2 * Copyright (C) International Business Machines Corp., 2000-2004
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of the GNU General Public License as published by
6 * the Free Software Foundation; either version 2 of the License, or
7 * (at your option) any later version.
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See
12 * the GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20 * jfs_imap.c: inode allocation map manager
23 * Each AG has a simple lock which is used to control the serialization of
24 * the AG level lists. This lock should be taken first whenever an AG
25 * level list will be modified or accessed.
27 * Each IAG is locked by obtaining the buffer for the IAG page.
29 * There is also a inode lock for the inode map inode. A read lock needs to
30 * be taken whenever an IAG is read from the map or the global level
31 * information is read. A write lock needs to be taken whenever the global
32 * level information is modified or an atomic operation needs to be used.
34 * If more than one IAG is read at one time, the read lock may not
35 * be given up until all of the IAG's are read. Otherwise, a deadlock
36 * may occur when trying to obtain the read lock while another thread
37 * holding the read lock is waiting on the IAG already being held.
39 * The control page of the inode map is read into memory by diMount().
40 * Thereafter it should only be modified in memory and then it will be
41 * written out when the filesystem is unmounted by diUnmount().
45 #include <linux/buffer_head.h>
46 #include <linux/pagemap.h>
47 #include <linux/quotaops.h>
49 #include "jfs_incore.h"
50 #include "jfs_inode.h"
51 #include "jfs_filsys.h"
52 #include "jfs_dinode.h"
55 #include "jfs_metapage.h"
56 #include "jfs_superblock.h"
57 #include "jfs_debug.h"
60 * __mark_inode_dirty expects inodes to be hashed. Since we don't want
61 * special inodes in the fileset inode space, we hash them to a dummy head
63 static HLIST_HEAD(aggregate_hash
);
68 /* iag free list lock */
69 #define IAGFREE_LOCK_INIT(imap) mutex_init(&imap->im_freelock)
70 #define IAGFREE_LOCK(imap) mutex_lock(&imap->im_freelock)
71 #define IAGFREE_UNLOCK(imap) mutex_unlock(&imap->im_freelock)
73 /* per ag iag list locks */
74 #define AG_LOCK_INIT(imap,index) mutex_init(&(imap->im_aglock[index]))
75 #define AG_LOCK(imap,agno) mutex_lock(&imap->im_aglock[agno])
76 #define AG_UNLOCK(imap,agno) mutex_unlock(&imap->im_aglock[agno])
81 static int diAllocAG(struct inomap
*, int, bool, struct inode
*);
82 static int diAllocAny(struct inomap
*, int, bool, struct inode
*);
83 static int diAllocBit(struct inomap
*, struct iag
*, int);
84 static int diAllocExt(struct inomap
*, int, struct inode
*);
85 static int diAllocIno(struct inomap
*, int, struct inode
*);
86 static int diFindFree(u32
, int);
87 static int diNewExt(struct inomap
*, struct iag
*, int);
88 static int diNewIAG(struct inomap
*, int *, int, struct metapage
**);
89 static void duplicateIXtree(struct super_block
*, s64
, int, s64
*);
91 static int diIAGRead(struct inomap
* imap
, int, struct metapage
**);
92 static int copy_from_dinode(struct dinode
*, struct inode
*);
93 static void copy_to_dinode(struct dinode
*, struct inode
*);
98 * FUNCTION: initialize the incore inode map control structures for
99 * a fileset or aggregate init time.
101 * the inode map's control structure (dinomap) is
102 * brought in from disk and placed in virtual memory.
105 * ipimap - pointer to inode map inode for the aggregate or fileset.
109 * -ENOMEM - insufficient free virtual memory.
112 int diMount(struct inode
*ipimap
)
117 struct dinomap_disk
*dinom_le
;
120 * allocate/initialize the in-memory inode map control structure
122 /* allocate the in-memory inode map control structure. */
123 imap
= kmalloc(sizeof(struct inomap
), GFP_KERNEL
);
125 jfs_err("diMount: kmalloc returned NULL!");
129 /* read the on-disk inode map control structure. */
131 mp
= read_metapage(ipimap
,
132 IMAPBLKNO
<< JFS_SBI(ipimap
->i_sb
)->l2nbperpage
,
139 /* copy the on-disk version to the in-memory version. */
140 dinom_le
= (struct dinomap_disk
*) mp
->data
;
141 imap
->im_freeiag
= le32_to_cpu(dinom_le
->in_freeiag
);
142 imap
->im_nextiag
= le32_to_cpu(dinom_le
->in_nextiag
);
143 atomic_set(&imap
->im_numinos
, le32_to_cpu(dinom_le
->in_numinos
));
144 atomic_set(&imap
->im_numfree
, le32_to_cpu(dinom_le
->in_numfree
));
145 imap
->im_nbperiext
= le32_to_cpu(dinom_le
->in_nbperiext
);
146 imap
->im_l2nbperiext
= le32_to_cpu(dinom_le
->in_l2nbperiext
);
147 for (index
= 0; index
< MAXAG
; index
++) {
148 imap
->im_agctl
[index
].inofree
=
149 le32_to_cpu(dinom_le
->in_agctl
[index
].inofree
);
150 imap
->im_agctl
[index
].extfree
=
151 le32_to_cpu(dinom_le
->in_agctl
[index
].extfree
);
152 imap
->im_agctl
[index
].numinos
=
153 le32_to_cpu(dinom_le
->in_agctl
[index
].numinos
);
154 imap
->im_agctl
[index
].numfree
=
155 le32_to_cpu(dinom_le
->in_agctl
[index
].numfree
);
158 /* release the buffer. */
159 release_metapage(mp
);
162 * allocate/initialize inode allocation map locks
164 /* allocate and init iag free list lock */
165 IAGFREE_LOCK_INIT(imap
);
167 /* allocate and init ag list locks */
168 for (index
= 0; index
< MAXAG
; index
++) {
169 AG_LOCK_INIT(imap
, index
);
172 /* bind the inode map inode and inode map control structure
175 imap
->im_ipimap
= ipimap
;
176 JFS_IP(ipimap
)->i_imap
= imap
;
185 * FUNCTION: write to disk the incore inode map control structures for
186 * a fileset or aggregate at unmount time.
189 * ipimap - pointer to inode map inode for the aggregate or fileset.
193 * -ENOMEM - insufficient free virtual memory.
196 int diUnmount(struct inode
*ipimap
, int mounterror
)
198 struct inomap
*imap
= JFS_IP(ipimap
)->i_imap
;
201 * update the on-disk inode map control structure
204 if (!(mounterror
|| isReadOnly(ipimap
)))
208 * Invalidate the page cache buffers
210 truncate_inode_pages(ipimap
->i_mapping
, 0);
213 * free in-memory control structure
224 int diSync(struct inode
*ipimap
)
226 struct dinomap_disk
*dinom_le
;
227 struct inomap
*imp
= JFS_IP(ipimap
)->i_imap
;
232 * write imap global conrol page
234 /* read the on-disk inode map control structure */
235 mp
= get_metapage(ipimap
,
236 IMAPBLKNO
<< JFS_SBI(ipimap
->i_sb
)->l2nbperpage
,
239 jfs_err("diSync: get_metapage failed!");
243 /* copy the in-memory version to the on-disk version */
244 dinom_le
= (struct dinomap_disk
*) mp
->data
;
245 dinom_le
->in_freeiag
= cpu_to_le32(imp
->im_freeiag
);
246 dinom_le
->in_nextiag
= cpu_to_le32(imp
->im_nextiag
);
247 dinom_le
->in_numinos
= cpu_to_le32(atomic_read(&imp
->im_numinos
));
248 dinom_le
->in_numfree
= cpu_to_le32(atomic_read(&imp
->im_numfree
));
249 dinom_le
->in_nbperiext
= cpu_to_le32(imp
->im_nbperiext
);
250 dinom_le
->in_l2nbperiext
= cpu_to_le32(imp
->im_l2nbperiext
);
251 for (index
= 0; index
< MAXAG
; index
++) {
252 dinom_le
->in_agctl
[index
].inofree
=
253 cpu_to_le32(imp
->im_agctl
[index
].inofree
);
254 dinom_le
->in_agctl
[index
].extfree
=
255 cpu_to_le32(imp
->im_agctl
[index
].extfree
);
256 dinom_le
->in_agctl
[index
].numinos
=
257 cpu_to_le32(imp
->im_agctl
[index
].numinos
);
258 dinom_le
->in_agctl
[index
].numfree
=
259 cpu_to_le32(imp
->im_agctl
[index
].numfree
);
262 /* write out the control structure */
266 * write out dirty pages of imap
268 filemap_write_and_wait(ipimap
->i_mapping
);
270 diWriteSpecial(ipimap
, 0);
279 * FUNCTION: initialize an incore inode from disk.
281 * on entry, the specifed incore inode should itself
282 * specify the disk inode number corresponding to the
283 * incore inode (i.e. i_number should be initialized).
285 * this routine handles incore inode initialization for
286 * both "special" and "regular" inodes. special inodes
287 * are those required early in the mount process and
288 * require special handling since much of the file system
289 * is not yet initialized. these "special" inodes are
290 * identified by a NULL inode map inode pointer and are
291 * actually initialized by a call to diReadSpecial().
293 * for regular inodes, the iag describing the disk inode
294 * is read from disk to determine the inode extent address
295 * for the disk inode. with the inode extent address in
296 * hand, the page of the extent that contains the disk
297 * inode is read and the disk inode is copied to the
301 * ip - pointer to incore inode to be initialized from disk.
306 * -ENOMEM - insufficient memory
309 int diRead(struct inode
*ip
)
311 struct jfs_sb_info
*sbi
= JFS_SBI(ip
->i_sb
);
312 int iagno
, ino
, extno
, rc
;
313 struct inode
*ipimap
;
321 unsigned long pageno
;
324 jfs_info("diRead: ino = %ld", ip
->i_ino
);
326 ipimap
= sbi
->ipimap
;
327 JFS_IP(ip
)->ipimap
= ipimap
;
329 /* determine the iag number for this inode (number) */
330 iagno
= INOTOIAG(ip
->i_ino
);
333 imap
= JFS_IP(ipimap
)->i_imap
;
334 IREAD_LOCK(ipimap
, RDWRLOCK_IMAP
);
335 rc
= diIAGRead(imap
, iagno
, &mp
);
336 IREAD_UNLOCK(ipimap
);
338 jfs_err("diRead: diIAGRead returned %d", rc
);
342 iagp
= (struct iag
*) mp
->data
;
344 /* determine inode extent that holds the disk inode */
345 ino
= ip
->i_ino
& (INOSPERIAG
- 1);
346 extno
= ino
>> L2INOSPEREXT
;
348 if ((lengthPXD(&iagp
->inoext
[extno
]) != imap
->im_nbperiext
) ||
349 (addressPXD(&iagp
->inoext
[extno
]) == 0)) {
350 release_metapage(mp
);
354 /* get disk block number of the page within the inode extent
355 * that holds the disk inode.
357 blkno
= INOPBLK(&iagp
->inoext
[extno
], ino
, sbi
->l2nbperpage
);
359 /* get the ag for the iag */
360 agstart
= le64_to_cpu(iagp
->agstart
);
362 release_metapage(mp
);
364 rel_inode
= (ino
& (INOSPERPAGE
- 1));
365 pageno
= blkno
>> sbi
->l2nbperpage
;
367 if ((block_offset
= ((u32
) blkno
& (sbi
->nbperpage
- 1)))) {
369 * OS/2 didn't always align inode extents on page boundaries
372 (sbi
->nbperpage
- block_offset
) << sbi
->l2niperblk
;
374 if (rel_inode
< inodes_left
)
375 rel_inode
+= block_offset
<< sbi
->l2niperblk
;
378 rel_inode
-= inodes_left
;
382 /* read the page of disk inode */
383 mp
= read_metapage(ipimap
, pageno
<< sbi
->l2nbperpage
, PSIZE
, 1);
385 jfs_err("diRead: read_metapage failed");
389 /* locate the disk inode requested */
390 dp
= (struct dinode
*) mp
->data
;
393 if (ip
->i_ino
!= le32_to_cpu(dp
->di_number
)) {
394 jfs_error(ip
->i_sb
, "diRead: i_ino != di_number");
396 } else if (le32_to_cpu(dp
->di_nlink
) == 0)
399 /* copy the disk inode to the in-memory inode */
400 rc
= copy_from_dinode(dp
, ip
);
402 release_metapage(mp
);
404 /* set the ag for the inode */
405 JFS_IP(ip
)->agno
= BLKTOAG(agstart
, sbi
);
406 JFS_IP(ip
)->active_ag
= -1;
413 * NAME: diReadSpecial()
415 * FUNCTION: initialize a 'special' inode from disk.
