projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
[XFS] Endianess annotations for various allocator data structures
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / xfs / xfs_mount.c
CommitLineData
1da177e4 1/*
7b718769
NS		 2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
1da177e4 4 *
7b718769
NS		 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
1da177e4
LT		 7 * published by the Free Software Foundation.
8 *
7b718769
NS		 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.
1da177e4 13 *
7b718769
NS		 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
1da177e4 17 */
1da177e4 18#include "xfs.h"
a844f451 19#include "xfs_fs.h"
1da177e4 20#include "xfs_types.h"
a844f451 21#include "xfs_bit.h"
1da177e4 22#include "xfs_log.h"
a844f451 23#include "xfs_inum.h"
1da177e4
LT		 24#include "xfs_trans.h"
25#include "xfs_sb.h"
26#include "xfs_ag.h"
27#include "xfs_dir.h"
28#include "xfs_dir2.h"
29#include "xfs_dmapi.h"
30#include "xfs_mount.h"
1da177e4 31#include "xfs_bmap_btree.h"
a844f451 32#include "xfs_alloc_btree.h"
1da177e4 33#include "xfs_ialloc_btree.h"
1da177e4
LT		 34#include "xfs_dir_sf.h"
35#include "xfs_dir2_sf.h"
a844f451 36#include "xfs_attr_sf.h"
1da177e4
LT		 37#include "xfs_dinode.h"
38#include "xfs_inode.h"
a844f451
NS		 39#include "xfs_btree.h"
40#include "xfs_ialloc.h"
1da177e4
LT		 41#include "xfs_alloc.h"
42#include "xfs_rtalloc.h"
43#include "xfs_bmap.h"
44#include "xfs_error.h"
1da177e4
LT		 45#include "xfs_rw.h"
46#include "xfs_quota.h"
47#include "xfs_fsops.h"
48
49STATIC void xfs_mount_log_sbunit(xfs_mount_t *, __int64_t);
50STATIC int xfs_uuid_mount(xfs_mount_t *);
51STATIC void xfs_uuid_unmount(xfs_mount_t *mp);
ba0f32d4 52STATIC void xfs_unmountfs_wait(xfs_mount_t *);
1da177e4
LT		 53
54static struct {
55 short offset;
56 short type; /* 0 = integer
57 * 1 = binary / string (no translation)
58 */
59} xfs_sb_info[] = {
60 { offsetof(xfs_sb_t, sb_magicnum), 0 },
61 { offsetof(xfs_sb_t, sb_blocksize), 0 },
62 { offsetof(xfs_sb_t, sb_dblocks), 0 },
63 { offsetof(xfs_sb_t, sb_rblocks), 0 },
64 { offsetof(xfs_sb_t, sb_rextents), 0 },
65 { offsetof(xfs_sb_t, sb_uuid), 1 },
66 { offsetof(xfs_sb_t, sb_logstart), 0 },
67 { offsetof(xfs_sb_t, sb_rootino), 0 },
68 { offsetof(xfs_sb_t, sb_rbmino), 0 },
69 { offsetof(xfs_sb_t, sb_rsumino), 0 },
70 { offsetof(xfs_sb_t, sb_rextsize), 0 },
71 { offsetof(xfs_sb_t, sb_agblocks), 0 },
72 { offsetof(xfs_sb_t, sb_agcount), 0 },
73 { offsetof(xfs_sb_t, sb_rbmblocks), 0 },
74 { offsetof(xfs_sb_t, sb_logblocks), 0 },
75 { offsetof(xfs_sb_t, sb_versionnum), 0 },
76 { offsetof(xfs_sb_t, sb_sectsize), 0 },
77 { offsetof(xfs_sb_t, sb_inodesize), 0 },
78 { offsetof(xfs_sb_t, sb_inopblock), 0 },
79 { offsetof(xfs_sb_t, sb_fname[0]), 1 },
80 { offsetof(xfs_sb_t, sb_blocklog), 0 },
81 { offsetof(xfs_sb_t, sb_sectlog), 0 },
82 { offsetof(xfs_sb_t, sb_inodelog), 0 },
83 { offsetof(xfs_sb_t, sb_inopblog), 0 },
84 { offsetof(xfs_sb_t, sb_agblklog), 0 },
85 { offsetof(xfs_sb_t, sb_rextslog), 0 },
86 { offsetof(xfs_sb_t, sb_inprogress), 0 },
87 { offsetof(xfs_sb_t, sb_imax_pct), 0 },
88 { offsetof(xfs_sb_t, sb_icount), 0 },
89 { offsetof(xfs_sb_t, sb_ifree), 0 },
90 { offsetof(xfs_sb_t, sb_fdblocks), 0 },
91 { offsetof(xfs_sb_t, sb_frextents), 0 },
92 { offsetof(xfs_sb_t, sb_uquotino), 0 },
93 { offsetof(xfs_sb_t, sb_gquotino), 0 },
94 { offsetof(xfs_sb_t, sb_qflags), 0 },
95 { offsetof(xfs_sb_t, sb_flags), 0 },
96 { offsetof(xfs_sb_t, sb_shared_vn), 0 },
97 { offsetof(xfs_sb_t, sb_inoalignmt), 0 },
98 { offsetof(xfs_sb_t, sb_unit), 0 },
99 { offsetof(xfs_sb_t, sb_width), 0 },
100 { offsetof(xfs_sb_t, sb_dirblklog), 0 },
101 { offsetof(xfs_sb_t, sb_logsectlog), 0 },
102 { offsetof(xfs_sb_t, sb_logsectsize),0 },
103 { offsetof(xfs_sb_t, sb_logsunit), 0 },
104 { offsetof(xfs_sb_t, sb_features2), 0 },
105 { sizeof(xfs_sb_t), 0 }
106};
107
108/*
109 * Return a pointer to an initialized xfs_mount structure.
110 */
111xfs_mount_t *
112xfs_mount_init(void)
113{
114 xfs_mount_t *mp;
115
116 mp = kmem_zalloc(sizeof(*mp), KM_SLEEP);
117
118 AIL_LOCKINIT(&mp->m_ail_lock, "xfs_ail");
119 spinlock_init(&mp->m_sb_lock, "xfs_sb");
120 mutex_init(&mp->m_ilock, MUTEX_DEFAULT, "xfs_ilock");
121 initnsema(&mp->m_growlock, 1, "xfs_grow");
122 /*
123 * Initialize the AIL.
124 */
125 xfs_trans_ail_init(mp);
126
127 atomic_set(&mp->m_active_trans, 0);
128
129 return mp;
130}
131
132/*
133 * Free up the resources associated with a mount structure. Assume that
134 * the structure was initially zeroed, so we can tell which fields got
135 * initialized.
136 */
137void
138xfs_mount_free(
139 xfs_mount_t *mp,
140 int remove_bhv)
141{
142 if (mp->m_ihash)
143 xfs_ihash_free(mp);
144 if (mp->m_chash)
145 xfs_chash_free(mp);
146
147 if (mp->m_perag) {
148 int agno;
149
150 for (agno = 0; agno < mp->m_maxagi; agno++)
151 if (mp->m_perag[agno].pagb_list)
152 kmem_free(mp->m_perag[agno].pagb_list,
153 sizeof(xfs_perag_busy_t) *
154 XFS_PAGB_NUM_SLOTS);
155 kmem_free(mp->m_perag,
156 sizeof(xfs_perag_t) * mp->m_sb.sb_agcount);
157 }
158
159 AIL_LOCK_DESTROY(&mp->m_ail_lock);
160 spinlock_destroy(&mp->m_sb_lock);
161 mutex_destroy(&mp->m_ilock);
162 freesema(&mp->m_growlock);
163 if (mp->m_quotainfo)
164 XFS_QM_DONE(mp);
165
166 if (mp->m_fsname != NULL)
167 kmem_free(mp->m_fsname, mp->m_fsname_len);
fc1f8c1c
NS		 168 if (mp->m_rtname != NULL)
169 kmem_free(mp->m_rtname, strlen(mp->m_rtname) + 1);
170 if (mp->m_logname != NULL)
171 kmem_free(mp->m_logname, strlen(mp->m_logname) + 1);
1da177e4
LT		 172
173 if (remove_bhv) {
174 struct vfs *vfsp = XFS_MTOVFS(mp);
175
176 bhv_remove_all_vfsops(vfsp, 0);
177 VFS_REMOVEBHV(vfsp, &mp->m_bhv);
178 }
179
180 kmem_free(mp, sizeof(xfs_mount_t));
181}
182
183
184/*
185 * Check the validity of the SB found.
