Commit | Line | Data |
---|---|---|
0b61f8a4 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 | 2 | /* |
7b718769 NS |
3 | * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc. |
4 | * All Rights Reserved. | |
1da177e4 | 5 | */ |
1da177e4 | 6 | #include "xfs.h" |
a844f451 | 7 | #include "xfs_fs.h" |
5467b34b | 8 | #include "xfs_shared.h" |
a4fbe6ab | 9 | #include "xfs_format.h" |
239880ef DC |
10 | #include "xfs_log_format.h" |
11 | #include "xfs_trans_resv.h" | |
1da177e4 | 12 | #include "xfs_mount.h" |
1da177e4 | 13 | #include "xfs_inode.h" |
239880ef | 14 | #include "xfs_trans.h" |
a844f451 | 15 | #include "xfs_inode_item.h" |
0b1b213f | 16 | #include "xfs_trace.h" |
239880ef | 17 | #include "xfs_trans_priv.h" |
d3a304b6 | 18 | #include "xfs_buf_item.h" |
1234351c | 19 | #include "xfs_log.h" |
01728b44 | 20 | #include "xfs_log_priv.h" |
a5155b87 | 21 | #include "xfs_error.h" |
1da177e4 | 22 | |
f0e28280 | 23 | #include <linux/iversion.h> |
1da177e4 | 24 | |
182696fb | 25 | struct kmem_cache *xfs_ili_cache; /* inode log item */ |
1da177e4 | 26 | |
7bfa31d8 CH |
27 | static inline struct xfs_inode_log_item *INODE_ITEM(struct xfs_log_item *lip) |
28 | { | |
29 | return container_of(lip, struct xfs_inode_log_item, ili_item); | |
30 | } | |
31 | ||
19f4e7cc DC |
32 | /* |
33 | * The logged size of an inode fork is always the current size of the inode | |
34 | * fork. This means that when an inode fork is relogged, the size of the logged | |
35 | * region is determined by the current state, not the combination of the | |
36 | * previously logged state + the current state. This is different relogging | |
37 | * behaviour to most other log items which will retain the size of the | |
38 | * previously logged changes when smaller regions are relogged. | |
39 | * | |
40 | * Hence operations that remove data from the inode fork (e.g. shortform | |
41 | * dir/attr remove, extent form extent removal, etc), the size of the relogged | |
42 | * inode gets -smaller- rather than stays the same size as the previously logged | |
43 | * size and this can result in the committing transaction reducing the amount of | |
44 | * space being consumed by the CIL. | |
45 | */ | |
166d1368 | 46 | STATIC void |
ce9641d6 CH |
47 | xfs_inode_item_data_fork_size( |
48 | struct xfs_inode_log_item *iip, | |
166d1368 DC |
49 | int *nvecs, |
50 | int *nbytes) | |
1da177e4 | 51 | { |
7bfa31d8 | 52 | struct xfs_inode *ip = iip->ili_inode; |
166d1368 | 53 | |
f7e67b20 | 54 | switch (ip->i_df.if_format) { |
1da177e4 | 55 | case XFS_DINODE_FMT_EXTENTS: |
f5d8d5c4 | 56 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
daf83964 | 57 | ip->i_df.if_nextents > 0 && |
166d1368 DC |
58 | ip->i_df.if_bytes > 0) { |
59 | /* worst case, doesn't subtract delalloc extents */ | |
60 | *nbytes += XFS_IFORK_DSIZE(ip); | |
61 | *nvecs += 1; | |
62 | } | |
1da177e4 | 63 | break; |
1da177e4 | 64 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 65 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
166d1368 DC |
66 | ip->i_df.if_broot_bytes > 0) { |
67 | *nbytes += ip->i_df.if_broot_bytes; | |
68 | *nvecs += 1; | |
69 | } | |
1da177e4 | 70 | break; |
1da177e4 | 71 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 72 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
166d1368 | 73 | ip->i_df.if_bytes > 0) { |
b2c28035 | 74 | *nbytes += xlog_calc_iovec_len(ip->i_df.if_bytes); |
166d1368 DC |
75 | *nvecs += 1; |
76 | } | |
1da177e4 LT |
77 | break; |
78 | ||
79 | case XFS_DINODE_FMT_DEV: | |
1da177e4 | 80 | break; |
1da177e4 LT |
81 | default: |
82 | ASSERT(0); | |
83 | break; | |
84 | } | |
ce9641d6 | 85 | } |
1da177e4 | 86 | |
ce9641d6 CH |
87 | STATIC void |
88 | xfs_inode_item_attr_fork_size( | |
89 | struct xfs_inode_log_item *iip, | |
90 | int *nvecs, | |
91 | int *nbytes) | |
92 | { | |
93 | struct xfs_inode *ip = iip->ili_inode; | |
1da177e4 | 94 | |
2ed5b09b | 95 | switch (ip->i_af.if_format) { |
1da177e4 | 96 | case XFS_DINODE_FMT_EXTENTS: |
f5d8d5c4 | 97 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
2ed5b09b DW |
98 | ip->i_af.if_nextents > 0 && |
99 | ip->i_af.if_bytes > 0) { | |
166d1368 DC |
100 | /* worst case, doesn't subtract unused space */ |
101 | *nbytes += XFS_IFORK_ASIZE(ip); | |
102 | *nvecs += 1; | |
103 | } | |
1da177e4 | 104 | break; |
1da177e4 | 105 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 106 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
2ed5b09b DW |
107 | ip->i_af.if_broot_bytes > 0) { |
108 | *nbytes += ip->i_af.if_broot_bytes; | |
166d1368 DC |
109 | *nvecs += 1; |
110 | } | |
1da177e4 | 111 | break; |
1da177e4 | 112 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 113 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
2ed5b09b DW |
114 | ip->i_af.if_bytes > 0) { |
115 | *nbytes += xlog_calc_iovec_len(ip->i_af.if_bytes); | |
166d1368 DC |
116 | *nvecs += 1; |
117 | } | |
1da177e4 | 118 | break; |
1da177e4 LT |
119 | default: |
120 | ASSERT(0); | |
121 | break; | |
122 | } | |
1da177e4 LT |
123 | } |
124 | ||
ce9641d6 CH |
125 | /* |
126 | * This returns the number of iovecs needed to log the given inode item. | |
127 | * | |
128 | * We need one iovec for the inode log format structure, one for the | |
