blob: e39b02351b4a257e92f9e2506a8af8f31aecc688 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Nathan Scott7b718762005-11-02 14:58:39 +11002 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 *
Nathan Scott7b718762005-11-02 14:58:39 +11005 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 * published by the Free Software Foundation.
8 *
Nathan Scott7b718762005-11-02 14:58:39 +11009 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
Linus Torvalds1da177e2005-04-16 15:20:36 -070013 *
Nathan Scott7b718762005-11-02 14:58:39 +110014 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
Linus Torvalds1da177e2005-04-16 15:20:36 -070017 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include "xfs.h"
Nathan Scotta844f452005-11-02 14:38:42 +110019#include "xfs_fs.h"
Dave Chinner70a98832013-10-23 10:36:05 +110020#include "xfs_shared.h"
Dave Chinner239880e2013-10-23 10:50:10 +110021#include "xfs_format.h"
22#include "xfs_log_format.h"
23#include "xfs_trans_resv.h"
Nathan Scotta844f452005-11-02 14:38:42 +110024#include "xfs_bit.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include "xfs_sb.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include "xfs_mount.h"
Dave Chinner57062782013-10-15 09:17:51 +110027#include "xfs_da_format.h"
Dave Chinner9a2cc412014-12-04 09:43:17 +110028#include "xfs_da_btree.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include "xfs_inode.h"
Dave Chinnera4fbe6a2013-10-23 10:51:50 +110030#include "xfs_dir2.h"
Nathan Scotta844f452005-11-02 14:38:42 +110031#include "xfs_ialloc.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include "xfs_alloc.h"
33#include "xfs_rtalloc.h"
34#include "xfs_bmap.h"
Dave Chinnera4fbe6a2013-10-23 10:51:50 +110035#include "xfs_trans.h"
36#include "xfs_trans_priv.h"
37#include "xfs_log.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#include "xfs_error.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070039#include "xfs_quota.h"
40#include "xfs_fsops.h"
Christoph Hellwig0b1b2132009-12-14 23:14:59 +000041#include "xfs_trace.h"
Dave Chinner6d8b79c2012-10-08 21:56:09 +110042#include "xfs_icache.h"
Brian Fostera31b1d32014-07-15 08:07:01 +100043#include "xfs_sysfs.h"
Christoph Hellwig0b1b2132009-12-14 23:14:59 +000044
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
Christoph Hellwig27174202009-03-30 10:21:31 +020046static DEFINE_MUTEX(xfs_uuid_table_mutex);
47static int xfs_uuid_table_size;
48static uuid_t *xfs_uuid_table;
49
Darrick J. Wongaf3b6382015-11-03 13:06:34 +110050void
51xfs_uuid_table_free(void)
52{
53 if (xfs_uuid_table_size == 0)
54 return;
55 kmem_free(xfs_uuid_table);
56 xfs_uuid_table = NULL;
57 xfs_uuid_table_size = 0;
58}
59
Christoph Hellwig27174202009-03-30 10:21:31 +020060/*
61 * See if the UUID is unique among mounted XFS filesystems.
62 * Mount fails if UUID is nil or a FS with the same UUID is already mounted.
63 */
64STATIC int
65xfs_uuid_mount(
66 struct xfs_mount *mp)
67{
68 uuid_t *uuid = &mp->m_sb.sb_uuid;
69 int hole, i;
70
71 if (mp->m_flags & XFS_MOUNT_NOUUID)
72 return 0;
73
74 if (uuid_is_nil(uuid)) {
Dave Chinner0b932cc2011-03-07 10:08:35 +110075 xfs_warn(mp, "Filesystem has nil UUID - can't mount");
Dave Chinner24513372014-06-25 14:58:08 +100076 return -EINVAL;
Christoph Hellwig27174202009-03-30 10:21:31 +020077 }
78
79 mutex_lock(&xfs_uuid_table_mutex);
80 for (i = 0, hole = -1; i < xfs_uuid_table_size; i++) {
81 if (uuid_is_nil(&xfs_uuid_table[i])) {
82 hole = i;
83 continue;
84 }
85 if (uuid_equal(uuid, &xfs_uuid_table[i]))
86 goto out_duplicate;
87 }
88
89 if (hole < 0) {
90 xfs_uuid_table = kmem_realloc(xfs_uuid_table,
91 (xfs_uuid_table_size + 1) * sizeof(*xfs_uuid_table),
Christoph Hellwig27174202009-03-30 10:21:31 +020092 KM_SLEEP);
93 hole = xfs_uuid_table_size++;
94 }
95 xfs_uuid_table[hole] = *uuid;
96 mutex_unlock(&xfs_uuid_table_mutex);
97
98 return 0;
99
100 out_duplicate:
101 mutex_unlock(&xfs_uuid_table_mutex);
Mitsuo Hayasaka021000e2012-01-13 05:58:39 +0000102 xfs_warn(mp, "Filesystem has duplicate UUID %pU - can't mount", uuid);
Dave Chinner24513372014-06-25 14:58:08 +1000103 return -EINVAL;
Christoph Hellwig27174202009-03-30 10:21:31 +0200104}
105
106STATIC void
107xfs_uuid_unmount(
108 struct xfs_mount *mp)
109{
110 uuid_t *uuid = &mp->m_sb.sb_uuid;
111 int i;
112
113 if (mp->m_flags & XFS_MOUNT_NOUUID)
114 return;
115
116 mutex_lock(&xfs_uuid_table_mutex);
117 for (i = 0; i < xfs_uuid_table_size; i++) {
118 if (uuid_is_nil(&xfs_uuid_table[i]))
119 continue;
120 if (!uuid_equal(uuid, &xfs_uuid_table[i]))
121 continue;
122 memset(&xfs_uuid_table[i], 0, sizeof(uuid_t));
123 break;
124 }
125 ASSERT(i < xfs_uuid_table_size);
126 mutex_unlock(&xfs_uuid_table_mutex);
127}
128
129
Dave Chinnere1765792010-09-22 10:47:20 +1000130STATIC void
131__xfs_free_perag(
132 struct rcu_head *head)
133{
134 struct xfs_perag *pag = container_of(head, struct xfs_perag, rcu_head);
135
136 ASSERT(atomic_read(&pag->pag_ref) == 0);
137 kmem_free(pag);
138}
139
Dave Chinner0fa800f2010-01-11 11:47:46 +0000140/*
Dave Chinnere1765792010-09-22 10:47:20 +1000141 * Free up the per-ag resources associated with the mount structure.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700142 */
Christoph Hellwigc962fb72008-05-20 15:10:52 +1000143STATIC void
Christoph Hellwigff4f0382008-08-13 16:50:47 +1000144xfs_free_perag(
Christoph Hellwig745f6912007-08-30 17:20:39 +1000145 xfs_mount_t *mp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146{
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000147 xfs_agnumber_t agno;
148 struct xfs_perag *pag;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000150 for (agno = 0; agno < mp->m_sb.sb_agcount; agno++) {
151 spin_lock(&mp->m_perag_lock);
152 pag = radix_tree_delete(&mp->m_perag_tree, agno);
153 spin_unlock(&mp->m_perag_lock);
Dave Chinnere1765792010-09-22 10:47:20 +1000154 ASSERT(pag);
Dave Chinnerf83282a2010-11-08 08:55:04 +0000155 ASSERT(atomic_read(&pag->pag_ref) == 0);
Dave Chinnere1765792010-09-22 10:47:20 +1000156 call_rcu(&pag->rcu_head, __xfs_free_perag);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158}
159
Nathan Scott4cc929e2007-05-14 18:24:02 +1000160/*
161 * Check size of device based on the (data/realtime) block count.
162 * Note: this check is used by the growfs code as well as mount.
163 */
164int
165xfs_sb_validate_fsb_count(
166 xfs_sb_t *sbp,
167 __uint64_t nblocks)
168{
169 ASSERT(PAGE_SHIFT >= sbp->sb_blocklog);
170 ASSERT(sbp->sb_blocklog >= BBSHIFT);
171
Christoph Hellwigd5cf09b2014-07-30 09:12:05 +1000172 /* Limited by ULONG_MAX of page cache index */
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300173 if (nblocks >> (PAGE_SHIFT - sbp->sb_blocklog) > ULONG_MAX)
Dave Chinner24513372014-06-25 14:58:08 +1000174 return -EFBIG;
Nathan Scott4cc929e2007-05-14 18:24:02 +1000175 return 0;
176}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000178int
Nathan Scottc11e2c32005-11-02 15:11:45 +1100179xfs_initialize_perag(
Nathan Scottc11e2c32005-11-02 15:11:45 +1100180 xfs_mount_t *mp,
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000181 xfs_agnumber_t agcount,
182 xfs_agnumber_t *maxagi)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700183{
Carlos Maiolino2d2194f2012-09-20 10:32:38 -0300184 xfs_agnumber_t index;
Dave Chinner8b26c582010-01-11 11:47:48 +0000185 xfs_agnumber_t first_initialised = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 xfs_perag_t *pag;
Dave Chinner8b26c582010-01-11 11:47:48 +0000187 int error = -ENOMEM;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000189 /*
190 * Walk the current per-ag tree so we don't try to initialise AGs
191 * that already exist (growfs case). Allocate and insert all the
192 * AGs we don't find ready for initialisation.
