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 | |
1da177e4 LT |
39 | switch (ip->i_d.di_format) { |
40 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 41 | if ((iip->ili_fields & XFS_ILOG_DEXT) && |
339a5f5d | 42 | ip->i_d.di_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 | |
1da177e4 LT |
80 | switch (ip->i_d.di_aformat) { |
81 | case XFS_DINODE_FMT_EXTENTS: | |
f5d8d5c4 | 82 | if ((iip->ili_fields & XFS_ILOG_AEXT) && |
339a5f5d | 83 | ip->i_d.di_anextents > 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) + | |
f8d55aa0 | 128 | xfs_log_dinode_size(ip->i_d.di_version); |
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 LT |
144 | |
145 | switch (ip->i_d.di_format) { | |
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) && |
339a5f5d CH |
151 | ip->i_d.di_nextents > 0 && |
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 LT |
229 | |
230 | switch (ip->i_d.di_aformat) { | |
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) && |
339a5f5d CH |
236 | ip->i_d.di_anextents > 0 && |
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) == |
339a5f5d | 241 | ip->i_d.di_anextents); |
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 DC |
307 | to->di_magic = XFS_DINODE_MAGIC; |
308 | ||
f8d55aa0 DC |
309 | to->di_version = from->di_version; |
310 | to->di_format = from->di_format; | |
54295159 CH |
311 | to->di_uid = xfs_kuid_to_uid(inode->i_uid); |
312 | to->di_gid = xfs_kgid_to_gid(inode->i_gid); | |
de7a866f CH |
313 | to->di_projid_lo = from->di_projid & 0xffff; |
314 | to->di_projid_hi = from->di_projid >> 16; | |
f8d55aa0 | 315 | |
93f958f9 | 316 | memset(to->di_pad, 0, sizeof(to->di_pad)); |
faeb4e47 | 317 | memset(to->di_pad3, 0, sizeof(to->di_pad3)); |
3987848c DC |
318 | to->di_atime.t_sec = inode->i_atime.tv_sec; |
319 | to->di_atime.t_nsec = inode->i_atime.tv_nsec; | |
320 | to->di_mtime.t_sec = inode->i_mtime.tv_sec; | |
321 | to->di_mtime.t_nsec = inode->i_mtime.tv_nsec; | |
322 | to->di_ctime.t_sec = inode->i_ctime.tv_sec; | |
323 | to->di_ctime.t_nsec = inode->i_ctime.tv_nsec; | |
54d7b5c1 | 324 | to->di_nlink = inode->i_nlink; |
9e9a2674 | 325 | to->di_gen = inode->i_generation; |
c19b3b05 | 326 | to->di_mode = inode->i_mode; |
f8d55aa0 | 327 | |
f8d55aa0 DC |
328 | to->di_size = from->di_size; |
329 | to->di_nblocks = from->di_nblocks; | |
330 | to->di_extsize = from->di_extsize; | |
331 | to->di_nextents = from->di_nextents; | |
332 | to->di_anextents = from->di_anextents; | |
333 | to->di_forkoff = from->di_forkoff; | |
334 | to->di_aformat = from->di_aformat; | |
335 | to->di_dmevmask = from->di_dmevmask; | |
336 | to->di_dmstate = from->di_dmstate; | |
337 | to->di_flags = from->di_flags; | |
f8d55aa0 | 338 | |
20413e37 DC |
339 | /* log a dummy value to ensure log structure is fully initialised */ |
340 | to->di_next_unlinked = NULLAGINO; | |
341 | ||
f8d55aa0 | 342 | if (from->di_version == 3) { |
f0e28280 | 343 | to->di_changecount = inode_peek_iversion(inode); |
8d2d878d CH |
344 | to->di_crtime.t_sec = from->di_crtime.tv_sec; |
345 | to->di_crtime.t_nsec = from->di_crtime.tv_nsec; | |
f8d55aa0 | 346 | to->di_flags2 = from->di_flags2; |
