xref: /openbmc/linux/fs/xfs/xfs_fsops.c (revision 3252b11f)
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_hasasciici(&mp->m_sb) ?
99 				XFS_FSOP_GEOM_FLAGS_DIRV2CI : 0) |
100 			(xfs_sb_version_haslazysbcount(&mp->m_sb) ?
101 				XFS_FSOP_GEOM_FLAGS_LAZYSB : 0) |
102 			(xfs_sb_version_hasattr2(&mp->m_sb) ?
103 				XFS_FSOP_GEOM_FLAGS_ATTR2 : 0);
104 		geo->logsectsize = xfs_sb_version_hassector(&mp->m_sb) ?
105 				mp->m_sb.sb_logsectsize : BBSIZE;
106 		geo->rtsectsize = mp->m_sb.sb_blocksize;
107 		geo->dirblocksize = mp->m_dirblksize;
108 	}
109 	if (new_version >= 4) {
110 		geo->flags |=
111 			(xfs_sb_version_haslogv2(&mp->m_sb) ?
112 				XFS_FSOP_GEOM_FLAGS_LOGV2 : 0);
113 		geo->logsunit = mp->m_sb.sb_logsunit;
114 	}
115 	return 0;
116 }
117 
118 static int
119 xfs_growfs_data_private(
120 	xfs_mount_t		*mp,		/* mount point for filesystem */
121 	xfs_growfs_data_t	*in)		/* growfs data input struct */
122 {
123 	xfs_agf_t		*agf;
124 	xfs_agi_t		*agi;
125 	xfs_agnumber_t		agno;
126 	xfs_extlen_t		agsize;
127 	xfs_extlen_t		tmpsize;
128 	xfs_alloc_rec_t		*arec;
129 	struct xfs_btree_block	*block;
130 	xfs_buf_t		*bp;
131 	int			bucket;
132 	int			dpct;
133 	int			error;
134 	xfs_agnumber_t		nagcount;
135 	xfs_agnumber_t		nagimax = 0;
136 	xfs_rfsblock_t		nb, nb_mod;
137 	xfs_rfsblock_t		new;
138 	xfs_rfsblock_t		nfree;
139 	xfs_agnumber_t		oagcount;
140 	int			pct;
141 	xfs_trans_t		*tp;
142 
143 	nb = in->newblocks;
144 	pct = in->imaxpct;
145 	if (nb < mp->m_sb.sb_dblocks || pct < 0 || pct > 100)
146 		return XFS_ERROR(EINVAL);
147 	if ((error = xfs_sb_validate_fsb_count(&mp->m_sb, nb)))
148 		return error;
149 	dpct = pct - mp->m_sb.sb_imax_pct;
150 	error = xfs_read_buf(mp, mp->m_ddev_targp,
151 			XFS_FSB_TO_BB(mp, nb) - XFS_FSS_TO_BB(mp, 1),
152 			XFS_FSS_TO_BB(mp, 1), 0, &bp);
153 	if (error)
154 		return error;
155 	ASSERT(bp);
156 	xfs_buf_relse(bp);
157 
158 	new = nb;	/* use new as a temporary here */
159 	nb_mod = do_div(new, mp->m_sb.sb_agblocks);
160 	nagcount = new + (nb_mod != 0);
161 	if (nb_mod && nb_mod < XFS_MIN_AG_BLOCKS) {
162 		nagcount--;
163 		nb = (xfs_rfsblock_t)nagcount * mp->m_sb.sb_agblocks;
164 		if (nb < mp->m_sb.sb_dblocks)
165 			return XFS_ERROR(EINVAL);
166 	}
167 	new = nb - mp->m_sb.sb_dblocks;
168 	oagcount = mp->m_sb.sb_agcount;
169 	if (nagcount > oagcount) {
170 		void *new_perag, *old_perag;
171 
172 		xfs_filestream_flush(mp);
173 
174 		new_perag = kmem_zalloc(sizeof(xfs_perag_t) * nagcount,
175 					KM_MAYFAIL);
176 		if (!new_perag)
177 			return XFS_ERROR(ENOMEM);
178 
179 		down_write(&mp->m_peraglock);
180 		memcpy(new_perag, mp->m_perag, sizeof(xfs_perag_t) * oagcount);
181 		old_perag = mp->m_perag;
182 		mp->m_perag = new_perag;
183 
184 		mp->m_flags |= XFS_MOUNT_32BITINODES;
