Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
fe6c991c | 3 | * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved. |
b3b94faa DT |
4 | * |
5 | * This copyrighted material is made available to anyone wishing to use, | |
6 | * modify, copy, or redistribute it subject to the terms and conditions | |
e9fc2aa0 | 7 | * of the GNU General Public License version 2. |
b3b94faa DT |
8 | */ |
9 | ||
b3b94faa DT |
10 | #include <linux/slab.h> |
11 | #include <linux/spinlock.h> | |
12 | #include <linux/completion.h> | |
13 | #include <linux/buffer_head.h> | |
f42faf4f | 14 | #include <linux/fs.h> |
5c676f6d | 15 | #include <linux/gfs2_ondisk.h> |
1f466a47 | 16 | #include <linux/prefetch.h> |
f15ab561 | 17 | #include <linux/blkdev.h> |
7c9ca621 | 18 | #include <linux/rbtree.h> |
b3b94faa DT |
19 | |
20 | #include "gfs2.h" | |
5c676f6d | 21 | #include "incore.h" |
b3b94faa DT |
22 | #include "glock.h" |
23 | #include "glops.h" | |
b3b94faa DT |
24 | #include "lops.h" |
25 | #include "meta_io.h" | |
26 | #include "quota.h" | |
27 | #include "rgrp.h" | |
28 | #include "super.h" | |
29 | #include "trans.h" | |
5c676f6d | 30 | #include "util.h" |
172e045a | 31 | #include "log.h" |
c8cdf479 | 32 | #include "inode.h" |
63997775 | 33 | #include "trace_gfs2.h" |
b3b94faa | 34 | |
2c1e52aa | 35 | #define BFITNOENT ((u32)~0) |
6760bdcd | 36 | #define NO_BLOCK ((u64)~0) |
88c8ab1f | 37 | |
1f466a47 BP |
38 | #if BITS_PER_LONG == 32 |
39 | #define LBITMASK (0x55555555UL) | |
40 | #define LBITSKIP55 (0x55555555UL) | |
41 | #define LBITSKIP00 (0x00000000UL) | |
42 | #else | |
43 | #define LBITMASK (0x5555555555555555UL) | |
44 | #define LBITSKIP55 (0x5555555555555555UL) | |
45 | #define LBITSKIP00 (0x0000000000000000UL) | |
46 | #endif | |
47 | ||
88c8ab1f SW |
48 | /* |
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 | |
feaa7bba SW |
51 | * bits. So, each byte represents GFS2_NBBY (i.e. 4) blocks. |
52 | * | |
53 | * 0 = Free | |
54 | * 1 = Used (not metadata) | |
55 | * 2 = Unlinked (still in use) inode | |
56 | * 3 = Used (metadata) | |
88c8ab1f SW |
57 | */ |
58 | ||
59 | static const char valid_change[16] = { | |
60 | /* current */ | |
feaa7bba | 61 | /* n */ 0, 1, 1, 1, |
88c8ab1f | 62 | /* e */ 1, 0, 0, 0, |
feaa7bba | 63 | /* w */ 0, 0, 0, 1, |
88c8ab1f SW |
64 | 1, 0, 0, 0 |
65 | }; | |
66 | ||
ff7f4cb4 SW |
67 | static int gfs2_rbm_find(struct gfs2_rbm *rbm, u8 state, u32 minext, |
68 | const struct gfs2_inode *ip, bool nowrap); | |
69 | ||
70 | ||
88c8ab1f SW |
71 | /** |
72 | * gfs2_setbit - Set a bit in the bitmaps | |
3e6339dd SW |
73 | * @rbm: The position of the bit to set |
74 | * @do_clone: Also set the clone bitmap, if it exists | |
88c8ab1f SW |
75 | * @new_state: the new state of the block |
76 | * | |
77 | */ | |
78 | ||
3e6339dd | 79 | static inline void gfs2_setbit(const struct gfs2_rbm *rbm, bool do_clone, |
06344b91 | 80 | unsigned char new_state) |
88c8ab1f | 81 | { |
b45e41d7 | 82 | unsigned char *byte1, *byte2, *end, cur_state; |
3e6339dd SW |
83 | unsigned int buflen = rbm->bi->bi_len; |
84 | const unsigned int bit = (rbm->offset % GFS2_NBBY) * GFS2_BIT_SIZE; | |
88c8ab1f | 85 | |
3e6339dd SW |
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; | |
88c8ab1f | 88 | |
b45e41d7 | 89 | BUG_ON(byte1 >= end); |
88c8ab1f | 90 | |
b45e41d7 | 91 | cur_state = (*byte1 >> bit) & GFS2_BIT_MASK; |
88c8ab1f | 92 | |
b45e41d7 | 93 | if (unlikely(!valid_change[new_state * 4 + cur_state])) { |
3e6339dd SW |
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, | |
98 | rbm->bi->bi_start); | |
99 | printk(KERN_WARNING "GFS2: bi_offset=0x%x bi_len=0x%x\n", | |
100 | rbm->bi->bi_offset, rbm->bi->bi_len); | |
95c8e17f | 101 | dump_stack(); |
3e6339dd | 102 | gfs2_consist_rgrpd(rbm->rgd); |
b45e41d7 SW |
103 | return; |
104 | } | |
105 | *byte1 ^= (cur_state ^ new_state) << bit; | |
106 | ||
3e6339dd SW |
107 | if (do_clone && rbm->bi->bi_clone) { |
108 | byte2 = rbm->bi->bi_clone + rbm->bi->bi_offset + (rbm->offset / GFS2_NBBY); | |
b45e41d7 SW |
109 | cur_state = (*byte2 >> bit) & GFS2_BIT_MASK; |
110 | *byte2 ^= (cur_state ^ new_state) << bit; | |
111 | } | |
88c8ab1f SW |
112 | } |
113 | ||
114 | /** | |
115 | * gfs2_testbit - test a bit in the bitmaps | |
c04a2ef3 | 116 | * @rbm: The bit to test |
88c8ab1f | 117 | * |
c04a2ef3 | 118 | * Returns: The two bit block state of the requested bit |
88c8ab1f SW |
119 | */ |
120 | ||
c04a2ef3 | 121 | static inline u8 gfs2_testbit(const struct gfs2_rbm *rbm) |
88c8ab1f | 122 | { |
c04a2ef3 SW |
123 | const u8 *buffer = rbm->bi->bi_bh->b_data + rbm->bi->bi_offset; |
124 | const u8 *byte; | |
88c8ab1f SW |
125 | unsigned int bit; |
126 | ||
c04a2ef3 SW |
127 | byte = buffer + (rbm->offset / GFS2_NBBY); |
128 | bit = (rbm->offset % GFS2_NBBY) * GFS2_BIT_SIZE; | |
88c8ab1f | 129 | |
c04a2ef3 | 130 | return (*byte >> bit) & GFS2_BIT_MASK; |
88c8ab1f SW |
131 | } |
132 | ||
223b2b88 SW |
133 | /** |
134 | * gfs2_bit_search | |
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 | |
138 | * | |
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 | |
145 | * odd bit positions. | |
146 | * | |
147 | * This allows searching of a whole u64 at once (32 blocks) with a | |
148 | * single test (on 64 bit arches). | |
149 | */ | |
150 | ||
151 | static inline u64 gfs2_bit_search(const __le64 *ptr, u64 mask, u8 state) | |
152 | { | |
153 | u64 tmp; | |
154 | static const u64 search[] = { | |
075ac448 HE |
155 | [0] = 0xffffffffffffffffULL, |
156 | [1] = 0xaaaaaaaaaaaaaaaaULL, | |
157 | [2] = 0x5555555555555555ULL, | |
158 | [3] = 0x0000000000000000ULL, | |
223b2b88 SW |
159 | }; |
160 | tmp = le64_to_cpu(*ptr) ^ search[state]; | |
161 | tmp &= (tmp >> 1); | |
162 | tmp &= mask; | |
163 | return tmp; | |
164 | } | |
165 | ||
8e2e0047 BP |
166 | /** |
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 | |
171 | * | |
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 | |
175 | */ | |
176 | static inline int rs_cmp(u64 blk, u32 len, struct gfs2_blkreserv *rs) | |
177 | { | |
4a993fb1 | 178 | u64 startblk = gfs2_rbm_to_block(&rs->rs_rbm); |
8e2e0047 BP |
179 | |
180 | if (blk >= startblk + rs->rs_free) | |
181 | return 1; | |
182 | if (blk + len - 1 < startblk) | |
183 | return -1; | |
184 | return 0; | |
185 | } | |
186 | ||
88c8ab1f SW |
187 | /** |
188 | * gfs2_bitfit - Search an rgrp's bitmap buffer to find a bit-pair representing | |
189 | * a block in a given allocation state. | |
886b1416 | 190 | * @buf: the buffer that holds the bitmaps |
223b2b88 | 191 | * @len: the length (in bytes) of the buffer |
88c8ab1f | 192 | * @goal: start search at this block's bit-pair (within @buffer) |
223b2b88 | 193 | * @state: GFS2_BLKST_XXX the state of the block we're looking for. |
88c8ab1f SW |
194 | * |
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 | |
223b2b88 SW |
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. | |
200 | * | |
201 | * The size of the buffer is in bytes, but is it assumed that it is | |
fd589a8f | 202 | * always ok to read a complete multiple of 64 bits at the end |
223b2b88 | 203 | * of the block in case the end is no aligned to a natural boundary. |
88c8ab1f SW |
204 | * |
205 | * Return: the block number (bitmap buffer scope) that was found | |
206 | */ | |
207 | ||
02ab1721 HE |
208 | static u32 gfs2_bitfit(const u8 *buf, const unsigned int len, |
209 | u32 goal, u8 state) | |
88c8ab1f | 210 | { |
223b2b88 SW |
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))); | |
214 | u64 tmp; | |
075ac448 | 215 | u64 mask = 0x5555555555555555ULL; |
223b2b88 SW |
216 | u32 bit; |
217 | ||
223b2b88 SW |
218 | /* Mask off bits we don't care about at the start of the search */ |
219 | mask <<= spoint; | |
220 | tmp = gfs2_bit_search(ptr, mask, state); | |
221 | ptr++; | |
222 | while(tmp == 0 && ptr < end) { | |
075ac448 | 223 | tmp = gfs2_bit_search(ptr, 0x5555555555555555ULL, state); |
223b2b88 | 224 | ptr++; |
1f466a47 | 225 | } |
223b2b88 SW |
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 */ | |
230 | if (tmp == 0) | |
231 | return BFITNOENT; | |
232 | ptr--; | |
d8bd504a | 233 | bit = __ffs64(tmp); |
223b2b88 SW |
234 | bit /= 2; /* two bits per entry in the bitmap */ |
235 | return (((const unsigned char *)ptr - buf) * GFS2_NBBY) + bit; | |
88c8ab1f SW |
236 | } |
237 | ||
ff7f4cb4 SW |
238 | /** |
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) | |
242 | * | |
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. | |
247 | * | |
248 | * Returns: 0 on success, or an error code | |
249 | */ | |
250 | ||
251 | static int gfs2_rbm_from_block(struct gfs2_rbm *rbm, u64 block) | |
252 | { | |
253 | u64 rblock = block - rbm->rgd->rd_data0; | |
a68a0a35 | 254 | u32 x; |
ff7f4cb4 SW |
255 | |
256 | if (WARN_ON_ONCE(rblock > UINT_MAX)) | |
257 | return -EINVAL; | |
258 | if (block >= rbm->rgd->rd_data0 + rbm->rgd->rd_data) | |
259 | return -E2BIG; | |
260 | ||
a68a0a35 BP |
261 | rbm->bi = rbm->rgd->rd_bits; |
262 | rbm->offset = (u32)(rblock); | |
263 | /* Check if the block is within the first block */ | |
264 | if (rbm->offset < (rbm->bi->bi_start + rbm->bi->bi_len) * GFS2_NBBY) | |
265 | return 0; | |
266 | ||
267 | /* Adjust for the size diff between gfs2_meta_header and gfs2_rgrp */ | |
268 | rbm->offset += (sizeof(struct gfs2_rgrp) - | |
269 | sizeof(struct gfs2_meta_header)) * GFS2_NBBY; | |
270 | x = rbm->offset / rbm->rgd->rd_sbd->sd_blocks_per_bitmap; | |
271 | rbm->offset -= x * rbm->rgd->rd_sbd->sd_blocks_per_bitmap; | |
272 | rbm->bi += x; | |
ff7f4cb4 SW |
273 | return 0; |
274 | } | |
275 | ||
276 | /** | |
277 | * gfs2_unaligned_extlen - Look for free blocks which are not byte aligned | |
278 | * @rbm: Position to search (value/result) | |
279 | * @n_unaligned: Number of unaligned blocks to check | |
280 | * @len: Decremented for each block found (terminate on zero) | |
281 | * | |
282 | * Returns: true if a non-free block is encountered | |
283 | */ | |
284 | ||
285 | static bool gfs2_unaligned_extlen(struct gfs2_rbm *rbm, u32 n_unaligned, u32 *len) | |
286 | { | |
287 | u64 block; | |
288 | u32 n; | |
289 | u8 res; | |
290 | ||
291 | for (n = 0; n < n_unaligned; n++) { | |
292 | res = gfs2_testbit(rbm); | |
293 | if (res != GFS2_BLKST_FREE) | |
294 | return true; | |
295 | (*len)--; | |
296 | if (*len == 0) | |
297 | return true; | |
298 | block = gfs2_rbm_to_block(rbm); | |
299 | if (gfs2_rbm_from_block(rbm, block + 1)) | |
300 | return true; | |
301 | } | |
302 | ||
303 | return false; | |
304 | } | |
305 | ||
306 | /** | |
307 | * gfs2_free_extlen - Return extent length of free blocks | |
308 | * @rbm: Starting position | |
309 | * @len: Max length to check | |
310 | * | |
311 | * Starting at the block specified by the rbm, see how many free blocks | |
312 | * there are, not reading more than len blocks ahead. This can be done | |
313 | * using memchr_inv when the blocks are byte aligned, but has to be done | |
314 | * on a block by block basis in case of unaligned blocks. Also this | |
315 | * function can cope with bitmap boundaries (although it must stop on | |
