]>
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" |
1da177e4 | 38 | |
dddde68b | 39 | #include <linux/magic.h> |
73e5fff9 IK |
40 | #include <linux/fs_context.h> |
41 | #include <linux/fs_parser.h> | |
1da177e4 | 42 | |
b87221de | 43 | static const struct super_operations xfs_super_operations; |
65b65735 | 44 | |
e3aed1a0 | 45 | static struct kset *xfs_kset; /* top-level xfs sysfs dir */ |
65b65735 BF |
46 | #ifdef DEBUG |
47 | static struct xfs_kobj xfs_dbg_kobj; /* global debug sysfs attrs */ | |
48 | #endif | |
1da177e4 | 49 | |
8d6c3446 IW |
50 | enum xfs_dax_mode { |
51 | XFS_DAX_INODE = 0, | |
52 | XFS_DAX_ALWAYS = 1, | |
53 | XFS_DAX_NEVER = 2, | |
54 | }; | |
55 | ||
56 | static void | |
57 | xfs_mount_set_dax_mode( | |
58 | struct xfs_mount *mp, | |
59 | enum xfs_dax_mode mode) | |
60 | { | |
61 | switch (mode) { | |
62 | case XFS_DAX_INODE: | |
63 | mp->m_flags &= ~(XFS_MOUNT_DAX_ALWAYS | XFS_MOUNT_DAX_NEVER); | |
64 | break; | |
65 | case XFS_DAX_ALWAYS: | |
66 | mp->m_flags |= XFS_MOUNT_DAX_ALWAYS; | |
67 | mp->m_flags &= ~XFS_MOUNT_DAX_NEVER; | |
68 | break; | |
69 | case XFS_DAX_NEVER: | |
70 | mp->m_flags |= XFS_MOUNT_DAX_NEVER; | |
71 | mp->m_flags &= ~XFS_MOUNT_DAX_ALWAYS; | |
72 | break; | |
73 | } | |
74 | } | |
75 | ||
76 | static const struct constant_table dax_param_enums[] = { | |
77 | {"inode", XFS_DAX_INODE }, | |
78 | {"always", XFS_DAX_ALWAYS }, | |
79 | {"never", XFS_DAX_NEVER }, | |
80 | {} | |
81 | }; | |
82 | ||
62a877e3 CH |
83 | /* |
84 | * Table driven mount option parser. | |
62a877e3 CH |
85 | */ |
86 | enum { | |
8da57c5c | 87 | Opt_logbufs, Opt_logbsize, Opt_logdev, Opt_rtdev, |
2e74af0e | 88 | Opt_wsync, Opt_noalign, Opt_swalloc, Opt_sunit, Opt_swidth, Opt_nouuid, |
94079285 | 89 | Opt_grpid, Opt_nogrpid, Opt_bsdgroups, Opt_sysvgroups, |
1c02d502 ES |
90 | Opt_allocsize, Opt_norecovery, Opt_inode64, Opt_inode32, Opt_ikeep, |
91 | Opt_noikeep, Opt_largeio, Opt_nolargeio, Opt_attr2, Opt_noattr2, | |
92 | Opt_filestreams, Opt_quota, Opt_noquota, Opt_usrquota, Opt_grpquota, | |
93 | Opt_prjquota, Opt_uquota, Opt_gquota, Opt_pquota, | |
2e74af0e | 94 | Opt_uqnoenforce, Opt_gqnoenforce, Opt_pqnoenforce, Opt_qnoenforce, |
8d6c3446 | 95 | Opt_discard, Opt_nodiscard, Opt_dax, Opt_dax_enum, |
62a877e3 CH |
96 | }; |
97 | ||
d7167b14 | 98 | static const struct fs_parameter_spec xfs_fs_parameters[] = { |
73e5fff9 IK |
99 | fsparam_u32("logbufs", Opt_logbufs), |
100 | fsparam_string("logbsize", Opt_logbsize), | |
101 | fsparam_string("logdev", Opt_logdev), | |
102 | fsparam_string("rtdev", Opt_rtdev), | |
103 | fsparam_flag("wsync", Opt_wsync), | |
104 | fsparam_flag("noalign", Opt_noalign), | |
105 | fsparam_flag("swalloc", Opt_swalloc), | |
106 | fsparam_u32("sunit", Opt_sunit), | |
107 | fsparam_u32("swidth", Opt_swidth), | |
108 | fsparam_flag("nouuid", Opt_nouuid), | |
109 | fsparam_flag("grpid", Opt_grpid), | |
110 | fsparam_flag("nogrpid", Opt_nogrpid), | |
111 | fsparam_flag("bsdgroups", Opt_bsdgroups), | |
112 | fsparam_flag("sysvgroups", Opt_sysvgroups), | |
113 | fsparam_string("allocsize", Opt_allocsize), | |
114 | fsparam_flag("norecovery", Opt_norecovery), | |
115 | fsparam_flag("inode64", Opt_inode64), | |
116 | fsparam_flag("inode32", Opt_inode32), | |
117 | fsparam_flag("ikeep", Opt_ikeep), | |
118 | fsparam_flag("noikeep", Opt_noikeep), | |
119 | fsparam_flag("largeio", Opt_largeio), | |
120 | fsparam_flag("nolargeio", Opt_nolargeio), | |
121 | fsparam_flag("attr2", Opt_attr2), | |
122 | fsparam_flag("noattr2", Opt_noattr2), | |
123 | fsparam_flag("filestreams", Opt_filestreams), | |
124 | fsparam_flag("quota", Opt_quota), | |
125 | fsparam_flag("noquota", Opt_noquota), | |
126 | fsparam_flag("usrquota", Opt_usrquota), | |
127 | fsparam_flag("grpquota", Opt_grpquota), | |
128 | fsparam_flag("prjquota", Opt_prjquota), | |
129 | fsparam_flag("uquota", Opt_uquota), | |
130 | fsparam_flag("gquota", Opt_gquota), | |
131 | fsparam_flag("pquota", Opt_pquota), | |
132 | fsparam_flag("uqnoenforce", Opt_uqnoenforce), | |
133 | fsparam_flag("gqnoenforce", Opt_gqnoenforce), | |
134 | fsparam_flag("pqnoenforce", Opt_pqnoenforce), | |
135 | fsparam_flag("qnoenforce", Opt_qnoenforce), | |
136 | fsparam_flag("discard", Opt_discard), | |
137 | fsparam_flag("nodiscard", Opt_nodiscard), | |
138 | fsparam_flag("dax", Opt_dax), | |
8d6c3446 | 139 | fsparam_enum("dax", Opt_dax_enum, dax_param_enums), |
73e5fff9 | 140 | {} |
62a877e3 CH |
141 | }; |
142 | ||
a67d7c5f | 143 | struct proc_xfs_info { |
cbe4dab1 DC |
144 | uint64_t flag; |
145 | char *str; | |
a67d7c5f DC |
146 | }; |
147 | ||
21f55993 CH |
148 | static int |
149 | xfs_fs_show_options( | |
150 | struct seq_file *m, | |
151 | struct dentry *root) | |
a67d7c5f DC |
152 | { |
153 | static struct proc_xfs_info xfs_info_set[] = { | |
154 | /* the few simple ones we can get from the mount struct */ | |
2e74af0e ES |
155 | { XFS_MOUNT_IKEEP, ",ikeep" }, |
156 | { XFS_MOUNT_WSYNC, ",wsync" }, | |
157 | { XFS_MOUNT_NOALIGN, ",noalign" }, | |
158 | { XFS_MOUNT_SWALLOC, ",swalloc" }, | |
159 | { XFS_MOUNT_NOUUID, ",nouuid" }, | |
160 | { XFS_MOUNT_NORECOVERY, ",norecovery" }, | |
161 | { XFS_MOUNT_ATTR2, ",attr2" }, | |
162 | { XFS_MOUNT_FILESTREAMS, ",filestreams" }, | |
163 | { XFS_MOUNT_GRPID, ",grpid" }, | |
164 | { XFS_MOUNT_DISCARD, ",discard" }, | |
7c6b94b1 | 165 | { XFS_MOUNT_LARGEIO, ",largeio" }, |
8d6c3446 IW |
166 | { XFS_MOUNT_DAX_ALWAYS, ",dax=always" }, |
167 | { XFS_MOUNT_DAX_NEVER, ",dax=never" }, | |
a67d7c5f DC |
168 | { 0, NULL } |
169 | }; | |
21f55993 | 170 | struct xfs_mount *mp = XFS_M(root->d_sb); |
a67d7c5f DC |
171 | struct proc_xfs_info *xfs_infop; |
172 | ||
173 | for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) { | |
174 | if (mp->m_flags & xfs_infop->flag) | |
175 | seq_puts(m, xfs_infop->str); | |
176 | } | |
1775c506 CH |
177 | |
178 | seq_printf(m, ",inode%d", | |
179 | (mp->m_flags & XFS_MOUNT_SMALL_INUMS) ? 32 : 64); | |
a67d7c5f | 180 | |
3274d008 | 181 | if (mp->m_flags & XFS_MOUNT_ALLOCSIZE) |
2e74af0e | 182 | seq_printf(m, ",allocsize=%dk", |
aa58d445 | 183 | (1 << mp->m_allocsize_log) >> 10); |
a67d7c5f DC |
184 | |
185 | if (mp->m_logbufs > 0) | |
2e74af0e | 186 | seq_printf(m, ",logbufs=%d", mp->m_logbufs); |
a67d7c5f | 187 | if (mp->m_logbsize > 0) |
2e74af0e | 188 | seq_printf(m, ",logbsize=%dk", mp->m_logbsize >> 10); |
a67d7c5f DC |
189 | |
190 | if (mp->m_logname) | |
2e74af0e | 191 | seq_show_option(m, "logdev", mp->m_logname); |
a67d7c5f | 192 | if (mp->m_rtname) |
2e74af0e | 193 | seq_show_option(m, "rtdev", mp->m_rtname); |
a67d7c5f DC |
194 | |
195 | if (mp->m_dalign > 0) | |
2e74af0e | 196 | seq_printf(m, ",sunit=%d", |
a67d7c5f DC |
197 | (int)XFS_FSB_TO_BB(mp, mp->m_dalign)); |
198 | if (mp->m_swidth > 0) | |
2e74af0e | 199 | seq_printf(m, ",swidth=%d", |
a67d7c5f DC |
200 | (int)XFS_FSB_TO_BB(mp, mp->m_swidth)); |
201 | ||
202 | if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD)) | |
2e74af0e | 203 | seq_puts(m, ",usrquota"); |
a67d7c5f | 204 | else if (mp->m_qflags & XFS_UQUOTA_ACCT) |
2e74af0e | 205 | seq_puts(m, ",uqnoenforce"); |
a67d7c5f | 206 | |
988abe40 | 207 | if (mp->m_qflags & XFS_PQUOTA_ACCT) { |
83e782e1 | 208 | if (mp->m_qflags & XFS_PQUOTA_ENFD) |
2e74af0e | 209 | seq_puts(m, ",prjquota"); |
988abe40 | 210 | else |
2e74af0e | 211 | seq_puts(m, ",pqnoenforce"); |
d892d586 CS |
212 | } |
213 | if (mp->m_qflags & XFS_GQUOTA_ACCT) { | |
83e782e1 | 214 | if (mp->m_qflags & XFS_GQUOTA_ENFD) |
2e74af0e | 215 | seq_puts(m, ",grpquota"); |
988abe40 | 216 | else |
2e74af0e | 217 | seq_puts(m, ",gqnoenforce"); |
988abe40 | 218 | } |
a67d7c5f DC |
219 | |
220 | if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT)) | |
2e74af0e | 221 | seq_puts(m, ",noquota"); |
21f55993 CH |
222 | |
223 | return 0; | |
a67d7c5f | 224 | } |
91083269 | 225 | |
9de67c3b | 226 | /* |
12c3f05c ES |
227 | * Set parameters for inode allocation heuristics, taking into account |
228 | * filesystem size and inode32/inode64 mount options; i.e. specifically | |
229 | * whether or not XFS_MOUNT_SMALL_INUMS is set. | |
230 | * | |
231 | * Inode allocation patterns are altered only if inode32 is requested | |
232 | * (XFS_MOUNT_SMALL_INUMS), and the filesystem is sufficiently large. | |
233 | * If altered, XFS_MOUNT_32BITINODES is set as well. | |
234 | * | |
235 | * An agcount independent of that in the mount structure is provided | |
236 | * because in the growfs case, mp->m_sb.sb_agcount is not yet updated | |
237 | * to the potentially higher ag count. | |
238 | * | |
239 | * Returns the maximum AG index which may contain inodes. | |
9de67c3b | 240 | */ |
2d2194f6 | 241 | xfs_agnumber_t |
12c3f05c ES |
242 | xfs_set_inode_alloc( |
243 | struct xfs_mount *mp, | |
244 | xfs_agnumber_t agcount) | |
2d2194f6 | 245 | { |
12c3f05c | 246 | xfs_agnumber_t index; |
4056c1d0 | 247 | xfs_agnumber_t maxagi = 0; |
2d2194f6 CM |
248 | xfs_sb_t *sbp = &mp->m_sb; |
249 | xfs_agnumber_t max_metadata; | |
54aa61f8 ES |
250 | xfs_agino_t agino; |
251 | xfs_ino_t ino; | |
2d2194f6 | 252 | |
12c3f05c ES |
253 | /* |
254 | * Calculate how much should be reserved for inodes to meet | |
255 | * the max inode percentage. Used only for inode32. | |
2d2194f6 | 256 | */ |
ef325959 | 257 | if (M_IGEO(mp)->maxicount) { |
c8ce540d | 258 | uint64_t icount; |
2d2194f6 CM |
259 | |
260 | icount = sbp->sb_dblocks * sbp->sb_imax_pct; | |
261 | do_div(icount, 100); | |
262 | icount += sbp->sb_agblocks - 1; | |
263 | do_div(icount, sbp->sb_agblocks); | |
264 | max_metadata = icount; | |
265 | } else { | |
9de67c3b | 266 | max_metadata = agcount; |
2d2194f6 CM |
267 | } |
268 | ||
12c3f05c | 269 | /* Get the last possible inode in the filesystem */ |
43004b2a | 270 | agino = XFS_AGB_TO_AGINO(mp, sbp->sb_agblocks - 1); |
12c3f05c ES |
271 | ino = XFS_AGINO_TO_INO(mp, agcount - 1, agino); |
272 | ||
273 | /* | |
274 | * If user asked for no more than 32-bit inodes, and the fs is | |
275 | * sufficiently large, set XFS_MOUNT_32BITINODES if we must alter | |
276 | * the allocator to accommodate the request. | |
277 | */ | |
278 | if ((mp->m_flags & XFS_MOUNT_SMALL_INUMS) && ino > XFS_MAXINUMBER_32) | |
279 | mp->m_flags |= XFS_MOUNT_32BITINODES; | |
280 | else | |
281 | mp->m_flags &= ~XFS_MOUNT_32BITINODES; | |
54aa61f8 | 282 | |
9de67c3b | 283 | for (index = 0; index < agcount; index++) { |
12c3f05c | 284 | struct xfs_perag *pag; |
4056c1d0 | 285 | |
12c3f05c | 286 | ino = XFS_AGINO_TO_INO(mp, index, agino); |
2d2194f6 CM |
287 | |
288 | pag = xfs_perag_get(mp, index); | |
2d2194f6 | 289 | |
12c3f05c ES |
290 | if (mp->m_flags & XFS_MOUNT_32BITINODES) { |