185 		nagimax = xfs_initialize_perag(mp, nagcount);
186 		up_write(&mp->m_peraglock);
187 
188 		kmem_free(old_perag);
189 	}
190 	tp = xfs_trans_alloc(mp, XFS_TRANS_GROWFS);
191 	tp->t_flags |= XFS_TRANS_RESERVE;
192 	if ((error = xfs_trans_reserve(tp, XFS_GROWFS_SPACE_RES(mp),
193 			XFS_GROWDATA_LOG_RES(mp), 0, 0, 0))) {
194 		xfs_trans_cancel(tp, 0);
195 		return error;
196 	}
197 
198 	nfree = 0;
199 	for (agno = nagcount - 1; agno >= oagcount; agno--, new -= agsize) {
200 		/*
201 		 * AG freelist header block
202 		 */
203 		bp = xfs_buf_get(mp->m_ddev_targp,
204 				 XFS_AG_DADDR(mp, agno, XFS_AGF_DADDR(mp)),
205 				 XFS_FSS_TO_BB(mp, 1), XBF_LOCK | XBF_MAPPED);
206 		agf = XFS_BUF_TO_AGF(bp);
207 		memset(agf, 0, mp->m_sb.sb_sectsize);
208 		agf->agf_magicnum = cpu_to_be32(XFS_AGF_MAGIC);
209 		agf->agf_versionnum = cpu_to_be32(XFS_AGF_VERSION);
210 		agf->agf_seqno = cpu_to_be32(agno);
211 		if (agno == nagcount - 1)
212 			agsize =
213 				nb -
214 				(agno * (xfs_rfsblock_t)mp->m_sb.sb_agblocks);
215 		else
216 			agsize = mp->m_sb.sb_agblocks;
217 		agf->agf_length = cpu_to_be32(agsize);
218 		agf->agf_roots[XFS_BTNUM_BNOi] = cpu_to_be32(XFS_BNO_BLOCK(mp));
219 		agf->agf_roots[XFS_BTNUM_CNTi] = cpu_to_be32(XFS_CNT_BLOCK(mp));
220 		agf->agf_levels[XFS_BTNUM_BNOi] = cpu_to_be32(1);
221 		agf->agf_levels[XFS_BTNUM_CNTi] = cpu_to_be32(1);
222 		agf->agf_flfirst = 0;
223 		agf->agf_fllast = cpu_to_be32(XFS_AGFL_SIZE(mp) - 1);
224 		agf->agf_flcount = 0;
225 		tmpsize = agsize - XFS_PREALLOC_BLOCKS(mp);
226 		agf->agf_freeblks = cpu_to_be32(tmpsize);
227 		agf->agf_longest = cpu_to_be32(tmpsize);
228 		error = xfs_bwrite(mp, bp);
229 		if (error) {
230 			goto error0;
231 		}
232 		/*
233 		 * AG inode header block
234 		 */
235 		bp = xfs_buf_get(mp->m_ddev_targp,
236 				 XFS_AG_DADDR(mp, agno, XFS_AGI_DADDR(mp)),
237 				 XFS_FSS_TO_BB(mp, 1), XBF_LOCK | XBF_MAPPED);
238 		agi = XFS_BUF_TO_AGI(bp);
239 		memset(agi, 0, mp->m_sb.sb_sectsize);
240 		agi->agi_magicnum = cpu_to_be32(XFS_AGI_MAGIC);
241 		agi->agi_versionnum = cpu_to_be32(XFS_AGI_VERSION);
242 		agi->agi_seqno = cpu_to_be32(agno);
243 		agi->agi_length = cpu_to_be32(agsize);
244 		agi->agi_count = 0;
245 		agi->agi_root = cpu_to_be32(XFS_IBT_BLOCK(mp));
246 		agi->agi_level = cpu_to_be32(1);
247 		agi->agi_freecount = 0;
248 		agi->agi_newino = cpu_to_be32(NULLAGINO);
249 		agi->agi_dirino = cpu_to_be32(NULLAGINO);
250 		for (bucket = 0; bucket < XFS_AGI_UNLINKED_BUCKETS; bucket++)
251 			agi->agi_unlinked[bucket] = cpu_to_be32(NULLAGINO);
252 		error = xfs_bwrite(mp, bp);
253 		if (error) {
254 			goto error0;
255 		}
256 		/*
257 		 * BNO btree root block
258 		 */
259 		bp = xfs_buf_get(mp->m_ddev_targp,
260 				 XFS_AGB_TO_DADDR(mp, agno, XFS_BNO_BLOCK(mp)),