129 | * inode core, and possibly one for the inode data/extents/b-tree root | |
130 | * and one for the inode attribute data/extents/b-tree root. | |
131 | */ | |
132 | STATIC void | |
133 | xfs_inode_item_size( | |
134 | struct xfs_log_item *lip, | |
135 | int *nvecs, | |
136 | int *nbytes) | |
137 | { | |
138 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
139 | struct xfs_inode *ip = iip->ili_inode; | |
140 | ||
141 | *nvecs += 2; | |
142 | *nbytes += sizeof(struct xfs_inode_log_format) + | |
e9e2eae8 | 143 | xfs_log_dinode_size(ip->i_mount); |
ce9641d6 CH |
144 | |
145 | xfs_inode_item_data_fork_size(iip, nvecs, nbytes); | |
932b42c6 | 146 | if (xfs_inode_has_attr_fork(ip)) |
ce9641d6 CH |
147 | xfs_inode_item_attr_fork_size(iip, nvecs, nbytes); |
148 | } | |
149 | ||
1234351c | 150 | STATIC void |
3de559fb CH |
151 | xfs_inode_item_format_data_fork( |
152 | struct xfs_inode_log_item *iip, | |
bde7cff6 CH |
153 | struct xfs_inode_log_format *ilf, |
154 | struct xfs_log_vec *lv, | |
155 | struct xfs_log_iovec **vecp) | |
1da177e4 | 156 | { |
7bfa31d8 | 157 | struct xfs_inode *ip = iip->ili_inode; |
1da177e4 | 158 | size_t data_bytes; |
1da177e4 | 159 | |
f7e67b20 | 160 | switch (ip->i_df.if_format) { |
1da177e4 | 161 | case XFS_DINODE_FMT_EXTENTS: |
f5d8d5c4 | 162 | iip->ili_fields &= |
42b67dc6 | 163 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | XFS_ILOG_DEV); |
339a5f5d | 164 | |
f5d8d5c4 | 165 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
daf83964 | 166 | ip->i_df.if_nextents > 0 && |
339a5f5d | 167 | ip->i_df.if_bytes > 0) { |
da776503 CH |
168 | struct xfs_bmbt_rec *p; |
169 | ||
5d829300 | 170 | ASSERT(xfs_iext_count(&ip->i_df) > 0); |
da776503 CH |
171 | |
172 | p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IEXT); | |
173 | data_bytes = xfs_iextents_copy(ip, p, XFS_DATA_FORK); | |
174 | xlog_finish_iovec(lv, *vecp, data_bytes); | |
175 | ||
176 | ASSERT(data_bytes <= ip->i_df.if_bytes); | |
177 | ||
178 | ilf->ilf_dsize = data_bytes; | |
bde7cff6 | 179 | ilf->ilf_size++; |
339a5f5d | 180 | } else { |
f5d8d5c4 | 181 | iip->ili_fields &= ~XFS_ILOG_DEXT; |
1da177e4 LT |
182 | } |
183 | break; | |
1da177e4 | 184 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 185 | iip->ili_fields &= |
42b67dc6 | 186 | ~(XFS_ILOG_DDATA | XFS_ILOG_DEXT | XFS_ILOG_DEV); |
339a5f5d | 187 | |
f5d8d5c4 | 188 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
339a5f5d | 189 | ip->i_df.if_broot_bytes > 0) { |
1da177e4 | 190 | ASSERT(ip->i_df.if_broot != NULL); |
bde7cff6 | 191 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IBROOT, |
1234351c CH |
192 | ip->i_df.if_broot, |
193 | ip->i_df.if_broot_bytes); | |
bde7cff6 CH |
194 | ilf->ilf_dsize = ip->i_df.if_broot_bytes; |
195 | ilf->ilf_size++; | |
339a5f5d | 196 | } else { |
f5d8d5c4 | 197 | ASSERT(!(iip->ili_fields & |
339a5f5d | 198 | XFS_ILOG_DBROOT)); |
f5d8d5c4 | 199 | iip->ili_fields &= ~XFS_ILOG_DBROOT; |
1da177e4 LT |
200 | } |
201 | break; | |
1da177e4 | 202 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 203 | iip->ili_fields &= |
42b67dc6 | 204 | ~(XFS_ILOG_DEXT | XFS_ILOG_DBROOT | XFS_ILOG_DEV); |
f5d8d5c4 | 205 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
339a5f5d | 206 | ip->i_df.if_bytes > 0) { |
1234351c | 207 | ASSERT(ip->i_df.if_u1.if_data != NULL); |
13d2c10b | 208 | ASSERT(ip->i_disk_size > 0); |
bde7cff6 | 209 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_ILOCAL, |
b2c28035 DC |
210 | ip->i_df.if_u1.if_data, |
211 | ip->i_df.if_bytes); | |
212 | ilf->ilf_dsize = (unsigned)ip->i_df.if_bytes; | |
bde7cff6 | 213 | ilf->ilf_size++; |
339a5f5d | 214 | } else { |
f5d8d5c4 | 215 | iip->ili_fields &= ~XFS_ILOG_DDATA; |
1da177e4 LT |
216 | } |
217 | break; | |
1da177e4 | 218 | case XFS_DINODE_FMT_DEV: |
f5d8d5c4 | 219 | iip->ili_fields &= |
42b67dc6 | 220 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | XFS_ILOG_DEXT); |
bde7cff6 | 221 | if (iip->ili_fields & XFS_ILOG_DEV) |
66f36464 | 222 | ilf->ilf_u.ilfu_rdev = sysv_encode_dev(VFS_I(ip)->i_rdev); |
1da177e4 | 223 | break; |
1da177e4 LT |
224 | default: |
225 | ASSERT(0); | |
226 | break; | |
227 | } | |
3de559fb CH |
228 | } |
229 | ||
1234351c | 230 | STATIC void |
3de559fb CH |
231 | xfs_inode_item_format_attr_fork( |
232 | struct xfs_inode_log_item *iip, | |
bde7cff6 CH |
233 | struct xfs_inode_log_format *ilf, |
234 | struct xfs_log_vec *lv, | |
235 | struct xfs_log_iovec **vecp) | |
3de559fb CH |
236 | { |
237 | struct xfs_inode *ip = iip->ili_inode; | |
238 | size_t data_bytes; | |
1da177e4 | 239 | |
2ed5b09b | 240 | switch (ip->i_af.if_format) { |
1da177e4 | 241 | case XFS_DINODE_FMT_EXTENTS: |
f5d8d5c4 | 242 | iip->ili_fields &= |
339a5f5d CH |
243 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT); |
244 | ||
f5d8d5c4 | 245 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
2ed5b09b DW |
246 | ip->i_af.if_nextents > 0 && |
247 | ip->i_af.if_bytes > 0) { | |
da776503 CH |
248 | struct xfs_bmbt_rec *p; |
249 | ||
2ed5b09b DW |
250 | ASSERT(xfs_iext_count(&ip->i_af) == |
251 | ip->i_af.if_nextents); | |
da776503 CH |
252 | |
253 | p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_EXT); | |
254 | data_bytes = xfs_iextents_copy(ip, p, XFS_ATTR_FORK); | |
