]>
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 | { | |
141 | vnode_t *vp = LINVFS_GET_VP(inode); | |
142 | ||
143 | if (vp->v_type == VNON) { | |
144 | vn_mark_bad(vp); | |
145 | } else if (S_ISREG(inode->i_mode)) { | |
146 | inode->i_op = &linvfs_file_inode_operations; | |
147 | inode->i_fop = &linvfs_file_operations; | |
148 | inode->i_mapping->a_ops = &linvfs_aops; | |
149 | } else if (S_ISDIR(inode->i_mode)) { | |
150 | inode->i_op = &linvfs_dir_inode_operations; | |
151 | inode->i_fop = &linvfs_dir_operations; | |
152 | } else if (S_ISLNK(inode->i_mode)) { | |
153 | inode->i_op = &linvfs_symlink_inode_operations; | |
154 | if (inode->i_blocks) | |
155 | inode->i_mapping->a_ops = &linvfs_aops; | |
156 | } else { | |
157 | inode->i_op = &linvfs_file_inode_operations; | |
158 | init_special_inode(inode, inode->i_mode, inode->i_rdev); | |
159 | } | |
160 | } | |
161 | ||
162 | STATIC __inline__ void | |
163 | xfs_revalidate_inode( | |
164 | xfs_mount_t *mp, | |
165 | vnode_t *vp, | |
166 | xfs_inode_t *ip) | |
167 | { | |
168 | struct inode *inode = LINVFS_GET_IP(vp); | |
169 | ||
170 | inode->i_mode = (ip->i_d.di_mode & MODEMASK) | VTTOIF(vp->v_type); | |
171 | inode->i_nlink = ip->i_d.di_nlink; | |
172 | inode->i_uid = ip->i_d.di_uid; | |
173 | inode->i_gid = ip->i_d.di_gid; | |
174 | if (((1 << vp->v_type) & ((1<<VBLK) | (1<<VCHR))) == 0) { | |
175 | inode->i_rdev = 0; | |
176 | } else { | |
177 | xfs_dev_t dev = ip->i_df.if_u2.if_rdev; | |
178 | inode->i_rdev = MKDEV(sysv_major(dev) & 0x1ff, sysv_minor(dev)); | |
179 | } | |
180 | inode->i_blksize = PAGE_CACHE_SIZE; | |
181 | inode->i_generation = ip->i_d.di_gen; | |
182 | i_size_write(inode, ip->i_d.di_size); | |
183 | inode->i_blocks = | |
184 | XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks); | |
185 | inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec; | |
186 | inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec; | |
187 | inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec; | |
188 | inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec; | |
189 | inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec; | |
190 | inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec; | |
191 | if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE) | |
192 | inode->i_flags |= S_IMMUTABLE; | |
193 | else | |
194 | inode->i_flags &= ~S_IMMUTABLE; | |
195 | if (ip->i_d.di_flags & XFS_DIFLAG_APPEND) | |
196 | inode->i_flags |= S_APPEND; | |
197 | else | |
198 | inode->i_flags &= ~S_APPEND; | |
199 | if (ip->i_d.di_flags & XFS_DIFLAG_SYNC) | |
200 | inode->i_flags |= S_SYNC; | |
201 | else | |
202 | inode->i_flags &= ~S_SYNC; | |
203 | if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME) | |
204 | inode->i_flags |= S_NOATIME; | |
205 | else | |
206 | inode->i_flags &= ~S_NOATIME; | |
207 | vp->v_flag &= ~VMODIFIED; | |
208 | } | |
209 | ||
210 | void | |
211 | xfs_initialize_vnode( | |
212 | bhv_desc_t *bdp, | |
213 | vnode_t *vp, | |
