Commit | Line | Data |
---|---|---|
1c6fdbd8 KO |
1 | /* SPDX-License-Identifier: GPL-2.0 */ |
2 | #ifndef _BCACHEFS_FORMAT_H | |
3 | #define _BCACHEFS_FORMAT_H | |
4 | ||
5 | /* | |
6 | * bcachefs on disk data structures | |
7 | * | |
8 | * OVERVIEW: | |
9 | * | |
10 | * There are three main types of on disk data structures in bcachefs (this is | |
11 | * reduced from 5 in bcache) | |
12 | * | |
13 | * - superblock | |
14 | * - journal | |
15 | * - btree | |
16 | * | |
17 | * The btree is the primary structure; most metadata exists as keys in the | |
18 | * various btrees. There are only a small number of btrees, they're not | |
19 | * sharded - we have one btree for extents, another for inodes, et cetera. | |
20 | * | |
21 | * SUPERBLOCK: | |
22 | * | |
23 | * The superblock contains the location of the journal, the list of devices in | |
24 | * the filesystem, and in general any metadata we need in order to decide | |
25 | * whether we can start a filesystem or prior to reading the journal/btree | |
26 | * roots. | |
27 | * | |
28 | * The superblock is extensible, and most of the contents of the superblock are | |
29 | * in variable length, type tagged fields; see struct bch_sb_field. | |
30 | * | |
31 | * Backup superblocks do not reside in a fixed location; also, superblocks do | |
32 | * not have a fixed size. To locate backup superblocks we have struct | |
33 | * bch_sb_layout; we store a copy of this inside every superblock, and also | |
34 | * before the first superblock. | |
35 | * | |
36 | * JOURNAL: | |
37 | * | |
38 | * The journal primarily records btree updates in the order they occurred; | |
39 | * journal replay consists of just iterating over all the keys in the open | |
40 | * journal entries and re-inserting them into the btrees. | |
41 | * | |
42 | * The journal also contains entry types for the btree roots, and blacklisted | |
43 | * journal sequence numbers (see journal_seq_blacklist.c). | |
44 | * | |
45 | * BTREE: | |
46 | * | |
47 | * bcachefs btrees are copy on write b+ trees, where nodes are big (typically | |
48 | * 128k-256k) and log structured. We use struct btree_node for writing the first | |
49 | * entry in a given node (offset 0), and struct btree_node_entry for all | |
50 | * subsequent writes. | |
51 | * | |
52 | * After the header, btree node entries contain a list of keys in sorted order. | |
53 | * Values are stored inline with the keys; since values are variable length (and | |
54 | * keys effectively are variable length too, due to packing) we can't do random | |
55 | * access without building up additional in memory tables in the btree node read | |
56 | * path. | |
57 | * | |
58 | * BTREE KEYS (struct bkey): | |
59 | * | |
60 | * The various btrees share a common format for the key - so as to avoid | |
61 | * switching in fastpath lookup/comparison code - but define their own | |
62 | * structures for the key values. | |
63 | * | |
64 | * The size of a key/value pair is stored as a u8 in units of u64s, so the max | |
65 | * size is just under 2k. The common part also contains a type tag for the | |
66 | * value, and a format field indicating whether the key is packed or not (and | |
67 | * also meant to allow adding new key fields in the future, if desired). | |
68 | * | |
69 | * bkeys, when stored within a btree node, may also be packed. In that case, the | |
70 | * bkey_format in that node is used to unpack it. Packed bkeys mean that we can | |
71 | * be generous with field sizes in the common part of the key format (64 bit | |
72 | * inode number, 64 bit offset, 96 bit version field, etc.) for negligible cost. | |
73 | */ | |
74 | ||
75 | #include <asm/types.h> | |
76 | #include <asm/byteorder.h> | |
7121643e | 77 | #include <linux/kernel.h> |
1c6fdbd8 | 78 | #include <linux/uuid.h> |
528b18e6 | 79 | #include "vstructs.h" |
1c6fdbd8 KO |
80 | |
81 | #ifdef __KERNEL__ | |
82 | typedef uuid_t __uuid_t; | |
83 | #endif | |
84 | ||
3d48a7f8 | 85 | #define BITMASK(name, type, field, offset, end) \ |
96dea3d5 KO |
86 | static const __maybe_unused unsigned name##_OFFSET = offset; \ |
87 | static const __maybe_unused unsigned name##_BITS = (end - offset); \ | |
3d48a7f8 KO |
88 | \ |
89 | static inline __u64 name(const type *k) \ | |
90 | { \ | |
91 | return (k->field >> offset) & ~(~0ULL << (end - offset)); \ | |
92 | } \ | |
93 | \ | |
94 | static inline void SET_##name(type *k, __u64 v) \ | |
95 | { \ | |
96 | k->field &= ~(~(~0ULL << (end - offset)) << offset); \ | |
97 | k->field |= (v & ~(~0ULL << (end - offset))) << offset; \ | |
98 | } | |
99 | ||
1c6fdbd8 | 100 | #define LE_BITMASK(_bits, name, type, field, offset, end) \ |
96dea3d5 KO |
101 | static const __maybe_unused unsigned name##_OFFSET = offset; \ |
102 | static const __maybe_unused unsigned name##_BITS = (end - offset); \ | |
103 | static const __maybe_unused __u##_bits name##_MAX = (1ULL << (end - offset)) - 1;\ | |
1c6fdbd8 KO |
104 | \ |
105 | static inline __u64 name(const type *k) \ | |
106 | { \ | |
107 | return (__le##_bits##_to_cpu(k->field) >> offset) & \ | |
108 | ~(~0ULL << (end - offset)); \ | |
109 | } \ | |
110 | \ | |
111 | static inline void SET_##name(type *k, __u64 v) \ | |
112 | { \ | |
113 | __u##_bits new = __le##_bits##_to_cpu(k->field); \ | |
114 | \ | |
115 | new &= ~(~(~0ULL << (end - offset)) << offset); \ | |
116 | new |= (v & ~(~0ULL << (end - offset))) << offset; \ | |
117 | k->field = __cpu_to_le##_bits(new); \ | |
118 | } | |
119 | ||
120 | #define LE16_BITMASK(n, t, f, o, e) LE_BITMASK(16, n, t, f, o, e) | |
121 | #define LE32_BITMASK(n, t, f, o, e) LE_BITMASK(32, n, t, f, o, e) | |
122 | #define LE64_BITMASK(n, t, f, o, e) LE_BITMASK(64, n, t, f, o, e) | |
123 | ||
124 | struct bkey_format { | |
125 | __u8 key_u64s; | |
126 | __u8 nr_fields; | |
127 | /* One unused slot for now: */ | |
128 | __u8 bits_per_field[6]; | |
129 | __le64 field_offset[6]; | |
130 | }; | |
131 | ||
132 | /* Btree keys - all units are in sectors */ | |
133 | ||
134 | struct bpos { | |
135 | /* | |
136 | * Word order matches machine byte order - btree code treats a bpos as a | |
137 | * single large integer, for search/comparison purposes | |
138 | * | |
139 | * Note that wherever a bpos is embedded in another on disk data | |
140 | * structure, it has to be byte swabbed when reading in metadata that | |
141 | * wasn't written in native endian order: | |
142 | */ | |
143 | #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ | |
144 | __u32 snapshot; | |
145 | __u64 offset; | |
146 | __u64 inode; | |
147 | #elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ | |
148 | __u64 inode; | |
149 | __u64 offset; /* Points to end of extent - sectors */ | |
150 | __u32 snapshot; | |
151 | #else | |
152 | #error edit for your odd byteorder. | |
153 | #endif | |
3f3ae125 KO |
154 | } __packed |
155 | #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ | |
156 | __aligned(4) | |
157 | #endif | |
158 | ; | |
1c6fdbd8 KO |
159 | |
160 | #define KEY_INODE_MAX ((__u64)~0ULL) | |
161 | #define KEY_OFFSET_MAX ((__u64)~0ULL) | |
162 | #define KEY_SNAPSHOT_MAX ((__u32)~0U) | |
163 | #define KEY_SIZE_MAX ((__u32)~0U) | |
164 | ||
e751c01a | 165 | static inline struct bpos SPOS(__u64 inode, __u64 offset, __u32 snapshot) |
1c6fdbd8 | 166 | { |
e751c01a KO |
167 | return (struct bpos) { |
168 | .inode = inode, | |
169 | .offset = offset, | |
170 | .snapshot = snapshot, | |
171 | }; | |
1c6fdbd8 KO |
172 | } |
173 | ||
e751c01a | 174 | #define POS_MIN SPOS(0, 0, 0) |
618b1c0e KO |
175 | #define POS_MAX SPOS(KEY_INODE_MAX, KEY_OFFSET_MAX, 0) |
176 | #define SPOS_MAX SPOS(KEY_INODE_MAX, KEY_OFFSET_MAX, KEY_SNAPSHOT_MAX) | |
e751c01a | 177 | #define POS(_inode, _offset) SPOS(_inode, _offset, 0) |
1c6fdbd8 KO |
178 | |
179 | /* Empty placeholder struct, for container_of() */ | |
180 | struct bch_val { | |
181 | __u64 __nothing[0]; | |
182 | }; | |
183 | ||
184 | struct bversion { | |
185 | #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ | |
186 | __u64 lo; | |
187 | __u32 hi; | |
188 | #elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ | |
189 | __u32 hi; | |
190 | __u64 lo; | |
191 | #endif | |
fd0c7679 | 192 | } __packed __aligned(4); |
1c6fdbd8 KO |
193 | |
194 | struct bkey { | |
195 | /* Size of combined key and value, in u64s */ | |
196 | __u8 u64s; | |
197 | ||
198 | /* Format of key (0 for format local to btree node) */ | |
199 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
200 | __u8 format:7, | |
201 | needs_whiteout:1; | |
202 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
203 | __u8 needs_whiteout:1, | |
204 | format:7; | |
205 | #else | |
206 | #error edit for your odd byteorder. | |
207 | #endif | |
208 | ||
209 | /* Type of the value */ | |
210 | __u8 type; | |
211 | ||
212 | #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ | |
213 | __u8 pad[1]; | |
214 | ||
215 | struct bversion version; | |
216 | __u32 size; /* extent size, in sectors */ | |
217 | struct bpos p; | |
218 | #elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ | |
219 | struct bpos p; | |
220 | __u32 size; /* extent size, in sectors */ | |
221 | struct bversion version; | |
222 | ||
223 | __u8 pad[1]; | |
224 | #endif | |
fd0c7679 | 225 | } __packed __aligned(8); |
1c6fdbd8 KO |
226 | |
227 | struct bkey_packed { | |
228 | __u64 _data[0]; | |
229 | ||
230 | /* Size of combined key and value, in u64s */ | |
231 | __u8 u64s; | |
232 | ||
233 | /* Format of key (0 for format local to btree node) */ | |
234 | ||
235 | /* | |
236 | * XXX: next incompat on disk format change, switch format and | |
237 | * needs_whiteout - bkey_packed() will be cheaper if format is the high | |
238 | * bits of the bitfield | |
239 | */ | |
240 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
241 | __u8 format:7, | |
242 | needs_whiteout:1; | |
243 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
244 | __u8 needs_whiteout:1, | |
245 | format:7; | |
246 | #endif | |
247 | ||
248 | /* Type of the value */ | |
249 | __u8 type; | |
250 | __u8 key_start[0]; | |
251 | ||
252 | /* | |
253 | * We copy bkeys with struct assignment in various places, and while | |
254 | * that shouldn't be done with packed bkeys we can't disallow it in C, | |
255 | * and it's legal to cast a bkey to a bkey_packed - so padding it out | |
256 | * to the same size as struct bkey should hopefully be safest. | |
257 | */ | |
258 | __u8 pad[sizeof(struct bkey) - 3]; | |
fd0c7679 | 259 | } __packed __aligned(8); |
1c6fdbd8 | 260 | |
653693be KO |
261 | typedef struct { |
262 | __le64 lo; | |
263 | __le64 hi; | |
264 | } bch_le128; | |
265 | ||
1c6fdbd8 | 266 | #define BKEY_U64s (sizeof(struct bkey) / sizeof(__u64)) |
cd575ddf KO |
267 | #define BKEY_U64s_MAX U8_MAX |
268 | #define BKEY_VAL_U64s_MAX (BKEY_U64s_MAX - BKEY_U64s) | |
269 | ||
1c6fdbd8 KO |
270 | #define KEY_PACKED_BITS_START 24 |
271 | ||
272 | #define KEY_FORMAT_LOCAL_BTREE 0 | |
273 | #define KEY_FORMAT_CURRENT 1 | |
274 | ||
275 | enum bch_bkey_fields { | |
276 | BKEY_FIELD_INODE, | |
277 | BKEY_FIELD_OFFSET, | |
278 | BKEY_FIELD_SNAPSHOT, | |
279 | BKEY_FIELD_SIZE, | |
280 | BKEY_FIELD_VERSION_HI, | |
281 | BKEY_FIELD_VERSION_LO, | |
282 | BKEY_NR_FIELDS, | |
283 | }; | |
284 | ||
285 | #define bkey_format_field(name, field) \ | |
286 | [BKEY_FIELD_##name] = (sizeof(((struct bkey *) NULL)->field) * 8) | |
287 | ||
288 | #define BKEY_FORMAT_CURRENT \ | |
289 | ((struct bkey_format) { \ | |
290 | .key_u64s = BKEY_U64s, \ | |
