1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
4 * Copyright (c) 2013 Red Hat, Inc.
5 * All Rights Reserved.
6 */
7 #include "xfs.h"
8 #include "xfs_fs.h"
9 #include "xfs_shared.h"
10 #include "xfs_format.h"
11 #include "xfs_log_format.h"
12 #include "xfs_trans_resv.h"
13 #include "xfs_bit.h"
14 #include "xfs_mount.h"
15 #include "xfs_da_format.h"
16 #include "xfs_da_btree.h"
17 #include "xfs_inode.h"
18 #include "xfs_attr.h"
19 #include "xfs_attr_remote.h"
20 #include "xfs_trans.h"
21 #include "xfs_bmap.h"
22 #include "xfs_attr_leaf.h"
23 #include "xfs_quota.h"
24 #include "xfs_dir2.h"
25 #include "xfs_error.h"
26
27 /*
28 * Invalidate any incore buffers associated with this remote attribute value
29 * extent. We never log remote attribute value buffers, which means that they
30 * won't be attached to a transaction and are therefore safe to mark stale.
31 * The actual bunmapi will be taken care of later.
32 */
33 STATIC int
xfs_attr3_rmt_stale(struct xfs_inode * dp,xfs_dablk_t blkno,int blkcnt)34 xfs_attr3_rmt_stale(
35 struct xfs_inode *dp,
36 xfs_dablk_t blkno,
37 int blkcnt)
38 {
39 struct xfs_bmbt_irec map;
40 int nmap;
41 int error;
42
43 /*
44 * Roll through the "value", invalidating the attribute value's
45 * blocks.
46 */
47 while (blkcnt > 0) {
48 /*
49 * Try to remember where we decided to put the value.
50 */
51 nmap = 1;
52 error = xfs_bmapi_read(dp, (xfs_fileoff_t)blkno, blkcnt,
53 &map, &nmap, XFS_BMAPI_ATTRFORK);
54 if (error)
55 return error;
56 if (XFS_IS_CORRUPT(dp->i_mount, nmap != 1))
57 return -EFSCORRUPTED;
58
59 /*
60 * Mark any incore buffers for the remote value as stale. We
61 * never log remote attr value buffers, so the buffer should be
62 * easy to kill.
63 */
64 error = xfs_attr_rmtval_stale(dp, &map, 0);
65 if (error)
66 return error;
67
68 blkno += map.br_blockcount;
69 blkcnt -= map.br_blockcount;
70 }
71
72 return 0;
73 }
74
75 /*
76 * Invalidate all of the "remote" value regions pointed to by a particular
77 * leaf block.
78 * Note that we must release the lock on the buffer so that we are not
79 * caught holding something that the logging code wants to flush to disk.
80 */
81 STATIC int
xfs_attr3_leaf_inactive(struct xfs_trans ** trans,struct xfs_inode * dp,struct xfs_buf * bp)82 xfs_attr3_leaf_inactive(
83 struct xfs_trans **trans,
84 struct xfs_inode *dp,
85 struct xfs_buf *bp)
86 {
87 struct xfs_attr3_icleaf_hdr ichdr;
88 struct xfs_mount *mp = bp->b_mount;
89 struct xfs_attr_leafblock *leaf = bp->b_addr;
90 struct xfs_attr_leaf_entry *entry;
91 struct xfs_attr_leaf_name_remote *name_rmt;
92 int error = 0;
93 int i;
94
95 xfs_attr3_leaf_hdr_from_disk(mp->m_attr_geo, &ichdr, leaf);
96
97 /*
98 * Find the remote value extents for this leaf and invalidate their
99 * incore buffers.
100 */
101 entry = xfs_attr3_leaf_entryp(leaf);
102 for (i = 0; i < ichdr.count; entry++, i++) {
103 int blkcnt;
104
105 if (!entry->nameidx || (entry->flags & XFS_ATTR_LOCAL))
106 continue;
107
108 name_rmt = xfs_attr3_leaf_name_remote(leaf, i);
109 if (!name_rmt->valueblk)
110 continue;
111
112 blkcnt = xfs_attr3_rmt_blocks(dp->i_mount,
113 be32_to_cpu(name_rmt->valuelen));
114 error = xfs_attr3_rmt_stale(dp,
115 be32_to_cpu(name_rmt->valueblk), blkcnt);
116 if (error)
117 goto err;
118 }
119
120 xfs_trans_brelse(*trans, bp);
121 err:
122 return error;
123 }
124
125 /*
126 * Recurse (gasp!) through the attribute nodes until we find leaves.
127 * We're doing a depth-first traversal in order to invalidate everything.