f7ca3522 | 347 | to->di_cowextsize = from->di_cowextsize; |
93f958f9 DC |
348 | to->di_ino = ip->i_ino; |
349 | to->di_lsn = lsn; | |
350 | memset(to->di_pad2, 0, sizeof(to->di_pad2)); | |
351 | uuid_copy(&to->di_uuid, &ip->i_mount->m_sb.sb_meta_uuid); | |
f8d55aa0 DC |
352 | to->di_flushiter = 0; |
353 | } else { | |
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); |
f8d55aa0 DC |
373 | xlog_finish_iovec(lv, *vecp, xfs_log_dinode_size(ip->i_d.di_version)); |
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 | |
263997a6 DC |
398 | ASSERT(ip->i_d.di_version > 1); |
399 | ||
2f251293 CH |
400 | ilf = xlog_prepare_iovec(lv, &vecp, XLOG_REG_TYPE_IFORMAT); |
401 | ilf->ilf_type = XFS_LI_INODE; | |
402 | ilf->ilf_ino = ip->i_ino; | |
403 | ilf->ilf_blkno = ip->i_imap.im_blkno; | |
404 | ilf->ilf_len = ip->i_imap.im_len; | |
405 | ilf->ilf_boffset = ip->i_imap.im_boffset; | |
406 | ilf->ilf_fields = XFS_ILOG_CORE; | |
407 | ilf->ilf_size = 2; /* format + core */ | |
20413e37 DC |
408 | |
409 | /* | |
410 | * make sure we don't leak uninitialised data into the log in the case | |
411 | * when we don't log every field in the inode. | |
412 | */ | |
413 | ilf->ilf_dsize = 0; | |
414 | ilf->ilf_asize = 0; | |
415 | ilf->ilf_pad = 0; | |
42b67dc6 | 416 | memset(&ilf->ilf_u, 0, sizeof(ilf->ilf_u)); |
20413e37 DC |
417 | |
418 | xlog_finish_iovec(lv, vecp, sizeof(*ilf)); | |
3de559fb | 419 | |
f8d55aa0 | 420 | xfs_inode_item_format_core(ip, lv, &vecp); |
bde7cff6 | 421 | xfs_inode_item_format_data_fork(iip, ilf, lv, &vecp); |
3de559fb | 422 | if (XFS_IFORK_Q(ip)) { |
bde7cff6 | 423 | xfs_inode_item_format_attr_fork(iip, ilf, lv, &vecp); |
3de559fb CH |
424 | } else { |
425 | iip->ili_fields &= | |
426 | ~(XFS_ILOG_ADATA | XFS_ILOG_ABROOT | XFS_ILOG_AEXT); | |
427 | } | |
428 | ||
2f251293 CH |
429 | /* update the format with the exact fields we actually logged */ |
430 | ilf->ilf_fields |= (iip->ili_fields & ~XFS_ILOG_TIMESTAMP); | |
1da177e4 LT |
431 | } |
432 | ||
1da177e4 LT |
433 | /* |
434 | * This is called to pin the inode associated with the inode log | |
a14a5ab5 | 435 | * item in memory so it cannot be written out. |
1da177e4 LT |
436 | */ |
437 | STATIC void | |
438 | xfs_inode_item_pin( | |
7bfa31d8 | 439 | struct xfs_log_item *lip) |
1da177e4 | 440 | { |
7bfa31d8 | 441 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 442 | |
7bfa31d8 CH |
443 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); |
444 | ||
445 | trace_xfs_inode_pin(ip, _RET_IP_); | |
446 | atomic_inc(&ip->i_pincount); | |
1da177e4 LT |
447 | } |
448 | ||
449 | ||
450 | /* | |
451 | * This is called to unpin the inode associated with the inode log | |
452 | * item which was previously pinned with a call to xfs_inode_item_pin(). | |
a14a5ab5 CH |
453 | * |
454 | * Also wake up anyone in xfs_iunpin_wait() if the count goes to 0. | |
1da177e4 | 455 | */ |
1da177e4 LT |
456 | STATIC void |
457 | xfs_inode_item_unpin( | |
7bfa31d8 | 458 | struct xfs_log_item *lip, |
9412e318 | 459 | int remove) |
1da177e4 | 460 | { |
7bfa31d8 | 461 | struct xfs_inode *ip = INODE_ITEM(lip)->ili_inode; |
