Commit | Line | Data |
---|---|---|
b3b94faa DT |
1 | /* |
2 | * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. | |
fe6c991c | 3 | * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved. |
b3b94faa DT |
4 | * |
5 | * This copyrighted material is made available to anyone wishing to use, | |
6 | * modify, copy, or redistribute it subject to the terms and conditions | |
e9fc2aa0 | 7 | * of the GNU General Public License version 2. |
b3b94faa DT |
8 | */ |
9 | ||
b3b94faa DT |
10 | #include <linux/slab.h> |
11 | #include <linux/spinlock.h> | |
12 | #include <linux/completion.h> | |
13 | #include <linux/buffer_head.h> | |
f42faf4f | 14 | #include <linux/fs.h> |
5c676f6d | 15 | #include <linux/gfs2_ondisk.h> |
1f466a47 | 16 | #include <linux/prefetch.h> |
f15ab561 | 17 | #include <linux/blkdev.h> |
7c9ca621 | 18 | #include <linux/rbtree.h> |
b3b94faa DT |
19 | |
20 | #include "gfs2.h" | |
5c676f6d | 21 | #include "incore.h" |
b3b94faa DT |
22 | #include "glock.h" |
23 | #include "glops.h" | |
b3b94faa DT |
24 | #include "lops.h" |
25 | #include "meta_io.h" | |
26 | #include "quota.h" | |
27 | #include "rgrp.h" | |
28 | #include "super.h" | |
29 | #include "trans.h" | |
5c676f6d | 30 | #include "util.h" |
172e045a | 31 | #include "log.h" |
c8cdf479 | 32 | #include "inode.h" |
63997775 | 33 | #include "trace_gfs2.h" |
b3b94faa | 34 | |
2c1e52aa | 35 | #define BFITNOENT ((u32)~0) |
6760bdcd | 36 | #define NO_BLOCK ((u64)~0) |
88c8ab1f | 37 | |
1f466a47 BP |
38 | #if BITS_PER_LONG == 32 |
39 | #define LBITMASK (0x55555555UL) | |
40 | #define LBITSKIP55 (0x55555555UL) | |
41 | #define LBITSKIP00 (0x00000000UL) | |
42 | #else | |
43 | #define LBITMASK (0x5555555555555555UL) | |
44 | #define LBITSKIP55 (0x5555555555555555UL) | |
45 | #define LBITSKIP00 (0x0000000000000000UL) | |
46 | #endif | |
47 | ||
88c8ab1f SW |
48 | /* |
49 | * These routines are used by the resource group routines (rgrp.c) | |
50 | * to keep track of block allocation. Each block is represented by two | |
feaa7bba SW |
51 | * bits. So, each byte represents GFS2_NBBY (i.e. 4) blocks. |
52 | * | |
53 | * 0 = Free | |
54 | * 1 = Used (not metadata) | |
55 | * 2 = Unlinked (still in use) inode | |
56 | * 3 = Used (metadata) | |
88c8ab1f SW |
57 | */ |
58 | ||
59 | static const char valid_change[16] = { | |
60 | /* current */ | |
feaa7bba | 61 | /* n */ 0, 1, 1, 1, |
88c8ab1f | 62 | /* e */ 1, 0, 0, 0, |
feaa7bba | 63 | /* w */ 0, 0, 0, 1, |
88c8ab1f SW |
64 | 1, 0, 0, 0 |
65 | }; | |
66 | ||
c8cdf479 | 67 | static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal, |
6a8099ed | 68 | unsigned char old_state, |
b3e47ca0 | 69 | struct gfs2_bitmap **rbi); |
c8cdf479 | 70 | |
88c8ab1f SW |
71 | /** |
72 | * gfs2_setbit - Set a bit in the bitmaps | |
29c578f5 BP |
73 | * @rgd: the resource group descriptor |
74 | * @buf1: the primary buffer that holds the bitmaps | |
75 | * @buf2: the clone buffer that holds the bitmaps | |
76 | * @bi: the bitmap structure | |
88c8ab1f SW |
77 | * @block: the block to set |
78 | * @new_state: the new state of the block | |
79 | * | |
80 | */ | |
81 | ||
b45e41d7 | 82 | static inline void gfs2_setbit(struct gfs2_rgrpd *rgd, unsigned char *buf1, |
29c578f5 BP |
83 | unsigned char *buf2, struct gfs2_bitmap *bi, |
84 | u32 block, unsigned char new_state) | |
88c8ab1f | 85 | { |
b45e41d7 | 86 | unsigned char *byte1, *byte2, *end, cur_state; |
95c8e17f | 87 | unsigned int buflen = bi->bi_len; |
b45e41d7 | 88 | const unsigned int bit = (block % GFS2_NBBY) * GFS2_BIT_SIZE; |
88c8ab1f | 89 | |
29c578f5 BP |
90 | byte1 = buf1 + bi->bi_offset + (block / GFS2_NBBY); |
91 | end = buf1 + bi->bi_offset + buflen; | |
88c8ab1f | 92 | |
b45e41d7 | 93 | BUG_ON(byte1 >= end); |
88c8ab1f | 94 | |
b45e41d7 | 95 | cur_state = (*byte1 >> bit) & GFS2_BIT_MASK; |
88c8ab1f | 96 | |
b45e41d7 | 97 | if (unlikely(!valid_change[new_state * 4 + cur_state])) { |
95c8e17f BP |
98 | printk(KERN_WARNING "GFS2: buf_blk = 0x%llx old_state=%d, " |
99 | "new_state=%d\n", | |
100 | (unsigned long long)block, cur_state, new_state); | |
101 | printk(KERN_WARNING "GFS2: rgrp=0x%llx bi_start=0x%lx\n", | |
102 | (unsigned long long)rgd->rd_addr, | |
103 | (unsigned long)bi->bi_start); | |
104 | printk(KERN_WARNING "GFS2: bi_offset=0x%lx bi_len=0x%lx\n", | |
105 | (unsigned long)bi->bi_offset, | |
106 | (unsigned long)bi->bi_len); | |
107 | dump_stack(); | |
88c8ab1f | 108 | gfs2_consist_rgrpd(rgd); |
b45e41d7 SW |
109 | return; |
110 | } | |
111 | *byte1 ^= (cur_state ^ new_state) << bit; | |
112 | ||
113 | if (buf2) { | |
29c578f5 | 114 | byte2 = buf2 + bi->bi_offset + (block / GFS2_NBBY); |
b45e41d7 SW |
115 | cur_state = (*byte2 >> bit) & GFS2_BIT_MASK; |
116 | *byte2 ^= (cur_state ^ new_state) << bit; | |
117 | } | |
88c8ab1f SW |
118 | } |
119 | ||
120 | /** | |
121 | * gfs2_testbit - test a bit in the bitmaps | |
122 | * @buffer: the buffer that holds the bitmaps | |
123 | * @buflen: the length (in bytes) of the buffer | |
124 | * @block: the block to read | |
125 | * | |
126 | */ | |
127 | ||
b45e41d7 SW |
128 | static inline unsigned char gfs2_testbit(struct gfs2_rgrpd *rgd, |
129 | const unsigned char *buffer, | |
130 | unsigned int buflen, u32 block) | |
88c8ab1f | 131 | { |
b45e41d7 SW |
132 | const unsigned char *byte, *end; |
133 | unsigned char cur_state; | |
88c8ab1f SW |
134 | unsigned int bit; |
135 | ||
136 | byte = buffer + (block / GFS2_NBBY); | |
137 | bit = (block % GFS2_NBBY) * GFS2_BIT_SIZE; | |
138 | end = buffer + buflen; | |
139 | ||
140 | gfs2_assert(rgd->rd_sbd, byte < end); | |
141 | ||
142 | cur_state = (*byte >> bit) & GFS2_BIT_MASK; | |
143 | ||
144 | return cur_state; | |
145 | } | |
146 | ||
223b2b88 SW |
147 | /** |
148 | * gfs2_bit_search | |
149 | * @ptr: Pointer to bitmap data | |
150 | * @mask: Mask to use (normally 0x55555.... but adjusted for search start) | |
151 | * @state: The state we are searching for | |
152 | * | |
153 | * We xor the bitmap data with a patter which is the bitwise opposite | |
154 | * of what we are looking for, this gives rise to a pattern of ones | |
155 | * wherever there is a match. Since we have two bits per entry, we | |
156 | * take this pattern, shift it down by one place and then and it with | |
157 | * the original. All the even bit positions (0,2,4, etc) then represent | |
158 | * successful matches, so we mask with 0x55555..... to remove the unwanted | |
159 | * odd bit positions. | |
160 | * | |
161 | * This allows searching of a whole u64 at once (32 blocks) with a | |
162 | * single test (on 64 bit arches). | |
163 | */ | |
164 | ||
165 | static inline u64 gfs2_bit_search(const __le64 *ptr, u64 mask, u8 state) | |
166 | { | |
167 | u64 tmp; | |
168 | static const u64 search[] = { | |
075ac448 HE |
169 | [0] = 0xffffffffffffffffULL, |
170 | [1] = 0xaaaaaaaaaaaaaaaaULL, | |
171 | [2] = 0x5555555555555555ULL, | |
172 | [3] = 0x0000000000000000ULL, | |
223b2b88 SW |
173 | }; |
174 | tmp = le64_to_cpu(*ptr) ^ search[state]; | |
175 | tmp &= (tmp >> 1); | |
176 | tmp &= mask; | |
177 | return tmp; | |
178 | } | |
179 | ||
88c8ab1f SW |
180 | /** |
181 | * gfs2_bitfit - Search an rgrp's bitmap buffer to find a bit-pair representing | |
182 | * a block in a given allocation state. | |
183 | * @buffer: the buffer that holds the bitmaps | |
223b2b88 | 184 | * @len: the length (in bytes) of the buffer |
88c8ab1f | 185 | * @goal: start search at this block's bit-pair (within @buffer) |
223b2b88 | 186 | * @state: GFS2_BLKST_XXX the state of the block we're looking for. |
88c8ab1f SW |
187 | * |
188 | * Scope of @goal and returned block number is only within this bitmap buffer, | |
189 | * not entire rgrp or filesystem. @buffer will be offset from the actual | |
223b2b88 SW |
190 | * beginning of a bitmap block buffer, skipping any header structures, but |
191 | * headers are always a multiple of 64 bits long so that the buffer is | |
192 | * always aligned to a 64 bit boundary. | |
193 | * | |
194 | * The size of the buffer is in bytes, but is it assumed that it is | |
fd589a8f | 195 | * always ok to read a complete multiple of 64 bits at the end |
223b2b88 | 196 | * of the block in case the end is no aligned to a natural boundary. |
88c8ab1f SW |
197 | * |
198 | * Return: the block number (bitmap buffer scope) that was found | |
199 | */ | |
200 | ||
02ab1721 HE |
201 | static u32 gfs2_bitfit(const u8 *buf, const unsigned int len, |
202 | u32 goal, u8 state) | |
88c8ab1f | 203 | { |
223b2b88 SW |
204 | u32 spoint = (goal << 1) & ((8*sizeof(u64)) - 1); |
205 | const __le64 *ptr = ((__le64 *)buf) + (goal >> 5); | |
