barriers(xfs_mount_t *mp) | |
296 | { | |
297 | int error; | |
298 | ||
299 | if (mp->m_logdev_targp != mp->m_ddev_targp) { | |
300 | xfs_fs_cmn_err(CE_NOTE, mp, | |
301 | "Disabling barriers, not supported with external log device"); | |
302 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 303 | return; |
f538d4da CH |
304 | } |
305 | ||
b2ea401b NS |
306 | if (xfs_readonly_buftarg(mp->m_ddev_targp)) { |
307 | xfs_fs_cmn_err(CE_NOTE, mp, | |
308 | "Disabling barriers, underlying device is readonly"); | |
309 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
310 | return; | |
311 | } | |
312 | ||
f538d4da CH |
313 | error = xfs_barrier_test(mp); |
314 | if (error) { | |
315 | xfs_fs_cmn_err(CE_NOTE, mp, | |
316 | "Disabling barriers, trial barrier write failed"); | |
317 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 318 | return; |
f538d4da CH |
319 | } |
320 | } | |
321 | ||
322 | void | |
323 | xfs_blkdev_issue_flush( | |
324 | xfs_buftarg_t *buftarg) | |
325 | { | |
ce8e922c | 326 | blkdev_issue_flush(buftarg->bt_bdev, NULL); |
f538d4da | 327 | } |
1da177e4 LT |
328 | |
329 | STATIC struct inode * | |
a50cd269 | 330 | xfs_fs_alloc_inode( |
1da177e4 LT |
331 | struct super_block *sb) |
332 | { | |
67fcaa73 | 333 | bhv_vnode_t *vp; |
1da177e4 | 334 | |
8758280f NS |
335 | vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP); |
336 | if (unlikely(!vp)) | |
1da177e4 | 337 | return NULL; |
ec86dc02 | 338 | return vn_to_inode(vp); |
1da177e4 LT |
339 | } |
340 | ||
341 | STATIC void | |
a50cd269 | 342 | xfs_fs_destroy_inode( |
1da177e4 LT |
343 | struct inode *inode) |
344 | { | |
ec86dc02 | 345 | kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode)); |
1da177e4 LT |
346 | } |
347 | ||
348 | STATIC void | |
a50cd269 | 349 | xfs_fs_inode_init_once( |
8758280f NS |
350 | void *vnode, |
351 | kmem_zone_t *zonep, | |
1da177e4 LT |
352 | unsigned long flags) |
353 | { | |
a35afb83 | 354 | inode_init_once(vn_to_inode((bhv_vnode_t *)vnode)); |
1da177e4 LT |
355 | } |
356 | ||
357 | STATIC int | |
8758280f | 358 | xfs_init_zones(void) |
1da177e4 | 359 | { |
67fcaa73 | 360 | xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode", |
e0cc2325 NS |
361 | KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | |
362 | KM_ZONE_SPREAD, | |
a50cd269 | 363 | xfs_fs_inode_init_once); |
0829c360 CH |
364 | if (!xfs_vnode_zone) |
365 | goto out; | |
366 | ||
367 | xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend"); | |
368 | if (!xfs_ioend_zone) | |
369 | goto out_destroy_vnode_zone; | |
370 | ||
93d2341c MD |
371 | xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE, |
372 | xfs_ioend_zone); | |
0829c360 CH |
373 | if (!xfs_ioend_pool) |
374 | goto out_free_ioend_zone; | |
1da177e4 | 375 | return 0; |
0829c360 | 376 | |
0829c360 CH |
377 | out_free_ioend_zone: |
378 | kmem_zone_destroy(xfs_ioend_zone); | |