291 | .nr_fields = BKEY_NR_FIELDS, \ | |
292 | .bits_per_field = { \ | |
293 | bkey_format_field(INODE, p.inode), \ | |
294 | bkey_format_field(OFFSET, p.offset), \ | |
295 | bkey_format_field(SNAPSHOT, p.snapshot), \ | |
296 | bkey_format_field(SIZE, size), \ | |
297 | bkey_format_field(VERSION_HI, version.hi), \ | |
298 | bkey_format_field(VERSION_LO, version.lo), \ | |
299 | }, \ | |
300 | }) | |
301 | ||
302 | /* bkey with inline value */ | |
303 | struct bkey_i { | |
304 | __u64 _data[0]; | |
305 | ||
ac2ccddc KO |
306 | struct bkey k; |
307 | struct bch_val v; | |
1c6fdbd8 KO |
308 | }; |
309 | ||
1ae8a090 KO |
310 | #define POS_KEY(_pos) \ |
311 | ((struct bkey) { \ | |
312 | .u64s = BKEY_U64s, \ | |
313 | .format = KEY_FORMAT_CURRENT, \ | |
314 | .p = _pos, \ | |
315 | }) | |
316 | ||
1c6fdbd8 KO |
317 | #define KEY(_inode, _offset, _size) \ |
318 | ((struct bkey) { \ | |
319 | .u64s = BKEY_U64s, \ | |
320 | .format = KEY_FORMAT_CURRENT, \ | |
321 | .p = POS(_inode, _offset), \ | |
322 | .size = _size, \ | |
323 | }) | |
324 | ||
325 | static inline void bkey_init(struct bkey *k) | |
326 | { | |
327 | *k = KEY(0, 0, 0); | |
328 | } | |
329 | ||
330 | #define bkey_bytes(_k) ((_k)->u64s * sizeof(__u64)) | |
331 | ||
332 | #define __BKEY_PADDED(key, pad) \ | |
45dd05b3 | 333 | struct bkey_i key; __u64 key ## _pad[pad] |
1c6fdbd8 | 334 | |
1c6fdbd8 KO |
335 | /* |
336 | * - DELETED keys are used internally to mark keys that should be ignored but | |
337 | * override keys in composition order. Their version number is ignored. | |
338 | * | |
339 | * - DISCARDED keys indicate that the data is all 0s because it has been | |
340 | * discarded. DISCARDs may have a version; if the version is nonzero the key | |
341 | * will be persistent, otherwise the key will be dropped whenever the btree | |
342 | * node is rewritten (like DELETED keys). | |
343 | * | |
344 | * - ERROR: any read of the data returns a read error, as the data was lost due | |
345 | * to a failing device. Like DISCARDED keys, they can be removed (overridden) | |
346 | * by new writes or cluster-wide GC. Node repair can also overwrite them with | |
347 | * the same or a more recent version number, but not with an older version | |
348 | * number. | |
26609b61 KO |
349 | * |
350 | * - WHITEOUT: for hash table btrees | |
3e3e02e6 | 351 | */ |
26609b61 KO |
352 | #define BCH_BKEY_TYPES() \ |
353 | x(deleted, 0) \ | |
7a7d17b2 | 354 | x(whiteout, 1) \ |
26609b61 KO |
355 | x(error, 2) \ |
356 | x(cookie, 3) \ | |
79f88eba | 357 | x(hash_whiteout, 4) \ |
26609b61 KO |
358 | x(btree_ptr, 5) \ |
359 | x(extent, 6) \ | |
360 | x(reservation, 7) \ | |
361 | x(inode, 8) \ | |
362 | x(inode_generation, 9) \ | |
363 | x(dirent, 10) \ | |
364 | x(xattr, 11) \ | |
365 | x(alloc, 12) \ | |
366 | x(quota, 13) \ | |
76426098 KO |
367 | x(stripe, 14) \ |
368 | x(reflink_p, 15) \ | |
4be1a412 | 369 | x(reflink_v, 16) \ |
548b3d20 | 370 | x(inline_data, 17) \ |
801a3de6 | 371 | x(btree_ptr_v2, 18) \ |
7f4e1d5d | 372 | x(indirect_inline_data, 19) \ |
14b393ee KO |
373 | x(alloc_v2, 20) \ |
374 | x(subvolume, 21) \ | |
3e52c222 KO |
375 | x(snapshot, 22) \ |
376 | x(inode_v2, 23) \ | |
179e3434 | 377 | x(alloc_v3, 24) \ |
d326ab2f | 378 | x(set, 25) \ |
3d48a7f8 | 379 | x(lru, 26) \ |
a8c752bb | 380 | x(alloc_v4, 27) \ |
8dd69d9f | 381 | x(backpointer, 28) \ |
5250b74d | 382 | x(inode_v3, 29) \ |
1c59b483 | 383 | x(bucket_gens, 30) \ |
b030e262 | 384 | x(snapshot_tree, 31) \ |
f3e374ef KO |
385 | x(logged_op_truncate, 32) \ |
386 | x(logged_op_finsert, 33) | |
26609b61 KO |
387 | |
388 | enum bch_bkey_type { | |
389 | #define x(name, nr) KEY_TYPE_##name = nr, | |
390 | BCH_BKEY_TYPES() | |
391 | #undef x | |
392 | KEY_TYPE_MAX, | |
393 | }; | |
1c6fdbd8 | 394 | |
79f88eba KO |
395 | struct bch_deleted { |
396 | struct bch_val v; | |
397 | }; | |
398 | ||
7a7d17b2 | 399 | struct bch_whiteout { |
79f88eba KO |
400 | struct bch_val v; |
401 | }; | |
402 | ||
403 | struct bch_error { | |
404 | struct bch_val v; | |
405 | }; | |
406 | ||
1c6fdbd8 KO |
407 | struct bch_cookie { |
408 | struct bch_val v; | |
409 | __le64 cookie; | |
410 | }; | |
1c6fdbd8 | 411 | |
79f88eba KO |
412 | struct bch_hash_whiteout { |
413 | struct bch_val v; | |
414 | }; | |
415 | ||
179e3434 KO |
416 | struct bch_set { |
417 | struct bch_val v; | |
418 | }; | |
419 | ||
1c6fdbd8 KO |
420 | /* Extents */ |
421 | ||
422 | /* | |
423 | * In extent bkeys, the value is a list of pointers (bch_extent_ptr), optionally | |
424 | * preceded by checksum/compression information (bch_extent_crc32 or | |
425 | * bch_extent_crc64). | |
426 | * | |
427 | * One major determining factor in the format of extents is how we handle and | |
428 | * represent extents that have been partially overwritten and thus trimmed: | |
429 | * | |
430 | * If an extent is not checksummed or compressed, when the extent is trimmed we | |
431 | * don't have to remember the extent we originally allocated and wrote: we can | |
432 | * merely adjust ptr->offset to point to the start of the data that is currently | |
433 | * live. The size field in struct bkey records the current (live) size of the | |
434 | * extent, and is also used to mean "size of region on disk that we point to" in | |
435 | * this case. | |
436 | * | |
437 | * Thus an extent that is not checksummed or compressed will consist only of a | |
438 | * list of bch_extent_ptrs, with none of the fields in | |
439 | * bch_extent_crc32/bch_extent_crc64. | |
440 | * | |
441 | * When an extent is checksummed or compressed, it's not possible to read only | |
442 | * the data that is currently live: we have to read the entire extent that was | |
443 | * originally written, and then return only the part of the extent that is | |
444 | * currently live. | |
445 | * | |
446 | * Thus, in addition to the current size of the extent in struct bkey, we need | |
447 | * to store the size of the originally allocated space - this is the | |
448 | * compressed_size and uncompressed_size fields in bch_extent_crc32/64. Also, | |
449 | * when the extent is trimmed, instead of modifying the offset field of the | |
450 | * pointer, we keep a second smaller offset field - "offset into the original | |
451 | * extent of the currently live region". | |
452 | * | |
453 | * The other major determining factor is replication and data migration: | |
454 | * | |
455 | * Each pointer may have its own bch_extent_crc32/64. When doing a replicated | |
456 | * write, we will initially write all the replicas in the same format, with the | |
457 | * same checksum type and compression format - however, when copygc runs later (or | |
458 | * tiering/cache promotion, anything that moves data), it is not in general | |
459 | * going to rewrite all the pointers at once - one of the replicas may be in a | |
460 | * bucket on one device that has very little fragmentation while another lives | |
461 | * in a bucket that has become heavily fragmented, and thus is being rewritten | |
462 | * sooner than the rest. | |
463 | * | |
464 | * Thus it will only move a subset of the pointers (or in the case of | |
465 | * tiering/cache promotion perhaps add a single pointer without dropping any | |
466 | * current pointers), and if the extent has been partially overwritten it must | |
467 | * write only the currently live portion (or copygc would not be able to reduce | |
468 | * fragmentation!) - which necessitates a different bch_extent_crc format for | |
469 | * the new pointer. | |
470 | * | |
471 | * But in the interests of space efficiency, we don't want to store one | |
472 | * bch_extent_crc for each pointer if we don't have to. | |
473 | * | |
474 | * Thus, a bch_extent consists of bch_extent_crc32s, bch_extent_crc64s, and | |
475 | * bch_extent_ptrs appended arbitrarily one after the other. We determine the | |
476 | * type of a given entry with a scheme similar to utf8 (except we're encoding a | |
477 | * type, not a size), encoding the type in the position of the first set bit: | |
478 | * | |
479 | * bch_extent_crc32 - 0b1 | |
480 | * bch_extent_ptr - 0b10 | |
481 | * bch_extent_crc64 - 0b100 | |
482 | * | |
483 | * We do it this way because bch_extent_crc32 is _very_ constrained on bits (and | |
484 | * bch_extent_crc64 is the least constrained). | |
485 | * | |
486 | * Then, each bch_extent_crc32/64 applies to the pointers that follow after it, | |
487 | * until the next bch_extent_crc32/64. | |
488 | * | |
489 | * If there are no bch_extent_crcs preceding a bch_extent_ptr, then that pointer | |
490 | * is neither checksummed nor compressed. | |
491 | */ | |
492 | ||
493 | /* 128 bits, sufficient for cryptographic MACs: */ | |
494 | struct bch_csum { | |
495 | __le64 lo; | |
496 | __le64 hi; | |
fd0c7679 | 497 | } __packed __aligned(8); |
1c6fdbd8 | 498 | |
abce30b7 KO |
499 | #define BCH_EXTENT_ENTRY_TYPES() \ |
500 | x(ptr, 0) \ | |
501 | x(crc32, 1) \ | |
502 | x(crc64, 2) \ | |
cd575ddf | 503 | x(crc128, 3) \ |
2766876d KO |
504 | x(stripe_ptr, 4) \ |
505 | x(rebalance, 5) | |
506 | #define BCH_EXTENT_ENTRY_MAX 6 | |
abce30b7 | 507 | |
1c6fdbd8 | 508 | enum bch_extent_entry_type { |
abce30b7 KO |
509 | #define x(f, n) BCH_EXTENT_ENTRY_##f = n, |
510 | BCH_EXTENT_ENTRY_TYPES() | |
511 | #undef x | |
1c6fdbd8 KO |
512 | }; |
513 | ||
1c6fdbd8 KO |
514 | /* Compressed/uncompressed size are stored biased by 1: */ |
515 | struct bch_extent_crc32 { | |
516 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
517 | __u32 type:2, | |
518 | _compressed_size:7, | |
519 | _uncompressed_size:7, | |
520 | offset:7, | |
521 | _unused:1, | |
522 | csum_type:4, | |
523 | compression_type:4; | |
524 | __u32 csum; | |
525 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
526 | __u32 csum; | |
527 | __u32 compression_type:4, | |
528 | csum_type:4, | |
529 | _unused:1, | |
530 | offset:7, | |
531 | _uncompressed_size:7, | |
532 | _compressed_size:7, | |
533 | type:2; | |
534 | #endif | |
fd0c7679 | 535 | } __packed __aligned(8); |
1c6fdbd8 KO |
536 | |
537 | #define CRC32_SIZE_MAX (1U << 7) | |
538 | #define CRC32_NONCE_MAX 0 | |
539 | ||
540 | struct bch_extent_crc64 { | |
541 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
542 | __u64 type:3, | |
543 | _compressed_size:9, | |
544 | _uncompressed_size:9, | |
545 | offset:9, | |
546 | nonce:10, | |
547 | csum_type:4, | |
548 | compression_type:4, | |
549 | csum_hi:16; | |
550 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
551 | __u64 csum_hi:16, | |
552 | compression_type:4, | |
553 | csum_type:4, | |
554 | nonce:10, | |
555 | offset:9, | |
556 | _uncompressed_size:9, | |
557 | _compressed_size:9, | |
558 | type:3; | |
559 | #endif | |
560 | __u64 csum_lo; | |
fd0c7679 | 561 | } __packed __aligned(8); |
1c6fdbd8 KO |
562 | |
563 | #define CRC64_SIZE_MAX (1U << 9) | |
564 | #define CRC64_NONCE_MAX ((1U << 10) - 1) | |
565 | ||
566 | struct bch_extent_crc128 { | |
567 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
568 | __u64 type:4, | |
569 | _compressed_size:13, | |
570 | _uncompressed_size:13, | |
571 | offset:13, | |
572 | nonce:13, | |
573 | csum_type:4, | |
574 | compression_type:4; | |
575 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
576 | __u64 compression_type:4, | |
577 | csum_type:4, | |
578 | nonce:13, | |
579 | offset:13, | |
580 | _uncompressed_size:13, | |
581 | _compressed_size:13, | |
582 | type:4; | |
583 | #endif | |
584 | struct bch_csum csum; | |