255 | xlog_finish_iovec(lv, *vecp, data_bytes); | |
256 | ||
257 | ilf->ilf_asize = data_bytes; | |
bde7cff6 | 258 | ilf->ilf_size++; |
339a5f5d | 259 | } else { |
f5d8d5c4 | 260 | iip->ili_fields &= ~XFS_ILOG_AEXT; |
1da177e4 LT |
261 | } |
262 | break; | |
1da177e4 | 263 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 264 | iip->ili_fields &= |
339a5f5d CH |
265 | ~(XFS_ILOG_ADATA | XFS_ILOG_AEXT); |
266 | ||
f5d8d5c4 | 267 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
2ed5b09b DW |
268 | ip->i_af.if_broot_bytes > 0) { |
269 | ASSERT(ip->i_af.if_broot != NULL); | |
339a5f5d | 270 | |
bde7cff6 | 271 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_BROOT, |
2ed5b09b DW |
272 | ip->i_af.if_broot, |
273 | ip->i_af.if_broot_bytes); | |
274 | ilf->ilf_asize = ip->i_af.if_broot_bytes; | |
bde7cff6 | 275 | ilf->ilf_size++; |
339a5f5d | 276 | } else { |
f5d8d5c4 | 277 | iip->ili_fields &= ~XFS_ILOG_ABROOT; |
1da177e4 LT |
278 | } |
279 | break; | |
1da177e4 | 280 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 281 | iip->ili_fields &= |
339a5f5d CH |
282 | ~(XFS_ILOG_AEXT | XFS_ILOG_ABROOT); |
283 | ||
f5d8d5c4 | 284 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
2ed5b09b DW |
285 | ip->i_af.if_bytes > 0) { |
286 | ASSERT(ip->i_af.if_u1.if_data != NULL); | |
bde7cff6 | 287 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_LOCAL, |
2ed5b09b DW |
288 | ip->i_af.if_u1.if_data, |
289 | ip->i_af.if_bytes); | |
290 | ilf->ilf_asize = (unsigned)ip->i_af.if_bytes; | |
bde7cff6 | 291 | ilf->ilf_size++; |
339a5f5d | 292 | } else { |
f5d8d5c4 | 293 | iip->ili_fields &= ~XFS_ILOG_ADATA; |
1da177e4 LT |
294 | } |
295 | break; | |
1da177e4 LT |
296 | default: |
297 | ASSERT(0); | |
298 | break; | |
299 | } | |
3de559fb CH |
300 | } |
301 | ||
30e05599 DW |
302 | /* |
303 | * Convert an incore timestamp to a log timestamp. Note that the log format | |
304 | * specifies host endian format! | |
305 | */ | |
6fc277c7 | 306 | static inline xfs_log_timestamp_t |
30e05599 | 307 | xfs_inode_to_log_dinode_ts( |
f93e5436 | 308 | struct xfs_inode *ip, |
30e05599 DW |
309 | const struct timespec64 tv) |
310 | { | |
732de7db | 311 | struct xfs_log_legacy_timestamp *lits; |
6fc277c7 | 312 | xfs_log_timestamp_t its; |
30e05599 | 313 | |
f93e5436 DW |
314 | if (xfs_inode_has_bigtime(ip)) |
315 | return xfs_inode_encode_bigtime(tv); | |
316 | ||
732de7db | 317 | lits = (struct xfs_log_legacy_timestamp *)&its; |
30e05599 DW |
318 | lits->t_sec = tv.tv_sec; |
319 | lits->t_nsec = tv.tv_nsec; | |
320 | ||
321 | return its; | |
322 | } | |
323 | ||
9b3beb02 CH |
324 | /* |
325 | * The legacy DMAPI fields are only present in the on-disk and in-log inodes, | |
326 | * but not in the in-memory one. But we are guaranteed to have an inode buffer | |
327 | * in memory when logging an inode, so we can just copy it from the on-disk | |
328 | * inode to the in-log inode here so that recovery of file system with these | |
329 | * fields set to non-zero values doesn't lose them. For all other cases we zero | |
330 | * the fields. | |
331 | */ | |
332 | static void | |
333 | xfs_copy_dm_fields_to_log_dinode( | |
334 | struct xfs_inode *ip, | |
335 | struct xfs_log_dinode *to) | |
336 | { | |
337 | struct xfs_dinode *dip; | |
338 | ||
339 | dip = xfs_buf_offset(ip->i_itemp->ili_item.li_buf, | |
340 | ip->i_imap.im_boffset); | |
341 | ||
342 | if (xfs_iflags_test(ip, XFS_IPRESERVE_DM_FIELDS)) { | |
343 | to->di_dmevmask = be32_to_cpu(dip->di_dmevmask); | |
344 | to->di_dmstate = be16_to_cpu(dip->di_dmstate); | |
345 | } else { | |
346 | to->di_dmevmask = 0; | |
347 | to->di_dmstate = 0; | |
348 | } | |
349 | } | |
350 | ||
52a4a148 CB |
351 | static inline void |
352 | xfs_inode_to_log_dinode_iext_counters( | |
353 | struct xfs_inode *ip, | |
354 | struct xfs_log_dinode *to) | |
355 | { | |
356 | if (xfs_inode_has_large_extent_counts(ip)) { | |
357 | to->di_big_nextents = xfs_ifork_nextents(&ip->i_df); | |
2ed5b09b | 358 | to->di_big_anextents = xfs_ifork_nextents(&ip->i_af); |
52a4a148 CB |
359 | to->di_nrext64_pad = 0; |
360 | } else { | |
361 | to->di_nextents = xfs_ifork_nextents(&ip->i_df); | |
2ed5b09b | 362 | to->di_anextents = xfs_ifork_nextents(&ip->i_af); |
52a4a148 CB |
363 | } |
364 | } | |
365 | ||
f8d55aa0 | 366 | static void |
3987848c DC |
367 | xfs_inode_to_log_dinode( |
368 | struct xfs_inode *ip, | |
93f958f9 DC |
369 | struct xfs_log_dinode *to, |
370 | xfs_lsn_t lsn) | |
f8d55aa0 | 371 | { |
3987848c DC |
372 | struct inode *inode = VFS_I(ip); |
373 | ||
93f958f9 | 374 | to->di_magic = XFS_DINODE_MAGIC; |
f7e67b20 | 375 | to->di_format = xfs_ifork_format(&ip->i_df); |
ba8adad5 CH |
376 | to->di_uid = i_uid_read(inode); |
377 | to->di_gid = i_gid_read(inode); | |
ceaf603c CH |
378 | to->di_projid_lo = ip->i_projid & 0xffff; |
379 | to->di_projid_hi = ip->i_projid >> 16; | |
f8d55aa0 | 380 | |
faeb4e47 | 381 | memset(to->di_pad3, 0, sizeof(to->di_pad3)); |
f93e5436 DW |
382 | to->di_atime = xfs_inode_to_log_dinode_ts(ip, inode->i_atime); |
383 | to->di_mtime = xfs_inode_to_log_dinode_ts(ip, inode->i_mtime); | |
384 | to->di_ctime = xfs_inode_to_log_dinode_ts(ip, inode->i_ctime); | |
54d7b5c1 | 385 | to->di_nlink = inode->i_nlink; |
