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"
45 #include "xfs_rtbitmap.h"
46 #include "scrub/stats.h"
48 #include <linux/magic.h>
49 #include <linux/fs_context.h>
50 #include <linux/fs_parser.h>
52 static const struct super_operations xfs_super_operations
;
54 static struct dentry
*xfs_debugfs
; /* top-level xfs debugfs dir */
55 static struct kset
*xfs_kset
; /* top-level xfs sysfs dir */
57 static struct xfs_kobj xfs_dbg_kobj
; /* global debug sysfs attrs */
67 xfs_mount_set_dax_mode(
69 enum xfs_dax_mode mode
)
73 mp
->m_features
&= ~(XFS_FEAT_DAX_ALWAYS
| XFS_FEAT_DAX_NEVER
);
76 mp
->m_features
|= XFS_FEAT_DAX_ALWAYS
;
77 mp
->m_features
&= ~XFS_FEAT_DAX_NEVER
;
80 mp
->m_features
|= XFS_FEAT_DAX_NEVER
;
81 mp
->m_features
&= ~XFS_FEAT_DAX_ALWAYS
;
86 static const struct constant_table dax_param_enums
[] = {
87 {"inode", XFS_DAX_INODE
},
88 {"always", XFS_DAX_ALWAYS
},
89 {"never", XFS_DAX_NEVER
},
94 * Table driven mount option parser.
97 Opt_logbufs
, Opt_logbsize
, Opt_logdev
, Opt_rtdev
,
98 Opt_wsync
, Opt_noalign
, Opt_swalloc
, Opt_sunit
, Opt_swidth
, Opt_nouuid
,
99 Opt_grpid
, Opt_nogrpid
, Opt_bsdgroups
, Opt_sysvgroups
,
100 Opt_allocsize
, Opt_norecovery
, Opt_inode64
, Opt_inode32
, Opt_ikeep
,
101 Opt_noikeep
, Opt_largeio
, Opt_nolargeio
, Opt_attr2
, Opt_noattr2
,
102 Opt_filestreams
, Opt_quota
, Opt_noquota
, Opt_usrquota
, Opt_grpquota
,
103 Opt_prjquota
, Opt_uquota
, Opt_gquota
, Opt_pquota
,
104 Opt_uqnoenforce
, Opt_gqnoenforce
, Opt_pqnoenforce
, Opt_qnoenforce
,
105 Opt_discard
, Opt_nodiscard
, Opt_dax
, Opt_dax_enum
,
108 static const struct fs_parameter_spec xfs_fs_parameters
[] = {
109 fsparam_u32("logbufs", Opt_logbufs
),
110 fsparam_string("logbsize", Opt_logbsize
),
111 fsparam_string("logdev", Opt_logdev
),
112 fsparam_string("rtdev", Opt_rtdev
),
113 fsparam_flag("wsync", Opt_wsync
),
114 fsparam_flag("noalign", Opt_noalign
),
115 fsparam_flag("swalloc", Opt_swalloc
),
116 fsparam_u32("sunit", Opt_sunit
),
117 fsparam_u32("swidth", Opt_swidth
),
118 fsparam_flag("nouuid", Opt_nouuid
),
119 fsparam_flag("grpid", Opt_grpid
),
120 fsparam_flag("nogrpid", Opt_nogrpid
),
121 fsparam_flag("bsdgroups", Opt_bsdgroups
),
122 fsparam_flag("sysvgroups", Opt_sysvgroups
),
123 fsparam_string("allocsize", Opt_allocsize
),
124 fsparam_flag("norecovery", Opt_norecovery
),
125 fsparam_flag("inode64", Opt_inode64
),
126 fsparam_flag("inode32", Opt_inode32
),
127 fsparam_flag("ikeep", Opt_ikeep
),
128 fsparam_flag("noikeep", Opt_noikeep
),
129 fsparam_flag("largeio", Opt_largeio
),
130 fsparam_flag("nolargeio", Opt_nolargeio
),
131 fsparam_flag("attr2", Opt_attr2
),
132 fsparam_flag("noattr2", Opt_noattr2
),
133 fsparam_flag("filestreams", Opt_filestreams
),
134 fsparam_flag("quota", Opt_quota
),
135 fsparam_flag("noquota", Opt_noquota
),
136 fsparam_flag("usrquota", Opt_usrquota
),
137 fsparam_flag("grpquota", Opt_grpquota
),
138 fsparam_flag("prjquota", Opt_prjquota
),
139 fsparam_flag("uquota", Opt_uquota
),
140 fsparam_flag("gquota", Opt_gquota
),
141 fsparam_flag("pquota", Opt_pquota
),
142 fsparam_flag("uqnoenforce", Opt_uqnoenforce
),
143 fsparam_flag("gqnoenforce", Opt_gqnoenforce
),
144 fsparam_flag("pqnoenforce", Opt_pqnoenforce
),
145 fsparam_flag("qnoenforce", Opt_qnoenforce
),
146 fsparam_flag("discard", Opt_discard
),
147 fsparam_flag("nodiscard", Opt_nodiscard
),
148 fsparam_flag("dax", Opt_dax
),
149 fsparam_enum("dax", Opt_dax_enum
, dax_param_enums
),
153 struct proc_xfs_info
{
163 static struct proc_xfs_info xfs_info_set
[] = {
164 /* the few simple ones we can get from the mount struct */
165 { XFS_FEAT_IKEEP
, ",ikeep" },
166 { XFS_FEAT_WSYNC
, ",wsync" },
167 { XFS_FEAT_NOALIGN
, ",noalign" },
168 { XFS_FEAT_SWALLOC
, ",swalloc" },
169 { XFS_FEAT_NOUUID
, ",nouuid" },
170 { XFS_FEAT_NORECOVERY
, ",norecovery" },
171 { XFS_FEAT_ATTR2
, ",attr2" },
172 { XFS_FEAT_FILESTREAMS
, ",filestreams" },
173 { XFS_FEAT_GRPID
, ",grpid" },
174 { XFS_FEAT_DISCARD
, ",discard" },
175 { XFS_FEAT_LARGE_IOSIZE
, ",largeio" },
176 { XFS_FEAT_DAX_ALWAYS
, ",dax=always" },
177 { XFS_FEAT_DAX_NEVER
, ",dax=never" },
180 struct xfs_mount
*mp
= XFS_M(root
->d_sb
);
181 struct proc_xfs_info
*xfs_infop
;
183 for (xfs_infop
= xfs_info_set
; xfs_infop
->flag
; xfs_infop
++) {
184 if (mp
->m_features
& xfs_infop
->flag
)
185 seq_puts(m
, xfs_infop
->str
);
188 seq_printf(m
, ",inode%d", xfs_has_small_inums(mp
) ? 32 : 64);
190 if (xfs_has_allocsize(mp
))
191 seq_printf(m
, ",allocsize=%dk",
192 (1 << mp
->m_allocsize_log
) >> 10);
194 if (mp
->m_logbufs
> 0)
195 seq_printf(m
, ",logbufs=%d", mp
->m_logbufs
);
196 if (mp
->m_logbsize
> 0)
197 seq_printf(m
, ",logbsize=%dk", mp
->m_logbsize
>> 10);
200 seq_show_option(m
, "logdev", mp
->m_logname
);
202 seq_show_option(m
, "rtdev", mp
->m_rtname
);
204 if (mp
->m_dalign
> 0)
205 seq_printf(m
, ",sunit=%d",
206 (int)XFS_FSB_TO_BB(mp
, mp
->m_dalign
));
207 if (mp
->m_swidth
> 0)
208 seq_printf(m
, ",swidth=%d",
209 (int)XFS_FSB_TO_BB(mp
, mp
->m_swidth
));
211 if (mp
->m_qflags
& XFS_UQUOTA_ENFD
)
212 seq_puts(m
, ",usrquota");
213 else if (mp
->m_qflags
& XFS_UQUOTA_ACCT
)
214 seq_puts(m
, ",uqnoenforce");
216 if (mp
->m_qflags
& XFS_PQUOTA_ENFD
)
217 seq_puts(m
, ",prjquota");
218 else if (mp
->m_qflags
& XFS_PQUOTA_ACCT
)
219 seq_puts(m
, ",pqnoenforce");
221 if (mp
->m_qflags
& XFS_GQUOTA_ENFD
)
222 seq_puts(m
, ",grpquota");
223 else if (mp
->m_qflags
& XFS_GQUOTA_ACCT
)
224 seq_puts(m
, ",gqnoenforce");
226 if (!(mp
->m_qflags
& XFS_ALL_QUOTA_ACCT
))
227 seq_puts(m
, ",noquota");
233 xfs_set_inode_alloc_perag(
234 struct xfs_perag
*pag
,
236 xfs_agnumber_t max_metadata
)
238 if (!xfs_is_inode32(pag
->pag_mount
)) {
239 set_bit(XFS_AGSTATE_ALLOWS_INODES
, &pag
->pag_opstate
);
240 clear_bit(XFS_AGSTATE_PREFERS_METADATA
, &pag
->pag_opstate
);
244 if (ino
> XFS_MAXINUMBER_32
) {
245 clear_bit(XFS_AGSTATE_ALLOWS_INODES
, &pag
->pag_opstate
);
246 clear_bit(XFS_AGSTATE_PREFERS_METADATA
, &pag
->pag_opstate
);
250 set_bit(XFS_AGSTATE_ALLOWS_INODES
, &pag
->pag_opstate
);
251 if (pag
->pag_agno
< max_metadata
)
252 set_bit(XFS_AGSTATE_PREFERS_METADATA
, &pag
->pag_opstate
);
254 clear_bit(XFS_AGSTATE_PREFERS_METADATA
, &pag
->pag_opstate
);
259 * Set parameters for inode allocation heuristics, taking into account
260 * filesystem size and inode32/inode64 mount options; i.e. specifically
261 * whether or not XFS_FEAT_SMALL_INUMS is set.
263 * Inode allocation patterns are altered only if inode32 is requested
264 * (XFS_FEAT_SMALL_INUMS), and the filesystem is sufficiently large.