193 */
194 for (index = 0; index < agcount; index++) {
195 pag = xfs_perag_get(mp, index);
196 if (pag) {
197 xfs_perag_put(pag);
198 continue;
199 }
Dave Chinner8b26c582010-01-11 11:47:48 +0000200 if (!first_initialised)
201 first_initialised = index;
Christoph Hellwigfb3b5042010-05-28 19:03:10 +0000202
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000203 pag = kmem_zalloc(sizeof(*pag), KM_MAYFAIL);
204 if (!pag)
Dave Chinner8b26c582010-01-11 11:47:48 +0000205 goto out_unwind;
Christoph Hellwigfb3b5042010-05-28 19:03:10 +0000206 pag->pag_agno = index;
207 pag->pag_mount = mp;
Dave Chinner1a427ab2010-12-16 17:08:41 +1100208 spin_lock_init(&pag->pag_ici_lock);
Dave Chinner69b491c2010-09-27 11:09:51 +1000209 mutex_init(&pag->pag_ici_reclaim_lock);
Christoph Hellwigfb3b5042010-05-28 19:03:10 +0000210 INIT_RADIX_TREE(&pag->pag_ici_root, GFP_ATOMIC);
Dave Chinner74f75a02010-09-24 19:59:04 +1000211 spin_lock_init(&pag->pag_buf_lock);
212 pag->pag_buf_tree = RB_ROOT;
Christoph Hellwigfb3b5042010-05-28 19:03:10 +0000213
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000214 if (radix_tree_preload(GFP_NOFS))
Dave Chinner8b26c582010-01-11 11:47:48 +0000215 goto out_unwind;
Christoph Hellwigfb3b5042010-05-28 19:03:10 +0000216
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000217 spin_lock(&mp->m_perag_lock);
218 if (radix_tree_insert(&mp->m_perag_tree, index, pag)) {
219 BUG();
220 spin_unlock(&mp->m_perag_lock);
Dave Chinner8b26c582010-01-11 11:47:48 +0000221 radix_tree_preload_end();
222 error = -EEXIST;
223 goto out_unwind;
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000224 }
225 spin_unlock(&mp->m_perag_lock);
226 radix_tree_preload_end();
227 }
228
Eric Sandeen12c3f052016-03-02 09:58:09 +1100229 index = xfs_set_inode_alloc(mp, agcount);
Christoph Hellwigfb3b5042010-05-28 19:03:10 +0000230
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000231 if (maxagi)
232 *maxagi = index;
233 return 0;
Dave Chinner8b26c582010-01-11 11:47:48 +0000234
235out_unwind:
236 kmem_free(pag);
237 for (; index > first_initialised; index--) {
238 pag = radix_tree_delete(&mp->m_perag_tree, index);
239 kmem_free(pag);
240 }
241 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700242}
243
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244/*
245 * xfs_readsb
246 *
247 * Does the initial read of the superblock.
248 */
249int
Dave Chinnerff550682013-08-12 20:49:41 +1000250xfs_readsb(
251 struct xfs_mount *mp,
252 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253{
254 unsigned int sector_size;
Dave Chinner04a1e6c2013-04-03 16:11:31 +1100255 struct xfs_buf *bp;
256 struct xfs_sb *sbp = &mp->m_sb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 int error;
Dave Chinneraf34e092011-03-07 10:04:35 +1100258 int loud = !(flags & XFS_MFSI_QUIET);
Eric Sandeendaba5422014-02-19 15:39:16 +1100259 const struct xfs_buf_ops *buf_ops;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260
261 ASSERT(mp->m_sb_bp == NULL);
262 ASSERT(mp->m_ddev_targp != NULL);
263
264 /*
Eric Sandeendaba5422014-02-19 15:39:16 +1100265 * For the initial read, we must guess at the sector
266 * size based on the block device. It's enough to
267 * get the sb_sectsize out of the superblock and
268 * then reread with the proper length.
269 * We don't verify it yet, because it may not be complete.
270 */
271 sector_size = xfs_getsize_buftarg(mp->m_ddev_targp);
272 buf_ops = NULL;
273
274 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 * Allocate a (locked) buffer to hold the superblock.
276 * This will be kept around at all times to optimize
277 * access to the superblock.
278 */
Dave Chinner26af6552010-09-22 10:47:20 +1000279reread:
Dave Chinnerba372672014-10-02 09:05:32 +1000280 error = xfs_buf_read_uncached(mp->m_ddev_targp, XFS_SB_DADDR,
281 BTOBB(sector_size), 0, &bp, buf_ops);
282 if (error) {
Dave Chinnereab4e632012-11-12 22:54:02 +1100283 if (loud)
Dave Chinnere721f502013-04-03 16:11:32 +1100284 xfs_warn(mp, "SB validate failed with error %d.", error);
Dave Chinnerac75a1f2014-03-07 16:19:14 +1100285 /* bad CRC means corrupted metadata */
Dave Chinner24513372014-06-25 14:58:08 +1000286 if (error == -EFSBADCRC)
287 error = -EFSCORRUPTED;
Dave Chinnerba372672014-10-02 09:05:32 +1000288 return error;
Dave Chinnereab4e632012-11-12 22:54:02 +1100289 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290
291 /*
292 * Initialize the mount structure from the superblock.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 */
Dave Chinner556b8882014-06-06 16:00:43 +1000294 xfs_sb_from_disk(sbp, XFS_BUF_TO_SBP(bp));
Dave Chinner556b8882014-06-06 16:00:43 +1000295
296 /*
297 * If we haven't validated the superblock, do so now before we try
298 * to check the sector size and reread the superblock appropriately.
299 */
300 if (sbp->sb_magicnum != XFS_SB_MAGIC) {
301 if (loud)
302 xfs_warn(mp, "Invalid superblock magic number");
Dave Chinner24513372014-06-25 14:58:08 +1000303 error = -EINVAL;
Dave Chinner556b8882014-06-06 16:00:43 +1000304 goto release_buf;
305 }
Dave Chinnerff550682013-08-12 20:49:41 +1000306
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 /*
308 * We must be able to do sector-sized and sector-aligned IO.
309 */
Dave Chinner04a1e6c2013-04-03 16:11:31 +1100310 if (sector_size > sbp->sb_sectsize) {
Dave Chinneraf34e092011-03-07 10:04:35 +1100311 if (loud)
312 xfs_warn(mp, "device supports %u byte sectors (not %u)",
Dave Chinner04a1e6c2013-04-03 16:11:31 +1100313 sector_size, sbp->sb_sectsize);
Dave Chinner24513372014-06-25 14:58:08 +1000314 error = -ENOSYS;
Dave Chinner26af6552010-09-22 10:47:20 +1000315 goto release_buf;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700316 }
317
Eric Sandeendaba5422014-02-19 15:39:16 +1100318 if (buf_ops == NULL) {
Dave Chinner556b8882014-06-06 16:00:43 +1000319 /*
320 * Re-read the superblock so the buffer is correctly sized,
321 * and properly verified.
322 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 xfs_buf_relse(bp);
Dave Chinner04a1e6c2013-04-03 16:11:31 +1100324 sector_size = sbp->sb_sectsize;
Eric Sandeendaba5422014-02-19 15:39:16 +1100325 buf_ops = loud ? &xfs_sb_buf_ops : &xfs_sb_quiet_buf_ops;
Dave Chinner26af6552010-09-22 10:47:20 +1000326 goto reread;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327 }
328
Dave Chinner5681ca42015-02-23 21:22:31 +1100329 xfs_reinit_percpu_counters(mp);
David Chinner8d280b92006-03-14 13:13:09 +1100330
Dave Chinner04a1e6c2013-04-03 16:11:31 +1100331 /* no need to be quiet anymore, so reset the buf ops */
332 bp->b_ops = &xfs_sb_buf_ops;
333
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 mp->m_sb_bp = bp;
Dave Chinner26af6552010-09-22 10:47:20 +1000335 xfs_buf_unlock(bp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 return 0;
337
Dave Chinner26af6552010-09-22 10:47:20 +1000338release_buf:
339 xfs_buf_relse(bp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 return error;
341}
342
Linus Torvalds1da177e2005-04-16 15:20:36 -0700343/*
Eric Sandeen0771fb42007-10-12 11:03:40 +1000344 * Update alignment values based on mount options and sb values
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 */
Eric Sandeen0771fb42007-10-12 11:03:40 +1000346STATIC int
Christoph Hellwig7884bc82009-01-19 02:04:07 +0100347xfs_update_alignment(xfs_mount_t *mp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348{
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 xfs_sb_t *sbp = &(mp->m_sb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350
Christoph Hellwig42490232008-08-13 16:49:32 +1000351 if (mp->m_dalign) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 /*
353 * If stripe unit and stripe width are not multiples
354 * of the fs blocksize turn off alignment.