fd0c7679 | 585 | } __packed __aligned(8); |
1c6fdbd8 KO |
586 | |
587 | #define CRC128_SIZE_MAX (1U << 13) | |
588 | #define CRC128_NONCE_MAX ((1U << 13) - 1) | |
589 | ||
590 | /* | |
591 | * @reservation - pointer hasn't been written to, just reserved | |
592 | */ | |
593 | struct bch_extent_ptr { | |
594 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
595 | __u64 type:1, | |
596 | cached:1, | |
cd575ddf | 597 | unused:1, |
79203111 | 598 | unwritten:1, |
1c6fdbd8 KO |
599 | offset:44, /* 8 petabytes */ |
600 | dev:8, | |
601 | gen:8; | |
602 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
603 | __u64 gen:8, | |
604 | dev:8, | |
605 | offset:44, | |
79203111 | 606 | unwritten:1, |
cd575ddf | 607 | unused:1, |
1c6fdbd8 KO |
608 | cached:1, |
609 | type:1; | |
610 | #endif | |
fd0c7679 | 611 | } __packed __aligned(8); |
1c6fdbd8 | 612 | |
cd575ddf | 613 | struct bch_extent_stripe_ptr { |
1c6fdbd8 KO |
614 | #if defined(__LITTLE_ENDIAN_BITFIELD) |
615 | __u64 type:5, | |
cd575ddf | 616 | block:8, |
7f4e1d5d KO |
617 | redundancy:4, |
618 | idx:47; | |
cd575ddf | 619 | #elif defined (__BIG_ENDIAN_BITFIELD) |
7f4e1d5d KO |
620 | __u64 idx:47, |
621 | redundancy:4, | |
cd575ddf KO |
622 | block:8, |
623 | type:5; | |
624 | #endif | |
625 | }; | |
626 | ||
2766876d KO |
627 | struct bch_extent_rebalance { |
628 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
fb3f57bb KO |
629 | __u64 type:6, |
630 | unused:34, | |
631 | compression:8, /* enum bch_compression_opt */ | |
2766876d KO |
632 | target:16; |
633 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
634 | __u64 target:16, | |
635 | compression:8, | |
fb3f57bb KO |
636 | unused:34, |
637 | type:6; | |
2766876d KO |
638 | #endif |
639 | }; | |
640 | ||
1c6fdbd8 KO |
641 | union bch_extent_entry { |
642 | #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ || __BITS_PER_LONG == 64 | |
643 | unsigned long type; | |
644 | #elif __BITS_PER_LONG == 32 | |
645 | struct { | |
646 | unsigned long pad; | |
647 | unsigned long type; | |
648 | }; | |
649 | #else | |
650 | #error edit for your odd byteorder. | |
651 | #endif | |
abce30b7 KO |
652 | |
653 | #define x(f, n) struct bch_extent_##f f; | |
654 | BCH_EXTENT_ENTRY_TYPES() | |
655 | #undef x | |
1c6fdbd8 KO |
656 | }; |
657 | ||
26609b61 KO |
658 | struct bch_btree_ptr { |
659 | struct bch_val v; | |
1c6fdbd8 | 660 | |
26609b61 KO |
661 | __u64 _data[0]; |
662 | struct bch_extent_ptr start[]; | |
fd0c7679 | 663 | } __packed __aligned(8); |
1c6fdbd8 | 664 | |
548b3d20 KO |
665 | struct bch_btree_ptr_v2 { |
666 | struct bch_val v; | |
667 | ||
668 | __u64 mem_ptr; | |
669 | __le64 seq; | |
670 | __le16 sectors_written; | |
51d2dfb8 | 671 | __le16 flags; |
548b3d20 KO |
672 | struct bpos min_key; |
673 | __u64 _data[0]; | |
674 | struct bch_extent_ptr start[]; | |
fd0c7679 | 675 | } __packed __aligned(8); |
548b3d20 | 676 | |
51d2dfb8 KO |
677 | LE16_BITMASK(BTREE_PTR_RANGE_UPDATED, struct bch_btree_ptr_v2, flags, 0, 1); |
678 | ||
1c6fdbd8 KO |
679 | struct bch_extent { |
680 | struct bch_val v; | |
681 | ||
682 | __u64 _data[0]; | |
683 | union bch_extent_entry start[]; | |
fd0c7679 | 684 | } __packed __aligned(8); |
1c6fdbd8 KO |
685 | |
686 | struct bch_reservation { | |
687 | struct bch_val v; | |
688 | ||
689 | __le32 generation; | |
690 | __u8 nr_replicas; | |
691 | __u8 pad[3]; | |
fd0c7679 | 692 | } __packed __aligned(8); |
1c6fdbd8 KO |
693 | |
694 | /* Maximum size (in u64s) a single pointer could be: */ | |
695 | #define BKEY_EXTENT_PTR_U64s_MAX\ | |
696 | ((sizeof(struct bch_extent_crc128) + \ | |
a5cf5a4b | 697 | sizeof(struct bch_extent_ptr)) / sizeof(__u64)) |
1c6fdbd8 KO |
698 | |
699 | /* Maximum possible size of an entire extent value: */ | |
700 | #define BKEY_EXTENT_VAL_U64s_MAX \ | |
5055b509 | 701 | (1 + BKEY_EXTENT_PTR_U64s_MAX * (BCH_REPLICAS_MAX + 1)) |
1c6fdbd8 | 702 | |
1c6fdbd8 KO |
703 | /* * Maximum possible size of an entire extent, key + value: */ |
704 | #define BKEY_EXTENT_U64s_MAX (BKEY_U64s + BKEY_EXTENT_VAL_U64s_MAX) | |
705 | ||
706 | /* Btree pointers don't carry around checksums: */ | |
707 | #define BKEY_BTREE_PTR_VAL_U64s_MAX \ | |
548b3d20 | 708 | ((sizeof(struct bch_btree_ptr_v2) + \ |
a5cf5a4b | 709 | sizeof(struct bch_extent_ptr) * BCH_REPLICAS_MAX) / sizeof(__u64)) |
1c6fdbd8 KO |
710 | #define BKEY_BTREE_PTR_U64s_MAX \ |
711 | (BKEY_U64s + BKEY_BTREE_PTR_VAL_U64s_MAX) | |
712 | ||
713 | /* Inodes */ | |
714 | ||
715 | #define BLOCKDEV_INODE_MAX 4096 | |
716 | ||
717 | #define BCACHEFS_ROOT_INO 4096 | |
718 | ||
1c6fdbd8 KO |
719 | struct bch_inode { |
720 | struct bch_val v; | |
721 | ||
722 | __le64 bi_hash_seed; | |
723 | __le32 bi_flags; | |
724 | __le16 bi_mode; | |
5cfd6977 | 725 | __u8 fields[]; |
fd0c7679 | 726 | } __packed __aligned(8); |
1c6fdbd8 | 727 | |
3e52c222 KO |
728 | struct bch_inode_v2 { |
729 | struct bch_val v; | |
730 | ||
731 | __le64 bi_journal_seq; | |
732 | __le64 bi_hash_seed; | |
733 | __le64 bi_flags; | |
734 | __le16 bi_mode; | |
5cfd6977 | 735 | __u8 fields[]; |
fd0c7679 | 736 | } __packed __aligned(8); |
3e52c222 | 737 | |
8dd69d9f KO |
738 | struct bch_inode_v3 { |
739 | struct bch_val v; | |
740 | ||
741 | __le64 bi_journal_seq; | |
742 | __le64 bi_hash_seed; | |
743 | __le64 bi_flags; | |
744 | __le64 bi_sectors; | |
745 | __le64 bi_size; | |
746 | __le64 bi_version; | |
5cfd6977 | 747 | __u8 fields[]; |
8dd69d9f KO |
748 | } __packed __aligned(8); |
749 | ||
750 | #define INODEv3_FIELDS_START_INITIAL 6 | |
a5cf5a4b | 751 | #define INODEv3_FIELDS_START_CUR (offsetof(struct bch_inode_v3, fields) / sizeof(__u64)) |
8dd69d9f | 752 | |
1c6fdbd8 KO |
753 | struct bch_inode_generation { |
754 | struct bch_val v; | |
755 | ||
756 | __le32 bi_generation; | |
757 | __le32 pad; | |
fd0c7679 | 758 | } __packed __aligned(8); |
1c6fdbd8 | 759 | |
14b393ee KO |
760 | /* |
761 | * bi_subvol and bi_parent_subvol are only set for subvolume roots: | |
762 | */ | |
763 | ||
8dd69d9f | 764 | #define BCH_INODE_FIELDS_v2() \ |
a3e72262 KO |
765 | x(bi_atime, 96) \ |
766 | x(bi_ctime, 96) \ | |
767 | x(bi_mtime, 96) \ | |
768 | x(bi_otime, 96) \ | |
a3e70fb2 KO |
769 | x(bi_size, 64) \ |
770 | x(bi_sectors, 64) \ | |
771 | x(bi_uid, 32) \ | |
772 | x(bi_gid, 32) \ | |
773 | x(bi_nlink, 32) \ | |
774 | x(bi_generation, 32) \ | |
775 | x(bi_dev, 32) \ | |
776 | x(bi_data_checksum, 8) \ | |
777 | x(bi_compression, 8) \ | |
778 | x(bi_project, 32) \ | |
779 | x(bi_background_compression, 8) \ | |
780 | x(bi_data_replicas, 8) \ | |
781 | x(bi_promote_target, 16) \ | |
782 | x(bi_foreground_target, 16) \ | |
783 | x(bi_background_target, 16) \ | |
721d4ad8 | 784 | x(bi_erasure_code, 16) \ |
ab2a29cc KO |
785 | x(bi_fields_set, 16) \ |
786 | x(bi_dir, 64) \ | |
14b393ee KO |
787 | x(bi_dir_offset, 64) \ |
788 | x(bi_subvol, 32) \ | |
789 | x(bi_parent_subvol, 32) | |
a3e70fb2 | 790 | |
8dd69d9f KO |
791 | #define BCH_INODE_FIELDS_v3() \ |
792 | x(bi_atime, 96) \ | |
793 | x(bi_ctime, 96) \ | |
794 | x(bi_mtime, 96) \ | |
795 | x(bi_otime, 96) \ | |
796 | x(bi_uid, 32) \ | |
797 | x(bi_gid, 32) \ | |
798 | x(bi_nlink, 32) \ | |
799 | x(bi_generation, 32) \ | |
800 | x(bi_dev, 32) \ | |
801 | x(bi_data_checksum, 8) \ | |
802 | x(bi_compression, 8) \ | |
803 | x(bi_project, 32) \ | |
804 | x(bi_background_compression, 8) \ | |
805 | x(bi_data_replicas, 8) \ | |
806 | x(bi_promote_target, 16) \ | |
807 | x(bi_foreground_target, 16) \ | |
808 | x(bi_background_target, 16) \ | |
809 | x(bi_erasure_code, 16) \ | |
810 | x(bi_fields_set, 16) \ | |
811 | x(bi_dir, 64) \ | |
812 | x(bi_dir_offset, 64) \ | |
813 | x(bi_subvol, 32) \ | |
a8b3a677 KO |
814 | x(bi_parent_subvol, 32) \ |
815 | x(bi_nocow, 8) | |
8dd69d9f | 816 | |
d42dd4ad KO |
817 | /* subset of BCH_INODE_FIELDS */ |
818 | #define BCH_INODE_OPTS() \ | |
819 | x(data_checksum, 8) \ | |
820 | x(compression, 8) \ | |
821 | x(project, 32) \ | |
822 | x(background_compression, 8) \ | |
823 | x(data_replicas, 8) \ | |
824 | x(promote_target, 16) \ | |
825 | x(foreground_target, 16) \ | |
826 | x(background_target, 16) \ | |
a8b3a677 KO |
827 | x(erasure_code, 16) \ |
828 | x(nocow, 8) | |
1c6fdbd8 | 829 | |
721d4ad8 KO |
830 | enum inode_opt_id { |
831 | #define x(name, ...) \ | |
832 | Inode_opt_##name, | |
833 | BCH_INODE_OPTS() | |
834 | #undef x | |
835 | Inode_opt_nr, | |
836 | }; | |
837 | ||
103ffe9a KO |
838 | #define BCH_INODE_FLAGS() \ |
839 | x(sync, 0) \ | |
840 | x(immutable, 1) \ | |
841 | x(append, 2) \ | |
842 | x(nodump, 3) \ | |
843 | x(noatime, 4) \ | |
844 | x(i_size_dirty, 5) \ | |
845 | x(i_sectors_dirty, 6) \ | |
846 | x(unlinked, 7) \ | |
847 | x(backptr_untrusted, 8) | |
848 | ||
849 | /* bits 20+ reserved for packed fields below: */ | |
850 | ||
851 | enum bch_inode_flags { | |
852 | #define x(t, n) BCH_INODE_##t = 1U << n, | |
853 | BCH_INODE_FLAGS() | |
854 | #undef x | |
1c6fdbd8 KO |
855 | }; |
856 | ||
103ffe9a KO |
857 | enum __bch_inode_flags { |
858 | #define x(t, n) __BCH_INODE_##t = n, | |
859 | BCH_INODE_FLAGS() | |
860 | #undef x | |
861 | }; | |
1c6fdbd8 KO |
862 | |
863 | LE32_BITMASK(INODE_STR_HASH, struct bch_inode, bi_flags, 20, 24); | |
a3e72262 KO |
864 | LE32_BITMASK(INODE_NR_FIELDS, struct bch_inode, bi_flags, 24, 31); |
865 | LE32_BITMASK(INODE_NEW_VARINT, struct bch_inode, bi_flags, 31, 32); | |
1c6fdbd8 | 866 | |
3e52c222 KO |
867 | LE64_BITMASK(INODEv2_STR_HASH, struct bch_inode_v2, bi_flags, 20, 24); |
868 | LE64_BITMASK(INODEv2_NR_FIELDS, struct bch_inode_v2, bi_flags, 24, 31); | |
869 | ||
8dd69d9f KO |
870 | LE64_BITMASK(INODEv3_STR_HASH, struct bch_inode_v3, bi_flags, 20, 24); |
871 | LE64_BITMASK(INODEv3_NR_FIELDS, struct bch_inode_v3, bi_flags, 24, 31); | |
872 | ||
873 | LE64_BITMASK(INODEv3_FIELDS_START, | |
874 | struct bch_inode_v3, bi_flags, 31, 36); | |
875 | LE64_BITMASK(INODEv3_MODE, struct bch_inode_v3, bi_flags, 36, 52); | |
876 | ||
1c6fdbd8 KO |
877 | /* Dirents */ |
878 | ||
879 | /* | |
880 | * Dirents (and xattrs) have to implement string lookups; since our b-tree | |
881 | * doesn't support arbitrary length strings for the key, we instead index by a | |
882 | * 64 bit hash (currently truncated sha1) of the string, stored in the offset | |
883 | * field of the key - using linear probing to resolve hash collisions. This also | |
884 | * provides us with the readdir cookie posix requires. | |
885 | * | |
886 | * Linear probing requires us to use whiteouts for deletions, in the event of a | |
887 | * collision: | |
888 | */ | |
889 | ||
1c6fdbd8 KO |
890 | struct bch_dirent { |
891 | struct bch_val v; | |
892 | ||
893 | /* Target inode number: */ | |
4db65027 | 894 | union { |
1c6fdbd8 | 895 | __le64 d_inum; |
4db65027 KO |
896 | struct { /* DT_SUBVOL */ |
897 | __le32 d_child_subvol; | |
898 | __le32 d_parent_subvol; | |
899 | }; | |
900 | }; | |
1c6fdbd8 KO |
901 | |
902 | /* | |
903 | * Copy of mode bits 12-15 from the target inode - so userspace can get | |
904 | * the filetype without having to do a stat() | |
905 | */ | |
906 | __u8 d_type; | |
907 | ||
908 | __u8 d_name[]; | |
fd0c7679 | 909 | } __packed __aligned(8); |
1c6fdbd8 | 910 | |
14b393ee KO |
911 | #define DT_SUBVOL 16 |
912 | #define BCH_DT_MAX 17 | |
913 | ||
a125c074 | 914 | #define BCH_NAME_MAX 512 |
1c6fdbd8 KO |
915 | |
916 | /* Xattrs */ | |
917 | ||
26609b61 KO |
918 | #define KEY_TYPE_XATTR_INDEX_USER 0 |
919 | #define KEY_TYPE_XATTR_INDEX_POSIX_ACL_ACCESS 1 | |
