2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
5 * This copyrighted material is made available to anyone wishing to use,
6 * modify, copy, or redistribute it subject to the terms and conditions
7 * of the GNU General Public License version 2.
10 #include <linux/slab.h>
11 #include <linux/spinlock.h>
12 #include <linux/completion.h>
13 #include <linux/buffer_head.h>
15 #include <linux/gfs2_ondisk.h>
16 #include <linux/prefetch.h>
17 #include <linux/blkdev.h>
18 #include <linux/rbtree.h>
33 #include "trace_gfs2.h"
35 #define BFITNOENT ((u32)~0)
36 #define NO_BLOCK ((u64)~0)
38 #if BITS_PER_LONG == 32
39 #define LBITMASK (0x55555555UL)
40 #define LBITSKIP55 (0x55555555UL)
41 #define LBITSKIP00 (0x00000000UL)
43 #define LBITMASK (0x5555555555555555UL)
44 #define LBITSKIP55 (0x5555555555555555UL)
45 #define LBITSKIP00 (0x0000000000000000UL)
49 * These routines are used by the resource group routines (rgrp.c)
50 * to keep track of block allocation. Each block is represented by two
51 * bits. So, each byte represents GFS2_NBBY (i.e. 4) blocks.
54 * 1 = Used (not metadata)
55 * 2 = Unlinked (still in use) inode
59 static const char valid_change
[16] = {
67 static u32
rgblk_search(struct gfs2_rgrpd
*rgd
, u32 goal
,
68 unsigned char old_state
,
69 struct gfs2_bitmap
**rbi
);
72 * gfs2_setbit - Set a bit in the bitmaps
73 * @buffer: the buffer that holds the bitmaps
74 * @buflen: the length (in bytes) of the buffer
75 * @block: the block to set
76 * @new_state: the new state of the block
80 static inline void gfs2_setbit(struct gfs2_rgrpd
*rgd
, unsigned char *buf1
,
81 unsigned char *buf2
, unsigned int offset
,
82 struct gfs2_bitmap
*bi
, u32 block
,
83 unsigned char new_state
)
85 unsigned char *byte1
, *byte2
, *end
, cur_state
;
86 unsigned int buflen
= bi
->bi_len
;
87 const unsigned int bit
= (block
% GFS2_NBBY
) * GFS2_BIT_SIZE
;
89 byte1
= buf1
+ offset
+ (block
/ GFS2_NBBY
);
90 end
= buf1
+ offset
+ buflen
;
94 cur_state
= (*byte1
>> bit
) & GFS2_BIT_MASK
;
96 if (unlikely(!valid_change
[new_state
* 4 + cur_state
])) {
97 printk(KERN_WARNING
"GFS2: buf_blk = 0x%llx old_state=%d, "
99 (unsigned long long)block
, cur_state
, new_state
);
100 printk(KERN_WARNING
"GFS2: rgrp=0x%llx bi_start=0x%lx\n",
101 (unsigned long long)rgd
->rd_addr
,
102 (unsigned long)bi
->bi_start
);
103 printk(KERN_WARNING
"GFS2: bi_offset=0x%lx bi_len=0x%lx\n",
104 (unsigned long)bi
->bi_offset
,
105 (unsigned long)bi
->bi_len
);
107 gfs2_consist_rgrpd(rgd
);
110 *byte1
^= (cur_state
^ new_state
) << bit
;
113 byte2
= buf2
+ offset
+ (block
/ GFS2_NBBY
);
114 cur_state
= (*byte2
>> bit
) & GFS2_BIT_MASK
;
115 *byte2
^= (cur_state
^ new_state
) << bit
;
120 * gfs2_testbit - test a bit in the bitmaps
121 * @buffer: the buffer that holds the bitmaps
122 * @buflen: the length (in bytes) of the buffer
123 * @block: the block to read
127 static inline unsigned char gfs2_testbit(struct gfs2_rgrpd
*rgd
,
128 const unsigned char *buffer
,
129 unsigned int buflen
, u32 block
)
131 const unsigned char *byte
, *end
;
132 unsigned char cur_state
;
135 byte
= buffer
+ (block
/ GFS2_NBBY
);
136 bit
= (block
% GFS2_NBBY
) * GFS2_BIT_SIZE
;
137 end
= buffer
+ buflen
;
139 gfs2_assert(rgd
->rd_sbd
, byte
< end
);
141 cur_state
= (*byte
>> bit
) & GFS2_BIT_MASK
;
148 * @ptr: Pointer to bitmap data
149 * @mask: Mask to use (normally 0x55555.... but adjusted for search start)
150 * @state: The state we are searching for
152 * We xor the bitmap data with a patter which is the bitwise opposite
153 * of what we are looking for, this gives rise to a pattern of ones
154 * wherever there is a match. Since we have two bits per entry, we
155 * take this pattern, shift it down by one place and then and it with
156 * the original. All the even bit positions (0,2,4, etc) then represent
157 * successful matches, so we mask with 0x55555..... to remove the unwanted
160 * This allows searching of a whole u64 at once (32 blocks) with a
161 * single test (on 64 bit arches).
164 static inline u64
gfs2_bit_search(const __le64
*ptr
, u64 mask
, u8 state
)
167 static const u64 search
[] = {
168 [0] = 0xffffffffffffffffULL
,
169 [1] = 0xaaaaaaaaaaaaaaaaULL
,
170 [2] = 0x5555555555555555ULL
,
171 [3] = 0x0000000000000000ULL
,
173 tmp
= le64_to_cpu(*ptr
) ^ search
[state
];
180 * gfs2_bitfit - Search an rgrp's bitmap buffer to find a bit-pair representing
181 * a block in a given allocation state.
182 * @buffer: the buffer that holds the bitmaps
183 * @len: the length (in bytes) of the buffer
184 * @goal: start search at this block's bit-pair (within @buffer)
185 * @state: GFS2_BLKST_XXX the state of the block we're looking for.
187 * Scope of @goal and returned block number is only within this bitmap buffer,
188 * not entire rgrp or filesystem. @buffer will be offset from the actual
189 * beginning of a bitmap block buffer, skipping any header structures, but
190 * headers are always a multiple of 64 bits long so that the buffer is
191 * always aligned to a 64 bit boundary.
193 * The size of the buffer is in bytes, but is it assumed that it is
194 * always ok to read a complete multiple of 64 bits at the end
195 * of the block in case the end is no aligned to a natural boundary.