316 | * a resource group boundary) | |
317 | * | |
318 | * Returns: Number of free blocks in the extent | |
319 | */ | |
320 | ||
321 | static u32 gfs2_free_extlen(const struct gfs2_rbm *rrbm, u32 len) | |
322 | { | |
323 | struct gfs2_rbm rbm = *rrbm; | |
324 | u32 n_unaligned = rbm.offset & 3; | |
325 | u32 size = len; | |
326 | u32 bytes; | |
327 | u32 chunk_size; | |
328 | u8 *ptr, *start, *end; | |
329 | u64 block; | |
330 | ||
331 | if (n_unaligned && | |
332 | gfs2_unaligned_extlen(&rbm, 4 - n_unaligned, &len)) | |
333 | goto out; | |
334 | ||
3701530a | 335 | n_unaligned = len & 3; |
ff7f4cb4 SW |
336 | /* Start is now byte aligned */ |
337 | while (len > 3) { | |
338 | start = rbm.bi->bi_bh->b_data; | |
339 | if (rbm.bi->bi_clone) | |
340 | start = rbm.bi->bi_clone; | |
341 | end = start + rbm.bi->bi_bh->b_size; | |
342 | start += rbm.bi->bi_offset; | |
343 | BUG_ON(rbm.offset & 3); | |
344 | start += (rbm.offset / GFS2_NBBY); | |
345 | bytes = min_t(u32, len / GFS2_NBBY, (end - start)); | |
346 | ptr = memchr_inv(start, 0, bytes); | |
347 | chunk_size = ((ptr == NULL) ? bytes : (ptr - start)); | |
348 | chunk_size *= GFS2_NBBY; | |
349 | BUG_ON(len < chunk_size); | |
350 | len -= chunk_size; | |
351 | block = gfs2_rbm_to_block(&rbm); | |
352 | gfs2_rbm_from_block(&rbm, block + chunk_size); | |
353 | n_unaligned = 3; | |
354 | if (ptr) | |
355 | break; | |
356 | n_unaligned = len & 3; | |
357 | } | |
358 | ||
359 | /* Deal with any bits left over at the end */ | |
360 | if (n_unaligned) | |
361 | gfs2_unaligned_extlen(&rbm, n_unaligned, &len); | |
362 | out: | |
363 | return size - len; | |
364 | } | |
365 | ||
88c8ab1f SW |
366 | /** |
367 | * gfs2_bitcount - count the number of bits in a certain state | |
886b1416 | 368 | * @rgd: the resource group descriptor |
88c8ab1f SW |
369 | * @buffer: the buffer that holds the bitmaps |
370 | * @buflen: the length (in bytes) of the buffer | |
371 | * @state: the state of the block we're looking for | |
372 | * | |
373 | * Returns: The number of bits | |
374 | */ | |
375 | ||
110acf38 SW |
376 | static u32 gfs2_bitcount(struct gfs2_rgrpd *rgd, const u8 *buffer, |
377 | unsigned int buflen, u8 state) | |
88c8ab1f | 378 | { |
110acf38 SW |
379 | const u8 *byte = buffer; |
380 | const u8 *end = buffer + buflen; | |
381 | const u8 state1 = state << 2; | |
382 | const u8 state2 = state << 4; | |
383 | const u8 state3 = state << 6; | |
cd915493 | 384 | u32 count = 0; |
88c8ab1f SW |
385 | |
386 | for (; byte < end; byte++) { | |
387 | if (((*byte) & 0x03) == state) | |
388 | count++; | |
389 | if (((*byte) & 0x0C) == state1) | |
390 | count++; | |
391 | if (((*byte) & 0x30) == state2) | |
392 | count++; | |
393 | if (((*byte) & 0xC0) == state3) | |
394 | count++; | |
395 | } | |
396 | ||
397 | return count; | |
398 | } | |
399 | ||
b3b94faa DT |
400 | /** |
401 | * gfs2_rgrp_verify - Verify that a resource group is consistent | |
b3b94faa DT |
402 | * @rgd: the rgrp |
403 | * | |
404 | */ | |
405 | ||
406 | void gfs2_rgrp_verify(struct gfs2_rgrpd *rgd) | |
407 | { | |
408 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
409 | struct gfs2_bitmap *bi = NULL; | |
bb8d8a6f | 410 | u32 length = rgd->rd_length; |
cd915493 | 411 | u32 count[4], tmp; |
b3b94faa DT |
412 | int buf, x; |
413 | ||
cd915493 | 414 | memset(count, 0, 4 * sizeof(u32)); |
b3b94faa DT |
415 | |
416 | /* Count # blocks in each of 4 possible allocation states */ | |
417 | for (buf = 0; buf < length; buf++) { | |
418 | bi = rgd->rd_bits + buf; | |
419 | for (x = 0; x < 4; x++) | |
420 | count[x] += gfs2_bitcount(rgd, | |
421 | bi->bi_bh->b_data + | |
422 | bi->bi_offset, | |
423 | bi->bi_len, x); | |
424 | } | |
425 | ||
cfc8b549 | 426 | if (count[0] != rgd->rd_free) { |
b3b94faa DT |
427 | if (gfs2_consist_rgrpd(rgd)) |
428 | fs_err(sdp, "free data mismatch: %u != %u\n", | |
cfc8b549 | 429 | count[0], rgd->rd_free); |
b3b94faa DT |
430 | return; |
431 | } | |
432 | ||
73f74948 | 433 | tmp = rgd->rd_data - rgd->rd_free - rgd->rd_dinodes; |
6b946170 | 434 | if (count[1] != tmp) { |
b3b94faa DT |
435 | if (gfs2_consist_rgrpd(rgd)) |
436 | fs_err(sdp, "used data mismatch: %u != %u\n", | |
437 | count[1], tmp); | |
438 | return; | |
439 | } | |
440 | ||
6b946170 | 441 | if (count[2] + count[3] != rgd->rd_dinodes) { |
b3b94faa | 442 | if (gfs2_consist_rgrpd(rgd)) |
feaa7bba | 443 | fs_err(sdp, "used metadata mismatch: %u != %u\n", |
6b946170 | 444 | count[2] + count[3], rgd->rd_dinodes); |
b3b94faa DT |
445 | return; |
446 | } | |
b3b94faa DT |
447 | } |
448 | ||
bb8d8a6f | 449 | static inline int rgrp_contains_block(struct gfs2_rgrpd *rgd, u64 block) |
b3b94faa | 450 | { |
bb8d8a6f SW |
451 | u64 first = rgd->rd_data0; |
452 | u64 last = first + rgd->rd_data; | |
16910427 | 453 | return first <= block && block < last; |
b3b94faa DT |
454 | } |
455 | ||
456 | /** | |
457 | * gfs2_blk2rgrpd - Find resource group for a given data/meta block number | |
458 | * @sdp: The GFS2 superblock | |
886b1416 BP |
459 | * @blk: The data block number |
460 | * @exact: True if this needs to be an exact match | |
b3b94faa DT |
461 | * |
462 | * Returns: The resource group, or NULL if not found | |
463 | */ | |
464 | ||
66fc061b | 465 | struct gfs2_rgrpd *gfs2_blk2rgrpd(struct gfs2_sbd *sdp, u64 blk, bool exact) |
b3b94faa | 466 | { |
66fc061b | 467 | struct rb_node *n, *next; |
f75bbfb4 | 468 | struct gfs2_rgrpd *cur; |
b3b94faa DT |
469 | |
470 | spin_lock(&sdp->sd_rindex_spin); | |
66fc061b SW |
471 | n = sdp->sd_rindex_tree.rb_node; |
472 | while (n) { | |
473 | cur = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
474 | next = NULL; | |
7c9ca621 | 475 | if (blk < cur->rd_addr) |
66fc061b | 476 | next = n->rb_left; |
f75bbfb4 | 477 | else if (blk >= cur->rd_data0 + cur->rd_data) |
66fc061b SW |
478 | next = n->rb_right; |
479 | if (next == NULL) { | |
b3b94faa | 480 | spin_unlock(&sdp->sd_rindex_spin); |
66fc061b SW |
481 | if (exact) { |
482 | if (blk < cur->rd_addr) | |
483 | return NULL; | |
484 | if (blk >= cur->rd_data0 + cur->rd_data) | |
485 | return NULL; | |
486 | } | |
7c9ca621 | 487 | return cur; |
b3b94faa | 488 | } |
66fc061b | 489 | n = next; |
b3b94faa | 490 | } |
b3b94faa DT |
491 | spin_unlock(&sdp->sd_rindex_spin); |
492 | ||
493 | return NULL; | |
494 | } | |
495 | ||
496 | /** | |
497 | * gfs2_rgrpd_get_first - get the first Resource Group in the filesystem | |
498 | * @sdp: The GFS2 superblock | |
499 | * | |
500 | * Returns: The first rgrp in the filesystem | |
501 | */ | |
502 | ||
503 | struct gfs2_rgrpd *gfs2_rgrpd_get_first(struct gfs2_sbd *sdp) | |
504 | { | |
7c9ca621 BP |
505 | const struct rb_node *n; |
506 | struct gfs2_rgrpd *rgd; | |
507 | ||
8339ee54 | 508 | spin_lock(&sdp->sd_rindex_spin); |
7c9ca621 BP |
509 | n = rb_first(&sdp->sd_rindex_tree); |
510 | rgd = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
8339ee54 | 511 | spin_unlock(&sdp->sd_rindex_spin); |
7c9ca621 BP |
512 | |
513 | return rgd; | |
b3b94faa DT |
514 | } |
515 | ||
516 | /** | |
517 | * gfs2_rgrpd_get_next - get the next RG | |
886b1416 | 518 | * @rgd: the resource group descriptor |
b3b94faa DT |
519 | * |
520 | * Returns: The next rgrp | |
521 | */ | |
522 | ||
523 | struct gfs2_rgrpd *gfs2_rgrpd_get_next(struct gfs2_rgrpd *rgd) | |
524 | { | |
7c9ca621 BP |
525 | struct gfs2_sbd *sdp = rgd->rd_sbd; |
526 | const struct rb_node *n; | |
527 | ||
528 | spin_lock(&sdp->sd_rindex_spin); | |
529 | n = rb_next(&rgd->rd_node); | |
530 | if (n == NULL) | |
531 | n = rb_first(&sdp->sd_rindex_tree); | |
532 | ||
533 | if (unlikely(&rgd->rd_node == n)) { | |
534 | spin_unlock(&sdp->sd_rindex_spin); | |
b3b94faa | 535 | return NULL; |
7c9ca621 BP |
536 | } |
537 | rgd = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
538 | spin_unlock(&sdp->sd_rindex_spin); | |
539 | return rgd; | |
b3b94faa DT |
540 | } |
541 | ||
8339ee54 SW |
542 | void gfs2_free_clones(struct gfs2_rgrpd *rgd) |
543 | { | |
544 | int x; | |
545 | ||
546 | for (x = 0; x < rgd->rd_length; x++) { | |
547 | struct gfs2_bitmap *bi = rgd->rd_bits + x; | |
548 | kfree(bi->bi_clone); | |
549 | bi->bi_clone = NULL; | |
550 | } | |
551 | } | |
552 | ||
0a305e49 BP |
553 | /** |
554 | * gfs2_rs_alloc - make sure we have a reservation assigned to the inode | |
555 | * @ip: the inode for this reservation | |
556 | */ | |
557 | int gfs2_rs_alloc(struct gfs2_inode *ip) | |
558 | { | |
8e2e0047 BP |
559 | struct gfs2_blkreserv *res; |
560 | ||
561 | if (ip->i_res) | |
562 | return 0; | |
563 | ||
564 | res = kmem_cache_zalloc(gfs2_rsrv_cachep, GFP_NOFS); | |
565 | if (!res) | |
cd0ed19f | 566 | return -ENOMEM; |
0a305e49 | 567 | |
24d634e8 | 568 | RB_CLEAR_NODE(&res->rs_node); |
4a993fb1 | 569 | |
0a305e49 | 570 | down_write(&ip->i_rw_mutex); |
8e2e0047 BP |
571 | if (ip->i_res) |
572 | kmem_cache_free(gfs2_rsrv_cachep, res); | |
573 | else | |
574 | ip->i_res = res; | |
0a305e49 | 575 | up_write(&ip->i_rw_mutex); |
cd0ed19f | 576 | return 0; |
0a305e49 BP |
577 | } |
578 | ||
9e733d39 | 579 | static void dump_rs(struct seq_file *seq, const struct gfs2_blkreserv *rs) |
8e2e0047 | 580 | { |
9e733d39 SW |
581 | gfs2_print_dbg(seq, " B: n:%llu s:%llu b:%u f:%u\n", |
582 | (unsigned long long)rs->rs_inum, | |
583 | (unsigned long long)gfs2_rbm_to_block(&rs->rs_rbm), | |
4a993fb1 | 584 | rs->rs_rbm.offset, rs->rs_free); |
8e2e0047 BP |
585 | } |
586 | ||
0a305e49 | 587 | /** |
8e2e0047 BP |
588 | * __rs_deltree - remove a multi-block reservation from the rgd tree |
589 | * @rs: The reservation to remove | |
590 | * | |
591 | */ | |
4a993fb1 | 592 | static void __rs_deltree(struct gfs2_inode *ip, struct gfs2_blkreserv *rs) |
8e2e0047 BP |
593 | { |
594 | struct gfs2_rgrpd *rgd; | |
595 | ||
596 | if (!gfs2_rs_active(rs)) | |
597 | return; | |
598 | ||
4a993fb1 | 599 | rgd = rs->rs_rbm.rgd; |
9e733d39 | 600 | trace_gfs2_rs(rs, TRACE_RS_TREEDEL); |
4a993fb1 | 601 | rb_erase(&rs->rs_node, &rgd->rd_rstree); |
24d634e8 | 602 | RB_CLEAR_NODE(&rs->rs_node); |
8e2e0047 BP |
603 | |
604 | if (rs->rs_free) { | |
605 | /* return reserved blocks to the rgrp and the ip */ | |
4a993fb1 SW |
606 | BUG_ON(rs->rs_rbm.rgd->rd_reserved < rs->rs_free); |
607 | rs->rs_rbm.rgd->rd_reserved -= rs->rs_free; | |
8e2e0047 | 608 | rs->rs_free = 0; |
4a993fb1 | 609 | clear_bit(GBF_FULL, &rs->rs_rbm.bi->bi_flags); |
8e2e0047 BP |
610 | smp_mb__after_clear_bit(); |
611 | } | |
8e2e0047 BP |
612 | } |
613 | ||
614 | /** | |
615 | * gfs2_rs_deltree - remove a multi-block reservation from the rgd tree | |
616 | * @rs: The reservation to remove | |
617 | * | |
618 | */ | |
4a993fb1 | 619 | void gfs2_rs_deltree(struct gfs2_inode *ip, struct gfs2_blkreserv *rs) |
8e2e0047 BP |
620 | { |
621 | struct gfs2_rgrpd *rgd; | |
622 | ||
4a993fb1 SW |
623 | rgd = rs->rs_rbm.rgd; |
624 | if (rgd) { | |
625 | spin_lock(&rgd->rd_rsspin); | |
626 | __rs_deltree(ip, rs); | |
627 | spin_unlock(&rgd->rd_rsspin); | |
628 | } | |
8e2e0047 BP |
629 | } |
630 | ||
631 | /** | |
