xref: /linux/fs/bcachefs/super-io.c (revision 3ea5eb68b9d624935108b5e696859304edfac202)
1 // SPDX-License-Identifier: GPL-2.0
2 
3 #include "bcachefs.h"
4 #include "checksum.h"
5 #include "disk_groups.h"
6 #include "ec.h"
7 #include "error.h"
8 #include "journal.h"
9 #include "journal_sb.h"
10 #include "journal_seq_blacklist.h"
11 #include "recovery_passes.h"
12 #include "replicas.h"
13 #include "quota.h"
14 #include "sb-clean.h"
15 #include "sb-counters.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 };
34 
35 static const struct bch2_metadata_version bch2_metadata_versions[] = {
36 #define x(n, v) {		\
37 	.version = v,				\
38 	.name = #n,				\
39 },
40 	BCH_METADATA_VERSIONS()
41 #undef x
42 };
43 
44 void bch2_version_to_text(struct printbuf *out, unsigned v)
45 {
46 	const char *str = "(unknown version)";
47 
48 	for (unsigned i = 0; i < ARRAY_SIZE(bch2_metadata_versions); i++)
49 		if (bch2_metadata_versions[i].version == v) {
50 			str = bch2_metadata_versions[i].name;
51 			break;
52 		}
53 
54 	prt_printf(out, "%u.%u: %s", BCH_VERSION_MAJOR(v), BCH_VERSION_MINOR(v), str);
55 }
56 
57 unsigned bch2_latest_compatible_version(unsigned v)
58 {
59 	if (!BCH_VERSION_MAJOR(v))
60 		return v;
61 
62 	for (unsigned i = 0; i < ARRAY_SIZE(bch2_metadata_versions); i++)
63 		if (bch2_metadata_versions[i].version > v &&
64 		    BCH_VERSION_MAJOR(bch2_metadata_versions[i].version) ==
65 		    BCH_VERSION_MAJOR(v))
66 			v = bch2_metadata_versions[i].version;
67 
68 	return v;
69 }
70 
71 const char * const bch2_sb_fields[] = {
72 #define x(name, nr)	#name,
73 	BCH_SB_FIELDS()
74 #undef x
75 	NULL
76 };
77 
78 static int bch2_sb_field_validate(struct bch_sb *, struct bch_sb_field *,
79 				  enum bch_validate_flags, struct printbuf *);
80 
81 struct bch_sb_field *bch2_sb_field_get_id(struct bch_sb *sb,
82 				      enum bch_sb_field_type type)
83 {
84 	/* XXX: need locking around superblock to access optional fields */
85 
86 	vstruct_for_each(sb, f)
87 		if (le32_to_cpu(f->type) == type)
88 			return f;
89 	return NULL;
90 }
91 
92 static struct bch_sb_field *__bch2_sb_field_resize(struct bch_sb_handle *sb,
93 						   struct bch_sb_field *f,
94 						   unsigned u64s)
95 {
96 	unsigned old_u64s = f ? le32_to_cpu(f->u64s) : 0;
97 	unsigned sb_u64s = le32_to_cpu(sb->sb->u64s) + u64s - old_u64s;
98 
99 	BUG_ON(__vstruct_bytes(struct bch_sb, sb_u64s) > sb->buffer_size);
100 
101 	if (!f && !u64s) {
102 		/* nothing to do: */
103 	} else if (!f) {
104 		f = vstruct_last(sb->sb);
105 		memset(f, 0, sizeof(u64) * u64s);
106 		f->u64s = cpu_to_le32(u64s);
107 		f->type = 0;
108 	} else {
109 		void *src, *dst;
110 
111 		src = vstruct_end(f);
112 
113 		if (u64s) {
114 			f->u64s = cpu_to_le32(u64s);
115 			dst = vstruct_end(f);
116 		} else {
117 			dst = f;
118 		}
119 
120 		memmove(dst, src, vstruct_end(sb->sb) - src);
121 
122 		if (dst > src)
123 			memset(src, 0, dst - src);
124 	}
125 
126 	sb->sb->u64s = cpu_to_le32(sb_u64s);
127 
128 	return u64s ? f : NULL;
129 }
130 
131 void bch2_sb_field_delete(struct bch_sb_handle *sb,
132 			  enum bch_sb_field_type type)
133 {
134 	struct bch_sb_field *f = bch2_sb_field_get_id(sb->sb, type);
135 
136 	if (f)
137 		__bch2_sb_field_resize(sb, f, 0);
138 }
139 
140 /* Superblock realloc/free: */
141 
142 void bch2_free_super(struct bch_sb_handle *sb)
143 {
144 	kfree(sb->bio);
145 	if (!IS_ERR_OR_NULL(sb->s_bdev_file))
146 		bdev_fput(sb->s_bdev_file);
147 	kfree(sb->holder);
148 	kfree(sb->sb_name);
149 
150 	kfree(sb->sb);
151 	memset(sb, 0, sizeof(*sb));
152 }
153 
154 int bch2_sb_realloc(struct bch_sb_handle *sb, unsigned u64s)
155 {
156 	size_t new_bytes = __vstruct_bytes(struct bch_sb, u64s);
157 	size_t new_buffer_size;
158 	struct bch_sb *new_sb;
159 	struct bio *bio;
160 
161 	if (sb->bdev)
162 		new_bytes = max_t(size_t, new_bytes, bdev_logical_block_size(sb->bdev));
163 
164 	new_buffer_size = roundup_pow_of_two(new_bytes);
165 
166 	if (sb->sb && sb->buffer_size >= new_buffer_size)
167 		return 0;
168 
169 	if (sb->sb && sb->have_layout) {
170 		u64 max_bytes = 512 << sb->sb->layout.sb_max_size_bits;
171 
172 		if (new_bytes > max_bytes) {
173 			struct printbuf buf = PRINTBUF;
174 
175 			prt_bdevname(&buf, sb->bdev);
176 			prt_printf(&buf, ": superblock too big: want %zu but have %llu", new_bytes, max_bytes);
177 			pr_err("%s", buf.buf);
178 			printbuf_exit(&buf);
179 			return -BCH_ERR_ENOSPC_sb;
180 		}
181 	}
182 