197 * Return: the block number (bitmap buffer scope) that was found
200 static u32
gfs2_bitfit(const u8
*buf
, const unsigned int len
,
203 u32 spoint
= (goal
<< 1) & ((8*sizeof(u64
)) - 1);
204 const __le64
*ptr
= ((__le64
*)buf
) + (goal
>> 5);
205 const __le64
*end
= (__le64
*)(buf
+ ALIGN(len
, sizeof(u64
)));
207 u64 mask
= 0x5555555555555555ULL
;
212 /* Mask off bits we don't care about at the start of the search */
214 tmp
= gfs2_bit_search(ptr
, mask
, state
);
216 while(tmp
== 0 && ptr
< end
) {
217 tmp
= gfs2_bit_search(ptr
, 0x5555555555555555ULL
, state
);
220 /* Mask off any bits which are more than len bytes from the start */
221 if (ptr
== end
&& (len
& (sizeof(u64
) - 1)))
222 tmp
&= (((u64
)~0) >> (64 - 8*(len
& (sizeof(u64
) - 1))));
223 /* Didn't find anything, so return */
228 bit
/= 2; /* two bits per entry in the bitmap */
229 return (((const unsigned char *)ptr
- buf
) * GFS2_NBBY
) + bit
;
233 * gfs2_bitcount - count the number of bits in a certain state
234 * @buffer: the buffer that holds the bitmaps
235 * @buflen: the length (in bytes) of the buffer
236 * @state: the state of the block we're looking for
238 * Returns: The number of bits
241 static u32
gfs2_bitcount(struct gfs2_rgrpd
*rgd
, const u8
*buffer
,
242 unsigned int buflen
, u8 state
)
244 const u8
*byte
= buffer
;
245 const u8
*end
= buffer
+ buflen
;
246 const u8 state1
= state
<< 2;
247 const u8 state2
= state
<< 4;
248 const u8 state3
= state
<< 6;
251 for (; byte
< end
; byte
++) {
252 if (((*byte
) & 0x03) == state
)
254 if (((*byte
) & 0x0C) == state1
)
256 if (((*byte
) & 0x30) == state2
)
258 if (((*byte
) & 0xC0) == state3
)
266 * gfs2_rgrp_verify - Verify that a resource group is consistent
267 * @sdp: the filesystem
272 void gfs2_rgrp_verify(struct gfs2_rgrpd
*rgd
)
274 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
275 struct gfs2_bitmap
*bi
= NULL
;
276 u32 length
= rgd
->rd_length
;
280 memset(count
, 0, 4 * sizeof(u32
));
282 /* Count # blocks in each of 4 possible allocation states */
283 for (buf
= 0; buf
< length
; buf
++) {
284 bi
= rgd
->rd_bits
+ buf
;
285 for (x
= 0; x
< 4; x
++)
286 count
[x
] += gfs2_bitcount(rgd
,
292 if (count
[0] != rgd
->rd_free
) {
293 if (gfs2_consist_rgrpd(rgd
))
294 fs_err(sdp
, "free data mismatch: %u != %u\n",
295 count
[0], rgd
->rd_free
);
299 tmp
= rgd
->rd_data
- rgd
->rd_free
- rgd
->rd_dinodes
;
300 if (count
[1] != tmp
) {
301 if (gfs2_consist_rgrpd(rgd
))
302 fs_err(sdp
, "used data mismatch: %u != %u\n",
307 if (count
[2] + count
[3] != rgd
->rd_dinodes
) {
308 if (gfs2_consist_rgrpd(rgd
))
309 fs_err(sdp
, "used metadata mismatch: %u != %u\n",
310 count
[2] + count
[3], rgd
->rd_dinodes
);
315 static inline int rgrp_contains_block(struct gfs2_rgrpd
*rgd
, u64 block
)
317 u64 first
= rgd
->rd_data0
;
318 u64 last
= first
+ rgd
->rd_data
;
319 return first
<= block
&& block
< last
;
323 * gfs2_blk2rgrpd - Find resource group for a given data/meta block number
324 * @sdp: The GFS2 superblock
325 * @n: The data block number
327 * Returns: The resource group, or NULL if not found
330 struct gfs2_rgrpd
*gfs2_blk2rgrpd(struct gfs2_sbd
*sdp
, u64 blk
, bool exact
)
332 struct rb_node
*n
, *next
;
333 struct gfs2_rgrpd
*cur
;
335 spin_lock(&sdp
->sd_rindex_spin
);
336 n
= sdp
->sd_rindex_tree
.rb_node
;
338 cur
= rb_entry(n
, struct gfs2_rgrpd
, rd_node
);
340 if (blk
< cur
->rd_addr
)
342 else if (blk
>= cur
->rd_data0
+ cur
->rd_data
)
345 spin_unlock(&sdp
->sd_rindex_spin
);
347 if (blk
< cur
->rd_addr
)
349 if (blk
>= cur
->rd_data0
+ cur
->rd_data
)
356 spin_unlock(&sdp
->sd_rindex_spin
);
362 * gfs2_rgrpd_get_first - get the first Resource Group in the filesystem
363 * @sdp: The GFS2 superblock
365 * Returns: The first rgrp in the filesystem
368 struct gfs2_rgrpd
*gfs2_rgrpd_get_first(struct gfs2_sbd
*sdp
)
370 const struct rb_node
*n
;
371 struct gfs2_rgrpd
*rgd
;
373 spin_lock(&sdp
->sd_rindex_spin
);
374 n
= rb_first(&sdp
->sd_rindex_tree
);
375 rgd
= rb_entry(n
, struct gfs2_rgrpd
, rd_node
);
376 spin_unlock(&sdp
->sd_rindex_spin
);
382 * gfs2_rgrpd_get_next - get the next RG
385 * Returns: The next rgrp
388 struct gfs2_rgrpd
*gfs2_rgrpd_get_next(struct gfs2_rgrpd
*rgd
)
390 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
391 const struct rb_node
*n
;
393 spin_lock(&sdp
->sd_rindex_spin
);
394 n
= rb_next(&rgd
->rd_node
);
396 n
= rb_first(&sdp
->sd_rindex_tree
);
398 if (unlikely(&rgd
->rd_node
== n
)) {
399 spin_unlock(&sdp
->sd_rindex_spin
);
402 rgd
= rb_entry(n
, struct gfs2_rgrpd
, rd_node
);
403 spin_unlock(&sdp
->sd_rindex_spin
);
407 void gfs2_free_clones(struct gfs2_rgrpd
*rgd
)
411 for (x
= 0; x
< rgd
->rd_length
; x
++) {
412 struct gfs2_bitmap
*bi
= rgd
->rd_bits
+ x
;
418 void gfs2_clear_rgrpd(struct gfs2_sbd
*sdp
)
421 struct gfs2_rgrpd
*rgd
;
422 struct gfs2_glock
*gl
;
424 while ((n
= rb_first(&sdp
->sd_rindex_tree
))) {
425 rgd
= rb_entry(n
, struct gfs2_rgrpd
, rd_node
);
428 rb_erase(n
, &sdp
->sd_rindex_tree
);
431 spin_lock(&gl
->gl_spin
);
432 gl
->gl_object
= NULL
;
433 spin_unlock(&gl
->gl_spin
);
434 gfs2_glock_add_to_lru(gl
);
438 gfs2_free_clones(rgd
);
440 kmem_cache_free(gfs2_rgrpd_cachep
, rgd
);
444 static void gfs2_rindex_print(const struct gfs2_rgrpd
*rgd
)
446 printk(KERN_INFO
" ri_addr = %llu\n", (unsigned long long)rgd
->rd_addr
);
447 printk(KERN_INFO
" ri_length = %u\n", rgd
->rd_length
);
448 printk(KERN_INFO
" ri_data0 = %llu\n", (unsigned long long)rgd
->rd_data0
);
449 printk(KERN_INFO
" ri_data = %u\n", rgd
->rd_data
);
450 printk(KERN_INFO
" ri_bitbytes = %u\n", rgd
->rd_bitbytes
);
454 * gfs2_compute_bitstructs - Compute the bitmap sizes
455 * @rgd: The resource group descriptor
457 * Calculates bitmap descriptors, one for each block that contains bitmap data
462 static int compute_bitstructs(struct gfs2_rgrpd
*rgd
)
464 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
465 struct gfs2_bitmap
*bi
;
466 u32 length
= rgd
->rd_length
; /* # blocks in hdr & bitmap */
467 u32 bytes_left
, bytes
;
473 rgd
->rd_bits
= kcalloc(length
, sizeof(struct gfs2_bitmap
), GFP_NOFS
);
477 bytes_left
= rgd
->rd_bitbytes
;
479 for (x
= 0; x
< length
; x
++) {
480 bi
= rgd
->rd_bits
+ x
;
483 /* small rgrp; bitmap stored completely in header block */
486 bi
->bi_offset
= sizeof(struct gfs2_rgrp
);
491 bytes
= sdp
->sd_sb
.sb_bsize
- sizeof(struct gfs2_rgrp
);
492 bi
->bi_offset
= sizeof(struct gfs2_rgrp
);
496 } else if (x
+ 1 == length
) {
498 bi
->bi_offset
= sizeof(struct gfs2_meta_header
);
499 bi
->bi_start
= rgd
->rd_bitbytes
- bytes_left
;
503 bytes
= sdp
->sd_sb
.sb_bsize
-
504 sizeof(struct gfs2_meta_header
);
505 bi
->bi_offset
= sizeof(struct gfs2_meta_header
);
506 bi
->bi_start
= rgd
->rd_bitbytes
- bytes_left
;
514 gfs2_consist_rgrpd(rgd
);
517 bi
= rgd
->rd_bits
+ (length
- 1);
518 if ((bi
->bi_start
+ bi
->bi_len
) * GFS2_NBBY
!= rgd
->rd_data
) {
519 if (gfs2_consist_rgrpd(rgd
)) {
520 gfs2_rindex_print(rgd
);
521 fs_err(sdp
, "start=%u len=%u offset=%u\n",
522 bi
->bi_start
, bi
->bi_len
, bi
->bi_offset
);
531 * gfs2_ri_total - Total up the file system space, according to the rindex.
