xref: /openbmc/linux/fs/xfs/xfs_super.c (revision 93d90ad7)
1 /*
2  * Copyright (c) 2000-2006 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 
19 #include "xfs.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_inode.h"
28 #include "xfs_btree.h"
29 #include "xfs_bmap.h"
30 #include "xfs_alloc.h"
31 #include "xfs_error.h"
32 #include "xfs_fsops.h"
33 #include "xfs_trans.h"
34 #include "xfs_buf_item.h"
35 #include "xfs_log.h"
36 #include "xfs_log_priv.h"
37 #include "xfs_da_btree.h"
38 #include "xfs_dir2.h"
39 #include "xfs_extfree_item.h"
40 #include "xfs_mru_cache.h"
41 #include "xfs_inode_item.h"
42 #include "xfs_icache.h"
43 #include "xfs_trace.h"
44 #include "xfs_icreate_item.h"
45 #include "xfs_filestream.h"
46 #include "xfs_quota.h"
47 #include "xfs_sysfs.h"
48 
49 #include <linux/namei.h>
50 #include <linux/init.h>
51 #include <linux/slab.h>
52 #include <linux/mount.h>
53 #include <linux/mempool.h>
54 #include <linux/writeback.h>
55 #include <linux/kthread.h>
56 #include <linux/freezer.h>
57 #include <linux/parser.h>
58 
59 static const struct super_operations xfs_super_operations;
60 static kmem_zone_t *xfs_ioend_zone;
61 mempool_t *xfs_ioend_pool;
62 
63 static struct kset *xfs_kset;		/* top-level xfs sysfs dir */
64 #ifdef DEBUG
65 static struct xfs_kobj xfs_dbg_kobj;	/* global debug sysfs attrs */
66 #endif
67 
68 #define MNTOPT_LOGBUFS	"logbufs"	/* number of XFS log buffers */
69 #define MNTOPT_LOGBSIZE	"logbsize"	/* size of XFS log buffers */
70 #define MNTOPT_LOGDEV	"logdev"	/* log device */
71 #define MNTOPT_RTDEV	"rtdev"		/* realtime I/O device */
72 #define MNTOPT_BIOSIZE	"biosize"	/* log2 of preferred buffered io size */
73 #define MNTOPT_WSYNC	"wsync"		/* safe-mode nfs compatible mount */
74 #define MNTOPT_NOALIGN	"noalign"	/* turn off stripe alignment */
75 #define MNTOPT_SWALLOC	"swalloc"	/* turn on stripe width allocation */
76 #define MNTOPT_SUNIT	"sunit"		/* data volume stripe unit */
77 #define MNTOPT_SWIDTH	"swidth"	/* data volume stripe width */
78 #define MNTOPT_NOUUID	"nouuid"	/* ignore filesystem UUID */
79 #define MNTOPT_MTPT	"mtpt"		/* filesystem mount point */
80 #define MNTOPT_GRPID	"grpid"		/* group-ID from parent directory */
81 #define MNTOPT_NOGRPID	"nogrpid"	/* group-ID from current process */
82 #define MNTOPT_BSDGROUPS    "bsdgroups"    /* group-ID from parent directory */
83 #define MNTOPT_SYSVGROUPS   "sysvgroups"   /* group-ID from current process */
84 #define MNTOPT_ALLOCSIZE    "allocsize"    /* preferred allocation size */
85 #define MNTOPT_NORECOVERY   "norecovery"   /* don't run XFS recovery */
86 #define MNTOPT_BARRIER	"barrier"	/* use writer barriers for log write and
87 					 * unwritten extent conversion */
88 #define MNTOPT_NOBARRIER "nobarrier"	/* .. disable */
89 #define MNTOPT_64BITINODE   "inode64"	/* inodes can be allocated anywhere */
90 #define MNTOPT_32BITINODE   "inode32"	/* inode allocation limited to
91 					 * XFS_MAXINUMBER_32 */
92 #define MNTOPT_IKEEP	"ikeep"		/* do not free empty inode clusters */
93 #define MNTOPT_NOIKEEP	"noikeep"	/* free empty inode clusters */
94 #define MNTOPT_LARGEIO	   "largeio"	/* report large I/O sizes in stat() */
95 #define MNTOPT_NOLARGEIO   "nolargeio"	/* do not report large I/O sizes
96 					 * in stat(). */
97 #define MNTOPT_ATTR2	"attr2"		/* do use attr2 attribute format */
98 #define MNTOPT_NOATTR2	"noattr2"	/* do not use attr2 attribute format */
99 #define MNTOPT_FILESTREAM  "filestreams" /* use filestreams allocator */
100 #define MNTOPT_QUOTA	"quota"		/* disk quotas (user) */
101 #define MNTOPT_NOQUOTA	"noquota"	/* no quotas */
102 #define MNTOPT_USRQUOTA	"usrquota"	/* user quota enabled */
103 #define MNTOPT_GRPQUOTA	"grpquota"	/* group quota enabled */
104 #define MNTOPT_PRJQUOTA	"prjquota"	/* project quota enabled */
105 #define MNTOPT_UQUOTA	"uquota"	/* user quota (IRIX variant) */
106 #define MNTOPT_GQUOTA	"gquota"	/* group quota (IRIX variant) */
107 #define MNTOPT_PQUOTA	"pquota"	/* project quota (IRIX variant) */
108 #define MNTOPT_UQUOTANOENF "uqnoenforce"/* user quota limit enforcement */
109 #define MNTOPT_GQUOTANOENF "gqnoenforce"/* group quota limit enforcement */
110 #define MNTOPT_PQUOTANOENF "pqnoenforce"/* project quota limit enforcement */
111 #define MNTOPT_QUOTANOENF  "qnoenforce"	/* same as uqnoenforce */
112 #define MNTOPT_DELAYLOG    "delaylog"	/* Delayed logging enabled */
113 #define MNTOPT_NODELAYLOG  "nodelaylog"	/* Delayed logging disabled */
114 #define MNTOPT_DISCARD	   "discard"	/* Discard unused blocks */
115 #define MNTOPT_NODISCARD   "nodiscard"	/* Do not discard unused blocks */
116 
117 /*
118  * Table driven mount option parser.
119  *
120  * Currently only used for remount, but it will be used for mount
121  * in the future, too.
122  */
123 enum {
124 	Opt_barrier,
125 	Opt_nobarrier,
126 	Opt_inode64,
127 	Opt_inode32,
128 	Opt_err
129 };
130 
131 static const match_table_t tokens = {
132 	{Opt_barrier, "barrier"},
133 	{Opt_nobarrier, "nobarrier"},
134 	{Opt_inode64, "inode64"},
135 	{Opt_inode32, "inode32"},
136 	{Opt_err, NULL}
137 };
138 
139 
140 STATIC unsigned long
141 suffix_kstrtoint(char *s, unsigned int base, int *res)
142 {
143 	int	last, shift_left_factor = 0, _res;
144 	char	*value = s;
145 
146 	last = strlen(value) - 1;
147 	if (value[last] == 'K' || value[last] == 'k') {
148 		shift_left_factor = 10;
149 		value[last] = '\0';
150 	}
151 	if (value[last] == 'M' || value[last] == 'm') {
152 		shift_left_factor = 20;
153 		value[last] = '\0';
154 	}
155 	if (value[last] == 'G' || value[last] == 'g') {
156 		shift_left_factor = 30;
157 		value[last] = '\0';
158 	}
159 
160 	if (kstrtoint(s, base, &_res))
161 		return -EINVAL;
162 	*res = _res << shift_left_factor;
163 	return 0;
164 }
165 
166 /*
167  * This function fills in xfs_mount_t fields based on mount args.
168  * Note: the superblock has _not_ yet been read in.
169  *
170  * Note that this function leaks the various device name allocations on
171  * failure.  The caller takes care of them.
172  */
173 STATIC int
174 xfs_parseargs(
175 	struct xfs_mount	*mp,
176 	char			*options)
177 {
178 	struct super_block	*sb = mp->m_super;
179 	char			*this_char, *value;
180 	int			dsunit = 0;
181 	int			dswidth = 0;
182 	int			iosize = 0;
183 	__uint8_t		iosizelog = 0;
184 
185 	/*
186 	 * set up the mount name first so all the errors will refer to the
187 	 * correct device.
188 	 */
189 	mp->m_fsname = kstrndup(sb->s_id, MAXNAMELEN, GFP_KERNEL);
190 	if (!mp->m_fsname)
191 		return -ENOMEM;
192 	mp->m_fsname_len = strlen(mp->m_fsname) + 1;
193 
194 	/*
195 	 * Copy binary VFS mount flags we are interested in.
196 	 */
197 	if (sb->s_flags & MS_RDONLY)
198 		mp->m_flags |= XFS_MOUNT_RDONLY;
199 	if (sb->s_flags & MS_DIRSYNC)
200 		mp->m_flags |= XFS_MOUNT_DIRSYNC;
201 	if (sb->s_flags & MS_SYNCHRONOUS)
202 		mp->m_flags |= XFS_MOUNT_WSYNC;
203 
204 	/*
205 	 * Set some default flags that could be cleared by the mount option
206 	 * parsing.
207 	 */
208 	mp->m_flags |= XFS_MOUNT_BARRIER;
209 	mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
210 
211 	/*
212 	 * These can be overridden by the mount option parsing.
