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 LT |
61 | |
62 | STATIC struct quotactl_ops linvfs_qops; | |
63 | STATIC struct super_operations linvfs_sops; | |
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( | |
70 | struct super_block *sb) | |
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; | |
83 | ||
84 | /* Default to 32 bit inodes on Linux all the time */ | |
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: | |
1da177e4 LT |
130 | inode->i_op = &linvfs_file_inode_operations; |
131 | inode->i_fop = &linvfs_file_operations; | |
132 | inode->i_mapping->a_ops = &linvfs_aops; | |
0432dab2 CH |
133 | break; |
134 | case S_IFDIR: | |
1da177e4 LT |
135 | inode->i_op = &linvfs_dir_inode_operations; |
136 | inode->i_fop = &linvfs_dir_operations; | |
0432dab2 CH |
137 | break; |
138 | case S_IFLNK: | |
1da177e4 LT |
139 | inode->i_op = &linvfs_symlink_inode_operations; |
140 | if (inode->i_blocks) | |
141 | inode->i_mapping->a_ops = &linvfs_aops; | |
0432dab2 CH |
142 | break; |
143 | default: | |
1da177e4 LT |
144 | inode->i_op = &linvfs_file_inode_operations; |
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, | |
153 | vnode_t *vp, | |
154 | xfs_inode_t *ip) | |
155 | { | |
156 | struct inode *inode = LINVFS_GET_IP(vp); | |
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 | |
e8c8b3a7 | 175 | inode->i_blksize = xfs_preferred_iosize(mp); |
1da177e4 LT |
176 | inode->i_generation = ip->i_d.di_gen; |
177 | i_size_write(inode, ip->i_d.di_size); | |
178 | inode->i_blocks = | |
179 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
180 | inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec; | |
181 | inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec; | |
182 | inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec; | |
183 | inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec; | |
184 | inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec; | |
185 | inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec; | |
186 | if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE) | |
187 | inode->i_flags |= S_IMMUTABLE; | |
188 | else | |
189 | inode->i_flags &= ~S_IMMUTABLE; | |
190 | if (ip->i_d.di_flags & XFS_DIFLAG_APPEND) | |
191 | inode->i_flags |= S_APPEND; | |
192 | else | |
193 | inode->i_flags &= ~S_APPEND; | |
194 | if (ip->i_d.di_flags & XFS_DIFLAG_SYNC) | |
195 | inode->i_flags |= S_SYNC; | |
196 | else | |
197 | inode->i_flags &= ~S_SYNC; | |
198 | if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME) | |
199 | inode->i_flags |= S_NOATIME; | |
200 | else | |
201 | inode->i_flags &= ~S_NOATIME; | |
202 | vp->v_flag &= ~VMODIFIED; | |
203 | } | |
204 | ||
205 | void | |
206 | xfs_initialize_vnode( | |
207 | bhv_desc_t *bdp, | |
208 | vnode_t *vp, | |
209 | bhv_desc_t *inode_bhv, | |
210 | int unlock) | |
211 | { | |
212 | xfs_inode_t *ip = XFS_BHVTOI(inode_bhv); | |
213 | struct inode *inode = LINVFS_GET_IP(vp); | |
214 | ||
215 | if (!inode_bhv->bd_vobj) { | |
216 | vp->v_vfsp = bhvtovfs(bdp); | |
217 | bhv_desc_init(inode_bhv, ip, vp, &xfs_vnodeops); | |
218 | bhv_insert(VN_BHV_HEAD(vp), inode_bhv); | |
219 | } | |
220 | ||
221 | /* | |
222 | * We need to set the ops vectors, and unlock the inode, but if | |
223 | * we have been called during the new inode create process, it is | |
224 | * too early to fill in the Linux inode. We will get called a | |
225 | * second time once the inode is properly set up, and then we can | |