265 * If altered, XFS_OPSTATE_INODE32 is set as well.
267 * An agcount independent of that in the mount structure is provided
268 * because in the growfs case, mp->m_sb.sb_agcount is not yet updated
269 * to the potentially higher ag count.
271 * Returns the maximum AG index which may contain inodes.
275 struct xfs_mount
*mp
,
276 xfs_agnumber_t agcount
)
278 xfs_agnumber_t index
;
279 xfs_agnumber_t maxagi
= 0;
280 xfs_sb_t
*sbp
= &mp
->m_sb
;
281 xfs_agnumber_t max_metadata
;
286 * Calculate how much should be reserved for inodes to meet
287 * the max inode percentage. Used only for inode32.
289 if (M_IGEO(mp
)->maxicount
) {
292 icount
= sbp
->sb_dblocks
* sbp
->sb_imax_pct
;
294 icount
+= sbp
->sb_agblocks
- 1;
295 do_div(icount
, sbp
->sb_agblocks
);
296 max_metadata
= icount
;
298 max_metadata
= agcount
;
301 /* Get the last possible inode in the filesystem */
302 agino
= XFS_AGB_TO_AGINO(mp
, sbp
->sb_agblocks
- 1);
303 ino
= XFS_AGINO_TO_INO(mp
, agcount
- 1, agino
);
306 * If user asked for no more than 32-bit inodes, and the fs is
307 * sufficiently large, set XFS_OPSTATE_INODE32 if we must alter
308 * the allocator to accommodate the request.
310 if (xfs_has_small_inums(mp
) && ino
> XFS_MAXINUMBER_32
)
311 set_bit(XFS_OPSTATE_INODE32
, &mp
->m_opstate
);
313 clear_bit(XFS_OPSTATE_INODE32
, &mp
->m_opstate
);
315 for (index
= 0; index
< agcount
; index
++) {
316 struct xfs_perag
*pag
;
318 ino
= XFS_AGINO_TO_INO(mp
, index
, agino
);
320 pag
= xfs_perag_get(mp
, index
);
321 if (xfs_set_inode_alloc_perag(pag
, ino
, max_metadata
))
326 return xfs_is_inode32(mp
) ? maxagi
: agcount
;
330 xfs_setup_dax_always(
331 struct xfs_mount
*mp
)
333 if (!mp
->m_ddev_targp
->bt_daxdev
&&
334 (!mp
->m_rtdev_targp
|| !mp
->m_rtdev_targp
->bt_daxdev
)) {
336 "DAX unsupported by block device. Turning off DAX.");
340 if (mp
->m_super
->s_blocksize
!= PAGE_SIZE
) {
342 "DAX not supported for blocksize. Turning off DAX.");
346 if (xfs_has_reflink(mp
) &&
347 bdev_is_partition(mp
->m_ddev_targp
->bt_bdev
)) {
349 "DAX and reflink cannot work with multi-partitions!");
356 xfs_mount_set_dax_mode(mp
, XFS_DAX_NEVER
);
364 struct file
**bdev_filep
)
368 *bdev_filep
= bdev_file_open_by_path(name
,
369 BLK_OPEN_READ
| BLK_OPEN_WRITE
| BLK_OPEN_RESTRICT_WRITES
,
370 mp
->m_super
, &fs_holder_ops
);
371 if (IS_ERR(*bdev_filep
)) {
372 error
= PTR_ERR(*bdev_filep
);
374 xfs_warn(mp
, "Invalid device [%s], error=%d", name
, error
);
381 xfs_shutdown_devices(
382 struct xfs_mount
*mp
)
385 * Udev is triggered whenever anyone closes a block device or unmounts
386 * a file systemm on a block device.
387 * The default udev rules invoke blkid to read the fs super and create
388 * symlinks to the bdev under /dev/disk. For this, it uses buffered
389 * reads through the page cache.
391 * xfs_db also uses buffered reads to examine metadata. There is no
392 * coordination between xfs_db and udev, which means that they can run
393 * concurrently. Note there is no coordination between the kernel and
396 * On a system with 64k pages, the page cache can cache the superblock
397 * and the root inode (and hence the root directory) with the same 64k
398 * page. If udev spawns blkid after the mkfs and the system is busy
399 * enough that it is still running when xfs_db starts up, they'll both
400 * read from the same page in the pagecache.
402 * The unmount writes updated inode metadata to disk directly. The XFS
403 * buffer cache does not use the bdev pagecache, so it needs to
404 * invalidate that pagecache on unmount. If the above scenario occurs,
405 * the pagecache no longer reflects what's on disk, xfs_db reads the
406 * stale metadata, and fails to find /a. Most of the time this succeeds
407 * because closing a bdev invalidates the page cache, but when processes
408 * race, everyone loses.
410 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
411 blkdev_issue_flush(mp
->m_logdev_targp
->bt_bdev
);
412 invalidate_bdev(mp
->m_logdev_targp
->bt_bdev
);
414 if (mp
->m_rtdev_targp
) {
415 blkdev_issue_flush(mp
->m_rtdev_targp
->bt_bdev
);
416 invalidate_bdev(mp
->m_rtdev_targp
->bt_bdev
);
418 blkdev_issue_flush(mp
->m_ddev_targp
->bt_bdev
);
419 invalidate_bdev(mp
->m_ddev_targp
->bt_bdev
);
423 * The file system configurations are:
424 * (1) device (partition) with data and internal log
425 * (2) logical volume with data and log subvolumes.
426 * (3) logical volume with data, log, and realtime subvolumes.
428 * We only have to handle opening the log and realtime volumes here if
429 * they are present. The data subvolume has already been opened by
430 * get_sb_bdev() and is stored in sb->s_bdev.
434 struct xfs_mount
*mp
)
436 struct super_block
*sb
= mp
->m_super
;
437 struct block_device
*ddev
= sb
->s_bdev
;
438 struct file
*logdev_file
= NULL
, *rtdev_file
= NULL
;
442 * Open real time and log devices - order is important.
445 error
= xfs_blkdev_get(mp
, mp
->m_logname
, &logdev_file
);
451 error
= xfs_blkdev_get(mp
, mp
->m_rtname
, &rtdev_file
);
453 goto out_close_logdev
;
455 if (file_bdev(rtdev_file
) == ddev
||
457 file_bdev(rtdev_file
) == file_bdev(logdev_file
))) {
459 "Cannot mount filesystem with identical rtdev and ddev/logdev.");
461 goto out_close_rtdev
;
466 * Setup xfs_mount buffer target pointers
469 mp
->m_ddev_targp
= xfs_alloc_buftarg(mp
, sb
->s_bdev_file
);
470 if (!mp
->m_ddev_targp
)
471 goto out_close_rtdev
;
474 mp
->m_rtdev_targp
= xfs_alloc_buftarg(mp
, rtdev_file
);
475 if (!mp
->m_rtdev_targp
)
476 goto out_free_ddev_targ
;
479 if (logdev_file
&& file_bdev(logdev_file
) != ddev
) {
480 mp
->m_logdev_targp
= xfs_alloc_buftarg(mp
, logdev_file
);
481 if (!mp
->m_logdev_targp
)
482 goto out_free_rtdev_targ
;
484 mp
->m_logdev_targp
= mp
->m_ddev_targp
;
485 /* Handle won't be used, drop it */
493 if (mp
->m_rtdev_targp
)
494 xfs_free_buftarg(mp
->m_rtdev_targp
);
496 xfs_free_buftarg(mp
->m_ddev_targp
);
507 * Setup xfs_mount buffer target pointers based on superblock
511 struct xfs_mount
*mp
)
515 error
= xfs_setsize_buftarg(mp
->m_ddev_targp
, mp
->m_sb
.sb_sectsize
);
519 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
) {
520 unsigned int log_sector_size
= BBSIZE
;
522 if (xfs_has_sector(mp
))
523 log_sector_size
= mp
->m_sb
.sb_logsectsize
;
524 error
= xfs_setsize_buftarg(mp
->m_logdev_targp
,
529 if (mp
->m_rtdev_targp
) {
530 error
= xfs_setsize_buftarg(mp
->m_rtdev_targp
,
531 mp
->m_sb
.sb_sectsize
);
540 xfs_init_mount_workqueues(
541 struct xfs_mount
*mp
)
543 mp
->m_buf_workqueue
= alloc_workqueue("xfs-buf/%s",
544 XFS_WQFLAGS(WQ_FREEZABLE
| WQ_MEM_RECLAIM
),
545 1, mp
->m_super
->s_id
);
546 if (!mp
->m_buf_workqueue
)
549 mp
->m_unwritten_workqueue
= alloc_workqueue("xfs-conv/%s",
550 XFS_WQFLAGS(WQ_FREEZABLE
| WQ_MEM_RECLAIM
),
551 0, mp
->m_super
->s_id
);
552 if (!mp
->m_unwritten_workqueue
)
553 goto out_destroy_buf
;
555 mp
->m_reclaim_workqueue
= alloc_workqueue("xfs-reclaim/%s",
556 XFS_WQFLAGS(WQ_FREEZABLE
| WQ_MEM_RECLAIM
),
557 0, mp
->m_super
->s_id
);
558 if (!mp
->m_reclaim_workqueue
)
559 goto out_destroy_unwritten
;
561 mp
->m_blockgc_wq
= alloc_workqueue("xfs-blockgc/%s",
562 XFS_WQFLAGS(WQ_UNBOUND
| WQ_FREEZABLE
| WQ_MEM_RECLAIM
),
563 0, mp
->m_super
->s_id
);
564 if (!mp
->m_blockgc_wq
)
565 goto out_destroy_reclaim
;
567 mp
->m_inodegc_wq
= alloc_workqueue("xfs-inodegc/%s",
568 XFS_WQFLAGS(WQ_FREEZABLE
| WQ_MEM_RECLAIM
),
569 1, mp
->m_super
->s_id
);
570 if (!mp
->m_inodegc_wq
)
571 goto out_destroy_blockgc
;
573 mp
->m_sync_workqueue
= alloc_workqueue("xfs-sync/%s",
574 XFS_WQFLAGS(WQ_FREEZABLE
), 0, mp
->m_super
->s_id
);
575 if (!mp
->m_sync_workqueue
)
576 goto out_destroy_inodegc
;
581 destroy_workqueue(mp
->m_inodegc_wq
);
583 destroy_workqueue(mp
->m_blockgc_wq
);
585 destroy_workqueue(mp
->m_reclaim_workqueue
);
586 out_destroy_unwritten
:
587 destroy_workqueue(mp
->m_unwritten_workqueue
);
589 destroy_workqueue(mp
->m_buf_workqueue
);
595 xfs_destroy_mount_workqueues(
596 struct xfs_mount
*mp
)
598 destroy_workqueue(mp
->m_sync_workqueue
);
599 destroy_workqueue(mp
->m_blockgc_wq
);
600 destroy_workqueue(mp
->m_inodegc_wq
);
601 destroy_workqueue(mp
->m_reclaim_workqueue
);
602 destroy_workqueue(mp
->m_unwritten_workqueue
);
603 destroy_workqueue(mp
->m_buf_workqueue
);
607 xfs_flush_inodes_worker(
608 struct work_struct
*work
)
610 struct xfs_mount
*mp
= container_of(work
, struct xfs_mount
,
611 m_flush_inodes_work
);
612 struct super_block
*sb
= mp
->m_super
;
614 if (down_read_trylock(&sb
->s_umount
)) {
616 up_read(&sb
->s_umount
);
621 * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
622 * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
623 * for IO to complete so that we effectively throttle multiple callers to the
624 * rate at which IO is completing.
