Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
a805bad5 | 2 | * Copyright (c) 2000-2006 Silicon Graphics, Inc. |
7b718769 | 3 | * All Rights Reserved. |
1da177e4 | 4 | * |
7b718769 NS |
5 | * This program is free software; you can redistribute it and/or |
6 | * modify it under the terms of the GNU General Public License as | |
1da177e4 LT |
7 | * published by the Free Software Foundation. |
8 | * | |
7b718769 NS |
9 | * This program is distributed in the hope that it would be useful, |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
1da177e4 | 13 | * |
7b718769 NS |
14 | * You should have received a copy of the GNU General Public License |
15 | * along with this program; if not, write the Free Software Foundation, | |
16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
1da177e4 | 17 | */ |
1da177e4 | 18 | #include "xfs.h" |
a844f451 | 19 | #include "xfs_bit.h" |
1da177e4 LT |
20 | #include "xfs_log.h" |
21 | #include "xfs_clnt.h" | |
a844f451 | 22 | #include "xfs_inum.h" |
1da177e4 LT |
23 | #include "xfs_trans.h" |
24 | #include "xfs_sb.h" | |
a844f451 | 25 | #include "xfs_ag.h" |
1da177e4 LT |
26 | #include "xfs_dir2.h" |
27 | #include "xfs_alloc.h" | |
28 | #include "xfs_dmapi.h" | |
29 | #include "xfs_quota.h" | |
30 | #include "xfs_mount.h" | |
1da177e4 | 31 | #include "xfs_bmap_btree.h" |
a844f451 | 32 | #include "xfs_alloc_btree.h" |
1da177e4 | 33 | #include "xfs_ialloc_btree.h" |
1da177e4 | 34 | #include "xfs_dir2_sf.h" |
a844f451 | 35 | #include "xfs_attr_sf.h" |
1da177e4 LT |
36 | #include "xfs_dinode.h" |
37 | #include "xfs_inode.h" | |
a844f451 NS |
38 | #include "xfs_btree.h" |
39 | #include "xfs_ialloc.h" | |
1da177e4 | 40 | #include "xfs_bmap.h" |
1da177e4 LT |
41 | #include "xfs_rtalloc.h" |
42 | #include "xfs_error.h" | |
43 | #include "xfs_itable.h" | |
44 | #include "xfs_rw.h" | |
45 | #include "xfs_acl.h" | |
1da177e4 LT |
46 | #include "xfs_attr.h" |
47 | #include "xfs_buf_item.h" | |
48 | #include "xfs_utils.h" | |
739bfb2a | 49 | #include "xfs_vnodeops.h" |
745f6919 | 50 | #include "xfs_vfsops.h" |
1da177e4 | 51 | #include "xfs_version.h" |
1da177e4 LT |
52 | |
53 | #include <linux/namei.h> | |
54 | #include <linux/init.h> | |
55 | #include <linux/mount.h> | |
0829c360 | 56 | #include <linux/mempool.h> |
1da177e4 | 57 | #include <linux/writeback.h> |
4df08c52 | 58 | #include <linux/kthread.h> |
7dfb7103 | 59 | #include <linux/freezer.h> |
1da177e4 | 60 | |
7989cb8e DC |
61 | static struct quotactl_ops xfs_quotactl_operations; |
62 | static struct super_operations xfs_super_operations; | |
63 | static kmem_zone_t *xfs_vnode_zone; | |
64 | static kmem_zone_t *xfs_ioend_zone; | |
0829c360 | 65 | mempool_t *xfs_ioend_pool; |
1da177e4 LT |
66 | |
67 | STATIC struct xfs_mount_args * | |
68 | xfs_args_allocate( | |
764d1f89 NS |
69 | struct super_block *sb, |
70 | int silent) | |
1da177e4 LT |
71 | { |
72 | struct xfs_mount_args *args; | |
73 | ||
74 | args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP); | |
75 | args->logbufs = args->logbufsize = -1; | |
76 | strncpy(args->fsname, sb->s_id, MAXNAMELEN); | |
77 | ||
78 | /* Copy the already-parsed mount(2) flags we're interested in */ | |
1da177e4 LT |
79 | if (sb->s_flags & MS_DIRSYNC) |
80 | args->flags |= XFSMNT_DIRSYNC; | |
81 | if (sb->s_flags & MS_SYNCHRONOUS) | |
82 | args->flags |= XFSMNT_WSYNC; | |
764d1f89 NS |
83 | if (silent) |
84 | args->flags |= XFSMNT_QUIET; | |
1da177e4 LT |
85 | args->flags |= XFSMNT_32BITINODES; |
86 | ||
87 | return args; | |
88 | } | |
89 | ||
90 | __uint64_t | |
91 | xfs_max_file_offset( | |
92 | unsigned int blockshift) | |
93 | { | |
94 | unsigned int pagefactor = 1; | |
95 | unsigned int bitshift = BITS_PER_LONG - 1; | |
96 | ||
97 | /* Figure out maximum filesize, on Linux this can depend on | |
98 | * the filesystem blocksize (on 32 bit platforms). | |
99 | * __block_prepare_write does this in an [unsigned] long... | |
100 | * page->index << (PAGE_CACHE_SHIFT - bbits) | |
101 | * So, for page sized blocks (4K on 32 bit platforms), | |
102 | * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is | |
103 | * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1) | |
