xref: /linux/fs/xfs/xfs_fsops.c (revision 2b8232ce512105e28453f301d1510de8363bccd1)
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_types.h"
21 #include "xfs_bit.h"
22 #include "xfs_inum.h"
23 #include "xfs_log.h"
24 #include "xfs_trans.h"
25 #include "xfs_sb.h"
26 #include "xfs_ag.h"
27 #include "xfs_dir2.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_mount.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_alloc_btree.h"
32 #include "xfs_ialloc_btree.h"
33 #include "xfs_dir2_sf.h"
34 #include "xfs_attr_sf.h"
35 #include "xfs_dinode.h"
36 #include "xfs_inode.h"
37 #include "xfs_inode_item.h"
38 #include "xfs_btree.h"
39 #include "xfs_error.h"
40 #include "xfs_alloc.h"
41 #include "xfs_ialloc.h"
42 #include "xfs_fsops.h"
43 #include "xfs_itable.h"
44 #include "xfs_trans_space.h"
45 #include "xfs_rtalloc.h"
46 #include "xfs_rw.h"
47 #include "xfs_filestream.h"
48 
49 /*
50  * File system operations
51  */
52 
53 int
54 xfs_fs_geometry(
55 	xfs_mount_t		*mp,
56 	xfs_fsop_geom_t		*geo,
57 	int			new_version)
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 =
80 			(XFS_SB_VERSION_HASATTR(&mp->m_sb) ?
81 				XFS_FSOP_GEOM_FLAGS_ATTR : 0) |
82 			(XFS_SB_VERSION_HASNLINK(&mp->m_sb) ?
83 				XFS_FSOP_GEOM_FLAGS_NLINK : 0) |
84 			(XFS_SB_VERSION_HASQUOTA(&mp->m_sb) ?
85 				XFS_FSOP_GEOM_FLAGS_QUOTA : 0) |
86 			(XFS_SB_VERSION_HASALIGN(&mp->m_sb) ?
87 				XFS_FSOP_GEOM_FLAGS_IALIGN : 0) |
88 			(XFS_SB_VERSION_HASDALIGN(&mp->m_sb) ?
89 				XFS_FSOP_GEOM_FLAGS_DALIGN : 0) |
90 			(XFS_SB_VERSION_HASSHARED(&mp->m_sb) ?
91 				XFS_FSOP_GEOM_FLAGS_SHARED : 0) |
92 			(XFS_SB_VERSION_HASEXTFLGBIT(&mp->m_sb) ?
93 				XFS_FSOP_GEOM_FLAGS_EXTFLG : 0) |
94 			(XFS_SB_VERSION_HASDIRV2(&mp->m_sb) ?
95 				XFS_FSOP_GEOM_FLAGS_DIRV2 : 0) |
96 			(XFS_SB_VERSION_HASSECTOR(&mp->m_sb) ?
97 				XFS_FSOP_GEOM_FLAGS_SECTOR : 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 		geo->logsectsize = XFS_SB_VERSION_HASSECTOR(&mp->m_sb) ?
103 				mp->m_sb.sb_logsectsize : BBSIZE;
104 		geo->rtsectsize = mp->m_sb.sb_blocksize;
105 		geo->dirblocksize = mp->m_dirblksize;
106 	}
107 	if (new_version >= 4) {
108 		geo->flags |=
109 			(XFS_SB_VERSION_HASLOGV2(&mp->m_sb) ?
