]>
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_sb.h" |
a844f451 | 24 | #include "xfs_ag.h" |
1da177e4 | 25 | #include "xfs_mount.h" |
1da177e4 | 26 | #include "xfs_inode.h" |
239880ef | 27 | #include "xfs_trans.h" |
a844f451 | 28 | #include "xfs_inode_item.h" |
db7a19f2 | 29 | #include "xfs_error.h" |
0b1b213f | 30 | #include "xfs_trace.h" |
239880ef | 31 | #include "xfs_trans_priv.h" |
a4fbe6ab | 32 | #include "xfs_dinode.h" |
1da177e4 LT |
33 | |
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 | case XFS_DINODE_FMT_UUID: |
1da177e4 | 77 | break; |
1da177e4 LT |
78 | default: |
79 | ASSERT(0); | |
80 | break; | |
81 | } | |
ce9641d6 | 82 | } |
1da177e4 | 83 | |
ce9641d6 CH |
84 | STATIC void |
85 | xfs_inode_item_attr_fork_size( | |
86 | struct xfs_inode_log_item *iip, | |
87 | int *nvecs, | |
88 | int *nbytes) | |
89 | { | |
90 | struct xfs_inode *ip = iip->ili_inode; | |
1da177e4 | 91 | |
1da177e4 LT |
92 | switch (ip->i_d.di_aformat) { |
93 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 94 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
339a5f5d | 95 | ip->i_d.di_anextents > 0 && |
166d1368 DC |
96 | ip->i_afp->if_bytes > 0) { |
97 | /* worst case, doesn't subtract unused space */ | |
98 | *nbytes += XFS_IFORK_ASIZE(ip); | |
99 | *nvecs += 1; | |
100 | } | |
1da177e4 | 101 | break; |
1da177e4 | 102 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 103 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
166d1368 DC |
104 | ip->i_afp->if_broot_bytes > 0) { |
105 | *nbytes += ip->i_afp->if_broot_bytes; | |
106 | *nvecs += 1; | |
107 | } | |
1da177e4 | 108 | break; |
1da177e4 | 109 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 110 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
166d1368 DC |
111 | ip->i_afp->if_bytes > 0) { |
112 | *nbytes += roundup(ip->i_afp->if_bytes, 4); | |
113 | *nvecs += 1; | |
114 | } | |
1da177e4 | 115 | break; |
1da177e4 LT |
116 | default: |
117 | ASSERT(0); | |
118 | break; | |
119 | } | |
1da177e4 LT |
120 | } |
121 | ||
ce9641d6 CH |
122 | /* |
123 | * This returns the number of iovecs needed to log the given inode item. | |
124 | * | |
125 | * We need one iovec for the inode log format structure, one for the | |
126 | * inode core, and possibly one for the inode data/extents/b-tree root | |
127 | * and one for the inode attribute data/extents/b-tree root. | |
128 | */ | |
129 | STATIC void | |
130 | xfs_inode_item_size( | |
131 | struct xfs_log_item *lip, | |
132 | int *nvecs, | |
133 | int *nbytes) | |
134 | { | |
135 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
136 | struct xfs_inode *ip = iip->ili_inode; | |
137 | ||
138 | *nvecs += 2; | |
139 | *nbytes += sizeof(struct xfs_inode_log_format) + | |
140 | xfs_icdinode_size(ip->i_d.di_version); | |
141 | ||
142 | xfs_inode_item_data_fork_size(iip, nvecs, nbytes); | |
143 | if (XFS_IFORK_Q(ip)) | |
144 | xfs_inode_item_attr_fork_size(iip, nvecs, nbytes); | |
145 | } | |
146 | ||
e828776a DC |
147 | /* |
148 | * xfs_inode_item_format_extents - convert in-core extents to on-disk form | |
149 | * | |
150 | * For either the data or attr fork in extent format, we need to endian convert | |
151 | * the in-core extent as we place them into the on-disk inode. In this case, we | |
152 | * need to do this conversion before we write the extents into the log. Because | |
153 | * we don't have the disk inode to write into here, we allocate a buffer and | |
154 | * format the extents into it via xfs_iextents_copy(). We free the buffer in | |
155 | * the unlock routine after the copy for the log has been made. | |
156 | * | |
157 | * In the case of the data fork, the in-core and on-disk fork sizes can be | |
158 | * different due to delayed allocation extents. We only log on-disk extents | |
159 | * here, so always use the physical fork size to determine the size of the | |
160 | * buffer we need to allocate. | |
161 | */ | |
162 | STATIC void | |
163 | xfs_inode_item_format_extents( | |
164 | struct xfs_inode *ip, | |
165 | struct xfs_log_iovec *vecp, | |
166 | int whichfork, | |
167 | int type) | |
168 | { | |
169 | xfs_bmbt_rec_t *ext_buffer; | |
