]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
7b718769 NS |
2 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. |
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_dir.h" |
27 | #include "xfs_dir2.h" | |
28 | #include "xfs_alloc.h" | |
29 | #include "xfs_dmapi.h" | |
30 | #include "xfs_quota.h" | |
31 | #include "xfs_mount.h" | |
1da177e4 | 32 | #include "xfs_bmap_btree.h" |
a844f451 | 33 | #include "xfs_alloc_btree.h" |
1da177e4 | 34 | #include "xfs_ialloc_btree.h" |
1da177e4 LT |
35 | #include "xfs_dir_sf.h" |
36 | #include "xfs_dir2_sf.h" | |
a844f451 | 37 | #include "xfs_attr_sf.h" |
1da177e4 LT |
38 | #include "xfs_dinode.h" |
39 | #include "xfs_inode.h" | |
a844f451 NS |
40 | #include "xfs_btree.h" |
41 | #include "xfs_ialloc.h" | |
1da177e4 | 42 | #include "xfs_bmap.h" |
1da177e4 LT |
43 | #include "xfs_rtalloc.h" |
44 | #include "xfs_error.h" | |
45 | #include "xfs_itable.h" | |
46 | #include "xfs_rw.h" | |
47 | #include "xfs_acl.h" | |
48 | #include "xfs_cap.h" | |
49 | #include "xfs_mac.h" | |
50 | #include "xfs_attr.h" | |
51 | #include "xfs_buf_item.h" | |
52 | #include "xfs_utils.h" | |
53 | #include "xfs_version.h" | |
1da177e4 LT |
54 | |
55 | #include <linux/namei.h> | |
56 | #include <linux/init.h> | |
57 | #include <linux/mount.h> | |
0829c360 | 58 | #include <linux/mempool.h> |
1da177e4 | 59 | #include <linux/writeback.h> |
4df08c52 | 60 | #include <linux/kthread.h> |
1da177e4 | 61 | |
a50cd269 NS |
62 | STATIC struct quotactl_ops xfs_quotactl_operations; |
63 | STATIC struct super_operations xfs_super_operations; | |
0829c360 CH |
64 | STATIC kmem_zone_t *xfs_vnode_zone; |
65 | STATIC kmem_zone_t *xfs_ioend_zone; | |
66 | mempool_t *xfs_ioend_pool; | |
1da177e4 LT |
67 | |
68 | STATIC struct xfs_mount_args * | |
69 | xfs_args_allocate( | |
764d1f89 NS |
70 | struct super_block *sb, |
71 | int silent) | |
1da177e4 LT |
72 | { |
73 | struct xfs_mount_args *args; | |
74 | ||
75 | args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP); | |
76 | args->logbufs = args->logbufsize = -1; | |
77 | strncpy(args->fsname, sb->s_id, MAXNAMELEN); | |
78 | ||
79 | /* Copy the already-parsed mount(2) flags we're interested in */ | |
1da177e4 LT |
80 | if (sb->s_flags & MS_DIRSYNC) |
81 | args->flags |= XFSMNT_DIRSYNC; | |
82 | if (sb->s_flags & MS_SYNCHRONOUS) | |
83 | args->flags |= XFSMNT_WSYNC; | |
764d1f89 NS |
84 | if (silent) |
85 | args->flags |= XFSMNT_QUIET; | |
1da177e4 LT |
86 | args->flags |= XFSMNT_32BITINODES; |
87 | ||
88 | return args; | |
89 | } | |
90 | ||
91 | __uint64_t | |
92 | xfs_max_file_offset( | |
93 | unsigned int blockshift) | |
94 | { | |
95 | unsigned int pagefactor = 1; | |
96 | unsigned int bitshift = BITS_PER_LONG - 1; | |
97 | ||
98 | /* Figure out maximum filesize, on Linux this can depend on | |
99 | * the filesystem blocksize (on 32 bit platforms). | |
100 | * __block_prepare_write does this in an [unsigned] long... | |
101 | * page->index << (PAGE_CACHE_SHIFT - bbits) | |
102 | * So, for page sized blocks (4K on 32 bit platforms), | |
103 | * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is | |
104 | * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1) | |
105 | * but for smaller blocksizes it is less (bbits = log2 bsize). | |
106 | * Note1: get_block_t takes a long (implicit cast from above) | |
107 | * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch | |
108 | * can optionally convert the [unsigned] long from above into | |
109 | * an [unsigned] long long. | |
110 | */ | |
111 | ||
112 | #if BITS_PER_LONG == 32 | |
113 | # if defined(CONFIG_LBD) | |
114 | ASSERT(sizeof(sector_t) == 8); | |
115 | pagefactor = PAGE_CACHE_SIZE; | |
116 | bitshift = BITS_PER_LONG; | |
117 | # else | |
118 | pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift); | |
119 | # endif | |
120 | #endif | |
121 | ||
122 | return (((__uint64_t)pagefactor) << bitshift) - 1; | |
123 | } | |
124 | ||
125 | STATIC __inline__ void | |
126 | xfs_set_inodeops( | |
127 | struct inode *inode) | |
128 | { | |
0432dab2 CH |
129 | switch (inode->i_mode & S_IFMT) { |
130 | case S_IFREG: | |
