]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - fs/xfs/xfs_inode_item.c
xfs: don't leak uuid table on rmmod
[mirror_ubuntu-zesty-kernel.git] / fs / xfs / xfs_inode_item.c
CommitLineData
1da177e4 1/*
7b718769
NS
2 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
1da177e4 4 *
7b718769
NS
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
1da177e4
LT
7 * published by the Free Software Foundation.
8 *
7b718769
NS
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
1da177e4 13 *
7b718769
NS
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
1da177e4 17 */
1da177e4 18#include "xfs.h"
a844f451 19#include "xfs_fs.h"
a4fbe6ab 20#include "xfs_format.h"
239880ef
DC
21#include "xfs_log_format.h"
22#include "xfs_trans_resv.h"
1da177e4 23#include "xfs_mount.h"
1da177e4 24#include "xfs_inode.h"
239880ef 25#include "xfs_trans.h"
a844f451 26#include "xfs_inode_item.h"
db7a19f2 27#include "xfs_error.h"
0b1b213f 28#include "xfs_trace.h"
239880ef 29#include "xfs_trans_priv.h"
1234351c 30#include "xfs_log.h"
1da177e4
LT
31
32
33kmem_zone_t *xfs_ili_zone; /* inode log item zone */
34
7bfa31d8
CH
35static inline struct xfs_inode_log_item *INODE_ITEM(struct xfs_log_item *lip)
36{
37 return container_of(lip, struct xfs_inode_log_item, ili_item);
38}
39
166d1368 40STATIC void
ce9641d6
CH
41xfs_inode_item_data_fork_size(
42 struct xfs_inode_log_item *iip,
166d1368
DC
43 int *nvecs,
44 int *nbytes)
1da177e4 45{
7bfa31d8 46 struct xfs_inode *ip = iip->ili_inode;
166d1368 47
1da177e4
LT
48 switch (ip->i_d.di_format) {
49 case XFS_DINODE_FMT_EXTENTS:
f5d8d5c4 50 if ((iip->ili_fields & XFS_ILOG_DEXT) &&
339a5f5d 51 ip->i_d.di_nextents > 0 &&
166d1368
DC
52 ip->i_df.if_bytes > 0) {
53 /* worst case, doesn't subtract delalloc extents */
54 *nbytes += XFS_IFORK_DSIZE(ip);
55 *nvecs += 1;
56 }
1da177e4 57 break;
1da177e4 58 case XFS_DINODE_FMT_BTREE:
f5d8d5c4 59 if ((iip->ili_fields & XFS_ILOG_DBROOT) &&
166d1368
DC
60 ip->i_df.if_broot_bytes > 0) {
61 *nbytes += ip->i_df.if_broot_bytes;
62 *nvecs += 1;
63 }
1da177e4 64 break;
1da177e4 65 case XFS_DINODE_FMT_LOCAL:
f5d8d5c4 66 if ((iip->ili_fields & XFS_ILOG_DDATA) &&
166d1368
DC
67 ip->i_df.if_bytes > 0) {
68 *nbytes += roundup(ip->i_df.if_bytes, 4);
69 *nvecs += 1;
70 }
1da177e4
LT
71 break;
72
73 case XFS_DINODE_FMT_DEV:
1da177e4 74 case XFS_DINODE_FMT_UUID:
1da177e4 75 break;
1da177e4
LT
76 default:
77 ASSERT(0);
78 break;
79 }
ce9641d6 80}
1da177e4 81
ce9641d6
CH
82STATIC void
83xfs_inode_item_attr_fork_size(
84 struct xfs_inode_log_item *iip,
85 int *nvecs,
86 int *nbytes)
87{
88 struct xfs_inode *ip = iip->ili_inode;
1da177e4 89
1da177e4
LT
90 switch (ip->i_d.di_aformat) {
91 case XFS_DINODE_FMT_EXTENTS:
f5d8d5c4 92 if ((iip->ili_fields & XFS_ILOG_AEXT) &&
339a5f5d 93 ip->i_d.di_anextents > 0 &&
166d1368
DC
94 ip->i_afp->if_bytes > 0) {
95 /* worst case, doesn't subtract unused space */
96 *nbytes += XFS_IFORK_ASIZE(ip);
97 *nvecs += 1;
98 }
1da177e4 99 break;
1da177e4 100 case XFS_DINODE_FMT_BTREE:
f5d8d5c4 101 if ((iip->ili_fields & XFS_ILOG_ABROOT) &&
166d1368
DC
102 ip->i_afp->if_broot_bytes > 0) {
103 *nbytes += ip->i_afp->if_broot_bytes;
104 *nvecs += 1;
105 }
1da177e4 106 break;
1da177e4 107 case XFS_DINODE_FMT_LOCAL:
f5d8d5c4 108 if ((iip->ili_fields & XFS_ILOG_ADATA) &&
166d1368
DC
109 ip->i_afp->if_bytes > 0) {
110 *nbytes += roundup(ip->i_afp->if_bytes, 4);
111 *nvecs += 1;
112 }
1da177e4 113 break;
1da177e4
LT
114 default:
115 ASSERT(0);
116 break;
117 }
1da177e4
LT
118}
119
ce9641d6
CH
120/*
121 * This returns the number of iovecs needed to log the given inode item.
