]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
d3870398 | 2 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. All Rights Reserved. |
1da177e4 LT |
3 | * |
4 | * This program is free software; you can redistribute it and/or modify it | |
5 | * under the terms of version 2 of the GNU General Public License as | |
6 | * published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it would be useful, but | |
9 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. | |
11 | * | |
12 | * Further, this software is distributed without any warranty that it is | |
13 | * free of the rightful claim of any third person regarding infringement | |
14 | * or the like. Any license provided herein, whether implied or | |
15 | * otherwise, applies only to this software file. Patent licenses, if | |
16 | * any, provided herein do not apply to combinations of this program with | |
17 | * other software, or any other product whatsoever. | |
18 | * | |
19 | * You should have received a copy of the GNU General Public License along | |
20 | * with this program; if not, write the Free Software Foundation, Inc., 59 | |
21 | * Temple Place - Suite 330, Boston MA 02111-1307, USA. | |
22 | * | |
23 | * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy, | |
24 | * Mountain View, CA 94043, or: | |
25 | * | |
26 | * http://www.sgi.com | |
27 | * | |
28 | * For further information regarding this notice, see: | |
29 | * | |
30 | * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/ | |
31 | */ | |
32 | ||
33 | #include "xfs.h" | |
34 | ||
35 | #include "xfs_inum.h" | |
36 | #include "xfs_log.h" | |
37 | #include "xfs_clnt.h" | |
38 | #include "xfs_trans.h" | |
39 | #include "xfs_sb.h" | |
40 | #include "xfs_dir.h" | |
41 | #include "xfs_dir2.h" | |
42 | #include "xfs_alloc.h" | |
43 | #include "xfs_dmapi.h" | |
44 | #include "xfs_quota.h" | |
45 | #include "xfs_mount.h" | |
46 | #include "xfs_alloc_btree.h" | |
47 | #include "xfs_bmap_btree.h" | |
48 | #include "xfs_ialloc_btree.h" | |
49 | #include "xfs_btree.h" | |
50 | #include "xfs_ialloc.h" | |
51 | #include "xfs_attr_sf.h" | |
52 | #include "xfs_dir_sf.h" | |
53 | #include "xfs_dir2_sf.h" | |
54 | #include "xfs_dinode.h" | |
55 | #include "xfs_inode.h" | |
56 | #include "xfs_bmap.h" | |
57 | #include "xfs_bit.h" | |
58 | #include "xfs_rtalloc.h" | |
59 | #include "xfs_error.h" | |
60 | #include "xfs_itable.h" | |
61 | #include "xfs_rw.h" | |
62 | #include "xfs_acl.h" | |
63 | #include "xfs_cap.h" | |
64 | #include "xfs_mac.h" | |
65 | #include "xfs_attr.h" | |
66 | #include "xfs_buf_item.h" | |
67 | #include "xfs_utils.h" | |
68 | #include "xfs_version.h" | |
1da177e4 LT |
69 | |
70 | #include <linux/namei.h> | |
71 | #include <linux/init.h> | |
72 | #include <linux/mount.h> | |
73 | #include <linux/writeback.h> | |
74 | ||
75 | STATIC struct quotactl_ops linvfs_qops; | |
76 | STATIC struct super_operations linvfs_sops; | |
77 | STATIC kmem_zone_t *linvfs_inode_zone; | |
78 | ||
79 | STATIC struct xfs_mount_args * | |
80 | xfs_args_allocate( | |
81 | struct super_block *sb) | |
82 | { | |
83 | struct xfs_mount_args *args; | |
84 | ||
85 | args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP); | |
86 | args->logbufs = args->logbufsize = -1; | |
87 | strncpy(args->fsname, sb->s_id, MAXNAMELEN); | |
88 | ||
89 | /* Copy the already-parsed mount(2) flags we're interested in */ | |
90 | if (sb->s_flags & MS_NOATIME) | |
91 | args->flags |= XFSMNT_NOATIME; | |
92 | if (sb->s_flags & MS_DIRSYNC) | |
93 | args->flags |= XFSMNT_DIRSYNC; | |
94 | if (sb->s_flags & MS_SYNCHRONOUS) | |
95 | args->flags |= XFSMNT_WSYNC; | |
96 | ||
97 | /* Default to 32 bit inodes on Linux all the time */ | |
