xref: /openbmc/linux/fs/xfs/xfs_fsops.c (revision 6d3ebaae7c20128bfa6965a8d5cee0d1deea8486)
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_filestream.h"
44 
45 /*
46  * File system operations
47  */
48 
49 int
50 xfs_fs_geometry(
51 	xfs_mount_t		*mp,
52 	xfs_fsop_geom_t		*geo,
53 	int			new_version)
54 {
55 
56 	memset(geo, 0, sizeof(*geo));
57 
58 	geo->blocksize = mp->m_sb.sb_blocksize;
59 	geo->rtextsize = mp->m_sb.sb_rextsize;
60 	geo->agblocks = mp->m_sb.sb_agblocks;
61 	geo->agcount = mp->m_sb.sb_agcount;
62 	geo->logblocks = mp->m_sb.sb_logblocks;
63 	geo->sectsize = mp->m_sb.sb_sectsize;
64 	geo->inodesize = mp->m_sb.sb_inodesize;
65 	geo->imaxpct = mp->m_sb.sb_imax_pct;
66 	geo->datablocks = mp->m_sb.sb_dblocks;
67 	geo->rtblocks = mp->m_sb.sb_rblocks;
68 	geo->rtextents = mp->m_sb.sb_rextents;
69 	geo->logstart = mp->m_sb.sb_logstart;
70 	ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
71 	memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
72 	if (new_version >= 2) {
73 		geo->sunit = mp->m_sb.sb_unit;
74 		geo->swidth = mp->m_sb.sb_width;
75 	}
76 	if (new_version >= 3) {
77 		geo->version = XFS_FSOP_GEOM_VERSION;
78 		geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
79 			     XFS_FSOP_GEOM_FLAGS_DIRV2 |
80 			(xfs_sb_version_hasattr(&mp->m_sb) ?
81 				XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
82 			(xfs_sb_version_hasquota(&mp->m_sb) ?
83 				XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
84 			(xfs_sb_version_hasalign(&mp->m_sb) ?
85 				XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
86 			(xfs_sb_version_hasdalign(&mp->m_sb) ?
87 				XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
88 			(xfs_sb_version_hasextflgbit(&mp->m_sb) ?
89 				XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
90 			(xfs_sb_version_hassector(&mp->m_sb) ?
91 				XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
92 			(xfs_sb_version_hasasciici(&mp->m_sb) ?
93 				XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
94 			(xfs_sb_version_haslazysbcount(&mp->m_sb) ?
95 				XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
96 			(xfs_sb_version_hasattr2(&mp->m_sb) ?
97 				XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) |
98 			(xfs_sb_version_hasprojid32bit(&mp->m_sb) ?
99 				XFS_FSOP_GEOM_FLAGS_PROJID32 : 0) |
100 			(xfs_sb_version_hascrc(&mp->m_sb) ?
101 				XFS_FSOP_GEOM_FLAGS_V5SB : 0) |
102 			(xfs_sb_version_hasftype(&mp->m_sb) ?
103 				XFS_FSOP_GEOM_FLAGS_FTYPE : 0) |
104 			(xfs_sb_version_hasfinobt(&mp->m_sb) ?
105 				XFS_FSOP_GEOM_FLAGS_FINOBT : 0);
106 		geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
107 				mp->m_sb.sb_logsectsize : BBSIZE;
108 		geo->rtsectsize = mp->m_sb.sb_blocksize;
109 		geo->dirblocksize = mp->m_dir_geo->blksize;
110 	}
111 	if (new_version >= 4) {
112 		geo->flags |=
113 			(xfs_sb_version_haslogv2(&mp->m_sb) ?
