]>
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 | ||
237d7887 KX |
202 | if (mp->m_qflags & XFS_UQUOTA_ACCT) { |
203 | if (mp->m_qflags & XFS_UQUOTA_ENFD) | |
204 | seq_puts(m, ",usrquota"); | |
205 | else | |
206 | seq_puts(m, ",uqnoenforce"); | |
207 | } | |
a67d7c5f | 208 | |
988abe40 | 209 | if (mp->m_qflags & XFS_PQUOTA_ACCT) { |
83e782e1 | 210 | if (mp->m_qflags & XFS_PQUOTA_ENFD) |
2e74af0e | 211 | seq_puts(m, ",prjquota"); |
988abe40 | 212 | else |
2e74af0e | 213 | seq_puts(m, ",pqnoenforce"); |
d892d586 CS |
214 | } |
215 | if (mp->m_qflags & XFS_GQUOTA_ACCT) { | |
83e782e1 | 216 | if (mp->m_qflags & XFS_GQUOTA_ENFD) |
2e74af0e | 217 | seq_puts(m, ",grpquota"); |
988abe40 | 218 | else |
2e74af0e | 219 | seq_puts(m, ",gqnoenforce"); |
988abe40 | 220 | } |
a67d7c5f DC |
221 | |
222 | if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT)) | |
2e74af0e | 223 | seq_puts(m, ",noquota"); |
21f55993 CH |
224 | |
225 | return 0; | |
a67d7c5f | 226 | } |
91083269 | 227 | |
9de67c3b | 228 | /* |
12c3f05c ES |
229 | * Set parameters for inode allocation heuristics, taking into account |
230 | * filesystem size and inode32/inode64 mount options; i.e. specifically | |
231 | * whether or not XFS_MOUNT_SMALL_INUMS is set. | |
232 | * | |
233 | * Inode allocation patterns are altered only if inode32 is requested | |
234 | * (XFS_MOUNT_SMALL_INUMS), and the filesystem is sufficiently large. | |
235 | * If altered, XFS_MOUNT_32BITINODES is set as well. | |
236 | * | |
237 | * An agcount independent of that in the mount structure is provided | |
238 | * because in the growfs case, mp->m_sb.sb_agcount is not yet updated | |
239 | * to the potentially higher ag count. | |
240 | * | |
241 | * Returns the maximum AG index which may contain inodes. | |
9de67c3b | 242 | */ |
2d2194f6 | 243 | xfs_agnumber_t |
12c3f05c ES |
244 | xfs_set_inode_alloc( |
245 | struct xfs_mount *mp, | |
246 | xfs_agnumber_t agcount) | |
2d2194f6 | 247 | { |
12c3f05c | 248 | xfs_agnumber_t index; |
4056c1d0 | 249 | xfs_agnumber_t maxagi = 0; |
2d2194f6 CM |
250 | xfs_sb_t *sbp = &mp->m_sb; |
251 | xfs_agnumber_t max_metadata; | |
54aa61f8 ES |
252 | xfs_agino_t agino; |
253 | xfs_ino_t ino; | |
2d2194f6 | 254 | |
12c3f05c ES |
255 | /* |
256 | * Calculate how much should be reserved for inodes to meet | |
257 | * the max inode percentage. Used only for inode32. | |
2d2194f6 | 258 | */ |
ef325959 | 259 | if (M_IGEO(mp)->maxicount) { |
c8ce540d | 260 | uint64_t icount; |
2d2194f6 CM |
261 | |
262 | icount = sbp->sb_dblocks * sbp->sb_imax_pct; | |
263 | do_div(icount, 100); | |
264 | icount += sbp->sb_agblocks - 1; | |
265 | do_div(icount, sbp->sb_agblocks); | |
266 | max_metadata = icount; | |
267 | } else { | |
9de67c3b | 268 | max_metadata = agcount; |
2d2194f6 CM |
269 | } |
270 | ||
12c3f05c | 271 | /* Get the last possible inode in the filesystem */ |
43004b2a | 272 | agino = XFS_AGB_TO_AGINO(mp, sbp->sb_agblocks - 1); |
12c3f05c ES |
273 | ino = XFS_AGINO_TO_INO(mp, agcount - 1, agino); |
274 | ||
275 | /* | |
276 | * If user asked for no more than 32-bit inodes, and the fs is | |
277 | * sufficiently large, set XFS_MOUNT_32BITINODES if we must alter | |
278 | * the allocator to accommodate the request. | |
279 | */ | |
280 | if ((mp->m_flags & XFS_MOUNT_SMALL_INUMS) && ino > XFS_MAXINUMBER_32) | |
281 | mp->m_flags |= XFS_MOUNT_32BITINODES; | |
282 | else | |
283 | mp->m_flags &= ~XFS_MOUNT_32BITINODES; | |
54aa61f8 | 284 | |
9de67c3b | 285 | for (index = 0; index < agcount; index++) { |
12c3f05c | 286 | struct xfs_perag *pag; |
4056c1d0 | 287 | |
12c3f05c | 288 | ino = XFS_AGINO_TO_INO(mp, index, agino); |
2d2194f6 CM |
289 | |
290 | pag = xfs_perag_get(mp, index); | |
2d2194f6 | 291 | |
12c3f05c ES |
292 | if (mp->m_flags & XFS_MOUNT_32BITINODES) { |
293 | if (ino > XFS_MAXINUMBER_32) { | |
294 | pag->pagi_inodeok = 0; | |
295 | pag->pagf_metadata = 0; | |
296 | } else { | |
297 | pag->pagi_inodeok = 1; | |
298 | maxagi++; | |
299 | if (index < max_metadata) | |
300 | pag->pagf_metadata = 1; | |
301 | else | |
302 | pag->pagf_metadata = 0; | |
303 | } | |
304 | } else { | |
305 | pag->pagi_inodeok = 1; | |
306 | pag->pagf_metadata = 0; | |
307 | } | |
2d2194f6 | 308 | |
2d2194f6 CM |
309 | xfs_perag_put(pag); |
310 | } | |
311 | ||
12c3f05c | 312 | return (mp->m_flags & XFS_MOUNT_32BITINODES) ? maxagi : agcount; |
2d2194f6 CM |
313 | } |
314 | ||
3180e66d | 315 | STATIC int |
1da177e4 LT |
316 | xfs_blkdev_get( |
317 | xfs_mount_t *mp, | |
318 | const char *name, | |
319 | struct block_device **bdevp) | |
320 | { | |
321 | int error = 0; | |
322 | ||
d4d77629 TH |
323 | *bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL, |
324 | mp); | |
1da177e4 LT |
325 | if (IS_ERR(*bdevp)) { |
326 | error = PTR_ERR(*bdevp); | |
77af574e | 327 | xfs_warn(mp, "Invalid device [%s], error=%d", name, error); |
1da177e4 LT |
328 | } |
329 | ||
2451337d | 330 | return error; |
1da177e4 LT |
331 | } |
332 | ||
3180e66d | 333 | STATIC void |
1da177e4 LT |
334 | xfs_blkdev_put( |
335 | struct block_device *bdev) | |
336 | { | |
337 | if (bdev) | |
e525fd89 | 338 | blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); |
1da177e4 LT |
339 | } |
340 | ||
f538d4da CH |
341 | void |
342 | xfs_blkdev_issue_flush( | |
343 | xfs_buftarg_t *buftarg) | |
344 | { | |
9398554f | 345 | blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS); |
f538d4da | 346 | } |
1da177e4 | 347 | |
19f354d4 CH |
348 | STATIC void |
349 | xfs_close_devices( | |
350 | struct xfs_mount *mp) | |
351 | { | |
486aff5e DW |
352 | struct dax_device *dax_ddev = mp->m_ddev_targp->bt_daxdev; |
353 | ||
19f354d4 | 354 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { |
c032bfcf | 355 | struct block_device *logdev = mp->m_logdev_targp->bt_bdev; |
486aff5e DW |
356 | struct dax_device *dax_logdev = mp->m_logdev_targp->bt_daxdev; |
357 | ||
a1f69417 | 358 | xfs_free_buftarg(mp->m_logdev_targp); |
c032bfcf | 359 | xfs_blkdev_put(logdev); |
486aff5e | 360 | fs_put_dax(dax_logdev); |
19f354d4 CH |
361 | } |
362 | if (mp->m_rtdev_targp) { | |
c032bfcf | 363 | struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev; |
486aff5e DW |
364 | struct dax_device *dax_rtdev = mp->m_rtdev_targp->bt_daxdev; |
365 | ||
a1f69417 | 366 | xfs_free_buftarg(mp->m_rtdev_targp); |
c032bfcf | 367 | xfs_blkdev_put(rtdev); |
486aff5e | 368 | fs_put_dax(dax_rtdev); |
19f354d4 | 369 | } |
a1f69417 | 370 | xfs_free_buftarg(mp->m_ddev_targp); |
486aff5e | 371 | fs_put_dax(dax_ddev); |
19f354d4 CH |
372 | } |
373 | ||
374 | /* | |
375 | * The file system configurations are: | |
376 | * (1) device (partition) with data and internal log | |
377 | * (2) logical volume with data and log subvolumes. | |
378 | * (3) logical volume with data, log, and realtime subvolumes. | |
379 | * | |
380 | * We only have to handle opening the log and realtime volumes here if | |
381 | * they are present. The data subvolume has already been opened by | |
382 | * get_sb_bdev() and is stored in sb->s_bdev. | |
383 | */ | |
384 | STATIC int | |
385 | xfs_open_devices( | |
9d565ffa | 386 | struct xfs_mount *mp) |
19f354d4 CH |
387 | { |
388 | struct block_device *ddev = mp->m_super->s_bdev; | |
486aff5e DW |
389 | struct dax_device *dax_ddev = fs_dax_get_by_bdev(ddev); |
390 | struct dax_device *dax_logdev = NULL, *dax_rtdev = NULL; | |
19f354d4 CH |
391 | struct block_device *logdev = NULL, *rtdev = NULL; |
392 | int error; | |
393 | ||
394 | /* | |
395 | * Open real time and log devices - order is important. | |
396 | */ | |
9d565ffa CH |
397 | if (mp->m_logname) { |
398 | error = xfs_blkdev_get(mp, mp->m_logname, &logdev); | |
19f354d4 CH |
399 | if (error) |
400 | goto out; | |
486aff5e | 401 | dax_logdev = fs_dax_get_by_bdev(logdev); |
19f354d4 CH |
402 | } |
403 | ||
9d565ffa CH |
404 | if (mp->m_rtname) { |
405 | error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev); | |
19f354d4 CH |
406 | if (error) |
407 | goto out_close_logdev; | |
408 | ||
409 | if (rtdev == ddev || rtdev == logdev) { | |
4f10700a DC |
410 | xfs_warn(mp, |
411 | "Cannot mount filesystem with identical rtdev and ddev/logdev."); | |
2451337d | 412 | error = -EINVAL; |
19f354d4 CH |
413 | goto out_close_rtdev; |
414 | } | |
486aff5e | 415 | dax_rtdev = fs_dax_get_by_bdev(rtdev); |
19f354d4 CH |
416 | } |
417 | ||
418 | /* | |
419 | * Setup xfs_mount buffer target pointers | |
420 | */ | |
2451337d | 421 | error = -ENOMEM; |
486aff5e | 422 | mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev, dax_ddev); |
19f354d4 CH |
423 | if (!mp->m_ddev_targp) |
424 | goto out_close_rtdev; | |
425 | ||
426 | if (rtdev) { | |
486aff5e | 427 | mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev, dax_rtdev); |
19f354d4 CH |
428 | if (!mp->m_rtdev_targp) |
429 | goto out_free_ddev_targ; | |
430 | } | |
431 | ||
432 | if (logdev && logdev != ddev) { | |
486aff5e | 433 | mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev, dax_logdev); |
19f354d4 CH |
434 | if (!mp->m_logdev_targp) |
435 | goto out_free_rtdev_targ; | |
436 | } else { | |
437 | mp->m_logdev_targp = mp->m_ddev_targp; | |
438 | } | |
439 | ||
440 | return 0; | |
441 | ||
442 | out_free_rtdev_targ: | |
443 | if (mp->m_rtdev_targp) | |
a1f69417 | 444 | xfs_free_buftarg(mp->m_rtdev_targp); |
19f354d4 | 445 | out_free_ddev_targ: |
a1f69417 | 446 | xfs_free_buftarg(mp->m_ddev_targp); |
19f354d4 | 447 | out_close_rtdev: |
d2a5e3c6 | 448 | xfs_blkdev_put(rtdev); |
486aff5e | 449 | fs_put_dax(dax_rtdev); |
19f354d4 | 450 | out_close_logdev: |
486aff5e | 451 | if (logdev && logdev != ddev) { |
19f354d4 | 452 | xfs_blkdev_put(logdev); |
486aff5e DW |
