xref: /openbmc/linux/fs/xfs/xfs_fsops.c (revision 930beb5a)
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_inode.h"
28 #include "xfs_trans.h"
29 #include "xfs_inode_item.h"
30 #include "xfs_error.h"
31 #include "xfs_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_alloc.h"
34 #include "xfs_ialloc.h"
35 #include "xfs_fsops.h"
36 #include "xfs_itable.h"
37 #include "xfs_trans_space.h"
38 #include "xfs_rtalloc.h"
39 #include "xfs_trace.h"
40 #include "xfs_log.h"
41 #include "xfs_dinode.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 =
78 			(xfs_sb_version_hasattr(&mp->m_sb) ?
79 				XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
80 			(xfs_sb_version_hasnlink(&mp->m_sb) ?
81 				XFS_FSOP_GEOM_FLAGS_NLINK : 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_hasshared(&mp->m_sb) ?
89 				XFS_FSOP_GEOM_FLAGS_SHARED : 0) |
90 			(xfs_sb_version_hasextflgbit(&mp->m_sb) ?
91 				XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
92 			(xfs_sb_version_hasdirv2(&mp->m_sb) ?
93 				XFS_FSOP_GEOM_FLAGS_DIRV2 : 0) |
94 			(xfs_sb_version_hassector(&mp->m_sb) ?
95 				XFS_FSOP_GEOM_FLAGS_SECTOR : 0) |
96 			(xfs_sb_version_hasasciici(&mp->m_sb) ?
97 				XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
98 			(xfs_sb_version_haslazysbcount(&mp->m_sb) ?
99 				XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
100 			(xfs_sb_version_hasattr2(&mp->m_sb) ?
101 				XFS_FSOP_GEOM_FLAGS_ATTR2 : 0) |
102 			(xfs_sb_version_hasprojid32bit(&mp->m_sb) ?
103 				XFS_FSOP_GEOM_FLAGS_PROJID32 : 0) |
104 			(xfs_sb_version_hascrc(&mp->m_sb) ?
105 				XFS_FSOP_GEOM_FLAGS_V5SB : 0) |
106 			(xfs_sb_version_hasftype(&mp->m_sb) ?
107 				XFS_FSOP_GEOM_FLAGS_FTYPE : 0);
108 		geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
109 				mp->m_sb.sb_logsectsize : BBSIZE;
110 		geo->rtsectsize = mp->m_sb.sb_blocksize;
111 		geo->dirblocksize = mp->m_dirblksize;
112 	}
113 	if (new_version >= 4) {
114 		geo->flags |=
115 			(xfs_sb_version_haslogv2(&mp->m_sb) ?
116 				XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
117 		geo->logsunit = mp->m_sb.sb_logsunit;
118 	}
119 	return 0;
120 }
121 
122 static struct xfs_buf *
123 xfs_growfs_get_hdr_buf(
124 	struct xfs_mount	*mp,
125 	xfs_daddr_t		blkno,
126 	size_t			numblks,
127 	int			flags,
128 	const struct xfs_buf_ops *ops)
129 {
130 	struct xfs_buf		*bp;
131 
132 	bp = xfs_buf_get_uncached(mp->m_ddev_targp, numblks, flags);
133 	if (!bp)
134 		return NULL;
135 
136 	xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
137 	bp->b_bn = blkno;
138 	bp->b_maps[0].bm_bn = blkno;
139 	bp->b_ops = ops;
140 
141 	return bp;
142 }
143 
144 static int
145 xfs_growfs_data_private(
146 	xfs_mount_t		*mp,		/* mount point for filesystem */
147 	xfs_growfs_data_t	*in)		/* growfs data input struct */
148 {
149 	xfs_agf_t		*agf;
150 	struct xfs_agfl		*agfl;
151 	xfs_agi_t		*agi;
152 	xfs_agnumber_t		agno;
153 	xfs_extlen_t		agsize;
154 	xfs_extlen_t		tmpsize;
155 	xfs_alloc_rec_t		*arec;
156 	xfs_buf_t		*bp;
157 	int			bucket;
158 	int			dpct;
159 	int			error, saved_error = 0;
160 	xfs_agnumber_t		nagcount;
161 	xfs_agnumber_t		nagimax = 0;
162 	xfs_rfsblock_t		nb, nb_mod;
163 	xfs_rfsblock_t		new;
164 	xfs_rfsblock_t		nfree;
165 	xfs_agnumber_t		oagcount;
166 	int			pct;
167 	xfs_trans_t		*tp;
168 