114 				XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
115 		geo->logsunit = mp->m_sb.sb_logsunit;
116 	}
117 	return 0;
118 }
119 
120 static struct xfs_buf *
121 xfs_growfs_get_hdr_buf(
122 	struct xfs_mount	*mp,
123 	xfs_daddr_t		blkno,
124 	size_t			numblks,
125 	int			flags,
126 	const struct xfs_buf_ops *ops)
127 {
128 	struct xfs_buf		*bp;
129 
130 	bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags);
131 	if (!bp)
132 		return NULL;
133 
134 	xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
135 	bp->b_bn = blkno;
136 	bp->b_maps[0].bm_bn = blkno;
137 	bp->b_ops = ops;
138 
139 	return bp;
140 }
141 
142 static int
143 xfs_growfs_data_private(
144 	xfs_mount_t		*mp,		/* mount point for filesystem */
145 	xfs_growfs_data_t	*in)		/* growfs data input struct */
146 {
147 	xfs_agf_t		*agf;
148 	struct xfs_agfl		*agfl;
149 	xfs_agi_t		*agi;
150 	xfs_agnumber_t		agno;
151 	xfs_extlen_t		agsize;
152 	xfs_extlen_t		tmpsize;
153 	xfs_alloc_rec_t		*arec;
154 	xfs_buf_t		*bp;
155 	int			bucket;
156 	int			dpct;
157 	int			error, saved_error = 0;
158 	xfs_agnumber_t		nagcount;
159 	xfs_agnumber_t		nagimax = 0;
160 	xfs_rfsblock_t		nb, nb_mod;
161 	xfs_rfsblock_t		new;
162 	xfs_rfsblock_t		nfree;
163 	xfs_agnumber_t		oagcount;
164 	int			pct;
165 	xfs_trans_t		*tp;
166 
167 	nb = in->newblocks;
168 	pct = in->imaxpct;
169 	if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
170 		return -EINVAL;
171 	if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
172 		return error;
173 	dpct = pct - mp->m_sb.sb_imax_pct;
174 	error = xfs_buf_read_uncached(mp->m_ddev_targp,
175 				XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
176 				XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
177 	if (error)
178 		return error;
179 	xfs_buf_relse(bp);
180 
181 	new = nb;	/* use new as a temporary here */
182 	nb_mod = do_div(new, mp->m_sb.sb_agblocks);
183 	nagcount = new + (nb_mod != 0);
184 	if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
185 		nagcount--;
186 		nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
187 		if (nb < mp->m_sb.sb_dblocks)
188 			return -EINVAL;
189 	}
190 	new = nb - mp->m_sb.sb_dblocks;
191 	oagcount = mp->m_sb.sb_agcount;
192 
193 	/* allocate the new per-ag structures */
194 	if (nagcount > oagcount) {
195 		error = xfs_initialize_perag(mp, nagcount, &nagimax);
196 		if (error)
197 			return error;
198 	}
199 
200 	tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
201 	tp->t_flags |= XFS_TRANS_RESERVE;
202 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_growdata,
203 				  XFS_GROWFS_SPACE_RES(mp), 0);
204 	if (error) {
205 		xfs_trans_cancel(tp, 0);
206 		return error;
207 	}
208 
209 	/*
210 	 * Write new AG headers to disk. Non-transactional, but written
211 	 * synchronously so they are completed prior to the growfs transaction
212 	 * being logged.
213 	 */
214 	nfree = 0;
215 	for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
216 		__be32	*agfl_bno;
217 
218 		/*
219 		 * AG freespace header block
220 		 */
221 		bp = xfs_growfs_get_hdr_buf(mp,
222 				XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
223 				XFS_FSS_TO_BB(mp, 1), 0,
224 				&xfs_agf_buf_ops);
225 		if (!bp) {
226 			error = -ENOMEM;
227 			goto error0;
228 		}
229 
230 		agf = XFS_BUF_TO_AGF(bp);
231 		agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
232 		agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
233 		agf->agf_seqno = cpu_to_be32(agno);
234 		if (agno == nagcount - 1)
235 			agsize =
236 				nb -
237 				(agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
238 		else
239 			agsize = mp->m_sb.sb_agblocks;
240 		agf->agf_length = cpu_to_be32(agsize);
