xref: /linux/fs/xfs/xfs_fsops.c (revision 4949009eb8d40a441dcddcd96e101e77d31cf1b2)
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_mount.h"
26 #include "xfs_da_format.h"
27 #include "xfs_da_btree.h"
28 #include "xfs_inode.h"
29 #include "xfs_trans.h"
30 #include "xfs_inode_item.h"
31 #include "xfs_error.h"
32 #include "xfs_btree.h"
33 #include "xfs_alloc_btree.h"
34 #include "xfs_alloc.h"
35 #include "xfs_ialloc.h"
36 #include "xfs_fsops.h"
37 #include "xfs_itable.h"
38 #include "xfs_trans_space.h"
39 #include "xfs_rtalloc.h"
40 #include "xfs_trace.h"
41 #include "xfs_log.h"
42 #include "xfs_filestream.h"
43 
44 /*
45  * File system operations
46  */
47 
48 int
49 xfs_fs_geometry(
50 	xfs_mount_t		*mp,
51 	xfs_fsop_geom_t		*geo,
52 	int			new_version)
53 {
54 
55 	memset(geo, 0, sizeof(*geo));
56 
57 	geo->blocksize = mp->m_sb.sb_blocksize;
58 	geo->rtextsize = mp->m_sb.sb_rextsize;
59 	geo->agblocks = mp->m_sb.sb_agblocks;
60 	geo->agcount = mp->m_sb.sb_agcount;
61 	geo->logblocks = mp->m_sb.sb_logblocks;
62 	geo->sectsize = mp->m_sb.sb_sectsize;
63 	geo->inodesize = mp->m_sb.sb_inodesize;
64 	geo->imaxpct = mp->m_sb.sb_imax_pct;
65 	geo->datablocks = mp->m_sb.sb_dblocks;
66 	geo->rtblocks = mp->m_sb.sb_rblocks;
67 	geo->rtextents = mp->m_sb.sb_rextents;
68 	geo->logstart = mp->m_sb.sb_logstart;
69 	ASSERT(sizeof(geo->uuid)==sizeof(mp->m_sb.sb_uuid));
70 	memcpy(geo->uuid, &mp->m_sb.sb_uuid, sizeof(mp->m_sb.sb_uuid));
71 	if (new_version >= 2) {
72 		geo->sunit = mp->m_sb.sb_unit;
73 		geo->swidth = mp->m_sb.sb_width;
74 	}
75 	if (new_version >= 3) {
76 		geo->version = XFS_FSOP_GEOM_VERSION;
77 		geo->flags = XFS_FSOP_GEOM_FLAGS_NLINK |
78 			     XFS_FSOP_GEOM_FLAGS_DIRV2 |
79 			(xfs_sb_version_hasattr(&mp->m_sb) ?
80 				XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
81 			(xfs_sb_version_hasquota(&mp->m_sb) ?
82 				XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
83 			(xfs_sb_version_hasalign(&mp->m_sb) ?
84 				XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
85 			(xfs_sb_version_hasdalign(&mp->m_sb) ?
86 				XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
87 			(xfs_sb_version_hasextflgbit(&mp->m_sb) ?
88 				XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
89 			(xfs_sb_version_hassector(&mp->m_sb) ?
90 				XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
91 			(xfs_sb_version_hasasciici(&mp->m_sb) ?
92 				XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
93 			(xfs_sb_version_haslazysbcount(&mp->m_sb) ?
94 				XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
95 			(xfs_sb_version_hasattr2(&mp->m_sb) ?
96 				XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) |
97 			(xfs_sb_version_hasprojid32bit(&mp->m_sb) ?
98 				XFS_FSOP_GEOM_FLAGS_PROJID32 : 0) |
99 			(xfs_sb_version_hascrc(&mp->m_sb) ?
100 				XFS_FSOP_GEOM_FLAGS_V5SB : 0) |
101 			(xfs_sb_version_hasftype(&mp->m_sb) ?
102 				XFS_FSOP_GEOM_FLAGS_FTYPE : 0) |
103 			(xfs_sb_version_hasfinobt(&mp->m_sb) ?
104 				XFS_FSOP_GEOM_FLAGS_FINOBT : 0);
105 		geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
106 				mp->m_sb.sb_logsectsize : BBSIZE;
107 		geo->rtsectsize = mp->m_sb.sb_blocksize;
108 		geo->dirblocksize = mp->m_dir_geo->blksize;
109 	}
110 	if (new_version >= 4) {
111 		geo->flags |=
112 			(xfs_sb_version_haslogv2(&mp->m_sb) ?
