]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/xfs/xfs_inode_item.c
xfs: keep sb_bad_features2 the same a sb_features2
[mirror_ubuntu-bionic-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_sb.h"
a844f451 24#include "xfs_ag.h"
1da177e4 25#include "xfs_mount.h"
1da177e4 26#include "xfs_inode.h"
239880ef 27#include "xfs_trans.h"
a844f451 28#include "xfs_inode_item.h"
db7a19f2 29#include "xfs_error.h"
0b1b213f 30#include "xfs_trace.h"
239880ef 31#include "xfs_trans_priv.h"
a4fbe6ab 32#include "xfs_dinode.h"
1234351c 33#include "xfs_log.h"
1da177e4
LT
34
35
36kmem_zone_t *xfs_ili_zone; /* inode log item zone */
37
7bfa31d8
CH
38static inline struct xfs_inode_log_item *INODE_ITEM(struct xfs_log_item *lip)
39{
40 return container_of(lip, struct xfs_inode_log_item, ili_item);
41}
42
166d1368 43STATIC void
ce9641d6
CH
44xfs_inode_item_data_fork_size(
45 struct xfs_inode_log_item *iip,
166d1368
DC
46 int *nvecs,
47 int *nbytes)
1da177e4 48{
7bfa31d8 49 struct xfs_inode *ip = iip->ili_inode;
166d1368 50
1da177e4
LT
51 switch (ip->i_d.di_format) {
52 case XFS_DINODE_FMT_EXTENTS:
f5d8d5c4 53 if ((iip->ili_fields & XFS_ILOG_DEXT) &&
339a5f5d 54 ip->i_d.di_nextents > 0 &&
166d1368
DC
55 ip->i_df.if_bytes > 0) {
56 /* worst case, doesn't subtract delalloc extents */
57 *nbytes += XFS_IFORK_DSIZE(ip);
58 *nvecs += 1;
59 }
1da177e4 60 break;
1da177e4 61 case XFS_DINODE_FMT_BTREE:
f5d8d5c4 62 if ((iip->ili_fields & XFS_ILOG_DBROOT) &&
166d1368
DC
63 ip->i_df.if_broot_bytes > 0) {
64 *nbytes += ip->i_df.if_broot_bytes;
65 *nvecs += 1;
66 }
1da177e4 67 break;
1da177e4 68 case XFS_DINODE_FMT_LOCAL:
f5d8d5c4 69 if ((iip->ili_fields & XFS_ILOG_DDATA) &&
166d1368
DC
70 ip->i_df.if_bytes > 0) {
71 *nbytes += roundup(ip->i_df.if_bytes, 4);
72 *nvecs += 1;
73 }
1da177e4
LT
74 break;
75
76 case XFS_DINODE_FMT_DEV:
1da177e4 77 case XFS_DINODE_FMT_UUID:
1da177e4 78 break;
1da177e4
LT
79 default:
80 ASSERT(0);
81 break;
82 }
ce9641d6 83}
1da177e4 84
ce9641d6
CH
85STATIC void
86xfs_inode_item_attr_fork_size(
87 struct xfs_inode_log_item *iip,
88 int *nvecs,
89 int *nbytes)
90{
91 struct xfs_inode *ip = iip->ili_inode;
1da177e4 92
1da177e4
LT
93 switch (ip->i_d.di_aformat) {
94 case XFS_DINODE_FMT_EXTENTS:
f5d8d5c4 95 if ((iip->ili_fields & XFS_ILOG_AEXT) &&
339a5f5d 96 ip->i_d.di_anextents > 0 &&
166d1368
DC
97 ip->i_afp->if_bytes > 0) {
98 /* worst case, doesn't subtract unused space */
99 *nbytes += XFS_IFORK_ASIZE(ip);
100 *nvecs += 1;
101 }
1da177e4 102 break;
1da177e4 103 case XFS_DINODE_FMT_BTREE:
f5d8d5c4 104 if ((iip->ili_fields & XFS_ILOG_ABROOT) &&
166d1368
DC
105 ip->i_afp->if_broot_bytes > 0) {
106 *nbytes += ip->i_afp->if_broot_bytes;
107 *nvecs += 1;
108 }
1da177e4 109 break;
1da177e4 110 case XFS_DINODE_FMT_LOCAL:
f5d8d5c4 111 if ((iip->ili_fields & XFS_ILOG_ADATA) &&
166d1368
DC
112 ip->i_afp->if_bytes > 0) {
113 *nbytes += roundup(ip->i_afp->if_bytes, 4);
114 *nvecs += 1;
115 }
1da177e4 116 break;
1da177e4
LT
117 default:
118 ASSERT(0);
119 break;
120 }
1da177e4
LT
121}
122
ce9641d6
CH
123/*
124 * This returns the number of iovecs needed to log the given inode item.