186 */
187STATIC int
188xfs_mount_validate_sb(
189 xfs_mount_t *mp,
190 xfs_sb_t *sbp)
191{
192 /*
193 * If the log device and data device have the
194 * same device number, the log is internal.
195 * Consequently, the sb_logstart should be non-zero. If
196 * we have a zero sb_logstart in this case, we may be trying to mount
197 * a volume filesystem in a non-volume manner.
198 */
199 if (sbp->sb_magicnum != XFS_SB_MAGIC) {
200 cmn_err(CE_WARN, "XFS: bad magic number");
201 return XFS_ERROR(EWRONGFS);
202 }
203
204 if (!XFS_SB_GOOD_VERSION(sbp)) {
205 cmn_err(CE_WARN, "XFS: bad version");
206 return XFS_ERROR(EWRONGFS);
207 }
208
209 if (unlikely(
210 sbp->sb_logstart == 0 && mp->m_logdev_targp == mp->m_ddev_targp)) {
211 cmn_err(CE_WARN,
212 "XFS: filesystem is marked as having an external log; "
213 "specify logdev on the\nmount command line.");
214 XFS_CORRUPTION_ERROR("xfs_mount_validate_sb(1)",
215 XFS_ERRLEVEL_HIGH, mp, sbp);
216 return XFS_ERROR(EFSCORRUPTED);
217 }
218
219 if (unlikely(
220 sbp->sb_logstart != 0 && mp->m_logdev_targp != mp->m_ddev_targp)) {
221 cmn_err(CE_WARN,
222 "XFS: filesystem is marked as having an internal log; "
223 "don't specify logdev on\nthe mount command line.");
224 XFS_CORRUPTION_ERROR("xfs_mount_validate_sb(2)",
225 XFS_ERRLEVEL_HIGH, mp, sbp);
226 return XFS_ERROR(EFSCORRUPTED);
227 }
228
229 /*
230 * More sanity checking. These were stolen directly from
231 * xfs_repair.
232 */
233 if (unlikely(
234 sbp->sb_agcount <= 0 ||
235 sbp->sb_sectsize < XFS_MIN_SECTORSIZE ||
236 sbp->sb_sectsize > XFS_MAX_SECTORSIZE ||
237 sbp->sb_sectlog < XFS_MIN_SECTORSIZE_LOG ||
238 sbp->sb_sectlog > XFS_MAX_SECTORSIZE_LOG ||
239 sbp->sb_blocksize < XFS_MIN_BLOCKSIZE ||
240 sbp->sb_blocksize > XFS_MAX_BLOCKSIZE ||
241 sbp->sb_blocklog < XFS_MIN_BLOCKSIZE_LOG ||
242 sbp->sb_blocklog > XFS_MAX_BLOCKSIZE_LOG ||
243 sbp->sb_inodesize < XFS_DINODE_MIN_SIZE ||
244 sbp->sb_inodesize > XFS_DINODE_MAX_SIZE ||
245 (sbp->sb_rextsize * sbp->sb_blocksize > XFS_MAX_RTEXTSIZE) ||
246 (sbp->sb_rextsize * sbp->sb_blocksize < XFS_MIN_RTEXTSIZE) ||
247 sbp->sb_imax_pct > 100)) {
248 cmn_err(CE_WARN, "XFS: SB sanity check 1 failed");
249 XFS_CORRUPTION_ERROR("xfs_mount_validate_sb(3)",
250 XFS_ERRLEVEL_LOW, mp, sbp);
251 return XFS_ERROR(EFSCORRUPTED);
252 }
253
254 /*
255 * Sanity check AG count, size fields against data size field
256 */
257 if (unlikely(
258 sbp->sb_dblocks == 0 ||
259 sbp->sb_dblocks >
260 (xfs_drfsbno_t)sbp->sb_agcount * sbp->sb_agblocks ||
261 sbp->sb_dblocks < (xfs_drfsbno_t)(sbp->sb_agcount - 1) *
262 sbp->sb_agblocks + XFS_MIN_AG_BLOCKS)) {
263 cmn_err(CE_WARN, "XFS: SB sanity check 2 failed");
264 XFS_ERROR_REPORT("xfs_mount_validate_sb(4)",
265 XFS_ERRLEVEL_LOW, mp);
266 return XFS_ERROR(EFSCORRUPTED);
267 }
268
269 ASSERT(PAGE_SHIFT >= sbp->sb_blocklog);
270 ASSERT(sbp->sb_blocklog >= BBSHIFT);
271
272#if XFS_BIG_BLKNOS /* Limited by ULONG_MAX of page cache index */
273 if (unlikely(
274 (sbp->sb_dblocks >> (PAGE_SHIFT - sbp->sb_blocklog)) > ULONG_MAX ||
275 (sbp->sb_rblocks >> (PAGE_SHIFT - sbp->sb_blocklog)) > ULONG_MAX)) {
276#else /* Limited by UINT_MAX of sectors */
277 if (unlikely(
278 (sbp->sb_dblocks << (sbp->sb_blocklog - BBSHIFT)) > UINT_MAX ||
279 (sbp->sb_rblocks << (sbp->sb_blocklog - BBSHIFT)) > UINT_MAX)) {
280#endif
281 cmn_err(CE_WARN,
282 "XFS: File system is too large to be mounted on this system.");
283 return XFS_ERROR(E2BIG);
284 }
285
286 if (unlikely(sbp->sb_inprogress)) {
287 cmn_err(CE_WARN, "XFS: file system busy");
288 XFS_ERROR_REPORT("xfs_mount_validate_sb(5)",
289 XFS_ERRLEVEL_LOW, mp);
290 return XFS_ERROR(EFSCORRUPTED);
291 }
292
de20614b
NS		 293 /*
294 * Version 1 directory format has never worked on Linux.
295 */
296 if (unlikely(!XFS_SB_VERSION_HASDIRV2(sbp))) {
297 cmn_err(CE_WARN,
298 "XFS: Attempted to mount file system using version 1 directory format");
299 return XFS_ERROR(ENOSYS);
300 }
301
1da177e4
LT		 302 /*
303 * Until this is fixed only page-sized or smaller data blocks work.
304 */
305 if (unlikely(sbp->sb_blocksize > PAGE_SIZE)) {
306 cmn_err(CE_WARN,
307 "XFS: Attempted to mount file system with blocksize %d bytes",
308 sbp->sb_blocksize);
309 cmn_err(CE_WARN,
da1650a5 310 "XFS: Only page-sized (%ld) or less blocksizes currently work.",
1da177e4
LT		 311 PAGE_SIZE);
312 return XFS_ERROR(ENOSYS);
313 }
314
315 return 0;
316}
317
318xfs_agnumber_t
319xfs_initialize_perag(xfs_mount_t *mp, xfs_agnumber_t agcount)
320{
321 xfs_agnumber_t index, max_metadata;
322 xfs_perag_t *pag;
323 xfs_agino_t agino;
324 xfs_ino_t ino;
325 xfs_sb_t *sbp = &mp->m_sb;
326 xfs_ino_t max_inum = XFS_MAXINUMBER_32;
327
328 /* Check to see if the filesystem can overflow 32 bit inodes */
329 agino = XFS_OFFBNO_TO_AGINO(mp, sbp->sb_agblocks - 1, 0);
330 ino = XFS_AGINO_TO_INO(mp, agcount - 1, agino);
331
332 /* Clear the mount flag if no inode can overflow 32 bits
333 * on this filesystem, or if specifically requested..