213 	 */
214 	mp->m_logbufs = -1;
215 	mp->m_logbsize = -1;
216 
217 	if (!options)
218 		goto done;
219 
220 	while ((this_char = strsep(&options, ",")) != NULL) {
221 		if (!*this_char)
222 			continue;
223 		if ((value = strchr(this_char, '=')) != NULL)
224 			*value++ = 0;
225 
226 		if (!strcmp(this_char, MNTOPT_LOGBUFS)) {
227 			if (!value || !*value) {
228 				xfs_warn(mp, "%s option requires an argument",
229 					this_char);
230 				return -EINVAL;
231 			}
232 			if (kstrtoint(value, 10, &mp->m_logbufs))
233 				return -EINVAL;
234 		} else if (!strcmp(this_char, MNTOPT_LOGBSIZE)) {
235 			if (!value || !*value) {
236 				xfs_warn(mp, "%s option requires an argument",
237 					this_char);
238 				return -EINVAL;
239 			}
240 			if (suffix_kstrtoint(value, 10, &mp->m_logbsize))
241 				return -EINVAL;
242 		} else if (!strcmp(this_char, MNTOPT_LOGDEV)) {
243 			if (!value || !*value) {
244 				xfs_warn(mp, "%s option requires an argument",
245 					this_char);
246 				return -EINVAL;
247 			}
248 			mp->m_logname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
249 			if (!mp->m_logname)
250 				return -ENOMEM;
251 		} else if (!strcmp(this_char, MNTOPT_MTPT)) {
252 			xfs_warn(mp, "%s option not allowed on this system",
253 				this_char);
254 			return -EINVAL;
255 		} else if (!strcmp(this_char, MNTOPT_RTDEV)) {
256 			if (!value || !*value) {
257 				xfs_warn(mp, "%s option requires an argument",
258 					this_char);
259 				return -EINVAL;
260 			}
261 			mp->m_rtname = kstrndup(value, MAXNAMELEN, GFP_KERNEL);
262 			if (!mp->m_rtname)
263 				return -ENOMEM;
264 		} else if (!strcmp(this_char, MNTOPT_BIOSIZE)) {
265 			if (!value || !*value) {
266 				xfs_warn(mp, "%s option requires an argument",
267 					this_char);
268 				return -EINVAL;
269 			}
270 			if (kstrtoint(value, 10, &iosize))
271 				return -EINVAL;
272 			iosizelog = ffs(iosize) - 1;
273 		} else if (!strcmp(this_char, MNTOPT_ALLOCSIZE)) {
274 			if (!value || !*value) {
275 				xfs_warn(mp, "%s option requires an argument",
276 					this_char);
277 				return -EINVAL;
278 			}
279 			if (suffix_kstrtoint(value, 10, &iosize))
280 				return -EINVAL;
281 			iosizelog = ffs(iosize) - 1;
282 		} else if (!strcmp(this_char, MNTOPT_GRPID) ||
283 			   !strcmp(this_char, MNTOPT_BSDGROUPS)) {
284 			mp->m_flags |= XFS_MOUNT_GRPID;
285 		} else if (!strcmp(this_char, MNTOPT_NOGRPID) ||
286 			   !strcmp(this_char, MNTOPT_SYSVGROUPS)) {
287 			mp->m_flags &= ~XFS_MOUNT_GRPID;
288 		} else if (!strcmp(this_char, MNTOPT_WSYNC)) {
289 			mp->m_flags |= XFS_MOUNT_WSYNC;
290 		} else if (!strcmp(this_char, MNTOPT_NORECOVERY)) {
291 			mp->m_flags |= XFS_MOUNT_NORECOVERY;
292 		} else if (!strcmp(this_char, MNTOPT_NOALIGN)) {
293 			mp->m_flags |= XFS_MOUNT_NOALIGN;
294 		} else if (!strcmp(this_char, MNTOPT_SWALLOC)) {
295 			mp->m_flags |= XFS_MOUNT_SWALLOC;
296 		} else if (!strcmp(this_char, MNTOPT_SUNIT)) {
297 			if (!value || !*value) {
298 				xfs_warn(mp, "%s option requires an argument",
299 					this_char);
300 				return -EINVAL;
301 			}
302 			if (kstrtoint(value, 10, &dsunit))
303 				return -EINVAL;
304 		} else if (!strcmp(this_char, MNTOPT_SWIDTH)) {
305 			if (!value || !*value) {
306 				xfs_warn(mp, "%s option requires an argument",
307 					this_char);
308 				return -EINVAL;
309 			}
310 			if (kstrtoint(value, 10, &dswidth))
311 				return -EINVAL;
312 		} else if (!strcmp(this_char, MNTOPT_32BITINODE)) {
313 			mp->m_flags |= XFS_MOUNT_SMALL_INUMS;
314 		} else if (!strcmp(this_char, MNTOPT_64BITINODE)) {
315 			mp->m_flags &= ~XFS_MOUNT_SMALL_INUMS;
316 		} else if (!strcmp(this_char, MNTOPT_NOUUID)) {
317 			mp->m_flags |= XFS_MOUNT_NOUUID;
318 		} else if (!strcmp(this_char, MNTOPT_BARRIER)) {
319 			mp->m_flags |= XFS_MOUNT_BARRIER;
320 		} else if (!strcmp(this_char, MNTOPT_NOBARRIER)) {
321 			mp->m_flags &= ~XFS_MOUNT_BARRIER;
322 		} else if (!strcmp(this_char, MNTOPT_IKEEP)) {
323 			mp->m_flags |= XFS_MOUNT_IKEEP;
324 		} else if (!strcmp(this_char, MNTOPT_NOIKEEP)) {
325 			mp->m_flags &= ~XFS_MOUNT_IKEEP;
326 		} else if (!strcmp(this_char, MNTOPT_LARGEIO)) {
327 			mp->m_flags &= ~XFS_MOUNT_COMPAT_IOSIZE;
328 		} else if (!strcmp(this_char, MNTOPT_NOLARGEIO)) {
329 			mp->m_flags |= XFS_MOUNT_COMPAT_IOSIZE;
330 		} else if (!strcmp(this_char, MNTOPT_ATTR2)) {
331 			mp->m_flags |= XFS_MOUNT_ATTR2;
332 		} else if (!strcmp(this_char, MNTOPT_NOATTR2)) {
333 			mp->m_flags &= ~XFS_MOUNT_ATTR2;
334 			mp->m_flags |= XFS_MOUNT_NOATTR2;
335 		} else if (!strcmp(this_char, MNTOPT_FILESTREAM)) {
336 			mp->m_flags |= XFS_MOUNT_FILESTREAMS;
337 		} else if (!strcmp(this_char, MNTOPT_NOQUOTA)) {
338 			mp->m_qflags &= ~XFS_ALL_QUOTA_ACCT;
339 			mp->m_qflags &= ~XFS_ALL_QUOTA_ENFD;
340 			mp->m_qflags &= ~XFS_ALL_QUOTA_ACTIVE;
341 		} else if (!strcmp(this_char, MNTOPT_QUOTA) ||
342 			   !strcmp(this_char, MNTOPT_UQUOTA) ||
343 			   !strcmp(this_char, MNTOPT_USRQUOTA)) {
344 			mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE |
345 					 XFS_UQUOTA_ENFD);
346 		} else if (!strcmp(this_char, MNTOPT_QUOTANOENF) ||
347 			   !strcmp(this_char, MNTOPT_UQUOTANOENF)) {
348 			mp->m_qflags |= (XFS_UQUOTA_ACCT | XFS_UQUOTA_ACTIVE);
349 			mp->m_qflags &= ~XFS_UQUOTA_ENFD;
350 		} else if (!strcmp(this_char, MNTOPT_PQUOTA) ||
351 			   !strcmp(this_char, MNTOPT_PRJQUOTA)) {
352 			mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE |
353 					 XFS_PQUOTA_ENFD);
354 		} else if (!strcmp(this_char, MNTOPT_PQUOTANOENF)) {
355 			mp->m_qflags |= (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE);
356 			mp->m_qflags &= ~XFS_PQUOTA_ENFD;
357 		} else if (!strcmp(this_char, MNTOPT_GQUOTA) ||
358 			   !strcmp(this_char, MNTOPT_GRPQUOTA)) {
359 			mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE |
360 					 XFS_GQUOTA_ENFD);
361 		} else if (!strcmp(this_char, MNTOPT_GQUOTANOENF)) {
362 			mp->m_qflags |= (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE);
363 			mp->m_qflags &= ~XFS_GQUOTA_ENFD;
364 		} else if (!strcmp(this_char, MNTOPT_DELAYLOG)) {
365 			xfs_warn(mp,
366 	"delaylog is the default now, option is deprecated.");
367 		} else if (!strcmp(this_char, MNTOPT_NODELAYLOG)) {
368 			xfs_warn(mp,
369 	"nodelaylog support has been removed, option is deprecated.");
370 		} else if (!strcmp(this_char, MNTOPT_DISCARD)) {
371 			mp->m_flags |= XFS_MOUNT_DISCARD;
372 		} else if (!strcmp(this_char, MNTOPT_NODISCARD)) {
373 			mp->m_flags &= ~XFS_MOUNT_DISCARD;
374 		} else if (!strcmp(this_char, "ihashsize")) {
375 			xfs_warn(mp,
376 	"ihashsize no longer used, option is deprecated.");
377 		} else if (!strcmp(this_char, "osyncisdsync")) {
378 			xfs_warn(mp,
379 	"osyncisdsync has no effect, option is deprecated.");
380 		} else if (!strcmp(this_char, "osyncisosync")) {
381 			xfs_warn(mp,
382 	"osyncisosync has no effect, option is deprecated.");
383 		} else if (!strcmp(this_char, "irixsgid")) {
384 			xfs_warn(mp,
385 	"irixsgid is now a sysctl(2) variable, option is deprecated.");
386 		} else {
387 			xfs_warn(mp, "unknown mount option [%s].", this_char);
388 			return -EINVAL;
389 		}
390 	}
391 
392 	/*
393 	 * no recovery flag requires a read-only mount
394 	 */
395 	if ((mp->m_flags & XFS_MOUNT_NORECOVERY) &&
396 	    !(mp->m_flags & XFS_MOUNT_RDONLY)) {
397 		xfs_warn(mp, "no-recovery mounts must be read-only.");
398 		return -EINVAL;
399 	}
400 
401 	if ((mp->m_flags & XFS_MOUNT_NOALIGN) && (dsunit || dswidth)) {
402 		xfs_warn(mp,
403 	"sunit and swidth options incompatible with the noalign option");
404 		return -EINVAL;
405 	}
406 
407 #ifndef CONFIG_XFS_QUOTA
408 	if (XFS_IS_QUOTA_RUNNING(mp)) {
409 		xfs_warn(mp, "quota support not available in this kernel.");
410 		return -EINVAL;
411 	}
412 #endif
413 
414 	if ((dsunit && !dswidth) || (!dsunit && dswidth)) {
415 		xfs_warn(mp, "sunit and swidth must be specified together");
416 		return -EINVAL;
417 	}
418 
419 	if (dsunit && (dswidth % dsunit != 0)) {
420 		xfs_warn(mp,
421 	"stripe width (%d) must be a multiple of the stripe unit (%d)",
422 			dswidth, dsunit);
423 		return -EINVAL;
424 	}
425 
426 done:
427 	if (dsunit && !(mp->m_flags & XFS_MOUNT_NOALIGN)) {
428 		/*
429 		 * At this point the superblock has not been read
430 		 * in, therefore we do not know the block size.
