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" |
1234351c | 33 | #include "xfs_log.h" |
1da177e4 LT |
34 | |
35 | ||
36 | kmem_zone_t *xfs_ili_zone; /* inode log item zone */ | |
37 | ||
7bfa31d8 CH |
38 | static inline struct xfs_inode_log_item *INODE_ITEM(struct xfs_log_item *lip) |
39 | { | |
40 | return container_of(lip, struct xfs_inode_log_item, ili_item); | |
41 | } | |
42 | ||
166d1368 | 43 | STATIC void |
ce9641d6 CH |
44 | xfs_inode_item_data_fork_size( |
45 | struct xfs_inode_log_item *iip, | |
166d1368 DC |
46 | int *nvecs, |
47 | int *nbytes) | |
1da177e4 | 48 | { |
7bfa31d8 | 49 | struct xfs_inode *ip = iip->ili_inode; |
166d1368 | 50 | |
1da177e4 LT |
51 | switch (ip->i_d.di_format) { |
52 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 53 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
339a5f5d | 54 | ip->i_d.di_nextents > 0 && |
166d1368 DC |
55 | ip->i_df.if_bytes > 0) { |
56 | /* worst case, doesn't subtract delalloc extents */ | |
57 | *nbytes += XFS_IFORK_DSIZE(ip); | |
58 | *nvecs += 1; | |
59 | } | |
1da177e4 | 60 | break; |
1da177e4 | 61 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 62 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
166d1368 DC |
63 | ip->i_df.if_broot_bytes > 0) { |
64 | *nbytes += ip->i_df.if_broot_bytes; | |
65 | *nvecs += 1; | |
66 | } | |
1da177e4 | 67 | break; |
1da177e4 | 68 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 69 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
166d1368 DC |
70 | ip->i_df.if_bytes > 0) { |
71 | *nbytes += roundup(ip->i_df.if_bytes, 4); | |
72 | *nvecs += 1; | |
73 | } | |
1da177e4 LT |
74 | break; |
75 | ||
76 | case XFS_DINODE_FMT_DEV: | |
1da177e4 | 77 | case XFS_DINODE_FMT_UUID: |
1da177e4 | 78 | break; |
1da177e4 LT |
79 | default: |
80 | ASSERT(0); | |
81 | break; | |
82 | } | |
ce9641d6 | 83 | } |
1da177e4 | 84 | |
ce9641d6 CH |
85 | STATIC void |
86 | xfs_inode_item_attr_fork_size( | |
87 | struct xfs_inode_log_item *iip, | |
88 | int *nvecs, | |
89 | int *nbytes) | |
90 | { | |
91 | struct xfs_inode *ip = iip->ili_inode; | |
1da177e4 | 92 | |
1da177e4 LT |
93 | switch (ip->i_d.di_aformat) { |
94 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 95 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
339a5f5d | 96 | ip->i_d.di_anextents > 0 && |
166d1368 DC |
97 | ip->i_afp->if_bytes > 0) { |
98 | /* worst case, doesn't subtract unused space */ | |
99 | *nbytes += XFS_IFORK_ASIZE(ip); | |
100 | *nvecs += 1; | |
101 | } | |
1da177e4 | 102 | break; |
1da177e4 | 103 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 104 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
166d1368 DC |
105 | ip->i_afp->if_broot_bytes > 0) { |
106 | *nbytes += ip->i_afp->if_broot_bytes; | |
107 | *nvecs += 1; | |
108 | } | |
1da177e4 | 109 | break; |
1da177e4 | 110 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 111 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
166d1368 DC |
112 | ip->i_afp->if_bytes > 0) { |
113 | *nbytes += roundup(ip->i_afp->if_bytes, 4); | |
114 | *nvecs += 1; | |
115 | } | |
1da177e4 | 116 | break; |
1da177e4 LT |
117 | default: |
118 | ASSERT(0); | |
119 | break; | |
120 | } | |
1da177e4 LT |
121 | } |
122 | ||
ce9641d6 CH |
123 | /* |
124 | * This returns the number of iovecs needed to log the given inode item. | |
125 | * | |
126 | * We need one iovec for the inode log format structure, one for the | |
127 | * inode core, and possibly one for the inode data/extents/b-tree root | |
128 | * and one for the inode attribute data/extents/b-tree root. | |
129 | */ | |
130 | STATIC void | |
131 | xfs_inode_item_size( | |
132 | struct xfs_log_item *lip, | |
133 | int *nvecs, | |
134 | int *nbytes) | |
135 | { | |
