Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
7b718769 NS |
2 | * Copyright (c) 2000-2003,2005 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" |
1da177e4 | 20 | #include "xfs_types.h" |
a844f451 | 21 | #include "xfs_bit.h" |
1da177e4 | 22 | #include "xfs_log.h" |
a844f451 | 23 | #include "xfs_inum.h" |
1da177e4 LT |
24 | #include "xfs_trans.h" |
25 | #include "xfs_sb.h" | |
26 | #include "xfs_ag.h" | |
27 | #include "xfs_dir.h" | |
28 | #include "xfs_dir2.h" | |
29 | #include "xfs_dmapi.h" | |
30 | #include "xfs_mount.h" | |
1da177e4 | 31 | #include "xfs_bmap_btree.h" |
a844f451 | 32 | #include "xfs_alloc_btree.h" |
1da177e4 | 33 | #include "xfs_ialloc_btree.h" |
a844f451 NS |
34 | #include "xfs_dir_sf.h" |
35 | #include "xfs_dir2_sf.h" | |
36 | #include "xfs_attr_sf.h" | |
37 | #include "xfs_dinode.h" | |
38 | #include "xfs_inode.h" | |
39 | #include "xfs_inode_item.h" | |
1da177e4 LT |
40 | #include "xfs_itable.h" |
41 | #include "xfs_btree.h" | |
42 | #include "xfs_alloc.h" | |
43 | #include "xfs_ialloc.h" | |
44 | #include "xfs_attr.h" | |
1da177e4 LT |
45 | #include "xfs_bmap.h" |
46 | #include "xfs_acl.h" | |
47 | #include "xfs_mac.h" | |
48 | #include "xfs_error.h" | |
49 | #include "xfs_buf_item.h" | |
50 | #include "xfs_rw.h" | |
51 | ||
52 | /* | |
53 | * This is a subroutine for xfs_write() and other writers (xfs_ioctl) | |
54 | * which clears the setuid and setgid bits when a file is written. | |
55 | */ | |
56 | int | |
57 | xfs_write_clear_setuid( | |
58 | xfs_inode_t *ip) | |
59 | { | |
60 | xfs_mount_t *mp; | |
61 | xfs_trans_t *tp; | |
62 | int error; | |
63 | ||
64 | mp = ip->i_mount; | |
65 | tp = xfs_trans_alloc(mp, XFS_TRANS_WRITEID); | |
66 | if ((error = xfs_trans_reserve(tp, 0, | |
67 | XFS_WRITEID_LOG_RES(mp), | |
68 | 0, 0, 0))) { | |
69 | xfs_trans_cancel(tp, 0); | |
70 | return error; | |
71 | } | |
72 | xfs_ilock(ip, XFS_ILOCK_EXCL); | |
73 | xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL); | |
74 | xfs_trans_ihold(tp, ip); | |
75 | ip->i_d.di_mode &= ~S_ISUID; | |
76 | ||
77 | /* | |
78 | * Note that we don't have to worry about mandatory | |
79 | * file locking being disabled here because we only | |
80 | * clear the S_ISGID bit if the Group execute bit is | |
81 | * on, but if it was on then mandatory locking wouldn't | |
82 | * have been enabled. | |
83 | */ | |
84 | if (ip->i_d.di_mode & S_IXGRP) { | |
85 | ip->i_d.di_mode &= ~S_ISGID; | |
86 | } | |
87 | xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE); | |
88 | xfs_trans_set_sync(tp); | |
89 | error = xfs_trans_commit(tp, 0, NULL); | |
90 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
91 | return 0; | |
92 | } | |
93 | ||
94 | /* | |
95 | * Force a shutdown of the filesystem instantly while keeping | |
96 | * the filesystem consistent. We don't do an unmount here; just shutdown | |