110 				XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
111 		geo->logsunit = mp->m_sb.sb_logsunit;
112 	}
113 	return 0;
114 }
115 
116 static int
117 xfs_growfs_data_private(
118 	xfs_mount_t		*mp,		/* mount point for filesystem */
119 	xfs_growfs_data_t	*in)		/* growfs data input struct */
120 {
121 	xfs_agf_t		*agf;
122 	xfs_agi_t		*agi;
123 	xfs_agnumber_t		agno;
124 	xfs_extlen_t		agsize;
125 	xfs_extlen_t		tmpsize;
126 	xfs_alloc_rec_t		*arec;
127 	xfs_btree_sblock_t	*block;
128 	xfs_buf_t		*bp;
129 	int			bucket;
130 	int			dpct;
131 	int			error;
132 	xfs_agnumber_t		nagcount;
133 	xfs_agnumber_t		nagimax = 0;
134 	xfs_rfsblock_t		nb, nb_mod;
135 	xfs_rfsblock_t		new;
136 	xfs_rfsblock_t		nfree;
137 	xfs_agnumber_t		oagcount;
138 	int			pct;
139 	xfs_sb_t		*sbp;
140 	xfs_trans_t		*tp;
141 
142 	nb = in->newblocks;
143 	pct = in->imaxpct;
144 	if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
145 		return XFS_ERROR(EINVAL);
146 	if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
147 		return error;
148 	dpct = pct - mp->m_sb.sb_imax_pct;
149 	error = xfs_read_buf(mp, mp->m_ddev_targp,
150 			XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
151 			XFS_FSS_TO_BB(mp, 1), 0, &bp);
152 	if (error)
153 		return error;
154 	ASSERT(bp);
155 	xfs_buf_relse(bp);
156 
157 	new = nb;	/* use new as a temporary here */
158 	nb_mod = do_div(new, mp->m_sb.sb_agblocks);
159 	nagcount = new + (nb_mod != 0);
160 	if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
161 		nagcount--;
162 		nb = nagcount * mp->m_sb.sb_agblocks;
163 		if (nb < mp->m_sb.sb_dblocks)
164 			return XFS_ERROR(EINVAL);
165 	}
166 	new = nb - mp->m_sb.sb_dblocks;
167 	oagcount = mp->m_sb.sb_agcount;
168 	if (nagcount > oagcount) {
169 		xfs_filestream_flush(mp);
170 		down_write(&mp->m_peraglock);
171 		mp->m_perag = kmem_realloc(mp->m_perag,
172 			sizeof(xfs_perag_t) * nagcount,
173 			sizeof(xfs_perag_t) * oagcount,
174 			KM_SLEEP);
175 		memset(&mp->m_perag[oagcount], 0,
176 			(nagcount - oagcount) * sizeof(xfs_perag_t));
177 		mp->m_flags |= XFS_MOUNT_32BITINODES;
178 		nagimax = xfs_initialize_perag(XFS_MTOVFS(mp), mp, nagcount);
179 		up_write(&mp->m_peraglock);
180 	}
181 	tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
182 	tp->t_flags |= XFS_TRANS_RESERVE;
183 	if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
184 			XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
185 		xfs_trans_cancel(tp, 0);
186 		return error;
187 	}
188 
189 	nfree = 0;
190 	for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
191 		/*
192 		 * AG freelist header block
193 		 */
194 		bp = xfs_buf_get(mp->m_ddev_targp,
195 				  XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
196 				  XFS_FSS_TO_BB(mp, 1), 0);
197 		agf = XFS_BUF_TO_AGF(bp);
198 		memset(agf, 0, mp->m_sb.sb_sectsize);
199 		agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
200 		agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
201 		agf->agf_seqno = cpu_to_be32(agno);
202 		if (agno == nagcount - 1)
203 			agsize =
204 				nb -
205 				(agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
206 		else
207 			agsize = mp->m_sb.sb_agblocks;
208 		agf->agf_length = cpu_to_be32(agsize);
209 		agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
210 		agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
211 		agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
212 		agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
213 		agf->agf_flfirst = 0;
214 		agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
215 		agf->agf_flcount = 0;
216 		tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
217 		agf->agf_freeblks = cpu_to_be32(tmpsize);
218 		agf->agf_longest = cpu_to_be32(tmpsize);
219 		error = xfs_bwrite(mp, bp);
220 		if (error) {
221 			goto error0;
222 		}
223 		/*
224 		 * AG inode header block
225 		 */
226 		bp = xfs_buf_get(mp->m_ddev_targp,
227 				  XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
228 				  XFS_FSS_TO_BB(mp, 1), 0);
229 		agi = XFS_BUF_TO_AGI(bp);
230 		memset(agi, 0, mp->m_sb.sb_sectsize);
231 		agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
232 		agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
233 		agi->agi_seqno = cpu_to_be32(agno);
234 		agi->agi_length = cpu_to_be32(agsize);
235 		agi->agi_count = 0;
236 		agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
237 		agi->agi_level = cpu_to_be32(1);
238 		agi->agi_freecount = 0;
239 		agi->agi_newino = cpu_to_be32(NULLAGINO);
240 		agi->agi_dirino = cpu_to_be32(NULLAGINO);
241 		for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
242 			agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
243 		error = xfs_bwrite(mp, bp);
244 		if (error) {
245 			goto error0;
246 		}
247 		/*
248 		 * BNO btree root block
249 		 */
