1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (c) 2000-2006 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"
13 #include "xfs_mount.h"
14 #include "xfs_inode.h"
15 #include "xfs_btree.h"
17 #include "xfs_alloc.h"
18 #include "xfs_fsops.h"
19 #include "xfs_trans.h"
20 #include "xfs_buf_item.h"
22 #include "xfs_log_priv.h"
24 #include "xfs_extfree_item.h"
25 #include "xfs_mru_cache.h"
26 #include "xfs_inode_item.h"
27 #include "xfs_icache.h"
28 #include "xfs_trace.h"
29 #include "xfs_icreate_item.h"
30 #include "xfs_filestream.h"
31 #include "xfs_quota.h"
32 #include "xfs_sysfs.h"
33 #include "xfs_ondisk.h"
34 #include "xfs_rmap_item.h"
35 #include "xfs_refcount_item.h"
36 #include "xfs_bmap_item.h"
37 #include "xfs_reflink.h"
38 #include "xfs_pwork.h"
40 #include "xfs_defer.h"
41 #include "xfs_attr_item.h"
42 #include "xfs_xattr.h"
43 #include "xfs_iunlink_item.h"
44 #include "xfs_dahash_test.h"
46 #include <linux/magic.h>
47 #include <linux/fs_context.h>
48 #include <linux/fs_parser.h>
50 static const struct super_operations xfs_super_operations;
52 static struct kset *xfs_kset; /* top-level xfs sysfs dir */
54 static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */
57 #ifdef CONFIG_HOTPLUG_CPU
58 static LIST_HEAD(xfs_mount_list);
59 static DEFINE_SPINLOCK(xfs_mount_list_lock);
61 static inline void xfs_mount_list_add(struct xfs_mount *mp)
63 spin_lock(&xfs_mount_list_lock);
64 list_add(&mp->m_mount_list, &xfs_mount_list);
65 spin_unlock(&xfs_mount_list_lock);
68 static inline void xfs_mount_list_del(struct xfs_mount *mp)
70 spin_lock(&xfs_mount_list_lock);
71 list_del(&mp->m_mount_list);
72 spin_unlock(&xfs_mount_list_lock);
74 #else /* !CONFIG_HOTPLUG_CPU */
75 static inline void xfs_mount_list_add(struct xfs_mount *mp) {}
76 static inline void xfs_mount_list_del(struct xfs_mount *mp) {}
86 xfs_mount_set_dax_mode(
88 enum xfs_dax_mode mode)
92 mp->m_features &= ~(XFS_FEAT_DAX_ALWAYS | XFS_FEAT_DAX_NEVER);
95 mp->m_features |= XFS_FEAT_DAX_ALWAYS;
96 mp->m_features &= ~XFS_FEAT_DAX_NEVER;
99 mp->m_features |= XFS_FEAT_DAX_NEVER;
100 mp->m_features &= ~XFS_FEAT_DAX_ALWAYS;
105 static const struct constant_table dax_param_enums[] = {
106 {"inode", XFS_DAX_INODE },
107 {"always", XFS_DAX_ALWAYS },
108 {"never", XFS_DAX_NEVER },
113 * Table driven mount option parser.
116 Opt_logbufs, Opt_logbsize, Opt_logdev, Opt_rtdev,
117 Opt_wsync, Opt_noalign, Opt_swalloc, Opt_sunit, Opt_swidth, Opt_nouuid,
118 Opt_grpid, Opt_nogrpid, Opt_bsdgroups, Opt_sysvgroups,
119 Opt_allocsize, Opt_norecovery, Opt_inode64, Opt_inode32, Opt_ikeep,
120 Opt_noikeep, Opt_largeio, Opt_nolargeio, Opt_attr2, Opt_noattr2,
121 Opt_filestreams, Opt_quota, Opt_noquota, Opt_usrquota, Opt_grpquota,
122 Opt_prjquota, Opt_uquota, Opt_gquota, Opt_pquota,
123 Opt_uqnoenforce, Opt_gqnoenforce, Opt_pqnoenforce, Opt_qnoenforce,
124 Opt_discard, Opt_nodiscard, Opt_dax, Opt_dax_enum,
127 static const struct fs_parameter_spec xfs_fs_parameters[] = {
128 fsparam_u32("logbufs", Opt_logbufs),
129 fsparam_string("logbsize", Opt_logbsize),
130 fsparam_string("logdev", Opt_logdev),
131 fsparam_string("rtdev", Opt_rtdev),
132 fsparam_flag("wsync", Opt_wsync),
133 fsparam_flag("noalign", Opt_noalign),
134 fsparam_flag("swalloc", Opt_swalloc),
135 fsparam_u32("sunit", Opt_sunit),
136 fsparam_u32("swidth", Opt_swidth),
137 fsparam_flag("nouuid", Opt_nouuid),
138 fsparam_flag("grpid", Opt_grpid),
139 fsparam_flag("nogrpid", Opt_nogrpid),
140 fsparam_flag("bsdgroups", Opt_bsdgroups),
141 fsparam_flag("sysvgroups", Opt_sysvgroups),
142 fsparam_string("allocsize", Opt_allocsize),
143 fsparam_flag("norecovery", Opt_norecovery),
144 fsparam_flag("inode64", Opt_inode64),
145 fsparam_flag("inode32", Opt_inode32),
146 fsparam_flag("ikeep", Opt_ikeep),
147 fsparam_flag("noikeep", Opt_noikeep),
148 fsparam_flag("largeio", Opt_largeio),
149 fsparam_flag("nolargeio", Opt_nolargeio),
150 fsparam_flag("attr2", Opt_attr2),
151 fsparam_flag("noattr2", Opt_noattr2),
152 fsparam_flag("filestreams", Opt_filestreams),
153 fsparam_flag("quota", Opt_quota),
154 fsparam_flag("noquota", Opt_noquota),
155 fsparam_flag("usrquota", Opt_usrquota),
156 fsparam_flag("grpquota", Opt_grpquota),
157 fsparam_flag("prjquota", Opt_prjquota),
158 fsparam_flag("uquota", Opt_uquota),
159 fsparam_flag("gquota", Opt_gquota),
160 fsparam_flag("pquota", Opt_pquota),
161 fsparam_flag("uqnoenforce", Opt_uqnoenforce),
162 fsparam_flag("gqnoenforce", Opt_gqnoenforce),
163 fsparam_flag("pqnoenforce", Opt_pqnoenforce),
164 fsparam_flag("qnoenforce", Opt_qnoenforce),
165 fsparam_flag("discard", Opt_discard),
166 fsparam_flag("nodiscard", Opt_nodiscard),
167 fsparam_flag("dax", Opt_dax),
168 fsparam_enum("dax", Opt_dax_enum, dax_param_enums),
172 struct proc_xfs_info {
182 static struct proc_xfs_info xfs_info_set[] = {
183 /* the few simple ones we can get from the mount struct */
184 { XFS_FEAT_IKEEP, ",ikeep" },
185 { XFS_FEAT_WSYNC, ",wsync" },
186 { XFS_FEAT_NOALIGN, ",noalign" },
187 { XFS_FEAT_SWALLOC, ",swalloc" },
188 { XFS_FEAT_NOUUID, ",nouuid" },
189 { XFS_FEAT_NORECOVERY, ",norecovery" },
190 { XFS_FEAT_ATTR2, ",attr2" },
191 { XFS_FEAT_FILESTREAMS, ",filestreams" },
192 { XFS_FEAT_GRPID, ",grpid" },
193 { XFS_FEAT_DISCARD, ",discard" },
194 { XFS_FEAT_LARGE_IOSIZE, ",largeio" },
195 { XFS_FEAT_DAX_ALWAYS, ",dax=always" },
196 { XFS_FEAT_DAX_NEVER, ",dax=never" },
199 struct xfs_mount *mp = XFS_M(root->d_sb);
200 struct proc_xfs_info *xfs_infop;
202 for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
203 if (mp->m_features & xfs_infop->flag)
204 seq_puts(m, xfs_infop->str);
207 seq_printf(m, ",inode%d", xfs_has_small_inums(mp) ? 32 : 64);
209 if (xfs_has_allocsize(mp))
210 seq_printf(m, ",allocsize=%dk",
211 (1 << mp->m_allocsize_log) >> 10);
213 if (mp->m_logbufs > 0)
214 seq_printf(m, ",logbufs=%d", mp->m_logbufs);
215 if (mp->m_logbsize > 0)
216 seq_printf(m, ",logbsize=%dk", mp->m_logbsize >> 10);
219 seq_show_option(m, "logdev", mp->m_logname);
221 seq_show_option(m, "rtdev", mp->m_rtname);
223 if (mp->m_dalign > 0)
224 seq_printf(m, ",sunit=%d",
225 (int)XFS_FSB_TO_BB(mp, mp->m_dalign));
226 if (mp->m_swidth > 0)
227 seq_printf(m, ",swidth=%d",
228 (int)XFS_FSB_TO_BB(mp, mp->m_swidth));
230 if (mp->m_qflags & XFS_UQUOTA_ENFD)
231 seq_puts(m, ",usrquota");
232 else if (mp->m_qflags & XFS_UQUOTA_ACCT)
233 seq_puts(m, ",uqnoenforce");
235 if (mp->m_qflags & XFS_PQUOTA_ENFD)
236 seq_puts(m, ",prjquota");
237 else if (mp->m_qflags & XFS_PQUOTA_ACCT)
238 seq_puts(m, ",pqnoenforce");
240 if (mp->m_qflags & XFS_GQUOTA_ENFD)
241 seq_puts(m, ",grpquota");
242 else if (mp->m_qflags & XFS_GQUOTA_ACCT)
243 seq_puts(m, ",gqnoenforce");
245 if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
246 seq_puts(m, ",noquota");
252 xfs_set_inode_alloc_perag(
253 struct xfs_perag *pag,
255 xfs_agnumber_t max_metadata)
257 if (!xfs_is_inode32(pag->pag_mount)) {
258 set_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate);
259 clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate);
263 if (ino > XFS_MAXINUMBER_32) {
264 clear_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate);
265 clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate);
269 set_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate);
270 if (pag->pag_agno < max_metadata)
271 set_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate);
273 clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate);
278 * Set parameters for inode allocation heuristics, taking into account
279 * filesystem size and inode32/inode64 mount options; i.e. specifically
280 * whether or not XFS_FEAT_SMALL_INUMS is set.