628 struct xfs_mount
*mp
)
631 * If flush_work() returns true then that means we waited for a flush
632 * which was already in progress. Don't bother running another scan.
634 if (flush_work(&mp
->m_flush_inodes_work
))
637 queue_work(mp
->m_sync_workqueue
, &mp
->m_flush_inodes_work
);
638 flush_work(&mp
->m_flush_inodes_work
);
641 /* Catch misguided souls that try to use this interface on XFS */
642 STATIC
struct inode
*
644 struct super_block
*sb
)
651 * Now that the generic code is guaranteed not to be accessing
652 * the linux inode, we can inactivate and reclaim the inode.
655 xfs_fs_destroy_inode(
658 struct xfs_inode
*ip
= XFS_I(inode
);
660 trace_xfs_destroy_inode(ip
);
662 ASSERT(!rwsem_is_locked(&inode
->i_rwsem
));
663 XFS_STATS_INC(ip
->i_mount
, vn_rele
);
664 XFS_STATS_INC(ip
->i_mount
, vn_remove
);
665 xfs_inode_mark_reclaimable(ip
);
673 struct xfs_inode
*ip
= XFS_I(inode
);
674 struct xfs_mount
*mp
= ip
->i_mount
;
675 struct xfs_trans
*tp
;
677 if (!(inode
->i_sb
->s_flags
& SB_LAZYTIME
))
681 * Only do the timestamp update if the inode is dirty (I_DIRTY_SYNC)
682 * and has dirty timestamp (I_DIRTY_TIME). I_DIRTY_TIME can be passed
683 * in flags possibly together with I_DIRTY_SYNC.
685 if ((flags
& ~I_DIRTY_TIME
) != I_DIRTY_SYNC
|| !(flags
& I_DIRTY_TIME
))
688 if (xfs_trans_alloc(mp
, &M_RES(mp
)->tr_fsyncts
, 0, 0, 0, &tp
))
690 xfs_ilock(ip
, XFS_ILOCK_EXCL
);
691 xfs_trans_ijoin(tp
, ip
, XFS_ILOCK_EXCL
);
692 xfs_trans_log_inode(tp
, ip
, XFS_ILOG_TIMESTAMP
);
693 xfs_trans_commit(tp
);
697 * Slab object creation initialisation for the XFS inode.
698 * This covers only the idempotent fields in the XFS inode;
699 * all other fields need to be initialised on allocation
700 * from the slab. This avoids the need to repeatedly initialise
701 * fields in the xfs inode that left in the initialise state
702 * when freeing the inode.
705 xfs_fs_inode_init_once(
708 struct xfs_inode
*ip
= inode
;
710 memset(ip
, 0, sizeof(struct xfs_inode
));
713 inode_init_once(VFS_I(ip
));
716 atomic_set(&ip
->i_pincount
, 0);
717 spin_lock_init(&ip
->i_flags_lock
);
719 mrlock_init(&ip
->i_lock
, MRLOCK_ALLOW_EQUAL_PRI
|MRLOCK_BARRIER
,
720 "xfsino", ip
->i_ino
);
724 * We do an unlocked check for XFS_IDONTCACHE here because we are already
725 * serialised against cache hits here via the inode->i_lock and igrab() in
726 * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
727 * racing with us, and it avoids needing to grab a spinlock here for every inode
728 * we drop the final reference on.
734 struct xfs_inode
*ip
= XFS_I(inode
);
737 * If this unlinked inode is in the middle of recovery, don't
738 * drop the inode just yet; log recovery will take care of
739 * that. See the comment for this inode flag.
741 if (ip
->i_flags
& XFS_IRECOVERY
) {
742 ASSERT(xlog_recovery_needed(ip
->i_mount
->m_log
));
746 return generic_drop_inode(inode
);
751 struct xfs_mount
*mp
)
753 if (mp
->m_logdev_targp
&& mp
->m_logdev_targp
!= mp
->m_ddev_targp
)
754 xfs_free_buftarg(mp
->m_logdev_targp
);
755 if (mp
->m_rtdev_targp
)
756 xfs_free_buftarg(mp
->m_rtdev_targp
);
757 if (mp
->m_ddev_targp
)
758 xfs_free_buftarg(mp
->m_ddev_targp
);
760 debugfs_remove(mp
->m_debugfs
);
762 kfree(mp
->m_logname
);
768 struct super_block
*sb
,
771 struct xfs_mount
*mp
= XFS_M(sb
);
774 trace_xfs_fs_sync_fs(mp
, __return_address
);
777 * Doing anything during the async pass would be counterproductive.
782 error
= xfs_log_force(mp
, XFS_LOG_SYNC
);
788 * The disk must be active because we're syncing.
789 * We schedule log work now (now that the disk is
790 * active) instead of later (when it might not be).
792 flush_delayed_work(&mp
->m_log
->l_work
);
796 * If we are called with page faults frozen out, it means we are about
797 * to freeze the transaction subsystem. Take the opportunity to shut
798 * down inodegc because once SB_FREEZE_FS is set it's too late to
799 * prevent inactivation races with freeze. The fs doesn't get called
800 * again by the freezing process until after SB_FREEZE_FS has been set,
801 * so it's now or never. Same logic applies to speculative allocation
802 * garbage collection.
804 * We don't care if this is a normal syncfs call that does this or
805 * freeze that does this - we can run this multiple times without issue
806 * and we won't race with a restart because a restart can only occur
807 * when the state is either SB_FREEZE_FS or SB_FREEZE_COMPLETE.
809 if (sb
->s_writers
.frozen
== SB_FREEZE_PAGEFAULT
) {
810 xfs_inodegc_stop(mp
);
811 xfs_blockgc_stop(mp
);
819 struct dentry
*dentry
,
820 struct kstatfs
*statp
)
822 struct xfs_mount
*mp
= XFS_M(dentry
->d_sb
);
823 xfs_sb_t
*sbp
= &mp
->m_sb
;
824 struct xfs_inode
*ip
= XFS_I(d_inode(dentry
));
825 uint64_t fakeinos
, id
;
833 * Expedite background inodegc but don't wait. We do not want to block
834 * here waiting hours for a billion extent file to be truncated.
836 xfs_inodegc_push(mp
);
838 statp
->f_type
= XFS_SUPER_MAGIC
;
839 statp
->f_namelen
= MAXNAMELEN
- 1;
841 id
= huge_encode_dev(mp
->m_ddev_targp
->bt_dev
);
842 statp
->f_fsid
= u64_to_fsid(id
);
844 icount
= percpu_counter_sum(&mp
->m_icount
);
845 ifree
= percpu_counter_sum(&mp
->m_ifree
);
846 fdblocks
= percpu_counter_sum(&mp
->m_fdblocks
);
848 spin_lock(&mp
->m_sb_lock
);
849 statp
->f_bsize
= sbp
->sb_blocksize
;
850 lsize
= sbp
->sb_logstart
? sbp
->sb_logblocks
: 0;
851 statp
->f_blocks
= sbp
->sb_dblocks
- lsize
;
852 spin_unlock(&mp
->m_sb_lock
);
854 /* make sure statp->f_bfree does not underflow */
855 statp
->f_bfree
= max_t(int64_t, 0,
856 fdblocks
- xfs_fdblocks_unavailable(mp
));
857 statp
->f_bavail
= statp
->f_bfree
;
859 fakeinos
= XFS_FSB_TO_INO(mp
, statp
->f_bfree
);
860 statp
->f_files
= min(icount
+ fakeinos
, (uint64_t)XFS_MAXINUMBER
);
861 if (M_IGEO(mp
)->maxicount
)
862 statp
->f_files
= min_t(typeof(statp
->f_files
),
864 M_IGEO(mp
)->maxicount
);
866 /* If sb_icount overshot maxicount, report actual allocation */
867 statp
->f_files
= max_t(typeof(statp
->f_files
),
871 /* make sure statp->f_ffree does not underflow */
872 ffree
= statp
->f_files
- (icount
- ifree
);
873 statp
->f_ffree
= max_t(int64_t, ffree
, 0);
876 if ((ip
->i_diflags
& XFS_DIFLAG_PROJINHERIT
) &&
877 ((mp
->m_qflags
& (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))) ==
878 (XFS_PQUOTA_ACCT
|XFS_PQUOTA_ENFD
))
879 xfs_qm_statvfs(ip
, statp
);
881 if (XFS_IS_REALTIME_MOUNT(mp
) &&
882 (ip
->i_diflags
& (XFS_DIFLAG_RTINHERIT
| XFS_DIFLAG_REALTIME
))) {
885 statp
->f_blocks
= sbp
->sb_rblocks
;
886 freertx
= percpu_counter_sum_positive(&mp
->m_frextents
);
887 statp
->f_bavail
= statp
->f_bfree
= xfs_rtx_to_rtb(mp
, freertx
);
894 xfs_save_resvblks(struct xfs_mount
*mp
)
896 mp
->m_resblks_save
= mp
->m_resblks
;
897 xfs_reserve_blocks(mp
, 0);
901 xfs_restore_resvblks(struct xfs_mount
*mp
)
905 if (mp
->m_resblks_save
) {
906 resblks
= mp
->m_resblks_save
;
907 mp
->m_resblks_save
= 0;
909 resblks
= xfs_default_resblks(mp
);
911 xfs_reserve_blocks(mp
, resblks
);
915 * Second stage of a freeze. The data is already frozen so we only
916 * need to take care of the metadata. Once that's done sync the superblock
917 * to the log to dirty it in case of a crash while frozen. This ensures that we
918 * will recover the unlinked inode lists on the next mount.