632 | * gfs2_rs_delete - delete a multi-block reservation | |
0a305e49 BP |
633 | * @ip: The inode for this reservation |
634 | * | |
635 | */ | |
636 | void gfs2_rs_delete(struct gfs2_inode *ip) | |
637 | { | |
638 | down_write(&ip->i_rw_mutex); | |
639 | if (ip->i_res) { | |
4a993fb1 | 640 | gfs2_rs_deltree(ip, ip->i_res); |
8e2e0047 | 641 | BUG_ON(ip->i_res->rs_free); |
0a305e49 BP |
642 | kmem_cache_free(gfs2_rsrv_cachep, ip->i_res); |
643 | ip->i_res = NULL; | |
644 | } | |
645 | up_write(&ip->i_rw_mutex); | |
646 | } | |
647 | ||
8e2e0047 BP |
648 | /** |
649 | * return_all_reservations - return all reserved blocks back to the rgrp. | |
650 | * @rgd: the rgrp that needs its space back | |
651 | * | |
652 | * We previously reserved a bunch of blocks for allocation. Now we need to | |
653 | * give them back. This leave the reservation structures in tact, but removes | |
654 | * all of their corresponding "no-fly zones". | |
655 | */ | |
656 | static void return_all_reservations(struct gfs2_rgrpd *rgd) | |
657 | { | |
658 | struct rb_node *n; | |
659 | struct gfs2_blkreserv *rs; | |
660 | ||
661 | spin_lock(&rgd->rd_rsspin); | |
662 | while ((n = rb_first(&rgd->rd_rstree))) { | |
663 | rs = rb_entry(n, struct gfs2_blkreserv, rs_node); | |
4a993fb1 | 664 | __rs_deltree(NULL, rs); |
8e2e0047 BP |
665 | } |
666 | spin_unlock(&rgd->rd_rsspin); | |
667 | } | |
668 | ||
8339ee54 | 669 | void gfs2_clear_rgrpd(struct gfs2_sbd *sdp) |
b3b94faa | 670 | { |
7c9ca621 | 671 | struct rb_node *n; |
b3b94faa DT |
672 | struct gfs2_rgrpd *rgd; |
673 | struct gfs2_glock *gl; | |
674 | ||
7c9ca621 BP |
675 | while ((n = rb_first(&sdp->sd_rindex_tree))) { |
676 | rgd = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
b3b94faa DT |
677 | gl = rgd->rd_gl; |
678 | ||
7c9ca621 | 679 | rb_erase(n, &sdp->sd_rindex_tree); |
b3b94faa DT |
680 | |
681 | if (gl) { | |
8339ee54 | 682 | spin_lock(&gl->gl_spin); |
5c676f6d | 683 | gl->gl_object = NULL; |
8339ee54 | 684 | spin_unlock(&gl->gl_spin); |
29687a2a | 685 | gfs2_glock_add_to_lru(gl); |
b3b94faa DT |
686 | gfs2_glock_put(gl); |
687 | } | |
688 | ||
8339ee54 | 689 | gfs2_free_clones(rgd); |
b3b94faa | 690 | kfree(rgd->rd_bits); |
8e2e0047 | 691 | return_all_reservations(rgd); |
6bdd9be6 | 692 | kmem_cache_free(gfs2_rgrpd_cachep, rgd); |
b3b94faa DT |
693 | } |
694 | } | |
695 | ||
bb8d8a6f SW |
696 | static void gfs2_rindex_print(const struct gfs2_rgrpd *rgd) |
697 | { | |
698 | printk(KERN_INFO " ri_addr = %llu\n", (unsigned long long)rgd->rd_addr); | |
699 | printk(KERN_INFO " ri_length = %u\n", rgd->rd_length); | |
700 | printk(KERN_INFO " ri_data0 = %llu\n", (unsigned long long)rgd->rd_data0); | |
701 | printk(KERN_INFO " ri_data = %u\n", rgd->rd_data); | |
702 | printk(KERN_INFO " ri_bitbytes = %u\n", rgd->rd_bitbytes); | |
703 | } | |
704 | ||
b3b94faa DT |
705 | /** |
706 | * gfs2_compute_bitstructs - Compute the bitmap sizes | |
707 | * @rgd: The resource group descriptor | |
708 | * | |
709 | * Calculates bitmap descriptors, one for each block that contains bitmap data | |
710 | * | |
711 | * Returns: errno | |
712 | */ | |
713 | ||
714 | static int compute_bitstructs(struct gfs2_rgrpd *rgd) | |
715 | { | |
716 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
717 | struct gfs2_bitmap *bi; | |
bb8d8a6f | 718 | u32 length = rgd->rd_length; /* # blocks in hdr & bitmap */ |
cd915493 | 719 | u32 bytes_left, bytes; |
b3b94faa DT |
720 | int x; |
721 | ||
feaa7bba SW |
722 | if (!length) |
723 | return -EINVAL; | |
724 | ||
dd894be8 | 725 | rgd->rd_bits = kcalloc(length, sizeof(struct gfs2_bitmap), GFP_NOFS); |
b3b94faa DT |
726 | if (!rgd->rd_bits) |
727 | return -ENOMEM; | |
728 | ||
bb8d8a6f | 729 | bytes_left = rgd->rd_bitbytes; |
b3b94faa DT |
730 | |
731 | for (x = 0; x < length; x++) { | |
732 | bi = rgd->rd_bits + x; | |
733 | ||
60a0b8f9 | 734 | bi->bi_flags = 0; |
b3b94faa DT |
735 | /* small rgrp; bitmap stored completely in header block */ |
736 | if (length == 1) { | |
737 | bytes = bytes_left; | |
738 | bi->bi_offset = sizeof(struct gfs2_rgrp); | |
739 | bi->bi_start = 0; | |
740 | bi->bi_len = bytes; | |
741 | /* header block */ | |
742 | } else if (x == 0) { | |
743 | bytes = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_rgrp); | |
744 | bi->bi_offset = sizeof(struct gfs2_rgrp); | |
745 | bi->bi_start = 0; | |
746 | bi->bi_len = bytes; | |
747 | /* last block */ | |
748 | } else if (x + 1 == length) { | |
749 | bytes = bytes_left; | |
750 | bi->bi_offset = sizeof(struct gfs2_meta_header); | |
bb8d8a6f | 751 | bi->bi_start = rgd->rd_bitbytes - bytes_left; |
b3b94faa DT |
752 | bi->bi_len = bytes; |
753 | /* other blocks */ | |
754 | } else { | |
568f4c96 SW |
755 | bytes = sdp->sd_sb.sb_bsize - |
756 | sizeof(struct gfs2_meta_header); | |
b3b94faa | 757 | bi->bi_offset = sizeof(struct gfs2_meta_header); |
bb8d8a6f | 758 | bi->bi_start = rgd->rd_bitbytes - bytes_left; |
b3b94faa DT |
759 | bi->bi_len = bytes; |
760 | } | |
761 | ||
762 | bytes_left -= bytes; | |
763 | } | |
764 | ||
765 | if (bytes_left) { | |
766 | gfs2_consist_rgrpd(rgd); | |
767 | return -EIO; | |
768 | } | |
769 | bi = rgd->rd_bits + (length - 1); | |
bb8d8a6f | 770 | if ((bi->bi_start + bi->bi_len) * GFS2_NBBY != rgd->rd_data) { |
b3b94faa | 771 | if (gfs2_consist_rgrpd(rgd)) { |
bb8d8a6f | 772 | gfs2_rindex_print(rgd); |
b3b94faa DT |
773 | fs_err(sdp, "start=%u len=%u offset=%u\n", |
774 | bi->bi_start, bi->bi_len, bi->bi_offset); | |
775 | } | |
776 | return -EIO; | |
777 | } | |
778 | ||
779 | return 0; | |
780 | } | |
781 | ||
7ae8fa84 RP |
782 | /** |
783 | * gfs2_ri_total - Total up the file system space, according to the rindex. | |
886b1416 | 784 | * @sdp: the filesystem |
7ae8fa84 RP |
785 | * |
786 | */ | |
787 | u64 gfs2_ri_total(struct gfs2_sbd *sdp) | |
788 | { | |
789 | u64 total_data = 0; | |
790 | struct inode *inode = sdp->sd_rindex; | |
791 | struct gfs2_inode *ip = GFS2_I(inode); | |
7ae8fa84 | 792 | char buf[sizeof(struct gfs2_rindex)]; |
7ae8fa84 RP |
793 | int error, rgrps; |
794 | ||
7ae8fa84 RP |
795 | for (rgrps = 0;; rgrps++) { |
796 | loff_t pos = rgrps * sizeof(struct gfs2_rindex); | |
797 | ||
bcd7278d | 798 | if (pos + sizeof(struct gfs2_rindex) > i_size_read(inode)) |
7ae8fa84 | 799 | break; |
4306629e | 800 | error = gfs2_internal_read(ip, buf, &pos, |
7ae8fa84 RP |
801 | sizeof(struct gfs2_rindex)); |
802 | if (error != sizeof(struct gfs2_rindex)) | |
803 | break; | |
bb8d8a6f | 804 | total_data += be32_to_cpu(((struct gfs2_rindex *)buf)->ri_data); |
7ae8fa84 | 805 | } |
7ae8fa84 RP |
806 | return total_data; |
807 | } | |
808 | ||
6aad1c3d | 809 | static int rgd_insert(struct gfs2_rgrpd *rgd) |
7c9ca621 BP |
810 | { |
811 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
812 | struct rb_node **newn = &sdp->sd_rindex_tree.rb_node, *parent = NULL; | |
813 | ||
814 | /* Figure out where to put new node */ | |
815 | while (*newn) { | |
816 | struct gfs2_rgrpd *cur = rb_entry(*newn, struct gfs2_rgrpd, | |
817 | rd_node); | |
818 | ||
819 | parent = *newn; | |
820 | if (rgd->rd_addr < cur->rd_addr) | |
821 | newn = &((*newn)->rb_left); | |
822 | else if (rgd->rd_addr > cur->rd_addr) | |
823 | newn = &((*newn)->rb_right); | |
824 | else | |
6aad1c3d | 825 | return -EEXIST; |
7c9ca621 BP |
826 | } |
827 | ||
828 | rb_link_node(&rgd->rd_node, parent, newn); | |
829 | rb_insert_color(&rgd->rd_node, &sdp->sd_rindex_tree); | |
6aad1c3d BP |
830 | sdp->sd_rgrps++; |
831 | return 0; | |
7c9ca621 BP |
832 | } |
833 | ||
b3b94faa | 834 | /** |
6c53267f | 835 | * read_rindex_entry - Pull in a new resource index entry from the disk |
4306629e | 836 | * @ip: Pointer to the rindex inode |
b3b94faa | 837 | * |
8339ee54 | 838 | * Returns: 0 on success, > 0 on EOF, error code otherwise |
6c53267f RP |
839 | */ |
840 | ||
4306629e | 841 | static int read_rindex_entry(struct gfs2_inode *ip) |
6c53267f RP |
842 | { |
843 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
844 | loff_t pos = sdp->sd_rgrps * sizeof(struct gfs2_rindex); | |
8339ee54 | 845 | struct gfs2_rindex buf; |
6c53267f RP |
846 | int error; |
847 | struct gfs2_rgrpd *rgd; | |
848 | ||
8339ee54 SW |
849 | if (pos >= i_size_read(&ip->i_inode)) |
850 | return 1; | |
851 | ||
4306629e | 852 | error = gfs2_internal_read(ip, (char *)&buf, &pos, |
6c53267f | 853 | sizeof(struct gfs2_rindex)); |
8339ee54 SW |
854 | |
855 | if (error != sizeof(struct gfs2_rindex)) | |
856 | return (error == 0) ? 1 : error; | |
6c53267f | 857 | |
6bdd9be6 | 858 | rgd = kmem_cache_zalloc(gfs2_rgrpd_cachep, GFP_NOFS); |
6c53267f RP |
859 | error = -ENOMEM; |
860 | if (!rgd) | |
861 | return error; | |
862 | ||
6c53267f | 863 | rgd->rd_sbd = sdp; |
8339ee54 SW |
864 | rgd->rd_addr = be64_to_cpu(buf.ri_addr); |
865 | rgd->rd_length = be32_to_cpu(buf.ri_length); | |
866 | rgd->rd_data0 = be64_to_cpu(buf.ri_data0); | |
867 | rgd->rd_data = be32_to_cpu(buf.ri_data); | |
868 | rgd->rd_bitbytes = be32_to_cpu(buf.ri_bitbytes); | |
8e2e0047 | 869 | spin_lock_init(&rgd->rd_rsspin); |
7c9ca621 | 870 | |
6c53267f RP |
871 | error = compute_bitstructs(rgd); |
872 | if (error) | |
8339ee54 | 873 | goto fail; |
6c53267f | 874 | |
bb8d8a6f | 875 | error = gfs2_glock_get(sdp, rgd->rd_addr, |
6c53267f RP |
876 | &gfs2_rgrp_glops, CREATE, &rgd->rd_gl); |
877 | if (error) | |
8339ee54 | 878 | goto fail; |
6c53267f RP |
879 | |
880 | rgd->rd_gl->gl_object = rgd; | |
90306c41 | 881 | rgd->rd_rgl = (struct gfs2_rgrp_lvb *)rgd->rd_gl->gl_lvb; |
cf45b752 | 882 | rgd->rd_flags &= ~GFS2_RDF_UPTODATE; |
7c9ca621 BP |
883 | if (rgd->rd_data > sdp->sd_max_rg_data) |
884 | sdp->sd_max_rg_data = rgd->rd_data; | |
8339ee54 | 885 | spin_lock(&sdp->sd_rindex_spin); |
6aad1c3d | 886 | error = rgd_insert(rgd); |
8339ee54 | 887 | spin_unlock(&sdp->sd_rindex_spin); |
6aad1c3d BP |
888 | if (!error) |
889 | return 0; | |
890 | ||
891 | error = 0; /* someone else read in the rgrp; free it and ignore it */ | |
c1ac539e | 892 | gfs2_glock_put(rgd->rd_gl); |
8339ee54 SW |
893 | |
894 | fail: | |
895 | kfree(rgd->rd_bits); | |
896 | kmem_cache_free(gfs2_rgrpd_cachep, rgd); | |
6c53267f RP |
897 | return error; |
898 | } | |
899 | ||
900 | /** | |
901 | * gfs2_ri_update - Pull in a new resource index from the disk | |
902 | * @ip: pointer to the rindex inode | |
903 | * | |
b3b94faa DT |
904 | * Returns: 0 on successful update, error code otherwise |
905 | */ | |
906 | ||
8339ee54 | 907 | static int gfs2_ri_update(struct gfs2_inode *ip) |
b3b94faa | 908 | { |
feaa7bba | 909 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
910 | int error; |
911 | ||
8339ee54 | 912 | do { |
4306629e | 913 | error = read_rindex_entry(ip); |
8339ee54 SW |
914 | } while (error == 0); |
915 | ||
916 | if (error < 0) | |
917 | return error; | |
b3b94faa | 918 | |
cf45b752 | 919 | sdp->sd_rindex_uptodate = 1; |
6c53267f RP |
920 | return 0; |
921 | } | |
b3b94faa | 922 | |
b3b94faa | 923 | /** |
8339ee54 | 924 | * gfs2_rindex_update - Update the rindex if required |
b3b94faa | 925 | * @sdp: The GFS2 superblock |
