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 KO |
85 | #define BITMASK(name, type, field, offset, end) \ |
86 | static const unsigned name##_OFFSET = offset; \ | |
87 | static const unsigned name##_BITS = (end - offset); \ | |
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 KO |
100 | #define LE_BITMASK(_bits, name, type, field, offset, end) \ |
101 | static const unsigned name##_OFFSET = offset; \ | |
102 | static const unsigned name##_BITS = (end - offset); \ | |
103 | static const __u##_bits name##_MAX = (1ULL << (end - offset)) - 1; \ | |
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 | |
fd0c7679 | 154 | } __packed __aligned(4); |
1c6fdbd8 KO |
155 | |
156 | #define KEY_INODE_MAX ((__u64)~0ULL) | |
157 | #define KEY_OFFSET_MAX ((__u64)~0ULL) | |
158 | #define KEY_SNAPSHOT_MAX ((__u32)~0U) | |
159 | #define KEY_SIZE_MAX ((__u32)~0U) | |
160 | ||
e751c01a | 161 | static inline struct bpos SPOS(__u64 inode, __u64 offset, __u32 snapshot) |
1c6fdbd8 | 162 | { |
e751c01a KO |
163 | return (struct bpos) { |
164 | .inode = inode, | |
165 | .offset = offset, | |
166 | .snapshot = snapshot, | |
167 | }; | |
1c6fdbd8 KO |
168 | } |
169 | ||
e751c01a | 170 | #define POS_MIN SPOS(0, 0, 0) |
618b1c0e KO |
171 | #define POS_MAX SPOS(KEY_INODE_MAX, KEY_OFFSET_MAX, 0) |
172 | #define SPOS_MAX SPOS(KEY_INODE_MAX, KEY_OFFSET_MAX, KEY_SNAPSHOT_MAX) | |
e751c01a | 173 | #define POS(_inode, _offset) SPOS(_inode, _offset, 0) |
1c6fdbd8 KO |
174 | |
175 | /* Empty placeholder struct, for container_of() */ | |
176 | struct bch_val { | |
177 | __u64 __nothing[0]; | |
178 | }; | |
179 | ||
180 | struct bversion { | |
181 | #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ | |
182 | __u64 lo; | |
183 | __u32 hi; | |
184 | #elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ | |
185 | __u32 hi; | |
186 | __u64 lo; | |
187 | #endif | |
fd0c7679 | 188 | } __packed __aligned(4); |
1c6fdbd8 KO |
189 | |
190 | struct bkey { | |
191 | /* Size of combined key and value, in u64s */ | |
192 | __u8 u64s; | |
193 | ||
194 | /* Format of key (0 for format local to btree node) */ | |
195 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
196 | __u8 format:7, | |
197 | needs_whiteout:1; | |
198 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
199 | __u8 needs_whiteout:1, | |
200 | format:7; | |
201 | #else | |
202 | #error edit for your odd byteorder. | |
203 | #endif | |
204 | ||
205 | /* Type of the value */ | |
206 | __u8 type; | |
207 | ||
208 | #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ | |
209 | __u8 pad[1]; | |
210 | ||
211 | struct bversion version; | |
212 | __u32 size; /* extent size, in sectors */ | |
213 | struct bpos p; | |
214 | #elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ | |
215 | struct bpos p; | |
216 | __u32 size; /* extent size, in sectors */ | |
217 | struct bversion version; | |
218 | ||
219 | __u8 pad[1]; | |
220 | #endif | |
fd0c7679 | 221 | } __packed __aligned(8); |
1c6fdbd8 KO |
222 | |
223 | struct bkey_packed { | |
224 | __u64 _data[0]; | |
225 | ||
226 | /* Size of combined key and value, in u64s */ | |
227 | __u8 u64s; | |
228 | ||
229 | /* Format of key (0 for format local to btree node) */ | |
230 | ||
231 | /* | |
232 | * XXX: next incompat on disk format change, switch format and | |
233 | * needs_whiteout - bkey_packed() will be cheaper if format is the high | |
234 | * bits of the bitfield | |
235 | */ | |
236 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
237 | __u8 format:7, | |
238 | needs_whiteout:1; | |
239 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
240 | __u8 needs_whiteout:1, | |
241 | format:7; | |
242 | #endif | |
243 | ||
244 | /* Type of the value */ | |
245 | __u8 type; | |
246 | __u8 key_start[0]; | |
247 | ||
248 | /* | |
249 | * We copy bkeys with struct assignment in various places, and while | |
250 | * that shouldn't be done with packed bkeys we can't disallow it in C, | |
251 | * and it's legal to cast a bkey to a bkey_packed - so padding it out | |
252 | * to the same size as struct bkey should hopefully be safest. | |
253 | */ | |
254 | __u8 pad[sizeof(struct bkey) - 3]; | |
fd0c7679 | 255 | } __packed __aligned(8); |
1c6fdbd8 | 256 | |
653693be KO |
257 | typedef struct { |
258 | __le64 lo; | |
259 | __le64 hi; | |
260 | } bch_le128; | |
261 | ||
1c6fdbd8 | 262 | #define BKEY_U64s (sizeof(struct bkey) / sizeof(__u64)) |
cd575ddf KO |
263 | #define BKEY_U64s_MAX U8_MAX |
264 | #define BKEY_VAL_U64s_MAX (BKEY_U64s_MAX - BKEY_U64s) | |
265 | ||
1c6fdbd8 KO |
266 | #define KEY_PACKED_BITS_START 24 |
267 | ||
268 | #define KEY_FORMAT_LOCAL_BTREE 0 | |
269 | #define KEY_FORMAT_CURRENT 1 | |
270 | ||
271 | enum bch_bkey_fields { | |
272 | BKEY_FIELD_INODE, | |
273 | BKEY_FIELD_OFFSET, | |
274 | BKEY_FIELD_SNAPSHOT, | |
275 | BKEY_FIELD_SIZE, | |
276 | BKEY_FIELD_VERSION_HI, | |
277 | BKEY_FIELD_VERSION_LO, | |
278 | BKEY_NR_FIELDS, | |
279 | }; | |
280 | ||
281 | #define bkey_format_field(name, field) \ | |
282 | [BKEY_FIELD_##name] = (sizeof(((struct bkey *) NULL)->field) * 8) | |
283 | ||
284 | #define BKEY_FORMAT_CURRENT \ | |
285 | ((struct bkey_format) { \ | |
286 | .key_u64s = BKEY_U64s, \ | |
287 | .nr_fields = BKEY_NR_FIELDS, \ | |
288 | .bits_per_field = { \ | |
289 | bkey_format_field(INODE, p.inode), \ | |
290 | bkey_format_field(OFFSET, p.offset), \ | |
291 | bkey_format_field(SNAPSHOT, p.snapshot), \ | |
292 | bkey_format_field(SIZE, size), \ | |
293 | bkey_format_field(VERSION_HI, version.hi), \ | |
294 | bkey_format_field(VERSION_LO, version.lo), \ | |
295 | }, \ | |
296 | }) | |
297 | ||
298 | /* bkey with inline value */ | |
299 | struct bkey_i { | |
300 | __u64 _data[0]; | |
301 | ||
ac2ccddc KO |
302 | struct bkey k; |
303 | struct bch_val v; | |
1c6fdbd8 KO |
304 | }; |
305 | ||
306 | #define KEY(_inode, _offset, _size) \ | |
307 | ((struct bkey) { \ | |
308 | .u64s = BKEY_U64s, \ | |
309 | .format = KEY_FORMAT_CURRENT, \ | |
310 | .p = POS(_inode, _offset), \ | |
311 | .size = _size, \ | |
312 | }) | |
313 | ||
314 | static inline void bkey_init(struct bkey *k) | |
315 | { | |
316 | *k = KEY(0, 0, 0); | |
317 | } | |
318 | ||
319 | #define bkey_bytes(_k) ((_k)->u64s * sizeof(__u64)) | |
320 | ||
321 | #define __BKEY_PADDED(key, pad) \ | |
45dd05b3 | 322 | struct bkey_i key; __u64 key ## _pad[pad] |
1c6fdbd8 | 323 | |
1c6fdbd8 KO |
324 | /* |
325 | * - DELETED keys are used internally to mark keys that should be ignored but | |
326 | * override keys in composition order. Their version number is ignored. | |
327 | * | |
328 | * - DISCARDED keys indicate that the data is all 0s because it has been | |
329 | * discarded. DISCARDs may have a version; if the version is nonzero the key | |
330 | * will be persistent, otherwise the key will be dropped whenever the btree | |
331 | * node is rewritten (like DELETED keys). | |
332 | * | |
333 | * - ERROR: any read of the data returns a read error, as the data was lost due | |
334 | * to a failing device. Like DISCARDED keys, they can be removed (overridden) | |
335 | * by new writes or cluster-wide GC. Node repair can also overwrite them with | |
336 | * the same or a more recent version number, but not with an older version | |
337 | * number. | |
26609b61 KO |
338 | * |
339 | * - WHITEOUT: for hash table btrees | |
3e3e02e6 | 340 | */ |
26609b61 KO |
341 | #define BCH_BKEY_TYPES() \ |
342 | x(deleted, 0) \ | |
7a7d17b2 | 343 | x(whiteout, 1) \ |
26609b61 KO |
344 | x(error, 2) \ |
345 | x(cookie, 3) \ | |
79f88eba | 346 | x(hash_whiteout, 4) \ |
26609b61 KO |
347 | x(btree_ptr, 5) \ |
348 | x(extent, 6) \ | |
349 | x(reservation, 7) \ | |
350 | x(inode, 8) \ | |
351 | x(inode_generation, 9) \ | |
352 | x(dirent, 10) \ | |
353 | x(xattr, 11) \ | |
354 | x(alloc, 12) \ | |
355 | x(quota, 13) \ | |
76426098 KO |
356 | x(stripe, 14) \ |
357 | x(reflink_p, 15) \ | |
4be1a412 | 358 | x(reflink_v, 16) \ |
548b3d20 | 359 | x(inline_data, 17) \ |
801a3de6 | 360 | x(btree_ptr_v2, 18) \ |
7f4e1d5d | 361 | x(indirect_inline_data, 19) \ |
14b393ee KO |
362 | x(alloc_v2, 20) \ |
363 | x(subvolume, 21) \ | |
3e52c222 KO |
364 | x(snapshot, 22) \ |
365 | x(inode_v2, 23) \ | |
179e3434 | 366 | x(alloc_v3, 24) \ |
d326ab2f | 367 | x(set, 25) \ |
3d48a7f8 | 368 | x(lru, 26) \ |
a8c752bb | 369 | x(alloc_v4, 27) \ |
8dd69d9f | 370 | x(backpointer, 28) \ |
5250b74d | 371 | x(inode_v3, 29) \ |
1c59b483 KO |
372 | x(bucket_gens, 30) \ |
373 | x(snapshot_tree, 31) | |
26609b61 KO |
374 | |
375 | enum bch_bkey_type { | |
376 | #define x(name, nr) KEY_TYPE_##name = nr, | |
377 | BCH_BKEY_TYPES() | |
378 | #undef x | |
379 | KEY_TYPE_MAX, | |
380 | }; | |
1c6fdbd8 | 381 | |
79f88eba KO |
382 | struct bch_deleted { |
383 | struct bch_val v; | |
384 | }; | |
385 | ||
7a7d17b2 | 386 | struct bch_whiteout { |
79f88eba KO |
387 | struct bch_val v; |
388 | }; | |
389 | ||
390 | struct bch_error { | |
391 | struct bch_val v; | |
392 | }; | |
393 | ||
1c6fdbd8 KO |
394 | struct bch_cookie { |
395 | struct bch_val v; | |
396 | __le64 cookie; | |
397 | }; | |
1c6fdbd8 | 398 | |
79f88eba KO |
399 | struct bch_hash_whiteout { |
400 | struct bch_val v; | |
401 | }; | |
402 | ||
179e3434 KO |
403 | struct bch_set { |
404 | struct bch_val v; | |
405 | }; | |
406 | ||
1c6fdbd8 KO |
407 | /* Extents */ |
408 | ||
409 | /* | |
410 | * In extent bkeys, the value is a list of pointers (bch_extent_ptr), optionally | |
411 | * preceded by checksum/compression information (bch_extent_crc32 or | |
412 | * bch_extent_crc64). | |
