xref: /linux/fs/bcachefs/super-io.c (revision 3c471b65889aa22d6ba4f8e3d2e5006bb70fdb58)
1 // SPDX-License-Identifier: GPL-2.0
2 
3 #include "bcachefs.h"
4 #include "checksum.h"
5 #include "counters.h"
6 #include "disk_groups.h"
7 #include "ec.h"
8 #include "error.h"
9 #include "journal.h"
10 #include "journal_sb.h"
11 #include "journal_seq_blacklist.h"
12 #include "recovery.h"
13 #include "replicas.h"
14 #include "quota.h"
15 #include "sb-clean.h"
16 #include "sb-downgrade.h"
17 #include "sb-errors.h"
18 #include "sb-members.h"
19 #include "super-io.h"
20 #include "super.h"
21 #include "trace.h"
22 #include "vstructs.h"
23 
24 #include <linux/backing-dev.h>
25 #include <linux/sort.h>
26 
27 static const struct blk_holder_ops bch2_sb_handle_bdev_ops = {
28 };
29 
30 struct bch2_metadata_version {
31 	u16		version;
32 	const char	*name;
33 	u64		recovery_passes;
34 };
35 
36 static const struct bch2_metadata_version bch2_metadata_versions[] = {
37 #define x(n, v, _recovery_passes) {		\
38 	.version = v,				\
39 	.name = #n,				\
40 	.recovery_passes = _recovery_passes,	\
41 },
42 	BCH_METADATA_VERSIONS()
43 #undef x
44 };
45 
46 void bch2_version_to_text(struct printbuf *out, unsigned v)
47 {
48 	const char *str = "(unknown version)";
49 
50 	for (unsigned i = 0; i < ARRAY_SIZE(bch2_metadata_versions); i++)
51 		if (bch2_metadata_versions[i].version == v) {
52 			str = bch2_metadata_versions[i].name;
53 			break;
54 		}
55 
56 	prt_printf(out, "%u.%u: %s", BCH_VERSION_MAJOR(v), BCH_VERSION_MINOR(v), str);
57 }
58 
59 unsigned bch2_latest_compatible_version(unsigned v)
60 {
61 	if (!BCH_VERSION_MAJOR(v))
62 		return v;
63 
64 	for (unsigned i = 0; i < ARRAY_SIZE(bch2_metadata_versions); i++)
65 		if (bch2_metadata_versions[i].version > v &&
66 		    BCH_VERSION_MAJOR(bch2_metadata_versions[i].version) ==
67 		    BCH_VERSION_MAJOR(v))
68 			v = bch2_metadata_versions[i].version;
69 
70 	return v;
71 }
72 
73 u64 bch2_upgrade_recovery_passes(struct bch_fs *c,
74 				 unsigned old_version,
75 				 unsigned new_version)
76 {
77 	u64 ret = 0;
78 
79 	for (const struct bch2_metadata_version *i = bch2_metadata_versions;
80 	     i < bch2_metadata_versions + ARRAY_SIZE(bch2_metadata_versions);
81 	     i++)
82 		if (i->version > old_version && i->version <= new_version) {
83 			if (i->recovery_passes & RECOVERY_PASS_ALL_FSCK)
84 				ret |= bch2_fsck_recovery_passes();
85 			ret |= i->recovery_passes;
86 		}
87 
88 	return ret &= ~RECOVERY_PASS_ALL_FSCK;
89 }
90 
91 const char * const bch2_sb_fields[] = {
92 #define x(name, nr)	#name,
93 	BCH_SB_FIELDS()
94 #undef x
95 	NULL
96 };
97 
98 static int bch2_sb_field_validate(struct bch_sb *, struct bch_sb_field *,
99 				  struct printbuf *);
100 
101 struct bch_sb_field *bch2_sb_field_get_id(struct bch_sb *sb,
102 				      enum bch_sb_field_type type)
103 {
104 	struct bch_sb_field *f;
105 
106 	/* XXX: need locking around superblock to access optional fields */
107 
108 	vstruct_for_each(sb, f)
109 		if (le32_to_cpu(f->type) == type)
110 			return f;
111 	return NULL;
112 }
113 
114 static struct bch_sb_field *__bch2_sb_field_resize(struct bch_sb_handle *sb,
115 						   struct bch_sb_field *f,
116 						   unsigned u64s)
117 {
118 	unsigned old_u64s = f ? le32_to_cpu(f->u64s) : 0;
119 	unsigned sb_u64s = le32_to_cpu(sb->sb->u64s) + u64s - old_u64s;
120 
121 	BUG_ON(__vstruct_bytes(struct bch_sb, sb_u64s) > sb->buffer_size);
122 
123 	if (!f && !u64s) {
124 		/* nothing to do: */
125 	} else if (!f) {
126 		f = vstruct_last(sb->sb);
127 		memset(f, 0, sizeof(u64) * u64s);
128 		f->u64s = cpu_to_le32(u64s);
129 		f->type = 0;
130 	} else {
131 		void *src, *dst;
132 
133 		src = vstruct_end(f);
134 
135 		if (u64s) {
136 			f->u64s = cpu_to_le32(u64s);
137 			dst = vstruct_end(f);
138 		} else {
139 			dst = f;
140 		}
141 
142 		memmove(dst, src, vstruct_end(sb->sb) - src);
143 
144 		if (dst > src)
145 			memset(src, 0, dst - src);
146 	}
147 
148 	sb->sb->u64s = cpu_to_le32(sb_u64s);
149 
150 	return u64s ? f : NULL;
151 }
152 
153 void bch2_sb_field_delete(struct bch_sb_handle *sb,
154 			  enum bch_sb_field_type type)
155 {
156 	struct bch_sb_field *f = bch2_sb_field_get_id(sb->sb, type);
157 
158 	if (f)
159 		__bch2_sb_field_resize(sb, f, 0);
160 }
161 
162 /* Superblock realloc/free: */
163 
164 void bch2_free_super(struct bch_sb_handle *sb)
165 {
166 	kfree(sb->bio);
167 	if (!IS_ERR_OR_NULL(sb->bdev))
168 		blkdev_put(sb->bdev, sb->holder);
169 	kfree(sb->holder);
170 	kfree(sb->sb_name);
171 
172 	kfree(sb->sb);
173 	memset(sb, 0, sizeof(*sb));
174 }
175 
176 int bch2_sb_realloc(struct bch_sb_handle *sb, unsigned u64s)
177 {
178 	size_t new_bytes = __vstruct_bytes(struct bch_sb, u64s);
179 	size_t new_buffer_size;
180 	struct bch_sb *new_sb;
181 	struct bio *bio;
182 
183 	if (sb->bdev)
184 		new_bytes = max_t(size_t, new_bytes, bdev_logical_block_size(sb->bdev));
185 
186 	new_buffer_size = roundup_pow_of_two(new_bytes);
187 
188 	if (sb->sb && sb->buffer_size >= new_buffer_size)
189 		return 0;
190 
191 	if (sb->sb && sb->have_layout) {
192 		u64 max_bytes = 512 << sb->sb->layout.sb_max_size_bits;