226 | * finish our work. | |
227 | */ | |
228 | if (ip->i_d.di_mode != 0 && unlock && (inode->i_state & I_NEW)) { | |
1da177e4 LT |
229 | xfs_revalidate_inode(XFS_BHVTOM(bdp), vp, ip); |
230 | xfs_set_inodeops(inode); | |
231 | ||
232 | ip->i_flags &= ~XFS_INEW; | |
233 | barrier(); | |
234 | ||
235 | unlock_new_inode(inode); | |
236 | } | |
237 | } | |
238 | ||
239 | int | |
240 | xfs_blkdev_get( | |
241 | xfs_mount_t *mp, | |
242 | const char *name, | |
243 | struct block_device **bdevp) | |
244 | { | |
245 | int error = 0; | |
246 | ||
247 | *bdevp = open_bdev_excl(name, 0, mp); | |
248 | if (IS_ERR(*bdevp)) { | |
249 | error = PTR_ERR(*bdevp); | |
250 | printk("XFS: Invalid device [%s], error=%d\n", name, error); | |
251 | } | |
252 | ||
253 | return -error; | |
254 | } | |
255 | ||
256 | void | |
257 | xfs_blkdev_put( | |
258 | struct block_device *bdev) | |
259 | { | |
260 | if (bdev) | |
261 | close_bdev_excl(bdev); | |
262 | } | |
263 | ||
f538d4da CH |
264 | /* |
265 | * Try to write out the superblock using barriers. | |
266 | */ | |
267 | STATIC int | |
268 | xfs_barrier_test( | |
269 | xfs_mount_t *mp) | |
270 | { | |
271 | xfs_buf_t *sbp = xfs_getsb(mp, 0); | |
272 | int error; | |
273 | ||
274 | XFS_BUF_UNDONE(sbp); | |
275 | XFS_BUF_UNREAD(sbp); | |
276 | XFS_BUF_UNDELAYWRITE(sbp); | |
277 | XFS_BUF_WRITE(sbp); | |
278 | XFS_BUF_UNASYNC(sbp); | |
279 | XFS_BUF_ORDERED(sbp); | |
280 | ||
281 | xfsbdstrat(mp, sbp); | |
282 | error = xfs_iowait(sbp); | |
283 | ||
284 | /* | |
285 | * Clear all the flags we set and possible error state in the | |
286 | * buffer. We only did the write to try out whether barriers | |
287 | * worked and shouldn't leave any traces in the superblock | |
288 | * buffer. | |
289 | */ | |
290 | XFS_BUF_DONE(sbp); | |
291 | XFS_BUF_ERROR(sbp, 0); | |
292 | XFS_BUF_UNORDERED(sbp); | |
293 | ||
294 | xfs_buf_relse(sbp); | |
295 | return error; | |
296 | } | |
297 | ||
298 | void | |
299 | xfs_mountfs_check_barriers(xfs_mount_t *mp) | |
300 | { | |
301 | int error; | |
302 | ||
303 | if (mp->m_logdev_targp != mp->m_ddev_targp) { | |
304 | xfs_fs_cmn_err(CE_NOTE, mp, | |
305 | "Disabling barriers, not supported with external log device"); | |
306 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 307 | return; |
f538d4da CH |
308 | } |
309 | ||
ce8e922c | 310 | if (mp->m_ddev_targp->bt_bdev->bd_disk->queue->ordered == |
f538d4da CH |
311 | QUEUE_ORDERED_NONE) { |
312 | xfs_fs_cmn_err(CE_NOTE, mp, | |
313 | "Disabling barriers, not supported by the underlying device"); | |
314 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 315 | return; |
f538d4da CH |
316 | } |
317 | ||
318 | error = xfs_barrier_test(mp); | |
319 | if (error) { | |
320 | xfs_fs_cmn_err(CE_NOTE, mp, | |
321 | "Disabling barriers, trial barrier write failed"); | |
322 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 323 | return; |
f538d4da CH |
324 | } |
325 | } | |
326 | ||
327 | void | |
328 | xfs_blkdev_issue_flush( | |
329 | xfs_buftarg_t *buftarg) | |
330 | { | |
ce8e922c | 331 | blkdev_issue_flush(buftarg->bt_bdev, NULL); |
f538d4da | 332 | } |
1da177e4 LT |
333 | |
334 | STATIC struct inode * | |
335 | linvfs_alloc_inode( | |
336 | struct super_block *sb) | |
337 | { | |
338 | vnode_t *vp; | |
339 | ||
8758280f NS |
340 | vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP); |
341 | if (unlikely(!vp)) | |
1da177e4 LT |
342 | return NULL; |