355 */
356 if ((BBTOB(mp->m_dalign) & mp->m_blockmask) ||
357 (BBTOB(mp->m_swidth) & mp->m_blockmask)) {
Jie Liu39a45d82013-05-02 19:27:47 +0800358 xfs_warn(mp,
359 "alignment check failed: sunit/swidth vs. blocksize(%d)",
360 sbp->sb_blocksize);
Dave Chinner24513372014-06-25 14:58:08 +1000361 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 } else {
363 /*
364 * Convert the stripe unit and width to FSBs.
365 */
366 mp->m_dalign = XFS_BB_TO_FSBT(mp, mp->m_dalign);
367 if (mp->m_dalign && (sbp->sb_agblocks % mp->m_dalign)) {
Dave Chinner53487782011-03-07 10:05:35 +1100368 xfs_warn(mp,
Jie Liu39a45d82013-05-02 19:27:47 +0800369 "alignment check failed: sunit/swidth vs. agsize(%d)",
370 sbp->sb_agblocks);
Dave Chinner24513372014-06-25 14:58:08 +1000371 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372 } else if (mp->m_dalign) {
373 mp->m_swidth = XFS_BB_TO_FSBT(mp, mp->m_swidth);
374 } else {
Jie Liu39a45d82013-05-02 19:27:47 +0800375 xfs_warn(mp,
376 "alignment check failed: sunit(%d) less than bsize(%d)",
377 mp->m_dalign, sbp->sb_blocksize);
Dave Chinner24513372014-06-25 14:58:08 +1000378 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700379 }
380 }
381
382 /*
383 * Update superblock with new values
384 * and log changes
385 */
Eric Sandeen62118702008-03-06 13:44:28 +1100386 if (xfs_sb_version_hasdalign(sbp)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700387 if (sbp->sb_unit != mp->m_dalign) {
388 sbp->sb_unit = mp->m_dalign;
Dave Chinner61e63ec2015-01-22 09:10:31 +1100389 mp->m_update_sb = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 }
391 if (sbp->sb_width != mp->m_swidth) {
392 sbp->sb_width = mp->m_swidth;
Dave Chinner61e63ec2015-01-22 09:10:31 +1100393 mp->m_update_sb = true;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700394 }
Jie Liu34d7f602013-05-02 19:27:53 +0800395 } else {
396 xfs_warn(mp,
397 "cannot change alignment: superblock does not support data alignment");
Dave Chinner24513372014-06-25 14:58:08 +1000398 return -EINVAL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 }
400 } else if ((mp->m_flags & XFS_MOUNT_NOALIGN) != XFS_MOUNT_NOALIGN &&
Eric Sandeen62118702008-03-06 13:44:28 +1100401 xfs_sb_version_hasdalign(&mp->m_sb)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 mp->m_dalign = sbp->sb_unit;
403 mp->m_swidth = sbp->sb_width;
404 }
405
Eric Sandeen0771fb42007-10-12 11:03:40 +1000406 return 0;
407}
408
409/*
410 * Set the maximum inode count for this filesystem
411 */
412STATIC void
413xfs_set_maxicount(xfs_mount_t *mp)
414{
415 xfs_sb_t *sbp = &(mp->m_sb);
416 __uint64_t icount;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417
418 if (sbp->sb_imax_pct) {
Eric Sandeen0771fb42007-10-12 11:03:40 +1000419 /*
420 * Make sure the maximum inode count is a multiple
421 * of the units we allocate inodes in.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
424 do_div(icount, 100);
425 do_div(icount, mp->m_ialloc_blks);
426 mp->m_maxicount = (icount * mp->m_ialloc_blks) <<
427 sbp->sb_inopblog;
Eric Sandeen0771fb42007-10-12 11:03:40 +1000428 } else {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700429 mp->m_maxicount = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 }
Eric Sandeen0771fb42007-10-12 11:03:40 +1000431}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432
Eric Sandeen0771fb42007-10-12 11:03:40 +1000433/*
434 * Set the default minimum read and write sizes unless
435 * already specified in a mount option.
436 * We use smaller I/O sizes when the file system
437 * is being used for NFS service (wsync mount option).
438 */
439STATIC void
440xfs_set_rw_sizes(xfs_mount_t *mp)
441{
442 xfs_sb_t *sbp = &(mp->m_sb);
443 int readio_log, writeio_log;
444
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 if (!(mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)) {
446 if (mp->m_flags & XFS_MOUNT_WSYNC) {
447 readio_log = XFS_WSYNC_READIO_LOG;
448 writeio_log = XFS_WSYNC_WRITEIO_LOG;
449 } else {
450 readio_log = XFS_READIO_LOG_LARGE;
451 writeio_log = XFS_WRITEIO_LOG_LARGE;
452 }
453 } else {
454 readio_log = mp->m_readio_log;
455 writeio_log = mp->m_writeio_log;
456 }
457
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458 if (sbp->sb_blocklog > readio_log) {
459 mp->m_readio_log = sbp->sb_blocklog;
460 } else {
461 mp->m_readio_log = readio_log;
462 }
463 mp->m_readio_blocks = 1 << (mp->m_readio_log - sbp->sb_blocklog);
464 if (sbp->sb_blocklog > writeio_log) {
465 mp->m_writeio_log = sbp->sb_blocklog;
466 } else {
467 mp->m_writeio_log = writeio_log;
468 }
469 mp->m_writeio_blocks = 1 << (mp->m_writeio_log - sbp->sb_blocklog);
Eric Sandeen0771fb42007-10-12 11:03:40 +1000470}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471
Eric Sandeen0771fb42007-10-12 11:03:40 +1000472/*
Dave Chinner055388a2011-01-04 11:35:03 +1100473 * precalculate the low space thresholds for dynamic speculative preallocation.
474 */
475void
476xfs_set_low_space_thresholds(
477 struct xfs_mount *mp)
478{
479 int i;
480
481 for (i = 0; i < XFS_LOWSP_MAX; i++) {
482 __uint64_t space = mp->m_sb.sb_dblocks;
483
484 do_div(space, 100);
485 mp->m_low_space[i] = space * (i + 1);
486 }
487}
488
489
490/*
Eric Sandeen0771fb42007-10-12 11:03:40 +1000491 * Set whether we're using inode alignment.
492 */
493STATIC void
494xfs_set_inoalignment(xfs_mount_t *mp)
495{
Eric Sandeen62118702008-03-06 13:44:28 +1100496 if (xfs_sb_version_hasalign(&mp->m_sb) &&
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 mp->m_sb.sb_inoalignmt >=
498 XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size))
499 mp->m_inoalign_mask = mp->m_sb.sb_inoalignmt - 1;
500 else
501 mp->m_inoalign_mask = 0;
502 /*
503 * If we are using stripe alignment, check whether
504 * the stripe unit is a multiple of the inode alignment
505 */
506 if (mp->m_dalign && mp->m_inoalign_mask &&
507 !(mp->m_dalign & mp->m_inoalign_mask))
508 mp->m_sinoalign = mp->m_dalign;
509 else
510 mp->m_sinoalign = 0;
Eric Sandeen0771fb42007-10-12 11:03:40 +1000511}
512
513/*
Zhi Yong Wu0471f622013-08-07 10:10:58 +0000514 * Check that the data (and log if separate) is an ok size.