206 | const __le64 *end = (__le64 *)(buf + ALIGN(len, sizeof(u64))); | |
207 | u64 tmp; | |
075ac448 | 208 | u64 mask = 0x5555555555555555ULL; |
223b2b88 SW |
209 | u32 bit; |
210 | ||
211 | BUG_ON(state > 3); | |
212 | ||
213 | /* Mask off bits we don't care about at the start of the search */ | |
214 | mask <<= spoint; | |
215 | tmp = gfs2_bit_search(ptr, mask, state); | |
216 | ptr++; | |
217 | while(tmp == 0 && ptr < end) { | |
075ac448 | 218 | tmp = gfs2_bit_search(ptr, 0x5555555555555555ULL, state); |
223b2b88 | 219 | ptr++; |
1f466a47 | 220 | } |
223b2b88 SW |
221 | /* Mask off any bits which are more than len bytes from the start */ |
222 | if (ptr == end && (len & (sizeof(u64) - 1))) | |
223 | tmp &= (((u64)~0) >> (64 - 8*(len & (sizeof(u64) - 1)))); | |
224 | /* Didn't find anything, so return */ | |
225 | if (tmp == 0) | |
226 | return BFITNOENT; | |
227 | ptr--; | |
d8bd504a | 228 | bit = __ffs64(tmp); |
223b2b88 SW |
229 | bit /= 2; /* two bits per entry in the bitmap */ |
230 | return (((const unsigned char *)ptr - buf) * GFS2_NBBY) + bit; | |
88c8ab1f SW |
231 | } |
232 | ||
233 | /** | |
234 | * gfs2_bitcount - count the number of bits in a certain state | |
235 | * @buffer: the buffer that holds the bitmaps | |
236 | * @buflen: the length (in bytes) of the buffer | |
237 | * @state: the state of the block we're looking for | |
238 | * | |
239 | * Returns: The number of bits | |
240 | */ | |
241 | ||
110acf38 SW |
242 | static u32 gfs2_bitcount(struct gfs2_rgrpd *rgd, const u8 *buffer, |
243 | unsigned int buflen, u8 state) | |
88c8ab1f | 244 | { |
110acf38 SW |
245 | const u8 *byte = buffer; |
246 | const u8 *end = buffer + buflen; | |
247 | const u8 state1 = state << 2; | |
248 | const u8 state2 = state << 4; | |
249 | const u8 state3 = state << 6; | |
cd915493 | 250 | u32 count = 0; |
88c8ab1f SW |
251 | |
252 | for (; byte < end; byte++) { | |
253 | if (((*byte) & 0x03) == state) | |
254 | count++; | |
255 | if (((*byte) & 0x0C) == state1) | |
256 | count++; | |
257 | if (((*byte) & 0x30) == state2) | |
258 | count++; | |
259 | if (((*byte) & 0xC0) == state3) | |
260 | count++; | |
261 | } | |
262 | ||
263 | return count; | |
264 | } | |
265 | ||
b3b94faa DT |
266 | /** |
267 | * gfs2_rgrp_verify - Verify that a resource group is consistent | |
268 | * @sdp: the filesystem | |
269 | * @rgd: the rgrp | |
270 | * | |
271 | */ | |
272 | ||
273 | void gfs2_rgrp_verify(struct gfs2_rgrpd *rgd) | |
274 | { | |
275 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
276 | struct gfs2_bitmap *bi = NULL; | |
bb8d8a6f | 277 | u32 length = rgd->rd_length; |
cd915493 | 278 | u32 count[4], tmp; |
b3b94faa DT |
279 | int buf, x; |
280 | ||
cd915493 | 281 | memset(count, 0, 4 * sizeof(u32)); |
b3b94faa DT |
282 | |
283 | /* Count # blocks in each of 4 possible allocation states */ | |
284 | for (buf = 0; buf < length; buf++) { | |
285 | bi = rgd->rd_bits + buf; | |
286 | for (x = 0; x < 4; x++) | |
287 | count[x] += gfs2_bitcount(rgd, | |
288 | bi->bi_bh->b_data + | |
289 | bi->bi_offset, | |
290 | bi->bi_len, x); | |
291 | } | |
292 | ||
cfc8b549 | 293 | if (count[0] != rgd->rd_free) { |
b3b94faa DT |
294 | if (gfs2_consist_rgrpd(rgd)) |
295 | fs_err(sdp, "free data mismatch: %u != %u\n", | |
cfc8b549 | 296 | count[0], rgd->rd_free); |
b3b94faa DT |
297 | return; |
298 | } | |
299 | ||
73f74948 | 300 | tmp = rgd->rd_data - rgd->rd_free - rgd->rd_dinodes; |
6b946170 | 301 | if (count[1] != tmp) { |
b3b94faa DT |
302 | if (gfs2_consist_rgrpd(rgd)) |
303 | fs_err(sdp, "used data mismatch: %u != %u\n", | |
304 | count[1], tmp); | |
305 | return; | |
306 | } | |
307 | ||
6b946170 | 308 | if (count[2] + count[3] != rgd->rd_dinodes) { |
b3b94faa | 309 | if (gfs2_consist_rgrpd(rgd)) |
feaa7bba | 310 | fs_err(sdp, "used metadata mismatch: %u != %u\n", |
6b946170 | 311 | count[2] + count[3], rgd->rd_dinodes); |
b3b94faa DT |
312 | return; |
313 | } | |
b3b94faa DT |
314 | } |
315 | ||
bb8d8a6f | 316 | static inline int rgrp_contains_block(struct gfs2_rgrpd *rgd, u64 block) |
b3b94faa | 317 | { |
bb8d8a6f SW |
318 | u64 first = rgd->rd_data0; |
319 | u64 last = first + rgd->rd_data; | |
16910427 | 320 | return first <= block && block < last; |
b3b94faa DT |
321 | } |
322 | ||
323 | /** | |
324 | * gfs2_blk2rgrpd - Find resource group for a given data/meta block number | |
325 | * @sdp: The GFS2 superblock | |
326 | * @n: The data block number | |
327 | * | |
328 | * Returns: The resource group, or NULL if not found | |
329 | */ | |
330 | ||
66fc061b | 331 | struct gfs2_rgrpd *gfs2_blk2rgrpd(struct gfs2_sbd *sdp, u64 blk, bool exact) |
b3b94faa | 332 | { |
66fc061b | 333 | struct rb_node *n, *next; |
f75bbfb4 | 334 | struct gfs2_rgrpd *cur; |
b3b94faa DT |
335 | |
336 | spin_lock(&sdp->sd_rindex_spin); | |
66fc061b SW |
337 | n = sdp->sd_rindex_tree.rb_node; |
338 | while (n) { | |
339 | cur = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
340 | next = NULL; | |
7c9ca621 | 341 | if (blk < cur->rd_addr) |
66fc061b | 342 | next = n->rb_left; |
f75bbfb4 | 343 | else if (blk >= cur->rd_data0 + cur->rd_data) |
66fc061b SW |
344 | next = n->rb_right; |
345 | if (next == NULL) { | |
b3b94faa | 346 | spin_unlock(&sdp->sd_rindex_spin); |
66fc061b SW |
347 | if (exact) { |
348 | if (blk < cur->rd_addr) | |
349 | return NULL; | |
350 | if (blk >= cur->rd_data0 + cur->rd_data) | |
351 | return NULL; | |
352 | } | |
7c9ca621 | 353 | return cur; |
b3b94faa | 354 | } |
66fc061b | 355 | n = next; |
b3b94faa | 356 | } |
b3b94faa DT |
357 | spin_unlock(&sdp->sd_rindex_spin); |
358 | ||
359 | return NULL; | |
360 | } | |
361 | ||
362 | /** | |
363 | * gfs2_rgrpd_get_first - get the first Resource Group in the filesystem | |
364 | * @sdp: The GFS2 superblock | |
365 | * | |
366 | * Returns: The first rgrp in the filesystem | |
367 | */ | |
368 | ||
369 | struct gfs2_rgrpd *gfs2_rgrpd_get_first(struct gfs2_sbd *sdp) | |
370 | { | |
7c9ca621 BP |
371 | const struct rb_node *n; |
372 | struct gfs2_rgrpd *rgd; | |
373 | ||
8339ee54 | 374 | spin_lock(&sdp->sd_rindex_spin); |
7c9ca621 BP |
375 | n = rb_first(&sdp->sd_rindex_tree); |
376 | rgd = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
8339ee54 | 377 | spin_unlock(&sdp->sd_rindex_spin); |
7c9ca621 BP |
378 | |
379 | return rgd; | |
b3b94faa DT |
380 | } |
381 | ||
382 | /** | |
383 | * gfs2_rgrpd_get_next - get the next RG | |
384 | * @rgd: A RG | |
385 | * | |
386 | * Returns: The next rgrp | |
387 | */ | |
388 | ||
389 | struct gfs2_rgrpd *gfs2_rgrpd_get_next(struct gfs2_rgrpd *rgd) | |
390 | { | |
7c9ca621 BP |
391 | struct gfs2_sbd *sdp = rgd->rd_sbd; |
392 | const struct rb_node *n; | |
393 | ||
394 | spin_lock(&sdp->sd_rindex_spin); | |
395 | n = rb_next(&rgd->rd_node); | |
396 | if (n == NULL) | |
397 | n = rb_first(&sdp->sd_rindex_tree); | |
398 | ||
399 | if (unlikely(&rgd->rd_node == n)) { | |
400 | spin_unlock(&sdp->sd_rindex_spin); | |
b3b94faa | 401 | return NULL; |
7c9ca621 BP |
402 | } |
403 | rgd = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
404 | spin_unlock(&sdp->sd_rindex_spin); | |
405 | return rgd; | |
b3b94faa DT |
406 | } |
407 | ||
8339ee54 SW |
408 | void gfs2_free_clones(struct gfs2_rgrpd *rgd) |
409 | { | |
410 | int x; | |
411 | ||
412 | for (x = 0; x < rgd->rd_length; x++) { | |
413 | struct gfs2_bitmap *bi = rgd->rd_bits + x; | |
414 | kfree(bi->bi_clone); | |
415 | bi->bi_clone = NULL; | |
416 | } | |
417 | } | |
418 | ||
419 | void gfs2_clear_rgrpd(struct gfs2_sbd *sdp) | |
b3b94faa | 420 | { |
7c9ca621 | 421 | struct rb_node *n; |
b3b94faa DT |
422 | struct gfs2_rgrpd *rgd; |
423 | struct gfs2_glock *gl; | |
424 | ||
7c9ca621 BP |
425 | while ((n = rb_first(&sdp->sd_rindex_tree))) { |
426 | rgd = rb_entry(n, struct gfs2_rgrpd, rd_node); | |
b3b94faa DT |
427 | gl = rgd->rd_gl; |
428 | ||
7c9ca621 | 429 | rb_erase(n, &sdp->sd_rindex_tree); |
b3b94faa DT |
430 | |
431 | if (gl) { | |
8339ee54 | 432 | spin_lock(&gl->gl_spin); |
5c676f6d | 433 | gl->gl_object = NULL; |
8339ee54 | 434 | spin_unlock(&gl->gl_spin); |
29687a2a | 435 | gfs2_glock_add_to_lru(gl); |
b3b94faa DT |
436 | gfs2_glock_put(gl); |
437 | } | |
438 | ||
8339ee54 | 439 | gfs2_free_clones(rgd); |
b3b94faa | 440 | kfree(rgd->rd_bits); |
6bdd9be6 | 441 | kmem_cache_free(gfs2_rgrpd_cachep, rgd); |
b3b94faa DT |
442 | } |
443 | } | |
444 | ||
bb8d8a6f SW |
445 | static void gfs2_rindex_print(const struct gfs2_rgrpd *rgd) |
446 | { | |
447 | printk(KERN_INFO " ri_addr = %llu\n", (unsigned long long)rgd->rd_addr); | |
448 | printk(KERN_INFO " ri_length = %u\n", rgd->rd_length); | |
449 | printk(KERN_INFO " ri_data0 = %llu\n", (unsigned long long)rgd->rd_data0); | |