193 
194 		if (new_bytes > max_bytes) {
195 			pr_err("%pg: superblock too big: want %zu but have %llu",
196 			       sb->bdev, new_bytes, max_bytes);
197 			return -BCH_ERR_ENOSPC_sb;
198 		}
199 	}
200 
201 	if (sb->buffer_size >= new_buffer_size && sb->sb)
202 		return 0;
203 
204 	if (dynamic_fault("bcachefs:add:super_realloc"))
205 		return -BCH_ERR_ENOMEM_sb_realloc_injected;
206 
207 	new_sb = krealloc(sb->sb, new_buffer_size, GFP_NOFS|__GFP_ZERO);
208 	if (!new_sb)
209 		return -BCH_ERR_ENOMEM_sb_buf_realloc;
210 
211 	sb->sb = new_sb;
212 
213 	if (sb->have_bio) {
214 		unsigned nr_bvecs = buf_pages(sb->sb, new_buffer_size);
215 
216 		bio = bio_kmalloc(nr_bvecs, GFP_KERNEL);
217 		if (!bio)
218 			return -BCH_ERR_ENOMEM_sb_bio_realloc;
219 
220 		bio_init(bio, NULL, bio->bi_inline_vecs, nr_bvecs, 0);
221 
222 		kfree(sb->bio);
223 		sb->bio = bio;
224 	}
225 
226 	sb->buffer_size = new_buffer_size;
227 
228 	return 0;
229 }
230 
231 struct bch_sb_field *bch2_sb_field_resize_id(struct bch_sb_handle *sb,
232 					  enum bch_sb_field_type type,
233 					  unsigned u64s)
234 {
235 	struct bch_sb_field *f = bch2_sb_field_get_id(sb->sb, type);
236 	ssize_t old_u64s = f ? le32_to_cpu(f->u64s) : 0;
237 	ssize_t d = -old_u64s + u64s;
238 
239 	if (bch2_sb_realloc(sb, le32_to_cpu(sb->sb->u64s) + d))
240 		return NULL;
241 
242 	if (sb->fs_sb) {
243 		struct bch_fs *c = container_of(sb, struct bch_fs, disk_sb);
244 		struct bch_dev *ca;
245 		unsigned i;
246 
247 		lockdep_assert_held(&c->sb_lock);
248 
249 		/* XXX: we're not checking that offline device have enough space */
250 
251 		for_each_online_member(ca, c, i) {
252 			struct bch_sb_handle *dev_sb = &ca->disk_sb;
253 
254 			if (bch2_sb_realloc(dev_sb, le32_to_cpu(dev_sb->sb->u64s) + d)) {
255 				percpu_ref_put(&ca->ref);
256 				return NULL;
257 			}
258 		}
259 	}
260 
261 	f = bch2_sb_field_get_id(sb->sb, type);
262 	f = __bch2_sb_field_resize(sb, f, u64s);
263 	if (f)
264 		f->type = cpu_to_le32(type);
265 	return f;
266 }
267 
268 struct bch_sb_field *bch2_sb_field_get_minsize_id(struct bch_sb_handle *sb,
269 						  enum bch_sb_field_type type,
270 						  unsigned u64s)
271 {
272 	struct bch_sb_field *f = bch2_sb_field_get_id(sb->sb, type);
273 
274 	if (!f || le32_to_cpu(f->u64s) < u64s)
275 		f = bch2_sb_field_resize_id(sb, type, u64s);
276 	return f;
277 }
278 
279 /* Superblock validate: */
280 
281 static int validate_sb_layout(struct bch_sb_layout *layout, struct printbuf *out)
282 {
283 	u64 offset, prev_offset, max_sectors;
284 	unsigned i;
285 
286 	BUILD_BUG_ON(sizeof(struct bch_sb_layout) != 512);
287 
288 	if (!uuid_equal(&layout->magic, &BCACHE_MAGIC) &&
289 	    !uuid_equal(&layout->magic, &BCHFS_MAGIC)) {
290 		prt_printf(out, "Not a bcachefs superblock layout");
291 		return -BCH_ERR_invalid_sb_layout;
292 	}
293 
294 	if (layout->layout_type != 0) {
295 		prt_printf(out, "Invalid superblock layout type %u",
296 		       layout->layout_type);
297 		return -BCH_ERR_invalid_sb_layout_type;
298 	}
299 
300 	if (!layout->nr_superblocks) {
301 		prt_printf(out, "Invalid superblock layout: no superblocks");
302 		return -BCH_ERR_invalid_sb_layout_nr_superblocks;
303 	}
304 
305 	if (layout->nr_superblocks > ARRAY_SIZE(layout->sb_offset)) {
306 		prt_printf(out, "Invalid superblock layout: too many superblocks");
307 		return -BCH_ERR_invalid_sb_layout_nr_superblocks;
308 	}
309 
310 	max_sectors = 1 << layout->sb_max_size_bits;
311 
312 	prev_offset = le64_to_cpu(layout->sb_offset[0]);
313 
314 	for (i = 1; i < layout->nr_superblocks; i++) {
315 		offset = le64_to_cpu(layout->sb_offset[i]);
316 
317 		if (offset < prev_offset + max_sectors) {
318 			prt_printf(out, "Invalid superblock layout: superblocks overlap\n"
319 			       "  (sb %u ends at %llu next starts at %llu",
320 			       i - 1, prev_offset + max_sectors, offset);
321 			return -BCH_ERR_invalid_sb_layout_superblocks_overlap;
322 		}
323 		prev_offset = offset;
324 	}
325 
326 	return 0;
327 }
328 
329 static int bch2_sb_compatible(struct bch_sb *sb, struct printbuf *out)
330 {
331 	u16 version		= le16_to_cpu(sb->version);
332 	u16 version_min		= le16_to_cpu(sb->version_min);
333 
334 	if (!bch2_version_compatible(version)) {
335 		prt_str(out, "Unsupported superblock version ");
336 		bch2_version_to_text(out, version);
337 		prt_str(out, " (min ");
338 		bch2_version_to_text(out, bcachefs_metadata_version_min);
339 		prt_str(out, ", max ");
340 		bch2_version_to_text(out, bcachefs_metadata_version_current);
341 		prt_str(out, ")");
342 		return -BCH_ERR_invalid_sb_version;
343 	}
344 
345 	if (!bch2_version_compatible(version_min)) {
346 		prt_str(out, "Unsupported superblock version_min ");
347 		bch2_version_to_text(out, version_min);
348 		prt_str(out, " (min ");
349 		bch2_version_to_text(out, bcachefs_metadata_version_min);
350 		prt_str(out, ", max ");
351 		bch2_version_to_text(out, bcachefs_metadata_version_current);
352 		prt_str(out, ")");
353 		return -BCH_ERR_invalid_sb_version;
354 	}
355 
356 	if (version_min > version) {
357 		prt_str(out, "Bad minimum version ");