Eric Sandeen0771fb42007-10-12 11:03:40 +1000515 */
516STATIC int
Dave Chinnerba372672014-10-02 09:05:32 +1000517xfs_check_sizes(
518 struct xfs_mount *mp)
Eric Sandeen0771fb42007-10-12 11:03:40 +1000519{
Dave Chinnerba372672014-10-02 09:05:32 +1000520 struct xfs_buf *bp;
Eric Sandeen0771fb42007-10-12 11:03:40 +1000521 xfs_daddr_t d;
Dave Chinnerba372672014-10-02 09:05:32 +1000522 int error;
Eric Sandeen0771fb42007-10-12 11:03:40 +1000523
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524 d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
525 if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_dblocks) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100526 xfs_warn(mp, "filesystem size mismatch detected");
Dave Chinner24513372014-06-25 14:58:08 +1000527 return -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 }
Dave Chinnerba372672014-10-02 09:05:32 +1000529 error = xfs_buf_read_uncached(mp->m_ddev_targp,
Dave Chinner1922c942010-09-22 10:47:20 +1000530 d - XFS_FSS_TO_BB(mp, 1),
Dave Chinnerba372672014-10-02 09:05:32 +1000531 XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
532 if (error) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100533 xfs_warn(mp, "last sector read failed");
Dave Chinnerba372672014-10-02 09:05:32 +1000534 return error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 }
Dave Chinner1922c942010-09-22 10:47:20 +1000536 xfs_buf_relse(bp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
Dave Chinnerba372672014-10-02 09:05:32 +1000538 if (mp->m_logdev_targp == mp->m_ddev_targp)
539 return 0;
540
541 d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_logblocks);
542 if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_logblocks) {
543 xfs_warn(mp, "log size mismatch detected");
544 return -EFBIG;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 }
Dave Chinnerba372672014-10-02 09:05:32 +1000546 error = xfs_buf_read_uncached(mp->m_logdev_targp,
547 d - XFS_FSB_TO_BB(mp, 1),
548 XFS_FSB_TO_BB(mp, 1), 0, &bp, NULL);
549 if (error) {
550 xfs_warn(mp, "log device read failed");
551 return error;
552 }
553 xfs_buf_relse(bp);
Eric Sandeen0771fb42007-10-12 11:03:40 +1000554 return 0;
555}
556
557/*
Christoph Hellwig7d095252009-06-08 15:33:32 +0200558 * Clear the quotaflags in memory and in the superblock.
559 */
560int
561xfs_mount_reset_sbqflags(
562 struct xfs_mount *mp)
563{
Christoph Hellwig7d095252009-06-08 15:33:32 +0200564 mp->m_qflags = 0;
565
Dave Chinner61e63ec2015-01-22 09:10:31 +1100566 /* It is OK to look at sb_qflags in the mount path without m_sb_lock. */
Christoph Hellwig7d095252009-06-08 15:33:32 +0200567 if (mp->m_sb.sb_qflags == 0)
568 return 0;
569 spin_lock(&mp->m_sb_lock);
570 mp->m_sb.sb_qflags = 0;
571 spin_unlock(&mp->m_sb_lock);
572
Dave Chinner61e63ec2015-01-22 09:10:31 +1100573 if (!xfs_fs_writable(mp, SB_FREEZE_WRITE))
Christoph Hellwig7d095252009-06-08 15:33:32 +0200574 return 0;
575
Dave Chinner61e63ec2015-01-22 09:10:31 +1100576 return xfs_sync_sb(mp, false);
Christoph Hellwig7d095252009-06-08 15:33:32 +0200577}
578
Eric Sandeend5db0f92010-02-05 22:59:53 +0000579__uint64_t
580xfs_default_resblks(xfs_mount_t *mp)
581{
582 __uint64_t resblks;
583
584 /*
Dave Chinner8babd8a2010-03-04 01:46:25 +0000585 * We default to 5% or 8192 fsbs of space reserved, whichever is
586 * smaller. This is intended to cover concurrent allocation
587 * transactions when we initially hit enospc. These each require a 4
588 * block reservation. Hence by default we cover roughly 2000 concurrent
589 * allocation reservations.
Eric Sandeend5db0f92010-02-05 22:59:53 +0000590 */
591 resblks = mp->m_sb.sb_dblocks;
592 do_div(resblks, 20);
Dave Chinner8babd8a2010-03-04 01:46:25 +0000593 resblks = min_t(__uint64_t, resblks, 8192);
Eric Sandeend5db0f92010-02-05 22:59:53 +0000594 return resblks;
595}
596
Christoph Hellwig7d095252009-06-08 15:33:32 +0200597/*
Eric Sandeen0771fb42007-10-12 11:03:40 +1000598 * This function does the following on an initial mount of a file system:
599 * - reads the superblock from disk and init the mount struct
600 * - if we're a 32-bit kernel, do a size check on the superblock
601 * so we don't mount terabyte filesystems
602 * - init mount struct realtime fields
603 * - allocate inode hash table for fs
604 * - init directory manager
605 * - perform recovery and init the log manager
606 */
607int
608xfs_mountfs(
Brian Fosterf0b2efa2015-08-19 09:58:36 +1000609 struct xfs_mount *mp)
Eric Sandeen0771fb42007-10-12 11:03:40 +1000610{
Brian Fosterf0b2efa2015-08-19 09:58:36 +1000611 struct xfs_sb *sbp = &(mp->m_sb);
612 struct xfs_inode *rip;
613 __uint64_t resblks;
614 uint quotamount = 0;
615 uint quotaflags = 0;
616 int error = 0;
Eric Sandeen0771fb42007-10-12 11:03:40 +1000617
Dave Chinnerff550682013-08-12 20:49:41 +1000618 xfs_sb_mount_common(mp, sbp);
Eric Sandeen0771fb42007-10-12 11:03:40 +1000619
620 /*
Dave Chinner074e4272015-01-22 09:10:33 +1100621 * Check for a mismatched features2 values. Older kernels read & wrote
622 * into the wrong sb offset for sb_features2 on some platforms due to
623 * xfs_sb_t not being 64bit size aligned when sb_features2 was added,
624 * which made older superblock reading/writing routines swap it as a
625 * 64-bit value.
David Chinneree1c0902008-03-06 13:45:50 +1100626 *
Eric Sandeene6957ea2008-04-10 12:19:34 +1000627 * For backwards compatibility, we make both slots equal.
628 *
Dave Chinner074e4272015-01-22 09:10:33 +1100629 * If we detect a mismatched field, we OR the set bits into the existing
630 * features2 field in case it has already been modified; we don't want
631 * to lose any features. We then update the bad location with the ORed
632 * value so that older kernels will see any features2 flags. The
633 * superblock writeback code ensures the new sb_features2 is copied to
634 * sb_bad_features2 before it is logged or written to disk.
David Chinneree1c0902008-03-06 13:45:50 +1100635 */
Eric Sandeene6957ea2008-04-10 12:19:34 +1000636 if (xfs_sb_has_mismatched_features2(sbp)) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100637 xfs_warn(mp, "correcting sb_features alignment problem");
David Chinneree1c0902008-03-06 13:45:50 +1100638 sbp->sb_features2 |= sbp->sb_bad_features2;
Dave Chinner61e63ec2015-01-22 09:10:31 +1100639 mp->m_update_sb = true;
Eric Sandeene6957ea2008-04-10 12:19:34 +1000640
641 /*
642 * Re-check for ATTR2 in case it was found in bad_features2
643 * slot.
644 */
Tim Shimmin7c12f292008-04-30 18:15:28 +1000645 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
646 !(mp->m_flags & XFS_MOUNT_NOATTR2))
Eric Sandeene6957ea2008-04-10 12:19:34 +1000647 mp->m_flags |= XFS_MOUNT_ATTR2;
Tim Shimmin7c12f292008-04-30 18:15:28 +1000648 }
Eric Sandeene6957ea2008-04-10 12:19:34 +1000649
Tim Shimmin7c12f292008-04-30 18:15:28 +1000650 if (xfs_sb_version_hasattr2(&mp->m_sb) &&
651 (mp->m_flags & XFS_MOUNT_NOATTR2)) {
652 xfs_sb_version_removeattr2(&mp->m_sb);
Dave Chinner61e63ec2015-01-22 09:10:31 +1100653 mp->m_update_sb = true;
Tim Shimmin7c12f292008-04-30 18:15:28 +1000654
655 /* update sb_versionnum for the clearing of the morebits */
656 if (!sbp->sb_features2)
Dave Chinner61e63ec2015-01-22 09:10:31 +1100657 mp->m_update_sb = true;
David Chinneree1c0902008-03-06 13:45:50 +1100658 }
659
Dave Chinner263997a2014-05-20 07:46:40 +1000660 /* always use v2 inodes by default now */
661 if (!(mp->m_sb.sb_versionnum & XFS_SB_VERSION_NLINKBIT)) {
662 mp->m_sb.sb_versionnum |= XFS_SB_VERSION_NLINKBIT;
Dave Chinner61e63ec2015-01-22 09:10:31 +1100663 mp->m_update_sb = true;
Dave Chinner263997a2014-05-20 07:46:40 +1000664 }
665
David Chinneree1c0902008-03-06 13:45:50 +1100666 /*
Eric Sandeen0771fb42007-10-12 11:03:40 +1000667 * Check if sb_agblocks is aligned at stripe boundary
668 * If sb_agblocks is NOT aligned turn off m_dalign since
669 * allocator alignment is within an ag, therefore ag has
670 * to be aligned at stripe boundary.
