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" | |
49 | #include "xfs_version.h" | |
1da177e4 LT |
50 | |
51 | #include <linux/namei.h> | |
52 | #include <linux/init.h> | |
53 | #include <linux/mount.h> | |
0829c360 | 54 | #include <linux/mempool.h> |
1da177e4 | 55 | #include <linux/writeback.h> |
4df08c52 | 56 | #include <linux/kthread.h> |
7dfb7103 | 57 | #include <linux/freezer.h> |
1da177e4 | 58 | |
7989cb8e DC |
59 | static struct quotactl_ops xfs_quotactl_operations; |
60 | static struct super_operations xfs_super_operations; | |
61 | static kmem_zone_t *xfs_vnode_zone; | |
62 | static kmem_zone_t *xfs_ioend_zone; | |
0829c360 | 63 | mempool_t *xfs_ioend_pool; |
1da177e4 LT |
64 | |
65 | STATIC struct xfs_mount_args * | |
66 | xfs_args_allocate( | |
764d1f89 NS |
67 | struct super_block *sb, |
68 | int silent) | |
1da177e4 LT |
69 | { |
70 | struct xfs_mount_args *args; | |
71 | ||
72 | args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP); | |
73 | args->logbufs = args->logbufsize = -1; | |
74 | strncpy(args->fsname, sb->s_id, MAXNAMELEN); | |
75 | ||
76 | /* Copy the already-parsed mount(2) flags we're interested in */ | |
1da177e4 LT |
77 | if (sb->s_flags & MS_DIRSYNC) |
78 | args->flags |= XFSMNT_DIRSYNC; | |
79 | if (sb->s_flags & MS_SYNCHRONOUS) | |
80 | args->flags |= XFSMNT_WSYNC; | |
764d1f89 NS |
81 | if (silent) |
82 | args->flags |= XFSMNT_QUIET; | |
1da177e4 LT |
83 | args->flags |= XFSMNT_32BITINODES; |
84 | ||
85 | return args; | |
86 | } | |
87 | ||
88 | __uint64_t | |
89 | xfs_max_file_offset( | |
90 | unsigned int blockshift) | |
91 | { | |
92 | unsigned int pagefactor = 1; | |
93 | unsigned int bitshift = BITS_PER_LONG - 1; | |
94 | ||
95 | /* Figure out maximum filesize, on Linux this can depend on | |
96 | * the filesystem blocksize (on 32 bit platforms). | |
97 | * __block_prepare_write does this in an [unsigned] long... | |
98 | * page->index << (PAGE_CACHE_SHIFT - bbits) | |
99 | * So, for page sized blocks (4K on 32 bit platforms), | |
100 | * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is | |
101 | * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1) | |
102 | * but for smaller blocksizes it is less (bbits = log2 bsize). | |
103 | * Note1: get_block_t takes a long (implicit cast from above) | |
104 | * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch | |
105 | * can optionally convert the [unsigned] long from above into | |
106 | * an [unsigned] long long. | |
107 | */ | |
108 | ||
109 | #if BITS_PER_LONG == 32 | |
110 | # if defined(CONFIG_LBD) | |
111 | ASSERT(sizeof(sector_t) == 8); | |
112 | pagefactor = PAGE_CACHE_SIZE; | |
113 | bitshift = BITS_PER_LONG; | |
114 | # else | |
115 | pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift); | |
116 | # endif | |
117 | #endif | |
118 | ||
119 | return (((__uint64_t)pagefactor) << bitshift) - 1; | |
120 | } | |
121 | ||
7989cb8e | 122 | STATIC_INLINE void |
1da177e4 LT |
123 | xfs_set_inodeops( |
124 | struct inode *inode) | |
125 | { | |
0432dab2 CH |
126 | switch (inode->i_mode & S_IFMT) { |
127 | case S_IFREG: | |
416c6d5b | 128 | inode->i_op = &xfs_inode_operations; |
3562fd45 | 129 | inode->i_fop = &xfs_file_operations; |
e4c573bb | 130 | inode->i_mapping->a_ops = &xfs_address_space_operations; |
0432dab2 CH |
131 | break; |
132 | case S_IFDIR: | |
416c6d5b | 133 | inode->i_op = &xfs_dir_inode_operations; |
3562fd45 | 134 | inode->i_fop = &xfs_dir_file_operations; |
0432dab2 CH |
135 | break; |
136 | case S_IFLNK: | |
416c6d5b | 137 | inode->i_op = &xfs_symlink_inode_operations; |
1da177e4 | 138 | if (inode->i_blocks) |
e4c573bb | 139 | inode->i_mapping->a_ops = &xfs_address_space_operations; |
0432dab2 CH |
140 | break; |
141 | default: | |
416c6d5b | 142 | inode->i_op = &xfs_inode_operations; |
1da177e4 | 143 | init_special_inode(inode, inode->i_mode, inode->i_rdev); |
0432dab2 | 144 | break; |
1da177e4 LT |
145 | } |
146 | } | |
147 | ||
7989cb8e | 148 | STATIC_INLINE void |
1da177e4 LT |
149 | xfs_revalidate_inode( |
