]>
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 | ||
d77d1b58 JP |
10 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
11 | ||
b3b94faa DT |
12 | #include <linux/slab.h> |
13 | #include <linux/spinlock.h> | |
14 | #include <linux/completion.h> | |
15 | #include <linux/buffer_head.h> | |
f42faf4f | 16 | #include <linux/fs.h> |
5c676f6d | 17 | #include <linux/gfs2_ondisk.h> |
1f466a47 | 18 | #include <linux/prefetch.h> |
f15ab561 | 19 | #include <linux/blkdev.h> |
7c9ca621 | 20 | #include <linux/rbtree.h> |
9dbe9610 | 21 | #include <linux/random.h> |
b3b94faa DT |
22 | |
23 | #include "gfs2.h" | |
5c676f6d | 24 | #include "incore.h" |
b3b94faa DT |
25 | #include "glock.h" |
26 | #include "glops.h" | |
b3b94faa DT |
27 | #include "lops.h" |
28 | #include "meta_io.h" | |
29 | #include "quota.h" | |
30 | #include "rgrp.h" | |
31 | #include "super.h" | |
32 | #include "trans.h" | |
5c676f6d | 33 | #include "util.h" |
172e045a | 34 | #include "log.h" |
c8cdf479 | 35 | #include "inode.h" |
63997775 | 36 | #include "trace_gfs2.h" |
850d2d91 | 37 | #include "dir.h" |
b3b94faa | 38 | |
2c1e52aa | 39 | #define BFITNOENT ((u32)~0) |
6760bdcd | 40 | #define NO_BLOCK ((u64)~0) |
88c8ab1f | 41 | |
1f466a47 BP |
42 | #if BITS_PER_LONG == 32 |
43 | #define LBITMASK (0x55555555UL) | |
44 | #define LBITSKIP55 (0x55555555UL) | |
45 | #define LBITSKIP00 (0x00000000UL) | |
46 | #else | |
47 | #define LBITMASK (0x5555555555555555UL) | |
48 | #define LBITSKIP55 (0x5555555555555555UL) | |
49 | #define LBITSKIP00 (0x0000000000000000UL) | |
50 | #endif | |
51 | ||
88c8ab1f SW |
52 | /* |
53 | * These routines are used by the resource group routines (rgrp.c) | |
54 | * to keep track of block allocation. Each block is represented by two | |
feaa7bba SW |
55 | * bits. So, each byte represents GFS2_NBBY (i.e. 4) blocks. |
56 | * | |
57 | * 0 = Free | |
58 | * 1 = Used (not metadata) | |
59 | * 2 = Unlinked (still in use) inode | |
60 | * 3 = Used (metadata) | |
88c8ab1f SW |
61 | */ |
62 | ||
5ce13431 BP |
63 | struct gfs2_extent { |
64 | struct gfs2_rbm rbm; | |
65 | u32 len; | |
66 | }; | |
67 | ||
88c8ab1f SW |
68 | static const char valid_change[16] = { |
69 | /* current */ | |
feaa7bba | 70 | /* n */ 0, 1, 1, 1, |
88c8ab1f | 71 | /* e */ 1, 0, 0, 0, |
feaa7bba | 72 | /* w */ 0, 0, 0, 1, |
88c8ab1f SW |
73 | 1, 0, 0, 0 |
74 | }; | |
75 | ||
5ce13431 | 76 | static int gfs2_rbm_find(struct gfs2_rbm *rbm, u8 state, u32 *minext, |
8381e602 | 77 | const struct gfs2_inode *ip, bool nowrap); |
ff7f4cb4 SW |
78 | |
79 | ||
88c8ab1f SW |
80 | /** |
81 | * gfs2_setbit - Set a bit in the bitmaps | |
3e6339dd SW |
82 | * @rbm: The position of the bit to set |
83 | * @do_clone: Also set the clone bitmap, if it exists | |
88c8ab1f SW |
84 | * @new_state: the new state of the block |
85 | * | |
86 | */ | |
87 | ||
3e6339dd | 88 | static inline void gfs2_setbit(const struct gfs2_rbm *rbm, bool do_clone, |
06344b91 | 89 | unsigned char new_state) |
88c8ab1f | 90 | { |
b45e41d7 | 91 | unsigned char *byte1, *byte2, *end, cur_state; |
e579ed4f BP |
92 | struct gfs2_bitmap *bi = rbm_bi(rbm); |
93 | unsigned int buflen = bi->bi_len; | |
3e6339dd | 94 | const unsigned int bit = (rbm->offset % GFS2_NBBY) * GFS2_BIT_SIZE; |
88c8ab1f | 95 | |
e579ed4f BP |
96 | byte1 = bi->bi_bh->b_data + bi->bi_offset + (rbm->offset / GFS2_NBBY); |
97 | end = bi->bi_bh->b_data + bi->bi_offset + buflen; | |
88c8ab1f | 98 | |
b45e41d7 | 99 | BUG_ON(byte1 >= end); |
88c8ab1f | 100 | |
b45e41d7 | 101 | cur_state = (*byte1 >> bit) & GFS2_BIT_MASK; |
88c8ab1f | 102 | |
b45e41d7 | 103 | if (unlikely(!valid_change[new_state * 4 + cur_state])) { |
e54c78a2 BP |
104 | struct gfs2_sbd *sdp = rbm->rgd->rd_sbd; |
105 | ||
106 | fs_warn(sdp, "buf_blk = 0x%x old_state=%d, new_state=%d\n", | |
d77d1b58 | 107 | rbm->offset, cur_state, new_state); |
e54c78a2 BP |
108 | fs_warn(sdp, "rgrp=0x%llx bi_start=0x%x biblk: 0x%llx\n", |
109 | (unsigned long long)rbm->rgd->rd_addr, bi->bi_start, | |
110 | (unsigned long long)bi->bi_bh->b_blocknr); | |
111 | fs_warn(sdp, "bi_offset=0x%x bi_len=0x%x block=0x%llx\n", | |
112 | bi->bi_offset, bi->bi_len, | |
113 | (unsigned long long)gfs2_rbm_to_block(rbm)); | |
95c8e17f | 114 | dump_stack(); |
3e6339dd | 115 | gfs2_consist_rgrpd(rbm->rgd); |
b45e41d7 SW |
116 | return; |
117 | } | |
118 | *byte1 ^= (cur_state ^ new_state) << bit; | |
119 | ||
e579ed4f BP |
120 | if (do_clone && bi->bi_clone) { |
121 | byte2 = bi->bi_clone + bi->bi_offset + (rbm->offset / GFS2_NBBY); | |
b45e41d7 SW |
122 | cur_state = (*byte2 >> bit) & GFS2_BIT_MASK; |
123 | *byte2 ^= (cur_state ^ new_state) << bit; | |
124 | } | |
88c8ab1f SW |
125 | } |
126 | ||
127 | /** | |
128 | * gfs2_testbit - test a bit in the bitmaps | |
c04a2ef3 | 129 | * @rbm: The bit to test |
dffe12a8 BP |
130 | * @use_clone: If true, test the clone bitmap, not the official bitmap. |
131 | * | |
132 | * Some callers like gfs2_unaligned_extlen need to test the clone bitmaps, | |
133 | * not the "real" bitmaps, to avoid allocating recently freed blocks. | |
88c8ab1f | 134 | * |
c04a2ef3 | 135 | * Returns: The two bit block state of the requested bit |
88c8ab1f SW |
136 | */ |
137 | ||
dffe12a8 | 138 | static inline u8 gfs2_testbit(const struct gfs2_rbm *rbm, bool use_clone) |
88c8ab1f | 139 | { |
e579ed4f | 140 | struct gfs2_bitmap *bi = rbm_bi(rbm); |
dffe12a8 | 141 | const u8 *buffer; |
c04a2ef3 | 142 | const u8 *byte; |
88c8ab1f SW |
143 | unsigned int bit; |
144 | ||
dffe12a8 BP |
145 | if (use_clone && bi->bi_clone) |
146 | buffer = bi->bi_clone; | |
147 | else | |
148 | buffer = bi->bi_bh->b_data; | |
149 | buffer += bi->bi_offset; | |
c04a2ef3 SW |
150 | byte = buffer + (rbm->offset / GFS2_NBBY); |
151 | bit = (rbm->offset % GFS2_NBBY) * GFS2_BIT_SIZE; | |
88c8ab1f | 152 | |
c04a2ef3 | 153 | return (*byte >> bit) & GFS2_BIT_MASK; |
88c8ab1f SW |
154 | } |
155 | ||
223b2b88 SW |
156 | /** |
157 | * gfs2_bit_search | |
158 | * @ptr: Pointer to bitmap data | |
159 | * @mask: Mask to use (normally 0x55555.... but adjusted for search start) | |
160 | * @state: The state we are searching for | |
161 | * | |
162 | * We xor the bitmap data with a patter which is the bitwise opposite | |
163 | * of what we are looking for, this gives rise to a pattern of ones | |
164 | * wherever there is a match. Since we have two bits per entry, we | |
165 | * take this pattern, shift it down by one place and then and it with | |
166 | * the original. All the even bit positions (0,2,4, etc) then represent | |
167 | * successful matches, so we mask with 0x55555..... to remove the unwanted | |
168 | * odd bit positions. | |
169 | * | |
170 | * This allows searching of a whole u64 at once (32 blocks) with a | |
171 | * single test (on 64 bit arches). | |
172 | */ | |
173 | ||
174 | static inline u64 gfs2_bit_search(const __le64 *ptr, u64 mask, u8 state) | |
175 | { | |
176 | u64 tmp; | |
177 | static const u64 search[] = { | |
075ac448 HE |
178 | [0] = 0xffffffffffffffffULL, |
179 | [1] = 0xaaaaaaaaaaaaaaaaULL, | |
180 | [2] = 0x5555555555555555ULL, | |
181 | [3] = 0x0000000000000000ULL, | |
223b2b88 SW |
182 | }; |
183 | tmp = le64_to_cpu(*ptr) ^ search[state]; | |
184 | tmp &= (tmp >> 1); | |
185 | tmp &= mask; | |
186 | return tmp; | |
187 | } | |
188 | ||
8e2e0047 BP |
189 | /** |
190 | * rs_cmp - multi-block reservation range compare | |
191 | * @blk: absolute file system block number of the new reservation | |
192 | * @len: number of blocks in the new reservation | |
193 | * @rs: existing reservation to compare against | |
194 | * | |
195 | * returns: 1 if the block range is beyond the reach of the reservation | |
196 | * -1 if the block range is before the start of the reservation | |
197 | * 0 if the block range overlaps with the reservation | |
198 | */ | |
199 | static inline int rs_cmp(u64 blk, u32 len, struct gfs2_blkreserv *rs) | |
200 | { | |
4a993fb1 | 201 | u64 startblk = gfs2_rbm_to_block(&rs->rs_rbm); |
8e2e0047 BP |
202 | |
203 | if (blk >= startblk + rs->rs_free) | |
204 | return 1; | |
205 | if (blk + len - 1 < startblk) | |
206 | return -1; | |
207 | return 0; | |
208 | } | |
209 | ||
88c8ab1f SW |
210 | /** |
211 | * gfs2_bitfit - Search an rgrp's bitmap buffer to find a bit-pair representing | |
212 | * a block in a given allocation state. | |
886b1416 | 213 | * @buf: the buffer that holds the bitmaps |
223b2b88 | 214 | * @len: the length (in bytes) of the buffer |
88c8ab1f | 215 | * @goal: start search at this block's bit-pair (within @buffer) |
223b2b88 | 216 | * @state: GFS2_BLKST_XXX the state of the block we're looking for. |
88c8ab1f SW |
217 | * |
218 | * Scope of @goal and returned block number is only within this bitmap buffer, | |
219 | * not entire rgrp or filesystem. @buffer will be offset from the actual | |
223b2b88 SW |
220 | * beginning of a bitmap block buffer, skipping any header structures, but |
221 | * headers are always a multiple of 64 bits long so that the buffer is | |
222 | * always aligned to a 64 bit boundary. | |
223 | * | |
224 | * The size of the buffer is in bytes, but is it assumed that it is | |
fd589a8f | 225 | * always ok to read a complete multiple of 64 bits at the end |
223b2b88 | 226 | * of the block in case the end is no aligned to a natural boundary. |
88c8ab1f SW |
227 | * |
228 | * Return: the block number (bitmap buffer scope) that was found | |
229 | */ | |
230 | ||
02ab1721 HE |
231 | static u32 gfs2_bitfit(const u8 *buf, const unsigned int len, |
232 | u32 goal, u8 state) | |
88c8ab1f | 233 | { |
223b2b88 SW |
234 | u32 spoint = (goal << 1) & ((8*sizeof(u64)) - 1); |
235 | const __le64 *ptr = ((__le64 *)buf) + (goal >> 5); | |
236 | const __le64 *end = (__le64 *)(buf + ALIGN(len, sizeof(u64))); | |
237 | u64 tmp; | |
075ac448 | 238 | u64 mask = 0x5555555555555555ULL; |
223b2b88 SW |
239 | u32 bit; |
240 | ||
223b2b88 SW |
241 | /* Mask off bits we don't care about at the start of the search */ |
242 | mask <<= spoint; | |
243 | tmp = gfs2_bit_search(ptr, mask, state); | |
244 | ptr++; | |
245 | while(tmp == 0 && ptr < end) { | |
075ac448 | 246 | tmp = gfs2_bit_search(ptr, 0x5555555555555555ULL, state); |
223b2b88 | 247 | ptr++; |
1f466a47 | 248 | } |
223b2b88 SW |
249 | /* Mask off any bits which are more than len bytes from the start */ |
250 | if (ptr == end && (len & (sizeof(u64) - 1))) | |
251 | tmp &= (((u64)~0) >> (64 - 8*(len & (sizeof(u64) - 1)))); | |
252 | /* Didn't find anything, so return */ | |
253 | if (tmp == 0) | |
254 | return BFITNOENT; | |
255 | ptr--; | |
d8bd504a | 256 | bit = __ffs64(tmp); |
223b2b88 SW |
257 | bit /= 2; /* two bits per entry in the bitmap */ |
258 | return (((const unsigned char *)ptr - buf) * GFS2_NBBY) + bit; | |
88c8ab1f SW |
259 | } |
260 | ||
ff7f4cb4 SW |
261 | /** |
262 | * gfs2_rbm_from_block - Set the rbm based upon rgd and block number | |
263 | * @rbm: The rbm with rgd already set correctly | |
264 | * @block: The block number (filesystem relative) | |
265 | * | |
266 | * This sets the bi and offset members of an rbm based on a | |
267 | * resource group and a filesystem relative block number. The | |
268 | * resource group must be set in the rbm on entry, the bi and | |
269 | * offset members will be set by this function. | |
270 | * | |
271 | * Returns: 0 on success, or an error code | |
272 | */ | |
273 | ||
274 | static int gfs2_rbm_from_block(struct gfs2_rbm *rbm, u64 block) | |
275 | { | |
276 | u64 rblock = block - rbm->rgd->rd_data0; | |
ff7f4cb4 SW |
277 | |
278 | if (WARN_ON_ONCE(rblock > UINT_MAX)) | |
279 | return -EINVAL; | |
280 | if (block >= rbm->rgd->rd_data0 + rbm->rgd->rd_data) | |
281 | return -E2BIG; | |
282 | ||
e579ed4f | 283 | rbm->bii = 0; |
a68a0a35 BP |
284 | rbm->offset = (u32)(rblock); |
285 | /* Check if the block is within the first block */ | |
e579ed4f | 286 | if (rbm->offset < rbm_bi(rbm)->bi_blocks) |
a68a0a35 BP |
287 | return 0; |
288 | ||
289 | /* Adjust for the size diff between gfs2_meta_header and gfs2_rgrp */ | |
290 | rbm->offset += (sizeof(struct gfs2_rgrp) - | |
291 | sizeof(struct gfs2_meta_header)) * GFS2_NBBY; | |
e579ed4f BP |
292 | rbm->bii = rbm->offset / rbm->rgd->rd_sbd->sd_blocks_per_bitmap; |
293 | rbm->offset -= rbm->bii * rbm->rgd->rd_sbd->sd_blocks_per_bitmap; | |
ff7f4cb4 SW |
294 | return 0; |
295 | } | |
296 | ||
149ed7f5 BP |
297 | /** |
298 | * gfs2_rbm_incr - increment an rbm structure | |
299 | * @rbm: The rbm with rgd already set correctly | |
300 | * | |
301 | * This function takes an existing rbm structure and increments it to the next | |
302 | * viable block offset. | |
303 | * | |
304 | * Returns: If incrementing the offset would cause the rbm to go past the | |
305 | * end of the rgrp, true is returned, otherwise false. | |
306 | * | |
307 | */ | |
308 | ||
309 | static bool gfs2_rbm_incr(struct gfs2_rbm *rbm) | |
310 | { | |
311 | if (rbm->offset + 1 < rbm_bi(rbm)->bi_blocks) { /* in the same bitmap */ | |
312 | rbm->offset++; | |
313 | return false; | |
314 | } | |
315 | if (rbm->bii == rbm->rgd->rd_length - 1) /* at the last bitmap */ | |
316 | return true; | |
317 | ||
318 | rbm->offset = 0; | |
319 | rbm->bii++; | |
320 | return false; | |
321 | } | |
322 | ||
ff7f4cb4 SW |
323 | /** |
324 | * gfs2_unaligned_extlen - Look for free blocks which are not byte aligned | |
325 | * @rbm: Position to search (value/result) | |
326 | * @n_unaligned: Number of unaligned blocks to check | |
327 | * @len: Decremented for each block found (terminate on zero) | |
328 | * | |