343 | return LINVFS_GET_IP(vp); | |
344 | } | |
345 | ||
346 | STATIC void | |
347 | linvfs_destroy_inode( | |
348 | struct inode *inode) | |
349 | { | |
0829c360 | 350 | kmem_zone_free(xfs_vnode_zone, LINVFS_GET_VP(inode)); |
1da177e4 LT |
351 | } |
352 | ||
353 | STATIC void | |
0829c360 | 354 | linvfs_inode_init_once( |
8758280f NS |
355 | void *vnode, |
356 | kmem_zone_t *zonep, | |
1da177e4 LT |
357 | unsigned long flags) |
358 | { | |
1da177e4 | 359 | if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == |
8758280f NS |
360 | SLAB_CTOR_CONSTRUCTOR) |
361 | inode_init_once(LINVFS_GET_IP((vnode_t *)vnode)); | |
1da177e4 LT |
362 | } |
363 | ||
364 | STATIC int | |
8758280f | 365 | xfs_init_zones(void) |
1da177e4 | 366 | { |
8758280f NS |
367 | xfs_vnode_zone = kmem_zone_init_flags(sizeof(vnode_t), "xfs_vnode_t", |
368 | KM_ZONE_HWALIGN | KM_ZONE_RECLAIM, | |
369 | linvfs_inode_init_once); | |
0829c360 CH |
370 | if (!xfs_vnode_zone) |
371 | goto out; | |
372 | ||
373 | xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend"); | |
374 | if (!xfs_ioend_zone) | |
375 | goto out_destroy_vnode_zone; | |
376 | ||
377 | xfs_ioend_pool = mempool_create(4 * MAX_BUF_PER_PAGE, | |
8758280f NS |
378 | mempool_alloc_slab, mempool_free_slab, |
379 | xfs_ioend_zone); | |
0829c360 CH |
380 | if (!xfs_ioend_pool) |
381 | goto out_free_ioend_zone; | |
1da177e4 | 382 | return 0; |
0829c360 | 383 | |
0829c360 CH |
384 | out_free_ioend_zone: |
385 | kmem_zone_destroy(xfs_ioend_zone); | |
386 | out_destroy_vnode_zone: | |
387 | kmem_zone_destroy(xfs_vnode_zone); | |
388 | out: | |
389 | return -ENOMEM; | |
1da177e4 LT |
390 | } |
391 | ||
392 | STATIC void | |
8758280f | 393 | xfs_destroy_zones(void) |
1da177e4 | 394 | { |
0829c360 CH |
395 | mempool_destroy(xfs_ioend_pool); |
396 | kmem_zone_destroy(xfs_vnode_zone); | |
397 | kmem_zone_destroy(xfs_ioend_zone); | |
1da177e4 LT |
398 | } |
399 | ||
400 | /* | |
401 | * Attempt to flush the inode, this will actually fail | |
402 | * if the inode is pinned, but we dirty the inode again | |
403 | * at the point when it is unpinned after a log write, | |
8758280f | 404 | * since this is when the inode itself becomes flushable. |
1da177e4 LT |
405 | */ |
406 | STATIC int | |
407 | linvfs_write_inode( | |
408 | struct inode *inode, | |
409 | int sync) | |
410 | { | |
411 | vnode_t *vp = LINVFS_GET_VP(inode); | |
412 | int error = 0, flags = FLUSH_INODE; | |
413 | ||
414 | if (vp) { | |
415 | vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); | |
416 | if (sync) | |
417 | flags |= FLUSH_SYNC; | |
418 | VOP_IFLUSH(vp, flags, error); | |
419 | if (error == EAGAIN) { | |
420 | if (sync) | |
421 | VOP_IFLUSH(vp, flags | FLUSH_LOG, error); | |
422 | else | |
423 | error = 0; | |
424 | } | |
425 | } | |
426 | ||
427 | return -error; | |
428 | } | |
429 | ||
430 | STATIC void | |
431 | linvfs_clear_inode( | |
432 | struct inode *inode) | |
433 | { | |
434 | vnode_t *vp = LINVFS_GET_VP(inode); | |
56d433e4 | 435 | int error, cache; |
1da177e4 | 436 | |
56d433e4 CH |
437 | vn_trace_entry(vp, "clear_inode", (inst_t *)__return_address); |
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 | */ | |
448 | if (vp->v_fbhv) | |
449 | VOP_INACTIVE(vp, NULL, cache); | |
1da177e4 | 450 | |
56d433e4 CH |
451 | VN_LOCK(vp); |
452 | vp->v_flag &= ~VMODIFIED; | |
453 | VN_UNLOCK(vp, 0); | |
454 | ||
0c147f9a FB |
455 | if (vp->v_fbhv) { |
456 | VOP_RECLAIM(vp, error); | |