183 	if (sb->buffer_size >= new_buffer_size && sb->sb)
184 		return 0;
185 
186 	if (dynamic_fault("bcachefs:add:super_realloc"))
187 		return -BCH_ERR_ENOMEM_sb_realloc_injected;
188 
189 	new_sb = krealloc(sb->sb, new_buffer_size, GFP_NOFS|__GFP_ZERO);
190 	if (!new_sb)
191 		return -BCH_ERR_ENOMEM_sb_buf_realloc;
192 
193 	sb->sb = new_sb;
194 
195 	if (sb->have_bio) {
196 		unsigned nr_bvecs = buf_pages(sb->sb, new_buffer_size);
197 
198 		bio = bio_kmalloc(nr_bvecs, GFP_KERNEL);
199 		if (!bio)
200 			return -BCH_ERR_ENOMEM_sb_bio_realloc;
201 
202 		bio_init(bio, NULL, bio->bi_inline_vecs, nr_bvecs, 0);
203 
204 		kfree(sb->bio);
205 		sb->bio = bio;
206 	}
207 
208 	sb->buffer_size = new_buffer_size;
209 
210 	return 0;
211 }
212 
213 struct bch_sb_field *bch2_sb_field_resize_id(struct bch_sb_handle *sb,
214 					  enum bch_sb_field_type type,
215 					  unsigned u64s)
216 {
217 	struct bch_sb_field *f = bch2_sb_field_get_id(sb->sb, type);
218 	ssize_t old_u64s = f ? le32_to_cpu(f->u64s) : 0;
219 	ssize_t d = -old_u64s + u64s;
220 
221 	if (bch2_sb_realloc(sb, le32_to_cpu(sb->sb->u64s) + d))
222 		return NULL;
223 
224 	if (sb->fs_sb) {
225 		struct bch_fs *c = container_of(sb, struct bch_fs, disk_sb);
226 
227 		lockdep_assert_held(&c->sb_lock);
228 
229 		/* XXX: we're not checking that offline device have enough space */
230 
231 		for_each_online_member(c, ca) {
232 			struct bch_sb_handle *dev_sb = &ca->disk_sb;
233 
234 			if (bch2_sb_realloc(dev_sb, le32_to_cpu(dev_sb->sb->u64s) + d)) {
235 				percpu_ref_put(&ca->io_ref);
236 				return NULL;
237 			}
238 		}
239 	}
240 
241 	f = bch2_sb_field_get_id(sb->sb, type);
242 	f = __bch2_sb_field_resize(sb, f, u64s);
243 	if (f)
244 		f->type = cpu_to_le32(type);
245 	return f;
246 }
247 
248 struct bch_sb_field *bch2_sb_field_get_minsize_id(struct bch_sb_handle *sb,
249 						  enum bch_sb_field_type type,
250 						  unsigned u64s)
251 {
252 	struct bch_sb_field *f = bch2_sb_field_get_id(sb->sb, type);
253 
254 	if (!f || le32_to_cpu(f->u64s) < u64s)
255 		f = bch2_sb_field_resize_id(sb, type, u64s);
256 	return f;
257 }
258 
259 /* Superblock validate: */
260 
261 static int validate_sb_layout(struct bch_sb_layout *layout, struct printbuf *out)
262 {
263 	u64 offset, prev_offset, max_sectors;
264 	unsigned i;
265 
266 	BUILD_BUG_ON(sizeof(struct bch_sb_layout) != 512);
267 
268 	if (!uuid_equal(&layout->magic, &BCACHE_MAGIC) &&
269 	    !uuid_equal(&layout->magic, &BCHFS_MAGIC)) {
270 		prt_printf(out, "Not a bcachefs superblock layout");
271 		return -BCH_ERR_invalid_sb_layout;
272 	}
273 
274 	if (layout->layout_type != 0) {
275 		prt_printf(out, "Invalid superblock layout type %u",
276 		       layout->layout_type);
277 		return -BCH_ERR_invalid_sb_layout_type;
278 	}
279 
280 	if (!layout->nr_superblocks) {
281 		prt_printf(out, "Invalid superblock layout: no superblocks");
282 		return -BCH_ERR_invalid_sb_layout_nr_superblocks;
283 	}
284 
285 	if (layout->nr_superblocks > ARRAY_SIZE(layout->sb_offset)) {
286 		prt_printf(out, "Invalid superblock layout: too many superblocks");
287 		return -BCH_ERR_invalid_sb_layout_nr_superblocks;
288 	}
289 
290 	max_sectors = 1 << layout->sb_max_size_bits;
291 
292 	prev_offset = le64_to_cpu(layout->sb_offset[0]);
293 
294 	for (i = 1; i < layout->nr_superblocks; i++) {
295 		offset = le64_to_cpu(layout->sb_offset[i]);
296 
297 		if (offset < prev_offset + max_sectors) {
298 			prt_printf(out, "Invalid superblock layout: superblocks overlap\n"
299 			       "  (sb %u ends at %llu next starts at %llu",
300 			       i - 1, prev_offset + max_sectors, offset);
301 			return -BCH_ERR_invalid_sb_layout_superblocks_overlap;
302 		}
303 		prev_offset = offset;
304 	}
305 
306 	return 0;
307 }
308 
309 static int bch2_sb_compatible(struct bch_sb *sb, struct printbuf *out)
310 {
311 	u16 version		= le16_to_cpu(sb->version);
312 	u16 version_min		= le16_to_cpu(sb->version_min);
313 
314 	if (!bch2_version_compatible(version)) {
315 		prt_str(out, "Unsupported superblock version ");
316 		bch2_version_to_text(out, version);
317 		prt_str(out, " (min ");
318 		bch2_version_to_text(out, bcachefs_metadata_version_min);
319 		prt_str(out, ", max ");
320 		bch2_version_to_text(out, bcachefs_metadata_version_current);
321 		prt_str(out, ")");
322 		return -BCH_ERR_invalid_sb_version;
323 	}
324 
325 	if (!bch2_version_compatible(version_min)) {
326 		prt_str(out, "Unsupported superblock version_min ");