358 		bch2_version_to_text(out, version_min);
359 		prt_str(out, ", greater than version field ");
360 		bch2_version_to_text(out, version);
361 		return -BCH_ERR_invalid_sb_version;
362 	}
363 
364 	return 0;
365 }
366 
367 static int bch2_sb_validate(struct bch_sb_handle *disk_sb, struct printbuf *out,
368 			    int rw)
369 {
370 	struct bch_sb *sb = disk_sb->sb;
371 	struct bch_sb_field *f;
372 	struct bch_sb_field_members_v1 *mi;
373 	enum bch_opt_id opt_id;
374 	u16 block_size;
375 	int ret;
376 
377 	ret = bch2_sb_compatible(sb, out);
378 	if (ret)
379 		return ret;
380 
381 	if (sb->features[1] ||
382 	    (le64_to_cpu(sb->features[0]) & (~0ULL << BCH_FEATURE_NR))) {
383 		prt_printf(out, "Filesystem has incompatible features");
384 		return -BCH_ERR_invalid_sb_features;
385 	}
386 
387 	block_size = le16_to_cpu(sb->block_size);
388 
389 	if (block_size > PAGE_SECTORS) {
390 		prt_printf(out, "Block size too big (got %u, max %u)",
391 		       block_size, PAGE_SECTORS);
392 		return -BCH_ERR_invalid_sb_block_size;
393 	}
394 
395 	if (bch2_is_zero(sb->user_uuid.b, sizeof(sb->user_uuid))) {
396 		prt_printf(out, "Bad user UUID (got zeroes)");
397 		return -BCH_ERR_invalid_sb_uuid;
398 	}
399 
400 	if (bch2_is_zero(sb->uuid.b, sizeof(sb->uuid))) {
401 		prt_printf(out, "Bad internal UUID (got zeroes)");
402 		return -BCH_ERR_invalid_sb_uuid;
403 	}
404 
405 	if (!sb->nr_devices ||
406 	    sb->nr_devices > BCH_SB_MEMBERS_MAX) {
407 		prt_printf(out, "Bad number of member devices %u (max %u)",
408 		       sb->nr_devices, BCH_SB_MEMBERS_MAX);
409 		return -BCH_ERR_invalid_sb_too_many_members;
410 	}
411 
412 	if (sb->dev_idx >= sb->nr_devices) {
413 		prt_printf(out, "Bad dev_idx (got %u, nr_devices %u)",
414 		       sb->dev_idx, sb->nr_devices);
415 		return -BCH_ERR_invalid_sb_dev_idx;
416 	}
417 
418 	if (!sb->time_precision ||
419 	    le32_to_cpu(sb->time_precision) > NSEC_PER_SEC) {
420 		prt_printf(out, "Invalid time precision: %u (min 1, max %lu)",
421 		       le32_to_cpu(sb->time_precision), NSEC_PER_SEC);
422 		return -BCH_ERR_invalid_sb_time_precision;
423 	}
424 
425 	if (rw == READ) {
426 		/*
427 		 * Been seeing a bug where these are getting inexplicably
428 		 * zeroed, so we're now validating them, but we have to be
429 		 * careful not to preven people's filesystems from mounting:
430 		 */
431 		if (!BCH_SB_JOURNAL_FLUSH_DELAY(sb))
432 			SET_BCH_SB_JOURNAL_FLUSH_DELAY(sb, 1000);
433 		if (!BCH_SB_JOURNAL_RECLAIM_DELAY(sb))
434 			SET_BCH_SB_JOURNAL_RECLAIM_DELAY(sb, 1000);
435 
436 		if (!BCH_SB_VERSION_UPGRADE_COMPLETE(sb))
437 			SET_BCH_SB_VERSION_UPGRADE_COMPLETE(sb, le16_to_cpu(sb->version));
438 	}
439 
440 	for (opt_id = 0; opt_id < bch2_opts_nr; opt_id++) {
441 		const struct bch_option *opt = bch2_opt_table + opt_id;
442 
443 		if (opt->get_sb != BCH2_NO_SB_OPT) {
444 			u64 v = bch2_opt_from_sb(sb, opt_id);
445 
446 			prt_printf(out, "Invalid option ");
447 			ret = bch2_opt_validate(opt, v, out);
448 			if (ret)
449 				return ret;
450 
451 			printbuf_reset(out);
452 		}
453 	}
454 
455 	/* validate layout */
456 	ret = validate_sb_layout(&sb->layout, out);
457 	if (ret)
458 		return ret;
459 
460 	vstruct_for_each(sb, f) {
461 		if (!f->u64s) {
462 			prt_printf(out, "Invalid superblock: optional field with size 0 (type %u)",
463 			       le32_to_cpu(f->type));
464 			return -BCH_ERR_invalid_sb_field_size;
465 		}
466 
467 		if (vstruct_next(f) > vstruct_last(sb)) {
468 			prt_printf(out, "Invalid superblock: optional field extends past end of superblock (type %u)",
469 			       le32_to_cpu(f->type));
470 			return -BCH_ERR_invalid_sb_field_size;
471 		}
472 	}
473 
474 	/* members must be validated first: */
475 	mi = bch2_sb_field_get(sb, members_v1);
476 	if (!mi) {
477 		prt_printf(out, "Invalid superblock: member info area missing");
478 		return -BCH_ERR_invalid_sb_members_missing;
479 	}
480 
481 	ret = bch2_sb_field_validate(sb, &mi->field, out);
482 	if (ret)
483 		return ret;
484 
485 	vstruct_for_each(sb, f) {
486 		if (le32_to_cpu(f->type) == BCH_SB_FIELD_members_v1)
487 			continue;
488 
489 		ret = bch2_sb_field_validate(sb, f, out);
490 		if (ret)
491 			return ret;
492 	}
493 
494 	return 0;
495 }
496 
497 /* device open: */
498 
499 static unsigned long le_ulong_to_cpu(unsigned long v)
500 {
501 	return sizeof(unsigned long) == 8
502 		? le64_to_cpu(v)
503 		: le32_to_cpu(v);
504 }
505 
506 static void le_bitvector_to_cpu(unsigned long *dst, unsigned long *src, unsigned nr)
507 {
508 	BUG_ON(nr & (BITS_PER_TYPE(long) - 1));
509 
510 	for (unsigned i = 0; i < BITS_TO_LONGS(nr); i++)
511 		dst[i] = le_ulong_to_cpu(src[i]);
512 }
513 
514 static void bch2_sb_update(struct bch_fs *c)
515 {
516 	struct bch_sb *src = c->disk_sb.sb;
517 	struct bch_dev *ca;
518 	unsigned i;
519 
520 	lockdep_assert_held(&c->sb_lock);
521 
522 	c->sb.uuid		= src->uuid;
523 	c->sb.user_uuid		= src->user_uuid;
524 	c->sb.version		= le16_to_cpu(src->version);
525 	c->sb.version_min	= le16_to_cpu(src->version_min);
526 	c->sb.version_upgrade_complete = BCH_SB_VERSION_UPGRADE_COMPLETE(src);
527 	c->sb.nr_devices	= src->nr_devices;
