]>
git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - fs/gfs2/rgrp.c
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 int gfs2_rbm_find(struct gfs2_rbm
*rbm
, u8 state
, u32 minext
,
68 const struct gfs2_inode
*ip
, bool nowrap
);
72 * gfs2_setbit - Set a bit in the bitmaps
73 * @rbm: The position of the bit to set
74 * @do_clone: Also set the clone bitmap, if it exists
75 * @new_state: the new state of the block
79 static inline void gfs2_setbit(const struct gfs2_rbm
*rbm
, bool do_clone
,
80 unsigned char new_state
)
82 unsigned char *byte1
, *byte2
, *end
, cur_state
;
83 unsigned int buflen
= rbm
->bi
->bi_len
;
84 const unsigned int bit
= (rbm
->offset
% GFS2_NBBY
) * GFS2_BIT_SIZE
;
86 byte1
= rbm
->bi
->bi_bh
->b_data
+ rbm
->bi
->bi_offset
+ (rbm
->offset
/ GFS2_NBBY
);
87 end
= rbm
->bi
->bi_bh
->b_data
+ rbm
->bi
->bi_offset
+ buflen
;
91 cur_state
= (*byte1
>> bit
) & GFS2_BIT_MASK
;
93 if (unlikely(!valid_change
[new_state
* 4 + cur_state
])) {
94 printk(KERN_WARNING
"GFS2: buf_blk = 0x%x old_state=%d, "
95 "new_state=%d\n", rbm
->offset
, cur_state
, new_state
);
96 printk(KERN_WARNING
"GFS2: rgrp=0x%llx bi_start=0x%x\n",
97 (unsigned long long)rbm
->rgd
->rd_addr
,
99 printk(KERN_WARNING
"GFS2: bi_offset=0x%x bi_len=0x%x\n",
100 rbm
->bi
->bi_offset
, rbm
->bi
->bi_len
);
102 gfs2_consist_rgrpd(rbm
->rgd
);
105 *byte1
^= (cur_state
^ new_state
) << bit
;
107 if (do_clone
&& rbm
->bi
->bi_clone
) {
108 byte2
= rbm
->bi
->bi_clone
+ rbm
->bi
->bi_offset
+ (rbm
->offset
/ GFS2_NBBY
);
109 cur_state
= (*byte2
>> bit
) & GFS2_BIT_MASK
;
110 *byte2
^= (cur_state
^ new_state
) << bit
;
115 * gfs2_testbit - test a bit in the bitmaps
116 * @rbm: The bit to test
118 * Returns: The two bit block state of the requested bit
121 static inline u8
gfs2_testbit(const struct gfs2_rbm
*rbm
)
123 const u8
*buffer
= rbm
->bi
->bi_bh
->b_data
+ rbm
->bi
->bi_offset
;
127 byte
= buffer
+ (rbm
->offset
/ GFS2_NBBY
);
128 bit
= (rbm
->offset
% GFS2_NBBY
) * GFS2_BIT_SIZE
;
130 return (*byte
>> bit
) & GFS2_BIT_MASK
;
135 * @ptr: Pointer to bitmap data
136 * @mask: Mask to use (normally 0x55555.... but adjusted for search start)
137 * @state: The state we are searching for
139 * We xor the bitmap data with a patter which is the bitwise opposite
140 * of what we are looking for, this gives rise to a pattern of ones
141 * wherever there is a match. Since we have two bits per entry, we
142 * take this pattern, shift it down by one place and then and it with
143 * the original. All the even bit positions (0,2,4, etc) then represent
144 * successful matches, so we mask with 0x55555..... to remove the unwanted
147 * This allows searching of a whole u64 at once (32 blocks) with a
148 * single test (on 64 bit arches).
151 static inline u64
gfs2_bit_search(const __le64
*ptr
, u64 mask
, u8 state
)
154 static const u64 search
[] = {
155 [0] = 0xffffffffffffffffULL
,
156 [1] = 0xaaaaaaaaaaaaaaaaULL
,
157 [2] = 0x5555555555555555ULL
,
158 [3] = 0x0000000000000000ULL
,
160 tmp
= le64_to_cpu(*ptr
) ^ search
[state
];
167 * rs_cmp - multi-block reservation range compare
168 * @blk: absolute file system block number of the new reservation
169 * @len: number of blocks in the new reservation
170 * @rs: existing reservation to compare against
172 * returns: 1 if the block range is beyond the reach of the reservation
173 * -1 if the block range is before the start of the reservation
174 * 0 if the block range overlaps with the reservation
176 static inline int rs_cmp(u64 blk
, u32 len
, struct gfs2_blkreserv
*rs
)
178 u64 startblk
= gfs2_rbm_to_block(&rs
->rs_rbm
);
180 if (blk
>= startblk
+ rs
->rs_free
)
182 if (blk
+ len
- 1 < startblk
)
188 * gfs2_bitfit - Search an rgrp's bitmap buffer to find a bit-pair representing
189 * a block in a given allocation state.
190 * @buf: the buffer that holds the bitmaps
191 * @len: the length (in bytes) of the buffer
192 * @goal: start search at this block's bit-pair (within @buffer)
193 * @state: GFS2_BLKST_XXX the state of the block we're looking for.
195 * Scope of @goal and returned block number is only within this bitmap buffer,
196 * not entire rgrp or filesystem. @buffer will be offset from the actual
197 * beginning of a bitmap block buffer, skipping any header structures, but
198 * headers are always a multiple of 64 bits long so that the buffer is
199 * always aligned to a 64 bit boundary.
201 * The size of the buffer is in bytes, but is it assumed that it is
202 * always ok to read a complete multiple of 64 bits at the end
203 * of the block in case the end is no aligned to a natural boundary.
205 * Return: the block number (bitmap buffer scope) that was found
208 static u32
gfs2_bitfit(const u8
*buf
, const unsigned int len
,
211 u32 spoint
= (goal
<< 1) & ((8*sizeof(u64
)) - 1);
212 const __le64
*ptr
= ((__le64
*)buf
) + (goal
>> 5);
213 const __le64
*end
= (__le64
*)(buf
+ ALIGN(len
, sizeof(u64
)));
215 u64 mask
= 0x5555555555555555ULL
;
218 /* Mask off bits we don't care about at the start of the search */
220 tmp
= gfs2_bit_search(ptr
, mask
, state
);
222 while(tmp
== 0 && ptr
< end
) {
223 tmp
= gfs2_bit_search(ptr
, 0x5555555555555555ULL
, state
);
226 /* Mask off any bits which are more than len bytes from the start */
227 if (ptr
== end
&& (len
& (sizeof(u64
) - 1)))
228 tmp
&= (((u64
)~0) >> (64 - 8*(len
& (sizeof(u64
) - 1))));
229 /* Didn't find anything, so return */
234 bit
/= 2; /* two bits per entry in the bitmap */
235 return (((const unsigned char *)ptr
- buf
) * GFS2_NBBY
) + bit
;
239 * gfs2_rbm_from_block - Set the rbm based upon rgd and block number
240 * @rbm: The rbm with rgd already set correctly
241 * @block: The block number (filesystem relative)
243 * This sets the bi and offset members of an rbm based on a
244 * resource group and a filesystem relative block number. The
245 * resource group must be set in the rbm on entry, the bi and
246 * offset members will be set by this function.
