]>
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 | 32 | |
f0e28280 | 33 | #include <linux/iversion.h> |
1da177e4 LT |
34 | |
35 | kmem_zone_t *xfs_ili_zone; /* inode log item zone */ | |
36 | ||
7bfa31d8 CH |
37 | static inline struct xfs_inode_log_item *INODE_ITEM(struct xfs_log_item *lip) |
38 | { | |
39 | return container_of(lip, struct xfs_inode_log_item, ili_item); | |
40 | } | |
41 | ||
166d1368 | 42 | STATIC void |
ce9641d6 CH |
43 | xfs_inode_item_data_fork_size( |
44 | struct xfs_inode_log_item *iip, | |
166d1368 DC |
45 | int *nvecs, |
46 | int *nbytes) | |
1da177e4 | 47 | { |
7bfa31d8 | 48 | struct xfs_inode *ip = iip->ili_inode; |
166d1368 | 49 | |
1da177e4 LT |
50 | switch (ip->i_d.di_format) { |
51 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 52 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
339a5f5d | 53 | ip->i_d.di_nextents > 0 && |
166d1368 DC |
54 | ip->i_df.if_bytes > 0) { |
55 | /* worst case, doesn't subtract delalloc extents */ | |
56 | *nbytes += XFS_IFORK_DSIZE(ip); | |
57 | *nvecs += 1; | |
58 | } | |
1da177e4 | 59 | break; |
1da177e4 | 60 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 61 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
166d1368 DC |
62 | ip->i_df.if_broot_bytes > 0) { |
63 | *nbytes += ip->i_df.if_broot_bytes; | |
64 | *nvecs += 1; | |
65 | } | |
1da177e4 | 66 | break; |
1da177e4 | 67 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 68 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
166d1368 DC |
69 | ip->i_df.if_bytes > 0) { |
70 | *nbytes += roundup(ip->i_df.if_bytes, 4); | |
71 | *nvecs += 1; | |
72 | } | |
1da177e4 LT |
73 | break; |
74 | ||
75 | case XFS_DINODE_FMT_DEV: | |
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 &= |
42b67dc6 | 159 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | XFS_ILOG_DEV); |
339a5f5d | 160 | |
f5d8d5c4 | 161 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
339a5f5d CH |
162 | ip->i_d.di_nextents > 0 && |
163 | ip->i_df.if_bytes > 0) { | |
da776503 CH |
164 | struct xfs_bmbt_rec *p; |
165 | ||
5d829300 | 166 | ASSERT(xfs_iext_count(&ip->i_df) > 0); |
da776503 CH |
167 | |
168 | p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IEXT); | |
169 | data_bytes = xfs_iextents_copy(ip, p, XFS_DATA_FORK); | |
170 | xlog_finish_iovec(lv, *vecp, data_bytes); | |
171 | ||
172 | ASSERT(data_bytes <= ip->i_df.if_bytes); | |
173 | ||
174 | ilf->ilf_dsize = data_bytes; | |
bde7cff6 | 175 | ilf->ilf_size++; |
339a5f5d | 176 | } else { |
f5d8d5c4 | 177 | iip->ili_fields &= ~XFS_ILOG_DEXT; |
1da177e4 LT |
178 | } |
179 | break; | |
1da177e4 | 180 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 181 | iip->ili_fields &= |
42b67dc6 | 182 | ~(XFS_ILOG_DDATA | XFS_ILOG_DEXT | XFS_ILOG_DEV); |
339a5f5d | 183 | |
f5d8d5c4 | 184 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
339a5f5d | 185 | ip->i_df.if_broot_bytes > 0) { |
1da177e4 | 186 | ASSERT(ip->i_df.if_broot != NULL); |
bde7cff6 | 187 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IBROOT, |
1234351c CH |
188 | ip->i_df.if_broot, |
189 | ip->i_df.if_broot_bytes); | |
bde7cff6 CH |
190 | ilf->ilf_dsize = ip->i_df.if_broot_bytes; |
191 | ilf->ilf_size++; | |
339a5f5d | 192 | } else { |
f5d8d5c4 | 193 | ASSERT(!(iip->ili_fields & |
339a5f5d | 194 | XFS_ILOG_DBROOT)); |
f5d8d5c4 | 195 | iip->ili_fields &= ~XFS_ILOG_DBROOT; |
1da177e4 LT |
196 | } |
197 | break; | |
1da177e4 | 198 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 199 | iip->ili_fields &= |
42b67dc6 | 200 | ~(XFS_ILOG_DEXT | XFS_ILOG_DBROOT | XFS_ILOG_DEV); |
f5d8d5c4 | 201 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
339a5f5d | 202 | ip->i_df.if_bytes > 0) { |
1da177e4 LT |
203 | /* |
204 | * Round i_bytes up to a word boundary. | |
205 | * The underlying memory is guaranteed to | |
206 | * to be there by xfs_idata_realloc(). | |
207 | */ | |
208 | data_bytes = roundup(ip->i_df.if_bytes, 4); | |
