xref: /linux/fs/xfs/xfs_fsops.c (revision a4f174dee4ae842e07cab7eeec194a3e60925c8d)
1 /*
2  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3  * All Rights Reserved.
4  *
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
7  * published by the Free Software Foundation.
8  *
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.
13  *
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
17  */
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
24 #include "xfs_sb.h"
25 #include "xfs_ag.h"
26 #include "xfs_mount.h"
27 #include "xfs_da_format.h"
28 #include "xfs_da_btree.h"
29 #include "xfs_inode.h"
30 #include "xfs_trans.h"
31 #include "xfs_inode_item.h"
32 #include "xfs_error.h"
33 #include "xfs_btree.h"
34 #include "xfs_alloc_btree.h"
35 #include "xfs_alloc.h"
36 #include "xfs_ialloc.h"
37 #include "xfs_fsops.h"
38 #include "xfs_itable.h"
39 #include "xfs_trans_space.h"
40 #include "xfs_rtalloc.h"
41 #include "xfs_trace.h"
42 #include "xfs_log.h"
43 #include "xfs_dinode.h"
44 #include "xfs_filestream.h"
45 
46 /*
47  * File system operations
48  */
49 
50 int
51 xfs_fs_geometry(
52 	xfs_mount_t		*mp,
53 	xfs_fsop_geom_t		*geo,
54 	int			new_version)
55 {
56 
57 	memset(geo, 0, sizeof(*geo));
58 
59 	geo->blocksize = mp->m_sb.sb_blocksize;
60 	geo->rtextsize = mp->m_sb.sb_rextsize;
61 	geo->agblocks = mp->m_sb.sb_agblocks;
62 	geo->agcount = mp->m_sb.sb_agcount;
63 	geo->logblocks = mp->m_sb.sb_logblocks;
64 	geo->sectsize = mp->m_sb.sb_sectsize;
65 	geo->inodesize = mp->m_sb.sb_inodesize;
66 	geo->imaxpct = mp->m_sb.sb_imax_pct;
67 	geo->datablocks = mp->m_sb.sb_dblocks;
68 	geo->rtblocks = mp->m_sb.sb_rblocks;
69 	geo->rtextents = mp->m_sb.sb_rextents;
70 	geo->logstart = mp->m_sb.sb_logstart;
71 	ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
72 	memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
73 	if (new_version >= 2) {
74 		geo->sunit = mp->m_sb.sb_unit;
75 		geo->swidth = mp->m_sb.sb_width;
76 	}
77 	if (new_version >= 3) {
78 		geo->version = XFS_FSOP_GEOM_VERSION;
79 		geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
80 			     XFS_FSOP_GEOM_FLAGS_DIRV2 |
81 			(xfs_sb_version_hasattr(&mp->m_sb) ?
82 				XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
83 			(xfs_sb_version_hasquota(&mp->m_sb) ?
84 				XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
85 			(xfs_sb_version_hasalign(&mp->m_sb) ?
86 				XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
87 			(xfs_sb_version_hasdalign(&mp->m_sb) ?
88 				XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
89 			(xfs_sb_version_hasextflgbit(&mp->m_sb) ?
90 				XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
91 			(xfs_sb_version_hassector(&mp->m_sb) ?
92 				XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
93 			(xfs_sb_version_hasasciici(&mp->m_sb) ?
94 				XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
95 			(xfs_sb_version_haslazysbcount(&mp->m_sb) ?
96 				XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
97 			(xfs_sb_version_hasattr2(&mp->m_sb) ?
98 				XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) |
99 			(xfs_sb_version_hasprojid32bit(&mp->m_sb) ?
100 				XFS_FSOP_GEOM_FLAGS_PROJID32 : 0) |
101 			(xfs_sb_version_hascrc(&mp->m_sb) ?
102 				XFS_FSOP_GEOM_FLAGS_V5SB : 0) |
103 			(xfs_sb_version_hasftype(&mp->m_sb) ?
104 				XFS_FSOP_GEOM_FLAGS_FTYPE : 0) |
105 			(xfs_sb_version_hasfinobt(&mp->m_sb) ?