98 | args->flags |= XFSMNT_32BITINODES; | |
99 | ||
100 | return args; | |
101 | } | |
102 | ||
103 | __uint64_t | |
104 | xfs_max_file_offset( | |
105 | unsigned int blockshift) | |
106 | { | |
107 | unsigned int pagefactor = 1; | |
108 | unsigned int bitshift = BITS_PER_LONG - 1; | |
109 | ||
110 | /* Figure out maximum filesize, on Linux this can depend on | |
111 | * the filesystem blocksize (on 32 bit platforms). | |
112 | * __block_prepare_write does this in an [unsigned] long... | |
113 | * page->index << (PAGE_CACHE_SHIFT - bbits) | |
114 | * So, for page sized blocks (4K on 32 bit platforms), | |
115 | * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is | |
116 | * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1) | |
117 | * but for smaller blocksizes it is less (bbits = log2 bsize). | |
118 | * Note1: get_block_t takes a long (implicit cast from above) | |
119 | * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch | |
120 | * can optionally convert the [unsigned] long from above into | |
121 | * an [unsigned] long long. | |
122 | */ | |
123 | ||
124 | #if BITS_PER_LONG == 32 | |
125 | # if defined(CONFIG_LBD) | |
126 | ASSERT(sizeof(sector_t) == 8); | |
127 | pagefactor = PAGE_CACHE_SIZE; | |
128 | bitshift = BITS_PER_LONG; | |
129 | # else | |
130 | pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift); | |
131 | # endif | |
132 | #endif | |
133 | ||
134 | return (((__uint64_t)pagefactor) << bitshift) - 1; | |
135 | } | |
136 | ||
137 | STATIC __inline__ void | |
138 | xfs_set_inodeops( | |
139 | struct inode *inode) | |
140 | { | |
0432dab2 CH |
141 | switch (inode->i_mode & S_IFMT) { |
142 | case S_IFREG: | |
1da177e4 LT |
143 | inode->i_op = &linvfs_file_inode_operations; |
144 | inode->i_fop = &linvfs_file_operations; | |
145 | inode->i_mapping->a_ops = &linvfs_aops; | |
0432dab2 CH |
146 | break; |
147 | case S_IFDIR: | |
1da177e4 LT |
148 | inode->i_op = &linvfs_dir_inode_operations; |
149 | inode->i_fop = &linvfs_dir_operations; | |
0432dab2 CH |
150 | break; |
151 | case S_IFLNK: | |
1da177e4 LT |
152 | inode->i_op = &linvfs_symlink_inode_operations; |
153 | if (inode->i_blocks) | |
154 | inode->i_mapping->a_ops = &linvfs_aops; | |
0432dab2 CH |
155 | break; |
156 | default: | |
1da177e4 LT |
157 | inode->i_op = &linvfs_file_inode_operations; |
158 | init_special_inode(inode, inode->i_mode, inode->i_rdev); | |
0432dab2 | 159 | break; |
1da177e4 LT |
160 | } |
161 | } | |
162 | ||
163 | STATIC __inline__ void | |
164 | xfs_revalidate_inode( | |
165 | xfs_mount_t *mp, | |
166 | vnode_t *vp, | |
167 | xfs_inode_t *ip) | |
168 | { | |
169 | struct inode *inode = LINVFS_GET_IP(vp); | |
170 | ||
0432dab2 | 171 | inode->i_mode = ip->i_d.di_mode; |
1da177e4 LT |
172 | inode->i_nlink = ip->i_d.di_nlink; |
173 | inode->i_uid = ip->i_d.di_uid; | |
174 | inode->i_gid = ip->i_d.di_gid; | |
0432dab2 CH |
175 | |
176 | switch (inode->i_mode & S_IFMT) { | |
177 | case S_IFBLK: | |
178 | case S_IFCHR: | |
179 | inode->i_rdev = | |
180 | MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff, | |
181 | sysv_minor(ip->i_df.if_u2.if_rdev)); | |
182 | break; | |
183 | default: | |
1da177e4 | 184 | inode->i_rdev = 0; |
0432dab2 | 185 | break; |
1da177e4 | 186 | } |
0432dab2 | 187 | |
1da177e4 LT |
188 | inode->i_blksize = PAGE_CACHE_SIZE; |
189 | inode->i_generation = ip->i_d.di_gen; | |
190 | i_size_write(inode, ip->i_d.di_size); | |
191 | inode->i_blocks = | |
192 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
193 | inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec; | |
194 | inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec; | |