113 				XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
114 		geo->logsunit = mp->m_sb.sb_logsunit;
115 	}
116 	return 0;
117 }
118 
119 static struct xfs_buf *
120 xfs_growfs_get_hdr_buf(
121 	struct xfs_mount	*mp,
122 	xfs_daddr_t		blkno,
123 	size_t			numblks,
124 	int			flags,
125 	const struct xfs_buf_ops *ops)
126 {
127 	struct xfs_buf		*bp;
128 
129 	bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags);
130 	if (!bp)
131 		return NULL;
132 
133 	xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
134 	bp->b_bn = blkno;
135 	bp->b_maps[0].bm_bn = blkno;
136 	bp->b_ops = ops;
137 
138 	return bp;
139 }
140 
141 static int
142 xfs_growfs_data_private(
143 	xfs_mount_t		*mp,		/* mount point for filesystem */
144 	xfs_growfs_data_t	*in)		/* growfs data input struct */
145 {
146 	xfs_agf_t		*agf;
147 	struct xfs_agfl		*agfl;
148 	xfs_agi_t		*agi;
149 	xfs_agnumber_t		agno;
150 	xfs_extlen_t		agsize;
151 	xfs_extlen_t		tmpsize;
152 	xfs_alloc_rec_t		*arec;
153 	xfs_buf_t		*bp;
154 	int			bucket;
155 	int			dpct;
156 	int			error, saved_error = 0;
157 	xfs_agnumber_t		nagcount;
158 	xfs_agnumber_t		nagimax = 0;
159 	xfs_rfsblock_t		nb, nb_mod;
160 	xfs_rfsblock_t		new;
161 	xfs_rfsblock_t		nfree;
162 	xfs_agnumber_t		oagcount;
163 	int			pct;
164 	xfs_trans_t		*tp;
165 
166 	nb = in->newblocks;
167 	pct = in->imaxpct;
168 	if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
169 		return -EINVAL;
170 	if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
171 		return error;
172 	dpct = pct - mp->m_sb.sb_imax_pct;
173 	error = xfs_buf_read_uncached(mp->m_ddev_targp,
174 				XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
175 				XFS_FSS_TO_BB(mp, 1), 0, &bp, NULL);
176 	if (error)
177 		return error;
178 	xfs_buf_relse(bp);
179 
180 	new = nb;	/* use new as a temporary here */
181 	nb_mod = do_div(new, mp->m_sb.sb_agblocks);
182 	nagcount = new + (nb_mod != 0);
183 	if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
184 		nagcount--;
185 		nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
186 		if (nb < mp->m_sb.sb_dblocks)
187 			return -EINVAL;
188 	}
189 	new = nb - mp->m_sb.sb_dblocks;
190 	oagcount = mp->m_sb.sb_agcount;
191 
192 	/* allocate the new per-ag structures */
193 	if (nagcount > oagcount) {
194 		error = xfs_initialize_perag(mp, nagcount, &nagimax);
195 		if (error)
196 			return error;
197 	}
198 
199 	tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
200 	tp->t_flags |= XFS_TRANS_RESERVE;
201 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_growdata,
202 				  XFS_GROWFS_SPACE_RES(mp), 0);
203 	if (error) {
204 		xfs_trans_cancel(tp, 0);
205 		return error;
206 	}
207 
208 	/*
209 	 * Write new AG headers to disk. Non-transactional, but written
210 	 * synchronously so they are completed prior to the growfs transaction
211 	 * being logged.