9e9a2674 | 386 | to->di_gen = inode->i_generation; |
c19b3b05 | 387 | to->di_mode = inode->i_mode; |
f8d55aa0 | 388 | |
13d2c10b | 389 | to->di_size = ip->i_disk_size; |
6e73a545 | 390 | to->di_nblocks = ip->i_nblocks; |
031474c2 | 391 | to->di_extsize = ip->i_extsize; |
7821ea30 | 392 | to->di_forkoff = ip->i_forkoff; |
2ed5b09b | 393 | to->di_aformat = xfs_ifork_format(&ip->i_af); |
db07349d | 394 | to->di_flags = ip->i_diflags; |
f8d55aa0 | 395 | |
9b3beb02 CH |
396 | xfs_copy_dm_fields_to_log_dinode(ip, to); |
397 | ||
20413e37 DC |
398 | /* log a dummy value to ensure log structure is fully initialised */ |
399 | to->di_next_unlinked = NULLAGINO; | |
400 | ||
38c26bfd | 401 | if (xfs_has_v3inodes(ip->i_mount)) { |
6471e9c5 | 402 | to->di_version = 3; |
f0e28280 | 403 | to->di_changecount = inode_peek_iversion(inode); |
e98d5e88 | 404 | to->di_crtime = xfs_inode_to_log_dinode_ts(ip, ip->i_crtime); |
3e09ab8f | 405 | to->di_flags2 = ip->i_diflags2; |
b33ce57d | 406 | to->di_cowextsize = ip->i_cowextsize; |
93f958f9 DC |
407 | to->di_ino = ip->i_ino; |
408 | to->di_lsn = lsn; | |
409 | memset(to->di_pad2, 0, sizeof(to->di_pad2)); | |
410 | uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid); | |
52a4a148 | 411 | to->di_v3_pad = 0; |
f8d55aa0 | 412 | } else { |
6471e9c5 | 413 | to->di_version = 2; |
965e0a1a | 414 | to->di_flushiter = ip->i_flushiter; |
52a4a148 | 415 | memset(to->di_v2_pad, 0, sizeof(to->di_v2_pad)); |
f8d55aa0 | 416 | } |
52a4a148 CB |
417 | |
418 | xfs_inode_to_log_dinode_iext_counters(ip, to); | |
f8d55aa0 DC |
419 | } |
420 | ||
421 | /* | |
422 | * Format the inode core. Current timestamp data is only in the VFS inode | |
423 | * fields, so we need to grab them from there. Hence rather than just copying | |
424 | * the XFS inode core structure, format the fields directly into the iovec. | |
425 | */ | |
426 | static void | |
427 | xfs_inode_item_format_core( | |
428 | struct xfs_inode *ip, | |
429 | struct xfs_log_vec *lv, | |
430 | struct xfs_log_iovec **vecp) | |
431 | { | |
432 | struct xfs_log_dinode *dic; | |
433 | ||
434 | dic = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_ICORE); | |
93f958f9 | 435 | xfs_inode_to_log_dinode(ip, dic, ip->i_itemp->ili_item.li_lsn); |
e9e2eae8 | 436 | xlog_finish_iovec(lv, *vecp, xfs_log_dinode_size(ip->i_mount)); |
f8d55aa0 DC |
437 | } |
438 | ||
3de559fb CH |
439 | /* |
440 | * This is called to fill in the vector of log iovecs for the given inode | |
441 | * log item. It fills the first item with an inode log format structure, | |
442 | * the second with the on-disk inode structure, and a possible third and/or | |
443 | * fourth with the inode data/extents/b-tree root and inode attributes | |
444 | * data/extents/b-tree root. | |
20413e37 DC |
445 | * |
446 | * Note: Always use the 64 bit inode log format structure so we don't | |
447 | * leave an uninitialised hole in the format item on 64 bit systems. Log | |
448 | * recovery on 32 bit systems handles this just fine, so there's no reason | |
449 | * for not using an initialising the properly padded structure all the time. | |
3de559fb CH |
450 | */ |
451 | STATIC void | |
452 | xfs_inode_item_format( | |
453 | struct xfs_log_item *lip, | |
bde7cff6 | 454 | struct xfs_log_vec *lv) |
3de559fb CH |
455 | { |
456 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
457 | struct xfs_inode *ip = iip->ili_inode; | |
bde7cff6 | 458 | struct xfs_log_iovec *vecp = NULL; |
20413e37 | 459 | struct xfs_inode_log_format *ilf; |
3de559fb | 460 | |
2f251293 CH |
461 | ilf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_IFORMAT); |
462 | ilf->ilf_type = XFS_LI_INODE; | |
463 | ilf->ilf_ino = ip->i_ino; | |
464 | ilf->ilf_blkno = ip->i_imap.im_blkno; | |
465 | ilf->ilf_len = ip->i_imap.im_len; | |
466 | ilf->ilf_boffset = ip->i_imap.im_boffset; | |
467 | ilf->ilf_fields = XFS_ILOG_CORE; | |
468 | ilf->ilf_size = 2; /* format + core */ | |
20413e37 DC |
469 | |
470 | /* | |
471 | * make sure we don't leak uninitialised data into the log in the case | |
472 | * when we don't log every field in the inode. | |
473 | */ | |
474 | ilf->ilf_dsize = 0; | |
475 | ilf->ilf_asize = 0; | |
476 | ilf->ilf_pad = 0; | |
42b67dc6 | 477 | memset(&ilf->ilf_u, 0, sizeof(ilf->ilf_u)); |
20413e37 DC |
478 | |
479 | xlog_finish_iovec(lv, vecp, sizeof(*ilf)); | |
3de559fb | 480 | |
f8d55aa0 | 481 | xfs_inode_item_format_core(ip, lv, &vecp); |
bde7cff6 | 482 | xfs_inode_item_format_data_fork(iip, ilf, lv, &vecp); |
932b42c6 | 483 | if (xfs_inode_has_attr_fork(ip)) { |
bde7cff6 | 484 | xfs_inode_item_format_attr_fork(iip, ilf, lv, &vecp); |
3de559fb CH |
485 | } else { |
486 | iip->ili_fields &= | |
487 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT); | |
488 | } | |
489 | ||
2f251293 CH |
490 | /* update the format with the exact fields we actually logged */ |
491 | ilf->ilf_fields |= (iip->ili_fields & ~XFS_ILOG_TIMESTAMP); | |
1da177e4 LT |
492 | } |
493 | ||
1da177e4 LT |
494 | /* |
495 | * This is called to pin the inode associated with the inode log | |
a14a5ab5 | 496 | * item in memory so it cannot be written out. |
1da177e4 LT |
497 | */ |
498 | STATIC void | |
499 | xfs_inode_item_pin( | |