282 * Inode allocation patterns are altered only if inode32 is requested
283 * (XFS_FEAT_SMALL_INUMS), and the filesystem is sufficiently large.
284 * If altered, XFS_OPSTATE_INODE32 is set as well.
286 * An agcount independent of that in the mount structure is provided
287 * because in the growfs case, mp->m_sb.sb_agcount is not yet updated
288 * to the potentially higher ag count.
290 * Returns the maximum AG index which may contain inodes.
294 struct xfs_mount *mp,
295 xfs_agnumber_t agcount)
297 xfs_agnumber_t index;
298 xfs_agnumber_t maxagi = 0;
299 xfs_sb_t *sbp = &mp->m_sb;
300 xfs_agnumber_t max_metadata;
305 * Calculate how much should be reserved for inodes to meet
306 * the max inode percentage. Used only for inode32.
308 if (M_IGEO(mp)->maxicount) {
311 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
313 icount += sbp->sb_agblocks - 1;
314 do_div(icount, sbp->sb_agblocks);
315 max_metadata = icount;
317 max_metadata = agcount;
320 /* Get the last possible inode in the filesystem */
321 agino = XFS_AGB_TO_AGINO(mp, sbp->sb_agblocks - 1);
322 ino = XFS_AGINO_TO_INO(mp, agcount - 1, agino);
325 * If user asked for no more than 32-bit inodes, and the fs is
326 * sufficiently large, set XFS_OPSTATE_INODE32 if we must alter
327 * the allocator to accommodate the request.
329 if (xfs_has_small_inums(mp) && ino > XFS_MAXINUMBER_32)
330 set_bit(XFS_OPSTATE_INODE32, &mp->m_opstate);
332 clear_bit(XFS_OPSTATE_INODE32, &mp->m_opstate);
334 for (index = 0; index < agcount; index++) {
335 struct xfs_perag *pag;
337 ino = XFS_AGINO_TO_INO(mp, index, agino);
339 pag = xfs_perag_get(mp, index);
340 if (xfs_set_inode_alloc_perag(pag, ino, max_metadata))
345 return xfs_is_inode32(mp) ? maxagi : agcount;
349 xfs_setup_dax_always(
350 struct xfs_mount *mp)
352 if (!mp->m_ddev_targp->bt_daxdev &&
353 (!mp->m_rtdev_targp || !mp->m_rtdev_targp->bt_daxdev)) {
355 "DAX unsupported by block device. Turning off DAX.");
359 if (mp->m_super->s_blocksize != PAGE_SIZE) {
361 "DAX not supported for blocksize. Turning off DAX.");
365 if (xfs_has_reflink(mp) &&
366 bdev_is_partition(mp->m_ddev_targp->bt_bdev)) {
368 "DAX and reflink cannot work with multi-partitions!");
372 xfs_warn(mp, "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
376 xfs_mount_set_dax_mode(mp, XFS_DAX_NEVER);
382 struct block_device *bdev)
384 xfs_force_shutdown(bdev->bd_holder, SHUTDOWN_DEVICE_REMOVED);
387 static const struct blk_holder_ops xfs_holder_ops = {
388 .mark_dead = xfs_bdev_mark_dead,
395 struct block_device **bdevp)
399 *bdevp = blkdev_get_by_path(name, BLK_OPEN_READ | BLK_OPEN_WRITE, mp,
401 if (IS_ERR(*bdevp)) {
402 error = PTR_ERR(*bdevp);
403 xfs_warn(mp, "Invalid device [%s], error=%d", name, error);
410 xfs_shutdown_devices(
411 struct xfs_mount *mp)
414 * Udev is triggered whenever anyone closes a block device or unmounts
415 * a file systemm on a block device.
416 * The default udev rules invoke blkid to read the fs super and create
417 * symlinks to the bdev under /dev/disk. For this, it uses buffered
418 * reads through the page cache.
420 * xfs_db also uses buffered reads to examine metadata. There is no
421 * coordination between xfs_db and udev, which means that they can run
422 * concurrently. Note there is no coordination between the kernel and
425 * On a system with 64k pages, the page cache can cache the superblock
426 * and the root inode (and hence the root directory) with the same 64k
427 * page. If udev spawns blkid after the mkfs and the system is busy
428 * enough that it is still running when xfs_db starts up, they'll both
429 * read from the same page in the pagecache.
431 * The unmount writes updated inode metadata to disk directly. The XFS
432 * buffer cache does not use the bdev pagecache, so it needs to
433 * invalidate that pagecache on unmount. If the above scenario occurs,
434 * the pagecache no longer reflects what's on disk, xfs_db reads the
435 * stale metadata, and fails to find /a. Most of the time this succeeds
436 * because closing a bdev invalidates the page cache, but when processes
437 * race, everyone loses.
439 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
440 blkdev_issue_flush(mp->m_logdev_targp->bt_bdev);
441 invalidate_bdev(mp->m_logdev_targp->bt_bdev);
443 if (mp->m_rtdev_targp) {
444 blkdev_issue_flush(mp->m_rtdev_targp->bt_bdev);
445 invalidate_bdev(mp->m_rtdev_targp->bt_bdev);
447 blkdev_issue_flush(mp->m_ddev_targp->bt_bdev);
448 invalidate_bdev(mp->m_ddev_targp->bt_bdev);
452 * The file system configurations are:
453 * (1) device (partition) with data and internal log
454 * (2) logical volume with data and log subvolumes.
455 * (3) logical volume with data, log, and realtime subvolumes.
457 * We only have to handle opening the log and realtime volumes here if
458 * they are present. The data subvolume has already been opened by
459 * get_sb_bdev() and is stored in sb->s_bdev.
463 struct xfs_mount *mp)
465 struct block_device *ddev = mp->m_super->s_bdev;
466 struct block_device *logdev = NULL, *rtdev = NULL;
470 * Open real time and log devices - order is important.
