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 print_hex_dump(KERN_ERR
, "imap: ", DUMP_PREFIX_ADDRESS
, 16, 4,
896 "diFree: inum = %d, iagno = %d, nextiag = %d",
897 (uint
) inum
, iagno
, imap
->im_nextiag
);
901 /* get the allocation group for this ino.
903 agno
= JFS_IP(ip
)->agno
;
905 /* Lock the AG specific inode map information
909 /* Obtain read lock in imap inode. Don't release it until we have
910 * read all of the IAG's that we are going to.
912 IREAD_LOCK(ipimap
, RDWRLOCK_IMAP
);
916 if ((rc
= diIAGRead(imap
, iagno
, &mp
))) {
917 IREAD_UNLOCK(ipimap
);
918 AG_UNLOCK(imap
, agno
);
921 iagp
= (struct iag
*) mp
->data
;
923 /* get the inode number and extent number of the inode within
924 * the iag and the inode number within the extent.
926 ino
= inum
& (INOSPERIAG
- 1);
927 extno
= ino
>> L2INOSPEREXT
;
928 bitno
= ino
& (INOSPEREXT
- 1);
929 mask
= HIGHORDER
>> bitno
;
931 if (!(le32_to_cpu(iagp
->wmap
[extno
]) & mask
)) {
933 "diFree: wmap shows inode already free");
936 if (!addressPXD(&iagp
->inoext
[extno
])) {
937 release_metapage(mp
);
938 IREAD_UNLOCK(ipimap
);
939 AG_UNLOCK(imap
, agno
);
940 jfs_error(ip
->i_sb
, "diFree: invalid inoext");
944 /* compute the bitmap for the extent reflecting the freed inode.
946 bitmap
= le32_to_cpu(iagp
->wmap
[extno
]) & ~mask
;
948 if (imap
->im_agctl
[agno
].numfree
> imap
->im_agctl
[agno
].numinos
) {
949 release_metapage(mp
);
950 IREAD_UNLOCK(ipimap
);
951 AG_UNLOCK(imap
, agno
);
952 jfs_error(ip
->i_sb
, "diFree: numfree > numinos");
956 * inode extent still has some inodes or below low water mark:
957 * keep the inode extent;
960 imap
->im_agctl
[agno
].numfree
< 96 ||
961 (imap
->im_agctl
[agno
].numfree
< 288 &&
962 (((imap
->im_agctl
[agno
].numfree
* 100) /
963 imap
->im_agctl
[agno
].numinos
) <= 25))) {
964 /* if the iag currently has no free inodes (i.e.,
965 * the inode being freed is the first free inode of iag),
966 * insert the iag at head of the inode free list for the ag.
968 if (iagp
->nfreeinos
== 0) {
969 /* check if there are any iags on the ag inode
970 * free list. if so, read the first one so that
971 * we can link the current iag onto the list at
974 if ((fwd
= imap
->im_agctl
[agno
].inofree
) >= 0) {
975 /* read the iag that currently is the head
978 if ((rc
= diIAGRead(imap
, fwd
, &
))) {
979 IREAD_UNLOCK(ipimap
);
980 AG_UNLOCK(imap
, agno
);
981 release_metapage(mp
);
984 aiagp
= (struct iag
*) amp
->data
;
986 /* make current head point back to the iag.
988 aiagp
->inofreeback
= cpu_to_le32(iagno
);
993 /* iag points forward to current head and iag
994 * becomes the new head of the list.
997 cpu_to_le32(imap
->im_agctl
[agno
].inofree
);
998 iagp
->inofreeback
= cpu_to_le32(-1);
999 imap
->im_agctl
[agno
].inofree
= iagno
;
1001 IREAD_UNLOCK(ipimap
);
1003 /* update the free inode summary map for the extent if
1004 * freeing the inode means the extent will now have free
1005 * inodes (i.e., the inode being freed is the first free
1008 if (iagp
->wmap
[extno
] == cpu_to_le32(ONES
)) {
1009 sword
= extno
>> L2EXTSPERSUM
;
1010 bitno
= extno
& (EXTSPERSUM
- 1);
1011 iagp
->inosmap
[sword
] &=
1012 cpu_to_le32(~(HIGHORDER
>> bitno
));
1015 /* update the bitmap.
1017 iagp
->wmap
[extno
] = cpu_to_le32(bitmap
);
1019 /* update the free inode counts at the iag, ag and
1023 cpu_to_le32(le32_to_cpu(iagp
->nfreeinos
) + 1);
1024 imap
->im_agctl
[agno
].numfree
+= 1;
1025 atomic_inc(&imap
->im_numfree
);
1027 /* release the AG inode map lock
1029 AG_UNLOCK(imap
, agno
);
1039 * inode extent has become free and above low water mark:
1040 * free the inode extent;
1044 * prepare to update iag list(s) (careful update step 1)
1046 amp
= bmp
= cmp
= dmp
= NULL
;
1049 /* check if the iag currently has no free extents. if so,
1050 * it will be placed on the head of the ag extent free list.
1052 if (iagp
->nfreeexts
== 0) {
1053 /* check if the ag extent free list has any iags.
1054 * if so, read the iag at the head of the list now.
1055 * this (head) iag will be updated later to reflect
1056 * the addition of the current iag at the head of
1059 if ((fwd
= imap
->im_agctl
[agno
].extfree
) >= 0) {
1060 if ((rc
= diIAGRead(imap
, fwd
, &
)))
1062 aiagp
= (struct iag
*) amp
->data
;
1065 /* iag has free extents. check if the addition of a free
1066 * extent will cause all extents to be free within this
1067 * iag. if so, the iag will be removed from the ag extent
1068 * free list and placed on the inode map's free iag list.
1070 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
- 1)) {
1071 /* in preparation for removing the iag from the
1072 * ag extent free list, read the iags preceeding
1073 * and following the iag on the ag extent free
1076 if ((fwd
= le32_to_cpu(iagp
->extfreefwd
)) >= 0) {
1077 if ((rc
= diIAGRead(imap
, fwd
, &
)))
1079 aiagp
= (struct iag
*) amp
->data
;
1082 if ((back
= le32_to_cpu(iagp
->extfreeback
)) >= 0) {
1083 if ((rc
= diIAGRead(imap
, back
, &bmp
)))
1085 biagp
= (struct iag
*) bmp
->data
;
1090 /* remove the iag from the ag inode free list if freeing
1091 * this extent cause the iag to have no free inodes.
1093 if (iagp
->nfreeinos
== cpu_to_le32(INOSPEREXT
- 1)) {
1094 int inofreeback
= le32_to_cpu(iagp
->inofreeback
);
1095 int inofreefwd
= le32_to_cpu(iagp
->inofreefwd
);
1097 /* in preparation for removing the iag from the
1098 * ag inode free list, read the iags preceeding
1099 * and following the iag on the ag inode free
1100 * list. before reading these iags, we must make
1101 * sure that we already don't have them in hand
1102 * from up above, since re-reading an iag (buffer)
1103 * we are currently holding would cause a deadlock.
1105 if (inofreefwd
>= 0) {
1107 if (inofreefwd
== fwd
)
1108 ciagp
= (struct iag
*) amp
->data
;
1109 else if (inofreefwd
== back
)
1110 ciagp
= (struct iag
*) bmp
->data
;
1113 diIAGRead(imap
, inofreefwd
, &cmp
)))
1115 ciagp
= (struct iag
*) cmp
->data
;
1117 assert(ciagp
!= NULL
);
1120 if (inofreeback
>= 0) {
1121 if (inofreeback
== fwd
)
1122 diagp
= (struct iag
*) amp
->data
;
1123 else if (inofreeback
== back
)
1124 diagp
= (struct iag
*) bmp
->data
;
1127 diIAGRead(imap
, inofreeback
, &dmp
)))
1129 diagp
= (struct iag
*) dmp
->data
;
1131 assert(diagp
!= NULL
);
1135 IREAD_UNLOCK(ipimap
);
1138 * invalidate any page of the inode extent freed from buffer cache;
1140 freepxd
= iagp
->inoext
[extno
];
1141 invalidate_pxd_metapages(ip
, freepxd
);
1144 * update iag list(s) (careful update step 2)
1146 /* add the iag to the ag extent free list if this is the
1147 * first free extent for the iag.
1149 if (iagp
->nfreeexts
== 0) {
1151 aiagp
->extfreeback
= cpu_to_le32(iagno
);
1154 cpu_to_le32(imap
->im_agctl
[agno
].extfree
);
1155 iagp
->extfreeback
= cpu_to_le32(-1);
1156 imap
->im_agctl
[agno
].extfree
= iagno
;
1158 /* remove the iag from the ag extent list if all extents
1159 * are now free and place it on the inode map iag free list.
1161 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
- 1)) {
1163 aiagp
->extfreeback
= iagp
->extfreeback
;
1166 biagp
->extfreefwd
= iagp
->extfreefwd
;
1168 imap
->im_agctl
[agno
].extfree
=
1169 le32_to_cpu(iagp
->extfreefwd
);
1171 iagp
->extfreefwd
= iagp
->extfreeback
= cpu_to_le32(-1);
1174 iagp
->iagfree
= cpu_to_le32(imap
->im_freeiag
);
1175 imap
->im_freeiag
= iagno
;
1176 IAGFREE_UNLOCK(imap
);
1180 /* remove the iag from the ag inode free list if freeing
1181 * this extent causes the iag to have no free inodes.
