Commit | Line | Data |
---|---|---|
0b61f8a4 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 | 2 | /* |
7b718769 NS |
3 | * Copyright (c) 2000-2005 Silicon Graphics, Inc. |
4 | * All Rights Reserved. | |
1da177e4 | 5 | */ |
1da177e4 | 6 | #include "xfs.h" |
a844f451 | 7 | #include "xfs_fs.h" |
70a9883c | 8 | #include "xfs_shared.h" |
239880ef | 9 | #include "xfs_format.h" |
a4fbe6ab | 10 | #include "xfs_log_format.h" |
239880ef | 11 | #include "xfs_trans_resv.h" |
1da177e4 | 12 | #include "xfs_sb.h" |
1da177e4 | 13 | #include "xfs_mount.h" |
239880ef | 14 | #include "xfs_trans.h" |
1da177e4 LT |
15 | #include "xfs_error.h" |
16 | #include "xfs_alloc.h" | |
1da177e4 | 17 | #include "xfs_fsops.h" |
1da177e4 | 18 | #include "xfs_trans_space.h" |
239880ef | 19 | #include "xfs_log.h" |
b16817b6 | 20 | #include "xfs_ag.h" |
84d69619 | 21 | #include "xfs_ag_resv.h" |
7f89c838 | 22 | #include "xfs_trace.h" |
1da177e4 | 23 | |
c789c83c GX |
24 | /* |
25 | * Write new AG headers to disk. Non-transactional, but need to be | |
26 | * written and completed prior to the growfs transaction being logged. | |
27 | * To do this, we use a delayed write buffer list and wait for | |
28 | * submission and IO completion of the list as a whole. This allows the | |
29 | * IO subsystem to merge all the AG headers in a single AG into a single | |
30 | * IO and hide most of the latency of the IO from us. | |
31 | * | |
32 | * This also means that if we get an error whilst building the buffer | |
33 | * list to write, we can cancel the entire list without having written | |
34 | * anything. | |
35 | */ | |
36 | static int | |
37 | xfs_resizefs_init_new_ags( | |
38 | struct xfs_trans *tp, | |
39 | struct aghdr_init_data *id, | |
40 | xfs_agnumber_t oagcount, | |
41 | xfs_agnumber_t nagcount, | |
42 | xfs_rfsblock_t delta, | |
43 | bool *lastag_extended) | |
44 | { | |
45 | struct xfs_mount *mp = tp->t_mountp; | |
46 | xfs_rfsblock_t nb = mp->m_sb.sb_dblocks + delta; | |
47 | int error; | |
48 | ||
49 | *lastag_extended = false; | |
50 | ||
51 | INIT_LIST_HEAD(&id->buffer_list); | |
52 | for (id->agno = nagcount - 1; | |
53 | id->agno >= oagcount; | |
54 | id->agno--, delta -= id->agsize) { | |
55 | ||
56 | if (id->agno == nagcount - 1) | |
57 | id->agsize = nb - (id->agno * | |
58 | (xfs_rfsblock_t)mp->m_sb.sb_agblocks); | |
59 | else | |
60 | id->agsize = mp->m_sb.sb_agblocks; | |
61 | ||
62 | error = xfs_ag_init_headers(mp, id); | |
63 | if (error) { | |
64 | xfs_buf_delwri_cancel(&id->buffer_list); | |
65 | return error; | |
66 | } | |
67 | } | |
68 | ||
69 | error = xfs_buf_delwri_submit(&id->buffer_list); | |
70 | if (error) | |
71 | return error; | |
72 | ||
c789c83c GX |
73 | if (delta) { |
74 | *lastag_extended = true; | |
75 | error = xfs_ag_extend_space(mp, tp, id, delta); | |
76 | } | |
77 | return error; | |
78 | } | |
79 | ||
1da177e4 | 80 | /* |
b16817b6 | 81 | * growfs operations |
1da177e4 | 82 | */ |
1da177e4 LT |
83 | static int |
84 | xfs_growfs_data_private( | |
07aabd9c GX |
85 | struct xfs_mount *mp, /* mount point for filesystem */ |
86 | struct xfs_growfs_data *in) /* growfs data input struct */ | |
1da177e4 | 87 | { |
e8222613 | 88 | struct xfs_buf *bp; |
83a7f86e | 89 | int error; |
1da177e4 LT |
90 | xfs_agnumber_t nagcount; |