453 | fs_put_dax(dax_logdev); |
454 | } | |
19f354d4 | 455 | out: |
486aff5e | 456 | fs_put_dax(dax_ddev); |
19f354d4 CH |
457 | return error; |
458 | } | |
459 | ||
e34b562c CH |
460 | /* |
461 | * Setup xfs_mount buffer target pointers based on superblock | |
462 | */ | |
463 | STATIC int | |
464 | xfs_setup_devices( | |
465 | struct xfs_mount *mp) | |
466 | { | |
467 | int error; | |
19f354d4 | 468 | |
a96c4151 | 469 | error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize); |
e34b562c CH |
470 | if (error) |
471 | return error; | |
472 | ||
473 | if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) { | |
474 | unsigned int log_sector_size = BBSIZE; | |
475 | ||
476 | if (xfs_sb_version_hassector(&mp->m_sb)) | |
477 | log_sector_size = mp->m_sb.sb_logsectsize; | |
478 | error = xfs_setsize_buftarg(mp->m_logdev_targp, | |
e34b562c CH |
479 | log_sector_size); |
480 | if (error) | |
481 | return error; | |
482 | } | |
483 | if (mp->m_rtdev_targp) { | |
484 | error = xfs_setsize_buftarg(mp->m_rtdev_targp, | |
e34b562c CH |
485 | mp->m_sb.sb_sectsize); |
486 | if (error) | |
487 | return error; | |
488 | } | |
489 | ||
490 | return 0; | |
491 | } | |
19f354d4 | 492 | |
aa6bf01d CH |
493 | STATIC int |
494 | xfs_init_mount_workqueues( | |
495 | struct xfs_mount *mp) | |
496 | { | |
78c931b8 | 497 | mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s", |
05a302a1 DW |
498 | XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM), |
499 | 1, mp->m_super->s_id); | |
78c931b8 BF |
500 | if (!mp->m_buf_workqueue) |
501 | goto out; | |
502 | ||
aa6bf01d | 503 | mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s", |
05a302a1 DW |
504 | XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM), |
505 | 0, mp->m_super->s_id); | |
aa6bf01d | 506 | if (!mp->m_unwritten_workqueue) |
28408243 | 507 | goto out_destroy_buf; |
aa6bf01d | 508 | |
4c2d542f | 509 | mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s", |
05a302a1 | 510 | XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM | WQ_UNBOUND), |
e1d3d218 | 511 | 0, mp->m_super->s_id); |
4c2d542f DC |
512 | if (!mp->m_cil_workqueue) |
513 | goto out_destroy_unwritten; | |
5889608d DC |
514 | |
515 | mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s", | |
05a302a1 DW |
516 | XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM), |
517 | 0, mp->m_super->s_id); | |
5889608d DC |
518 | if (!mp->m_reclaim_workqueue) |
519 | goto out_destroy_cil; | |
520 | ||
9669f51d | 521 | mp->m_blockgc_workqueue = alloc_workqueue("xfs-blockgc/%s", |
05a302a1 DW |
522 | XFS_WQFLAGS(WQ_FREEZABLE | WQ_MEM_RECLAIM), |
523 | 0, mp->m_super->s_id); | |
9669f51d | 524 | if (!mp->m_blockgc_workqueue) |
1058d0f5 | 525 | goto out_destroy_reclaim; |
579b62fa | 526 | |
05a302a1 DW |
527 | mp->m_sync_workqueue = alloc_workqueue("xfs-sync/%s", |
528 | XFS_WQFLAGS(WQ_FREEZABLE), 0, mp->m_super->s_id); | |
696a5620 BF |
529 | if (!mp->m_sync_workqueue) |
530 | goto out_destroy_eofb; | |
531 | ||
aa6bf01d CH |
532 | return 0; |
533 | ||
696a5620 | 534 | out_destroy_eofb: |
9669f51d | 535 | destroy_workqueue(mp->m_blockgc_workqueue); |
5889608d DC |
536 | out_destroy_reclaim: |
537 | destroy_workqueue(mp->m_reclaim_workqueue); | |
538 | out_destroy_cil: | |
539 | destroy_workqueue(mp->m_cil_workqueue); | |
4c2d542f DC |
540 | out_destroy_unwritten: |
541 | destroy_workqueue(mp->m_unwritten_workqueue); | |
78c931b8 BF |
542 | out_destroy_buf: |
543 | destroy_workqueue(mp->m_buf_workqueue); | |
aa6bf01d CH |
544 | out: |
545 | return -ENOMEM; | |
546 | } | |
547 | ||
548 | STATIC void | |
549 | xfs_destroy_mount_workqueues( | |
550 | struct xfs_mount *mp) | |
551 | { | |
696a5620 | 552 | destroy_workqueue(mp->m_sync_workqueue); |
9669f51d | 553 | destroy_workqueue(mp->m_blockgc_workqueue); |
5889608d | 554 | destroy_workqueue(mp->m_reclaim_workqueue); |
4c2d542f | 555 | destroy_workqueue(mp->m_cil_workqueue); |
aa6bf01d | 556 | destroy_workqueue(mp->m_unwritten_workqueue); |
78c931b8 | 557 | destroy_workqueue(mp->m_buf_workqueue); |
aa6bf01d CH |
558 | } |
559 | ||
f0f7a674 DW |
560 | static void |
561 | xfs_flush_inodes_worker( | |
562 | struct work_struct *work) | |
563 | { | |
564 | struct xfs_mount *mp = container_of(work, struct xfs_mount, | |
565 | m_flush_inodes_work); | |
566 | struct super_block *sb = mp->m_super; | |
567 | ||
568 | if (down_read_trylock(&sb->s_umount)) { | |
569 | sync_inodes_sb(sb); | |
570 | up_read(&sb->s_umount); | |
571 | } | |
572 | } | |
573 | ||
9aa05000 DC |
574 | /* |
575 | * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK | |
576 | * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting | |
577 | * for IO to complete so that we effectively throttle multiple callers to the | |
578 | * rate at which IO is completing. | |
579 | */ | |
580 | void | |
581 | xfs_flush_inodes( | |
582 | struct xfs_mount *mp) | |
583 | { | |
f0f7a674 DW |
584 | /* |
585 | * If flush_work() returns true then that means we waited for a flush | |
586 | * which was already in progress. Don't bother running another scan. | |
587 | */ | |
588 | if (flush_work(&mp->m_flush_inodes_work)) | |
c6425702 DW |
589 | return; |
590 | ||
f0f7a674 DW |
591 | queue_work(mp->m_sync_workqueue, &mp->m_flush_inodes_work); |
592 | flush_work(&mp->m_flush_inodes_work); | |
9aa05000 DC |
593 | } |
594 | ||
bf904248 | 595 | /* Catch misguided souls that try to use this interface on XFS */ |
1da177e4 | 596 | STATIC struct inode * |
a50cd269 | 597 | xfs_fs_alloc_inode( |
1da177e4 LT |
598 | struct super_block *sb) |
599 | { | |
bf904248 | 600 | BUG(); |
493dca61 | 601 | return NULL; |
1da177e4 LT |
602 | } |
603 | ||
4831822f CH |
604 | #ifdef DEBUG |
605 | static void | |
606 | xfs_check_delalloc( | |
607 | struct xfs_inode *ip, | |
608 | int whichfork) | |
609 | { | |
610 | struct xfs_ifork *ifp = XFS_IFORK_PTR(ip, whichfork); | |
611 | struct xfs_bmbt_irec got; | |
612 | struct xfs_iext_cursor icur; | |
613 | ||
614 | if (!ifp || !xfs_iext_lookup_extent(ip, ifp, 0, &icur, &got)) | |
615 | return; | |
616 | do { | |
617 | if (isnullstartblock(got.br_startblock)) { | |
618 | xfs_warn(ip->i_mount, | |
619 | "ino %llx %s fork has delalloc extent at [0x%llx:0x%llx]", | |
620 | ip->i_ino, | |
621 | whichfork == XFS_DATA_FORK ? "data" : "cow", | |
622 | got.br_startoff, got.br_blockcount); | |
623 | } | |
624 | } while (xfs_iext_next_extent(ifp, &icur, &got)); | |
625 | } | |
626 | #else | |
627 | #define xfs_check_delalloc(ip, whichfork) do { } while (0) | |
628 | #endif | |
629 | ||
bf904248 | 630 | /* |
99fa8cb3 | 631 | * Now that the generic code is guaranteed not to be accessing |
8179c036 | 632 | * the linux inode, we can inactivate and reclaim the inode. |
bf904248 | 633 | */ |
1da177e4 | 634 | STATIC void |
a50cd269 | 635 | xfs_fs_destroy_inode( |
848ce8f7 | 636 | struct inode *inode) |
1da177e4 | 637 | { |
848ce8f7 CH |
638 | struct xfs_inode *ip = XFS_I(inode); |
639 | ||
cca28fb8 | 640 | trace_xfs_destroy_inode(ip); |
99fa8cb3 | 641 | |
65523218 | 642 | ASSERT(!rwsem_is_locked(&inode->i_rwsem)); |
8179c036 DC |
643 | XFS_STATS_INC(ip->i_mount, vn_rele); |
644 | XFS_STATS_INC(ip->i_mount, vn_remove); | |
645 | ||
646 | xfs_inactive(ip); | |
848ce8f7 | 647 | |
4831822f CH |
648 | if (!XFS_FORCED_SHUTDOWN(ip->i_mount) && ip->i_delayed_blks) { |
649 | xfs_check_delalloc(ip, XFS_DATA_FORK); | |
650 | xfs_check_delalloc(ip, XFS_COW_FORK); | |
651 | ASSERT(0); | |
652 | } | |
653 | ||
8179c036 | 654 | XFS_STATS_INC(ip->i_mount, vn_reclaim); |
848ce8f7 CH |
655 | |
656 | /* | |
657 | * We should never get here with one of the reclaim flags already set. | |
658 | */ | |
659 | ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE)); | |
660 | ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM)); | |
661 | ||
662 | /* | |
718ecc50 DC |
663 | * We always use background reclaim here because even if the inode is |
664 | * clean, it still may be under IO and hence we have wait for IO | |
665 | * completion to occur before we can reclaim the inode. The background | |
666 | * reclaim path handles this more efficiently than we can here, so | |
667 | * simply let background reclaim tear down all inodes. | |
848ce8f7 | 668 | */ |
57817c68 | 669 | xfs_inode_set_reclaim_tag(ip); |
1da177e4 LT |
670 | } |
671 | ||
c3b1b131 CH |
672 | static void |
673 | xfs_fs_dirty_inode( | |
674 | struct inode *inode, | |
675 | int flag) | |
676 | { | |
677 | struct xfs_inode *ip = XFS_I(inode); | |
678 | struct xfs_mount *mp = ip->i_mount; | |
679 | struct xfs_trans *tp; | |
680 | ||
681 | if (!(inode->i_sb->s_flags & SB_LAZYTIME)) | |
682 | return; | |
683 | if (flag != I_DIRTY_SYNC || !(inode->i_state & I_DIRTY_TIME)) | |
684 | return; | |
685 | ||
686 | if (xfs_trans_alloc(mp, &M_RES(mp)->tr_fsyncts, 0, 0, 0, &tp)) | |
687 | return; | |
688 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
689 | xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); | |
690 | xfs_trans_log_inode(tp, ip, XFS_ILOG_TIMESTAMP); | |
691 | xfs_trans_commit(tp); | |
692 | } | |
693 | ||
07c8f675 DC |
694 | /* |
695 | * Slab object creation initialisation for the XFS inode. | |
696 | * This covers only the idempotent fields in the XFS inode; | |
697 | * all other fields need to be initialised on allocation | |
b595076a | 698 | * from the slab. This avoids the need to repeatedly initialise |
07c8f675 DC |
699 | * fields in the xfs inode that left in the initialise state |
700 | * when freeing the inode. | |
701 | */ | |
bf904248 DC |
702 | STATIC void |
703 | xfs_fs_inode_init_once( | |
07c8f675 DC |
704 | void *inode) |
705 | { | |
706 | struct xfs_inode *ip = inode; | |
707 | ||