170 | ||
171 | ext_buffer = kmem_alloc(XFS_IFORK_SIZE(ip, whichfork), KM_SLEEP); | |
172 | if (whichfork == XFS_DATA_FORK) | |
173 | ip->i_itemp->ili_extents_buf = ext_buffer; | |
174 | else | |
175 | ip->i_itemp->ili_aextents_buf = ext_buffer; | |
176 | ||
177 | vecp->i_addr = ext_buffer; | |
178 | vecp->i_len = xfs_iextents_copy(ip, ext_buffer, whichfork); | |
179 | vecp->i_type = type; | |
180 | } | |
181 | ||
1da177e4 LT |
182 | /* |
183 | * This is called to fill in the vector of log iovecs for the | |
184 | * given inode log item. It fills the first item with an inode | |
185 | * log format structure, the second with the on-disk inode structure, | |
186 | * and a possible third and/or fourth with the inode data/extents/b-tree | |
187 | * root and inode attributes data/extents/b-tree root. | |
188 | */ | |
189 | STATIC void | |
190 | xfs_inode_item_format( | |
7bfa31d8 CH |
191 | struct xfs_log_item *lip, |
192 | struct xfs_log_iovec *vecp) | |
1da177e4 | 193 | { |
7bfa31d8 CH |
194 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
195 | struct xfs_inode *ip = iip->ili_inode; | |
1da177e4 | 196 | uint nvecs; |
1da177e4 | 197 | size_t data_bytes; |
1da177e4 LT |
198 | xfs_mount_t *mp; |
199 | ||
4e0d5f92 | 200 | vecp->i_addr = &iip->ili_format; |
1da177e4 | 201 | vecp->i_len = sizeof(xfs_inode_log_format_t); |
4139b3b3 | 202 | vecp->i_type = XLOG_REG_TYPE_IFORMAT; |
1da177e4 LT |
203 | vecp++; |
204 | nvecs = 1; | |
205 | ||
4e0d5f92 | 206 | vecp->i_addr = &ip->i_d; |
93848a99 | 207 | vecp->i_len = xfs_icdinode_size(ip->i_d.di_version); |
4139b3b3 | 208 | vecp->i_type = XLOG_REG_TYPE_ICORE; |
1da177e4 LT |
209 | vecp++; |
210 | nvecs++; | |
1da177e4 LT |
211 | |
212 | /* | |
213 | * If this is really an old format inode, then we need to | |
214 | * log it as such. This means that we have to copy the link | |
215 | * count from the new field to the old. We don't have to worry | |
216 | * about the new fields, because nothing trusts them as long as | |
217 | * the old inode version number is there. If the superblock already | |
218 | * has a new version number, then we don't bother converting back. | |
219 | */ | |
220 | mp = ip->i_mount; | |
51ce16d5 CH |
221 | ASSERT(ip->i_d.di_version == 1 || xfs_sb_version_hasnlink(&mp->m_sb)); |
222 | if (ip->i_d.di_version == 1) { | |
62118709 | 223 | if (!xfs_sb_version_hasnlink(&mp->m_sb)) { |
1da177e4 LT |
224 | /* |
225 | * Convert it back. | |
226 | */ | |
227 | ASSERT(ip->i_d.di_nlink <= XFS_MAXLINK_1); | |
228 | ip->i_d.di_onlink = ip->i_d.di_nlink; | |
229 | } else { | |
230 | /* | |
231 | * The superblock version has already been bumped, | |
232 | * so just make the conversion to the new inode | |
233 | * format permanent. | |
234 | */ | |
51ce16d5 | 235 | ip->i_d.di_version = 2; |
1da177e4 LT |
236 | ip->i_d.di_onlink = 0; |
237 | memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad)); | |
238 | } | |
239 | } | |
240 | ||
241 | switch (ip->i_d.di_format) { | |
242 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 243 | iip->ili_fields &= |
339a5f5d CH |
244 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | |
245 | XFS_ILOG_DEV | XFS_ILOG_UUID); | |
246 | ||
f5d8d5c4 | 247 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
339a5f5d CH |
248 | ip->i_d.di_nextents > 0 && |
249 | ip->i_df.if_bytes > 0) { | |
1da177e4 | 250 | ASSERT(ip->i_df.if_u1.if_extents != NULL); |
339a5f5d | 251 | ASSERT(ip->i_df.if_bytes / sizeof(xfs_bmbt_rec_t) > 0); |
1da177e4 | 252 | ASSERT(iip->ili_extents_buf == NULL); |
339a5f5d | 253 | |
f016bad6 | 254 | #ifdef XFS_NATIVE_HOST |
696123fc DC |
255 | if (ip->i_d.di_nextents == ip->i_df.if_bytes / |
256 | (uint)sizeof(xfs_bmbt_rec_t)) { | |
1da177e4 LT |
257 | /* |
258 | * There are no delayed allocation | |
259 | * extents, so just point to the | |
260 | * real extents array. | |
261 | */ | |
4e0d5f92 | 262 | vecp->i_addr = ip->i_df.if_u1.if_extents; |
1da177e4 | 263 | vecp->i_len = ip->i_df.if_bytes; |
4139b3b3 | 264 | vecp->i_type = XLOG_REG_TYPE_IEXT; |
1da177e4 LT |
265 | } else |
266 | #endif | |
267 | { | |
e828776a DC |
268 | xfs_inode_item_format_extents(ip, vecp, |