150 | xfs_mount_t *mp, | |
67fcaa73 | 151 | bhv_vnode_t *vp, |
1da177e4 LT |
152 | xfs_inode_t *ip) |
153 | { | |
ec86dc02 | 154 | struct inode *inode = vn_to_inode(vp); |
1da177e4 | 155 | |
0432dab2 | 156 | inode->i_mode = ip->i_d.di_mode; |
1da177e4 LT |
157 | inode->i_nlink = ip->i_d.di_nlink; |
158 | inode->i_uid = ip->i_d.di_uid; | |
159 | inode->i_gid = ip->i_d.di_gid; | |
0432dab2 CH |
160 | |
161 | switch (inode->i_mode & S_IFMT) { | |
162 | case S_IFBLK: | |
163 | case S_IFCHR: | |
164 | inode->i_rdev = | |
165 | MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff, | |
166 | sysv_minor(ip->i_df.if_u2.if_rdev)); | |
167 | break; | |
168 | default: | |
1da177e4 | 169 | inode->i_rdev = 0; |
0432dab2 | 170 | break; |
1da177e4 | 171 | } |
0432dab2 | 172 | |
1da177e4 LT |
173 | inode->i_generation = ip->i_d.di_gen; |
174 | i_size_write(inode, ip->i_d.di_size); | |
175 | inode->i_blocks = | |
176 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
177 | inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec; | |
178 | inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec; | |
179 | inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec; | |
180 | inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec; | |
181 | inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec; | |
182 | inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec; | |
183 | if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE) | |
184 | inode->i_flags |= S_IMMUTABLE; | |
185 | else | |
186 | inode->i_flags &= ~S_IMMUTABLE; | |
187 | if (ip->i_d.di_flags & XFS_DIFLAG_APPEND) | |
188 | inode->i_flags |= S_APPEND; | |
189 | else | |
190 | inode->i_flags &= ~S_APPEND; | |
191 | if (ip->i_d.di_flags & XFS_DIFLAG_SYNC) | |
192 | inode->i_flags |= S_SYNC; | |
193 | else | |
194 | inode->i_flags &= ~S_SYNC; | |
195 | if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME) | |
196 | inode->i_flags |= S_NOATIME; | |
197 | else | |
198 | inode->i_flags &= ~S_NOATIME; | |
199 | vp->v_flag &= ~VMODIFIED; | |
200 | } | |
201 | ||
202 | void | |
203 | xfs_initialize_vnode( | |
204 | bhv_desc_t *bdp, | |
67fcaa73 | 205 | bhv_vnode_t *vp, |
1da177e4 LT |
206 | bhv_desc_t *inode_bhv, |
207 | int unlock) | |
208 | { | |
209 | xfs_inode_t *ip = XFS_BHVTOI(inode_bhv); | |
ec86dc02 | 210 | struct inode *inode = vn_to_inode(vp); |
1da177e4 LT |
211 | |
212 | if (!inode_bhv->bd_vobj) { | |
213 | vp->v_vfsp = bhvtovfs(bdp); | |
214 | bhv_desc_init(inode_bhv, ip, vp, &xfs_vnodeops); | |
215 | bhv_insert(VN_BHV_HEAD(vp), inode_bhv); | |
216 | } | |
217 | ||
218 | /* | |
219 | * We need to set the ops vectors, and unlock the inode, but if | |
220 | * we have been called during the new inode create process, it is | |
221 | * too early to fill in the Linux inode. We will get called a | |
222 | * second time once the inode is properly set up, and then we can | |
223 | * finish our work. | |
224 | */ | |
225 | if (ip->i_d.di_mode != 0 && unlock && (inode->i_state & I_NEW)) { | |
1da177e4 LT |
226 | xfs_revalidate_inode(XFS_BHVTOM(bdp), vp, ip); |
227 | xfs_set_inodeops(inode); | |
ec86dc02 | 228 | |
7a18c386 | 229 | xfs_iflags_clear(ip, XFS_INEW); |
1da177e4 LT |
230 | barrier(); |
231 | ||
232 | unlock_new_inode(inode); | |
233 | } | |
234 | } | |
235 | ||
236 | int | |
237 | xfs_blkdev_get( | |
238 | xfs_mount_t *mp, | |
239 | const char *name, | |
240 | struct block_device **bdevp) | |
241 | { | |
242 | int error = 0; | |
243 | ||
244 | *bdevp = open_bdev_excl(name, 0, mp); | |
245 | if (IS_ERR(*bdevp)) { | |
246 | error = PTR_ERR(*bdevp); | |
247 | printk("XFS: Invalid device [%s], error=%d\n", name, error); | |
248 | } | |
249 | ||
250 | return -error; | |
251 | } | |
252 | ||
253 | void | |
254 | xfs_blkdev_put( | |
255 | struct block_device *bdev) | |
256 | { | |
257 | if (bdev) | |
258 | close_bdev_excl(bdev); | |
259 | } | |
260 | ||
f538d4da CH |
261 | /* |
262 | * Try to write out the superblock using barriers. | |
263 | */ | |
264 | STATIC int | |
265 | xfs_barrier_test( | |