104 | * but for smaller blocksizes it is less (bbits = log2 bsize). | |
105 | * Note1: get_block_t takes a long (implicit cast from above) | |
106 | * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch | |
107 | * can optionally convert the [unsigned] long from above into | |
108 | * an [unsigned] long long. | |
109 | */ | |
110 | ||
111 | #if BITS_PER_LONG == 32 | |
112 | # if defined(CONFIG_LBD) | |
113 | ASSERT(sizeof(sector_t) == 8); | |
114 | pagefactor = PAGE_CACHE_SIZE; | |
115 | bitshift = BITS_PER_LONG; | |
116 | # else | |
117 | pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift); | |
118 | # endif | |
119 | #endif | |
120 | ||
121 | return (((__uint64_t)pagefactor) << bitshift) - 1; | |
122 | } | |
123 | ||
7989cb8e | 124 | STATIC_INLINE void |
1da177e4 LT |
125 | xfs_set_inodeops( |
126 | struct inode *inode) | |
127 | { | |
0432dab2 CH |
128 | switch (inode->i_mode & S_IFMT) { |
129 | case S_IFREG: | |
416c6d5b | 130 | inode->i_op = &xfs_inode_operations; |
3562fd45 | 131 | inode->i_fop = &xfs_file_operations; |
e4c573bb | 132 | inode->i_mapping->a_ops = &xfs_address_space_operations; |
0432dab2 CH |
133 | break; |
134 | case S_IFDIR: | |
416c6d5b | 135 | inode->i_op = &xfs_dir_inode_operations; |
3562fd45 | 136 | inode->i_fop = &xfs_dir_file_operations; |
0432dab2 CH |
137 | break; |
138 | case S_IFLNK: | |
416c6d5b | 139 | inode->i_op = &xfs_symlink_inode_operations; |
1da177e4 | 140 | if (inode->i_blocks) |
e4c573bb | 141 | inode->i_mapping->a_ops = &xfs_address_space_operations; |
0432dab2 CH |
142 | break; |
143 | default: | |
416c6d5b | 144 | inode->i_op = &xfs_inode_operations; |
1da177e4 | 145 | init_special_inode(inode, inode->i_mode, inode->i_rdev); |
0432dab2 | 146 | break; |
1da177e4 LT |
147 | } |
148 | } | |
149 | ||
7989cb8e | 150 | STATIC_INLINE void |
1da177e4 LT |
151 | xfs_revalidate_inode( |
152 | xfs_mount_t *mp, | |
67fcaa73 | 153 | bhv_vnode_t *vp, |
1da177e4 LT |
154 | xfs_inode_t *ip) |
155 | { | |
ec86dc02 | 156 | struct inode *inode = vn_to_inode(vp); |
1da177e4 | 157 | |
0432dab2 | 158 | inode->i_mode = ip->i_d.di_mode; |
1da177e4 LT |
159 | inode->i_nlink = ip->i_d.di_nlink; |
160 | inode->i_uid = ip->i_d.di_uid; | |
161 | inode->i_gid = ip->i_d.di_gid; | |
0432dab2 CH |
162 | |
163 | switch (inode->i_mode & S_IFMT) { | |
164 | case S_IFBLK: | |
165 | case S_IFCHR: | |
166 | inode->i_rdev = | |
167 | MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff, | |
168 | sysv_minor(ip->i_df.if_u2.if_rdev)); | |
169 | break; | |
170 | default: | |
1da177e4 | 171 | inode->i_rdev = 0; |
0432dab2 | 172 | break; |
1da177e4 | 173 | } |
0432dab2 | 174 | |
1da177e4 LT |
175 | inode->i_generation = ip->i_d.di_gen; |
176 | i_size_write(inode, ip->i_d.di_size); | |
177 | inode->i_blocks = | |
178 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
179 | inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec; | |
180 | inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec; | |
181 | inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec; | |
182 | inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec; | |
183 | inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec; | |
184 | inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec; | |
185 | if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE) | |
186 | inode->i_flags |= S_IMMUTABLE; | |
187 | else | |
188 | inode->i_flags &= ~S_IMMUTABLE; | |
189 | if (ip->i_d.di_flags & XFS_DIFLAG_APPEND) | |
190 | inode->i_flags |= S_APPEND; | |
191 | else | |
192 | inode->i_flags &= ~S_APPEND; | |
193 | if (ip->i_d.di_flags & XFS_DIFLAG_SYNC) | |
194 | inode->i_flags |= S_SYNC; | |
195 | else | |
196 | inode->i_flags &= ~S_SYNC; | |
197 | if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME) | |
198 | inode->i_flags |= S_NOATIME; | |
199 | else | |
200 | inode->i_flags &= ~S_NOATIME; | |
b3aea4ed | 201 | xfs_iflags_clear(ip, XFS_IMODIFIED); |
1da177e4 LT |
202 | } |
203 | ||
204 | void | |
205 | xfs_initialize_vnode( | |
48c872a9 | 206 | struct xfs_mount *mp, |
67fcaa73 | 207 | bhv_vnode_t *vp, |
745f6919 | 208 | struct xfs_inode *ip) |
1da177e4 | 209 | { |
ec86dc02 | 210 | struct inode *inode = vn_to_inode(vp); |
1da177e4 | 211 | |