334 */
335 if ((mp->m_flags & XFS_MOUNT_32BITINOOPT) && ino > max_inum) {
336 mp->m_flags |= XFS_MOUNT_32BITINODES;
337 } else {
338 mp->m_flags &= ~XFS_MOUNT_32BITINODES;
339 }
340
341 /* If we can overflow then setup the ag headers accordingly */
342 if (mp->m_flags & XFS_MOUNT_32BITINODES) {
343 /* Calculate how much should be reserved for inodes to
344 * meet the max inode percentage.
345 */
346 if (mp->m_maxicount) {
347 __uint64_t icount;
348
349 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
350 do_div(icount, 100);
351 icount += sbp->sb_agblocks - 1;
352 do_div(icount, mp->m_ialloc_blks);
353 max_metadata = icount;
354 } else {
355 max_metadata = agcount;
356 }
357 for (index = 0; index < agcount; index++) {
358 ino = XFS_AGINO_TO_INO(mp, index, agino);
359 if (ino > max_inum) {
360 index++;
361 break;
362 }
363
364 /* This ag is prefered for inodes */
365 pag = &mp->m_perag[index];
366 pag->pagi_inodeok = 1;
367 if (index < max_metadata)
368 pag->pagf_metadata = 1;
369 }
370 } else {
371 /* Setup default behavior for smaller filesystems */
372 for (index = 0; index < agcount; index++) {
373 pag = &mp->m_perag[index];
374 pag->pagi_inodeok = 1;
375 }
376 }
377 return index;
378}
379
380/*
381 * xfs_xlatesb
382 *
383 * data - on disk version of sb
384 * sb - a superblock
385 * dir - conversion direction: <0 - convert sb to buf
386 * >0 - convert buf to sb
387 * fields - which fields to copy (bitmask)
388 */
389void
390xfs_xlatesb(
391 void *data,
392 xfs_sb_t *sb,
393 int dir,
394 __int64_t fields)
395{
396 xfs_caddr_t buf_ptr;
397 xfs_caddr_t mem_ptr;
398 xfs_sb_field_t f;
399 int first;
400 int size;
401
402 ASSERT(dir);
403 ASSERT(fields);
404
405 if (!fields)
406 return;
407
408 buf_ptr = (xfs_caddr_t)data;
409 mem_ptr = (xfs_caddr_t)sb;
410
411 while (fields) {
412 f = (xfs_sb_field_t)xfs_lowbit64((__uint64_t)fields);
413 first = xfs_sb_info[f].offset;
414 size = xfs_sb_info[f + 1].offset - first;
415
416 ASSERT(xfs_sb_info[f].type == 0 || xfs_sb_info[f].type == 1);
417
418 if (size == 1 || xfs_sb_info[f].type == 1) {
419 if (dir > 0) {
420 memcpy(mem_ptr + first, buf_ptr + first, size);
421 } else {
422 memcpy(buf_ptr + first, mem_ptr + first, size);
423 }
424 } else {
425 switch (size) {
426 case 2:
427 INT_XLATE(*(__uint16_t*)(buf_ptr+first),
428 *(__uint16_t*)(mem_ptr+first),
429 dir, ARCH_CONVERT);
430 break;
431 case 4:
432 INT_XLATE(*(__uint32_t*)(buf_ptr+first),
433 *(__uint32_t*)(mem_ptr+first),
434 dir, ARCH_CONVERT);
435 break;
436 case 8:
437 INT_XLATE(*(__uint64_t*)(buf_ptr+first),
438 *(__uint64_t*)(mem_ptr+first), dir, ARCH_CONVERT);
439 break;
440 default:
441 ASSERT(0);
442 }
443 }
444
445 fields &= ~(1LL << f);
446 }
447}
448
449/*
450 * xfs_readsb
451 *
452 * Does the initial read of the superblock.
453 */
454int
455xfs_readsb(xfs_mount_t *mp)
456{
457 unsigned int sector_size;
458 unsigned int extra_flags;
459 xfs_buf_t *bp;
460 xfs_sb_t *sbp;
461 int error;
462
463 ASSERT(mp->m_sb_bp == NULL);
464 ASSERT(mp->m_ddev_targp != NULL);
465
466 /*
467 * Allocate a (locked) buffer to hold the superblock.
468 * This will be kept around at all times to optimize
469 * access to the superblock.
470 */
471 sector_size = xfs_getsize_buftarg(mp->m_ddev_targp);
472 extra_flags = XFS_BUF_LOCK | XFS_BUF_MANAGE | XFS_BUF_MAPPED;
473
474 bp = xfs_buf_read_flags(mp->m_ddev_targp, XFS_SB_DADDR,
475 BTOBB(sector_size), extra_flags);
476 if (!bp || XFS_BUF_ISERROR(bp)) {
477 cmn_err(CE_WARN, "XFS: SB read failed");
478 error = bp ? XFS_BUF_GETERROR(bp) : ENOMEM;
479 goto fail;
480 }
481 ASSERT(XFS_BUF_ISBUSY(bp));
482 ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
483
484 /*
485 * Initialize the mount structure from the superblock.
486 * But first do some basic consistency checking.
487 */
488 sbp = XFS_BUF_TO_SBP(bp);
489 xfs_xlatesb(XFS_BUF_PTR(bp), &(mp->m_sb), 1, XFS_SB_ALL_BITS);
490
491 error = xfs_mount_validate_sb(mp, &(mp->m_sb));
492 if (error) {
493 cmn_err(CE_WARN, "XFS: SB validate failed");
494 goto fail;
495 }
496
497 /*
498 * We must be able to do sector-sized and sector-aligned IO.
499 */
500 if (sector_size > mp->m_sb.sb_sectsize) {
501 cmn_err(CE_WARN,
502 "XFS: device supports only %u byte sectors (not %u)",
503 sector_size, mp->m_sb.sb_sectsize);
504 error = ENOSYS;
505 goto fail;
506 }
507
508 /*
509 * If device sector size is smaller than the superblock size,
510 * re-read the superblock so the buffer is correctly sized.
511 */
512 if (sector_size < mp->m_sb.sb_sectsize) {
513 XFS_BUF_UNMANAGE(bp);
514 xfs_buf_relse(bp);
515 sector_size = mp->m_sb.sb_sectsize;
516 bp = xfs_buf_read_flags(mp->m_ddev_targp, XFS_SB_DADDR,
517 BTOBB(sector_size), extra_flags);
518 if (!bp || XFS_BUF_ISERROR(bp)) {
519 cmn_err(CE_WARN, "XFS: SB re-read failed");
520 error = bp ? XFS_BUF_GETERROR(bp) : ENOMEM;
521 goto fail;
522 }
523 ASSERT(XFS_BUF_ISBUSY(bp));
524 ASSERT(XFS_BUF_VALUSEMA(bp) <= 0);
525 }
526
527 mp->m_sb_bp = bp;
528 xfs_buf_relse(bp);
529 ASSERT(XFS_BUF_VALUSEMA(bp) > 0);
530 return 0;
531
532 fail:
533 if (bp) {
534 XFS_BUF_UNMANAGE(bp);
535 xfs_buf_relse(bp);
536 }
537 return error;
538}
539
540
541/*
542 * xfs_mount_common
543 *
544 * Mount initialization code establishing various mount
545 * fields from the superblock associated with the given
546 * mount structure
547 */
ba0f32d4 548STATIC void
1da177e4
LT		 549xfs_mount_common(xfs_mount_t *mp, xfs_sb_t *sbp)
550{
551 int i;
552
553 mp->m_agfrotor = mp->m_agirotor = 0;
554 spinlock_init(&mp->m_agirotor_lock, "m_agirotor_lock");
555 mp->m_maxagi = mp->m_sb.sb_agcount;
556 mp->m_blkbit_log = sbp->sb_blocklog + XFS_NBBYLOG;
557 mp->m_blkbb_log = sbp->sb_blocklog - BBSHIFT;
558 mp->m_sectbb_log = sbp->sb_sectlog - BBSHIFT;
559 mp->m_agno_log = xfs_highbit32(sbp->sb_agcount - 1) + 1;
560 mp->m_agino_log = sbp->sb_inopblog + sbp->sb_agblklog;
561 mp->m_litino = sbp->sb_inodesize -
562 ((uint)sizeof(xfs_dinode_core_t) + (uint)sizeof(xfs_agino_t));
563 mp->m_blockmask = sbp->sb_blocksize - 1;
564 mp->m_blockwsize = sbp->sb_blocksize >> XFS_WORDLOG;
565 mp->m_blockwmask = mp->m_blockwsize - 1;
566 INIT_LIST_HEAD(&mp->m_del_inodes);
567
568 /*
569 * Setup for attributes, in case they get created.
