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, | |
67fcaa73 | 154 | bhv_vnode_t *vp, |
1da177e4 LT |
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, | |
67fcaa73 | 209 | bhv_vnode_t *vp, |
1da177e4 LT |
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 | { | |
67fcaa73 | 339 | bhv_vnode_t *vp; |
1da177e4 | 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) |
67fcaa73 | 362 | inode_init_once(vn_to_inode((bhv_vnode_t *)vnode)); |
1da177e4 LT |
363 | } |
364 | ||
365 | STATIC int | |
8758280f | 366 | xfs_init_zones(void) |
1da177e4 | 367 | { |
67fcaa73 | 368 | xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode", |
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 | { | |
67fcaa73 | 412 | bhv_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; | |
67fcaa73 NS |
419 | error = bhv_vop_iflush(vp, flags); |
420 | if (error == EAGAIN) | |
421 | error = sync? bhv_vop_iflush(vp, flags | FLUSH_LOG) : 0; | |
1da177e4 | 422 | } |
1da177e4 LT |
423 | return -error; |
424 | } | |
425 | ||
426 | STATIC void | |
a50cd269 | 427 | xfs_fs_clear_inode( |
1da177e4 LT |
428 | struct inode *inode) |
429 | { | |
67fcaa73 | 430 | bhv_vnode_t *vp = vn_from_inode(inode); |
1da177e4 | 431 | |
a50cd269 | 432 | vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); |
56d433e4 | 433 | |
56d433e4 CH |
434 | XFS_STATS_INC(vn_rele); |
435 | XFS_STATS_INC(vn_remove); | |
436 | XFS_STATS_INC(vn_reclaim); | |
437 | XFS_STATS_DEC(vn_active); | |
438 | ||
02ba71de CH |
439 | /* |
440 | * This can happen because xfs_iget_core calls xfs_idestroy if we | |
441 | * find an inode with di_mode == 0 but without IGET_CREATE set. | |
442 | */ | |
67fcaa73 NS |
443 | if (VNHEAD(vp)) |
444 | bhv_vop_inactive(vp, NULL); | |
1da177e4 | 445 | |
56d433e4 CH |
446 | VN_LOCK(vp); |
447 | vp->v_flag &= ~VMODIFIED; | |
448 | VN_UNLOCK(vp, 0); | |
449 | ||
67fcaa73 NS |
450 | if (VNHEAD(vp)) |
451 | if (bhv_vop_reclaim(vp)) | |
452 | panic("%s: cannot reclaim 0x%p\n", __FUNCTION__, vp); | |
56d433e4 | 453 | |
67fcaa73 | 454 | ASSERT(VNHEAD(vp) == NULL); |
56d433e4 CH |
455 | |
456 | #ifdef XFS_VNODE_TRACE | |
457 | ktrace_free(vp->v_trace); | |
458 | #endif | |
459 | } | |
1da177e4 LT |
460 | |
461 | /* | |
462 | * Enqueue a work item to be picked up by the vfs xfssyncd thread. | |
463 | * Doing this has two advantages: | |
464 | * - It saves on stack space, which is tight in certain situations | |
465 | * - It can be used (with care) as a mechanism to avoid deadlocks. | |
466 | * Flushing while allocating in a full filesystem requires both. | |
467 | */ | |
468 | STATIC void | |
469 | xfs_syncd_queue_work( | |
b83bd138 | 470 | struct bhv_vfs *vfs, |
1da177e4 | 471 | void *data, |
b83bd138 | 472 | void (*syncer)(bhv_vfs_t *, void *)) |