473 error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
479 error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
481 goto out_close_logdev;
483 if (rtdev == ddev || rtdev == logdev) {
485 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
487 goto out_close_rtdev;
492 * Setup xfs_mount buffer target pointers
495 mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev);
496 if (!mp->m_ddev_targp)
497 goto out_close_rtdev;
500 mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev);
501 if (!mp->m_rtdev_targp)
502 goto out_free_ddev_targ;
505 if (logdev && logdev != ddev) {
506 mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev);
507 if (!mp->m_logdev_targp)
508 goto out_free_rtdev_targ;
510 mp->m_logdev_targp = mp->m_ddev_targp;
516 if (mp->m_rtdev_targp)
517 xfs_free_buftarg(mp->m_rtdev_targp);
519 xfs_free_buftarg(mp->m_ddev_targp);
522 blkdev_put(rtdev, mp);
524 if (logdev && logdev != ddev)
525 blkdev_put(logdev, mp);
530 * Setup xfs_mount buffer target pointers based on superblock
534 struct xfs_mount *mp)
538 error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize);
542 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
543 unsigned int log_sector_size = BBSIZE;
545 if (xfs_has_sector(mp))
546 log_sector_size = mp->m_sb.sb_logsectsize;
547 error = xfs_setsize_buftarg(mp->m_logdev_targp,
552 if (mp->m_rtdev_targp) {
553 error = xfs_setsize_buftarg(mp->m_rtdev_targp,
554 mp->m_sb.sb_sectsize);
563 xfs_init_mount_workqueues(
564 struct xfs_mount *mp)
566 mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s",
567 XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM),
568 1, mp->m_super->s_id);
569 if (!mp->m_buf_workqueue)
572 mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
573 XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM),
574 0, mp->m_super->s_id);
575 if (!mp->m_unwritten_workqueue)
576 goto out_destroy_buf;
578 mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
579 XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM),
580 0, mp->m_super->s_id);
581 if (!mp->m_reclaim_workqueue)
582 goto out_destroy_unwritten;
584 mp->m_blockgc_wq = alloc_workqueue("xfs-blockgc/%s",
585 XFS_WQFLAGS(WQ_UNBOUND | WQ_FREEZABLE | WQ_MEM_RECLAIM),
586 0, mp->m_super->s_id);
587 if (!mp->m_blockgc_wq)
588 goto out_destroy_reclaim;
590 mp->m_inodegc_wq = alloc_workqueue("xfs-inodegc/%s",
591 XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM),
592 1, mp->m_super->s_id);
593 if (!mp->m_inodegc_wq)
594 goto out_destroy_blockgc;
596 mp->m_sync_workqueue = alloc_workqueue("xfs-sync/%s",
597 XFS_WQFLAGS(WQ_FREEZABLE), 0, mp->m_super->s_id);
598 if (!mp->m_sync_workqueue)
599 goto out_destroy_inodegc;
604 destroy_workqueue(mp->m_inodegc_wq);
606 destroy_workqueue(mp->m_blockgc_wq);
608 destroy_workqueue(mp->m_reclaim_workqueue);
609 out_destroy_unwritten:
610 destroy_workqueue(mp->m_unwritten_workqueue);
612 destroy_workqueue(mp->m_buf_workqueue);
618 xfs_destroy_mount_workqueues(
619 struct xfs_mount *mp)
621 destroy_workqueue(mp->m_sync_workqueue);
622 destroy_workqueue(mp->m_blockgc_wq);
623 destroy_workqueue(mp->m_inodegc_wq);
624 destroy_workqueue(mp->m_reclaim_workqueue);
625 destroy_workqueue(mp->m_unwritten_workqueue);
626 destroy_workqueue(mp->m_buf_workqueue);
630 xfs_flush_inodes_worker(
631 struct work_struct *work)
633 struct xfs_mount *mp = container_of(work, struct xfs_mount,
634 m_flush_inodes_work);
635 struct super_block *sb = mp->m_super;
637 if (down_read_trylock(&sb->s_umount)) {
639 up_read(&sb->s_umount);
644 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
645 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
646 * for IO to complete so that we effectively throttle multiple callers to the
647 * rate at which IO is completing.
651 struct xfs_mount *mp)
654 * If flush_work() returns true then that means we waited for a flush
655 * which was already in progress. Don't bother running another scan.
657 if (flush_work(&mp->m_flush_inodes_work))
660 queue_work(mp->m_sync_workqueue, &mp->m_flush_inodes_work);
661 flush_work(&mp->m_flush_inodes_work);
664 /* Catch misguided souls that try to use this interface on XFS */
665 STATIC struct inode *
667 struct super_block *sb)
674 * Now that the generic code is guaranteed not to be accessing
675 * the linux inode, we can inactivate and reclaim the inode.
678 xfs_fs_destroy_inode(
681 struct xfs_inode *ip = XFS_I(inode);
683 trace_xfs_destroy_inode(ip);
685 ASSERT(!rwsem_is_locked(&inode->i_rwsem));
686 XFS_STATS_INC(ip->i_mount, vn_rele);
687 XFS_STATS_INC(ip->i_mount, vn_remove);
688 xfs_inode_mark_reclaimable(ip);
696 struct xfs_inode *ip = XFS_I(inode);
697 struct xfs_mount *mp = ip->i_mount;
698 struct xfs_trans *tp;
700 if (!(inode->i_sb->s_flags & SB_LAZYTIME))
704 * Only do the timestamp update if the inode is dirty (I_DIRTY_SYNC)
705 * and has dirty timestamp (I_DIRTY_TIME). I_DIRTY_TIME can be passed
706 * in flags possibly together with I_DIRTY_SYNC.
708 if ((flags & ~I_DIRTY_TIME) != I_DIRTY_SYNC || !(flags & I_DIRTY_TIME))
711 if (xfs_trans_alloc(mp, &M_RES(mp)->tr_fsyncts, 0, 0, 0, &tp))
713 xfs_ilock(ip, XFS_ILOCK_EXCL);
714 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
715 xfs_trans_log_inode(tp, ip, XFS_ILOG_TIMESTAMP);
716 xfs_trans_commit(tp);
720 * Slab object creation initialisation for the XFS inode.
721 * This covers only the idempotent fields in the XFS inode;
722 * all other fields need to be initialised on allocation
723 * from the slab. This avoids the need to repeatedly initialise
724 * fields in the xfs inode that left in the initialise state
725 * when freeing the inode.
728 xfs_fs_inode_init_once(
731 struct xfs_inode *ip = inode;
733 memset(ip, 0, sizeof(struct xfs_inode));
736 inode_init_once(VFS_I(ip));
739 atomic_set(&ip->i_pincount, 0);
740 spin_lock_init(&ip->i_flags_lock);
742 mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
743 "xfsino", ip->i_ino);
747 * We do an unlocked check for XFS_IDONTCACHE here because we are already
748 * serialised against cache hits here via the inode->i_lock and igrab() in
749 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
750 * racing with us, and it avoids needing to grab a spinlock here for every inode
751 * we drop the final reference on.
757 struct xfs_inode *ip = XFS_I(inode);
760 * If this unlinked inode is in the middle of recovery, don't
761 * drop the inode just yet; log recovery will take care of
762 * that. See the comment for this inode flag.
764 if (ip->i_flags & XFS_IRECOVERY) {
765 ASSERT(xlog_recovery_needed(ip->i_mount->m_log));
769 return generic_drop_inode(inode);
774 struct xfs_mount *mp)
777 * Free the buftargs here because blkdev_put needs to be called outside
778 * of sb->s_umount, which is held around the call to ->put_super.
780 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp)
781 xfs_free_buftarg(mp->m_logdev_targp);
782 if (mp->m_rtdev_targp)
783 xfs_free_buftarg(mp->m_rtdev_targp);
784 if (mp->m_ddev_targp)
785 xfs_free_buftarg(mp->m_ddev_targp);
788 kfree(mp->m_logname);
794 struct super_block *sb,
797 struct xfs_mount *mp = XFS_M(sb);
800 trace_xfs_fs_sync_fs(mp, __return_address);
803 * Doing anything during the async pass would be counterproductive.
808 error = xfs_log_force(mp, XFS_LOG_SYNC);
814 * The disk must be active because we're syncing.
815 * We schedule log work now (now that the disk is
816 * active) instead of later (when it might not be).
818 flush_delayed_work(&mp->m_log->l_work);
822 * If we are called with page faults frozen out, it means we are about
823 * to freeze the transaction subsystem. Take the opportunity to shut
824 * down inodegc because once SB_FREEZE_FS is set it's too late to
825 * prevent inactivation races with freeze. The fs doesn't get called
826 * again by the freezing process until after SB_FREEZE_FS has been set,
827 * so it's now or never. Same logic applies to speculative allocation
828 * garbage collection.
830 * We don't care if this is a normal syncfs call that does this or
831 * freeze that does this - we can run this multiple times without issue
832 * and we won't race with a restart because a restart can only occur
833 * when the state is either SB_FREEZE_FS or SB_FREEZE_COMPLETE.
835 if (sb->s_writers.frozen == SB_FREEZE_PAGEFAULT) {
836 xfs_inodegc_stop(mp);
837 xfs_blockgc_stop(mp);
845 struct dentry *dentry,
846 struct kstatfs *statp)
848 struct xfs_mount *mp = XFS_M(dentry->d_sb);
849 xfs_sb_t *sbp = &mp->m_sb;
850 struct xfs_inode *ip = XFS_I(d_inode(dentry));
851 uint64_t fakeinos, id;
859 * Expedite background inodegc but don't wait. We do not want to block
860 * here waiting hours for a billion extent file to be truncated.
