]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
7b718769 NS |
2 | * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc. |
3 | * All Rights Reserved. | |
1da177e4 | 4 | * |
7b718769 NS |
5 | * This program is free software; you can redistribute it and/or |
6 | * modify it under the terms of the GNU General Public License as | |
1da177e4 LT |
7 | * published by the Free Software Foundation. |
8 | * | |
7b718769 NS |
9 | * This program is distributed in the hope that it would be useful, |
10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
12 | * GNU General Public License for more details. | |
1da177e4 | 13 | * |
7b718769 NS |
14 | * You should have received a copy of the GNU General Public License |
15 | * along with this program; if not, write the Free Software Foundation, | |
16 | * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
1da177e4 | 17 | */ |
1da177e4 | 18 | #include "xfs.h" |
a844f451 | 19 | #include "xfs_fs.h" |
a4fbe6ab | 20 | #include "xfs_format.h" |
239880ef DC |
21 | #include "xfs_log_format.h" |
22 | #include "xfs_trans_resv.h" | |
1da177e4 | 23 | #include "xfs_mount.h" |
1da177e4 | 24 | #include "xfs_inode.h" |
239880ef | 25 | #include "xfs_trans.h" |
a844f451 | 26 | #include "xfs_inode_item.h" |
db7a19f2 | 27 | #include "xfs_error.h" |
0b1b213f | 28 | #include "xfs_trace.h" |
239880ef | 29 | #include "xfs_trans_priv.h" |
d3a304b6 | 30 | #include "xfs_buf_item.h" |
1234351c | 31 | #include "xfs_log.h" |
1da177e4 LT |
32 | |
33 | ||
34 | kmem_zone_t *xfs_ili_zone; /* inode log item zone */ | |
35 | ||
7bfa31d8 CH |
36 | static inline struct xfs_inode_log_item *INODE_ITEM(struct xfs_log_item *lip) |
37 | { | |
38 | return container_of(lip, struct xfs_inode_log_item, ili_item); | |
39 | } | |
40 | ||
166d1368 | 41 | STATIC void |
ce9641d6 CH |
42 | xfs_inode_item_data_fork_size( |
43 | struct xfs_inode_log_item *iip, | |
166d1368 DC |
44 | int *nvecs, |
45 | int *nbytes) | |
1da177e4 | 46 | { |
7bfa31d8 | 47 | struct xfs_inode *ip = iip->ili_inode; |
166d1368 | 48 | |
1da177e4 LT |
49 | switch (ip->i_d.di_format) { |
50 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 51 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
339a5f5d | 52 | ip->i_d.di_nextents > 0 && |
166d1368 DC |
53 | ip->i_df.if_bytes > 0) { |
54 | /* worst case, doesn't subtract delalloc extents */ | |
55 | *nbytes += XFS_IFORK_DSIZE(ip); | |
56 | *nvecs += 1; | |
57 | } | |
1da177e4 | 58 | break; |
1da177e4 | 59 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 60 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
166d1368 DC |
61 | ip->i_df.if_broot_bytes > 0) { |
62 | *nbytes += ip->i_df.if_broot_bytes; | |
63 | *nvecs += 1; | |
64 | } | |
1da177e4 | 65 | break; |
1da177e4 | 66 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 67 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
166d1368 DC |
68 | ip->i_df.if_bytes > 0) { |
69 | *nbytes += roundup(ip->i_df.if_bytes, 4); | |
70 | *nvecs += 1; | |
71 | } | |
1da177e4 LT |
72 | break; |
73 | ||
74 | case XFS_DINODE_FMT_DEV: | |
1da177e4 | 75 | case XFS_DINODE_FMT_UUID: |
1da177e4 | 76 | break; |
1da177e4 LT |
77 | default: |
78 | ASSERT(0); | |
79 | break; | |
80 | } | |
ce9641d6 | 81 | } |
1da177e4 | 82 | |
ce9641d6 CH |
83 | STATIC void |
84 | xfs_inode_item_attr_fork_size( | |
85 | struct xfs_inode_log_item *iip, | |
86 | int *nvecs, | |
87 | int *nbytes) | |
88 | { | |
89 | struct xfs_inode *ip = iip->ili_inode; | |
1da177e4 | 90 | |
1da177e4 LT |
91 | switch (ip->i_d.di_aformat) { |
92 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 93 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
339a5f5d | 94 | ip->i_d.di_anextents > 0 && |
166d1368 DC |
95 | ip->i_afp->if_bytes > 0) { |
96 | /* worst case, doesn't subtract unused space */ | |
97 | *nbytes += XFS_IFORK_ASIZE(ip); | |
98 | *nvecs += 1; | |
99 | } | |
1da177e4 | 100 | break; |
1da177e4 | 101 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 102 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
166d1368 DC |
103 | ip->i_afp->if_broot_bytes > 0) { |
104 | *nbytes += ip->i_afp->if_broot_bytes; | |
105 | *nvecs += 1; | |
106 | } | |
1da177e4 | 107 | break; |
1da177e4 | 108 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 109 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
166d1368 DC |
110 | ip->i_afp->if_bytes > 0) { |
111 | *nbytes += roundup(ip->i_afp->if_bytes, 4); | |
112 | *nvecs += 1; | |
113 | } | |
1da177e4 | 114 | break; |
1da177e4 LT |
115 | default: |
116 | ASSERT(0); | |