413 | * | |
414 | * One major determining factor in the format of extents is how we handle and | |
415 | * represent extents that have been partially overwritten and thus trimmed: | |
416 | * | |
417 | * If an extent is not checksummed or compressed, when the extent is trimmed we | |
418 | * don't have to remember the extent we originally allocated and wrote: we can | |
419 | * merely adjust ptr->offset to point to the start of the data that is currently | |
420 | * live. The size field in struct bkey records the current (live) size of the | |
421 | * extent, and is also used to mean "size of region on disk that we point to" in | |
422 | * this case. | |
423 | * | |
424 | * Thus an extent that is not checksummed or compressed will consist only of a | |
425 | * list of bch_extent_ptrs, with none of the fields in | |
426 | * bch_extent_crc32/bch_extent_crc64. | |
427 | * | |
428 | * When an extent is checksummed or compressed, it's not possible to read only | |
429 | * the data that is currently live: we have to read the entire extent that was | |
430 | * originally written, and then return only the part of the extent that is | |
431 | * currently live. | |
432 | * | |
433 | * Thus, in addition to the current size of the extent in struct bkey, we need | |
434 | * to store the size of the originally allocated space - this is the | |
435 | * compressed_size and uncompressed_size fields in bch_extent_crc32/64. Also, | |
436 | * when the extent is trimmed, instead of modifying the offset field of the | |
437 | * pointer, we keep a second smaller offset field - "offset into the original | |
438 | * extent of the currently live region". | |
439 | * | |
440 | * The other major determining factor is replication and data migration: | |
441 | * | |
442 | * Each pointer may have its own bch_extent_crc32/64. When doing a replicated | |
443 | * write, we will initially write all the replicas in the same format, with the | |
444 | * same checksum type and compression format - however, when copygc runs later (or | |
445 | * tiering/cache promotion, anything that moves data), it is not in general | |
446 | * going to rewrite all the pointers at once - one of the replicas may be in a | |
447 | * bucket on one device that has very little fragmentation while another lives | |
448 | * in a bucket that has become heavily fragmented, and thus is being rewritten | |
449 | * sooner than the rest. | |
450 | * | |
451 | * Thus it will only move a subset of the pointers (or in the case of | |
452 | * tiering/cache promotion perhaps add a single pointer without dropping any | |
453 | * current pointers), and if the extent has been partially overwritten it must | |
454 | * write only the currently live portion (or copygc would not be able to reduce | |
455 | * fragmentation!) - which necessitates a different bch_extent_crc format for | |
456 | * the new pointer. | |
457 | * | |
458 | * But in the interests of space efficiency, we don't want to store one | |
459 | * bch_extent_crc for each pointer if we don't have to. | |
460 | * | |
461 | * Thus, a bch_extent consists of bch_extent_crc32s, bch_extent_crc64s, and | |
462 | * bch_extent_ptrs appended arbitrarily one after the other. We determine the | |
463 | * type of a given entry with a scheme similar to utf8 (except we're encoding a | |
464 | * type, not a size), encoding the type in the position of the first set bit: | |
465 | * | |
466 | * bch_extent_crc32 - 0b1 | |
467 | * bch_extent_ptr - 0b10 | |
468 | * bch_extent_crc64 - 0b100 | |
469 | * | |
470 | * We do it this way because bch_extent_crc32 is _very_ constrained on bits (and | |
471 | * bch_extent_crc64 is the least constrained). | |
472 | * | |
473 | * Then, each bch_extent_crc32/64 applies to the pointers that follow after it, | |
474 | * until the next bch_extent_crc32/64. | |
475 | * | |
476 | * If there are no bch_extent_crcs preceding a bch_extent_ptr, then that pointer | |
477 | * is neither checksummed nor compressed. | |
478 | */ | |
479 | ||
480 | /* 128 bits, sufficient for cryptographic MACs: */ | |
481 | struct bch_csum { | |
482 | __le64 lo; | |
483 | __le64 hi; | |
fd0c7679 | 484 | } __packed __aligned(8); |
1c6fdbd8 | 485 | |
abce30b7 KO |
486 | #define BCH_EXTENT_ENTRY_TYPES() \ |
487 | x(ptr, 0) \ | |
488 | x(crc32, 1) \ | |
489 | x(crc64, 2) \ | |
cd575ddf | 490 | x(crc128, 3) \ |
2766876d KO |
491 | x(stripe_ptr, 4) \ |
492 | x(rebalance, 5) | |
493 | #define BCH_EXTENT_ENTRY_MAX 6 | |
abce30b7 | 494 | |
1c6fdbd8 | 495 | enum bch_extent_entry_type { |
abce30b7 KO |
496 | #define x(f, n) BCH_EXTENT_ENTRY_##f = n, |
497 | BCH_EXTENT_ENTRY_TYPES() | |
498 | #undef x | |
1c6fdbd8 KO |
499 | }; |
500 | ||
1c6fdbd8 KO |
501 | /* Compressed/uncompressed size are stored biased by 1: */ |
502 | struct bch_extent_crc32 { | |
503 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
504 | __u32 type:2, | |
505 | _compressed_size:7, | |
506 | _uncompressed_size:7, | |
507 | offset:7, | |
508 | _unused:1, | |
509 | csum_type:4, | |
510 | compression_type:4; | |
511 | __u32 csum; | |
512 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
513 | __u32 csum; | |
514 | __u32 compression_type:4, | |
515 | csum_type:4, | |
516 | _unused:1, | |
517 | offset:7, | |
518 | _uncompressed_size:7, | |
519 | _compressed_size:7, | |
520 | type:2; | |
521 | #endif | |
fd0c7679 | 522 | } __packed __aligned(8); |
1c6fdbd8 KO |
523 | |
524 | #define CRC32_SIZE_MAX (1U << 7) | |
525 | #define CRC32_NONCE_MAX 0 | |
526 | ||
527 | struct bch_extent_crc64 { | |
528 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
529 | __u64 type:3, | |
530 | _compressed_size:9, | |
531 | _uncompressed_size:9, | |
532 | offset:9, | |
533 | nonce:10, | |
534 | csum_type:4, | |
535 | compression_type:4, | |
536 | csum_hi:16; | |
537 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
538 | __u64 csum_hi:16, | |
539 | compression_type:4, | |
540 | csum_type:4, | |
541 | nonce:10, | |
542 | offset:9, | |
543 | _uncompressed_size:9, | |
544 | _compressed_size:9, | |
545 | type:3; | |
546 | #endif | |
547 | __u64 csum_lo; | |
fd0c7679 | 548 | } __packed __aligned(8); |
1c6fdbd8 KO |
549 | |
550 | #define CRC64_SIZE_MAX (1U << 9) | |
551 | #define CRC64_NONCE_MAX ((1U << 10) - 1) | |
552 | ||
553 | struct bch_extent_crc128 { | |
554 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
555 | __u64 type:4, | |
556 | _compressed_size:13, | |
557 | _uncompressed_size:13, | |
558 | offset:13, | |
559 | nonce:13, | |
560 | csum_type:4, | |
561 | compression_type:4; | |
562 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
563 | __u64 compression_type:4, | |
564 | csum_type:4, | |
565 | nonce:13, | |
566 | offset:13, | |
567 | _uncompressed_size:13, | |
568 | _compressed_size:13, | |
569 | type:4; | |
570 | #endif | |
571 | struct bch_csum csum; | |
fd0c7679 | 572 | } __packed __aligned(8); |
1c6fdbd8 KO |
573 | |
574 | #define CRC128_SIZE_MAX (1U << 13) | |
575 | #define CRC128_NONCE_MAX ((1U << 13) - 1) | |
576 | ||
577 | /* | |
578 | * @reservation - pointer hasn't been written to, just reserved | |
579 | */ | |
580 | struct bch_extent_ptr { | |
581 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
582 | __u64 type:1, | |
583 | cached:1, | |
cd575ddf | 584 | unused:1, |
79203111 | 585 | unwritten:1, |
1c6fdbd8 KO |
586 | offset:44, /* 8 petabytes */ |
587 | dev:8, | |
588 | gen:8; | |
589 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
590 | __u64 gen:8, | |
591 | dev:8, | |
592 | offset:44, | |
79203111 | 593 | unwritten:1, |
cd575ddf | 594 | unused:1, |
1c6fdbd8 KO |
595 | cached:1, |
596 | type:1; | |
597 | #endif | |
fd0c7679 | 598 | } __packed __aligned(8); |
1c6fdbd8 | 599 | |
cd575ddf | 600 | struct bch_extent_stripe_ptr { |
1c6fdbd8 KO |
601 | #if defined(__LITTLE_ENDIAN_BITFIELD) |
602 | __u64 type:5, | |
cd575ddf | 603 | block:8, |
7f4e1d5d KO |
604 | redundancy:4, |
605 | idx:47; | |
cd575ddf | 606 | #elif defined (__BIG_ENDIAN_BITFIELD) |
7f4e1d5d KO |
607 | __u64 idx:47, |
608 | redundancy:4, | |
cd575ddf KO |
609 | block:8, |
610 | type:5; | |
611 | #endif | |
612 | }; | |
613 | ||
614 | struct bch_extent_reservation { | |
615 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
616 | __u64 type:6, | |
617 | unused:22, | |
1c6fdbd8 KO |
618 | replicas:4, |
619 | generation:32; | |
620 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
621 | __u64 generation:32, | |
622 | replicas:4, | |
cd575ddf KO |
623 | unused:22, |
624 | type:6; | |
1c6fdbd8 KO |
625 | #endif |
626 | }; | |
627 | ||
2766876d KO |
628 | struct bch_extent_rebalance { |
629 | #if defined(__LITTLE_ENDIAN_BITFIELD) | |
630 | __u64 type:7, | |
631 | unused:33, | |
632 | compression:8, | |
633 | target:16; | |
634 | #elif defined (__BIG_ENDIAN_BITFIELD) | |
635 | __u64 target:16, | |
636 | compression:8, | |
637 | unused:33, | |
638 | type:7; | |
639 | #endif | |
640 | }; | |
641 | ||
1c6fdbd8 KO |
642 | union bch_extent_entry { |
643 | #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ || __BITS_PER_LONG == 64 | |
644 | unsigned long type; | |
645 | #elif __BITS_PER_LONG == 32 | |
646 | struct { | |
647 | unsigned long pad; | |
648 | unsigned long type; | |
649 | }; | |
650 | #else | |
651 | #error edit for your odd byteorder. | |
652 | #endif | |
abce30b7 KO |
653 | |
654 | #define x(f, n) struct bch_extent_##f f; | |
655 | BCH_EXTENT_ENTRY_TYPES() | |
656 | #undef x | |
1c6fdbd8 KO |
657 | }; |
658 | ||
26609b61 KO |
659 | struct bch_btree_ptr { |
660 | struct bch_val v; | |