1234351c | 209 | ASSERT(ip->i_df.if_real_bytes == 0 || |
30ee052e | 210 | ip->i_df.if_real_bytes >= data_bytes); |
1234351c CH |
211 | ASSERT(ip->i_df.if_u1.if_data != NULL); |
212 | ASSERT(ip->i_d.di_size > 0); | |
bde7cff6 | 213 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_ILOCAL, |
1234351c | 214 | ip->i_df.if_u1.if_data, data_bytes); |
bde7cff6 CH |
215 | ilf->ilf_dsize = (unsigned)data_bytes; |
216 | ilf->ilf_size++; | |
339a5f5d | 217 | } else { |
f5d8d5c4 | 218 | iip->ili_fields &= ~XFS_ILOG_DDATA; |
1da177e4 LT |
219 | } |
220 | break; | |
1da177e4 | 221 | case XFS_DINODE_FMT_DEV: |
f5d8d5c4 | 222 | iip->ili_fields &= |
42b67dc6 | 223 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | XFS_ILOG_DEXT); |
bde7cff6 | 224 | if (iip->ili_fields & XFS_ILOG_DEV) |
66f36464 | 225 | ilf->ilf_u.ilfu_rdev = sysv_encode_dev(VFS_I(ip)->i_rdev); |
1da177e4 | 226 | break; |
1da177e4 LT |
227 | default: |
228 | ASSERT(0); | |
229 | break; | |
230 | } | |
3de559fb CH |
231 | } |
232 | ||
1234351c | 233 | STATIC void |
3de559fb CH |
234 | xfs_inode_item_format_attr_fork( |
235 | struct xfs_inode_log_item *iip, | |
bde7cff6 CH |
236 | struct xfs_inode_log_format *ilf, |
237 | struct xfs_log_vec *lv, | |
238 | struct xfs_log_iovec **vecp) | |
3de559fb CH |
239 | { |
240 | struct xfs_inode *ip = iip->ili_inode; | |
241 | size_t data_bytes; | |
1da177e4 LT |
242 | |
243 | switch (ip->i_d.di_aformat) { | |
244 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 245 | iip->ili_fields &= |
339a5f5d CH |
246 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT); |
247 | ||
f5d8d5c4 | 248 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
339a5f5d CH |
249 | ip->i_d.di_anextents > 0 && |
250 | ip->i_afp->if_bytes > 0) { | |
da776503 CH |
251 | struct xfs_bmbt_rec *p; |
252 | ||
5d829300 | 253 | ASSERT(xfs_iext_count(ip->i_afp) == |
339a5f5d | 254 | ip->i_d.di_anextents); |
da776503 CH |
255 | |
256 | p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_EXT); | |
257 | data_bytes = xfs_iextents_copy(ip, p, XFS_ATTR_FORK); | |
258 | xlog_finish_iovec(lv, *vecp, data_bytes); | |
259 | ||
260 | ilf->ilf_asize = data_bytes; | |
bde7cff6 | 261 | ilf->ilf_size++; |
339a5f5d | 262 | } else { |
f5d8d5c4 | 263 | iip->ili_fields &= ~XFS_ILOG_AEXT; |
1da177e4 LT |
264 | } |
265 | break; | |
1da177e4 | 266 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 267 | iip->ili_fields &= |
339a5f5d CH |
268 | ~(XFS_ILOG_ADATA | XFS_ILOG_AEXT); |
269 | ||
f5d8d5c4 | 270 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
339a5f5d | 271 | ip->i_afp->if_broot_bytes > 0) { |
1da177e4 | 272 | ASSERT(ip->i_afp->if_broot != NULL); |
339a5f5d | 273 | |
bde7cff6 | 274 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_BROOT, |
1234351c CH |
275 | ip->i_afp->if_broot, |
276 | ip->i_afp->if_broot_bytes); | |
bde7cff6 CH |
277 | ilf->ilf_asize = ip->i_afp->if_broot_bytes; |
278 | ilf->ilf_size++; | |
339a5f5d | 279 | } else { |
f5d8d5c4 | 280 | iip->ili_fields &= ~XFS_ILOG_ABROOT; |
1da177e4 LT |
281 | } |
282 | break; | |
1da177e4 | 283 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 284 | iip->ili_fields &= |
339a5f5d CH |
285 | ~(XFS_ILOG_AEXT | XFS_ILOG_ABROOT); |
286 | ||
f5d8d5c4 | 287 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
339a5f5d | 288 | ip->i_afp->if_bytes > 0) { |
1da177e4 LT |
289 | /* |
290 | * Round i_bytes up to a word boundary. | |
291 | * The underlying memory is guaranteed to | |
292 | * to be there by xfs_idata_realloc(). | |
293 | */ | |
294 | data_bytes = roundup(ip->i_afp->if_bytes, 4); | |
1234351c | 295 | ASSERT(ip->i_afp->if_real_bytes == 0 || |
30ee052e | 296 | ip->i_afp->if_real_bytes >= data_bytes); |
1234351c | 297 | ASSERT(ip->i_afp->if_u1.if_data != NULL); |
bde7cff6 | 298 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_LOCAL, |
1234351c CH |
299 | ip->i_afp->if_u1.if_data, |
300 | data_bytes); | |