739bfb2a | 212 | if (!ip->i_vnode) { |
739bfb2a CH |
213 | ip->i_vnode = vp; |
214 | inode->i_private = ip; | |
1da177e4 LT |
215 | } |
216 | ||
217 | /* | |
218 | * We need to set the ops vectors, and unlock the inode, but if | |
219 | * we have been called during the new inode create process, it is | |
220 | * too early to fill in the Linux inode. We will get called a | |
221 | * second time once the inode is properly set up, and then we can | |
222 | * finish our work. | |
223 | */ | |
745f6919 | 224 | if (ip->i_d.di_mode != 0 && (inode->i_state & I_NEW)) { |
48c872a9 | 225 | xfs_revalidate_inode(mp, vp, ip); |
1da177e4 | 226 | xfs_set_inodeops(inode); |
ec86dc02 | 227 | |
7a18c386 | 228 | xfs_iflags_clear(ip, XFS_INEW); |
1da177e4 LT |
229 | barrier(); |
230 | ||
231 | unlock_new_inode(inode); | |
232 | } | |
233 | } | |
234 | ||
235 | int | |
236 | xfs_blkdev_get( | |
237 | xfs_mount_t *mp, | |
238 | const char *name, | |
239 | struct block_device **bdevp) | |
240 | { | |
241 | int error = 0; | |
242 | ||
243 | *bdevp = open_bdev_excl(name, 0, mp); | |
244 | if (IS_ERR(*bdevp)) { | |
245 | error = PTR_ERR(*bdevp); | |
246 | printk("XFS: Invalid device [%s], error=%d\n", name, error); | |
247 | } | |
248 | ||
249 | return -error; | |
250 | } | |
251 | ||
252 | void | |
253 | xfs_blkdev_put( | |
254 | struct block_device *bdev) | |
255 | { | |
256 | if (bdev) | |
257 | close_bdev_excl(bdev); | |
258 | } | |
259 | ||
f538d4da CH |
260 | /* |
261 | * Try to write out the superblock using barriers. | |
262 | */ | |
263 | STATIC int | |
264 | xfs_barrier_test( | |
265 | xfs_mount_t *mp) | |
266 | { | |
267 | xfs_buf_t *sbp = xfs_getsb(mp, 0); | |
268 | int error; | |
269 | ||
270 | XFS_BUF_UNDONE(sbp); | |
271 | XFS_BUF_UNREAD(sbp); | |
272 | XFS_BUF_UNDELAYWRITE(sbp); | |
273 | XFS_BUF_WRITE(sbp); | |
274 | XFS_BUF_UNASYNC(sbp); | |
275 | XFS_BUF_ORDERED(sbp); | |
276 | ||
277 | xfsbdstrat(mp, sbp); | |
278 | error = xfs_iowait(sbp); | |
279 | ||
280 | /* | |
281 | * Clear all the flags we set and possible error state in the | |
282 | * buffer. We only did the write to try out whether barriers | |
283 | * worked and shouldn't leave any traces in the superblock | |
284 | * buffer. | |
285 | */ | |
286 | XFS_BUF_DONE(sbp); | |
287 | XFS_BUF_ERROR(sbp, 0); | |
288 | XFS_BUF_UNORDERED(sbp); | |
289 | ||
290 | xfs_buf_relse(sbp); | |
291 | return error; | |
292 | } | |
293 | ||
294 | void | |
295 | xfs_mountfs_check_barriers(xfs_mount_t *mp) | |
296 | { | |
297 | int error; | |
298 | ||
299 | if (mp->m_logdev_targp != mp->m_ddev_targp) { | |
300 | xfs_fs_cmn_err(CE_NOTE, mp, | |
301 | "Disabling barriers, not supported with external log device"); | |
302 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 303 | return; |
f538d4da CH |
304 | } |
305 | ||
ce8e922c | 306 | if (mp->m_ddev_targp->bt_bdev->bd_disk->queue->ordered == |
f538d4da CH |
307 | QUEUE_ORDERED_NONE) { |
308 | xfs_fs_cmn_err(CE_NOTE, mp, | |
309 | "Disabling barriers, not supported by the underlying device"); | |
310 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 311 | return; |
f538d4da CH |
312 | } |
313 | ||
b2ea401b NS |
314 | if (xfs_readonly_buftarg(mp->m_ddev_targp)) { |
315 | xfs_fs_cmn_err(CE_NOTE, mp, | |
316 | "Disabling barriers, underlying device is readonly"); | |
317 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
318 | return; | |
319 | } | |
320 | ||
f538d4da CH |
321 | error = xfs_barrier_test(mp); |
322 | if (error) { | |
323 | xfs_fs_cmn_err(CE_NOTE, mp, | |
324 | "Disabling barriers, trial barrier write failed"); | |
325 | mp->m_flags &= ~XFS_MOUNT_BARRIER; | |
4ef19ddd | 326 | return; |
f538d4da CH |
327 | } |
328 | } | |
329 | ||
330 | void | |
331 | xfs_blkdev_issue_flush( | |
332 | xfs_buftarg_t *buftarg) | |
333 | { | |
ce8e922c | 334 | blkdev_issue_flush(buftarg->bt_bdev, NULL); |
f538d4da | 335 | } |
1da177e4 LT |
336 | |
337 | STATIC struct inode * | |
a50cd269 | 338 | xfs_fs_alloc_inode( |
1da177e4 LT |
339 | struct super_block *sb) |
340 | { | |
67fcaa73 | 341 | bhv_vnode_t *vp; |
1da177e4 | 342 | |
8758280f NS |
343 | vp = kmem_zone_alloc(xfs_vnode_zone, KM_SLEEP); |
344 | if (unlikely(!vp)) | |
1da177e4 | 345 | return NULL; |
