Commit | Line | Data |
---|---|---|
fde2227c DC |
1 | /* |
2 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. | |
3 | * Copyright (c) 2013 Red Hat, Inc. | |
4 | * All Rights Reserved. | |
5 | * | |
6 | * This program is free software; you can redistribute it and/or | |
7 | * modify it under the terms of the GNU General Public License as | |
8 | * published by the Free Software Foundation. | |
9 | * | |
10 | * This program is distributed in the hope that it would be useful, | |
11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
13 | * GNU General Public License for more details. | |
14 | * | |
15 | * You should have received a copy of the GNU General Public License | |
16 | * along with this program; if not, write the Free Software Foundation, | |
17 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
18 | */ | |
19 | #include "xfs.h" | |
20 | #include "xfs_fs.h" | |
21 | #include "xfs_format.h" | |
70a9883c | 22 | #include "xfs_shared.h" |
fde2227c DC |
23 | #include "xfs_bit.h" |
24 | #include "xfs_log.h" | |
25 | #include "xfs_trans.h" | |
26 | #include "xfs_sb.h" | |
27 | #include "xfs_ag.h" | |
28 | #include "xfs_mount.h" | |
29 | #include "xfs_da_btree.h" | |
30 | #include "xfs_bmap_btree.h" | |
31 | #include "xfs_alloc_btree.h" | |
32 | #include "xfs_ialloc_btree.h" | |
33 | #include "xfs_alloc.h" | |
34 | #include "xfs_btree.h" | |
35 | #include "xfs_attr_remote.h" | |
36 | #include "xfs_dinode.h" | |
37 | #include "xfs_inode.h" | |
38 | #include "xfs_inode_item.h" | |
39 | #include "xfs_bmap.h" | |
40 | #include "xfs_attr.h" | |
41 | #include "xfs_attr_leaf.h" | |
42 | #include "xfs_error.h" | |
43 | #include "xfs_quota.h" | |
44 | #include "xfs_trace.h" | |
45 | #include "xfs_trans_priv.h" | |
46 | ||
47 | /* | |
48 | * Look at all the extents for this logical region, | |
49 | * invalidate any buffers that are incore/in transactions. | |
50 | */ | |
51 | STATIC int | |
52 | xfs_attr3_leaf_freextent( | |
53 | struct xfs_trans **trans, | |
54 | struct xfs_inode *dp, | |
55 | xfs_dablk_t blkno, | |
56 | int blkcnt) | |
57 | { | |
58 | struct xfs_bmbt_irec map; | |
59 | struct xfs_buf *bp; | |
60 | xfs_dablk_t tblkno; | |
61 | xfs_daddr_t dblkno; | |
62 | int tblkcnt; | |
63 | int dblkcnt; | |
64 | int nmap; | |
65 | int error; | |
66 | ||
67 | /* | |
68 | * Roll through the "value", invalidating the attribute value's | |
69 | * blocks. | |
70 | */ | |
71 | tblkno = blkno; | |
72 | tblkcnt = blkcnt; | |
73 | while (tblkcnt > 0) { | |
74 | /* | |
75 | * Try to remember where we decided to put the value. | |
76 | */ | |
77 | nmap = 1; | |
78 | error = xfs_bmapi_read(dp, (xfs_fileoff_t)tblkno, tblkcnt, | |
79 | &map, &nmap, XFS_BMAPI_ATTRFORK); | |
80 | if (error) { | |
81 | return(error); | |
82 | } | |
83 | ASSERT(nmap == 1); | |
84 | ASSERT(map.br_startblock != DELAYSTARTBLOCK); | |
85 | ||
86 | /* | |
87 | * If it's a hole, these are already unmapped | |
88 | * so there's nothing to invalidate. | |
89 | */ | |
