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_ictimestamp_t
303 xfs_inode_to_log_dinode_ts(
304 struct xfs_inode *ip,
305 const struct timespec64 tv)
307 struct xfs_legacy_ictimestamp *lits;
308 xfs_ictimestamp_t its;
310 if (xfs_inode_has_bigtime(ip))
311 return xfs_inode_encode_bigtime(tv);
313 lits = (struct xfs_legacy_ictimestamp *)&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 xfs_icdinode *from = &ip->i_d;
354 struct inode *inode = VFS_I(ip);
356 to->di_magic = XFS_DINODE_MAGIC;
357 to->di_format = xfs_ifork_format(&ip->i_df);
358 to->di_uid = i_uid_read(inode);
359 to->di_gid = i_gid_read(inode);
360 to->di_projid_lo = ip->i_projid & 0xffff;
361 to->di_projid_hi = ip->i_projid >> 16;
363 memset(to->di_pad, 0, sizeof(to->di_pad));
364 memset(to->di_pad3, 0, sizeof(to->di_pad3));
365 to->di_atime = xfs_inode_to_log_dinode_ts(ip, inode->i_atime);
366 to->di_mtime = xfs_inode_to_log_dinode_ts(ip, inode->i_mtime);
367 to->di_ctime = xfs_inode_to_log_dinode_ts(ip, inode->i_ctime);
368 to->di_nlink = inode->i_nlink;
369 to->di_gen = inode->i_generation;
370 to->di_mode = inode->i_mode;
372 to->di_size = ip->i_disk_size;
373 to->di_nblocks = ip->i_nblocks;
374 to->di_extsize = from->di_extsize;
375 to->di_nextents = xfs_ifork_nextents(&ip->i_df);
376 to->di_anextents = xfs_ifork_nextents(ip->i_afp);
377 to->di_forkoff = from->di_forkoff;
378 to->di_aformat = xfs_ifork_format(ip->i_afp);
379 to->di_flags = from->di_flags;
381 xfs_copy_dm_fields_to_log_dinode(ip, to);
383 /* log a dummy value to ensure log structure is fully initialised */
384 to->di_next_unlinked = NULLAGINO;
386 if (xfs_sb_version_has_v3inode(&ip->i_mount->m_sb)) {
388 to->di_changecount = inode_peek_iversion(inode);
389 to->di_crtime = xfs_inode_to_log_dinode_ts(ip, from->di_crtime);
390 to->di_flags2 = from->di_flags2;
391 to->di_cowextsize = from->di_cowextsize;
392 to->di_ino = ip->i_ino;
394 memset(to->di_pad2, 0, sizeof(to->di_pad2));
395 uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid);
396 to->di_flushiter = 0;
399 to->di_flushiter = from->di_flushiter;
404 * Format the inode core. Current timestamp data is only in the VFS inode
405 * fields, so we need to grab them from there. Hence rather than just copying
406 * the XFS inode core structure, format the fields directly into the iovec.
409 xfs_inode_item_format_core(
410 struct xfs_inode *ip,
411 struct xfs_log_vec *lv,
412 struct xfs_log_iovec **vecp)
414 struct xfs_log_dinode *dic;
416 dic = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_ICORE);
417 xfs_inode_to_log_dinode(ip, dic, ip->i_itemp->ili_item.li_lsn);
418 xlog_finish_iovec(lv, *vecp, xfs_log_dinode_size(ip->i_mount));
422 * This is called to fill in the vector of log iovecs for the given inode
423 * log item. It fills the first item with an inode log format structure,
424 * the second with the on-disk inode structure, and a possible third and/or
425 * fourth with the inode data/extents/b-tree root and inode attributes
426 * data/extents/b-tree root.
428 * Note: Always use the 64 bit inode log format structure so we don't
429 * leave an uninitialised hole in the format item on 64 bit systems. Log
430 * recovery on 32 bit systems handles this just fine, so there's no reason
431 * for not using an initialising the properly padded structure all the time.
434 xfs_inode_item_format(
435 struct xfs_log_item *lip,
436 struct xfs_log_vec *lv)
438 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
439 struct xfs_inode *ip = iip->ili_inode;
440 struct xfs_log_iovec *vecp = NULL;
441 struct xfs_inode_log_format *ilf;
443 ilf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_IFORMAT);
444 ilf->ilf_type = XFS_LI_INODE;
445 ilf->ilf_ino = ip->i_ino;
446 ilf->ilf_blkno = ip->i_imap.im_blkno;
447 ilf->ilf_len = ip->i_imap.im_len;
448 ilf->ilf_boffset = ip->i_imap.im_boffset;
449 ilf->ilf_fields = XFS_ILOG_CORE;
450 ilf->ilf_size = 2; /* format + core */
453 * make sure we don't leak uninitialised data into the log in the case
454 * when we don't log every field in the inode.