431 		 * Before the mount call ends we will convert
432 		 * these to FSBs.
433 		 */
434 		mp->m_dalign = dsunit;
435 		mp->m_swidth = dswidth;
436 	}
437 
438 	if (mp->m_logbufs != -1 &&
439 	    mp->m_logbufs != 0 &&
440 	    (mp->m_logbufs < XLOG_MIN_ICLOGS ||
441 	     mp->m_logbufs > XLOG_MAX_ICLOGS)) {
442 		xfs_warn(mp, "invalid logbufs value: %d [not %d-%d]",
443 			mp->m_logbufs, XLOG_MIN_ICLOGS, XLOG_MAX_ICLOGS);
444 		return -EINVAL;
445 	}
446 	if (mp->m_logbsize != -1 &&
447 	    mp->m_logbsize !=  0 &&
448 	    (mp->m_logbsize < XLOG_MIN_RECORD_BSIZE ||
449 	     mp->m_logbsize > XLOG_MAX_RECORD_BSIZE ||
450 	     !is_power_of_2(mp->m_logbsize))) {
451 		xfs_warn(mp,
452 			"invalid logbufsize: %d [not 16k,32k,64k,128k or 256k]",
453 			mp->m_logbsize);
454 		return -EINVAL;
455 	}
456 
457 	if (iosizelog) {
458 		if (iosizelog > XFS_MAX_IO_LOG ||
459 		    iosizelog < XFS_MIN_IO_LOG) {
460 			xfs_warn(mp, "invalid log iosize: %d [not %d-%d]",
461 				iosizelog, XFS_MIN_IO_LOG,
462 				XFS_MAX_IO_LOG);
463 			return -EINVAL;
464 		}
465 
466 		mp->m_flags |= XFS_MOUNT_DFLT_IOSIZE;
467 		mp->m_readio_log = iosizelog;
468 		mp->m_writeio_log = iosizelog;
469 	}
470 
471 	return 0;
472 }
473 
474 struct proc_xfs_info {
475 	int	flag;
476 	char	*str;
477 };
478 
479 STATIC int
480 xfs_showargs(
481 	struct xfs_mount	*mp,
482 	struct seq_file		*m)
483 {
484 	static struct proc_xfs_info xfs_info_set[] = {
485 		/* the few simple ones we can get from the mount struct */
486 		{ XFS_MOUNT_IKEEP,		"," MNTOPT_IKEEP },
487 		{ XFS_MOUNT_WSYNC,		"," MNTOPT_WSYNC },
488 		{ XFS_MOUNT_NOALIGN,		"," MNTOPT_NOALIGN },
489 		{ XFS_MOUNT_SWALLOC,		"," MNTOPT_SWALLOC },
490 		{ XFS_MOUNT_NOUUID,		"," MNTOPT_NOUUID },
491 		{ XFS_MOUNT_NORECOVERY,		"," MNTOPT_NORECOVERY },
492 		{ XFS_MOUNT_ATTR2,		"," MNTOPT_ATTR2 },
493 		{ XFS_MOUNT_FILESTREAMS,	"," MNTOPT_FILESTREAM },
494 		{ XFS_MOUNT_GRPID,		"," MNTOPT_GRPID },
495 		{ XFS_MOUNT_DISCARD,		"," MNTOPT_DISCARD },
496 		{ XFS_MOUNT_SMALL_INUMS,	"," MNTOPT_32BITINODE },
497 		{ 0, NULL }
498 	};
499 	static struct proc_xfs_info xfs_info_unset[] = {
500 		/* the few simple ones we can get from the mount struct */
501 		{ XFS_MOUNT_COMPAT_IOSIZE,	"," MNTOPT_LARGEIO },
502 		{ XFS_MOUNT_BARRIER,		"," MNTOPT_NOBARRIER },
503 		{ XFS_MOUNT_SMALL_INUMS,	"," MNTOPT_64BITINODE },
504 		{ 0, NULL }
505 	};
506 	struct proc_xfs_info	*xfs_infop;
507 
508 	for (xfs_infop = xfs_info_set; xfs_infop->flag; xfs_infop++) {
509 		if (mp->m_flags & xfs_infop->flag)
510 			seq_puts(m, xfs_infop->str);
511 	}
512 	for (xfs_infop = xfs_info_unset; xfs_infop->flag; xfs_infop++) {
513 		if (!(mp->m_flags & xfs_infop->flag))
514 			seq_puts(m, xfs_infop->str);
515 	}
516 
517 	if (mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)
518 		seq_printf(m, "," MNTOPT_ALLOCSIZE "=%dk",
519 				(int)(1 << mp->m_writeio_log) >> 10);
520 
521 	if (mp->m_logbufs > 0)
522 		seq_printf(m, "," MNTOPT_LOGBUFS "=%d", mp->m_logbufs);
523 	if (mp->m_logbsize > 0)
524 		seq_printf(m, "," MNTOPT_LOGBSIZE "=%dk", mp->m_logbsize >> 10);
525 
526 	if (mp->m_logname)
527 		seq_printf(m, "," MNTOPT_LOGDEV "=%s", mp->m_logname);
528 	if (mp->m_rtname)
529 		seq_printf(m, "," MNTOPT_RTDEV "=%s", mp->m_rtname);
530 
531 	if (mp->m_dalign > 0)
532 		seq_printf(m, "," MNTOPT_SUNIT "=%d",
533 				(int)XFS_FSB_TO_BB(mp, mp->m_dalign));
534 	if (mp->m_swidth > 0)
535 		seq_printf(m, "," MNTOPT_SWIDTH "=%d",
536 				(int)XFS_FSB_TO_BB(mp, mp->m_swidth));
537 
538 	if (mp->m_qflags & (XFS_UQUOTA_ACCT|XFS_UQUOTA_ENFD))
539 		seq_puts(m, "," MNTOPT_USRQUOTA);
540 	else if (mp->m_qflags & XFS_UQUOTA_ACCT)
541 		seq_puts(m, "," MNTOPT_UQUOTANOENF);
542 
543 	if (mp->m_qflags & XFS_PQUOTA_ACCT) {
544 		if (mp->m_qflags & XFS_PQUOTA_ENFD)
545 			seq_puts(m, "," MNTOPT_PRJQUOTA);
546 		else
547 			seq_puts(m, "," MNTOPT_PQUOTANOENF);
548 	}
549 	if (mp->m_qflags & XFS_GQUOTA_ACCT) {
550 		if (mp->m_qflags & XFS_GQUOTA_ENFD)
551 			seq_puts(m, "," MNTOPT_GRPQUOTA);
552 		else
553 			seq_puts(m, "," MNTOPT_GQUOTANOENF);
554 	}
555 
556 	if (!(mp->m_qflags & XFS_ALL_QUOTA_ACCT))
557 		seq_puts(m, "," MNTOPT_NOQUOTA);
558 
559 	return 0;
560 }
561 __uint64_t
562 xfs_max_file_offset(
563 	unsigned int		blockshift)
564 {
565 	unsigned int		pagefactor = 1;
566 	unsigned int		bitshift = BITS_PER_LONG - 1;
567 
568 	/* Figure out maximum filesize, on Linux this can depend on
569 	 * the filesystem blocksize (on 32 bit platforms).
570 	 * __block_write_begin does this in an [unsigned] long...
571 	 *      page->index << (PAGE_CACHE_SHIFT - bbits)
572 	 * So, for page sized blocks (4K on 32 bit platforms),
573 	 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
574 	 *      (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
575 	 * but for smaller blocksizes it is less (bbits = log2 bsize).
576 	 * Note1: get_block_t takes a long (implicit cast from above)
577 	 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
578 	 * can optionally convert the [unsigned] long from above into
579 	 * an [unsigned] long long.
580 	 */
581 
582 #if BITS_PER_LONG == 32
583 # if defined(CONFIG_LBDAF)
584 	ASSERT(sizeof(sector_t) == 8);
585 	pagefactor = PAGE_CACHE_SIZE;
586 	bitshift = BITS_PER_LONG;
587 # else
588 	pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
589 # endif
590 #endif
591 
592 	return (((__uint64_t)pagefactor) << bitshift) - 1;
593 }
594 
595 /*
596  * xfs_set_inode32() and xfs_set_inode64() are passed an agcount
597  * because in the growfs case, mp->m_sb.sb_agcount is not updated
598  * yet to the potentially higher ag count.
