Commit | Line | Data |
---|---|---|
0b61f8a4 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 | 2 | /* |
7b718769 NS |
3 | * Copyright (c) 2000-2002,2005 Silicon Graphics, Inc. |
4 | * All Rights Reserved. | |
1da177e4 | 5 | */ |
1da177e4 | 6 | #include "xfs.h" |
a844f451 | 7 | #include "xfs_fs.h" |
5467b34b | 8 | #include "xfs_shared.h" |
a4fbe6ab | 9 | #include "xfs_format.h" |
239880ef DC |
10 | #include "xfs_log_format.h" |
11 | #include "xfs_trans_resv.h" | |
1da177e4 | 12 | #include "xfs_mount.h" |
1da177e4 | 13 | #include "xfs_inode.h" |
239880ef | 14 | #include "xfs_trans.h" |
a844f451 | 15 | #include "xfs_inode_item.h" |
0b1b213f | 16 | #include "xfs_trace.h" |
239880ef | 17 | #include "xfs_trans_priv.h" |
d3a304b6 | 18 | #include "xfs_buf_item.h" |
1234351c | 19 | #include "xfs_log.h" |
a5155b87 | 20 | #include "xfs_error.h" |
1da177e4 | 21 | |
f0e28280 | 22 | #include <linux/iversion.h> |
1da177e4 LT |
23 | |
24 | kmem_zone_t *xfs_ili_zone; /* inode log item zone */ | |
25 | ||
7bfa31d8 CH |
26 | static inline struct xfs_inode_log_item *INODE_ITEM(struct xfs_log_item *lip) |
27 | { | |
28 | return container_of(lip, struct xfs_inode_log_item, ili_item); | |
29 | } | |
30 | ||
166d1368 | 31 | STATIC void |
ce9641d6 CH |
32 | xfs_inode_item_data_fork_size( |
33 | struct xfs_inode_log_item *iip, | |
166d1368 DC |
34 | int *nvecs, |
35 | int *nbytes) | |
1da177e4 | 36 | { |
7bfa31d8 | 37 | struct xfs_inode *ip = iip->ili_inode; |
166d1368 | 38 | |
f7e67b20 | 39 | switch (ip->i_df.if_format) { |
1da177e4 | 40 | case XFS_DINODE_FMT_EXTENTS: |
f5d8d5c4 | 41 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
daf83964 | 42 | ip->i_df.if_nextents > 0 && |
166d1368 DC |
43 | ip->i_df.if_bytes > 0) { |
44 | /* worst case, doesn't subtract delalloc extents */ | |
45 | *nbytes += XFS_IFORK_DSIZE(ip); | |
46 | *nvecs += 1; | |
47 | } | |
1da177e4 | 48 | break; |
1da177e4 | 49 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 50 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
166d1368 DC |
51 | ip->i_df.if_broot_bytes > 0) { |
52 | *nbytes += ip->i_df.if_broot_bytes; | |
53 | *nvecs += 1; | |
54 | } | |
1da177e4 | 55 | break; |
1da177e4 | 56 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 57 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
166d1368 DC |
58 | ip->i_df.if_bytes > 0) { |
59 | *nbytes += roundup(ip->i_df.if_bytes, 4); | |
60 | *nvecs += 1; | |
61 | } | |
1da177e4 LT |
62 | break; |
63 | ||
64 | case XFS_DINODE_FMT_DEV: | |
1da177e4 | 65 | break; |
1da177e4 LT |
66 | default: |
67 | ASSERT(0); | |
68 | break; | |
69 | } | |
ce9641d6 | 70 | } |
1da177e4 | 71 | |
ce9641d6 CH |
72 | STATIC void |
73 | xfs_inode_item_attr_fork_size( | |
74 | struct xfs_inode_log_item *iip, | |
75 | int *nvecs, | |
76 | int *nbytes) | |
77 | { | |
78 | struct xfs_inode *ip = iip->ili_inode; | |
1da177e4 | 79 | |
f7e67b20 | 80 | switch (ip->i_afp->if_format) { |
1da177e4 | 81 | case XFS_DINODE_FMT_EXTENTS: |
f5d8d5c4 | 82 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
daf83964 | 83 | ip->i_afp->if_nextents > 0 && |
166d1368 DC |
84 | ip->i_afp->if_bytes > 0) { |
85 | /* worst case, doesn't subtract unused space */ | |
86 | *nbytes += XFS_IFORK_ASIZE(ip); | |
87 | *nvecs += 1; | |
88 | } | |
1da177e4 | 89 | break; |
1da177e4 | 90 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 91 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
166d1368 DC |
92 | ip->i_afp->if_broot_bytes > 0) { |
93 | *nbytes += ip->i_afp->if_broot_bytes; | |
94 | *nvecs += 1; | |
95 | } | |
1da177e4 | 96 | break; |
1da177e4 | 97 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 98 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
166d1368 DC |
99 | ip->i_afp->if_bytes > 0) { |
100 | *nbytes += roundup(ip->i_afp->if_bytes, 4); | |
101 | *nvecs += 1; | |
102 | } | |
1da177e4 | 103 | break; |
1da177e4 LT |
104 | default: |
105 | ASSERT(0); | |
106 | break; | |
107 | } | |
1da177e4 LT |
108 | } |
109 | ||
ce9641d6 CH |
110 | /* |