327 		bch2_version_to_text(out, version_min);
328 		prt_str(out, " (min ");
329 		bch2_version_to_text(out, bcachefs_metadata_version_min);
330 		prt_str(out, ", max ");
331 		bch2_version_to_text(out, bcachefs_metadata_version_current);
332 		prt_str(out, ")");
333 		return -BCH_ERR_invalid_sb_version;
334 	}
335 
336 	if (version_min > version) {
337 		prt_str(out, "Bad minimum version ");
338 		bch2_version_to_text(out, version_min);
339 		prt_str(out, ", greater than version field ");
340 		bch2_version_to_text(out, version);
341 		return -BCH_ERR_invalid_sb_version;
342 	}
343 
344 	return 0;
345 }
346 
347 static int bch2_sb_validate(struct bch_sb_handle *disk_sb,
348 			    enum bch_validate_flags flags, struct printbuf *out)
349 {
350 	struct bch_sb *sb = disk_sb->sb;
351 	struct bch_sb_field_members_v1 *mi;
352 	enum bch_opt_id opt_id;
353 	u16 block_size;
354 	int ret;
355 
356 	ret = bch2_sb_compatible(sb, out);
357 	if (ret)
358 		return ret;
359 
360 	if (sb->features[1] ||
361 	    (le64_to_cpu(sb->features[0]) & (~0ULL << BCH_FEATURE_NR))) {
362 		prt_printf(out, "Filesystem has incompatible features");
363 		return -BCH_ERR_invalid_sb_features;
364 	}
365 
366 	block_size = le16_to_cpu(sb->block_size);
367 
368 	if (block_size > PAGE_SECTORS) {
369 		prt_printf(out, "Block size too big (got %u, max %u)",
370 		       block_size, PAGE_SECTORS);
371 		return -BCH_ERR_invalid_sb_block_size;
372 	}
373 
374 	if (bch2_is_zero(sb->user_uuid.b, sizeof(sb->user_uuid))) {
375 		prt_printf(out, "Bad user UUID (got zeroes)");
376 		return -BCH_ERR_invalid_sb_uuid;
377 	}
378 
379 	if (bch2_is_zero(sb->uuid.b, sizeof(sb->uuid))) {
380 		prt_printf(out, "Bad internal UUID (got zeroes)");
381 		return -BCH_ERR_invalid_sb_uuid;
382 	}
383 
384 	if (!sb->nr_devices ||
385 	    sb->nr_devices > BCH_SB_MEMBERS_MAX) {
386 		prt_printf(out, "Bad number of member devices %u (max %u)",
387 		       sb->nr_devices, BCH_SB_MEMBERS_MAX);
388 		return -BCH_ERR_invalid_sb_too_many_members;
389 	}
390 
391 	if (sb->dev_idx >= sb->nr_devices) {
392 		prt_printf(out, "Bad dev_idx (got %u, nr_devices %u)",
393 		       sb->dev_idx, sb->nr_devices);
394 		return -BCH_ERR_invalid_sb_dev_idx;
395 	}
396 
397 	if (!sb->time_precision ||
398 	    le32_to_cpu(sb->time_precision) > NSEC_PER_SEC) {
399 		prt_printf(out, "Invalid time precision: %u (min 1, max %lu)",
400 		       le32_to_cpu(sb->time_precision), NSEC_PER_SEC);
401 		return -BCH_ERR_invalid_sb_time_precision;
402 	}
403 
404 	if (!flags) {
405 		/*
406 		 * Been seeing a bug where these are getting inexplicably
407 		 * zeroed, so we're now validating them, but we have to be
408 		 * careful not to preven people's filesystems from mounting:
409 		 */
410 		if (!BCH_SB_JOURNAL_FLUSH_DELAY(sb))
411 			SET_BCH_SB_JOURNAL_FLUSH_DELAY(sb, 1000);
412 		if (!BCH_SB_JOURNAL_RECLAIM_DELAY(sb))
413 			SET_BCH_SB_JOURNAL_RECLAIM_DELAY(sb, 1000);
414 
415 		if (!BCH_SB_VERSION_UPGRADE_COMPLETE(sb))
416 			SET_BCH_SB_VERSION_UPGRADE_COMPLETE(sb, le16_to_cpu(sb->version));
417 
418 		if (le16_to_cpu(sb->version) <= bcachefs_metadata_version_disk_accounting_v2 &&
419 		    !BCH_SB_ALLOCATOR_STUCK_TIMEOUT(sb))
420 			SET_BCH_SB_ALLOCATOR_STUCK_TIMEOUT(sb, 30);
421 
422 		if (le16_to_cpu(sb->version) <= bcachefs_metadata_version_disk_accounting_v2)
423 			SET_BCH_SB_PROMOTE_WHOLE_EXTENTS(sb, true);
424 	}
425 
426 	for (opt_id = 0; opt_id < bch2_opts_nr; opt_id++) {
427 		const struct bch_option *opt = bch2_opt_table + opt_id;
428 
429 		if (opt->get_sb != BCH2_NO_SB_OPT) {
430 			u64 v = bch2_opt_from_sb(sb, opt_id);
431 
432 			prt_printf(out, "Invalid option ");
433 			ret = bch2_opt_validate(opt, v, out);
434 			if (ret)
435 				return ret;
436 
437 			printbuf_reset(out);
438 		}
439 	}
440 
441 	/* validate layout */
442 	ret = validate_sb_layout(&sb->layout, out);
443 	if (ret)
444 		return ret;
445 
446 	vstruct_for_each(sb, f) {
447 		if (!f->u64s) {
448 			prt_printf(out, "Invalid superblock: optional field with size 0 (type %u)",
449 			       le32_to_cpu(f->type));
450 			return -BCH_ERR_invalid_sb_field_size;
451 		}
452 
453 		if (vstruct_next(f) > vstruct_last(sb)) {
454 			prt_printf(out, "Invalid superblock: optional field extends past end of superblock (type %u)",
455 			       le32_to_cpu(f->type));
456 			return -BCH_ERR_invalid_sb_field_size;
457 		}
458 	}
459 
460 	/* members must be validated first: */
461 	mi = bch2_sb_field_get(sb, members_v1);