214 | bhv_desc_t *inode_bhv, | |
215 | int unlock) | |
216 | { | |
217 | xfs_inode_t *ip = XFS_BHVTOI(inode_bhv); | |
218 | struct inode *inode = LINVFS_GET_IP(vp); | |
219 | ||
220 | if (!inode_bhv->bd_vobj) { | |
221 | vp->v_vfsp = bhvtovfs(bdp); | |
222 | bhv_desc_init(inode_bhv, ip, vp, &xfs_vnodeops); | |
223 | bhv_insert(VN_BHV_HEAD(vp), inode_bhv); | |
224 | } | |
225 | ||
226 | /* | |
227 | * We need to set the ops vectors, and unlock the inode, but if | |
228 | * we have been called during the new inode create process, it is | |
229 | * too early to fill in the Linux inode. We will get called a | |
230 | * second time once the inode is properly set up, and then we can | |
231 | * finish our work. | |
232 | */ | |
233 | if (ip->i_d.di_mode != 0 && unlock && (inode->i_state & I_NEW)) { | |
234 | vp->v_type = IFTOVT(ip->i_d.di_mode); | |
235 | xfs_revalidate_inode(XFS_BHVTOM(bdp), vp, ip); | |
236 | xfs_set_inodeops(inode); | |
237 | ||
238 | ip->i_flags &= ~XFS_INEW; | |
239 | barrier(); | |
240 | ||
241 | unlock_new_inode(inode); | |
242 | } | |
243 | } | |
244 | ||
245 | int | |
246 | xfs_blkdev_get( | |
247 | xfs_mount_t *mp, | |
248 | const char *name, | |
249 | struct block_device **bdevp) | |
250 | { | |
251 | int error = 0; | |
252 | ||
253 | *bdevp = open_bdev_excl(name, 0, mp); | |
254 | if (IS_ERR(*bdevp)) { | |
255 | error = PTR_ERR(*bdevp); | |
256 | printk("XFS: Invalid device [%s], error=%d\n", name, error); | |
257 | } | |
258 | ||
259 | return -error; | |
260 | } | |
261 | ||
262 | void | |
263 | xfs_blkdev_put( | |
264 | struct block_device *bdev) | |
265 | { | |
266 | if (bdev) | |
267 | close_bdev_excl(bdev); | |
268 | } | |
269 | ||
270 | ||
271 | STATIC struct inode * | |
272 | linvfs_alloc_inode( | |
273 | struct super_block *sb) | |
274 | { | |
275 | vnode_t *vp; | |
276 | ||
277 | vp = (vnode_t *)kmem_cache_alloc(linvfs_inode_zone, | |
278 | kmem_flags_convert(KM_SLEEP)); | |
279 | if (!vp) | |
280 | return NULL; | |
281 | return LINVFS_GET_IP(vp); | |
282 | } | |
283 | ||
284 | STATIC void | |
285 | linvfs_destroy_inode( | |
286 | struct inode *inode) | |
287 | { | |
288 | kmem_cache_free(linvfs_inode_zone, LINVFS_GET_VP(inode)); | |
289 | } | |
290 | ||
291 | STATIC void | |
292 | init_once( | |
293 | void *data, | |
294 | kmem_cache_t *cachep, | |
295 | unsigned long flags) | |
296 | { | |
297 | vnode_t *vp = (vnode_t *)data; | |
298 | ||
299 | if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) == | |
300 | SLAB_CTOR_CONSTRUCTOR) | |
301 | inode_init_once(LINVFS_GET_IP(vp)); | |
302 | } | |
303 | ||
304 | STATIC int | |
305 | init_inodecache( void ) | |
306 | { | |
307 | linvfs_inode_zone = kmem_cache_create("linvfs_icache", | |
308 | sizeof(vnode_t), 0, SLAB_RECLAIM_ACCOUNT, | |
309 | init_once, NULL); | |
310 | if (linvfs_inode_zone == NULL) | |
311 | return -ENOMEM; | |
312 | return 0; | |
313 | } | |
314 | ||
315 | STATIC void | |
316 | destroy_inodecache( void ) | |
317 | { | |
318 | if (kmem_cache_destroy(linvfs_inode_zone)) | |