bde7cff6 CH |
301 | ilf->ilf_asize = (unsigned)data_bytes; |
302 | ilf->ilf_size++; | |
339a5f5d | 303 | } else { |
f5d8d5c4 | 304 | iip->ili_fields &= ~XFS_ILOG_ADATA; |
1da177e4 LT |
305 | } |
306 | break; | |
1da177e4 LT |
307 | default: |
308 | ASSERT(0); | |
309 | break; | |
310 | } | |
3de559fb CH |
311 | } |
312 | ||
f8d55aa0 | 313 | static void |
3987848c DC |
314 | xfs_inode_to_log_dinode( |
315 | struct xfs_inode *ip, | |
93f958f9 DC |
316 | struct xfs_log_dinode *to, |
317 | xfs_lsn_t lsn) | |
f8d55aa0 | 318 | { |
3987848c DC |
319 | struct xfs_icdinode *from = &ip->i_d; |
320 | struct inode *inode = VFS_I(ip); | |
321 | ||
93f958f9 DC |
322 | to->di_magic = XFS_DINODE_MAGIC; |
323 | ||
f8d55aa0 DC |
324 | to->di_version = from->di_version; |
325 | to->di_format = from->di_format; | |
f8d55aa0 DC |
326 | to->di_uid = from->di_uid; |
327 | to->di_gid = from->di_gid; | |
f8d55aa0 DC |
328 | to->di_projid_lo = from->di_projid_lo; |
329 | to->di_projid_hi = from->di_projid_hi; | |
f8d55aa0 | 330 | |
93f958f9 | 331 | memset(to->di_pad, 0, sizeof(to->di_pad)); |
faeb4e47 | 332 | memset(to->di_pad3, 0, sizeof(to->di_pad3)); |
3987848c DC |
333 | to->di_atime.t_sec = inode->i_atime.tv_sec; |
334 | to->di_atime.t_nsec = inode->i_atime.tv_nsec; | |
335 | to->di_mtime.t_sec = inode->i_mtime.tv_sec; | |
336 | to->di_mtime.t_nsec = inode->i_mtime.tv_nsec; | |
337 | to->di_ctime.t_sec = inode->i_ctime.tv_sec; | |
338 | to->di_ctime.t_nsec = inode->i_ctime.tv_nsec; | |
54d7b5c1 | 339 | to->di_nlink = inode->i_nlink; |
9e9a2674 | 340 | to->di_gen = inode->i_generation; |
c19b3b05 | 341 | to->di_mode = inode->i_mode; |
f8d55aa0 | 342 | |
f8d55aa0 DC |
343 | to->di_size = from->di_size; |
344 | to->di_nblocks = from->di_nblocks; | |
345 | to->di_extsize = from->di_extsize; | |
346 | to->di_nextents = from->di_nextents; | |
347 | to->di_anextents = from->di_anextents; | |
348 | to->di_forkoff = from->di_forkoff; | |
349 | to->di_aformat = from->di_aformat; | |
350 | to->di_dmevmask = from->di_dmevmask; | |
351 | to->di_dmstate = from->di_dmstate; | |
352 | to->di_flags = from->di_flags; | |
f8d55aa0 | 353 | |
20413e37 DC |
354 | /* log a dummy value to ensure log structure is fully initialised */ |
355 | to->di_next_unlinked = NULLAGINO; | |
356 | ||
f8d55aa0 | 357 | if (from->di_version == 3) { |
f0e28280 | 358 | to->di_changecount = inode_peek_iversion(inode); |
f8d55aa0 DC |
359 | to->di_crtime.t_sec = from->di_crtime.t_sec; |
360 | to->di_crtime.t_nsec = from->di_crtime.t_nsec; | |
361 | to->di_flags2 = from->di_flags2; | |
f7ca3522 | 362 | to->di_cowextsize = from->di_cowextsize; |
93f958f9 DC |
363 | to->di_ino = ip->i_ino; |
364 | to->di_lsn = lsn; | |
365 | memset(to->di_pad2, 0, sizeof(to->di_pad2)); | |
366 | uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid); | |
f8d55aa0 DC |
367 | to->di_flushiter = 0; |
368 | } else { | |
369 | to->di_flushiter = from->di_flushiter; | |
370 | } | |
371 | } | |
372 | ||
373 | /* | |
374 | * Format the inode core. Current timestamp data is only in the VFS inode | |
375 | * fields, so we need to grab them from there. Hence rather than just copying | |
376 | * the XFS inode core structure, format the fields directly into the iovec. | |
377 | */ | |
378 | static void | |
379 | xfs_inode_item_format_core( | |
380 | struct xfs_inode *ip, | |
381 | struct xfs_log_vec *lv, | |
382 | struct xfs_log_iovec **vecp) | |
383 | { | |
384 | struct xfs_log_dinode *dic; | |
385 | ||
386 | dic = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_ICORE); | |
93f958f9 | 387 | xfs_inode_to_log_dinode(ip, dic, ip->i_itemp->ili_item.li_lsn); |
f8d55aa0 DC |
388 | xlog_finish_iovec(lv, *vecp, xfs_log_dinode_size(ip->i_d.di_version)); |
389 | } | |
390 | ||
3de559fb CH |
391 | /* |
392 | * This is called to fill in the vector of log iovecs for the given inode | |
393 | * log item. It fills the first item with an inode log format structure, | |