416c6d5b | 131 | inode->i_op = &xfs_inode_operations; |
3562fd45 | 132 | inode->i_fop = &xfs_file_operations; |
e4c573bb | 133 | inode->i_mapping->a_ops = &xfs_address_space_operations; |
0432dab2 CH |
134 | break; |
135 | case S_IFDIR: | |
416c6d5b | 136 | inode->i_op = &xfs_dir_inode_operations; |
3562fd45 | 137 | inode->i_fop = &xfs_dir_file_operations; |
0432dab2 CH |
138 | break; |
139 | case S_IFLNK: | |
416c6d5b | 140 | inode->i_op = &xfs_symlink_inode_operations; |
1da177e4 | 141 | if (inode->i_blocks) |
e4c573bb | 142 | inode->i_mapping->a_ops = &xfs_address_space_operations; |
0432dab2 CH |
143 | break; |
144 | default: | |
416c6d5b | 145 | inode->i_op = &xfs_inode_operations; |
1da177e4 | 146 | init_special_inode(inode, inode->i_mode, inode->i_rdev); |
0432dab2 | 147 | break; |
1da177e4 LT |
148 | } |
149 | } | |
150 | ||
151 | STATIC __inline__ void | |
152 | xfs_revalidate_inode( | |
153 | xfs_mount_t *mp, | |
154 | vnode_t *vp, | |
155 | xfs_inode_t *ip) | |
156 | { | |
ec86dc02 | 157 | struct inode *inode = vn_to_inode(vp); |
1da177e4 | 158 | |
0432dab2 | 159 | inode->i_mode = ip->i_d.di_mode; |
1da177e4 LT |
160 | inode->i_nlink = ip->i_d.di_nlink; |
161 | inode->i_uid = ip->i_d.di_uid; | |
162 | inode->i_gid = ip->i_d.di_gid; | |
0432dab2 CH |
163 | |
164 | switch (inode->i_mode & S_IFMT) { | |
165 | case S_IFBLK: | |
166 | case S_IFCHR: | |
167 | inode->i_rdev = | |
168 | MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff, | |
169 | sysv_minor(ip->i_df.if_u2.if_rdev)); | |
170 | break; | |
171 | default: | |
1da177e4 | 172 | inode->i_rdev = 0; |
0432dab2 | 173 | break; |
1da177e4 | 174 | } |
0432dab2 | 175 | |
e8c8b3a7 | 176 | inode->i_blksize = xfs_preferred_iosize(mp); |
1da177e4 LT |
177 | inode->i_generation = ip->i_d.di_gen; |
178 | i_size_write(inode, ip->i_d.di_size); | |
179 | inode->i_blocks = | |
180 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
181 | inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec; | |
182 | inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec; | |
183 | inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec; | |
184 | inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec; | |
185 | inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec; | |
186 | inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec; | |
187 | if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE) | |
188 | inode->i_flags |= S_IMMUTABLE; | |
189 | else | |
190 | inode->i_flags &= ~S_IMMUTABLE; | |
191 | if (ip->i_d.di_flags & XFS_DIFLAG_APPEND) | |
192 | inode->i_flags |= S_APPEND; | |
193 | else | |
194 | inode->i_flags &= ~S_APPEND; | |
195 | if (ip->i_d.di_flags & XFS_DIFLAG_SYNC) | |
196 | inode->i_flags |= S_SYNC; | |
197 | else | |
198 | inode->i_flags &= ~S_SYNC; | |
199 | if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME) | |
200 | inode->i_flags |= S_NOATIME; | |
201 | else | |
202 | inode->i_flags &= ~S_NOATIME; | |
203 | vp->v_flag &= ~VMODIFIED; | |
204 | } | |
205 | ||
206 | void | |
207 | xfs_initialize_vnode( | |
208 | bhv_desc_t *bdp, | |
209 | vnode_t *vp, | |
210 | bhv_desc_t *inode_bhv, | |
211 | int unlock) | |
212 | { | |
213 | xfs_inode_t *ip = XFS_BHVTOI(inode_bhv); | |
ec86dc02 | 214 | struct inode *inode = vn_to_inode(vp); |
1da177e4 LT |
215 | |
216 | if (!inode_bhv->bd_vobj) { | |
217 | vp->v_vfsp = bhvtovfs(bdp); | |
218 | bhv_desc_init(inode_bhv, ip, vp, &xfs_vnodeops); | |
219 | bhv_insert(VN_BHV_HEAD(vp), inode_bhv); | |
220 | } | |
221 | ||
222 | /* | |
223 | * We need to set the ops vectors, and unlock the inode, but if | |
224 | * we have been called during the new inode create process, it is | |
225 | * too early to fill in the Linux inode. We will get called a | |
226 | * second time once the inode is properly set up, and then we can | |
227 | * finish our work. | |
228 | */ | |
229 | if (ip->i_d.di_mode != 0 && unlock && (inode->i_state & I_NEW)) { | |
1da177e4 LT |
230 | xfs_revalidate_inode(XFS_BHVTOM(bdp), vp, ip); |
231 | xfs_set_inodeops(inode); | |
ec86dc02 | 232 | |
1da177e4 LT |
233 | ip->i_flags &= ~XFS_INEW; |
234 | barrier(); | |
235 | ||