599  */
600 xfs_agnumber_t
601 xfs_set_inode32(struct xfs_mount *mp, xfs_agnumber_t agcount)
602 {
603 	xfs_agnumber_t	index = 0;
604 	xfs_agnumber_t	maxagi = 0;
605 	xfs_sb_t	*sbp = &mp->m_sb;
606 	xfs_agnumber_t	max_metadata;
607 	xfs_agino_t	agino;
608 	xfs_ino_t	ino;
609 	xfs_perag_t	*pag;
610 
611 	/* Calculate how much should be reserved for inodes to meet
612 	 * the max inode percentage.
613 	 */
614 	if (mp->m_maxicount) {
615 		__uint64_t	icount;
616 
617 		icount = sbp->sb_dblocks * sbp->sb_imax_pct;
618 		do_div(icount, 100);
619 		icount += sbp->sb_agblocks - 1;
620 		do_div(icount, sbp->sb_agblocks);
621 		max_metadata = icount;
622 	} else {
623 		max_metadata = agcount;
624 	}
625 
626 	agino =	XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks - 1, 0);
627 
628 	for (index = 0; index < agcount; index++) {
629 		ino = XFS_AGINO_TO_INO(mp, index, agino);
630 
631 		if (ino > XFS_MAXINUMBER_32) {
632 			pag = xfs_perag_get(mp, index);
633 			pag->pagi_inodeok = 0;
634 			pag->pagf_metadata = 0;
635 			xfs_perag_put(pag);
636 			continue;
637 		}
638 
639 		pag = xfs_perag_get(mp, index);
640 		pag->pagi_inodeok = 1;
641 		maxagi++;
642 		if (index < max_metadata)
643 			pag->pagf_metadata = 1;
644 		xfs_perag_put(pag);
645 	}
646 	mp->m_flags |= (XFS_MOUNT_32BITINODES |
647 			XFS_MOUNT_SMALL_INUMS);
648 
649 	return maxagi;
650 }
651 
652 xfs_agnumber_t
653 xfs_set_inode64(struct xfs_mount *mp, xfs_agnumber_t agcount)
654 {
655 	xfs_agnumber_t index = 0;
656 
657 	for (index = 0; index < agcount; index++) {
658 		struct xfs_perag	*pag;
659 
660 		pag = xfs_perag_get(mp, index);
661 		pag->pagi_inodeok = 1;
662 		pag->pagf_metadata = 0;
663 		xfs_perag_put(pag);
664 	}
665 
666 	/* There is no need for lock protection on m_flags,
667 	 * the rw_semaphore of the VFS superblock is locked
668 	 * during mount/umount/remount operations, so this is
669 	 * enough to avoid concurency on the m_flags field
670 	 */
671 	mp->m_flags &= ~(XFS_MOUNT_32BITINODES |
672 			 XFS_MOUNT_SMALL_INUMS);
673 	return index;
674 }
675 
676 STATIC int
677 xfs_blkdev_get(
678 	xfs_mount_t		*mp,
679 	const char		*name,
680 	struct block_device	**bdevp)
681 {
682 	int			error = 0;
683 
684 	*bdevp = blkdev_get_by_path(name, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
685 				    mp);
686 	if (IS_ERR(*bdevp)) {
687 		error = PTR_ERR(*bdevp);
688 		xfs_warn(mp, "Invalid device [%s], error=%d\n", name, error);
689 	}
690 
691 	return error;
692 }
693 
694 STATIC void
695 xfs_blkdev_put(
696 	struct block_device	*bdev)
697 {
698 	if (bdev)
699 		blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
700 }
701 
702 void
703 xfs_blkdev_issue_flush(
704 	xfs_buftarg_t		*buftarg)
705 {
706 	blkdev_issue_flush(buftarg->bt_bdev, GFP_NOFS, NULL);
707 }
708 
709 STATIC void
710 xfs_close_devices(
711 	struct xfs_mount	*mp)
712 {
713 	if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
714 		struct block_device *logdev = mp->m_logdev_targp->bt_bdev;
715 		xfs_free_buftarg(mp, mp->m_logdev_targp);
716 		xfs_blkdev_put(logdev);
717 	}
718 	if (mp->m_rtdev_targp) {
719 		struct block_device *rtdev = mp->m_rtdev_targp->bt_bdev;
720 		xfs_free_buftarg(mp, mp->m_rtdev_targp);
721 		xfs_blkdev_put(rtdev);
722 	}
723 	xfs_free_buftarg(mp, mp->m_ddev_targp);
724 }
725 
726 /*
727  * The file system configurations are:
728  *	(1) device (partition) with data and internal log
729  *	(2) logical volume with data and log subvolumes.
730  *	(3) logical volume with data, log, and realtime subvolumes.
731  *
732  * We only have to handle opening the log and realtime volumes here if
733  * they are present.  The data subvolume has already been opened by
734  * get_sb_bdev() and is stored in sb->s_bdev.
735  */
736 STATIC int
737 xfs_open_devices(
738 	struct xfs_mount	*mp)
739 {
740 	struct block_device	*ddev = mp->m_super->s_bdev;
741 	struct block_device	*logdev = NULL, *rtdev = NULL;
742 	int			error;
743 
744 	/*
745 	 * Open real time and log devices - order is important.
746 	 */
747 	if (mp->m_logname) {
748 		error = xfs_blkdev_get(mp, mp->m_logname, &logdev);
749 		if (error)
750 			goto out;
751 	}
752 
753 	if (mp->m_rtname) {
754 		error = xfs_blkdev_get(mp, mp->m_rtname, &rtdev);
755 		if (error)
756 			goto out_close_logdev;
757 
758 		if (rtdev == ddev || rtdev == logdev) {
759 			xfs_warn(mp,
760 	"Cannot mount filesystem with identical rtdev and ddev/logdev.");
761 			error = -EINVAL;
762 			goto out_close_rtdev;
763 		}
764 	}
765 
766 	/*
767 	 * Setup xfs_mount buffer target pointers
768 	 */
769 	error = -ENOMEM;
770 	mp->m_ddev_targp = xfs_alloc_buftarg(mp, ddev);
771 	if (!mp->m_ddev_targp)
772 		goto out_close_rtdev;
773 
774 	if (rtdev) {
775 		mp->m_rtdev_targp = xfs_alloc_buftarg(mp, rtdev);
776 		if (!mp->m_rtdev_targp)
777 			goto out_free_ddev_targ;
778 	}
779 
780 	if (logdev && logdev != ddev) {
781 		mp->m_logdev_targp = xfs_alloc_buftarg(mp, logdev);
782 		if (!mp->m_logdev_targp)
783 			goto out_free_rtdev_targ;
784 	} else {
785 		mp->m_logdev_targp = mp->m_ddev_targp;
786 	}
787 
788 	return 0;
789 
790  out_free_rtdev_targ:
791 	if (mp->m_rtdev_targp)
792 		xfs_free_buftarg(mp, mp->m_rtdev_targp);
793  out_free_ddev_targ:
794 	xfs_free_buftarg(mp, mp->m_ddev_targp);
795  out_close_rtdev:
796 	xfs_blkdev_put(rtdev);
797  out_close_logdev:
798 	if (logdev && logdev != ddev)
799 		xfs_blkdev_put(logdev);
800  out:
801 	return error;
802 }
803 
804 /*
805  * Setup xfs_mount buffer target pointers based on superblock
806  */
807 STATIC int
808 xfs_setup_devices(
809 	struct xfs_mount	*mp)
810 {
811 	int			error;
812 
813 	error = xfs_setsize_buftarg(mp->m_ddev_targp, mp->m_sb.sb_sectsize);
814 	if (error)
815 		return error;
816 
817 	if (mp->m_logdev_targp && mp->m_logdev_targp != mp->m_ddev_targp) {
818 		unsigned int	log_sector_size = BBSIZE;
819 
820 		if (xfs_sb_version_hassector(&mp->m_sb))
821 			log_sector_size = mp->m_sb.sb_logsectsize;
822 		error = xfs_setsize_buftarg(mp->m_logdev_targp,
823 					    log_sector_size);
824 		if (error)
825 			return error;
826 	}
827 	if (mp->m_rtdev_targp) {
828 		error = xfs_setsize_buftarg(mp->m_rtdev_targp,
829 					    mp->m_sb.sb_sectsize);
830 		if (error)
831 			return error;
832 	}
833 
834 	return 0;
835 }
836 
837 STATIC int
838 xfs_init_mount_workqueues(
839 	struct xfs_mount	*mp)
840 {
841 	mp->m_buf_workqueue = alloc_workqueue("xfs-buf/%s",
842 			WQ_MEM_RECLAIM|WQ_FREEZABLE, 1, mp->m_fsname);
843 	if (!mp->m_buf_workqueue)
844 		goto out;
845 
846 	mp->m_data_workqueue = alloc_workqueue("xfs-data/%s",
847 			WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
848 	if (!mp->m_data_workqueue)
849 		goto out_destroy_buf;
850 
851 	mp->m_unwritten_workqueue = alloc_workqueue("xfs-conv/%s",
852 			WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
853 	if (!mp->m_unwritten_workqueue)
854 		goto out_destroy_data_iodone_queue;
855 
856 	mp->m_cil_workqueue = alloc_workqueue("xfs-cil/%s",
857 			WQ_MEM_RECLAIM|WQ_FREEZABLE, 0, mp->m_fsname);
858 	if (!mp->m_cil_workqueue)
859 		goto out_destroy_unwritten;
860 
861 	mp->m_reclaim_workqueue = alloc_workqueue("xfs-reclaim/%s",
862 			WQ_FREEZABLE, 0, mp->m_fsname);
863 	if (!mp->m_reclaim_workqueue)
864 		goto out_destroy_cil;
865 
866 	mp->m_log_workqueue = alloc_workqueue("xfs-log/%s",
867 			WQ_FREEZABLE|WQ_HIGHPRI, 0, mp->m_fsname);
868 	if (!mp->m_log_workqueue)
869 		goto out_destroy_reclaim;
870 
871 	mp->m_eofblocks_workqueue = alloc_workqueue("xfs-eofblocks/%s",
872 			WQ_FREEZABLE, 0, mp->m_fsname);
873 	if (!mp->m_eofblocks_workqueue)
874 		goto out_destroy_log;
875 
876 	return 0;
877 
878 out_destroy_log:
879 	destroy_workqueue(mp->m_log_workqueue);
880 out_destroy_reclaim:
881 	destroy_workqueue(mp->m_reclaim_workqueue);
882 out_destroy_cil:
883 	destroy_workqueue(mp->m_cil_workqueue);
884 out_destroy_unwritten:
885 	destroy_workqueue(mp->m_unwritten_workqueue);
886 out_destroy_data_iodone_queue:
887 	destroy_workqueue(mp->m_data_workqueue);
888 out_destroy_buf:
889 	destroy_workqueue(mp->m_buf_workqueue);
890 out:
891 	return -ENOMEM;
892 }
893 
894 STATIC void
895 xfs_destroy_mount_workqueues(
896 	struct xfs_mount	*mp)
897 {
898 	destroy_workqueue(mp->m_eofblocks_workqueue);
899 	destroy_workqueue(mp->m_log_workqueue);
900 	destroy_workqueue(mp->m_reclaim_workqueue);
901 	destroy_workqueue(mp->m_cil_workqueue);
902 	destroy_workqueue(mp->m_data_workqueue);
903 	destroy_workqueue(mp->m_unwritten_workqueue);
904 	destroy_workqueue(mp->m_buf_workqueue);
905 }
906 
907 /*
908  * Flush all dirty data to disk. Must not be called while holding an XFS_ILOCK
909  * or a page lock. We use sync_inodes_sb() here to ensure we block while waiting
910  * for IO to complete so that we effectively throttle multiple callers to the
911  * rate at which IO is completing.