329 | * Returns: true if a non-free block is encountered | |
330 | */ | |
331 | ||
332 | static bool gfs2_unaligned_extlen(struct gfs2_rbm *rbm, u32 n_unaligned, u32 *len) | |
333 | { | |
ff7f4cb4 SW |
334 | u32 n; |
335 | u8 res; | |
336 | ||
337 | for (n = 0; n < n_unaligned; n++) { | |
dffe12a8 | 338 | res = gfs2_testbit(rbm, true); |
ff7f4cb4 SW |
339 | if (res != GFS2_BLKST_FREE) |
340 | return true; | |
341 | (*len)--; | |
342 | if (*len == 0) | |
343 | return true; | |
149ed7f5 | 344 | if (gfs2_rbm_incr(rbm)) |
ff7f4cb4 SW |
345 | return true; |
346 | } | |
347 | ||
348 | return false; | |
349 | } | |
350 | ||
351 | /** | |
352 | * gfs2_free_extlen - Return extent length of free blocks | |
27ff6a0f | 353 | * @rrbm: Starting position |
ff7f4cb4 SW |
354 | * @len: Max length to check |
355 | * | |
356 | * Starting at the block specified by the rbm, see how many free blocks | |
357 | * there are, not reading more than len blocks ahead. This can be done | |
358 | * using memchr_inv when the blocks are byte aligned, but has to be done | |
359 | * on a block by block basis in case of unaligned blocks. Also this | |
360 | * function can cope with bitmap boundaries (although it must stop on | |
361 | * a resource group boundary) | |
362 | * | |
363 | * Returns: Number of free blocks in the extent | |
364 | */ | |
365 | ||
366 | static u32 gfs2_free_extlen(const struct gfs2_rbm *rrbm, u32 len) | |
367 | { | |
368 | struct gfs2_rbm rbm = *rrbm; | |
369 | u32 n_unaligned = rbm.offset & 3; | |
370 | u32 size = len; | |
371 | u32 bytes; | |
372 | u32 chunk_size; | |
373 | u8 *ptr, *start, *end; | |
374 | u64 block; | |
e579ed4f | 375 | struct gfs2_bitmap *bi; |
ff7f4cb4 SW |
376 | |
377 | if (n_unaligned && | |
378 | gfs2_unaligned_extlen(&rbm, 4 - n_unaligned, &len)) | |
379 | goto out; | |
380 | ||
3701530a | 381 | n_unaligned = len & 3; |
ff7f4cb4 SW |
382 | /* Start is now byte aligned */ |
383 | while (len > 3) { | |
e579ed4f BP |
384 | bi = rbm_bi(&rbm); |
385 | start = bi->bi_bh->b_data; | |
386 | if (bi->bi_clone) | |
387 | start = bi->bi_clone; | |
e579ed4f | 388 | start += bi->bi_offset; |
dc8fbb03 | 389 | end = start + bi->bi_len; |
ff7f4cb4 SW |
390 | BUG_ON(rbm.offset & 3); |
391 | start += (rbm.offset / GFS2_NBBY); | |
392 | bytes = min_t(u32, len / GFS2_NBBY, (end - start)); | |
393 | ptr = memchr_inv(start, 0, bytes); | |
394 | chunk_size = ((ptr == NULL) ? bytes : (ptr - start)); | |
395 | chunk_size *= GFS2_NBBY; | |
396 | BUG_ON(len < chunk_size); | |
397 | len -= chunk_size; | |
398 | block = gfs2_rbm_to_block(&rbm); | |
15bd50ad BP |
399 | if (gfs2_rbm_from_block(&rbm, block + chunk_size)) { |
400 | n_unaligned = 0; | |
ff7f4cb4 | 401 | break; |
15bd50ad BP |
402 | } |
403 | if (ptr) { | |
404 | n_unaligned = 3; | |
405 | break; | |
406 | } | |
ff7f4cb4 SW |
407 | n_unaligned = len & 3; |
408 | } | |
409 | ||
410 | /* Deal with any bits left over at the end */ | |
411 | if (n_unaligned) | |
412 | gfs2_unaligned_extlen(&rbm, n_unaligned, &len); | |
413 | out: | |
414 | return size - len; | |
415 | } | |
416 | ||
88c8ab1f SW |
417 | /** |
418 | * gfs2_bitcount - count the number of bits in a certain state | |
886b1416 | 419 | * @rgd: the resource group descriptor |
88c8ab1f SW |
420 | * @buffer: the buffer that holds the bitmaps |
421 | * @buflen: the length (in bytes) of the buffer | |
422 | * @state: the state of the block we're looking for | |
423 | * | |
424 | * Returns: The number of bits | |
425 | */ | |
426 | ||
110acf38 SW |
427 | static u32 gfs2_bitcount(struct gfs2_rgrpd *rgd, const u8 *buffer, |
428 | unsigned int buflen, u8 state) | |
88c8ab1f | 429 | { |
110acf38 SW |
430 | const u8 *byte = buffer; |
431 | const u8 *end = buffer + buflen; | |
432 | const u8 state1 = state << 2; | |
433 | const u8 state2 = state << 4; | |
434 | const u8 state3 = state << 6; | |
cd915493 | 435 | u32 count = 0; |
88c8ab1f SW |
436 | |
437 | for (; byte < end; byte++) { | |
438 | if (((*byte) & 0x03) == state) | |
439 | count++; | |
440 | if (((*byte) & 0x0C) == state1) | |
441 | count++; | |
442 | if (((*byte) & 0x30) == state2) | |
443 | count++; | |
444 | if (((*byte) & 0xC0) == state3) | |
445 | count++; | |
446 | } | |
447 | ||
448 | return count; | |
449 | } | |
450 | ||
b3b94faa DT |
451 | /** |
452 | * gfs2_rgrp_verify - Verify that a resource group is consistent | |
b3b94faa DT |
453 | * @rgd: the rgrp |
454 | * | |
455 | */ | |
456 | ||
457 | void gfs2_rgrp_verify(struct gfs2_rgrpd *rgd) | |
458 | { | |
459 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
460 | struct gfs2_bitmap *bi = NULL; | |
bb8d8a6f | 461 | u32 length = rgd->rd_length; |
cd915493 | 462 | u32 count[4], tmp; |
b3b94faa DT |
463 | int buf, x; |
464 | ||
cd915493 | 465 | memset(count, 0, 4 * sizeof(u32)); |
b3b94faa DT |
466 | |
467 | /* Count # blocks in each of 4 possible allocation states */ | |
468 | for (buf = 0; buf < length; buf++) { | |
469 | bi = rgd->rd_bits + buf; | |
470 | for (x = 0; x < 4; x++) | |
471 | count[x] += gfs2_bitcount(rgd, | |
472 | bi->bi_bh->b_data + | |
473 | bi->bi_offset, | |
474 | bi->bi_len, x); | |
475 | } | |
476 | ||
cfc8b549 | 477 | if (count[0] != rgd->rd_free) { |
b3b94faa DT |
478 | if (gfs2_consist_rgrpd(rgd)) |
479 | fs_err(sdp, "free data mismatch: %u != %u\n", | |
cfc8b549 | 480 | count[0], rgd->rd_free); |
b3b94faa DT |
481 | return; |
482 | } | |
483 | ||
73f74948 | 484 | tmp = rgd->rd_data - rgd->rd_free - rgd->rd_dinodes; |
6b946170 | 485 | if (count[1] != tmp) { |
b3b94faa DT |
486 | if (gfs2_consist_rgrpd(rgd)) |
487 | fs_err(sdp, "used data mismatch: %u != %u\n", | |
488 | count[1], tmp); | |
489 | return; | |
490 | } | |
491 | ||
6b946170 | 492 | if (count[2] + count[3] != rgd->rd_dinodes) { |
b3b94faa | 493 | if (gfs2_consist_rgrpd(rgd)) |
feaa7bba | 494 | fs_err(sdp, "used metadata mismatch: %u != %u\n", |
6b946170 | 495 | count[2] + count[3], rgd->rd_dinodes); |
b3b94faa DT |
496 | return; |
497 | } | |
b3b94faa DT |
498 | } |
499 | ||
b3b94faa DT |
500 | /** |
501 | * gfs2_blk2rgrpd - Find resource group for a given data/meta block number | |
502 | * @sdp: The GFS2 superblock | |
886b1416 BP |
503 | * @blk: The data block number |
504 | * @exact: True if this needs to be an exact match | |
b3b94faa | 505 | * |
90bcab99 SW |
506 | * The @exact argument should be set to true by most callers. The exception |
507 | * is when we need to match blocks which are not represented by the rgrp | |
508 | * bitmap, but which are part of the rgrp (i.e. padding blocks) which are | |
509 | * there for alignment purposes. Another way of looking at it is that @exact | |
510 | * matches only valid data/metadata blocks, but with @exact false, it will | |
511 | * match any block within the extent of the rgrp. | |
512 | * | |
b3b94faa DT |
513 | * Returns: The resource group, or NULL if not found |
514 | */ | |
515 | ||
66fc061b | 516 | struct gfs2_rgrpd *gfs2_blk2rgrpd(struct gfs2_sbd *sdp, u64 blk, bool exact) |
b3b94faa | 517 | { |
66fc061b | 518 | struct rb_node *n, *next; |
f75bbfb4 | 519 | struct gfs2_rgrpd *cur; |
b3b94faa DT |
520 | |
521 | spin_lock(&sdp->sd_rindex_spin); | |
66fc061b SW |
522 | n = sdp->sd_rindex_tree.rb_node; |
523 | while (n) { | |
524 | cur = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
525 | next = NULL; | |
7c9ca621 | 526 | if (blk < cur->rd_addr) |
66fc061b | 527 | next = n->rb_left; |
f75bbfb4 | 528 | else if (blk >= cur->rd_data0 + cur->rd_data) |
66fc061b SW |
529 | next = n->rb_right; |
530 | if (next == NULL) { | |
b3b94faa | 531 | spin_unlock(&sdp->sd_rindex_spin); |
66fc061b SW |
532 | if (exact) { |
533 | if (blk < cur->rd_addr) | |
534 | return NULL; | |
535 | if (blk >= cur->rd_data0 + cur->rd_data) | |
536 | return NULL; | |
537 | } | |
7c9ca621 | 538 | return cur; |
b3b94faa | 539 | } |
66fc061b | 540 | n = next; |
b3b94faa | 541 | } |
b3b94faa DT |
542 | spin_unlock(&sdp->sd_rindex_spin); |
543 | ||
544 | return NULL; | |
545 | } | |
546 | ||
547 | /** | |
548 | * gfs2_rgrpd_get_first - get the first Resource Group in the filesystem | |
549 | * @sdp: The GFS2 superblock | |
550 | * | |
551 | * Returns: The first rgrp in the filesystem | |
552 | */ | |
553 | ||
554 | struct gfs2_rgrpd *gfs2_rgrpd_get_first(struct gfs2_sbd *sdp) | |
555 | { | |
7c9ca621 BP |
556 | const struct rb_node *n; |
557 | struct gfs2_rgrpd *rgd; | |
558 | ||
8339ee54 | 559 | spin_lock(&sdp->sd_rindex_spin); |
7c9ca621 BP |
560 | n = rb_first(&sdp->sd_rindex_tree); |
561 | rgd = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
8339ee54 | 562 | spin_unlock(&sdp->sd_rindex_spin); |
7c9ca621 BP |
563 | |
564 | return rgd; | |
b3b94faa DT |
565 | } |
566 | ||
567 | /** | |
568 | * gfs2_rgrpd_get_next - get the next RG | |
886b1416 | 569 | * @rgd: the resource group descriptor |
b3b94faa DT |
570 | * |
571 | * Returns: The next rgrp | |
572 | */ | |
573 | ||
574 | struct gfs2_rgrpd *gfs2_rgrpd_get_next(struct gfs2_rgrpd *rgd) | |
575 | { | |
7c9ca621 BP |
576 | struct gfs2_sbd *sdp = rgd->rd_sbd; |
577 | const struct rb_node *n; | |
578 | ||
579 | spin_lock(&sdp->sd_rindex_spin); | |
580 | n = rb_next(&rgd->rd_node); | |
581 | if (n == NULL) | |
582 | n = rb_first(&sdp->sd_rindex_tree); | |
583 | ||
584 | if (unlikely(&rgd->rd_node == n)) { | |
585 | spin_unlock(&sdp->sd_rindex_spin); | |
b3b94faa | 586 | return NULL; |
7c9ca621 BP |
587 | } |
588 | rgd = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
589 | spin_unlock(&sdp->sd_rindex_spin); | |
590 | return rgd; | |
b3b94faa DT |
591 | } |
592 | ||
00a158be AD |
593 | void check_and_update_goal(struct gfs2_inode *ip) |
594 | { | |
595 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
596 | if (!ip->i_goal || gfs2_blk2rgrpd(sdp, ip->i_goal, 1) == NULL) | |
597 | ip->i_goal = ip->i_no_addr; | |
598 | } | |
599 | ||
8339ee54 SW |
600 | void gfs2_free_clones(struct gfs2_rgrpd *rgd) |
601 | { | |
602 | int x; | |
603 | ||
604 | for (x = 0; x < rgd->rd_length; x++) { | |
605 | struct gfs2_bitmap *bi = rgd->rd_bits + x; | |
606 | kfree(bi->bi_clone); | |
607 | bi->bi_clone = NULL; | |
608 | } | |
609 | } | |
610 | ||
0a305e49 | 611 | /** |
b54e9a0b BP |
612 | * gfs2_rsqa_alloc - make sure we have a reservation assigned to the inode |
613 | * plus a quota allocations data structure, if necessary | |
0a305e49 BP |
614 | * @ip: the inode for this reservation |
615 | */ | |
b54e9a0b | 616 | int gfs2_rsqa_alloc(struct gfs2_inode *ip) |
0a305e49 | 617 | { |
a097dc7e | 618 | return gfs2_qa_alloc(ip); |
0a305e49 BP |
619 | } |
620 | ||
9e733d39 | 621 | static void dump_rs(struct seq_file *seq, const struct gfs2_blkreserv *rs) |
8e2e0047 | 622 | { |
f85c10e2 BP |
623 | struct gfs2_inode *ip = container_of(rs, struct gfs2_inode, i_res); |
624 | ||
9e733d39 | 625 | gfs2_print_dbg(seq, " B: n:%llu s:%llu b:%u f:%u\n", |
f85c10e2 | 626 | (unsigned long long)ip->i_no_addr, |
9e733d39 | 627 | (unsigned long long)gfs2_rbm_to_block(&rs->rs_rbm), |
4a993fb1 | 628 | rs->rs_rbm.offset, rs->rs_free); |
8e2e0047 BP |
629 | } |
630 | ||
0a305e49 | 631 | /** |
8e2e0047 BP |
632 | * __rs_deltree - remove a multi-block reservation from the rgd tree |
633 | * @rs: The reservation to remove | |
634 | * | |
635 | */ | |
20095218 | 636 | static void __rs_deltree(struct gfs2_blkreserv *rs) |
8e2e0047 BP |
637 | { |
638 | struct gfs2_rgrpd *rgd; | |
639 | ||
640 | if (!gfs2_rs_active(rs)) | |
641 | return; | |
642 | ||
4a993fb1 | 643 | rgd = rs->rs_rbm.rgd; |
9e733d39 | 644 | trace_gfs2_rs(rs, TRACE_RS_TREEDEL); |
4a993fb1 | 645 | rb_erase(&rs->rs_node, &rgd->rd_rstree); |
24d634e8 | 646 | RB_CLEAR_NODE(&rs->rs_node); |
8e2e0047 BP |
647 | |
648 | if (rs->rs_free) { | |
e579ed4f BP |
649 | struct gfs2_bitmap *bi = rbm_bi(&rs->rs_rbm); |
650 | ||
20095218 | 651 | /* return reserved blocks to the rgrp */ |
4a993fb1 SW |
652 | BUG_ON(rs->rs_rbm.rgd->rd_reserved < rs->rs_free); |
653 | rs->rs_rbm.rgd->rd_reserved -= rs->rs_free; | |
5ea5050c BP |
654 | /* The rgrp extent failure point is likely not to increase; |
655 | it will only do so if the freed blocks are somehow | |
656 | contiguous with a span of free blocks that follows. Still, | |
657 | it will force the number to be recalculated later. */ | |
658 | rgd->rd_extfail_pt += rs->rs_free; | |
8e2e0047 | 659 | rs->rs_free = 0; |
e579ed4f | 660 | clear_bit(GBF_FULL, &bi->bi_flags); |
8e2e0047 | 661 | } |
8e2e0047 BP |
662 | } |
663 | ||
664 | /** | |
665 | * gfs2_rs_deltree - remove a multi-block reservation from the rgd tree | |
666 | * @rs: The reservation to remove | |
667 | * | |
668 | */ | |
20095218 | 669 | void gfs2_rs_deltree(struct gfs2_blkreserv *rs) |
8e2e0047 BP |
670 | { |
671 | struct gfs2_rgrpd *rgd; | |
672 | ||
4a993fb1 SW |
673 | rgd = rs->rs_rbm.rgd; |
674 | if (rgd) { | |
675 | spin_lock(&rgd->rd_rsspin); | |
20095218 | 676 | __rs_deltree(rs); |
44f52122 | 677 | BUG_ON(rs->rs_free); |
4a993fb1 SW |
678 | spin_unlock(&rgd->rd_rsspin); |
679 | } | |
8e2e0047 BP |
680 | } |
681 | ||
682 | /** | |
b54e9a0b | 683 | * gfs2_rsqa_delete - delete a multi-block reservation and quota allocation |
0a305e49 | 684 | * @ip: The inode for this reservation |
af5c2697 | 685 | * @wcount: The inode's write count, or NULL |
0a305e49 BP |
686 | * |
687 | */ | |
b54e9a0b | 688 | void gfs2_rsqa_delete(struct gfs2_inode *ip, atomic_t *wcount) |
0a305e49 BP |
689 | { |
690 | down_write(&ip->i_rw_mutex); | |
44f52122 | 691 | if ((wcount == NULL) || (atomic_read(wcount) <= 1)) |
a097dc7e | 692 | gfs2_rs_deltree(&ip->i_res); |
0a305e49 | 693 | up_write(&ip->i_rw_mutex); |