236 | unlock_new_inode(inode); | |
237 | } | |
238 | } | |
239 | ||
240 | int | |
241 | xfs_blkdev_get( | |
242 | xfs_mount_t *mp, | |
243 | const char *name, | |
244 | struct block_device **bdevp) | |
245 | { | |
246 | int error = 0; | |
247 | ||
248 | *bdevp = open_bdev_excl(name, 0, mp); | |
249 | if (IS_ERR(*bdevp)) { | |
250 | error = PTR_ERR(*bdevp); | |
251 | printk("XFS: Invalid device [%s], error=%d\n", name, error); | |
252 | } | |
253 | ||
254 | return -error; | |
255 | } | |
256 | ||
257 | void | |
258 | xfs_blkdev_put( | |
259 | struct block_device *bdev) | |
260 | { | |
261 | if (bdev) | |
262 | close_bdev_excl(bdev); | |
263 | } | |
264 | ||
f538d4da CH |
265 | /* |
266 | * Try to write out the superblock using barriers. | |
267 | */ | |
268 | STATIC int | |
269 | xfs_barrier_test( | |
270 | xfs_mount_t *mp) | |
271 | { | |
272 | xfs_buf_t *sbp = xfs_getsb(mp, 0); | |
273 | int error; | |
274 | ||
275 | XFS_BUF_UNDONE(sbp); | |
276 | XFS_BUF_UNREAD(sbp); | |
277 | XFS_BUF_UNDELAYWRITE(sbp); | |
278 | XFS_BUF_WRITE(sbp); | |
279 | XFS_BUF_UNASYNC(sbp); | |
280 | XFS_BUF_ORDERED(sbp); | |
281 | ||
282 | xfsbdstrat(mp, sbp); | |
283 | error = xfs_iowait(sbp); | |
284 | ||
285 | /* | |
286 | * Clear all the flags we set and possible error state in the | |
287 | * buffer. We only did the write to try out whether barriers | |
288 | * worked and shouldn't leave any traces in the superblock | |
289 | * buffer. | |
290 | */ | |
291 | XFS_BUF_DONE(sbp); | |
292 | XFS_BUF_ERROR(sbp, 0); | |
293 | XFS_BUF_UNORDERED(sbp); | |
294 | ||
295 | xfs_buf_relse(sbp); | |
296 | return error; | |
297 | } | |
298 | ||
299 | void | |
300 | xfs_mountfs_check_barriers(xfs_mount_t *mp) | |
301 | { | |
302 | int error; | |
303 | ||
304 | if (mp->m_logdev_targp != mp->m_ddev_targp) { | |
305 | xfs_fs_cmn_err(CE_NOTE, mp, | |
306 | "Disabling barriers, not supported with external log device"); | |
307 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 308 | return; |
f538d4da CH |
309 | } |
310 | ||
ce8e922c | 311 | if (mp->m_ddev_targp->bt_bdev->bd_disk->queue->ordered == |
f538d4da CH |
312 | QUEUE_ORDERED_NONE) { |
313 | xfs_fs_cmn_err(CE_NOTE, mp, | |
314 | "Disabling barriers, not supported by the underlying device"); | |
315 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 316 | return; |
f538d4da CH |
317 | } |
318 | ||
319 | error = xfs_barrier_test(mp); | |
320 | if (error) { | |
321 | xfs_fs_cmn_err(CE_NOTE, mp, | |
322 | "Disabling barriers, trial barrier write failed"); | |
323 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 324 | return; |
f538d4da CH |
325 | } |
326 | } | |
327 | ||
328 | void | |
329 | xfs_blkdev_issue_flush( | |
330 | xfs_buftarg_t *buftarg) | |
331 | { | |
ce8e922c | 332 | blkdev_issue_flush(buftarg->bt_bdev, NULL); |
f538d4da | 333 | } |
1da177e4 LT |
334 | |
335 | STATIC struct inode * | |
a50cd269 | 336 | xfs_fs_alloc_inode( |
1da177e4 LT |
337 | struct super_block *sb) |
338 | { | |
339 | vnode_t *vp; | |
340 | ||
8758280f NS |
341 | vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP); |
342 | if (unlikely(!vp)) | |
1da177e4 | 343 | return NULL; |
ec86dc02 | 344 | return vn_to_inode(vp); |
1da177e4 LT |
345 | } |
346 | ||
347 | STATIC void | |
a50cd269 | 348 | xfs_fs_destroy_inode( |
1da177e4 LT |
349 | struct inode *inode) |
350 | { | |
ec86dc02 | 351 | kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode)); |
1da177e4 LT |
352 | } |
353 | ||
354 | STATIC void | |
a50cd269 | 355 | xfs_fs_inode_init_once( |
8758280f NS |
356 | void *vnode, |
357 | kmem_zone_t *zonep, | |
1da177e4 LT |
358 | unsigned long flags) |
359 | { | |
1da177e4 | 360 | if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == |
8758280f | 361 | SLAB_CTOR_CONSTRUCTOR) |
ec86dc02 | 362 | inode_init_once(vn_to_inode((vnode_t *)vnode)); |
1da177e4 LT |
363 | } |
364 | ||
365 | STATIC int | |
8758280f | 366 | xfs_init_zones(void) |
1da177e4 | 367 | { |
8758280f | 368 | xfs_vnode_zone = kmem_zone_init_flags(sizeof(vnode_t), "xfs_vnode_t", |
e0cc2325 NS |
369 | KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | |
370 | KM_ZONE_SPREAD, | |