450 | printk(KERN_INFO " ri_data = %u\n", rgd->rd_data); | |
451 | printk(KERN_INFO " ri_bitbytes = %u\n", rgd->rd_bitbytes); | |
452 | } | |
453 | ||
b3b94faa DT |
454 | /** |
455 | * gfs2_compute_bitstructs - Compute the bitmap sizes | |
456 | * @rgd: The resource group descriptor | |
457 | * | |
458 | * Calculates bitmap descriptors, one for each block that contains bitmap data | |
459 | * | |
460 | * Returns: errno | |
461 | */ | |
462 | ||
463 | static int compute_bitstructs(struct gfs2_rgrpd *rgd) | |
464 | { | |
465 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
466 | struct gfs2_bitmap *bi; | |
bb8d8a6f | 467 | u32 length = rgd->rd_length; /* # blocks in hdr & bitmap */ |
cd915493 | 468 | u32 bytes_left, bytes; |
b3b94faa DT |
469 | int x; |
470 | ||
feaa7bba SW |
471 | if (!length) |
472 | return -EINVAL; | |
473 | ||
dd894be8 | 474 | rgd->rd_bits = kcalloc(length, sizeof(struct gfs2_bitmap), GFP_NOFS); |
b3b94faa DT |
475 | if (!rgd->rd_bits) |
476 | return -ENOMEM; | |
477 | ||
bb8d8a6f | 478 | bytes_left = rgd->rd_bitbytes; |
b3b94faa DT |
479 | |
480 | for (x = 0; x < length; x++) { | |
481 | bi = rgd->rd_bits + x; | |
482 | ||
60a0b8f9 | 483 | bi->bi_flags = 0; |
b3b94faa DT |
484 | /* small rgrp; bitmap stored completely in header block */ |
485 | if (length == 1) { | |
486 | bytes = bytes_left; | |
487 | bi->bi_offset = sizeof(struct gfs2_rgrp); | |
488 | bi->bi_start = 0; | |
489 | bi->bi_len = bytes; | |
490 | /* header block */ | |
491 | } else if (x == 0) { | |
492 | bytes = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_rgrp); | |
493 | bi->bi_offset = sizeof(struct gfs2_rgrp); | |
494 | bi->bi_start = 0; | |
495 | bi->bi_len = bytes; | |
496 | /* last block */ | |
497 | } else if (x + 1 == length) { | |
498 | bytes = bytes_left; | |
499 | bi->bi_offset = sizeof(struct gfs2_meta_header); | |
bb8d8a6f | 500 | bi->bi_start = rgd->rd_bitbytes - bytes_left; |
b3b94faa DT |
501 | bi->bi_len = bytes; |
502 | /* other blocks */ | |
503 | } else { | |
568f4c96 SW |
504 | bytes = sdp->sd_sb.sb_bsize - |
505 | sizeof(struct gfs2_meta_header); | |
b3b94faa | 506 | bi->bi_offset = sizeof(struct gfs2_meta_header); |
bb8d8a6f | 507 | bi->bi_start = rgd->rd_bitbytes - bytes_left; |
b3b94faa DT |
508 | bi->bi_len = bytes; |
509 | } | |
510 | ||
511 | bytes_left -= bytes; | |
512 | } | |
513 | ||
514 | if (bytes_left) { | |
515 | gfs2_consist_rgrpd(rgd); | |
516 | return -EIO; | |
517 | } | |
518 | bi = rgd->rd_bits + (length - 1); | |
bb8d8a6f | 519 | if ((bi->bi_start + bi->bi_len) * GFS2_NBBY != rgd->rd_data) { |
b3b94faa | 520 | if (gfs2_consist_rgrpd(rgd)) { |
bb8d8a6f | 521 | gfs2_rindex_print(rgd); |
b3b94faa DT |
522 | fs_err(sdp, "start=%u len=%u offset=%u\n", |
523 | bi->bi_start, bi->bi_len, bi->bi_offset); | |
524 | } | |
525 | return -EIO; | |
526 | } | |
527 | ||
528 | return 0; | |
529 | } | |
530 | ||
7ae8fa84 RP |
531 | /** |
532 | * gfs2_ri_total - Total up the file system space, according to the rindex. | |
533 | * | |
534 | */ | |
535 | u64 gfs2_ri_total(struct gfs2_sbd *sdp) | |
536 | { | |
537 | u64 total_data = 0; | |
538 | struct inode *inode = sdp->sd_rindex; | |
539 | struct gfs2_inode *ip = GFS2_I(inode); | |
7ae8fa84 RP |
540 | char buf[sizeof(struct gfs2_rindex)]; |
541 | struct file_ra_state ra_state; | |
542 | int error, rgrps; | |
543 | ||
7ae8fa84 RP |
544 | file_ra_state_init(&ra_state, inode->i_mapping); |
545 | for (rgrps = 0;; rgrps++) { | |
546 | loff_t pos = rgrps * sizeof(struct gfs2_rindex); | |
547 | ||
bcd7278d | 548 | if (pos + sizeof(struct gfs2_rindex) > i_size_read(inode)) |
7ae8fa84 RP |
549 | break; |
550 | error = gfs2_internal_read(ip, &ra_state, buf, &pos, | |
551 | sizeof(struct gfs2_rindex)); | |
552 | if (error != sizeof(struct gfs2_rindex)) | |
553 | break; | |
bb8d8a6f | 554 | total_data += be32_to_cpu(((struct gfs2_rindex *)buf)->ri_data); |
7ae8fa84 | 555 | } |
7ae8fa84 RP |
556 | return total_data; |
557 | } | |
558 | ||
6aad1c3d | 559 | static int rgd_insert(struct gfs2_rgrpd *rgd) |
7c9ca621 BP |
560 | { |
561 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
562 | struct rb_node **newn = &sdp->sd_rindex_tree.rb_node, *parent = NULL; | |
563 | ||
564 | /* Figure out where to put new node */ | |
565 | while (*newn) { | |
566 | struct gfs2_rgrpd *cur = rb_entry(*newn, struct gfs2_rgrpd, | |
567 | rd_node); | |
568 | ||
569 | parent = *newn; | |
570 | if (rgd->rd_addr < cur->rd_addr) | |
571 | newn = &((*newn)->rb_left); | |
572 | else if (rgd->rd_addr > cur->rd_addr) | |
573 | newn = &((*newn)->rb_right); | |
574 | else | |
6aad1c3d | 575 | return -EEXIST; |
7c9ca621 BP |
576 | } |
577 | ||
578 | rb_link_node(&rgd->rd_node, parent, newn); | |
579 | rb_insert_color(&rgd->rd_node, &sdp->sd_rindex_tree); | |
6aad1c3d BP |
580 | sdp->sd_rgrps++; |
581 | return 0; | |
7c9ca621 BP |
582 | } |
583 | ||
b3b94faa | 584 | /** |
6c53267f | 585 | * read_rindex_entry - Pull in a new resource index entry from the disk |
b3b94faa DT |
586 | * @gl: The glock covering the rindex inode |
587 | * | |
8339ee54 | 588 | * Returns: 0 on success, > 0 on EOF, error code otherwise |
6c53267f RP |
589 | */ |
590 | ||
591 | static int read_rindex_entry(struct gfs2_inode *ip, | |
592 | struct file_ra_state *ra_state) | |
593 | { | |
594 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
595 | loff_t pos = sdp->sd_rgrps * sizeof(struct gfs2_rindex); | |
8339ee54 | 596 | struct gfs2_rindex buf; |
6c53267f RP |
597 | int error; |
598 | struct gfs2_rgrpd *rgd; | |
599 | ||
8339ee54 SW |
600 | if (pos >= i_size_read(&ip->i_inode)) |
601 | return 1; | |
602 | ||
603 | error = gfs2_internal_read(ip, ra_state, (char *)&buf, &pos, | |
6c53267f | 604 | sizeof(struct gfs2_rindex)); |
8339ee54 SW |
605 | |
606 | if (error != sizeof(struct gfs2_rindex)) | |
607 | return (error == 0) ? 1 : error; | |
6c53267f | 608 | |
6bdd9be6 | 609 | rgd = kmem_cache_zalloc(gfs2_rgrpd_cachep, GFP_NOFS); |
6c53267f RP |
610 | error = -ENOMEM; |
611 | if (!rgd) | |
612 | return error; | |
613 | ||
6c53267f | 614 | rgd->rd_sbd = sdp; |
8339ee54 SW |
615 | rgd->rd_addr = be64_to_cpu(buf.ri_addr); |
616 | rgd->rd_length = be32_to_cpu(buf.ri_length); | |
617 | rgd->rd_data0 = be64_to_cpu(buf.ri_data0); | |
618 | rgd->rd_data = be32_to_cpu(buf.ri_data); | |
619 | rgd->rd_bitbytes = be32_to_cpu(buf.ri_bitbytes); | |
7c9ca621 | 620 | |
6c53267f RP |
621 | error = compute_bitstructs(rgd); |
622 | if (error) | |
8339ee54 | 623 | goto fail; |
6c53267f | 624 | |
bb8d8a6f | 625 | error = gfs2_glock_get(sdp, rgd->rd_addr, |
6c53267f RP |
626 | &gfs2_rgrp_glops, CREATE, &rgd->rd_gl); |
627 | if (error) | |
8339ee54 | 628 | goto fail; |
6c53267f RP |
629 | |
630 | rgd->rd_gl->gl_object = rgd; | |
cf45b752 | 631 | rgd->rd_flags &= ~GFS2_RDF_UPTODATE; |
7c9ca621 BP |
632 | if (rgd->rd_data > sdp->sd_max_rg_data) |
633 | sdp->sd_max_rg_data = rgd->rd_data; | |
8339ee54 | 634 | spin_lock(&sdp->sd_rindex_spin); |
6aad1c3d | 635 | error = rgd_insert(rgd); |
8339ee54 | 636 | spin_unlock(&sdp->sd_rindex_spin); |
6aad1c3d BP |
637 | if (!error) |
638 | return 0; | |
639 | ||
640 | error = 0; /* someone else read in the rgrp; free it and ignore it */ | |
c1ac539e | 641 | gfs2_glock_put(rgd->rd_gl); |
8339ee54 SW |
642 | |
643 | fail: | |
644 | kfree(rgd->rd_bits); | |
645 | kmem_cache_free(gfs2_rgrpd_cachep, rgd); | |
6c53267f RP |
646 | return error; |
647 | } | |
648 | ||
649 | /** | |
650 | * gfs2_ri_update - Pull in a new resource index from the disk | |
651 | * @ip: pointer to the rindex inode | |
652 | * | |
b3b94faa DT |
653 | * Returns: 0 on successful update, error code otherwise |
654 | */ | |
655 | ||
8339ee54 | 656 | static int gfs2_ri_update(struct gfs2_inode *ip) |
b3b94faa | 657 | { |
feaa7bba SW |
658 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
659 | struct inode *inode = &ip->i_inode; | |
f42faf4f | 660 | struct file_ra_state ra_state; |
b3b94faa DT |
661 | int error; |
662 | ||
f42faf4f | 663 | file_ra_state_init(&ra_state, inode->i_mapping); |
8339ee54 | 664 | do { |
6c53267f | 665 | error = read_rindex_entry(ip, &ra_state); |
8339ee54 SW |
666 | } while (error == 0); |
667 | ||
668 | if (error < 0) | |
669 | return error; | |
b3b94faa | 670 | |
cf45b752 | 671 | sdp->sd_rindex_uptodate = 1; |
6c53267f RP |
672 | return 0; |
673 | } | |
b3b94faa | 674 | |
b3b94faa | 675 | /** |
8339ee54 | 676 | * gfs2_rindex_update - Update the rindex if required |
b3b94faa | 677 | * @sdp: The GFS2 superblock |
b3b94faa DT |
678 | * |
679 | * We grab a lock on the rindex inode to make sure that it doesn't | |
680 | * change whilst we are performing an operation. We keep this lock | |