269 | XFS_DATA_FORK, XLOG_REG_TYPE_IEXT); | |
1da177e4 LT |
270 | } |
271 | ASSERT(vecp->i_len <= ip->i_df.if_bytes); | |
272 | iip->ili_format.ilf_dsize = vecp->i_len; | |
273 | vecp++; | |
274 | nvecs++; | |
339a5f5d | 275 | } else { |
f5d8d5c4 | 276 | iip->ili_fields &= ~XFS_ILOG_DEXT; |
1da177e4 LT |
277 | } |
278 | break; | |
279 | ||
280 | case XFS_DINODE_FMT_BTREE: | |
f5d8d5c4 | 281 | iip->ili_fields &= |
339a5f5d CH |
282 | ~(XFS_ILOG_DDATA | XFS_ILOG_DEXT | |
283 | XFS_ILOG_DEV | XFS_ILOG_UUID); | |
284 | ||
f5d8d5c4 | 285 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
339a5f5d | 286 | ip->i_df.if_broot_bytes > 0) { |
1da177e4 | 287 | ASSERT(ip->i_df.if_broot != NULL); |
4e0d5f92 | 288 | vecp->i_addr = ip->i_df.if_broot; |
1da177e4 | 289 | vecp->i_len = ip->i_df.if_broot_bytes; |
4139b3b3 | 290 | vecp->i_type = XLOG_REG_TYPE_IBROOT; |
1da177e4 LT |
291 | vecp++; |
292 | nvecs++; | |
293 | iip->ili_format.ilf_dsize = ip->i_df.if_broot_bytes; | |
339a5f5d | 294 | } else { |
f5d8d5c4 | 295 | ASSERT(!(iip->ili_fields & |
339a5f5d | 296 | XFS_ILOG_DBROOT)); |
f5d8d5c4 | 297 | iip->ili_fields &= ~XFS_ILOG_DBROOT; |
1da177e4 LT |
298 | } |
299 | break; | |
300 | ||
301 | case XFS_DINODE_FMT_LOCAL: | |
f5d8d5c4 | 302 | iip->ili_fields &= |
339a5f5d CH |
303 | ~(XFS_ILOG_DEXT | XFS_ILOG_DBROOT | |
304 | XFS_ILOG_DEV | XFS_ILOG_UUID); | |
f5d8d5c4 | 305 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
339a5f5d | 306 | ip->i_df.if_bytes > 0) { |
1da177e4 LT |
307 | ASSERT(ip->i_df.if_u1.if_data != NULL); |
308 | ASSERT(ip->i_d.di_size > 0); | |
309 | ||
4e0d5f92 | 310 | vecp->i_addr = ip->i_df.if_u1.if_data; |
1da177e4 LT |
311 | /* |
312 | * Round i_bytes up to a word boundary. | |
313 | * The underlying memory is guaranteed to | |
314 | * to be there by xfs_idata_realloc(). | |
315 | */ | |
316 | data_bytes = roundup(ip->i_df.if_bytes, 4); | |
317 | ASSERT((ip->i_df.if_real_bytes == 0) || | |
318 | (ip->i_df.if_real_bytes == data_bytes)); | |
319 | vecp->i_len = (int)data_bytes; | |
4139b3b3 | 320 | vecp->i_type = XLOG_REG_TYPE_ILOCAL; |
1da177e4 LT |
321 | vecp++; |
322 | nvecs++; | |
323 | iip->ili_format.ilf_dsize = (unsigned)data_bytes; | |
339a5f5d | 324 | } else { |
f5d8d5c4 | 325 | iip->ili_fields &= ~XFS_ILOG_DDATA; |
1da177e4 LT |
326 | } |
327 | break; | |
328 | ||
329 | case XFS_DINODE_FMT_DEV: | |
f5d8d5c4 | 330 | iip->ili_fields &= |
339a5f5d CH |
331 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | |
332 | XFS_ILOG_DEXT | XFS_ILOG_UUID); | |
f5d8d5c4 | 333 | if (iip->ili_fields & XFS_ILOG_DEV) { |
1da177e4 LT |
334 | iip->ili_format.ilf_u.ilfu_rdev = |
335 | ip->i_df.if_u2.if_rdev; | |
336 | } | |
337 | break; | |
338 | ||
339 | case XFS_DINODE_FMT_UUID: | |
f5d8d5c4 | 340 | iip->ili_fields &= |
339a5f5d CH |
341 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | |
342 | XFS_ILOG_DEXT | XFS_ILOG_DEV); | |
f5d8d5c4 | 343 | if (iip->ili_fields & XFS_ILOG_UUID) { |
1da177e4 LT |
344 | iip->ili_format.ilf_u.ilfu_uuid = |
345 | ip->i_df.if_u2.if_uuid; | |
346 | } | |
347 | break; | |
348 | ||
349 | default: | |
350 | ASSERT(0); | |
351 | break; | |
352 | } | |
353 | ||
354 | /* | |
339a5f5d | 355 | * If there are no attributes associated with the file, then we're done. |
1da177e4 LT |
356 | */ |
357 | if (!XFS_IFORK_Q(ip)) { | |
f5d8d5c4 | 358 | iip->ili_fields &= |
339a5f5d | 359 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT); |
f5d8d5c4 | 360 | goto out; |
1da177e4 LT |
361 | } |
362 | ||
363 | switch (ip->i_d.di_aformat) { | |
364 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 365 | iip->ili_fields &= |
339a5f5d CH |
366 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT); |
367 | ||
f5d8d5c4 | 368 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
339a5f5d CH |
369 | ip->i_d.di_anextents > 0 && |
370 | ip->i_afp->if_bytes > 0) { | |
371 | ASSERT(ip->i_afp->if_bytes / sizeof(xfs_bmbt_rec_t) == | |
372 | ip->i_d.di_anextents); | |
73523a2e | 373 | ASSERT(ip->i_afp->if_u1.if_extents != NULL); |
f016bad6 | 374 | #ifdef XFS_NATIVE_HOST |
1da177e4 LT |
375 | /* |