117 | break; | |
118 | } | |
1da177e4 LT |
119 | } |
120 | ||
ce9641d6 CH |
121 | /* |
122 | * This returns the number of iovecs needed to log the given inode item. | |
123 | * | |
124 | * We need one iovec for the inode log format structure, one for the | |
125 | * inode core, and possibly one for the inode data/extents/b-tree root | |
126 | * and one for the inode attribute data/extents/b-tree root. | |
127 | */ | |
128 | STATIC void | |
129 | xfs_inode_item_size( | |
130 | struct xfs_log_item *lip, | |
131 | int *nvecs, | |
132 | int *nbytes) | |
133 | { | |
134 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
135 | struct xfs_inode *ip = iip->ili_inode; | |
136 | ||
137 | *nvecs += 2; | |
138 | *nbytes += sizeof(struct xfs_inode_log_format) + | |
f8d55aa0 | 139 | xfs_log_dinode_size(ip->i_d.di_version); |
ce9641d6 CH |
140 | |
141 | xfs_inode_item_data_fork_size(iip, nvecs, nbytes); | |
142 | if (XFS_IFORK_Q(ip)) | |
143 | xfs_inode_item_attr_fork_size(iip, nvecs, nbytes); | |
144 | } | |
145 | ||
1234351c | 146 | STATIC void |
3de559fb CH |
147 | xfs_inode_item_format_data_fork( |
148 | struct xfs_inode_log_item *iip, | |
bde7cff6 CH |
149 | struct xfs_inode_log_format *ilf, |
150 | struct xfs_log_vec *lv, | |
151 | struct xfs_log_iovec **vecp) | |
1da177e4 | 152 | { |
7bfa31d8 | 153 | struct xfs_inode *ip = iip->ili_inode; |
1da177e4 | 154 | size_t data_bytes; |
1da177e4 LT |
155 | |
156 | switch (ip->i_d.di_format) { | |
157 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 158 | iip->ili_fields &= |
339a5f5d CH |
159 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | |
160 | XFS_ILOG_DEV | XFS_ILOG_UUID); | |
161 | ||
f5d8d5c4 | 162 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
339a5f5d CH |
163 | ip->i_d.di_nextents > 0 && |
164 | ip->i_df.if_bytes > 0) { | |
da776503 CH |
165 | struct xfs_bmbt_rec *p; |
166 | ||
1da177e4 | 167 | ASSERT(ip->i_df.if_u1.if_extents != NULL); |
5d829300 | 168 | ASSERT(xfs_iext_count(&ip->i_df) > 0); |
da776503 CH |
169 | |
170 | p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IEXT); | |
171 | data_bytes = xfs_iextents_copy(ip, p, XFS_DATA_FORK); | |
172 | xlog_finish_iovec(lv, *vecp, data_bytes); | |
173 | ||
174 | ASSERT(data_bytes <= ip->i_df.if_bytes); | |
175 | ||
176 | ilf->ilf_dsize = data_bytes; | |
bde7cff6 | 177 | ilf->ilf_size++; |
339a5f5d | 178 | } else { |
f5d8d5c4 | 179 | iip->ili_fields &= ~XFS_ILOG_DEXT; |
1da177e4 LT |
180 | } |
181 | break; | |
1da177e4 | 182 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 183 | iip->ili_fields &= |
339a5f5d CH |
184 | ~(XFS_ILOG_DDATA | XFS_ILOG_DEXT | |
185 | XFS_ILOG_DEV | XFS_ILOG_UUID); | |
186 | ||
f5d8d5c4 | 187 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
339a5f5d | 188 | ip->i_df.if_broot_bytes > 0) { |
1da177e4 | 189 | ASSERT(ip->i_df.if_broot != NULL); |
bde7cff6 | 190 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IBROOT, |
1234351c CH |
191 | ip->i_df.if_broot, |
192 | ip->i_df.if_broot_bytes); | |
bde7cff6 CH |
193 | ilf->ilf_dsize = ip->i_df.if_broot_bytes; |
194 | ilf->ilf_size++; | |
339a5f5d | 195 | } else { |
f5d8d5c4 | 196 | ASSERT(!(iip->ili_fields & |
339a5f5d | 197 | XFS_ILOG_DBROOT)); |
f5d8d5c4 | 198 | iip->ili_fields &= ~XFS_ILOG_DBROOT; |
1da177e4 LT |
199 | } |
200 | break; | |
1da177e4 | 201 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 202 | iip->ili_fields &= |
339a5f5d CH |
203 | ~(XFS_ILOG_DEXT | XFS_ILOG_DBROOT | |
204 | XFS_ILOG_DEV | XFS_ILOG_UUID); | |
f5d8d5c4 | 205 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
339a5f5d | 206 | ip->i_df.if_bytes > 0) { |
1da177e4 LT |
207 | /* |
208 | * Round i_bytes up to a word boundary. | |
209 | * The underlying memory is guaranteed to | |
210 | * to be there by xfs_idata_realloc(). | |
211 | */ | |
212 | data_bytes = roundup(ip->i_df.if_bytes, 4); | |
1234351c | 213 | ASSERT(ip->i_df.if_real_bytes == 0 || |
30ee052e | 214 | ip->i_df.if_real_bytes >= data_bytes); |
1234351c CH |
215 | ASSERT(ip->i_df.if_u1.if_data != NULL); |
216 | ASSERT(ip->i_d.di_size > 0); | |
bde7cff6 | 217 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_ILOCAL, |
1234351c | 218 | ip->i_df.if_u1.if_data, data_bytes); |
bde7cff6 CH |
219 | ilf->ilf_dsize = (unsigned)data_bytes; |
220 | ilf->ilf_size++; | |
339a5f5d | 221 | } else { |
f5d8d5c4 | 222 | iip->ili_fields &= ~XFS_ILOG_DDATA; |
1da177e4 LT |
223 | } |
224 | break; | |
1da177e4 | 225 | case XFS_DINODE_FMT_DEV: |
f5d8d5c4 | 226 | iip->ili_fields &= |
339a5f5d CH |