394 | * the second with the on-disk inode structure, and a possible third and/or | |
395 | * fourth with the inode data/extents/b-tree root and inode attributes | |
396 | * data/extents/b-tree root. | |
20413e37 DC |
397 | * |
398 | * Note: Always use the 64 bit inode log format structure so we don't | |
399 | * leave an uninitialised hole in the format item on 64 bit systems. Log | |
400 | * recovery on 32 bit systems handles this just fine, so there's no reason | |
401 | * for not using an initialising the properly padded structure all the time. | |
3de559fb CH |
402 | */ |
403 | STATIC void | |
404 | xfs_inode_item_format( | |
405 | struct xfs_log_item *lip, | |
bde7cff6 | 406 | struct xfs_log_vec *lv) |
3de559fb CH |
407 | { |
408 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
409 | struct xfs_inode *ip = iip->ili_inode; | |
bde7cff6 | 410 | struct xfs_log_iovec *vecp = NULL; |
20413e37 | 411 | struct xfs_inode_log_format *ilf; |
3de559fb | 412 | |
263997a6 DC |
413 | ASSERT(ip->i_d.di_version > 1); |
414 | ||
2f251293 CH |
415 | ilf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_IFORMAT); |
416 | ilf->ilf_type = XFS_LI_INODE; | |
417 | ilf->ilf_ino = ip->i_ino; | |
418 | ilf->ilf_blkno = ip->i_imap.im_blkno; | |
419 | ilf->ilf_len = ip->i_imap.im_len; | |
420 | ilf->ilf_boffset = ip->i_imap.im_boffset; | |
421 | ilf->ilf_fields = XFS_ILOG_CORE; | |
422 | ilf->ilf_size = 2; /* format + core */ | |
20413e37 DC |
423 | |
424 | /* | |
425 | * make sure we don't leak uninitialised data into the log in the case | |
426 | * when we don't log every field in the inode. | |
427 | */ | |
428 | ilf->ilf_dsize = 0; | |
429 | ilf->ilf_asize = 0; | |
430 | ilf->ilf_pad = 0; | |
42b67dc6 | 431 | memset(&ilf->ilf_u, 0, sizeof(ilf->ilf_u)); |
20413e37 DC |
432 | |
433 | xlog_finish_iovec(lv, vecp, sizeof(*ilf)); | |
3de559fb | 434 | |
f8d55aa0 | 435 | xfs_inode_item_format_core(ip, lv, &vecp); |
bde7cff6 | 436 | xfs_inode_item_format_data_fork(iip, ilf, lv, &vecp); |
3de559fb | 437 | if (XFS_IFORK_Q(ip)) { |
bde7cff6 | 438 | xfs_inode_item_format_attr_fork(iip, ilf, lv, &vecp); |
3de559fb CH |
439 | } else { |
440 | iip->ili_fields &= | |
441 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT); | |
442 | } | |
443 | ||
2f251293 CH |
444 | /* update the format with the exact fields we actually logged */ |
445 | ilf->ilf_fields |= (iip->ili_fields & ~XFS_ILOG_TIMESTAMP); | |
1da177e4 LT |
446 | } |
447 | ||
1da177e4 LT |
448 | /* |
449 | * This is called to pin the inode associated with the inode log | |
a14a5ab5 | 450 | * item in memory so it cannot be written out. |
1da177e4 LT |
451 | */ |
452 | STATIC void | |
453 | xfs_inode_item_pin( | |
7bfa31d8 | 454 | struct xfs_log_item *lip) |
1da177e4 | 455 | { |
7bfa31d8 | 456 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 457 | |
7bfa31d8 CH |
458 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); |
459 | ||
460 | trace_xfs_inode_pin(ip, _RET_IP_); | |
461 | atomic_inc(&ip->i_pincount); | |
1da177e4 LT |
462 | } |
463 | ||
464 | ||
465 | /* | |
466 | * This is called to unpin the inode associated with the inode log | |
467 | * item which was previously pinned with a call to xfs_inode_item_pin(). | |
a14a5ab5 CH |
468 | * |
469 | * Also wake up anyone in xfs_iunpin_wait() if the count goes to 0. | |
1da177e4 | 470 | */ |
1da177e4 LT |
471 | STATIC void |
472 | xfs_inode_item_unpin( | |
7bfa31d8 | 473 | struct xfs_log_item *lip, |
9412e318 | 474 | int remove) |
1da177e4 | 475 | { |
7bfa31d8 | 476 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 477 | |
4aaf15d1 | 478 | trace_xfs_inode_unpin(ip, _RET_IP_); |
a14a5ab5 CH |
479 | ASSERT(atomic_read(&ip->i_pincount) > 0); |
480 | if (atomic_dec_and_test(&ip->i_pincount)) | |
f392e631 | 481 | wake_up_bit(&ip->i_flags, __XFS_IPINNED_BIT); |
1da177e4 LT |
482 | } |
483 | ||
d3a304b6 CM |
484 | /* |