a097dc7e | 694 | gfs2_qa_delete(ip, wcount); |
0a305e49 BP |
695 | } |
696 | ||
8e2e0047 BP |
697 | /** |
698 | * return_all_reservations - return all reserved blocks back to the rgrp. | |
699 | * @rgd: the rgrp that needs its space back | |
700 | * | |
701 | * We previously reserved a bunch of blocks for allocation. Now we need to | |
702 | * give them back. This leave the reservation structures in tact, but removes | |
703 | * all of their corresponding "no-fly zones". | |
704 | */ | |
705 | static void return_all_reservations(struct gfs2_rgrpd *rgd) | |
706 | { | |
707 | struct rb_node *n; | |
708 | struct gfs2_blkreserv *rs; | |
709 | ||
710 | spin_lock(&rgd->rd_rsspin); | |
711 | while ((n = rb_first(&rgd->rd_rstree))) { | |
712 | rs = rb_entry(n, struct gfs2_blkreserv, rs_node); | |
20095218 | 713 | __rs_deltree(rs); |
8e2e0047 BP |
714 | } |
715 | spin_unlock(&rgd->rd_rsspin); | |
716 | } | |
717 | ||
8339ee54 | 718 | void gfs2_clear_rgrpd(struct gfs2_sbd *sdp) |
b3b94faa | 719 | { |
7c9ca621 | 720 | struct rb_node *n; |
b3b94faa DT |
721 | struct gfs2_rgrpd *rgd; |
722 | struct gfs2_glock *gl; | |
723 | ||
7c9ca621 BP |
724 | while ((n = rb_first(&sdp->sd_rindex_tree))) { |
725 | rgd = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
b3b94faa DT |
726 | gl = rgd->rd_gl; |
727 | ||
7c9ca621 | 728 | rb_erase(n, &sdp->sd_rindex_tree); |
b3b94faa DT |
729 | |
730 | if (gl) { | |
7023a0b1 | 731 | glock_clear_object(gl, rgd); |
b3b94faa DT |
732 | gfs2_glock_put(gl); |
733 | } | |
734 | ||
8339ee54 | 735 | gfs2_free_clones(rgd); |
b3b94faa | 736 | kfree(rgd->rd_bits); |
36e4ad03 | 737 | rgd->rd_bits = NULL; |
8e2e0047 | 738 | return_all_reservations(rgd); |
6bdd9be6 | 739 | kmem_cache_free(gfs2_rgrpd_cachep, rgd); |
b3b94faa DT |
740 | } |
741 | } | |
742 | ||
bb8d8a6f SW |
743 | static void gfs2_rindex_print(const struct gfs2_rgrpd *rgd) |
744 | { | |
e54c78a2 BP |
745 | struct gfs2_sbd *sdp = rgd->rd_sbd; |
746 | ||
747 | fs_info(sdp, "ri_addr = %llu\n", (unsigned long long)rgd->rd_addr); | |
748 | fs_info(sdp, "ri_length = %u\n", rgd->rd_length); | |
749 | fs_info(sdp, "ri_data0 = %llu\n", (unsigned long long)rgd->rd_data0); | |
750 | fs_info(sdp, "ri_data = %u\n", rgd->rd_data); | |
751 | fs_info(sdp, "ri_bitbytes = %u\n", rgd->rd_bitbytes); | |
bb8d8a6f SW |
752 | } |
753 | ||
b3b94faa DT |
754 | /** |
755 | * gfs2_compute_bitstructs - Compute the bitmap sizes | |
756 | * @rgd: The resource group descriptor | |
757 | * | |
758 | * Calculates bitmap descriptors, one for each block that contains bitmap data | |
759 | * | |
760 | * Returns: errno | |
761 | */ | |
762 | ||
763 | static int compute_bitstructs(struct gfs2_rgrpd *rgd) | |
764 | { | |
765 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
766 | struct gfs2_bitmap *bi; | |
bb8d8a6f | 767 | u32 length = rgd->rd_length; /* # blocks in hdr & bitmap */ |
cd915493 | 768 | u32 bytes_left, bytes; |
b3b94faa DT |
769 | int x; |
770 | ||
feaa7bba SW |
771 | if (!length) |
772 | return -EINVAL; | |
773 | ||
dd894be8 | 774 | rgd->rd_bits = kcalloc(length, sizeof(struct gfs2_bitmap), GFP_NOFS); |
b3b94faa DT |
775 | if (!rgd->rd_bits) |
776 | return -ENOMEM; | |
777 | ||
bb8d8a6f | 778 | bytes_left = rgd->rd_bitbytes; |
b3b94faa DT |
779 | |
780 | for (x = 0; x < length; x++) { | |
781 | bi = rgd->rd_bits + x; | |
782 | ||
60a0b8f9 | 783 | bi->bi_flags = 0; |
b3b94faa DT |
784 | /* small rgrp; bitmap stored completely in header block */ |
785 | if (length == 1) { | |
786 | bytes = bytes_left; | |
787 | bi->bi_offset = sizeof(struct gfs2_rgrp); | |
788 | bi->bi_start = 0; | |
789 | bi->bi_len = bytes; | |
7e230f57 | 790 | bi->bi_blocks = bytes * GFS2_NBBY; |
b3b94faa DT |
791 | /* header block */ |
792 | } else if (x == 0) { | |
793 | bytes = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_rgrp); | |
794 | bi->bi_offset = sizeof(struct gfs2_rgrp); | |
795 | bi->bi_start = 0; | |
796 | bi->bi_len = bytes; | |
7e230f57 | 797 | bi->bi_blocks = bytes * GFS2_NBBY; |
b3b94faa DT |
798 | /* last block */ |
799 | } else if (x + 1 == length) { | |
800 | bytes = bytes_left; | |
801 | bi->bi_offset = sizeof(struct gfs2_meta_header); | |
bb8d8a6f | 802 | bi->bi_start = rgd->rd_bitbytes - bytes_left; |
b3b94faa | 803 | bi->bi_len = bytes; |
7e230f57 | 804 | bi->bi_blocks = bytes * GFS2_NBBY; |
b3b94faa DT |
805 | /* other blocks */ |
806 | } else { | |
568f4c96 SW |
807 | bytes = sdp->sd_sb.sb_bsize - |
808 | sizeof(struct gfs2_meta_header); | |
b3b94faa | 809 | bi->bi_offset = sizeof(struct gfs2_meta_header); |
bb8d8a6f | 810 | bi->bi_start = rgd->rd_bitbytes - bytes_left; |
b3b94faa | 811 | bi->bi_len = bytes; |
7e230f57 | 812 | bi->bi_blocks = bytes * GFS2_NBBY; |
b3b94faa DT |
813 | } |
814 | ||
815 | bytes_left -= bytes; | |
816 | } | |
817 | ||
818 | if (bytes_left) { | |
819 | gfs2_consist_rgrpd(rgd); | |
820 | return -EIO; | |
821 | } | |
822 | bi = rgd->rd_bits + (length - 1); | |
bb8d8a6f | 823 | if ((bi->bi_start + bi->bi_len) * GFS2_NBBY != rgd->rd_data) { |
b3b94faa | 824 | if (gfs2_consist_rgrpd(rgd)) { |
bb8d8a6f | 825 | gfs2_rindex_print(rgd); |
b3b94faa DT |
826 | fs_err(sdp, "start=%u len=%u offset=%u\n", |
827 | bi->bi_start, bi->bi_len, bi->bi_offset); | |
828 | } | |
829 | return -EIO; | |
830 | } | |
831 | ||
832 | return 0; | |
833 | } | |
834 | ||
7ae8fa84 RP |
835 | /** |
836 | * gfs2_ri_total - Total up the file system space, according to the rindex. | |
886b1416 | 837 | * @sdp: the filesystem |
7ae8fa84 RP |
838 | * |
839 | */ | |
840 | u64 gfs2_ri_total(struct gfs2_sbd *sdp) | |
841 | { | |
842 | u64 total_data = 0; | |
843 | struct inode *inode = sdp->sd_rindex; | |
844 | struct gfs2_inode *ip = GFS2_I(inode); | |
7ae8fa84 | 845 | char buf[sizeof(struct gfs2_rindex)]; |
7ae8fa84 RP |
846 | int error, rgrps; |
847 | ||
7ae8fa84 RP |
848 | for (rgrps = 0;; rgrps++) { |
849 | loff_t pos = rgrps * sizeof(struct gfs2_rindex); | |
850 | ||
bcd7278d | 851 | if (pos + sizeof(struct gfs2_rindex) > i_size_read(inode)) |
7ae8fa84 | 852 | break; |
4306629e | 853 | error = gfs2_internal_read(ip, buf, &pos, |
7ae8fa84 RP |
854 | sizeof(struct gfs2_rindex)); |
855 | if (error != sizeof(struct gfs2_rindex)) | |
856 | break; | |
bb8d8a6f | 857 | total_data += be32_to_cpu(((struct gfs2_rindex *)buf)->ri_data); |
7ae8fa84 | 858 | } |
7ae8fa84 RP |
859 | return total_data; |
860 | } | |
861 | ||
6aad1c3d | 862 | static int rgd_insert(struct gfs2_rgrpd *rgd) |
7c9ca621 BP |
863 | { |
864 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
865 | struct rb_node **newn = &sdp->sd_rindex_tree.rb_node, *parent = NULL; | |
866 | ||
867 | /* Figure out where to put new node */ | |
868 | while (*newn) { | |
869 | struct gfs2_rgrpd *cur = rb_entry(*newn, struct gfs2_rgrpd, | |
870 | rd_node); | |
871 | ||
872 | parent = *newn; | |
873 | if (rgd->rd_addr < cur->rd_addr) | |
874 | newn = &((*newn)->rb_left); | |
875 | else if (rgd->rd_addr > cur->rd_addr) | |
876 | newn = &((*newn)->rb_right); | |
877 | else | |
6aad1c3d | 878 | return -EEXIST; |
7c9ca621 BP |
879 | } |
880 | ||
881 | rb_link_node(&rgd->rd_node, parent, newn); | |
882 | rb_insert_color(&rgd->rd_node, &sdp->sd_rindex_tree); | |
6aad1c3d BP |
883 | sdp->sd_rgrps++; |
884 | return 0; | |
7c9ca621 BP |
885 | } |
886 | ||
b3b94faa | 887 | /** |
6c53267f | 888 | * read_rindex_entry - Pull in a new resource index entry from the disk |
4306629e | 889 | * @ip: Pointer to the rindex inode |
b3b94faa | 890 | * |
8339ee54 | 891 | * Returns: 0 on success, > 0 on EOF, error code otherwise |
6c53267f RP |
892 | */ |
893 | ||
4306629e | 894 | static int read_rindex_entry(struct gfs2_inode *ip) |
6c53267f RP |
895 | { |
896 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
7005c3e4 | 897 | const unsigned bsize = sdp->sd_sb.sb_bsize; |
6c53267f | 898 | loff_t pos = sdp->sd_rgrps * sizeof(struct gfs2_rindex); |
8339ee54 | 899 | struct gfs2_rindex buf; |
6c53267f RP |
900 | int error; |
901 | struct gfs2_rgrpd *rgd; | |
902 | ||
8339ee54 SW |
903 | if (pos >= i_size_read(&ip->i_inode)) |
904 | return 1; | |
905 | ||
4306629e | 906 | error = gfs2_internal_read(ip, (char *)&buf, &pos, |
6c53267f | 907 | sizeof(struct gfs2_rindex)); |
8339ee54 SW |
908 | |
909 | if (error != sizeof(struct gfs2_rindex)) | |
910 | return (error == 0) ? 1 : error; | |
6c53267f | 911 | |
6bdd9be6 | 912 | rgd = kmem_cache_zalloc(gfs2_rgrpd_cachep, GFP_NOFS); |
6c53267f RP |
913 | error = -ENOMEM; |
914 | if (!rgd) | |
915 | return error; | |
916 | ||
6c53267f | 917 | rgd->rd_sbd = sdp; |
8339ee54 SW |
918 | rgd->rd_addr = be64_to_cpu(buf.ri_addr); |
919 | rgd->rd_length = be32_to_cpu(buf.ri_length); | |
920 | rgd->rd_data0 = be64_to_cpu(buf.ri_data0); | |
921 | rgd->rd_data = be32_to_cpu(buf.ri_data); | |
922 | rgd->rd_bitbytes = be32_to_cpu(buf.ri_bitbytes); | |
8e2e0047 | 923 | spin_lock_init(&rgd->rd_rsspin); |
7c9ca621 | 924 | |
6c53267f RP |
925 | error = compute_bitstructs(rgd); |
926 | if (error) | |
8339ee54 | 927 | goto fail; |
6c53267f | 928 | |
bb8d8a6f | 929 | error = gfs2_glock_get(sdp, rgd->rd_addr, |
6c53267f RP |
930 | &gfs2_rgrp_glops, CREATE, &rgd->rd_gl); |
931 | if (error) | |
8339ee54 | 932 | goto fail; |
6c53267f | 933 | |
4e2f8849 | 934 | rgd->rd_rgl = (struct gfs2_rgrp_lvb *)rgd->rd_gl->gl_lksb.sb_lvbptr; |
0e27c18c | 935 | rgd->rd_flags &= ~(GFS2_RDF_UPTODATE | GFS2_RDF_PREFERRED); |
7c9ca621 BP |
936 | if (rgd->rd_data > sdp->sd_max_rg_data) |
937 | sdp->sd_max_rg_data = rgd->rd_data; | |
8339ee54 | 938 | spin_lock(&sdp->sd_rindex_spin); |
6aad1c3d | 939 | error = rgd_insert(rgd); |
8339ee54 | 940 | spin_unlock(&sdp->sd_rindex_spin); |
36e4ad03 | 941 | if (!error) { |
6f6597ba | 942 | glock_set_object(rgd->rd_gl, rgd); |
36e4ad03 BP |
943 | rgd->rd_gl->gl_vm.start = (rgd->rd_addr * bsize) & PAGE_MASK; |
944 | rgd->rd_gl->gl_vm.end = PAGE_ALIGN((rgd->rd_addr + | |
945 | rgd->rd_length) * bsize) - 1; | |
6aad1c3d | 946 | return 0; |
36e4ad03 | 947 | } |
6aad1c3d BP |
948 | |
949 | error = 0; /* someone else read in the rgrp; free it and ignore it */ | |
c1ac539e | 950 | gfs2_glock_put(rgd->rd_gl); |
8339ee54 SW |
951 | |
952 | fail: | |
953 | kfree(rgd->rd_bits); | |
36e4ad03 | 954 | rgd->rd_bits = NULL; |
8339ee54 | 955 | kmem_cache_free(gfs2_rgrpd_cachep, rgd); |
6c53267f RP |
956 | return error; |
957 | } | |
958 | ||
0e27c18c BP |
959 | /** |
960 | * set_rgrp_preferences - Run all the rgrps, selecting some we prefer to use | |
961 | * @sdp: the GFS2 superblock | |
962 | * | |
963 | * The purpose of this function is to select a subset of the resource groups | |
964 | * and mark them as PREFERRED. We do it in such a way that each node prefers | |
965 | * to use a unique set of rgrps to minimize glock contention. | |
966 | */ | |
967 | static void set_rgrp_preferences(struct gfs2_sbd *sdp) | |
968 | { | |
969 | struct gfs2_rgrpd *rgd, *first; | |
970 | int i; | |
971 | ||
972 | /* Skip an initial number of rgrps, based on this node's journal ID. | |
973 | That should start each node out on its own set. */ | |
974 | rgd = gfs2_rgrpd_get_first(sdp); | |
975 | for (i = 0; i < sdp->sd_lockstruct.ls_jid; i++) | |
976 | rgd = gfs2_rgrpd_get_next(rgd); | |
977 | first = rgd; | |
978 | ||
979 | do { | |
980 | rgd->rd_flags |= GFS2_RDF_PREFERRED; | |
981 | for (i = 0; i < sdp->sd_journals; i++) { | |
982 | rgd = gfs2_rgrpd_get_next(rgd); | |
959b6717 | 983 | if (!rgd || rgd == first) |
0e27c18c BP |
984 | break; |
985 | } | |
959b6717 | 986 | } while (rgd && rgd != first); |
0e27c18c BP |
987 | } |
988 | ||
6c53267f RP |
989 | /** |
990 | * gfs2_ri_update - Pull in a new resource index from the disk | |
991 | * @ip: pointer to the rindex inode | |
992 | * | |
b3b94faa DT |
993 | * Returns: 0 on successful update, error code otherwise |
994 | */ | |
995 | ||
8339ee54 | 996 | static int gfs2_ri_update(struct gfs2_inode *ip) |
b3b94faa | 997 | { |
feaa7bba | 998 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
999 | int error; |
1000 | ||
8339ee54 | 1001 | do { |
4306629e | 1002 | error = read_rindex_entry(ip); |
8339ee54 SW |
1003 | } while (error == 0); |
1004 | ||
1005 | if (error < 0) | |
1006 | return error; | |
b3b94faa | 1007 | |
0e27c18c BP |
1008 | set_rgrp_preferences(sdp); |
1009 | ||
cf45b752 | 1010 | sdp->sd_rindex_uptodate = 1; |
6c53267f RP |
1011 | return 0; |
1012 | } | |
b3b94faa | 1013 | |
b3b94faa | 1014 | /** |
8339ee54 | 1015 | * gfs2_rindex_update - Update the rindex if required |
b3b94faa | 1016 | * @sdp: The GFS2 superblock |
b3b94faa DT |
1017 | * |
1018 | * We grab a lock on the rindex inode to make sure that it doesn't | |
1019 | * change whilst we are performing an operation. We keep this lock | |
1020 | * for quite long periods of time compared to other locks. This | |
1021 | * doesn't matter, since it is shared and it is very, very rarely | |
1022 | * accessed in the exclusive mode (i.e. only when expanding the filesystem). | |
1023 | * | |
1024 | * This makes sure that we're using the latest copy of the resource index | |
1025 | * special file, which might have been updated if someone expanded the | |
1026 | * filesystem (via gfs2_grow utility), which adds new resource groups. | |
1027 | * | |
8339ee54 | 1028 | * Returns: 0 on succeess, error code otherwise |
b3b94faa DT |
1029 | */ |
1030 | ||
8339ee54 | 1031 | int gfs2_rindex_update(struct gfs2_sbd *sdp) |
b3b94faa | 1032 | { |
feaa7bba | 1033 | struct gfs2_inode *ip = GFS2_I(sdp->sd_rindex); |
b3b94faa | 1034 | struct gfs2_glock *gl = ip->i_gl; |
8339ee54 SW |