462 	if (!mi) {
463 		prt_printf(out, "Invalid superblock: member info area missing");
464 		return -BCH_ERR_invalid_sb_members_missing;
465 	}
466 
467 	ret = bch2_sb_field_validate(sb, &mi->field, flags, out);
468 	if (ret)
469 		return ret;
470 
471 	vstruct_for_each(sb, f) {
472 		if (le32_to_cpu(f->type) == BCH_SB_FIELD_members_v1)
473 			continue;
474 
475 		ret = bch2_sb_field_validate(sb, f, flags, out);
476 		if (ret)
477 			return ret;
478 	}
479 
480 	if ((flags & BCH_VALIDATE_write) &&
481 	    bch2_sb_member_get(sb, sb->dev_idx).seq != sb->seq) {
482 		prt_printf(out, "Invalid superblock: member seq %llu != sb seq %llu",
483 			   le64_to_cpu(bch2_sb_member_get(sb, sb->dev_idx).seq),
484 			   le64_to_cpu(sb->seq));
485 		return -BCH_ERR_invalid_sb_members_missing;
486 	}
487 
488 	return 0;
489 }
490 
491 /* device open: */
492 
493 static unsigned long le_ulong_to_cpu(unsigned long v)
494 {
495 	return sizeof(unsigned long) == 8
496 		? le64_to_cpu(v)
497 		: le32_to_cpu(v);
498 }
499 
500 static void le_bitvector_to_cpu(unsigned long *dst, unsigned long *src, unsigned nr)
501 {
502 	BUG_ON(nr & (BITS_PER_TYPE(long) - 1));
503 
504 	for (unsigned i = 0; i < BITS_TO_LONGS(nr); i++)
505 		dst[i] = le_ulong_to_cpu(src[i]);
506 }
507 
508 static void bch2_sb_update(struct bch_fs *c)
509 {
510 	struct bch_sb *src = c->disk_sb.sb;
511 
512 	lockdep_assert_held(&c->sb_lock);
513 
514 	c->sb.uuid		= src->uuid;
515 	c->sb.user_uuid		= src->user_uuid;
516 	c->sb.version		= le16_to_cpu(src->version);
517 	c->sb.version_min	= le16_to_cpu(src->version_min);
518 	c->sb.version_upgrade_complete = BCH_SB_VERSION_UPGRADE_COMPLETE(src);
519 	c->sb.nr_devices	= src->nr_devices;
520 	c->sb.clean		= BCH_SB_CLEAN(src);
521 	c->sb.encryption_type	= BCH_SB_ENCRYPTION_TYPE(src);
522 
523 	c->sb.nsec_per_time_unit = le32_to_cpu(src->time_precision);
524 	c->sb.time_units_per_sec = NSEC_PER_SEC / c->sb.nsec_per_time_unit;
525 
526 	/* XXX this is wrong, we need a 96 or 128 bit integer type */
527 	c->sb.time_base_lo	= div_u64(le64_to_cpu(src->time_base_lo),
528 					  c->sb.nsec_per_time_unit);
529 	c->sb.time_base_hi	= le32_to_cpu(src->time_base_hi);
530 
531 	c->sb.features		= le64_to_cpu(src->features[0]);
532 	c->sb.compat		= le64_to_cpu(src->compat[0]);
533 
534 	memset(c->sb.errors_silent, 0, sizeof(c->sb.errors_silent));
535 
536 	struct bch_sb_field_ext *ext = bch2_sb_field_get(src, ext);
537 	if (ext) {
538 		le_bitvector_to_cpu(c->sb.errors_silent, (void *) ext->errors_silent,
539 				    sizeof(c->sb.errors_silent) * 8);
540 		c->sb.btrees_lost_data = le64_to_cpu(ext->btrees_lost_data);
541 	}
542 
543 	for_each_member_device(c, ca) {
544 		struct bch_member m = bch2_sb_member_get(src, ca->dev_idx);
545 		ca->mi = bch2_mi_to_cpu(&m);
546 	}
547 }
548 
549 static int __copy_super(struct bch_sb_handle *dst_handle, struct bch_sb *src)
550 {
551 	struct bch_sb_field *src_f, *dst_f;
552 	struct bch_sb *dst = dst_handle->sb;
553 	unsigned i;
554 
555 	dst->version		= src->version;
556 	dst->version_min	= src->version_min;
557 	dst->seq		= src->seq;
558 	dst->uuid		= src->uuid;
559 	dst->user_uuid		= src->user_uuid;
560 	memcpy(dst->label,	src->label, sizeof(dst->label));
561 
562 	dst->block_size		= src->block_size;
563 	dst->nr_devices		= src->nr_devices;
564 
565 	dst->time_base_lo	= src->time_base_lo;
566 	dst->time_base_hi	= src->time_base_hi;
567 	dst->time_precision	= src->time_precision;
568 	dst->write_time		= src->write_time;
569 
570 	memcpy(dst->flags,	src->flags,	sizeof(dst->flags));
571 	memcpy(dst->features,	src->features,	sizeof(dst->features));
572 	memcpy(dst->compat,	src->compat,	sizeof(dst->compat));
573 
574 	for (i = 0; i < BCH_SB_FIELD_NR; i++) {
575 		int d;
576 
577 		if ((1U << i) & BCH_SINGLE_DEVICE_SB_FIELDS)
578 			continue;
579 
580 		src_f = bch2_sb_field_get_id(src, i);
581 		dst_f = bch2_sb_field_get_id(dst, i);
582 
583 		d = (src_f ? le32_to_cpu(src_f->u64s) : 0) -
584 		    (dst_f ? le32_to_cpu(dst_f->u64s) : 0);
585 		if (d > 0) {
586 			int ret = bch2_sb_realloc(dst_handle,
587 					le32_to_cpu(dst_handle->sb->u64s) + d);
588 
589 			if (ret)
590 				return ret;
591 
592 			dst = dst_handle->sb;
593 			dst_f = bch2_sb_field_get_id(dst, i);
594 		}
595 
596 		dst_f = __bch2_sb_field_resize(dst_handle, dst_f,
597 				src_f ? le32_to_cpu(src_f->u64s) : 0);
598 
599 		if (src_f)
600 			memcpy(dst_f, src_f, vstruct_bytes(src_f));
601 	}
602 
603 	return 0;
604 }
605 