1183 if (iagp
->nfreeinos
== cpu_to_le32(INOSPEREXT
- 1)) {
1184 if ((int) le32_to_cpu(iagp
->inofreefwd
) >= 0)
1185 ciagp
->inofreeback
= iagp
->inofreeback
;
1187 if ((int) le32_to_cpu(iagp
->inofreeback
) >= 0)
1188 diagp
->inofreefwd
= iagp
->inofreefwd
;
1190 imap
->im_agctl
[agno
].inofree
=
1191 le32_to_cpu(iagp
->inofreefwd
);
1193 iagp
->inofreefwd
= iagp
->inofreeback
= cpu_to_le32(-1);
1196 /* update the inode extent address and working map
1197 * to reflect the free extent.
1198 * the permanent map should have been updated already
1199 * for the inode being freed.
1201 if (iagp
->pmap
[extno
] != 0) {
1202 jfs_error(ip
->i_sb
, "diFree: the pmap does not show inode free");
1204 iagp
->wmap
[extno
] = 0;
1205 PXDlength(&iagp
->inoext
[extno
], 0);
1206 PXDaddress(&iagp
->inoext
[extno
], 0);
1208 /* update the free extent and free inode summary maps
1209 * to reflect the freed extent.
1210 * the inode summary map is marked to indicate no inodes
1211 * available for the freed extent.
1213 sword
= extno
>> L2EXTSPERSUM
;
1214 bitno
= extno
& (EXTSPERSUM
- 1);
1215 mask
= HIGHORDER
>> bitno
;
1216 iagp
->inosmap
[sword
] |= cpu_to_le32(mask
);
1217 iagp
->extsmap
[sword
] &= cpu_to_le32(~mask
);
1219 /* update the number of free inodes and number of free extents
1222 iagp
->nfreeinos
= cpu_to_le32(le32_to_cpu(iagp
->nfreeinos
) -
1224 iagp
->nfreeexts
= cpu_to_le32(le32_to_cpu(iagp
->nfreeexts
) + 1);
1226 /* update the number of free inodes and backed inodes
1227 * at the ag and inode map level.
1229 imap
->im_agctl
[agno
].numfree
-= (INOSPEREXT
- 1);
1230 imap
->im_agctl
[agno
].numinos
-= INOSPEREXT
;
1231 atomic_sub(INOSPEREXT
- 1, &imap
->im_numfree
);
1232 atomic_sub(INOSPEREXT
, &imap
->im_numinos
);
1235 write_metapage(amp
);
1237 write_metapage(bmp
);
1239 write_metapage(cmp
);
1241 write_metapage(dmp
);
1244 * start transaction to update block allocation map
1245 * for the inode extent freed;
1247 * N.B. AG_LOCK is released and iag will be released below, and
1248 * other thread may allocate inode from/reusing the ixad freed
1249 * BUT with new/different backing inode extent from the extent
1250 * to be freed by the transaction;
1252 tid
= txBegin(ipimap
->i_sb
, COMMIT_FORCE
);
1253 mutex_lock(&JFS_IP(ipimap
)->commit_mutex
);
1255 /* acquire tlock of the iag page of the freed ixad
1256 * to force the page NOHOMEOK (even though no data is
1257 * logged from the iag page) until NOREDOPAGE|FREEXTENT log
1258 * for the free of the extent is committed;
1259 * write FREEXTENT|NOREDOPAGE log record
1260 * N.B. linelock is overlaid as freed extent descriptor;
1262 tlck
= txLock(tid
, ipimap
, mp
, tlckINODE
| tlckFREE
);
1263 pxdlock
= (struct pxd_lock
*) & tlck
->lock
;
1264 pxdlock
->flag
= mlckFREEPXD
;
1265 pxdlock
->pxd
= freepxd
;
1273 * logredo needs the IAG number and IAG extent index in order
1274 * to ensure that the IMap is consistent. The least disruptive
1275 * way to pass these values through to the transaction manager
1276 * is in the iplist array.
1278 * It's not pretty, but it works.
1280 iplist
[1] = (struct inode
*) (size_t)iagno
;
1281 iplist
[2] = (struct inode
*) (size_t)extno
;
1283 rc
= txCommit(tid
, 1, &iplist
[0], COMMIT_FORCE
);
1286 mutex_unlock(&JFS_IP(ipimap
)->commit_mutex
);
1288 /* unlock the AG inode map information */
1289 AG_UNLOCK(imap
, agno
);
1294 IREAD_UNLOCK(ipimap
);
1297 release_metapage(amp
);
1299 release_metapage(bmp
);
1301 release_metapage(cmp
);
1303 release_metapage(dmp
);
1305 AG_UNLOCK(imap
, agno
);
1307 release_metapage(mp
);
1313 * There are several places in the diAlloc* routines where we initialize
1317 diInitInode(struct inode
*ip
, int iagno
, int ino
, int extno
, struct iag
* iagp
)
1319 struct jfs_sb_info
*sbi
= JFS_SBI(ip
->i_sb
);
1320 struct jfs_inode_info
*jfs_ip
= JFS_IP(ip
);
1322 ip
->i_ino
= (iagno
<< L2INOSPERIAG
) + ino
;
1323 jfs_ip
->ixpxd
= iagp
->inoext
[extno
];
1324 jfs_ip
->agno
= BLKTOAG(le64_to_cpu(iagp
->agstart
), sbi
);
1325 jfs_ip
->active_ag
= -1;
1330 * NAME: diAlloc(pip,dir,ip)
1332 * FUNCTION: allocate a disk inode from the inode working map
1333 * for a fileset or aggregate.
1336 * pip - pointer to incore inode for the parent inode.
1337 * dir - 'true' if the new disk inode is for a directory.
1338 * ip - pointer to a new inode
1342 * -ENOSPC - insufficient disk resources.
1345 int diAlloc(struct inode
*pip
, bool dir
, struct inode
*ip
)
1347 int rc
, ino
, iagno
, addext
, extno
, bitno
, sword
;
1348 int nwords
, rem
, i
, agno
;
1349 u32 mask
, inosmap
, extsmap
;
1350 struct inode
*ipimap
;
1351 struct metapage
*mp
;
1354 struct inomap
*imap
;
1356 /* get the pointers to the inode map inode and the
1357 * corresponding imap control structure.
1359 ipimap
= JFS_SBI(pip
->i_sb
)->ipimap
;
1360 imap
= JFS_IP(ipimap
)->i_imap
;
1361 JFS_IP(ip
)->ipimap
= ipimap
;
1362 JFS_IP(ip
)->fileset
= FILESYSTEM_I
;
1364 /* for a directory, the allocation policy is to start
1365 * at the ag level using the preferred ag.
1368 agno
= dbNextAG(JFS_SBI(pip
->i_sb
)->ipbmap
);
1369 AG_LOCK(imap
, agno
);
1373 /* for files, the policy starts off by trying to allocate from
1374 * the same iag containing the parent disk inode:
1375 * try to allocate the new disk inode close to the parent disk
1376 * inode, using parent disk inode number + 1 as the allocation
1377 * hint. (we use a left-to-right policy to attempt to avoid
1378 * moving backward on the disk.) compute the hint within the
1379 * file system and the iag.
1382 /* get the ag number of this iag */
1383 agno
= JFS_IP(pip
)->agno
;
1385 if (atomic_read(&JFS_SBI(pip
->i_sb
)->bmap
->db_active
[agno
])) {
1387 * There is an open file actively growing. We want to
1388 * allocate new inodes from a different ag to avoid
1389 * fragmentation problems.
1391 agno
= dbNextAG(JFS_SBI(pip
->i_sb
)->ipbmap
);
1392 AG_LOCK(imap
, agno
);
1396 inum
= pip
->i_ino
+ 1;
1397 ino
= inum
& (INOSPERIAG
- 1);
1399 /* back off the hint if it is outside of the iag */
1403 /* lock the AG inode map information */
1404 AG_LOCK(imap
, agno
);
1406 /* Get read lock on imap inode */
1407 IREAD_LOCK(ipimap
, RDWRLOCK_IMAP
);
1409 /* get the iag number and read the iag */
1410 iagno
= INOTOIAG(inum
);
1411 if ((rc
= diIAGRead(imap
, iagno
, &mp
))) {
1412 IREAD_UNLOCK(ipimap
);
1413 AG_UNLOCK(imap
, agno
);
1416 iagp
= (struct iag
*) mp
->data
;
1418 /* determine if new inode extent is allowed to be added to the iag.
1419 * new inode extent can be added to the iag if the ag
1420 * has less than 32 free disk inodes and the iag has free extents.
1422 addext
= (imap
->im_agctl
[agno
].numfree
< 32 && iagp
->nfreeexts
);
1425 * try to allocate from the IAG
1427 /* check if the inode may be allocated from the iag
1428 * (i.e. the inode has free inodes or new extent can be added).
1430 if (iagp
->nfreeinos
|| addext
) {
1431 /* determine the extent number of the hint.
1433 extno
= ino
>> L2INOSPEREXT
;
1435 /* check if the extent containing the hint has backed
1436 * inodes. if so, try to allocate within this extent.
1438 if (addressPXD(&iagp
->inoext
[extno
])) {
1439 bitno
= ino
& (INOSPEREXT
- 1);
1441 diFindFree(le32_to_cpu(iagp
->wmap
[extno
]),
1444 ino
= (extno
<< L2INOSPEREXT
) + bitno
;
1446 /* a free inode (bit) was found within this
1447 * extent, so allocate it.