459 memset(&ilf->ilf_u, 0, sizeof(ilf->ilf_u));
461 xlog_finish_iovec(lv, vecp, sizeof(*ilf));
463 xfs_inode_item_format_core(ip, lv, &vecp);
464 xfs_inode_item_format_data_fork(iip, ilf, lv, &vecp);
465 if (XFS_IFORK_Q(ip)) {
466 xfs_inode_item_format_attr_fork(iip, ilf, lv, &vecp);
469 ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT);
472 /* update the format with the exact fields we actually logged */
473 ilf->ilf_fields |= (iip->ili_fields & ~XFS_ILOG_TIMESTAMP);
477 * This is called to pin the inode associated with the inode log
478 * item in memory so it cannot be written out.
482 struct xfs_log_item *lip)
484 struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode;
486 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
489 trace_xfs_inode_pin(ip, _RET_IP_);
490 atomic_inc(&ip->i_pincount);
495 * This is called to unpin the inode associated with the inode log
496 * item which was previously pinned with a call to xfs_inode_item_pin().
498 * Also wake up anyone in xfs_iunpin_wait() if the count goes to 0.
500 * Note that unpin can race with inode cluster buffer freeing marking the buffer
501 * stale. In that case, flush completions are run from the buffer unpin call,
502 * which may happen before the inode is unpinned. If we lose the race, there
503 * will be no buffer attached to the log item, but the inode will be marked
507 xfs_inode_item_unpin(
508 struct xfs_log_item *lip,
511 struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode;
513 trace_xfs_inode_unpin(ip, _RET_IP_);
514 ASSERT(lip->li_buf || xfs_iflags_test(ip, XFS_ISTALE));
515 ASSERT(atomic_read(&ip->i_pincount) > 0);
516 if (atomic_dec_and_test(&ip->i_pincount))
517 wake_up_bit(&ip->i_flags, __XFS_IPINNED_BIT);
522 struct xfs_log_item *lip,
523 struct list_head *buffer_list)
524 __releases(&lip->li_ailp->ail_lock)
525 __acquires(&lip->li_ailp->ail_lock)
527 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
528 struct xfs_inode *ip = iip->ili_inode;
529 struct xfs_buf *bp = lip->li_buf;
530 uint rval = XFS_ITEM_SUCCESS;
533 ASSERT(iip->ili_item.li_buf);
535 if (xfs_ipincount(ip) > 0 || xfs_buf_ispinned(bp) ||
536 (ip->i_flags & XFS_ISTALE))
537 return XFS_ITEM_PINNED;
539 if (xfs_iflags_test(ip, XFS_IFLUSHING))
540 return XFS_ITEM_FLUSHING;
542 if (!xfs_buf_trylock(bp))
543 return XFS_ITEM_LOCKED;
545 spin_unlock(&lip->li_ailp->ail_lock);
548 * We need to hold a reference for flushing the cluster buffer as it may
549 * fail the buffer without IO submission. In which case, we better get a
550 * reference for that completion because otherwise we don't get a
551 * reference for IO until we queue the buffer for delwri submission.
554 error = xfs_iflush_cluster(bp);
556 if (!xfs_buf_delwri_queue(bp, buffer_list))
557 rval = XFS_ITEM_FLUSHING;
561 * Release the buffer if we were unable to flush anything. On
562 * any other error, the buffer has already been released.
564 if (error == -EAGAIN)
566 rval = XFS_ITEM_LOCKED;
569 spin_lock(&lip->li_ailp->ail_lock);
574 * Unlock the inode associated with the inode log item.
577 xfs_inode_item_release(
578 struct xfs_log_item *lip)
580 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
581 struct xfs_inode *ip = iip->ili_inode;
582 unsigned short lock_flags;
584 ASSERT(ip->i_itemp != NULL);
585 ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL));
587 lock_flags = iip->ili_lock_flags;
588 iip->ili_lock_flags = 0;
590 xfs_iunlock(ip, lock_flags);
594 * This is called to find out where the oldest active copy of the inode log
595 * item in the on disk log resides now that the last log write of it completed
596 * at the given lsn. Since we always re-log all dirty data in an inode, the
597 * latest copy in the on disk log is the only one that matters. Therefore,
598 * simply return the given lsn.
