Commit | Line | Data |
---|---|---|
0b61f8a4 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 | 2 | /* |
a805bad5 | 3 | * Copyright (c) 2000-2006 Silicon Graphics, Inc. |
7b718769 | 4 | * All Rights Reserved. |
1da177e4 | 5 | */ |
0b1b213f | 6 | |
1da177e4 | 7 | #include "xfs.h" |
70a9883c | 8 | #include "xfs_shared.h" |
6ca1c906 | 9 | #include "xfs_format.h" |
239880ef DC |
10 | #include "xfs_log_format.h" |
11 | #include "xfs_trans_resv.h" | |
1da177e4 | 12 | #include "xfs_sb.h" |
1da177e4 | 13 | #include "xfs_mount.h" |
1da177e4 | 14 | #include "xfs_inode.h" |
a844f451 | 15 | #include "xfs_btree.h" |
1da177e4 | 16 | #include "xfs_bmap.h" |
a4fbe6ab | 17 | #include "xfs_alloc.h" |
9909c4aa | 18 | #include "xfs_fsops.h" |
239880ef | 19 | #include "xfs_trans.h" |
1da177e4 | 20 | #include "xfs_buf_item.h" |
239880ef | 21 | #include "xfs_log.h" |
a67d7c5f | 22 | #include "xfs_log_priv.h" |
2b9ab5ab | 23 | #include "xfs_dir2.h" |
9f8868ff CH |
24 | #include "xfs_extfree_item.h" |
25 | #include "xfs_mru_cache.h" | |
26 | #include "xfs_inode_item.h" | |
6d8b79cf | 27 | #include "xfs_icache.h" |
0b1b213f | 28 | #include "xfs_trace.h" |
3ebe7d2d | 29 | #include "xfs_icreate_item.h" |
a4fbe6ab DC |
30 | #include "xfs_filestream.h" |
31 | #include "xfs_quota.h" | |
65b65735 | 32 | #include "xfs_sysfs.h" |
30cbc591 | 33 | #include "xfs_ondisk.h" |
5880f2d7 | 34 | #include "xfs_rmap_item.h" |
baf4bcac | 35 | #include "xfs_refcount_item.h" |
6413a014 | 36 | #include "xfs_bmap_item.h" |
5e7e605c | 37 | #include "xfs_reflink.h" |
894ecacf | 38 | #include "xfs_pwork.h" |
9bbafc71 | 39 | #include "xfs_ag.h" |
f3c799c2 | 40 | #include "xfs_defer.h" |
4136e38a | 41 | #include "xfs_attr_item.h" |
d9c61ccb | 42 | #include "xfs_xattr.h" |
784eb7d8 | 43 | #include "xfs_iunlink_item.h" |
3cfb9290 | 44 | #include "xfs_dahash_test.h" |
d7a74cad | 45 | #include "scrub/stats.h" |
1da177e4 | 46 | |
dddde68b | 47 | #include <linux/magic.h> |
73e5fff9 IK |
48 | #include <linux/fs_context.h> |
49 | #include <linux/fs_parser.h> | |
1da177e4 | 50 | |
b87221de | 51 | static const struct super_operations xfs_super_operations; |
65b65735 | 52 | |
a76dba3b | 53 | static struct dentry *xfs_debugfs; /* top-level xfs debugfs dir */ |
e3aed1a0 | 54 | static struct kset *xfs_kset; /* top-level xfs sysfs dir */ |
65b65735 BF |
55 | #ifdef DEBUG |
56 | static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */ | |
57 | #endif | |
1da177e4 | 58 | |
8d6c3446 IW |
59 | enum xfs_dax_mode { |
60 | XFS_DAX_INODE = 0, | |
61 | XFS_DAX_ALWAYS = 1, | |
62 | XFS_DAX_NEVER = 2, | |
63 | }; | |
64 | ||
65 | static void | |
66 | xfs_mount_set_dax_mode( | |
67 | struct xfs_mount *mp, | |
68 | enum xfs_dax_mode mode) | |
69 | { | |
70 | switch (mode) { | |
71 | case XFS_DAX_INODE: | |
0560f31a | 72 | mp->m_features &= ~(XFS_FEAT_DAX_ALWAYS | XFS_FEAT_DAX_NEVER); |
8d6c3446 IW |
73 | break; |
74 | case XFS_DAX_ALWAYS: | |
0560f31a DC |
75 | mp->m_features |= XFS_FEAT_DAX_ALWAYS; |
76 | mp->m_features &= ~XFS_FEAT_DAX_NEVER; | |
8d6c3446 IW |
77 | break; |
78 | case XFS_DAX_NEVER: | |
0560f31a DC |
79 | mp->m_features |= XFS_FEAT_DAX_NEVER; |
80 | mp->m_features &= ~XFS_FEAT_DAX_ALWAYS; | |
8d6c3446 IW |
81 | break; |
82 | } | |
83 | } | |
84 | ||
85 | static const struct constant_table dax_param_enums[] = { | |
86 | {"inode", XFS_DAX_INODE }, | |
87 | {"always", XFS_DAX_ALWAYS }, | |
88 | {"never", XFS_DAX_NEVER }, | |
89 | {} | |
90 | }; | |
91 | ||
62a877e3 CH |
92 | /* |
93 | * Table driven mount option parser. | |
62a877e3 CH |
94 | */ |
95 | enum { | |
8da57c5c | 96 | Opt_logbufs, Opt_logbsize, Opt_logdev, Opt_rtdev, |
2e74af0e | 97 | Opt_wsync, Opt_noalign, Opt_swalloc, Opt_sunit, Opt_swidth, Opt_nouuid, |
94079285 | 98 | Opt_grpid, Opt_nogrpid, Opt_bsdgroups, Opt_sysvgroups, |
1c02d502 ES |
99 | Opt_allocsize, Opt_norecovery, Opt_inode64, Opt_inode32, Opt_ikeep, |
100 | Opt_noikeep, Opt_largeio, Opt_nolargeio, Opt_attr2, Opt_noattr2, | |
101 | Opt_filestreams, Opt_quota, Opt_noquota, Opt_usrquota, Opt_grpquota, | |
102 | Opt_prjquota, Opt_uquota, Opt_gquota, Opt_pquota, | |
2e74af0e | 103 | Opt_uqnoenforce, Opt_gqnoenforce, Opt_pqnoenforce, Opt_qnoenforce, |
8d6c3446 | 104 | Opt_discard, Opt_nodiscard, Opt_dax, Opt_dax_enum, |
62a877e3 CH |
105 | }; |
106 | ||
d7167b14 | 107 | static const struct fs_parameter_spec xfs_fs_parameters[] = { |
73e5fff9 IK |
108 | fsparam_u32("logbufs", Opt_logbufs), |
109 | fsparam_string("logbsize", Opt_logbsize), | |
110 | fsparam_string("logdev", Opt_logdev), | |
111 | fsparam_string("rtdev", Opt_rtdev), | |
112 | fsparam_flag("wsync", Opt_wsync), | |
113 | fsparam_flag("noalign", Opt_noalign), | |
114 | fsparam_flag("swalloc", Opt_swalloc), | |
115 | fsparam_u32("sunit", Opt_sunit), | |
116 | fsparam_u32("swidth", Opt_swidth), | |
117 | fsparam_flag("nouuid", Opt_nouuid), | |
118 | fsparam_flag("grpid", Opt_grpid), | |
119 | fsparam_flag("nogrpid", Opt_nogrpid), | |
120 | fsparam_flag("bsdgroups", Opt_bsdgroups), | |
121 | fsparam_flag("sysvgroups", Opt_sysvgroups), | |
122 | fsparam_string("allocsize", Opt_allocsize), | |
123 | fsparam_flag("norecovery", Opt_norecovery), | |
124 | fsparam_flag("inode64", Opt_inode64), | |
125 | fsparam_flag("inode32", Opt_inode32), | |
126 | fsparam_flag("ikeep", Opt_ikeep), | |
127 | fsparam_flag("noikeep", Opt_noikeep), | |
128 | fsparam_flag("largeio", Opt_largeio), | |
129 | fsparam_flag("nolargeio", Opt_nolargeio), | |
130 | fsparam_flag("attr2", Opt_attr2), | |
131 | fsparam_flag("noattr2", Opt_noattr2), | |
132 | fsparam_flag("filestreams", Opt_filestreams), | |
133 | fsparam_flag("quota", Opt_quota), | |
134 | fsparam_flag("noquota", Opt_noquota), | |
135 | fsparam_flag("usrquota", Opt_usrquota), | |
136 | fsparam_flag("grpquota", Opt_grpquota), | |
137 | fsparam_flag("prjquota", Opt_prjquota), | |
138 | fsparam_flag("uquota", Opt_uquota), | |
139 | fsparam_flag("gquota", Opt_gquota), | |
140 | fsparam_flag("pquota", Opt_pquota), | |
141 | fsparam_flag("uqnoenforce", Opt_uqnoenforce), | |
142 | fsparam_flag("gqnoenforce", Opt_gqnoenforce), | |
143 | fsparam_flag("pqnoenforce", Opt_pqnoenforce), | |
144 | fsparam_flag("qnoenforce", Opt_qnoenforce), | |
145 | fsparam_flag("discard", Opt_discard), | |
146 | fsparam_flag("nodiscard", Opt_nodiscard), | |
147 | fsparam_flag("dax", Opt_dax), | |
8d6c3446 | 148 | fsparam_enum("dax", Opt_dax_enum, dax_param_enums), |
73e5fff9 | 149 | {} |
62a877e3 CH |
150 | }; |
151 | ||
a67d7c5f | 152 | struct proc_xfs_info { |
cbe4dab1 DC |
153 | uint64_t flag; |
154 | char *str; | |
a67d7c5f DC |
155 | }; |
156 | ||
21f55993 CH |
157 | static int |
158 | xfs_fs_show_options( | |
159 | struct seq_file *m, | |
160 | struct dentry *root) | |
a67d7c5f DC |
161 | { |
162 | static struct proc_xfs_info xfs_info_set[] = { | |
163 | /* the few simple ones we can get from the mount struct */ | |
0560f31a DC |
164 | { XFS_FEAT_IKEEP, ",ikeep" }, |
165 | { XFS_FEAT_WSYNC, ",wsync" }, | |
166 | { XFS_FEAT_NOALIGN, ",noalign" }, | |
167 | { XFS_FEAT_SWALLOC, ",swalloc" }, | |
168 | { XFS_FEAT_NOUUID, ",nouuid" }, | |
169 | { XFS_FEAT_NORECOVERY, ",norecovery" }, | |
170 | { XFS_FEAT_ATTR2, ",attr2" }, | |
171 | { XFS_FEAT_FILESTREAMS, ",filestreams" }, | |
172 | { XFS_FEAT_GRPID, ",grpid" }, | |
173 | { XFS_FEAT_DISCARD, ",discard" }, | |
174 | { XFS_FEAT_LARGE_IOSIZE, ",largeio" }, | |
175 | { XFS_FEAT_DAX_ALWAYS, ",dax=always" }, | |
176 | { XFS_FEAT_DAX_NEVER, ",dax=never" }, | |
a67d7c5f DC |
177 | { 0, NULL } |
178 | }; | |
21f55993 | 179 | struct xfs_mount *mp = XFS_M(root->d_sb); |
a67d7c5f DC |
180 | struct proc_xfs_info *xfs_infop; |
181 | ||
182 | for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) { | |
0560f31a | 183 | if (mp->m_features & xfs_infop->flag) |
a67d7c5f DC |
184 | seq_puts(m, xfs_infop->str); |
185 | } | |
1775c506 | 186 | |
0560f31a | 187 | seq_printf(m, ",inode%d", xfs_has_small_inums(mp) ? 32 : 64); |
a67d7c5f | 188 | |
0560f31a | 189 | if (xfs_has_allocsize(mp)) |
2e74af0e | 190 | seq_printf(m, ",allocsize=%dk", |
aa58d445 | 191 | (1 << mp->m_allocsize_log) >> 10); |
a67d7c5f DC |
192 | |
193 | if (mp->m_logbufs > 0) | |
2e74af0e | 194 | seq_printf(m, ",logbufs=%d", mp->m_logbufs); |
a67d7c5f | 195 | if (mp->m_logbsize > 0) |
2e74af0e | 196 | seq_printf(m, ",logbsize=%dk", mp->m_logbsize >> 10); |
a67d7c5f DC |
197 | |
198 | if (mp->m_logname) | |
2e74af0e | 199 | seq_show_option(m, "logdev", mp->m_logname); |
a67d7c5f | 200 | if (mp->m_rtname) |
2e74af0e | 201 | seq_show_option(m, "rtdev", mp->m_rtname); |
a67d7c5f DC |
202 | |
203 | if (mp->m_dalign > 0) | |
2e74af0e | 204 | seq_printf(m, ",sunit=%d", |
a67d7c5f DC |
205 | (int)XFS_FSB_TO_BB(mp, mp->m_dalign)); |
206 | if (mp->m_swidth > 0) | |
2e74af0e | 207 | seq_printf(m, ",swidth=%d", |
a67d7c5f DC |
208 | (int)XFS_FSB_TO_BB(mp, mp->m_swidth)); |
209 | ||
149e53af CH |
210 | if (mp->m_qflags & XFS_UQUOTA_ENFD) |
211 | seq_puts(m, ",usrquota"); | |
212 | else if (mp->m_qflags & XFS_UQUOTA_ACCT) | |
213 | seq_puts(m, ",uqnoenforce"); | |
a67d7c5f | 214 | |
149e53af CH |
215 | if (mp->m_qflags & XFS_PQUOTA_ENFD) |
216 | seq_puts(m, ",prjquota"); | |
217 | else if (mp->m_qflags & XFS_PQUOTA_ACCT) | |
218 | seq_puts(m, ",pqnoenforce"); | |
219 | ||
220 | if (mp->m_qflags & XFS_GQUOTA_ENFD) | |
221 | seq_puts(m, ",grpquota"); | |
222 | else if (mp->m_qflags & XFS_GQUOTA_ACCT) | |
223 | seq_puts(m, ",gqnoenforce"); | |
a67d7c5f DC |
224 | |
225 | if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT)) | |
2e74af0e | 226 | seq_puts(m, ",noquota"); |
21f55993 CH |
227 | |
228 | return 0; | |
a67d7c5f | 229 | } |
91083269 | 230 | |
7ac2ff8b DC |
231 | static bool |
232 | xfs_set_inode_alloc_perag( | |
233 | struct xfs_perag *pag, | |
234 | xfs_ino_t ino, | |
235 | xfs_agnumber_t max_metadata) | |
236 | { | |
237 | if (!xfs_is_inode32(pag->pag_mount)) { | |
238 | set_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate); | |
239 | clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate); | |
240 | return false; | |
241 | } | |
242 | ||
243 | if (ino > XFS_MAXINUMBER_32) { | |
244 | clear_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate); | |
245 | clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate); | |
246 | return false; | |
247 | } | |
248 | ||
249 | set_bit(XFS_AGSTATE_ALLOWS_INODES, &pag->pag_opstate); | |
250 | if (pag->pag_agno < max_metadata) | |
251 | set_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate); | |
252 | else | |
253 | clear_bit(XFS_AGSTATE_PREFERS_METADATA, &pag->pag_opstate); | |
254 | return true; | |
255 | } | |
256 | ||
9de67c3b | 257 | /* |
12c3f05c ES |
258 | * Set parameters for inode allocation heuristics, taking into account |
259 | * filesystem size and inode32/inode64 mount options; i.e. specifically | |
0560f31a | 260 | * whether or not XFS_FEAT_SMALL_INUMS is set. |
12c3f05c ES |
261 | * |
262 | * Inode allocation patterns are altered only if inode32 is requested | |
0560f31a | 263 | * (XFS_FEAT_SMALL_INUMS), and the filesystem is sufficiently large. |
2e973b2c | 264 | * If altered, XFS_OPSTATE_INODE32 is set as well. |
12c3f05c ES |
265 | * |
266 | * An agcount independent of that in the mount structure is provided | |
267 | * because in the growfs case, mp->m_sb.sb_agcount is not yet updated | |
268 | * to the potentially higher ag count. | |
269 | * | |
270 | * Returns the maximum AG index which may contain inodes. | |
9de67c3b | 271 | */ |
2d2194f6 | 272 | xfs_agnumber_t |
12c3f05c ES |
273 | xfs_set_inode_alloc( |
274 | struct xfs_mount *mp, | |
275 | xfs_agnumber_t agcount) | |
2d2194f6 | 276 | { |
12c3f05c | 277 | xfs_agnumber_t index; |
4056c1d0 | 278 | xfs_agnumber_t maxagi = 0; |
2d2194f6 CM |
279 | xfs_sb_t *sbp = &mp->m_sb; |
280 | xfs_agnumber_t max_metadata; | |
54aa61f8 ES |
281 | xfs_agino_t agino; |
282 | xfs_ino_t ino; | |
2d2194f6 | 283 | |
12c3f05c ES |
284 | /* |
285 | * Calculate how much should be reserved for inodes to meet | |