862 xfs_inodegc_push(mp);
864 statp->f_type = XFS_SUPER_MAGIC;
865 statp->f_namelen = MAXNAMELEN - 1;
867 id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
868 statp->f_fsid = u64_to_fsid(id);
870 icount = percpu_counter_sum(&mp->m_icount);
871 ifree = percpu_counter_sum(&mp->m_ifree);
872 fdblocks = percpu_counter_sum(&mp->m_fdblocks);
874 spin_lock(&mp->m_sb_lock);
875 statp->f_bsize = sbp->sb_blocksize;
876 lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
877 statp->f_blocks = sbp->sb_dblocks - lsize;
878 spin_unlock(&mp->m_sb_lock);
880 /* make sure statp->f_bfree does not underflow */
881 statp->f_bfree = max_t(int64_t, 0,
882 fdblocks - xfs_fdblocks_unavailable(mp));
883 statp->f_bavail = statp->f_bfree;
885 fakeinos = XFS_FSB_TO_INO(mp, statp->f_bfree);
886 statp->f_files = min(icount + fakeinos, (uint64_t)XFS_MAXINUMBER);
887 if (M_IGEO(mp)->maxicount)
888 statp->f_files = min_t(typeof(statp->f_files),
890 M_IGEO(mp)->maxicount);
892 /* If sb_icount overshot maxicount, report actual allocation */
893 statp->f_files = max_t(typeof(statp->f_files),
897 /* make sure statp->f_ffree does not underflow */
898 ffree = statp->f_files - (icount - ifree);
899 statp->f_ffree = max_t(int64_t, ffree, 0);
902 if ((ip->i_diflags & XFS_DIFLAG_PROJINHERIT) &&
903 ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
904 (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
905 xfs_qm_statvfs(ip, statp);
907 if (XFS_IS_REALTIME_MOUNT(mp) &&
908 (ip->i_diflags & (XFS_DIFLAG_RTINHERIT | XFS_DIFLAG_REALTIME))) {
911 statp->f_blocks = sbp->sb_rblocks;
912 freertx = percpu_counter_sum_positive(&mp->m_frextents);
913 statp->f_bavail = statp->f_bfree = freertx * sbp->sb_rextsize;
920 xfs_save_resvblks(struct xfs_mount *mp)
922 uint64_t resblks = 0;
924 mp->m_resblks_save = mp->m_resblks;
925 xfs_reserve_blocks(mp, &resblks, NULL);
929 xfs_restore_resvblks(struct xfs_mount *mp)
933 if (mp->m_resblks_save) {
934 resblks = mp->m_resblks_save;
935 mp->m_resblks_save = 0;
937 resblks = xfs_default_resblks(mp);
939 xfs_reserve_blocks(mp, &resblks, NULL);
943 * Second stage of a freeze. The data is already frozen so we only
944 * need to take care of the metadata. Once that's done sync the superblock
945 * to the log to dirty it in case of a crash while frozen. This ensures that we
946 * will recover the unlinked inode lists on the next mount.
950 struct super_block *sb)
952 struct xfs_mount *mp = XFS_M(sb);
957 * The filesystem is now frozen far enough that memory reclaim
958 * cannot safely operate on the filesystem. Hence we need to
959 * set a GFP_NOFS context here to avoid recursion deadlocks.
961 flags = memalloc_nofs_save();
962 xfs_save_resvblks(mp);
963 ret = xfs_log_quiesce(mp);
964 memalloc_nofs_restore(flags);
967 * For read-write filesystems, we need to restart the inodegc on error
968 * because we stopped it at SB_FREEZE_PAGEFAULT level and a thaw is not
969 * going to be run to restart it now. We are at SB_FREEZE_FS level
970 * here, so we can restart safely without racing with a stop in
973 if (ret && !xfs_is_readonly(mp)) {
974 xfs_blockgc_start(mp);
975 xfs_inodegc_start(mp);
983 struct super_block *sb)
985 struct xfs_mount *mp = XFS_M(sb);
987 xfs_restore_resvblks(mp);
988 xfs_log_work_queue(mp);
991 * Don't reactivate the inodegc worker on a readonly filesystem because
992 * inodes are sent directly to reclaim. Don't reactivate the blockgc
993 * worker because there are no speculative preallocations on a readonly
996 if (!xfs_is_readonly(mp)) {
997 xfs_blockgc_start(mp);
998 xfs_inodegc_start(mp);
1005 * This function fills in xfs_mount_t fields based on mount args.
1006 * Note: the superblock _has_ now been read in.
1010 struct xfs_mount *mp)
1012 /* Fail a mount where the logbuf is smaller than the log stripe */
1013 if (xfs_has_logv2(mp)) {
1014 if (mp->m_logbsize <= 0 &&
1015 mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1016 mp->m_logbsize = mp->m_sb.sb_logsunit;
1017 } else if (mp->m_logbsize > 0 &&
1018 mp->m_logbsize < mp->m_sb.sb_logsunit) {
1020 "logbuf size must be greater than or equal to log stripe size");
1024 /* Fail a mount if the logbuf is larger than 32K */
1025 if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1027 "logbuf size for version 1 logs must be 16K or 32K");
1033 * V5 filesystems always use attr2 format for attributes.
1035 if (xfs_has_crc(mp) && xfs_has_noattr2(mp)) {
1036 xfs_warn(mp, "Cannot mount a V5 filesystem as noattr2. "
1037 "attr2 is always enabled for V5 filesystems.");
1042 * prohibit r/w mounts of read-only filesystems
1044 if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !xfs_is_readonly(mp)) {
1046 "cannot mount a read-only filesystem as read-write");
1050 if ((mp->m_qflags & XFS_GQUOTA_ACCT) &&
1051 (mp->m_qflags & XFS_PQUOTA_ACCT) &&
1052 !xfs_has_pquotino(mp)) {
1054 "Super block does not support project and group quota together");
1062 xfs_init_percpu_counters(
1063 struct xfs_mount *mp)
1067 error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL);
1071 error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL);
1075 error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL);
1079 error = percpu_counter_init(&mp->m_delalloc_blks, 0, GFP_KERNEL);
1083 error = percpu_counter_init(&mp->m_frextents, 0, GFP_KERNEL);
1090 percpu_counter_destroy(&mp->m_delalloc_blks);
1092 percpu_counter_destroy(&mp->m_fdblocks);
1094 percpu_counter_destroy(&mp->m_ifree);
1096 percpu_counter_destroy(&mp->m_icount);
1101 xfs_reinit_percpu_counters(
1102 struct xfs_mount *mp)
1104 percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount);
1105 percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree);
1106 percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks);
1107 percpu_counter_set(&mp->m_frextents, mp->m_sb.sb_frextents);
1111 xfs_destroy_percpu_counters(
1112 struct xfs_mount *mp)
1114 percpu_counter_destroy(&mp->m_icount);
1115 percpu_counter_destroy(&mp->m_ifree);
1116 percpu_counter_destroy(&mp->m_fdblocks);
1117 ASSERT(xfs_is_shutdown(mp) ||
1118 percpu_counter_sum(&mp->m_delalloc_blks) == 0);
1119 percpu_counter_destroy(&mp->m_delalloc_blks);
1120 percpu_counter_destroy(&mp->m_frextents);
1124 xfs_inodegc_init_percpu(
1125 struct xfs_mount *mp)
1127 struct xfs_inodegc *gc;
1130 mp->m_inodegc = alloc_percpu(struct xfs_inodegc);
1134 for_each_possible_cpu(cpu) {
1135 gc = per_cpu_ptr(mp->m_inodegc, cpu);
1136 #if defined(DEBUG) || defined(XFS_WARN)
1139 init_llist_head(&gc->list);
1142 INIT_DELAYED_WORK(&gc->work, xfs_inodegc_worker);
1148 xfs_inodegc_free_percpu(
1149 struct xfs_mount *mp)
1153 free_percpu(mp->m_inodegc);
1158 struct super_block *sb)
1160 struct xfs_mount *mp = XFS_M(sb);
1162 xfs_notice(mp, "Unmounting Filesystem %pU", &mp->m_sb.sb_uuid);
1163 xfs_filestream_unmount(mp);
1167 free_percpu(mp->m_stats.xs_stats);
1168 xfs_mount_list_del(mp);
1169 xfs_inodegc_free_percpu(mp);
1170 xfs_destroy_percpu_counters(mp);
1171 xfs_destroy_mount_workqueues(mp);
1172 xfs_shutdown_devices(mp);
1176 xfs_fs_nr_cached_objects(
1177 struct super_block *sb,
1178 struct shrink_control *sc)
1180 /* Paranoia: catch incorrect calls during mount setup or teardown */
1181 if (WARN_ON_ONCE(!sb->s_fs_info))
1183 return xfs_reclaim_inodes_count(XFS_M(sb));
1187 xfs_fs_free_cached_objects(
1188 struct super_block *sb,
1189 struct shrink_control *sc)
1191 return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan);
1196 struct super_block *sb)
1198 xfs_force_shutdown(XFS_M(sb), SHUTDOWN_DEVICE_REMOVED);
1201 static const struct super_operations xfs_super_operations = {
1202 .alloc_inode = xfs_fs_alloc_inode,
1203 .destroy_inode = xfs_fs_destroy_inode,
1204 .dirty_inode = xfs_fs_dirty_inode,
1205 .drop_inode = xfs_fs_drop_inode,
1206 .put_super = xfs_fs_put_super,
1207 .sync_fs = xfs_fs_sync_fs,
1208 .freeze_fs = xfs_fs_freeze,
1209 .unfreeze_fs = xfs_fs_unfreeze,
1210 .statfs = xfs_fs_statfs,
1211 .show_options = xfs_fs_show_options,
1212 .nr_cached_objects = xfs_fs_nr_cached_objects,
1213 .free_cached_objects = xfs_fs_free_cached_objects,
1214 .shutdown = xfs_fs_shutdown,
1223 int last, shift_left_factor = 0, _res;
1227 value = kstrdup(s, GFP_KERNEL);
1231 last = strlen(value) - 1;
1232 if (value[last] == 'K' || value[last] == 'k') {
1233 shift_left_factor = 10;
1236 if (value[last] == 'M' || value[last] == 'm') {
1237 shift_left_factor = 20;
1240 if (value[last] == 'G' || value[last] == 'g') {
1241 shift_left_factor = 30;
1245 if (kstrtoint(value, base, &_res))
1248 *res = _res << shift_left_factor;
1253 xfs_fs_warn_deprecated(
1254 struct fs_context *fc,
1255 struct fs_parameter *param,
1259 /* Don't print the warning if reconfiguring and current mount point
1260 * already had the flag set
1262 if ((fc->purpose & FS_CONTEXT_FOR_RECONFIGURE) &&
1263 !!(XFS_M(fc->root->d_sb)->m_features & flag) == value)
1265 xfs_warn(fc->s_fs_info, "%s mount option is deprecated.", param->key);
1269 * Set mount state from a mount option.