534 u64
gfs2_ri_total(struct gfs2_sbd
*sdp
)
537 struct inode
*inode
= sdp
->sd_rindex
;
538 struct gfs2_inode
*ip
= GFS2_I(inode
);
539 char buf
[sizeof(struct gfs2_rindex
)];
540 struct file_ra_state ra_state
;
543 file_ra_state_init(&ra_state
, inode
->i_mapping
);
544 for (rgrps
= 0;; rgrps
++) {
545 loff_t pos
= rgrps
* sizeof(struct gfs2_rindex
);
547 if (pos
+ sizeof(struct gfs2_rindex
) > i_size_read(inode
))
549 error
= gfs2_internal_read(ip
, &ra_state
, buf
, &pos
,
550 sizeof(struct gfs2_rindex
));
551 if (error
!= sizeof(struct gfs2_rindex
))
553 total_data
+= be32_to_cpu(((struct gfs2_rindex
*)buf
)->ri_data
);
558 static int rgd_insert(struct gfs2_rgrpd
*rgd
)
560 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
561 struct rb_node
**newn
= &sdp
->sd_rindex_tree
.rb_node
, *parent
= NULL
;
563 /* Figure out where to put new node */
565 struct gfs2_rgrpd
*cur
= rb_entry(*newn
, struct gfs2_rgrpd
,
569 if (rgd
->rd_addr
< cur
->rd_addr
)
570 newn
= &((*newn
)->rb_left
);
571 else if (rgd
->rd_addr
> cur
->rd_addr
)
572 newn
= &((*newn
)->rb_right
);
577 rb_link_node(&rgd
->rd_node
, parent
, newn
);
578 rb_insert_color(&rgd
->rd_node
, &sdp
->sd_rindex_tree
);
584 * read_rindex_entry - Pull in a new resource index entry from the disk
585 * @gl: The glock covering the rindex inode
587 * Returns: 0 on success, > 0 on EOF, error code otherwise
590 static int read_rindex_entry(struct gfs2_inode
*ip
,
591 struct file_ra_state
*ra_state
)
593 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
594 loff_t pos
= sdp
->sd_rgrps
* sizeof(struct gfs2_rindex
);
595 struct gfs2_rindex buf
;
597 struct gfs2_rgrpd
*rgd
;
599 if (pos
>= i_size_read(&ip
->i_inode
))
602 error
= gfs2_internal_read(ip
, ra_state
, (char *)&buf
, &pos
,
603 sizeof(struct gfs2_rindex
));
605 if (error
!= sizeof(struct gfs2_rindex
))
606 return (error
== 0) ? 1 : error
;
608 rgd
= kmem_cache_zalloc(gfs2_rgrpd_cachep
, GFP_NOFS
);
614 rgd
->rd_addr
= be64_to_cpu(buf
.ri_addr
);
615 rgd
->rd_length
= be32_to_cpu(buf
.ri_length
);
616 rgd
->rd_data0
= be64_to_cpu(buf
.ri_data0
);
617 rgd
->rd_data
= be32_to_cpu(buf
.ri_data
);
618 rgd
->rd_bitbytes
= be32_to_cpu(buf
.ri_bitbytes
);
620 error
= compute_bitstructs(rgd
);
624 error
= gfs2_glock_get(sdp
, rgd
->rd_addr
,
625 &gfs2_rgrp_glops
, CREATE
, &rgd
->rd_gl
);
629 rgd
->rd_gl
->gl_object
= rgd
;
630 rgd
->rd_flags
&= ~GFS2_RDF_UPTODATE
;
631 if (rgd
->rd_data
> sdp
->sd_max_rg_data
)
632 sdp
->sd_max_rg_data
= rgd
->rd_data
;
633 spin_lock(&sdp
->sd_rindex_spin
);
634 error
= rgd_insert(rgd
);
635 spin_unlock(&sdp
->sd_rindex_spin
);
639 error
= 0; /* someone else read in the rgrp; free it and ignore it */
640 gfs2_glock_put(rgd
->rd_gl
);
644 kmem_cache_free(gfs2_rgrpd_cachep
, rgd
);
649 * gfs2_ri_update - Pull in a new resource index from the disk
650 * @ip: pointer to the rindex inode
652 * Returns: 0 on successful update, error code otherwise
655 static int gfs2_ri_update(struct gfs2_inode
*ip
)
657 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
658 struct inode
*inode
= &ip
->i_inode
;
659 struct file_ra_state ra_state
;
662 file_ra_state_init(&ra_state
, inode
->i_mapping
);
664 error
= read_rindex_entry(ip
, &ra_state
);
665 } while (error
== 0);
670 sdp
->sd_rindex_uptodate
= 1;
675 * gfs2_rindex_update - Update the rindex if required
676 * @sdp: The GFS2 superblock
678 * We grab a lock on the rindex inode to make sure that it doesn't
679 * change whilst we are performing an operation. We keep this lock
680 * for quite long periods of time compared to other locks. This
681 * doesn't matter, since it is shared and it is very, very rarely
682 * accessed in the exclusive mode (i.e. only when expanding the filesystem).
684 * This makes sure that we're using the latest copy of the resource index
685 * special file, which might have been updated if someone expanded the
686 * filesystem (via gfs2_grow utility), which adds new resource groups.