291 | if (ino > XFS_MAXINUMBER_32) { | |
292 | pag->pagi_inodeok = 0; | |
293 | pag->pagf_metadata = 0; | |
294 | } else { | |
295 | pag->pagi_inodeok = 1; | |
296 | maxagi++; | |
297 | if (index < max_metadata) | |
298 | pag->pagf_metadata = 1; | |
299 | else | |
300 | pag->pagf_metadata = 0; | |
301 | } | |
302 | } else { | |
303 | pag->pagi_inodeok = 1; | |
304 | pag->pagf_metadata = 0; | |
305 | } | |
2d2194f6 | 306 | |
2d2194f6 CM |
307 | xfs_perag_put(pag); |
308 | } | |
309 | ||
12c3f05c | 310 | return (mp->m_flags & XFS_MOUNT_32BITINODES) ? maxagi : agcount; |
2d2194f6 CM |
311 | } |
312 | ||
3180e66d | 313 | STATIC int |
1da177e4 LT |
314 | xfs_blkdev_get( |
315 | xfs_mount_t *mp, | |
316 | const char *name, | |
317 | struct block_device **bdevp) | |
318 | { | |
319 | int error = 0; | |
320 | ||
d4d77629 TH |
321 | *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL, |
322 | mp); | |
1da177e4 LT |
323 | if (IS_ERR(*bdevp)) { |
324 | error = PTR_ERR(*bdevp); | |
77af574e | 325 | xfs_warn(mp, "Invalid device [%s], error=%d", name, error); |
1da177e4 LT |
326 | } |
327 | ||
2451337d | 328 | return error; |
1da177e4 LT |
329 | } |
330 | ||
3180e66d | 331 | STATIC void |
1da177e4 LT |
332 | xfs_blkdev_put( |
333 | struct block_device *bdev) | |
334 | { | |
335 | if (bdev) | |
e525fd89 | 336 | blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); |
1da177e4 LT |
337 | } |
338 | ||
f538d4da CH |
339 | void |
340 | xfs_blkdev_issue_flush( | |
341 | xfs_buftarg_t *buftarg) | |
342 | { | |
9398554f | 343 | blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS); |
f538d4da | 344 | } |
1da177e4 | 345 | |
19f354d4 CH |
346 | STATIC void |
347 | xfs_close_devices( | |
348 | struct xfs_mount *mp) | |
349 | { | |
486aff5e DW |
350 | struct dax_device *dax_ddev = mp->m_ddev_targp->bt_daxdev; |
351 | ||
19f354d4 | 352 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { |
c032bfcf | 353 | struct block_device *logdev = mp->m_logdev_targp->bt_bdev; |
486aff5e DW |
354 | struct dax_device *dax_logdev = mp->m_logdev_targp->bt_daxdev; |
355 | ||
a1f69417 | 356 | xfs_free_buftarg(mp->m_logdev_targp); |
c032bfcf | 357 | xfs_blkdev_put(logdev); |
486aff5e | 358 | fs_put_dax(dax_logdev); |
19f354d4 CH |
359 | } |
360 | if (mp->m_rtdev_targp) { | |
c032bfcf | 361 | struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev; |
486aff5e DW |
362 | struct dax_device *dax_rtdev = mp->m_rtdev_targp->bt_daxdev; |
363 | ||
a1f69417 | 364 | xfs_free_buftarg(mp->m_rtdev_targp); |
c032bfcf | 365 | xfs_blkdev_put(rtdev); |
486aff5e | 366 | fs_put_dax(dax_rtdev); |
19f354d4 | 367 | } |
a1f69417 | 368 | xfs_free_buftarg(mp->m_ddev_targp); |
486aff5e | 369 | fs_put_dax(dax_ddev); |
19f354d4 CH |
370 | } |
371 | ||
372 | /* | |
373 | * The file system configurations are: | |
374 | * (1) device (partition) with data and internal log | |
375 | * (2) logical volume with data and log subvolumes. | |
376 | * (3) logical volume with data, log, and realtime subvolumes. | |
377 | * | |
378 | * We only have to handle opening the log and realtime volumes here if | |
379 | * they are present. The data subvolume has already been opened by | |
380 | * get_sb_bdev() and is stored in sb->s_bdev. | |
381 | */ | |
382 | STATIC int | |
383 | xfs_open_devices( | |
9d565ffa | 384 | struct xfs_mount *mp) |
19f354d4 CH |
385 | { |
386 | struct block_device *ddev = mp->m_super->s_bdev; | |
486aff5e DW |
387 | struct dax_device *dax_ddev = fs_dax_get_by_bdev(ddev); |
388 | struct dax_device *dax_logdev = NULL, *dax_rtdev = NULL; | |
19f354d4 CH |
389 | struct block_device *logdev = NULL, *rtdev = NULL; |
390 | int error; | |
391 | ||
392 | /* | |
393 | * Open real time and log devices - order is important. | |
394 | */ | |
9d565ffa CH |
395 | if (mp->m_logname) { |
396 | error = xfs_blkdev_get(mp, mp->m_logname, &logdev); | |
19f354d4 CH |
397 | if (error) |
398 | goto out; | |
486aff5e | 399 | dax_logdev = fs_dax_get_by_bdev(logdev); |
19f354d4 CH |
400 | } |
401 | ||
9d565ffa CH |
402 | if (mp->m_rtname) { |
403 | error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev); | |
19f354d4 CH |
404 | if (error) |
405 | goto out_close_logdev; | |
406 | ||
407 | if (rtdev == ddev || rtdev == logdev) { | |
4f10700a DC |
408 | xfs_warn(mp, |
409 | "Cannot mount filesystem with identical rtdev and ddev/logdev."); | |
2451337d | 410 | error = -EINVAL; |
19f354d4 CH |
411 | goto out_close_rtdev; |
412 | } | |
486aff5e | 413 | dax_rtdev = fs_dax_get_by_bdev(rtdev); |
19f354d4 CH |
414 | } |
415 | ||
416 | /* | |
417 | * Setup xfs_mount buffer target pointers | |
418 | */ | |
2451337d | 419 | error = -ENOMEM; |
486aff5e | 420 | mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, dax_ddev); |
19f354d4 CH |
421 | if (!mp->m_ddev_targp) |
422 | goto out_close_rtdev; | |
423 | ||
424 | if (rtdev) { | |
486aff5e | 425 | mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, dax_rtdev); |
19f354d4 CH |
426 | if (!mp->m_rtdev_targp) |
427 | goto out_free_ddev_targ; | |
428 | } | |
429 | ||
430 | if (logdev && logdev != ddev) { | |
486aff5e | 431 | mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, dax_logdev); |
19f354d4 CH |
432 | if (!mp->m_logdev_targp) |
433 | goto out_free_rtdev_targ; | |
434 | } else { | |
435 | mp->m_logdev_targp = mp->m_ddev_targp; | |
436 | } | |
437 | ||
438 | return 0; | |
439 | ||
440 | out_free_rtdev_targ: | |
441 | if (mp->m_rtdev_targp) | |
a1f69417 | 442 | xfs_free_buftarg(mp->m_rtdev_targp); |
19f354d4 | 443 | out_free_ddev_targ: |
a1f69417 | 444 | xfs_free_buftarg(mp->m_ddev_targp); |
19f354d4 | 445 | out_close_rtdev: |
d2a5e3c6 | 446 | xfs_blkdev_put(rtdev); |
486aff5e | 447 | fs_put_dax(dax_rtdev); |
19f354d4 | 448 | out_close_logdev: |
486aff5e | 449 | if (logdev && logdev != ddev) { |
19f354d4 | 450 | xfs_blkdev_put(logdev); |
486aff5e DW |
451 | fs_put_dax(dax_logdev); |
452 | } | |
19f354d4 | 453 | out: |
486aff5e | 454 | fs_put_dax(dax_ddev); |
19f354d4 CH |
455 | return error; |
456 | } | |
457 | ||
e34b562c CH |
458 | /* |
459 | * Setup xfs_mount buffer target pointers based on superblock | |
460 | */ | |
461 | STATIC int | |
462 | xfs_setup_devices( | |
463 | struct xfs_mount *mp) | |
464 | { | |
465 | int error; | |
19f354d4 | 466 | |
a96c4151 | 467 | error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize); |
e34b562c CH |
468 | if (error) |
469 | return error; | |
470 | ||
471 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { | |
472 | unsigned int log_sector_size = BBSIZE; | |
473 | ||
474 | if (xfs_sb_version_hassector(&mp->m_sb)) | |
475 | log_sector_size = mp->m_sb.sb_logsectsize; | |
476 | error = xfs_setsize_buftarg(mp->m_logdev_targp, | |
e34b562c CH |
477 | log_sector_size); |
478 | if (error) | |
479 | return error; | |
480 | } | |
481 | if (mp->m_rtdev_targp) { | |
482 | error = xfs_setsize_buftarg(mp->m_rtdev_targp, | |
e34b562c CH |
483 | mp->m_sb.sb_sectsize); |
484 | if (error) | |
485 | return error; | |
486 | } | |
487 | ||
488 | return 0; | |
489 | } | |
19f354d4 | 490 | |
aa6bf01d CH |
491 | STATIC int |
492 | xfs_init_mount_workqueues( | |
493 | struct xfs_mount *mp) | |
494 | { | |
78c931b8 | 495 | mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s", |
e1d3d218 | 496 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 1, mp->m_super->s_id); |
78c931b8 BF |
497 | if (!mp->m_buf_workqueue) |
498 | goto out; | |
499 | ||
aa6bf01d | 500 | mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s", |
e1d3d218 | 501 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_super->s_id); |
aa6bf01d | 502 | if (!mp->m_unwritten_workqueue) |
28408243 | 503 | goto out_destroy_buf; |
aa6bf01d | 504 | |
4c2d542f | 505 | mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s", |
8ab39f11 | 506 | WQ_MEM_RECLAIM | WQ_FREEZABLE | WQ_UNBOUND, |
e1d3d218 | 507 | 0, mp->m_super->s_id); |
4c2d542f DC |
508 | if (!mp->m_cil_workqueue) |
509 | goto out_destroy_unwritten; | |
5889608d DC |
510 | |
511 | mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s", | |
e1d3d218 | 512 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_super->s_id); |
5889608d DC |
513 | if (!mp->m_reclaim_workqueue) |
514 | goto out_destroy_cil; | |
515 | ||
579b62fa | 516 | mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s", |
e1d3d218 | 517 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_super->s_id); |
579b62fa | 518 | if (!mp->m_eofblocks_workqueue) |
1058d0f5 | 519 | goto out_destroy_reclaim; |
579b62fa | 520 | |
696a5620 | 521 | mp->m_sync_workqueue = alloc_workqueue("xfs-sync/%s", WQ_FREEZABLE, 0, |
e1d3d218 | 522 | mp->m_super->s_id); |
696a5620 BF |
523 | if (!mp->m_sync_workqueue) |
524 | goto out_destroy_eofb; | |
525 | ||
aa6bf01d CH |
526 | return 0; |
527 | ||
696a5620 BF |
528 | out_destroy_eofb: |
529 | destroy_workqueue(mp->m_eofblocks_workqueue); | |
5889608d DC |
530 | out_destroy_reclaim: |
531 | destroy_workqueue(mp->m_reclaim_workqueue); | |
532 | out_destroy_cil: | |
533 | destroy_workqueue(mp->m_cil_workqueue); | |
4c2d542f DC |
534 | out_destroy_unwritten: |
535 | destroy_workqueue(mp->m_unwritten_workqueue); | |
78c931b8 BF |
536 | out_destroy_buf: |
537 | destroy_workqueue(mp->m_buf_workqueue); | |
aa6bf01d CH |
538 | out: |
539 | return -ENOMEM; | |
540 | } | |
541 | ||
542 | STATIC void | |
543 | xfs_destroy_mount_workqueues( | |
544 | struct xfs_mount *mp) | |
545 | { | |
696a5620 | 546 | destroy_workqueue(mp->m_sync_workqueue); |
579b62fa | 547 | destroy_workqueue(mp->m_eofblocks_workqueue); |
5889608d | 548 | destroy_workqueue(mp->m_reclaim_workqueue); |
4c2d542f | 549 | destroy_workqueue(mp->m_cil_workqueue); |
aa6bf01d | 550 | destroy_workqueue(mp->m_unwritten_workqueue); |
78c931b8 | 551 | destroy_workqueue(mp->m_buf_workqueue); |
aa6bf01d CH |
552 | } |
553 | ||
f0f7a674 DW |
554 | static void |
555 | xfs_flush_inodes_worker( | |
556 | struct work_struct *work) | |
557 | { | |
558 | struct xfs_mount *mp = container_of(work, struct xfs_mount, | |
559 | m_flush_inodes_work); | |
560 | struct super_block *sb = mp->m_super; | |
561 | ||
562 | if (down_read_trylock(&sb->s_umount)) { | |
563 | sync_inodes_sb(sb); | |
564 | up_read(&sb->s_umount); | |
565 | } | |
566 | } | |
567 | ||
9aa05000 DC |
568 | /* |
569 | * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK | |
570 | * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting | |
571 | * for IO to complete so that we effectively throttle multiple callers to the | |
572 | * rate at which IO is completing. | |
573 | */ | |
574 | void | |
575 | xfs_flush_inodes( | |
576 | struct xfs_mount *mp) | |
577 | { | |
f0f7a674 DW |
578 | /* |