417 * this routines handles aggregate level inodes. The
418 * inode cache cannot differentiate between the
419 * aggregate inodes and the filesystem inodes, so we
420 * handle these here. We don't actually use the aggregate
421 * inode map, since these inodes are at a fixed location
422 * and in some cases the aggregate inode map isn't initialized
426 * sb - filesystem superblock
427 * inum - aggregate inode number
428 * secondary - 1 if secondary aggregate inode table
431 * new inode - success
434 struct inode
*diReadSpecial(struct super_block
*sb
, ino_t inum
, int secondary
)
436 struct jfs_sb_info
*sbi
= JFS_SBI(sb
);
444 jfs_err("diReadSpecial: new_inode returned NULL!");
449 address
= addressPXD(&sbi
->ait2
) >> sbi
->l2nbperpage
;
450 JFS_IP(ip
)->ipimap
= sbi
->ipaimap2
;
452 address
= AITBL_OFF
>> L2PSIZE
;
453 JFS_IP(ip
)->ipimap
= sbi
->ipaimap
;
456 ASSERT(inum
< INOSPEREXT
);
460 address
+= inum
>> 3; /* 8 inodes per 4K page */
462 /* read the page of fixed disk inode (AIT) in raw mode */
463 mp
= read_metapage(ip
, address
<< sbi
->l2nbperpage
, PSIZE
, 1);
465 ip
->i_nlink
= 1; /* Don't want iput() deleting it */
470 /* get the pointer to the disk inode of interest */
471 dp
= (struct dinode
*) (mp
->data
);
472 dp
+= inum
% 8; /* 8 inodes per 4K page */
474 /* copy on-disk inode to in-memory inode */
475 if ((copy_from_dinode(dp
, ip
)) != 0) {
476 /* handle bad return by returning NULL for ip */
477 ip
->i_nlink
= 1; /* Don't want iput() deleting it */
479 /* release the page */
480 release_metapage(mp
);
485 ip
->i_mapping
->a_ops
= &jfs_metapage_aops
;
486 mapping_set_gfp_mask(ip
->i_mapping
, GFP_NOFS
);
488 /* Allocations to metadata inodes should not affect quotas */
489 ip
->i_flags
|= S_NOQUOTA
;
491 if ((inum
== FILESYSTEM_I
) && (JFS_IP(ip
)->ipimap
== sbi
->ipaimap
)) {
492 sbi
->gengen
= le32_to_cpu(dp
->di_gengen
);
493 sbi
->inostamp
= le32_to_cpu(dp
->di_inostamp
);
496 /* release the page */
497 release_metapage(mp
);
499 hlist_add_head(&ip
->i_hash
, &aggregate_hash
);
505 * NAME: diWriteSpecial()
507 * FUNCTION: Write the special inode to disk
511 * secondary - 1 if secondary aggregate inode table
513 * RETURN VALUES: none
516 void diWriteSpecial(struct inode
*ip
, int secondary
)
518 struct jfs_sb_info
*sbi
= JFS_SBI(ip
->i_sb
);
521 ino_t inum
= ip
->i_ino
;
525 address
= addressPXD(&sbi
->ait2
) >> sbi
->l2nbperpage
;
527 address
= AITBL_OFF
>> L2PSIZE
;
529 ASSERT(inum
< INOSPEREXT
);
531 address
+= inum
>> 3; /* 8 inodes per 4K page */
533 /* read the page of fixed disk inode (AIT) in raw mode */
534 mp
= read_metapage(ip
, address
<< sbi
->l2nbperpage
, PSIZE
, 1);
536 jfs_err("diWriteSpecial: failed to read aggregate inode "
541 /* get the pointer to the disk inode of interest */
542 dp
= (struct dinode
*) (mp
->data
);
543 dp
+= inum
% 8; /* 8 inodes per 4K page */
545 /* copy on-disk inode to in-memory inode */
546 copy_to_dinode(dp
, ip
);
547 memcpy(&dp
->di_xtroot
, &JFS_IP(ip
)->i_xtroot
, 288);
549 if (inum
== FILESYSTEM_I
)
550 dp
->di_gengen
= cpu_to_le32(sbi
->gengen
);
557 * NAME: diFreeSpecial()
559 * FUNCTION: Free allocated space for special inode
561 void diFreeSpecial(struct inode
*ip
)
564 jfs_err("diFreeSpecial called with NULL ip!");
567 filemap_write_and_wait(ip
->i_mapping
);
568 truncate_inode_pages(ip
->i_mapping
, 0);
577 * FUNCTION: write the on-disk inode portion of the in-memory inode
578 * to its corresponding on-disk inode.
580 * on entry, the specifed incore inode should itself
581 * specify the disk inode number corresponding to the
582 * incore inode (i.e. i_number should be initialized).
584 * the inode contains the inode extent address for the disk
585 * inode. with the inode extent address in hand, the
586 * page of the extent that contains the disk inode is
587 * read and the disk inode portion of the incore inode
588 * is copied to the disk inode.
591 * tid - transacation id
592 * ip - pointer to incore inode to be written to the inode extent.
598 int diWrite(tid_t tid
, struct inode
*ip
)
600 struct jfs_sb_info
*sbi
= JFS_SBI(ip
->i_sb
);
601 struct jfs_inode_info
*jfs_ip
= JFS_IP(ip
);
609 unsigned long pageno
;
612 struct inode
*ipimap
;
615 struct tlock
*ditlck
, *tlck
;
616 struct linelock
*dilinelock
, *ilinelock
;
620 ipimap
= jfs_ip
->ipimap
;
622 ino
= ip
->i_ino
& (INOSPERIAG
- 1);
624 if (!addressPXD(&(jfs_ip
->ixpxd
)) ||
625 (lengthPXD(&(jfs_ip
->ixpxd
)) !=
626 JFS_IP(ipimap
)->i_imap
->im_nbperiext
)) {
627 jfs_error(ip
->i_sb
, "diWrite: ixpxd invalid");
632 * read the page of disk inode containing the specified inode:
634 /* compute the block address of the page */
635 blkno
= INOPBLK(&(jfs_ip
->ixpxd
), ino
, sbi
->l2nbperpage
);
637 rel_inode
= (ino
& (INOSPERPAGE
- 1));
638 pageno
= blkno
>> sbi
->l2nbperpage
;
640 if ((block_offset
= ((u32
) blkno
& (sbi
->nbperpage
- 1)))) {
642 * OS/2 didn't always align inode extents on page boundaries
645 (sbi
->nbperpage
- block_offset
) << sbi
->l2niperblk
;
647 if (rel_inode
< inodes_left
)
648 rel_inode
+= block_offset
<< sbi
->l2niperblk
;
651 rel_inode
-= inodes_left
;
654 /* read the page of disk inode */
656 mp
= read_metapage(ipimap
, pageno
<< sbi
->l2nbperpage
, PSIZE
, 1);
660 /* get the pointer to the disk inode */
661 dp
= (struct dinode
*) mp
->data
;
664 dioffset
= (ino
& (INOSPERPAGE
- 1)) << L2DISIZE
;
667 * acquire transaction lock on the on-disk inode;
668 * N.B. tlock is acquired on ipimap not ip;
671 txLock(tid
, ipimap
, mp
, tlckINODE
| tlckENTRY
)) == NULL
)
673 dilinelock
= (struct linelock
*) & ditlck
->lock
;
676 * copy btree root from in-memory inode to on-disk inode
678 * (tlock is taken from inline B+-tree root in in-memory
679 * inode when the B+-tree root is updated, which is pointed
680 * by jfs_ip->blid as well as being on tx tlock list)
682 * further processing of btree root is based on the copy
683 * in in-memory inode, where txLog() will log from, and,
684 * for xtree root, txUpdateMap() will update map and reset
688 if (S_ISDIR(ip
->i_mode
) && (lid
= jfs_ip
->xtlid
)) {
690 * This is the special xtree inside the directory for storing
691 * the directory table
697 tlck
= lid_to_tlock(lid
);
698 assert(tlck
->type
& tlckXTREE
);
699 tlck
->type
|= tlckBTROOT
;
701 ilinelock
= (struct linelock
*) & tlck
->lock
;
704 * copy xtree root from inode to dinode:
706 p
= &jfs_ip
->i_xtroot
;
707 xp
= (xtpage_t
*) &dp
->di_dirtable
;
709 for (n
= 0; n
< ilinelock
->index
; n
++, lv
++) {
710 memcpy(&xp
->xad
[lv
->offset
], &p
->xad
[lv
->offset
],
711 lv
->length
<< L2XTSLOTSIZE
);
714 /* reset on-disk (metadata page) xtree XAD_NEW bit */
715 xad
= &xp
->xad
[XTENTRYSTART
];
716 for (n
= XTENTRYSTART
;
717 n
< le16_to_cpu(xp
->header
.nextindex
); n
++, xad
++)
718 if (xad
->flag
& (XAD_NEW
| XAD_EXTENDED
))
719 xad
->flag
&= ~(XAD_NEW
| XAD_EXTENDED
);
722 if ((lid
= jfs_ip
->blid
) == 0)
726 tlck
= lid_to_tlock(lid
);
728 tlck
->type
|= tlckBTROOT
;
730 ilinelock
= (struct linelock
*) & tlck
->lock
;
733 * regular file: 16 byte (XAD slot) granularity
735 if (type
& tlckXTREE
) {
740 * copy xtree root from inode to dinode:
742 p
= &jfs_ip
->i_xtroot
;
745 for (n
= 0; n
< ilinelock
->index
; n
++, lv
++) {
746 memcpy(&xp
->xad
[lv
->offset
], &p
->xad
[lv
->offset
],
747 lv
->length
<< L2XTSLOTSIZE
);
750 /* reset on-disk (metadata page) xtree XAD_NEW bit */
751 xad
= &xp
->xad
[XTENTRYSTART
];
752 for (n
= XTENTRYSTART
;
753 n
< le16_to_cpu(xp
->header
.nextindex
); n
++, xad
++)
754 if (xad
->flag
& (XAD_NEW
| XAD_EXTENDED
))
755 xad
->flag
&= ~(XAD_NEW
| XAD_EXTENDED
);
758 * directory: 32 byte (directory entry slot) granularity
760 else if (type
& tlckDTREE
) {
764 * copy dtree root from inode to dinode:
766 p
= (dtpage_t
*) &jfs_ip
->i_dtroot
;
767 xp
= (dtpage_t
*) & dp
->di_dtroot
;
769 for (n
= 0; n
< ilinelock
->index
; n
++, lv
++) {
770 memcpy(&xp
->slot
[lv
->offset
], &p
->slot
[lv
->offset
],
771 lv
->length
<< L2DTSLOTSIZE
);
774 jfs_err("diWrite: UFO tlock");
779 * copy inline symlink from in-memory inode to on-disk inode
781 if (S_ISLNK(ip
->i_mode
) && ip
->i_size
< IDATASIZE
) {
782 lv
= & dilinelock
->lv
[dilinelock
->index
];
783 lv
->offset
= (dioffset
+ 2 * 128) >> L2INODESLOTSIZE
;
785 memcpy(&dp
->di_fastsymlink
, jfs_ip
->i_inline
, IDATASIZE
);
789 * copy inline data from in-memory inode to on-disk inode:
790 * 128 byte slot granularity
792 if (test_cflag(COMMIT_Inlineea
, ip
)) {
793 lv
= & dilinelock
->lv
[dilinelock
->index
];
794 lv
->offset
= (dioffset
+ 3 * 128) >> L2INODESLOTSIZE
;
796 memcpy(&dp
->di_inlineea
, jfs_ip
->i_inline_ea
, INODESLOTSIZE
);
799 clear_cflag(COMMIT_Inlineea
, ip
);
803 * lock/copy inode base: 128 byte slot granularity
805 lv
= & dilinelock
->lv
[dilinelock
->index
];
806 lv
->offset
= dioffset
>> L2INODESLOTSIZE
;
807 copy_to_dinode(dp
, ip
);
808 if (test_and_clear_cflag(COMMIT_Dirtable
, ip
)) {
810 memcpy(&dp
->di_dirtable
, &jfs_ip
->i_dirtable
, 96);
815 /* release the buffer holding the updated on-disk inode.
816 * the buffer will be later written by commit processing.
827 * FUNCTION: free a specified inode from the inode working map
828 * for a fileset or aggregate.
830 * if the inode to be freed represents the first (only)
831 * free inode within the iag, the iag will be placed on
832 * the ag free inode list.
834 * freeing the inode will cause the inode extent to be
835 * freed if the inode is the only allocated inode within
836 * the extent. in this case all the disk resource backing
837 * up the inode extent will be freed. in addition, the iag
838 * will be placed on the ag extent free list if the extent
839 * is the first free extent in the iag. if freeing the
840 * extent also means that no free inodes will exist for
841 * the iag, the iag will also be removed from the ag free
844 * the iag describing the inode will be freed if the extent
845 * is to be freed and it is the only backed extent within
846 * the iag. in this case, the iag will be removed from the
847 * ag free extent list and ag free inode list and placed on
848 * the inode map's free iag list.