122 *
123 * We need one iovec for the inode log format structure, one for the
124 * inode core, and possibly one for the inode data/extents/b-tree root
125 * and one for the inode attribute data/extents/b-tree root.
126 */
127STATIC void
128xfs_inode_item_size(
129 struct xfs_log_item *lip,
130 int *nvecs,
131 int *nbytes)
132{
133 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
134 struct xfs_inode *ip = iip->ili_inode;
135
136 *nvecs += 2;
137 *nbytes += sizeof(struct xfs_inode_log_format) +
138 xfs_icdinode_size(ip->i_d.di_version);
139
140 xfs_inode_item_data_fork_size(iip, nvecs, nbytes);
141 if (XFS_IFORK_Q(ip))
142 xfs_inode_item_attr_fork_size(iip, nvecs, nbytes);
143}
144
1234351c 145STATIC void
3de559fb
CH
146xfs_inode_item_format_data_fork(
147 struct xfs_inode_log_item *iip,
bde7cff6
CH
148 struct xfs_inode_log_format *ilf,
149 struct xfs_log_vec *lv,
150 struct xfs_log_iovec **vecp)
1da177e4 151{
7bfa31d8 152 struct xfs_inode *ip = iip->ili_inode;
1da177e4 153 size_t data_bytes;
1da177e4
LT
154
155 switch (ip->i_d.di_format) {
156 case XFS_DINODE_FMT_EXTENTS:
f5d8d5c4 157 iip->ili_fields &=
339a5f5d
CH
158 ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT |
159 XFS_ILOG_DEV | XFS_ILOG_UUID);
160
f5d8d5c4 161 if ((iip->ili_fields & XFS_ILOG_DEXT) &&
339a5f5d
CH
162 ip->i_d.di_nextents > 0 &&
163 ip->i_df.if_bytes > 0) {
da776503
CH
164 struct xfs_bmbt_rec *p;
165
1da177e4 166 ASSERT(ip->i_df.if_u1.if_extents != NULL);
339a5f5d 167 ASSERT(ip->i_df.if_bytes / sizeof(xfs_bmbt_rec_t) > 0);
da776503
CH
168
169 p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IEXT);
170 data_bytes = xfs_iextents_copy(ip, p, XFS_DATA_FORK);
171 xlog_finish_iovec(lv, *vecp, data_bytes);
172
173 ASSERT(data_bytes <= ip->i_df.if_bytes);
174
175 ilf->ilf_dsize = data_bytes;
bde7cff6 176 ilf->ilf_size++;
339a5f5d 177 } else {
f5d8d5c4 178 iip->ili_fields &= ~XFS_ILOG_DEXT;
1da177e4
LT
179 }
180 break;
1da177e4 181 case XFS_DINODE_FMT_BTREE:
f5d8d5c4 182 iip->ili_fields &=
339a5f5d
CH
183 ~(XFS_ILOG_DDATA | XFS_ILOG_DEXT |
184 XFS_ILOG_DEV | XFS_ILOG_UUID);
185
f5d8d5c4 186 if ((iip->ili_fields & XFS_ILOG_DBROOT) &&
339a5f5d 187 ip->i_df.if_broot_bytes > 0) {
1da177e4 188 ASSERT(ip->i_df.if_broot != NULL);
bde7cff6 189 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IBROOT,
1234351c
CH
190 ip->i_df.if_broot,
191 ip->i_df.if_broot_bytes);
bde7cff6
CH
192 ilf->ilf_dsize = ip->i_df.if_broot_bytes;
193 ilf->ilf_size++;
339a5f5d 194 } else {
f5d8d5c4 195 ASSERT(!(iip->ili_fields &
339a5f5d 196 XFS_ILOG_DBROOT));
f5d8d5c4 197 iip->ili_fields &= ~XFS_ILOG_DBROOT;
1da177e4
LT
198 }
199 break;
1da177e4 200 case XFS_DINODE_FMT_LOCAL:
f5d8d5c4 201 iip->ili_fields &=
339a5f5d
CH
202 ~(XFS_ILOG_DEXT | XFS_ILOG_DBROOT |
203 XFS_ILOG_DEV | XFS_ILOG_UUID);
f5d8d5c4 204 if ((iip->ili_fields & XFS_ILOG_DDATA) &&
339a5f5d 205 ip->i_df.if_bytes > 0) {
1da177e4
LT
206 /*
207 * Round i_bytes up to a word boundary.