319 | printk(KERN_WARNING "%s: cache still in use!\n", __FUNCTION__); | |
320 | } | |
321 | ||
322 | /* | |
323 | * Attempt to flush the inode, this will actually fail | |
324 | * if the inode is pinned, but we dirty the inode again | |
325 | * at the point when it is unpinned after a log write, | |
326 | * since this is when the inode itself becomes flushable. | |
327 | */ | |
328 | STATIC int | |
329 | linvfs_write_inode( | |
330 | struct inode *inode, | |
331 | int sync) | |
332 | { | |
333 | vnode_t *vp = LINVFS_GET_VP(inode); | |
334 | int error = 0, flags = FLUSH_INODE; | |
335 | ||
336 | if (vp) { | |
337 | vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); | |
338 | if (sync) | |
339 | flags |= FLUSH_SYNC; | |
340 | VOP_IFLUSH(vp, flags, error); | |
341 | if (error == EAGAIN) { | |
342 | if (sync) | |
343 | VOP_IFLUSH(vp, flags | FLUSH_LOG, error); | |
344 | else | |
345 | error = 0; | |
346 | } | |
347 | } | |
348 | ||
349 | return -error; | |
350 | } | |
351 | ||
352 | STATIC void | |
353 | linvfs_clear_inode( | |
354 | struct inode *inode) | |
355 | { | |
356 | vnode_t *vp = LINVFS_GET_VP(inode); | |
357 | ||
358 | if (vp) { | |
359 | vn_rele(vp); | |
360 | vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address); | |
361 | /* | |
362 | * Do all our cleanup, and remove this vnode. | |
363 | */ | |
364 | vn_remove(vp); | |
365 | } | |
366 | } | |
367 | ||
368 | ||
369 | /* | |
370 | * Enqueue a work item to be picked up by the vfs xfssyncd thread. | |
371 | * Doing this has two advantages: | |
372 | * - It saves on stack space, which is tight in certain situations | |
373 | * - It can be used (with care) as a mechanism to avoid deadlocks. | |
374 | * Flushing while allocating in a full filesystem requires both. | |
375 | */ | |
376 | STATIC void | |
377 | xfs_syncd_queue_work( | |
378 | struct vfs *vfs, | |
379 | void *data, | |
380 | void (*syncer)(vfs_t *, void *)) | |
381 | { | |
382 | vfs_sync_work_t *work; | |
383 | ||
384 | work = kmem_alloc(sizeof(struct vfs_sync_work), KM_SLEEP); | |
385 | INIT_LIST_HEAD(&work->w_list); | |
386 | work->w_syncer = syncer; | |
387 | work->w_data = data; | |
388 | work->w_vfs = vfs; | |
389 | spin_lock(&vfs->vfs_sync_lock); | |
390 | list_add_tail(&work->w_list, &vfs->vfs_sync_list); | |
391 | spin_unlock(&vfs->vfs_sync_lock); | |
392 | wake_up_process(vfs->vfs_sync_task); | |
393 | } | |
394 | ||
395 | /* | |
396 | * Flush delayed allocate data, attempting to free up reserved space | |
397 | * from existing allocations. At this point a new allocation attempt | |
398 | * has failed with ENOSPC and we are in the process of scratching our | |
399 | * heads, looking about for more room... | |
400 | */ | |
401 | STATIC void | |
402 | xfs_flush_inode_work( | |
403 | vfs_t *vfs, | |
404 | void *inode) | |
405 | { | |
406 | filemap_flush(((struct inode *)inode)->i_mapping); | |
407 | iput((struct inode *)inode); | |
408 | } | |
409 | ||
410 | void | |
411 | xfs_flush_inode( | |