681 | * for quite long periods of time compared to other locks. This | |
682 | * doesn't matter, since it is shared and it is very, very rarely | |
683 | * accessed in the exclusive mode (i.e. only when expanding the filesystem). | |
684 | * | |
685 | * This makes sure that we're using the latest copy of the resource index | |
686 | * special file, which might have been updated if someone expanded the | |
687 | * filesystem (via gfs2_grow utility), which adds new resource groups. | |
688 | * | |
8339ee54 | 689 | * Returns: 0 on succeess, error code otherwise |
b3b94faa DT |
690 | */ |
691 | ||
8339ee54 | 692 | int gfs2_rindex_update(struct gfs2_sbd *sdp) |
b3b94faa | 693 | { |
feaa7bba | 694 | struct gfs2_inode *ip = GFS2_I(sdp->sd_rindex); |
b3b94faa | 695 | struct gfs2_glock *gl = ip->i_gl; |
8339ee54 SW |
696 | struct gfs2_holder ri_gh; |
697 | int error = 0; | |
a365fbf3 | 698 | int unlock_required = 0; |
b3b94faa DT |
699 | |
700 | /* Read new copy from disk if we don't have the latest */ | |
cf45b752 | 701 | if (!sdp->sd_rindex_uptodate) { |
a365fbf3 SW |
702 | if (!gfs2_glock_is_locked_by_me(gl)) { |
703 | error = gfs2_glock_nq_init(gl, LM_ST_SHARED, 0, &ri_gh); | |
704 | if (error) | |
6aad1c3d | 705 | return error; |
a365fbf3 SW |
706 | unlock_required = 1; |
707 | } | |
8339ee54 | 708 | if (!sdp->sd_rindex_uptodate) |
b3b94faa | 709 | error = gfs2_ri_update(ip); |
a365fbf3 SW |
710 | if (unlock_required) |
711 | gfs2_glock_dq_uninit(&ri_gh); | |
b3b94faa DT |
712 | } |
713 | ||
714 | return error; | |
715 | } | |
716 | ||
42d52e38 | 717 | static void gfs2_rgrp_in(struct gfs2_rgrpd *rgd, const void *buf) |
bb8d8a6f SW |
718 | { |
719 | const struct gfs2_rgrp *str = buf; | |
42d52e38 | 720 | u32 rg_flags; |
bb8d8a6f | 721 | |
42d52e38 | 722 | rg_flags = be32_to_cpu(str->rg_flags); |
09010978 | 723 | rg_flags &= ~GFS2_RDF_MASK; |
1ce97e56 SW |
724 | rgd->rd_flags &= GFS2_RDF_MASK; |
725 | rgd->rd_flags |= rg_flags; | |
cfc8b549 | 726 | rgd->rd_free = be32_to_cpu(str->rg_free); |
73f74948 | 727 | rgd->rd_dinodes = be32_to_cpu(str->rg_dinodes); |
d8b71f73 | 728 | rgd->rd_igeneration = be64_to_cpu(str->rg_igeneration); |
bb8d8a6f SW |
729 | } |
730 | ||
42d52e38 | 731 | static void gfs2_rgrp_out(struct gfs2_rgrpd *rgd, void *buf) |
bb8d8a6f SW |
732 | { |
733 | struct gfs2_rgrp *str = buf; | |
734 | ||
09010978 | 735 | str->rg_flags = cpu_to_be32(rgd->rd_flags & ~GFS2_RDF_MASK); |
cfc8b549 | 736 | str->rg_free = cpu_to_be32(rgd->rd_free); |
73f74948 | 737 | str->rg_dinodes = cpu_to_be32(rgd->rd_dinodes); |
bb8d8a6f | 738 | str->__pad = cpu_to_be32(0); |
d8b71f73 | 739 | str->rg_igeneration = cpu_to_be64(rgd->rd_igeneration); |
bb8d8a6f SW |
740 | memset(&str->rg_reserved, 0, sizeof(str->rg_reserved)); |
741 | } | |
742 | ||
b3b94faa | 743 | /** |
7c9ca621 | 744 | * gfs2_rgrp_go_lock - Read in a RG's header and bitmaps |
b3b94faa DT |
745 | * @rgd: the struct gfs2_rgrpd describing the RG to read in |
746 | * | |
747 | * Read in all of a Resource Group's header and bitmap blocks. | |
748 | * Caller must eventually call gfs2_rgrp_relse() to free the bitmaps. | |
749 | * | |
750 | * Returns: errno | |
751 | */ | |
752 | ||
7c9ca621 | 753 | int gfs2_rgrp_go_lock(struct gfs2_holder *gh) |
b3b94faa | 754 | { |
7c9ca621 | 755 | struct gfs2_rgrpd *rgd = gh->gh_gl->gl_object; |
b3b94faa DT |
756 | struct gfs2_sbd *sdp = rgd->rd_sbd; |
757 | struct gfs2_glock *gl = rgd->rd_gl; | |
bb8d8a6f | 758 | unsigned int length = rgd->rd_length; |
b3b94faa DT |
759 | struct gfs2_bitmap *bi; |
760 | unsigned int x, y; | |
761 | int error; | |
762 | ||
b3b94faa DT |
763 | for (x = 0; x < length; x++) { |
764 | bi = rgd->rd_bits + x; | |
bb8d8a6f | 765 | error = gfs2_meta_read(gl, rgd->rd_addr + x, 0, &bi->bi_bh); |
b3b94faa DT |
766 | if (error) |
767 | goto fail; | |
768 | } | |
769 | ||
770 | for (y = length; y--;) { | |
771 | bi = rgd->rd_bits + y; | |
7276b3b0 | 772 | error = gfs2_meta_wait(sdp, bi->bi_bh); |
b3b94faa DT |
773 | if (error) |
774 | goto fail; | |
feaa7bba | 775 | if (gfs2_metatype_check(sdp, bi->bi_bh, y ? GFS2_METATYPE_RB : |
b3b94faa DT |
776 | GFS2_METATYPE_RG)) { |
777 | error = -EIO; | |
778 | goto fail; | |
779 | } | |
780 | } | |
781 | ||
cf45b752 | 782 | if (!(rgd->rd_flags & GFS2_RDF_UPTODATE)) { |
60a0b8f9 SW |
783 | for (x = 0; x < length; x++) |
784 | clear_bit(GBF_FULL, &rgd->rd_bits[x].bi_flags); | |
42d52e38 | 785 | gfs2_rgrp_in(rgd, (rgd->rd_bits[0].bi_bh)->b_data); |
1ce97e56 | 786 | rgd->rd_flags |= (GFS2_RDF_UPTODATE | GFS2_RDF_CHECK); |
7c9ca621 | 787 | rgd->rd_free_clone = rgd->rd_free; |
b3b94faa DT |
788 | } |
789 | ||
b3b94faa DT |
790 | return 0; |
791 | ||
feaa7bba | 792 | fail: |
b3b94faa DT |
793 | while (x--) { |
794 | bi = rgd->rd_bits + x; | |
795 | brelse(bi->bi_bh); | |
796 | bi->bi_bh = NULL; | |
797 | gfs2_assert_warn(sdp, !bi->bi_clone); | |
798 | } | |
b3b94faa DT |
799 | |
800 | return error; | |
801 | } | |
802 | ||
b3b94faa | 803 | /** |
7c9ca621 | 804 | * gfs2_rgrp_go_unlock - Release RG bitmaps read in with gfs2_rgrp_bh_get() |
b3b94faa DT |
805 | * @rgd: the struct gfs2_rgrpd describing the RG to read in |
806 | * | |
807 | */ | |
808 | ||
7c9ca621 | 809 | void gfs2_rgrp_go_unlock(struct gfs2_holder *gh) |
b3b94faa | 810 | { |
7c9ca621 | 811 | struct gfs2_rgrpd *rgd = gh->gh_gl->gl_object; |
bb8d8a6f | 812 | int x, length = rgd->rd_length; |
b3b94faa | 813 | |
b3b94faa DT |
814 | for (x = 0; x < length; x++) { |
815 | struct gfs2_bitmap *bi = rgd->rd_bits + x; | |
b3b94faa DT |
816 | brelse(bi->bi_bh); |
817 | bi->bi_bh = NULL; | |
818 | } | |
819 | ||
b3b94faa DT |
820 | } |
821 | ||
66fc061b | 822 | int gfs2_rgrp_send_discards(struct gfs2_sbd *sdp, u64 offset, |
7c9ca621 | 823 | struct buffer_head *bh, |
66fc061b | 824 | const struct gfs2_bitmap *bi, unsigned minlen, u64 *ptrimmed) |
f15ab561 SW |
825 | { |
826 | struct super_block *sb = sdp->sd_vfs; | |
827 | struct block_device *bdev = sb->s_bdev; | |
828 | const unsigned int sects_per_blk = sdp->sd_sb.sb_bsize / | |
e1defc4f | 829 | bdev_logical_block_size(sb->s_bdev); |
f15ab561 | 830 | u64 blk; |
64d576ba | 831 | sector_t start = 0; |
f15ab561 SW |
832 | sector_t nr_sects = 0; |
833 | int rv; | |
834 | unsigned int x; | |
66fc061b SW |
835 | u32 trimmed = 0; |
836 | u8 diff; | |
f15ab561 SW |
837 | |
838 | for (x = 0; x < bi->bi_len; x++) { | |
66fc061b SW |
839 | const u8 *clone = bi->bi_clone ? bi->bi_clone : bi->bi_bh->b_data; |
840 | clone += bi->bi_offset; | |
841 | clone += x; | |
842 | if (bh) { | |
843 | const u8 *orig = bh->b_data + bi->bi_offset + x; | |
844 | diff = ~(*orig | (*orig >> 1)) & (*clone | (*clone >> 1)); | |
845 | } else { | |
846 | diff = ~(*clone | (*clone >> 1)); | |
847 | } | |
f15ab561 SW |
848 | diff &= 0x55; |
849 | if (diff == 0) | |
850 | continue; | |
851 | blk = offset + ((bi->bi_start + x) * GFS2_NBBY); | |
852 | blk *= sects_per_blk; /* convert to sectors */ | |
853 | while(diff) { | |
854 | if (diff & 1) { | |
855 | if (nr_sects == 0) | |
856 | goto start_new_extent; | |
857 | if ((start + nr_sects) != blk) { | |
66fc061b SW |
858 | if (nr_sects >= minlen) { |
859 | rv = blkdev_issue_discard(bdev, | |
860 | start, nr_sects, | |
861 | GFP_NOFS, 0); | |
862 | if (rv) | |
863 | goto fail; | |
864 | trimmed += nr_sects; | |
865 | } | |
f15ab561 SW |
866 | nr_sects = 0; |
867 | start_new_extent: | |
868 | start = blk; | |
869 | } | |
870 | nr_sects += sects_per_blk; | |
871 | } | |
872 | diff >>= 2; | |
873 | blk += sects_per_blk; | |
874 | } | |
875 | } | |
66fc061b | 876 | if (nr_sects >= minlen) { |
dd3932ed | 877 | rv = blkdev_issue_discard(bdev, start, nr_sects, GFP_NOFS, 0); |
f15ab561 SW |
878 | if (rv) |
879 | goto fail; | |
66fc061b | 880 | trimmed += nr_sects; |
f15ab561 | 881 | } |
66fc061b SW |
882 | if (ptrimmed) |
883 | *ptrimmed = trimmed; | |
884 | return 0; | |
885 | ||
f15ab561 | 886 | fail: |
66fc061b SW |
887 | if (sdp->sd_args.ar_discard) |
888 | fs_warn(sdp, "error %d on discard request, turning discards off for this filesystem", rv); | |
f15ab561 | 889 | sdp->sd_args.ar_discard = 0; |
66fc061b SW |
890 | return -EIO; |
891 | } | |
892 | ||
893 | /** | |
894 | * gfs2_fitrim - Generate discard requests for unused bits of the filesystem | |
895 | * @filp: Any file on the filesystem | |
896 | * @argp: Pointer to the arguments (also used to pass result) | |
897 | * | |
898 | * Returns: 0 on success, otherwise error code | |
899 | */ | |
900 | ||
901 | int gfs2_fitrim(struct file *filp, void __user *argp) | |
902 | { | |
903 | struct inode *inode = filp->f_dentry->d_inode; | |