136 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
137 | struct xfs_inode *ip = iip->ili_inode; | |
138 | ||
139 | *nvecs += 2; | |
140 | *nbytes += sizeof(struct xfs_inode_log_format) + | |
141 | xfs_icdinode_size(ip->i_d.di_version); | |
142 | ||
143 | xfs_inode_item_data_fork_size(iip, nvecs, nbytes); | |
144 | if (XFS_IFORK_Q(ip)) | |
145 | xfs_inode_item_attr_fork_size(iip, nvecs, nbytes); | |
146 | } | |
147 | ||
e828776a DC |
148 | /* |
149 | * xfs_inode_item_format_extents - convert in-core extents to on-disk form | |
150 | * | |
151 | * For either the data or attr fork in extent format, we need to endian convert | |
152 | * the in-core extent as we place them into the on-disk inode. In this case, we | |
153 | * need to do this conversion before we write the extents into the log. Because | |
154 | * we don't have the disk inode to write into here, we allocate a buffer and | |
155 | * format the extents into it via xfs_iextents_copy(). We free the buffer in | |
156 | * the unlock routine after the copy for the log has been made. | |
157 | * | |
158 | * In the case of the data fork, the in-core and on-disk fork sizes can be | |
159 | * different due to delayed allocation extents. We only log on-disk extents | |
160 | * here, so always use the physical fork size to determine the size of the | |
161 | * buffer we need to allocate. | |
162 | */ | |
1234351c | 163 | STATIC int |
e828776a DC |
164 | xfs_inode_item_format_extents( |
165 | struct xfs_inode *ip, | |
1234351c | 166 | struct xfs_log_iovec **vecp, |
e828776a DC |
167 | int whichfork, |
168 | int type) | |
169 | { | |
170 | xfs_bmbt_rec_t *ext_buffer; | |
1234351c | 171 | int len; |
e828776a DC |
172 | |
173 | ext_buffer = kmem_alloc(XFS_IFORK_SIZE(ip, whichfork), KM_SLEEP); | |
174 | if (whichfork == XFS_DATA_FORK) | |
175 | ip->i_itemp->ili_extents_buf = ext_buffer; | |
176 | else | |
177 | ip->i_itemp->ili_aextents_buf = ext_buffer; | |
178 | ||
1234351c CH |
179 | len = xfs_iextents_copy(ip, ext_buffer, whichfork); |
180 | xlog_copy_iovec(vecp, type, ext_buffer, len); | |
181 | return len; | |
e828776a DC |
182 | } |
183 | ||
1da177e4 | 184 | /* |
3de559fb CH |
185 | * If this is a v1 format inode, then we need to log it as such. This means |
186 | * that we have to copy the link count from the new field to the old. We | |
187 | * don't have to worry about the new fields, because nothing trusts them as | |
188 | * long as the old inode version number is there. | |
1da177e4 LT |
189 | */ |
190 | STATIC void | |
3de559fb CH |
191 | xfs_inode_item_format_v1_inode( |
192 | struct xfs_inode *ip) | |
193 | { | |
194 | if (!xfs_sb_version_hasnlink(&ip->i_mount->m_sb)) { | |
195 | /* | |
196 | * Convert it back. | |
197 | */ | |
198 | ASSERT(ip->i_d.di_nlink <= XFS_MAXLINK_1); | |
199 | ip->i_d.di_onlink = ip->i_d.di_nlink; | |
200 | } else { | |
201 | /* | |
202 | * The superblock version has already been bumped, | |
203 | * so just make the conversion to the new inode | |
204 | * format permanent. | |
205 | */ | |
206 | ip->i_d.di_version = 2; | |
207 | ip->i_d.di_onlink = 0; | |
208 | memset(&(ip->i_d.di_pad[0]), 0, sizeof(ip->i_d.di_pad)); | |
209 | } | |
210 | } | |
211 | ||
1234351c | 212 | STATIC void |
3de559fb CH |
213 | xfs_inode_item_format_data_fork( |
214 | struct xfs_inode_log_item *iip, | |
1234351c | 215 | struct xfs_log_iovec **vecp, |
3de559fb | 216 | int *nvecs) |
1da177e4 | 217 | { |
7bfa31d8 | 218 | struct xfs_inode *ip = iip->ili_inode; |
1da177e4 | 219 | size_t data_bytes; |
1da177e4 LT |
220 | |
221 | switch (ip->i_d.di_format) { | |
222 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 223 | iip->ili_fields &= |
339a5f5d CH |
224 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | |
225 | XFS_ILOG_DEV | XFS_ILOG_UUID); | |
226 | ||
f5d8d5c4 | 227 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
339a5f5d CH |
228 | ip->i_d.di_nextents > 0 && |
229 | ip->i_df.if_bytes > 0) { | |
1da177e4 | 230 | ASSERT(ip->i_df.if_u1.if_extents != NULL); |
339a5f5d | 231 | ASSERT(ip->i_df.if_bytes / sizeof(xfs_bmbt_rec_t) > 0); |
1da177e4 | 232 | ASSERT(iip->ili_extents_buf == NULL); |
339a5f5d | 233 | |
f016bad6 | 234 | #ifdef XFS_NATIVE_HOST |
696123fc DC |
235 | if (ip->i_d.di_nextents == ip->i_df.if_bytes / |
236 | (uint)sizeof(xfs_bmbt_rec_t)) { | |
1da177e4 LT |
237 | /* |
238 | * There are no delayed allocation | |
239 | * extents, so just point to the | |
240 | * real extents array. | |
241 | */ | |
1234351c CH |
242 | xlog_copy_iovec(vecp, XLOG_REG_TYPE_IEXT, |
243 | ip->i_df.if_u1.if_extents, | |
244 | ip->i_df.if_bytes); | |
245 | iip->ili_format.ilf_dsize = ip->i_df.if_bytes; | |
1da177e4 LT |
246 | } else |
247 | #endif | |
248 | { | |
1234351c CH |
249 | iip->ili_format.ilf_dsize = |
250 | xfs_inode_item_format_extents(ip, vecp, | |
251 | XFS_DATA_FORK, XLOG_REG_TYPE_IEXT); | |
252 | ASSERT(iip->ili_format.ilf_dsize <= ip->i_df.if_bytes); | |
1da177e4 | 253 | } |
3de559fb | 254 | (*nvecs)++; |
339a5f5d | 255 | } else { |
f5d8d5c4 | 256 | iip->ili_fields &= ~XFS_ILOG_DEXT; |
1da177e4 LT |
257 | } |
258 | break; | |
1da177e4 | 259 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 260 | iip->ili_fields &= |
339a5f5d CH |
261 | ~(XFS_ILOG_DDATA | XFS_ILOG_DEXT | |
262 | XFS_ILOG_DEV | XFS_ILOG_UUID); | |
263 | ||
f5d8d5c4 | 264 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
339a5f5d | 265 | ip->i_df.if_broot_bytes > 0) { |
1da177e4 | 266 | ASSERT(ip->i_df.if_broot != NULL); |
1234351c CH |
267 | xlog_copy_iovec(vecp, XLOG_REG_TYPE_IBROOT, |
268 | ip->i_df.if_broot, | |
269 | ip->i_df.if_broot_bytes); | |
3de559fb | 270 | (*nvecs)++; |
1da177e4 | 271 | iip->ili_format.ilf_dsize = ip->i_df.if_broot_bytes; |
339a5f5d | 272 | } else { |
f5d8d5c4 | 273 | ASSERT(!(iip->ili_fields & |
339a5f5d | 274 | XFS_ILOG_DBROOT)); |
f5d8d5c4 | 275 | iip->ili_fields &= ~XFS_ILOG_DBROOT; |
1da177e4 LT |
276 | } |
277 | break; | |
1da177e4 | 278 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 279 | iip->ili_fields &= |
339a5f5d CH |
280 | ~(XFS_ILOG_DEXT | XFS_ILOG_DBROOT | |
281 | XFS_ILOG_DEV | XFS_ILOG_UUID); | |
f5d8d5c4 | 282 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
339a5f5d | 283 | ip->i_df.if_bytes > 0) { |
1da177e4 LT |
284 | /* |
285 | * Round i_bytes up to a word boundary. | |
286 | * The underlying memory is guaranteed to | |
287 | * to be there by xfs_idata_realloc(). | |
288 | */ | |
289 | data_bytes = roundup(ip->i_df.if_bytes, 4); | |
1234351c CH |
290 | ASSERT(ip->i_df.if_real_bytes == 0 || |
291 | ip->i_df.if_real_bytes == data_bytes); | |
292 | ASSERT(ip->i_df.if_u1.if_data != NULL); | |
293 | ASSERT(ip->i_d.di_size > 0); | |
294 | xlog_copy_iovec(vecp, XLOG_REG_TYPE_ILOCAL, | |
295 | ip->i_df.if_u1.if_data, data_bytes); | |
3de559fb | 296 | (*nvecs)++; |
1da177e4 | 297 | iip->ili_format.ilf_dsize = (unsigned)data_bytes; |
339a5f5d | 298 | } else { |
f5d8d5c4 | 299 | iip->ili_fields &= ~XFS_ILOG_DDATA; |
1da177e4 LT |
300 | } |
301 | break; | |
1da177e4 | 302 | case XFS_DINODE_FMT_DEV: |
f5d8d5c4 | 303 | iip->ili_fields &= |
339a5f5d CH |
304 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | |
305 | XFS_ILOG_DEXT | XFS_ILOG_UUID); | |
f5d8d5c4 | 306 | if (iip->ili_fields & XFS_ILOG_DEV) { |
1da177e4 LT |
307 | iip->ili_format.ilf_u.ilfu_rdev = |
308 | ip->i_df.if_u2.if_rdev; | |
309 | } | |
310 | break; | |
1da177e4 | 311 | case XFS_DINODE_FMT_UUID: |
f5d8d5c4 | 312 | iip->ili_fields &= |
339a5f5d CH |