227 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | |
228 | XFS_ILOG_DEXT | XFS_ILOG_UUID); | |
bde7cff6 CH |
229 | if (iip->ili_fields & XFS_ILOG_DEV) |
230 | ilf->ilf_u.ilfu_rdev = ip->i_df.if_u2.if_rdev; | |
1da177e4 | 231 | break; |
1da177e4 | 232 | case XFS_DINODE_FMT_UUID: |
f5d8d5c4 | 233 | iip->ili_fields &= |
339a5f5d CH |
234 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | |
235 | XFS_ILOG_DEXT | XFS_ILOG_DEV); | |
bde7cff6 CH |
236 | if (iip->ili_fields & XFS_ILOG_UUID) |
237 | ilf->ilf_u.ilfu_uuid = ip->i_df.if_u2.if_uuid; | |
1da177e4 | 238 | break; |
1da177e4 LT |
239 | default: |
240 | ASSERT(0); | |
241 | break; | |
242 | } | |
3de559fb CH |
243 | } |
244 | ||
1234351c | 245 | STATIC void |
3de559fb CH |
246 | xfs_inode_item_format_attr_fork( |
247 | struct xfs_inode_log_item *iip, | |
bde7cff6 CH |
248 | struct xfs_inode_log_format *ilf, |
249 | struct xfs_log_vec *lv, | |
250 | struct xfs_log_iovec **vecp) | |
3de559fb CH |
251 | { |
252 | struct xfs_inode *ip = iip->ili_inode; | |
253 | size_t data_bytes; | |
1da177e4 LT |
254 | |
255 | switch (ip->i_d.di_aformat) { | |
256 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 257 | iip->ili_fields &= |
339a5f5d CH |
258 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT); |
259 | ||
f5d8d5c4 | 260 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
339a5f5d CH |
261 | ip->i_d.di_anextents > 0 && |
262 | ip->i_afp->if_bytes > 0) { | |
da776503 CH |
263 | struct xfs_bmbt_rec *p; |
264 | ||
5d829300 | 265 | ASSERT(xfs_iext_count(ip->i_afp) == |
339a5f5d | 266 | ip->i_d.di_anextents); |
73523a2e | 267 | ASSERT(ip->i_afp->if_u1.if_extents != NULL); |
da776503 CH |
268 | |
269 | p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_EXT); | |
270 | data_bytes = xfs_iextents_copy(ip, p, XFS_ATTR_FORK); | |
271 | xlog_finish_iovec(lv, *vecp, data_bytes); | |
272 | ||
273 | ilf->ilf_asize = data_bytes; | |
bde7cff6 | 274 | ilf->ilf_size++; |
339a5f5d | 275 | } else { |
f5d8d5c4 | 276 | iip->ili_fields &= ~XFS_ILOG_AEXT; |
1da177e4 LT |
277 | } |
278 | break; | |
1da177e4 | 279 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 280 | iip->ili_fields &= |
339a5f5d CH |
281 | ~(XFS_ILOG_ADATA | XFS_ILOG_AEXT); |
282 | ||
f5d8d5c4 | 283 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
339a5f5d | 284 | ip->i_afp->if_broot_bytes > 0) { |
1da177e4 | 285 | ASSERT(ip->i_afp->if_broot != NULL); |
339a5f5d | 286 | |
bde7cff6 | 287 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_BROOT, |
1234351c CH |
288 | ip->i_afp->if_broot, |
289 | ip->i_afp->if_broot_bytes); | |
bde7cff6 CH |
290 | ilf->ilf_asize = ip->i_afp->if_broot_bytes; |
291 | ilf->ilf_size++; | |
339a5f5d | 292 | } else { |
f5d8d5c4 | 293 | iip->ili_fields &= ~XFS_ILOG_ABROOT; |
1da177e4 LT |
294 | } |
295 | break; | |
1da177e4 | 296 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 297 | iip->ili_fields &= |
339a5f5d CH |
298 | ~(XFS_ILOG_AEXT | XFS_ILOG_ABROOT); |
299 | ||
f5d8d5c4 | 300 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
339a5f5d | 301 | ip->i_afp->if_bytes > 0) { |
1da177e4 LT |
302 | /* |
303 | * Round i_bytes up to a word boundary. | |
304 | * The underlying memory is guaranteed to | |
305 | * to be there by xfs_idata_realloc(). | |
306 | */ | |
307 | data_bytes = roundup(ip->i_afp->if_bytes, 4); | |
1234351c | 308 | ASSERT(ip->i_afp->if_real_bytes == 0 || |
30ee052e | 309 | ip->i_afp->if_real_bytes >= data_bytes); |
1234351c | 310 | ASSERT(ip->i_afp->if_u1.if_data != NULL); |
bde7cff6 | 311 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_LOCAL, |
1234351c CH |
312 | ip->i_afp->if_u1.if_data, |
313 | data_bytes); | |
bde7cff6 CH |
314 | ilf->ilf_asize = (unsigned)data_bytes; |
315 | ilf->ilf_size++; | |
339a5f5d | 316 | } else { |
f5d8d5c4 | 317 | iip->ili_fields &= ~XFS_ILOG_ADATA; |
1da177e4 LT |
318 | } |
319 | break; | |
1da177e4 LT |
320 | default: |
321 | ASSERT(0); | |
322 | break; | |
323 | } | |
3de559fb CH |
324 | } |
325 | ||
f8d55aa0 | 326 | static void |
3987848c DC |
327 | xfs_inode_to_log_dinode( |
328 | struct xfs_inode *ip, | |
93f958f9 DC |
329 | struct xfs_log_dinode *to, |
330 | xfs_lsn_t lsn) | |
f8d55aa0 | 331 | { |
3987848c DC |
332 | struct xfs_icdinode *from = &ip->i_d; |
333 | struct inode *inode = VFS_I(ip); | |
334 | ||
93f958f9 DC |
335 | to->di_magic = XFS_DINODE_MAGIC; |
336 | ||
f8d55aa0 DC |
337 | to->di_version = from->di_version; |
338 | to->di_format = from->di_format; | |
f8d55aa0 DC |
339 | to->di_uid = from->di_uid; |
340 | to->di_gid = from->di_gid; | |
