]>
Commit | Line | Data |
---|---|---|
5212e11f VV |
1 | /* |
2 | * Block Translation Table | |
3 | * Copyright (c) 2014-2015, Intel Corporation. | |
4 | * | |
5 | * This program is free software; you can redistribute it and/or modify it | |
6 | * under the terms and conditions of the GNU General Public License, | |
7 | * version 2, as published by the Free Software Foundation. | |
8 | * | |
9 | * This program is distributed in the hope it will be useful, but WITHOUT | |
10 | * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or | |
11 | * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for | |
12 | * more details. | |
13 | */ | |
14 | #include <linux/highmem.h> | |
15 | #include <linux/debugfs.h> | |
16 | #include <linux/blkdev.h> | |
17 | #include <linux/module.h> | |
18 | #include <linux/device.h> | |
19 | #include <linux/mutex.h> | |
20 | #include <linux/hdreg.h> | |
21 | #include <linux/genhd.h> | |
22 | #include <linux/sizes.h> | |
23 | #include <linux/ndctl.h> | |
24 | #include <linux/fs.h> | |
25 | #include <linux/nd.h> | |
23c47d2a | 26 | #include <linux/backing-dev.h> |
5212e11f VV |
27 | #include "btt.h" |
28 | #include "nd.h" | |
29 | ||
30 | enum log_ent_request { | |
31 | LOG_NEW_ENT = 0, | |
32 | LOG_OLD_ENT | |
33 | }; | |
34 | ||
86652d2e VV |
35 | static struct device *to_dev(struct arena_info *arena) |
36 | { | |
37 | return &arena->nd_btt->dev; | |
38 | } | |
39 | ||
d9b83c75 VV |
40 | static u64 adjust_initial_offset(struct nd_btt *nd_btt, u64 offset) |
41 | { | |
42 | return offset + nd_btt->initial_offset; | |
43 | } | |
44 | ||
5212e11f | 45 | static int arena_read_bytes(struct arena_info *arena, resource_size_t offset, |
3ae3d67b | 46 | void *buf, size_t n, unsigned long flags) |
5212e11f VV |
47 | { |
48 | struct nd_btt *nd_btt = arena->nd_btt; | |
49 | struct nd_namespace_common *ndns = nd_btt->ndns; | |
50 | ||
14e49454 | 51 | /* arena offsets may be shifted from the base of the device */ |
d9b83c75 | 52 | offset = adjust_initial_offset(nd_btt, offset); |
3ae3d67b | 53 | return nvdimm_read_bytes(ndns, offset, buf, n, flags); |
5212e11f VV |
54 | } |
55 | ||
56 | static int arena_write_bytes(struct arena_info *arena, resource_size_t offset, | |
3ae3d67b | 57 | void *buf, size_t n, unsigned long flags) |
5212e11f VV |
58 | { |
59 | struct nd_btt *nd_btt = arena->nd_btt; | |
60 | struct nd_namespace_common *ndns = nd_btt->ndns; | |
61 | ||
14e49454 | 62 | /* arena offsets may be shifted from the base of the device */ |
d9b83c75 | 63 | offset = adjust_initial_offset(nd_btt, offset); |
3ae3d67b | 64 | return nvdimm_write_bytes(ndns, offset, buf, n, flags); |
5212e11f VV |
65 | } |
66 | ||
67 | static int btt_info_write(struct arena_info *arena, struct btt_sb *super) | |
68 | { | |
69 | int ret; | |
70 | ||
b177fe85 VV |
71 | /* |
72 | * infooff and info2off should always be at least 512B aligned. | |
73 | * We rely on that to make sure rw_bytes does error clearing | |
74 | * correctly, so make sure that is the case. | |
75 | */ | |
86652d2e VV |
76 | dev_WARN_ONCE(to_dev(arena), !IS_ALIGNED(arena->infooff, 512), |
77 | "arena->infooff: %#llx is unaligned\n", arena->infooff); | |
78 | dev_WARN_ONCE(to_dev(arena), !IS_ALIGNED(arena->info2off, 512), | |
79 | "arena->info2off: %#llx is unaligned\n", arena->info2off); | |
b177fe85 | 80 | |
5212e11f | 81 | ret = arena_write_bytes(arena, arena->info2off, super, |
3ae3d67b | 82 | sizeof(struct btt_sb), 0); |
5212e11f VV |
83 | if (ret) |
84 | return ret; | |
85 | ||
86 | return arena_write_bytes(arena, arena->infooff, super, | |
3ae3d67b | 87 | sizeof(struct btt_sb), 0); |
5212e11f VV |
88 | } |
89 | ||
90 | static int btt_info_read(struct arena_info *arena, struct btt_sb *super) | |
91 | { | |
5212e11f | 92 | return arena_read_bytes(arena, arena->infooff, super, |
3ae3d67b | 93 | sizeof(struct btt_sb), 0); |
5212e11f VV |
94 | } |
95 | ||
96 | /* | |
97 | * 'raw' version of btt_map write | |
98 | * Assumptions: | |
99 | * mapping is in little-endian | |
100 | * mapping contains 'E' and 'Z' flags as desired | |
101 | */ | |
3ae3d67b VV |
102 | static int __btt_map_write(struct arena_info *arena, u32 lba, __le32 mapping, |
103 | unsigned long flags) | |
5212e11f VV |
104 | { |
105 | u64 ns_off = arena->mapoff + (lba * MAP_ENT_SIZE); | |
106 | ||
86652d2e VV |
107 | if (unlikely(lba >= arena->external_nlba)) |
108 | dev_err_ratelimited(to_dev(arena), | |
109 | "%s: lba %#x out of range (max: %#x)\n", | |
110 | __func__, lba, arena->external_nlba); | |
3ae3d67b | 111 | return arena_write_bytes(arena, ns_off, &mapping, MAP_ENT_SIZE, flags); |
5212e11f VV |
112 | } |
113 | ||
114 | static int btt_map_write(struct arena_info *arena, u32 lba, u32 mapping, | |
3ae3d67b | 115 | u32 z_flag, u32 e_flag, unsigned long rwb_flags) |
5212e11f VV |
116 | { |
117 | u32 ze; | |
118 | __le32 mapping_le; | |
119 | ||
120 | /* | |
121 | * This 'mapping' is supposed to be just the LBA mapping, without | |
122 | * any flags set, so strip the flag bits. | |
123 | */ | |
0595d539 | 124 | mapping = ent_lba(mapping); |
5212e11f VV |
125 | |
126 | ze = (z_flag << 1) + e_flag; | |
127 | switch (ze) { | |
128 | case 0: | |
129 | /* | |
130 | * We want to set neither of the Z or E flags, and | |
131 | * in the actual layout, this means setting the bit | |
132 | * positions of both to '1' to indicate a 'normal' | |
133 | * map entry | |
134 | */ | |
135 | mapping |= MAP_ENT_NORMAL; | |
136 | break; | |
137 | case 1: | |
138 | mapping |= (1 << MAP_ERR_SHIFT); | |
139 | break; | |
140 | case 2: | |
141 | mapping |= (1 << MAP_TRIM_SHIFT); | |
142 | break; | |
143 | default: | |
144 | /* | |
145 | * The case where Z and E are both sent in as '1' could be | |
146 | * construed as a valid 'normal' case, but we decide not to, | |
147 | * to avoid confusion | |
148 | */ | |
86652d2e VV |
149 | dev_err_ratelimited(to_dev(arena), |
150 | "Invalid use of Z and E flags\n"); | |
5212e11f VV |
151 | return -EIO; |
152 | } | |
153 | ||
154 | mapping_le = cpu_to_le32(mapping); | |
3ae3d67b | 155 | return __btt_map_write(arena, lba, mapping_le, rwb_flags); |
5212e11f VV |
156 | } |
157 | ||
158 | static int btt_map_read(struct arena_info *arena, u32 lba, u32 *mapping, | |
3ae3d67b | 159 | int *trim, int *error, unsigned long rwb_flags) |
5212e11f VV |
160 | { |
161 | int ret; | |
162 | __le32 in; | |
163 | u32 raw_mapping, postmap, ze, z_flag, e_flag; | |
164 | u64 ns_off = arena->mapoff + (lba * MAP_ENT_SIZE); | |
165 | ||
86652d2e VV |
166 | if (unlikely(lba >= arena->external_nlba)) |
167 | dev_err_ratelimited(to_dev(arena), | |
168 | "%s: lba %#x out of range (max: %#x)\n", | |
169 | __func__, lba, arena->external_nlba); | |
5212e11f | 170 | |
3ae3d67b | 171 | ret = arena_read_bytes(arena, ns_off, &in, MAP_ENT_SIZE, rwb_flags); |
5212e11f VV |
172 | if (ret) |
173 | return ret; | |
174 | ||
175 | raw_mapping = le32_to_cpu(in); | |
176 | ||
0595d539 VV |
177 | z_flag = ent_z_flag(raw_mapping); |
178 | e_flag = ent_e_flag(raw_mapping); | |
5212e11f | 179 | ze = (z_flag << 1) + e_flag; |
0595d539 | 180 | postmap = ent_lba(raw_mapping); |
5212e11f VV |
181 | |
182 | /* Reuse the {z,e}_flag variables for *trim and *error */ | |
183 | z_flag = 0; | |
184 | e_flag = 0; | |
185 | ||
186 | switch (ze) { | |
187 | case 0: | |
188 | /* Initial state. Return postmap = premap */ | |
189 | *mapping = lba; | |
190 | break; | |
191 | case 1: | |
192 | *mapping = postmap; | |
193 | e_flag = 1; | |
194 | break; | |
195 | case 2: | |
196 | *mapping = postmap; | |
197 | z_flag = 1; | |
198 | break; | |
199 | case 3: | |
200 | *mapping = postmap; | |
201 | break; | |
202 | default: | |
203 | return -EIO; | |
204 | } | |
205 | ||
206 | if (trim) | |
207 | *trim = z_flag; | |
208 | if (error) | |
209 | *error = e_flag; | |
210 | ||
211 | return ret; | |
212 | } | |
213 | ||
24e3a7fb VV |