286 | * the max inode percentage. Used only for inode32. | |
2d2194f6 | 287 | */ |
ef325959 | 288 | if (M_IGEO(mp)->maxicount) { |
c8ce540d | 289 | uint64_t icount; |
2d2194f6 CM |
290 | |
291 | icount = sbp->sb_dblocks * sbp->sb_imax_pct; | |
292 | do_div(icount, 100); | |
293 | icount += sbp->sb_agblocks - 1; | |
294 | do_div(icount, sbp->sb_agblocks); | |
295 | max_metadata = icount; | |
296 | } else { | |
9de67c3b | 297 | max_metadata = agcount; |
2d2194f6 CM |
298 | } |
299 | ||
12c3f05c | 300 | /* Get the last possible inode in the filesystem */ |
43004b2a | 301 | agino = XFS_AGB_TO_AGINO(mp, sbp->sb_agblocks - 1); |
12c3f05c ES |
302 | ino = XFS_AGINO_TO_INO(mp, agcount - 1, agino); |
303 | ||
304 | /* | |
305 | * If user asked for no more than 32-bit inodes, and the fs is | |
2e973b2c | 306 | * sufficiently large, set XFS_OPSTATE_INODE32 if we must alter |
12c3f05c ES |
307 | * the allocator to accommodate the request. |
308 | */ | |
0560f31a | 309 | if (xfs_has_small_inums(mp) && ino > XFS_MAXINUMBER_32) |
2e973b2c | 310 | set_bit(XFS_OPSTATE_INODE32, &mp->m_opstate); |
12c3f05c | 311 | else |
2e973b2c | 312 | clear_bit(XFS_OPSTATE_INODE32, &mp->m_opstate); |
54aa61f8 | 313 | |
9de67c3b | 314 | for (index = 0; index < agcount; index++) { |
12c3f05c | 315 | struct xfs_perag *pag; |
4056c1d0 | 316 | |
12c3f05c | 317 | ino = XFS_AGINO_TO_INO(mp, index, agino); |
2d2194f6 CM |
318 | |
319 | pag = xfs_perag_get(mp, index); | |
7ac2ff8b DC |
320 | if (xfs_set_inode_alloc_perag(pag, ino, max_metadata)) |
321 | maxagi++; | |
2d2194f6 CM |
322 | xfs_perag_put(pag); |
323 | } | |
324 | ||
2e973b2c | 325 | return xfs_is_inode32(mp) ? maxagi : agcount; |
2d2194f6 CM |
326 | } |
327 | ||
679a9949 CH |
328 | static int |
329 | xfs_setup_dax_always( | |
330 | struct xfs_mount *mp) | |
a384f088 | 331 | { |
7b0800d0 CH |
332 | if (!mp->m_ddev_targp->bt_daxdev && |
333 | (!mp->m_rtdev_targp || !mp->m_rtdev_targp->bt_daxdev)) { | |
679a9949 CH |
334 | xfs_alert(mp, |
335 | "DAX unsupported by block device. Turning off DAX."); | |
336 | goto disable_dax; | |
7b0800d0 CH |
337 | } |
338 | ||
339 | if (mp->m_super->s_blocksize != PAGE_SIZE) { | |
340 | xfs_alert(mp, | |
341 | "DAX not supported for blocksize. Turning off DAX."); | |
342 | goto disable_dax; | |
679a9949 CH |
343 | } |
344 | ||
35fcd75a SR |
345 | if (xfs_has_reflink(mp) && |
346 | bdev_is_partition(mp->m_ddev_targp->bt_bdev)) { | |
347 | xfs_alert(mp, | |
348 | "DAX and reflink cannot work with multi-partitions!"); | |
679a9949 CH |
349 | return -EINVAL; |
350 | } | |
351 | ||
352 | xfs_warn(mp, "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); | |
353 | return 0; | |
354 | ||
355 | disable_dax: | |
356 | xfs_mount_set_dax_mode(mp, XFS_DAX_NEVER); | |
357 | return 0; | |
a384f088 CH |
358 | } |
359 | ||
3180e66d | 360 | STATIC int |
1da177e4 LT |
361 | xfs_blkdev_get( |
362 | xfs_mount_t *mp, | |
363 | const char *name, | |
364 | struct block_device **bdevp) | |
365 | { | |
366 | int error = 0; | |
367 | ||
8ffa54e3 CH |
368 | *bdevp = blkdev_get_by_path(name, BLK_OPEN_READ | BLK_OPEN_WRITE, |
369 | mp->m_super, &fs_holder_ops); | |
1da177e4 LT |
370 | if (IS_ERR(*bdevp)) { |
371 | error = PTR_ERR(*bdevp); | |
77af574e | 372 | xfs_warn(mp, "Invalid device [%s], error=%d", name, error); |
1da177e4 LT |
373 | } |
374 | ||
2451337d | 375 | return error; |
1da177e4 LT |
376 | } |
377 | ||
3180e66d | 378 | STATIC void |
35a93b14 | 379 | xfs_shutdown_devices( |
19f354d4 CH |
380 | struct xfs_mount *mp) |
381 | { | |
1a0a5dad CH |
382 | /* |
383 | * Udev is triggered whenever anyone closes a block device or unmounts | |
384 | * a file systemm on a block device. | |
385 | * The default udev rules invoke blkid to read the fs super and create | |
386 | * symlinks to the bdev under /dev/disk. For this, it uses buffered | |
387 | * reads through the page cache. | |
388 | * | |
389 | * xfs_db also uses buffered reads to examine metadata. There is no | |
390 | * coordination between xfs_db and udev, which means that they can run | |
391 | * concurrently. Note there is no coordination between the kernel and | |
392 | * blkid either. | |
393 | * | |
394 | * On a system with 64k pages, the page cache can cache the superblock | |
395 | * and the root inode (and hence the root directory) with the same 64k | |
396 | * page. If udev spawns blkid after the mkfs and the system is busy | |
397 | * enough that it is still running when xfs_db starts up, they'll both | |
398 | * read from the same page in the pagecache. | |
399 | * | |
400 | * The unmount writes updated inode metadata to disk directly. The XFS | |
401 | * buffer cache does not use the bdev pagecache, so it needs to | |
402 | * invalidate that pagecache on unmount. If the above scenario occurs, | |
403 | * the pagecache no longer reflects what's on disk, xfs_db reads the | |
404 | * stale metadata, and fails to find /a. Most of the time this succeeds | |
405 | * because closing a bdev invalidates the page cache, but when processes | |
406 | * race, everyone loses. | |
407 | */ | |
19f354d4 | 408 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { |
35a93b14 CH |
409 | blkdev_issue_flush(mp->m_logdev_targp->bt_bdev); |
410 | invalidate_bdev(mp->m_logdev_targp->bt_bdev); | |
19f354d4 CH |
411 | } |
412 | if (mp->m_rtdev_targp) { | |
35a93b14 CH |
413 | blkdev_issue_flush(mp->m_rtdev_targp->bt_bdev); |
414 | invalidate_bdev(mp->m_rtdev_targp->bt_bdev); | |
19f354d4 | 415 | } |
35a93b14 CH |
416 | blkdev_issue_flush(mp->m_ddev_targp->bt_bdev); |
417 | invalidate_bdev(mp->m_ddev_targp->bt_bdev); | |
19f354d4 CH |
418 | } |
419 | ||
420 | /* | |
421 | * The file system configurations are: | |
422 | * (1) device (partition) with data and internal log | |
423 | * (2) logical volume with data and log subvolumes. | |
424 | * (3) logical volume with data, log, and realtime subvolumes. | |
425 | * | |
426 | * We only have to handle opening the log and realtime volumes here if | |
427 | * they are present. The data subvolume has already been opened by | |
428 | * get_sb_bdev() and is stored in sb->s_bdev. | |
429 | */ | |
430 | STATIC int | |
431 | xfs_open_devices( | |
9d565ffa | 432 | struct xfs_mount *mp) |
19f354d4 | 433 | { |
8d945b59 CH |
434 | struct super_block *sb = mp->m_super; |
435 | struct block_device *ddev = sb->s_bdev; | |
19f354d4 CH |
436 | struct block_device *logdev = NULL, *rtdev = NULL; |
437 | int error; | |
438 | ||
8d945b59 CH |
439 | /* |
440 | * blkdev_put() can't be called under s_umount, see the comment | |
441 | * in get_tree_bdev() for more details | |
442 | */ | |
443 | up_write(&sb->s_umount); | |
444 | ||
19f354d4 CH |
445 | /* |
446 | * Open real time and log devices - order is important. | |
447 | */ | |
9d565ffa CH |
448 | if (mp->m_logname) { |
449 | error = xfs_blkdev_get(mp, mp->m_logname, &logdev); | |
19f354d4 | 450 | if (error) |
8d945b59 | 451 | goto out_relock; |
19f354d4 CH |
452 | } |
453 | ||
9d565ffa CH |
454 | if (mp->m_rtname) { |
455 | error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev); | |
19f354d4 CH |
456 | if (error) |
457 | goto out_close_logdev; | |
458 | ||
459 | if (rtdev == ddev || rtdev == logdev) { | |
4f10700a DC |
460 | xfs_warn(mp, |
461 | "Cannot mount filesystem with identical rtdev and ddev/logdev."); | |
2451337d | 462 | error = -EINVAL; |
19f354d4 CH |
463 | goto out_close_rtdev; |
464 | } | |
465 | } | |
466 | ||
467 | /* | |
468 | * Setup xfs_mount buffer target pointers | |
469 | */ | |
2451337d | 470 | error = -ENOMEM; |
5b5abbef | 471 | mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev); |
19f354d4 CH |
472 | if (!mp->m_ddev_targp) |
473 | goto out_close_rtdev; | |
474 | ||
475 | if (rtdev) { | |
5b5abbef | 476 | mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev); |
19f354d4 CH |
477 | if (!mp->m_rtdev_targp) |
478 | goto out_free_ddev_targ; | |
479 | } | |
480 | ||
481 | if (logdev && logdev != ddev) { | |
5b5abbef | 482 | mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev); |
19f354d4 CH |
483 | if (!mp->m_logdev_targp) |
484 | goto out_free_rtdev_targ; | |
485 | } else { | |
486 | mp->m_logdev_targp = mp->m_ddev_targp; | |
487 | } | |
488 | ||
8d945b59 CH |
489 | error = 0; |
490 | out_relock: | |
491 | down_write(&sb->s_umount); | |
492 | return error; | |
19f354d4 CH |
493 | |
494 | out_free_rtdev_targ: | |
495 | if (mp->m_rtdev_targp) | |
a1f69417 | 496 | xfs_free_buftarg(mp->m_rtdev_targp); |
19f354d4 | 497 | out_free_ddev_targ: |
a1f69417 | 498 | xfs_free_buftarg(mp->m_ddev_targp); |
19f354d4 | 499 | out_close_rtdev: |
d3ef7e94 | 500 | if (rtdev) |
8ffa54e3 | 501 | blkdev_put(rtdev, sb); |
19f354d4 | 502 | out_close_logdev: |
5b5abbef | 503 | if (logdev && logdev != ddev) |
8ffa54e3 | 504 | blkdev_put(logdev, sb); |
8d945b59 | 505 | goto out_relock; |
19f354d4 CH |
506 | } |
507 | ||
e34b562c CH |
508 | /* |
509 | * Setup xfs_mount buffer target pointers based on superblock | |
510 | */ | |
511 | STATIC int | |
512 | xfs_setup_devices( | |
513 | struct xfs_mount *mp) | |
514 | { | |
515 | int error; | |
19f354d4 | 516 | |
a96c4151 | 517 | error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize); |
e34b562c CH |
518 | if (error) |
519 | return error; | |
520 | ||
521 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { | |
522 | unsigned int log_sector_size = BBSIZE; | |
523 | ||
38c26bfd | 524 | if (xfs_has_sector(mp)) |
e34b562c CH |
525 | log_sector_size = mp->m_sb.sb_logsectsize; |
526 | error = xfs_setsize_buftarg(mp->m_logdev_targp, | |
e34b562c CH |
527 | log_sector_size); |
528 | if (error) | |
529 | return error; | |
530 | } | |
531 | if (mp->m_rtdev_targp) { | |
532 | error = xfs_setsize_buftarg(mp->m_rtdev_targp, | |
e34b562c CH |
533 | mp->m_sb.sb_sectsize); |
534 | if (error) | |
535 | return error; | |
536 | } | |
537 | ||
538 | return 0; | |
539 | } | |
19f354d4 | 540 | |
aa6bf01d CH |
541 | STATIC int |
542 | xfs_init_mount_workqueues( | |
543 | struct xfs_mount *mp) | |
544 | { | |
78c931b8 | 545 | mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s", |
05a302a1 DW |
546 | XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM), |
547 | 1, mp->m_super->s_id); | |
78c931b8 BF |
548 | if (!mp->m_buf_workqueue) |
549 | goto out; | |
550 | ||
aa6bf01d | 551 | mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s", |
05a302a1 DW |
552 | XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM), |
553 | 0, mp->m_super->s_id); | |
aa6bf01d | 554 | if (!mp->m_unwritten_workqueue) |
28408243 | 555 | goto out_destroy_buf; |
aa6bf01d | 556 | |
5889608d | 557 | mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s", |
05a302a1 DW |
558 | XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM), |
559 | 0, mp->m_super->s_id); | |
5889608d | 560 | if (!mp->m_reclaim_workqueue) |
33c0dd78 | 561 | goto out_destroy_unwritten; |
5889608d | 562 | |
ab23a776 DC |
563 | mp->m_blockgc_wq = alloc_workqueue("xfs-blockgc/%s", |
564 | XFS_WQFLAGS(WQ_UNBOUND | WQ_FREEZABLE | WQ_MEM_RECLAIM), | |
05a302a1 | 565 | 0, mp->m_super->s_id); |
ab23a776 | 566 | if (!mp->m_blockgc_wq) |
1058d0f5 | 567 | goto out_destroy_reclaim; |
579b62fa | 568 | |
ab23a776 DC |
569 | mp->m_inodegc_wq = alloc_workqueue("xfs-inodegc/%s", |
570 | XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM), | |
571 | 1, mp->m_super->s_id); | |
572 | if (!mp->m_inodegc_wq) | |
573 | goto out_destroy_blockgc; | |
574 | ||
05a302a1 DW |
575 | mp->m_sync_workqueue = alloc_workqueue("xfs-sync/%s", |
576 | XFS_WQFLAGS(WQ_FREEZABLE), 0, mp->m_super->s_id); | |
696a5620 | 577 | if (!mp->m_sync_workqueue) |
ab23a776 | 578 | goto out_destroy_inodegc; |
696a5620 | 579 | |
aa6bf01d CH |
580 | return 0; |
581 | ||
ab23a776 DC |
582 | out_destroy_inodegc: |
583 | destroy_workqueue(mp->m_inodegc_wq); | |
584 | out_destroy_blockgc: | |
585 | destroy_workqueue(mp->m_blockgc_wq); | |
5889608d DC |
586 | out_destroy_reclaim: |