920 | #define KEY_TYPE_XATTR_INDEX_POSIX_ACL_DEFAULT 2 | |
921 | #define KEY_TYPE_XATTR_INDEX_TRUSTED 3 | |
922 | #define KEY_TYPE_XATTR_INDEX_SECURITY 4 | |
1c6fdbd8 KO |
923 | |
924 | struct bch_xattr { | |
925 | struct bch_val v; | |
926 | __u8 x_type; | |
927 | __u8 x_name_len; | |
928 | __le16 x_val_len; | |
929 | __u8 x_name[]; | |
fd0c7679 | 930 | } __packed __aligned(8); |
1c6fdbd8 KO |
931 | |
932 | /* Bucket/allocation information: */ | |
933 | ||
1c6fdbd8 KO |
934 | struct bch_alloc { |
935 | struct bch_val v; | |
936 | __u8 fields; | |
937 | __u8 gen; | |
938 | __u8 data[]; | |
fd0c7679 | 939 | } __packed __aligned(8); |
1c6fdbd8 | 940 | |
7f4e1d5d | 941 | #define BCH_ALLOC_FIELDS_V1() \ |
8fe826f9 KO |
942 | x(read_time, 16) \ |
943 | x(write_time, 16) \ | |
944 | x(data_type, 8) \ | |
945 | x(dirty_sectors, 16) \ | |
946 | x(cached_sectors, 16) \ | |
7f4e1d5d KO |
947 | x(oldest_gen, 8) \ |
948 | x(stripe, 32) \ | |
949 | x(stripe_redundancy, 8) | |
950 | ||
a8c752bb KO |
951 | enum { |
952 | #define x(name, _bits) BCH_ALLOC_FIELD_V1_##name, | |
953 | BCH_ALLOC_FIELDS_V1() | |
954 | #undef x | |
955 | }; | |
956 | ||
7f4e1d5d KO |
957 | struct bch_alloc_v2 { |
958 | struct bch_val v; | |
959 | __u8 nr_fields; | |
960 | __u8 gen; | |
961 | __u8 oldest_gen; | |
962 | __u8 data_type; | |
963 | __u8 data[]; | |
fd0c7679 | 964 | } __packed __aligned(8); |
7f4e1d5d KO |
965 | |
966 | #define BCH_ALLOC_FIELDS_V2() \ | |
967 | x(read_time, 64) \ | |
968 | x(write_time, 64) \ | |
66d90823 KO |
969 | x(dirty_sectors, 32) \ |
970 | x(cached_sectors, 32) \ | |
7f4e1d5d KO |
971 | x(stripe, 32) \ |
972 | x(stripe_redundancy, 8) | |
90541a74 | 973 | |
3e52c222 KO |
974 | struct bch_alloc_v3 { |
975 | struct bch_val v; | |
976 | __le64 journal_seq; | |
977 | __le32 flags; | |
978 | __u8 nr_fields; | |
979 | __u8 gen; | |
980 | __u8 oldest_gen; | |
981 | __u8 data_type; | |
982 | __u8 data[]; | |
fd0c7679 | 983 | } __packed __aligned(8); |
3e52c222 | 984 | |
a8c752bb KO |
985 | LE32_BITMASK(BCH_ALLOC_V3_NEED_DISCARD,struct bch_alloc_v3, flags, 0, 1) |
986 | LE32_BITMASK(BCH_ALLOC_V3_NEED_INC_GEN,struct bch_alloc_v3, flags, 1, 2) | |
987 | ||
3d48a7f8 KO |
988 | struct bch_alloc_v4 { |
989 | struct bch_val v; | |
990 | __u64 journal_seq; | |
991 | __u32 flags; | |
992 | __u8 gen; | |
993 | __u8 oldest_gen; | |
994 | __u8 data_type; | |
995 | __u8 stripe_redundancy; | |
996 | __u32 dirty_sectors; | |
997 | __u32 cached_sectors; | |
998 | __u64 io_time[2]; | |
999 | __u32 stripe; | |
1000 | __u32 nr_external_backpointers; | |
80c33085 | 1001 | __u64 fragmentation_lru; |
fd0c7679 | 1002 | } __packed __aligned(8); |
3d48a7f8 | 1003 | |
19a614d2 | 1004 | #define BCH_ALLOC_V4_U64s_V0 6 |
a5cf5a4b | 1005 | #define BCH_ALLOC_V4_U64s (sizeof(struct bch_alloc_v4) / sizeof(__u64)) |
19a614d2 | 1006 | |
3d48a7f8 KO |
1007 | BITMASK(BCH_ALLOC_V4_NEED_DISCARD, struct bch_alloc_v4, flags, 0, 1) |
1008 | BITMASK(BCH_ALLOC_V4_NEED_INC_GEN, struct bch_alloc_v4, flags, 1, 2) | |
1009 | BITMASK(BCH_ALLOC_V4_BACKPOINTERS_START,struct bch_alloc_v4, flags, 2, 8) | |
1010 | BITMASK(BCH_ALLOC_V4_NR_BACKPOINTERS, struct bch_alloc_v4, flags, 8, 14) | |
1011 | ||
a8c752bb KO |
1012 | #define BCH_ALLOC_V4_NR_BACKPOINTERS_MAX 40 |
1013 | ||
1014 | struct bch_backpointer { | |
1015 | struct bch_val v; | |
1016 | __u8 btree_id; | |
1017 | __u8 level; | |
1018 | __u8 data_type; | |
1019 | __u64 bucket_offset:40; | |
1020 | __u32 bucket_len; | |
1021 | struct bpos pos; | |
1022 | } __packed __aligned(8); | |
90541a74 | 1023 | |
5250b74d KO |
1024 | #define KEY_TYPE_BUCKET_GENS_BITS 8 |
1025 | #define KEY_TYPE_BUCKET_GENS_NR (1U << KEY_TYPE_BUCKET_GENS_BITS) | |
1026 | #define KEY_TYPE_BUCKET_GENS_MASK (KEY_TYPE_BUCKET_GENS_NR - 1) | |
1027 | ||
1028 | struct bch_bucket_gens { | |
1029 | struct bch_val v; | |
1030 | u8 gens[KEY_TYPE_BUCKET_GENS_NR]; | |
1031 | } __packed __aligned(8); | |
1032 | ||
1c6fdbd8 KO |
1033 | /* Quotas: */ |
1034 | ||
1c6fdbd8 KO |
1035 | enum quota_types { |
1036 | QTYP_USR = 0, | |
1037 | QTYP_GRP = 1, | |
1038 | QTYP_PRJ = 2, | |
1039 | QTYP_NR = 3, | |
1040 | }; | |
1041 | ||
1042 | enum quota_counters { | |
1043 | Q_SPC = 0, | |
1044 | Q_INO = 1, | |
1045 | Q_COUNTERS = 2, | |
1046 | }; | |
1047 | ||
1048 | struct bch_quota_counter { | |
1049 | __le64 hardlimit; | |
1050 | __le64 softlimit; | |
1051 | }; | |
1052 | ||
1053 | struct bch_quota { | |
1054 | struct bch_val v; | |
1055 | struct bch_quota_counter c[Q_COUNTERS]; | |
fd0c7679 | 1056 | } __packed __aligned(8); |
1c6fdbd8 | 1057 | |
cd575ddf KO |
1058 | /* Erasure coding */ |
1059 | ||
cd575ddf KO |
1060 | struct bch_stripe { |
1061 | struct bch_val v; | |
1062 | __le16 sectors; | |
1063 | __u8 algorithm; | |
1064 | __u8 nr_blocks; | |
1065 | __u8 nr_redundant; | |
1066 | ||
1067 | __u8 csum_granularity_bits; | |
1068 | __u8 csum_type; | |
1069 | __u8 pad; | |
1070 | ||
81d8599e | 1071 | struct bch_extent_ptr ptrs[]; |
fd0c7679 | 1072 | } __packed __aligned(8); |
cd575ddf | 1073 | |
76426098 KO |
1074 | /* Reflink: */ |
1075 | ||
1076 | struct bch_reflink_p { | |
1077 | struct bch_val v; | |
1078 | __le64 idx; | |
6d76aefe KO |
1079 | /* |
1080 | * A reflink pointer might point to an indirect extent which is then | |
1081 | * later split (by copygc or rebalance). If we only pointed to part of | |
1082 | * the original indirect extent, and then one of the fragments is | |
1083 | * outside the range we point to, we'd leak a refcount: so when creating | |
1084 | * reflink pointers, we need to store pad values to remember the full | |
1085 | * range we were taking a reference on. | |
1086 | */ | |
1087 | __le32 front_pad; | |
1088 | __le32 back_pad; | |
fd0c7679 | 1089 | } __packed __aligned(8); |
76426098 KO |
1090 | |
1091 | struct bch_reflink_v { | |
1092 | struct bch_val v; | |
1093 | __le64 refcount; | |
1094 | union bch_extent_entry start[0]; | |
5cfd6977 | 1095 | __u64 _data[]; |
fd0c7679 | 1096 | } __packed __aligned(8); |
76426098 | 1097 | |
801a3de6 KO |
1098 | struct bch_indirect_inline_data { |
1099 | struct bch_val v; | |
1100 | __le64 refcount; | |
5cfd6977 | 1101 | u8 data[]; |
801a3de6 KO |
1102 | }; |
1103 | ||
4be1a412 KO |
1104 | /* Inline data */ |
1105 | ||
1106 | struct bch_inline_data { | |
1107 | struct bch_val v; | |
5cfd6977 | 1108 | u8 data[]; |
4be1a412 KO |
1109 | }; |
1110 | ||
14b393ee KO |
1111 | /* Subvolumes: */ |
1112 | ||
1113 | #define SUBVOL_POS_MIN POS(0, 1) | |
1114 | #define SUBVOL_POS_MAX POS(0, S32_MAX) | |
1115 | #define BCACHEFS_ROOT_SUBVOL 1 | |
1116 | ||
1117 | struct bch_subvolume { | |
1118 | struct bch_val v; | |
1119 | __le32 flags; | |
1120 | __le32 snapshot; | |
1121 | __le64 inode; | |
8e877caa KO |
1122 | /* |
1123 | * Snapshot subvolumes form a tree, separate from the snapshot nodes | |
1124 | * tree - if this subvolume is a snapshot, this is the ID of the | |
1125 | * subvolume it was created from: | |
1126 | */ | |
653693be KO |
1127 | __le32 parent; |
1128 | __le32 pad; | |
1129 | bch_le128 otime; | |
14b393ee KO |
1130 | }; |
1131 | ||
1132 | LE32_BITMASK(BCH_SUBVOLUME_RO, struct bch_subvolume, flags, 0, 1) | |
1133 | /* | |
1134 | * We need to know whether a subvolume is a snapshot so we can know whether we | |
1135 | * can delete it (or whether it should just be rm -rf'd) | |
1136 | */ | |
1137 | LE32_BITMASK(BCH_SUBVOLUME_SNAP, struct bch_subvolume, flags, 1, 2) | |
2027875b | 1138 | LE32_BITMASK(BCH_SUBVOLUME_UNLINKED, struct bch_subvolume, flags, 2, 3) |
14b393ee KO |
1139 | |
1140 | /* Snapshots */ | |
1141 | ||
1142 | struct bch_snapshot { | |
1143 | struct bch_val v; | |
1144 | __le32 flags; | |
1145 | __le32 parent; | |
1146 | __le32 children[2]; | |
1147 | __le32 subvol; | |
f55d6e07 | 1148 | /* corresponds to a bch_snapshot_tree in BTREE_ID_snapshot_trees */ |
1c59b483 | 1149 | __le32 tree; |
f26c67f4 KO |
1150 | __le32 depth; |
1151 | __le32 skip[3]; | |
14b393ee KO |
1152 | }; |
1153 | ||
1154 | LE32_BITMASK(BCH_SNAPSHOT_DELETED, struct bch_snapshot, flags, 0, 1) | |
1155 | ||
1156 | /* True if a subvolume points to this snapshot node: */ | |
1157 | LE32_BITMASK(BCH_SNAPSHOT_SUBVOL, struct bch_snapshot, flags, 1, 2) | |
1158 | ||
1c59b483 KO |
1159 | /* |
1160 | * Snapshot trees: | |
1161 | * | |
1162 | * The snapshot_trees btree gives us persistent indentifier for each tree of | |
1163 | * bch_snapshot nodes, and allow us to record and easily find the root/master | |
1164 | * subvolume that other snapshots were created from: | |
1165 | */ | |
1166 | struct bch_snapshot_tree { | |
1167 | struct bch_val v; | |
1168 | __le32 master_subvol; | |
1169 | __le32 root_snapshot; | |
1170 | }; | |
1171 | ||
d326ab2f KO |
1172 | /* LRU btree: */ |
1173 | ||
1174 | struct bch_lru { | |
1175 | struct bch_val v; | |
1176 | __le64 idx; | |
fd0c7679 | 1177 | } __packed __aligned(8); |
d326ab2f KO |
1178 | |
1179 | #define LRU_ID_STRIPES (1U << 16) | |
1180 | ||
b030e262 KO |
1181 | /* Logged operations btree: */ |
1182 | ||
1183 | struct bch_logged_op_truncate { | |
1184 | struct bch_val v; | |
1185 | __le32 subvol; | |
1186 | __le32 pad; | |
1187 | __le64 inum; | |
1188 | __le64 new_i_size; | |
1189 | }; | |
1190 | ||
f3e374ef KO |
1191 | enum logged_op_finsert_state { |
1192 | LOGGED_OP_FINSERT_start, | |
1193 | LOGGED_OP_FINSERT_shift_extents, | |
1194 | LOGGED_OP_FINSERT_finish, | |
1195 | }; | |
1196 | ||
1197 | struct bch_logged_op_finsert { | |
1198 | struct bch_val v; | |
1199 | __u8 state; | |
1200 | __u8 pad[3]; | |
1201 | __le32 subvol; | |
1202 | __le64 inum; | |
1203 | __le64 dst_offset; | |
1204 | __le64 src_offset; | |
1205 | __le64 pos; | |
1206 | }; | |
1207 | ||
1c6fdbd8 KO |
1208 | /* Optional/variable size superblock sections: */ |
1209 | ||
1210 | struct bch_sb_field { | |
1211 | __u64 _data[0]; | |
1212 | __le32 u64s; | |
1213 | __le32 type; | |
1214 | }; | |
1215 | ||
25be2e5d | 1216 | #define BCH_SB_FIELDS() \ |
8b16413c KO |
1217 | x(journal, 0) \ |
1218 | x(members_v1, 1) \ | |
1219 | x(crypt, 2) \ | |
1220 | x(replicas_v0, 3) \ | |
1221 | x(quota, 4) \ | |
1222 | x(disk_groups, 5) \ | |
1223 | x(clean, 6) \ | |
1224 | x(replicas, 7) \ | |
1225 | x(journal_seq_blacklist, 8) \ | |
1226 | x(journal_v2, 9) \ | |
1227 | x(counters, 10) \ | |
1228 | x(members_v2, 11) \ | |
1229 | x(errors, 12) \ | |
84f16387 KO |
1230 | x(ext, 13) \ |
1231 | x(downgrade, 14) | |
1c6fdbd8 KO |
1232 | |
1233 | enum bch_sb_field_type { | |
1234 | #define x(f, nr) BCH_SB_FIELD_##f = nr, | |
1235 | BCH_SB_FIELDS() | |
1236 | #undef x | |
1237 | BCH_SB_FIELD_NR | |
1238 | }; | |
1239 | ||
25be2e5d KO |
1240 | /* |
1241 | * Most superblock fields are replicated in all device's superblocks - a few are | |
1242 | * not: | |
1243 | */ | |
1244 | #define BCH_SINGLE_DEVICE_SB_FIELDS \ | |
1245 | ((1U << BCH_SB_FIELD_journal)| \ | |
1246 | (1U << BCH_SB_FIELD_journal_v2)) | |
1247 | ||
1c6fdbd8 KO |
1248 | /* BCH_SB_FIELD_journal: */ |
1249 | ||
1250 | struct bch_sb_field_journal { | |
1251 | struct bch_sb_field field; | |
5cfd6977 | 1252 | __le64 buckets[]; |
1c6fdbd8 KO |
1253 | }; |
1254 | ||
25be2e5d KO |
1255 | struct bch_sb_field_journal_v2 { |
1256 | struct bch_sb_field field; | |
1257 | ||
1258 | struct bch_sb_field_journal_v2_entry { | |
1259 | __le64 start; | |
1260 | __le64 nr; | |
5cfd6977 | 1261 | } d[]; |
25be2e5d KO |
1262 | }; |
1263 | ||
9af26120 | 1264 | /* BCH_SB_FIELD_members_v1: */ |
1c6fdbd8 | 1265 | |
8b335bae KO |