600 * If the inode has been marked stale because the cluster is being freed, we
601 * don't want to (re-)insert this inode into the AIL. There is a race condition
602 * where the cluster buffer may be unpinned before the inode is inserted into
603 * the AIL during transaction committed processing. If the buffer is unpinned
604 * before the inode item has been committed and inserted, then it is possible
605 * for the buffer to be written and IO completes before the inode is inserted
606 * into the AIL. In that case, we'd be inserting a clean, stale inode into the
607 * AIL which will never get removed. It will, however, get reclaimed which
608 * triggers an assert in xfs_inode_free() complaining about freein an inode
611 * To avoid this, just unpin the inode directly and return a LSN of -1 so the
612 * transaction committed code knows that it does not need to do any further
613 * processing on the item.
616 xfs_inode_item_committed(
617 struct xfs_log_item *lip,
620 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
621 struct xfs_inode *ip = iip->ili_inode;
623 if (xfs_iflags_test(ip, XFS_ISTALE)) {
624 xfs_inode_item_unpin(lip, 0);
631 xfs_inode_item_committing(
632 struct xfs_log_item *lip,
633 xfs_lsn_t commit_lsn)
635 INODE_ITEM(lip)->ili_last_lsn = commit_lsn;
636 return xfs_inode_item_release(lip);
639 static const struct xfs_item_ops xfs_inode_item_ops = {
640 .iop_size = xfs_inode_item_size,
641 .iop_format = xfs_inode_item_format,
642 .iop_pin = xfs_inode_item_pin,
643 .iop_unpin = xfs_inode_item_unpin,
644 .iop_release = xfs_inode_item_release,
645 .iop_committed = xfs_inode_item_committed,
646 .iop_push = xfs_inode_item_push,
647 .iop_committing = xfs_inode_item_committing,
652 * Initialize the inode log item for a newly allocated (in-core) inode.
656 struct xfs_inode *ip,
657 struct xfs_mount *mp)
659 struct xfs_inode_log_item *iip;
661 ASSERT(ip->i_itemp == NULL);
662 iip = ip->i_itemp = kmem_cache_zalloc(xfs_ili_zone,
663 GFP_KERNEL | __GFP_NOFAIL);
666 spin_lock_init(&iip->ili_lock);
667 xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE,
668 &xfs_inode_item_ops);
672 * Free the inode log item and any memory hanging off of it.
675 xfs_inode_item_destroy(
676 struct xfs_inode *ip)
678 struct xfs_inode_log_item *iip = ip->i_itemp;
680 ASSERT(iip->ili_item.li_buf == NULL);
683 kmem_free(iip->ili_item.li_lv_shadow);
684 kmem_cache_free(xfs_ili_zone, iip);
689 * We only want to pull the item from the AIL if it is actually there
690 * and its location in the log has not changed since we started the
691 * flush. Thus, we only bother if the inode's lsn has not changed.
694 xfs_iflush_ail_updates(
695 struct xfs_ail *ailp,
696 struct list_head *list)
698 struct xfs_log_item *lip;
699 xfs_lsn_t tail_lsn = 0;
701 /* this is an opencoded batch version of xfs_trans_ail_delete */
702 spin_lock(&ailp->ail_lock);
703 list_for_each_entry(lip, list, li_bio_list) {
706 clear_bit(XFS_LI_FAILED, &lip->li_flags);
707 if (INODE_ITEM(lip)->ili_flush_lsn != lip->li_lsn)
710 lsn = xfs_ail_delete_one(ailp, lip);
711 if (!tail_lsn && lsn)
714 xfs_ail_update_finish(ailp, tail_lsn);
718 * Walk the list of inodes that have completed their IOs. If they are clean
719 * remove them from the list and dissociate them from the buffer. Buffers that
720 * are still dirty remain linked to the buffer and on the list. Caller must
721 * handle them appropriately.
726 struct list_head *list)
728 struct xfs_log_item *lip, *n;
730 list_for_each_entry_safe(lip, n, list, li_bio_list) {
731 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
732 bool drop_buffer = false;
734 spin_lock(&iip->ili_lock);
737 * Remove the reference to the cluster buffer if the inode is
738 * clean in memory and drop the buffer reference once we've
739 * dropped the locks we hold.