a14a5ab5 | 462 | |
4aaf15d1 | 463 | trace_xfs_inode_unpin(ip, _RET_IP_); |
a14a5ab5 CH |
464 | ASSERT(atomic_read(&ip->i_pincount) > 0); |
465 | if (atomic_dec_and_test(&ip->i_pincount)) | |
f392e631 | 466 | wake_up_bit(&ip->i_flags, __XFS_IPINNED_BIT); |
1da177e4 LT |
467 | } |
468 | ||
d3a304b6 CM |
469 | /* |
470 | * Callback used to mark a buffer with XFS_LI_FAILED when items in the buffer | |
471 | * have been failed during writeback | |
472 | * | |
473 | * This informs the AIL that the inode is already flush locked on the next push, | |
474 | * and acquires a hold on the buffer to ensure that it isn't reclaimed before | |
475 | * dirty data makes it to disk. | |
476 | */ | |
477 | STATIC void | |
478 | xfs_inode_item_error( | |
479 | struct xfs_log_item *lip, | |
480 | struct xfs_buf *bp) | |
481 | { | |
482 | ASSERT(xfs_isiflocked(INODE_ITEM(lip)->ili_inode)); | |
483 | xfs_set_li_failed(lip, bp); | |
484 | } | |
485 | ||
1da177e4 | 486 | STATIC uint |
43ff2122 CH |
487 | xfs_inode_item_push( |
488 | struct xfs_log_item *lip, | |
489 | struct list_head *buffer_list) | |
57e80956 MW |
490 | __releases(&lip->li_ailp->ail_lock) |
491 | __acquires(&lip->li_ailp->ail_lock) | |
1da177e4 | 492 | { |
7bfa31d8 CH |
493 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
494 | struct xfs_inode *ip = iip->ili_inode; | |
d3a304b6 | 495 | struct xfs_buf *bp = lip->li_buf; |
43ff2122 CH |
496 | uint rval = XFS_ITEM_SUCCESS; |
497 | int error; | |
1da177e4 | 498 | |
7bfa31d8 | 499 | if (xfs_ipincount(ip) > 0) |
1da177e4 | 500 | return XFS_ITEM_PINNED; |
1da177e4 | 501 | |
d3a304b6 CM |
502 | /* |
503 | * The buffer containing this item failed to be written back | |
504 | * previously. Resubmit the buffer for IO. | |
505 | */ | |
22525c17 | 506 | if (test_bit(XFS_LI_FAILED, &lip->li_flags)) { |
d3a304b6 CM |
507 | if (!xfs_buf_trylock(bp)) |
508 | return XFS_ITEM_LOCKED; | |
509 | ||
643c8c05 | 510 | if (!xfs_buf_resubmit_failed_buffers(bp, buffer_list)) |
d3a304b6 CM |
511 | rval = XFS_ITEM_FLUSHING; |
512 | ||
513 | xfs_buf_unlock(bp); | |
514 | return rval; | |
515 | } | |
516 | ||
7bfa31d8 | 517 | if (!xfs_ilock_nowait(ip, XFS_ILOCK_SHARED)) |
1da177e4 | 518 | return XFS_ITEM_LOCKED; |
1da177e4 | 519 | |
4c46819a CH |
520 | /* |
521 | * Re-check the pincount now that we stabilized the value by | |
522 | * taking the ilock. | |
523 | */ | |
524 | if (xfs_ipincount(ip) > 0) { | |
43ff2122 CH |
525 | rval = XFS_ITEM_PINNED; |
526 | goto out_unlock; | |
4c46819a CH |
527 | } |
528 | ||
9a3a5dab BF |
529 | /* |
530 | * Stale inode items should force out the iclog. | |
531 | */ | |
532 | if (ip->i_flags & XFS_ISTALE) { | |
533 | rval = XFS_ITEM_PINNED; | |
534 | goto out_unlock; | |
535 | } | |
536 | ||
43ff2122 CH |
537 | /* |
538 | * Someone else is already flushing the inode. Nothing we can do | |
539 | * here but wait for the flush to finish and remove the item from | |
540 | * the AIL. | |
541 | */ | |
1da177e4 | 542 | if (!xfs_iflock_nowait(ip)) { |
43ff2122 CH |
543 | rval = XFS_ITEM_FLUSHING; |
544 | goto out_unlock; | |
1da177e4 LT |
545 | } |
546 | ||
43ff2122 CH |
