2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * Copyright (c) 2013 Red Hat, Inc.
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.
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.
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
21 #include "xfs_format.h"
24 #include "xfs_trans.h"
27 #include "xfs_mount.h"
28 #include "xfs_da_btree.h"
29 #include "xfs_bmap_btree.h"
30 #include "xfs_alloc_btree.h"
31 #include "xfs_ialloc_btree.h"
32 #include "xfs_alloc.h"
33 #include "xfs_btree.h"
34 #include "xfs_attr_remote.h"
35 #include "xfs_dinode.h"
36 #include "xfs_inode.h"
37 #include "xfs_inode_item.h"
40 #include "xfs_attr_leaf.h"
41 #include "xfs_error.h"
42 #include "xfs_quota.h"
43 #include "xfs_trace.h"
44 #include "xfs_trans_priv.h"
47 * Look at all the extents for this logical region,
48 * invalidate any buffers that are incore/in transactions.
51 xfs_attr3_leaf_freextent(
52 struct xfs_trans **trans,
57 struct xfs_bmbt_irec map;
67 * Roll through the "value", invalidating the attribute value's
74 * Try to remember where we decided to put the value.
77 error = xfs_bmapi_read(dp, (xfs_fileoff_t)tblkno, tblkcnt,
78 &map, &nmap, XFS_BMAPI_ATTRFORK);
83 ASSERT(map.br_startblock != DELAYSTARTBLOCK);
86 * If it's a hole, these are already unmapped
87 * so there's nothing to invalidate.
89 if (map.br_startblock != HOLESTARTBLOCK) {
91 dblkno = XFS_FSB_TO_DADDR(dp->i_mount,
93 dblkcnt = XFS_FSB_TO_BB(dp->i_mount,
95 bp = xfs_trans_get_buf(*trans,
96 dp->i_mount->m_ddev_targp,
100 xfs_trans_binval(*trans, bp);
102 * Roll to next transaction.
104 error = xfs_trans_roll(trans, dp);
109 tblkno += map.br_blockcount;
110 tblkcnt -= map.br_blockcount;
117 * Invalidate all of the "remote" value regions pointed to by a particular
119 * Note that we must release the lock on the buffer so that we are not
120 * caught holding something that the logging code wants to flush to disk.
123 xfs_attr3_leaf_inactive(
124 struct xfs_trans **trans,
125 struct xfs_inode *dp,
128 struct xfs_attr_leafblock *leaf;
129 struct xfs_attr3_icleaf_hdr ichdr;
130 struct xfs_attr_leaf_entry *entry;
131 struct xfs_attr_leaf_name_remote *name_rmt;
132 struct xfs_attr_inactive_list *list;
133 struct xfs_attr_inactive_list *lp;
141 xfs_attr3_leaf_hdr_from_disk(&ichdr, leaf);
144 * Count the number of "remote" value extents.
147 entry = xfs_attr3_leaf_entryp(leaf);
148 for (i = 0; i < ichdr.count; entry++, i++) {
149 if (be16_to_cpu(entry->nameidx) &&
150 ((entry->flags & XFS_ATTR_LOCAL) == 0)) {
151 name_rmt = xfs_attr3_leaf_name_remote(leaf, i);
152 if (name_rmt->valueblk)
158 * If there are no "remote" values, we're done.
161 xfs_trans_brelse(*trans, bp);
166 * Allocate storage for a list of all the "remote" value extents.
168 size = count * sizeof(xfs_attr_inactive_list_t);
169 list = kmem_alloc(size, KM_SLEEP);
172 * Identify each of the "remote" value extents.
175 entry = xfs_attr3_leaf_entryp(leaf);
176 for (i = 0; i < ichdr.count; entry++, i++) {
177 if (be16_to_cpu(entry->nameidx) &&
178 ((entry->flags & XFS_ATTR_LOCAL) == 0)) {
179 name_rmt = xfs_attr3_leaf_name_remote(leaf, i);
180 if (name_rmt->valueblk) {
181 lp->valueblk = be32_to_cpu(name_rmt->valueblk);
182 lp->valuelen = xfs_attr3_rmt_blocks(dp->i_mount,
183 be32_to_cpu(name_rmt->valuelen));
188 xfs_trans_brelse(*trans, bp); /* unlock for trans. in freextent() */
191 * Invalidate each of the "remote" value extents.
194 for (lp = list, i = 0; i < count; i++, lp++) {
195 tmp = xfs_attr3_leaf_freextent(trans, dp,
196 lp->valueblk, lp->valuelen);
199 error = tmp; /* save only the 1st errno */
207 * Recurse (gasp!) through the attribute nodes until we find leaves.
208 * We're doing a depth-first traversal in order to invalidate everything.
211 xfs_attr3_node_inactive(
212 struct xfs_trans **trans,
213 struct xfs_inode *dp,
217 xfs_da_blkinfo_t *info;
218 xfs_da_intnode_t *node;
219 xfs_dablk_t child_fsb;
220 xfs_daddr_t parent_blkno, child_blkno;
222 struct xfs_buf *child_bp;
223 struct xfs_da_node_entry *btree;
224 struct xfs_da3_icnode_hdr ichdr;
227 * Since this code is recursive (gasp!) we must protect ourselves.
229 if (level > XFS_DA_NODE_MAXDEPTH) {
230 xfs_trans_brelse(*trans, bp); /* no locks for later trans */
231 return XFS_ERROR(EIO);
235 xfs_da3_node_hdr_from_disk(&ichdr, node);
236 parent_blkno = bp->b_bn;
238 xfs_trans_brelse(*trans, bp);
241 btree = xfs_da3_node_tree_p(node);
242 child_fsb = be32_to_cpu(btree[0].before);
243 xfs_trans_brelse(*trans, bp); /* no locks for later trans */
246 * If this is the node level just above the leaves, simply loop
247 * over the leaves removing all of them. If this is higher up
248 * in the tree, recurse downward.