579 | * If flush_work() returns true then that means we waited for a flush | |
580 | * which was already in progress. Don't bother running another scan. | |
581 | */ | |
582 | if (flush_work(&mp->m_flush_inodes_work)) | |
c6425702 DW |
583 | return; |
584 | ||
f0f7a674 DW |
585 | queue_work(mp->m_sync_workqueue, &mp->m_flush_inodes_work); |
586 | flush_work(&mp->m_flush_inodes_work); | |
9aa05000 DC |
587 | } |
588 | ||
bf904248 | 589 | /* Catch misguided souls that try to use this interface on XFS */ |
1da177e4 | 590 | STATIC struct inode * |
a50cd269 | 591 | xfs_fs_alloc_inode( |
1da177e4 LT |
592 | struct super_block *sb) |
593 | { | |
bf904248 | 594 | BUG(); |
493dca61 | 595 | return NULL; |
1da177e4 LT |
596 | } |
597 | ||
4831822f CH |
598 | #ifdef DEBUG |
599 | static void | |
600 | xfs_check_delalloc( | |
601 | struct xfs_inode *ip, | |
602 | int whichfork) | |
603 | { | |
604 | struct xfs_ifork *ifp = XFS_IFORK_PTR(ip, whichfork); | |
605 | struct xfs_bmbt_irec got; | |
606 | struct xfs_iext_cursor icur; | |
607 | ||
608 | if (!ifp || !xfs_iext_lookup_extent(ip, ifp, 0, &icur, &got)) | |
609 | return; | |
610 | do { | |
611 | if (isnullstartblock(got.br_startblock)) { | |
612 | xfs_warn(ip->i_mount, | |
613 | "ino %llx %s fork has delalloc extent at [0x%llx:0x%llx]", | |
614 | ip->i_ino, | |
615 | whichfork == XFS_DATA_FORK ? "data" : "cow", | |
616 | got.br_startoff, got.br_blockcount); | |
617 | } | |
618 | } while (xfs_iext_next_extent(ifp, &icur, &got)); | |
619 | } | |
620 | #else | |
621 | #define xfs_check_delalloc(ip, whichfork) do { } while (0) | |
622 | #endif | |
623 | ||
bf904248 | 624 | /* |
99fa8cb3 | 625 | * Now that the generic code is guaranteed not to be accessing |
8179c036 | 626 | * the linux inode, we can inactivate and reclaim the inode. |
bf904248 | 627 | */ |
1da177e4 | 628 | STATIC void |
a50cd269 | 629 | xfs_fs_destroy_inode( |
848ce8f7 | 630 | struct inode *inode) |
1da177e4 | 631 | { |
848ce8f7 CH |
632 | struct xfs_inode *ip = XFS_I(inode); |
633 | ||
cca28fb8 | 634 | trace_xfs_destroy_inode(ip); |
99fa8cb3 | 635 | |
65523218 | 636 | ASSERT(!rwsem_is_locked(&inode->i_rwsem)); |
8179c036 DC |
637 | XFS_STATS_INC(ip->i_mount, vn_rele); |
638 | XFS_STATS_INC(ip->i_mount, vn_remove); | |
639 | ||
640 | xfs_inactive(ip); | |
848ce8f7 | 641 | |
4831822f CH |
642 | if (!XFS_FORCED_SHUTDOWN(ip->i_mount) && ip->i_delayed_blks) { |
643 | xfs_check_delalloc(ip, XFS_DATA_FORK); | |
644 | xfs_check_delalloc(ip, XFS_COW_FORK); | |
645 | ASSERT(0); | |
646 | } | |
647 | ||
8179c036 | 648 | XFS_STATS_INC(ip->i_mount, vn_reclaim); |
848ce8f7 CH |
649 | |
650 | /* | |
651 | * We should never get here with one of the reclaim flags already set. | |
652 | */ | |
653 | ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE)); | |
654 | ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM)); | |
655 | ||
656 | /* | |
718ecc50 DC |
657 | * We always use background reclaim here because even if the inode is |
658 | * clean, it still may be under IO and hence we have wait for IO | |
659 | * completion to occur before we can reclaim the inode. The background | |
660 | * reclaim path handles this more efficiently than we can here, so | |
661 | * simply let background reclaim tear down all inodes. | |
848ce8f7 | 662 | */ |
57817c68 | 663 | xfs_inode_set_reclaim_tag(ip); |
1da177e4 LT |
664 | } |
665 | ||
c3b1b131 CH |
666 | static void |
667 | xfs_fs_dirty_inode( | |
668 | struct inode *inode, | |
669 | int flag) | |
670 | { | |
671 | struct xfs_inode *ip = XFS_I(inode); | |
672 | struct xfs_mount *mp = ip->i_mount; | |
673 | struct xfs_trans *tp; | |
674 | ||
675 | if (!(inode->i_sb->s_flags & SB_LAZYTIME)) | |
676 | return; | |
677 | if (flag != I_DIRTY_SYNC || !(inode->i_state & I_DIRTY_TIME)) | |
678 | return; | |
679 | ||
680 | if (xfs_trans_alloc(mp, &M_RES(mp)->tr_fsyncts, 0, 0, 0, &tp)) | |
681 | return; | |
682 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
683 | xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); | |
684 | xfs_trans_log_inode(tp, ip, XFS_ILOG_TIMESTAMP); | |
685 | xfs_trans_commit(tp); | |
686 | } | |
687 | ||
07c8f675 DC |
688 | /* |
689 | * Slab object creation initialisation for the XFS inode. | |
690 | * This covers only the idempotent fields in the XFS inode; | |
691 | * all other fields need to be initialised on allocation | |
b595076a | 692 | * from the slab. This avoids the need to repeatedly initialise |
07c8f675 DC |
693 | * fields in the xfs inode that left in the initialise state |
694 | * when freeing the inode. | |
695 | */ | |
bf904248 DC |
696 | STATIC void |
697 | xfs_fs_inode_init_once( | |
07c8f675 DC |
698 | void *inode) |
699 | { | |
700 | struct xfs_inode *ip = inode; | |
701 | ||
702 | memset(ip, 0, sizeof(struct xfs_inode)); | |
bf904248 DC |
703 | |
704 | /* vfs inode */ | |
705 | inode_init_once(VFS_I(ip)); | |
706 | ||
707 | /* xfs inode */ | |
07c8f675 DC |
708 | atomic_set(&ip->i_pincount, 0); |
709 | spin_lock_init(&ip->i_flags_lock); | |
07c8f675 | 710 | |
653c60b6 DC |
711 | mrlock_init(&ip->i_mmaplock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, |
712 | "xfsino", ip->i_ino); | |
07c8f675 DC |
713 | mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, |
714 | "xfsino", ip->i_ino); | |
07c8f675 DC |
715 | } |
716 | ||
5132ba8f DC |
717 | /* |
718 | * We do an unlocked check for XFS_IDONTCACHE here because we are already | |
719 | * serialised against cache hits here via the inode->i_lock and igrab() in | |
720 | * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be | |
721 | * racing with us, and it avoids needing to grab a spinlock here for every inode | |
722 | * we drop the final reference on. | |
723 | */ | |
724 | STATIC int | |
725 | xfs_fs_drop_inode( | |
726 | struct inode *inode) | |
727 | { | |
728 | struct xfs_inode *ip = XFS_I(inode); | |
729 | ||
17c12bcd DW |
730 | /* |
731 | * If this unlinked inode is in the middle of recovery, don't | |
732 | * drop the inode just yet; log recovery will take care of | |
733 | * that. See the comment for this inode flag. | |
734 | */ | |
735 | if (ip->i_flags & XFS_IRECOVERY) { | |
736 | ASSERT(ip->i_mount->m_log->l_flags & XLOG_RECOVERY_NEEDED); | |
737 | return 0; | |
738 | } | |
739 | ||
dae2f8ed | 740 | return generic_drop_inode(inode); |
5132ba8f DC |
741 | } |
742 | ||
a943f372 IK |
743 | static void |
744 | xfs_mount_free( | |
a738159d CH |
745 | struct xfs_mount *mp) |
746 | { | |
a738159d CH |
747 | kfree(mp->m_rtname); |
748 | kfree(mp->m_logname); | |
a943f372 | 749 | kmem_free(mp); |
a738159d CH |
750 | } |
751 | ||
1da177e4 | 752 | STATIC int |
69961a26 | 753 | xfs_fs_sync_fs( |
1da177e4 LT |
754 | struct super_block *sb, |
755 | int wait) | |
756 | { | |
745f6919 | 757 | struct xfs_mount *mp = XFS_M(sb); |
1da177e4 | 758 | |
e893bffd | 759 | /* |
34625c66 | 760 | * Doing anything during the async pass would be counterproductive. |
e893bffd | 761 | */ |
34625c66 | 762 | if (!wait) |
69961a26 | 763 | return 0; |
69961a26 | 764 | |
34061f5c | 765 | xfs_log_force(mp, XFS_LOG_SYNC); |
69961a26 | 766 | if (laptop_mode) { |
1da177e4 LT |
767 | /* |
768 | * The disk must be active because we're syncing. | |
f661f1e0 | 769 | * We schedule log work now (now that the disk is |
1da177e4 LT |
770 | * active) instead of later (when it might not be). |
771 | */ | |
f661f1e0 | 772 | flush_delayed_work(&mp->m_log->l_work); |
1da177e4 LT |
773 | } |
774 | ||
69961a26 | 775 | return 0; |
1da177e4 LT |
776 | } |
777 | ||
778 | STATIC int | |
a50cd269 | 779 | xfs_fs_statfs( |
726c3342 | 780 | struct dentry *dentry, |
1da177e4 LT |
781 | struct kstatfs *statp) |
782 | { | |
4ca488eb CH |
783 | struct xfs_mount *mp = XFS_M(dentry->d_sb); |
784 | xfs_sb_t *sbp = &mp->m_sb; | |
2b0143b5 | 785 | struct xfs_inode *ip = XFS_I(d_inode(dentry)); |
c8ce540d DW |
786 | uint64_t fakeinos, id; |
787 | uint64_t icount; | |
788 | uint64_t ifree; | |
789 | uint64_t fdblocks; | |
4ca488eb | 790 | xfs_extlen_t lsize; |
c8ce540d | 791 | int64_t ffree; |
4ca488eb | 792 | |
dddde68b | 793 | statp->f_type = XFS_SUPER_MAGIC; |
4ca488eb CH |
794 | statp->f_namelen = MAXNAMELEN - 1; |
795 | ||
796 | id = huge_encode_dev(mp->m_ddev_targp->bt_dev); | |
6d1349c7 | 797 | statp->f_fsid = u64_to_fsid(id); |
4ca488eb | 798 | |
501ab323 | 799 | icount = percpu_counter_sum(&mp->m_icount); |
e88b64ea | 800 | ifree = percpu_counter_sum(&mp->m_ifree); |
0d485ada | 801 | fdblocks = percpu_counter_sum(&mp->m_fdblocks); |
4ca488eb CH |
802 | |
803 | spin_lock(&mp->m_sb_lock); | |
804 | statp->f_bsize = sbp->sb_blocksize; | |
805 | lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0; | |
806 | statp->f_blocks = sbp->sb_dblocks - lsize; | |
0d485ada DC |
807 | spin_unlock(&mp->m_sb_lock); |
808 | ||
237aac46 ZB |
809 | /* make sure statp->f_bfree does not underflow */ |
810 | statp->f_bfree = max_t(int64_t, fdblocks - mp->m_alloc_set_aside, 0); | |
0d485ada DC |
811 | statp->f_bavail = statp->f_bfree; |
812 | ||
43004b2a | 813 | fakeinos = XFS_FSB_TO_INO(mp, statp->f_bfree); |
9bb54cb5 | 814 | statp->f_files = min(icount + fakeinos, (uint64_t)XFS_MAXINUMBER); |
ef325959 | 815 | if (M_IGEO(mp)->maxicount) |
a19d9f88 CH |
816 | statp->f_files = min_t(typeof(statp->f_files), |
817 | statp->f_files, | |
ef325959 | 818 | M_IGEO(mp)->maxicount); |
2fe33661 | 819 | |
01f9882e ES |
820 | /* If sb_icount overshot maxicount, report actual allocation */ |
821 | statp->f_files = max_t(typeof(statp->f_files), | |
822 | statp->f_files, | |
823 | sbp->sb_icount); | |
824 | ||
2fe33661 | 825 | /* make sure statp->f_ffree does not underflow */ |
e88b64ea | 826 | ffree = statp->f_files - (icount - ifree); |
c8ce540d | 827 | statp->f_ffree = max_t(int64_t, ffree, 0); |
2fe33661 | 828 | |
4ca488eb | 829 | |
da5bf95e | 830 | if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && |
83e782e1 CS |
831 | ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) == |
832 | (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD)) | |
7d095257 | 833 | xfs_qm_statvfs(ip, statp); |
a0158315 RW |
834 | |
835 | if (XFS_IS_REALTIME_MOUNT(mp) && | |
836 | (ip->i_d.di_flags & (XFS_DIFLAG_RTINHERIT | XFS_DIFLAG_REALTIME))) { | |
837 | statp->f_blocks = sbp->sb_rblocks; | |
838 | statp->f_bavail = statp->f_bfree = | |
839 | sbp->sb_frextents * sbp->sb_rextsize; | |
840 | } | |
841 | ||
4ca488eb | 842 | return 0; |
1da177e4 LT |
843 | } |
844 | ||
d5db0f97 ES |
845 | STATIC void |
846 | xfs_save_resvblks(struct xfs_mount *mp) | |
847 | { | |
c8ce540d | 848 | uint64_t resblks = 0; |
d5db0f97 ES |
849 | |
850 | mp->m_resblks_save = mp->m_resblks; | |
851 | xfs_reserve_blocks(mp, &resblks, NULL); | |
852 | } | |
853 | ||