313 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | |
314 | XFS_ILOG_DEXT | XFS_ILOG_DEV); | |
f5d8d5c4 | 315 | if (iip->ili_fields & XFS_ILOG_UUID) { |
1da177e4 LT |
316 | iip->ili_format.ilf_u.ilfu_uuid = |
317 | ip->i_df.if_u2.if_uuid; | |
318 | } | |
319 | break; | |
1da177e4 LT |
320 | default: |
321 | ASSERT(0); | |
322 | break; | |
323 | } | |
3de559fb CH |
324 | } |
325 | ||
1234351c | 326 | STATIC void |
3de559fb CH |
327 | xfs_inode_item_format_attr_fork( |
328 | struct xfs_inode_log_item *iip, | |
1234351c | 329 | struct xfs_log_iovec **vecp, |
3de559fb CH |
330 | int *nvecs) |
331 | { | |
332 | struct xfs_inode *ip = iip->ili_inode; | |
333 | size_t data_bytes; | |
1da177e4 LT |
334 | |
335 | switch (ip->i_d.di_aformat) { | |
336 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 337 | iip->ili_fields &= |
339a5f5d CH |
338 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT); |
339 | ||
f5d8d5c4 | 340 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
339a5f5d CH |
341 | ip->i_d.di_anextents > 0 && |
342 | ip->i_afp->if_bytes > 0) { | |
343 | ASSERT(ip->i_afp->if_bytes / sizeof(xfs_bmbt_rec_t) == | |
344 | ip->i_d.di_anextents); | |
73523a2e | 345 | ASSERT(ip->i_afp->if_u1.if_extents != NULL); |
f016bad6 | 346 | #ifdef XFS_NATIVE_HOST |
1da177e4 LT |
347 | /* |
348 | * There are not delayed allocation extents | |
349 | * for attributes, so just point at the array. | |
350 | */ | |
1234351c CH |
351 | xlog_copy_iovec(vecp, XLOG_REG_TYPE_IATTR_EXT, |
352 | ip->i_afp->if_u1.if_extents, | |
353 | ip->i_afp->if_bytes); | |
354 | iip->ili_format.ilf_asize = ip->i_afp->if_bytes; | |
1da177e4 LT |
355 | #else |
356 | ASSERT(iip->ili_aextents_buf == NULL); | |
1234351c CH |
357 | iip->ili_format.ilf_asize = |
358 | xfs_inode_item_format_extents(ip, vecp, | |
e828776a | 359 | XFS_ATTR_FORK, XLOG_REG_TYPE_IATTR_EXT); |
1da177e4 | 360 | #endif |
3de559fb | 361 | (*nvecs)++; |
339a5f5d | 362 | } else { |
f5d8d5c4 | 363 | iip->ili_fields &= ~XFS_ILOG_AEXT; |
1da177e4 LT |
364 | } |
365 | break; | |
1da177e4 | 366 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 367 | iip->ili_fields &= |
339a5f5d CH |
368 | ~(XFS_ILOG_ADATA | XFS_ILOG_AEXT); |
369 | ||
f5d8d5c4 | 370 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
339a5f5d | 371 | ip->i_afp->if_broot_bytes > 0) { |
1da177e4 | 372 | ASSERT(ip->i_afp->if_broot != NULL); |
339a5f5d | 373 | |
1234351c CH |
374 | xlog_copy_iovec(vecp, XLOG_REG_TYPE_IATTR_BROOT, |
375 | ip->i_afp->if_broot, | |
376 | ip->i_afp->if_broot_bytes); | |
3de559fb | 377 | (*nvecs)++; |
1da177e4 | 378 | iip->ili_format.ilf_asize = ip->i_afp->if_broot_bytes; |
339a5f5d | 379 | } else { |
f5d8d5c4 | 380 | iip->ili_fields &= ~XFS_ILOG_ABROOT; |
1da177e4 LT |
381 | } |
382 | break; | |
1da177e4 | 383 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 384 | iip->ili_fields &= |
339a5f5d CH |
385 | ~(XFS_ILOG_AEXT | XFS_ILOG_ABROOT); |
386 | ||
f5d8d5c4 | 387 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
339a5f5d | 388 | ip->i_afp->if_bytes > 0) { |
1da177e4 LT |
389 | /* |
390 | * Round i_bytes up to a word boundary. | |
391 | * The underlying memory is guaranteed to | |
392 | * to be there by xfs_idata_realloc(). | |
393 | */ | |
394 | data_bytes = roundup(ip->i_afp->if_bytes, 4); | |
1234351c CH |
395 | ASSERT(ip->i_afp->if_real_bytes == 0 || |
396 | ip->i_afp->if_real_bytes == data_bytes); | |
397 | ASSERT(ip->i_afp->if_u1.if_data != NULL); | |
398 | xlog_copy_iovec(vecp, XLOG_REG_TYPE_IATTR_LOCAL, | |
399 | ip->i_afp->if_u1.if_data, | |
400 | data_bytes); | |
3de559fb | 401 | (*nvecs)++; |
1da177e4 | 402 | iip->ili_format.ilf_asize = (unsigned)data_bytes; |
339a5f5d | 403 | } else { |
