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 */ | |
79 | if (sb->s_flags & MS_NOATIME) | |
80 | args->flags |= XFSMNT_NOATIME; | |
81 | if (sb->s_flags & MS_DIRSYNC) | |
82 | args->flags |= XFSMNT_DIRSYNC; | |
83 | if (sb->s_flags & MS_SYNCHRONOUS) | |
84 | args->flags |= XFSMNT_WSYNC; | |
85 | ||
86 | /* Default to 32 bit inodes on Linux all the time */ | |
87 | args->flags |= XFSMNT_32BITINODES; | |
88 | ||
89 | return args; | |
90 | } | |
91 | ||
92 | __uint64_t | |
93 | xfs_max_file_offset( | |
94 | unsigned int blockshift) | |
95 | { | |
96 | unsigned int pagefactor = 1; | |
97 | unsigned int bitshift = BITS_PER_LONG - 1; | |
98 | ||
99 | /* Figure out maximum filesize, on Linux this can depend on | |
100 | * the filesystem blocksize (on 32 bit platforms). | |
101 | * __block_prepare_write does this in an [unsigned] long... | |
102 | * page->index << (PAGE_CACHE_SHIFT - bbits) | |
103 | * So, for page sized blocks (4K on 32 bit platforms), | |
104 | * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is | |
105 | * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1) | |
106 | * but for smaller blocksizes it is less (bbits = log2 bsize). | |
107 | * Note1: get_block_t takes a long (implicit cast from above) | |
108 | * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch | |
109 | * can optionally convert the [unsigned] long from above into | |
110 | * an [unsigned] long long. | |
111 | */ | |
112 | ||
113 | #if BITS_PER_LONG == 32 | |
114 | # if defined(CONFIG_LBD) | |
115 | ASSERT(sizeof(sector_t) == 8); | |
116 | pagefactor = PAGE_CACHE_SIZE; | |
117 | bitshift = BITS_PER_LONG; | |
118 | # else | |
119 | pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift); | |
120 | # endif | |
121 | #endif | |
122 | ||
123 | return (((__uint64_t)pagefactor) << bitshift) - 1; | |
124 | } | |
125 | ||
126 | STATIC __inline__ void | |
127 | xfs_set_inodeops( | |
128 | struct inode *inode) | |
129 | { | |
0432dab2 CH |
130 | switch (inode->i_mode & S_IFMT) { |
131 | case S_IFREG: | |
1da177e4 LT |
132 | inode->i_op = &linvfs_file_inode_operations; |
133 | inode->i_fop = &linvfs_file_operations; | |
134 | inode->i_mapping->a_ops = &linvfs_aops; | |
0432dab2 CH |
135 | break; |
136 | case S_IFDIR: | |
1da177e4 LT |
137 | inode->i_op = &linvfs_dir_inode_operations; |
138 | inode->i_fop = &linvfs_dir_operations; | |
0432dab2 CH |
139 | break; |
140 | case S_IFLNK: | |
1da177e4 LT |
141 | inode->i_op = &linvfs_symlink_inode_operations; |
142 | if (inode->i_blocks) | |
143 | inode->i_mapping->a_ops = &linvfs_aops; | |
0432dab2 CH |
144 | break; |
145 | default: | |
1da177e4 LT |
146 | inode->i_op = &linvfs_file_inode_operations; |
147 | init_special_inode(inode, inode->i_mode, inode->i_rdev); | |
0432dab2 | 148 | break; |
1da177e4 LT |
149 | } |
150 | } | |
151 | ||
152 | STATIC __inline__ void | |
153 | xfs_revalidate_inode( | |
154 | xfs_mount_t *mp, | |
155 | vnode_t *vp, | |
156 | xfs_inode_t *ip) | |
157 | { | |
158 | struct inode *inode = LINVFS_GET_IP(vp); | |
159 | ||
0432dab2 | 160 | inode->i_mode = ip->i_d.di_mode; |
1da177e4 LT |
161 | inode->i_nlink = ip->i_d.di_nlink; |
162 | inode->i_uid = ip->i_d.di_uid; | |
163 | inode->i_gid = ip->i_d.di_gid; | |
0432dab2 CH |
164 | |
165 | switch (inode->i_mode & S_IFMT) { | |
166 | case S_IFBLK: | |
167 | case S_IFCHR: | |
168 | inode->i_rdev = | |
169 | MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff, | |
170 | sysv_minor(ip->i_df.if_u2.if_rdev)); | |
171 | break; | |
172 | default: | |
1da177e4 | 173 | inode->i_rdev = 0; |
0432dab2 | 174 | break; |
1da177e4 | 175 | } |
0432dab2 | 176 | |
e8c8b3a7 | 177 | inode->i_blksize = xfs_preferred_iosize(mp); |