912  */
913 void
914 xfs_flush_inodes(
915 	struct xfs_mount	*mp)
916 {
917 	struct super_block	*sb = mp->m_super;
918 
919 	if (down_read_trylock(&sb->s_umount)) {
920 		sync_inodes_sb(sb);
921 		up_read(&sb->s_umount);
922 	}
923 }
924 
925 /* Catch misguided souls that try to use this interface on XFS */
926 STATIC struct inode *
927 xfs_fs_alloc_inode(
928 	struct super_block	*sb)
929 {
930 	BUG();
931 	return NULL;
932 }
933 
934 /*
935  * Now that the generic code is guaranteed not to be accessing
936  * the linux inode, we can reclaim the inode.
937  */
938 STATIC void
939 xfs_fs_destroy_inode(
940 	struct inode		*inode)
941 {
942 	struct xfs_inode	*ip = XFS_I(inode);
943 
944 	trace_xfs_destroy_inode(ip);
945 
946 	XFS_STATS_INC(vn_reclaim);
947 
948 	ASSERT(XFS_FORCED_SHUTDOWN(ip->i_mount) || ip->i_delayed_blks == 0);
949 
950 	/*
951 	 * We should never get here with one of the reclaim flags already set.
952 	 */
953 	ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIMABLE));
954 	ASSERT_ALWAYS(!xfs_iflags_test(ip, XFS_IRECLAIM));
955 
956 	/*
957 	 * We always use background reclaim here because even if the
958 	 * inode is clean, it still may be under IO and hence we have
959 	 * to take the flush lock. The background reclaim path handles
960 	 * this more efficiently than we can here, so simply let background
961 	 * reclaim tear down all inodes.
962 	 */
963 	xfs_inode_set_reclaim_tag(ip);
964 }
965 
966 /*
967  * Slab object creation initialisation for the XFS inode.
968  * This covers only the idempotent fields in the XFS inode;
969  * all other fields need to be initialised on allocation
970  * from the slab. This avoids the need to repeatedly initialise
971  * fields in the xfs inode that left in the initialise state
972  * when freeing the inode.
973  */
974 STATIC void
975 xfs_fs_inode_init_once(
976 	void			*inode)
977 {
978 	struct xfs_inode	*ip = inode;
979 
980 	memset(ip, 0, sizeof(struct xfs_inode));
981 
982 	/* vfs inode */
983 	inode_init_once(VFS_I(ip));
984 
985 	/* xfs inode */
986 	atomic_set(&ip->i_pincount, 0);
987 	spin_lock_init(&ip->i_flags_lock);
988 
989 	mrlock_init(&ip->i_lock, MRLOCK_ALLOW_EQUAL_PRI|MRLOCK_BARRIER,
990 		     "xfsino", ip->i_ino);
991 }
992 
993 STATIC void
994 xfs_fs_evict_inode(
995 	struct inode		*inode)
996 {
997 	xfs_inode_t		*ip = XFS_I(inode);
998 
999 	ASSERT(!rwsem_is_locked(&ip->i_iolock.mr_lock));
1000 
1001 	trace_xfs_evict_inode(ip);
1002 
1003 	truncate_inode_pages_final(&inode->i_data);
1004 	clear_inode(inode);
1005 	XFS_STATS_INC(vn_rele);
1006 	XFS_STATS_INC(vn_remove);
1007 
1008 	xfs_inactive(ip);
1009 }
1010 
1011 /*
1012  * We do an unlocked check for XFS_IDONTCACHE here because we are already
1013  * serialised against cache hits here via the inode->i_lock and igrab() in
1014  * xfs_iget_cache_hit(). Hence a lookup that might clear this flag will not be
1015  * racing with us, and it avoids needing to grab a spinlock here for every inode
1016  * we drop the final reference on.
1017  */
1018 STATIC int
1019 xfs_fs_drop_inode(
1020 	struct inode		*inode)
1021 {
1022 	struct xfs_inode	*ip = XFS_I(inode);
1023 
1024 	return generic_drop_inode(inode) || (ip->i_flags & XFS_IDONTCACHE);
1025 }
1026 
1027 STATIC void
1028 xfs_free_fsname(
1029 	struct xfs_mount	*mp)
1030 {
1031 	kfree(mp->m_fsname);
1032 	kfree(mp->m_rtname);
1033 	kfree(mp->m_logname);
1034 }
1035 
1036 STATIC void
1037 xfs_fs_put_super(
1038 	struct super_block	*sb)
1039 {
1040 	struct xfs_mount	*mp = XFS_M(sb);
1041 
1042 	xfs_filestream_unmount(mp);
1043 	xfs_unmountfs(mp);
1044 
1045 	xfs_freesb(mp);
1046 	xfs_icsb_destroy_counters(mp);
1047 	xfs_destroy_mount_workqueues(mp);
1048 	xfs_close_devices(mp);
1049 	xfs_free_fsname(mp);
1050 	kfree(mp);
1051 }
1052 
1053 STATIC int
1054 xfs_fs_sync_fs(
1055 	struct super_block	*sb,
1056 	int			wait)
1057 {
1058 	struct xfs_mount	*mp = XFS_M(sb);
1059 
1060 	/*
1061 	 * Doing anything during the async pass would be counterproductive.
1062 	 */
1063 	if (!wait)
1064 		return 0;
1065 
1066 	xfs_log_force(mp, XFS_LOG_SYNC);
1067 	if (laptop_mode) {
1068 		/*
1069 		 * The disk must be active because we're syncing.
1070 		 * We schedule log work now (now that the disk is
1071 		 * active) instead of later (when it might not be).
1072 		 */
1073 		flush_delayed_work(&mp->m_log->l_work);
1074 	}
1075 
1076 	return 0;
1077 }
1078 
1079 STATIC int
1080 xfs_fs_statfs(
1081 	struct dentry		*dentry,
1082 	struct kstatfs		*statp)
1083 {
1084 	struct xfs_mount	*mp = XFS_M(dentry->d_sb);
1085 	xfs_sb_t		*sbp = &mp->m_sb;
1086 	struct xfs_inode	*ip = XFS_I(dentry->d_inode);
1087 	__uint64_t		fakeinos, id;
1088 	xfs_extlen_t		lsize;
1089 	__int64_t		ffree;
1090 
1091 	statp->f_type = XFS_SB_MAGIC;
1092 	statp->f_namelen = MAXNAMELEN - 1;
1093 
1094 	id = huge_encode_dev(mp->m_ddev_targp->bt_dev);
1095 	statp->f_fsid.val[0] = (u32)id;
1096 	statp->f_fsid.val[1] = (u32)(id >> 32);
1097 
1098 	xfs_icsb_sync_counters(mp, XFS_ICSB_LAZY_COUNT);
1099 
1100 	spin_lock(&mp->m_sb_lock);
1101 	statp->f_bsize = sbp->sb_blocksize;
1102 	lsize = sbp->sb_logstart ? sbp->sb_logblocks : 0;
1103 	statp->f_blocks = sbp->sb_dblocks - lsize;
1104 	statp->f_bfree = statp->f_bavail =
1105 				sbp->sb_fdblocks - XFS_ALLOC_SET_ASIDE(mp);
1106 	fakeinos = statp->f_bfree << sbp->sb_inopblog;
1107 	statp->f_files =
1108 	    MIN(sbp->sb_icount + fakeinos, (__uint64_t)XFS_MAXINUMBER);
1109 	if (mp->m_maxicount)
1110 		statp->f_files = min_t(typeof(statp->f_files),
1111 					statp->f_files,
1112 					mp->m_maxicount);
1113 
1114 	/* make sure statp->f_ffree does not underflow */
1115 	ffree = statp->f_files - (sbp->sb_icount - sbp->sb_ifree);
1116 	statp->f_ffree = max_t(__int64_t, ffree, 0);
1117 
1118 	spin_unlock(&mp->m_sb_lock);
1119 
1120 	if ((ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT) &&
1121 	    ((mp->m_qflags & (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))) ==
1122 			      (XFS_PQUOTA_ACCT|XFS_PQUOTA_ENFD))
1123 		xfs_qm_statvfs(ip, statp);
1124 	return 0;
1125 }
1126 
1127 STATIC void
1128 xfs_save_resvblks(struct xfs_mount *mp)
1129 {
1130 	__uint64_t resblks = 0;
1131 
1132 	mp->m_resblks_save = mp->m_resblks;
1133 	xfs_reserve_blocks(mp, &resblks, NULL);
1134 }
1135 
1136 STATIC void
1137 xfs_restore_resvblks(struct xfs_mount *mp)
1138 {
1139 	__uint64_t resblks;
1140 
1141 	if (mp->m_resblks_save) {
1142 		resblks = mp->m_resblks_save;
1143 		mp->m_resblks_save = 0;
1144 	} else
1145 		resblks = xfs_default_resblks(mp);
1146 
1147 	xfs_reserve_blocks(mp, &resblks, NULL);
1148 }
1149 
1150 /*
1151  * Trigger writeback of all the dirty metadata in the file system.