125 *
126 * We need one iovec for the inode log format structure, one for the
127 * inode core, and possibly one for the inode data/extents/b-tree root
128 * and one for the inode attribute data/extents/b-tree root.
129 */
130STATIC void
131xfs_inode_item_size(
132 struct xfs_log_item *lip,
133 int *nvecs,
134 int *nbytes)
135{
136 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
137 struct xfs_inode *ip = iip->ili_inode;
138
139 *nvecs += 2;
140 *nbytes += sizeof(struct xfs_inode_log_format) +
141 xfs_icdinode_size(ip->i_d.di_version);
142
143 xfs_inode_item_data_fork_size(iip, nvecs, nbytes);
144 if (XFS_IFORK_Q(ip))
145 xfs_inode_item_attr_fork_size(iip, nvecs, nbytes);
146}
147
1da177e4 148/*
3de559fb
CH
149 * If this is a v1 format inode, then we need to log it as such. This means
150 * that we have to copy the link count from the new field to the old. We
151 * don't have to worry about the new fields, because nothing trusts them as
152 * long as the old inode version number is there.
1da177e4
LT
153 */
154STATIC void
3de559fb
CH
155xfs_inode_item_format_v1_inode(
156 struct xfs_inode *ip)
157{
158 if (!xfs_sb_version_hasnlink(&ip->i_mount->m_sb)) {
159 /*
160 * Convert it back.
161 */
162 ASSERT(ip->i_d.di_nlink <= XFS_MAXLINK_1);
163 ip->i_d.di_onlink = ip->i_d.di_nlink;
164 } else {
165 /*
166 * The superblock version has already been bumped,
167 * so just make the conversion to the new inode
168 * format permanent.
169 */
170 ip->i_d.di_version = 2;
171 ip->i_d.di_onlink = 0;
172 memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad));
173 }
174}
175
1234351c 176STATIC void
3de559fb
CH
177xfs_inode_item_format_data_fork(
178 struct xfs_inode_log_item *iip,
bde7cff6
CH
179 struct xfs_inode_log_format *ilf,
180 struct xfs_log_vec *lv,
181 struct xfs_log_iovec **vecp)
1da177e4 182{
7bfa31d8 183 struct xfs_inode *ip = iip->ili_inode;
1da177e4 184 size_t data_bytes;
1da177e4
LT
185
186 switch (ip->i_d.di_format) {
187 case XFS_DINODE_FMT_EXTENTS:
f5d8d5c4 188 iip->ili_fields &=
339a5f5d
CH
189 ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT |
190 XFS_ILOG_DEV | XFS_ILOG_UUID);
191
f5d8d5c4 192 if ((iip->ili_fields & XFS_ILOG_DEXT) &&
339a5f5d
CH
193 ip->i_d.di_nextents > 0 &&
194 ip->i_df.if_bytes > 0) {
da776503
CH
195 struct xfs_bmbt_rec *p;
196
1da177e4 197 ASSERT(ip->i_df.if_u1.if_extents != NULL);
339a5f5d 198 ASSERT(ip->i_df.if_bytes / sizeof(xfs_bmbt_rec_t) > 0);
da776503
CH
199
200 p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IEXT);
201 data_bytes = xfs_iextents_copy(ip, p, XFS_DATA_FORK);
202 xlog_finish_iovec(lv, *vecp, data_bytes);
203
204 ASSERT(data_bytes <= ip->i_df.if_bytes);
205
206 ilf->ilf_dsize = data_bytes;
bde7cff6 207 ilf->ilf_size++;
339a5f5d 208 } else {
f5d8d5c4 209 iip->ili_fields &= ~XFS_ILOG_DEXT;
1da177e4
LT
210 }
211 break;
1da177e4 212 case XFS_DINODE_FMT_BTREE:
f5d8d5c4 213 iip->ili_fields &=
339a5f5d
CH
214 ~(XFS_ILOG_DDATA | XFS_ILOG_DEXT |
215 XFS_ILOG_DEV | XFS_ILOG_UUID);
216
f5d8d5c4 217 if ((iip->ili_fields & XFS_ILOG_DBROOT) &&
339a5f5d 218 ip->i_df.if_broot_bytes > 0) {
1da177e4 219 ASSERT(ip->i_df.if_broot != NULL);
bde7cff6 220 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IBROOT,
1234351c
CH
221 ip->i_df.if_broot,
222 ip->i_df.if_broot_bytes);
bde7cff6
CH
223 ilf->ilf_dsize = ip->i_df.if_broot_bytes;
224 ilf->ilf_size++;
339a5f5d 225 } else {
f5d8d5c4 226 ASSERT(!(iip->ili_fields &
339a5f5d 227 XFS_ILOG_DBROOT));
f5d8d5c4 228 iip->ili_fields &= ~XFS_ILOG_DBROOT;
1da177e4
LT
229 }
230 break;
1da177e4 231 case XFS_DINODE_FMT_LOCAL:
f5d8d5c4 232 iip->ili_fields &=
339a5f5d
CH
233 ~(XFS_ILOG_DEXT | XFS_ILOG_DBROOT |
234 XFS_ILOG_DEV | XFS_ILOG_UUID);
f5d8d5c4 235 if ((iip->ili_fields & XFS_ILOG_DDATA) &&
339a5f5d 236 ip->i_df.if_bytes > 0) {
1da177e4
LT
237 /*
238 * Round i_bytes up to a word boundary.
