1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc.
8 #include "xfs_shared.h"
9 #include "xfs_format.h"
10 #include "xfs_log_format.h"
11 #include "xfs_trans_resv.h"
12 #include "xfs_mount.h"
13 #include "xfs_inode.h"
14 #include "xfs_trans.h"
15 #include "xfs_inode_item.h"
16 #include "xfs_trace.h"
17 #include "xfs_trans_priv.h"
18 #include "xfs_buf_item.h"
20 #include "xfs_error.h"
22 #include <linux/iversion.h>
24 kmem_zone_t *xfs_ili_zone; /* inode log item zone */
26 static inline struct xfs_inode_log_item *INODE_ITEM(struct xfs_log_item *lip)
28 return container_of(lip, struct xfs_inode_log_item, ili_item);
32 xfs_inode_item_data_fork_size(
33 struct xfs_inode_log_item *iip,
37 struct xfs_inode *ip = iip->ili_inode;
39 switch (ip->i_df.if_format) {
40 case XFS_DINODE_FMT_EXTENTS:
41 if ((iip->ili_fields & XFS_ILOG_DEXT) &&
42 ip->i_df.if_nextents > 0 &&
43 ip->i_df.if_bytes > 0) {
44 /* worst case, doesn't subtract delalloc extents */
45 *nbytes += XFS_IFORK_DSIZE(ip);
49 case XFS_DINODE_FMT_BTREE:
50 if ((iip->ili_fields & XFS_ILOG_DBROOT) &&
51 ip->i_df.if_broot_bytes > 0) {
52 *nbytes += ip->i_df.if_broot_bytes;
56 case XFS_DINODE_FMT_LOCAL:
57 if ((iip->ili_fields & XFS_ILOG_DDATA) &&
58 ip->i_df.if_bytes > 0) {
59 *nbytes += roundup(ip->i_df.if_bytes, 4);
64 case XFS_DINODE_FMT_DEV:
73 xfs_inode_item_attr_fork_size(
74 struct xfs_inode_log_item *iip,
78 struct xfs_inode *ip = iip->ili_inode;
80 switch (ip->i_afp->if_format) {
81 case XFS_DINODE_FMT_EXTENTS:
82 if ((iip->ili_fields & XFS_ILOG_AEXT) &&
83 ip->i_afp->if_nextents > 0 &&
84 ip->i_afp->if_bytes > 0) {
85 /* worst case, doesn't subtract unused space */
86 *nbytes += XFS_IFORK_ASIZE(ip);
90 case XFS_DINODE_FMT_BTREE:
91 if ((iip->ili_fields & XFS_ILOG_ABROOT) &&
92 ip->i_afp->if_broot_bytes > 0) {
93 *nbytes += ip->i_afp->if_broot_bytes;
97 case XFS_DINODE_FMT_LOCAL:
98 if ((iip->ili_fields & XFS_ILOG_ADATA) &&
99 ip->i_afp->if_bytes > 0) {
100 *nbytes += roundup(ip->i_afp->if_bytes, 4);
111 * This returns the number of iovecs needed to log the given inode item.
113 * We need one iovec for the inode log format structure, one for the
114 * inode core, and possibly one for the inode data/extents/b-tree root
115 * and one for the inode attribute data/extents/b-tree root.