b3b94faa DT |
926 | * |
927 | * We grab a lock on the rindex inode to make sure that it doesn't | |
928 | * change whilst we are performing an operation. We keep this lock | |
929 | * for quite long periods of time compared to other locks. This | |
930 | * doesn't matter, since it is shared and it is very, very rarely | |
931 | * accessed in the exclusive mode (i.e. only when expanding the filesystem). | |
932 | * | |
933 | * This makes sure that we're using the latest copy of the resource index | |
934 | * special file, which might have been updated if someone expanded the | |
935 | * filesystem (via gfs2_grow utility), which adds new resource groups. | |
936 | * | |
8339ee54 | 937 | * Returns: 0 on succeess, error code otherwise |
b3b94faa DT |
938 | */ |
939 | ||
8339ee54 | 940 | int gfs2_rindex_update(struct gfs2_sbd *sdp) |
b3b94faa | 941 | { |
feaa7bba | 942 | struct gfs2_inode *ip = GFS2_I(sdp->sd_rindex); |
b3b94faa | 943 | struct gfs2_glock *gl = ip->i_gl; |
8339ee54 SW |
944 | struct gfs2_holder ri_gh; |
945 | int error = 0; | |
a365fbf3 | 946 | int unlock_required = 0; |
b3b94faa DT |
947 | |
948 | /* Read new copy from disk if we don't have the latest */ | |
cf45b752 | 949 | if (!sdp->sd_rindex_uptodate) { |
a365fbf3 SW |
950 | if (!gfs2_glock_is_locked_by_me(gl)) { |
951 | error = gfs2_glock_nq_init(gl, LM_ST_SHARED, 0, &ri_gh); | |
952 | if (error) | |
6aad1c3d | 953 | return error; |
a365fbf3 SW |
954 | unlock_required = 1; |
955 | } | |
8339ee54 | 956 | if (!sdp->sd_rindex_uptodate) |
b3b94faa | 957 | error = gfs2_ri_update(ip); |
a365fbf3 SW |
958 | if (unlock_required) |
959 | gfs2_glock_dq_uninit(&ri_gh); | |
b3b94faa DT |
960 | } |
961 | ||
962 | return error; | |
963 | } | |
964 | ||
42d52e38 | 965 | static void gfs2_rgrp_in(struct gfs2_rgrpd *rgd, const void *buf) |
bb8d8a6f SW |
966 | { |
967 | const struct gfs2_rgrp *str = buf; | |
42d52e38 | 968 | u32 rg_flags; |
bb8d8a6f | 969 | |
42d52e38 | 970 | rg_flags = be32_to_cpu(str->rg_flags); |
09010978 | 971 | rg_flags &= ~GFS2_RDF_MASK; |
1ce97e56 SW |
972 | rgd->rd_flags &= GFS2_RDF_MASK; |
973 | rgd->rd_flags |= rg_flags; | |
cfc8b549 | 974 | rgd->rd_free = be32_to_cpu(str->rg_free); |
73f74948 | 975 | rgd->rd_dinodes = be32_to_cpu(str->rg_dinodes); |
d8b71f73 | 976 | rgd->rd_igeneration = be64_to_cpu(str->rg_igeneration); |
bb8d8a6f SW |
977 | } |
978 | ||
42d52e38 | 979 | static void gfs2_rgrp_out(struct gfs2_rgrpd *rgd, void *buf) |
bb8d8a6f SW |
980 | { |
981 | struct gfs2_rgrp *str = buf; | |
982 | ||
09010978 | 983 | str->rg_flags = cpu_to_be32(rgd->rd_flags & ~GFS2_RDF_MASK); |
cfc8b549 | 984 | str->rg_free = cpu_to_be32(rgd->rd_free); |
73f74948 | 985 | str->rg_dinodes = cpu_to_be32(rgd->rd_dinodes); |
bb8d8a6f | 986 | str->__pad = cpu_to_be32(0); |
d8b71f73 | 987 | str->rg_igeneration = cpu_to_be64(rgd->rd_igeneration); |
bb8d8a6f SW |
988 | memset(&str->rg_reserved, 0, sizeof(str->rg_reserved)); |
989 | } | |
990 | ||
90306c41 BM |
991 | static int gfs2_rgrp_lvb_valid(struct gfs2_rgrpd *rgd) |
992 | { | |
993 | struct gfs2_rgrp_lvb *rgl = rgd->rd_rgl; | |
994 | struct gfs2_rgrp *str = (struct gfs2_rgrp *)rgd->rd_bits[0].bi_bh->b_data; | |
995 | ||
996 | if (rgl->rl_flags != str->rg_flags || rgl->rl_free != str->rg_free || | |
997 | rgl->rl_dinodes != str->rg_dinodes || | |
998 | rgl->rl_igeneration != str->rg_igeneration) | |
999 | return 0; | |
1000 | return 1; | |
1001 | } | |
1002 | ||
1003 | static void gfs2_rgrp_ondisk2lvb(struct gfs2_rgrp_lvb *rgl, const void *buf) | |
1004 | { | |
1005 | const struct gfs2_rgrp *str = buf; | |
1006 | ||
1007 | rgl->rl_magic = cpu_to_be32(GFS2_MAGIC); | |
1008 | rgl->rl_flags = str->rg_flags; | |
1009 | rgl->rl_free = str->rg_free; | |
1010 | rgl->rl_dinodes = str->rg_dinodes; | |
1011 | rgl->rl_igeneration = str->rg_igeneration; | |
1012 | rgl->__pad = 0UL; | |
1013 | } | |
1014 | ||
1015 | static void update_rgrp_lvb_unlinked(struct gfs2_rgrpd *rgd, u32 change) | |
1016 | { | |
1017 | struct gfs2_rgrp_lvb *rgl = rgd->rd_rgl; | |
1018 | u32 unlinked = be32_to_cpu(rgl->rl_unlinked) + change; | |
1019 | rgl->rl_unlinked = cpu_to_be32(unlinked); | |
1020 | } | |
1021 | ||
1022 | static u32 count_unlinked(struct gfs2_rgrpd *rgd) | |
1023 | { | |
1024 | struct gfs2_bitmap *bi; | |
1025 | const u32 length = rgd->rd_length; | |
1026 | const u8 *buffer = NULL; | |
1027 | u32 i, goal, count = 0; | |
1028 | ||
1029 | for (i = 0, bi = rgd->rd_bits; i < length; i++, bi++) { | |
1030 | goal = 0; | |
1031 | buffer = bi->bi_bh->b_data + bi->bi_offset; | |
1032 | WARN_ON(!buffer_uptodate(bi->bi_bh)); | |
1033 | while (goal < bi->bi_len * GFS2_NBBY) { | |
1034 | goal = gfs2_bitfit(buffer, bi->bi_len, goal, | |
1035 | GFS2_BLKST_UNLINKED); | |
1036 | if (goal == BFITNOENT) | |
1037 | break; | |
1038 | count++; | |
1039 | goal++; | |
1040 | } | |
1041 | } | |
1042 | ||
1043 | return count; | |
1044 | } | |
1045 | ||
1046 | ||
b3b94faa | 1047 | /** |
90306c41 BM |
1048 | * gfs2_rgrp_bh_get - Read in a RG's header and bitmaps |
1049 | * @rgd: the struct gfs2_rgrpd describing the RG to read in | |
b3b94faa DT |
1050 | * |
1051 | * Read in all of a Resource Group's header and bitmap blocks. | |
1052 | * Caller must eventually call gfs2_rgrp_relse() to free the bitmaps. | |
1053 | * | |
1054 | * Returns: errno | |
1055 | */ | |
1056 | ||
90306c41 | 1057 | int gfs2_rgrp_bh_get(struct gfs2_rgrpd *rgd) |
b3b94faa DT |
1058 | { |
1059 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
1060 | struct gfs2_glock *gl = rgd->rd_gl; | |
bb8d8a6f | 1061 | unsigned int length = rgd->rd_length; |
b3b94faa DT |
1062 | struct gfs2_bitmap *bi; |
1063 | unsigned int x, y; | |
1064 | int error; | |
1065 | ||
90306c41 BM |
1066 | if (rgd->rd_bits[0].bi_bh != NULL) |
1067 | return 0; | |
1068 | ||
b3b94faa DT |
1069 | for (x = 0; x < length; x++) { |
1070 | bi = rgd->rd_bits + x; | |
bb8d8a6f | 1071 | error = gfs2_meta_read(gl, rgd->rd_addr + x, 0, &bi->bi_bh); |
b3b94faa DT |
1072 | if (error) |
1073 | goto fail; | |
1074 | } | |
1075 | ||
1076 | for (y = length; y--;) { | |
1077 | bi = rgd->rd_bits + y; | |
7276b3b0 | 1078 | error = gfs2_meta_wait(sdp, bi->bi_bh); |
b3b94faa DT |
1079 | if (error) |
1080 | goto fail; | |
feaa7bba | 1081 | if (gfs2_metatype_check(sdp, bi->bi_bh, y ? GFS2_METATYPE_RB : |
b3b94faa DT |
1082 | GFS2_METATYPE_RG)) { |
1083 | error = -EIO; | |
1084 | goto fail; | |
1085 | } | |
1086 | } | |
1087 | ||
cf45b752 | 1088 | if (!(rgd->rd_flags & GFS2_RDF_UPTODATE)) { |
60a0b8f9 SW |
1089 | for (x = 0; x < length; x++) |
1090 | clear_bit(GBF_FULL, &rgd->rd_bits[x].bi_flags); | |
42d52e38 | 1091 | gfs2_rgrp_in(rgd, (rgd->rd_bits[0].bi_bh)->b_data); |
1ce97e56 | 1092 | rgd->rd_flags |= (GFS2_RDF_UPTODATE | GFS2_RDF_CHECK); |
7c9ca621 | 1093 | rgd->rd_free_clone = rgd->rd_free; |
b3b94faa | 1094 | } |
90306c41 BM |
1095 | if (be32_to_cpu(GFS2_MAGIC) != rgd->rd_rgl->rl_magic) { |
1096 | rgd->rd_rgl->rl_unlinked = cpu_to_be32(count_unlinked(rgd)); | |
1097 | gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, | |
1098 | rgd->rd_bits[0].bi_bh->b_data); | |
1099 | } | |
1100 | else if (sdp->sd_args.ar_rgrplvb) { | |
1101 | if (!gfs2_rgrp_lvb_valid(rgd)){ | |
1102 | gfs2_consist_rgrpd(rgd); | |
1103 | error = -EIO; | |
1104 | goto fail; | |
1105 | } | |
1106 | if (rgd->rd_rgl->rl_unlinked == 0) | |
1107 | rgd->rd_flags &= ~GFS2_RDF_CHECK; | |
1108 | } | |
b3b94faa DT |
1109 | return 0; |
1110 | ||
feaa7bba | 1111 | fail: |
b3b94faa DT |
1112 | while (x--) { |
1113 | bi = rgd->rd_bits + x; | |
1114 | brelse(bi->bi_bh); | |
1115 | bi->bi_bh = NULL; | |
1116 | gfs2_assert_warn(sdp, !bi->bi_clone); | |
1117 | } | |
b3b94faa DT |
1118 | |
1119 | return error; | |
1120 | } | |
1121 | ||
90306c41 BM |
1122 | int update_rgrp_lvb(struct gfs2_rgrpd *rgd) |
1123 | { | |
1124 | u32 rl_flags; | |
1125 | ||
1126 | if (rgd->rd_flags & GFS2_RDF_UPTODATE) | |
1127 | return 0; | |
1128 | ||
1129 | if (be32_to_cpu(GFS2_MAGIC) != rgd->rd_rgl->rl_magic) | |
1130 | return gfs2_rgrp_bh_get(rgd); | |
1131 | ||
1132 | rl_flags = be32_to_cpu(rgd->rd_rgl->rl_flags); | |
1133 | rl_flags &= ~GFS2_RDF_MASK; | |
1134 | rgd->rd_flags &= GFS2_RDF_MASK; | |
1135 | rgd->rd_flags |= (rl_flags | GFS2_RDF_UPTODATE | GFS2_RDF_CHECK); | |
1136 | if (rgd->rd_rgl->rl_unlinked == 0) | |
1137 | rgd->rd_flags &= ~GFS2_RDF_CHECK; | |
1138 | rgd->rd_free = be32_to_cpu(rgd->rd_rgl->rl_free); | |
1139 | rgd->rd_free_clone = rgd->rd_free; | |
1140 | rgd->rd_dinodes = be32_to_cpu(rgd->rd_rgl->rl_dinodes); | |
1141 | rgd->rd_igeneration = be64_to_cpu(rgd->rd_rgl->rl_igeneration); | |
1142 | return 0; | |
1143 | } | |
1144 | ||
1145 | int gfs2_rgrp_go_lock(struct gfs2_holder *gh) | |
1146 | { | |
1147 | struct gfs2_rgrpd *rgd = gh->gh_gl->gl_object; | |
1148 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
1149 | ||
1150 | if (gh->gh_flags & GL_SKIP && sdp->sd_args.ar_rgrplvb) | |
1151 | return 0; | |
1152 | return gfs2_rgrp_bh_get((struct gfs2_rgrpd *)gh->gh_gl->gl_object); | |
1153 | } | |
1154 | ||
b3b94faa | 1155 | /** |
7c9ca621 | 1156 | * gfs2_rgrp_go_unlock - Release RG bitmaps read in with gfs2_rgrp_bh_get() |
886b1416 | 1157 | * @gh: The glock holder for the resource group |
b3b94faa DT |
1158 | * |
1159 | */ | |
1160 | ||
7c9ca621 | 1161 | void gfs2_rgrp_go_unlock(struct gfs2_holder *gh) |
b3b94faa | 1162 | { |
7c9ca621 | 1163 | struct gfs2_rgrpd *rgd = gh->gh_gl->gl_object; |
bb8d8a6f | 1164 | int x, length = rgd->rd_length; |
b3b94faa | 1165 | |
b3b94faa DT |
1166 | for (x = 0; x < length; x++) { |
1167 | struct gfs2_bitmap *bi = rgd->rd_bits + x; | |
90306c41 BM |
1168 | if (bi->bi_bh) { |
1169 | brelse(bi->bi_bh); | |
1170 | bi->bi_bh = NULL; | |
1171 | } | |
b3b94faa DT |
1172 | } |
1173 | ||
b3b94faa DT |
1174 | } |
1175 | ||
66fc061b | 1176 | int gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset, |
7c9ca621 | 1177 | struct buffer_head *bh, |
66fc061b | 1178 | const struct gfs2_bitmap *bi, unsigned minlen, u64 *ptrimmed) |
f15ab561 SW |
1179 | { |
1180 | struct super_block *sb = sdp->sd_vfs; | |
1181 | struct block_device *bdev = sb->s_bdev; | |
1182 | const unsigned int sects_per_blk = sdp->sd_sb.sb_bsize / | |
e1defc4f | 1183 | bdev_logical_block_size(sb->s_bdev); |
f15ab561 | 1184 | u64 blk; |
64d576ba | 1185 | sector_t start = 0; |
f15ab561 SW |
1186 | sector_t nr_sects = 0; |
1187 | int rv; | |
1188 | unsigned int x; | |
66fc061b SW |
1189 | u32 trimmed = 0; |
1190 | u8 diff; | |
f15ab561 SW |
1191 | |
1192 | for (x = 0; x < bi->bi_len; x++) { | |
66fc061b SW |
1193 | const u8 *clone = bi->bi_clone ? bi->bi_clone : bi->bi_bh->b_data; |
1194 | clone += bi->bi_offset; | |
1195 | clone += x; | |
1196 | if (bh) { | |
1197 | const u8 *orig = bh->b_data + bi->bi_offset + x; | |
1198 | diff = ~(*orig | (*orig >> 1)) & (*clone | (*clone >> 1)); | |
1199 | } else { | |
1200 | diff = ~(*clone | (*clone >> 1)); | |
1201 | } | |
f15ab561 SW |
1202 | diff &= 0x55; |
1203 | if (diff == 0) | |
1204 | continue; | |
1205 | blk = offset + ((bi->bi_start + x) * GFS2_NBBY); | |