239 * The underlying memory is guaranteed to
240 * to be there by xfs_idata_realloc().
241 */
242 data_bytes = roundup(ip->i_df.if_bytes, 4);
1234351c
CH
243 ASSERT(ip->i_df.if_real_bytes == 0 ||
244 ip->i_df.if_real_bytes == data_bytes);
245 ASSERT(ip->i_df.if_u1.if_data != NULL);
246 ASSERT(ip->i_d.di_size > 0);
bde7cff6 247 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_ILOCAL,
1234351c 248 ip->i_df.if_u1.if_data, data_bytes);
bde7cff6
CH
249 ilf->ilf_dsize = (unsigned)data_bytes;
250 ilf->ilf_size++;
339a5f5d 251 } else {
f5d8d5c4 252 iip->ili_fields &= ~XFS_ILOG_DDATA;
1da177e4
LT
253 }
254 break;
1da177e4 255 case XFS_DINODE_FMT_DEV:
f5d8d5c4 256 iip->ili_fields &=
339a5f5d
CH
257 ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT |
258 XFS_ILOG_DEXT | XFS_ILOG_UUID);
bde7cff6
CH
259 if (iip->ili_fields & XFS_ILOG_DEV)
260 ilf->ilf_u.ilfu_rdev = ip->i_df.if_u2.if_rdev;
1da177e4 261 break;
1da177e4 262 case XFS_DINODE_FMT_UUID:
f5d8d5c4 263 iip->ili_fields &=
339a5f5d
CH
264 ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT |
265 XFS_ILOG_DEXT | XFS_ILOG_DEV);
bde7cff6
CH
266 if (iip->ili_fields & XFS_ILOG_UUID)
267 ilf->ilf_u.ilfu_uuid = ip->i_df.if_u2.if_uuid;
1da177e4 268 break;
1da177e4
LT
269 default:
270 ASSERT(0);
271 break;
272 }
3de559fb
CH
273}
274
1234351c 275STATIC void
3de559fb
CH
276xfs_inode_item_format_attr_fork(
277 struct xfs_inode_log_item *iip,
bde7cff6
CH
278 struct xfs_inode_log_format *ilf,
279 struct xfs_log_vec *lv,
280 struct xfs_log_iovec **vecp)
3de559fb
CH
281{
282 struct xfs_inode *ip = iip->ili_inode;
283 size_t data_bytes;
1da177e4
LT
284
285 switch (ip->i_d.di_aformat) {
286 case XFS_DINODE_FMT_EXTENTS:
f5d8d5c4 287 iip->ili_fields &=
339a5f5d
CH
288 ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT);
289
f5d8d5c4 290 if ((iip->ili_fields & XFS_ILOG_AEXT) &&
339a5f5d
CH
291 ip->i_d.di_anextents > 0 &&
292 ip->i_afp->if_bytes > 0) {
da776503
CH
293 struct xfs_bmbt_rec *p;
294
339a5f5d
CH
295 ASSERT(ip->i_afp->if_bytes / sizeof(xfs_bmbt_rec_t) ==
296 ip->i_d.di_anextents);
73523a2e 297 ASSERT(ip->i_afp->if_u1.if_extents != NULL);
da776503
CH
298
299 p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_EXT);
300 data_bytes = xfs_iextents_copy(ip, p, XFS_ATTR_FORK);
301 xlog_finish_iovec(lv, *vecp, data_bytes);
302
303 ilf->ilf_asize = data_bytes;
bde7cff6 304 ilf->ilf_size++;
339a5f5d 305 } else {
f5d8d5c4 306 iip->ili_fields &= ~XFS_ILOG_AEXT;
1da177e4
LT
307 }
308 break;
1da177e4 309 case XFS_DINODE_FMT_BTREE:
f5d8d5c4 310 iip->ili_fields &=
339a5f5d
CH
311 ~(XFS_ILOG_ADATA | XFS_ILOG_AEXT);
312
f5d8d5c4 313 if ((iip->ili_fields & XFS_ILOG_ABROOT) &&
339a5f5d 314 ip->i_afp->if_broot_bytes > 0) {
1da177e4 315 ASSERT(ip->i_afp->if_broot != NULL);
339a5f5d 316
bde7cff6 317 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_BROOT,
1234351c
CH
318 ip->i_afp->if_broot,
319 ip->i_afp->if_broot_bytes);
bde7cff6
CH
320 ilf->ilf_asize = ip->i_afp->if_broot_bytes;
321 ilf->ilf_size++;