528 	c->sb.clean		= BCH_SB_CLEAN(src);
529 	c->sb.encryption_type	= BCH_SB_ENCRYPTION_TYPE(src);
530 
531 	c->sb.nsec_per_time_unit = le32_to_cpu(src->time_precision);
532 	c->sb.time_units_per_sec = NSEC_PER_SEC / c->sb.nsec_per_time_unit;
533 
534 	/* XXX this is wrong, we need a 96 or 128 bit integer type */
535 	c->sb.time_base_lo	= div_u64(le64_to_cpu(src->time_base_lo),
536 					  c->sb.nsec_per_time_unit);
537 	c->sb.time_base_hi	= le32_to_cpu(src->time_base_hi);
538 
539 	c->sb.features		= le64_to_cpu(src->features[0]);
540 	c->sb.compat		= le64_to_cpu(src->compat[0]);
541 
542 	memset(c->sb.errors_silent, 0, sizeof(c->sb.errors_silent));
543 
544 	struct bch_sb_field_ext *ext = bch2_sb_field_get(src, ext);
545 	if (ext)
546 		le_bitvector_to_cpu(c->sb.errors_silent, (void *) ext->errors_silent,
547 				    sizeof(c->sb.errors_silent) * 8);
548 
549 	for_each_member_device(ca, c, i) {
550 		struct bch_member m = bch2_sb_member_get(src, ca->dev_idx);
551 		ca->mi = bch2_mi_to_cpu(&m);
552 	}
553 }
554 
555 static int __copy_super(struct bch_sb_handle *dst_handle, struct bch_sb *src)
556 {
557 	struct bch_sb_field *src_f, *dst_f;
558 	struct bch_sb *dst = dst_handle->sb;
559 	unsigned i;
560 
561 	dst->version		= src->version;
562 	dst->version_min	= src->version_min;
563 	dst->seq		= src->seq;
564 	dst->uuid		= src->uuid;
565 	dst->user_uuid		= src->user_uuid;
566 	memcpy(dst->label,	src->label, sizeof(dst->label));
567 
568 	dst->block_size		= src->block_size;
569 	dst->nr_devices		= src->nr_devices;
570 
571 	dst->time_base_lo	= src->time_base_lo;
572 	dst->time_base_hi	= src->time_base_hi;
573 	dst->time_precision	= src->time_precision;
574 
575 	memcpy(dst->flags,	src->flags,	sizeof(dst->flags));
576 	memcpy(dst->features,	src->features,	sizeof(dst->features));
577 	memcpy(dst->compat,	src->compat,	sizeof(dst->compat));
578 
579 	for (i = 0; i < BCH_SB_FIELD_NR; i++) {
580 		int d;
581 
582 		if ((1U << i) & BCH_SINGLE_DEVICE_SB_FIELDS)
583 			continue;
584 
585 		src_f = bch2_sb_field_get_id(src, i);
586 		dst_f = bch2_sb_field_get_id(dst, i);
587 
588 		d = (src_f ? le32_to_cpu(src_f->u64s) : 0) -
589 		    (dst_f ? le32_to_cpu(dst_f->u64s) : 0);
590 		if (d > 0) {
591 			int ret = bch2_sb_realloc(dst_handle,
592 					le32_to_cpu(dst_handle->sb->u64s) + d);
593 
594 			if (ret)
595 				return ret;
596 
597 			dst = dst_handle->sb;
598 			dst_f = bch2_sb_field_get_id(dst, i);
599 		}
600 
601 		dst_f = __bch2_sb_field_resize(dst_handle, dst_f,
602 				src_f ? le32_to_cpu(src_f->u64s) : 0);
603 
604 		if (src_f)
605 			memcpy(dst_f, src_f, vstruct_bytes(src_f));
606 	}
607 
608 	return 0;
609 }
610 
611 int bch2_sb_to_fs(struct bch_fs *c, struct bch_sb *src)
612 {
613 	int ret;
614 
615 	lockdep_assert_held(&c->sb_lock);
616 
617 	ret =   bch2_sb_realloc(&c->disk_sb, 0) ?:
618 		__copy_super(&c->disk_sb, src) ?:
619 		bch2_sb_replicas_to_cpu_replicas(c) ?:
620 		bch2_sb_disk_groups_to_cpu(c);
621 	if (ret)
622 		return ret;
623 
624 	bch2_sb_update(c);
625 	return 0;
626 }
627 
628 int bch2_sb_from_fs(struct bch_fs *c, struct bch_dev *ca)
629 {
630 	return __copy_super(&ca->disk_sb, c->disk_sb.sb);
631 }
632 
633 /* read superblock: */
634 
635 static int read_one_super(struct bch_sb_handle *sb, u64 offset, struct printbuf *err)
636 {
637 	struct bch_csum csum;
638 	size_t bytes;
639 	int ret;
640 reread:
641 	bio_reset(sb->bio, sb->bdev, REQ_OP_READ|REQ_SYNC|REQ_META);
642 	sb->bio->bi_iter.bi_sector = offset;
643 	bch2_bio_map(sb->bio, sb->sb, sb->buffer_size);
644 
645 	ret = submit_bio_wait(sb->bio);
646 	if (ret) {
647 		prt_printf(err, "IO error: %i", ret);
648 		return ret;
649 	}
650 
651 	if (!uuid_equal(&sb->sb->magic, &BCACHE_MAGIC) &&
652 	    !uuid_equal(&sb->sb->magic, &BCHFS_MAGIC)) {
653 		prt_printf(err, "Not a bcachefs superblock");
654 		return -BCH_ERR_invalid_sb_magic;
655 	}
656 
657 	ret = bch2_sb_compatible(sb->sb, err);
658 	if (ret)
659 		return ret;
660 
661 	bytes = vstruct_bytes(sb->sb);
662 
663 	if (bytes > 512 << sb->sb->layout.sb_max_size_bits) {
664 		prt_printf(err, "Invalid superblock: too big (got %zu bytes, layout max %lu)",
665 		       bytes, 512UL << sb->sb->layout.sb_max_size_bits);
666 		return -BCH_ERR_invalid_sb_too_big;
667 	}
668 
669 	if (bytes > sb->buffer_size) {
670 		ret = bch2_sb_realloc(sb, le32_to_cpu(sb->sb->u64s));
671 		if (ret)
672 			return ret;
673 		goto reread;
674 	}
675 
676 	if (BCH_SB_CSUM_TYPE(sb->sb) >= BCH_CSUM_NR) {
677 		prt_printf(err, "unknown checksum type %llu", BCH_SB_CSUM_TYPE(sb->sb));
678 		return -BCH_ERR_invalid_sb_csum_type;
679 	}
680 
681 	/* XXX: verify MACs */
682 	csum = csum_vstruct(NULL, BCH_SB_CSUM_TYPE(sb->sb),
683 			    null_nonce(), sb->sb);
684 
685 	if (bch2_crc_cmp(csum, sb->sb->csum)) {
686 		prt_printf(err, "bad checksum");
687 		return -BCH_ERR_invalid_sb_csum;
688 	}
689 
690 	sb->seq = le64_to_cpu(sb->sb->seq);
691 
692 	return 0;
693 }
694 
695 static int __bch2_read_super(const char *path, struct bch_opts *opts,
696 		    struct bch_sb_handle *sb, bool ignore_notbchfs_msg)
697 {
698 	u64 offset = opt_get(*opts, sb);