1da177e4 | 473 | { |
b83bd138 | 474 | struct bhv_vfs_sync_work *work; |
1da177e4 | 475 | |
b83bd138 | 476 | work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP); |
1da177e4 LT |
477 | INIT_LIST_HEAD(&work->w_list); |
478 | work->w_syncer = syncer; | |
479 | work->w_data = data; | |
480 | work->w_vfs = vfs; | |
481 | spin_lock(&vfs->vfs_sync_lock); | |
482 | list_add_tail(&work->w_list, &vfs->vfs_sync_list); | |
483 | spin_unlock(&vfs->vfs_sync_lock); | |
484 | wake_up_process(vfs->vfs_sync_task); | |
485 | } | |
486 | ||
487 | /* | |
488 | * Flush delayed allocate data, attempting to free up reserved space | |
489 | * from existing allocations. At this point a new allocation attempt | |
490 | * has failed with ENOSPC and we are in the process of scratching our | |
491 | * heads, looking about for more room... | |
492 | */ | |
493 | STATIC void | |
494 | xfs_flush_inode_work( | |
b83bd138 | 495 | bhv_vfs_t *vfs, |
1da177e4 LT |
496 | void *inode) |
497 | { | |
498 | filemap_flush(((struct inode *)inode)->i_mapping); | |
499 | iput((struct inode *)inode); | |
500 | } | |
501 | ||
502 | void | |
503 | xfs_flush_inode( | |
504 | xfs_inode_t *ip) | |
505 | { | |
ec86dc02 | 506 | struct inode *inode = vn_to_inode(XFS_ITOV(ip)); |
b83bd138 | 507 | struct bhv_vfs *vfs = XFS_MTOVFS(ip->i_mount); |
1da177e4 LT |
508 | |
509 | igrab(inode); | |
510 | xfs_syncd_queue_work(vfs, inode, xfs_flush_inode_work); | |
041e0e3b | 511 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
512 | } |
513 | ||
514 | /* | |
515 | * This is the "bigger hammer" version of xfs_flush_inode_work... | |
516 | * (IOW, "If at first you don't succeed, use a Bigger Hammer"). | |
517 | */ | |
518 | STATIC void | |
519 | xfs_flush_device_work( | |
b83bd138 | 520 | bhv_vfs_t *vfs, |
1da177e4 LT |
521 | void *inode) |
522 | { | |
523 | sync_blockdev(vfs->vfs_super->s_bdev); | |
524 | iput((struct inode *)inode); | |
525 | } | |
526 | ||
527 | void | |
528 | xfs_flush_device( | |
529 | xfs_inode_t *ip) | |
530 | { | |
ec86dc02 | 531 | struct inode *inode = vn_to_inode(XFS_ITOV(ip)); |
b83bd138 | 532 | struct bhv_vfs *vfs = XFS_MTOVFS(ip->i_mount); |
1da177e4 LT |
533 | |
534 | igrab(inode); | |
535 | xfs_syncd_queue_work(vfs, inode, xfs_flush_device_work); | |
041e0e3b | 536 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
537 | xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC); |
538 | } | |
539 | ||
1da177e4 LT |
540 | STATIC void |
541 | vfs_sync_worker( | |
b83bd138 | 542 | bhv_vfs_t *vfsp, |
1da177e4 LT |
543 | void *unused) |
544 | { | |
545 | int error; | |
546 | ||
547 | if (!(vfsp->vfs_flag & VFS_RDONLY)) | |
b83bd138 NS |
548 | error = bhv_vfs_sync(vfsp, SYNC_FSDATA | SYNC_BDFLUSH | \ |