485 | * Callback used to mark a buffer with XFS_LI_FAILED when items in the buffer | |
486 | * have been failed during writeback | |
487 | * | |
488 | * This informs the AIL that the inode is already flush locked on the next push, | |
489 | * and acquires a hold on the buffer to ensure that it isn't reclaimed before | |
490 | * dirty data makes it to disk. | |
491 | */ | |
492 | STATIC void | |
493 | xfs_inode_item_error( | |
494 | struct xfs_log_item *lip, | |
495 | struct xfs_buf *bp) | |
496 | { | |
497 | ASSERT(xfs_isiflocked(INODE_ITEM(lip)->ili_inode)); | |
498 | xfs_set_li_failed(lip, bp); | |
499 | } | |
500 | ||
1da177e4 | 501 | STATIC uint |
43ff2122 CH |
502 | xfs_inode_item_push( |
503 | struct xfs_log_item *lip, | |
504 | struct list_head *buffer_list) | |
6e3e6d55 EG |
505 | __releases(&lip->li_ailp->xa_lock) |
506 | __acquires(&lip->li_ailp->xa_lock) | |
1da177e4 | 507 | { |
7bfa31d8 CH |
508 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
509 | struct xfs_inode *ip = iip->ili_inode; | |
d3a304b6 | 510 | struct xfs_buf *bp = lip->li_buf; |
43ff2122 CH |
511 | uint rval = XFS_ITEM_SUCCESS; |
512 | int error; | |
1da177e4 | 513 | |
7bfa31d8 | 514 | if (xfs_ipincount(ip) > 0) |
1da177e4 | 515 | return XFS_ITEM_PINNED; |
1da177e4 | 516 | |
d3a304b6 CM |
517 | /* |
518 | * The buffer containing this item failed to be written back | |
519 | * previously. Resubmit the buffer for IO. | |
520 | */ | |
521 | if (lip->li_flags & XFS_LI_FAILED) { | |
522 | if (!xfs_buf_trylock(bp)) | |
523 | return XFS_ITEM_LOCKED; | |
524 | ||
525 | if (!xfs_buf_resubmit_failed_buffers(bp, lip, buffer_list)) | |
526 | rval = XFS_ITEM_FLUSHING; | |
527 | ||
528 | xfs_buf_unlock(bp); | |
529 | return rval; | |
530 | } | |
531 | ||
7bfa31d8 | 532 | if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED)) |
1da177e4 | 533 | return XFS_ITEM_LOCKED; |
1da177e4 | 534 | |
4c46819a CH |
535 | /* |
536 | * Re-check the pincount now that we stabilized the value by | |
537 | * taking the ilock. | |
538 | */ | |
539 | if (xfs_ipincount(ip) > 0) { | |
43ff2122 CH |
540 | rval = XFS_ITEM_PINNED; |
541 | goto out_unlock; | |
4c46819a CH |
542 | } |
543 | ||
9a3a5dab BF |
544 | /* |
545 | * Stale inode items should force out the iclog. | |
546 | */ | |
547 | if (ip->i_flags & XFS_ISTALE) { | |
548 | rval = XFS_ITEM_PINNED; | |
549 | goto out_unlock; | |
550 | } | |
551 | ||
43ff2122 CH |
552 | /* |
553 | * Someone else is already flushing the inode. Nothing we can do | |
554 | * here but wait for the flush to finish and remove the item from | |
555 | * the AIL. | |
556 | */ | |
1da177e4 | 557 | if (!xfs_iflock_nowait(ip)) { |
43ff2122 CH |
558 | rval = XFS_ITEM_FLUSHING; |
559 | goto out_unlock; | |
1da177e4 LT |
560 | } |
561 | ||
43ff2122 CH |
562 | ASSERT(iip->ili_fields != 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); |
563 | ASSERT(iip->ili_logged == 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); | |
564 | ||
565 | spin_unlock(&lip->li_ailp->xa_lock); | |
566 | ||
567 | error = xfs_iflush(ip, &bp); | |
568 | if (!error) { | |
569 | if (!xfs_buf_delwri_queue(bp, buffer_list)) | |
570 | rval = XFS_ITEM_FLUSHING; | |
571 | xfs_buf_relse(bp); | |
1da177e4 | 572 | } |
43ff2122 CH |
573 | |
574 | spin_lock(&lip->li_ailp->xa_lock); | |
575 | out_unlock: | |
576 | xfs_iunlock(ip, XFS_ILOCK_SHARED); | |
577 | return rval; | |
1da177e4 LT |
578 | } |
579 | ||
580 | /* | |
581 | * Unlock the inode associated with the inode log item. | |
582 | * Clear the fields of the inode and inode log item that | |
583 | * are specific to the current transaction. If the | |
584 | * hold flags is set, do not unlock the inode. | |
585 | */ | |
586 | STATIC void | |
587 | xfs_inode_item_unlock( | |
7bfa31d8 | 588 | struct xfs_log_item *lip) |
1da177e4 | 589 | { |
7bfa31d8 CH |
590 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
591 | struct xfs_inode *ip = iip->ili_inode; | |
898621d5 | 592 | unsigned short lock_flags; |