7bfa31d8 | 500 | struct xfs_log_item *lip) |
1da177e4 | 501 | { |
7bfa31d8 | 502 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 503 | |
7bfa31d8 | 504 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); |
298f7bec | 505 | ASSERT(lip->li_buf); |
7bfa31d8 CH |
506 | |
507 | trace_xfs_inode_pin(ip, _RET_IP_); | |
508 | atomic_inc(&ip->i_pincount); | |
1da177e4 LT |
509 | } |
510 | ||
511 | ||
512 | /* | |
513 | * This is called to unpin the inode associated with the inode log | |
514 | * item which was previously pinned with a call to xfs_inode_item_pin(). | |
a14a5ab5 CH |
515 | * |
516 | * Also wake up anyone in xfs_iunpin_wait() if the count goes to 0. | |
298f7bec DC |
517 | * |
518 | * Note that unpin can race with inode cluster buffer freeing marking the buffer | |
519 | * stale. In that case, flush completions are run from the buffer unpin call, | |
520 | * which may happen before the inode is unpinned. If we lose the race, there | |
521 | * will be no buffer attached to the log item, but the inode will be marked | |
522 | * XFS_ISTALE. | |
1da177e4 | 523 | */ |
1da177e4 LT |
524 | STATIC void |
525 | xfs_inode_item_unpin( | |
7bfa31d8 | 526 | struct xfs_log_item *lip, |
9412e318 | 527 | int remove) |
1da177e4 | 528 | { |
7bfa31d8 | 529 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 530 | |
4aaf15d1 | 531 | trace_xfs_inode_unpin(ip, _RET_IP_); |
298f7bec | 532 | ASSERT(lip->li_buf || xfs_iflags_test(ip, XFS_ISTALE)); |
a14a5ab5 CH |
533 | ASSERT(atomic_read(&ip->i_pincount) > 0); |
534 | if (atomic_dec_and_test(&ip->i_pincount)) | |
f392e631 | 535 | wake_up_bit(&ip->i_flags, __XFS_IPINNED_BIT); |
1da177e4 LT |
536 | } |
537 | ||
1da177e4 | 538 | STATIC uint |
43ff2122 CH |
539 | xfs_inode_item_push( |
540 | struct xfs_log_item *lip, | |
541 | struct list_head *buffer_list) | |
57e80956 MW |
542 | __releases(&lip->li_ailp->ail_lock) |
543 | __acquires(&lip->li_ailp->ail_lock) | |
1da177e4 | 544 | { |
7bfa31d8 CH |
545 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
546 | struct xfs_inode *ip = iip->ili_inode; | |
d3a304b6 | 547 | struct xfs_buf *bp = lip->li_buf; |
43ff2122 CH |
548 | uint rval = XFS_ITEM_SUCCESS; |
549 | int error; | |
1da177e4 | 550 | |
d2d7c047 DC |
551 | if (!bp || (ip->i_flags & XFS_ISTALE)) { |
552 | /* | |
553 | * Inode item/buffer is being being aborted due to cluster | |
554 | * buffer deletion. Trigger a log force to have that operation | |
555 | * completed and items removed from the AIL before the next push | |
556 | * attempt. | |
557 | */ | |
558 | return XFS_ITEM_PINNED; | |
559 | } | |
90c60e16 | 560 | |
d2d7c047 | 561 | if (xfs_ipincount(ip) > 0 || xfs_buf_ispinned(bp)) |
1da177e4 | 562 | return XFS_ITEM_PINNED; |
1da177e4 | 563 | |
718ecc50 | 564 | if (xfs_iflags_test(ip, XFS_IFLUSHING)) |
90c60e16 | 565 | return XFS_ITEM_FLUSHING; |
1da177e4 | 566 | |
90c60e16 DC |
567 | if (!xfs_buf_trylock(bp)) |
568 | return XFS_ITEM_LOCKED; | |
4c46819a | 569 | |
90c60e16 | 570 | spin_unlock(&lip->li_ailp->ail_lock); |
9a3a5dab | 571 | |
43ff2122 | 572 | /* |
90c60e16 DC |
573 | * We need to hold a reference for flushing the cluster buffer as it may |
574 | * fail the buffer without IO submission. In which case, we better get a | |
575 | * reference for that completion because otherwise we don't get a | |
576 | * reference for IO until we queue the buffer for delwri submission. | |
43ff2122 | 577 | */ |
90c60e16 | 578 | xfs_buf_hold(bp); |
5717ea4d | 579 | error = xfs_iflush_cluster(bp); |
43ff2122 CH |
580 | if (!error) { |
581 | if (!xfs_buf_delwri_queue(bp, buffer_list)) | |
582 | rval = XFS_ITEM_FLUSHING; | |
583 | xfs_buf_relse(bp); | |
90c60e16 | 584 | } else { |
5717ea4d DC |
585 | /* |
586 | * Release the buffer if we were unable to flush anything. On | |
587 | * any other error, the buffer has already been released. | |
588 | */ | |
589 | if (error == -EAGAIN) | |
590 | xfs_buf_relse(bp); | |
d4bc4c5f | 591 | rval = XFS_ITEM_LOCKED; |
90c60e16 | 592 | } |
43ff2122 | 593 | |
57e80956 | 594 | spin_lock(&lip->li_ailp->ail_lock); |
43ff2122 | 595 | return rval; |
1da177e4 LT |
596 | } |
597 | ||
598 | /* | |
599 | * Unlock the inode associated with the inode log item. | |
1da177e4 LT |
600 | */ |
601 | STATIC void | |
ddf92053 | 602 | xfs_inode_item_release( |
7bfa31d8 | 603 | struct xfs_log_item *lip) |
1da177e4 | 604 | { |
7bfa31d8 CH |
605 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
606 | struct xfs_inode *ip = iip->ili_inode; | |
898621d5 | 607 | unsigned short lock_flags; |
1da177e4 | 608 | |
f3ca8738 CH |
609 | ASSERT(ip->i_itemp != NULL); |
610 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); | |
1da177e4 | 611 | |
898621d5 CH |
612 | lock_flags = iip->ili_lock_flags; |
613 | iip->ili_lock_flags = 0; | |
ddc3415a | 614 | if (lock_flags) |
f3ca8738 | 615 | xfs_iunlock(ip, lock_flags); |
1da177e4 LT |
616 | } |
617 | ||
618 | /* | |
de25c181 DC |
619 | * This is called to find out where the oldest active copy of the inode log |
620 | * item in the on disk log resides now that the last log write of it completed | |
621 | * at the given lsn. Since we always re-log all dirty data in an inode, the | |
622 | * latest copy in the on disk log is the only one that matters. Therefore, | |