671 */
Christoph Hellwig7884bc82009-01-19 02:04:07 +0100672 error = xfs_update_alignment(mp);
Eric Sandeen0771fb42007-10-12 11:03:40 +1000673 if (error)
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100674 goto out;
Eric Sandeen0771fb42007-10-12 11:03:40 +1000675
676 xfs_alloc_compute_maxlevels(mp);
677 xfs_bmap_compute_maxlevels(mp, XFS_DATA_FORK);
678 xfs_bmap_compute_maxlevels(mp, XFS_ATTR_FORK);
679 xfs_ialloc_compute_maxlevels(mp);
680
681 xfs_set_maxicount(mp);
682
Carlos Maiolinoe6b3bb72016-05-18 11:11:27 +1000683 /* enable fail_at_unmount as default */
684 mp->m_fail_unmount = 1;
685
Brian Fostera31b1d32014-07-15 08:07:01 +1000686 error = xfs_sysfs_init(&mp->m_kobj, &xfs_mp_ktype, NULL, mp->m_fsname);
Christoph Hellwig27174202009-03-30 10:21:31 +0200687 if (error)
688 goto out;
Eric Sandeen0771fb42007-10-12 11:03:40 +1000689
Bill O'Donnell225e4632015-10-12 18:21:19 +1100690 error = xfs_sysfs_init(&mp->m_stats.xs_kobj, &xfs_stats_ktype,
691 &mp->m_kobj, "stats");
Brian Fostera31b1d32014-07-15 08:07:01 +1000692 if (error)
693 goto out_remove_sysfs;
694
Carlos Maiolino192852b2016-05-18 10:58:51 +1000695 error = xfs_error_sysfs_init(mp);
Bill O'Donnell225e4632015-10-12 18:21:19 +1100696 if (error)
697 goto out_del_stats;
698
Carlos Maiolino192852b2016-05-18 10:58:51 +1000699
700 error = xfs_uuid_mount(mp);
701 if (error)
702 goto out_remove_error_sysfs;
703
Eric Sandeen0771fb42007-10-12 11:03:40 +1000704 /*
705 * Set the minimum read and write sizes
706 */
707 xfs_set_rw_sizes(mp);
708
Dave Chinner055388a2011-01-04 11:35:03 +1100709 /* set the low space thresholds for dynamic preallocation */
710 xfs_set_low_space_thresholds(mp);
711
Eric Sandeen0771fb42007-10-12 11:03:40 +1000712 /*
713 * Set the inode cluster size.
714 * This may still be overridden by the file system
715 * block size if it is larger than the chosen cluster size.
Dave Chinner8f805872013-11-01 15:27:20 +1100716 *
717 * For v5 filesystems, scale the cluster size with the inode size to
718 * keep a constant ratio of inode per cluster buffer, but only if mkfs
719 * has set the inode alignment value appropriately for larger cluster
720 * sizes.
Eric Sandeen0771fb42007-10-12 11:03:40 +1000721 */
722 mp->m_inode_cluster_size = XFS_INODE_BIG_CLUSTER_SIZE;
Dave Chinner8f805872013-11-01 15:27:20 +1100723 if (xfs_sb_version_hascrc(&mp->m_sb)) {
724 int new_size = mp->m_inode_cluster_size;
725
726 new_size *= mp->m_sb.sb_inodesize / XFS_DINODE_MIN_SIZE;
727 if (mp->m_sb.sb_inoalignmt >= XFS_B_TO_FSBT(mp, new_size))
728 mp->m_inode_cluster_size = new_size;
Dave Chinner8f805872013-11-01 15:27:20 +1100729 }
Eric Sandeen0771fb42007-10-12 11:03:40 +1000730
731 /*
Brian Fostere5376fc2015-05-29 08:57:27 +1000732 * If enabled, sparse inode chunk alignment is expected to match the
733 * cluster size. Full inode chunk alignment must match the chunk size,
734 * but that is checked on sb read verification...
735 */
736 if (xfs_sb_version_hassparseinodes(&mp->m_sb) &&
737 mp->m_sb.sb_spino_align !=
738 XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size)) {
739 xfs_warn(mp,
740 "Sparse inode block alignment (%u) must match cluster size (%llu).",
741 mp->m_sb.sb_spino_align,
742 XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size));
743 error = -EINVAL;
744 goto out_remove_uuid;
745 }
746
747 /*
Eric Sandeen0771fb42007-10-12 11:03:40 +1000748 * Set inode alignment fields
749 */
750 xfs_set_inoalignment(mp);
751
752 /*
Zhi Yong Wuc2bfbc92013-08-12 03:15:03 +0000753 * Check that the data (and log if separate) is an ok size.
Eric Sandeen0771fb42007-10-12 11:03:40 +1000754 */
Christoph Hellwig42490232008-08-13 16:49:32 +1000755 error = xfs_check_sizes(mp);
Eric Sandeen0771fb42007-10-12 11:03:40 +1000756 if (error)
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100757 goto out_remove_uuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758
759 /*
760 * Initialize realtime fields in the mount structure
761 */
Eric Sandeen0771fb42007-10-12 11:03:40 +1000762 error = xfs_rtmount_init(mp);
763 if (error) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100764 xfs_warn(mp, "RT mount failed");
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100765 goto out_remove_uuid;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766 }
767
768 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 * Copies the low order bits of the timestamp and the randomly
770 * set "sequence" number out of a UUID.
771 */
772 uuid_getnodeuniq(&sbp->sb_uuid, mp->m_fixedfsid);
773
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 mp->m_dmevmask = 0; /* not persistent; set after each mount */
775
Dave Chinner0650b552014-06-06 15:01:58 +1000776 error = xfs_da_mount(mp);
777 if (error) {
778 xfs_warn(mp, "Failed dir/attr init: %d", error);
779 goto out_remove_uuid;
780 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700781
782 /*
783 * Initialize the precomputed transaction reservations values.
784 */
785 xfs_trans_init(mp);
786
787 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700788 * Allocate and initialize the per-ag data.
789 */
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000790 spin_lock_init(&mp->m_perag_lock);
Dave Chinner9b98b6f2010-05-27 01:58:13 +0000791 INIT_RADIX_TREE(&mp->m_perag_tree, GFP_ATOMIC);
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000792 error = xfs_initialize_perag(mp, sbp->sb_agcount, &mp->m_maxagi);
793 if (error) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100794 xfs_warn(mp, "Failed per-ag init: %d", error);
Dave Chinner0650b552014-06-06 15:01:58 +1000795 goto out_free_dir;
Dave Chinner1c1c6eb2010-01-11 11:47:44 +0000796 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100798 if (!sbp->sb_logblocks) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100799 xfs_warn(mp, "no log defined");
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100800 XFS_ERROR_REPORT("xfs_mountfs", XFS_ERRLEVEL_LOW, mp);
Dave Chinner24513372014-06-25 14:58:08 +1000801 error = -EFSCORRUPTED;
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100802 goto out_free_perag;
803 }
804
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805 /*
Brian Fosterf0b2efa2015-08-19 09:58:36 +1000806 * Log's mount-time initialization. The first part of recovery can place
807 * some items on the AIL, to be handled when recovery is finished or
808 * cancelled.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809 */
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100810 error = xfs_log_mount(mp, mp->m_logdev_targp,
811 XFS_FSB_TO_DADDR(mp, sbp->sb_logstart),
812 XFS_FSB_TO_BB(mp, sbp->sb_logblocks));
813 if (error) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100814 xfs_warn(mp, "log mount failed");
Dave Chinnerd4f35122012-04-23 15:59:06 +1000815 goto out_fail_wait;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816 }
817
818 /*
David Chinner92821e22007-05-24 15:26:31 +1000819 * Now the log is mounted, we know if it was an unclean shutdown or
820 * not. If it was, with the first phase of recovery has completed, we
821 * have consistent AG blocks on disk. We have not recovered EFIs yet,
822 * but they are recovered transactionally in the second recovery phase
823 * later.
824 *
825 * Hence we can safely re-initialise incore superblock counters from
826 * the per-ag data. These may not be correct if the filesystem was not
827 * cleanly unmounted, so we need to wait for recovery to finish before
828 * doing this.
829 *
830 * If the filesystem was cleanly unmounted, then we can trust the
831 * values in the superblock to be correct and we don't need to do
832 * anything here.
833 *
834 * If we are currently making the filesystem, the initialisation will
835 * fail as the perag data is in an undefined state.
836 */
David Chinner92821e22007-05-24 15:26:31 +1000837 if (xfs_sb_version_haslazysbcount(&mp->m_sb) &&
838 !XFS_LAST_UNMOUNT_WAS_CLEAN(mp) &&
839 !mp->m_sb.sb_inprogress) {
840 error = xfs_initialize_perag_data(mp, sbp->sb_agcount);
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100841 if (error)
kbuild test robot6eee8972014-08-04 13:49:40 +1000842 goto out_log_dealloc;
David Chinner92821e22007-05-24 15:26:31 +1000843 }
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100844
David Chinner92821e22007-05-24 15:26:31 +1000845 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846 * Get and sanity-check the root inode.