261 				 BTOBB(mp->m_sb.sb_blocksize),
262 				 XBF_LOCK | XBF_MAPPED);
263 		block = XFS_BUF_TO_BLOCK(bp);
264 		memset(block, 0, mp->m_sb.sb_blocksize);
265 		block->bb_magic = cpu_to_be32(XFS_ABTB_MAGIC);
266 		block->bb_level = 0;
267 		block->bb_numrecs = cpu_to_be16(1);
268 		block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
269 		block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
270 		arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
271 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
272 		arec->ar_blockcount = cpu_to_be32(
273 			agsize - be32_to_cpu(arec->ar_startblock));
274 		error = xfs_bwrite(mp, bp);
275 		if (error) {
276 			goto error0;
277 		}
278 		/*
279 		 * CNT btree root block
280 		 */
281 		bp = xfs_buf_get(mp->m_ddev_targp,
282 				 XFS_AGB_TO_DADDR(mp, agno, XFS_CNT_BLOCK(mp)),
283 				 BTOBB(mp->m_sb.sb_blocksize),
284 				 XBF_LOCK | XBF_MAPPED);
285 		block = XFS_BUF_TO_BLOCK(bp);
286 		memset(block, 0, mp->m_sb.sb_blocksize);
287 		block->bb_magic = cpu_to_be32(XFS_ABTC_MAGIC);
288 		block->bb_level = 0;
289 		block->bb_numrecs = cpu_to_be16(1);
290 		block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
291 		block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
292 		arec = XFS_ALLOC_REC_ADDR(mp, block, 1);
293 		arec->ar_startblock = cpu_to_be32(XFS_PREALLOC_BLOCKS(mp));
294 		arec->ar_blockcount = cpu_to_be32(
295 			agsize - be32_to_cpu(arec->ar_startblock));
296 		nfree += be32_to_cpu(arec->ar_blockcount);
297 		error = xfs_bwrite(mp, bp);
298 		if (error) {
299 			goto error0;
300 		}
301 		/*
302 		 * INO btree root block
303 		 */
304 		bp = xfs_buf_get(mp->m_ddev_targp,
305 				 XFS_AGB_TO_DADDR(mp, agno, XFS_IBT_BLOCK(mp)),
306 				 BTOBB(mp->m_sb.sb_blocksize),
307 				 XBF_LOCK | XBF_MAPPED);
308 		block = XFS_BUF_TO_BLOCK(bp);
309 		memset(block, 0, mp->m_sb.sb_blocksize);
310 		block->bb_magic = cpu_to_be32(XFS_IBT_MAGIC);
311 		block->bb_level = 0;
312 		block->bb_numrecs = 0;
313 		block->bb_u.s.bb_leftsib = cpu_to_be32(NULLAGBLOCK);
314 		block->bb_u.s.bb_rightsib = cpu_to_be32(NULLAGBLOCK);
315 		error = xfs_bwrite(mp, bp);
316 		if (error) {
317 			goto error0;
318 		}
319 	}
320 	xfs_trans_agblocks_delta(tp, nfree);
321 	/*
322 	 * There are new blocks in the old last a.g.
323 	 */
324 	if (new) {
325 		/*
326 		 * Change the agi length.
327 		 */
328 		error = xfs_ialloc_read_agi(mp, tp, agno, &bp);
329 		if (error) {
330 			goto error0;
331 		}
332 		ASSERT(bp);
333 		agi = XFS_BUF_TO_AGI(bp);
334 		be32_add_cpu(&agi->agi_length, new);
335 		ASSERT(nagcount == oagcount ||
336 		       be32_to_cpu(agi->agi_length) == mp->m_sb.sb_agblocks);
337 		xfs_ialloc_log_agi(tp, bp, XFS_AGI_LENGTH);
338 		/*
339 		 * Change agf length.