212 	 */
213 	nfree = 0;
214 	for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
215 		__be32	*agfl_bno;
216 
217 		/*
218 		 * AG freespace header block
219 		 */
220 		bp = xfs_growfs_get_hdr_buf(mp,
221 				XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
222 				XFS_FSS_TO_BB(mp, 1), 0,
223 				&xfs_agf_buf_ops);
224 		if (!bp) {
225 			error = -ENOMEM;
226 			goto error0;
227 		}
228 
229 		agf = XFS_BUF_TO_AGF(bp);
230 		agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
231 		agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
232 		agf->agf_seqno = cpu_to_be32(agno);
233 		if (agno == nagcount - 1)
234 			agsize =
235 				nb -
236 				(agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
237 		else
238 			agsize = mp->m_sb.sb_agblocks;
239 		agf->agf_length = cpu_to_be32(agsize);
240 		agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
241 		agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
242 		agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
243 		agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
244 		agf->agf_flfirst = 0;
245 		agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
246 		agf->agf_flcount = 0;
247 		tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
248 		agf->agf_freeblks = cpu_to_be32(tmpsize);
249 		agf->agf_longest = cpu_to_be32(tmpsize);
250 		if (xfs_sb_version_hascrc(&mp->m_sb))
251 			uuid_copy(&agf->agf_uuid, &mp->m_sb.sb_uuid);
252 
253 		error = xfs_bwrite(bp);
254 		xfs_buf_relse(bp);
255 		if (error)
256 			goto error0;
257 
258 		/*
259 		 * AG freelist header block
260 		 */
261 		bp = xfs_growfs_get_hdr_buf(mp,
262 				XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
263 				XFS_FSS_TO_BB(mp, 1), 0,
264 				&xfs_agfl_buf_ops);
265 		if (!bp) {
266 			error = -ENOMEM;
267 			goto error0;
268 		}
269 
270 		agfl = XFS_BUF_TO_AGFL(bp);
271 		if (xfs_sb_version_hascrc(&mp->m_sb)) {
272 			agfl->agfl_magicnum = cpu_to_be32(XFS_AGFL_MAGIC);
273 			agfl->agfl_seqno = cpu_to_be32(agno);
274 			uuid_copy(&agfl->agfl_uuid, &mp->m_sb.sb_uuid);
275 		}
276 
277 		agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, bp);
278 		for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++)
279 			agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK);
280 
281 		error = xfs_bwrite(bp);
282 		xfs_buf_relse(bp);
283 		if (error)
284 			goto error0;
285 
286 		/*
287 		 * AG inode header block
288 		 */
289 		bp = xfs_growfs_get_hdr_buf(mp,
290 				XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
291 				XFS_FSS_TO_BB(mp, 1), 0,
292 				&xfs_agi_buf_ops);
293 		if (!bp) {
294 			error = -ENOMEM;
295 			goto error0;
296 		}
297 
298 		agi = XFS_BUF_TO_AGI(bp);
299 		agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
300 		agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
301 		agi->agi_seqno = cpu_to_be32(agno);
302 		agi->agi_length = cpu_to_be32(agsize);
303 		agi->agi_count = 0;
304 		agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
305 		agi->agi_level = cpu_to_be32(1);
306 		agi->agi_freecount = 0;
307 		agi->agi_newino = cpu_to_be32(NULLAGINO);
308 		agi->agi_dirino = cpu_to_be32(NULLAGINO);
309 		if (xfs_sb_version_hascrc(&mp->m_sb))
310 			uuid_copy(&agi->agi_uuid, &mp->m_sb.sb_uuid);
311 		if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
312 			agi->agi_free_root = cpu_to_be32(XFS_FIBT_BLOCK(mp));
313 			agi->agi_free_level = cpu_to_be32(1);
314 		}
315 		for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
316 			agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
317 
318 		error = xfs_bwrite(bp);
319 		xfs_buf_relse(bp);
320 		if (error)
321 			goto error0;
322 
323 		/*
324 		 * BNO btree root block
325 		 */
326 		bp = xfs_growfs_get_hdr_buf(mp,
327 				XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
328 				BTOBB(mp->m_sb.sb_blocksize), 0,
329 				&xfs_allocbt_buf_ops);
330 
331 		if (!bp) {
332 			error = -ENOMEM;
333 			goto error0;
334 		}
335 
336 		if (xfs_sb_version_hascrc(&mp->m_sb))
337 			xfs_btree_init_block(mp, bp, XFS_ABTB_CRC_MAGIC, 0, 1,