111 | * This returns the number of iovecs needed to log the given inode item. | |
112 | * | |
113 | * We need one iovec for the inode log format structure, one for the | |
114 | * inode core, and possibly one for the inode data/extents/b-tree root | |
115 | * and one for the inode attribute data/extents/b-tree root. | |
116 | */ | |
117 | STATIC void | |
118 | xfs_inode_item_size( | |
119 | struct xfs_log_item *lip, | |
120 | int *nvecs, | |
121 | int *nbytes) | |
122 | { | |
123 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
124 | struct xfs_inode *ip = iip->ili_inode; | |
125 | ||
126 | *nvecs += 2; | |
127 | *nbytes += sizeof(struct xfs_inode_log_format) + | |
e9e2eae8 | 128 | xfs_log_dinode_size(ip->i_mount); |
ce9641d6 CH |
129 | |
130 | xfs_inode_item_data_fork_size(iip, nvecs, nbytes); | |
131 | if (XFS_IFORK_Q(ip)) | |
132 | xfs_inode_item_attr_fork_size(iip, nvecs, nbytes); | |
133 | } | |
134 | ||
1234351c | 135 | STATIC void |
3de559fb CH |
136 | xfs_inode_item_format_data_fork( |
137 | struct xfs_inode_log_item *iip, | |
bde7cff6 CH |
138 | struct xfs_inode_log_format *ilf, |
139 | struct xfs_log_vec *lv, | |
140 | struct xfs_log_iovec **vecp) | |
1da177e4 | 141 | { |
7bfa31d8 | 142 | struct xfs_inode *ip = iip->ili_inode; |
1da177e4 | 143 | size_t data_bytes; |
1da177e4 | 144 | |
f7e67b20 | 145 | switch (ip->i_df.if_format) { |
1da177e4 | 146 | case XFS_DINODE_FMT_EXTENTS: |
f5d8d5c4 | 147 | iip->ili_fields &= |
42b67dc6 | 148 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | XFS_ILOG_DEV); |
339a5f5d | 149 | |
f5d8d5c4 | 150 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
daf83964 | 151 | ip->i_df.if_nextents > 0 && |
339a5f5d | 152 | ip->i_df.if_bytes > 0) { |
da776503 CH |
153 | struct xfs_bmbt_rec *p; |
154 | ||
5d829300 | 155 | ASSERT(xfs_iext_count(&ip->i_df) > 0); |
da776503 CH |
156 | |
157 | p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IEXT); | |
158 | data_bytes = xfs_iextents_copy(ip, p, XFS_DATA_FORK); | |
159 | xlog_finish_iovec(lv, *vecp, data_bytes); | |
160 | ||
161 | ASSERT(data_bytes <= ip->i_df.if_bytes); | |
162 | ||
163 | ilf->ilf_dsize = data_bytes; | |
bde7cff6 | 164 | ilf->ilf_size++; |
339a5f5d | 165 | } else { |
f5d8d5c4 | 166 | iip->ili_fields &= ~XFS_ILOG_DEXT; |
1da177e4 LT |
167 | } |
168 | break; | |
1da177e4 | 169 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 170 | iip->ili_fields &= |
42b67dc6 | 171 | ~(XFS_ILOG_DDATA | XFS_ILOG_DEXT | XFS_ILOG_DEV); |
339a5f5d | 172 | |
f5d8d5c4 | 173 | if ((iip->ili_fields & XFS_ILOG_DBROOT) && |
339a5f5d | 174 | ip->i_df.if_broot_bytes > 0) { |
1da177e4 | 175 | ASSERT(ip->i_df.if_broot != NULL); |
bde7cff6 | 176 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IBROOT, |
1234351c CH |
177 | ip->i_df.if_broot, |
178 | ip->i_df.if_broot_bytes); | |
bde7cff6 CH |
179 | ilf->ilf_dsize = ip->i_df.if_broot_bytes; |
180 | ilf->ilf_size++; | |
339a5f5d | 181 | } else { |
f5d8d5c4 | 182 | ASSERT(!(iip->ili_fields & |
339a5f5d | 183 | XFS_ILOG_DBROOT)); |
f5d8d5c4 | 184 | iip->ili_fields &= ~XFS_ILOG_DBROOT; |
1da177e4 LT |
185 | } |
186 | break; | |
1da177e4 | 187 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 188 | iip->ili_fields &= |
42b67dc6 | 189 | ~(XFS_ILOG_DEXT | XFS_ILOG_DBROOT | XFS_ILOG_DEV); |
f5d8d5c4 | 190 | if ((iip->ili_fields & XFS_ILOG_DDATA) && |
339a5f5d | 191 | ip->i_df.if_bytes > 0) { |
1da177e4 LT |
192 | /* |
193 | * Round i_bytes up to a word boundary. | |
194 | * The underlying memory is guaranteed to | |
195 | * to be there by xfs_idata_realloc(). | |
196 | */ | |
197 | data_bytes = roundup(ip->i_df.if_bytes, 4); | |
1234351c CH |
198 | ASSERT(ip->i_df.if_u1.if_data != NULL); |
199 | ASSERT(ip->i_d.di_size > 0); | |
bde7cff6 | 200 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_ILOCAL, |
1234351c | 201 | ip->i_df.if_u1.if_data, data_bytes); |
bde7cff6 CH |
202 | ilf->ilf_dsize = (unsigned)data_bytes; |
203 | ilf->ilf_size++; | |
339a5f5d | 204 | } else { |
f5d8d5c4 | 205 | iip->ili_fields &= ~XFS_ILOG_DDATA; |