250 		bp = xfs_buf_get(mp->m_ddev_targp,
251 			XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
252 			BTOBB(mp->m_sb.sb_blocksize), 0);
253 		block = XFS_BUF_TO_SBLOCK(bp);
254 		memset(block, 0, mp->m_sb.sb_blocksize);
255 		block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
256 		block->bb_level = 0;
257 		block->bb_numrecs = cpu_to_be16(1);
258 		block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
259 		block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
260 		arec = XFS_BTREE_REC_ADDR(xfs_alloc, block, 1);
261 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
262 		arec->ar_blockcount = cpu_to_be32(
263 			agsize - be32_to_cpu(arec->ar_startblock));
264 		error = xfs_bwrite(mp, bp);
265 		if (error) {
266 			goto error0;
267 		}
268 		/*
269 		 * CNT btree root block
270 		 */
271 		bp = xfs_buf_get(mp->m_ddev_targp,
272 			XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
273 			BTOBB(mp->m_sb.sb_blocksize), 0);
274 		block = XFS_BUF_TO_SBLOCK(bp);
275 		memset(block, 0, mp->m_sb.sb_blocksize);
276 		block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
277 		block->bb_level = 0;
278 		block->bb_numrecs = cpu_to_be16(1);
279 		block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
280 		block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
281 		arec = XFS_BTREE_REC_ADDR(xfs_alloc, block, 1);
282 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
283 		arec->ar_blockcount = cpu_to_be32(
284 			agsize - be32_to_cpu(arec->ar_startblock));
285 		nfree += be32_to_cpu(arec->ar_blockcount);
286 		error = xfs_bwrite(mp, bp);
287 		if (error) {
288 			goto error0;
289 		}
290 		/*
291 		 * INO btree root block
292 		 */
293 		bp = xfs_buf_get(mp->m_ddev_targp,
294 			XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
295 			BTOBB(mp->m_sb.sb_blocksize), 0);
296 		block = XFS_BUF_TO_SBLOCK(bp);
297 		memset(block, 0, mp->m_sb.sb_blocksize);
298 		block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
299 		block->bb_level = 0;
300 		block->bb_numrecs = 0;
301 		block->bb_leftsib = cpu_to_be32(NULLAGBLOCK);
302 		block->bb_rightsib = cpu_to_be32(NULLAGBLOCK);
303 		error = xfs_bwrite(mp, bp);
304 		if (error) {
305 			goto error0;
306 		}
307 	}
308 	xfs_trans_agblocks_delta(tp, nfree);
309 	/*
310 	 * There are new blocks in the old last a.g.
311 	 */
312 	if (new) {
313 		/*
314 		 * Change the agi length.
315 		 */
316 		error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
317 		if (error) {
318 			goto error0;
319 		}
320 		ASSERT(bp);
321 		agi = XFS_BUF_TO_AGI(bp);
322 		be32_add(&agi->agi_length, new);
323 		ASSERT(nagcount == oagcount ||
324 		       be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
325 		xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
326 		/*
327 		 * Change agf length.
328 		 */
329 		error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
330 		if (error) {
331 			goto error0;
332 		}
333 		ASSERT(bp);
334 		agf = XFS_BUF_TO_AGF(bp);
335 		be32_add(&agf->agf_length, new);
336 		ASSERT(be32_to_cpu(agf->agf_length) ==
337 		       be32_to_cpu(agi->agi_length));
338 		xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
339 		/*
340 		 * Free the new space.
341 		 */
342 		error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
343 			be32_to_cpu(agf->agf_length) - new), new);
344 		if (error) {
345 			goto error0;
346 		}
347 	}
348 	if (nagcount > oagcount)
349 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
350 	if (nb > mp->m_sb.sb_dblocks)
351 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
352 				 nb - mp->m_sb.sb_dblocks);
353 	if (nfree)
354 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
355 	if (dpct)
356 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
357 	error = xfs_trans_commit(tp, 0);
358 	if (error) {
359 		return error;
360 	}
361 	/* New allocation groups fully initialized, so update mount struct */
362 	if (nagimax)
363 		mp->m_maxagi = nagimax;
364 	if (mp->m_sb.sb_imax_pct) {
365 		__uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
366 		do_div(icount, 100);
367 		mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
368 	} else
369 		mp->m_maxicount = 0;
370 	for (agno = 1; agno < nagcount; agno++) {
371 		error = xfs_read_buf(mp, mp->m_ddev_targp,
372 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
373 				  XFS_FSS_TO_BB(mp, 1), 0, &bp);
374 		if (error) {
375 			xfs_fs_cmn_err(CE_WARN, mp,
376 			"error %d reading secondary superblock for ag %d",
377 				error, agno);
378 			break;
379 		}
380 		sbp = XFS_BUF_TO_SBP(bp);
381 		xfs_xlatesb(sbp, &mp->m_sb, -1, XFS_SB_ALL_BITS);
382 		/*
383 		 * If we get an error writing out the alternate superblocks,
384 		 * just issue a warning and continue.  The real work is
385 		 * already done and committed.