549 | SYNC_ATTR | SYNC_REFCACHE, NULL); | |
1da177e4 LT |
550 | vfsp->vfs_sync_seq++; |
551 | wmb(); | |
552 | wake_up(&vfsp->vfs_wait_single_sync_task); | |
553 | } | |
554 | ||
555 | STATIC int | |
556 | xfssyncd( | |
557 | void *arg) | |
558 | { | |
559 | long timeleft; | |
b83bd138 NS |
560 | bhv_vfs_t *vfsp = (bhv_vfs_t *) arg; |
561 | bhv_vfs_sync_work_t *work, *n; | |
4df08c52 | 562 | LIST_HEAD (tmp); |
1da177e4 | 563 | |
041e0e3b | 564 | timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10); |
1da177e4 | 565 | for (;;) { |
041e0e3b | 566 | timeleft = schedule_timeout_interruptible(timeleft); |
1da177e4 | 567 | /* swsusp */ |
3e1d1d28 | 568 | try_to_freeze(); |
71df099d | 569 | if (kthread_should_stop() && list_empty(&vfsp->vfs_sync_list)) |
1da177e4 LT |
570 | break; |
571 | ||
572 | spin_lock(&vfsp->vfs_sync_lock); | |
573 | /* | |
574 | * We can get woken by laptop mode, to do a sync - | |
575 | * that's the (only!) case where the list would be | |
576 | * empty with time remaining. | |
577 | */ | |
578 | if (!timeleft || list_empty(&vfsp->vfs_sync_list)) { | |
579 | if (!timeleft) | |
041e0e3b NA |
580 | timeleft = xfs_syncd_centisecs * |
581 | msecs_to_jiffies(10); | |
1da177e4 LT |
582 | INIT_LIST_HEAD(&vfsp->vfs_sync_work.w_list); |
583 | list_add_tail(&vfsp->vfs_sync_work.w_list, | |
584 | &vfsp->vfs_sync_list); | |
585 | } | |
586 | list_for_each_entry_safe(work, n, &vfsp->vfs_sync_list, w_list) | |
587 | list_move(&work->w_list, &tmp); | |
588 | spin_unlock(&vfsp->vfs_sync_lock); | |
589 | ||
590 | list_for_each_entry_safe(work, n, &tmp, w_list) { | |
591 | (*work->w_syncer)(vfsp, work->w_data); | |
592 | list_del(&work->w_list); | |
593 | if (work == &vfsp->vfs_sync_work) | |
594 | continue; | |
b83bd138 | 595 | kmem_free(work, sizeof(struct bhv_vfs_sync_work)); |
1da177e4 LT |
596 | } |
597 | } | |
598 | ||
1da177e4 LT |
599 | return 0; |
600 | } | |
601 | ||
602 | STATIC int | |
a50cd269 | 603 | xfs_fs_start_syncd( |
b83bd138 | 604 | bhv_vfs_t *vfsp) |
1da177e4 | 605 | { |
4df08c52 CH |
606 | vfsp->vfs_sync_work.w_syncer = vfs_sync_worker; |
607 | vfsp->vfs_sync_work.w_vfs = vfsp; | |
608 | vfsp->vfs_sync_task = kthread_run(xfssyncd, vfsp, "xfssyncd"); | |
609 | if (IS_ERR(vfsp->vfs_sync_task)) | |
610 | return -PTR_ERR(vfsp->vfs_sync_task); | |
1da177e4 LT |
611 | return 0; |
612 | } | |
613 | ||
614 | STATIC void | |
a50cd269 | 615 | xfs_fs_stop_syncd( |
b83bd138 | 616 | bhv_vfs_t *vfsp) |
1da177e4 | 617 | { |
4df08c52 | 618 | kthread_stop(vfsp->vfs_sync_task); |
1da177e4 LT |
619 | } |
620 | ||
621 | STATIC void | |
a50cd269 | 622 | xfs_fs_put_super( |
1da177e4 LT |
623 | struct super_block *sb) |
624 | { | |
b83bd138 | 625 | bhv_vfs_t *vfsp = vfs_from_sb(sb); |
1da177e4 LT |
626 | int error; |