91 | xfs_agnumber_t nagimax = 0; | |
ce5e1062 | 92 | xfs_rfsblock_t nb, nb_div, nb_mod; |
fb2fc172 | 93 | int64_t delta; |
c789c83c | 94 | bool lastag_extended; |
1da177e4 | 95 | xfs_agnumber_t oagcount; |
07aabd9c | 96 | struct xfs_trans *tp; |
0410c3bb | 97 | struct aghdr_init_data id = {}; |
1da177e4 LT |
98 | |
99 | nb = in->newblocks; | |
fb2fc172 GX |
100 | error = xfs_sb_validate_fsb_count(&mp->m_sb, nb); |
101 | if (error) | |
4cc929ee | 102 | return error; |
fb2fc172 GX |
103 | |
104 | if (nb > mp->m_sb.sb_dblocks) { | |
105 | error = xfs_buf_read_uncached(mp->m_ddev_targp, | |
1922c949 | 106 | XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1), |
ba372674 | 107 | XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL); |
fb2fc172 GX |
108 | if (error) |
109 | return error; | |
110 | xfs_buf_relse(bp); | |
111 | } | |
1da177e4 | 112 | |
ce5e1062 GX |
113 | nb_div = nb; |
114 | nb_mod = do_div(nb_div, mp->m_sb.sb_agblocks); | |
115 | nagcount = nb_div + (nb_mod != 0); | |
1da177e4 LT |
116 | if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) { |
117 | nagcount--; | |
e6da7c9f | 118 | nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks; |
1da177e4 | 119 | } |
ce5e1062 | 120 | delta = nb - mp->m_sb.sb_dblocks; |
fb2fc172 GX |
121 | /* |
122 | * Reject filesystems with a single AG because they are not | |
123 | * supported, and reject a shrink operation that would cause a | |
124 | * filesystem to become unsupported. | |
125 | */ | |
126 | if (delta < 0 && nagcount < 2) | |
127 | return -EINVAL; | |
128 | ||
1da177e4 | 129 | oagcount = mp->m_sb.sb_agcount; |
0cc6eee1 | 130 | |
1c1c6ebc DC |
131 | /* allocate the new per-ag structures */ |
132 | if (nagcount > oagcount) { | |
133 | error = xfs_initialize_perag(mp, nagcount, &nagimax); | |
134 | if (error) | |
135 | return error; | |
fb2fc172 GX |
136 | } else if (nagcount < oagcount) { |
137 | /* TODO: shrinking the entire AGs hasn't yet completed */ | |
138 | return -EINVAL; | |
1da177e4 | 139 | } |
1c1c6ebc | 140 | |
253f4911 | 141 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_growdata, |
fb2fc172 GX |
142 | (delta > 0 ? XFS_GROWFS_SPACE_RES(mp) : -delta), 0, |
143 | XFS_TRANS_RESERVE, &tp); | |
253f4911 | 144 | if (error) |
1da177e4 | 145 | return error; |
1da177e4 | 146 | |
fb2fc172 GX |
147 | if (delta > 0) { |
148 | error = xfs_resizefs_init_new_ags(tp, &id, oagcount, nagcount, | |
149 | delta, &lastag_extended); | |
150 | } else { | |
151 | static struct ratelimit_state shrink_warning = \ | |
152 | RATELIMIT_STATE_INIT("shrink_warning", 86400 * HZ, 1); | |
153 | ratelimit_set_flags(&shrink_warning, RATELIMIT_MSG_ON_RELEASE); | |
154 | ||
155 | if (__ratelimit(&shrink_warning)) | |
156 | xfs_alert(mp, | |
157 | "EXPERIMENTAL online shrink feature in use. Use at your own risk!"); | |
158 | ||
159 | error = xfs_ag_shrink_space(mp, &tp, nagcount - 1, -delta); | |
160 | } | |
9aebe805 | 161 | if (error) |
83a7f86e | 162 | goto out_trans_cancel; |
9aebe805 | 163 | |
1c1c6ebc DC |
164 | /* |
165 | * Update changed superblock fields transactionally. These are not | |
166 | * seen by the rest of the world until the transaction commit applies | |
167 | * them atomically to the superblock. | |
168 | */ | |
1da177e4 LT |
169 | if (nagcount > oagcount) |