1449 rc
= diAllocBit(imap
, iagp
, ino
);
1450 IREAD_UNLOCK(ipimap
);
1454 /* set the results of the allocation
1455 * and write the iag.
1457 diInitInode(ip
, iagno
, ino
, extno
,
1459 mark_metapage_dirty(mp
);
1461 release_metapage(mp
);
1463 /* free the AG lock and return.
1465 AG_UNLOCK(imap
, agno
);
1472 EXTSPERIAG
- 1) ? 0 : extno
+ 1;
1476 * no free inodes within the extent containing the hint.
1478 * try to allocate from the backed extents following
1479 * hint or, if appropriate (i.e. addext is true), allocate
1480 * an extent of free inodes at or following the extent
1481 * containing the hint.
1483 * the free inode and free extent summary maps are used
1484 * here, so determine the starting summary map position
1485 * and the number of words we'll have to examine. again,
1486 * the approach is to allocate following the hint, so we
1487 * might have to initially ignore prior bits of the summary
1488 * map that represent extents prior to the extent containing
1489 * the hint and later revisit these bits.
1491 bitno
= extno
& (EXTSPERSUM
- 1);
1492 nwords
= (bitno
== 0) ? SMAPSZ
: SMAPSZ
+ 1;
1493 sword
= extno
>> L2EXTSPERSUM
;
1495 /* mask any prior bits for the starting words of the
1498 mask
= ONES
<< (EXTSPERSUM
- bitno
);
1499 inosmap
= le32_to_cpu(iagp
->inosmap
[sword
]) | mask
;
1500 extsmap
= le32_to_cpu(iagp
->extsmap
[sword
]) | mask
;
1502 /* scan the free inode and free extent summary maps for
1505 for (i
= 0; i
< nwords
; i
++) {
1506 /* check if this word of the free inode summary
1507 * map describes an extent with free inodes.
1510 /* an extent with free inodes has been
1511 * found. determine the extent number
1512 * and the inode number within the extent.
1514 rem
= diFindFree(inosmap
, 0);
1515 extno
= (sword
<< L2EXTSPERSUM
) + rem
;
1516 rem
= diFindFree(le32_to_cpu(iagp
->wmap
[extno
]),
1518 if (rem
>= INOSPEREXT
) {
1519 IREAD_UNLOCK(ipimap
);
1520 release_metapage(mp
);
1521 AG_UNLOCK(imap
, agno
);
1523 "diAlloc: can't find free bit "
1528 /* determine the inode number within the
1529 * iag and allocate the inode from the
1532 ino
= (extno
<< L2INOSPEREXT
) + rem
;
1533 rc
= diAllocBit(imap
, iagp
, ino
);
1534 IREAD_UNLOCK(ipimap
);
1538 /* set the results of the allocation
1539 * and write the iag.
1541 diInitInode(ip
, iagno
, ino
, extno
,
1543 mark_metapage_dirty(mp
);
1545 release_metapage(mp
);
1547 /* free the AG lock and return.
1549 AG_UNLOCK(imap
, agno
);
1554 /* check if we may allocate an extent of free
1555 * inodes and whether this word of the free
1556 * extents summary map describes a free extent.
1558 if (addext
&& ~extsmap
) {
1559 /* a free extent has been found. determine
1560 * the extent number.
1562 rem
= diFindFree(extsmap
, 0);
1563 extno
= (sword
<< L2EXTSPERSUM
) + rem
;
1565 /* allocate an extent of free inodes.
1567 if ((rc
= diNewExt(imap
, iagp
, extno
))) {
1568 /* if there is no disk space for a
1569 * new extent, try to allocate the
1570 * disk inode from somewhere else.
1577 /* set the results of the allocation
1578 * and write the iag.
1580 diInitInode(ip
, iagno
,
1581 extno
<< L2INOSPEREXT
,
1583 mark_metapage_dirty(mp
);
1585 release_metapage(mp
);
1586 /* free the imap inode & the AG lock & return.
1588 IREAD_UNLOCK(ipimap
);
1589 AG_UNLOCK(imap
, agno
);
1593 /* move on to the next set of summary map words.
1595 sword
= (sword
== SMAPSZ
- 1) ? 0 : sword
+ 1;
1596 inosmap
= le32_to_cpu(iagp
->inosmap
[sword
]);
1597 extsmap
= le32_to_cpu(iagp
->extsmap
[sword
]);
1600 /* unlock imap inode */
1601 IREAD_UNLOCK(ipimap
);
1603 /* nothing doing in this iag, so release it. */
1604 release_metapage(mp
);
1608 * try to allocate anywhere within the same AG as the parent inode.
1610 rc
= diAllocAG(imap
, agno
, dir
, ip
);
1612 AG_UNLOCK(imap
, agno
);
1618 * try to allocate in any AG.
1620 return (diAllocAny(imap
, agno
, dir
, ip
));
1625 * NAME: diAllocAG(imap,agno,dir,ip)
1627 * FUNCTION: allocate a disk inode from the allocation group.
1629 * this routine first determines if a new extent of free
1630 * inodes should be added for the allocation group, with
1631 * the current request satisfied from this extent. if this
1632 * is the case, an attempt will be made to do just that. if
1633 * this attempt fails or it has been determined that a new
1634 * extent should not be added, an attempt is made to satisfy
1635 * the request by allocating an existing (backed) free inode
1636 * from the allocation group.
1638 * PRE CONDITION: Already have the AG lock for this AG.
1641 * imap - pointer to inode map control structure.
1642 * agno - allocation group to allocate from.
1643 * dir - 'true' if the new disk inode is for a directory.
1644 * ip - pointer to the new inode to be filled in on successful return
1645 * with the disk inode number allocated, its extent address
1646 * and the start of the ag.
1650 * -ENOSPC - insufficient disk resources.
1654 diAllocAG(struct inomap
* imap
, int agno
, bool dir
, struct inode
*ip
)
1656 int rc
, addext
, numfree
, numinos
;
1658 /* get the number of free and the number of backed disk
1659 * inodes currently within the ag.
1661 numfree
= imap
->im_agctl
[agno
].numfree
;
1662 numinos
= imap
->im_agctl
[agno
].numinos
;
1664 if (numfree
> numinos
) {
1665 jfs_error(ip
->i_sb
, "diAllocAG: numfree > numinos");
1669 /* determine if we should allocate a new extent of free inodes
1670 * within the ag: for directory inodes, add a new extent
1671 * if there are a small number of free inodes or number of free
1672 * inodes is a small percentage of the number of backed inodes.
1675 addext
= (numfree
< 64 ||
1677 && ((numfree
* 100) / numinos
) <= 20));
1679 addext
= (numfree
== 0);
1682 * try to allocate a new extent of free inodes.
1685 /* if free space is not avaliable for this new extent, try
1686 * below to allocate a free and existing (already backed)
1687 * inode from the ag.
1689 if ((rc
= diAllocExt(imap
, agno
, ip
)) != -ENOSPC
)
1694 * try to allocate an existing free inode from the ag.
1696 return (diAllocIno(imap
, agno
, ip
));
1701 * NAME: diAllocAny(imap,agno,dir,iap)
1703 * FUNCTION: allocate a disk inode from any other allocation group.
1705 * this routine is called when an allocation attempt within
1706 * the primary allocation group has failed. if attempts to
1707 * allocate an inode from any allocation group other than the
1708 * specified primary group.
1711 * imap - pointer to inode map control structure.
1712 * agno - primary allocation group (to avoid).
1713 * dir - 'true' if the new disk inode is for a directory.
1714 * ip - pointer to a new inode to be filled in on successful return
1715 * with the disk inode number allocated, its extent address
1716 * and the start of the ag.
1720 * -ENOSPC - insufficient disk resources.
1724 diAllocAny(struct inomap
* imap
, int agno
, bool dir
, struct inode
*ip
)
1727 int maxag
= JFS_SBI(imap
->im_ipimap
->i_sb
)->bmap
->db_maxag
;
1730 /* try to allocate from the ags following agno up to
1731 * the maximum ag number.
1733 for (ag
= agno
+ 1; ag
<= maxag
; ag
++) {
1736 rc
= diAllocAG(imap
, ag
, dir
, ip
);
1738 AG_UNLOCK(imap
, ag
);
1744 /* try to allocate from the ags in front of agno.
1746 for (ag
= 0; ag
< agno
; ag
++) {
1749 rc
= diAllocAG(imap
, ag
, dir
, ip
);
1751 AG_UNLOCK(imap
, ag
);
1757 /* no free disk inodes.
1764 * NAME: diAllocIno(imap,agno,ip)
1766 * FUNCTION: allocate a disk inode from the allocation group's free
1767 * inode list, returning an error if this free list is
1768 * empty (i.e. no iags on the list).
1770 * allocation occurs from the first iag on the list using
1771 * the iag's free inode summary map to find the leftmost
1772 * free inode in the iag.
1774 * PRE CONDITION: Already have AG lock for this AG.
1777 * imap - pointer to inode map control structure.
1778 * agno - allocation group.
1779 * ip - pointer to new inode to be filled in on successful return
1780 * with the disk inode number allocated, its extent address
1781 * and the start of the ag.
1785 * -ENOSPC - insufficient disk resources.
1788 static int diAllocIno(struct inomap
* imap
, int agno
, struct inode
*ip
)
1790 int iagno
, ino
, rc
, rem
, extno
, sword
;
1791 struct metapage
*mp
;
1794 /* check if there are iags on the ag's free inode list.