106 				XFS_FSOP_GEOM_FLAGS_FINOBT : 0);
107 		geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
108 				mp->m_sb.sb_logsectsize : BBSIZE;
109 		geo->rtsectsize = mp->m_sb.sb_blocksize;
110 		geo->dirblocksize = mp->m_dir_geo->blksize;
111 	}
112 	if (new_version >= 4) {
113 		geo->flags |=
114 			(xfs_sb_version_haslogv2(&mp->m_sb) ?
115 				XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
116 		geo->logsunit = mp->m_sb.sb_logsunit;
117 	}
118 	return 0;
119 }
120 
121 static struct xfs_buf *
122 xfs_growfs_get_hdr_buf(
123 	struct xfs_mount	*mp,
124 	xfs_daddr_t		blkno,
125 	size_t			numblks,
126 	int			flags,
127 	const struct xfs_buf_ops *ops)
128 {
129 	struct xfs_buf		*bp;
130 
131 	bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags);
132 	if (!bp)
133 		return NULL;
134 
135 	xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
136 	bp->b_bn = blkno;
137 	bp->b_maps[0].bm_bn = blkno;
138 	bp->b_ops = ops;
139 
140 	return bp;
141 }
142 
143 static int
144 xfs_growfs_data_private(
145 	xfs_mount_t		*mp,		/* mount point for filesystem */
146 	xfs_growfs_data_t	*in)		/* growfs data input struct */
147 {
148 	xfs_agf_t		*agf;
149 	struct xfs_agfl		*agfl;
150 	xfs_agi_t		*agi;
151 	xfs_agnumber_t		agno;
152 	xfs_extlen_t		agsize;
153 	xfs_extlen_t		tmpsize;
154 	xfs_alloc_rec_t		*arec;
155 	xfs_buf_t		*bp;
156 	int			bucket;
157 	int			dpct;
158 	int			error, saved_error = 0;
159 	xfs_agnumber_t		nagcount;
160 	xfs_agnumber_t		nagimax = 0;
161 	xfs_rfsblock_t		nb, nb_mod;
162 	xfs_rfsblock_t		new;
163 	xfs_rfsblock_t		nfree;
164 	xfs_agnumber_t		oagcount;
165 	int			pct;
166 	xfs_trans_t		*tp;
167 
168 	nb = in->newblocks;
169 	pct = in->imaxpct;
170 	if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
171 		return -EINVAL;
172 	if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
173 		return error;
174 	dpct = pct - mp->m_sb.sb_imax_pct;
175 	error = xfs_buf_read_uncached(mp->m_ddev_targp,
176 				XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
177 				XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
178 	if (error)
179 		return error;
180 	xfs_buf_relse(bp);
181 
182 	new = nb;	/* use new as a temporary here */
183 	nb_mod = do_div(new, mp->m_sb.sb_agblocks);
184 	nagcount = new + (nb_mod != 0);
185 	if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
186 		nagcount--;
187 		nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
188 		if (nb < mp->m_sb.sb_dblocks)
189 			return -EINVAL;
190 	}
191 	new = nb - mp->m_sb.sb_dblocks;
192 	oagcount = mp->m_sb.sb_agcount;
193 
194 	/* allocate the new per-ag structures */
195 	if (nagcount > oagcount) {
196 		error = xfs_initialize_perag(mp, nagcount, &nagimax);
197 		if (error)
198 			return error;
199 	}
200 
201 	tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
202 	tp->t_flags |= XFS_TRANS_RESERVE;
203 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_growdata,
204 				  XFS_GROWFS_SPACE_RES(mp), 0);
205 	if (error) {
206 		xfs_trans_cancel(tp, 0);
207 		return error;
208 	}
209 
210 	/*
211 	 * Write new AG headers to disk. Non-transactional, but written
212 	 * synchronously so they are completed prior to the growfs transaction
213 	 * being logged.