340 		 */
341 		error = xfs_alloc_read_agf(mp, tp, agno, 0, &bp);
342 		if (error) {
343 			goto error0;
344 		}
345 		ASSERT(bp);
346 		agf = XFS_BUF_TO_AGF(bp);
347 		be32_add_cpu(&agf->agf_length, new);
348 		ASSERT(be32_to_cpu(agf->agf_length) ==
349 		       be32_to_cpu(agi->agi_length));
350 		xfs_alloc_log_agf(tp, bp, XFS_AGF_LENGTH);
351 		/*
352 		 * Free the new space.
353 		 */
354 		error = xfs_free_extent(tp, XFS_AGB_TO_FSB(mp, agno,
355 			be32_to_cpu(agf->agf_length) - new), new);
356 		if (error) {
357 			goto error0;
358 		}
359 	}
360 	if (nagcount > oagcount)
361 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_AGCOUNT, nagcount - oagcount);
362 	if (nb > mp->m_sb.sb_dblocks)
363 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_DBLOCKS,
364 				 nb - mp->m_sb.sb_dblocks);
365 	if (nfree)
366 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_FDBLOCKS, nfree);
367 	if (dpct)
368 		xfs_trans_mod_sb(tp, XFS_TRANS_SB_IMAXPCT, dpct);
369 	error = xfs_trans_commit(tp, 0);
370 	if (error) {
371 		return error;
372 	}
373 	/* New allocation groups fully initialized, so update mount struct */
374 	if (nagimax)
375 		mp->m_maxagi = nagimax;
376 	if (mp->m_sb.sb_imax_pct) {
377 		__uint64_t icount = mp->m_sb.sb_dblocks * mp->m_sb.sb_imax_pct;
378 		do_div(icount, 100);
379 		mp->m_maxicount = icount << mp->m_sb.sb_inopblog;
380 	} else
381 		mp->m_maxicount = 0;
382 	for (agno = 1; agno < nagcount; agno++) {
383 		error = xfs_read_buf(mp, mp->m_ddev_targp,
384 				  XFS_AGB_TO_DADDR(mp, agno, XFS_SB_BLOCK(mp)),
385 				  XFS_FSS_TO_BB(mp, 1), 0, &bp);
386 		if (error) {
387 			xfs_fs_cmn_err(CE_WARN, mp,
388 			"error %d reading secondary superblock for ag %d",
389 				error, agno);
390 			break;
391 		}
392 		xfs_sb_to_disk(XFS_BUF_TO_SBP(bp), &mp->m_sb, XFS_SB_ALL_BITS);
393 		/*
394 		 * If we get an error writing out the alternate superblocks,
395 		 * just issue a warning and continue.  The real work is
396 		 * already done and committed.
397 		 */
398 		if (!(error = xfs_bwrite(mp, bp))) {
399 			continue;
400 		} else {
401 			xfs_fs_cmn_err(CE_WARN, mp,
402 		"write error %d updating secondary superblock for ag %d",
403 				error, agno);
404 			break; /* no point in continuing */
405 		}
406 	}
407 	return 0;
408 
409  error0:
410 	xfs_trans_cancel(tp, XFS_TRANS_ABORT);
411 	return error;
412 }
413 
414 static int
415 xfs_growfs_log_private(
416 	xfs_mount_t		*mp,	/* mount point for filesystem */
417 	xfs_growfs_log_t	*in)	/* growfs log input struct */
418 {
419 	xfs_extlen_t		nb;
420 
421 	nb = in->newblocks;
422 	if (nb < XFS_MIN_LOG_BLOCKS || nb < XFS_B_TO_FSB(mp, XFS_MIN_LOG_BYTES))
423 		return XFS_ERROR(EINVAL);
424 	if (nb == mp->m_sb.sb_logblocks &&
425 	    in->isint == (mp->m_sb.sb_logstart != 0))
426 		return XFS_ERROR(EINVAL);
427 	/*
428 	 * Moving the log is hard, need new interfaces to sync
429 	 * the log first, hold off all activity while moving it.
430 	 * Can have shorter or longer log in the same space,
431 	 * or transform internal to external log or vice versa.