708 | memset(ip, 0, sizeof(struct xfs_inode)); | |
bf904248 DC |
709 | |
710 | /* vfs inode */ | |
711 | inode_init_once(VFS_I(ip)); | |
712 | ||
713 | /* xfs inode */ | |
07c8f675 DC |
714 | atomic_set(&ip->i_pincount, 0); |
715 | spin_lock_init(&ip->i_flags_lock); | |
07c8f675 | 716 | |
653c60b6 DC |
717 | mrlock_init(&ip->i_mmaplock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, |
718 | "xfsino", ip->i_ino); | |
07c8f675 DC |
719 | mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER, |
720 | "xfsino", ip->i_ino); | |
07c8f675 DC |
721 | } |
722 | ||
5132ba8f DC |
723 | /* |
724 | * We do an unlocked check for XFS_IDONTCACHE here because we are already | |
725 | * serialised against cache hits here via the inode->i_lock and igrab() in | |
726 | * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be | |
727 | * racing with us, and it avoids needing to grab a spinlock here for every inode | |
728 | * we drop the final reference on. | |
729 | */ | |
730 | STATIC int | |
731 | xfs_fs_drop_inode( | |
732 | struct inode *inode) | |
733 | { | |
734 | struct xfs_inode *ip = XFS_I(inode); | |
735 | ||
17c12bcd DW |
736 | /* |
737 | * If this unlinked inode is in the middle of recovery, don't | |
738 | * drop the inode just yet; log recovery will take care of | |
739 | * that. See the comment for this inode flag. | |
740 | */ | |
741 | if (ip->i_flags & XFS_IRECOVERY) { | |
742 | ASSERT(ip->i_mount->m_log->l_flags & XLOG_RECOVERY_NEEDED); | |
743 | return 0; | |
744 | } | |
745 | ||
dae2f8ed | 746 | return generic_drop_inode(inode); |
5132ba8f DC |
747 | } |
748 | ||
a943f372 IK |
749 | static void |
750 | xfs_mount_free( | |
a738159d CH |
751 | struct xfs_mount *mp) |
752 | { | |
a738159d CH |
753 | kfree(mp->m_rtname); |
754 | kfree(mp->m_logname); | |
a943f372 | 755 | kmem_free(mp); |
a738159d CH |
756 | } |
757 | ||
1da177e4 | 758 | STATIC int |
69961a26 | 759 | xfs_fs_sync_fs( |
1da177e4 LT |
760 | struct super_block *sb, |
761 | int wait) | |
762 | { | |
745f6919 | 763 | struct xfs_mount *mp = XFS_M(sb); |
1da177e4 | 764 | |
e893bffd | 765 | /* |
34625c66 | 766 | * Doing anything during the async pass would be counterproductive. |
e893bffd | 767 | */ |
34625c66 | 768 | if (!wait) |
69961a26 | 769 | return 0; |
69961a26 | 770 | |
34061f5c | 771 | xfs_log_force(mp, XFS_LOG_SYNC); |
69961a26 | 772 | if (laptop_mode) { |
1da177e4 LT |
773 | /* |
774 | * The disk must be active because we're syncing. | |
f661f1e0 | 775 | * We schedule log work now (now that the disk is |
1da177e4 LT |
776 | * active) instead of later (when it might not be). |
777 | */ | |
f661f1e0 | 778 | flush_delayed_work(&mp->m_log->l_work); |
1da177e4 LT |
779 | } |
780 | ||
69961a26 | 781 | return 0; |
1da177e4 LT |
782 | } |
783 | ||
784 | STATIC int | |
a50cd269 | 785 | xfs_fs_statfs( |
726c3342 | 786 | struct dentry *dentry, |
1da177e4 LT |
787 | struct kstatfs *statp) |
788 | { | |
4ca488eb CH |
789 | struct xfs_mount *mp = XFS_M(dentry->d_sb); |
790 | xfs_sb_t *sbp = &mp->m_sb; | |
2b0143b5 | 791 | struct xfs_inode *ip = XFS_I(d_inode(dentry)); |
c8ce540d DW |
792 | uint64_t fakeinos, id; |
793 | uint64_t icount; | |
794 | uint64_t ifree; | |
795 | uint64_t fdblocks; | |
4ca488eb | 796 | xfs_extlen_t lsize; |
c8ce540d | 797 | int64_t ffree; |
4ca488eb | 798 | |
dddde68b | 799 | statp->f_type = XFS_SUPER_MAGIC; |
4ca488eb CH |
800 | statp->f_namelen = MAXNAMELEN - 1; |
801 | ||
802 | id = huge_encode_dev(mp->m_ddev_targp->bt_dev); | |
6d1349c7 | 803 | statp->f_fsid = u64_to_fsid(id); |
4ca488eb | 804 | |
501ab323 | 805 | icount = percpu_counter_sum(&mp->m_icount); |
e88b64ea | 806 | ifree = percpu_counter_sum(&mp->m_ifree); |
0d485ada | 807 | fdblocks = percpu_counter_sum(&mp->m_fdblocks); |
4ca488eb CH |
808 | |
809 | spin_lock(&mp->m_sb_lock); | |
810 | statp->f_bsize = sbp->sb_blocksize; | |
811 | lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0; | |
812 | statp->f_blocks = sbp->sb_dblocks - lsize; | |
0d485ada DC |
813 | spin_unlock(&mp->m_sb_lock); |
814 | ||
237aac46 ZB |
815 | /* make sure statp->f_bfree does not underflow */ |
816 | statp->f_bfree = max_t(int64_t, fdblocks - mp->m_alloc_set_aside, 0); | |
0d485ada DC |
817 | statp->f_bavail = statp->f_bfree; |
818 | ||
43004b2a | 819 | fakeinos = XFS_FSB_TO_INO(mp, statp->f_bfree); |
9bb54cb5 | 820 | statp->f_files = min(icount + fakeinos, (uint64_t)XFS_MAXINUMBER); |
ef325959 | 821 | if (M_IGEO(mp)->maxicount) |
a19d9f88 CH |
822 | statp->f_files = min_t(typeof(statp->f_files), |
823 | statp->f_files, | |
ef325959 | 824 | M_IGEO(mp)->maxicount); |
2fe33661 | 825 | |
01f9882e ES |
826 | /* If sb_icount overshot maxicount, report actual allocation */ |
827 | statp->f_files = max_t(typeof(statp->f_files), | |
828 | statp->f_files, | |
829 | sbp->sb_icount); | |
830 | ||
2fe33661 | 831 | /* make sure statp->f_ffree does not underflow */ |
e88b64ea | 832 | ffree = statp->f_files - (icount - ifree); |
c8ce540d | 833 | statp->f_ffree = max_t(int64_t, ffree, 0); |
2fe33661 | 834 | |
4ca488eb | 835 | |
da5bf95e | 836 | if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) && |
83e782e1 CS |
837 | ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) == |
838 | (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD)) | |
7d095257 | 839 | xfs_qm_statvfs(ip, statp); |
a0158315 RW |
840 | |
841 | if (XFS_IS_REALTIME_MOUNT(mp) && | |
842 | (ip->i_d.di_flags & (XFS_DIFLAG_RTINHERIT | XFS_DIFLAG_REALTIME))) { | |
843 | statp->f_blocks = sbp->sb_rblocks; | |
844 | statp->f_bavail = statp->f_bfree = | |
845 | sbp->sb_frextents * sbp->sb_rextsize; | |
846 | } | |
847 | ||
4ca488eb | 848 | return 0; |
1da177e4 LT |
849 | } |
850 | ||
d5db0f97 ES |
851 | STATIC void |
852 | xfs_save_resvblks(struct xfs_mount *mp) | |
853 | { | |
c8ce540d | 854 | uint64_t resblks = 0; |
d5db0f97 ES |
855 | |
856 | mp->m_resblks_save = mp->m_resblks; | |
857 | xfs_reserve_blocks(mp, &resblks, NULL); | |
858 | } | |
859 | ||
860 | STATIC void | |
861 | xfs_restore_resvblks(struct xfs_mount *mp) | |
862 | { | |
c8ce540d | 863 | uint64_t resblks; |
d5db0f97 ES |
864 | |
865 | if (mp->m_resblks_save) { | |
866 | resblks = mp->m_resblks_save; | |
867 | mp->m_resblks_save = 0; | |
868 | } else | |
869 | resblks = xfs_default_resblks(mp); | |
870 | ||
871 | xfs_reserve_blocks(mp, &resblks, NULL); | |
872 | } | |
873 | ||
8757c38f IK |
874 | /* |
875 | * Second stage of a freeze. The data is already frozen so we only | |
876 | * need to take care of the metadata. Once that's done sync the superblock | |
877 | * to the log to dirty it in case of a crash while frozen. This ensures that we | |
878 | * will recover the unlinked inode lists on the next mount. | |
879 | */ | |
880 | STATIC int | |
881 | xfs_fs_freeze( | |
882 | struct super_block *sb) | |
883 | { | |
884 | struct xfs_mount *mp = XFS_M(sb); | |
c3f2375b WL |
885 | unsigned int flags; |
886 | int ret; | |
8757c38f | 887 | |
c3f2375b WL |
888 | /* |
889 | * The filesystem is now frozen far enough that memory reclaim | |
890 | * cannot safely operate on the filesystem. Hence we need to | |
891 | * set a GFP_NOFS context here to avoid recursion deadlocks. | |
892 | */ | |
893 | flags = memalloc_nofs_save(); | |
c9a6526f | 894 | xfs_blockgc_stop(mp); |
8757c38f | 895 | xfs_save_resvblks(mp); |
5b0ad7c2 | 896 | ret = xfs_log_quiesce(mp); |
c3f2375b WL |
897 | memalloc_nofs_restore(flags); |
898 | return ret; | |
8757c38f IK |
899 | } |
900 | ||
901 | STATIC int | |
902 | xfs_fs_unfreeze( | |
903 | struct super_block *sb) | |
904 | { | |
905 | struct xfs_mount *mp = XFS_M(sb); | |
906 | ||
907 | xfs_restore_resvblks(mp); | |
908 | xfs_log_work_queue(mp); | |
c9a6526f | 909 | xfs_blockgc_start(mp); |
8757c38f IK |
910 | return 0; |
911 | } | |
912 | ||
913 | /* | |
914 | * This function fills in xfs_mount_t fields based on mount args. | |
915 | * Note: the superblock _has_ now been read in. | |
916 | */ | |
917 | STATIC int | |
918 | xfs_finish_flags( | |
919 | struct xfs_mount *mp) | |
920 | { | |
921 | int ronly = (mp->m_flags & XFS_MOUNT_RDONLY); | |
922 | ||
923 | /* Fail a mount where the logbuf is smaller than the log stripe */ | |
924 | if (xfs_sb_version_haslogv2(&mp->m_sb)) { | |
925 | if (mp->m_logbsize <= 0 && | |
926 | mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) { | |
927 | mp->m_logbsize = mp->m_sb.sb_logsunit; | |
928 | } else if (mp->m_logbsize > 0 && | |
929 | mp->m_logbsize < mp->m_sb.sb_logsunit) { | |
930 | xfs_warn(mp, | |
931 | "logbuf size must be greater than or equal to log stripe size"); | |
932 | return -EINVAL; | |
933 | } | |
934 | } else { | |
935 | /* Fail a mount if the logbuf is larger than 32K */ | |
936 | if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) { | |
937 | xfs_warn(mp, | |
938 | "logbuf size for version 1 logs must be 16K or 32K"); | |
939 | return -EINVAL; | |
940 | } | |
941 | } | |
942 | ||
943 | /* | |
944 | * V5 filesystems always use attr2 format for attributes. | |
945 | */ | |
946 | if (xfs_sb_version_hascrc(&mp->m_sb) && | |
947 | (mp->m_flags & XFS_MOUNT_NOATTR2)) { | |
948 | xfs_warn(mp, "Cannot mount a V5 filesystem as noattr2. " | |
949 | "attr2 is always enabled for V5 filesystems."); | |
950 | return -EINVAL; | |
951 | } | |
952 | ||
953 | /* | |
954 | * mkfs'ed attr2 will turn on attr2 mount unless explicitly | |
955 | * told by noattr2 to turn it off | |
956 | */ | |
957 | if (xfs_sb_version_hasattr2(&mp->m_sb) && | |
958 | !(mp->m_flags & XFS_MOUNT_NOATTR2)) | |
959 | mp->m_flags |= XFS_MOUNT_ATTR2; | |
960 | ||
961 | /* | |
962 | * prohibit r/w mounts of read-only filesystems | |
963 | */ | |
964 | if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) { | |