1da177e4 LT |
178 | inode->i_generation = ip->i_d.di_gen; |
179 | i_size_write(inode, ip->i_d.di_size); | |
180 | inode->i_blocks = | |
181 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
182 | inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec; | |
183 | inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec; | |
184 | inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec; | |
185 | inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec; | |
186 | inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec; | |
187 | inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec; | |
188 | if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE) | |
189 | inode->i_flags |= S_IMMUTABLE; | |
190 | else | |
191 | inode->i_flags &= ~S_IMMUTABLE; | |
192 | if (ip->i_d.di_flags & XFS_DIFLAG_APPEND) | |
193 | inode->i_flags |= S_APPEND; | |
194 | else | |
195 | inode->i_flags &= ~S_APPEND; | |
196 | if (ip->i_d.di_flags & XFS_DIFLAG_SYNC) | |
197 | inode->i_flags |= S_SYNC; | |
198 | else | |
199 | inode->i_flags &= ~S_SYNC; | |
200 | if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME) | |
201 | inode->i_flags |= S_NOATIME; | |
202 | else | |
203 | inode->i_flags &= ~S_NOATIME; | |
204 | vp->v_flag &= ~VMODIFIED; | |
205 | } | |
206 | ||
207 | void | |
208 | xfs_initialize_vnode( | |
209 | bhv_desc_t *bdp, | |
210 | vnode_t *vp, | |
211 | bhv_desc_t *inode_bhv, | |
212 | int unlock) | |
213 | { | |
214 | xfs_inode_t *ip = XFS_BHVTOI(inode_bhv); | |
215 | struct inode *inode = LINVFS_GET_IP(vp); | |
216 | ||
217 | if (!inode_bhv->bd_vobj) { | |
218 | vp->v_vfsp = bhvtovfs(bdp); | |
219 | bhv_desc_init(inode_bhv, ip, vp, &xfs_vnodeops); | |
220 | bhv_insert(VN_BHV_HEAD(vp), inode_bhv); | |
221 | } | |
222 | ||
223 | /* | |
224 | * We need to set the ops vectors, and unlock the inode, but if | |
225 | * we have been called during the new inode create process, it is | |
226 | * too early to fill in the Linux inode. We will get called a | |
227 | * second time once the inode is properly set up, and then we can | |
228 | * finish our work. | |
229 | */ | |
230 | if (ip->i_d.di_mode != 0 && unlock && (inode->i_state & I_NEW)) { | |
1da177e4 LT |
231 | xfs_revalidate_inode(XFS_BHVTOM(bdp), vp, ip); |
232 | xfs_set_inodeops(inode); | |
233 | ||
234 | ip->i_flags &= ~XFS_INEW; | |
235 | barrier(); | |
236 | ||
237 | unlock_new_inode(inode); | |
238 | } | |
239 | } | |
240 | ||
241 | int | |
242 | xfs_blkdev_get( | |
243 | xfs_mount_t *mp, | |
244 | const char *name, | |
245 | struct block_device **bdevp) | |
246 | { | |
247 | int error = 0; | |
248 | ||
249 | *bdevp = open_bdev_excl(name, 0, mp); | |
250 | if (IS_ERR(*bdevp)) { | |
251 | error = PTR_ERR(*bdevp); | |
252 | printk("XFS: Invalid device [%s], error=%d\n", name, error); | |
253 | } | |
254 | ||
255 | return -error; | |
256 | } | |
257 | ||
258 | void | |
259 | xfs_blkdev_put( | |
260 | struct block_device *bdev) | |
261 | { | |
262 | if (bdev) | |
263 | close_bdev_excl(bdev); | |
264 | } | |
265 | ||
f538d4da CH |
266 | /* |
267 | * Try to write out the superblock using barriers. | |
268 | */ | |
269 | STATIC int | |
270 | xfs_barrier_test( | |
271 | xfs_mount_t *mp) | |
272 | { | |
273 | xfs_buf_t *sbp = xfs_getsb(mp, 0); | |
274 | int error; | |
275 | ||
276 | XFS_BUF_UNDONE(sbp); | |
277 | XFS_BUF_UNREAD(sbp); | |
278 | XFS_BUF_UNDELAYWRITE(sbp); | |
279 | XFS_BUF_WRITE(sbp); | |
280 | XFS_BUF_UNASYNC(sbp); | |
281 | XFS_BUF_ORDERED(sbp); | |
282 | ||
283 | xfsbdstrat(mp, sbp); | |
284 | error = xfs_iowait(sbp); | |
285 | ||
286 | /* | |
287 | * Clear all the flags we set and possible error state in the | |