ec86dc02 | 346 | return vn_to_inode(vp); |
1da177e4 LT |
347 | } |
348 | ||
349 | STATIC void | |
a50cd269 | 350 | xfs_fs_destroy_inode( |
1da177e4 LT |
351 | struct inode *inode) |
352 | { | |
ec86dc02 | 353 | kmem_zone_free(xfs_vnode_zone, vn_from_inode(inode)); |
1da177e4 LT |
354 | } |
355 | ||
356 | STATIC void | |
a50cd269 | 357 | xfs_fs_inode_init_once( |
8758280f | 358 | kmem_zone_t *zonep, |
4ba9b9d0 | 359 | void *vnode) |
1da177e4 | 360 | { |
a35afb83 | 361 | inode_init_once(vn_to_inode((bhv_vnode_t *)vnode)); |
1da177e4 LT |
362 | } |
363 | ||
364 | STATIC int | |
8758280f | 365 | xfs_init_zones(void) |
1da177e4 | 366 | { |
67fcaa73 | 367 | xfs_vnode_zone = kmem_zone_init_flags(sizeof(bhv_vnode_t), "xfs_vnode", |
e0cc2325 NS |
368 | KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | |
369 | KM_ZONE_SPREAD, | |
a50cd269 | 370 | xfs_fs_inode_init_once); |
0829c360 CH |
371 | if (!xfs_vnode_zone) |
372 | goto out; | |
373 | ||
374 | xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend"); | |
375 | if (!xfs_ioend_zone) | |
376 | goto out_destroy_vnode_zone; | |
377 | ||
93d2341c MD |
378 | xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE, |
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 | |
a50cd269 | 407 | xfs_fs_write_inode( |
1da177e4 LT |
408 | struct inode *inode, |
409 | int sync) | |
410 | { | |
1da177e4 LT |
411 | int error = 0, flags = FLUSH_INODE; |
412 | ||
1543d79c | 413 | vn_trace_entry(XFS_I(inode), __FUNCTION__, |
739bfb2a CH |
414 | (inst_t *)__return_address); |
415 | if (sync) { | |
416 | filemap_fdatawait(inode->i_mapping); | |
417 | flags |= FLUSH_SYNC; | |
1da177e4 | 418 | } |
739bfb2a | 419 | error = xfs_inode_flush(XFS_I(inode), flags); |
e893bffd LM |
420 | /* |
421 | * if we failed to write out the inode then mark | |
422 | * it dirty again so we'll try again later. | |
423 | */ | |
424 | if (error) | |
425 | mark_inode_dirty_sync(inode); | |
739bfb2a | 426 | |
1da177e4 LT |
427 | return -error; |
428 | } | |
429 | ||
430 | STATIC void | |
a50cd269 | 431 | xfs_fs_clear_inode( |
1da177e4 LT |
432 | struct inode *inode) |
433 | { | |
1543d79c | 434 | xfs_inode_t *ip = XFS_I(inode); |
56d433e4 | 435 | |
02ba71de | 436 | /* |
1543d79c | 437 | * ip can be null when xfs_iget_core calls xfs_idestroy if we |
02ba71de CH |
438 | * find an inode with di_mode == 0 but without IGET_CREATE set. |
439 | */ | |
1543d79c CH |
440 | if (ip) { |
441 | vn_trace_entry(ip, __FUNCTION__, (inst_t *)__return_address); | |
442 | ||
443 | XFS_STATS_INC(vn_rele); | |
444 | XFS_STATS_INC(vn_remove); | |
445 | XFS_STATS_INC(vn_reclaim); | |
446 | XFS_STATS_DEC(vn_active); | |
447 | ||
448 | xfs_inactive(ip); | |
449 | xfs_iflags_clear(ip, XFS_IMODIFIED); | |
450 | if (xfs_reclaim(ip)) | |
451 | panic("%s: cannot reclaim 0x%p\n", __FUNCTION__, inode); | |
b3aea4ed | 452 | } |
56d433e4 | 453 | |
739bfb2a | 454 | ASSERT(XFS_I(inode) == NULL); |
56d433e4 | 455 | } |
1da177e4 LT |
456 | |
457 | /* | |
458 | * Enqueue a work item to be picked up by the vfs xfssyncd thread. | |
459 | * Doing this has two advantages: | |
460 | * - It saves on stack space, which is tight in certain situations | |
461 | * - It can be used (with care) as a mechanism to avoid deadlocks. | |
462 | * Flushing while allocating in a full filesystem requires both. | |
463 | */ | |
464 | STATIC void | |
465 | xfs_syncd_queue_work( | |
74394496 | 466 | struct xfs_mount *mp, |
1da177e4 | 467 | void *data, |
74394496 | 468 | void (*syncer)(struct xfs_mount *, void *)) |
1da177e4 | 469 | { |
b83bd138 | 470 | struct bhv_vfs_sync_work *work; |
1da177e4 | 471 | |
b83bd138 | 472 | work = kmem_alloc(sizeof(struct bhv_vfs_sync_work), KM_SLEEP); |
1da177e4 LT |
473 | INIT_LIST_HEAD(&work->w_list); |
474 | work->w_syncer = syncer; | |
475 | work->w_data = data; | |
74394496 CH |
476 | work->w_mount = mp; |
477 | spin_lock(&mp->m_sync_lock); | |
478 | list_add_tail(&work->w_list, &mp->m_sync_list); | |
479 | spin_unlock(&mp->m_sync_lock); | |
480 | wake_up_process(mp->m_sync_task); | |
1da177e4 LT |
481 | } |
482 | ||
483 | /* | |
484 | * Flush delayed allocate data, attempting to free up reserved space | |