965 | xfs_warn(mp, | |
966 | "cannot mount a read-only filesystem as read-write"); | |
967 | return -EROFS; | |
968 | } | |
969 | ||
970 | if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) && | |
971 | (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) && | |
972 | !xfs_sb_version_has_pquotino(&mp->m_sb)) { | |
973 | xfs_warn(mp, | |
974 | "Super block does not support project and group quota together"); | |
975 | return -EINVAL; | |
976 | } | |
977 | ||
978 | return 0; | |
979 | } | |
980 | ||
981 | static int | |
982 | xfs_init_percpu_counters( | |
983 | struct xfs_mount *mp) | |
984 | { | |
985 | int error; | |
986 | ||
987 | error = percpu_counter_init(&mp->m_icount, 0, GFP_KERNEL); | |
988 | if (error) | |
989 | return -ENOMEM; | |
990 | ||
991 | error = percpu_counter_init(&mp->m_ifree, 0, GFP_KERNEL); | |
992 | if (error) | |
993 | goto free_icount; | |
994 | ||
995 | error = percpu_counter_init(&mp->m_fdblocks, 0, GFP_KERNEL); | |
996 | if (error) | |
997 | goto free_ifree; | |
998 | ||
999 | error = percpu_counter_init(&mp->m_delalloc_blks, 0, GFP_KERNEL); | |
1000 | if (error) | |
1001 | goto free_fdblocks; | |
1002 | ||
1003 | return 0; | |
1004 | ||
1005 | free_fdblocks: | |
1006 | percpu_counter_destroy(&mp->m_fdblocks); | |
1007 | free_ifree: | |
1008 | percpu_counter_destroy(&mp->m_ifree); | |
1009 | free_icount: | |
1010 | percpu_counter_destroy(&mp->m_icount); | |
1011 | return -ENOMEM; | |
1012 | } | |
1013 | ||
1014 | void | |
1015 | xfs_reinit_percpu_counters( | |
1016 | struct xfs_mount *mp) | |
1017 | { | |
1018 | percpu_counter_set(&mp->m_icount, mp->m_sb.sb_icount); | |
1019 | percpu_counter_set(&mp->m_ifree, mp->m_sb.sb_ifree); | |
1020 | percpu_counter_set(&mp->m_fdblocks, mp->m_sb.sb_fdblocks); | |
1021 | } | |
1022 | ||
1023 | static void | |
1024 | xfs_destroy_percpu_counters( | |
1025 | struct xfs_mount *mp) | |
1026 | { | |
1027 | percpu_counter_destroy(&mp->m_icount); | |
1028 | percpu_counter_destroy(&mp->m_ifree); | |
1029 | percpu_counter_destroy(&mp->m_fdblocks); | |
1030 | ASSERT(XFS_FORCED_SHUTDOWN(mp) || | |
1031 | percpu_counter_sum(&mp->m_delalloc_blks) == 0); | |
1032 | percpu_counter_destroy(&mp->m_delalloc_blks); | |
1033 | } | |
1034 | ||
1035 | static void | |
1036 | xfs_fs_put_super( | |
1037 | struct super_block *sb) | |
1038 | { | |
1039 | struct xfs_mount *mp = XFS_M(sb); | |
1040 | ||
1041 | /* if ->fill_super failed, we have no mount to tear down */ | |
1042 | if (!sb->s_fs_info) | |
1043 | return; | |
1044 | ||
1045 | xfs_notice(mp, "Unmounting Filesystem"); | |
1046 | xfs_filestream_unmount(mp); | |
1047 | xfs_unmountfs(mp); | |
1048 | ||
1049 | xfs_freesb(mp); | |
1050 | free_percpu(mp->m_stats.xs_stats); | |
1051 | xfs_destroy_percpu_counters(mp); | |
1052 | xfs_destroy_mount_workqueues(mp); | |
1053 | xfs_close_devices(mp); | |
1054 | ||
1055 | sb->s_fs_info = NULL; | |
1056 | xfs_mount_free(mp); | |
1057 | } | |
1058 | ||
1059 | static long | |
1060 | xfs_fs_nr_cached_objects( | |
1061 | struct super_block *sb, | |
1062 | struct shrink_control *sc) | |
1063 | { | |
1064 | /* Paranoia: catch incorrect calls during mount setup or teardown */ | |
1065 | if (WARN_ON_ONCE(!sb->s_fs_info)) | |
1066 | return 0; | |
1067 | return xfs_reclaim_inodes_count(XFS_M(sb)); | |
1068 | } | |
1069 | ||
1070 | static long | |
1071 | xfs_fs_free_cached_objects( | |
1072 | struct super_block *sb, | |
1073 | struct shrink_control *sc) | |
c7eea6f7 | 1074 | { |
8757c38f IK |
1075 | return xfs_reclaim_inodes_nr(XFS_M(sb), sc->nr_to_scan); |
1076 | } | |
c7eea6f7 | 1077 | |
8757c38f IK |
1078 | static const struct super_operations xfs_super_operations = { |
1079 | .alloc_inode = xfs_fs_alloc_inode, | |
1080 | .destroy_inode = xfs_fs_destroy_inode, | |
1081 | .dirty_inode = xfs_fs_dirty_inode, | |
1082 | .drop_inode = xfs_fs_drop_inode, | |
1083 | .put_super = xfs_fs_put_super, | |
1084 | .sync_fs = xfs_fs_sync_fs, | |
1085 | .freeze_fs = xfs_fs_freeze, | |
1086 | .unfreeze_fs = xfs_fs_unfreeze, | |
1087 | .statfs = xfs_fs_statfs, | |
1088 | .show_options = xfs_fs_show_options, | |
1089 | .nr_cached_objects = xfs_fs_nr_cached_objects, | |
1090 | .free_cached_objects = xfs_fs_free_cached_objects, | |
1091 | }; | |
c7eea6f7 | 1092 | |
8757c38f IK |
1093 | static int |
1094 | suffix_kstrtoint( | |
1095 | const char *s, | |
1096 | unsigned int base, | |
1097 | int *res) | |
1098 | { | |
1099 | int last, shift_left_factor = 0, _res; | |
1100 | char *value; | |
1101 | int ret = 0; | |
1102 | ||
1103 | value = kstrdup(s, GFP_KERNEL); | |
1104 | if (!value) | |
1105 | return -ENOMEM; | |
1106 | ||
1107 | last = strlen(value) - 1; | |
1108 | if (value[last] == 'K' || value[last] == 'k') { | |
1109 | shift_left_factor = 10; | |
1110 | value[last] = '\0'; | |
1111 | } | |
1112 | if (value[last] == 'M' || value[last] == 'm') { | |
1113 | shift_left_factor = 20; | |
1114 | value[last] = '\0'; | |
1115 | } | |
1116 | if (value[last] == 'G' || value[last] == 'g') { | |
1117 | shift_left_factor = 30; | |
1118 | value[last] = '\0'; | |
1119 | } | |
1120 | ||
1121 | if (kstrtoint(value, base, &_res)) | |
1122 | ret = -EINVAL; | |
1123 | kfree(value); | |
1124 | *res = _res << shift_left_factor; | |
1125 | return ret; | |
c7eea6f7 DC |
1126 | } |
1127 | ||
9909c4aa | 1128 | /* |
8757c38f IK |
1129 | * Set mount state from a mount option. |
1130 | * | |
1131 | * NOTE: mp->m_super is NULL here! | |
9909c4aa | 1132 | */ |
8757c38f | 1133 | static int |
1e5c39df | 1134 | xfs_fs_parse_param( |
8757c38f IK |
1135 | struct fs_context *fc, |
1136 | struct fs_parameter *param) | |
1da177e4 | 1137 | { |
8757c38f IK |
1138 | struct xfs_mount *mp = fc->s_fs_info; |
1139 | struct fs_parse_result result; | |
1140 | int size = 0; | |
1141 | int opt; | |
9909c4aa | 1142 | |
d7167b14 | 1143 | opt = fs_parse(fc, xfs_fs_parameters, param, &result); |
8757c38f IK |
1144 | if (opt < 0) |
1145 | return opt; | |
1da177e4 | 1146 | |
8757c38f IK |
1147 | switch (opt) { |
1148 | case Opt_logbufs: | |
1149 | mp->m_logbufs = result.uint_32; | |
1150 | return 0; | |
1151 | case Opt_logbsize: | |
1152 | if (suffix_kstrtoint(param->string, 10, &mp->m_logbsize)) | |
1153 | return -EINVAL; | |
1154 | return 0; | |
1155 | case Opt_logdev: | |
1156 | kfree(mp->m_logname); | |
1157 | mp->m_logname = kstrdup(param->string, GFP_KERNEL); | |
1158 | if (!mp->m_logname) | |
1159 | return -ENOMEM; | |
1160 | return 0; | |
1161 | case Opt_rtdev: | |
1162 | kfree(mp->m_rtname); | |
1163 | mp->m_rtname = kstrdup(param->string, GFP_KERNEL); | |
1164 | if (!mp->m_rtname) | |
1165 | return -ENOMEM; | |
1166 | return 0; | |
1167 | case Opt_allocsize: | |
1168 | if (suffix_kstrtoint(param->string, 10, &size)) | |
1169 | return -EINVAL; | |
1170 | mp->m_allocsize_log = ffs(size) - 1; | |
1171 | mp->m_flags |= XFS_MOUNT_ALLOCSIZE; | |
1172 | return 0; | |
1173 | case Opt_grpid: | |
1174 | case Opt_bsdgroups: | |
1175 | mp->m_flags |= XFS_MOUNT_GRPID; | |
1176 | return 0; | |
1177 | case Opt_nogrpid: | |
1178 | case Opt_sysvgroups: | |
1179 | mp->m_flags &= ~XFS_MOUNT_GRPID; | |
1180 | return 0; | |
1181 | case Opt_wsync: | |
1182 | mp->m_flags |= XFS_MOUNT_WSYNC; | |
1183 | return 0; | |
1184 | case Opt_norecovery: | |
1185 | mp->m_flags |= XFS_MOUNT_NORECOVERY; | |
1186 | return 0; | |
1187 | case Opt_noalign: | |
1188 | mp->m_flags |= XFS_MOUNT_NOALIGN; | |
1189 | return 0; | |
1190 | case Opt_swalloc: | |
1191 | mp->m_flags |= XFS_MOUNT_SWALLOC; | |
1192 | return 0; | |
1193 | case Opt_sunit: | |
1194 | mp->m_dalign = result.uint_32; | |
1195 | return 0; | |
1196 | case Opt_swidth: | |
1197 | mp->m_swidth = result.uint_32; | |
1198 | return 0; | |
1199 | case Opt_inode32: | |
1200 | mp->m_flags |= XFS_MOUNT_SMALL_INUMS; | |
1201 | return 0; | |
1202 | case Opt_inode64: | |
1203 | mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS; | |
1204 | return 0; | |
1205 | case Opt_nouuid: | |
1206 | mp->m_flags |= XFS_MOUNT_NOUUID; | |
1207 | return 0; | |
8757c38f IK |
1208 | case Opt_largeio: |
1209 | mp->m_flags |= XFS_MOUNT_LARGEIO; | |
1210 | return 0; | |
1211 | case Opt_nolargeio: | |
1212 | mp->m_flags &= ~XFS_MOUNT_LARGEIO; | |
1213 | return 0; | |
8757c38f IK |
1214 | case Opt_filestreams: |
1215 | mp->m_flags |= XFS_MOUNT_FILESTREAMS; | |
1216 | return 0; | |
1217 | case Opt_noquota: | |
1218 | mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT; | |
1219 | mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD; | |
1220 | mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE; | |
1221 | return 0; | |
1222 | case Opt_quota: | |
1223 | case Opt_uquota: | |
1224 | case Opt_usrquota: | |
1225 | mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE | | |
1226 | XFS_UQUOTA_ENFD); | |
1227 | return 0; | |
1228 | case Opt_qnoenforce: | |
1229 | case Opt_uqnoenforce: | |
1230 | mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE); | |
1231 | mp->m_qflags &= ~XFS_UQUOTA_ENFD; | |
1232 | return 0; | |
1233 | case Opt_pquota: | |
1234 | case Opt_prjquota: | |
1235 | mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE | | |
1236 | XFS_PQUOTA_ENFD); | |
1237 | return 0; | |
1238 | case Opt_pqnoenforce: | |
1239 | mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE); | |
1240 | mp->m_qflags &= ~XFS_PQUOTA_ENFD; | |
1241 | return 0; | |
1242 | case Opt_gquota: | |
1243 | case Opt_grpquota: | |
1244 | mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE | | |
1245 | XFS_GQUOTA_ENFD); | |
1246 | return 0; | |
1247 | case Opt_gqnoenforce: | |
1248 | mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE); | |
1249 | mp->m_qflags &= ~XFS_GQUOTA_ENFD; | |