850 * a careful update approach is used to provide consistency
851 * in the face of updates to multiple buffers. under this
852 * approach, all required buffers are obtained before making
853 * any updates and are held until all updates are complete.
856 * ip - inode to be freed.
862 int diFree(struct inode
*ip
)
865 ino_t inum
= ip
->i_ino
;
866 struct iag
*iagp
, *aiagp
, *biagp
, *ciagp
, *diagp
;
867 struct metapage
*mp
, *amp
, *bmp
, *cmp
, *dmp
;
868 int iagno
, ino
, extno
, bitno
, sword
, agno
;
871 struct inode
*ipimap
= JFS_SBI(ip
->i_sb
)->ipimap
;
872 struct inomap
*imap
= JFS_IP(ipimap
)->i_imap
;
875 struct inode
*iplist
[3];
877 struct pxd_lock
*pxdlock
;
880 * This is just to suppress compiler warnings. The same logic that
881 * references these variables is used to initialize them.
883 aiagp
= biagp
= ciagp
= diagp
= NULL
;
885 /* get the iag number containing the inode.
887 iagno
= INOTOIAG(inum
);
889 /* make sure that the iag is contained within
892 if (iagno
>= imap
->im_nextiag
) {
893 dump_mem("imap", imap
, 32);
895 "diFree: inum = %d, iagno = %d, nextiag = %d",
896 (uint
) inum
, iagno
, imap
->im_nextiag
);
900 /* get the allocation group for this ino.
902 agno
= JFS_IP(ip
)->agno
;
904 /* Lock the AG specific inode map information
908 /* Obtain read lock in imap inode. Don't release it until we have
909 * read all of the IAG's that we are going to.
911 IREAD_LOCK(ipimap
, RDWRLOCK_IMAP
);
915 if ((rc
= diIAGRead(imap
, iagno
, &mp
))) {
916 IREAD_UNLOCK(ipimap
);
917 AG_UNLOCK(imap
, agno
);
920 iagp
= (struct iag
*) mp
->data
;
922 /* get the inode number and extent number of the inode within
923 * the iag and the inode number within the extent.
925 ino
= inum
& (INOSPERIAG
- 1);
926 extno
= ino
>> L2INOSPEREXT
;
927 bitno
= ino
& (INOSPEREXT
- 1);
928 mask
= HIGHORDER
>> bitno
;
930 if (!(le32_to_cpu(iagp
->wmap
[extno
]) & mask
)) {
932 "diFree: wmap shows inode already free");
935 if (!addressPXD(&iagp
->inoext
[extno
])) {
936 release_metapage(mp
);
937 IREAD_UNLOCK(ipimap
);
938 AG_UNLOCK(imap
, agno
);
939 jfs_error(ip
->i_sb
, "diFree: invalid inoext");
943 /* compute the bitmap for the extent reflecting the freed inode.
945 bitmap
= le32_to_cpu(iagp
->wmap
[extno
]) & ~mask
;
947 if (imap
->im_agctl
[agno
].numfree
> imap
->im_agctl
[agno
].numinos
) {
948 release_metapage(mp
);
949 IREAD_UNLOCK(ipimap
);
950 AG_UNLOCK(imap
, agno
);
951 jfs_error(ip
->i_sb
, "diFree: numfree > numinos");
955 * inode extent still has some inodes or below low water mark:
956 * keep the inode extent;
959 imap
->im_agctl
[agno
].numfree
< 96 ||
960 (imap
->im_agctl
[agno
].numfree
< 288 &&
961 (((imap
->im_agctl
[agno
].numfree
* 100) /
962 imap
->im_agctl
[agno
].numinos
) <= 25))) {
963 /* if the iag currently has no free inodes (i.e.,
964 * the inode being freed is the first free inode of iag),
965 * insert the iag at head of the inode free list for the ag.
967 if (iagp
->nfreeinos
== 0) {
968 /* check if there are any iags on the ag inode
969 * free list. if so, read the first one so that
970 * we can link the current iag onto the list at
973 if ((fwd
= imap
->im_agctl
[agno
].inofree
) >= 0) {
974 /* read the iag that currently is the head
977 if ((rc
= diIAGRead(imap
, fwd
, &
))) {
978 IREAD_UNLOCK(ipimap
);
979 AG_UNLOCK(imap
, agno
);
980 release_metapage(mp
);
983 aiagp
= (struct iag
*) amp
->data
;
985 /* make current head point back to the iag.
987 aiagp
->inofreeback
= cpu_to_le32(iagno
);
992 /* iag points forward to current head and iag
993 * becomes the new head of the list.
996 cpu_to_le32(imap
->im_agctl
[agno
].inofree
);
997 iagp
->inofreeback
= cpu_to_le32(-1);
998 imap
->im_agctl
[agno
].inofree
= iagno
;
1000 IREAD_UNLOCK(ipimap
);
1002 /* update the free inode summary map for the extent if
1003 * freeing the inode means the extent will now have free
1004 * inodes (i.e., the inode being freed is the first free
1007 if (iagp
->wmap
[extno
] == cpu_to_le32(ONES
)) {
1008 sword
= extno
>> L2EXTSPERSUM
;
1009 bitno
= extno
& (EXTSPERSUM
- 1);
1010 iagp
->inosmap
[sword
] &=
1011 cpu_to_le32(~(HIGHORDER
>> bitno
));
1014 /* update the bitmap.
1016 iagp
->wmap
[extno
] = cpu_to_le32(bitmap
);
1018 /* update the free inode counts at the iag, ag and
1022 cpu_to_le32(le32_to_cpu(iagp
->nfreeinos
) + 1);
1023 imap
->im_agctl
[agno
].numfree
+= 1;
1024 atomic_inc(&imap
->im_numfree
);
1026 /* release the AG inode map lock
1028 AG_UNLOCK(imap
, agno
);
1038 * inode extent has become free and above low water mark:
1039 * free the inode extent;
1043 * prepare to update iag list(s) (careful update step 1)
1045 amp
= bmp
= cmp
= dmp
= NULL
;
1048 /* check if the iag currently has no free extents. if so,
1049 * it will be placed on the head of the ag extent free list.
1051 if (iagp
->nfreeexts
== 0) {
1052 /* check if the ag extent free list has any iags.
1053 * if so, read the iag at the head of the list now.
1054 * this (head) iag will be updated later to reflect
1055 * the addition of the current iag at the head of
1058 if ((fwd
= imap
->im_agctl
[agno
].extfree
) >= 0) {
1059 if ((rc
= diIAGRead(imap
, fwd
, &
)))
1061 aiagp
= (struct iag
*) amp
->data
;
1064 /* iag has free extents. check if the addition of a free
1065 * extent will cause all extents to be free within this
1066 * iag. if so, the iag will be removed from the ag extent
1067 * free list and placed on the inode map's free iag list.
1069 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
- 1)) {
1070 /* in preparation for removing the iag from the
1071 * ag extent free list, read the iags preceeding
1072 * and following the iag on the ag extent free
1075 if ((fwd
= le32_to_cpu(iagp
->extfreefwd
)) >= 0) {
1076 if ((rc
= diIAGRead(imap
, fwd
, &
)))
1078 aiagp
= (struct iag
*) amp
->data
;
1081 if ((back
= le32_to_cpu(iagp
->extfreeback
)) >= 0) {
1082 if ((rc
= diIAGRead(imap
, back
, &bmp
)))
1084 biagp
= (struct iag
*) bmp
->data
;
1089 /* remove the iag from the ag inode free list if freeing
1090 * this extent cause the iag to have no free inodes.
1092 if (iagp
->nfreeinos
== cpu_to_le32(INOSPEREXT
- 1)) {
1093 int inofreeback
= le32_to_cpu(iagp
->inofreeback
);
1094 int inofreefwd
= le32_to_cpu(iagp
->inofreefwd
);
1096 /* in preparation for removing the iag from the
1097 * ag inode free list, read the iags preceeding
1098 * and following the iag on the ag inode free
1099 * list. before reading these iags, we must make
1100 * sure that we already don't have them in hand
1101 * from up above, since re-reading an iag (buffer)
1102 * we are currently holding would cause a deadlock.
1104 if (inofreefwd
>= 0) {
1106 if (inofreefwd
== fwd
)
1107 ciagp
= (struct iag
*) amp
->data
;
1108 else if (inofreefwd
== back
)
1109 ciagp
= (struct iag
*) bmp
->data
;
1112 diIAGRead(imap
, inofreefwd
, &cmp
)))
1114 ciagp
= (struct iag
*) cmp
->data
;
1116 assert(ciagp
!= NULL
);
1119 if (inofreeback
>= 0) {
1120 if (inofreeback
== fwd
)
1121 diagp
= (struct iag
*) amp
->data
;
1122 else if (inofreeback
== back
)
1123 diagp
= (struct iag
*) bmp
->data
;
1126 diIAGRead(imap
, inofreeback
, &dmp
)))
1128 diagp
= (struct iag
*) dmp
->data
;
1130 assert(diagp
!= NULL
);
1134 IREAD_UNLOCK(ipimap
);
1137 * invalidate any page of the inode extent freed from buffer cache;
1139 freepxd
= iagp
->inoext
[extno
];
1140 invalidate_pxd_metapages(ip
, freepxd
);
1143 * update iag list(s) (careful update step 2)
1145 /* add the iag to the ag extent free list if this is the
1146 * first free extent for the iag.
1148 if (iagp
->nfreeexts
== 0) {
1150 aiagp
->extfreeback
= cpu_to_le32(iagno
);
1153 cpu_to_le32(imap
->im_agctl
[agno
].extfree
);
1154 iagp
->extfreeback
= cpu_to_le32(-1);
1155 imap
->im_agctl
[agno
].extfree
= iagno
;
1157 /* remove the iag from the ag extent list if all extents
1158 * are now free and place it on the inode map iag free list.
1160 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
- 1)) {
1162 aiagp
->extfreeback
= iagp
->extfreeback
;
1165 biagp
->extfreefwd
= iagp
->extfreefwd
;
1167 imap
->im_agctl
[agno
].extfree
=
1168 le32_to_cpu(iagp
->extfreefwd
);
1170 iagp
->extfreefwd
= iagp
->extfreeback
= cpu_to_le32(-1);
1173 iagp
->iagfree
= cpu_to_le32(imap
->im_freeiag
);
1174 imap
->im_freeiag
= iagno
;
1175 IAGFREE_UNLOCK(imap
);
1179 /* remove the iag from the ag inode free list if freeing
1180 * this extent causes the iag to have no free inodes.
1182 if (iagp
->nfreeinos
== cpu_to_le32(INOSPEREXT
- 1)) {
1183 if ((int) le32_to_cpu(iagp
->inofreefwd
) >= 0)
1184 ciagp
->inofreeback
= iagp
->inofreeback
;
1186 if ((int) le32_to_cpu(iagp
->inofreeback
) >= 0)
1187 diagp
->inofreefwd
= iagp
->inofreefwd
;
1189 imap
->im_agctl
[agno
].inofree
=
1190 le32_to_cpu(iagp
->inofreefwd
);
1192 iagp
->inofreefwd
= iagp
->inofreeback
= cpu_to_le32(-1);
1195 /* update the inode extent address and working map
1196 * to reflect the free extent.
1197 * the permanent map should have been updated already
1198 * for the inode being freed.
1200 if (iagp
->pmap
[extno
] != 0) {
1201 jfs_error(ip
->i_sb
, "diFree: the pmap does not show inode free");
1203 iagp
->wmap
[extno
] = 0;
1204 PXDlength(&iagp
->inoext
[extno
], 0);
1205 PXDaddress(&iagp
->inoext
[extno
], 0);
1207 /* update the free extent and free inode summary maps
1208 * to reflect the freed extent.
1209 * the inode summary map is marked to indicate no inodes
1210 * available for the freed extent.
1212 sword
= extno
>> L2EXTSPERSUM
;
1213 bitno
= extno
& (EXTSPERSUM
- 1);
1214 mask
= HIGHORDER
>> bitno
;
1215 iagp
->inosmap
[sword
] |= cpu_to_le32(mask
);
1216 iagp
->extsmap
[sword
] &= cpu_to_le32(~mask
);
1218 /* update the number of free inodes and number of free extents
1221 iagp
->nfreeinos
= cpu_to_le32(le32_to_cpu(iagp
->nfreeinos
) -
1223 iagp
->nfreeexts
= cpu_to_le32(le32_to_cpu(iagp
->nfreeexts
) + 1);
1225 /* update the number of free inodes and backed inodes
1226 * at the ag and inode map level.