248 * Returns: 0 on success, or an error code
251 static int gfs2_rbm_from_block(struct gfs2_rbm
*rbm
, u64 block
)
253 u64 rblock
= block
- rbm
->rgd
->rd_data0
;
254 u32 goal
= (u32
)rblock
;
257 if (WARN_ON_ONCE(rblock
> UINT_MAX
))
259 if (block
>= rbm
->rgd
->rd_data0
+ rbm
->rgd
->rd_data
)
262 for (x
= 0; x
< rbm
->rgd
->rd_length
; x
++) {
263 rbm
->bi
= rbm
->rgd
->rd_bits
+ x
;
264 if (goal
< (rbm
->bi
->bi_start
+ rbm
->bi
->bi_len
) * GFS2_NBBY
) {
265 rbm
->offset
= goal
- (rbm
->bi
->bi_start
* GFS2_NBBY
);
274 * gfs2_unaligned_extlen - Look for free blocks which are not byte aligned
275 * @rbm: Position to search (value/result)
276 * @n_unaligned: Number of unaligned blocks to check
277 * @len: Decremented for each block found (terminate on zero)
279 * Returns: true if a non-free block is encountered
282 static bool gfs2_unaligned_extlen(struct gfs2_rbm
*rbm
, u32 n_unaligned
, u32
*len
)
288 for (n
= 0; n
< n_unaligned
; n
++) {
289 res
= gfs2_testbit(rbm
);
290 if (res
!= GFS2_BLKST_FREE
)
295 block
= gfs2_rbm_to_block(rbm
);
296 if (gfs2_rbm_from_block(rbm
, block
+ 1))
304 * gfs2_free_extlen - Return extent length of free blocks
305 * @rbm: Starting position
306 * @len: Max length to check
308 * Starting at the block specified by the rbm, see how many free blocks
309 * there are, not reading more than len blocks ahead. This can be done
310 * using memchr_inv when the blocks are byte aligned, but has to be done
311 * on a block by block basis in case of unaligned blocks. Also this
312 * function can cope with bitmap boundaries (although it must stop on
313 * a resource group boundary)
315 * Returns: Number of free blocks in the extent
318 static u32
gfs2_free_extlen(const struct gfs2_rbm
*rrbm
, u32 len
)
320 struct gfs2_rbm rbm
= *rrbm
;
321 u32 n_unaligned
= rbm
.offset
& 3;
325 u8
*ptr
, *start
, *end
;
329 gfs2_unaligned_extlen(&rbm
, 4 - n_unaligned
, &len
))
332 n_unaligned
= len
& 3;
333 /* Start is now byte aligned */
335 start
= rbm
.bi
->bi_bh
->b_data
;
336 if (rbm
.bi
->bi_clone
)
337 start
= rbm
.bi
->bi_clone
;
338 end
= start
+ rbm
.bi
->bi_bh
->b_size
;
339 start
+= rbm
.bi
->bi_offset
;
340 BUG_ON(rbm
.offset
& 3);
341 start
+= (rbm
.offset
/ GFS2_NBBY
);
342 bytes
= min_t(u32
, len
/ GFS2_NBBY
, (end
- start
));
343 ptr
= memchr_inv(start
, 0, bytes
);
344 chunk_size
= ((ptr
== NULL
) ? bytes
: (ptr
- start
));
345 chunk_size
*= GFS2_NBBY
;
346 BUG_ON(len
< chunk_size
);
348 block
= gfs2_rbm_to_block(&rbm
);
349 gfs2_rbm_from_block(&rbm
, block
+ chunk_size
);
353 n_unaligned
= len
& 3;
356 /* Deal with any bits left over at the end */
358 gfs2_unaligned_extlen(&rbm
, n_unaligned
, &len
);
364 * gfs2_bitcount - count the number of bits in a certain state
365 * @rgd: the resource group descriptor
366 * @buffer: the buffer that holds the bitmaps
367 * @buflen: the length (in bytes) of the buffer
368 * @state: the state of the block we're looking for
370 * Returns: The number of bits
373 static u32
gfs2_bitcount(struct gfs2_rgrpd
*rgd
, const u8
*buffer
,
374 unsigned int buflen
, u8 state
)
376 const u8
*byte
= buffer
;
377 const u8
*end
= buffer
+ buflen
;
378 const u8 state1
= state
<< 2;
379 const u8 state2
= state
<< 4;
380 const u8 state3
= state
<< 6;
383 for (; byte
< end
; byte
++) {
384 if (((*byte
) & 0x03) == state
)
386 if (((*byte
) & 0x0C) == state1
)
388 if (((*byte
) & 0x30) == state2
)
390 if (((*byte
) & 0xC0) == state3
)
398 * gfs2_rgrp_verify - Verify that a resource group is consistent
403 void gfs2_rgrp_verify(struct gfs2_rgrpd
*rgd
)
405 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
406 struct gfs2_bitmap
*bi
= NULL
;
407 u32 length
= rgd
->rd_length
;
411 memset(count
, 0, 4 * sizeof(u32
));
413 /* Count # blocks in each of 4 possible allocation states */
414 for (buf
= 0; buf
< length
; buf
++) {
415 bi
= rgd
->rd_bits
+ buf
;
416 for (x
= 0; x
< 4; x
++)
417 count
[x
] += gfs2_bitcount(rgd
,
423 if (count
[0] != rgd
->rd_free
) {
424 if (gfs2_consist_rgrpd(rgd
))
425 fs_err(sdp
, "free data mismatch: %u != %u\n",
426 count
[0], rgd
->rd_free
);
430 tmp
= rgd
->rd_data
- rgd
->rd_free
- rgd
->rd_dinodes
;
431 if (count
[1] != tmp
) {
432 if (gfs2_consist_rgrpd(rgd
))
433 fs_err(sdp
, "used data mismatch: %u != %u\n",
438 if (count
[2] + count
[3] != rgd
->rd_dinodes
) {
439 if (gfs2_consist_rgrpd(rgd
))
440 fs_err(sdp
, "used metadata mismatch: %u != %u\n",
441 count
[2] + count
[3], rgd
->rd_dinodes
);
446 static inline int rgrp_contains_block(struct gfs2_rgrpd
*rgd
, u64 block
)
448 u64 first
= rgd
->rd_data0
;
449 u64 last
= first
+ rgd
->rd_data
;
450 return first
<= block
&& block
< last
;
454 * gfs2_blk2rgrpd - Find resource group for a given data/meta block number
455 * @sdp: The GFS2 superblock
456 * @blk: The data block number
457 * @exact: True if this needs to be an exact match
459 * Returns: The resource group, or NULL if not found
462 struct gfs2_rgrpd
*gfs2_blk2rgrpd(struct gfs2_sbd
*sdp
, u64 blk
, bool exact
)
464 struct rb_node
*n
, *next
;
465 struct gfs2_rgrpd
*cur
;
467 spin_lock(&sdp
->sd_rindex_spin
);
468 n
= sdp
->sd_rindex_tree
.rb_node
;
470 cur
= rb_entry(n
, struct gfs2_rgrpd
, rd_node
);
472 if (blk
< cur
->rd_addr
)
474 else if (blk
>= cur
->rd_data0
+ cur
->rd_data
)
477 spin_unlock(&sdp
->sd_rindex_spin
);
479 if (blk
< cur
->rd_addr
)
481 if (blk
>= cur
->rd_data0
+ cur
->rd_data
)
488 spin_unlock(&sdp
->sd_rindex_spin
);
494 * gfs2_rgrpd_get_first - get the first Resource Group in the filesystem
495 * @sdp: The GFS2 superblock
497 * Returns: The first rgrp in the filesystem
500 struct gfs2_rgrpd
*gfs2_rgrpd_get_first(struct gfs2_sbd
*sdp
)
502 const struct rb_node
*n
;
503 struct gfs2_rgrpd
*rgd
;
505 spin_lock(&sdp
->sd_rindex_spin
);
506 n
= rb_first(&sdp
->sd_rindex_tree
);
507 rgd
= rb_entry(n
, struct gfs2_rgrpd
, rd_node
);
508 spin_unlock(&sdp
->sd_rindex_spin
);
514 * gfs2_rgrpd_get_next - get the next RG
515 * @rgd: the resource group descriptor
517 * Returns: The next rgrp
520 struct gfs2_rgrpd
*gfs2_rgrpd_get_next(struct gfs2_rgrpd
*rgd
)
522 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
523 const struct rb_node
*n
;
525 spin_lock(&sdp
->sd_rindex_spin
);
526 n
= rb_next(&rgd
->rd_node
);
528 n
= rb_first(&sdp
->sd_rindex_tree
);
530 if (unlikely(&rgd
->rd_node
== n
)) {
531 spin_unlock(&sdp
->sd_rindex_spin
);
534 rgd
= rb_entry(n
, struct gfs2_rgrpd
, rd_node
);
535 spin_unlock(&sdp
->sd_rindex_spin
);
539 void gfs2_free_clones(struct gfs2_rgrpd
*rgd
)
543 for (x
= 0; x
< rgd
->rd_length
; x
++) {
544 struct gfs2_bitmap
*bi
= rgd
->rd_bits
+ x
;
551 * gfs2_rs_alloc - make sure we have a reservation assigned to the inode
552 * @ip: the inode for this reservation
554 int gfs2_rs_alloc(struct gfs2_inode
*ip
)
557 struct gfs2_blkreserv
*res
;
562 res
= kmem_cache_zalloc(gfs2_rsrv_cachep
, GFP_NOFS
);
566 RB_CLEAR_NODE(&res
->rs_node
);
568 down_write(&ip
->i_rw_mutex
);
570 kmem_cache_free(gfs2_rsrv_cachep
, res
);
573 up_write(&ip
->i_rw_mutex
);
577 static void dump_rs(struct seq_file
*seq
, const struct gfs2_blkreserv
*rs
)
579 gfs2_print_dbg(seq
, " B: n:%llu s:%llu b:%u f:%u\n",
580 (unsigned long long)rs
->rs_inum
,
581 (unsigned long long)gfs2_rbm_to_block(&rs
->rs_rbm
),
582 rs
->rs_rbm
.offset
, rs
->rs_free
);
586 * __rs_deltree - remove a multi-block reservation from the rgd tree
587 * @rs: The reservation to remove
590 static void __rs_deltree(struct gfs2_inode
*ip
, struct gfs2_blkreserv
*rs
)
592 struct gfs2_rgrpd
*rgd
;
594 if (!gfs2_rs_active(rs
))
597 rgd
= rs
->rs_rbm
.rgd
;
598 trace_gfs2_rs(rs
, TRACE_RS_TREEDEL
);
599 rb_erase(&rs
->rs_node
, &rgd
->rd_rstree
);
600 RB_CLEAR_NODE(&rs
->rs_node
);
603 /* return reserved blocks to the rgrp and the ip */
604 BUG_ON(rs
->rs_rbm
.rgd
->rd_reserved
< rs
->rs_free
);
605 rs
->rs_rbm
.rgd
->rd_reserved
-= rs
->rs_free
;
607 clear_bit(GBF_FULL
, &rs
->rs_rbm
.bi
->bi_flags
);
608 smp_mb__after_clear_bit();
613 * gfs2_rs_deltree - remove a multi-block reservation from the rgd tree
614 * @rs: The reservation to remove
617 void gfs2_rs_deltree(struct gfs2_inode
*ip
, struct gfs2_blkreserv
*rs
)
619 struct gfs2_rgrpd
*rgd
;
621 rgd
= rs
->rs_rbm
.rgd
;
623 spin_lock(&rgd
->rd_rsspin
);
624 __rs_deltree(ip
, rs
);
625 spin_unlock(&rgd
->rd_rsspin
);
630 * gfs2_rs_delete - delete a multi-block reservation
631 * @ip: The inode for this reservation
634 void gfs2_rs_delete(struct gfs2_inode
*ip
)
636 down_write(&ip
->i_rw_mutex
);
638 gfs2_rs_deltree(ip
, ip
->i_res
);
639 BUG_ON(ip
->i_res
->rs_free
);
640 kmem_cache_free(gfs2_rsrv_cachep
, ip
->i_res
);
643 up_write(&ip
->i_rw_mutex
);
647 * return_all_reservations - return all reserved blocks back to the rgrp.
648 * @rgd: the rgrp that needs its space back
650 * We previously reserved a bunch of blocks for allocation. Now we need to
651 * give them back. This leave the reservation structures in tact, but removes
652 * all of their corresponding "no-fly zones".