854 | STATIC void | |
855 | xfs_restore_resvblks(struct xfs_mount *mp) | |
856 | { | |
c8ce540d | 857 | uint64_t resblks; |
d5db0f97 ES |
858 | |
859 | if (mp->m_resblks_save) { | |
860 | resblks = mp->m_resblks_save; | |
861 | mp->m_resblks_save = 0; | |
862 | } else | |
863 | resblks = xfs_default_resblks(mp); | |
864 | ||
865 | xfs_reserve_blocks(mp, &resblks, NULL); | |
866 | } | |
867 | ||
8757c38f IK |
868 | /* |
869 | * Trigger writeback of all the dirty metadata in the file system. | |
870 | * | |
871 | * This ensures that the metadata is written to their location on disk rather | |
872 | * than just existing in transactions in the log. This means after a quiesce | |
873 | * there is no log replay required to write the inodes to disk - this is the | |
874 | * primary difference between a sync and a quiesce. | |
875 | * | |
b41b46c2 DC |
876 | * We cancel log work early here to ensure all transactions the log worker may |
877 | * run have finished before we clean up and log the superblock and write an | |
878 | * unmount record. The unfreeze process is responsible for restarting the log | |
879 | * worker correctly. | |
8757c38f IK |
880 | */ |
881 | void | |
882 | xfs_quiesce_attr( | |
883 | struct xfs_mount *mp) | |
884 | { | |
885 | int error = 0; | |
886 | ||
b41b46c2 | 887 | cancel_delayed_work_sync(&mp->m_log->l_work); |
8757c38f IK |
888 | |
889 | /* force the log to unpin objects from the now complete transactions */ | |
890 | xfs_log_force(mp, XFS_LOG_SYNC); | |
891 | ||
8757c38f IK |
892 | |
893 | /* Push the superblock and write an unmount record */ | |
894 | error = xfs_log_sbcount(mp); | |
895 | if (error) | |
896 | xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. " | |
897 | "Frozen image may not be consistent."); | |
8757c38f IK |
898 | xfs_log_quiesce(mp); |
899 | } | |
900 | ||
901 | /* | |
902 | * Second stage of a freeze. The data is already frozen so we only | |
903 | * need to take care of the metadata. Once that's done sync the superblock | |
904 | * to the log to dirty it in case of a crash while frozen. This ensures that we | |
905 | * will recover the unlinked inode lists on the next mount. | |
906 | */ | |
907 | STATIC int | |
908 | xfs_fs_freeze( | |
909 | struct super_block *sb) | |
910 | { | |
911 | struct xfs_mount *mp = XFS_M(sb); | |
c3f2375b WL |
912 | unsigned int flags; |
913 | int ret; | |
8757c38f | 914 | |
c3f2375b WL |
915 | /* |
916 | * The filesystem is now frozen far enough that memory reclaim | |
917 | * cannot safely operate on the filesystem. Hence we need to | |
918 | * set a GFP_NOFS context here to avoid recursion deadlocks. | |
919 | */ | |
920 | flags = memalloc_nofs_save(); | |
8757c38f IK |
921 | xfs_stop_block_reaping(mp); |
922 | xfs_save_resvblks(mp); | |
923 | xfs_quiesce_attr(mp); | |
c3f2375b WL |
924 | ret = xfs_sync_sb(mp, true); |
925 | memalloc_nofs_restore(flags); | |
926 | return ret; | |
8757c38f IK |
927 | } |
928 | ||
929 | STATIC int | |
930 | xfs_fs_unfreeze( | |
931 | struct super_block *sb) | |
932 | { | |
933 | struct xfs_mount *mp = XFS_M(sb); | |
934 | ||
935 | xfs_restore_resvblks(mp); | |
936 | xfs_log_work_queue(mp); | |
937 | xfs_start_block_reaping(mp); | |
938 | return 0; | |
939 | } | |
940 | ||
941 | /* | |
942 | * This function fills in xfs_mount_t fields based on mount args. | |
943 | * Note: the superblock _has_ now been read in. | |
944 | */ | |
945 | STATIC int | |
946 | xfs_finish_flags( | |
947 | struct xfs_mount *mp) | |
948 | { | |
949 | int ronly = (mp->m_flags & XFS_MOUNT_RDONLY); | |
950 | ||
951 | /* Fail a mount where the logbuf is smaller than the log stripe */ | |
952 | if (xfs_sb_version_haslogv2(&mp->m_sb)) { | |
953 | if (mp->m_logbsize <= 0 && | |
954 | mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) { | |
955 | mp->m_logbsize = mp->m_sb.sb_logsunit; | |
956 | } else if (mp->m_logbsize > 0 && | |
957 | mp->m_logbsize < mp->m_sb.sb_logsunit) { | |
958 | xfs_warn(mp, | |
959 | "logbuf size must be greater than or equal to log stripe size"); | |
960 | return -EINVAL; | |
961 | } | |
962 | } else { | |
963 | /* Fail a mount if the logbuf is larger than 32K */ | |
964 | if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) { | |
965 | xfs_warn(mp, | |
966 | "logbuf size for version 1 logs must be 16K or 32K"); | |
967 | return -EINVAL; | |
968 | } | |
969 | } | |
970 | ||
971 | /* | |
972 | * V5 filesystems always use attr2 format for attributes. | |
973 | */ | |
974 | if (xfs_sb_version_hascrc(&mp->m_sb) && | |
975 | (mp->m_flags & XFS_MOUNT_NOATTR2)) { | |
976 | xfs_warn(mp, "Cannot mount a V5 filesystem as noattr2. " | |
977 | "attr2 is always enabled for V5 filesystems."); | |
978 | return -EINVAL; | |
979 | } | |
980 | ||
981 | /* | |
982 | * mkfs'ed attr2 will turn on attr2 mount unless explicitly | |
983 | * told by noattr2 to turn it off | |
984 | */ | |
985 | if (xfs_sb_version_hasattr2(&mp->m_sb) && | |
986 | !(mp->m_flags & XFS_MOUNT_NOATTR2)) | |
987 | mp->m_flags |= XFS_MOUNT_ATTR2; | |
988 | ||
989 | /* | |
990 | * prohibit r/w mounts of read-only filesystems | |
991 | */ | |
992 | if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) { | |
993 | xfs_warn(mp, | |
994 | "cannot mount a read-only filesystem as read-write"); | |
995 | return -EROFS; | |
996 | } | |
997 | ||
998 | if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) && | |
999 | (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) && | |
1000 | !xfs_sb_version_has_pquotino(&mp->m_sb)) { | |
1001 | xfs_warn(mp, | |
1002 | "Super block does not support project and group quota together"); | |
1003 | return -EINVAL; | |
1004 | } | |
1005 | ||
1006 | return 0; | |
1007 | } | |
1008 | ||
1009 | static int | |
1010 | xfs_init_percpu_counters( | |
1011 | struct xfs_mount *mp) | |
1012 | { | |
1013 | int error; | |
1014 | ||
1015 | error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL); | |
1016 | if (error) | |
1017 | return -ENOMEM; | |
1018 | ||
1019 | error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL); | |
1020 | if (error) | |
1021 | goto free_icount; | |
1022 | ||
1023 | error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL); | |
1024 | if (error) | |
1025 | goto free_ifree; | |
1026 | ||
1027 | error = percpu_counter_init(&mp->m_delalloc_blks, 0, GFP_KERNEL); | |
1028 | if (error) | |
1029 | goto free_fdblocks; | |
1030 | ||
1031 | return 0; | |
1032 | ||
1033 | free_fdblocks: | |
1034 | percpu_counter_destroy(&mp->m_fdblocks); | |
1035 | free_ifree: | |
1036 | percpu_counter_destroy(&mp->m_ifree); | |
1037 | free_icount: | |
1038 | percpu_counter_destroy(&mp->m_icount); | |
1039 | return -ENOMEM; | |
1040 | } | |
1041 | ||
1042 | void | |
1043 | xfs_reinit_percpu_counters( | |
1044 | struct xfs_mount *mp) | |
1045 | { | |
1046 | percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount); | |
1047 | percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree); | |
1048 | percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks); | |
1049 | } | |
1050 | ||
1051 | static void | |
1052 | xfs_destroy_percpu_counters( | |
1053 | struct xfs_mount *mp) | |
1054 | { | |
1055 | percpu_counter_destroy(&mp->m_icount); | |
1056 | percpu_counter_destroy(&mp->m_ifree); | |
1057 | percpu_counter_destroy(&mp->m_fdblocks); | |
1058 | ASSERT(XFS_FORCED_SHUTDOWN(mp) || | |
1059 | percpu_counter_sum(&mp->m_delalloc_blks) == 0); | |
1060 | percpu_counter_destroy(&mp->m_delalloc_blks); | |
1061 | } | |
1062 | ||
1063 | static void | |
1064 | xfs_fs_put_super( | |
1065 | struct super_block *sb) | |
1066 | { | |
1067 | struct xfs_mount *mp = XFS_M(sb); | |
1068 | ||
1069 | /* if ->fill_super failed, we have no mount to tear down */ | |
1070 | if (!sb->s_fs_info) | |
1071 | return; | |
1072 | ||
1073 | xfs_notice(mp, "Unmounting Filesystem"); | |
1074 | xfs_filestream_unmount(mp); | |
1075 | xfs_unmountfs(mp); | |
1076 | ||
1077 | xfs_freesb(mp); | |
1078 | free_percpu(mp->m_stats.xs_stats); | |
1079 | xfs_destroy_percpu_counters(mp); | |
1080 | xfs_destroy_mount_workqueues(mp); | |
1081 | xfs_close_devices(mp); | |
1082 | ||
1083 | sb->s_fs_info = NULL; | |
1084 | xfs_mount_free(mp); | |
1085 | } | |
1086 | ||
1087 | static long | |
1088 | xfs_fs_nr_cached_objects( | |
1089 | struct super_block *sb, | |
1090 | struct shrink_control *sc) | |
1091 | { | |
1092 | /* Paranoia: catch incorrect calls during mount setup or teardown */ | |
1093 | if (WARN_ON_ONCE(!sb->s_fs_info)) | |
1094 | return 0; | |
1095 | return xfs_reclaim_inodes_count(XFS_M(sb)); | |
1096 | } | |
1097 | ||
1098 | static long | |
1099 | xfs_fs_free_cached_objects( | |
1100 | struct super_block *sb, | |
1101 | struct shrink_control *sc) | |
c7eea6f7 | 1102 | { |
8757c38f IK |
1103 | return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan); |
1104 | } | |
c7eea6f7 | 1105 | |
8757c38f IK |
1106 | static const struct super_operations xfs_super_operations = { |
1107 | .alloc_inode = xfs_fs_alloc_inode, | |
1108 | .destroy_inode = xfs_fs_destroy_inode, | |
1109 | .dirty_inode = xfs_fs_dirty_inode, | |
1110 | .drop_inode = xfs_fs_drop_inode, | |
1111 | .put_super = xfs_fs_put_super, | |
1112 | .sync_fs = xfs_fs_sync_fs, | |
1113 | .freeze_fs = xfs_fs_freeze, | |
1114 | .unfreeze_fs = xfs_fs_unfreeze, | |
1115 | .statfs = xfs_fs_statfs, | |
1116 | .show_options = xfs_fs_show_options, | |
1117 | .nr_cached_objects = xfs_fs_nr_cached_objects, | |
1118 | .free_cached_objects = xfs_fs_free_cached_objects, | |
1119 | }; | |
c7eea6f7 | 1120 | |
8757c38f IK |
1121 | static int |
1122 | suffix_kstrtoint( | |
1123 | const char *s, | |
1124 | unsigned int base, | |
1125 | int *res) | |
1126 | { | |
1127 | int last, shift_left_factor = 0, _res; | |
1128 | char *value; | |
1129 | int ret = 0; | |
1130 | ||
1131 | value = kstrdup(s, GFP_KERNEL); | |
1132 | if (!value) | |
1133 | return -ENOMEM; | |
1134 | ||
1135 | last = strlen(value) - 1; | |
1136 | if (value[last] == 'K' || value[last] == 'k') { | |
1137 | shift_left_factor = 10; | |
1138 | value[last] = '\0'; | |
1139 | } | |
1140 | if (value[last] == 'M' || value[last] == 'm') { | |
1141 | shift_left_factor = 20; | |
1142 | value[last] = '\0'; | |
1143 | } | |
1144 | if (value[last] == 'G' || value[last] == 'g') { | |
1145 | shift_left_factor = 30; | |
1146 | value[last] = '\0'; | |
1147 | } | |
1148 | ||
1149 | if (kstrtoint(value, base, &_res)) | |
1150 | ret = -EINVAL; | |
1151 | kfree(value); | |
1152 | *res = _res << shift_left_factor; | |
1153 | return ret; | |
c7eea6f7 DC |
1154 | } |
1155 | ||
9909c4aa | 1156 | /* |
8757c38f IK |
1157 | * Set mount state from a mount option. |
1158 | * | |
1159 | * NOTE: mp->m_super is NULL here! | |