214 	 */
215 	nfree = 0;
216 	for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
217 		__be32	*agfl_bno;
218 
219 		/*
220 		 * AG freespace header block
221 		 */
222 		bp = xfs_growfs_get_hdr_buf(mp,
223 				XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
224 				XFS_FSS_TO_BB(mp, 1), 0,
225 				&xfs_agf_buf_ops);
226 		if (!bp) {
227 			error = -ENOMEM;
228 			goto error0;
229 		}
230 
231 		agf = XFS_BUF_TO_AGF(bp);
232 		agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
233 		agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
234 		agf->agf_seqno = cpu_to_be32(agno);
235 		if (agno == nagcount - 1)
236 			agsize =
237 				nb -
238 				(agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
239 		else
240 			agsize = mp->m_sb.sb_agblocks;
241 		agf->agf_length = cpu_to_be32(agsize);
242 		agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
243 		agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
244 		agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
245 		agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
246 		agf->agf_flfirst = 0;
247 		agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
248 		agf->agf_flcount = 0;
249 		tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
250 		agf->agf_freeblks = cpu_to_be32(tmpsize);
251 		agf->agf_longest = cpu_to_be32(tmpsize);
252 		if (xfs_sb_version_hascrc(&mp->m_sb))
253 			uuid_copy(&agf->agf_uuid, &mp->m_sb.sb_uuid);
254 
255 		error = xfs_bwrite(bp);
256 		xfs_buf_relse(bp);
257 		if (error)
258 			goto error0;
259 
260 		/*
261 		 * AG freelist header block
262 		 */
263 		bp = xfs_growfs_get_hdr_buf(mp,
264 				XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
265 				XFS_FSS_TO_BB(mp, 1), 0,
266 				&xfs_agfl_buf_ops);
267 		if (!bp) {
268 			error = -ENOMEM;
269 			goto error0;
270 		}
271 
272 		agfl = XFS_BUF_TO_AGFL(bp);
273 		if (xfs_sb_version_hascrc(&mp->m_sb)) {
274 			agfl->agfl_magicnum = cpu_to_be32(XFS_AGFL_MAGIC);
275 			agfl->agfl_seqno = cpu_to_be32(agno);
276 			uuid_copy(&agfl->agfl_uuid, &mp->m_sb.sb_uuid);
277 		}
278 
279 		agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, bp);
280 		for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++)
281 			agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK);
282 
283 		error = xfs_bwrite(bp);
284 		xfs_buf_relse(bp);
285 		if (error)
286 			goto error0;
287 
288 		/*
289 		 * AG inode header block
290 		 */
291 		bp = xfs_growfs_get_hdr_buf(mp,
292 				XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
293 				XFS_FSS_TO_BB(mp, 1), 0,
294 				&xfs_agi_buf_ops);
295 		if (!bp) {
296 			error = -ENOMEM;
297 			goto error0;
298 		}
299 
300 		agi = XFS_BUF_TO_AGI(bp);
301 		agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
302 		agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
303 		agi->agi_seqno = cpu_to_be32(agno);
304 		agi->agi_length = cpu_to_be32(agsize);
305 		agi->agi_count = 0;
306 		agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
307 		agi->agi_level = cpu_to_be32(1);
308 		agi->agi_freecount = 0;
309 		agi->agi_newino = cpu_to_be32(NULLAGINO);
310 		agi->agi_dirino = cpu_to_be32(NULLAGINO);
311 		if (xfs_sb_version_hascrc(&mp->m_sb))
312 			uuid_copy(&agi->agi_uuid, &mp->m_sb.sb_uuid);
313 		if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
314 			agi->agi_free_root = cpu_to_be32(XFS_FIBT_BLOCK(mp));
315 			agi->agi_free_level = cpu_to_be32(1);
316 		}
317 		for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
318 			agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
319 
320 		error = xfs_bwrite(bp);
321 		xfs_buf_relse(bp);
322 		if (error)
323 			goto error0;
324 
325 		/*
326 		 * BNO btree root block
327 		 */
328 		bp = xfs_growfs_get_hdr_buf(mp,
329 				XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
330 				BTOBB(mp->m_sb.sb_blocksize), 0,
331 				&xfs_allocbt_buf_ops);
332 
333 		if (!bp) {
334 			error = -ENOMEM;
335 			goto error0;
336 		}
337 
338 		if (xfs_sb_version_hascrc(&mp->m_sb))
339 			xfs_btree_init_block(mp, bp, XFS_ABTB_CRC_MAGIC, 0, 1,