485 | * from existing allocations. At this point a new allocation attempt | |
486 | * has failed with ENOSPC and we are in the process of scratching our | |
487 | * heads, looking about for more room... | |
488 | */ | |
489 | STATIC void | |
490 | xfs_flush_inode_work( | |
74394496 CH |
491 | struct xfs_mount *mp, |
492 | void *arg) | |
1da177e4 | 493 | { |
74394496 CH |
494 | struct inode *inode = arg; |
495 | filemap_flush(inode->i_mapping); | |
496 | iput(inode); | |
1da177e4 LT |
497 | } |
498 | ||
499 | void | |
500 | xfs_flush_inode( | |
501 | xfs_inode_t *ip) | |
502 | { | |
74394496 | 503 | struct inode *inode = ip->i_vnode; |
1da177e4 LT |
504 | |
505 | igrab(inode); | |
74394496 | 506 | xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_inode_work); |
041e0e3b | 507 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
508 | } |
509 | ||
510 | /* | |
511 | * This is the "bigger hammer" version of xfs_flush_inode_work... | |
512 | * (IOW, "If at first you don't succeed, use a Bigger Hammer"). | |
513 | */ | |
514 | STATIC void | |
515 | xfs_flush_device_work( | |
74394496 CH |
516 | struct xfs_mount *mp, |
517 | void *arg) | |
1da177e4 | 518 | { |
74394496 | 519 | struct inode *inode = arg; |
b267ce99 | 520 | sync_blockdev(mp->m_super->s_bdev); |
74394496 | 521 | iput(inode); |
1da177e4 LT |
522 | } |
523 | ||
524 | void | |
525 | xfs_flush_device( | |
526 | xfs_inode_t *ip) | |
527 | { | |
ec86dc02 | 528 | struct inode *inode = vn_to_inode(XFS_ITOV(ip)); |
1da177e4 LT |
529 | |
530 | igrab(inode); | |
74394496 | 531 | xfs_syncd_queue_work(ip->i_mount, inode, xfs_flush_device_work); |
041e0e3b | 532 | delay(msecs_to_jiffies(500)); |
1da177e4 LT |
533 | xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC); |
534 | } | |
535 | ||
1da177e4 | 536 | STATIC void |
74394496 CH |
537 | xfs_sync_worker( |
538 | struct xfs_mount *mp, | |
1da177e4 LT |
539 | void *unused) |
540 | { | |
541 | int error; | |
542 | ||
74394496 CH |
543 | if (!(mp->m_flags & XFS_MOUNT_RDONLY)) |
544 | error = xfs_sync(mp, SYNC_FSDATA | SYNC_BDFLUSH | SYNC_ATTR | | |
545 | SYNC_REFCACHE | SYNC_SUPER); | |
546 | mp->m_sync_seq++; | |
547 | wake_up(&mp->m_wait_single_sync_task); | |
1da177e4 LT |
548 | } |
549 | ||
550 | STATIC int | |
551 | xfssyncd( | |
552 | void *arg) | |
553 | { | |
74394496 | 554 | struct xfs_mount *mp = arg; |
1da177e4 | 555 | long timeleft; |
b83bd138 | 556 | bhv_vfs_sync_work_t *work, *n; |
4df08c52 | 557 | LIST_HEAD (tmp); |
1da177e4 | 558 | |
83144186 | 559 | set_freezable(); |
041e0e3b | 560 | timeleft = xfs_syncd_centisecs * msecs_to_jiffies(10); |
1da177e4 | 561 | for (;;) { |
041e0e3b | 562 | timeleft = schedule_timeout_interruptible(timeleft); |
1da177e4 | 563 | /* swsusp */ |
3e1d1d28 | 564 | try_to_freeze(); |
74394496 | 565 | if (kthread_should_stop() && list_empty(&mp->m_sync_list)) |
1da177e4 LT |
566 | break; |
567 | ||
74394496 | 568 | spin_lock(&mp->m_sync_lock); |
1da177e4 LT |
569 | /* |
570 | * We can get woken by laptop mode, to do a sync - | |
571 | * that's the (only!) case where the list would be | |
572 | * empty with time remaining. | |
573 | */ | |
74394496 | 574 | if (!timeleft || list_empty(&mp->m_sync_list)) { |
1da177e4 | 575 | if (!timeleft) |
041e0e3b NA |
576 | timeleft = xfs_syncd_centisecs * |
577 | msecs_to_jiffies(10); | |
74394496 CH |
578 | INIT_LIST_HEAD(&mp->m_sync_work.w_list); |
579 | list_add_tail(&mp->m_sync_work.w_list, | |
580 | &mp->m_sync_list); | |
1da177e4 | 581 | } |
74394496 | 582 | list_for_each_entry_safe(work, n, &mp->m_sync_list, w_list) |
1da177e4 | 583 | list_move(&work->w_list, &tmp); |
74394496 | 584 | spin_unlock(&mp->m_sync_lock); |
1da177e4 LT |
585 | |
586 | list_for_each_entry_safe(work, n, &tmp, w_list) { | |
74394496 | 587 | (*work->w_syncer)(mp, work->w_data); |
1da177e4 | 588 | list_del(&work->w_list); |
74394496 | 589 | if (work == &mp->m_sync_work) |
1da177e4 | 590 | continue; |
b83bd138 | 591 | kmem_free(work, sizeof(struct bhv_vfs_sync_work)); |
1da177e4 LT |
592 | } |
593 | } | |
594 | ||
1da177e4 LT |
595 | return 0; |
596 | } | |
597 | ||
1da177e4 | 598 | STATIC void |
a50cd269 | 599 | xfs_fs_put_super( |
1da177e4 LT |
600 | struct super_block *sb) |
601 | { | |