1da177e4 | 593 | |
f3ca8738 CH |
594 | ASSERT(ip->i_itemp != NULL); |
595 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); | |
1da177e4 | 596 | |
898621d5 CH |
597 | lock_flags = iip->ili_lock_flags; |
598 | iip->ili_lock_flags = 0; | |
ddc3415a | 599 | if (lock_flags) |
f3ca8738 | 600 | xfs_iunlock(ip, lock_flags); |
1da177e4 LT |
601 | } |
602 | ||
603 | /* | |
de25c181 DC |
604 | * This is called to find out where the oldest active copy of the inode log |
605 | * item in the on disk log resides now that the last log write of it completed | |
606 | * at the given lsn. Since we always re-log all dirty data in an inode, the | |
607 | * latest copy in the on disk log is the only one that matters. Therefore, | |
608 | * simply return the given lsn. | |
609 | * | |
610 | * If the inode has been marked stale because the cluster is being freed, we | |
611 | * don't want to (re-)insert this inode into the AIL. There is a race condition | |
612 | * where the cluster buffer may be unpinned before the inode is inserted into | |
613 | * the AIL during transaction committed processing. If the buffer is unpinned | |
614 | * before the inode item has been committed and inserted, then it is possible | |
1316d4da | 615 | * for the buffer to be written and IO completes before the inode is inserted |
de25c181 DC |
616 | * into the AIL. In that case, we'd be inserting a clean, stale inode into the |
617 | * AIL which will never get removed. It will, however, get reclaimed which | |
618 | * triggers an assert in xfs_inode_free() complaining about freein an inode | |
619 | * still in the AIL. | |
620 | * | |
1316d4da DC |
621 | * To avoid this, just unpin the inode directly and return a LSN of -1 so the |
622 | * transaction committed code knows that it does not need to do any further | |
623 | * processing on the item. | |
1da177e4 | 624 | */ |
1da177e4 LT |
625 | STATIC xfs_lsn_t |
626 | xfs_inode_item_committed( | |
7bfa31d8 | 627 | struct xfs_log_item *lip, |
1da177e4 LT |
628 | xfs_lsn_t lsn) |
629 | { | |
de25c181 DC |
630 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
631 | struct xfs_inode *ip = iip->ili_inode; | |
632 | ||
1316d4da DC |
633 | if (xfs_iflags_test(ip, XFS_ISTALE)) { |
634 | xfs_inode_item_unpin(lip, 0); | |
635 | return -1; | |
636 | } | |
7bfa31d8 | 637 | return lsn; |
1da177e4 LT |
638 | } |
639 | ||
1da177e4 LT |
640 | /* |
641 | * XXX rcc - this one really has to do something. Probably needs | |
642 | * to stamp in a new field in the incore inode. | |
643 | */ | |
1da177e4 LT |
644 | STATIC void |
645 | xfs_inode_item_committing( | |
7bfa31d8 | 646 | struct xfs_log_item *lip, |
1da177e4 LT |
647 | xfs_lsn_t lsn) |
648 | { | |
7bfa31d8 | 649 | INODE_ITEM(lip)->ili_last_lsn = lsn; |
1da177e4 LT |
650 | } |
651 | ||
652 | /* | |
653 | * This is the ops vector shared by all buf log items. | |
654 | */ | |
272e42b2 | 655 | static const struct xfs_item_ops xfs_inode_item_ops = { |
7bfa31d8 CH |
656 | .iop_size = xfs_inode_item_size, |
657 | .iop_format = xfs_inode_item_format, | |
658 | .iop_pin = xfs_inode_item_pin, | |
659 | .iop_unpin = xfs_inode_item_unpin, | |
7bfa31d8 CH |
660 | .iop_unlock = xfs_inode_item_unlock, |
661 | .iop_committed = xfs_inode_item_committed, | |
662 | .iop_push = xfs_inode_item_push, | |
d3a304b6 CM |
663 | .iop_committing = xfs_inode_item_committing, |
664 | .iop_error = xfs_inode_item_error | |
1da177e4 LT |
665 | }; |
666 | ||
667 | ||
668 | /* | |
669 | * Initialize the inode log item for a newly allocated (in-core) inode. | |
670 | */ | |
671 | void | |
672 | xfs_inode_item_init( | |
7bfa31d8 CH |
673 | struct xfs_inode *ip, |
674 | struct xfs_mount *mp) | |
1da177e4 | 675 | { |
7bfa31d8 | 676 | struct xfs_inode_log_item *iip; |
1da177e4 LT |
677 | |
678 | ASSERT(ip->i_itemp == NULL); | |
679 | iip = ip->i_itemp = kmem_zone_zalloc(xfs_ili_zone, KM_SLEEP); | |
680 | ||