699 	struct bch_sb_layout layout;
700 	struct printbuf err = PRINTBUF;
701 	struct printbuf err2 = PRINTBUF;
702 	__le64 *i;
703 	int ret;
704 #ifndef __KERNEL__
705 retry:
706 #endif
707 	memset(sb, 0, sizeof(*sb));
708 	sb->mode	= BLK_OPEN_READ;
709 	sb->have_bio	= true;
710 	sb->holder	= kmalloc(1, GFP_KERNEL);
711 	if (!sb->holder)
712 		return -ENOMEM;
713 
714 	sb->sb_name = kstrdup(path, GFP_KERNEL);
715 	if (!sb->sb_name)
716 		return -ENOMEM;
717 
718 #ifndef __KERNEL__
719 	if (opt_get(*opts, direct_io) == false)
720 		sb->mode |= BLK_OPEN_BUFFERED;
721 #endif
722 
723 	if (!opt_get(*opts, noexcl))
724 		sb->mode |= BLK_OPEN_EXCL;
725 
726 	if (!opt_get(*opts, nochanges))
727 		sb->mode |= BLK_OPEN_WRITE;
728 
729 	sb->bdev = blkdev_get_by_path(path, sb->mode, sb->holder, &bch2_sb_handle_bdev_ops);
730 	if (IS_ERR(sb->bdev) &&
731 	    PTR_ERR(sb->bdev) == -EACCES &&
732 	    opt_get(*opts, read_only)) {
733 		sb->mode &= ~BLK_OPEN_WRITE;
734 
735 		sb->bdev = blkdev_get_by_path(path, sb->mode, sb->holder, &bch2_sb_handle_bdev_ops);
736 		if (!IS_ERR(sb->bdev))
737 			opt_set(*opts, nochanges, true);
738 	}
739 
740 	if (IS_ERR(sb->bdev)) {
741 		ret = PTR_ERR(sb->bdev);
742 		goto out;
743 	}
744 
745 	ret = bch2_sb_realloc(sb, 0);
746 	if (ret) {
747 		prt_printf(&err, "error allocating memory for superblock");
748 		goto err;
749 	}
750 
751 	if (bch2_fs_init_fault("read_super")) {
752 		prt_printf(&err, "dynamic fault");
753 		ret = -EFAULT;
754 		goto err;
755 	}
756 
757 	ret = read_one_super(sb, offset, &err);
758 	if (!ret)
759 		goto got_super;
760 
761 	if (opt_defined(*opts, sb))
762 		goto err;
763 
764 	prt_printf(&err2, "bcachefs (%s): error reading default superblock: %s\n",
765 	       path, err.buf);
766 	if (ret == -BCH_ERR_invalid_sb_magic && ignore_notbchfs_msg)
767 		printk(KERN_INFO "%s", err2.buf);
768 	else
769 		printk(KERN_ERR "%s", err2.buf);
770 
771 	printbuf_exit(&err2);
772 	printbuf_reset(&err);
773 
774 	/*
775 	 * Error reading primary superblock - read location of backup
776 	 * superblocks:
777 	 */
778 	bio_reset(sb->bio, sb->bdev, REQ_OP_READ|REQ_SYNC|REQ_META);
779 	sb->bio->bi_iter.bi_sector = BCH_SB_LAYOUT_SECTOR;
780 	/*
781 	 * use sb buffer to read layout, since sb buffer is page aligned but
782 	 * layout won't be:
783 	 */
784 	bch2_bio_map(sb->bio, sb->sb, sizeof(struct bch_sb_layout));
785 
786 	ret = submit_bio_wait(sb->bio);
787 	if (ret) {
788 		prt_printf(&err, "IO error: %i", ret);
789 		goto err;
790 	}
791 
792 	memcpy(&layout, sb->sb, sizeof(layout));
793 	ret = validate_sb_layout(&layout, &err);
794 	if (ret)
795 		goto err;
796 
797 	for (i = layout.sb_offset;
798 	     i < layout.sb_offset + layout.nr_superblocks; i++) {
799 		offset = le64_to_cpu(*i);
800 
801 		if (offset == opt_get(*opts, sb))
802 			continue;
803 
804 		ret = read_one_super(sb, offset, &err);
805 		if (!ret)
806 			goto got_super;
807 	}
808 
809 	goto err;
810 
811 got_super:
812 	if (le16_to_cpu(sb->sb->block_size) << 9 <
813 	    bdev_logical_block_size(sb->bdev) &&
814 	    opt_get(*opts, direct_io)) {
815 #ifndef __KERNEL__
816 		opt_set(*opts, direct_io, false);
817 		bch2_free_super(sb);
818 		goto retry;
819 #endif
820 		prt_printf(&err, "block size (%u) smaller than device block size (%u)",
821 		       le16_to_cpu(sb->sb->block_size) << 9,
822 		       bdev_logical_block_size(sb->bdev));
823 		ret = -BCH_ERR_block_size_too_small;
824 		goto err;
825 	}
826 
827 	ret = 0;
828 	sb->have_layout = true;
829 
830 	ret = bch2_sb_validate(sb, &err, READ);
831 	if (ret) {
832 		printk(KERN_ERR "bcachefs (%s): error validating superblock: %s\n",
833 		       path, err.buf);
834 		goto err_no_print;
835 	}
836 out:
837 	printbuf_exit(&err);
838 	return ret;
839 err:
840 	printk(KERN_ERR "bcachefs (%s): error reading superblock: %s\n",
841 	       path, err.buf);
842 err_no_print:
843 	bch2_free_super(sb);
844 	goto out;
845 }
846 
847 int bch2_read_super(const char *path, struct bch_opts *opts,
848 		    struct bch_sb_handle *sb)
849 {
850 	return __bch2_read_super(path, opts, sb, false);
851 }
852 
853 /* provide a silenced version for mount.bcachefs */
854 
855 int bch2_read_super_silent(const char *path, struct bch_opts *opts,
856 		    struct bch_sb_handle *sb)
857 {
858 	return __bch2_read_super(path, opts, sb, true);
859 }
860 
861 /* write superblock: */
862 
863 static void write_super_endio(struct bio *bio)
864 {
865 	struct bch_dev *ca = bio->bi_private;
866 
867 	/* XXX: return errors directly */
868 
869 	if (bch2_dev_io_err_on(bio->bi_status, ca,
870 			       bio_data_dir(bio)
871 			       ? BCH_MEMBER_ERROR_write
872 			       : BCH_MEMBER_ERROR_read,
873 			       "superblock %s error: %s",
874 			       bio_data_dir(bio) ? "write" : "read",
875 			       bch2_blk_status_to_str(bio->bi_status)))
876 		ca->sb_write_error = 1;
877 
878 	closure_put(&ca->fs->sb_write);
879 	percpu_ref_put(&ca->io_ref);
880 }
881 
882 static void read_back_super(struct bch_fs *c, struct bch_dev *ca)
883 {
884 	struct bch_sb *sb = ca->disk_sb.sb;
885 	struct bio *bio = ca->disk_sb.bio;
886 