340 						agno, XFS_BTREE_CRC_BLOCKS);
341 		else
342 			xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1,
343 						agno, 0);
344 
345 		arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
346 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
347 		arec->ar_blockcount = cpu_to_be32(
348 			agsize - be32_to_cpu(arec->ar_startblock));
349 
350 		error = xfs_bwrite(bp);
351 		xfs_buf_relse(bp);
352 		if (error)
353 			goto error0;
354 
355 		/*
356 		 * CNT btree root block
357 		 */
358 		bp = xfs_growfs_get_hdr_buf(mp,
359 				XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
360 				BTOBB(mp->m_sb.sb_blocksize), 0,
361 				&xfs_allocbt_buf_ops);
362 		if (!bp) {
363 			error = -ENOMEM;
364 			goto error0;
365 		}
366 
367 		if (xfs_sb_version_hascrc(&mp->m_sb))
368 			xfs_btree_init_block(mp, bp, XFS_ABTC_CRC_MAGIC, 0, 1,
369 						agno, XFS_BTREE_CRC_BLOCKS);
370 		else
371 			xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1,
372 						agno, 0);
373 
374 		arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
375 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
376 		arec->ar_blockcount = cpu_to_be32(
377 			agsize - be32_to_cpu(arec->ar_startblock));
378 		nfree += be32_to_cpu(arec->ar_blockcount);
379 
380 		error = xfs_bwrite(bp);
381 		xfs_buf_relse(bp);
382 		if (error)
383 			goto error0;
384 
385 		/*
386 		 * INO btree root block
387 		 */
388 		bp = xfs_growfs_get_hdr_buf(mp,
389 				XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
390 				BTOBB(mp->m_sb.sb_blocksize), 0,
391 				&xfs_inobt_buf_ops);
392 		if (!bp) {
393 			error = -ENOMEM;
394 			goto error0;
395 		}
396 
397 		if (xfs_sb_version_hascrc(&mp->m_sb))
398 			xfs_btree_init_block(mp, bp, XFS_IBT_CRC_MAGIC, 0, 0,
399 						agno, XFS_BTREE_CRC_BLOCKS);
400 		else
401 			xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0,
402 						agno, 0);
403 
404 		error = xfs_bwrite(bp);
405 		xfs_buf_relse(bp);
406 		if (error)
407 			goto error0;
408 
409 		/*
410 		 * FINO btree root block
411 		 */
412 		if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
413 			bp = xfs_growfs_get_hdr_buf(mp,
414 				XFS_AGB_TO_DADDR(mp, agno, XFS_FIBT_BLOCK(mp)),
415 				BTOBB(mp->m_sb.sb_blocksize), 0,
416 				&xfs_inobt_buf_ops);
417 			if (!bp) {
418 				error = -ENOMEM;
419 				goto error0;
420 			}
421 
422 			if (xfs_sb_version_hascrc(&mp->m_sb))
423 				xfs_btree_init_block(mp, bp, XFS_FIBT_CRC_MAGIC,
424 						     0, 0, agno,
425 						     XFS_BTREE_CRC_BLOCKS);
426 			else
427 				xfs_btree_init_block(mp, bp, XFS_FIBT_MAGIC, 0,
428 						     0, agno, 0);
429 
430 			error = xfs_bwrite(bp);
431 			xfs_buf_relse(bp);
432 			if (error)
433 				goto error0;
434 		}
435 
436 	}
437 	xfs_trans_agblocks_delta(tp, nfree);
438 	/*
439 	 * There are new blocks in the old last a.g.
440 	 */
441 	if (new) {
442 		/*
443 		 * Change the agi length.
444 		 */
445 		error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
446 		if (error) {
447 			goto error0;
448 		}
449 		ASSERT(bp);
450 		agi = XFS_BUF_TO_AGI(bp);
451 		be32_add_cpu(&agi->agi_length, new);
452 		ASSERT(nagcount == oagcount ||
453 		       be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
454 		xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
455 		/*
456 		 * Change agf length.
457 		 */
458 		error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
459 		if (error) {
460 			goto error0;
461 		}
462 		ASSERT(bp);
463 		agf = XFS_BUF_TO_AGF(bp);
464 		be32_add_cpu(&agf->agf_length, new);
465 		ASSERT(be32_to_cpu(agf->agf_length) ==
466 		       be32_to_cpu(agi->agi_length));
467 
468 		xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
469 		/*
470 		 * Free the new space.
471 		 */
472 		error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
473 			be32_to_cpu(agf->agf_length) - new), new);
474 		if (error) {
475 			goto error0;
476 		}
477 	}
478 
479 	/*
480 	 * Update changed superblock fields transactionally. These are not
481 	 * seen by the rest of the world until the transaction commit applies
482 	 * them atomically to the superblock.
