1/* SPDX-License-Identifier: GPL-2.0 WITH Linux-syscall-note */
2#ifndef _LINUX_BCACHE_H
3#define _LINUX_BCACHE_H
4
5/*
6 * Bcache on disk data structures
7 */
8
9#include <linux/types.h>
10
11#define BITMASK(name, type, field, offset, size)		\
12static inline __u64 name(const type *k)				\
13{ return (k->field >> offset) & ~(~0ULL << size); }		\
14								\
15static inline void SET_##name(type *k, __u64 v)			\
16{								\
17	k->field &= ~(~(~0ULL << size) << offset);		\
18	k->field |= (v & ~(~0ULL << size)) << offset;		\
19}
20
21/* Btree keys - all units are in sectors */
22
23struct bkey {
24	__u64	high;
25	__u64	low;
26	__u64	ptr[];
27};
28
29#define KEY_FIELD(name, field, offset, size)				\
30	BITMASK(name, struct bkey, field, offset, size)
31
32#define PTR_FIELD(name, offset, size)					\
33static inline __u64 name(const struct bkey *k, unsigned int i)		\
34{ return (k->ptr[i] >> offset) & ~(~0ULL << size); }			\
35									\
36static inline void SET_##name(struct bkey *k, unsigned int i, __u64 v)	\
37{									\
38	k->ptr[i] &= ~(~(~0ULL << size) << offset);			\
39	k->ptr[i] |= (v & ~(~0ULL << size)) << offset;			\
40}
41
42#define KEY_SIZE_BITS		16
43#define KEY_MAX_U64S		8
44
45KEY_FIELD(KEY_PTRS,	high, 60, 3)
46KEY_FIELD(HEADER_SIZE,	high, 58, 2)
47KEY_FIELD(KEY_CSUM,	high, 56, 2)
48KEY_FIELD(KEY_PINNED,	high, 55, 1)
49KEY_FIELD(KEY_DIRTY,	high, 36, 1)
50
51KEY_FIELD(KEY_SIZE,	high, 20, KEY_SIZE_BITS)
52KEY_FIELD(KEY_INODE,	high, 0,  20)
53
54/* Next time I change the on disk format, KEY_OFFSET() won't be 64 bits */
55
56static inline __u64 KEY_OFFSET(const struct bkey *k)
57{
58	return k->low;
59}
60
61static inline void SET_KEY_OFFSET(struct bkey *k, __u64 v)
62{
63	k->low = v;
64}
65
66/*
67 * The high bit being set is a relic from when we used it to do binary
68 * searches - it told you where a key started. It's not used anymore,
69 * and can probably be safely dropped.
70 */
71#define KEY(inode, offset, size)					\
72((struct bkey) {							\
73	.high = (1ULL << 63) | ((__u64) (size) << 20) | (inode),	\
74	.low = (offset)							\
75})
76
77#define ZERO_KEY			KEY(0, 0, 0)
78
79#define MAX_KEY_INODE			(~(~0 << 20))
80#define MAX_KEY_OFFSET			(~0ULL >> 1)
81#define MAX_KEY				KEY(MAX_KEY_INODE, MAX_KEY_OFFSET, 0)
82
83#define KEY_START(k)			(KEY_OFFSET(k) - KEY_SIZE(k))
84#define START_KEY(k)			KEY(KEY_INODE(k), KEY_START(k), 0)
85
86#define PTR_DEV_BITS			12
87
88PTR_FIELD(PTR_DEV,			51, PTR_DEV_BITS)
89PTR_FIELD(PTR_OFFSET,			8,  43)
90PTR_FIELD(PTR_GEN,			0,  8)
91
92#define PTR_CHECK_DEV			((1 << PTR_DEV_BITS) - 1)
93
94#define MAKE_PTR(gen, offset, dev)					\
95	((((__u64) dev) << 51) | ((__u64) offset) << 8 | gen)
96
97/* Bkey utility code */
98
99static inline unsigned long bkey_u64s(const struct bkey *k)
100{
101	return (sizeof(struct bkey) / sizeof(__u64)) + KEY_PTRS(k);
102}
103
104static inline unsigned long bkey_bytes(const struct bkey *k)
105{
106	return bkey_u64s(k) * sizeof(__u64);
107}
108
109#define bkey_copy(_dest, _src)	memcpy(_dest, _src, bkey_bytes(_src))
110
111static inline void bkey_copy_key(struct bkey *dest, const struct bkey *src)
112{
113	SET_KEY_INODE(dest, KEY_INODE(src));
114	SET_KEY_OFFSET(dest, KEY_OFFSET(src));
115}
116
117static inline struct bkey *bkey_next(const struct bkey *k)
118{
119	__u64 *d = (void *) k;
120
121	return (struct bkey *) (d + bkey_u64s(k));
122}
123
124static inline struct bkey *bkey_idx(const struct bkey *k, unsigned int nr_keys)
125{
126	__u64 *d = (void *) k;
127
128	return (struct bkey *) (d + nr_keys);
129}
130/* Enough for a key with 6 pointers */
131#define BKEY_PAD		8
132
133#define BKEY_PADDED(key)					\
134	union { struct bkey key; __u64 key ## _pad[BKEY_PAD]; }
135