119 struct xfs_log_item *lip,
123 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
124 struct xfs_inode *ip = iip->ili_inode;
127 *nbytes += sizeof(struct xfs_inode_log_format) +
128 xfs_log_dinode_size(ip->i_mount);
130 xfs_inode_item_data_fork_size(iip, nvecs, nbytes);
132 xfs_inode_item_attr_fork_size(iip, nvecs, nbytes);
136 xfs_inode_item_format_data_fork(
137 struct xfs_inode_log_item *iip,
138 struct xfs_inode_log_format *ilf,
139 struct xfs_log_vec *lv,
140 struct xfs_log_iovec **vecp)
142 struct xfs_inode *ip = iip->ili_inode;
145 switch (ip->i_df.if_format) {
146 case XFS_DINODE_FMT_EXTENTS:
148 ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | XFS_ILOG_DEV);
150 if ((iip->ili_fields & XFS_ILOG_DEXT) &&
151 ip->i_df.if_nextents > 0 &&
152 ip->i_df.if_bytes > 0) {
153 struct xfs_bmbt_rec *p;
155 ASSERT(xfs_iext_count(&ip->i_df) > 0);
157 p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IEXT);
158 data_bytes = xfs_iextents_copy(ip, p, XFS_DATA_FORK);
159 xlog_finish_iovec(lv, *vecp, data_bytes);
161 ASSERT(data_bytes <= ip->i_df.if_bytes);
163 ilf->ilf_dsize = data_bytes;
166 iip->ili_fields &= ~XFS_ILOG_DEXT;
169 case XFS_DINODE_FMT_BTREE:
171 ~(XFS_ILOG_DDATA | XFS_ILOG_DEXT | XFS_ILOG_DEV);
173 if ((iip->ili_fields & XFS_ILOG_DBROOT) &&
174 ip->i_df.if_broot_bytes > 0) {
175 ASSERT(ip->i_df.if_broot != NULL);
176 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IBROOT,
178 ip->i_df.if_broot_bytes);
179 ilf->ilf_dsize = ip->i_df.if_broot_bytes;
182 ASSERT(!(iip->ili_fields &
184 iip->ili_fields &= ~XFS_ILOG_DBROOT;
187 case XFS_DINODE_FMT_LOCAL:
189 ~(XFS_ILOG_DEXT | XFS_ILOG_DBROOT | XFS_ILOG_DEV);
190 if ((iip->ili_fields & XFS_ILOG_DDATA) &&
191 ip->i_df.if_bytes > 0) {
193 * Round i_bytes up to a word boundary.
194 * The underlying memory is guaranteed
195 * to be there by xfs_idata_realloc().
197 data_bytes = roundup(ip->i_df.if_bytes, 4);
198 ASSERT(ip->i_df.if_u1.if_data != NULL);
199 ASSERT(ip->i_disk_size > 0);
200 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_ILOCAL,
201 ip->i_df.if_u1.if_data, data_bytes);
202 ilf->ilf_dsize = (unsigned)data_bytes;
205 iip->ili_fields &= ~XFS_ILOG_DDATA;
208 case XFS_DINODE_FMT_DEV:
210 ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | XFS_ILOG_DEXT);
211 if (iip->ili_fields & XFS_ILOG_DEV)
212 ilf->ilf_u.ilfu_rdev = sysv_encode_dev(VFS_I(ip)->i_rdev);
221 xfs_inode_item_format_attr_fork(
222 struct xfs_inode_log_item *iip,
223 struct xfs_inode_log_format *ilf,
224 struct xfs_log_vec *lv,
225 struct xfs_log_iovec **vecp)
227 struct xfs_inode *ip = iip->ili_inode;
230 switch (ip->i_afp->if_format) {
231 case XFS_DINODE_FMT_EXTENTS:
233 ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT);
235 if ((iip->ili_fields & XFS_ILOG_AEXT) &&
236 ip->i_afp->if_nextents > 0 &&
237 ip->i_afp->if_bytes > 0) {
238 struct xfs_bmbt_rec *p;
240 ASSERT(xfs_iext_count(ip->i_afp) ==
241 ip->i_afp->if_nextents);
243 p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_EXT);
244 data_bytes = xfs_iextents_copy(ip, p, XFS_ATTR_FORK);
245 xlog_finish_iovec(lv, *vecp, data_bytes);
247 ilf->ilf_asize = data_bytes;
250 iip->ili_fields &= ~XFS_ILOG_AEXT;
253 case XFS_DINODE_FMT_BTREE:
255 ~(XFS_ILOG_ADATA | XFS_ILOG_AEXT);
257 if ((iip->ili_fields & XFS_ILOG_ABROOT) &&
258 ip->i_afp->if_broot_bytes > 0) {
259 ASSERT(ip->i_afp->if_broot != NULL);
261 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_BROOT,
263 ip->i_afp->if_broot_bytes);
264 ilf->ilf_asize = ip->i_afp->if_broot_bytes;
267 iip->ili_fields &= ~XFS_ILOG_ABROOT;
270 case XFS_DINODE_FMT_LOCAL:
272 ~(XFS_ILOG_AEXT | XFS_ILOG_ABROOT);
274 if ((iip->ili_fields & XFS_ILOG_ADATA) &&
275 ip->i_afp->if_bytes > 0) {
277 * Round i_bytes up to a word boundary.