379 | out_destroy_vnode_zone: | |
380 | kmem_zone_destroy(xfs_vnode_zone); | |
381 | out: | |
382 | return -ENOMEM; | |
1da177e4 LT |
383 | } |
384 | ||
385 | STATIC void | |
8758280f | 386 | xfs_destroy_zones(void) |
1da177e4 | 387 | { |
0829c360 CH |
388 | mempool_destroy(xfs_ioend_pool); |
389 | kmem_zone_destroy(xfs_vnode_zone); | |
390 | kmem_zone_destroy(xfs_ioend_zone); | |
1da177e4 LT |
391 | } |
392 | ||
393 | /* | |
394 | * Attempt to flush the inode, this will actually fail | |
395 | * if the inode is pinned, but we dirty the inode again | |
396 | * at the point when it is unpinned after a log write, | |
8758280f | 397 | * since this is when the inode itself becomes flushable. |
1da177e4 LT |
398 | */ |
399 | STATIC int | |
a50cd269 | 400 | xfs_fs_write_inode( |
1da177e4 LT |
401 | struct inode *inode, |
402 | int sync) | |
403 | { | |
1da177e4 LT |
404 | int error = 0, flags = FLUSH_INODE; |
405 | ||
1543d79c | 406 | vn_trace_entry(XFS_I(inode), __FUNCTION__, |
739bfb2a CH |
407 | (inst_t *)__return_address); |
408 | if (sync) { | |
409 | filemap_fdatawait(inode->i_mapping); | |
410 | flags |= FLUSH_SYNC; | |
411 | } | |
412 | error = xfs_inode_flush(XFS_I(inode), flags); | |
413 | if (error == EAGAIN) { | |
414 | if (sync) | |
415 | error = xfs_inode_flush(XFS_I(inode), | |
416 | flags | FLUSH_LOG); | |
417 | else | |
418 | error = 0; | |
1da177e4 | 419 | } |
739bfb2a | 420 | |
1da177e4 LT |
421 | return -error; |
422 | } | |
423 | ||
424 | STATIC void | |
a50cd269 | 425 | xfs_fs_clear_inode( |
1da177e4 LT |
426 | struct inode *inode) |
427 | { | |
1543d79c | 428 | xfs_inode_t *ip = XFS_I(inode); |
56d433e4 | 429 | |
02ba71de | 430 | /* |
1543d79c | 431 | * ip can be null when xfs_iget_core calls xfs_idestroy if we |
02ba71de CH |
432 | * find an inode with di_mode == 0 but without IGET_CREATE set. |
433 | */ | |
1543d79c CH |
434 | if (ip) { |
435 | vn_trace_entry(ip, __FUNCTION__, (inst_t *)__return_address); | |
436 | ||
437 | XFS_STATS_INC(vn_rele); | |
438 | XFS_STATS_INC(vn_remove); | |
439 | XFS_STATS_INC(vn_reclaim); | |
440 | XFS_STATS_DEC(vn_active); | |
441 | ||
442 | xfs_inactive(ip); | |
443 | xfs_iflags_clear(ip, XFS_IMODIFIED); | |
444 | if (xfs_reclaim(ip)) | |
445 | panic("%s: cannot reclaim 0x%p\n", __FUNCTION__, inode); | |
b3aea4ed | 446 | } |
56d433e4 | 447 | |
739bfb2a | 448 | ASSERT(XFS_I(inode) == NULL); |
56d433e4 | 449 | } |
1da177e4 LT |
450 | |
451 | /* | |
452 | * Enqueue a work item to be picked up by the vfs xfssyncd thread. | |
453 | * Doing this has two advantages: | |
454 | * - It saves on stack space, which is tight in certain situations | |
455 | * - It can be used (with care) as a mechanism to avoid deadlocks. | |
456 | * Flushing while allocating in a full filesystem requires both. | |
457 | */ | |
458 | STATIC void | |
459 | xfs_syncd_queue_work( | |