483 	 */
484 	if (nagcount > oagcount)
485 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
486 	if (nb > mp->m_sb.sb_dblocks)
487 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
488 				 nb - mp->m_sb.sb_dblocks);
489 	if (nfree)
490 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
491 	if (dpct)
492 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
493 	error = xfs_trans_commit(tp, 0);
494 	if (error)
495 		return error;
496 
497 	/* New allocation groups fully initialized, so update mount struct */
498 	if (nagimax)
499 		mp->m_maxagi = nagimax;
500 	if (mp->m_sb.sb_imax_pct) {
501 		__uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
502 		do_div(icount, 100);
503 		mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
504 	} else
505 		mp->m_maxicount = 0;
506 	xfs_set_low_space_thresholds(mp);
507 
508 	/* update secondary superblocks. */
509 	for (agno = 1; agno < nagcount; agno++) {
510 		error = 0;
511 		/*
512 		 * new secondary superblocks need to be zeroed, not read from
513 		 * disk as the contents of the new area we are growing into is
514 		 * completely unknown.
515 		 */
516 		if (agno < oagcount) {
517 			error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
518 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
519 				  XFS_FSS_TO_BB(mp, 1), 0, &bp,
520 				  &xfs_sb_buf_ops);
521 		} else {
522 			bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
523 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
524 				  XFS_FSS_TO_BB(mp, 1), 0);
525 			if (bp) {
526 				bp->b_ops = &xfs_sb_buf_ops;
527 				xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
528 			} else
529 				error = -ENOMEM;
530 		}
531 
532 		/*
533 		 * If we get an error reading or writing alternate superblocks,
534 		 * continue.  xfs_repair chooses the "best" superblock based
535 		 * on most matches; if we break early, we'll leave more
536 		 * superblocks un-updated than updated, and xfs_repair may
537 		 * pick them over the properly-updated primary.
538 		 */
539 		if (error) {
540 			xfs_warn(mp,
541 		"error %d reading secondary superblock for ag %d",
542 				error, agno);
543 			saved_error = error;
544 			continue;
545 		}
546 		xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
547 
548 		error = xfs_bwrite(bp);
549 		xfs_buf_relse(bp);
550 		if (error) {
551 			xfs_warn(mp,
552 		"write error %d updating secondary superblock for ag %d",
553 				error, agno);
554 			saved_error = error;
555 			continue;
556 		}
557 	}
558 	return saved_error ? saved_error : error;
559 
560  error0:
561 	xfs_trans_cancel(tp, XFS_TRANS_ABORT);
562 	return error;
563 }
564 
565 static int
566 xfs_growfs_log_private(
567 	xfs_mount_t		*mp,	/* mount point for filesystem */
568 	xfs_growfs_log_t	*in)	/* growfs log input struct */
569 {
570 	xfs_extlen_t		nb;
571 
572 	nb = in->newblocks;
573 	if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
574 		return -EINVAL;
575 	if (nb == mp->m_sb.sb_logblocks &&
576 	    in->isint == (mp->m_sb.sb_logstart != 0))
577 		return -EINVAL;
578 	/*
579 	 * Moving the log is hard, need new interfaces to sync
580 	 * the log first, hold off all activity while moving it.
581 	 * Can have shorter or longer log in the same space,
582 	 * or transform internal to external log or vice versa.
583 	 */
584 	return -ENOSYS;
585 }
586 
587 /*
588  * protected versions of growfs function acquire and release locks on the mount
589  * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
590  * XFS_IOC_FSGROWFSRT
591  */
592 
593 
594 int
595 xfs_growfs_data(
596 	xfs_mount_t		*mp,
597 	xfs_growfs_data_t	*in)
598 {
599 	int error;
600 
601 	if (!capable(CAP_SYS_ADMIN))
602 		return -EPERM;
603 	if (!mutex_trylock(&mp->m_growlock))
604 		return -EWOULDBLOCK;
605 	error = xfs_growfs_data_private(mp, in);
606 	mutex_unlock(&mp->m_growlock);
607 	return error;
608 }
609 
610 int
611 xfs_growfs_log(
612 	xfs_mount_t		*mp,
613 	xfs_growfs_log_t	*in)
614 {
615 	int error;
616 
617 	if (!capable(CAP_SYS_ADMIN))
618 		return -EPERM;
619 	if (!mutex_trylock(&mp->m_growlock))
620 		return -EWOULDBLOCK;
621 	error = xfs_growfs_log_private(mp, in);
622 	mutex_unlock(&mp->m_growlock);
623 	return error;
624 }
625 
626 /*
627  * exported through ioctl XFS_IOC_FSCOUNTS
628  */
629 
630 int
631 xfs_fs_counts(
632 	xfs_mount_t		*mp,
633 	xfs_fsop_counts_t	*cnt)
634 {
635 	xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
636 	spin_lock(&mp->m_sb_lock);
637 	cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
638 	cnt->freertx = mp->m_sb.sb_frextents;
639 	cnt->freeino = mp->m_sb.sb_ifree;
640 	cnt->allocino = mp->m_sb.sb_icount;
641 	spin_unlock(&mp->m_sb_lock);
642 	return 0;
643 }
644 
645 /*
646  * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
647  *
648  * xfs_reserve_blocks is called to set m_resblks
649  * in the in-core mount table. The number of unused reserved blocks
650  * is kept in m_resblks_avail.