136/* Superblock */
137
138/* Version 0: Cache device
139 * Version 1: Backing device
140 * Version 2: Seed pointer into btree node checksum
141 * Version 3: Cache device with new UUID format
142 * Version 4: Backing device with data offset
143 */
144#define BCACHE_SB_VERSION_CDEV			0
145#define BCACHE_SB_VERSION_BDEV			1
146#define BCACHE_SB_VERSION_CDEV_WITH_UUID	3
147#define BCACHE_SB_VERSION_BDEV_WITH_OFFSET	4
148#define BCACHE_SB_VERSION_CDEV_WITH_FEATURES	5
149#define BCACHE_SB_VERSION_BDEV_WITH_FEATURES	6
150#define BCACHE_SB_MAX_VERSION			6
151
152#define SB_SECTOR			8
153#define SB_OFFSET			(SB_SECTOR << SECTOR_SHIFT)
154#define SB_SIZE				4096
155#define SB_LABEL_SIZE			32
156#define SB_JOURNAL_BUCKETS		256U
157/* SB_JOURNAL_BUCKETS must be divisible by BITS_PER_LONG */
158#define MAX_CACHES_PER_SET		8
159
160#define BDEV_DATA_START_DEFAULT		16	/* sectors */
161
162struct cache_sb_disk {
163	__le64			csum;
164	__le64			offset;	/* sector where this sb was written */
165	__le64			version;
166
167	__u8			magic[16];
168
169	__u8			uuid[16];
170	union {
171		__u8		set_uuid[16];
172		__le64		set_magic;
173	};
174	__u8			label[SB_LABEL_SIZE];
175
176	__le64			flags;
177	__le64			seq;
178
179	__le64			feature_compat;
180	__le64			feature_incompat;
181	__le64			feature_ro_compat;
182
183	__le64			pad[5];
184
185	union {
186	struct {
187		/* Cache devices */
188		__le64		nbuckets;	/* device size */
189
190		__le16		block_size;	/* sectors */
191		__le16		bucket_size;	/* sectors */
192
193		__le16		nr_in_set;
194		__le16		nr_this_dev;
195	};
196	struct {
197		/* Backing devices */
198		__le64		data_offset;
199
200		/*
201		 * block_size from the cache device section is still used by
202		 * backing devices, so don't add anything here until we fix
203		 * things to not need it for backing devices anymore
204		 */
205	};
206	};
207
208	__le32			last_mount;	/* time overflow in y2106 */
209
210	__le16			first_bucket;
211	union {
212		__le16		njournal_buckets;
213		__le16		keys;
214	};
215	__le64			d[SB_JOURNAL_BUCKETS];	/* journal buckets */
216	__le16			obso_bucket_size_hi;	/* obsoleted */
217};
218
219/*
220 * This is for in-memory bcache super block.
221 * NOTE: cache_sb is NOT exactly mapping to cache_sb_disk, the member
222 *       size, ordering and even whole struct size may be different
223 *       from cache_sb_disk.
224 */
225struct cache_sb {
226	__u64			offset;	/* sector where this sb was written */
227	__u64			version;
228
229	__u8			magic[16];
230
231	__u8			uuid[16];
232	union {
233		__u8		set_uuid[16];
234		__u64		set_magic;
235	};
236	__u8			label[SB_LABEL_SIZE];
237
238	__u64			flags;
239	__u64			seq;
240
241	__u64			feature_compat;
242	__u64			feature_incompat;
243	__u64			feature_ro_compat;
244
245	union {
246	struct {
247		/* Cache devices */
248		__u64		nbuckets;	/* device size */
249
250		__u16		block_size;	/* sectors */
251		__u16		nr_in_set;
252		__u16		nr_this_dev;
253		__u32		bucket_size;	/* sectors */
254	};
255	struct {
256		/* Backing devices */
257		__u64		data_offset;
258
259		/*
260		 * block_size from the cache device section is still used by
261		 * backing devices, so don't add anything here until we fix
262		 * things to not need it for backing devices anymore
263		 */
264	};
265	};
266
267	__u32			last_mount;	/* time overflow in y2106 */
268
269	__u16			first_bucket;
270	union {
271		__u16		njournal_buckets;
272		__u16		keys;
273	};
274	__u64			d[SB_JOURNAL_BUCKETS];	/* journal buckets */
275};
276
277static inline _Bool SB_IS_BDEV(const struct cache_sb *sb)
278{
279	return sb->version == BCACHE_SB_VERSION_BDEV
280		|| sb->version == BCACHE_SB_VERSION_BDEV_WITH_OFFSET
281		|| sb->version == BCACHE_SB_VERSION_BDEV_WITH_FEATURES;