887 	bio_reset(bio, ca->disk_sb.bdev, REQ_OP_READ|REQ_SYNC|REQ_META);
888 	bio->bi_iter.bi_sector	= le64_to_cpu(sb->layout.sb_offset[0]);
889 	bio->bi_end_io		= write_super_endio;
890 	bio->bi_private		= ca;
891 	bch2_bio_map(bio, ca->sb_read_scratch, PAGE_SIZE);
892 
893 	this_cpu_add(ca->io_done->sectors[READ][BCH_DATA_sb],
894 		     bio_sectors(bio));
895 
896 	percpu_ref_get(&ca->io_ref);
897 	closure_bio_submit(bio, &c->sb_write);
898 }
899 
900 static void write_one_super(struct bch_fs *c, struct bch_dev *ca, unsigned idx)
901 {
902 	struct bch_sb *sb = ca->disk_sb.sb;
903 	struct bio *bio = ca->disk_sb.bio;
904 
905 	sb->offset = sb->layout.sb_offset[idx];
906 
907 	SET_BCH_SB_CSUM_TYPE(sb, bch2_csum_opt_to_type(c->opts.metadata_checksum, false));
908 	sb->csum = csum_vstruct(c, BCH_SB_CSUM_TYPE(sb),
909 				null_nonce(), sb);
910 
911 	bio_reset(bio, ca->disk_sb.bdev, REQ_OP_WRITE|REQ_SYNC|REQ_META);
912 	bio->bi_iter.bi_sector	= le64_to_cpu(sb->offset);
913 	bio->bi_end_io		= write_super_endio;
914 	bio->bi_private		= ca;
915 	bch2_bio_map(bio, sb,
916 		     roundup((size_t) vstruct_bytes(sb),
917 			     bdev_logical_block_size(ca->disk_sb.bdev)));
918 
919 	this_cpu_add(ca->io_done->sectors[WRITE][BCH_DATA_sb],
920 		     bio_sectors(bio));
921 
922 	percpu_ref_get(&ca->io_ref);
923 	closure_bio_submit(bio, &c->sb_write);
924 }
925 
926 int bch2_write_super(struct bch_fs *c)
927 {
928 	struct closure *cl = &c->sb_write;
929 	struct bch_dev *ca;
930 	struct printbuf err = PRINTBUF;
931 	unsigned i, sb = 0, nr_wrote;
932 	struct bch_devs_mask sb_written;
933 	bool wrote, can_mount_without_written, can_mount_with_written;
934 	unsigned degraded_flags = BCH_FORCE_IF_DEGRADED;
935 	int ret = 0;
936 
937 	trace_and_count(c, write_super, c, _RET_IP_);
938 
939 	if (c->opts.very_degraded)
940 		degraded_flags |= BCH_FORCE_IF_LOST;
941 
942 	lockdep_assert_held(&c->sb_lock);
943 
944 	closure_init_stack(cl);
945 	memset(&sb_written, 0, sizeof(sb_written));
946 
947 	/* Make sure we're using the new magic numbers: */
948 	c->disk_sb.sb->magic = BCHFS_MAGIC;
949 	c->disk_sb.sb->layout.magic = BCHFS_MAGIC;
950 
951 	le64_add_cpu(&c->disk_sb.sb->seq, 1);
952 
953 	if (test_bit(BCH_FS_error, &c->flags))
954 		SET_BCH_SB_HAS_ERRORS(c->disk_sb.sb, 1);
955 	if (test_bit(BCH_FS_topology_error, &c->flags))
956 		SET_BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb, 1);
957 
958 	SET_BCH_SB_BIG_ENDIAN(c->disk_sb.sb, CPU_BIG_ENDIAN);
959 
960 	bch2_sb_counters_from_cpu(c);
961 	bch2_sb_members_from_cpu(c);
962 	bch2_sb_members_cpy_v2_v1(&c->disk_sb);
963 	bch2_sb_errors_from_cpu(c);
964 	bch2_sb_downgrade_update(c);
965 
966 	for_each_online_member(ca, c, i)
967 		bch2_sb_from_fs(c, ca);
968 
969 	for_each_online_member(ca, c, i) {
970 		printbuf_reset(&err);
971 
972 		ret = bch2_sb_validate(&ca->disk_sb, &err, WRITE);
973 		if (ret) {
974 			bch2_fs_inconsistent(c, "sb invalid before write: %s", err.buf);
975 			percpu_ref_put(&ca->io_ref);
976 			goto out;
977 		}
978 	}
979 
980 	if (c->opts.nochanges)
981 		goto out;
982 
983 	/*
984 	 * Defer writing the superblock until filesystem initialization is
985 	 * complete - don't write out a partly initialized superblock:
986 	 */
987 	if (!BCH_SB_INITIALIZED(c->disk_sb.sb))
988 		goto out;
989 
990 	for_each_online_member(ca, c, i) {
991 		__set_bit(ca->dev_idx, sb_written.d);
992 		ca->sb_write_error = 0;
993 	}
994 
995 	for_each_online_member(ca, c, i)
996 		read_back_super(c, ca);
997 	closure_sync(cl);
998 
999 	for_each_online_member(ca, c, i) {
1000 		if (ca->sb_write_error)
1001 			continue;
1002 
1003 		if (le64_to_cpu(ca->sb_read_scratch->seq) < ca->disk_sb.seq) {
1004 			bch2_fs_fatal_error(c,
1005 				"Superblock write was silently dropped! (seq %llu expected %llu)",
1006 				le64_to_cpu(ca->sb_read_scratch->seq),
1007 				ca->disk_sb.seq);
1008 			percpu_ref_put(&ca->io_ref);
1009 			ret = -BCH_ERR_erofs_sb_err;
1010 			goto out;
1011 		}
1012 
1013 		if (le64_to_cpu(ca->sb_read_scratch->seq) > ca->disk_sb.seq) {
1014 			bch2_fs_fatal_error(c,
1015 				"Superblock modified by another process (seq %llu expected %llu)",
1016 				le64_to_cpu(ca->sb_read_scratch->seq),
1017 				ca->disk_sb.seq);
1018 			percpu_ref_put(&ca->io_ref);
1019 			ret = -BCH_ERR_erofs_sb_err;
1020 			goto out;
1021 		}
1022 	}
1023 
1024 	do {
1025 		wrote = false;
1026 		for_each_online_member(ca, c, i)
1027 			if (!ca->sb_write_error &&
1028 			    sb < ca->disk_sb.sb->layout.nr_superblocks) {
1029 				write_one_super(c, ca, sb);
1030 				wrote = true;
1031 			}
1032 		closure_sync(cl);
1033 		sb++;
1034 	} while (wrote);
1035 
1036 	for_each_online_member(ca, c, i) {
1037 		if (ca->sb_write_error)
1038 			__clear_bit(ca->dev_idx, sb_written.d);
1039 		else
1040 			ca->disk_sb.seq = le64_to_cpu(ca->disk_sb.sb->seq);
1041 	}
1042 
1043 	nr_wrote = dev_mask_nr(&sb_written);
1044 
1045 	can_mount_with_written =
1046 		bch2_have_enough_devs(c, sb_written, degraded_flags, false);
1047 
1048 	for (i = 0; i < ARRAY_SIZE(sb_written.d); i++)
1049 		sb_written.d[i] = ~sb_written.d[i];
1050 
1051 	can_mount_without_written =