266 | xfs_mount_t *mp) | |
267 | { | |
268 | xfs_buf_t *sbp = xfs_getsb(mp, 0); | |
269 | int error; | |
270 | ||
271 | XFS_BUF_UNDONE(sbp); | |
272 | XFS_BUF_UNREAD(sbp); | |
273 | XFS_BUF_UNDELAYWRITE(sbp); | |
274 | XFS_BUF_WRITE(sbp); | |
275 | XFS_BUF_UNASYNC(sbp); | |
276 | XFS_BUF_ORDERED(sbp); | |
277 | ||
278 | xfsbdstrat(mp, sbp); | |
279 | error = xfs_iowait(sbp); | |
280 | ||
281 | /* | |
282 | * Clear all the flags we set and possible error state in the | |
283 | * buffer. We only did the write to try out whether barriers | |
284 | * worked and shouldn't leave any traces in the superblock | |
285 | * buffer. | |
286 | */ | |
287 | XFS_BUF_DONE(sbp); | |
288 | XFS_BUF_ERROR(sbp, 0); | |
289 | XFS_BUF_UNORDERED(sbp); | |
290 | ||
291 | xfs_buf_relse(sbp); | |
292 | return error; | |
293 | } | |
294 | ||
295 | void | |
296 | xfs_mountfs_check_barriers(xfs_mount_t *mp) | |
297 | { | |
298 | int error; | |
299 | ||
300 | if (mp->m_logdev_targp != mp->m_ddev_targp) { | |
301 | xfs_fs_cmn_err(CE_NOTE, mp, | |
302 | "Disabling barriers, not supported with external log device"); | |
303 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 304 | return; |
f538d4da CH |
305 | } |
306 | ||
b2ea401b NS |
307 | if (xfs_readonly_buftarg(mp->m_ddev_targp)) { |
308 | xfs_fs_cmn_err(CE_NOTE, mp, | |
309 | "Disabling barriers, underlying device is readonly"); | |
310 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
311 | return; | |
312 | } | |
313 | ||
f538d4da CH |
314 | error = xfs_barrier_test(mp); |
315 | if (error) { | |
316 | xfs_fs_cmn_err(CE_NOTE, mp, | |
317 | "Disabling barriers, trial barrier write failed"); | |
318 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 319 | return; |
f538d4da CH |
320 | } |
321 | } | |
322 | ||
323 | void | |
324 | xfs_blkdev_issue_flush( | |
325 | xfs_buftarg_t *buftarg) | |
326 | { | |
ce8e922c | 327 | blkdev_issue_flush(buftarg->bt_bdev, NULL); |
f538d4da | 328 | } |
1da177e4 LT |
329 | |
330 | STATIC struct inode * | |
a50cd269 | 331 | xfs_fs_alloc_inode( |
1da177e4 LT |
332 | struct super_block *sb) |
333 | { | |
67fcaa73 | 334 | bhv_vnode_t *vp; |
1da177e4 | 335 | |
8758280f NS |
336 | vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP); |
337 | if (unlikely(!vp)) | |
1da177e4 | 338 | return NULL; |
ec86dc02 | 339 | return vn_to_inode(vp); |
1da177e4 LT |
340 | } |
341 | ||
342 | STATIC void | |
a50cd269 | 343 | xfs_fs_destroy_inode( |
1da177e4 LT |
344 | struct inode *inode) |
345 | { | |
ec86dc02 | 346 | kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode)); |
1da177e4 LT |
347 | } |
348 | ||
349 | STATIC void | |
a50cd269 | 350 | xfs_fs_inode_init_once( |
8758280f NS |
351 | void *vnode, |
352 | kmem_zone_t *zonep, | |
1da177e4 LT |
353 | unsigned long flags) |
354 | { | |
a35afb83 | 355 | inode_init_once(vn_to_inode((bhv_vnode_t *)vnode)); |
1da177e4 LT |
356 | } |
357 | ||
358 | STATIC int | |
8758280f | 359 | xfs_init_zones(void) |
1da177e4 | 360 | { |
67fcaa73 | 361 | xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode", |
e0cc2325 NS |
362 | KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | |
363 | KM_ZONE_SPREAD, | |
a50cd269 | 364 | xfs_fs_inode_init_once); |
0829c360 CH |
365 | if (!xfs_vnode_zone) |
366 | goto out; | |
367 | ||
368 | xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend"); | |
369 | if (!xfs_ioend_zone) | |
370 | goto out_destroy_vnode_zone; | |
371 | ||
93d2341c MD |
372 | xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE, |
373 | xfs_ioend_zone); | |
0829c360 CH |
374 | if (!xfs_ioend_pool) |
375 | goto out_free_ioend_zone; | |
1da177e4 | 376 | return 0; |
0829c360 | 377 | |
0829c360 CH |
378 | out_free_ioend_zone: |
379 | kmem_zone_destroy(xfs_ioend_zone); | |
380 | out_destroy_vnode_zone: | |
381 | kmem_zone_destroy(xfs_vnode_zone); | |
382 | out: | |
383 | return -ENOMEM; | |
1da177e4 LT |
384 | } |
385 | ||
386 | STATIC void | |
8758280f | 387 | xfs_destroy_zones(void) |
1da177e4 | 388 | { |
0829c360 CH |