547 | ASSERT(iip->ili_fields != 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); |
548 | ASSERT(iip->ili_logged == 0 || XFS_FORCED_SHUTDOWN(ip->i_mount)); | |
549 | ||
57e80956 | 550 | spin_unlock(&lip->li_ailp->ail_lock); |
43ff2122 CH |
551 | |
552 | error = xfs_iflush(ip, &bp); | |
553 | if (!error) { | |
554 | if (!xfs_buf_delwri_queue(bp, buffer_list)) | |
555 | rval = XFS_ITEM_FLUSHING; | |
556 | xfs_buf_relse(bp); | |
1da177e4 | 557 | } |
43ff2122 | 558 | |
57e80956 | 559 | spin_lock(&lip->li_ailp->ail_lock); |
43ff2122 CH |
560 | out_unlock: |
561 | xfs_iunlock(ip, XFS_ILOCK_SHARED); | |
562 | return rval; | |
1da177e4 LT |
563 | } |
564 | ||
565 | /* | |
566 | * Unlock the inode associated with the inode log item. | |
1da177e4 LT |
567 | */ |
568 | STATIC void | |
ddf92053 | 569 | xfs_inode_item_release( |
7bfa31d8 | 570 | struct xfs_log_item *lip) |
1da177e4 | 571 | { |
7bfa31d8 CH |
572 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
573 | struct xfs_inode *ip = iip->ili_inode; | |
898621d5 | 574 | unsigned short lock_flags; |
1da177e4 | 575 | |
f3ca8738 CH |
576 | ASSERT(ip->i_itemp != NULL); |
577 | ASSERT(xfs_isilocked(ip, XFS_ILOCK_EXCL)); | |
1da177e4 | 578 | |
898621d5 CH |
579 | lock_flags = iip->ili_lock_flags; |
580 | iip->ili_lock_flags = 0; | |
ddc3415a | 581 | if (lock_flags) |
f3ca8738 | 582 | xfs_iunlock(ip, lock_flags); |
1da177e4 LT |
583 | } |
584 | ||
585 | /* | |
de25c181 DC |
586 | * This is called to find out where the oldest active copy of the inode log |
587 | * item in the on disk log resides now that the last log write of it completed | |
588 | * at the given lsn. Since we always re-log all dirty data in an inode, the | |
589 | * latest copy in the on disk log is the only one that matters. Therefore, | |
590 | * simply return the given lsn. | |
591 | * | |
592 | * If the inode has been marked stale because the cluster is being freed, we | |
593 | * don't want to (re-)insert this inode into the AIL. There is a race condition | |
594 | * where the cluster buffer may be unpinned before the inode is inserted into | |
595 | * the AIL during transaction committed processing. If the buffer is unpinned | |
596 | * before the inode item has been committed and inserted, then it is possible | |
1316d4da | 597 | * for the buffer to be written and IO completes before the inode is inserted |
de25c181 DC |
598 | * into the AIL. In that case, we'd be inserting a clean, stale inode into the |
599 | * AIL which will never get removed. It will, however, get reclaimed which | |
600 | * triggers an assert in xfs_inode_free() complaining about freein an inode | |
601 | * still in the AIL. | |
602 | * | |
1316d4da DC |
603 | * To avoid this, just unpin the inode directly and return a LSN of -1 so the |
604 | * transaction committed code knows that it does not need to do any further | |
605 | * processing on the item. | |
1da177e4 | 606 | */ |
1da177e4 LT |
607 | STATIC xfs_lsn_t |
608 | xfs_inode_item_committed( | |
7bfa31d8 | 609 | struct xfs_log_item *lip, |
1da177e4 LT |
610 | xfs_lsn_t lsn) |
611 | { | |
de25c181 DC |
612 | struct xfs_inode_log_item *iip = INODE_ITEM(lip); |