1052 		bch2_have_enough_devs(c, sb_written, degraded_flags, false);
1053 
1054 	/*
1055 	 * If we would be able to mount _without_ the devices we successfully
1056 	 * wrote superblocks to, we weren't able to write to enough devices:
1057 	 *
1058 	 * Exception: if we can mount without the successes because we haven't
1059 	 * written anything (new filesystem), we continue if we'd be able to
1060 	 * mount with the devices we did successfully write to:
1061 	 */
1062 	if (bch2_fs_fatal_err_on(!nr_wrote ||
1063 				 !can_mount_with_written ||
1064 				 (can_mount_without_written &&
1065 				  !can_mount_with_written), c,
1066 		"Unable to write superblock to sufficient devices (from %ps)",
1067 		(void *) _RET_IP_))
1068 		ret = -1;
1069 out:
1070 	/* Make new options visible after they're persistent: */
1071 	bch2_sb_update(c);
1072 	printbuf_exit(&err);
1073 	return ret;
1074 }
1075 
1076 void __bch2_check_set_feature(struct bch_fs *c, unsigned feat)
1077 {
1078 	mutex_lock(&c->sb_lock);
1079 	if (!(c->sb.features & (1ULL << feat))) {
1080 		c->disk_sb.sb->features[0] |= cpu_to_le64(1ULL << feat);
1081 
1082 		bch2_write_super(c);
1083 	}
1084 	mutex_unlock(&c->sb_lock);
1085 }
1086 
1087 /* Downgrade if superblock is at a higher version than currently supported: */
1088 bool bch2_check_version_downgrade(struct bch_fs *c)
1089 {
1090 	bool ret = bcachefs_metadata_version_current < c->sb.version;
1091 
1092 	lockdep_assert_held(&c->sb_lock);
1093 
1094 	/*
1095 	 * Downgrade, if superblock is at a higher version than currently
1096 	 * supported:
1097 	 */
1098 	if (BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb) > bcachefs_metadata_version_current)
1099 		SET_BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb, bcachefs_metadata_version_current);
1100 	if (c->sb.version > bcachefs_metadata_version_current)
1101 		c->disk_sb.sb->version = cpu_to_le16(bcachefs_metadata_version_current);
1102 	if (c->sb.version_min > bcachefs_metadata_version_current)
1103 		c->disk_sb.sb->version_min = cpu_to_le16(bcachefs_metadata_version_current);
1104 	c->disk_sb.sb->compat[0] &= cpu_to_le64((1ULL << BCH_COMPAT_NR) - 1);
1105 	return ret;
1106 }
1107 
1108 void bch2_sb_upgrade(struct bch_fs *c, unsigned new_version)
1109 {
1110 	lockdep_assert_held(&c->sb_lock);
1111 
1112 	if (BCH_VERSION_MAJOR(new_version) >
1113 	    BCH_VERSION_MAJOR(le16_to_cpu(c->disk_sb.sb->version)))
1114 		bch2_sb_field_resize(&c->disk_sb, downgrade, 0);
1115 
1116 	c->disk_sb.sb->version = cpu_to_le16(new_version);
1117 	c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALL);
1118 }
1119 
1120 static int bch2_sb_ext_validate(struct bch_sb *sb, struct bch_sb_field *f,
1121 				struct printbuf *err)
1122 {
1123 	if (vstruct_bytes(f) < 88) {
1124 		prt_printf(err, "field too small (%zu < %u)", vstruct_bytes(f), 88);
1125 		return -BCH_ERR_invalid_sb_ext;
1126 	}
1127 
1128 	return 0;
1129 }
1130 
1131 static void bch2_sb_ext_to_text(struct printbuf *out, struct bch_sb *sb,
1132 				struct bch_sb_field *f)
1133 {
1134 	struct bch_sb_field_ext *e = field_to_type(f, ext);
1135 
1136 	prt_printf(out, "Recovery passes required:");
1137 	prt_tab(out);
1138 	prt_bitflags(out, bch2_recovery_passes,
1139 		     bch2_recovery_passes_from_stable(le64_to_cpu(e->recovery_passes_required[0])));
1140 	prt_newline(out);
1141 
1142 	unsigned long *errors_silent = kmalloc(sizeof(e->errors_silent), GFP_KERNEL);
1143 	if (errors_silent) {
1144 		le_bitvector_to_cpu(errors_silent, (void *) e->errors_silent, sizeof(e->errors_silent) * 8);
1145 
1146 		prt_printf(out, "Errors to silently fix:");
1147 		prt_tab(out);
1148 		prt_bitflags_vector(out, bch2_sb_error_strs, errors_silent, sizeof(e->errors_silent) * 8);
1149 		prt_newline(out);
1150 
1151 		kfree(errors_silent);
1152 	}
1153 }
1154 
1155 static const struct bch_sb_field_ops bch_sb_field_ops_ext = {
1156 	.validate	= bch2_sb_ext_validate,
1157 	.to_text	= bch2_sb_ext_to_text,
1158 };
1159 
1160 static const struct bch_sb_field_ops *bch2_sb_field_ops[] = {
1161 #define x(f, nr)					\
1162 	[BCH_SB_FIELD_##f] = &bch_sb_field_ops_##f,
1163 	BCH_SB_FIELDS()
1164 #undef x
1165 };
1166 
1167 static const struct bch_sb_field_ops bch2_sb_field_null_ops;
1168 
1169 static const struct bch_sb_field_ops *bch2_sb_field_type_ops(unsigned type)
1170 {
1171 	return likely(type < ARRAY_SIZE(bch2_sb_field_ops))
1172 		? bch2_sb_field_ops[type]
1173 		: &bch2_sb_field_null_ops;
1174 }
1175 
1176 static int bch2_sb_field_validate(struct bch_sb *sb, struct bch_sb_field *f,
1177 				  struct printbuf *err)
1178 {
1179 	unsigned type = le32_to_cpu(f->type);
1180 	struct printbuf field_err = PRINTBUF;
1181 	const struct bch_sb_field_ops *ops = bch2_sb_field_type_ops(type);
1182 	int ret;
1183 
1184 	ret = ops->validate ? ops->validate(sb, f, &field_err) : 0;
1185 	if (ret) {
1186 		prt_printf(err, "Invalid superblock section %s: %s",
1187 			   bch2_sb_fields[type], field_err.buf);
1188 		prt_newline(err);
1189 		bch2_sb_field_to_text(err, sb, f);
1190 	}
1191 
1192 	printbuf_exit(&field_err);
1193 	return ret;
1194 }
1195 
1196 void bch2_sb_field_to_text(struct printbuf *out, struct bch_sb *sb,