208 * The underlying memory is guaranteed to
209 * to be there by xfs_idata_realloc().
210 */
211 data_bytes = roundup(ip->i_df.if_bytes, 4);
1234351c
CH
212 ASSERT(ip->i_df.if_real_bytes == 0 ||
213 ip->i_df.if_real_bytes == data_bytes);
214 ASSERT(ip->i_df.if_u1.if_data != NULL);
215 ASSERT(ip->i_d.di_size > 0);
bde7cff6 216 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_ILOCAL,
1234351c 217 ip->i_df.if_u1.if_data, data_bytes);
bde7cff6
CH
218 ilf->ilf_dsize = (unsigned)data_bytes;
219 ilf->ilf_size++;
339a5f5d 220 } else {
f5d8d5c4 221 iip->ili_fields &= ~XFS_ILOG_DDATA;
1da177e4
LT
222 }
223 break;
1da177e4 224 case XFS_DINODE_FMT_DEV:
f5d8d5c4 225 iip->ili_fields &=
339a5f5d
CH
226 ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT |
227 XFS_ILOG_DEXT | XFS_ILOG_UUID);
bde7cff6
CH
228 if (iip->ili_fields & XFS_ILOG_DEV)
229 ilf->ilf_u.ilfu_rdev = ip->i_df.if_u2.if_rdev;
1da177e4 230 break;
1da177e4 231 case XFS_DINODE_FMT_UUID:
f5d8d5c4 232 iip->ili_fields &=
339a5f5d
CH
233 ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT |
234 XFS_ILOG_DEXT | XFS_ILOG_DEV);
bde7cff6
CH
235 if (iip->ili_fields & XFS_ILOG_UUID)
236 ilf->ilf_u.ilfu_uuid = ip->i_df.if_u2.if_uuid;
1da177e4 237 break;
1da177e4
LT
238 default:
239 ASSERT(0);
240 break;
241 }
3de559fb
CH
242}
243
1234351c 244STATIC void
3de559fb
CH
245xfs_inode_item_format_attr_fork(
246 struct xfs_inode_log_item *iip,
bde7cff6
CH
247 struct xfs_inode_log_format *ilf,
248 struct xfs_log_vec *lv,
249 struct xfs_log_iovec **vecp)
3de559fb
CH
250{
251 struct xfs_inode *ip = iip->ili_inode;
252 size_t data_bytes;
1da177e4
LT
253
254 switch (ip->i_d.di_aformat) {
255 case XFS_DINODE_FMT_EXTENTS:
f5d8d5c4 256 iip->ili_fields &=
339a5f5d
CH
257 ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT);
258
f5d8d5c4 259 if ((iip->ili_fields & XFS_ILOG_AEXT) &&
339a5f5d
CH
260 ip->i_d.di_anextents > 0 &&
261 ip->i_afp->if_bytes > 0) {
da776503
CH
262 struct xfs_bmbt_rec *p;
263
339a5f5d
CH
264 ASSERT(ip->i_afp->if_bytes / sizeof(xfs_bmbt_rec_t) ==
265 ip->i_d.di_anextents);
73523a2e 266 ASSERT(ip->i_afp->if_u1.if_extents != NULL);
da776503
CH
267
268 p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_EXT);
269 data_bytes = xfs_iextents_copy(ip, p, XFS_ATTR_FORK);
270 xlog_finish_iovec(lv, *vecp, data_bytes);
271
272 ilf->ilf_asize = data_bytes;
bde7cff6 273 ilf->ilf_size++;
339a5f5d 274 } else {
f5d8d5c4 275 iip->ili_fields &= ~XFS_ILOG_AEXT;
1da177e4
LT
276 }
277 break;
1da177e4 278 case XFS_DINODE_FMT_BTREE:
f5d8d5c4 279 iip->ili_fields &=
339a5f5d
CH
280 ~(XFS_ILOG_ADATA | XFS_ILOG_AEXT);
281
f5d8d5c4 282 if ((iip->ili_fields & XFS_ILOG_ABROOT) &&
339a5f5d 283 ip->i_afp->if_broot_bytes > 0) {
1da177e4 284 ASSERT(ip->i_afp->if_broot != NULL);
339a5f5d 285
bde7cff6 286 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_BROOT,
1234351c
CH
287 ip->i_afp->if_broot,
288 ip->i_afp->if_broot_bytes);
bde7cff6
CH
289 ilf->ilf_asize = ip->i_afp->if_broot_bytes;