587 | destroy_workqueue(mp->m_reclaim_workqueue); | |
4c2d542f DC |
588 | out_destroy_unwritten: |
589 | destroy_workqueue(mp->m_unwritten_workqueue); | |
78c931b8 BF |
590 | out_destroy_buf: |
591 | destroy_workqueue(mp->m_buf_workqueue); | |
aa6bf01d CH |
592 | out: |
593 | return -ENOMEM; | |
594 | } | |
595 | ||
596 | STATIC void | |
597 | xfs_destroy_mount_workqueues( | |
598 | struct xfs_mount *mp) | |
599 | { | |
696a5620 | 600 | destroy_workqueue(mp->m_sync_workqueue); |
ab23a776 DC |
601 | destroy_workqueue(mp->m_blockgc_wq); |
602 | destroy_workqueue(mp->m_inodegc_wq); | |
5889608d | 603 | destroy_workqueue(mp->m_reclaim_workqueue); |
aa6bf01d | 604 | destroy_workqueue(mp->m_unwritten_workqueue); |
78c931b8 | 605 | destroy_workqueue(mp->m_buf_workqueue); |
aa6bf01d CH |
606 | } |
607 | ||
f0f7a674 DW |
608 | static void |
609 | xfs_flush_inodes_worker( | |
610 | struct work_struct *work) | |
611 | { | |
612 | struct xfs_mount *mp = container_of(work, struct xfs_mount, | |
613 | m_flush_inodes_work); | |
614 | struct super_block *sb = mp->m_super; | |
615 | ||
616 | if (down_read_trylock(&sb->s_umount)) { | |
617 | sync_inodes_sb(sb); | |
618 | up_read(&sb->s_umount); | |
619 | } | |
620 | } | |
621 | ||
9aa05000 DC |
622 | /* |
623 | * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK | |
624 | * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting | |
625 | * for IO to complete so that we effectively throttle multiple callers to the | |
626 | * rate at which IO is completing. | |
627 | */ | |
628 | void | |
629 | xfs_flush_inodes( | |
630 | struct xfs_mount *mp) | |
631 | { | |
f0f7a674 DW |
632 | /* |
633 | * If flush_work() returns true then that means we waited for a flush | |
634 | * which was already in progress. Don't bother running another scan. | |
635 | */ | |
636 | if (flush_work(&mp->m_flush_inodes_work)) | |
c6425702 DW |
637 | return; |
638 | ||
f0f7a674 DW |
639 | queue_work(mp->m_sync_workqueue, &mp->m_flush_inodes_work); |
640 | flush_work(&mp->m_flush_inodes_work); | |
9aa05000 DC |
641 | } |
642 | ||
bf904248 | 643 | /* Catch misguided souls that try to use this interface on XFS */ |
1da177e4 | 644 | STATIC struct inode * |
a50cd269 | 645 | xfs_fs_alloc_inode( |
1da177e4 LT |
646 | struct super_block *sb) |
647 | { | |
bf904248 | 648 | BUG(); |
493dca61 | 649 | return NULL; |
1da177e4 LT |
650 | } |
651 | ||
bf904248 | 652 | /* |
99fa8cb3 | 653 | * Now that the generic code is guaranteed not to be accessing |
8179c036 | 654 | * the linux inode, we can inactivate and reclaim the inode. |
bf904248 | 655 | */ |
1da177e4 | 656 | STATIC void |
a50cd269 | 657 | xfs_fs_destroy_inode( |
848ce8f7 | 658 | struct inode *inode) |
1da177e4 | 659 | { |
848ce8f7 CH |
660 | struct xfs_inode *ip = XFS_I(inode); |
661 | ||
cca28fb8 | 662 | trace_xfs_destroy_inode(ip); |
99fa8cb3 | 663 | |
65523218 | 664 | ASSERT(!rwsem_is_locked(&inode->i_rwsem)); |
8179c036 DC |
665 | XFS_STATS_INC(ip->i_mount, vn_rele); |
666 | XFS_STATS_INC(ip->i_mount, vn_remove); | |
c076ae7a | 667 | xfs_inode_mark_reclaimable(ip); |
1da177e4 LT |
668 | } |
669 | ||
c3b1b131 CH |
670 | static void |
671 | xfs_fs_dirty_inode( | |
672 | struct inode *inode, | |
cbfecb92 | 673 | int flags) |
c3b1b131 CH |
674 | { |
675 | struct xfs_inode *ip = XFS_I(inode); | |
676 | struct xfs_mount *mp = ip->i_mount; | |
677 | struct xfs_trans *tp; | |
678 | ||
679 | if (!(inode->i_sb->s_flags & SB_LAZYTIME)) | |
680 | return; | |
cbfecb92 LC |
681 | |
682 | /* | |
683 | * Only do the timestamp update if the inode is dirty (I_DIRTY_SYNC) | |
684 | * and has dirty timestamp (I_DIRTY_TIME). I_DIRTY_TIME can be passed | |
685 | * in flags possibly together with I_DIRTY_SYNC. | |
686 | */ | |
687 | if ((flags & ~I_DIRTY_TIME) != I_DIRTY_SYNC || !(flags & I_DIRTY_TIME)) | |
c3b1b131 CH |
688 | return; |
689 | ||
690 | if (xfs_trans_alloc(mp, &M_RES(mp)->tr_fsyncts, 0, 0, 0, &tp)) | |
691 | return; | |
692 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
693 | xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); | |
694 | xfs_trans_log_inode(tp, ip, XFS_ILOG_TIMESTAMP); | |
695 | xfs_trans_commit(tp); | |
696 | } | |
697 | ||
07c8f675 DC |
698 | /* |
699 | * Slab object creation initialisation for the XFS inode. | |
700 | * This covers only the idempotent fields in the XFS inode; | |
701 | * all other fields need to be initialised on allocation | |
b595076a | 702 | * from the slab. This avoids the need to repeatedly initialise |
07c8f675 DC |
703 | * fields in the xfs inode that left in the initialise state |
704 | * when freeing the inode. | |
705 | */ | |
bf904248 DC |
706 | STATIC void |
707 | xfs_fs_inode_init_once( | |
07c8f675 DC |
708 | void *inode) |
709 | { | |
710 | struct xfs_inode *ip = inode; | |
711 | ||
712 | memset(ip, 0, sizeof(struct xfs_inode)); | |
bf904248 DC |
713 | |
714 | /* vfs inode */ | |
715 | inode_init_once(VFS_I(ip)); | |
716 | ||
717 | /* xfs inode */ | |
07c8f675 DC |
718 | atomic_set(&ip->i_pincount, 0); |
719 | spin_lock_init(&ip->i_flags_lock); | |
07c8f675 DC |
720 | |
721 | mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, | |
722 | "xfsino", ip->i_ino); | |
07c8f675 DC |
723 | } |
724 | ||
5132ba8f DC |
725 | /* |
726 | * We do an unlocked check for XFS_IDONTCACHE here because we are already | |
727 | * serialised against cache hits here via the inode->i_lock and igrab() in | |
728 | * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be | |
729 | * racing with us, and it avoids needing to grab a spinlock here for every inode | |
730 | * we drop the final reference on. | |
731 | */ | |
732 | STATIC int | |
733 | xfs_fs_drop_inode( | |
734 | struct inode *inode) | |
735 | { | |
736 | struct xfs_inode *ip = XFS_I(inode); | |
737 | ||
17c12bcd DW |
738 | /* |
739 | * If this unlinked inode is in the middle of recovery, don't | |
740 | * drop the inode just yet; log recovery will take care of | |
741 | * that. See the comment for this inode flag. | |
742 | */ | |
743 | if (ip->i_flags & XFS_IRECOVERY) { | |
e1d06e5f | 744 | ASSERT(xlog_recovery_needed(ip->i_mount->m_log)); |
17c12bcd DW |
745 | return 0; |
746 | } | |
747 | ||
dae2f8ed | 748 | return generic_drop_inode(inode); |
5132ba8f DC |
749 | } |
750 | ||
a943f372 IK |
751 | static void |
752 | xfs_mount_free( | |
a738159d CH |
753 | struct xfs_mount *mp) |
754 | { | |
35a93b14 CH |
755 | /* |
756 | * Free the buftargs here because blkdev_put needs to be called outside | |
757 | * of sb->s_umount, which is held around the call to ->put_super. | |
758 | */ | |
759 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) | |
760 | xfs_free_buftarg(mp->m_logdev_targp); | |
761 | if (mp->m_rtdev_targp) | |
762 | xfs_free_buftarg(mp->m_rtdev_targp); | |
763 | if (mp->m_ddev_targp) | |
764 | xfs_free_buftarg(mp->m_ddev_targp); | |
765 | ||
a76dba3b | 766 | debugfs_remove(mp->m_debugfs); |
a738159d CH |
767 | kfree(mp->m_rtname); |
768 | kfree(mp->m_logname); | |
a943f372 | 769 | kmem_free(mp); |
a738159d CH |
770 | } |
771 | ||
1da177e4 | 772 | STATIC int |
69961a26 | 773 | xfs_fs_sync_fs( |
1da177e4 LT |
774 | struct super_block *sb, |
775 | int wait) | |
776 | { | |
745f6919 | 777 | struct xfs_mount *mp = XFS_M(sb); |
2d86293c | 778 | int error; |
1da177e4 | 779 | |
ab23a776 DC |
780 | trace_xfs_fs_sync_fs(mp, __return_address); |
781 | ||
e893bffd | 782 | /* |
34625c66 | 783 | * Doing anything during the async pass would be counterproductive. |
e893bffd | 784 | */ |
34625c66 | 785 | if (!wait) |
69961a26 | 786 | return 0; |
69961a26 | 787 | |
2d86293c DW |
788 | error = xfs_log_force(mp, XFS_LOG_SYNC); |
789 | if (error) | |
790 | return error; | |
791 | ||
69961a26 | 792 | if (laptop_mode) { |
1da177e4 LT |
793 | /* |
794 | * The disk must be active because we're syncing. | |
f661f1e0 | 795 | * We schedule log work now (now that the disk is |
1da177e4 LT |
796 | * active) instead of later (when it might not be). |
797 | */ | |
f661f1e0 | 798 | flush_delayed_work(&mp->m_log->l_work); |
1da177e4 LT |
799 | } |
800 | ||
ab23a776 DC |
801 | /* |
802 | * If we are called with page faults frozen out, it means we are about | |
803 | * to freeze the transaction subsystem. Take the opportunity to shut | |
804 | * down inodegc because once SB_FREEZE_FS is set it's too late to | |
805 | * prevent inactivation races with freeze. The fs doesn't get called | |
806 | * again by the freezing process until after SB_FREEZE_FS has been set, | |
6f649091 DW |
807 | * so it's now or never. Same logic applies to speculative allocation |
808 | * garbage collection. | |
ab23a776 DC |
809 | * |
810 | * We don't care if this is a normal syncfs call that does this or | |
811 | * freeze that does this - we can run this multiple times without issue | |
812 | * and we won't race with a restart because a restart can only occur | |
813 | * when the state is either SB_FREEZE_FS or SB_FREEZE_COMPLETE. | |
814 | */ | |
6f649091 | 815 | if (sb->s_writers.frozen == SB_FREEZE_PAGEFAULT) { |
ab23a776 | 816 | xfs_inodegc_stop(mp); |
6f649091 DW |
817 | xfs_blockgc_stop(mp); |
818 | } | |
ab23a776 | 819 | |
69961a26 | 820 | return 0; |
1da177e4 LT |
821 | } |
822 | ||
823 | STATIC int | |
a50cd269 | 824 | xfs_fs_statfs( |
726c3342 | 825 | struct dentry *dentry, |
1da177e4 LT |
826 | struct kstatfs *statp) |
827 | { | |
4ca488eb CH |
828 | struct xfs_mount *mp = XFS_M(dentry->d_sb); |
829 | xfs_sb_t *sbp = &mp->m_sb; | |
2b0143b5 | 830 | struct xfs_inode *ip = XFS_I(d_inode(dentry)); |
c8ce540d DW |
831 | uint64_t fakeinos, id; |
832 | uint64_t icount; | |
833 | uint64_t ifree; | |
834 | uint64_t fdblocks; | |
4ca488eb | 835 | xfs_extlen_t lsize; |
c8ce540d | 836 | int64_t ffree; |
4ca488eb | 837 | |
5e672cd6 DC |
838 | /* |
839 | * Expedite background inodegc but don't wait. We do not want to block | |
840 | * here waiting hours for a billion extent file to be truncated. | |
841 | */ | |
842 | xfs_inodegc_push(mp); | |
01e8f379 | 843 | |
dddde68b | 844 | statp->f_type = XFS_SUPER_MAGIC; |
4ca488eb CH |
845 | statp->f_namelen = MAXNAMELEN - 1; |
846 | ||
847 | id = huge_encode_dev(mp->m_ddev_targp->bt_dev); | |
6d1349c7 | 848 | statp->f_fsid = u64_to_fsid(id); |
4ca488eb | 849 | |
501ab323 | 850 | icount = percpu_counter_sum(&mp->m_icount); |
e88b64ea | 851 | ifree = percpu_counter_sum(&mp->m_ifree); |
0d485ada | 852 | fdblocks = percpu_counter_sum(&mp->m_fdblocks); |
4ca488eb CH |
853 | |
854 | spin_lock(&mp->m_sb_lock); | |
855 | statp->f_bsize = sbp->sb_blocksize; | |
856 | lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0; | |
857 | statp->f_blocks = sbp->sb_dblocks - lsize; | |
0d485ada DC |
858 | spin_unlock(&mp->m_sb_lock); |
859 | ||
237aac46 | 860 | /* make sure statp->f_bfree does not underflow */ |
85bcfa26 DW |
861 | statp->f_bfree = max_t(int64_t, 0, |
862 | fdblocks - xfs_fdblocks_unavailable(mp)); | |
0d485ada DC |
863 | statp->f_bavail = statp->f_bfree; |
864 | ||
43004b2a | 865 | fakeinos = XFS_FSB_TO_INO(mp, statp->f_bfree); |
9bb54cb5 | 866 | statp->f_files = min(icount + fakeinos, (uint64_t)XFS_MAXINUMBER); |
ef325959 | 867 | if (M_IGEO(mp)->maxicount) |
a19d9f88 CH |
868 | statp->f_files = min_t(typeof(statp->f_files), |
869 | statp->f_files, | |
ef325959 | 870 | M_IGEO(mp)->maxicount); |
2fe33661 | 871 | |
01f9882e ES |
872 | /* If sb_icount overshot maxicount, report actual allocation */ |
873 | statp->f_files = max_t(typeof(statp->f_files), | |
874 | statp->f_files, | |
875 | sbp->sb_icount); | |
876 | ||
2fe33661 | 877 | /* make sure statp->f_ffree does not underflow */ |
e88b64ea | 878 | ffree = statp->f_files - (icount - ifree); |
c8ce540d | 879 | statp->f_ffree = max_t(int64_t, ffree, 0); |
2fe33661 | 880 | |
4ca488eb | 881 | |
db07349d | 882 | if ((ip->i_diflags & XFS_DIFLAG_PROJINHERIT) && |
83e782e1 CS |
883 | ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) == |
884 | (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD)) | |
7d095257 | 885 | xfs_qm_statvfs(ip, statp); |
a0158315 RW |