1035 | struct gfs2_holder ri_gh; |
1036 | int error = 0; | |
a365fbf3 | 1037 | int unlock_required = 0; |
b3b94faa DT |
1038 | |
1039 | /* Read new copy from disk if we don't have the latest */ | |
cf45b752 | 1040 | if (!sdp->sd_rindex_uptodate) { |
a365fbf3 SW |
1041 | if (!gfs2_glock_is_locked_by_me(gl)) { |
1042 | error = gfs2_glock_nq_init(gl, LM_ST_SHARED, 0, &ri_gh); | |
1043 | if (error) | |
6aad1c3d | 1044 | return error; |
a365fbf3 SW |
1045 | unlock_required = 1; |
1046 | } | |
8339ee54 | 1047 | if (!sdp->sd_rindex_uptodate) |
b3b94faa | 1048 | error = gfs2_ri_update(ip); |
a365fbf3 SW |
1049 | if (unlock_required) |
1050 | gfs2_glock_dq_uninit(&ri_gh); | |
b3b94faa DT |
1051 | } |
1052 | ||
1053 | return error; | |
1054 | } | |
1055 | ||
42d52e38 | 1056 | static void gfs2_rgrp_in(struct gfs2_rgrpd *rgd, const void *buf) |
bb8d8a6f SW |
1057 | { |
1058 | const struct gfs2_rgrp *str = buf; | |
42d52e38 | 1059 | u32 rg_flags; |
bb8d8a6f | 1060 | |
42d52e38 | 1061 | rg_flags = be32_to_cpu(str->rg_flags); |
09010978 | 1062 | rg_flags &= ~GFS2_RDF_MASK; |
1ce97e56 SW |
1063 | rgd->rd_flags &= GFS2_RDF_MASK; |
1064 | rgd->rd_flags |= rg_flags; | |
cfc8b549 | 1065 | rgd->rd_free = be32_to_cpu(str->rg_free); |
73f74948 | 1066 | rgd->rd_dinodes = be32_to_cpu(str->rg_dinodes); |
d8b71f73 | 1067 | rgd->rd_igeneration = be64_to_cpu(str->rg_igeneration); |
166725d9 | 1068 | /* rd_data0, rd_data and rd_bitbytes already set from rindex */ |
bb8d8a6f SW |
1069 | } |
1070 | ||
3f30f929 BP |
1071 | static void gfs2_rgrp_ondisk2lvb(struct gfs2_rgrp_lvb *rgl, const void *buf) |
1072 | { | |
1073 | const struct gfs2_rgrp *str = buf; | |
1074 | ||
1075 | rgl->rl_magic = cpu_to_be32(GFS2_MAGIC); | |
1076 | rgl->rl_flags = str->rg_flags; | |
1077 | rgl->rl_free = str->rg_free; | |
1078 | rgl->rl_dinodes = str->rg_dinodes; | |
1079 | rgl->rl_igeneration = str->rg_igeneration; | |
1080 | rgl->__pad = 0UL; | |
1081 | } | |
1082 | ||
42d52e38 | 1083 | static void gfs2_rgrp_out(struct gfs2_rgrpd *rgd, void *buf) |
bb8d8a6f | 1084 | { |
65adc273 | 1085 | struct gfs2_rgrpd *next = gfs2_rgrpd_get_next(rgd); |
bb8d8a6f | 1086 | struct gfs2_rgrp *str = buf; |
850d2d91 | 1087 | u32 crc; |
bb8d8a6f | 1088 | |
09010978 | 1089 | str->rg_flags = cpu_to_be32(rgd->rd_flags & ~GFS2_RDF_MASK); |
cfc8b549 | 1090 | str->rg_free = cpu_to_be32(rgd->rd_free); |
73f74948 | 1091 | str->rg_dinodes = cpu_to_be32(rgd->rd_dinodes); |
65adc273 AP |
1092 | if (next == NULL) |
1093 | str->rg_skip = 0; | |
1094 | else if (next->rd_addr > rgd->rd_addr) | |
1095 | str->rg_skip = cpu_to_be32(next->rd_addr - rgd->rd_addr); | |
d8b71f73 | 1096 | str->rg_igeneration = cpu_to_be64(rgd->rd_igeneration); |
166725d9 AP |
1097 | str->rg_data0 = cpu_to_be64(rgd->rd_data0); |
1098 | str->rg_data = cpu_to_be32(rgd->rd_data); | |
1099 | str->rg_bitbytes = cpu_to_be32(rgd->rd_bitbytes); | |
850d2d91 AP |
1100 | str->rg_crc = 0; |
1101 | crc = gfs2_disk_hash(buf, sizeof(struct gfs2_rgrp)); | |
1102 | str->rg_crc = cpu_to_be32(crc); | |
166725d9 | 1103 | |
bb8d8a6f | 1104 | memset(&str->rg_reserved, 0, sizeof(str->rg_reserved)); |
3f30f929 | 1105 | gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, buf); |
bb8d8a6f SW |
1106 | } |
1107 | ||
90306c41 BM |
1108 | static int gfs2_rgrp_lvb_valid(struct gfs2_rgrpd *rgd) |
1109 | { | |
1110 | struct gfs2_rgrp_lvb *rgl = rgd->rd_rgl; | |
1111 | struct gfs2_rgrp *str = (struct gfs2_rgrp *)rgd->rd_bits[0].bi_bh->b_data; | |
72244b6b | 1112 | int valid = 1; |
90306c41 | 1113 | |
72244b6b BP |
1114 | if (rgl->rl_flags != str->rg_flags) { |
1115 | printk(KERN_WARNING "GFS2: rgd: %llu lvb flag mismatch %u/%u", | |
1116 | (unsigned long long)rgd->rd_addr, | |
1117 | be32_to_cpu(rgl->rl_flags), be32_to_cpu(str->rg_flags)); | |
1118 | valid = 0; | |
1119 | } | |
1120 | if (rgl->rl_free != str->rg_free) { | |
1121 | printk(KERN_WARNING "GFS2: rgd: %llu lvb free mismatch %u/%u", | |
1122 | (unsigned long long)rgd->rd_addr, | |
1123 | be32_to_cpu(rgl->rl_free), be32_to_cpu(str->rg_free)); | |
1124 | valid = 0; | |
1125 | } | |
1126 | if (rgl->rl_dinodes != str->rg_dinodes) { | |
1127 | printk(KERN_WARNING "GFS2: rgd: %llu lvb dinode mismatch %u/%u", | |
1128 | (unsigned long long)rgd->rd_addr, | |
1129 | be32_to_cpu(rgl->rl_dinodes), | |
1130 | be32_to_cpu(str->rg_dinodes)); | |
1131 | valid = 0; | |
1132 | } | |
1133 | if (rgl->rl_igeneration != str->rg_igeneration) { | |
1134 | printk(KERN_WARNING "GFS2: rgd: %llu lvb igen mismatch " | |
1135 | "%llu/%llu", (unsigned long long)rgd->rd_addr, | |
1136 | (unsigned long long)be64_to_cpu(rgl->rl_igeneration), | |
1137 | (unsigned long long)be64_to_cpu(str->rg_igeneration)); | |
1138 | valid = 0; | |
1139 | } | |
1140 | return valid; | |
90306c41 BM |
1141 | } |
1142 | ||
90306c41 BM |
1143 | static u32 count_unlinked(struct gfs2_rgrpd *rgd) |
1144 | { | |
1145 | struct gfs2_bitmap *bi; | |
1146 | const u32 length = rgd->rd_length; | |
1147 | const u8 *buffer = NULL; | |
1148 | u32 i, goal, count = 0; | |
1149 | ||
1150 | for (i = 0, bi = rgd->rd_bits; i < length; i++, bi++) { | |
1151 | goal = 0; | |
1152 | buffer = bi->bi_bh->b_data + bi->bi_offset; | |
1153 | WARN_ON(!buffer_uptodate(bi->bi_bh)); | |
1154 | while (goal < bi->bi_len * GFS2_NBBY) { | |
1155 | goal = gfs2_bitfit(buffer, bi->bi_len, goal, | |
1156 | GFS2_BLKST_UNLINKED); | |
1157 | if (goal == BFITNOENT) | |
1158 | break; | |
1159 | count++; | |
1160 | goal++; | |
1161 | } | |
1162 | } | |
1163 | ||
1164 | return count; | |
1165 | } | |
1166 | ||
1167 | ||
b3b94faa | 1168 | /** |
90306c41 BM |
1169 | * gfs2_rgrp_bh_get - Read in a RG's header and bitmaps |
1170 | * @rgd: the struct gfs2_rgrpd describing the RG to read in | |
b3b94faa DT |
1171 | * |
1172 | * Read in all of a Resource Group's header and bitmap blocks. | |
1173 | * Caller must eventually call gfs2_rgrp_relse() to free the bitmaps. | |
1174 | * | |
1175 | * Returns: errno | |
1176 | */ | |
1177 | ||
c2b0b30e | 1178 | static int gfs2_rgrp_bh_get(struct gfs2_rgrpd *rgd) |
b3b94faa DT |
1179 | { |
1180 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
1181 | struct gfs2_glock *gl = rgd->rd_gl; | |
bb8d8a6f | 1182 | unsigned int length = rgd->rd_length; |
b3b94faa DT |
1183 | struct gfs2_bitmap *bi; |
1184 | unsigned int x, y; | |
1185 | int error; | |
1186 | ||
90306c41 BM |
1187 | if (rgd->rd_bits[0].bi_bh != NULL) |
1188 | return 0; | |
1189 | ||
b3b94faa DT |
1190 | for (x = 0; x < length; x++) { |
1191 | bi = rgd->rd_bits + x; | |
c8d57703 | 1192 | error = gfs2_meta_read(gl, rgd->rd_addr + x, 0, 0, &bi->bi_bh); |
b3b94faa DT |
1193 | if (error) |
1194 | goto fail; | |
1195 | } | |
1196 | ||
1197 | for (y = length; y--;) { | |
1198 | bi = rgd->rd_bits + y; | |
7276b3b0 | 1199 | error = gfs2_meta_wait(sdp, bi->bi_bh); |
b3b94faa DT |
1200 | if (error) |
1201 | goto fail; | |
feaa7bba | 1202 | if (gfs2_metatype_check(sdp, bi->bi_bh, y ? GFS2_METATYPE_RB : |
b3b94faa DT |
1203 | GFS2_METATYPE_RG)) { |
1204 | error = -EIO; | |
1205 | goto fail; | |
1206 | } | |
1207 | } | |
1208 | ||
cf45b752 | 1209 | if (!(rgd->rd_flags & GFS2_RDF_UPTODATE)) { |
60a0b8f9 SW |
1210 | for (x = 0; x < length; x++) |
1211 | clear_bit(GBF_FULL, &rgd->rd_bits[x].bi_flags); | |
42d52e38 | 1212 | gfs2_rgrp_in(rgd, (rgd->rd_bits[0].bi_bh)->b_data); |
1ce97e56 | 1213 | rgd->rd_flags |= (GFS2_RDF_UPTODATE | GFS2_RDF_CHECK); |
7c9ca621 | 1214 | rgd->rd_free_clone = rgd->rd_free; |
5ea5050c BP |
1215 | /* max out the rgrp allocation failure point */ |
1216 | rgd->rd_extfail_pt = rgd->rd_free; | |
b3b94faa | 1217 | } |
951b4bd5 | 1218 | if (cpu_to_be32(GFS2_MAGIC) != rgd->rd_rgl->rl_magic) { |
90306c41 BM |
1219 | rgd->rd_rgl->rl_unlinked = cpu_to_be32(count_unlinked(rgd)); |
1220 | gfs2_rgrp_ondisk2lvb(rgd->rd_rgl, | |
1221 | rgd->rd_bits[0].bi_bh->b_data); | |
1222 | } | |
1223 | else if (sdp->sd_args.ar_rgrplvb) { | |
1224 | if (!gfs2_rgrp_lvb_valid(rgd)){ | |
1225 | gfs2_consist_rgrpd(rgd); | |
1226 | error = -EIO; | |
1227 | goto fail; | |
1228 | } | |
1229 | if (rgd->rd_rgl->rl_unlinked == 0) | |
1230 | rgd->rd_flags &= ~GFS2_RDF_CHECK; | |
1231 | } | |
b3b94faa DT |
1232 | return 0; |
1233 | ||
feaa7bba | 1234 | fail: |
b3b94faa DT |
1235 | while (x--) { |
1236 | bi = rgd->rd_bits + x; | |
1237 | brelse(bi->bi_bh); | |
1238 | bi->bi_bh = NULL; | |
1239 | gfs2_assert_warn(sdp, !bi->bi_clone); | |
1240 | } | |
b3b94faa DT |
1241 | |
1242 | return error; | |
1243 | } | |
1244 | ||
c2b0b30e | 1245 | static int update_rgrp_lvb(struct gfs2_rgrpd *rgd) |
90306c41 BM |
1246 | { |
1247 | u32 rl_flags; | |
1248 | ||
1249 | if (rgd->rd_flags & GFS2_RDF_UPTODATE) | |
1250 | return 0; | |
1251 | ||
951b4bd5 | 1252 | if (cpu_to_be32(GFS2_MAGIC) != rgd->rd_rgl->rl_magic) |
90306c41 BM |
1253 | return gfs2_rgrp_bh_get(rgd); |
1254 | ||
1255 | rl_flags = be32_to_cpu(rgd->rd_rgl->rl_flags); | |
1256 | rl_flags &= ~GFS2_RDF_MASK; | |
1257 | rgd->rd_flags &= GFS2_RDF_MASK; | |
4f36cb36 | 1258 | rgd->rd_flags |= (rl_flags | GFS2_RDF_CHECK); |
90306c41 BM |
1259 | if (rgd->rd_rgl->rl_unlinked == 0) |
1260 | rgd->rd_flags &= ~GFS2_RDF_CHECK; | |
1261 | rgd->rd_free = be32_to_cpu(rgd->rd_rgl->rl_free); | |
1262 | rgd->rd_free_clone = rgd->rd_free; | |
1263 | rgd->rd_dinodes = be32_to_cpu(rgd->rd_rgl->rl_dinodes); | |
1264 | rgd->rd_igeneration = be64_to_cpu(rgd->rd_rgl->rl_igeneration); | |
1265 | return 0; | |
1266 | } | |
1267 | ||
1268 | int gfs2_rgrp_go_lock(struct gfs2_holder *gh) | |
1269 | { | |
1270 | struct gfs2_rgrpd *rgd = gh->gh_gl->gl_object; | |
1271 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
1272 | ||
1273 | if (gh->gh_flags & GL_SKIP && sdp->sd_args.ar_rgrplvb) | |
1274 | return 0; | |
8b127d04 | 1275 | return gfs2_rgrp_bh_get(rgd); |
90306c41 BM |
1276 | } |
1277 | ||
b3b94faa | 1278 | /** |
39b0f1e9 BP |
1279 | * gfs2_rgrp_brelse - Release RG bitmaps read in with gfs2_rgrp_bh_get() |
1280 | * @rgd: The resource group | |
b3b94faa DT |
1281 | * |
1282 | */ | |
1283 | ||
39b0f1e9 | 1284 | void gfs2_rgrp_brelse(struct gfs2_rgrpd *rgd) |
b3b94faa | 1285 | { |
bb8d8a6f | 1286 | int x, length = rgd->rd_length; |
b3b94faa | 1287 | |
b3b94faa DT |
1288 | for (x = 0; x < length; x++) { |
1289 | struct gfs2_bitmap *bi = rgd->rd_bits + x; | |
90306c41 BM |
1290 | if (bi->bi_bh) { |
1291 | brelse(bi->bi_bh); | |
1292 | bi->bi_bh = NULL; | |
1293 | } | |
b3b94faa DT |
1294 | } |
1295 | ||
b3b94faa DT |
1296 | } |
1297 | ||
39b0f1e9 BP |
1298 | /** |
1299 | * gfs2_rgrp_go_unlock - Unlock a rgrp glock | |
1300 | * @gh: The glock holder for the resource group | |
1301 | * | |
1302 | */ | |
1303 | ||
1304 | void gfs2_rgrp_go_unlock(struct gfs2_holder *gh) | |
1305 | { | |
1306 | struct gfs2_rgrpd *rgd = gh->gh_gl->gl_object; | |
1307 | int demote_requested = test_bit(GLF_DEMOTE, &gh->gh_gl->gl_flags) | | |
1308 | test_bit(GLF_PENDING_DEMOTE, &gh->gh_gl->gl_flags); | |
1309 | ||
1310 | if (rgd && demote_requested) | |
1311 | gfs2_rgrp_brelse(rgd); | |
1312 | } | |
1313 | ||
66fc061b | 1314 | int gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset, |
7c9ca621 | 1315 | struct buffer_head *bh, |
66fc061b | 1316 | const struct gfs2_bitmap *bi, unsigned minlen, u64 *ptrimmed) |
f15ab561 SW |
1317 | { |
1318 | struct super_block *sb = sdp->sd_vfs; | |
f15ab561 | 1319 | u64 blk; |
64d576ba | 1320 | sector_t start = 0; |
b2c87cae | 1321 | sector_t nr_blks = 0; |
f15ab561 SW |
1322 | int rv; |
1323 | unsigned int x; | |
66fc061b SW |
1324 | u32 trimmed = 0; |
1325 | u8 diff; | |
f15ab561 SW |
1326 | |
1327 | for (x = 0; x < bi->bi_len; x++) { | |
66fc061b SW |
1328 | const u8 *clone = bi->bi_clone ? bi->bi_clone : bi->bi_bh->b_data; |
1329 | clone += bi->bi_offset; | |
1330 | clone += x; | |
1331 | if (bh) { | |
1332 | const u8 *orig = bh->b_data + bi->bi_offset + x; | |
1333 | diff = ~(*orig | (*orig >> 1)) & (*clone | (*clone >> 1)); | |
1334 | } else { | |
1335 | diff = ~(*clone | (*clone >> 1)); | |
1336 | } | |
f15ab561 SW |
1337 | diff &= 0x55; |
1338 | if (diff == 0) | |
1339 | continue; | |
1340 | blk = offset + ((bi->bi_start + x) * GFS2_NBBY); | |
f15ab561 SW |
1341 | while(diff) { |
1342 | if (diff & 1) { | |
b2c87cae | 1343 | if (nr_blks == 0) |
f15ab561 | 1344 | goto start_new_extent; |
b2c87cae BP |
1345 | if ((start + nr_blks) != blk) { |
1346 | if (nr_blks >= minlen) { | |
1347 | rv = sb_issue_discard(sb, | |
1348 | start, nr_blks, | |
66fc061b SW |
1349 | GFP_NOFS, 0); |
1350 | if (rv) | |
1351 | goto fail; | |
b2c87cae | 1352 | trimmed += nr_blks; |
66fc061b | 1353 | } |
b2c87cae | 1354 | nr_blks = 0; |
f15ab561 SW |
1355 | start_new_extent: |
1356 | start = blk; | |
1357 | } | |
b2c87cae | 1358 | nr_blks++; |
f15ab561 SW |
1359 | } |
1360 | diff >>= 2; | |
b2c87cae | 1361 | blk++; |
f15ab561 SW |
1362 | } |
1363 | } | |
b2c87cae BP |
1364 | if (nr_blks >= minlen) { |
1365 | rv = sb_issue_discard(sb, start, nr_blks, GFP_NOFS, 0); | |
f15ab561 SW |
1366 | if (rv) |
1367 | goto fail; | |
b2c87cae | 1368 | trimmed += nr_blks; |
f15ab561 | 1369 | } |
66fc061b SW |
1370 | if (ptrimmed) |
1371 | *ptrimmed = trimmed; | |
1372 | return 0; | |
1373 | ||
f15ab561 | 1374 | fail: |
66fc061b | 1375 | if (sdp->sd_args.ar_discard) |
af38816e | 1376 | fs_warn(sdp, "error %d on discard request, turning discards off for this filesystem\n", rv); |
f15ab561 | 1377 | sdp->sd_args.ar_discard = 0; |
66fc061b SW |
1378 | return -EIO; |
1379 | } | |
1380 | ||
1381 | /** | |
1382 | * gfs2_fitrim - Generate discard requests for unused bits of the filesystem | |