a50cd269 | 371 | xfs_fs_inode_init_once); |
0829c360 CH |
372 | if (!xfs_vnode_zone) |
373 | goto out; | |
374 | ||
375 | xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend"); | |
376 | if (!xfs_ioend_zone) | |
377 | goto out_destroy_vnode_zone; | |
378 | ||
93d2341c MD |
379 | xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE, |
380 | xfs_ioend_zone); | |
0829c360 CH |
381 | if (!xfs_ioend_pool) |
382 | goto out_free_ioend_zone; | |
1da177e4 | 383 | return 0; |
0829c360 | 384 | |
0829c360 CH |
385 | out_free_ioend_zone: |
386 | kmem_zone_destroy(xfs_ioend_zone); | |
387 | out_destroy_vnode_zone: | |
388 | kmem_zone_destroy(xfs_vnode_zone); | |
389 | out: | |
390 | return -ENOMEM; | |
1da177e4 LT |
391 | } |
392 | ||
393 | STATIC void | |
8758280f | 394 | xfs_destroy_zones(void) |
1da177e4 | 395 | { |
0829c360 CH |
396 | mempool_destroy(xfs_ioend_pool); |
397 | kmem_zone_destroy(xfs_vnode_zone); | |
398 | kmem_zone_destroy(xfs_ioend_zone); | |
1da177e4 LT |
399 | } |
400 | ||
401 | /* | |
402 | * Attempt to flush the inode, this will actually fail | |
403 | * if the inode is pinned, but we dirty the inode again | |
404 | * at the point when it is unpinned after a log write, | |
8758280f | 405 | * since this is when the inode itself becomes flushable. |
1da177e4 LT |
406 | */ |
407 | STATIC int | |
a50cd269 | 408 | xfs_fs_write_inode( |
1da177e4 LT |
409 | struct inode *inode, |
410 | int sync) | |
411 | { | |
ec86dc02 | 412 | vnode_t *vp = vn_from_inode(inode); |
1da177e4 LT |
413 | int error = 0, flags = FLUSH_INODE; |
414 | ||
415 | if (vp) { | |
416 | vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); | |
417 | if (sync) | |
418 | flags |= FLUSH_SYNC; | |
419 | VOP_IFLUSH(vp, flags, error); | |
420 | if (error == EAGAIN) { | |
421 | if (sync) | |
422 | VOP_IFLUSH(vp, flags | FLUSH_LOG, error); | |
423 | else | |
424 | error = 0; | |
425 | } | |
426 | } | |
427 | ||
428 | return -error; | |
429 | } | |
430 | ||
431 | STATIC void | |
a50cd269 | 432 | xfs_fs_clear_inode( |
1da177e4 LT |
433 | struct inode *inode) |
434 | { | |
ec86dc02 | 435 | vnode_t *vp = vn_from_inode(inode); |
56d433e4 | 436 | int error, cache; |
1da177e4 | 437 | |
a50cd269 | 438 | vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); |
56d433e4 | 439 | |
56d433e4 CH |
440 | XFS_STATS_INC(vn_rele); |
441 | XFS_STATS_INC(vn_remove); | |
442 | XFS_STATS_INC(vn_reclaim); | |
443 | XFS_STATS_DEC(vn_active); | |
444 | ||
02ba71de CH |
445 | /* |
446 | * This can happen because xfs_iget_core calls xfs_idestroy if we | |
447 | * find an inode with di_mode == 0 but without IGET_CREATE set. | |
448 | */ | |
449 | if (vp->v_fbhv) | |
450 | VOP_INACTIVE(vp, NULL, cache); | |
1da177e4 | 451 | |
56d433e4 CH |
452 | VN_LOCK(vp); |
453 | vp->v_flag &= ~VMODIFIED; | |
454 | VN_UNLOCK(vp, 0); | |
455 | ||
0c147f9a FB |
456 | if (vp->v_fbhv) { |
457 | VOP_RECLAIM(vp, error); | |
458 | if (error) | |
459 | panic("vn_purge: cannot reclaim"); | |
460 | } | |
56d433e4 CH |
461 | |
462 | ASSERT(vp->v_fbhv == NULL); | |
463 | ||
464 | #ifdef XFS_VNODE_TRACE | |
465 | ktrace_free(vp->v_trace); | |
466 | #endif | |
467 | } | |
1da177e4 LT |
468 | |
469 | /* | |
470 | * Enqueue a work item to be picked up by the vfs xfssyncd thread. | |
471 | * Doing this has two advantages: | |
472 | * - It saves on stack space, which is tight in certain situations | |
473 | * - It can be used (with care) as a mechanism to avoid deadlocks. | |
474 | * Flushing while allocating in a full filesystem requires both. | |
475 | */ | |
476 | STATIC void | |
477 | xfs_syncd_queue_work( | |
478 | struct vfs *vfs, | |
479 | void *data, | |
480 | void (*syncer)(vfs_t *, void *)) | |
481 | { | |
482 | vfs_sync_work_t *work; | |
483 | ||
484 | work = kmem_alloc(sizeof(struct vfs_sync_work), KM_SLEEP); | |
485 | INIT_LIST_HEAD(&work->w_list); | |
486 | work->w_syncer = syncer; | |
487 | work->w_data = data; | |
488 | work->w_vfs = vfs; | |
489 | spin_lock(&vfs->vfs_sync_lock); | |
490 | list_add_tail(&work->w_list, &vfs->vfs_sync_list); | |
491 | spin_unlock(&vfs->vfs_sync_lock); | |
492 | wake_up_process(vfs->vfs_sync_task); | |