90 | if (map.br_startblock != HOLESTARTBLOCK) { | |
91 | ||
92 | dblkno = XFS_FSB_TO_DADDR(dp->i_mount, | |
93 | map.br_startblock); | |
94 | dblkcnt = XFS_FSB_TO_BB(dp->i_mount, | |
95 | map.br_blockcount); | |
96 | bp = xfs_trans_get_buf(*trans, | |
97 | dp->i_mount->m_ddev_targp, | |
98 | dblkno, dblkcnt, 0); | |
99 | if (!bp) | |
100 | return ENOMEM; | |
101 | xfs_trans_binval(*trans, bp); | |
102 | /* | |
103 | * Roll to next transaction. | |
104 | */ | |
105 | error = xfs_trans_roll(trans, dp); | |
106 | if (error) | |
107 | return (error); | |
108 | } | |
109 | ||
110 | tblkno += map.br_blockcount; | |
111 | tblkcnt -= map.br_blockcount; | |
112 | } | |
113 | ||
114 | return(0); | |
115 | } | |
116 | ||
117 | /* | |
118 | * Invalidate all of the "remote" value regions pointed to by a particular | |
119 | * leaf block. | |
120 | * Note that we must release the lock on the buffer so that we are not | |
121 | * caught holding something that the logging code wants to flush to disk. | |
122 | */ | |
123 | STATIC int | |
124 | xfs_attr3_leaf_inactive( | |
125 | struct xfs_trans **trans, | |
126 | struct xfs_inode *dp, | |
127 | struct xfs_buf *bp) | |
128 | { | |
129 | struct xfs_attr_leafblock *leaf; | |
130 | struct xfs_attr3_icleaf_hdr ichdr; | |
131 | struct xfs_attr_leaf_entry *entry; | |
132 | struct xfs_attr_leaf_name_remote *name_rmt; | |
133 | struct xfs_attr_inactive_list *list; | |
134 | struct xfs_attr_inactive_list *lp; | |
135 | int error; | |
136 | int count; | |
137 | int size; | |
138 | int tmp; | |
139 | int i; | |
140 | ||
141 | leaf = bp->b_addr; | |
142 | xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf); | |
143 | ||
144 | /* | |
145 | * Count the number of "remote" value extents. | |
146 | */ | |
147 | count = 0; | |
148 | entry = xfs_attr3_leaf_entryp(leaf); | |
149 | for (i = 0; i < ichdr.count; entry++, i++) { | |
150 | if (be16_to_cpu(entry->nameidx) && | |
151 | ((entry->flags & XFS_ATTR_LOCAL) == 0)) { | |
152 | name_rmt = xfs_attr3_leaf_name_remote(leaf, i); | |
153 | if (name_rmt->valueblk) | |
154 | count++; | |
155 | } | |
156 | } | |
157 | ||
158 | /* | |
159 | * If there are no "remote" values, we're done. | |
160 | */ | |
161 | if (count == 0) { | |
162 | xfs_trans_brelse(*trans, bp); | |
163 | return 0; | |
164 | } | |
165 | ||
166 | /* | |
167 | * Allocate storage for a list of all the "remote" value extents. | |
168 | */ | |
169 | size = count * sizeof(xfs_attr_inactive_list_t); | |
170 | list = kmem_alloc(size, KM_SLEEP); | |
171 | ||
172 | /* | |
173 | * Identify each of the "remote" value extents. | |
174 | */ | |
175 | lp = list; | |
176 | entry = xfs_attr3_leaf_entryp(leaf); | |
177 | for (i = 0; i < ichdr.count; entry++, i++) { | |
178 | if (be16_to_cpu(entry->nameidx) && | |
179 | ((entry->flags & XFS_ATTR_LOCAL) == 0)) { | |
180 | name_rmt = xfs_attr3_leaf_name_remote(leaf, i); | |
181 | if (name_rmt->valueblk) { | |
182 | lp->valueblk = be32_to_cpu(name_rmt->valueblk); | |