651  *
652  * Reserve the requested number of blocks if available. Otherwise return
653  * as many as possible to satisfy the request. The actual number
654  * reserved are returned in outval
655  *
656  * A null inval pointer indicates that only the current reserved blocks
657  * available  should  be returned no settings are changed.
658  */
659 
660 int
661 xfs_reserve_blocks(
662 	xfs_mount_t             *mp,
663 	__uint64_t              *inval,
664 	xfs_fsop_resblks_t      *outval)
665 {
666 	__int64_t		lcounter, delta, fdblks_delta;
667 	__uint64_t		request;
668 
669 	/* If inval is null, report current values and return */
670 	if (inval == (__uint64_t *)NULL) {
671 		if (!outval)
672 			return -EINVAL;
673 		outval->resblks = mp->m_resblks;
674 		outval->resblks_avail = mp->m_resblks_avail;
675 		return 0;
676 	}
677 
678 	request = *inval;
679 
680 	/*
681 	 * With per-cpu counters, this becomes an interesting
682 	 * problem. we needto work out if we are freeing or allocation
683 	 * blocks first, then we can do the modification as necessary.
684 	 *
685 	 * We do this under the m_sb_lock so that if we are near
686 	 * ENOSPC, we will hold out any changes while we work out
687 	 * what to do. This means that the amount of free space can
688 	 * change while we do this, so we need to retry if we end up
689 	 * trying to reserve more space than is available.
690 	 *
691 	 * We also use the xfs_mod_incore_sb() interface so that we
692 	 * don't have to care about whether per cpu counter are
693 	 * enabled, disabled or even compiled in....
694 	 */
695 retry:
696 	spin_lock(&mp->m_sb_lock);
697 	xfs_icsb_sync_counters_locked(mp, 0);
698 
699 	/*
700 	 * If our previous reservation was larger than the current value,
701 	 * then move any unused blocks back to the free pool.
702 	 */
703 	fdblks_delta = 0;
704 	if (mp->m_resblks > request) {
705 		lcounter = mp->m_resblks_avail - request;
706 		if (lcounter  > 0) {		/* release unused blocks */
707 			fdblks_delta = lcounter;
708 			mp->m_resblks_avail -= lcounter;
709 		}
710 		mp->m_resblks = request;
711 	} else {
712 		__int64_t	free;
713 
714 		free =  mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
715 		if (!free)
716 			goto out; /* ENOSPC and fdblks_delta = 0 */
717 
718 		delta = request - mp->m_resblks;
719 		lcounter = free - delta;
720 		if (lcounter < 0) {
721 			/* We can't satisfy the request, just get what we can */
722 			mp->m_resblks += free;
723 			mp->m_resblks_avail += free;
724 			fdblks_delta = -free;
725 		} else {
726 			fdblks_delta = -delta;
727 			mp->m_resblks = request;
728 			mp->m_resblks_avail += delta;
729 		}
730 	}
731 out:
732 	if (outval) {
733 		outval->resblks = mp->m_resblks;
734 		outval->resblks_avail = mp->m_resblks_avail;
735 	}
736 	spin_unlock(&mp->m_sb_lock);
737 
738 	if (fdblks_delta) {
739 		/*
740 		 * If we are putting blocks back here, m_resblks_avail is
741 		 * already at its max so this will put it in the free pool.