1796 if ((iagno
= imap
->im_agctl
[agno
].inofree
) < 0)
1799 /* obtain read lock on imap inode */
1800 IREAD_LOCK(imap
->im_ipimap
, RDWRLOCK_IMAP
);
1802 /* read the iag at the head of the list.
1804 if ((rc
= diIAGRead(imap
, iagno
, &mp
))) {
1805 IREAD_UNLOCK(imap
->im_ipimap
);
1808 iagp
= (struct iag
*) mp
->data
;
1810 /* better be free inodes in this iag if it is on the
1813 if (!iagp
->nfreeinos
) {
1814 IREAD_UNLOCK(imap
->im_ipimap
);
1815 release_metapage(mp
);
1817 "diAllocIno: nfreeinos = 0, but iag on freelist");
1821 /* scan the free inode summary map to find an extent
1824 for (sword
= 0;; sword
++) {
1825 if (sword
>= SMAPSZ
) {
1826 IREAD_UNLOCK(imap
->im_ipimap
);
1827 release_metapage(mp
);
1829 "diAllocIno: free inode not found in summary map");
1833 if (~iagp
->inosmap
[sword
])
1837 /* found a extent with free inodes. determine
1838 * the extent number.
1840 rem
= diFindFree(le32_to_cpu(iagp
->inosmap
[sword
]), 0);
1841 if (rem
>= EXTSPERSUM
) {
1842 IREAD_UNLOCK(imap
->im_ipimap
);
1843 release_metapage(mp
);
1844 jfs_error(ip
->i_sb
, "diAllocIno: no free extent found");
1847 extno
= (sword
<< L2EXTSPERSUM
) + rem
;
1849 /* find the first free inode in the extent.
1851 rem
= diFindFree(le32_to_cpu(iagp
->wmap
[extno
]), 0);
1852 if (rem
>= INOSPEREXT
) {
1853 IREAD_UNLOCK(imap
->im_ipimap
);
1854 release_metapage(mp
);
1855 jfs_error(ip
->i_sb
, "diAllocIno: free inode not found");
1859 /* compute the inode number within the iag.
1861 ino
= (extno
<< L2INOSPEREXT
) + rem
;
1863 /* allocate the inode.
1865 rc
= diAllocBit(imap
, iagp
, ino
);
1866 IREAD_UNLOCK(imap
->im_ipimap
);
1868 release_metapage(mp
);
1872 /* set the results of the allocation and write the iag.
1874 diInitInode(ip
, iagno
, ino
, extno
, iagp
);
1882 * NAME: diAllocExt(imap,agno,ip)
1884 * FUNCTION: add a new extent of free inodes to an iag, allocating
1885 * an inode from this extent to satisfy the current allocation
1888 * this routine first tries to find an existing iag with free
1889 * extents through the ag free extent list. if list is not
1890 * empty, the head of the list will be selected as the home
1891 * of the new extent of free inodes. otherwise (the list is
1892 * empty), a new iag will be allocated for the ag to contain
1895 * once an iag has been selected, the free extent summary map
1896 * is used to locate a free extent within the iag and diNewExt()
1897 * is called to initialize the extent, with initialization
1898 * including the allocation of the first inode of the extent
1899 * for the purpose of satisfying this request.
1902 * imap - pointer to inode map control structure.
1903 * agno - allocation group number.
1904 * ip - pointer to new inode to be filled in on successful return
1905 * with the disk inode number allocated, its extent address
1906 * and the start of the ag.
1910 * -ENOSPC - insufficient disk resources.
1913 static int diAllocExt(struct inomap
* imap
, int agno
, struct inode
*ip
)
1915 int rem
, iagno
, sword
, extno
, rc
;
1916 struct metapage
*mp
;
1919 /* check if the ag has any iags with free extents. if not,
1920 * allocate a new iag for the ag.
1922 if ((iagno
= imap
->im_agctl
[agno
].extfree
) < 0) {
1923 /* If successful, diNewIAG will obtain the read lock on the
1926 if ((rc
= diNewIAG(imap
, &iagno
, agno
, &mp
))) {
1929 iagp
= (struct iag
*) mp
->data
;
1931 /* set the ag number if this a brand new iag
1934 cpu_to_le64(AGTOBLK(agno
, imap
->im_ipimap
));
1938 IREAD_LOCK(imap
->im_ipimap
, RDWRLOCK_IMAP
);
1939 if ((rc
= diIAGRead(imap
, iagno
, &mp
))) {
1940 IREAD_UNLOCK(imap
->im_ipimap
);
1941 jfs_error(ip
->i_sb
, "diAllocExt: error reading iag");
1944 iagp
= (struct iag
*) mp
->data
;
1947 /* using the free extent summary map, find a free extent.
1949 for (sword
= 0;; sword
++) {
1950 if (sword
>= SMAPSZ
) {
1951 release_metapage(mp
);
1952 IREAD_UNLOCK(imap
->im_ipimap
);
1954 "diAllocExt: free ext summary map not found");
1957 if (~iagp
->extsmap
[sword
])
1961 /* determine the extent number of the free extent.
1963 rem
= diFindFree(le32_to_cpu(iagp
->extsmap
[sword
]), 0);
1964 if (rem
>= EXTSPERSUM
) {
1965 release_metapage(mp
);
1966 IREAD_UNLOCK(imap
->im_ipimap
);
1967 jfs_error(ip
->i_sb
, "diAllocExt: free extent not found");
1970 extno
= (sword
<< L2EXTSPERSUM
) + rem
;
1972 /* initialize the new extent.
1974 rc
= diNewExt(imap
, iagp
, extno
);
1975 IREAD_UNLOCK(imap
->im_ipimap
);
1977 /* something bad happened. if a new iag was allocated,
1978 * place it back on the inode map's iag free list, and
1979 * clear the ag number information.
1981 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
)) {
1983 iagp
->iagfree
= cpu_to_le32(imap
->im_freeiag
);
1984 imap
->im_freeiag
= iagno
;
1985 IAGFREE_UNLOCK(imap
);
1991 /* set the results of the allocation and write the iag.
1993 diInitInode(ip
, iagno
, extno
<< L2INOSPEREXT
, extno
, iagp
);
2002 * NAME: diAllocBit(imap,iagp,ino)
2004 * FUNCTION: allocate a backed inode from an iag.
2006 * this routine performs the mechanics of allocating a
2007 * specified inode from a backed extent.
2009 * if the inode to be allocated represents the last free
2010 * inode within the iag, the iag will be removed from the
2011 * ag free inode list.
2013 * a careful update approach is used to provide consistency
2014 * in the face of updates to multiple buffers. under this
2015 * approach, all required buffers are obtained before making
2016 * any updates and are held all are updates are complete.
2018 * PRE CONDITION: Already have buffer lock on iagp. Already have AG lock on
2019 * this AG. Must have read lock on imap inode.
2022 * imap - pointer to inode map control structure.
2023 * iagp - pointer to iag.
2024 * ino - inode number to be allocated within the iag.
2028 * -ENOSPC - insufficient disk resources.
2031 static int diAllocBit(struct inomap
* imap
, struct iag
* iagp
, int ino
)
2033 int extno
, bitno
, agno
, sword
, rc
;
2034 struct metapage
*amp
= NULL
, *bmp
= NULL
;
2035 struct iag
*aiagp
= NULL
, *biagp
= NULL
;
2038 /* check if this is the last free inode within the iag.
2039 * if so, it will have to be removed from the ag free
2040 * inode list, so get the iags preceeding and following
2043 if (iagp
->nfreeinos
== cpu_to_le32(1)) {
2044 if ((int) le32_to_cpu(iagp
->inofreefwd
) >= 0) {
2046 diIAGRead(imap
, le32_to_cpu(iagp
->inofreefwd
),
2049 aiagp
= (struct iag
*) amp
->data
;
2052 if ((int) le32_to_cpu(iagp
->inofreeback
) >= 0) {
2055 le32_to_cpu(iagp
->inofreeback
),
2058 release_metapage(amp
);
2061 biagp
= (struct iag
*) bmp
->data
;
2065 /* get the ag number, extent number, inode number within
2068 agno
= BLKTOAG(le64_to_cpu(iagp
->agstart
), JFS_SBI(imap
->im_ipimap
->i_sb
));
2069 extno
= ino
>> L2INOSPEREXT
;
2070 bitno
= ino
& (INOSPEREXT
- 1);
2072 /* compute the mask for setting the map.
2074 mask
= HIGHORDER
>> bitno
;
2076 /* the inode should be free and backed.
2078 if (((le32_to_cpu(iagp
->pmap
[extno
]) & mask
) != 0) ||
2079 ((le32_to_cpu(iagp
->wmap
[extno
]) & mask
) != 0) ||
2080 (addressPXD(&iagp
->inoext
[extno
]) == 0)) {
2082 release_metapage(amp
);
2084 release_metapage(bmp
);
2086 jfs_error(imap
->im_ipimap
->i_sb
,
2087 "diAllocBit: iag inconsistent");
2091 /* mark the inode as allocated in the working map.
2093 iagp
->wmap
[extno
] |= cpu_to_le32(mask
);
2095 /* check if all inodes within the extent are now
2096 * allocated. if so, update the free inode summary
2097 * map to reflect this.
2099 if (iagp
->wmap
[extno
] == cpu_to_le32(ONES
)) {
2100 sword
= extno
>> L2EXTSPERSUM
;
2101 bitno
= extno
& (EXTSPERSUM
- 1);
2102 iagp
->inosmap
[sword
] |= cpu_to_le32(HIGHORDER
>> bitno
);
2105 /* if this was the last free inode in the iag, remove the
2106 * iag from the ag free inode list.