214 | static int btt_log_group_read(struct arena_info *arena, u32 lane, |
215 | struct log_group *log) | |
5212e11f | 216 | { |
5212e11f | 217 | return arena_read_bytes(arena, |
24e3a7fb VV |
218 | arena->logoff + (lane * LOG_GRP_SIZE), log, |
219 | LOG_GRP_SIZE, 0); | |
5212e11f VV |
220 | } |
221 | ||
222 | static struct dentry *debugfs_root; | |
223 | ||
224 | static void arena_debugfs_init(struct arena_info *a, struct dentry *parent, | |
225 | int idx) | |
226 | { | |
227 | char dirname[32]; | |
228 | struct dentry *d; | |
229 | ||
230 | /* If for some reason, parent bttN was not created, exit */ | |
231 | if (!parent) | |
232 | return; | |
233 | ||
234 | snprintf(dirname, 32, "arena%d", idx); | |
235 | d = debugfs_create_dir(dirname, parent); | |
236 | if (IS_ERR_OR_NULL(d)) | |
237 | return; | |
238 | a->debugfs_dir = d; | |
239 | ||
240 | debugfs_create_x64("size", S_IRUGO, d, &a->size); | |
241 | debugfs_create_x64("external_lba_start", S_IRUGO, d, | |
242 | &a->external_lba_start); | |
243 | debugfs_create_x32("internal_nlba", S_IRUGO, d, &a->internal_nlba); | |
244 | debugfs_create_u32("internal_lbasize", S_IRUGO, d, | |
245 | &a->internal_lbasize); | |
246 | debugfs_create_x32("external_nlba", S_IRUGO, d, &a->external_nlba); | |
247 | debugfs_create_u32("external_lbasize", S_IRUGO, d, | |
248 | &a->external_lbasize); | |
249 | debugfs_create_u32("nfree", S_IRUGO, d, &a->nfree); | |
250 | debugfs_create_u16("version_major", S_IRUGO, d, &a->version_major); | |
251 | debugfs_create_u16("version_minor", S_IRUGO, d, &a->version_minor); | |
252 | debugfs_create_x64("nextoff", S_IRUGO, d, &a->nextoff); | |
253 | debugfs_create_x64("infooff", S_IRUGO, d, &a->infooff); | |
254 | debugfs_create_x64("dataoff", S_IRUGO, d, &a->dataoff); | |
255 | debugfs_create_x64("mapoff", S_IRUGO, d, &a->mapoff); | |
256 | debugfs_create_x64("logoff", S_IRUGO, d, &a->logoff); | |
257 | debugfs_create_x64("info2off", S_IRUGO, d, &a->info2off); | |
258 | debugfs_create_x32("flags", S_IRUGO, d, &a->flags); | |
24e3a7fb VV |
259 | debugfs_create_u32("log_index_0", S_IRUGO, d, &a->log_index[0]); |
260 | debugfs_create_u32("log_index_1", S_IRUGO, d, &a->log_index[1]); | |
5212e11f VV |
261 | } |
262 | ||
263 | static void btt_debugfs_init(struct btt *btt) | |
264 | { | |
265 | int i = 0; | |
266 | struct arena_info *arena; | |
267 | ||
268 | btt->debugfs_dir = debugfs_create_dir(dev_name(&btt->nd_btt->dev), | |
269 | debugfs_root); | |
270 | if (IS_ERR_OR_NULL(btt->debugfs_dir)) | |
271 | return; | |
272 | ||
273 | list_for_each_entry(arena, &btt->arena_list, list) { | |
274 | arena_debugfs_init(arena, btt->debugfs_dir, i); | |
275 | i++; | |
276 | } | |
277 | } | |
278 | ||
24e3a7fb VV |
279 | static u32 log_seq(struct log_group *log, int log_idx) |
280 | { | |
281 | return le32_to_cpu(log->ent[log_idx].seq); | |
282 | } | |
283 | ||
5212e11f VV |
284 | /* |
285 | * This function accepts two log entries, and uses the | |
286 | * sequence number to find the 'older' entry. | |
287 | * It also updates the sequence number in this old entry to | |
288 | * make it the 'new' one if the mark_flag is set. | |
289 | * Finally, it returns which of the entries was the older one. | |
290 | * | |
291 | * TODO The logic feels a bit kludge-y. make it better.. | |
292 | */ | |
24e3a7fb | 293 | static int btt_log_get_old(struct arena_info *a, struct log_group *log) |
5212e11f | 294 | { |
24e3a7fb VV |
295 | int idx0 = a->log_index[0]; |
296 | int idx1 = a->log_index[1]; | |
5212e11f VV |
297 | int old; |
298 | ||
299 | /* | |
300 | * the first ever time this is seen, the entry goes into [0] | |
301 | * the next time, the following logic works out to put this | |
302 | * (next) entry into [1] | |
303 | */ | |
24e3a7fb VV |
304 | if (log_seq(log, idx0) == 0) { |
305 | log->ent[idx0].seq = cpu_to_le32(1); | |
5212e11f VV |
306 | return 0; |
307 | } | |
308 | ||
24e3a7fb | 309 | if (log_seq(log, idx0) == log_seq(log, idx1)) |
5212e11f | 310 | return -EINVAL; |
24e3a7fb | 311 | if (log_seq(log, idx0) + log_seq(log, idx1) > 5) |
5212e11f VV |
312 | return -EINVAL; |
313 | ||
24e3a7fb VV |
314 | if (log_seq(log, idx0) < log_seq(log, idx1)) { |
315 | if ((log_seq(log, idx1) - log_seq(log, idx0)) == 1) | |
5212e11f VV |
316 | old = 0; |
317 | else | |
318 | old = 1; | |
319 | } else { | |
24e3a7fb | 320 | if ((log_seq(log, idx0) - log_seq(log, idx1)) == 1) |
5212e11f VV |
321 | old = 1; |
322 | else | |
323 | old = 0; | |
324 | } | |
325 | ||
326 | return old; | |
327 | } | |
328 | ||
5212e11f VV |
329 | /* |
330 | * This function copies the desired (old/new) log entry into ent if | |
331 | * it is not NULL. It returns the sub-slot number (0 or 1) | |
332 | * where the desired log entry was found. Negative return values | |
333 | * indicate errors. | |
334 | */ | |
335 | static int btt_log_read(struct arena_info *arena, u32 lane, | |
336 | struct log_entry *ent, int old_flag) | |
337 | { | |
338 | int ret; | |
339 | int old_ent, ret_ent; | |
24e3a7fb | 340 | struct log_group log; |
5212e11f | 341 | |
24e3a7fb | 342 | ret = btt_log_group_read(arena, lane, &log); |
5212e11f VV |
343 | if (ret) |
344 | return -EIO; | |
345 | ||
24e3a7fb | 346 | old_ent = btt_log_get_old(arena, &log); |
5212e11f | 347 | if (old_ent < 0 || old_ent > 1) { |
e6be2dcb | 348 | dev_err(to_dev(arena), |
5212e11f | 349 | "log corruption (%d): lane %d seq [%d, %d]\n", |
24e3a7fb VV |
350 | old_ent, lane, log.ent[arena->log_index[0]].seq, |
351 | log.ent[arena->log_index[1]].seq); | |
5212e11f VV |
352 | /* TODO set error state? */ |
353 | return -EIO; | |
354 | } | |
355 | ||
356 | ret_ent = (old_flag ? old_ent : (1 - old_ent)); | |
357 | ||
358 | if (ent != NULL) | |
24e3a7fb | 359 | memcpy(ent, &log.ent[arena->log_index[ret_ent]], LOG_ENT_SIZE); |
5212e11f VV |
360 | |
361 | return ret_ent; | |
362 | } | |
363 | ||
364 | /* | |
365 | * This function commits a log entry to media | |
366 | * It does _not_ prepare the freelist entry for the next write | |
367 | * btt_flog_write is the wrapper for updating the freelist elements | |
368 | */ | |
369 | static int __btt_log_write(struct arena_info *arena, u32 lane, | |
3ae3d67b | 370 | u32 sub, struct log_entry *ent, unsigned long flags) |
5212e11f VV |
371 | { |
372 | int ret; | |
24e3a7fb VV |
373 | u32 group_slot = arena->log_index[sub]; |
374 | unsigned int log_half = LOG_ENT_SIZE / 2; | |
5212e11f | 375 | void *src = ent; |
24e3a7fb | 376 | u64 ns_off; |
5212e11f | 377 | |
24e3a7fb VV |
378 | ns_off = arena->logoff + (lane * LOG_GRP_SIZE) + |
379 | (group_slot * LOG_ENT_SIZE); | |
5212e11f | 380 | /* split the 16B write into atomic, durable halves */ |
3ae3d67b | 381 | ret = arena_write_bytes(arena, ns_off, src, log_half, flags); |
5212e11f VV |
382 | if (ret) |
383 | return ret; | |
384 | ||
385 | ns_off += log_half; | |
386 | src += log_half; | |
3ae3d67b | 387 | return arena_write_bytes(arena, ns_off, src, log_half, flags); |
5212e11f VV |
388 | } |
389 | ||
390 | static int btt_flog_write(struct arena_info *arena, u32 lane, u32 sub, | |
391 | struct log_entry *ent) | |
392 | { | |
393 | int ret; | |
394 | ||
3ae3d67b | 395 | ret = __btt_log_write(arena, lane, sub, ent, NVDIMM_IO_ATOMIC); |
5212e11f VV |
396 | if (ret) |
397 | return ret; | |
398 | ||
399 | /* prepare the next free entry */ | |
400 | arena->freelist[lane].sub = 1 - arena->freelist[lane].sub; | |
401 | if (++(arena->freelist[lane].seq) == 4) | |
402 | arena->freelist[lane].seq = 1; | |
d9b83c75 VV |
403 | if (ent_e_flag(ent->old_map)) |
404 | arena->freelist[lane].has_err = 1; | |
405 | arena->freelist[lane].block = le32_to_cpu(ent_lba(ent->old_map)); | |
5212e11f VV |
406 | |
407 | return ret; | |
408 | } | |
409 | ||
410 | /* | |
411 | * This function initializes the BTT map to the initial state, which is | |
412 | * all-zeroes, and indicates an identity mapping | |