170 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount); | |
c789c83c GX |
171 | if (delta) |
172 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS, delta); | |
0410c3bb DC |
173 | if (id.nfree) |
174 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, id.nfree); | |
014695c0 GX |
175 | |
176 | /* | |
177 | * Sync sb counters now to reflect the updated values. This is | |
178 | * particularly important for shrink because the write verifier | |
179 | * will fail if sb_fdblocks is ever larger than sb_dblocks. | |
180 | */ | |
38c26bfd | 181 | if (xfs_has_lazysbcount(mp)) |
014695c0 GX |
182 | xfs_log_sb(tp); |
183 | ||
f8079b85 | 184 | xfs_trans_set_sync(tp); |
70393313 | 185 | error = xfs_trans_commit(tp); |
1c1c6ebc | 186 | if (error) |
1da177e4 | 187 | return error; |
1c1c6ebc | 188 | |
1da177e4 LT |
189 | /* New allocation groups fully initialized, so update mount struct */ |
190 | if (nagimax) | |
191 | mp->m_maxagi = nagimax; | |
055388a3 | 192 | xfs_set_low_space_thresholds(mp); |
52548852 | 193 | mp->m_alloc_set_aside = xfs_alloc_set_aside(mp); |
1c1c6ebc | 194 | |
fb2fc172 GX |
195 | if (delta > 0) { |
196 | /* | |
197 | * If we expanded the last AG, free the per-AG reservation | |
198 | * so we can reinitialize it with the new size. | |
199 | */ | |
200 | if (lastag_extended) { | |
201 | struct xfs_perag *pag; | |
202 | ||
203 | pag = xfs_perag_get(mp, id.agno); | |
204 | error = xfs_ag_resv_free(pag); | |
205 | xfs_perag_put(pag); | |
206 | if (error) | |
207 | return error; | |
208 | } | |
209 | /* | |
210 | * Reserve AG metadata blocks. ENOSPC here does not mean there | |
211 | * was a growfs failure, just that there still isn't space for | |
212 | * new user data after the grow has been run. | |
213 | */ | |
214 | error = xfs_fs_reserve_ag_blocks(mp); | |
215 | if (error == -ENOSPC) | |
216 | error = 0; | |
20e73b00 | 217 | } |
83a7f86e DC |
218 | return error; |
219 | ||
220 | out_trans_cancel: | |
221 | xfs_trans_cancel(tp); | |
222 | return error; | |
223 | } | |
224 | ||
225 | static int | |
226 | xfs_growfs_log_private( | |
07aabd9c GX |
227 | struct xfs_mount *mp, /* mount point for filesystem */ |
228 | struct xfs_growfs_log *in) /* growfs log input struct */ | |
83a7f86e DC |
229 | { |
230 | xfs_extlen_t nb; | |
231 | ||
232 | nb = in->newblocks; | |
233 | if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES)) | |
234 | return -EINVAL; | |
235 | if (nb == mp->m_sb.sb_logblocks && | |
236 | in->isint == (mp->m_sb.sb_logstart != 0)) | |
237 | return -EINVAL; | |
238 | /* | |
239 | * Moving the log is hard, need new interfaces to sync | |
240 | * the log first, hold off all activity while moving it. | |
241 | * Can have shorter or longer log in the same space, | |
242 | * or transform internal to external log or vice versa. | |
243 | */ | |
244 | return -ENOSYS; | |
245 | } | |
246 | ||
247 | static int | |
248 | xfs_growfs_imaxpct( | |
249 | struct xfs_mount *mp, | |
250 | __u32 imaxpct) | |
251 | { | |
252 | struct xfs_trans *tp; | |
253 | int dpct; | |
254 | int error; | |
255 | ||
256 | if (imaxpct > 100) | |
257 | return -EINVAL; | |
258 | ||
259 | error = xfs_trans_alloc(mp, &M_RES(mp)->tr_growdata, | |
260 | XFS_GROWFS_SPACE_RES(mp), 0, XFS_TRANS_RESERVE, &tp); | |