1da177e4 LT |
206 | } |
207 | break; | |
1da177e4 | 208 | case XFS_DINODE_FMT_DEV: |
f5d8d5c4 | 209 | iip->ili_fields &= |
42b67dc6 | 210 | ~(XFS_ILOG_DDATA | XFS_ILOG_DBROOT | XFS_ILOG_DEXT); |
bde7cff6 | 211 | if (iip->ili_fields & XFS_ILOG_DEV) |
66f36464 | 212 | ilf->ilf_u.ilfu_rdev = sysv_encode_dev(VFS_I(ip)->i_rdev); |
1da177e4 | 213 | break; |
1da177e4 LT |
214 | default: |
215 | ASSERT(0); | |
216 | break; | |
217 | } | |
3de559fb CH |
218 | } |
219 | ||
1234351c | 220 | STATIC void |
3de559fb CH |
221 | xfs_inode_item_format_attr_fork( |
222 | struct xfs_inode_log_item *iip, | |
bde7cff6 CH |
223 | struct xfs_inode_log_format *ilf, |
224 | struct xfs_log_vec *lv, | |
225 | struct xfs_log_iovec **vecp) | |
3de559fb CH |
226 | { |
227 | struct xfs_inode *ip = iip->ili_inode; | |
228 | size_t data_bytes; | |
1da177e4 | 229 | |
f7e67b20 | 230 | switch (ip->i_afp->if_format) { |
1da177e4 | 231 | case XFS_DINODE_FMT_EXTENTS: |
f5d8d5c4 | 232 | iip->ili_fields &= |
339a5f5d CH |
233 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT); |
234 | ||
f5d8d5c4 | 235 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
daf83964 | 236 | ip->i_afp->if_nextents > 0 && |
339a5f5d | 237 | ip->i_afp->if_bytes > 0) { |
da776503 CH |
238 | struct xfs_bmbt_rec *p; |
239 | ||
5d829300 | 240 | ASSERT(xfs_iext_count(ip->i_afp) == |
daf83964 | 241 | ip->i_afp->if_nextents); |
da776503 CH |
242 | |
243 | p = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_EXT); | |
244 | data_bytes = xfs_iextents_copy(ip, p, XFS_ATTR_FORK); | |
245 | xlog_finish_iovec(lv, *vecp, data_bytes); | |
246 | ||
247 | ilf->ilf_asize = data_bytes; | |
bde7cff6 | 248 | ilf->ilf_size++; |
339a5f5d | 249 | } else { |
f5d8d5c4 | 250 | iip->ili_fields &= ~XFS_ILOG_AEXT; |
1da177e4 LT |
251 | } |
252 | break; | |
1da177e4 | 253 | case XFS_DINODE_FMT_BTREE: |
f5d8d5c4 | 254 | iip->ili_fields &= |
339a5f5d CH |
255 | ~(XFS_ILOG_ADATA | XFS_ILOG_AEXT); |
256 | ||
f5d8d5c4 | 257 | if ((iip->ili_fields & XFS_ILOG_ABROOT) && |
339a5f5d | 258 | ip->i_afp->if_broot_bytes > 0) { |
1da177e4 | 259 | ASSERT(ip->i_afp->if_broot != NULL); |
339a5f5d | 260 | |
bde7cff6 | 261 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_BROOT, |
1234351c CH |
262 | ip->i_afp->if_broot, |
263 | ip->i_afp->if_broot_bytes); | |
bde7cff6 CH |
264 | ilf->ilf_asize = ip->i_afp->if_broot_bytes; |
265 | ilf->ilf_size++; | |
339a5f5d | 266 | } else { |
f5d8d5c4 | 267 | iip->ili_fields &= ~XFS_ILOG_ABROOT; |
1da177e4 LT |
268 | } |
269 | break; | |
1da177e4 | 270 | case XFS_DINODE_FMT_LOCAL: |
f5d8d5c4 | 271 | iip->ili_fields &= |
339a5f5d CH |
272 | ~(XFS_ILOG_AEXT | XFS_ILOG_ABROOT); |
273 | ||
f5d8d5c4 | 274 | if ((iip->ili_fields & XFS_ILOG_ADATA) && |
339a5f5d | 275 | ip->i_afp->if_bytes > 0) { |
1da177e4 LT |
276 | /* |
277 | * Round i_bytes up to a word boundary. | |
278 | * The underlying memory is guaranteed to | |
279 | * to be there by xfs_idata_realloc(). | |
280 | */ | |
281 | data_bytes = roundup(ip->i_afp->if_bytes, 4); | |
1234351c | 282 | ASSERT(ip->i_afp->if_u1.if_data != NULL); |
bde7cff6 | 283 | xlog_copy_iovec(lv, vecp, XLOG_REG_TYPE_IATTR_LOCAL, |
1234351c CH |
284 | ip->i_afp->if_u1.if_data, |
285 | data_bytes); | |
bde7cff6 CH |
286 | ilf->ilf_asize = (unsigned)data_bytes; |
287 | ilf->ilf_size++; | |
339a5f5d | 288 | } else { |
f5d8d5c4 | 289 | iip->ili_fields &= ~XFS_ILOG_ADATA; |
1da177e4 LT |
290 | } |
291 | break; | |
1da177e4 LT |
292 | default: |
293 | ASSERT(0); | |
294 | break; | |
295 | } | |
3de559fb CH |
296 | } |
297 | ||
f8d55aa0 | 298 | static void |
3987848c DC |
299 | xfs_inode_to_log_dinode( |
300 | struct xfs_inode *ip, | |
93f958f9 DC |
301 | struct xfs_log_dinode *to, |
302 | xfs_lsn_t lsn) | |
f8d55aa0 | 303 | { |
3987848c DC |
304 | struct xfs_icdinode *from = &ip->i_d; |
305 | struct inode *inode = VFS_I(ip); | |
306 | ||
93f958f9 | 307 | to->di_magic = XFS_DINODE_MAGIC; |
f7e67b20 | 308 | to->di_format = xfs_ifork_format(&ip->i_df); |
ba8adad5 CH |