74394496 | 460 | struct xfs_mount *mp, |
1da177e4 | 461 | void *data, |
74394496 | 462 | void (*syncer)(struct xfs_mount *, void *)) |
1da177e4 | 463 | { |
b83bd138 | 464 | struct bhv_vfs_sync_work *work; |
1da177e4 | 465 | |
b83bd138 | 466 | work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP); |
1da177e4 LT |
467 | INIT_LIST_HEAD(&work->w_list); |
468 | work->w_syncer = syncer; | |
469 | work->w_data = data; | |
74394496 CH |
470 | work->w_mount = mp; |
471 | spin_lock(&mp->m_sync_lock); | |
472 | list_add_tail(&work->w_list, &mp->m_sync_list); | |
473 | spin_unlock(&mp->m_sync_lock); | |
474 | wake_up_process(mp->m_sync_task); | |
1da177e4 LT |
475 | } |
476 | ||
477 | /* | |
478 | * Flush delayed allocate data, attempting to free up reserved space | |
479 | * from existing allocations. At this point a new allocation attempt | |
480 | * has failed with ENOSPC and we are in the process of scratching our | |
481 | * heads, looking about for more room... | |
482 | */ | |
483 | STATIC void | |
484 | xfs_flush_inode_work( | |
74394496 CH |
485 | struct xfs_mount *mp, |
486 | void *arg) | |
1da177e4 | 487 | { |
74394496 CH |
488 | struct inode *inode = arg; |
489 | filemap_flush(inode->i_mapping); | |
490 | iput(inode); | |
1da177e4 LT |
491 | } |
492 | ||
493 | void | |
494 | xfs_flush_inode( | |
495 | xfs_inode_t *ip) | |
496 | { | |
74394496 | 497 | struct inode *inode = ip->i_vnode; |
1da177e4 LT |
498 | |
499 | igrab(inode); | |
74394496 | 500 | xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work); |
041e0e3b | 501 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
502 | } |
503 | ||
504 | /* | |
505 | * This is the "bigger hammer" version of xfs_flush_inode_work... | |
506 | * (IOW, "If at first you don't succeed, use a Bigger Hammer"). | |
507 | */ | |
508 | STATIC void | |
509 | xfs_flush_device_work( | |
74394496 CH |
510 | struct xfs_mount *mp, |
511 | void *arg) | |
1da177e4 | 512 | { |
74394496 CH |
513 | struct inode *inode = arg; |
514 | sync_blockdev(mp->m_vfsp->vfs_super->s_bdev); | |
515 | iput(inode); | |
1da177e4 LT |
516 | } |
517 | ||
518 | void | |
519 | xfs_flush_device( | |
520 | xfs_inode_t *ip) | |
521 | { | |
ec86dc02 | 522 | struct inode *inode = vn_to_inode(XFS_ITOV(ip)); |
1da177e4 LT |
523 | |
524 | igrab(inode); | |
74394496 | 525 | xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work); |
041e0e3b | 526 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
527 | xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC); |
528 | } | |
529 | ||
1da177e4 | 530 | STATIC void |
74394496 CH |
531 | xfs_sync_worker( |
532 | struct xfs_mount *mp, | |
1da177e4 LT |
533 | void *unused) |
534 | { | |
535 | int error; | |
536 | ||
74394496 CH |
537 | if (!(mp->m_flags & XFS_MOUNT_RDONLY)) |
538 | error = xfs_sync(mp, SYNC_FSDATA | SYNC_BDFLUSH | SYNC_ATTR | | |
539 | SYNC_REFCACHE | SYNC_SUPER); | |