389 | mempool_destroy(xfs_ioend_pool); |
390 | kmem_zone_destroy(xfs_vnode_zone); | |
391 | kmem_zone_destroy(xfs_ioend_zone); | |
1da177e4 LT |
392 | } |
393 | ||
394 | /* | |
395 | * Attempt to flush the inode, this will actually fail | |
396 | * if the inode is pinned, but we dirty the inode again | |
397 | * at the point when it is unpinned after a log write, | |
8758280f | 398 | * since this is when the inode itself becomes flushable. |
1da177e4 LT |
399 | */ |
400 | STATIC int | |
a50cd269 | 401 | xfs_fs_write_inode( |
1da177e4 LT |
402 | struct inode *inode, |
403 | int sync) | |
404 | { | |
67fcaa73 | 405 | bhv_vnode_t *vp = vn_from_inode(inode); |
1da177e4 LT |
406 | int error = 0, flags = FLUSH_INODE; |
407 | ||
408 | if (vp) { | |
409 | vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); | |
776a75fa LM |
410 | if (sync) { |
411 | filemap_fdatawait(inode->i_mapping); | |
1da177e4 | 412 | flags |= FLUSH_SYNC; |
776a75fa | 413 | } |
67fcaa73 NS |
414 | error = bhv_vop_iflush(vp, flags); |
415 | if (error == EAGAIN) | |
416 | error = sync? bhv_vop_iflush(vp, flags | FLUSH_LOG) : 0; | |
1da177e4 | 417 | } |
1da177e4 LT |
418 | return -error; |
419 | } | |
420 | ||
421 | STATIC void | |
a50cd269 | 422 | xfs_fs_clear_inode( |
1da177e4 LT |
423 | struct inode *inode) |
424 | { | |
67fcaa73 | 425 | bhv_vnode_t *vp = vn_from_inode(inode); |
1da177e4 | 426 | |
a50cd269 | 427 | vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); |
56d433e4 | 428 | |
56d433e4 CH |
429 | XFS_STATS_INC(vn_rele); |
430 | XFS_STATS_INC(vn_remove); | |
431 | XFS_STATS_INC(vn_reclaim); | |
432 | XFS_STATS_DEC(vn_active); | |
433 | ||
02ba71de CH |
434 | /* |
435 | * This can happen because xfs_iget_core calls xfs_idestroy if we | |
436 | * find an inode with di_mode == 0 but without IGET_CREATE set. | |
437 | */ | |
67fcaa73 NS |
438 | if (VNHEAD(vp)) |
439 | bhv_vop_inactive(vp, NULL); | |
1da177e4 | 440 | |
56d433e4 CH |
441 | VN_LOCK(vp); |
442 | vp->v_flag &= ~VMODIFIED; | |
443 | VN_UNLOCK(vp, 0); | |
444 | ||
67fcaa73 NS |
445 | if (VNHEAD(vp)) |
446 | if (bhv_vop_reclaim(vp)) | |
447 | panic("%s: cannot reclaim 0x%p\n", __FUNCTION__, vp); | |
56d433e4 | 448 | |
67fcaa73 | 449 | ASSERT(VNHEAD(vp) == NULL); |
56d433e4 CH |
450 | |
451 | #ifdef XFS_VNODE_TRACE | |
452 | ktrace_free(vp->v_trace); | |
453 | #endif | |
454 | } | |
1da177e4 LT |
455 | |
456 | /* | |
457 | * Enqueue a work item to be picked up by the vfs xfssyncd thread. | |
458 | * Doing this has two advantages: | |
459 | * - It saves on stack space, which is tight in certain situations | |
460 | * - It can be used (with care) as a mechanism to avoid deadlocks. | |
461 | * Flushing while allocating in a full filesystem requires both. | |
462 | */ | |
463 | STATIC void | |
464 | xfs_syncd_queue_work( | |
b83bd138 | 465 | struct bhv_vfs *vfs, |
1da177e4 | 466 | void *data, |
b83bd138 | 467 | void (*syncer)(bhv_vfs_t *, void *)) |
1da177e4 | 468 | { |
b83bd138 | 469 | struct bhv_vfs_sync_work *work; |
1da177e4 | 470 | |
b83bd138 | 471 | work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP); |
1da177e4 LT |
472 | INIT_LIST_HEAD(&work->w_list); |
473 | work->w_syncer = syncer; | |
474 | work->w_data = data; | |
475 | work->w_vfs = vfs; | |
476 | spin_lock(&vfs->vfs_sync_lock); | |
477 | list_add_tail(&work->w_list, &vfs->vfs_sync_list); | |
478 | spin_unlock(&vfs->vfs_sync_lock); | |
479 | wake_up_process(vfs->vfs_sync_task); | |
480 | } | |
481 | ||
482 | /* | |
483 | * Flush delayed allocate data, attempting to free up reserved space | |
484 | * from existing allocations. At this point a new allocation attempt | |
485 | * has failed with ENOSPC and we are in the process of scratching our | |
486 | * heads, looking about for more room... | |
487 | */ | |
488 | STATIC void | |
489 | xfs_flush_inode_work( | |
b83bd138 | 490 | bhv_vfs_t *vfs, |
1da177e4 LT |
491 | void *inode) |
492 | { | |
493 | filemap_flush(((struct inode *)inode)->i_mapping); | |
494 | iput((struct inode *)inode); | |
495 | } | |