457 | if (error) | |
458 | panic("vn_purge: cannot reclaim"); | |
459 | } | |
56d433e4 CH |
460 | |
461 | ASSERT(vp->v_fbhv == NULL); | |
462 | ||
463 | #ifdef XFS_VNODE_TRACE | |
464 | ktrace_free(vp->v_trace); | |
465 | #endif | |
466 | } | |
1da177e4 LT |
467 | |
468 | /* | |
469 | * Enqueue a work item to be picked up by the vfs xfssyncd thread. | |
470 | * Doing this has two advantages: | |
471 | * - It saves on stack space, which is tight in certain situations | |
472 | * - It can be used (with care) as a mechanism to avoid deadlocks. | |
473 | * Flushing while allocating in a full filesystem requires both. | |
474 | */ | |
475 | STATIC void | |
476 | xfs_syncd_queue_work( | |
477 | struct vfs *vfs, | |
478 | void *data, | |
479 | void (*syncer)(vfs_t *, void *)) | |
480 | { | |
481 | vfs_sync_work_t *work; | |
482 | ||
483 | work = kmem_alloc(sizeof(struct vfs_sync_work), KM_SLEEP); | |
484 | INIT_LIST_HEAD(&work->w_list); | |
485 | work->w_syncer = syncer; | |
486 | work->w_data = data; | |
487 | work->w_vfs = vfs; | |
488 | spin_lock(&vfs->vfs_sync_lock); | |
489 | list_add_tail(&work->w_list, &vfs->vfs_sync_list); | |
490 | spin_unlock(&vfs->vfs_sync_lock); | |
491 | wake_up_process(vfs->vfs_sync_task); | |
492 | } | |
493 | ||
494 | /* | |
495 | * Flush delayed allocate data, attempting to free up reserved space | |
496 | * from existing allocations. At this point a new allocation attempt | |
497 | * has failed with ENOSPC and we are in the process of scratching our | |
498 | * heads, looking about for more room... | |
499 | */ | |
500 | STATIC void | |
501 | xfs_flush_inode_work( | |
502 | vfs_t *vfs, | |
503 | void *inode) | |
504 | { | |
505 | filemap_flush(((struct inode *)inode)->i_mapping); | |
506 | iput((struct inode *)inode); | |
507 | } | |
508 | ||
509 | void | |
510 | xfs_flush_inode( | |
511 | xfs_inode_t *ip) | |
512 | { | |
513 | struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); | |
514 | struct vfs *vfs = XFS_MTOVFS(ip->i_mount); | |
515 | ||
516 | igrab(inode); | |
517 | xfs_syncd_queue_work(vfs, inode, xfs_flush_inode_work); | |
041e0e3b | 518 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
519 | } |
520 | ||
521 | /* | |
522 | * This is the "bigger hammer" version of xfs_flush_inode_work... | |
523 | * (IOW, "If at first you don't succeed, use a Bigger Hammer"). | |
524 | */ | |
525 | STATIC void | |
526 | xfs_flush_device_work( | |
527 | vfs_t *vfs, | |
528 | void *inode) | |
529 | { | |
530 | sync_blockdev(vfs->vfs_super->s_bdev); | |
531 | iput((struct inode *)inode); | |
532 | } | |
533 | ||
534 | void | |
535 | xfs_flush_device( | |
536 | xfs_inode_t *ip) | |
537 | { | |
538 | struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); | |
539 | struct vfs *vfs = XFS_MTOVFS(ip->i_mount); | |
540 | ||
541 | igrab(inode); | |
542 | xfs_syncd_queue_work(vfs, inode, xfs_flush_device_work); | |
041e0e3b | 543 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
544 | xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC); |
545 | } | |
546 | ||
547 | #define SYNCD_FLAGS (SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR) | |
548 | STATIC void | |
549 | vfs_sync_worker( | |
550 | vfs_t *vfsp, | |
551 | void *unused) | |
552 | { | |
553 | int error; | |
554 | ||
555 | if (!(vfsp->vfs_flag & VFS_RDONLY)) | |
556 | VFS_SYNC(vfsp, SYNCD_FLAGS, NULL, error); | |
557 | vfsp->vfs_sync_seq++; | |
558 | wmb(); | |