540 | mp->m_sync_seq++; | |
541 | wake_up(&mp->m_wait_single_sync_task); | |
1da177e4 LT |
542 | } |
543 | ||
544 | STATIC int | |
545 | xfssyncd( | |
546 | void *arg) | |
547 | { | |
74394496 | 548 | struct xfs_mount *mp = arg; |
1da177e4 | 549 | long timeleft; |
b83bd138 | 550 | bhv_vfs_sync_work_t *work, *n; |
4df08c52 | 551 | LIST_HEAD (tmp); |
1da177e4 | 552 | |
83144186 | 553 | set_freezable(); |
041e0e3b | 554 | timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10); |
1da177e4 | 555 | for (;;) { |
041e0e3b | 556 | timeleft = schedule_timeout_interruptible(timeleft); |
1da177e4 | 557 | /* swsusp */ |
3e1d1d28 | 558 | try_to_freeze(); |
74394496 | 559 | if (kthread_should_stop() && list_empty(&mp->m_sync_list)) |
1da177e4 LT |
560 | break; |
561 | ||
74394496 | 562 | spin_lock(&mp->m_sync_lock); |
1da177e4 LT |
563 | /* |
564 | * We can get woken by laptop mode, to do a sync - | |
565 | * that's the (only!) case where the list would be | |
566 | * empty with time remaining. | |
567 | */ | |
74394496 | 568 | if (!timeleft || list_empty(&mp->m_sync_list)) { |
1da177e4 | 569 | if (!timeleft) |
041e0e3b NA |
570 | timeleft = xfs_syncd_centisecs * |
571 | msecs_to_jiffies(10); | |
74394496 CH |
572 | INIT_LIST_HEAD(&mp->m_sync_work.w_list); |
573 | list_add_tail(&mp->m_sync_work.w_list, | |
574 | &mp->m_sync_list); | |
1da177e4 | 575 | } |
74394496 | 576 | list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list) |
1da177e4 | 577 | list_move(&work->w_list, &tmp); |
74394496 | 578 | spin_unlock(&mp->m_sync_lock); |
1da177e4 LT |
579 | |
580 | list_for_each_entry_safe(work, n, &tmp, w_list) { | |
74394496 | 581 | (*work->w_syncer)(mp, work->w_data); |
1da177e4 | 582 | list_del(&work->w_list); |
74394496 | 583 | if (work == &mp->m_sync_work) |
1da177e4 | 584 | continue; |
b83bd138 | 585 | kmem_free(work, sizeof(struct bhv_vfs_sync_work)); |
1da177e4 LT |
586 | } |
587 | } | |
588 | ||
1da177e4 LT |
589 | return 0; |
590 | } | |
591 | ||
1da177e4 | 592 | STATIC void |
a50cd269 | 593 | xfs_fs_put_super( |
1da177e4 LT |
594 | struct super_block *sb) |
595 | { | |
b83bd138 | 596 | bhv_vfs_t *vfsp = vfs_from_sb(sb); |
745f6919 | 597 | struct xfs_mount *mp = XFS_M(sb); |
1da177e4 LT |
598 | int error; |
599 | ||
74394496 CH |
600 | kthread_stop(mp->m_sync_task); |
601 | ||
745f6919 CH |
602 | xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI); |
603 | error = xfs_unmount(mp, 0, NULL); | |
1da177e4 | 604 | if (error) { |
b83bd138 NS |
605 | printk("XFS: unmount got error=%d\n", error); |
606 | printk("%s: vfs=0x%p left dangling!\n", __FUNCTION__, vfsp); | |
607 | } else { | |
608 | vfs_deallocate(vfsp); | |
1da177e4 | 609 | } |
1da177e4 LT |
610 | } |
611 | ||
612 | STATIC void | |
a50cd269 | 613 | xfs_fs_write_super( |
1da177e4 LT |
614 | struct super_block *sb) |
615 | { | |
b83bd138 | 616 | if (!(sb->s_flags & MS_RDONLY)) |