1c6fdbd8 | 661 | |
26609b61 KO |
662 | __u64 _data[0]; |
663 | struct bch_extent_ptr start[]; | |
fd0c7679 | 664 | } __packed __aligned(8); |
1c6fdbd8 | 665 | |
548b3d20 KO |
666 | struct bch_btree_ptr_v2 { |
667 | struct bch_val v; | |
668 | ||
669 | __u64 mem_ptr; | |
670 | __le64 seq; | |
671 | __le16 sectors_written; | |
51d2dfb8 | 672 | __le16 flags; |
548b3d20 KO |
673 | struct bpos min_key; |
674 | __u64 _data[0]; | |
675 | struct bch_extent_ptr start[]; | |
fd0c7679 | 676 | } __packed __aligned(8); |
548b3d20 | 677 | |
51d2dfb8 KO |
678 | LE16_BITMASK(BTREE_PTR_RANGE_UPDATED, struct bch_btree_ptr_v2, flags, 0, 1); |
679 | ||
1c6fdbd8 KO |
680 | struct bch_extent { |
681 | struct bch_val v; | |
682 | ||
683 | __u64 _data[0]; | |
684 | union bch_extent_entry start[]; | |
fd0c7679 | 685 | } __packed __aligned(8); |
1c6fdbd8 KO |
686 | |
687 | struct bch_reservation { | |
688 | struct bch_val v; | |
689 | ||
690 | __le32 generation; | |
691 | __u8 nr_replicas; | |
692 | __u8 pad[3]; | |
fd0c7679 | 693 | } __packed __aligned(8); |
1c6fdbd8 KO |
694 | |
695 | /* Maximum size (in u64s) a single pointer could be: */ | |
696 | #define BKEY_EXTENT_PTR_U64s_MAX\ | |
697 | ((sizeof(struct bch_extent_crc128) + \ | |
698 | sizeof(struct bch_extent_ptr)) / sizeof(u64)) | |
699 | ||
700 | /* Maximum possible size of an entire extent value: */ | |
701 | #define BKEY_EXTENT_VAL_U64s_MAX \ | |
5055b509 | 702 | (1 + BKEY_EXTENT_PTR_U64s_MAX * (BCH_REPLICAS_MAX + 1)) |
1c6fdbd8 | 703 | |
1c6fdbd8 KO |
704 | /* * Maximum possible size of an entire extent, key + value: */ |
705 | #define BKEY_EXTENT_U64s_MAX (BKEY_U64s + BKEY_EXTENT_VAL_U64s_MAX) | |
706 | ||
707 | /* Btree pointers don't carry around checksums: */ | |
708 | #define BKEY_BTREE_PTR_VAL_U64s_MAX \ | |
548b3d20 KO |
709 | ((sizeof(struct bch_btree_ptr_v2) + \ |
710 | sizeof(struct bch_extent_ptr) * BCH_REPLICAS_MAX) / sizeof(u64)) | |
1c6fdbd8 KO |
711 | #define BKEY_BTREE_PTR_U64s_MAX \ |
712 | (BKEY_U64s + BKEY_BTREE_PTR_VAL_U64s_MAX) | |
713 | ||
714 | /* Inodes */ | |
715 | ||
716 | #define BLOCKDEV_INODE_MAX 4096 | |
717 | ||
718 | #define BCACHEFS_ROOT_INO 4096 | |
719 | ||
1c6fdbd8 KO |
720 | struct bch_inode { |
721 | struct bch_val v; | |
722 | ||
723 | __le64 bi_hash_seed; | |
724 | __le32 bi_flags; | |
725 | __le16 bi_mode; | |
726 | __u8 fields[0]; | |
fd0c7679 | 727 | } __packed __aligned(8); |
1c6fdbd8 | 728 | |
3e52c222 KO |
729 | struct bch_inode_v2 { |
730 | struct bch_val v; | |
731 | ||
732 | __le64 bi_journal_seq; | |
733 | __le64 bi_hash_seed; | |
734 | __le64 bi_flags; | |
735 | __le16 bi_mode; | |
736 | __u8 fields[0]; | |
fd0c7679 | 737 | } __packed __aligned(8); |
3e52c222 | 738 | |
8dd69d9f KO |
739 | struct bch_inode_v3 { |
740 | struct bch_val v; | |
741 | ||
742 | __le64 bi_journal_seq; | |
743 | __le64 bi_hash_seed; | |
744 | __le64 bi_flags; | |
745 | __le64 bi_sectors; | |
746 | __le64 bi_size; | |
747 | __le64 bi_version; | |
748 | __u8 fields[0]; | |
749 | } __packed __aligned(8); | |
750 | ||
751 | #define INODEv3_FIELDS_START_INITIAL 6 | |
752 | #define INODEv3_FIELDS_START_CUR (offsetof(struct bch_inode_v3, fields) / sizeof(u64)) | |
753 | ||
1c6fdbd8 KO |
754 | struct bch_inode_generation { |
755 | struct bch_val v; | |
756 | ||
757 | __le32 bi_generation; | |
758 | __le32 pad; | |
fd0c7679 | 759 | } __packed __aligned(8); |
1c6fdbd8 | 760 | |
14b393ee KO |
761 | /* |
762 | * bi_subvol and bi_parent_subvol are only set for subvolume roots: | |
763 | */ | |
764 | ||
8dd69d9f | 765 | #define BCH_INODE_FIELDS_v2() \ |
a3e72262 KO |
766 | x(bi_atime, 96) \ |
767 | x(bi_ctime, 96) \ | |
768 | x(bi_mtime, 96) \ | |
769 | x(bi_otime, 96) \ | |
a3e70fb2 KO |
770 | x(bi_size, 64) \ |
771 | x(bi_sectors, 64) \ | |
772 | x(bi_uid, 32) \ | |
773 | x(bi_gid, 32) \ | |
774 | x(bi_nlink, 32) \ | |
775 | x(bi_generation, 32) \ | |
776 | x(bi_dev, 32) \ | |
777 | x(bi_data_checksum, 8) \ | |
778 | x(bi_compression, 8) \ | |
779 | x(bi_project, 32) \ | |
780 | x(bi_background_compression, 8) \ | |
781 | x(bi_data_replicas, 8) \ | |
782 | x(bi_promote_target, 16) \ | |
783 | x(bi_foreground_target, 16) \ | |
784 | x(bi_background_target, 16) \ | |
721d4ad8 | 785 | x(bi_erasure_code, 16) \ |
ab2a29cc KO |
786 | x(bi_fields_set, 16) \ |
787 | x(bi_dir, 64) \ | |
14b393ee KO |
788 | x(bi_dir_offset, 64) \ |
789 | x(bi_subvol, 32) \ | |
790 | x(bi_parent_subvol, 32) | |
a3e70fb2 | 791 | |
8dd69d9f KO |
792 | #define BCH_INODE_FIELDS_v3() \ |
793 | x(bi_atime, 96) \ | |
794 | x(bi_ctime, 96) \ | |
795 | x(bi_mtime, 96) \ | |
796 | x(bi_otime, 96) \ | |
797 | x(bi_uid, 32) \ | |
798 | x(bi_gid, 32) \ | |
799 | x(bi_nlink, 32) \ | |
800 | x(bi_generation, 32) \ | |
801 | x(bi_dev, 32) \ | |
802 | x(bi_data_checksum, 8) \ | |
803 | x(bi_compression, 8) \ | |
804 | x(bi_project, 32) \ | |
805 | x(bi_background_compression, 8) \ | |
806 | x(bi_data_replicas, 8) \ | |
807 | x(bi_promote_target, 16) \ | |
808 | x(bi_foreground_target, 16) \ | |
809 | x(bi_background_target, 16) \ | |
810 | x(bi_erasure_code, 16) \ | |
811 | x(bi_fields_set, 16) \ | |
812 | x(bi_dir, 64) \ | |
813 | x(bi_dir_offset, 64) \ | |
814 | x(bi_subvol, 32) \ | |
a8b3a677 KO |
815 | x(bi_parent_subvol, 32) \ |
816 | x(bi_nocow, 8) | |
8dd69d9f | 817 | |
d42dd4ad KO |
818 | /* subset of BCH_INODE_FIELDS */ |
819 | #define BCH_INODE_OPTS() \ | |
820 | x(data_checksum, 8) \ | |
821 | x(compression, 8) \ | |
822 | x(project, 32) \ | |
823 | x(background_compression, 8) \ | |
824 | x(data_replicas, 8) \ | |
825 | x(promote_target, 16) \ | |
826 | x(foreground_target, 16) \ | |
827 | x(background_target, 16) \ | |
a8b3a677 KO |
828 | x(erasure_code, 16) \ |
829 | x(nocow, 8) | |
1c6fdbd8 | 830 | |
721d4ad8 KO |
831 | enum inode_opt_id { |
832 | #define x(name, ...) \ | |
833 | Inode_opt_##name, | |
834 | BCH_INODE_OPTS() | |
835 | #undef x | |
836 | Inode_opt_nr, | |
837 | }; | |
838 | ||
1c6fdbd8 KO |
839 | enum { |
840 | /* | |
841 | * User flags (get/settable with FS_IOC_*FLAGS, correspond to FS_*_FL | |
842 | * flags) | |
843 | */ | |
3e3e02e6 KO |
844 | __BCH_INODE_SYNC = 0, |
845 | __BCH_INODE_IMMUTABLE = 1, | |
846 | __BCH_INODE_APPEND = 2, | |
847 | __BCH_INODE_NODUMP = 3, | |
848 | __BCH_INODE_NOATIME = 4, | |
849 | ||
850 | __BCH_INODE_I_SIZE_DIRTY = 5, | |
851 | __BCH_INODE_I_SECTORS_DIRTY = 6, | |
852 | __BCH_INODE_UNLINKED = 7, | |
853 | __BCH_INODE_BACKPTR_UNTRUSTED = 8, | |
1c6fdbd8 KO |
854 | |
855 | /* bits 20+ reserved for packed fields below: */ | |
856 | }; | |
857 | ||
858 | #define BCH_INODE_SYNC (1 << __BCH_INODE_SYNC) | |
859 | #define BCH_INODE_IMMUTABLE (1 << __BCH_INODE_IMMUTABLE) | |
860 | #define BCH_INODE_APPEND (1 << __BCH_INODE_APPEND) | |
861 | #define BCH_INODE_NODUMP (1 << __BCH_INODE_NODUMP) | |
862 | #define BCH_INODE_NOATIME (1 << __BCH_INODE_NOATIME) | |
863 | #define BCH_INODE_I_SIZE_DIRTY (1 << __BCH_INODE_I_SIZE_DIRTY) | |
864 | #define BCH_INODE_I_SECTORS_DIRTY (1 << __BCH_INODE_I_SECTORS_DIRTY) | |
865 | #define BCH_INODE_UNLINKED (1 << __BCH_INODE_UNLINKED) | |
ab2a29cc | 866 | #define BCH_INODE_BACKPTR_UNTRUSTED (1 << __BCH_INODE_BACKPTR_UNTRUSTED) |
1c6fdbd8 KO |
867 | |
868 | LE32_BITMASK(INODE_STR_HASH, struct bch_inode, bi_flags, 20, 24); | |
a3e72262 KO |
869 | LE32_BITMASK(INODE_NR_FIELDS, struct bch_inode, bi_flags, 24, 31); |
870 | LE32_BITMASK(INODE_NEW_VARINT, struct bch_inode, bi_flags, 31, 32); | |
1c6fdbd8 | 871 | |
3e52c222 KO |
872 | LE64_BITMASK(INODEv2_STR_HASH, struct bch_inode_v2, bi_flags, 20, 24); |
873 | LE64_BITMASK(INODEv2_NR_FIELDS, struct bch_inode_v2, bi_flags, 24, 31); | |
874 | ||
8dd69d9f KO |
875 | LE64_BITMASK(INODEv3_STR_HASH, struct bch_inode_v3, bi_flags, 20, 24); |
876 | LE64_BITMASK(INODEv3_NR_FIELDS, struct bch_inode_v3, bi_flags, 24, 31); | |
877 | ||
878 | LE64_BITMASK(INODEv3_FIELDS_START, | |
879 | struct bch_inode_v3, bi_flags, 31, 36); | |
880 | LE64_BITMASK(INODEv3_MODE, struct bch_inode_v3, bi_flags, 36, 52); | |
881 | ||
1c6fdbd8 KO |
882 | /* Dirents */ |
883 | ||
884 | /* | |
885 | * Dirents (and xattrs) have to implement string lookups; since our b-tree | |
886 | * doesn't support arbitrary length strings for the key, we instead index by a | |
887 | * 64 bit hash (currently truncated sha1) of the string, stored in the offset | |
888 | * field of the key - using linear probing to resolve hash collisions. This also | |
889 | * provides us with the readdir cookie posix requires. | |
890 | * | |
891 | * Linear probing requires us to use whiteouts for deletions, in the event of a | |
892 | * collision: | |
893 | */ | |
894 | ||
1c6fdbd8 KO |
895 | struct bch_dirent { |
896 | struct bch_val v; | |
897 | ||
898 | /* Target inode number: */ | |
4db65027 | 899 | union { |
1c6fdbd8 | 900 | __le64 d_inum; |
4db65027 KO |
901 | struct { /* DT_SUBVOL */ |
902 | __le32 d_child_subvol; | |
903 | __le32 d_parent_subvol; | |
904 | }; | |
905 | }; | |
1c6fdbd8 KO |
906 | |
907 | /* | |
908 | * Copy of mode bits 12-15 from the target inode - so userspace can get | |
909 | * the filetype without having to do a stat() | |
910 | */ | |
911 | __u8 d_type; | |
912 | ||
913 | __u8 d_name[]; | |
fd0c7679 | 914 | } __packed __aligned(8); |
1c6fdbd8 | 915 | |
14b393ee KO |
916 | #define DT_SUBVOL 16 |
917 | #define BCH_DT_MAX 17 | |
918 | ||
502f973d | 919 | #define BCH_NAME_MAX ((unsigned) (U8_MAX * sizeof(u64) - \ |
1c6fdbd8 | 920 | sizeof(struct bkey) - \ |
502f973d | 921 | offsetof(struct bch_dirent, d_name))) |
1c6fdbd8 KO |
922 | |
923 | /* Xattrs */ | |