169 	nb = in->newblocks;
170 	pct = in->imaxpct;
171 	if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
172 		return XFS_ERROR(EINVAL);
173 	if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
174 		return error;
175 	dpct = pct - mp->m_sb.sb_imax_pct;
176 	bp = xfs_buf_read_uncached(mp->m_ddev_targp,
177 				XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
178 				XFS_FSS_TO_BB(mp, 1), 0, NULL);
179 	if (!bp)
180 		return EIO;
181 	if (bp->b_error) {
182 		error = bp->b_error;
183 		xfs_buf_relse(bp);
184 		return error;
185 	}
186 	xfs_buf_relse(bp);
187 
188 	new = nb;	/* use new as a temporary here */
189 	nb_mod = do_div(new, mp->m_sb.sb_agblocks);
190 	nagcount = new + (nb_mod != 0);
191 	if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
192 		nagcount--;
193 		nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
194 		if (nb < mp->m_sb.sb_dblocks)
195 			return XFS_ERROR(EINVAL);
196 	}
197 	new = nb - mp->m_sb.sb_dblocks;
198 	oagcount = mp->m_sb.sb_agcount;
199 
200 	/* allocate the new per-ag structures */
201 	if (nagcount > oagcount) {
202 		error = xfs_initialize_perag(mp, nagcount, &nagimax);
203 		if (error)
204 			return error;
205 	}
206 
207 	tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
208 	tp->t_flags |= XFS_TRANS_RESERVE;
209 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_growdata,
210 				  XFS_GROWFS_SPACE_RES(mp), 0);
211 	if (error) {
212 		xfs_trans_cancel(tp, 0);
213 		return error;
214 	}
215 
216 	/*
217 	 * Write new AG headers to disk. Non-transactional, but written
218 	 * synchronously so they are completed prior to the growfs transaction
219 	 * being logged.
220 	 */
221 	nfree = 0;
222 	for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
223 		__be32	*agfl_bno;
224 
225 		/*
226 		 * AG freespace header block
227 		 */
228 		bp = xfs_growfs_get_hdr_buf(mp,
229 				XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
230 				XFS_FSS_TO_BB(mp, 1), 0,
231 				&xfs_agf_buf_ops);
232 		if (!bp) {
233 			error = ENOMEM;
234 			goto error0;
235 		}
236 
237 		agf = XFS_BUF_TO_AGF(bp);
238 		agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
239 		agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
240 		agf->agf_seqno = cpu_to_be32(agno);
241 		if (agno == nagcount - 1)
242 			agsize =
243 				nb -
244 				(agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
245 		else
246 			agsize = mp->m_sb.sb_agblocks;
247 		agf->agf_length = cpu_to_be32(agsize);
248 		agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
249 		agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
250 		agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
251 		agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
252 		agf->agf_flfirst = 0;
253 		agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
254 		agf->agf_flcount = 0;
255 		tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
256 		agf->agf_freeblks = cpu_to_be32(tmpsize);
257 		agf->agf_longest = cpu_to_be32(tmpsize);
258 		if (xfs_sb_version_hascrc(&mp->m_sb))
259 			uuid_copy(&agf->agf_uuid, &mp->m_sb.sb_uuid);
260 
261 		error = xfs_bwrite(bp);
262 		xfs_buf_relse(bp);
263 		if (error)
264 			goto error0;
265 
266 		/*
267 		 * AG freelist header block
268 		 */
269 		bp = xfs_growfs_get_hdr_buf(mp,
270 				XFS_AG_DADDR(mp, agno, XFS_AGFL_DADDR(mp)),