559 | wake_up(&vfsp->vfs_wait_single_sync_task); | |
560 | } | |
561 | ||
562 | STATIC int | |
563 | xfssyncd( | |
564 | void *arg) | |
565 | { | |
566 | long timeleft; | |
567 | vfs_t *vfsp = (vfs_t *) arg; | |
1da177e4 | 568 | struct vfs_sync_work *work, *n; |
4df08c52 | 569 | LIST_HEAD (tmp); |
1da177e4 | 570 | |
041e0e3b | 571 | timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10); |
1da177e4 | 572 | for (;;) { |
041e0e3b | 573 | timeleft = schedule_timeout_interruptible(timeleft); |
1da177e4 | 574 | /* swsusp */ |
3e1d1d28 | 575 | try_to_freeze(); |
71df099d | 576 | if (kthread_should_stop() && list_empty(&vfsp->vfs_sync_list)) |
1da177e4 LT |
577 | break; |
578 | ||
579 | spin_lock(&vfsp->vfs_sync_lock); | |
580 | /* | |
581 | * We can get woken by laptop mode, to do a sync - | |
582 | * that's the (only!) case where the list would be | |
583 | * empty with time remaining. | |
584 | */ | |
585 | if (!timeleft || list_empty(&vfsp->vfs_sync_list)) { | |
586 | if (!timeleft) | |
041e0e3b NA |
587 | timeleft = xfs_syncd_centisecs * |
588 | msecs_to_jiffies(10); | |
1da177e4 LT |
589 | INIT_LIST_HEAD(&vfsp->vfs_sync_work.w_list); |
590 | list_add_tail(&vfsp->vfs_sync_work.w_list, | |
591 | &vfsp->vfs_sync_list); | |
592 | } | |
593 | list_for_each_entry_safe(work, n, &vfsp->vfs_sync_list, w_list) | |
594 | list_move(&work->w_list, &tmp); | |
595 | spin_unlock(&vfsp->vfs_sync_lock); | |
596 | ||
597 | list_for_each_entry_safe(work, n, &tmp, w_list) { | |
598 | (*work->w_syncer)(vfsp, work->w_data); | |
599 | list_del(&work->w_list); | |
600 | if (work == &vfsp->vfs_sync_work) | |
601 | continue; | |
602 | kmem_free(work, sizeof(struct vfs_sync_work)); | |
603 | } | |
604 | } | |
605 | ||
1da177e4 LT |
606 | return 0; |
607 | } | |
608 | ||
609 | STATIC int | |
610 | linvfs_start_syncd( | |
611 | vfs_t *vfsp) | |
612 | { | |
4df08c52 CH |
613 | vfsp->vfs_sync_work.w_syncer = vfs_sync_worker; |
614 | vfsp->vfs_sync_work.w_vfs = vfsp; | |
615 | vfsp->vfs_sync_task = kthread_run(xfssyncd, vfsp, "xfssyncd"); | |
616 | if (IS_ERR(vfsp->vfs_sync_task)) | |
617 | return -PTR_ERR(vfsp->vfs_sync_task); | |
1da177e4 LT |
618 | return 0; |
619 | } | |
620 | ||
621 | STATIC void | |
622 | linvfs_stop_syncd( | |
623 | vfs_t *vfsp) | |
624 | { | |
4df08c52 | 625 | kthread_stop(vfsp->vfs_sync_task); |
1da177e4 LT |
626 | } |
627 | ||
628 | STATIC void | |
629 | linvfs_put_super( | |
630 | struct super_block *sb) | |
631 | { | |
632 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
633 | int error; | |
634 | ||
635 | linvfs_stop_syncd(vfsp); | |
636 | VFS_SYNC(vfsp, SYNC_ATTR|SYNC_DELWRI, NULL, error); | |
637 | if (!error) | |
638 | VFS_UNMOUNT(vfsp, 0, NULL, error); | |
639 | if (error) { | |
640 | printk("XFS unmount got error %d\n", error); | |
641 | printk("%s: vfsp/0x%p left dangling!\n", __FUNCTION__, vfsp); | |
642 | return; | |
643 | } | |
644 | ||
645 | vfs_deallocate(vfsp); | |
646 | } | |
647 | ||
648 | STATIC void | |
649 | linvfs_write_super( | |
650 | struct super_block *sb) | |
651 | { | |
652 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
653 | int error; | |
654 | ||
655 | if (sb->s_flags & MS_RDONLY) { | |
656 | sb->s_dirt = 0; /* paranoia */ | |
657 | return; | |
658 | } | |
659 | /* Push the log and superblock a little */ | |
660 | VFS_SYNC(vfsp, SYNC_FSDATA, NULL, error); | |
661 | sb->s_dirt = 0; | |
662 | } | |
663 | ||
664 | STATIC int | |
665 | linvfs_sync_super( | |