688 * Returns: 0 on succeess, error code otherwise
691 int gfs2_rindex_update(struct gfs2_sbd
*sdp
)
693 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_rindex
);
694 struct gfs2_glock
*gl
= ip
->i_gl
;
695 struct gfs2_holder ri_gh
;
697 int unlock_required
= 0;
699 /* Read new copy from disk if we don't have the latest */
700 if (!sdp
->sd_rindex_uptodate
) {
701 if (!gfs2_glock_is_locked_by_me(gl
)) {
702 error
= gfs2_glock_nq_init(gl
, LM_ST_SHARED
, 0, &ri_gh
);
707 if (!sdp
->sd_rindex_uptodate
)
708 error
= gfs2_ri_update(ip
);
710 gfs2_glock_dq_uninit(&ri_gh
);
716 static void gfs2_rgrp_in(struct gfs2_rgrpd
*rgd
, const void *buf
)
718 const struct gfs2_rgrp
*str
= buf
;
721 rg_flags
= be32_to_cpu(str
->rg_flags
);
722 rg_flags
&= ~GFS2_RDF_MASK
;
723 rgd
->rd_flags
&= GFS2_RDF_MASK
;
724 rgd
->rd_flags
|= rg_flags
;
725 rgd
->rd_free
= be32_to_cpu(str
->rg_free
);
726 rgd
->rd_dinodes
= be32_to_cpu(str
->rg_dinodes
);
727 rgd
->rd_igeneration
= be64_to_cpu(str
->rg_igeneration
);
730 static void gfs2_rgrp_out(struct gfs2_rgrpd
*rgd
, void *buf
)
732 struct gfs2_rgrp
*str
= buf
;
734 str
->rg_flags
= cpu_to_be32(rgd
->rd_flags
& ~GFS2_RDF_MASK
);
735 str
->rg_free
= cpu_to_be32(rgd
->rd_free
);
736 str
->rg_dinodes
= cpu_to_be32(rgd
->rd_dinodes
);
737 str
->__pad
= cpu_to_be32(0);
738 str
->rg_igeneration
= cpu_to_be64(rgd
->rd_igeneration
);
739 memset(&str
->rg_reserved
, 0, sizeof(str
->rg_reserved
));
743 * gfs2_rgrp_go_lock - Read in a RG's header and bitmaps
744 * @rgd: the struct gfs2_rgrpd describing the RG to read in
746 * Read in all of a Resource Group's header and bitmap blocks.
747 * Caller must eventually call gfs2_rgrp_relse() to free the bitmaps.
752 int gfs2_rgrp_go_lock(struct gfs2_holder
*gh
)
754 struct gfs2_rgrpd
*rgd
= gh
->gh_gl
->gl_object
;
755 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
756 struct gfs2_glock
*gl
= rgd
->rd_gl
;
757 unsigned int length
= rgd
->rd_length
;
758 struct gfs2_bitmap
*bi
;
762 for (x
= 0; x
< length
; x
++) {
763 bi
= rgd
->rd_bits
+ x
;
764 error
= gfs2_meta_read(gl
, rgd
->rd_addr
+ x
, 0, &bi
->bi_bh
);
769 for (y
= length
; y
--;) {
770 bi
= rgd
->rd_bits
+ y
;
771 error
= gfs2_meta_wait(sdp
, bi
->bi_bh
);
774 if (gfs2_metatype_check(sdp
, bi
->bi_bh
, y
? GFS2_METATYPE_RB
:
781 if (!(rgd
->rd_flags
& GFS2_RDF_UPTODATE
)) {
782 for (x
= 0; x
< length
; x
++)
783 clear_bit(GBF_FULL
, &rgd
->rd_bits
[x
].bi_flags
);
784 gfs2_rgrp_in(rgd
, (rgd
->rd_bits
[0].bi_bh
)->b_data
);
785 rgd
->rd_flags
|= (GFS2_RDF_UPTODATE
| GFS2_RDF_CHECK
);
786 rgd
->rd_free_clone
= rgd
->rd_free
;
793 bi
= rgd
->rd_bits
+ x
;
796 gfs2_assert_warn(sdp
, !bi
->bi_clone
);
803 * gfs2_rgrp_go_unlock - Release RG bitmaps read in with gfs2_rgrp_bh_get()
804 * @rgd: the struct gfs2_rgrpd describing the RG to read in
808 void gfs2_rgrp_go_unlock(struct gfs2_holder
*gh
)
810 struct gfs2_rgrpd
*rgd
= gh
->gh_gl
->gl_object
;
811 int x
, length
= rgd
->rd_length
;
813 for (x
= 0; x
< length
; x
++) {
814 struct gfs2_bitmap
*bi
= rgd
->rd_bits
+ x
;
821 int gfs2_rgrp_send_discards(struct gfs2_sbd
*sdp
, u64 offset
,
822 struct buffer_head
*bh
,
823 const struct gfs2_bitmap
*bi
, unsigned minlen
, u64
*ptrimmed
)
825 struct super_block
*sb
= sdp
->sd_vfs
;
826 struct block_device
*bdev
= sb
->s_bdev
;
827 const unsigned int sects_per_blk
= sdp
->sd_sb
.sb_bsize
/
828 bdev_logical_block_size(sb
->s_bdev
);
831 sector_t nr_sects
= 0;
837 for (x
= 0; x
< bi
->bi_len
; x
++) {
838 const u8
*clone
= bi
->bi_clone
? bi
->bi_clone
: bi
->bi_bh
->b_data
;
839 clone
+= bi
->bi_offset
;
842 const u8
*orig
= bh
->b_data
+ bi
->bi_offset
+ x
;
843 diff
= ~(*orig
| (*orig
>> 1)) & (*clone
| (*clone
>> 1));
845 diff
= ~(*clone
| (*clone
>> 1));
850 blk
= offset
+ ((bi
->bi_start
+ x
) * GFS2_NBBY
);
851 blk
*= sects_per_blk
; /* convert to sectors */
855 goto start_new_extent
;
856 if ((start
+ nr_sects
) != blk
) {
857 if (nr_sects
>= minlen
) {
858 rv
= blkdev_issue_discard(bdev
,
869 nr_sects
+= sects_per_blk
;
872 blk
+= sects_per_blk
;
875 if (nr_sects
>= minlen
) {
876 rv
= blkdev_issue_discard(bdev
, start
, nr_sects
, GFP_NOFS
, 0);
886 if (sdp
->sd_args
.ar_discard
)
887 fs_warn(sdp
, "error %d on discard request, turning discards off for this filesystem", rv
);
888 sdp
->sd_args
.ar_discard
= 0;
893 * gfs2_fitrim - Generate discard requests for unused bits of the filesystem
894 * @filp: Any file on the filesystem
895 * @argp: Pointer to the arguments (also used to pass result)
897 * Returns: 0 on success, otherwise error code
900 int gfs2_fitrim(struct file
*filp
, void __user
*argp
)
902 struct inode
*inode
= filp
->f_dentry
->d_inode
;
903 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
904 struct request_queue
*q
= bdev_get_queue(sdp
->sd_vfs
->s_bdev
);
905 struct buffer_head
*bh
;
906 struct gfs2_rgrpd
*rgd
;
907 struct gfs2_rgrpd
*rgd_end
;
908 struct gfs2_holder gh
;
909 struct fstrim_range r
;
915 if (!capable(CAP_SYS_ADMIN
))
918 if (!blk_queue_discard(q
))
925 } else if (copy_from_user(&r
, argp
, sizeof(r
)))
928 ret
= gfs2_rindex_update(sdp
);
932 rgd
= gfs2_blk2rgrpd(sdp
, r
.start
, 0);
933 rgd_end
= gfs2_blk2rgrpd(sdp
, r
.start
+ r
.len
, 0);
937 ret
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
941 if (!(rgd
->rd_flags
& GFS2_RGF_TRIMMED
)) {
942 /* Trim each bitmap in the rgrp */
943 for (x
= 0; x
< rgd
->rd_length
; x
++) {
944 struct gfs2_bitmap
*bi
= rgd
->rd_bits
+ x
;
945 ret
= gfs2_rgrp_send_discards(sdp
, rgd
->rd_data0
, NULL
, bi
, r
.minlen
, &amt
);
947 gfs2_glock_dq_uninit(&gh