1250 | return 0; | |
1251 | case Opt_discard: | |
1252 | mp->m_flags |= XFS_MOUNT_DISCARD; | |
1253 | return 0; | |
1254 | case Opt_nodiscard: | |
1255 | mp->m_flags &= ~XFS_MOUNT_DISCARD; | |
1256 | return 0; | |
1257 | #ifdef CONFIG_FS_DAX | |
1258 | case Opt_dax: | |
8d6c3446 IW |
1259 | xfs_mount_set_dax_mode(mp, XFS_DAX_ALWAYS); |
1260 | return 0; | |
1261 | case Opt_dax_enum: | |
1262 | xfs_mount_set_dax_mode(mp, result.uint_32); | |
8757c38f IK |
1263 | return 0; |
1264 | #endif | |
c23c393e PR |
1265 | /* Following mount options will be removed in September 2025 */ |
1266 | case Opt_ikeep: | |
1267 | xfs_warn(mp, "%s mount option is deprecated.", param->key); | |
1268 | mp->m_flags |= XFS_MOUNT_IKEEP; | |
1269 | return 0; | |
1270 | case Opt_noikeep: | |
1271 | xfs_warn(mp, "%s mount option is deprecated.", param->key); | |
1272 | mp->m_flags &= ~XFS_MOUNT_IKEEP; | |
1273 | return 0; | |
1274 | case Opt_attr2: | |
1275 | xfs_warn(mp, "%s mount option is deprecated.", param->key); | |
1276 | mp->m_flags |= XFS_MOUNT_ATTR2; | |
1277 | return 0; | |
1278 | case Opt_noattr2: | |
1279 | xfs_warn(mp, "%s mount option is deprecated.", param->key); | |
1280 | mp->m_flags &= ~XFS_MOUNT_ATTR2; | |
1281 | mp->m_flags |= XFS_MOUNT_NOATTR2; | |
1282 | return 0; | |
8757c38f IK |
1283 | default: |
1284 | xfs_warn(mp, "unknown mount option [%s].", param->key); | |
1285 | return -EINVAL; | |
1286 | } | |
d5db0f97 | 1287 | |
d5db0f97 ES |
1288 | return 0; |
1289 | } | |
1290 | ||
8757c38f | 1291 | static int |
1e5c39df | 1292 | xfs_fs_validate_params( |
f8f15e42 CH |
1293 | struct xfs_mount *mp) |
1294 | { | |
d3eaace8 | 1295 | /* |
8757c38f | 1296 | * no recovery flag requires a read-only mount |
d3eaace8 | 1297 | */ |
8757c38f IK |
1298 | if ((mp->m_flags & XFS_MOUNT_NORECOVERY) && |
1299 | !(mp->m_flags & XFS_MOUNT_RDONLY)) { | |
1300 | xfs_warn(mp, "no-recovery mounts must be read-only."); | |
2451337d | 1301 | return -EINVAL; |
d3eaace8 DC |
1302 | } |
1303 | ||
8757c38f IK |
1304 | if ((mp->m_flags & XFS_MOUNT_NOALIGN) && |
1305 | (mp->m_dalign || mp->m_swidth)) { | |
4f10700a | 1306 | xfs_warn(mp, |
8757c38f IK |
1307 | "sunit and swidth options incompatible with the noalign option"); |
1308 | return -EINVAL; | |
f8f15e42 CH |
1309 | } |
1310 | ||
8757c38f IK |
1311 | if (!IS_ENABLED(CONFIG_XFS_QUOTA) && mp->m_qflags != 0) { |
1312 | xfs_warn(mp, "quota support not available in this kernel."); | |
2451337d | 1313 | return -EINVAL; |
d892d586 CS |
1314 | } |
1315 | ||
8757c38f IK |
1316 | if ((mp->m_dalign && !mp->m_swidth) || |
1317 | (!mp->m_dalign && mp->m_swidth)) { | |
1318 | xfs_warn(mp, "sunit and swidth must be specified together"); | |
1319 | return -EINVAL; | |
1320 | } | |
5681ca40 | 1321 | |
8757c38f IK |
1322 | if (mp->m_dalign && (mp->m_swidth % mp->m_dalign != 0)) { |
1323 | xfs_warn(mp, | |
1324 | "stripe width (%d) must be a multiple of the stripe unit (%d)", | |
1325 | mp->m_swidth, mp->m_dalign); | |
1326 | return -EINVAL; | |
1327 | } | |
5681ca40 | 1328 | |
8757c38f IK |
1329 | if (mp->m_logbufs != -1 && |
1330 | mp->m_logbufs != 0 && | |
1331 | (mp->m_logbufs < XLOG_MIN_ICLOGS || | |
1332 | mp->m_logbufs > XLOG_MAX_ICLOGS)) { | |
1333 | xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]", | |
1334 | mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS); | |
1335 | return -EINVAL; | |
1336 | } | |
5681ca40 | 1337 | |
8757c38f IK |
1338 | if (mp->m_logbsize != -1 && |
1339 | mp->m_logbsize != 0 && | |
1340 | (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE || | |
1341 | mp->m_logbsize > XLOG_MAX_RECORD_BSIZE || | |
1342 | !is_power_of_2(mp->m_logbsize))) { | |
1343 | xfs_warn(mp, | |
1344 | "invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]", | |
1345 | mp->m_logbsize); | |
1346 | return -EINVAL; | |
1347 | } | |
5681ca40 | 1348 | |
8757c38f IK |
1349 | if ((mp->m_flags & XFS_MOUNT_ALLOCSIZE) && |
1350 | (mp->m_allocsize_log > XFS_MAX_IO_LOG || | |
1351 | mp->m_allocsize_log < XFS_MIN_IO_LOG)) { | |
1352 | xfs_warn(mp, "invalid log iosize: %d [not %d-%d]", | |
1353 | mp->m_allocsize_log, XFS_MIN_IO_LOG, XFS_MAX_IO_LOG); | |
1354 | return -EINVAL; | |
1355 | } | |
9fe82b8c | 1356 | |
5681ca40 | 1357 | return 0; |
72c44e35 | 1358 | } |
74394496 | 1359 | |
73e5fff9 | 1360 | static int |
1e5c39df | 1361 | xfs_fs_fill_super( |
72c44e35 | 1362 | struct super_block *sb, |
73e5fff9 | 1363 | struct fs_context *fc) |
72c44e35 | 1364 | { |
73e5fff9 | 1365 | struct xfs_mount *mp = sb->s_fs_info; |
72c44e35 | 1366 | struct inode *root; |
0279c71f | 1367 | int flags = 0, error; |
72c44e35 | 1368 | |
7c89fcb2 | 1369 | mp->m_super = sb; |
1da177e4 | 1370 | |
1e5c39df | 1371 | error = xfs_fs_validate_params(mp); |
745f6919 | 1372 | if (error) |
e1d3d218 | 1373 | goto out_free_names; |
1da177e4 LT |
1374 | |
1375 | sb_min_blocksize(sb, BBSIZE); | |
0ec58516 | 1376 | sb->s_xattr = xfs_xattr_handlers; |
a50cd269 | 1377 | sb->s_export_op = &xfs_export_operations; |
fcafb71b | 1378 | #ifdef CONFIG_XFS_QUOTA |
a50cd269 | 1379 | sb->s_qcop = &xfs_quotactl_operations; |
17ef4fdd | 1380 | sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; |
fcafb71b | 1381 | #endif |
a50cd269 | 1382 | sb->s_op = &xfs_super_operations; |
1da177e4 | 1383 | |
dae5cd81 DC |
1384 | /* |
1385 | * Delay mount work if the debug hook is set. This is debug | |
1386 | * instrumention to coordinate simulation of xfs mount failures with | |
1387 | * VFS superblock operations | |
1388 | */ | |
1389 | if (xfs_globals.mount_delay) { | |
1390 | xfs_notice(mp, "Delaying mount for %d seconds.", | |
1391 | xfs_globals.mount_delay); | |
1392 | msleep(xfs_globals.mount_delay * 1000); | |
1393 | } | |
1394 | ||
73e5fff9 | 1395 | if (fc->sb_flags & SB_SILENT) |
f8f15e42 CH |
1396 | flags |= XFS_MFSI_QUIET; |
1397 | ||
9d565ffa | 1398 | error = xfs_open_devices(mp); |
19f354d4 | 1399 | if (error) |
e1d3d218 | 1400 | goto out_free_names; |
f8f15e42 | 1401 | |
2451337d | 1402 | error = xfs_init_mount_workqueues(mp); |
61ba35de CH |
1403 | if (error) |
1404 | goto out_close_devices; | |
c962fb79 | 1405 | |
5681ca40 | 1406 | error = xfs_init_percpu_counters(mp); |
aa6bf01d CH |
1407 | if (error) |
1408 | goto out_destroy_workqueues; | |
1409 | ||
225e4635 BD |
1410 | /* Allocate stats memory before we do operations that might use it */ |
1411 | mp->m_stats.xs_stats = alloc_percpu(struct xfsstats); | |
1412 | if (!mp->m_stats.xs_stats) { | |
f9d460b3 | 1413 | error = -ENOMEM; |
225e4635 BD |
1414 | goto out_destroy_counters; |
1415 | } | |
1416 | ||
f8f15e42 CH |
1417 | error = xfs_readsb(mp, flags); |
1418 | if (error) | |
225e4635 | 1419 | goto out_free_stats; |
9d565ffa CH |
1420 | |
1421 | error = xfs_finish_flags(mp); | |
f8f15e42 | 1422 | if (error) |
effa2eda | 1423 | goto out_free_sb; |
f8f15e42 | 1424 | |
e34b562c | 1425 | error = xfs_setup_devices(mp); |
19f354d4 | 1426 | if (error) |
effa2eda | 1427 | goto out_free_sb; |
f8f15e42 | 1428 | |
b96cb835 DW |
1429 | /* V4 support is undergoing deprecation. */ |
1430 | if (!xfs_sb_version_hascrc(&mp->m_sb)) { | |
1431 | #ifdef CONFIG_XFS_SUPPORT_V4 | |
1432 | xfs_warn_once(mp, | |
1433 | "Deprecated V4 format (crc=0) will not be supported after September 2030."); | |
1434 | #else | |
1435 | xfs_warn(mp, | |
1436 | "Deprecated V4 format (crc=0) not supported by kernel."); | |
1437 | error = -EINVAL; | |
1438 | goto out_free_sb; | |
1439 | #endif | |
1440 | } | |
1441 | ||
80c720b8 DW |
1442 | /* Filesystem claims it needs repair, so refuse the mount. */ |
1443 | if (xfs_sb_version_needsrepair(&mp->m_sb)) { | |
1444 | xfs_warn(mp, "Filesystem needs repair. Please run xfs_repair."); | |
1445 | error = -EFSCORRUPTED; | |
1446 | goto out_free_sb; | |
1447 | } | |
1448 | ||
3945ae03 DW |
1449 | /* |
1450 | * Don't touch the filesystem if a user tool thinks it owns the primary | |
1451 | * superblock. mkfs doesn't clear the flag from secondary supers, so | |
1452 | * we don't check them at all. | |
1453 | */ | |
1454 | if (mp->m_sb.sb_inprogress) { | |
1455 | xfs_warn(mp, "Offline file system operation in progress!"); | |
1456 | error = -EFSCORRUPTED; | |
1457 | goto out_free_sb; | |
1458 | } | |
1459 | ||
1460 | /* | |
1461 | * Until this is fixed only page-sized or smaller data blocks work. | |
1462 | */ | |
1463 | if (mp->m_sb.sb_blocksize > PAGE_SIZE) { | |
1464 | xfs_warn(mp, | |
1465 | "File system with blocksize %d bytes. " | |
1466 | "Only pagesize (%ld) or less will currently work.", | |
1467 | mp->m_sb.sb_blocksize, PAGE_SIZE); | |
1468 | error = -ENOSYS; | |
1469 | goto out_free_sb; | |
1470 | } | |
1471 | ||
1472 | /* Ensure this filesystem fits in the page cache limits */ | |
1473 | if (xfs_sb_validate_fsb_count(&mp->m_sb, mp->m_sb.sb_dblocks) || | |
1474 | xfs_sb_validate_fsb_count(&mp->m_sb, mp->m_sb.sb_rblocks)) { | |
1475 | xfs_warn(mp, | |
1476 | "file system too large to be mounted on this system."); | |
1477 | error = -EFBIG; | |
1478 | goto out_free_sb; | |
1479 | } | |
1480 | ||
932befe3 DW |
1481 | /* |
1482 | * XFS block mappings use 54 bits to store the logical block offset. | |
1483 | * This should suffice to handle the maximum file size that the VFS | |
1484 | * supports (currently 2^63 bytes on 64-bit and ULONG_MAX << PAGE_SHIFT | |
1485 | * bytes on 32-bit), but as XFS and VFS have gotten the s_maxbytes | |
1486 | * calculation wrong on 32-bit kernels in the past, we'll add a WARN_ON | |
1487 | * to check this assertion. | |
1488 | * | |
1489 | * Avoid integer overflow by comparing the maximum bmbt offset to the | |
1490 | * maximum pagecache offset in units of fs blocks. | |
1491 | */ | |