309 | to->di_uid = i_uid_read(inode); |
310 | to->di_gid = i_gid_read(inode); | |
de7a866f CH |
311 | to->di_projid_lo = from->di_projid & 0xffff; |
312 | to->di_projid_hi = from->di_projid >> 16; | |
f8d55aa0 | 313 | |
93f958f9 | 314 | memset(to->di_pad, 0, sizeof(to->di_pad)); |
faeb4e47 | 315 | memset(to->di_pad3, 0, sizeof(to->di_pad3)); |
3987848c DC |
316 | to->di_atime.t_sec = inode->i_atime.tv_sec; |
317 | to->di_atime.t_nsec = inode->i_atime.tv_nsec; | |
318 | to->di_mtime.t_sec = inode->i_mtime.tv_sec; | |
319 | to->di_mtime.t_nsec = inode->i_mtime.tv_nsec; | |
320 | to->di_ctime.t_sec = inode->i_ctime.tv_sec; | |
321 | to->di_ctime.t_nsec = inode->i_ctime.tv_nsec; | |
54d7b5c1 | 322 | to->di_nlink = inode->i_nlink; |
9e9a2674 | 323 | to->di_gen = inode->i_generation; |
c19b3b05 | 324 | to->di_mode = inode->i_mode; |
f8d55aa0 | 325 | |
f8d55aa0 DC |
326 | to->di_size = from->di_size; |
327 | to->di_nblocks = from->di_nblocks; | |
328 | to->di_extsize = from->di_extsize; | |
daf83964 CH |
329 | to->di_nextents = xfs_ifork_nextents(&ip->i_df); |
330 | to->di_anextents = xfs_ifork_nextents(ip->i_afp); | |
f8d55aa0 | 331 | to->di_forkoff = from->di_forkoff; |
f7e67b20 | 332 | to->di_aformat = xfs_ifork_format(ip->i_afp); |
f8d55aa0 DC |
333 | to->di_dmevmask = from->di_dmevmask; |
334 | to->di_dmstate = from->di_dmstate; | |
335 | to->di_flags = from->di_flags; | |
f8d55aa0 | 336 | |
20413e37 DC |
337 | /* log a dummy value to ensure log structure is fully initialised */ |
338 | to->di_next_unlinked = NULLAGINO; | |
339 | ||
6471e9c5 CH |
340 | if (xfs_sb_version_has_v3inode(&ip->i_mount->m_sb)) { |
341 | to->di_version = 3; | |
f0e28280 | 342 | to->di_changecount = inode_peek_iversion(inode); |
8d2d878d CH |
343 | to->di_crtime.t_sec = from->di_crtime.tv_sec; |
344 | to->di_crtime.t_nsec = from->di_crtime.tv_nsec; | |
f8d55aa0 | 345 | to->di_flags2 = from->di_flags2; |
f7ca3522 | 346 | to->di_cowextsize = from->di_cowextsize; |
93f958f9 DC |
347 | to->di_ino = ip->i_ino; |
348 | to->di_lsn = lsn; | |
349 | memset(to->di_pad2, 0, sizeof(to->di_pad2)); | |
350 | uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid); | |
f8d55aa0 DC |
351 | to->di_flushiter = 0; |
352 | } else { | |
6471e9c5 | 353 | to->di_version = 2; |
f8d55aa0 DC |
354 | to->di_flushiter = from->di_flushiter; |
355 | } | |
356 | } | |
357 | ||
358 | /* | |
359 | * Format the inode core. Current timestamp data is only in the VFS inode | |
360 | * fields, so we need to grab them from there. Hence rather than just copying | |
361 | * the XFS inode core structure, format the fields directly into the iovec. | |
362 | */ | |
363 | static void | |
364 | xfs_inode_item_format_core( | |
365 | struct xfs_inode *ip, | |
366 | struct xfs_log_vec *lv, | |
367 | struct xfs_log_iovec **vecp) | |
368 | { | |
369 | struct xfs_log_dinode *dic; | |
370 | ||
371 | dic = xlog_prepare_iovec(lv, vecp, XLOG_REG_TYPE_ICORE); | |
93f958f9 | 372 | xfs_inode_to_log_dinode(ip, dic, ip->i_itemp->ili_item.li_lsn); |
e9e2eae8 | 373 | xlog_finish_iovec(lv, *vecp, xfs_log_dinode_size(ip->i_mount)); |
f8d55aa0 DC |
374 | } |
375 | ||
3de559fb CH |
376 | /* |
377 | * This is called to fill in the vector of log iovecs for the given inode | |
378 | * log item. It fills the first item with an inode log format structure, | |
379 | * the second with the on-disk inode structure, and a possible third and/or | |
380 | * fourth with the inode data/extents/b-tree root and inode attributes | |
381 | * data/extents/b-tree root. | |
20413e37 DC |
382 | * |
383 | * Note: Always use the 64 bit inode log format structure so we don't | |
384 | * leave an uninitialised hole in the format item on 64 bit systems. Log | |
385 | * recovery on 32 bit systems handles this just fine, so there's no reason | |
386 | * for not using an initialising the properly padded structure all the time. | |
3de559fb CH |
387 | */ |
388 | STATIC void | |
389 | xfs_inode_item_format( | |
390 | struct xfs_log_item *lip, | |