97 | * the shop, make sure that absolutely nothing persistent happens to | |
98 | * this filesystem after this point. | |
99 | */ | |
100 | ||
101 | void | |
102 | xfs_do_force_shutdown( | |
103 | bhv_desc_t *bdp, | |
104 | int flags, | |
105 | char *fname, | |
106 | int lnnum) | |
107 | { | |
108 | int logerror; | |
109 | xfs_mount_t *mp; | |
110 | ||
111 | mp = XFS_BHVTOM(bdp); | |
7d04a335 | 112 | logerror = flags & SHUTDOWN_LOG_IO_ERROR; |
1da177e4 | 113 | |
7d04a335 NS |
114 | if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { |
115 | cmn_err(CE_NOTE, "xfs_force_shutdown(%s,0x%x) called from " | |
116 | "line %d of file %s. Return address = 0x%p", | |
117 | mp->m_fsname, flags, lnnum, fname, __return_address); | |
1da177e4 LT |
118 | } |
119 | /* | |
120 | * No need to duplicate efforts. | |
121 | */ | |
122 | if (XFS_FORCED_SHUTDOWN(mp) && !logerror) | |
123 | return; | |
124 | ||
125 | /* | |
126 | * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't | |
127 | * queue up anybody new on the log reservations, and wakes up | |
7d04a335 NS |
128 | * everybody who's sleeping on log reservations to tell them |
129 | * the bad news. | |
1da177e4 LT |
130 | */ |
131 | if (xfs_log_force_umount(mp, logerror)) | |
132 | return; | |
133 | ||
7d04a335 | 134 | if (flags & SHUTDOWN_CORRUPT_INCORE) { |
1da177e4 LT |
135 | xfs_cmn_err(XFS_PTAG_SHUTDOWN_CORRUPT, CE_ALERT, mp, |
136 | "Corruption of in-memory data detected. Shutting down filesystem: %s", | |
137 | mp->m_fsname); | |
138 | if (XFS_ERRLEVEL_HIGH <= xfs_error_level) { | |
139 | xfs_stack_trace(); | |
140 | } | |
7d04a335 | 141 | } else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { |
1da177e4 LT |
142 | if (logerror) { |
143 | xfs_cmn_err(XFS_PTAG_SHUTDOWN_LOGERROR, CE_ALERT, mp, | |
7d04a335 | 144 | "Log I/O Error Detected. Shutting down filesystem: %s", |
1da177e4 | 145 | mp->m_fsname); |
7d04a335 | 146 | } else if (flags & SHUTDOWN_DEVICE_REQ) { |
1da177e4 | 147 | xfs_cmn_err(XFS_PTAG_SHUTDOWN_IOERROR, CE_ALERT, mp, |
7d04a335 NS |
148 | "All device paths lost. Shutting down filesystem: %s", |
149 | mp->m_fsname); | |
150 | } else if (!(flags & SHUTDOWN_REMOTE_REQ)) { | |
151 | xfs_cmn_err(XFS_PTAG_SHUTDOWN_IOERROR, CE_ALERT, mp, | |
152 | "I/O Error Detected. Shutting down filesystem: %s", | |
1da177e4 LT |
153 | mp->m_fsname); |
154 | } | |
155 | } | |
7d04a335 NS |
156 | if (!(flags & SHUTDOWN_FORCE_UMOUNT)) { |
157 | cmn_err(CE_ALERT, "Please umount the filesystem, " | |
158 | "and rectify the problem(s)"); | |
1da177e4 LT |
159 | } |
160 | } | |
161 | ||
162 | ||
163 | /* | |
164 | * Called when we want to stop a buffer from getting written or read. | |
165 | * We attach the EIO error, muck with its flags, and call biodone | |
166 | * so that the proper iodone callbacks get called. | |