1266 | #define BCH_MIN_NR_NBUCKETS (1 << 6) |
1267 | ||
40f7914e HS |
1268 | #define BCH_IOPS_MEASUREMENTS() \ |
1269 | x(seqread, 0) \ | |
1270 | x(seqwrite, 1) \ | |
1271 | x(randread, 2) \ | |
1272 | x(randwrite, 3) | |
1273 | ||
1274 | enum bch_iops_measurement { | |
1275 | #define x(t, n) BCH_IOPS_##t = n, | |
1276 | BCH_IOPS_MEASUREMENTS() | |
1277 | #undef x | |
1278 | BCH_IOPS_NR | |
1279 | }; | |
1280 | ||
94119eeb KO |
1281 | #define BCH_MEMBER_ERROR_TYPES() \ |
1282 | x(read, 0) \ | |
1283 | x(write, 1) \ | |
1284 | x(checksum, 2) | |
1285 | ||
1286 | enum bch_member_error_type { | |
1287 | #define x(t, n) BCH_MEMBER_ERROR_##t = n, | |
1288 | BCH_MEMBER_ERROR_TYPES() | |
1289 | #undef x | |
1290 | BCH_MEMBER_ERROR_NR | |
1291 | }; | |
1292 | ||
1c6fdbd8 KO |
1293 | struct bch_member { |
1294 | __uuid_t uuid; | |
1295 | __le64 nbuckets; /* device size */ | |
1296 | __le16 first_bucket; /* index of first bucket used */ | |
1297 | __le16 bucket_size; /* sectors */ | |
1298 | __le32 pad; | |
1299 | __le64 last_mount; /* time_t */ | |
1300 | ||
40f7914e HS |
1301 | __le64 flags; |
1302 | __le32 iops[4]; | |
94119eeb KO |
1303 | __le64 errors[BCH_MEMBER_ERROR_NR]; |
1304 | __le64 errors_at_reset[BCH_MEMBER_ERROR_NR]; | |
1305 | __le64 errors_reset_time; | |
1c6fdbd8 KO |
1306 | }; |
1307 | ||
3f7b9713 HS |
1308 | #define BCH_MEMBER_V1_BYTES 56 |
1309 | ||
40f7914e | 1310 | LE64_BITMASK(BCH_MEMBER_STATE, struct bch_member, flags, 0, 4) |
7243498d | 1311 | /* 4-14 unused, was TIER, HAS_(META)DATA, REPLACEMENT */ |
40f7914e HS |
1312 | LE64_BITMASK(BCH_MEMBER_DISCARD, struct bch_member, flags, 14, 15) |
1313 | LE64_BITMASK(BCH_MEMBER_DATA_ALLOWED, struct bch_member, flags, 15, 20) | |
1314 | LE64_BITMASK(BCH_MEMBER_GROUP, struct bch_member, flags, 20, 28) | |
1315 | LE64_BITMASK(BCH_MEMBER_DURABILITY, struct bch_member, flags, 28, 30) | |
c6b2826c | 1316 | LE64_BITMASK(BCH_MEMBER_FREESPACE_INITIALIZED, |
40f7914e | 1317 | struct bch_member, flags, 30, 31) |
1c6fdbd8 | 1318 | |
1c6fdbd8 KO |
1319 | #if 0 |
1320 | LE64_BITMASK(BCH_MEMBER_NR_READ_ERRORS, struct bch_member, flags[1], 0, 20); | |
1321 | LE64_BITMASK(BCH_MEMBER_NR_WRITE_ERRORS,struct bch_member, flags[1], 20, 40); | |
1322 | #endif | |
1323 | ||
2436cb9f KO |
1324 | #define BCH_MEMBER_STATES() \ |
1325 | x(rw, 0) \ | |
1326 | x(ro, 1) \ | |
1327 | x(failed, 2) \ | |
1328 | x(spare, 3) | |
1329 | ||
1c6fdbd8 | 1330 | enum bch_member_state { |
2436cb9f KO |
1331 | #define x(t, n) BCH_MEMBER_STATE_##t = n, |
1332 | BCH_MEMBER_STATES() | |
1333 | #undef x | |
1334 | BCH_MEMBER_STATE_NR | |
1c6fdbd8 KO |
1335 | }; |
1336 | ||
9af26120 | 1337 | struct bch_sb_field_members_v1 { |
1c6fdbd8 | 1338 | struct bch_sb_field field; |
3f7b9713 HS |
1339 | struct bch_member _members[]; //Members are now variable size |
1340 | }; | |
1341 | ||
1342 | struct bch_sb_field_members_v2 { | |
1343 | struct bch_sb_field field; | |
1344 | __le16 member_bytes; //size of single member entry | |
1345 | u8 pad[6]; | |
1346 | struct bch_member _members[]; | |
1c6fdbd8 KO |
1347 | }; |
1348 | ||
1349 | /* BCH_SB_FIELD_crypt: */ | |
1350 | ||
1351 | struct nonce { | |
1352 | __le32 d[4]; | |
1353 | }; | |
1354 | ||
1355 | struct bch_key { | |
1356 | __le64 key[4]; | |
1357 | }; | |
1358 | ||
1359 | #define BCH_KEY_MAGIC \ | |
a5cf5a4b KO |
1360 | (((__u64) 'b' << 0)|((__u64) 'c' << 8)| \ |
1361 | ((__u64) 'h' << 16)|((__u64) '*' << 24)| \ | |
1362 | ((__u64) '*' << 32)|((__u64) 'k' << 40)| \ | |
1363 | ((__u64) 'e' << 48)|((__u64) 'y' << 56)) | |
1c6fdbd8 KO |
1364 | |
1365 | struct bch_encrypted_key { | |
1366 | __le64 magic; | |
1367 | struct bch_key key; | |
1368 | }; | |
1369 | ||
1370 | /* | |
1371 | * If this field is present in the superblock, it stores an encryption key which | |
1372 | * is used encrypt all other data/metadata. The key will normally be encrypted | |
1373 | * with the key userspace provides, but if encryption has been turned off we'll | |
1374 | * just store the master key unencrypted in the superblock so we can access the | |
1375 | * previously encrypted data. | |
1376 | */ | |
1377 | struct bch_sb_field_crypt { | |
1378 | struct bch_sb_field field; | |
1379 | ||
1380 | __le64 flags; | |
1381 | __le64 kdf_flags; | |
1382 | struct bch_encrypted_key key; | |
1383 | }; | |
1384 | ||
1385 | LE64_BITMASK(BCH_CRYPT_KDF_TYPE, struct bch_sb_field_crypt, flags, 0, 4); | |
1386 | ||
1387 | enum bch_kdf_types { | |
1388 | BCH_KDF_SCRYPT = 0, | |
1389 | BCH_KDF_NR = 1, | |
1390 | }; | |
1391 | ||
1392 | /* stored as base 2 log of scrypt params: */ | |
1393 | LE64_BITMASK(BCH_KDF_SCRYPT_N, struct bch_sb_field_crypt, kdf_flags, 0, 16); | |
1394 | LE64_BITMASK(BCH_KDF_SCRYPT_R, struct bch_sb_field_crypt, kdf_flags, 16, 32); | |
1395 | LE64_BITMASK(BCH_KDF_SCRYPT_P, struct bch_sb_field_crypt, kdf_flags, 32, 48); | |
1396 | ||
1397 | /* BCH_SB_FIELD_replicas: */ | |
1398 | ||
89fd25be | 1399 | #define BCH_DATA_TYPES() \ |
822835ff | 1400 | x(free, 0) \ |
89fd25be KO |
1401 | x(sb, 1) \ |
1402 | x(journal, 2) \ | |
1403 | x(btree, 3) \ | |
1404 | x(user, 4) \ | |
af4d05c4 | 1405 | x(cached, 5) \ |
822835ff KO |
1406 | x(parity, 6) \ |
1407 | x(stripe, 7) \ | |
1408 | x(need_gc_gens, 8) \ | |
1409 | x(need_discard, 9) | |
89fd25be | 1410 | |
1c6fdbd8 | 1411 | enum bch_data_type { |
89fd25be KO |
1412 | #define x(t, n) BCH_DATA_##t, |
1413 | BCH_DATA_TYPES() | |
1414 | #undef x | |
1415 | BCH_DATA_NR | |
1c6fdbd8 KO |
1416 | }; |
1417 | ||
822835ff KO |
1418 | static inline bool data_type_is_empty(enum bch_data_type type) |
1419 | { | |
1420 | switch (type) { | |
1421 | case BCH_DATA_free: | |
1422 | case BCH_DATA_need_gc_gens: | |
1423 | case BCH_DATA_need_discard: | |
1424 | return true; | |
1425 | default: | |
1426 | return false; | |
1427 | } | |
1428 | } | |
1429 | ||
1430 | static inline bool data_type_is_hidden(enum bch_data_type type) | |
1431 | { | |
1432 | switch (type) { | |
1433 | case BCH_DATA_sb: | |
1434 | case BCH_DATA_journal: | |
1435 | return true; | |
1436 | default: | |
1437 | return false; | |
1438 | } | |
1439 | } | |
1440 | ||
af9d3bc2 KO |
1441 | struct bch_replicas_entry_v0 { |
1442 | __u8 data_type; | |
1443 | __u8 nr_devs; | |
5cfd6977 | 1444 | __u8 devs[]; |
fd0c7679 | 1445 | } __packed; |
af9d3bc2 KO |
1446 | |
1447 | struct bch_sb_field_replicas_v0 { | |
1448 | struct bch_sb_field field; | |
5cfd6977 | 1449 | struct bch_replicas_entry_v0 entries[]; |
fd0c7679 | 1450 | } __packed __aligned(8); |
af9d3bc2 | 1451 | |
086a52f7 | 1452 | struct bch_replicas_entry_v1 { |
7a920560 KO |
1453 | __u8 data_type; |
1454 | __u8 nr_devs; | |
af9d3bc2 | 1455 | __u8 nr_required; |
5cfd6977 | 1456 | __u8 devs[]; |
fd0c7679 | 1457 | } __packed; |
1c6fdbd8 | 1458 | |
22502ac2 KO |
1459 | #define replicas_entry_bytes(_i) \ |
1460 | (offsetof(typeof(*(_i)), devs) + (_i)->nr_devs) | |
1461 | ||
1c6fdbd8 KO |
1462 | struct bch_sb_field_replicas { |
1463 | struct bch_sb_field field; | |
086a52f7 | 1464 | struct bch_replicas_entry_v1 entries[]; |
fd0c7679 | 1465 | } __packed __aligned(8); |
1c6fdbd8 KO |
1466 | |
1467 | /* BCH_SB_FIELD_quota: */ | |
1468 | ||
1469 | struct bch_sb_quota_counter { | |
1470 | __le32 timelimit; | |
1471 | __le32 warnlimit; | |
1472 | }; | |
1473 | ||
1474 | struct bch_sb_quota_type { | |
1475 | __le64 flags; | |
1476 | struct bch_sb_quota_counter c[Q_COUNTERS]; | |
1477 | }; | |
1478 | ||
1479 | struct bch_sb_field_quota { | |
1480 | struct bch_sb_field field; | |
1481 | struct bch_sb_quota_type q[QTYP_NR]; | |
fd0c7679 | 1482 | } __packed __aligned(8); |
1c6fdbd8 KO |
1483 | |
1484 | /* BCH_SB_FIELD_disk_groups: */ | |
1485 | ||
1486 | #define BCH_SB_LABEL_SIZE 32 | |
1487 | ||
1488 | struct bch_disk_group { | |
1489 | __u8 label[BCH_SB_LABEL_SIZE]; | |
1490 | __le64 flags[2]; | |
fd0c7679 | 1491 | } __packed __aligned(8); |
1c6fdbd8 KO |
1492 | |
1493 | LE64_BITMASK(BCH_GROUP_DELETED, struct bch_disk_group, flags[0], 0, 1) | |
1494 | LE64_BITMASK(BCH_GROUP_DATA_ALLOWED, struct bch_disk_group, flags[0], 1, 6) | |
1495 | LE64_BITMASK(BCH_GROUP_PARENT, struct bch_disk_group, flags[0], 6, 24) | |
1496 | ||
1497 | struct bch_sb_field_disk_groups { | |
1498 | struct bch_sb_field field; | |
5cfd6977 | 1499 | struct bch_disk_group entries[]; |
fd0c7679 | 1500 | } __packed __aligned(8); |
1c6fdbd8 | 1501 | |
104c6974 DH |
1502 | /* BCH_SB_FIELD_counters */ |
1503 | ||
674cfc26 KO |
1504 | #define BCH_PERSISTENT_COUNTERS() \ |
1505 | x(io_read, 0) \ | |
1506 | x(io_write, 1) \ | |
1507 | x(io_move, 2) \ | |
1508 | x(bucket_invalidate, 3) \ | |
1509 | x(bucket_discard, 4) \ | |
1510 | x(bucket_alloc, 5) \ | |
1511 | x(bucket_alloc_fail, 6) \ | |
1512 | x(btree_cache_scan, 7) \ | |
1513 | x(btree_cache_reap, 8) \ | |
1514 | x(btree_cache_cannibalize, 9) \ | |
1515 | x(btree_cache_cannibalize_lock, 10) \ | |
1516 | x(btree_cache_cannibalize_lock_fail, 11) \ | |
1517 | x(btree_cache_cannibalize_unlock, 12) \ | |
1518 | x(btree_node_write, 13) \ | |
1519 | x(btree_node_read, 14) \ | |
1520 | x(btree_node_compact, 15) \ | |
1521 | x(btree_node_merge, 16) \ | |
1522 | x(btree_node_split, 17) \ | |
1523 | x(btree_node_rewrite, 18) \ | |
1524 | x(btree_node_alloc, 19) \ | |
1525 | x(btree_node_free, 20) \ | |
1526 | x(btree_node_set_root, 21) \ | |
1527 | x(btree_path_relock_fail, 22) \ | |
1528 | x(btree_path_upgrade_fail, 23) \ | |
1529 | x(btree_reserve_get_fail, 24) \ | |
1530 | x(journal_entry_full, 25) \ | |
1531 | x(journal_full, 26) \ | |
1532 | x(journal_reclaim_finish, 27) \ | |
1533 | x(journal_reclaim_start, 28) \ | |
1534 | x(journal_write, 29) \ | |
1535 | x(read_promote, 30) \ | |
1536 | x(read_bounce, 31) \ | |
1537 | x(read_split, 33) \ | |
1538 | x(read_retry, 32) \ | |
1539 | x(read_reuse_race, 34) \ | |
1540 | x(move_extent_read, 35) \ | |
1541 | x(move_extent_write, 36) \ | |
1542 | x(move_extent_finish, 37) \ | |
1543 | x(move_extent_fail, 38) \ | |
ae4d612c | 1544 | x(move_extent_start_fail, 39) \ |
674cfc26 KO |
1545 | x(copygc, 40) \ |
1546 | x(copygc_wait, 41) \ | |
1547 | x(gc_gens_end, 42) \ | |
1548 | x(gc_gens_start, 43) \ | |
1549 | x(trans_blocked_journal_reclaim, 44) \ | |
1550 | x(trans_restart_btree_node_reused, 45) \ | |
1551 | x(trans_restart_btree_node_split, 46) \ | |
1552 | x(trans_restart_fault_inject, 47) \ | |
1553 | x(trans_restart_iter_upgrade, 48) \ | |
1554 | x(trans_restart_journal_preres_get, 49) \ | |
1555 | x(trans_restart_journal_reclaim, 50) \ | |
1556 | x(trans_restart_journal_res_get, 51) \ | |
1557 | x(trans_restart_key_cache_key_realloced, 52) \ | |
1558 | x(trans_restart_key_cache_raced, 53) \ | |
1559 | x(trans_restart_mark_replicas, 54) \ | |
1560 | x(trans_restart_mem_realloced, 55) \ | |
1561 | x(trans_restart_memory_allocation_failure, 56) \ | |
1562 | x(trans_restart_relock, 57) \ | |
1563 | x(trans_restart_relock_after_fill, 58) \ | |
1564 | x(trans_restart_relock_key_cache_fill, 59) \ | |
1565 | x(trans_restart_relock_next_node, 60) \ | |
1566 | x(trans_restart_relock_parent_for_fill, 61) \ | |
1567 | x(trans_restart_relock_path, 62) \ | |
1568 | x(trans_restart_relock_path_intent, 63) \ | |
1569 | x(trans_restart_too_many_iters, 64) \ | |
1570 | x(trans_restart_traverse, 65) \ | |
1571 | x(trans_restart_upgrade, 66) \ | |
1572 | x(trans_restart_would_deadlock, 67) \ | |
1573 | x(trans_restart_would_deadlock_write, 68) \ | |
1574 | x(trans_restart_injected, 69) \ | |
1575 | x(trans_restart_key_cache_upgrade, 70) \ | |