413 | */ | |
414 | static int btt_map_init(struct arena_info *arena) | |
415 | { | |
416 | int ret = -EINVAL; | |
417 | void *zerobuf; | |
418 | size_t offset = 0; | |
419 | size_t chunk_size = SZ_2M; | |
420 | size_t mapsize = arena->logoff - arena->mapoff; | |
421 | ||
422 | zerobuf = kzalloc(chunk_size, GFP_KERNEL); | |
423 | if (!zerobuf) | |
424 | return -ENOMEM; | |
425 | ||
b177fe85 VV |
426 | /* |
427 | * mapoff should always be at least 512B aligned. We rely on that to | |
428 | * make sure rw_bytes does error clearing correctly, so make sure that | |
429 | * is the case. | |
430 | */ | |
86652d2e VV |
431 | dev_WARN_ONCE(to_dev(arena), !IS_ALIGNED(arena->mapoff, 512), |
432 | "arena->mapoff: %#llx is unaligned\n", arena->mapoff); | |
b177fe85 | 433 | |
5212e11f VV |
434 | while (mapsize) { |
435 | size_t size = min(mapsize, chunk_size); | |
436 | ||
86652d2e | 437 | dev_WARN_ONCE(to_dev(arena), size < 512, |
04c3c982 | 438 | "chunk size: %#zx is unaligned\n", size); |
5212e11f | 439 | ret = arena_write_bytes(arena, arena->mapoff + offset, zerobuf, |
3ae3d67b | 440 | size, 0); |
5212e11f VV |
441 | if (ret) |
442 | goto free; | |
443 | ||
444 | offset += size; | |
445 | mapsize -= size; | |
446 | cond_resched(); | |
447 | } | |
448 | ||
449 | free: | |
450 | kfree(zerobuf); | |
451 | return ret; | |
452 | } | |
453 | ||
454 | /* | |
455 | * This function initializes the BTT log with 'fake' entries pointing | |
456 | * to the initial reserved set of blocks as being free | |
457 | */ | |
458 | static int btt_log_init(struct arena_info *arena) | |
459 | { | |
b177fe85 VV |
460 | size_t logsize = arena->info2off - arena->logoff; |
461 | size_t chunk_size = SZ_4K, offset = 0; | |
24e3a7fb | 462 | struct log_entry ent; |
b177fe85 | 463 | void *zerobuf; |
5212e11f VV |
464 | int ret; |
465 | u32 i; | |
5212e11f | 466 | |
b177fe85 VV |
467 | zerobuf = kzalloc(chunk_size, GFP_KERNEL); |
468 | if (!zerobuf) | |
469 | return -ENOMEM; | |
470 | /* | |
471 | * logoff should always be at least 512B aligned. We rely on that to | |
472 | * make sure rw_bytes does error clearing correctly, so make sure that | |
473 | * is the case. | |
474 | */ | |
86652d2e VV |
475 | dev_WARN_ONCE(to_dev(arena), !IS_ALIGNED(arena->logoff, 512), |
476 | "arena->logoff: %#llx is unaligned\n", arena->logoff); | |
b177fe85 VV |
477 | |
478 | while (logsize) { | |
479 | size_t size = min(logsize, chunk_size); | |
480 | ||
86652d2e | 481 | dev_WARN_ONCE(to_dev(arena), size < 512, |
04c3c982 | 482 | "chunk size: %#zx is unaligned\n", size); |
b177fe85 VV |
483 | ret = arena_write_bytes(arena, arena->logoff + offset, zerobuf, |
484 | size, 0); | |
485 | if (ret) | |
486 | goto free; | |
487 | ||
488 | offset += size; | |
489 | logsize -= size; | |
490 | cond_resched(); | |
491 | } | |
5212e11f VV |
492 | |
493 | for (i = 0; i < arena->nfree; i++) { | |
24e3a7fb VV |
494 | ent.lba = cpu_to_le32(i); |
495 | ent.old_map = cpu_to_le32(arena->external_nlba + i); | |
496 | ent.new_map = cpu_to_le32(arena->external_nlba + i); | |
497 | ent.seq = cpu_to_le32(LOG_SEQ_INIT); | |
498 | ret = __btt_log_write(arena, i, 0, &ent, 0); | |
5212e11f | 499 | if (ret) |
b177fe85 | 500 | goto free; |
5212e11f VV |
501 | } |
502 | ||
b177fe85 VV |
503 | free: |
504 | kfree(zerobuf); | |
505 | return ret; | |
5212e11f VV |
506 | } |
507 | ||
d9b83c75 VV |
508 | static u64 to_namespace_offset(struct arena_info *arena, u64 lba) |
509 | { | |
510 | return arena->dataoff + ((u64)lba * arena->internal_lbasize); | |
511 | } | |
512 | ||
513 | static int arena_clear_freelist_error(struct arena_info *arena, u32 lane) | |
514 | { | |
515 | int ret = 0; | |
516 | ||
517 | if (arena->freelist[lane].has_err) { | |
518 | void *zero_page = page_address(ZERO_PAGE(0)); | |
519 | u32 lba = arena->freelist[lane].block; | |
520 | u64 nsoff = to_namespace_offset(arena, lba); | |
521 | unsigned long len = arena->sector_size; | |
522 | ||
523 | mutex_lock(&arena->err_lock); | |
524 | ||
525 | while (len) { | |
526 | unsigned long chunk = min(len, PAGE_SIZE); | |
527 | ||
528 | ret = arena_write_bytes(arena, nsoff, zero_page, | |
529 | chunk, 0); | |
530 | if (ret) | |
531 | break; | |
532 | len -= chunk; | |
533 | nsoff += chunk; | |
534 | if (len == 0) | |
535 | arena->freelist[lane].has_err = 0; | |
536 | } | |
537 | mutex_unlock(&arena->err_lock); | |
538 | } | |
539 | return ret; | |
540 | } | |
541 | ||
5212e11f VV |
542 | static int btt_freelist_init(struct arena_info *arena) |
543 | { | |
544 | int old, new, ret; | |
545 | u32 i, map_entry; | |
546 | struct log_entry log_new, log_old; | |
547 | ||
548 | arena->freelist = kcalloc(arena->nfree, sizeof(struct free_entry), | |
549 | GFP_KERNEL); | |
550 | if (!arena->freelist) | |
551 | return -ENOMEM; | |
552 | ||
553 | for (i = 0; i < arena->nfree; i++) { | |
554 | old = btt_log_read(arena, i, &log_old, LOG_OLD_ENT); | |
555 | if (old < 0) | |
556 | return old; | |
557 | ||
558 | new = btt_log_read(arena, i, &log_new, LOG_NEW_ENT); | |
559 | if (new < 0) | |
560 | return new; | |
561 | ||
562 | /* sub points to the next one to be overwritten */ | |
563 | arena->freelist[i].sub = 1 - new; | |
564 | arena->freelist[i].seq = nd_inc_seq(le32_to_cpu(log_new.seq)); | |
565 | arena->freelist[i].block = le32_to_cpu(log_new.old_map); | |
566 | ||
d9b83c75 VV |
567 | /* |
568 | * FIXME: if error clearing fails during init, we want to make | |
569 | * the BTT read-only | |
570 | */ | |
571 | if (ent_e_flag(log_new.old_map)) { | |
572 | ret = arena_clear_freelist_error(arena, i); | |
573 | if (ret) | |
86652d2e VV |
574 | dev_err_ratelimited(to_dev(arena), |
575 | "Unable to clear known errors\n"); | |
d9b83c75 VV |
576 | } |
577 | ||
5212e11f VV |
578 | /* This implies a newly created or untouched flog entry */ |
579 | if (log_new.old_map == log_new.new_map) | |
580 | continue; | |
581 | ||
582 | /* Check if map recovery is needed */ | |
583 | ret = btt_map_read(arena, le32_to_cpu(log_new.lba), &map_entry, | |
3ae3d67b | 584 | NULL, NULL, 0); |
5212e11f VV |
585 | if (ret) |
586 | return ret; | |
587 | if ((le32_to_cpu(log_new.new_map) != map_entry) && | |
588 | (le32_to_cpu(log_new.old_map) == map_entry)) { | |
589 | /* | |
590 | * Last transaction wrote the flog, but wasn't able | |
591 | * to complete the map write. So fix up the map. | |
592 | */ | |
593 | ret = btt_map_write(arena, le32_to_cpu(log_new.lba), | |
3ae3d67b | 594 | le32_to_cpu(log_new.new_map), 0, 0, 0); |
5212e11f VV |
595 | if (ret) |
596 | return ret; | |
597 | } | |
5212e11f VV |
598 | } |
599 | ||
600 | return 0; | |
601 | } | |
602 | ||
24e3a7fb VV |
603 | static bool ent_is_padding(struct log_entry *ent) |
604 | { | |
605 | return (ent->lba == 0) && (ent->old_map == 0) && (ent->new_map == 0) | |
606 | && (ent->seq == 0); | |
607 | } | |
608 | ||
609 | /* | |
610 | * Detecting valid log indices: We read a log group (see the comments in btt.h | |
611 | * for a description of a 'log_group' and its 'slots'), and iterate over its | |
612 | * four slots. We expect that a padding slot will be all-zeroes, and use this | |
613 | * to detect a padding slot vs. an actual entry. | |
614 | * | |
615 | * If a log_group is in the initial state, i.e. hasn't been used since the | |
616 | * creation of this BTT layout, it will have three of the four slots with | |
617 | * zeroes. We skip over these log_groups for the detection of log_index. If | |
618 | * all log_groups are in the initial state (i.e. the BTT has never been | |
619 | * written to), it is safe to assume the 'new format' of log entries in slots | |
620 | * (0, 1). | |
621 | */ | |
622 | static int log_set_indices(struct arena_info *arena) | |
623 | { | |
624 | bool idx_set = false, initial_state = true; | |
625 | int ret, log_index[2] = {-1, -1}; | |
626 | u32 i, j, next_idx = 0; | |