241 		agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
242 		agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
243 		agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
244 		agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
245 		agf->agf_flfirst = 0;
246 		agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
247 		agf->agf_flcount = 0;
248 		tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
249 		agf->agf_freeblks = cpu_to_be32(tmpsize);
250 		agf->agf_longest = cpu_to_be32(tmpsize);
251 		if (xfs_sb_version_hascrc(&mp->m_sb))
252 			uuid_copy(&agf->agf_uuid, &mp->m_sb.sb_uuid);
253 
254 		error = xfs_bwrite(bp);
255 		xfs_buf_relse(bp);
256 		if (error)
257 			goto error0;
258 
259 		/*
260 		 * AG freelist header block
261 		 */
262 		bp = xfs_growfs_get_hdr_buf(mp,
263 				XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
264 				XFS_FSS_TO_BB(mp, 1), 0,
265 				&xfs_agfl_buf_ops);
266 		if (!bp) {
267 			error = -ENOMEM;
268 			goto error0;
269 		}
270 
271 		agfl = XFS_BUF_TO_AGFL(bp);
272 		if (xfs_sb_version_hascrc(&mp->m_sb)) {
273 			agfl->agfl_magicnum = cpu_to_be32(XFS_AGFL_MAGIC);
274 			agfl->agfl_seqno = cpu_to_be32(agno);
275 			uuid_copy(&agfl->agfl_uuid, &mp->m_sb.sb_uuid);
276 		}
277 
278 		agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, bp);
279 		for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++)
280 			agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK);
281 
282 		error = xfs_bwrite(bp);
283 		xfs_buf_relse(bp);
284 		if (error)
285 			goto error0;
286 
287 		/*
288 		 * AG inode header block
289 		 */
290 		bp = xfs_growfs_get_hdr_buf(mp,
291 				XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
292 				XFS_FSS_TO_BB(mp, 1), 0,
293 				&xfs_agi_buf_ops);
294 		if (!bp) {
295 			error = -ENOMEM;
296 			goto error0;
297 		}
298 
299 		agi = XFS_BUF_TO_AGI(bp);
300 		agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
301 		agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
302 		agi->agi_seqno = cpu_to_be32(agno);
303 		agi->agi_length = cpu_to_be32(agsize);
304 		agi->agi_count = 0;
305 		agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
306 		agi->agi_level = cpu_to_be32(1);
307 		agi->agi_freecount = 0;
308 		agi->agi_newino = cpu_to_be32(NULLAGINO);
309 		agi->agi_dirino = cpu_to_be32(NULLAGINO);
310 		if (xfs_sb_version_hascrc(&mp->m_sb))
311 			uuid_copy(&agi->agi_uuid, &mp->m_sb.sb_uuid);
312 		if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
313 			agi->agi_free_root = cpu_to_be32(XFS_FIBT_BLOCK(mp));
314 			agi->agi_free_level = cpu_to_be32(1);
315 		}
316 		for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
317 			agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
318 
319 		error = xfs_bwrite(bp);
320 		xfs_buf_relse(bp);
321 		if (error)
322 			goto error0;
323 
324 		/*
325 		 * BNO btree root block
326 		 */
327 		bp = xfs_growfs_get_hdr_buf(mp,
328 				XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
329 				BTOBB(mp->m_sb.sb_blocksize), 0,
330 				&xfs_allocbt_buf_ops);
331 
332 		if (!bp) {
333 			error = -ENOMEM;
334 			goto error0;
335 		}
336 
337 		if (xfs_sb_version_hascrc(&mp->m_sb))
338 			xfs_btree_init_block(mp, bp, XFS_ABTB_CRC_MAGIC, 0, 1,
339 						agno, XFS_BTREE_CRC_BLOCKS);
340 		else
341 			xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1,
342 						agno, 0);
343 
344 		arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
345 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
346 		arec->ar_blockcount = cpu_to_be32(
347 			agsize - be32_to_cpu(arec->ar_startblock));
348 
349 		error = xfs_bwrite(bp);
350 		xfs_buf_relse(bp);
351 		if (error)