183 | lp->valuelen = xfs_attr3_rmt_blocks(dp->i_mount, | |
184 | be32_to_cpu(name_rmt->valuelen)); | |
185 | lp++; | |
186 | } | |
187 | } | |
188 | } | |
189 | xfs_trans_brelse(*trans, bp); /* unlock for trans. in freextent() */ | |
190 | ||
191 | /* | |
192 | * Invalidate each of the "remote" value extents. | |
193 | */ | |
194 | error = 0; | |
195 | for (lp = list, i = 0; i < count; i++, lp++) { | |
196 | tmp = xfs_attr3_leaf_freextent(trans, dp, | |
197 | lp->valueblk, lp->valuelen); | |
198 | ||
199 | if (error == 0) | |
200 | error = tmp; /* save only the 1st errno */ | |
201 | } | |
202 | ||
203 | kmem_free(list); | |
204 | return error; | |
205 | } | |
206 | ||
207 | /* | |
208 | * Recurse (gasp!) through the attribute nodes until we find leaves. | |
209 | * We're doing a depth-first traversal in order to invalidate everything. | |
210 | */ | |
211 | STATIC int | |
212 | xfs_attr3_node_inactive( | |
213 | struct xfs_trans **trans, | |
214 | struct xfs_inode *dp, | |
215 | struct xfs_buf *bp, | |
216 | int level) | |
217 | { | |
218 | xfs_da_blkinfo_t *info; | |
219 | xfs_da_intnode_t *node; | |
220 | xfs_dablk_t child_fsb; | |
221 | xfs_daddr_t parent_blkno, child_blkno; | |
222 | int error, i; | |
223 | struct xfs_buf *child_bp; | |
224 | struct xfs_da_node_entry *btree; | |
225 | struct xfs_da3_icnode_hdr ichdr; | |
226 | ||
227 | /* | |
228 | * Since this code is recursive (gasp!) we must protect ourselves. | |
229 | */ | |
230 | if (level > XFS_DA_NODE_MAXDEPTH) { | |
231 | xfs_trans_brelse(*trans, bp); /* no locks for later trans */ | |
232 | return XFS_ERROR(EIO); | |
233 | } | |
234 | ||
235 | node = bp->b_addr; | |
236 | xfs_da3_node_hdr_from_disk(&ichdr, node); | |
237 | parent_blkno = bp->b_bn; | |
238 | if (!ichdr.count) { | |
239 | xfs_trans_brelse(*trans, bp); | |
240 | return 0; | |
241 | } | |
242 | btree = xfs_da3_node_tree_p(node); | |
243 | child_fsb = be32_to_cpu(btree[0].before); | |
244 | xfs_trans_brelse(*trans, bp); /* no locks for later trans */ | |
245 | ||
246 | /* | |
247 | * If this is the node level just above the leaves, simply loop | |
248 | * over the leaves removing all of them. If this is higher up | |
249 | * in the tree, recurse downward. | |
250 | */ | |
251 | for (i = 0; i < ichdr.count; i++) { | |
252 | /* | |
253 | * Read the subsidiary block to see what we have to work with. | |
254 | * Don't do this in a transaction. This is a depth-first | |
255 | * traversal of the tree so we may deal with many blocks | |
256 | * before we come back to this one. | |
257 | */ | |
258 | error = xfs_da3_node_read(*trans, dp, child_fsb, -2, &child_bp, | |
259 | XFS_ATTR_FORK); | |
260 | if (error) | |
261 | return(error); | |
262 | if (child_bp) { | |
263 | /* save for re-read later */ | |
264 | child_blkno = XFS_BUF_ADDR(child_bp); | |
265 | ||
266 | /* | |
267 | * Invalidate the subtree, however we have to. | |
268 | */ | |
269 | info = child_bp->b_addr; | |
270 | switch (info->magic) { | |
271 | case cpu_to_be16(XFS_DA_NODE_MAGIC): | |