496 | ||
497 | void | |
498 | xfs_flush_inode( | |
499 | xfs_inode_t *ip) | |
500 | { | |
ec86dc02 | 501 | struct inode *inode = vn_to_inode(XFS_ITOV(ip)); |
b83bd138 | 502 | struct bhv_vfs *vfs = XFS_MTOVFS(ip->i_mount); |
1da177e4 LT |
503 | |
504 | igrab(inode); | |
505 | xfs_syncd_queue_work(vfs, inode, xfs_flush_inode_work); | |
041e0e3b | 506 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
507 | } |
508 | ||
509 | /* | |
510 | * This is the "bigger hammer" version of xfs_flush_inode_work... | |
511 | * (IOW, "If at first you don't succeed, use a Bigger Hammer"). | |
512 | */ | |
513 | STATIC void | |
514 | xfs_flush_device_work( | |
b83bd138 | 515 | bhv_vfs_t *vfs, |
1da177e4 LT |
516 | void *inode) |
517 | { | |
518 | sync_blockdev(vfs->vfs_super->s_bdev); | |
519 | iput((struct inode *)inode); | |
520 | } | |
521 | ||
522 | void | |
523 | xfs_flush_device( | |
524 | xfs_inode_t *ip) | |
525 | { | |
ec86dc02 | 526 | struct inode *inode = vn_to_inode(XFS_ITOV(ip)); |
b83bd138 | 527 | struct bhv_vfs *vfs = XFS_MTOVFS(ip->i_mount); |
1da177e4 LT |
528 | |
529 | igrab(inode); | |
530 | xfs_syncd_queue_work(vfs, inode, xfs_flush_device_work); | |
041e0e3b | 531 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
532 | xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC); |
533 | } | |
534 | ||
1da177e4 LT |
535 | STATIC void |
536 | vfs_sync_worker( | |
b83bd138 | 537 | bhv_vfs_t *vfsp, |
1da177e4 LT |
538 | void *unused) |
539 | { | |
540 | int error; | |
541 | ||
542 | if (!(vfsp->vfs_flag & VFS_RDONLY)) | |
b83bd138 | 543 | error = bhv_vfs_sync(vfsp, SYNC_FSDATA | SYNC_BDFLUSH | \ |
92821e2b DC |
544 | SYNC_ATTR | SYNC_REFCACHE | SYNC_SUPER, |
545 | NULL); | |
1da177e4 | 546 | vfsp->vfs_sync_seq++; |
1da177e4 LT |
547 | wake_up(&vfsp->vfs_wait_single_sync_task); |
548 | } | |
549 | ||
550 | STATIC int | |
551 | xfssyncd( | |
552 | void *arg) | |
553 | { | |
554 | long timeleft; | |
b83bd138 NS |
555 | bhv_vfs_t *vfsp = (bhv_vfs_t *) arg; |
556 | bhv_vfs_sync_work_t *work, *n; | |
4df08c52 | 557 | LIST_HEAD (tmp); |
1da177e4 | 558 | |
83144186 | 559 | set_freezable(); |
041e0e3b | 560 | timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10); |
1da177e4 | 561 | for (;;) { |
041e0e3b | 562 | timeleft = schedule_timeout_interruptible(timeleft); |
1da177e4 | 563 | /* swsusp */ |
3e1d1d28 | 564 | try_to_freeze(); |
71df099d | 565 | if (kthread_should_stop() && list_empty(&vfsp->vfs_sync_list)) |
1da177e4 LT |
566 | break; |
567 | ||
568 | spin_lock(&vfsp->vfs_sync_lock); | |
569 | /* | |
570 | * We can get woken by laptop mode, to do a sync - | |
571 | * that's the (only!) case where the list would be | |
572 | * empty with time remaining. | |
573 | */ | |
574 | if (!timeleft || list_empty(&vfsp->vfs_sync_list)) { | |
575 | if (!timeleft) | |
041e0e3b NA |
576 | timeleft = xfs_syncd_centisecs * |
577 | msecs_to_jiffies(10); | |
1da177e4 LT |
578 | INIT_LIST_HEAD(&vfsp->vfs_sync_work.w_list); |
579 | list_add_tail(&vfsp->vfs_sync_work.w_list, | |
580 | &vfsp->vfs_sync_list); | |
581 | } | |
582 | list_for_each_entry_safe(work, n, &vfsp->vfs_sync_list, w_list) | |
583 | list_move(&work->w_list, &tmp); | |
584 | spin_unlock(&vfsp->vfs_sync_lock); | |
585 | ||
586 | list_for_each_entry_safe(work, n, &tmp, w_list) { | |
587 | (*work->w_syncer)(vfsp, work->w_data); | |
588 | list_del(&work->w_list); | |
589 | if (work == &vfsp->vfs_sync_work) | |
590 | continue; | |
b83bd138 | 591 | kmem_free(work, sizeof(struct bhv_vfs_sync_work)); |
1da177e4 LT |
592 | } |
593 | } | |
594 | ||
1da177e4 LT |
595 | return 0; |
596 | } | |
597 | ||
598 | STATIC int | |
a50cd269 | 599 | xfs_fs_start_syncd( |
b83bd138 | 600 | bhv_vfs_t *vfsp) |
1da177e4 | 601 | { |
4df08c52 CH |
602 | vfsp->vfs_sync_work.w_syncer = vfs_sync_worker; |
603 | vfsp->vfs_sync_work.w_vfs = vfsp; | |
604 | vfsp->vfs_sync_task = kthread_run(xfssyncd, vfsp, "xfssyncd"); | |
605 | if (IS_ERR(vfsp->vfs_sync_task)) | |
606 | return -PTR_ERR(vfsp->vfs_sync_task); | |
1da177e4 LT |