1271 * NOTE: mp->m_super is NULL here!
1275 struct fs_context *fc,
1276 struct fs_parameter *param)
1278 struct xfs_mount *parsing_mp = fc->s_fs_info;
1279 struct fs_parse_result result;
1283 opt = fs_parse(fc, xfs_fs_parameters, param, &result);
1289 parsing_mp->m_logbufs = result.uint_32;
1292 if (suffix_kstrtoint(param->string, 10, &parsing_mp->m_logbsize))
1296 kfree(parsing_mp->m_logname);
1297 parsing_mp->m_logname = kstrdup(param->string, GFP_KERNEL);
1298 if (!parsing_mp->m_logname)
1302 kfree(parsing_mp->m_rtname);
1303 parsing_mp->m_rtname = kstrdup(param->string, GFP_KERNEL);
1304 if (!parsing_mp->m_rtname)
1308 if (suffix_kstrtoint(param->string, 10, &size))
1310 parsing_mp->m_allocsize_log = ffs(size) - 1;
1311 parsing_mp->m_features |= XFS_FEAT_ALLOCSIZE;
1315 parsing_mp->m_features |= XFS_FEAT_GRPID;
1318 case Opt_sysvgroups:
1319 parsing_mp->m_features &= ~XFS_FEAT_GRPID;
1322 parsing_mp->m_features |= XFS_FEAT_WSYNC;
1324 case Opt_norecovery:
1325 parsing_mp->m_features |= XFS_FEAT_NORECOVERY;
1328 parsing_mp->m_features |= XFS_FEAT_NOALIGN;
1331 parsing_mp->m_features |= XFS_FEAT_SWALLOC;
1334 parsing_mp->m_dalign = result.uint_32;
1337 parsing_mp->m_swidth = result.uint_32;
1340 parsing_mp->m_features |= XFS_FEAT_SMALL_INUMS;
1343 parsing_mp->m_features &= ~XFS_FEAT_SMALL_INUMS;
1346 parsing_mp->m_features |= XFS_FEAT_NOUUID;
1349 parsing_mp->m_features |= XFS_FEAT_LARGE_IOSIZE;
1352 parsing_mp->m_features &= ~XFS_FEAT_LARGE_IOSIZE;
1354 case Opt_filestreams:
1355 parsing_mp->m_features |= XFS_FEAT_FILESTREAMS;
1358 parsing_mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
1359 parsing_mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
1364 parsing_mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ENFD);
1366 case Opt_qnoenforce:
1367 case Opt_uqnoenforce:
1368 parsing_mp->m_qflags |= XFS_UQUOTA_ACCT;
1369 parsing_mp->m_qflags &= ~XFS_UQUOTA_ENFD;
1373 parsing_mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ENFD);
1375 case Opt_pqnoenforce:
1376 parsing_mp->m_qflags |= XFS_PQUOTA_ACCT;
1377 parsing_mp->m_qflags &= ~XFS_PQUOTA_ENFD;
1381 parsing_mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ENFD);
1383 case Opt_gqnoenforce:
1384 parsing_mp->m_qflags |= XFS_GQUOTA_ACCT;
1385 parsing_mp->m_qflags &= ~XFS_GQUOTA_ENFD;
1388 parsing_mp->m_features |= XFS_FEAT_DISCARD;
1391 parsing_mp->m_features &= ~XFS_FEAT_DISCARD;
1393 #ifdef CONFIG_FS_DAX
1395 xfs_mount_set_dax_mode(parsing_mp, XFS_DAX_ALWAYS);
1398 xfs_mount_set_dax_mode(parsing_mp, result.uint_32);
1401 /* Following mount options will be removed in September 2025 */
1403 xfs_fs_warn_deprecated(fc, param, XFS_FEAT_IKEEP, true);
1404 parsing_mp->m_features |= XFS_FEAT_IKEEP;
1407 xfs_fs_warn_deprecated(fc, param, XFS_FEAT_IKEEP, false);
1408 parsing_mp->m_features &= ~XFS_FEAT_IKEEP;
1411 xfs_fs_warn_deprecated(fc, param, XFS_FEAT_ATTR2, true);
1412 parsing_mp->m_features |= XFS_FEAT_ATTR2;
1415 xfs_fs_warn_deprecated(fc, param, XFS_FEAT_NOATTR2, true);
1416 parsing_mp->m_features |= XFS_FEAT_NOATTR2;
1419 xfs_warn(parsing_mp, "unknown mount option [%s].", param->key);
1427 xfs_fs_validate_params(
1428 struct xfs_mount *mp)
1430 /* No recovery flag requires a read-only mount */
1431 if (xfs_has_norecovery(mp) && !xfs_is_readonly(mp)) {
1432 xfs_warn(mp, "no-recovery mounts must be read-only.");
1437 * We have not read the superblock at this point, so only the attr2
1438 * mount option can set the attr2 feature by this stage.
1440 if (xfs_has_attr2(mp) && xfs_has_noattr2(mp)) {
1441 xfs_warn(mp, "attr2 and noattr2 cannot both be specified.");
1446 if (xfs_has_noalign(mp) && (mp->m_dalign || mp->m_swidth)) {
1448 "sunit and swidth options incompatible with the noalign option");
1452 if (!IS_ENABLED(CONFIG_XFS_QUOTA) && mp->m_qflags != 0) {
1453 xfs_warn(mp, "quota support not available in this kernel.");
1457 if ((mp->m_dalign && !mp->m_swidth) ||
1458 (!mp->m_dalign && mp->m_swidth)) {
1459 xfs_warn(mp, "sunit and swidth must be specified together");
1463 if (mp->m_dalign && (mp->m_swidth % mp->m_dalign != 0)) {
1465 "stripe width (%d) must be a multiple of the stripe unit (%d)",
1466 mp->m_swidth, mp->m_dalign);
1470 if (mp->m_logbufs != -1 &&
1471 mp->m_logbufs != 0 &&
1472 (mp->m_logbufs < XLOG_MIN_ICLOGS ||
1473 mp->m_logbufs > XLOG_MAX_ICLOGS)) {
1474 xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
1475 mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
1479 if (mp->m_logbsize != -1 &&
1480 mp->m_logbsize != 0 &&
1481 (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
1482 mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
1483 !is_power_of_2(mp->m_logbsize))) {
1485 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
1490 if (xfs_has_allocsize(mp) &&
1491 (mp->m_allocsize_log > XFS_MAX_IO_LOG ||
1492 mp->m_allocsize_log < XFS_MIN_IO_LOG)) {
1493 xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
1494 mp->m_allocsize_log, XFS_MIN_IO_LOG, XFS_MAX_IO_LOG);
1503 struct super_block *sb,
1504 struct fs_context *fc)
1506 struct xfs_mount *mp = sb->s_fs_info;
1508 int flags = 0, error;
1512 error = xfs_fs_validate_params(mp);
1516 sb_min_blocksize(sb, BBSIZE);
1517 sb->s_xattr = xfs_xattr_handlers;
1518 sb->s_export_op = &xfs_export_operations;
1519 #ifdef CONFIG_XFS_QUOTA
1520 sb->s_qcop = &xfs_quotactl_operations;
1521 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
1523 sb->s_op = &xfs_super_operations;
1526 * Delay mount work if the debug hook is set. This is debug
1527 * instrumention to coordinate simulation of xfs mount failures with
1528 * VFS superblock operations
1530 if (xfs_globals.mount_delay) {
1531 xfs_notice(mp, "Delaying mount for %d seconds.",
1532 xfs_globals.mount_delay);
1533 msleep(xfs_globals.mount_delay * 1000);
1536 if (fc->sb_flags & SB_SILENT)
1537 flags |= XFS_MFSI_QUIET;
1539 error = xfs_open_devices(mp);
1543 error = xfs_init_mount_workqueues(mp);
1545 goto out_shutdown_devices;
1547 error = xfs_init_percpu_counters(mp);
1549 goto out_destroy_workqueues;
1551 error = xfs_inodegc_init_percpu(mp);
1553 goto out_destroy_counters;
1556 * All percpu data structures requiring cleanup when a cpu goes offline
1557 * must be allocated before adding this @mp to the cpu-dead handler's
1560 xfs_mount_list_add(mp);
1562 /* Allocate stats memory before we do operations that might use it */
1563 mp->m_stats.xs_stats = alloc_percpu(struct xfsstats);
1564 if (!mp->m_stats.xs_stats) {
1566 goto out_destroy_inodegc;
1569 error = xfs_readsb(mp, flags);
1571 goto out_free_stats;
1573 error = xfs_finish_flags(mp);
1577 error = xfs_setup_devices(mp);
1581 /* V4 support is undergoing deprecation. */
1582 if (!xfs_has_crc(mp)) {
1583 #ifdef CONFIG_XFS_SUPPORT_V4
1585 "Deprecated V4 format (crc=0) will not be supported after September 2030.");
1588 "Deprecated V4 format (crc=0) not supported by kernel.");
1594 /* ASCII case insensitivity is undergoing deprecation. */
1595 if (xfs_has_asciici(mp)) {
1596 #ifdef CONFIG_XFS_SUPPORT_ASCII_CI
1598 "Deprecated ASCII case-insensitivity feature (ascii-ci=1) will not be supported after September 2030.");
1601 "Deprecated ASCII case-insensitivity feature (ascii-ci=1) not supported by kernel.");
1607 /* Filesystem claims it needs repair, so refuse the mount. */
1608 if (xfs_has_needsrepair(mp)) {
1609 xfs_warn(mp, "Filesystem needs repair. Please run xfs_repair.");
1610 error = -EFSCORRUPTED;
1615 * Don't touch the filesystem if a user tool thinks it owns the primary
1616 * superblock. mkfs doesn't clear the flag from secondary supers, so
1617 * we don't check them at all.