666 | struct super_block *sb, | |
667 | int wait) | |
668 | { | |
669 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
670 | int error; | |
671 | int flags = SYNC_FSDATA; | |
672 | ||
f898d6c0 CH |
673 | if (unlikely(sb->s_frozen == SB_FREEZE_WRITE)) |
674 | flags = SYNC_QUIESCE; | |
675 | else | |
676 | flags = SYNC_FSDATA | (wait ? SYNC_WAIT : 0); | |
1da177e4 LT |
677 | |
678 | VFS_SYNC(vfsp, flags, NULL, error); | |
679 | sb->s_dirt = 0; | |
680 | ||
681 | if (unlikely(laptop_mode)) { | |
682 | int prev_sync_seq = vfsp->vfs_sync_seq; | |
683 | ||
684 | /* | |
685 | * The disk must be active because we're syncing. | |
686 | * We schedule xfssyncd now (now that the disk is | |
687 | * active) instead of later (when it might not be). | |
688 | */ | |
689 | wake_up_process(vfsp->vfs_sync_task); | |
690 | /* | |
691 | * We have to wait for the sync iteration to complete. | |
692 | * If we don't, the disk activity caused by the sync | |
693 | * will come after the sync is completed, and that | |
694 | * triggers another sync from laptop mode. | |
695 | */ | |
696 | wait_event(vfsp->vfs_wait_single_sync_task, | |
697 | vfsp->vfs_sync_seq != prev_sync_seq); | |
698 | } | |
699 | ||
700 | return -error; | |
701 | } | |
702 | ||
703 | STATIC int | |
704 | linvfs_statfs( | |
705 | struct super_block *sb, | |
706 | struct kstatfs *statp) | |
707 | { | |
708 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
709 | int error; | |
710 | ||
711 | VFS_STATVFS(vfsp, statp, NULL, error); | |
712 | return -error; | |
713 | } | |
714 | ||
715 | STATIC int | |
716 | linvfs_remount( | |
717 | struct super_block *sb, | |
718 | int *flags, | |
719 | char *options) | |
720 | { | |
721 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
722 | struct xfs_mount_args *args = xfs_args_allocate(sb); | |
723 | int error; | |
724 | ||
725 | VFS_PARSEARGS(vfsp, options, args, 1, error); | |
726 | if (!error) | |
727 | VFS_MNTUPDATE(vfsp, flags, args, error); | |
728 | kmem_free(args, sizeof(*args)); | |
729 | return -error; | |
730 | } | |
731 | ||
732 | STATIC void | |
733 | linvfs_freeze_fs( | |
734 | struct super_block *sb) | |
735 | { | |
736 | VFS_FREEZE(LINVFS_GET_VFS(sb)); | |
737 | } | |
738 | ||
739 | STATIC int | |
740 | linvfs_show_options( | |
741 | struct seq_file *m, | |
742 | struct vfsmount *mnt) | |
743 | { | |
744 | struct vfs *vfsp = LINVFS_GET_VFS(mnt->mnt_sb); | |
745 | int error; | |
746 | ||
747 | VFS_SHOWARGS(vfsp, m, error); | |
748 | return error; | |
749 | } | |
750 | ||
ee34807a NS |
751 | STATIC int |
752 | linvfs_quotasync( | |
753 | struct super_block *sb, | |
754 | int type) | |
755 | { | |
756 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
757 | int error; | |
758 | ||
759 | VFS_QUOTACTL(vfsp, Q_XQUOTASYNC, 0, (caddr_t)NULL, error); | |
760 | return -error; | |
761 | } | |
762 | ||
1da177e4 LT |
763 | STATIC int |
764 | linvfs_getxstate( | |
765 | struct super_block *sb, | |
766 | struct fs_quota_stat *fqs) | |
767 | { | |
768 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
769 | int error; | |
770 | ||
771 | VFS_QUOTACTL(vfsp, Q_XGETQSTAT, 0, (caddr_t)fqs, error); | |
772 | return -error; | |
773 | } | |
774 | ||
775 | STATIC int | |
776 | linvfs_setxstate( | |
777 | struct super_block *sb, | |
778 | unsigned int flags, | |
779 | int op) | |
780 | { | |
781 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
782 | int error; | |
783 | ||