570 * This value is for inodes getting attributes for the first time,
571 * the per-inode value is for old attribute values.
572 */
573 ASSERT(sbp->sb_inodesize >= 256 && sbp->sb_inodesize <= 2048);
574 switch (sbp->sb_inodesize) {
575 case 256:
d8cc890d
NS		 576 mp->m_attroffset = XFS_LITINO(mp) -
577 XFS_BMDR_SPACE_CALC(MINABTPTRS);
1da177e4
LT		 578 break;
579 case 512:
580 case 1024:
581 case 2048:
d8cc890d 582 mp->m_attroffset = XFS_BMDR_SPACE_CALC(6 * MINABTPTRS);
1da177e4
LT		 583 break;
584 default:
585 ASSERT(0);
586 }
587 ASSERT(mp->m_attroffset < XFS_LITINO(mp));
588
589 for (i = 0; i < 2; i++) {
590 mp->m_alloc_mxr[i] = XFS_BTREE_BLOCK_MAXRECS(sbp->sb_blocksize,
591 xfs_alloc, i == 0);
592 mp->m_alloc_mnr[i] = XFS_BTREE_BLOCK_MINRECS(sbp->sb_blocksize,
593 xfs_alloc, i == 0);
594 }
595 for (i = 0; i < 2; i++) {
596 mp->m_bmap_dmxr[i] = XFS_BTREE_BLOCK_MAXRECS(sbp->sb_blocksize,
597 xfs_bmbt, i == 0);
598 mp->m_bmap_dmnr[i] = XFS_BTREE_BLOCK_MINRECS(sbp->sb_blocksize,
599 xfs_bmbt, i == 0);
600 }
601 for (i = 0; i < 2; i++) {
602 mp->m_inobt_mxr[i] = XFS_BTREE_BLOCK_MAXRECS(sbp->sb_blocksize,
603 xfs_inobt, i == 0);
604 mp->m_inobt_mnr[i] = XFS_BTREE_BLOCK_MINRECS(sbp->sb_blocksize,
605 xfs_inobt, i == 0);
606 }
607
608 mp->m_bsize = XFS_FSB_TO_BB(mp, 1);
609 mp->m_ialloc_inos = (int)MAX((__uint16_t)XFS_INODES_PER_CHUNK,
610 sbp->sb_inopblock);
611 mp->m_ialloc_blks = mp->m_ialloc_inos >> sbp->sb_inopblog;
612}
613/*
614 * xfs_mountfs
615 *
616 * This function does the following on an initial mount of a file system:
617 * - reads the superblock from disk and init the mount struct
618 * - if we're a 32-bit kernel, do a size check on the superblock
619 * so we don't mount terabyte filesystems
620 * - init mount struct realtime fields
621 * - allocate inode hash table for fs
622 * - init directory manager
623 * - perform recovery and init the log manager
624 */
625int
626xfs_mountfs(
627 vfs_t *vfsp,
628 xfs_mount_t *mp,
629 int mfsi_flags)
630{
631 xfs_buf_t *bp;
632 xfs_sb_t *sbp = &(mp->m_sb);
633 xfs_inode_t *rip;
634 vnode_t *rvp = NULL;
635 int readio_log, writeio_log;
636 xfs_daddr_t d;
637 __uint64_t ret64;
638 __int64_t update_flags;
639 uint quotamount, quotaflags;
640 int agno;
641 int uuid_mounted = 0;
642 int error = 0;
643
644 if (mp->m_sb_bp == NULL) {
645 if ((error = xfs_readsb(mp))) {
646 return (error);
647 }
648 }
649 xfs_mount_common(mp, sbp);
650
651 /*
652 * Check if sb_agblocks is aligned at stripe boundary
653 * If sb_agblocks is NOT aligned turn off m_dalign since
654 * allocator alignment is within an ag, therefore ag has
655 * to be aligned at stripe boundary.
656 */
657 update_flags = 0LL;
658 if (mp->m_dalign && !(mfsi_flags & XFS_MFSI_SECOND)) {
659 /*
660 * If stripe unit and stripe width are not multiples
661 * of the fs blocksize turn off alignment.
662 */
663 if ((BBTOB(mp->m_dalign) & mp->m_blockmask) ||
664 (BBTOB(mp->m_swidth) & mp->m_blockmask)) {
665 if (mp->m_flags & XFS_MOUNT_RETERR) {
666 cmn_err(CE_WARN,
667 "XFS: alignment check 1 failed");
668 error = XFS_ERROR(EINVAL);
669 goto error1;
670 }
671 mp->m_dalign = mp->m_swidth = 0;
672 } else {
673 /*
674 * Convert the stripe unit and width to FSBs.
675 */
676 mp->m_dalign = XFS_BB_TO_FSBT(mp, mp->m_dalign);
677 if (mp->m_dalign && (sbp->sb_agblocks % mp->m_dalign)) {
678 if (mp->m_flags & XFS_MOUNT_RETERR) {
679 error = XFS_ERROR(EINVAL);
680 goto error1;
681 }
682 xfs_fs_cmn_err(CE_WARN, mp,
683"stripe alignment turned off: sunit(%d)/swidth(%d) incompatible with agsize(%d)",
684 mp->m_dalign, mp->m_swidth,
685 sbp->sb_agblocks);
686
687 mp->m_dalign = 0;
688 mp->m_swidth = 0;
689 } else if (mp->m_dalign) {
690 mp->m_swidth = XFS_BB_TO_FSBT(mp, mp->m_swidth);
691 } else {
692 if (mp->m_flags & XFS_MOUNT_RETERR) {
693 xfs_fs_cmn_err(CE_WARN, mp,
694"stripe alignment turned off: sunit(%d) less than bsize(%d)",
695 mp->m_dalign,
696 mp->m_blockmask +1);
697 error = XFS_ERROR(EINVAL);
698 goto error1;
699 }
700 mp->m_swidth = 0;
701 }
702 }
703
704 /*
705 * Update superblock with new values
706 * and log changes
707 */
708 if (XFS_SB_VERSION_HASDALIGN(sbp)) {
709 if (sbp->sb_unit != mp->m_dalign) {
710 sbp->sb_unit = mp->m_dalign;
711 update_flags |= XFS_SB_UNIT;
712 }
713 if (sbp->sb_width != mp->m_swidth) {
714 sbp->sb_width = mp->m_swidth;
715 update_flags |= XFS_SB_WIDTH;
716 }
717 }
718 } else if ((mp->m_flags & XFS_MOUNT_NOALIGN) != XFS_MOUNT_NOALIGN &&
719 XFS_SB_VERSION_HASDALIGN(&mp->m_sb)) {
720 mp->m_dalign = sbp->sb_unit;
721 mp->m_swidth = sbp->sb_width;
722 }
723
724 xfs_alloc_compute_maxlevels(mp);
725 xfs_bmap_compute_maxlevels(mp, XFS_DATA_FORK);
726 xfs_bmap_compute_maxlevels(mp, XFS_ATTR_FORK);
727 xfs_ialloc_compute_maxlevels(mp);
728
729 if (sbp->sb_imax_pct) {
730 __uint64_t icount;
731
732 /* Make sure the maximum inode count is a multiple of the
733 * units we allocate inodes in.
734 */
735
736 icount = sbp->sb_dblocks * sbp->sb_imax_pct;
737 do_div(icount, 100);
738 do_div(icount, mp->m_ialloc_blks);
739 mp->m_maxicount = (icount * mp->m_ialloc_blks) <<
740 sbp->sb_inopblog;
741 } else
742 mp->m_maxicount = 0;
743
744 mp->m_maxioffset = xfs_max_file_offset(sbp->sb_blocklog);
745
746 /*
747 * XFS uses the uuid from the superblock as the unique
748 * identifier for fsid. We can not use the uuid from the volume
749 * since a single partition filesystem is identical to a single
750 * partition volume/filesystem.