195 | inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec; | |
196 | inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec; | |
197 | inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec; | |
198 | inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec; | |
199 | if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE) | |
200 | inode->i_flags |= S_IMMUTABLE; | |
201 | else | |
202 | inode->i_flags &= ~S_IMMUTABLE; | |
203 | if (ip->i_d.di_flags & XFS_DIFLAG_APPEND) | |
204 | inode->i_flags |= S_APPEND; | |
205 | else | |
206 | inode->i_flags &= ~S_APPEND; | |
207 | if (ip->i_d.di_flags & XFS_DIFLAG_SYNC) | |
208 | inode->i_flags |= S_SYNC; | |
209 | else | |
210 | inode->i_flags &= ~S_SYNC; | |
211 | if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME) | |
212 | inode->i_flags |= S_NOATIME; | |
213 | else | |
214 | inode->i_flags &= ~S_NOATIME; | |
215 | vp->v_flag &= ~VMODIFIED; | |
216 | } | |
217 | ||
218 | void | |
219 | xfs_initialize_vnode( | |
220 | bhv_desc_t *bdp, | |
221 | vnode_t *vp, | |
222 | bhv_desc_t *inode_bhv, | |
223 | int unlock) | |
224 | { | |
225 | xfs_inode_t *ip = XFS_BHVTOI(inode_bhv); | |
226 | struct inode *inode = LINVFS_GET_IP(vp); | |
227 | ||
228 | if (!inode_bhv->bd_vobj) { | |
229 | vp->v_vfsp = bhvtovfs(bdp); | |
230 | bhv_desc_init(inode_bhv, ip, vp, &xfs_vnodeops); | |
231 | bhv_insert(VN_BHV_HEAD(vp), inode_bhv); | |
232 | } | |
233 | ||
234 | /* | |
235 | * We need to set the ops vectors, and unlock the inode, but if | |
236 | * we have been called during the new inode create process, it is | |
237 | * too early to fill in the Linux inode. We will get called a | |
238 | * second time once the inode is properly set up, and then we can | |
239 | * finish our work. | |
240 | */ | |
241 | if (ip->i_d.di_mode != 0 && unlock && (inode->i_state & I_NEW)) { | |
1da177e4 LT |
242 | xfs_revalidate_inode(XFS_BHVTOM(bdp), vp, ip); |
243 | xfs_set_inodeops(inode); | |
244 | ||
245 | ip->i_flags &= ~XFS_INEW; | |
246 | barrier(); | |
247 | ||
248 | unlock_new_inode(inode); | |
249 | } | |
250 | } | |
251 | ||
252 | int | |
253 | xfs_blkdev_get( | |
254 | xfs_mount_t *mp, | |
255 | const char *name, | |
256 | struct block_device **bdevp) | |
257 | { | |
258 | int error = 0; | |
259 | ||
260 | *bdevp = open_bdev_excl(name, 0, mp); | |
261 | if (IS_ERR(*bdevp)) { | |
262 | error = PTR_ERR(*bdevp); | |
263 | printk("XFS: Invalid device [%s], error=%d\n", name, error); | |
264 | } | |
265 | ||
266 | return -error; | |
267 | } | |
268 | ||
269 | void | |
270 | xfs_blkdev_put( | |
271 | struct block_device *bdev) | |
272 | { | |
273 | if (bdev) | |
274 | close_bdev_excl(bdev); | |
275 | } | |
276 | ||
277 | ||
278 | STATIC struct inode * | |
279 | linvfs_alloc_inode( | |
280 | struct super_block *sb) | |
281 | { | |
282 | vnode_t *vp; | |
283 | ||
284 | vp = (vnode_t *)kmem_cache_alloc(linvfs_inode_zone, | |
285 | kmem_flags_convert(KM_SLEEP)); | |
286 | if (!vp) | |
287 | return NULL; | |
288 | return LINVFS_GET_IP(vp); | |
289 | } | |
290 | ||
291 | STATIC void | |
292 | linvfs_destroy_inode( | |
293 | struct inode *inode) | |
294 | { | |
295 | kmem_cache_free(linvfs_inode_zone, LINVFS_GET_VP(inode)); | |
296 | } | |
297 | ||
298 | STATIC void | |
299 | init_once( | |
300 | void *data, | |
301 | kmem_cache_t *cachep, | |
302 | unsigned long flags) | |
303 | { | |
304 | vnode_t *vp = (vnode_t *)data; | |
305 | ||
306 | if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == | |
307 | SLAB_CTOR_CONSTRUCTOR) | |
308 | inode_init_once(LINVFS_GET_IP(vp)); | |
309 | } | |
310 | ||
311 | STATIC int | |
312 | init_inodecache( void ) | |
313 | { | |