745f6919 | 602 | struct xfs_mount *mp = XFS_M(sb); |
1da177e4 LT |
603 | int error; |
604 | ||
74394496 CH |
605 | kthread_stop(mp->m_sync_task); |
606 | ||
745f6919 CH |
607 | xfs_sync(mp, SYNC_ATTR | SYNC_DELWRI); |
608 | error = xfs_unmount(mp, 0, NULL); | |
b267ce99 | 609 | if (error) |
b83bd138 | 610 | printk("XFS: unmount got error=%d\n", error); |
1da177e4 LT |
611 | } |
612 | ||
613 | STATIC void | |
a50cd269 | 614 | xfs_fs_write_super( |
1da177e4 LT |
615 | struct super_block *sb) |
616 | { | |
b83bd138 | 617 | if (!(sb->s_flags & MS_RDONLY)) |
745f6919 | 618 | xfs_sync(XFS_M(sb), SYNC_FSDATA); |
1da177e4 LT |
619 | sb->s_dirt = 0; |
620 | } | |
621 | ||
622 | STATIC int | |
a50cd269 | 623 | xfs_fs_sync_super( |
1da177e4 LT |
624 | struct super_block *sb, |
625 | int wait) | |
626 | { | |
745f6919 | 627 | struct xfs_mount *mp = XFS_M(sb); |
b83bd138 NS |
628 | int error; |
629 | int flags; | |
1da177e4 | 630 | |
e893bffd LM |
631 | /* |
632 | * Treat a sync operation like a freeze. This is to work | |
633 | * around a race in sync_inodes() which works in two phases | |
634 | * - an asynchronous flush, which can write out an inode | |
635 | * without waiting for file size updates to complete, and a | |
636 | * synchronous flush, which wont do anything because the | |
637 | * async flush removed the inode's dirty flag. Also | |
638 | * sync_inodes() will not see any files that just have | |
639 | * outstanding transactions to be flushed because we don't | |
640 | * dirty the Linux inode until after the transaction I/O | |
641 | * completes. | |
642 | */ | |
643 | if (wait || unlikely(sb->s_frozen == SB_FREEZE_WRITE)) { | |
2823945f DC |
644 | /* |
645 | * First stage of freeze - no more writers will make progress | |
646 | * now we are here, so we flush delwri and delalloc buffers | |
647 | * here, then wait for all I/O to complete. Data is frozen at | |
648 | * that point. Metadata is not frozen, transactions can still | |
649 | * occur here so don't bother flushing the buftarg (i.e | |
650 | * SYNC_QUIESCE) because it'll just get dirty again. | |
651 | */ | |
516b2e7c | 652 | flags = SYNC_DATA_QUIESCE; |
2823945f | 653 | } else |
e893bffd | 654 | flags = SYNC_FSDATA; |
1da177e4 | 655 | |
745f6919 | 656 | error = xfs_sync(mp, flags); |
1da177e4 LT |
657 | sb->s_dirt = 0; |
658 | ||
659 | if (unlikely(laptop_mode)) { | |
74394496 | 660 | int prev_sync_seq = mp->m_sync_seq; |
1da177e4 LT |
661 | |
662 | /* | |
663 | * The disk must be active because we're syncing. | |
664 | * We schedule xfssyncd now (now that the disk is | |
665 | * active) instead of later (when it might not be). | |
666 | */ | |
74394496 | 667 | wake_up_process(mp->m_sync_task); |
1da177e4 LT |
668 | /* |
669 | * We have to wait for the sync iteration to complete. | |
670 | * If we don't, the disk activity caused by the sync | |
671 | * will come after the sync is completed, and that | |
672 | * triggers another sync from laptop mode. | |
673 | */ | |
74394496 CH |
674 | wait_event(mp->m_wait_single_sync_task, |
675 | mp->m_sync_seq != prev_sync_seq); | |
1da177e4 LT |
676 | } |
677 | ||
678 | return -error; | |
679 | } | |
680 | ||
681 | STATIC int | |
a50cd269 | 682 | xfs_fs_statfs( |
726c3342 | 683 | struct dentry *dentry, |
1da177e4 LT |
684 | struct kstatfs *statp) |
685 | { | |
745f6919 | 686 | return -xfs_statvfs(XFS_M(dentry->d_sb), statp, |
d6938d1b | 687 | vn_from_inode(dentry->d_inode)); |
1da177e4 LT |
688 | } |
689 | ||
690 | STATIC int | |
a50cd269 | 691 | xfs_fs_remount( |
1da177e4 LT |
692 | struct super_block *sb, |
693 | int *flags, | |
694 | char *options) | |
695 | { | |
745f6919 | 696 | struct xfs_mount *mp = XFS_M(sb); |
764d1f89 | 697 | struct xfs_mount_args *args = xfs_args_allocate(sb, 0); |
1da177e4 LT |
698 | int error; |
699 | ||
745f6919 | 700 | error = xfs_parseargs(mp, options, args, 1); |
1da177e4 | 701 | if (!error) |
745f6919 | 702 | error = xfs_mntupdate(mp, flags, args); |
1da177e4 LT |
703 | kmem_free(args, sizeof(*args)); |
704 | return -error; | |
705 | } | |
706 | ||
707 | STATIC void | |
a50cd269 | 708 | xfs_fs_lockfs( |
1da177e4 LT |
709 | struct super_block *sb) |
710 | { | |
745f6919 | 711 | xfs_freeze(XFS_M(sb)); |