9909c4aa | 1160 | */ |
8757c38f IK |
1161 | static int |
1162 | xfs_fc_parse_param( | |
1163 | struct fs_context *fc, | |
1164 | struct fs_parameter *param) | |
1da177e4 | 1165 | { |
8757c38f IK |
1166 | struct xfs_mount *mp = fc->s_fs_info; |
1167 | struct fs_parse_result result; | |
1168 | int size = 0; | |
1169 | int opt; | |
9909c4aa | 1170 | |
d7167b14 | 1171 | opt = fs_parse(fc, xfs_fs_parameters, param, &result); |
8757c38f IK |
1172 | if (opt < 0) |
1173 | return opt; | |
1da177e4 | 1174 | |
8757c38f IK |
1175 | switch (opt) { |
1176 | case Opt_logbufs: | |
1177 | mp->m_logbufs = result.uint_32; | |
1178 | return 0; | |
1179 | case Opt_logbsize: | |
1180 | if (suffix_kstrtoint(param->string, 10, &mp->m_logbsize)) | |
1181 | return -EINVAL; | |
1182 | return 0; | |
1183 | case Opt_logdev: | |
1184 | kfree(mp->m_logname); | |
1185 | mp->m_logname = kstrdup(param->string, GFP_KERNEL); | |
1186 | if (!mp->m_logname) | |
1187 | return -ENOMEM; | |
1188 | return 0; | |
1189 | case Opt_rtdev: | |
1190 | kfree(mp->m_rtname); | |
1191 | mp->m_rtname = kstrdup(param->string, GFP_KERNEL); | |
1192 | if (!mp->m_rtname) | |
1193 | return -ENOMEM; | |
1194 | return 0; | |
1195 | case Opt_allocsize: | |
1196 | if (suffix_kstrtoint(param->string, 10, &size)) | |
1197 | return -EINVAL; | |
1198 | mp->m_allocsize_log = ffs(size) - 1; | |
1199 | mp->m_flags |= XFS_MOUNT_ALLOCSIZE; | |
1200 | return 0; | |
1201 | case Opt_grpid: | |
1202 | case Opt_bsdgroups: | |
1203 | mp->m_flags |= XFS_MOUNT_GRPID; | |
1204 | return 0; | |
1205 | case Opt_nogrpid: | |
1206 | case Opt_sysvgroups: | |
1207 | mp->m_flags &= ~XFS_MOUNT_GRPID; | |
1208 | return 0; | |
1209 | case Opt_wsync: | |
1210 | mp->m_flags |= XFS_MOUNT_WSYNC; | |
1211 | return 0; | |
1212 | case Opt_norecovery: | |
1213 | mp->m_flags |= XFS_MOUNT_NORECOVERY; | |
1214 | return 0; | |
1215 | case Opt_noalign: | |
1216 | mp->m_flags |= XFS_MOUNT_NOALIGN; | |
1217 | return 0; | |
1218 | case Opt_swalloc: | |
1219 | mp->m_flags |= XFS_MOUNT_SWALLOC; | |
1220 | return 0; | |
1221 | case Opt_sunit: | |
1222 | mp->m_dalign = result.uint_32; | |
1223 | return 0; | |
1224 | case Opt_swidth: | |
1225 | mp->m_swidth = result.uint_32; | |
1226 | return 0; | |
1227 | case Opt_inode32: | |
1228 | mp->m_flags |= XFS_MOUNT_SMALL_INUMS; | |
1229 | return 0; | |
1230 | case Opt_inode64: | |
1231 | mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS; | |
1232 | return 0; | |
1233 | case Opt_nouuid: | |
1234 | mp->m_flags |= XFS_MOUNT_NOUUID; | |
1235 | return 0; | |
8757c38f IK |
1236 | case Opt_largeio: |
1237 | mp->m_flags |= XFS_MOUNT_LARGEIO; | |
1238 | return 0; | |
1239 | case Opt_nolargeio: | |
1240 | mp->m_flags &= ~XFS_MOUNT_LARGEIO; | |
1241 | return 0; | |
8757c38f IK |
1242 | case Opt_filestreams: |
1243 | mp->m_flags |= XFS_MOUNT_FILESTREAMS; | |
1244 | return 0; | |
1245 | case Opt_noquota: | |
1246 | mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT; | |
1247 | mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD; | |
1248 | mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE; | |
1249 | return 0; | |
1250 | case Opt_quota: | |
1251 | case Opt_uquota: | |
1252 | case Opt_usrquota: | |
1253 | mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE | | |
1254 | XFS_UQUOTA_ENFD); | |
1255 | return 0; | |
1256 | case Opt_qnoenforce: | |
1257 | case Opt_uqnoenforce: | |
1258 | mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE); | |
1259 | mp->m_qflags &= ~XFS_UQUOTA_ENFD; | |
1260 | return 0; | |
1261 | case Opt_pquota: | |
1262 | case Opt_prjquota: | |
1263 | mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE | | |
1264 | XFS_PQUOTA_ENFD); | |
1265 | return 0; | |
1266 | case Opt_pqnoenforce: | |
1267 | mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE); | |
1268 | mp->m_qflags &= ~XFS_PQUOTA_ENFD; | |
1269 | return 0; | |
1270 | case Opt_gquota: | |
1271 | case Opt_grpquota: | |
1272 | mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE | | |
1273 | XFS_GQUOTA_ENFD); | |
1274 | return 0; | |
1275 | case Opt_gqnoenforce: | |
1276 | mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE); | |
1277 | mp->m_qflags &= ~XFS_GQUOTA_ENFD; | |
1278 | return 0; | |
1279 | case Opt_discard: | |
1280 | mp->m_flags |= XFS_MOUNT_DISCARD; | |
1281 | return 0; | |
1282 | case Opt_nodiscard: | |
1283 | mp->m_flags &= ~XFS_MOUNT_DISCARD; | |
1284 | return 0; | |
1285 | #ifdef CONFIG_FS_DAX | |
1286 | case Opt_dax: | |
8d6c3446 IW |
1287 | xfs_mount_set_dax_mode(mp, XFS_DAX_ALWAYS); |
1288 | return 0; | |
1289 | case Opt_dax_enum: | |
1290 | xfs_mount_set_dax_mode(mp, result.uint_32); | |
8757c38f IK |
1291 | return 0; |
1292 | #endif | |
c23c393e PR |
1293 | /* Following mount options will be removed in September 2025 */ |
1294 | case Opt_ikeep: | |
1295 | xfs_warn(mp, "%s mount option is deprecated.", param->key); | |
1296 | mp->m_flags |= XFS_MOUNT_IKEEP; | |
1297 | return 0; | |
1298 | case Opt_noikeep: | |
1299 | xfs_warn(mp, "%s mount option is deprecated.", param->key); | |
1300 | mp->m_flags &= ~XFS_MOUNT_IKEEP; | |
1301 | return 0; | |
1302 | case Opt_attr2: | |
1303 | xfs_warn(mp, "%s mount option is deprecated.", param->key); | |
1304 | mp->m_flags |= XFS_MOUNT_ATTR2; | |
1305 | return 0; | |
1306 | case Opt_noattr2: | |
1307 | xfs_warn(mp, "%s mount option is deprecated.", param->key); | |
1308 | mp->m_flags &= ~XFS_MOUNT_ATTR2; | |
1309 | mp->m_flags |= XFS_MOUNT_NOATTR2; | |
1310 | return 0; | |
8757c38f IK |
1311 | default: |
1312 | xfs_warn(mp, "unknown mount option [%s].", param->key); | |
1313 | return -EINVAL; | |
1314 | } | |
d5db0f97 | 1315 | |
d5db0f97 ES |
1316 | return 0; |
1317 | } | |
1318 | ||
8757c38f IK |
1319 | static int |
1320 | xfs_fc_validate_params( | |
f8f15e42 CH |
1321 | struct xfs_mount *mp) |
1322 | { | |
d3eaace8 | 1323 | /* |
8757c38f | 1324 | * no recovery flag requires a read-only mount |
d3eaace8 | 1325 | */ |
8757c38f IK |
1326 | if ((mp->m_flags & XFS_MOUNT_NORECOVERY) && |
1327 | !(mp->m_flags & XFS_MOUNT_RDONLY)) { | |
1328 | xfs_warn(mp, "no-recovery mounts must be read-only."); | |
2451337d | 1329 | return -EINVAL; |
d3eaace8 DC |
1330 | } |
1331 | ||
8757c38f IK |
1332 | if ((mp->m_flags & XFS_MOUNT_NOALIGN) && |
1333 | (mp->m_dalign || mp->m_swidth)) { | |
4f10700a | 1334 | xfs_warn(mp, |
8757c38f IK |
1335 | "sunit and swidth options incompatible with the noalign option"); |
1336 | return -EINVAL; | |
f8f15e42 CH |
1337 | } |
1338 | ||
8757c38f IK |
1339 | if (!IS_ENABLED(CONFIG_XFS_QUOTA) && mp->m_qflags != 0) { |
1340 | xfs_warn(mp, "quota support not available in this kernel."); | |
2451337d | 1341 | return -EINVAL; |
d892d586 CS |
1342 | } |
1343 | ||
8757c38f IK |
1344 | if ((mp->m_dalign && !mp->m_swidth) || |
1345 | (!mp->m_dalign && mp->m_swidth)) { | |
1346 | xfs_warn(mp, "sunit and swidth must be specified together"); | |
1347 | return -EINVAL; | |
1348 | } | |
5681ca40 | 1349 | |
8757c38f IK |
1350 | if (mp->m_dalign && (mp->m_swidth % mp->m_dalign != 0)) { |
1351 | xfs_warn(mp, | |
1352 | "stripe width (%d) must be a multiple of the stripe unit (%d)", | |
1353 | mp->m_swidth, mp->m_dalign); | |
1354 | return -EINVAL; | |
1355 | } | |
5681ca40 | 1356 | |
8757c38f IK |
1357 | if (mp->m_logbufs != -1 && |
1358 | mp->m_logbufs != 0 && | |
1359 | (mp->m_logbufs < XLOG_MIN_ICLOGS || | |
1360 | mp->m_logbufs > XLOG_MAX_ICLOGS)) { | |
1361 | xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]", | |
1362 | mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS); | |
1363 | return -EINVAL; | |
1364 | } | |
5681ca40 | 1365 | |
8757c38f IK |
1366 | if (mp->m_logbsize != -1 && |
1367 | mp->m_logbsize != 0 && | |
1368 | (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE || | |
1369 | mp->m_logbsize > XLOG_MAX_RECORD_BSIZE || | |
1370 | !is_power_of_2(mp->m_logbsize))) { | |
1371 | xfs_warn(mp, | |
1372 | "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]", | |
1373 | mp->m_logbsize); | |
1374 | return -EINVAL; | |
1375 | } | |
5681ca40 | 1376 | |
8757c38f IK |
1377 | if ((mp->m_flags & XFS_MOUNT_ALLOCSIZE) && |
1378 | (mp->m_allocsize_log > XFS_MAX_IO_LOG || | |
1379 | mp->m_allocsize_log < XFS_MIN_IO_LOG)) { | |
1380 | xfs_warn(mp, "invalid log iosize: %d [not %d-%d]", | |
1381 | mp->m_allocsize_log, XFS_MIN_IO_LOG, XFS_MAX_IO_LOG); | |
1382 | return -EINVAL; | |
1383 | } | |
9fe82b8c | 1384 | |
5681ca40 | 1385 | return 0; |
72c44e35 | 1386 | } |
74394496 | 1387 | |
73e5fff9 IK |
1388 | static int |
1389 | xfs_fc_fill_super( | |
72c44e35 | 1390 | struct super_block *sb, |
73e5fff9 | 1391 | struct fs_context *fc) |
72c44e35 | 1392 | { |
73e5fff9 | 1393 | struct xfs_mount *mp = sb->s_fs_info; |
72c44e35 | 1394 | struct inode *root; |
0279c71f | 1395 | int flags = 0, error; |
72c44e35 | 1396 | |
7c89fcb2 | 1397 | mp->m_super = sb; |
1da177e4 | 1398 | |
73e5fff9 | 1399 | error = xfs_fc_validate_params(mp); |
745f6919 | 1400 | if (error) |
e1d3d218 | 1401 | goto out_free_names; |
1da177e4 LT |
1402 | |
1403 | sb_min_blocksize(sb, BBSIZE); | |
0ec58516 | 1404 | sb->s_xattr = xfs_xattr_handlers; |
a50cd269 | 1405 | sb->s_export_op = &xfs_export_operations; |
fcafb71b | 1406 | #ifdef CONFIG_XFS_QUOTA |
a50cd269 | 1407 | sb->s_qcop = &xfs_quotactl_operations; |
17ef4fdd | 1408 | sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; |
fcafb71b | 1409 | #endif |
a50cd269 | 1410 | sb->s_op = &xfs_super_operations; |
1da177e4 | 1411 | |
dae5cd81 DC |
1412 | /* |
1413 | * Delay mount work if the debug hook is set. This is debug | |
1414 | * instrumention to coordinate simulation of xfs mount failures with | |
1415 | * VFS superblock operations | |
1416 | */ | |
1417 | if (xfs_globals.mount_delay) { | |
1418 | xfs_notice(mp, "Delaying mount for %d seconds.", | |
1419 | xfs_globals.mount_delay); | |
1420 | msleep(xfs_globals.mount_delay * 1000); | |
1421 | } | |
1422 | ||
73e5fff9 | 1423 | if (fc->sb_flags & SB_SILENT) |
f8f15e42 CH |
1424 | flags |= XFS_MFSI_QUIET; |
1425 | ||
9d565ffa | 1426 | error = xfs_open_devices(mp); |
19f354d4 | 1427 | if (error) |
e1d3d218 | 1428 | goto out_free_names; |
f8f15e42 | 1429 | |
2451337d | 1430 | error = xfs_init_mount_workqueues(mp); |
61ba35de CH |
1431 | if (error) |
1432 | goto out_close_devices; | |
c962fb79 | 1433 | |
5681ca40 | 1434 | error = xfs_init_percpu_counters(mp); |
aa6bf01d CH |
1435 | if (error) |
1436 | goto out_destroy_workqueues; | |
1437 | ||
225e4635 BD |
1438 | /* Allocate stats memory before we do operations that might use it */ |
1439 | mp->m_stats.xs_stats = alloc_percpu(struct xfsstats); | |
1440 | if (!mp->m_stats.xs_stats) { | |
f9d460b3 | 1441 | error = -ENOMEM; |
225e4635 BD |