f5d8d5c4 | 404 | iip->ili_fields &= ~XFS_ILOG_ADATA; |
1da177e4 LT |
405 | } |
406 | break; | |
1da177e4 LT |
407 | default: |
408 | ASSERT(0); | |
409 | break; | |
410 | } | |
3de559fb CH |
411 | } |
412 | ||
413 | /* | |
414 | * This is called to fill in the vector of log iovecs for the given inode | |
415 | * log item. It fills the first item with an inode log format structure, | |
416 | * the second with the on-disk inode structure, and a possible third and/or | |
417 | * fourth with the inode data/extents/b-tree root and inode attributes | |
418 | * data/extents/b-tree root. | |
419 | */ | |
420 | STATIC void | |
421 | xfs_inode_item_format( | |
422 | struct xfs_log_item *lip, | |
423 | struct xfs_log_iovec *vecp) | |
424 | { | |
425 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
426 | struct xfs_inode *ip = iip->ili_inode; | |
427 | uint nvecs; | |
428 | ||
1234351c CH |
429 | xlog_copy_iovec(&vecp, XLOG_REG_TYPE_IFORMAT, |
430 | &iip->ili_format, | |
431 | sizeof(struct xfs_inode_log_format)); | |
432 | nvecs = 1; | |
3de559fb | 433 | |
1234351c CH |
434 | xlog_copy_iovec(&vecp, XLOG_REG_TYPE_ICORE, |
435 | &ip->i_d, | |
436 | xfs_icdinode_size(ip->i_d.di_version)); | |
3de559fb CH |
437 | nvecs++; |
438 | ||
439 | if (ip->i_d.di_version == 1) | |
440 | xfs_inode_item_format_v1_inode(ip); | |
441 | ||
1234351c | 442 | xfs_inode_item_format_data_fork(iip, &vecp, &nvecs); |
3de559fb | 443 | if (XFS_IFORK_Q(ip)) { |
1234351c | 444 | xfs_inode_item_format_attr_fork(iip, &vecp, &nvecs); |
3de559fb CH |
445 | } else { |
446 | iip->ili_fields &= | |
447 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT); | |
448 | } | |
449 | ||
f5d8d5c4 CH |
450 | /* |
451 | * Now update the log format that goes out to disk from the in-core | |
452 | * values. We always write the inode core to make the arithmetic | |
453 | * games in recovery easier, which isn't a big deal as just about any | |
454 | * transaction would dirty it anyway. | |
455 | */ | |
8f639dde CH |
456 | iip->ili_format.ilf_fields = XFS_ILOG_CORE | |
457 | (iip->ili_fields & ~XFS_ILOG_TIMESTAMP); | |
1da177e4 LT |
458 | iip->ili_format.ilf_size = nvecs; |
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 | ||
1da177e4 | 497 | STATIC uint |
43ff2122 CH |
498 | xfs_inode_item_push( |
499 | struct xfs_log_item *lip, | |
500 | struct list_head *buffer_list) | |
1da177e4 | 501 | { |
7bfa31d8 CH |
502 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
503 | struct xfs_inode *ip = iip->ili_inode; | |
43ff2122 CH |
504 | struct xfs_buf *bp = NULL; |
505 | uint rval = XFS_ITEM_SUCCESS; | |
506 | int error; | |
1da177e4 | 507 | |
7bfa31d8 | 508 | if (xfs_ipincount(ip) > 0) |
1da177e4 | 509 | return XFS_ITEM_PINNED; |
1da177e4 | 510 | |
7bfa31d8 | 511 | if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED)) |
1da177e4 | 512 | return XFS_ITEM_LOCKED; |
1da177e4 | 513 | |
4c46819a CH |
514 | /* |
515 | * Re-check the pincount now that we stabilized the value by | |
516 | * taking the ilock. | |
517 | */ | |
518 | if (xfs_ipincount(ip) > 0) { | |
43ff2122 CH |
519 | rval = XFS_ITEM_PINNED; |
520 | goto out_unlock; | |
4c46819a CH |
521 | } |
522 | ||
9a3a5dab BF |
523 | /* |
524 | * Stale inode items should force out the iclog. | |
525 | */ | |
526 | if (ip->i_flags & XFS_ISTALE) { | |
527 | rval = XFS_ITEM_PINNED; | |
528 | goto out_unlock; | |
529 | } | |
530 | ||
43ff2122 CH |
531 | /* |
532 | * Someone else is already flushing the inode. Nothing we can do | |
533 | * here but wait for the flush to finish and remove the item from | |
534 | * the AIL. | |
535 | */ | |
1da177e4 | 536 | if (!xfs_iflock_nowait(ip)) { |
43ff2122 CH |
537 | rval = XFS_ITEM_FLUSHING; |
538 | goto out_unlock; | |
1da177e4 LT |
539 | } |