627 | struct log_group log; | |
628 | u32 pad_count = 0; | |
629 | ||
630 | for (i = 0; i < arena->nfree; i++) { | |
631 | ret = btt_log_group_read(arena, i, &log); | |
632 | if (ret < 0) | |
633 | return ret; | |
634 | ||
635 | for (j = 0; j < 4; j++) { | |
636 | if (!idx_set) { | |
637 | if (ent_is_padding(&log.ent[j])) { | |
638 | pad_count++; | |
639 | continue; | |
640 | } else { | |
641 | /* Skip if index has been recorded */ | |
642 | if ((next_idx == 1) && | |
643 | (j == log_index[0])) | |
644 | continue; | |
645 | /* valid entry, record index */ | |
646 | log_index[next_idx] = j; | |
647 | next_idx++; | |
648 | } | |
649 | if (next_idx == 2) { | |
650 | /* two valid entries found */ | |
651 | idx_set = true; | |
652 | } else if (next_idx > 2) { | |
653 | /* too many valid indices */ | |
654 | return -ENXIO; | |
655 | } | |
656 | } else { | |
657 | /* | |
658 | * once the indices have been set, just verify | |
659 | * that all subsequent log groups are either in | |
660 | * their initial state or follow the same | |
661 | * indices. | |
662 | */ | |
663 | if (j == log_index[0]) { | |
664 | /* entry must be 'valid' */ | |
665 | if (ent_is_padding(&log.ent[j])) | |
666 | return -ENXIO; | |
667 | } else if (j == log_index[1]) { | |
668 | ; | |
669 | /* | |
670 | * log_index[1] can be padding if the | |
671 | * lane never got used and it is still | |
672 | * in the initial state (three 'padding' | |
673 | * entries) | |
674 | */ | |
675 | } else { | |
676 | /* entry must be invalid (padding) */ | |
677 | if (!ent_is_padding(&log.ent[j])) | |
678 | return -ENXIO; | |
679 | } | |
680 | } | |
681 | } | |
682 | /* | |
683 | * If any of the log_groups have more than one valid, | |
684 | * non-padding entry, then the we are no longer in the | |
685 | * initial_state | |
686 | */ | |
687 | if (pad_count < 3) | |
688 | initial_state = false; | |
689 | pad_count = 0; | |
690 | } | |
691 | ||
692 | if (!initial_state && !idx_set) | |
693 | return -ENXIO; | |
694 | ||
695 | /* | |
696 | * If all the entries in the log were in the initial state, | |
697 | * assume new padding scheme | |
698 | */ | |
699 | if (initial_state) | |
700 | log_index[1] = 1; | |
701 | ||
702 | /* | |
703 | * Only allow the known permutations of log/padding indices, | |
704 | * i.e. (0, 1), and (0, 2) | |
705 | */ | |
706 | if ((log_index[0] == 0) && ((log_index[1] == 1) || (log_index[1] == 2))) | |
707 | ; /* known index possibilities */ | |
708 | else { | |
709 | dev_err(to_dev(arena), "Found an unknown padding scheme\n"); | |
710 | return -ENXIO; | |
711 | } | |
712 | ||
713 | arena->log_index[0] = log_index[0]; | |
714 | arena->log_index[1] = log_index[1]; | |
715 | dev_dbg(to_dev(arena), "log_index_0 = %d\n", log_index[0]); | |
716 | dev_dbg(to_dev(arena), "log_index_1 = %d\n", log_index[1]); | |
717 | return 0; | |
718 | } | |
719 | ||
5212e11f VV |
720 | static int btt_rtt_init(struct arena_info *arena) |
721 | { | |
722 | arena->rtt = kcalloc(arena->nfree, sizeof(u32), GFP_KERNEL); | |
723 | if (arena->rtt == NULL) | |
724 | return -ENOMEM; | |
725 | ||
726 | return 0; | |
727 | } | |
728 | ||
729 | static int btt_maplocks_init(struct arena_info *arena) | |
730 | { | |
731 | u32 i; | |
732 | ||
733 | arena->map_locks = kcalloc(arena->nfree, sizeof(struct aligned_lock), | |
734 | GFP_KERNEL); | |
735 | if (!arena->map_locks) | |
736 | return -ENOMEM; | |
737 | ||
738 | for (i = 0; i < arena->nfree; i++) | |
739 | spin_lock_init(&arena->map_locks[i].lock); | |
740 | ||
741 | return 0; | |
742 | } | |
743 | ||
744 | static struct arena_info *alloc_arena(struct btt *btt, size_t size, | |
745 | size_t start, size_t arena_off) | |
746 | { | |
747 | struct arena_info *arena; | |
748 | u64 logsize, mapsize, datasize; | |
749 | u64 available = size; | |
750 | ||
751 | arena = kzalloc(sizeof(struct arena_info), GFP_KERNEL); | |
752 | if (!arena) | |
753 | return NULL; | |
754 | arena->nd_btt = btt->nd_btt; | |
75892004 | 755 | arena->sector_size = btt->sector_size; |
d08cd5e0 | 756 | mutex_init(&arena->err_lock); |
5212e11f VV |
757 | |
758 | if (!size) | |
759 | return arena; | |
760 | ||
761 | arena->size = size; | |
762 | arena->external_lba_start = start; | |
763 | arena->external_lbasize = btt->lbasize; | |
764 | arena->internal_lbasize = roundup(arena->external_lbasize, | |
765 | INT_LBASIZE_ALIGNMENT); | |
766 | arena->nfree = BTT_DEFAULT_NFREE; | |
14e49454 VV |
767 | arena->version_major = btt->nd_btt->version_major; |
768 | arena->version_minor = btt->nd_btt->version_minor; | |
5212e11f VV |
769 | |
770 | if (available % BTT_PG_SIZE) | |
771 | available -= (available % BTT_PG_SIZE); | |
772 | ||
773 | /* Two pages are reserved for the super block and its copy */ | |
774 | available -= 2 * BTT_PG_SIZE; | |
775 | ||
776 | /* The log takes a fixed amount of space based on nfree */ | |
24e3a7fb | 777 | logsize = roundup(arena->nfree * LOG_GRP_SIZE, BTT_PG_SIZE); |
5212e11f VV |
778 | available -= logsize; |
779 | ||
780 | /* Calculate optimal split between map and data area */ | |
781 | arena->internal_nlba = div_u64(available - BTT_PG_SIZE, | |
782 | arena->internal_lbasize + MAP_ENT_SIZE); | |
783 | arena->external_nlba = arena->internal_nlba - arena->nfree; | |
784 | ||
785 | mapsize = roundup((arena->external_nlba * MAP_ENT_SIZE), BTT_PG_SIZE); | |
786 | datasize = available - mapsize; | |
787 | ||
788 | /* 'Absolute' values, relative to start of storage space */ | |
789 | arena->infooff = arena_off; | |
790 | arena->dataoff = arena->infooff + BTT_PG_SIZE; | |
791 | arena->mapoff = arena->dataoff + datasize; | |
792 | arena->logoff = arena->mapoff + mapsize; | |
793 | arena->info2off = arena->logoff + logsize; | |
24e3a7fb VV |
794 | |
795 | /* Default log indices are (0,1) */ | |
796 | arena->log_index[0] = 0; | |
797 | arena->log_index[1] = 1; | |
5212e11f VV |
798 | return arena; |
799 | } | |
800 | ||
801 | static void free_arenas(struct btt *btt) | |
802 | { | |
803 | struct arena_info *arena, *next; | |
804 | ||
805 | list_for_each_entry_safe(arena, next, &btt->arena_list, list) { | |
806 | list_del(&arena->list); | |
807 | kfree(arena->rtt); | |
808 | kfree(arena->map_locks); | |
809 | kfree(arena->freelist); | |
810 | debugfs_remove_recursive(arena->debugfs_dir); | |
811 | kfree(arena); | |
812 | } | |
813 | } | |
814 | ||
5212e11f VV |
815 | /* |
816 | * This function reads an existing valid btt superblock and | |
817 | * populates the corresponding arena_info struct | |
818 | */ | |
819 | static void parse_arena_meta(struct arena_info *arena, struct btt_sb *super, | |
820 | u64 arena_off) | |
821 | { | |
822 | arena->internal_nlba = le32_to_cpu(super->internal_nlba); | |
823 | arena->internal_lbasize = le32_to_cpu(super->internal_lbasize); | |
824 | arena->external_nlba = le32_to_cpu(super->external_nlba); | |
825 | arena->external_lbasize = le32_to_cpu(super->external_lbasize); | |
826 | arena->nfree = le32_to_cpu(super->nfree); | |
827 | arena->version_major = le16_to_cpu(super->version_major); | |
828 | arena->version_minor = le16_to_cpu(super->version_minor); | |
829 | ||
830 | arena->nextoff = (super->nextoff == 0) ? 0 : (arena_off + | |
831 | le64_to_cpu(super->nextoff)); | |
832 | arena->infooff = arena_off; | |
833 | arena->dataoff = arena_off + le64_to_cpu(super->dataoff); | |
834 | arena->mapoff = arena_off + le64_to_cpu(super->mapoff); | |
835 | arena->logoff = arena_off + le64_to_cpu(super->logoff); | |
836 | arena->info2off = arena_off + le64_to_cpu(super->info2off); | |
837 | ||
5e329406 DW |
838 | arena->size = (le64_to_cpu(super->nextoff) > 0) |
839 | ? (le64_to_cpu(super->nextoff)) | |
840 | : (arena->info2off - arena->infooff + BTT_PG_SIZE); | |
5212e11f VV |
841 | |
842 | arena->flags = le32_to_cpu(super->flags); | |
843 | } | |
844 | ||
845 | static int discover_arenas(struct btt *btt) | |