386 		 */
387 		if (!(error = xfs_bwrite(mp, bp))) {
388 			continue;
389 		} else {
390 			xfs_fs_cmn_err(CE_WARN, mp,
391 		"write error %d updating secondary superblock for ag %d",
392 				error, agno);
393 			break; /* no point in continuing */
394 		}
395 	}
396 	return 0;
397 
398  error0:
399 	xfs_trans_cancel(tp, XFS_TRANS_ABORT);
400 	return error;
401 }
402 
403 static int
404 xfs_growfs_log_private(
405 	xfs_mount_t		*mp,	/* mount point for filesystem */
406 	xfs_growfs_log_t	*in)	/* growfs log input struct */
407 {
408 	xfs_extlen_t		nb;
409 
410 	nb = in->newblocks;
411 	if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
412 		return XFS_ERROR(EINVAL);
413 	if (nb == mp->m_sb.sb_logblocks &&
414 	    in->isint == (mp->m_sb.sb_logstart != 0))
415 		return XFS_ERROR(EINVAL);
416 	/*
417 	 * Moving the log is hard, need new interfaces to sync
418 	 * the log first, hold off all activity while moving it.
419 	 * Can have shorter or longer log in the same space,
420 	 * or transform internal to external log or vice versa.
421 	 */
422 	return XFS_ERROR(ENOSYS);
423 }
424 
425 /*
426  * protected versions of growfs function acquire and release locks on the mount
427  * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
428  * XFS_IOC_FSGROWFSRT
429  */
430 
431 
432 int
433 xfs_growfs_data(
434 	xfs_mount_t		*mp,
435 	xfs_growfs_data_t	*in)
436 {
437 	int error;
438 	if (!cpsema(&mp->m_growlock))
439 		return XFS_ERROR(EWOULDBLOCK);
440 	error = xfs_growfs_data_private(mp, in);
441 	vsema(&mp->m_growlock);
442 	return error;
443 }
444 
445 int
446 xfs_growfs_log(
447 	xfs_mount_t		*mp,
448 	xfs_growfs_log_t	*in)
449 {
450 	int error;
451 	if (!cpsema(&mp->m_growlock))
452 		return XFS_ERROR(EWOULDBLOCK);
453 	error = xfs_growfs_log_private(mp, in);
454 	vsema(&mp->m_growlock);
455 	return error;
456 }
457 
458 /*
459  * exported through ioctl XFS_IOC_FSCOUNTS
460  */
461 
462 int
463 xfs_fs_counts(
464 	xfs_mount_t		*mp,
465 	xfs_fsop_counts_t	*cnt)
466 {
467 	unsigned long	s;
468 
469 	xfs_icsb_sync_counters_flags(mp, XFS_ICSB_LAZY_COUNT);
470 	s = XFS_SB_LOCK(mp);
471 	cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
472 	cnt->freertx = mp->m_sb.sb_frextents;
473 	cnt->freeino = mp->m_sb.sb_ifree;
474 	cnt->allocino = mp->m_sb.sb_icount;
475 	XFS_SB_UNLOCK(mp, s);
476 	return 0;
477 }
478 
479 /*
480  * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
481  *
482  * xfs_reserve_blocks is called to set m_resblks
483  * in the in-core mount table. The number of unused reserved blocks
484  * is kept in m_resblks_avail.
485  *
486  * Reserve the requested number of blocks if available. Otherwise return
487  * as many as possible to satisfy the request. The actual number
488  * reserved are returned in outval
489  *
490  * A null inval pointer indicates that only the current reserved blocks
491  * available  should  be returned no settings are changed.
492  */
493 
494 int
495 xfs_reserve_blocks(
496 	xfs_mount_t             *mp,
497 	__uint64_t              *inval,
498 	xfs_fsop_resblks_t      *outval)
499 {
500 	__int64_t		lcounter, delta, fdblks_delta;
501 	__uint64_t		request;
502 	unsigned long		s;
503 
504 	/* If inval is null, report current values and return */
505 	if (inval == (__uint64_t *)NULL) {
506 		if (!outval)
507 			return EINVAL;
508 		outval->resblks = mp->m_resblks;
509 		outval->resblks_avail = mp->m_resblks_avail;
510 		return 0;
511 	}
512 
513 	request = *inval;
514 
515 	/*
516 	 * With per-cpu counters, this becomes an interesting
517 	 * problem. we needto work out if we are freeing or allocation
518 	 * blocks first, then we can do the modification as necessary.