627 | ||
a50cd269 | 628 | xfs_fs_stop_syncd(vfsp); |
b83bd138 NS |
629 | bhv_vfs_sync(vfsp, SYNC_ATTR | SYNC_DELWRI, NULL); |
630 | error = bhv_vfs_unmount(vfsp, 0, NULL); | |
1da177e4 | 631 | if (error) { |
b83bd138 NS |
632 | printk("XFS: unmount got error=%d\n", error); |
633 | printk("%s: vfs=0x%p left dangling!\n", __FUNCTION__, vfsp); | |
634 | } else { | |
635 | vfs_deallocate(vfsp); | |
1da177e4 | 636 | } |
1da177e4 LT |
637 | } |
638 | ||
639 | STATIC void | |
a50cd269 | 640 | xfs_fs_write_super( |
1da177e4 LT |
641 | struct super_block *sb) |
642 | { | |
b83bd138 NS |
643 | if (!(sb->s_flags & MS_RDONLY)) |
644 | bhv_vfs_sync(vfs_from_sb(sb), SYNC_FSDATA, NULL); | |
1da177e4 LT |
645 | sb->s_dirt = 0; |
646 | } | |
647 | ||
648 | STATIC int | |
a50cd269 | 649 | xfs_fs_sync_super( |
1da177e4 LT |
650 | struct super_block *sb, |
651 | int wait) | |
652 | { | |
b83bd138 NS |
653 | bhv_vfs_t *vfsp = vfs_from_sb(sb); |
654 | int error; | |
655 | int flags; | |
1da177e4 | 656 | |
f898d6c0 CH |
657 | if (unlikely(sb->s_frozen == SB_FREEZE_WRITE)) |
658 | flags = SYNC_QUIESCE; | |
659 | else | |
660 | flags = SYNC_FSDATA | (wait ? SYNC_WAIT : 0); | |
1da177e4 | 661 | |
b83bd138 | 662 | error = bhv_vfs_sync(vfsp, flags, NULL); |
1da177e4 LT |
663 | sb->s_dirt = 0; |
664 | ||
665 | if (unlikely(laptop_mode)) { | |
666 | int prev_sync_seq = vfsp->vfs_sync_seq; | |
667 | ||
668 | /* | |
669 | * The disk must be active because we're syncing. | |
670 | * We schedule xfssyncd now (now that the disk is | |
671 | * active) instead of later (when it might not be). | |
672 | */ | |
673 | wake_up_process(vfsp->vfs_sync_task); | |
674 | /* | |
675 | * We have to wait for the sync iteration to complete. | |
676 | * If we don't, the disk activity caused by the sync | |
677 | * will come after the sync is completed, and that | |
678 | * triggers another sync from laptop mode. | |
679 | */ | |
680 | wait_event(vfsp->vfs_wait_single_sync_task, | |
681 | vfsp->vfs_sync_seq != prev_sync_seq); | |
682 | } | |
683 | ||
684 | return -error; | |
685 | } | |
686 | ||
687 | STATIC int | |
a50cd269 | 688 | xfs_fs_statfs( |
1da177e4 LT |
689 | struct super_block *sb, |
690 | struct kstatfs *statp) | |
691 | { | |
b83bd138 | 692 | return -bhv_vfs_statvfs(vfs_from_sb(sb), statp, NULL); |
1da177e4 LT |
693 | } |
694 | ||
695 | STATIC int | |
a50cd269 | 696 | xfs_fs_remount( |
1da177e4 LT |
697 | struct super_block *sb, |
698 | int *flags, | |
699 | char *options) | |
700 | { | |
b83bd138 | 701 | bhv_vfs_t *vfsp = vfs_from_sb(sb); |
764d1f89 | 702 | struct xfs_mount_args *args = xfs_args_allocate(sb, 0); |
1da177e4 LT |
703 | int error; |
704 | ||
b83bd138 | 705 | error = bhv_vfs_parseargs(vfsp, options, args, 1); |
1da177e4 | 706 | if (!error) |