654 static void return_all_reservations(struct gfs2_rgrpd
*rgd
)
657 struct gfs2_blkreserv
*rs
;
659 spin_lock(&rgd
->rd_rsspin
);
660 while ((n
= rb_first(&rgd
->rd_rstree
))) {
661 rs
= rb_entry(n
, struct gfs2_blkreserv
, rs_node
);
662 __rs_deltree(NULL
, rs
);
664 spin_unlock(&rgd
->rd_rsspin
);
667 void gfs2_clear_rgrpd(struct gfs2_sbd
*sdp
)
670 struct gfs2_rgrpd
*rgd
;
671 struct gfs2_glock
*gl
;
673 while ((n
= rb_first(&sdp
->sd_rindex_tree
))) {
674 rgd
= rb_entry(n
, struct gfs2_rgrpd
, rd_node
);
677 rb_erase(n
, &sdp
->sd_rindex_tree
);
680 spin_lock(&gl
->gl_spin
);
681 gl
->gl_object
= NULL
;
682 spin_unlock(&gl
->gl_spin
);
683 gfs2_glock_add_to_lru(gl
);
687 gfs2_free_clones(rgd
);
689 return_all_reservations(rgd
);
690 kmem_cache_free(gfs2_rgrpd_cachep
, rgd
);
694 static void gfs2_rindex_print(const struct gfs2_rgrpd
*rgd
)
696 printk(KERN_INFO
" ri_addr = %llu\n", (unsigned long long)rgd
->rd_addr
);
697 printk(KERN_INFO
" ri_length = %u\n", rgd
->rd_length
);
698 printk(KERN_INFO
" ri_data0 = %llu\n", (unsigned long long)rgd
->rd_data0
);
699 printk(KERN_INFO
" ri_data = %u\n", rgd
->rd_data
);
700 printk(KERN_INFO
" ri_bitbytes = %u\n", rgd
->rd_bitbytes
);
704 * gfs2_compute_bitstructs - Compute the bitmap sizes
705 * @rgd: The resource group descriptor
707 * Calculates bitmap descriptors, one for each block that contains bitmap data
712 static int compute_bitstructs(struct gfs2_rgrpd
*rgd
)
714 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
715 struct gfs2_bitmap
*bi
;
716 u32 length
= rgd
->rd_length
; /* # blocks in hdr & bitmap */
717 u32 bytes_left
, bytes
;
723 rgd
->rd_bits
= kcalloc(length
, sizeof(struct gfs2_bitmap
), GFP_NOFS
);
727 bytes_left
= rgd
->rd_bitbytes
;
729 for (x
= 0; x
< length
; x
++) {
730 bi
= rgd
->rd_bits
+ x
;
733 /* small rgrp; bitmap stored completely in header block */
736 bi
->bi_offset
= sizeof(struct gfs2_rgrp
);
741 bytes
= sdp
->sd_sb
.sb_bsize
- sizeof(struct gfs2_rgrp
);
742 bi
->bi_offset
= sizeof(struct gfs2_rgrp
);
746 } else if (x
+ 1 == length
) {
748 bi
->bi_offset
= sizeof(struct gfs2_meta_header
);
749 bi
->bi_start
= rgd
->rd_bitbytes
- bytes_left
;
753 bytes
= sdp
->sd_sb
.sb_bsize
-
754 sizeof(struct gfs2_meta_header
);
755 bi
->bi_offset
= sizeof(struct gfs2_meta_header
);
756 bi
->bi_start
= rgd
->rd_bitbytes
- bytes_left
;
764 gfs2_consist_rgrpd(rgd
);
767 bi
= rgd
->rd_bits
+ (length
- 1);
768 if ((bi
->bi_start
+ bi
->bi_len
) * GFS2_NBBY
!= rgd
->rd_data
) {
769 if (gfs2_consist_rgrpd(rgd
)) {
770 gfs2_rindex_print(rgd
);
771 fs_err(sdp
, "start=%u len=%u offset=%u\n",
772 bi
->bi_start
, bi
->bi_len
, bi
->bi_offset
);
781 * gfs2_ri_total - Total up the file system space, according to the rindex.
782 * @sdp: the filesystem
785 u64
gfs2_ri_total(struct gfs2_sbd
*sdp
)
788 struct inode
*inode
= sdp
->sd_rindex
;
789 struct gfs2_inode
*ip
= GFS2_I(inode
);
790 char buf
[sizeof(struct gfs2_rindex
)];
793 for (rgrps
= 0;; rgrps
++) {
794 loff_t pos
= rgrps
* sizeof(struct gfs2_rindex
);
796 if (pos
+ sizeof(struct gfs2_rindex
) > i_size_read(inode
))
798 error
= gfs2_internal_read(ip
, buf
, &pos
,
799 sizeof(struct gfs2_rindex
));
800 if (error
!= sizeof(struct gfs2_rindex
))
802 total_data
+= be32_to_cpu(((struct gfs2_rindex
*)buf
)->ri_data
);
807 static int rgd_insert(struct gfs2_rgrpd
*rgd
)
809 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
810 struct rb_node
**newn
= &sdp
->sd_rindex_tree
.rb_node
, *parent
= NULL
;
812 /* Figure out where to put new node */
814 struct gfs2_rgrpd
*cur
= rb_entry(*newn
, struct gfs2_rgrpd
,
818 if (rgd
->rd_addr
< cur
->rd_addr
)
819 newn
= &((*newn
)->rb_left
);
820 else if (rgd
->rd_addr
> cur
->rd_addr
)
821 newn
= &((*newn
)->rb_right
);
826 rb_link_node(&rgd
->rd_node
, parent
, newn
);
827 rb_insert_color(&rgd
->rd_node
, &sdp
->sd_rindex_tree
);
833 * read_rindex_entry - Pull in a new resource index entry from the disk
834 * @ip: Pointer to the rindex inode
836 * Returns: 0 on success, > 0 on EOF, error code otherwise
839 static int read_rindex_entry(struct gfs2_inode
*ip
)
841 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
842 loff_t pos
= sdp
->sd_rgrps
* sizeof(struct gfs2_rindex
);
843 struct gfs2_rindex buf
;
845 struct gfs2_rgrpd
*rgd
;
847 if (pos
>= i_size_read(&ip
->i_inode
))
850 error
= gfs2_internal_read(ip
, (char *)&buf
, &pos
,
851 sizeof(struct gfs2_rindex
));
853 if (error
!= sizeof(struct gfs2_rindex
))
854 return (error
== 0) ? 1 : error
;
856 rgd
= kmem_cache_zalloc(gfs2_rgrpd_cachep
, GFP_NOFS
);
862 rgd
->rd_addr
= be64_to_cpu(buf
.ri_addr
);
863 rgd
->rd_length
= be32_to_cpu(buf
.ri_length
);
864 rgd
->rd_data0
= be64_to_cpu(buf
.ri_data0
);
865 rgd
->rd_data
= be32_to_cpu(buf
.ri_data
);
866 rgd
->rd_bitbytes
= be32_to_cpu(buf
.ri_bitbytes
);
867 spin_lock_init(&rgd
->rd_rsspin
);
869 error
= compute_bitstructs(rgd
);
873 error
= gfs2_glock_get(sdp
, rgd
->rd_addr
,
874 &gfs2_rgrp_glops
, CREATE
, &rgd
->rd_gl
);
878 rgd
->rd_gl
->gl_object
= rgd
;
879 rgd
->rd_rgl
= (struct gfs2_rgrp_lvb
*)rgd
->rd_gl
->gl_lvb
;
880 rgd
->rd_flags
&= ~GFS2_RDF_UPTODATE
;
881 if (rgd
->rd_data
> sdp
->sd_max_rg_data
)
882 sdp
->sd_max_rg_data
= rgd
->rd_data
;
883 spin_lock(&sdp
->sd_rindex_spin
);
884 error
= rgd_insert(rgd
);
885 spin_unlock(&sdp
->sd_rindex_spin
);
889 error
= 0; /* someone else read in the rgrp; free it and ignore it */
890 gfs2_glock_put(rgd
->rd_gl
);
894 kmem_cache_free(gfs2_rgrpd_cachep
, rgd
);
899 * gfs2_ri_update - Pull in a new resource index from the disk
900 * @ip: pointer to the rindex inode
902 * Returns: 0 on successful update, error code otherwise
905 static int gfs2_ri_update(struct gfs2_inode
*ip
)
907 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
911 error
= read_rindex_entry(ip
);
912 } while (error
== 0);
917 sdp
->sd_rindex_uptodate
= 1;
922 * gfs2_rindex_update - Update the rindex if required
923 * @sdp: The GFS2 superblock
925 * We grab a lock on the rindex inode to make sure that it doesn't
926 * change whilst we are performing an operation. We keep this lock
927 * for quite long periods of time compared to other locks. This
928 * doesn't matter, since it is shared and it is very, very rarely
929 * accessed in the exclusive mode (i.e. only when expanding the filesystem).
931 * This makes sure that we're using the latest copy of the resource index
932 * special file, which might have been updated if someone expanded the
933 * filesystem (via gfs2_grow utility), which adds new resource groups.