751 */
752 if ((mfsi_flags & XFS_MFSI_SECOND) == 0 &&
753 (mp->m_flags & XFS_MOUNT_NOUUID) == 0) {
754 if (xfs_uuid_mount(mp)) {
755 error = XFS_ERROR(EINVAL);
756 goto error1;
757 }
758 uuid_mounted=1;
759 ret64 = uuid_hash64(&sbp->sb_uuid);
760 memcpy(&vfsp->vfs_fsid, &ret64, sizeof(ret64));
761 }
762
763 /*
764 * Set the default minimum read and write sizes unless
765 * already specified in a mount option.
766 * We use smaller I/O sizes when the file system
767 * is being used for NFS service (wsync mount option).
768 */
769 if (!(mp->m_flags & XFS_MOUNT_DFLT_IOSIZE)) {
770 if (mp->m_flags & XFS_MOUNT_WSYNC) {
771 readio_log = XFS_WSYNC_READIO_LOG;
772 writeio_log = XFS_WSYNC_WRITEIO_LOG;
773 } else {
774 readio_log = XFS_READIO_LOG_LARGE;
775 writeio_log = XFS_WRITEIO_LOG_LARGE;
776 }
777 } else {
778 readio_log = mp->m_readio_log;
779 writeio_log = mp->m_writeio_log;
780 }
781
782 /*
783 * Set the number of readahead buffers to use based on
784 * physical memory size.
785 */
786 if (xfs_physmem <= 4096) /* <= 16MB */
787 mp->m_nreadaheads = XFS_RW_NREADAHEAD_16MB;
788 else if (xfs_physmem <= 8192) /* <= 32MB */
789 mp->m_nreadaheads = XFS_RW_NREADAHEAD_32MB;
790 else
791 mp->m_nreadaheads = XFS_RW_NREADAHEAD_K32;
792 if (sbp->sb_blocklog > readio_log) {
793 mp->m_readio_log = sbp->sb_blocklog;
794 } else {
795 mp->m_readio_log = readio_log;
796 }
797 mp->m_readio_blocks = 1 << (mp->m_readio_log - sbp->sb_blocklog);
798 if (sbp->sb_blocklog > writeio_log) {
799 mp->m_writeio_log = sbp->sb_blocklog;
800 } else {
801 mp->m_writeio_log = writeio_log;
802 }
803 mp->m_writeio_blocks = 1 << (mp->m_writeio_log - sbp->sb_blocklog);
804
805 /*
806 * Set the inode cluster size based on the physical memory
807 * size. This may still be overridden by the file system
808 * block size if it is larger than the chosen cluster size.
809 */
810 if (xfs_physmem <= btoc(32 * 1024 * 1024)) { /* <= 32 MB */
811 mp->m_inode_cluster_size = XFS_INODE_SMALL_CLUSTER_SIZE;
812 } else {
813 mp->m_inode_cluster_size = XFS_INODE_BIG_CLUSTER_SIZE;
814 }
815 /*
816 * Set whether we're using inode alignment.
817 */
818 if (XFS_SB_VERSION_HASALIGN(&mp->m_sb) &&
819 mp->m_sb.sb_inoalignmt >=
820 XFS_B_TO_FSBT(mp, mp->m_inode_cluster_size))
821 mp->m_inoalign_mask = mp->m_sb.sb_inoalignmt - 1;
822 else
823 mp->m_inoalign_mask = 0;
824 /*
825 * If we are using stripe alignment, check whether
826 * the stripe unit is a multiple of the inode alignment
827 */
828 if (mp->m_dalign && mp->m_inoalign_mask &&
829 !(mp->m_dalign & mp->m_inoalign_mask))
830 mp->m_sinoalign = mp->m_dalign;
831 else
832 mp->m_sinoalign = 0;
833 /*
834 * Check that the data (and log if separate) are an ok size.
835 */
836 d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_dblocks);
837 if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_dblocks) {
838 cmn_err(CE_WARN, "XFS: size check 1 failed");
839 error = XFS_ERROR(E2BIG);
840 goto error1;
841 }
842 error = xfs_read_buf(mp, mp->m_ddev_targp,
843 d - XFS_FSS_TO_BB(mp, 1),
844 XFS_FSS_TO_BB(mp, 1), 0, &bp);
845 if (!error) {
846 xfs_buf_relse(bp);
847 } else {
848 cmn_err(CE_WARN, "XFS: size check 2 failed");
849 if (error == ENOSPC) {
850 error = XFS_ERROR(E2BIG);
851 }
852 goto error1;
853 }
854
855 if (((mfsi_flags & XFS_MFSI_CLIENT) == 0) &&
856 mp->m_logdev_targp != mp->m_ddev_targp) {
857 d = (xfs_daddr_t)XFS_FSB_TO_BB(mp, mp->m_sb.sb_logblocks);
858 if (XFS_BB_TO_FSB(mp, d) != mp->m_sb.sb_logblocks) {
859 cmn_err(CE_WARN, "XFS: size check 3 failed");
860 error = XFS_ERROR(E2BIG);
861 goto error1;
862 }
863 error = xfs_read_buf(mp, mp->m_logdev_targp,
864 d - XFS_FSB_TO_BB(mp, 1),
865 XFS_FSB_TO_BB(mp, 1), 0, &bp);
866 if (!error) {
867 xfs_buf_relse(bp);
868 } else {
869 cmn_err(CE_WARN, "XFS: size check 3 failed");
870 if (error == ENOSPC) {
871 error = XFS_ERROR(E2BIG);
872 }
873 goto error1;
874 }
875 }
876
877 /*
878 * Initialize realtime fields in the mount structure
879 */
880 if ((error = xfs_rtmount_init(mp))) {
881 cmn_err(CE_WARN, "XFS: RT mount failed");
882 goto error1;
883 }
884
885 /*
886 * For client case we are done now
887 */
888 if (mfsi_flags & XFS_MFSI_CLIENT) {
889 return(0);
890 }
891
892 /*
893 * Copies the low order bits of the timestamp and the randomly
894 * set "sequence" number out of a UUID.
895 */
896 uuid_getnodeuniq(&sbp->sb_uuid, mp->m_fixedfsid);
897
898 /*
899 * The vfs structure needs to have a file system independent
900 * way of checking for the invariant file system ID. Since it
901 * can't look at mount structures it has a pointer to the data
902 * in the mount structure.
903 *
904 * File systems that don't support user level file handles (i.e.
905 * all of them except for XFS) will leave vfs_altfsid as NULL.
906 */
907 vfsp->vfs_altfsid = (xfs_fsid_t *)mp->m_fixedfsid;
908 mp->m_dmevmask = 0; /* not persistent; set after each mount */
909
910 /*
911 * Select the right directory manager.
912 */
913 mp->m_dirops =
914 XFS_SB_VERSION_HASDIRV2(&mp->m_sb) ?
915 xfsv2_dirops :
916 xfsv1_dirops;
917
918 /*
919 * Initialize directory manager's entries.
920 */
921 XFS_DIR_MOUNT(mp);
922
923 /*
924 * Initialize the attribute manager's entries.
925 */
926 mp->m_attr_magicpct = (mp->m_sb.sb_blocksize * 37) / 100;
927
928 /*
929 * Initialize the precomputed transaction reservations values.
930 */
931 xfs_trans_init(mp);
932
933 /*
934 * Allocate and initialize the inode hash table for this
935 * file system.
936 */
937 xfs_ihash_init(mp);
938 xfs_chash_init(mp);
939
940 /*
941 * Allocate and initialize the per-ag data.
