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" | |
70a9883c | 21 | #include "xfs_shared.h" |
239880ef DC |
22 | #include "xfs_format.h" |
23 | #include "xfs_log_format.h" | |
24 | #include "xfs_trans_resv.h" | |
fde2227c | 25 | #include "xfs_bit.h" |
fde2227c | 26 | #include "xfs_mount.h" |
57062787 | 27 | #include "xfs_da_format.h" |
fde2227c | 28 | #include "xfs_da_btree.h" |
a4fbe6ab | 29 | #include "xfs_inode.h" |
fde2227c | 30 | #include "xfs_alloc.h" |
fde2227c | 31 | #include "xfs_attr_remote.h" |
239880ef | 32 | #include "xfs_trans.h" |
fde2227c DC |
33 | #include "xfs_inode_item.h" |
34 | #include "xfs_bmap.h" | |
35 | #include "xfs_attr.h" | |
36 | #include "xfs_attr_leaf.h" | |
37 | #include "xfs_error.h" | |
38 | #include "xfs_quota.h" | |
39 | #include "xfs_trace.h" | |
4bceb18f | 40 | #include "xfs_dir2.h" |
fde2227c DC |
41 | |
42 | /* | |
43 | * Look at all the extents for this logical region, | |
44 | * invalidate any buffers that are incore/in transactions. | |
45 | */ | |
46 | STATIC int | |
47 | xfs_attr3_leaf_freextent( | |
48 | struct xfs_trans **trans, | |
49 | struct xfs_inode *dp, | |
50 | xfs_dablk_t blkno, | |
51 | int blkcnt) | |
52 | { | |
53 | struct xfs_bmbt_irec map; | |
54 | struct xfs_buf *bp; | |
55 | xfs_dablk_t tblkno; | |
56 | xfs_daddr_t dblkno; | |
57 | int tblkcnt; | |
58 | int dblkcnt; | |
59 | int nmap; | |
60 | int error; | |
61 | ||
62 | /* | |
63 | * Roll through the "value", invalidating the attribute value's | |
64 | * blocks. | |
65 | */ | |
66 | tblkno = blkno; | |
67 | tblkcnt = blkcnt; | |
68 | while (tblkcnt > 0) { | |
69 | /* | |
70 | * Try to remember where we decided to put the value. | |
71 | */ | |
72 | nmap = 1; | |
73 | error = xfs_bmapi_read(dp, (xfs_fileoff_t)tblkno, tblkcnt, | |
74 | &map, &nmap, XFS_BMAPI_ATTRFORK); | |
75 | if (error) { | |
d99831ff | 76 | return error; |
fde2227c DC |
77 | } |
78 | ASSERT(nmap == 1); | |
79 | ASSERT(map.br_startblock != DELAYSTARTBLOCK); | |
80 | ||
81 | /* | |
82 | * If it's a hole, these are already unmapped | |
83 | * so there's nothing to invalidate. | |
84 | */ | |
85 | if (map.br_startblock != HOLESTARTBLOCK) { | |
86 | ||
87 | dblkno = XFS_FSB_TO_DADDR(dp->i_mount, | |
88 | map.br_startblock); | |
89 | dblkcnt = XFS_FSB_TO_BB(dp->i_mount, | |
90 | map.br_blockcount); | |
91 | bp = xfs_trans_get_buf(*trans, | |
92 | dp->i_mount->m_ddev_targp, | |
93 | dblkno, dblkcnt, 0); | |
94 | if (!bp) | |
2451337d | 95 | return -ENOMEM; |
fde2227c DC |
96 | xfs_trans_binval(*trans, bp); |
97 | /* | |
98 | * Roll to next transaction. | |
99 | */ | |
100 | error = xfs_trans_roll(trans, dp); | |
101 | if (error) | |
d99831ff | 102 | return error; |
fde2227c DC |
103 | } |
104 | ||
105 | tblkno += map.br_blockcount; | |
106 | tblkcnt -= map.br_blockcount; | |
107 | } | |
108 | ||
d99831ff | 109 | return 0; |
fde2227c DC |
110 | } |
111 | ||
112 | /* | |
113 | * Invalidate all of the "remote" value regions pointed to by a particular | |
114 | * leaf block. | |
115 | * Note that we must release the lock on the buffer so that we are not | |
116 | * caught holding something that the logging code wants to flush to disk. | |
117 | */ | |
118 | STATIC int | |
119 | xfs_attr3_leaf_inactive( | |
120 | struct xfs_trans **trans, | |
121 | struct xfs_inode *dp, | |
122 | struct xfs_buf *bp) | |
123 | { | |
124 | struct xfs_attr_leafblock *leaf; | |
125 | struct xfs_attr3_icleaf_hdr ichdr; | |
126 | struct xfs_attr_leaf_entry *entry; | |
127 | struct xfs_attr_leaf_name_remote *name_rmt; | |
128 | struct xfs_attr_inactive_list *list; | |