f8d55aa0 DC |
341 | to->di_projid_lo = from->di_projid_lo; |
342 | to->di_projid_hi = from->di_projid_hi; | |
f8d55aa0 | 343 | |
93f958f9 | 344 | memset(to->di_pad, 0, sizeof(to->di_pad)); |
faeb4e47 | 345 | memset(to->di_pad3, 0, sizeof(to->di_pad3)); |
3987848c DC |
346 | to->di_atime.t_sec = inode->i_atime.tv_sec; |
347 | to->di_atime.t_nsec = inode->i_atime.tv_nsec; | |
348 | to->di_mtime.t_sec = inode->i_mtime.tv_sec; | |
349 | to->di_mtime.t_nsec = inode->i_mtime.tv_nsec; | |
350 | to->di_ctime.t_sec = inode->i_ctime.tv_sec; | |
351 | to->di_ctime.t_nsec = inode->i_ctime.tv_nsec; | |
54d7b5c1 | 352 | to->di_nlink = inode->i_nlink; |
9e9a2674 | 353 | to->di_gen = inode->i_generation; |
c19b3b05 | 354 | to->di_mode = inode->i_mode; |
f8d55aa0 | 355 | |
f8d55aa0 DC |
356 | to->di_size = from->di_size; |
357 | to->di_nblocks = from->di_nblocks; | |
358 | to->di_extsize = from->di_extsize; | |
359 | to->di_nextents = from->di_nextents; | |
360 | to->di_anextents = from->di_anextents; | |
361 | to->di_forkoff = from->di_forkoff; | |
362 | to->di_aformat = from->di_aformat; | |
363 | to->di_dmevmask = from->di_dmevmask; | |
364 | to->di_dmstate = from->di_dmstate; | |
365 | to->di_flags = from->di_flags; | |
f8d55aa0 | 366 | |
20413e37 DC |
367 | /* log a dummy value to ensure log structure is fully initialised */ |
368 | to->di_next_unlinked = NULLAGINO; | |
369 | ||
f8d55aa0 | 370 | if (from->di_version == 3) { |
83e06f21 | 371 | to->di_changecount = inode->i_version; |
f8d55aa0 DC |
372 | to->di_crtime.t_sec = from->di_crtime.t_sec; |
373 | to->di_crtime.t_nsec = from->di_crtime.t_nsec; | |
374 | to->di_flags2 = from->di_flags2; | |
f7ca3522 | 375 | to->di_cowextsize = from->di_cowextsize; |
93f958f9 DC |
376 | to->di_ino = ip->i_ino; |
377 | to->di_lsn = lsn; | |
378 | memset(to->di_pad2, 0, sizeof(to->di_pad2)); | |
379 | uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid); | |
f8d55aa0 DC |
380 | to->di_flushiter = 0; |
381 | } else { | |
382 | to->di_flushiter = from->di_flushiter; | |
383 | } | |
384 | } | |
385 | ||
386 | /* | |
387 | * Format the inode core. Current timestamp data is only in the VFS inode | |
388 | * fields, so we need to grab them from there. Hence rather than just copying | |
389 | * the XFS inode core structure, format the fields directly into the iovec. | |
390 | */ | |
391 | static void | |
392 | xfs_inode_item_format_core( | |
393 | struct xfs_inode *ip, | |
394 | struct xfs_log_vec *lv, | |
395 | struct xfs_log_iovec **vecp) | |
396 | { | |
397 | struct xfs_log_dinode *dic; | |
398 | ||
399 | dic = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_ICORE); | |
93f958f9 | 400 | xfs_inode_to_log_dinode(ip, dic, ip->i_itemp->ili_item.li_lsn); |
f8d55aa0 DC |
401 | xlog_finish_iovec(lv, *vecp, xfs_log_dinode_size(ip->i_d.di_version)); |
402 | } | |
403 | ||
3de559fb CH |
404 | /* |
405 | * This is called to fill in the vector of log iovecs for the given inode | |
406 | * log item. It fills the first item with an inode log format structure, | |
407 | * the second with the on-disk inode structure, and a possible third and/or | |
408 | * fourth with the inode data/extents/b-tree root and inode attributes | |
409 | * data/extents/b-tree root. | |
20413e37 DC |
410 | * |
411 | * Note: Always use the 64 bit inode log format structure so we don't | |
412 | * leave an uninitialised hole in the format item on 64 bit systems. Log | |
413 | * recovery on 32 bit systems handles this just fine, so there's no reason | |
414 | * for not using an initialising the properly padded structure all the time. | |
3de559fb CH |
415 | */ |
416 | STATIC void | |
417 | xfs_inode_item_format( | |
418 | struct xfs_log_item *lip, | |
bde7cff6 | 419 | struct xfs_log_vec *lv) |
3de559fb CH |
420 | { |
421 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
422 | struct xfs_inode *ip = iip->ili_inode; | |
bde7cff6 | 423 | struct xfs_log_iovec *vecp = NULL; |
20413e37 | 424 | struct xfs_inode_log_format *ilf; |
3de559fb | 425 | |
263997a6 DC |
426 | ASSERT(ip->i_d.di_version > 1); |
427 | ||
2f251293 CH |
428 | ilf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_IFORMAT); |
429 | ilf->ilf_type = XFS_LI_INODE; | |
430 | ilf->ilf_ino = ip->i_ino; | |
431 | ilf->ilf_blkno = ip->i_imap.im_blkno; | |
432 | ilf->ilf_len = ip->i_imap.im_len; | |
433 | ilf->ilf_boffset = ip->i_imap.im_boffset; | |
434 | ilf->ilf_fields = XFS_ILOG_CORE; | |
435 | ilf->ilf_size = 2; /* format + core */ | |
20413e37 DC |
436 | |
437 | /* | |