bde7cff6 | 391 | struct xfs_log_vec *lv) |
3de559fb CH |
392 | { |
393 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); | |
394 | struct xfs_inode *ip = iip->ili_inode; | |
bde7cff6 | 395 | struct xfs_log_iovec *vecp = NULL; |
20413e37 | 396 | struct xfs_inode_log_format *ilf; |
3de559fb | 397 | |
2f251293 CH |
398 | ilf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_IFORMAT); |
399 | ilf->ilf_type = XFS_LI_INODE; | |
400 | ilf->ilf_ino = ip->i_ino; | |
401 | ilf->ilf_blkno = ip->i_imap.im_blkno; | |
402 | ilf->ilf_len = ip->i_imap.im_len; | |
403 | ilf->ilf_boffset = ip->i_imap.im_boffset; | |
404 | ilf->ilf_fields = XFS_ILOG_CORE; | |
405 | ilf->ilf_size = 2; /* format + core */ | |
20413e37 DC |
406 | |
407 | /* | |
408 | * make sure we don't leak uninitialised data into the log in the case | |
409 | * when we don't log every field in the inode. | |
410 | */ | |
411 | ilf->ilf_dsize = 0; | |
412 | ilf->ilf_asize = 0; | |
413 | ilf->ilf_pad = 0; | |
42b67dc6 | 414 | memset(&ilf->ilf_u, 0, sizeof(ilf->ilf_u)); |
20413e37 DC |
415 | |
416 | xlog_finish_iovec(lv, vecp, sizeof(*ilf)); | |
3de559fb | 417 | |
f8d55aa0 | 418 | xfs_inode_item_format_core(ip, lv, &vecp); |
bde7cff6 | 419 | xfs_inode_item_format_data_fork(iip, ilf, lv, &vecp); |
3de559fb | 420 | if (XFS_IFORK_Q(ip)) { |
bde7cff6 | 421 | xfs_inode_item_format_attr_fork(iip, ilf, lv, &vecp); |
3de559fb CH |
422 | } else { |
423 | iip->ili_fields &= | |
424 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT); | |
425 | } | |
426 | ||
2f251293 CH |
427 | /* update the format with the exact fields we actually logged */ |
428 | ilf->ilf_fields |= (iip->ili_fields & ~XFS_ILOG_TIMESTAMP); | |
1da177e4 LT |
429 | } |
430 | ||
1da177e4 LT |
431 | /* |
432 | * This is called to pin the inode associated with the inode log | |
a14a5ab5 | 433 | * item in memory so it cannot be written out. |
1da177e4 LT |
434 | */ |
435 | STATIC void | |
436 | xfs_inode_item_pin( | |
7bfa31d8 | 437 | struct xfs_log_item *lip) |
1da177e4 | 438 | { |
7bfa31d8 | 439 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 440 | |
7bfa31d8 CH |
441 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); |
442 | ||
443 | trace_xfs_inode_pin(ip, _RET_IP_); | |
444 | atomic_inc(&ip->i_pincount); | |
1da177e4 LT |
445 | } |
446 | ||
447 | ||
448 | /* | |
449 | * This is called to unpin the inode associated with the inode log | |
450 | * item which was previously pinned with a call to xfs_inode_item_pin(). | |
a14a5ab5 CH |
451 | * |
452 | * Also wake up anyone in xfs_iunpin_wait() if the count goes to 0. | |
1da177e4 | 453 | */ |
1da177e4 LT |
454 | STATIC void |
455 | xfs_inode_item_unpin( | |
7bfa31d8 | 456 | struct xfs_log_item *lip, |
9412e318 | 457 | int remove) |
1da177e4 | 458 | { |
7bfa31d8 | 459 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 460 | |
4aaf15d1 | 461 | trace_xfs_inode_unpin(ip, _RET_IP_); |
a14a5ab5 CH |
462 | ASSERT(atomic_read(&ip->i_pincount) > 0); |
463 | if (atomic_dec_and_test(&ip->i_pincount)) | |
f392e631 | 464 | wake_up_bit(&ip->i_flags, __XFS_IPINNED_BIT); |
1da177e4 LT |
465 | } |
466 | ||
1da177e4 | 467 | STATIC uint |
43ff2122 CH |
468 | xfs_inode_item_push( |
469 | struct xfs_log_item *lip, | |
470 | struct list_head *buffer_list) | |
57e80956 MW |
471 | __releases(&lip->li_ailp->ail_lock) |
472 | __acquires(&lip->li_ailp->ail_lock) | |
1da177e4 | 473 | { |
7bfa31d8 CH |
474 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
475 | struct xfs_inode *ip = iip->ili_inode; | |
d3a304b6 | 476 | struct xfs_buf *bp = lip->li_buf; |
43ff2122 CH |
477 | uint rval = XFS_ITEM_SUCCESS; |
478 | int error; | |
1da177e4 | 479 | |
7bfa31d8 | 480 | if (xfs_ipincount(ip) > 0) |
1da177e4 | 481 | return XFS_ITEM_PINNED; |
1da177e4 | 482 | |
7bfa31d8 | 483 | if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED)) |
1da177e4 | 484 | return XFS_ITEM_LOCKED; |
1da177e4 | 485 | |
4c46819a CH |
486 | /* |
487 | * Re-check the pincount now that we stabilized the value by | |
488 | * taking the ilock. | |