607 | return 0; |
608 | } | |
609 | ||
610 | STATIC void | |
a50cd269 | 611 | xfs_fs_stop_syncd( |
b83bd138 | 612 | bhv_vfs_t *vfsp) |
1da177e4 | 613 | { |
4df08c52 | 614 | kthread_stop(vfsp->vfs_sync_task); |
1da177e4 LT |
615 | } |
616 | ||
617 | STATIC void | |
a50cd269 | 618 | xfs_fs_put_super( |
1da177e4 LT |
619 | struct super_block *sb) |
620 | { | |
b83bd138 | 621 | bhv_vfs_t *vfsp = vfs_from_sb(sb); |
1da177e4 LT |
622 | int error; |
623 | ||
a50cd269 | 624 | xfs_fs_stop_syncd(vfsp); |
b83bd138 NS |
625 | bhv_vfs_sync(vfsp, SYNC_ATTR | SYNC_DELWRI, NULL); |
626 | error = bhv_vfs_unmount(vfsp, 0, NULL); | |
1da177e4 | 627 | if (error) { |
b83bd138 NS |
628 | printk("XFS: unmount got error=%d\n", error); |
629 | printk("%s: vfs=0x%p left dangling!\n", __FUNCTION__, vfsp); | |
630 | } else { | |
631 | vfs_deallocate(vfsp); | |
1da177e4 | 632 | } |
1da177e4 LT |
633 | } |
634 | ||
635 | STATIC void | |
a50cd269 | 636 | xfs_fs_write_super( |
1da177e4 LT |
637 | struct super_block *sb) |
638 | { | |
b83bd138 NS |
639 | if (!(sb->s_flags & MS_RDONLY)) |
640 | bhv_vfs_sync(vfs_from_sb(sb), SYNC_FSDATA, NULL); | |
1da177e4 LT |
641 | sb->s_dirt = 0; |
642 | } | |
643 | ||
644 | STATIC int | |
a50cd269 | 645 | xfs_fs_sync_super( |
1da177e4 LT |
646 | struct super_block *sb, |
647 | int wait) | |
648 | { | |
b83bd138 NS |
649 | bhv_vfs_t *vfsp = vfs_from_sb(sb); |
650 | int error; | |
651 | int flags; | |
1da177e4 | 652 | |
2823945f DC |
653 | if (unlikely(sb->s_frozen == SB_FREEZE_WRITE)) { |
654 | /* | |
655 | * First stage of freeze - no more writers will make progress | |
656 | * now we are here, so we flush delwri and delalloc buffers | |
657 | * here, then wait for all I/O to complete. Data is frozen at | |
658 | * that point. Metadata is not frozen, transactions can still | |
659 | * occur here so don't bother flushing the buftarg (i.e | |
660 | * SYNC_QUIESCE) because it'll just get dirty again. | |
661 | */ | |
516b2e7c | 662 | flags = SYNC_DATA_QUIESCE; |
2823945f | 663 | } else |
f898d6c0 | 664 | flags = SYNC_FSDATA | (wait ? SYNC_WAIT : 0); |
1da177e4 | 665 | |
b83bd138 | 666 | error = bhv_vfs_sync(vfsp, flags, NULL); |
1da177e4 LT |
667 | sb->s_dirt = 0; |
668 | ||
669 | if (unlikely(laptop_mode)) { | |
670 | int prev_sync_seq = vfsp->vfs_sync_seq; | |
671 | ||
672 | /* | |
673 | * The disk must be active because we're syncing. | |
674 | * We schedule xfssyncd now (now that the disk is | |
675 | * active) instead of later (when it might not be). | |
676 | */ | |
677 | wake_up_process(vfsp->vfs_sync_task); | |
678 | /* | |
679 | * We have to wait for the sync iteration to complete. | |
680 | * If we don't, the disk activity caused by the sync | |
681 | * will come after the sync is completed, and that | |
682 | * triggers another sync from laptop mode. | |
683 | */ | |
684 | wait_event(vfsp->vfs_wait_single_sync_task, | |
685 | vfsp->vfs_sync_seq != prev_sync_seq); | |
686 | } | |
687 | ||
688 | return -error; | |
689 | } | |
690 | ||
691 | STATIC int | |
a50cd269 | 692 | xfs_fs_statfs( |
726c3342 | 693 | struct dentry *dentry, |
1da177e4 LT |
694 | struct kstatfs *statp) |
695 | { | |
d6938d1b DH |
696 | return -bhv_vfs_statvfs(vfs_from_sb(dentry->d_sb), statp, |
697 | vn_from_inode(dentry->d_inode)); | |
1da177e4 LT |
698 | } |
699 | ||
700 | STATIC int | |
a50cd269 | 701 | xfs_fs_remount( |
1da177e4 LT |
702 | struct super_block *sb, |
703 | int *flags, | |
704 | char *options) | |
705 | { | |
b83bd138 | 706 | bhv_vfs_t *vfsp = vfs_from_sb(sb); |
764d1f89 | 707 | struct xfs_mount_args *args = xfs_args_allocate(sb, 0); |
1da177e4 LT |
708 | int error; |
709 | ||
b83bd138 | 710 | error = bhv_vfs_parseargs(vfsp, options, args, 1); |
1da177e4 | 711 | if (!error) |
b83bd138 | 712 | error = bhv_vfs_mntupdate(vfsp, flags, args); |
1da177e4 LT |
713 | kmem_free(args, sizeof(*args)); |
714 | return -error; | |
715 | } | |
716 | ||
717 | STATIC void | |
a50cd269 | 718 | xfs_fs_lockfs( |
1da177e4 LT |
719 | struct super_block *sb) |