1576 | x(trans_traverse_all, 71) \ | |
1577 | x(transaction_commit, 72) \ | |
33bd5d06 | 1578 | x(write_super, 73) \ |
920e69bc | 1579 | x(trans_restart_would_deadlock_recursion_limit, 74) \ |
e151580d | 1580 | x(trans_restart_write_buffer_flush, 75) \ |
56db2429 KO |
1581 | x(trans_restart_split_race, 76) \ |
1582 | x(write_buffer_flush_slowpath, 77) \ | |
1583 | x(write_buffer_flush_sync, 78) | |
104c6974 DH |
1584 | |
1585 | enum bch_persistent_counters { | |
1586 | #define x(t, n, ...) BCH_COUNTER_##t, | |
1587 | BCH_PERSISTENT_COUNTERS() | |
1588 | #undef x | |
1589 | BCH_COUNTER_NR | |
1590 | }; | |
1591 | ||
1592 | struct bch_sb_field_counters { | |
1593 | struct bch_sb_field field; | |
5cfd6977 | 1594 | __le64 d[]; |
104c6974 DH |
1595 | }; |
1596 | ||
1c6fdbd8 KO |
1597 | /* |
1598 | * On clean shutdown, store btree roots and current journal sequence number in | |
1599 | * the superblock: | |
1600 | */ | |
1601 | struct jset_entry { | |
1602 | __le16 u64s; | |
1603 | __u8 btree_id; | |
1604 | __u8 level; | |
1605 | __u8 type; /* designates what this jset holds */ | |
1606 | __u8 pad[3]; | |
1607 | ||
5cfd6977 KO |
1608 | struct bkey_i start[0]; |
1609 | __u64 _data[]; | |
1c6fdbd8 KO |
1610 | }; |
1611 | ||
1612 | struct bch_sb_field_clean { | |
1613 | struct bch_sb_field field; | |
1614 | ||
1615 | __le32 flags; | |
2abe5420 KO |
1616 | __le16 _read_clock; /* no longer used */ |
1617 | __le16 _write_clock; | |
1c6fdbd8 KO |
1618 | __le64 journal_seq; |
1619 | ||
5cfd6977 KO |
1620 | struct jset_entry start[0]; |
1621 | __u64 _data[]; | |
1c6fdbd8 KO |
1622 | }; |
1623 | ||
1dd7f9d9 KO |
1624 | struct journal_seq_blacklist_entry { |
1625 | __le64 start; | |
1626 | __le64 end; | |
1627 | }; | |
1628 | ||
1629 | struct bch_sb_field_journal_seq_blacklist { | |
1630 | struct bch_sb_field field; | |
6dfa10ab | 1631 | struct journal_seq_blacklist_entry start[]; |
1dd7f9d9 KO |
1632 | }; |
1633 | ||
f5d26fa3 KO |
1634 | struct bch_sb_field_errors { |
1635 | struct bch_sb_field field; | |
1636 | struct bch_sb_field_error_entry { | |
1637 | __le64 v; | |
1638 | __le64 last_error_time; | |
1639 | } entries[]; | |
1640 | }; | |
1641 | ||
1642 | LE64_BITMASK(BCH_SB_ERROR_ENTRY_ID, struct bch_sb_field_error_entry, v, 0, 16); | |
1643 | LE64_BITMASK(BCH_SB_ERROR_ENTRY_NR, struct bch_sb_field_error_entry, v, 16, 64); | |
1644 | ||
8b16413c KO |
1645 | struct bch_sb_field_ext { |
1646 | struct bch_sb_field field; | |
1647 | __le64 recovery_passes_required[2]; | |
1648 | __le64 errors_silent[8]; | |
1649 | }; | |
1650 | ||
84f16387 KO |
1651 | struct bch_sb_field_downgrade_entry { |
1652 | __le16 version; | |
1653 | __le64 recovery_passes[2]; | |
1654 | __le16 nr_errors; | |
1655 | __le16 errors[] __counted_by(nr_errors); | |
1656 | } __packed __aligned(2); | |
1657 | ||
1658 | struct bch_sb_field_downgrade { | |
1659 | struct bch_sb_field field; | |
1660 | struct bch_sb_field_downgrade_entry entries[]; | |
1661 | }; | |
1662 | ||
1c6fdbd8 KO |
1663 | /* Superblock: */ |
1664 | ||
1665 | /* | |
26609b61 KO |
1666 | * New versioning scheme: |
1667 | * One common version number for all on disk data structures - superblock, btree | |
1668 | * nodes, journal entries | |
1c6fdbd8 | 1669 | */ |
ba8eeae8 KO |
1670 | #define BCH_VERSION_MAJOR(_v) ((__u16) ((_v) >> 10)) |
1671 | #define BCH_VERSION_MINOR(_v) ((__u16) ((_v) & ~(~0U << 10))) | |
1672 | #define BCH_VERSION(_major, _minor) (((_major) << 10)|(_minor) << 0) | |
26609b61 | 1673 | |
065bd335 KO |
1674 | #define RECOVERY_PASS_ALL_FSCK (1ULL << 63) |
1675 | ||
84f16387 KO |
1676 | /* |
1677 | * field 1: version name | |
1678 | * field 2: BCH_VERSION(major, minor) | |
1679 | * field 3: recovery passess required on upgrade | |
1680 | */ | |
065bd335 KO |
1681 | #define BCH_METADATA_VERSIONS() \ |
1682 | x(bkey_renumber, BCH_VERSION(0, 10), \ | |
1683 | RECOVERY_PASS_ALL_FSCK) \ | |
1684 | x(inode_btree_change, BCH_VERSION(0, 11), \ | |
1685 | RECOVERY_PASS_ALL_FSCK) \ | |
1686 | x(snapshot, BCH_VERSION(0, 12), \ | |
1687 | RECOVERY_PASS_ALL_FSCK) \ | |
1688 | x(inode_backpointers, BCH_VERSION(0, 13), \ | |
1689 | RECOVERY_PASS_ALL_FSCK) \ | |
1690 | x(btree_ptr_sectors_written, BCH_VERSION(0, 14), \ | |
1691 | RECOVERY_PASS_ALL_FSCK) \ | |
1692 | x(snapshot_2, BCH_VERSION(0, 15), \ | |
1693 | BIT_ULL(BCH_RECOVERY_PASS_fs_upgrade_for_subvolumes)| \ | |
1694 | BIT_ULL(BCH_RECOVERY_PASS_initialize_subvolumes)| \ | |
1695 | RECOVERY_PASS_ALL_FSCK) \ | |
1696 | x(reflink_p_fix, BCH_VERSION(0, 16), \ | |
1697 | BIT_ULL(BCH_RECOVERY_PASS_fix_reflink_p)) \ | |
1698 | x(subvol_dirent, BCH_VERSION(0, 17), \ | |
1699 | RECOVERY_PASS_ALL_FSCK) \ | |
1700 | x(inode_v2, BCH_VERSION(0, 18), \ | |
1701 | RECOVERY_PASS_ALL_FSCK) \ | |
1702 | x(freespace, BCH_VERSION(0, 19), \ | |
1703 | RECOVERY_PASS_ALL_FSCK) \ | |
1704 | x(alloc_v4, BCH_VERSION(0, 20), \ | |
1705 | RECOVERY_PASS_ALL_FSCK) \ | |
1706 | x(new_data_types, BCH_VERSION(0, 21), \ | |
1707 | RECOVERY_PASS_ALL_FSCK) \ | |
1708 | x(backpointers, BCH_VERSION(0, 22), \ | |
1709 | RECOVERY_PASS_ALL_FSCK) \ | |
1710 | x(inode_v3, BCH_VERSION(0, 23), \ | |
1711 | RECOVERY_PASS_ALL_FSCK) \ | |
1712 | x(unwritten_extents, BCH_VERSION(0, 24), \ | |
1713 | RECOVERY_PASS_ALL_FSCK) \ | |
1714 | x(bucket_gens, BCH_VERSION(0, 25), \ | |
1715 | BIT_ULL(BCH_RECOVERY_PASS_bucket_gens_init)| \ | |
1716 | RECOVERY_PASS_ALL_FSCK) \ | |
1717 | x(lru_v2, BCH_VERSION(0, 26), \ | |
1718 | RECOVERY_PASS_ALL_FSCK) \ | |
1719 | x(fragmentation_lru, BCH_VERSION(0, 27), \ | |
1720 | RECOVERY_PASS_ALL_FSCK) \ | |
1721 | x(no_bps_in_alloc_keys, BCH_VERSION(0, 28), \ | |
1722 | RECOVERY_PASS_ALL_FSCK) \ | |
1723 | x(snapshot_trees, BCH_VERSION(0, 29), \ | |
1724 | RECOVERY_PASS_ALL_FSCK) \ | |
1725 | x(major_minor, BCH_VERSION(1, 0), \ | |
f26c67f4 KO |
1726 | 0) \ |
1727 | x(snapshot_skiplists, BCH_VERSION(1, 1), \ | |
dde8cb11 KO |
1728 | BIT_ULL(BCH_RECOVERY_PASS_check_snapshots)) \ |
1729 | x(deleted_inodes, BCH_VERSION(1, 2), \ | |
fb3f57bb KO |
1730 | BIT_ULL(BCH_RECOVERY_PASS_check_inodes)) \ |
1731 | x(rebalance_work, BCH_VERSION(1, 3), \ | |
1732 | BIT_ULL(BCH_RECOVERY_PASS_set_fs_needs_rebalance)) | |
74b33393 | 1733 | |
26609b61 | 1734 | enum bcachefs_metadata_version { |
74b33393 | 1735 | bcachefs_metadata_version_min = 9, |
065bd335 | 1736 | #define x(t, n, upgrade_passes) bcachefs_metadata_version_##t = n, |
74b33393 KO |
1737 | BCH_METADATA_VERSIONS() |
1738 | #undef x | |
1739 | bcachefs_metadata_version_max | |
26609b61 | 1740 | }; |
1c6fdbd8 | 1741 | |
96dea3d5 | 1742 | static const __maybe_unused |
fb3f57bb | 1743 | unsigned bcachefs_metadata_required_upgrade_below = bcachefs_metadata_version_rebalance_work; |
1c59b483 | 1744 | |
26609b61 | 1745 | #define bcachefs_metadata_version_current (bcachefs_metadata_version_max - 1) |
1c6fdbd8 KO |
1746 | |
1747 | #define BCH_SB_SECTOR 8 | |
1748 | #define BCH_SB_MEMBERS_MAX 64 /* XXX kill */ | |
1749 | ||
1750 | struct bch_sb_layout { | |
1751 | __uuid_t magic; /* bcachefs superblock UUID */ | |
1752 | __u8 layout_type; | |
1753 | __u8 sb_max_size_bits; /* base 2 of 512 byte sectors */ | |
1754 | __u8 nr_superblocks; | |
1755 | __u8 pad[5]; | |
1756 | __le64 sb_offset[61]; | |
fd0c7679 | 1757 | } __packed __aligned(8); |
1c6fdbd8 KO |
1758 | |
1759 | #define BCH_SB_LAYOUT_SECTOR 7 | |
1760 | ||
1761 | /* | |
1762 | * @offset - sector where this sb was written | |
1763 | * @version - on disk format version | |
26609b61 KO |
1764 | * @version_min - Oldest metadata version this filesystem contains; so we can |
1765 | * safely drop compatibility code and refuse to mount filesystems | |
1766 | * we'd need it for | |
e1538212 | 1767 | * @magic - identifies as a bcachefs superblock (BCHFS_MAGIC) |
1c6fdbd8 KO |
1768 | * @seq - incremented each time superblock is written |
1769 | * @uuid - used for generating various magic numbers and identifying | |
1770 | * member devices, never changes | |
1771 | * @user_uuid - user visible UUID, may be changed | |
1772 | * @label - filesystem label | |
1773 | * @seq - identifies most recent superblock, incremented each time | |
1774 | * superblock is written | |
1775 | * @features - enabled incompatible features | |
1776 | */ | |
1777 | struct bch_sb { | |
1778 | struct bch_csum csum; | |
1779 | __le16 version; | |
1780 | __le16 version_min; | |
1781 | __le16 pad[2]; | |
1782 | __uuid_t magic; | |
1783 | __uuid_t uuid; | |
1784 | __uuid_t user_uuid; | |
1785 | __u8 label[BCH_SB_LABEL_SIZE]; | |
1786 | __le64 offset; | |
1787 | __le64 seq; | |
1788 | ||
1789 | __le16 block_size; | |
1790 | __u8 dev_idx; | |
1791 | __u8 nr_devices; | |
1792 | __le32 u64s; | |
1793 | ||
1794 | __le64 time_base_lo; | |
1795 | __le32 time_base_hi; | |
1796 | __le32 time_precision; | |
1797 | ||
1798 | __le64 flags[8]; | |
1799 | __le64 features[2]; | |
1800 | __le64 compat[2]; | |
1801 | ||
1802 | struct bch_sb_layout layout; | |
1803 | ||
5cfd6977 KO |
1804 | struct bch_sb_field start[0]; |
1805 | __le64 _data[]; | |
fd0c7679 | 1806 | } __packed __aligned(8); |
1c6fdbd8 KO |
1807 | |
1808 | /* | |
1809 | * Flags: | |
1810 | * BCH_SB_INITALIZED - set on first mount | |
1811 | * BCH_SB_CLEAN - did we shut down cleanly? Just a hint, doesn't affect | |
1812 | * behaviour of mount/recovery path: | |
1813 | * BCH_SB_INODE_32BIT - limit inode numbers to 32 bits | |
1814 | * BCH_SB_128_BIT_MACS - 128 bit macs instead of 80 | |
1815 | * BCH_SB_ENCRYPTION_TYPE - if nonzero encryption is enabled; overrides | |
1816 | * DATA/META_CSUM_TYPE. Also indicates encryption | |
1817 | * algorithm in use, if/when we get more than one | |
1818 | */ | |
1819 | ||
1820 | LE16_BITMASK(BCH_SB_BLOCK_SIZE, struct bch_sb, block_size, 0, 16); | |
1821 | ||
1822 | LE64_BITMASK(BCH_SB_INITIALIZED, struct bch_sb, flags[0], 0, 1); | |
1823 | LE64_BITMASK(BCH_SB_CLEAN, struct bch_sb, flags[0], 1, 2); | |
1824 | LE64_BITMASK(BCH_SB_CSUM_TYPE, struct bch_sb, flags[0], 2, 8); | |
1825 | LE64_BITMASK(BCH_SB_ERROR_ACTION, struct bch_sb, flags[0], 8, 12); | |
1826 | ||
1827 | LE64_BITMASK(BCH_SB_BTREE_NODE_SIZE, struct bch_sb, flags[0], 12, 28); | |
1828 | ||
1829 | LE64_BITMASK(BCH_SB_GC_RESERVE, struct bch_sb, flags[0], 28, 33); | |
1830 | LE64_BITMASK(BCH_SB_ROOT_RESERVE, struct bch_sb, flags[0], 33, 40); | |
1831 | ||
1832 | LE64_BITMASK(BCH_SB_META_CSUM_TYPE, struct bch_sb, flags[0], 40, 44); | |
1833 | LE64_BITMASK(BCH_SB_DATA_CSUM_TYPE, struct bch_sb, flags[0], 44, 48); | |
1834 | ||
1835 | LE64_BITMASK(BCH_SB_META_REPLICAS_WANT, struct bch_sb, flags[0], 48, 52); | |
1836 | LE64_BITMASK(BCH_SB_DATA_REPLICAS_WANT, struct bch_sb, flags[0], 52, 56); | |
1837 | ||
1838 | LE64_BITMASK(BCH_SB_POSIX_ACL, struct bch_sb, flags[0], 56, 57); | |
1839 | LE64_BITMASK(BCH_SB_USRQUOTA, struct bch_sb, flags[0], 57, 58); | |
1840 | LE64_BITMASK(BCH_SB_GRPQUOTA, struct bch_sb, flags[0], 58, 59); | |
1841 | LE64_BITMASK(BCH_SB_PRJQUOTA, struct bch_sb, flags[0], 59, 60); | |
1842 | ||
0bc166ff | 1843 | LE64_BITMASK(BCH_SB_HAS_ERRORS, struct bch_sb, flags[0], 60, 61); |
aae15aaf | 1844 | LE64_BITMASK(BCH_SB_HAS_TOPOLOGY_ERRORS,struct bch_sb, flags[0], 61, 62); |
0bc166ff | 1845 | |
7d6f07ed | 1846 | LE64_BITMASK(BCH_SB_BIG_ENDIAN, struct bch_sb, flags[0], 62, 63); |
36b8372b | 1847 | |
1c6fdbd8 | 1848 | LE64_BITMASK(BCH_SB_STR_HASH_TYPE, struct bch_sb, flags[1], 0, 4); |
e86e9124 | 1849 | LE64_BITMASK(BCH_SB_COMPRESSION_TYPE_LO,struct bch_sb, flags[1], 4, 8); |
1c6fdbd8 KO |