339a5f5d 322 } else {
f5d8d5c4 323 iip->ili_fields &= ~XFS_ILOG_ABROOT;
1da177e4
LT
324 }
325 break;
1da177e4 326 case XFS_DINODE_FMT_LOCAL:
f5d8d5c4 327 iip->ili_fields &=
339a5f5d
CH
328 ~(XFS_ILOG_AEXT | XFS_ILOG_ABROOT);
329
f5d8d5c4 330 if ((iip->ili_fields & XFS_ILOG_ADATA) &&
339a5f5d 331 ip->i_afp->if_bytes > 0) {
1da177e4
LT
332 /*
333 * Round i_bytes up to a word boundary.
334 * The underlying memory is guaranteed to
335 * to be there by xfs_idata_realloc().
336 */
337 data_bytes = roundup(ip->i_afp->if_bytes, 4);
1234351c
CH
338 ASSERT(ip->i_afp->if_real_bytes == 0 ||
339 ip->i_afp->if_real_bytes == data_bytes);
340 ASSERT(ip->i_afp->if_u1.if_data != NULL);
bde7cff6 341 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_LOCAL,
1234351c
CH
342 ip->i_afp->if_u1.if_data,
343 data_bytes);
bde7cff6
CH
344 ilf->ilf_asize = (unsigned)data_bytes;
345 ilf->ilf_size++;
339a5f5d 346 } else {
f5d8d5c4 347 iip->ili_fields &= ~XFS_ILOG_ADATA;
1da177e4
LT
348 }
349 break;
1da177e4
LT
350 default:
351 ASSERT(0);
352 break;
353 }
3de559fb
CH
354}
355
356/*
357 * This is called to fill in the vector of log iovecs for the given inode
358 * log item. It fills the first item with an inode log format structure,
359 * the second with the on-disk inode structure, and a possible third and/or
360 * fourth with the inode data/extents/b-tree root and inode attributes
361 * data/extents/b-tree root.
362 */
363STATIC void
364xfs_inode_item_format(
365 struct xfs_log_item *lip,
bde7cff6 366 struct xfs_log_vec *lv)
3de559fb
CH
367{
368 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
369 struct xfs_inode *ip = iip->ili_inode;
bde7cff6
CH
370 struct xfs_inode_log_format *ilf;
371 struct xfs_log_iovec *vecp = NULL;
3de559fb 372
2f251293
CH
373 ilf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_IFORMAT);
374 ilf->ilf_type = XFS_LI_INODE;
375 ilf->ilf_ino = ip->i_ino;
376 ilf->ilf_blkno = ip->i_imap.im_blkno;
377 ilf->ilf_len = ip->i_imap.im_len;
378 ilf->ilf_boffset = ip->i_imap.im_boffset;
379 ilf->ilf_fields = XFS_ILOG_CORE;
380 ilf->ilf_size = 2; /* format + core */
381 xlog_finish_iovec(lv, vecp, sizeof(struct xfs_inode_log_format));
3de559fb
CH
382
383 if (ip->i_d.di_version == 1)
384 xfs_inode_item_format_v1_inode(ip);
bde7cff6
CH
385 xlog_copy_iovec(lv, &vecp, XLOG_REG_TYPE_ICORE,
386 &ip->i_d,
387 xfs_icdinode_size(ip->i_d.di_version));
3de559fb 388
bde7cff6 389 xfs_inode_item_format_data_fork(iip, ilf, lv, &vecp);
3de559fb 390 if (XFS_IFORK_Q(ip)) {
bde7cff6 391 xfs_inode_item_format_attr_fork(iip, ilf, lv, &vecp);
3de559fb
CH
392 } else {
393 iip->ili_fields &=
394 ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT);
395 }
396
2f251293
CH
397 /* update the format with the exact fields we actually logged */
398 ilf->ilf_fields |= (iip->ili_fields & ~XFS_ILOG_TIMESTAMP);
1da177e4
LT
399}
400
1da177e4
LT
401/*
402 * This is called to pin the inode associated with the inode log
a14a5ab5 403 * item in memory so it cannot be written out.