b83bd138 | 707 | error = bhv_vfs_mntupdate(vfsp, flags, args); |
1da177e4 LT |
708 | kmem_free(args, sizeof(*args)); |
709 | return -error; | |
710 | } | |
711 | ||
712 | STATIC void | |
a50cd269 | 713 | xfs_fs_lockfs( |
1da177e4 LT |
714 | struct super_block *sb) |
715 | { | |
b83bd138 | 716 | bhv_vfs_freeze(vfs_from_sb(sb)); |
1da177e4 LT |
717 | } |
718 | ||
719 | STATIC int | |
a50cd269 | 720 | xfs_fs_show_options( |
1da177e4 LT |
721 | struct seq_file *m, |
722 | struct vfsmount *mnt) | |
723 | { | |
b83bd138 | 724 | return -bhv_vfs_showargs(vfs_from_sb(mnt->mnt_sb), m); |
1da177e4 LT |
725 | } |
726 | ||
ee34807a | 727 | STATIC int |
a50cd269 | 728 | xfs_fs_quotasync( |
ee34807a NS |
729 | struct super_block *sb, |
730 | int type) | |
731 | { | |
b83bd138 | 732 | return -bhv_vfs_quotactl(vfs_from_sb(sb), Q_XQUOTASYNC, 0, NULL); |
ee34807a NS |
733 | } |
734 | ||
1da177e4 | 735 | STATIC int |
a50cd269 | 736 | xfs_fs_getxstate( |
1da177e4 LT |
737 | struct super_block *sb, |
738 | struct fs_quota_stat *fqs) | |
739 | { | |
b83bd138 | 740 | return -bhv_vfs_quotactl(vfs_from_sb(sb), Q_XGETQSTAT, 0, (caddr_t)fqs); |
1da177e4 LT |
741 | } |
742 | ||
743 | STATIC int | |
a50cd269 | 744 | xfs_fs_setxstate( |
1da177e4 LT |
745 | struct super_block *sb, |
746 | unsigned int flags, | |
747 | int op) | |
748 | { | |
b83bd138 | 749 | return -bhv_vfs_quotactl(vfs_from_sb(sb), op, 0, (caddr_t)&flags); |
1da177e4 LT |
750 | } |
751 | ||
752 | STATIC int | |
a50cd269 | 753 | xfs_fs_getxquota( |
1da177e4 LT |
754 | struct super_block *sb, |
755 | int type, | |
756 | qid_t id, | |
757 | struct fs_disk_quota *fdq) | |
758 | { | |
b83bd138 NS |
759 | return -bhv_vfs_quotactl(vfs_from_sb(sb), |
760 | (type == USRQUOTA) ? Q_XGETQUOTA : | |
761 | ((type == GRPQUOTA) ? Q_XGETGQUOTA : | |
762 | Q_XGETPQUOTA), id, (caddr_t)fdq); | |
1da177e4 LT |
763 | } |
764 | ||
765 | STATIC int | |
a50cd269 | 766 | xfs_fs_setxquota( |
1da177e4 LT |
767 | struct super_block *sb, |
768 | int type, | |
769 | qid_t id, | |
770 | struct fs_disk_quota *fdq) | |
771 | { | |
b83bd138 NS |
772 | return -bhv_vfs_quotactl(vfs_from_sb(sb), |
773 | (type == USRQUOTA) ? Q_XSETQLIM : | |
774 | ((type == GRPQUOTA) ? Q_XSETGQLIM : | |
775 | Q_XSETPQLIM), id, (caddr_t)fdq); | |
1da177e4 LT |
776 | } |
777 | ||
778 | STATIC int | |
a50cd269 | 779 | xfs_fs_fill_super( |
1da177e4 LT |
780 | struct super_block *sb, |
781 | void *data, | |
782 | int silent) | |
783 | { | |
67fcaa73 | 784 | struct bhv_vnode *rootvp; |
b83bd138 | 785 | struct bhv_vfs *vfsp = vfs_allocate(sb); |
764d1f89 | 786 | struct xfs_mount_args *args = xfs_args_allocate(sb, silent); |
1da177e4 | 787 | struct kstatfs statvfs; |
b83bd138 | 788 | int error; |
1da177e4 | 789 | |