288 | * buffer. We only did the write to try out whether barriers | |
289 | * worked and shouldn't leave any traces in the superblock | |
290 | * buffer. | |
291 | */ | |
292 | XFS_BUF_DONE(sbp); | |
293 | XFS_BUF_ERROR(sbp, 0); | |
294 | XFS_BUF_UNORDERED(sbp); | |
295 | ||
296 | xfs_buf_relse(sbp); | |
297 | return error; | |
298 | } | |
299 | ||
300 | void | |
301 | xfs_mountfs_check_barriers(xfs_mount_t *mp) | |
302 | { | |
303 | int error; | |
304 | ||
305 | if (mp->m_logdev_targp != mp->m_ddev_targp) { | |
306 | xfs_fs_cmn_err(CE_NOTE, mp, | |
307 | "Disabling barriers, not supported with external log device"); | |
308 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 309 | return; |
f538d4da CH |
310 | } |
311 | ||
312 | if (mp->m_ddev_targp->pbr_bdev->bd_disk->queue->ordered == | |
313 | QUEUE_ORDERED_NONE) { | |
314 | xfs_fs_cmn_err(CE_NOTE, mp, | |
315 | "Disabling barriers, not supported by the underlying device"); | |
316 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 317 | return; |
f538d4da CH |
318 | } |
319 | ||
320 | error = xfs_barrier_test(mp); | |
321 | if (error) { | |
322 | xfs_fs_cmn_err(CE_NOTE, mp, | |
323 | "Disabling barriers, trial barrier write failed"); | |
324 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 325 | return; |
f538d4da CH |
326 | } |
327 | } | |
328 | ||
329 | void | |
330 | xfs_blkdev_issue_flush( | |
331 | xfs_buftarg_t *buftarg) | |
332 | { | |
333 | blkdev_issue_flush(buftarg->pbr_bdev, NULL); | |
334 | } | |
1da177e4 LT |
335 | |
336 | STATIC struct inode * | |
337 | linvfs_alloc_inode( | |
338 | struct super_block *sb) | |
339 | { | |
340 | vnode_t *vp; | |
341 | ||
0829c360 | 342 | vp = kmem_cache_alloc(xfs_vnode_zone, kmem_flags_convert(KM_SLEEP)); |
1da177e4 LT |
343 | if (!vp) |
344 | return NULL; | |
345 | return LINVFS_GET_IP(vp); | |
346 | } | |
347 | ||
348 | STATIC void | |
349 | linvfs_destroy_inode( | |
350 | struct inode *inode) | |
351 | { | |
0829c360 | 352 | kmem_zone_free(xfs_vnode_zone, LINVFS_GET_VP(inode)); |
1da177e4 LT |
353 | } |
354 | ||
355 | STATIC void | |
0829c360 | 356 | linvfs_inode_init_once( |
1da177e4 LT |
357 | void *data, |
358 | kmem_cache_t *cachep, | |
359 | unsigned long flags) | |
360 | { | |
361 | vnode_t *vp = (vnode_t *)data; | |
362 | ||
363 | if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == | |
364 | SLAB_CTOR_CONSTRUCTOR) | |
365 | inode_init_once(LINVFS_GET_IP(vp)); | |
366 | } | |
367 | ||
368 | STATIC int | |
0829c360 | 369 | linvfs_init_zones(void) |
1da177e4 | 370 | { |
0829c360 | 371 | xfs_vnode_zone = kmem_cache_create("xfs_vnode", |
1da177e4 | 372 | sizeof(vnode_t), 0, SLAB_RECLAIM_ACCOUNT, |
0829c360 CH |
373 | linvfs_inode_init_once, NULL); |
374 | if (!xfs_vnode_zone) | |
375 | goto out; | |
376 | ||
377 | xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend"); | |
378 | if (!xfs_ioend_zone) | |
379 | goto out_destroy_vnode_zone; | |
380 | ||
381 | xfs_ioend_pool = mempool_create(4 * MAX_BUF_PER_PAGE, | |
382 | mempool_alloc_slab, mempool_free_slab, | |
383 | xfs_ioend_zone); | |
384 | if (!xfs_ioend_pool) | |
385 | goto out_free_ioend_zone; | |
386 | ||
1da177e4 | 387 | return 0; |
0829c360 CH |
388 | |
389 | ||
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 | |
0829c360 | 399 | linvfs_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, | |
410 | * since this is when the inode itself becomes flushable. | |
411 | */ | |
412 | STATIC int | |
413 | linvfs_write_inode( | |
414 | struct inode *inode, | |
415 | int sync) | |
416 | { | |
417 | vnode_t *vp = LINVFS_GET_VP(inode); | |
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; | |