623 | * simply return the given lsn. | |
624 | * | |
625 | * If the inode has been marked stale because the cluster is being freed, we | |
626 | * don't want to (re-)insert this inode into the AIL. There is a race condition | |
627 | * where the cluster buffer may be unpinned before the inode is inserted into | |
628 | * the AIL during transaction committed processing. If the buffer is unpinned | |
629 | * before the inode item has been committed and inserted, then it is possible | |
1316d4da | 630 | * for the buffer to be written and IO completes before the inode is inserted |
de25c181 DC |
631 | * into the AIL. In that case, we'd be inserting a clean, stale inode into the |
632 | * AIL which will never get removed. It will, however, get reclaimed which | |
633 | * triggers an assert in xfs_inode_free() complaining about freein an inode | |
634 | * still in the AIL. | |
635 | * | |
1316d4da DC |
636 | * To avoid this, just unpin the inode directly and return a LSN of -1 so the |
637 | * transaction committed code knows that it does not need to do any further | |
638 | * processing on the item. | |
1da177e4 | 639 | */ |
1da177e4 LT |
640 | STATIC xfs_lsn_t |
641 | xfs_inode_item_committed( | |
7bfa31d8 | 642 | struct xfs_log_item *lip, |
1da177e4 LT |
643 | xfs_lsn_t lsn) |
644 | { | |
de25c181 DC |
645 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
646 | struct xfs_inode *ip = iip->ili_inode; | |
647 | ||
1316d4da DC |
648 | if (xfs_iflags_test(ip, XFS_ISTALE)) { |
649 | xfs_inode_item_unpin(lip, 0); | |
650 | return -1; | |
651 | } | |
7bfa31d8 | 652 | return lsn; |
1da177e4 LT |
653 | } |
654 | ||
1da177e4 LT |
655 | STATIC void |
656 | xfs_inode_item_committing( | |
7bfa31d8 | 657 | struct xfs_log_item *lip, |
5f9b4b0d | 658 | xfs_csn_t seq) |
1da177e4 | 659 | { |
5f9b4b0d | 660 | INODE_ITEM(lip)->ili_commit_seq = seq; |
ddf92053 | 661 | return xfs_inode_item_release(lip); |
1da177e4 LT |
662 | } |
663 | ||
272e42b2 | 664 | static const struct xfs_item_ops xfs_inode_item_ops = { |
7bfa31d8 CH |
665 | .iop_size = xfs_inode_item_size, |
666 | .iop_format = xfs_inode_item_format, | |
667 | .iop_pin = xfs_inode_item_pin, | |
668 | .iop_unpin = xfs_inode_item_unpin, | |
ddf92053 | 669 | .iop_release = xfs_inode_item_release, |
7bfa31d8 CH |
670 | .iop_committed = xfs_inode_item_committed, |
671 | .iop_push = xfs_inode_item_push, | |
ddf92053 | 672 | .iop_committing = xfs_inode_item_committing, |
1da177e4 LT |
673 | }; |
674 | ||
675 | ||
676 | /* | |
677 | * Initialize the inode log item for a newly allocated (in-core) inode. | |
678 | */ | |
679 | void | |
680 | xfs_inode_item_init( | |
7bfa31d8 CH |
681 | struct xfs_inode *ip, |
682 | struct xfs_mount *mp) | |
1da177e4 | 683 | { |
7bfa31d8 | 684 | struct xfs_inode_log_item *iip; |
1da177e4 LT |
685 | |
686 | ASSERT(ip->i_itemp == NULL); | |
182696fb | 687 | iip = ip->i_itemp = kmem_cache_zalloc(xfs_ili_cache, |
32a2b11f | 688 | GFP_KERNEL | __GFP_NOFAIL); |
1da177e4 | 689 | |
1da177e4 | 690 | iip->ili_inode = ip; |
1319ebef | 691 | spin_lock_init(&iip->ili_lock); |
43f5efc5 DC |
692 | xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE, |
693 | &xfs_inode_item_ops); | |
1da177e4 LT |
694 | } |
695 | ||
696 | /* | |
697 | * Free the inode log item and any memory hanging off of it. | |
698 | */ | |
699 | void | |
700 | xfs_inode_item_destroy( | |
298f7bec | 701 | struct xfs_inode *ip) |
1da177e4 | 702 | { |
298f7bec DC |
703 | struct xfs_inode_log_item *iip = ip->i_itemp; |
704 | ||
705 | ASSERT(iip->ili_item.li_buf == NULL); | |
706 | ||
707 | ip->i_itemp = NULL; | |
708 | kmem_free(iip->ili_item.li_lv_shadow); | |
182696fb | 709 | kmem_cache_free(xfs_ili_cache, iip); |
1da177e4 LT |
710 | } |
711 | ||
712 | ||
713 | /* | |
a69a1dc2 DC |
714 | * We only want to pull the item from the AIL if it is actually there |
715 | * and its location in the log has not changed since we started the | |
716 | * flush. Thus, we only bother if the inode's lsn has not changed. | |
1da177e4 | 717 | */ |
a69a1dc2 DC |
718 | static void |
719 | xfs_iflush_ail_updates( | |
720 | struct xfs_ail *ailp, | |
721 | struct list_head *list) | |
1da177e4 | 722 | { |
a69a1dc2 DC |
723 | struct xfs_log_item *lip; |
724 | xfs_lsn_t tail_lsn = 0; | |
30136832 | 725 | |
a69a1dc2 DC |
726 | /* this is an opencoded batch version of xfs_trans_ail_delete */ |
727 | spin_lock(&ailp->ail_lock); | |
728 | list_for_each_entry(lip, list, li_bio_list) { | |
729 | xfs_lsn_t lsn; | |
30136832 | 730 | |
a69a1dc2 DC |
731 | clear_bit(XFS_LI_FAILED, &lip->li_flags); |
732 | if (INODE_ITEM(lip)->ili_flush_lsn != lip->li_lsn) | |
48d55e2a DC |
733 | continue; |
734 | ||
01728b44 DC |
735 | /* |
736 | * dgc: Not sure how this happens, but it happens very | |
737 | * occassionaly via generic/388. xfs_iflush_abort() also | |
738 | * silently handles this same "under writeback but not in AIL at | |
739 | * shutdown" condition via xfs_trans_ail_delete(). | |
740 | */ | |
741 | if (!test_bit(XFS_LI_IN_AIL, &lip->li_flags)) { | |
742 | ASSERT(xlog_is_shutdown(lip->li_log)); | |
743 | continue; | |
744 | } | |
745 | ||
a69a1dc2 DC |
746 | lsn = xfs_ail_delete_one(ailp, lip); |
747 | if (!tail_lsn && lsn) | |