1206 | blk *= sects_per_blk; /* convert to sectors */ | |
1207 | while(diff) { | |
1208 | if (diff & 1) { | |
1209 | if (nr_sects == 0) | |
1210 | goto start_new_extent; | |
1211 | if ((start + nr_sects) != blk) { | |
66fc061b SW |
1212 | if (nr_sects >= minlen) { |
1213 | rv = blkdev_issue_discard(bdev, | |
1214 | start, nr_sects, | |
1215 | GFP_NOFS, 0); | |
1216 | if (rv) | |
1217 | goto fail; | |
1218 | trimmed += nr_sects; | |
1219 | } | |
f15ab561 SW |
1220 | nr_sects = 0; |
1221 | start_new_extent: | |
1222 | start = blk; | |
1223 | } | |
1224 | nr_sects += sects_per_blk; | |
1225 | } | |
1226 | diff >>= 2; | |
1227 | blk += sects_per_blk; | |
1228 | } | |
1229 | } | |
66fc061b | 1230 | if (nr_sects >= minlen) { |
dd3932ed | 1231 | rv = blkdev_issue_discard(bdev, start, nr_sects, GFP_NOFS, 0); |
f15ab561 SW |
1232 | if (rv) |
1233 | goto fail; | |
66fc061b | 1234 | trimmed += nr_sects; |
f15ab561 | 1235 | } |
66fc061b SW |
1236 | if (ptrimmed) |
1237 | *ptrimmed = trimmed; | |
1238 | return 0; | |
1239 | ||
f15ab561 | 1240 | fail: |
66fc061b SW |
1241 | if (sdp->sd_args.ar_discard) |
1242 | fs_warn(sdp, "error %d on discard request, turning discards off for this filesystem", rv); | |
f15ab561 | 1243 | sdp->sd_args.ar_discard = 0; |
66fc061b SW |
1244 | return -EIO; |
1245 | } | |
1246 | ||
1247 | /** | |
1248 | * gfs2_fitrim - Generate discard requests for unused bits of the filesystem | |
1249 | * @filp: Any file on the filesystem | |
1250 | * @argp: Pointer to the arguments (also used to pass result) | |
1251 | * | |
1252 | * Returns: 0 on success, otherwise error code | |
1253 | */ | |
1254 | ||
1255 | int gfs2_fitrim(struct file *filp, void __user *argp) | |
1256 | { | |
1257 | struct inode *inode = filp->f_dentry->d_inode; | |
1258 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
1259 | struct request_queue *q = bdev_get_queue(sdp->sd_vfs->s_bdev); | |
1260 | struct buffer_head *bh; | |
1261 | struct gfs2_rgrpd *rgd; | |
1262 | struct gfs2_rgrpd *rgd_end; | |
1263 | struct gfs2_holder gh; | |
1264 | struct fstrim_range r; | |
1265 | int ret = 0; | |
1266 | u64 amt; | |
1267 | u64 trimmed = 0; | |
076f0faa | 1268 | u64 start, end, minlen; |
66fc061b | 1269 | unsigned int x; |
076f0faa | 1270 | unsigned bs_shift = sdp->sd_sb.sb_bsize_shift; |
66fc061b SW |
1271 | |
1272 | if (!capable(CAP_SYS_ADMIN)) | |
1273 | return -EPERM; | |
1274 | ||
1275 | if (!blk_queue_discard(q)) | |
1276 | return -EOPNOTSUPP; | |
1277 | ||
3a238ade | 1278 | if (copy_from_user(&r, argp, sizeof(r))) |
66fc061b SW |
1279 | return -EFAULT; |
1280 | ||
5e2f7d61 BP |
1281 | ret = gfs2_rindex_update(sdp); |
1282 | if (ret) | |
1283 | return ret; | |
1284 | ||
076f0faa LC |
1285 | start = r.start >> bs_shift; |
1286 | end = start + (r.len >> bs_shift); | |
1287 | minlen = max_t(u64, r.minlen, | |
1288 | q->limits.discard_granularity) >> bs_shift; | |
1289 | ||
1290 | rgd = gfs2_blk2rgrpd(sdp, start, 0); | |
1291 | rgd_end = gfs2_blk2rgrpd(sdp, end - 1, 0); | |
1292 | ||
1293 | if (end <= start || | |
1294 | minlen > sdp->sd_max_rg_data || | |
1295 | start > rgd_end->rd_data0 + rgd_end->rd_data) | |
1296 | return -EINVAL; | |
66fc061b SW |
1297 | |
1298 | while (1) { | |
1299 | ||
1300 | ret = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
1301 | if (ret) | |
1302 | goto out; | |
1303 | ||
1304 | if (!(rgd->rd_flags & GFS2_RGF_TRIMMED)) { | |
1305 | /* Trim each bitmap in the rgrp */ | |
1306 | for (x = 0; x < rgd->rd_length; x++) { | |
1307 | struct gfs2_bitmap *bi = rgd->rd_bits + x; | |
076f0faa LC |
1308 | ret = gfs2_rgrp_send_discards(sdp, |
1309 | rgd->rd_data0, NULL, bi, minlen, | |
1310 | &amt); | |
66fc061b SW |
1311 | if (ret) { |
1312 | gfs2_glock_dq_uninit(&gh); | |
1313 | goto out; | |
1314 | } | |
1315 | trimmed += amt; | |
1316 | } | |
1317 | ||
1318 | /* Mark rgrp as having been trimmed */ | |
1319 | ret = gfs2_trans_begin(sdp, RES_RG_HDR, 0); | |
1320 | if (ret == 0) { | |
1321 | bh = rgd->rd_bits[0].bi_bh; | |
1322 | rgd->rd_flags |= GFS2_RGF_TRIMMED; | |
1323 | gfs2_trans_add_bh(rgd->rd_gl, bh, 1); | |
1324 | gfs2_rgrp_out(rgd, bh->b_data); | |
90306c41 | 1325 | gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, bh->b_data); |
66fc061b SW |
1326 | gfs2_trans_end(sdp); |
1327 | } | |
1328 | } | |
1329 | gfs2_glock_dq_uninit(&gh); | |
1330 | ||
1331 | if (rgd == rgd_end) | |
1332 | break; | |
1333 | ||
1334 | rgd = gfs2_rgrpd_get_next(rgd); | |
1335 | } | |
1336 | ||
1337 | out: | |
1338 | r.len = trimmed << 9; | |
3a238ade | 1339 | if (copy_to_user(argp, &r, sizeof(r))) |
66fc061b SW |
1340 | return -EFAULT; |
1341 | ||
1342 | return ret; | |
f15ab561 SW |
1343 | } |
1344 | ||
8e2e0047 BP |
1345 | /** |
1346 | * rs_insert - insert a new multi-block reservation into the rgrp's rb_tree | |
8e2e0047 | 1347 | * @ip: the inode structure |
8e2e0047 | 1348 | * |
8e2e0047 | 1349 | */ |
ff7f4cb4 | 1350 | static void rs_insert(struct gfs2_inode *ip) |
8e2e0047 BP |
1351 | { |
1352 | struct rb_node **newn, *parent = NULL; | |
1353 | int rc; | |
1354 | struct gfs2_blkreserv *rs = ip->i_res; | |
4a993fb1 | 1355 | struct gfs2_rgrpd *rgd = rs->rs_rbm.rgd; |
ff7f4cb4 | 1356 | u64 fsblock = gfs2_rbm_to_block(&rs->rs_rbm); |
8e2e0047 | 1357 | |
8e2e0047 | 1358 | BUG_ON(gfs2_rs_active(rs)); |
c743ffd0 | 1359 | |
ff7f4cb4 SW |
1360 | spin_lock(&rgd->rd_rsspin); |
1361 | newn = &rgd->rd_rstree.rb_node; | |
8e2e0047 BP |
1362 | while (*newn) { |
1363 | struct gfs2_blkreserv *cur = | |
1364 | rb_entry(*newn, struct gfs2_blkreserv, rs_node); | |
1365 | ||
1366 | parent = *newn; | |
ff7f4cb4 | 1367 | rc = rs_cmp(fsblock, rs->rs_free, cur); |
8e2e0047 BP |
1368 | if (rc > 0) |
1369 | newn = &((*newn)->rb_right); | |
1370 | else if (rc < 0) | |
1371 | newn = &((*newn)->rb_left); | |
1372 | else { | |
1373 | spin_unlock(&rgd->rd_rsspin); | |
ff7f4cb4 SW |
1374 | WARN_ON(1); |
1375 | return; | |
8e2e0047 BP |
1376 | } |
1377 | } | |
1378 | ||
8e2e0047 BP |
1379 | rb_link_node(&rs->rs_node, parent, newn); |
1380 | rb_insert_color(&rs->rs_node, &rgd->rd_rstree); | |
1381 | ||
8e2e0047 | 1382 | /* Do our rgrp accounting for the reservation */ |
ff7f4cb4 | 1383 | rgd->rd_reserved += rs->rs_free; /* blocks reserved */ |
8e2e0047 | 1384 | spin_unlock(&rgd->rd_rsspin); |
9e733d39 | 1385 | trace_gfs2_rs(rs, TRACE_RS_INSERT); |
8e2e0047 BP |
1386 | } |
1387 | ||
1388 | /** | |
ff7f4cb4 | 1389 | * rg_mblk_search - find a group of multiple free blocks to form a reservation |
8e2e0047 | 1390 | * @rgd: the resource group descriptor |
8e2e0047 | 1391 | * @ip: pointer to the inode for which we're reserving blocks |
c743ffd0 | 1392 | * @requested: number of blocks required for this allocation |
8e2e0047 | 1393 | * |
8e2e0047 BP |
1394 | */ |
1395 | ||
ff7f4cb4 SW |
1396 | static void rg_mblk_search(struct gfs2_rgrpd *rgd, struct gfs2_inode *ip, |
1397 | unsigned requested) | |
8e2e0047 | 1398 | { |
ff7f4cb4 SW |
1399 | struct gfs2_rbm rbm = { .rgd = rgd, }; |
1400 | u64 goal; | |
1401 | struct gfs2_blkreserv *rs = ip->i_res; | |
1402 | u32 extlen; | |
1403 | u32 free_blocks = rgd->rd_free_clone - rgd->rd_reserved; | |
1404 | int ret; | |
8e2e0047 | 1405 | |
ff7f4cb4 SW |
1406 | extlen = max_t(u32, atomic_read(&rs->rs_sizehint), requested); |
1407 | extlen = clamp(extlen, RGRP_RSRV_MINBLKS, free_blocks); | |
1408 | if ((rgd->rd_free_clone < rgd->rd_reserved) || (free_blocks < extlen)) | |
c743ffd0 SW |
1409 | return; |
1410 | ||
8e2e0047 BP |
1411 | /* Find bitmap block that contains bits for goal block */ |
1412 | if (rgrp_contains_block(rgd, ip->i_goal)) | |
ff7f4cb4 | 1413 | goal = ip->i_goal; |
8e2e0047 | 1414 | else |
ff7f4cb4 | 1415 | goal = rgd->rd_last_alloc + rgd->rd_data0; |
8e2e0047 | 1416 | |
ff7f4cb4 SW |
1417 | if (WARN_ON(gfs2_rbm_from_block(&rbm, goal))) |
1418 | return; | |
8e2e0047 | 1419 | |
ff7f4cb4 SW |
1420 | ret = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, extlen, ip, true); |
1421 | if (ret == 0) { | |
1422 | rs->rs_rbm = rbm; | |
1423 | rs->rs_free = extlen; | |
1424 | rs->rs_inum = ip->i_no_addr; | |
1425 | rs_insert(ip); | |
8e2e0047 | 1426 | } |
b3e47ca0 BP |
1427 | } |
1428 | ||
5b924ae2 SW |
1429 | /** |
1430 | * gfs2_next_unreserved_block - Return next block that is not reserved | |
1431 | * @rgd: The resource group | |
1432 | * @block: The starting block | |
ff7f4cb4 | 1433 | * @length: The required length |
5b924ae2 SW |
1434 | * @ip: Ignore any reservations for this inode |
1435 | * | |
1436 | * If the block does not appear in any reservation, then return the | |
1437 | * block number unchanged. If it does appear in the reservation, then | |
1438 | * keep looking through the tree of reservations in order to find the | |
1439 | * first block number which is not reserved. | |
1440 | */ | |
1441 | ||
1442 | static u64 gfs2_next_unreserved_block(struct gfs2_rgrpd *rgd, u64 block, | |
ff7f4cb4 | 1443 | u32 length, |
5b924ae2 SW |
1444 | const struct gfs2_inode *ip) |
1445 | { | |
1446 | struct gfs2_blkreserv *rs; | |
1447 | struct rb_node *n; | |
1448 | int rc; | |
1449 | ||
1450 | spin_lock(&rgd->rd_rsspin); | |
ff7f4cb4 | 1451 | n = rgd->rd_rstree.rb_node; |
5b924ae2 SW |
1452 | while (n) { |
1453 | rs = rb_entry(n, struct gfs2_blkreserv, rs_node); | |
ff7f4cb4 | 1454 | rc = rs_cmp(block, length, rs); |
5b924ae2 SW |
1455 | if (rc < 0) |
1456 | n = n->rb_left; | |
1457 | else if (rc > 0) | |
1458 | n = n->rb_right; | |
1459 | else | |
1460 | break; | |
1461 | } | |
1462 | ||
1463 | if (n) { | |
ff7f4cb4 | 1464 | while ((rs_cmp(block, length, rs) == 0) && (ip->i_res != rs)) { |
5b924ae2 | 1465 | block = gfs2_rbm_to_block(&rs->rs_rbm) + rs->rs_free; |
ff7f4cb4 | 1466 | n = n->rb_right; |
5b924ae2 SW |
1467 | if (n == NULL) |
1468 | break; | |
1469 | rs = rb_entry(n, struct gfs2_blkreserv, rs_node); | |
1470 | } | |
1471 | } | |
1472 | ||
1473 | spin_unlock(&rgd->rd_rsspin); | |
1474 | return block; | |
1475 | } | |
1476 | ||
5b924ae2 SW |
1477 | /** |
1478 | * gfs2_reservation_check_and_update - Check for reservations during block alloc | |
1479 | * @rbm: The current position in the resource group | |
ff7f4cb4 SW |
1480 | * @ip: The inode for which we are searching for blocks |
1481 | * @minext: The minimum extent length | |
5b924ae2 SW |
1482 | * |
1483 | * This checks the current position in the rgrp to see whether there is | |
1484 | * a reservation covering this block. If not then this function is a | |
1485 | * no-op. If there is, then the position is moved to the end of the | |
1486 | * contiguous reservation(s) so that we are pointing at the first | |
1487 | * non-reserved block. | |
1488 | * | |
1489 | * Returns: 0 if no reservation, 1 if @rbm has changed, otherwise an error | |
1490 | */ | |
1491 | ||
1492 | static int gfs2_reservation_check_and_update(struct gfs2_rbm *rbm, | |
ff7f4cb4 SW |
1493 | const struct gfs2_inode *ip, |
1494 | u32 minext) | |
5b924ae2 SW |
1495 | { |
1496 | u64 block = gfs2_rbm_to_block(rbm); | |
ff7f4cb4 | 1497 | u32 extlen = 1; |
5b924ae2 SW |