33005fd0 | 1492 | if (!xfs_verify_fileoff(mp, XFS_B_TO_FSBT(mp, MAX_LFS_FILESIZE))) { |
932befe3 DW |
1493 | xfs_warn(mp, |
1494 | "MAX_LFS_FILESIZE block offset (%llu) exceeds extent map maximum (%llu)!", | |
1495 | XFS_B_TO_FSBT(mp, MAX_LFS_FILESIZE), | |
1496 | XFS_MAX_FILEOFF); | |
1497 | error = -EINVAL; | |
1498 | goto out_free_sb; | |
1499 | } | |
1500 | ||
f8f15e42 CH |
1501 | error = xfs_filestream_mount(mp); |
1502 | if (error) | |
effa2eda | 1503 | goto out_free_sb; |
f8f15e42 | 1504 | |
704b2907 DC |
1505 | /* |
1506 | * we must configure the block size in the superblock before we run the | |
1507 | * full mount process as the mount process can lookup and cache inodes. | |
704b2907 | 1508 | */ |
dddde68b | 1509 | sb->s_magic = XFS_SUPER_MAGIC; |
4ca488eb CH |
1510 | sb->s_blocksize = mp->m_sb.sb_blocksize; |
1511 | sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1; | |
932befe3 | 1512 | sb->s_maxbytes = MAX_LFS_FILESIZE; |
8de52778 | 1513 | sb->s_max_links = XFS_MAXLINK; |
1da177e4 | 1514 | sb->s_time_gran = 1; |
f93e5436 DW |
1515 | if (xfs_sb_version_hasbigtime(&mp->m_sb)) { |
1516 | sb->s_time_min = xfs_bigtime_to_unix(XFS_BIGTIME_TIME_MIN); | |
1517 | sb->s_time_max = xfs_bigtime_to_unix(XFS_BIGTIME_TIME_MAX); | |
1518 | } else { | |
1519 | sb->s_time_min = XFS_LEGACY_TIME_MIN; | |
1520 | sb->s_time_max = XFS_LEGACY_TIME_MAX; | |
1521 | } | |
06dbf82b | 1522 | trace_xfs_inode_timestamp_range(mp, sb->s_time_min, sb->s_time_max); |
adfb5fb4 CH |
1523 | sb->s_iflags |= SB_I_CGROUPWB; |
1524 | ||
1da177e4 LT |
1525 | set_posix_acl_flag(sb); |
1526 | ||
dc037ad7 DC |
1527 | /* version 5 superblocks support inode version counters. */ |
1528 | if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5) | |
357fdad0 | 1529 | sb->s_flags |= SB_I_VERSION; |
dc037ad7 | 1530 | |
f93e5436 DW |
1531 | if (xfs_sb_version_hasbigtime(&mp->m_sb)) |
1532 | xfs_warn(mp, | |
1533 | "EXPERIMENTAL big timestamp feature in use. Use at your own risk!"); | |
1534 | ||
606723d9 | 1535 | if (mp->m_flags & XFS_MOUNT_DAX_ALWAYS) { |
80660f20 | 1536 | bool rtdev_is_dax = false, datadev_is_dax; |
ba23cba9 | 1537 | |
cbe4dab1 | 1538 | xfs_warn(mp, |
1e937cdd TK |
1539 | "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); |
1540 | ||
80660f20 DJ |
1541 | datadev_is_dax = bdev_dax_supported(mp->m_ddev_targp->bt_bdev, |
1542 | sb->s_blocksize); | |
ba23cba9 | 1543 | if (mp->m_rtdev_targp) |
80660f20 DJ |
1544 | rtdev_is_dax = bdev_dax_supported( |
1545 | mp->m_rtdev_targp->bt_bdev, sb->s_blocksize); | |
1546 | if (!rtdev_is_dax && !datadev_is_dax) { | |
cbe4dab1 | 1547 | xfs_alert(mp, |
1e937cdd | 1548 | "DAX unsupported by block device. Turning off DAX."); |
8d6c3446 | 1549 | xfs_mount_set_dax_mode(mp, XFS_DAX_NEVER); |
cbe4dab1 | 1550 | } |
b6e03c10 | 1551 | if (xfs_sb_version_hasreflink(&mp->m_sb)) { |
e54b5bf9 | 1552 | xfs_alert(mp, |
1e369b0e | 1553 | "DAX and reflink cannot be used together!"); |
b6e03c10 DW |
1554 | error = -EINVAL; |
1555 | goto out_filestream_unmount; | |
1556 | } | |
cbe4dab1 DC |
1557 | } |
1558 | ||
1e6fa688 KN |
1559 | if (mp->m_flags & XFS_MOUNT_DISCARD) { |
1560 | struct request_queue *q = bdev_get_queue(sb->s_bdev); | |
1561 | ||
1562 | if (!blk_queue_discard(q)) { | |
1563 | xfs_warn(mp, "mounting with \"discard\" option, but " | |
1564 | "the device does not support discard"); | |
1565 | mp->m_flags &= ~XFS_MOUNT_DISCARD; | |
1566 | } | |
1567 | } | |
1568 | ||
66ae56a5 CH |
1569 | if (xfs_sb_version_hasreflink(&mp->m_sb)) { |
1570 | if (mp->m_sb.sb_rblocks) { | |
1571 | xfs_alert(mp, | |
c14632dd | 1572 | "reflink not compatible with realtime device!"); |
66ae56a5 CH |
1573 | error = -EINVAL; |
1574 | goto out_filestream_unmount; | |
1575 | } | |
1576 | ||
1577 | if (xfs_globals.always_cow) { | |
1578 | xfs_info(mp, "using DEBUG-only always_cow mode."); | |
1579 | mp->m_always_cow = true; | |
1580 | } | |
c14632dd DW |
1581 | } |
1582 | ||
76883f79 | 1583 | if (xfs_sb_version_hasrmapbt(&mp->m_sb) && mp->m_sb.sb_rblocks) { |
1c0607ac | 1584 | xfs_alert(mp, |
76883f79 DW |
1585 | "reverse mapping btree not compatible with realtime device!"); |
1586 | error = -EINVAL; | |
1587 | goto out_filestream_unmount; | |
738f57c1 | 1588 | } |
1c0607ac | 1589 | |
2a39946c DW |
1590 | if (xfs_sb_version_hasinobtcounts(&mp->m_sb)) |
1591 | xfs_warn(mp, | |
1592 | "EXPERIMENTAL inode btree counters feature in use. Use at your own risk!"); | |
1593 | ||
8a00ebe4 | 1594 | error = xfs_mountfs(mp); |
2bcf6e97 | 1595 | if (error) |
7e18530b | 1596 | goto out_filestream_unmount; |
704b2907 | 1597 | |
01651646 | 1598 | root = igrab(VFS_I(mp->m_rootip)); |
f3dcc13f | 1599 | if (!root) { |
2451337d | 1600 | error = -ENOENT; |
8a00ebe4 | 1601 | goto out_unmount; |
cbc89dcf | 1602 | } |
48fde701 | 1603 | sb->s_root = d_make_root(root); |
f3dcc13f | 1604 | if (!sb->s_root) { |
2451337d | 1605 | error = -ENOMEM; |
8a00ebe4 | 1606 | goto out_unmount; |
1da177e4 | 1607 | } |
74394496 | 1608 | |
1da177e4 | 1609 | return 0; |
33c7a2bc | 1610 | |
7e18530b | 1611 | out_filestream_unmount: |
120226c1 | 1612 | xfs_filestream_unmount(mp); |
effa2eda CH |
1613 | out_free_sb: |
1614 | xfs_freesb(mp); | |
225e4635 BD |
1615 | out_free_stats: |
1616 | free_percpu(mp->m_stats.xs_stats); | |
9d565ffa | 1617 | out_destroy_counters: |
5681ca40 | 1618 | xfs_destroy_percpu_counters(mp); |
225e4635 | 1619 | out_destroy_workqueues: |
aa6bf01d | 1620 | xfs_destroy_mount_workqueues(mp); |
61ba35de | 1621 | out_close_devices: |
19f354d4 | 1622 | xfs_close_devices(mp); |
e1d3d218 | 1623 | out_free_names: |
c9fbd7bb | 1624 | sb->s_fs_info = NULL; |
a943f372 | 1625 | xfs_mount_free(mp); |
2451337d | 1626 | return error; |
f8f15e42 | 1627 | |
2bcf6e97 | 1628 | out_unmount: |
e48ad316 | 1629 | xfs_filestream_unmount(mp); |
19f354d4 | 1630 | xfs_unmountfs(mp); |
6203300e | 1631 | goto out_free_sb; |
1da177e4 LT |
1632 | } |
1633 | ||
73e5fff9 | 1634 | static int |
1e5c39df | 1635 | xfs_fs_get_tree( |
73e5fff9 IK |
1636 | struct fs_context *fc) |
1637 | { | |
1e5c39df | 1638 | return get_tree_bdev(fc, xfs_fs_fill_super); |
73e5fff9 IK |
1639 | } |
1640 | ||
63cd1e9b IK |
1641 | static int |
1642 | xfs_remount_rw( | |
1643 | struct xfs_mount *mp) | |
1644 | { | |
1645 | struct xfs_sb *sbp = &mp->m_sb; | |
1646 | int error; | |
1647 | ||
1648 | if (mp->m_flags & XFS_MOUNT_NORECOVERY) { | |
1649 | xfs_warn(mp, | |
1650 | "ro->rw transition prohibited on norecovery mount"); | |
1651 | return -EINVAL; | |
1652 | } | |
1653 | ||
1654 | if (XFS_SB_VERSION_NUM(sbp) == XFS_SB_VERSION_5 && | |
1655 | xfs_sb_has_ro_compat_feature(sbp, XFS_SB_FEAT_RO_COMPAT_UNKNOWN)) { | |
1656 | xfs_warn(mp, | |
1657 | "ro->rw transition prohibited on unknown (0x%x) ro-compat filesystem", | |
1658 | (sbp->sb_features_ro_compat & | |
1659 | XFS_SB_FEAT_RO_COMPAT_UNKNOWN)); | |
1660 | return -EINVAL; | |
1661 | } | |
1662 | ||
1663 | mp->m_flags &= ~XFS_MOUNT_RDONLY; | |
1664 | ||
1665 | /* | |
1666 | * If this is the first remount to writeable state we might have some | |
1667 | * superblock changes to update. | |
1668 | */ | |
1669 | if (mp->m_update_sb) { | |
1670 | error = xfs_sync_sb(mp, false); | |
1671 | if (error) { | |
1672 | xfs_warn(mp, "failed to write sb changes"); | |
1673 | return error; | |
1674 | } | |
1675 | mp->m_update_sb = false; | |
1676 | } | |
1677 | ||
1678 | /* | |
1679 | * Fill out the reserve pool if it is empty. Use the stashed value if | |
1680 | * it is non-zero, otherwise go with the default. | |
1681 | */ | |
1682 | xfs_restore_resvblks(mp); | |
1683 | xfs_log_work_queue(mp); | |
1684 | ||
1685 | /* Recover any CoW blocks that never got remapped. */ | |
1686 | error = xfs_reflink_recover_cow(mp); | |
1687 | if (error) { | |
1688 | xfs_err(mp, | |
1689 | "Error %d recovering leftover CoW allocations.", error); | |
7f6bcf7c | 1690 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); |
63cd1e9b IK |
1691 | return error; |
1692 | } | |
c9a6526f | 1693 | xfs_blockgc_start(mp); |
63cd1e9b IK |
1694 | |
1695 | /* Create the per-AG metadata reservation pool .*/ | |
1696 | error = xfs_fs_reserve_ag_blocks(mp); | |
1697 | if (error && error != -ENOSPC) | |
1698 | return error; | |
1699 | ||
1700 | return 0; | |
1701 | } | |
1702 | ||
1703 | static int | |
1704 | xfs_remount_ro( | |
1705 | struct xfs_mount *mp) | |
1706 | { | |
1707 | int error; | |
1708 | ||
1709 | /* | |
1710 | * Cancel background eofb scanning so it cannot race with the final | |
1711 | * log force+buftarg wait and deadlock the remount. | |
1712 | */ | |
c9a6526f | 1713 | xfs_blockgc_stop(mp); |
63cd1e9b IK |
1714 | |
1715 | /* Get rid of any leftover CoW reservations... */ | |
b943c0cd | 1716 | error = xfs_blockgc_free_space(mp, NULL); |
63cd1e9b IK |
1717 | if (error) { |
1718 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); | |
1719 | return error; | |
1720 | } | |
1721 | ||
1722 | /* Free the per-AG metadata reservation pool. */ | |
1723 | error = xfs_fs_unreserve_ag_blocks(mp); | |
1724 | if (error) { | |
1725 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); | |
1726 | return error; | |
1727 | } | |
1728 | ||
1729 | /* | |
1730 | * Before we sync the metadata, we need to free up the reserve block | |
1731 | * pool so that the used block count in the superblock on disk is | |
1732 | * correct at the end of the remount. Stash the current* reserve pool | |
1733 | * size so that if we get remounted rw, we can return it to the same | |
1734 | * size. | |
1735 | */ | |
1736 | xfs_save_resvblks(mp); | |
1737 | ||
ea2064da | 1738 | xfs_log_clean(mp); |
63cd1e9b IK |
1739 | mp->m_flags |= XFS_MOUNT_RDONLY; |
1740 | ||
1741 | return 0; | |
1742 | } | |
1743 | ||