338 						agno, XFS_BTREE_CRC_BLOCKS);
339 		else
340 			xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1,
341 						agno, 0);
342 
343 		arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
344 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
345 		arec->ar_blockcount = cpu_to_be32(
346 			agsize - be32_to_cpu(arec->ar_startblock));
347 
348 		error = xfs_bwrite(bp);
349 		xfs_buf_relse(bp);
350 		if (error)
351 			goto error0;
352 
353 		/*
354 		 * CNT btree root block
355 		 */
356 		bp = xfs_growfs_get_hdr_buf(mp,
357 				XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
358 				BTOBB(mp->m_sb.sb_blocksize), 0,
359 				&xfs_allocbt_buf_ops);
360 		if (!bp) {
361 			error = -ENOMEM;
362 			goto error0;
363 		}
364 
365 		if (xfs_sb_version_hascrc(&mp->m_sb))
366 			xfs_btree_init_block(mp, bp, XFS_ABTC_CRC_MAGIC, 0, 1,
367 						agno, XFS_BTREE_CRC_BLOCKS);
368 		else
369 			xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1,
370 						agno, 0);
371 
372 		arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
373 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
374 		arec->ar_blockcount = cpu_to_be32(
375 			agsize - be32_to_cpu(arec->ar_startblock));
376 		nfree += be32_to_cpu(arec->ar_blockcount);
377 
378 		error = xfs_bwrite(bp);
379 		xfs_buf_relse(bp);
380 		if (error)
381 			goto error0;
382 
383 		/*
384 		 * INO btree root block
385 		 */
386 		bp = xfs_growfs_get_hdr_buf(mp,
387 				XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
388 				BTOBB(mp->m_sb.sb_blocksize), 0,
389 				&xfs_inobt_buf_ops);
390 		if (!bp) {
391 			error = -ENOMEM;
392 			goto error0;
393 		}
394 
395 		if (xfs_sb_version_hascrc(&mp->m_sb))
396 			xfs_btree_init_block(mp, bp, XFS_IBT_CRC_MAGIC, 0, 0,
397 						agno, XFS_BTREE_CRC_BLOCKS);
398 		else
399 			xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0,
400 						agno, 0);
401 
402 		error = xfs_bwrite(bp);
403 		xfs_buf_relse(bp);
404 		if (error)
405 			goto error0;
406 
407 		/*
408 		 * FINO btree root block
409 		 */
410 		if (xfs_sb_version_hasfinobt(&mp->m_sb)) {
411 			bp = xfs_growfs_get_hdr_buf(mp,
412 				XFS_AGB_TO_DADDR(mp, agno, XFS_FIBT_BLOCK(mp)),
413 				BTOBB(mp->m_sb.sb_blocksize), 0,
414 				&xfs_inobt_buf_ops);
415 			if (!bp) {
416 				error = -ENOMEM;
417 				goto error0;
418 			}
419 
420 			if (xfs_sb_version_hascrc(&mp->m_sb))
421 				xfs_btree_init_block(mp, bp, XFS_FIBT_CRC_MAGIC,
422 						     0, 0, agno,
423 						     XFS_BTREE_CRC_BLOCKS);
424 			else
425 				xfs_btree_init_block(mp, bp, XFS_FIBT_MAGIC, 0,
426 						     0, agno, 0);
427 
428 			error = xfs_bwrite(bp);
429 			xfs_buf_relse(bp);
430 			if (error)
431 				goto error0;
432 		}
433 
434 	}
435 	xfs_trans_agblocks_delta(tp, nfree);
436 	/*
437 	 * There are new blocks in the old last a.g.
438 	 */
439 	if (new) {
440 		/*
441 		 * Change the agi length.
442 		 */
443 		error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
444 		if (error) {
445 			goto error0;
446 		}
447 		ASSERT(bp);
448 		agi = XFS_BUF_TO_AGI(bp);
449 		be32_add_cpu(&agi->agi_length, new);
450 		ASSERT(nagcount == oagcount ||
451 		       be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
452 		xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
453 		/*
454 		 * Change agf length.
455 		 */
456 		error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
457 		if (error) {
458 			goto error0;
459 		}
460 		ASSERT(bp);
461 		agf = XFS_BUF_TO_AGF(bp);
462 		be32_add_cpu(&agf->agf_length, new);
463 		ASSERT(be32_to_cpu(agf->agf_length) ==
464 		       be32_to_cpu(agi->agi_length));
465 
466 		xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
467 		/*
468 		 * Free the new space.
469 		 */
470 		error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
471 			be32_to_cpu(agf->agf_length) - new), new);
472 		if (error) {
473 			goto error0;
474 		}
475 	}
476 
477 	/*
478 	 * Update changed superblock fields transactionally. These are not
479 	 * seen by the rest of the world until the transaction commit applies
480 	 * them atomically to the superblock.