376 | * There are not delayed allocation extents | |
377 | * for attributes, so just point at the array. | |
378 | */ | |
4e0d5f92 | 379 | vecp->i_addr = ip->i_afp->if_u1.if_extents; |
1da177e4 | 380 | vecp->i_len = ip->i_afp->if_bytes; |
e828776a | 381 | vecp->i_type = XLOG_REG_TYPE_IATTR_EXT; |
1da177e4 LT |
382 | #else |
383 | ASSERT(iip->ili_aextents_buf == NULL); | |
e828776a DC |
384 | xfs_inode_item_format_extents(ip, vecp, |
385 | XFS_ATTR_FORK, XLOG_REG_TYPE_IATTR_EXT); | |
1da177e4 LT |
386 | #endif |
387 | iip->ili_format.ilf_asize = vecp->i_len; | |
388 | vecp++; | |
389 | nvecs++; | |
339a5f5d | 390 | } else { |
f5d8d5c4 | 391 | iip->ili_fields &= ~XFS_ILOG_AEXT; |
1da177e4 LT |
392 | } |
393 | break; | |
394 | ||
395 | case XFS_DINODE_FMT_BTREE: | |
f5d8d5c4 | 396 | iip->ili_fields &= |
339a5f5d CH |
397 | ~(XFS_ILOG_ADATA | XFS_ILOG_AEXT); |
398 | ||
f5d8d5c4 | 399 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
339a5f5d | 400 | ip->i_afp->if_broot_bytes > 0) { |
1da177e4 | 401 | ASSERT(ip->i_afp->if_broot != NULL); |
339a5f5d | 402 | |
4e0d5f92 | 403 | vecp->i_addr = ip->i_afp->if_broot; |
1da177e4 | 404 | vecp->i_len = ip->i_afp->if_broot_bytes; |
4139b3b3 | 405 | vecp->i_type = XLOG_REG_TYPE_IATTR_BROOT; |
1da177e4 LT |
406 | vecp++; |
407 | nvecs++; | |
408 | iip->ili_format.ilf_asize = ip->i_afp->if_broot_bytes; | |
339a5f5d | 409 | } else { |
f5d8d5c4 | 410 | iip->ili_fields &= ~XFS_ILOG_ABROOT; |
1da177e4 LT |
411 | } |
412 | break; | |
413 | ||
414 | case XFS_DINODE_FMT_LOCAL: | |
f5d8d5c4 | 415 | iip->ili_fields &= |
339a5f5d CH |
416 | ~(XFS_ILOG_AEXT | XFS_ILOG_ABROOT); |
417 | ||
f5d8d5c4 | 418 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
339a5f5d | 419 | ip->i_afp->if_bytes > 0) { |
1da177e4 LT |
420 | ASSERT(ip->i_afp->if_u1.if_data != NULL); |
421 | ||
4e0d5f92 | 422 | vecp->i_addr = ip->i_afp->if_u1.if_data; |
1da177e4 LT |
423 | /* |
424 | * Round i_bytes up to a word boundary. | |
425 | * The underlying memory is guaranteed to | |
426 | * to be there by xfs_idata_realloc(). | |
427 | */ | |
428 | data_bytes = roundup(ip->i_afp->if_bytes, 4); | |
429 | ASSERT((ip->i_afp->if_real_bytes == 0) || | |
430 | (ip->i_afp->if_real_bytes == data_bytes)); | |
431 | vecp->i_len = (int)data_bytes; | |
4139b3b3 | 432 | vecp->i_type = XLOG_REG_TYPE_IATTR_LOCAL; |
1da177e4 LT |
433 | vecp++; |
434 | nvecs++; | |
435 | iip->ili_format.ilf_asize = (unsigned)data_bytes; | |
339a5f5d | 436 | } else { |
f5d8d5c4 | 437 | iip->ili_fields &= ~XFS_ILOG_ADATA; |
1da177e4 LT |
438 | } |
439 | break; | |
440 | ||
441 | default: | |
442 | ASSERT(0); | |
443 | break; | |
444 | } | |
445 | ||
f5d8d5c4 CH |
446 | out: |
447 | /* | |
448 | * Now update the log format that goes out to disk from the in-core | |
449 | * values. We always write the inode core to make the arithmetic | |
450 | * games in recovery easier, which isn't a big deal as just about any | |
451 | * transaction would dirty it anyway. | |
452 | */ | |
8f639dde CH |
453 | iip->ili_format.ilf_fields = XFS_ILOG_CORE | |
454 | (iip->ili_fields & ~XFS_ILOG_TIMESTAMP); | |
1da177e4 LT |
455 | iip->ili_format.ilf_size = nvecs; |
456 | } | |
457 | ||
458 | ||
459 | /* | |
460 | * This is called to pin the inode associated with the inode log | |
a14a5ab5 | 461 | * item in memory so it cannot be written out. |
1da177e4 LT |
462 | */ |
463 | STATIC void | |
464 | xfs_inode_item_pin( | |
7bfa31d8 | 465 | struct xfs_log_item *lip) |
1da177e4 | 466 | { |
7bfa31d8 | 467 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 468 | |
7bfa31d8 CH |
469 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); |
470 | ||
471 | trace_xfs_inode_pin(ip, _RET_IP_); | |
472 | atomic_inc(&ip->i_pincount); | |
1da177e4 LT |
473 | } |
474 | ||
475 | ||
476 | /* | |
477 | * This is called to unpin the inode associated with the inode log | |
478 | * item which was previously pinned with a call to xfs_inode_item_pin(). | |
a14a5ab5 CH |
479 | * |
480 | * Also wake up anyone in xfs_iunpin_wait() if the count goes to 0. | |
1da177e4 | 481 | */ |
1da177e4 LT |
482 | STATIC void |
483 | xfs_inode_item_unpin( | |
7bfa31d8 | 484 | struct xfs_log_item *lip, |