412 | xfs_inode_t *ip) | |
413 | { | |
414 | struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); | |
415 | struct vfs *vfs = XFS_MTOVFS(ip->i_mount); | |
416 | ||
417 | igrab(inode); | |
418 | xfs_syncd_queue_work(vfs, inode, xfs_flush_inode_work); | |
419 | delay(HZ/2); | |
420 | } | |
421 | ||
422 | /* | |
423 | * This is the "bigger hammer" version of xfs_flush_inode_work... | |
424 | * (IOW, "If at first you don't succeed, use a Bigger Hammer"). | |
425 | */ | |
426 | STATIC void | |
427 | xfs_flush_device_work( | |
428 | vfs_t *vfs, | |
429 | void *inode) | |
430 | { | |
431 | sync_blockdev(vfs->vfs_super->s_bdev); | |
432 | iput((struct inode *)inode); | |
433 | } | |
434 | ||
435 | void | |
436 | xfs_flush_device( | |
437 | xfs_inode_t *ip) | |
438 | { | |
439 | struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip)); | |
440 | struct vfs *vfs = XFS_MTOVFS(ip->i_mount); | |
441 | ||
442 | igrab(inode); | |
443 | xfs_syncd_queue_work(vfs, inode, xfs_flush_device_work); | |
444 | delay(HZ/2); | |
445 | xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC); | |
446 | } | |
447 | ||
448 | #define SYNCD_FLAGS (SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR) | |
449 | STATIC void | |
450 | vfs_sync_worker( | |
451 | vfs_t *vfsp, | |
452 | void *unused) | |
453 | { | |
454 | int error; | |
455 | ||
456 | if (!(vfsp->vfs_flag & VFS_RDONLY)) | |
457 | VFS_SYNC(vfsp, SYNCD_FLAGS, NULL, error); | |
458 | vfsp->vfs_sync_seq++; | |
459 | wmb(); | |
460 | wake_up(&vfsp->vfs_wait_single_sync_task); | |
461 | } | |
462 | ||
463 | STATIC int | |
464 | xfssyncd( | |
465 | void *arg) | |
466 | { | |
467 | long timeleft; | |
468 | vfs_t *vfsp = (vfs_t *) arg; | |
469 | struct list_head tmp; | |
470 | struct vfs_sync_work *work, *n; | |
471 | ||
472 | daemonize("xfssyncd"); | |
473 | ||
474 | vfsp->vfs_sync_work.w_vfs = vfsp; | |
475 | vfsp->vfs_sync_work.w_syncer = vfs_sync_worker; | |
476 | vfsp->vfs_sync_task = current; | |
477 | wmb(); | |
478 | wake_up(&vfsp->vfs_wait_sync_task); | |
479 | ||
480 | INIT_LIST_HEAD(&tmp); | |
481 | timeleft = (xfs_syncd_centisecs * HZ) / 100; | |
482 | for (;;) { | |
483 | set_current_state(TASK_INTERRUPTIBLE); | |
484 | timeleft = schedule_timeout(timeleft); | |
485 | /* swsusp */ | |
3e1d1d28 | 486 | try_to_freeze(); |
1da177e4 LT |
487 | if (vfsp->vfs_flag & VFS_UMOUNT) |
488 | break; | |
489 | ||
490 | spin_lock(&vfsp->vfs_sync_lock); | |
491 | /* | |
492 | * We can get woken by laptop mode, to do a sync - | |
493 | * that's the (only!) case where the list would be | |
494 | * empty with time remaining. | |
495 | */ | |
496 | if (!timeleft || list_empty(&vfsp->vfs_sync_list)) { | |
497 | if (!timeleft) | |
498 | timeleft = (xfs_syncd_centisecs * HZ) / 100; | |
499 | INIT_LIST_HEAD(&vfsp->vfs_sync_work.w_list); | |
500 | list_add_tail(&vfsp->vfs_sync_work.w_list, | |
501 | &vfsp->vfs_sync_list); | |
502 | } | |
503 | list_for_each_entry_safe(work, n, &vfsp->vfs_sync_list, w_list) | |