1152  *
1153  * This ensures that the metadata is written to their location on disk rather
1154  * than just existing in transactions in the log. This means after a quiesce
1155  * there is no log replay required to write the inodes to disk - this is the
1156  * primary difference between a sync and a quiesce.
1157  *
1158  * Note: xfs_log_quiesce() stops background log work - the callers must ensure
1159  * it is started again when appropriate.
1160  */
1161 static void
1162 xfs_quiesce_attr(
1163 	struct xfs_mount	*mp)
1164 {
1165 	int	error = 0;
1166 
1167 	/* wait for all modifications to complete */
1168 	while (atomic_read(&mp->m_active_trans) > 0)
1169 		delay(100);
1170 
1171 	/* force the log to unpin objects from the now complete transactions */
1172 	xfs_log_force(mp, XFS_LOG_SYNC);
1173 
1174 	/* reclaim inodes to do any IO before the freeze completes */
1175 	xfs_reclaim_inodes(mp, 0);
1176 	xfs_reclaim_inodes(mp, SYNC_WAIT);
1177 
1178 	/* Push the superblock and write an unmount record */
1179 	error = xfs_log_sbcount(mp);
1180 	if (error)
1181 		xfs_warn(mp, "xfs_attr_quiesce: failed to log sb changes. "
1182 				"Frozen image may not be consistent.");
1183 	/*
1184 	 * Just warn here till VFS can correctly support
1185 	 * read-only remount without racing.
1186 	 */
1187 	WARN_ON(atomic_read(&mp->m_active_trans) != 0);
1188 
1189 	xfs_log_quiesce(mp);
1190 }
1191 
1192 STATIC int
1193 xfs_fs_remount(
1194 	struct super_block	*sb,
1195 	int			*flags,
1196 	char			*options)
1197 {
1198 	struct xfs_mount	*mp = XFS_M(sb);
1199 	xfs_sb_t		*sbp = &mp->m_sb;
1200 	substring_t		args[MAX_OPT_ARGS];
1201 	char			*p;
1202 	int			error;
1203 
1204 	sync_filesystem(sb);
1205 	while ((p = strsep(&options, ",")) != NULL) {
1206 		int token;
1207 
1208 		if (!*p)
1209 			continue;
1210 
1211 		token = match_token(p, tokens, args);
1212 		switch (token) {
1213 		case Opt_barrier:
1214 			mp->m_flags |= XFS_MOUNT_BARRIER;
1215 			break;
1216 		case Opt_nobarrier:
1217 			mp->m_flags &= ~XFS_MOUNT_BARRIER;
1218 			break;
1219 		case Opt_inode64:
1220 			mp->m_maxagi = xfs_set_inode64(mp, sbp->sb_agcount);
1221 			break;
1222 		case Opt_inode32:
1223 			mp->m_maxagi = xfs_set_inode32(mp, sbp->sb_agcount);
1224 			break;
1225 		default:
1226 			/*
1227 			 * Logically we would return an error here to prevent
1228 			 * users from believing they might have changed
1229 			 * mount options using remount which can't be changed.
1230 			 *
1231 			 * But unfortunately mount(8) adds all options from
1232 			 * mtab and fstab to the mount arguments in some cases
1233 			 * so we can't blindly reject options, but have to
1234 			 * check for each specified option if it actually
1235 			 * differs from the currently set option and only
1236 			 * reject it if that's the case.
1237 			 *
1238 			 * Until that is implemented we return success for
1239 			 * every remount request, and silently ignore all
1240 			 * options that we can't actually change.
1241 			 */
1242 #if 0
1243 			xfs_info(mp,
1244 		"mount option \"%s\" not supported for remount", p);
1245 			return -EINVAL;
1246 #else
1247 			break;
1248 #endif
1249 		}
1250 	}
1251 
1252 	/* ro -> rw */
1253 	if ((mp->m_flags & XFS_MOUNT_RDONLY) && !(*flags & MS_RDONLY)) {
1254 		mp->m_flags &= ~XFS_MOUNT_RDONLY;
1255 
1256 		/*
1257 		 * If this is the first remount to writeable state we
1258 		 * might have some superblock changes to update.
1259 		 */
1260 		if (mp->m_update_flags) {
1261 			error = xfs_mount_log_sb(mp, mp->m_update_flags);
1262 			if (error) {
1263 				xfs_warn(mp, "failed to write sb changes");
1264 				return error;
1265 			}
1266 			mp->m_update_flags = 0;
1267 		}
1268 
1269 		/*
1270 		 * Fill out the reserve pool if it is empty. Use the stashed
1271 		 * value if it is non-zero, otherwise go with the default.
1272 		 */
1273 		xfs_restore_resvblks(mp);
1274 		xfs_log_work_queue(mp);
1275 	}
1276 
1277 	/* rw -> ro */
1278 	if (!(mp->m_flags & XFS_MOUNT_RDONLY) && (*flags & MS_RDONLY)) {
1279 		/*
1280 		 * Before we sync the metadata, we need to free up the reserve
1281 		 * block pool so that the used block count in the superblock on
1282 		 * disk is correct at the end of the remount. Stash the current
1283 		 * reserve pool size so that if we get remounted rw, we can
1284 		 * return it to the same size.
1285 		 */
1286 		xfs_save_resvblks(mp);
1287 		xfs_quiesce_attr(mp);
1288 		mp->m_flags |= XFS_MOUNT_RDONLY;
1289 	}
1290 
1291 	return 0;
1292 }
1293 
1294 /*
1295  * Second stage of a freeze. The data is already frozen so we only
1296  * need to take care of the metadata. Once that's done write a dummy
1297  * record to dirty the log in case of a crash while frozen.
1298  */
1299 STATIC int
1300 xfs_fs_freeze(
1301 	struct super_block	*sb)
1302 {
1303 	struct xfs_mount	*mp = XFS_M(sb);
1304 
1305 	xfs_save_resvblks(mp);
1306 	xfs_quiesce_attr(mp);
1307 	return xfs_fs_log_dummy(mp);
1308 }
1309 
1310 STATIC int
1311 xfs_fs_unfreeze(
1312 	struct super_block	*sb)
1313 {
1314 	struct xfs_mount	*mp = XFS_M(sb);
1315 
1316 	xfs_restore_resvblks(mp);
1317 	xfs_log_work_queue(mp);
1318 	return 0;
1319 }
1320 
1321 STATIC int
1322 xfs_fs_show_options(
1323 	struct seq_file		*m,
1324 	struct dentry		*root)
1325 {
1326 	return xfs_showargs(XFS_M(root->d_sb), m);
1327 }
1328 
1329 /*
1330  * This function fills in xfs_mount_t fields based on mount args.
1331  * Note: the superblock _has_ now been read in.
1332  */
1333 STATIC int
1334 xfs_finish_flags(
1335 	struct xfs_mount	*mp)
1336 {
1337 	int			ronly = (mp->m_flags & XFS_MOUNT_RDONLY);
1338 
1339 	/* Fail a mount where the logbuf is smaller than the log stripe */
1340 	if (xfs_sb_version_haslogv2(&mp->m_sb)) {
1341 		if (mp->m_logbsize <= 0 &&
1342 		    mp->m_sb.sb_logsunit > XLOG_BIG_RECORD_BSIZE) {
1343 			mp->m_logbsize = mp->m_sb.sb_logsunit;
1344 		} else if (mp->m_logbsize > 0 &&
1345 			   mp->m_logbsize < mp->m_sb.sb_logsunit) {
1346 			xfs_warn(mp,
1347 		"logbuf size must be greater than or equal to log stripe size");
1348 			return -EINVAL;
1349 		}
1350 	} else {
1351 		/* Fail a mount if the logbuf is larger than 32K */
1352 		if (mp->m_logbsize > XLOG_BIG_RECORD_BSIZE) {
1353 			xfs_warn(mp,
1354 		"logbuf size for version 1 logs must be 16K or 32K");
1355 			return -EINVAL;
1356 		}
1357 	}
1358 
1359 	/*
1360 	 * V5 filesystems always use attr2 format for attributes.