922 struct super_block
*sb
)
924 struct xfs_mount
*mp
= XFS_M(sb
);
929 * The filesystem is now frozen far enough that memory reclaim
930 * cannot safely operate on the filesystem. Hence we need to
931 * set a GFP_NOFS context here to avoid recursion deadlocks.
933 flags
= memalloc_nofs_save();
934 xfs_save_resvblks(mp
);
935 ret
= xfs_log_quiesce(mp
);
936 memalloc_nofs_restore(flags
);
939 * For read-write filesystems, we need to restart the inodegc on error
940 * because we stopped it at SB_FREEZE_PAGEFAULT level and a thaw is not
941 * going to be run to restart it now. We are at SB_FREEZE_FS level
942 * here, so we can restart safely without racing with a stop in
945 if (ret
&& !xfs_is_readonly(mp
)) {
946 xfs_blockgc_start(mp
);
947 xfs_inodegc_start(mp
);
955 struct super_block
*sb
)
957 struct xfs_mount
*mp
= XFS_M(sb
);
959 xfs_restore_resvblks(mp
);
960 xfs_log_work_queue(mp
);
963 * Don't reactivate the inodegc worker on a readonly filesystem because
964 * inodes are sent directly to reclaim. Don't reactivate the blockgc
965 * worker because there are no speculative preallocations on a readonly
968 if (!xfs_is_readonly(mp
)) {
969 xfs_blockgc_start(mp
);
970 xfs_inodegc_start(mp
);
977 * This function fills in xfs_mount_t fields based on mount args.
978 * Note: the superblock _has_ now been read in.
982 struct xfs_mount
*mp
)
984 /* Fail a mount where the logbuf is smaller than the log stripe */
985 if (xfs_has_logv2(mp
)) {
986 if (mp
->m_logbsize
<= 0 &&
987 mp
->m_sb
.sb_logsunit
> XLOG_BIG_RECORD_BSIZE
) {
988 mp
->m_logbsize
= mp
->m_sb
.sb_logsunit
;
989 } else if (mp
->m_logbsize
> 0 &&
990 mp
->m_logbsize
< mp
->m_sb
.sb_logsunit
) {
992 "logbuf size must be greater than or equal to log stripe size");
996 /* Fail a mount if the logbuf is larger than 32K */
997 if (mp
->m_logbsize
> XLOG_BIG_RECORD_BSIZE
) {
999 "logbuf size for version 1 logs must be 16K or 32K");
1005 * V5 filesystems always use attr2 format for attributes.
1007 if (xfs_has_crc(mp
) && xfs_has_noattr2(mp
)) {
1008 xfs_warn(mp
, "Cannot mount a V5 filesystem as noattr2. "
1009 "attr2 is always enabled for V5 filesystems.");
1014 * prohibit r/w mounts of read-only filesystems
1016 if ((mp
->m_sb
.sb_flags
& XFS_SBF_READONLY
) && !xfs_is_readonly(mp
)) {
1018 "cannot mount a read-only filesystem as read-write");
1022 if ((mp
->m_qflags
& XFS_GQUOTA_ACCT
) &&
1023 (mp
->m_qflags
& XFS_PQUOTA_ACCT
) &&
1024 !xfs_has_pquotino(mp
)) {
1026 "Super block does not support project and group quota together");
1034 xfs_init_percpu_counters(
1035 struct xfs_mount
*mp
)
1039 error
= percpu_counter_init(&mp
->m_icount
, 0, GFP_KERNEL
);
1043 error
= percpu_counter_init(&mp
->m_ifree
, 0, GFP_KERNEL
);
1047 error
= percpu_counter_init(&mp
->m_fdblocks
, 0, GFP_KERNEL
);
1051 error
= percpu_counter_init(&mp
->m_delalloc_blks
, 0, GFP_KERNEL
);
1055 error
= percpu_counter_init(&mp
->m_frextents
, 0, GFP_KERNEL
);
1062 percpu_counter_destroy(&mp
->m_delalloc_blks
);
1064 percpu_counter_destroy(&mp
->m_fdblocks
);
1066 percpu_counter_destroy(&mp
->m_ifree
);
1068 percpu_counter_destroy(&mp
->m_icount
);
1073 xfs_reinit_percpu_counters(
1074 struct xfs_mount
*mp
)
1076 percpu_counter_set(&mp
->m_icount
, mp
->m_sb
.sb_icount
);
1077 percpu_counter_set(&mp
->m_ifree
, mp
->m_sb
.sb_ifree
);
1078 percpu_counter_set(&mp
->m_fdblocks
, mp
->m_sb
.sb_fdblocks
);
1079 percpu_counter_set(&mp
->m_frextents
, mp
->m_sb
.sb_frextents
);
1083 xfs_destroy_percpu_counters(
1084 struct xfs_mount
*mp
)
1086 percpu_counter_destroy(&mp
->m_icount
);
1087 percpu_counter_destroy(&mp
->m_ifree
);
1088 percpu_counter_destroy(&mp
->m_fdblocks
);
1089 ASSERT(xfs_is_shutdown(mp
) ||
1090 percpu_counter_sum(&mp
->m_delalloc_blks
) == 0);
1091 percpu_counter_destroy(&mp
->m_delalloc_blks
);
1092 percpu_counter_destroy(&mp
->m_frextents
);
1096 xfs_inodegc_init_percpu(
1097 struct xfs_mount
*mp
)
1099 struct xfs_inodegc
*gc
;
1102 mp
->m_inodegc
= alloc_percpu(struct xfs_inodegc
);
1106 for_each_possible_cpu(cpu
) {
1107 gc
= per_cpu_ptr(mp
->m_inodegc
, cpu
);
1110 init_llist_head(&gc
->list
);
1113 INIT_DELAYED_WORK(&gc
->work
, xfs_inodegc_worker
);
1119 xfs_inodegc_free_percpu(
1120 struct xfs_mount
*mp
)
1124 free_percpu(mp
->m_inodegc
);
1129 struct super_block
*sb
)
1131 struct xfs_mount
*mp
= XFS_M(sb
);
1133 xfs_notice(mp
, "Unmounting Filesystem %pU", &mp
->m_sb
.sb_uuid
);
1134 xfs_filestream_unmount(mp
);
1138 xchk_mount_stats_free(mp
);
1139 free_percpu(mp
->m_stats
.xs_stats
);
1140 xfs_inodegc_free_percpu(mp
);
1141 xfs_destroy_percpu_counters(mp
);
1142 xfs_destroy_mount_workqueues(mp
);
1143 xfs_shutdown_devices(mp
);
1147 xfs_fs_nr_cached_objects(
1148 struct super_block
*sb
,
1149 struct shrink_control
*sc
)
1151 /* Paranoia: catch incorrect calls during mount setup or teardown */
1152 if (WARN_ON_ONCE(!sb
->s_fs_info
))
1154 return xfs_reclaim_inodes_count(XFS_M(sb
));
1158 xfs_fs_free_cached_objects(
1159 struct super_block
*sb
,
1160 struct shrink_control
*sc
)
1162 return xfs_reclaim_inodes_nr(XFS_M(sb
), sc
->nr_to_scan
);
1167 struct super_block
*sb
)
1169 xfs_force_shutdown(XFS_M(sb
), SHUTDOWN_DEVICE_REMOVED
);
1172 static const struct super_operations xfs_super_operations
= {
1173 .alloc_inode
= xfs_fs_alloc_inode
,
1174 .destroy_inode
= xfs_fs_destroy_inode
,
1175 .dirty_inode
= xfs_fs_dirty_inode
,
1176 .drop_inode
= xfs_fs_drop_inode
,
1177 .put_super
= xfs_fs_put_super
,
1178 .sync_fs
= xfs_fs_sync_fs
,
1179 .freeze_fs
= xfs_fs_freeze
,
1180 .unfreeze_fs
= xfs_fs_unfreeze
,
1181 .statfs
= xfs_fs_statfs
,
1182 .show_options
= xfs_fs_show_options
,
1183 .nr_cached_objects
= xfs_fs_nr_cached_objects
,
1184 .free_cached_objects
= xfs_fs_free_cached_objects
,
1185 .shutdown
= xfs_fs_shutdown
,
1194 int last
, shift_left_factor
= 0, _res
;
1198 value
= kstrdup(s
, GFP_KERNEL
);
1202 last
= strlen(value
) - 1;
1203 if (value
[last
] == 'K' || value
[last
] == 'k') {
1204 shift_left_factor
= 10;
1207 if (value
[last
] == 'M' || value
[last
] == 'm') {
1208 shift_left_factor
= 20;
1211 if (value
[last
] == 'G' || value
[last
] == 'g') {
1212 shift_left_factor
= 30;
1216 if (kstrtoint(value
, base
, &_res
))
1219 *res
= _res
<< shift_left_factor
;
1224 xfs_fs_warn_deprecated(
1225 struct fs_context
*fc
,
1226 struct fs_parameter
*param
,
1230 /* Don't print the warning if reconfiguring and current mount point
1231 * already had the flag set
1233 if ((fc
->purpose
& FS_CONTEXT_FOR_RECONFIGURE
) &&
1234 !!(XFS_M(fc
->root
->d_sb
)->m_features
& flag
) == value
)
1236 xfs_warn(fc
->s_fs_info
, "%s mount option is deprecated.", param
->key
);
1240 * Set mount state from a mount option.