504 | list_move(&work->w_list, &tmp); | |
505 | spin_unlock(&vfsp->vfs_sync_lock); | |
506 | ||
507 | list_for_each_entry_safe(work, n, &tmp, w_list) { | |
508 | (*work->w_syncer)(vfsp, work->w_data); | |
509 | list_del(&work->w_list); | |
510 | if (work == &vfsp->vfs_sync_work) | |
511 | continue; | |
512 | kmem_free(work, sizeof(struct vfs_sync_work)); | |
513 | } | |
514 | } | |
515 | ||
516 | vfsp->vfs_sync_task = NULL; | |
517 | wmb(); | |
518 | wake_up(&vfsp->vfs_wait_sync_task); | |
519 | ||
520 | return 0; | |
521 | } | |
522 | ||
523 | STATIC int | |
524 | linvfs_start_syncd( | |
525 | vfs_t *vfsp) | |
526 | { | |
527 | int pid; | |
528 | ||
529 | pid = kernel_thread(xfssyncd, (void *) vfsp, | |
530 | CLONE_VM | CLONE_FS | CLONE_FILES); | |
531 | if (pid < 0) | |
532 | return -pid; | |
533 | wait_event(vfsp->vfs_wait_sync_task, vfsp->vfs_sync_task); | |
534 | return 0; | |
535 | } | |
536 | ||
537 | STATIC void | |
538 | linvfs_stop_syncd( | |
539 | vfs_t *vfsp) | |
540 | { | |
541 | vfsp->vfs_flag |= VFS_UMOUNT; | |
542 | wmb(); | |
543 | ||
544 | wake_up_process(vfsp->vfs_sync_task); | |
545 | wait_event(vfsp->vfs_wait_sync_task, !vfsp->vfs_sync_task); | |
546 | } | |
547 | ||
548 | STATIC void | |
549 | linvfs_put_super( | |
550 | struct super_block *sb) | |
551 | { | |
552 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
553 | int error; | |
554 | ||
555 | linvfs_stop_syncd(vfsp); | |
556 | VFS_SYNC(vfsp, SYNC_ATTR|SYNC_DELWRI, NULL, error); | |
557 | if (!error) | |
558 | VFS_UNMOUNT(vfsp, 0, NULL, error); | |
559 | if (error) { | |
560 | printk("XFS unmount got error %d\n", error); | |
561 | printk("%s: vfsp/0x%p left dangling!\n", __FUNCTION__, vfsp); | |
562 | return; | |
563 | } | |
564 | ||
565 | vfs_deallocate(vfsp); | |
566 | } | |
567 | ||
568 | STATIC void | |
569 | linvfs_write_super( | |
570 | struct super_block *sb) | |
571 | { | |
572 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
573 | int error; | |
574 | ||
575 | if (sb->s_flags & MS_RDONLY) { | |
576 | sb->s_dirt = 0; /* paranoia */ | |
577 | return; | |
578 | } | |
579 | /* Push the log and superblock a little */ | |
580 | VFS_SYNC(vfsp, SYNC_FSDATA, NULL, error); | |
581 | sb->s_dirt = 0; | |
582 | } | |
583 | ||
584 | STATIC int | |
585 | linvfs_sync_super( | |
586 | struct super_block *sb, | |
587 | int wait) | |
588 | { | |
589 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
590 | int error; | |
591 | int flags = SYNC_FSDATA; | |
592 | ||
f898d6c0 CH |
593 | if (unlikely(sb->s_frozen == SB_FREEZE_WRITE)) |
594 | flags = SYNC_QUIESCE; | |
595 | else | |
596 | flags = SYNC_FSDATA | (wait ? SYNC_WAIT : 0); | |
1da177e4 LT |
597 | |
598 | VFS_SYNC(vfsp, flags, NULL, error); | |
599 | sb->s_dirt = 0; | |
600 | ||
601 | if (unlikely(laptop_mode)) { | |
602 | int prev_sync_seq = vfsp->vfs_sync_seq; | |
603 | ||
604 | /* | |
605 | * The disk must be active because we're syncing. | |
606 | * We schedule xfssyncd now (now that the disk is | |