886 | |
887 | if (XFS_IS_REALTIME_MOUNT(mp) && | |
db07349d | 888 | (ip->i_diflags & (XFS_DIFLAG_RTINHERIT | XFS_DIFLAG_REALTIME))) { |
2229276c DW |
889 | s64 freertx; |
890 | ||
a0158315 | 891 | statp->f_blocks = sbp->sb_rblocks; |
2229276c DW |
892 | freertx = percpu_counter_sum_positive(&mp->m_frextents); |
893 | statp->f_bavail = statp->f_bfree = freertx * sbp->sb_rextsize; | |
a0158315 RW |
894 | } |
895 | ||
4ca488eb | 896 | return 0; |
1da177e4 LT |
897 | } |
898 | ||
d5db0f97 ES |
899 | STATIC void |
900 | xfs_save_resvblks(struct xfs_mount *mp) | |
901 | { | |
c8ce540d | 902 | uint64_t resblks = 0; |
d5db0f97 ES |
903 | |
904 | mp->m_resblks_save = mp->m_resblks; | |
905 | xfs_reserve_blocks(mp, &resblks, NULL); | |
906 | } | |
907 | ||
908 | STATIC void | |
909 | xfs_restore_resvblks(struct xfs_mount *mp) | |
910 | { | |
c8ce540d | 911 | uint64_t resblks; |
d5db0f97 ES |
912 | |
913 | if (mp->m_resblks_save) { | |
914 | resblks = mp->m_resblks_save; | |
915 | mp->m_resblks_save = 0; | |
916 | } else | |
917 | resblks = xfs_default_resblks(mp); | |
918 | ||
919 | xfs_reserve_blocks(mp, &resblks, NULL); | |
920 | } | |
921 | ||
8757c38f IK |
922 | /* |
923 | * Second stage of a freeze. The data is already frozen so we only | |
924 | * need to take care of the metadata. Once that's done sync the superblock | |
925 | * to the log to dirty it in case of a crash while frozen. This ensures that we | |
926 | * will recover the unlinked inode lists on the next mount. | |
927 | */ | |
928 | STATIC int | |
929 | xfs_fs_freeze( | |
930 | struct super_block *sb) | |
931 | { | |
932 | struct xfs_mount *mp = XFS_M(sb); | |
c3f2375b WL |
933 | unsigned int flags; |
934 | int ret; | |
8757c38f | 935 | |
c3f2375b WL |
936 | /* |
937 | * The filesystem is now frozen far enough that memory reclaim | |
938 | * cannot safely operate on the filesystem. Hence we need to | |
939 | * set a GFP_NOFS context here to avoid recursion deadlocks. | |
940 | */ | |
941 | flags = memalloc_nofs_save(); | |
8757c38f | 942 | xfs_save_resvblks(mp); |
5b0ad7c2 | 943 | ret = xfs_log_quiesce(mp); |
c3f2375b | 944 | memalloc_nofs_restore(flags); |
ab23a776 DC |
945 | |
946 | /* | |
947 | * For read-write filesystems, we need to restart the inodegc on error | |
948 | * because we stopped it at SB_FREEZE_PAGEFAULT level and a thaw is not | |
949 | * going to be run to restart it now. We are at SB_FREEZE_FS level | |
950 | * here, so we can restart safely without racing with a stop in | |
951 | * xfs_fs_sync_fs(). | |
952 | */ | |
2e973b2c | 953 | if (ret && !xfs_is_readonly(mp)) { |
6f649091 | 954 | xfs_blockgc_start(mp); |
ab23a776 | 955 | xfs_inodegc_start(mp); |
6f649091 | 956 | } |
ab23a776 | 957 | |
c3f2375b | 958 | return ret; |
8757c38f IK |
959 | } |
960 | ||
961 | STATIC int | |
962 | xfs_fs_unfreeze( | |
963 | struct super_block *sb) | |
964 | { | |
965 | struct xfs_mount *mp = XFS_M(sb); | |
966 | ||
967 | xfs_restore_resvblks(mp); | |
968 | xfs_log_work_queue(mp); | |
ab23a776 DC |
969 | |
970 | /* | |
971 | * Don't reactivate the inodegc worker on a readonly filesystem because | |
6f649091 DW |
972 | * inodes are sent directly to reclaim. Don't reactivate the blockgc |
973 | * worker because there are no speculative preallocations on a readonly | |
974 | * filesystem. | |
ab23a776 | 975 | */ |
2e973b2c | 976 | if (!xfs_is_readonly(mp)) { |
6f649091 | 977 | xfs_blockgc_start(mp); |
ab23a776 | 978 | xfs_inodegc_start(mp); |
6f649091 | 979 | } |
ab23a776 | 980 | |
8757c38f IK |
981 | return 0; |
982 | } | |
983 | ||
984 | /* | |
985 | * This function fills in xfs_mount_t fields based on mount args. | |
986 | * Note: the superblock _has_ now been read in. | |
987 | */ | |
988 | STATIC int | |
989 | xfs_finish_flags( | |
990 | struct xfs_mount *mp) | |
991 | { | |
8757c38f | 992 | /* Fail a mount where the logbuf is smaller than the log stripe */ |
38c26bfd | 993 | if (xfs_has_logv2(mp)) { |
8757c38f IK |
994 | if (mp->m_logbsize <= 0 && |
995 | mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) { | |
996 | mp->m_logbsize = mp->m_sb.sb_logsunit; | |
997 | } else if (mp->m_logbsize > 0 && | |
998 | mp->m_logbsize < mp->m_sb.sb_logsunit) { | |
999 | xfs_warn(mp, | |
1000 | "logbuf size must be greater than or equal to log stripe size"); | |
1001 | return -EINVAL; | |
1002 | } | |
1003 | } else { | |
1004 | /* Fail a mount if the logbuf is larger than 32K */ | |
1005 | if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) { | |
1006 | xfs_warn(mp, | |
1007 | "logbuf size for version 1 logs must be 16K or 32K"); | |
1008 | return -EINVAL; | |
1009 | } | |
1010 | } | |
1011 | ||
1012 | /* | |
1013 | * V5 filesystems always use attr2 format for attributes. | |
1014 | */ | |
0560f31a | 1015 | if (xfs_has_crc(mp) && xfs_has_noattr2(mp)) { |
8757c38f IK |
1016 | xfs_warn(mp, "Cannot mount a V5 filesystem as noattr2. " |
1017 | "attr2 is always enabled for V5 filesystems."); | |
1018 | return -EINVAL; | |
1019 | } | |
1020 | ||
8757c38f IK |
1021 | /* |
1022 | * prohibit r/w mounts of read-only filesystems | |
1023 | */ | |
2e973b2c | 1024 | if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !xfs_is_readonly(mp)) { |
8757c38f IK |
1025 | xfs_warn(mp, |
1026 | "cannot mount a read-only filesystem as read-write"); | |
1027 | return -EROFS; | |
1028 | } | |
1029 | ||
149e53af CH |
1030 | if ((mp->m_qflags & XFS_GQUOTA_ACCT) && |
1031 | (mp->m_qflags & XFS_PQUOTA_ACCT) && | |
38c26bfd | 1032 | !xfs_has_pquotino(mp)) { |
8757c38f IK |
1033 | xfs_warn(mp, |
1034 | "Super block does not support project and group quota together"); | |
1035 | return -EINVAL; | |
1036 | } | |
1037 | ||
1038 | return 0; | |
1039 | } | |
1040 | ||
1041 | static int | |
1042 | xfs_init_percpu_counters( | |
1043 | struct xfs_mount *mp) | |
1044 | { | |
1045 | int error; | |
1046 | ||
1047 | error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL); | |
1048 | if (error) | |
1049 | return -ENOMEM; | |
1050 | ||
1051 | error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL); | |
1052 | if (error) | |
1053 | goto free_icount; | |
1054 | ||
1055 | error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL); | |
1056 | if (error) | |
1057 | goto free_ifree; | |
1058 | ||
1059 | error = percpu_counter_init(&mp->m_delalloc_blks, 0, GFP_KERNEL); | |
1060 | if (error) | |
1061 | goto free_fdblocks; | |
1062 | ||
2229276c DW |
1063 | error = percpu_counter_init(&mp->m_frextents, 0, GFP_KERNEL); |
1064 | if (error) | |
1065 | goto free_delalloc; | |
1066 | ||
8757c38f IK |
1067 | return 0; |
1068 | ||
2229276c DW |
1069 | free_delalloc: |
1070 | percpu_counter_destroy(&mp->m_delalloc_blks); | |
8757c38f IK |
1071 | free_fdblocks: |
1072 | percpu_counter_destroy(&mp->m_fdblocks); | |
1073 | free_ifree: | |
1074 | percpu_counter_destroy(&mp->m_ifree); | |
1075 | free_icount: | |
1076 | percpu_counter_destroy(&mp->m_icount); | |
1077 | return -ENOMEM; | |
1078 | } | |
1079 | ||
1080 | void | |
1081 | xfs_reinit_percpu_counters( | |
1082 | struct xfs_mount *mp) | |
1083 | { | |
1084 | percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount); | |
1085 | percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree); | |
1086 | percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks); | |
2229276c | 1087 | percpu_counter_set(&mp->m_frextents, mp->m_sb.sb_frextents); |
8757c38f IK |
1088 | } |
1089 | ||
1090 | static void | |
1091 | xfs_destroy_percpu_counters( | |
1092 | struct xfs_mount *mp) | |
1093 | { | |
1094 | percpu_counter_destroy(&mp->m_icount); | |
1095 | percpu_counter_destroy(&mp->m_ifree); | |
1096 | percpu_counter_destroy(&mp->m_fdblocks); | |
75c8c50f | 1097 | ASSERT(xfs_is_shutdown(mp) || |
8757c38f IK |
1098 | percpu_counter_sum(&mp->m_delalloc_blks) == 0); |
1099 | percpu_counter_destroy(&mp->m_delalloc_blks); | |
2229276c | 1100 | percpu_counter_destroy(&mp->m_frextents); |
8757c38f IK |
1101 | } |
1102 | ||
ab23a776 DC |
1103 | static int |
1104 | xfs_inodegc_init_percpu( | |
1105 | struct xfs_mount *mp) | |
1106 | { | |
1107 | struct xfs_inodegc *gc; | |
1108 | int cpu; | |
1109 | ||
1110 | mp->m_inodegc = alloc_percpu(struct xfs_inodegc); | |
1111 | if (!mp->m_inodegc) | |
1112 | return -ENOMEM; | |
1113 | ||
1114 | for_each_possible_cpu(cpu) { | |
1115 | gc = per_cpu_ptr(mp->m_inodegc, cpu); | |
b37c4c83 | 1116 | gc->cpu = cpu; |
62334fab | 1117 | gc->mp = mp; |
ab23a776 DC |
1118 | init_llist_head(&gc->list); |
1119 | gc->items = 0; | |
d4d12c02 | 1120 | gc->error = 0; |
7cf2b0f9 | 1121 | INIT_DELAYED_WORK(&gc->work, xfs_inodegc_worker); |
ab23a776 DC |
1122 | } |
1123 | return 0; | |
1124 | } | |
1125 | ||
1126 | static void | |
1127 | xfs_inodegc_free_percpu( | |
1128 | struct xfs_mount *mp) | |
1129 | { | |
1130 | if (!mp->m_inodegc) | |
1131 | return; | |
1132 | free_percpu(mp->m_inodegc); | |
1133 | } | |
1134 | ||
8757c38f IK |
1135 | static void |
1136 | xfs_fs_put_super( | |
1137 | struct super_block *sb) | |
1138 | { | |
1139 | struct xfs_mount *mp = XFS_M(sb); | |
1140 | ||
64c80dfd | 1141 | xfs_notice(mp, "Unmounting Filesystem %pU", &mp->m_sb.sb_uuid); |
8757c38f IK |
1142 | xfs_filestream_unmount(mp); |
1143 | xfs_unmountfs(mp); | |
1144 | ||
1145 | xfs_freesb(mp); | |
d7a74cad | 1146 | xchk_mount_stats_free(mp); |
8757c38f | 1147 | free_percpu(mp->m_stats.xs_stats); |
ab23a776 | 1148 | xfs_inodegc_free_percpu(mp); |
8757c38f IK |
1149 | xfs_destroy_percpu_counters(mp); |
1150 | xfs_destroy_mount_workqueues(mp); | |
35a93b14 | 1151 | xfs_shutdown_devices(mp); |
8757c38f IK |
1152 | } |
1153 | ||
1154 | static long | |
1155 | xfs_fs_nr_cached_objects( | |
1156 | struct super_block *sb, | |
1157 | struct shrink_control *sc) | |
1158 | { | |
1159 | /* Paranoia: catch incorrect calls during mount setup or teardown */ | |
1160 | if (WARN_ON_ONCE(!sb->s_fs_info)) | |
1161 | return 0; | |
1162 | return xfs_reclaim_inodes_count(XFS_M(sb)); | |
1163 | } | |
1164 | ||
1165 | static long | |
1166 | xfs_fs_free_cached_objects( | |
1167 | struct super_block *sb, | |
1168 | struct shrink_control *sc) | |
c7eea6f7 | 1169 | { |
8757c38f IK |
1170 | return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan); |
1171 | } | |
c7eea6f7 | 1172 | |
e7caa877 CH |
1173 | static void |
1174 | xfs_fs_shutdown( | |
1175 | struct super_block *sb) | |
1176 | { | |
1177 | xfs_force_shutdown(XFS_M(sb), SHUTDOWN_DEVICE_REMOVED); | |
1178 | } | |
1179 | ||
8757c38f IK |
1180 | static const struct super_operations xfs_super_operations = { |
1181 | .alloc_inode = xfs_fs_alloc_inode, | |
1182 | .destroy_inode = xfs_fs_destroy_inode, | |
1183 | .dirty_inode = xfs_fs_dirty_inode, | |
1184 | .drop_inode = xfs_fs_drop_inode, | |
1185 | .put_super = xfs_fs_put_super, | |
1186 | .sync_fs = xfs_fs_sync_fs, | |
1187 | .freeze_fs = xfs_fs_freeze, | |
1188 | .unfreeze_fs = xfs_fs_unfreeze, | |
1189 | .statfs = xfs_fs_statfs, | |
1190 | .show_options = xfs_fs_show_options, | |
1191 | .nr_cached_objects = xfs_fs_nr_cached_objects, | |
1192 | .free_cached_objects = xfs_fs_free_cached_objects, | |
e7caa877 | 1193 | .shutdown = xfs_fs_shutdown, |
8757c38f | 1194 | }; |
c7eea6f7 | 1195 | |
8757c38f IK |
1196 | static int |
1197 | suffix_kstrtoint( | |
1198 | const char *s, | |
1199 | unsigned int base, | |
1200 | int *res) | |
1201 | { | |
1202 | int last, shift_left_factor = 0, _res; | |
1203 | char *value; | |
1204 | int ret = 0; | |
1205 | ||
1206 | value = kstrdup(s, GFP_KERNEL); | |
1207 | if (!value) | |
1208 | return -ENOMEM; | |
1209 | ||
1210 | last = strlen(value) - 1; | |
1211 | if (value[last] == 'K' || value[last] == 'k') { | |
1212 | shift_left_factor = 10; | |
1213 | value[last] = '\0'; | |
1214 | } | |
1215 | if (value[last] == 'M' || value[last] == 'm') { | |
1216 | shift_left_factor = 20; | |
1217 | value[last] = '\0'; | |
1218 | } | |
1219 | if (value[last] == 'G' || value[last] == 'g') { | |
1220 | shift_left_factor = 30; | |
1221 | value[last] = '\0'; | |
1222 | } | |