1da177e4 LT |
712 | } |
713 | ||
714 | STATIC int | |
a50cd269 | 715 | xfs_fs_show_options( |
1da177e4 LT |
716 | struct seq_file *m, |
717 | struct vfsmount *mnt) | |
718 | { | |
745f6919 | 719 | return -xfs_showargs(XFS_M(mnt->mnt_sb), m); |
1da177e4 LT |
720 | } |
721 | ||
ee34807a | 722 | STATIC int |
a50cd269 | 723 | xfs_fs_quotasync( |
ee34807a NS |
724 | struct super_block *sb, |
725 | int type) | |
726 | { | |
b09cc771 | 727 | return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XQUOTASYNC, 0, NULL); |
ee34807a NS |
728 | } |
729 | ||
1da177e4 | 730 | STATIC int |
a50cd269 | 731 | xfs_fs_getxstate( |
1da177e4 LT |
732 | struct super_block *sb, |
733 | struct fs_quota_stat *fqs) | |
734 | { | |
b09cc771 | 735 | return -XFS_QM_QUOTACTL(XFS_M(sb), Q_XGETQSTAT, 0, (caddr_t)fqs); |
1da177e4 LT |
736 | } |
737 | ||
738 | STATIC int | |
a50cd269 | 739 | xfs_fs_setxstate( |
1da177e4 LT |
740 | struct super_block *sb, |
741 | unsigned int flags, | |
742 | int op) | |
743 | { | |
b09cc771 | 744 | return -XFS_QM_QUOTACTL(XFS_M(sb), op, 0, (caddr_t)&flags); |
1da177e4 LT |
745 | } |
746 | ||
747 | STATIC int | |
a50cd269 | 748 | xfs_fs_getxquota( |
1da177e4 LT |
749 | struct super_block *sb, |
750 | int type, | |
751 | qid_t id, | |
752 | struct fs_disk_quota *fdq) | |
753 | { | |
b09cc771 | 754 | return -XFS_QM_QUOTACTL(XFS_M(sb), |
b83bd138 NS |
755 | (type == USRQUOTA) ? Q_XGETQUOTA : |
756 | ((type == GRPQUOTA) ? Q_XGETGQUOTA : | |
757 | Q_XGETPQUOTA), id, (caddr_t)fdq); | |
1da177e4 LT |
758 | } |
759 | ||
760 | STATIC int | |
a50cd269 | 761 | xfs_fs_setxquota( |
1da177e4 LT |
762 | struct super_block *sb, |
763 | int type, | |
764 | qid_t id, | |
765 | struct fs_disk_quota *fdq) | |
766 | { | |
b09cc771 | 767 | return -XFS_QM_QUOTACTL(XFS_M(sb), |
b83bd138 NS |
768 | (type == USRQUOTA) ? Q_XSETQLIM : |
769 | ((type == GRPQUOTA) ? Q_XSETGQLIM : | |
770 | Q_XSETPQLIM), id, (caddr_t)fdq); | |
1da177e4 LT |
771 | } |
772 | ||
773 | STATIC int | |
a50cd269 | 774 | xfs_fs_fill_super( |
1da177e4 LT |
775 | struct super_block *sb, |
776 | void *data, | |
777 | int silent) | |
778 | { | |
0a74cd19 | 779 | struct inode *rootvp; |
745f6919 | 780 | struct xfs_mount *mp = NULL; |
764d1f89 | 781 | struct xfs_mount_args *args = xfs_args_allocate(sb, silent); |
1da177e4 | 782 | struct kstatfs statvfs; |
b83bd138 | 783 | int error; |
1da177e4 | 784 | |
745f6919 | 785 | mp = xfs_mount_init(); |
1da177e4 | 786 | |
74394496 CH |
787 | INIT_LIST_HEAD(&mp->m_sync_list); |
788 | spin_lock_init(&mp->m_sync_lock); | |
789 | init_waitqueue_head(&mp->m_wait_single_sync_task); | |
790 | ||
b267ce99 CH |
791 | mp->m_super = sb; |
792 | sb->s_fs_info = mp; | |
1da177e4 | 793 | |
bd186aa9 CH |
794 | if (sb->s_flags & MS_RDONLY) |
795 | mp->m_flags |= XFS_MOUNT_RDONLY; | |
796 | ||
745f6919 CH |
797 | error = xfs_parseargs(mp, (char *)data, args, 0); |
798 | if (error) | |
1da177e4 | 799 | goto fail_vfsop; |
1da177e4 LT |
800 | |
801 | sb_min_blocksize(sb, BBSIZE); | |
a50cd269 | 802 | sb->s_export_op = &xfs_export_operations; |
a50cd269 NS |
803 | sb->s_qcop = &xfs_quotactl_operations; |
804 | sb->s_op = &xfs_super_operations; | |
1da177e4 | 805 | |
745f6919 CH |
806 | error = xfs_mount(mp, args, NULL); |
807 | if (error) | |
1da177e4 | 808 | goto fail_vfsop; |
1da177e4 | 809 | |
745f6919 | 810 | error = xfs_statvfs(mp, &statvfs, NULL); |
1da177e4 LT |
811 | if (error) |
812 | goto fail_unmount; | |
813 | ||
814 | sb->s_dirt = 1; | |
815 | sb->s_magic = statvfs.f_type; | |
816 | sb->s_blocksize = statvfs.f_bsize; | |
817 | sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1; | |
818 | sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); | |
819 | sb->s_time_gran = 1; | |
820 | set_posix_acl_flag(sb); | |
821 | ||
745f6919 | 822 | error = xfs_root(mp, &rootvp); |
1da177e4 LT |
823 | if (error) |
824 | goto fail_unmount; | |
825 | ||
ec86dc02 | 826 | sb->s_root = d_alloc_root(vn_to_inode(rootvp)); |
1da177e4 LT |
827 | if (!sb->s_root) { |
828 | error = ENOMEM; | |
829 | goto fail_vnrele; | |
830 | } | |
831 | if (is_bad_inode(sb->s_root->d_inode)) { | |
832 | error = EINVAL; | |
833 | goto fail_vnrele; | |
834 | } | |
74394496 CH |
835 | |