613 | struct xfs_inode *ip = iip->ili_inode; | |
614 | ||
1316d4da DC |
615 | if (xfs_iflags_test(ip, XFS_ISTALE)) { |
616 | xfs_inode_item_unpin(lip, 0); | |
617 | return -1; | |
618 | } | |
7bfa31d8 | 619 | return lsn; |
1da177e4 LT |
620 | } |
621 | ||
1da177e4 LT |
622 | STATIC void |
623 | xfs_inode_item_committing( | |
7bfa31d8 | 624 | struct xfs_log_item *lip, |
ddf92053 | 625 | xfs_lsn_t commit_lsn) |
1da177e4 | 626 | { |
ddf92053 CH |
627 | INODE_ITEM(lip)->ili_last_lsn = commit_lsn; |
628 | return xfs_inode_item_release(lip); | |
1da177e4 LT |
629 | } |
630 | ||
272e42b2 | 631 | static const struct xfs_item_ops xfs_inode_item_ops = { |
7bfa31d8 CH |
632 | .iop_size = xfs_inode_item_size, |
633 | .iop_format = xfs_inode_item_format, | |
634 | .iop_pin = xfs_inode_item_pin, | |
635 | .iop_unpin = xfs_inode_item_unpin, | |
ddf92053 | 636 | .iop_release = xfs_inode_item_release, |
7bfa31d8 CH |
637 | .iop_committed = xfs_inode_item_committed, |
638 | .iop_push = xfs_inode_item_push, | |
ddf92053 | 639 | .iop_committing = xfs_inode_item_committing, |
d3a304b6 | 640 | .iop_error = xfs_inode_item_error |
1da177e4 LT |
641 | }; |
642 | ||
643 | ||
644 | /* | |
645 | * Initialize the inode log item for a newly allocated (in-core) inode. | |
646 | */ | |
647 | void | |
648 | xfs_inode_item_init( | |
7bfa31d8 CH |
649 | struct xfs_inode *ip, |
650 | struct xfs_mount *mp) | |
1da177e4 | 651 | { |
7bfa31d8 | 652 | struct xfs_inode_log_item *iip; |
1da177e4 LT |
653 | |
654 | ASSERT(ip->i_itemp == NULL); | |
707e0dda | 655 | iip = ip->i_itemp = kmem_zone_zalloc(xfs_ili_zone, 0); |
1da177e4 | 656 | |
1da177e4 | 657 | iip->ili_inode = ip; |
43f5efc5 DC |
658 | xfs_log_item_init(mp, &iip->ili_item, XFS_LI_INODE, |
659 | &xfs_inode_item_ops); | |
1da177e4 LT |
660 | } |
661 | ||
662 | /* | |
663 | * Free the inode log item and any memory hanging off of it. | |
664 | */ | |
665 | void | |
666 | xfs_inode_item_destroy( | |
667 | xfs_inode_t *ip) | |
668 | { | |
b1c5ebb2 | 669 | kmem_free(ip->i_itemp->ili_item.li_lv_shadow); |
377bcd5f | 670 | kmem_cache_free(xfs_ili_zone, ip->i_itemp); |
1da177e4 LT |
671 | } |
672 | ||
673 | ||
674 | /* | |
675 | * This is the inode flushing I/O completion routine. It is called | |
676 | * from interrupt level when the buffer containing the inode is | |
677 | * flushed to disk. It is responsible for removing the inode item | |
678 | * from the AIL if it has not been re-logged, and unlocking the inode's | |
679 | * flush lock. | |
30136832 DC |
680 | * |
681 | * To reduce AIL lock traffic as much as possible, we scan the buffer log item | |
682 | * list for other inodes that will run this function. We remove them from the | |
683 | * buffer list so we can process all the inode IO completions in one AIL lock | |
684 | * traversal. | |
1da177e4 | 685 | */ |
1da177e4 LT |
686 | void |
687 | xfs_iflush_done( | |
ca30b2a7 CH |
688 | struct xfs_buf *bp, |
689 | struct xfs_log_item *lip) | |
1da177e4 | 690 | { |
30136832 | 691 | struct xfs_inode_log_item *iip; |
643c8c05 | 692 | struct xfs_log_item *blip, *n; |
ca30b2a7 | 693 | struct xfs_ail *ailp = lip->li_ailp; |