1223 | ||
1224 | if (kstrtoint(value, base, &_res)) | |
1225 | ret = -EINVAL; | |
1226 | kfree(value); | |
1227 | *res = _res << shift_left_factor; | |
1228 | return ret; | |
c7eea6f7 DC |
1229 | } |
1230 | ||
92cf7d36 PR |
1231 | static inline void |
1232 | xfs_fs_warn_deprecated( | |
1233 | struct fs_context *fc, | |
1234 | struct fs_parameter *param, | |
1235 | uint64_t flag, | |
1236 | bool value) | |
1237 | { | |
1238 | /* Don't print the warning if reconfiguring and current mount point | |
1239 | * already had the flag set | |
1240 | */ | |
1241 | if ((fc->purpose & FS_CONTEXT_FOR_RECONFIGURE) && | |
0560f31a | 1242 | !!(XFS_M(fc->root->d_sb)->m_features & flag) == value) |
92cf7d36 PR |
1243 | return; |
1244 | xfs_warn(fc->s_fs_info, "%s mount option is deprecated.", param->key); | |
1245 | } | |
1246 | ||
9909c4aa | 1247 | /* |
8757c38f IK |
1248 | * Set mount state from a mount option. |
1249 | * | |
1250 | * NOTE: mp->m_super is NULL here! | |
9909c4aa | 1251 | */ |
8757c38f | 1252 | static int |
1e5c39df | 1253 | xfs_fs_parse_param( |
8757c38f IK |
1254 | struct fs_context *fc, |
1255 | struct fs_parameter *param) | |
1da177e4 | 1256 | { |
0f98b4ec | 1257 | struct xfs_mount *parsing_mp = fc->s_fs_info; |
8757c38f IK |
1258 | struct fs_parse_result result; |
1259 | int size = 0; | |
1260 | int opt; | |
9909c4aa | 1261 | |
d7167b14 | 1262 | opt = fs_parse(fc, xfs_fs_parameters, param, &result); |
8757c38f IK |
1263 | if (opt < 0) |
1264 | return opt; | |
1da177e4 | 1265 | |
8757c38f IK |
1266 | switch (opt) { |
1267 | case Opt_logbufs: | |
0f98b4ec | 1268 | parsing_mp->m_logbufs = result.uint_32; |
8757c38f IK |
1269 | return 0; |
1270 | case Opt_logbsize: | |
0f98b4ec | 1271 | if (suffix_kstrtoint(param->string, 10, &parsing_mp->m_logbsize)) |
8757c38f IK |
1272 | return -EINVAL; |
1273 | return 0; | |
1274 | case Opt_logdev: | |
0f98b4ec PR |
1275 | kfree(parsing_mp->m_logname); |
1276 | parsing_mp->m_logname = kstrdup(param->string, GFP_KERNEL); | |
1277 | if (!parsing_mp->m_logname) | |
8757c38f IK |
1278 | return -ENOMEM; |
1279 | return 0; | |
1280 | case Opt_rtdev: | |
0f98b4ec PR |
1281 | kfree(parsing_mp->m_rtname); |
1282 | parsing_mp->m_rtname = kstrdup(param->string, GFP_KERNEL); | |
1283 | if (!parsing_mp->m_rtname) | |
8757c38f IK |
1284 | return -ENOMEM; |
1285 | return 0; | |
1286 | case Opt_allocsize: | |
1287 | if (suffix_kstrtoint(param->string, 10, &size)) | |
1288 | return -EINVAL; | |
0f98b4ec | 1289 | parsing_mp->m_allocsize_log = ffs(size) - 1; |
0560f31a | 1290 | parsing_mp->m_features |= XFS_FEAT_ALLOCSIZE; |
8757c38f IK |
1291 | return 0; |
1292 | case Opt_grpid: | |
1293 | case Opt_bsdgroups: | |
0560f31a | 1294 | parsing_mp->m_features |= XFS_FEAT_GRPID; |
8757c38f IK |
1295 | return 0; |
1296 | case Opt_nogrpid: | |
1297 | case Opt_sysvgroups: | |
0560f31a | 1298 | parsing_mp->m_features &= ~XFS_FEAT_GRPID; |
8757c38f IK |
1299 | return 0; |
1300 | case Opt_wsync: | |
0560f31a | 1301 | parsing_mp->m_features |= XFS_FEAT_WSYNC; |
8757c38f IK |
1302 | return 0; |
1303 | case Opt_norecovery: | |
0560f31a | 1304 | parsing_mp->m_features |= XFS_FEAT_NORECOVERY; |
8757c38f IK |
1305 | return 0; |
1306 | case Opt_noalign: | |
0560f31a | 1307 | parsing_mp->m_features |= XFS_FEAT_NOALIGN; |
8757c38f IK |
1308 | return 0; |
1309 | case Opt_swalloc: | |
0560f31a | 1310 | parsing_mp->m_features |= XFS_FEAT_SWALLOC; |
8757c38f IK |
1311 | return 0; |
1312 | case Opt_sunit: | |
0f98b4ec | 1313 | parsing_mp->m_dalign = result.uint_32; |
8757c38f IK |
1314 | return 0; |
1315 | case Opt_swidth: | |
0f98b4ec | 1316 | parsing_mp->m_swidth = result.uint_32; |
8757c38f IK |
1317 | return 0; |
1318 | case Opt_inode32: | |
0560f31a | 1319 | parsing_mp->m_features |= XFS_FEAT_SMALL_INUMS; |
8757c38f IK |
1320 | return 0; |
1321 | case Opt_inode64: | |
0560f31a | 1322 | parsing_mp->m_features &= ~XFS_FEAT_SMALL_INUMS; |
8757c38f IK |
1323 | return 0; |
1324 | case Opt_nouuid: | |
0560f31a | 1325 | parsing_mp->m_features |= XFS_FEAT_NOUUID; |
8757c38f | 1326 | return 0; |
8757c38f | 1327 | case Opt_largeio: |
0560f31a | 1328 | parsing_mp->m_features |= XFS_FEAT_LARGE_IOSIZE; |
8757c38f IK |
1329 | return 0; |
1330 | case Opt_nolargeio: | |
0560f31a | 1331 | parsing_mp->m_features &= ~XFS_FEAT_LARGE_IOSIZE; |
8757c38f | 1332 | return 0; |
8757c38f | 1333 | case Opt_filestreams: |
0560f31a | 1334 | parsing_mp->m_features |= XFS_FEAT_FILESTREAMS; |
8757c38f IK |
1335 | return 0; |
1336 | case Opt_noquota: | |
0f98b4ec PR |
1337 | parsing_mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT; |
1338 | parsing_mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD; | |
8757c38f IK |
1339 | return 0; |
1340 | case Opt_quota: | |
1341 | case Opt_uquota: | |
1342 | case Opt_usrquota: | |
149e53af | 1343 | parsing_mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ENFD); |
8757c38f IK |
1344 | return 0; |
1345 | case Opt_qnoenforce: | |
1346 | case Opt_uqnoenforce: | |
149e53af | 1347 | parsing_mp->m_qflags |= XFS_UQUOTA_ACCT; |
0f98b4ec | 1348 | parsing_mp->m_qflags &= ~XFS_UQUOTA_ENFD; |
8757c38f IK |
1349 | return 0; |
1350 | case Opt_pquota: | |
1351 | case Opt_prjquota: | |
149e53af | 1352 | parsing_mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ENFD); |
8757c38f IK |
1353 | return 0; |
1354 | case Opt_pqnoenforce: | |
149e53af | 1355 | parsing_mp->m_qflags |= XFS_PQUOTA_ACCT; |
0f98b4ec | 1356 | parsing_mp->m_qflags &= ~XFS_PQUOTA_ENFD; |
8757c38f IK |
1357 | return 0; |
1358 | case Opt_gquota: | |
1359 | case Opt_grpquota: | |
149e53af | 1360 | parsing_mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ENFD); |
8757c38f IK |
1361 | return 0; |
1362 | case Opt_gqnoenforce: | |
149e53af | 1363 | parsing_mp->m_qflags |= XFS_GQUOTA_ACCT; |
0f98b4ec | 1364 | parsing_mp->m_qflags &= ~XFS_GQUOTA_ENFD; |
8757c38f IK |
1365 | return 0; |
1366 | case Opt_discard: | |
0560f31a | 1367 | parsing_mp->m_features |= XFS_FEAT_DISCARD; |
8757c38f IK |
1368 | return 0; |
1369 | case Opt_nodiscard: | |
0560f31a | 1370 | parsing_mp->m_features &= ~XFS_FEAT_DISCARD; |
8757c38f IK |
1371 | return 0; |
1372 | #ifdef CONFIG_FS_DAX | |
1373 | case Opt_dax: | |
0f98b4ec | 1374 | xfs_mount_set_dax_mode(parsing_mp, XFS_DAX_ALWAYS); |
8d6c3446 IW |
1375 | return 0; |
1376 | case Opt_dax_enum: | |
0f98b4ec | 1377 | xfs_mount_set_dax_mode(parsing_mp, result.uint_32); |
8757c38f IK |
1378 | return 0; |
1379 | #endif | |
c23c393e PR |
1380 | /* Following mount options will be removed in September 2025 */ |
1381 | case Opt_ikeep: | |
0560f31a DC |
1382 | xfs_fs_warn_deprecated(fc, param, XFS_FEAT_IKEEP, true); |
1383 | parsing_mp->m_features |= XFS_FEAT_IKEEP; | |
c23c393e PR |
1384 | return 0; |
1385 | case Opt_noikeep: | |
0560f31a DC |
1386 | xfs_fs_warn_deprecated(fc, param, XFS_FEAT_IKEEP, false); |
1387 | parsing_mp->m_features &= ~XFS_FEAT_IKEEP; | |
c23c393e PR |
1388 | return 0; |
1389 | case Opt_attr2: | |
0560f31a DC |
1390 | xfs_fs_warn_deprecated(fc, param, XFS_FEAT_ATTR2, true); |
1391 | parsing_mp->m_features |= XFS_FEAT_ATTR2; | |
c23c393e PR |
1392 | return 0; |
1393 | case Opt_noattr2: | |
0560f31a DC |
1394 | xfs_fs_warn_deprecated(fc, param, XFS_FEAT_NOATTR2, true); |
1395 | parsing_mp->m_features |= XFS_FEAT_NOATTR2; | |
c23c393e | 1396 | return 0; |
8757c38f | 1397 | default: |
0f98b4ec | 1398 | xfs_warn(parsing_mp, "unknown mount option [%s].", param->key); |
8757c38f IK |
1399 | return -EINVAL; |
1400 | } | |
d5db0f97 | 1401 | |
d5db0f97 ES |
1402 | return 0; |
1403 | } | |
1404 | ||
8757c38f | 1405 | static int |
1e5c39df | 1406 | xfs_fs_validate_params( |
f8f15e42 CH |
1407 | struct xfs_mount *mp) |
1408 | { | |
0560f31a | 1409 | /* No recovery flag requires a read-only mount */ |
2e973b2c | 1410 | if (xfs_has_norecovery(mp) && !xfs_is_readonly(mp)) { |
8757c38f | 1411 | xfs_warn(mp, "no-recovery mounts must be read-only."); |
2451337d | 1412 | return -EINVAL; |
d3eaace8 DC |
1413 | } |
1414 | ||
d3eaace8 | 1415 | /* |
0560f31a DC |
1416 | * We have not read the superblock at this point, so only the attr2 |
1417 | * mount option can set the attr2 feature by this stage. | |
d3eaace8 | 1418 | */ |
0560f31a | 1419 | if (xfs_has_attr2(mp) && xfs_has_noattr2(mp)) { |
e23b55d5 | 1420 | xfs_warn(mp, "attr2 and noattr2 cannot both be specified."); |
2451337d | 1421 | return -EINVAL; |
d3eaace8 DC |
1422 | } |
1423 | ||
e23b55d5 | 1424 | |
0560f31a | 1425 | if (xfs_has_noalign(mp) && (mp->m_dalign || mp->m_swidth)) { |
4f10700a | 1426 | xfs_warn(mp, |
8757c38f IK |
1427 | "sunit and swidth options incompatible with the noalign option"); |
1428 | return -EINVAL; | |
f8f15e42 CH |
1429 | } |
1430 | ||
8757c38f IK |
1431 | if (!IS_ENABLED(CONFIG_XFS_QUOTA) && mp->m_qflags != 0) { |
1432 | xfs_warn(mp, "quota support not available in this kernel."); | |
2451337d | 1433 | return -EINVAL; |
d892d586 CS |
1434 | } |
1435 | ||
8757c38f IK |
1436 | if ((mp->m_dalign && !mp->m_swidth) || |
1437 | (!mp->m_dalign && mp->m_swidth)) { | |
1438 | xfs_warn(mp, "sunit and swidth must be specified together"); | |
1439 | return -EINVAL; | |
1440 | } | |
5681ca40 | 1441 | |
8757c38f IK |
1442 | if (mp->m_dalign && (mp->m_swidth % mp->m_dalign != 0)) { |
1443 | xfs_warn(mp, | |
1444 | "stripe width (%d) must be a multiple of the stripe unit (%d)", | |
1445 | mp->m_swidth, mp->m_dalign); | |
1446 | return -EINVAL; | |
1447 | } | |
5681ca40 | 1448 | |
8757c38f IK |
1449 | if (mp->m_logbufs != -1 && |
1450 | mp->m_logbufs != 0 && | |
1451 | (mp->m_logbufs < XLOG_MIN_ICLOGS || | |
1452 | mp->m_logbufs > XLOG_MAX_ICLOGS)) { | |
1453 | xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]", | |
1454 | mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS); | |
1455 | return -EINVAL; | |
1456 | } | |
5681ca40 | 1457 | |
8757c38f IK |
1458 | if (mp->m_logbsize != -1 && |
1459 | mp->m_logbsize != 0 && | |
1460 | (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE || | |
1461 | mp->m_logbsize > XLOG_MAX_RECORD_BSIZE || | |
1462 | !is_power_of_2(mp->m_logbsize))) { | |
1463 | xfs_warn(mp, | |
1464 | "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]", | |
1465 | mp->m_logbsize); | |
1466 | return -EINVAL; | |
1467 | } | |
5681ca40 | 1468 | |
0560f31a | 1469 | if (xfs_has_allocsize(mp) && |
8757c38f IK |
1470 | (mp->m_allocsize_log > XFS_MAX_IO_LOG || |
1471 | mp->m_allocsize_log < XFS_MIN_IO_LOG)) { | |
1472 | xfs_warn(mp, "invalid log iosize: %d [not %d-%d]", | |
1473 | mp->m_allocsize_log, XFS_MIN_IO_LOG, XFS_MAX_IO_LOG); | |
1474 | return -EINVAL; | |
1475 | } | |
9fe82b8c | 1476 | |
5681ca40 | 1477 | return 0; |
72c44e35 | 1478 | } |
74394496 | 1479 | |
a76dba3b DW |
1480 | struct dentry * |
1481 | xfs_debugfs_mkdir( | |
1482 | const char *name, | |
1483 | struct dentry *parent) | |
1484 | { | |
1485 | struct dentry *child; | |
1486 | ||
1487 | /* Apparently we're expected to ignore error returns?? */ | |
1488 | child = debugfs_create_dir(name, parent); | |
1489 | if (IS_ERR(child)) | |
1490 | return NULL; | |
1491 | ||
1492 | return child; | |
1493 | } | |
1494 | ||
73e5fff9 | 1495 | static int |
1e5c39df | 1496 | xfs_fs_fill_super( |
72c44e35 | 1497 | struct super_block *sb, |
73e5fff9 | 1498 | struct fs_context *fc) |
72c44e35 | 1499 | { |
73e5fff9 | 1500 | struct xfs_mount *mp = sb->s_fs_info; |
72c44e35 | 1501 | struct inode *root; |
0279c71f | 1502 | int flags = 0, error; |
72c44e35 | 1503 | |
7c89fcb2 | 1504 | mp->m_super = sb; |
1da177e4 | 1505 | |
1e5c39df | 1506 | error = xfs_fs_validate_params(mp); |
745f6919 | 1507 | if (error) |
2a9311ad | 1508 | return error; |
1da177e4 LT |
1509 | |
1510 | sb_min_blocksize(sb, BBSIZE); | |
0ec58516 | 1511 | sb->s_xattr = xfs_xattr_handlers; |
a50cd269 | 1512 | sb->s_export_op = &xfs_export_operations; |
fcafb71b | 1513 | #ifdef CONFIG_XFS_QUOTA |
a50cd269 | 1514 | sb->s_qcop = &xfs_quotactl_operations; |
17ef4fdd | 1515 | sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; |