);
953 /* Mark rgrp as having been trimmed */
954 ret
= gfs2_trans_begin(sdp
, RES_RG_HDR
, 0);
956 bh
= rgd
->rd_bits
[0].bi_bh
;
957 rgd
->rd_flags
|= GFS2_RGF_TRIMMED
;
958 gfs2_trans_add_bh(rgd
->rd_gl
, bh
, 1);
959 gfs2_rgrp_out(rgd
, bh
->b_data
);
963 gfs2_glock_dq_uninit(&gh
);
968 rgd
= gfs2_rgrpd_get_next(rgd
);
972 r
.len
= trimmed
<< 9;
973 if (argp
&& copy_to_user(argp
, &r
, sizeof(r
)))
980 * gfs2_qadata_get - get the struct gfs2_qadata structure for an inode
981 * @ip: the incore GFS2 inode structure
983 * Returns: the struct gfs2_qadata
986 struct gfs2_qadata
*gfs2_qadata_get(struct gfs2_inode
*ip
)
988 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
990 BUG_ON(ip
->i_qadata
!= NULL
);
991 ip
->i_qadata
= kzalloc(sizeof(struct gfs2_qadata
), GFP_NOFS
);
992 error
= gfs2_rindex_update(sdp
);
994 fs_warn(sdp
, "rindex update returns %d\n", error
);
999 * gfs2_blkrsv_get - get the struct gfs2_blkreserv structure for an inode
1000 * @ip: the incore GFS2 inode structure
1002 * Returns: the struct gfs2_qadata
1005 static struct gfs2_blkreserv
*gfs2_blkrsv_get(struct gfs2_inode
*ip
)
1007 BUG_ON(ip
->i_res
!= NULL
);
1008 ip
->i_res
= kzalloc(sizeof(struct gfs2_blkreserv
), GFP_NOFS
);
1013 * try_rgrp_fit - See if a given reservation will fit in a given RG
1017 * If there's room for the requested blocks to be allocated from the RG:
1019 * Returns: 1 on success (it fits), 0 on failure (it doesn't fit)
1022 static int try_rgrp_fit(const struct gfs2_rgrpd
*rgd
, const struct gfs2_inode
*ip
)
1024 const struct gfs2_blkreserv
*rs
= ip
->i_res
;
1026 if (rgd
->rd_flags
& (GFS2_RGF_NOALLOC
| GFS2_RDF_ERROR
))
1028 if (rgd
->rd_free_clone
>= rs
->rs_requested
)
1033 static inline u32
gfs2_bi2rgd_blk(struct gfs2_bitmap
*bi
, u32 blk
)
1035 return (bi
->bi_start
* GFS2_NBBY
) + blk
;
1039 * try_rgrp_unlink - Look for any unlinked, allocated, but unused inodes
1042 * Returns: 0 if no error
1043 * The inode, if one has been found, in inode.
1046 static void try_rgrp_unlink(struct gfs2_rgrpd
*rgd
, u64
*last_unlinked
, u64 skip
)
1048 u32 goal
= 0, block
;
1050 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
1051 struct gfs2_glock
*gl
;
1052 struct gfs2_inode
*ip
;
1055 struct gfs2_bitmap
*bi
;
1057 while (goal
< rgd
->rd_data
) {
1058 down_write(&sdp
->sd_log_flush_lock
);
1059 block
= rgblk_search(rgd
, goal
, GFS2_BLKST_UNLINKED
, &bi
);
1060 up_write(&sdp
->sd_log_flush_lock
);
1061 if (block
== BFITNOENT
)
1064 block
= gfs2_bi2rgd_blk(bi
, block
);
1065 /* rgblk_search can return a block < goal, so we need to
1066 keep it marching forward. */
1067 no_addr
= block
+ rgd
->rd_data0
;
1068 goal
= max(block
+ 1, goal
+ 1);
1069 if (*last_unlinked
!= NO_BLOCK
&& no_addr
<= *last_unlinked
)
1071 if (no_addr
== skip
)
1073 *last_unlinked
= no_addr
;
1075 error
= gfs2_glock_get(sdp
, no_addr
, &gfs2_inode_glops
, CREATE
, &gl
);
1079 /* If the inode is already in cache, we can ignore it here
1080 * because the existing inode disposal code will deal with
1081 * it when all refs have gone away. Accessing gl_object like
1082 * this is not safe in general. Here it is ok because we do
1083 * not dereference the pointer, and we only need an approx
1084 * answer to whether it is NULL or not.
1088 if (ip
|| queue_work(gfs2_delete_workqueue
, &gl
->gl_delete
) == 0)
1093 /* Limit reclaim to sensible number of tasks */
1094 if (found
> NR_CPUS
)
1098 rgd
->rd_flags
&= ~GFS2_RDF_CHECK
;
1103 * get_local_rgrp - Choose and lock a rgrp for allocation
1104 * @ip: the inode to reserve space for
1105 * @rgp: the chosen and locked rgrp
1107 * Try to acquire rgrp in way which avoids contending with others.
1112 static int get_local_rgrp(struct gfs2_inode
*ip
, u64
*last_unlinked
)
1114 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
1115 struct gfs2_rgrpd
*rgd
, *begin
= NULL
;
1116 struct gfs2_blkreserv
*rs
= ip
->i_res
;
1117 int error
, rg_locked
, flags
= LM_FLAG_TRY
;
1120 if (ip
->i_rgd
&& rgrp_contains_block(ip
->i_rgd
, ip
->i_goal
))
1121 rgd
= begin
= ip
->i_rgd
;
1123 rgd
= begin
= gfs2_blk2rgrpd(sdp
, ip
->i_goal
, 1);
1131 if (gfs2_glock_is_locked_by_me(rgd
->rd_gl
)) {
1135 error
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_EXCLUSIVE
,
1136 flags
, &rs
->rs_rgd_gh
);
1140 if (try_rgrp_fit(rgd
, ip
)) {
1144 if (rgd
->rd_flags
& GFS2_RDF_CHECK
)
1145 try_rgrp_unlink(rgd
, last_unlinked
, ip
->i_no_addr
);
1147 gfs2_glock_dq_uninit(&rs
->rs_rgd_gh
);
1150 rgd
= gfs2_rgrpd_get_next(rgd
);
1164 static void gfs2_blkrsv_put(struct gfs2_inode
*ip
)
1166 BUG_ON(ip
->i_res
== NULL
);
1172 * gfs2_inplace_reserve - Reserve space in the filesystem
1173 * @ip: the inode to reserve space for
1178 int gfs2_inplace_reserve(struct gfs2_inode
*ip
, u32 requested
)
1180 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
1181 struct gfs2_blkreserv
*rs
;
1183 u64 last_unlinked
= NO_BLOCK
;
1186 rs
= gfs2_blkrsv_get(ip
);
1190 rs
->rs_requested
= requested
;
1191 if (gfs2_assert_warn(sdp
, requested
)) {
1197 error
= get_local_rgrp(ip
, &last_unlinked
);
1198 if (error
!= -ENOSPC
)
1200 /* Check that fs hasn't grown if writing to rindex */
1201 if (ip
== GFS2_I(sdp
->sd_rindex
) && !sdp
->sd_rindex_uptodate
) {
1202 error
= gfs2_ri_update(ip
);
1207 /* Flushing the log may release space */
1208 gfs2_log_flush(sdp
, NULL
);
1209 } while (tries
++ < 3);
1213 gfs2_blkrsv_put(ip
);
1218 * gfs2_inplace_release - release an inplace reservation
1219 * @ip: the inode the reservation was taken out on
1221 * Release a reservation made by gfs2_inplace_reserve().