720 | { | |
b83bd138 | 721 | bhv_vfs_freeze(vfs_from_sb(sb)); |
1da177e4 LT |
722 | } |
723 | ||
724 | STATIC int | |
a50cd269 | 725 | xfs_fs_show_options( |
1da177e4 LT |
726 | struct seq_file *m, |
727 | struct vfsmount *mnt) | |
728 | { | |
b83bd138 | 729 | return -bhv_vfs_showargs(vfs_from_sb(mnt->mnt_sb), m); |
1da177e4 LT |
730 | } |
731 | ||
ee34807a | 732 | STATIC int |
a50cd269 | 733 | xfs_fs_quotasync( |
ee34807a NS |
734 | struct super_block *sb, |
735 | int type) | |
736 | { | |
b83bd138 | 737 | return -bhv_vfs_quotactl(vfs_from_sb(sb), Q_XQUOTASYNC, 0, NULL); |
ee34807a NS |
738 | } |
739 | ||
1da177e4 | 740 | STATIC int |
a50cd269 | 741 | xfs_fs_getxstate( |
1da177e4 LT |
742 | struct super_block *sb, |
743 | struct fs_quota_stat *fqs) | |
744 | { | |
b83bd138 | 745 | return -bhv_vfs_quotactl(vfs_from_sb(sb), Q_XGETQSTAT, 0, (caddr_t)fqs); |
1da177e4 LT |
746 | } |
747 | ||
748 | STATIC int | |
a50cd269 | 749 | xfs_fs_setxstate( |
1da177e4 LT |
750 | struct super_block *sb, |
751 | unsigned int flags, | |
752 | int op) | |
753 | { | |
b83bd138 | 754 | return -bhv_vfs_quotactl(vfs_from_sb(sb), op, 0, (caddr_t)&flags); |
1da177e4 LT |
755 | } |
756 | ||
757 | STATIC int | |
a50cd269 | 758 | xfs_fs_getxquota( |
1da177e4 LT |
759 | struct super_block *sb, |
760 | int type, | |
761 | qid_t id, | |
762 | struct fs_disk_quota *fdq) | |
763 | { | |
b83bd138 NS |
764 | return -bhv_vfs_quotactl(vfs_from_sb(sb), |
765 | (type == USRQUOTA) ? Q_XGETQUOTA : | |
766 | ((type == GRPQUOTA) ? Q_XGETGQUOTA : | |
767 | Q_XGETPQUOTA), id, (caddr_t)fdq); | |
1da177e4 LT |
768 | } |
769 | ||
770 | STATIC int | |
a50cd269 | 771 | xfs_fs_setxquota( |
1da177e4 LT |
772 | struct super_block *sb, |
773 | int type, | |
774 | qid_t id, | |
775 | struct fs_disk_quota *fdq) | |
776 | { | |
b83bd138 NS |
777 | return -bhv_vfs_quotactl(vfs_from_sb(sb), |
778 | (type == USRQUOTA) ? Q_XSETQLIM : | |
779 | ((type == GRPQUOTA) ? Q_XSETGQLIM : | |
780 | Q_XSETPQLIM), id, (caddr_t)fdq); | |
1da177e4 LT |
781 | } |
782 | ||
783 | STATIC int | |
a50cd269 | 784 | xfs_fs_fill_super( |
1da177e4 LT |
785 | struct super_block *sb, |
786 | void *data, | |
787 | int silent) | |
788 | { | |
67fcaa73 | 789 | struct bhv_vnode *rootvp; |
b83bd138 | 790 | struct bhv_vfs *vfsp = vfs_allocate(sb); |
764d1f89 | 791 | struct xfs_mount_args *args = xfs_args_allocate(sb, silent); |
1da177e4 | 792 | struct kstatfs statvfs; |
b83bd138 | 793 | int error; |
1da177e4 | 794 | |
1da177e4 LT |
795 | bhv_insert_all_vfsops(vfsp); |
796 | ||
b83bd138 | 797 | error = bhv_vfs_parseargs(vfsp, (char *)data, args, 0); |
1da177e4 LT |
798 | if (error) { |
799 | bhv_remove_all_vfsops(vfsp, 1); | |
800 | goto fail_vfsop; | |
801 | } | |
802 | ||
803 | sb_min_blocksize(sb, BBSIZE); | |
a50cd269 | 804 | sb->s_export_op = &xfs_export_operations; |
a50cd269 NS |
805 | sb->s_qcop = &xfs_quotactl_operations; |
806 | sb->s_op = &xfs_super_operations; | |
1da177e4 | 807 | |
b83bd138 | 808 | error = bhv_vfs_mount(vfsp, args, NULL); |
1da177e4 LT |
809 | if (error) { |
810 | bhv_remove_all_vfsops(vfsp, 1); | |
811 | goto fail_vfsop; | |
812 | } | |
813 | ||
b83bd138 | 814 | error = bhv_vfs_statvfs(vfsp, &statvfs, NULL); |
1da177e4 LT |
815 | if (error) |
816 | goto fail_unmount; | |
817 | ||
818 | sb->s_dirt = 1; | |
819 | sb->s_magic = statvfs.f_type; | |
820 | sb->s_blocksize = statvfs.f_bsize; | |
821 | sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1; | |
822 | sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); | |
823 | sb->s_time_gran = 1; | |
824 | set_posix_acl_flag(sb); | |
825 | ||
b83bd138 | 826 | error = bhv_vfs_root(vfsp, &rootvp); |
1da177e4 LT |
827 | if (error) |
828 | goto fail_unmount; | |
829 | ||
ec86dc02 | 830 | sb->s_root = d_alloc_root(vn_to_inode(rootvp)); |
1da177e4 LT |
831 | if (!sb->s_root) { |
832 | error = ENOMEM; | |
833 | goto fail_vnrele; | |
834 | } | |
835 | if (is_bad_inode(sb->s_root->d_inode)) { | |
836 | error = EINVAL; | |
837 | goto fail_vnrele; | |
838 | } | |