2108 if (iagp
->nfreeinos
== cpu_to_le32(1)) {
2110 aiagp
->inofreeback
= iagp
->inofreeback
;
2111 write_metapage(amp
);
2115 biagp
->inofreefwd
= iagp
->inofreefwd
;
2116 write_metapage(bmp
);
2118 imap
->im_agctl
[agno
].inofree
=
2119 le32_to_cpu(iagp
->inofreefwd
);
2121 iagp
->inofreefwd
= iagp
->inofreeback
= cpu_to_le32(-1);
2124 /* update the free inode count at the iag, ag, inode
2127 iagp
->nfreeinos
= cpu_to_le32(le32_to_cpu(iagp
->nfreeinos
) - 1);
2128 imap
->im_agctl
[agno
].numfree
-= 1;
2129 atomic_dec(&imap
->im_numfree
);
2136 * NAME: diNewExt(imap,iagp,extno)
2138 * FUNCTION: initialize a new extent of inodes for an iag, allocating
2139 * the first inode of the extent for use for the current
2140 * allocation request.
2142 * disk resources are allocated for the new extent of inodes
2143 * and the inodes themselves are initialized to reflect their
2144 * existence within the extent (i.e. their inode numbers and
2145 * inode extent addresses are set) and their initial state
2146 * (mode and link count are set to zero).
2148 * if the iag is new, it is not yet on an ag extent free list
2149 * but will now be placed on this list.
2151 * if the allocation of the new extent causes the iag to
2152 * have no free extent, the iag will be removed from the
2153 * ag extent free list.
2155 * if the iag has no free backed inodes, it will be placed
2156 * on the ag free inode list, since the addition of the new
2157 * extent will now cause it to have free inodes.
2159 * a careful update approach is used to provide consistency
2160 * (i.e. list consistency) in the face of updates to multiple
2161 * buffers. under this approach, all required buffers are
2162 * obtained before making any updates and are held until all
2163 * updates are complete.
2165 * PRE CONDITION: Already have buffer lock on iagp. Already have AG lock on
2166 * this AG. Must have read lock on imap inode.
2169 * imap - pointer to inode map control structure.
2170 * iagp - pointer to iag.
2171 * extno - extent number.
2175 * -ENOSPC - insufficient disk resources.
2178 static int diNewExt(struct inomap
* imap
, struct iag
* iagp
, int extno
)
2180 int agno
, iagno
, fwd
, back
, freei
= 0, sword
, rc
;
2181 struct iag
*aiagp
= NULL
, *biagp
= NULL
, *ciagp
= NULL
;
2182 struct metapage
*amp
, *bmp
, *cmp
, *dmp
;
2183 struct inode
*ipimap
;
2189 struct jfs_sb_info
*sbi
;
2191 /* better have free extents.
2193 if (!iagp
->nfreeexts
) {
2194 jfs_error(imap
->im_ipimap
->i_sb
, "diNewExt: no free extents");
2198 /* get the inode map inode.
2200 ipimap
= imap
->im_ipimap
;
2201 sbi
= JFS_SBI(ipimap
->i_sb
);
2203 amp
= bmp
= cmp
= NULL
;
2205 /* get the ag and iag numbers for this iag.
2207 agno
= BLKTOAG(le64_to_cpu(iagp
->agstart
), sbi
);
2208 iagno
= le32_to_cpu(iagp
->iagnum
);
2210 /* check if this is the last free extent within the
2211 * iag. if so, the iag must be removed from the ag
2212 * free extent list, so get the iags preceeding and
2213 * following the iag on this list.
2215 if (iagp
->nfreeexts
== cpu_to_le32(1)) {
2216 if ((fwd
= le32_to_cpu(iagp
->extfreefwd
)) >= 0) {
2217 if ((rc
= diIAGRead(imap
, fwd
, &
)))
2219 aiagp
= (struct iag
*) amp
->data
;
2222 if ((back
= le32_to_cpu(iagp
->extfreeback
)) >= 0) {
2223 if ((rc
= diIAGRead(imap
, back
, &bmp
)))
2225 biagp
= (struct iag
*) bmp
->data
;
2228 /* the iag has free extents. if all extents are free
2229 * (as is the case for a newly allocated iag), the iag
2230 * must be added to the ag free extent list, so get
2231 * the iag at the head of the list in preparation for
2232 * adding this iag to this list.
2235 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
)) {
2236 if ((fwd
= imap
->im_agctl
[agno
].extfree
) >= 0) {
2237 if ((rc
= diIAGRead(imap
, fwd
, &
)))
2239 aiagp
= (struct iag
*) amp
->data
;
2244 /* check if the iag has no free inodes. if so, the iag
2245 * will have to be added to the ag free inode list, so get
2246 * the iag at the head of the list in preparation for
2247 * adding this iag to this list. in doing this, we must
2248 * check if we already have the iag at the head of
2251 if (iagp
->nfreeinos
== 0) {
2252 freei
= imap
->im_agctl
[agno
].inofree
;
2257 } else if (freei
== back
) {
2260 if ((rc
= diIAGRead(imap
, freei
, &cmp
)))
2262 ciagp
= (struct iag
*) cmp
->data
;
2264 if (ciagp
== NULL
) {
2265 jfs_error(imap
->im_ipimap
->i_sb
,
2266 "diNewExt: ciagp == NULL");
2273 /* allocate disk space for the inode extent.
2275 if ((extno
== 0) || (addressPXD(&iagp
->inoext
[extno
- 1]) == 0))
2276 hint
= ((s64
) agno
<< sbi
->bmap
->db_agl2size
) - 1;
2278 hint
= addressPXD(&iagp
->inoext
[extno
- 1]) +
2279 lengthPXD(&iagp
->inoext
[extno
- 1]) - 1;
2281 if ((rc
= dbAlloc(ipimap
, hint
, (s64
) imap
->im_nbperiext
, &blkno
)))
2284 /* compute the inode number of the first inode within the
2287 ino
= (iagno
<< L2INOSPERIAG
) + (extno
<< L2INOSPEREXT
);
2289 /* initialize the inodes within the newly allocated extent a
2292 for (i
= 0; i
< imap
->im_nbperiext
; i
+= sbi
->nbperpage
) {
2293 /* get a buffer for this page of disk inodes.
2295 dmp
= get_metapage(ipimap
, blkno
+ i
, PSIZE
, 1);
2300 dp
= (struct dinode
*) dmp
->data
;
2302 /* initialize the inode number, mode, link count and
2303 * inode extent address.
2305 for (j
= 0; j
< INOSPERPAGE
; j
++, dp
++, ino
++) {
2306 dp
->di_inostamp
= cpu_to_le32(sbi
->inostamp
);
2307 dp
->di_number
= cpu_to_le32(ino
);
2308 dp
->di_fileset
= cpu_to_le32(FILESYSTEM_I
);
2311 PXDaddress(&(dp
->di_ixpxd
), blkno
);
2312 PXDlength(&(dp
->di_ixpxd
), imap
->im_nbperiext
);
2314 write_metapage(dmp
);
2317 /* if this is the last free extent within the iag, remove the
2318 * iag from the ag free extent list.
2320 if (iagp
->nfreeexts
== cpu_to_le32(1)) {
2322 aiagp
->extfreeback
= iagp
->extfreeback
;
2325 biagp
->extfreefwd
= iagp
->extfreefwd
;
2327 imap
->im_agctl
[agno
].extfree
=
2328 le32_to_cpu(iagp
->extfreefwd
);
2330 iagp
->extfreefwd
= iagp
->extfreeback
= cpu_to_le32(-1);
2332 /* if the iag has all free extents (newly allocated iag),
2333 * add the iag to the ag free extent list.
2335 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
)) {
2337 aiagp
->extfreeback
= cpu_to_le32(iagno
);
2339 iagp
->extfreefwd
= cpu_to_le32(fwd
);
2340 iagp
->extfreeback
= cpu_to_le32(-1);
2341 imap
->im_agctl
[agno
].extfree
= iagno
;
2345 /* if the iag has no free inodes, add the iag to the
2346 * ag free inode list.
2348 if (iagp
->nfreeinos
== 0) {
2350 ciagp
->inofreeback
= cpu_to_le32(iagno
);
2353 cpu_to_le32(imap
->im_agctl
[agno
].inofree
);
2354 iagp
->inofreeback
= cpu_to_le32(-1);
2355 imap
->im_agctl
[agno
].inofree
= iagno
;
2358 /* initialize the extent descriptor of the extent. */
2359 PXDlength(&iagp
->inoext
[extno
], imap
->im_nbperiext
);
2360 PXDaddress(&iagp
->inoext
[extno
], blkno
);
2362 /* initialize the working and persistent map of the extent.
2363 * the working map will be initialized such that
2364 * it indicates the first inode of the extent is allocated.
2366 iagp
->wmap
[extno
] = cpu_to_le32(HIGHORDER
);
2367 iagp
->pmap
[extno
] = 0;
2369 /* update the free inode and free extent summary maps
2370 * for the extent to indicate the extent has free inodes
2371 * and no longer represents a free extent.
2373 sword
= extno
>> L2EXTSPERSUM
;
2374 mask
= HIGHORDER
>> (extno
& (EXTSPERSUM
- 1));
2375 iagp
->extsmap
[sword
] |= cpu_to_le32(mask
);
2376 iagp
->inosmap
[sword
] &= cpu_to_le32(~mask
);
2378 /* update the free inode and free extent counts for the
2381 iagp
->nfreeinos
= cpu_to_le32(le32_to_cpu(iagp
->nfreeinos
) +
2383 iagp
->nfreeexts
= cpu_to_le32(le32_to_cpu(iagp
->nfreeexts
) - 1);
2385 /* update the free and backed inode counts for the ag.