606 int bch2_sb_to_fs(struct bch_fs *c, struct bch_sb *src)
607 {
608 	int ret;
609 
610 	lockdep_assert_held(&c->sb_lock);
611 
612 	ret =   bch2_sb_realloc(&c->disk_sb, 0) ?:
613 		__copy_super(&c->disk_sb, src) ?:
614 		bch2_sb_replicas_to_cpu_replicas(c) ?:
615 		bch2_sb_disk_groups_to_cpu(c);
616 	if (ret)
617 		return ret;
618 
619 	bch2_sb_update(c);
620 	return 0;
621 }
622 
623 int bch2_sb_from_fs(struct bch_fs *c, struct bch_dev *ca)
624 {
625 	return __copy_super(&ca->disk_sb, c->disk_sb.sb);
626 }
627 
628 /* read superblock: */
629 
630 static int read_one_super(struct bch_sb_handle *sb, u64 offset, struct printbuf *err)
631 {
632 	size_t bytes;
633 	int ret;
634 reread:
635 	bio_reset(sb->bio, sb->bdev, REQ_OP_READ|REQ_SYNC|REQ_META);
636 	sb->bio->bi_iter.bi_sector = offset;
637 	bch2_bio_map(sb->bio, sb->sb, sb->buffer_size);
638 
639 	ret = submit_bio_wait(sb->bio);
640 	if (ret) {
641 		prt_printf(err, "IO error: %i", ret);
642 		return ret;
643 	}
644 
645 	if (!uuid_equal(&sb->sb->magic, &BCACHE_MAGIC) &&
646 	    !uuid_equal(&sb->sb->magic, &BCHFS_MAGIC)) {
647 		prt_str(err, "Not a bcachefs superblock (got magic ");
648 		pr_uuid(err, sb->sb->magic.b);
649 		prt_str(err, ")");
650 		return -BCH_ERR_invalid_sb_magic;
651 	}
652 
653 	ret = bch2_sb_compatible(sb->sb, err);
654 	if (ret)
655 		return ret;
656 
657 	bytes = vstruct_bytes(sb->sb);
658 
659 	u64 sb_size = 512ULL << min(BCH_SB_LAYOUT_SIZE_BITS_MAX, sb->sb->layout.sb_max_size_bits);
660 	if (bytes > sb_size) {
661 		prt_printf(err, "Invalid superblock: too big (got %zu bytes, layout max %llu)",
662 			   bytes, sb_size);
663 		return -BCH_ERR_invalid_sb_too_big;
664 	}
665 
666 	if (bytes > sb->buffer_size) {
667 		ret = bch2_sb_realloc(sb, le32_to_cpu(sb->sb->u64s));
668 		if (ret)
669 			return ret;
670 		goto reread;
671 	}
672 
673 	enum bch_csum_type csum_type = BCH_SB_CSUM_TYPE(sb->sb);
674 	if (csum_type >= BCH_CSUM_NR) {
675 		prt_printf(err, "unknown checksum type %llu", BCH_SB_CSUM_TYPE(sb->sb));
676 		return -BCH_ERR_invalid_sb_csum_type;
677 	}
678 
679 	/* XXX: verify MACs */
680 	struct bch_csum csum = csum_vstruct(NULL, csum_type, null_nonce(), sb->sb);
681 	if (bch2_crc_cmp(csum, sb->sb->csum)) {
682 		bch2_csum_err_msg(err, csum_type, sb->sb->csum, csum);
683 		return -BCH_ERR_invalid_sb_csum;
684 	}
685 
686 	sb->seq = le64_to_cpu(sb->sb->seq);
687 
688 	return 0;
689 }
690 
691 static int __bch2_read_super(const char *path, struct bch_opts *opts,
692 		    struct bch_sb_handle *sb, bool ignore_notbchfs_msg)
693 {
694 	u64 offset = opt_get(*opts, sb);
695 	struct bch_sb_layout layout;
696 	struct printbuf err = PRINTBUF;
697 	struct printbuf err2 = PRINTBUF;
698 	__le64 *i;
699 	int ret;
700 #ifndef __KERNEL__
701 retry:
702 #endif
703 	memset(sb, 0, sizeof(*sb));
704 	sb->mode	= BLK_OPEN_READ;
705 	sb->have_bio	= true;
706 	sb->holder	= kmalloc(1, GFP_KERNEL);
707 	if (!sb->holder)
708 		return -ENOMEM;
709 
710 	sb->sb_name = kstrdup(path, GFP_KERNEL);
711 	if (!sb->sb_name) {
712 		ret = -ENOMEM;
713 		prt_printf(&err, "error allocating memory for sb_name");
714 		goto err;
715 	}
716 
717 #ifndef __KERNEL__
718 	if (opt_get(*opts, direct_io) == false)
719 		sb->mode |= BLK_OPEN_BUFFERED;
720 #endif
721 
722 	if (!opt_get(*opts, noexcl))
723 		sb->mode |= BLK_OPEN_EXCL;
724 
725 	if (!opt_get(*opts, nochanges))
726 		sb->mode |= BLK_OPEN_WRITE;
727 
728 	sb->s_bdev_file = bdev_file_open_by_path(path, sb->mode, sb->holder, &bch2_sb_handle_bdev_ops);
729 	if (IS_ERR(sb->s_bdev_file) &&
730 	    PTR_ERR(sb->s_bdev_file) == -EACCES &&
731 	    opt_get(*opts, read_only)) {
732 		sb->mode &= ~BLK_OPEN_WRITE;
733 
734 		sb->s_bdev_file = bdev_file_open_by_path(path, sb->mode, sb->holder, &bch2_sb_handle_bdev_ops);
735 		if (!IS_ERR(sb->s_bdev_file))
736 			opt_set(*opts, nochanges, true);
737 	}
738 
739 	if (IS_ERR(sb->s_bdev_file)) {
740 		ret = PTR_ERR(sb->s_bdev_file);
741 		prt_printf(&err, "error opening %s: %s", path, bch2_err_str(ret));
742 		goto err;
743 	}
744 	sb->bdev = file_bdev(sb->s_bdev_file);
745 
746 	ret = bch2_sb_realloc(sb, 0);
747 	if (ret) {
748 		prt_printf(&err, "error allocating memory for superblock");
749 		goto err;
750 	}
751 
752 	if (bch2_fs_init_fault("read_super")) {
753 		prt_printf(&err, "dynamic fault");
754 		ret = -EFAULT;
755 		goto err;
756 	}
757 
758 	ret = read_one_super(sb, offset, &err);
759 	if (!ret)
760 		goto got_super;
761 
762 	if (opt_defined(*opts, sb))