1228 imap
->im_agctl
[agno
].numfree
-= (INOSPEREXT
- 1);
1229 imap
->im_agctl
[agno
].numinos
-= INOSPEREXT
;
1230 atomic_sub(INOSPEREXT
- 1, &imap
->im_numfree
);
1231 atomic_sub(INOSPEREXT
, &imap
->im_numinos
);
1234 write_metapage(amp
);
1236 write_metapage(bmp
);
1238 write_metapage(cmp
);
1240 write_metapage(dmp
);
1243 * start transaction to update block allocation map
1244 * for the inode extent freed;
1246 * N.B. AG_LOCK is released and iag will be released below, and
1247 * other thread may allocate inode from/reusing the ixad freed
1248 * BUT with new/different backing inode extent from the extent
1249 * to be freed by the transaction;
1251 tid
= txBegin(ipimap
->i_sb
, COMMIT_FORCE
);
1252 mutex_lock(&JFS_IP(ipimap
)->commit_mutex
);
1254 /* acquire tlock of the iag page of the freed ixad
1255 * to force the page NOHOMEOK (even though no data is
1256 * logged from the iag page) until NOREDOPAGE|FREEXTENT log
1257 * for the free of the extent is committed;
1258 * write FREEXTENT|NOREDOPAGE log record
1259 * N.B. linelock is overlaid as freed extent descriptor;
1261 tlck
= txLock(tid
, ipimap
, mp
, tlckINODE
| tlckFREE
);
1262 pxdlock
= (struct pxd_lock
*) & tlck
->lock
;
1263 pxdlock
->flag
= mlckFREEPXD
;
1264 pxdlock
->pxd
= freepxd
;
1272 * logredo needs the IAG number and IAG extent index in order
1273 * to ensure that the IMap is consistent. The least disruptive
1274 * way to pass these values through to the transaction manager
1275 * is in the iplist array.
1277 * It's not pretty, but it works.
1279 iplist
[1] = (struct inode
*) (size_t)iagno
;
1280 iplist
[2] = (struct inode
*) (size_t)extno
;
1282 rc
= txCommit(tid
, 1, &iplist
[0], COMMIT_FORCE
);
1285 mutex_unlock(&JFS_IP(ipimap
)->commit_mutex
);
1287 /* unlock the AG inode map information */
1288 AG_UNLOCK(imap
, agno
);
1293 IREAD_UNLOCK(ipimap
);
1296 release_metapage(amp
);
1298 release_metapage(bmp
);
1300 release_metapage(cmp
);
1302 release_metapage(dmp
);
1304 AG_UNLOCK(imap
, agno
);
1306 release_metapage(mp
);
1312 * There are several places in the diAlloc* routines where we initialize
1316 diInitInode(struct inode
*ip
, int iagno
, int ino
, int extno
, struct iag
* iagp
)
1318 struct jfs_sb_info
*sbi
= JFS_SBI(ip
->i_sb
);
1319 struct jfs_inode_info
*jfs_ip
= JFS_IP(ip
);
1321 ip
->i_ino
= (iagno
<< L2INOSPERIAG
) + ino
;
1322 jfs_ip
->ixpxd
= iagp
->inoext
[extno
];
1323 jfs_ip
->agno
= BLKTOAG(le64_to_cpu(iagp
->agstart
), sbi
);
1324 jfs_ip
->active_ag
= -1;
1329 * NAME: diAlloc(pip,dir,ip)
1331 * FUNCTION: allocate a disk inode from the inode working map
1332 * for a fileset or aggregate.
1335 * pip - pointer to incore inode for the parent inode.
1336 * dir - 'true' if the new disk inode is for a directory.
1337 * ip - pointer to a new inode
1341 * -ENOSPC - insufficient disk resources.
1344 int diAlloc(struct inode
*pip
, bool dir
, struct inode
*ip
)
1346 int rc
, ino
, iagno
, addext
, extno
, bitno
, sword
;
1347 int nwords
, rem
, i
, agno
;
1348 u32 mask
, inosmap
, extsmap
;
1349 struct inode
*ipimap
;
1350 struct metapage
*mp
;
1353 struct inomap
*imap
;
1355 /* get the pointers to the inode map inode and the
1356 * corresponding imap control structure.
1358 ipimap
= JFS_SBI(pip
->i_sb
)->ipimap
;
1359 imap
= JFS_IP(ipimap
)->i_imap
;
1360 JFS_IP(ip
)->ipimap
= ipimap
;
1361 JFS_IP(ip
)->fileset
= FILESYSTEM_I
;
1363 /* for a directory, the allocation policy is to start
1364 * at the ag level using the preferred ag.
1367 agno
= dbNextAG(JFS_SBI(pip
->i_sb
)->ipbmap
);
1368 AG_LOCK(imap
, agno
);
1372 /* for files, the policy starts off by trying to allocate from
1373 * the same iag containing the parent disk inode:
1374 * try to allocate the new disk inode close to the parent disk
1375 * inode, using parent disk inode number + 1 as the allocation
1376 * hint. (we use a left-to-right policy to attempt to avoid
1377 * moving backward on the disk.) compute the hint within the
1378 * file system and the iag.
1381 /* get the ag number of this iag */
1382 agno
= JFS_IP(pip
)->agno
;
1384 if (atomic_read(&JFS_SBI(pip
->i_sb
)->bmap
->db_active
[agno
])) {
1386 * There is an open file actively growing. We want to
1387 * allocate new inodes from a different ag to avoid
1388 * fragmentation problems.
1390 agno
= dbNextAG(JFS_SBI(pip
->i_sb
)->ipbmap
);
1391 AG_LOCK(imap
, agno
);
1395 inum
= pip
->i_ino
+ 1;
1396 ino
= inum
& (INOSPERIAG
- 1);
1398 /* back off the hint if it is outside of the iag */
1402 /* lock the AG inode map information */
1403 AG_LOCK(imap
, agno
);
1405 /* Get read lock on imap inode */
1406 IREAD_LOCK(ipimap
, RDWRLOCK_IMAP
);
1408 /* get the iag number and read the iag */
1409 iagno
= INOTOIAG(inum
);
1410 if ((rc
= diIAGRead(imap
, iagno
, &mp
))) {
1411 IREAD_UNLOCK(ipimap
);
1412 AG_UNLOCK(imap
, agno
);
1415 iagp
= (struct iag
*) mp
->data
;
1417 /* determine if new inode extent is allowed to be added to the iag.
1418 * new inode extent can be added to the iag if the ag
1419 * has less than 32 free disk inodes and the iag has free extents.
1421 addext
= (imap
->im_agctl
[agno
].numfree
< 32 && iagp
->nfreeexts
);
1424 * try to allocate from the IAG
1426 /* check if the inode may be allocated from the iag
1427 * (i.e. the inode has free inodes or new extent can be added).
1429 if (iagp
->nfreeinos
|| addext
) {
1430 /* determine the extent number of the hint.
1432 extno
= ino
>> L2INOSPEREXT
;
1434 /* check if the extent containing the hint has backed
1435 * inodes. if so, try to allocate within this extent.
1437 if (addressPXD(&iagp
->inoext
[extno
])) {
1438 bitno
= ino
& (INOSPEREXT
- 1);
1440 diFindFree(le32_to_cpu(iagp
->wmap
[extno
]),
1443 ino
= (extno
<< L2INOSPEREXT
) + bitno
;
1445 /* a free inode (bit) was found within this
1446 * extent, so allocate it.
1448 rc
= diAllocBit(imap
, iagp
, ino
);
1449 IREAD_UNLOCK(ipimap
);
1453 /* set the results of the allocation
1454 * and write the iag.
1456 diInitInode(ip
, iagno
, ino
, extno
,
1458 mark_metapage_dirty(mp
);
1460 release_metapage(mp
);
1462 /* free the AG lock and return.
1464 AG_UNLOCK(imap
, agno
);
1471 EXTSPERIAG
- 1) ? 0 : extno
+ 1;
1475 * no free inodes within the extent containing the hint.
1477 * try to allocate from the backed extents following
1478 * hint or, if appropriate (i.e. addext is true), allocate
1479 * an extent of free inodes at or following the extent
1480 * containing the hint.
1482 * the free inode and free extent summary maps are used
1483 * here, so determine the starting summary map position
1484 * and the number of words we'll have to examine. again,
1485 * the approach is to allocate following the hint, so we
1486 * might have to initially ignore prior bits of the summary
1487 * map that represent extents prior to the extent containing
1488 * the hint and later revisit these bits.
1490 bitno
= extno
& (EXTSPERSUM
- 1);
1491 nwords
= (bitno
== 0) ? SMAPSZ
: SMAPSZ
+ 1;
1492 sword
= extno
>> L2EXTSPERSUM
;
1494 /* mask any prior bits for the starting words of the
1497 mask
= ONES
<< (EXTSPERSUM
- bitno
);
1498 inosmap
= le32_to_cpu(iagp
->inosmap
[sword
]) | mask
;
1499 extsmap
= le32_to_cpu(iagp
->extsmap
[sword
]) | mask
;
1501 /* scan the free inode and free extent summary maps for
1504 for (i
= 0; i
< nwords
; i
++) {
1505 /* check if this word of the free inode summary
1506 * map describes an extent with free inodes.
1509 /* an extent with free inodes has been
1510 * found. determine the extent number
1511 * and the inode number within the extent.
1513 rem
= diFindFree(inosmap
, 0);
1514 extno
= (sword
<< L2EXTSPERSUM
) + rem
;
1515 rem
= diFindFree(le32_to_cpu(iagp
->wmap
[extno
]),
1517 if (rem
>= INOSPEREXT
) {
1518 IREAD_UNLOCK(ipimap
);
1519 release_metapage(mp
);
1520 AG_UNLOCK(imap
, agno
);
1522 "diAlloc: can't find free bit "
1527 /* determine the inode number within the
1528 * iag and allocate the inode from the
1531 ino
= (extno
<< L2INOSPEREXT
) + rem
;
1532 rc
= diAllocBit(imap
, iagp
, ino
);
1533 IREAD_UNLOCK(ipimap
);
1537 /* set the results of the allocation
1538 * and write the iag.
1540 diInitInode(ip
, iagno
, ino
, extno
,
1542 mark_metapage_dirty(mp
);
1544 release_metapage(mp
);
1546 /* free the AG lock and return.
1548 AG_UNLOCK(imap
, agno
);
1553 /* check if we may allocate an extent of free
1554 * inodes and whether this word of the free
1555 * extents summary map describes a free extent.
1557 if (addext
&& ~extsmap
) {
1558 /* a free extent has been found. determine
1559 * the extent number.
1561 rem
= diFindFree(extsmap
, 0);
1562 extno
= (sword
<< L2EXTSPERSUM
) + rem
;
1564 /* allocate an extent of free inodes.
1566 if ((rc
= diNewExt(imap
, iagp
, extno
))) {
1567 /* if there is no disk space for a
1568 * new extent, try to allocate the
1569 * disk inode from somewhere else.
1576 /* set the results of the allocation
1577 * and write the iag.
1579 diInitInode(ip
, iagno
,
1580 extno
<< L2INOSPEREXT
,
1582 mark_metapage_dirty(mp
);
1584 release_metapage(mp
);
1585 /* free the imap inode & the AG lock & return.
1587 IREAD_UNLOCK(ipimap
);
1588 AG_UNLOCK(imap
, agno
);
1592 /* move on to the next set of summary map words.
1594 sword
= (sword
== SMAPSZ
- 1) ? 0 : sword
+ 1;
1595 inosmap
= le32_to_cpu(iagp
->inosmap
[sword
]);
1596 extsmap
= le32_to_cpu(iagp
->extsmap
[sword
]);
1599 /* unlock imap inode */
1600 IREAD_UNLOCK(ipimap
);
1602 /* nothing doing in this iag, so release it. */
1603 release_metapage(mp
);
1607 * try to allocate anywhere within the same AG as the parent inode.
1609 rc
= diAllocAG(imap
, agno
, dir
, ip
);
1611 AG_UNLOCK(imap
, agno
);
1617 * try to allocate in any AG.
1619 return (diAllocAny(imap
, agno
, dir
, ip
));
1624 * NAME: diAllocAG(imap,agno,dir,ip)
1626 * FUNCTION: allocate a disk inode from the allocation group.
1628 * this routine first determines if a new extent of free
1629 * inodes should be added for the allocation group, with
1630 * the current request satisfied from this extent. if this
1631 * is the case, an attempt will be made to do just that. if
1632 * this attempt fails or it has been determined that a new
1633 * extent should not be added, an attempt is made to satisfy
1634 * the request by allocating an existing (backed) free inode
1635 * from the allocation group.
1637 * PRE CONDITION: Already have the AG lock for this AG.
1640 * imap - pointer to inode map control structure.
1641 * agno - allocation group to allocate from.