424 | VOP_IFLUSH(vp, flags, error); | |
425 | if (error == EAGAIN) { | |
426 | if (sync) | |
427 | VOP_IFLUSH(vp, flags | FLUSH_LOG, error); | |
428 | else | |
429 | error = 0; | |
430 | } | |
431 | } | |
432 | ||
433 | return -error; | |
434 | } | |
435 | ||
436 | STATIC void | |
437 | linvfs_clear_inode( | |
438 | struct inode *inode) | |
439 | { | |
440 | vnode_t *vp = LINVFS_GET_VP(inode); | |
56d433e4 | 441 | int error, cache; |
1da177e4 | 442 | |
56d433e4 CH |
443 | vn_trace_entry(vp, "clear_inode", (inst_t *)__return_address); |
444 | ||
56d433e4 CH |
445 | XFS_STATS_INC(vn_rele); |
446 | XFS_STATS_INC(vn_remove); | |
447 | XFS_STATS_INC(vn_reclaim); | |
448 | XFS_STATS_DEC(vn_active); | |
449 | ||
02ba71de CH |
450 | /* |
451 | * This can happen because xfs_iget_core calls xfs_idestroy if we | |
452 | * find an inode with di_mode == 0 but without IGET_CREATE set. | |
453 | */ | |
454 | if (vp->v_fbhv) | |
455 | VOP_INACTIVE(vp, NULL, cache); | |
1da177e4 | 456 | |
56d433e4 CH |
457 | VN_LOCK(vp); |
458 | vp->v_flag &= ~VMODIFIED; | |
459 | VN_UNLOCK(vp, 0); | |
460 | ||
0c147f9a FB |
461 | if (vp->v_fbhv) { |
462 | VOP_RECLAIM(vp, error); | |
463 | if (error) | |
464 | panic("vn_purge: cannot reclaim"); | |
465 | } | |
56d433e4 CH |
466 | |
467 | ASSERT(vp->v_fbhv == NULL); | |
468 | ||
469 | #ifdef XFS_VNODE_TRACE | |
470 | ktrace_free(vp->v_trace); | |
471 | #endif | |
472 | } | |
1da177e4 LT |
473 | |
474 | /* | |
475 | * Enqueue a work item to be picked up by the vfs xfssyncd thread. | |
476 | * Doing this has two advantages: | |
477 | * - It saves on stack space, which is tight in certain situations | |
478 | * - It can be used (with care) as a mechanism to avoid deadlocks. | |
479 | * Flushing while allocating in a full filesystem requires both. | |
480 | */ | |
481 | STATIC void | |
482 | xfs_syncd_queue_work( | |
483 | struct vfs *vfs, | |
484 | void *data, | |
485 | void (*syncer)(vfs_t *, void *)) | |
486 | { | |
487 | vfs_sync_work_t *work; | |
488 | ||
489 | work = kmem_alloc(sizeof(struct vfs_sync_work), KM_SLEEP); | |
490 | INIT_LIST_HEAD(&work->w_list); | |
491 | work->w_syncer = syncer; | |
492 | work->w_data = data; | |
493 | work->w_vfs = vfs; | |
494 | spin_lock(&vfs->vfs_sync_lock); | |
495 | list_add_tail(&work->w_list, &vfs->vfs_sync_list); | |
496 | spin_unlock(&vfs->vfs_sync_lock); | |
497 | wake_up_process(vfs->vfs_sync_task); | |
498 | } | |
499 | ||
500 | /* | |
501 | * Flush delayed allocate data, attempting to free up reserved space | |
502 | * from existing allocations. At this point a new allocation attempt | |
503 | * has failed with ENOSPC and we are in the process of scratching our | |
504 | * heads, looking about for more room... | |
505 | */ | |
506 | STATIC void | |
507 | xfs_flush_inode_work( | |
508 | vfs_t *vfs, | |
509 | void *inode) | |
510 | { | |
511 | filemap_flush(((struct inode *)inode)->i_mapping); | |
512 | iput((struct inode *)inode); | |
513 | } | |
514 | ||
515 | void | |
516 | xfs_flush_inode( | |
517 | xfs_inode_t *ip) | |
518 | { | |
519 | struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); | |
520 | struct vfs *vfs = XFS_MTOVFS(ip->i_mount); | |
521 | ||
522 | igrab(inode); | |
523 | xfs_syncd_queue_work(vfs, inode, xfs_flush_inode_work); | |
041e0e3b | 524 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
525 | } |
526 | ||
527 | /* | |
528 | * This is the "bigger hammer" version of xfs_flush_inode_work... | |
529 | * (IOW, "If at first you don't succeed, use a Bigger Hammer"). | |
530 | */ | |
531 | STATIC void | |
532 | xfs_flush_device_work( | |
533 | vfs_t *vfs, | |
534 | void *inode) | |
535 | { | |