1224 void gfs2_inplace_release(struct gfs2_inode
*ip
)
1226 struct gfs2_blkreserv
*rs
= ip
->i_res
;
1228 if (rs
->rs_rgd_gh
.gh_gl
)
1229 gfs2_glock_dq_uninit(&rs
->rs_rgd_gh
);
1230 gfs2_blkrsv_put(ip
);
1234 * gfs2_get_block_type - Check a block in a RG is of given type
1235 * @rgd: the resource group holding the block
1236 * @block: the block number
1238 * Returns: The block type (GFS2_BLKST_*)
1241 static unsigned char gfs2_get_block_type(struct gfs2_rgrpd
*rgd
, u64 block
)
1243 struct gfs2_bitmap
*bi
= NULL
;
1244 u32 length
, rgrp_block
, buf_block
;
1248 length
= rgd
->rd_length
;
1249 rgrp_block
= block
- rgd
->rd_data0
;
1251 for (buf
= 0; buf
< length
; buf
++) {
1252 bi
= rgd
->rd_bits
+ buf
;
1253 if (rgrp_block
< (bi
->bi_start
+ bi
->bi_len
) * GFS2_NBBY
)
1257 gfs2_assert(rgd
->rd_sbd
, buf
< length
);
1258 buf_block
= rgrp_block
- bi
->bi_start
* GFS2_NBBY
;
1260 type
= gfs2_testbit(rgd
, bi
->bi_bh
->b_data
+ bi
->bi_offset
,
1261 bi
->bi_len
, buf_block
);
1267 * rgblk_search - find a block in @state
1268 * @rgd: the resource group descriptor
1269 * @goal: the goal block within the RG (start here to search for avail block)
1270 * @state: GFS2_BLKST_XXX the before-allocation state to find
1271 * @dinode: TRUE if the first block we allocate is for a dinode
1272 * @rbi: address of the pointer to the bitmap containing the block found
1274 * Walk rgrp's bitmap to find bits that represent a block in @state.
1276 * This function never fails, because we wouldn't call it unless we
1277 * know (from reservation results, etc.) that a block is available.
1279 * Scope of @goal is just within rgrp, not the whole filesystem.
1280 * Scope of @returned block is just within bitmap, not the whole filesystem.
1282 * Returns: the block number found relative to the bitmap rbi
1285 static u32
rgblk_search(struct gfs2_rgrpd
*rgd
, u32 goal
,
1286 unsigned char state
,
1287 struct gfs2_bitmap
**rbi
)
1289 struct gfs2_bitmap
*bi
= NULL
;
1290 const u32 length
= rgd
->rd_length
;
1291 u32 blk
= BFITNOENT
;
1292 unsigned int buf
, x
;
1293 const u8
*buffer
= NULL
;
1296 /* Find bitmap block that contains bits for goal block */
1297 for (buf
= 0; buf
< length
; buf
++) {
1298 bi
= rgd
->rd_bits
+ buf
;
1299 /* Convert scope of "goal" from rgrp-wide to within found bit block */
1300 if (goal
< (bi
->bi_start
+ bi
->bi_len
) * GFS2_NBBY
) {
1301 goal
-= bi
->bi_start
* GFS2_NBBY
;
1309 /* Search (up to entire) bitmap in this rgrp for allocatable block.
1310 "x <= length", instead of "x < length", because we typically start
1311 the search in the middle of a bit block, but if we can't find an
1312 allocatable block anywhere else, we want to be able wrap around and
1313 search in the first part of our first-searched bit block. */
1314 for (x
= 0; x
<= length
; x
++) {
1315 bi
= rgd
->rd_bits
+ buf
;
1317 if (test_bit(GBF_FULL
, &bi
->bi_flags
) &&
1318 (state
== GFS2_BLKST_FREE
))
1321 /* The GFS2_BLKST_UNLINKED state doesn't apply to the clone
1322 bitmaps, so we must search the originals for that. */
1323 buffer
= bi
->bi_bh
->b_data
+ bi
->bi_offset
;
1324 WARN_ON(!buffer_uptodate(bi
->bi_bh
));
1325 if (state
!= GFS2_BLKST_UNLINKED
&& bi
->bi_clone
)
1326 buffer
= bi
->bi_clone
+ bi
->bi_offset
;
1328 blk
= gfs2_bitfit(buffer
, bi
->bi_len
, goal
, state
);
1329 if (blk
!= BFITNOENT
)
1332 if ((goal
== 0) && (state
== GFS2_BLKST_FREE
))
1333 set_bit(GBF_FULL
, &bi
->bi_flags
);
1335 /* Try next bitmap block (wrap back to rgrp header if at end) */
1342 if (blk
!= BFITNOENT
)
1349 * gfs2_alloc_extent - allocate an extent from a given bitmap
1350 * @rgd: the resource group descriptor
1351 * @bi: the bitmap within the rgrp
1352 * @blk: the block within the bitmap
1353 * @dinode: TRUE if the first block we allocate is for a dinode
1354 * @n: The extent length
1356 * Add the found bitmap buffer to the transaction.
1357 * Set the found bits to @new_state to change block's allocation state.
1358 * Returns: starting block number of the extent (fs scope)
1360 static u64
gfs2_alloc_extent(struct gfs2_rgrpd
*rgd
, struct gfs2_bitmap
*bi
,
1361 u32 blk
, bool dinode
, unsigned int *n
)
1363 const unsigned int elen
= *n
;
1365 const u8
*buffer
= NULL
;
1368 buffer
= bi
->bi_bh
->b_data
+ bi
->bi_offset
;
1369 gfs2_trans_add_bh(rgd
->rd_gl
, bi
->bi_bh
, 1);
1370 gfs2_setbit(rgd
, bi
->bi_bh
->b_data
, bi
->bi_clone
, bi
->bi_offset
,
1371 bi
, blk
, dinode
? GFS2_BLKST_DINODE
: GFS2_BLKST_USED
);
1376 if (goal
>= (bi
->bi_len
* GFS2_NBBY
))
1378 if (gfs2_testbit(rgd
, buffer
, bi
->bi_len
, goal
) !=
1381 gfs2_setbit(rgd
, bi
->bi_bh
->b_data
, bi
->bi_clone
, bi
->bi_offset
,
1382 bi
, goal
, GFS2_BLKST_USED
);
1385 blk
= gfs2_bi2rgd_blk(bi
, blk
);
1386 rgd
->rd_last_alloc
= blk
+ *n
- 1;
1387 return rgd
->rd_data0
+ blk
;
1391 * rgblk_free - Change alloc state of given block(s)
1392 * @sdp: the filesystem
1393 * @bstart: the start of a run of blocks to free
1394 * @blen: the length of the block run (all must lie within ONE RG!)