846 | { | |
847 | int ret = 0; | |
848 | struct arena_info *arena; | |
849 | struct btt_sb *super; | |
850 | size_t remaining = btt->rawsize; | |
851 | u64 cur_nlba = 0; | |
852 | size_t cur_off = 0; | |
853 | int num_arenas = 0; | |
854 | ||
855 | super = kzalloc(sizeof(*super), GFP_KERNEL); | |
856 | if (!super) | |
857 | return -ENOMEM; | |
858 | ||
859 | while (remaining) { | |
860 | /* Alloc memory for arena */ | |
861 | arena = alloc_arena(btt, 0, 0, 0); | |
862 | if (!arena) { | |
863 | ret = -ENOMEM; | |
864 | goto out_super; | |
865 | } | |
866 | ||
867 | arena->infooff = cur_off; | |
868 | ret = btt_info_read(arena, super); | |
869 | if (ret) | |
870 | goto out; | |
871 | ||
ab45e763 | 872 | if (!nd_btt_arena_is_valid(btt->nd_btt, super)) { |
5212e11f VV |
873 | if (remaining == btt->rawsize) { |
874 | btt->init_state = INIT_NOTFOUND; | |
875 | dev_info(to_dev(arena), "No existing arenas\n"); | |
876 | goto out; | |
877 | } else { | |
e6be2dcb | 878 | dev_err(to_dev(arena), |
5212e11f VV |
879 | "Found corrupted metadata!\n"); |
880 | ret = -ENODEV; | |
881 | goto out; | |
882 | } | |
883 | } | |
884 | ||
885 | arena->external_lba_start = cur_nlba; | |
886 | parse_arena_meta(arena, super, cur_off); | |
887 | ||
24e3a7fb VV |
888 | ret = log_set_indices(arena); |
889 | if (ret) { | |
890 | dev_err(to_dev(arena), | |
891 | "Unable to deduce log/padding indices\n"); | |
892 | goto out; | |
893 | } | |
894 | ||
5212e11f VV |
895 | ret = btt_freelist_init(arena); |
896 | if (ret) | |
897 | goto out; | |
898 | ||
899 | ret = btt_rtt_init(arena); | |
900 | if (ret) | |
901 | goto out; | |
902 | ||
903 | ret = btt_maplocks_init(arena); | |
904 | if (ret) | |
905 | goto out; | |
906 | ||
907 | list_add_tail(&arena->list, &btt->arena_list); | |
908 | ||
909 | remaining -= arena->size; | |
910 | cur_off += arena->size; | |
911 | cur_nlba += arena->external_nlba; | |
912 | num_arenas++; | |
913 | ||
914 | if (arena->nextoff == 0) | |
915 | break; | |
916 | } | |
917 | btt->num_arenas = num_arenas; | |
918 | btt->nlba = cur_nlba; | |
919 | btt->init_state = INIT_READY; | |
920 | ||
921 | kfree(super); | |
922 | return ret; | |
923 | ||
924 | out: | |
925 | kfree(arena); | |
926 | free_arenas(btt); | |
927 | out_super: | |
928 | kfree(super); | |
929 | return ret; | |
930 | } | |
931 | ||
932 | static int create_arenas(struct btt *btt) | |
933 | { | |
934 | size_t remaining = btt->rawsize; | |
935 | size_t cur_off = 0; | |
936 | ||
937 | while (remaining) { | |
938 | struct arena_info *arena; | |
939 | size_t arena_size = min_t(u64, ARENA_MAX_SIZE, remaining); | |
940 | ||
941 | remaining -= arena_size; | |
942 | if (arena_size < ARENA_MIN_SIZE) | |
943 | break; | |
944 | ||
945 | arena = alloc_arena(btt, arena_size, btt->nlba, cur_off); | |
946 | if (!arena) { | |
947 | free_arenas(btt); | |
948 | return -ENOMEM; | |
949 | } | |
950 | btt->nlba += arena->external_nlba; | |
951 | if (remaining >= ARENA_MIN_SIZE) | |
952 | arena->nextoff = arena->size; | |
953 | else | |
954 | arena->nextoff = 0; | |
955 | cur_off += arena_size; | |
956 | list_add_tail(&arena->list, &btt->arena_list); | |
957 | } | |
958 | ||
959 | return 0; | |
960 | } | |
961 | ||
962 | /* | |
963 | * This function completes arena initialization by writing | |
964 | * all the metadata. | |
965 | * It is only called for an uninitialized arena when a write | |
966 | * to that arena occurs for the first time. | |
967 | */ | |
fbde1414 | 968 | static int btt_arena_write_layout(struct arena_info *arena) |
5212e11f VV |
969 | { |
970 | int ret; | |
e1455744 | 971 | u64 sum; |
5212e11f | 972 | struct btt_sb *super; |
fbde1414 | 973 | struct nd_btt *nd_btt = arena->nd_btt; |
6ec68954 | 974 | const u8 *parent_uuid = nd_dev_to_uuid(&nd_btt->ndns->dev); |
5212e11f VV |
975 | |
976 | ret = btt_map_init(arena); | |
977 | if (ret) | |
978 | return ret; | |
979 | ||
980 | ret = btt_log_init(arena); | |
981 | if (ret) | |
982 | return ret; | |
983 | ||
984 | super = kzalloc(sizeof(struct btt_sb), GFP_NOIO); | |
985 | if (!super) | |
986 | return -ENOMEM; | |
987 | ||
988 | strncpy(super->signature, BTT_SIG, BTT_SIG_LEN); | |
fbde1414 | 989 | memcpy(super->uuid, nd_btt->uuid, 16); |
6ec68954 | 990 | memcpy(super->parent_uuid, parent_uuid, 16); |
5212e11f VV |
991 | super->flags = cpu_to_le32(arena->flags); |
992 | super->version_major = cpu_to_le16(arena->version_major); | |
993 | super->version_minor = cpu_to_le16(arena->version_minor); | |
994 | super->external_lbasize = cpu_to_le32(arena->external_lbasize); | |
995 | super->external_nlba = cpu_to_le32(arena->external_nlba); | |
996 | super->internal_lbasize = cpu_to_le32(arena->internal_lbasize); | |
997 | super->internal_nlba = cpu_to_le32(arena->internal_nlba); | |
998 | super->nfree = cpu_to_le32(arena->nfree); | |
999 | super->infosize = cpu_to_le32(sizeof(struct btt_sb)); | |
1000 | super->nextoff = cpu_to_le64(arena->nextoff); | |
1001 | /* | |
1002 | * Subtract arena->infooff (arena start) so numbers are relative | |
1003 | * to 'this' arena | |
1004 | */ | |
1005 | super->dataoff = cpu_to_le64(arena->dataoff - arena->infooff); | |
1006 | super->mapoff = cpu_to_le64(arena->mapoff - arena->infooff); | |
1007 | super->logoff = cpu_to_le64(arena->logoff - arena->infooff); | |
1008 | super->info2off = cpu_to_le64(arena->info2off - arena->infooff); | |
1009 | ||
1010 | super->flags = 0; | |
e1455744 DW |
1011 | sum = nd_sb_checksum((struct nd_gen_sb *) super); |
1012 | super->checksum = cpu_to_le64(sum); | |
5212e11f VV |
1013 | |
1014 | ret = btt_info_write(arena, super); | |
1015 | ||
1016 | kfree(super); | |
1017 | return ret; | |
1018 | } | |
1019 | ||
1020 | /* | |
1021 | * This function completes the initialization for the BTT namespace | |
1022 | * such that it is ready to accept IOs | |
1023 | */ | |
1024 | static int btt_meta_init(struct btt *btt) | |
1025 | { | |
1026 | int ret = 0; | |
1027 | struct arena_info *arena; | |
1028 | ||
1029 | mutex_lock(&btt->init_lock); | |
1030 | list_for_each_entry(arena, &btt->arena_list, list) { | |
fbde1414 | 1031 | ret = btt_arena_write_layout(arena); |
5212e11f VV |
1032 | if (ret) |
1033 | goto unlock; | |
1034 | ||
1035 | ret = btt_freelist_init(arena); | |
1036 | if (ret) | |
1037 | goto unlock; | |
1038 | ||
1039 | ret = btt_rtt_init(arena); | |
1040 | if (ret) | |
1041 | goto unlock; | |
1042 | ||
1043 | ret = btt_maplocks_init(arena); | |
1044 | if (ret) | |
1045 | goto unlock; | |
1046 | } | |
1047 | ||
1048 | btt->init_state = INIT_READY; | |
1049 | ||
1050 | unlock: | |
1051 | mutex_unlock(&btt->init_lock); | |
1052 | return ret; | |
1053 | } | |
1054 | ||
41cd8b70 VV |
1055 | static u32 btt_meta_size(struct btt *btt) |
1056 | { | |
1057 | return btt->lbasize - btt->sector_size; | |
1058 | } | |
1059 | ||
5212e11f VV |
1060 | /* |
1061 | * This function calculates the arena in which the given LBA lies | |
1062 | * by doing a linear walk. This is acceptable since we expect only | |
1063 | * a few arenas. If we have backing devices that get much larger, | |
1064 | * we can construct a balanced binary tree of arenas at init time | |
1065 | * so that this range search becomes faster. | |
1066 | */ | |
1067 | static int lba_to_arena(struct btt *btt, sector_t sector, __u32 *premap, | |
1068 | struct arena_info **arena) | |
1069 | { | |
1070 | struct arena_info *arena_list; | |
1071 | __u64 lba = div_u64(sector << SECTOR_SHIFT, btt->sector_size); | |
1072 | ||
1073 | list_for_each_entry(arena_list, &btt->arena_list, list) { | |
1074 | if (lba < arena_list->external_nlba) { | |
1075 | *arena = arena_list; | |
1076 | *premap = lba; | |
1077 | return 0; | |
1078 | } | |
1079 | lba -= arena_list->external_nlba; | |
1080 | } | |
1081 | ||
1082 | return -EIO; | |
1083 | } | |
1084 | ||
1085 | /* | |