2387 imap
->im_agctl
[agno
].numfree
+= (INOSPEREXT
- 1);
2388 imap
->im_agctl
[agno
].numinos
+= INOSPEREXT
;
2390 /* update the free and backed inode counts for the inode map.
2392 atomic_add(INOSPEREXT
- 1, &imap
->im_numfree
);
2393 atomic_add(INOSPEREXT
, &imap
->im_numinos
);
2398 write_metapage(amp
);
2400 write_metapage(bmp
);
2402 write_metapage(cmp
);
2408 /* release the iags.
2411 release_metapage(amp
);
2413 release_metapage(bmp
);
2415 release_metapage(cmp
);
2422 * NAME: diNewIAG(imap,iagnop,agno)
2424 * FUNCTION: allocate a new iag for an allocation group.
2426 * first tries to allocate the iag from the inode map
2428 * if the list has free iags, the head of the list is removed
2429 * and returned to satisfy the request.
2430 * if the inode map's iag free list is empty, the inode map
2431 * is extended to hold a new iag. this new iag is initialized
2432 * and returned to satisfy the request.
2435 * imap - pointer to inode map control structure.
2436 * iagnop - pointer to an iag number set with the number of the
2437 * newly allocated iag upon successful return.
2438 * agno - allocation group number.
2439 * bpp - Buffer pointer to be filled in with new IAG's buffer
2443 * -ENOSPC - insufficient disk resources.
2447 * AG lock held on entry/exit;
2448 * write lock on the map is held inside;
2449 * read lock on the map is held on successful completion;
2451 * note: new iag transaction:
2452 * . synchronously write iag;
2453 * . write log of xtree and inode of imap;
2455 * . synchronous write of xtree (right to left, bottom to top);
2456 * . at start of logredo(): init in-memory imap with one additional iag page;
2457 * . at end of logredo(): re-read imap inode to determine
2461 diNewIAG(struct inomap
* imap
, int *iagnop
, int agno
, struct metapage
** mpp
)
2465 struct inode
*ipimap
;
2466 struct super_block
*sb
;
2467 struct jfs_sb_info
*sbi
;
2468 struct metapage
*mp
;
2473 struct inode
*iplist
[1];
2475 /* pick up pointers to the inode map and mount inodes */
2476 ipimap
= imap
->im_ipimap
;
2480 /* acquire the free iag lock */
2483 /* if there are any iags on the inode map free iag list,
2484 * allocate the iag from the head of the list.
2486 if (imap
->im_freeiag
>= 0) {
2487 /* pick up the iag number at the head of the list */
2488 iagno
= imap
->im_freeiag
;
2490 /* determine the logical block number of the iag */
2491 blkno
= IAGTOLBLK(iagno
, sbi
->l2nbperpage
);
2493 /* no free iags. the inode map will have to be extented
2494 * to include a new iag.
2497 /* acquire inode map lock */
2498 IWRITE_LOCK(ipimap
, RDWRLOCK_IMAP
);
2500 if (ipimap
->i_size
>> L2PSIZE
!= imap
->im_nextiag
+ 1) {
2501 IWRITE_UNLOCK(ipimap
);
2502 IAGFREE_UNLOCK(imap
);
2503 jfs_error(imap
->im_ipimap
->i_sb
,
2504 "diNewIAG: ipimap->i_size is wrong");
2509 /* get the next avaliable iag number */
2510 iagno
= imap
->im_nextiag
;
2512 /* make sure that we have not exceeded the maximum inode
2515 if (iagno
> (MAXIAGS
- 1)) {
2516 /* release the inode map lock */
2517 IWRITE_UNLOCK(ipimap
);
2524 * synchronously append new iag page.
2526 /* determine the logical address of iag page to append */
2527 blkno
= IAGTOLBLK(iagno
, sbi
->l2nbperpage
);
2529 /* Allocate extent for new iag page */
2530 xlen
= sbi
->nbperpage
;
2531 if ((rc
= dbAlloc(ipimap
, 0, (s64
) xlen
, &xaddr
))) {
2532 /* release the inode map lock */
2533 IWRITE_UNLOCK(ipimap
);
2539 * start transaction of update of the inode map
2540 * addressing structure pointing to the new iag page;
2542 tid
= txBegin(sb
, COMMIT_FORCE
);
2543 mutex_lock(&JFS_IP(ipimap
)->commit_mutex
);
2545 /* update the inode map addressing structure to point to it */
2547 xtInsert(tid
, ipimap
, 0, blkno
, xlen
, &xaddr
, 0))) {
2549 mutex_unlock(&JFS_IP(ipimap
)->commit_mutex
);
2550 /* Free the blocks allocated for the iag since it was
2551 * not successfully added to the inode map
2553 dbFree(ipimap
, xaddr
, (s64
) xlen
);
2555 /* release the inode map lock */
2556 IWRITE_UNLOCK(ipimap
);
2561 /* update the inode map's inode to reflect the extension */
2562 ipimap
->i_size
+= PSIZE
;
2563 inode_add_bytes(ipimap
, PSIZE
);
2565 /* assign a buffer for the page */
2566 mp
= get_metapage(ipimap
, blkno
, PSIZE
, 0);
2569 * This is very unlikely since we just created the
2570 * extent, but let's try to handle it correctly
2572 xtTruncate(tid
, ipimap
, ipimap
->i_size
- PSIZE
,
2578 /* release the inode map lock */
2579 IWRITE_UNLOCK(ipimap
);
2584 iagp
= (struct iag
*) mp
->data
;
2587 memset(iagp
, 0, sizeof(struct iag
));
2588 iagp
->iagnum
= cpu_to_le32(iagno
);
2589 iagp
->inofreefwd
= iagp
->inofreeback
= cpu_to_le32(-1);
2590 iagp
->extfreefwd
= iagp
->extfreeback
= cpu_to_le32(-1);
2591 iagp
->iagfree
= cpu_to_le32(-1);
2592 iagp
->nfreeinos
= 0;
2593 iagp
->nfreeexts
= cpu_to_le32(EXTSPERIAG
);
2595 /* initialize the free inode summary map (free extent
2596 * summary map initialization handled by bzero).
2598 for (i
= 0; i
< SMAPSZ
; i
++)
2599 iagp
->inosmap
[i
] = cpu_to_le32(ONES
);
2602 * Write and sync the metapage
2607 * txCommit(COMMIT_FORCE) will synchronously write address
2608 * index pages and inode after commit in careful update order
2609 * of address index pages (right to left, bottom up);
2612 rc
= txCommit(tid
, 1, &iplist
[0], COMMIT_FORCE
);
2615 mutex_unlock(&JFS_IP(ipimap
)->commit_mutex
);
2617 duplicateIXtree(sb
, blkno
, xlen
, &xaddr
);
2619 /* update the next avaliable iag number */
2620 imap
->im_nextiag
+= 1;
2622 /* Add the iag to the iag free list so we don't lose the iag
2623 * if a failure happens now.
2625 imap
->im_freeiag
= iagno
;
2627 /* Until we have logredo working, we want the imap inode &
2628 * control page to be up to date.
2632 /* release the inode map lock */
2633 IWRITE_UNLOCK(ipimap
);
2636 /* obtain read lock on map */
2637 IREAD_LOCK(ipimap
, RDWRLOCK_IMAP
);
2640 if ((rc
= diIAGRead(imap
, iagno
, &mp
))) {
2641 IREAD_UNLOCK(ipimap
);
2645 iagp
= (struct iag
*) mp
->data
;
2647 /* remove the iag from the iag free list */
2648 imap
->im_freeiag
= le32_to_cpu(iagp
->iagfree
);
2649 iagp
->iagfree
= cpu_to_le32(-1);
2651 /* set the return iag number and buffer pointer */
2656 /* release the iag free lock */
2657 IAGFREE_UNLOCK(imap
);
2665 * FUNCTION: get the buffer for the specified iag within a fileset
2666 * or aggregate inode map.
2669 * imap - pointer to inode map control structure.
2670 * iagno - iag number.
2671 * bpp - point to buffer pointer to be filled in on successful
2675 * must have read lock on imap inode
2676 * (When called by diExtendFS, the filesystem is quiesced, therefore
2677 * the read lock is unnecessary.)
2683 static int diIAGRead(struct inomap
* imap
, int iagno
, struct metapage
** mpp
)
2685 struct inode
*ipimap
= imap
->im_ipimap
;
2688 /* compute the logical block number of the iag. */
2689 blkno
= IAGTOLBLK(iagno
, JFS_SBI(ipimap
->i_sb
)->l2nbperpage
);
2692 *mpp
= read_metapage(ipimap
, blkno
, PSIZE
, 0);
2701 * NAME: diFindFree()
2703 * FUNCTION: find the first free bit in a word starting at
2704 * the specified bit position.
2707 * word - word to be examined.
2708 * start - starting bit position.
2711 * bit position of first free bit in the word or 32 if
2712 * no free bits were found.
2714 static int diFindFree(u32 word
, int start
)
2718 /* scan the word for the first free bit. */
2719 for (word
<<= start
, bitno
= start
; bitno
< 32;
2720 bitno
++, word
<<= 1) {
2721 if ((word
& HIGHORDER
) == 0)
2728 * NAME: diUpdatePMap()
2730 * FUNCTION: Update the persistent map in an IAG for the allocation or
2731 * freeing of the specified inode.