1498 | u64 nblock; |
1499 | int ret; | |
1500 | ||
ff7f4cb4 SW |
1501 | /* |
1502 | * If we have a minimum extent length, then skip over any extent | |
1503 | * which is less than the min extent length in size. | |
1504 | */ | |
1505 | if (minext) { | |
1506 | extlen = gfs2_free_extlen(rbm, minext); | |
1507 | nblock = block + extlen; | |
1508 | if (extlen < minext) | |
1509 | goto fail; | |
1510 | } | |
1511 | ||
1512 | /* | |
1513 | * Check the extent which has been found against the reservations | |
1514 | * and skip if parts of it are already reserved | |
1515 | */ | |
1516 | nblock = gfs2_next_unreserved_block(rbm->rgd, block, extlen, ip); | |
5b924ae2 SW |
1517 | if (nblock == block) |
1518 | return 0; | |
ff7f4cb4 | 1519 | fail: |
5b924ae2 SW |
1520 | ret = gfs2_rbm_from_block(rbm, nblock); |
1521 | if (ret < 0) | |
1522 | return ret; | |
1523 | return 1; | |
1524 | } | |
1525 | ||
1526 | /** | |
1527 | * gfs2_rbm_find - Look for blocks of a particular state | |
1528 | * @rbm: Value/result starting position and final position | |
1529 | * @state: The state which we want to find | |
ff7f4cb4 | 1530 | * @minext: The requested extent length (0 for a single block) |
5b924ae2 SW |
1531 | * @ip: If set, check for reservations |
1532 | * @nowrap: Stop looking at the end of the rgrp, rather than wrapping | |
1533 | * around until we've reached the starting point. | |
1534 | * | |
1535 | * Side effects: | |
1536 | * - If looking for free blocks, we set GBF_FULL on each bitmap which | |
1537 | * has no free blocks in it. | |
1538 | * | |
1539 | * Returns: 0 on success, -ENOSPC if there is no block of the requested state | |
1540 | */ | |
1541 | ||
ff7f4cb4 | 1542 | static int gfs2_rbm_find(struct gfs2_rbm *rbm, u8 state, u32 minext, |
5b924ae2 SW |
1543 | const struct gfs2_inode *ip, bool nowrap) |
1544 | { | |
1545 | struct buffer_head *bh; | |
1546 | struct gfs2_bitmap *initial_bi; | |
1547 | u32 initial_offset; | |
1548 | u32 offset; | |
1549 | u8 *buffer; | |
1550 | int index; | |
1551 | int n = 0; | |
1552 | int iters = rbm->rgd->rd_length; | |
1553 | int ret; | |
1554 | ||
1555 | /* If we are not starting at the beginning of a bitmap, then we | |
1556 | * need to add one to the bitmap count to ensure that we search | |
1557 | * the starting bitmap twice. | |
1558 | */ | |
1559 | if (rbm->offset != 0) | |
1560 | iters++; | |
1561 | ||
1562 | while(1) { | |
1563 | if (test_bit(GBF_FULL, &rbm->bi->bi_flags) && | |
1564 | (state == GFS2_BLKST_FREE)) | |
1565 | goto next_bitmap; | |
1566 | ||
1567 | bh = rbm->bi->bi_bh; | |
1568 | buffer = bh->b_data + rbm->bi->bi_offset; | |
1569 | WARN_ON(!buffer_uptodate(bh)); | |
1570 | if (state != GFS2_BLKST_UNLINKED && rbm->bi->bi_clone) | |
1571 | buffer = rbm->bi->bi_clone + rbm->bi->bi_offset; | |
5b924ae2 SW |
1572 | initial_offset = rbm->offset; |
1573 | offset = gfs2_bitfit(buffer, rbm->bi->bi_len, rbm->offset, state); | |
1574 | if (offset == BFITNOENT) | |
1575 | goto bitmap_full; | |
1576 | rbm->offset = offset; | |
1577 | if (ip == NULL) | |
1578 | return 0; | |
1579 | ||
1580 | initial_bi = rbm->bi; | |
ff7f4cb4 | 1581 | ret = gfs2_reservation_check_and_update(rbm, ip, minext); |
5b924ae2 SW |
1582 | if (ret == 0) |
1583 | return 0; | |
1584 | if (ret > 0) { | |
1585 | n += (rbm->bi - initial_bi); | |
8d8b752a | 1586 | goto next_iter; |
5b924ae2 | 1587 | } |
5d50d532 SW |
1588 | if (ret == -E2BIG) { |
1589 | index = 0; | |
1590 | rbm->offset = 0; | |
1591 | n += (rbm->bi - initial_bi); | |
1592 | goto res_covered_end_of_rgrp; | |
1593 | } | |
5b924ae2 SW |
1594 | return ret; |
1595 | ||
1596 | bitmap_full: /* Mark bitmap as full and fall through */ | |
1597 | if ((state == GFS2_BLKST_FREE) && initial_offset == 0) | |
1598 | set_bit(GBF_FULL, &rbm->bi->bi_flags); | |
1599 | ||
1600 | next_bitmap: /* Find next bitmap in the rgrp */ | |
1601 | rbm->offset = 0; | |
1602 | index = rbm->bi - rbm->rgd->rd_bits; | |
1603 | index++; | |
1604 | if (index == rbm->rgd->rd_length) | |
1605 | index = 0; | |
5d50d532 | 1606 | res_covered_end_of_rgrp: |
5b924ae2 SW |
1607 | rbm->bi = &rbm->rgd->rd_bits[index]; |
1608 | if ((index == 0) && nowrap) | |
1609 | break; | |
1610 | n++; | |
8d8b752a | 1611 | next_iter: |
5b924ae2 SW |
1612 | if (n >= iters) |
1613 | break; | |
1614 | } | |
1615 | ||
1616 | return -ENOSPC; | |
1617 | } | |
1618 | ||
c8cdf479 SW |
1619 | /** |
1620 | * try_rgrp_unlink - Look for any unlinked, allocated, but unused inodes | |
1621 | * @rgd: The rgrp | |
886b1416 BP |
1622 | * @last_unlinked: block address of the last dinode we unlinked |
1623 | * @skip: block address we should explicitly not unlink | |
c8cdf479 | 1624 | * |
1a0eae88 BP |
1625 | * Returns: 0 if no error |
1626 | * The inode, if one has been found, in inode. | |
c8cdf479 SW |
1627 | */ |
1628 | ||
044b9414 | 1629 | static void try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked, u64 skip) |
c8cdf479 | 1630 | { |
5b924ae2 | 1631 | u64 block; |
5f3eae75 | 1632 | struct gfs2_sbd *sdp = rgd->rd_sbd; |
044b9414 SW |
1633 | struct gfs2_glock *gl; |
1634 | struct gfs2_inode *ip; | |
1635 | int error; | |
1636 | int found = 0; | |
5b924ae2 | 1637 | struct gfs2_rbm rbm = { .rgd = rgd, .bi = rgd->rd_bits, .offset = 0 }; |
c8cdf479 | 1638 | |
5b924ae2 | 1639 | while (1) { |
5f3eae75 | 1640 | down_write(&sdp->sd_log_flush_lock); |
ff7f4cb4 | 1641 | error = gfs2_rbm_find(&rbm, GFS2_BLKST_UNLINKED, 0, NULL, true); |
5f3eae75 | 1642 | up_write(&sdp->sd_log_flush_lock); |
5b924ae2 SW |
1643 | if (error == -ENOSPC) |
1644 | break; | |
1645 | if (WARN_ON_ONCE(error)) | |
24c73873 | 1646 | break; |
b3e47ca0 | 1647 | |
5b924ae2 SW |
1648 | block = gfs2_rbm_to_block(&rbm); |
1649 | if (gfs2_rbm_from_block(&rbm, block + 1)) | |
1650 | break; | |
1651 | if (*last_unlinked != NO_BLOCK && block <= *last_unlinked) | |
c8cdf479 | 1652 | continue; |
5b924ae2 | 1653 | if (block == skip) |
1e19a195 | 1654 | continue; |
5b924ae2 | 1655 | *last_unlinked = block; |
044b9414 | 1656 | |
5b924ae2 | 1657 | error = gfs2_glock_get(sdp, block, &gfs2_inode_glops, CREATE, &gl); |
044b9414 SW |
1658 | if (error) |
1659 | continue; | |
1660 | ||
1661 | /* If the inode is already in cache, we can ignore it here | |
1662 | * because the existing inode disposal code will deal with | |
1663 | * it when all refs have gone away. Accessing gl_object like | |
1664 | * this is not safe in general. Here it is ok because we do | |
1665 | * not dereference the pointer, and we only need an approx | |
1666 | * answer to whether it is NULL or not. | |
1667 | */ | |
1668 | ip = gl->gl_object; | |
1669 | ||
1670 | if (ip || queue_work(gfs2_delete_workqueue, &gl->gl_delete) == 0) | |
1671 | gfs2_glock_put(gl); | |
1672 | else | |
1673 | found++; | |
1674 | ||
1675 | /* Limit reclaim to sensible number of tasks */ | |
44ad37d6 | 1676 | if (found > NR_CPUS) |
044b9414 | 1677 | return; |
c8cdf479 SW |
1678 | } |
1679 | ||
1680 | rgd->rd_flags &= ~GFS2_RDF_CHECK; | |
044b9414 | 1681 | return; |
c8cdf479 SW |
1682 | } |
1683 | ||
c743ffd0 SW |
1684 | static bool gfs2_select_rgrp(struct gfs2_rgrpd **pos, const struct gfs2_rgrpd *begin) |
1685 | { | |
1686 | struct gfs2_rgrpd *rgd = *pos; | |
1687 | ||
1688 | rgd = gfs2_rgrpd_get_next(rgd); | |
1689 | if (rgd == NULL) | |
1690 | rgd = gfs2_rgrpd_get_next(NULL); | |
1691 | *pos = rgd; | |
1692 | if (rgd != begin) /* If we didn't wrap */ | |
1693 | return true; | |
1694 | return false; | |
1695 | } | |
1696 | ||
b3b94faa | 1697 | /** |
666d1d8a | 1698 | * gfs2_inplace_reserve - Reserve space in the filesystem |
b3b94faa | 1699 | * @ip: the inode to reserve space for |
666d1d8a | 1700 | * @requested: the number of blocks to be reserved |
b3b94faa DT |
1701 | * |
1702 | * Returns: errno | |
1703 | */ | |
1704 | ||
666d1d8a | 1705 | int gfs2_inplace_reserve(struct gfs2_inode *ip, u32 requested) |
b3b94faa | 1706 | { |
feaa7bba | 1707 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
8e2e0047 | 1708 | struct gfs2_rgrpd *begin = NULL; |
564e12b1 | 1709 | struct gfs2_blkreserv *rs = ip->i_res; |
666d1d8a BP |
1710 | int error = 0, rg_locked, flags = LM_FLAG_TRY; |
1711 | u64 last_unlinked = NO_BLOCK; | |
7c9ca621 | 1712 | int loops = 0; |
b3b94faa | 1713 | |
90306c41 BM |
1714 | if (sdp->sd_args.ar_rgrplvb) |
1715 | flags |= GL_SKIP; | |
c743ffd0 SW |
1716 | if (gfs2_assert_warn(sdp, requested)) |
1717 | return -EINVAL; | |
8e2e0047 | 1718 | if (gfs2_rs_active(rs)) { |
4a993fb1 | 1719 | begin = rs->rs_rbm.rgd; |
8e2e0047 BP |
1720 | flags = 0; /* Yoda: Do or do not. There is no try */ |
1721 | } else if (ip->i_rgd && rgrp_contains_block(ip->i_rgd, ip->i_goal)) { | |
4a993fb1 | 1722 | rs->rs_rbm.rgd = begin = ip->i_rgd; |
8e2e0047 | 1723 | } else { |
4a993fb1 | 1724 | rs->rs_rbm.rgd = begin = gfs2_blk2rgrpd(sdp, ip->i_goal, 1); |
8e2e0047 | 1725 | } |
4a993fb1 | 1726 | if (rs->rs_rbm.rgd == NULL) |
7c9ca621 BP |
1727 | return -EBADSLT; |
1728 | ||
1729 | while (loops < 3) { | |
c743ffd0 SW |
1730 | rg_locked = 1; |
1731 | ||
1732 | if (!gfs2_glock_is_locked_by_me(rs->rs_rbm.rgd->rd_gl)) { | |
1733 | rg_locked = 0; | |
4a993fb1 | 1734 | error = gfs2_glock_nq_init(rs->rs_rbm.rgd->rd_gl, |
8e2e0047 BP |
1735 | LM_ST_EXCLUSIVE, flags, |
1736 | &rs->rs_rgd_gh); | |
c743ffd0 SW |
1737 | if (error == GLR_TRYFAILED) |
1738 | goto next_rgrp; | |
1739 | if (unlikely(error)) | |
1740 | return error; | |
1741 | if (sdp->sd_args.ar_rgrplvb) { | |
4a993fb1 | 1742 | error = update_rgrp_lvb(rs->rs_rbm.rgd); |
c743ffd0 | 1743 | if (unlikely(error)) { |
90306c41 BM |
1744 | gfs2_glock_dq_uninit(&rs->rs_rgd_gh); |
1745 | return error; | |
1746 | } | |
1747 | } | |
292c8c14 | 1748 | } |
666d1d8a | 1749 | |
c743ffd0 SW |
1750 | /* Skip unuseable resource groups */ |
1751 | if (rs->rs_rbm.rgd->rd_flags & (GFS2_RGF_NOALLOC | GFS2_RDF_ERROR)) | |
1752 | goto skip_rgrp; | |
1753 | ||
1754 | if (sdp->sd_args.ar_rgrplvb) | |
1755 | gfs2_rgrp_bh_get(rs->rs_rbm.rgd); | |
1756 | ||
1757 | /* Get a reservation if we don't already have one */ | |
1758 | if (!gfs2_rs_active(rs)) | |
1759 | rg_mblk_search(rs->rs_rbm.rgd, ip, requested); | |
1760 | ||
1761 | /* Skip rgrps when we can't get a reservation on first pass */ | |
1762 | if (!gfs2_rs_active(rs) && (loops < 1)) | |
1763 | goto check_rgrp; | |
1764 | ||
1765 | /* If rgrp has enough free space, use it */ | |
1766 | if (rs->rs_rbm.rgd->rd_free_clone >= requested) { | |
1767 | ip->i_rgd = rs->rs_rbm.rgd; | |
1768 | return 0; | |
b3b94faa | 1769 | } |
c743ffd0 SW |
1770 | |
1771 | /* Drop reservation, if we couldn't use reserved rgrp */ | |
1772 | if (gfs2_rs_active(rs)) | |
1773 | gfs2_rs_deltree(ip, rs); | |
1774 | check_rgrp: | |
1775 | /* Check for unlinked inodes which can be reclaimed */ | |
1776 | if (rs->rs_rbm.rgd->rd_flags & GFS2_RDF_CHECK) | |
1777 | try_rgrp_unlink(rs->rs_rbm.rgd, &last_unlinked, | |
1778 | ip->i_no_addr); | |
1779 | skip_rgrp: | |
1780 | /* Unlock rgrp if required */ | |
1781 | if (!rg_locked) | |
1782 | gfs2_glock_dq_uninit(&rs->rs_rgd_gh); | |
1783 | next_rgrp: | |