1242 * NOTE: mp->m_super is NULL here!
1246 struct fs_context
*fc
,
1247 struct fs_parameter
*param
)
1249 struct xfs_mount
*parsing_mp
= fc
->s_fs_info
;
1250 struct fs_parse_result result
;
1254 opt
= fs_parse(fc
, xfs_fs_parameters
, param
, &result
);
1260 parsing_mp
->m_logbufs
= result
.uint_32
;
1263 if (suffix_kstrtoint(param
->string
, 10, &parsing_mp
->m_logbsize
))
1267 kfree(parsing_mp
->m_logname
);
1268 parsing_mp
->m_logname
= kstrdup(param
->string
, GFP_KERNEL
);
1269 if (!parsing_mp
->m_logname
)
1273 kfree(parsing_mp
->m_rtname
);
1274 parsing_mp
->m_rtname
= kstrdup(param
->string
, GFP_KERNEL
);
1275 if (!parsing_mp
->m_rtname
)
1279 if (suffix_kstrtoint(param
->string
, 10, &size
))
1281 parsing_mp
->m_allocsize_log
= ffs(size
) - 1;
1282 parsing_mp
->m_features
|= XFS_FEAT_ALLOCSIZE
;
1286 parsing_mp
->m_features
|= XFS_FEAT_GRPID
;
1289 case Opt_sysvgroups
:
1290 parsing_mp
->m_features
&= ~XFS_FEAT_GRPID
;
1293 parsing_mp
->m_features
|= XFS_FEAT_WSYNC
;
1295 case Opt_norecovery
:
1296 parsing_mp
->m_features
|= XFS_FEAT_NORECOVERY
;
1299 parsing_mp
->m_features
|= XFS_FEAT_NOALIGN
;
1302 parsing_mp
->m_features
|= XFS_FEAT_SWALLOC
;
1305 parsing_mp
->m_dalign
= result
.uint_32
;
1308 parsing_mp
->m_swidth
= result
.uint_32
;
1311 parsing_mp
->m_features
|= XFS_FEAT_SMALL_INUMS
;
1314 parsing_mp
->m_features
&= ~XFS_FEAT_SMALL_INUMS
;
1317 parsing_mp
->m_features
|= XFS_FEAT_NOUUID
;
1320 parsing_mp
->m_features
|= XFS_FEAT_LARGE_IOSIZE
;
1323 parsing_mp
->m_features
&= ~XFS_FEAT_LARGE_IOSIZE
;
1325 case Opt_filestreams
:
1326 parsing_mp
->m_features
|= XFS_FEAT_FILESTREAMS
;
1329 parsing_mp
->m_qflags
&= ~XFS_ALL_QUOTA_ACCT
;
1330 parsing_mp
->m_qflags
&= ~XFS_ALL_QUOTA_ENFD
;
1335 parsing_mp
->m_qflags
|= (XFS_UQUOTA_ACCT
| XFS_UQUOTA_ENFD
);
1337 case Opt_qnoenforce
:
1338 case Opt_uqnoenforce
:
1339 parsing_mp
->m_qflags
|= XFS_UQUOTA_ACCT
;
1340 parsing_mp
->m_qflags
&= ~XFS_UQUOTA_ENFD
;
1344 parsing_mp
->m_qflags
|= (XFS_PQUOTA_ACCT
| XFS_PQUOTA_ENFD
);
1346 case Opt_pqnoenforce
:
1347 parsing_mp
->m_qflags
|= XFS_PQUOTA_ACCT
;
1348 parsing_mp
->m_qflags
&= ~XFS_PQUOTA_ENFD
;
1352 parsing_mp
->m_qflags
|= (XFS_GQUOTA_ACCT
| XFS_GQUOTA_ENFD
);
1354 case Opt_gqnoenforce
:
1355 parsing_mp
->m_qflags
|= XFS_GQUOTA_ACCT
;
1356 parsing_mp
->m_qflags
&= ~XFS_GQUOTA_ENFD
;
1359 parsing_mp
->m_features
|= XFS_FEAT_DISCARD
;
1362 parsing_mp
->m_features
&= ~XFS_FEAT_DISCARD
;
1364 #ifdef CONFIG_FS_DAX
1366 xfs_mount_set_dax_mode(parsing_mp
, XFS_DAX_ALWAYS
);
1369 xfs_mount_set_dax_mode(parsing_mp
, result
.uint_32
);
1372 /* Following mount options will be removed in September 2025 */
1374 xfs_fs_warn_deprecated(fc
, param
, XFS_FEAT_IKEEP
, true);
1375 parsing_mp
->m_features
|= XFS_FEAT_IKEEP
;
1378 xfs_fs_warn_deprecated(fc
, param
, XFS_FEAT_IKEEP
, false);
1379 parsing_mp
->m_features
&= ~XFS_FEAT_IKEEP
;
1382 xfs_fs_warn_deprecated(fc
, param
, XFS_FEAT_ATTR2
, true);
1383 parsing_mp
->m_features
|= XFS_FEAT_ATTR2
;
1386 xfs_fs_warn_deprecated(fc
, param
, XFS_FEAT_NOATTR2
, true);
1387 parsing_mp
->m_features
|= XFS_FEAT_NOATTR2
;
1390 xfs_warn(parsing_mp
, "unknown mount option [%s].", param
->key
);
1398 xfs_fs_validate_params(
1399 struct xfs_mount
*mp
)
1401 /* No recovery flag requires a read-only mount */
1402 if (xfs_has_norecovery(mp
) && !xfs_is_readonly(mp
)) {
1403 xfs_warn(mp
, "no-recovery mounts must be read-only.");
1408 * We have not read the superblock at this point, so only the attr2
1409 * mount option can set the attr2 feature by this stage.
1411 if (xfs_has_attr2(mp
) && xfs_has_noattr2(mp
)) {
1412 xfs_warn(mp
, "attr2 and noattr2 cannot both be specified.");
1417 if (xfs_has_noalign(mp
) && (mp
->m_dalign
|| mp
->m_swidth
)) {
1419 "sunit and swidth options incompatible with the noalign option");
1423 if (!IS_ENABLED(CONFIG_XFS_QUOTA
) && mp
->m_qflags
!= 0) {
1424 xfs_warn(mp
, "quota support not available in this kernel.");
1428 if ((mp
->m_dalign
&& !mp
->m_swidth
) ||
1429 (!mp
->m_dalign
&& mp
->m_swidth
)) {
1430 xfs_warn(mp
, "sunit and swidth must be specified together");
1434 if (mp
->m_dalign
&& (mp
->m_swidth
% mp
->m_dalign
!= 0)) {
1436 "stripe width (%d) must be a multiple of the stripe unit (%d)",
1437 mp
->m_swidth
, mp
->m_dalign
);
1441 if (mp
->m_logbufs
!= -1 &&
1442 mp
->m_logbufs
!= 0 &&
1443 (mp
->m_logbufs
< XLOG_MIN_ICLOGS
||
1444 mp
->m_logbufs
> XLOG_MAX_ICLOGS
)) {
1445 xfs_warn(mp
, "invalid logbufs value: %d [not %d-%d]",
1446 mp
->m_logbufs
, XLOG_MIN_ICLOGS
, XLOG_MAX_ICLOGS
);
1450 if (mp
->m_logbsize
!= -1 &&
1451 mp
->m_logbsize
!= 0 &&
1452 (mp
->m_logbsize
< XLOG_MIN_RECORD_BSIZE
||
1453 mp
->m_logbsize
> XLOG_MAX_RECORD_BSIZE
||
1454 !is_power_of_2(mp
->m_logbsize
))) {
1456 "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
1461 if (xfs_has_allocsize(mp
) &&
1462 (mp
->m_allocsize_log
> XFS_MAX_IO_LOG
||
1463 mp
->m_allocsize_log
< XFS_MIN_IO_LOG
)) {
1464 xfs_warn(mp
, "invalid log iosize: %d [not %d-%d]",
1465 mp
->m_allocsize_log
, XFS_MIN_IO_LOG
, XFS_MAX_IO_LOG
);
1475 struct dentry
*parent
)
1477 struct dentry
*child
;
1479 /* Apparently we're expected to ignore error returns?? */
1480 child
= debugfs_create_dir(name
, parent
);
1489 struct super_block
*sb
,
1490 struct fs_context
*fc
)
1492 struct xfs_mount
*mp
= sb
->s_fs_info
;
1494 int flags
= 0, error
;
1499 * Copy VFS mount flags from the context now that all parameter parsing
1500 * is guaranteed to have been completed by either the old mount API or
1501 * the newer fsopen/fsconfig API.