742 		 *
743 		 * If we need space, we'll either succeed in getting it
744 		 * from the free block count or we'll get an enospc. If
745 		 * we get a ENOSPC, it means things changed while we were
746 		 * calculating fdblks_delta and so we should try again to
747 		 * see if there is anything left to reserve.
748 		 *
749 		 * Don't set the reserved flag here - we don't want to reserve
750 		 * the extra reserve blocks from the reserve.....
751 		 */
752 		int error;
753 		error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
754 						 fdblks_delta, 0);
755 		if (error == -ENOSPC)
756 			goto retry;
757 	}
758 	return 0;
759 }
760 
761 /*
762  * Dump a transaction into the log that contains no real change. This is needed
763  * to be able to make the log dirty or stamp the current tail LSN into the log
764  * during the covering operation.
765  *
766  * We cannot use an inode here for this - that will push dirty state back up
767  * into the VFS and then periodic inode flushing will prevent log covering from
768  * making progress. Hence we log a field in the superblock instead and use a
769  * synchronous transaction to ensure the superblock is immediately unpinned
770  * and can be written back.
771  */
772 int
773 xfs_fs_log_dummy(
774 	xfs_mount_t	*mp)
775 {
776 	xfs_trans_t	*tp;
777 	int		error;
778 
779 	tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
780 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_sb, 0, 0);
781 	if (error) {
782 		xfs_trans_cancel(tp, 0);
783 		return error;
784 	}
785 
786 	/* log the UUID because it is an unchanging field */
787 	xfs_mod_sb(tp, XFS_SB_UUID);
788 	xfs_trans_set_sync(tp);
789 	return xfs_trans_commit(tp, 0);
790 }
791 
792 int
793 xfs_fs_goingdown(
794 	xfs_mount_t	*mp,
795 	__uint32_t	inflags)
796 {
797 	switch (inflags) {
798 	case XFS_FSOP_GOING_FLAGS_DEFAULT: {
799 		struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
800 
801 		if (sb && !IS_ERR(sb)) {
802 			xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
803 			thaw_bdev(sb->s_bdev, sb);
804 		}
805 
806 		break;
807 	}
808 	case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
809 		xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
810 		break;
811 	case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
812 		xfs_force_shutdown(mp,
813 				SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
814 		break;
815 	default:
816 		return -EINVAL;
817 	}
818 
819 	return 0;
820 }
821 
822 /*
823  * Force a shutdown of the filesystem instantly while keeping the filesystem
824  * consistent. We don't do an unmount here; just shutdown the shop, make sure
825  * that absolutely nothing persistent happens to this filesystem after this
826  * point.
827  */
828 void
829 xfs_do_force_shutdown(
830 	xfs_mount_t	*mp,
831 	int		flags,
832 	char		*fname,
833 	int		lnnum)
834 {
835 	int		logerror;
836 
837 	logerror = flags & SHUTDOWN_LOG_IO_ERROR;
838 
839 	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
840 		xfs_notice(mp,
841 	"%s(0x%x) called from line %d of file %s.  Return address = 0x%p",
842 			__func__, flags, lnnum, fname, __return_address);
843 	}
844 	/*
845 	 * No need to duplicate efforts.
846 	 */
847 	if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
848 		return;
849 
850 	/*
851 	 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
852 	 * queue up anybody new on the log reservations, and wakes up
853 	 * everybody who's sleeping on log reservations to tell them
854 	 * the bad news.
855 	 */
856 	if (xfs_log_force_umount(mp, logerror))
857 		return;
858 
859 	if (flags & SHUTDOWN_CORRUPT_INCORE) {
860 		xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT,
861     "Corruption of in-memory data detected.  Shutting down filesystem");
862 		if (XFS_ERRLEVEL_HIGH <= xfs_error_level)
863 			xfs_stack_trace();
864 	} else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
865 		if (logerror) {
866 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR,
867 		"Log I/O Error Detected.  Shutting down filesystem");
868 		} else if (flags & SHUTDOWN_DEVICE_REQ) {
869 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
870 		"All device paths lost.  Shutting down filesystem");
871 		} else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
872 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
873 		"I/O Error Detected. Shutting down filesystem");
874 		}
875 	}
876 	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
877 		xfs_alert(mp,
878 	"Please umount the filesystem and rectify the problem(s)");
879 	}
880 }
881