607 | * active) instead of later (when it might not be). | |
608 | */ | |
609 | wake_up_process(vfsp->vfs_sync_task); | |
610 | /* | |
611 | * We have to wait for the sync iteration to complete. | |
612 | * If we don't, the disk activity caused by the sync | |
613 | * will come after the sync is completed, and that | |
614 | * triggers another sync from laptop mode. | |
615 | */ | |
616 | wait_event(vfsp->vfs_wait_single_sync_task, | |
617 | vfsp->vfs_sync_seq != prev_sync_seq); | |
618 | } | |
619 | ||
620 | return -error; | |
621 | } | |
622 | ||
623 | STATIC int | |
624 | linvfs_statfs( | |
625 | struct super_block *sb, | |
626 | struct kstatfs *statp) | |
627 | { | |
628 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
629 | int error; | |
630 | ||
631 | VFS_STATVFS(vfsp, statp, NULL, error); | |
632 | return -error; | |
633 | } | |
634 | ||
635 | STATIC int | |
636 | linvfs_remount( | |
637 | struct super_block *sb, | |
638 | int *flags, | |
639 | char *options) | |
640 | { | |
641 | vfs_t *vfsp = LINVFS_GET_VFS(sb); | |
642 | struct xfs_mount_args *args = xfs_args_allocate(sb); | |
643 | int error; | |
644 | ||
645 | VFS_PARSEARGS(vfsp, options, args, 1, error); | |
646 | if (!error) | |
647 | VFS_MNTUPDATE(vfsp, flags, args, error); | |
648 | kmem_free(args, sizeof(*args)); | |
649 | return -error; | |
650 | } | |
651 | ||
652 | STATIC void | |
653 | linvfs_freeze_fs( | |
654 | struct super_block *sb) | |
655 | { | |
656 | VFS_FREEZE(LINVFS_GET_VFS(sb)); | |
657 | } | |
658 | ||
659 | STATIC int | |
660 | linvfs_show_options( | |
661 | struct seq_file *m, | |
662 | struct vfsmount *mnt) | |
663 | { | |
664 | struct vfs *vfsp = LINVFS_GET_VFS(mnt->mnt_sb); | |
665 | int error; | |
666 | ||
667 | VFS_SHOWARGS(vfsp, m, error); | |
668 | return error; | |
669 | } | |
670 | ||
671 | STATIC int | |
672 | linvfs_getxstate( | |
673 | struct super_block *sb, | |
674 | struct fs_quota_stat *fqs) | |
675 | { | |
676 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
677 | int error; | |
678 | ||
679 | VFS_QUOTACTL(vfsp, Q_XGETQSTAT, 0, (caddr_t)fqs, error); | |
680 | return -error; | |
681 | } | |
682 | ||
683 | STATIC int | |
684 | linvfs_setxstate( | |
685 | struct super_block *sb, | |
686 | unsigned int flags, | |
687 | int op) | |
688 | { | |
689 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
690 | int error; | |
691 | ||
692 | VFS_QUOTACTL(vfsp, op, 0, (caddr_t)&flags, error); | |
693 | return -error; | |
694 | } | |
695 | ||
696 | STATIC int | |
697 | linvfs_getxquota( | |
698 | struct super_block *sb, | |
699 | int type, | |
700 | qid_t id, | |
701 | struct fs_disk_quota *fdq) | |
702 | { | |
703 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
704 | int error, getmode; | |
705 | ||
c8ad20ff NS |
706 | getmode = (type == USRQUOTA) ? Q_XGETQUOTA : |
707 | ((type == GRPQUOTA) ? Q_XGETGQUOTA : Q_XGETPQUOTA); | |
1da177e4 LT |
708 | VFS_QUOTACTL(vfsp, getmode, id, (caddr_t)fdq, error); |
709 | return -error; | |
710 | } | |
711 | ||
712 | STATIC int | |