942 */
943 init_rwsem(&mp->m_peraglock);
944 mp->m_perag =
945 kmem_zalloc(sbp->sb_agcount * sizeof(xfs_perag_t), KM_SLEEP);
946
947 mp->m_maxagi = xfs_initialize_perag(mp, sbp->sb_agcount);
948
949 /*
950 * log's mount-time initialization. Perform 1st part recovery if needed
951 */
952 if (likely(sbp->sb_logblocks > 0)) { /* check for volume case */
953 error = xfs_log_mount(mp, mp->m_logdev_targp,
954 XFS_FSB_TO_DADDR(mp, sbp->sb_logstart),
955 XFS_FSB_TO_BB(mp, sbp->sb_logblocks));
956 if (error) {
957 cmn_err(CE_WARN, "XFS: log mount failed");
958 goto error2;
959 }
960 } else { /* No log has been defined */
961 cmn_err(CE_WARN, "XFS: no log defined");
962 XFS_ERROR_REPORT("xfs_mountfs_int(1)", XFS_ERRLEVEL_LOW, mp);
963 error = XFS_ERROR(EFSCORRUPTED);
964 goto error2;
965 }
966
967 /*
968 * Get and sanity-check the root inode.
969 * Save the pointer to it in the mount structure.
970 */
971 error = xfs_iget(mp, NULL, sbp->sb_rootino, 0, XFS_ILOCK_EXCL, &rip, 0);
972 if (error) {
973 cmn_err(CE_WARN, "XFS: failed to read root inode");
974 goto error3;
975 }
976
977 ASSERT(rip != NULL);
978 rvp = XFS_ITOV(rip);
979
980 if (unlikely((rip->i_d.di_mode & S_IFMT) != S_IFDIR)) {
981 cmn_err(CE_WARN, "XFS: corrupted root inode");
982 prdev("Root inode %llu is not a directory",
983 mp->m_ddev_targp, (unsigned long long)rip->i_ino);
984 xfs_iunlock(rip, XFS_ILOCK_EXCL);
985 XFS_ERROR_REPORT("xfs_mountfs_int(2)", XFS_ERRLEVEL_LOW,
986 mp);
987 error = XFS_ERROR(EFSCORRUPTED);
988 goto error4;
989 }
990 mp->m_rootip = rip; /* save it */
991
992 xfs_iunlock(rip, XFS_ILOCK_EXCL);
993
994 /*
995 * Initialize realtime inode pointers in the mount structure
996 */
997 if ((error = xfs_rtmount_inodes(mp))) {
998 /*
999 * Free up the root inode.
1000 */
1001 cmn_err(CE_WARN, "XFS: failed to read RT inodes");
1002 goto error4;
1003 }
1004
1005 /*
1006 * If fs is not mounted readonly, then update the superblock
1007 * unit and width changes.
1008 */
1009 if (update_flags && !(vfsp->vfs_flag & VFS_RDONLY))
1010 xfs_mount_log_sbunit(mp, update_flags);
1011
1012 /*
1013 * Initialise the XFS quota management subsystem for this mount
1014 */
1015 if ((error = XFS_QM_INIT(mp, &quotamount, &quotaflags)))
1016 goto error4;
1017
1018 /*
1019 * Finish recovering the file system. This part needed to be
1020 * delayed until after the root and real-time bitmap inodes
1021 * were consistently read in.
1022 */
1023 error = xfs_log_mount_finish(mp, mfsi_flags);
1024 if (error) {
1025 cmn_err(CE_WARN, "XFS: log mount finish failed");
1026 goto error4;
1027 }
1028
1029 /*
1030 * Complete the quota initialisation, post-log-replay component.
1031 */
1032 if ((error = XFS_QM_MOUNT(mp, quotamount, quotaflags, mfsi_flags)))
1033 goto error4;
1034
1035 return 0;
1036
1037 error4:
1038 /*
1039 * Free up the root inode.
1040 */
1041 VN_RELE(rvp);
1042 error3:
1043 xfs_log_unmount_dealloc(mp);
1044 error2:
1045 xfs_ihash_free(mp);
1046 xfs_chash_free(mp);
1047 for (agno = 0; agno < sbp->sb_agcount; agno++)
1048 if (mp->m_perag[agno].pagb_list)
1049 kmem_free(mp->m_perag[agno].pagb_list,
1050 sizeof(xfs_perag_busy_t) * XFS_PAGB_NUM_SLOTS);
1051 kmem_free(mp->m_perag, sbp->sb_agcount * sizeof(xfs_perag_t));
1052 mp->m_perag = NULL;
1053 /* FALLTHROUGH */
1054 error1:
1055 if (uuid_mounted)
1056 xfs_uuid_unmount(mp);
1057 xfs_freesb(mp);
1058 return error;
1059}
1060
1061/*
1062 * xfs_unmountfs
1063 *
1064 * This flushes out the inodes,dquots and the superblock, unmounts the
1065 * log and makes sure that incore structures are freed.
1066 */
1067int
1068xfs_unmountfs(xfs_mount_t *mp, struct cred *cr)
1069{
1070 struct vfs *vfsp = XFS_MTOVFS(mp);
1071#if defined(DEBUG) || defined(INDUCE_IO_ERROR)
1072 int64_t fsid;
1073#endif
1074
efa80278 1075 xfs_iflush_all(mp);
1da177e4
LT		 1076
1077 XFS_QM_DQPURGEALL(mp,
1078 XFS_QMOPT_UQUOTA | XFS_QMOPT_GQUOTA | XFS_QMOPT_UMOUNTING);
1079
1080 /*
1081 * Flush out the log synchronously so that we know for sure
1082 * that nothing is pinned. This is important because bflush()
1083 * will skip pinned buffers.
1084 */
1085 xfs_log_force(mp, (xfs_lsn_t)0, XFS_LOG_FORCE | XFS_LOG_SYNC);
1086
1087 xfs_binval(mp->m_ddev_targp);
1088 if (mp->m_rtdev_targp) {
1089 xfs_binval(mp->m_rtdev_targp);
1090 }
1091
1092 xfs_unmountfs_writesb(mp);
1093
1094 xfs_unmountfs_wait(mp); /* wait for async bufs */
1095
1096 xfs_log_unmount(mp); /* Done! No more fs ops. */
1097
1098 xfs_freesb(mp);
1099
1100 /*
1101 * All inodes from this mount point should be freed.
1102 */
1103 ASSERT(mp->m_inodes == NULL);
1104
1da177e4
LT		 1105 xfs_unmountfs_close(mp, cr);
1106 if ((mp->m_flags & XFS_MOUNT_NOUUID) == 0)
1107 xfs_uuid_unmount(mp);
1108
1109#if defined(DEBUG) || defined(INDUCE_IO_ERROR)
1110 /*
1111 * clear all error tags on this filesystem
1112 */
1113 memcpy(&fsid, &vfsp->vfs_fsid, sizeof(int64_t));
1114 xfs_errortag_clearall_umount(fsid, mp->m_fsname, 0);
1115#endif
1116 XFS_IODONE(vfsp);
1117 xfs_mount_free(mp, 1);
1118 return 0;
1119}
1120
1121void
1122xfs_unmountfs_close(xfs_mount_t *mp, struct cred *cr)
1123{
1124 if (mp->m_logdev_targp != mp->m_ddev_targp)
1125 xfs_free_buftarg(mp->m_logdev_targp, 1);
1126 if (mp->m_rtdev_targp)
1127 xfs_free_buftarg(mp->m_rtdev_targp, 1);
1128 xfs_free_buftarg(mp->m_ddev_targp, 0);
1129}
1130
ba0f32d4 1131STATIC void
1da177e4
LT		 1132xfs_unmountfs_wait(xfs_mount_t *mp)
1133{
1134 if (mp->m_logdev_targp != mp->m_ddev_targp)
1135 xfs_wait_buftarg(mp->m_logdev_targp);
1136 if (mp->m_rtdev_targp)
1137 xfs_wait_buftarg(mp->m_rtdev_targp);
1138 xfs_wait_buftarg(mp->m_ddev_targp);
1139}
1140
1141int
1142xfs_unmountfs_writesb(xfs_mount_t *mp)
1143{
1144 xfs_buf_t *sbp;
1145 xfs_sb_t *sb;
1146 int error = 0;
1147
1148 /*
1149 * skip superblock write if fs is read-only, or
1150 * if we are doing a forced umount.