278 * The underlying memory is guaranteed
279 * to be there by xfs_idata_realloc().
281 data_bytes = roundup(ip->i_afp->if_bytes, 4);
282 ASSERT(ip->i_afp->if_u1.if_data != NULL);
283 xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_LOCAL,
284 ip->i_afp->if_u1.if_data,
286 ilf->ilf_asize = (unsigned)data_bytes;
289 iip->ili_fields &= ~XFS_ILOG_ADATA;
299 * Convert an incore timestamp to a log timestamp. Note that the log format
300 * specifies host endian format!
302 static inline xfs_log_timestamp_t
303 xfs_inode_to_log_dinode_ts(
304 struct xfs_inode *ip,
305 const struct timespec64 tv)
307 struct xfs_log_legacy_timestamp *lits;
308 xfs_log_timestamp_t its;
310 if (xfs_inode_has_bigtime(ip))
311 return xfs_inode_encode_bigtime(tv);
313 lits = (struct xfs_log_legacy_timestamp *)&its;
314 lits->t_sec = tv.tv_sec;
315 lits->t_nsec = tv.tv_nsec;
321 * The legacy DMAPI fields are only present in the on-disk and in-log inodes,
322 * but not in the in-memory one. But we are guaranteed to have an inode buffer
323 * in memory when logging an inode, so we can just copy it from the on-disk
324 * inode to the in-log inode here so that recovery of file system with these
325 * fields set to non-zero values doesn't lose them. For all other cases we zero
329 xfs_copy_dm_fields_to_log_dinode(
330 struct xfs_inode *ip,
331 struct xfs_log_dinode *to)
333 struct xfs_dinode *dip;
335 dip = xfs_buf_offset(ip->i_itemp->ili_item.li_buf,
336 ip->i_imap.im_boffset);
338 if (xfs_iflags_test(ip, XFS_IPRESERVE_DM_FIELDS)) {
339 to->di_dmevmask = be32_to_cpu(dip->di_dmevmask);
340 to->di_dmstate = be16_to_cpu(dip->di_dmstate);
348 xfs_inode_to_log_dinode(
349 struct xfs_inode *ip,
350 struct xfs_log_dinode *to,
353 struct inode *inode = VFS_I(ip);
355 to->di_magic = XFS_DINODE_MAGIC;
356 to->di_format = xfs_ifork_format(&ip->i_df);
357 to->di_uid = i_uid_read(inode);
358 to->di_gid = i_gid_read(inode);
359 to->di_projid_lo = ip->i_projid & 0xffff;
360 to->di_projid_hi = ip->i_projid >> 16;
362 memset(to->di_pad, 0, sizeof(to->di_pad));
363 memset(to->di_pad3, 0, sizeof(to->di_pad3));
364 to->di_atime = xfs_inode_to_log_dinode_ts(ip, inode->i_atime);
365 to->di_mtime = xfs_inode_to_log_dinode_ts(ip, inode->i_mtime);
366 to->di_ctime = xfs_inode_to_log_dinode_ts(ip, inode->i_ctime);
367 to->di_nlink = inode->i_nlink;
368 to->di_gen = inode->i_generation;
369 to->di_mode = inode->i_mode;
371 to->di_size = ip->i_disk_size;
372 to->di_nblocks = ip->i_nblocks;
373 to->di_extsize = ip->i_extsize;
374 to->di_nextents = xfs_ifork_nextents(&ip->i_df);
375 to->di_anextents = xfs_ifork_nextents(ip->i_afp);
376 to->di_forkoff = ip->i_forkoff;
377 to->di_aformat = xfs_ifork_format(ip->i_afp);
378 to->di_flags = ip->i_diflags;
380 xfs_copy_dm_fields_to_log_dinode(ip, to);
382 /* log a dummy value to ensure log structure is fully initialised */
383 to->di_next_unlinked = NULLAGINO;
385 if (xfs_sb_version_has_v3inode(&ip->i_mount->m_sb)) {
387 to->di_changecount = inode_peek_iversion(inode);
388 to->di_crtime = xfs_inode_to_log_dinode_ts(ip, ip->i_crtime);
389 to->di_flags2 = ip->i_diflags2;
390 to->di_cowextsize = ip->i_cowextsize;
391 to->di_ino = ip->i_ino;
393 memset(to->di_pad2, 0, sizeof(to->di_pad2));
394 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
395 to->di_flushiter = 0;
398 to->di_flushiter = ip->i_flushiter;
403 * Format the inode core. Current timestamp data is only in the VFS inode
404 * fields, so we need to grab them from there. Hence rather than just copying
405 * the XFS inode core structure, format the fields directly into the iovec.