2733 * PRE CONDITIONS: Working map has already been updated for allocate.
2736 * ipimap - Incore inode map inode
2737 * inum - Number of inode to mark in permanent map
2738 * is_free - If 'true' indicates inode should be marked freed, otherwise
2739 * indicates inode should be marked allocated.
2745 diUpdatePMap(struct inode
*ipimap
,
2746 unsigned long inum
, bool is_free
, struct tblock
* tblk
)
2750 struct metapage
*mp
;
2751 int iagno
, ino
, extno
, bitno
;
2752 struct inomap
*imap
;
2754 struct jfs_log
*log
;
2755 int lsn
, difft
, diffp
;
2756 unsigned long flags
;
2758 imap
= JFS_IP(ipimap
)->i_imap
;
2759 /* get the iag number containing the inode */
2760 iagno
= INOTOIAG(inum
);
2761 /* make sure that the iag is contained within the map */
2762 if (iagno
>= imap
->im_nextiag
) {
2763 jfs_error(ipimap
->i_sb
,
2764 "diUpdatePMap: the iag is outside the map");
2768 IREAD_LOCK(ipimap
, RDWRLOCK_IMAP
);
2769 rc
= diIAGRead(imap
, iagno
, &mp
);
2770 IREAD_UNLOCK(ipimap
);
2773 metapage_wait_for_io(mp
);
2774 iagp
= (struct iag
*) mp
->data
;
2775 /* get the inode number and extent number of the inode within
2776 * the iag and the inode number within the extent.
2778 ino
= inum
& (INOSPERIAG
- 1);
2779 extno
= ino
>> L2INOSPEREXT
;
2780 bitno
= ino
& (INOSPEREXT
- 1);
2781 mask
= HIGHORDER
>> bitno
;
2783 * mark the inode free in persistent map:
2786 /* The inode should have been allocated both in working
2787 * map and in persistent map;
2788 * the inode will be freed from working map at the release
2789 * of last reference release;
2791 if (!(le32_to_cpu(iagp
->wmap
[extno
]) & mask
)) {
2792 jfs_error(ipimap
->i_sb
,
2793 "diUpdatePMap: inode %ld not marked as "
2794 "allocated in wmap!", inum
);
2796 if (!(le32_to_cpu(iagp
->pmap
[extno
]) & mask
)) {
2797 jfs_error(ipimap
->i_sb
,
2798 "diUpdatePMap: inode %ld not marked as "
2799 "allocated in pmap!", inum
);
2801 /* update the bitmap for the extent of the freed inode */
2802 iagp
->pmap
[extno
] &= cpu_to_le32(~mask
);
2805 * mark the inode allocated in persistent map:
2808 /* The inode should be already allocated in the working map
2809 * and should be free in persistent map;
2811 if (!(le32_to_cpu(iagp
->wmap
[extno
]) & mask
)) {
2812 release_metapage(mp
);
2813 jfs_error(ipimap
->i_sb
,
2814 "diUpdatePMap: the inode is not allocated in "
2818 if ((le32_to_cpu(iagp
->pmap
[extno
]) & mask
) != 0) {
2819 release_metapage(mp
);
2820 jfs_error(ipimap
->i_sb
,
2821 "diUpdatePMap: the inode is not free in the "
2825 /* update the bitmap for the extent of the allocated inode */
2826 iagp
->pmap
[extno
] |= cpu_to_le32(mask
);
2832 log
= JFS_SBI(tblk
->sb
)->log
;
2833 LOGSYNC_LOCK(log
, flags
);
2835 /* inherit older/smaller lsn */
2836 logdiff(difft
, lsn
, log
);
2837 logdiff(diffp
, mp
->lsn
, log
);
2838 if (difft
< diffp
) {
2840 /* move mp after tblock in logsync list */
2841 list_move(&mp
->synclist
, &tblk
->synclist
);
2843 /* inherit younger/larger clsn */
2845 logdiff(difft
, tblk
->clsn
, log
);
2846 logdiff(diffp
, mp
->clsn
, log
);
2848 mp
->clsn
= tblk
->clsn
;
2852 /* insert mp after tblock in logsync list */
2854 list_add(&mp
->synclist
, &tblk
->synclist
);
2855 mp
->clsn
= tblk
->clsn
;
2857 LOGSYNC_UNLOCK(log
, flags
);
2865 * function: update imap for extendfs();
2867 * note: AG size has been increased s.t. each k old contiguous AGs are
2868 * coalesced into a new AG;
2870 int diExtendFS(struct inode
*ipimap
, struct inode
*ipbmap
)
2873 struct inomap
*imap
= JFS_IP(ipimap
)->i_imap
;
2874 struct iag
*iagp
= NULL
, *hiagp
= NULL
;
2875 struct bmap
*mp
= JFS_SBI(ipbmap
->i_sb
)->bmap
;
2876 struct metapage
*bp
, *hbp
;
2878 int numinos
, xnuminos
= 0, xnumfree
= 0;
2881 jfs_info("diExtendFS: nextiag:%d numinos:%d numfree:%d",
2882 imap
->im_nextiag
, atomic_read(&imap
->im_numinos
),
2883 atomic_read(&imap
->im_numfree
));
2888 * coalesce contiguous k (newAGSize/oldAGSize) AGs;
2889 * i.e., (AGi, ..., AGj) where i = k*n and j = k*(n+1) - 1 to AGn;
2890 * note: new AG size = old AG size * (2**x).
2893 /* init per AG control information im_agctl[] */
2894 for (i
= 0; i
< MAXAG
; i
++) {
2895 imap
->im_agctl
[i
].inofree
= -1;
2896 imap
->im_agctl
[i
].extfree
= -1;
2897 imap
->im_agctl
[i
].numinos
= 0; /* number of backed inodes */
2898 imap
->im_agctl
[i
].numfree
= 0; /* number of free backed inodes */
2902 * process each iag page of the map.
2904 * rebuild AG Free Inode List, AG Free Inode Extent List;
2906 for (i
= 0; i
< imap
->im_nextiag
; i
++) {
2907 if ((rc
= diIAGRead(imap
, i
, &bp
))) {
2911 iagp
= (struct iag
*) bp
->data
;
2912 if (le32_to_cpu(iagp
->iagnum
) != i
) {
2913 release_metapage(bp
);
2914 jfs_error(ipimap
->i_sb
,
2915 "diExtendFs: unexpected value of iagnum");
2919 /* leave free iag in the free iag list */
2920 if (iagp
->nfreeexts
== cpu_to_le32(EXTSPERIAG
)) {
2921 release_metapage(bp
);
2925 /* agstart that computes to the same ag is treated as same; */
2926 agstart
= le64_to_cpu(iagp
->agstart
);
2927 /* iagp->agstart = agstart & ~(mp->db_agsize - 1); */
2928 n
= agstart
>> mp
->db_agl2size
;
2930 /* compute backed inodes */
2931 numinos
= (EXTSPERIAG
- le32_to_cpu(iagp
->nfreeexts
))
2934 /* merge AG backed inodes */
2935 imap
->im_agctl
[n
].numinos
+= numinos
;
2936 xnuminos
+= numinos
;
2939 /* if any backed free inodes, insert at AG free inode list */
2940 if ((int) le32_to_cpu(iagp
->nfreeinos
) > 0) {
2941 if ((head
= imap
->im_agctl
[n
].inofree
) == -1) {
2942 iagp
->inofreefwd
= cpu_to_le32(-1);
2943 iagp
->inofreeback
= cpu_to_le32(-1);
2945 if ((rc
= diIAGRead(imap
, head
, &hbp
))) {
2949 hiagp
= (struct iag
*) hbp
->data
;
2950 hiagp
->inofreeback
= iagp
->iagnum
;
2951 iagp
->inofreefwd
= cpu_to_le32(head
);
2952 iagp
->inofreeback
= cpu_to_le32(-1);
2953 write_metapage(hbp
);
2956 imap
->im_agctl
[n
].inofree
=
2957 le32_to_cpu(iagp
->iagnum
);
2959 /* merge AG backed free inodes */
2960 imap
->im_agctl
[n
].numfree
+=
2961 le32_to_cpu(iagp
->nfreeinos
);
2962 xnumfree
+= le32_to_cpu(iagp
->nfreeinos
);
2965 /* if any free extents, insert at AG free extent list */
2966 if (le32_to_cpu(iagp
->nfreeexts
) > 0) {
2967 if ((head
= imap
->im_agctl
[n
].extfree
) == -1) {
2968 iagp
->extfreefwd
= cpu_to_le32(-1);
2969 iagp
->extfreeback
= cpu_to_le32(-1);
2971 if ((rc
= diIAGRead(imap