847 * Save the pointer to it in the mount structure.
848 */
Dave Chinner7b6259e2010-06-24 11:35:17 +1000849 error = xfs_iget(mp, NULL, sbp->sb_rootino, 0, XFS_ILOCK_EXCL, &rip);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850 if (error) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100851 xfs_warn(mp, "failed to read root inode");
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100852 goto out_log_dealloc;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700853 }
854
855 ASSERT(rip != NULL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856
Dave Chinnerc19b3b052016-02-09 16:54:58 +1100857 if (unlikely(!S_ISDIR(VFS_I(rip)->i_mode))) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100858 xfs_warn(mp, "corrupted root inode %llu: not a directory",
Nathan Scottb6574522006-06-09 15:29:40 +1000859 (unsigned long long)rip->i_ino);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860 xfs_iunlock(rip, XFS_ILOCK_EXCL);
861 XFS_ERROR_REPORT("xfs_mountfs_int(2)", XFS_ERRLEVEL_LOW,
862 mp);
Dave Chinner24513372014-06-25 14:58:08 +1000863 error = -EFSCORRUPTED;
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100864 goto out_rele_rip;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865 }
866 mp->m_rootip = rip; /* save it */
867
868 xfs_iunlock(rip, XFS_ILOCK_EXCL);
869
870 /*
871 * Initialize realtime inode pointers in the mount structure
872 */
Eric Sandeen0771fb42007-10-12 11:03:40 +1000873 error = xfs_rtmount_inodes(mp);
874 if (error) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700875 /*
876 * Free up the root inode.
877 */
Dave Chinner0b932cc2011-03-07 10:08:35 +1100878 xfs_warn(mp, "failed to read RT inodes");
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100879 goto out_rele_rip;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 }
881
882 /*
Christoph Hellwig7884bc82009-01-19 02:04:07 +0100883 * If this is a read-only mount defer the superblock updates until
884 * the next remount into writeable mode. Otherwise we would never
885 * perform the update e.g. for the root filesystem.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 */
Dave Chinner61e63ec2015-01-22 09:10:31 +1100887 if (mp->m_update_sb && !(mp->m_flags & XFS_MOUNT_RDONLY)) {
888 error = xfs_sync_sb(mp, false);
David Chinnere5720ee2008-04-10 12:21:18 +1000889 if (error) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100890 xfs_warn(mp, "failed to write sb changes");
Christoph Hellwigb93b6e42009-02-04 09:33:58 +0100891 goto out_rtunmount;
David Chinnere5720ee2008-04-10 12:21:18 +1000892 }
893 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894
895 /*
896 * Initialise the XFS quota management subsystem for this mount
897 */
Christoph Hellwig7d095252009-06-08 15:33:32 +0200898 if (XFS_IS_QUOTA_RUNNING(mp)) {
899 error = xfs_qm_newmount(mp, &quotamount, &quotaflags);
900 if (error)
901 goto out_rtunmount;
902 } else {
903 ASSERT(!XFS_IS_QUOTA_ON(mp));
904
905 /*
906 * If a file system had quotas running earlier, but decided to
907 * mount without -o uquota/pquota/gquota options, revoke the
908 * quotachecked license.
909 */
910 if (mp->m_sb.sb_qflags & XFS_ALL_QUOTA_ACCT) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100911 xfs_notice(mp, "resetting quota flags");
Christoph Hellwig7d095252009-06-08 15:33:32 +0200912 error = xfs_mount_reset_sbqflags(mp);
913 if (error)
Brian Fostera70a4fa2014-07-15 07:41:25 +1000914 goto out_rtunmount;
Christoph Hellwig7d095252009-06-08 15:33:32 +0200915 }
916 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700917
918 /*
Brian Fosterf0b2efa2015-08-19 09:58:36 +1000919 * Finish recovering the file system. This part needed to be delayed
920 * until after the root and real-time bitmap inodes were consistently
921 * read in.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 */
Christoph Hellwig42490232008-08-13 16:49:32 +1000923 error = xfs_log_mount_finish(mp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 if (error) {
Dave Chinner0b932cc2011-03-07 10:08:35 +1100925 xfs_warn(mp, "log mount finish failed");
Christoph Hellwigb93b6e42009-02-04 09:33:58 +0100926 goto out_rtunmount;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927 }
928
929 /*
930 * Complete the quota initialisation, post-log-replay component.
931 */
Christoph Hellwig7d095252009-06-08 15:33:32 +0200932 if (quotamount) {
933 ASSERT(mp->m_qflags == 0);
934 mp->m_qflags = quotaflags;
935
936 xfs_qm_mount_quotas(mp);
937 }
938
David Chinner84e1e992007-06-18 16:50:27 +1000939 /*
940 * Now we are mounted, reserve a small amount of unused space for
941 * privileged transactions. This is needed so that transaction
942 * space required for critical operations can dip into this pool
943 * when at ENOSPC. This is needed for operations like create with
944 * attr, unwritten extent conversion at ENOSPC, etc. Data allocations
945 * are not allowed to use this reserved space.
Dave Chinner8babd8a2010-03-04 01:46:25 +0000946 *
947 * This may drive us straight to ENOSPC on mount, but that implies
948 * we were already there on the last unmount. Warn if this occurs.
David Chinner84e1e992007-06-18 16:50:27 +1000949 */
Eric Sandeend5db0f92010-02-05 22:59:53 +0000950 if (!(mp->m_flags & XFS_MOUNT_RDONLY)) {
951 resblks = xfs_default_resblks(mp);
952 error = xfs_reserve_blocks(mp, &resblks, NULL);
953 if (error)
Dave Chinner0b932cc2011-03-07 10:08:35 +1100954 xfs_warn(mp,
955 "Unable to allocate reserve blocks. Continuing without reserve pool.");
Eric Sandeend5db0f92010-02-05 22:59:53 +0000956 }
David Chinner84e1e992007-06-18 16:50:27 +1000957
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 return 0;
959
Christoph Hellwigb93b6e42009-02-04 09:33:58 +0100960 out_rtunmount:
961 xfs_rtunmount_inodes(mp);
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100962 out_rele_rip:
Christoph Hellwig43355092008-03-27 18:01:08 +1100963 IRELE(rip);
Brian Foster0ae120f2015-08-19 10:00:28 +1000964 cancel_delayed_work_sync(&mp->m_reclaim_work);
965 xfs_reclaim_inodes(mp, SYNC_WAIT);
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100966 out_log_dealloc:
Carlos Maiolinoe6b3bb72016-05-18 11:11:27 +1000967 mp->m_flags |= XFS_MOUNT_UNMOUNTING;
Brian Fosterf0b2efa2015-08-19 09:58:36 +1000968 xfs_log_mount_cancel(mp);
Dave Chinnerd4f35122012-04-23 15:59:06 +1000969 out_fail_wait:
970 if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp)
971 xfs_wait_buftarg(mp->m_logdev_targp);
972 xfs_wait_buftarg(mp->m_ddev_targp);
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100973 out_free_perag:
Christoph Hellwigff4f0382008-08-13 16:50:47 +1000974 xfs_free_perag(mp);
Dave Chinner0650b552014-06-06 15:01:58 +1000975 out_free_dir:
976 xfs_da_unmount(mp);
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100977 out_remove_uuid:
Christoph Hellwig27174202009-03-30 10:21:31 +0200978 xfs_uuid_unmount(mp);
Carlos Maiolino192852b2016-05-18 10:58:51 +1000979 out_remove_error_sysfs:
980 xfs_error_sysfs_del(mp);
Bill O'Donnell225e4632015-10-12 18:21:19 +1100981 out_del_stats:
982 xfs_sysfs_del(&mp->m_stats.xs_kobj);
Brian Fostera31b1d32014-07-15 08:07:01 +1000983 out_remove_sysfs:
984 xfs_sysfs_del(&mp->m_kobj);
Christoph Hellwigf9057e32009-02-04 09:31:52 +0100985 out:
Linus Torvalds1da177e2005-04-16 15:20:36 -0700986 return error;
987}
988
989/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 * This flushes out the inodes,dquots and the superblock, unmounts the
991 * log and makes sure that incore structures are freed.