128 */
129 STATIC int
xfs_attr3_node_inactive(struct xfs_trans ** trans,struct xfs_inode * dp,struct xfs_buf * bp,int level)130 xfs_attr3_node_inactive(
131 struct xfs_trans **trans,
132 struct xfs_inode *dp,
133 struct xfs_buf *bp,
134 int level)
135 {
136 struct xfs_mount *mp = dp->i_mount;
137 struct xfs_da_blkinfo *info;
138 xfs_dablk_t child_fsb;
139 xfs_daddr_t parent_blkno, child_blkno;
140 struct xfs_buf *child_bp;
141 struct xfs_da3_icnode_hdr ichdr;
142 int error, i;
143
144 /*
145 * Since this code is recursive (gasp!) we must protect ourselves.
146 */
147 if (level > XFS_DA_NODE_MAXDEPTH) {
148 xfs_buf_mark_corrupt(bp);
149 xfs_trans_brelse(*trans, bp); /* no locks for later trans */
150 return -EFSCORRUPTED;
151 }
152
153 xfs_da3_node_hdr_from_disk(dp->i_mount, &ichdr, bp->b_addr);
154 parent_blkno = xfs_buf_daddr(bp);
155 if (!ichdr.count) {
156 xfs_trans_brelse(*trans, bp);
157 return 0;
158 }
159 child_fsb = be32_to_cpu(ichdr.btree[0].before);
160 xfs_trans_brelse(*trans, bp); /* no locks for later trans */
161 bp = NULL;
162
163 /*
164 * If this is the node level just above the leaves, simply loop
165 * over the leaves removing all of them. If this is higher up
166 * in the tree, recurse downward.
167 */
168 for (i = 0; i < ichdr.count; i++) {
169 /*
170 * Read the subsidiary block to see what we have to work with.
171 * Don't do this in a transaction. This is a depth-first
172 * traversal of the tree so we may deal with many blocks
173 * before we come back to this one.
174 */
175 error = xfs_da3_node_read(*trans, dp, child_fsb, &child_bp,
176 XFS_ATTR_FORK);
177 if (error)
178 return error;
179
180 /* save for re-read later */
181 child_blkno = xfs_buf_daddr(child_bp);
182
183 /*
184 * Invalidate the subtree, however we have to.
185 */
186 info = child_bp->b_addr;
187 switch (info->magic) {
188 case cpu_to_be16(XFS_DA_NODE_MAGIC):
189 case cpu_to_be16(XFS_DA3_NODE_MAGIC):
190 error = xfs_attr3_node_inactive(trans, dp, child_bp,
191 level + 1);
192 break;
193 case cpu_to_be16(XFS_ATTR_LEAF_MAGIC):
194 case cpu_to_be16(XFS_ATTR3_LEAF_MAGIC):
195 error = xfs_attr3_leaf_inactive(trans, dp, child_bp);
196 break;
197 default:
198 xfs_buf_mark_corrupt(child_bp);
199 xfs_trans_brelse(*trans, child_bp);
200 error = -EFSCORRUPTED;
201 break;
202 }
203 if (error)
204 return error;
205
206 /*
207 * Remove the subsidiary block from the cache and from the log.
208 */
209 error = xfs_trans_get_buf(*trans, mp->m_ddev_targp,
210 child_blkno,
211 XFS_FSB_TO_BB(mp, mp->m_attr_geo->fsbcount), 0,
212 &child_bp);
213 if (error)
214 return error;
215 xfs_trans_binval(*trans, child_bp);
216 child_bp = NULL;
217
218 /*
219 * If we're not done, re-read the parent to get the next
220 * child block number.
221 */
222 if (i + 1 < ichdr.count) {
223 struct xfs_da3_icnode_hdr phdr;
224
225 error = xfs_da3_node_read_mapped(*trans, dp,
226 parent_blkno, &bp, XFS_ATTR_FORK);
227 if (error)
228 return error;
229 xfs_da3_node_hdr_from_disk(dp->i_mount, &phdr,
230 bp->b_addr);
231 child_fsb = be32_to_cpu(phdr.btree[i + 1].before);
232 xfs_trans_brelse(*trans, bp);
233 bp = NULL;
234 }
235 /*
236 * Atomically commit the whole invalidate stuff.
237 */
238 error = xfs_trans_roll_inode(trans, dp);
239 if (error)
240 return error;
241 }
242
243 return 0;
244 }
245
246 /*
247 * Indiscriminately delete the entire attribute fork
248 *
249 * Recurse (gasp!) through the attribute nodes until we find leaves.
250 * We're doing a depth-first traversal in order to invalidate everything.
251 */
252 static int
xfs_attr3_root_inactive(struct xfs_trans ** trans,struct xfs_inode * dp)253 xfs_attr3_root_inactive(
254 struct xfs_trans **trans,
255 struct xfs_inode *dp)
256 {
257 struct xfs_mount *mp = dp->i_mount;
258 struct xfs_da_blkinfo *info;
259 struct xfs_buf *bp;
260 xfs_daddr_t blkno;
261 int error;
262
263 /*
264 * Read block 0 to see what we have to work with.