924 | ||
26609b61 KO |
925 | #define KEY_TYPE_XATTR_INDEX_USER 0 |
926 | #define KEY_TYPE_XATTR_INDEX_POSIX_ACL_ACCESS 1 | |
927 | #define KEY_TYPE_XATTR_INDEX_POSIX_ACL_DEFAULT 2 | |
928 | #define KEY_TYPE_XATTR_INDEX_TRUSTED 3 | |
929 | #define KEY_TYPE_XATTR_INDEX_SECURITY 4 | |
1c6fdbd8 KO |
930 | |
931 | struct bch_xattr { | |
932 | struct bch_val v; | |
933 | __u8 x_type; | |
934 | __u8 x_name_len; | |
935 | __le16 x_val_len; | |
936 | __u8 x_name[]; | |
fd0c7679 | 937 | } __packed __aligned(8); |
1c6fdbd8 KO |
938 | |
939 | /* Bucket/allocation information: */ | |
940 | ||
1c6fdbd8 KO |
941 | struct bch_alloc { |
942 | struct bch_val v; | |
943 | __u8 fields; | |
944 | __u8 gen; | |
945 | __u8 data[]; | |
fd0c7679 | 946 | } __packed __aligned(8); |
1c6fdbd8 | 947 | |
7f4e1d5d | 948 | #define BCH_ALLOC_FIELDS_V1() \ |
8fe826f9 KO |
949 | x(read_time, 16) \ |
950 | x(write_time, 16) \ | |
951 | x(data_type, 8) \ | |
952 | x(dirty_sectors, 16) \ | |
953 | x(cached_sectors, 16) \ | |
7f4e1d5d KO |
954 | x(oldest_gen, 8) \ |
955 | x(stripe, 32) \ | |
956 | x(stripe_redundancy, 8) | |
957 | ||
a8c752bb KO |
958 | enum { |
959 | #define x(name, _bits) BCH_ALLOC_FIELD_V1_##name, | |
960 | BCH_ALLOC_FIELDS_V1() | |
961 | #undef x | |
962 | }; | |
963 | ||
7f4e1d5d KO |
964 | struct bch_alloc_v2 { |
965 | struct bch_val v; | |
966 | __u8 nr_fields; | |
967 | __u8 gen; | |
968 | __u8 oldest_gen; | |
969 | __u8 data_type; | |
970 | __u8 data[]; | |
fd0c7679 | 971 | } __packed __aligned(8); |
7f4e1d5d KO |
972 | |
973 | #define BCH_ALLOC_FIELDS_V2() \ | |
974 | x(read_time, 64) \ | |
975 | x(write_time, 64) \ | |
66d90823 KO |
976 | x(dirty_sectors, 32) \ |
977 | x(cached_sectors, 32) \ | |
7f4e1d5d KO |
978 | x(stripe, 32) \ |
979 | x(stripe_redundancy, 8) | |
90541a74 | 980 | |
3e52c222 KO |
981 | struct bch_alloc_v3 { |
982 | struct bch_val v; | |
983 | __le64 journal_seq; | |
984 | __le32 flags; | |
985 | __u8 nr_fields; | |
986 | __u8 gen; | |
987 | __u8 oldest_gen; | |
988 | __u8 data_type; | |
989 | __u8 data[]; | |
fd0c7679 | 990 | } __packed __aligned(8); |
3e52c222 | 991 | |
a8c752bb KO |
992 | LE32_BITMASK(BCH_ALLOC_V3_NEED_DISCARD,struct bch_alloc_v3, flags, 0, 1) |
993 | LE32_BITMASK(BCH_ALLOC_V3_NEED_INC_GEN,struct bch_alloc_v3, flags, 1, 2) | |
994 | ||
3d48a7f8 KO |
995 | struct bch_alloc_v4 { |
996 | struct bch_val v; | |
997 | __u64 journal_seq; | |
998 | __u32 flags; | |
999 | __u8 gen; | |
1000 | __u8 oldest_gen; | |
1001 | __u8 data_type; | |
1002 | __u8 stripe_redundancy; | |
1003 | __u32 dirty_sectors; | |
1004 | __u32 cached_sectors; | |
1005 | __u64 io_time[2]; | |
1006 | __u32 stripe; | |
1007 | __u32 nr_external_backpointers; | |
80c33085 | 1008 | __u64 fragmentation_lru; |
fd0c7679 | 1009 | } __packed __aligned(8); |
3d48a7f8 | 1010 | |
19a614d2 KO |
1011 | #define BCH_ALLOC_V4_U64s_V0 6 |
1012 | #define BCH_ALLOC_V4_U64s (sizeof(struct bch_alloc_v4) / sizeof(u64)) | |
1013 | ||
3d48a7f8 KO |
1014 | BITMASK(BCH_ALLOC_V4_NEED_DISCARD, struct bch_alloc_v4, flags, 0, 1) |
1015 | BITMASK(BCH_ALLOC_V4_NEED_INC_GEN, struct bch_alloc_v4, flags, 1, 2) | |
1016 | BITMASK(BCH_ALLOC_V4_BACKPOINTERS_START,struct bch_alloc_v4, flags, 2, 8) | |
1017 | BITMASK(BCH_ALLOC_V4_NR_BACKPOINTERS, struct bch_alloc_v4, flags, 8, 14) | |
1018 | ||
a8c752bb KO |
1019 | #define BCH_ALLOC_V4_NR_BACKPOINTERS_MAX 40 |
1020 | ||
1021 | struct bch_backpointer { | |
1022 | struct bch_val v; | |
1023 | __u8 btree_id; | |
1024 | __u8 level; | |
1025 | __u8 data_type; | |
1026 | __u64 bucket_offset:40; | |
1027 | __u32 bucket_len; | |
1028 | struct bpos pos; | |
1029 | } __packed __aligned(8); | |
90541a74 | 1030 | |
5250b74d KO |
1031 | #define KEY_TYPE_BUCKET_GENS_BITS 8 |
1032 | #define KEY_TYPE_BUCKET_GENS_NR (1U << KEY_TYPE_BUCKET_GENS_BITS) | |
1033 | #define KEY_TYPE_BUCKET_GENS_MASK (KEY_TYPE_BUCKET_GENS_NR - 1) | |
1034 | ||
1035 | struct bch_bucket_gens { | |
1036 | struct bch_val v; | |
1037 | u8 gens[KEY_TYPE_BUCKET_GENS_NR]; | |
1038 | } __packed __aligned(8); | |
1039 | ||
1c6fdbd8 KO |
1040 | /* Quotas: */ |
1041 | ||
1c6fdbd8 KO |
1042 | enum quota_types { |
1043 | QTYP_USR = 0, | |
1044 | QTYP_GRP = 1, | |
1045 | QTYP_PRJ = 2, | |
1046 | QTYP_NR = 3, | |
1047 | }; | |
1048 | ||
1049 | enum quota_counters { | |
1050 | Q_SPC = 0, | |
1051 | Q_INO = 1, | |
1052 | Q_COUNTERS = 2, | |
1053 | }; | |
1054 | ||
1055 | struct bch_quota_counter { | |
1056 | __le64 hardlimit; | |
1057 | __le64 softlimit; | |
1058 | }; | |
1059 | ||
1060 | struct bch_quota { | |
1061 | struct bch_val v; | |
1062 | struct bch_quota_counter c[Q_COUNTERS]; | |
fd0c7679 | 1063 | } __packed __aligned(8); |
1c6fdbd8 | 1064 | |
cd575ddf KO |
1065 | /* Erasure coding */ |
1066 | ||
cd575ddf KO |
1067 | struct bch_stripe { |
1068 | struct bch_val v; | |
1069 | __le16 sectors; | |
1070 | __u8 algorithm; | |
1071 | __u8 nr_blocks; | |
1072 | __u8 nr_redundant; | |
1073 | ||
1074 | __u8 csum_granularity_bits; | |
1075 | __u8 csum_type; | |
1076 | __u8 pad; | |
1077 | ||
81d8599e | 1078 | struct bch_extent_ptr ptrs[]; |
fd0c7679 | 1079 | } __packed __aligned(8); |
cd575ddf | 1080 | |
76426098 KO |
1081 | /* Reflink: */ |
1082 | ||
1083 | struct bch_reflink_p { | |
1084 | struct bch_val v; | |
1085 | __le64 idx; | |
6d76aefe KO |
1086 | /* |
1087 | * A reflink pointer might point to an indirect extent which is then | |
1088 | * later split (by copygc or rebalance). If we only pointed to part of | |
1089 | * the original indirect extent, and then one of the fragments is | |
1090 | * outside the range we point to, we'd leak a refcount: so when creating | |
1091 | * reflink pointers, we need to store pad values to remember the full | |
1092 | * range we were taking a reference on. | |
1093 | */ | |
1094 | __le32 front_pad; | |
1095 | __le32 back_pad; | |
fd0c7679 | 1096 | } __packed __aligned(8); |
76426098 KO |
1097 | |
1098 | struct bch_reflink_v { | |
1099 | struct bch_val v; | |
1100 | __le64 refcount; | |
1101 | union bch_extent_entry start[0]; | |
1102 | __u64 _data[0]; | |
fd0c7679 | 1103 | } __packed __aligned(8); |
76426098 | 1104 | |
801a3de6 KO |
1105 | struct bch_indirect_inline_data { |
1106 | struct bch_val v; | |
1107 | __le64 refcount; | |
1108 | u8 data[0]; | |
1109 | }; | |
1110 | ||
4be1a412 KO |
1111 | /* Inline data */ |
1112 | ||
1113 | struct bch_inline_data { | |
1114 | struct bch_val v; | |
1115 | u8 data[0]; | |
1116 | }; | |
1117 | ||
14b393ee KO |
1118 | /* Subvolumes: */ |
1119 | ||
1120 | #define SUBVOL_POS_MIN POS(0, 1) | |
1121 | #define SUBVOL_POS_MAX POS(0, S32_MAX) | |
1122 | #define BCACHEFS_ROOT_SUBVOL 1 | |
1123 | ||
1124 | struct bch_subvolume { | |
1125 | struct bch_val v; | |
1126 | __le32 flags; | |
1127 | __le32 snapshot; | |
1128 | __le64 inode; | |
653693be KO |
1129 | __le32 parent; |
1130 | __le32 pad; | |
1131 | bch_le128 otime; | |
14b393ee KO |
1132 | }; |
1133 | ||
1134 | LE32_BITMASK(BCH_SUBVOLUME_RO, struct bch_subvolume, flags, 0, 1) | |
1135 | /* | |
1136 | * We need to know whether a subvolume is a snapshot so we can know whether we | |
1137 | * can delete it (or whether it should just be rm -rf'd) | |
1138 | */ | |
1139 | LE32_BITMASK(BCH_SUBVOLUME_SNAP, struct bch_subvolume, flags, 1, 2) | |
2027875b | 1140 | LE32_BITMASK(BCH_SUBVOLUME_UNLINKED, struct bch_subvolume, flags, 2, 3) |
14b393ee KO |
1141 | |
1142 | /* Snapshots */ | |
1143 | ||
1144 | struct bch_snapshot { | |
1145 | struct bch_val v; | |
1146 | __le32 flags; | |
1147 | __le32 parent; | |
1148 | __le32 children[2]; | |
1149 | __le32 subvol; | |
1c59b483 | 1150 | __le32 tree; |
14b393ee KO |
1151 | }; |
1152 | ||
1153 | LE32_BITMASK(BCH_SNAPSHOT_DELETED, struct bch_snapshot, flags, 0, 1) | |
1154 | ||
1155 | /* True if a subvolume points to this snapshot node: */ | |
1156 | LE32_BITMASK(BCH_SNAPSHOT_SUBVOL, struct bch_snapshot, flags, 1, 2) | |
1157 | ||
1c59b483 KO |
1158 | /* |
1159 | * Snapshot trees: | |
1160 | * | |
1161 | * The snapshot_trees btree gives us persistent indentifier for each tree of | |
1162 | * bch_snapshot nodes, and allow us to record and easily find the root/master | |
1163 | * subvolume that other snapshots were created from: | |
1164 | */ | |
1165 | struct bch_snapshot_tree { | |
1166 | struct bch_val v; | |
1167 | __le32 master_subvol; | |
1168 | __le32 root_snapshot; | |
1169 | }; | |
1170 | ||
d326ab2f KO |
1171 | /* LRU btree: */ |
1172 | ||
1173 | struct bch_lru { | |
1174 | struct bch_val v; | |
1175 | __le64 idx; | |
fd0c7679 | 1176 | } __packed __aligned(8); |
d326ab2f KO |
1177 | |
1178 | #define LRU_ID_STRIPES (1U << 16) | |
1179 | ||
1c6fdbd8 KO |
1180 | /* Optional/variable size superblock sections: */ |
1181 | ||
1182 | struct bch_sb_field { | |
1183 | __u64 _data[0]; | |
1184 | __le32 u64s; | |
1185 | __le32 type; | |
1186 | }; | |
1187 | ||
25be2e5d KO |
1188 | #define BCH_SB_FIELDS() \ |
1189 | x(journal, 0) \ | |
1190 | x(members, 1) \ | |
1191 | x(crypt, 2) \ | |
1192 | x(replicas_v0, 3) \ | |
1193 | x(quota, 4) \ | |
1194 | x(disk_groups, 5) \ | |
1195 | x(clean, 6) \ | |
1196 | x(replicas, 7) \ | |
1197 | x(journal_seq_blacklist, 8) \ | |
104c6974 DH |
1198 | x(journal_v2, 9) \ |
1199 | x(counters, 10) | |
1c6fdbd8 KO |
1200 | |
1201 | enum bch_sb_field_type { | |
1202 | #define x(f, nr) BCH_SB_FIELD_##f = nr, | |
1203 | BCH_SB_FIELDS() | |
1204 | #undef x | |
1205 | BCH_SB_FIELD_NR | |
1206 | }; | |
1207 | ||
25be2e5d KO |
1208 | /* |
1209 | * Most superblock fields are replicated in all device's superblocks - a few are | |