1da177e4 LT |
790 | bhv_insert_all_vfsops(vfsp); |
791 | ||
b83bd138 | 792 | error = bhv_vfs_parseargs(vfsp, (char *)data, args, 0); |
1da177e4 LT |
793 | if (error) { |
794 | bhv_remove_all_vfsops(vfsp, 1); | |
795 | goto fail_vfsop; | |
796 | } | |
797 | ||
798 | sb_min_blocksize(sb, BBSIZE); | |
799 | #ifdef CONFIG_XFS_EXPORT | |
a50cd269 | 800 | sb->s_export_op = &xfs_export_operations; |
1da177e4 | 801 | #endif |
a50cd269 NS |
802 | sb->s_qcop = &xfs_quotactl_operations; |
803 | sb->s_op = &xfs_super_operations; | |
1da177e4 | 804 | |
b83bd138 | 805 | error = bhv_vfs_mount(vfsp, args, NULL); |
1da177e4 LT |
806 | if (error) { |
807 | bhv_remove_all_vfsops(vfsp, 1); | |
808 | goto fail_vfsop; | |
809 | } | |
810 | ||
b83bd138 | 811 | error = bhv_vfs_statvfs(vfsp, &statvfs, NULL); |
1da177e4 LT |
812 | if (error) |
813 | goto fail_unmount; | |
814 | ||
815 | sb->s_dirt = 1; | |
816 | sb->s_magic = statvfs.f_type; | |
817 | sb->s_blocksize = statvfs.f_bsize; | |
818 | sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1; | |
819 | sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); | |
820 | sb->s_time_gran = 1; | |
821 | set_posix_acl_flag(sb); | |
822 | ||
b83bd138 | 823 | error = bhv_vfs_root(vfsp, &rootvp); |
1da177e4 LT |
824 | if (error) |
825 | goto fail_unmount; | |
826 | ||
ec86dc02 | 827 | sb->s_root = d_alloc_root(vn_to_inode(rootvp)); |
1da177e4 LT |
828 | if (!sb->s_root) { |
829 | error = ENOMEM; | |
830 | goto fail_vnrele; | |
831 | } | |
832 | if (is_bad_inode(sb->s_root->d_inode)) { | |
833 | error = EINVAL; | |
834 | goto fail_vnrele; | |
835 | } | |
a50cd269 | 836 | if ((error = xfs_fs_start_syncd(vfsp))) |
1da177e4 LT |
837 | goto fail_vnrele; |
838 | vn_trace_exit(rootvp, __FUNCTION__, (inst_t *)__return_address); | |
839 | ||
840 | kmem_free(args, sizeof(*args)); | |
841 | return 0; | |
842 | ||
843 | fail_vnrele: | |
844 | if (sb->s_root) { | |
845 | dput(sb->s_root); | |
846 | sb->s_root = NULL; | |
847 | } else { | |
848 | VN_RELE(rootvp); | |
849 | } | |
850 | ||
851 | fail_unmount: | |
b83bd138 | 852 | bhv_vfs_unmount(vfsp, 0, NULL); |
1da177e4 LT |
853 | |
854 | fail_vfsop: | |
855 | vfs_deallocate(vfsp); | |
856 | kmem_free(args, sizeof(*args)); | |
857 | return -error; | |
858 | } | |
859 | ||
860 | STATIC struct super_block * | |
a50cd269 | 861 | xfs_fs_get_sb( |
1da177e4 LT |
862 | struct file_system_type *fs_type, |
863 | int flags, | |
864 | const char *dev_name, | |
865 | void *data) | |
866 | { | |
a50cd269 NS |
867 | return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super); |
868 | } | |
869 | ||
870 | STATIC struct super_operations xfs_super_operations = { | |
871 | .alloc_inode = xfs_fs_alloc_inode, | |
872 | .destroy_inode = xfs_fs_destroy_inode, | |