fcafb71b | 1516 | #endif |
a50cd269 | 1517 | sb->s_op = &xfs_super_operations; |
1da177e4 | 1518 | |
dae5cd81 DC |
1519 | /* |
1520 | * Delay mount work if the debug hook is set. This is debug | |
1521 | * instrumention to coordinate simulation of xfs mount failures with | |
1522 | * VFS superblock operations | |
1523 | */ | |
1524 | if (xfs_globals.mount_delay) { | |
1525 | xfs_notice(mp, "Delaying mount for %d seconds.", | |
1526 | xfs_globals.mount_delay); | |
1527 | msleep(xfs_globals.mount_delay * 1000); | |
1528 | } | |
1529 | ||
73e5fff9 | 1530 | if (fc->sb_flags & SB_SILENT) |
f8f15e42 CH |
1531 | flags |= XFS_MFSI_QUIET; |
1532 | ||
9d565ffa | 1533 | error = xfs_open_devices(mp); |
19f354d4 | 1534 | if (error) |
2a9311ad | 1535 | return error; |
f8f15e42 | 1536 | |
a76dba3b DW |
1537 | if (xfs_debugfs) { |
1538 | mp->m_debugfs = xfs_debugfs_mkdir(mp->m_super->s_id, | |
1539 | xfs_debugfs); | |
1540 | } else { | |
1541 | mp->m_debugfs = NULL; | |
1542 | } | |
1543 | ||
2451337d | 1544 | error = xfs_init_mount_workqueues(mp); |
61ba35de | 1545 | if (error) |
35a93b14 | 1546 | goto out_shutdown_devices; |
c962fb79 | 1547 | |
5681ca40 | 1548 | error = xfs_init_percpu_counters(mp); |
aa6bf01d CH |
1549 | if (error) |
1550 | goto out_destroy_workqueues; | |
1551 | ||
ab23a776 DC |
1552 | error = xfs_inodegc_init_percpu(mp); |
1553 | if (error) | |
1554 | goto out_destroy_counters; | |
1555 | ||
225e4635 BD |
1556 | /* Allocate stats memory before we do operations that might use it */ |
1557 | mp->m_stats.xs_stats = alloc_percpu(struct xfsstats); | |
1558 | if (!mp->m_stats.xs_stats) { | |
f9d460b3 | 1559 | error = -ENOMEM; |
ab23a776 | 1560 | goto out_destroy_inodegc; |
225e4635 BD |
1561 | } |
1562 | ||
d7a74cad | 1563 | error = xchk_mount_stats_alloc(mp); |
f8f15e42 | 1564 | if (error) |
225e4635 | 1565 | goto out_free_stats; |
9d565ffa | 1566 | |
d7a74cad DW |
1567 | error = xfs_readsb(mp, flags); |
1568 | if (error) | |
1569 | goto out_free_scrub_stats; | |
1570 | ||
9d565ffa | 1571 | error = xfs_finish_flags(mp); |
f8f15e42 | 1572 | if (error) |
effa2eda | 1573 | goto out_free_sb; |
f8f15e42 | 1574 | |
e34b562c | 1575 | error = xfs_setup_devices(mp); |
19f354d4 | 1576 | if (error) |
effa2eda | 1577 | goto out_free_sb; |
f8f15e42 | 1578 | |
b96cb835 | 1579 | /* V4 support is undergoing deprecation. */ |
38c26bfd | 1580 | if (!xfs_has_crc(mp)) { |
b96cb835 DW |
1581 | #ifdef CONFIG_XFS_SUPPORT_V4 |
1582 | xfs_warn_once(mp, | |
1583 | "Deprecated V4 format (crc=0) will not be supported after September 2030."); | |
1584 | #else | |
1585 | xfs_warn(mp, | |
1586 | "Deprecated V4 format (crc=0) not supported by kernel."); | |
1587 | error = -EINVAL; | |
1588 | goto out_free_sb; | |
1589 | #endif | |
1590 | } | |
1591 | ||
7ba83850 DW |
1592 | /* ASCII case insensitivity is undergoing deprecation. */ |
1593 | if (xfs_has_asciici(mp)) { | |
1594 | #ifdef CONFIG_XFS_SUPPORT_ASCII_CI | |
1595 | xfs_warn_once(mp, | |
1596 | "Deprecated ASCII case-insensitivity feature (ascii-ci=1) will not be supported after September 2030."); | |
1597 | #else | |
1598 | xfs_warn(mp, | |
1599 | "Deprecated ASCII case-insensitivity feature (ascii-ci=1) not supported by kernel."); | |
1600 | error = -EINVAL; | |
1601 | goto out_free_sb; | |
1602 | #endif | |
1603 | } | |
1604 | ||
80c720b8 | 1605 | /* Filesystem claims it needs repair, so refuse the mount. */ |
ebd9027d | 1606 | if (xfs_has_needsrepair(mp)) { |
80c720b8 DW |
1607 | xfs_warn(mp, "Filesystem needs repair. Please run xfs_repair."); |
1608 | error = -EFSCORRUPTED; | |
1609 | goto out_free_sb; | |
1610 | } | |
1611 | ||
3945ae03 DW |
1612 | /* |
1613 | * Don't touch the filesystem if a user tool thinks it owns the primary | |
1614 | * superblock. mkfs doesn't clear the flag from secondary supers, so | |
1615 | * we don't check them at all. | |
1616 | */ | |
1617 | if (mp->m_sb.sb_inprogress) { | |
1618 | xfs_warn(mp, "Offline file system operation in progress!"); | |
1619 | error = -EFSCORRUPTED; | |
1620 | goto out_free_sb; | |
1621 | } | |
1622 | ||
1623 | /* | |
1624 | * Until this is fixed only page-sized or smaller data blocks work. | |
1625 | */ | |
1626 | if (mp->m_sb.sb_blocksize > PAGE_SIZE) { | |
1627 | xfs_warn(mp, | |
1628 | "File system with blocksize %d bytes. " | |
1629 | "Only pagesize (%ld) or less will currently work.", | |
1630 | mp->m_sb.sb_blocksize, PAGE_SIZE); | |
1631 | error = -ENOSYS; | |
1632 | goto out_free_sb; | |
1633 | } | |
1634 | ||
1635 | /* Ensure this filesystem fits in the page cache limits */ | |
1636 | if (xfs_sb_validate_fsb_count(&mp->m_sb, mp->m_sb.sb_dblocks) || | |
1637 | xfs_sb_validate_fsb_count(&mp->m_sb, mp->m_sb.sb_rblocks)) { | |
1638 | xfs_warn(mp, | |
1639 | "file system too large to be mounted on this system."); | |
1640 | error = -EFBIG; | |
1641 | goto out_free_sb; | |
1642 | } | |
1643 | ||
932befe3 DW |
1644 | /* |
1645 | * XFS block mappings use 54 bits to store the logical block offset. | |
1646 | * This should suffice to handle the maximum file size that the VFS | |
1647 | * supports (currently 2^63 bytes on 64-bit and ULONG_MAX << PAGE_SHIFT | |
1648 | * bytes on 32-bit), but as XFS and VFS have gotten the s_maxbytes | |
1649 | * calculation wrong on 32-bit kernels in the past, we'll add a WARN_ON | |
1650 | * to check this assertion. | |
1651 | * | |
1652 | * Avoid integer overflow by comparing the maximum bmbt offset to the | |
1653 | * maximum pagecache offset in units of fs blocks. | |
1654 | */ | |
33005fd0 | 1655 | if (!xfs_verify_fileoff(mp, XFS_B_TO_FSBT(mp, MAX_LFS_FILESIZE))) { |
932befe3 DW |
1656 | xfs_warn(mp, |
1657 | "MAX_LFS_FILESIZE block offset (%llu) exceeds extent map maximum (%llu)!", | |
1658 | XFS_B_TO_FSBT(mp, MAX_LFS_FILESIZE), | |
1659 | XFS_MAX_FILEOFF); | |
1660 | error = -EINVAL; | |
1661 | goto out_free_sb; | |
1662 | } | |
1663 | ||
f8f15e42 CH |
1664 | error = xfs_filestream_mount(mp); |
1665 | if (error) | |
effa2eda | 1666 | goto out_free_sb; |
f8f15e42 | 1667 | |
704b2907 DC |
1668 | /* |
1669 | * we must configure the block size in the superblock before we run the | |
1670 | * full mount process as the mount process can lookup and cache inodes. | |
704b2907 | 1671 | */ |
dddde68b | 1672 | sb->s_magic = XFS_SUPER_MAGIC; |
4ca488eb CH |
1673 | sb->s_blocksize = mp->m_sb.sb_blocksize; |
1674 | sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1; | |
932befe3 | 1675 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
8de52778 | 1676 | sb->s_max_links = XFS_MAXLINK; |
1da177e4 | 1677 | sb->s_time_gran = 1; |
38c26bfd | 1678 | if (xfs_has_bigtime(mp)) { |
f93e5436 DW |
1679 | sb->s_time_min = xfs_bigtime_to_unix(XFS_BIGTIME_TIME_MIN); |
1680 | sb->s_time_max = xfs_bigtime_to_unix(XFS_BIGTIME_TIME_MAX); | |
1681 | } else { | |
1682 | sb->s_time_min = XFS_LEGACY_TIME_MIN; | |
1683 | sb->s_time_max = XFS_LEGACY_TIME_MAX; | |
1684 | } | |
06dbf82b | 1685 | trace_xfs_inode_timestamp_range(mp, sb->s_time_min, sb->s_time_max); |
adfb5fb4 CH |
1686 | sb->s_iflags |= SB_I_CGROUPWB; |
1687 | ||
1da177e4 LT |
1688 | set_posix_acl_flag(sb); |
1689 | ||
dc037ad7 | 1690 | /* version 5 superblocks support inode version counters. */ |
d6837c1a | 1691 | if (xfs_has_crc(mp)) |
357fdad0 | 1692 | sb->s_flags |= SB_I_VERSION; |
dc037ad7 | 1693 | |
0560f31a | 1694 | if (xfs_has_dax_always(mp)) { |
679a9949 CH |
1695 | error = xfs_setup_dax_always(mp); |
1696 | if (error) | |
b6e03c10 | 1697 | goto out_filestream_unmount; |
cbe4dab1 DC |
1698 | } |
1699 | ||
70200574 CH |
1700 | if (xfs_has_discard(mp) && !bdev_max_discard_sectors(sb->s_bdev)) { |
1701 | xfs_warn(mp, | |
1702 | "mounting with \"discard\" option, but the device does not support discard"); | |
1703 | mp->m_features &= ~XFS_FEAT_DISCARD; | |
1e6fa688 KN |
1704 | } |
1705 | ||
38c26bfd | 1706 | if (xfs_has_reflink(mp)) { |
66ae56a5 CH |
1707 | if (mp->m_sb.sb_rblocks) { |
1708 | xfs_alert(mp, | |
c14632dd | 1709 | "reflink not compatible with realtime device!"); |
66ae56a5 CH |
1710 | error = -EINVAL; |
1711 | goto out_filestream_unmount; | |
1712 | } | |
1713 | ||
1714 | if (xfs_globals.always_cow) { | |
1715 | xfs_info(mp, "using DEBUG-only always_cow mode."); | |
1716 | mp->m_always_cow = true; | |
1717 | } | |
c14632dd DW |
1718 | } |
1719 | ||
38c26bfd | 1720 | if (xfs_has_rmapbt(mp) && mp->m_sb.sb_rblocks) { |
1c0607ac | 1721 | xfs_alert(mp, |
76883f79 DW |
1722 | "reverse mapping btree not compatible with realtime device!"); |
1723 | error = -EINVAL; | |
1724 | goto out_filestream_unmount; | |
738f57c1 | 1725 | } |
1c0607ac | 1726 | |
8a00ebe4 | 1727 | error = xfs_mountfs(mp); |
2bcf6e97 | 1728 | if (error) |
7e18530b | 1729 | goto out_filestream_unmount; |
704b2907 | 1730 | |
01651646 | 1731 | root = igrab(VFS_I(mp->m_rootip)); |
f3dcc13f | 1732 | if (!root) { |
2451337d | 1733 | error = -ENOENT; |
8a00ebe4 | 1734 | goto out_unmount; |
cbc89dcf | 1735 | } |
48fde701 | 1736 | sb->s_root = d_make_root(root); |
f3dcc13f | 1737 | if (!sb->s_root) { |
2451337d | 1738 | error = -ENOMEM; |
8a00ebe4 | 1739 | goto out_unmount; |
1da177e4 | 1740 | } |
74394496 | 1741 | |
1da177e4 | 1742 | return 0; |
33c7a2bc | 1743 | |
7e18530b | 1744 | out_filestream_unmount: |
120226c1 | 1745 | xfs_filestream_unmount(mp); |
effa2eda CH |
1746 | out_free_sb: |
1747 | xfs_freesb(mp); | |
d7a74cad DW |
1748 | out_free_scrub_stats: |
1749 | xchk_mount_stats_free(mp); | |
225e4635 BD |
1750 | out_free_stats: |
1751 | free_percpu(mp->m_stats.xs_stats); | |
ab23a776 | 1752 | out_destroy_inodegc: |
ab23a776 | 1753 | xfs_inodegc_free_percpu(mp); |
9d565ffa | 1754 | out_destroy_counters: |
5681ca40 | 1755 | xfs_destroy_percpu_counters(mp); |
225e4635 | 1756 | out_destroy_workqueues: |
aa6bf01d | 1757 | xfs_destroy_mount_workqueues(mp); |
35a93b14 CH |
1758 | out_shutdown_devices: |
1759 | xfs_shutdown_devices(mp); | |
2451337d | 1760 | return error; |
f8f15e42 | 1761 | |
2bcf6e97 | 1762 | out_unmount: |
e48ad316 | 1763 | xfs_filestream_unmount(mp); |
19f354d4 | 1764 | xfs_unmountfs(mp); |
6203300e | 1765 | goto out_free_sb; |
1da177e4 LT |
1766 | } |
1767 | ||
73e5fff9 | 1768 | static int |
1e5c39df | 1769 | xfs_fs_get_tree( |
73e5fff9 IK |
1770 | struct fs_context *fc) |
1771 | { | |
1e5c39df | 1772 | return get_tree_bdev(fc, xfs_fs_fill_super); |
73e5fff9 IK |
1773 | } |
1774 | ||
63cd1e9b IK |
1775 | static int |
1776 | xfs_remount_rw( | |
1777 | struct xfs_mount *mp) | |
1778 | { | |
1779 | struct xfs_sb *sbp = &mp->m_sb; | |
1780 | int error; | |
1781 | ||
0560f31a | 1782 | if (xfs_has_norecovery(mp)) { |
63cd1e9b IK |
1783 | xfs_warn(mp, |
1784 | "ro->rw transition prohibited on norecovery mount"); | |
1785 | return -EINVAL; | |
1786 | } | |
1787 | ||
d6837c1a | 1788 | if (xfs_sb_is_v5(sbp) && |
63cd1e9b IK |
1789 | xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) { |
1790 | xfs_warn(mp, | |
1791 | "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem", | |
1792 | (sbp->sb_features_ro_compat & | |
1793 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)); | |
1794 | return -EINVAL; | |
1795 | } | |
1796 | ||
2e973b2c | 1797 | clear_bit(XFS_OPSTATE_READONLY, &mp->m_opstate); |
63cd1e9b IK |
1798 | |
1799 | /* | |
1800 | * If this is the first remount to writeable state we might have some | |
1801 | * superblock changes to update. | |
1802 | */ | |
1803 | if (mp->m_update_sb) { | |
1804 | error = xfs_sync_sb(mp, false); | |
1805 | if (error) { | |
1806 | xfs_warn(mp, "failed to write sb changes"); | |
1807 | return error; | |
1808 | } | |
1809 | mp->m_update_sb = false; | |
1810 | } | |
1811 | ||
1812 | /* | |
1813 | * Fill out the reserve pool if it is empty. Use the stashed value if | |
1814 | * it is non-zero, otherwise go with the default. | |
1815 | */ | |
1816 | xfs_restore_resvblks(mp); | |
1817 | xfs_log_work_queue(mp); | |
c9a6526f | 1818 | xfs_blockgc_start(mp); |
63cd1e9b IK |
1819 | |
1820 | /* Create the per-AG metadata reservation pool .*/ | |