935 * Returns: 0 on succeess, error code otherwise
938 int gfs2_rindex_update(struct gfs2_sbd
*sdp
)
940 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_rindex
);
941 struct gfs2_glock
*gl
= ip
->i_gl
;
942 struct gfs2_holder ri_gh
;
944 int unlock_required
= 0;
946 /* Read new copy from disk if we don't have the latest */
947 if (!sdp
->sd_rindex_uptodate
) {
948 if (!gfs2_glock_is_locked_by_me(gl
)) {
949 error
= gfs2_glock_nq_init(gl
, LM_ST_SHARED
, 0, &ri_gh
);
954 if (!sdp
->sd_rindex_uptodate
)
955 error
= gfs2_ri_update(ip
);
957 gfs2_glock_dq_uninit(&ri_gh
);
963 static void gfs2_rgrp_in(struct gfs2_rgrpd
*rgd
, const void *buf
)
965 const struct gfs2_rgrp
*str
= buf
;
968 rg_flags
= be32_to_cpu(str
->rg_flags
);
969 rg_flags
&= ~GFS2_RDF_MASK
;
970 rgd
->rd_flags
&= GFS2_RDF_MASK
;
971 rgd
->rd_flags
|= rg_flags
;
972 rgd
->rd_free
= be32_to_cpu(str
->rg_free
);
973 rgd
->rd_dinodes
= be32_to_cpu(str
->rg_dinodes
);
974 rgd
->rd_igeneration
= be64_to_cpu(str
->rg_igeneration
);
977 static void gfs2_rgrp_out(struct gfs2_rgrpd
*rgd
, void *buf
)
979 struct gfs2_rgrp
*str
= buf
;
981 str
->rg_flags
= cpu_to_be32(rgd
->rd_flags
& ~GFS2_RDF_MASK
);
982 str
->rg_free
= cpu_to_be32(rgd
->rd_free
);
983 str
->rg_dinodes
= cpu_to_be32(rgd
->rd_dinodes
);
984 str
->__pad
= cpu_to_be32(0);
985 str
->rg_igeneration
= cpu_to_be64(rgd
->rd_igeneration
);
986 memset(&str
->rg_reserved
, 0, sizeof(str
->rg_reserved
));
989 static int gfs2_rgrp_lvb_valid(struct gfs2_rgrpd
*rgd
)
991 struct gfs2_rgrp_lvb
*rgl
= rgd
->rd_rgl
;
992 struct gfs2_rgrp
*str
= (struct gfs2_rgrp
*)rgd
->rd_bits
[0].bi_bh
->b_data
;
994 if (rgl
->rl_flags
!= str
->rg_flags
|| rgl
->rl_free
!= str
->rg_free
||
995 rgl
->rl_dinodes
!= str
->rg_dinodes
||
996 rgl
->rl_igeneration
!= str
->rg_igeneration
)
1001 static void gfs2_rgrp_ondisk2lvb(struct gfs2_rgrp_lvb
*rgl
, const void *buf
)
1003 const struct gfs2_rgrp
*str
= buf
;
1005 rgl
->rl_magic
= cpu_to_be32(GFS2_MAGIC
);
1006 rgl
->rl_flags
= str
->rg_flags
;
1007 rgl
->rl_free
= str
->rg_free
;
1008 rgl
->rl_dinodes
= str
->rg_dinodes
;
1009 rgl
->rl_igeneration
= str
->rg_igeneration
;
1013 static void update_rgrp_lvb_unlinked(struct gfs2_rgrpd
*rgd
, u32 change
)
1015 struct gfs2_rgrp_lvb
*rgl
= rgd
->rd_rgl
;
1016 u32 unlinked
= be32_to_cpu(rgl
->rl_unlinked
) + change
;
1017 rgl
->rl_unlinked
= cpu_to_be32(unlinked
);
1020 static u32
count_unlinked(struct gfs2_rgrpd
*rgd
)
1022 struct gfs2_bitmap
*bi
;
1023 const u32 length
= rgd
->rd_length
;
1024 const u8
*buffer
= NULL
;
1025 u32 i
, goal
, count
= 0;
1027 for (i
= 0, bi
= rgd
->rd_bits
; i
< length
; i
++, bi
++) {
1029 buffer
= bi
->bi_bh
->b_data
+ bi
->bi_offset
;
1030 WARN_ON(!buffer_uptodate(bi
->bi_bh
));
1031 while (goal
< bi
->bi_len
* GFS2_NBBY
) {
1032 goal
= gfs2_bitfit(buffer
, bi
->bi_len
, goal
,
1033 GFS2_BLKST_UNLINKED
);
1034 if (goal
== BFITNOENT
)
1046 * gfs2_rgrp_bh_get - Read in a RG's header and bitmaps
1047 * @rgd: the struct gfs2_rgrpd describing the RG to read in
1049 * Read in all of a Resource Group's header and bitmap blocks.
1050 * Caller must eventually call gfs2_rgrp_relse() to free the bitmaps.
1055 int gfs2_rgrp_bh_get(struct gfs2_rgrpd
*rgd
)
1057 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
1058 struct gfs2_glock
*gl
= rgd
->rd_gl
;
1059 unsigned int length
= rgd
->rd_length
;
1060 struct gfs2_bitmap
*bi
;
1064 if (rgd
->rd_bits
[0].bi_bh
!= NULL
)
1067 for (x
= 0; x
< length
; x
++) {
1068 bi
= rgd
->rd_bits
+ x
;
1069 error
= gfs2_meta_read(gl
, rgd
->rd_addr
+ x
, 0, &bi
->bi_bh
);
1074 for (y
= length
; y
--;) {
1075 bi
= rgd
->rd_bits
+ y
;
1076 error
= gfs2_meta_wait(sdp
, bi
->bi_bh
);
1079 if (gfs2_metatype_check(sdp
, bi
->bi_bh
, y
? GFS2_METATYPE_RB
:
1080 GFS2_METATYPE_RG
)) {
1086 if (!(rgd
->rd_flags
& GFS2_RDF_UPTODATE
)) {
1087 for (x
= 0; x
< length
; x
++)
1088 clear_bit(GBF_FULL
, &rgd
->rd_bits
[x
].bi_flags
);
1089 gfs2_rgrp_in(rgd
, (rgd
->rd_bits
[0].bi_bh
)->b_data
);
1090 rgd
->rd_flags
|= (GFS2_RDF_UPTODATE
| GFS2_RDF_CHECK
);
1091 rgd
->rd_free_clone
= rgd
->rd_free
;
1093 if (be32_to_cpu(GFS2_MAGIC
) != rgd
->rd_rgl
->rl_magic
) {
1094 rgd
->rd_rgl
->rl_unlinked
= cpu_to_be32(count_unlinked(rgd
));
1095 gfs2_rgrp_ondisk2lvb(rgd
->rd_rgl
,
1096 rgd
->rd_bits
[0].bi_bh
->b_data
);
1098 else if (sdp
->sd_args
.ar_rgrplvb
) {
1099 if (!gfs2_rgrp_lvb_valid(rgd
)){
1100 gfs2_consist_rgrpd(rgd
);
1104 if (rgd
->rd_rgl
->rl_unlinked
== 0)
1105 rgd
->rd_flags
&= ~GFS2_RDF_CHECK
;
1111 bi
= rgd
->rd_bits
+ x
;
1114 gfs2_assert_warn(sdp
, !bi
->bi_clone
);
1120 int update_rgrp_lvb(struct gfs2_rgrpd
*rgd
)
1124 if (rgd
->rd_flags
& GFS2_RDF_UPTODATE
)
1127 if (be32_to_cpu(GFS2_MAGIC
) != rgd
->rd_rgl
->rl_magic
)
1128 return gfs2_rgrp_bh_get(rgd
);
1130 rl_flags
= be32_to_cpu(rgd
->rd_rgl
->rl_flags
);
1131 rl_flags
&= ~GFS2_RDF_MASK
;
1132 rgd
->rd_flags
&= GFS2_RDF_MASK
;
1133 rgd
->rd_flags
|= (rl_flags
| GFS2_RDF_UPTODATE
| GFS2_RDF_CHECK
);
1134 if (rgd
->rd_rgl
->rl_unlinked
== 0)
1135 rgd
->rd_flags
&= ~GFS2_RDF_CHECK
;
1136 rgd
->rd_free
= be32_to_cpu(rgd
->rd_rgl
->rl_free
);
1137 rgd
->rd_free_clone
= rgd
->rd_free
;
1138 rgd
->rd_dinodes
= be32_to_cpu(rgd
->rd_rgl
->rl_dinodes
);
1139 rgd
->rd_igeneration
= be64_to_cpu(rgd
->rd_rgl
->rl_igeneration
);
1143 int gfs2_rgrp_go_lock(struct gfs2_holder
*gh
)
1145 struct gfs2_rgrpd
*rgd
= gh
->gh_gl
->gl_object
;
1146 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
1148 if (gh
->gh_flags
& GL_SKIP
&& sdp
->sd_args
.ar_rgrplvb
)
1150 return gfs2_rgrp_bh_get((struct gfs2_rgrpd
*)gh
->gh_gl
->gl_object
);
1154 * gfs2_rgrp_go_unlock - Release RG bitmaps read in with gfs2_rgrp_bh_get()
1155 * @gh: The glock holder for the resource group
1159 void gfs2_rgrp_go_unlock(struct gfs2_holder
*gh
)
1161 struct gfs2_rgrpd
*rgd
= gh
->gh_gl
->gl_object
;
1162 int x
, length
= rgd
->rd_length
;
1164 for (x
= 0; x
< length
; x
++) {
1165 struct gfs2_bitmap
*bi
= rgd
->rd_bits
+ x
;
1174 int gfs2_rgrp_send_discards(struct gfs2_sbd
*sdp
, u64 offset
,
1175 struct buffer_head
*bh
,
1176 const struct gfs2_bitmap
*bi
, unsigned minlen
, u64
*ptrimmed
)
1178 struct super_block
*sb
= sdp
->sd_vfs
;
1179 struct block_device
*bdev
= sb
->s_bdev
;
1180 const unsigned int sects_per_blk
= sdp
->sd_sb
.sb_bsize
/
1181 bdev_logical_block_size(sb
->s_bdev
);
1184 sector_t nr_sects
= 0;
1190 for (x
= 0; x
< bi
->bi_len
; x
++) {
1191 const u8
*clone
= bi
->bi_clone
? bi
->bi_clone
: bi
->bi_bh
->b_data
;
1192 clone
+= bi
->bi_offset
;
1195 const u8
*orig
= bh
->b_data
+ bi
->bi_offset
+ x
;
1196 diff
= ~(*orig
| (*orig
>> 1)) & (*clone
| (*clone
>> 1));
1198 diff
= ~(*clone
| (*clone
>> 1));
1203 blk
= offset
+ ((bi
->bi_start
+ x
) * GFS2_NBBY
);
1204 blk
*= sects_per_blk
; /* convert to sectors */
1208 goto start_new_extent
;
1209 if ((start
+ nr_sects
) != blk
) {
1210 if (nr_sects
>= minlen
) {
1211 rv
= blkdev_issue_discard(bdev
,
1216 trimmed
+= nr_sects
;
1222 nr_sects