272 | case cpu_to_be16(XFS_DA3_NODE_MAGIC): | |
273 | error = xfs_attr3_node_inactive(trans, dp, | |
274 | child_bp, level + 1); | |
275 | break; | |
276 | case cpu_to_be16(XFS_ATTR_LEAF_MAGIC): | |
277 | case cpu_to_be16(XFS_ATTR3_LEAF_MAGIC): | |
278 | error = xfs_attr3_leaf_inactive(trans, dp, | |
279 | child_bp); | |
280 | break; | |
281 | default: | |
282 | error = XFS_ERROR(EIO); | |
283 | xfs_trans_brelse(*trans, child_bp); | |
284 | break; | |
285 | } | |
286 | if (error) | |
287 | return error; | |
288 | ||
289 | /* | |
290 | * Remove the subsidiary block from the cache | |
291 | * and from the log. | |
292 | */ | |
293 | error = xfs_da_get_buf(*trans, dp, 0, child_blkno, | |
294 | &child_bp, XFS_ATTR_FORK); | |
295 | if (error) | |
296 | return error; | |
297 | xfs_trans_binval(*trans, child_bp); | |
298 | } | |
299 | ||
300 | /* | |
301 | * If we're not done, re-read the parent to get the next | |
302 | * child block number. | |
303 | */ | |
304 | if (i + 1 < ichdr.count) { | |
305 | error = xfs_da3_node_read(*trans, dp, 0, parent_blkno, | |
306 | &bp, XFS_ATTR_FORK); | |
307 | if (error) | |
308 | return error; | |
309 | child_fsb = be32_to_cpu(btree[i + 1].before); | |
310 | xfs_trans_brelse(*trans, bp); | |
311 | } | |
312 | /* | |
313 | * Atomically commit the whole invalidate stuff. | |
314 | */ | |
315 | error = xfs_trans_roll(trans, dp); | |
316 | if (error) | |
317 | return error; | |
318 | } | |
319 | ||
320 | return 0; | |
321 | } | |
322 | ||
323 | /* | |
324 | * Indiscriminately delete the entire attribute fork | |
325 | * | |
326 | * Recurse (gasp!) through the attribute nodes until we find leaves. | |
327 | * We're doing a depth-first traversal in order to invalidate everything. | |
328 | */ | |
329 | int | |
330 | xfs_attr3_root_inactive( | |
331 | struct xfs_trans **trans, | |
332 | struct xfs_inode *dp) | |
333 | { | |
334 | struct xfs_da_blkinfo *info; | |
335 | struct xfs_buf *bp; | |
336 | xfs_daddr_t blkno; | |
337 | int error; | |
338 | ||
339 | /* | |
340 | * Read block 0 to see what we have to work with. | |
341 | * We only get here if we have extents, since we remove | |
342 | * the extents in reverse order the extent containing | |
343 | * block 0 must still be there. | |
344 | */ | |
345 | error = xfs_da3_node_read(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK); | |
346 | if (error) | |
347 | return error; | |
348 | blkno = bp->b_bn; | |
349 | ||
350 | /* | |
351 | * Invalidate the tree, even if the "tree" is only a single leaf block. | |
352 | * This is a depth-first traversal! | |
353 | */ | |
354 | info = bp->b_addr; | |
355 | switch (info->magic) { | |
356 | case cpu_to_be16(XFS_DA_NODE_MAGIC): | |
357 | case cpu_to_be16(XFS_DA3_NODE_MAGIC): | |
358 | error = xfs_attr3_node_inactive(trans, dp, bp, 1); | |
359 | break; | |
360 | case cpu_to_be16(XFS_ATTR_LEAF_MAGIC): | |
361 | case cpu_to_be16(XFS_ATTR3_LEAF_MAGIC): | |
362 | error = xfs_attr3_leaf_inactive(trans, dp, bp); | |