167 | */ | |
168 | int | |
169 | xfs_bioerror( | |
170 | xfs_buf_t *bp) | |
171 | { | |
172 | ||
173 | #ifdef XFSERRORDEBUG | |
174 | ASSERT(XFS_BUF_ISREAD(bp) || bp->b_iodone); | |
175 | #endif | |
176 | ||
177 | /* | |
178 | * No need to wait until the buffer is unpinned. | |
179 | * We aren't flushing it. | |
180 | */ | |
181 | xfs_buftrace("XFS IOERROR", bp); | |
182 | XFS_BUF_ERROR(bp, EIO); | |
183 | /* | |
184 | * We're calling biodone, so delete B_DONE flag. Either way | |
185 | * we have to call the iodone callback, and calling biodone | |
186 | * probably is the best way since it takes care of | |
187 | * GRIO as well. | |
188 | */ | |
189 | XFS_BUF_UNREAD(bp); | |
190 | XFS_BUF_UNDELAYWRITE(bp); | |
191 | XFS_BUF_UNDONE(bp); | |
192 | XFS_BUF_STALE(bp); | |
193 | ||
194 | XFS_BUF_CLR_BDSTRAT_FUNC(bp); | |
195 | xfs_biodone(bp); | |
196 | ||
197 | return (EIO); | |
198 | } | |
199 | ||
200 | /* | |
201 | * Same as xfs_bioerror, except that we are releasing the buffer | |
202 | * here ourselves, and avoiding the biodone call. | |
203 | * This is meant for userdata errors; metadata bufs come with | |
204 | * iodone functions attached, so that we can track down errors. | |
205 | */ | |
206 | int | |
207 | xfs_bioerror_relse( | |
208 | xfs_buf_t *bp) | |
209 | { | |
210 | int64_t fl; | |
211 | ||
212 | ASSERT(XFS_BUF_IODONE_FUNC(bp) != xfs_buf_iodone_callbacks); | |
213 | ASSERT(XFS_BUF_IODONE_FUNC(bp) != xlog_iodone); | |
214 | ||
215 | xfs_buftrace("XFS IOERRELSE", bp); | |
216 | fl = XFS_BUF_BFLAGS(bp); | |
217 | /* | |
218 | * No need to wait until the buffer is unpinned. | |
219 | * We aren't flushing it. | |
220 | * | |
221 | * chunkhold expects B_DONE to be set, whether | |
222 | * we actually finish the I/O or not. We don't want to | |
223 | * change that interface. | |
224 | */ | |
225 | XFS_BUF_UNREAD(bp); | |
226 | XFS_BUF_UNDELAYWRITE(bp); | |
227 | XFS_BUF_DONE(bp); | |
228 | XFS_BUF_STALE(bp); | |
229 | XFS_BUF_CLR_IODONE_FUNC(bp); | |
230 | XFS_BUF_CLR_BDSTRAT_FUNC(bp); | |
231 | if (!(fl & XFS_B_ASYNC)) { | |
232 | /* | |
233 | * Mark b_error and B_ERROR _both_. | |
234 | * Lot's of chunkcache code assumes that. | |
235 | * There's no reason to mark error for | |
236 | * ASYNC buffers. | |
237 | */ | |
238 | XFS_BUF_ERROR(bp, EIO); | |
239 | XFS_BUF_V_IODONESEMA(bp); | |
240 | } else { | |
241 | xfs_buf_relse(bp); | |
242 | } | |
243 | return (EIO); | |
244 | } | |
ce8e922c | 245 | |
1da177e4 LT |
246 | /* |
247 | * Prints out an ALERT message about I/O error. | |
248 | */ | |
249 | void | |
250 | xfs_ioerror_alert( | |
251 | char *func, | |
252 | struct xfs_mount *mp, | |
253 | xfs_buf_t *bp, | |
254 | xfs_daddr_t blkno) | |
255 | { | |
256 | cmn_err(CE_ALERT, | |
257 | "I/O error in filesystem (\"%s\") meta-data dev %s block 0x%llx" | |