261 | if (error) | |
262 | return error; | |
263 | ||
264 | dpct = imaxpct - mp->m_sb.sb_imax_pct; | |
265 | xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct); | |
266 | xfs_trans_set_sync(tp); | |
267 | return xfs_trans_commit(tp); | |
268 | } | |
269 | ||
1da177e4 LT |
270 | /* |
271 | * protected versions of growfs function acquire and release locks on the mount | |
272 | * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG, | |
273 | * XFS_IOC_FSGROWFSRT | |
274 | */ | |
1da177e4 LT |
275 | int |
276 | xfs_growfs_data( | |
87444b8c DC |
277 | struct xfs_mount *mp, |
278 | struct xfs_growfs_data *in) | |
1da177e4 | 279 | { |
87444b8c | 280 | int error = 0; |
743bb465 | 281 | |
282 | if (!capable(CAP_SYS_ADMIN)) | |
2451337d | 283 | return -EPERM; |
cc92e7ac | 284 | if (!mutex_trylock(&mp->m_growlock)) |
2451337d | 285 | return -EWOULDBLOCK; |
87444b8c DC |
286 | |
287 | /* update imaxpct separately to the physical grow of the filesystem */ | |
288 | if (in->imaxpct != mp->m_sb.sb_imax_pct) { | |
289 | error = xfs_growfs_imaxpct(mp, in->imaxpct); | |
290 | if (error) | |
291 | goto out_error; | |
292 | } | |
293 | ||
294 | if (in->newblocks != mp->m_sb.sb_dblocks) { | |
295 | error = xfs_growfs_data_private(mp, in); | |
296 | if (error) | |
297 | goto out_error; | |
298 | } | |
299 | ||
300 | /* Post growfs calculations needed to reflect new state in operations */ | |
301 | if (mp->m_sb.sb_imax_pct) { | |
302 | uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct; | |
303 | do_div(icount, 100); | |
ef325959 | 304 | M_IGEO(mp)->maxicount = XFS_FSB_TO_INO(mp, icount); |
87444b8c | 305 | } else |
ef325959 | 306 | M_IGEO(mp)->maxicount = 0; |
87444b8c | 307 | |
83a7f86e | 308 | /* Update secondary superblocks now the physical grow has completed */ |
b16817b6 | 309 | error = xfs_update_secondary_sbs(mp); |
83a7f86e | 310 | |
87444b8c | 311 | out_error: |
52785112 CH |
312 | /* |
313 | * Increment the generation unconditionally, the error could be from | |
314 | * updating the secondary superblocks, in which case the new size | |
315 | * is live already. | |
316 | */ | |
317 | mp->m_generation++; | |
cc92e7ac | 318 | mutex_unlock(&mp->m_growlock); |
1da177e4 LT |
319 | return error; |
320 | } | |
321 | ||
322 | int | |
323 | xfs_growfs_log( | |
324 | xfs_mount_t *mp, | |
07aabd9c | 325 | struct xfs_growfs_log *in) |
1da177e4 LT |
326 | { |
327 | int error; | |
743bb465 | 328 | |
329 | if (!capable(CAP_SYS_ADMIN)) | |
2451337d | 330 | return -EPERM; |
cc92e7ac | 331 | if (!mutex_trylock(&mp->m_growlock)) |
2451337d | 332 | return -EWOULDBLOCK; |
1da177e4 | 333 | error = xfs_growfs_log_private(mp, in); |
cc92e7ac | 334 | mutex_unlock(&mp->m_growlock); |
1da177e4 LT |
335 | return error; |
336 | } | |
337 | ||
338 | /* | |
339 | * exported through ioctl XFS_IOC_FSCOUNTS | |
340 | */ | |
341 | ||
91083269 | 342 | void |
1da177e4 LT |
343 | xfs_fs_counts( |
344 | xfs_mount_t *mp, | |
345 | xfs_fsop_counts_t *cnt) | |
346 | { | |
501ab323 | 347 | cnt->allocino = percpu_counter_read_positive(&mp->m_icount); |
e88b64ea | 348 | cnt->freeino = percpu_counter_read_positive(&mp->m_ifree); |
0d485ada | 349 | cnt->freedata = percpu_counter_read_positive(&mp->m_fdblocks) - |
52548852 | 350 | mp->m_alloc_set_aside; |