290 ilf->ilf_size++;
339a5f5d 291 } else {
f5d8d5c4 292 iip->ili_fields &= ~XFS_ILOG_ABROOT;
1da177e4
LT
293 }
294 break;
1da177e4 295 case XFS_DINODE_FMT_LOCAL:
f5d8d5c4 296 iip->ili_fields &=
339a5f5d
CH
297 ~(XFS_ILOG_AEXT | XFS_ILOG_ABROOT);
298
f5d8d5c4 299 if ((iip->ili_fields & XFS_ILOG_ADATA) &&
339a5f5d 300 ip->i_afp->if_bytes > 0) {
1da177e4
LT
301 /*
302 * Round i_bytes up to a word boundary.
303 * The underlying memory is guaranteed to
304 * to be there by xfs_idata_realloc().
305 */
306 data_bytes = roundup(ip->i_afp->if_bytes, 4);
1234351c
CH
307 ASSERT(ip->i_afp->if_real_bytes == 0 ||
308 ip->i_afp->if_real_bytes == data_bytes);
309 ASSERT(ip->i_afp->if_u1.if_data != NULL);
bde7cff6 310 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_LOCAL,
1234351c
CH
311 ip->i_afp->if_u1.if_data,
312 data_bytes);
bde7cff6
CH
313 ilf->ilf_asize = (unsigned)data_bytes;
314 ilf->ilf_size++;
339a5f5d 315 } else {
f5d8d5c4 316 iip->ili_fields &= ~XFS_ILOG_ADATA;
1da177e4
LT
317 }
318 break;
1da177e4
LT
319 default:
320 ASSERT(0);
321 break;
322 }
3de559fb
CH
323}
324
325/*
326 * This is called to fill in the vector of log iovecs for the given inode
327 * log item. It fills the first item with an inode log format structure,
328 * the second with the on-disk inode structure, and a possible third and/or
329 * fourth with the inode data/extents/b-tree root and inode attributes
330 * data/extents/b-tree root.
331 */
332STATIC void
333xfs_inode_item_format(
334 struct xfs_log_item *lip,
bde7cff6 335 struct xfs_log_vec *lv)
3de559fb
CH
336{
337 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
338 struct xfs_inode *ip = iip->ili_inode;
bde7cff6
CH
339 struct xfs_inode_log_format *ilf;
340 struct xfs_log_iovec *vecp = NULL;
3de559fb 341
263997a6
DC
342 ASSERT(ip->i_d.di_version > 1);
343
2f251293
CH
344 ilf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_IFORMAT);
345 ilf->ilf_type = XFS_LI_INODE;
346 ilf->ilf_ino = ip->i_ino;
347 ilf->ilf_blkno = ip->i_imap.im_blkno;
348 ilf->ilf_len = ip->i_imap.im_len;
349 ilf->ilf_boffset = ip->i_imap.im_boffset;
350 ilf->ilf_fields = XFS_ILOG_CORE;
351 ilf->ilf_size = 2; /* format + core */
352 xlog_finish_iovec(lv, vecp, sizeof(struct xfs_inode_log_format));
3de559fb 353
bde7cff6
CH
354 xlog_copy_iovec(lv, &vecp, XLOG_REG_TYPE_ICORE,
355 &ip->i_d,
356 xfs_icdinode_size(ip->i_d.di_version));
3de559fb 357
bde7cff6 358 xfs_inode_item_format_data_fork(iip, ilf, lv, &vecp);
3de559fb 359 if (XFS_IFORK_Q(ip)) {
bde7cff6 360 xfs_inode_item_format_attr_fork(iip, ilf, lv, &vecp);
3de559fb
CH
361 } else {
362 iip->ili_fields &=
363 ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT);
364 }
365
2f251293
CH
366 /* update the format with the exact fields we actually logged */
367 ilf->ilf_fields |= (iip->ili_fields & ~XFS_ILOG_TIMESTAMP);
1da177e4
LT
368}
369
1da177e4
LT
370/*
371 * This is called to pin the inode associated with the inode log
a14a5ab5 372 * item in memory so it cannot be written out.