+= sects_per_blk
;
1225 blk
+= sects_per_blk
;
1228 if (nr_sects
>= minlen
) {
1229 rv
= blkdev_issue_discard(bdev
, start
, nr_sects
, GFP_NOFS
, 0);
1232 trimmed
+= nr_sects
;
1235 *ptrimmed
= trimmed
;
1239 if (sdp
->sd_args
.ar_discard
)
1240 fs_warn(sdp
, "error %d on discard request, turning discards off for this filesystem", rv
);
1241 sdp
->sd_args
.ar_discard
= 0;
1246 * gfs2_fitrim - Generate discard requests for unused bits of the filesystem
1247 * @filp: Any file on the filesystem
1248 * @argp: Pointer to the arguments (also used to pass result)
1250 * Returns: 0 on success, otherwise error code
1253 int gfs2_fitrim(struct file
*filp
, void __user
*argp
)
1255 struct inode
*inode
= filp
->f_dentry
->d_inode
;
1256 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
1257 struct request_queue
*q
= bdev_get_queue(sdp
->sd_vfs
->s_bdev
);
1258 struct buffer_head
*bh
;
1259 struct gfs2_rgrpd
*rgd
;
1260 struct gfs2_rgrpd
*rgd_end
;
1261 struct gfs2_holder gh
;
1262 struct fstrim_range r
;
1268 if (!capable(CAP_SYS_ADMIN
))
1271 if (!blk_queue_discard(q
))
1278 } else if (copy_from_user(&r
, argp
, sizeof(r
)))
1281 ret
= gfs2_rindex_update(sdp
);
1285 rgd
= gfs2_blk2rgrpd(sdp
, r
.start
, 0);
1286 rgd_end
= gfs2_blk2rgrpd(sdp
, r
.start
+ r
.len
, 0);
1290 ret
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
1294 if (!(rgd
->rd_flags
& GFS2_RGF_TRIMMED
)) {
1295 /* Trim each bitmap in the rgrp */
1296 for (x
= 0; x
< rgd
->rd_length
; x
++) {
1297 struct gfs2_bitmap
*bi
= rgd
->rd_bits
+ x
;
1298 ret
= gfs2_rgrp_send_discards(sdp
, rgd
->rd_data0
, NULL
, bi
, r
.minlen
, &amt
);
1300 gfs2_glock_dq_uninit(&gh
);
1306 /* Mark rgrp as having been trimmed */
1307 ret
= gfs2_trans_begin(sdp
, RES_RG_HDR
, 0);
1309 bh
= rgd
->rd_bits
[0].bi_bh
;
1310 rgd
->rd_flags
|= GFS2_RGF_TRIMMED
;
1311 gfs2_trans_add_bh(rgd
->rd_gl
, bh
, 1);
1312 gfs2_rgrp_out(rgd
, bh
->b_data
);
1313 gfs2_rgrp_ondisk2lvb(rgd
->rd_rgl
, bh
->b_data
);
1314 gfs2_trans_end(sdp
);
1317 gfs2_glock_dq_uninit(&gh
);
1322 rgd
= gfs2_rgrpd_get_next(rgd
);
1326 r
.len
= trimmed
<< 9;
1327 if (argp
&& copy_to_user(argp
, &r
, sizeof(r
)))
1334 * rs_insert - insert a new multi-block reservation into the rgrp's rb_tree
1335 * @ip: the inode structure
1338 static void rs_insert(struct gfs2_inode
*ip
)
1340 struct rb_node
**newn
, *parent
= NULL
;
1342 struct gfs2_blkreserv
*rs
= ip
->i_res
;
1343 struct gfs2_rgrpd
*rgd
= rs
->rs_rbm
.rgd
;
1344 u64 fsblock
= gfs2_rbm_to_block(&rs
->rs_rbm
);
1346 BUG_ON(gfs2_rs_active(rs
));
1348 spin_lock(&rgd
->rd_rsspin
);
1349 newn
= &rgd
->rd_rstree
.rb_node
;
1351 struct gfs2_blkreserv
*cur
=
1352 rb_entry(*newn
, struct gfs2_blkreserv
, rs_node
);
1355 rc
= rs_cmp(fsblock
, rs
->rs_free
, cur
);
1357 newn
= &((*newn
)->rb_right
);
1359 newn
= &((*newn
)->rb_left
);
1361 spin_unlock(&rgd
->rd_rsspin
);
1367 rb_link_node(&rs
->rs_node
, parent
, newn
);
1368 rb_insert_color(&rs
->rs_node
, &rgd
->rd_rstree
);
1370 /* Do our rgrp accounting for the reservation */
1371 rgd
->rd_reserved
+= rs
->rs_free
; /* blocks reserved */
1372 spin_unlock(&rgd
->rd_rsspin
);
1373 trace_gfs2_rs(rs
, TRACE_RS_INSERT
);
1377 * rg_mblk_search - find a group of multiple free blocks to form a reservation
1378 * @rgd: the resource group descriptor
1379 * @ip: pointer to the inode for which we're reserving blocks
1380 * @requested: number of blocks required for this allocation
1384 static void rg_mblk_search(struct gfs2_rgrpd
*rgd
, struct gfs2_inode
*ip
,
1387 struct gfs2_rbm rbm
= { .rgd
= rgd
, };
1389 struct gfs2_blkreserv
*rs
= ip
->i_res
;
1391 u32 free_blocks
= rgd
->rd_free_clone
- rgd
->rd_reserved
;
1394 extlen
= max_t(u32
, atomic_read(&rs
->rs_sizehint
), requested
);
1395 extlen
= clamp(extlen
, RGRP_RSRV_MINBLKS
, free_blocks
);
1396 if ((rgd
->rd_free_clone
< rgd
->rd_reserved
) || (free_blocks
< extlen
))
1399 /* Find bitmap block that contains bits for goal block */
1400 if (rgrp_contains_block(rgd
, ip
->i_goal
))
1403 goal
= rgd
->rd_last_alloc
+ rgd
->rd_data0
;
1405 if (WARN_ON(gfs2_rbm_from_block(&rbm
, goal
)))
1408 ret
= gfs2_rbm_find(&rbm
, GFS2_BLKST_FREE
, extlen
, ip
, true);
1411 rs
->rs_free
= extlen
;
1412 rs
->rs_inum
= ip
->i_no_addr
;
1418 * gfs2_next_unreserved_block - Return next block that is not reserved
1419 * @rgd: The resource group
1420 * @block: The starting block
1421 * @length: The required length
1422 * @ip: Ignore any reservations for this inode
1424 * If the block does not appear in any reservation, then return the
1425 * block number unchanged. If it does appear in the reservation, then
1426 * keep looking through the tree of reservations in order to find the
1427 * first block number which is not reserved.
1430 static u64
gfs2_next_unreserved_block(struct gfs2_rgrpd
*rgd
, u64 block
,
1432 const struct gfs2_inode
*ip
)
1434 struct gfs2_blkreserv
*rs
;
1438 spin_lock(&rgd
->rd_rsspin
);
1439 n
= rgd
->rd_rstree
.rb_node
;
1441 rs
= rb_entry(n
, struct gfs2_blkreserv
, rs_node
);
1442 rc
= rs_cmp(block
, length
, rs
);
1452 while ((rs_cmp(block
, length
, rs
) == 0) && (ip
->i_res
!= rs
)) {
1453 block
= gfs2_rbm_to_block(&rs
->rs_rbm
) + rs
->rs_free
;
1457 rs
= rb_entry(n
, struct gfs2_blkreserv
, rs_node
);
1461 spin_unlock(&rgd
->rd_rsspin
);
1466 * gfs2_reservation_check_and_update - Check for reservations during block alloc
1467 * @rbm: The current position in the resource group
1468 * @ip: The inode for which we are searching for blocks
1469 * @minext: The minimum extent length
1471 * This checks the current position in the rgrp to see whether there is
1472 * a reservation covering this block. If not then this function is a
1473 * no-op. If there is, then the position is moved to the end of the
1474 * contiguous reservation(s) so that we are pointing at the first
1475 * non-reserved block.
1477 * Returns: 0 if no reservation, 1 if @rbm has changed, otherwise an error
1480 static int gfs2_reservation_check_and_update(struct gfs2_rbm
*rbm
,
1481 const struct gfs2_inode
*ip
,
1484 u64 block
= gfs2_rbm_to_block(rbm
);
1490 * If we have a minimum extent length, then skip over any extent
1491 * which is less than the min extent length in size.
1494 extlen
= gfs2_free_extlen(rbm
, minext
);
1495 nblock
= block
+ extlen
;
1496 if (extlen
< minext
)
1501 * Check the extent which has been found against the reservations
1502 * and skip if parts of it are already reserved
1504 nblock
= gfs2_next_unreserved_block(rbm
->rgd
, block
, extlen
, ip
);
1505 if (nblock
== block
)
1508 ret
= gfs2_rbm_from_block(rbm
, nblock
);
1515 * gfs2_rbm_find - Look for blocks of a particular state
1516 * @rbm: Value/result starting position and final position
1517 * @state: The state which we want to find
1518 * @minext: The requested extent length (0 for a single block)
1519 * @ip: If set, check for reservations
1520 * @nowrap: Stop looking at the end of the rgrp, rather than wrapping
1521 * around until we've reached the starting point.