763 		goto err;
764 
765 	prt_printf(&err2, "bcachefs (%s): error reading default superblock: %s\n",
766 	       path, err.buf);
767 	if (ret == -BCH_ERR_invalid_sb_magic && ignore_notbchfs_msg)
768 		bch2_print_opts(opts, KERN_INFO "%s", err2.buf);
769 	else
770 		bch2_print_opts(opts, KERN_ERR "%s", err2.buf);
771 
772 	printbuf_exit(&err2);
773 	printbuf_reset(&err);
774 
775 	/*
776 	 * Error reading primary superblock - read location of backup
777 	 * superblocks:
778 	 */
779 	bio_reset(sb->bio, sb->bdev, REQ_OP_READ|REQ_SYNC|REQ_META);
780 	sb->bio->bi_iter.bi_sector = BCH_SB_LAYOUT_SECTOR;
781 	/*
782 	 * use sb buffer to read layout, since sb buffer is page aligned but
783 	 * layout won't be:
784 	 */
785 	bch2_bio_map(sb->bio, sb->sb, sizeof(struct bch_sb_layout));
786 
787 	ret = submit_bio_wait(sb->bio);
788 	if (ret) {
789 		prt_printf(&err, "IO error: %i", ret);
790 		goto err;
791 	}
792 
793 	memcpy(&layout, sb->sb, sizeof(layout));
794 	ret = validate_sb_layout(&layout, &err);
795 	if (ret)
796 		goto err;
797 
798 	for (i = layout.sb_offset;
799 	     i < layout.sb_offset + layout.nr_superblocks; i++) {
800 		offset = le64_to_cpu(*i);
801 
802 		if (offset == opt_get(*opts, sb)) {
803 			ret = -BCH_ERR_invalid;
804 			continue;
805 		}
806 
807 		ret = read_one_super(sb, offset, &err);
808 		if (!ret)
809 			goto got_super;
810 	}
811 
812 	goto err;
813 
814 got_super:
815 	if (le16_to_cpu(sb->sb->block_size) << 9 <
816 	    bdev_logical_block_size(sb->bdev) &&
817 	    opt_get(*opts, direct_io)) {
818 #ifndef __KERNEL__
819 		opt_set(*opts, direct_io, false);
820 		bch2_free_super(sb);
821 		goto retry;
822 #endif
823 		prt_printf(&err, "block size (%u) smaller than device block size (%u)",
824 		       le16_to_cpu(sb->sb->block_size) << 9,
825 		       bdev_logical_block_size(sb->bdev));
826 		ret = -BCH_ERR_block_size_too_small;
827 		goto err;
828 	}
829 
830 	sb->have_layout = true;
831 
832 	ret = bch2_sb_validate(sb, 0, &err);
833 	if (ret) {
834 		bch2_print_opts(opts, KERN_ERR "bcachefs (%s): error validating superblock: %s\n",
835 				path, err.buf);
836 		goto err_no_print;
837 	}
838 out:
839 	printbuf_exit(&err);
840 	return ret;
841 err:
842 	bch2_print_opts(opts, KERN_ERR "bcachefs (%s): error reading superblock: %s\n",
843 			path, err.buf);
844 err_no_print:
845 	bch2_free_super(sb);
846 	goto out;
847 }
848 
849 int bch2_read_super(const char *path, struct bch_opts *opts,
850 		    struct bch_sb_handle *sb)
851 {
852 	return __bch2_read_super(path, opts, sb, false);
853 }
854 
855 /* provide a silenced version for mount.bcachefs */
856 
857 int bch2_read_super_silent(const char *path, struct bch_opts *opts,
858 		    struct bch_sb_handle *sb)
859 {
860 	return __bch2_read_super(path, opts, sb, true);
861 }
862 
863 /* write superblock: */
864 
865 static void write_super_endio(struct bio *bio)
866 {
867 	struct bch_dev *ca = bio->bi_private;
868 
869 	/* XXX: return errors directly */
870 
871 	if (bch2_dev_io_err_on(bio->bi_status, ca,
872 			       bio_data_dir(bio)
873 			       ? BCH_MEMBER_ERROR_write
874 			       : BCH_MEMBER_ERROR_read,
875 			       "superblock %s error: %s",
876 			       bio_data_dir(bio) ? "write" : "read",
877 			       bch2_blk_status_to_str(bio->bi_status)))
878 		ca->sb_write_error = 1;
879 
880 	closure_put(&ca->fs->sb_write);
881 	percpu_ref_put(&ca->io_ref);
882 }
883 
884 static void read_back_super(struct bch_fs *c, struct bch_dev *ca)
885 {
886 	struct bch_sb *sb = ca->disk_sb.sb;
887 	struct bio *bio = ca->disk_sb.bio;
888 
889 	bio_reset(bio, ca->disk_sb.bdev, REQ_OP_READ|REQ_SYNC|REQ_META);
890 	bio->bi_iter.bi_sector	= le64_to_cpu(sb->layout.sb_offset[0]);
891 	bio->bi_end_io		= write_super_endio;
892 	bio->bi_private		= ca;
893 	bch2_bio_map(bio, ca->sb_read_scratch, PAGE_SIZE);
894 
895 	this_cpu_add(ca->io_done->sectors[READ][BCH_DATA_sb],
896 		     bio_sectors(bio));
897 
898 	percpu_ref_get(&ca->io_ref);
899 	closure_bio_submit(bio, &c->sb_write);
900 }
901 
902 static void write_one_super(struct bch_fs *c, struct bch_dev *ca, unsigned idx)
903 {
904 	struct bch_sb *sb = ca->disk_sb.sb;
905 	struct bio *bio = ca->disk_sb.bio;
906 
907 	sb->offset = sb->layout.sb_offset[idx];
908 
909 	SET_BCH_SB_CSUM_TYPE(sb, bch2_csum_opt_to_type(c->opts.metadata_checksum, false));
910 	sb->csum = csum_vstruct(c, BCH_SB_CSUM_TYPE(sb),
911 				null_nonce(), sb);
912 
913 	bio_reset(bio, ca->disk_sb.bdev, REQ_OP_WRITE|REQ_SYNC|REQ_META);