408 xfs_inode_item_format_core(
409 struct xfs_inode *ip,
410 struct xfs_log_vec *lv,
411 struct xfs_log_iovec **vecp)
413 struct xfs_log_dinode *dic;
415 dic = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_ICORE);
416 xfs_inode_to_log_dinode(ip, dic, ip->i_itemp->ili_item.li_lsn);
417 xlog_finish_iovec(lv, *vecp, xfs_log_dinode_size(ip->i_mount));
421 * This is called to fill in the vector of log iovecs for the given inode
422 * log item. It fills the first item with an inode log format structure,
423 * the second with the on-disk inode structure, and a possible third and/or
424 * fourth with the inode data/extents/b-tree root and inode attributes
425 * data/extents/b-tree root.
427 * Note: Always use the 64 bit inode log format structure so we don't
428 * leave an uninitialised hole in the format item on 64 bit systems. Log
429 * recovery on 32 bit systems handles this just fine, so there's no reason
430 * for not using an initialising the properly padded structure all the time.
433 xfs_inode_item_format(
434 struct xfs_log_item *lip,
435 struct xfs_log_vec *lv)
437 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
438 struct xfs_inode *ip = iip->ili_inode;
439 struct xfs_log_iovec *vecp = NULL;
440 struct xfs_inode_log_format *ilf;
442 ilf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_IFORMAT);
443 ilf->ilf_type = XFS_LI_INODE;
444 ilf->ilf_ino = ip->i_ino;
445 ilf->ilf_blkno = ip->i_imap.im_blkno;
446 ilf->ilf_len = ip->i_imap.im_len;
447 ilf->ilf_boffset = ip->i_imap.im_boffset;
448 ilf->ilf_fields = XFS_ILOG_CORE;
449 ilf->ilf_size = 2; /* format + core */
452 * make sure we don't leak uninitialised data into the log in the case
453 * when we don't log every field in the inode.
458 memset(&ilf->ilf_u, 0, sizeof(ilf->ilf_u));
460 xlog_finish_iovec(lv, vecp, sizeof(*ilf));
462 xfs_inode_item_format_core(ip, lv, &vecp);
463 xfs_inode_item_format_data_fork(iip, ilf, lv, &vecp);
464 if (XFS_IFORK_Q(ip)) {
465 xfs_inode_item_format_attr_fork(iip, ilf, lv, &vecp);
468 ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT);
471 /* update the format with the exact fields we actually logged */
472 ilf->ilf_fields |= (iip->ili_fields & ~XFS_ILOG_TIMESTAMP);
476 * This is called to pin the inode associated with the inode log
477 * item in memory so it cannot be written out.
481 struct xfs_log_item *lip)
483 struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode;
485 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
488 trace_xfs_inode_pin(ip, _RET_IP_);
489 atomic_inc(&ip->i_pincount);
494 * This is called to unpin the inode associated with the inode log
495 * item which was previously pinned with a call to xfs_inode_item_pin().