1642 * dir - 'true' if the new disk inode is for a directory.
1643 * ip - pointer to the new inode to be filled in on successful return
1644 * with the disk inode number allocated, its extent address
1645 * and the start of the ag.
1649 * -ENOSPC - insufficient disk resources.
1653 diAllocAG(struct inomap
* imap
, int agno
, bool dir
, struct inode
*ip
)
1655 int rc
, addext
, numfree
, numinos
;
1657 /* get the number of free and the number of backed disk
1658 * inodes currently within the ag.
1660 numfree
= imap
->im_agctl
[agno
].numfree
;
1661 numinos
= imap
->im_agctl
[agno
].numinos
;
1663 if (numfree
> numinos
) {
1664 jfs_error(ip
->i_sb
, "diAllocAG: numfree > numinos");
1668 /* determine if we should allocate a new extent of free inodes
1669 * within the ag: for directory inodes, add a new extent
1670 * if there are a small number of free inodes or number of free
1671 * inodes is a small percentage of the number of backed inodes.
1674 addext
= (numfree
< 64 ||
1676 && ((numfree
* 100) / numinos
) <= 20));
1678 addext
= (numfree
== 0);
1681 * try to allocate a new extent of free inodes.
1684 /* if free space is not avaliable for this new extent, try
1685 * below to allocate a free and existing (already backed)
1686 * inode from the ag.
1688 if ((rc
= diAllocExt(imap
, agno
, ip
)) != -ENOSPC
)
1693 * try to allocate an existing free inode from the ag.
1695 return (diAllocIno(imap
, agno
, ip
));
1700 * NAME: diAllocAny(imap,agno,dir,iap)
1702 * FUNCTION: allocate a disk inode from any other allocation group.
1704 * this routine is called when an allocation attempt within
1705 * the primary allocation group has failed. if attempts to
1706 * allocate an inode from any allocation group other than the
1707 * specified primary group.
1710 * imap - pointer to inode map control structure.
1711 * agno - primary allocation group (to avoid).
1712 * dir - 'true' if the new disk inode is for a directory.
1713 * ip - pointer to a new inode to be filled in on successful return
1714 * with the disk inode number allocated, its extent address
1715 * and the start of the ag.
1719 * -ENOSPC - insufficient disk resources.
1723 diAllocAny(struct inomap
* imap
, int agno
, bool dir
, struct inode
*ip
)
1726 int maxag
= JFS_SBI(imap
->im_ipimap
->i_sb
)->bmap
->db_maxag
;
1729 /* try to allocate from the ags following agno up to
1730 * the maximum ag number.
1732 for (ag
= agno
+ 1; ag
<= maxag
; ag
++) {
1735 rc
= diAllocAG(imap
, ag
, dir
, ip
);
1737 AG_UNLOCK(imap
, ag
);
1743 /* try to allocate from the ags in front of agno.
1745 for (ag
= 0; ag
< agno
; ag
++) {
1748 rc
= diAllocAG(imap
, ag
, dir
, ip
);
1750 AG_UNLOCK(imap
, ag
);
1756 /* no free disk inodes.
1763 * NAME: diAllocIno(imap,agno,ip)
1765 * FUNCTION: allocate a disk inode from the allocation group's free
1766 * inode list, returning an error if this free list is
1767 * empty (i.e. no iags on the list).
1769 * allocation occurs from the first iag on the list using
1770 * the iag's free inode summary map to find the leftmost
1771 * free inode in the iag.
1773 * PRE CONDITION: Already have AG lock for this AG.
1776 * imap - pointer to inode map control structure.
1777 * agno - allocation group.
1778 * ip - pointer to new inode to be filled in on successful return
1779 * with the disk inode number allocated, its extent address
1780 * and the start of the ag.
1784 * -ENOSPC - insufficient disk resources.
1787 static int diAllocIno(struct inomap
* imap
, int agno
, struct inode
*ip
)
1789 int iagno
, ino
, rc
, rem
, extno
, sword
;
1790 struct metapage
*mp
;
1793 /* check if there are iags on the ag's free inode list.
1795 if ((iagno
= imap
->im_agctl
[agno
].inofree
) < 0)
1798 /* obtain read lock on imap inode */
1799 IREAD_LOCK(imap
->im_ipimap
, RDWRLOCK_IMAP
);
1801 /* read the iag at the head of the list.
1803 if ((rc
= diIAGRead(imap
, iagno
, &mp
))) {
1804 IREAD_UNLOCK(imap
->im_ipimap
);
1807 iagp
= (struct iag
*) mp
->data
;
1809 /* better be free inodes in this iag if it is on the
1812 if (!iagp
->nfreeinos
) {
1813 IREAD_UNLOCK(imap
->im_ipimap
);
1814 release_metapage(mp
);
1816 "diAllocIno: nfreeinos = 0, but iag on freelist");
1820 /* scan the free inode summary map to find an extent
1823 for (sword
= 0;; sword
++) {
1824 if (sword
>= SMAPSZ
) {
1825 IREAD_UNLOCK(imap
->im_ipimap
);
1826 release_metapage(mp
);
1828 "diAllocIno: free inode not found in summary map");
1832 if (~iagp
->inosmap
[sword
])
1836 /* found a extent with free inodes. determine
1837 * the extent number.
1839 rem
= diFindFree(le32_to_cpu(iagp
->inosmap
[sword
]), 0);
1840 if (rem
>= EXTSPERSUM
) {
1841 IREAD_UNLOCK(imap
->im_ipimap
);
1842 release_metapage(mp
);
1843 jfs_error(ip
->i_sb
, "diAllocIno: no free extent found");
1846 extno
= (sword
<< L2EXTSPERSUM
) + rem
;
1848 /* find the first free inode in the extent.
1850 rem
= diFindFree(le32_to_cpu(iagp
->wmap
[extno
]), 0);
1851 if (rem
>= INOSPEREXT
) {
1852 IREAD_UNLOCK(imap
->im_ipimap
);
1853 release_metapage(mp
);
1854 jfs_error(ip
->i_sb
, "diAllocIno: free inode not found");
1858 /* compute the inode number within the iag.
1860 ino
= (extno
<< L2INOSPEREXT
) + rem
;
1862 /* allocate the inode.
1864 rc
= diAllocBit(imap
, iagp
, ino
);
1865 IREAD_UNLOCK(imap
->im_ipimap
);
1867 release_metapage(mp
);
1871 /* set the results of the allocation and write the iag.
1873 diInitInode(ip
, iagno
, ino
, extno
, iagp
);
1881 * NAME: diAllocExt(imap,agno,ip)
1883 * FUNCTION: add a new extent of free inodes to an iag, allocating
1884 * an inode from this extent to satisfy the current allocation
1887 * this routine first tries to find an existing iag with free
1888 * extents through the ag free extent list. if list is not
1889 * empty, the head of the list will be selected as the home
1890 * of the new extent of free inodes. otherwise (the list is
1891 * empty), a new iag will be allocated for the ag to contain
1894 * once an iag has been selected, the free extent summary map
1895 * is used to locate a free extent within the iag and diNewExt()
1896 * is called to initialize the extent, with initialization
1897 * including the allocation of the first inode of the extent
1898 * for the purpose of satisfying this request.
1901 * imap - pointer to inode map control structure.
1902 * agno - allocation group number.
1903 * ip - pointer to new inode to be filled in on successful return
1904 * with the disk inode number allocated, its extent address
1905 * and the start of the ag.
1909 * -ENOSPC - insufficient disk resources.
1912 static int diAllocExt(struct inomap
* imap
, int agno
, struct inode
*ip
)
1914 int rem
, iagno
, sword
, extno
, rc
;
1915 struct metapage
*mp
;
1918 /* check if the ag has any iags with free extents. if not,
1919 * allocate a new iag for the ag.
1921 if ((iagno
= imap
->im_agctl
[agno
].extfree
) < 0) {
1922 /* If successful, diNewIAG will obtain the read lock on the
1925 if ((rc
= diNewIAG(imap
, &iagno
, agno
, &mp
))) {
1928 iagp
= (struct iag
*) mp
->data
;
1930 /* set the ag number if this a brand new iag
1933 cpu_to_le64(AGTOBLK(agno
, imap
->im_ipimap
));
1937 IREAD_LOCK(imap
->im_ipimap
, RDWRLOCK_IMAP
);
1938 if ((rc
= diIAGRead(imap
, iagno
, &mp
))) {
1939 IREAD_UNLOCK(imap
->im_ipimap
);
1940 jfs_error(ip
->i_sb
, "diAllocExt: error reading iag");
1943 iagp
= (struct iag
*) mp
->data
;
1946 /* using the free extent summary map, find a free extent.
1948 for (sword
= 0;; sword
++) {
1949 if (sword
>= SMAPSZ
) {
1950 release_metapage(mp
);
1951 IREAD_UNLOCK(imap
->im_ipimap
);
1953 "diAllocExt: free ext summary map not found");
1956 if (~iagp
->extsmap
[sword
])
1960 /* determine the extent number of the free extent.
1962 rem
= diFindFree(le32_to_cpu(iagp
->extsmap
[sword
]), 0);
1963 if (rem
>= EXTSPERSUM
) {
1964 release_metapage(mp
);
1965 IREAD_UNLOCK(imap
->im_ipimap
);
1966 jfs_error(ip
->i_sb
, "diAllocExt: free extent not found");
1969 extno
= (sword
<< L2EXTSPERSUM
) + rem
;
1971 /* initialize the new extent.
1973 rc
= diNewExt(imap
, iagp
, extno
);
1974 IREAD_UNLOCK(imap
->im_ipimap
);
1976 /* something bad happened. if a new iag was allocated,
1977 * place it back on the inode map's iag free list, and
1978 * clear the ag number information.
1980 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
)) {
1982 iagp
->iagfree
= cpu_to_le32(imap
->im_freeiag
);
1983 imap
->im_freeiag
= iagno
;
1984 IAGFREE_UNLOCK(imap
);
1990 /* set the results of the allocation and write the iag.
1992 diInitInode(ip
, iagno
, extno
<< L2INOSPEREXT
, extno
, iagp
);
2001 * NAME: diAllocBit(imap,iagp,ino)
2003 * FUNCTION: allocate a backed inode from an iag.
2005 * this routine performs the mechanics of allocating a
2006 * specified inode from a backed extent.
2008 * if the inode to be allocated represents the last free
2009 * inode within the iag, the iag will be removed from the
2010 * ag free inode list.
2012 * a careful update approach is used to provide consistency
2013 * in the face of updates to multiple buffers. under this
2014 * approach, all required buffers are obtained before making
2015 * any updates and are held all are updates are complete.
2017 * PRE CONDITION: Already have buffer lock on iagp. Already have AG lock on
2018 * this AG. Must have read lock on imap inode.
2021 * imap - pointer to inode map control structure.
2022 * iagp - pointer to iag.
2023 * ino - inode number to be allocated within the iag.
2027 * -ENOSPC - insufficient disk resources.
2030 static int diAllocBit(struct inomap
* imap
, struct iag
* iagp
, int ino
)
2032 int extno
, bitno
, agno
, sword
, rc
;
2033 struct metapage
*amp
= NULL
, *bmp
= NULL
;
2034 struct iag
*aiagp
= NULL
, *biagp
= NULL
;
2037 /* check if this is the last free inode within the iag.
2038 * if so, it will have to be removed from the ag free
2039 * inode list, so get the iags preceeding and following
2042 if (iagp
->nfreeinos
== cpu_to_le32(1)) {
2043 if ((int) le32_to_cpu(iagp
->inofreefwd
) >= 0) {
2045 diIAGRead(imap
, le32_to_cpu(iagp
->inofreefwd
),
2048 aiagp
= (struct iag
*) amp
->data
;
2051 if ((int) le32_to_cpu(iagp
->inofreeback
) >= 0) {
2054 le32_to_cpu(iagp
->inofreeback
),
2057 release_metapage(amp
);
2060 biagp
= (struct iag
*) bmp
->data
;
2064 /* get the ag number, extent number, inode number within
2067 agno
= BLKTOAG(le64_to_cpu(iagp
->agstart
), JFS_SBI(imap
->im_ipimap
->i_sb
));
2068 extno
= ino
>> L2INOSPEREXT
;
2069 bitno
= ino
& (INOSPEREXT
- 1);
2071 /* compute the mask for setting the map.
2073 mask
= HIGHORDER
>> bitno
;
2075 /* the inode should be free and backed.
2077 if (((le32_to_cpu(iagp
->pmap
[extno
]) & mask
) != 0) ||
2078 ((le32_to_cpu(iagp
->wmap
[extno
]) & mask
) != 0) ||
2079 (addressPXD(&iagp
->inoext
[extno
]) == 0)) {
2081 release_metapage(amp
);
2083 release_metapage(bmp
);
2085 jfs_error(imap
->im_ipimap
->i_sb
,
2086 "diAllocBit: iag inconsistent");
2090 /* mark the inode as allocated in the working map.