1383 | * @filp: Any file on the filesystem | |
1384 | * @argp: Pointer to the arguments (also used to pass result) | |
1385 | * | |
1386 | * Returns: 0 on success, otherwise error code | |
1387 | */ | |
1388 | ||
1389 | int gfs2_fitrim(struct file *filp, void __user *argp) | |
1390 | { | |
496ad9aa | 1391 | struct inode *inode = file_inode(filp); |
66fc061b SW |
1392 | struct gfs2_sbd *sdp = GFS2_SB(inode); |
1393 | struct request_queue *q = bdev_get_queue(sdp->sd_vfs->s_bdev); | |
1394 | struct buffer_head *bh; | |
1395 | struct gfs2_rgrpd *rgd; | |
1396 | struct gfs2_rgrpd *rgd_end; | |
1397 | struct gfs2_holder gh; | |
1398 | struct fstrim_range r; | |
1399 | int ret = 0; | |
1400 | u64 amt; | |
1401 | u64 trimmed = 0; | |
076f0faa | 1402 | u64 start, end, minlen; |
66fc061b | 1403 | unsigned int x; |
076f0faa | 1404 | unsigned bs_shift = sdp->sd_sb.sb_bsize_shift; |
66fc061b SW |
1405 | |
1406 | if (!capable(CAP_SYS_ADMIN)) | |
1407 | return -EPERM; | |
1408 | ||
1409 | if (!blk_queue_discard(q)) | |
1410 | return -EOPNOTSUPP; | |
1411 | ||
3a238ade | 1412 | if (copy_from_user(&r, argp, sizeof(r))) |
66fc061b SW |
1413 | return -EFAULT; |
1414 | ||
5e2f7d61 BP |
1415 | ret = gfs2_rindex_update(sdp); |
1416 | if (ret) | |
1417 | return ret; | |
1418 | ||
076f0faa LC |
1419 | start = r.start >> bs_shift; |
1420 | end = start + (r.len >> bs_shift); | |
1421 | minlen = max_t(u64, r.minlen, | |
1422 | q->limits.discard_granularity) >> bs_shift; | |
1423 | ||
6a98c333 AD |
1424 | if (end <= start || minlen > sdp->sd_max_rg_data) |
1425 | return -EINVAL; | |
1426 | ||
076f0faa | 1427 | rgd = gfs2_blk2rgrpd(sdp, start, 0); |
6a98c333 | 1428 | rgd_end = gfs2_blk2rgrpd(sdp, end, 0); |
076f0faa | 1429 | |
6a98c333 AD |
1430 | if ((gfs2_rgrpd_get_first(sdp) == gfs2_rgrpd_get_next(rgd_end)) |
1431 | && (start > rgd_end->rd_data0 + rgd_end->rd_data)) | |
1432 | return -EINVAL; /* start is beyond the end of the fs */ | |
66fc061b SW |
1433 | |
1434 | while (1) { | |
1435 | ||
1436 | ret = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
1437 | if (ret) | |
1438 | goto out; | |
1439 | ||
1440 | if (!(rgd->rd_flags & GFS2_RGF_TRIMMED)) { | |
1441 | /* Trim each bitmap in the rgrp */ | |
1442 | for (x = 0; x < rgd->rd_length; x++) { | |
1443 | struct gfs2_bitmap *bi = rgd->rd_bits + x; | |
076f0faa LC |
1444 | ret = gfs2_rgrp_send_discards(sdp, |
1445 | rgd->rd_data0, NULL, bi, minlen, | |
1446 | &amt); | |
66fc061b SW |
1447 | if (ret) { |
1448 | gfs2_glock_dq_uninit(&gh); | |
1449 | goto out; | |
1450 | } | |
1451 | trimmed += amt; | |
1452 | } | |
1453 | ||
1454 | /* Mark rgrp as having been trimmed */ | |
1455 | ret = gfs2_trans_begin(sdp, RES_RG_HDR, 0); | |
1456 | if (ret == 0) { | |
1457 | bh = rgd->rd_bits[0].bi_bh; | |
1458 | rgd->rd_flags |= GFS2_RGF_TRIMMED; | |
350a9b0a | 1459 | gfs2_trans_add_meta(rgd->rd_gl, bh); |
66fc061b SW |
1460 | gfs2_rgrp_out(rgd, bh->b_data); |
1461 | gfs2_trans_end(sdp); | |
1462 | } | |
1463 | } | |
1464 | gfs2_glock_dq_uninit(&gh); | |
1465 | ||
1466 | if (rgd == rgd_end) | |
1467 | break; | |
1468 | ||
1469 | rgd = gfs2_rgrpd_get_next(rgd); | |
1470 | } | |
1471 | ||
1472 | out: | |
6a98c333 | 1473 | r.len = trimmed << bs_shift; |
3a238ade | 1474 | if (copy_to_user(argp, &r, sizeof(r))) |
66fc061b SW |
1475 | return -EFAULT; |
1476 | ||
1477 | return ret; | |
f15ab561 SW |
1478 | } |
1479 | ||
8e2e0047 BP |
1480 | /** |
1481 | * rs_insert - insert a new multi-block reservation into the rgrp's rb_tree | |
8e2e0047 | 1482 | * @ip: the inode structure |
8e2e0047 | 1483 | * |
8e2e0047 | 1484 | */ |
ff7f4cb4 | 1485 | static void rs_insert(struct gfs2_inode *ip) |
8e2e0047 BP |
1486 | { |
1487 | struct rb_node **newn, *parent = NULL; | |
1488 | int rc; | |
a097dc7e | 1489 | struct gfs2_blkreserv *rs = &ip->i_res; |
4a993fb1 | 1490 | struct gfs2_rgrpd *rgd = rs->rs_rbm.rgd; |
ff7f4cb4 | 1491 | u64 fsblock = gfs2_rbm_to_block(&rs->rs_rbm); |
8e2e0047 | 1492 | |
8e2e0047 | 1493 | BUG_ON(gfs2_rs_active(rs)); |
c743ffd0 | 1494 | |
ff7f4cb4 SW |
1495 | spin_lock(&rgd->rd_rsspin); |
1496 | newn = &rgd->rd_rstree.rb_node; | |
8e2e0047 BP |
1497 | while (*newn) { |
1498 | struct gfs2_blkreserv *cur = | |
1499 | rb_entry(*newn, struct gfs2_blkreserv, rs_node); | |
1500 | ||
1501 | parent = *newn; | |
ff7f4cb4 | 1502 | rc = rs_cmp(fsblock, rs->rs_free, cur); |
8e2e0047 BP |
1503 | if (rc > 0) |
1504 | newn = &((*newn)->rb_right); | |
1505 | else if (rc < 0) | |
1506 | newn = &((*newn)->rb_left); | |
1507 | else { | |
1508 | spin_unlock(&rgd->rd_rsspin); | |
ff7f4cb4 SW |
1509 | WARN_ON(1); |
1510 | return; | |
8e2e0047 BP |
1511 | } |
1512 | } | |
1513 | ||
8e2e0047 BP |
1514 | rb_link_node(&rs->rs_node, parent, newn); |
1515 | rb_insert_color(&rs->rs_node, &rgd->rd_rstree); | |
1516 | ||
8e2e0047 | 1517 | /* Do our rgrp accounting for the reservation */ |
ff7f4cb4 | 1518 | rgd->rd_reserved += rs->rs_free; /* blocks reserved */ |
8e2e0047 | 1519 | spin_unlock(&rgd->rd_rsspin); |
9e733d39 | 1520 | trace_gfs2_rs(rs, TRACE_RS_INSERT); |
8e2e0047 BP |
1521 | } |
1522 | ||
f6753df3 BP |
1523 | /** |
1524 | * rgd_free - return the number of free blocks we can allocate. | |
1525 | * @rgd: the resource group | |
1526 | * | |
1527 | * This function returns the number of free blocks for an rgrp. | |
1528 | * That's the clone-free blocks (blocks that are free, not including those | |
1529 | * still being used for unlinked files that haven't been deleted.) | |
1530 | * | |
1531 | * It also subtracts any blocks reserved by someone else, but does not | |
1532 | * include free blocks that are still part of our current reservation, | |
1533 | * because obviously we can (and will) allocate them. | |
1534 | */ | |
1535 | static inline u32 rgd_free(struct gfs2_rgrpd *rgd, struct gfs2_blkreserv *rs) | |
1536 | { | |
1537 | u32 tot_reserved, tot_free; | |
1538 | ||
1539 | if (WARN_ON_ONCE(rgd->rd_reserved < rs->rs_free)) | |
1540 | return 0; | |
1541 | tot_reserved = rgd->rd_reserved - rs->rs_free; | |
1542 | ||
1543 | if (rgd->rd_free_clone < tot_reserved) | |
1544 | tot_reserved = 0; | |
1545 | ||
1546 | tot_free = rgd->rd_free_clone - tot_reserved; | |
1547 | ||
1548 | return tot_free; | |
1549 | } | |
1550 | ||
8e2e0047 | 1551 | /** |
ff7f4cb4 | 1552 | * rg_mblk_search - find a group of multiple free blocks to form a reservation |
8e2e0047 | 1553 | * @rgd: the resource group descriptor |
8e2e0047 | 1554 | * @ip: pointer to the inode for which we're reserving blocks |
7b9cff46 | 1555 | * @ap: the allocation parameters |
8e2e0047 | 1556 | * |
8e2e0047 BP |
1557 | */ |
1558 | ||
ff7f4cb4 | 1559 | static void rg_mblk_search(struct gfs2_rgrpd *rgd, struct gfs2_inode *ip, |
7b9cff46 | 1560 | const struct gfs2_alloc_parms *ap) |
8e2e0047 | 1561 | { |
ff7f4cb4 SW |
1562 | struct gfs2_rbm rbm = { .rgd = rgd, }; |
1563 | u64 goal; | |
a097dc7e | 1564 | struct gfs2_blkreserv *rs = &ip->i_res; |
ff7f4cb4 | 1565 | u32 extlen; |
f6753df3 | 1566 | u32 free_blocks = rgd_free(rgd, rs); |
ff7f4cb4 | 1567 | int ret; |
af21ca8e | 1568 | struct inode *inode = &ip->i_inode; |
8e2e0047 | 1569 | |
af21ca8e BP |
1570 | if (S_ISDIR(inode->i_mode)) |
1571 | extlen = 1; | |
1572 | else { | |
7b9cff46 | 1573 | extlen = max_t(u32, atomic_read(&rs->rs_sizehint), ap->target); |
af21ca8e BP |
1574 | extlen = clamp(extlen, RGRP_RSRV_MINBLKS, free_blocks); |
1575 | } | |
ff7f4cb4 | 1576 | if ((rgd->rd_free_clone < rgd->rd_reserved) || (free_blocks < extlen)) |
c743ffd0 SW |
1577 | return; |
1578 | ||
8e2e0047 BP |
1579 | /* Find bitmap block that contains bits for goal block */ |
1580 | if (rgrp_contains_block(rgd, ip->i_goal)) | |
ff7f4cb4 | 1581 | goal = ip->i_goal; |
8e2e0047 | 1582 | else |
ff7f4cb4 | 1583 | goal = rgd->rd_last_alloc + rgd->rd_data0; |
8e2e0047 | 1584 | |
ff7f4cb4 SW |
1585 | if (WARN_ON(gfs2_rbm_from_block(&rbm, goal))) |
1586 | return; | |
8e2e0047 | 1587 | |
8381e602 | 1588 | ret = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, &extlen, ip, true); |
ff7f4cb4 SW |
1589 | if (ret == 0) { |
1590 | rs->rs_rbm = rbm; | |
1591 | rs->rs_free = extlen; | |
ff7f4cb4 | 1592 | rs_insert(ip); |
13d2eb01 BP |
1593 | } else { |
1594 | if (goal == rgd->rd_last_alloc + rgd->rd_data0) | |
1595 | rgd->rd_last_alloc = 0; | |
8e2e0047 | 1596 | } |
b3e47ca0 BP |
1597 | } |
1598 | ||
5b924ae2 SW |
1599 | /** |
1600 | * gfs2_next_unreserved_block - Return next block that is not reserved | |
1601 | * @rgd: The resource group | |
1602 | * @block: The starting block | |
ff7f4cb4 | 1603 | * @length: The required length |
5b924ae2 SW |
1604 | * @ip: Ignore any reservations for this inode |
1605 | * | |
1606 | * If the block does not appear in any reservation, then return the | |
1607 | * block number unchanged. If it does appear in the reservation, then | |
1608 | * keep looking through the tree of reservations in order to find the | |
1609 | * first block number which is not reserved. | |
1610 | */ | |
1611 | ||
1612 | static u64 gfs2_next_unreserved_block(struct gfs2_rgrpd *rgd, u64 block, | |
ff7f4cb4 | 1613 | u32 length, |
5b924ae2 SW |
1614 | const struct gfs2_inode *ip) |
1615 | { | |
1616 | struct gfs2_blkreserv *rs; | |
1617 | struct rb_node *n; | |
1618 | int rc; | |
1619 | ||
1620 | spin_lock(&rgd->rd_rsspin); | |
ff7f4cb4 | 1621 | n = rgd->rd_rstree.rb_node; |
5b924ae2 SW |
1622 | while (n) { |
1623 | rs = rb_entry(n, struct gfs2_blkreserv, rs_node); | |
ff7f4cb4 | 1624 | rc = rs_cmp(block, length, rs); |
5b924ae2 SW |
1625 | if (rc < 0) |
1626 | n = n->rb_left; | |
1627 | else if (rc > 0) | |
1628 | n = n->rb_right; | |
1629 | else | |
1630 | break; | |
1631 | } | |
1632 | ||
1633 | if (n) { | |
a097dc7e | 1634 | while ((rs_cmp(block, length, rs) == 0) && (&ip->i_res != rs)) { |
5b924ae2 | 1635 | block = gfs2_rbm_to_block(&rs->rs_rbm) + rs->rs_free; |
ff7f4cb4 | 1636 | n = n->rb_right; |
5b924ae2 SW |
1637 | if (n == NULL) |
1638 | break; | |
1639 | rs = rb_entry(n, struct gfs2_blkreserv, rs_node); | |
1640 | } | |
1641 | } | |
1642 | ||
1643 | spin_unlock(&rgd->rd_rsspin); | |
1644 | return block; | |
1645 | } | |
1646 | ||
5b924ae2 SW |
1647 | /** |
1648 | * gfs2_reservation_check_and_update - Check for reservations during block alloc | |
1649 | * @rbm: The current position in the resource group | |
ff7f4cb4 SW |
1650 | * @ip: The inode for which we are searching for blocks |
1651 | * @minext: The minimum extent length | |
5ce13431 | 1652 | * @maxext: A pointer to the maximum extent structure |
5b924ae2 SW |
1653 | * |
1654 | * This checks the current position in the rgrp to see whether there is | |
1655 | * a reservation covering this block. If not then this function is a | |
1656 | * no-op. If there is, then the position is moved to the end of the | |
1657 | * contiguous reservation(s) so that we are pointing at the first | |
1658 | * non-reserved block. | |
1659 | * | |
1660 | * Returns: 0 if no reservation, 1 if @rbm has changed, otherwise an error | |
1661 | */ | |
1662 | ||
1663 | static int gfs2_reservation_check_and_update(struct gfs2_rbm *rbm, | |
ff7f4cb4 | 1664 | const struct gfs2_inode *ip, |
5ce13431 BP |
1665 | u32 minext, |
1666 | struct gfs2_extent *maxext) | |
5b924ae2 SW |
1667 | { |
1668 | u64 block = gfs2_rbm_to_block(rbm); | |
ff7f4cb4 | 1669 | u32 extlen = 1; |
5b924ae2 SW |
1670 | u64 nblock; |
1671 | int ret; | |
1672 | ||
ff7f4cb4 SW |
1673 | /* |
1674 | * If we have a minimum extent length, then skip over any extent | |
1675 | * which is less than the min extent length in size. | |
1676 | */ | |
1677 | if (minext) { | |
1678 | extlen = gfs2_free_extlen(rbm, minext); | |
5ce13431 | 1679 | if (extlen <= maxext->len) |
ff7f4cb4 SW |
1680 | goto fail; |
1681 | } | |
1682 | ||
1683 | /* | |
1684 | * Check the extent which has been found against the reservations | |
1685 | * and skip if parts of it are already reserved | |
1686 | */ | |
1687 | nblock = gfs2_next_unreserved_block(rbm->rgd, block, extlen, ip); | |
5ce13431 BP |
1688 | if (nblock == block) { |
1689 | if (!minext || extlen >= minext) | |
1690 | return 0; | |
1691 | ||
1692 | if (extlen > maxext->len) { | |
1693 | maxext->len = extlen; | |
1694 | maxext->rbm = *rbm; | |
1695 | } | |
ff7f4cb4 | 1696 | fail: |
5ce13431 BP |
1697 | nblock = block + extlen; |
1698 | } | |
5b924ae2 SW |
1699 | ret = gfs2_rbm_from_block(rbm, nblock); |
1700 | if (ret < 0) | |
1701 | return ret; | |
1702 | return 1; | |
1703 | } | |
1704 | ||
1705 | /** | |
1706 | * gfs2_rbm_find - Look for blocks of a particular state | |
1707 | * @rbm: Value/result starting position and final position | |
1708 | * @state: The state which we want to find | |
5ce13431 BP |
1709 | * @minext: Pointer to the requested extent length (NULL for a single block) |
1710 | * This is updated to be the actual reservation size. | |
5b924ae2 SW |
1711 | * @ip: If set, check for reservations |
1712 | * @nowrap: Stop looking at the end of the rgrp, rather than wrapping | |
1713 | * around until we've reached the starting point. | |
1714 | * | |
1715 | * Side effects: | |
1716 | * - If looking for free blocks, we set GBF_FULL on each bitmap which | |
1717 | * has no free blocks in it. | |
5ea5050c BP |
1718 | * - If looking for free blocks, we set rd_extfail_pt on each rgrp which |
1719 | * has come up short on a free block search. | |
5b924ae2 SW |
1720 | * |
1721 | * Returns: 0 on success, -ENOSPC if there is no block of the requested state | |
1722 | */ | |
1723 | ||
5ce13431 | 1724 | static int gfs2_rbm_find(struct gfs2_rbm *rbm, u8 state, u32 *minext, |