540 | ||
43ff2122 CH |
541 | ASSERT(iip->ili_fields != 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); |
542 | ASSERT(iip->ili_logged == 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); | |
543 | ||
544 | spin_unlock(&lip->li_ailp->xa_lock); | |
545 | ||
546 | error = xfs_iflush(ip, &bp); | |
547 | if (!error) { | |
548 | if (!xfs_buf_delwri_queue(bp, buffer_list)) | |
549 | rval = XFS_ITEM_FLUSHING; | |
550 | xfs_buf_relse(bp); | |
1da177e4 | 551 | } |
43ff2122 CH |
552 | |
553 | spin_lock(&lip->li_ailp->xa_lock); | |
554 | out_unlock: | |
555 | xfs_iunlock(ip, XFS_ILOCK_SHARED); | |
556 | return rval; | |
1da177e4 LT |
557 | } |
558 | ||
559 | /* | |
560 | * Unlock the inode associated with the inode log item. | |
561 | * Clear the fields of the inode and inode log item that | |
562 | * are specific to the current transaction. If the | |
563 | * hold flags is set, do not unlock the inode. | |
564 | */ | |
565 | STATIC void | |
566 | xfs_inode_item_unlock( | |
7bfa31d8 | 567 | struct xfs_log_item *lip) |
1da177e4 | 568 | { |
7bfa31d8 CH |
569 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
570 | struct xfs_inode *ip = iip->ili_inode; | |
898621d5 | 571 | unsigned short lock_flags; |
1da177e4 | 572 | |
f3ca8738 CH |
573 | ASSERT(ip->i_itemp != NULL); |
574 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); | |
1da177e4 LT |
575 | |
576 | /* | |
577 | * If the inode needed a separate buffer with which to log | |
578 | * its extents, then free it now. | |
579 | */ | |
580 | if (iip->ili_extents_buf != NULL) { | |
581 | ASSERT(ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS); | |
582 | ASSERT(ip->i_d.di_nextents > 0); | |
f5d8d5c4 | 583 | ASSERT(iip->ili_fields & XFS_ILOG_DEXT); |
1da177e4 | 584 | ASSERT(ip->i_df.if_bytes > 0); |
f0e2d93c | 585 | kmem_free(iip->ili_extents_buf); |
1da177e4 LT |
586 | iip->ili_extents_buf = NULL; |
587 | } | |
588 | if (iip->ili_aextents_buf != NULL) { | |
589 | ASSERT(ip->i_d.di_aformat == XFS_DINODE_FMT_EXTENTS); | |
590 | ASSERT(ip->i_d.di_anextents > 0); | |
f5d8d5c4 | 591 | ASSERT(iip->ili_fields & XFS_ILOG_AEXT); |
1da177e4 | 592 | ASSERT(ip->i_afp->if_bytes > 0); |
f0e2d93c | 593 | kmem_free(iip->ili_aextents_buf); |
1da177e4 LT |
594 | iip->ili_aextents_buf = NULL; |
595 | } | |
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, | |
7bfa31d8 | 663 | .iop_committing = xfs_inode_item_committing |
1da177e4 LT |
664 | }; |
665 | ||
666 | ||
667 | /* | |
668 | * Initialize the inode log item for a newly allocated (in-core) inode. | |
669 | */ | |
670 | void | |
671 | xfs_inode_item_init( | |
7bfa31d8 CH |
672 | struct xfs_inode *ip, |
673 | struct xfs_mount *mp) | |
1da177e4 | 674 | { |
7bfa31d8 | 675 | struct xfs_inode_log_item *iip; |
1da177e4 LT |
676 | |
677 | ASSERT(ip->i_itemp == NULL); | |
678 | iip = ip->i_itemp = kmem_zone_zalloc(xfs_ili_zone, KM_SLEEP); | |
679 | ||
1da177e4 | 680 | iip->ili_inode = ip; |
43f5efc5 DC |
681 | xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE, |
682 | &xfs_inode_item_ops); | |
1da177e4 LT |
683 | iip->ili_format.ilf_type = XFS_LI_INODE; |
684 | iip->ili_format.ilf_ino = ip->i_ino; | |
92bfc6e7 CH |
685 | iip->ili_format.ilf_blkno = ip->i_imap.im_blkno; |
686 | iip->ili_format.ilf_len = ip->i_imap.im_len; | |
687 | iip->ili_format.ilf_boffset = ip->i_imap.im_boffset; | |
1da177e4 LT |
688 | } |
689 | ||
690 | /* | |
691 | * Free the inode log item and any memory hanging off of it. | |
692 | */ | |
693 | void | |
694 | xfs_inode_item_destroy( | |
695 | xfs_inode_t *ip) | |
696 | { | |
1da177e4 LT |
697 | kmem_zone_free(xfs_ili_zone, ip->i_itemp); |
698 | } | |
699 | ||
700 | ||
701 | /* | |
702 | * This is the inode flushing I/O completion routine. It is called | |