314 | linvfs_inode_zone = kmem_cache_create("linvfs_icache", | |
315 | sizeof(vnode_t), 0, SLAB_RECLAIM_ACCOUNT, | |
316 | init_once, NULL); | |
317 | if (linvfs_inode_zone == NULL) | |
318 | return -ENOMEM; | |
319 | return 0; | |
320 | } | |
321 | ||
322 | STATIC void | |
323 | destroy_inodecache( void ) | |
324 | { | |
325 | if (kmem_cache_destroy(linvfs_inode_zone)) | |
326 | printk(KERN_WARNING "%s: cache still in use!\n", __FUNCTION__); | |
327 | } | |
328 | ||
329 | /* | |
330 | * Attempt to flush the inode, this will actually fail | |
331 | * if the inode is pinned, but we dirty the inode again | |
332 | * at the point when it is unpinned after a log write, | |
333 | * since this is when the inode itself becomes flushable. | |
334 | */ | |
335 | STATIC int | |
336 | linvfs_write_inode( | |
337 | struct inode *inode, | |
338 | int sync) | |
339 | { | |
340 | vnode_t *vp = LINVFS_GET_VP(inode); | |
341 | int error = 0, flags = FLUSH_INODE; | |
342 | ||
343 | if (vp) { | |
344 | vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); | |
345 | if (sync) | |
346 | flags |= FLUSH_SYNC; | |
347 | VOP_IFLUSH(vp, flags, error); | |
348 | if (error == EAGAIN) { | |
349 | if (sync) | |
350 | VOP_IFLUSH(vp, flags | FLUSH_LOG, error); | |
351 | else | |
352 | error = 0; | |
353 | } | |
354 | } | |
355 | ||
356 | return -error; | |
357 | } | |
358 | ||
359 | STATIC void | |
360 | linvfs_clear_inode( | |
361 | struct inode *inode) | |
362 | { | |
363 | vnode_t *vp = LINVFS_GET_VP(inode); | |
364 | ||
365 | if (vp) { | |
366 | vn_rele(vp); | |
367 | vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); | |
368 | /* | |
369 | * Do all our cleanup, and remove this vnode. | |
370 | */ | |
371 | vn_remove(vp); | |
372 | } | |
373 | } | |
374 | ||
375 | ||
376 | /* | |
377 | * Enqueue a work item to be picked up by the vfs xfssyncd thread. | |
378 | * Doing this has two advantages: | |
379 | * - It saves on stack space, which is tight in certain situations | |
380 | * - It can be used (with care) as a mechanism to avoid deadlocks. | |
381 | * Flushing while allocating in a full filesystem requires both. | |
382 | */ | |
383 | STATIC void | |
384 | xfs_syncd_queue_work( | |
385 | struct vfs *vfs, | |
386 | void *data, | |
387 | void (*syncer)(vfs_t *, void *)) | |
388 | { | |
389 | vfs_sync_work_t *work; | |
390 | ||
391 | work = kmem_alloc(sizeof(struct vfs_sync_work), KM_SLEEP); | |
392 | INIT_LIST_HEAD(&work->w_list); | |
393 | work->w_syncer = syncer; | |
394 | work->w_data = data; | |
395 | work->w_vfs = vfs; | |
396 | spin_lock(&vfs->vfs_sync_lock); | |
397 | list_add_tail(&work->w_list, &vfs->vfs_sync_list); | |
398 | spin_unlock(&vfs->vfs_sync_lock); | |
399 | wake_up_process(vfs->vfs_sync_task); | |
400 | } | |
401 | ||
402 | /* | |
403 | * Flush delayed allocate data, attempting to free up reserved space | |
404 | * from existing allocations. At this point a new allocation attempt | |
405 | * has failed with ENOSPC and we are in the process of scratching our | |
406 | * heads, looking about for more room... | |
407 | */ | |
408 | STATIC void | |
409 | xfs_flush_inode_work( | |
410 | vfs_t *vfs, | |
411 | void *inode) | |
412 | { | |
413 | filemap_flush(((struct inode *)inode)->i_mapping); | |
414 | iput((struct inode *)inode); | |
415 | } | |
416 | ||
417 | void | |
418 | xfs_flush_inode( | |
419 | xfs_inode_t *ip) | |
420 | { | |
421 | struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); | |
422 | struct vfs *vfs = XFS_MTOVFS(ip->i_mount); | |
423 | ||
424 | igrab(inode); | |
425 | xfs_syncd_queue_work(vfs, inode, xfs_flush_inode_work); | |
426 | delay(HZ/2); | |
427 | } | |
428 | ||
429 | /* | |