129 | struct xfs_attr_inactive_list *lp; | |
130 | int error; | |
131 | int count; | |
132 | int size; | |
133 | int tmp; | |
134 | int i; | |
2f661241 | 135 | struct xfs_mount *mp = bp->b_target->bt_mount; |
fde2227c DC |
136 | |
137 | leaf = bp->b_addr; | |
2f661241 | 138 | xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, leaf); |
fde2227c DC |
139 | |
140 | /* | |
141 | * Count the number of "remote" value extents. | |
142 | */ | |
143 | count = 0; | |
144 | entry = xfs_attr3_leaf_entryp(leaf); | |
145 | for (i = 0; i < ichdr.count; entry++, i++) { | |
146 | if (be16_to_cpu(entry->nameidx) && | |
147 | ((entry->flags & XFS_ATTR_LOCAL) == 0)) { | |
148 | name_rmt = xfs_attr3_leaf_name_remote(leaf, i); | |
149 | if (name_rmt->valueblk) | |
150 | count++; | |
151 | } | |
152 | } | |
153 | ||
154 | /* | |
155 | * If there are no "remote" values, we're done. | |
156 | */ | |
157 | if (count == 0) { | |
158 | xfs_trans_brelse(*trans, bp); | |
159 | return 0; | |
160 | } | |
161 | ||
162 | /* | |
163 | * Allocate storage for a list of all the "remote" value extents. | |
164 | */ | |
165 | size = count * sizeof(xfs_attr_inactive_list_t); | |
166 | list = kmem_alloc(size, KM_SLEEP); | |
167 | ||
168 | /* | |
169 | * Identify each of the "remote" value extents. | |
170 | */ | |
171 | lp = list; | |
172 | entry = xfs_attr3_leaf_entryp(leaf); | |
173 | for (i = 0; i < ichdr.count; entry++, i++) { | |
174 | if (be16_to_cpu(entry->nameidx) && | |
175 | ((entry->flags & XFS_ATTR_LOCAL) == 0)) { | |
176 | name_rmt = xfs_attr3_leaf_name_remote(leaf, i); | |
177 | if (name_rmt->valueblk) { | |
178 | lp->valueblk = be32_to_cpu(name_rmt->valueblk); | |
179 | lp->valuelen = xfs_attr3_rmt_blocks(dp->i_mount, | |
180 | be32_to_cpu(name_rmt->valuelen)); | |
181 | lp++; | |
182 | } | |
183 | } | |
184 | } | |
185 | xfs_trans_brelse(*trans, bp); /* unlock for trans. in freextent() */ | |
186 | ||
187 | /* | |
188 | * Invalidate each of the "remote" value extents. | |
189 | */ | |
190 | error = 0; | |
191 | for (lp = list, i = 0; i < count; i++, lp++) { | |
192 | tmp = xfs_attr3_leaf_freextent(trans, dp, | |
193 | lp->valueblk, lp->valuelen); | |
194 | ||
195 | if (error == 0) | |
196 | error = tmp; /* save only the 1st errno */ | |
197 | } | |
198 | ||
199 | kmem_free(list); | |
200 | return error; | |
201 | } | |
202 | ||
203 | /* | |
204 | * Recurse (gasp!) through the attribute nodes until we find leaves. | |
205 | * We're doing a depth-first traversal in order to invalidate everything. | |
206 | */ | |
207 | STATIC int | |
208 | xfs_attr3_node_inactive( | |
209 | struct xfs_trans **trans, | |
210 | struct xfs_inode *dp, | |
211 | struct xfs_buf *bp, | |
212 | int level) | |
213 | { | |
214 | xfs_da_blkinfo_t *info; | |
215 | xfs_da_intnode_t *node; | |
216 | xfs_dablk_t child_fsb; | |
217 | xfs_daddr_t parent_blkno, child_blkno; | |
218 | int error, i; | |
219 | struct xfs_buf *child_bp; | |
220 | struct xfs_da_node_entry *btree; | |
221 | struct xfs_da3_icnode_hdr ichdr; | |
222 | ||
223 | /* | |
224 | * Since this code is recursive (gasp!) we must protect ourselves. | |
225 | */ | |
226 | if (level > XFS_DA_NODE_MAXDEPTH) { | |
227 | xfs_trans_brelse(*trans, bp); /* no locks for later trans */ | |
2451337d | 228 | return -EIO; |
fde2227c DC |
229 | } |
230 | ||
231 | node = bp->b_addr; | |
01ba43b8 | 232 | dp->d_ops->node_hdr_from_disk(&ichdr, node); |
fde2227c DC |
233 | parent_blkno = bp->b_bn; |
234 | if (!ichdr.count) { | |
235 | xfs_trans_brelse(*trans, bp); | |
236 | return 0; | |
237 | } | |
4bceb18f | 238 | btree = dp->d_ops->node_tree_p(node); |
fde2227c DC |