489 | */ | |
490 | if (xfs_ipincount(ip) > 0) { | |
43ff2122 CH |
491 | rval = XFS_ITEM_PINNED; |
492 | goto out_unlock; | |
4c46819a CH |
493 | } |
494 | ||
9a3a5dab BF |
495 | /* |
496 | * Stale inode items should force out the iclog. | |
497 | */ | |
498 | if (ip->i_flags & XFS_ISTALE) { | |
499 | rval = XFS_ITEM_PINNED; | |
500 | goto out_unlock; | |
501 | } | |
502 | ||
43ff2122 CH |
503 | /* |
504 | * Someone else is already flushing the inode. Nothing we can do | |
505 | * here but wait for the flush to finish and remove the item from | |
506 | * the AIL. | |
507 | */ | |
1da177e4 | 508 | if (!xfs_iflock_nowait(ip)) { |
43ff2122 CH |
509 | rval = XFS_ITEM_FLUSHING; |
510 | goto out_unlock; | |
1da177e4 LT |
511 | } |
512 | ||
43ff2122 | 513 | ASSERT(iip->ili_fields != 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); |
57e80956 | 514 | spin_unlock(&lip->li_ailp->ail_lock); |
43ff2122 CH |
515 | |
516 | error = xfs_iflush(ip, &bp); | |
517 | if (!error) { | |
518 | if (!xfs_buf_delwri_queue(bp, buffer_list)) | |
519 | rval = XFS_ITEM_FLUSHING; | |
520 | xfs_buf_relse(bp); | |
d4bc4c5f BF |
521 | } else if (error == -EAGAIN) |
522 | rval = XFS_ITEM_LOCKED; | |
43ff2122 | 523 | |
57e80956 | 524 | spin_lock(&lip->li_ailp->ail_lock); |
43ff2122 CH |
525 | out_unlock: |
526 | xfs_iunlock(ip, XFS_ILOCK_SHARED); | |
527 | return rval; | |
1da177e4 LT |
528 | } |
529 | ||
530 | /* | |
531 | * Unlock the inode associated with the inode log item. | |
1da177e4 LT |
532 | */ |
533 | STATIC void | |
ddf92053 | 534 | xfs_inode_item_release( |
7bfa31d8 | 535 | struct xfs_log_item *lip) |
1da177e4 | 536 | { |
7bfa31d8 CH |
537 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
538 | struct xfs_inode *ip = iip->ili_inode; | |
898621d5 | 539 | unsigned short lock_flags; |
1da177e4 | 540 | |
f3ca8738 CH |
541 | ASSERT(ip->i_itemp != NULL); |
542 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); | |
1da177e4 | 543 | |
898621d5 CH |
544 | lock_flags = iip->ili_lock_flags; |
545 | iip->ili_lock_flags = 0; | |
ddc3415a | 546 | if (lock_flags) |
f3ca8738 | 547 | xfs_iunlock(ip, lock_flags); |
1da177e4 LT |
548 | } |
549 | ||
550 | /* | |
de25c181 DC |
551 | * This is called to find out where the oldest active copy of the inode log |
552 | * item in the on disk log resides now that the last log write of it completed | |
553 | * at the given lsn. Since we always re-log all dirty data in an inode, the | |
554 | * latest copy in the on disk log is the only one that matters. Therefore, | |
555 | * simply return the given lsn. | |
556 | * | |
557 | * If the inode has been marked stale because the cluster is being freed, we | |
558 | * don't want to (re-)insert this inode into the AIL. There is a race condition | |
559 | * where the cluster buffer may be unpinned before the inode is inserted into | |
560 | * the AIL during transaction committed processing. If the buffer is unpinned | |
561 | * before the inode item has been committed and inserted, then it is possible | |
1316d4da | 562 | * for the buffer to be written and IO completes before the inode is inserted |
de25c181 DC |
563 | * into the AIL. In that case, we'd be inserting a clean, stale inode into the |
564 | * AIL which will never get removed. It will, however, get reclaimed which | |
565 | * triggers an assert in xfs_inode_free() complaining about freein an inode | |
566 | * still in the AIL. | |
567 | * | |
1316d4da DC |
568 | * To avoid this, just unpin the inode directly and return a LSN of -1 so the |
569 | * transaction committed code knows that it does not need to do any further | |
570 | * processing on the item. | |
1da177e4 | 571 | */ |
1da177e4 LT |
572 | STATIC xfs_lsn_t |
573 | xfs_inode_item_committed( | |
7bfa31d8 | 574 | struct xfs_log_item *lip, |
1da177e4 LT |
575 | xfs_lsn_t lsn) |
576 | { | |
de25c181 DC |
577 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
578 | struct xfs_inode *ip = iip->ili_inode; | |
579 | ||
1316d4da DC |