1210 | * not: | |
1211 | */ | |
1212 | #define BCH_SINGLE_DEVICE_SB_FIELDS \ | |
1213 | ((1U << BCH_SB_FIELD_journal)| \ | |
1214 | (1U << BCH_SB_FIELD_journal_v2)) | |
1215 | ||
1c6fdbd8 KO |
1216 | /* BCH_SB_FIELD_journal: */ |
1217 | ||
1218 | struct bch_sb_field_journal { | |
1219 | struct bch_sb_field field; | |
1220 | __le64 buckets[0]; | |
1221 | }; | |
1222 | ||
25be2e5d KO |
1223 | struct bch_sb_field_journal_v2 { |
1224 | struct bch_sb_field field; | |
1225 | ||
1226 | struct bch_sb_field_journal_v2_entry { | |
1227 | __le64 start; | |
1228 | __le64 nr; | |
1229 | } d[0]; | |
1230 | }; | |
1231 | ||
1c6fdbd8 KO |
1232 | /* BCH_SB_FIELD_members: */ |
1233 | ||
8b335bae KO |
1234 | #define BCH_MIN_NR_NBUCKETS (1 << 6) |
1235 | ||
1c6fdbd8 KO |
1236 | struct bch_member { |
1237 | __uuid_t uuid; | |
1238 | __le64 nbuckets; /* device size */ | |
1239 | __le16 first_bucket; /* index of first bucket used */ | |
1240 | __le16 bucket_size; /* sectors */ | |
1241 | __le32 pad; | |
1242 | __le64 last_mount; /* time_t */ | |
1243 | ||
1244 | __le64 flags[2]; | |
1245 | }; | |
1246 | ||
1247 | LE64_BITMASK(BCH_MEMBER_STATE, struct bch_member, flags[0], 0, 4) | |
7243498d | 1248 | /* 4-14 unused, was TIER, HAS_(META)DATA, REPLACEMENT */ |
1c6fdbd8 KO |
1249 | LE64_BITMASK(BCH_MEMBER_DISCARD, struct bch_member, flags[0], 14, 15) |
1250 | LE64_BITMASK(BCH_MEMBER_DATA_ALLOWED, struct bch_member, flags[0], 15, 20) | |
1251 | LE64_BITMASK(BCH_MEMBER_GROUP, struct bch_member, flags[0], 20, 28) | |
1252 | LE64_BITMASK(BCH_MEMBER_DURABILITY, struct bch_member, flags[0], 28, 30) | |
c6b2826c KO |
1253 | LE64_BITMASK(BCH_MEMBER_FREESPACE_INITIALIZED, |
1254 | struct bch_member, flags[0], 30, 31) | |
1c6fdbd8 | 1255 | |
1c6fdbd8 KO |
1256 | #if 0 |
1257 | LE64_BITMASK(BCH_MEMBER_NR_READ_ERRORS, struct bch_member, flags[1], 0, 20); | |
1258 | LE64_BITMASK(BCH_MEMBER_NR_WRITE_ERRORS,struct bch_member, flags[1], 20, 40); | |
1259 | #endif | |
1260 | ||
2436cb9f KO |
1261 | #define BCH_MEMBER_STATES() \ |
1262 | x(rw, 0) \ | |
1263 | x(ro, 1) \ | |
1264 | x(failed, 2) \ | |
1265 | x(spare, 3) | |
1266 | ||
1c6fdbd8 | 1267 | enum bch_member_state { |
2436cb9f KO |
1268 | #define x(t, n) BCH_MEMBER_STATE_##t = n, |
1269 | BCH_MEMBER_STATES() | |
1270 | #undef x | |
1271 | BCH_MEMBER_STATE_NR | |
1c6fdbd8 KO |
1272 | }; |
1273 | ||
1c6fdbd8 KO |
1274 | struct bch_sb_field_members { |
1275 | struct bch_sb_field field; | |
1276 | struct bch_member members[0]; | |
1277 | }; | |
1278 | ||
1279 | /* BCH_SB_FIELD_crypt: */ | |
1280 | ||
1281 | struct nonce { | |
1282 | __le32 d[4]; | |
1283 | }; | |
1284 | ||
1285 | struct bch_key { | |
1286 | __le64 key[4]; | |
1287 | }; | |
1288 | ||
1289 | #define BCH_KEY_MAGIC \ | |
1290 | (((u64) 'b' << 0)|((u64) 'c' << 8)| \ | |
1291 | ((u64) 'h' << 16)|((u64) '*' << 24)| \ | |
1292 | ((u64) '*' << 32)|((u64) 'k' << 40)| \ | |
1293 | ((u64) 'e' << 48)|((u64) 'y' << 56)) | |
1294 | ||
1295 | struct bch_encrypted_key { | |
1296 | __le64 magic; | |
1297 | struct bch_key key; | |
1298 | }; | |
1299 | ||
1300 | /* | |
1301 | * If this field is present in the superblock, it stores an encryption key which | |
1302 | * is used encrypt all other data/metadata. The key will normally be encrypted | |
1303 | * with the key userspace provides, but if encryption has been turned off we'll | |
1304 | * just store the master key unencrypted in the superblock so we can access the | |
1305 | * previously encrypted data. | |
1306 | */ | |
1307 | struct bch_sb_field_crypt { | |
1308 | struct bch_sb_field field; | |
1309 | ||
1310 | __le64 flags; | |
1311 | __le64 kdf_flags; | |
1312 | struct bch_encrypted_key key; | |
1313 | }; | |
1314 | ||
1315 | LE64_BITMASK(BCH_CRYPT_KDF_TYPE, struct bch_sb_field_crypt, flags, 0, 4); | |
1316 | ||
1317 | enum bch_kdf_types { | |
1318 | BCH_KDF_SCRYPT = 0, | |
1319 | BCH_KDF_NR = 1, | |
1320 | }; | |
1321 | ||
1322 | /* stored as base 2 log of scrypt params: */ | |
1323 | LE64_BITMASK(BCH_KDF_SCRYPT_N, struct bch_sb_field_crypt, kdf_flags, 0, 16); | |
1324 | LE64_BITMASK(BCH_KDF_SCRYPT_R, struct bch_sb_field_crypt, kdf_flags, 16, 32); | |
1325 | LE64_BITMASK(BCH_KDF_SCRYPT_P, struct bch_sb_field_crypt, kdf_flags, 32, 48); | |
1326 | ||
1327 | /* BCH_SB_FIELD_replicas: */ | |
1328 | ||
89fd25be | 1329 | #define BCH_DATA_TYPES() \ |
822835ff | 1330 | x(free, 0) \ |
89fd25be KO |
1331 | x(sb, 1) \ |
1332 | x(journal, 2) \ | |
1333 | x(btree, 3) \ | |
1334 | x(user, 4) \ | |
af4d05c4 | 1335 | x(cached, 5) \ |
822835ff KO |
1336 | x(parity, 6) \ |
1337 | x(stripe, 7) \ | |
1338 | x(need_gc_gens, 8) \ | |
1339 | x(need_discard, 9) | |
89fd25be | 1340 | |
1c6fdbd8 | 1341 | enum bch_data_type { |
89fd25be KO |
1342 | #define x(t, n) BCH_DATA_##t, |
1343 | BCH_DATA_TYPES() | |
1344 | #undef x | |
1345 | BCH_DATA_NR | |
1c6fdbd8 KO |
1346 | }; |
1347 | ||
822835ff KO |
1348 | static inline bool data_type_is_empty(enum bch_data_type type) |
1349 | { | |
1350 | switch (type) { | |
1351 | case BCH_DATA_free: | |
1352 | case BCH_DATA_need_gc_gens: | |
1353 | case BCH_DATA_need_discard: | |
1354 | return true; | |
1355 | default: | |
1356 | return false; | |
1357 | } | |
1358 | } | |
1359 | ||
1360 | static inline bool data_type_is_hidden(enum bch_data_type type) | |
1361 | { | |
1362 | switch (type) { | |
1363 | case BCH_DATA_sb: | |
1364 | case BCH_DATA_journal: | |
1365 | return true; | |
1366 | default: | |
1367 | return false; | |
1368 | } | |
1369 | } | |
1370 | ||
af9d3bc2 KO |
1371 | struct bch_replicas_entry_v0 { |
1372 | __u8 data_type; | |
1373 | __u8 nr_devs; | |
73bd774d | 1374 | __u8 devs[0]; |
fd0c7679 | 1375 | } __packed; |
af9d3bc2 KO |
1376 | |
1377 | struct bch_sb_field_replicas_v0 { | |
1378 | struct bch_sb_field field; | |
73bd774d | 1379 | struct bch_replicas_entry_v0 entries[0]; |
fd0c7679 | 1380 | } __packed __aligned(8); |
af9d3bc2 | 1381 | |
1c6fdbd8 | 1382 | struct bch_replicas_entry { |
7a920560 KO |
1383 | __u8 data_type; |
1384 | __u8 nr_devs; | |
af9d3bc2 | 1385 | __u8 nr_required; |
73bd774d | 1386 | __u8 devs[0]; |
fd0c7679 | 1387 | } __packed; |
1c6fdbd8 | 1388 | |
22502ac2 KO |
1389 | #define replicas_entry_bytes(_i) \ |
1390 | (offsetof(typeof(*(_i)), devs) + (_i)->nr_devs) | |
1391 | ||
1c6fdbd8 KO |
1392 | struct bch_sb_field_replicas { |
1393 | struct bch_sb_field field; | |
73bd774d | 1394 | struct bch_replicas_entry entries[0]; |
fd0c7679 | 1395 | } __packed __aligned(8); |
1c6fdbd8 KO |
1396 | |
1397 | /* BCH_SB_FIELD_quota: */ | |
1398 | ||
1399 | struct bch_sb_quota_counter { | |
1400 | __le32 timelimit; | |
1401 | __le32 warnlimit; | |
1402 | }; | |
1403 | ||
1404 | struct bch_sb_quota_type { | |
1405 | __le64 flags; | |
1406 | struct bch_sb_quota_counter c[Q_COUNTERS]; | |
1407 | }; | |
1408 | ||
1409 | struct bch_sb_field_quota { | |
1410 | struct bch_sb_field field; | |
1411 | struct bch_sb_quota_type q[QTYP_NR]; | |
fd0c7679 | 1412 | } __packed __aligned(8); |
1c6fdbd8 KO |
1413 | |
1414 | /* BCH_SB_FIELD_disk_groups: */ | |
1415 | ||
1416 | #define BCH_SB_LABEL_SIZE 32 | |
1417 | ||
1418 | struct bch_disk_group { | |
1419 | __u8 label[BCH_SB_LABEL_SIZE]; | |
1420 | __le64 flags[2]; | |
fd0c7679 | 1421 | } __packed __aligned(8); |
1c6fdbd8 KO |
1422 | |
1423 | LE64_BITMASK(BCH_GROUP_DELETED, struct bch_disk_group, flags[0], 0, 1) | |
1424 | LE64_BITMASK(BCH_GROUP_DATA_ALLOWED, struct bch_disk_group, flags[0], 1, 6) | |
1425 | LE64_BITMASK(BCH_GROUP_PARENT, struct bch_disk_group, flags[0], 6, 24) | |
1426 | ||
1427 | struct bch_sb_field_disk_groups { | |
1428 | struct bch_sb_field field; | |
1429 | struct bch_disk_group entries[0]; | |
fd0c7679 | 1430 | } __packed __aligned(8); |
1c6fdbd8 | 1431 | |
104c6974 DH |
1432 | /* BCH_SB_FIELD_counters */ |
1433 | ||
674cfc26 KO |
1434 | #define BCH_PERSISTENT_COUNTERS() \ |
1435 | x(io_read, 0) \ | |
1436 | x(io_write, 1) \ | |
1437 | x(io_move, 2) \ | |
1438 | x(bucket_invalidate, 3) \ | |
1439 | x(bucket_discard, 4) \ | |
1440 | x(bucket_alloc, 5) \ | |
1441 | x(bucket_alloc_fail, 6) \ | |
1442 | x(btree_cache_scan, 7) \ | |
1443 | x(btree_cache_reap, 8) \ | |
1444 | x(btree_cache_cannibalize, 9) \ | |
1445 | x(btree_cache_cannibalize_lock, 10) \ | |
1446 | x(btree_cache_cannibalize_lock_fail, 11) \ | |
1447 | x(btree_cache_cannibalize_unlock, 12) \ | |
1448 | x(btree_node_write, 13) \ | |
1449 | x(btree_node_read, 14) \ | |
1450 | x(btree_node_compact, 15) \ | |
1451 | x(btree_node_merge, 16) \ | |
1452 | x(btree_node_split, 17) \ | |
1453 | x(btree_node_rewrite, 18) \ | |
1454 | x(btree_node_alloc, 19) \ | |
1455 | x(btree_node_free, 20) \ | |
1456 | x(btree_node_set_root, 21) \ | |
1457 | x(btree_path_relock_fail, 22) \ | |
1458 | x(btree_path_upgrade_fail, 23) \ | |
1459 | x(btree_reserve_get_fail, 24) \ | |
1460 | x(journal_entry_full, 25) \ | |
1461 | x(journal_full, 26) \ | |
1462 | x(journal_reclaim_finish, 27) \ | |
1463 | x(journal_reclaim_start, 28) \ | |
1464 | x(journal_write, 29) \ | |
1465 | x(read_promote, 30) \ | |
1466 | x(read_bounce, 31) \ | |
1467 | x(read_split, 33) \ | |
1468 | x(read_retry, 32) \ | |
1469 | x(read_reuse_race, 34) \ | |
1470 | x(move_extent_read, 35) \ | |
1471 | x(move_extent_write, 36) \ | |
1472 | x(move_extent_finish, 37) \ | |
1473 | x(move_extent_fail, 38) \ | |
1474 | x(move_extent_alloc_mem_fail, 39) \ | |
1475 | x(copygc, 40) \ | |
1476 | x(copygc_wait, 41) \ | |
1477 | x(gc_gens_end, 42) \ | |
1478 | x(gc_gens_start, 43) \ | |
1479 | x(trans_blocked_journal_reclaim, 44) \ | |