1da177e4
LT
404 */
405STATIC void
406xfs_inode_item_pin(
7bfa31d8 407 struct xfs_log_item *lip)
1da177e4 408{
7bfa31d8 409 struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode;
a14a5ab5 410
7bfa31d8
CH
411 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
412
413 trace_xfs_inode_pin(ip, _RET_IP_);
414 atomic_inc(&ip->i_pincount);
1da177e4
LT
415}
416
417
418/*
419 * This is called to unpin the inode associated with the inode log
420 * item which was previously pinned with a call to xfs_inode_item_pin().
a14a5ab5
CH
421 *
422 * Also wake up anyone in xfs_iunpin_wait() if the count goes to 0.
1da177e4 423 */
1da177e4
LT
424STATIC void
425xfs_inode_item_unpin(
7bfa31d8 426 struct xfs_log_item *lip,
9412e318 427 int remove)
1da177e4 428{
7bfa31d8 429 struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode;
a14a5ab5 430
4aaf15d1 431 trace_xfs_inode_unpin(ip, _RET_IP_);
a14a5ab5
CH
432 ASSERT(atomic_read(&ip->i_pincount) > 0);
433 if (atomic_dec_and_test(&ip->i_pincount))
f392e631 434 wake_up_bit(&ip->i_flags, __XFS_IPINNED_BIT);
1da177e4
LT
435}
436
1da177e4 437STATIC uint
43ff2122
CH
438xfs_inode_item_push(
439 struct xfs_log_item *lip,
440 struct list_head *buffer_list)
1da177e4 441{
7bfa31d8
CH
442 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
443 struct xfs_inode *ip = iip->ili_inode;
43ff2122
CH
444 struct xfs_buf *bp = NULL;
445 uint rval = XFS_ITEM_SUCCESS;
446 int error;
1da177e4 447
7bfa31d8 448 if (xfs_ipincount(ip) > 0)
1da177e4 449 return XFS_ITEM_PINNED;
1da177e4 450
7bfa31d8 451 if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED))
1da177e4 452 return XFS_ITEM_LOCKED;
1da177e4 453
4c46819a
CH
454 /*
455 * Re-check the pincount now that we stabilized the value by
456 * taking the ilock.
457 */
458 if (xfs_ipincount(ip) > 0) {
43ff2122
CH
459 rval = XFS_ITEM_PINNED;
460 goto out_unlock;
4c46819a
CH
461 }
462
9a3a5dab
BF
463 /*
464 * Stale inode items should force out the iclog.
465 */
466 if (ip->i_flags & XFS_ISTALE) {
467 rval = XFS_ITEM_PINNED;
468 goto out_unlock;
469 }
470
43ff2122
CH
471 /*
472 * Someone else is already flushing the inode. Nothing we can do
473 * here but wait for the flush to finish and remove the item from
474 * the AIL.
475 */
1da177e4 476 if (!xfs_iflock_nowait(ip)) {
43ff2122
CH
477 rval = XFS_ITEM_FLUSHING;
478 goto out_unlock;
1da177e4
LT
479 }
480
43ff2122
CH
481 ASSERT(iip->ili_fields != 0 || XFS_FORCED_SHUTDOWN(ip->i_mount));
482 ASSERT(iip->ili_logged == 0 || XFS_FORCED_SHUTDOWN(ip->i_mount));
483
484 spin_unlock(&lip->li_ailp->xa_lock);
485
486 error = xfs_iflush(ip, &bp);
487 if (!error) {
488 if (!xfs_buf_delwri_queue(bp, buffer_list))
489 rval = XFS_ITEM_FLUSHING;
490 xfs_buf_relse(bp);
1da177e4 491 }
43ff2122
CH
492
493 spin_lock(&lip->li_ailp->xa_lock);
494out_unlock:
495 xfs_iunlock(ip, XFS_ILOCK_SHARED);
496 return rval;
1da177e4
LT
497}
498
499/*
500 * Unlock the inode associated with the inode log item.