1361 	 */
1362 	if (xfs_sb_version_hascrc(&mp->m_sb) &&
1363 	    (mp->m_flags & XFS_MOUNT_NOATTR2)) {
1364 		xfs_warn(mp,
1365 "Cannot mount a V5 filesystem as %s. %s is always enabled for V5 filesystems.",
1366 			MNTOPT_NOATTR2, MNTOPT_ATTR2);
1367 		return -EINVAL;
1368 	}
1369 
1370 	/*
1371 	 * mkfs'ed attr2 will turn on attr2 mount unless explicitly
1372 	 * told by noattr2 to turn it off
1373 	 */
1374 	if (xfs_sb_version_hasattr2(&mp->m_sb) &&
1375 	    !(mp->m_flags & XFS_MOUNT_NOATTR2))
1376 		mp->m_flags |= XFS_MOUNT_ATTR2;
1377 
1378 	/*
1379 	 * prohibit r/w mounts of read-only filesystems
1380 	 */
1381 	if ((mp->m_sb.sb_flags & XFS_SBF_READONLY) && !ronly) {
1382 		xfs_warn(mp,
1383 			"cannot mount a read-only filesystem as read-write");
1384 		return -EROFS;
1385 	}
1386 
1387 	if ((mp->m_qflags & (XFS_GQUOTA_ACCT | XFS_GQUOTA_ACTIVE)) &&
1388 	    (mp->m_qflags & (XFS_PQUOTA_ACCT | XFS_PQUOTA_ACTIVE)) &&
1389 	    !xfs_sb_version_has_pquotino(&mp->m_sb)) {
1390 		xfs_warn(mp,
1391 		  "Super block does not support project and group quota together");
1392 		return -EINVAL;
1393 	}
1394 
1395 	return 0;
1396 }
1397 
1398 STATIC int
1399 xfs_fs_fill_super(
1400 	struct super_block	*sb,
1401 	void			*data,
1402 	int			silent)
1403 {
1404 	struct inode		*root;
1405 	struct xfs_mount	*mp = NULL;
1406 	int			flags = 0, error = -ENOMEM;
1407 
1408 	mp = kzalloc(sizeof(struct xfs_mount), GFP_KERNEL);
1409 	if (!mp)
1410 		goto out;
1411 
1412 	spin_lock_init(&mp->m_sb_lock);
1413 	mutex_init(&mp->m_growlock);
1414 	atomic_set(&mp->m_active_trans, 0);
1415 	INIT_DELAYED_WORK(&mp->m_reclaim_work, xfs_reclaim_worker);
1416 	INIT_DELAYED_WORK(&mp->m_eofblocks_work, xfs_eofblocks_worker);
1417 	mp->m_kobj.kobject.kset = xfs_kset;
1418 
1419 	mp->m_super = sb;
1420 	sb->s_fs_info = mp;
1421 
1422 	error = xfs_parseargs(mp, (char *)data);
1423 	if (error)
1424 		goto out_free_fsname;
1425 
1426 	sb_min_blocksize(sb, BBSIZE);
1427 	sb->s_xattr = xfs_xattr_handlers;
1428 	sb->s_export_op = &xfs_export_operations;
1429 #ifdef CONFIG_XFS_QUOTA
1430 	sb->s_qcop = &xfs_quotactl_operations;
1431 	sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
1432 #endif
1433 	sb->s_op = &xfs_super_operations;
1434 
1435 	if (silent)
1436 		flags |= XFS_MFSI_QUIET;
1437 
1438 	error = xfs_open_devices(mp);
1439 	if (error)
1440 		goto out_free_fsname;
1441 
1442 	error = xfs_init_mount_workqueues(mp);
1443 	if (error)
1444 		goto out_close_devices;
1445 
1446 	error = xfs_icsb_init_counters(mp);
1447 	if (error)
1448 		goto out_destroy_workqueues;
1449 
1450 	error = xfs_readsb(mp, flags);
1451 	if (error)
1452 		goto out_destroy_counters;
1453 
1454 	error = xfs_finish_flags(mp);
1455 	if (error)
1456 		goto out_free_sb;
1457 
1458 	error = xfs_setup_devices(mp);
1459 	if (error)
1460 		goto out_free_sb;
1461 
1462 	error = xfs_filestream_mount(mp);
1463 	if (error)
1464 		goto out_free_sb;
1465 
1466 	/*
1467 	 * we must configure the block size in the superblock before we run the
1468 	 * full mount process as the mount process can lookup and cache inodes.
1469 	 */
1470 	sb->s_magic = XFS_SB_MAGIC;
1471 	sb->s_blocksize = mp->m_sb.sb_blocksize;
1472 	sb->s_blocksize_bits = ffs(sb->s_blocksize) - 1;
1473 	sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
1474 	sb->s_max_links = XFS_MAXLINK;
1475 	sb->s_time_gran = 1;
1476 	set_posix_acl_flag(sb);
1477 
1478 	/* version 5 superblocks support inode version counters. */
1479 	if (XFS_SB_VERSION_NUM(&mp->m_sb) == XFS_SB_VERSION_5)
1480 		sb->s_flags |= MS_I_VERSION;
1481 
1482 	error = xfs_mountfs(mp);
1483 	if (error)
1484 		goto out_filestream_unmount;
1485 
1486 	root = igrab(VFS_I(mp->m_rootip));
1487 	if (!root) {
1488 		error = -ENOENT;
1489 		goto out_unmount;
1490 	}
1491 	sb->s_root = d_make_root(root);
1492 	if (!sb->s_root) {
1493 		error = -ENOMEM;
1494 		goto out_unmount;
1495 	}
1496 
1497 	return 0;
1498 
1499  out_filestream_unmount:
1500 	xfs_filestream_unmount(mp);
1501  out_free_sb:
1502 	xfs_freesb(mp);
1503  out_destroy_counters:
1504 	xfs_icsb_destroy_counters(mp);
1505 out_destroy_workqueues:
1506 	xfs_destroy_mount_workqueues(mp);
1507  out_close_devices:
1508 	xfs_close_devices(mp);
1509  out_free_fsname:
1510 	xfs_free_fsname(mp);
1511 	kfree(mp);
1512  out:
1513 	return error;
1514 
1515  out_unmount:
1516 	xfs_filestream_unmount(mp);
1517 	xfs_unmountfs(mp);
1518 	goto out_free_sb;
1519 }
1520 
1521 STATIC struct dentry *
1522 xfs_fs_mount(
1523 	struct file_system_type	*fs_type,
1524 	int			flags,
1525 	const char		*dev_name,
1526 	void			*data)
1527 {
1528 	return mount_bdev(fs_type, flags, dev_name, data, xfs_fs_fill_super);
1529 }
1530 
1531 static long
1532 xfs_fs_nr_cached_objects(
1533 	struct super_block	*sb,
1534 	int			nid)
1535 {
1536 	return xfs_reclaim_inodes_count(XFS_M(sb));
1537 }
1538 
1539 static long
1540 xfs_fs_free_cached_objects(
1541 	struct super_block	*sb,
1542 	long			nr_to_scan,
1543 	int			nid)
1544 {
1545 	return xfs_reclaim_inodes_nr(XFS_M(sb), nr_to_scan);
1546 }
1547 
1548 static const struct super_operations xfs_super_operations = {
1549 	.alloc_inode		= xfs_fs_alloc_inode,
1550 	.destroy_inode		= xfs_fs_destroy_inode,
1551 	.evict_inode		= xfs_fs_evict_inode,
1552 	.drop_inode		= xfs_fs_drop_inode,
1553 	.put_super		= xfs_fs_put_super,
1554 	.sync_fs		= xfs_fs_sync_fs,
1555 	.freeze_fs		= xfs_fs_freeze,
1556 	.unfreeze_fs		= xfs_fs_unfreeze,
1557 	.statfs			= xfs_fs_statfs,
1558 	.remount_fs		= xfs_fs_remount,
1559 	.show_options		= xfs_fs_show_options,
1560 	.nr_cached_objects	= xfs_fs_nr_cached_objects,
1561 	.free_cached_objects	= xfs_fs_free_cached_objects,
1562 };
1563 
1564 static struct file_system_type xfs_fs_type = {
1565 	.owner			= THIS_MODULE,
1566 	.name			= "xfs",
1567 	.mount			= xfs_fs_mount,
1568 	.kill_sb		= kill_block_super,
1569 	.fs_flags		= FS_REQUIRES_DEV,
1570 };
1571 MODULE_ALIAS_FS("xfs");
1572 
1573 STATIC int __init
1574 xfs_init_zones(void)
1575 {
1576 
1577 	xfs_ioend_zone = kmem_zone_init(sizeof(xfs_ioend_t), "xfs_ioend");
1578 	if (!xfs_ioend_zone)
1579 		goto out;
1580 
1581 	xfs_ioend_pool = mempool_create_slab_pool(4 * MAX_BUF_PER_PAGE,
1582 						  xfs_ioend_zone);
1583 	if (!xfs_ioend_pool)
1584 		goto out_destroy_ioend_zone;
1585 
1586 	xfs_log_ticket_zone = kmem_zone_init(sizeof(xlog_ticket_t),
1587 						"xfs_log_ticket");
1588 	if (!xfs_log_ticket_zone)
1589 		goto out_destroy_ioend_pool;
1590 
1591 	xfs_bmap_free_item_zone = kmem_zone_init(sizeof(xfs_bmap_free_item_t),
1592 						"xfs_bmap_free_item");
1593 	if (!xfs_bmap_free_item_zone)
1594 		goto out_destroy_log_ticket_zone;
1595 
1596 	xfs_btree_cur_zone = kmem_zone_init(sizeof(xfs_btree_cur_t),
1597 						"xfs_btree_cur");
1598 	if (!xfs_btree_cur_zone)
1599 		goto out_destroy_bmap_free_item_zone;
1600 
1601 	xfs_da_state_zone = kmem_zone_init(sizeof(xfs_da_state_t),
1602 						"xfs_da_state");
1603 	if (!xfs_da_state_zone)
1604 		goto out_destroy_btree_cur_zone;
1605 
1606 	xfs_ifork_zone = kmem_zone_init(sizeof(xfs_ifork_t), "xfs_ifork");
1607 	if (!xfs_ifork_zone)
1608 		goto out_destroy_da_state_zone;
1609 
1610 	xfs_trans_zone = kmem_zone_init(sizeof(xfs_trans_t), "xfs_trans");
1611 	if (!xfs_trans_zone)
1612 		goto out_destroy_ifork_zone;
1613 
1614 	xfs_log_item_desc_zone =
1615 		kmem_zone_init(sizeof(struct xfs_log_item_desc),
1616 			       "xfs_log_item_desc");
1617 	if (!xfs_log_item_desc_zone)
1618 		goto out_destroy_trans_zone;
1619 
1620 	/*
1621 	 * The size of the zone allocated buf log item is the maximum
1622 	 * size possible under XFS.  This wastes a little bit of memory,
1623 	 * but it is much faster.