1503 if (fc
->sb_flags
& SB_RDONLY
)
1504 set_bit(XFS_OPSTATE_READONLY
, &mp
->m_opstate
);
1505 if (fc
->sb_flags
& SB_DIRSYNC
)
1506 mp
->m_features
|= XFS_FEAT_DIRSYNC
;
1507 if (fc
->sb_flags
& SB_SYNCHRONOUS
)
1508 mp
->m_features
|= XFS_FEAT_WSYNC
;
1510 error
= xfs_fs_validate_params(mp
);
1514 sb_min_blocksize(sb
, BBSIZE
);
1515 sb
->s_xattr
= xfs_xattr_handlers
;
1516 sb
->s_export_op
= &xfs_export_operations
;
1517 #ifdef CONFIG_XFS_QUOTA
1518 sb
->s_qcop
= &xfs_quotactl_operations
;
1519 sb
->s_quota_types
= QTYPE_MASK_USR
| QTYPE_MASK_GRP
| QTYPE_MASK_PRJ
;
1521 sb
->s_op
= &xfs_super_operations
;
1524 * Delay mount work if the debug hook is set. This is debug
1525 * instrumention to coordinate simulation of xfs mount failures with
1526 * VFS superblock operations
1528 if (xfs_globals
.mount_delay
) {
1529 xfs_notice(mp
, "Delaying mount for %d seconds.",
1530 xfs_globals
.mount_delay
);
1531 msleep(xfs_globals
.mount_delay
* 1000);
1534 if (fc
->sb_flags
& SB_SILENT
)
1535 flags
|= XFS_MFSI_QUIET
;
1537 error
= xfs_open_devices(mp
);
1542 mp
->m_debugfs
= xfs_debugfs_mkdir(mp
->m_super
->s_id
,
1545 mp
->m_debugfs
= NULL
;
1548 error
= xfs_init_mount_workqueues(mp
);
1550 goto out_shutdown_devices
;
1552 error
= xfs_init_percpu_counters(mp
);
1554 goto out_destroy_workqueues
;
1556 error
= xfs_inodegc_init_percpu(mp
);
1558 goto out_destroy_counters
;
1560 /* Allocate stats memory before we do operations that might use it */
1561 mp
->m_stats
.xs_stats
= alloc_percpu(struct xfsstats
);
1562 if (!mp
->m_stats
.xs_stats
) {
1564 goto out_destroy_inodegc
;
1567 error
= xchk_mount_stats_alloc(mp
);
1569 goto out_free_stats
;
1571 error
= xfs_readsb(mp
, flags
);
1573 goto out_free_scrub_stats
;
1575 error
= xfs_finish_flags(mp
);
1579 error
= xfs_setup_devices(mp
);
1583 /* V4 support is undergoing deprecation. */
1584 if (!xfs_has_crc(mp
)) {
1585 #ifdef CONFIG_XFS_SUPPORT_V4
1587 "Deprecated V4 format (crc=0) will not be supported after September 2030.");
1590 "Deprecated V4 format (crc=0) not supported by kernel.");
1596 /* ASCII case insensitivity is undergoing deprecation. */
1597 if (xfs_has_asciici(mp
)) {
1598 #ifdef CONFIG_XFS_SUPPORT_ASCII_CI
1600 "Deprecated ASCII case-insensitivity feature (ascii-ci=1) will not be supported after September 2030.");
1603 "Deprecated ASCII case-insensitivity feature (ascii-ci=1) not supported by kernel.");
1609 /* Filesystem claims it needs repair, so refuse the mount. */
1610 if (xfs_has_needsrepair(mp
)) {
1611 xfs_warn(mp
, "Filesystem needs repair. Please run xfs_repair.");
1612 error
= -EFSCORRUPTED
;
1617 * Don't touch the filesystem if a user tool thinks it owns the primary
1618 * superblock. mkfs doesn't clear the flag from secondary supers, so
1619 * we don't check them at all.
1621 if (mp
->m_sb
.sb_inprogress
) {
1622 xfs_warn(mp
, "Offline file system operation in progress!");
1623 error
= -EFSCORRUPTED
;
1628 * Until this is fixed only page-sized or smaller data blocks work.
1630 if (mp
->m_sb
.sb_blocksize
> PAGE_SIZE
) {
1632 "File system with blocksize %d bytes. "
1633 "Only pagesize (%ld) or less will currently work.",
1634 mp
->m_sb
.sb_blocksize
, PAGE_SIZE
);
1639 /* Ensure this filesystem fits in the page cache limits */
1640 if (xfs_sb_validate_fsb_count(&mp
->m_sb
, mp
->m_sb
.sb_dblocks
) ||
1641 xfs_sb_validate_fsb_count(&mp
->m_sb
, mp
->m_sb
.sb_rblocks
)) {
1643 "file system too large to be mounted on this system.");
1649 * XFS block mappings use 54 bits to store the logical block offset.
1650 * This should suffice to handle the maximum file size that the VFS
1651 * supports (currently 2^63 bytes on 64-bit and ULONG_MAX << PAGE_SHIFT
1652 * bytes on 32-bit), but as XFS and VFS have gotten the s_maxbytes
1653 * calculation wrong on 32-bit kernels in the past, we'll add a WARN_ON
1654 * to check this assertion.
1656 * Avoid integer overflow by comparing the maximum bmbt offset to the
1657 * maximum pagecache offset in units of fs blocks.
1659 if (!xfs_verify_fileoff(mp
, XFS_B_TO_FSBT(mp
, MAX_LFS_FILESIZE
))) {
1661 "MAX_LFS_FILESIZE block offset (%llu) exceeds extent map maximum (%llu)!",
1662 XFS_B_TO_FSBT(mp
, MAX_LFS_FILESIZE
),
1668 error
= xfs_filestream_mount(mp
);
1673 * we must configure the block size in the superblock before we run the
1674 * full mount process as the mount process can lookup and cache inodes.
1676 sb
->s_magic
= XFS_SUPER_MAGIC
;
1677 sb
->s_blocksize
= mp
->m_sb
.sb_blocksize
;
1678 sb
->s_blocksize_bits
= ffs(sb
->s_blocksize
) - 1;
1679 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
1680 sb
->s_max_links
= XFS_MAXLINK
;
1681 sb
->s_time_gran
= 1;
1682 if (xfs_has_bigtime(mp
)) {
1683 sb
->s_time_min
= xfs_bigtime_to_unix(XFS_BIGTIME_TIME_MIN
);
1684 sb
->s_time_max
= xfs_bigtime_to_unix(XFS_BIGTIME_TIME_MAX
);
1686 sb
->s_time_min
= XFS_LEGACY_TIME_MIN
;
1687 sb
->s_time_max
= XFS_LEGACY_TIME_MAX
;
1689 trace_xfs_inode_timestamp_range(mp
, sb
->s_time_min
, sb
->s_time_max
);
1690 sb
->s_iflags
|= SB_I_CGROUPWB
;
1692 set_posix_acl_flag(sb
);
1694 /* version 5 superblocks support inode version counters. */
1695 if (xfs_has_crc(mp
))
1696 sb
->s_flags
|= SB_I_VERSION
;
1698 if (xfs_has_dax_always(mp
)) {
1699 error
= xfs_setup_dax_always(mp
);
1701 goto out_filestream_unmount
;
1704 if (xfs_has_discard(mp
) && !bdev_max_discard_sectors(sb
->s_bdev
)) {
1706 "mounting with \"discard\" option, but the device does not support discard");
1707 mp
->m_features
&= ~XFS_FEAT_DISCARD
;
1710 if (xfs_has_reflink(mp
)) {
1711 if (mp
->m_sb
.sb_rblocks
) {
1713 "reflink not compatible with realtime device!");
1715 goto out_filestream_unmount
;
1718 if (xfs_globals
.always_cow
) {
1719 xfs_info(mp
, "using DEBUG-only always_cow mode.");
1720 mp
->m_always_cow
= true;
1724 if (xfs_has_rmapbt(mp
) && mp
->m_sb
.sb_rblocks
) {
1726 "reverse mapping btree not compatible with realtime device!");
1728 goto out_filestream_unmount
;
1731 error
= xfs_mountfs(mp
);
1733 goto out_filestream_unmount
;
1735 root
= igrab(VFS_I(mp
->m_rootip
));
1740 sb
->s_root
= d_make_root(root
);
1748 out_filestream_unmount
:
1749 xfs_filestream_unmount(mp
);
1752 out_free_scrub_stats
:
1753 xchk_mount_stats_free(mp
);
1755 free_percpu(mp
->m_stats
.xs_stats
);
1756 out_destroy_inodegc
:
1757 xfs_inodegc_free_percpu(mp
);
1758 out_destroy_counters
:
1759 xfs_destroy_percpu_counters(mp
);
1760 out_destroy_workqueues
:
1761 xfs_destroy_mount_workqueues(mp
);
1762 out_shutdown_devices
:
1763 xfs_shutdown_devices(mp
);
1767 xfs_filestream_unmount(mp
);
1774 struct fs_context
*fc
)
1776 return get_tree_bdev(fc
, xfs_fs_fill_super
);
1781 struct xfs_mount
*mp
)
1783 struct xfs_sb
*sbp
= &mp
->m_sb
;
1786 if (xfs_has_norecovery(mp
)) {
1788 "ro->rw transition prohibited on norecovery mount");
1792 if (xfs_sb_is_v5(sbp
) &&
1793 xfs_sb_has_ro_compat_feature(sbp
, XFS_SB_FEAT_RO_COMPAT_UNKNOWN
)) {
1795 "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem",
1796 (sbp
->sb_features_ro_compat
&
1797 XFS_SB_FEAT_RO_COMPAT_UNKNOWN
));
1801 clear_bit(XFS_OPSTATE_READONLY
, &mp
->m_opstate
);
1804 * If this is the first remount to writeable state we might have some
1805 * superblock changes to update.
1807 if (mp
->m_update_sb
) {
1808 error
= xfs_sync_sb(mp
, false);
1810 xfs_warn(mp
, "failed to write sb changes");
1813 mp
->m_update_sb
= false;
1817 * Fill out the reserve pool if it is empty. Use the stashed value if
1818 * it is non-zero, otherwise go with the default.
1820 xfs_restore_resvblks(mp
);
1821 xfs_log_work_queue(mp
);
1822 xfs_blockgc_start(mp
);
1824 /* Create the per-AG metadata reservation pool .*/
1825 error
= xfs_fs_reserve_ag_blocks(mp
);
1826 if (error
&& error
!= -ENOSPC
)
1829 /* Re-enable the background inode inactivation worker. */
1830 xfs_inodegc_start(mp
);
1837 struct xfs_mount
*mp
)
1839 struct xfs_icwalk icw
= {
1840 .icw_flags
= XFS_ICWALK_FLAG_SYNC
,
1844 /* Flush all the dirty data to disk. */
1845 error
= sync_filesystem(mp
->m_super
);
1850 * Cancel background eofb scanning so it cannot race with the final
1851 * log force+buftarg wait and deadlock the remount.