501 * Clear the fields of the inode and inode log item that
502 * are specific to the current transaction. If the
503 * hold flags is set, do not unlock the inode.
504 */
505STATIC void
506xfs_inode_item_unlock(
7bfa31d8 507 struct xfs_log_item *lip)
1da177e4 508{
7bfa31d8
CH
509 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
510 struct xfs_inode *ip = iip->ili_inode;
898621d5 511 unsigned short lock_flags;
1da177e4 512
f3ca8738
CH
513 ASSERT(ip->i_itemp != NULL);
514 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
1da177e4 515
898621d5
CH
516 lock_flags = iip->ili_lock_flags;
517 iip->ili_lock_flags = 0;
ddc3415a 518 if (lock_flags)
f3ca8738 519 xfs_iunlock(ip, lock_flags);
1da177e4
LT
520}
521
522/*
de25c181
DC
523 * This is called to find out where the oldest active copy of the inode log
524 * item in the on disk log resides now that the last log write of it completed
525 * at the given lsn. Since we always re-log all dirty data in an inode, the
526 * latest copy in the on disk log is the only one that matters. Therefore,
527 * simply return the given lsn.
528 *
529 * If the inode has been marked stale because the cluster is being freed, we
530 * don't want to (re-)insert this inode into the AIL. There is a race condition
531 * where the cluster buffer may be unpinned before the inode is inserted into
532 * the AIL during transaction committed processing. If the buffer is unpinned
533 * before the inode item has been committed and inserted, then it is possible
1316d4da 534 * for the buffer to be written and IO completes before the inode is inserted
de25c181
DC
535 * into the AIL. In that case, we'd be inserting a clean, stale inode into the
536 * AIL which will never get removed. It will, however, get reclaimed which
537 * triggers an assert in xfs_inode_free() complaining about freein an inode
538 * still in the AIL.
539 *
1316d4da
DC
540 * To avoid this, just unpin the inode directly and return a LSN of -1 so the
541 * transaction committed code knows that it does not need to do any further
542 * processing on the item.
1da177e4 543 */
1da177e4
LT
544STATIC xfs_lsn_t
545xfs_inode_item_committed(
7bfa31d8 546 struct xfs_log_item *lip,
1da177e4
LT
547 xfs_lsn_t lsn)
548{
de25c181
DC
549 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
550 struct xfs_inode *ip = iip->ili_inode;
551
1316d4da
DC
552 if (xfs_iflags_test(ip, XFS_ISTALE)) {
553 xfs_inode_item_unpin(lip, 0);
554 return -1;
555 }
7bfa31d8 556 return lsn;
1da177e4
LT
557}
558
1da177e4
LT
559/*
560 * XXX rcc - this one really has to do something. Probably needs
561 * to stamp in a new field in the incore inode.
562 */
1da177e4
LT
563STATIC void
564xfs_inode_item_committing(
7bfa31d8 565 struct xfs_log_item *lip,
1da177e4
LT
566 xfs_lsn_t lsn)
567{
7bfa31d8 568 INODE_ITEM(lip)->ili_last_lsn = lsn;
1da177e4
LT
569}
570
571/*
572 * This is the ops vector shared by all buf log items.
573 */
272e42b2 574static const struct xfs_item_ops xfs_inode_item_ops = {
7bfa31d8
CH
575 .iop_size = xfs_inode_item_size,
576 .iop_format = xfs_inode_item_format,
577 .iop_pin = xfs_inode_item_pin,
578 .iop_unpin = xfs_inode_item_unpin,
7bfa31d8
CH
579 .iop_unlock = xfs_inode_item_unlock,
580 .iop_committed = xfs_inode_item_committed,
581 .iop_push = xfs_inode_item_push,
7bfa31d8 582 .iop_committing = xfs_inode_item_committing
1da177e4
LT
583};
584
585
586/*
587 * Initialize the inode log item for a newly allocated (in-core) inode.
588 */
589void
590xfs_inode_item_init(
7bfa31d8
CH
591 struct xfs_inode *ip,
592 struct xfs_mount *mp)
1da177e4 593{
7bfa31d8 594 struct xfs_inode_log_item *iip;
1da177e4
LT
595
596 ASSERT(ip->i_itemp == NULL);
597 iip = ip->i_itemp = kmem_zone_zalloc(xfs_ili_zone, KM_SLEEP);
598
1da177e4 599 iip->ili_inode = ip;
43f5efc5
DC
600 xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE,
601 &xfs_inode_item_ops);
1da177e4
LT
602}
603
604/*
605 * Free the inode log item and any memory hanging off of it.