1086 | * The following (lock_map, unlock_map) are mostly just to improve | |
1087 | * readability, since they index into an array of locks | |
1088 | */ | |
1089 | static void lock_map(struct arena_info *arena, u32 premap) | |
1090 | __acquires(&arena->map_locks[idx].lock) | |
1091 | { | |
1092 | u32 idx = (premap * MAP_ENT_SIZE / L1_CACHE_BYTES) % arena->nfree; | |
1093 | ||
1094 | spin_lock(&arena->map_locks[idx].lock); | |
1095 | } | |
1096 | ||
1097 | static void unlock_map(struct arena_info *arena, u32 premap) | |
1098 | __releases(&arena->map_locks[idx].lock) | |
1099 | { | |
1100 | u32 idx = (premap * MAP_ENT_SIZE / L1_CACHE_BYTES) % arena->nfree; | |
1101 | ||
1102 | spin_unlock(&arena->map_locks[idx].lock); | |
1103 | } | |
1104 | ||
5212e11f VV |
1105 | static int btt_data_read(struct arena_info *arena, struct page *page, |
1106 | unsigned int off, u32 lba, u32 len) | |
1107 | { | |
1108 | int ret; | |
1109 | u64 nsoff = to_namespace_offset(arena, lba); | |
1110 | void *mem = kmap_atomic(page); | |
1111 | ||
3ae3d67b | 1112 | ret = arena_read_bytes(arena, nsoff, mem + off, len, NVDIMM_IO_ATOMIC); |
5212e11f VV |
1113 | kunmap_atomic(mem); |
1114 | ||
1115 | return ret; | |
1116 | } | |
1117 | ||
1118 | static int btt_data_write(struct arena_info *arena, u32 lba, | |
1119 | struct page *page, unsigned int off, u32 len) | |
1120 | { | |
1121 | int ret; | |
1122 | u64 nsoff = to_namespace_offset(arena, lba); | |
1123 | void *mem = kmap_atomic(page); | |
1124 | ||
3ae3d67b | 1125 | ret = arena_write_bytes(arena, nsoff, mem + off, len, NVDIMM_IO_ATOMIC); |
5212e11f VV |
1126 | kunmap_atomic(mem); |
1127 | ||
1128 | return ret; | |
1129 | } | |
1130 | ||
1131 | static void zero_fill_data(struct page *page, unsigned int off, u32 len) | |
1132 | { | |
1133 | void *mem = kmap_atomic(page); | |
1134 | ||
1135 | memset(mem + off, 0, len); | |
1136 | kunmap_atomic(mem); | |
1137 | } | |
1138 | ||
41cd8b70 VV |
1139 | #ifdef CONFIG_BLK_DEV_INTEGRITY |
1140 | static int btt_rw_integrity(struct btt *btt, struct bio_integrity_payload *bip, | |
1141 | struct arena_info *arena, u32 postmap, int rw) | |
1142 | { | |
1143 | unsigned int len = btt_meta_size(btt); | |
1144 | u64 meta_nsoff; | |
1145 | int ret = 0; | |
1146 | ||
1147 | if (bip == NULL) | |
1148 | return 0; | |
1149 | ||
1150 | meta_nsoff = to_namespace_offset(arena, postmap) + btt->sector_size; | |
1151 | ||
1152 | while (len) { | |
1153 | unsigned int cur_len; | |
1154 | struct bio_vec bv; | |
1155 | void *mem; | |
1156 | ||
1157 | bv = bvec_iter_bvec(bip->bip_vec, bip->bip_iter); | |
1158 | /* | |
1159 | * The 'bv' obtained from bvec_iter_bvec has its .bv_len and | |
1160 | * .bv_offset already adjusted for iter->bi_bvec_done, and we | |
1161 | * can use those directly | |
1162 | */ | |
1163 | ||
1164 | cur_len = min(len, bv.bv_len); | |
1165 | mem = kmap_atomic(bv.bv_page); | |
1166 | if (rw) | |
1167 | ret = arena_write_bytes(arena, meta_nsoff, | |
3ae3d67b VV |
1168 | mem + bv.bv_offset, cur_len, |
1169 | NVDIMM_IO_ATOMIC); | |
41cd8b70 VV |
1170 | else |
1171 | ret = arena_read_bytes(arena, meta_nsoff, | |
3ae3d67b VV |
1172 | mem + bv.bv_offset, cur_len, |
1173 | NVDIMM_IO_ATOMIC); | |
41cd8b70 VV |
1174 | |
1175 | kunmap_atomic(mem); | |
1176 | if (ret) | |
1177 | return ret; | |
1178 | ||
1179 | len -= cur_len; | |
1180 | meta_nsoff += cur_len; | |
b1fb2c52 DM |
1181 | if (!bvec_iter_advance(bip->bip_vec, &bip->bip_iter, cur_len)) |
1182 | return -EIO; | |
41cd8b70 VV |
1183 | } |
1184 | ||
1185 | return ret; | |
1186 | } | |
1187 | ||
1188 | #else /* CONFIG_BLK_DEV_INTEGRITY */ | |
1189 | static int btt_rw_integrity(struct btt *btt, struct bio_integrity_payload *bip, | |
1190 | struct arena_info *arena, u32 postmap, int rw) | |
1191 | { | |
1192 | return 0; | |
1193 | } | |
1194 | #endif | |
1195 | ||
1196 | static int btt_read_pg(struct btt *btt, struct bio_integrity_payload *bip, | |
1197 | struct page *page, unsigned int off, sector_t sector, | |
1198 | unsigned int len) | |
5212e11f VV |
1199 | { |
1200 | int ret = 0; | |
1201 | int t_flag, e_flag; | |
1202 | struct arena_info *arena = NULL; | |
1203 | u32 lane = 0, premap, postmap; | |
1204 | ||
1205 | while (len) { | |
1206 | u32 cur_len; | |
1207 | ||
1208 | lane = nd_region_acquire_lane(btt->nd_region); | |
1209 | ||
1210 | ret = lba_to_arena(btt, sector, &premap, &arena); | |
1211 | if (ret) | |
1212 | goto out_lane; | |
1213 | ||
1214 | cur_len = min(btt->sector_size, len); | |
1215 | ||
3ae3d67b VV |
1216 | ret = btt_map_read(arena, premap, &postmap, &t_flag, &e_flag, |
1217 | NVDIMM_IO_ATOMIC); | |
5212e11f VV |
1218 | if (ret) |
1219 | goto out_lane; | |
1220 | ||
1221 | /* | |
1222 | * We loop to make sure that the post map LBA didn't change | |
1223 | * from under us between writing the RTT and doing the actual | |
1224 | * read. | |
1225 | */ | |
1226 | while (1) { | |
1227 | u32 new_map; | |
1398199d | 1228 | int new_t, new_e; |
5212e11f VV |
1229 | |
1230 | if (t_flag) { | |
1231 | zero_fill_data(page, off, cur_len); | |
1232 | goto out_lane; | |
1233 | } | |
1234 | ||
1235 | if (e_flag) { | |
1236 | ret = -EIO; | |
1237 | goto out_lane; | |
1238 | } | |
1239 | ||
1240 | arena->rtt[lane] = RTT_VALID | postmap; | |
1241 | /* | |
1242 | * Barrier to make sure this write is not reordered | |
1243 | * to do the verification map_read before the RTT store | |
1244 | */ | |
1245 | barrier(); | |
1246 | ||
1398199d VV |
1247 | ret = btt_map_read(arena, premap, &new_map, &new_t, |
1248 | &new_e, NVDIMM_IO_ATOMIC); | |
5212e11f VV |
1249 | if (ret) |
1250 | goto out_rtt; | |
1251 | ||
1398199d VV |
1252 | if ((postmap == new_map) && (t_flag == new_t) && |
1253 | (e_flag == new_e)) | |
5212e11f VV |
1254 | break; |
1255 | ||
1256 | postmap = new_map; | |
1398199d VV |
1257 | t_flag = new_t; |
1258 | e_flag = new_e; | |
5212e11f VV |
1259 | } |
1260 | ||
1261 | ret = btt_data_read(arena, page, off, postmap, cur_len); | |
d9b83c75 VV |
1262 | if (ret) { |
1263 | int rc; | |
1264 | ||
1265 | /* Media error - set the e_flag */ | |
1266 | rc = btt_map_write(arena, premap, postmap, 0, 1, | |
1267 | NVDIMM_IO_ATOMIC); | |
5212e11f | 1268 | goto out_rtt; |
d9b83c75 | 1269 | } |
5212e11f | 1270 | |
41cd8b70 VV |
1271 | if (bip) { |
1272 | ret = btt_rw_integrity(btt, bip, arena, postmap, READ); | |
1273 | if (ret) | |
1274 | goto out_rtt; | |
1275 | } | |
1276 | ||
5212e11f VV |
1277 | arena->rtt[lane] = RTT_INVALID; |
1278 | nd_region_release_lane(btt->nd_region, lane); | |
1279 | ||
1280 | len -= cur_len; | |
1281 | off += cur_len; | |
1282 | sector += btt->sector_size >> SECTOR_SHIFT; | |
1283 | } | |
1284 | ||
1285 | return 0; | |
1286 | ||
1287 | out_rtt: | |
1288 | arena->rtt[lane] = RTT_INVALID; | |
1289 | out_lane: | |
1290 | nd_region_release_lane(btt->nd_region, lane); | |
1291 | return ret; | |
1292 | } | |
1293 | ||
d9b83c75 VV |
1294 | /* |
1295 | * Normally, arena_{read,write}_bytes will take care of the initial offset | |
1296 | * adjustment, but in the case of btt_is_badblock, where we query is_bad_pmem, | |
1297 | * we need the final, raw namespace offset here | |
1298 | */ | |
1299 | static bool btt_is_badblock(struct btt *btt, struct arena_info *arena, | |
1300 | u32 postmap) | |
1301 | { | |
1302 | u64 nsoff = adjust_initial_offset(arena->nd_btt, | |
1303 | to_namespace_offset(arena, postmap)); | |
1304 | sector_t phys_sector = nsoff >> 9; | |
1305 | ||
1306 | return is_bad_pmem(btt->phys_bb, phys_sector, arena->internal_lbasize); | |
1307 | } | |
1308 | ||
41cd8b70 VV |
1309 | static int btt_write_pg(struct btt *btt, struct bio_integrity_payload *bip, |
1310 | sector_t sector, struct page *page, unsigned int off, | |
1311 | unsigned int len) | |
5212e11f VV |
1312 | { |
1313 | int ret = 0; | |
1314 | struct arena_info *arena = NULL; | |
1315 | u32 premap = 0, old_postmap, new_postmap, lane = 0, i; | |