1744 | /* | |
1745 | * Logically we would return an error here to prevent users from believing | |
1746 | * they might have changed mount options using remount which can't be changed. | |
1747 | * | |
1748 | * But unfortunately mount(8) adds all options from mtab and fstab to the mount | |
1749 | * arguments in some cases so we can't blindly reject options, but have to | |
1750 | * check for each specified option if it actually differs from the currently | |
1751 | * set option and only reject it if that's the case. | |
1752 | * | |
1753 | * Until that is implemented we return success for every remount request, and | |
1754 | * silently ignore all options that we can't actually change. | |
1755 | */ | |
1756 | static int | |
1e5c39df | 1757 | xfs_fs_reconfigure( |
63cd1e9b IK |
1758 | struct fs_context *fc) |
1759 | { | |
1760 | struct xfs_mount *mp = XFS_M(fc->root->d_sb); | |
1761 | struct xfs_mount *new_mp = fc->s_fs_info; | |
1762 | xfs_sb_t *sbp = &mp->m_sb; | |
1763 | int flags = fc->sb_flags; | |
1764 | int error; | |
1765 | ||
4750a171 ES |
1766 | /* version 5 superblocks always support version counters. */ |
1767 | if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5) | |
1768 | fc->sb_flags |= SB_I_VERSION; | |
1769 | ||
1e5c39df | 1770 | error = xfs_fs_validate_params(new_mp); |
63cd1e9b IK |
1771 | if (error) |
1772 | return error; | |
1773 | ||
1774 | sync_filesystem(mp->m_super); | |
1775 | ||
1776 | /* inode32 -> inode64 */ | |
1777 | if ((mp->m_flags & XFS_MOUNT_SMALL_INUMS) && | |
1778 | !(new_mp->m_flags & XFS_MOUNT_SMALL_INUMS)) { | |
1779 | mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS; | |
1780 | mp->m_maxagi = xfs_set_inode_alloc(mp, sbp->sb_agcount); | |
1781 | } | |
1782 | ||
1783 | /* inode64 -> inode32 */ | |
1784 | if (!(mp->m_flags & XFS_MOUNT_SMALL_INUMS) && | |
1785 | (new_mp->m_flags & XFS_MOUNT_SMALL_INUMS)) { | |
1786 | mp->m_flags |= XFS_MOUNT_SMALL_INUMS; | |
1787 | mp->m_maxagi = xfs_set_inode_alloc(mp, sbp->sb_agcount); | |
1788 | } | |
1789 | ||
1790 | /* ro -> rw */ | |
1791 | if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(flags & SB_RDONLY)) { | |
1792 | error = xfs_remount_rw(mp); | |
1793 | if (error) | |
1794 | return error; | |
1795 | } | |
1796 | ||
1797 | /* rw -> ro */ | |
1798 | if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (flags & SB_RDONLY)) { | |
1799 | error = xfs_remount_ro(mp); | |
1800 | if (error) | |
1801 | return error; | |
1802 | } | |
1803 | ||
1804 | return 0; | |
1805 | } | |
1806 | ||
1e5c39df | 1807 | static void xfs_fs_free( |
73e5fff9 IK |
1808 | struct fs_context *fc) |
1809 | { | |
1810 | struct xfs_mount *mp = fc->s_fs_info; | |
1811 | ||
1812 | /* | |
1813 | * mp is stored in the fs_context when it is initialized. | |
1814 | * mp is transferred to the superblock on a successful mount, | |
1815 | * but if an error occurs before the transfer we have to free | |
1816 | * it here. | |
1817 | */ | |
1818 | if (mp) | |
1819 | xfs_mount_free(mp); | |
1820 | } | |
1821 | ||
1822 | static const struct fs_context_operations xfs_context_ops = { | |
1e5c39df DW |
1823 | .parse_param = xfs_fs_parse_param, |
1824 | .get_tree = xfs_fs_get_tree, | |
1825 | .reconfigure = xfs_fs_reconfigure, | |
1826 | .free = xfs_fs_free, | |
73e5fff9 IK |
1827 | }; |
1828 | ||
1829 | static int xfs_init_fs_context( | |
1830 | struct fs_context *fc) | |
1831 | { | |
1832 | struct xfs_mount *mp; | |
1833 | ||
50f83009 | 1834 | mp = kmem_alloc(sizeof(struct xfs_mount), KM_ZERO); |
73e5fff9 IK |
1835 | if (!mp) |
1836 | return -ENOMEM; | |
1837 | ||
50f83009 IK |
1838 | spin_lock_init(&mp->m_sb_lock); |
1839 | spin_lock_init(&mp->m_agirotor_lock); | |
1840 | INIT_RADIX_TREE(&mp->m_perag_tree, GFP_ATOMIC); | |
1841 | spin_lock_init(&mp->m_perag_lock); | |
1842 | mutex_init(&mp->m_growlock); | |
f0f7a674 | 1843 | INIT_WORK(&mp->m_flush_inodes_work, xfs_flush_inodes_worker); |
50f83009 | 1844 | INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker); |
9669f51d | 1845 | INIT_DELAYED_WORK(&mp->m_blockgc_work, xfs_blockgc_worker); |
50f83009 IK |
1846 | mp->m_kobj.kobject.kset = xfs_kset; |
1847 | /* | |
1848 | * We don't create the finobt per-ag space reservation until after log | |
1849 | * recovery, so we must set this to true so that an ifree transaction | |
1850 | * started during log recovery will not depend on space reservations | |
1851 | * for finobt expansion. | |
1852 | */ | |
1853 | mp->m_finobt_nores = true; | |
1854 | ||
73e5fff9 IK |
1855 | /* |
1856 | * These can be overridden by the mount option parsing. | |
1857 | */ | |
1858 | mp->m_logbufs = -1; | |
1859 | mp->m_logbsize = -1; | |
1860 | mp->m_allocsize_log = 16; /* 64k */ | |
1861 | ||
1862 | /* | |
1863 | * Copy binary VFS mount flags we are interested in. | |
1864 | */ | |
1865 | if (fc->sb_flags & SB_RDONLY) | |
1866 | mp->m_flags |= XFS_MOUNT_RDONLY; | |
1867 | if (fc->sb_flags & SB_DIRSYNC) | |
1868 | mp->m_flags |= XFS_MOUNT_DIRSYNC; | |
1869 | if (fc->sb_flags & SB_SYNCHRONOUS) | |
1870 | mp->m_flags |= XFS_MOUNT_WSYNC; | |
1871 | ||
1872 | fc->s_fs_info = mp; | |
1873 | fc->ops = &xfs_context_ops; | |
1874 | ||
1875 | return 0; | |
1876 | } | |
1877 | ||
5085b607 | 1878 | static struct file_system_type xfs_fs_type = { |
1da177e4 LT |
1879 | .owner = THIS_MODULE, |
1880 | .name = "xfs", | |
73e5fff9 | 1881 | .init_fs_context = xfs_init_fs_context, |
d7167b14 | 1882 | .parameters = xfs_fs_parameters, |
1da177e4 LT |
1883 | .kill_sb = kill_block_super, |
1884 | .fs_flags = FS_REQUIRES_DEV, | |
1885 | }; | |
7f78e035 | 1886 | MODULE_ALIAS_FS("xfs"); |
1da177e4 | 1887 | |
9f8868ff CH |
1888 | STATIC int __init |
1889 | xfs_init_zones(void) | |
1890 | { | |
b1231760 CM |
1891 | xfs_log_ticket_zone = kmem_cache_create("xfs_log_ticket", |
1892 | sizeof(struct xlog_ticket), | |
1893 | 0, 0, NULL); | |
9f8868ff | 1894 | if (!xfs_log_ticket_zone) |
598ecfba | 1895 | goto out; |
9f8868ff | 1896 | |
b1231760 CM |
1897 | xfs_bmap_free_item_zone = kmem_cache_create("xfs_bmap_free_item", |
1898 | sizeof(struct xfs_extent_free_item), | |
1899 | 0, 0, NULL); | |
9f8868ff CH |
1900 | if (!xfs_bmap_free_item_zone) |
1901 | goto out_destroy_log_ticket_zone; | |
bf904248 | 1902 | |
b1231760 CM |
1903 | xfs_btree_cur_zone = kmem_cache_create("xfs_btree_cur", |
1904 | sizeof(struct xfs_btree_cur), | |
1905 | 0, 0, NULL); | |
9f8868ff CH |
1906 | if (!xfs_btree_cur_zone) |
1907 | goto out_destroy_bmap_free_item_zone; | |
1908 | ||
b1231760 CM |
1909 | xfs_da_state_zone = kmem_cache_create("xfs_da_state", |
1910 | sizeof(struct xfs_da_state), | |
1911 | 0, 0, NULL); | |
9f8868ff CH |
1912 | if (!xfs_da_state_zone) |
1913 | goto out_destroy_btree_cur_zone; | |
1914 | ||
b1231760 CM |
1915 | xfs_ifork_zone = kmem_cache_create("xfs_ifork", |
1916 | sizeof(struct xfs_ifork), | |
1917 | 0, 0, NULL); | |
9f8868ff | 1918 | if (!xfs_ifork_zone) |
1d9025e5 | 1919 | goto out_destroy_da_state_zone; |
9f8868ff | 1920 | |
b1231760 CM |
1921 | xfs_trans_zone = kmem_cache_create("xf_trans", |
1922 | sizeof(struct xfs_trans), | |
1923 | 0, 0, NULL); | |
9f8868ff CH |
1924 | if (!xfs_trans_zone) |
1925 | goto out_destroy_ifork_zone; | |
1926 | ||
e98c414f | 1927 | |
9f8868ff CH |
1928 | /* |
1929 | * The size of the zone allocated buf log item is the maximum | |
1930 | * size possible under XFS. This wastes a little bit of memory, | |
1931 | * but it is much faster. | |
1932 | */ | |
b1231760 CM |
1933 | xfs_buf_item_zone = kmem_cache_create("xfs_buf_item", |
1934 | sizeof(struct xfs_buf_log_item), | |
1935 | 0, 0, NULL); | |
9f8868ff | 1936 | if (!xfs_buf_item_zone) |
e6631f85 | 1937 | goto out_destroy_trans_zone; |
9f8868ff | 1938 | |
b1231760 CM |
1939 | xfs_efd_zone = kmem_cache_create("xfs_efd_item", |
1940 | (sizeof(struct xfs_efd_log_item) + | |
1941 | (XFS_EFD_MAX_FAST_EXTENTS - 1) * | |
1942 | sizeof(struct xfs_extent)), | |
1943 | 0, 0, NULL); | |
9f8868ff CH |
1944 | if (!xfs_efd_zone) |
1945 | goto out_destroy_buf_item_zone; | |
1946 | ||
b1231760 CM |
1947 | xfs_efi_zone = kmem_cache_create("xfs_efi_item", |
1948 | (sizeof(struct xfs_efi_log_item) + | |
1949 | (XFS_EFI_MAX_FAST_EXTENTS - 1) * | |
1950 | sizeof(struct xfs_extent)), | |
1951 | 0, 0, NULL); | |
9f8868ff CH |
1952 | if (!xfs_efi_zone) |
1953 | goto out_destroy_efd_zone; | |
1954 | ||
b1231760 CM |
1955 | xfs_inode_zone = kmem_cache_create("xfs_inode", |
1956 | sizeof(struct xfs_inode), 0, | |
1957 | (SLAB_HWCACHE_ALIGN | | |
1958 | SLAB_RECLAIM_ACCOUNT | | |
1959 | SLAB_MEM_SPREAD | SLAB_ACCOUNT), | |
1960 | xfs_fs_inode_init_once); | |
9f8868ff CH |
1961 | if (!xfs_inode_zone) |
1962 | goto out_destroy_efi_zone; | |
1963 | ||
b1231760 CM |
1964 | xfs_ili_zone = kmem_cache_create("xfs_ili", |
1965 | sizeof(struct xfs_inode_log_item), 0, | |
d59eadae DC |
1966 | SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, |
1967 | NULL); | |
9f8868ff CH |
1968 | if (!xfs_ili_zone) |
1969 | goto out_destroy_inode_zone; | |
b1231760 CM |
1970 | |
1971 | xfs_icreate_zone = kmem_cache_create("xfs_icr", | |
1972 | sizeof(struct xfs_icreate_item), | |
1973 | 0, 0, NULL); | |
3ebe7d2d DC |
1974 | if (!xfs_icreate_zone) |
1975 | goto out_destroy_ili_zone; | |
9f8868ff | 1976 | |
b1231760 CM |
1977 | xfs_rud_zone = kmem_cache_create("xfs_rud_item", |
1978 | sizeof(struct xfs_rud_log_item), | |
1979 | 0, 0, NULL); | |
5880f2d7 DW |
1980 | if (!xfs_rud_zone) |
1981 | goto out_destroy_icreate_zone; | |
1982 | ||
b1231760 | 1983 | xfs_rui_zone = kmem_cache_create("xfs_rui_item", |
cd00158c | 1984 | xfs_rui_log_item_sizeof(XFS_RUI_MAX_FAST_EXTENTS), |
b1231760 | 1985 | 0, 0, NULL); |
5880f2d7 DW |
1986 | if (!xfs_rui_zone) |