1da177e4 | 681 | iip->ili_inode = ip; |
43f5efc5 DC |
682 | xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE, |
683 | &xfs_inode_item_ops); | |
1da177e4 LT |
684 | } |
685 | ||
686 | /* | |
687 | * Free the inode log item and any memory hanging off of it. | |
688 | */ | |
689 | void | |
690 | xfs_inode_item_destroy( | |
691 | xfs_inode_t *ip) | |
692 | { | |
b1c5ebb2 | 693 | kmem_free(ip->i_itemp->ili_item.li_lv_shadow); |
1da177e4 LT |
694 | kmem_zone_free(xfs_ili_zone, ip->i_itemp); |
695 | } | |
696 | ||
697 | ||
698 | /* | |
699 | * This is the inode flushing I/O completion routine. It is called | |
700 | * from interrupt level when the buffer containing the inode is | |
701 | * flushed to disk. It is responsible for removing the inode item | |
702 | * from the AIL if it has not been re-logged, and unlocking the inode's | |
703 | * flush lock. | |
30136832 DC |
704 | * |
705 | * To reduce AIL lock traffic as much as possible, we scan the buffer log item | |
706 | * list for other inodes that will run this function. We remove them from the | |
707 | * buffer list so we can process all the inode IO completions in one AIL lock | |
708 | * traversal. | |
1da177e4 | 709 | */ |
1da177e4 LT |
710 | void |
711 | xfs_iflush_done( | |
ca30b2a7 CH |
712 | struct xfs_buf *bp, |
713 | struct xfs_log_item *lip) | |
1da177e4 | 714 | { |
30136832 DC |
715 | struct xfs_inode_log_item *iip; |
716 | struct xfs_log_item *blip; | |
717 | struct xfs_log_item *next; | |
718 | struct xfs_log_item *prev; | |
ca30b2a7 | 719 | struct xfs_ail *ailp = lip->li_ailp; |
30136832 DC |
720 | int need_ail = 0; |
721 | ||
722 | /* | |
723 | * Scan the buffer IO completions for other inodes being completed and | |
724 | * attach them to the current inode log item. | |
725 | */ | |
adadbeef | 726 | blip = bp->b_fspriv; |
30136832 DC |
727 | prev = NULL; |
728 | while (blip != NULL) { | |
52177937 | 729 | if (blip->li_cb != xfs_iflush_done) { |
30136832 DC |
730 | prev = blip; |
731 | blip = blip->li_bio_list; | |
732 | continue; | |
733 | } | |
734 | ||
735 | /* remove from list */ | |
736 | next = blip->li_bio_list; | |
737 | if (!prev) { | |
adadbeef | 738 | bp->b_fspriv = next; |
30136832 DC |
739 | } else { |
740 | prev->li_bio_list = next; | |
741 | } | |
742 | ||
743 | /* add to current list */ | |
744 | blip->li_bio_list = lip->li_bio_list; | |
745 | lip->li_bio_list = blip; | |
746 | ||
747 | /* | |
748 | * while we have the item, do the unlocked check for needing | |
749 | * the AIL lock. | |
750 | */ | |
751 | iip = INODE_ITEM(blip); | |
d3a304b6 | 752 | if ((iip->ili_logged && blip->li_lsn == iip->ili_flush_lsn) || |
842f6e9f | 753 | (blip->li_flags & XFS_LI_FAILED)) |
30136832 DC |
754 | need_ail++; |
755 | ||
756 | blip = next; | |
757 | } | |
758 | ||
759 | /* make sure we capture the state of the initial inode. */ | |
760 | iip = INODE_ITEM(lip); | |
d3a304b6 CM |
761 | if ((iip->ili_logged && lip->li_lsn == iip->ili_flush_lsn) || |
762 | lip->li_flags & XFS_LI_FAILED) | |
30136832 | 763 | need_ail++; |
1da177e4 LT |
764 | |
765 | /* | |
766 | * We only want to pull the item from the AIL if it is | |
767 | * actually there and its location in the log has not | |
768 | * changed since we started the flush. Thus, we only bother | |
769 | * if the ili_logged flag is set and the inode's lsn has not | |
770 | * changed. First we check the lsn outside | |
771 | * the lock since it's cheaper, and then we recheck while | |
772 | * holding the lock before removing the inode from the AIL. | |
773 | */ | |
30136832 | 774 | if (need_ail) { |
27af1bbf CH |
775 | bool mlip_changed = false; |
776 | ||
777 | /* this is an opencoded batch version of xfs_trans_ail_delete */ | |
783a2f65 | 778 | spin_lock(&ailp->xa_lock); |
30136832 | 779 | for (blip = lip; blip; blip = blip->li_bio_list) { |
27af1bbf CH |
780 | if (INODE_ITEM(blip)->ili_logged && |
781 | blip->li_lsn == INODE_ITEM(blip)->ili_flush_lsn) | |