703 | * from interrupt level when the buffer containing the inode is | |
704 | * flushed to disk. It is responsible for removing the inode item | |
705 | * from the AIL if it has not been re-logged, and unlocking the inode's | |
706 | * flush lock. | |
30136832 DC |
707 | * |
708 | * To reduce AIL lock traffic as much as possible, we scan the buffer log item | |
709 | * list for other inodes that will run this function. We remove them from the | |
710 | * buffer list so we can process all the inode IO completions in one AIL lock | |
711 | * traversal. | |
1da177e4 | 712 | */ |
1da177e4 LT |
713 | void |
714 | xfs_iflush_done( | |
ca30b2a7 CH |
715 | struct xfs_buf *bp, |
716 | struct xfs_log_item *lip) | |
1da177e4 | 717 | { |
30136832 DC |
718 | struct xfs_inode_log_item *iip; |
719 | struct xfs_log_item *blip; | |
720 | struct xfs_log_item *next; | |
721 | struct xfs_log_item *prev; | |
ca30b2a7 | 722 | struct xfs_ail *ailp = lip->li_ailp; |
30136832 DC |
723 | int need_ail = 0; |
724 | ||
725 | /* | |
726 | * Scan the buffer IO completions for other inodes being completed and | |
727 | * attach them to the current inode log item. | |
728 | */ | |
adadbeef | 729 | blip = bp->b_fspriv; |
30136832 DC |
730 | prev = NULL; |
731 | while (blip != NULL) { | |
732 | if (lip->li_cb != xfs_iflush_done) { | |
733 | prev = blip; | |
734 | blip = blip->li_bio_list; | |
735 | continue; | |
736 | } | |
737 | ||
738 | /* remove from list */ | |
739 | next = blip->li_bio_list; | |
740 | if (!prev) { | |
adadbeef | 741 | bp->b_fspriv = next; |
30136832 DC |
742 | } else { |
743 | prev->li_bio_list = next; | |
744 | } | |
745 | ||
746 | /* add to current list */ | |
747 | blip->li_bio_list = lip->li_bio_list; | |
748 | lip->li_bio_list = blip; | |
749 | ||
750 | /* | |
751 | * while we have the item, do the unlocked check for needing | |
752 | * the AIL lock. | |
753 | */ | |
754 | iip = INODE_ITEM(blip); | |
755 | if (iip->ili_logged && blip->li_lsn == iip->ili_flush_lsn) | |
756 | need_ail++; | |
757 | ||
758 | blip = next; | |
759 | } | |
760 | ||
761 | /* make sure we capture the state of the initial inode. */ | |
762 | iip = INODE_ITEM(lip); | |
763 | if (iip->ili_logged && lip->li_lsn == iip->ili_flush_lsn) | |
764 | need_ail++; | |
1da177e4 LT |
765 | |
766 | /* | |
767 | * We only want to pull the item from the AIL if it is | |
768 | * actually there and its location in the log has not | |
769 | * changed since we started the flush. Thus, we only bother | |
770 | * if the ili_logged flag is set and the inode's lsn has not | |
771 | * changed. First we check the lsn outside | |
772 | * the lock since it's cheaper, and then we recheck while | |
773 | * holding the lock before removing the inode from the AIL. | |
774 | */ | |
30136832 DC |
775 | if (need_ail) { |
776 | struct xfs_log_item *log_items[need_ail]; | |
777 | int i = 0; | |
783a2f65 | 778 | spin_lock(&ailp->xa_lock); |
30136832 DC |
779 | for (blip = lip; blip; blip = blip->li_bio_list) { |
780 | iip = INODE_ITEM(blip); | |
781 | if (iip->ili_logged && | |
782 | blip->li_lsn == iip->ili_flush_lsn) { | |
783 | log_items[i++] = blip; | |
784 | } | |
785 | ASSERT(i <= need_ail); | |
1da177e4 | 786 | } |
30136832 | 787 | /* xfs_trans_ail_delete_bulk() drops the AIL lock. */ |
04913fdd DC |
788 | xfs_trans_ail_delete_bulk(ailp, log_items, i, |
789 | SHUTDOWN_CORRUPT_INCORE); | |
1da177e4 LT |
790 | } |
791 | ||
1da177e4 LT |
792 | |
793 | /* | |
30136832 DC |
794 | * clean up and unlock the flush lock now we are done. We can clear the |
795 | * ili_last_fields bits now that we know that the data corresponding to | |
796 | * them is safely on disk. | |
1da177e4 | 797 | */ |
30136832 DC |