519 	 *
520 	 * We do this under the XFS_SB_LOCK so that if we are near
521 	 * ENOSPC, we will hold out any changes while we work out
522 	 * what to do. This means that the amount of free space can
523 	 * change while we do this, so we need to retry if we end up
524 	 * trying to reserve more space than is available.
525 	 *
526 	 * We also use the xfs_mod_incore_sb() interface so that we
527 	 * don't have to care about whether per cpu counter are
528 	 * enabled, disabled or even compiled in....
529 	 */
530 retry:
531 	s = XFS_SB_LOCK(mp);
532 	xfs_icsb_sync_counters_flags(mp, XFS_ICSB_SB_LOCKED);
533 
534 	/*
535 	 * If our previous reservation was larger than the current value,
536 	 * then move any unused blocks back to the free pool.
537 	 */
538 	fdblks_delta = 0;
539 	if (mp->m_resblks > request) {
540 		lcounter = mp->m_resblks_avail - request;
541 		if (lcounter  > 0) {		/* release unused blocks */
542 			fdblks_delta = lcounter;
543 			mp->m_resblks_avail -= lcounter;
544 		}
545 		mp->m_resblks = request;
546 	} else {
547 		__int64_t	free;
548 
549 		free =  mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
550 		if (!free)
551 			goto out; /* ENOSPC and fdblks_delta = 0 */
552 
553 		delta = request - mp->m_resblks;
554 		lcounter = free - delta;
555 		if (lcounter < 0) {
556 			/* We can't satisfy the request, just get what we can */
557 			mp->m_resblks += free;
558 			mp->m_resblks_avail += free;
559 			fdblks_delta = -free;
560 			mp->m_sb.sb_fdblocks = XFS_ALLOC_SET_ASIDE(mp);
561 		} else {
562 			fdblks_delta = -delta;
563 			mp->m_sb.sb_fdblocks =
564 				lcounter + XFS_ALLOC_SET_ASIDE(mp);
565 			mp->m_resblks = request;
566 			mp->m_resblks_avail += delta;
567 		}
568 	}
569 out:
570 	if (outval) {
571 		outval->resblks = mp->m_resblks;
572 		outval->resblks_avail = mp->m_resblks_avail;
573 	}
574 	XFS_SB_UNLOCK(mp, s);
575 
576 	if (fdblks_delta) {
577 		/*
578 		 * If we are putting blocks back here, m_resblks_avail is
579 		 * already at it's max so this will put it in the free pool.
580 		 *
581 		 * If we need space, we'll either succeed in getting it
582 		 * from the free block count or we'll get an enospc. If
583 		 * we get a ENOSPC, it means things changed while we were
584 		 * calculating fdblks_delta and so we should try again to
585 		 * see if there is anything left to reserve.
586 		 *
587 		 * Don't set the reserved flag here - we don't want to reserve
588 		 * the extra reserve blocks from the reserve.....
589 		 */
590 		int error;
591 		error = xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS, fdblks_delta, 0);
592 		if (error == ENOSPC)
593 			goto retry;
594 	}
595 
596 	return 0;
597 }
598 
599 void
600 xfs_fs_log_dummy(
601 	xfs_mount_t	*mp)
602 {
603 	xfs_trans_t	*tp;
604 	xfs_inode_t	*ip;
605 
606 	tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1);
607 	if (xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0)) {
608 		xfs_trans_cancel(tp, 0);
609 		return;
610 	}
611 
612 	ip = mp->m_rootip;
613 	xfs_ilock(ip, XFS_ILOCK_EXCL);
614 
615 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
616 	xfs_trans_ihold(tp, ip);
617 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
618 	xfs_trans_set_sync(tp);
619 	xfs_trans_commit(tp, 0);
620 
621 	xfs_iunlock(ip, XFS_ILOCK_EXCL);
622 }
623 
624 int
625 xfs_fs_goingdown(
626 	xfs_mount_t	*mp,
627 	__uint32_t	inflags)
628 {
629 	switch (inflags) {
630 	case XFS_FSOP_GOING_FLAGS_DEFAULT: {
631 		struct bhv_vfs *vfsp = XFS_MTOVFS(mp);
632 		struct super_block *sb = freeze_bdev(vfsp->vfs_super->s_bdev);
633 
634 		if (sb && !IS_ERR(sb)) {
635 			xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
636 			thaw_bdev(sb->s_bdev, sb);
637 		}
638 
639 		break;
640 	}
641 	case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
642 		xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
643 		break;
644 	case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
645 		xfs_force_shutdown(mp,
646 				SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
647 		break;
648 	default:
649 		return XFS_ERROR(EINVAL);
650 	}
651 
652 	return 0;
653 }
654