904 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
905 | struct request_queue *q = bdev_get_queue(sdp->sd_vfs->s_bdev); | |
906 | struct buffer_head *bh; | |
907 | struct gfs2_rgrpd *rgd; | |
908 | struct gfs2_rgrpd *rgd_end; | |
909 | struct gfs2_holder gh; | |
910 | struct fstrim_range r; | |
911 | int ret = 0; | |
912 | u64 amt; | |
913 | u64 trimmed = 0; | |
914 | unsigned int x; | |
915 | ||
916 | if (!capable(CAP_SYS_ADMIN)) | |
917 | return -EPERM; | |
918 | ||
919 | if (!blk_queue_discard(q)) | |
920 | return -EOPNOTSUPP; | |
921 | ||
66fc061b SW |
922 | if (argp == NULL) { |
923 | r.start = 0; | |
924 | r.len = ULLONG_MAX; | |
925 | r.minlen = 0; | |
926 | } else if (copy_from_user(&r, argp, sizeof(r))) | |
927 | return -EFAULT; | |
928 | ||
5e2f7d61 BP |
929 | ret = gfs2_rindex_update(sdp); |
930 | if (ret) | |
931 | return ret; | |
932 | ||
66fc061b SW |
933 | rgd = gfs2_blk2rgrpd(sdp, r.start, 0); |
934 | rgd_end = gfs2_blk2rgrpd(sdp, r.start + r.len, 0); | |
935 | ||
936 | while (1) { | |
937 | ||
938 | ret = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, 0, &gh); | |
939 | if (ret) | |
940 | goto out; | |
941 | ||
942 | if (!(rgd->rd_flags & GFS2_RGF_TRIMMED)) { | |
943 | /* Trim each bitmap in the rgrp */ | |
944 | for (x = 0; x < rgd->rd_length; x++) { | |
945 | struct gfs2_bitmap *bi = rgd->rd_bits + x; | |
946 | ret = gfs2_rgrp_send_discards(sdp, rgd->rd_data0, NULL, bi, r.minlen, &amt); | |
947 | if (ret) { | |
948 | gfs2_glock_dq_uninit(&gh); | |
949 | goto out; | |
950 | } | |
951 | trimmed += amt; | |
952 | } | |
953 | ||
954 | /* Mark rgrp as having been trimmed */ | |
955 | ret = gfs2_trans_begin(sdp, RES_RG_HDR, 0); | |
956 | if (ret == 0) { | |
957 | bh = rgd->rd_bits[0].bi_bh; | |
958 | rgd->rd_flags |= GFS2_RGF_TRIMMED; | |
959 | gfs2_trans_add_bh(rgd->rd_gl, bh, 1); | |
960 | gfs2_rgrp_out(rgd, bh->b_data); | |
961 | gfs2_trans_end(sdp); | |
962 | } | |
963 | } | |
964 | gfs2_glock_dq_uninit(&gh); | |
965 | ||
966 | if (rgd == rgd_end) | |
967 | break; | |
968 | ||
969 | rgd = gfs2_rgrpd_get_next(rgd); | |
970 | } | |
971 | ||
972 | out: | |
973 | r.len = trimmed << 9; | |
974 | if (argp && copy_to_user(argp, &r, sizeof(r))) | |
975 | return -EFAULT; | |
976 | ||
977 | return ret; | |
f15ab561 SW |
978 | } |
979 | ||
b3b94faa | 980 | /** |
564e12b1 | 981 | * gfs2_qadata_get - get the struct gfs2_qadata structure for an inode |
b3b94faa DT |
982 | * @ip: the incore GFS2 inode structure |
983 | * | |
564e12b1 | 984 | * Returns: the struct gfs2_qadata |
b3b94faa DT |
985 | */ |
986 | ||
564e12b1 | 987 | struct gfs2_qadata *gfs2_qadata_get(struct gfs2_inode *ip) |
b3b94faa | 988 | { |
8339ee54 SW |
989 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
990 | int error; | |
564e12b1 BP |
991 | BUG_ON(ip->i_qadata != NULL); |
992 | ip->i_qadata = kzalloc(sizeof(struct gfs2_qadata), GFP_NOFS); | |
8339ee54 SW |
993 | error = gfs2_rindex_update(sdp); |
994 | if (error) | |
995 | fs_warn(sdp, "rindex update returns %d\n", error); | |
564e12b1 BP |
996 | return ip->i_qadata; |
997 | } | |
998 | ||
999 | /** | |
1000 | * gfs2_blkrsv_get - get the struct gfs2_blkreserv structure for an inode | |
1001 | * @ip: the incore GFS2 inode structure | |
1002 | * | |
1003 | * Returns: the struct gfs2_qadata | |
1004 | */ | |
1005 | ||
36f5580b | 1006 | static int gfs2_blkrsv_get(struct gfs2_inode *ip) |
564e12b1 BP |
1007 | { |
1008 | BUG_ON(ip->i_res != NULL); | |
36f5580b BP |
1009 | ip->i_res = kmem_cache_zalloc(gfs2_rsrv_cachep, GFP_NOFS); |
1010 | if (!ip->i_res) | |
1011 | return -ENOMEM; | |
1012 | return 0; | |
b3b94faa DT |
1013 | } |
1014 | ||
b3b94faa DT |
1015 | /** |
1016 | * try_rgrp_fit - See if a given reservation will fit in a given RG | |
1017 | * @rgd: the RG data | |
54335b1f | 1018 | * @ip: the inode |
b3b94faa DT |
1019 | * |
1020 | * If there's room for the requested blocks to be allocated from the RG: | |
b3b94faa DT |
1021 | * |
1022 | * Returns: 1 on success (it fits), 0 on failure (it doesn't fit) | |
1023 | */ | |
1024 | ||
54335b1f | 1025 | static int try_rgrp_fit(const struct gfs2_rgrpd *rgd, const struct gfs2_inode *ip) |
b3b94faa | 1026 | { |
564e12b1 | 1027 | const struct gfs2_blkreserv *rs = ip->i_res; |
54335b1f | 1028 | |
09010978 | 1029 | if (rgd->rd_flags & (GFS2_RGF_NOALLOC | GFS2_RDF_ERROR)) |
a43a4906 | 1030 | return 0; |
564e12b1 | 1031 | if (rgd->rd_free_clone >= rs->rs_requested) |
7c9ca621 | 1032 | return 1; |
7c9ca621 | 1033 | return 0; |
b3b94faa DT |
1034 | } |
1035 | ||
b3e47ca0 BP |
1036 | static inline u32 gfs2_bi2rgd_blk(struct gfs2_bitmap *bi, u32 blk) |
1037 | { | |
1038 | return (bi->bi_start * GFS2_NBBY) + blk; | |
1039 | } | |
1040 | ||
c8cdf479 SW |
1041 | /** |
1042 | * try_rgrp_unlink - Look for any unlinked, allocated, but unused inodes | |
1043 | * @rgd: The rgrp | |
1044 | * | |
1a0eae88 BP |
1045 | * Returns: 0 if no error |
1046 | * The inode, if one has been found, in inode. | |
c8cdf479 SW |
1047 | */ |
1048 | ||
044b9414 | 1049 | static void try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked, u64 skip) |
c8cdf479 | 1050 | { |
6760bdcd | 1051 | u32 goal = 0, block; |
bb9bcf06 | 1052 | u64 no_addr; |
5f3eae75 | 1053 | struct gfs2_sbd *sdp = rgd->rd_sbd; |
044b9414 SW |
1054 | struct gfs2_glock *gl; |
1055 | struct gfs2_inode *ip; | |
1056 | int error; | |
1057 | int found = 0; | |
b3e47ca0 | 1058 | struct gfs2_bitmap *bi; |
c8cdf479 | 1059 | |
044b9414 | 1060 | while (goal < rgd->rd_data) { |
5f3eae75 | 1061 | down_write(&sdp->sd_log_flush_lock); |
6a8099ed | 1062 | block = rgblk_search(rgd, goal, GFS2_BLKST_UNLINKED, &bi); |
5f3eae75 | 1063 | up_write(&sdp->sd_log_flush_lock); |
6760bdcd | 1064 | if (block == BFITNOENT) |
24c73873 | 1065 | break; |
b3e47ca0 BP |
1066 | |
1067 | block = gfs2_bi2rgd_blk(bi, block); | |
6760bdcd BP |
1068 | /* rgblk_search can return a block < goal, so we need to |
1069 | keep it marching forward. */ | |
1070 | no_addr = block + rgd->rd_data0; | |
44ad37d6 | 1071 | goal = max(block + 1, goal + 1); |
6760bdcd | 1072 | if (*last_unlinked != NO_BLOCK && no_addr <= *last_unlinked) |
c8cdf479 | 1073 | continue; |
1e19a195 SW |
1074 | if (no_addr == skip) |
1075 | continue; | |
bb9bcf06 | 1076 | *last_unlinked = no_addr; |
044b9414 SW |
1077 | |
1078 | error = gfs2_glock_get(sdp, no_addr, &gfs2_inode_glops, CREATE, &gl); | |
1079 | if (error) | |
1080 | continue; | |
1081 | ||
1082 | /* If the inode is already in cache, we can ignore it here | |
1083 | * because the existing inode disposal code will deal with | |
1084 | * it when all refs have gone away. Accessing gl_object like | |
1085 | * this is not safe in general. Here it is ok because we do | |
1086 | * not dereference the pointer, and we only need an approx | |
1087 | * answer to whether it is NULL or not. | |
1088 | */ | |
1089 | ip = gl->gl_object; | |
1090 | ||
1091 | if (ip || queue_work(gfs2_delete_workqueue, &gl->gl_delete) == 0) | |
1092 | gfs2_glock_put(gl); | |
1093 | else | |
1094 | found++; | |
1095 | ||
1096 | /* Limit reclaim to sensible number of tasks */ | |
44ad37d6 | 1097 | if (found > NR_CPUS) |
044b9414 | 1098 | return; |
c8cdf479 SW |
1099 | } |
1100 | ||
1101 | rgd->rd_flags &= ~GFS2_RDF_CHECK; | |
044b9414 | 1102 | return; |
c8cdf479 SW |
1103 | } |
1104 | ||
b3b94faa DT |
1105 | /** |
1106 | * get_local_rgrp - Choose and lock a rgrp for allocation | |
1107 | * @ip: the inode to reserve space for | |
1108 | * @rgp: the chosen and locked rgrp | |
1109 | * | |
1110 | * Try to acquire rgrp in way which avoids contending with others. | |
1111 | * | |
1112 | * Returns: errno | |
1113 | */ | |
1114 | ||
044b9414 | 1115 | static int get_local_rgrp(struct gfs2_inode *ip, u64 *last_unlinked) |
b3b94faa | 1116 | { |
feaa7bba | 1117 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa | 1118 | struct gfs2_rgrpd *rgd, *begin = NULL; |
564e12b1 | 1119 | struct gfs2_blkreserv *rs = ip->i_res; |
c688b8b3 | 1120 | int error, rg_locked, flags = LM_FLAG_TRY; |
7c9ca621 | 1121 | int loops = 0; |
b3b94faa | 1122 | |
54335b1f SW |
1123 | if (ip->i_rgd && rgrp_contains_block(ip->i_rgd, ip->i_goal)) |
1124 | rgd = begin = ip->i_rgd; | |
1125 | else | |
66fc061b | 1126 | rgd = begin = gfs2_blk2rgrpd(sdp, ip->i_goal, 1); |
b3b94faa | 1127 | |
7c9ca621 BP |
1128 | if (rgd == NULL) |
1129 | return -EBADSLT; | |
1130 | ||
1131 | while (loops < 3) { | |
292c8c14 AD |
1132 | rg_locked = 0; |
1133 | ||
1134 | if (gfs2_glock_is_locked_by_me(rgd->rd_gl)) { | |
1135 | rg_locked = 1; | |
1136 | error = 0; | |
1137 | } else { | |
1138 | error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_EXCLUSIVE, | |