430 | * This is the "bigger hammer" version of xfs_flush_inode_work... | |
431 | * (IOW, "If at first you don't succeed, use a Bigger Hammer"). | |
432 | */ | |
433 | STATIC void | |
434 | xfs_flush_device_work( | |
435 | vfs_t *vfs, | |
436 | void *inode) | |
437 | { | |
438 | sync_blockdev(vfs->vfs_super->s_bdev); | |
439 | iput((struct inode *)inode); | |
440 | } | |
441 | ||
442 | void | |
443 | xfs_flush_device( | |
444 | xfs_inode_t *ip) | |
445 | { | |
446 | struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); | |
447 | struct vfs *vfs = XFS_MTOVFS(ip->i_mount); | |
448 | ||
449 | igrab(inode); | |
450 | xfs_syncd_queue_work(vfs, inode, xfs_flush_device_work); | |
451 | delay(HZ/2); | |
452 | xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC); | |
453 | } | |
454 | ||
455 | #define SYNCD_FLAGS (SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR) | |
456 | STATIC void | |
457 | vfs_sync_worker( | |
458 | vfs_t *vfsp, | |
459 | void *unused) | |
460 | { | |
461 | int error; | |
462 | ||
463 | if (!(vfsp->vfs_flag & VFS_RDONLY)) | |
464 | VFS_SYNC(vfsp, SYNCD_FLAGS, NULL, error); | |
465 | vfsp->vfs_sync_seq++; | |
466 | wmb(); | |
467 | wake_up(&vfsp->vfs_wait_single_sync_task); | |
468 | } | |
469 | ||
470 | STATIC int | |
471 | xfssyncd( | |
472 | void *arg) | |
473 | { | |
474 | long timeleft; | |
475 | vfs_t *vfsp = (vfs_t *) arg; | |
476 | struct list_head tmp; | |
477 | struct vfs_sync_work *work, *n; | |
478 | ||
479 | daemonize("xfssyncd"); | |
480 | ||
481 | vfsp->vfs_sync_work.w_vfs = vfsp; | |
482 | vfsp->vfs_sync_work.w_syncer = vfs_sync_worker; | |
483 | vfsp->vfs_sync_task = current; | |
484 | wmb(); | |
485 | wake_up(&vfsp->vfs_wait_sync_task); | |
486 | ||
487 | INIT_LIST_HEAD(&tmp); | |
488 | timeleft = (xfs_syncd_centisecs * HZ) / 100; | |
489 | for (;;) { | |
490 | set_current_state(TASK_INTERRUPTIBLE); | |
491 | timeleft = schedule_timeout(timeleft); | |
492 | /* swsusp */ | |
3e1d1d28 | 493 | try_to_freeze(); |
1da177e4 LT |
494 | if (vfsp->vfs_flag & VFS_UMOUNT) |
495 | break; | |
496 | ||
497 | spin_lock(&vfsp->vfs_sync_lock); | |
498 | /* | |
499 | * We can get woken by laptop mode, to do a sync - | |
500 | * that's the (only!) case where the list would be | |
501 | * empty with time remaining. | |
502 | */ | |
503 | if (!timeleft || list_empty(&vfsp->vfs_sync_list)) { | |
504 | if (!timeleft) | |
505 | timeleft = (xfs_syncd_centisecs * HZ) / 100; | |
506 | INIT_LIST_HEAD(&vfsp->vfs_sync_work.w_list); | |
507 | list_add_tail(&vfsp->vfs_sync_work.w_list, | |
508 | &vfsp->vfs_sync_list); | |
509 | } | |
510 | list_for_each_entry_safe(work, n, &vfsp->vfs_sync_list, w_list) | |
511 | list_move(&work->w_list, &tmp); | |
512 | spin_unlock(&vfsp->vfs_sync_lock); | |
513 | ||
514 | list_for_each_entry_safe(work, n, &tmp, w_list) { | |
515 | (*work->w_syncer)(vfsp, work->w_data); | |
516 | list_del(&work->w_list); | |
517 | if (work == &vfsp->vfs_sync_work) | |
518 | continue; | |
519 | kmem_free(work, sizeof(struct vfs_sync_work)); | |
520 | } | |
521 | } | |
522 | ||
523 | vfsp->vfs_sync_task = NULL; | |
524 | wmb(); | |
525 | wake_up(&vfsp->vfs_wait_sync_task); | |
526 | ||
527 | return 0; | |
528 | } | |
529 | ||
530 | STATIC int | |
531 | linvfs_start_syncd( | |
532 | vfs_t *vfsp) | |
533 | { | |
534 | int pid; | |
535 | ||
536 | pid = kernel_thread(xfssyncd, (void *) vfsp, | |
537 | CLONE_VM | CLONE_FS | CLONE_FILES); | |
538 | if (pid < 0) | |
539 | return -pid; | |
540 | wait_event(vfsp->vfs_wait_sync_task, vfsp->vfs_sync_task); | |
541 | return 0; | |
542 | } | |
543 | ||
544 | STATIC void | |
545 | linvfs_stop_syncd( | |
546 | vfs_t *vfsp) | |
547 | { | |