9412e318 | 485 | int remove) |
1da177e4 | 486 | { |
7bfa31d8 | 487 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 488 | |
4aaf15d1 | 489 | trace_xfs_inode_unpin(ip, _RET_IP_); |
a14a5ab5 CH |
490 | ASSERT(atomic_read(&ip->i_pincount) > 0); |
491 | if (atomic_dec_and_test(&ip->i_pincount)) | |
f392e631 | 492 | wake_up_bit(&ip->i_flags, __XFS_IPINNED_BIT); |
1da177e4 LT |
493 | } |
494 | ||
1da177e4 | 495 | STATIC uint |
43ff2122 CH |
496 | xfs_inode_item_push( |
497 | struct xfs_log_item *lip, | |
498 | struct list_head *buffer_list) | |
1da177e4 | 499 | { |
7bfa31d8 CH |
500 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
501 | struct xfs_inode *ip = iip->ili_inode; | |
43ff2122 CH |
502 | struct xfs_buf *bp = NULL; |
503 | uint rval = XFS_ITEM_SUCCESS; | |
504 | int error; | |
1da177e4 | 505 | |
7bfa31d8 | 506 | if (xfs_ipincount(ip) > 0) |
1da177e4 | 507 | return XFS_ITEM_PINNED; |
1da177e4 | 508 | |
7bfa31d8 | 509 | if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED)) |
1da177e4 | 510 | return XFS_ITEM_LOCKED; |
1da177e4 | 511 | |
4c46819a CH |
512 | /* |
513 | * Re-check the pincount now that we stabilized the value by | |
514 | * taking the ilock. | |
515 | */ | |
516 | if (xfs_ipincount(ip) > 0) { | |
43ff2122 CH |
517 | rval = XFS_ITEM_PINNED; |
518 | goto out_unlock; | |
4c46819a CH |
519 | } |
520 | ||
9a3a5dab BF |
521 | /* |
522 | * Stale inode items should force out the iclog. | |
523 | */ | |
524 | if (ip->i_flags & XFS_ISTALE) { | |
525 | rval = XFS_ITEM_PINNED; | |
526 | goto out_unlock; | |
527 | } | |
528 | ||
43ff2122 CH |
529 | /* |
530 | * Someone else is already flushing the inode. Nothing we can do | |
531 | * here but wait for the flush to finish and remove the item from | |
532 | * the AIL. | |
533 | */ | |
1da177e4 | 534 | if (!xfs_iflock_nowait(ip)) { |
43ff2122 CH |
535 | rval = XFS_ITEM_FLUSHING; |
536 | goto out_unlock; | |
1da177e4 LT |
537 | } |
538 | ||
43ff2122 CH |
539 | ASSERT(iip->ili_fields != 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); |
540 | ASSERT(iip->ili_logged == 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); | |
541 | ||
542 | spin_unlock(&lip->li_ailp->xa_lock); | |
543 | ||
544 | error = xfs_iflush(ip, &bp); | |
545 | if (!error) { | |
546 | if (!xfs_buf_delwri_queue(bp, buffer_list)) | |
547 | rval = XFS_ITEM_FLUSHING; | |
548 | xfs_buf_relse(bp); | |
1da177e4 | 549 | } |
43ff2122 CH |
550 | |
551 | spin_lock(&lip->li_ailp->xa_lock); | |
552 | out_unlock: | |
553 | xfs_iunlock(ip, XFS_ILOCK_SHARED); | |
554 | return rval; | |
1da177e4 LT |
555 | } |
556 | ||
557 | /* | |
558 | * Unlock the inode associated with the inode log item. | |
559 | * Clear the fields of the inode and inode log item that | |
560 | * are specific to the current transaction. If the | |
561 | * hold flags is set, do not unlock the inode. | |
562 | */ | |
563 | STATIC void | |
564 | xfs_inode_item_unlock( | |
7bfa31d8 | 565 | struct xfs_log_item *lip) |
1da177e4 | 566 | { |
7bfa31d8 CH |
567 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
568 | struct xfs_inode *ip = iip->ili_inode; | |
898621d5 | 569 | unsigned short lock_flags; |
1da177e4 | 570 | |
f3ca8738 CH |
571 | ASSERT(ip->i_itemp != NULL); |
572 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); | |
1da177e4 LT |
573 | |
574 | /* | |
575 | * If the inode needed a separate buffer with which to log | |
576 | * its extents, then free it now. | |
577 | */ | |
578 | if (iip->ili_extents_buf != NULL) { | |
579 | ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS); | |
580 | ASSERT(ip->i_d.di_nextents > 0); | |
f5d8d5c4 | 581 | ASSERT(iip->ili_fields & XFS_ILOG_DEXT); |
1da177e4 | 582 | ASSERT(ip->i_df.if_bytes > 0); |
f0e2d93c | 583 | kmem_free(iip->ili_extents_buf); |
1da177e4 LT |
584 | iip->ili_extents_buf = NULL; |
585 | } | |
586 | if (iip->ili_aextents_buf != NULL) { | |
587 | ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS); | |
588 | ASSERT(ip->i_d.di_anextents > 0); | |
f5d8d5c4 | 589 | ASSERT(iip->ili_fields & XFS_ILOG_AEXT); |
1da177e4 | 590 | ASSERT(ip->i_afp->if_bytes > 0); |