745f6919 | 617 | xfs_sync(XFS_M(sb), SYNC_FSDATA); |
1da177e4 LT |
618 | sb->s_dirt = 0; |
619 | } | |
620 | ||
621 | STATIC int | |
a50cd269 | 622 | xfs_fs_sync_super( |
1da177e4 LT |
623 | struct super_block *sb, |
624 | int wait) | |
625 | { | |
745f6919 | 626 | struct xfs_mount *mp = XFS_M(sb); |
b83bd138 NS |
627 | int error; |
628 | int flags; | |
1da177e4 | 629 | |
2823945f DC |
630 | if (unlikely(sb->s_frozen == SB_FREEZE_WRITE)) { |
631 | /* | |
632 | * First stage of freeze - no more writers will make progress | |
633 | * now we are here, so we flush delwri and delalloc buffers | |
634 | * here, then wait for all I/O to complete. Data is frozen at | |
635 | * that point. Metadata is not frozen, transactions can still | |
636 | * occur here so don't bother flushing the buftarg (i.e | |
637 | * SYNC_QUIESCE) because it'll just get dirty again. | |
638 | */ | |
516b2e7c | 639 | flags = SYNC_DATA_QUIESCE; |
2823945f | 640 | } else |
f898d6c0 | 641 | flags = SYNC_FSDATA | (wait ? SYNC_WAIT : 0); |
1da177e4 | 642 | |
745f6919 | 643 | error = xfs_sync(mp, flags); |
1da177e4 LT |
644 | sb->s_dirt = 0; |
645 | ||
646 | if (unlikely(laptop_mode)) { | |
74394496 | 647 | int prev_sync_seq = mp->m_sync_seq; |
1da177e4 LT |
648 | |
649 | /* | |
650 | * The disk must be active because we're syncing. | |
651 | * We schedule xfssyncd now (now that the disk is | |
652 | * active) instead of later (when it might not be). | |
653 | */ | |
74394496 | 654 | wake_up_process(mp->m_sync_task); |
1da177e4 LT |
655 | /* |
656 | * We have to wait for the sync iteration to complete. | |
657 | * If we don't, the disk activity caused by the sync | |
658 | * will come after the sync is completed, and that | |
659 | * triggers another sync from laptop mode. | |
660 | */ | |
74394496 CH |
661 | wait_event(mp->m_wait_single_sync_task, |
662 | mp->m_sync_seq != prev_sync_seq); | |
1da177e4 LT |
663 | } |
664 | ||
665 | return -error; | |
666 | } | |
667 | ||
668 | STATIC int | |
a50cd269 | 669 | xfs_fs_statfs( |
726c3342 | 670 | struct dentry *dentry, |
1da177e4 LT |
671 | struct kstatfs *statp) |
672 | { | |
745f6919 | 673 | return -xfs_statvfs(XFS_M(dentry->d_sb), statp, |
d6938d1b | 674 | vn_from_inode(dentry->d_inode)); |
1da177e4 LT |
675 | } |
676 | ||
677 | STATIC int | |
a50cd269 | 678 | xfs_fs_remount( |
1da177e4 LT |
679 | struct super_block *sb, |
680 | int *flags, | |
681 | char *options) | |
682 | { | |
745f6919 | 683 | struct xfs_mount *mp = XFS_M(sb); |
764d1f89 | 684 | struct xfs_mount_args *args = xfs_args_allocate(sb, 0); |
1da177e4 LT |
685 | int error; |
686 | ||
745f6919 | 687 | error = xfs_parseargs(mp, options, args, 1); |
1da177e4 | 688 | if (!error) |
745f6919 | 689 | error = xfs_mntupdate(mp, flags, args); |
1da177e4 LT |
690 | kmem_free(args, sizeof(*args)); |
691 | return -error; | |
692 | } | |