1395 * @new_state: GFS2_BLKST_XXX the after-allocation block state
1397 * Returns: Resource group containing the block(s)
1400 static struct gfs2_rgrpd
*rgblk_free(struct gfs2_sbd
*sdp
, u64 bstart
,
1401 u32 blen
, unsigned char new_state
)
1403 struct gfs2_rgrpd
*rgd
;
1404 struct gfs2_bitmap
*bi
= NULL
;
1405 u32 length
, rgrp_blk
, buf_blk
;
1408 rgd
= gfs2_blk2rgrpd(sdp
, bstart
, 1);
1410 if (gfs2_consist(sdp
))
1411 fs_err(sdp
, "block = %llu\n", (unsigned long long)bstart
);
1415 length
= rgd
->rd_length
;
1417 rgrp_blk
= bstart
- rgd
->rd_data0
;
1420 for (buf
= 0; buf
< length
; buf
++) {
1421 bi
= rgd
->rd_bits
+ buf
;
1422 if (rgrp_blk
< (bi
->bi_start
+ bi
->bi_len
) * GFS2_NBBY
)
1426 gfs2_assert(rgd
->rd_sbd
, buf
< length
);
1428 buf_blk
= rgrp_blk
- bi
->bi_start
* GFS2_NBBY
;
1431 if (!bi
->bi_clone
) {
1432 bi
->bi_clone
= kmalloc(bi
->bi_bh
->b_size
,
1433 GFP_NOFS
| __GFP_NOFAIL
);
1434 memcpy(bi
->bi_clone
+ bi
->bi_offset
,
1435 bi
->bi_bh
->b_data
+ bi
->bi_offset
,
1438 gfs2_trans_add_bh(rgd
->rd_gl
, bi
->bi_bh
, 1);
1439 gfs2_setbit(rgd
, bi
->bi_bh
->b_data
, NULL
, bi
->bi_offset
,
1440 bi
, buf_blk
, new_state
);
1447 * gfs2_rgrp_dump - print out an rgrp
1448 * @seq: The iterator
1449 * @gl: The glock in question
1453 int gfs2_rgrp_dump(struct seq_file
*seq
, const struct gfs2_glock
*gl
)
1455 const struct gfs2_rgrpd
*rgd
= gl
->gl_object
;
1458 gfs2_print_dbg(seq
, " R: n:%llu f:%02x b:%u/%u i:%u\n",
1459 (unsigned long long)rgd
->rd_addr
, rgd
->rd_flags
,
1460 rgd
->rd_free
, rgd
->rd_free_clone
, rgd
->rd_dinodes
);
1464 static void gfs2_rgrp_error(struct gfs2_rgrpd
*rgd
)
1466 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
1467 fs_warn(sdp
, "rgrp %llu has an error, marking it readonly until umount\n",
1468 (unsigned long long)rgd
->rd_addr
);
1469 fs_warn(sdp
, "umount on all nodes and run fsck.gfs2 to fix the error\n");
1470 gfs2_rgrp_dump(NULL
, rgd
->rd_gl
);
1471 rgd
->rd_flags
|= GFS2_RDF_ERROR
;
1475 * gfs2_alloc_blocks - Allocate one or more blocks of data and/or a dinode
1476 * @ip: the inode to allocate the block for
1477 * @bn: Used to return the starting block number
1478 * @ndata: requested number of blocks/extent length (value/result)
1479 * @dinode: 1 if we're allocating a dinode block, else 0
1480 * @generation: the generation number of the inode
1482 * Returns: 0 or error
1485 int gfs2_alloc_blocks(struct gfs2_inode
*ip
, u64
*bn
, unsigned int *nblocks
,
1486 bool dinode
, u64
*generation
)
1488 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
1489 struct buffer_head
*dibh
;
1490 struct gfs2_rgrpd
*rgd
;
1492 u32 goal
, blk
; /* block, within the rgrp scope */
1493 u64 block
; /* block, within the file system scope */
1495 struct gfs2_bitmap
*bi
;
1497 /* Only happens if there is a bug in gfs2, return something distinctive
1498 * to ensure that it is noticed.
1500 if (ip
->i_res
== NULL
)
1505 if (!dinode
&& rgrp_contains_block(rgd
, ip
->i_goal
))
1506 goal
= ip
->i_goal
- rgd
->rd_data0
;
1508 goal
= rgd
->rd_last_alloc
;
1510 blk
= rgblk_search(rgd
, goal
, GFS2_BLKST_FREE
, &bi
);
1512 /* Since all blocks are reserved in advance, this shouldn't happen */
1513 if (blk
== BFITNOENT
)
1516 block
= gfs2_alloc_extent(rgd
, bi
, blk
, dinode
, nblocks
);
1522 ip
->i_goal
= block
+ ndata
- 1;
1523 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
1525 struct gfs2_dinode
*di
=
1526 (struct gfs2_dinode
*)dibh
->b_data
;
1527 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
1528 di
->di_goal_meta
= di
->di_goal_data
=
1529 cpu_to_be64(ip
->i_goal
);
1533 if (rgd
->rd_free
< *nblocks
)
1536 rgd
->rd_free
-= *nblocks
;
1539 *generation
= rgd
->rd_igeneration
++;
1540 if (*generation
== 0)
1541 *generation
= rgd
->rd_igeneration
++;
1544 gfs2_trans_add_bh(rgd
->rd_gl
, rgd
->rd_bits
[0].bi_bh
, 1);
1545 gfs2_rgrp_out(rgd
, rgd
->rd_bits
[0].bi_bh
->b_data
);
1547 gfs2_statfs_change(sdp
, 0, -(s64
)*nblocks
, dinode
? 1 : 0);
1549 gfs2_trans_add_unrevoke(sdp
, block
, 1);
1552 * This needs reviewing to see why we cannot do the quota change
1553 * at this point in the dinode case.