497 * Also wake up anyone in xfs_iunpin_wait() if the count goes to 0.
499 * Note that unpin can race with inode cluster buffer freeing marking the buffer
500 * stale. In that case, flush completions are run from the buffer unpin call,
501 * which may happen before the inode is unpinned. If we lose the race, there
502 * will be no buffer attached to the log item, but the inode will be marked
506 xfs_inode_item_unpin(
507 struct xfs_log_item *lip,
510 struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode;
512 trace_xfs_inode_unpin(ip, _RET_IP_);
513 ASSERT(lip->li_buf || xfs_iflags_test(ip, XFS_ISTALE));
514 ASSERT(atomic_read(&ip->i_pincount) > 0);
515 if (atomic_dec_and_test(&ip->i_pincount))
516 wake_up_bit(&ip->i_flags, __XFS_IPINNED_BIT);
521 struct xfs_log_item *lip,
522 struct list_head *buffer_list)
523 __releases(&lip->li_ailp->ail_lock)
524 __acquires(&lip->li_ailp->ail_lock)
526 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
527 struct xfs_inode *ip = iip->ili_inode;
528 struct xfs_buf *bp = lip->li_buf;
529 uint rval = XFS_ITEM_SUCCESS;
532 ASSERT(iip->ili_item.li_buf);
534 if (xfs_ipincount(ip) > 0 || xfs_buf_ispinned(bp) ||
535 (ip->i_flags & XFS_ISTALE))
536 return XFS_ITEM_PINNED;
538 if (xfs_iflags_test(ip, XFS_IFLUSHING))
539 return XFS_ITEM_FLUSHING;
541 if (!xfs_buf_trylock(bp))
542 return XFS_ITEM_LOCKED;
544 spin_unlock(&lip->li_ailp->ail_lock);
547 * We need to hold a reference for flushing the cluster buffer as it may
548 * fail the buffer without IO submission. In which case, we better get a
549 * reference for that completion because otherwise we don't get a
550 * reference for IO until we queue the buffer for delwri submission.
553 error = xfs_iflush_cluster(bp);
555 if (!xfs_buf_delwri_queue(bp, buffer_list))
556 rval = XFS_ITEM_FLUSHING;
560 * Release the buffer if we were unable to flush anything. On
561 * any other error, the buffer has already been released.
563 if (error == -EAGAIN)
565 rval = XFS_ITEM_LOCKED;
568 spin_lock(&lip->li_ailp->ail_lock);
573 * Unlock the inode associated with the inode log item.
576 xfs_inode_item_release(
577 struct xfs_log_item *lip)
579 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
580 struct xfs_inode *ip = iip->ili_inode;
581 unsigned short lock_flags;
583 ASSERT(ip->i_itemp != NULL);
584 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
586 lock_flags = iip->ili_lock_flags;
587 iip->ili_lock_flags = 0;
589 xfs_iunlock(ip, lock_flags);
593 * This is called to find out where the oldest active copy of the inode log
594 * item in the on disk log resides now that the last log write of it completed
595 * at the given lsn. Since we always re-log all dirty data in an inode, the
596 * latest copy in the on disk log is the only one that matters. Therefore,
597 * simply return the given lsn.
599 * If the inode has been marked stale because the cluster is being freed, we
600 * don't want to (re-)insert this inode into the AIL. There is a race condition
601 * where the cluster buffer may be unpinned before the inode is inserted into
602 * the AIL during transaction committed processing. If the buffer is unpinned
603 * before the inode item has been committed and inserted, then it is possible
604 * for the buffer to be written and IO completes before the inode is inserted
605 * into the AIL. In that case, we'd be inserting a clean, stale inode into the
606 * AIL which will never get removed. It will, however, get reclaimed which
607 * triggers an assert in xfs_inode_free() complaining about freein an inode
610 * To avoid this, just unpin the inode directly and return a LSN of -1 so the
611 * transaction committed code knows that it does not need to do any further
612 * processing on the item.