992 */
Christoph Hellwig41b5c2e2008-08-13 16:49:57 +1000993void
994xfs_unmountfs(
995 struct xfs_mount *mp)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996{
Christoph Hellwig41b5c2e2008-08-13 16:49:57 +1000997 __uint64_t resblks;
998 int error;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
Brian Foster579b62f2012-11-06 09:50:47 -05001000 cancel_delayed_work_sync(&mp->m_eofblocks_work);
1001
Christoph Hellwig7d095252009-06-08 15:33:32 +02001002 xfs_qm_unmount_quotas(mp);
Christoph Hellwigb93b6e42009-02-04 09:33:58 +01001003 xfs_rtunmount_inodes(mp);
Christoph Hellwig77508ec2008-08-13 16:49:04 +10001004 IRELE(mp->m_rootip);
1005
David Chinner641c56f2007-06-18 16:50:17 +10001006 /*
1007 * We can potentially deadlock here if we have an inode cluster
Malcolm Parsons9da096f2009-03-29 09:55:42 +02001008 * that has been freed has its buffer still pinned in memory because
David Chinner641c56f2007-06-18 16:50:17 +10001009 * the transaction is still sitting in a iclog. The stale inodes
1010 * on that buffer will have their flush locks held until the
1011 * transaction hits the disk and the callbacks run. the inode
1012 * flush takes the flush lock unconditionally and with nothing to
1013 * push out the iclog we will never get that unlocked. hence we
1014 * need to force the log first.
1015 */
Christoph Hellwiga14a3482010-01-19 09:56:46 +00001016 xfs_log_force(mp, XFS_LOG_SYNC);
Dave Chinnerc8543632010-02-06 12:39:36 +11001017
1018 /*
Carlos Maiolinoe6b3bb72016-05-18 11:11:27 +10001019 * We now need to tell the world we are unmounting. This will allow
1020 * us to detect that the filesystem is going away and we should error
1021 * out anything that we have been retrying in the background. This will
1022 * prevent neverending retries in AIL pushing from hanging the unmount.
1023 */
1024 mp->m_flags |= XFS_MOUNT_UNMOUNTING;
1025
1026 /*
Christoph Hellwig211e4d42012-04-23 15:58:34 +10001027 * Flush all pending changes from the AIL.
Dave Chinnerc8543632010-02-06 12:39:36 +11001028 */
Christoph Hellwig211e4d42012-04-23 15:58:34 +10001029 xfs_ail_push_all_sync(mp->m_ail);
1030
1031 /*
1032 * And reclaim all inodes. At this point there should be no dirty
Dave Chinner7e185302012-10-08 21:56:00 +11001033 * inodes and none should be pinned or locked, but use synchronous
1034 * reclaim just to be sure. We can stop background inode reclaim
1035 * here as well if it is still running.
Christoph Hellwig211e4d42012-04-23 15:58:34 +10001036 */
Dave Chinner7e185302012-10-08 21:56:00 +11001037 cancel_delayed_work_sync(&mp->m_reclaim_work);
Dave Chinnerc8543632010-02-06 12:39:36 +11001038 xfs_reclaim_inodes(mp, SYNC_WAIT);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039
Christoph Hellwig7d095252009-06-08 15:33:32 +02001040 xfs_qm_unmount(mp);
Lachlan McIlroya357a122008-10-30 16:53:25 +11001041
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 /*
David Chinner84e1e992007-06-18 16:50:27 +10001043 * Unreserve any blocks we have so that when we unmount we don't account
1044 * the reserved free space as used. This is really only necessary for
1045 * lazy superblock counting because it trusts the incore superblock
Malcolm Parsons9da096f2009-03-29 09:55:42 +02001046 * counters to be absolutely correct on clean unmount.
David Chinner84e1e992007-06-18 16:50:27 +10001047 *
1048 * We don't bother correcting this elsewhere for lazy superblock
1049 * counting because on mount of an unclean filesystem we reconstruct the
1050 * correct counter value and this is irrelevant.
1051 *
1052 * For non-lazy counter filesystems, this doesn't matter at all because
1053 * we only every apply deltas to the superblock and hence the incore
1054 * value does not matter....
1055 */
1056 resblks = 0;
David Chinner714082b2008-04-10 12:20:03 +10001057 error = xfs_reserve_blocks(mp, &resblks, NULL);
1058 if (error)
Dave Chinner0b932cc2011-03-07 10:08:35 +11001059 xfs_warn(mp, "Unable to free reserved block pool. "
David Chinner714082b2008-04-10 12:20:03 +10001060 "Freespace may not be correct on next mount.");
1061
Chandra Seetharamanadab0f62011-06-29 22:10:14 +00001062 error = xfs_log_sbcount(mp);
David Chinnere5720ee2008-04-10 12:21:18 +10001063 if (error)
Dave Chinner0b932cc2011-03-07 10:08:35 +11001064 xfs_warn(mp, "Unable to update superblock counters. "
David Chinnere5720ee2008-04-10 12:21:18 +10001065 "Freespace may not be correct on next mount.");
Christoph Hellwig87c7bec2011-09-14 14:08:26 +00001066
Bill O'Donnell225e4632015-10-12 18:21:19 +11001067
Christoph Hellwig21b699c2009-03-16 08:19:29 +01001068 xfs_log_unmount(mp);
Dave Chinner0650b552014-06-06 15:01:58 +10001069 xfs_da_unmount(mp);
Christoph Hellwig27174202009-03-30 10:21:31 +02001070 xfs_uuid_unmount(mp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
Christoph Hellwig1550d0b2008-08-13 16:17:37 +10001072#if defined(DEBUG)
Christoph Hellwig0ce4cfd2007-08-30 17:20:53 +10001073 xfs_errortag_clearall(mp, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074#endif
Christoph Hellwigff4f0382008-08-13 16:50:47 +10001075 xfs_free_perag(mp);
Brian Fostera31b1d32014-07-15 08:07:01 +10001076
Carlos Maiolino192852b2016-05-18 10:58:51 +10001077 xfs_error_sysfs_del(mp);
Bill O'Donnell225e4632015-10-12 18:21:19 +11001078 xfs_sysfs_del(&mp->m_stats.xs_kobj);
Brian Fostera31b1d32014-07-15 08:07:01 +10001079 xfs_sysfs_del(&mp->m_kobj);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080}
1081
Brian Foster91ee5752014-11-28 14:02:59 +11001082/*
1083 * Determine whether modifications can proceed. The caller specifies the minimum
1084 * freeze level for which modifications should not be allowed. This allows
1085 * certain operations to proceed while the freeze sequence is in progress, if
1086 * necessary.
1087 */
1088bool
1089xfs_fs_writable(
1090 struct xfs_mount *mp,
1091 int level)
David Chinner92821e22007-05-24 15:26:31 +10001092{
Brian Foster91ee5752014-11-28 14:02:59 +11001093 ASSERT(level > SB_UNFROZEN);
1094 if ((mp->m_super->s_writers.frozen >= level) ||
1095 XFS_FORCED_SHUTDOWN(mp) || (mp->m_flags & XFS_MOUNT_RDONLY))
1096 return false;
1097
1098 return true;
David Chinner92821e22007-05-24 15:26:31 +10001099}
1100
1101/*
Alex Elderb2ce3972011-07-11 09:51:44 -05001102 * xfs_log_sbcount
1103 *
Chandra Seetharamanadab0f62011-06-29 22:10:14 +00001104 * Sync the superblock counters to disk.
Alex Elderb2ce3972011-07-11 09:51:44 -05001105 *
Brian Foster91ee5752014-11-28 14:02:59 +11001106 * Note this code can be called during the process of freezing, so we use the
1107 * transaction allocator that does not block when the transaction subsystem is
1108 * in its frozen state.
David Chinner92821e22007-05-24 15:26:31 +10001109 */
1110int
Chandra Seetharamanadab0f62011-06-29 22:10:14 +00001111xfs_log_sbcount(xfs_mount_t *mp)
David Chinner92821e22007-05-24 15:26:31 +10001112{
Brian Foster91ee5752014-11-28 14:02:59 +11001113 /* allow this to proceed during the freeze sequence... */
1114 if (!xfs_fs_writable(mp, SB_FREEZE_COMPLETE))
David Chinner92821e22007-05-24 15:26:31 +10001115 return 0;
1116
David Chinner92821e22007-05-24 15:26:31 +10001117 /*
1118 * we don't need to do this if we are updating the superblock
1119 * counters on every modification.
1120 */
1121 if (!xfs_sb_version_haslazysbcount(&mp->m_sb))
1122 return 0;
1123
Dave Chinner61e63ec2015-01-22 09:10:31 +11001124 return xfs_sync_sb(mp, true);
David Chinner92821e22007-05-24 15:26:31 +10001125}
1126
Dave Chinner8c1903d2015-05-29 07:39:34 +10001127/*
1128 * Deltas for the inode count are +/-64, hence we use a large batch size
1129 * of 128 so we don't need to take the counter lock on every update.
