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