1442 | goto out_destroy_counters; |
1443 | } | |
1444 | ||
f8f15e42 CH |
1445 | error = xfs_readsb(mp, flags); |
1446 | if (error) | |
225e4635 | 1447 | goto out_free_stats; |
9d565ffa CH |
1448 | |
1449 | error = xfs_finish_flags(mp); | |
f8f15e42 | 1450 | if (error) |
effa2eda | 1451 | goto out_free_sb; |
f8f15e42 | 1452 | |
e34b562c | 1453 | error = xfs_setup_devices(mp); |
19f354d4 | 1454 | if (error) |
effa2eda | 1455 | goto out_free_sb; |
f8f15e42 | 1456 | |
b96cb835 DW |
1457 | /* V4 support is undergoing deprecation. */ |
1458 | if (!xfs_sb_version_hascrc(&mp->m_sb)) { | |
1459 | #ifdef CONFIG_XFS_SUPPORT_V4 | |
1460 | xfs_warn_once(mp, | |
1461 | "Deprecated V4 format (crc=0) will not be supported after September 2030."); | |
1462 | #else | |
1463 | xfs_warn(mp, | |
1464 | "Deprecated V4 format (crc=0) not supported by kernel."); | |
1465 | error = -EINVAL; | |
1466 | goto out_free_sb; | |
1467 | #endif | |
1468 | } | |
1469 | ||
80c720b8 DW |
1470 | /* Filesystem claims it needs repair, so refuse the mount. */ |
1471 | if (xfs_sb_version_needsrepair(&mp->m_sb)) { | |
1472 | xfs_warn(mp, "Filesystem needs repair. Please run xfs_repair."); | |
1473 | error = -EFSCORRUPTED; | |
1474 | goto out_free_sb; | |
1475 | } | |
1476 | ||
3945ae03 DW |
1477 | /* |
1478 | * Don't touch the filesystem if a user tool thinks it owns the primary | |
1479 | * superblock. mkfs doesn't clear the flag from secondary supers, so | |
1480 | * we don't check them at all. | |
1481 | */ | |
1482 | if (mp->m_sb.sb_inprogress) { | |
1483 | xfs_warn(mp, "Offline file system operation in progress!"); | |
1484 | error = -EFSCORRUPTED; | |
1485 | goto out_free_sb; | |
1486 | } | |
1487 | ||
1488 | /* | |
1489 | * Until this is fixed only page-sized or smaller data blocks work. | |
1490 | */ | |
1491 | if (mp->m_sb.sb_blocksize > PAGE_SIZE) { | |
1492 | xfs_warn(mp, | |
1493 | "File system with blocksize %d bytes. " | |
1494 | "Only pagesize (%ld) or less will currently work.", | |
1495 | mp->m_sb.sb_blocksize, PAGE_SIZE); | |
1496 | error = -ENOSYS; | |
1497 | goto out_free_sb; | |
1498 | } | |
1499 | ||
1500 | /* Ensure this filesystem fits in the page cache limits */ | |
1501 | if (xfs_sb_validate_fsb_count(&mp->m_sb, mp->m_sb.sb_dblocks) || | |
1502 | xfs_sb_validate_fsb_count(&mp->m_sb, mp->m_sb.sb_rblocks)) { | |
1503 | xfs_warn(mp, | |
1504 | "file system too large to be mounted on this system."); | |
1505 | error = -EFBIG; | |
1506 | goto out_free_sb; | |
1507 | } | |
1508 | ||
932befe3 DW |
1509 | /* |
1510 | * XFS block mappings use 54 bits to store the logical block offset. | |
1511 | * This should suffice to handle the maximum file size that the VFS | |
1512 | * supports (currently 2^63 bytes on 64-bit and ULONG_MAX << PAGE_SHIFT | |
1513 | * bytes on 32-bit), but as XFS and VFS have gotten the s_maxbytes | |
1514 | * calculation wrong on 32-bit kernels in the past, we'll add a WARN_ON | |
1515 | * to check this assertion. | |
1516 | * | |
1517 | * Avoid integer overflow by comparing the maximum bmbt offset to the | |
1518 | * maximum pagecache offset in units of fs blocks. | |
1519 | */ | |
1520 | if (XFS_B_TO_FSBT(mp, MAX_LFS_FILESIZE) > XFS_MAX_FILEOFF) { | |
1521 | xfs_warn(mp, | |
1522 | "MAX_LFS_FILESIZE block offset (%llu) exceeds extent map maximum (%llu)!", | |
1523 | XFS_B_TO_FSBT(mp, MAX_LFS_FILESIZE), | |
1524 | XFS_MAX_FILEOFF); | |
1525 | error = -EINVAL; | |
1526 | goto out_free_sb; | |
1527 | } | |
1528 | ||
f8f15e42 CH |
1529 | error = xfs_filestream_mount(mp); |
1530 | if (error) | |
effa2eda | 1531 | goto out_free_sb; |
f8f15e42 | 1532 | |
704b2907 DC |
1533 | /* |
1534 | * we must configure the block size in the superblock before we run the | |
1535 | * full mount process as the mount process can lookup and cache inodes. | |
704b2907 | 1536 | */ |
dddde68b | 1537 | sb->s_magic = XFS_SUPER_MAGIC; |
4ca488eb CH |
1538 | sb->s_blocksize = mp->m_sb.sb_blocksize; |
1539 | sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1; | |
932befe3 | 1540 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
8de52778 | 1541 | sb->s_max_links = XFS_MAXLINK; |
1da177e4 | 1542 | sb->s_time_gran = 1; |
f93e5436 DW |
1543 | if (xfs_sb_version_hasbigtime(&mp->m_sb)) { |
1544 | sb->s_time_min = xfs_bigtime_to_unix(XFS_BIGTIME_TIME_MIN); | |
1545 | sb->s_time_max = xfs_bigtime_to_unix(XFS_BIGTIME_TIME_MAX); | |
1546 | } else { | |
1547 | sb->s_time_min = XFS_LEGACY_TIME_MIN; | |
1548 | sb->s_time_max = XFS_LEGACY_TIME_MAX; | |
1549 | } | |
06dbf82b | 1550 | trace_xfs_inode_timestamp_range(mp, sb->s_time_min, sb->s_time_max); |
adfb5fb4 CH |
1551 | sb->s_iflags |= SB_I_CGROUPWB; |
1552 | ||
1da177e4 LT |
1553 | set_posix_acl_flag(sb); |
1554 | ||
dc037ad7 DC |
1555 | /* version 5 superblocks support inode version counters. */ |
1556 | if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5) | |
357fdad0 | 1557 | sb->s_flags |= SB_I_VERSION; |
dc037ad7 | 1558 | |
f93e5436 DW |
1559 | if (xfs_sb_version_hasbigtime(&mp->m_sb)) |
1560 | xfs_warn(mp, | |
1561 | "EXPERIMENTAL big timestamp feature in use. Use at your own risk!"); | |
1562 | ||
606723d9 | 1563 | if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS) { |
80660f20 | 1564 | bool rtdev_is_dax = false, datadev_is_dax; |
ba23cba9 | 1565 | |
cbe4dab1 | 1566 | xfs_warn(mp, |
1e937cdd TK |
1567 | "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); |
1568 | ||
80660f20 DJ |
1569 | datadev_is_dax = bdev_dax_supported(mp->m_ddev_targp->bt_bdev, |
1570 | sb->s_blocksize); | |
ba23cba9 | 1571 | if (mp->m_rtdev_targp) |
80660f20 DJ |
1572 | rtdev_is_dax = bdev_dax_supported( |
1573 | mp->m_rtdev_targp->bt_bdev, sb->s_blocksize); | |
1574 | if (!rtdev_is_dax && !datadev_is_dax) { | |
cbe4dab1 | 1575 | xfs_alert(mp, |
1e937cdd | 1576 | "DAX unsupported by block device. Turning off DAX."); |
8d6c3446 | 1577 | xfs_mount_set_dax_mode(mp, XFS_DAX_NEVER); |
cbe4dab1 | 1578 | } |
b6e03c10 | 1579 | if (xfs_sb_version_hasreflink(&mp->m_sb)) { |
e54b5bf9 | 1580 | xfs_alert(mp, |
1e369b0e | 1581 | "DAX and reflink cannot be used together!"); |
b6e03c10 DW |
1582 | error = -EINVAL; |
1583 | goto out_filestream_unmount; | |
1584 | } | |
cbe4dab1 DC |
1585 | } |
1586 | ||
1e6fa688 KN |
1587 | if (mp->m_flags & XFS_MOUNT_DISCARD) { |
1588 | struct request_queue *q = bdev_get_queue(sb->s_bdev); | |
1589 | ||
1590 | if (!blk_queue_discard(q)) { | |
1591 | xfs_warn(mp, "mounting with \"discard\" option, but " | |
1592 | "the device does not support discard"); | |
1593 | mp->m_flags &= ~XFS_MOUNT_DISCARD; | |
1594 | } | |
1595 | } | |
1596 | ||
66ae56a5 CH |
1597 | if (xfs_sb_version_hasreflink(&mp->m_sb)) { |
1598 | if (mp->m_sb.sb_rblocks) { | |
1599 | xfs_alert(mp, | |
c14632dd | 1600 | "reflink not compatible with realtime device!"); |
66ae56a5 CH |
1601 | error = -EINVAL; |
1602 | goto out_filestream_unmount; | |
1603 | } | |
1604 | ||
1605 | if (xfs_globals.always_cow) { | |
1606 | xfs_info(mp, "using DEBUG-only always_cow mode."); | |
1607 | mp->m_always_cow = true; | |
1608 | } | |
c14632dd DW |
1609 | } |
1610 | ||
76883f79 | 1611 | if (xfs_sb_version_hasrmapbt(&mp->m_sb) && mp->m_sb.sb_rblocks) { |
1c0607ac | 1612 | xfs_alert(mp, |
76883f79 DW |
1613 | "reverse mapping btree not compatible with realtime device!"); |
1614 | error = -EINVAL; | |
1615 | goto out_filestream_unmount; | |
738f57c1 | 1616 | } |
1c0607ac | 1617 | |
2a39946c DW |
1618 | if (xfs_sb_version_hasinobtcounts(&mp->m_sb)) |
1619 | xfs_warn(mp, | |
1620 | "EXPERIMENTAL inode btree counters feature in use. Use at your own risk!"); | |
1621 | ||
8a00ebe4 | 1622 | error = xfs_mountfs(mp); |
2bcf6e97 | 1623 | if (error) |
7e18530b | 1624 | goto out_filestream_unmount; |
704b2907 | 1625 | |
01651646 | 1626 | root = igrab(VFS_I(mp->m_rootip)); |
f3dcc13f | 1627 | if (!root) { |
2451337d | 1628 | error = -ENOENT; |
8a00ebe4 | 1629 | goto out_unmount; |
cbc89dcf | 1630 | } |
48fde701 | 1631 | sb->s_root = d_make_root(root); |
f3dcc13f | 1632 | if (!sb->s_root) { |
2451337d | 1633 | error = -ENOMEM; |
8a00ebe4 | 1634 | goto out_unmount; |
1da177e4 | 1635 | } |
74394496 | 1636 | |
1da177e4 | 1637 | return 0; |
33c7a2bc | 1638 | |
7e18530b | 1639 | out_filestream_unmount: |
120226c1 | 1640 | xfs_filestream_unmount(mp); |
effa2eda CH |
1641 | out_free_sb: |
1642 | xfs_freesb(mp); | |
225e4635 BD |
1643 | out_free_stats: |
1644 | free_percpu(mp->m_stats.xs_stats); | |
9d565ffa | 1645 | out_destroy_counters: |
5681ca40 | 1646 | xfs_destroy_percpu_counters(mp); |
225e4635 | 1647 | out_destroy_workqueues: |
aa6bf01d | 1648 | xfs_destroy_mount_workqueues(mp); |
61ba35de | 1649 | out_close_devices: |
19f354d4 | 1650 | xfs_close_devices(mp); |
e1d3d218 | 1651 | out_free_names: |
c9fbd7bb | 1652 | sb->s_fs_info = NULL; |
a943f372 | 1653 | xfs_mount_free(mp); |
2451337d | 1654 | return error; |
f8f15e42 | 1655 | |
2bcf6e97 | 1656 | out_unmount: |
e48ad316 | 1657 | xfs_filestream_unmount(mp); |
19f354d4 | 1658 | xfs_unmountfs(mp); |
6203300e | 1659 | goto out_free_sb; |
1da177e4 LT |
1660 | } |
1661 | ||
73e5fff9 IK |
1662 | static int |
1663 | xfs_fc_get_tree( | |
1664 | struct fs_context *fc) | |
1665 | { | |
1666 | return get_tree_bdev(fc, xfs_fc_fill_super); | |
1667 | } | |
1668 | ||
63cd1e9b IK |
1669 | static int |
1670 | xfs_remount_rw( | |
1671 | struct xfs_mount *mp) | |
1672 | { | |
1673 | struct xfs_sb *sbp = &mp->m_sb; | |
1674 | int error; | |
1675 | ||
1676 | if (mp->m_flags & XFS_MOUNT_NORECOVERY) { | |
1677 | xfs_warn(mp, | |
1678 | "ro->rw transition prohibited on norecovery mount"); | |
1679 | return -EINVAL; | |
1680 | } | |
1681 | ||
1682 | if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 && | |
1683 | xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) { | |
1684 | xfs_warn(mp, | |
1685 | "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem", | |
1686 | (sbp->sb_features_ro_compat & | |
1687 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)); | |
1688 | return -EINVAL; | |
1689 | } | |
1690 | ||
1691 | mp->m_flags &= ~XFS_MOUNT_RDONLY; | |
1692 | ||
1693 | /* | |
1694 | * If this is the first remount to writeable state we might have some | |
1695 | * superblock changes to update. | |
1696 | */ | |
1697 | if (mp->m_update_sb) { | |
1698 | error = xfs_sync_sb(mp, false); | |
1699 | if (error) { | |
1700 | xfs_warn(mp, "failed to write sb changes"); | |
1701 | return error; | |
1702 | } | |
1703 | mp->m_update_sb = false; | |
1704 | } | |
1705 | ||
1706 | /* | |
1707 | * Fill out the reserve pool if it is empty. Use the stashed value if | |
1708 | * it is non-zero, otherwise go with the default. | |
1709 | */ | |
1710 | xfs_restore_resvblks(mp); | |