536 | sync_blockdev(vfs->vfs_super->s_bdev); | |
537 | iput((struct inode *)inode); | |
538 | } | |
539 | ||
540 | void | |
541 | xfs_flush_device( | |
542 | xfs_inode_t *ip) | |
543 | { | |
544 | struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); | |
545 | struct vfs *vfs = XFS_MTOVFS(ip->i_mount); | |
546 | ||
547 | igrab(inode); | |
548 | xfs_syncd_queue_work(vfs, inode, xfs_flush_device_work); | |
041e0e3b | 549 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
550 | xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC); |
551 | } | |
552 | ||
553 | #define SYNCD_FLAGS (SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR) | |
554 | STATIC void | |
555 | vfs_sync_worker( | |
556 | vfs_t *vfsp, | |
557 | void *unused) | |
558 | { | |
559 | int error; | |
560 | ||
561 | if (!(vfsp->vfs_flag & VFS_RDONLY)) | |
562 | VFS_SYNC(vfsp, SYNCD_FLAGS, NULL, error); | |
563 | vfsp->vfs_sync_seq++; | |
564 | wmb(); | |
565 | wake_up(&vfsp->vfs_wait_single_sync_task); | |
566 | } | |
567 | ||
568 | STATIC int | |
569 | xfssyncd( | |
570 | void *arg) | |
571 | { | |
572 | long timeleft; | |
573 | vfs_t *vfsp = (vfs_t *) arg; | |
1da177e4 | 574 | struct vfs_sync_work *work, *n; |
4df08c52 | 575 | LIST_HEAD (tmp); |
1da177e4 | 576 | |
041e0e3b | 577 | timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10); |
1da177e4 | 578 | for (;;) { |
041e0e3b | 579 | timeleft = schedule_timeout_interruptible(timeleft); |
1da177e4 | 580 | /* swsusp */ |
3e1d1d28 | 581 | try_to_freeze(); |
4df08c52 | 582 | if (kthread_should_stop()) |
1da177e4 LT |
583 | break; |
584 | ||
585 | spin_lock(&vfsp->vfs_sync_lock); | |
586 | /* | |
587 | * We can get woken by laptop mode, to do a sync - | |
588 | * that's the (only!) case where the list would be | |
589 | * empty with time remaining. | |
590 | */ | |
591 | if (!timeleft || list_empty(&vfsp->vfs_sync_list)) { | |
592 | if (!timeleft) | |
041e0e3b NA |
593 | timeleft = xfs_syncd_centisecs * |
594 | msecs_to_jiffies(10); | |
1da177e4 LT |
595 | INIT_LIST_HEAD(&vfsp->vfs_sync_work.w_list); |
596 | list_add_tail(&vfsp->vfs_sync_work.w_list, | |
597 | &vfsp->vfs_sync_list); | |
598 | } | |
599 | list_for_each_entry_safe(work, n, &vfsp->vfs_sync_list, w_list) | |
600 | list_move(&work->w_list, &tmp); | |
601 | spin_unlock(&vfsp->vfs_sync_lock); | |
602 | ||
603 | list_for_each_entry_safe(work, n, &tmp, w_list) { | |
604 | (*work->w_syncer)(vfsp, work->w_data); | |
605 | list_del(&work->w_list); | |
606 | if (work == &vfsp->vfs_sync_work) | |
607 | continue; | |
608 | kmem_free(work, sizeof(struct vfs_sync_work)); | |
609 | } | |
610 | } | |
611 | ||
1da177e4 LT |
612 | return 0; |
613 | } | |
614 | ||
615 | STATIC int | |
616 | linvfs_start_syncd( | |
617 | vfs_t *vfsp) | |
618 | { | |
4df08c52 CH |
619 | vfsp->vfs_sync_work.w_syncer = vfs_sync_worker; |
620 | vfsp->vfs_sync_work.w_vfs = vfsp; | |
621 | vfsp->vfs_sync_task = kthread_run(xfssyncd, vfsp, "xfssyncd"); | |
622 | if (IS_ERR(vfsp->vfs_sync_task)) | |
623 | return -PTR_ERR(vfsp->vfs_sync_task); | |
1da177e4 LT |
624 | return 0; |
625 | } | |
626 | ||
627 | STATIC void | |
628 | linvfs_stop_syncd( | |
629 | vfs_t *vfsp) | |
630 | { | |
4df08c52 | 631 | kthread_stop(vfsp->vfs_sync_task); |
1da177e4 LT |
632 | } |
633 | ||
634 | STATIC void | |
635 | linvfs_put_super( | |
636 | struct super_block *sb) | |
637 | { | |
638 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
639 | int error; | |
640 | ||
641 | linvfs_stop_syncd(vfsp); | |
642 | VFS_SYNC(vfsp, SYNC_ATTR|SYNC_DELWRI, NULL, error); | |
643 | if (!error) | |
644 | VFS_UNMOUNT(vfsp, 0, NULL, error); | |
645 | if (error) { | |