1619 if (mp->m_sb.sb_inprogress) {
1620 xfs_warn(mp, "Offline file system operation in progress!");
1621 error = -EFSCORRUPTED;
1626 * Until this is fixed only page-sized or smaller data blocks work.
1628 if (mp->m_sb.sb_blocksize > PAGE_SIZE) {
1630 "File system with blocksize %d bytes. "
1631 "Only pagesize (%ld) or less will currently work.",
1632 mp->m_sb.sb_blocksize, PAGE_SIZE);
1637 /* Ensure this filesystem fits in the page cache limits */
1638 if (xfs_sb_validate_fsb_count(&mp->m_sb, mp->m_sb.sb_dblocks) ||
1639 xfs_sb_validate_fsb_count(&mp->m_sb, mp->m_sb.sb_rblocks)) {
1641 "file system too large to be mounted on this system.");
1647 * XFS block mappings use 54 bits to store the logical block offset.
1648 * This should suffice to handle the maximum file size that the VFS
1649 * supports (currently 2^63 bytes on 64-bit and ULONG_MAX << PAGE_SHIFT
1650 * bytes on 32-bit), but as XFS and VFS have gotten the s_maxbytes
1651 * calculation wrong on 32-bit kernels in the past, we'll add a WARN_ON
1652 * to check this assertion.
1654 * Avoid integer overflow by comparing the maximum bmbt offset to the
1655 * maximum pagecache offset in units of fs blocks.
1657 if (!xfs_verify_fileoff(mp, XFS_B_TO_FSBT(mp, MAX_LFS_FILESIZE))) {
1659 "MAX_LFS_FILESIZE block offset (%llu) exceeds extent map maximum (%llu)!",
1660 XFS_B_TO_FSBT(mp, MAX_LFS_FILESIZE),
1666 error = xfs_filestream_mount(mp);
1671 * we must configure the block size in the superblock before we run the
1672 * full mount process as the mount process can lookup and cache inodes.
1674 sb->s_magic = XFS_SUPER_MAGIC;
1675 sb->s_blocksize = mp->m_sb.sb_blocksize;
1676 sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1677 sb->s_maxbytes = MAX_LFS_FILESIZE;
1678 sb->s_max_links = XFS_MAXLINK;
1679 sb->s_time_gran = 1;
1680 if (xfs_has_bigtime(mp)) {
1681 sb->s_time_min = xfs_bigtime_to_unix(XFS_BIGTIME_TIME_MIN);
1682 sb->s_time_max = xfs_bigtime_to_unix(XFS_BIGTIME_TIME_MAX);
1684 sb->s_time_min = XFS_LEGACY_TIME_MIN;
1685 sb->s_time_max = XFS_LEGACY_TIME_MAX;
1687 trace_xfs_inode_timestamp_range(mp, sb->s_time_min, sb->s_time_max);
1688 sb->s_iflags |= SB_I_CGROUPWB;
1690 set_posix_acl_flag(sb);
1692 /* version 5 superblocks support inode version counters. */
1693 if (xfs_has_crc(mp))
1694 sb->s_flags |= SB_I_VERSION;
1696 if (xfs_has_dax_always(mp)) {
1697 error = xfs_setup_dax_always(mp);
1699 goto out_filestream_unmount;
1702 if (xfs_has_discard(mp) && !bdev_max_discard_sectors(sb->s_bdev)) {
1704 "mounting with \"discard\" option, but the device does not support discard");
1705 mp->m_features &= ~XFS_FEAT_DISCARD;
1708 if (xfs_has_reflink(mp)) {
1709 if (mp->m_sb.sb_rblocks) {
1711 "reflink not compatible with realtime device!");
1713 goto out_filestream_unmount;
1716 if (xfs_globals.always_cow) {
1717 xfs_info(mp, "using DEBUG-only always_cow mode.");
1718 mp->m_always_cow = true;
1722 if (xfs_has_rmapbt(mp) && mp->m_sb.sb_rblocks) {
1724 "reverse mapping btree not compatible with realtime device!");
1726 goto out_filestream_unmount;
1729 error = xfs_mountfs(mp);
1731 goto out_filestream_unmount;
1733 root = igrab(VFS_I(mp->m_rootip));
1738 sb->s_root = d_make_root(root);
1746 out_filestream_unmount:
1747 xfs_filestream_unmount(mp);
1751 free_percpu(mp->m_stats.xs_stats);
1752 out_destroy_inodegc:
1753 xfs_mount_list_del(mp);
1754 xfs_inodegc_free_percpu(mp);
1755 out_destroy_counters:
1756 xfs_destroy_percpu_counters(mp);
1757 out_destroy_workqueues:
1758 xfs_destroy_mount_workqueues(mp);
1759 out_shutdown_devices:
1760 xfs_shutdown_devices(mp);
1764 xfs_filestream_unmount(mp);
1771 struct fs_context *fc)
1773 return get_tree_bdev(fc, xfs_fs_fill_super);
1778 struct xfs_mount *mp)
1780 struct xfs_sb *sbp = &mp->m_sb;
1783 if (xfs_has_norecovery(mp)) {
1785 "ro->rw transition prohibited on norecovery mount");
1789 if (xfs_sb_is_v5(sbp) &&
1790 xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) {
1792 "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem",
1793 (sbp->sb_features_ro_compat &
1794 XFS_SB_FEAT_RO_COMPAT_UNKNOWN));
1798 clear_bit(XFS_OPSTATE_READONLY, &mp->m_opstate);
1801 * If this is the first remount to writeable state we might have some
1802 * superblock changes to update.
1804 if (mp->m_update_sb) {
1805 error = xfs_sync_sb(mp, false);
1807 xfs_warn(mp, "failed to write sb changes");
1810 mp->m_update_sb = false;
1814 * Fill out the reserve pool if it is empty. Use the stashed value if
1815 * it is non-zero, otherwise go with the default.
1817 xfs_restore_resvblks(mp);
1818 xfs_log_work_queue(mp);
1819 xfs_blockgc_start(mp);
1821 /* Create the per-AG metadata reservation pool .*/
1822 error = xfs_fs_reserve_ag_blocks(mp);
1823 if (error && error != -ENOSPC)
1826 /* Re-enable the background inode inactivation worker. */
1827 xfs_inodegc_start(mp);
1834 struct xfs_mount *mp)
1836 struct xfs_icwalk icw = {
1837 .icw_flags = XFS_ICWALK_FLAG_SYNC,
1841 /* Flush all the dirty data to disk. */
1842 error = sync_filesystem(mp->m_super);
1847 * Cancel background eofb scanning so it cannot race with the final
1848 * log force+buftarg wait and deadlock the remount.
1850 xfs_blockgc_stop(mp);
1853 * Clear out all remaining COW staging extents and speculative post-EOF
1854 * preallocations so that we don't leave inodes requiring inactivation
1855 * cleanups during reclaim on a read-only mount. We must process every
1856 * cached inode, so this requires a synchronous cache scan.
1858 error = xfs_blockgc_free_space(mp, &icw);
1860 xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
1865 * Stop the inodegc background worker. xfs_fs_reconfigure already
1866 * flushed all pending inodegc work when it sync'd the filesystem.
1867 * The VFS holds s_umount, so we know that inodes cannot enter
1868 * xfs_fs_destroy_inode during a remount operation. In readonly mode
1869 * we send inodes straight to reclaim, so no inodes will be queued.
1871 xfs_inodegc_stop(mp);
1873 /* Free the per-AG metadata reservation pool. */
1874 error = xfs_fs_unreserve_ag_blocks(mp);
1876 xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE);
1881 * Before we sync the metadata, we need to free up the reserve block
1882 * pool so that the used block count in the superblock on disk is
1883 * correct at the end of the remount. Stash the current* reserve pool
1884 * size so that if we get remounted rw, we can return it to the same
1887 xfs_save_resvblks(mp);
1890 set_bit(XFS_OPSTATE_READONLY, &mp->m_opstate);
1896 * Logically we would return an error here to prevent users from believing
1897 * they might have changed mount options using remount which can't be changed.