836 | mp->m_sync_work.w_syncer = xfs_sync_worker; | |
837 | mp->m_sync_work.w_mount = mp; | |
838 | mp->m_sync_task = kthread_run(xfssyncd, mp, "xfssyncd"); | |
839 | if (IS_ERR(mp->m_sync_task)) { | |
840 | error = -PTR_ERR(mp->m_sync_task); | |
1da177e4 | 841 | goto fail_vnrele; |
74394496 CH |
842 | } |
843 | ||
1543d79c CH |
844 | vn_trace_exit(XFS_I(sb->s_root->d_inode), __FUNCTION__, |
845 | (inst_t *)__return_address); | |
1da177e4 LT |
846 | |
847 | kmem_free(args, sizeof(*args)); | |
848 | return 0; | |
849 | ||
850 | fail_vnrele: | |
851 | if (sb->s_root) { | |
852 | dput(sb->s_root); | |
853 | sb->s_root = NULL; | |
854 | } else { | |
855 | VN_RELE(rootvp); | |
856 | } | |
857 | ||
858 | fail_unmount: | |
745f6919 | 859 | xfs_unmount(mp, 0, NULL); |
1da177e4 LT |
860 | |
861 | fail_vfsop: | |
1da177e4 LT |
862 | kmem_free(args, sizeof(*args)); |
863 | return -error; | |
864 | } | |
865 | ||
454e2398 | 866 | STATIC int |
a50cd269 | 867 | xfs_fs_get_sb( |
1da177e4 LT |
868 | struct file_system_type *fs_type, |
869 | int flags, | |
870 | const char *dev_name, | |
454e2398 DH |
871 | void *data, |
872 | struct vfsmount *mnt) | |
1da177e4 | 873 | { |
454e2398 DH |
874 | return get_sb_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super, |
875 | mnt); | |
a50cd269 NS |
876 | } |
877 | ||
7989cb8e | 878 | static struct super_operations xfs_super_operations = { |
a50cd269 NS |
879 | .alloc_inode = xfs_fs_alloc_inode, |
880 | .destroy_inode = xfs_fs_destroy_inode, | |
881 | .write_inode = xfs_fs_write_inode, | |
882 | .clear_inode = xfs_fs_clear_inode, | |
883 | .put_super = xfs_fs_put_super, | |
884 | .write_super = xfs_fs_write_super, | |
885 | .sync_fs = xfs_fs_sync_super, | |
886 | .write_super_lockfs = xfs_fs_lockfs, | |
887 | .statfs = xfs_fs_statfs, | |
888 | .remount_fs = xfs_fs_remount, | |
889 | .show_options = xfs_fs_show_options, | |
1da177e4 LT |
890 | }; |
891 | ||
7989cb8e | 892 | static struct quotactl_ops xfs_quotactl_operations = { |
a50cd269 NS |
893 | .quota_sync = xfs_fs_quotasync, |
894 | .get_xstate = xfs_fs_getxstate, | |
895 | .set_xstate = xfs_fs_setxstate, | |
896 | .get_xquota = xfs_fs_getxquota, | |
897 | .set_xquota = xfs_fs_setxquota, | |
1da177e4 LT |
898 | }; |
899 | ||
5085b607 | 900 | static struct file_system_type xfs_fs_type = { |
1da177e4 LT |
901 | .owner = THIS_MODULE, |
902 | .name = "xfs", | |
a50cd269 | 903 | .get_sb = xfs_fs_get_sb, |
1da177e4 LT |
904 | .kill_sb = kill_block_super, |
905 | .fs_flags = FS_REQUIRES_DEV, | |
906 | }; | |
907 | ||
908 | ||
909 | STATIC int __init | |
910 | init_xfs_fs( void ) | |
911 | { | |
912 | int error; | |
1da177e4 LT |
913 | static char message[] __initdata = KERN_INFO \ |
914 | XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n"; | |
915 | ||
916 | printk(message); | |
917 | ||
1da177e4 LT |
918 | ktrace_init(64); |
919 | ||
8758280f | 920 | error = xfs_init_zones(); |
1da177e4 | 921 | if (error < 0) |
0829c360 | 922 | goto undo_zones; |
1da177e4 | 923 | |
ce8e922c | 924 | error = xfs_buf_init(); |
1da177e4 | 925 | if (error < 0) |
ce8e922c | 926 | goto undo_buffers; |
1da177e4 LT |
927 | |
928 | vn_init(); | |
929 | xfs_init(); | |
930 | uuid_init(); | |
931 | vfs_initquota(); | |
932 | ||
933 | error = register_filesystem(&xfs_fs_type); | |
934 | if (error) | |
935 | goto undo_register; | |
1da177e4 LT |
936 | return 0; |
937 | ||
938 | undo_register: | |
ce8e922c | 939 | xfs_buf_terminate(); |
1da177e4 | 940 | |
ce8e922c | 941 | undo_buffers: |
8758280f | 942 | xfs_destroy_zones(); |
1da177e4 | 943 | |
0829c360 | 944 | undo_zones: |
1da177e4 LT |
945 | return error; |
946 | } | |
947 | ||
948 | STATIC void __exit | |
949 | exit_xfs_fs( void ) | |
950 | { | |
951 | vfs_exitquota(); | |
1da177e4 LT |
952 | unregister_filesystem(&xfs_fs_type); |
953 | xfs_cleanup(); | |
ce8e922c | 954 | xfs_buf_terminate(); |
8758280f | 955 | xfs_destroy_zones(); |
1da177e4 LT |
956 | ktrace_uninit(); |
957 | } | |
958 | ||
959 | module_init(init_xfs_fs); | |
960 | module_exit(exit_xfs_fs); | |
961 | ||
962 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
963 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
964 | MODULE_LICENSE("GPL"); |