606 */
607void
608xfs_inode_item_destroy(
609 xfs_inode_t *ip)
610{
1da177e4
LT
611 kmem_zone_free(xfs_ili_zone, ip->i_itemp);
612}
613
614
615/*
616 * This is the inode flushing I/O completion routine. It is called
617 * from interrupt level when the buffer containing the inode is
618 * flushed to disk. It is responsible for removing the inode item
619 * from the AIL if it has not been re-logged, and unlocking the inode's
620 * flush lock.
30136832
DC
621 *
622 * To reduce AIL lock traffic as much as possible, we scan the buffer log item
623 * list for other inodes that will run this function. We remove them from the
624 * buffer list so we can process all the inode IO completions in one AIL lock
625 * traversal.
1da177e4 626 */
1da177e4
LT
627void
628xfs_iflush_done(
ca30b2a7
CH
629 struct xfs_buf *bp,
630 struct xfs_log_item *lip)
1da177e4 631{
30136832
DC
632 struct xfs_inode_log_item *iip;
633 struct xfs_log_item *blip;
634 struct xfs_log_item *next;
635 struct xfs_log_item *prev;
ca30b2a7 636 struct xfs_ail *ailp = lip->li_ailp;
30136832
DC
637 int need_ail = 0;
638
639 /*
640 * Scan the buffer IO completions for other inodes being completed and
641 * attach them to the current inode log item.
642 */
adadbeef 643 blip = bp->b_fspriv;
30136832
DC
644 prev = NULL;
645 while (blip != NULL) {
646 if (lip->li_cb != xfs_iflush_done) {
647 prev = blip;
648 blip = blip->li_bio_list;
649 continue;
650 }
651
652 /* remove from list */
653 next = blip->li_bio_list;
654 if (!prev) {
adadbeef 655 bp->b_fspriv = next;
30136832
DC
656 } else {
657 prev->li_bio_list = next;
658 }
659
660 /* add to current list */
661 blip->li_bio_list = lip->li_bio_list;
662 lip->li_bio_list = blip;
663
664 /*
665 * while we have the item, do the unlocked check for needing
666 * the AIL lock.
667 */
668 iip = INODE_ITEM(blip);
669 if (iip->ili_logged && blip->li_lsn == iip->ili_flush_lsn)
670 need_ail++;
671
672 blip = next;
673 }
674
675 /* make sure we capture the state of the initial inode. */
676 iip = INODE_ITEM(lip);
677 if (iip->ili_logged && lip->li_lsn == iip->ili_flush_lsn)
678 need_ail++;
1da177e4
LT
679
680 /*
681 * We only want to pull the item from the AIL if it is
682 * actually there and its location in the log has not
683 * changed since we started the flush. Thus, we only bother
684 * if the ili_logged flag is set and the inode's lsn has not
685 * changed. First we check the lsn outside
686 * the lock since it's cheaper, and then we recheck while
687 * holding the lock before removing the inode from the AIL.
688 */
30136832
DC
689 if (need_ail) {
690 struct xfs_log_item *log_items[need_ail];
691 int i = 0;
783a2f65 692 spin_lock(&ailp->xa_lock);
30136832
DC
693 for (blip = lip; blip; blip = blip->li_bio_list) {
694 iip = INODE_ITEM(blip);
695 if (iip->ili_logged &&
696 blip->li_lsn == iip->ili_flush_lsn) {
697 log_items[i++] = blip;
698 }
699 ASSERT(i <= need_ail);
1da177e4 700 }
30136832 701 /* xfs_trans_ail_delete_bulk() drops the AIL lock. */
04913fdd
DC
702 xfs_trans_ail_delete_bulk(ailp, log_items, i,
703 SHUTDOWN_CORRUPT_INCORE);
1da177e4
LT
704 }
705
1da177e4
LT
706
707 /*
30136832
DC
708 * clean up and unlock the flush lock now we are done. We can clear the
709 * ili_last_fields bits now that we know that the data corresponding to
710 * them is safely on disk.
1da177e4 711 */
30136832
DC
712 for (blip = lip; blip; blip = next) {
713 next = blip->li_bio_list;
714 blip->li_bio_list = NULL;
715
716 iip = INODE_ITEM(blip);
717 iip->ili_logged = 0;
718 iip->ili_last_fields = 0;
719 xfs_ifunlock(iip->ili_inode);
720 }
1da177e4
LT
721}
722
723/*
04913fdd
DC
724 * This is the inode flushing abort routine. It is called from xfs_iflush when
725 * the filesystem is shutting down to clean up the inode state. It is
726 * responsible for removing the inode item from the AIL if it has not been
727 * re-logged, and unlocking the inode's flush lock.