352 			goto error0;
353 
354 		/*
355 		 * CNT btree root block
356 		 */
357 		bp = xfs_growfs_get_hdr_buf(mp,
358 				XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
359 				BTOBB(mp->m_sb.sb_blocksize), 0,
360 				&xfs_allocbt_buf_ops);
361 		if (!bp) {
362 			error = -ENOMEM;
363 			goto error0;
364 		}
365 
366 		if (xfs_sb_version_hascrc(&mp->m_sb))
367 			xfs_btree_init_block(mp, bp, XFS_ABTC_CRC_MAGIC, 0, 1,
368 						agno, XFS_BTREE_CRC_BLOCKS);
369 		else
370 			xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1,
371 						agno, 0);
372 
373 		arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
374 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
375 		arec->ar_blockcount = cpu_to_be32(
376 			agsize - be32_to_cpu(arec->ar_startblock));
377 		nfree += be32_to_cpu(arec->ar_blockcount);
378 
379 		error = xfs_bwrite(bp);
380 		xfs_buf_relse(bp);
381 		if (error)
382 			goto error0;
383 
384 		/*
385 		 * INO btree root block
386 		 */
387 		bp = xfs_growfs_get_hdr_buf(mp,
388 				XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
389 				BTOBB(mp->m_sb.sb_blocksize), 0,
390 				&xfs_inobt_buf_ops);
391 		if (!bp) {
392 			error = -ENOMEM;
393 			goto error0;
394 		}
395 
396 		if (xfs_sb_version_hascrc(&mp->m_sb))
397 			xfs_btree_init_block(mp, bp, XFS_IBT_CRC_MAGIC, 0, 0,
398 						agno, XFS_BTREE_CRC_BLOCKS);
399 		else
400 			xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0,
401 						agno, 0);
402 
403 		error = xfs_bwrite(bp);
404 		xfs_buf_relse(bp);
405 		if (error)
406 			goto error0;
407 
408 		/*
409 		 * FINO btree root block
410 		 */
411 		if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
412 			bp = xfs_growfs_get_hdr_buf(mp,
413 				XFS_AGB_TO_DADDR(mp, agno, XFS_FIBT_BLOCK(mp)),
414 				BTOBB(mp->m_sb.sb_blocksize), 0,
415 				&xfs_inobt_buf_ops);
416 			if (!bp) {
417 				error = -ENOMEM;
418 				goto error0;
419 			}
420 
421 			if (xfs_sb_version_hascrc(&mp->m_sb))
422 				xfs_btree_init_block(mp, bp, XFS_FIBT_CRC_MAGIC,
423 						     0, 0, agno,
424 						     XFS_BTREE_CRC_BLOCKS);
425 			else
426 				xfs_btree_init_block(mp, bp, XFS_FIBT_MAGIC, 0,
427 						     0, agno, 0);
428 
429 			error = xfs_bwrite(bp);
430 			xfs_buf_relse(bp);
431 			if (error)
432 				goto error0;
433 		}
434 
435 	}
436 	xfs_trans_agblocks_delta(tp, nfree);
437 	/*
438 	 * There are new blocks in the old last a.g.
439 	 */
440 	if (new) {
441 		/*
442 		 * Change the agi length.
443 		 */
444 		error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
445 		if (error) {
446 			goto error0;
447 		}
448 		ASSERT(bp);
449 		agi = XFS_BUF_TO_AGI(bp);
450 		be32_add_cpu(&agi->agi_length, new);
451 		ASSERT(nagcount == oagcount ||
452 		       be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
453 		xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
454 		/*
455 		 * Change agf length.
456 		 */
457 		error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
458 		if (error) {
459 			goto error0;
460 		}
461 		ASSERT(bp);
462 		agf = XFS_BUF_TO_AGF(bp);
463 		be32_add_cpu(&agf->agf_length, new);
464 		ASSERT(be32_to_cpu(agf->agf_length) ==
465 		       be32_to_cpu(agi->agi_length));
466 
467 		xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
468 		/*
469 		 * Free the new space.
470 		 */
471 		error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
472 			be32_to_cpu(agf->agf_length) - new), new);
473 		if (error) {
474 			goto error0;
475 		}
476 	}
477 
478 	/*
479 	 * Update changed superblock fields transactionally. These are not
480 	 * seen by the rest of the world until the transaction commit applies
481 	 * them atomically to the superblock.