615 xfs_inode_item_committed(
616 struct xfs_log_item *lip,
619 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
620 struct xfs_inode *ip = iip->ili_inode;
622 if (xfs_iflags_test(ip, XFS_ISTALE)) {
623 xfs_inode_item_unpin(lip, 0);
630 xfs_inode_item_committing(
631 struct xfs_log_item *lip,
632 xfs_lsn_t commit_lsn)
634 INODE_ITEM(lip)->ili_last_lsn = commit_lsn;
635 return xfs_inode_item_release(lip);
638 static const struct xfs_item_ops xfs_inode_item_ops = {
639 .iop_size = xfs_inode_item_size,
640 .iop_format = xfs_inode_item_format,
641 .iop_pin = xfs_inode_item_pin,
642 .iop_unpin = xfs_inode_item_unpin,
643 .iop_release = xfs_inode_item_release,
644 .iop_committed = xfs_inode_item_committed,
645 .iop_push = xfs_inode_item_push,
646 .iop_committing = xfs_inode_item_committing,
651 * Initialize the inode log item for a newly allocated (in-core) inode.
655 struct xfs_inode *ip,
656 struct xfs_mount *mp)
658 struct xfs_inode_log_item *iip;
660 ASSERT(ip->i_itemp == NULL);
661 iip = ip->i_itemp = kmem_cache_zalloc(xfs_ili_zone,
662 GFP_KERNEL | __GFP_NOFAIL);
665 spin_lock_init(&iip->ili_lock);
666 xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE,
667 &xfs_inode_item_ops);
671 * Free the inode log item and any memory hanging off of it.
674 xfs_inode_item_destroy(
675 struct xfs_inode *ip)
677 struct xfs_inode_log_item *iip = ip->i_itemp;
679 ASSERT(iip->ili_item.li_buf == NULL);
682 kmem_free(iip->ili_item.li_lv_shadow);
683 kmem_cache_free(xfs_ili_zone, iip);
688 * We only want to pull the item from the AIL if it is actually there
689 * and its location in the log has not changed since we started the
690 * flush. Thus, we only bother if the inode's lsn has not changed.
693 xfs_iflush_ail_updates(
694 struct xfs_ail *ailp,
695 struct list_head *list)
697 struct xfs_log_item *lip;
698 xfs_lsn_t tail_lsn = 0;
700 /* this is an opencoded batch version of xfs_trans_ail_delete */
701 spin_lock(&ailp->ail_lock);
702 list_for_each_entry(lip, list, li_bio_list) {
705 clear_bit(XFS_LI_FAILED, &lip->li_flags);
706 if (INODE_ITEM(lip)->ili_flush_lsn != lip->li_lsn)
709 lsn = xfs_ail_delete_one(ailp, lip);
710 if (!tail_lsn && lsn)
713 xfs_ail_update_finish(ailp, tail_lsn);
717 * Walk the list of inodes that have completed their IOs. If they are clean
718 * remove them from the list and dissociate them from the buffer. Buffers that
719 * are still dirty remain linked to the buffer and on the list. Caller must
720 * handle them appropriately.
725 struct list_head *list)
727 struct xfs_log_item *lip, *n;
729 list_for_each_entry_safe(lip, n, list, li_bio_list) {
730 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
731 bool drop_buffer = false;
733 spin_lock(&iip->ili_lock);
736 * Remove the reference to the cluster buffer if the inode is
737 * clean in memory and drop the buffer reference once we've
738 * dropped the locks we hold.
740 ASSERT(iip->ili_item.li_buf == bp);
741 if (!iip->ili_fields) {
742 iip->ili_item.li_buf = NULL;
743 list_del_init(&lip->li_bio_list);
746 iip->ili_last_fields = 0;
747 iip->ili_flush_lsn = 0;
748 spin_unlock(&iip->ili_lock);
749 xfs_iflags_clear(iip->ili_inode, XFS_IFLUSHING);
756 * Inode buffer IO completion routine. It is responsible for removing inodes
757 * attached to the buffer from the AIL if they have not been re-logged and
758 * completing the inode flush.