548 | vfsp->vfs_flag |= VFS_UMOUNT; | |
549 | wmb(); | |
550 | ||
551 | wake_up_process(vfsp->vfs_sync_task); | |
552 | wait_event(vfsp->vfs_wait_sync_task, !vfsp->vfs_sync_task); | |
553 | } | |
554 | ||
555 | STATIC void | |
556 | linvfs_put_super( | |
557 | struct super_block *sb) | |
558 | { | |
559 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
560 | int error; | |
561 | ||
562 | linvfs_stop_syncd(vfsp); | |
563 | VFS_SYNC(vfsp, SYNC_ATTR|SYNC_DELWRI, NULL, error); | |
564 | if (!error) | |
565 | VFS_UNMOUNT(vfsp, 0, NULL, error); | |
566 | if (error) { | |
567 | printk("XFS unmount got error %d\n", error); | |
568 | printk("%s: vfsp/0x%p left dangling!\n", __FUNCTION__, vfsp); | |
569 | return; | |
570 | } | |
571 | ||
572 | vfs_deallocate(vfsp); | |
573 | } | |
574 | ||
575 | STATIC void | |
576 | linvfs_write_super( | |
577 | struct super_block *sb) | |
578 | { | |
579 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
580 | int error; | |
581 | ||
582 | if (sb->s_flags & MS_RDONLY) { | |
583 | sb->s_dirt = 0; /* paranoia */ | |
584 | return; | |
585 | } | |
586 | /* Push the log and superblock a little */ | |
587 | VFS_SYNC(vfsp, SYNC_FSDATA, NULL, error); | |
588 | sb->s_dirt = 0; | |
589 | } | |
590 | ||
591 | STATIC int | |
592 | linvfs_sync_super( | |
593 | struct super_block *sb, | |
594 | int wait) | |
595 | { | |
596 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
597 | int error; | |
598 | int flags = SYNC_FSDATA; | |
599 | ||
f898d6c0 CH |
600 | if (unlikely(sb->s_frozen == SB_FREEZE_WRITE)) |
601 | flags = SYNC_QUIESCE; | |
602 | else | |
603 | flags = SYNC_FSDATA | (wait ? SYNC_WAIT : 0); | |
1da177e4 LT |
604 | |
605 | VFS_SYNC(vfsp, flags, NULL, error); | |
606 | sb->s_dirt = 0; | |
607 | ||
608 | if (unlikely(laptop_mode)) { | |
609 | int prev_sync_seq = vfsp->vfs_sync_seq; | |
610 | ||
611 | /* | |
612 | * The disk must be active because we're syncing. | |
613 | * We schedule xfssyncd now (now that the disk is | |
614 | * active) instead of later (when it might not be). | |
615 | */ | |
616 | wake_up_process(vfsp->vfs_sync_task); | |
617 | /* | |
618 | * We have to wait for the sync iteration to complete. | |
619 | * If we don't, the disk activity caused by the sync | |
620 | * will come after the sync is completed, and that | |
621 | * triggers another sync from laptop mode. | |
622 | */ | |
623 | wait_event(vfsp->vfs_wait_single_sync_task, | |
624 | vfsp->vfs_sync_seq != prev_sync_seq); | |
625 | } | |
626 | ||
627 | return -error; | |
628 | } | |
629 | ||
630 | STATIC int | |
631 | linvfs_statfs( | |
632 | struct super_block *sb, | |
633 | struct kstatfs *statp) | |
634 | { | |
635 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
636 | int error; | |
637 | ||
638 | VFS_STATVFS(vfsp, statp, NULL, error); | |
639 | return -error; | |
640 | } | |
641 | ||
642 | STATIC int | |
643 | linvfs_remount( | |
644 | struct super_block *sb, | |
645 | int *flags, | |
646 | char *options) | |
647 | { | |
648 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
649 | struct xfs_mount_args *args = xfs_args_allocate(sb); | |
650 | int error; | |
651 | ||
652 | VFS_PARSEARGS(vfsp, options, args, 1, error); | |
653 | if (!error) | |
654 | VFS_MNTUPDATE(vfsp, flags, args, error); | |
655 | kmem_free(args, sizeof(*args)); | |
656 | return -error; | |
657 | } | |
658 | ||
659 | STATIC void | |
660 | linvfs_freeze_fs( | |
661 | struct super_block *sb) | |
662 | { | |
663 | VFS_FREEZE(LINVFS_GET_VFS(sb)); | |
664 | } | |
665 | ||
666 | STATIC int | |
667 | linvfs_show_options( | |
668 | struct seq_file *m, | |
669 | struct vfsmount *mnt) | |
670 | { | |
671 | struct vfs *vfsp = LINVFS_GET_VFS(mnt->mnt_sb); | |