1da177e4
LT
373 */
374STATIC void
375xfs_inode_item_pin(
7bfa31d8 376 struct xfs_log_item *lip)
1da177e4 377{
7bfa31d8 378 struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode;
a14a5ab5 379
7bfa31d8
CH
380 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
381
382 trace_xfs_inode_pin(ip, _RET_IP_);
383 atomic_inc(&ip->i_pincount);
1da177e4
LT
384}
385
386
387/*
388 * This is called to unpin the inode associated with the inode log
389 * item which was previously pinned with a call to xfs_inode_item_pin().
a14a5ab5
CH
390 *
391 * Also wake up anyone in xfs_iunpin_wait() if the count goes to 0.
1da177e4 392 */
1da177e4
LT
393STATIC void
394xfs_inode_item_unpin(
7bfa31d8 395 struct xfs_log_item *lip,
9412e318 396 int remove)
1da177e4 397{
7bfa31d8 398 struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode;
a14a5ab5 399
4aaf15d1 400 trace_xfs_inode_unpin(ip, _RET_IP_);
a14a5ab5
CH
401 ASSERT(atomic_read(&ip->i_pincount) > 0);
402 if (atomic_dec_and_test(&ip->i_pincount))
f392e631 403 wake_up_bit(&ip->i_flags, __XFS_IPINNED_BIT);
1da177e4
LT
404}
405
1da177e4 406STATIC uint
43ff2122
CH
407xfs_inode_item_push(
408 struct xfs_log_item *lip,
409 struct list_head *buffer_list)
1da177e4 410{
7bfa31d8
CH
411 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
412 struct xfs_inode *ip = iip->ili_inode;
43ff2122
CH
413 struct xfs_buf *bp = NULL;
414 uint rval = XFS_ITEM_SUCCESS;
415 int error;
1da177e4 416
7bfa31d8 417 if (xfs_ipincount(ip) > 0)
1da177e4 418 return XFS_ITEM_PINNED;
1da177e4 419
7bfa31d8 420 if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
1da177e4 421 return XFS_ITEM_LOCKED;
1da177e4 422
4c46819a
CH
423 /*
424 * Re-check the pincount now that we stabilized the value by
425 * taking the ilock.
426 */
427 if (xfs_ipincount(ip) > 0) {
43ff2122
CH
428 rval = XFS_ITEM_PINNED;
429 goto out_unlock;
4c46819a
CH
430 }
431
9a3a5dab
BF
432 /*
433 * Stale inode items should force out the iclog.
434 */
435 if (ip->i_flags & XFS_ISTALE) {
436 rval = XFS_ITEM_PINNED;
437 goto out_unlock;
438 }
439
43ff2122
CH
440 /*
441 * Someone else is already flushing the inode. Nothing we can do
442 * here but wait for the flush to finish and remove the item from
443 * the AIL.
444 */
1da177e4 445 if (!xfs_iflock_nowait(ip)) {
43ff2122
CH
446 rval = XFS_ITEM_FLUSHING;
447 goto out_unlock;
1da177e4
LT
448 }
449
43ff2122
CH
450 ASSERT(iip->ili_fields != 0 || XFS_FORCED_SHUTDOWN(ip->i_mount));
451 ASSERT(iip->ili_logged == 0 || XFS_FORCED_SHUTDOWN(ip->i_mount));
452
453 spin_unlock(&lip->li_ailp->xa_lock);
454
455 error = xfs_iflush(ip, &bp);
456 if (!error) {
457 if (!xfs_buf_delwri_queue(bp, buffer_list))
458 rval = XFS_ITEM_FLUSHING;
459 xfs_buf_relse(bp);
1da177e4 460 }
43ff2122
CH
461
462 spin_lock(&lip->li_ailp->xa_lock);
463out_unlock:
464 xfs_iunlock(ip, XFS_ILOCK_SHARED);
465 return rval;
1da177e4
LT
466}
467
468/*
469 * Unlock the inode associated with the inode log item.