271 				XFS_FSS_TO_BB(mp, 1), 0,
272 				&xfs_agfl_buf_ops);
273 		if (!bp) {
274 			error = ENOMEM;
275 			goto error0;
276 		}
277 
278 		agfl = XFS_BUF_TO_AGFL(bp);
279 		if (xfs_sb_version_hascrc(&mp->m_sb)) {
280 			agfl->agfl_magicnum = cpu_to_be32(XFS_AGFL_MAGIC);
281 			agfl->agfl_seqno = cpu_to_be32(agno);
282 			uuid_copy(&agfl->agfl_uuid, &mp->m_sb.sb_uuid);
283 		}
284 
285 		agfl_bno = XFS_BUF_TO_AGFL_BNO(mp, bp);
286 		for (bucket = 0; bucket < XFS_AGFL_SIZE(mp); bucket++)
287 			agfl_bno[bucket] = cpu_to_be32(NULLAGBLOCK);
288 
289 		error = xfs_bwrite(bp);
290 		xfs_buf_relse(bp);
291 		if (error)
292 			goto error0;
293 
294 		/*
295 		 * AG inode header block
296 		 */
297 		bp = xfs_growfs_get_hdr_buf(mp,
298 				XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
299 				XFS_FSS_TO_BB(mp, 1), 0,
300 				&xfs_agi_buf_ops);
301 		if (!bp) {
302 			error = ENOMEM;
303 			goto error0;
304 		}
305 
306 		agi = XFS_BUF_TO_AGI(bp);
307 		agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
308 		agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
309 		agi->agi_seqno = cpu_to_be32(agno);
310 		agi->agi_length = cpu_to_be32(agsize);
311 		agi->agi_count = 0;
312 		agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
313 		agi->agi_level = cpu_to_be32(1);
314 		agi->agi_freecount = 0;
315 		agi->agi_newino = cpu_to_be32(NULLAGINO);
316 		agi->agi_dirino = cpu_to_be32(NULLAGINO);
317 		if (xfs_sb_version_hascrc(&mp->m_sb))
318 			uuid_copy(&agi->agi_uuid, &mp->m_sb.sb_uuid);
319 		for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
320 			agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
321 
322 		error = xfs_bwrite(bp);
323 		xfs_buf_relse(bp);
324 		if (error)
325 			goto error0;
326 
327 		/*
328 		 * BNO btree root block
329 		 */
330 		bp = xfs_growfs_get_hdr_buf(mp,
331 				XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
332 				BTOBB(mp->m_sb.sb_blocksize), 0,
333 				&xfs_allocbt_buf_ops);
334 
335 		if (!bp) {
336 			error = ENOMEM;
337 			goto error0;
338 		}
339 
340 		if (xfs_sb_version_hascrc(&mp->m_sb))
341 			xfs_btree_init_block(mp, bp, XFS_ABTB_CRC_MAGIC, 0, 1,
342 						agno, XFS_BTREE_CRC_BLOCKS);
343 		else
344 			xfs_btree_init_block(mp, bp, XFS_ABTB_MAGIC, 0, 1,
345 						agno, 0);
346 
347 		arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
348 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
349 		arec->ar_blockcount = cpu_to_be32(
350 			agsize - be32_to_cpu(arec->ar_startblock));
351 
352 		error = xfs_bwrite(bp);
353 		xfs_buf_relse(bp);
354 		if (error)
355 			goto error0;
356 
357 		/*
358 		 * CNT btree root block
359 		 */
360 		bp = xfs_growfs_get_hdr_buf(mp,
361 				XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
362 				BTOBB(mp->m_sb.sb_blocksize), 0,
363 				&xfs_allocbt_buf_ops);
364 		if (!bp) {
365 			error = ENOMEM;
366 			goto error0;
367 		}
368 
369 		if (xfs_sb_version_hascrc(&mp->m_sb))
370 			xfs_btree_init_block(mp, bp, XFS_ABTC_CRC_MAGIC, 0, 1,
371 						agno, XFS_BTREE_CRC_BLOCKS);
372 		else
373 			xfs_btree_init_block(mp, bp, XFS_ABTC_MAGIC, 0, 1,
374 						agno, 0);
375 
376 		arec = XFS_ALLOC_REC_ADDR(mp, XFS_BUF_TO_BLOCK(bp), 1);
377 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
378 		arec->ar_blockcount = cpu_to_be32(