438 | * make sure we don't leak uninitialised data into the log in the case | |
439 | * when we don't log every field in the inode. | |
440 | */ | |
441 | ilf->ilf_dsize = 0; | |
442 | ilf->ilf_asize = 0; | |
443 | ilf->ilf_pad = 0; | |
444 | uuid_copy(&ilf->ilf_u.ilfu_uuid, &uuid_null); | |
445 | ||
446 | xlog_finish_iovec(lv, vecp, sizeof(*ilf)); | |
3de559fb | 447 | |
f8d55aa0 | 448 | xfs_inode_item_format_core(ip, lv, &vecp); |
bde7cff6 | 449 | xfs_inode_item_format_data_fork(iip, ilf, lv, &vecp); |
3de559fb | 450 | if (XFS_IFORK_Q(ip)) { |
bde7cff6 | 451 | xfs_inode_item_format_attr_fork(iip, ilf, lv, &vecp); |
3de559fb CH |
452 | } else { |
453 | iip->ili_fields &= | |
454 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT); | |
455 | } | |
456 | ||
2f251293 CH |
457 | /* update the format with the exact fields we actually logged */ |
458 | ilf->ilf_fields |= (iip->ili_fields & ~XFS_ILOG_TIMESTAMP); | |
1da177e4 LT |
459 | } |
460 | ||
1da177e4 LT |
461 | /* |
462 | * This is called to pin the inode associated with the inode log | |
a14a5ab5 | 463 | * item in memory so it cannot be written out. |
1da177e4 LT |
464 | */ |
465 | STATIC void | |
466 | xfs_inode_item_pin( | |
7bfa31d8 | 467 | struct xfs_log_item *lip) |
1da177e4 | 468 | { |
7bfa31d8 | 469 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 470 | |
7bfa31d8 CH |
471 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); |
472 | ||
473 | trace_xfs_inode_pin(ip, _RET_IP_); | |
474 | atomic_inc(&ip->i_pincount); | |
1da177e4 LT |
475 | } |
476 | ||
477 | ||
478 | /* | |
479 | * This is called to unpin the inode associated with the inode log | |
480 | * item which was previously pinned with a call to xfs_inode_item_pin(). | |
a14a5ab5 CH |
481 | * |
482 | * Also wake up anyone in xfs_iunpin_wait() if the count goes to 0. | |
1da177e4 | 483 | */ |
1da177e4 LT |
484 | STATIC void |
485 | xfs_inode_item_unpin( | |
7bfa31d8 | 486 | struct xfs_log_item *lip, |
9412e318 | 487 | int remove) |
1da177e4 | 488 | { |
7bfa31d8 | 489 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 490 | |
4aaf15d1 | 491 | trace_xfs_inode_unpin(ip, _RET_IP_); |
a14a5ab5 CH |
492 | ASSERT(atomic_read(&ip->i_pincount) > 0); |
493 | if (atomic_dec_and_test(&ip->i_pincount)) | |
f392e631 | 494 | wake_up_bit(&ip->i_flags, __XFS_IPINNED_BIT); |
1da177e4 LT |
495 | } |
496 | ||
d3a304b6 CM |
497 | /* |
498 | * Callback used to mark a buffer with XFS_LI_FAILED when items in the buffer | |
499 | * have been failed during writeback | |
500 | * | |
501 | * This informs the AIL that the inode is already flush locked on the next push, | |
502 | * and acquires a hold on the buffer to ensure that it isn't reclaimed before | |
503 | * dirty data makes it to disk. | |
504 | */ | |
505 | STATIC void | |
506 | xfs_inode_item_error( | |
507 | struct xfs_log_item *lip, | |
508 | struct xfs_buf *bp) | |
509 | { | |
510 | ASSERT(xfs_isiflocked(INODE_ITEM(lip)->ili_inode)); | |
511 | xfs_set_li_failed(lip, bp); | |
512 | } | |
513 | ||
1da177e4 | 514 | STATIC uint |
43ff2122 CH |
515 | xfs_inode_item_push( |
516 | struct xfs_log_item *lip, | |
517 | struct list_head *buffer_list) | |
6e3e6d55 EG |
518 | __releases(&lip->li_ailp->xa_lock) |
519 | __acquires(&lip->li_ailp->xa_lock) | |
1da177e4 | 520 | { |
7bfa31d8 CH |
521 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
522 | struct xfs_inode *ip = iip->ili_inode; | |
d3a304b6 | 523 | struct xfs_buf *bp = lip->li_buf; |
43ff2122 CH |
524 | uint rval = XFS_ITEM_SUCCESS; |
525 | int error; | |
1da177e4 | 526 | |
7bfa31d8 | 527 | if (xfs_ipincount(ip) > 0) |
1da177e4 | 528 | return XFS_ITEM_PINNED; |
1da177e4 | 529 | |
d3a304b6 CM |
530 | /* |
531 | * The buffer containing this item failed to be written back | |
532 | * previously. Resubmit the buffer for IO. | |
533 | */ | |
534 | if (lip->li_flags & XFS_LI_FAILED) { | |
535 | if (!xfs_buf_trylock(bp)) | |
536 | return XFS_ITEM_LOCKED; | |
537 | ||
538 | if (!xfs_buf_resubmit_failed_buffers(bp, lip, buffer_list)) | |
539 | rval = XFS_ITEM_FLUSHING; | |
540 | ||
541 | xfs_buf_unlock(bp); | |
542 | return rval; | |
543 | } | |
544 | ||
7bfa31d8 | 545 | if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED)) |
1da177e4 | 546 | return XFS_ITEM_LOCKED; |
1da177e4 | 547 | |
4c46819a CH |
548 | /* |
549 | * Re-check the pincount now that we stabilized the value by | |
550 | * taking the ilock. | |
551 | */ | |
552 | if (xfs_ipincount(ip) > 0) { | |
43ff2122 CH |
553 | rval = XFS_ITEM_PINNED; |
554 | goto out_unlock; | |
4c46819a CH |
555 | } |