482 	 */
483 	if (nagcount > oagcount)
484 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
485 	if (nb > mp->m_sb.sb_dblocks)
486 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
487 				 nb - mp->m_sb.sb_dblocks);
488 	if (nfree)
489 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
490 	if (dpct)
491 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
492 	error = xfs_trans_commit(tp, 0);
493 	if (error)
494 		return error;
495 
496 	/* New allocation groups fully initialized, so update mount struct */
497 	if (nagimax)
498 		mp->m_maxagi = nagimax;
499 	if (mp->m_sb.sb_imax_pct) {
500 		__uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
501 		do_div(icount, 100);
502 		mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
503 	} else
504 		mp->m_maxicount = 0;
505 	xfs_set_low_space_thresholds(mp);
506 
507 	/* update secondary superblocks. */
508 	for (agno = 1; agno < nagcount; agno++) {
509 		error = 0;
510 		/*
511 		 * new secondary superblocks need to be zeroed, not read from
512 		 * disk as the contents of the new area we are growing into is
513 		 * completely unknown.
514 		 */
515 		if (agno < oagcount) {
516 			error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
517 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
518 				  XFS_FSS_TO_BB(mp, 1), 0, &bp,
519 				  &xfs_sb_buf_ops);
520 		} else {
521 			bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
522 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
523 				  XFS_FSS_TO_BB(mp, 1), 0);
524 			if (bp) {
525 				bp->b_ops = &xfs_sb_buf_ops;
526 				xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
527 			} else
528 				error = -ENOMEM;
529 		}
530 
531 		/*
532 		 * If we get an error reading or writing alternate superblocks,
533 		 * continue.  xfs_repair chooses the "best" superblock based
534 		 * on most matches; if we break early, we'll leave more
535 		 * superblocks un-updated than updated, and xfs_repair may
536 		 * pick them over the properly-updated primary.
537 		 */
538 		if (error) {
539 			xfs_warn(mp,
540 		"error %d reading secondary superblock for ag %d",
541 				error, agno);
542 			saved_error = error;
543 			continue;
544 		}
545 		xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
546 
547 		error = xfs_bwrite(bp);
548 		xfs_buf_relse(bp);
549 		if (error) {
550 			xfs_warn(mp,
551 		"write error %d updating secondary superblock for ag %d",
552 				error, agno);
553 			saved_error = error;
554 			continue;
555 		}
556 	}
557 	return saved_error ? saved_error : error;
558 
559  error0:
560 	xfs_trans_cancel(tp, XFS_TRANS_ABORT);
561 	return error;
562 }
563 
564 static int
565 xfs_growfs_log_private(
566 	xfs_mount_t		*mp,	/* mount point for filesystem */
567 	xfs_growfs_log_t	*in)	/* growfs log input struct */
568 {
569 	xfs_extlen_t		nb;
570 
571 	nb = in->newblocks;
572 	if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
573 		return -EINVAL;
574 	if (nb == mp->m_sb.sb_logblocks &&
575 	    in->isint == (mp->m_sb.sb_logstart != 0))
576 		return -EINVAL;
577 	/*
578 	 * Moving the log is hard, need new interfaces to sync
579 	 * the log first, hold off all activity while moving it.
580 	 * Can have shorter or longer log in the same space,
581 	 * or transform internal to external log or vice versa.
582 	 */
583 	return -ENOSYS;
584 }
585 
586 /*
587  * protected versions of growfs function acquire and release locks on the mount
588  * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
589  * XFS_IOC_FSGROWFSRT
590  */
591 
592 
593 int
594 xfs_growfs_data(
595 	xfs_mount_t		*mp,
596 	xfs_growfs_data_t	*in)
597 {
598 	int error;
599 
600 	if (!capable(CAP_SYS_ADMIN))
601 		return -EPERM;
602 	if (!mutex_trylock(&mp->m_growlock))
603 		return -EWOULDBLOCK;
604 	error = xfs_growfs_data_private(mp, in);
605 	mutex_unlock(&mp->m_growlock);
606 	return error;
607 }
608 
609 int
610 xfs_growfs_log(
611 	xfs_mount_t		*mp,
612 	xfs_growfs_log_t	*in)
613 {
614 	int error;
615 
616 	if (!capable(CAP_SYS_ADMIN))
617 		return -EPERM;
618 	if (!mutex_trylock(&mp->m_growlock))
619 		return -EWOULDBLOCK;
620 	error = xfs_growfs_log_private(mp, in);
621 	mutex_unlock(&mp->m_growlock);
622 	return error;
623 }
624 
625 /*
626  * exported through ioctl XFS_IOC_FSCOUNTS
627  */
628 
629 int
630 xfs_fs_counts(
631 	xfs_mount_t		*mp,
632 	xfs_fsop_counts_t	*cnt)
633 {
634 	xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
635 	spin_lock(&mp->m_sb_lock);
636 	cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
637 	cnt->freertx = mp->m_sb.sb_frextents;
638 	cnt->freeino = mp->m_sb.sb_ifree;
639 	cnt->allocino = mp->m_sb.sb_icount;
640 	spin_unlock(&mp->m_sb_lock);
641 	return 0;
642 }
643 
644 /*
645  * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
646  *
647  * xfs_reserve_blocks is called to set m_resblks
648  * in the in-core mount table. The number of unused reserved blocks
649  * is kept in m_resblks_avail.