481 	 */
482 	if (nagcount > oagcount)
483 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
484 	if (nb > mp->m_sb.sb_dblocks)
485 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
486 				 nb - mp->m_sb.sb_dblocks);
487 	if (nfree)
488 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
489 	if (dpct)
490 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
491 	error = xfs_trans_commit(tp, 0);
492 	if (error)
493 		return error;
494 
495 	/* New allocation groups fully initialized, so update mount struct */
496 	if (nagimax)
497 		mp->m_maxagi = nagimax;
498 	if (mp->m_sb.sb_imax_pct) {
499 		__uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
500 		do_div(icount, 100);
501 		mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
502 	} else
503 		mp->m_maxicount = 0;
504 	xfs_set_low_space_thresholds(mp);
505 
506 	/* update secondary superblocks. */
507 	for (agno = 1; agno < nagcount; agno++) {
508 		error = 0;
509 		/*
510 		 * new secondary superblocks need to be zeroed, not read from
511 		 * disk as the contents of the new area we are growing into is
512 		 * completely unknown.
513 		 */
514 		if (agno < oagcount) {
515 			error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
516 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
517 				  XFS_FSS_TO_BB(mp, 1), 0, &bp,
518 				  &xfs_sb_buf_ops);
519 		} else {
520 			bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
521 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
522 				  XFS_FSS_TO_BB(mp, 1), 0);
523 			if (bp) {
524 				bp->b_ops = &xfs_sb_buf_ops;
525 				xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
526 			} else
527 				error = -ENOMEM;
528 		}
529 
530 		/*
531 		 * If we get an error reading or writing alternate superblocks,
532 		 * continue.  xfs_repair chooses the "best" superblock based
533 		 * on most matches; if we break early, we'll leave more
534 		 * superblocks un-updated than updated, and xfs_repair may
535 		 * pick them over the properly-updated primary.
536 		 */
537 		if (error) {
538 			xfs_warn(mp,
539 		"error %d reading secondary superblock for ag %d",
540 				error, agno);
541 			saved_error = error;
542 			continue;
543 		}
544 		xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
545 
546 		error = xfs_bwrite(bp);
547 		xfs_buf_relse(bp);
548 		if (error) {
549 			xfs_warn(mp,
550 		"write error %d updating secondary superblock for ag %d",
551 				error, agno);
552 			saved_error = error;
553 			continue;
554 		}
555 	}
556 	return saved_error ? saved_error : error;
557 
558  error0:
559 	xfs_trans_cancel(tp, XFS_TRANS_ABORT);
560 	return error;
561 }
562 
563 static int
564 xfs_growfs_log_private(
565 	xfs_mount_t		*mp,	/* mount point for filesystem */
566 	xfs_growfs_log_t	*in)	/* growfs log input struct */
567 {
568 	xfs_extlen_t		nb;
569 
570 	nb = in->newblocks;
571 	if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
572 		return -EINVAL;
573 	if (nb == mp->m_sb.sb_logblocks &&
574 	    in->isint == (mp->m_sb.sb_logstart != 0))
575 		return -EINVAL;
576 	/*
577 	 * Moving the log is hard, need new interfaces to sync
578 	 * the log first, hold off all activity while moving it.
579 	 * Can have shorter or longer log in the same space,
580 	 * or transform internal to external log or vice versa.