432 	 */
433 	return XFS_ERROR(ENOSYS);
434 }
435 
436 /*
437  * protected versions of growfs function acquire and release locks on the mount
438  * point - exported through ioctls: XFS_IOC_FSGROWFSDATA, XFS_IOC_FSGROWFSLOG,
439  * XFS_IOC_FSGROWFSRT
440  */
441 
442 
443 int
444 xfs_growfs_data(
445 	xfs_mount_t		*mp,
446 	xfs_growfs_data_t	*in)
447 {
448 	int error;
449 
450 	if (!capable(CAP_SYS_ADMIN))
451 		return XFS_ERROR(EPERM);
452 	if (!mutex_trylock(&mp->m_growlock))
453 		return XFS_ERROR(EWOULDBLOCK);
454 	error = xfs_growfs_data_private(mp, in);
455 	mutex_unlock(&mp->m_growlock);
456 	return error;
457 }
458 
459 int
460 xfs_growfs_log(
461 	xfs_mount_t		*mp,
462 	xfs_growfs_log_t	*in)
463 {
464 	int error;
465 
466 	if (!capable(CAP_SYS_ADMIN))
467 		return XFS_ERROR(EPERM);
468 	if (!mutex_trylock(&mp->m_growlock))
469 		return XFS_ERROR(EWOULDBLOCK);
470 	error = xfs_growfs_log_private(mp, in);
471 	mutex_unlock(&mp->m_growlock);
472 	return error;
473 }
474 
475 /*
476  * exported through ioctl XFS_IOC_FSCOUNTS
477  */
478 
479 int
480 xfs_fs_counts(
481 	xfs_mount_t		*mp,
482 	xfs_fsop_counts_t	*cnt)
483 {
484 	xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
485 	spin_lock(&mp->m_sb_lock);
486 	cnt->freedata = mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
487 	cnt->freertx = mp->m_sb.sb_frextents;
488 	cnt->freeino = mp->m_sb.sb_ifree;
489 	cnt->allocino = mp->m_sb.sb_icount;
490 	spin_unlock(&mp->m_sb_lock);
491 	return 0;
492 }
493 
494 /*
495  * exported through ioctl XFS_IOC_SET_RESBLKS & XFS_IOC_GET_RESBLKS
496  *
497  * xfs_reserve_blocks is called to set m_resblks
498  * in the in-core mount table. The number of unused reserved blocks
499  * is kept in m_resblks_avail.
500  *
501  * Reserve the requested number of blocks if available. Otherwise return
502  * as many as possible to satisfy the request. The actual number
503  * reserved are returned in outval
504  *
505  * A null inval pointer indicates that only the current reserved blocks
506  * available  should  be returned no settings are changed.
507  */
508 
509 int
510 xfs_reserve_blocks(
511 	xfs_mount_t             *mp,
512 	__uint64_t              *inval,
513 	xfs_fsop_resblks_t      *outval)
514 {
515 	__int64_t		lcounter, delta, fdblks_delta;
516 	__uint64_t		request;
517 
518 	/* If inval is null, report current values and return */
519 	if (inval == (__uint64_t *)NULL) {
520 		if (!outval)
521 			return EINVAL;
522 		outval->resblks = mp->m_resblks;
523 		outval->resblks_avail = mp->m_resblks_avail;
524 		return 0;
525 	}
526 
527 	request = *inval;
528 
529 	/*
530 	 * With per-cpu counters, this becomes an interesting
531 	 * problem. we needto work out if we are freeing or allocation
532 	 * blocks first, then we can do the modification as necessary.
533 	 *
534 	 * We do this under the m_sb_lock so that if we are near
535 	 * ENOSPC, we will hold out any changes while we work out
536 	 * what to do. This means that the amount of free space can
537 	 * change while we do this, so we need to retry if we end up
538 	 * trying to reserve more space than is available.
539 	 *
540 	 * We also use the xfs_mod_incore_sb() interface so that we
541 	 * don't have to care about whether per cpu counter are
542 	 * enabled, disabled or even compiled in....
543 	 */
544 retry:
545 	spin_lock(&mp->m_sb_lock);
546 	xfs_icsb_sync_counters_locked(mp, 0);
547 
548 	/*
549 	 * If our previous reservation was larger than the current value,
550 	 * then move any unused blocks back to the free pool.