564e12b1 | 1139 | flags, &rs->rs_rgd_gh); |
292c8c14 | 1140 | } |
b3b94faa DT |
1141 | switch (error) { |
1142 | case 0: | |
54335b1f SW |
1143 | if (try_rgrp_fit(rgd, ip)) { |
1144 | ip->i_rgd = rgd; | |
7c9ca621 | 1145 | return 0; |
54335b1f | 1146 | } |
044b9414 SW |
1147 | if (rgd->rd_flags & GFS2_RDF_CHECK) |
1148 | try_rgrp_unlink(rgd, last_unlinked, ip->i_no_addr); | |
292c8c14 | 1149 | if (!rg_locked) |
564e12b1 | 1150 | gfs2_glock_dq_uninit(&rs->rs_rgd_gh); |
9cabcdbd | 1151 | /* fall through */ |
b3b94faa | 1152 | case GLR_TRYFAILED: |
7c9ca621 | 1153 | rgd = gfs2_rgrpd_get_next(rgd); |
c688b8b3 BP |
1154 | if (rgd == begin) { |
1155 | flags = 0; | |
7c9ca621 | 1156 | loops++; |
c688b8b3 | 1157 | } |
b3b94faa | 1158 | break; |
b3b94faa | 1159 | default: |
cc0581bd | 1160 | return error; |
b3b94faa | 1161 | } |
b3b94faa DT |
1162 | } |
1163 | ||
7c9ca621 | 1164 | return -ENOSPC; |
b3b94faa DT |
1165 | } |
1166 | ||
564e12b1 BP |
1167 | static void gfs2_blkrsv_put(struct gfs2_inode *ip) |
1168 | { | |
1169 | BUG_ON(ip->i_res == NULL); | |
36f5580b | 1170 | kmem_cache_free(gfs2_rsrv_cachep, ip->i_res); |
564e12b1 BP |
1171 | ip->i_res = NULL; |
1172 | } | |
1173 | ||
b3b94faa | 1174 | /** |
9ae32429 | 1175 | * gfs2_inplace_reserve - Reserve space in the filesystem |
b3b94faa DT |
1176 | * @ip: the inode to reserve space for |
1177 | * | |
1178 | * Returns: errno | |
1179 | */ | |
1180 | ||
564e12b1 | 1181 | int gfs2_inplace_reserve(struct gfs2_inode *ip, u32 requested) |
b3b94faa | 1182 | { |
feaa7bba | 1183 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
564e12b1 | 1184 | struct gfs2_blkreserv *rs; |
36f5580b | 1185 | int error; |
044b9414 SW |
1186 | u64 last_unlinked = NO_BLOCK; |
1187 | int tries = 0; | |
b3b94faa | 1188 | |
36f5580b BP |
1189 | error = gfs2_blkrsv_get(ip); |
1190 | if (error) | |
1191 | return error; | |
564e12b1 | 1192 | |
36f5580b | 1193 | rs = ip->i_res; |
564e12b1 BP |
1194 | rs->rs_requested = requested; |
1195 | if (gfs2_assert_warn(sdp, requested)) { | |
1196 | error = -EINVAL; | |
1197 | goto out; | |
1198 | } | |
b3b94faa | 1199 | |
044b9414 SW |
1200 | do { |
1201 | error = get_local_rgrp(ip, &last_unlinked); | |
54335b1f SW |
1202 | if (error != -ENOSPC) |
1203 | break; | |
1204 | /* Check that fs hasn't grown if writing to rindex */ | |
1205 | if (ip == GFS2_I(sdp->sd_rindex) && !sdp->sd_rindex_uptodate) { | |
1206 | error = gfs2_ri_update(ip); | |
1207 | if (error) | |
1208 | break; | |
1209 | continue; | |
0489b3f5 | 1210 | } |
54335b1f SW |
1211 | /* Flushing the log may release space */ |
1212 | gfs2_log_flush(sdp, NULL); | |
1213 | } while (tries++ < 3); | |
b3b94faa | 1214 | |
564e12b1 BP |
1215 | out: |
1216 | if (error) | |
1217 | gfs2_blkrsv_put(ip); | |
9ae32429 | 1218 | return error; |
b3b94faa DT |
1219 | } |
1220 | ||
1221 | /** | |
1222 | * gfs2_inplace_release - release an inplace reservation | |
1223 | * @ip: the inode the reservation was taken out on | |
1224 | * | |
1225 | * Release a reservation made by gfs2_inplace_reserve(). | |
1226 | */ | |
1227 | ||
1228 | void gfs2_inplace_release(struct gfs2_inode *ip) | |
1229 | { | |
564e12b1 | 1230 | struct gfs2_blkreserv *rs = ip->i_res; |
b3b94faa | 1231 | |
564e12b1 BP |
1232 | if (rs->rs_rgd_gh.gh_gl) |
1233 | gfs2_glock_dq_uninit(&rs->rs_rgd_gh); | |
49528b4e | 1234 | gfs2_blkrsv_put(ip); |
b3b94faa DT |
1235 | } |
1236 | ||
1237 | /** | |
1238 | * gfs2_get_block_type - Check a block in a RG is of given type | |
1239 | * @rgd: the resource group holding the block | |
1240 | * @block: the block number | |
1241 | * | |
1242 | * Returns: The block type (GFS2_BLKST_*) | |
1243 | */ | |
1244 | ||
acf7e244 | 1245 | static unsigned char gfs2_get_block_type(struct gfs2_rgrpd *rgd, u64 block) |
b3b94faa DT |
1246 | { |
1247 | struct gfs2_bitmap *bi = NULL; | |
cd915493 | 1248 | u32 length, rgrp_block, buf_block; |
b3b94faa DT |
1249 | unsigned int buf; |
1250 | unsigned char type; | |
1251 | ||
bb8d8a6f SW |
1252 | length = rgd->rd_length; |
1253 | rgrp_block = block - rgd->rd_data0; | |
b3b94faa DT |
1254 | |
1255 | for (buf = 0; buf < length; buf++) { | |
1256 | bi = rgd->rd_bits + buf; | |
1257 | if (rgrp_block < (bi->bi_start + bi->bi_len) * GFS2_NBBY) | |
1258 | break; | |
1259 | } | |
1260 | ||
1261 | gfs2_assert(rgd->rd_sbd, buf < length); | |
1262 | buf_block = rgrp_block - bi->bi_start * GFS2_NBBY; | |
1263 | ||
feaa7bba | 1264 | type = gfs2_testbit(rgd, bi->bi_bh->b_data + bi->bi_offset, |
b3b94faa DT |
1265 | bi->bi_len, buf_block); |
1266 | ||
1267 | return type; | |
1268 | } | |
1269 | ||
1270 | /** | |
6a8099ed | 1271 | * rgblk_search - find a block in @state |
b3b94faa DT |
1272 | * @rgd: the resource group descriptor |
1273 | * @goal: the goal block within the RG (start here to search for avail block) | |
6a8099ed | 1274 | * @state: GFS2_BLKST_XXX the before-allocation state to find |
6e87ed0f | 1275 | * @dinode: TRUE if the first block we allocate is for a dinode |
b3e47ca0 | 1276 | * @rbi: address of the pointer to the bitmap containing the block found |
b3b94faa | 1277 | * |
6a8099ed | 1278 | * Walk rgrp's bitmap to find bits that represent a block in @state. |
b3b94faa DT |
1279 | * |
1280 | * This function never fails, because we wouldn't call it unless we | |
1281 | * know (from reservation results, etc.) that a block is available. | |
1282 | * | |
b3e47ca0 BP |
1283 | * Scope of @goal is just within rgrp, not the whole filesystem. |
1284 | * Scope of @returned block is just within bitmap, not the whole filesystem. | |
b3b94faa | 1285 | * |
b3e47ca0 | 1286 | * Returns: the block number found relative to the bitmap rbi |
b3b94faa DT |
1287 | */ |
1288 | ||
cd915493 | 1289 | static u32 rgblk_search(struct gfs2_rgrpd *rgd, u32 goal, |
6a8099ed | 1290 | unsigned char state, |
b3e47ca0 | 1291 | struct gfs2_bitmap **rbi) |
b3b94faa DT |
1292 | { |
1293 | struct gfs2_bitmap *bi = NULL; | |
b45e41d7 | 1294 | const u32 length = rgd->rd_length; |
60a0b8f9 | 1295 | u32 blk = BFITNOENT; |
b3b94faa | 1296 | unsigned int buf, x; |
60a0b8f9 | 1297 | const u8 *buffer = NULL; |
b3b94faa | 1298 | |
b3e47ca0 | 1299 | *rbi = NULL; |
b3b94faa DT |
1300 | /* Find bitmap block that contains bits for goal block */ |
1301 | for (buf = 0; buf < length; buf++) { | |
1302 | bi = rgd->rd_bits + buf; | |
60a0b8f9 SW |
1303 | /* Convert scope of "goal" from rgrp-wide to within found bit block */ |
1304 | if (goal < (bi->bi_start + bi->bi_len) * GFS2_NBBY) { | |
1305 | goal -= bi->bi_start * GFS2_NBBY; | |
1306 | goto do_search; | |
1307 | } | |
b3b94faa | 1308 | } |
60a0b8f9 SW |
1309 | buf = 0; |
1310 | goal = 0; | |
b3b94faa | 1311 | |
60a0b8f9 | 1312 | do_search: |
b3b94faa DT |
1313 | /* Search (up to entire) bitmap in this rgrp for allocatable block. |
1314 | "x <= length", instead of "x < length", because we typically start | |
1315 | the search in the middle of a bit block, but if we can't find an | |
1316 | allocatable block anywhere else, we want to be able wrap around and | |
1317 | search in the first part of our first-searched bit block. */ | |
1318 | for (x = 0; x <= length; x++) { | |
60a0b8f9 SW |
1319 | bi = rgd->rd_bits + buf; |
1320 | ||
1321 | if (test_bit(GBF_FULL, &bi->bi_flags) && | |
6a8099ed | 1322 | (state == GFS2_BLKST_FREE)) |
60a0b8f9 SW |
1323 | goto skip; |
1324 | ||
5f3eae75 BP |
1325 | /* The GFS2_BLKST_UNLINKED state doesn't apply to the clone |
1326 | bitmaps, so we must search the originals for that. */ | |
b45e41d7 | 1327 | buffer = bi->bi_bh->b_data + bi->bi_offset; |
7c9ca621 | 1328 | WARN_ON(!buffer_uptodate(bi->bi_bh)); |
6a8099ed | 1329 | if (state != GFS2_BLKST_UNLINKED && bi->bi_clone) |
110acf38 SW |
1330 | buffer = bi->bi_clone + bi->bi_offset; |
1331 | ||
6a8099ed | 1332 | blk = gfs2_bitfit(buffer, bi->bi_len, goal, state); |
b3b94faa DT |
1333 | if (blk != BFITNOENT) |
1334 | break; | |
1335 | ||
6a8099ed | 1336 | if ((goal == 0) && (state == GFS2_BLKST_FREE)) |
60a0b8f9 SW |
1337 | set_bit(GBF_FULL, &bi->bi_flags); |
1338 | ||
b3b94faa | 1339 | /* Try next bitmap block (wrap back to rgrp header if at end) */ |
60a0b8f9 SW |
1340 | skip: |
1341 | buf++; | |
1342 | buf %= length; | |
b3b94faa DT |
1343 | goal = 0; |
1344 | } | |
1345 | ||
b3e47ca0 BP |
1346 | if (blk != BFITNOENT) |
1347 | *rbi = bi; | |
7c9ca621 | 1348 | |
b3e47ca0 BP |
1349 | return blk; |
1350 | } | |
60a0b8f9 | 1351 | |
b3e47ca0 BP |
1352 | /** |
1353 | * gfs2_alloc_extent - allocate an extent from a given bitmap | |
1354 | * @rgd: the resource group descriptor | |
1355 | * @bi: the bitmap within the rgrp | |
1356 | * @blk: the block within the bitmap | |