581 	 */
582 	return -ENOSYS;
583 }
584 
585 /*
586  * protected versions of growfs function acquire and release locks on the mount
587  * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
588  * XFS_IOC_FSGROWFSRT
589  */
590 
591 
592 int
593 xfs_growfs_data(
594 	xfs_mount_t		*mp,
595 	xfs_growfs_data_t	*in)
596 {
597 	int error;
598 
599 	if (!capable(CAP_SYS_ADMIN))
600 		return -EPERM;
601 	if (!mutex_trylock(&mp->m_growlock))
602 		return -EWOULDBLOCK;
603 	error = xfs_growfs_data_private(mp, in);
604 	mutex_unlock(&mp->m_growlock);
605 	return error;
606 }
607 
608 int
609 xfs_growfs_log(
610 	xfs_mount_t		*mp,
611 	xfs_growfs_log_t	*in)
612 {
613 	int error;
614 
615 	if (!capable(CAP_SYS_ADMIN))
616 		return -EPERM;
617 	if (!mutex_trylock(&mp->m_growlock))
618 		return -EWOULDBLOCK;
619 	error = xfs_growfs_log_private(mp, in);
620 	mutex_unlock(&mp->m_growlock);
621 	return error;
622 }
623 
624 /*
625  * exported through ioctl XFS_IOC_FSCOUNTS
626  */
627 
628 int
629 xfs_fs_counts(
630 	xfs_mount_t		*mp,
631 	xfs_fsop_counts_t	*cnt)
632 {
633 	xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
634 	spin_lock(&mp->m_sb_lock);
635 	cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
636 	cnt->freertx = mp->m_sb.sb_frextents;
637 	cnt->freeino = mp->m_sb.sb_ifree;
638 	cnt->allocino = mp->m_sb.sb_icount;
639 	spin_unlock(&mp->m_sb_lock);
640 	return 0;
641 }
642 
643 /*
644  * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
645  *
646  * xfs_reserve_blocks is called to set m_resblks
647  * in the in-core mount table. The number of unused reserved blocks
648  * is kept in m_resblks_avail.
649  *
650  * Reserve the requested number of blocks if available. Otherwise return
651  * as many as possible to satisfy the request. The actual number
652  * reserved are returned in outval
653  *
654  * A null inval pointer indicates that only the current reserved blocks
655  * available  should  be returned no settings are changed.
656  */
657 
658 int
659 xfs_reserve_blocks(
660 	xfs_mount_t             *mp,
661 	__uint64_t              *inval,
662 	xfs_fsop_resblks_t      *outval)
663 {
664 	__int64_t		lcounter, delta, fdblks_delta;
665 	__uint64_t		request;
666 
667 	/* If inval is null, report current values and return */
668 	if (inval == (__uint64_t *)NULL) {
669 		if (!outval)
670 			return -EINVAL;
671 		outval->resblks = mp->m_resblks;
672 		outval->resblks_avail = mp->m_resblks_avail;
673 		return 0;
674 	}
675 
676 	request = *inval;
677 
678 	/*
679 	 * With per-cpu counters, this becomes an interesting
680 	 * problem. we needto work out if we are freeing or allocation
681 	 * blocks first, then we can do the modification as necessary.
682 	 *
683 	 * We do this under the m_sb_lock so that if we are near
684 	 * ENOSPC, we will hold out any changes while we work out
685 	 * what to do. This means that the amount of free space can
686 	 * change while we do this, so we need to retry if we end up
687 	 * trying to reserve more space than is available.
688 	 *
689 	 * We also use the xfs_mod_incore_sb() interface so that we
690 	 * don't have to care about whether per cpu counter are
691 	 * enabled, disabled or even compiled in....
692 	 */
693 retry:
694 	spin_lock(&mp->m_sb_lock);
695 	xfs_icsb_sync_counters_locked(mp, 0);
696 
697 	/*
698 	 * If our previous reservation was larger than the current value,
699 	 * then move any unused blocks back to the free pool.
700 	 */
701 	fdblks_delta = 0;
702 	if (mp->m_resblks > request) {
703 		lcounter = mp->m_resblks_avail - request;
704 		if (lcounter  > 0) {		/* release unused blocks */
705 			fdblks_delta = lcounter;
706 			mp->m_resblks_avail -= lcounter;
707 		}
708 		mp->m_resblks = request;
709 	} else {
710 		__int64_t	free;
711 
712 		free =  mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
713 		if (!free)
714 			goto out; /* ENOSPC and fdblks_delta = 0 */
715 
716 		delta = request - mp->m_resblks;
717 		lcounter = free - delta;
718 		if (lcounter < 0) {
719 			/* We can't satisfy the request, just get what we can */
720 			mp->m_resblks += free;
721 			mp->m_resblks_avail += free;
722 			fdblks_delta = -free;
723 		} else {
724 			fdblks_delta = -delta;
725 			mp->m_resblks = request;
726 			mp->m_resblks_avail += delta;
727 		}
728 	}
729 out:
730 	if (outval) {
731 		outval->resblks = mp->m_resblks;
732 		outval->resblks_avail = mp->m_resblks_avail;
733 	}
734 	spin_unlock(&mp->m_sb_lock);
735 
736 	if (fdblks_delta) {
737 		/*
738 		 * If we are putting blocks back here, m_resblks_avail is
739 		 * already at its max so this will put it in the free pool.