1899 * But unfortunately mount(8) adds all options from mtab and fstab to the mount
1900 * arguments in some cases so we can't blindly reject options, but have to
1901 * check for each specified option if it actually differs from the currently
1902 * set option and only reject it if that's the case.
1904 * Until that is implemented we return success for every remount request, and
1905 * silently ignore all options that we can't actually change.
1909 struct fs_context *fc)
1911 struct xfs_mount *mp = XFS_M(fc->root->d_sb);
1912 struct xfs_mount *new_mp = fc->s_fs_info;
1913 int flags = fc->sb_flags;
1916 /* version 5 superblocks always support version counters. */
1917 if (xfs_has_crc(mp))
1918 fc->sb_flags |= SB_I_VERSION;
1920 error = xfs_fs_validate_params(new_mp);
1924 /* inode32 -> inode64 */
1925 if (xfs_has_small_inums(mp) && !xfs_has_small_inums(new_mp)) {
1926 mp->m_features &= ~XFS_FEAT_SMALL_INUMS;
1927 mp->m_maxagi = xfs_set_inode_alloc(mp, mp->m_sb.sb_agcount);
1930 /* inode64 -> inode32 */
1931 if (!xfs_has_small_inums(mp) && xfs_has_small_inums(new_mp)) {
1932 mp->m_features |= XFS_FEAT_SMALL_INUMS;
1933 mp->m_maxagi = xfs_set_inode_alloc(mp, mp->m_sb.sb_agcount);
1937 if (xfs_is_readonly(mp) && !(flags & SB_RDONLY)) {
1938 error = xfs_remount_rw(mp);
1944 if (!xfs_is_readonly(mp) && (flags & SB_RDONLY)) {
1945 error = xfs_remount_ro(mp);
1955 struct fs_context *fc)
1957 struct xfs_mount *mp = fc->s_fs_info;
1960 * mp is stored in the fs_context when it is initialized.
1961 * mp is transferred to the superblock on a successful mount,
1962 * but if an error occurs before the transfer we have to free
1969 static const struct fs_context_operations xfs_context_ops = {
1970 .parse_param = xfs_fs_parse_param,
1971 .get_tree = xfs_fs_get_tree,
1972 .reconfigure = xfs_fs_reconfigure,
1973 .free = xfs_fs_free,
1976 static int xfs_init_fs_context(
1977 struct fs_context *fc)
1979 struct xfs_mount *mp;
1981 mp = kmem_alloc(sizeof(struct xfs_mount), KM_ZERO);
1985 spin_lock_init(&mp->m_sb_lock);
1986 INIT_RADIX_TREE(&mp->m_perag_tree, GFP_ATOMIC);
1987 spin_lock_init(&mp->m_perag_lock);
1988 mutex_init(&mp->m_growlock);
1989 INIT_WORK(&mp->m_flush_inodes_work, xfs_flush_inodes_worker);
1990 INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1991 mp->m_kobj.kobject.kset = xfs_kset;
1993 * We don't create the finobt per-ag space reservation until after log
1994 * recovery, so we must set this to true so that an ifree transaction
1995 * started during log recovery will not depend on space reservations
1996 * for finobt expansion.
1998 mp->m_finobt_nores = true;
2001 * These can be overridden by the mount option parsing.
2004 mp->m_logbsize = -1;
2005 mp->m_allocsize_log = 16; /* 64k */
2008 * Copy binary VFS mount flags we are interested in.
2010 if (fc->sb_flags & SB_RDONLY)
2011 set_bit(XFS_OPSTATE_READONLY, &mp->m_opstate);
2012 if (fc->sb_flags & SB_DIRSYNC)
2013 mp->m_features |= XFS_FEAT_DIRSYNC;
2014 if (fc->sb_flags & SB_SYNCHRONOUS)
2015 mp->m_features |= XFS_FEAT_WSYNC;
2018 fc->ops = &xfs_context_ops;
2025 struct super_block *sb)
2027 kill_block_super(sb);
2028 xfs_mount_free(XFS_M(sb));
2031 static struct file_system_type xfs_fs_type = {
2032 .owner = THIS_MODULE,
2034 .init_fs_context = xfs_init_fs_context,
2035 .parameters = xfs_fs_parameters,
2036 .kill_sb = xfs_kill_sb,
2037 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
2039 MODULE_ALIAS_FS("xfs");
2042 xfs_init_caches(void)
2046 xfs_buf_cache = kmem_cache_create("xfs_buf", sizeof(struct xfs_buf), 0,
2047 SLAB_HWCACHE_ALIGN |
2048 SLAB_RECLAIM_ACCOUNT |
2054 xfs_log_ticket_cache = kmem_cache_create("xfs_log_ticket",
2055 sizeof(struct xlog_ticket),
2057 if (!xfs_log_ticket_cache)
2058 goto out_destroy_buf_cache;
2060 error = xfs_btree_init_cur_caches();
2062 goto out_destroy_log_ticket_cache;
2064 error = xfs_defer_init_item_caches();
2066 goto out_destroy_btree_cur_cache;
2068 xfs_da_state_cache = kmem_cache_create("xfs_da_state",
2069 sizeof(struct xfs_da_state),
2071 if (!xfs_da_state_cache)
2072 goto out_destroy_defer_item_cache;
2074 xfs_ifork_cache = kmem_cache_create("xfs_ifork",
2075 sizeof(struct xfs_ifork),
2077 if (!xfs_ifork_cache)
2078 goto out_destroy_da_state_cache;
2080 xfs_trans_cache = kmem_cache_create("xfs_trans",
2081 sizeof(struct xfs_trans),
2083 if (!xfs_trans_cache)
2084 goto out_destroy_ifork_cache;
2088 * The size of the cache-allocated buf log item is the maximum
2089 * size possible under XFS. This wastes a little bit of memory,
2090 * but it is much faster.
2092 xfs_buf_item_cache = kmem_cache_create("xfs_buf_item",
2093 sizeof(struct xfs_buf_log_item),
2095 if (!xfs_buf_item_cache)
2096 goto out_destroy_trans_cache;
2098 xfs_efd_cache = kmem_cache_create("xfs_efd_item",
2099 xfs_efd_log_item_sizeof(XFS_EFD_MAX_FAST_EXTENTS),
2102 goto out_destroy_buf_item_cache;
2104 xfs_efi_cache = kmem_cache_create("xfs_efi_item",
2105 xfs_efi_log_item_sizeof(XFS_EFI_MAX_FAST_EXTENTS),
2108 goto out_destroy_efd_cache;
2110 xfs_inode_cache = kmem_cache_create("xfs_inode",
2111 sizeof(struct xfs_inode), 0,
2112 (SLAB_HWCACHE_ALIGN |
2113 SLAB_RECLAIM_ACCOUNT |
2114 SLAB_MEM_SPREAD | SLAB_ACCOUNT),
2115 xfs_fs_inode_init_once);
2116 if (!xfs_inode_cache)
2117 goto out_destroy_efi_cache;
2119 xfs_ili_cache = kmem_cache_create("xfs_ili",
2120 sizeof(struct xfs_inode_log_item), 0,
2121 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD,
2124 goto out_destroy_inode_cache;
2126 xfs_icreate_cache = kmem_cache_create("xfs_icr",
2127 sizeof(struct xfs_icreate_item),
2129 if (!xfs_icreate_cache)
2130 goto out_destroy_ili_cache;
2132 xfs_rud_cache = kmem_cache_create("xfs_rud_item",
2133 sizeof(struct xfs_rud_log_item),
2136 goto out_destroy_icreate_cache;
2138 xfs_rui_cache = kmem_cache_create("xfs_rui_item",
2139 xfs_rui_log_item_sizeof(XFS_RUI_MAX_FAST_EXTENTS),
2142 goto out_destroy_rud_cache;
2144 xfs_cud_cache = kmem_cache_create("xfs_cud_item",
2145 sizeof(struct xfs_cud_log_item),
2148 goto out_destroy_rui_cache;
2150 xfs_cui_cache = kmem_cache_create("xfs_cui_item",
2151 xfs_cui_log_item_sizeof(XFS_CUI_MAX_FAST_EXTENTS),
2154 goto out_destroy_cud_cache;
2156 xfs_bud_cache = kmem_cache_create("xfs_bud_item",
2157 sizeof(struct xfs_bud_log_item),
2160 goto out_destroy_cui_cache;
2162 xfs_bui_cache = kmem_cache_create("xfs_bui_item",
2163 xfs_bui_log_item_sizeof(XFS_BUI_MAX_FAST_EXTENTS),
2166 goto out_destroy_bud_cache;
2168 xfs_attrd_cache = kmem_cache_create("xfs_attrd_item",
2169 sizeof(struct xfs_attrd_log_item),