1da177e4
LT
728 */
729void
730xfs_iflush_abort(
04913fdd
DC
731 xfs_inode_t *ip,
732 bool stale)
1da177e4 733{
783a2f65 734 xfs_inode_log_item_t *iip = ip->i_itemp;
1da177e4 735
1da177e4 736 if (iip) {
783a2f65 737 struct xfs_ail *ailp = iip->ili_item.li_ailp;
1da177e4 738 if (iip->ili_item.li_flags & XFS_LI_IN_AIL) {
783a2f65 739 spin_lock(&ailp->xa_lock);
1da177e4 740 if (iip->ili_item.li_flags & XFS_LI_IN_AIL) {
783a2f65 741 /* xfs_trans_ail_delete() drops the AIL lock. */
04913fdd
DC
742 xfs_trans_ail_delete(ailp, &iip->ili_item,
743 stale ?
744 SHUTDOWN_LOG_IO_ERROR :
745 SHUTDOWN_CORRUPT_INCORE);
1da177e4 746 } else
783a2f65 747 spin_unlock(&ailp->xa_lock);
1da177e4
LT
748 }
749 iip->ili_logged = 0;
750 /*
751 * Clear the ili_last_fields bits now that we know that the
752 * data corresponding to them is safely on disk.
753 */
754 iip->ili_last_fields = 0;
755 /*
756 * Clear the inode logging fields so no more flushes are
757 * attempted.
758 */
f5d8d5c4 759 iip->ili_fields = 0;
1da177e4
LT
760 }
761 /*
762 * Release the inode's flush lock since we're done with it.
763 */
764 xfs_ifunlock(ip);
765}
766
767void
768xfs_istale_done(
ca30b2a7
CH
769 struct xfs_buf *bp,
770 struct xfs_log_item *lip)
1da177e4 771{
04913fdd 772 xfs_iflush_abort(INODE_ITEM(lip)->ili_inode, true);
1da177e4 773}
6d192a9b
TS
774
775/*
776 * convert an xfs_inode_log_format struct from either 32 or 64 bit versions
777 * (which can have different field alignments) to the native version
778 */
779int
780xfs_inode_item_format_convert(
781 xfs_log_iovec_t *buf,
782 xfs_inode_log_format_t *in_f)
783{
784 if (buf->i_len == sizeof(xfs_inode_log_format_32_t)) {
4e0d5f92 785 xfs_inode_log_format_32_t *in_f32 = buf->i_addr;
6d192a9b 786
6d192a9b
TS
787 in_f->ilf_type = in_f32->ilf_type;
788 in_f->ilf_size = in_f32->ilf_size;
789 in_f->ilf_fields = in_f32->ilf_fields;
790 in_f->ilf_asize = in_f32->ilf_asize;
791 in_f->ilf_dsize = in_f32->ilf_dsize;
792 in_f->ilf_ino = in_f32->ilf_ino;
793 /* copy biggest field of ilf_u */
794 memcpy(in_f->ilf_u.ilfu_uuid.__u_bits,
795 in_f32->ilf_u.ilfu_uuid.__u_bits,
796 sizeof(uuid_t));
797 in_f->ilf_blkno = in_f32->ilf_blkno;
798 in_f->ilf_len = in_f32->ilf_len;
799 in_f->ilf_boffset = in_f32->ilf_boffset;
800 return 0;
801 } else if (buf->i_len == sizeof(xfs_inode_log_format_64_t)){
4e0d5f92 802 xfs_inode_log_format_64_t *in_f64 = buf->i_addr;
6d192a9b 803
6d192a9b
TS
804 in_f->ilf_type = in_f64->ilf_type;
805 in_f->ilf_size = in_f64->ilf_size;
806 in_f->ilf_fields = in_f64->ilf_fields;
807 in_f->ilf_asize = in_f64->ilf_asize;
808 in_f->ilf_dsize = in_f64->ilf_dsize;
809 in_f->ilf_ino = in_f64->ilf_ino;
810 /* copy biggest field of ilf_u */
811 memcpy(in_f->ilf_u.ilfu_uuid.__u_bits,
812 in_f64->ilf_u.ilfu_uuid.__u_bits,
813 sizeof(uuid_t));
814 in_f->ilf_blkno = in_f64->ilf_blkno;
815 in_f->ilf_len = in_f64->ilf_len;
816 in_f->ilf_boffset = in_f64->ilf_boffset;
817 return 0;
818 }
819 return EFSCORRUPTED;
820}