493 | } | |
494 | ||
495 | /* | |
496 | * Flush delayed allocate data, attempting to free up reserved space | |
497 | * from existing allocations. At this point a new allocation attempt | |
498 | * has failed with ENOSPC and we are in the process of scratching our | |
499 | * heads, looking about for more room... | |
500 | */ | |
501 | STATIC void | |
502 | xfs_flush_inode_work( | |
503 | vfs_t *vfs, | |
504 | void *inode) | |
505 | { | |
506 | filemap_flush(((struct inode *)inode)->i_mapping); | |
507 | iput((struct inode *)inode); | |
508 | } | |
509 | ||
510 | void | |
511 | xfs_flush_inode( | |
512 | xfs_inode_t *ip) | |
513 | { | |
ec86dc02 | 514 | struct inode *inode = vn_to_inode(XFS_ITOV(ip)); |
1da177e4 LT |
515 | struct vfs *vfs = XFS_MTOVFS(ip->i_mount); |
516 | ||
517 | igrab(inode); | |
518 | xfs_syncd_queue_work(vfs, inode, xfs_flush_inode_work); | |
041e0e3b | 519 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
520 | } |
521 | ||
522 | /* | |
523 | * This is the "bigger hammer" version of xfs_flush_inode_work... | |
524 | * (IOW, "If at first you don't succeed, use a Bigger Hammer"). | |
525 | */ | |
526 | STATIC void | |
527 | xfs_flush_device_work( | |
528 | vfs_t *vfs, | |
529 | void *inode) | |
530 | { | |
531 | sync_blockdev(vfs->vfs_super->s_bdev); | |
532 | iput((struct inode *)inode); | |
533 | } | |
534 | ||
535 | void | |
536 | xfs_flush_device( | |
537 | xfs_inode_t *ip) | |
538 | { | |
ec86dc02 | 539 | struct inode *inode = vn_to_inode(XFS_ITOV(ip)); |
1da177e4 LT |
540 | struct vfs *vfs = XFS_MTOVFS(ip->i_mount); |
541 | ||
542 | igrab(inode); | |
543 | xfs_syncd_queue_work(vfs, inode, xfs_flush_device_work); | |
041e0e3b | 544 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
545 | xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC); |
546 | } | |
547 | ||
bb19fba1 | 548 | #define SYNCD_FLAGS (SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR|SYNC_REFCACHE) |
1da177e4 LT |
549 | STATIC void |
550 | vfs_sync_worker( | |
551 | vfs_t *vfsp, | |
552 | void *unused) | |
553 | { | |
554 | int error; | |
555 | ||
556 | if (!(vfsp->vfs_flag & VFS_RDONLY)) | |
557 | VFS_SYNC(vfsp, SYNCD_FLAGS, NULL, error); | |
558 | vfsp->vfs_sync_seq++; | |
559 | wmb(); | |
560 | wake_up(&vfsp->vfs_wait_single_sync_task); | |
561 | } | |
562 | ||
563 | STATIC int | |
564 | xfssyncd( | |
565 | void *arg) | |
566 | { | |
567 | long timeleft; | |
568 | vfs_t *vfsp = (vfs_t *) arg; | |
1da177e4 | 569 | struct vfs_sync_work *work, *n; |
4df08c52 | 570 | LIST_HEAD (tmp); |
1da177e4 | 571 | |
041e0e3b | 572 | timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10); |
1da177e4 | 573 | for (;;) { |
041e0e3b | 574 | timeleft = schedule_timeout_interruptible(timeleft); |
1da177e4 | 575 | /* swsusp */ |
3e1d1d28 | 576 | try_to_freeze(); |
71df099d | 577 | if (kthread_should_stop() && list_empty(&vfsp->vfs_sync_list)) |
1da177e4 LT |
578 | break; |
579 | ||
580 | spin_lock(&vfsp->vfs_sync_lock); | |
581 | /* | |
582 | * We can get woken by laptop mode, to do a sync - | |
583 | * that's the (only!) case where the list would be | |
584 | * empty with time remaining. | |
585 | */ | |
586 | if (!timeleft || list_empty(&vfsp->vfs_sync_list)) { | |
587 | if (!timeleft) | |
041e0e3b NA |
588 | timeleft = xfs_syncd_centisecs * |
589 | msecs_to_jiffies(10); | |
1da177e4 LT |
590 | INIT_LIST_HEAD(&vfsp->vfs_sync_work.w_list); |
591 | list_add_tail(&vfsp->vfs_sync_work.w_list, | |
592 | &vfsp->vfs_sync_list); | |
593 | } | |
594 | list_for_each_entry_safe(work, n, &vfsp->vfs_sync_list, w_list) | |
595 | list_move(&work->w_list, &tmp); | |
596 | spin_unlock(&vfsp->vfs_sync_lock); | |
597 | ||
598 | list_for_each_entry_safe(work, n, &tmp, w_list) { | |
599 | (*work->w_syncer)(vfsp, work->w_data); | |
600 | list_del(&work->w_list); | |
601 | if (work == &vfsp->vfs_sync_work) | |
602 | continue; | |
603 | kmem_free(work, sizeof(struct vfs_sync_work)); | |
604 | } | |
605 | } | |
606 | ||
1da177e4 LT |
607 | return 0; |
608 | } | |
609 | ||
610 | STATIC int | |
a50cd269 | 611 | xfs_fs_start_syncd( |
1da177e4 LT |
612 | vfs_t *vfsp) |