1197 			   struct bch_sb_field *f)
1198 {
1199 	unsigned type = le32_to_cpu(f->type);
1200 	const struct bch_sb_field_ops *ops = bch2_sb_field_type_ops(type);
1201 
1202 	if (!out->nr_tabstops)
1203 		printbuf_tabstop_push(out, 32);
1204 
1205 	if (type < BCH_SB_FIELD_NR)
1206 		prt_printf(out, "%s", bch2_sb_fields[type]);
1207 	else
1208 		prt_printf(out, "(unknown field %u)", type);
1209 
1210 	prt_printf(out, " (size %zu):", vstruct_bytes(f));
1211 	prt_newline(out);
1212 
1213 	if (ops->to_text) {
1214 		printbuf_indent_add(out, 2);
1215 		ops->to_text(out, sb, f);
1216 		printbuf_indent_sub(out, 2);
1217 	}
1218 }
1219 
1220 void bch2_sb_layout_to_text(struct printbuf *out, struct bch_sb_layout *l)
1221 {
1222 	unsigned i;
1223 
1224 	prt_printf(out, "Type:                    %u", l->layout_type);
1225 	prt_newline(out);
1226 
1227 	prt_str(out, "Superblock max size:     ");
1228 	prt_units_u64(out, 512 << l->sb_max_size_bits);
1229 	prt_newline(out);
1230 
1231 	prt_printf(out, "Nr superblocks:          %u", l->nr_superblocks);
1232 	prt_newline(out);
1233 
1234 	prt_str(out, "Offsets:                 ");
1235 	for (i = 0; i < l->nr_superblocks; i++) {
1236 		if (i)
1237 			prt_str(out, ", ");
1238 		prt_printf(out, "%llu", le64_to_cpu(l->sb_offset[i]));
1239 	}
1240 	prt_newline(out);
1241 }
1242 
1243 void bch2_sb_to_text(struct printbuf *out, struct bch_sb *sb,
1244 		     bool print_layout, unsigned fields)
1245 {
1246 	struct bch_sb_field *f;
1247 	u64 fields_have = 0;
1248 	unsigned nr_devices = 0;
1249 
1250 	if (!out->nr_tabstops)
1251 		printbuf_tabstop_push(out, 44);
1252 
1253 	for (int i = 0; i < sb->nr_devices; i++)
1254 		nr_devices += bch2_dev_exists(sb, i);
1255 
1256 	prt_printf(out, "External UUID:");
1257 	prt_tab(out);
1258 	pr_uuid(out, sb->user_uuid.b);
1259 	prt_newline(out);
1260 
1261 	prt_printf(out, "Internal UUID:");
1262 	prt_tab(out);
1263 	pr_uuid(out, sb->uuid.b);
1264 	prt_newline(out);
1265 
1266 	prt_str(out, "Device index:");
1267 	prt_tab(out);
1268 	prt_printf(out, "%u", sb->dev_idx);
1269 	prt_newline(out);
1270 
1271 	prt_str(out, "Label:");
1272 	prt_tab(out);
1273 	prt_printf(out, "%.*s", (int) sizeof(sb->label), sb->label);
1274 	prt_newline(out);
1275 
1276 	prt_str(out, "Version:");
1277 	prt_tab(out);
1278 	bch2_version_to_text(out, le16_to_cpu(sb->version));
1279 	prt_newline(out);
1280 
1281 	prt_str(out, "Version upgrade complete:");
1282 	prt_tab(out);
1283 	bch2_version_to_text(out, BCH_SB_VERSION_UPGRADE_COMPLETE(sb));
1284 	prt_newline(out);
1285 
1286 	prt_printf(out, "Oldest version on disk:");
1287 	prt_tab(out);
1288 	bch2_version_to_text(out, le16_to_cpu(sb->version_min));
1289 	prt_newline(out);
1290 
1291 	prt_printf(out, "Created:");
1292 	prt_tab(out);
1293 	if (sb->time_base_lo)
1294 		bch2_prt_datetime(out, div_u64(le64_to_cpu(sb->time_base_lo), NSEC_PER_SEC));
1295 	else
1296 		prt_printf(out, "(not set)");
1297 	prt_newline(out);
1298 
1299 	prt_printf(out, "Sequence number:");
1300 	prt_tab(out);
1301 	prt_printf(out, "%llu", le64_to_cpu(sb->seq));
1302 	prt_newline(out);
1303 
1304 	prt_printf(out, "Superblock size:");
1305 	prt_tab(out);
1306 	prt_printf(out, "%zu", vstruct_bytes(sb));
1307 	prt_newline(out);
1308 
1309 	prt_printf(out, "Clean:");
1310 	prt_tab(out);
1311 	prt_printf(out, "%llu", BCH_SB_CLEAN(sb));
1312 	prt_newline(out);
1313 
1314 	prt_printf(out, "Devices:");
1315 	prt_tab(out);
1316 	prt_printf(out, "%u", nr_devices);
1317 	prt_newline(out);
1318 
1319 	prt_printf(out, "Sections:");
1320 	vstruct_for_each(sb, f)
1321 		fields_have |= 1 << le32_to_cpu(f->type);
1322 	prt_tab(out);
1323 	prt_bitflags(out, bch2_sb_fields, fields_have);
1324 	prt_newline(out);
1325 
1326 	prt_printf(out, "Features:");
1327 	prt_tab(out);
1328 	prt_bitflags(out, bch2_sb_features, le64_to_cpu(sb->features[0]));
1329 	prt_newline(out);
1330 
1331 	prt_printf(out, "Compat features:");
1332 	prt_tab(out);
1333 	prt_bitflags(out, bch2_sb_compat, le64_to_cpu(sb->compat[0]));
1334 	prt_newline(out);
1335 
1336 	prt_newline(out);
1337 	prt_printf(out, "Options:");
1338 	prt_newline(out);
1339 	printbuf_indent_add(out, 2);
1340 	{
1341 		enum bch_opt_id id;
1342 
1343 		for (id = 0; id < bch2_opts_nr; id++) {
1344 			const struct bch_option *opt = bch2_opt_table + id;
1345 
1346 			if (opt->get_sb != BCH2_NO_SB_OPT) {
1347 				u64 v = bch2_opt_from_sb(sb, id);
1348 
1349 				prt_printf(out, "%s:", opt->attr.name);
1350 				prt_tab(out);
1351 				bch2_opt_to_text(out, NULL, sb, opt, v,
1352 						 OPT_HUMAN_READABLE|OPT_SHOW_FULL_LIST);
1353 				prt_newline(out);
1354 			}
1355 		}
1356 	}
1357 
1358 	printbuf_indent_sub(out, 2);
1359 
1360 	if (print_layout) {
1361 		prt_newline(out);
1362 		prt_printf(out, "layout:");
1363 		prt_newline(out);
1364 		printbuf_indent_add(out, 2);
1365 		bch2_sb_layout_to_text(out, &sb->layout);
1366 		printbuf_indent_sub(out, 2);
1367 	}
1368 
1369 	vstruct_for_each(sb, f)
1370 		if (fields & (1 << le32_to_cpu(f->type))) {
1371 			prt_newline(out);
1372 			bch2_sb_field_to_text(out, sb, f);
1373 		}
1374 }
1375