265 * We only get here if we have extents, since we remove
266 * the extents in reverse order the extent containing
267 * block 0 must still be there.
268 */
269 error = xfs_da3_node_read(*trans, dp, 0, &bp, XFS_ATTR_FORK);
270 if (error)
271 return error;
272 blkno = xfs_buf_daddr(bp);
273
274 /*
275 * Invalidate the tree, even if the "tree" is only a single leaf block.
276 * This is a depth-first traversal!
277 */
278 info = bp->b_addr;
279 switch (info->magic) {
280 case cpu_to_be16(XFS_DA_NODE_MAGIC):
281 case cpu_to_be16(XFS_DA3_NODE_MAGIC):
282 error = xfs_attr3_node_inactive(trans, dp, bp, 1);
283 break;
284 case cpu_to_be16(XFS_ATTR_LEAF_MAGIC):
285 case cpu_to_be16(XFS_ATTR3_LEAF_MAGIC):
286 error = xfs_attr3_leaf_inactive(trans, dp, bp);
287 break;
288 default:
289 error = -EFSCORRUPTED;
290 xfs_buf_mark_corrupt(bp);
291 xfs_trans_brelse(*trans, bp);
292 break;
293 }
294 if (error)
295 return error;
296
297 /*
298 * Invalidate the incore copy of the root block.
299 */
300 error = xfs_trans_get_buf(*trans, mp->m_ddev_targp, blkno,
301 XFS_FSB_TO_BB(mp, mp->m_attr_geo->fsbcount), 0, &bp);
302 if (error)
303 return error;
304 error = bp->b_error;
305 if (error) {
306 xfs_trans_brelse(*trans, bp);
307 return error;
308 }
309 xfs_trans_binval(*trans, bp); /* remove from cache */
310 /*
311 * Commit the invalidate and start the next transaction.
312 */
313 error = xfs_trans_roll_inode(trans, dp);
314
315 return error;
316 }
317
318 /*
319 * xfs_attr_inactive kills all traces of an attribute fork on an inode. It
320 * removes both the on-disk and in-memory inode fork. Note that this also has to
321 * handle the condition of inodes without attributes but with an attribute fork
322 * configured, so we can't use xfs_inode_hasattr() here.
323 *
324 * The in-memory attribute fork is removed even on error.
325 */
326 int
xfs_attr_inactive(struct xfs_inode * dp)327 xfs_attr_inactive(
328 struct xfs_inode *dp)
329 {
330 struct xfs_trans *trans;
331 struct xfs_mount *mp;
332 int lock_mode = XFS_ILOCK_SHARED;
333 int error = 0;
334
335 mp = dp->i_mount;
336 ASSERT(! XFS_NOT_DQATTACHED(mp, dp));
337
338 xfs_ilock(dp, lock_mode);
339 if (!xfs_inode_has_attr_fork(dp))
340 goto out_destroy_fork;
341 xfs_iunlock(dp, lock_mode);
342
343 lock_mode = 0;
344
345 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_attrinval, 0, 0, 0, &trans);
346 if (error)
347 goto out_destroy_fork;
348
349 lock_mode = XFS_ILOCK_EXCL;
350 xfs_ilock(dp, lock_mode);
351
352 if (!xfs_inode_has_attr_fork(dp))
353 goto out_cancel;
354
355 /*
356 * No need to make quota reservations here. We expect to release some
357 * blocks, not allocate, in the common case.
358 */
359 xfs_trans_ijoin(trans, dp, 0);
360
361 /*
362 * Invalidate and truncate the attribute fork extents. Make sure the
363 * fork actually has xattr blocks as otherwise the invalidation has no
364 * blocks to read and returns an error. In this case, just do the fork
365 * removal below.
366 */
367 if (dp->i_af.if_nextents > 0) {
368 error = xfs_attr3_root_inactive(&trans, dp);
369 if (error)
370 goto out_cancel;
371
372 error = xfs_itruncate_extents(&trans, dp, XFS_ATTR_FORK, 0);
373 if (error)
374 goto out_cancel;
375 }
376
377 /* Reset the attribute fork - this also destroys the in-core fork */
378 xfs_attr_fork_remove(dp, trans);
379
380 error = xfs_trans_commit(trans);
381 xfs_iunlock(dp, lock_mode);
382 return error;
383
384 out_cancel:
385 xfs_trans_cancel(trans);
386 out_destroy_fork:
387 /* kill the in-core attr fork before we drop the inode lock */
388 xfs_ifork_zap_attr(dp);
389 if (lock_mode)
390 xfs_iunlock(dp, lock_mode);
391 return error;
392 }
393