613 | { | |
4df08c52 CH |
614 | vfsp->vfs_sync_work.w_syncer = vfs_sync_worker; |
615 | vfsp->vfs_sync_work.w_vfs = vfsp; | |
616 | vfsp->vfs_sync_task = kthread_run(xfssyncd, vfsp, "xfssyncd"); | |
617 | if (IS_ERR(vfsp->vfs_sync_task)) | |
618 | return -PTR_ERR(vfsp->vfs_sync_task); | |
1da177e4 LT |
619 | return 0; |
620 | } | |
621 | ||
622 | STATIC void | |
a50cd269 | 623 | xfs_fs_stop_syncd( |
1da177e4 LT |
624 | vfs_t *vfsp) |
625 | { | |
4df08c52 | 626 | kthread_stop(vfsp->vfs_sync_task); |
1da177e4 LT |
627 | } |
628 | ||
629 | STATIC void | |
a50cd269 | 630 | xfs_fs_put_super( |
1da177e4 LT |
631 | struct super_block *sb) |
632 | { | |
ec86dc02 | 633 | vfs_t *vfsp = vfs_from_sb(sb); |
1da177e4 LT |
634 | int error; |
635 | ||
a50cd269 | 636 | xfs_fs_stop_syncd(vfsp); |
1da177e4 LT |
637 | VFS_SYNC(vfsp, SYNC_ATTR|SYNC_DELWRI, NULL, error); |
638 | if (!error) | |
639 | VFS_UNMOUNT(vfsp, 0, NULL, error); | |
640 | if (error) { | |
641 | printk("XFS unmount got error %d\n", error); | |
642 | printk("%s: vfsp/0x%p left dangling!\n", __FUNCTION__, vfsp); | |
643 | return; | |
644 | } | |
645 | ||
646 | vfs_deallocate(vfsp); | |
647 | } | |
648 | ||
649 | STATIC void | |
a50cd269 | 650 | xfs_fs_write_super( |
1da177e4 LT |
651 | struct super_block *sb) |
652 | { | |
ec86dc02 | 653 | vfs_t *vfsp = vfs_from_sb(sb); |
1da177e4 LT |
654 | int error; |
655 | ||
656 | if (sb->s_flags & MS_RDONLY) { | |
657 | sb->s_dirt = 0; /* paranoia */ | |
658 | return; | |
659 | } | |
660 | /* Push the log and superblock a little */ | |
661 | VFS_SYNC(vfsp, SYNC_FSDATA, NULL, error); | |
662 | sb->s_dirt = 0; | |
663 | } | |
664 | ||
665 | STATIC int | |
a50cd269 | 666 | xfs_fs_sync_super( |
1da177e4 LT |
667 | struct super_block *sb, |
668 | int wait) | |
669 | { | |
ec86dc02 | 670 | vfs_t *vfsp = vfs_from_sb(sb); |
1da177e4 LT |
671 | int error; |
672 | int flags = SYNC_FSDATA; | |
673 | ||
f898d6c0 CH |
674 | if (unlikely(sb->s_frozen == SB_FREEZE_WRITE)) |
675 | flags = SYNC_QUIESCE; | |
676 | else | |
677 | flags = SYNC_FSDATA | (wait ? SYNC_WAIT : 0); | |
1da177e4 LT |
678 | |
679 | VFS_SYNC(vfsp, flags, NULL, error); | |
680 | sb->s_dirt = 0; | |
681 | ||
682 | if (unlikely(laptop_mode)) { | |
683 | int prev_sync_seq = vfsp->vfs_sync_seq; | |
684 | ||
685 | /* | |
686 | * The disk must be active because we're syncing. | |
687 | * We schedule xfssyncd now (now that the disk is | |
688 | * active) instead of later (when it might not be). | |
689 | */ | |
690 | wake_up_process(vfsp->vfs_sync_task); | |
691 | /* | |
692 | * We have to wait for the sync iteration to complete. | |
693 | * If we don't, the disk activity caused by the sync | |
694 | * will come after the sync is completed, and that | |
695 | * triggers another sync from laptop mode. | |
696 | */ | |
697 | wait_event(vfsp->vfs_wait_single_sync_task, | |
698 | vfsp->vfs_sync_seq != prev_sync_seq); | |
699 | } | |
700 | ||
701 | return -error; | |
702 | } | |
703 | ||
704 | STATIC int | |
a50cd269 | 705 | xfs_fs_statfs( |
1da177e4 LT |
706 | struct super_block *sb, |
707 | struct kstatfs *statp) | |
708 | { | |
ec86dc02 | 709 | vfs_t *vfsp = vfs_from_sb(sb); |
1da177e4 LT |
710 | int error; |
711 | ||
712 | VFS_STATVFS(vfsp, statp, NULL, error); | |
713 | return -error; | |
714 | } | |
715 | ||
716 | STATIC int | |
a50cd269 | 717 | xfs_fs_remount( |
1da177e4 LT |
718 | struct super_block *sb, |
719 | int *flags, | |
720 | char *options) | |
721 | { | |
ec86dc02 | 722 | vfs_t *vfsp = vfs_from_sb(sb); |
764d1f89 | 723 | struct xfs_mount_args *args = xfs_args_allocate(sb, 0); |
1da177e4 LT |
724 | int error; |
725 | ||
726 | VFS_PARSEARGS(vfsp, options, args, 1, error); | |
727 | if (!error) | |
728 | VFS_MNTUPDATE(vfsp, flags, args, error); | |
729 | kmem_free(args, sizeof(*args)); | |
730 | return -error; | |
731 | } | |
732 | ||
733 | STATIC void | |
a50cd269 | 734 | xfs_fs_lockfs( |
1da177e4 LT |
735 | struct super_block *sb) |
736 | { | |
ec86dc02 | 737 | VFS_FREEZE(vfs_from_sb(sb)); |
1da177e4 LT |
738 | } |
739 | ||
740 | STATIC int | |
a50cd269 | 741 | xfs_fs_show_options( |
1da177e4 LT |