914 	bio->bi_iter.bi_sector	= le64_to_cpu(sb->offset);
915 	bio->bi_end_io		= write_super_endio;
916 	bio->bi_private		= ca;
917 	bch2_bio_map(bio, sb,
918 		     roundup((size_t) vstruct_bytes(sb),
919 			     bdev_logical_block_size(ca->disk_sb.bdev)));
920 
921 	this_cpu_add(ca->io_done->sectors[WRITE][BCH_DATA_sb],
922 		     bio_sectors(bio));
923 
924 	percpu_ref_get(&ca->io_ref);
925 	closure_bio_submit(bio, &c->sb_write);
926 }
927 
928 int bch2_write_super(struct bch_fs *c)
929 {
930 	struct closure *cl = &c->sb_write;
931 	struct printbuf err = PRINTBUF;
932 	unsigned sb = 0, nr_wrote;
933 	struct bch_devs_mask sb_written;
934 	bool wrote, can_mount_without_written, can_mount_with_written;
935 	unsigned degraded_flags = BCH_FORCE_IF_DEGRADED;
936 	DARRAY(struct bch_dev *) online_devices = {};
937 	int ret = 0;
938 
939 	trace_and_count(c, write_super, c, _RET_IP_);
940 
941 	if (c->opts.very_degraded)
942 		degraded_flags |= BCH_FORCE_IF_LOST;
943 
944 	lockdep_assert_held(&c->sb_lock);
945 
946 	closure_init_stack(cl);
947 	memset(&sb_written, 0, sizeof(sb_written));
948 
949 	for_each_online_member(c, ca) {
950 		ret = darray_push(&online_devices, ca);
951 		if (bch2_fs_fatal_err_on(ret, c, "%s: error allocating online devices", __func__)) {
952 			percpu_ref_put(&ca->io_ref);
953 			goto out;
954 		}
955 		percpu_ref_get(&ca->io_ref);
956 	}
957 
958 	/* Make sure we're using the new magic numbers: */
959 	c->disk_sb.sb->magic = BCHFS_MAGIC;
960 	c->disk_sb.sb->layout.magic = BCHFS_MAGIC;
961 
962 	le64_add_cpu(&c->disk_sb.sb->seq, 1);
963 
964 	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
965 	darray_for_each(online_devices, ca)
966 		__bch2_members_v2_get_mut(mi, (*ca)->dev_idx)->seq = c->disk_sb.sb->seq;
967 	c->disk_sb.sb->write_time = cpu_to_le64(ktime_get_real_seconds());
968 
969 	if (test_bit(BCH_FS_error, &c->flags))
970 		SET_BCH_SB_HAS_ERRORS(c->disk_sb.sb, 1);
971 	if (test_bit(BCH_FS_topology_error, &c->flags))
972 		SET_BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb, 1);
973 
974 	SET_BCH_SB_BIG_ENDIAN(c->disk_sb.sb, CPU_BIG_ENDIAN);
975 
976 	bch2_sb_counters_from_cpu(c);
977 	bch2_sb_members_from_cpu(c);
978 	bch2_sb_members_cpy_v2_v1(&c->disk_sb);
979 	bch2_sb_errors_from_cpu(c);
980 	bch2_sb_downgrade_update(c);
981 
982 	darray_for_each(online_devices, ca)
983 		bch2_sb_from_fs(c, (*ca));
984 
985 	darray_for_each(online_devices, ca) {
986 		printbuf_reset(&err);
987 
988 		ret = bch2_sb_validate(&(*ca)->disk_sb, BCH_VALIDATE_write, &err);
989 		if (ret) {
990 			bch2_fs_inconsistent(c, "sb invalid before write: %s", err.buf);
991 			goto out;
992 		}
993 	}
994 
995 	if (c->opts.nochanges)
996 		goto out;
997 
998 	/*
999 	 * Defer writing the superblock until filesystem initialization is
1000 	 * complete - don't write out a partly initialized superblock:
1001 	 */
1002 	if (!BCH_SB_INITIALIZED(c->disk_sb.sb))
1003 		goto out;
1004 
1005 	if (le16_to_cpu(c->disk_sb.sb->version) > bcachefs_metadata_version_current) {
1006 		struct printbuf buf = PRINTBUF;
1007 		prt_printf(&buf, "attempting to write superblock that wasn't version downgraded (");
1008 		bch2_version_to_text(&buf, le16_to_cpu(c->disk_sb.sb->version));
1009 		prt_str(&buf, " > ");
1010 		bch2_version_to_text(&buf, bcachefs_metadata_version_current);
1011 		prt_str(&buf, ")");
1012 		bch2_fs_fatal_error(c, ": %s", buf.buf);
1013 		printbuf_exit(&buf);
1014 		return -BCH_ERR_sb_not_downgraded;
1015 	}
1016 
1017 	darray_for_each(online_devices, ca) {
1018 		__set_bit((*ca)->dev_idx, sb_written.d);
1019 		(*ca)->sb_write_error = 0;
1020 	}
1021 
1022 	darray_for_each(online_devices, ca)
1023 		read_back_super(c, *ca);
1024 	closure_sync(cl);
1025 
1026 	darray_for_each(online_devices, cap) {
1027 		struct bch_dev *ca = *cap;
1028 
1029 		if (ca->sb_write_error)
1030 			continue;
1031 
1032 		if (le64_to_cpu(ca->sb_read_scratch->seq) < ca->disk_sb.seq) {
1033 			struct printbuf buf = PRINTBUF;
1034 			prt_char(&buf, ' ');
1035 			prt_bdevname(&buf, ca->disk_sb.bdev);
1036 			prt_printf(&buf,
1037 				": Superblock write was silently dropped! (seq %llu expected %llu)",
1038 				le64_to_cpu(ca->sb_read_scratch->seq),
1039 				ca->disk_sb.seq);
1040 			bch2_fs_fatal_error(c, "%s", buf.buf);
1041 			printbuf_exit(&buf);
1042 			ret = -BCH_ERR_erofs_sb_err;
1043 		}
1044 
1045 		if (le64_to_cpu(ca->sb_read_scratch->seq) > ca->disk_sb.seq) {