470 * Clear the fields of the inode and inode log item that
471 * are specific to the current transaction. If the
472 * hold flags is set, do not unlock the inode.
473 */
474STATIC void
475xfs_inode_item_unlock(
7bfa31d8 476 struct xfs_log_item *lip)
1da177e4 477{
7bfa31d8
CH
478 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
479 struct xfs_inode *ip = iip->ili_inode;
898621d5 480 unsigned short lock_flags;
1da177e4 481
f3ca8738
CH
482 ASSERT(ip->i_itemp != NULL);
483 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
1da177e4 484
898621d5
CH
485 lock_flags = iip->ili_lock_flags;
486 iip->ili_lock_flags = 0;
ddc3415a 487 if (lock_flags)
f3ca8738 488 xfs_iunlock(ip, lock_flags);
1da177e4
LT
489}
490
491/*
de25c181
DC
492 * This is called to find out where the oldest active copy of the inode log
493 * item in the on disk log resides now that the last log write of it completed
494 * at the given lsn. Since we always re-log all dirty data in an inode, the
495 * latest copy in the on disk log is the only one that matters. Therefore,
496 * simply return the given lsn.
497 *
498 * If the inode has been marked stale because the cluster is being freed, we
499 * don't want to (re-)insert this inode into the AIL. There is a race condition
500 * where the cluster buffer may be unpinned before the inode is inserted into
501 * the AIL during transaction committed processing. If the buffer is unpinned
502 * before the inode item has been committed and inserted, then it is possible
1316d4da 503 * for the buffer to be written and IO completes before the inode is inserted
de25c181
DC
504 * into the AIL. In that case, we'd be inserting a clean, stale inode into the
505 * AIL which will never get removed. It will, however, get reclaimed which
506 * triggers an assert in xfs_inode_free() complaining about freein an inode
507 * still in the AIL.
508 *
1316d4da
DC
509 * To avoid this, just unpin the inode directly and return a LSN of -1 so the
510 * transaction committed code knows that it does not need to do any further
511 * processing on the item.
1da177e4 512 */
1da177e4
LT
513STATIC xfs_lsn_t
514xfs_inode_item_committed(
7bfa31d8 515 struct xfs_log_item *lip,
1da177e4
LT
516 xfs_lsn_t lsn)
517{
de25c181
DC
518 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
519 struct xfs_inode *ip = iip->ili_inode;
520
1316d4da
DC
521 if (xfs_iflags_test(ip, XFS_ISTALE)) {
522 xfs_inode_item_unpin(lip, 0);
523 return -1;
524 }
7bfa31d8 525 return lsn;
1da177e4
LT
526}
527
1da177e4
LT
528/*
529 * XXX rcc - this one really has to do something. Probably needs
530 * to stamp in a new field in the incore inode.
531 */
1da177e4
LT
532STATIC void
533xfs_inode_item_committing(
7bfa31d8 534 struct xfs_log_item *lip,
1da177e4
LT
535 xfs_lsn_t lsn)
536{
7bfa31d8 537 INODE_ITEM(lip)->ili_last_lsn = lsn;
1da177e4
LT
538}
539
540/*
541 * This is the ops vector shared by all buf log items.
542 */
272e42b2 543static const struct xfs_item_ops xfs_inode_item_ops = {
7bfa31d8
CH
544 .iop_size = xfs_inode_item_size,
545 .iop_format = xfs_inode_item_format,
546 .iop_pin = xfs_inode_item_pin,
547 .iop_unpin = xfs_inode_item_unpin,
7bfa31d8
CH
548 .iop_unlock = xfs_inode_item_unlock,
549 .iop_committed = xfs_inode_item_committed,
550 .iop_push = xfs_inode_item_push,
7bfa31d8 551 .iop_committing = xfs_inode_item_committing
1da177e4
LT
552};
553
554
555/*
556 * Initialize the inode log item for a newly allocated (in-core) inode.
557 */
558void
559xfs_inode_item_init(
7bfa31d8
CH
560 struct xfs_inode *ip,
561 struct xfs_mount *mp)
1da177e4 562{
7bfa31d8 563 struct xfs_inode_log_item *iip;
1da177e4
LT
564
565 ASSERT(ip->i_itemp == NULL);
566 iip = ip->i_itemp = kmem_zone_zalloc(xfs_ili_zone, KM_SLEEP);
567
1da177e4 568 iip->ili_inode = ip;
43f5efc5
DC
569 xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE,
570 &xfs_inode_item_ops);
1da177e4
LT
571}
572
573/*
574 * Free the inode log item and any memory hanging off of it.