1850 | LE64_BITMASK(BCH_SB_INODE_32BIT, struct bch_sb, flags[1], 8, 9); |
1851 | ||
1852 | LE64_BITMASK(BCH_SB_128_BIT_MACS, struct bch_sb, flags[1], 9, 10); | |
1853 | LE64_BITMASK(BCH_SB_ENCRYPTION_TYPE, struct bch_sb, flags[1], 10, 14); | |
1854 | ||
1855 | /* | |
1856 | * Max size of an extent that may require bouncing to read or write | |
1857 | * (checksummed, compressed): 64k | |
1858 | */ | |
1859 | LE64_BITMASK(BCH_SB_ENCODED_EXTENT_MAX_BITS, | |
1860 | struct bch_sb, flags[1], 14, 20); | |
1861 | ||
1862 | LE64_BITMASK(BCH_SB_META_REPLICAS_REQ, struct bch_sb, flags[1], 20, 24); | |
1863 | LE64_BITMASK(BCH_SB_DATA_REPLICAS_REQ, struct bch_sb, flags[1], 24, 28); | |
1864 | ||
1865 | LE64_BITMASK(BCH_SB_PROMOTE_TARGET, struct bch_sb, flags[1], 28, 40); | |
1866 | LE64_BITMASK(BCH_SB_FOREGROUND_TARGET, struct bch_sb, flags[1], 40, 52); | |
1867 | LE64_BITMASK(BCH_SB_BACKGROUND_TARGET, struct bch_sb, flags[1], 52, 64); | |
1868 | ||
e86e9124 | 1869 | LE64_BITMASK(BCH_SB_BACKGROUND_COMPRESSION_TYPE_LO, |
1c6fdbd8 | 1870 | struct bch_sb, flags[2], 0, 4); |
a50ed7c8 | 1871 | LE64_BITMASK(BCH_SB_GC_RESERVE_BYTES, struct bch_sb, flags[2], 4, 64); |
1c6fdbd8 | 1872 | |
cd575ddf | 1873 | LE64_BITMASK(BCH_SB_ERASURE_CODE, struct bch_sb, flags[3], 0, 16); |
d042b040 | 1874 | LE64_BITMASK(BCH_SB_METADATA_TARGET, struct bch_sb, flags[3], 16, 28); |
b282a74f | 1875 | LE64_BITMASK(BCH_SB_SHARD_INUMS, struct bch_sb, flags[3], 28, 29); |
996fb577 | 1876 | LE64_BITMASK(BCH_SB_INODES_USE_KEY_CACHE,struct bch_sb, flags[3], 29, 30); |
2430e72f KO |
1877 | LE64_BITMASK(BCH_SB_JOURNAL_FLUSH_DELAY,struct bch_sb, flags[3], 30, 62); |
1878 | LE64_BITMASK(BCH_SB_JOURNAL_FLUSH_DISABLED,struct bch_sb, flags[3], 62, 63); | |
1879 | LE64_BITMASK(BCH_SB_JOURNAL_RECLAIM_DELAY,struct bch_sb, flags[4], 0, 32); | |
fb64f3fd | 1880 | LE64_BITMASK(BCH_SB_JOURNAL_TRANSACTION_NAMES,struct bch_sb, flags[4], 32, 33); |
a8b3a677 | 1881 | LE64_BITMASK(BCH_SB_NOCOW, struct bch_sb, flags[4], 33, 34); |
920e69bc | 1882 | LE64_BITMASK(BCH_SB_WRITE_BUFFER_SIZE, struct bch_sb, flags[4], 34, 54); |
3045bb95 | 1883 | LE64_BITMASK(BCH_SB_VERSION_UPGRADE, struct bch_sb, flags[4], 54, 56); |
cd575ddf | 1884 | |
e86e9124 KO |
1885 | LE64_BITMASK(BCH_SB_COMPRESSION_TYPE_HI,struct bch_sb, flags[4], 56, 60); |
1886 | LE64_BITMASK(BCH_SB_BACKGROUND_COMPRESSION_TYPE_HI, | |
1887 | struct bch_sb, flags[4], 60, 64); | |
24964e1c KO |
1888 | |
1889 | LE64_BITMASK(BCH_SB_VERSION_UPGRADE_COMPLETE, | |
1890 | struct bch_sb, flags[5], 0, 16); | |
1891 | ||
e86e9124 KO |
1892 | static inline __u64 BCH_SB_COMPRESSION_TYPE(const struct bch_sb *sb) |
1893 | { | |
1894 | return BCH_SB_COMPRESSION_TYPE_LO(sb) | (BCH_SB_COMPRESSION_TYPE_HI(sb) << 4); | |
1895 | } | |
1896 | ||
1897 | static inline void SET_BCH_SB_COMPRESSION_TYPE(struct bch_sb *sb, __u64 v) | |
1898 | { | |
1899 | SET_BCH_SB_COMPRESSION_TYPE_LO(sb, v); | |
1900 | SET_BCH_SB_COMPRESSION_TYPE_HI(sb, v >> 4); | |
1901 | } | |
1902 | ||
1903 | static inline __u64 BCH_SB_BACKGROUND_COMPRESSION_TYPE(const struct bch_sb *sb) | |
1904 | { | |
1905 | return BCH_SB_BACKGROUND_COMPRESSION_TYPE_LO(sb) | | |
1906 | (BCH_SB_BACKGROUND_COMPRESSION_TYPE_HI(sb) << 4); | |
1907 | } | |
1908 | ||
1909 | static inline void SET_BCH_SB_BACKGROUND_COMPRESSION_TYPE(struct bch_sb *sb, __u64 v) | |
1910 | { | |
1911 | SET_BCH_SB_BACKGROUND_COMPRESSION_TYPE_LO(sb, v); | |
1912 | SET_BCH_SB_BACKGROUND_COMPRESSION_TYPE_HI(sb, v >> 4); | |
1913 | } | |
1914 | ||
1c3ff72c KO |
1915 | /* |
1916 | * Features: | |
1917 | * | |
1918 | * journal_seq_blacklist_v3: gates BCH_SB_FIELD_journal_seq_blacklist | |
1919 | * reflink: gates KEY_TYPE_reflink | |
1920 | * inline_data: gates KEY_TYPE_inline_data | |
6404dcc9 | 1921 | * new_siphash: gates BCH_STR_HASH_siphash |
bcd6f3e0 | 1922 | * new_extent_overwrite: gates BTREE_NODE_NEW_EXTENT_OVERWRITE |
1c3ff72c KO |
1923 | */ |
1924 | #define BCH_SB_FEATURES() \ | |
1925 | x(lz4, 0) \ | |
1926 | x(gzip, 1) \ | |
1927 | x(zstd, 2) \ | |
1928 | x(atomic_nlink, 3) \ | |
1929 | x(ec, 4) \ | |
1930 | x(journal_seq_blacklist_v3, 5) \ | |
1931 | x(reflink, 6) \ | |
1932 | x(new_siphash, 7) \ | |
bcd6f3e0 | 1933 | x(inline_data, 8) \ |
ab05de4c | 1934 | x(new_extent_overwrite, 9) \ |
548b3d20 | 1935 | x(incompressible, 10) \ |
e3e464ac | 1936 | x(btree_ptr_v2, 11) \ |
6357d607 | 1937 | x(extents_above_btree_updates, 12) \ |
801a3de6 | 1938 | x(btree_updates_journalled, 13) \ |
a3e72262 | 1939 | x(reflink_inline_data, 14) \ |
adbcada4 | 1940 | x(new_varint, 15) \ |
7f4e1d5d | 1941 | x(journal_no_flush, 16) \ |
8042b5b7 KO |
1942 | x(alloc_v2, 17) \ |
1943 | x(extents_across_btree_nodes, 18) | |
1944 | ||
1945 | #define BCH_SB_FEATURES_ALWAYS \ | |
1946 | ((1ULL << BCH_FEATURE_new_extent_overwrite)| \ | |
1947 | (1ULL << BCH_FEATURE_extents_above_btree_updates)|\ | |
1948 | (1ULL << BCH_FEATURE_btree_updates_journalled)|\ | |
73590619 | 1949 | (1ULL << BCH_FEATURE_alloc_v2)|\ |
8042b5b7 | 1950 | (1ULL << BCH_FEATURE_extents_across_btree_nodes)) |
1c3ff72c | 1951 | |
b807a0c8 | 1952 | #define BCH_SB_FEATURES_ALL \ |
8042b5b7 KO |
1953 | (BCH_SB_FEATURES_ALWAYS| \ |
1954 | (1ULL << BCH_FEATURE_new_siphash)| \ | |
e3e464ac | 1955 | (1ULL << BCH_FEATURE_btree_ptr_v2)| \ |
adbcada4 | 1956 | (1ULL << BCH_FEATURE_new_varint)| \ |
73590619 | 1957 | (1ULL << BCH_FEATURE_journal_no_flush)) |
b807a0c8 | 1958 | |
1c3ff72c KO |
1959 | enum bch_sb_feature { |
1960 | #define x(f, n) BCH_FEATURE_##f, | |
1961 | BCH_SB_FEATURES() | |
1962 | #undef x | |
c258f28e | 1963 | BCH_FEATURE_NR, |
1c6fdbd8 KO |
1964 | }; |
1965 | ||
19dd3172 KO |
1966 | #define BCH_SB_COMPAT() \ |
1967 | x(alloc_info, 0) \ | |
1968 | x(alloc_metadata, 1) \ | |
1969 | x(extents_above_btree_updates_done, 2) \ | |
1970 | x(bformat_overflow_done, 3) | |
1971 | ||
1df42b57 | 1972 | enum bch_sb_compat { |
19dd3172 KO |
1973 | #define x(f, n) BCH_COMPAT_##f, |
1974 | BCH_SB_COMPAT() | |
1975 | #undef x | |
1976 | BCH_COMPAT_NR, | |
1df42b57 KO |
1977 | }; |
1978 | ||
1c6fdbd8 KO |
1979 | /* options: */ |
1980 | ||
3045bb95 KO |
1981 | #define BCH_VERSION_UPGRADE_OPTS() \ |
1982 | x(compatible, 0) \ | |
1983 | x(incompatible, 1) \ | |
1984 | x(none, 2) | |
1985 | ||
1986 | enum bch_version_upgrade_opts { | |
1987 | #define x(t, n) BCH_VERSION_UPGRADE_##t = n, | |
1988 | BCH_VERSION_UPGRADE_OPTS() | |
1989 | #undef x | |
1990 | }; | |
1991 | ||
1c6fdbd8 KO |
1992 | #define BCH_REPLICAS_MAX 4U |
1993 | ||
ffb7c3d3 KO |
1994 | #define BCH_BKEY_PTRS_MAX 16U |
1995 | ||
2436cb9f KO |
1996 | #define BCH_ERROR_ACTIONS() \ |
1997 | x(continue, 0) \ | |
1998 | x(ro, 1) \ | |
1999 | x(panic, 2) | |
2000 | ||
1c6fdbd8 | 2001 | enum bch_error_actions { |
2436cb9f KO |
2002 | #define x(t, n) BCH_ON_ERROR_##t = n, |
2003 | BCH_ERROR_ACTIONS() | |
2004 | #undef x | |
2005 | BCH_ON_ERROR_NR | |
1c6fdbd8 KO |
2006 | }; |
2007 | ||
6404dcc9 KO |
2008 | #define BCH_STR_HASH_TYPES() \ |
2009 | x(crc32c, 0) \ | |
2010 | x(crc64, 1) \ | |
2011 | x(siphash_old, 2) \ | |
2012 | x(siphash, 3) | |
2013 | ||
73501ab8 | 2014 | enum bch_str_hash_type { |
6404dcc9 KO |
2015 | #define x(t, n) BCH_STR_HASH_##t = n, |
2016 | BCH_STR_HASH_TYPES() | |
2017 | #undef x | |
2018 | BCH_STR_HASH_NR | |
73501ab8 KO |
2019 | }; |
2020 | ||
2436cb9f KO |
2021 | #define BCH_STR_HASH_OPTS() \ |
2022 | x(crc32c, 0) \ | |
2023 | x(crc64, 1) \ | |
2024 | x(siphash, 2) | |
2025 | ||
73501ab8 | 2026 | enum bch_str_hash_opts { |
2436cb9f KO |
2027 | #define x(t, n) BCH_STR_HASH_OPT_##t = n, |
2028 | BCH_STR_HASH_OPTS() | |
2029 | #undef x | |
2030 | BCH_STR_HASH_OPT_NR | |
1c6fdbd8 KO |
2031 | }; |
2032 | ||
6404dcc9 KO |
2033 | #define BCH_CSUM_TYPES() \ |
2034 | x(none, 0) \ | |
2035 | x(crc32c_nonzero, 1) \ | |
2036 | x(crc64_nonzero, 2) \ | |
2037 | x(chacha20_poly1305_80, 3) \ | |
2038 | x(chacha20_poly1305_128, 4) \ | |
2039 | x(crc32c, 5) \ | |
2040 | x(crc64, 6) \ | |
2041 | x(xxhash, 7) | |
2042 | ||
1c3ff72c | 2043 | enum bch_csum_type { |
6404dcc9 KO |
2044 | #define x(t, n) BCH_CSUM_##t = n, |
2045 | BCH_CSUM_TYPES() | |
2046 | #undef x | |
2047 | BCH_CSUM_NR | |
1c3ff72c KO |
2048 | }; |
2049 | ||
96dea3d5 | 2050 | static const __maybe_unused unsigned bch_crc_bytes[] = { |
6404dcc9 KO |
2051 | [BCH_CSUM_none] = 0, |
2052 | [BCH_CSUM_crc32c_nonzero] = 4, | |
2053 | [BCH_CSUM_crc32c] = 4, | |
2054 | [BCH_CSUM_crc64_nonzero] = 8, | |
2055 | [BCH_CSUM_crc64] = 8, | |
2056 | [BCH_CSUM_xxhash] = 8, | |
2057 | [BCH_CSUM_chacha20_poly1305_80] = 10, | |
2058 | [BCH_CSUM_chacha20_poly1305_128] = 16, | |
1c3ff72c KO |
2059 | }; |
2060 | ||
2061 | static inline _Bool bch2_csum_type_is_encryption(enum bch_csum_type type) | |
2062 | { | |
2063 | switch (type) { | |
6404dcc9 KO |
2064 | case BCH_CSUM_chacha20_poly1305_80: |
2065 | case BCH_CSUM_chacha20_poly1305_128: | |
1c3ff72c KO |
2066 | return true; |
2067 | default: | |
2068 | return false; | |
2069 | } | |
2070 | } | |
2071 | ||
2436cb9f KO |
2072 | #define BCH_CSUM_OPTS() \ |
2073 | x(none, 0) \ | |
2074 | x(crc32c, 1) \ | |
41e63382 | 2075 | x(crc64, 2) \ |
2076 | x(xxhash, 3) | |
2436cb9f | 2077 | |
1c3ff72c | 2078 | enum bch_csum_opts { |
2436cb9f KO |
2079 | #define x(t, n) BCH_CSUM_OPT_##t = n, |
2080 | BCH_CSUM_OPTS() | |
2081 | #undef x | |
2082 | BCH_CSUM_OPT_NR | |
1c3ff72c KO |
2083 | }; |
2084 | ||
1c6fdbd8 | 2085 | #define BCH_COMPRESSION_TYPES() \ |
ab05de4c KO |
2086 | x(none, 0) \ |
2087 | x(lz4_old, 1) \ | |
2088 | x(gzip, 2) \ | |
2089 | x(lz4, 3) \ | |
2090 | x(zstd, 4) \ | |
2091 | x(incompressible, 5) | |
1c6fdbd8 | 2092 | |
1c3ff72c | 2093 | enum bch_compression_type { |
2436cb9f | 2094 | #define x(t, n) BCH_COMPRESSION_TYPE_##t = n, |
1c6fdbd8 | 2095 | BCH_COMPRESSION_TYPES() |
1c3ff72c KO |
2096 | #undef x |
2097 | BCH_COMPRESSION_TYPE_NR | |
2098 | }; | |
2099 | ||
2100 | #define BCH_COMPRESSION_OPTS() \ | |
2101 | x(none, 0) \ | |
2102 | x(lz4, 1) \ | |
2103 | x(gzip, 2) \ | |
2104 | x(zstd, 3) | |
2105 | ||
2106 | enum bch_compression_opts { | |
2436cb9f | 2107 | #define x(t, n) BCH_COMPRESSION_OPT_##t = n, |
1c3ff72c | 2108 | BCH_COMPRESSION_OPTS() |
1c6fdbd8 KO |
2109 | #undef x |
2110 | BCH_COMPRESSION_OPT_NR | |
2111 | }; | |
2112 | ||
2113 | /* | |
2114 | * Magic numbers | |
2115 | * | |
2116 | * The various other data structures have their own magic numbers, which are | |
2117 | * xored with the first part of the cache set's UUID | |
2118 | */ | |
2119 | ||
2120 | #define BCACHE_MAGIC \ | |
2121 | UUID_INIT(0xc68573f6, 0x4e1a, 0x45ca, \ | |
2122 | 0x82, 0x65, 0xf5, 0x7f, 0x48, 0xba, 0x6d, 0x81) | |
2123 | #define BCHFS_MAGIC \ | |
2124 | UUID_INIT(0xc68573f6, 0x66ce, 0x90a9, \ | |
2125 | 0xd9, 0x6a, 0x60, 0xcf, 0x80, 0x3d, 0xf7, 0xef) | |
2126 | ||
2127 | #define BCACHEFS_STATFS_MAGIC 0xca451a4e | |
2128 | ||
2129 | #define JSET_MAGIC __cpu_to_le64(0x245235c1a3625032ULL) | |
2130 | #define BSET_MAGIC __cpu_to_le64(0x90135c78b99e07f5ULL) | |
2131 | ||
2132 | static inline __le64 __bch2_sb_magic(struct bch_sb *sb) | |
2133 | { | |
2134 | __le64 ret; | |
a1019576 | 2135 | |
1c6fdbd8 KO |
2136 | memcpy(&ret, &sb->uuid, sizeof(ret)); |
2137 | return ret; | |
2138 | } | |
2139 | ||
2140 | static inline __u64 __jset_magic(struct bch_sb *sb) | |
2141 | { | |
2142 | return __le64_to_cpu(__bch2_sb_magic(sb) ^ JSET_MAGIC); | |
2143 | } | |
2144 | ||
2145 | static inline __u64 __bset_magic(struct bch_sb *sb) | |
2146 | { | |
2147 | return __le64_to_cpu(__bch2_sb_magic(sb) ^ BSET_MAGIC); | |