693 | ||
694 | STATIC void | |
a50cd269 | 695 | xfs_fs_lockfs( |
1da177e4 LT |
696 | struct super_block *sb) |
697 | { | |
745f6919 | 698 | xfs_freeze(XFS_M(sb)); |
1da177e4 LT |
699 | } |
700 | ||
701 | STATIC int | |
a50cd269 | 702 | xfs_fs_show_options( |
1da177e4 LT |
703 | struct seq_file *m, |
704 | struct vfsmount *mnt) | |
705 | { | |
745f6919 | 706 | return -xfs_showargs(XFS_M(mnt->mnt_sb), m); |
1da177e4 LT |
707 | } |
708 | ||
ee34807a | 709 | STATIC int |
a50cd269 | 710 | xfs_fs_quotasync( |
ee34807a NS |
711 | struct super_block *sb, |
712 | int type) | |
713 | { | |
b09cc771 | 714 | return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL); |
ee34807a NS |
715 | } |
716 | ||
1da177e4 | 717 | STATIC int |
a50cd269 | 718 | xfs_fs_getxstate( |
1da177e4 LT |
719 | struct super_block *sb, |
720 | struct fs_quota_stat *fqs) | |
721 | { | |
b09cc771 | 722 | return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs); |
1da177e4 LT |
723 | } |
724 | ||
725 | STATIC int | |
a50cd269 | 726 | xfs_fs_setxstate( |
1da177e4 LT |
727 | struct super_block *sb, |
728 | unsigned int flags, | |
729 | int op) | |
730 | { | |
b09cc771 | 731 | return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags); |
1da177e4 LT |
732 | } |
733 | ||
734 | STATIC int | |
a50cd269 | 735 | xfs_fs_getxquota( |
1da177e4 LT |
736 | struct super_block *sb, |
737 | int type, | |
738 | qid_t id, | |
739 | struct fs_disk_quota *fdq) | |
740 | { | |
b09cc771 | 741 | return -XFS_QM_QUOTACTL(XFS_M(sb), |
b83bd138 NS |
742 | (type == USRQUOTA) ? Q_XGETQUOTA : |
743 | ((type == GRPQUOTA) ? Q_XGETGQUOTA : | |
744 | Q_XGETPQUOTA), id, (caddr_t)fdq); | |
1da177e4 LT |
745 | } |
746 | ||
747 | STATIC int | |
a50cd269 | 748 | xfs_fs_setxquota( |
1da177e4 LT |
749 | struct super_block *sb, |
750 | int type, | |
751 | qid_t id, | |
752 | struct fs_disk_quota *fdq) | |
753 | { | |
b09cc771 | 754 | return -XFS_QM_QUOTACTL(XFS_M(sb), |
b83bd138 NS |
755 | (type == USRQUOTA) ? Q_XSETQLIM : |
756 | ((type == GRPQUOTA) ? Q_XSETGQLIM : | |
757 | Q_XSETPQLIM), id, (caddr_t)fdq); | |
1da177e4 LT |
758 | } |
759 | ||
760 | STATIC int | |
a50cd269 | 761 | xfs_fs_fill_super( |
1da177e4 LT |
762 | struct super_block *sb, |
763 | void *data, | |
764 | int silent) | |
765 | { | |
0a74cd19 | 766 | struct inode *rootvp; |
b83bd138 | 767 | struct bhv_vfs *vfsp = vfs_allocate(sb); |
745f6919 | 768 | struct xfs_mount *mp = NULL; |
764d1f89 | 769 | struct xfs_mount_args *args = xfs_args_allocate(sb, silent); |
1da177e4 | 770 | struct kstatfs statvfs; |
b83bd138 | 771 | int error; |
1da177e4 | 772 | |
745f6919 | 773 | mp = xfs_mount_init(); |
74394496 CH |
774 | |
775 | INIT_LIST_HEAD(&mp->m_sync_list); | |
776 | spin_lock_init(&mp->m_sync_lock); | |
777 | init_waitqueue_head(&mp->m_wait_single_sync_task); | |
778 | ||
745f6919 CH |
779 | mp->m_vfsp = vfsp; |
780 | vfsp->vfs_mount = mp; | |