282}
283
284BITMASK(CACHE_SYNC,			struct cache_sb, flags, 0, 1);
285BITMASK(CACHE_DISCARD,			struct cache_sb, flags, 1, 1);
286BITMASK(CACHE_REPLACEMENT,		struct cache_sb, flags, 2, 3);
287#define CACHE_REPLACEMENT_LRU		0U
288#define CACHE_REPLACEMENT_FIFO		1U
289#define CACHE_REPLACEMENT_RANDOM	2U
290
291BITMASK(BDEV_CACHE_MODE,		struct cache_sb, flags, 0, 4);
292#define CACHE_MODE_WRITETHROUGH		0U
293#define CACHE_MODE_WRITEBACK		1U
294#define CACHE_MODE_WRITEAROUND		2U
295#define CACHE_MODE_NONE			3U
296BITMASK(BDEV_STATE,			struct cache_sb, flags, 61, 2);
297#define BDEV_STATE_NONE			0U
298#define BDEV_STATE_CLEAN		1U
299#define BDEV_STATE_DIRTY		2U
300#define BDEV_STATE_STALE		3U
301
302/*
303 * Magic numbers
304 *
305 * The various other data structures have their own magic numbers, which are
306 * xored with the first part of the cache set's UUID
307 */
308
309#define JSET_MAGIC			0x245235c1a3625032ULL
310#define PSET_MAGIC			0x6750e15f87337f91ULL
311#define BSET_MAGIC			0x90135c78b99e07f5ULL
312
313static inline __u64 jset_magic(struct cache_sb *sb)
314{
315	return sb->set_magic ^ JSET_MAGIC;
316}
317
318static inline __u64 pset_magic(struct cache_sb *sb)
319{
320	return sb->set_magic ^ PSET_MAGIC;
321}
322
323static inline __u64 bset_magic(struct cache_sb *sb)
324{
325	return sb->set_magic ^ BSET_MAGIC;
326}
327
328/*
329 * Journal
330 *
331 * On disk format for a journal entry:
332 * seq is monotonically increasing; every journal entry has its own unique
333 * sequence number.
334 *
335 * last_seq is the oldest journal entry that still has keys the btree hasn't
336 * flushed to disk yet.
337 *
338 * version is for on disk format changes.
339 */
340
341#define BCACHE_JSET_VERSION_UUIDv1	1
342#define BCACHE_JSET_VERSION_UUID	1	/* Always latest UUID format */
343#define BCACHE_JSET_VERSION		1
344
345struct jset {
346	__u64			csum;
347	__u64			magic;
348	__u64			seq;
349	__u32			version;
350	__u32			keys;
351
352	__u64			last_seq;
353
354	BKEY_PADDED(uuid_bucket);
355	BKEY_PADDED(btree_root);
356	__u16			btree_level;
357	__u16			pad[3];
358
359	__u64			prio_bucket[MAX_CACHES_PER_SET];
360
361	union {
362		struct bkey	start[0];
363		__u64		d[0];
364	};
365};
366
367/* Bucket prios/gens */
368
369struct prio_set {
370	__u64			csum;
371	__u64			magic;
372	__u64			seq;
373	__u32			version;
374	__u32			pad;
375
376	__u64			next_bucket;
377
378	struct bucket_disk {
379		__u16		prio;
380		__u8		gen;
381	} __attribute((packed)) data[];
382};
383
384/* UUIDS - per backing device/flash only volume metadata */
385
386struct uuid_entry {
387	union {
388		struct {
389			__u8	uuid[16];
390			__u8	label[32];
391			__u32	first_reg; /* time overflow in y2106 */
392			__u32	last_reg;
393			__u32	invalidated;
394
395			__u32	flags;
396			/* Size of flash only volumes */
397			__u64	sectors;
398		};
399
400		__u8		pad[128];
401	};
402};
403
404BITMASK(UUID_FLASH_ONLY,	struct uuid_entry, flags, 0, 1);
405
406/* Btree nodes */
407
408/* Version 1: Seed pointer into btree node checksum
409 */
410#define BCACHE_BSET_CSUM		1
411#define BCACHE_BSET_VERSION		1
412
413/*
414 * Btree nodes
415 *
416 * On disk a btree node is a list/log of these; within each set the keys are
417 * sorted
418 */
419struct bset {
420	__u64			csum;
421	__u64			magic;
422	__u64			seq;
423	__u32			version;
424	__u32			keys;
425
426	union {
427		struct bkey	start[0];
428		__u64		d[0];
429	};
430};
431
432/* OBSOLETE */
433
434/* UUIDS - per backing device/flash only volume metadata */
435
436struct uuid_entry_v0 {
437	__u8		uuid[16];
438	__u8		label[32];
439	__u32		first_reg;
440	__u32		last_reg;
441	__u32		invalidated;
442	__u32		pad;
443};
444
445#endif /* _LINUX_BCACHE_H */
446