1151 */
1152 sbp = xfs_getsb(mp, 0);
1153 if (!(XFS_MTOVFS(mp)->vfs_flag & VFS_RDONLY ||
1154 XFS_FORCED_SHUTDOWN(mp))) {
1155 /*
1156 * mark shared-readonly if desired
1157 */
1158 sb = XFS_BUF_TO_SBP(sbp);
1159 if (mp->m_mk_sharedro) {
1160 if (!(sb->sb_flags & XFS_SBF_READONLY))
1161 sb->sb_flags |= XFS_SBF_READONLY;
1162 if (!XFS_SB_VERSION_HASSHARED(sb))
1163 XFS_SB_VERSION_ADDSHARED(sb);
1164 xfs_fs_cmn_err(CE_NOTE, mp,
1165 "Unmounting, marking shared read-only");
1166 }
1167 XFS_BUF_UNDONE(sbp);
1168 XFS_BUF_UNREAD(sbp);
1169 XFS_BUF_UNDELAYWRITE(sbp);
1170 XFS_BUF_WRITE(sbp);
1171 XFS_BUF_UNASYNC(sbp);
1172 ASSERT(XFS_BUF_TARGET(sbp) == mp->m_ddev_targp);
1173 xfsbdstrat(mp, sbp);
1174 /* Nevermind errors we might get here. */
1175 error = xfs_iowait(sbp);
1176 if (error)
1177 xfs_ioerror_alert("xfs_unmountfs_writesb",
1178 mp, sbp, XFS_BUF_ADDR(sbp));
1179 if (error && mp->m_mk_sharedro)
1180 xfs_fs_cmn_err(CE_ALERT, mp, "Superblock write error detected while unmounting. Filesystem may not be marked shared readonly");
1181 }
1182 xfs_buf_relse(sbp);
1183 return (error);
1184}
1185
1186/*
1187 * xfs_mod_sb() can be used to copy arbitrary changes to the
1188 * in-core superblock into the superblock buffer to be logged.
1189 * It does not provide the higher level of locking that is
1190 * needed to protect the in-core superblock from concurrent
1191 * access.
1192 */
1193void
1194xfs_mod_sb(xfs_trans_t *tp, __int64_t fields)
1195{
1196 xfs_buf_t *bp;
1197 int first;
1198 int last;
1199 xfs_mount_t *mp;
1200 xfs_sb_t *sbp;
1201 xfs_sb_field_t f;
1202
1203 ASSERT(fields);
1204 if (!fields)
1205 return;
1206 mp = tp->t_mountp;
1207 bp = xfs_trans_getsb(tp, mp, 0);
1208 sbp = XFS_BUF_TO_SBP(bp);
1209 first = sizeof(xfs_sb_t);
1210 last = 0;
1211
1212 /* translate/copy */
1213
1214 xfs_xlatesb(XFS_BUF_PTR(bp), &(mp->m_sb), -1, fields);
1215
1216 /* find modified range */
1217
1218 f = (xfs_sb_field_t)xfs_lowbit64((__uint64_t)fields);
1219 ASSERT((1LL << f) & XFS_SB_MOD_BITS);
1220 first = xfs_sb_info[f].offset;
1221
1222 f = (xfs_sb_field_t)xfs_highbit64((__uint64_t)fields);
1223 ASSERT((1LL << f) & XFS_SB_MOD_BITS);
1224 last = xfs_sb_info[f + 1].offset - 1;
1225
1226 xfs_trans_log_buf(tp, bp, first, last);
1227}
1228
1229/*
1230 * xfs_mod_incore_sb_unlocked() is a utility routine common used to apply
1231 * a delta to a specified field in the in-core superblock. Simply
1232 * switch on the field indicated and apply the delta to that field.
1233 * Fields are not allowed to dip below zero, so if the delta would
1234 * do this do not apply it and return EINVAL.
1235 *
1236 * The SB_LOCK must be held when this routine is called.
1237 */
1238STATIC int
1239xfs_mod_incore_sb_unlocked(xfs_mount_t *mp, xfs_sb_field_t field,
1240 int delta, int rsvd)
1241{
1242 int scounter; /* short counter for 32 bit fields */
1243 long long lcounter; /* long counter for 64 bit fields */
1244 long long res_used, rem;
1245
1246 /*
1247 * With the in-core superblock spin lock held, switch
1248 * on the indicated field. Apply the delta to the
1249 * proper field. If the fields value would dip below
1250 * 0, then do not apply the delta and return EINVAL.
1251 */
1252 switch (field) {
1253 case XFS_SBS_ICOUNT:
1254 lcounter = (long long)mp->m_sb.sb_icount;
1255 lcounter += delta;
1256 if (lcounter < 0) {
1257 ASSERT(0);
1258 return (XFS_ERROR(EINVAL));
1259 }
1260 mp->m_sb.sb_icount = lcounter;
1261 return (0);
1262 case XFS_SBS_IFREE:
1263 lcounter = (long long)mp->m_sb.sb_ifree;
1264 lcounter += delta;
1265 if (lcounter < 0) {
1266 ASSERT(0);
1267 return (XFS_ERROR(EINVAL));
1268 }
1269 mp->m_sb.sb_ifree = lcounter;
1270 return (0);
1271 case XFS_SBS_FDBLOCKS:
1272
1273 lcounter = (long long)mp->m_sb.sb_fdblocks;
1274 res_used = (long long)(mp->m_resblks - mp->m_resblks_avail);
1275
1276 if (delta > 0) { /* Putting blocks back */
1277 if (res_used > delta) {
1278 mp->m_resblks_avail += delta;
1279 } else {
1280 rem = delta - res_used;
1281 mp->m_resblks_avail = mp->m_resblks;
1282 lcounter += rem;
1283 }
1284 } else { /* Taking blocks away */
1285
1286 lcounter += delta;
1287
1288 /*
1289 * If were out of blocks, use any available reserved blocks if
1290 * were allowed to.
1291 */
1292
1293 if (lcounter < 0) {
1294 if (rsvd) {
1295 lcounter = (long long)mp->m_resblks_avail + delta;
1296 if (lcounter < 0) {
1297 return (XFS_ERROR(ENOSPC));
1298 }
1299 mp->m_resblks_avail = lcounter;
1300 return (0);
1301 } else { /* not reserved */
1302 return (XFS_ERROR(ENOSPC));
1303 }
1304 }
1305 }
1306
1307 mp->m_sb.sb_fdblocks = lcounter;
1308 return (0);
1309 case XFS_SBS_FREXTENTS:
1310 lcounter = (long long)mp->m_sb.sb_frextents;
1311 lcounter += delta;
1312 if (lcounter < 0) {
1313 return (XFS_ERROR(ENOSPC));
1314 }
1315 mp->m_sb.sb_frextents = lcounter;
1316 return (0);
1317 case XFS_SBS_DBLOCKS:
1318 lcounter = (long long)mp->m_sb.sb_dblocks;
1319 lcounter += delta;
1320 if (lcounter < 0) {
1321 ASSERT(0);
1322 return (XFS_ERROR(EINVAL));
1323 }
1324 mp->m_sb.sb_dblocks = lcounter;
1325 return (0);
1326 case XFS_SBS_AGCOUNT:
1327 scounter = mp->m_sb.sb_agcount;
1328 scounter += delta;
1329 if (scounter < 0) {
1330 ASSERT(0);
1331 return (XFS_ERROR(EINVAL));
1332 }
1333 mp->m_sb.sb_agcount = scounter;
1334 return (0);
1335 case XFS_SBS_IMAX_PCT:
1336 scounter = mp->m_sb.sb_imax_pct;
1337 scounter += delta;
1338 if (scounter < 0) {
1339 ASSERT(0);
1340 return (XFS_ERROR(EINVAL));
1341 }
1342 mp->m_sb.sb_imax_pct = scounter;
1343 return (0);
1344 case XFS_SBS_REXTSIZE:
1345 scounter = mp->m_sb.sb_rextsize;
1346 scounter += delta;
1347 if (scounter < 0) {
1348 ASSERT(0);
1349 return (XFS_ERROR(EINVAL));
1350 }
1351 mp->m_sb.sb_rextsize = scounter;
1352 return (0);
1353 case XFS_SBS_RBMBLOCKS:
1354 scounter = mp->m_sb.sb_rbmblocks;
1355 scounter += delta;
1356 if (scounter < 0) {
1357 ASSERT(0);
1358 return (XFS_ERROR(EINVAL));
1359 }
1360 mp->m_sb.sb_rbmblocks = scounter;
1361 return (0);
1362 case XFS_SBS_RBLOCKS:
1363 lcounter = (long long)mp->m_sb.sb_rblocks;
1364 lcounter += delta;
1365 if (lcounter < 0) {
1366 ASSERT(0);
1367 return (XFS_ERROR(EINVAL));
1368 }
1369 mp->m_sb.sb_rblocks = lcounter;
1370 return (0);
1371 case XFS_SBS_REXTENTS:
1372 lcounter = (long long)mp->m_sb.sb_rextents;
1373 lcounter += delta;
1374 if (lcounter < 0) {
1375 ASSERT(0);
1376 return (XFS_ERROR(EINVAL));
1377 }
1378 mp->m_sb.sb_rextents = lcounter;
1379 return (0);
1380 case XFS_SBS_REXTSLOG:
1381 scounter = mp->m_sb.sb_rextslog;
1382 scounter += delta;
1383 if (scounter < 0) {
1384 ASSERT(0);
1385 return (XFS_ERROR(EINVAL));
1386 }
1387 mp->m_sb.sb_rextslog = scounter;
1388 return (0);
1389 default:
1390 ASSERT(0);
1391 return (XFS_ERROR(EINVAL));
1392 }
1393}
1394
1395/*
1396 * xfs_mod_incore_sb() is used to change a field in the in-core
1397 * superblock structure by the specified delta. This modification
1398 * is protected by the SB_LOCK. Just use the xfs_mod_incore_sb_unlocked()
1399 * routine to do the work.