501ab323 | 351 | |
3685c2a1 | 352 | spin_lock(&mp->m_sb_lock); |
1da177e4 | 353 | cnt->freertx = mp->m_sb.sb_frextents; |
3685c2a1 | 354 | spin_unlock(&mp->m_sb_lock); |
1da177e4 LT |
355 | } |
356 | ||
357 | /* | |
358 | * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS | |
359 | * | |
360 | * xfs_reserve_blocks is called to set m_resblks | |
361 | * in the in-core mount table. The number of unused reserved blocks | |
c41564b5 | 362 | * is kept in m_resblks_avail. |
1da177e4 LT |
363 | * |
364 | * Reserve the requested number of blocks if available. Otherwise return | |
365 | * as many as possible to satisfy the request. The actual number | |
366 | * reserved are returned in outval | |
367 | * | |
368 | * A null inval pointer indicates that only the current reserved blocks | |
369 | * available should be returned no settings are changed. | |
370 | */ | |
371 | ||
372 | int | |
373 | xfs_reserve_blocks( | |
374 | xfs_mount_t *mp, | |
c8ce540d | 375 | uint64_t *inval, |
1da177e4 LT |
376 | xfs_fsop_resblks_t *outval) |
377 | { | |
c8ce540d DW |
378 | int64_t lcounter, delta; |
379 | int64_t fdblks_delta = 0; | |
380 | uint64_t request; | |
381 | int64_t free; | |
408fd484 | 382 | int error = 0; |
1da177e4 LT |
383 | |
384 | /* If inval is null, report current values and return */ | |
c8ce540d | 385 | if (inval == (uint64_t *)NULL) { |
84e1e99f | 386 | if (!outval) |
2451337d | 387 | return -EINVAL; |
1da177e4 LT |
388 | outval->resblks = mp->m_resblks; |
389 | outval->resblks_avail = mp->m_resblks_avail; | |
014c2544 | 390 | return 0; |
1da177e4 LT |
391 | } |
392 | ||
393 | request = *inval; | |
dbcabad1 DC |
394 | |
395 | /* | |
408fd484 BF |
396 | * With per-cpu counters, this becomes an interesting problem. we need |
397 | * to work out if we are freeing or allocation blocks first, then we can | |
398 | * do the modification as necessary. | |
dbcabad1 | 399 | * |
408fd484 BF |
400 | * We do this under the m_sb_lock so that if we are near ENOSPC, we will |
401 | * hold out any changes while we work out what to do. This means that | |
402 | * the amount of free space can change while we do this, so we need to | |
403 | * retry if we end up trying to reserve more space than is available. | |
dbcabad1 | 404 | */ |
3685c2a1 | 405 | spin_lock(&mp->m_sb_lock); |
1da177e4 LT |
406 | |
407 | /* | |
408 | * If our previous reservation was larger than the current value, | |
408fd484 BF |
409 | * then move any unused blocks back to the free pool. Modify the resblks |
410 | * counters directly since we shouldn't have any problems unreserving | |
411 | * space. | |
1da177e4 | 412 | */ |
1da177e4 LT |
413 | if (mp->m_resblks > request) { |
414 | lcounter = mp->m_resblks_avail - request; | |
415 | if (lcounter > 0) { /* release unused blocks */ | |
dbcabad1 | 416 | fdblks_delta = lcounter; |
1da177e4 LT |
417 | mp->m_resblks_avail -= lcounter; |
418 | } | |
419 | mp->m_resblks = request; | |
408fd484 BF |
420 | if (fdblks_delta) { |
421 | spin_unlock(&mp->m_sb_lock); | |
422 | error = xfs_mod_fdblocks(mp, fdblks_delta, 0); | |
423 | spin_lock(&mp->m_sb_lock); | |
424 | } | |
425 | ||
426 | goto out; | |
427 | } | |
4be536de | 428 | |
408fd484 BF |
429 | /* |
430 | * If the request is larger than the current reservation, reserve the | |