741 ASSERT(iip->ili_item.li_buf == bp);
742 if (!iip->ili_fields) {
743 iip->ili_item.li_buf = NULL;
744 list_del_init(&lip->li_bio_list);
747 iip->ili_last_fields = 0;
748 iip->ili_flush_lsn = 0;
749 spin_unlock(&iip->ili_lock);
750 xfs_iflags_clear(iip->ili_inode, XFS_IFLUSHING);
757 * Inode buffer IO completion routine. It is responsible for removing inodes
758 * attached to the buffer from the AIL if they have not been re-logged and
759 * completing the inode flush.
762 xfs_buf_inode_iodone(
765 struct xfs_log_item *lip, *n;
766 LIST_HEAD(flushed_inodes);
767 LIST_HEAD(ail_updates);
770 * Pull the attached inodes from the buffer one at a time and take the
771 * appropriate action on them.
773 list_for_each_entry_safe(lip, n, &bp->b_li_list, li_bio_list) {
774 struct xfs_inode_log_item *iip = INODE_ITEM(lip);
776 if (xfs_iflags_test(iip->ili_inode, XFS_ISTALE)) {
777 xfs_iflush_abort(iip->ili_inode);
780 if (!iip->ili_last_fields)
783 /* Do an unlocked check for needing the AIL lock. */
784 if (iip->ili_flush_lsn == lip->li_lsn ||
785 test_bit(XFS_LI_FAILED, &lip->li_flags))
786 list_move_tail(&lip->li_bio_list, &ail_updates);
788 list_move_tail(&lip->li_bio_list, &flushed_inodes);
791 if (!list_empty(&ail_updates)) {
792 xfs_iflush_ail_updates(bp->b_mount->m_ail, &ail_updates);
793 list_splice_tail(&ail_updates, &flushed_inodes);
796 xfs_iflush_finish(bp, &flushed_inodes);
797 if (!list_empty(&flushed_inodes))
798 list_splice_tail(&flushed_inodes, &bp->b_li_list);
802 xfs_buf_inode_io_fail(
805 struct xfs_log_item *lip;
807 list_for_each_entry(lip, &bp->b_li_list, li_bio_list)
808 set_bit(XFS_LI_FAILED, &lip->li_flags);
812 * This is the inode flushing abort routine. It is called when
813 * the filesystem is shutting down to clean up the inode state. It is
814 * responsible for removing the inode item from the AIL if it has not been
815 * re-logged and clearing the inode's flush state.
819 struct xfs_inode *ip)
821 struct xfs_inode_log_item *iip = ip->i_itemp;
822 struct xfs_buf *bp = NULL;
826 * Clear the failed bit before removing the item from the AIL so
827 * xfs_trans_ail_delete() doesn't try to clear and release the
828 * buffer attached to the log item before we are done with it.
830 clear_bit(XFS_LI_FAILED, &iip->ili_item.li_flags);
831 xfs_trans_ail_delete(&iip->ili_item, 0);
834 * Clear the inode logging fields so no more flushes are
837 spin_lock(&iip->ili_lock);
838 iip->ili_last_fields = 0;
840 iip->ili_fsync_fields = 0;
841 iip->ili_flush_lsn = 0;
842 bp = iip->ili_item.li_buf;
843 iip->ili_item.li_buf = NULL;
844 list_del_init(&iip->ili_item.li_bio_list);
845 spin_unlock(&iip->ili_lock);
847 xfs_iflags_clear(ip, XFS_IFLUSHING);
853 * convert an xfs_inode_log_format struct from the old 32 bit version
854 * (which can have different field alignments) to the native 64 bit version
857 xfs_inode_item_format_convert(
858 struct xfs_log_iovec *buf,
859 struct xfs_inode_log_format *in_f)
861 struct xfs_inode_log_format_32 *in_f32 = buf->i_addr;
863 if (buf->i_len != sizeof(*in_f32)) {
864 XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_LOW, NULL);
865 return -EFSCORRUPTED;
868 in_f->ilf_type = in_f32->ilf_type;
869 in_f->ilf_size = in_f32->ilf_size;
870 in_f->ilf_fields = in_f32->ilf_fields;
871 in_f->ilf_asize = in_f32->ilf_asize;
872 in_f->ilf_dsize = in_f32->ilf_dsize;
873 in_f->ilf_ino = in_f32->ilf_ino;
874 memcpy(&in_f->ilf_u, &in_f32->ilf_u, sizeof(in_f->ilf_u));
875 in_f->ilf_blkno = in_f32->ilf_blkno;
876 in_f->ilf_len = in_f32->ilf_len;
877 in_f->ilf_boffset = in_f32->ilf_boffset;