1357 | * @dinode: TRUE if the first block we allocate is for a dinode | |
1358 | * @n: The extent length | |
1359 | * | |
1360 | * Add the found bitmap buffer to the transaction. | |
1361 | * Set the found bits to @new_state to change block's allocation state. | |
1362 | * Returns: starting block number of the extent (fs scope) | |
1363 | */ | |
1364 | static u64 gfs2_alloc_extent(struct gfs2_rgrpd *rgd, struct gfs2_bitmap *bi, | |
1365 | u32 blk, bool dinode, unsigned int *n) | |
1366 | { | |
1367 | const unsigned int elen = *n; | |
1368 | u32 goal; | |
1369 | const u8 *buffer = NULL; | |
1370 | ||
6a8099ed | 1371 | *n = 0; |
b3e47ca0 | 1372 | buffer = bi->bi_bh->b_data + bi->bi_offset; |
60a0b8f9 | 1373 | gfs2_trans_add_bh(rgd->rd_gl, bi->bi_bh, 1); |
29c578f5 | 1374 | gfs2_setbit(rgd, bi->bi_bh->b_data, bi->bi_clone, |
6e87ed0f | 1375 | bi, blk, dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED); |
6a8099ed | 1376 | (*n)++; |
60a0b8f9 SW |
1377 | goal = blk; |
1378 | while (*n < elen) { | |
1379 | goal++; | |
1380 | if (goal >= (bi->bi_len * GFS2_NBBY)) | |
1381 | break; | |
1382 | if (gfs2_testbit(rgd, buffer, bi->bi_len, goal) != | |
1383 | GFS2_BLKST_FREE) | |
1384 | break; | |
29c578f5 | 1385 | gfs2_setbit(rgd, bi->bi_bh->b_data, bi->bi_clone, |
6e87ed0f | 1386 | bi, goal, GFS2_BLKST_USED); |
60a0b8f9 | 1387 | (*n)++; |
c8cdf479 | 1388 | } |
b3e47ca0 | 1389 | blk = gfs2_bi2rgd_blk(bi, blk); |
6a8099ed | 1390 | rgd->rd_last_alloc = blk + *n - 1; |
b3e47ca0 | 1391 | return rgd->rd_data0 + blk; |
b3b94faa DT |
1392 | } |
1393 | ||
1394 | /** | |
1395 | * rgblk_free - Change alloc state of given block(s) | |
1396 | * @sdp: the filesystem | |
1397 | * @bstart: the start of a run of blocks to free | |
1398 | * @blen: the length of the block run (all must lie within ONE RG!) | |
1399 | * @new_state: GFS2_BLKST_XXX the after-allocation block state | |
1400 | * | |
1401 | * Returns: Resource group containing the block(s) | |
1402 | */ | |
1403 | ||
cd915493 SW |
1404 | static struct gfs2_rgrpd *rgblk_free(struct gfs2_sbd *sdp, u64 bstart, |
1405 | u32 blen, unsigned char new_state) | |
b3b94faa DT |
1406 | { |
1407 | struct gfs2_rgrpd *rgd; | |
1408 | struct gfs2_bitmap *bi = NULL; | |
cd915493 | 1409 | u32 length, rgrp_blk, buf_blk; |
b3b94faa DT |
1410 | unsigned int buf; |
1411 | ||
66fc061b | 1412 | rgd = gfs2_blk2rgrpd(sdp, bstart, 1); |
b3b94faa DT |
1413 | if (!rgd) { |
1414 | if (gfs2_consist(sdp)) | |
382066da | 1415 | fs_err(sdp, "block = %llu\n", (unsigned long long)bstart); |
b3b94faa DT |
1416 | return NULL; |
1417 | } | |
1418 | ||
bb8d8a6f | 1419 | length = rgd->rd_length; |
b3b94faa | 1420 | |
bb8d8a6f | 1421 | rgrp_blk = bstart - rgd->rd_data0; |
b3b94faa DT |
1422 | |
1423 | while (blen--) { | |
1424 | for (buf = 0; buf < length; buf++) { | |
1425 | bi = rgd->rd_bits + buf; | |
1426 | if (rgrp_blk < (bi->bi_start + bi->bi_len) * GFS2_NBBY) | |
1427 | break; | |
1428 | } | |
1429 | ||
1430 | gfs2_assert(rgd->rd_sbd, buf < length); | |
1431 | ||
1432 | buf_blk = rgrp_blk - bi->bi_start * GFS2_NBBY; | |
1433 | rgrp_blk++; | |
1434 | ||
1435 | if (!bi->bi_clone) { | |
1436 | bi->bi_clone = kmalloc(bi->bi_bh->b_size, | |
dd894be8 | 1437 | GFP_NOFS | __GFP_NOFAIL); |
b3b94faa DT |
1438 | memcpy(bi->bi_clone + bi->bi_offset, |
1439 | bi->bi_bh->b_data + bi->bi_offset, | |
1440 | bi->bi_len); | |
1441 | } | |
d4e9c4c3 | 1442 | gfs2_trans_add_bh(rgd->rd_gl, bi->bi_bh, 1); |
29c578f5 | 1443 | gfs2_setbit(rgd, bi->bi_bh->b_data, NULL, |
95c8e17f | 1444 | bi, buf_blk, new_state); |
b3b94faa DT |
1445 | } |
1446 | ||
1447 | return rgd; | |
1448 | } | |
1449 | ||
1450 | /** | |
09010978 SW |
1451 | * gfs2_rgrp_dump - print out an rgrp |
1452 | * @seq: The iterator | |
1453 | * @gl: The glock in question | |
1454 | * | |
1455 | */ | |
1456 | ||
1457 | int gfs2_rgrp_dump(struct seq_file *seq, const struct gfs2_glock *gl) | |
1458 | { | |
1459 | const struct gfs2_rgrpd *rgd = gl->gl_object; | |
1460 | if (rgd == NULL) | |
1461 | return 0; | |
1462 | gfs2_print_dbg(seq, " R: n:%llu f:%02x b:%u/%u i:%u\n", | |
1463 | (unsigned long long)rgd->rd_addr, rgd->rd_flags, | |
1464 | rgd->rd_free, rgd->rd_free_clone, rgd->rd_dinodes); | |
1465 | return 0; | |
1466 | } | |
1467 | ||
6050b9c7 SW |
1468 | static void gfs2_rgrp_error(struct gfs2_rgrpd *rgd) |
1469 | { | |
1470 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
1471 | fs_warn(sdp, "rgrp %llu has an error, marking it readonly until umount\n", | |
86d00636 | 1472 | (unsigned long long)rgd->rd_addr); |
6050b9c7 SW |
1473 | fs_warn(sdp, "umount on all nodes and run fsck.gfs2 to fix the error\n"); |
1474 | gfs2_rgrp_dump(NULL, rgd->rd_gl); | |
1475 | rgd->rd_flags |= GFS2_RDF_ERROR; | |
1476 | } | |
1477 | ||
09010978 | 1478 | /** |
6e87ed0f | 1479 | * gfs2_alloc_blocks - Allocate one or more blocks of data and/or a dinode |
1639431a | 1480 | * @ip: the inode to allocate the block for |
09010978 | 1481 | * @bn: Used to return the starting block number |
6a8099ed | 1482 | * @ndata: requested number of blocks/extent length (value/result) |
6e87ed0f | 1483 | * @dinode: 1 if we're allocating a dinode block, else 0 |
3c5d785a | 1484 | * @generation: the generation number of the inode |
b3b94faa | 1485 | * |
09010978 | 1486 | * Returns: 0 or error |
b3b94faa DT |
1487 | */ |
1488 | ||
6a8099ed | 1489 | int gfs2_alloc_blocks(struct gfs2_inode *ip, u64 *bn, unsigned int *nblocks, |
6e87ed0f | 1490 | bool dinode, u64 *generation) |
b3b94faa | 1491 | { |
feaa7bba | 1492 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
d9ba7615 | 1493 | struct buffer_head *dibh; |
9a3f236d | 1494 | struct gfs2_rgrpd *rgd; |
6a8099ed SW |
1495 | unsigned int ndata; |
1496 | u32 goal, blk; /* block, within the rgrp scope */ | |
3c5d785a | 1497 | u64 block; /* block, within the file system scope */ |
d9ba7615 | 1498 | int error; |
b3e47ca0 | 1499 | struct gfs2_bitmap *bi; |
b3b94faa | 1500 | |
9a3f236d SW |
1501 | /* Only happens if there is a bug in gfs2, return something distinctive |
1502 | * to ensure that it is noticed. | |
1503 | */ | |
564e12b1 | 1504 | if (ip->i_res == NULL) |
9a3f236d SW |
1505 | return -ECANCELED; |
1506 | ||
54335b1f | 1507 | rgd = ip->i_rgd; |
9a3f236d | 1508 | |
3c5d785a | 1509 | if (!dinode && rgrp_contains_block(rgd, ip->i_goal)) |
ce276b06 | 1510 | goal = ip->i_goal - rgd->rd_data0; |
b3b94faa | 1511 | else |
ac576cc5 | 1512 | goal = rgd->rd_last_alloc; |
b3b94faa | 1513 | |
6a8099ed | 1514 | blk = rgblk_search(rgd, goal, GFS2_BLKST_FREE, &bi); |
09010978 SW |
1515 | |
1516 | /* Since all blocks are reserved in advance, this shouldn't happen */ | |
1517 | if (blk == BFITNOENT) | |
1518 | goto rgrp_error; | |
b3b94faa | 1519 | |
6a8099ed SW |
1520 | block = gfs2_alloc_extent(rgd, bi, blk, dinode, nblocks); |
1521 | ndata = *nblocks; | |
1522 | if (dinode) | |
1523 | ndata--; | |
b3e47ca0 | 1524 | |
3c5d785a | 1525 | if (!dinode) { |
6a8099ed | 1526 | ip->i_goal = block + ndata - 1; |
3c5d785a BP |
1527 | error = gfs2_meta_inode_buffer(ip, &dibh); |
1528 | if (error == 0) { | |
1529 | struct gfs2_dinode *di = | |
1530 | (struct gfs2_dinode *)dibh->b_data; | |
1531 | gfs2_trans_add_bh(ip->i_gl, dibh, 1); | |
1532 | di->di_goal_meta = di->di_goal_data = | |
1533 | cpu_to_be64(ip->i_goal); | |
1534 | brelse(dibh); | |
1535 | } | |
d9ba7615 | 1536 | } |
6a8099ed | 1537 | if (rgd->rd_free < *nblocks) |
09010978 SW |
1538 | goto rgrp_error; |
1539 | ||
6a8099ed | 1540 | rgd->rd_free -= *nblocks; |
3c5d785a BP |
1541 | if (dinode) { |
1542 | rgd->rd_dinodes++; | |
1543 | *generation = rgd->rd_igeneration++; | |
1544 | if (*generation == 0) | |
1545 | *generation = rgd->rd_igeneration++; | |
1546 | } | |
b3b94faa | 1547 | |
d4e9c4c3 | 1548 | gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1); |
42d52e38 | 1549 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
b3b94faa | 1550 | |
6a8099ed | 1551 | gfs2_statfs_change(sdp, 0, -(s64)*nblocks, dinode ? 1 : 0); |
3c5d785a BP |
1552 | if (dinode) |
1553 | gfs2_trans_add_unrevoke(sdp, block, 1); | |
6a8099ed SW |
1554 | |
1555 | /* | |
1556 | * This needs reviewing to see why we cannot do the quota change | |
1557 | * at this point in the dinode case. | |
1558 | */ | |
1559 | if (ndata) | |
1560 | gfs2_quota_change(ip, ndata, ip->i_inode.i_uid, | |
3c5d785a | 1561 | ip->i_inode.i_gid); |
b3b94faa | 1562 | |
6a8099ed SW |
1563 | rgd->rd_free_clone -= *nblocks; |
1564 | trace_gfs2_block_alloc(ip, block, *nblocks, | |
6e87ed0f | 1565 | dinode ? GFS2_BLKST_DINODE : GFS2_BLKST_USED); |
6050b9c7 SW |
1566 | *bn = block; |
1567 | return 0; | |