379 			agsize - be32_to_cpu(arec->ar_startblock));
380 		nfree += be32_to_cpu(arec->ar_blockcount);
381 
382 		error = xfs_bwrite(bp);
383 		xfs_buf_relse(bp);
384 		if (error)
385 			goto error0;
386 
387 		/*
388 		 * INO btree root block
389 		 */
390 		bp = xfs_growfs_get_hdr_buf(mp,
391 				XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
392 				BTOBB(mp->m_sb.sb_blocksize), 0,
393 				&xfs_inobt_buf_ops);
394 		if (!bp) {
395 			error = ENOMEM;
396 			goto error0;
397 		}
398 
399 		if (xfs_sb_version_hascrc(&mp->m_sb))
400 			xfs_btree_init_block(mp, bp, XFS_IBT_CRC_MAGIC, 0, 0,
401 						agno, XFS_BTREE_CRC_BLOCKS);
402 		else
403 			xfs_btree_init_block(mp, bp, XFS_IBT_MAGIC, 0, 0,
404 						agno, 0);
405 
406 		error = xfs_bwrite(bp);
407 		xfs_buf_relse(bp);
408 		if (error)
409 			goto error0;
410 	}
411 	xfs_trans_agblocks_delta(tp, nfree);
412 	/*
413 	 * There are new blocks in the old last a.g.
414 	 */
415 	if (new) {
416 		/*
417 		 * Change the agi length.
418 		 */
419 		error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
420 		if (error) {
421 			goto error0;
422 		}
423 		ASSERT(bp);
424 		agi = XFS_BUF_TO_AGI(bp);
425 		be32_add_cpu(&agi->agi_length, new);
426 		ASSERT(nagcount == oagcount ||
427 		       be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
428 		xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
429 		/*
430 		 * Change agf length.
431 		 */
432 		error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
433 		if (error) {
434 			goto error0;
435 		}
436 		ASSERT(bp);
437 		agf = XFS_BUF_TO_AGF(bp);
438 		be32_add_cpu(&agf->agf_length, new);
439 		ASSERT(be32_to_cpu(agf->agf_length) ==
440 		       be32_to_cpu(agi->agi_length));
441 
442 		xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
443 		/*
444 		 * Free the new space.
445 		 */
446 		error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
447 			be32_to_cpu(agf->agf_length) - new), new);
448 		if (error) {
449 			goto error0;
450 		}
451 	}
452 
453 	/*
454 	 * Update changed superblock fields transactionally. These are not
455 	 * seen by the rest of the world until the transaction commit applies
456 	 * them atomically to the superblock.
457 	 */
458 	if (nagcount > oagcount)
459 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
460 	if (nb > mp->m_sb.sb_dblocks)
461 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
462 				 nb - mp->m_sb.sb_dblocks);
463 	if (nfree)
464 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
465 	if (dpct)
466 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
467 	error = xfs_trans_commit(tp, 0);
468 	if (error)
469 		return error;
470 
471 	/* New allocation groups fully initialized, so update mount struct */
472 	if (nagimax)
473 		mp->m_maxagi = nagimax;
474 	if (mp->m_sb.sb_imax_pct) {
475 		__uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
476 		do_div(icount, 100);
477 		mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
478 	} else
479 		mp->m_maxicount = 0;
480 	xfs_set_low_space_thresholds(mp);
481 
482 	/* update secondary superblocks. */
483 	for (agno = 1; agno < nagcount; agno++) {
484 		error = 0;
485 		/*
486 		 * new secondary superblocks need to be zeroed, not read from
487 		 * disk as the contents of the new area we are growing into is
488 		 * completely unknown.