1821 | error = xfs_fs_reserve_ag_blocks(mp); | |
1822 | if (error && error != -ENOSPC) | |
1823 | return error; | |
1824 | ||
ab23a776 DC |
1825 | /* Re-enable the background inode inactivation worker. */ |
1826 | xfs_inodegc_start(mp); | |
1827 | ||
63cd1e9b IK |
1828 | return 0; |
1829 | } | |
1830 | ||
1831 | static int | |
1832 | xfs_remount_ro( | |
1833 | struct xfs_mount *mp) | |
1834 | { | |
089558bc DW |
1835 | struct xfs_icwalk icw = { |
1836 | .icw_flags = XFS_ICWALK_FLAG_SYNC, | |
1837 | }; | |
1838 | int error; | |
63cd1e9b | 1839 | |
b97cca3b DW |
1840 | /* Flush all the dirty data to disk. */ |
1841 | error = sync_filesystem(mp->m_super); | |
1842 | if (error) | |
1843 | return error; | |
1844 | ||
63cd1e9b IK |
1845 | /* |
1846 | * Cancel background eofb scanning so it cannot race with the final | |
1847 | * log force+buftarg wait and deadlock the remount. | |
1848 | */ | |
c9a6526f | 1849 | xfs_blockgc_stop(mp); |
63cd1e9b | 1850 | |
089558bc DW |
1851 | /* |
1852 | * Clear out all remaining COW staging extents and speculative post-EOF | |
1853 | * preallocations so that we don't leave inodes requiring inactivation | |
1854 | * cleanups during reclaim on a read-only mount. We must process every | |
1855 | * cached inode, so this requires a synchronous cache scan. | |
1856 | */ | |
1857 | error = xfs_blockgc_free_space(mp, &icw); | |
63cd1e9b IK |
1858 | if (error) { |
1859 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); | |
1860 | return error; | |
1861 | } | |
1862 | ||
ab23a776 DC |
1863 | /* |
1864 | * Stop the inodegc background worker. xfs_fs_reconfigure already | |
1865 | * flushed all pending inodegc work when it sync'd the filesystem. | |
1866 | * The VFS holds s_umount, so we know that inodes cannot enter | |
1867 | * xfs_fs_destroy_inode during a remount operation. In readonly mode | |
1868 | * we send inodes straight to reclaim, so no inodes will be queued. | |
1869 | */ | |
1870 | xfs_inodegc_stop(mp); | |
1871 | ||
63cd1e9b IK |
1872 | /* Free the per-AG metadata reservation pool. */ |
1873 | error = xfs_fs_unreserve_ag_blocks(mp); | |
1874 | if (error) { | |
1875 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); | |
1876 | return error; | |
1877 | } | |
1878 | ||
1879 | /* | |
1880 | * Before we sync the metadata, we need to free up the reserve block | |
1881 | * pool so that the used block count in the superblock on disk is | |
1882 | * correct at the end of the remount. Stash the current* reserve pool | |
1883 | * size so that if we get remounted rw, we can return it to the same | |
1884 | * size. | |
1885 | */ | |
1886 | xfs_save_resvblks(mp); | |
1887 | ||
ea2064da | 1888 | xfs_log_clean(mp); |
2e973b2c | 1889 | set_bit(XFS_OPSTATE_READONLY, &mp->m_opstate); |
63cd1e9b IK |
1890 | |
1891 | return 0; | |
1892 | } | |
1893 | ||
1894 | /* | |
1895 | * Logically we would return an error here to prevent users from believing | |
1896 | * they might have changed mount options using remount which can't be changed. | |
1897 | * | |
1898 | * But unfortunately mount(8) adds all options from mtab and fstab to the mount | |
1899 | * arguments in some cases so we can't blindly reject options, but have to | |
1900 | * check for each specified option if it actually differs from the currently | |
1901 | * set option and only reject it if that's the case. | |
1902 | * | |
1903 | * Until that is implemented we return success for every remount request, and | |
1904 | * silently ignore all options that we can't actually change. | |
1905 | */ | |
1906 | static int | |
1e5c39df | 1907 | xfs_fs_reconfigure( |
63cd1e9b IK |
1908 | struct fs_context *fc) |
1909 | { | |
1910 | struct xfs_mount *mp = XFS_M(fc->root->d_sb); | |
1911 | struct xfs_mount *new_mp = fc->s_fs_info; | |
63cd1e9b IK |
1912 | int flags = fc->sb_flags; |
1913 | int error; | |
1914 | ||
4750a171 | 1915 | /* version 5 superblocks always support version counters. */ |
d6837c1a | 1916 | if (xfs_has_crc(mp)) |
4750a171 ES |
1917 | fc->sb_flags |= SB_I_VERSION; |
1918 | ||
1e5c39df | 1919 | error = xfs_fs_validate_params(new_mp); |
63cd1e9b IK |
1920 | if (error) |
1921 | return error; | |
1922 | ||
63cd1e9b | 1923 | /* inode32 -> inode64 */ |
0560f31a DC |
1924 | if (xfs_has_small_inums(mp) && !xfs_has_small_inums(new_mp)) { |
1925 | mp->m_features &= ~XFS_FEAT_SMALL_INUMS; | |
d6837c1a | 1926 | mp->m_maxagi = xfs_set_inode_alloc(mp, mp->m_sb.sb_agcount); |
63cd1e9b IK |
1927 | } |
1928 | ||
1929 | /* inode64 -> inode32 */ | |
0560f31a DC |
1930 | if (!xfs_has_small_inums(mp) && xfs_has_small_inums(new_mp)) { |
1931 | mp->m_features |= XFS_FEAT_SMALL_INUMS; | |
d6837c1a | 1932 | mp->m_maxagi = xfs_set_inode_alloc(mp, mp->m_sb.sb_agcount); |
63cd1e9b IK |
1933 | } |
1934 | ||
1935 | /* ro -> rw */ | |
2e973b2c | 1936 | if (xfs_is_readonly(mp) && !(flags & SB_RDONLY)) { |
63cd1e9b IK |
1937 | error = xfs_remount_rw(mp); |
1938 | if (error) | |
1939 | return error; | |
1940 | } | |
1941 | ||
1942 | /* rw -> ro */ | |
2e973b2c | 1943 | if (!xfs_is_readonly(mp) && (flags & SB_RDONLY)) { |
63cd1e9b IK |
1944 | error = xfs_remount_ro(mp); |
1945 | if (error) | |
1946 | return error; | |
1947 | } | |
1948 | ||
1949 | return 0; | |
1950 | } | |
1951 | ||
dbbff489 CH |
1952 | static void |
1953 | xfs_fs_free( | |
73e5fff9 IK |
1954 | struct fs_context *fc) |
1955 | { | |
1956 | struct xfs_mount *mp = fc->s_fs_info; | |
1957 | ||
1958 | /* | |
1959 | * mp is stored in the fs_context when it is initialized. | |
1960 | * mp is transferred to the superblock on a successful mount, | |
1961 | * but if an error occurs before the transfer we have to free | |
1962 | * it here. | |
1963 | */ | |
1964 | if (mp) | |
1965 | xfs_mount_free(mp); | |
1966 | } | |
1967 | ||
1968 | static const struct fs_context_operations xfs_context_ops = { | |
1e5c39df DW |
1969 | .parse_param = xfs_fs_parse_param, |
1970 | .get_tree = xfs_fs_get_tree, | |
1971 | .reconfigure = xfs_fs_reconfigure, | |
1972 | .free = xfs_fs_free, | |
73e5fff9 IK |
1973 | }; |
1974 | ||
1975 | static int xfs_init_fs_context( | |
1976 | struct fs_context *fc) | |
1977 | { | |
1978 | struct xfs_mount *mp; | |
1979 | ||
50f83009 | 1980 | mp = kmem_alloc(sizeof(struct xfs_mount), KM_ZERO); |
73e5fff9 IK |
1981 | if (!mp) |
1982 | return -ENOMEM; | |
1983 | ||
50f83009 | 1984 | spin_lock_init(&mp->m_sb_lock); |
50f83009 IK |
1985 | INIT_RADIX_TREE(&mp->m_perag_tree, GFP_ATOMIC); |
1986 | spin_lock_init(&mp->m_perag_lock); | |
1987 | mutex_init(&mp->m_growlock); | |
f0f7a674 | 1988 | INIT_WORK(&mp->m_flush_inodes_work, xfs_flush_inodes_worker); |
50f83009 | 1989 | INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker); |
50f83009 IK |
1990 | mp->m_kobj.kobject.kset = xfs_kset; |
1991 | /* | |
1992 | * We don't create the finobt per-ag space reservation until after log | |
1993 | * recovery, so we must set this to true so that an ifree transaction | |
1994 | * started during log recovery will not depend on space reservations | |
1995 | * for finobt expansion. | |
1996 | */ | |
1997 | mp->m_finobt_nores = true; | |
1998 | ||
73e5fff9 IK |
1999 | /* |
2000 | * These can be overridden by the mount option parsing. | |
2001 | */ | |
2002 | mp->m_logbufs = -1; | |
2003 | mp->m_logbsize = -1; | |
2004 | mp->m_allocsize_log = 16; /* 64k */ | |
2005 | ||
2006 | /* | |
2007 | * Copy binary VFS mount flags we are interested in. | |
2008 | */ | |
2009 | if (fc->sb_flags & SB_RDONLY) | |
2e973b2c | 2010 | set_bit(XFS_OPSTATE_READONLY, &mp->m_opstate); |
73e5fff9 | 2011 | if (fc->sb_flags & SB_DIRSYNC) |
0560f31a | 2012 | mp->m_features |= XFS_FEAT_DIRSYNC; |
73e5fff9 | 2013 | if (fc->sb_flags & SB_SYNCHRONOUS) |
0560f31a | 2014 | mp->m_features |= XFS_FEAT_WSYNC; |
73e5fff9 IK |
2015 | |
2016 | fc->s_fs_info = mp; | |
2017 | fc->ops = &xfs_context_ops; | |
2018 | ||
2019 | return 0; | |
2020 | } | |
2021 | ||
2a9311ad CH |
2022 | static void |
2023 | xfs_kill_sb( | |
2024 | struct super_block *sb) | |
2025 | { | |
2026 | kill_block_super(sb); | |
2027 | xfs_mount_free(XFS_M(sb)); | |
2028 | } | |
2029 | ||
5085b607 | 2030 | static struct file_system_type xfs_fs_type = { |
1da177e4 LT |
2031 | .owner = THIS_MODULE, |
2032 | .name = "xfs", | |
73e5fff9 | 2033 | .init_fs_context = xfs_init_fs_context, |
d7167b14 | 2034 | .parameters = xfs_fs_parameters, |
2a9311ad | 2035 | .kill_sb = xfs_kill_sb, |
f798accd | 2036 | .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP, |
1da177e4 | 2037 | }; |
7f78e035 | 2038 | MODULE_ALIAS_FS("xfs"); |
1da177e4 | 2039 | |
9f8868ff | 2040 | STATIC int __init |
182696fb | 2041 | xfs_init_caches(void) |
9f8868ff | 2042 | { |
9fa47bdc DW |
2043 | int error; |
2044 | ||
231f91ab DC |
2045 | xfs_buf_cache = kmem_cache_create("xfs_buf", sizeof(struct xfs_buf), 0, |
2046 | SLAB_HWCACHE_ALIGN | | |
2047 | SLAB_RECLAIM_ACCOUNT | | |
2048 | SLAB_MEM_SPREAD, | |
2049 | NULL); | |
2050 | if (!xfs_buf_cache) | |
2051 | goto out; | |
2052 | ||
182696fb | 2053 | xfs_log_ticket_cache = kmem_cache_create("xfs_log_ticket", |
b1231760 CM |
2054 | sizeof(struct xlog_ticket), |
2055 | 0, 0, NULL); | |
182696fb | 2056 | if (!xfs_log_ticket_cache) |
231f91ab | 2057 | goto out_destroy_buf_cache; |
9f8868ff | 2058 | |
9fa47bdc DW |
2059 | error = xfs_btree_init_cur_caches(); |
2060 | if (error) | |
c201d9ca | 2061 | goto out_destroy_log_ticket_cache; |
9f8868ff | 2062 | |
f3c799c2 DW |
2063 | error = xfs_defer_init_item_caches(); |
2064 | if (error) | |
2065 | goto out_destroy_btree_cur_cache; | |
2066 | ||
182696fb | 2067 | xfs_da_state_cache = kmem_cache_create("xfs_da_state", |
b1231760 CM |
2068 | sizeof(struct xfs_da_state), |
2069 | 0, 0, NULL); | |
182696fb | 2070 | if (!xfs_da_state_cache) |
f3c799c2 | 2071 | goto out_destroy_defer_item_cache; |
9f8868ff | 2072 | |
182696fb | 2073 | xfs_ifork_cache = kmem_cache_create("xfs_ifork", |
b1231760 CM |
2074 | sizeof(struct xfs_ifork), |
2075 | 0, 0, NULL); | |
182696fb DW |
2076 | if (!xfs_ifork_cache) |
2077 | goto out_destroy_da_state_cache; | |
9f8868ff | 2078 | |
182696fb | 2079 | xfs_trans_cache = kmem_cache_create("xfs_trans", |
b1231760 CM |
2080 | sizeof(struct xfs_trans), |
2081 | 0, 0, NULL); | |
182696fb DW |
2082 | if (!xfs_trans_cache) |
2083 | goto out_destroy_ifork_cache; | |
9f8868ff | 2084 | |
e98c414f | 2085 | |
9f8868ff | 2086 | /* |
182696fb | 2087 | * The size of the cache-allocated buf log item is the maximum |
9f8868ff CH |
2088 | * size possible under XFS. This wastes a little bit of memory, |
2089 | * but it is much faster. | |
2090 | */ | |
182696fb | 2091 | xfs_buf_item_cache = kmem_cache_create("xfs_buf_item", |
b1231760 CM |
2092 | sizeof(struct xfs_buf_log_item), |
2093 | 0, 0, NULL); | |
182696fb DW |
2094 | if (!xfs_buf_item_cache) |
2095 | goto out_destroy_trans_cache; | |
9f8868ff | 2096 | |
182696fb | 2097 | xfs_efd_cache = kmem_cache_create("xfs_efd_item", |
3c5aaace DW |
2098 | xfs_efd_log_item_sizeof(XFS_EFD_MAX_FAST_EXTENTS), |
2099 | 0, 0, NULL); | |
182696fb DW |
2100 | if (!xfs_efd_cache) |
2101 | goto out_destroy_buf_item_cache; | |
9f8868ff | 2102 | |
182696fb | 2103 | xfs_efi_cache = kmem_cache_create("xfs_efi_item", |
3c5aaace DW |
2104 | xfs_efi_log_item_sizeof(XFS_EFI_MAX_FAST_EXTENTS), |
2105 | 0, 0, NULL); | |
182696fb DW |
2106 | if (!xfs_efi_cache) |
2107 | goto out_destroy_efd_cache; | |
9f8868ff | 2108 | |
182696fb | 2109 | xfs_inode_cache = kmem_cache_create("xfs_inode", |
b1231760 CM |
2110 | sizeof(struct xfs_inode), 0, |
2111 | (SLAB_HWCACHE_ALIGN | | |
2112 | SLAB_RECLAIM_ACCOUNT | | |
2113 | SLAB_MEM_SPREAD | SLAB_ACCOUNT), | |
2114 | xfs_fs_inode_init_once); | |
182696fb DW |
2115 | if (!xfs_inode_cache) |
2116 | goto out_destroy_efi_cache; | |
9f8868ff | 2117 | |
182696fb | 2118 | xfs_ili_cache = kmem_cache_create("xfs_ili", |
b1231760 | 2119 | sizeof(struct xfs_inode_log_item), 0, |
d59eadae DC |
2120 | SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, |
2121 | NULL); | |
182696fb DW |
