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