489 		 */
490 		if (agno < oagcount) {
491 			error = xfs_trans_read_buf(mp, NULL, mp->m_ddev_targp,
492 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
493 				  XFS_FSS_TO_BB(mp, 1), 0, &bp,
494 				  &xfs_sb_buf_ops);
495 		} else {
496 			bp = xfs_trans_get_buf(NULL, mp->m_ddev_targp,
497 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
498 				  XFS_FSS_TO_BB(mp, 1), 0);
499 			if (bp) {
500 				bp->b_ops = &xfs_sb_buf_ops;
501 				xfs_buf_zero(bp, 0, BBTOB(bp->b_length));
502 			} else
503 				error = ENOMEM;
504 		}
505 
506 		/*
507 		 * If we get an error reading or writing alternate superblocks,
508 		 * continue.  xfs_repair chooses the "best" superblock based
509 		 * on most matches; if we break early, we'll leave more
510 		 * superblocks un-updated than updated, and xfs_repair may
511 		 * pick them over the properly-updated primary.
512 		 */
513 		if (error) {
514 			xfs_warn(mp,
515 		"error %d reading secondary superblock for ag %d",
516 				error, agno);
517 			saved_error = error;
518 			continue;
519 		}
520 		xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
521 
522 		error = xfs_bwrite(bp);
523 		xfs_buf_relse(bp);
524 		if (error) {
525 			xfs_warn(mp,
526 		"write error %d updating secondary superblock for ag %d",
527 				error, agno);
528 			saved_error = error;
529 			continue;
530 		}
531 	}
532 	return saved_error ? saved_error : error;
533 
534  error0:
535 	xfs_trans_cancel(tp, XFS_TRANS_ABORT);
536 	return error;
537 }
538 
539 static int
540 xfs_growfs_log_private(
541 	xfs_mount_t		*mp,	/* mount point for filesystem */
542 	xfs_growfs_log_t	*in)	/* growfs log input struct */
543 {
544 	xfs_extlen_t		nb;
545 
546 	nb = in->newblocks;
547 	if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
548 		return XFS_ERROR(EINVAL);
549 	if (nb == mp->m_sb.sb_logblocks &&
550 	    in->isint == (mp->m_sb.sb_logstart != 0))
551 		return XFS_ERROR(EINVAL);
552 	/*
553 	 * Moving the log is hard, need new interfaces to sync
554 	 * the log first, hold off all activity while moving it.
555 	 * Can have shorter or longer log in the same space,
556 	 * or transform internal to external log or vice versa.
557 	 */
558 	return XFS_ERROR(ENOSYS);
559 }
560 
561 /*
562  * protected versions of growfs function acquire and release locks on the mount
563  * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
564  * XFS_IOC_FSGROWFSRT
565  */
566 
567 
568 int
569 xfs_growfs_data(
570 	xfs_mount_t		*mp,
571 	xfs_growfs_data_t	*in)
572 {
573 	int error;
574 
575 	if (!capable(CAP_SYS_ADMIN))
576 		return XFS_ERROR(EPERM);
577 	if (!mutex_trylock(&mp->m_growlock))
578 		return XFS_ERROR(EWOULDBLOCK);
579 	error = xfs_growfs_data_private(mp, in);
580 	mutex_unlock(&mp->m_growlock);
581 	return error;
582 }
583 
584 int
585 xfs_growfs_log(
586 	xfs_mount_t		*mp,
587 	xfs_growfs_log_t	*in)
588 {
589 	int error;
590 
591 	if (!capable(CAP_SYS_ADMIN))
592 		return XFS_ERROR(EPERM);
593 	if (!mutex_trylock(&mp->m_growlock))
594 		return XFS_ERROR(EWOULDBLOCK);
595 	error = xfs_growfs_log_private(mp, in);
596 	mutex_unlock(&mp->m_growlock);
597 	return error;
598 }
599 
600 /*
601  * exported through ioctl XFS_IOC_FSCOUNTS
602  */
603 
604 int
605 xfs_fs_counts(
606 	xfs_mount_t		*mp,
607 	xfs_fsop_counts_t	*cnt)
608 {
609 	xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
610 	spin_lock(&mp->m_sb_lock);
611 	cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
612 	cnt->freertx = mp->m_sb.sb_frextents;
613 	cnt->freeino = mp->m_sb.sb_ifree;
614 	cnt->allocino = mp->m_sb.sb_icount;
615 	spin_unlock(&mp->m_sb_lock);
616 	return 0;
617 }
618 
619 /*
620  * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
621  *
622  * xfs_reserve_blocks is called to set m_resblks
623  * in the in-core mount table. The number of unused reserved blocks
624  * is kept in m_resblks_avail.