742 | struct seq_file *m, |
743 | struct vfsmount *mnt) | |
744 | { | |
ec86dc02 | 745 | struct vfs *vfsp = vfs_from_sb(mnt->mnt_sb); |
1da177e4 LT |
746 | int error; |
747 | ||
748 | VFS_SHOWARGS(vfsp, m, error); | |
749 | return error; | |
750 | } | |
751 | ||
ee34807a | 752 | STATIC int |
a50cd269 | 753 | xfs_fs_quotasync( |
ee34807a NS |
754 | struct super_block *sb, |
755 | int type) | |
756 | { | |
ec86dc02 | 757 | struct vfs *vfsp = vfs_from_sb(sb); |
ee34807a NS |
758 | int error; |
759 | ||
760 | VFS_QUOTACTL(vfsp, Q_XQUOTASYNC, 0, (caddr_t)NULL, error); | |
761 | return -error; | |
762 | } | |
763 | ||
1da177e4 | 764 | STATIC int |
a50cd269 | 765 | xfs_fs_getxstate( |
1da177e4 LT |
766 | struct super_block *sb, |
767 | struct fs_quota_stat *fqs) | |
768 | { | |
ec86dc02 | 769 | struct vfs *vfsp = vfs_from_sb(sb); |
1da177e4 LT |
770 | int error; |
771 | ||
772 | VFS_QUOTACTL(vfsp, Q_XGETQSTAT, 0, (caddr_t)fqs, error); | |
773 | return -error; | |
774 | } | |
775 | ||
776 | STATIC int | |
a50cd269 | 777 | xfs_fs_setxstate( |
1da177e4 LT |
778 | struct super_block *sb, |
779 | unsigned int flags, | |
780 | int op) | |
781 | { | |
ec86dc02 | 782 | struct vfs *vfsp = vfs_from_sb(sb); |
1da177e4 LT |
783 | int error; |
784 | ||
785 | VFS_QUOTACTL(vfsp, op, 0, (caddr_t)&flags, error); | |
786 | return -error; | |
787 | } | |
788 | ||
789 | STATIC int | |
a50cd269 | 790 | xfs_fs_getxquota( |
1da177e4 LT |
791 | struct super_block *sb, |
792 | int type, | |
793 | qid_t id, | |
794 | struct fs_disk_quota *fdq) | |
795 | { | |
ec86dc02 | 796 | struct vfs *vfsp = vfs_from_sb(sb); |
1da177e4 LT |
797 | int error, getmode; |
798 | ||
c8ad20ff NS |
799 | getmode = (type == USRQUOTA) ? Q_XGETQUOTA : |
800 | ((type == GRPQUOTA) ? Q_XGETGQUOTA : Q_XGETPQUOTA); | |
1da177e4 LT |
801 | VFS_QUOTACTL(vfsp, getmode, id, (caddr_t)fdq, error); |
802 | return -error; | |
803 | } | |
804 | ||
805 | STATIC int | |
a50cd269 | 806 | xfs_fs_setxquota( |
1da177e4 LT |
807 | struct super_block *sb, |
808 | int type, | |
809 | qid_t id, | |
810 | struct fs_disk_quota *fdq) | |
811 | { | |
ec86dc02 | 812 | struct vfs *vfsp = vfs_from_sb(sb); |
1da177e4 LT |
813 | int error, setmode; |
814 | ||
c8ad20ff NS |
815 | setmode = (type == USRQUOTA) ? Q_XSETQLIM : |
816 | ((type == GRPQUOTA) ? Q_XSETGQLIM : Q_XSETPQLIM); | |
1da177e4 LT |
817 | VFS_QUOTACTL(vfsp, setmode, id, (caddr_t)fdq, error); |
818 | return -error; | |
819 | } | |
820 | ||
821 | STATIC int | |
a50cd269 | 822 | xfs_fs_fill_super( |
1da177e4 LT |
823 | struct super_block *sb, |
824 | void *data, | |
825 | int silent) | |
826 | { | |
827 | vnode_t *rootvp; | |
ec86dc02 | 828 | struct vfs *vfsp = vfs_allocate(sb); |
764d1f89 | 829 | struct xfs_mount_args *args = xfs_args_allocate(sb, silent); |
1da177e4 LT |
830 | struct kstatfs statvfs; |
831 | int error, error2; | |
832 | ||
1da177e4 LT |
833 | bhv_insert_all_vfsops(vfsp); |
834 | ||
835 | VFS_PARSEARGS(vfsp, (char *)data, args, 0, error); | |
836 | if (error) { | |
837 | bhv_remove_all_vfsops(vfsp, 1); | |
838 | goto fail_vfsop; | |
839 | } | |
840 | ||
841 | sb_min_blocksize(sb, BBSIZE); | |
842 | #ifdef CONFIG_XFS_EXPORT | |
a50cd269 | 843 | sb->s_export_op = &xfs_export_operations; |
1da177e4 | 844 | #endif |
a50cd269 NS |
845 | sb->s_qcop = &xfs_quotactl_operations; |
846 | sb->s_op = &xfs_super_operations; | |
1da177e4 LT |
847 | |
848 | VFS_MOUNT(vfsp, args, NULL, error); | |
849 | if (error) { | |
850 | bhv_remove_all_vfsops(vfsp, 1); | |
851 | goto fail_vfsop; | |
852 | } | |
853 | ||
854 | VFS_STATVFS(vfsp, &statvfs, NULL, error); | |
855 | if (error) | |
856 | goto fail_unmount; | |
857 | ||
858 | sb->s_dirt = 1; | |
859 | sb->s_magic = statvfs.f_type; | |
860 | sb->s_blocksize = statvfs.f_bsize; | |
861 | sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1; | |
862 | sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); | |
863 | sb->s_time_gran = 1; | |
864 | set_posix_acl_flag(sb); | |
865 | ||
866 | VFS_ROOT(vfsp, &rootvp, error); | |
867 | if (error) | |
868 | goto fail_unmount; | |
869 | ||
ec86dc02 | 870 | sb->s_root = d_alloc_root(vn_to_inode(rootvp)); |