784 | VFS_QUOTACTL(vfsp, op, 0, (caddr_t)&flags, error); | |
785 | return -error; | |
786 | } | |
787 | ||
788 | STATIC int | |
789 | linvfs_getxquota( | |
790 | struct super_block *sb, | |
791 | int type, | |
792 | qid_t id, | |
793 | struct fs_disk_quota *fdq) | |
794 | { | |
795 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
796 | int error, getmode; | |
797 | ||
c8ad20ff NS |
798 | getmode = (type == USRQUOTA) ? Q_XGETQUOTA : |
799 | ((type == GRPQUOTA) ? Q_XGETGQUOTA : Q_XGETPQUOTA); | |
1da177e4 LT |
800 | VFS_QUOTACTL(vfsp, getmode, id, (caddr_t)fdq, error); |
801 | return -error; | |
802 | } | |
803 | ||
804 | STATIC int | |
805 | linvfs_setxquota( | |
806 | struct super_block *sb, | |
807 | int type, | |
808 | qid_t id, | |
809 | struct fs_disk_quota *fdq) | |
810 | { | |
811 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
812 | int error, setmode; | |
813 | ||
c8ad20ff NS |
814 | setmode = (type == USRQUOTA) ? Q_XSETQLIM : |
815 | ((type == GRPQUOTA) ? Q_XSETGQLIM : Q_XSETPQLIM); | |
1da177e4 LT |
816 | VFS_QUOTACTL(vfsp, setmode, id, (caddr_t)fdq, error); |
817 | return -error; | |
818 | } | |
819 | ||
820 | STATIC int | |
821 | linvfs_fill_super( | |
822 | struct super_block *sb, | |
823 | void *data, | |
824 | int silent) | |
825 | { | |
826 | vnode_t *rootvp; | |
827 | struct vfs *vfsp = vfs_allocate(); | |
828 | struct xfs_mount_args *args = xfs_args_allocate(sb); | |
829 | struct kstatfs statvfs; | |
830 | int error, error2; | |
831 | ||
832 | vfsp->vfs_super = sb; | |
833 | LINVFS_SET_VFS(sb, vfsp); | |
834 | if (sb->s_flags & MS_RDONLY) | |
835 | vfsp->vfs_flag |= VFS_RDONLY; | |
836 | bhv_insert_all_vfsops(vfsp); | |
837 | ||
838 | VFS_PARSEARGS(vfsp, (char *)data, args, 0, error); | |
839 | if (error) { | |
840 | bhv_remove_all_vfsops(vfsp, 1); | |
841 | goto fail_vfsop; | |
842 | } | |
843 | ||
844 | sb_min_blocksize(sb, BBSIZE); | |
845 | #ifdef CONFIG_XFS_EXPORT | |
846 | sb->s_export_op = &linvfs_export_ops; | |
847 | #endif | |
848 | sb->s_qcop = &linvfs_qops; | |
849 | sb->s_op = &linvfs_sops; | |
850 | ||
851 | VFS_MOUNT(vfsp, args, NULL, error); | |
852 | if (error) { | |
853 | bhv_remove_all_vfsops(vfsp, 1); | |
854 | goto fail_vfsop; | |
855 | } | |
856 | ||
857 | VFS_STATVFS(vfsp, &statvfs, NULL, error); | |
858 | if (error) | |
859 | goto fail_unmount; | |
860 | ||
861 | sb->s_dirt = 1; | |
862 | sb->s_magic = statvfs.f_type; | |
863 | sb->s_blocksize = statvfs.f_bsize; | |
864 | sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1; | |
865 | sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); | |
866 | sb->s_time_gran = 1; | |
867 | set_posix_acl_flag(sb); | |
868 | ||
869 | VFS_ROOT(vfsp, &rootvp, error); | |
870 | if (error) | |
871 | goto fail_unmount; | |
872 | ||
873 | sb->s_root = d_alloc_root(LINVFS_GET_IP(rootvp)); | |
874 | if (!sb->s_root) { | |
875 | error = ENOMEM; | |
876 | goto fail_vnrele; | |
877 | } | |
878 | if (is_bad_inode(sb->s_root->d_inode)) { | |
879 | error = EINVAL; | |
880 | goto fail_vnrele; | |
881 | } | |
882 | if ((error = linvfs_start_syncd(vfsp))) | |
883 | goto fail_vnrele; | |
884 | vn_trace_exit(rootvp, __FUNCTION__, (inst_t *)__return_address); | |
885 | ||
886 | kmem_free(args, sizeof(*args)); | |
887 | return 0; | |
888 | ||
889 | fail_vnrele: | |
890 | if (sb->s_root) { | |
891 | dput(sb->s_root); | |
892 | sb->s_root = NULL; | |
893 | } else { | |
894 | VN_RELE(rootvp); | |
895 | } | |