625  *
626  * Reserve the requested number of blocks if available. Otherwise return
627  * as many as possible to satisfy the request. The actual number
628  * reserved are returned in outval
629  *
630  * A null inval pointer indicates that only the current reserved blocks
631  * available  should  be returned no settings are changed.
632  */
633 
634 int
635 xfs_reserve_blocks(
636 	xfs_mount_t             *mp,
637 	__uint64_t              *inval,
638 	xfs_fsop_resblks_t      *outval)
639 {
640 	__int64_t		lcounter, delta, fdblks_delta;
641 	__uint64_t		request;
642 
643 	/* If inval is null, report current values and return */
644 	if (inval == (__uint64_t *)NULL) {
645 		if (!outval)
646 			return EINVAL;
647 		outval->resblks = mp->m_resblks;
648 		outval->resblks_avail = mp->m_resblks_avail;
649 		return 0;
650 	}
651 
652 	request = *inval;
653 
654 	/*
655 	 * With per-cpu counters, this becomes an interesting
656 	 * problem. we needto work out if we are freeing or allocation
657 	 * blocks first, then we can do the modification as necessary.
658 	 *
659 	 * We do this under the m_sb_lock so that if we are near
660 	 * ENOSPC, we will hold out any changes while we work out
661 	 * what to do. This means that the amount of free space can
662 	 * change while we do this, so we need to retry if we end up
663 	 * trying to reserve more space than is available.
664 	 *
665 	 * We also use the xfs_mod_incore_sb() interface so that we
666 	 * don't have to care about whether per cpu counter are
667 	 * enabled, disabled or even compiled in....
668 	 */
669 retry:
670 	spin_lock(&mp->m_sb_lock);
671 	xfs_icsb_sync_counters_locked(mp, 0);
672 
673 	/*
674 	 * If our previous reservation was larger than the current value,
675 	 * then move any unused blocks back to the free pool.
676 	 */
677 	fdblks_delta = 0;
678 	if (mp->m_resblks > request) {
679 		lcounter = mp->m_resblks_avail - request;
680 		if (lcounter  > 0) {		/* release unused blocks */
681 			fdblks_delta = lcounter;
682 			mp->m_resblks_avail -= lcounter;
683 		}
684 		mp->m_resblks = request;
685 	} else {
686 		__int64_t	free;
687 
688 		free =  mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
689 		if (!free)
690 			goto out; /* ENOSPC and fdblks_delta = 0 */
691 
692 		delta = request - mp->m_resblks;
693 		lcounter = free - delta;
694 		if (lcounter < 0) {
695 			/* We can't satisfy the request, just get what we can */
696 			mp->m_resblks += free;
697 			mp->m_resblks_avail += free;
698 			fdblks_delta = -free;
699 		} else {
700 			fdblks_delta = -delta;
701 			mp->m_resblks = request;
702 			mp->m_resblks_avail += delta;
703 		}
704 	}
705 out:
706 	if (outval) {
707 		outval->resblks = mp->m_resblks;
708 		outval->resblks_avail = mp->m_resblks_avail;
709 	}
710 	spin_unlock(&mp->m_sb_lock);
711 
712 	if (fdblks_delta) {
713 		/*
714 		 * If we are putting blocks back here, m_resblks_avail is
715 		 * already at its max so this will put it in the free pool.
716 		 *
717 		 * If we need space, we'll either succeed in getting it
718 		 * from the free block count or we'll get an enospc. If
719 		 * we get a ENOSPC, it means things changed while we were
720 		 * calculating fdblks_delta and so we should try again to
721 		 * see if there is anything left to reserve.
722 		 *
723 		 * Don't set the reserved flag here - we don't want to reserve
724 		 * the extra reserve blocks from the reserve.....
725 		 */
726 		int error;
727 		error = xfs_icsb_modify_counters(mp, XFS_SBS_FDBLOCKS,
728 						 fdblks_delta, 0);
729 		if (error == ENOSPC)
730 			goto retry;
731 	}
732 	return 0;
733 }
734 
735 /*
736  * Dump a transaction into the log that contains no real change. This is needed
737  * to be able to make the log dirty or stamp the current tail LSN into the log
738  * during the covering operation.