1624 	 */
1625 	xfs_buf_item_zone = kmem_zone_init(sizeof(struct xfs_buf_log_item),
1626 					   "xfs_buf_item");
1627 	if (!xfs_buf_item_zone)
1628 		goto out_destroy_log_item_desc_zone;
1629 
1630 	xfs_efd_zone = kmem_zone_init((sizeof(xfs_efd_log_item_t) +
1631 			((XFS_EFD_MAX_FAST_EXTENTS - 1) *
1632 				 sizeof(xfs_extent_t))), "xfs_efd_item");
1633 	if (!xfs_efd_zone)
1634 		goto out_destroy_buf_item_zone;
1635 
1636 	xfs_efi_zone = kmem_zone_init((sizeof(xfs_efi_log_item_t) +
1637 			((XFS_EFI_MAX_FAST_EXTENTS - 1) *
1638 				sizeof(xfs_extent_t))), "xfs_efi_item");
1639 	if (!xfs_efi_zone)
1640 		goto out_destroy_efd_zone;
1641 
1642 	xfs_inode_zone =
1643 		kmem_zone_init_flags(sizeof(xfs_inode_t), "xfs_inode",
1644 			KM_ZONE_HWALIGN | KM_ZONE_RECLAIM | KM_ZONE_SPREAD,
1645 			xfs_fs_inode_init_once);
1646 	if (!xfs_inode_zone)
1647 		goto out_destroy_efi_zone;
1648 
1649 	xfs_ili_zone =
1650 		kmem_zone_init_flags(sizeof(xfs_inode_log_item_t), "xfs_ili",
1651 					KM_ZONE_SPREAD, NULL);
1652 	if (!xfs_ili_zone)
1653 		goto out_destroy_inode_zone;
1654 	xfs_icreate_zone = kmem_zone_init(sizeof(struct xfs_icreate_item),
1655 					"xfs_icr");
1656 	if (!xfs_icreate_zone)
1657 		goto out_destroy_ili_zone;
1658 
1659 	return 0;
1660 
1661  out_destroy_ili_zone:
1662 	kmem_zone_destroy(xfs_ili_zone);
1663  out_destroy_inode_zone:
1664 	kmem_zone_destroy(xfs_inode_zone);
1665  out_destroy_efi_zone:
1666 	kmem_zone_destroy(xfs_efi_zone);
1667  out_destroy_efd_zone:
1668 	kmem_zone_destroy(xfs_efd_zone);
1669  out_destroy_buf_item_zone:
1670 	kmem_zone_destroy(xfs_buf_item_zone);
1671  out_destroy_log_item_desc_zone:
1672 	kmem_zone_destroy(xfs_log_item_desc_zone);
1673  out_destroy_trans_zone:
1674 	kmem_zone_destroy(xfs_trans_zone);
1675  out_destroy_ifork_zone:
1676 	kmem_zone_destroy(xfs_ifork_zone);
1677  out_destroy_da_state_zone:
1678 	kmem_zone_destroy(xfs_da_state_zone);
1679  out_destroy_btree_cur_zone:
1680 	kmem_zone_destroy(xfs_btree_cur_zone);
1681  out_destroy_bmap_free_item_zone:
1682 	kmem_zone_destroy(xfs_bmap_free_item_zone);
1683  out_destroy_log_ticket_zone:
1684 	kmem_zone_destroy(xfs_log_ticket_zone);
1685  out_destroy_ioend_pool:
1686 	mempool_destroy(xfs_ioend_pool);
1687  out_destroy_ioend_zone:
1688 	kmem_zone_destroy(xfs_ioend_zone);
1689  out:
1690 	return -ENOMEM;
1691 }
1692 
1693 STATIC void
1694 xfs_destroy_zones(void)
1695 {
1696 	/*
1697 	 * Make sure all delayed rcu free are flushed before we
1698 	 * destroy caches.
1699 	 */
1700 	rcu_barrier();
1701 	kmem_zone_destroy(xfs_icreate_zone);
1702 	kmem_zone_destroy(xfs_ili_zone);
1703 	kmem_zone_destroy(xfs_inode_zone);
1704 	kmem_zone_destroy(xfs_efi_zone);
1705 	kmem_zone_destroy(xfs_efd_zone);
1706 	kmem_zone_destroy(xfs_buf_item_zone);
1707 	kmem_zone_destroy(xfs_log_item_desc_zone);
1708 	kmem_zone_destroy(xfs_trans_zone);
1709 	kmem_zone_destroy(xfs_ifork_zone);
1710 	kmem_zone_destroy(xfs_da_state_zone);
1711 	kmem_zone_destroy(xfs_btree_cur_zone);
1712 	kmem_zone_destroy(xfs_bmap_free_item_zone);
1713 	kmem_zone_destroy(xfs_log_ticket_zone);
1714 	mempool_destroy(xfs_ioend_pool);
1715 	kmem_zone_destroy(xfs_ioend_zone);
1716 
1717 }
1718 
1719 STATIC int __init
1720 xfs_init_workqueues(void)
1721 {
1722 	/*
1723 	 * The allocation workqueue can be used in memory reclaim situations
1724 	 * (writepage path), and parallelism is only limited by the number of
1725 	 * AGs in all the filesystems mounted. Hence use the default large
1726 	 * max_active value for this workqueue.
1727 	 */
1728 	xfs_alloc_wq = alloc_workqueue("xfsalloc",
1729 			WQ_MEM_RECLAIM|WQ_FREEZABLE, 0);
1730 	if (!xfs_alloc_wq)
1731 		return -ENOMEM;
1732 
1733 	return 0;
1734 }
1735 
1736 STATIC void
1737 xfs_destroy_workqueues(void)
1738 {
1739 	destroy_workqueue(xfs_alloc_wq);
1740 }
1741 
1742 STATIC int __init
1743 init_xfs_fs(void)
1744 {
1745 	int			error;
1746 
1747 	printk(KERN_INFO XFS_VERSION_STRING " with "
1748 			 XFS_BUILD_OPTIONS " enabled\n");
1749 
1750 	xfs_dir_startup();
1751 
1752 	error = xfs_init_zones();
1753 	if (error)
1754 		goto out;
1755 
1756 	error = xfs_init_workqueues();
1757 	if (error)
1758 		goto out_destroy_zones;
1759 
1760 	error = xfs_mru_cache_init();
1761 	if (error)
1762 		goto out_destroy_wq;
1763 
1764 	error = xfs_buf_init();
1765 	if (error)
1766 		goto out_mru_cache_uninit;
1767 
1768 	error = xfs_init_procfs();
1769 	if (error)
1770 		goto out_buf_terminate;
1771 
1772 	error = xfs_sysctl_register();
1773 	if (error)
1774 		goto out_cleanup_procfs;
1775 
1776 	xfs_kset = kset_create_and_add("xfs", NULL, fs_kobj);
1777 	if (!xfs_kset) {
1778 		error = -ENOMEM;
1779 		goto out_sysctl_unregister;;
1780 	}
1781 
1782 #ifdef DEBUG
1783 	xfs_dbg_kobj.kobject.kset = xfs_kset;
1784 	error = xfs_sysfs_init(&xfs_dbg_kobj, &xfs_dbg_ktype, NULL, "debug");
1785 	if (error)
1786 		goto out_kset_unregister;
1787 #endif
1788 
1789 	error = xfs_qm_init();
1790 	if (error)
1791 		goto out_remove_kobj;
1792 
1793 	error = register_filesystem(&xfs_fs_type);
1794 	if (error)
1795 		goto out_qm_exit;
1796 	return 0;
1797 
1798  out_qm_exit:
1799 	xfs_qm_exit();
1800  out_remove_kobj:
1801 #ifdef DEBUG
1802 	xfs_sysfs_del(&xfs_dbg_kobj);
1803  out_kset_unregister:
1804 #endif
1805 	kset_unregister(xfs_kset);
1806  out_sysctl_unregister:
1807 	xfs_sysctl_unregister();
1808  out_cleanup_procfs:
1809 	xfs_cleanup_procfs();
1810  out_buf_terminate:
1811 	xfs_buf_terminate();
1812  out_mru_cache_uninit:
1813 	xfs_mru_cache_uninit();
1814  out_destroy_wq:
1815 	xfs_destroy_workqueues();
1816  out_destroy_zones:
1817 	xfs_destroy_zones();
1818  out:
1819 	return error;
1820 }
1821 
1822 STATIC void __exit
1823 exit_xfs_fs(void)
1824 {
1825 	xfs_qm_exit();
1826 	unregister_filesystem(&xfs_fs_type);
1827 #ifdef DEBUG
1828 	xfs_sysfs_del(&xfs_dbg_kobj);
1829 #endif
1830 	kset_unregister(xfs_kset);
1831 	xfs_sysctl_unregister();
1832 	xfs_cleanup_procfs();
1833 	xfs_buf_terminate();
1834 	xfs_mru_cache_uninit();
1835 	xfs_destroy_workqueues();
1836 	xfs_destroy_zones();
1837 }
1838 
1839 module_init(init_xfs_fs);
1840 module_exit(exit_xfs_fs);
1841 
1842 MODULE_AUTHOR("Silicon Graphics, Inc.");
1843 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
1844 MODULE_LICENSE("GPL");
1845