431 | * blocks before we update the reserve counters. Sample m_fdblocks and | |
432 | * perform a partial reservation if the request exceeds free space. | |
433 | */ | |
434 | error = -ENOSPC; | |
435 | do { | |
0d485ada | 436 | free = percpu_counter_sum(&mp->m_fdblocks) - |
52548852 | 437 | mp->m_alloc_set_aside; |
aafe12ce | 438 | if (free <= 0) |
408fd484 | 439 | break; |
dbcabad1 | 440 | |
1da177e4 | 441 | delta = request - mp->m_resblks; |
4be536de | 442 | lcounter = free - delta; |
408fd484 | 443 | if (lcounter < 0) |
1da177e4 | 444 | /* We can't satisfy the request, just get what we can */ |
408fd484 BF |
445 | fdblks_delta = free; |
446 | else | |
447 | fdblks_delta = delta; | |
dbcabad1 | 448 | |
dbcabad1 | 449 | /* |
408fd484 BF |
450 | * We'll either succeed in getting space from the free block |
451 | * count or we'll get an ENOSPC. If we get a ENOSPC, it means | |
452 | * things changed while we were calculating fdblks_delta and so | |
453 | * we should try again to see if there is anything left to | |
454 | * reserve. | |
dbcabad1 DC |
455 | * |
456 | * Don't set the reserved flag here - we don't want to reserve | |
457 | * the extra reserve blocks from the reserve..... | |
458 | */ | |
408fd484 BF |
459 | spin_unlock(&mp->m_sb_lock); |
460 | error = xfs_mod_fdblocks(mp, -fdblks_delta, 0); | |
461 | spin_lock(&mp->m_sb_lock); | |
462 | } while (error == -ENOSPC); | |
463 | ||
464 | /* | |
465 | * Update the reserve counters if blocks have been successfully | |
466 | * allocated. | |
467 | */ | |
468 | if (!error && fdblks_delta) { | |
469 | mp->m_resblks += fdblks_delta; | |
470 | mp->m_resblks_avail += fdblks_delta; | |
dbcabad1 | 471 | } |
408fd484 BF |
472 | |
473 | out: | |
474 | if (outval) { | |
475 | outval->resblks = mp->m_resblks; | |
476 | outval->resblks_avail = mp->m_resblks_avail; | |
477 | } | |
478 | ||
479 | spin_unlock(&mp->m_sb_lock); | |
480 | return error; | |
1da177e4 LT |
481 | } |
482 | ||
1da177e4 LT |
483 | int |
484 | xfs_fs_goingdown( | |
485 | xfs_mount_t *mp, | |
c8ce540d | 486 | uint32_t inflags) |
1da177e4 LT |
487 | { |
488 | switch (inflags) { | |
489 | case XFS_FSOP_GOING_FLAGS_DEFAULT: { | |
040f04bd | 490 | if (!freeze_bdev(mp->m_super->s_bdev)) { |
7d04a335 | 491 | xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT); |
040f04bd | 492 | thaw_bdev(mp->m_super->s_bdev); |
1da177e4 | 493 | } |
1da177e4 LT |
494 | break; |
495 | } | |
496 | case XFS_FSOP_GOING_FLAGS_LOGFLUSH: | |
7d04a335 | 497 | xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT); |
1da177e4 LT |
498 | break; |
499 | case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH: | |
7d04a335 NS |
500 | xfs_force_shutdown(mp, |
501 | SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR); | |
1da177e4 LT |
502 | break; |
503 | default: | |
2451337d | 504 | return -EINVAL; |
1da177e4 LT |
505 | } |
506 | ||
507 | return 0; | |
508 | } | |
2af51f3a DC |
509 | |
510 | /* | |
511 | * Force a shutdown of the filesystem instantly while keeping the filesystem | |
512 | * consistent. We don't do an unmount here; just shutdown the shop, make sure | |
513 | * that absolutely nothing persistent happens to this filesystem after this | |
514 | * point. | |
b36d4651 DC |
515 | * |