1480 | x(trans_restart_btree_node_reused, 45) \ | |
1481 | x(trans_restart_btree_node_split, 46) \ | |
1482 | x(trans_restart_fault_inject, 47) \ | |
1483 | x(trans_restart_iter_upgrade, 48) \ | |
1484 | x(trans_restart_journal_preres_get, 49) \ | |
1485 | x(trans_restart_journal_reclaim, 50) \ | |
1486 | x(trans_restart_journal_res_get, 51) \ | |
1487 | x(trans_restart_key_cache_key_realloced, 52) \ | |
1488 | x(trans_restart_key_cache_raced, 53) \ | |
1489 | x(trans_restart_mark_replicas, 54) \ | |
1490 | x(trans_restart_mem_realloced, 55) \ | |
1491 | x(trans_restart_memory_allocation_failure, 56) \ | |
1492 | x(trans_restart_relock, 57) \ | |
1493 | x(trans_restart_relock_after_fill, 58) \ | |
1494 | x(trans_restart_relock_key_cache_fill, 59) \ | |
1495 | x(trans_restart_relock_next_node, 60) \ | |
1496 | x(trans_restart_relock_parent_for_fill, 61) \ | |
1497 | x(trans_restart_relock_path, 62) \ | |
1498 | x(trans_restart_relock_path_intent, 63) \ | |
1499 | x(trans_restart_too_many_iters, 64) \ | |
1500 | x(trans_restart_traverse, 65) \ | |
1501 | x(trans_restart_upgrade, 66) \ | |
1502 | x(trans_restart_would_deadlock, 67) \ | |
1503 | x(trans_restart_would_deadlock_write, 68) \ | |
1504 | x(trans_restart_injected, 69) \ | |
1505 | x(trans_restart_key_cache_upgrade, 70) \ | |
1506 | x(trans_traverse_all, 71) \ | |
1507 | x(transaction_commit, 72) \ | |
33bd5d06 | 1508 | x(write_super, 73) \ |
920e69bc | 1509 | x(trans_restart_would_deadlock_recursion_limit, 74) \ |
e151580d KO |
1510 | x(trans_restart_write_buffer_flush, 75) \ |
1511 | x(trans_restart_split_race, 76) | |
104c6974 DH |
1512 | |
1513 | enum bch_persistent_counters { | |
1514 | #define x(t, n, ...) BCH_COUNTER_##t, | |
1515 | BCH_PERSISTENT_COUNTERS() | |
1516 | #undef x | |
1517 | BCH_COUNTER_NR | |
1518 | }; | |
1519 | ||
1520 | struct bch_sb_field_counters { | |
1521 | struct bch_sb_field field; | |
1522 | __le64 d[0]; | |
1523 | }; | |
1524 | ||
1c6fdbd8 KO |
1525 | /* |
1526 | * On clean shutdown, store btree roots and current journal sequence number in | |
1527 | * the superblock: | |
1528 | */ | |
1529 | struct jset_entry { | |
1530 | __le16 u64s; | |
1531 | __u8 btree_id; | |
1532 | __u8 level; | |
1533 | __u8 type; /* designates what this jset holds */ | |
1534 | __u8 pad[3]; | |
1535 | ||
1536 | union { | |
1537 | struct bkey_i start[0]; | |
1538 | __u64 _data[0]; | |
1539 | }; | |
1540 | }; | |
1541 | ||
1542 | struct bch_sb_field_clean { | |
1543 | struct bch_sb_field field; | |
1544 | ||
1545 | __le32 flags; | |
2abe5420 KO |
1546 | __le16 _read_clock; /* no longer used */ |
1547 | __le16 _write_clock; | |
1c6fdbd8 KO |
1548 | __le64 journal_seq; |
1549 | ||
1550 | union { | |
1551 | struct jset_entry start[0]; | |
1552 | __u64 _data[0]; | |
1553 | }; | |
1554 | }; | |
1555 | ||
1dd7f9d9 KO |
1556 | struct journal_seq_blacklist_entry { |
1557 | __le64 start; | |
1558 | __le64 end; | |
1559 | }; | |
1560 | ||
1561 | struct bch_sb_field_journal_seq_blacklist { | |
1562 | struct bch_sb_field field; | |
1563 | ||
1564 | union { | |
1565 | struct journal_seq_blacklist_entry start[0]; | |
1566 | __u64 _data[0]; | |
1567 | }; | |
1568 | }; | |
1569 | ||
1c6fdbd8 KO |
1570 | /* Superblock: */ |
1571 | ||
1572 | /* | |
26609b61 KO |
1573 | * New versioning scheme: |
1574 | * One common version number for all on disk data structures - superblock, btree | |
1575 | * nodes, journal entries | |
1c6fdbd8 | 1576 | */ |
26609b61 | 1577 | |
74b33393 KO |
1578 | #define BCH_METADATA_VERSIONS() \ |
1579 | x(bkey_renumber, 10) \ | |
1580 | x(inode_btree_change, 11) \ | |
1581 | x(snapshot, 12) \ | |
1582 | x(inode_backpointers, 13) \ | |
1583 | x(btree_ptr_sectors_written, 14) \ | |
1584 | x(snapshot_2, 15) \ | |
1585 | x(reflink_p_fix, 16) \ | |
1586 | x(subvol_dirent, 17) \ | |
c6b2826c KO |
1587 | x(inode_v2, 18) \ |
1588 | x(freespace, 19) \ | |
822835ff | 1589 | x(alloc_v4, 20) \ |
a8c752bb | 1590 | x(new_data_types, 21) \ |
8dd69d9f | 1591 | x(backpointers, 22) \ |
a8b3a677 | 1592 | x(inode_v3, 23) \ |
5250b74d | 1593 | x(unwritten_extents, 24) \ |
83f33d68 | 1594 | x(bucket_gens, 25) \ |
80c33085 | 1595 | x(lru_v2, 26) \ |
62a03559 | 1596 | x(fragmentation_lru, 27) \ |
1c59b483 KO |
1597 | x(no_bps_in_alloc_keys, 28) \ |
1598 | x(snapshot_trees, 29) | |
74b33393 | 1599 | |
26609b61 | 1600 | enum bcachefs_metadata_version { |
74b33393 KO |
1601 | bcachefs_metadata_version_min = 9, |
1602 | #define x(t, n) bcachefs_metadata_version_##t = n, | |
1603 | BCH_METADATA_VERSIONS() | |
1604 | #undef x | |
1605 | bcachefs_metadata_version_max | |
26609b61 | 1606 | }; |
1c6fdbd8 | 1607 | |
1c59b483 KO |
1608 | static const unsigned bcachefs_metadata_required_upgrade_below = bcachefs_metadata_version_snapshot_trees; |
1609 | ||
26609b61 | 1610 | #define bcachefs_metadata_version_current (bcachefs_metadata_version_max - 1) |
1c6fdbd8 KO |
1611 | |
1612 | #define BCH_SB_SECTOR 8 | |
1613 | #define BCH_SB_MEMBERS_MAX 64 /* XXX kill */ | |
1614 | ||
1615 | struct bch_sb_layout { | |
1616 | __uuid_t magic; /* bcachefs superblock UUID */ | |
1617 | __u8 layout_type; | |
1618 | __u8 sb_max_size_bits; /* base 2 of 512 byte sectors */ | |
1619 | __u8 nr_superblocks; | |
1620 | __u8 pad[5]; | |
1621 | __le64 sb_offset[61]; | |
fd0c7679 | 1622 | } __packed __aligned(8); |
1c6fdbd8 KO |
1623 | |
1624 | #define BCH_SB_LAYOUT_SECTOR 7 | |
1625 | ||
1626 | /* | |
1627 | * @offset - sector where this sb was written | |
1628 | * @version - on disk format version | |
26609b61 KO |
1629 | * @version_min - Oldest metadata version this filesystem contains; so we can |
1630 | * safely drop compatibility code and refuse to mount filesystems | |
1631 | * we'd need it for | |
e1538212 | 1632 | * @magic - identifies as a bcachefs superblock (BCHFS_MAGIC) |
1c6fdbd8 KO |
1633 | * @seq - incremented each time superblock is written |
1634 | * @uuid - used for generating various magic numbers and identifying | |
1635 | * member devices, never changes | |
1636 | * @user_uuid - user visible UUID, may be changed | |
1637 | * @label - filesystem label | |
1638 | * @seq - identifies most recent superblock, incremented each time | |
1639 | * superblock is written | |
1640 | * @features - enabled incompatible features | |
1641 | */ | |
1642 | struct bch_sb { | |
1643 | struct bch_csum csum; | |
1644 | __le16 version; | |
1645 | __le16 version_min; | |
1646 | __le16 pad[2]; | |
1647 | __uuid_t magic; | |
1648 | __uuid_t uuid; | |
1649 | __uuid_t user_uuid; | |
1650 | __u8 label[BCH_SB_LABEL_SIZE]; | |
1651 | __le64 offset; | |
1652 | __le64 seq; | |
1653 | ||
1654 | __le16 block_size; | |
1655 | __u8 dev_idx; | |
1656 | __u8 nr_devices; | |
1657 | __le32 u64s; | |
1658 | ||
1659 | __le64 time_base_lo; | |
1660 | __le32 time_base_hi; | |
1661 | __le32 time_precision; | |
1662 | ||
1663 | __le64 flags[8]; | |
1664 | __le64 features[2]; | |
1665 | __le64 compat[2]; | |
1666 | ||
1667 | struct bch_sb_layout layout; | |
1668 | ||
1669 | union { | |
1670 | struct bch_sb_field start[0]; | |
1671 | __le64 _data[0]; | |
1672 | }; | |
fd0c7679 | 1673 | } __packed __aligned(8); |
1c6fdbd8 KO |
1674 | |
1675 | /* | |
1676 | * Flags: | |
1677 | * BCH_SB_INITALIZED - set on first mount | |
1678 | * BCH_SB_CLEAN - did we shut down cleanly? Just a hint, doesn't affect | |
1679 | * behaviour of mount/recovery path: | |
1680 | * BCH_SB_INODE_32BIT - limit inode numbers to 32 bits | |
1681 | * BCH_SB_128_BIT_MACS - 128 bit macs instead of 80 | |
1682 | * BCH_SB_ENCRYPTION_TYPE - if nonzero encryption is enabled; overrides | |
1683 | * DATA/META_CSUM_TYPE. Also indicates encryption | |
1684 | * algorithm in use, if/when we get more than one | |
1685 | */ | |
1686 | ||
1687 | LE16_BITMASK(BCH_SB_BLOCK_SIZE, struct bch_sb, block_size, 0, 16); | |
1688 | ||
1689 | LE64_BITMASK(BCH_SB_INITIALIZED, struct bch_sb, flags[0], 0, 1); | |
1690 | LE64_BITMASK(BCH_SB_CLEAN, struct bch_sb, flags[0], 1, 2); | |
1691 | LE64_BITMASK(BCH_SB_CSUM_TYPE, struct bch_sb, flags[0], 2, 8); | |
1692 | LE64_BITMASK(BCH_SB_ERROR_ACTION, struct bch_sb, flags[0], 8, 12); | |
1693 | ||
1694 | LE64_BITMASK(BCH_SB_BTREE_NODE_SIZE, struct bch_sb, flags[0], 12, 28); | |
1695 | ||
1696 | LE64_BITMASK(BCH_SB_GC_RESERVE, struct bch_sb, flags[0], 28, 33); | |
1697 | LE64_BITMASK(BCH_SB_ROOT_RESERVE, struct bch_sb, flags[0], 33, 40); | |
1698 | ||
1699 | LE64_BITMASK(BCH_SB_META_CSUM_TYPE, struct bch_sb, flags[0], 40, 44); | |
1700 | LE64_BITMASK(BCH_SB_DATA_CSUM_TYPE, struct bch_sb, flags[0], 44, 48); | |
1701 | ||
1702 | LE64_BITMASK(BCH_SB_META_REPLICAS_WANT, struct bch_sb, flags[0], 48, 52); | |
1703 | LE64_BITMASK(BCH_SB_DATA_REPLICAS_WANT, struct bch_sb, flags[0], 52, 56); | |
1704 | ||
1705 | LE64_BITMASK(BCH_SB_POSIX_ACL, struct bch_sb, flags[0], 56, 57); | |
1706 | LE64_BITMASK(BCH_SB_USRQUOTA, struct bch_sb, flags[0], 57, 58); | |
1707 | LE64_BITMASK(BCH_SB_GRPQUOTA, struct bch_sb, flags[0], 58, 59); | |
1708 | LE64_BITMASK(BCH_SB_PRJQUOTA, struct bch_sb, flags[0], 59, 60); | |
1709 | ||
0bc166ff | 1710 | LE64_BITMASK(BCH_SB_HAS_ERRORS, struct bch_sb, flags[0], 60, 61); |
aae15aaf | 1711 | LE64_BITMASK(BCH_SB_HAS_TOPOLOGY_ERRORS,struct bch_sb, flags[0], 61, 62); |
0bc166ff | 1712 | |
7d6f07ed | 1713 | LE64_BITMASK(BCH_SB_BIG_ENDIAN, struct bch_sb, flags[0], 62, 63); |
36b8372b | 1714 | |
1c6fdbd8 KO |
1715 | LE64_BITMASK(BCH_SB_STR_HASH_TYPE, struct bch_sb, flags[1], 0, 4); |
1716 | LE64_BITMASK(BCH_SB_COMPRESSION_TYPE, struct bch_sb, flags[1], 4, 8); | |