1316 | struct log_entry log; | |
1317 | int sub; | |
1318 | ||
1319 | while (len) { | |
1320 | u32 cur_len; | |
d9b83c75 | 1321 | int e_flag; |
5212e11f | 1322 | |
d9b83c75 | 1323 | retry: |
5212e11f VV |
1324 | lane = nd_region_acquire_lane(btt->nd_region); |
1325 | ||
1326 | ret = lba_to_arena(btt, sector, &premap, &arena); | |
1327 | if (ret) | |
1328 | goto out_lane; | |
1329 | cur_len = min(btt->sector_size, len); | |
1330 | ||
1331 | if ((arena->flags & IB_FLAG_ERROR_MASK) != 0) { | |
1332 | ret = -EIO; | |
1333 | goto out_lane; | |
1334 | } | |
1335 | ||
d9b83c75 VV |
1336 | if (btt_is_badblock(btt, arena, arena->freelist[lane].block)) |
1337 | arena->freelist[lane].has_err = 1; | |
1338 | ||
1339 | if (mutex_is_locked(&arena->err_lock) | |
1340 | || arena->freelist[lane].has_err) { | |
1341 | nd_region_release_lane(btt->nd_region, lane); | |
1342 | ||
1343 | ret = arena_clear_freelist_error(arena, lane); | |
1344 | if (ret) | |
1345 | return ret; | |
1346 | ||
1347 | /* OK to acquire a different lane/free block */ | |
1348 | goto retry; | |
1349 | } | |
1350 | ||
5212e11f VV |
1351 | new_postmap = arena->freelist[lane].block; |
1352 | ||
1353 | /* Wait if the new block is being read from */ | |
1354 | for (i = 0; i < arena->nfree; i++) | |
1355 | while (arena->rtt[i] == (RTT_VALID | new_postmap)) | |
1356 | cpu_relax(); | |
1357 | ||
1358 | ||
1359 | if (new_postmap >= arena->internal_nlba) { | |
1360 | ret = -EIO; | |
1361 | goto out_lane; | |
41cd8b70 VV |
1362 | } |
1363 | ||
1364 | ret = btt_data_write(arena, new_postmap, page, off, cur_len); | |
5212e11f VV |
1365 | if (ret) |
1366 | goto out_lane; | |
1367 | ||
41cd8b70 VV |
1368 | if (bip) { |
1369 | ret = btt_rw_integrity(btt, bip, arena, new_postmap, | |
1370 | WRITE); | |
1371 | if (ret) | |
1372 | goto out_lane; | |
1373 | } | |
1374 | ||
5212e11f | 1375 | lock_map(arena, premap); |
d9b83c75 | 1376 | ret = btt_map_read(arena, premap, &old_postmap, NULL, &e_flag, |
3ae3d67b | 1377 | NVDIMM_IO_ATOMIC); |
5212e11f VV |
1378 | if (ret) |
1379 | goto out_map; | |
1380 | if (old_postmap >= arena->internal_nlba) { | |
1381 | ret = -EIO; | |
1382 | goto out_map; | |
1383 | } | |
d9b83c75 VV |
1384 | if (e_flag) |
1385 | set_e_flag(old_postmap); | |
5212e11f VV |
1386 | |
1387 | log.lba = cpu_to_le32(premap); | |
1388 | log.old_map = cpu_to_le32(old_postmap); | |
1389 | log.new_map = cpu_to_le32(new_postmap); | |
1390 | log.seq = cpu_to_le32(arena->freelist[lane].seq); | |
1391 | sub = arena->freelist[lane].sub; | |
1392 | ret = btt_flog_write(arena, lane, sub, &log); | |
1393 | if (ret) | |
1394 | goto out_map; | |
1395 | ||
1db1f3ce VV |
1396 | ret = btt_map_write(arena, premap, new_postmap, 0, 0, |
1397 | NVDIMM_IO_ATOMIC); | |
5212e11f VV |
1398 | if (ret) |
1399 | goto out_map; | |
1400 | ||
1401 | unlock_map(arena, premap); | |
1402 | nd_region_release_lane(btt->nd_region, lane); | |
1403 | ||
d9b83c75 VV |
1404 | if (e_flag) { |
1405 | ret = arena_clear_freelist_error(arena, lane); | |
1406 | if (ret) | |
1407 | return ret; | |
1408 | } | |
1409 | ||
5212e11f VV |
1410 | len -= cur_len; |
1411 | off += cur_len; | |
1412 | sector += btt->sector_size >> SECTOR_SHIFT; | |
1413 | } | |
1414 | ||
1415 | return 0; | |
1416 | ||
1417 | out_map: | |
1418 | unlock_map(arena, premap); | |
1419 | out_lane: | |
1420 | nd_region_release_lane(btt->nd_region, lane); | |
1421 | return ret; | |
1422 | } | |
1423 | ||
41cd8b70 VV |
1424 | static int btt_do_bvec(struct btt *btt, struct bio_integrity_payload *bip, |
1425 | struct page *page, unsigned int len, unsigned int off, | |
c11f0c0b | 1426 | bool is_write, sector_t sector) |
5212e11f VV |
1427 | { |
1428 | int ret; | |
1429 | ||
c11f0c0b | 1430 | if (!is_write) { |
41cd8b70 | 1431 | ret = btt_read_pg(btt, bip, page, off, sector, len); |
5212e11f VV |
1432 | flush_dcache_page(page); |
1433 | } else { | |
1434 | flush_dcache_page(page); | |
41cd8b70 | 1435 | ret = btt_write_pg(btt, bip, sector, page, off, len); |
5212e11f VV |
1436 | } |
1437 | ||
1438 | return ret; | |
1439 | } | |
1440 | ||
dece1635 | 1441 | static blk_qc_t btt_make_request(struct request_queue *q, struct bio *bio) |
5212e11f | 1442 | { |
41cd8b70 | 1443 | struct bio_integrity_payload *bip = bio_integrity(bio); |
5212e11f VV |
1444 | struct btt *btt = q->queuedata; |
1445 | struct bvec_iter iter; | |
f0dc089c | 1446 | unsigned long start; |
5212e11f | 1447 | struct bio_vec bvec; |
abf54548 | 1448 | int err = 0; |
f0dc089c | 1449 | bool do_acct; |
5212e11f | 1450 | |
e23947bd DM |
1451 | if (!bio_integrity_prep(bio)) |
1452 | return BLK_QC_T_NONE; | |
41cd8b70 | 1453 | |
f0dc089c | 1454 | do_acct = nd_iostat_start(bio, &start); |
5212e11f VV |
1455 | bio_for_each_segment(bvec, bio, iter) { |
1456 | unsigned int len = bvec.bv_len; | |
1457 | ||
86652d2e VV |
1458 | if (len > PAGE_SIZE || len < btt->sector_size || |
1459 | len % btt->sector_size) { | |
1460 | dev_err_ratelimited(&btt->nd_btt->dev, | |
1461 | "unaligned bio segment (len: %d)\n", len); | |
1462 | bio->bi_status = BLK_STS_IOERR; | |
1463 | break; | |
1464 | } | |
5212e11f | 1465 | |
41cd8b70 | 1466 | err = btt_do_bvec(btt, bip, bvec.bv_page, len, bvec.bv_offset, |
c11f0c0b | 1467 | op_is_write(bio_op(bio)), iter.bi_sector); |
5212e11f | 1468 | if (err) { |
e6be2dcb | 1469 | dev_err(&btt->nd_btt->dev, |
5212e11f | 1470 | "io error in %s sector %lld, len %d,\n", |
abf54548 MC |
1471 | (op_is_write(bio_op(bio))) ? "WRITE" : |
1472 | "READ", | |
5212e11f | 1473 | (unsigned long long) iter.bi_sector, len); |
4e4cbee9 | 1474 | bio->bi_status = errno_to_blk_status(err); |
f0dc089c | 1475 | break; |
5212e11f VV |
1476 | } |
1477 | } | |
f0dc089c DW |
1478 | if (do_acct) |
1479 | nd_iostat_end(bio, start); | |
5212e11f | 1480 | |
4246a0b6 | 1481 | bio_endio(bio); |
dece1635 | 1482 | return BLK_QC_T_NONE; |
5212e11f VV |
1483 | } |
1484 | ||
1485 | static int btt_rw_page(struct block_device *bdev, sector_t sector, | |
c11f0c0b | 1486 | struct page *page, bool is_write) |
5212e11f VV |
1487 | { |
1488 | struct btt *btt = bdev->bd_disk->private_data; | |
c13c43d5 | 1489 | int rc; |
98cc093c | 1490 | unsigned int len; |
5212e11f | 1491 | |
98cc093c HY |
1492 | len = hpage_nr_pages(page) * PAGE_SIZE; |
1493 | rc = btt_do_bvec(btt, NULL, page, len, 0, is_write, sector); | |
c13c43d5 VV |
1494 | if (rc == 0) |
1495 | page_endio(page, is_write, 0); | |
1496 | ||
1497 | return rc; | |
5212e11f VV |
1498 | } |
1499 | ||
1500 | ||
1501 | static int btt_getgeo(struct block_device *bd, struct hd_geometry *geo) | |
1502 | { | |
1503 | /* some standard values */ | |
1504 | geo->heads = 1 << 6; | |
1505 | geo->sectors = 1 << 5; | |
1506 | geo->cylinders = get_capacity(bd->bd_disk) >> 11; | |
1507 | return 0; | |
1508 | } | |
1509 | ||
1510 | static const struct block_device_operations btt_fops = { | |
1511 | .owner = THIS_MODULE, | |
1512 | .rw_page = btt_rw_page, | |
1513 | .getgeo = btt_getgeo, | |
58138820 | 1514 | .revalidate_disk = nvdimm_revalidate_disk, |
5212e11f VV |
1515 | }; |
1516 | ||
1517 | static int btt_blk_init(struct btt *btt) | |
1518 | { | |
1519 | struct nd_btt *nd_btt = btt->nd_btt; | |
1520 | struct nd_namespace_common *ndns = nd_btt->ndns; | |
1521 | ||
1522 | /* create a new disk and request queue for btt */ | |
1523 | btt->btt_queue = blk_alloc_queue(GFP_KERNEL); | |
1524 | if (!btt->btt_queue) | |
1525 | return -ENOMEM; | |
1526 | ||
1527 | btt->btt_disk = alloc_disk(0); | |
1528 | if (!btt->btt_disk) { | |
1529 | blk_cleanup_queue(btt->btt_queue); | |
1530 | return -ENOMEM; | |
1531 | } | |
1532 | ||
1533 | nvdimm_namespace_disk_name(ndns, btt->btt_disk->disk_name); | |
5212e11f VV |
1534 | btt->btt_disk->first_minor = 0; |
1535 | btt->btt_disk->fops = &btt_fops; | |
1536 | btt->btt_disk->private_data = btt; | |
1537 | btt->btt_disk->queue = btt->btt_queue; | |