580 | if (xfs_iflags_test(ip, XFS_ISTALE)) { |
581 | xfs_inode_item_unpin(lip, 0); | |
582 | return -1; | |
583 | } | |
7bfa31d8 | 584 | return lsn; |
1da177e4 LT |
585 | } |
586 | ||
1da177e4 LT |
587 | STATIC void |
588 | xfs_inode_item_committing( | |
7bfa31d8 | 589 | struct xfs_log_item *lip, |
ddf92053 | 590 | xfs_lsn_t commit_lsn) |
1da177e4 | 591 | { |
ddf92053 CH |
592 | INODE_ITEM(lip)->ili_last_lsn = commit_lsn; |
593 | return xfs_inode_item_release(lip); | |
1da177e4 LT |
594 | } |
595 | ||
272e42b2 | 596 | static const struct xfs_item_ops xfs_inode_item_ops = { |
7bfa31d8 CH |
597 | .iop_size = xfs_inode_item_size, |
598 | .iop_format = xfs_inode_item_format, | |
599 | .iop_pin = xfs_inode_item_pin, | |
600 | .iop_unpin = xfs_inode_item_unpin, | |
ddf92053 | 601 | .iop_release = xfs_inode_item_release, |
7bfa31d8 CH |
602 | .iop_committed = xfs_inode_item_committed, |
603 | .iop_push = xfs_inode_item_push, | |
ddf92053 | 604 | .iop_committing = xfs_inode_item_committing, |
1da177e4 LT |
605 | }; |
606 | ||
607 | ||
608 | /* | |
609 | * Initialize the inode log item for a newly allocated (in-core) inode. | |
610 | */ | |
611 | void | |
612 | xfs_inode_item_init( | |
7bfa31d8 CH |
613 | struct xfs_inode *ip, |
614 | struct xfs_mount *mp) | |
1da177e4 | 615 | { |
7bfa31d8 | 616 | struct xfs_inode_log_item *iip; |
1da177e4 LT |
617 | |
618 | ASSERT(ip->i_itemp == NULL); | |
707e0dda | 619 | iip = ip->i_itemp = kmem_zone_zalloc(xfs_ili_zone, 0); |
1da177e4 | 620 | |
1da177e4 | 621 | iip->ili_inode = ip; |
1319ebef | 622 | spin_lock_init(&iip->ili_lock); |
43f5efc5 DC |
623 | xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE, |
624 | &xfs_inode_item_ops); | |
1da177e4 LT |
625 | } |
626 | ||
627 | /* | |
628 | * Free the inode log item and any memory hanging off of it. | |
629 | */ | |
630 | void | |
631 | xfs_inode_item_destroy( | |
632 | xfs_inode_t *ip) | |
633 | { | |
b1c5ebb2 | 634 | kmem_free(ip->i_itemp->ili_item.li_lv_shadow); |
377bcd5f | 635 | kmem_cache_free(xfs_ili_zone, ip->i_itemp); |
1da177e4 LT |
636 | } |
637 | ||
638 | ||
639 | /* | |
640 | * This is the inode flushing I/O completion routine. It is called | |
641 | * from interrupt level when the buffer containing the inode is | |
642 | * flushed to disk. It is responsible for removing the inode item | |
643 | * from the AIL if it has not been re-logged, and unlocking the inode's | |
644 | * flush lock. | |
30136832 DC |
645 | * |
646 | * To reduce AIL lock traffic as much as possible, we scan the buffer log item | |
647 | * list for other inodes that will run this function. We remove them from the | |
648 | * buffer list so we can process all the inode IO completions in one AIL lock | |
649 | * traversal. | |
1da177e4 | 650 | */ |
1da177e4 LT |
651 | void |
652 | xfs_iflush_done( | |
aac855ab | 653 | struct xfs_buf *bp) |
1da177e4 | 654 | { |
30136832 | 655 | struct xfs_inode_log_item *iip; |
aac855ab DC |
656 | struct xfs_log_item *lip, *n; |
657 | struct xfs_ail *ailp = bp->b_mount->m_ail; | |
30136832 | 658 | int need_ail = 0; |
643c8c05 | 659 | LIST_HEAD(tmp); |
30136832 DC |
660 | |
661 | /* | |
aac855ab DC |
662 | * Pull the attached inodes from the buffer one at a time and take the |
663 | * appropriate action on them. | |
30136832 | 664 | */ |
aac855ab DC |
665 | list_for_each_entry_safe(lip, n, &bp->b_li_list, li_bio_list) { |
666 | iip = INODE_ITEM(lip); | |
667 | if (xfs_iflags_test(iip->ili_inode, XFS_ISTALE)) { | |
668 | list_del_init(&lip->li_bio_list); | |
669 | xfs_iflush_abort(iip->ili_inode); | |
643c8c05 | 670 | continue; |
aac855ab | 671 | } |
30136832 | 672 | |
aac855ab | 673 | list_move_tail(&lip->li_bio_list, &tmp); |
1dfde687 DC |
674 | |
675 | /* Do an unlocked check for needing the AIL lock. */ | |
aac855ab DC |
676 | if (lip->li_lsn == iip->ili_flush_lsn || |
677 | test_bit(XFS_LI_FAILED, &lip->li_flags)) | |
30136832 | 678 | need_ail++; |
30136832 | 679 | } |
aac855ab | 680 | ASSERT(list_empty(&bp->b_li_list)); |
1da177e4 LT |
681 | |
682 | /* | |