713 | linvfs_setxquota( | |
714 | struct super_block *sb, | |
715 | int type, | |
716 | qid_t id, | |
717 | struct fs_disk_quota *fdq) | |
718 | { | |
719 | struct vfs *vfsp = LINVFS_GET_VFS(sb); | |
720 | int error, setmode; | |
721 | ||
c8ad20ff NS |
722 | setmode = (type == USRQUOTA) ? Q_XSETQLIM : |
723 | ((type == GRPQUOTA) ? Q_XSETGQLIM : Q_XSETPQLIM); | |
1da177e4 LT |
724 | VFS_QUOTACTL(vfsp, setmode, id, (caddr_t)fdq, error); |
725 | return -error; | |
726 | } | |
727 | ||
728 | STATIC int | |
729 | linvfs_fill_super( | |
730 | struct super_block *sb, | |
731 | void *data, | |
732 | int silent) | |
733 | { | |
734 | vnode_t *rootvp; | |
735 | struct vfs *vfsp = vfs_allocate(); | |
736 | struct xfs_mount_args *args = xfs_args_allocate(sb); | |
737 | struct kstatfs statvfs; | |
738 | int error, error2; | |
739 | ||
740 | vfsp->vfs_super = sb; | |
741 | LINVFS_SET_VFS(sb, vfsp); | |
742 | if (sb->s_flags & MS_RDONLY) | |
743 | vfsp->vfs_flag |= VFS_RDONLY; | |
744 | bhv_insert_all_vfsops(vfsp); | |
745 | ||
746 | VFS_PARSEARGS(vfsp, (char *)data, args, 0, error); | |
747 | if (error) { | |
748 | bhv_remove_all_vfsops(vfsp, 1); | |
749 | goto fail_vfsop; | |
750 | } | |
751 | ||
752 | sb_min_blocksize(sb, BBSIZE); | |
753 | #ifdef CONFIG_XFS_EXPORT | |
754 | sb->s_export_op = &linvfs_export_ops; | |
755 | #endif | |
756 | sb->s_qcop = &linvfs_qops; | |
757 | sb->s_op = &linvfs_sops; | |
758 | ||
759 | VFS_MOUNT(vfsp, args, NULL, error); | |
760 | if (error) { | |
761 | bhv_remove_all_vfsops(vfsp, 1); | |
762 | goto fail_vfsop; | |
763 | } | |
764 | ||
765 | VFS_STATVFS(vfsp, &statvfs, NULL, error); | |
766 | if (error) | |
767 | goto fail_unmount; | |
768 | ||
769 | sb->s_dirt = 1; | |
770 | sb->s_magic = statvfs.f_type; | |
771 | sb->s_blocksize = statvfs.f_bsize; | |
772 | sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1; | |
773 | sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits); | |
774 | sb->s_time_gran = 1; | |
775 | set_posix_acl_flag(sb); | |
776 | ||
777 | VFS_ROOT(vfsp, &rootvp, error); | |
778 | if (error) | |
779 | goto fail_unmount; | |
780 | ||
781 | sb->s_root = d_alloc_root(LINVFS_GET_IP(rootvp)); | |
782 | if (!sb->s_root) { | |
783 | error = ENOMEM; | |
784 | goto fail_vnrele; | |
785 | } | |
786 | if (is_bad_inode(sb->s_root->d_inode)) { | |
787 | error = EINVAL; | |
788 | goto fail_vnrele; | |
789 | } | |
790 | if ((error = linvfs_start_syncd(vfsp))) | |
791 | goto fail_vnrele; | |
792 | vn_trace_exit(rootvp, __FUNCTION__, (inst_t *)__return_address); | |
793 | ||
794 | kmem_free(args, sizeof(*args)); | |
795 | return 0; | |
796 | ||
797 | fail_vnrele: | |
798 | if (sb->s_root) { | |
799 | dput(sb->s_root); | |
800 | sb->s_root = NULL; | |
801 | } else { | |
802 | VN_RELE(rootvp); | |
803 | } | |
804 | ||
805 | fail_unmount: | |
806 | VFS_UNMOUNT(vfsp, 0, NULL, error2); | |
807 | ||
808 | fail_vfsop: | |
809 | vfs_deallocate(vfsp); | |
810 | kmem_free(args, sizeof(*args)); | |
811 | return -error; | |