1711 | xfs_log_work_queue(mp); | |
1712 | ||
1713 | /* Recover any CoW blocks that never got remapped. */ | |
1714 | error = xfs_reflink_recover_cow(mp); | |
1715 | if (error) { | |
1716 | xfs_err(mp, | |
1717 | "Error %d recovering leftover CoW allocations.", error); | |
7f6bcf7c | 1718 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); |
63cd1e9b IK |
1719 | return error; |
1720 | } | |
1721 | xfs_start_block_reaping(mp); | |
1722 | ||
1723 | /* Create the per-AG metadata reservation pool .*/ | |
1724 | error = xfs_fs_reserve_ag_blocks(mp); | |
1725 | if (error && error != -ENOSPC) | |
1726 | return error; | |
1727 | ||
1728 | return 0; | |
1729 | } | |
1730 | ||
1731 | static int | |
1732 | xfs_remount_ro( | |
1733 | struct xfs_mount *mp) | |
1734 | { | |
1735 | int error; | |
1736 | ||
1737 | /* | |
1738 | * Cancel background eofb scanning so it cannot race with the final | |
1739 | * log force+buftarg wait and deadlock the remount. | |
1740 | */ | |
1741 | xfs_stop_block_reaping(mp); | |
1742 | ||
1743 | /* Get rid of any leftover CoW reservations... */ | |
1744 | error = xfs_icache_free_cowblocks(mp, NULL); | |
1745 | if (error) { | |
1746 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); | |
1747 | return error; | |
1748 | } | |
1749 | ||
1750 | /* Free the per-AG metadata reservation pool. */ | |
1751 | error = xfs_fs_unreserve_ag_blocks(mp); | |
1752 | if (error) { | |
1753 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); | |
1754 | return error; | |
1755 | } | |
1756 | ||
1757 | /* | |
1758 | * Before we sync the metadata, we need to free up the reserve block | |
1759 | * pool so that the used block count in the superblock on disk is | |
1760 | * correct at the end of the remount. Stash the current* reserve pool | |
1761 | * size so that if we get remounted rw, we can return it to the same | |
1762 | * size. | |
1763 | */ | |
1764 | xfs_save_resvblks(mp); | |
1765 | ||
1766 | xfs_quiesce_attr(mp); | |
1767 | mp->m_flags |= XFS_MOUNT_RDONLY; | |
1768 | ||
1769 | return 0; | |
1770 | } | |
1771 | ||
1772 | /* | |
1773 | * Logically we would return an error here to prevent users from believing | |
1774 | * they might have changed mount options using remount which can't be changed. | |
1775 | * | |
1776 | * But unfortunately mount(8) adds all options from mtab and fstab to the mount | |
1777 | * arguments in some cases so we can't blindly reject options, but have to | |
1778 | * check for each specified option if it actually differs from the currently | |
1779 | * set option and only reject it if that's the case. | |
1780 | * | |
1781 | * Until that is implemented we return success for every remount request, and | |
1782 | * silently ignore all options that we can't actually change. | |
1783 | */ | |
1784 | static int | |
1785 | xfs_fc_reconfigure( | |
1786 | struct fs_context *fc) | |
1787 | { | |
1788 | struct xfs_mount *mp = XFS_M(fc->root->d_sb); | |
1789 | struct xfs_mount *new_mp = fc->s_fs_info; | |
1790 | xfs_sb_t *sbp = &mp->m_sb; | |
1791 | int flags = fc->sb_flags; | |
1792 | int error; | |
1793 | ||
4750a171 ES |
1794 | /* version 5 superblocks always support version counters. */ |
1795 | if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5) | |
1796 | fc->sb_flags |= SB_I_VERSION; | |
1797 | ||
63cd1e9b IK |
1798 | error = xfs_fc_validate_params(new_mp); |
1799 | if (error) | |
1800 | return error; | |
1801 | ||
1802 | sync_filesystem(mp->m_super); | |
1803 | ||
1804 | /* inode32 -> inode64 */ | |
1805 | if ((mp->m_flags & XFS_MOUNT_SMALL_INUMS) && | |
1806 | !(new_mp->m_flags & XFS_MOUNT_SMALL_INUMS)) { | |
1807 | mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS; | |
1808 | mp->m_maxagi = xfs_set_inode_alloc(mp, sbp->sb_agcount); | |
1809 | } | |
1810 | ||
1811 | /* inode64 -> inode32 */ | |
1812 | if (!(mp->m_flags & XFS_MOUNT_SMALL_INUMS) && | |
1813 | (new_mp->m_flags & XFS_MOUNT_SMALL_INUMS)) { | |
1814 | mp->m_flags |= XFS_MOUNT_SMALL_INUMS; | |
1815 | mp->m_maxagi = xfs_set_inode_alloc(mp, sbp->sb_agcount); | |
1816 | } | |
1817 | ||
1818 | /* ro -> rw */ | |
1819 | if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(flags & SB_RDONLY)) { | |
1820 | error = xfs_remount_rw(mp); | |
1821 | if (error) | |
1822 | return error; | |
1823 | } | |
1824 | ||
1825 | /* rw -> ro */ | |
1826 | if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (flags & SB_RDONLY)) { | |
1827 | error = xfs_remount_ro(mp); | |
1828 | if (error) | |
1829 | return error; | |
1830 | } | |
1831 | ||
1832 | return 0; | |
1833 | } | |
1834 | ||
73e5fff9 IK |
1835 | static void xfs_fc_free( |
1836 | struct fs_context *fc) | |
1837 | { | |
1838 | struct xfs_mount *mp = fc->s_fs_info; | |
1839 | ||
1840 | /* | |
1841 | * mp is stored in the fs_context when it is initialized. | |
1842 | * mp is transferred to the superblock on a successful mount, | |
1843 | * but if an error occurs before the transfer we have to free | |
1844 | * it here. | |
1845 | */ | |
1846 | if (mp) | |
1847 | xfs_mount_free(mp); | |
1848 | } | |
1849 | ||
1850 | static const struct fs_context_operations xfs_context_ops = { | |
1851 | .parse_param = xfs_fc_parse_param, | |
1852 | .get_tree = xfs_fc_get_tree, | |
1853 | .reconfigure = xfs_fc_reconfigure, | |
1854 | .free = xfs_fc_free, | |
1855 | }; | |
1856 | ||
1857 | static int xfs_init_fs_context( | |
1858 | struct fs_context *fc) | |
1859 | { | |
1860 | struct xfs_mount *mp; | |
1861 | ||
50f83009 | 1862 | mp = kmem_alloc(sizeof(struct xfs_mount), KM_ZERO); |
73e5fff9 IK |
1863 | if (!mp) |
1864 | return -ENOMEM; | |
1865 | ||
50f83009 IK |
1866 | spin_lock_init(&mp->m_sb_lock); |
1867 | spin_lock_init(&mp->m_agirotor_lock); | |
1868 | INIT_RADIX_TREE(&mp->m_perag_tree, GFP_ATOMIC); | |
1869 | spin_lock_init(&mp->m_perag_lock); | |
1870 | mutex_init(&mp->m_growlock); | |
f0f7a674 | 1871 | INIT_WORK(&mp->m_flush_inodes_work, xfs_flush_inodes_worker); |
50f83009 IK |
1872 | INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker); |
1873 | INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker); | |
1874 | INIT_DELAYED_WORK(&mp->m_cowblocks_work, xfs_cowblocks_worker); | |
1875 | mp->m_kobj.kobject.kset = xfs_kset; | |
1876 | /* | |
1877 | * We don't create the finobt per-ag space reservation until after log | |
1878 | * recovery, so we must set this to true so that an ifree transaction | |
1879 | * started during log recovery will not depend on space reservations | |
1880 | * for finobt expansion. | |
1881 | */ | |
1882 | mp->m_finobt_nores = true; | |
1883 | ||
73e5fff9 IK |
1884 | /* |
1885 | * These can be overridden by the mount option parsing. | |
1886 | */ | |
1887 | mp->m_logbufs = -1; | |
1888 | mp->m_logbsize = -1; | |
1889 | mp->m_allocsize_log = 16; /* 64k */ | |
1890 | ||
1891 | /* | |
1892 | * Copy binary VFS mount flags we are interested in. | |
1893 | */ | |
1894 | if (fc->sb_flags & SB_RDONLY) | |
1895 | mp->m_flags |= XFS_MOUNT_RDONLY; | |
1896 | if (fc->sb_flags & SB_DIRSYNC) | |
1897 | mp->m_flags |= XFS_MOUNT_DIRSYNC; | |
1898 | if (fc->sb_flags & SB_SYNCHRONOUS) | |
1899 | mp->m_flags |= XFS_MOUNT_WSYNC; | |
1900 | ||
1901 | fc->s_fs_info = mp; | |
1902 | fc->ops = &xfs_context_ops; | |
1903 | ||
1904 | return 0; | |
1905 | } | |
1906 | ||
5085b607 | 1907 | static struct file_system_type xfs_fs_type = { |
1da177e4 LT |
1908 | .owner = THIS_MODULE, |
1909 | .name = "xfs", | |
73e5fff9 | 1910 | .init_fs_context = xfs_init_fs_context, |
d7167b14 | 1911 | .parameters = xfs_fs_parameters, |
1da177e4 LT |
1912 | .kill_sb = kill_block_super, |
1913 | .fs_flags = FS_REQUIRES_DEV, | |
1914 | }; | |
7f78e035 | 1915 | MODULE_ALIAS_FS("xfs"); |
1da177e4 | 1916 | |
9f8868ff CH |
1917 | STATIC int __init |
1918 | xfs_init_zones(void) | |
1919 | { | |
b1231760 CM |
1920 | xfs_log_ticket_zone = kmem_cache_create("xfs_log_ticket", |
1921 | sizeof(struct xlog_ticket), | |
1922 | 0, 0, NULL); | |
9f8868ff | 1923 | if (!xfs_log_ticket_zone) |
598ecfba | 1924 | goto out; |
9f8868ff | 1925 | |
b1231760 CM |
1926 | xfs_bmap_free_item_zone = kmem_cache_create("xfs_bmap_free_item", |
1927 | sizeof(struct xfs_extent_free_item), | |
1928 | 0, 0, NULL); | |
9f8868ff CH |
1929 | if (!xfs_bmap_free_item_zone) |
1930 | goto out_destroy_log_ticket_zone; | |
bf904248 | 1931 | |
b1231760 CM |
1932 | xfs_btree_cur_zone = kmem_cache_create("xfs_btree_cur", |
1933 | sizeof(struct xfs_btree_cur), | |
1934 | 0, 0, NULL); | |
9f8868ff CH |
1935 | if (!xfs_btree_cur_zone) |
1936 | goto out_destroy_bmap_free_item_zone; | |
1937 | ||
b1231760 CM |
1938 | xfs_da_state_zone = kmem_cache_create("xfs_da_state", |
1939 | sizeof(struct xfs_da_state), | |
1940 | 0, 0, NULL); | |
9f8868ff CH |
1941 | if (!xfs_da_state_zone) |
1942 | goto out_destroy_btree_cur_zone; | |
1943 | ||
b1231760 CM |
1944 | xfs_ifork_zone = kmem_cache_create("xfs_ifork", |
1945 | sizeof(struct xfs_ifork), | |
1946 | 0, 0, NULL); | |
9f8868ff | 1947 | if (!xfs_ifork_zone) |
1d9025e5 | 1948 | goto out_destroy_da_state_zone; |
9f8868ff | 1949 | |
b1231760 CM |
1950 | xfs_trans_zone = kmem_cache_create("xf_trans", |
1951 | sizeof(struct xfs_trans), | |
1952 | 0, 0, NULL); | |
9f8868ff CH |
1953 | if (!xfs_trans_zone) |
1954 | goto out_destroy_ifork_zone; | |
1955 | ||
e98c414f | 1956 | |
9f8868ff CH |
1957 | /* |
1958 | * The size of the zone allocated buf log item is the maximum | |
1959 | * size possible under XFS. This wastes a little bit of memory, | |
1960 | * but it is much faster. | |
1961 | */ | |
b1231760 CM |
1962 | xfs_buf_item_zone = kmem_cache_create("xfs_buf_item", |
1963 | sizeof(struct xfs_buf_log_item), | |
1964 | 0, 0, NULL); | |
9f8868ff | 1965 | if (!xfs_buf_item_zone) |
e6631f85 | 1966 | goto out_destroy_trans_zone; |
9f8868ff | 1967 | |
b1231760 CM |
1968 | xfs_efd_zone = kmem_cache_create("xfs_efd_item", |
1969 | (sizeof(struct xfs_efd_log_item) + | |
1970 | (XFS_EFD_MAX_FAST_EXTENTS - 1) * | |
1971 | sizeof(struct xfs_extent)), | |
1972 | 0, 0, NULL); | |
9f8868ff CH |
1973 | if (!xfs_efd_zone) |
1974 | goto out_destroy_buf_item_zone; | |
1975 | ||
b1231760 CM |
1976 | xfs_efi_zone = kmem_cache_create("xfs_efi_item", |
1977 | (sizeof(struct xfs_efi_log_item) + | |
1978 | (XFS_EFI_MAX_FAST_EXTENTS - 1) * | |
1979 | sizeof(struct xfs_extent)), | |
1980 | 0, 0, NULL); | |
9f8868ff CH |
1981 | if (!xfs_efi_zone) |
1982 | goto out_destroy_efd_zone; | |
1983 | ||
b1231760 CM |
1984 | xfs_inode_zone = kmem_cache_create("xfs_inode", |
1985 | sizeof(struct xfs_inode), 0, | |
1986 | (SLAB_HWCACHE_ALIGN | | |
1987 | SLAB_RECLAIM_ACCOUNT | | |
1988 | SLAB_MEM_SPREAD | SLAB_ACCOUNT), | |
1989 | xfs_fs_inode_init_once); | |
9f8868ff CH |