650  *
651  * Reserve the requested number of blocks if available. Otherwise return
652  * as many as possible to satisfy the request. The actual number
653  * reserved are returned in outval
654  *
655  * A null inval pointer indicates that only the current reserved blocks
656  * available  should  be returned no settings are changed.
657  */
658 
659 int
660 xfs_reserve_blocks(
661 	xfs_mount_t             *mp,
662 	__uint64_t              *inval,
663 	xfs_fsop_resblks_t      *outval)
664 {
665 	__int64_t		lcounter, delta, fdblks_delta;
666 	__uint64_t		request;
667 
668 	/* If inval is null, report current values and return */
669 	if (inval == (__uint64_t *)NULL) {
670 		if (!outval)
671 			return -EINVAL;
672 		outval->resblks = mp->m_resblks;
673 		outval->resblks_avail = mp->m_resblks_avail;
674 		return 0;
675 	}
676 
677 	request = *inval;
678 
679 	/*
680 	 * With per-cpu counters, this becomes an interesting
681 	 * problem. we needto work out if we are freeing or allocation
682 	 * blocks first, then we can do the modification as necessary.
683 	 *
684 	 * We do this under the m_sb_lock so that if we are near
685 	 * ENOSPC, we will hold out any changes while we work out
686 	 * what to do. This means that the amount of free space can
687 	 * change while we do this, so we need to retry if we end up
688 	 * trying to reserve more space than is available.
689 	 *
690 	 * We also use the xfs_mod_incore_sb() interface so that we
691 	 * don't have to care about whether per cpu counter are
692 	 * enabled, disabled or even compiled in....
693 	 */
694 retry:
695 	spin_lock(&mp->m_sb_lock);
696 	xfs_icsb_sync_counters_locked(mp, 0);
697 
698 	/*
699 	 * If our previous reservation was larger than the current value,
700 	 * then move any unused blocks back to the free pool.
701 	 */
702 	fdblks_delta = 0;
703 	if (mp->m_resblks > request) {
704 		lcounter = mp->m_resblks_avail - request;
705 		if (lcounter  > 0) {		/* release unused blocks */
706 			fdblks_delta = lcounter;
707 			mp->m_resblks_avail -= lcounter;
708 		}
709 		mp->m_resblks = request;
710 	} else {
711 		__int64_t	free;
712 
713 		free =  mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
714 		if (!free)
715 			goto out; /* ENOSPC and fdblks_delta = 0 */
716 
717 		delta = request - mp->m_resblks;
718 		lcounter = free - delta;
719 		if (lcounter < 0) {
720 			/* We can't satisfy the request, just get what we can */
721 			mp->m_resblks += free;
722 			mp->m_resblks_avail += free;
723 			fdblks_delta = -free;
724 		} else {
725 			fdblks_delta = -delta;
726 			mp->m_resblks = request;
727 			mp->m_resblks_avail += delta;
728 		}
729 	}
730 out:
731 	if (outval) {
732 		outval->resblks = mp->m_resblks;
733 		outval->resblks_avail = mp->m_resblks_avail;
734 	}
735 	spin_unlock(&mp->m_sb_lock);
736 
737 	if (fdblks_delta) {
738 		/*
739 		 * If we are putting blocks back here, m_resblks_avail is
740 		 * already at its max so this will put it in the free pool.