, head
, &hbp
))) {
2975 hiagp
= (struct iag
*) hbp
->data
;
2976 hiagp
->extfreeback
= iagp
->iagnum
;
2977 iagp
->extfreefwd
= cpu_to_le32(head
);
2978 iagp
->extfreeback
= cpu_to_le32(-1);
2979 write_metapage(hbp
);
2982 imap
->im_agctl
[n
].extfree
=
2983 le32_to_cpu(iagp
->iagnum
);
2990 if (xnuminos
!= atomic_read(&imap
->im_numinos
) ||
2991 xnumfree
!= atomic_read(&imap
->im_numfree
)) {
2992 jfs_error(ipimap
->i_sb
,
2993 "diExtendFs: numinos or numfree incorrect");
3004 * serialization: IWRITE_LOCK held on entry/exit
3006 * note: shadow page with regular inode (rel.2);
3008 static void duplicateIXtree(struct super_block
*sb
, s64 blkno
,
3009 int xlen
, s64
*xaddr
)
3011 struct jfs_superblock
*j_sb
;
3012 struct buffer_head
*bh
;
3016 /* if AIT2 ipmap2 is bad, do not try to update it */
3017 if (JFS_SBI(sb
)->mntflag
& JFS_BAD_SAIT
) /* s_flag */
3019 ip
= diReadSpecial(sb
, FILESYSTEM_I
, 1);
3021 JFS_SBI(sb
)->mntflag
|= JFS_BAD_SAIT
;
3022 if (readSuper(sb
, &bh
))
3024 j_sb
= (struct jfs_superblock
*)bh
->b_data
;
3025 j_sb
->s_flag
|= cpu_to_le32(JFS_BAD_SAIT
);
3027 mark_buffer_dirty(bh
);
3028 sync_dirty_buffer(bh
);
3033 /* start transaction */
3034 tid
= txBegin(sb
, COMMIT_FORCE
);
3035 /* update the inode map addressing structure to point to it */
3036 if (xtInsert(tid
, ip
, 0, blkno
, xlen
, xaddr
, 0)) {
3037 JFS_SBI(sb
)->mntflag
|= JFS_BAD_SAIT
;
3042 /* update the inode map's inode to reflect the extension */
3043 ip
->i_size
+= PSIZE
;
3044 inode_add_bytes(ip
, PSIZE
);
3045 txCommit(tid
, 1, &ip
, COMMIT_FORCE
);
3052 * NAME: copy_from_dinode()
3054 * FUNCTION: Copies inode info from disk inode to in-memory inode
3058 * -ENOMEM - insufficient memory
3060 static int copy_from_dinode(struct dinode
* dip
, struct inode
*ip
)
3062 struct jfs_inode_info
*jfs_ip
= JFS_IP(ip
);
3063 struct jfs_sb_info
*sbi
= JFS_SBI(ip
->i_sb
);
3065 jfs_ip
->fileset
= le32_to_cpu(dip
->di_fileset
);
3066 jfs_ip
->mode2
= le32_to_cpu(dip
->di_mode
);
3067 jfs_set_inode_flags(ip
);
3069 ip
->i_mode
= le32_to_cpu(dip
->di_mode
) & 0xffff;
3070 if (sbi
->umask
!= -1) {
3071 ip
->i_mode
= (ip
->i_mode
& ~0777) | (0777 & ~sbi
->umask
);
3072 /* For directories, add x permission if r is allowed by umask */
3073 if (S_ISDIR(ip
->i_mode
)) {
3074 if (ip
->i_mode
& 0400)
3076 if (ip
->i_mode
& 0040)
3078 if (ip
->i_mode
& 0004)
3082 ip
->i_nlink
= le32_to_cpu(dip
->di_nlink
);
3084 jfs_ip
->saved_uid
= le32_to_cpu(dip
->di_uid
);
3086 ip
->i_uid
= jfs_ip
->saved_uid
;
3088 ip
->i_uid
= sbi
->uid
;
3091 jfs_ip
->saved_gid
= le32_to_cpu(dip
->di_gid
);
3093 ip
->i_gid
= jfs_ip
->saved_gid
;
3095 ip
->i_gid
= sbi
->gid
;
3098 ip
->i_size
= le64_to_cpu(dip
->di_size
);
3099 ip
->i_atime
.tv_sec
= le32_to_cpu(dip
->di_atime
.tv_sec
);
3100 ip
->i_atime
.tv_nsec
= le32_to_cpu(dip
->di_atime
.tv_nsec
);
3101 ip
->i_mtime
.tv_sec
= le32_to_cpu(dip
->di_mtime
.tv_sec
);
3102 ip
->i_mtime
.tv_nsec
= le32_to_cpu(dip
->di_mtime
.tv_nsec
);
3103 ip
->i_ctime
.tv_sec
= le32_to_cpu(dip
->di_ctime
.tv_sec
);
3104 ip
->i_ctime
.tv_nsec
= le32_to_cpu(dip
->di_ctime
.tv_nsec
);
3105 ip
->i_blocks
= LBLK2PBLK(ip
->i_sb
, le64_to_cpu(dip
->di_nblocks
));
3106 ip
->i_generation
= le32_to_cpu(dip
->di_gen
);
3108 jfs_ip
->ixpxd
= dip
->di_ixpxd
; /* in-memory pxd's are little-endian */
3109 jfs_ip
->acl
= dip
->di_acl
; /* as are dxd's */
3110 jfs_ip
->ea
= dip
->di_ea
;
3111 jfs_ip
->next_index
= le32_to_cpu(dip
->di_next_index
);
3112 jfs_ip
->otime
= le32_to_cpu(dip
->di_otime
.tv_sec
);
3113 jfs_ip
->acltype
= le32_to_cpu(dip
->di_acltype
);
3115 if (S_ISCHR(ip
->i_mode
) || S_ISBLK(ip
->i_mode
)) {
3116 jfs_ip
->dev
= le32_to_cpu(dip
->di_rdev
);
3117 ip
->i_rdev
= new_decode_dev(jfs_ip
->dev
);
3120 if (S_ISDIR(ip
->i_mode
)) {
3121 memcpy(&jfs_ip
->i_dirtable
, &dip
->di_dirtable
, 384);
3122 } else if (S_ISREG(ip
->i_mode
) || S_ISLNK(ip
->i_mode
)) {
3123 memcpy(&jfs_ip
->i_xtroot
, &dip
->di_xtroot
, 288);
3125 memcpy(&jfs_ip
->i_inline_ea
, &dip
->di_inlineea
, 128);
3127 /* Zero the in-memory-only stuff */
3129 jfs_ip
->btindex
= 0;
3130 jfs_ip
->btorder
= 0;
3133 jfs_ip
->atlhead
= 0;
3134 jfs_ip
->atltail
= 0;
3140 * NAME: copy_to_dinode()
3142 * FUNCTION: Copies inode info from in-memory inode to disk inode
3144 static void copy_to_dinode(struct dinode
* dip
, struct inode
*ip
)
3146 struct jfs_inode_info
*jfs_ip
= JFS_IP(ip
);
3147 struct jfs_sb_info
*sbi
= JFS_SBI(ip
->i_sb
);
3149 dip
->di_fileset
= cpu_to_le32(jfs_ip
->fileset
);
3150 dip
->di_inostamp
= cpu_to_le32(sbi
->inostamp
);
3151 dip
->di_number
= cpu_to_le32(ip
->i_ino
);
3152 dip
->di_gen
= cpu_to_le32(ip
->i_generation
);
3153 dip
->di_size
= cpu_to_le64(ip
->i_size
);
3154 dip
->di_nblocks
= cpu_to_le64(PBLK2LBLK(ip
->i_sb
, ip
->i_blocks
));
3155 dip
->di_nlink
= cpu_to_le32(ip
->i_nlink
);
3157 dip
->di_uid
= cpu_to_le32(ip
->i_uid
);
3159 dip
->di_uid
= cpu_to_le32(jfs_ip
->saved_uid
);
3161 dip
->di_gid
= cpu_to_le32(ip
->i_gid
);
3163 dip
->di_gid
= cpu_to_le32(jfs_ip
->saved_gid
);
3164 jfs_get_inode_flags(jfs_ip
);
3166 * mode2 is only needed for storing the higher order bits.
3167 * Trust i_mode for the lower order ones
3169 if (sbi
->umask
== -1)
3170 dip
->di_mode
= cpu_to_le32((jfs_ip
->mode2
& 0xffff0000) |
3172 else /* Leave the original permissions alone */
3173 dip
->di_mode
= cpu_to_le32(jfs_ip
->mode2
);
3175 dip
->di_atime
.tv_sec
= cpu_to_le32(ip
->i_atime
.tv_sec
);
3176 dip
->di_atime
.tv_nsec
= cpu_to_le32(ip
->i_atime
.tv_nsec
);
3177 dip
->di_ctime
.tv_sec
= cpu_to_le32(ip
->i_ctime
.tv_sec
);
3178 dip
->di_ctime
.tv_nsec
= cpu_to_le32(ip
->i_ctime
.tv_nsec
);
3179 dip
->di_mtime
.tv_sec
= cpu_to_le32(ip
->i_mtime
.tv_sec
);
3180 dip
->di_mtime
.tv_nsec
= cpu_to_le32(ip
->i_mtime
.tv_nsec
);
3181 dip
->di_ixpxd
= jfs_ip
->ixpxd
; /* in-memory pxd's are little-endian */
3182 dip
->di_acl
= jfs_ip
->acl
; /* as are dxd's */
3183 dip
->di_ea
= jfs_ip
->ea
;
3184 dip
->di_next_index
= cpu_to_le32(jfs_ip
->next_index
);
3185 dip
->di_otime
.tv_sec
= cpu_to_le32(jfs_ip
->otime
);
3186 dip
->di_otime
.tv_nsec
= 0;
3187 dip
->di_acltype
= cpu_to_le32(jfs_ip
->acltype
);
3188 if (S_ISCHR(ip
->i_mode
) || S_ISBLK(ip
->i_mode
))
3189 dip
->di_rdev
= cpu_to_le32(jfs_ip
->dev
);
This page took 0.09059 seconds and 6 git commands to generate.