672 | int error; | |
673 | ||
674 | VFS_SHOWARGS(vfsp, m, error); | |
675 | return error; | |
676 | } | |
677 | ||
678 | STATIC int | |
679 | linvfs_getxstate( | |
680 | struct super_block *sb, | |
681 | struct fs_quota_stat *fqs) | |
682 | { | |
683 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
684 | int error; | |
685 | ||
686 | VFS_QUOTACTL(vfsp, Q_XGETQSTAT, 0, (caddr_t)fqs, error); | |
687 | return -error; | |
688 | } | |
689 | ||
690 | STATIC int | |
691 | linvfs_setxstate( | |
692 | struct super_block *sb, | |
693 | unsigned int flags, | |
694 | int op) | |
695 | { | |
696 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
697 | int error; | |
698 | ||
699 | VFS_QUOTACTL(vfsp, op, 0, (caddr_t)&flags, error); | |
700 | return -error; | |
701 | } | |
702 | ||
703 | STATIC int | |
704 | linvfs_getxquota( | |
705 | struct super_block *sb, | |
706 | int type, | |
707 | qid_t id, | |
708 | struct fs_disk_quota *fdq) | |
709 | { | |
710 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
711 | int error, getmode; | |
712 | ||
c8ad20ff NS |
713 | getmode = (type == USRQUOTA) ? Q_XGETQUOTA : |
714 | ((type == GRPQUOTA) ? Q_XGETGQUOTA : Q_XGETPQUOTA); | |
1da177e4 LT |
715 | VFS_QUOTACTL(vfsp, getmode, id, (caddr_t)fdq, error); |
716 | return -error; | |
717 | } | |
718 | ||
719 | STATIC int | |
720 | linvfs_setxquota( | |
721 | struct super_block *sb, | |
722 | int type, | |
723 | qid_t id, | |
724 | struct fs_disk_quota *fdq) | |
725 | { | |
726 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
727 | int error, setmode; | |
728 | ||
c8ad20ff NS |
729 | setmode = (type == USRQUOTA) ? Q_XSETQLIM : |
730 | ((type == GRPQUOTA) ? Q_XSETGQLIM : Q_XSETPQLIM); | |
1da177e4 LT |
731 | VFS_QUOTACTL(vfsp, setmode, id, (caddr_t)fdq, error); |
732 | return -error; | |
733 | } | |
734 | ||
735 | STATIC int | |
736 | linvfs_fill_super( | |
737 | struct super_block *sb, | |
738 | void *data, | |
739 | int silent) | |
740 | { | |
741 | vnode_t *rootvp; | |
742 | struct vfs *vfsp = vfs_allocate(); | |
743 | struct xfs_mount_args *args = xfs_args_allocate(sb); | |
744 | struct kstatfs statvfs; | |
745 | int error, error2; | |
746 | ||
747 | vfsp->vfs_super = sb; | |
748 | LINVFS_SET_VFS(sb, vfsp); | |
749 | if (sb->s_flags & MS_RDONLY) | |
750 | vfsp->vfs_flag |= VFS_RDONLY; | |
751 | bhv_insert_all_vfsops(vfsp); | |
752 | ||
753 | VFS_PARSEARGS(vfsp, (char *)data, args, 0, error); | |
754 | if (error) { | |
755 | bhv_remove_all_vfsops(vfsp, 1); | |
756 | goto fail_vfsop; | |
757 | } | |
758 | ||
759 | sb_min_blocksize(sb, BBSIZE); | |
760 | #ifdef CONFIG_XFS_EXPORT | |
761 | sb->s_export_op = &linvfs_export_ops; | |
762 | #endif | |
763 | sb->s_qcop = &linvfs_qops; | |
764 | sb->s_op = &linvfs_sops; | |
765 | ||
766 | VFS_MOUNT(vfsp, args, NULL, error); | |
767 | if (error) { | |
768 | bhv_remove_all_vfsops(vfsp, 1); | |
769 | goto fail_vfsop; | |
770 | } | |
771 | ||
772 | VFS_STATVFS(vfsp, &statvfs, NULL, error); | |
773 | if (error) | |
774 | goto fail_unmount; | |
775 | ||
776 | sb->s_dirt = 1; | |
777 | sb->s_magic = statvfs.f_type; | |
778 | sb->s_blocksize = statvfs.f_bsize; | |
779 | sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1; | |
780 | sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); | |
781 | sb->s_time_gran = 1; | |
782 | set_posix_acl_flag(sb); | |
783 | ||
784 | VFS_ROOT(vfsp, &rootvp, error); | |
785 | if (error) | |
786 | goto fail_unmount; | |
787 | ||
788 | sb->s_root = d_alloc_root(LINVFS_GET_IP(rootvp)); | |
789 | if (!sb->s_root) { | |
790 | error = ENOMEM; | |
791 | goto fail_vnrele; | |
792 | } | |
793 | if (is_bad_inode(sb->s_root->d_inode)) { | |