2092 iagp
->wmap
[extno
] |= cpu_to_le32(mask
);
2094 /* check if all inodes within the extent are now
2095 * allocated. if so, update the free inode summary
2096 * map to reflect this.
2098 if (iagp
->wmap
[extno
] == cpu_to_le32(ONES
)) {
2099 sword
= extno
>> L2EXTSPERSUM
;
2100 bitno
= extno
& (EXTSPERSUM
- 1);
2101 iagp
->inosmap
[sword
] |= cpu_to_le32(HIGHORDER
>> bitno
);
2104 /* if this was the last free inode in the iag, remove the
2105 * iag from the ag free inode list.
2107 if (iagp
->nfreeinos
== cpu_to_le32(1)) {
2109 aiagp
->inofreeback
= iagp
->inofreeback
;
2110 write_metapage(amp
);
2114 biagp
->inofreefwd
= iagp
->inofreefwd
;
2115 write_metapage(bmp
);
2117 imap
->im_agctl
[agno
].inofree
=
2118 le32_to_cpu(iagp
->inofreefwd
);
2120 iagp
->inofreefwd
= iagp
->inofreeback
= cpu_to_le32(-1);
2123 /* update the free inode count at the iag, ag, inode
2126 iagp
->nfreeinos
= cpu_to_le32(le32_to_cpu(iagp
->nfreeinos
) - 1);
2127 imap
->im_agctl
[agno
].numfree
-= 1;
2128 atomic_dec(&imap
->im_numfree
);
2135 * NAME: diNewExt(imap,iagp,extno)
2137 * FUNCTION: initialize a new extent of inodes for an iag, allocating
2138 * the first inode of the extent for use for the current
2139 * allocation request.
2141 * disk resources are allocated for the new extent of inodes
2142 * and the inodes themselves are initialized to reflect their
2143 * existence within the extent (i.e. their inode numbers and
2144 * inode extent addresses are set) and their initial state
2145 * (mode and link count are set to zero).
2147 * if the iag is new, it is not yet on an ag extent free list
2148 * but will now be placed on this list.
2150 * if the allocation of the new extent causes the iag to
2151 * have no free extent, the iag will be removed from the
2152 * ag extent free list.
2154 * if the iag has no free backed inodes, it will be placed
2155 * on the ag free inode list, since the addition of the new
2156 * extent will now cause it to have free inodes.
2158 * a careful update approach is used to provide consistency
2159 * (i.e. list consistency) in the face of updates to multiple
2160 * buffers. under this approach, all required buffers are
2161 * obtained before making any updates and are held until all
2162 * updates are complete.
2164 * PRE CONDITION: Already have buffer lock on iagp. Already have AG lock on
2165 * this AG. Must have read lock on imap inode.
2168 * imap - pointer to inode map control structure.
2169 * iagp - pointer to iag.
2170 * extno - extent number.
2174 * -ENOSPC - insufficient disk resources.
2177 static int diNewExt(struct inomap
* imap
, struct iag
* iagp
, int extno
)
2179 int agno
, iagno
, fwd
, back
, freei
= 0, sword
, rc
;
2180 struct iag
*aiagp
= NULL
, *biagp
= NULL
, *ciagp
= NULL
;
2181 struct metapage
*amp
, *bmp
, *cmp
, *dmp
;
2182 struct inode
*ipimap
;
2188 struct jfs_sb_info
*sbi
;
2190 /* better have free extents.
2192 if (!iagp
->nfreeexts
) {
2193 jfs_error(imap
->im_ipimap
->i_sb
, "diNewExt: no free extents");
2197 /* get the inode map inode.
2199 ipimap
= imap
->im_ipimap
;
2200 sbi
= JFS_SBI(ipimap
->i_sb
);
2202 amp
= bmp
= cmp
= NULL
;
2204 /* get the ag and iag numbers for this iag.
2206 agno
= BLKTOAG(le64_to_cpu(iagp
->agstart
), sbi
);
2207 iagno
= le32_to_cpu(iagp
->iagnum
);
2209 /* check if this is the last free extent within the
2210 * iag. if so, the iag must be removed from the ag
2211 * free extent list, so get the iags preceeding and
2212 * following the iag on this list.
2214 if (iagp
->nfreeexts
== cpu_to_le32(1)) {
2215 if ((fwd
= le32_to_cpu(iagp
->extfreefwd
)) >= 0) {
2216 if ((rc
= diIAGRead(imap
, fwd
, &
)))
2218 aiagp
= (struct iag
*) amp
->data
;
2221 if ((back
= le32_to_cpu(iagp
->extfreeback
)) >= 0) {
2222 if ((rc
= diIAGRead(imap
, back
, &bmp
)))
2224 biagp
= (struct iag
*) bmp
->data
;
2227 /* the iag has free extents. if all extents are free
2228 * (as is the case for a newly allocated iag), the iag
2229 * must be added to the ag free extent list, so get
2230 * the iag at the head of the list in preparation for
2231 * adding this iag to this list.
2234 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
)) {
2235 if ((fwd
= imap
->im_agctl
[agno
].extfree
) >= 0) {
2236 if ((rc
= diIAGRead(imap
, fwd
, &
)))
2238 aiagp
= (struct iag
*) amp
->data
;
2243 /* check if the iag has no free inodes. if so, the iag
2244 * will have to be added to the ag free inode list, so get
2245 * the iag at the head of the list in preparation for
2246 * adding this iag to this list. in doing this, we must
2247 * check if we already have the iag at the head of
2250 if (iagp
->nfreeinos
== 0) {
2251 freei
= imap
->im_agctl
[agno
].inofree
;
2256 } else if (freei
== back
) {
2259 if ((rc
= diIAGRead(imap
, freei
, &cmp
)))
2261 ciagp
= (struct iag
*) cmp
->data
;
2263 if (ciagp
== NULL
) {
2264 jfs_error(imap
->im_ipimap
->i_sb
,
2265 "diNewExt: ciagp == NULL");
2272 /* allocate disk space for the inode extent.
2274 if ((extno
== 0) || (addressPXD(&iagp
->inoext
[extno
- 1]) == 0))
2275 hint
= ((s64
) agno
<< sbi
->bmap
->db_agl2size
) - 1;
2277 hint
= addressPXD(&iagp
->inoext
[extno
- 1]) +
2278 lengthPXD(&iagp
->inoext
[extno
- 1]) - 1;
2280 if ((rc
= dbAlloc(ipimap
, hint
, (s64
) imap
->im_nbperiext
, &blkno
)))
2283 /* compute the inode number of the first inode within the
2286 ino
= (iagno
<< L2INOSPERIAG
) + (extno
<< L2INOSPEREXT
);
2288 /* initialize the inodes within the newly allocated extent a
2291 for (i
= 0; i
< imap
->im_nbperiext
; i
+= sbi
->nbperpage
) {
2292 /* get a buffer for this page of disk inodes.
2294 dmp
= get_metapage(ipimap
, blkno
+ i
, PSIZE
, 1);
2299 dp
= (struct dinode
*) dmp
->data
;
2301 /* initialize the inode number, mode, link count and
2302 * inode extent address.
2304 for (j
= 0; j
< INOSPERPAGE
; j
++, dp
++, ino
++) {
2305 dp
->di_inostamp
= cpu_to_le32(sbi
->inostamp
);
2306 dp
->di_number
= cpu_to_le32(ino
);
2307 dp
->di_fileset
= cpu_to_le32(FILESYSTEM_I
);
2310 PXDaddress(&(dp
->di_ixpxd
), blkno
);
2311 PXDlength(&(dp
->di_ixpxd
), imap
->im_nbperiext
);
2313 write_metapage(dmp
);
2316 /* if this is the last free extent within the iag, remove the
2317 * iag from the ag free extent list.
2319 if (iagp
->nfreeexts
== cpu_to_le32(1)) {
2321 aiagp
->extfreeback
= iagp
->extfreeback
;
2324 biagp
->extfreefwd
= iagp
->extfreefwd
;
2326 imap
->im_agctl
[agno
].extfree
=
2327 le32_to_cpu(iagp
->extfreefwd
);
2329 iagp
->extfreefwd
= iagp
->extfreeback
= cpu_to_le32(-1);
2331 /* if the iag has all free extents (newly allocated iag),
2332 * add the iag to the ag free extent list.
2334 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
)) {
2336 aiagp
->extfreeback
= cpu_to_le32(iagno
);
2338 iagp
->extfreefwd
= cpu_to_le32(fwd
);
2339 iagp
->extfreeback
= cpu_to_le32(-1);
2340 imap
->im_agctl
[agno
].extfree
= iagno
;
2344 /* if the iag has no free inodes, add the iag to the
2345 * ag free inode list.
2347 if (iagp
->nfreeinos
== 0) {
2349 ciagp
->inofreeback
= cpu_to_le32(iagno
);
2352 cpu_to_le32(imap
->im_agctl
[agno
].inofree
);
2353 iagp
->inofreeback
= cpu_to_le32(-1);
2354 imap
->im_agctl
[agno
].inofree
= iagno
;
2357 /* initialize the extent descriptor of the extent. */
2358 PXDlength(&iagp
->inoext
[extno
], imap
->im_nbperiext
);
2359 PXDaddress(&iagp
->inoext
[extno
], blkno
);
2361 /* initialize the working and persistent map of the extent.
2362 * the working map will be initialized such that
2363 * it indicates the first inode of the extent is allocated.
2365 iagp
->wmap
[extno
] = cpu_to_le32(HIGHORDER
);
2366 iagp
->pmap
[extno
] = 0;
2368 /* update the free inode and free extent summary maps
2369 * for the extent to indicate the extent has free inodes
2370 * and no longer represents a free extent.
2372 sword
= extno
>> L2EXTSPERSUM
;
2373 mask
= HIGHORDER
>> (extno
& (EXTSPERSUM
- 1));
2374 iagp
->extsmap
[sword
] |= cpu_to_le32(mask
);
2375 iagp
->inosmap
[sword
] &= cpu_to_le32(~mask
);
2377 /* update the free inode and free extent counts for the
2380 iagp
->nfreeinos
= cpu_to_le32(le32_to_cpu(iagp
->nfreeinos
) +
2382 iagp
->nfreeexts
= cpu_to_le32(le32_to_cpu(iagp
->nfreeexts
) - 1);
2384 /* update the free and backed inode counts for the ag.
2386 imap
->im_agctl
[agno
].numfree
+= (INOSPEREXT
- 1);
2387 imap
->im_agctl
[agno
].numinos
+= INOSPEREXT
;
2389 /* update the free and backed inode counts for the inode map.
2391 atomic_add(INOSPEREXT
- 1, &imap
->im_numfree
);
2392 atomic_add(INOSPEREXT
, &imap
->im_numinos
);
2397 write_metapage(amp
);
2399 write_metapage(bmp
);
2401 write_metapage(cmp
);
2407 /* release the iags.
2410 release_metapage(amp
);
2412 release_metapage(bmp
);
2414 release_metapage(cmp
);
2421 * NAME: diNewIAG(imap,iagnop,agno)
2423 * FUNCTION: allocate a new iag for an allocation group.
2425 * first tries to allocate the iag from the inode map
2427 * if the list has free iags, the head of the list is removed
2428 * and returned to satisfy the request.
2429 * if the inode map's iag free list is empty, the inode map
2430 * is extended to hold a new iag. this new iag is initialized
2431 * and returned to satisfy the request.
2434 * imap - pointer to inode map control structure.
2435 * iagnop - pointer to an iag number set with the number of the
2436 * newly allocated iag upon successful return.
2437 * agno - allocation group number.
2438 * bpp - Buffer pointer to be filled in with new IAG's buffer
2442 * -ENOSPC - insufficient disk resources.
2446 * AG lock held on entry/exit;
2447 * write lock on the map is held inside;
2448 * read lock on the map is held on successful completion;
2450 * note: new iag transaction:
2451 * . synchronously write iag;
2452 * . write log of xtree and inode of imap;
2454 * . synchronous write of xtree (right to left, bottom to top);
2455 * . at start of logredo(): init in-memory imap with one additional iag page;
2456 * . at end of logredo(): re-read imap inode to determine
2460 diNewIAG(struct inomap
* imap
, int *iagnop
, int agno
, struct metapage
** mpp
)
2464 struct inode
*ipimap
;
2465 struct super_block
*sb
;
2466 struct jfs_sb_info
*sbi
;
2467 struct metapage
*mp
;
2472 struct inode
*iplist
[1];
2474 /* pick up pointers to the inode map and mount inodes */
2475 ipimap
= imap
->im_ipimap
;
2479 /* acquire the free iag lock */
2482 /* if there are any iags on the inode map free iag list,
2483 * allocate the iag from the head of the list.