1717 | LE64_BITMASK(BCH_SB_INODE_32BIT, struct bch_sb, flags[1], 8, 9); | |
1718 | ||
1719 | LE64_BITMASK(BCH_SB_128_BIT_MACS, struct bch_sb, flags[1], 9, 10); | |
1720 | LE64_BITMASK(BCH_SB_ENCRYPTION_TYPE, struct bch_sb, flags[1], 10, 14); | |
1721 | ||
1722 | /* | |
1723 | * Max size of an extent that may require bouncing to read or write | |
1724 | * (checksummed, compressed): 64k | |
1725 | */ | |
1726 | LE64_BITMASK(BCH_SB_ENCODED_EXTENT_MAX_BITS, | |
1727 | struct bch_sb, flags[1], 14, 20); | |
1728 | ||
1729 | LE64_BITMASK(BCH_SB_META_REPLICAS_REQ, struct bch_sb, flags[1], 20, 24); | |
1730 | LE64_BITMASK(BCH_SB_DATA_REPLICAS_REQ, struct bch_sb, flags[1], 24, 28); | |
1731 | ||
1732 | LE64_BITMASK(BCH_SB_PROMOTE_TARGET, struct bch_sb, flags[1], 28, 40); | |
1733 | LE64_BITMASK(BCH_SB_FOREGROUND_TARGET, struct bch_sb, flags[1], 40, 52); | |
1734 | LE64_BITMASK(BCH_SB_BACKGROUND_TARGET, struct bch_sb, flags[1], 52, 64); | |
1735 | ||
1736 | LE64_BITMASK(BCH_SB_BACKGROUND_COMPRESSION_TYPE, | |
1737 | struct bch_sb, flags[2], 0, 4); | |
a50ed7c8 | 1738 | LE64_BITMASK(BCH_SB_GC_RESERVE_BYTES, struct bch_sb, flags[2], 4, 64); |
1c6fdbd8 | 1739 | |
cd575ddf | 1740 | LE64_BITMASK(BCH_SB_ERASURE_CODE, struct bch_sb, flags[3], 0, 16); |
d042b040 | 1741 | LE64_BITMASK(BCH_SB_METADATA_TARGET, struct bch_sb, flags[3], 16, 28); |
b282a74f | 1742 | LE64_BITMASK(BCH_SB_SHARD_INUMS, struct bch_sb, flags[3], 28, 29); |
996fb577 | 1743 | LE64_BITMASK(BCH_SB_INODES_USE_KEY_CACHE,struct bch_sb, flags[3], 29, 30); |
2430e72f KO |
1744 | LE64_BITMASK(BCH_SB_JOURNAL_FLUSH_DELAY,struct bch_sb, flags[3], 30, 62); |
1745 | LE64_BITMASK(BCH_SB_JOURNAL_FLUSH_DISABLED,struct bch_sb, flags[3], 62, 63); | |
1746 | LE64_BITMASK(BCH_SB_JOURNAL_RECLAIM_DELAY,struct bch_sb, flags[4], 0, 32); | |
fb64f3fd | 1747 | LE64_BITMASK(BCH_SB_JOURNAL_TRANSACTION_NAMES,struct bch_sb, flags[4], 32, 33); |
a8b3a677 | 1748 | LE64_BITMASK(BCH_SB_NOCOW, struct bch_sb, flags[4], 33, 34); |
920e69bc | 1749 | LE64_BITMASK(BCH_SB_WRITE_BUFFER_SIZE, struct bch_sb, flags[4], 34, 54); |
cd575ddf | 1750 | |
24964e1c KO |
1751 | /* flags[4] 56-64 unused: */ |
1752 | ||
1753 | LE64_BITMASK(BCH_SB_VERSION_UPGRADE_COMPLETE, | |
1754 | struct bch_sb, flags[5], 0, 16); | |
1755 | ||
1c3ff72c KO |
1756 | /* |
1757 | * Features: | |
1758 | * | |
1759 | * journal_seq_blacklist_v3: gates BCH_SB_FIELD_journal_seq_blacklist | |
1760 | * reflink: gates KEY_TYPE_reflink | |
1761 | * inline_data: gates KEY_TYPE_inline_data | |
6404dcc9 | 1762 | * new_siphash: gates BCH_STR_HASH_siphash |
bcd6f3e0 | 1763 | * new_extent_overwrite: gates BTREE_NODE_NEW_EXTENT_OVERWRITE |
1c3ff72c KO |
1764 | */ |
1765 | #define BCH_SB_FEATURES() \ | |
1766 | x(lz4, 0) \ | |
1767 | x(gzip, 1) \ | |
1768 | x(zstd, 2) \ | |
1769 | x(atomic_nlink, 3) \ | |
1770 | x(ec, 4) \ | |
1771 | x(journal_seq_blacklist_v3, 5) \ | |
1772 | x(reflink, 6) \ | |
1773 | x(new_siphash, 7) \ | |
bcd6f3e0 | 1774 | x(inline_data, 8) \ |
ab05de4c | 1775 | x(new_extent_overwrite, 9) \ |
548b3d20 | 1776 | x(incompressible, 10) \ |
e3e464ac | 1777 | x(btree_ptr_v2, 11) \ |
6357d607 | 1778 | x(extents_above_btree_updates, 12) \ |
801a3de6 | 1779 | x(btree_updates_journalled, 13) \ |
a3e72262 | 1780 | x(reflink_inline_data, 14) \ |
adbcada4 | 1781 | x(new_varint, 15) \ |
7f4e1d5d | 1782 | x(journal_no_flush, 16) \ |
8042b5b7 KO |
1783 | x(alloc_v2, 17) \ |
1784 | x(extents_across_btree_nodes, 18) | |
1785 | ||
1786 | #define BCH_SB_FEATURES_ALWAYS \ | |
1787 | ((1ULL << BCH_FEATURE_new_extent_overwrite)| \ | |
1788 | (1ULL << BCH_FEATURE_extents_above_btree_updates)|\ | |
1789 | (1ULL << BCH_FEATURE_btree_updates_journalled)|\ | |
73590619 | 1790 | (1ULL << BCH_FEATURE_alloc_v2)|\ |
8042b5b7 | 1791 | (1ULL << BCH_FEATURE_extents_across_btree_nodes)) |
1c3ff72c | 1792 | |
b807a0c8 | 1793 | #define BCH_SB_FEATURES_ALL \ |
8042b5b7 KO |
1794 | (BCH_SB_FEATURES_ALWAYS| \ |
1795 | (1ULL << BCH_FEATURE_new_siphash)| \ | |
e3e464ac | 1796 | (1ULL << BCH_FEATURE_btree_ptr_v2)| \ |
adbcada4 | 1797 | (1ULL << BCH_FEATURE_new_varint)| \ |
73590619 | 1798 | (1ULL << BCH_FEATURE_journal_no_flush)) |
b807a0c8 | 1799 | |
1c3ff72c KO |
1800 | enum bch_sb_feature { |
1801 | #define x(f, n) BCH_FEATURE_##f, | |
1802 | BCH_SB_FEATURES() | |
1803 | #undef x | |
c258f28e | 1804 | BCH_FEATURE_NR, |
1c6fdbd8 KO |
1805 | }; |
1806 | ||
19dd3172 KO |
1807 | #define BCH_SB_COMPAT() \ |
1808 | x(alloc_info, 0) \ | |
1809 | x(alloc_metadata, 1) \ | |
1810 | x(extents_above_btree_updates_done, 2) \ | |
1811 | x(bformat_overflow_done, 3) | |
1812 | ||
1df42b57 | 1813 | enum bch_sb_compat { |
19dd3172 KO |
1814 | #define x(f, n) BCH_COMPAT_##f, |
1815 | BCH_SB_COMPAT() | |
1816 | #undef x | |
1817 | BCH_COMPAT_NR, | |
1df42b57 KO |
1818 | }; |
1819 | ||
1c6fdbd8 KO |
1820 | /* options: */ |
1821 | ||
1822 | #define BCH_REPLICAS_MAX 4U | |
1823 | ||
ffb7c3d3 KO |
1824 | #define BCH_BKEY_PTRS_MAX 16U |
1825 | ||
2436cb9f KO |
1826 | #define BCH_ERROR_ACTIONS() \ |
1827 | x(continue, 0) \ | |
1828 | x(ro, 1) \ | |
1829 | x(panic, 2) | |
1830 | ||
1c6fdbd8 | 1831 | enum bch_error_actions { |
2436cb9f KO |
1832 | #define x(t, n) BCH_ON_ERROR_##t = n, |
1833 | BCH_ERROR_ACTIONS() | |
1834 | #undef x | |
1835 | BCH_ON_ERROR_NR | |
1c6fdbd8 KO |
1836 | }; |
1837 | ||
6404dcc9 KO |
1838 | #define BCH_STR_HASH_TYPES() \ |
1839 | x(crc32c, 0) \ | |
1840 | x(crc64, 1) \ | |
1841 | x(siphash_old, 2) \ | |
1842 | x(siphash, 3) | |
1843 | ||
73501ab8 | 1844 | enum bch_str_hash_type { |
6404dcc9 KO |
1845 | #define x(t, n) BCH_STR_HASH_##t = n, |
1846 | BCH_STR_HASH_TYPES() | |
1847 | #undef x | |
1848 | BCH_STR_HASH_NR | |
73501ab8 KO |
1849 | }; |
1850 | ||
2436cb9f KO |
1851 | #define BCH_STR_HASH_OPTS() \ |
1852 | x(crc32c, 0) \ | |
1853 | x(crc64, 1) \ | |
1854 | x(siphash, 2) | |
1855 | ||
73501ab8 | 1856 | enum bch_str_hash_opts { |
2436cb9f KO |
1857 | #define x(t, n) BCH_STR_HASH_OPT_##t = n, |
1858 | BCH_STR_HASH_OPTS() | |
1859 | #undef x | |
1860 | BCH_STR_HASH_OPT_NR | |
1c6fdbd8 KO |
1861 | }; |
1862 | ||
6404dcc9 KO |
1863 | #define BCH_CSUM_TYPES() \ |
1864 | x(none, 0) \ | |
1865 | x(crc32c_nonzero, 1) \ | |
1866 | x(crc64_nonzero, 2) \ | |
1867 | x(chacha20_poly1305_80, 3) \ | |
1868 | x(chacha20_poly1305_128, 4) \ | |
1869 | x(crc32c, 5) \ | |
1870 | x(crc64, 6) \ | |
1871 | x(xxhash, 7) | |
1872 | ||
1c3ff72c | 1873 | enum bch_csum_type { |
6404dcc9 KO |
1874 | #define x(t, n) BCH_CSUM_##t = n, |
1875 | BCH_CSUM_TYPES() | |
1876 | #undef x | |
1877 | BCH_CSUM_NR | |
1c3ff72c KO |
1878 | }; |
1879 | ||
1880 | static const unsigned bch_crc_bytes[] = { | |
6404dcc9 KO |
1881 | [BCH_CSUM_none] = 0, |
1882 | [BCH_CSUM_crc32c_nonzero] = 4, | |
1883 | [BCH_CSUM_crc32c] = 4, | |
1884 | [BCH_CSUM_crc64_nonzero] = 8, | |
1885 | [BCH_CSUM_crc64] = 8, | |
1886 | [BCH_CSUM_xxhash] = 8, | |
1887 | [BCH_CSUM_chacha20_poly1305_80] = 10, | |
1888 | [BCH_CSUM_chacha20_poly1305_128] = 16, | |
1c3ff72c KO |
1889 | }; |
1890 | ||
1891 | static inline _Bool bch2_csum_type_is_encryption(enum bch_csum_type type) | |
1892 | { | |
1893 | switch (type) { | |
6404dcc9 KO |
1894 | case BCH_CSUM_chacha20_poly1305_80: |
1895 | case BCH_CSUM_chacha20_poly1305_128: | |
1c3ff72c KO |
1896 | return true; |
1897 | default: | |
1898 | return false; | |
1899 | } | |
1900 | } | |
1901 | ||
2436cb9f KO |
1902 | #define BCH_CSUM_OPTS() \ |
1903 | x(none, 0) \ | |
1904 | x(crc32c, 1) \ | |
41e63382 | 1905 | x(crc64, 2) \ |
1906 | x(xxhash, 3) | |
2436cb9f | 1907 | |
1c3ff72c | 1908 | enum bch_csum_opts { |
2436cb9f KO |
1909 | #define x(t, n) BCH_CSUM_OPT_##t = n, |
1910 | BCH_CSUM_OPTS() | |
1911 | #undef x | |
1912 | BCH_CSUM_OPT_NR | |
1c3ff72c KO |
1913 | }; |
1914 | ||
1c6fdbd8 | 1915 | #define BCH_COMPRESSION_TYPES() \ |
ab05de4c KO |
1916 | x(none, 0) \ |
1917 | x(lz4_old, 1) \ | |
1918 | x(gzip, 2) \ | |
1919 | x(lz4, 3) \ | |
1920 | x(zstd, 4) \ | |
1921 | x(incompressible, 5) | |
1c6fdbd8 | 1922 | |
1c3ff72c | 1923 | enum bch_compression_type { |
2436cb9f | 1924 | #define x(t, n) BCH_COMPRESSION_TYPE_##t = n, |
1c6fdbd8 | 1925 | BCH_COMPRESSION_TYPES() |
1c3ff72c KO |
1926 | #undef x |
1927 | BCH_COMPRESSION_TYPE_NR | |
1928 | }; | |
1929 | ||
1930 | #define BCH_COMPRESSION_OPTS() \ | |
1931 | x(none, 0) \ | |
1932 | x(lz4, 1) \ | |
1933 | x(gzip, 2) \ | |
1934 | x(zstd, 3) | |
1935 | ||
1936 | enum bch_compression_opts { | |
2436cb9f | 1937 | #define x(t, n) BCH_COMPRESSION_OPT_##t = n, |
1c3ff72c | 1938 | BCH_COMPRESSION_OPTS() |
1c6fdbd8 KO |
1939 | #undef x |
1940 | BCH_COMPRESSION_OPT_NR | |
1941 | }; | |
1942 | ||
1943 | /* | |
1944 | * Magic numbers | |
1945 | * | |
1946 | * The various other data structures have their own magic numbers, which are | |
1947 | * xored with the first part of the cache set's UUID | |
1948 | */ | |
1949 | ||
1950 | #define BCACHE_MAGIC \ | |
1951 | UUID_INIT(0xc68573f6, 0x4e1a, 0x45ca, \ | |
1952 | 0x82, 0x65, 0xf5, 0x7f, 0x48, 0xba, 0x6d, 0x81) | |
1953 | #define BCHFS_MAGIC \ | |
1954 | UUID_INIT(0xc68573f6, 0x66ce, 0x90a9, \ | |
1955 | 0xd9, 0x6a, 0x60, 0xcf, 0x80, 0x3d, 0xf7, 0xef) | |
1956 | ||
1957 | #define BCACHEFS_STATFS_MAGIC 0xca451a4e | |
1958 | ||
1959 | #define JSET_MAGIC __cpu_to_le64(0x245235c1a3625032ULL) | |
1960 | #define BSET_MAGIC __cpu_to_le64(0x90135c78b99e07f5ULL) | |
1961 | ||
1962 | static inline __le64 __bch2_sb_magic(struct bch_sb *sb) | |
1963 | { | |
1964 | __le64 ret; | |
a1019576 | 1965 | |
1c6fdbd8 KO |