239 | child_fsb = be32_to_cpu(btree[0].before); |
240 | xfs_trans_brelse(*trans, bp); /* no locks for later trans */ | |
241 | ||
242 | /* | |
243 | * If this is the node level just above the leaves, simply loop | |
244 | * over the leaves removing all of them. If this is higher up | |
245 | * in the tree, recurse downward. | |
246 | */ | |
247 | for (i = 0; i < ichdr.count; i++) { | |
248 | /* | |
249 | * Read the subsidiary block to see what we have to work with. | |
250 | * Don't do this in a transaction. This is a depth-first | |
251 | * traversal of the tree so we may deal with many blocks | |
252 | * before we come back to this one. | |
253 | */ | |
254 | error = xfs_da3_node_read(*trans, dp, child_fsb, -2, &child_bp, | |
255 | XFS_ATTR_FORK); | |
256 | if (error) | |
d99831ff | 257 | return error; |
fde2227c DC |
258 | if (child_bp) { |
259 | /* save for re-read later */ | |
260 | child_blkno = XFS_BUF_ADDR(child_bp); | |
261 | ||
262 | /* | |
263 | * Invalidate the subtree, however we have to. | |
264 | */ | |
265 | info = child_bp->b_addr; | |
266 | switch (info->magic) { | |
267 | case cpu_to_be16(XFS_DA_NODE_MAGIC): | |
268 | case cpu_to_be16(XFS_DA3_NODE_MAGIC): | |
269 | error = xfs_attr3_node_inactive(trans, dp, | |
270 | child_bp, level + 1); | |
271 | break; | |
272 | case cpu_to_be16(XFS_ATTR_LEAF_MAGIC): | |
273 | case cpu_to_be16(XFS_ATTR3_LEAF_MAGIC): | |
274 | error = xfs_attr3_leaf_inactive(trans, dp, | |
275 | child_bp); | |
276 | break; | |
277 | default: | |
2451337d | 278 | error = -EIO; |
fde2227c DC |
279 | xfs_trans_brelse(*trans, child_bp); |
280 | break; | |
281 | } | |
282 | if (error) | |
283 | return error; | |
284 | ||
285 | /* | |
286 | * Remove the subsidiary block from the cache | |
287 | * and from the log. | |
288 | */ | |
289 | error = xfs_da_get_buf(*trans, dp, 0, child_blkno, | |
290 | &child_bp, XFS_ATTR_FORK); | |
291 | if (error) | |
292 | return error; | |
293 | xfs_trans_binval(*trans, child_bp); | |
294 | } | |
295 | ||
296 | /* | |
297 | * If we're not done, re-read the parent to get the next | |
298 | * child block number. | |
299 | */ | |
300 | if (i + 1 < ichdr.count) { | |
301 | error = xfs_da3_node_read(*trans, dp, 0, parent_blkno, | |
302 | &bp, XFS_ATTR_FORK); | |
303 | if (error) | |
304 | return error; | |
305 | child_fsb = be32_to_cpu(btree[i + 1].before); | |
306 | xfs_trans_brelse(*trans, bp); | |
307 | } | |
308 | /* | |
309 | * Atomically commit the whole invalidate stuff. | |
310 | */ | |
311 | error = xfs_trans_roll(trans, dp); | |
312 | if (error) | |
313 | return error; | |
314 | } | |
315 | ||
316 | return 0; | |
317 | } | |
318 | ||
319 | /* | |
320 | * Indiscriminately delete the entire attribute fork | |
321 | * | |
322 | * Recurse (gasp!) through the attribute nodes until we find leaves. | |
323 | * We're doing a depth-first traversal in order to invalidate everything. | |
324 | */ | |
325 | int | |
326 | xfs_attr3_root_inactive( | |
327 | struct xfs_trans **trans, | |
328 | struct xfs_inode *dp) | |
329 | { | |
330 | struct xfs_da_blkinfo *info; | |
331 | struct xfs_buf *bp; | |
332 | xfs_daddr_t blkno; | |
333 | int error; | |
334 | ||
335 | /* | |
336 | * Read block 0 to see what we have to work with. | |
337 | * We only get here if we have extents, since we remove | |
338 | * the extents in reverse order the extent containing | |
339 | * block 0 must still be there. | |
340 | */ | |
341 | error = xfs_da3_node_read(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK); | |
342 | if (error) | |
343 | return error; | |
344 | blkno = bp->b_bn; | |
345 | ||
346 | /* | |