646 | printk("XFS unmount got error %d\n", error); | |
647 | printk("%s: vfsp/0x%p left dangling!\n", __FUNCTION__, vfsp); | |
648 | return; | |
649 | } | |
650 | ||
651 | vfs_deallocate(vfsp); | |
652 | } | |
653 | ||
654 | STATIC void | |
655 | linvfs_write_super( | |
656 | struct super_block *sb) | |
657 | { | |
658 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
659 | int error; | |
660 | ||
661 | if (sb->s_flags & MS_RDONLY) { | |
662 | sb->s_dirt = 0; /* paranoia */ | |
663 | return; | |
664 | } | |
665 | /* Push the log and superblock a little */ | |
666 | VFS_SYNC(vfsp, SYNC_FSDATA, NULL, error); | |
667 | sb->s_dirt = 0; | |
668 | } | |
669 | ||
670 | STATIC int | |
671 | linvfs_sync_super( | |
672 | struct super_block *sb, | |
673 | int wait) | |
674 | { | |
675 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
676 | int error; | |
677 | int flags = SYNC_FSDATA; | |
678 | ||
f898d6c0 CH |
679 | if (unlikely(sb->s_frozen == SB_FREEZE_WRITE)) |
680 | flags = SYNC_QUIESCE; | |
681 | else | |
682 | flags = SYNC_FSDATA | (wait ? SYNC_WAIT : 0); | |
1da177e4 LT |
683 | |
684 | VFS_SYNC(vfsp, flags, NULL, error); | |
685 | sb->s_dirt = 0; | |
686 | ||
687 | if (unlikely(laptop_mode)) { | |
688 | int prev_sync_seq = vfsp->vfs_sync_seq; | |
689 | ||
690 | /* | |
691 | * The disk must be active because we're syncing. | |
692 | * We schedule xfssyncd now (now that the disk is | |
693 | * active) instead of later (when it might not be). | |
694 | */ | |
695 | wake_up_process(vfsp->vfs_sync_task); | |
696 | /* | |
697 | * We have to wait for the sync iteration to complete. | |
698 | * If we don't, the disk activity caused by the sync | |
699 | * will come after the sync is completed, and that | |
700 | * triggers another sync from laptop mode. | |
701 | */ | |
702 | wait_event(vfsp->vfs_wait_single_sync_task, | |
703 | vfsp->vfs_sync_seq != prev_sync_seq); | |
704 | } | |
705 | ||
706 | return -error; | |
707 | } | |
708 | ||
709 | STATIC int | |
710 | linvfs_statfs( | |
711 | struct super_block *sb, | |
712 | struct kstatfs *statp) | |
713 | { | |
714 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
715 | int error; | |
716 | ||
717 | VFS_STATVFS(vfsp, statp, NULL, error); | |
718 | return -error; | |
719 | } | |
720 | ||
721 | STATIC int | |
722 | linvfs_remount( | |
723 | struct super_block *sb, | |
724 | int *flags, | |
725 | char *options) | |
726 | { | |
727 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
728 | struct xfs_mount_args *args = xfs_args_allocate(sb); | |
729 | int error; | |
730 | ||
731 | VFS_PARSEARGS(vfsp, options, args, 1, error); | |
732 | if (!error) | |
733 | VFS_MNTUPDATE(vfsp, flags, args, error); | |
734 | kmem_free(args, sizeof(*args)); | |
735 | return -error; | |
736 | } | |
737 | ||
738 | STATIC void | |
739 | linvfs_freeze_fs( | |
740 | struct super_block *sb) | |
741 | { | |
742 | VFS_FREEZE(LINVFS_GET_VFS(sb)); | |
743 | } | |
744 | ||
745 | STATIC int | |
746 | linvfs_show_options( | |
747 | struct seq_file *m, | |
748 | struct vfsmount *mnt) | |
749 | { | |
750 | struct vfs *vfsp = LINVFS_GET_VFS(mnt->mnt_sb); | |
751 | int error; | |
752 | ||
753 | VFS_SHOWARGS(vfsp, m, error); | |
754 | return error; | |
755 | } | |
756 | ||
ee34807a NS |
757 | STATIC int |
758 | linvfs_quotasync( | |
759 | struct super_block *sb, | |
760 | int type) | |
761 | { | |
762 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
763 | int error; | |
764 | ||
765 | VFS_QUOTACTL(vfsp, Q_XQUOTASYNC, 0, (caddr_t)NULL, error); | |
766 | return -error; | |
767 | } | |
768 | ||
1da177e4 LT |
769 | STATIC int |
770 | linvfs_getxstate( | |
771 | struct super_block *sb, | |