363 | break; | |
364 | default: | |
365 | error = XFS_ERROR(EIO); | |
366 | xfs_trans_brelse(*trans, bp); | |
367 | break; | |
368 | } | |
369 | if (error) | |
370 | return error; | |
371 | ||
372 | /* | |
373 | * Invalidate the incore copy of the root block. | |
374 | */ | |
375 | error = xfs_da_get_buf(*trans, dp, 0, blkno, &bp, XFS_ATTR_FORK); | |
376 | if (error) | |
377 | return error; | |
378 | xfs_trans_binval(*trans, bp); /* remove from cache */ | |
379 | /* | |
380 | * Commit the invalidate and start the next transaction. | |
381 | */ | |
382 | error = xfs_trans_roll(trans, dp); | |
383 | ||
384 | return error; | |
385 | } | |
386 | ||
387 | int | |
388 | xfs_attr_inactive(xfs_inode_t *dp) | |
389 | { | |
390 | xfs_trans_t *trans; | |
391 | xfs_mount_t *mp; | |
392 | int error; | |
393 | ||
394 | mp = dp->i_mount; | |
395 | ASSERT(! XFS_NOT_DQATTACHED(mp, dp)); | |
396 | ||
397 | xfs_ilock(dp, XFS_ILOCK_SHARED); | |
398 | if (!xfs_inode_hasattr(dp) || | |
399 | dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) { | |
400 | xfs_iunlock(dp, XFS_ILOCK_SHARED); | |
401 | return 0; | |
402 | } | |
403 | xfs_iunlock(dp, XFS_ILOCK_SHARED); | |
404 | ||
405 | /* | |
406 | * Start our first transaction of the day. | |
407 | * | |
408 | * All future transactions during this code must be "chained" off | |
409 | * this one via the trans_dup() call. All transactions will contain | |
410 | * the inode, and the inode will always be marked with trans_ihold(). | |
411 | * Since the inode will be locked in all transactions, we must log | |
412 | * the inode in every transaction to let it float upward through | |
413 | * the log. | |
414 | */ | |
415 | trans = xfs_trans_alloc(mp, XFS_TRANS_ATTRINVAL); | |
3d3c8b52 JL |
416 | error = xfs_trans_reserve(trans, &M_RES(mp)->tr_attrinval, 0, 0); |
417 | if (error) { | |
fde2227c DC |
418 | xfs_trans_cancel(trans, 0); |
419 | return(error); | |
420 | } | |
421 | xfs_ilock(dp, XFS_ILOCK_EXCL); | |
422 | ||
423 | /* | |
424 | * No need to make quota reservations here. We expect to release some | |
425 | * blocks, not allocate, in the common case. | |
426 | */ | |
427 | xfs_trans_ijoin(trans, dp, 0); | |
428 | ||
429 | /* | |
430 | * Decide on what work routines to call based on the inode size. | |
431 | */ | |
432 | if (!xfs_inode_hasattr(dp) || | |
433 | dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) { | |
434 | error = 0; | |
435 | goto out; | |
436 | } | |
437 | error = xfs_attr3_root_inactive(&trans, dp); | |
438 | if (error) | |
439 | goto out; | |
440 | ||
441 | error = xfs_itruncate_extents(&trans, dp, XFS_ATTR_FORK, 0); | |
442 | if (error) | |
443 | goto out; | |
444 | ||
445 | error = xfs_trans_commit(trans, XFS_TRANS_RELEASE_LOG_RES); | |
446 | xfs_iunlock(dp, XFS_ILOCK_EXCL); | |
447 | ||
448 | return(error); | |
449 | ||
450 | out: | |
451 | xfs_trans_cancel(trans, XFS_TRANS_RELEASE_LOG_RES|XFS_TRANS_ABORT); | |
452 | xfs_iunlock(dp, XFS_ILOCK_EXCL); | |
453 | return(error); | |
454 | } |