556 | ||
9a3a5dab BF |
557 | /* |
558 | * Stale inode items should force out the iclog. | |
559 | */ | |
560 | if (ip->i_flags & XFS_ISTALE) { | |
561 | rval = XFS_ITEM_PINNED; | |
562 | goto out_unlock; | |
563 | } | |
564 | ||
43ff2122 CH |
565 | /* |
566 | * Someone else is already flushing the inode. Nothing we can do | |
567 | * here but wait for the flush to finish and remove the item from | |
568 | * the AIL. | |
569 | */ | |
1da177e4 | 570 | if (!xfs_iflock_nowait(ip)) { |
43ff2122 CH |
571 | rval = XFS_ITEM_FLUSHING; |
572 | goto out_unlock; | |
1da177e4 LT |
573 | } |
574 | ||
43ff2122 CH |
575 | ASSERT(iip->ili_fields != 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); |
576 | ASSERT(iip->ili_logged == 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); | |
577 | ||
578 | spin_unlock(&lip->li_ailp->xa_lock); | |
579 | ||
580 | error = xfs_iflush(ip, &bp); | |
581 | if (!error) { | |
582 | if (!xfs_buf_delwri_queue(bp, buffer_list)) | |
583 | rval = XFS_ITEM_FLUSHING; | |
584 | xfs_buf_relse(bp); | |
1da177e4 | 585 | } |
43ff2122 CH |
586 | |
587 | spin_lock(&lip->li_ailp->xa_lock); | |
588 | out_unlock: | |
589 | xfs_iunlock(ip, XFS_ILOCK_SHARED); | |
590 | return rval; | |
1da177e4 LT |
591 | } |
592 | ||
593 | /* | |
594 | * Unlock the inode associated with the inode log item. | |
595 | * Clear the fields of the inode and inode log item that | |
596 | * are specific to the current transaction. If the | |
597 | * hold flags is set, do not unlock the inode. | |
598 | */ | |
599 | STATIC void | |
600 | xfs_inode_item_unlock( | |
7bfa31d8 | 601 | struct xfs_log_item *lip) |
1da177e4 | 602 | { |
7bfa31d8 CH |
603 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
604 | struct xfs_inode *ip = iip->ili_inode; | |
898621d5 | 605 | unsigned short lock_flags; |
1da177e4 | 606 | |
f3ca8738 CH |
607 | ASSERT(ip->i_itemp != NULL); |
608 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); | |
1da177e4 | 609 | |
898621d5 CH |
610 | lock_flags = iip->ili_lock_flags; |
611 | iip->ili_lock_flags = 0; | |
ddc3415a | 612 | if (lock_flags) |
f3ca8738 | 613 | xfs_iunlock(ip, lock_flags); |
1da177e4 LT |
614 | } |
615 | ||
616 | /* | |
de25c181 DC |
617 | * This is called to find out where the oldest active copy of the inode log |
618 | * item in the on disk log resides now that the last log write of it completed | |
619 | * at the given lsn. Since we always re-log all dirty data in an inode, the | |
620 | * latest copy in the on disk log is the only one that matters. Therefore, | |
621 | * simply return the given lsn. | |
622 | * | |
623 | * If the inode has been marked stale because the cluster is being freed, we | |
624 | * don't want to (re-)insert this inode into the AIL. There is a race condition | |
625 | * where the cluster buffer may be unpinned before the inode is inserted into | |
626 | * the AIL during transaction committed processing. If the buffer is unpinned | |
627 | * before the inode item has been committed and inserted, then it is possible | |
1316d4da | 628 | * for the buffer to be written and IO completes before the inode is inserted |
de25c181 DC |
629 | * into the AIL. In that case, we'd be inserting a clean, stale inode into the |
630 | * AIL which will never get removed. It will, however, get reclaimed which | |
631 | * triggers an assert in xfs_inode_free() complaining about freein an inode | |
632 | * still in the AIL. | |
633 | * | |
1316d4da DC |
634 | * To avoid this, just unpin the inode directly and return a LSN of -1 so the |
635 | * transaction committed code knows that it does not need to do any further | |
636 | * processing on the item. | |
1da177e4 | 637 | */ |
1da177e4 LT |
638 | STATIC xfs_lsn_t |
639 | xfs_inode_item_committed( | |
7bfa31d8 | 640 | struct xfs_log_item *lip, |
1da177e4 LT |
641 | xfs_lsn_t lsn) |
642 | { | |
de25c181 DC |
643 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
644 | struct xfs_inode *ip = iip->ili_inode; | |
645 | ||
1316d4da DC |
646 | if (xfs_iflags_test(ip, XFS_ISTALE)) { |
647 | xfs_inode_item_unpin(lip, 0); | |
648 | return -1; | |
649 | } | |
7bfa31d8 | 650 | return lsn; |
1da177e4 LT |
651 | } |
652 | ||
1da177e4 LT |
653 | /* |
654 | * XXX rcc - this one really has to do something. Probably needs | |
655 | * to stamp in a new field in the incore inode. | |
656 | */ | |
1da177e4 LT |
657 | STATIC void |
658 | xfs_inode_item_committing( | |
7bfa31d8 | 659 | struct xfs_log_item *lip, |
1da177e4 LT |
660 | xfs_lsn_t lsn) |
661 | { | |