575 */
576void
577xfs_inode_item_destroy(
578 xfs_inode_t *ip)
579{
1da177e4
LT
580 kmem_zone_free(xfs_ili_zone, ip->i_itemp);
581}
582
583
584/*
585 * This is the inode flushing I/O completion routine. It is called
586 * from interrupt level when the buffer containing the inode is
587 * flushed to disk. It is responsible for removing the inode item
588 * from the AIL if it has not been re-logged, and unlocking the inode's
589 * flush lock.
30136832
DC
590 *
591 * To reduce AIL lock traffic as much as possible, we scan the buffer log item
592 * list for other inodes that will run this function. We remove them from the
593 * buffer list so we can process all the inode IO completions in one AIL lock
594 * traversal.
1da177e4 595 */
1da177e4
LT
596void
597xfs_iflush_done(
ca30b2a7
CH
598 struct xfs_buf *bp,
599 struct xfs_log_item *lip)
1da177e4 600{
30136832
DC
601 struct xfs_inode_log_item *iip;
602 struct xfs_log_item *blip;
603 struct xfs_log_item *next;
604 struct xfs_log_item *prev;
ca30b2a7 605 struct xfs_ail *ailp = lip->li_ailp;
30136832
DC
606 int need_ail = 0;
607
608 /*
609 * Scan the buffer IO completions for other inodes being completed and
610 * attach them to the current inode log item.
611 */
adadbeef 612 blip = bp->b_fspriv;
30136832
DC
613 prev = NULL;
614 while (blip != NULL) {
52177937 615 if (blip->li_cb != xfs_iflush_done) {
30136832
DC
616 prev = blip;
617 blip = blip->li_bio_list;
618 continue;
619 }
620
621 /* remove from list */
622 next = blip->li_bio_list;
623 if (!prev) {
adadbeef 624 bp->b_fspriv = next;
30136832
DC
625 } else {
626 prev->li_bio_list = next;
627 }
628
629 /* add to current list */
630 blip->li_bio_list = lip->li_bio_list;
631 lip->li_bio_list = blip;
632
633 /*
634 * while we have the item, do the unlocked check for needing
635 * the AIL lock.
636 */
637 iip = INODE_ITEM(blip);
638 if (iip->ili_logged && blip->li_lsn == iip->ili_flush_lsn)
639 need_ail++;
640
641 blip = next;
642 }
643
644 /* make sure we capture the state of the initial inode. */
645 iip = INODE_ITEM(lip);
646 if (iip->ili_logged && lip->li_lsn == iip->ili_flush_lsn)
647 need_ail++;
1da177e4
LT
648
649 /*
650 * We only want to pull the item from the AIL if it is
651 * actually there and its location in the log has not
652 * changed since we started the flush. Thus, we only bother
653 * if the ili_logged flag is set and the inode's lsn has not
654 * changed. First we check the lsn outside
655 * the lock since it's cheaper, and then we recheck while
656 * holding the lock before removing the inode from the AIL.
657 */
30136832
DC
658 if (need_ail) {
659 struct xfs_log_item *log_items[need_ail];
660 int i = 0;
783a2f65 661 spin_lock(&ailp->xa_lock);
30136832
DC
662 for (blip = lip; blip; blip = blip->li_bio_list) {
663 iip = INODE_ITEM(blip);
664 if (iip->ili_logged &&
665 blip->li_lsn == iip->ili_flush_lsn) {
666 log_items[i++] = blip;
667 }
668 ASSERT(i <= need_ail);
1da177e4 669 }
30136832 670 /* xfs_trans_ail_delete_bulk() drops the AIL lock. */
04913fdd
DC
671 xfs_trans_ail_delete_bulk(ailp, log_items, i,
672 SHUTDOWN_CORRUPT_INCORE);
1da177e4
LT
673 }
674
1da177e4
LT
675
676 /*
30136832
DC
677 * clean up and unlock the flush lock now we are done. We can clear the
678 * ili_last_fields bits now that we know that the data corresponding to
679 * them is safely on disk.