250 for (i = 0; i < ichdr.count; i++) {
252 * Read the subsidiary block to see what we have to work with.
253 * Don't do this in a transaction. This is a depth-first
254 * traversal of the tree so we may deal with many blocks
255 * before we come back to this one.
257 error = xfs_da3_node_read(*trans, dp, child_fsb, -2, &child_bp,
262 /* save for re-read later */
263 child_blkno = XFS_BUF_ADDR(child_bp);
266 * Invalidate the subtree, however we have to.
268 info = child_bp->b_addr;
269 switch (info->magic) {
270 case cpu_to_be16(XFS_DA_NODE_MAGIC):
271 case cpu_to_be16(XFS_DA3_NODE_MAGIC):
272 error = xfs_attr3_node_inactive(trans, dp,
273 child_bp, level + 1);
275 case cpu_to_be16(XFS_ATTR_LEAF_MAGIC):
276 case cpu_to_be16(XFS_ATTR3_LEAF_MAGIC):
277 error = xfs_attr3_leaf_inactive(trans, dp,
281 error = XFS_ERROR(EIO);
282 xfs_trans_brelse(*trans, child_bp);
289 * Remove the subsidiary block from the cache
292 error = xfs_da_get_buf(*trans, dp, 0, child_blkno,
293 &child_bp, XFS_ATTR_FORK);
296 xfs_trans_binval(*trans, child_bp);
300 * If we're not done, re-read the parent to get the next
301 * child block number.
303 if (i + 1 < ichdr.count) {
304 error = xfs_da3_node_read(*trans, dp, 0, parent_blkno,
308 child_fsb = be32_to_cpu(btree[i + 1].before);
309 xfs_trans_brelse(*trans, bp);
312 * Atomically commit the whole invalidate stuff.
314 error = xfs_trans_roll(trans, dp);
323 * Indiscriminately delete the entire attribute fork
325 * Recurse (gasp!) through the attribute nodes until we find leaves.
326 * We're doing a depth-first traversal in order to invalidate everything.
329 xfs_attr3_root_inactive(
330 struct xfs_trans **trans,
331 struct xfs_inode *dp)
333 struct xfs_da_blkinfo *info;
339 * Read block 0 to see what we have to work with.
340 * We only get here if we have extents, since we remove
341 * the extents in reverse order the extent containing
342 * block 0 must still be there.
344 error = xfs_da3_node_read(*trans, dp, 0, -1, &bp, XFS_ATTR_FORK);
350 * Invalidate the tree, even if the "tree" is only a single leaf block.
351 * This is a depth-first traversal!
354 switch (info->magic) {
355 case cpu_to_be16(XFS_DA_NODE_MAGIC):
356 case cpu_to_be16(XFS_DA3_NODE_MAGIC):
357 error = xfs_attr3_node_inactive(trans, dp, bp, 1);
359 case cpu_to_be16(XFS_ATTR_LEAF_MAGIC):
360 case cpu_to_be16(XFS_ATTR3_LEAF_MAGIC):
361 error = xfs_attr3_leaf_inactive(trans, dp, bp);
364 error = XFS_ERROR(EIO);
365 xfs_trans_brelse(*trans, bp);
372 * Invalidate the incore copy of the root block.
374 error = xfs_da_get_buf(*trans, dp, 0, blkno, &bp, XFS_ATTR_FORK);
377 xfs_trans_binval(*trans, bp); /* remove from cache */
379 * Commit the invalidate and start the next transaction.
381 error = xfs_trans_roll(trans, dp);
387 xfs_attr_inactive(xfs_inode_t *dp)
394 ASSERT(! XFS_NOT_DQATTACHED(mp, dp));
396 xfs_ilock(dp, XFS_ILOCK_SHARED);
397 if (!xfs_inode_hasattr(dp) ||
398 dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
399 xfs_iunlock(dp, XFS_ILOCK_SHARED);
402 xfs_iunlock(dp, XFS_ILOCK_SHARED);
405 * Start our first transaction of the day.
407 * All future transactions during this code must be "chained" off
408 * this one via the trans_dup() call. All transactions will contain
409 * the inode, and the inode will always be marked with trans_ihold().
410 * Since the inode will be locked in all transactions, we must log
411 * the inode in every transaction to let it float upward through
414 trans = xfs_trans_alloc(mp, XFS_TRANS_ATTRINVAL);
415 error = xfs_trans_reserve(trans, &M_RES(mp)->tr_attrinval, 0, 0);
417 xfs_trans_cancel(trans, 0);
420 xfs_ilock(dp, XFS_ILOCK_EXCL);
423 * No need to make quota reservations here. We expect to release some
424 * blocks, not allocate, in the common case.
426 xfs_trans_ijoin(trans, dp, 0);
429 * Decide on what work routines to call based on the inode size.
431 if (!xfs_inode_hasattr(dp) ||
432 dp->i_d.di_aformat == XFS_DINODE_FMT_LOCAL) {
436 error = xfs_attr3_root_inactive(&trans, dp);
440 error = xfs_itruncate_extents(&trans, dp, XFS_ATTR_FORK, 0);
444 error = xfs_trans_commit(trans, XFS_TRANS_RELEASE_LOG_RES);
445 xfs_iunlock(dp, XFS_ILOCK_EXCL);
450 xfs_trans_cancel(trans, XFS_TRANS_RELEASE_LOG_RES|XFS_TRANS_ABORT);
451 xfs_iunlock(dp, XFS_ILOCK_EXCL);