1524 * - If looking for free blocks, we set GBF_FULL on each bitmap which
1525 * has no free blocks in it.
1527 * Returns: 0 on success, -ENOSPC if there is no block of the requested state
1530 static int gfs2_rbm_find(struct gfs2_rbm
*rbm
, u8 state
, u32 minext
,
1531 const struct gfs2_inode
*ip
, bool nowrap
)
1533 struct buffer_head
*bh
;
1534 struct gfs2_bitmap
*initial_bi
;
1540 int iters
= rbm
->rgd
->rd_length
;
1543 /* If we are not starting at the beginning of a bitmap, then we
1544 * need to add one to the bitmap count to ensure that we search
1545 * the starting bitmap twice.
1547 if (rbm
->offset
!= 0)
1551 if (test_bit(GBF_FULL
, &rbm
->bi
->bi_flags
) &&
1552 (state
== GFS2_BLKST_FREE
))
1555 bh
= rbm
->bi
->bi_bh
;
1556 buffer
= bh
->b_data
+ rbm
->bi
->bi_offset
;
1557 WARN_ON(!buffer_uptodate(bh
));
1558 if (state
!= GFS2_BLKST_UNLINKED
&& rbm
->bi
->bi_clone
)
1559 buffer
= rbm
->bi
->bi_clone
+ rbm
->bi
->bi_offset
;
1560 initial_offset
= rbm
->offset
;
1561 offset
= gfs2_bitfit(buffer
, rbm
->bi
->bi_len
, rbm
->offset
, state
);
1562 if (offset
== BFITNOENT
)
1564 rbm
->offset
= offset
;
1568 initial_bi
= rbm
->bi
;
1569 ret
= gfs2_reservation_check_and_update(rbm
, ip
, minext
);
1573 n
+= (rbm
->bi
- initial_bi
);
1576 if (ret
== -E2BIG
) {
1579 n
+= (rbm
->bi
- initial_bi
);
1580 goto res_covered_end_of_rgrp
;
1584 bitmap_full
: /* Mark bitmap as full and fall through */
1585 if ((state
== GFS2_BLKST_FREE
) && initial_offset
== 0)
1586 set_bit(GBF_FULL
, &rbm
->bi
->bi_flags
);
1588 next_bitmap
: /* Find next bitmap in the rgrp */
1590 index
= rbm
->bi
- rbm
->rgd
->rd_bits
;
1592 if (index
== rbm
->rgd
->rd_length
)
1594 res_covered_end_of_rgrp
:
1595 rbm
->bi
= &rbm
->rgd
->rd_bits
[index
];
1596 if ((index
== 0) && nowrap
)
1608 * try_rgrp_unlink - Look for any unlinked, allocated, but unused inodes
1610 * @last_unlinked: block address of the last dinode we unlinked
1611 * @skip: block address we should explicitly not unlink
1613 * Returns: 0 if no error
1614 * The inode, if one has been found, in inode.
1617 static void try_rgrp_unlink(struct gfs2_rgrpd
*rgd
, u64
*last_unlinked
, u64 skip
)
1620 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
1621 struct gfs2_glock
*gl
;
1622 struct gfs2_inode
*ip
;
1625 struct gfs2_rbm rbm
= { .rgd
= rgd
, .bi
= rgd
->rd_bits
, .offset
= 0 };
1628 down_write(&sdp
->sd_log_flush_lock
);
1629 error
= gfs2_rbm_find(&rbm
, GFS2_BLKST_UNLINKED
, 0, NULL
, true);
1630 up_write(&sdp
->sd_log_flush_lock
);
1631 if (error
== -ENOSPC
)
1633 if (WARN_ON_ONCE(error
))
1636 block
= gfs2_rbm_to_block(&rbm
);
1637 if (gfs2_rbm_from_block(&rbm
, block
+ 1))
1639 if (*last_unlinked
!= NO_BLOCK
&& block
<= *last_unlinked
)
1643 *last_unlinked
= block
;
1645 error
= gfs2_glock_get(sdp
, block
, &gfs2_inode_glops
, CREATE
, &gl
);
1649 /* If the inode is already in cache, we can ignore it here
1650 * because the existing inode disposal code will deal with
1651 * it when all refs have gone away. Accessing gl_object like
1652 * this is not safe in general. Here it is ok because we do
1653 * not dereference the pointer, and we only need an approx
1654 * answer to whether it is NULL or not.
1658 if (ip
|| queue_work(gfs2_delete_workqueue
, &gl
->gl_delete
) == 0)
1663 /* Limit reclaim to sensible number of tasks */
1664 if (found
> NR_CPUS
)
1668 rgd
->rd_flags
&= ~GFS2_RDF_CHECK
;
1672 static bool gfs2_select_rgrp(struct gfs2_rgrpd
**pos
, const struct gfs2_rgrpd
*begin
)
1674 struct gfs2_rgrpd
*rgd
= *pos
;
1676 rgd
= gfs2_rgrpd_get_next(rgd
);
1678 rgd
= gfs2_rgrpd_get_next(NULL
);
1680 if (rgd
!= begin
) /* If we didn't wrap */
1686 * gfs2_inplace_reserve - Reserve space in the filesystem
1687 * @ip: the inode to reserve space for
1688 * @requested: the number of blocks to be reserved
1693 int gfs2_inplace_reserve(struct gfs2_inode
*ip
, u32 requested
)
1695 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
1696 struct gfs2_rgrpd
*begin
= NULL
;
1697 struct gfs2_blkreserv
*rs
= ip
->i_res
;
1698 int error
= 0, rg_locked
, flags
= LM_FLAG_TRY
;
1699 u64 last_unlinked
= NO_BLOCK
;
1702 if (sdp
->sd_args
.ar_rgrplvb
)
1704 if (gfs2_assert_warn(sdp
, requested
))
1706 if (gfs2_rs_active(rs
)) {
1707 begin
= rs
->rs_rbm
.rgd
;
1708 flags
= 0; /* Yoda: Do or do not. There is no try */
1709 } else if (ip
->i_rgd
&& rgrp_contains_block(ip
->i_rgd
, ip
->i_goal
)) {
1710 rs
->rs_rbm
.rgd
= begin
= ip
->i_rgd
;
1712 rs
->rs_rbm
.rgd
= begin
= gfs2_blk2rgrpd(sdp
, ip
->i_goal
, 1);
1714 if (rs
->rs_rbm
.rgd
== NULL
)
1720 if (!gfs2_glock_is_locked_by_me(rs
->rs_rbm
.rgd
->rd_gl
)) {
1722 error
= gfs2_glock_nq_init(rs
->rs_rbm
.rgd
->rd_gl
,
1723 LM_ST_EXCLUSIVE
, flags
,
1725 if (error
== GLR_TRYFAILED
)
1727 if (unlikely(error
))
1729 if (sdp
->sd_args
.ar_rgrplvb
) {
1730 error
= update_rgrp_lvb(rs
->rs_rbm
.rgd
);
1731 if (unlikely(error
)) {
1732 gfs2_glock_dq_uninit(&rs
->rs_rgd_gh
);
1738 /* Skip unuseable resource groups */
1739 if (rs
->rs_rbm
.rgd
->rd_flags
& (GFS2_RGF_NOALLOC
| GFS2_RDF_ERROR
))
1742 if (sdp
->sd_args
.ar_rgrplvb
)
1743 gfs2_rgrp_bh_get(rs
->rs_rbm
.rgd
);
1745 /* Get a reservation if we don't already have one */
1746 if (!gfs2_rs_active(rs
))
1747 rg_mblk_search(rs
->rs_rbm
.rgd
, ip
, requested
);
1749 /* Skip rgrps when we can't get a reservation on first pass */
1750 if (!gfs2_rs_active(rs
) && (loops
< 1))
1753 /* If rgrp has enough free space, use it */
1754 if (rs
->rs_rbm
.rgd
->rd_free_clone
>= requested
) {
1755 ip
->i_rgd
= rs
->rs_rbm
.rgd
;
1759 /* Drop reservation, if we couldn't use reserved rgrp */
1760 if (gfs2_rs_active(rs
))
1761 gfs2_rs_deltree(ip
, rs
);
1763 /* Check for unlinked inodes which can be reclaimed */
1764 if (rs
->rs_rbm
.rgd
->rd_flags
& GFS2_RDF_CHECK
)
1765 try_rgrp_unlink(rs
->rs_rbm
.rgd
, &last_unlinked
,
1768 /* Unlock rgrp if required */
1770 gfs2_glock_dq_uninit(&rs
->rs_rgd_gh
);
1772 /* Find the next rgrp, and continue looking */
1773 if (gfs2_select_rgrp(&rs
->rs_rbm
.rgd
, begin
))
1776 /* If we've scanned all the rgrps, but found no free blocks
1777 * then this checks for some less likely conditions before
1780 flags
&= ~LM_FLAG_TRY
;
1782 /* Check that fs hasn't grown if writing to rindex */
1783 if (ip
== GFS2_I(sdp
->sd_rindex
) && !sdp
->sd_rindex_uptodate
) {
1784 error
= gfs2_ri_update(ip
);
1788 /* Flushing the log may release space */
1790 gfs2_log_flush(sdp
, NULL
);
1797 * gfs2_inplace_release - release an inplace reservation
1798 * @ip: the inode the reservation was taken out on
1800 * Release a reservation made by gfs2_inplace_reserve().