30136832 | 694 | int need_ail = 0; |
643c8c05 | 695 | LIST_HEAD(tmp); |
30136832 DC |
696 | |
697 | /* | |
698 | * Scan the buffer IO completions for other inodes being completed and | |
699 | * attach them to the current inode log item. | |
700 | */ | |
30136832 | 701 | |
643c8c05 | 702 | list_add_tail(&lip->li_bio_list, &tmp); |
30136832 | 703 | |
643c8c05 CM |
704 | list_for_each_entry_safe(blip, n, &bp->b_li_list, li_bio_list) { |
705 | if (lip->li_cb != xfs_iflush_done) | |
706 | continue; | |
30136832 | 707 | |
643c8c05 | 708 | list_move_tail(&blip->li_bio_list, &tmp); |
30136832 DC |
709 | /* |
710 | * while we have the item, do the unlocked check for needing | |
711 | * the AIL lock. | |
712 | */ | |
713 | iip = INODE_ITEM(blip); | |
d3a304b6 | 714 | if ((iip->ili_logged && blip->li_lsn == iip->ili_flush_lsn) || |
22525c17 | 715 | test_bit(XFS_LI_FAILED, &blip->li_flags)) |
30136832 | 716 | need_ail++; |
30136832 DC |
717 | } |
718 | ||
719 | /* make sure we capture the state of the initial inode. */ | |
720 | iip = INODE_ITEM(lip); | |
d3a304b6 | 721 | if ((iip->ili_logged && lip->li_lsn == iip->ili_flush_lsn) || |
22525c17 | 722 | test_bit(XFS_LI_FAILED, &lip->li_flags)) |
30136832 | 723 | need_ail++; |
1da177e4 LT |
724 | |
725 | /* | |
726 | * We only want to pull the item from the AIL if it is | |
727 | * actually there and its location in the log has not | |
728 | * changed since we started the flush. Thus, we only bother | |
729 | * if the ili_logged flag is set and the inode's lsn has not | |
730 | * changed. First we check the lsn outside | |
731 | * the lock since it's cheaper, and then we recheck while | |
732 | * holding the lock before removing the inode from the AIL. | |
733 | */ | |
30136832 | 734 | if (need_ail) { |
27af1bbf CH |
735 | bool mlip_changed = false; |
736 | ||
737 | /* this is an opencoded batch version of xfs_trans_ail_delete */ | |
57e80956 | 738 | spin_lock(&ailp->ail_lock); |
643c8c05 | 739 | list_for_each_entry(blip, &tmp, li_bio_list) { |
27af1bbf CH |
740 | if (INODE_ITEM(blip)->ili_logged && |
741 | blip->li_lsn == INODE_ITEM(blip)->ili_flush_lsn) | |
742 | mlip_changed |= xfs_ail_delete_one(ailp, blip); | |
d3a304b6 CM |
743 | else { |
744 | xfs_clear_li_failed(blip); | |
745 | } | |
1da177e4 | 746 | } |
1da177e4 | 747 | |
27af1bbf | 748 | if (mlip_changed) { |
57e80956 MW |
749 | if (!XFS_FORCED_SHUTDOWN(ailp->ail_mount)) |
750 | xlog_assign_tail_lsn_locked(ailp->ail_mount); | |
751 | if (list_empty(&ailp->ail_head)) | |
752 | wake_up_all(&ailp->ail_empty); | |
27af1bbf | 753 | } |
57e80956 | 754 | spin_unlock(&ailp->ail_lock); |
27af1bbf CH |
755 | |
756 | if (mlip_changed) | |
57e80956 | 757 | xfs_log_space_wake(ailp->ail_mount); |
27af1bbf | 758 | } |
1da177e4 LT |
759 | |
760 | /* | |
30136832 DC |
761 | * clean up and unlock the flush lock now we are done. We can clear the |
762 | * ili_last_fields bits now that we know that the data corresponding to | |
763 | * them is safely on disk. | |
1da177e4 | 764 | */ |
643c8c05 CM |
765 | list_for_each_entry_safe(blip, n, &tmp, li_bio_list) { |
766 | list_del_init(&blip->li_bio_list); | |