2171 if (!xfs_attrd_cache)
2172 goto out_destroy_bui_cache;
2174 xfs_attri_cache = kmem_cache_create("xfs_attri_item",
2175 sizeof(struct xfs_attri_log_item),
2177 if (!xfs_attri_cache)
2178 goto out_destroy_attrd_cache;
2180 xfs_iunlink_cache = kmem_cache_create("xfs_iul_item",
2181 sizeof(struct xfs_iunlink_item),
2183 if (!xfs_iunlink_cache)
2184 goto out_destroy_attri_cache;
2188 out_destroy_attri_cache:
2189 kmem_cache_destroy(xfs_attri_cache);
2190 out_destroy_attrd_cache:
2191 kmem_cache_destroy(xfs_attrd_cache);
2192 out_destroy_bui_cache:
2193 kmem_cache_destroy(xfs_bui_cache);
2194 out_destroy_bud_cache:
2195 kmem_cache_destroy(xfs_bud_cache);
2196 out_destroy_cui_cache:
2197 kmem_cache_destroy(xfs_cui_cache);
2198 out_destroy_cud_cache:
2199 kmem_cache_destroy(xfs_cud_cache);
2200 out_destroy_rui_cache:
2201 kmem_cache_destroy(xfs_rui_cache);
2202 out_destroy_rud_cache:
2203 kmem_cache_destroy(xfs_rud_cache);
2204 out_destroy_icreate_cache:
2205 kmem_cache_destroy(xfs_icreate_cache);
2206 out_destroy_ili_cache:
2207 kmem_cache_destroy(xfs_ili_cache);
2208 out_destroy_inode_cache:
2209 kmem_cache_destroy(xfs_inode_cache);
2210 out_destroy_efi_cache:
2211 kmem_cache_destroy(xfs_efi_cache);
2212 out_destroy_efd_cache:
2213 kmem_cache_destroy(xfs_efd_cache);
2214 out_destroy_buf_item_cache:
2215 kmem_cache_destroy(xfs_buf_item_cache);
2216 out_destroy_trans_cache:
2217 kmem_cache_destroy(xfs_trans_cache);
2218 out_destroy_ifork_cache:
2219 kmem_cache_destroy(xfs_ifork_cache);
2220 out_destroy_da_state_cache:
2221 kmem_cache_destroy(xfs_da_state_cache);
2222 out_destroy_defer_item_cache:
2223 xfs_defer_destroy_item_caches();
2224 out_destroy_btree_cur_cache:
2225 xfs_btree_destroy_cur_caches();
2226 out_destroy_log_ticket_cache:
2227 kmem_cache_destroy(xfs_log_ticket_cache);
2228 out_destroy_buf_cache:
2229 kmem_cache_destroy(xfs_buf_cache);
2235 xfs_destroy_caches(void)
2238 * Make sure all delayed rcu free are flushed before we
2242 kmem_cache_destroy(xfs_iunlink_cache);
2243 kmem_cache_destroy(xfs_attri_cache);
2244 kmem_cache_destroy(xfs_attrd_cache);
2245 kmem_cache_destroy(xfs_bui_cache);
2246 kmem_cache_destroy(xfs_bud_cache);
2247 kmem_cache_destroy(xfs_cui_cache);
2248 kmem_cache_destroy(xfs_cud_cache);
2249 kmem_cache_destroy(xfs_rui_cache);
2250 kmem_cache_destroy(xfs_rud_cache);
2251 kmem_cache_destroy(xfs_icreate_cache);
2252 kmem_cache_destroy(xfs_ili_cache);
2253 kmem_cache_destroy(xfs_inode_cache);
2254 kmem_cache_destroy(xfs_efi_cache);
2255 kmem_cache_destroy(xfs_efd_cache);
2256 kmem_cache_destroy(xfs_buf_item_cache);
2257 kmem_cache_destroy(xfs_trans_cache);
2258 kmem_cache_destroy(xfs_ifork_cache);
2259 kmem_cache_destroy(xfs_da_state_cache);
2260 xfs_defer_destroy_item_caches();
2261 xfs_btree_destroy_cur_caches();
2262 kmem_cache_destroy(xfs_log_ticket_cache);
2263 kmem_cache_destroy(xfs_buf_cache);
2267 xfs_init_workqueues(void)
2270 * The allocation workqueue can be used in memory reclaim situations
2271 * (writepage path), and parallelism is only limited by the number of
2272 * AGs in all the filesystems mounted. Hence use the default large
2273 * max_active value for this workqueue.
2275 xfs_alloc_wq = alloc_workqueue("xfsalloc",
2276 XFS_WQFLAGS(WQ_MEM_RECLAIM | WQ_FREEZABLE), 0);
2280 xfs_discard_wq = alloc_workqueue("xfsdiscard", XFS_WQFLAGS(WQ_UNBOUND),
2282 if (!xfs_discard_wq)
2283 goto out_free_alloc_wq;
2287 destroy_workqueue(xfs_alloc_wq);
2292 xfs_destroy_workqueues(void)
2294 destroy_workqueue(xfs_discard_wq);
2295 destroy_workqueue(xfs_alloc_wq);
2298 #ifdef CONFIG_HOTPLUG_CPU
2303 struct xfs_mount *mp, *n;
2305 spin_lock(&xfs_mount_list_lock);
2306 list_for_each_entry_safe(mp, n, &xfs_mount_list, m_mount_list) {
2307 spin_unlock(&xfs_mount_list_lock);
2308 xfs_inodegc_cpu_dead(mp, cpu);
2309 xlog_cil_pcp_dead(mp->m_log, cpu);
2310 spin_lock(&xfs_mount_list_lock);
2312 spin_unlock(&xfs_mount_list_lock);
2317 xfs_cpu_hotplug_init(void)
2321 error = cpuhp_setup_state_nocalls(CPUHP_XFS_DEAD, "xfs:dead", NULL,
2325 "Failed to initialise CPU hotplug, error %d. XFS is non-functional.",
2331 xfs_cpu_hotplug_destroy(void)
2333 cpuhp_remove_state_nocalls(CPUHP_XFS_DEAD);
2336 #else /* !CONFIG_HOTPLUG_CPU */
2337 static inline int xfs_cpu_hotplug_init(void) { return 0; }
2338 static inline void xfs_cpu_hotplug_destroy(void) {}
2346 xfs_check_ondisk_structs();
2348 error = xfs_dahash_test();
2352 printk(KERN_INFO XFS_VERSION_STRING " with "
2353 XFS_BUILD_OPTIONS " enabled\n");
2357 error = xfs_cpu_hotplug_init();
2361 error = xfs_init_caches();
2363 goto out_destroy_hp;
2365 error = xfs_init_workqueues();
2367 goto out_destroy_caches;
2369 error = xfs_mru_cache_init();
2371 goto out_destroy_wq;
2373 error = xfs_init_procfs();
2375 goto out_mru_cache_uninit;
2377 error = xfs_sysctl_register();
2379 goto out_cleanup_procfs;
2381 xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj);
2384 goto out_sysctl_unregister;
2387 xfsstats.xs_kobj.kobject.kset = xfs_kset;
2389 xfsstats.xs_stats = alloc_percpu(struct xfsstats);
2390 if (!xfsstats.xs_stats) {
2392 goto out_kset_unregister;
2395 error = xfs_sysfs_init(&xfsstats.xs_kobj, &xfs_stats_ktype, NULL,
2398 goto out_free_stats;
2401 xfs_dbg_kobj.kobject.kset = xfs_kset;
2402 error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug");
2404 goto out_remove_stats_kobj;
2407 error = xfs_qm_init();
2409 goto out_remove_dbg_kobj;
2411 error = register_filesystem(&xfs_fs_type);
2418 out_remove_dbg_kobj:
2420 xfs_sysfs_del(&xfs_dbg_kobj);
2421 out_remove_stats_kobj:
2423 xfs_sysfs_del(&xfsstats.xs_kobj);
2425 free_percpu(xfsstats.xs_stats);
2426 out_kset_unregister:
2427 kset_unregister(xfs_kset);
2428 out_sysctl_unregister:
2429 xfs_sysctl_unregister();
2431 xfs_cleanup_procfs();
2432 out_mru_cache_uninit:
2433 xfs_mru_cache_uninit();
2435 xfs_destroy_workqueues();
2437 xfs_destroy_caches();
2439 xfs_cpu_hotplug_destroy();
2448 unregister_filesystem(&xfs_fs_type);
2450 xfs_sysfs_del(&xfs_dbg_kobj);
2452 xfs_sysfs_del(&xfsstats.xs_kobj);
2453 free_percpu(xfsstats.xs_stats);
2454 kset_unregister(xfs_kset);
2455 xfs_sysctl_unregister();
2456 xfs_cleanup_procfs();
2457 xfs_mru_cache_uninit();
2458 xfs_destroy_workqueues();
2459 xfs_destroy_caches();
2460 xfs_uuid_table_free();
2461 xfs_cpu_hotplug_destroy();
2464 module_init(init_xfs_fs);
2465 module_exit(exit_xfs_fs);
2467 MODULE_AUTHOR("Silicon Graphics, Inc.");
2468 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
2469 MODULE_LICENSE("GPL");