f0e2d93c | 591 | kmem_free(iip->ili_aextents_buf); |
1da177e4 LT |
592 | iip->ili_aextents_buf = NULL; |
593 | } | |
594 | ||
898621d5 CH |
595 | lock_flags = iip->ili_lock_flags; |
596 | iip->ili_lock_flags = 0; | |
ddc3415a | 597 | if (lock_flags) |
f3ca8738 | 598 | xfs_iunlock(ip, lock_flags); |
1da177e4 LT |
599 | } |
600 | ||
601 | /* | |
de25c181 DC |
602 | * This is called to find out where the oldest active copy of the inode log |
603 | * item in the on disk log resides now that the last log write of it completed | |
604 | * at the given lsn. Since we always re-log all dirty data in an inode, the | |
605 | * latest copy in the on disk log is the only one that matters. Therefore, | |
606 | * simply return the given lsn. | |
607 | * | |
608 | * If the inode has been marked stale because the cluster is being freed, we | |
609 | * don't want to (re-)insert this inode into the AIL. There is a race condition | |
610 | * where the cluster buffer may be unpinned before the inode is inserted into | |
611 | * the AIL during transaction committed processing. If the buffer is unpinned | |
612 | * before the inode item has been committed and inserted, then it is possible | |
1316d4da | 613 | * for the buffer to be written and IO completes before the inode is inserted |
de25c181 DC |
614 | * into the AIL. In that case, we'd be inserting a clean, stale inode into the |
615 | * AIL which will never get removed. It will, however, get reclaimed which | |
616 | * triggers an assert in xfs_inode_free() complaining about freein an inode | |
617 | * still in the AIL. | |
618 | * | |
1316d4da DC |
619 | * To avoid this, just unpin the inode directly and return a LSN of -1 so the |
620 | * transaction committed code knows that it does not need to do any further | |
621 | * processing on the item. | |
1da177e4 | 622 | */ |
1da177e4 LT |
623 | STATIC xfs_lsn_t |
624 | xfs_inode_item_committed( | |
7bfa31d8 | 625 | struct xfs_log_item *lip, |
1da177e4 LT |
626 | xfs_lsn_t lsn) |
627 | { | |
de25c181 DC |
628 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
629 | struct xfs_inode *ip = iip->ili_inode; | |
630 | ||
1316d4da DC |
631 | if (xfs_iflags_test(ip, XFS_ISTALE)) { |
632 | xfs_inode_item_unpin(lip, 0); | |
633 | return -1; | |
634 | } | |
7bfa31d8 | 635 | return lsn; |
1da177e4 LT |
636 | } |
637 | ||
1da177e4 LT |
638 | /* |
639 | * XXX rcc - this one really has to do something. Probably needs | |
640 | * to stamp in a new field in the incore inode. | |
641 | */ | |
1da177e4 LT |
642 | STATIC void |
643 | xfs_inode_item_committing( | |
7bfa31d8 | 644 | struct xfs_log_item *lip, |
1da177e4 LT |
645 | xfs_lsn_t lsn) |
646 | { | |
7bfa31d8 | 647 | INODE_ITEM(lip)->ili_last_lsn = lsn; |
1da177e4 LT |
648 | } |
649 | ||
650 | /* | |
651 | * This is the ops vector shared by all buf log items. | |
652 | */ | |
272e42b2 | 653 | static const struct xfs_item_ops xfs_inode_item_ops = { |
7bfa31d8 CH |
654 | .iop_size = xfs_inode_item_size, |
655 | .iop_format = xfs_inode_item_format, | |
656 | .iop_pin = xfs_inode_item_pin, | |
657 | .iop_unpin = xfs_inode_item_unpin, | |
7bfa31d8 CH |
658 | .iop_unlock = xfs_inode_item_unlock, |
659 | .iop_committed = xfs_inode_item_committed, | |
660 | .iop_push = xfs_inode_item_push, | |
7bfa31d8 | 661 | .iop_committing = xfs_inode_item_committing |
1da177e4 LT |
662 | }; |
663 | ||
664 | ||
665 | /* | |
666 | * Initialize the inode log item for a newly allocated (in-core) inode. | |
667 | */ | |
668 | void | |
669 | xfs_inode_item_init( | |
7bfa31d8 CH |
670 | struct xfs_inode *ip, |
671 | struct xfs_mount *mp) | |
1da177e4 | 672 | { |
7bfa31d8 | 673 | struct xfs_inode_log_item *iip; |
1da177e4 LT |
674 | |
675 | ASSERT(ip->i_itemp == NULL); | |
676 | iip = ip->i_itemp = kmem_zone_zalloc(xfs_ili_zone, KM_SLEEP); | |
677 | ||
1da177e4 | 678 | iip->ili_inode = ip; |
43f5efc5 DC |
679 | xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE, |
680 | &xfs_inode_item_ops); | |
1da177e4 LT |
681 | iip->ili_format.ilf_type = XFS_LI_INODE; |
682 | iip->ili_format.ilf_ino = ip->i_ino; | |
92bfc6e7 CH |
683 | iip->ili_format.ilf_blkno = ip->i_imap.im_blkno; |