1987 | goto out_destroy_rud_zone; | |
1988 | ||
b1231760 CM |
1989 | xfs_cud_zone = kmem_cache_create("xfs_cud_item", |
1990 | sizeof(struct xfs_cud_log_item), | |
1991 | 0, 0, NULL); | |
baf4bcac DW |
1992 | if (!xfs_cud_zone) |
1993 | goto out_destroy_rui_zone; | |
1994 | ||
b1231760 | 1995 | xfs_cui_zone = kmem_cache_create("xfs_cui_item", |
baf4bcac | 1996 | xfs_cui_log_item_sizeof(XFS_CUI_MAX_FAST_EXTENTS), |
b1231760 | 1997 | 0, 0, NULL); |
baf4bcac DW |
1998 | if (!xfs_cui_zone) |
1999 | goto out_destroy_cud_zone; | |
2000 | ||
b1231760 CM |
2001 | xfs_bud_zone = kmem_cache_create("xfs_bud_item", |
2002 | sizeof(struct xfs_bud_log_item), | |
2003 | 0, 0, NULL); | |
6413a014 DW |
2004 | if (!xfs_bud_zone) |
2005 | goto out_destroy_cui_zone; | |
2006 | ||
b1231760 | 2007 | xfs_bui_zone = kmem_cache_create("xfs_bui_item", |
6413a014 | 2008 | xfs_bui_log_item_sizeof(XFS_BUI_MAX_FAST_EXTENTS), |
b1231760 | 2009 | 0, 0, NULL); |
6413a014 DW |
2010 | if (!xfs_bui_zone) |
2011 | goto out_destroy_bud_zone; | |
2012 | ||
9f8868ff CH |
2013 | return 0; |
2014 | ||
6413a014 | 2015 | out_destroy_bud_zone: |
aaf54eb8 | 2016 | kmem_cache_destroy(xfs_bud_zone); |
6413a014 | 2017 | out_destroy_cui_zone: |
aaf54eb8 | 2018 | kmem_cache_destroy(xfs_cui_zone); |
baf4bcac | 2019 | out_destroy_cud_zone: |
aaf54eb8 | 2020 | kmem_cache_destroy(xfs_cud_zone); |
baf4bcac | 2021 | out_destroy_rui_zone: |
aaf54eb8 | 2022 | kmem_cache_destroy(xfs_rui_zone); |
5880f2d7 | 2023 | out_destroy_rud_zone: |
aaf54eb8 | 2024 | kmem_cache_destroy(xfs_rud_zone); |
5880f2d7 | 2025 | out_destroy_icreate_zone: |
aaf54eb8 | 2026 | kmem_cache_destroy(xfs_icreate_zone); |
3ebe7d2d | 2027 | out_destroy_ili_zone: |
aaf54eb8 | 2028 | kmem_cache_destroy(xfs_ili_zone); |
9f8868ff | 2029 | out_destroy_inode_zone: |
aaf54eb8 | 2030 | kmem_cache_destroy(xfs_inode_zone); |
9f8868ff | 2031 | out_destroy_efi_zone: |
aaf54eb8 | 2032 | kmem_cache_destroy(xfs_efi_zone); |
9f8868ff | 2033 | out_destroy_efd_zone: |
aaf54eb8 | 2034 | kmem_cache_destroy(xfs_efd_zone); |
9f8868ff | 2035 | out_destroy_buf_item_zone: |
aaf54eb8 | 2036 | kmem_cache_destroy(xfs_buf_item_zone); |
9f8868ff | 2037 | out_destroy_trans_zone: |
aaf54eb8 | 2038 | kmem_cache_destroy(xfs_trans_zone); |
9f8868ff | 2039 | out_destroy_ifork_zone: |
aaf54eb8 | 2040 | kmem_cache_destroy(xfs_ifork_zone); |
9f8868ff | 2041 | out_destroy_da_state_zone: |
aaf54eb8 | 2042 | kmem_cache_destroy(xfs_da_state_zone); |
9f8868ff | 2043 | out_destroy_btree_cur_zone: |
aaf54eb8 | 2044 | kmem_cache_destroy(xfs_btree_cur_zone); |
9f8868ff | 2045 | out_destroy_bmap_free_item_zone: |
aaf54eb8 | 2046 | kmem_cache_destroy(xfs_bmap_free_item_zone); |
9f8868ff | 2047 | out_destroy_log_ticket_zone: |
aaf54eb8 | 2048 | kmem_cache_destroy(xfs_log_ticket_zone); |
9f8868ff CH |
2049 | out: |
2050 | return -ENOMEM; | |
2051 | } | |
2052 | ||
2053 | STATIC void | |
2054 | xfs_destroy_zones(void) | |
2055 | { | |
8c0a8537 KS |
2056 | /* |
2057 | * Make sure all delayed rcu free are flushed before we | |
2058 | * destroy caches. | |
2059 | */ | |
2060 | rcu_barrier(); | |
aaf54eb8 CM |
2061 | kmem_cache_destroy(xfs_bui_zone); |
2062 | kmem_cache_destroy(xfs_bud_zone); | |
2063 | kmem_cache_destroy(xfs_cui_zone); | |
2064 | kmem_cache_destroy(xfs_cud_zone); | |
2065 | kmem_cache_destroy(xfs_rui_zone); | |
2066 | kmem_cache_destroy(xfs_rud_zone); | |
2067 | kmem_cache_destroy(xfs_icreate_zone); | |
2068 | kmem_cache_destroy(xfs_ili_zone); | |
2069 | kmem_cache_destroy(xfs_inode_zone); | |
2070 | kmem_cache_destroy(xfs_efi_zone); | |
2071 | kmem_cache_destroy(xfs_efd_zone); | |
2072 | kmem_cache_destroy(xfs_buf_item_zone); | |
2073 | kmem_cache_destroy(xfs_trans_zone); | |
2074 | kmem_cache_destroy(xfs_ifork_zone); | |
2075 | kmem_cache_destroy(xfs_da_state_zone); | |
2076 | kmem_cache_destroy(xfs_btree_cur_zone); | |
2077 | kmem_cache_destroy(xfs_bmap_free_item_zone); | |
2078 | kmem_cache_destroy(xfs_log_ticket_zone); | |
9f8868ff | 2079 | } |
1da177e4 | 2080 | |
0bf6a5bd DC |
2081 | STATIC int __init |
2082 | xfs_init_workqueues(void) | |
2083 | { | |
c999a223 DC |
2084 | /* |
2085 | * The allocation workqueue can be used in memory reclaim situations | |
2086 | * (writepage path), and parallelism is only limited by the number of | |
2087 | * AGs in all the filesystems mounted. Hence use the default large | |
2088 | * max_active value for this workqueue. | |
2089 | */ | |
8018ec08 | 2090 | xfs_alloc_wq = alloc_workqueue("xfsalloc", |
05a302a1 | 2091 | XFS_WQFLAGS(WQ_MEM_RECLAIM | WQ_FREEZABLE), 0); |
c999a223 | 2092 | if (!xfs_alloc_wq) |
5889608d | 2093 | return -ENOMEM; |
c999a223 | 2094 | |
05a302a1 DW |
2095 | xfs_discard_wq = alloc_workqueue("xfsdiscard", XFS_WQFLAGS(WQ_UNBOUND), |
2096 | 0); | |
4560e78f CH |
2097 | if (!xfs_discard_wq) |
2098 | goto out_free_alloc_wq; | |
2099 | ||
0bf6a5bd | 2100 | return 0; |
4560e78f CH |
2101 | out_free_alloc_wq: |
2102 | destroy_workqueue(xfs_alloc_wq); | |
2103 | return -ENOMEM; | |
0bf6a5bd DC |
2104 | } |
2105 | ||
39411f81 | 2106 | STATIC void |
0bf6a5bd DC |
2107 | xfs_destroy_workqueues(void) |
2108 | { | |
4560e78f | 2109 | destroy_workqueue(xfs_discard_wq); |
c999a223 | 2110 | destroy_workqueue(xfs_alloc_wq); |
0bf6a5bd DC |
2111 | } |
2112 | ||
1da177e4 | 2113 | STATIC int __init |
9f8868ff | 2114 | init_xfs_fs(void) |
1da177e4 LT |
2115 | { |
2116 | int error; | |
1da177e4 | 2117 | |
30cbc591 DW |
2118 | xfs_check_ondisk_structs(); |
2119 | ||
65795910 CH |
2120 | printk(KERN_INFO XFS_VERSION_STRING " with " |
2121 | XFS_BUILD_OPTIONS " enabled\n"); | |
1da177e4 | 2122 | |
9f8868ff | 2123 | xfs_dir_startup(); |
1da177e4 | 2124 | |
8758280f | 2125 | error = xfs_init_zones(); |
9f8868ff CH |
2126 | if (error) |
2127 | goto out; | |
2128 | ||
0bf6a5bd | 2129 | error = xfs_init_workqueues(); |
9f8868ff | 2130 | if (error) |
0b1b213f | 2131 | goto out_destroy_zones; |
9f8868ff | 2132 | |
0bf6a5bd DC |
2133 | error = xfs_mru_cache_init(); |
2134 | if (error) | |
2135 | goto out_destroy_wq; | |
2136 | ||
ce8e922c | 2137 | error = xfs_buf_init(); |
9f8868ff | 2138 | if (error) |
1919adda | 2139 | goto out_mru_cache_uninit; |
9f8868ff CH |
2140 | |
2141 | error = xfs_init_procfs(); | |
2142 | if (error) | |
2143 | goto out_buf_terminate; | |
2144 | ||
2145 | error = xfs_sysctl_register(); | |
2146 | if (error) | |
2147 | goto out_cleanup_procfs; | |
1da177e4 | 2148 | |
3d871226 BF |
2149 | xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj); |
2150 | if (!xfs_kset) { | |
2151 | error = -ENOMEM; | |
bb230c12 | 2152 | goto out_sysctl_unregister; |
3d871226 BF |
2153 | } |
2154 | ||
80529c45 BD |
2155 | xfsstats.xs_kobj.kobject.kset = xfs_kset; |
2156 | ||
2157 | xfsstats.xs_stats = alloc_percpu(struct xfsstats); | |
2158 | if (!xfsstats.xs_stats) { | |
2159 | error = -ENOMEM; | |
2160 | goto out_kset_unregister; | |
2161 | } | |
2162 | ||
2163 | error = xfs_sysfs_init(&xfsstats.xs_kobj, &xfs_stats_ktype, NULL, | |
bb230c12 BD |
2164 | "stats"); |
2165 | if (error) | |
80529c45 | 2166 | goto out_free_stats; |
bb230c12 | 2167 | |
65b65735 BF |
2168 | #ifdef DEBUG |
2169 | xfs_dbg_kobj.kobject.kset = xfs_kset; | |
2170 | error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug"); | |
a05931ce | 2171 | if (error) |
bb230c12 | 2172 | goto out_remove_stats_kobj; |
65b65735 BF |
2173 | #endif |
2174 | ||
2175 | error = xfs_qm_init(); | |
2176 | if (error) | |
bb230c12 | 2177 | goto out_remove_dbg_kobj; |
1da177e4 LT |
2178 | |
2179 | error = register_filesystem(&xfs_fs_type); | |
2180 | if (error) | |
a05931ce | 2181 | goto out_qm_exit; |
1da177e4 LT |
2182 | return 0; |
2183 | ||
a05931ce CH |
2184 | out_qm_exit: |
2185 | xfs_qm_exit(); | |
bb230c12 | 2186 | out_remove_dbg_kobj: |
65b65735 BF |
2187 | #ifdef DEBUG |
2188 | xfs_sysfs_del(&xfs_dbg_kobj); | |
bb230c12 | 2189 | out_remove_stats_kobj: |
65b65735 | 2190 | #endif |
80529c45 BD |
2191 | xfs_sysfs_del(&xfsstats.xs_kobj); |
2192 | out_free_stats: | |
2193 | free_percpu(xfsstats.xs_stats); | |
bb230c12 | 2194 | out_kset_unregister: |
3d871226 | 2195 | kset_unregister(xfs_kset); |
9f8868ff CH |
2196 | out_sysctl_unregister: |
2197 | xfs_sysctl_unregister(); | |
2198 | out_cleanup_procfs: | |
2199 | xfs_cleanup_procfs(); | |
2200 | out_buf_terminate: | |
ce8e922c | 2201 | xfs_buf_terminate(); |
9f8868ff CH |
2202 | out_mru_cache_uninit: |
2203 | xfs_mru_cache_uninit(); | |
0bf6a5bd DC |
2204 | out_destroy_wq: |
2205 | xfs_destroy_workqueues(); | |
9f8868ff | 2206 | out_destroy_zones: |
8758280f | 2207 | xfs_destroy_zones(); |
9f8868ff | 2208 | out: |
1da177e4 LT |
2209 | return error; |
2210 | } | |
2211 | ||
2212 | STATIC void __exit | |
9f8868ff | 2213 | exit_xfs_fs(void) |
1da177e4 | 2214 | { |
a05931ce | 2215 | xfs_qm_exit(); |
1da177e4 | 2216 | unregister_filesystem(&xfs_fs_type); |
65b65735 BF |
2217 | #ifdef DEBUG |
2218 | xfs_sysfs_del(&xfs_dbg_kobj); | |
2219 | #endif | |
80529c45 BD |
2220 | xfs_sysfs_del(&xfsstats.xs_kobj); |
2221 | free_percpu(xfsstats.xs_stats); | |
3d871226 | 2222 | kset_unregister(xfs_kset); |
9f8868ff CH |
2223 | xfs_sysctl_unregister(); |
2224 | xfs_cleanup_procfs(); | |
ce8e922c | 2225 | xfs_buf_terminate(); |
9f8868ff | 2226 | xfs_mru_cache_uninit(); |
0bf6a5bd | 2227 | xfs_destroy_workqueues(); |
8758280f | 2228 | xfs_destroy_zones(); |
af3b6382 | 2229 | xfs_uuid_table_free(); |
1da177e4 LT |
2230 | } |
2231 | ||
2232 | module_init(init_xfs_fs); | |
2233 | module_exit(exit_xfs_fs); | |
2234 | ||
2235 | MODULE_AUTHOR("Silicon Graphics, Inc."); | |
2236 | MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled"); | |
2237 | MODULE_LICENSE("GPL"); |