761 xfs_buf_inode_iodone(
764 struct xfs_log_item *lip, *n;
765 LIST_HEAD(flushed_inodes);
766 LIST_HEAD(ail_updates);
769 * Pull the attached inodes from the buffer one at a time and take the
770 * appropriate action on them.
772 list_for_each_entry_safe(lip, n, &bp->b_li_list, li_bio_list) {
773 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
775 if (xfs_iflags_test(iip->ili_inode, XFS_ISTALE)) {
776 xfs_iflush_abort(iip->ili_inode);
779 if (!iip->ili_last_fields)
782 /* Do an unlocked check for needing the AIL lock. */
783 if (iip->ili_flush_lsn == lip->li_lsn ||
784 test_bit(XFS_LI_FAILED, &lip->li_flags))
785 list_move_tail(&lip->li_bio_list, &ail_updates);
787 list_move_tail(&lip->li_bio_list, &flushed_inodes);
790 if (!list_empty(&ail_updates)) {
791 xfs_iflush_ail_updates(bp->b_mount->m_ail, &ail_updates);
792 list_splice_tail(&ail_updates, &flushed_inodes);
795 xfs_iflush_finish(bp, &flushed_inodes);
796 if (!list_empty(&flushed_inodes))
797 list_splice_tail(&flushed_inodes, &bp->b_li_list);
801 xfs_buf_inode_io_fail(
804 struct xfs_log_item *lip;
806 list_for_each_entry(lip, &bp->b_li_list, li_bio_list)
807 set_bit(XFS_LI_FAILED, &lip->li_flags);
811 * This is the inode flushing abort routine. It is called when
812 * the filesystem is shutting down to clean up the inode state. It is
813 * responsible for removing the inode item from the AIL if it has not been
814 * re-logged and clearing the inode's flush state.
818 struct xfs_inode *ip)
820 struct xfs_inode_log_item *iip = ip->i_itemp;
821 struct xfs_buf *bp = NULL;
825 * Clear the failed bit before removing the item from the AIL so
826 * xfs_trans_ail_delete() doesn't try to clear and release the
827 * buffer attached to the log item before we are done with it.
829 clear_bit(XFS_LI_FAILED, &iip->ili_item.li_flags);
830 xfs_trans_ail_delete(&iip->ili_item, 0);
833 * Clear the inode logging fields so no more flushes are
836 spin_lock(&iip->ili_lock);
837 iip->ili_last_fields = 0;
839 iip->ili_fsync_fields = 0;
840 iip->ili_flush_lsn = 0;
841 bp = iip->ili_item.li_buf;
842 iip->ili_item.li_buf = NULL;
843 list_del_init(&iip->ili_item.li_bio_list);
844 spin_unlock(&iip->ili_lock);
846 xfs_iflags_clear(ip, XFS_IFLUSHING);
852 * convert an xfs_inode_log_format struct from the old 32 bit version
853 * (which can have different field alignments) to the native 64 bit version
856 xfs_inode_item_format_convert(
857 struct xfs_log_iovec *buf,
858 struct xfs_inode_log_format *in_f)
860 struct xfs_inode_log_format_32 *in_f32 = buf->i_addr;
862 if (buf->i_len != sizeof(*in_f32)) {
863 XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_LOW, NULL);
864 return -EFSCORRUPTED;
867 in_f->ilf_type = in_f32->ilf_type;
868 in_f->ilf_size = in_f32->ilf_size;
869 in_f->ilf_fields = in_f32->ilf_fields;
870 in_f->ilf_asize = in_f32->ilf_asize;
871 in_f->ilf_dsize = in_f32->ilf_dsize;
872 in_f->ilf_ino = in_f32->ilf_ino;
873 memcpy(&in_f->ilf_u, &in_f32->ilf_u, sizeof(in_f->ilf_u));
874 in_f->ilf_blkno = in_f32->ilf_blkno;
875 in_f->ilf_len = in_f32->ilf_len;
876 in_f->ilf_boffset = in_f32->ilf_boffset;