1853 xfs_blockgc_stop(mp
);
1856 * Clear out all remaining COW staging extents and speculative post-EOF
1857 * preallocations so that we don't leave inodes requiring inactivation
1858 * cleanups during reclaim on a read-only mount. We must process every
1859 * cached inode, so this requires a synchronous cache scan.
1861 error
= xfs_blockgc_free_space(mp
, &icw
);
1863 xfs_force_shutdown(mp
, SHUTDOWN_CORRUPT_INCORE
);
1868 * Stop the inodegc background worker. xfs_fs_reconfigure already
1869 * flushed all pending inodegc work when it sync'd the filesystem.
1870 * The VFS holds s_umount, so we know that inodes cannot enter
1871 * xfs_fs_destroy_inode during a remount operation. In readonly mode
1872 * we send inodes straight to reclaim, so no inodes will be queued.
1874 xfs_inodegc_stop(mp
);
1876 /* Free the per-AG metadata reservation pool. */
1877 error
= xfs_fs_unreserve_ag_blocks(mp
);
1879 xfs_force_shutdown(mp
, SHUTDOWN_CORRUPT_INCORE
);
1884 * Before we sync the metadata, we need to free up the reserve block
1885 * pool so that the used block count in the superblock on disk is
1886 * correct at the end of the remount. Stash the current* reserve pool
1887 * size so that if we get remounted rw, we can return it to the same
1890 xfs_save_resvblks(mp
);
1893 set_bit(XFS_OPSTATE_READONLY
, &mp
->m_opstate
);
1899 * Logically we would return an error here to prevent users from believing
1900 * they might have changed mount options using remount which can't be changed.
1902 * But unfortunately mount(8) adds all options from mtab and fstab to the mount
1903 * arguments in some cases so we can't blindly reject options, but have to
1904 * check for each specified option if it actually differs from the currently
1905 * set option and only reject it if that's the case.
1907 * Until that is implemented we return success for every remount request, and
1908 * silently ignore all options that we can't actually change.
1912 struct fs_context
*fc
)
1914 struct xfs_mount
*mp
= XFS_M(fc
->root
->d_sb
);
1915 struct xfs_mount
*new_mp
= fc
->s_fs_info
;
1916 int flags
= fc
->sb_flags
;
1919 /* version 5 superblocks always support version counters. */
1920 if (xfs_has_crc(mp
))
1921 fc
->sb_flags
|= SB_I_VERSION
;
1923 error
= xfs_fs_validate_params(new_mp
);
1927 /* inode32 -> inode64 */
1928 if (xfs_has_small_inums(mp
) && !xfs_has_small_inums(new_mp
)) {
1929 mp
->m_features
&= ~XFS_FEAT_SMALL_INUMS
;
1930 mp
->m_maxagi
= xfs_set_inode_alloc(mp
, mp
->m_sb
.sb_agcount
);
1933 /* inode64 -> inode32 */
1934 if (!xfs_has_small_inums(mp
) && xfs_has_small_inums(new_mp
)) {
1935 mp
->m_features
|= XFS_FEAT_SMALL_INUMS
;
1936 mp
->m_maxagi
= xfs_set_inode_alloc(mp
, mp
->m_sb
.sb_agcount
);
1940 if (xfs_is_readonly(mp
) && !(flags
& SB_RDONLY
)) {
1941 error
= xfs_remount_rw(mp
);
1947 if (!xfs_is_readonly(mp
) && (flags
& SB_RDONLY
)) {
1948 error
= xfs_remount_ro(mp
);
1958 struct fs_context
*fc
)
1960 struct xfs_mount
*mp
= fc
->s_fs_info
;
1963 * mp is stored in the fs_context when it is initialized.
1964 * mp is transferred to the superblock on a successful mount,
1965 * but if an error occurs before the transfer we have to free
1972 static const struct fs_context_operations xfs_context_ops
= {
1973 .parse_param
= xfs_fs_parse_param
,
1974 .get_tree
= xfs_fs_get_tree
,
1975 .reconfigure
= xfs_fs_reconfigure
,
1976 .free
= xfs_fs_free
,
1980 * WARNING: do not initialise any parameters in this function that depend on
1981 * mount option parsing having already been performed as this can be called from
1982 * fsopen() before any parameters have been set.
1984 static int xfs_init_fs_context(
1985 struct fs_context
*fc
)
1987 struct xfs_mount
*mp
;
1989 mp
= kmem_alloc(sizeof(struct xfs_mount
), KM_ZERO
);
1993 spin_lock_init(&mp
->m_sb_lock
);
1994 INIT_RADIX_TREE(&mp
->m_perag_tree
, GFP_ATOMIC
);
1995 spin_lock_init(&mp
->m_perag_lock
);
1996 mutex_init(&mp
->m_growlock
);
1997 INIT_WORK(&mp
->m_flush_inodes_work
, xfs_flush_inodes_worker
);
1998 INIT_DELAYED_WORK(&mp
->m_reclaim_work
, xfs_reclaim_worker
);
1999 mp
->m_kobj
.kobject
.kset
= xfs_kset
;
2001 * We don't create the finobt per-ag space reservation until after log
2002 * recovery, so we must set this to true so that an ifree transaction
2003 * started during log recovery will not depend on space reservations
2004 * for finobt expansion.
2006 mp
->m_finobt_nores
= true;
2009 * These can be overridden by the mount option parsing.
2012 mp
->m_logbsize
= -1;
2013 mp
->m_allocsize_log
= 16; /* 64k */
2016 fc
->ops
= &xfs_context_ops
;
2023 struct super_block
*sb
)
2025 kill_block_super(sb
);
2026 xfs_mount_free(XFS_M(sb
));
2029 static struct file_system_type xfs_fs_type
= {
2030 .owner
= THIS_MODULE
,
2032 .init_fs_context
= xfs_init_fs_context
,
2033 .parameters
= xfs_fs_parameters
,
2034 .kill_sb
= xfs_kill_sb
,
2035 .fs_flags
= FS_REQUIRES_DEV
| FS_ALLOW_IDMAP
,
2037 MODULE_ALIAS_FS("xfs");
2040 xfs_init_caches(void)
2044 xfs_buf_cache
= kmem_cache_create("xfs_buf", sizeof(struct xfs_buf
), 0,
2045 SLAB_HWCACHE_ALIGN
|
2046 SLAB_RECLAIM_ACCOUNT
|
2052 xfs_log_ticket_cache
= kmem_cache_create("xfs_log_ticket",
2053 sizeof(struct xlog_ticket
),
2055 if (!xfs_log_ticket_cache
)
2056 goto out_destroy_buf_cache
;
2058 error
= xfs_btree_init_cur_caches();
2060 goto out_destroy_log_ticket_cache
;
2062 error
= xfs_defer_init_item_caches();
2064 goto out_destroy_btree_cur_cache
;
2066 xfs_da_state_cache
= kmem_cache_create("xfs_da_state",
2067 sizeof(struct xfs_da_state
),
2069 if (!xfs_da_state_cache
)
2070 goto out_destroy_defer_item_cache
;
2072 xfs_ifork_cache
= kmem_cache_create("xfs_ifork",
2073 sizeof(struct xfs_ifork
),
2075 if (!xfs_ifork_cache
)
2076 goto out_destroy_da_state_cache
;
2078 xfs_trans_cache
= kmem_cache_create("xfs_trans",
2079 sizeof(struct xfs_trans
),
2081 if (!xfs_trans_cache
)
2082 goto out_destroy_ifork_cache
;
2086 * The size of the cache-allocated buf log item is the maximum
2087 * size possible under XFS. This wastes a little bit of memory,
2088 * but it is much faster.