2122 | if (!xfs_ili_cache) |
2123 | goto out_destroy_inode_cache; | |
b1231760 | 2124 | |
182696fb | 2125 | xfs_icreate_cache = kmem_cache_create("xfs_icr", |
b1231760 CM |
2126 | sizeof(struct xfs_icreate_item), |
2127 | 0, 0, NULL); | |
182696fb DW |
2128 | if (!xfs_icreate_cache) |
2129 | goto out_destroy_ili_cache; | |
9f8868ff | 2130 | |
182696fb | 2131 | xfs_rud_cache = kmem_cache_create("xfs_rud_item", |
b1231760 CM |
2132 | sizeof(struct xfs_rud_log_item), |
2133 | 0, 0, NULL); | |
182696fb DW |
2134 | if (!xfs_rud_cache) |
2135 | goto out_destroy_icreate_cache; | |
5880f2d7 | 2136 | |
182696fb | 2137 | xfs_rui_cache = kmem_cache_create("xfs_rui_item", |
cd00158c | 2138 | xfs_rui_log_item_sizeof(XFS_RUI_MAX_FAST_EXTENTS), |
b1231760 | 2139 | 0, 0, NULL); |
182696fb DW |
2140 | if (!xfs_rui_cache) |
2141 | goto out_destroy_rud_cache; | |
5880f2d7 | 2142 | |
182696fb | 2143 | xfs_cud_cache = kmem_cache_create("xfs_cud_item", |
b1231760 CM |
2144 | sizeof(struct xfs_cud_log_item), |
2145 | 0, 0, NULL); | |
182696fb DW |
2146 | if (!xfs_cud_cache) |
2147 | goto out_destroy_rui_cache; | |
baf4bcac | 2148 | |
182696fb | 2149 | xfs_cui_cache = kmem_cache_create("xfs_cui_item", |
baf4bcac | 2150 | xfs_cui_log_item_sizeof(XFS_CUI_MAX_FAST_EXTENTS), |
b1231760 | 2151 | 0, 0, NULL); |
182696fb DW |
2152 | if (!xfs_cui_cache) |
2153 | goto out_destroy_cud_cache; | |
baf4bcac | 2154 | |
182696fb | 2155 | xfs_bud_cache = kmem_cache_create("xfs_bud_item", |
b1231760 CM |
2156 | sizeof(struct xfs_bud_log_item), |
2157 | 0, 0, NULL); | |
182696fb DW |
2158 | if (!xfs_bud_cache) |
2159 | goto out_destroy_cui_cache; | |
6413a014 | 2160 | |
182696fb | 2161 | xfs_bui_cache = kmem_cache_create("xfs_bui_item", |
6413a014 | 2162 | xfs_bui_log_item_sizeof(XFS_BUI_MAX_FAST_EXTENTS), |
b1231760 | 2163 | 0, 0, NULL); |
182696fb DW |
2164 | if (!xfs_bui_cache) |
2165 | goto out_destroy_bud_cache; | |
6413a014 | 2166 | |
4136e38a DW |
2167 | xfs_attrd_cache = kmem_cache_create("xfs_attrd_item", |
2168 | sizeof(struct xfs_attrd_log_item), | |
2169 | 0, 0, NULL); | |
2170 | if (!xfs_attrd_cache) | |
2171 | goto out_destroy_bui_cache; | |
2172 | ||
2173 | xfs_attri_cache = kmem_cache_create("xfs_attri_item", | |
2174 | sizeof(struct xfs_attri_log_item), | |
2175 | 0, 0, NULL); | |
2176 | if (!xfs_attri_cache) | |
2177 | goto out_destroy_attrd_cache; | |
2178 | ||
784eb7d8 DC |
2179 | xfs_iunlink_cache = kmem_cache_create("xfs_iul_item", |
2180 | sizeof(struct xfs_iunlink_item), | |
2181 | 0, 0, NULL); | |
2182 | if (!xfs_iunlink_cache) | |
2183 | goto out_destroy_attri_cache; | |
2184 | ||
9f8868ff CH |
2185 | return 0; |
2186 | ||
784eb7d8 DC |
2187 | out_destroy_attri_cache: |
2188 | kmem_cache_destroy(xfs_attri_cache); | |
4136e38a DW |
2189 | out_destroy_attrd_cache: |
2190 | kmem_cache_destroy(xfs_attrd_cache); | |
2191 | out_destroy_bui_cache: | |
2192 | kmem_cache_destroy(xfs_bui_cache); | |
182696fb DW |
2193 | out_destroy_bud_cache: |
2194 | kmem_cache_destroy(xfs_bud_cache); | |
2195 | out_destroy_cui_cache: | |
2196 | kmem_cache_destroy(xfs_cui_cache); | |
2197 | out_destroy_cud_cache: | |
2198 | kmem_cache_destroy(xfs_cud_cache); | |
2199 | out_destroy_rui_cache: | |
2200 | kmem_cache_destroy(xfs_rui_cache); | |
2201 | out_destroy_rud_cache: | |
2202 | kmem_cache_destroy(xfs_rud_cache); | |
2203 | out_destroy_icreate_cache: | |
2204 | kmem_cache_destroy(xfs_icreate_cache); | |
2205 | out_destroy_ili_cache: | |
2206 | kmem_cache_destroy(xfs_ili_cache); | |
2207 | out_destroy_inode_cache: | |
2208 | kmem_cache_destroy(xfs_inode_cache); | |
2209 | out_destroy_efi_cache: | |
2210 | kmem_cache_destroy(xfs_efi_cache); | |
2211 | out_destroy_efd_cache: | |
2212 | kmem_cache_destroy(xfs_efd_cache); | |
2213 | out_destroy_buf_item_cache: | |
2214 | kmem_cache_destroy(xfs_buf_item_cache); | |
2215 | out_destroy_trans_cache: | |
2216 | kmem_cache_destroy(xfs_trans_cache); | |
2217 | out_destroy_ifork_cache: | |
2218 | kmem_cache_destroy(xfs_ifork_cache); | |
2219 | out_destroy_da_state_cache: | |
2220 | kmem_cache_destroy(xfs_da_state_cache); | |
f3c799c2 DW |
2221 | out_destroy_defer_item_cache: |
2222 | xfs_defer_destroy_item_caches(); | |
182696fb | 2223 | out_destroy_btree_cur_cache: |
9fa47bdc | 2224 | xfs_btree_destroy_cur_caches(); |
182696fb DW |
2225 | out_destroy_log_ticket_cache: |
2226 | kmem_cache_destroy(xfs_log_ticket_cache); | |
231f91ab DC |
2227 | out_destroy_buf_cache: |
2228 | kmem_cache_destroy(xfs_buf_cache); | |
9f8868ff CH |
2229 | out: |
2230 | return -ENOMEM; | |
2231 | } | |
2232 | ||
2233 | STATIC void | |
182696fb | 2234 | xfs_destroy_caches(void) |
9f8868ff | 2235 | { |
8c0a8537 KS |
2236 | /* |
2237 | * Make sure all delayed rcu free are flushed before we | |
2238 | * destroy caches. | |
2239 | */ | |
2240 | rcu_barrier(); | |
784eb7d8 | 2241 | kmem_cache_destroy(xfs_iunlink_cache); |
4136e38a DW |
2242 | kmem_cache_destroy(xfs_attri_cache); |
2243 | kmem_cache_destroy(xfs_attrd_cache); | |
182696fb DW |
2244 | kmem_cache_destroy(xfs_bui_cache); |
2245 | kmem_cache_destroy(xfs_bud_cache); | |
2246 | kmem_cache_destroy(xfs_cui_cache); | |
2247 | kmem_cache_destroy(xfs_cud_cache); | |
2248 | kmem_cache_destroy(xfs_rui_cache); | |
2249 | kmem_cache_destroy(xfs_rud_cache); | |
2250 | kmem_cache_destroy(xfs_icreate_cache); | |
2251 | kmem_cache_destroy(xfs_ili_cache); | |
2252 | kmem_cache_destroy(xfs_inode_cache); | |
2253 | kmem_cache_destroy(xfs_efi_cache); | |
2254 | kmem_cache_destroy(xfs_efd_cache); | |
2255 | kmem_cache_destroy(xfs_buf_item_cache); | |
2256 | kmem_cache_destroy(xfs_trans_cache); | |
2257 | kmem_cache_destroy(xfs_ifork_cache); | |
2258 | kmem_cache_destroy(xfs_da_state_cache); | |
f3c799c2 | 2259 | xfs_defer_destroy_item_caches(); |
9fa47bdc | 2260 | xfs_btree_destroy_cur_caches(); |
182696fb | 2261 | kmem_cache_destroy(xfs_log_ticket_cache); |
231f91ab | 2262 | kmem_cache_destroy(xfs_buf_cache); |
9f8868ff | 2263 | } |
1da177e4 | 2264 | |
0bf6a5bd DC |
2265 | STATIC int __init |
2266 | xfs_init_workqueues(void) | |
2267 | { | |
c999a223 DC |
2268 | /* |
2269 | * The allocation workqueue can be used in memory reclaim situations | |
2270 | * (writepage path), and parallelism is only limited by the number of | |
2271 | * AGs in all the filesystems mounted. Hence use the default large | |
2272 | * max_active value for this workqueue. | |
2273 | */ | |
8018ec08 | 2274 | xfs_alloc_wq = alloc_workqueue("xfsalloc", |
05a302a1 | 2275 | XFS_WQFLAGS(WQ_MEM_RECLAIM | WQ_FREEZABLE), 0); |
c999a223 | 2276 | if (!xfs_alloc_wq) |
5889608d | 2277 | return -ENOMEM; |
c999a223 | 2278 | |
05a302a1 DW |
2279 | xfs_discard_wq = alloc_workqueue("xfsdiscard", XFS_WQFLAGS(WQ_UNBOUND), |
2280 | 0); | |
4560e78f CH |
2281 | if (!xfs_discard_wq) |
2282 | goto out_free_alloc_wq; | |
2283 | ||
0bf6a5bd | 2284 | return 0; |
4560e78f CH |
2285 | out_free_alloc_wq: |
2286 | destroy_workqueue(xfs_alloc_wq); | |
2287 | return -ENOMEM; | |
0bf6a5bd DC |
2288 | } |
2289 | ||
39411f81 | 2290 | STATIC void |
0bf6a5bd DC |
2291 | xfs_destroy_workqueues(void) |
2292 | { | |
4560e78f | 2293 | destroy_workqueue(xfs_discard_wq); |
c999a223 | 2294 | destroy_workqueue(xfs_alloc_wq); |
0bf6a5bd DC |
2295 | } |
2296 | ||
1da177e4 | 2297 | STATIC int __init |
9f8868ff | 2298 | init_xfs_fs(void) |
1da177e4 LT |
2299 | { |
2300 | int error; | |
1da177e4 | 2301 | |
30cbc591 DW |
2302 | xfs_check_ondisk_structs(); |
2303 | ||
3cfb9290 DW |
2304 | error = xfs_dahash_test(); |
2305 | if (error) | |
2306 | return error; | |
2307 | ||
65795910 CH |
2308 | printk(KERN_INFO XFS_VERSION_STRING " with " |
2309 | XFS_BUILD_OPTIONS " enabled\n"); | |
1da177e4 | 2310 | |
9f8868ff | 2311 | xfs_dir_startup(); |
1da177e4 | 2312 | |
182696fb | 2313 | error = xfs_init_caches(); |
f1653c2e | 2314 | if (error) |
ef7d9593 | 2315 | goto out; |
f1653c2e | 2316 | |
0bf6a5bd | 2317 | error = xfs_init_workqueues(); |
9f8868ff | 2318 | if (error) |
182696fb | 2319 | goto out_destroy_caches; |
9f8868ff | 2320 | |
0bf6a5bd DC |
2321 | error = xfs_mru_cache_init(); |
2322 | if (error) | |
2323 | goto out_destroy_wq; | |
2324 | ||
9f8868ff CH |
2325 | error = xfs_init_procfs(); |
2326 | if (error) | |
231f91ab | 2327 | goto out_mru_cache_uninit; |
9f8868ff CH |
2328 | |
2329 | error = xfs_sysctl_register(); | |
2330 | if (error) | |
2331 | goto out_cleanup_procfs; | |
1da177e4 | 2332 | |
a76dba3b DW |
2333 | xfs_debugfs = xfs_debugfs_mkdir("xfs", NULL); |
2334 | ||
3d871226 BF |
2335 | xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj); |
2336 | if (!xfs_kset) { | |
2337 | error = -ENOMEM; | |
a76dba3b | 2338 | goto out_debugfs_unregister; |
3d871226 BF |
2339 | } |
2340 | ||
80529c45 BD |
2341 | xfsstats.xs_kobj.kobject.kset = xfs_kset; |
2342 | ||
2343 | xfsstats.xs_stats = alloc_percpu(struct xfsstats); | |
2344 | if (!xfsstats.xs_stats) { | |
2345 | error = -ENOMEM; | |
2346 | goto out_kset_unregister; | |
2347 | } | |
2348 | ||
2349 | error = xfs_sysfs_init(&xfsstats.xs_kobj, &xfs_stats_ktype, NULL, | |
bb230c12 BD |
2350 | "stats"); |
2351 | if (error) | |
80529c45 | 2352 | goto out_free_stats; |
bb230c12 | 2353 | |
d7a74cad DW |
2354 | error = xchk_global_stats_setup(xfs_debugfs); |
2355 | if (error) | |
2356 | goto out_remove_stats_kobj; | |
2357 | ||
65b65735 BF |
2358 | #ifdef DEBUG |
2359 | xfs_dbg_kobj.kobject.kset = xfs_kset; | |
2360 | error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug"); | |
a05931ce | 2361 | if (error) |
d7a74cad | 2362 | goto out_remove_scrub_stats; |
65b65735 BF |
2363 | #endif |
2364 | ||
2365 | error = xfs_qm_init(); | |
2366 | if (error) | |
bb230c12 | 2367 | goto out_remove_dbg_kobj; |
1da177e4 LT |
2368 | |
2369 | error = register_filesystem(&xfs_fs_type); | |
2370 | if (error) | |
a05931ce | 2371 | goto out_qm_exit; |
1da177e4 LT |
2372 | return 0; |
2373 | ||
a05931ce CH |
2374 | out_qm_exit: |
2375 | xfs_qm_exit(); | |
bb230c12 | 2376 | out_remove_dbg_kobj: |
65b65735 BF |
2377 | #ifdef DEBUG |
2378 | xfs_sysfs_del(&xfs_dbg_kobj); | |
d7a74cad | 2379 | out_remove_scrub_stats: |
65b65735 | 2380 | #endif |
d7a74cad DW |
2381 | xchk_global_stats_teardown(); |
2382 | out_remove_stats_kobj: | |
80529c45 BD |
2383 | xfs_sysfs_del(&xfsstats.xs_kobj); |
2384 | out_free_stats: | |
2385 | free_percpu(xfsstats.xs_stats); | |
bb230c12 | 2386 | out_kset_unregister: |
3d871226 | 2387 | kset_unregister(xfs_kset); |
a76dba3b DW |
2388 | out_debugfs_unregister: |
2389 | debugfs_remove(xfs_debugfs); | |
9f8868ff CH |
2390 | xfs_sysctl_unregister(); |
2391 | out_cleanup_procfs: | |
2392 | xfs_cleanup_procfs(); | |
9f8868ff CH |
2393 | out_mru_cache_uninit: |
2394 | xfs_mru_cache_uninit(); | |
0bf6a5bd DC |
2395 | out_destroy_wq: |
2396 | xfs_destroy_workqueues(); | |
182696fb DW |
2397 | out_destroy_caches: |
2398 | xfs_destroy_caches(); | |
9f8868ff | 2399 | out: |
1da177e4 LT |
2400 | return error; |
2401 | } | |
2402 | ||
2403 | STATIC void __exit | |
9f8868ff | 2404 | exit_xfs_fs(void) |
1da177e4 | 2405 | { |
a05931ce | 2406 | xfs_qm_exit(); |
1da177e4 | 2407 | unregister_filesystem(&xfs_fs_type); |
65b65735 BF |
2408 | #ifdef DEBUG |
2409 | xfs_sysfs_del(&xfs_dbg_kobj); | |
2410 | #endif | |
d7a74cad | 2411 | xchk_global_stats_teardown(); |
80529c45 BD |
2412 | xfs_sysfs_del(&xfsstats.xs_kobj); |
2413 | free_percpu(xfsstats.xs_stats); | |
3d871226 | 2414 | kset_unregister(xfs_kset); |
a76dba3b | 2415 | debugfs_remove(xfs_debugfs); |
9f8868ff CH |
2416 | xfs_sysctl_unregister(); |
2417 | xfs_cleanup_procfs(); | |
9f8868ff | 2418 | xfs_mru_cache_uninit(); |
0bf6a5bd | 2419 | xfs_destroy_workqueues(); |
182696fb | 2420 | xfs_destroy_caches(); |
af3b6382 | 2421 | xfs_uuid_table_free(); |
1da177e4 LT |
2422 | } |
2423 | ||
2424 | module_init(init_xfs_fs); | |
2425 | module_exit(exit_xfs_fs); | |
2426 | ||
2427 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
2428 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
2429 | MODULE_LICENSE("GPL"); |