782 | mlip_changed |= xfs_ail_delete_one(ailp, blip); | |
d3a304b6 CM |
783 | else { |
784 | xfs_clear_li_failed(blip); | |
785 | } | |
1da177e4 | 786 | } |
1da177e4 | 787 | |
27af1bbf CH |
788 | if (mlip_changed) { |
789 | if (!XFS_FORCED_SHUTDOWN(ailp->xa_mount)) | |
790 | xlog_assign_tail_lsn_locked(ailp->xa_mount); | |
791 | if (list_empty(&ailp->xa_ail)) | |
792 | wake_up_all(&ailp->xa_empty); | |
793 | } | |
794 | spin_unlock(&ailp->xa_lock); | |
795 | ||
796 | if (mlip_changed) | |
797 | xfs_log_space_wake(ailp->xa_mount); | |
798 | } | |
1da177e4 LT |
799 | |
800 | /* | |
30136832 DC |
801 | * clean up and unlock the flush lock now we are done. We can clear the |
802 | * ili_last_fields bits now that we know that the data corresponding to | |
803 | * them is safely on disk. | |
1da177e4 | 804 | */ |
30136832 DC |
805 | for (blip = lip; blip; blip = next) { |
806 | next = blip->li_bio_list; | |
807 | blip->li_bio_list = NULL; | |
808 | ||
809 | iip = INODE_ITEM(blip); | |
810 | iip->ili_logged = 0; | |
811 | iip->ili_last_fields = 0; | |
812 | xfs_ifunlock(iip->ili_inode); | |
813 | } | |
1da177e4 LT |
814 | } |
815 | ||
816 | /* | |
04913fdd DC |
817 | * This is the inode flushing abort routine. It is called from xfs_iflush when |
818 | * the filesystem is shutting down to clean up the inode state. It is | |
819 | * responsible for removing the inode item from the AIL if it has not been | |
820 | * re-logged, and unlocking the inode's flush lock. | |
1da177e4 LT |
821 | */ |
822 | void | |
823 | xfs_iflush_abort( | |
04913fdd DC |
824 | xfs_inode_t *ip, |
825 | bool stale) | |
1da177e4 | 826 | { |
783a2f65 | 827 | xfs_inode_log_item_t *iip = ip->i_itemp; |
1da177e4 | 828 | |
1da177e4 LT |
829 | if (iip) { |
830 | if (iip->ili_item.li_flags & XFS_LI_IN_AIL) { | |
146e54b7 BF |
831 | xfs_trans_ail_remove(&iip->ili_item, |
832 | stale ? SHUTDOWN_LOG_IO_ERROR : | |
04913fdd | 833 | SHUTDOWN_CORRUPT_INCORE); |
1da177e4 LT |
834 | } |
835 | iip->ili_logged = 0; | |
836 | /* | |
837 | * Clear the ili_last_fields bits now that we know that the | |
838 | * data corresponding to them is safely on disk. | |
839 | */ | |
840 | iip->ili_last_fields = 0; | |
841 | /* | |
842 | * Clear the inode logging fields so no more flushes are | |
843 | * attempted. | |
844 | */ | |
f5d8d5c4 | 845 | iip->ili_fields = 0; |
fc0561ce | 846 | iip->ili_fsync_fields = 0; |
1da177e4 LT |
847 | } |
848 | /* | |
849 | * Release the inode's flush lock since we're done with it. | |
850 | */ | |
851 | xfs_ifunlock(ip); | |
852 | } | |
853 | ||
854 | void | |
855 | xfs_istale_done( | |
ca30b2a7 CH |
856 | struct xfs_buf *bp, |
857 | struct xfs_log_item *lip) | |
1da177e4 | 858 | { |
04913fdd | 859 | xfs_iflush_abort(INODE_ITEM(lip)->ili_inode, true); |
1da177e4 | 860 | } |
6d192a9b TS |
861 | |
862 | /* | |
20413e37 DC |
863 | * convert an xfs_inode_log_format struct from the old 32 bit version |
864 | * (which can have different field alignments) to the native 64 bit version | |
6d192a9b TS |
865 | */ |
866 | int | |
867 | xfs_inode_item_format_convert( | |
20413e37 DC |
868 | struct xfs_log_iovec *buf, |
869 | struct xfs_inode_log_format *in_f) | |
6d192a9b | 870 | { |
20413e37 DC |
871 | struct xfs_inode_log_format_32 *in_f32 = buf->i_addr; |
872 | ||
873 | if (buf->i_len != sizeof(*in_f32)) | |
874 | return -EFSCORRUPTED; | |
875 | ||
876 | in_f->ilf_type = in_f32->ilf_type; | |
877 | in_f->ilf_size = in_f32->ilf_size; | |
878 | in_f->ilf_fields = in_f32->ilf_fields; | |
879 | in_f->ilf_asize = in_f32->ilf_asize; | |
880 | in_f->ilf_dsize = in_f32->ilf_dsize; | |
881 | in_f->ilf_ino = in_f32->ilf_ino; | |
42b67dc6 | 882 | memcpy(&in_f->ilf_u, &in_f32->ilf_u, sizeof(in_f->ilf_u)); |
20413e37 DC |
883 | in_f->ilf_blkno = in_f32->ilf_blkno; |
884 | in_f->ilf_len = in_f32->ilf_len; | |
885 | in_f->ilf_boffset = in_f32->ilf_boffset; | |
886 | return 0; | |
6d192a9b | 887 | } |