1da177e4 | 781 | |
bd186aa9 CH |
782 | if (sb->s_flags & MS_RDONLY) |
783 | mp->m_flags |= XFS_MOUNT_RDONLY; | |
784 | ||
745f6919 CH |
785 | error = xfs_parseargs(mp, (char *)data, args, 0); |
786 | if (error) | |
1da177e4 | 787 | goto fail_vfsop; |
1da177e4 LT |
788 | |
789 | sb_min_blocksize(sb, BBSIZE); | |
a50cd269 | 790 | sb->s_export_op = &xfs_export_operations; |
a50cd269 NS |
791 | sb->s_qcop = &xfs_quotactl_operations; |
792 | sb->s_op = &xfs_super_operations; | |
1da177e4 | 793 | |
745f6919 CH |
794 | error = xfs_mount(mp, args, NULL); |
795 | if (error) | |
1da177e4 | 796 | goto fail_vfsop; |
1da177e4 | 797 | |
745f6919 | 798 | error = xfs_statvfs(mp, &statvfs, NULL); |
1da177e4 LT |
799 | if (error) |
800 | goto fail_unmount; | |
801 | ||
802 | sb->s_dirt = 1; | |
803 | sb->s_magic = statvfs.f_type; | |
804 | sb->s_blocksize = statvfs.f_bsize; | |
805 | sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1; | |
806 | sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); | |
807 | sb->s_time_gran = 1; | |
808 | set_posix_acl_flag(sb); | |
809 | ||
745f6919 | 810 | error = xfs_root(mp, &rootvp); |
1da177e4 LT |
811 | if (error) |
812 | goto fail_unmount; | |
813 | ||
ec86dc02 | 814 | sb->s_root = d_alloc_root(vn_to_inode(rootvp)); |
1da177e4 LT |
815 | if (!sb->s_root) { |
816 | error = ENOMEM; | |
817 | goto fail_vnrele; | |
818 | } | |
819 | if (is_bad_inode(sb->s_root->d_inode)) { | |
820 | error = EINVAL; | |
821 | goto fail_vnrele; | |
822 | } | |
74394496 CH |
823 | |
824 | mp->m_sync_work.w_syncer = xfs_sync_worker; | |
825 | mp->m_sync_work.w_mount = mp; | |
826 | mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd"); | |
827 | if (IS_ERR(mp->m_sync_task)) { | |
828 | error = -PTR_ERR(mp->m_sync_task); | |
1da177e4 | 829 | goto fail_vnrele; |
74394496 CH |
830 | } |
831 | ||
1543d79c CH |
832 | vn_trace_exit(XFS_I(sb->s_root->d_inode), __FUNCTION__, |
833 | (inst_t *)__return_address); | |
1da177e4 LT |
834 | |
835 | kmem_free(args, sizeof(*args)); | |
836 | return 0; | |
837 | ||
838 | fail_vnrele: | |
839 | if (sb->s_root) { | |
840 | dput(sb->s_root); | |
841 | sb->s_root = NULL; | |
842 | } else { | |
843 | VN_RELE(rootvp); | |
844 | } | |
845 | ||
846 | fail_unmount: | |
745f6919 | 847 | xfs_unmount(mp, 0, NULL); |
1da177e4 LT |
848 | |
849 | fail_vfsop: | |
850 | vfs_deallocate(vfsp); | |
851 | kmem_free(args, sizeof(*args)); | |
852 | return -error; | |
853 | } | |
854 | ||
454e2398 | 855 | STATIC int |
a50cd269 | 856 | xfs_fs_get_sb( |
1da177e4 LT |
857 | struct file_system_type *fs_type, |
858 | int flags, | |
859 | const char *dev_name, | |
454e2398 DH |
860 | void *data, |
861 | struct vfsmount *mnt) | |
1da177e4 | 862 | { |
454e2398 DH |
863 | return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super, |
864 | mnt); | |
a50cd269 NS |
865 | } |
866 | ||