7bfa31d8 | 662 | INODE_ITEM(lip)->ili_last_lsn = lsn; |
1da177e4 LT |
663 | } |
664 | ||
665 | /* | |
666 | * This is the ops vector shared by all buf log items. | |
667 | */ | |
272e42b2 | 668 | static const struct xfs_item_ops xfs_inode_item_ops = { |
7bfa31d8 CH |
669 | .iop_size = xfs_inode_item_size, |
670 | .iop_format = xfs_inode_item_format, | |
671 | .iop_pin = xfs_inode_item_pin, | |
672 | .iop_unpin = xfs_inode_item_unpin, | |
7bfa31d8 CH |
673 | .iop_unlock = xfs_inode_item_unlock, |
674 | .iop_committed = xfs_inode_item_committed, | |
675 | .iop_push = xfs_inode_item_push, | |
d3a304b6 CM |
676 | .iop_committing = xfs_inode_item_committing, |
677 | .iop_error = xfs_inode_item_error | |
1da177e4 LT |
678 | }; |
679 | ||
680 | ||
681 | /* | |
682 | * Initialize the inode log item for a newly allocated (in-core) inode. | |
683 | */ | |
684 | void | |
685 | xfs_inode_item_init( | |
7bfa31d8 CH |
686 | struct xfs_inode *ip, |
687 | struct xfs_mount *mp) | |
1da177e4 | 688 | { |
7bfa31d8 | 689 | struct xfs_inode_log_item *iip; |
1da177e4 LT |
690 | |
691 | ASSERT(ip->i_itemp == NULL); | |
692 | iip = ip->i_itemp = kmem_zone_zalloc(xfs_ili_zone, KM_SLEEP); | |
693 | ||
1da177e4 | 694 | iip->ili_inode = ip; |
43f5efc5 DC |
695 | xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE, |
696 | &xfs_inode_item_ops); | |
1da177e4 LT |
697 | } |
698 | ||
699 | /* | |
700 | * Free the inode log item and any memory hanging off of it. | |
701 | */ | |
702 | void | |
703 | xfs_inode_item_destroy( | |
704 | xfs_inode_t *ip) | |
705 | { | |
b1c5ebb2 | 706 | kmem_free(ip->i_itemp->ili_item.li_lv_shadow); |
1da177e4 LT |
707 | kmem_zone_free(xfs_ili_zone, ip->i_itemp); |
708 | } | |
709 | ||
710 | ||
711 | /* | |
712 | * This is the inode flushing I/O completion routine. It is called | |
713 | * from interrupt level when the buffer containing the inode is | |
714 | * flushed to disk. It is responsible for removing the inode item | |
715 | * from the AIL if it has not been re-logged, and unlocking the inode's | |
716 | * flush lock. | |
30136832 DC |
717 | * |
718 | * To reduce AIL lock traffic as much as possible, we scan the buffer log item | |
719 | * list for other inodes that will run this function. We remove them from the | |
720 | * buffer list so we can process all the inode IO completions in one AIL lock | |
721 | * traversal. | |
1da177e4 | 722 | */ |
1da177e4 LT |
723 | void |
724 | xfs_iflush_done( | |
ca30b2a7 CH |
725 | struct xfs_buf *bp, |
726 | struct xfs_log_item *lip) | |
1da177e4 | 727 | { |
30136832 DC |
728 | struct xfs_inode_log_item *iip; |
729 | struct xfs_log_item *blip; | |
730 | struct xfs_log_item *next; | |
731 | struct xfs_log_item *prev; | |
ca30b2a7 | 732 | struct xfs_ail *ailp = lip->li_ailp; |
30136832 DC |
733 | int need_ail = 0; |
734 | ||
735 | /* | |
736 | * Scan the buffer IO completions for other inodes being completed and | |
737 | * attach them to the current inode log item. | |
738 | */ | |
adadbeef | 739 | blip = bp->b_fspriv; |
30136832 DC |
740 | prev = NULL; |
741 | while (blip != NULL) { | |
52177937 | 742 | if (blip->li_cb != xfs_iflush_done) { |
30136832 DC |
743 | prev = blip; |
744 | blip = blip->li_bio_list; | |
745 | continue; | |
746 | } | |
747 | ||
748 | /* remove from list */ | |
749 | next = blip->li_bio_list; | |
750 | if (!prev) { | |
adadbeef | 751 | bp->b_fspriv = next; |
30136832 DC |
752 | } else { |
753 | prev->li_bio_list = next; | |
754 | } | |
755 | ||
756 | /* add to current list */ | |
757 | blip->li_bio_list = lip->li_bio_list; | |
758 | lip->li_bio_list = blip; | |
759 | ||
760 | /* | |
761 | * while we have the item, do the unlocked check for needing | |
762 | * the AIL lock. | |
763 | */ | |
764 | iip = INODE_ITEM(blip); | |
d3a304b6 | 765 | if ((iip->ili_logged && blip->li_lsn == iip->ili_flush_lsn) || |
842f6e9f | 766 | (blip->li_flags & XFS_LI_FAILED)) |
30136832 DC |
767 | need_ail++; |
768 | ||
769 | blip = next; | |
770 | } | |
771 | ||
772 | /* make sure we capture the state of the initial inode. */ | |
773 | iip = INODE_ITEM(lip); | |
d3a304b6 CM |
774 | if ((iip->ili_logged && lip->li_lsn == iip->ili_flush_lsn) || |
775 | lip->li_flags & XFS_LI_FAILED) | |
30136832 | 776 | need_ail++; |
1da177e4 LT |
777 | |
778 | /* | |
779 | * We only want to pull the item from the AIL if it is | |
780 | * actually there and its location in the log has not | |
781 | * changed since we started the flush. Thus, we only bother | |
782 | * if the ili_logged flag is set and the inode's lsn has not | |