1556 gfs2_quota_change(ip
, ndata
, ip
->i_inode
.i_uid
,
1559 rgd
->rd_free_clone
-= *nblocks
;
1560 trace_gfs2_block_alloc(ip
, block
, *nblocks
,
1561 dinode
? GFS2_BLKST_DINODE
: GFS2_BLKST_USED
);
1566 gfs2_rgrp_error(rgd
);
1571 * __gfs2_free_blocks - free a contiguous run of block(s)
1572 * @ip: the inode these blocks are being freed from
1573 * @bstart: first block of a run of contiguous blocks
1574 * @blen: the length of the block run
1575 * @meta: 1 if the blocks represent metadata
1579 void __gfs2_free_blocks(struct gfs2_inode
*ip
, u64 bstart
, u32 blen
, int meta
)
1581 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
1582 struct gfs2_rgrpd
*rgd
;
1584 rgd
= rgblk_free(sdp
, bstart
, blen
, GFS2_BLKST_FREE
);
1587 trace_gfs2_block_alloc(ip
, bstart
, blen
, GFS2_BLKST_FREE
);
1588 rgd
->rd_free
+= blen
;
1589 rgd
->rd_flags
&= ~GFS2_RGF_TRIMMED
;
1590 gfs2_trans_add_bh(rgd
->rd_gl
, rgd
->rd_bits
[0].bi_bh
, 1);
1591 gfs2_rgrp_out(rgd
, rgd
->rd_bits
[0].bi_bh
->b_data
);
1593 /* Directories keep their data in the metadata address space */
1594 if (meta
|| ip
->i_depth
)
1595 gfs2_meta_wipe(ip
, bstart
, blen
);
1599 * gfs2_free_meta - free a contiguous run of data block(s)
1600 * @ip: the inode these blocks are being freed from
1601 * @bstart: first block of a run of contiguous blocks
1602 * @blen: the length of the block run
1606 void gfs2_free_meta(struct gfs2_inode
*ip
, u64 bstart
, u32 blen
)
1608 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
1610 __gfs2_free_blocks(ip
, bstart
, blen
, 1);
1611 gfs2_statfs_change(sdp
, 0, +blen
, 0);
1612 gfs2_quota_change(ip
, -(s64
)blen
, ip
->i_inode
.i_uid
, ip
->i_inode
.i_gid
);
1615 void gfs2_unlink_di(struct inode
*inode
)
1617 struct gfs2_inode
*ip
= GFS2_I(inode
);
1618 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
1619 struct gfs2_rgrpd
*rgd
;
1620 u64 blkno
= ip
->i_no_addr
;
1622 rgd
= rgblk_free(sdp
, blkno
, 1, GFS2_BLKST_UNLINKED
);
1625 trace_gfs2_block_alloc(ip
, blkno
, 1, GFS2_BLKST_UNLINKED
);
1626 gfs2_trans_add_bh(rgd
->rd_gl
, rgd
->rd_bits
[0].bi_bh
, 1);
1627 gfs2_rgrp_out(rgd
, rgd
->rd_bits
[0].bi_bh
->b_data
);
1630 static void gfs2_free_uninit_di(struct gfs2_rgrpd
*rgd
, u64 blkno
)
1632 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
1633 struct gfs2_rgrpd
*tmp_rgd
;
1635 tmp_rgd
= rgblk_free(sdp
, blkno
, 1, GFS2_BLKST_FREE
);
1638 gfs2_assert_withdraw(sdp
, rgd
== tmp_rgd
);
1640 if (!rgd
->rd_dinodes
)
1641 gfs2_consist_rgrpd(rgd
);
1645 gfs2_trans_add_bh(rgd
->rd_gl
, rgd
->rd_bits
[0].bi_bh
, 1);
1646 gfs2_rgrp_out(rgd
, rgd
->rd_bits
[0].bi_bh
->b_data
);
1648 gfs2_statfs_change(sdp
, 0, +1, -1);
1652 void gfs2_free_di(struct gfs2_rgrpd
*rgd
, struct gfs2_inode
*ip
)
1654 gfs2_free_uninit_di(rgd
, ip
->i_no_addr
);
1655 trace_gfs2_block_alloc(ip
, ip
->i_no_addr
, 1, GFS2_BLKST_FREE
);
1656 gfs2_quota_change(ip
, -1, ip
->i_inode
.i_uid
, ip
->i_inode
.i_gid
);
1657 gfs2_meta_wipe(ip
, ip
->i_no_addr
, 1);
1661 * gfs2_check_blk_type - Check the type of a block
1662 * @sdp: The superblock
1663 * @no_addr: The block number to check
1664 * @type: The block type we are looking for
1666 * Returns: 0 if the block type matches the expected type
1667 * -ESTALE if it doesn't match
1668 * or -ve errno if something went wrong while checking
1671 int gfs2_check_blk_type(struct gfs2_sbd
*sdp
, u64 no_addr
, unsigned int type
)
1673 struct gfs2_rgrpd
*rgd
;
1674 struct gfs2_holder rgd_gh
;
1675 int error
= -EINVAL
;
1677 rgd
= gfs2_blk2rgrpd(sdp
, no_addr
, 1);
1681 error
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_SHARED
, 0, &rgd_gh
);
1685 if (gfs2_get_block_type(rgd
, no_addr
) != type
)
1688 gfs2_glock_dq_uninit(&rgd_gh
);
1694 * gfs2_rlist_add - add a RG to a list of RGs
1696 * @rlist: the list of resource groups
1699 * Figure out what RG a block belongs to and add that RG to the list
1701 * FIXME: Don't use NOFAIL
1705 void gfs2_rlist_add(struct gfs2_inode
*ip
, struct gfs2_rgrp_list
*rlist
,
1708 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
1709 struct gfs2_rgrpd
*rgd
;
1710 struct gfs2_rgrpd
**tmp
;
1711 unsigned int new_space
;
1714 if (gfs2_assert_warn(sdp
, !rlist
->rl_ghs
))
1717 if (ip
->i_rgd
&& rgrp_contains_block(ip
->i_rgd
, block
))
1720 rgd
= gfs2_blk2rgrpd(sdp
, block
, 1);
1722 fs_err(sdp
, "rlist_add: no rgrp for block %llu\n", (unsigned long long)block
);
1727 for (x
= 0; x
< rlist
->rl_rgrps
; x
++)
1728 if (rlist
->rl_rgd
[x
] == rgd
)
1731 if (rlist
->rl_rgrps
== rlist
->rl_space
) {
1732 new_space
= rlist
->rl_space
+ 10;
1734 tmp
= kcalloc(new_space
, sizeof(struct gfs2_rgrpd
*),
1735 GFP_NOFS
| __GFP_NOFAIL
);
1737 if (rlist
->rl_rgd
) {
1738 memcpy(tmp
, rlist
->rl_rgd
,
1739 rlist
->rl_space
* sizeof(struct gfs2_rgrpd
*));
1740 kfree(rlist
->rl_rgd
);
1743 rlist
->rl_space
= new_space
;
1744 rlist
->rl_rgd
= tmp
;
1747 rlist
->rl_rgd
[rlist
->rl_rgrps
++] = rgd
;
1751 * gfs2_rlist_alloc - all RGs have been added to the rlist, now allocate
1752 * and initialize an array of glock holders for them
1753 * @rlist: the list of resource groups
1754 * @state: the lock state to acquire the RG lock in
1755 * @flags: the modifier flags for the holder structures
1757 * FIXME: Don't use NOFAIL
1761 void gfs2_rlist_alloc(struct gfs2_rgrp_list
*rlist
, unsigned int state
)
1765 rlist
->rl_ghs
= kcalloc(rlist
->rl_rgrps
, sizeof(struct gfs2_holder
),
1766 GFP_NOFS
| __GFP_NOFAIL
);
1767 for (x
= 0; x
< rlist
->rl_rgrps
; x
++)
1768 gfs2_holder_init(rlist
->rl_rgd
[x
]->rd_gl
,
1774 * gfs2_rlist_free - free a resource group list
1775 * @list: the list of resource groups
1779 void gfs2_rlist_free(struct gfs2_rgrp_list
*rlist
)
1783 kfree(rlist
->rl_rgd
);
1785 if (rlist
->rl_ghs
) {
1786 for (x
= 0; x
< rlist
->rl_rgrps
; x
++)
1787 gfs2_holder_uninit(&rlist
->rl_ghs
[x
]);
1788 kfree(rlist
->rl_ghs
);