1538 | btt->btt_disk->flags = GENHD_FL_EXT_DEVT; | |
23c47d2a MK |
1539 | btt->btt_disk->queue->backing_dev_info->capabilities |= |
1540 | BDI_CAP_SYNCHRONOUS_IO; | |
5212e11f VV |
1541 | |
1542 | blk_queue_make_request(btt->btt_queue, btt_make_request); | |
1543 | blk_queue_logical_block_size(btt->btt_queue, btt->sector_size); | |
1544 | blk_queue_max_hw_sectors(btt->btt_queue, UINT_MAX); | |
5212e11f VV |
1545 | queue_flag_set_unlocked(QUEUE_FLAG_NONROT, btt->btt_queue); |
1546 | btt->btt_queue->queuedata = btt; | |
1547 | ||
41cd8b70 VV |
1548 | if (btt_meta_size(btt)) { |
1549 | int rc = nd_integrity_init(btt->btt_disk, btt_meta_size(btt)); | |
1550 | ||
1551 | if (rc) { | |
1552 | del_gendisk(btt->btt_disk); | |
1553 | put_disk(btt->btt_disk); | |
1554 | blk_cleanup_queue(btt->btt_queue); | |
1555 | return rc; | |
1556 | } | |
1557 | } | |
1558 | set_capacity(btt->btt_disk, btt->nlba * btt->sector_size >> 9); | |
3ffb0ba9 | 1559 | device_add_disk(&btt->nd_btt->dev, btt->btt_disk); |
abe8b4e3 | 1560 | btt->nd_btt->size = btt->nlba * (u64)btt->sector_size; |
58138820 | 1561 | revalidate_disk(btt->btt_disk); |
5212e11f VV |
1562 | |
1563 | return 0; | |
1564 | } | |
1565 | ||
1566 | static void btt_blk_cleanup(struct btt *btt) | |
1567 | { | |
1568 | del_gendisk(btt->btt_disk); | |
1569 | put_disk(btt->btt_disk); | |
1570 | blk_cleanup_queue(btt->btt_queue); | |
1571 | } | |
1572 | ||
1573 | /** | |
1574 | * btt_init - initialize a block translation table for the given device | |
1575 | * @nd_btt: device with BTT geometry and backing device info | |
1576 | * @rawsize: raw size in bytes of the backing device | |
1577 | * @lbasize: lba size of the backing device | |
1578 | * @uuid: A uuid for the backing device - this is stored on media | |
1579 | * @maxlane: maximum number of parallel requests the device can handle | |
1580 | * | |
1581 | * Initialize a Block Translation Table on a backing device to provide | |
1582 | * single sector power fail atomicity. | |
1583 | * | |
1584 | * Context: | |
1585 | * Might sleep. | |
1586 | * | |
1587 | * Returns: | |
1588 | * Pointer to a new struct btt on success, NULL on failure. | |
1589 | */ | |
1590 | static struct btt *btt_init(struct nd_btt *nd_btt, unsigned long long rawsize, | |
1591 | u32 lbasize, u8 *uuid, struct nd_region *nd_region) | |
1592 | { | |
1593 | int ret; | |
1594 | struct btt *btt; | |
d9b83c75 | 1595 | struct nd_namespace_io *nsio; |
5212e11f VV |
1596 | struct device *dev = &nd_btt->dev; |
1597 | ||
e32bc729 | 1598 | btt = devm_kzalloc(dev, sizeof(struct btt), GFP_KERNEL); |
5212e11f VV |
1599 | if (!btt) |
1600 | return NULL; | |
1601 | ||
1602 | btt->nd_btt = nd_btt; | |
1603 | btt->rawsize = rawsize; | |
1604 | btt->lbasize = lbasize; | |
1605 | btt->sector_size = ((lbasize >= 4096) ? 4096 : 512); | |
1606 | INIT_LIST_HEAD(&btt->arena_list); | |
1607 | mutex_init(&btt->init_lock); | |
1608 | btt->nd_region = nd_region; | |
d9b83c75 VV |
1609 | nsio = to_nd_namespace_io(&nd_btt->ndns->dev); |
1610 | btt->phys_bb = &nsio->bb; | |
5212e11f VV |
1611 | |
1612 | ret = discover_arenas(btt); | |
1613 | if (ret) { | |
1614 | dev_err(dev, "init: error in arena_discover: %d\n", ret); | |
e32bc729 | 1615 | return NULL; |
5212e11f VV |
1616 | } |
1617 | ||
58138820 | 1618 | if (btt->init_state != INIT_READY && nd_region->ro) { |
e6be2dcb | 1619 | dev_warn(dev, "%s is read-only, unable to init btt metadata\n", |
58138820 | 1620 | dev_name(&nd_region->dev)); |
e32bc729 | 1621 | return NULL; |
58138820 | 1622 | } else if (btt->init_state != INIT_READY) { |
5212e11f VV |
1623 | btt->num_arenas = (rawsize / ARENA_MAX_SIZE) + |
1624 | ((rawsize % ARENA_MAX_SIZE) ? 1 : 0); | |
1625 | dev_dbg(dev, "init: %d arenas for %llu rawsize\n", | |
1626 | btt->num_arenas, rawsize); | |
1627 | ||
1628 | ret = create_arenas(btt); | |
1629 | if (ret) { | |
1630 | dev_info(dev, "init: create_arenas: %d\n", ret); | |
e32bc729 | 1631 | return NULL; |
5212e11f VV |
1632 | } |
1633 | ||
1634 | ret = btt_meta_init(btt); | |
1635 | if (ret) { | |
1636 | dev_err(dev, "init: error in meta_init: %d\n", ret); | |
e32bc729 | 1637 | return NULL; |
5212e11f VV |
1638 | } |
1639 | } | |
1640 | ||
1641 | ret = btt_blk_init(btt); | |
1642 | if (ret) { | |
1643 | dev_err(dev, "init: error in blk_init: %d\n", ret); | |
e32bc729 | 1644 | return NULL; |
5212e11f VV |
1645 | } |
1646 | ||
1647 | btt_debugfs_init(btt); | |
1648 | ||
1649 | return btt; | |
5212e11f VV |
1650 | } |
1651 | ||
1652 | /** | |
1653 | * btt_fini - de-initialize a BTT | |
1654 | * @btt: the BTT handle that was generated by btt_init | |
1655 | * | |
1656 | * De-initialize a Block Translation Table on device removal | |
1657 | * | |
1658 | * Context: | |
1659 | * Might sleep. | |
1660 | */ | |
1661 | static void btt_fini(struct btt *btt) | |
1662 | { | |
1663 | if (btt) { | |
1664 | btt_blk_cleanup(btt); | |
1665 | free_arenas(btt); | |
1666 | debugfs_remove_recursive(btt->debugfs_dir); | |
5212e11f VV |
1667 | } |
1668 | } | |
1669 | ||
1670 | int nvdimm_namespace_attach_btt(struct nd_namespace_common *ndns) | |
1671 | { | |
1672 | struct nd_btt *nd_btt = to_nd_btt(ndns->claim); | |
1673 | struct nd_region *nd_region; | |
14e49454 | 1674 | struct btt_sb *btt_sb; |
5212e11f VV |
1675 | struct btt *btt; |
1676 | size_t rawsize; | |
1677 | ||
9dec4892 DW |
1678 | if (!nd_btt->uuid || !nd_btt->ndns || !nd_btt->lbasize) { |
1679 | dev_dbg(&nd_btt->dev, "incomplete btt configuration\n"); | |
5212e11f | 1680 | return -ENODEV; |
9dec4892 | 1681 | } |
5212e11f | 1682 | |
14e49454 | 1683 | btt_sb = devm_kzalloc(&nd_btt->dev, sizeof(*btt_sb), GFP_KERNEL); |
ed36b4db CJ |
1684 | if (!btt_sb) |
1685 | return -ENOMEM; | |
14e49454 VV |
1686 | |
1687 | /* | |
1688 | * If this returns < 0, that is ok as it just means there wasn't | |
1689 | * an existing BTT, and we're creating a new one. We still need to | |
1690 | * call this as we need the version dependent fields in nd_btt to be | |
1691 | * set correctly based on the holder class | |
1692 | */ | |
1693 | nd_btt_version(nd_btt, ndns, btt_sb); | |
1694 | ||
1695 | rawsize = nvdimm_namespace_capacity(ndns) - nd_btt->initial_offset; | |
5212e11f | 1696 | if (rawsize < ARENA_MIN_SIZE) { |
9dec4892 | 1697 | dev_dbg(&nd_btt->dev, "%s must be at least %ld bytes\n", |
14e49454 VV |
1698 | dev_name(&ndns->dev), |
1699 | ARENA_MIN_SIZE + nd_btt->initial_offset); | |
5212e11f VV |
1700 | return -ENXIO; |
1701 | } | |
1702 | nd_region = to_nd_region(nd_btt->dev.parent); | |
1703 | btt = btt_init(nd_btt, rawsize, nd_btt->lbasize, nd_btt->uuid, | |
1704 | nd_region); | |
1705 | if (!btt) | |
1706 | return -ENOMEM; | |
1707 | nd_btt->btt = btt; | |
1708 | ||
1709 | return 0; | |
1710 | } | |
1711 | EXPORT_SYMBOL(nvdimm_namespace_attach_btt); | |
1712 | ||
298f2bc5 | 1713 | int nvdimm_namespace_detach_btt(struct nd_btt *nd_btt) |
5212e11f | 1714 | { |
5212e11f VV |
1715 | struct btt *btt = nd_btt->btt; |
1716 | ||
1717 | btt_fini(btt); | |
1718 | nd_btt->btt = NULL; | |
1719 | ||
1720 | return 0; | |
1721 | } | |
1722 | EXPORT_SYMBOL(nvdimm_namespace_detach_btt); | |
1723 | ||
1724 | static int __init nd_btt_init(void) | |
1725 | { | |
ff8e92d5 | 1726 | int rc = 0; |
5212e11f VV |
1727 | |
1728 | debugfs_root = debugfs_create_dir("btt", NULL); | |
ff8e92d5 | 1729 | if (IS_ERR_OR_NULL(debugfs_root)) |
5212e11f | 1730 | rc = -ENXIO; |
5212e11f VV |
1731 | |
1732 | return rc; | |
1733 | } | |
1734 | ||
1735 | static void __exit nd_btt_exit(void) | |
1736 | { | |
1737 | debugfs_remove_recursive(debugfs_root); | |
5212e11f VV |
1738 | } |
1739 | ||
1740 | MODULE_ALIAS_ND_DEVICE(ND_DEVICE_BTT); | |
1741 | MODULE_AUTHOR("Vishal Verma <vishal.l.verma@linux.intel.com>"); | |
1742 | MODULE_LICENSE("GPL v2"); | |
1743 | module_init(nd_btt_init); | |
1744 | module_exit(nd_btt_exit); |