798 | for (blip = lip; blip; blip = next) { |
799 | next = blip->li_bio_list; | |
800 | blip->li_bio_list = NULL; | |
801 | ||
802 | iip = INODE_ITEM(blip); | |
803 | iip->ili_logged = 0; | |
804 | iip->ili_last_fields = 0; | |
805 | xfs_ifunlock(iip->ili_inode); | |
806 | } | |
1da177e4 LT |
807 | } |
808 | ||
809 | /* | |
04913fdd DC |
810 | * This is the inode flushing abort routine. It is called from xfs_iflush when |
811 | * the filesystem is shutting down to clean up the inode state. It is | |
812 | * responsible for removing the inode item from the AIL if it has not been | |
813 | * re-logged, and unlocking the inode's flush lock. | |
1da177e4 LT |
814 | */ |
815 | void | |
816 | xfs_iflush_abort( | |
04913fdd DC |
817 | xfs_inode_t *ip, |
818 | bool stale) | |
1da177e4 | 819 | { |
783a2f65 | 820 | xfs_inode_log_item_t *iip = ip->i_itemp; |
1da177e4 | 821 | |
1da177e4 | 822 | if (iip) { |
783a2f65 | 823 | struct xfs_ail *ailp = iip->ili_item.li_ailp; |
1da177e4 | 824 | if (iip->ili_item.li_flags & XFS_LI_IN_AIL) { |
783a2f65 | 825 | spin_lock(&ailp->xa_lock); |
1da177e4 | 826 | if (iip->ili_item.li_flags & XFS_LI_IN_AIL) { |
783a2f65 | 827 | /* xfs_trans_ail_delete() drops the AIL lock. */ |
04913fdd DC |
828 | xfs_trans_ail_delete(ailp, &iip->ili_item, |
829 | stale ? | |
830 | SHUTDOWN_LOG_IO_ERROR : | |
831 | SHUTDOWN_CORRUPT_INCORE); | |
1da177e4 | 832 | } else |
783a2f65 | 833 | spin_unlock(&ailp->xa_lock); |
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; |
1da177e4 LT |
846 | } |
847 | /* | |
848 | * Release the inode's flush lock since we're done with it. | |
849 | */ | |
850 | xfs_ifunlock(ip); | |
851 | } | |
852 | ||
853 | void | |
854 | xfs_istale_done( | |
ca30b2a7 CH |
855 | struct xfs_buf *bp, |
856 | struct xfs_log_item *lip) | |
1da177e4 | 857 | { |
04913fdd | 858 | xfs_iflush_abort(INODE_ITEM(lip)->ili_inode, true); |
1da177e4 | 859 | } |
6d192a9b TS |
860 | |
861 | /* | |
862 | * convert an xfs_inode_log_format struct from either 32 or 64 bit versions | |
863 | * (which can have different field alignments) to the native version | |
864 | */ | |
865 | int | |
866 | xfs_inode_item_format_convert( | |
867 | xfs_log_iovec_t *buf, | |
868 | xfs_inode_log_format_t *in_f) | |
869 | { | |
870 | if (buf->i_len == sizeof(xfs_inode_log_format_32_t)) { | |
4e0d5f92 | 871 | xfs_inode_log_format_32_t *in_f32 = buf->i_addr; |
6d192a9b | 872 | |
6d192a9b TS |
873 | in_f->ilf_type = in_f32->ilf_type; |
874 | in_f->ilf_size = in_f32->ilf_size; | |
875 | in_f->ilf_fields = in_f32->ilf_fields; | |
876 | in_f->ilf_asize = in_f32->ilf_asize; | |
877 | in_f->ilf_dsize = in_f32->ilf_dsize; | |
878 | in_f->ilf_ino = in_f32->ilf_ino; | |
879 | /* copy biggest field of ilf_u */ | |
880 | memcpy(in_f->ilf_u.ilfu_uuid.__u_bits, | |
881 | in_f32->ilf_u.ilfu_uuid.__u_bits, | |
882 | sizeof(uuid_t)); | |
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; | |
887 | } else if (buf->i_len == sizeof(xfs_inode_log_format_64_t)){ | |
4e0d5f92 | 888 | xfs_inode_log_format_64_t *in_f64 = buf->i_addr; |
6d192a9b | 889 | |
6d192a9b TS |
890 | in_f->ilf_type = in_f64->ilf_type; |
891 | in_f->ilf_size = in_f64->ilf_size; | |
892 | in_f->ilf_fields = in_f64->ilf_fields; | |
893 | in_f->ilf_asize = in_f64->ilf_asize; | |
894 | in_f->ilf_dsize = in_f64->ilf_dsize; | |
895 | in_f->ilf_ino = in_f64->ilf_ino; | |
896 | /* copy biggest field of ilf_u */ | |
897 | memcpy(in_f->ilf_u.ilfu_uuid.__u_bits, | |
898 | in_f64->ilf_u.ilfu_uuid.__u_bits, | |
899 | sizeof(uuid_t)); | |
900 | in_f->ilf_blkno = in_f64->ilf_blkno; | |
901 | in_f->ilf_len = in_f64->ilf_len; | |
902 | in_f->ilf_boffset = in_f64->ilf_boffset; | |
903 | return 0; | |
904 | } | |
905 | return EFSCORRUPTED; | |
906 | } |