684 | iip->ili_format.ilf_len = ip->i_imap.im_len; | |
685 | iip->ili_format.ilf_boffset = ip->i_imap.im_boffset; | |
1da177e4 LT |
686 | } |
687 | ||
688 | /* | |
689 | * Free the inode log item and any memory hanging off of it. | |
690 | */ | |
691 | void | |
692 | xfs_inode_item_destroy( | |
693 | xfs_inode_t *ip) | |
694 | { | |
1da177e4 LT |
695 | kmem_zone_free(xfs_ili_zone, ip->i_itemp); |
696 | } | |
697 | ||
698 | ||
699 | /* | |
700 | * This is the inode flushing I/O completion routine. It is called | |
701 | * from interrupt level when the buffer containing the inode is | |
702 | * flushed to disk. It is responsible for removing the inode item | |
703 | * from the AIL if it has not been re-logged, and unlocking the inode's | |
704 | * flush lock. | |
30136832 DC |
705 | * |
706 | * To reduce AIL lock traffic as much as possible, we scan the buffer log item | |
707 | * list for other inodes that will run this function. We remove them from the | |
708 | * buffer list so we can process all the inode IO completions in one AIL lock | |
709 | * traversal. | |
1da177e4 | 710 | */ |
1da177e4 LT |
711 | void |
712 | xfs_iflush_done( | |
ca30b2a7 CH |
713 | struct xfs_buf *bp, |
714 | struct xfs_log_item *lip) | |
1da177e4 | 715 | { |
30136832 DC |
716 | struct xfs_inode_log_item *iip; |
717 | struct xfs_log_item *blip; | |
718 | struct xfs_log_item *next; | |
719 | struct xfs_log_item *prev; | |
ca30b2a7 | 720 | struct xfs_ail *ailp = lip->li_ailp; |
30136832 DC |
721 | int need_ail = 0; |
722 | ||
723 | /* | |
724 | * Scan the buffer IO completions for other inodes being completed and | |
725 | * attach them to the current inode log item. | |
726 | */ | |
adadbeef | 727 | blip = bp->b_fspriv; |
30136832 DC |
728 | prev = NULL; |
729 | while (blip != NULL) { | |
730 | if (lip->li_cb != xfs_iflush_done) { | |
731 | prev = blip; | |
732 | blip = blip->li_bio_list; | |
733 | continue; | |
734 | } | |
735 | ||
736 | /* remove from list */ | |
737 | next = blip->li_bio_list; | |
738 | if (!prev) { | |
adadbeef | 739 | bp->b_fspriv = next; |
30136832 DC |
740 | } else { |
741 | prev->li_bio_list = next; | |
742 | } | |
743 | ||
744 | /* add to current list */ | |
745 | blip->li_bio_list = lip->li_bio_list; | |
746 | lip->li_bio_list = blip; | |
747 | ||
748 | /* | |
749 | * while we have the item, do the unlocked check for needing | |
750 | * the AIL lock. | |
751 | */ | |
752 | iip = INODE_ITEM(blip); | |
753 | if (iip->ili_logged && blip->li_lsn == iip->ili_flush_lsn) | |
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); | |
761 | if (iip->ili_logged && lip->li_lsn == iip->ili_flush_lsn) | |
762 | need_ail++; | |
1da177e4 LT |
763 | |
764 | /* | |
765 | * We only want to pull the item from the AIL if it is | |
766 | * actually there and its location in the log has not | |
767 | * changed since we started the flush. Thus, we only bother | |
768 | * if the ili_logged flag is set and the inode's lsn has not | |
769 | * changed. First we check the lsn outside | |
770 | * the lock since it's cheaper, and then we recheck while | |
771 | * holding the lock before removing the inode from the AIL. | |
772 | */ | |
30136832 DC |
773 | if (need_ail) { |
774 | struct xfs_log_item *log_items[need_ail]; | |
775 | int i = 0; | |
783a2f65 | 776 | spin_lock(&ailp->xa_lock); |
30136832 DC |
777 | for (blip = lip; blip; blip = blip->li_bio_list) { |
778 | iip = INODE_ITEM(blip); | |
779 | if (iip->ili_logged && | |
780 | blip->li_lsn == iip->ili_flush_lsn) { | |
781 | log_items[i++] = blip; | |
782 | } | |
783 | ASSERT(i <= need_ail); | |
1da177e4 | 784 | } |
30136832 | 785 | /* xfs_trans_ail_delete_bulk() drops the AIL lock. */ |
04913fdd DC |
786 | xfs_trans_ail_delete_bulk(ailp, log_items, i, |
787 | SHUTDOWN_CORRUPT_INCORE); | |
1da177e4 LT |
788 | } |
789 | ||
1da177e4 LT |
790 | |
791 | /* | |
30136832 DC |
792 | * clean up and unlock the flush lock now we are done. We can clear the |
793 | * ili_last_fields bits now that we know that the data corresponding to | |
794 | * them is safely on disk. | |
1da177e4 | 795 | */ |
30136832 DC |
796 | for (blip = lip; blip; blip = next) { |