748 | tail_lsn = lsn; | |
30136832 | 749 | } |
a69a1dc2 DC |
750 | xfs_ail_update_finish(ailp, tail_lsn); |
751 | } | |
1da177e4 | 752 | |
a69a1dc2 DC |
753 | /* |
754 | * Walk the list of inodes that have completed their IOs. If they are clean | |
755 | * remove them from the list and dissociate them from the buffer. Buffers that | |
756 | * are still dirty remain linked to the buffer and on the list. Caller must | |
757 | * handle them appropriately. | |
758 | */ | |
759 | static void | |
760 | xfs_iflush_finish( | |
761 | struct xfs_buf *bp, | |
762 | struct list_head *list) | |
763 | { | |
764 | struct xfs_log_item *lip, *n; | |
1da177e4 | 765 | |
a69a1dc2 DC |
766 | list_for_each_entry_safe(lip, n, list, li_bio_list) { |
767 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
298f7bec DC |
768 | bool drop_buffer = false; |
769 | ||
1319ebef | 770 | spin_lock(&iip->ili_lock); |
298f7bec DC |
771 | |
772 | /* | |
773 | * Remove the reference to the cluster buffer if the inode is | |
a69a1dc2 DC |
774 | * clean in memory and drop the buffer reference once we've |
775 | * dropped the locks we hold. | |
298f7bec DC |
776 | */ |
777 | ASSERT(iip->ili_item.li_buf == bp); | |
778 | if (!iip->ili_fields) { | |
779 | iip->ili_item.li_buf = NULL; | |
a69a1dc2 | 780 | list_del_init(&lip->li_bio_list); |
298f7bec DC |
781 | drop_buffer = true; |
782 | } | |
30136832 | 783 | iip->ili_last_fields = 0; |
298f7bec | 784 | iip->ili_flush_lsn = 0; |
1319ebef | 785 | spin_unlock(&iip->ili_lock); |
718ecc50 | 786 | xfs_iflags_clear(iip->ili_inode, XFS_IFLUSHING); |
298f7bec DC |
787 | if (drop_buffer) |
788 | xfs_buf_rele(bp); | |
30136832 | 789 | } |
1da177e4 LT |
790 | } |
791 | ||
a69a1dc2 DC |
792 | /* |
793 | * Inode buffer IO completion routine. It is responsible for removing inodes | |
718ecc50 DC |
794 | * attached to the buffer from the AIL if they have not been re-logged and |
795 | * completing the inode flush. | |
a69a1dc2 DC |
796 | */ |
797 | void | |
664ffb8a | 798 | xfs_buf_inode_iodone( |
a69a1dc2 DC |
799 | struct xfs_buf *bp) |
800 | { | |
801 | struct xfs_log_item *lip, *n; | |
802 | LIST_HEAD(flushed_inodes); | |
803 | LIST_HEAD(ail_updates); | |
804 | ||
805 | /* | |
806 | * Pull the attached inodes from the buffer one at a time and take the | |
807 | * appropriate action on them. | |
808 | */ | |
809 | list_for_each_entry_safe(lip, n, &bp->b_li_list, li_bio_list) { | |
810 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
811 | ||
812 | if (xfs_iflags_test(iip->ili_inode, XFS_ISTALE)) { | |
813 | xfs_iflush_abort(iip->ili_inode); | |
814 | continue; | |
815 | } | |
816 | if (!iip->ili_last_fields) | |
817 | continue; | |
818 | ||
819 | /* Do an unlocked check for needing the AIL lock. */ | |
820 | if (iip->ili_flush_lsn == lip->li_lsn || | |
821 | test_bit(XFS_LI_FAILED, &lip->li_flags)) | |
822 | list_move_tail(&lip->li_bio_list, &ail_updates); | |
823 | else | |
824 | list_move_tail(&lip->li_bio_list, &flushed_inodes); | |
825 | } | |
826 | ||
827 | if (!list_empty(&ail_updates)) { | |
828 | xfs_iflush_ail_updates(bp->b_mount->m_ail, &ail_updates); | |
829 | list_splice_tail(&ail_updates, &flushed_inodes); | |
830 | } | |
831 | ||
832 | xfs_iflush_finish(bp, &flushed_inodes); | |
833 | if (!list_empty(&flushed_inodes)) | |
834 | list_splice_tail(&flushed_inodes, &bp->b_li_list); | |
835 | } | |
836 | ||
664ffb8a CH |
837 | void |
838 | xfs_buf_inode_io_fail( | |
839 | struct xfs_buf *bp) | |
840 | { | |
841 | struct xfs_log_item *lip; | |
842 | ||
843 | list_for_each_entry(lip, &bp->b_li_list, li_bio_list) | |
844 | set_bit(XFS_LI_FAILED, &lip->li_flags); | |
845 | } | |
846 | ||
1da177e4 | 847 | /* |
d2d7c047 DC |
848 | * Clear the inode logging fields so no more flushes are attempted. If we are |
849 | * on a buffer list, it is now safe to remove it because the buffer is | |
850 | * guaranteed to be locked. The caller will drop the reference to the buffer | |
851 | * the log item held. | |
852 | */ | |
853 | static void | |
854 | xfs_iflush_abort_clean( | |
855 | struct xfs_inode_log_item *iip) | |
856 | { | |
857 | iip->ili_last_fields = 0; | |
858 | iip->ili_fields = 0; | |
859 | iip->ili_fsync_fields = 0; | |
860 | iip->ili_flush_lsn = 0; | |
861 | iip->ili_item.li_buf = NULL; | |
862 | list_del_init(&iip->ili_item.li_bio_list); | |
863 | } | |
864 | ||
865 | /* | |
866 | * Abort flushing the inode from a context holding the cluster buffer locked. | |
867 | * | |
868 | * This is the normal runtime method of aborting writeback of an inode that is | |
869 | * attached to a cluster buffer. It occurs when the inode and the backing | |
870 | * cluster buffer have been freed (i.e. inode is XFS_ISTALE), or when cluster | |
871 | * flushing or buffer IO completion encounters a log shutdown situation. | |
872 | * | |
873 | * If we need to abort inode writeback and we don't already hold the buffer | |
874 | * locked, call xfs_iflush_shutdown_abort() instead as this should only ever be | |
875 | * necessary in a shutdown situation. | |
1da177e4 LT |
876 | */ |
877 | void | |
878 | xfs_iflush_abort( | |
298f7bec | 879 | struct xfs_inode *ip) |
1da177e4 | 880 | { |
298f7bec | 881 | struct xfs_inode_log_item *iip = ip->i_itemp; |