347 | * Invalidate the tree, even if the "tree" is only a single leaf block. | |
348 | * This is a depth-first traversal! | |
349 | */ | |
350 | info = bp->b_addr; | |
351 | switch (info->magic) { | |
352 | case cpu_to_be16(XFS_DA_NODE_MAGIC): | |
353 | case cpu_to_be16(XFS_DA3_NODE_MAGIC): | |
354 | error = xfs_attr3_node_inactive(trans, dp, bp, 1); | |
355 | break; | |
356 | case cpu_to_be16(XFS_ATTR_LEAF_MAGIC): | |
357 | case cpu_to_be16(XFS_ATTR3_LEAF_MAGIC): | |
358 | error = xfs_attr3_leaf_inactive(trans, dp, bp); | |
359 | break; | |
360 | default: | |
2451337d | 361 | error = -EIO; |
fde2227c DC |
362 | xfs_trans_brelse(*trans, bp); |
363 | break; | |
364 | } | |
365 | if (error) | |
366 | return error; | |
367 | ||
368 | /* | |
369 | * Invalidate the incore copy of the root block. | |
370 | */ | |
371 | error = xfs_da_get_buf(*trans, dp, 0, blkno, &bp, XFS_ATTR_FORK); | |
372 | if (error) | |
373 | return error; | |
374 | xfs_trans_binval(*trans, bp); /* remove from cache */ | |
375 | /* | |
376 | * Commit the invalidate and start the next transaction. | |
377 | */ | |
378 | error = xfs_trans_roll(trans, dp); | |
379 | ||
380 | return error; | |
381 | } | |
382 | ||
383 | int | |
384 | xfs_attr_inactive(xfs_inode_t *dp) | |
385 | { | |
386 | xfs_trans_t *trans; | |
387 | xfs_mount_t *mp; | |
388 | int error; | |
389 | ||
390 | mp = dp->i_mount; | |
391 | ASSERT(! XFS_NOT_DQATTACHED(mp, dp)); | |
392 | ||
393 | xfs_ilock(dp, XFS_ILOCK_SHARED); | |
394 | if (!xfs_inode_hasattr(dp) || | |
395 | dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) { | |
396 | xfs_iunlock(dp, XFS_ILOCK_SHARED); | |
397 | return 0; | |
398 | } | |
399 | xfs_iunlock(dp, XFS_ILOCK_SHARED); | |
400 | ||
401 | /* | |
402 | * Start our first transaction of the day. | |
403 | * | |
404 | * All future transactions during this code must be "chained" off | |
405 | * this one via the trans_dup() call. All transactions will contain | |
406 | * the inode, and the inode will always be marked with trans_ihold(). | |
407 | * Since the inode will be locked in all transactions, we must log | |
408 | * the inode in every transaction to let it float upward through | |
409 | * the log. | |
410 | */ | |
411 | trans = xfs_trans_alloc(mp, XFS_TRANS_ATTRINVAL); | |
3d3c8b52 JL |
412 | error = xfs_trans_reserve(trans, &M_RES(mp)->tr_attrinval, 0, 0); |
413 | if (error) { | |
fde2227c | 414 | xfs_trans_cancel(trans, 0); |
d99831ff | 415 | return error; |
fde2227c DC |
416 | } |
417 | xfs_ilock(dp, XFS_ILOCK_EXCL); | |
418 | ||
419 | /* | |
420 | * No need to make quota reservations here. We expect to release some | |
421 | * blocks, not allocate, in the common case. | |
422 | */ | |
423 | xfs_trans_ijoin(trans, dp, 0); | |
424 | ||
425 | /* | |
426 | * Decide on what work routines to call based on the inode size. | |
427 | */ | |
428 | if (!xfs_inode_hasattr(dp) || | |
429 | dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) { | |
430 | error = 0; | |
431 | goto out; | |
432 | } | |
433 | error = xfs_attr3_root_inactive(&trans, dp); | |
434 | if (error) | |
435 | goto out; | |
436 | ||
437 | error = xfs_itruncate_extents(&trans, dp, XFS_ATTR_FORK, 0); | |
438 | if (error) | |
439 | goto out; | |
440 | ||
441 | error = xfs_trans_commit(trans, XFS_TRANS_RELEASE_LOG_RES); | |
442 | xfs_iunlock(dp, XFS_ILOCK_EXCL); | |
443 | ||
d99831ff | 444 | return error; |
fde2227c DC |
445 | |
446 | out: | |
447 | xfs_trans_cancel(trans, XFS_TRANS_RELEASE_LOG_RES|XFS_TRANS_ABORT); | |
448 | xfs_iunlock(dp, XFS_ILOCK_EXCL); | |
d99831ff | 449 | return error; |
fde2227c | 450 | } |