812 | } | |
813 | ||
814 | STATIC struct super_block * | |
815 | linvfs_get_sb( | |
816 | struct file_system_type *fs_type, | |
817 | int flags, | |
818 | const char *dev_name, | |
819 | void *data) | |
820 | { | |
821 | return get_sb_bdev(fs_type, flags, dev_name, data, linvfs_fill_super); | |
822 | } | |
823 | ||
824 | STATIC struct super_operations linvfs_sops = { | |
825 | .alloc_inode = linvfs_alloc_inode, | |
826 | .destroy_inode = linvfs_destroy_inode, | |
827 | .write_inode = linvfs_write_inode, | |
828 | .clear_inode = linvfs_clear_inode, | |
829 | .put_super = linvfs_put_super, | |
830 | .write_super = linvfs_write_super, | |
831 | .sync_fs = linvfs_sync_super, | |
832 | .write_super_lockfs = linvfs_freeze_fs, | |
833 | .statfs = linvfs_statfs, | |
834 | .remount_fs = linvfs_remount, | |
835 | .show_options = linvfs_show_options, | |
836 | }; | |
837 | ||
838 | STATIC struct quotactl_ops linvfs_qops = { | |
839 | .get_xstate = linvfs_getxstate, | |
840 | .set_xstate = linvfs_setxstate, | |
841 | .get_xquota = linvfs_getxquota, | |
842 | .set_xquota = linvfs_setxquota, | |
843 | }; | |
844 | ||
845 | STATIC struct file_system_type xfs_fs_type = { | |
846 | .owner = THIS_MODULE, | |
847 | .name = "xfs", | |
848 | .get_sb = linvfs_get_sb, | |
849 | .kill_sb = kill_block_super, | |
850 | .fs_flags = FS_REQUIRES_DEV, | |
851 | }; | |
852 | ||
853 | ||
854 | STATIC int __init | |
855 | init_xfs_fs( void ) | |
856 | { | |
857 | int error; | |
858 | struct sysinfo si; | |
859 | static char message[] __initdata = KERN_INFO \ | |
860 | XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n"; | |
861 | ||
862 | printk(message); | |
863 | ||
864 | si_meminfo(&si); | |
865 | xfs_physmem = si.totalram; | |
866 | ||
867 | ktrace_init(64); | |
868 | ||
869 | error = init_inodecache(); | |
870 | if (error < 0) | |
871 | goto undo_inodecache; | |
872 | ||
873 | error = pagebuf_init(); | |
874 | if (error < 0) | |
875 | goto undo_pagebuf; | |
876 | ||
877 | vn_init(); | |
878 | xfs_init(); | |
879 | uuid_init(); | |
880 | vfs_initquota(); | |
881 | ||
882 | error = register_filesystem(&xfs_fs_type); | |
883 | if (error) | |
884 | goto undo_register; | |
885 | XFS_DM_INIT(&xfs_fs_type); | |
886 | return 0; | |
887 | ||
888 | undo_register: | |
889 | pagebuf_terminate(); | |
890 | ||
891 | undo_pagebuf: | |
892 | destroy_inodecache(); | |
893 | ||
894 | undo_inodecache: | |
895 | return error; | |
896 | } | |
897 | ||
898 | STATIC void __exit | |
899 | exit_xfs_fs( void ) | |
900 | { | |
901 | vfs_exitquota(); | |
902 | XFS_DM_EXIT(&xfs_fs_type); | |
903 | unregister_filesystem(&xfs_fs_type); | |
904 | xfs_cleanup(); | |
905 | pagebuf_terminate(); | |
906 | destroy_inodecache(); | |
907 | ktrace_uninit(); | |
908 | } | |
909 | ||
910 | module_init(init_xfs_fs); | |
911 | module_exit(exit_xfs_fs); | |
912 | ||
913 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
914 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
915 | MODULE_LICENSE("GPL"); |