1966 | memcpy(&ret, &sb->uuid, sizeof(ret)); |
1967 | return ret; | |
1968 | } | |
1969 | ||
1970 | static inline __u64 __jset_magic(struct bch_sb *sb) | |
1971 | { | |
1972 | return __le64_to_cpu(__bch2_sb_magic(sb) ^ JSET_MAGIC); | |
1973 | } | |
1974 | ||
1975 | static inline __u64 __bset_magic(struct bch_sb *sb) | |
1976 | { | |
1977 | return __le64_to_cpu(__bch2_sb_magic(sb) ^ BSET_MAGIC); | |
1978 | } | |
1979 | ||
1980 | /* Journal */ | |
1981 | ||
1c6fdbd8 KO |
1982 | #define JSET_KEYS_U64s (sizeof(struct jset_entry) / sizeof(__u64)) |
1983 | ||
1984 | #define BCH_JSET_ENTRY_TYPES() \ | |
1985 | x(btree_keys, 0) \ | |
1986 | x(btree_root, 1) \ | |
1987 | x(prio_ptrs, 2) \ | |
1988 | x(blacklist, 3) \ | |
2c5af169 | 1989 | x(blacklist_v2, 4) \ |
3577df5f | 1990 | x(usage, 5) \ |
2abe5420 | 1991 | x(data_usage, 6) \ |
180fb49d | 1992 | x(clock, 7) \ |
fb64f3fd | 1993 | x(dev_usage, 8) \ |
cb685ce7 KO |
1994 | x(log, 9) \ |
1995 | x(overwrite, 10) | |
1c6fdbd8 KO |
1996 | |
1997 | enum { | |
1998 | #define x(f, nr) BCH_JSET_ENTRY_##f = nr, | |
1999 | BCH_JSET_ENTRY_TYPES() | |
2000 | #undef x | |
2001 | BCH_JSET_ENTRY_NR | |
2002 | }; | |
2003 | ||
2004 | /* | |
2005 | * Journal sequence numbers can be blacklisted: bsets record the max sequence | |
2006 | * number of all the journal entries they contain updates for, so that on | |
2007 | * recovery we can ignore those bsets that contain index updates newer that what | |
2008 | * made it into the journal. | |
2009 | * | |
2010 | * This means that we can't reuse that journal_seq - we have to skip it, and | |
2011 | * then record that we skipped it so that the next time we crash and recover we | |
2012 | * don't think there was a missing journal entry. | |
2013 | */ | |
2014 | struct jset_entry_blacklist { | |
2015 | struct jset_entry entry; | |
2016 | __le64 seq; | |
2017 | }; | |
2018 | ||
2019 | struct jset_entry_blacklist_v2 { | |
2020 | struct jset_entry entry; | |
2021 | __le64 start; | |
2022 | __le64 end; | |
2023 | }; | |
2024 | ||
528b18e6 KO |
2025 | #define BCH_FS_USAGE_TYPES() \ |
2026 | x(reserved, 0) \ | |
2027 | x(inodes, 1) \ | |
2028 | x(key_version, 2) | |
2029 | ||
2c5af169 | 2030 | enum { |
528b18e6 KO |
2031 | #define x(f, nr) BCH_FS_USAGE_##f = nr, |
2032 | BCH_FS_USAGE_TYPES() | |
2033 | #undef x | |
2034 | BCH_FS_USAGE_NR | |
2c5af169 KO |
2035 | }; |
2036 | ||
2037 | struct jset_entry_usage { | |
2038 | struct jset_entry entry; | |
3577df5f | 2039 | __le64 v; |
fd0c7679 | 2040 | } __packed; |
3577df5f KO |
2041 | |
2042 | struct jset_entry_data_usage { | |
2043 | struct jset_entry entry; | |
2044 | __le64 v; | |
2c5af169 | 2045 | struct bch_replicas_entry r; |
fd0c7679 | 2046 | } __packed; |
2c5af169 | 2047 | |
2abe5420 KO |
2048 | struct jset_entry_clock { |
2049 | struct jset_entry entry; | |
2050 | __u8 rw; | |
2051 | __u8 pad[7]; | |
2052 | __le64 time; | |
fd0c7679 | 2053 | } __packed; |
2abe5420 | 2054 | |
180fb49d KO |
2055 | struct jset_entry_dev_usage_type { |
2056 | __le64 buckets; | |
2057 | __le64 sectors; | |
2058 | __le64 fragmented; | |
fd0c7679 | 2059 | } __packed; |
180fb49d KO |
2060 | |
2061 | struct jset_entry_dev_usage { | |
2062 | struct jset_entry entry; | |
2063 | __le32 dev; | |
2064 | __u32 pad; | |
2065 | ||
2066 | __le64 buckets_ec; | |
822835ff | 2067 | __le64 _buckets_unavailable; /* No longer used */ |
180fb49d KO |
2068 | |
2069 | struct jset_entry_dev_usage_type d[]; | |
fd0c7679 | 2070 | } __packed; |
180fb49d | 2071 | |
528b18e6 KO |
2072 | static inline unsigned jset_entry_dev_usage_nr_types(struct jset_entry_dev_usage *u) |
2073 | { | |
2074 | return (vstruct_bytes(&u->entry) - sizeof(struct jset_entry_dev_usage)) / | |
2075 | sizeof(struct jset_entry_dev_usage_type); | |
2076 | } | |
2077 | ||
fb64f3fd KO |
2078 | struct jset_entry_log { |
2079 | struct jset_entry entry; | |
2080 | u8 d[]; | |
fd0c7679 | 2081 | } __packed; |
fb64f3fd | 2082 | |
1c6fdbd8 KO |
2083 | /* |
2084 | * On disk format for a journal entry: | |
2085 | * seq is monotonically increasing; every journal entry has its own unique | |
2086 | * sequence number. | |
2087 | * | |
2088 | * last_seq is the oldest journal entry that still has keys the btree hasn't | |
2089 | * flushed to disk yet. | |
2090 | * | |
2091 | * version is for on disk format changes. | |
2092 | */ | |
2093 | struct jset { | |
2094 | struct bch_csum csum; | |
2095 | ||
2096 | __le64 magic; | |
2097 | __le64 seq; | |
2098 | __le32 version; | |
2099 | __le32 flags; | |
2100 | ||
2101 | __le32 u64s; /* size of d[] in u64s */ | |
2102 | ||
2103 | __u8 encrypted_start[0]; | |
2104 | ||
2abe5420 KO |
2105 | __le16 _read_clock; /* no longer used */ |
2106 | __le16 _write_clock; | |
1c6fdbd8 KO |
2107 | |
2108 | /* Sequence number of oldest dirty journal entry */ | |
2109 | __le64 last_seq; | |
2110 | ||
2111 | ||
2112 | union { | |
2113 | struct jset_entry start[0]; | |
2114 | __u64 _data[0]; | |
2115 | }; | |
fd0c7679 | 2116 | } __packed __aligned(8); |
1c6fdbd8 KO |
2117 | |
2118 | LE32_BITMASK(JSET_CSUM_TYPE, struct jset, flags, 0, 4); | |
2119 | LE32_BITMASK(JSET_BIG_ENDIAN, struct jset, flags, 4, 5); | |
adbcada4 | 2120 | LE32_BITMASK(JSET_NO_FLUSH, struct jset, flags, 5, 6); |
1c6fdbd8 | 2121 | |
8b335bae | 2122 | #define BCH_JOURNAL_BUCKETS_MIN 8 |
1c6fdbd8 KO |
2123 | |
2124 | /* Btree: */ | |
2125 | ||
41f8b09e | 2126 | #define BCH_BTREE_IDS() \ |
a8c752bb KO |
2127 | x(extents, 0) \ |
2128 | x(inodes, 1) \ | |
2129 | x(dirents, 2) \ | |
2130 | x(xattrs, 3) \ | |
2131 | x(alloc, 4) \ | |
2132 | x(quotas, 5) \ | |
2133 | x(stripes, 6) \ | |
2134 | x(reflink, 7) \ | |
2135 | x(subvolumes, 8) \ | |
2136 | x(snapshots, 9) \ | |
2137 | x(lru, 10) \ | |
2138 | x(freespace, 11) \ | |
2139 | x(need_discard, 12) \ | |
5250b74d | 2140 | x(backpointers, 13) \ |
1c59b483 KO |
2141 | x(bucket_gens, 14) \ |
2142 | x(snapshot_trees, 15) | |
1c6fdbd8 KO |
2143 | |
2144 | enum btree_id { | |
41f8b09e | 2145 | #define x(kwd, val) BTREE_ID_##kwd = val, |
26609b61 KO |
2146 | BCH_BTREE_IDS() |
2147 | #undef x | |
1c6fdbd8 KO |
2148 | BTREE_ID_NR |
2149 | }; | |
2150 | ||
1c6fdbd8 KO |
2151 | #define BTREE_MAX_DEPTH 4U |
2152 | ||
2153 | /* Btree nodes */ | |
2154 | ||
1c6fdbd8 KO |
2155 | /* |
2156 | * Btree nodes | |
2157 | * | |
2158 | * On disk a btree node is a list/log of these; within each set the keys are | |
2159 | * sorted | |
2160 | */ | |
2161 | struct bset { | |
2162 | __le64 seq; | |
2163 | ||
2164 | /* | |
2165 | * Highest journal entry this bset contains keys for. | |
2166 | * If on recovery we don't see that journal entry, this bset is ignored: | |
2167 | * this allows us to preserve the order of all index updates after a | |
2168 | * crash, since the journal records a total order of all index updates | |
2169 | * and anything that didn't make it to the journal doesn't get used. | |
2170 | */ | |
2171 | __le64 journal_seq; | |
2172 | ||
2173 | __le32 flags; | |
2174 | __le16 version; | |
2175 | __le16 u64s; /* count of d[] in u64s */ | |
2176 | ||
2177 | union { | |
2178 | struct bkey_packed start[0]; | |
2179 | __u64 _data[0]; | |
2180 | }; | |
fd0c7679 | 2181 | } __packed __aligned(8); |
1c6fdbd8 KO |
2182 | |
2183 | LE32_BITMASK(BSET_CSUM_TYPE, struct bset, flags, 0, 4); | |
2184 | ||
2185 | LE32_BITMASK(BSET_BIG_ENDIAN, struct bset, flags, 4, 5); | |
2186 | LE32_BITMASK(BSET_SEPARATE_WHITEOUTS, | |
2187 | struct bset, flags, 5, 6); | |
2188 | ||
e719fc34 KO |
2189 | /* Sector offset within the btree node: */ |
2190 | LE32_BITMASK(BSET_OFFSET, struct bset, flags, 16, 32); | |
2191 | ||
1c6fdbd8 KO |
2192 | struct btree_node { |
2193 | struct bch_csum csum; | |
2194 | __le64 magic; | |
2195 | ||
2196 | /* this flags field is encrypted, unlike bset->flags: */ | |
2197 | __le64 flags; | |
2198 | ||
2199 | /* Closed interval: */ | |
2200 | struct bpos min_key; | |
2201 | struct bpos max_key; | |
e751c01a | 2202 | struct bch_extent_ptr _ptr; /* not used anymore */ |
1c6fdbd8 KO |
2203 | struct bkey_format format; |
2204 | ||
2205 | union { | |
2206 | struct bset keys; | |
2207 | struct { | |
2208 | __u8 pad[22]; | |
2209 | __le16 u64s; | |
2210 | __u64 _data[0]; | |
2211 | ||
2212 | }; | |
2213 | }; | |
fd0c7679 | 2214 | } __packed __aligned(8); |
1c6fdbd8 | 2215 | |
4e1430a7 | 2216 | LE64_BITMASK(BTREE_NODE_ID_LO, struct btree_node, flags, 0, 4); |
1c6fdbd8 | 2217 | LE64_BITMASK(BTREE_NODE_LEVEL, struct btree_node, flags, 4, 8); |
bcd6f3e0 KO |
2218 | LE64_BITMASK(BTREE_NODE_NEW_EXTENT_OVERWRITE, |
2219 | struct btree_node, flags, 8, 9); | |
4e1430a7 KO |
2220 | LE64_BITMASK(BTREE_NODE_ID_HI, struct btree_node, flags, 9, 25); |
2221 | /* 25-32 unused */ | |
1c6fdbd8 KO |
2222 | LE64_BITMASK(BTREE_NODE_SEQ, struct btree_node, flags, 32, 64); |
2223 | ||
4e1430a7 KO |
2224 | static inline __u64 BTREE_NODE_ID(struct btree_node *n) |
2225 | { | |
2226 | return BTREE_NODE_ID_LO(n) | (BTREE_NODE_ID_HI(n) << 4); | |
2227 | } | |
2228 | ||
2229 | static inline void SET_BTREE_NODE_ID(struct btree_node *n, u64 v) | |
2230 | { | |
2231 | SET_BTREE_NODE_ID_LO(n, v); | |
2232 | SET_BTREE_NODE_ID_HI(n, v >> 4); | |
2233 | } | |
2234 | ||
1c6fdbd8 KO |
2235 | struct btree_node_entry { |
2236 | struct bch_csum csum; | |
2237 | ||
2238 | union { | |
2239 | struct bset keys; | |
2240 | struct { | |
2241 | __u8 pad[22]; | |
2242 | __le16 u64s; | |
2243 | __u64 _data[0]; | |
1c6fdbd8 KO |
2244 | }; |
2245 | }; | |
fd0c7679 | 2246 | } __packed __aligned(8); |
1c6fdbd8 KO |
2247 | |
2248 | #endif /* _BCACHEFS_FORMAT_H */ |