8381e602 | 1725 | const struct gfs2_inode *ip, bool nowrap) |
5b924ae2 SW |
1726 | { |
1727 | struct buffer_head *bh; | |
e579ed4f | 1728 | int initial_bii; |
5b924ae2 | 1729 | u32 initial_offset; |
5ea5050c BP |
1730 | int first_bii = rbm->bii; |
1731 | u32 first_offset = rbm->offset; | |
5b924ae2 SW |
1732 | u32 offset; |
1733 | u8 *buffer; | |
5b924ae2 SW |
1734 | int n = 0; |
1735 | int iters = rbm->rgd->rd_length; | |
1736 | int ret; | |
e579ed4f | 1737 | struct gfs2_bitmap *bi; |
5ce13431 | 1738 | struct gfs2_extent maxext = { .rbm.rgd = rbm->rgd, }; |
5b924ae2 SW |
1739 | |
1740 | /* If we are not starting at the beginning of a bitmap, then we | |
1741 | * need to add one to the bitmap count to ensure that we search | |
1742 | * the starting bitmap twice. | |
1743 | */ | |
1744 | if (rbm->offset != 0) | |
1745 | iters++; | |
1746 | ||
1747 | while(1) { | |
e579ed4f | 1748 | bi = rbm_bi(rbm); |
e79e0e14 BP |
1749 | if ((ip == NULL || !gfs2_rs_active(&ip->i_res)) && |
1750 | test_bit(GBF_FULL, &bi->bi_flags) && | |
5b924ae2 SW |
1751 | (state == GFS2_BLKST_FREE)) |
1752 | goto next_bitmap; | |
1753 | ||
e579ed4f BP |
1754 | bh = bi->bi_bh; |
1755 | buffer = bh->b_data + bi->bi_offset; | |
5b924ae2 | 1756 | WARN_ON(!buffer_uptodate(bh)); |
e579ed4f BP |
1757 | if (state != GFS2_BLKST_UNLINKED && bi->bi_clone) |
1758 | buffer = bi->bi_clone + bi->bi_offset; | |
5b924ae2 | 1759 | initial_offset = rbm->offset; |
e579ed4f | 1760 | offset = gfs2_bitfit(buffer, bi->bi_len, rbm->offset, state); |
5b924ae2 SW |
1761 | if (offset == BFITNOENT) |
1762 | goto bitmap_full; | |
1763 | rbm->offset = offset; | |
1764 | if (ip == NULL) | |
1765 | return 0; | |
1766 | ||
e579ed4f | 1767 | initial_bii = rbm->bii; |
5ce13431 BP |
1768 | ret = gfs2_reservation_check_and_update(rbm, ip, |
1769 | minext ? *minext : 0, | |
1770 | &maxext); | |
5b924ae2 SW |
1771 | if (ret == 0) |
1772 | return 0; | |
1773 | if (ret > 0) { | |
e579ed4f | 1774 | n += (rbm->bii - initial_bii); |
8d8b752a | 1775 | goto next_iter; |
5b924ae2 | 1776 | } |
5d50d532 | 1777 | if (ret == -E2BIG) { |
e579ed4f | 1778 | rbm->bii = 0; |
5d50d532 | 1779 | rbm->offset = 0; |
e579ed4f | 1780 | n += (rbm->bii - initial_bii); |
5d50d532 SW |
1781 | goto res_covered_end_of_rgrp; |
1782 | } | |
5b924ae2 SW |
1783 | return ret; |
1784 | ||
1785 | bitmap_full: /* Mark bitmap as full and fall through */ | |
a3e32136 | 1786 | if ((state == GFS2_BLKST_FREE) && initial_offset == 0) |
e579ed4f | 1787 | set_bit(GBF_FULL, &bi->bi_flags); |
5b924ae2 SW |
1788 | |
1789 | next_bitmap: /* Find next bitmap in the rgrp */ | |
1790 | rbm->offset = 0; | |
e579ed4f BP |
1791 | rbm->bii++; |
1792 | if (rbm->bii == rbm->rgd->rd_length) | |
1793 | rbm->bii = 0; | |
5d50d532 | 1794 | res_covered_end_of_rgrp: |
e579ed4f | 1795 | if ((rbm->bii == 0) && nowrap) |
5b924ae2 SW |
1796 | break; |
1797 | n++; | |
8d8b752a | 1798 | next_iter: |
5b924ae2 SW |
1799 | if (n >= iters) |
1800 | break; | |
1801 | } | |
1802 | ||
5ce13431 BP |
1803 | if (minext == NULL || state != GFS2_BLKST_FREE) |
1804 | return -ENOSPC; | |
1805 | ||
5ea5050c BP |
1806 | /* If the extent was too small, and it's smaller than the smallest |
1807 | to have failed before, remember for future reference that it's | |
1808 | useless to search this rgrp again for this amount or more. */ | |
1809 | if ((first_offset == 0) && (first_bii == 0) && | |
1810 | (*minext < rbm->rgd->rd_extfail_pt)) | |
1811 | rbm->rgd->rd_extfail_pt = *minext; | |
1812 | ||
5ce13431 BP |
1813 | /* If the maximum extent we found is big enough to fulfill the |
1814 | minimum requirements, use it anyway. */ | |
1815 | if (maxext.len) { | |
1816 | *rbm = maxext.rbm; | |
1817 | *minext = maxext.len; | |
1818 | return 0; | |
1819 | } | |
1820 | ||
5b924ae2 SW |
1821 | return -ENOSPC; |
1822 | } | |
1823 | ||
c8cdf479 SW |
1824 | /** |
1825 | * try_rgrp_unlink - Look for any unlinked, allocated, but unused inodes | |
1826 | * @rgd: The rgrp | |
886b1416 BP |
1827 | * @last_unlinked: block address of the last dinode we unlinked |
1828 | * @skip: block address we should explicitly not unlink | |
c8cdf479 | 1829 | * |
1a0eae88 BP |
1830 | * Returns: 0 if no error |
1831 | * The inode, if one has been found, in inode. | |
c8cdf479 SW |
1832 | */ |
1833 | ||
044b9414 | 1834 | static void try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked, u64 skip) |
c8cdf479 | 1835 | { |
5b924ae2 | 1836 | u64 block; |
5f3eae75 | 1837 | struct gfs2_sbd *sdp = rgd->rd_sbd; |
044b9414 SW |
1838 | struct gfs2_glock *gl; |
1839 | struct gfs2_inode *ip; | |
1840 | int error; | |
1841 | int found = 0; | |
e579ed4f | 1842 | struct gfs2_rbm rbm = { .rgd = rgd, .bii = 0, .offset = 0 }; |
c8cdf479 | 1843 | |
5b924ae2 | 1844 | while (1) { |
5f3eae75 | 1845 | down_write(&sdp->sd_log_flush_lock); |
5ce13431 | 1846 | error = gfs2_rbm_find(&rbm, GFS2_BLKST_UNLINKED, NULL, NULL, |
8381e602 | 1847 | true); |
5f3eae75 | 1848 | up_write(&sdp->sd_log_flush_lock); |
5b924ae2 SW |
1849 | if (error == -ENOSPC) |
1850 | break; | |
1851 | if (WARN_ON_ONCE(error)) | |
24c73873 | 1852 | break; |
b3e47ca0 | 1853 | |
5b924ae2 SW |
1854 | block = gfs2_rbm_to_block(&rbm); |
1855 | if (gfs2_rbm_from_block(&rbm, block + 1)) | |
1856 | break; | |
1857 | if (*last_unlinked != NO_BLOCK && block <= *last_unlinked) | |
c8cdf479 | 1858 | continue; |
5b924ae2 | 1859 | if (block == skip) |
1e19a195 | 1860 | continue; |
5b924ae2 | 1861 | *last_unlinked = block; |
044b9414 | 1862 | |
5ea31bc0 | 1863 | error = gfs2_glock_get(sdp, block, &gfs2_iopen_glops, CREATE, &gl); |
044b9414 SW |
1864 | if (error) |
1865 | continue; | |
1866 | ||
1867 | /* If the inode is already in cache, we can ignore it here | |
1868 | * because the existing inode disposal code will deal with | |
1869 | * it when all refs have gone away. Accessing gl_object like | |
1870 | * this is not safe in general. Here it is ok because we do | |
1871 | * not dereference the pointer, and we only need an approx | |
1872 | * answer to whether it is NULL or not. | |
1873 | */ | |
1874 | ip = gl->gl_object; | |
1875 | ||
1876 | if (ip || queue_work(gfs2_delete_workqueue, &gl->gl_delete) == 0) | |
1877 | gfs2_glock_put(gl); | |
1878 | else | |
1879 | found++; | |
1880 | ||
1881 | /* Limit reclaim to sensible number of tasks */ | |
44ad37d6 | 1882 | if (found > NR_CPUS) |
044b9414 | 1883 | return; |
c8cdf479 SW |
1884 | } |
1885 | ||
1886 | rgd->rd_flags &= ~GFS2_RDF_CHECK; | |
044b9414 | 1887 | return; |
c8cdf479 SW |
1888 | } |
1889 | ||
bcd97c06 SW |
1890 | /** |
1891 | * gfs2_rgrp_congested - Use stats to figure out whether an rgrp is congested | |
1892 | * @rgd: The rgrp in question | |
1893 | * @loops: An indication of how picky we can be (0=very, 1=less so) | |
1894 | * | |
1895 | * This function uses the recently added glock statistics in order to | |
1896 | * figure out whether a parciular resource group is suffering from | |
1897 | * contention from multiple nodes. This is done purely on the basis | |
1898 | * of timings, since this is the only data we have to work with and | |
1899 | * our aim here is to reject a resource group which is highly contended | |
1900 | * but (very important) not to do this too often in order to ensure that | |
1901 | * we do not land up introducing fragmentation by changing resource | |
1902 | * groups when not actually required. | |
1903 | * | |
1904 | * The calculation is fairly simple, we want to know whether the SRTTB | |
1905 | * (i.e. smoothed round trip time for blocking operations) to acquire | |
1906 | * the lock for this rgrp's glock is significantly greater than the | |
1907 | * time taken for resource groups on average. We introduce a margin in | |
1908 | * the form of the variable @var which is computed as the sum of the two | |
1909 | * respective variences, and multiplied by a factor depending on @loops | |
1910 | * and whether we have a lot of data to base the decision on. This is | |
1911 | * then tested against the square difference of the means in order to | |
1912 | * decide whether the result is statistically significant or not. | |
1913 | * | |
1914 | * Returns: A boolean verdict on the congestion status | |
1915 | */ | |
1916 | ||
1917 | static bool gfs2_rgrp_congested(const struct gfs2_rgrpd *rgd, int loops) | |
1918 | { | |
1919 | const struct gfs2_glock *gl = rgd->rd_gl; | |
15562c43 | 1920 | const struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; |
bcd97c06 | 1921 | struct gfs2_lkstats *st; |
4d207133 BH |
1922 | u64 r_dcount, l_dcount; |
1923 | u64 l_srttb, a_srttb = 0; | |
bcd97c06 | 1924 | s64 srttb_diff; |
4d207133 BH |
1925 | u64 sqr_diff; |
1926 | u64 var; | |
0166b197 | 1927 | int cpu, nonzero = 0; |
bcd97c06 SW |
1928 | |
1929 | preempt_disable(); | |
f4a3ae93 BP |
1930 | for_each_present_cpu(cpu) { |
1931 | st = &per_cpu_ptr(sdp->sd_lkstats, cpu)->lkstats[LM_TYPE_RGRP]; | |
0166b197 BP |
1932 | if (st->stats[GFS2_LKS_SRTTB]) { |
1933 | a_srttb += st->stats[GFS2_LKS_SRTTB]; | |
1934 | nonzero++; | |
1935 | } | |
f4a3ae93 | 1936 | } |
bcd97c06 | 1937 | st = &this_cpu_ptr(sdp->sd_lkstats)->lkstats[LM_TYPE_RGRP]; |
0166b197 BP |
1938 | if (nonzero) |
1939 | do_div(a_srttb, nonzero); | |
bcd97c06 SW |
1940 | r_dcount = st->stats[GFS2_LKS_DCOUNT]; |
1941 | var = st->stats[GFS2_LKS_SRTTVARB] + | |
1942 | gl->gl_stats.stats[GFS2_LKS_SRTTVARB]; | |
1943 | preempt_enable(); | |
1944 | ||
1945 | l_srttb = gl->gl_stats.stats[GFS2_LKS_SRTTB]; | |
1946 | l_dcount = gl->gl_stats.stats[GFS2_LKS_DCOUNT]; | |
1947 | ||
f4a3ae93 | 1948 | if ((l_dcount < 1) || (r_dcount < 1) || (a_srttb == 0)) |
bcd97c06 SW |
1949 | return false; |
1950 | ||
f4a3ae93 | 1951 | srttb_diff = a_srttb - l_srttb; |
bcd97c06 SW |
1952 | sqr_diff = srttb_diff * srttb_diff; |
1953 | ||
1954 | var *= 2; | |
1955 | if (l_dcount < 8 || r_dcount < 8) | |
1956 | var *= 2; | |
1957 | if (loops == 1) | |
1958 | var *= 2; | |
1959 | ||
1960 | return ((srttb_diff < 0) && (sqr_diff > var)); | |
1961 | } | |
1962 | ||
1963 | /** | |
1964 | * gfs2_rgrp_used_recently | |
1965 | * @rs: The block reservation with the rgrp to test | |
1966 | * @msecs: The time limit in milliseconds | |
1967 | * | |
1968 | * Returns: True if the rgrp glock has been used within the time limit | |
1969 | */ | |
1970 | static bool gfs2_rgrp_used_recently(const struct gfs2_blkreserv *rs, | |
1971 | u64 msecs) | |
1972 | { | |
1973 | u64 tdiff; | |
1974 | ||
1975 | tdiff = ktime_to_ns(ktime_sub(ktime_get_real(), | |
1976 | rs->rs_rbm.rgd->rd_gl->gl_dstamp)); | |
1977 | ||
1978 | return tdiff > (msecs * 1000 * 1000); | |
1979 | } | |
1980 | ||
9dbe9610 SW |
1981 | static u32 gfs2_orlov_skip(const struct gfs2_inode *ip) |
1982 | { | |
1983 | const struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
1984 | u32 skip; | |
1985 | ||
1986 | get_random_bytes(&skip, sizeof(skip)); | |
1987 | return skip % sdp->sd_rgrps; | |
1988 | } | |
1989 | ||
c743ffd0 SW |
1990 | static bool gfs2_select_rgrp(struct gfs2_rgrpd **pos, const struct gfs2_rgrpd *begin) |
1991 | { | |
1992 | struct gfs2_rgrpd *rgd = *pos; | |
aa8920c9 | 1993 | struct gfs2_sbd *sdp = rgd->rd_sbd; |
c743ffd0 SW |
1994 | |
1995 | rgd = gfs2_rgrpd_get_next(rgd); | |
1996 | if (rgd == NULL) | |
aa8920c9 | 1997 | rgd = gfs2_rgrpd_get_first(sdp); |
c743ffd0 SW |
1998 | *pos = rgd; |
1999 | if (rgd != begin) /* If we didn't wrap */ | |
2000 | return true; | |
2001 | return false; | |
2002 | } | |
2003 | ||
0e27c18c BP |
2004 | /** |
2005 | * fast_to_acquire - determine if a resource group will be fast to acquire | |
2006 | * | |
2007 | * If this is one of our preferred rgrps, it should be quicker to acquire, | |
2008 | * because we tried to set ourselves up as dlm lock master. | |
2009 | */ | |
2010 | static inline int fast_to_acquire(struct gfs2_rgrpd *rgd) | |
2011 | { | |
2012 | struct gfs2_glock *gl = rgd->rd_gl; | |
2013 | ||
2014 | if (gl->gl_state != LM_ST_UNLOCKED && list_empty(&gl->gl_holders) && | |
2015 | !test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags) && | |
2016 | !test_bit(GLF_DEMOTE, &gl->gl_flags)) | |
2017 | return 1; | |
2018 | if (rgd->rd_flags & GFS2_RDF_PREFERRED) | |
2019 | return 1; | |
2020 | return 0; | |
2021 | } | |
2022 | ||
b3b94faa | 2023 | /** |
666d1d8a | 2024 | * gfs2_inplace_reserve - Reserve space in the filesystem |
b3b94faa | 2025 | * @ip: the inode to reserve space for |
7b9cff46 | 2026 | * @ap: the allocation parameters |
b3b94faa | 2027 | * |
25435e5e AD |
2028 | * We try our best to find an rgrp that has at least ap->target blocks |
2029 | * available. After a couple of passes (loops == 2), the prospects of finding | |
2030 | * such an rgrp diminish. At this stage, we return the first rgrp that has | |
2031 | * atleast ap->min_target blocks available. Either way, we set ap->allowed to | |
2032 | * the number of blocks available in the chosen rgrp. | |
2033 | * | |
2034 | * Returns: 0 on success, | |
2035 | * -ENOMEM if a suitable rgrp can't be found | |
2036 | * errno otherwise | |
b3b94faa DT |
2037 | */ |
2038 | ||
25435e5e | 2039 | int gfs2_inplace_reserve(struct gfs2_inode *ip, struct gfs2_alloc_parms *ap) |
b3b94faa | 2040 | { |
feaa7bba | 2041 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
8e2e0047 | 2042 | struct gfs2_rgrpd *begin = NULL; |
a097dc7e | 2043 | struct gfs2_blkreserv *rs = &ip->i_res; |
bcd97c06 | 2044 | int error = 0, rg_locked, flags = 0; |
666d1d8a | 2045 | u64 last_unlinked = NO_BLOCK; |
7c9ca621 | 2046 | int loops = 0; |
f6753df3 | 2047 | u32 free_blocks, skip = 0; |