794 | error = EINVAL; | |
795 | goto fail_vnrele; | |
796 | } | |
797 | if ((error = linvfs_start_syncd(vfsp))) | |
798 | goto fail_vnrele; | |
799 | vn_trace_exit(rootvp, __FUNCTION__, (inst_t *)__return_address); | |
800 | ||
801 | kmem_free(args, sizeof(*args)); | |
802 | return 0; | |
803 | ||
804 | fail_vnrele: | |
805 | if (sb->s_root) { | |
806 | dput(sb->s_root); | |
807 | sb->s_root = NULL; | |
808 | } else { | |
809 | VN_RELE(rootvp); | |
810 | } | |
811 | ||
812 | fail_unmount: | |
813 | VFS_UNMOUNT(vfsp, 0, NULL, error2); | |
814 | ||
815 | fail_vfsop: | |
816 | vfs_deallocate(vfsp); | |
817 | kmem_free(args, sizeof(*args)); | |
818 | return -error; | |
819 | } | |
820 | ||
821 | STATIC struct super_block * | |
822 | linvfs_get_sb( | |
823 | struct file_system_type *fs_type, | |
824 | int flags, | |
825 | const char *dev_name, | |
826 | void *data) | |
827 | { | |
828 | return get_sb_bdev(fs_type, flags, dev_name, data, linvfs_fill_super); | |
829 | } | |
830 | ||
831 | STATIC struct super_operations linvfs_sops = { | |
832 | .alloc_inode = linvfs_alloc_inode, | |
833 | .destroy_inode = linvfs_destroy_inode, | |
834 | .write_inode = linvfs_write_inode, | |
835 | .clear_inode = linvfs_clear_inode, | |
836 | .put_super = linvfs_put_super, | |
837 | .write_super = linvfs_write_super, | |
838 | .sync_fs = linvfs_sync_super, | |
839 | .write_super_lockfs = linvfs_freeze_fs, | |
840 | .statfs = linvfs_statfs, | |
841 | .remount_fs = linvfs_remount, | |
842 | .show_options = linvfs_show_options, | |
843 | }; | |
844 | ||
845 | STATIC struct quotactl_ops linvfs_qops = { | |
846 | .get_xstate = linvfs_getxstate, | |
847 | .set_xstate = linvfs_setxstate, | |
848 | .get_xquota = linvfs_getxquota, | |
849 | .set_xquota = linvfs_setxquota, | |
850 | }; | |
851 | ||
852 | STATIC struct file_system_type xfs_fs_type = { | |
853 | .owner = THIS_MODULE, | |
854 | .name = "xfs", | |
855 | .get_sb = linvfs_get_sb, | |
856 | .kill_sb = kill_block_super, | |
857 | .fs_flags = FS_REQUIRES_DEV, | |
858 | }; | |
859 | ||
860 | ||
861 | STATIC int __init | |
862 | init_xfs_fs( void ) | |
863 | { | |
864 | int error; | |
865 | struct sysinfo si; | |
866 | static char message[] __initdata = KERN_INFO \ | |
867 | XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n"; | |
868 | ||
869 | printk(message); | |
870 | ||
871 | si_meminfo(&si); | |
872 | xfs_physmem = si.totalram; | |
873 | ||
874 | ktrace_init(64); | |
875 | ||
876 | error = init_inodecache(); | |
877 | if (error < 0) | |
878 | goto undo_inodecache; | |
879 | ||
880 | error = pagebuf_init(); | |
881 | if (error < 0) | |
882 | goto undo_pagebuf; | |
883 | ||
884 | vn_init(); | |
885 | xfs_init(); | |
886 | uuid_init(); | |
887 | vfs_initquota(); | |
888 | ||
889 | error = register_filesystem(&xfs_fs_type); | |
890 | if (error) | |
891 | goto undo_register; | |
892 | XFS_DM_INIT(&xfs_fs_type); | |
893 | return 0; | |
894 | ||
895 | undo_register: | |
896 | pagebuf_terminate(); | |
897 | ||
898 | undo_pagebuf: | |
899 | destroy_inodecache(); | |
900 | ||
901 | undo_inodecache: | |
902 | return error; | |
903 | } | |
904 | ||
905 | STATIC void __exit | |
906 | exit_xfs_fs( void ) | |
907 | { | |
908 | vfs_exitquota(); | |
909 | XFS_DM_EXIT(&xfs_fs_type); | |
910 | unregister_filesystem(&xfs_fs_type); | |
911 | xfs_cleanup(); | |
912 | pagebuf_terminate(); | |
913 | destroy_inodecache(); | |
914 | ktrace_uninit(); | |
915 | } | |
916 | ||
917 | module_init(init_xfs_fs); | |
918 | module_exit(exit_xfs_fs); | |
919 | ||
920 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
921 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
922 | MODULE_LICENSE("GPL"); |