30136832 DC |
767 | iip = INODE_ITEM(blip); |
768 | iip->ili_logged = 0; | |
769 | iip->ili_last_fields = 0; | |
770 | xfs_ifunlock(iip->ili_inode); | |
771 | } | |
643c8c05 | 772 | list_del(&tmp); |
1da177e4 LT |
773 | } |
774 | ||
775 | /* | |
04913fdd DC |
776 | * This is the inode flushing abort routine. It is called from xfs_iflush when |
777 | * the filesystem is shutting down to clean up the inode state. It is | |
778 | * responsible for removing the inode item from the AIL if it has not been | |
779 | * re-logged, and unlocking the inode's flush lock. | |
1da177e4 LT |
780 | */ |
781 | void | |
782 | xfs_iflush_abort( | |
04913fdd DC |
783 | xfs_inode_t *ip, |
784 | bool stale) | |
1da177e4 | 785 | { |
783a2f65 | 786 | xfs_inode_log_item_t *iip = ip->i_itemp; |
1da177e4 | 787 | |
1da177e4 | 788 | if (iip) { |
22525c17 | 789 | if (test_bit(XFS_LI_IN_AIL, &iip->ili_item.li_flags)) { |
146e54b7 BF |
790 | xfs_trans_ail_remove(&iip->ili_item, |
791 | stale ? SHUTDOWN_LOG_IO_ERROR : | |
04913fdd | 792 | SHUTDOWN_CORRUPT_INCORE); |
1da177e4 LT |
793 | } |
794 | iip->ili_logged = 0; | |
795 | /* | |
796 | * Clear the ili_last_fields bits now that we know that the | |
797 | * data corresponding to them is safely on disk. | |
798 | */ | |
799 | iip->ili_last_fields = 0; | |
800 | /* | |
801 | * Clear the inode logging fields so no more flushes are | |
802 | * attempted. | |
803 | */ | |
f5d8d5c4 | 804 | iip->ili_fields = 0; |
fc0561ce | 805 | iip->ili_fsync_fields = 0; |
1da177e4 LT |
806 | } |
807 | /* | |
808 | * Release the inode's flush lock since we're done with it. | |
809 | */ | |
810 | xfs_ifunlock(ip); | |
811 | } | |
812 | ||
813 | void | |
814 | xfs_istale_done( | |
ca30b2a7 CH |
815 | struct xfs_buf *bp, |
816 | struct xfs_log_item *lip) | |
1da177e4 | 817 | { |
04913fdd | 818 | xfs_iflush_abort(INODE_ITEM(lip)->ili_inode, true); |
1da177e4 | 819 | } |
6d192a9b TS |
820 | |
821 | /* | |
20413e37 DC |
822 | * convert an xfs_inode_log_format struct from the old 32 bit version |
823 | * (which can have different field alignments) to the native 64 bit version | |
6d192a9b TS |
824 | */ |
825 | int | |
826 | xfs_inode_item_format_convert( | |
20413e37 DC |
827 | struct xfs_log_iovec *buf, |
828 | struct xfs_inode_log_format *in_f) | |
6d192a9b | 829 | { |
20413e37 DC |
830 | struct xfs_inode_log_format_32 *in_f32 = buf->i_addr; |
831 | ||
a5155b87 DW |
832 | if (buf->i_len != sizeof(*in_f32)) { |
833 | XFS_ERROR_REPORT(__func__, XFS_ERRLEVEL_LOW, NULL); | |
20413e37 | 834 | return -EFSCORRUPTED; |
a5155b87 | 835 | } |
20413e37 DC |
836 | |
837 | in_f->ilf_type = in_f32->ilf_type; | |
838 | in_f->ilf_size = in_f32->ilf_size; | |
839 | in_f->ilf_fields = in_f32->ilf_fields; | |
840 | in_f->ilf_asize = in_f32->ilf_asize; | |
841 | in_f->ilf_dsize = in_f32->ilf_dsize; | |
842 | in_f->ilf_ino = in_f32->ilf_ino; | |
42b67dc6 | 843 | memcpy(&in_f->ilf_u, &in_f32->ilf_u, sizeof(in_f->ilf_u)); |
20413e37 DC |
844 | in_f->ilf_blkno = in_f32->ilf_blkno; |
845 | in_f->ilf_len = in_f32->ilf_len; | |
846 | in_f->ilf_boffset = in_f32->ilf_boffset; | |
847 | return 0; | |
6d192a9b | 848 | } |