b3b94faa | 2048 | |
90306c41 BM |
2049 | if (sdp->sd_args.ar_rgrplvb) |
2050 | flags |= GL_SKIP; | |
7b9cff46 | 2051 | if (gfs2_assert_warn(sdp, ap->target)) |
c743ffd0 | 2052 | return -EINVAL; |
8e2e0047 | 2053 | if (gfs2_rs_active(rs)) { |
4a993fb1 | 2054 | begin = rs->rs_rbm.rgd; |
b7eba890 AG |
2055 | } else if (rs->rs_rbm.rgd && |
2056 | rgrp_contains_block(rs->rs_rbm.rgd, ip->i_goal)) { | |
2057 | begin = rs->rs_rbm.rgd; | |
8e2e0047 | 2058 | } else { |
00a158be | 2059 | check_and_update_goal(ip); |
4a993fb1 | 2060 | rs->rs_rbm.rgd = begin = gfs2_blk2rgrpd(sdp, ip->i_goal, 1); |
8e2e0047 | 2061 | } |
7b9cff46 | 2062 | if (S_ISDIR(ip->i_inode.i_mode) && (ap->aflags & GFS2_AF_ORLOV)) |
9dbe9610 | 2063 | skip = gfs2_orlov_skip(ip); |
4a993fb1 | 2064 | if (rs->rs_rbm.rgd == NULL) |
7c9ca621 BP |
2065 | return -EBADSLT; |
2066 | ||
2067 | while (loops < 3) { | |
c743ffd0 SW |
2068 | rg_locked = 1; |
2069 | ||
2070 | if (!gfs2_glock_is_locked_by_me(rs->rs_rbm.rgd->rd_gl)) { | |
2071 | rg_locked = 0; | |
9dbe9610 SW |
2072 | if (skip && skip--) |
2073 | goto next_rgrp; | |
0e27c18c BP |
2074 | if (!gfs2_rs_active(rs)) { |
2075 | if (loops == 0 && | |
2076 | !fast_to_acquire(rs->rs_rbm.rgd)) | |
2077 | goto next_rgrp; | |
2078 | if ((loops < 2) && | |
2079 | gfs2_rgrp_used_recently(rs, 1000) && | |
2080 | gfs2_rgrp_congested(rs->rs_rbm.rgd, loops)) | |
2081 | goto next_rgrp; | |
2082 | } | |
4a993fb1 | 2083 | error = gfs2_glock_nq_init(rs->rs_rbm.rgd->rd_gl, |
8e2e0047 BP |
2084 | LM_ST_EXCLUSIVE, flags, |
2085 | &rs->rs_rgd_gh); | |
c743ffd0 SW |
2086 | if (unlikely(error)) |
2087 | return error; | |
bcd97c06 SW |
2088 | if (!gfs2_rs_active(rs) && (loops < 2) && |
2089 | gfs2_rgrp_congested(rs->rs_rbm.rgd, loops)) | |
2090 | goto skip_rgrp; | |
c743ffd0 | 2091 | if (sdp->sd_args.ar_rgrplvb) { |
4a993fb1 | 2092 | error = update_rgrp_lvb(rs->rs_rbm.rgd); |
c743ffd0 | 2093 | if (unlikely(error)) { |
90306c41 BM |
2094 | gfs2_glock_dq_uninit(&rs->rs_rgd_gh); |
2095 | return error; | |
2096 | } | |
2097 | } | |
292c8c14 | 2098 | } |
666d1d8a | 2099 | |
c743ffd0 | 2100 | /* Skip unuseable resource groups */ |
5ea5050c BP |
2101 | if ((rs->rs_rbm.rgd->rd_flags & (GFS2_RGF_NOALLOC | |
2102 | GFS2_RDF_ERROR)) || | |
25435e5e | 2103 | (loops == 0 && ap->target > rs->rs_rbm.rgd->rd_extfail_pt)) |
c743ffd0 SW |
2104 | goto skip_rgrp; |
2105 | ||
2106 | if (sdp->sd_args.ar_rgrplvb) | |
2107 | gfs2_rgrp_bh_get(rs->rs_rbm.rgd); | |
2108 | ||
2109 | /* Get a reservation if we don't already have one */ | |
2110 | if (!gfs2_rs_active(rs)) | |
7b9cff46 | 2111 | rg_mblk_search(rs->rs_rbm.rgd, ip, ap); |
c743ffd0 SW |
2112 | |
2113 | /* Skip rgrps when we can't get a reservation on first pass */ | |
2114 | if (!gfs2_rs_active(rs) && (loops < 1)) | |
2115 | goto check_rgrp; | |
2116 | ||
2117 | /* If rgrp has enough free space, use it */ | |
f6753df3 BP |
2118 | free_blocks = rgd_free(rs->rs_rbm.rgd, rs); |
2119 | if (free_blocks >= ap->target || | |
25435e5e | 2120 | (loops == 2 && ap->min_target && |
f6753df3 BP |
2121 | free_blocks >= ap->min_target)) { |
2122 | ap->allowed = free_blocks; | |
c743ffd0 | 2123 | return 0; |
b3b94faa | 2124 | } |
c743ffd0 SW |
2125 | check_rgrp: |
2126 | /* Check for unlinked inodes which can be reclaimed */ | |
2127 | if (rs->rs_rbm.rgd->rd_flags & GFS2_RDF_CHECK) | |
2128 | try_rgrp_unlink(rs->rs_rbm.rgd, &last_unlinked, | |
2129 | ip->i_no_addr); | |
2130 | skip_rgrp: | |
1330edbe BP |
2131 | /* Drop reservation, if we couldn't use reserved rgrp */ |
2132 | if (gfs2_rs_active(rs)) | |
2133 | gfs2_rs_deltree(rs); | |
2134 | ||
c743ffd0 SW |
2135 | /* Unlock rgrp if required */ |
2136 | if (!rg_locked) | |
2137 | gfs2_glock_dq_uninit(&rs->rs_rgd_gh); | |
2138 | next_rgrp: | |
2139 | /* Find the next rgrp, and continue looking */ | |
2140 | if (gfs2_select_rgrp(&rs->rs_rbm.rgd, begin)) | |
2141 | continue; | |
9dbe9610 SW |
2142 | if (skip) |
2143 | continue; | |
c743ffd0 SW |
2144 | |
2145 | /* If we've scanned all the rgrps, but found no free blocks | |
2146 | * then this checks for some less likely conditions before | |
2147 | * trying again. | |
2148 | */ | |
c743ffd0 SW |
2149 | loops++; |
2150 | /* Check that fs hasn't grown if writing to rindex */ | |
2151 | if (ip == GFS2_I(sdp->sd_rindex) && !sdp->sd_rindex_uptodate) { | |
2152 | error = gfs2_ri_update(ip); | |
2153 | if (error) | |
2154 | return error; | |
2155 | } | |
2156 | /* Flushing the log may release space */ | |
2157 | if (loops == 2) | |
805c0907 BP |
2158 | gfs2_log_flush(sdp, NULL, GFS2_LOG_HEAD_FLUSH_NORMAL | |
2159 | GFS2_LFC_INPLACE_RESERVE); | |
b3b94faa | 2160 | } |
b3b94faa | 2161 | |
c743ffd0 | 2162 | return -ENOSPC; |
b3b94faa DT |
2163 | } |
2164 | ||
2165 | /** | |
2166 | * gfs2_inplace_release - release an inplace reservation | |
2167 | * @ip: the inode the reservation was taken out on | |
2168 | * | |
2169 | * Release a reservation made by gfs2_inplace_reserve(). | |
2170 | */ | |
2171 | ||
2172 | void gfs2_inplace_release(struct gfs2_inode *ip) | |
2173 | { | |
a097dc7e | 2174 | struct gfs2_blkreserv *rs = &ip->i_res; |
b3b94faa | 2175 | |
6df9f9a2 | 2176 | if (gfs2_holder_initialized(&rs->rs_rgd_gh)) |
564e12b1 | 2177 | gfs2_glock_dq_uninit(&rs->rs_rgd_gh); |
b3b94faa DT |
2178 | } |
2179 | ||
b3e47ca0 BP |
2180 | /** |
2181 | * gfs2_alloc_extent - allocate an extent from a given bitmap | |
4a993fb1 | 2182 | * @rbm: the resource group information |
b3e47ca0 | 2183 | * @dinode: TRUE if the first block we allocate is for a dinode |
c04a2ef3 | 2184 | * @n: The extent length (value/result) |
b3e47ca0 | 2185 | * |
c04a2ef3 | 2186 | * Add the bitmap buffer to the transaction. |
b3e47ca0 | 2187 | * Set the found bits to @new_state to change block's allocation state. |
b3e47ca0 | 2188 | */ |
c04a2ef3 | 2189 | static void gfs2_alloc_extent(const struct gfs2_rbm *rbm, bool dinode, |
4a993fb1 | 2190 | unsigned int *n) |
b3e47ca0 | 2191 | { |
c04a2ef3 | 2192 | struct gfs2_rbm pos = { .rgd = rbm->rgd, }; |
b3e47ca0 | 2193 | const unsigned int elen = *n; |
c04a2ef3 SW |
2194 | u64 block; |
2195 | int ret; | |
b3e47ca0 | 2196 | |
c04a2ef3 SW |
2197 | *n = 1; |
2198 | block = gfs2_rbm_to_block(rbm); | |
e579ed4f | 2199 | gfs2_trans_add_meta(rbm->rgd->rd_gl, rbm_bi(rbm)->bi_bh); |
3e6339dd | 2200 | gfs2_setbit(rbm, true, dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED); |
c04a2ef3 | 2201 | block++; |
60a0b8f9 | 2202 | while (*n < elen) { |
c04a2ef3 | 2203 | ret = gfs2_rbm_from_block(&pos, block); |
dffe12a8 | 2204 | if (ret || gfs2_testbit(&pos, true) != GFS2_BLKST_FREE) |
60a0b8f9 | 2205 | break; |
e579ed4f | 2206 | gfs2_trans_add_meta(pos.rgd->rd_gl, rbm_bi(&pos)->bi_bh); |
3e6339dd | 2207 | gfs2_setbit(&pos, true, GFS2_BLKST_USED); |
60a0b8f9 | 2208 | (*n)++; |
c04a2ef3 | 2209 | block++; |
c8cdf479 | 2210 | } |
b3b94faa DT |
2211 | } |
2212 | ||
2213 | /** | |
2214 | * rgblk_free - Change alloc state of given block(s) | |
2215 | * @sdp: the filesystem | |
2216 | * @bstart: the start of a run of blocks to free | |
2217 | * @blen: the length of the block run (all must lie within ONE RG!) | |
2218 | * @new_state: GFS2_BLKST_XXX the after-allocation block state | |
2219 | * | |
2220 | * Returns: Resource group containing the block(s) | |
2221 | */ | |
2222 | ||
cd915493 SW |
2223 | static struct gfs2_rgrpd *rgblk_free(struct gfs2_sbd *sdp, u64 bstart, |
2224 | u32 blen, unsigned char new_state) | |
b3b94faa | 2225 | { |
3b1d0b9d | 2226 | struct gfs2_rbm rbm; |
d24e0569 | 2227 | struct gfs2_bitmap *bi, *bi_prev = NULL; |
b3b94faa | 2228 | |
3b1d0b9d SW |
2229 | rbm.rgd = gfs2_blk2rgrpd(sdp, bstart, 1); |
2230 | if (!rbm.rgd) { | |
b3b94faa | 2231 | if (gfs2_consist(sdp)) |
382066da | 2232 | fs_err(sdp, "block = %llu\n", (unsigned long long)bstart); |
b3b94faa DT |
2233 | return NULL; |
2234 | } | |
2235 | ||
f654683d AG |
2236 | if (WARN_ON_ONCE(gfs2_rbm_from_block(&rbm, bstart))) |
2237 | return NULL; | |
b3b94faa | 2238 | while (blen--) { |
e579ed4f | 2239 | bi = rbm_bi(&rbm); |
d24e0569 BP |
2240 | if (bi != bi_prev) { |
2241 | if (!bi->bi_clone) { | |
2242 | bi->bi_clone = kmalloc(bi->bi_bh->b_size, | |
2243 | GFP_NOFS | __GFP_NOFAIL); | |
2244 | memcpy(bi->bi_clone + bi->bi_offset, | |
2245 | bi->bi_bh->b_data + bi->bi_offset, | |
2246 | bi->bi_len); | |
2247 | } | |
2248 | gfs2_trans_add_meta(rbm.rgd->rd_gl, bi->bi_bh); | |
2249 | bi_prev = bi; | |
b3b94faa | 2250 | } |
3e6339dd | 2251 | gfs2_setbit(&rbm, false, new_state); |
d24e0569 | 2252 | gfs2_rbm_incr(&rbm); |
b3b94faa DT |
2253 | } |
2254 | ||
3b1d0b9d | 2255 | return rbm.rgd; |
b3b94faa DT |
2256 | } |
2257 | ||
2258 | /** | |
09010978 SW |
2259 | * gfs2_rgrp_dump - print out an rgrp |
2260 | * @seq: The iterator | |
2261 | * @gl: The glock in question | |
2262 | * | |
2263 | */ | |
2264 | ||
ac3beb6a | 2265 | void gfs2_rgrp_dump(struct seq_file *seq, const struct gfs2_glock *gl) |
09010978 | 2266 | { |
8e2e0047 BP |
2267 | struct gfs2_rgrpd *rgd = gl->gl_object; |
2268 | struct gfs2_blkreserv *trs; | |
2269 | const struct rb_node *n; | |
2270 | ||
09010978 | 2271 | if (rgd == NULL) |
ac3beb6a | 2272 | return; |
5ea5050c | 2273 | gfs2_print_dbg(seq, " R: n:%llu f:%02x b:%u/%u i:%u r:%u e:%u\n", |
09010978 | 2274 | (unsigned long long)rgd->rd_addr, rgd->rd_flags, |
8e2e0047 | 2275 | rgd->rd_free, rgd->rd_free_clone, rgd->rd_dinodes, |
5ea5050c | 2276 | rgd->rd_reserved, rgd->rd_extfail_pt); |
72244b6b BP |
2277 | if (rgd->rd_sbd->sd_args.ar_rgrplvb) { |
2278 | struct gfs2_rgrp_lvb *rgl = rgd->rd_rgl; | |
2279 | ||
2280 | gfs2_print_dbg(seq, " L: f:%02x b:%u i:%u\n", | |
2281 | be32_to_cpu(rgl->rl_flags), | |
2282 | be32_to_cpu(rgl->rl_free), | |
2283 | be32_to_cpu(rgl->rl_dinodes)); | |
2284 | } | |
8e2e0047 BP |
2285 | spin_lock(&rgd->rd_rsspin); |
2286 | for (n = rb_first(&rgd->rd_rstree); n; n = rb_next(&trs->rs_node)) { | |
2287 | trs = rb_entry(n, struct gfs2_blkreserv, rs_node); | |
2288 | dump_rs(seq, trs); | |
2289 | } | |
2290 | spin_unlock(&rgd->rd_rsspin); | |
09010978 SW |
2291 | } |
2292 | ||
6050b9c7 SW |
2293 | static void gfs2_rgrp_error(struct gfs2_rgrpd *rgd) |
2294 | { | |
2295 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
2296 | fs_warn(sdp, "rgrp %llu has an error, marking it readonly until umount\n", | |
86d00636 | 2297 | (unsigned long long)rgd->rd_addr); |
6050b9c7 SW |
2298 | fs_warn(sdp, "umount on all nodes and run fsck.gfs2 to fix the error\n"); |
2299 | gfs2_rgrp_dump(NULL, rgd->rd_gl); | |
2300 | rgd->rd_flags |= GFS2_RDF_ERROR; | |
2301 | } | |
2302 | ||
8e2e0047 | 2303 | /** |
5b924ae2 SW |
2304 | * gfs2_adjust_reservation - Adjust (or remove) a reservation after allocation |
2305 | * @ip: The inode we have just allocated blocks for | |
2306 | * @rbm: The start of the allocated blocks | |
2307 | * @len: The extent length | |
8e2e0047 | 2308 | * |
5b924ae2 SW |
2309 | * Adjusts a reservation after an allocation has taken place. If the |
2310 | * reservation does not match the allocation, or if it is now empty | |
2311 | * then it is removed. | |
8e2e0047 | 2312 | */ |
5b924ae2 SW |
2313 | |
2314 | static void gfs2_adjust_reservation(struct gfs2_inode *ip, | |
2315 | const struct gfs2_rbm *rbm, unsigned len) | |
8e2e0047 | 2316 | { |
a097dc7e | 2317 | struct gfs2_blkreserv *rs = &ip->i_res; |
5b924ae2 SW |
2318 | struct gfs2_rgrpd *rgd = rbm->rgd; |
2319 | unsigned rlen; | |
2320 | u64 block; | |
2321 | int ret; | |
8e2e0047 | 2322 | |
5b924ae2 SW |
2323 | spin_lock(&rgd->rd_rsspin); |
2324 | if (gfs2_rs_active(rs)) { | |
2325 | if (gfs2_rbm_eq(&rs->rs_rbm, rbm)) { | |
2326 | block = gfs2_rbm_to_block(rbm); | |
2327 | ret = gfs2_rbm_from_block(&rs->rs_rbm, block + len); | |
2328 | rlen = min(rs->rs_free, len); | |
2329 | rs->rs_free -= rlen; | |
2330 | rgd->rd_reserved -= rlen; | |
9e733d39 | 2331 | trace_gfs2_rs(rs, TRACE_RS_CLAIM); |
5b924ae2 SW |
2332 | if (rs->rs_free && !ret) |
2333 | goto out; | |
1a855033 BP |
2334 | /* We used up our block reservation, so we should |
2335 | reserve more blocks next time. */ | |
2336 | atomic_add(RGRP_RSRV_ADDBLKS, &rs->rs_sizehint); | |
5b924ae2 | 2337 | } |
20095218 | 2338 | __rs_deltree(rs); |
8e2e0047 | 2339 | } |
5b924ae2 SW |
2340 | out: |
2341 | spin_unlock(&rgd->rd_rsspin); | |
8e2e0047 BP |
2342 | } |
2343 | ||
9e07f2cb SW |
2344 | /** |
2345 | * gfs2_set_alloc_start - Set starting point for block allocation | |
2346 | * @rbm: The rbm which will be set to the required location | |
2347 | * @ip: The gfs2 inode | |
2348 | * @dinode: Flag to say if allocation includes a new inode | |
2349 | * | |
2350 | * This sets the starting point from the reservation if one is active | |
2351 | * otherwise it falls back to guessing a start point based on the | |
2352 | * inode's goal block or the last allocation point in the rgrp. | |
2353 | */ | |
2354 | ||
2355 | static void gfs2_set_alloc_start(struct gfs2_rbm *rbm, | |
2356 | const struct gfs2_inode *ip, bool dinode) | |
2357 | { | |
2358 | u64 goal; | |
2359 | ||
a097dc7e BP |
2360 | if (gfs2_rs_active(&ip->i_res)) { |
2361 | *rbm = ip->i_res.rs_rbm; | |
9e07f2cb SW |
2362 | return; |
2363 | } | |
2364 | ||
2365 | if (!dinode && rgrp_contains_block(rbm->rgd, ip->i_goal)) | |
2366 | goal = ip->i_goal; | |
2367 | else | |
2368 | goal = rbm->rgd->rd_last_alloc + rbm->rgd->rd_data0; | |
2369 | ||
f654683d AG |
2370 | if (WARN_ON_ONCE(gfs2_rbm_from_block(rbm, goal))) { |
2371 | rbm->bii = 0; | |
2372 | rbm->offset = 0; | |
2373 | } | |
9e07f2cb SW |
2374 | } |
2375 | ||