772 | struct fs_quota_stat *fqs) | |
773 | { | |
774 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
775 | int error; | |
776 | ||
777 | VFS_QUOTACTL(vfsp, Q_XGETQSTAT, 0, (caddr_t)fqs, error); | |
778 | return -error; | |
779 | } | |
780 | ||
781 | STATIC int | |
782 | linvfs_setxstate( | |
783 | struct super_block *sb, | |
784 | unsigned int flags, | |
785 | int op) | |
786 | { | |
787 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
788 | int error; | |
789 | ||
790 | VFS_QUOTACTL(vfsp, op, 0, (caddr_t)&flags, error); | |
791 | return -error; | |
792 | } | |
793 | ||
794 | STATIC int | |
795 | linvfs_getxquota( | |
796 | struct super_block *sb, | |
797 | int type, | |
798 | qid_t id, | |
799 | struct fs_disk_quota *fdq) | |
800 | { | |
801 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
802 | int error, getmode; | |
803 | ||
c8ad20ff NS |
804 | getmode = (type == USRQUOTA) ? Q_XGETQUOTA : |
805 | ((type == GRPQUOTA) ? Q_XGETGQUOTA : Q_XGETPQUOTA); | |
1da177e4 LT |
806 | VFS_QUOTACTL(vfsp, getmode, id, (caddr_t)fdq, error); |
807 | return -error; | |
808 | } | |
809 | ||
810 | STATIC int | |
811 | linvfs_setxquota( | |
812 | struct super_block *sb, | |
813 | int type, | |
814 | qid_t id, | |
815 | struct fs_disk_quota *fdq) | |
816 | { | |
817 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
818 | int error, setmode; | |
819 | ||
c8ad20ff NS |
820 | setmode = (type == USRQUOTA) ? Q_XSETQLIM : |
821 | ((type == GRPQUOTA) ? Q_XSETGQLIM : Q_XSETPQLIM); | |
1da177e4 LT |
822 | VFS_QUOTACTL(vfsp, setmode, id, (caddr_t)fdq, error); |
823 | return -error; | |
824 | } | |
825 | ||
826 | STATIC int | |
827 | linvfs_fill_super( | |
828 | struct super_block *sb, | |
829 | void *data, | |
830 | int silent) | |
831 | { | |
832 | vnode_t *rootvp; | |
833 | struct vfs *vfsp = vfs_allocate(); | |
834 | struct xfs_mount_args *args = xfs_args_allocate(sb); | |
835 | struct kstatfs statvfs; | |
836 | int error, error2; | |
837 | ||
838 | vfsp->vfs_super = sb; | |
839 | LINVFS_SET_VFS(sb, vfsp); | |
840 | if (sb->s_flags & MS_RDONLY) | |
841 | vfsp->vfs_flag |= VFS_RDONLY; | |
842 | bhv_insert_all_vfsops(vfsp); | |
843 | ||
844 | VFS_PARSEARGS(vfsp, (char *)data, args, 0, error); | |
845 | if (error) { | |
846 | bhv_remove_all_vfsops(vfsp, 1); | |
847 | goto fail_vfsop; | |
848 | } | |
849 | ||
850 | sb_min_blocksize(sb, BBSIZE); | |
851 | #ifdef CONFIG_XFS_EXPORT | |
852 | sb->s_export_op = &linvfs_export_ops; | |
853 | #endif | |
854 | sb->s_qcop = &linvfs_qops; | |
855 | sb->s_op = &linvfs_sops; | |
856 | ||
857 | VFS_MOUNT(vfsp, args, NULL, error); | |
858 | if (error) { | |
859 | bhv_remove_all_vfsops(vfsp, 1); | |
860 | goto fail_vfsop; | |
861 | } | |
862 | ||
863 | VFS_STATVFS(vfsp, &statvfs, NULL, error); | |
864 | if (error) | |
865 | goto fail_unmount; | |
866 | ||
867 | sb->s_dirt = 1; | |
868 | sb->s_magic = statvfs.f_type; | |
869 | sb->s_blocksize = statvfs.f_bsize; | |
870 | sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1; | |
871 | sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); | |
872 | sb->s_time_gran = 1; | |
873 | set_posix_acl_flag(sb); | |
874 | ||
875 | VFS_ROOT(vfsp, &rootvp, error); | |
876 | if (error) | |
877 | goto fail_unmount; | |
878 | ||
879 | sb->s_root = d_alloc_root(LINVFS_GET_IP(rootvp)); | |
880 | if (!sb->s_root) { | |
881 | error = ENOMEM; | |
882 | goto fail_vnrele; | |
883 | } | |
884 | if (is_bad_inode(sb->s_root->d_inode)) { | |
885 | error = EINVAL; | |
886 | goto fail_vnrele; | |
887 | } | |
888 | if ((error = linvfs_start_syncd(vfsp))) | |
889 | goto fail_vnrele; | |
890 | vn_trace_exit(rootvp, __FUNCTION__, (inst_t *)__return_address); | |