873 | .write_inode = xfs_fs_write_inode, | |
874 | .clear_inode = xfs_fs_clear_inode, | |
875 | .put_super = xfs_fs_put_super, | |
876 | .write_super = xfs_fs_write_super, | |
877 | .sync_fs = xfs_fs_sync_super, | |
878 | .write_super_lockfs = xfs_fs_lockfs, | |
879 | .statfs = xfs_fs_statfs, | |
880 | .remount_fs = xfs_fs_remount, | |
881 | .show_options = xfs_fs_show_options, | |
1da177e4 LT |
882 | }; |
883 | ||
a50cd269 NS |
884 | STATIC struct quotactl_ops xfs_quotactl_operations = { |
885 | .quota_sync = xfs_fs_quotasync, | |
886 | .get_xstate = xfs_fs_getxstate, | |
887 | .set_xstate = xfs_fs_setxstate, | |
888 | .get_xquota = xfs_fs_getxquota, | |
889 | .set_xquota = xfs_fs_setxquota, | |
1da177e4 LT |
890 | }; |
891 | ||
892 | STATIC struct file_system_type xfs_fs_type = { | |
893 | .owner = THIS_MODULE, | |
894 | .name = "xfs", | |
a50cd269 | 895 | .get_sb = xfs_fs_get_sb, |
1da177e4 LT |
896 | .kill_sb = kill_block_super, |
897 | .fs_flags = FS_REQUIRES_DEV, | |
898 | }; | |
899 | ||
900 | ||
901 | STATIC int __init | |
902 | init_xfs_fs( void ) | |
903 | { | |
904 | int error; | |
905 | struct sysinfo si; | |
906 | static char message[] __initdata = KERN_INFO \ | |
907 | XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n"; | |
908 | ||
909 | printk(message); | |
910 | ||
911 | si_meminfo(&si); | |
912 | xfs_physmem = si.totalram; | |
913 | ||
914 | ktrace_init(64); | |
915 | ||
8758280f | 916 | error = xfs_init_zones(); |
1da177e4 | 917 | if (error < 0) |
0829c360 | 918 | goto undo_zones; |
1da177e4 | 919 | |
ce8e922c | 920 | error = xfs_buf_init(); |
1da177e4 | 921 | if (error < 0) |
ce8e922c | 922 | goto undo_buffers; |
1da177e4 LT |
923 | |
924 | vn_init(); | |
925 | xfs_init(); | |
926 | uuid_init(); | |
927 | vfs_initquota(); | |
928 | ||
929 | error = register_filesystem(&xfs_fs_type); | |
930 | if (error) | |
931 | goto undo_register; | |
1da177e4 LT |
932 | return 0; |
933 | ||
934 | undo_register: | |
ce8e922c | 935 | xfs_buf_terminate(); |
1da177e4 | 936 | |
ce8e922c | 937 | undo_buffers: |
8758280f | 938 | xfs_destroy_zones(); |
1da177e4 | 939 | |
0829c360 | 940 | undo_zones: |
1da177e4 LT |
941 | return error; |
942 | } | |
943 | ||
944 | STATIC void __exit | |
945 | exit_xfs_fs( void ) | |
946 | { | |
947 | vfs_exitquota(); | |
1da177e4 LT |
948 | unregister_filesystem(&xfs_fs_type); |
949 | xfs_cleanup(); | |
ce8e922c | 950 | xfs_buf_terminate(); |
8758280f | 951 | xfs_destroy_zones(); |
1da177e4 LT |
952 | ktrace_uninit(); |
953 | } | |
954 | ||
955 | module_init(init_xfs_fs); | |
956 | module_exit(exit_xfs_fs); | |
957 | ||
958 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
959 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
960 | MODULE_LICENSE("GPL"); |