516 | * The shutdown state change is atomic, resulting in the first and only the | |
517 | * first shutdown call processing the shutdown. This means we only shutdown the | |
518 | * log once as it requires, and we don't spam the logs when multiple concurrent | |
519 | * shutdowns race to set the shutdown flags. | |
2af51f3a DC |
520 | */ |
521 | void | |
522 | xfs_do_force_shutdown( | |
56668a5c | 523 | struct xfs_mount *mp, |
2af51f3a DC |
524 | int flags, |
525 | char *fname, | |
526 | int lnnum) | |
527 | { | |
b36d4651 DC |
528 | int tag; |
529 | const char *why; | |
2af51f3a | 530 | |
2e973b2c | 531 | if (test_and_set_bit(XFS_OPSTATE_SHUTDOWN, &mp->m_opstate)) |
56668a5c | 532 | return; |
b36d4651 DC |
533 | if (mp->m_sb_bp) |
534 | mp->m_sb_bp->b_flags |= XBF_DONE; | |
b36d4651 DC |
535 | |
536 | if (flags & SHUTDOWN_FORCE_UMOUNT) | |
537 | xfs_alert(mp, "User initiated shutdown received."); | |
56668a5c | 538 | |
b36d4651 DC |
539 | if (xlog_force_shutdown(mp->m_log, flags)) { |
540 | tag = XFS_PTAG_SHUTDOWN_LOGERROR; | |
541 | why = "Log I/O Error"; | |
542 | } else if (flags & SHUTDOWN_CORRUPT_INCORE) { | |
543 | tag = XFS_PTAG_SHUTDOWN_CORRUPT; | |
544 | why = "Corruption of in-memory data"; | |
28d84620 | 545 | } else { |
b36d4651 DC |
546 | tag = XFS_PTAG_SHUTDOWN_IOERROR; |
547 | why = "Metadata I/O Error"; | |
2af51f3a | 548 | } |
56668a5c | 549 | |
7f89c838 DW |
550 | trace_xfs_force_shutdown(mp, tag, flags, fname, lnnum); |
551 | ||
b36d4651 DC |
552 | xfs_alert_tag(mp, tag, |
553 | "%s (0x%x) detected at %pS (%s:%d). Shutting down filesystem.", | |
554 | why, flags, __return_address, fname, lnnum); | |
56668a5c DC |
555 | xfs_alert(mp, |
556 | "Please unmount the filesystem and rectify the problem(s)"); | |
b36d4651 DC |
557 | if (xfs_error_level >= XFS_ERRLEVEL_HIGH) |
558 | xfs_stack_trace(); | |
2af51f3a | 559 | } |
84d69619 DW |
560 | |
561 | /* | |
562 | * Reserve free space for per-AG metadata. | |
563 | */ | |
564 | int | |
565 | xfs_fs_reserve_ag_blocks( | |
566 | struct xfs_mount *mp) | |
567 | { | |
568 | xfs_agnumber_t agno; | |
569 | struct xfs_perag *pag; | |
570 | int error = 0; | |
571 | int err2; | |
572 | ||
15a268d9 | 573 | mp->m_finobt_nores = false; |
f250eedc | 574 | for_each_perag(mp, agno, pag) { |
ebcbef3a | 575 | err2 = xfs_ag_resv_init(pag, NULL); |
84d69619 DW |
576 | if (err2 && !error) |
577 | error = err2; | |
578 | } | |
579 | ||
580 | if (error && error != -ENOSPC) { | |
581 | xfs_warn(mp, | |
582 | "Error %d reserving per-AG metadata reserve pool.", error); | |
583 | xfs_force_shutdown(mp, SHUTDOWN_CORRUPT_INCORE); | |
584 | } | |
585 | ||
586 | return error; | |
587 | } | |
588 | ||
589 | /* | |
590 | * Free space reserved for per-AG metadata. | |
591 | */ | |
592 | int | |
593 | xfs_fs_unreserve_ag_blocks( | |
594 | struct xfs_mount *mp) | |
595 | { | |
596 | xfs_agnumber_t agno; | |
597 | struct xfs_perag *pag; | |
598 | int error = 0; | |
599 | int err2; | |
600 | ||
f250eedc | 601 | for_each_perag(mp, agno, pag) { |
84d69619 | 602 | err2 = xfs_ag_resv_free(pag); |
84d69619 DW |
603 | if (err2 && !error) |
604 | error = err2; | |
605 | } | |
606 | ||
607 | if (error) | |
608 | xfs_warn(mp, | |
609 | "Error %d freeing per-AG metadata reserve pool.", error); | |
610 | ||
611 | return error; | |
612 | } |