2148 | } | |
2149 | ||
2150 | /* Journal */ | |
2151 | ||
1c6fdbd8 KO |
2152 | #define JSET_KEYS_U64s (sizeof(struct jset_entry) / sizeof(__u64)) |
2153 | ||
2154 | #define BCH_JSET_ENTRY_TYPES() \ | |
2155 | x(btree_keys, 0) \ | |
2156 | x(btree_root, 1) \ | |
2157 | x(prio_ptrs, 2) \ | |
2158 | x(blacklist, 3) \ | |
2c5af169 | 2159 | x(blacklist_v2, 4) \ |
3577df5f | 2160 | x(usage, 5) \ |
2abe5420 | 2161 | x(data_usage, 6) \ |
180fb49d | 2162 | x(clock, 7) \ |
fb64f3fd | 2163 | x(dev_usage, 8) \ |
cb685ce7 | 2164 | x(log, 9) \ |
09caeabe KO |
2165 | x(overwrite, 10) \ |
2166 | x(write_buffer_keys, 11) | |
1c6fdbd8 KO |
2167 | |
2168 | enum { | |
2169 | #define x(f, nr) BCH_JSET_ENTRY_##f = nr, | |
2170 | BCH_JSET_ENTRY_TYPES() | |
2171 | #undef x | |
2172 | BCH_JSET_ENTRY_NR | |
2173 | }; | |
2174 | ||
2175 | /* | |
2176 | * Journal sequence numbers can be blacklisted: bsets record the max sequence | |
2177 | * number of all the journal entries they contain updates for, so that on | |
2178 | * recovery we can ignore those bsets that contain index updates newer that what | |
2179 | * made it into the journal. | |
2180 | * | |
2181 | * This means that we can't reuse that journal_seq - we have to skip it, and | |
2182 | * then record that we skipped it so that the next time we crash and recover we | |
2183 | * don't think there was a missing journal entry. | |
2184 | */ | |
2185 | struct jset_entry_blacklist { | |
2186 | struct jset_entry entry; | |
2187 | __le64 seq; | |
2188 | }; | |
2189 | ||
2190 | struct jset_entry_blacklist_v2 { | |
2191 | struct jset_entry entry; | |
2192 | __le64 start; | |
2193 | __le64 end; | |
2194 | }; | |
2195 | ||
528b18e6 KO |
2196 | #define BCH_FS_USAGE_TYPES() \ |
2197 | x(reserved, 0) \ | |
2198 | x(inodes, 1) \ | |
2199 | x(key_version, 2) | |
2200 | ||
2c5af169 | 2201 | enum { |
528b18e6 KO |
2202 | #define x(f, nr) BCH_FS_USAGE_##f = nr, |
2203 | BCH_FS_USAGE_TYPES() | |
2204 | #undef x | |
2205 | BCH_FS_USAGE_NR | |
2c5af169 KO |
2206 | }; |
2207 | ||
2208 | struct jset_entry_usage { | |
2209 | struct jset_entry entry; | |
3577df5f | 2210 | __le64 v; |
fd0c7679 | 2211 | } __packed; |
3577df5f KO |
2212 | |
2213 | struct jset_entry_data_usage { | |
2214 | struct jset_entry entry; | |
2215 | __le64 v; | |
086a52f7 | 2216 | struct bch_replicas_entry_v1 r; |
fd0c7679 | 2217 | } __packed; |
2c5af169 | 2218 | |
2abe5420 KO |
2219 | struct jset_entry_clock { |
2220 | struct jset_entry entry; | |
2221 | __u8 rw; | |
2222 | __u8 pad[7]; | |
2223 | __le64 time; | |
fd0c7679 | 2224 | } __packed; |
2abe5420 | 2225 | |
180fb49d KO |
2226 | struct jset_entry_dev_usage_type { |
2227 | __le64 buckets; | |
2228 | __le64 sectors; | |
2229 | __le64 fragmented; | |
fd0c7679 | 2230 | } __packed; |
180fb49d KO |
2231 | |
2232 | struct jset_entry_dev_usage { | |
2233 | struct jset_entry entry; | |
2234 | __le32 dev; | |
2235 | __u32 pad; | |
2236 | ||
9b34f02c KO |
2237 | __le64 _buckets_ec; /* No longer used */ |
2238 | __le64 _buckets_unavailable; /* No longer used */ | |
180fb49d KO |
2239 | |
2240 | struct jset_entry_dev_usage_type d[]; | |
bf5a261c | 2241 | }; |
180fb49d | 2242 | |
528b18e6 KO |
2243 | static inline unsigned jset_entry_dev_usage_nr_types(struct jset_entry_dev_usage *u) |
2244 | { | |
2245 | return (vstruct_bytes(&u->entry) - sizeof(struct jset_entry_dev_usage)) / | |
2246 | sizeof(struct jset_entry_dev_usage_type); | |
2247 | } | |
2248 | ||
fb64f3fd KO |
2249 | struct jset_entry_log { |
2250 | struct jset_entry entry; | |
2251 | u8 d[]; | |
fd0c7679 | 2252 | } __packed; |
fb64f3fd | 2253 | |
1c6fdbd8 KO |
2254 | /* |
2255 | * On disk format for a journal entry: | |
2256 | * seq is monotonically increasing; every journal entry has its own unique | |
2257 | * sequence number. | |
2258 | * | |
2259 | * last_seq is the oldest journal entry that still has keys the btree hasn't | |
2260 | * flushed to disk yet. | |
2261 | * | |
2262 | * version is for on disk format changes. | |
2263 | */ | |
2264 | struct jset { | |
2265 | struct bch_csum csum; | |
2266 | ||
2267 | __le64 magic; | |
2268 | __le64 seq; | |
2269 | __le32 version; | |
2270 | __le32 flags; | |
2271 | ||
2272 | __le32 u64s; /* size of d[] in u64s */ | |
2273 | ||
2274 | __u8 encrypted_start[0]; | |
2275 | ||
2abe5420 KO |
2276 | __le16 _read_clock; /* no longer used */ |
2277 | __le16 _write_clock; | |
1c6fdbd8 KO |
2278 | |
2279 | /* Sequence number of oldest dirty journal entry */ | |
2280 | __le64 last_seq; | |
2281 | ||
2282 | ||
5cfd6977 KO |
2283 | struct jset_entry start[0]; |
2284 | __u64 _data[]; | |
fd0c7679 | 2285 | } __packed __aligned(8); |
1c6fdbd8 KO |
2286 | |
2287 | LE32_BITMASK(JSET_CSUM_TYPE, struct jset, flags, 0, 4); | |
2288 | LE32_BITMASK(JSET_BIG_ENDIAN, struct jset, flags, 4, 5); | |
adbcada4 | 2289 | LE32_BITMASK(JSET_NO_FLUSH, struct jset, flags, 5, 6); |
1c6fdbd8 | 2290 | |
8b335bae | 2291 | #define BCH_JOURNAL_BUCKETS_MIN 8 |
1c6fdbd8 KO |
2292 | |
2293 | /* Btree: */ | |
2294 | ||
e8d2fe3b KO |
2295 | enum btree_id_flags { |
2296 | BTREE_ID_EXTENTS = BIT(0), | |
2297 | BTREE_ID_SNAPSHOTS = BIT(1), | |
d3c7727b KO |
2298 | BTREE_ID_SNAPSHOT_FIELD = BIT(2), |
2299 | BTREE_ID_DATA = BIT(3), | |
e8d2fe3b KO |
2300 | }; |
2301 | ||
2302 | #define BCH_BTREE_IDS() \ | |
2303 | x(extents, 0, BTREE_ID_EXTENTS|BTREE_ID_SNAPSHOTS|BTREE_ID_DATA,\ | |
2304 | BIT_ULL(KEY_TYPE_whiteout)| \ | |
2305 | BIT_ULL(KEY_TYPE_error)| \ | |
2306 | BIT_ULL(KEY_TYPE_cookie)| \ | |
2307 | BIT_ULL(KEY_TYPE_extent)| \ | |
2308 | BIT_ULL(KEY_TYPE_reservation)| \ | |
2309 | BIT_ULL(KEY_TYPE_reflink_p)| \ | |
2310 | BIT_ULL(KEY_TYPE_inline_data)) \ | |
2311 | x(inodes, 1, BTREE_ID_SNAPSHOTS, \ | |
2312 | BIT_ULL(KEY_TYPE_whiteout)| \ | |
2313 | BIT_ULL(KEY_TYPE_inode)| \ | |
2314 | BIT_ULL(KEY_TYPE_inode_v2)| \ | |
2315 | BIT_ULL(KEY_TYPE_inode_v3)| \ | |
2316 | BIT_ULL(KEY_TYPE_inode_generation)) \ | |
2317 | x(dirents, 2, BTREE_ID_SNAPSHOTS, \ | |
2318 | BIT_ULL(KEY_TYPE_whiteout)| \ | |
2319 | BIT_ULL(KEY_TYPE_hash_whiteout)| \ | |
2320 | BIT_ULL(KEY_TYPE_dirent)) \ | |
2321 | x(xattrs, 3, BTREE_ID_SNAPSHOTS, \ | |
2322 | BIT_ULL(KEY_TYPE_whiteout)| \ | |
2323 | BIT_ULL(KEY_TYPE_cookie)| \ | |
2324 | BIT_ULL(KEY_TYPE_hash_whiteout)| \ | |
2325 | BIT_ULL(KEY_TYPE_xattr)) \ | |
2326 | x(alloc, 4, 0, \ | |
2327 | BIT_ULL(KEY_TYPE_alloc)| \ | |
2328 | BIT_ULL(KEY_TYPE_alloc_v2)| \ | |
2329 | BIT_ULL(KEY_TYPE_alloc_v3)| \ | |
2330 | BIT_ULL(KEY_TYPE_alloc_v4)) \ | |
2331 | x(quotas, 5, 0, \ | |
2332 | BIT_ULL(KEY_TYPE_quota)) \ | |
2333 | x(stripes, 6, 0, \ | |
2334 | BIT_ULL(KEY_TYPE_stripe)) \ | |
2335 | x(reflink, 7, BTREE_ID_EXTENTS|BTREE_ID_DATA, \ | |
2336 | BIT_ULL(KEY_TYPE_reflink_v)| \ | |
2337 | BIT_ULL(KEY_TYPE_indirect_inline_data)) \ | |
2338 | x(subvolumes, 8, 0, \ | |
2339 | BIT_ULL(KEY_TYPE_subvolume)) \ | |
2340 | x(snapshots, 9, 0, \ | |
2341 | BIT_ULL(KEY_TYPE_snapshot)) \ | |
2342 | x(lru, 10, 0, \ | |
2343 | BIT_ULL(KEY_TYPE_set)) \ | |
2344 | x(freespace, 11, BTREE_ID_EXTENTS, \ | |
2345 | BIT_ULL(KEY_TYPE_set)) \ | |
2346 | x(need_discard, 12, 0, \ | |
2347 | BIT_ULL(KEY_TYPE_set)) \ | |
2348 | x(backpointers, 13, 0, \ | |
2349 | BIT_ULL(KEY_TYPE_backpointer)) \ | |
2350 | x(bucket_gens, 14, 0, \ | |
2351 | BIT_ULL(KEY_TYPE_bucket_gens)) \ | |
2352 | x(snapshot_trees, 15, 0, \ | |
dde8cb11 | 2353 | BIT_ULL(KEY_TYPE_snapshot_tree)) \ |
d3c7727b | 2354 | x(deleted_inodes, 16, BTREE_ID_SNAPSHOT_FIELD, \ |
aaad530a KO |
2355 | BIT_ULL(KEY_TYPE_set)) \ |
2356 | x(logged_ops, 17, 0, \ | |
f3e374ef | 2357 | BIT_ULL(KEY_TYPE_logged_op_truncate)| \ |
fb3f57bb | 2358 | BIT_ULL(KEY_TYPE_logged_op_finsert)) \ |
d3c7727b | 2359 | x(rebalance_work, 18, BTREE_ID_SNAPSHOT_FIELD, \ |
fb3f57bb | 2360 | BIT_ULL(KEY_TYPE_set)|BIT_ULL(KEY_TYPE_cookie)) |
1c6fdbd8 KO |
2361 | |
2362 | enum btree_id { | |
e8d2fe3b | 2363 | #define x(name, nr, ...) BTREE_ID_##name = nr, |
26609b61 KO |
2364 | BCH_BTREE_IDS() |
2365 | #undef x | |
1c6fdbd8 KO |
2366 | BTREE_ID_NR |
2367 | }; | |
2368 | ||
1c6fdbd8 KO |
2369 | #define BTREE_MAX_DEPTH 4U |
2370 | ||
2371 | /* Btree nodes */ | |
2372 | ||
1c6fdbd8 KO |
2373 | /* |
2374 | * Btree nodes | |
2375 | * | |
2376 | * On disk a btree node is a list/log of these; within each set the keys are | |
2377 | * sorted | |
2378 | */ | |
2379 | struct bset { | |
2380 | __le64 seq; | |
2381 | ||
2382 | /* | |
2383 | * Highest journal entry this bset contains keys for. | |
2384 | * If on recovery we don't see that journal entry, this bset is ignored: | |
2385 | * this allows us to preserve the order of all index updates after a | |
2386 | * crash, since the journal records a total order of all index updates | |
2387 | * and anything that didn't make it to the journal doesn't get used. | |
2388 | */ | |
2389 | __le64 journal_seq; | |
2390 | ||
2391 | __le32 flags; | |
2392 | __le16 version; | |
2393 | __le16 u64s; /* count of d[] in u64s */ | |
2394 | ||
5cfd6977 KO |
2395 | struct bkey_packed start[0]; |
2396 | __u64 _data[]; | |
fd0c7679 | 2397 | } __packed __aligned(8); |
1c6fdbd8 KO |
2398 | |
2399 | LE32_BITMASK(BSET_CSUM_TYPE, struct bset, flags, 0, 4); | |
2400 | ||
2401 | LE32_BITMASK(BSET_BIG_ENDIAN, struct bset, flags, 4, 5); | |
2402 | LE32_BITMASK(BSET_SEPARATE_WHITEOUTS, | |
2403 | struct bset, flags, 5, 6); | |
2404 | ||
e719fc34 KO |
2405 | /* Sector offset within the btree node: */ |
2406 | LE32_BITMASK(BSET_OFFSET, struct bset, flags, 16, 32); | |
2407 | ||
1c6fdbd8 KO |
2408 | struct btree_node { |
2409 | struct bch_csum csum; | |
2410 | __le64 magic; | |
2411 | ||
2412 | /* this flags field is encrypted, unlike bset->flags: */ | |
2413 | __le64 flags; | |
2414 | ||
2415 | /* Closed interval: */ | |
2416 | struct bpos min_key; | |
2417 | struct bpos max_key; | |
e751c01a | 2418 | struct bch_extent_ptr _ptr; /* not used anymore */ |
1c6fdbd8 KO |
2419 | struct bkey_format format; |
2420 | ||
2421 | union { | |
2422 | struct bset keys; | |
2423 | struct { | |
2424 | __u8 pad[22]; | |
2425 | __le16 u64s; | |
2426 | __u64 _data[0]; | |
2427 | ||
2428 | }; | |
2429 | }; | |
fd0c7679 | 2430 | } __packed __aligned(8); |
1c6fdbd8 | 2431 | |
4e1430a7 | 2432 | LE64_BITMASK(BTREE_NODE_ID_LO, struct btree_node, flags, 0, 4); |
1c6fdbd8 | 2433 | LE64_BITMASK(BTREE_NODE_LEVEL, struct btree_node, flags, 4, 8); |
bcd6f3e0 KO |
2434 | LE64_BITMASK(BTREE_NODE_NEW_EXTENT_OVERWRITE, |
2435 | struct btree_node, flags, 8, 9); | |
4e1430a7 KO |
2436 | LE64_BITMASK(BTREE_NODE_ID_HI, struct btree_node, flags, 9, 25); |
2437 | /* 25-32 unused */ | |
1c6fdbd8 KO |
2438 | LE64_BITMASK(BTREE_NODE_SEQ, struct btree_node, flags, 32, 64); |
2439 | ||
4e1430a7 KO |
2440 | static inline __u64 BTREE_NODE_ID(struct btree_node *n) |
2441 | { | |
2442 | return BTREE_NODE_ID_LO(n) | (BTREE_NODE_ID_HI(n) << 4); | |
2443 | } | |
2444 | ||
a5cf5a4b | 2445 | static inline void SET_BTREE_NODE_ID(struct btree_node *n, __u64 v) |
4e1430a7 KO |
2446 | { |
2447 | SET_BTREE_NODE_ID_LO(n, v); | |
2448 | SET_BTREE_NODE_ID_HI(n, v >> 4); | |
2449 | } | |
2450 | ||
1c6fdbd8 KO |
2451 | struct btree_node_entry { |
2452 | struct bch_csum csum; | |
2453 | ||
2454 | union { | |
2455 | struct bset keys; | |
2456 | struct { | |
2457 | __u8 pad[22]; | |
2458 | __le16 u64s; | |
2459 | __u64 _data[0]; | |
1c6fdbd8 KO |
2460 | }; |
2461 | }; | |
fd0c7679 | 2462 | } __packed __aligned(8); |
1c6fdbd8 KO |
2463 | |
2464 | #endif /* _BCACHEFS_FORMAT_H */ |