896 | ||
897 | fail_unmount: | |
898 | VFS_UNMOUNT(vfsp, 0, NULL, error2); | |
899 | ||
900 | fail_vfsop: | |
901 | vfs_deallocate(vfsp); | |
902 | kmem_free(args, sizeof(*args)); | |
903 | return -error; | |
904 | } | |
905 | ||
906 | STATIC struct super_block * | |
907 | linvfs_get_sb( | |
908 | struct file_system_type *fs_type, | |
909 | int flags, | |
910 | const char *dev_name, | |
911 | void *data) | |
912 | { | |
913 | return get_sb_bdev(fs_type, flags, dev_name, data, linvfs_fill_super); | |
914 | } | |
915 | ||
916 | STATIC struct super_operations linvfs_sops = { | |
917 | .alloc_inode = linvfs_alloc_inode, | |
918 | .destroy_inode = linvfs_destroy_inode, | |
919 | .write_inode = linvfs_write_inode, | |
920 | .clear_inode = linvfs_clear_inode, | |
921 | .put_super = linvfs_put_super, | |
922 | .write_super = linvfs_write_super, | |
923 | .sync_fs = linvfs_sync_super, | |
924 | .write_super_lockfs = linvfs_freeze_fs, | |
925 | .statfs = linvfs_statfs, | |
926 | .remount_fs = linvfs_remount, | |
927 | .show_options = linvfs_show_options, | |
928 | }; | |
929 | ||
930 | STATIC struct quotactl_ops linvfs_qops = { | |
ee34807a | 931 | .quota_sync = linvfs_quotasync, |
1da177e4 LT |
932 | .get_xstate = linvfs_getxstate, |
933 | .set_xstate = linvfs_setxstate, | |
934 | .get_xquota = linvfs_getxquota, | |
935 | .set_xquota = linvfs_setxquota, | |
936 | }; | |
937 | ||
938 | STATIC struct file_system_type xfs_fs_type = { | |
939 | .owner = THIS_MODULE, | |
940 | .name = "xfs", | |
941 | .get_sb = linvfs_get_sb, | |
942 | .kill_sb = kill_block_super, | |
943 | .fs_flags = FS_REQUIRES_DEV, | |
944 | }; | |
945 | ||
946 | ||
947 | STATIC int __init | |
948 | init_xfs_fs( void ) | |
949 | { | |
950 | int error; | |
951 | struct sysinfo si; | |
952 | static char message[] __initdata = KERN_INFO \ | |
953 | XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n"; | |
954 | ||
955 | printk(message); | |
956 | ||
957 | si_meminfo(&si); | |
958 | xfs_physmem = si.totalram; | |
959 | ||
960 | ktrace_init(64); | |
961 | ||
8758280f | 962 | error = xfs_init_zones(); |
1da177e4 | 963 | if (error < 0) |
0829c360 | 964 | goto undo_zones; |
1da177e4 | 965 | |
ce8e922c | 966 | error = xfs_buf_init(); |
1da177e4 | 967 | if (error < 0) |
ce8e922c | 968 | goto undo_buffers; |
1da177e4 LT |
969 | |
970 | vn_init(); | |
971 | xfs_init(); | |
972 | uuid_init(); | |
973 | vfs_initquota(); | |
974 | ||
975 | error = register_filesystem(&xfs_fs_type); | |
976 | if (error) | |
977 | goto undo_register; | |
978 | XFS_DM_INIT(&xfs_fs_type); | |
979 | return 0; | |
980 | ||
981 | undo_register: | |
ce8e922c | 982 | xfs_buf_terminate(); |
1da177e4 | 983 | |
ce8e922c | 984 | undo_buffers: |
8758280f | 985 | xfs_destroy_zones(); |
1da177e4 | 986 | |
0829c360 | 987 | undo_zones: |
1da177e4 LT |
988 | return error; |
989 | } | |
990 | ||
991 | STATIC void __exit | |
992 | exit_xfs_fs( void ) | |
993 | { | |
994 | vfs_exitquota(); | |
995 | XFS_DM_EXIT(&xfs_fs_type); | |
996 | unregister_filesystem(&xfs_fs_type); | |
997 | xfs_cleanup(); | |
ce8e922c | 998 | xfs_buf_terminate(); |
8758280f | 999 | xfs_destroy_zones(); |
1da177e4 LT |
1000 | ktrace_uninit(); |
1001 | } | |
1002 | ||
1003 | module_init(init_xfs_fs); | |
1004 | module_exit(exit_xfs_fs); | |
1005 | ||
1006 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
1007 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
1008 | MODULE_LICENSE("GPL"); |