740 		 *
741 		 * If we need space, we'll either succeed in getting it
742 		 * from the free block count or we'll get an enospc. If
743 		 * we get a ENOSPC, it means things changed while we were
744 		 * calculating fdblks_delta and so we should try again to
745 		 * see if there is anything left to reserve.
746 		 *
747 		 * Don't set the reserved flag here - we don't want to reserve
748 		 * the extra reserve blocks from the reserve.....
749 		 */
750 		int error;
751 		error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
752 						 fdblks_delta, 0);
753 		if (error == -ENOSPC)
754 			goto retry;
755 	}
756 	return 0;
757 }
758 
759 /*
760  * Dump a transaction into the log that contains no real change. This is needed
761  * to be able to make the log dirty or stamp the current tail LSN into the log
762  * during the covering operation.
763  *
764  * We cannot use an inode here for this - that will push dirty state back up
765  * into the VFS and then periodic inode flushing will prevent log covering from
766  * making progress. Hence we log a field in the superblock instead and use a
767  * synchronous transaction to ensure the superblock is immediately unpinned
768  * and can be written back.
769  */
770 int
771 xfs_fs_log_dummy(
772 	xfs_mount_t	*mp)
773 {
774 	xfs_trans_t	*tp;
775 	int		error;
776 
777 	tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
778 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_sb, 0, 0);
779 	if (error) {
780 		xfs_trans_cancel(tp, 0);
781 		return error;
782 	}
783 
784 	/* log the UUID because it is an unchanging field */
785 	xfs_mod_sb(tp, XFS_SB_UUID);
786 	xfs_trans_set_sync(tp);
787 	return xfs_trans_commit(tp, 0);
788 }
789 
790 int
791 xfs_fs_goingdown(
792 	xfs_mount_t	*mp,
793 	__uint32_t	inflags)
794 {
795 	switch (inflags) {
796 	case XFS_FSOP_GOING_FLAGS_DEFAULT: {
797 		struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
798 
799 		if (sb && !IS_ERR(sb)) {
800 			xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
801 			thaw_bdev(sb->s_bdev, sb);
802 		}
803 
804 		break;
805 	}
806 	case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
807 		xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
808 		break;
809 	case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
810 		xfs_force_shutdown(mp,
811 				SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
812 		break;
813 	default:
814 		return -EINVAL;
815 	}
816 
817 	return 0;
818 }
819 
820 /*
821  * Force a shutdown of the filesystem instantly while keeping the filesystem
822  * consistent. We don't do an unmount here; just shutdown the shop, make sure
823  * that absolutely nothing persistent happens to this filesystem after this
824  * point.
825  */
826 void
827 xfs_do_force_shutdown(
828 	xfs_mount_t	*mp,
829 	int		flags,
830 	char		*fname,
831 	int		lnnum)
832 {
833 	int		logerror;
834 
835 	logerror = flags & SHUTDOWN_LOG_IO_ERROR;
836 
837 	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
838 		xfs_notice(mp,
839 	"%s(0x%x) called from line %d of file %s.  Return address = 0x%p",
840 			__func__, flags, lnnum, fname, __return_address);
841 	}
842 	/*
843 	 * No need to duplicate efforts.
844 	 */
845 	if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
846 		return;
847 
848 	/*
849 	 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
850 	 * queue up anybody new on the log reservations, and wakes up
851 	 * everybody who's sleeping on log reservations to tell them
852 	 * the bad news.
853 	 */
854 	if (xfs_log_force_umount(mp, logerror))
855 		return;
856 
857 	if (flags & SHUTDOWN_CORRUPT_INCORE) {
858 		xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT,
859     "Corruption of in-memory data detected.  Shutting down filesystem");
860 		if (XFS_ERRLEVEL_HIGH <= xfs_error_level)
861 			xfs_stack_trace();
862 	} else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
863 		if (logerror) {
864 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR,
865 		"Log I/O Error Detected.  Shutting down filesystem");
866 		} else if (flags & SHUTDOWN_DEVICE_REQ) {
867 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
868 		"All device paths lost.  Shutting down filesystem");
869 		} else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
870 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
871 		"I/O Error Detected. Shutting down filesystem");
872 		}
873 	}
874 	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
875 		xfs_alert(mp,
876 	"Please umount the filesystem and rectify the problem(s)");
877 	}
878 }
879