a50cd269 | 839 | if ((error = xfs_fs_start_syncd(vfsp))) |
1da177e4 LT |
840 | goto fail_vnrele; |
841 | vn_trace_exit(rootvp, __FUNCTION__, (inst_t *)__return_address); | |
842 | ||
843 | kmem_free(args, sizeof(*args)); | |
844 | return 0; | |
845 | ||
846 | fail_vnrele: | |
847 | if (sb->s_root) { | |
848 | dput(sb->s_root); | |
849 | sb->s_root = NULL; | |
850 | } else { | |
851 | VN_RELE(rootvp); | |
852 | } | |
853 | ||
854 | fail_unmount: | |
b83bd138 | 855 | bhv_vfs_unmount(vfsp, 0, NULL); |
1da177e4 LT |
856 | |
857 | fail_vfsop: | |
858 | vfs_deallocate(vfsp); | |
859 | kmem_free(args, sizeof(*args)); | |
860 | return -error; | |
861 | } | |
862 | ||
454e2398 | 863 | STATIC int |
a50cd269 | 864 | xfs_fs_get_sb( |
1da177e4 LT |
865 | struct file_system_type *fs_type, |
866 | int flags, | |
867 | const char *dev_name, | |
454e2398 DH |
868 | void *data, |
869 | struct vfsmount *mnt) | |
1da177e4 | 870 | { |
454e2398 DH |
871 | return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super, |
872 | mnt); | |
a50cd269 NS |
873 | } |
874 | ||
7989cb8e | 875 | static struct super_operations xfs_super_operations = { |
a50cd269 NS |
876 | .alloc_inode = xfs_fs_alloc_inode, |
877 | .destroy_inode = xfs_fs_destroy_inode, | |
878 | .write_inode = xfs_fs_write_inode, | |
879 | .clear_inode = xfs_fs_clear_inode, | |
880 | .put_super = xfs_fs_put_super, | |
881 | .write_super = xfs_fs_write_super, | |
882 | .sync_fs = xfs_fs_sync_super, | |
883 | .write_super_lockfs = xfs_fs_lockfs, | |
884 | .statfs = xfs_fs_statfs, | |
885 | .remount_fs = xfs_fs_remount, | |
886 | .show_options = xfs_fs_show_options, | |
1da177e4 LT |
887 | }; |
888 | ||
7989cb8e | 889 | static struct quotactl_ops xfs_quotactl_operations = { |
a50cd269 NS |
890 | .quota_sync = xfs_fs_quotasync, |
891 | .get_xstate = xfs_fs_getxstate, | |
892 | .set_xstate = xfs_fs_setxstate, | |
893 | .get_xquota = xfs_fs_getxquota, | |
894 | .set_xquota = xfs_fs_setxquota, | |
1da177e4 LT |
895 | }; |
896 | ||
5085b607 | 897 | static struct file_system_type xfs_fs_type = { |
1da177e4 LT |
898 | .owner = THIS_MODULE, |
899 | .name = "xfs", | |
a50cd269 | 900 | .get_sb = xfs_fs_get_sb, |
1da177e4 LT |
901 | .kill_sb = kill_block_super, |
902 | .fs_flags = FS_REQUIRES_DEV, | |
903 | }; | |
904 | ||
905 | ||
906 | STATIC int __init | |
907 | init_xfs_fs( void ) | |
908 | { | |
909 | int error; | |
910 | struct sysinfo si; | |
911 | static char message[] __initdata = KERN_INFO \ | |
912 | XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n"; | |
913 | ||
914 | printk(message); | |
915 | ||
916 | si_meminfo(&si); | |
917 | xfs_physmem = si.totalram; | |
918 | ||
919 | ktrace_init(64); | |
920 | ||
8758280f | 921 | error = xfs_init_zones(); |
1da177e4 | 922 | if (error < 0) |
0829c360 | 923 | goto undo_zones; |
1da177e4 | 924 | |
ce8e922c | 925 | error = xfs_buf_init(); |
1da177e4 | 926 | if (error < 0) |
ce8e922c | 927 | goto undo_buffers; |
1da177e4 LT |
928 | |
929 | vn_init(); | |
930 | xfs_init(); | |
931 | uuid_init(); | |
932 | vfs_initquota(); | |
933 | ||
934 | error = register_filesystem(&xfs_fs_type); | |
935 | if (error) | |
936 | goto undo_register; | |
1da177e4 LT |
937 | return 0; |
938 | ||
939 | undo_register: | |
ce8e922c | 940 | xfs_buf_terminate(); |
1da177e4 | 941 | |
ce8e922c | 942 | undo_buffers: |
8758280f | 943 | xfs_destroy_zones(); |
1da177e4 | 944 | |
0829c360 | 945 | undo_zones: |
1da177e4 LT |
946 | return error; |
947 | } | |
948 | ||
949 | STATIC void __exit | |
950 | exit_xfs_fs( void ) | |
951 | { | |
952 | vfs_exitquota(); | |
1da177e4 LT |
953 | unregister_filesystem(&xfs_fs_type); |
954 | xfs_cleanup(); | |
ce8e922c | 955 | xfs_buf_terminate(); |
8758280f | 956 | xfs_destroy_zones(); |
1da177e4 LT |
957 | ktrace_uninit(); |
958 | } | |
959 | ||
960 | module_init(init_xfs_fs); | |
961 | module_exit(exit_xfs_fs); | |
962 | ||
963 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
964 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
965 | MODULE_LICENSE("GPL"); |