da1650a5 | 258 | " (\"%s\") error %d buf count %zd", |
1da177e4 | 259 | (!mp || !mp->m_fsname) ? "(fs name not set)" : mp->m_fsname, |
ce8e922c NS |
260 | XFS_BUFTARG_NAME(XFS_BUF_TARGET(bp)), |
261 | (__uint64_t)blkno, func, | |
262 | XFS_BUF_GETERROR(bp), XFS_BUF_COUNT(bp)); | |
1da177e4 LT |
263 | } |
264 | ||
265 | /* | |
266 | * This isn't an absolute requirement, but it is | |
267 | * just a good idea to call xfs_read_buf instead of | |
268 | * directly doing a read_buf call. For one, we shouldn't | |
269 | * be doing this disk read if we are in SHUTDOWN state anyway, | |
270 | * so this stops that from happening. Secondly, this does all | |
271 | * the error checking stuff and the brelse if appropriate for | |
272 | * the caller, so the code can be a little leaner. | |
273 | */ | |
274 | ||
275 | int | |
276 | xfs_read_buf( | |
277 | struct xfs_mount *mp, | |
278 | xfs_buftarg_t *target, | |
279 | xfs_daddr_t blkno, | |
280 | int len, | |
281 | uint flags, | |
282 | xfs_buf_t **bpp) | |
283 | { | |
284 | xfs_buf_t *bp; | |
285 | int error; | |
286 | ||
287 | if (flags) | |
288 | bp = xfs_buf_read_flags(target, blkno, len, flags); | |
289 | else | |
290 | bp = xfs_buf_read(target, blkno, len, flags); | |
291 | if (!bp) | |
292 | return XFS_ERROR(EIO); | |
293 | error = XFS_BUF_GETERROR(bp); | |
294 | if (bp && !error && !XFS_FORCED_SHUTDOWN(mp)) { | |
295 | *bpp = bp; | |
296 | } else { | |
297 | *bpp = NULL; | |
298 | if (error) { | |
299 | xfs_ioerror_alert("xfs_read_buf", mp, bp, XFS_BUF_ADDR(bp)); | |
300 | } else { | |
301 | error = XFS_ERROR(EIO); | |
302 | } | |
303 | if (bp) { | |
304 | XFS_BUF_UNDONE(bp); | |
305 | XFS_BUF_UNDELAYWRITE(bp); | |
306 | XFS_BUF_STALE(bp); | |
307 | /* | |
308 | * brelse clears B_ERROR and b_error | |
309 | */ | |
310 | xfs_buf_relse(bp); | |
311 | } | |
312 | } | |
313 | return (error); | |
314 | } | |
315 | ||
316 | /* | |
317 | * Wrapper around bwrite() so that we can trap | |
318 | * write errors, and act accordingly. | |
319 | */ | |
320 | int | |
321 | xfs_bwrite( | |
322 | struct xfs_mount *mp, | |
323 | struct xfs_buf *bp) | |
324 | { | |
325 | int error; | |
326 | ||
327 | /* | |
328 | * XXXsup how does this work for quotas. | |
329 | */ | |
330 | XFS_BUF_SET_BDSTRAT_FUNC(bp, xfs_bdstrat_cb); | |
331 | XFS_BUF_SET_FSPRIVATE3(bp, mp); | |
332 | XFS_BUF_WRITE(bp); | |
333 | ||
334 | if ((error = XFS_bwrite(bp))) { | |
335 | ASSERT(mp); | |
336 | /* | |
337 | * Cannot put a buftrace here since if the buffer is not | |
338 | * B_HOLD then we will brelse() the buffer before returning | |
339 | * from bwrite and we could be tracing a buffer that has | |
340 | * been reused. | |
341 | */ | |
7d04a335 | 342 | xfs_force_shutdown(mp, SHUTDOWN_META_IO_ERROR); |
1da177e4 LT |
343 | } |
344 | return (error); | |
345 | } |