551 	 */
552 	fdblks_delta = 0;
553 	if (mp->m_resblks > request) {
554 		lcounter = mp->m_resblks_avail - request;
555 		if (lcounter  > 0) {		/* release unused blocks */
556 			fdblks_delta = lcounter;
557 			mp->m_resblks_avail -= lcounter;
558 		}
559 		mp->m_resblks = request;
560 	} else {
561 		__int64_t	free;
562 
563 		free =  mp->m_sb.sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
564 		if (!free)
565 			goto out; /* ENOSPC and fdblks_delta = 0 */
566 
567 		delta = request - mp->m_resblks;
568 		lcounter = free - delta;
569 		if (lcounter < 0) {
570 			/* We can't satisfy the request, just get what we can */
571 			mp->m_resblks += free;
572 			mp->m_resblks_avail += free;
573 			fdblks_delta = -free;
574 		} else {
575 			fdblks_delta = -delta;
576 			mp->m_resblks = request;
577 			mp->m_resblks_avail += delta;
578 		}
579 	}
580 out:
581 	if (outval) {
582 		outval->resblks = mp->m_resblks;
583 		outval->resblks_avail = mp->m_resblks_avail;
584 	}
585 	spin_unlock(&mp->m_sb_lock);
586 
587 	if (fdblks_delta) {
588 		/*
589 		 * If we are putting blocks back here, m_resblks_avail is
590 		 * already at its max so this will put it in the free pool.
591 		 *
592 		 * If we need space, we'll either succeed in getting it
593 		 * from the free block count or we'll get an enospc. If
594 		 * we get a ENOSPC, it means things changed while we were
595 		 * calculating fdblks_delta and so we should try again to
596 		 * see if there is anything left to reserve.
597 		 *
598 		 * Don't set the reserved flag here - we don't want to reserve
599 		 * the extra reserve blocks from the reserve.....
600 		 */
601 		int error;
602 		error = xfs_mod_incore_sb(mp, XFS_SBS_FDBLOCKS, fdblks_delta, 0);
603 		if (error == ENOSPC)
604 			goto retry;
605 	}
606 	return 0;
607 }
608 
609 int
610 xfs_fs_log_dummy(
611 	xfs_mount_t	*mp)
612 {
613 	xfs_trans_t	*tp;
614 	xfs_inode_t	*ip;
615 	int		error;
616 
617 	tp = _xfs_trans_alloc(mp, XFS_TRANS_DUMMY1, KM_SLEEP);
618 	error = xfs_trans_reserve(tp, 0, XFS_ICHANGE_LOG_RES(mp), 0, 0, 0);
619 	if (error) {
620 		xfs_trans_cancel(tp, 0);
621 		return error;
622 	}
623 
624 	ip = mp->m_rootip;
625 	xfs_ilock(ip, XFS_ILOCK_EXCL);
626 
627 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
628 	xfs_trans_ihold(tp, ip);
629 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
630 	xfs_trans_set_sync(tp);
631 	error = xfs_trans_commit(tp, 0);
632 
633 	xfs_iunlock(ip, XFS_ILOCK_EXCL);
634 	return error;
635 }
636 
637 int
638 xfs_fs_goingdown(
639 	xfs_mount_t	*mp,
640 	__uint32_t	inflags)
641 {
642 	switch (inflags) {
643 	case XFS_FSOP_GOING_FLAGS_DEFAULT: {
644 		struct super_block *sb = freeze_bdev(mp->m_super->s_bdev);
645 
646 		if (sb && !IS_ERR(sb)) {
647 			xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
648 			thaw_bdev(sb->s_bdev, sb);
649 		}
650 
651 		break;
652 	}
653 	case XFS_FSOP_GOING_FLAGS_LOGFLUSH:
654 		xfs_force_shutdown(mp, SHUTDOWN_FORCE_UMOUNT);
655 		break;
656 	case XFS_FSOP_GOING_FLAGS_NOLOGFLUSH:
657 		xfs_force_shutdown(mp,
658 				SHUTDOWN_FORCE_UMOUNT | SHUTDOWN_LOG_IO_ERROR);
659 		break;
660 	default:
661 		return XFS_ERROR(EINVAL);
662 	}
663 
664 	return 0;
665 }
666