2090 xfs_buf_item_cache
= kmem_cache_create("xfs_buf_item",
2091 sizeof(struct xfs_buf_log_item
),
2093 if (!xfs_buf_item_cache
)
2094 goto out_destroy_trans_cache
;
2096 xfs_efd_cache
= kmem_cache_create("xfs_efd_item",
2097 xfs_efd_log_item_sizeof(XFS_EFD_MAX_FAST_EXTENTS
),
2100 goto out_destroy_buf_item_cache
;
2102 xfs_efi_cache
= kmem_cache_create("xfs_efi_item",
2103 xfs_efi_log_item_sizeof(XFS_EFI_MAX_FAST_EXTENTS
),
2106 goto out_destroy_efd_cache
;
2108 xfs_inode_cache
= kmem_cache_create("xfs_inode",
2109 sizeof(struct xfs_inode
), 0,
2110 (SLAB_HWCACHE_ALIGN
|
2111 SLAB_RECLAIM_ACCOUNT
|
2112 SLAB_MEM_SPREAD
| SLAB_ACCOUNT
),
2113 xfs_fs_inode_init_once
);
2114 if (!xfs_inode_cache
)
2115 goto out_destroy_efi_cache
;
2117 xfs_ili_cache
= kmem_cache_create("xfs_ili",
2118 sizeof(struct xfs_inode_log_item
), 0,
2119 SLAB_RECLAIM_ACCOUNT
| SLAB_MEM_SPREAD
,
2122 goto out_destroy_inode_cache
;
2124 xfs_icreate_cache
= kmem_cache_create("xfs_icr",
2125 sizeof(struct xfs_icreate_item
),
2127 if (!xfs_icreate_cache
)
2128 goto out_destroy_ili_cache
;
2130 xfs_rud_cache
= kmem_cache_create("xfs_rud_item",
2131 sizeof(struct xfs_rud_log_item
),
2134 goto out_destroy_icreate_cache
;
2136 xfs_rui_cache
= kmem_cache_create("xfs_rui_item",
2137 xfs_rui_log_item_sizeof(XFS_RUI_MAX_FAST_EXTENTS
),
2140 goto out_destroy_rud_cache
;
2142 xfs_cud_cache
= kmem_cache_create("xfs_cud_item",
2143 sizeof(struct xfs_cud_log_item
),
2146 goto out_destroy_rui_cache
;
2148 xfs_cui_cache
= kmem_cache_create("xfs_cui_item",
2149 xfs_cui_log_item_sizeof(XFS_CUI_MAX_FAST_EXTENTS
),
2152 goto out_destroy_cud_cache
;
2154 xfs_bud_cache
= kmem_cache_create("xfs_bud_item",
2155 sizeof(struct xfs_bud_log_item
),
2158 goto out_destroy_cui_cache
;
2160 xfs_bui_cache
= kmem_cache_create("xfs_bui_item",
2161 xfs_bui_log_item_sizeof(XFS_BUI_MAX_FAST_EXTENTS
),
2164 goto out_destroy_bud_cache
;
2166 xfs_attrd_cache
= kmem_cache_create("xfs_attrd_item",
2167 sizeof(struct xfs_attrd_log_item
),
2169 if (!xfs_attrd_cache
)
2170 goto out_destroy_bui_cache
;
2172 xfs_attri_cache
= kmem_cache_create("xfs_attri_item",
2173 sizeof(struct xfs_attri_log_item
),
2175 if (!xfs_attri_cache
)
2176 goto out_destroy_attrd_cache
;
2178 xfs_iunlink_cache
= kmem_cache_create("xfs_iul_item",
2179 sizeof(struct xfs_iunlink_item
),
2181 if (!xfs_iunlink_cache
)
2182 goto out_destroy_attri_cache
;
2186 out_destroy_attri_cache
:
2187 kmem_cache_destroy(xfs_attri_cache
);
2188 out_destroy_attrd_cache
:
2189 kmem_cache_destroy(xfs_attrd_cache
);
2190 out_destroy_bui_cache
:
2191 kmem_cache_destroy(xfs_bui_cache
);
2192 out_destroy_bud_cache
:
2193 kmem_cache_destroy(xfs_bud_cache
);
2194 out_destroy_cui_cache
:
2195 kmem_cache_destroy(xfs_cui_cache
);
2196 out_destroy_cud_cache
:
2197 kmem_cache_destroy(xfs_cud_cache
);
2198 out_destroy_rui_cache
:
2199 kmem_cache_destroy(xfs_rui_cache
);
2200 out_destroy_rud_cache
:
2201 kmem_cache_destroy(xfs_rud_cache
);
2202 out_destroy_icreate_cache
:
2203 kmem_cache_destroy(xfs_icreate_cache
);
2204 out_destroy_ili_cache
:
2205 kmem_cache_destroy(xfs_ili_cache
);
2206 out_destroy_inode_cache
:
2207 kmem_cache_destroy(xfs_inode_cache
);
2208 out_destroy_efi_cache
:
2209 kmem_cache_destroy(xfs_efi_cache
);
2210 out_destroy_efd_cache
:
2211 kmem_cache_destroy(xfs_efd_cache
);
2212 out_destroy_buf_item_cache
:
2213 kmem_cache_destroy(xfs_buf_item_cache
);
2214 out_destroy_trans_cache
:
2215 kmem_cache_destroy(xfs_trans_cache
);
2216 out_destroy_ifork_cache
:
2217 kmem_cache_destroy(xfs_ifork_cache
);
2218 out_destroy_da_state_cache
:
2219 kmem_cache_destroy(xfs_da_state_cache
);
2220 out_destroy_defer_item_cache
:
2221 xfs_defer_destroy_item_caches();
2222 out_destroy_btree_cur_cache
:
2223 xfs_btree_destroy_cur_caches();
2224 out_destroy_log_ticket_cache
:
2225 kmem_cache_destroy(xfs_log_ticket_cache
);
2226 out_destroy_buf_cache
:
2227 kmem_cache_destroy(xfs_buf_cache
);
2233 xfs_destroy_caches(void)
2236 * Make sure all delayed rcu free are flushed before we
2240 kmem_cache_destroy(xfs_iunlink_cache
);
2241 kmem_cache_destroy(xfs_attri_cache
);
2242 kmem_cache_destroy(xfs_attrd_cache
);
2243 kmem_cache_destroy(xfs_bui_cache
);
2244 kmem_cache_destroy(xfs_bud_cache
);
2245 kmem_cache_destroy(xfs_cui_cache
);
2246 kmem_cache_destroy(xfs_cud_cache
);
2247 kmem_cache_destroy(xfs_rui_cache
);
2248 kmem_cache_destroy(xfs_rud_cache
);
2249 kmem_cache_destroy(xfs_icreate_cache
);
2250 kmem_cache_destroy(xfs_ili_cache
);
2251 kmem_cache_destroy(xfs_inode_cache
);
2252 kmem_cache_destroy(xfs_efi_cache
);
2253 kmem_cache_destroy(xfs_efd_cache
);
2254 kmem_cache_destroy(xfs_buf_item_cache
);
2255 kmem_cache_destroy(xfs_trans_cache
);
2256 kmem_cache_destroy(xfs_ifork_cache
);
2257 kmem_cache_destroy(xfs_da_state_cache
);
2258 xfs_defer_destroy_item_caches();
2259 xfs_btree_destroy_cur_caches();
2260 kmem_cache_destroy(xfs_log_ticket_cache
);
2261 kmem_cache_destroy(xfs_buf_cache
);
2265 xfs_init_workqueues(void)
2268 * The allocation workqueue can be used in memory reclaim situations
2269 * (writepage path), and parallelism is only limited by the number of
2270 * AGs in all the filesystems mounted. Hence use the default large
2271 * max_active value for this workqueue.
2273 xfs_alloc_wq
= alloc_workqueue("xfsalloc",
2274 XFS_WQFLAGS(WQ_MEM_RECLAIM
| WQ_FREEZABLE
), 0);
2278 xfs_discard_wq
= alloc_workqueue("xfsdiscard", XFS_WQFLAGS(WQ_UNBOUND
),
2280 if (!xfs_discard_wq
)
2281 goto out_free_alloc_wq
;
2285 destroy_workqueue(xfs_alloc_wq
);
2290 xfs_destroy_workqueues(void)
2292 destroy_workqueue(xfs_discard_wq
);
2293 destroy_workqueue(xfs_alloc_wq
);
2301 xfs_check_ondisk_structs();
2303 error
= xfs_dahash_test();
2307 printk(KERN_INFO XFS_VERSION_STRING
" with "
2308 XFS_BUILD_OPTIONS
" enabled\n");
2312 error
= xfs_init_caches();
2316 error
= xfs_init_workqueues();
2318 goto out_destroy_caches
;
2320 error
= xfs_mru_cache_init();
2322 goto out_destroy_wq
;
2324 error
= xfs_init_procfs();
2326 goto out_mru_cache_uninit
;
2328 error
= xfs_sysctl_register();
2330 goto out_cleanup_procfs
;
2332 xfs_debugfs
= xfs_debugfs_mkdir("xfs", NULL
);
2334 xfs_kset
= kset_create_and_add("xfs", NULL
, fs_kobj
);
2337 goto out_debugfs_unregister
;
2340 xfsstats
.xs_kobj
.kobject
.kset
= xfs_kset
;
2342 xfsstats
.xs_stats
= alloc_percpu(struct xfsstats
);
2343 if (!xfsstats
.xs_stats
) {
2345 goto out_kset_unregister
;
2348 error
= xfs_sysfs_init(&xfsstats
.xs_kobj
, &xfs_stats_ktype
, NULL
,
2351 goto out_free_stats
;
2353 error
= xchk_global_stats_setup(xfs_debugfs
);
2355 goto out_remove_stats_kobj
;
2358 xfs_dbg_kobj
.kobject
.kset
= xfs_kset
;
2359 error
= xfs_sysfs_init(&xfs_dbg_kobj
, &xfs_dbg_ktype
, NULL
, "debug");
2361 goto out_remove_scrub_stats
;
2364 error
= xfs_qm_init();
2366 goto out_remove_dbg_kobj
;
2368 error
= register_filesystem(&xfs_fs_type
);
2375 out_remove_dbg_kobj
:
2377 xfs_sysfs_del(&xfs_dbg_kobj
);
2378 out_remove_scrub_stats
:
2380 xchk_global_stats_teardown();
2381 out_remove_stats_kobj
:
2382 xfs_sysfs_del(&xfsstats
.xs_kobj
);
2384 free_percpu(xfsstats
.xs_stats
);
2385 out_kset_unregister
:
2386 kset_unregister(xfs_kset
);
2387 out_debugfs_unregister
:
2388 debugfs_remove(xfs_debugfs
);
2389 xfs_sysctl_unregister();
2391 xfs_cleanup_procfs();
2392 out_mru_cache_uninit
:
2393 xfs_mru_cache_uninit();
2395 xfs_destroy_workqueues();
2397 xfs_destroy_caches();
2406 unregister_filesystem(&xfs_fs_type
);
2408 xfs_sysfs_del(&xfs_dbg_kobj
);
2410 xchk_global_stats_teardown();
2411 xfs_sysfs_del(&xfsstats
.xs_kobj
);
2412 free_percpu(xfsstats
.xs_stats
);
2413 kset_unregister(xfs_kset
);
2414 debugfs_remove(xfs_debugfs
);
2415 xfs_sysctl_unregister();
2416 xfs_cleanup_procfs();
2417 xfs_mru_cache_uninit();
2418 xfs_destroy_workqueues();
2419 xfs_destroy_caches();
2420 xfs_uuid_table_free();
2423 module_init(init_xfs_fs
);
2424 module_exit(exit_xfs_fs
);
2426 MODULE_AUTHOR("Silicon Graphics, Inc.");
2427 MODULE_DESCRIPTION(XFS_VERSION_STRING
" with " XFS_BUILD_OPTIONS
" enabled");
2428 MODULE_LICENSE("GPL");