1990 | if (!xfs_inode_zone) |
1991 | goto out_destroy_efi_zone; | |
1992 | ||
b1231760 CM |
1993 | xfs_ili_zone = kmem_cache_create("xfs_ili", |
1994 | sizeof(struct xfs_inode_log_item), 0, | |
d59eadae DC |
1995 | SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, |
1996 | NULL); | |
9f8868ff CH |
1997 | if (!xfs_ili_zone) |
1998 | goto out_destroy_inode_zone; | |
b1231760 CM |
1999 | |
2000 | xfs_icreate_zone = kmem_cache_create("xfs_icr", | |
2001 | sizeof(struct xfs_icreate_item), | |
2002 | 0, 0, NULL); | |
3ebe7d2d DC |
2003 | if (!xfs_icreate_zone) |
2004 | goto out_destroy_ili_zone; | |
9f8868ff | 2005 | |
b1231760 CM |
2006 | xfs_rud_zone = kmem_cache_create("xfs_rud_item", |
2007 | sizeof(struct xfs_rud_log_item), | |
2008 | 0, 0, NULL); | |
5880f2d7 DW |
2009 | if (!xfs_rud_zone) |
2010 | goto out_destroy_icreate_zone; | |
2011 | ||
b1231760 | 2012 | xfs_rui_zone = kmem_cache_create("xfs_rui_item", |
cd00158c | 2013 | xfs_rui_log_item_sizeof(XFS_RUI_MAX_FAST_EXTENTS), |
b1231760 | 2014 | 0, 0, NULL); |
5880f2d7 DW |
2015 | if (!xfs_rui_zone) |
2016 | goto out_destroy_rud_zone; | |
2017 | ||
b1231760 CM |
2018 | xfs_cud_zone = kmem_cache_create("xfs_cud_item", |
2019 | sizeof(struct xfs_cud_log_item), | |
2020 | 0, 0, NULL); | |
baf4bcac DW |
2021 | if (!xfs_cud_zone) |
2022 | goto out_destroy_rui_zone; | |
2023 | ||
b1231760 | 2024 | xfs_cui_zone = kmem_cache_create("xfs_cui_item", |
baf4bcac | 2025 | xfs_cui_log_item_sizeof(XFS_CUI_MAX_FAST_EXTENTS), |
b1231760 | 2026 | 0, 0, NULL); |
baf4bcac DW |
2027 | if (!xfs_cui_zone) |
2028 | goto out_destroy_cud_zone; | |
2029 | ||
b1231760 CM |
2030 | xfs_bud_zone = kmem_cache_create("xfs_bud_item", |
2031 | sizeof(struct xfs_bud_log_item), | |
2032 | 0, 0, NULL); | |
6413a014 DW |
2033 | if (!xfs_bud_zone) |
2034 | goto out_destroy_cui_zone; | |
2035 | ||
b1231760 | 2036 | xfs_bui_zone = kmem_cache_create("xfs_bui_item", |
6413a014 | 2037 | xfs_bui_log_item_sizeof(XFS_BUI_MAX_FAST_EXTENTS), |
b1231760 | 2038 | 0, 0, NULL); |
6413a014 DW |
2039 | if (!xfs_bui_zone) |
2040 | goto out_destroy_bud_zone; | |
2041 | ||
9f8868ff CH |
2042 | return 0; |
2043 | ||
6413a014 | 2044 | out_destroy_bud_zone: |
aaf54eb8 | 2045 | kmem_cache_destroy(xfs_bud_zone); |
6413a014 | 2046 | out_destroy_cui_zone: |
aaf54eb8 | 2047 | kmem_cache_destroy(xfs_cui_zone); |
baf4bcac | 2048 | out_destroy_cud_zone: |
aaf54eb8 | 2049 | kmem_cache_destroy(xfs_cud_zone); |
baf4bcac | 2050 | out_destroy_rui_zone: |
aaf54eb8 | 2051 | kmem_cache_destroy(xfs_rui_zone); |
5880f2d7 | 2052 | out_destroy_rud_zone: |
aaf54eb8 | 2053 | kmem_cache_destroy(xfs_rud_zone); |
5880f2d7 | 2054 | out_destroy_icreate_zone: |
aaf54eb8 | 2055 | kmem_cache_destroy(xfs_icreate_zone); |
3ebe7d2d | 2056 | out_destroy_ili_zone: |
aaf54eb8 | 2057 | kmem_cache_destroy(xfs_ili_zone); |
9f8868ff | 2058 | out_destroy_inode_zone: |
aaf54eb8 | 2059 | kmem_cache_destroy(xfs_inode_zone); |
9f8868ff | 2060 | out_destroy_efi_zone: |
aaf54eb8 | 2061 | kmem_cache_destroy(xfs_efi_zone); |
9f8868ff | 2062 | out_destroy_efd_zone: |
aaf54eb8 | 2063 | kmem_cache_destroy(xfs_efd_zone); |
9f8868ff | 2064 | out_destroy_buf_item_zone: |
aaf54eb8 | 2065 | kmem_cache_destroy(xfs_buf_item_zone); |
9f8868ff | 2066 | out_destroy_trans_zone: |
aaf54eb8 | 2067 | kmem_cache_destroy(xfs_trans_zone); |
9f8868ff | 2068 | out_destroy_ifork_zone: |
aaf54eb8 | 2069 | kmem_cache_destroy(xfs_ifork_zone); |
9f8868ff | 2070 | out_destroy_da_state_zone: |
aaf54eb8 | 2071 | kmem_cache_destroy(xfs_da_state_zone); |
9f8868ff | 2072 | out_destroy_btree_cur_zone: |
aaf54eb8 | 2073 | kmem_cache_destroy(xfs_btree_cur_zone); |
9f8868ff | 2074 | out_destroy_bmap_free_item_zone: |
aaf54eb8 | 2075 | kmem_cache_destroy(xfs_bmap_free_item_zone); |
9f8868ff | 2076 | out_destroy_log_ticket_zone: |
aaf54eb8 | 2077 | kmem_cache_destroy(xfs_log_ticket_zone); |
9f8868ff CH |
2078 | out: |
2079 | return -ENOMEM; | |
2080 | } | |
2081 | ||
2082 | STATIC void | |
2083 | xfs_destroy_zones(void) | |
2084 | { | |
8c0a8537 KS |
2085 | /* |
2086 | * Make sure all delayed rcu free are flushed before we | |
2087 | * destroy caches. | |
2088 | */ | |
2089 | rcu_barrier(); | |
aaf54eb8 CM |
2090 | kmem_cache_destroy(xfs_bui_zone); |
2091 | kmem_cache_destroy(xfs_bud_zone); | |
2092 | kmem_cache_destroy(xfs_cui_zone); | |
2093 | kmem_cache_destroy(xfs_cud_zone); | |
2094 | kmem_cache_destroy(xfs_rui_zone); | |
2095 | kmem_cache_destroy(xfs_rud_zone); | |
2096 | kmem_cache_destroy(xfs_icreate_zone); | |
2097 | kmem_cache_destroy(xfs_ili_zone); | |
2098 | kmem_cache_destroy(xfs_inode_zone); | |
2099 | kmem_cache_destroy(xfs_efi_zone); | |
2100 | kmem_cache_destroy(xfs_efd_zone); | |
2101 | kmem_cache_destroy(xfs_buf_item_zone); | |
2102 | kmem_cache_destroy(xfs_trans_zone); | |
2103 | kmem_cache_destroy(xfs_ifork_zone); | |
2104 | kmem_cache_destroy(xfs_da_state_zone); | |
2105 | kmem_cache_destroy(xfs_btree_cur_zone); | |
2106 | kmem_cache_destroy(xfs_bmap_free_item_zone); | |
2107 | kmem_cache_destroy(xfs_log_ticket_zone); | |
9f8868ff | 2108 | } |
1da177e4 | 2109 | |
0bf6a5bd DC |
2110 | STATIC int __init |
2111 | xfs_init_workqueues(void) | |
2112 | { | |
c999a223 DC |
2113 | /* |
2114 | * The allocation workqueue can be used in memory reclaim situations | |
2115 | * (writepage path), and parallelism is only limited by the number of | |
2116 | * AGs in all the filesystems mounted. Hence use the default large | |
2117 | * max_active value for this workqueue. | |
2118 | */ | |
8018ec08 BF |
2119 | xfs_alloc_wq = alloc_workqueue("xfsalloc", |
2120 | WQ_MEM_RECLAIM|WQ_FREEZABLE, 0); | |
c999a223 | 2121 | if (!xfs_alloc_wq) |
5889608d | 2122 | return -ENOMEM; |
c999a223 | 2123 | |
4560e78f CH |
2124 | xfs_discard_wq = alloc_workqueue("xfsdiscard", WQ_UNBOUND, 0); |
2125 | if (!xfs_discard_wq) | |
2126 | goto out_free_alloc_wq; | |
2127 | ||
0bf6a5bd | 2128 | return 0; |
4560e78f CH |
2129 | out_free_alloc_wq: |
2130 | destroy_workqueue(xfs_alloc_wq); | |
2131 | return -ENOMEM; | |
0bf6a5bd DC |
2132 | } |
2133 | ||
39411f81 | 2134 | STATIC void |
0bf6a5bd DC |
2135 | xfs_destroy_workqueues(void) |
2136 | { | |
4560e78f | 2137 | destroy_workqueue(xfs_discard_wq); |
c999a223 | 2138 | destroy_workqueue(xfs_alloc_wq); |
0bf6a5bd DC |
2139 | } |
2140 | ||
1da177e4 | 2141 | STATIC int __init |
9f8868ff | 2142 | init_xfs_fs(void) |
1da177e4 LT |
2143 | { |
2144 | int error; | |
1da177e4 | 2145 | |
30cbc591 DW |
2146 | xfs_check_ondisk_structs(); |
2147 | ||
65795910 CH |
2148 | printk(KERN_INFO XFS_VERSION_STRING " with " |
2149 | XFS_BUILD_OPTIONS " enabled\n"); | |
1da177e4 | 2150 | |
9f8868ff | 2151 | xfs_dir_startup(); |
1da177e4 | 2152 | |
8758280f | 2153 | error = xfs_init_zones(); |
9f8868ff CH |
2154 | if (error) |
2155 | goto out; | |
2156 | ||
0bf6a5bd | 2157 | error = xfs_init_workqueues(); |
9f8868ff | 2158 | if (error) |
0b1b213f | 2159 | goto out_destroy_zones; |
9f8868ff | 2160 | |
0bf6a5bd DC |
2161 | error = xfs_mru_cache_init(); |
2162 | if (error) | |
2163 | goto out_destroy_wq; | |
2164 | ||
ce8e922c | 2165 | error = xfs_buf_init(); |
9f8868ff | 2166 | if (error) |
1919adda | 2167 | goto out_mru_cache_uninit; |
9f8868ff CH |
2168 | |
2169 | error = xfs_init_procfs(); | |
2170 | if (error) | |
2171 | goto out_buf_terminate; | |
2172 | ||
2173 | error = xfs_sysctl_register(); | |
2174 | if (error) | |
2175 | goto out_cleanup_procfs; | |
1da177e4 | 2176 | |
3d871226 BF |
2177 | xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj); |
2178 | if (!xfs_kset) { | |
2179 | error = -ENOMEM; | |
bb230c12 | 2180 | goto out_sysctl_unregister; |
3d871226 BF |
2181 | } |
2182 | ||
80529c45 BD |
2183 | xfsstats.xs_kobj.kobject.kset = xfs_kset; |
2184 | ||
2185 | xfsstats.xs_stats = alloc_percpu(struct xfsstats); | |
2186 | if (!xfsstats.xs_stats) { | |
2187 | error = -ENOMEM; | |
2188 | goto out_kset_unregister; | |
2189 | } | |
2190 | ||
2191 | error = xfs_sysfs_init(&xfsstats.xs_kobj, &xfs_stats_ktype, NULL, | |
bb230c12 BD |
2192 | "stats"); |
2193 | if (error) | |
80529c45 | 2194 | goto out_free_stats; |
bb230c12 | 2195 | |
65b65735 BF |
2196 | #ifdef DEBUG |
2197 | xfs_dbg_kobj.kobject.kset = xfs_kset; | |
2198 | error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug"); | |
a05931ce | 2199 | if (error) |
bb230c12 | 2200 | goto out_remove_stats_kobj; |
65b65735 BF |
2201 | #endif |
2202 | ||
2203 | error = xfs_qm_init(); | |
2204 | if (error) | |
bb230c12 | 2205 | goto out_remove_dbg_kobj; |
1da177e4 LT |
2206 | |
2207 | error = register_filesystem(&xfs_fs_type); | |
2208 | if (error) | |
a05931ce | 2209 | goto out_qm_exit; |
1da177e4 LT |
2210 | return 0; |
2211 | ||
a05931ce CH |
2212 | out_qm_exit: |
2213 | xfs_qm_exit(); | |
bb230c12 | 2214 | out_remove_dbg_kobj: |
65b65735 BF |
2215 | #ifdef DEBUG |
2216 | xfs_sysfs_del(&xfs_dbg_kobj); | |
bb230c12 | 2217 | out_remove_stats_kobj: |
65b65735 | 2218 | #endif |
80529c45 BD |
2219 | xfs_sysfs_del(&xfsstats.xs_kobj); |
2220 | out_free_stats: | |
2221 | free_percpu(xfsstats.xs_stats); | |
bb230c12 | 2222 | out_kset_unregister: |
3d871226 | 2223 | kset_unregister(xfs_kset); |
9f8868ff CH |
2224 | out_sysctl_unregister: |
2225 | xfs_sysctl_unregister(); | |
2226 | out_cleanup_procfs: | |
2227 | xfs_cleanup_procfs(); | |
2228 | out_buf_terminate: | |
ce8e922c | 2229 | xfs_buf_terminate(); |
9f8868ff CH |
2230 | out_mru_cache_uninit: |
2231 | xfs_mru_cache_uninit(); | |
0bf6a5bd DC |
2232 | out_destroy_wq: |
2233 | xfs_destroy_workqueues(); | |
9f8868ff | 2234 | out_destroy_zones: |
8758280f | 2235 | xfs_destroy_zones(); |
9f8868ff | 2236 | out: |
1da177e4 LT |
2237 | return error; |
2238 | } | |
2239 | ||
2240 | STATIC void __exit | |
9f8868ff | 2241 | exit_xfs_fs(void) |
1da177e4 | 2242 | { |
a05931ce | 2243 | xfs_qm_exit(); |
1da177e4 | 2244 | unregister_filesystem(&xfs_fs_type); |
65b65735 BF |
2245 | #ifdef DEBUG |
2246 | xfs_sysfs_del(&xfs_dbg_kobj); | |
2247 | #endif | |
80529c45 BD |
2248 | xfs_sysfs_del(&xfsstats.xs_kobj); |
2249 | free_percpu(xfsstats.xs_stats); | |
3d871226 | 2250 | kset_unregister(xfs_kset); |
9f8868ff CH |
2251 | xfs_sysctl_unregister(); |
2252 | xfs_cleanup_procfs(); | |
ce8e922c | 2253 | xfs_buf_terminate(); |
9f8868ff | 2254 | xfs_mru_cache_uninit(); |
0bf6a5bd | 2255 | xfs_destroy_workqueues(); |
8758280f | 2256 | xfs_destroy_zones(); |
af3b6382 | 2257 | xfs_uuid_table_free(); |
1da177e4 LT |
2258 | } |
2259 | ||
2260 | module_init(init_xfs_fs); | |
2261 | module_exit(exit_xfs_fs); | |
2262 | ||
2263 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
2264 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
2265 | MODULE_LICENSE("GPL"); |