1dfde687 DC |
683 | * We only want to pull the item from the AIL if it is actually there |
684 | * and its location in the log has not changed since we started the | |
685 | * flush. Thus, we only bother if the inode's lsn has not changed. | |
1da177e4 | 686 | */ |
30136832 | 687 | if (need_ail) { |
8eb807bd | 688 | xfs_lsn_t tail_lsn = 0; |
27af1bbf CH |
689 | |
690 | /* this is an opencoded batch version of xfs_trans_ail_delete */ | |
57e80956 | 691 | spin_lock(&ailp->ail_lock); |
aac855ab | 692 | list_for_each_entry(lip, &tmp, li_bio_list) { |
e98084b8 | 693 | xfs_clear_li_failed(lip); |
aac855ab DC |
694 | if (lip->li_lsn == INODE_ITEM(lip)->ili_flush_lsn) { |
695 | xfs_lsn_t lsn = xfs_ail_delete_one(ailp, lip); | |
8eb807bd DC |
696 | if (!tail_lsn && lsn) |
697 | tail_lsn = lsn; | |
d3a304b6 | 698 | } |
1da177e4 | 699 | } |
8eb807bd | 700 | xfs_ail_update_finish(ailp, tail_lsn); |
27af1bbf | 701 | } |
1da177e4 LT |
702 | |
703 | /* | |
1dfde687 | 704 | * Clean up and unlock the flush lock now we are done. We can clear the |
30136832 DC |
705 | * ili_last_fields bits now that we know that the data corresponding to |
706 | * them is safely on disk. | |
1da177e4 | 707 | */ |
aac855ab DC |
708 | list_for_each_entry_safe(lip, n, &tmp, li_bio_list) { |
709 | list_del_init(&lip->li_bio_list); | |
710 | iip = INODE_ITEM(lip); | |
1319ebef DC |
711 | |
712 | spin_lock(&iip->ili_lock); | |
30136832 | 713 | iip->ili_last_fields = 0; |
1319ebef DC |
714 | spin_unlock(&iip->ili_lock); |
715 | ||
30136832 DC |
716 | xfs_ifunlock(iip->ili_inode); |
717 | } | |
1da177e4 LT |
718 | } |
719 | ||
720 | /* | |
04913fdd DC |
721 | * This is the inode flushing abort routine. It is called from xfs_iflush when |
722 | * the filesystem is shutting down to clean up the inode state. It is | |
723 | * responsible for removing the inode item from the AIL if it has not been | |
724 | * re-logged, and unlocking the inode's flush lock. | |
1da177e4 LT |
725 | */ |
726 | void | |
727 | xfs_iflush_abort( | |
88fc1879 | 728 | struct xfs_inode *ip) |
1da177e4 | 729 | { |
88fc1879 | 730 | struct xfs_inode_log_item *iip = ip->i_itemp; |
1da177e4 | 731 | |
1da177e4 | 732 | if (iip) { |
2b3cf093 | 733 | xfs_trans_ail_delete(&iip->ili_item, 0); |
1da177e4 LT |
734 | /* |
735 | * Clear the inode logging fields so no more flushes are | |
736 | * attempted. | |
737 | */ | |
1319ebef | 738 | spin_lock(&iip->ili_lock); |
1dfde687 | 739 | iip->ili_last_fields = 0; |
f5d8d5c4 | 740 | iip->ili_fields = 0; |
fc0561ce | 741 | iip->ili_fsync_fields = 0; |
1319ebef | 742 | spin_unlock(&iip->ili_lock); |
1da177e4 LT |
743 | } |
744 | /* | |
745 | * Release the inode's flush lock since we're done with it. | |
746 | */ | |
747 | xfs_ifunlock(ip); | |
748 | } | |
749 | ||
6d192a9b | 750 | /* |
20413e37 DC |
751 | * convert an xfs_inode_log_format struct from the old 32 bit version |
752 | * (which can have different field alignments) to the native 64 bit version | |
6d192a9b TS |
753 | */ |
754 | int | |
755 | xfs_inode_item_format_convert( | |
20413e37 DC |
756 | struct xfs_log_iovec *buf, |
757 | struct xfs_inode_log_format *in_f) | |
6d192a9b | 758 | { |
20413e37 DC |
759 | struct xfs_inode_log_format_32 *in_f32 = buf->i_addr; |
760 | ||
a5155b87 DW |
761 | if (buf->i_len != sizeof(*in_f32)) { |
762 | XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_LOW, NULL); | |
20413e37 | 763 | return -EFSCORRUPTED; |
a5155b87 | 764 | } |
20413e37 DC |
765 | |
766 | in_f->ilf_type = in_f32->ilf_type; | |
767 | in_f->ilf_size = in_f32->ilf_size; | |
768 | in_f->ilf_fields = in_f32->ilf_fields; | |
769 | in_f->ilf_asize = in_f32->ilf_asize; | |
770 | in_f->ilf_dsize = in_f32->ilf_dsize; | |
771 | in_f->ilf_ino = in_f32->ilf_ino; | |
42b67dc6 | 772 | memcpy(&in_f->ilf_u, &in_f32->ilf_u, sizeof(in_f->ilf_u)); |
20413e37 DC |
773 | in_f->ilf_blkno = in_f32->ilf_blkno; |
774 | in_f->ilf_len = in_f32->ilf_len; | |
775 | in_f->ilf_boffset = in_f32->ilf_boffset; | |
776 | return 0; | |
6d192a9b | 777 | } |