09010978 | 2376 | /** |
6e87ed0f | 2377 | * gfs2_alloc_blocks - Allocate one or more blocks of data and/or a dinode |
1639431a | 2378 | * @ip: the inode to allocate the block for |
09010978 | 2379 | * @bn: Used to return the starting block number |
8e2e0047 | 2380 | * @nblocks: requested number of blocks/extent length (value/result) |
6e87ed0f | 2381 | * @dinode: 1 if we're allocating a dinode block, else 0 |
3c5d785a | 2382 | * @generation: the generation number of the inode |
b3b94faa | 2383 | * |
09010978 | 2384 | * Returns: 0 or error |
b3b94faa DT |
2385 | */ |
2386 | ||
6a8099ed | 2387 | int gfs2_alloc_blocks(struct gfs2_inode *ip, u64 *bn, unsigned int *nblocks, |
6e87ed0f | 2388 | bool dinode, u64 *generation) |
b3b94faa | 2389 | { |
feaa7bba | 2390 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
d9ba7615 | 2391 | struct buffer_head *dibh; |
b7eba890 | 2392 | struct gfs2_rbm rbm = { .rgd = ip->i_res.rs_rbm.rgd, }; |
6a8099ed | 2393 | unsigned int ndata; |
3c5d785a | 2394 | u64 block; /* block, within the file system scope */ |
d9ba7615 | 2395 | int error; |
b3b94faa | 2396 | |
9e07f2cb | 2397 | gfs2_set_alloc_start(&rbm, ip, dinode); |
8381e602 | 2398 | error = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, NULL, ip, false); |
62e252ee | 2399 | |
137834a6 | 2400 | if (error == -ENOSPC) { |
9e07f2cb | 2401 | gfs2_set_alloc_start(&rbm, ip, dinode); |
8381e602 | 2402 | error = gfs2_rbm_find(&rbm, GFS2_BLKST_FREE, NULL, NULL, false); |
137834a6 SW |
2403 | } |
2404 | ||
62e252ee | 2405 | /* Since all blocks are reserved in advance, this shouldn't happen */ |
5b924ae2 | 2406 | if (error) { |
5ea5050c | 2407 | fs_warn(sdp, "inum=%llu error=%d, nblocks=%u, full=%d fail_pt=%d\n", |
9e733d39 | 2408 | (unsigned long long)ip->i_no_addr, error, *nblocks, |
5ea5050c BP |
2409 | test_bit(GBF_FULL, &rbm.rgd->rd_bits->bi_flags), |
2410 | rbm.rgd->rd_extfail_pt); | |
62e252ee | 2411 | goto rgrp_error; |
8e2e0047 | 2412 | } |
62e252ee | 2413 | |
c04a2ef3 SW |
2414 | gfs2_alloc_extent(&rbm, dinode, nblocks); |
2415 | block = gfs2_rbm_to_block(&rbm); | |
c743ffd0 | 2416 | rbm.rgd->rd_last_alloc = block - rbm.rgd->rd_data0; |
a097dc7e | 2417 | if (gfs2_rs_active(&ip->i_res)) |
5b924ae2 | 2418 | gfs2_adjust_reservation(ip, &rbm, *nblocks); |
6a8099ed SW |
2419 | ndata = *nblocks; |
2420 | if (dinode) | |
2421 | ndata--; | |
b3e47ca0 | 2422 | |
3c5d785a | 2423 | if (!dinode) { |
6a8099ed | 2424 | ip->i_goal = block + ndata - 1; |
3c5d785a BP |
2425 | error = gfs2_meta_inode_buffer(ip, &dibh); |
2426 | if (error == 0) { | |
2427 | struct gfs2_dinode *di = | |
2428 | (struct gfs2_dinode *)dibh->b_data; | |
350a9b0a | 2429 | gfs2_trans_add_meta(ip->i_gl, dibh); |
3c5d785a BP |
2430 | di->di_goal_meta = di->di_goal_data = |
2431 | cpu_to_be64(ip->i_goal); | |
2432 | brelse(dibh); | |
2433 | } | |
d9ba7615 | 2434 | } |
4a993fb1 | 2435 | if (rbm.rgd->rd_free < *nblocks) { |
e54c78a2 | 2436 | fs_warn(sdp, "nblocks=%u\n", *nblocks); |
09010978 | 2437 | goto rgrp_error; |
8e2e0047 | 2438 | } |
09010978 | 2439 | |
4a993fb1 | 2440 | rbm.rgd->rd_free -= *nblocks; |
3c5d785a | 2441 | if (dinode) { |
4a993fb1 SW |
2442 | rbm.rgd->rd_dinodes++; |
2443 | *generation = rbm.rgd->rd_igeneration++; | |
3c5d785a | 2444 | if (*generation == 0) |
4a993fb1 | 2445 | *generation = rbm.rgd->rd_igeneration++; |
3c5d785a | 2446 | } |
b3b94faa | 2447 | |
350a9b0a | 2448 | gfs2_trans_add_meta(rbm.rgd->rd_gl, rbm.rgd->rd_bits[0].bi_bh); |
4a993fb1 | 2449 | gfs2_rgrp_out(rbm.rgd, rbm.rgd->rd_bits[0].bi_bh->b_data); |
b3b94faa | 2450 | |
6a8099ed | 2451 | gfs2_statfs_change(sdp, 0, -(s64)*nblocks, dinode ? 1 : 0); |
3c5d785a | 2452 | if (dinode) |
b2c8b3ea | 2453 | gfs2_trans_add_unrevoke(sdp, block, *nblocks); |
6a8099ed | 2454 | |
fd4b4e04 | 2455 | gfs2_quota_change(ip, *nblocks, ip->i_inode.i_uid, ip->i_inode.i_gid); |
b3b94faa | 2456 | |
4a993fb1 SW |
2457 | rbm.rgd->rd_free_clone -= *nblocks; |
2458 | trace_gfs2_block_alloc(ip, rbm.rgd, block, *nblocks, | |
6e87ed0f | 2459 | dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED); |
6050b9c7 SW |
2460 | *bn = block; |
2461 | return 0; | |
2462 | ||
2463 | rgrp_error: | |
4a993fb1 | 2464 | gfs2_rgrp_error(rbm.rgd); |
6050b9c7 | 2465 | return -EIO; |
b3b94faa DT |
2466 | } |
2467 | ||
2468 | /** | |
46fcb2ed | 2469 | * __gfs2_free_blocks - free a contiguous run of block(s) |
b3b94faa DT |
2470 | * @ip: the inode these blocks are being freed from |
2471 | * @bstart: first block of a run of contiguous blocks | |
2472 | * @blen: the length of the block run | |
46fcb2ed | 2473 | * @meta: 1 if the blocks represent metadata |
b3b94faa DT |
2474 | * |
2475 | */ | |
2476 | ||
46fcb2ed | 2477 | void __gfs2_free_blocks(struct gfs2_inode *ip, u64 bstart, u32 blen, int meta) |
b3b94faa | 2478 | { |
feaa7bba | 2479 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
2480 | struct gfs2_rgrpd *rgd; |
2481 | ||
2482 | rgd = rgblk_free(sdp, bstart, blen, GFS2_BLKST_FREE); | |
2483 | if (!rgd) | |
2484 | return; | |
41db1ab9 | 2485 | trace_gfs2_block_alloc(ip, rgd, bstart, blen, GFS2_BLKST_FREE); |
cfc8b549 | 2486 | rgd->rd_free += blen; |
66fc061b | 2487 | rgd->rd_flags &= ~GFS2_RGF_TRIMMED; |
350a9b0a | 2488 | gfs2_trans_add_meta(rgd->rd_gl, rgd->rd_bits[0].bi_bh); |
42d52e38 | 2489 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
b3b94faa | 2490 | |
6d3117b4 | 2491 | /* Directories keep their data in the metadata address space */ |
46fcb2ed | 2492 | if (meta || ip->i_depth) |
6d3117b4 | 2493 | gfs2_meta_wipe(ip, bstart, blen); |
4c16c36a | 2494 | } |
b3b94faa | 2495 | |
4c16c36a BP |
2496 | /** |
2497 | * gfs2_free_meta - free a contiguous run of data block(s) | |
2498 | * @ip: the inode these blocks are being freed from | |
2499 | * @bstart: first block of a run of contiguous blocks | |
2500 | * @blen: the length of the block run | |
2501 | * | |
2502 | */ | |
2503 | ||
2504 | void gfs2_free_meta(struct gfs2_inode *ip, u64 bstart, u32 blen) | |
2505 | { | |
2506 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
2507 | ||
46fcb2ed | 2508 | __gfs2_free_blocks(ip, bstart, blen, 1); |
b3b94faa | 2509 | gfs2_statfs_change(sdp, 0, +blen, 0); |
2933f925 | 2510 | gfs2_quota_change(ip, -(s64)blen, ip->i_inode.i_uid, ip->i_inode.i_gid); |
b3b94faa DT |
2511 | } |
2512 | ||
feaa7bba SW |
2513 | void gfs2_unlink_di(struct inode *inode) |
2514 | { | |
2515 | struct gfs2_inode *ip = GFS2_I(inode); | |
2516 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
2517 | struct gfs2_rgrpd *rgd; | |
dbb7cae2 | 2518 | u64 blkno = ip->i_no_addr; |
feaa7bba SW |
2519 | |
2520 | rgd = rgblk_free(sdp, blkno, 1, GFS2_BLKST_UNLINKED); | |
2521 | if (!rgd) | |
2522 | return; | |
41db1ab9 | 2523 | trace_gfs2_block_alloc(ip, rgd, blkno, 1, GFS2_BLKST_UNLINKED); |
350a9b0a | 2524 | gfs2_trans_add_meta(rgd->rd_gl, rgd->rd_bits[0].bi_bh); |
42d52e38 | 2525 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
f5580d0f | 2526 | be32_add_cpu(&rgd->rd_rgl->rl_unlinked, 1); |
feaa7bba SW |
2527 | } |
2528 | ||
a18c78c5 | 2529 | void gfs2_free_di(struct gfs2_rgrpd *rgd, struct gfs2_inode *ip) |
b3b94faa DT |
2530 | { |
2531 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
2532 | struct gfs2_rgrpd *tmp_rgd; | |
2533 | ||
a18c78c5 | 2534 | tmp_rgd = rgblk_free(sdp, ip->i_no_addr, 1, GFS2_BLKST_FREE); |
b3b94faa DT |
2535 | if (!tmp_rgd) |
2536 | return; | |
2537 | gfs2_assert_withdraw(sdp, rgd == tmp_rgd); | |
2538 | ||
73f74948 | 2539 | if (!rgd->rd_dinodes) |
b3b94faa | 2540 | gfs2_consist_rgrpd(rgd); |
73f74948 | 2541 | rgd->rd_dinodes--; |
cfc8b549 | 2542 | rgd->rd_free++; |
b3b94faa | 2543 | |
350a9b0a | 2544 | gfs2_trans_add_meta(rgd->rd_gl, rgd->rd_bits[0].bi_bh); |
42d52e38 | 2545 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
f5580d0f | 2546 | be32_add_cpu(&rgd->rd_rgl->rl_unlinked, -1); |
b3b94faa DT |
2547 | |
2548 | gfs2_statfs_change(sdp, 0, +1, -1); | |
41db1ab9 | 2549 | trace_gfs2_block_alloc(ip, rgd, ip->i_no_addr, 1, GFS2_BLKST_FREE); |
2933f925 | 2550 | gfs2_quota_change(ip, -1, ip->i_inode.i_uid, ip->i_inode.i_gid); |
dbb7cae2 | 2551 | gfs2_meta_wipe(ip, ip->i_no_addr, 1); |
b3b94faa DT |
2552 | } |
2553 | ||
acf7e244 SW |
2554 | /** |
2555 | * gfs2_check_blk_type - Check the type of a block | |
2556 | * @sdp: The superblock | |
2557 | * @no_addr: The block number to check | |
2558 | * @type: The block type we are looking for | |
2559 | * | |
2560 | * Returns: 0 if the block type matches the expected type | |
2561 | * -ESTALE if it doesn't match | |
2562 | * or -ve errno if something went wrong while checking | |
2563 | */ | |
2564 | ||
2565 | int gfs2_check_blk_type(struct gfs2_sbd *sdp, u64 no_addr, unsigned int type) | |
2566 | { | |
2567 | struct gfs2_rgrpd *rgd; | |
8339ee54 | 2568 | struct gfs2_holder rgd_gh; |
dffe12a8 | 2569 | struct gfs2_rbm rbm; |
58884c4d | 2570 | int error = -EINVAL; |
acf7e244 | 2571 | |
66fc061b | 2572 | rgd = gfs2_blk2rgrpd(sdp, no_addr, 1); |
acf7e244 | 2573 | if (!rgd) |
8339ee54 | 2574 | goto fail; |
acf7e244 SW |
2575 | |
2576 | error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_SHARED, 0, &rgd_gh); | |
2577 | if (error) | |
8339ee54 | 2578 | goto fail; |
acf7e244 | 2579 | |
dffe12a8 BP |
2580 | rbm.rgd = rgd; |
2581 | error = gfs2_rbm_from_block(&rbm, no_addr); | |
f654683d AG |
2582 | if (WARN_ON_ONCE(error)) |
2583 | goto fail; | |
dffe12a8 BP |
2584 | |
2585 | if (gfs2_testbit(&rbm, false) != type) | |
acf7e244 SW |
2586 | error = -ESTALE; |
2587 | ||
2588 | gfs2_glock_dq_uninit(&rgd_gh); | |
acf7e244 SW |
2589 | fail: |
2590 | return error; | |
2591 | } | |
2592 | ||
b3b94faa DT |
2593 | /** |
2594 | * gfs2_rlist_add - add a RG to a list of RGs | |
70b0c365 | 2595 | * @ip: the inode |
b3b94faa DT |
2596 | * @rlist: the list of resource groups |
2597 | * @block: the block | |
2598 | * | |
2599 | * Figure out what RG a block belongs to and add that RG to the list | |
2600 | * | |
2601 | * FIXME: Don't use NOFAIL | |
2602 | * | |
2603 | */ | |
2604 | ||
70b0c365 | 2605 | void gfs2_rlist_add(struct gfs2_inode *ip, struct gfs2_rgrp_list *rlist, |
cd915493 | 2606 | u64 block) |
b3b94faa | 2607 | { |
70b0c365 | 2608 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
2609 | struct gfs2_rgrpd *rgd; |
2610 | struct gfs2_rgrpd **tmp; | |
2611 | unsigned int new_space; | |
2612 | unsigned int x; | |
2613 | ||
2614 | if (gfs2_assert_warn(sdp, !rlist->rl_ghs)) | |
2615 | return; | |
2616 | ||
03f8c41c AG |
2617 | /* |
2618 | * The resource group last accessed is kept in the last position. | |
2619 | */ | |
2620 | ||
2621 | if (rlist->rl_rgrps) { | |
2622 | rgd = rlist->rl_rgd[rlist->rl_rgrps - 1]; | |
2623 | if (rgrp_contains_block(rgd, block)) | |
2624 | return; | |
66fc061b | 2625 | rgd = gfs2_blk2rgrpd(sdp, block, 1); |
03f8c41c | 2626 | } else { |
b7eba890 | 2627 | rgd = ip->i_res.rs_rbm.rgd; |
03f8c41c AG |
2628 | if (!rgd || !rgrp_contains_block(rgd, block)) |
2629 | rgd = gfs2_blk2rgrpd(sdp, block, 1); | |
2630 | } | |
2631 | ||
b3b94faa | 2632 | if (!rgd) { |
03f8c41c AG |
2633 | fs_err(sdp, "rlist_add: no rgrp for block %llu\n", |
2634 | (unsigned long long)block); | |
b3b94faa DT |
2635 | return; |
2636 | } | |
2637 | ||
03f8c41c AG |
2638 | for (x = 0; x < rlist->rl_rgrps; x++) { |
2639 | if (rlist->rl_rgd[x] == rgd) { | |
2640 | swap(rlist->rl_rgd[x], | |
2641 | rlist->rl_rgd[rlist->rl_rgrps - 1]); | |
b3b94faa | 2642 | return; |
03f8c41c AG |
2643 | } |
2644 | } | |
b3b94faa DT |
2645 | |
2646 | if (rlist->rl_rgrps == rlist->rl_space) { | |
2647 | new_space = rlist->rl_space + 10; | |
2648 | ||
2649 | tmp = kcalloc(new_space, sizeof(struct gfs2_rgrpd *), | |
dd894be8 | 2650 | GFP_NOFS | __GFP_NOFAIL); |
b3b94faa DT |
2651 | |
2652 | if (rlist->rl_rgd) { | |
2653 | memcpy(tmp, rlist->rl_rgd, | |
2654 | rlist->rl_space * sizeof(struct gfs2_rgrpd *)); | |
2655 | kfree(rlist->rl_rgd); | |
2656 | } | |
2657 | ||
2658 | rlist->rl_space = new_space; | |
2659 | rlist->rl_rgd = tmp; | |
2660 | } | |
2661 | ||
2662 | rlist->rl_rgd[rlist->rl_rgrps++] = rgd; | |
2663 | } | |
2664 | ||
2665 | /** | |
2666 | * gfs2_rlist_alloc - all RGs have been added to the rlist, now allocate | |
2667 | * and initialize an array of glock holders for them | |
2668 | * @rlist: the list of resource groups | |
2669 | * @state: the lock state to acquire the RG lock in | |
b3b94faa DT |
2670 | * |
2671 | * FIXME: Don't use NOFAIL | |
2672 | * | |
2673 | */ | |
2674 | ||
fe6c991c | 2675 | void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state) |
b3b94faa DT |
2676 | { |
2677 | unsigned int x; | |
2678 | ||
6da2ec56 KC |
2679 | rlist->rl_ghs = kmalloc_array(rlist->rl_rgrps, |
2680 | sizeof(struct gfs2_holder), | |
2681 | GFP_NOFS | __GFP_NOFAIL); | |
b3b94faa DT |
2682 | for (x = 0; x < rlist->rl_rgrps; x++) |
2683 | gfs2_holder_init(rlist->rl_rgd[x]->rd_gl, | |
fe6c991c | 2684 | state, 0, |
b3b94faa DT |
2685 | &rlist->rl_ghs[x]); |
2686 | } | |
2687 | ||
2688 | /** | |
2689 | * gfs2_rlist_free - free a resource group list | |
27ff6a0f | 2690 | * @rlist: the list of resource groups |
b3b94faa DT |
2691 | * |
2692 | */ | |
2693 | ||
2694 | void gfs2_rlist_free(struct gfs2_rgrp_list *rlist) | |
2695 | { | |
2696 | unsigned int x; | |
2697 | ||
2698 | kfree(rlist->rl_rgd); | |
2699 | ||
2700 | if (rlist->rl_ghs) { | |
2701 | for (x = 0; x < rlist->rl_rgrps; x++) | |
2702 | gfs2_holder_uninit(&rlist->rl_ghs[x]); | |
2703 | kfree(rlist->rl_ghs); | |
8e2e0047 | 2704 | rlist->rl_ghs = NULL; |
b3b94faa DT |
2705 | } |
2706 | } | |
2707 |