1784 | /* Find the next rgrp, and continue looking */ | |
1785 | if (gfs2_select_rgrp(&rs->rs_rbm.rgd, begin)) | |
1786 | continue; | |
1787 | ||
1788 | /* If we've scanned all the rgrps, but found no free blocks | |
1789 | * then this checks for some less likely conditions before | |
1790 | * trying again. | |
1791 | */ | |
1792 | flags &= ~LM_FLAG_TRY; | |
1793 | loops++; | |
1794 | /* Check that fs hasn't grown if writing to rindex */ | |
1795 | if (ip == GFS2_I(sdp->sd_rindex) && !sdp->sd_rindex_uptodate) { | |
1796 | error = gfs2_ri_update(ip); | |
1797 | if (error) | |
1798 | return error; | |
1799 | } | |
1800 | /* Flushing the log may release space */ | |
1801 | if (loops == 2) | |
1802 | gfs2_log_flush(sdp, NULL); | |
b3b94faa | 1803 | } |
b3b94faa | 1804 | |
c743ffd0 | 1805 | return -ENOSPC; |
b3b94faa DT |
1806 | } |
1807 | ||
1808 | /** | |
1809 | * gfs2_inplace_release - release an inplace reservation | |
1810 | * @ip: the inode the reservation was taken out on | |
1811 | * | |
1812 | * Release a reservation made by gfs2_inplace_reserve(). | |
1813 | */ | |
1814 | ||
1815 | void gfs2_inplace_release(struct gfs2_inode *ip) | |
1816 | { | |
564e12b1 | 1817 | struct gfs2_blkreserv *rs = ip->i_res; |
b3b94faa | 1818 | |
564e12b1 BP |
1819 | if (rs->rs_rgd_gh.gh_gl) |
1820 | gfs2_glock_dq_uninit(&rs->rs_rgd_gh); | |
b3b94faa DT |
1821 | } |
1822 | ||
1823 | /** | |
1824 | * gfs2_get_block_type - Check a block in a RG is of given type | |
1825 | * @rgd: the resource group holding the block | |
1826 | * @block: the block number | |
1827 | * | |
1828 | * Returns: The block type (GFS2_BLKST_*) | |
1829 | */ | |
1830 | ||
acf7e244 | 1831 | static unsigned char gfs2_get_block_type(struct gfs2_rgrpd *rgd, u64 block) |
b3b94faa | 1832 | { |
3983903a SW |
1833 | struct gfs2_rbm rbm = { .rgd = rgd, }; |
1834 | int ret; | |
b3b94faa | 1835 | |
3983903a SW |
1836 | ret = gfs2_rbm_from_block(&rbm, block); |
1837 | WARN_ON_ONCE(ret != 0); | |
b3b94faa | 1838 | |
c04a2ef3 | 1839 | return gfs2_testbit(&rbm); |
b3b94faa DT |
1840 | } |
1841 | ||
60a0b8f9 | 1842 | |
b3e47ca0 BP |
1843 | /** |
1844 | * gfs2_alloc_extent - allocate an extent from a given bitmap | |
4a993fb1 | 1845 | * @rbm: the resource group information |
b3e47ca0 | 1846 | * @dinode: TRUE if the first block we allocate is for a dinode |
c04a2ef3 | 1847 | * @n: The extent length (value/result) |
b3e47ca0 | 1848 | * |
c04a2ef3 | 1849 | * Add the bitmap buffer to the transaction. |
b3e47ca0 | 1850 | * Set the found bits to @new_state to change block's allocation state. |
b3e47ca0 | 1851 | */ |
c04a2ef3 | 1852 | static void gfs2_alloc_extent(const struct gfs2_rbm *rbm, bool dinode, |
4a993fb1 | 1853 | unsigned int *n) |
b3e47ca0 | 1854 | { |
c04a2ef3 | 1855 | struct gfs2_rbm pos = { .rgd = rbm->rgd, }; |
b3e47ca0 | 1856 | const unsigned int elen = *n; |
c04a2ef3 SW |
1857 | u64 block; |
1858 | int ret; | |
b3e47ca0 | 1859 | |
c04a2ef3 SW |
1860 | *n = 1; |
1861 | block = gfs2_rbm_to_block(rbm); | |
1862 | gfs2_trans_add_bh(rbm->rgd->rd_gl, rbm->bi->bi_bh, 1); | |
3e6339dd | 1863 | gfs2_setbit(rbm, true, dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED); |
c04a2ef3 | 1864 | block++; |
60a0b8f9 | 1865 | while (*n < elen) { |
c04a2ef3 | 1866 | ret = gfs2_rbm_from_block(&pos, block); |
0688a5ec | 1867 | if (ret || gfs2_testbit(&pos) != GFS2_BLKST_FREE) |
60a0b8f9 | 1868 | break; |
c04a2ef3 | 1869 | gfs2_trans_add_bh(pos.rgd->rd_gl, pos.bi->bi_bh, 1); |
3e6339dd | 1870 | gfs2_setbit(&pos, true, GFS2_BLKST_USED); |
60a0b8f9 | 1871 | (*n)++; |
c04a2ef3 | 1872 | block++; |
c8cdf479 | 1873 | } |
b3b94faa DT |
1874 | } |
1875 | ||
1876 | /** | |
1877 | * rgblk_free - Change alloc state of given block(s) | |
1878 | * @sdp: the filesystem | |
1879 | * @bstart: the start of a run of blocks to free | |
1880 | * @blen: the length of the block run (all must lie within ONE RG!) | |
1881 | * @new_state: GFS2_BLKST_XXX the after-allocation block state | |
1882 | * | |
1883 | * Returns: Resource group containing the block(s) | |
1884 | */ | |
1885 | ||
cd915493 SW |
1886 | static struct gfs2_rgrpd *rgblk_free(struct gfs2_sbd *sdp, u64 bstart, |
1887 | u32 blen, unsigned char new_state) | |
b3b94faa | 1888 | { |
3b1d0b9d | 1889 | struct gfs2_rbm rbm; |
b3b94faa | 1890 | |
3b1d0b9d SW |
1891 | rbm.rgd = gfs2_blk2rgrpd(sdp, bstart, 1); |
1892 | if (!rbm.rgd) { | |
b3b94faa | 1893 | if (gfs2_consist(sdp)) |
382066da | 1894 | fs_err(sdp, "block = %llu\n", (unsigned long long)bstart); |
b3b94faa DT |
1895 | return NULL; |
1896 | } | |
1897 | ||
b3b94faa | 1898 | while (blen--) { |
3b1d0b9d SW |
1899 | gfs2_rbm_from_block(&rbm, bstart); |
1900 | bstart++; | |
1901 | if (!rbm.bi->bi_clone) { | |
1902 | rbm.bi->bi_clone = kmalloc(rbm.bi->bi_bh->b_size, | |
1903 | GFP_NOFS | __GFP_NOFAIL); | |
1904 | memcpy(rbm.bi->bi_clone + rbm.bi->bi_offset, | |
1905 | rbm.bi->bi_bh->b_data + rbm.bi->bi_offset, | |
1906 | rbm.bi->bi_len); | |
b3b94faa | 1907 | } |
3b1d0b9d | 1908 | gfs2_trans_add_bh(rbm.rgd->rd_gl, rbm.bi->bi_bh, 1); |
3e6339dd | 1909 | gfs2_setbit(&rbm, false, new_state); |
b3b94faa DT |
1910 | } |
1911 | ||
3b1d0b9d | 1912 | return rbm.rgd; |
b3b94faa DT |
1913 | } |
1914 | ||
1915 | /** | |
09010978 SW |
1916 | * gfs2_rgrp_dump - print out an rgrp |
1917 | * @seq: The iterator | |
1918 | * @gl: The glock in question | |
1919 | * | |
1920 | */ | |
1921 | ||
1922 | int gfs2_rgrp_dump(struct seq_file *seq, const struct gfs2_glock *gl) | |
1923 | { | |
8e2e0047 BP |
1924 | struct gfs2_rgrpd *rgd = gl->gl_object; |
1925 | struct gfs2_blkreserv *trs; | |
1926 | const struct rb_node *n; | |
1927 | ||
09010978 SW |
1928 | if (rgd == NULL) |
1929 | return 0; | |
8e2e0047 | 1930 | gfs2_print_dbg(seq, " R: n:%llu f:%02x b:%u/%u i:%u r:%u\n", |
09010978 | 1931 | (unsigned long long)rgd->rd_addr, rgd->rd_flags, |
8e2e0047 BP |
1932 | rgd->rd_free, rgd->rd_free_clone, rgd->rd_dinodes, |
1933 | rgd->rd_reserved); | |
1934 | spin_lock(&rgd->rd_rsspin); | |
1935 | for (n = rb_first(&rgd->rd_rstree); n; n = rb_next(&trs->rs_node)) { | |
1936 | trs = rb_entry(n, struct gfs2_blkreserv, rs_node); | |
1937 | dump_rs(seq, trs); | |
1938 | } | |
1939 | spin_unlock(&rgd->rd_rsspin); | |
09010978 SW |
1940 | return 0; |
1941 | } | |
1942 | ||
6050b9c7 SW |
1943 | static void gfs2_rgrp_error(struct gfs2_rgrpd *rgd) |
1944 | { | |
1945 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
1946 | fs_warn(sdp, "rgrp %llu has an error, marking it readonly until umount\n", | |
86d00636 | 1947 | (unsigned long long)rgd->rd_addr); |
6050b9c7 SW |
1948 | fs_warn(sdp, "umount on all nodes and run fsck.gfs2 to fix the error\n"); |
1949 | gfs2_rgrp_dump(NULL, rgd->rd_gl); | |
1950 | rgd->rd_flags |= GFS2_RDF_ERROR; | |
1951 | } | |
1952 | ||
8e2e0047 | 1953 | /** |
5b924ae2 SW |
1954 | * gfs2_adjust_reservation - Adjust (or remove) a reservation after allocation |
1955 | * @ip: The inode we have just allocated blocks for | |
1956 | * @rbm: The start of the allocated blocks | |
1957 | * @len: The extent length | |
8e2e0047 | 1958 | * |
5b924ae2 SW |
1959 | * Adjusts a reservation after an allocation has taken place. If the |
1960 | * reservation does not match the allocation, or if it is now empty | |
1961 | * then it is removed. | |
8e2e0047 | 1962 | */ |
5b924ae2 SW |
1963 | |
1964 | static void gfs2_adjust_reservation(struct gfs2_inode *ip, | |
1965 | const struct gfs2_rbm *rbm, unsigned len) | |
8e2e0047 BP |
1966 | { |
1967 | struct gfs2_blkreserv *rs = ip->i_res; | |
5b924ae2 SW |
1968 | struct gfs2_rgrpd *rgd = rbm->rgd; |
1969 | unsigned rlen; | |
1970 | u64 block; | |
1971 | int ret; | |
8e2e0047 | 1972 | |
5b924ae2 SW |
1973 | spin_lock(&rgd->rd_rsspin); |
1974 | if (gfs2_rs_active(rs)) { | |
1975 | if (gfs2_rbm_eq(&rs->rs_rbm, rbm)) { | |
1976 | block = gfs2_rbm_to_block(rbm); | |
1977 | ret = gfs2_rbm_from_block(&rs->rs_rbm, block + len); | |
1978 | rlen = min(rs->rs_free, len); | |
1979 | rs->rs_free -= rlen; | |
1980 | rgd->rd_reserved -= rlen; | |
9e733d39 | 1981 | trace_gfs2_rs(rs, TRACE_RS_CLAIM); |
5b924ae2 SW |
1982 | if (rs->rs_free && !ret) |
1983 | goto out; | |
1984 | } | |
1985 | __rs_deltree(ip, rs); | |
8e2e0047 | 1986 | } |
5b924ae2 SW |
1987 | out: |
1988 | spin_unlock(&rgd->rd_rsspin); | |
8e2e0047 BP |
1989 | } |
1990 | ||
09010978 | 1991 | /** |
6e87ed0f | 1992 | * gfs2_alloc_blocks - Allocate one or more blocks of data and/or a dinode |
1639431a | 1993 | * @ip: the inode to allocate the block for |
09010978 | 1994 | * @bn: Used to return the starting block number |
8e2e0047 | 1995 | * @nblocks: requested number of blocks/extent length (value/result) |
6e87ed0f | 1996 | * @dinode: 1 if we're allocating a dinode block, else 0 |
3c5d785a | 1997 | * @generation: the generation number of the inode |
b3b94faa | 1998 | * |
09010978 | 1999 | * Returns: 0 or error |
b3b94faa DT |
2000 | */ |
2001 | ||
6a8099ed | 2002 | int gfs2_alloc_blocks(struct gfs2_inode *ip, u64 *bn, unsigned int *nblocks, |
6e87ed0f | 2003 | bool dinode, u64 *generation) |
b3b94faa | 2004 | { |
feaa7bba | 2005 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
d9ba7615 | 2006 | struct buffer_head *dibh; |
4a993fb1 | 2007 | struct gfs2_rbm rbm = { .rgd = ip->i_rgd, }; |
6a8099ed | 2008 | unsigned int ndata; |
5b924ae2 | 2009 | u64 goal; |
3c5d785a | 2010 | u64 block; /* block, within the file system scope */ |
d9ba7615 | 2011 | int error; |
b3b94faa | 2012 | |
5b924ae2 SW |
2013 | if (gfs2_rs_active(ip->i_res)) |
2014 | goal = gfs2_rbm_to_block(&ip->i_res->rs_rbm); | |
2015 | else if (!dinode && rgrp_contains_block(rbm.rgd, ip->i_goal)) | |
2016 | goal = ip->i_goal; | |
62e252ee | 2017 | else |
5b924ae2 | 2018 | goal = rbm.rgd->rd_last_alloc + rbm.rgd->rd_data0; |
62e252ee | 2019 | |
5b924ae2 | 2020 | gfs2_rbm_from_block(&rbm, goal); |
ff7f4cb4 | 2021 | error = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, 0, ip, false); |
62e252ee | 2022 | |
137834a6 SW |
2023 | if (error == -ENOSPC) { |
2024 | gfs2_rbm_from_block(&rbm, goal); | |
ff7f4cb4 | 2025 | error = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, 0, NULL, false); |
137834a6 SW |
2026 | } |
2027 | ||
62e252ee | 2028 | /* Since all blocks are reserved in advance, this shouldn't happen */ |
5b924ae2 | 2029 | if (error) { |
9e733d39 SW |
2030 | fs_warn(sdp, "inum=%llu error=%d, nblocks=%u, full=%d\n", |
2031 | (unsigned long long)ip->i_no_addr, error, *nblocks, | |
5b924ae2 | 2032 | test_bit(GBF_FULL, &rbm.rgd->rd_bits->bi_flags)); |
62e252ee | 2033 | goto rgrp_error; |
8e2e0047 | 2034 | } |
62e252ee | 2035 | |
c04a2ef3 SW |
2036 | gfs2_alloc_extent(&rbm, dinode, nblocks); |
2037 | block = gfs2_rbm_to_block(&rbm); | |
c743ffd0 | 2038 | rbm.rgd->rd_last_alloc = block - rbm.rgd->rd_data0; |
5b924ae2 SW |
2039 | if (gfs2_rs_active(ip->i_res)) |
2040 | gfs2_adjust_reservation(ip, &rbm, *nblocks); | |
6a8099ed SW |
2041 | ndata = *nblocks; |
2042 | if (dinode) | |
2043 | ndata--; | |
b3e47ca0 | 2044 | |
3c5d785a | 2045 | if (!dinode) { |
6a8099ed | 2046 | ip->i_goal = block + ndata - 1; |