1130 */
1131#define XFS_ICOUNT_BATCH 128
Dave Chinner501ab322015-02-23 21:19:28 +11001132int
1133xfs_mod_icount(
1134 struct xfs_mount *mp,
1135 int64_t delta)
1136{
Dave Chinner8c1903d2015-05-29 07:39:34 +10001137 __percpu_counter_add(&mp->m_icount, delta, XFS_ICOUNT_BATCH);
1138 if (__percpu_counter_compare(&mp->m_icount, 0, XFS_ICOUNT_BATCH) < 0) {
Dave Chinner501ab322015-02-23 21:19:28 +11001139 ASSERT(0);
1140 percpu_counter_add(&mp->m_icount, -delta);
1141 return -EINVAL;
1142 }
1143 return 0;
1144}
1145
Dave Chinnere88b64e2015-02-23 21:19:53 +11001146int
1147xfs_mod_ifree(
1148 struct xfs_mount *mp,
1149 int64_t delta)
1150{
1151 percpu_counter_add(&mp->m_ifree, delta);
1152 if (percpu_counter_compare(&mp->m_ifree, 0) < 0) {
1153 ASSERT(0);
1154 percpu_counter_add(&mp->m_ifree, -delta);
1155 return -EINVAL;
1156 }
1157 return 0;
1158}
Dave Chinner0d485ad2015-02-23 21:22:03 +11001159
Dave Chinner8c1903d2015-05-29 07:39:34 +10001160/*
1161 * Deltas for the block count can vary from 1 to very large, but lock contention
1162 * only occurs on frequent small block count updates such as in the delayed
1163 * allocation path for buffered writes (page a time updates). Hence we set
1164 * a large batch count (1024) to minimise global counter updates except when
1165 * we get near to ENOSPC and we have to be very accurate with our updates.
1166 */
1167#define XFS_FDBLOCKS_BATCH 1024
Dave Chinner0d485ad2015-02-23 21:22:03 +11001168int
1169xfs_mod_fdblocks(
1170 struct xfs_mount *mp,
1171 int64_t delta,
1172 bool rsvd)
1173{
1174 int64_t lcounter;
1175 long long res_used;
1176 s32 batch;
1177
1178 if (delta > 0) {
1179 /*
1180 * If the reserve pool is depleted, put blocks back into it
1181 * first. Most of the time the pool is full.
1182 */
1183 if (likely(mp->m_resblks == mp->m_resblks_avail)) {
1184 percpu_counter_add(&mp->m_fdblocks, delta);
1185 return 0;
1186 }
1187
1188 spin_lock(&mp->m_sb_lock);
1189 res_used = (long long)(mp->m_resblks - mp->m_resblks_avail);
1190
1191 if (res_used > delta) {
1192 mp->m_resblks_avail += delta;
1193 } else {
1194 delta -= res_used;
1195 mp->m_resblks_avail = mp->m_resblks;
1196 percpu_counter_add(&mp->m_fdblocks, delta);
1197 }
1198 spin_unlock(&mp->m_sb_lock);
1199 return 0;
1200 }
1201
1202 /*
1203 * Taking blocks away, need to be more accurate the closer we
1204 * are to zero.
1205 *
Dave Chinner0d485ad2015-02-23 21:22:03 +11001206 * If the counter has a value of less than 2 * max batch size,
1207 * then make everything serialise as we are real close to
1208 * ENOSPC.
1209 */
Dave Chinner8c1903d2015-05-29 07:39:34 +10001210 if (__percpu_counter_compare(&mp->m_fdblocks, 2 * XFS_FDBLOCKS_BATCH,
1211 XFS_FDBLOCKS_BATCH) < 0)
Dave Chinner0d485ad2015-02-23 21:22:03 +11001212 batch = 1;
1213 else
Dave Chinner8c1903d2015-05-29 07:39:34 +10001214 batch = XFS_FDBLOCKS_BATCH;
Dave Chinner0d485ad2015-02-23 21:22:03 +11001215
1216 __percpu_counter_add(&mp->m_fdblocks, delta, batch);
Dave Chinner8c1903d2015-05-29 07:39:34 +10001217 if (__percpu_counter_compare(&mp->m_fdblocks, XFS_ALLOC_SET_ASIDE(mp),
1218 XFS_FDBLOCKS_BATCH) >= 0) {
Dave Chinner0d485ad2015-02-23 21:22:03 +11001219 /* we had space! */
1220 return 0;
1221 }
1222
1223 /*
1224 * lock up the sb for dipping into reserves before releasing the space
1225 * that took us to ENOSPC.
1226 */
1227 spin_lock(&mp->m_sb_lock);
1228 percpu_counter_add(&mp->m_fdblocks, -delta);
1229 if (!rsvd)
1230 goto fdblocks_enospc;
1231
1232 lcounter = (long long)mp->m_resblks_avail + delta;
1233 if (lcounter >= 0) {
1234 mp->m_resblks_avail = lcounter;
1235 spin_unlock(&mp->m_sb_lock);
1236 return 0;
1237 }
1238 printk_once(KERN_WARNING
1239 "Filesystem \"%s\": reserve blocks depleted! "
1240 "Consider increasing reserve pool size.",
1241 mp->m_fsname);
1242fdblocks_enospc:
1243 spin_unlock(&mp->m_sb_lock);
1244 return -ENOSPC;
1245}
1246
Dave Chinnerbab98bb2015-02-23 21:22:54 +11001247int
1248xfs_mod_frextents(
1249 struct xfs_mount *mp,
1250 int64_t delta)
1251{
1252 int64_t lcounter;
1253 int ret = 0;
1254
1255 spin_lock(&mp->m_sb_lock);
1256 lcounter = mp->m_sb.sb_frextents + delta;
1257 if (lcounter < 0)
1258 ret = -ENOSPC;
1259 else
1260 mp->m_sb.sb_frextents = lcounter;
1261 spin_unlock(&mp->m_sb_lock);
1262 return ret;
1263}
1264
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266 * xfs_getsb() is called to obtain the buffer for the superblock.
1267 * The buffer is returned locked and read in from disk.
1268 * The buffer should be released with a call to xfs_brelse().
1269 *
1270 * If the flags parameter is BUF_TRYLOCK, then we'll only return
1271 * the superblock buffer if it can be locked without sleeping.
1272 * If it can't then we'll return NULL.
1273 */
Christoph Hellwig0c842ad2011-07-08 14:36:19 +02001274struct xfs_buf *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275xfs_getsb(
Christoph Hellwig0c842ad2011-07-08 14:36:19 +02001276 struct xfs_mount *mp,
1277 int flags)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278{
Christoph Hellwig0c842ad2011-07-08 14:36:19 +02001279 struct xfs_buf *bp = mp->m_sb_bp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280
Christoph Hellwig0c842ad2011-07-08 14:36:19 +02001281 if (!xfs_buf_trylock(bp)) {
1282 if (flags & XBF_TRYLOCK)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 return NULL;
Christoph Hellwig0c842ad2011-07-08 14:36:19 +02001284 xfs_buf_lock(bp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001285 }
Christoph Hellwig0c842ad2011-07-08 14:36:19 +02001286
Chandra Seetharaman72790aa2011-07-22 23:40:04 +00001287 xfs_buf_hold(bp);
Dave Chinnerb0388bf2016-02-10 15:01:11 +11001288 ASSERT(bp->b_flags & XBF_DONE);
Jesper Juhl014c2542006-01-15 02:37:08 +01001289 return bp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290}
1291
1292/*
1293 * Used to free the superblock along various error paths.
1294 */
1295void
1296xfs_freesb(
Dave Chinner26af6552010-09-22 10:47:20 +10001297 struct xfs_mount *mp)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001298{
Dave Chinner26af6552010-09-22 10:47:20 +10001299 struct xfs_buf *bp = mp->m_sb_bp;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300
Dave Chinner26af6552010-09-22 10:47:20 +10001301 xfs_buf_lock(bp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302 mp->m_sb_bp = NULL;
Dave Chinner26af6552010-09-22 10:47:20 +10001303 xfs_buf_relse(bp);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304}
1305
1306/*
Christoph Hellwigdda35b82010-02-15 09:44:46 +00001307 * If the underlying (data/log/rt) device is readonly, there are some
1308 * operations that cannot proceed.
1309 */
1310int
1311xfs_dev_is_read_only(
1312 struct xfs_mount *mp,
1313 char *message)
1314{
1315 if (xfs_readonly_buftarg(mp->m_ddev_targp) ||
1316 xfs_readonly_buftarg(mp->m_logdev_targp) ||
1317 (mp->m_rtdev_targp && xfs_readonly_buftarg(mp->m_rtdev_targp))) {
Dave Chinner0b932cc2011-03-07 10:08:35 +11001318 xfs_notice(mp, "%s required on read-only device.", message);
1319 xfs_notice(mp, "write access unavailable, cannot proceed.");
Dave Chinner24513372014-06-25 14:58:08 +10001320 return -EROFS;
Christoph Hellwigdda35b82010-02-15 09:44:46 +00001321 }
1322 return 0;
1323}