797 | next = blip->li_bio_list; | |
798 | blip->li_bio_list = NULL; | |
799 | ||
800 | iip = INODE_ITEM(blip); | |
801 | iip->ili_logged = 0; | |
802 | iip->ili_last_fields = 0; | |
803 | xfs_ifunlock(iip->ili_inode); | |
804 | } | |
1da177e4 LT |
805 | } |
806 | ||
807 | /* | |
04913fdd DC |
808 | * This is the inode flushing abort routine. It is called from xfs_iflush when |
809 | * the filesystem is shutting down to clean up the inode state. It is | |
810 | * responsible for removing the inode item from the AIL if it has not been | |
811 | * re-logged, and unlocking the inode's flush lock. | |
1da177e4 LT |
812 | */ |
813 | void | |
814 | xfs_iflush_abort( | |
04913fdd DC |
815 | xfs_inode_t *ip, |
816 | bool stale) | |
1da177e4 | 817 | { |
783a2f65 | 818 | xfs_inode_log_item_t *iip = ip->i_itemp; |
1da177e4 | 819 | |
1da177e4 | 820 | if (iip) { |
783a2f65 | 821 | struct xfs_ail *ailp = iip->ili_item.li_ailp; |
1da177e4 | 822 | if (iip->ili_item.li_flags & XFS_LI_IN_AIL) { |
783a2f65 | 823 | spin_lock(&ailp->xa_lock); |
1da177e4 | 824 | if (iip->ili_item.li_flags & XFS_LI_IN_AIL) { |
783a2f65 | 825 | /* xfs_trans_ail_delete() drops the AIL lock. */ |
04913fdd DC |
826 | xfs_trans_ail_delete(ailp, &iip->ili_item, |
827 | stale ? | |
828 | SHUTDOWN_LOG_IO_ERROR : | |
829 | SHUTDOWN_CORRUPT_INCORE); | |
1da177e4 | 830 | } else |
783a2f65 | 831 | spin_unlock(&ailp->xa_lock); |
1da177e4 LT |
832 | } |
833 | iip->ili_logged = 0; | |
834 | /* | |
835 | * Clear the ili_last_fields bits now that we know that the | |
836 | * data corresponding to them is safely on disk. | |
837 | */ | |
838 | iip->ili_last_fields = 0; | |
839 | /* | |
840 | * Clear the inode logging fields so no more flushes are | |
841 | * attempted. | |
842 | */ | |
f5d8d5c4 | 843 | iip->ili_fields = 0; |
1da177e4 LT |
844 | } |
845 | /* | |
846 | * Release the inode's flush lock since we're done with it. | |
847 | */ | |
848 | xfs_ifunlock(ip); | |
849 | } | |
850 | ||
851 | void | |
852 | xfs_istale_done( | |
ca30b2a7 CH |
853 | struct xfs_buf *bp, |
854 | struct xfs_log_item *lip) | |
1da177e4 | 855 | { |
04913fdd | 856 | xfs_iflush_abort(INODE_ITEM(lip)->ili_inode, true); |
1da177e4 | 857 | } |
6d192a9b TS |
858 | |
859 | /* | |
860 | * convert an xfs_inode_log_format struct from either 32 or 64 bit versions | |
861 | * (which can have different field alignments) to the native version | |
862 | */ | |
863 | int | |
864 | xfs_inode_item_format_convert( | |
865 | xfs_log_iovec_t *buf, | |
866 | xfs_inode_log_format_t *in_f) | |
867 | { | |
868 | if (buf->i_len == sizeof(xfs_inode_log_format_32_t)) { | |
4e0d5f92 | 869 | xfs_inode_log_format_32_t *in_f32 = buf->i_addr; |
6d192a9b | 870 | |
6d192a9b TS |
871 | in_f->ilf_type = in_f32->ilf_type; |
872 | in_f->ilf_size = in_f32->ilf_size; | |
873 | in_f->ilf_fields = in_f32->ilf_fields; | |
874 | in_f->ilf_asize = in_f32->ilf_asize; | |
875 | in_f->ilf_dsize = in_f32->ilf_dsize; | |
876 | in_f->ilf_ino = in_f32->ilf_ino; | |
877 | /* copy biggest field of ilf_u */ | |
878 | memcpy(in_f->ilf_u.ilfu_uuid.__u_bits, | |
879 | in_f32->ilf_u.ilfu_uuid.__u_bits, | |
880 | sizeof(uuid_t)); | |
881 | in_f->ilf_blkno = in_f32->ilf_blkno; | |
882 | in_f->ilf_len = in_f32->ilf_len; | |
883 | in_f->ilf_boffset = in_f32->ilf_boffset; | |
884 | return 0; | |
885 | } else if (buf->i_len == sizeof(xfs_inode_log_format_64_t)){ | |
4e0d5f92 | 886 | xfs_inode_log_format_64_t *in_f64 = buf->i_addr; |
6d192a9b | 887 | |
6d192a9b TS |
888 | in_f->ilf_type = in_f64->ilf_type; |
889 | in_f->ilf_size = in_f64->ilf_size; | |
890 | in_f->ilf_fields = in_f64->ilf_fields; | |
891 | in_f->ilf_asize = in_f64->ilf_asize; | |
892 | in_f->ilf_dsize = in_f64->ilf_dsize; | |
893 | in_f->ilf_ino = in_f64->ilf_ino; | |
894 | /* copy biggest field of ilf_u */ | |
895 | memcpy(in_f->ilf_u.ilfu_uuid.__u_bits, | |
896 | in_f64->ilf_u.ilfu_uuid.__u_bits, | |
897 | sizeof(uuid_t)); | |
898 | in_f->ilf_blkno = in_f64->ilf_blkno; | |
899 | in_f->ilf_len = in_f64->ilf_len; | |
900 | in_f->ilf_boffset = in_f64->ilf_boffset; | |
901 | return 0; | |
902 | } | |
903 | return EFSCORRUPTED; | |
904 | } |