1da177e4 680 */
30136832
DC
681 for (blip = lip; blip; blip = next) {
682 next = blip->li_bio_list;
683 blip->li_bio_list = NULL;
684
685 iip = INODE_ITEM(blip);
686 iip->ili_logged = 0;
687 iip->ili_last_fields = 0;
688 xfs_ifunlock(iip->ili_inode);
689 }
1da177e4
LT
690}
691
692/*
04913fdd
DC
693 * This is the inode flushing abort routine. It is called from xfs_iflush when
694 * the filesystem is shutting down to clean up the inode state. It is
695 * responsible for removing the inode item from the AIL if it has not been
696 * re-logged, and unlocking the inode's flush lock.
1da177e4
LT
697 */
698void
699xfs_iflush_abort(
04913fdd
DC
700 xfs_inode_t *ip,
701 bool stale)
1da177e4 702{
783a2f65 703 xfs_inode_log_item_t *iip = ip->i_itemp;
1da177e4 704
1da177e4
LT
705 if (iip) {
706 if (iip->ili_item.li_flags & XFS_LI_IN_AIL) {
146e54b7
BF
707 xfs_trans_ail_remove(&iip->ili_item,
708 stale ? SHUTDOWN_LOG_IO_ERROR :
04913fdd 709 SHUTDOWN_CORRUPT_INCORE);
1da177e4
LT
710 }
711 iip->ili_logged = 0;
712 /*
713 * Clear the ili_last_fields bits now that we know that the
714 * data corresponding to them is safely on disk.
715 */
716 iip->ili_last_fields = 0;
717 /*
718 * Clear the inode logging fields so no more flushes are
719 * attempted.
720 */
f5d8d5c4 721 iip->ili_fields = 0;
1da177e4
LT
722 }
723 /*
724 * Release the inode's flush lock since we're done with it.
725 */
726 xfs_ifunlock(ip);
727}
728
729void
730xfs_istale_done(
ca30b2a7
CH
731 struct xfs_buf *bp,
732 struct xfs_log_item *lip)
1da177e4 733{
04913fdd 734 xfs_iflush_abort(INODE_ITEM(lip)->ili_inode, true);
1da177e4 735}
6d192a9b
TS
736
737/*
738 * convert an xfs_inode_log_format struct from either 32 or 64 bit versions
739 * (which can have different field alignments) to the native version
740 */
741int
742xfs_inode_item_format_convert(
743 xfs_log_iovec_t *buf,
744 xfs_inode_log_format_t *in_f)
745{
746 if (buf->i_len == sizeof(xfs_inode_log_format_32_t)) {
4e0d5f92 747 xfs_inode_log_format_32_t *in_f32 = buf->i_addr;
6d192a9b 748
6d192a9b
TS
749 in_f->ilf_type = in_f32->ilf_type;
750 in_f->ilf_size = in_f32->ilf_size;
751 in_f->ilf_fields = in_f32->ilf_fields;
752 in_f->ilf_asize = in_f32->ilf_asize;
753 in_f->ilf_dsize = in_f32->ilf_dsize;
754 in_f->ilf_ino = in_f32->ilf_ino;
755 /* copy biggest field of ilf_u */
756 memcpy(in_f->ilf_u.ilfu_uuid.__u_bits,
757 in_f32->ilf_u.ilfu_uuid.__u_bits,
758 sizeof(uuid_t));
759 in_f->ilf_blkno = in_f32->ilf_blkno;
760 in_f->ilf_len = in_f32->ilf_len;
761 in_f->ilf_boffset = in_f32->ilf_boffset;
762 return 0;
763 } else if (buf->i_len == sizeof(xfs_inode_log_format_64_t)){
4e0d5f92 764 xfs_inode_log_format_64_t *in_f64 = buf->i_addr;
6d192a9b 765
6d192a9b
TS
766 in_f->ilf_type = in_f64->ilf_type;
767 in_f->ilf_size = in_f64->ilf_size;
768 in_f->ilf_fields = in_f64->ilf_fields;
769 in_f->ilf_asize = in_f64->ilf_asize;
770 in_f->ilf_dsize = in_f64->ilf_dsize;
771 in_f->ilf_ino = in_f64->ilf_ino;
772 /* copy biggest field of ilf_u */
773 memcpy(in_f->ilf_u.ilfu_uuid.__u_bits,
774 in_f64->ilf_u.ilfu_uuid.__u_bits,
775 sizeof(uuid_t));
776 in_f->ilf_blkno = in_f64->ilf_blkno;
777 in_f->ilf_len = in_f64->ilf_len;
778 in_f->ilf_boffset = in_f64->ilf_boffset;
779 return 0;
780 }
2451337d 781 return -EFSCORRUPTED;
6d192a9b 782}