7989cb8e | 867 | static struct super_operations xfs_super_operations = { |
a50cd269 NS |
868 | .alloc_inode = xfs_fs_alloc_inode, |
869 | .destroy_inode = xfs_fs_destroy_inode, | |
870 | .write_inode = xfs_fs_write_inode, | |
871 | .clear_inode = xfs_fs_clear_inode, | |
872 | .put_super = xfs_fs_put_super, | |
873 | .write_super = xfs_fs_write_super, | |
874 | .sync_fs = xfs_fs_sync_super, | |
875 | .write_super_lockfs = xfs_fs_lockfs, | |
876 | .statfs = xfs_fs_statfs, | |
877 | .remount_fs = xfs_fs_remount, | |
878 | .show_options = xfs_fs_show_options, | |
1da177e4 LT |
879 | }; |
880 | ||
7989cb8e | 881 | static struct quotactl_ops xfs_quotactl_operations = { |
a50cd269 NS |
882 | .quota_sync = xfs_fs_quotasync, |
883 | .get_xstate = xfs_fs_getxstate, | |
884 | .set_xstate = xfs_fs_setxstate, | |
885 | .get_xquota = xfs_fs_getxquota, | |
886 | .set_xquota = xfs_fs_setxquota, | |
1da177e4 LT |
887 | }; |
888 | ||
5085b607 | 889 | static struct file_system_type xfs_fs_type = { |
1da177e4 LT |
890 | .owner = THIS_MODULE, |
891 | .name = "xfs", | |
a50cd269 | 892 | .get_sb = xfs_fs_get_sb, |
1da177e4 LT |
893 | .kill_sb = kill_block_super, |
894 | .fs_flags = FS_REQUIRES_DEV, | |
895 | }; | |
896 | ||
897 | ||
898 | STATIC int __init | |
899 | init_xfs_fs( void ) | |
900 | { | |
901 | int error; | |
1da177e4 LT |
902 | static char message[] __initdata = KERN_INFO \ |
903 | XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n"; | |
904 | ||
905 | printk(message); | |
906 | ||
1da177e4 LT |
907 | ktrace_init(64); |
908 | ||
8758280f | 909 | error = xfs_init_zones(); |
1da177e4 | 910 | if (error < 0) |
0829c360 | 911 | goto undo_zones; |
1da177e4 | 912 | |
ce8e922c | 913 | error = xfs_buf_init(); |
1da177e4 | 914 | if (error < 0) |
ce8e922c | 915 | goto undo_buffers; |
1da177e4 LT |
916 | |
917 | vn_init(); | |
918 | xfs_init(); | |
919 | uuid_init(); | |
920 | vfs_initquota(); | |
921 | ||
922 | error = register_filesystem(&xfs_fs_type); | |
923 | if (error) | |
924 | goto undo_register; | |
1da177e4 LT |
925 | return 0; |
926 | ||
927 | undo_register: | |
ce8e922c | 928 | xfs_buf_terminate(); |
1da177e4 | 929 | |
ce8e922c | 930 | undo_buffers: |
8758280f | 931 | xfs_destroy_zones(); |
1da177e4 | 932 | |
0829c360 | 933 | undo_zones: |
1da177e4 LT |
934 | return error; |
935 | } | |
936 | ||
937 | STATIC void __exit | |
938 | exit_xfs_fs( void ) | |
939 | { | |
940 | vfs_exitquota(); | |
1da177e4 LT |
941 | unregister_filesystem(&xfs_fs_type); |
942 | xfs_cleanup(); | |
ce8e922c | 943 | xfs_buf_terminate(); |
8758280f | 944 | xfs_destroy_zones(); |
1da177e4 LT |
945 | ktrace_uninit(); |
946 | } | |
947 | ||
948 | module_init(init_xfs_fs); | |
949 | module_exit(exit_xfs_fs); | |
950 | ||
951 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
952 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
953 | MODULE_LICENSE("GPL"); |