741 		 *
742 		 * If we need space, we'll either succeed in getting it
743 		 * from the free block count or we'll get an enospc. If
744 		 * we get a ENOSPC, it means things changed while we were
745 		 * calculating fdblks_delta and so we should try again to
746 		 * see if there is anything left to reserve.
747 		 *
748 		 * Don't set the reserved flag here - we don't want to reserve
749 		 * the extra reserve blocks from the reserve.....
750 		 */
751 		int error;
752 		error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
753 						 fdblks_delta, 0);
754 		if (error == -ENOSPC)
755 			goto retry;
756 	}
757 	return 0;
758 }
759 
760 /*
761  * Dump a transaction into the log that contains no real change. This is needed
762  * to be able to make the log dirty or stamp the current tail LSN into the log
763  * during the covering operation.
764  *
765  * We cannot use an inode here for this - that will push dirty state back up
766  * into the VFS and then periodic inode flushing will prevent log covering from
767  * making progress. Hence we log a field in the superblock instead and use a
768  * synchronous transaction to ensure the superblock is immediately unpinned
769  * and can be written back.
770  */
771 int
772 xfs_fs_log_dummy(
773 	xfs_mount_t	*mp)
774 {
775 	xfs_trans_t	*tp;
776 	int		error;
777 
778 	tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
779 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_sb, 0, 0);
780 	if (error) {
781 		xfs_trans_cancel(tp, 0);
782 		return error;
783 	}
784 
785 	/* log the UUID because it is an unchanging field */
786 	xfs_mod_sb(tp, XFS_SB_UUID);
787 	xfs_trans_set_sync(tp);
788 	return xfs_trans_commit(tp, 0);
789 }
790 
791 int
792 xfs_fs_goingdown(
793 	xfs_mount_t	*mp,
794 	__uint32_t	inflags)
795 {
796 	switch (inflags) {
797 	case XFS_FSOP_GOING_FLAGS_DEFAULT: {
798 		struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
799 
800 		if (sb && !IS_ERR(sb)) {
801 			xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
802 			thaw_bdev(sb->s_bdev, sb);
803 		}
804 
805 		break;
806 	}
807 	case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
808 		xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
809 		break;
810 	case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
811 		xfs_force_shutdown(mp,
812 				SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
813 		break;
814 	default:
815 		return -EINVAL;
816 	}
817 
818 	return 0;
819 }
820 
821 /*
822  * Force a shutdown of the filesystem instantly while keeping the filesystem
823  * consistent. We don't do an unmount here; just shutdown the shop, make sure
824  * that absolutely nothing persistent happens to this filesystem after this
825  * point.
826  */
827 void
828 xfs_do_force_shutdown(
829 	xfs_mount_t	*mp,
830 	int		flags,
831 	char		*fname,
832 	int		lnnum)
833 {
834 	int		logerror;
835 
836 	logerror = flags & SHUTDOWN_LOG_IO_ERROR;
837 
838 	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
839 		xfs_notice(mp,
840 	"%s(0x%x) called from line %d of file %s.  Return address = 0x%p",
841 			__func__, flags, lnnum, fname, __return_address);
842 	}
843 	/*
844 	 * No need to duplicate efforts.
845 	 */
846 	if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
847 		return;
848 
849 	/*
850 	 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
851 	 * queue up anybody new on the log reservations, and wakes up
852 	 * everybody who's sleeping on log reservations to tell them
853 	 * the bad news.
854 	 */
855 	if (xfs_log_force_umount(mp, logerror))
856 		return;
857 
858 	if (flags & SHUTDOWN_CORRUPT_INCORE) {
859 		xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT,
860     "Corruption of in-memory data detected.  Shutting down filesystem");
861 		if (XFS_ERRLEVEL_HIGH <= xfs_error_level)
862 			xfs_stack_trace();
863 	} else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
864 		if (logerror) {
865 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR,
866 		"Log I/O Error Detected.  Shutting down filesystem");
867 		} else if (flags & SHUTDOWN_DEVICE_REQ) {
868 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
869 		"All device paths lost.  Shutting down filesystem");
870 		} else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
871 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
872 		"I/O Error Detected. Shutting down filesystem");
873 		}
874 	}
875 	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
876 		xfs_alert(mp,
877 	"Please umount the filesystem and rectify the problem(s)");
878 	}
879 }
880