2485 if (imap
->im_freeiag
>= 0) {
2486 /* pick up the iag number at the head of the list */
2487 iagno
= imap
->im_freeiag
;
2489 /* determine the logical block number of the iag */
2490 blkno
= IAGTOLBLK(iagno
, sbi
->l2nbperpage
);
2492 /* no free iags. the inode map will have to be extented
2493 * to include a new iag.
2496 /* acquire inode map lock */
2497 IWRITE_LOCK(ipimap
, RDWRLOCK_IMAP
);
2499 if (ipimap
->i_size
>> L2PSIZE
!= imap
->im_nextiag
+ 1) {
2500 IWRITE_UNLOCK(ipimap
);
2501 IAGFREE_UNLOCK(imap
);
2502 jfs_error(imap
->im_ipimap
->i_sb
,
2503 "diNewIAG: ipimap->i_size is wrong");
2508 /* get the next avaliable iag number */
2509 iagno
= imap
->im_nextiag
;
2511 /* make sure that we have not exceeded the maximum inode
2514 if (iagno
> (MAXIAGS
- 1)) {
2515 /* release the inode map lock */
2516 IWRITE_UNLOCK(ipimap
);
2523 * synchronously append new iag page.
2525 /* determine the logical address of iag page to append */
2526 blkno
= IAGTOLBLK(iagno
, sbi
->l2nbperpage
);
2528 /* Allocate extent for new iag page */
2529 xlen
= sbi
->nbperpage
;
2530 if ((rc
= dbAlloc(ipimap
, 0, (s64
) xlen
, &xaddr
))) {
2531 /* release the inode map lock */
2532 IWRITE_UNLOCK(ipimap
);
2538 * start transaction of update of the inode map
2539 * addressing structure pointing to the new iag page;
2541 tid
= txBegin(sb
, COMMIT_FORCE
);
2542 mutex_lock(&JFS_IP(ipimap
)->commit_mutex
);
2544 /* update the inode map addressing structure to point to it */
2546 xtInsert(tid
, ipimap
, 0, blkno
, xlen
, &xaddr
, 0))) {
2548 mutex_unlock(&JFS_IP(ipimap
)->commit_mutex
);
2549 /* Free the blocks allocated for the iag since it was
2550 * not successfully added to the inode map
2552 dbFree(ipimap
, xaddr
, (s64
) xlen
);
2554 /* release the inode map lock */
2555 IWRITE_UNLOCK(ipimap
);
2560 /* update the inode map's inode to reflect the extension */
2561 ipimap
->i_size
+= PSIZE
;
2562 inode_add_bytes(ipimap
, PSIZE
);
2564 /* assign a buffer for the page */
2565 mp
= get_metapage(ipimap
, blkno
, PSIZE
, 0);
2568 * This is very unlikely since we just created the
2569 * extent, but let's try to handle it correctly
2571 xtTruncate(tid
, ipimap
, ipimap
->i_size
- PSIZE
,
2577 /* release the inode map lock */
2578 IWRITE_UNLOCK(ipimap
);
2583 iagp
= (struct iag
*) mp
->data
;
2586 memset(iagp
, 0, sizeof(struct iag
));
2587 iagp
->iagnum
= cpu_to_le32(iagno
);
2588 iagp
->inofreefwd
= iagp
->inofreeback
= cpu_to_le32(-1);
2589 iagp
->extfreefwd
= iagp
->extfreeback
= cpu_to_le32(-1);
2590 iagp
->iagfree
= cpu_to_le32(-1);
2591 iagp
->nfreeinos
= 0;
2592 iagp
->nfreeexts
= cpu_to_le32(EXTSPERIAG
);
2594 /* initialize the free inode summary map (free extent
2595 * summary map initialization handled by bzero).
2597 for (i
= 0; i
< SMAPSZ
; i
++)
2598 iagp
->inosmap
[i
] = cpu_to_le32(ONES
);
2601 * Write and sync the metapage
2606 * txCommit(COMMIT_FORCE) will synchronously write address
2607 * index pages and inode after commit in careful update order
2608 * of address index pages (right to left, bottom up);
2611 rc
= txCommit(tid
, 1, &iplist
[0], COMMIT_FORCE
);
2614 mutex_unlock(&JFS_IP(ipimap
)->commit_mutex
);
2616 duplicateIXtree(sb
, blkno
, xlen
, &xaddr
);
2618 /* update the next avaliable iag number */
2619 imap
->im_nextiag
+= 1;
2621 /* Add the iag to the iag free list so we don't lose the iag
2622 * if a failure happens now.
2624 imap
->im_freeiag
= iagno
;
2626 /* Until we have logredo working, we want the imap inode &
2627 * control page to be up to date.
2631 /* release the inode map lock */
2632 IWRITE_UNLOCK(ipimap
);
2635 /* obtain read lock on map */
2636 IREAD_LOCK(ipimap
, RDWRLOCK_IMAP
);
2639 if ((rc
= diIAGRead(imap
, iagno
, &mp
))) {
2640 IREAD_UNLOCK(ipimap
);
2644 iagp
= (struct iag
*) mp
->data
;
2646 /* remove the iag from the iag free list */
2647 imap
->im_freeiag
= le32_to_cpu(iagp
->iagfree
);
2648 iagp
->iagfree
= cpu_to_le32(-1);
2650 /* set the return iag number and buffer pointer */
2655 /* release the iag free lock */
2656 IAGFREE_UNLOCK(imap
);
2664 * FUNCTION: get the buffer for the specified iag within a fileset
2665 * or aggregate inode map.
2668 * imap - pointer to inode map control structure.
2669 * iagno - iag number.
2670 * bpp - point to buffer pointer to be filled in on successful
2674 * must have read lock on imap inode
2675 * (When called by diExtendFS, the filesystem is quiesced, therefore
2676 * the read lock is unnecessary.)
2682 static int diIAGRead(struct inomap
* imap
, int iagno
, struct metapage
** mpp
)
2684 struct inode
*ipimap
= imap
->im_ipimap
;
2687 /* compute the logical block number of the iag. */
2688 blkno
= IAGTOLBLK(iagno
, JFS_SBI(ipimap
->i_sb
)->l2nbperpage
);
2691 *mpp
= read_metapage(ipimap
, blkno
, PSIZE
, 0);
2700 * NAME: diFindFree()
2702 * FUNCTION: find the first free bit in a word starting at
2703 * the specified bit position.
2706 * word - word to be examined.
2707 * start - starting bit position.
2710 * bit position of first free bit in the word or 32 if
2711 * no free bits were found.
2713 static int diFindFree(u32 word
, int start
)
2717 /* scan the word for the first free bit. */
2718 for (word
<<= start
, bitno
= start
; bitno
< 32;
2719 bitno
++, word
<<= 1) {
2720 if ((word
& HIGHORDER
) == 0)
2727 * NAME: diUpdatePMap()
2729 * FUNCTION: Update the persistent map in an IAG for the allocation or
2730 * freeing of the specified inode.
2732 * PRE CONDITIONS: Working map has already been updated for allocate.
2735 * ipimap - Incore inode map inode
2736 * inum - Number of inode to mark in permanent map
2737 * is_free - If 'true' indicates inode should be marked freed, otherwise
2738 * indicates inode should be marked allocated.
2744 diUpdatePMap(struct inode
*ipimap
,
2745 unsigned long inum
, bool is_free
, struct tblock
* tblk
)
2749 struct metapage
*mp
;
2750 int iagno
, ino
, extno
, bitno
;
2751 struct inomap
*imap
;
2753 struct jfs_log
*log
;
2754 int lsn
, difft
, diffp
;
2755 unsigned long flags
;
2757 imap
= JFS_IP(ipimap
)->i_imap
;
2758 /* get the iag number containing the inode */
2759 iagno
= INOTOIAG(inum
);
2760 /* make sure that the iag is contained within the map */
2761 if (iagno
>= imap
->im_nextiag
) {
2762 jfs_error(ipimap
->i_sb
,
2763 "diUpdatePMap: the iag is outside the map");
2767 IREAD_LOCK(ipimap
, RDWRLOCK_IMAP
);
2768 rc
= diIAGRead(imap
, iagno
, &mp
);
2769 IREAD_UNLOCK(ipimap
);
2772 metapage_wait_for_io(mp
);
2773 iagp
= (struct iag
*) mp
->data
;
2774 /* get the inode number and extent number of the inode within
2775 * the iag and the inode number within the extent.
2777 ino
= inum
& (INOSPERIAG
- 1);
2778 extno
= ino
>> L2INOSPEREXT
;
2779 bitno
= ino
& (INOSPEREXT
- 1);
2780 mask
= HIGHORDER
>> bitno
;
2782 * mark the inode free in persistent map:
2785 /* The inode should have been allocated both in working
2786 * map and in persistent map;
2787 * the inode will be freed from working map at the release
2788 * of last reference release;
2790 if (!(le32_to_cpu(iagp
->wmap
[extno
]) & mask
)) {
2791 jfs_error(ipimap
->i_sb
,
2792 "diUpdatePMap: inode %ld not marked as "
2793 "allocated in wmap!", inum
);
2795 if (!(le32_to_cpu(iagp
->pmap
[extno
]) & mask
)) {
2796 jfs_error(ipimap
->i_sb
,
2797 "diUpdatePMap: inode %ld not marked as "
2798 "allocated in pmap!", inum
);
2800 /* update the bitmap for the extent of the freed inode */
2801 iagp
->pmap
[extno
] &= cpu_to_le32(~mask
);
2804 * mark the inode allocated in persistent map:
2807 /* The inode should be already allocated in the working map
2808 * and should be free in persistent map;
2810 if (!(le32_to_cpu(iagp
->wmap
[extno
]) & mask
)) {
2811 release_metapage(mp
);
2812 jfs_error(ipimap
->i_sb
,
2813 "diUpdatePMap: the inode is not allocated in "
2817 if ((le32_to_cpu(iagp
->pmap
[extno
]) & mask
) != 0) {
2818 release_metapage(mp
);
2819 jfs_error(ipimap
->i_sb
,
2820 "diUpdatePMap: the inode is not free in the "
2824 /* update the bitmap for the extent of the allocated inode */
2825 iagp
->pmap
[extno
] |= cpu_to_le32(mask
);
2831 log
= JFS_SBI(tblk
->sb
)->log
;
2832 LOGSYNC_LOCK(log
, flags
);
2834 /* inherit older/smaller lsn */
2835 logdiff(difft
, lsn
, log
);
2836 logdiff(diffp
, mp
->lsn
, log
);
2837 if (difft
< diffp
) {
2839 /* move mp after tblock in logsync list */
2840 list_move(&mp
->synclist
, &tblk
->synclist
);
2842 /* inherit younger/larger clsn */
2844 logdiff(difft
, tblk
->clsn
, log
);
2845 logdiff(diffp
, mp
->clsn
, log
);
2847 mp
->clsn
= tblk
->clsn
;
2851 /* insert mp after tblock in logsync list */
2853 list_add(&mp
->synclist
, &tblk
->synclist
);
2854 mp
->clsn
= tblk
->clsn
;
2856 LOGSYNC_UNLOCK(log
, flags
);
2864 * function: update imap for extendfs();
2866 * note: AG size has been increased s.t. each k old contiguous AGs are
2867 * coalesced into a new AG;
2869 int diExtendFS(struct inode
*ipimap
, struct inode
*ipbmap
)
2872 struct inomap
*imap
= JFS_IP(ipimap
)->i_imap
;
2873 struct iag
*iagp
= NULL
, *hiagp
= NULL
;
2874 struct bmap
*mp
= JFS_SBI(ipbmap
->i_sb
)->bmap
;
2875 struct metapage
*bp
, *hbp
;
2877 int numinos
, xnuminos
= 0, xnumfree
= 0;
2880 jfs_info("diExtendFS: nextiag:%d numinos:%d numfree:%d",
2881 imap
->im_nextiag
, atomic_read(&imap
->im_numinos
),
2882 atomic_read(&imap
->im_numfree
));
2887 * coalesce contiguous k (newAGSize/oldAGSize) AGs;
2888 * i.e., (AGi, ..., AGj) where i = k*n and j = k*(n+1) - 1 to AGn;
2889 * note: new AG size = old AG size * (2**x).
2892 /* init per AG control information im_agctl[] */
2893 for (i
= 0; i
< MAXAG
; i
++) {
2894 imap
->im_agctl
[i
].inofree
= -1;
2895 imap
->im_agctl
[i
].extfree
= -1;
2896 imap
->im_agctl
[i
].numinos
= 0; /* number of backed inodes */
2897 imap
->im_agctl
[i
].numfree
= 0; /* number of free backed inodes */
2901 * process each iag page of the map.