1da177e4 LT |
871 | if (!sb->s_root) { |
872 | error = ENOMEM; | |
873 | goto fail_vnrele; | |
874 | } | |
875 | if (is_bad_inode(sb->s_root->d_inode)) { | |
876 | error = EINVAL; | |
877 | goto fail_vnrele; | |
878 | } | |
a50cd269 | 879 | if ((error = xfs_fs_start_syncd(vfsp))) |
1da177e4 LT |
880 | goto fail_vnrele; |
881 | vn_trace_exit(rootvp, __FUNCTION__, (inst_t *)__return_address); | |
882 | ||
883 | kmem_free(args, sizeof(*args)); | |
884 | return 0; | |
885 | ||
886 | fail_vnrele: | |
887 | if (sb->s_root) { | |
888 | dput(sb->s_root); | |
889 | sb->s_root = NULL; | |
890 | } else { | |
891 | VN_RELE(rootvp); | |
892 | } | |
893 | ||
894 | fail_unmount: | |
895 | VFS_UNMOUNT(vfsp, 0, NULL, error2); | |
896 | ||
897 | fail_vfsop: | |
898 | vfs_deallocate(vfsp); | |
899 | kmem_free(args, sizeof(*args)); | |
900 | return -error; | |
901 | } | |
902 | ||
903 | STATIC struct super_block * | |
a50cd269 | 904 | xfs_fs_get_sb( |
1da177e4 LT |
905 | struct file_system_type *fs_type, |
906 | int flags, | |
907 | const char *dev_name, | |
908 | void *data) | |
909 | { | |
a50cd269 NS |
910 | return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super); |
911 | } | |
912 | ||
913 | STATIC struct super_operations xfs_super_operations = { | |
914 | .alloc_inode = xfs_fs_alloc_inode, | |
915 | .destroy_inode = xfs_fs_destroy_inode, | |
916 | .write_inode = xfs_fs_write_inode, | |
917 | .clear_inode = xfs_fs_clear_inode, | |
918 | .put_super = xfs_fs_put_super, | |
919 | .write_super = xfs_fs_write_super, | |
920 | .sync_fs = xfs_fs_sync_super, | |
921 | .write_super_lockfs = xfs_fs_lockfs, | |
922 | .statfs = xfs_fs_statfs, | |
923 | .remount_fs = xfs_fs_remount, | |
924 | .show_options = xfs_fs_show_options, | |
1da177e4 LT |
925 | }; |
926 | ||
a50cd269 NS |
927 | STATIC struct quotactl_ops xfs_quotactl_operations = { |
928 | .quota_sync = xfs_fs_quotasync, | |
929 | .get_xstate = xfs_fs_getxstate, | |
930 | .set_xstate = xfs_fs_setxstate, | |
931 | .get_xquota = xfs_fs_getxquota, | |
932 | .set_xquota = xfs_fs_setxquota, | |
1da177e4 LT |
933 | }; |
934 | ||
935 | STATIC struct file_system_type xfs_fs_type = { | |
936 | .owner = THIS_MODULE, | |
937 | .name = "xfs", | |
a50cd269 | 938 | .get_sb = xfs_fs_get_sb, |
1da177e4 LT |
939 | .kill_sb = kill_block_super, |
940 | .fs_flags = FS_REQUIRES_DEV, | |
941 | }; | |
942 | ||
943 | ||
944 | STATIC int __init | |
945 | init_xfs_fs( void ) | |
946 | { | |
947 | int error; | |
948 | struct sysinfo si; | |
949 | static char message[] __initdata = KERN_INFO \ | |
950 | XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n"; | |
951 | ||
952 | printk(message); | |
953 | ||
954 | si_meminfo(&si); | |
955 | xfs_physmem = si.totalram; | |
956 | ||
957 | ktrace_init(64); | |
958 | ||
8758280f | 959 | error = xfs_init_zones(); |
1da177e4 | 960 | if (error < 0) |
0829c360 | 961 | goto undo_zones; |
1da177e4 | 962 | |
ce8e922c | 963 | error = xfs_buf_init(); |
1da177e4 | 964 | if (error < 0) |
ce8e922c | 965 | goto undo_buffers; |
1da177e4 LT |
966 | |
967 | vn_init(); | |
968 | xfs_init(); | |
969 | uuid_init(); | |
970 | vfs_initquota(); | |
971 | ||
972 | error = register_filesystem(&xfs_fs_type); | |
973 | if (error) | |
974 | goto undo_register; | |
1da177e4 LT |
975 | return 0; |
976 | ||
977 | undo_register: | |
ce8e922c | 978 | xfs_buf_terminate(); |
1da177e4 | 979 | |
ce8e922c | 980 | undo_buffers: |
8758280f | 981 | xfs_destroy_zones(); |
1da177e4 | 982 | |
0829c360 | 983 | undo_zones: |
1da177e4 LT |
984 | return error; |
985 | } | |
986 | ||
987 | STATIC void __exit | |
988 | exit_xfs_fs( void ) | |
989 | { | |
990 | vfs_exitquota(); | |
1da177e4 LT |
991 | unregister_filesystem(&xfs_fs_type); |
992 | xfs_cleanup(); | |
ce8e922c | 993 | xfs_buf_terminate(); |
8758280f | 994 | xfs_destroy_zones(); |
1da177e4 LT |
995 | ktrace_uninit(); |
996 | } | |
997 | ||
998 | module_init(init_xfs_fs); | |
999 | module_exit(exit_xfs_fs); | |
1000 | ||
1001 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
1002 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
1003 | MODULE_LICENSE("GPL"); |