1568 | ||
1569 | rgrp_error: | |
1570 | gfs2_rgrp_error(rgd); | |
1571 | return -EIO; | |
b3b94faa DT |
1572 | } |
1573 | ||
1574 | /** | |
46fcb2ed | 1575 | * __gfs2_free_blocks - free a contiguous run of block(s) |
b3b94faa DT |
1576 | * @ip: the inode these blocks are being freed from |
1577 | * @bstart: first block of a run of contiguous blocks | |
1578 | * @blen: the length of the block run | |
46fcb2ed | 1579 | * @meta: 1 if the blocks represent metadata |
b3b94faa DT |
1580 | * |
1581 | */ | |
1582 | ||
46fcb2ed | 1583 | void __gfs2_free_blocks(struct gfs2_inode *ip, u64 bstart, u32 blen, int meta) |
b3b94faa | 1584 | { |
feaa7bba | 1585 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
1586 | struct gfs2_rgrpd *rgd; |
1587 | ||
1588 | rgd = rgblk_free(sdp, bstart, blen, GFS2_BLKST_FREE); | |
1589 | if (!rgd) | |
1590 | return; | |
63997775 | 1591 | trace_gfs2_block_alloc(ip, bstart, blen, GFS2_BLKST_FREE); |
cfc8b549 | 1592 | rgd->rd_free += blen; |
66fc061b | 1593 | rgd->rd_flags &= ~GFS2_RGF_TRIMMED; |
d4e9c4c3 | 1594 | gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1); |
42d52e38 | 1595 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
b3b94faa | 1596 | |
6d3117b4 | 1597 | /* Directories keep their data in the metadata address space */ |
46fcb2ed | 1598 | if (meta || ip->i_depth) |
6d3117b4 | 1599 | gfs2_meta_wipe(ip, bstart, blen); |
4c16c36a | 1600 | } |
b3b94faa | 1601 | |
4c16c36a BP |
1602 | /** |
1603 | * gfs2_free_meta - free a contiguous run of data block(s) | |
1604 | * @ip: the inode these blocks are being freed from | |
1605 | * @bstart: first block of a run of contiguous blocks | |
1606 | * @blen: the length of the block run | |
1607 | * | |
1608 | */ | |
1609 | ||
1610 | void gfs2_free_meta(struct gfs2_inode *ip, u64 bstart, u32 blen) | |
1611 | { | |
1612 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); | |
1613 | ||
46fcb2ed | 1614 | __gfs2_free_blocks(ip, bstart, blen, 1); |
b3b94faa | 1615 | gfs2_statfs_change(sdp, 0, +blen, 0); |
2933f925 | 1616 | gfs2_quota_change(ip, -(s64)blen, ip->i_inode.i_uid, ip->i_inode.i_gid); |
b3b94faa DT |
1617 | } |
1618 | ||
feaa7bba SW |
1619 | void gfs2_unlink_di(struct inode *inode) |
1620 | { | |
1621 | struct gfs2_inode *ip = GFS2_I(inode); | |
1622 | struct gfs2_sbd *sdp = GFS2_SB(inode); | |
1623 | struct gfs2_rgrpd *rgd; | |
dbb7cae2 | 1624 | u64 blkno = ip->i_no_addr; |
feaa7bba SW |
1625 | |
1626 | rgd = rgblk_free(sdp, blkno, 1, GFS2_BLKST_UNLINKED); | |
1627 | if (!rgd) | |
1628 | return; | |
63997775 | 1629 | trace_gfs2_block_alloc(ip, blkno, 1, GFS2_BLKST_UNLINKED); |
feaa7bba | 1630 | gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1); |
42d52e38 | 1631 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
feaa7bba SW |
1632 | } |
1633 | ||
cd915493 | 1634 | static void gfs2_free_uninit_di(struct gfs2_rgrpd *rgd, u64 blkno) |
b3b94faa DT |
1635 | { |
1636 | struct gfs2_sbd *sdp = rgd->rd_sbd; | |
1637 | struct gfs2_rgrpd *tmp_rgd; | |
1638 | ||
1639 | tmp_rgd = rgblk_free(sdp, blkno, 1, GFS2_BLKST_FREE); | |
1640 | if (!tmp_rgd) | |
1641 | return; | |
1642 | gfs2_assert_withdraw(sdp, rgd == tmp_rgd); | |
1643 | ||
73f74948 | 1644 | if (!rgd->rd_dinodes) |
b3b94faa | 1645 | gfs2_consist_rgrpd(rgd); |
73f74948 | 1646 | rgd->rd_dinodes--; |
cfc8b549 | 1647 | rgd->rd_free++; |
b3b94faa | 1648 | |
d4e9c4c3 | 1649 | gfs2_trans_add_bh(rgd->rd_gl, rgd->rd_bits[0].bi_bh, 1); |
42d52e38 | 1650 | gfs2_rgrp_out(rgd, rgd->rd_bits[0].bi_bh->b_data); |
b3b94faa DT |
1651 | |
1652 | gfs2_statfs_change(sdp, 0, +1, -1); | |
b3b94faa DT |
1653 | } |
1654 | ||
b3b94faa DT |
1655 | |
1656 | void gfs2_free_di(struct gfs2_rgrpd *rgd, struct gfs2_inode *ip) | |
1657 | { | |
dbb7cae2 | 1658 | gfs2_free_uninit_di(rgd, ip->i_no_addr); |
63997775 | 1659 | trace_gfs2_block_alloc(ip, ip->i_no_addr, 1, GFS2_BLKST_FREE); |
2933f925 | 1660 | gfs2_quota_change(ip, -1, ip->i_inode.i_uid, ip->i_inode.i_gid); |
dbb7cae2 | 1661 | gfs2_meta_wipe(ip, ip->i_no_addr, 1); |
b3b94faa DT |
1662 | } |
1663 | ||
acf7e244 SW |
1664 | /** |
1665 | * gfs2_check_blk_type - Check the type of a block | |
1666 | * @sdp: The superblock | |
1667 | * @no_addr: The block number to check | |
1668 | * @type: The block type we are looking for | |
1669 | * | |
1670 | * Returns: 0 if the block type matches the expected type | |
1671 | * -ESTALE if it doesn't match | |
1672 | * or -ve errno if something went wrong while checking | |
1673 | */ | |
1674 | ||
1675 | int gfs2_check_blk_type(struct gfs2_sbd *sdp, u64 no_addr, unsigned int type) | |
1676 | { | |
1677 | struct gfs2_rgrpd *rgd; | |
8339ee54 | 1678 | struct gfs2_holder rgd_gh; |
58884c4d | 1679 | int error = -EINVAL; |
acf7e244 | 1680 | |
66fc061b | 1681 | rgd = gfs2_blk2rgrpd(sdp, no_addr, 1); |
acf7e244 | 1682 | if (!rgd) |
8339ee54 | 1683 | goto fail; |
acf7e244 SW |
1684 | |
1685 | error = gfs2_glock_nq_init(rgd->rd_gl, LM_ST_SHARED, 0, &rgd_gh); | |
1686 | if (error) | |
8339ee54 | 1687 | goto fail; |
acf7e244 SW |
1688 | |
1689 | if (gfs2_get_block_type(rgd, no_addr) != type) | |
1690 | error = -ESTALE; | |
1691 | ||
1692 | gfs2_glock_dq_uninit(&rgd_gh); | |
acf7e244 SW |
1693 | fail: |
1694 | return error; | |
1695 | } | |
1696 | ||
b3b94faa DT |
1697 | /** |
1698 | * gfs2_rlist_add - add a RG to a list of RGs | |
70b0c365 | 1699 | * @ip: the inode |
b3b94faa DT |
1700 | * @rlist: the list of resource groups |
1701 | * @block: the block | |
1702 | * | |
1703 | * Figure out what RG a block belongs to and add that RG to the list | |
1704 | * | |
1705 | * FIXME: Don't use NOFAIL | |
1706 | * | |
1707 | */ | |
1708 | ||
70b0c365 | 1709 | void gfs2_rlist_add(struct gfs2_inode *ip, struct gfs2_rgrp_list *rlist, |
cd915493 | 1710 | u64 block) |
b3b94faa | 1711 | { |
70b0c365 | 1712 | struct gfs2_sbd *sdp = GFS2_SB(&ip->i_inode); |
b3b94faa DT |
1713 | struct gfs2_rgrpd *rgd; |
1714 | struct gfs2_rgrpd **tmp; | |
1715 | unsigned int new_space; | |
1716 | unsigned int x; | |
1717 | ||
1718 | if (gfs2_assert_warn(sdp, !rlist->rl_ghs)) | |
1719 | return; | |
1720 | ||
70b0c365 SW |
1721 | if (ip->i_rgd && rgrp_contains_block(ip->i_rgd, block)) |
1722 | rgd = ip->i_rgd; | |
1723 | else | |
66fc061b | 1724 | rgd = gfs2_blk2rgrpd(sdp, block, 1); |
b3b94faa | 1725 | if (!rgd) { |
70b0c365 | 1726 | fs_err(sdp, "rlist_add: no rgrp for block %llu\n", (unsigned long long)block); |
b3b94faa DT |
1727 | return; |
1728 | } | |
70b0c365 | 1729 | ip->i_rgd = rgd; |
b3b94faa DT |
1730 | |
1731 | for (x = 0; x < rlist->rl_rgrps; x++) | |
1732 | if (rlist->rl_rgd[x] == rgd) | |
1733 | return; | |
1734 | ||
1735 | if (rlist->rl_rgrps == rlist->rl_space) { | |
1736 | new_space = rlist->rl_space + 10; | |
1737 | ||
1738 | tmp = kcalloc(new_space, sizeof(struct gfs2_rgrpd *), | |
dd894be8 | 1739 | GFP_NOFS | __GFP_NOFAIL); |
b3b94faa DT |
1740 | |
1741 | if (rlist->rl_rgd) { | |
1742 | memcpy(tmp, rlist->rl_rgd, | |
1743 | rlist->rl_space * sizeof(struct gfs2_rgrpd *)); | |
1744 | kfree(rlist->rl_rgd); | |
1745 | } | |
1746 | ||
1747 | rlist->rl_space = new_space; | |
1748 | rlist->rl_rgd = tmp; | |
1749 | } | |
1750 | ||
1751 | rlist->rl_rgd[rlist->rl_rgrps++] = rgd; | |
1752 | } | |
1753 | ||
1754 | /** | |
1755 | * gfs2_rlist_alloc - all RGs have been added to the rlist, now allocate | |
1756 | * and initialize an array of glock holders for them | |
1757 | * @rlist: the list of resource groups | |
1758 | * @state: the lock state to acquire the RG lock in | |
1759 | * @flags: the modifier flags for the holder structures | |
1760 | * | |
1761 | * FIXME: Don't use NOFAIL | |
1762 | * | |
1763 | */ | |
1764 | ||
fe6c991c | 1765 | void gfs2_rlist_alloc(struct gfs2_rgrp_list *rlist, unsigned int state) |
b3b94faa DT |
1766 | { |
1767 | unsigned int x; | |
1768 | ||
1769 | rlist->rl_ghs = kcalloc(rlist->rl_rgrps, sizeof(struct gfs2_holder), | |
dd894be8 | 1770 | GFP_NOFS | __GFP_NOFAIL); |
b3b94faa DT |
1771 | for (x = 0; x < rlist->rl_rgrps; x++) |
1772 | gfs2_holder_init(rlist->rl_rgd[x]->rd_gl, | |
fe6c991c | 1773 | state, 0, |
b3b94faa DT |
1774 | &rlist->rl_ghs[x]); |
1775 | } | |
1776 | ||
1777 | /** | |
1778 | * gfs2_rlist_free - free a resource group list | |
1779 | * @list: the list of resource groups | |
1780 | * | |
1781 | */ | |
1782 | ||
1783 | void gfs2_rlist_free(struct gfs2_rgrp_list *rlist) | |
1784 | { | |
1785 | unsigned int x; | |
1786 | ||
1787 | kfree(rlist->rl_rgd); | |
1788 | ||
1789 | if (rlist->rl_ghs) { | |
1790 | for (x = 0; x < rlist->rl_rgrps; x++) | |
1791 | gfs2_holder_uninit(&rlist->rl_ghs[x]); | |
1792 | kfree(rlist->rl_ghs); | |
1793 | } | |
1794 | } | |
1795 |