783 | * changed. First we check the lsn outside | |
784 | * the lock since it's cheaper, and then we recheck while | |
785 | * holding the lock before removing the inode from the AIL. | |
786 | */ | |
30136832 | 787 | if (need_ail) { |
27af1bbf CH |
788 | bool mlip_changed = false; |
789 | ||
790 | /* this is an opencoded batch version of xfs_trans_ail_delete */ | |
783a2f65 | 791 | spin_lock(&ailp->xa_lock); |
30136832 | 792 | for (blip = lip; blip; blip = blip->li_bio_list) { |
27af1bbf CH |
793 | if (INODE_ITEM(blip)->ili_logged && |
794 | blip->li_lsn == INODE_ITEM(blip)->ili_flush_lsn) | |
795 | mlip_changed |= xfs_ail_delete_one(ailp, blip); | |
d3a304b6 CM |
796 | else { |
797 | xfs_clear_li_failed(blip); | |
798 | } | |
1da177e4 | 799 | } |
1da177e4 | 800 | |
27af1bbf CH |
801 | if (mlip_changed) { |
802 | if (!XFS_FORCED_SHUTDOWN(ailp->xa_mount)) | |
803 | xlog_assign_tail_lsn_locked(ailp->xa_mount); | |
804 | if (list_empty(&ailp->xa_ail)) | |
805 | wake_up_all(&ailp->xa_empty); | |
806 | } | |
807 | spin_unlock(&ailp->xa_lock); | |
808 | ||
809 | if (mlip_changed) | |
810 | xfs_log_space_wake(ailp->xa_mount); | |
811 | } | |
1da177e4 LT |
812 | |
813 | /* | |
30136832 DC |
814 | * clean up and unlock the flush lock now we are done. We can clear the |
815 | * ili_last_fields bits now that we know that the data corresponding to | |
816 | * them is safely on disk. | |
1da177e4 | 817 | */ |
30136832 DC |
818 | for (blip = lip; blip; blip = next) { |
819 | next = blip->li_bio_list; | |
820 | blip->li_bio_list = NULL; | |
821 | ||
822 | iip = INODE_ITEM(blip); | |
823 | iip->ili_logged = 0; | |
824 | iip->ili_last_fields = 0; | |
825 | xfs_ifunlock(iip->ili_inode); | |
826 | } | |
1da177e4 LT |
827 | } |
828 | ||
829 | /* | |
04913fdd DC |
830 | * This is the inode flushing abort routine. It is called from xfs_iflush when |
831 | * the filesystem is shutting down to clean up the inode state. It is | |
832 | * responsible for removing the inode item from the AIL if it has not been | |
833 | * re-logged, and unlocking the inode's flush lock. | |
1da177e4 LT |
834 | */ |
835 | void | |
836 | xfs_iflush_abort( | |
04913fdd DC |
837 | xfs_inode_t *ip, |
838 | bool stale) | |
1da177e4 | 839 | { |
783a2f65 | 840 | xfs_inode_log_item_t *iip = ip->i_itemp; |
1da177e4 | 841 | |
1da177e4 LT |
842 | if (iip) { |
843 | if (iip->ili_item.li_flags & XFS_LI_IN_AIL) { | |
146e54b7 BF |
844 | xfs_trans_ail_remove(&iip->ili_item, |
845 | stale ? SHUTDOWN_LOG_IO_ERROR : | |
04913fdd | 846 | SHUTDOWN_CORRUPT_INCORE); |
1da177e4 LT |
847 | } |
848 | iip->ili_logged = 0; | |
849 | /* | |
850 | * Clear the ili_last_fields bits now that we know that the | |
851 | * data corresponding to them is safely on disk. | |
852 | */ | |
853 | iip->ili_last_fields = 0; | |
854 | /* | |
855 | * Clear the inode logging fields so no more flushes are | |
856 | * attempted. | |
857 | */ | |
f5d8d5c4 | 858 | iip->ili_fields = 0; |
fc0561ce | 859 | iip->ili_fsync_fields = 0; |
1da177e4 LT |
860 | } |
861 | /* | |
862 | * Release the inode's flush lock since we're done with it. | |
863 | */ | |
864 | xfs_ifunlock(ip); | |
865 | } | |
866 | ||
867 | void | |
868 | xfs_istale_done( | |
ca30b2a7 CH |
869 | struct xfs_buf *bp, |
870 | struct xfs_log_item *lip) | |
1da177e4 | 871 | { |
04913fdd | 872 | xfs_iflush_abort(INODE_ITEM(lip)->ili_inode, true); |
1da177e4 | 873 | } |
6d192a9b TS |
874 | |
875 | /* | |
20413e37 DC |
876 | * convert an xfs_inode_log_format struct from the old 32 bit version |
877 | * (which can have different field alignments) to the native 64 bit version | |
6d192a9b TS |
878 | */ |
879 | int | |
880 | xfs_inode_item_format_convert( | |
20413e37 DC |
881 | struct xfs_log_iovec *buf, |
882 | struct xfs_inode_log_format *in_f) | |
6d192a9b | 883 | { |
20413e37 DC |
884 | struct xfs_inode_log_format_32 *in_f32 = buf->i_addr; |
885 | ||
886 | if (buf->i_len != sizeof(*in_f32)) | |
887 | return -EFSCORRUPTED; | |
888 | ||
889 | in_f->ilf_type = in_f32->ilf_type; | |
890 | in_f->ilf_size = in_f32->ilf_size; | |
891 | in_f->ilf_fields = in_f32->ilf_fields; | |
892 | in_f->ilf_asize = in_f32->ilf_asize; | |
893 | in_f->ilf_dsize = in_f32->ilf_dsize; | |
894 | in_f->ilf_ino = in_f32->ilf_ino; | |
895 | /* copy biggest field of ilf_u */ | |
896 | uuid_copy(&in_f->ilf_u.ilfu_uuid, &in_f32->ilf_u.ilfu_uuid); | |
897 | in_f->ilf_blkno = in_f32->ilf_blkno; | |
898 | in_f->ilf_len = in_f32->ilf_len; | |
899 | in_f->ilf_boffset = in_f32->ilf_boffset; | |
900 | return 0; | |
6d192a9b | 901 | } |