1803 void gfs2_inplace_release(struct gfs2_inode
*ip
)
1805 struct gfs2_blkreserv
*rs
= ip
->i_res
;
1807 if (rs
->rs_rgd_gh
.gh_gl
)
1808 gfs2_glock_dq_uninit(&rs
->rs_rgd_gh
);
1812 * gfs2_get_block_type - Check a block in a RG is of given type
1813 * @rgd: the resource group holding the block
1814 * @block: the block number
1816 * Returns: The block type (GFS2_BLKST_*)
1819 static unsigned char gfs2_get_block_type(struct gfs2_rgrpd
*rgd
, u64 block
)
1821 struct gfs2_rbm rbm
= { .rgd
= rgd
, };
1824 ret
= gfs2_rbm_from_block(&rbm
, block
);
1825 WARN_ON_ONCE(ret
!= 0);
1827 return gfs2_testbit(&rbm
);
1832 * gfs2_alloc_extent - allocate an extent from a given bitmap
1833 * @rbm: the resource group information
1834 * @dinode: TRUE if the first block we allocate is for a dinode
1835 * @n: The extent length (value/result)
1837 * Add the bitmap buffer to the transaction.
1838 * Set the found bits to @new_state to change block's allocation state.
1840 static void gfs2_alloc_extent(const struct gfs2_rbm
*rbm
, bool dinode
,
1843 struct gfs2_rbm pos
= { .rgd
= rbm
->rgd
, };
1844 const unsigned int elen
= *n
;
1849 block
= gfs2_rbm_to_block(rbm
);
1850 gfs2_trans_add_bh(rbm
->rgd
->rd_gl
, rbm
->bi
->bi_bh
, 1);
1851 gfs2_setbit(rbm
, true, dinode
? GFS2_BLKST_DINODE
: GFS2_BLKST_USED
);
1854 ret
= gfs2_rbm_from_block(&pos
, block
);
1855 if (ret
|| gfs2_testbit(&pos
) != GFS2_BLKST_FREE
)
1857 gfs2_trans_add_bh(pos
.rgd
->rd_gl
, pos
.bi
->bi_bh
, 1);
1858 gfs2_setbit(&pos
, true, GFS2_BLKST_USED
);
1865 * rgblk_free - Change alloc state of given block(s)
1866 * @sdp: the filesystem
1867 * @bstart: the start of a run of blocks to free
1868 * @blen: the length of the block run (all must lie within ONE RG!)
1869 * @new_state: GFS2_BLKST_XXX the after-allocation block state
1871 * Returns: Resource group containing the block(s)
1874 static struct gfs2_rgrpd
*rgblk_free(struct gfs2_sbd
*sdp
, u64 bstart
,
1875 u32 blen
, unsigned char new_state
)
1877 struct gfs2_rbm rbm
;
1879 rbm
.rgd
= gfs2_blk2rgrpd(sdp
, bstart
, 1);
1881 if (gfs2_consist(sdp
))
1882 fs_err(sdp
, "block = %llu\n", (unsigned long long)bstart
);
1887 gfs2_rbm_from_block(&rbm
, bstart
);
1889 if (!rbm
.bi
->bi_clone
) {
1890 rbm
.bi
->bi_clone
= kmalloc(rbm
.bi
->bi_bh
->b_size
,
1891 GFP_NOFS
| __GFP_NOFAIL
);
1892 memcpy(rbm
.bi
->bi_clone
+ rbm
.bi
->bi_offset
,
1893 rbm
.bi
->bi_bh
->b_data
+ rbm
.bi
->bi_offset
,
1896 gfs2_trans_add_bh(rbm
.rgd
->rd_gl
, rbm
.bi
->bi_bh
, 1);
1897 gfs2_setbit(&rbm
, false, new_state
);
1904 * gfs2_rgrp_dump - print out an rgrp
1905 * @seq: The iterator
1906 * @gl: The glock in question
1910 int gfs2_rgrp_dump(struct seq_file
*seq
, const struct gfs2_glock
*gl
)
1912 struct gfs2_rgrpd
*rgd
= gl
->gl_object
;
1913 struct gfs2_blkreserv
*trs
;
1914 const struct rb_node
*n
;
1918 gfs2_print_dbg(seq
, " R: n:%llu f:%02x b:%u/%u i:%u r:%u\n",
1919 (unsigned long long)rgd
->rd_addr
, rgd
->rd_flags
,
1920 rgd
->rd_free
, rgd
->rd_free_clone
, rgd
->rd_dinodes
,
1922 spin_lock(&rgd
->rd_rsspin
);
1923 for (n
= rb_first(&rgd
->rd_rstree
); n
; n
= rb_next(&trs
->rs_node
)) {
1924 trs
= rb_entry(n
, struct gfs2_blkreserv
, rs_node
);
1927 spin_unlock(&rgd
->rd_rsspin
);
1931 static void gfs2_rgrp_error(struct gfs2_rgrpd
*rgd
)
1933 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
1934 fs_warn(sdp
, "rgrp %llu has an error, marking it readonly until umount\n",
1935 (unsigned long long)rgd
->rd_addr
);
1936 fs_warn(sdp
, "umount on all nodes and run fsck.gfs2 to fix the error\n");
1937 gfs2_rgrp_dump(NULL
, rgd
->rd_gl
);
1938 rgd
->rd_flags
|= GFS2_RDF_ERROR
;
1942 * gfs2_adjust_reservation - Adjust (or remove) a reservation after allocation
1943 * @ip: The inode we have just allocated blocks for
1944 * @rbm: The start of the allocated blocks
1945 * @len: The extent length
1947 * Adjusts a reservation after an allocation has taken place. If the
1948 * reservation does not match the allocation, or if it is now empty
1949 * then it is removed.
1952 static void gfs2_adjust_reservation(struct gfs2_inode
*ip
,
1953 const struct gfs2_rbm
*rbm
, unsigned len
)
1955 struct gfs2_blkreserv
*rs
= ip
->i_res
;
1956 struct gfs2_rgrpd
*rgd
= rbm
->rgd
;
1961 spin_lock(&rgd
->rd_rsspin
);
1962 if (gfs2_rs_active(rs
)) {
1963 if (gfs2_rbm_eq(&rs
->rs_rbm
, rbm
)) {
1964 block
= gfs2_rbm_to_block(rbm
);
1965 ret
= gfs2_rbm_from_block(&rs
->rs_rbm
, block
+ len
);
1966 rlen
= min(rs
->rs_free
, len
);
1967 rs
->rs_free
-= rlen
;
1968 rgd
->rd_reserved
-= rlen
;
1969 trace_gfs2_rs(rs
, TRACE_RS_CLAIM
);
1970 if (rs
->rs_free
&& !ret
)
1973 __rs_deltree(ip
, rs
);
1976 spin_unlock(&rgd
->rd_rsspin
);
1980 * gfs2_alloc_blocks - Allocate one or more blocks of data and/or a dinode
1981 * @ip: the inode to allocate the block for
1982 * @bn: Used to return the starting block number
1983 * @nblocks: requested number of blocks/extent length (value/result)
1984 * @dinode: 1 if we're allocating a dinode block, else 0
1985 * @generation: the generation number of the inode
1987 * Returns: 0 or error
1990 int gfs2_alloc_blocks(struct gfs2_inode
*ip
, u64
*bn
, unsigned int *nblocks
,
1991 bool dinode
, u64
*generation
)
1993 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
1994 struct buffer_head
*dibh
;
1995 struct gfs2_rbm rbm
= { .rgd
= ip
->i_rgd
, };
1998 u64 block
; /* block, within the file system scope */
2001 if (gfs2_rs_active(ip
->i_res
))
2002 goal
= gfs2_rbm_to_block(&ip
->i_res
->rs_rbm
);
2003 else if (!dinode
&& rgrp_contains_block(rbm
.rgd
, ip
->i_goal
))
2006 goal
= rbm
.rgd
->rd_last_alloc
+ rbm
.rgd
->rd_data0
;
2008 gfs2_rbm_from_block(&rbm
, goal
);
2009 error
= gfs2_rbm_find(&rbm
, GFS2_BLKST_FREE
, 0, ip
, false);
2011 if (error
== -ENOSPC
) {
2012 gfs2_rbm_from_block(&rbm
, goal
);
2013 error
= gfs2_rbm_find(&rbm
, GFS2_BLKST_FREE
, 0, NULL
, false);
2016 /* Since all blocks are reserved in advance, this shouldn't happen */
2018 fs_warn(sdp
, "inum=%llu error=%d, nblocks=%u, full=%d\n",
2019 (unsigned long long)ip
->i_no_addr
, error
, *nblocks
,
2020 test_bit(GBF_FULL
, &rbm
.rgd
->rd_bits
->bi_flags
));
2024 gfs2_alloc_extent(&rbm
, dinode
, nblocks
);
2025 block
= gfs2_rbm_to_block(&rbm
);
2026 rbm
.rgd
->rd_last_alloc
= block
- rbm
.rgd
->rd_data0
;
2027 if (gfs2_rs_active(ip
->i_res
))
2028 gfs2_adjust_reservation(ip
, &rbm
, *nblocks
);
2034 ip
->i_goal
= block
+ ndata
- 1;
2035 error
= gfs2_meta_inode_buffer(ip
, &dibh
);
2037 struct gfs2_dinode
*di
=
2038 (struct gfs2_dinode
*)dibh
->b_data
;
2039 gfs2_trans_add_bh(ip
->i_gl
, dibh
, 1);
2040 di
->di_goal_meta
= di
->di_goal_data
=
2041 cpu_to_be64(ip
->i_goal
);
2045 if (rbm
.rgd
->rd_free
< *nblocks
) {
2046 printk(KERN_WARNING
"nblocks=%u\n", *nblocks
);
2050 rbm
.rgd
->rd_free
-= *nblocks
;
2052 rbm
.rgd
->rd_dinodes
++;
2053 *generation
= rbm
.rgd
->rd_igeneration
++;
2054 if (*generation
== 0)
2055 *generation
= rbm
.rgd
->rd_igeneration
++;
2058 gfs2_trans_add_bh(rbm
.rgd
->rd_gl
, rbm
.rgd
->rd_bits
[0].bi_bh
, 1);
2059 gfs2_rgrp_out(rbm
.rgd
, rbm
.rgd
->rd_bits
[0].bi_bh
->b_data
);
2060 gfs2_rgrp_ondisk2lvb(rbm
.rgd
->rd_rgl
, rbm
.rgd
->rd_bits
[0].bi_bh
->b_data
);
2062 gfs2_statfs_change(sdp
, 0, -(s64
)*nblocks
, dinode
? 1 : 0);
2064 gfs2_trans_add_unrevoke(sdp
, block
, 1);
2067 * This needs reviewing to see why we cannot do the quota change
2068 * at this point in the dinode case.