d2d7c047 | 882 | struct xfs_buf *bp; |
1da177e4 | 883 | |
d2d7c047 DC |
884 | if (!iip) { |
885 | /* clean inode, nothing to do */ | |
886 | xfs_iflags_clear(ip, XFS_IFLUSHING); | |
887 | return; | |
888 | } | |
889 | ||
890 | /* | |
891 | * Remove the inode item from the AIL before we clear its internal | |
892 | * state. Whilst the inode is in the AIL, it should have a valid buffer | |
893 | * pointer for push operations to access - it is only safe to remove the | |
894 | * inode from the buffer once it has been removed from the AIL. | |
895 | * | |
896 | * We also clear the failed bit before removing the item from the AIL | |
897 | * as xfs_trans_ail_delete()->xfs_clear_li_failed() will release buffer | |
898 | * references the inode item owns and needs to hold until we've fully | |
899 | * aborted the inode log item and detached it from the buffer. | |
900 | */ | |
901 | clear_bit(XFS_LI_FAILED, &iip->ili_item.li_flags); | |
902 | xfs_trans_ail_delete(&iip->ili_item, 0); | |
903 | ||
904 | /* | |
905 | * Grab the inode buffer so can we release the reference the inode log | |
906 | * item holds on it. | |
907 | */ | |
908 | spin_lock(&iip->ili_lock); | |
909 | bp = iip->ili_item.li_buf; | |
910 | xfs_iflush_abort_clean(iip); | |
911 | spin_unlock(&iip->ili_lock); | |
298f7bec | 912 | |
d2d7c047 DC |
913 | xfs_iflags_clear(ip, XFS_IFLUSHING); |
914 | if (bp) | |
915 | xfs_buf_rele(bp); | |
916 | } | |
917 | ||
918 | /* | |
919 | * Abort an inode flush in the case of a shutdown filesystem. This can be called | |
920 | * from anywhere with just an inode reference and does not require holding the | |
921 | * inode cluster buffer locked. If the inode is attached to a cluster buffer, | |
922 | * it will grab and lock it safely, then abort the inode flush. | |
923 | */ | |
924 | void | |
925 | xfs_iflush_shutdown_abort( | |
926 | struct xfs_inode *ip) | |
927 | { | |
928 | struct xfs_inode_log_item *iip = ip->i_itemp; | |
929 | struct xfs_buf *bp; | |
930 | ||
931 | if (!iip) { | |
932 | /* clean inode, nothing to do */ | |
933 | xfs_iflags_clear(ip, XFS_IFLUSHING); | |
934 | return; | |
935 | } | |
936 | ||
937 | spin_lock(&iip->ili_lock); | |
938 | bp = iip->ili_item.li_buf; | |
939 | if (!bp) { | |
940 | spin_unlock(&iip->ili_lock); | |
941 | xfs_iflush_abort(ip); | |
942 | return; | |
943 | } | |
944 | ||
945 | /* | |
946 | * We have to take a reference to the buffer so that it doesn't get | |
947 | * freed when we drop the ili_lock and then wait to lock the buffer. | |
948 | * We'll clean up the extra reference after we pick up the ili_lock | |
949 | * again. | |
950 | */ | |
951 | xfs_buf_hold(bp); | |
952 | spin_unlock(&iip->ili_lock); | |
953 | xfs_buf_lock(bp); | |
954 | ||
955 | spin_lock(&iip->ili_lock); | |
956 | if (!iip->ili_item.li_buf) { | |
1da177e4 | 957 | /* |
d2d7c047 DC |
958 | * Raced with another removal, hold the only reference |
959 | * to bp now. Inode should not be in the AIL now, so just clean | |
960 | * up and return; | |
1da177e4 | 961 | */ |
d2d7c047 DC |
962 | ASSERT(list_empty(&iip->ili_item.li_bio_list)); |
963 | ASSERT(!test_bit(XFS_LI_IN_AIL, &iip->ili_item.li_flags)); | |
964 | xfs_iflush_abort_clean(iip); | |
1319ebef | 965 | spin_unlock(&iip->ili_lock); |
d2d7c047 DC |
966 | xfs_iflags_clear(ip, XFS_IFLUSHING); |
967 | xfs_buf_relse(bp); | |
968 | return; | |
1da177e4 | 969 | } |
d2d7c047 DC |
970 | |
971 | /* | |
972 | * Got two references to bp. The first will get dropped by | |
973 | * xfs_iflush_abort() when the item is removed from the buffer list, but | |
974 | * we can't drop our reference until _abort() returns because we have to | |
975 | * unlock the buffer as well. Hence we abort and then unlock and release | |
976 | * our reference to the buffer. | |
977 | */ | |
978 | ASSERT(iip->ili_item.li_buf == bp); | |
979 | spin_unlock(&iip->ili_lock); | |
980 | xfs_iflush_abort(ip); | |
981 | xfs_buf_relse(bp); | |
1da177e4 LT |
982 | } |
983 | ||
d2d7c047 | 984 | |
6d192a9b | 985 | /* |
20413e37 DC |
986 | * convert an xfs_inode_log_format struct from the old 32 bit version |
987 | * (which can have different field alignments) to the native 64 bit version | |
6d192a9b TS |
988 | */ |
989 | int | |
990 | xfs_inode_item_format_convert( | |
20413e37 DC |
991 | struct xfs_log_iovec *buf, |
992 | struct xfs_inode_log_format *in_f) | |
6d192a9b | 993 | { |
20413e37 DC |
994 | struct xfs_inode_log_format_32 *in_f32 = buf->i_addr; |
995 | ||
a5155b87 DW |
996 | if (buf->i_len != sizeof(*in_f32)) { |
997 | XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_LOW, NULL); | |
20413e37 | 998 | return -EFSCORRUPTED; |
a5155b87 | 999 | } |
20413e37 DC |
1000 | |
1001 | in_f->ilf_type = in_f32->ilf_type; | |
1002 | in_f->ilf_size = in_f32->ilf_size; | |
1003 | in_f->ilf_fields = in_f32->ilf_fields; | |
1004 | in_f->ilf_asize = in_f32->ilf_asize; | |
1005 | in_f->ilf_dsize = in_f32->ilf_dsize; | |
1006 | in_f->ilf_ino = in_f32->ilf_ino; | |
42b67dc6 | 1007 | memcpy(&in_f->ilf_u, &in_f32->ilf_u, sizeof(in_f->ilf_u)); |
20413e37 DC |
1008 | in_f->ilf_blkno = in_f32->ilf_blkno; |
1009 | in_f->ilf_len = in_f32->ilf_len; | |
1010 | in_f->ilf_boffset = in_f32->ilf_boffset; | |
1011 | return 0; | |
6d192a9b | 1012 | } |