891 | ||
892 | kmem_free(args, sizeof(*args)); | |
893 | return 0; | |
894 | ||
895 | fail_vnrele: | |
896 | if (sb->s_root) { | |
897 | dput(sb->s_root); | |
898 | sb->s_root = NULL; | |
899 | } else { | |
900 | VN_RELE(rootvp); | |
901 | } | |
902 | ||
903 | fail_unmount: | |
904 | VFS_UNMOUNT(vfsp, 0, NULL, error2); | |
905 | ||
906 | fail_vfsop: | |
907 | vfs_deallocate(vfsp); | |
908 | kmem_free(args, sizeof(*args)); | |
909 | return -error; | |
910 | } | |
911 | ||
912 | STATIC struct super_block * | |
913 | linvfs_get_sb( | |
914 | struct file_system_type *fs_type, | |
915 | int flags, | |
916 | const char *dev_name, | |
917 | void *data) | |
918 | { | |
919 | return get_sb_bdev(fs_type, flags, dev_name, data, linvfs_fill_super); | |
920 | } | |
921 | ||
922 | STATIC struct super_operations linvfs_sops = { | |
923 | .alloc_inode = linvfs_alloc_inode, | |
924 | .destroy_inode = linvfs_destroy_inode, | |
925 | .write_inode = linvfs_write_inode, | |
926 | .clear_inode = linvfs_clear_inode, | |
927 | .put_super = linvfs_put_super, | |
928 | .write_super = linvfs_write_super, | |
929 | .sync_fs = linvfs_sync_super, | |
930 | .write_super_lockfs = linvfs_freeze_fs, | |
931 | .statfs = linvfs_statfs, | |
932 | .remount_fs = linvfs_remount, | |
933 | .show_options = linvfs_show_options, | |
934 | }; | |
935 | ||
936 | STATIC struct quotactl_ops linvfs_qops = { | |
ee34807a | 937 | .quota_sync = linvfs_quotasync, |
1da177e4 LT |
938 | .get_xstate = linvfs_getxstate, |
939 | .set_xstate = linvfs_setxstate, | |
940 | .get_xquota = linvfs_getxquota, | |
941 | .set_xquota = linvfs_setxquota, | |
942 | }; | |
943 | ||
944 | STATIC struct file_system_type xfs_fs_type = { | |
945 | .owner = THIS_MODULE, | |
946 | .name = "xfs", | |
947 | .get_sb = linvfs_get_sb, | |
948 | .kill_sb = kill_block_super, | |
949 | .fs_flags = FS_REQUIRES_DEV, | |
950 | }; | |
951 | ||
952 | ||
953 | STATIC int __init | |
954 | init_xfs_fs( void ) | |
955 | { | |
956 | int error; | |
957 | struct sysinfo si; | |
958 | static char message[] __initdata = KERN_INFO \ | |
959 | XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n"; | |
960 | ||
961 | printk(message); | |
962 | ||
963 | si_meminfo(&si); | |
964 | xfs_physmem = si.totalram; | |
965 | ||
966 | ktrace_init(64); | |
967 | ||
0829c360 | 968 | error = linvfs_init_zones(); |
1da177e4 | 969 | if (error < 0) |
0829c360 | 970 | goto undo_zones; |
1da177e4 LT |
971 | |
972 | error = pagebuf_init(); | |
973 | if (error < 0) | |
974 | goto undo_pagebuf; | |
975 | ||
976 | vn_init(); | |
977 | xfs_init(); | |
978 | uuid_init(); | |
979 | vfs_initquota(); | |
980 | ||
981 | error = register_filesystem(&xfs_fs_type); | |
982 | if (error) | |
983 | goto undo_register; | |
984 | XFS_DM_INIT(&xfs_fs_type); | |
985 | return 0; | |
986 | ||
987 | undo_register: | |
988 | pagebuf_terminate(); | |
989 | ||
990 | undo_pagebuf: | |
0829c360 | 991 | linvfs_destroy_zones(); |
1da177e4 | 992 | |
0829c360 | 993 | undo_zones: |
1da177e4 LT |
994 | return error; |
995 | } | |
996 | ||
997 | STATIC void __exit | |
998 | exit_xfs_fs( void ) | |
999 | { | |
1000 | vfs_exitquota(); | |
1001 | XFS_DM_EXIT(&xfs_fs_type); | |
1002 | unregister_filesystem(&xfs_fs_type); | |
1003 | xfs_cleanup(); | |
1004 | pagebuf_terminate(); | |
0829c360 | 1005 | linvfs_destroy_zones(); |
1da177e4 LT |
1006 | ktrace_uninit(); |
1007 | } | |
1008 | ||
1009 | module_init(init_xfs_fs); | |
1010 | module_exit(exit_xfs_fs); | |
1011 | ||
1012 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
1013 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
1014 | MODULE_LICENSE("GPL"); |