2071 gfs2_quota_change(ip
, ndata
, ip
->i_inode
.i_uid
,
2074 rbm
.rgd
->rd_free_clone
-= *nblocks
;
2075 trace_gfs2_block_alloc(ip
, rbm
.rgd
, block
, *nblocks
,
2076 dinode
? GFS2_BLKST_DINODE
: GFS2_BLKST_USED
);
2081 gfs2_rgrp_error(rbm
.rgd
);
2086 * __gfs2_free_blocks - free a contiguous run of block(s)
2087 * @ip: the inode these blocks are being freed from
2088 * @bstart: first block of a run of contiguous blocks
2089 * @blen: the length of the block run
2090 * @meta: 1 if the blocks represent metadata
2094 void __gfs2_free_blocks(struct gfs2_inode
*ip
, u64 bstart
, u32 blen
, int meta
)
2096 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
2097 struct gfs2_rgrpd
*rgd
;
2099 rgd
= rgblk_free(sdp
, bstart
, blen
, GFS2_BLKST_FREE
);
2102 trace_gfs2_block_alloc(ip
, rgd
, bstart
, blen
, GFS2_BLKST_FREE
);
2103 rgd
->rd_free
+= blen
;
2104 rgd
->rd_flags
&= ~GFS2_RGF_TRIMMED
;
2105 gfs2_trans_add_bh(rgd
->rd_gl
, rgd
->rd_bits
[0].bi_bh
, 1);
2106 gfs2_rgrp_out(rgd
, rgd
->rd_bits
[0].bi_bh
->b_data
);
2107 gfs2_rgrp_ondisk2lvb(rgd
->rd_rgl
, rgd
->rd_bits
[0].bi_bh
->b_data
);
2109 /* Directories keep their data in the metadata address space */
2110 if (meta
|| ip
->i_depth
)
2111 gfs2_meta_wipe(ip
, bstart
, blen
);
2115 * gfs2_free_meta - free a contiguous run of data block(s)
2116 * @ip: the inode these blocks are being freed from
2117 * @bstart: first block of a run of contiguous blocks
2118 * @blen: the length of the block run
2122 void gfs2_free_meta(struct gfs2_inode
*ip
, u64 bstart
, u32 blen
)
2124 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
2126 __gfs2_free_blocks(ip
, bstart
, blen
, 1);
2127 gfs2_statfs_change(sdp
, 0, +blen
, 0);
2128 gfs2_quota_change(ip
, -(s64
)blen
, ip
->i_inode
.i_uid
, ip
->i_inode
.i_gid
);
2131 void gfs2_unlink_di(struct inode
*inode
)
2133 struct gfs2_inode
*ip
= GFS2_I(inode
);
2134 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
2135 struct gfs2_rgrpd
*rgd
;
2136 u64 blkno
= ip
->i_no_addr
;
2138 rgd
= rgblk_free(sdp
, blkno
, 1, GFS2_BLKST_UNLINKED
);
2141 trace_gfs2_block_alloc(ip
, rgd
, blkno
, 1, GFS2_BLKST_UNLINKED
);
2142 gfs2_trans_add_bh(rgd
->rd_gl
, rgd
->rd_bits
[0].bi_bh
, 1);
2143 gfs2_rgrp_out(rgd
, rgd
->rd_bits
[0].bi_bh
->b_data
);
2144 gfs2_rgrp_ondisk2lvb(rgd
->rd_rgl
, rgd
->rd_bits
[0].bi_bh
->b_data
);
2145 update_rgrp_lvb_unlinked(rgd
, 1);
2148 static void gfs2_free_uninit_di(struct gfs2_rgrpd
*rgd
, u64 blkno
)
2150 struct gfs2_sbd
*sdp
= rgd
->rd_sbd
;
2151 struct gfs2_rgrpd
*tmp_rgd
;
2153 tmp_rgd
= rgblk_free(sdp
, blkno
, 1, GFS2_BLKST_FREE
);
2156 gfs2_assert_withdraw(sdp
, rgd
== tmp_rgd
);
2158 if (!rgd
->rd_dinodes
)
2159 gfs2_consist_rgrpd(rgd
);
2163 gfs2_trans_add_bh(rgd
->rd_gl
, rgd
->rd_bits
[0].bi_bh
, 1);
2164 gfs2_rgrp_out(rgd
, rgd
->rd_bits
[0].bi_bh
->b_data
);
2165 gfs2_rgrp_ondisk2lvb(rgd
->rd_rgl
, rgd
->rd_bits
[0].bi_bh
->b_data
);
2166 update_rgrp_lvb_unlinked(rgd
, -1);
2168 gfs2_statfs_change(sdp
, 0, +1, -1);
2172 void gfs2_free_di(struct gfs2_rgrpd
*rgd
, struct gfs2_inode
*ip
)
2174 gfs2_free_uninit_di(rgd
, ip
->i_no_addr
);
2175 trace_gfs2_block_alloc(ip
, rgd
, ip
->i_no_addr
, 1, GFS2_BLKST_FREE
);
2176 gfs2_quota_change(ip
, -1, ip
->i_inode
.i_uid
, ip
->i_inode
.i_gid
);
2177 gfs2_meta_wipe(ip
, ip
->i_no_addr
, 1);
2181 * gfs2_check_blk_type - Check the type of a block
2182 * @sdp: The superblock
2183 * @no_addr: The block number to check
2184 * @type: The block type we are looking for
2186 * Returns: 0 if the block type matches the expected type
2187 * -ESTALE if it doesn't match
2188 * or -ve errno if something went wrong while checking
2191 int gfs2_check_blk_type(struct gfs2_sbd
*sdp
, u64 no_addr
, unsigned int type
)
2193 struct gfs2_rgrpd
*rgd
;
2194 struct gfs2_holder rgd_gh
;
2195 int error
= -EINVAL
;
2197 rgd
= gfs2_blk2rgrpd(sdp
, no_addr
, 1);
2201 error
= gfs2_glock_nq_init(rgd
->rd_gl
, LM_ST_SHARED
, 0, &rgd_gh
);
2205 if (gfs2_get_block_type(rgd
, no_addr
) != type
)
2208 gfs2_glock_dq_uninit(&rgd_gh
);
2214 * gfs2_rlist_add - add a RG to a list of RGs
2216 * @rlist: the list of resource groups
2219 * Figure out what RG a block belongs to and add that RG to the list
2221 * FIXME: Don't use NOFAIL
2225 void gfs2_rlist_add(struct gfs2_inode
*ip
, struct gfs2_rgrp_list
*rlist
,
2228 struct gfs2_sbd
*sdp
= GFS2_SB(&ip
->i_inode
);
2229 struct gfs2_rgrpd
*rgd
;
2230 struct gfs2_rgrpd
**tmp
;
2231 unsigned int new_space
;
2234 if (gfs2_assert_warn(sdp
, !rlist
->rl_ghs
))
2237 if (ip
->i_rgd
&& rgrp_contains_block(ip
->i_rgd
, block
))
2240 rgd
= gfs2_blk2rgrpd(sdp
, block
, 1);
2242 fs_err(sdp
, "rlist_add: no rgrp for block %llu\n", (unsigned long long)block
);
2247 for (x
= 0; x
< rlist
->rl_rgrps
; x
++)
2248 if (rlist
->rl_rgd
[x
] == rgd
)
2251 if (rlist
->rl_rgrps
== rlist
->rl_space
) {
2252 new_space
= rlist
->rl_space
+ 10;
2254 tmp
= kcalloc(new_space
, sizeof(struct gfs2_rgrpd
*),
2255 GFP_NOFS
| __GFP_NOFAIL
);
2257 if (rlist
->rl_rgd
) {
2258 memcpy(tmp
, rlist
->rl_rgd
,
2259 rlist
->rl_space
* sizeof(struct gfs2_rgrpd
*));
2260 kfree(rlist
->rl_rgd
);
2263 rlist
->rl_space
= new_space
;
2264 rlist
->rl_rgd
= tmp
;
2267 rlist
->rl_rgd
[rlist
->rl_rgrps
++] = rgd
;
2271 * gfs2_rlist_alloc - all RGs have been added to the rlist, now allocate
2272 * and initialize an array of glock holders for them
2273 * @rlist: the list of resource groups
2274 * @state: the lock state to acquire the RG lock in
2276 * FIXME: Don't use NOFAIL
2280 void gfs2_rlist_alloc(struct gfs2_rgrp_list
*rlist
, unsigned int state
)
2284 rlist
->rl_ghs
= kcalloc(rlist
->rl_rgrps
, sizeof(struct gfs2_holder
),
2285 GFP_NOFS
| __GFP_NOFAIL
);
2286 for (x
= 0; x
< rlist
->rl_rgrps
; x
++)
2287 gfs2_holder_init(rlist
->rl_rgd
[x
]->rd_gl
,
2293 * gfs2_rlist_free - free a resource group list
2294 * @list: the list of resource groups
2298 void gfs2_rlist_free(struct gfs2_rgrp_list
*rlist
)
2302 kfree(rlist
->rl_rgd
);
2304 if (rlist
->rl_ghs
) {
2305 for (x
= 0; x
< rlist
->rl_rgrps
; x
++)
2306 gfs2_holder_uninit(&rlist
->rl_ghs
[x
]);
2307 kfree(rlist
->rl_ghs
);
2308 rlist
->rl_ghs
= NULL
;