739  *
740  * We cannot use an inode here for this - that will push dirty state back up
741  * into the VFS and then periodic inode flushing will prevent log covering from
742  * making progress. Hence we log a field in the superblock instead and use a
743  * synchronous transaction to ensure the superblock is immediately unpinned
744  * and can be written back.
745  */
746 int
747 xfs_fs_log_dummy(
748 	xfs_mount_t	*mp)
749 {
750 	xfs_trans_t	*tp;
751 	int		error;
752 
753 	tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
754 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_sb, 0, 0);
755 	if (error) {
756 		xfs_trans_cancel(tp, 0);
757 		return error;
758 	}
759 
760 	/* log the UUID because it is an unchanging field */
761 	xfs_mod_sb(tp, XFS_SB_UUID);
762 	xfs_trans_set_sync(tp);
763 	return xfs_trans_commit(tp, 0);
764 }
765 
766 int
767 xfs_fs_goingdown(
768 	xfs_mount_t	*mp,
769 	__uint32_t	inflags)
770 {
771 	switch (inflags) {
772 	case XFS_FSOP_GOING_FLAGS_DEFAULT: {
773 		struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
774 
775 		if (sb && !IS_ERR(sb)) {
776 			xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
777 			thaw_bdev(sb->s_bdev, sb);
778 		}
779 
780 		break;
781 	}
782 	case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
783 		xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
784 		break;
785 	case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
786 		xfs_force_shutdown(mp,
787 				SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
788 		break;
789 	default:
790 		return XFS_ERROR(EINVAL);
791 	}
792 
793 	return 0;
794 }
795 
796 /*
797  * Force a shutdown of the filesystem instantly while keeping the filesystem
798  * consistent. We don't do an unmount here; just shutdown the shop, make sure
799  * that absolutely nothing persistent happens to this filesystem after this
800  * point.
801  */
802 void
803 xfs_do_force_shutdown(
804 	xfs_mount_t	*mp,
805 	int		flags,
806 	char		*fname,
807 	int		lnnum)
808 {
809 	int		logerror;
810 
811 	logerror = flags & SHUTDOWN_LOG_IO_ERROR;
812 
813 	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
814 		xfs_notice(mp,
815 	"%s(0x%x) called from line %d of file %s.  Return address = 0x%p",
816 			__func__, flags, lnnum, fname, __return_address);
817 	}
818 	/*
819 	 * No need to duplicate efforts.
820 	 */
821 	if (XFS_FORCED_SHUTDOWN(mp) && !logerror)
822 		return;
823 
824 	/*
825 	 * This flags XFS_MOUNT_FS_SHUTDOWN, makes sure that we don't
826 	 * queue up anybody new on the log reservations, and wakes up
827 	 * everybody who's sleeping on log reservations to tell them
828 	 * the bad news.
829 	 */
830 	if (xfs_log_force_umount(mp, logerror))
831 		return;
832 
833 	if (flags & SHUTDOWN_CORRUPT_INCORE) {
834 		xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_CORRUPT,
835     "Corruption of in-memory data detected.  Shutting down filesystem");
836 		if (XFS_ERRLEVEL_HIGH <= xfs_error_level)
837 			xfs_stack_trace();
838 	} else if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
839 		if (logerror) {
840 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_LOGERROR,
841 		"Log I/O Error Detected.  Shutting down filesystem");
842 		} else if (flags & SHUTDOWN_DEVICE_REQ) {
843 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
844 		"All device paths lost.  Shutting down filesystem");
845 		} else if (!(flags & SHUTDOWN_REMOTE_REQ)) {
846 			xfs_alert_tag(mp, XFS_PTAG_SHUTDOWN_IOERROR,
847 		"I/O Error Detected. Shutting down filesystem");
848 		}
849 	}
850 	if (!(flags & SHUTDOWN_FORCE_UMOUNT)) {
851 		xfs_alert(mp,
852 	"Please umount the filesystem and rectify the problem(s)");
853 	}
854 }
855