2903 * rebuild AG Free Inode List, AG Free Inode Extent List;
2905 for (i
= 0; i
< imap
->im_nextiag
; i
++) {
2906 if ((rc
= diIAGRead(imap
, i
, &bp
))) {
2910 iagp
= (struct iag
*) bp
->data
;
2911 if (le32_to_cpu(iagp
->iagnum
) != i
) {
2912 release_metapage(bp
);
2913 jfs_error(ipimap
->i_sb
,
2914 "diExtendFs: unexpected value of iagnum");
2918 /* leave free iag in the free iag list */
2919 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
)) {
2920 release_metapage(bp
);
2924 /* agstart that computes to the same ag is treated as same; */
2925 agstart
= le64_to_cpu(iagp
->agstart
);
2926 /* iagp->agstart = agstart & ~(mp->db_agsize - 1); */
2927 n
= agstart
>> mp
->db_agl2size
;
2929 /* compute backed inodes */
2930 numinos
= (EXTSPERIAG
- le32_to_cpu(iagp
->nfreeexts
))
2933 /* merge AG backed inodes */
2934 imap
->im_agctl
[n
].numinos
+= numinos
;
2935 xnuminos
+= numinos
;
2938 /* if any backed free inodes, insert at AG free inode list */
2939 if ((int) le32_to_cpu(iagp
->nfreeinos
) > 0) {
2940 if ((head
= imap
->im_agctl
[n
].inofree
) == -1) {
2941 iagp
->inofreefwd
= cpu_to_le32(-1);
2942 iagp
->inofreeback
= cpu_to_le32(-1);
2944 if ((rc
= diIAGRead(imap
, head
, &hbp
))) {
2948 hiagp
= (struct iag
*) hbp
->data
;
2949 hiagp
->inofreeback
= iagp
->iagnum
;
2950 iagp
->inofreefwd
= cpu_to_le32(head
);
2951 iagp
->inofreeback
= cpu_to_le32(-1);
2952 write_metapage(hbp
);
2955 imap
->im_agctl
[n
].inofree
=
2956 le32_to_cpu(iagp
->iagnum
);
2958 /* merge AG backed free inodes */
2959 imap
->im_agctl
[n
].numfree
+=
2960 le32_to_cpu(iagp
->nfreeinos
);
2961 xnumfree
+= le32_to_cpu(iagp
->nfreeinos
);
2964 /* if any free extents, insert at AG free extent list */
2965 if (le32_to_cpu(iagp
->nfreeexts
) > 0) {
2966 if ((head
= imap
->im_agctl
[n
].extfree
) == -1) {
2967 iagp
->extfreefwd
= cpu_to_le32(-1);
2968 iagp
->extfreeback
= cpu_to_le32(-1);
2970 if ((rc
= diIAGRead(imap
, head
, &hbp
))) {
2974 hiagp
= (struct iag
*) hbp
->data
;
2975 hiagp
->extfreeback
= iagp
->iagnum
;
2976 iagp
->extfreefwd
= cpu_to_le32(head
);
2977 iagp
->extfreeback
= cpu_to_le32(-1);
2978 write_metapage(hbp
);
2981 imap
->im_agctl
[n
].extfree
=
2982 le32_to_cpu(iagp
->iagnum
);
2989 if (xnuminos
!= atomic_read(&imap
->im_numinos
) ||
2990 xnumfree
!= atomic_read(&imap
->im_numfree
)) {
2991 jfs_error(ipimap
->i_sb
,
2992 "diExtendFs: numinos or numfree incorrect");
3003 * serialization: IWRITE_LOCK held on entry/exit
3005 * note: shadow page with regular inode (rel.2);
3007 static void duplicateIXtree(struct super_block
*sb
, s64 blkno
,
3008 int xlen
, s64
*xaddr
)
3010 struct jfs_superblock
*j_sb
;
3011 struct buffer_head
*bh
;
3015 /* if AIT2 ipmap2 is bad, do not try to update it */
3016 if (JFS_SBI(sb
)->mntflag
& JFS_BAD_SAIT
) /* s_flag */
3018 ip
= diReadSpecial(sb
, FILESYSTEM_I
, 1);
3020 JFS_SBI(sb
)->mntflag
|= JFS_BAD_SAIT
;
3021 if (readSuper(sb
, &bh
))
3023 j_sb
= (struct jfs_superblock
*)bh
->b_data
;
3024 j_sb
->s_flag
|= cpu_to_le32(JFS_BAD_SAIT
);
3026 mark_buffer_dirty(bh
);
3027 sync_dirty_buffer(bh
);
3032 /* start transaction */
3033 tid
= txBegin(sb
, COMMIT_FORCE
);
3034 /* update the inode map addressing structure to point to it */
3035 if (xtInsert(tid
, ip
, 0, blkno
, xlen
, xaddr
, 0)) {
3036 JFS_SBI(sb
)->mntflag
|= JFS_BAD_SAIT
;
3041 /* update the inode map's inode to reflect the extension */
3042 ip
->i_size
+= PSIZE
;
3043 inode_add_bytes(ip
, PSIZE
);
3044 txCommit(tid
, 1, &ip
, COMMIT_FORCE
);
3051 * NAME: copy_from_dinode()
3053 * FUNCTION: Copies inode info from disk inode to in-memory inode
3057 * -ENOMEM - insufficient memory
3059 static int copy_from_dinode(struct dinode
* dip
, struct inode
*ip
)
3061 struct jfs_inode_info
*jfs_ip
= JFS_IP(ip
);
3062 struct jfs_sb_info
*sbi
= JFS_SBI(ip
->i_sb
);
3064 jfs_ip
->fileset
= le32_to_cpu(dip
->di_fileset
);
3065 jfs_ip
->mode2
= le32_to_cpu(dip
->di_mode
);
3066 jfs_set_inode_flags(ip
);
3068 ip
->i_mode
= le32_to_cpu(dip
->di_mode
) & 0xffff;
3069 if (sbi
->umask
!= -1) {
3070 ip
->i_mode
= (ip
->i_mode
& ~0777) | (0777 & ~sbi
->umask
);
3071 /* For directories, add x permission if r is allowed by umask */
3072 if (S_ISDIR(ip
->i_mode
)) {
3073 if (ip
->i_mode
& 0400)
3075 if (ip
->i_mode
& 0040)
3077 if (ip
->i_mode
& 0004)
3081 ip
->i_nlink
= le32_to_cpu(dip
->di_nlink
);
3083 jfs_ip
->saved_uid
= le32_to_cpu(dip
->di_uid
);
3085 ip
->i_uid
= jfs_ip
->saved_uid
;
3087 ip
->i_uid
= sbi
->uid
;
3090 jfs_ip
->saved_gid
= le32_to_cpu(dip
->di_gid
);
3092 ip
->i_gid
= jfs_ip
->saved_gid
;
3094 ip
->i_gid
= sbi
->gid
;
3097 ip
->i_size
= le64_to_cpu(dip
->di_size
);
3098 ip
->i_atime
.tv_sec
= le32_to_cpu(dip
->di_atime
.tv_sec
);
3099 ip
->i_atime
.tv_nsec
= le32_to_cpu(dip
->di_atime
.tv_nsec
);
3100 ip
->i_mtime
.tv_sec
= le32_to_cpu(dip
->di_mtime
.tv_sec
);
3101 ip
->i_mtime
.tv_nsec
= le32_to_cpu(dip
->di_mtime
.tv_nsec
);
3102 ip
->i_ctime
.tv_sec
= le32_to_cpu(dip
->di_ctime
.tv_sec
);
3103 ip
->i_ctime
.tv_nsec
= le32_to_cpu(dip
->di_ctime
.tv_nsec
);
3104 ip
->i_blocks
= LBLK2PBLK(ip
->i_sb
, le64_to_cpu(dip
->di_nblocks
));
3105 ip
->i_generation
= le32_to_cpu(dip
->di_gen
);
3107 jfs_ip
->ixpxd
= dip
->di_ixpxd
; /* in-memory pxd's are little-endian */
3108 jfs_ip
->acl
= dip
->di_acl
; /* as are dxd's */
3109 jfs_ip
->ea
= dip
->di_ea
;
3110 jfs_ip
->next_index
= le32_to_cpu(dip
->di_next_index
);
3111 jfs_ip
->otime
= le32_to_cpu(dip
->di_otime
.tv_sec
);
3112 jfs_ip
->acltype
= le32_to_cpu(dip
->di_acltype
);
3114 if (S_ISCHR(ip
->i_mode
) || S_ISBLK(ip
->i_mode
)) {
3115 jfs_ip
->dev
= le32_to_cpu(dip
->di_rdev
);
3116 ip
->i_rdev
= new_decode_dev(jfs_ip
->dev
);
3119 if (S_ISDIR(ip
->i_mode
)) {
3120 memcpy(&jfs_ip
->i_dirtable
, &dip
->di_dirtable
, 384);
3121 } else if (S_ISREG(ip
->i_mode
) || S_ISLNK(ip
->i_mode
)) {
3122 memcpy(&jfs_ip
->i_xtroot
, &dip
->di_xtroot
, 288);
3124 memcpy(&jfs_ip
->i_inline_ea
, &dip
->di_inlineea
, 128);
3126 /* Zero the in-memory-only stuff */
3128 jfs_ip
->btindex
= 0;
3129 jfs_ip
->btorder
= 0;
3132 jfs_ip
->atlhead
= 0;
3133 jfs_ip
->atltail
= 0;
3139 * NAME: copy_to_dinode()
3141 * FUNCTION: Copies inode info from in-memory inode to disk inode
3143 static void copy_to_dinode(struct dinode
* dip
, struct inode
*ip
)
3145 struct jfs_inode_info
*jfs_ip
= JFS_IP(ip
);
3146 struct jfs_sb_info
*sbi
= JFS_SBI(ip
->i_sb
);
3148 dip
->di_fileset
= cpu_to_le32(jfs_ip
->fileset
);
3149 dip
->di_inostamp
= cpu_to_le32(sbi
->inostamp
);
3150 dip
->di_number
= cpu_to_le32(ip
->i_ino
);
3151 dip
->di_gen
= cpu_to_le32(ip
->i_generation
);
3152 dip
->di_size
= cpu_to_le64(ip
->i_size
);
3153 dip
->di_nblocks
= cpu_to_le64(PBLK2LBLK(ip
->i_sb
, ip
->i_blocks
));
3154 dip
->di_nlink
= cpu_to_le32(ip
->i_nlink
);
3156 dip
->di_uid
= cpu_to_le32(ip
->i_uid
);
3158 dip
->di_uid
= cpu_to_le32(jfs_ip
->saved_uid
);
3160 dip
->di_gid
= cpu_to_le32(ip
->i_gid
);
3162 dip
->di_gid
= cpu_to_le32(jfs_ip
->saved_gid
);
3163 jfs_get_inode_flags(jfs_ip
);
3165 * mode2 is only needed for storing the higher order bits.
3166 * Trust i_mode for the lower order ones
3168 if (sbi
->umask
== -1)
3169 dip
->di_mode
= cpu_to_le32((jfs_ip
->mode2
& 0xffff0000) |
3171 else /* Leave the original permissions alone */
3172 dip
->di_mode
= cpu_to_le32(jfs_ip
->mode2
);
3174 dip
->di_atime
.tv_sec
= cpu_to_le32(ip
->i_atime
.tv_sec
);
3175 dip
->di_atime
.tv_nsec
= cpu_to_le32(ip
->i_atime
.tv_nsec
);
3176 dip
->di_ctime
.tv_sec
= cpu_to_le32(ip
->i_ctime
.tv_sec
);
3177 dip
->di_ctime
.tv_nsec
= cpu_to_le32(ip
->i_ctime
.tv_nsec
);
3178 dip
->di_mtime
.tv_sec
= cpu_to_le32(ip
->i_mtime
.tv_sec
);
3179 dip
->di_mtime
.tv_nsec
= cpu_to_le32(ip
->i_mtime
.tv_nsec
);
3180 dip
->di_ixpxd
= jfs_ip
->ixpxd
; /* in-memory pxd's are little-endian */
3181 dip
->di_acl
= jfs_ip
->acl
; /* as are dxd's */
3182 dip
->di_ea
= jfs_ip
->ea
;
3183 dip
->di_next_index
= cpu_to_le32(jfs_ip
->next_index
);
3184 dip
->di_otime
.tv_sec
= cpu_to_le32(jfs_ip
->otime
);
3185 dip
->di_otime
.tv_nsec
= 0;
3186 dip
->di_acltype
= cpu_to_le32(jfs_ip
->acltype
);
3187 if (S_ISCHR(ip
->i_mode
) || S_ISBLK(ip
->i_mode
))
3188 dip
->di_rdev
= cpu_to_le32(jfs_ip
->dev
);
This page took 0.15725 seconds and 5 git commands to generate.