3c5d785a BP |
2047 | error = gfs2_meta_inode_buffer(ip, &dibh); |
2048 | if (error == 0) { | |
2049 | struct gfs2_dinode *di = | |
2050 | (struct gfs2_dinode *)dibh->b_data; | |
2051 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
2052 | di->di_goal_meta = di->di_goal_data = | |
2053 | cpu_to_be64(ip->i_goal); | |
2054 | brelse(dibh); | |
2055 | } | |
d9ba7615 | 2056 | } |
4a993fb1 | 2057 | if (rbm.rgd->rd_free < *nblocks) { |
8e2e0047 | 2058 | printk(KERN_WARNING "nblocks=%u\n", *nblocks); |
09010978 | 2059 | goto rgrp_error; |
8e2e0047 | 2060 | } |
09010978 | 2061 | |
4a993fb1 | 2062 | rbm.rgd->rd_free -= *nblocks; |
3c5d785a | 2063 | if (dinode) { |
4a993fb1 SW |
2064 | rbm.rgd->rd_dinodes++; |
2065 | *generation = rbm.rgd->rd_igeneration++; | |
3c5d785a | 2066 | if (*generation == 0) |
4a993fb1 | 2067 | *generation = rbm.rgd->rd_igeneration++; |
3c5d785a | 2068 | } |
b3b94faa | 2069 | |
4a993fb1 SW |
2070 | gfs2_trans_add_bh(rbm.rgd->rd_gl, rbm.rgd->rd_bits[0].bi_bh, 1); |
2071 | gfs2_rgrp_out(rbm.rgd, rbm.rgd->rd_bits[0].bi_bh->b_data); | |
2072 | gfs2_rgrp_ondisk2lvb(rbm.rgd->rd_rgl, rbm.rgd->rd_bits[0].bi_bh->b_data); | |
b3b94faa | 2073 | |
6a8099ed | 2074 | gfs2_statfs_change(sdp, 0, -(s64)*nblocks, dinode ? 1 : 0); |
3c5d785a BP |
2075 | if (dinode) |
2076 | gfs2_trans_add_unrevoke(sdp, block, 1); | |
6a8099ed SW |
2077 | |
2078 | /* | |
2079 | * This needs reviewing to see why we cannot do the quota change | |
2080 | * at this point in the dinode case. | |
2081 | */ | |
2082 | if (ndata) | |
2083 | gfs2_quota_change(ip, ndata, ip->i_inode.i_uid, | |
3c5d785a | 2084 | ip->i_inode.i_gid); |
b3b94faa | 2085 | |
4a993fb1 SW |
2086 | rbm.rgd->rd_free_clone -= *nblocks; |
2087 | trace_gfs2_block_alloc(ip, rbm.rgd, block, *nblocks, | |
6e87ed0f | 2088 | dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED); |
6050b9c7 SW |
2089 | *bn = block; |
2090 | return 0; | |
2091 | ||
2092 | rgrp_error: | |
4a993fb1 | 2093 | gfs2_rgrp_error(rbm.rgd); |
6050b9c7 | 2094 | return -EIO; |
b3b94faa DT |
2095 | } |
2096 | ||
2097 | /** | |
46fcb2ed | 2098 | * __gfs2_free_blocks - free a contiguous run of block(s) |
b3b94faa DT |
2099 | * @ip: the inode these blocks are being freed from |
2100 | * @bstart: first block of a run of contiguous blocks | |
2101 | * @blen: the length of the block run | |
46fcb2ed | 2102 | * @meta: 1 if the blocks represent metadata |
b3b94faa DT |
2103 | * |
2104 | */ | |
2105 | ||
46fcb2ed | 2106 | void __gfs2_free_blocks(struct gfs2_inode *ip, u64 bstart, u32 blen, int meta) |
b3b94faa | 2107 | { |
feaa7bba | 2108 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
2109 | struct gfs2_rgrpd *rgd; |
2110 | ||
2111 | rgd = rgblk_free(sdp, bstart, blen, GFS2_BLKST_FREE); | |
2112 | if (!rgd) | |
2113 | return; | |
41db1ab9 | 2114 | trace_gfs2_block_alloc(ip, rgd, bstart, blen, GFS2_BLKST_FREE); |
cfc8b549 | 2115 | rgd->rd_free += blen; |
66fc061b | 2116 | rgd->rd_flags &= ~GFS2_RGF_TRIMMED; |
d4e9c4c3 | 2117 | gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1); |
42d52e38 | 2118 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
90306c41 | 2119 | gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, rgd->rd_bits[0].bi_bh->b_data); |
b3b94faa | 2120 | |
6d3117b4 | 2121 | /* Directories keep their data in the metadata address space */ |
46fcb2ed | 2122 | if (meta || ip->i_depth) |
6d3117b4 | 2123 | gfs2_meta_wipe(ip, bstart, blen); |
4c16c36a | 2124 | } |
b3b94faa | 2125 | |
4c16c36a BP |
2126 | /** |
2127 | * gfs2_free_meta - free a contiguous run of data block(s) | |
2128 | * @ip: the inode these blocks are being freed from | |
2129 | * @bstart: first block of a run of contiguous blocks | |
2130 | * @blen: the length of the block run | |
2131 | * | |
2132 | */ | |
2133 | ||
2134 | void gfs2_free_meta(struct gfs2_inode *ip, u64 bstart, u32 blen) | |
2135 | { | |
2136 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
2137 | ||
46fcb2ed | 2138 | __gfs2_free_blocks(ip, bstart, blen, 1); |
b3b94faa | 2139 | gfs2_statfs_change(sdp, 0, +blen, 0); |
2933f925 | 2140 | gfs2_quota_change(ip, -(s64)blen, ip->i_inode.i_uid, ip->i_inode.i_gid); |
b3b94faa DT |
2141 | } |
2142 | ||
feaa7bba SW |
2143 | void gfs2_unlink_di(struct inode *inode) |
2144 | { | |
2145 | struct gfs2_inode *ip = GFS2_I(inode); | |
2146 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
2147 | struct gfs2_rgrpd *rgd; | |
dbb7cae2 | 2148 | u64 blkno = ip->i_no_addr; |
feaa7bba SW |
2149 | |
2150 | rgd = rgblk_free(sdp, blkno, 1, GFS2_BLKST_UNLINKED); | |
2151 | if (!rgd) | |
2152 | return; | |
41db1ab9 | 2153 | trace_gfs2_block_alloc(ip, rgd, blkno, 1, GFS2_BLKST_UNLINKED); |
feaa7bba | 2154 | gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1); |
42d52e38 | 2155 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
90306c41 BM |
2156 | gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, rgd->rd_bits[0].bi_bh->b_data); |
2157 | update_rgrp_lvb_unlinked(rgd, 1); | |
feaa7bba SW |
2158 | } |
2159 | ||
cd915493 | 2160 | static void gfs2_free_uninit_di(struct gfs2_rgrpd *rgd, u64 blkno) |
b3b94faa DT |
2161 | { |
2162 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
2163 | struct gfs2_rgrpd *tmp_rgd; | |
2164 | ||
2165 | tmp_rgd = rgblk_free(sdp, blkno, 1, GFS2_BLKST_FREE); | |
2166 | if (!tmp_rgd) | |
2167 | return; | |
2168 | gfs2_assert_withdraw(sdp, rgd == tmp_rgd); | |
2169 | ||
73f74948 | 2170 | if (!rgd->rd_dinodes) |
b3b94faa | 2171 | gfs2_consist_rgrpd(rgd); |
73f74948 | 2172 | rgd->rd_dinodes--; |
cfc8b549 | 2173 | rgd->rd_free++; |
b3b94faa | 2174 | |
d4e9c4c3 | 2175 | gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1); |
42d52e38 | 2176 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
90306c41 BM |
2177 | gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, rgd->rd_bits[0].bi_bh->b_data); |
2178 | update_rgrp_lvb_unlinked(rgd, -1); | |
b3b94faa DT |
2179 | |
2180 | gfs2_statfs_change(sdp, 0, +1, -1); | |
b3b94faa DT |
2181 | } |
2182 | ||
b3b94faa DT |
2183 | |
2184 | void gfs2_free_di(struct gfs2_rgrpd *rgd, struct gfs2_inode *ip) | |
2185 | { | |
dbb7cae2 | 2186 | gfs2_free_uninit_di(rgd, ip->i_no_addr); |
41db1ab9 | 2187 | trace_gfs2_block_alloc(ip, rgd, ip->i_no_addr, 1, GFS2_BLKST_FREE); |
2933f925 | 2188 | gfs2_quota_change(ip, -1, ip->i_inode.i_uid, ip->i_inode.i_gid); |
dbb7cae2 | 2189 | gfs2_meta_wipe(ip, ip->i_no_addr, 1); |
b3b94faa DT |
2190 | } |
2191 | ||
acf7e244 SW |
2192 | /** |
2193 | * gfs2_check_blk_type - Check the type of a block | |
2194 | * @sdp: The superblock | |
2195 | * @no_addr: The block number to check | |
2196 | * @type: The block type we are looking for | |
2197 | * | |
2198 | * Returns: 0 if the block type matches the expected type | |
2199 | * -ESTALE if it doesn't match | |
2200 | * or -ve errno if something went wrong while checking | |
2201 | */ | |
2202 | ||
2203 | int gfs2_check_blk_type(struct gfs2_sbd *sdp, u64 no_addr, unsigned int type) | |
2204 | { | |
2205 | struct gfs2_rgrpd *rgd; | |
8339ee54 | 2206 | struct gfs2_holder rgd_gh; |
58884c4d | 2207 | int error = -EINVAL; |
acf7e244 | 2208 | |
66fc061b | 2209 | rgd = gfs2_blk2rgrpd(sdp, no_addr, 1); |
acf7e244 | 2210 | if (!rgd) |
8339ee54 | 2211 | goto fail; |
acf7e244 SW |
2212 | |
2213 | error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_SHARED, 0, &rgd_gh); | |
2214 | if (error) | |
8339ee54 | 2215 | goto fail; |
acf7e244 SW |
2216 | |
2217 | if (gfs2_get_block_type(rgd, no_addr) != type) | |
2218 | error = -ESTALE; | |
2219 | ||
2220 | gfs2_glock_dq_uninit(&rgd_gh); | |
acf7e244 SW |
2221 | fail: |
2222 | return error; | |
2223 | } | |
2224 | ||
b3b94faa DT |
2225 | /** |
2226 | * gfs2_rlist_add - add a RG to a list of RGs | |
70b0c365 | 2227 | * @ip: the inode |
b3b94faa DT |
2228 | * @rlist: the list of resource groups |
2229 | * @block: the block | |
2230 | * | |
2231 | * Figure out what RG a block belongs to and add that RG to the list | |
2232 | * | |
2233 | * FIXME: Don't use NOFAIL | |
2234 | * | |
2235 | */ | |
2236 | ||
70b0c365 | 2237 | void gfs2_rlist_add(struct gfs2_inode *ip, struct gfs2_rgrp_list *rlist, |
cd915493 | 2238 | u64 block) |
b3b94faa | 2239 | { |
70b0c365 | 2240 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
2241 | struct gfs2_rgrpd *rgd; |
2242 | struct gfs2_rgrpd **tmp; | |
2243 | unsigned int new_space; | |
2244 | unsigned int x; | |
2245 | ||
2246 | if (gfs2_assert_warn(sdp, !rlist->rl_ghs)) | |
2247 | return; | |
2248 | ||
70b0c365 SW |
2249 | if (ip->i_rgd && rgrp_contains_block(ip->i_rgd, block)) |
2250 | rgd = ip->i_rgd; | |
2251 | else | |
66fc061b | 2252 | rgd = gfs2_blk2rgrpd(sdp, block, 1); |
b3b94faa | 2253 | if (!rgd) { |
70b0c365 | 2254 | fs_err(sdp, "rlist_add: no rgrp for block %llu\n", (unsigned long long)block); |
b3b94faa DT |
2255 | return; |
2256 | } | |
70b0c365 | 2257 | ip->i_rgd = rgd; |
b3b94faa DT |
2258 | |
2259 | for (x = 0; x < rlist->rl_rgrps; x++) | |
2260 | if (rlist->rl_rgd[x] == rgd) | |
2261 | return; | |
2262 | ||
2263 | if (rlist->rl_rgrps == rlist->rl_space) { | |
2264 | new_space = rlist->rl_space + 10; | |
2265 | ||
2266 | tmp = kcalloc(new_space, sizeof(struct gfs2_rgrpd *), | |
dd894be8 | 2267 | GFP_NOFS | __GFP_NOFAIL); |
b3b94faa DT |
2268 | |
2269 | if (rlist->rl_rgd) { | |
2270 | memcpy(tmp, rlist->rl_rgd, | |
2271 | rlist->rl_space * sizeof(struct gfs2_rgrpd *)); | |
2272 | kfree(rlist->rl_rgd); | |
2273 | } | |
2274 | ||
2275 | rlist->rl_space = new_space; | |
2276 | rlist->rl_rgd = tmp; | |
2277 | } | |
2278 | ||
2279 | rlist->rl_rgd[rlist->rl_rgrps++] = rgd; | |
2280 | } | |
2281 | ||
2282 | /** | |
2283 | * gfs2_rlist_alloc - all RGs have been added to the rlist, now allocate | |
2284 | * and initialize an array of glock holders for them | |
2285 | * @rlist: the list of resource groups | |
2286 | * @state: the lock state to acquire the RG lock in | |
b3b94faa DT |
2287 | * |
2288 | * FIXME: Don't use NOFAIL | |
2289 | * | |
2290 | */ | |
2291 | ||
fe6c991c | 2292 | void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state) |
b3b94faa DT |
2293 | { |
2294 | unsigned int x; | |
2295 | ||
2296 | rlist->rl_ghs = kcalloc(rlist->rl_rgrps, sizeof(struct gfs2_holder), | |
dd894be8 | 2297 | GFP_NOFS | __GFP_NOFAIL); |
b3b94faa DT |
2298 | for (x = 0; x < rlist->rl_rgrps; x++) |
2299 | gfs2_holder_init(rlist->rl_rgd[x]->rd_gl, | |
fe6c991c | 2300 | state, 0, |
b3b94faa DT |
2301 | &rlist->rl_ghs[x]); |
2302 | } | |
2303 | ||
2304 | /** | |
2305 | * gfs2_rlist_free - free a resource group list | |
2306 | * @list: the list of resource groups | |
2307 | * | |
2308 | */ | |
2309 | ||
2310 | void gfs2_rlist_free(struct gfs2_rgrp_list *rlist) | |
2311 | { | |
2312 | unsigned int x; | |
2313 | ||
2314 | kfree(rlist->rl_rgd); | |
2315 | ||
2316 | if (rlist->rl_ghs) { | |
2317 | for (x = 0; x < rlist->rl_rgrps; x++) | |
2318 | gfs2_holder_uninit(&rlist->rl_ghs[x]); | |
2319 | kfree(rlist->rl_ghs); | |
8e2e0047 | 2320 | rlist->rl_ghs = NULL; |
b3b94faa DT |
2321 | } |
2322 | } | |
2323 |