1400 */
1401int
1402xfs_mod_incore_sb(xfs_mount_t *mp, xfs_sb_field_t field, int delta, int rsvd)
1403{
1404 unsigned long s;
1405 int status;
1406
1407 s = XFS_SB_LOCK(mp);
1408 status = xfs_mod_incore_sb_unlocked(mp, field, delta, rsvd);
1409 XFS_SB_UNLOCK(mp, s);
1410 return (status);
1411}
1412
1413/*
1414 * xfs_mod_incore_sb_batch() is used to change more than one field
1415 * in the in-core superblock structure at a time. This modification
1416 * is protected by a lock internal to this module. The fields and
1417 * changes to those fields are specified in the array of xfs_mod_sb
1418 * structures passed in.
1419 *
1420 * Either all of the specified deltas will be applied or none of
1421 * them will. If any modified field dips below 0, then all modifications
1422 * will be backed out and EINVAL will be returned.
1423 */
1424int
1425xfs_mod_incore_sb_batch(xfs_mount_t *mp, xfs_mod_sb_t *msb, uint nmsb, int rsvd)
1426{
1427 unsigned long s;
1428 int status=0;
1429 xfs_mod_sb_t *msbp;
1430
1431 /*
1432 * Loop through the array of mod structures and apply each
1433 * individually. If any fail, then back out all those
1434 * which have already been applied. Do all of this within
1435 * the scope of the SB_LOCK so that all of the changes will
1436 * be atomic.
1437 */
1438 s = XFS_SB_LOCK(mp);
1439 msbp = &msb[0];
1440 for (msbp = &msbp[0]; msbp < (msb + nmsb); msbp++) {
1441 /*
1442 * Apply the delta at index n. If it fails, break
1443 * from the loop so we'll fall into the undo loop
1444 * below.
1445 */
1446 status = xfs_mod_incore_sb_unlocked(mp, msbp->msb_field,
1447 msbp->msb_delta, rsvd);
1448 if (status != 0) {
1449 break;
1450 }
1451 }
1452
1453 /*
1454 * If we didn't complete the loop above, then back out
1455 * any changes made to the superblock. If you add code
1456 * between the loop above and here, make sure that you
1457 * preserve the value of status. Loop back until
1458 * we step below the beginning of the array. Make sure
1459 * we don't touch anything back there.
1460 */
1461 if (status != 0) {
1462 msbp--;
1463 while (msbp >= msb) {
1464 status = xfs_mod_incore_sb_unlocked(mp,
1465 msbp->msb_field, -(msbp->msb_delta), rsvd);
1466 ASSERT(status == 0);
1467 msbp--;
1468 }
1469 }
1470 XFS_SB_UNLOCK(mp, s);
1471 return (status);
1472}
1473
1474/*
1475 * xfs_getsb() is called to obtain the buffer for the superblock.
1476 * The buffer is returned locked and read in from disk.
1477 * The buffer should be released with a call to xfs_brelse().
1478 *
1479 * If the flags parameter is BUF_TRYLOCK, then we'll only return
1480 * the superblock buffer if it can be locked without sleeping.
1481 * If it can't then we'll return NULL.
1482 */
1483xfs_buf_t *
1484xfs_getsb(
1485 xfs_mount_t *mp,
1486 int flags)
1487{
1488 xfs_buf_t *bp;
1489
1490 ASSERT(mp->m_sb_bp != NULL);
1491 bp = mp->m_sb_bp;
1492 if (flags & XFS_BUF_TRYLOCK) {
1493 if (!XFS_BUF_CPSEMA(bp)) {
1494 return NULL;
1495 }
1496 } else {
1497 XFS_BUF_PSEMA(bp, PRIBIO);
1498 }
1499 XFS_BUF_HOLD(bp);
1500 ASSERT(XFS_BUF_ISDONE(bp));
1501 return (bp);
1502}
1503
1504/*
1505 * Used to free the superblock along various error paths.
1506 */
1507void
1508xfs_freesb(
1509 xfs_mount_t *mp)
1510{
1511 xfs_buf_t *bp;
1512
1513 /*
1514 * Use xfs_getsb() so that the buffer will be locked
1515 * when we call xfs_buf_relse().
1516 */
1517 bp = xfs_getsb(mp, 0);
1518 XFS_BUF_UNMANAGE(bp);
1519 xfs_buf_relse(bp);
1520 mp->m_sb_bp = NULL;
1521}
1522
1523/*
1524 * See if the UUID is unique among mounted XFS filesystems.
1525 * Mount fails if UUID is nil or a FS with the same UUID is already mounted.
1526 */
1527STATIC int
1528xfs_uuid_mount(
1529 xfs_mount_t *mp)
1530{
1531 if (uuid_is_nil(&mp->m_sb.sb_uuid)) {
1532 cmn_err(CE_WARN,
1533 "XFS: Filesystem %s has nil UUID - can't mount",
1534 mp->m_fsname);
1535 return -1;
1536 }
1537 if (!uuid_table_insert(&mp->m_sb.sb_uuid)) {
1538 cmn_err(CE_WARN,
1539 "XFS: Filesystem %s has duplicate UUID - can't mount",
1540 mp->m_fsname);
1541 return -1;
1542 }
1543 return 0;
1544}
1545
1546/*
1547 * Remove filesystem from the UUID table.
1548 */
1549STATIC void
1550xfs_uuid_unmount(
1551 xfs_mount_t *mp)
1552{
1553 uuid_table_remove(&mp->m_sb.sb_uuid);
1554}
1555
1556/*
1557 * Used to log changes to the superblock unit and width fields which could
1558 * be altered by the mount options. Only the first superblock is updated.
1559 */
1560STATIC void
1561xfs_mount_log_sbunit(
1562 xfs_mount_t *mp,
1563 __int64_t fields)
1564{
1565 xfs_trans_t *tp;
1566
1567 ASSERT(fields & (XFS_SB_UNIT|XFS_SB_WIDTH|XFS_SB_UUID));
1568
1569 tp = xfs_trans_alloc(mp, XFS_TRANS_SB_UNIT);
1570 if (xfs_trans_reserve(tp, 0, mp->m_sb.sb_sectsize + 128, 0, 0,
1571 XFS_DEFAULT_LOG_COUNT)) {
1572 xfs_trans_cancel(tp, 0);
1573 return;
1574 }
1575 xfs_mod_sb(tp, fields);
1576 xfs_trans_commit(tp, 0, NULL);
1577}
kernel source for telekom puls (alcatel/mediatek)