1046 			struct printbuf buf = PRINTBUF;
1047 			prt_char(&buf, ' ');
1048 			prt_bdevname(&buf, ca->disk_sb.bdev);
1049 			prt_printf(&buf,
1050 				": Superblock modified by another process (seq %llu expected %llu)",
1051 				le64_to_cpu(ca->sb_read_scratch->seq),
1052 				ca->disk_sb.seq);
1053 			bch2_fs_fatal_error(c, "%s", buf.buf);
1054 			printbuf_exit(&buf);
1055 			ret = -BCH_ERR_erofs_sb_err;
1056 		}
1057 	}
1058 
1059 	if (ret)
1060 		goto out;
1061 
1062 	do {
1063 		wrote = false;
1064 		darray_for_each(online_devices, cap) {
1065 			struct bch_dev *ca = *cap;
1066 			if (!ca->sb_write_error &&
1067 			    sb < ca->disk_sb.sb->layout.nr_superblocks) {
1068 				write_one_super(c, ca, sb);
1069 				wrote = true;
1070 			}
1071 		}
1072 		closure_sync(cl);
1073 		sb++;
1074 	} while (wrote);
1075 
1076 	darray_for_each(online_devices, cap) {
1077 		struct bch_dev *ca = *cap;
1078 		if (ca->sb_write_error)
1079 			__clear_bit(ca->dev_idx, sb_written.d);
1080 		else
1081 			ca->disk_sb.seq = le64_to_cpu(ca->disk_sb.sb->seq);
1082 	}
1083 
1084 	nr_wrote = dev_mask_nr(&sb_written);
1085 
1086 	can_mount_with_written =
1087 		bch2_have_enough_devs(c, sb_written, degraded_flags, false);
1088 
1089 	for (unsigned i = 0; i < ARRAY_SIZE(sb_written.d); i++)
1090 		sb_written.d[i] = ~sb_written.d[i];
1091 
1092 	can_mount_without_written =
1093 		bch2_have_enough_devs(c, sb_written, degraded_flags, false);
1094 
1095 	/*
1096 	 * If we would be able to mount _without_ the devices we successfully
1097 	 * wrote superblocks to, we weren't able to write to enough devices:
1098 	 *
1099 	 * Exception: if we can mount without the successes because we haven't
1100 	 * written anything (new filesystem), we continue if we'd be able to
1101 	 * mount with the devices we did successfully write to:
1102 	 */
1103 	if (bch2_fs_fatal_err_on(!nr_wrote ||
1104 				 !can_mount_with_written ||
1105 				 (can_mount_without_written &&
1106 				  !can_mount_with_written), c,
1107 		": Unable to write superblock to sufficient devices (from %ps)",
1108 		(void *) _RET_IP_))
1109 		ret = -1;
1110 out:
1111 	/* Make new options visible after they're persistent: */
1112 	bch2_sb_update(c);
1113 	darray_for_each(online_devices, ca)
1114 		percpu_ref_put(&(*ca)->io_ref);
1115 	darray_exit(&online_devices);
1116 	printbuf_exit(&err);
1117 	return ret;
1118 }
1119 
1120 void __bch2_check_set_feature(struct bch_fs *c, unsigned feat)
1121 {
1122 	mutex_lock(&c->sb_lock);
1123 	if (!(c->sb.features & (1ULL << feat))) {
1124 		c->disk_sb.sb->features[0] |= cpu_to_le64(1ULL << feat);
1125 
1126 		bch2_write_super(c);
1127 	}
1128 	mutex_unlock(&c->sb_lock);
1129 }
1130 
1131 /* Downgrade if superblock is at a higher version than currently supported: */
1132 bool bch2_check_version_downgrade(struct bch_fs *c)
1133 {
1134 	bool ret = bcachefs_metadata_version_current < c->sb.version;
1135 
1136 	lockdep_assert_held(&c->sb_lock);
1137 
1138 	/*
1139 	 * Downgrade, if superblock is at a higher version than currently
1140 	 * supported:
1141 	 *
1142 	 * c->sb will be checked before we write the superblock, so update it as
1143 	 * well:
1144 	 */
1145 	if (BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb) > bcachefs_metadata_version_current)
1146 		SET_BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb, bcachefs_metadata_version_current);
1147 	if (c->sb.version > bcachefs_metadata_version_current)
1148 		c->disk_sb.sb->version = cpu_to_le16(bcachefs_metadata_version_current);
1149 	if (c->sb.version_min > bcachefs_metadata_version_current)
1150 		c->disk_sb.sb->version_min = cpu_to_le16(bcachefs_metadata_version_current);
1151 	c->disk_sb.sb->compat[0] &= cpu_to_le64((1ULL << BCH_COMPAT_NR) - 1);
1152 	return ret;
1153 }
1154 
1155 void bch2_sb_upgrade(struct bch_fs *c, unsigned new_version)
1156 {
1157 	lockdep_assert_held(&c->sb_lock);
1158 
1159 	if (BCH_VERSION_MAJOR(new_version) >
1160 	    BCH_VERSION_MAJOR(le16_to_cpu(c->disk_sb.sb->version)))
1161 		bch2_sb_field_resize(&c->disk_sb, downgrade, 0);
1162 
1163 	c->disk_sb.sb->version = cpu_to_le16(new_version);
1164 	c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALL);
1165 }
1166 
1167 static int bch2_sb_ext_validate(struct bch_sb *sb, struct bch_sb_field *f,
1168 				enum bch_validate_flags flags, struct printbuf *err)
1169 {
1170 	if (vstruct_bytes(f) < 88) {
1171 		prt_printf(err, "field too small (%zu < %u)", vstruct_bytes(f), 88);
1172 		return -BCH_ERR_invalid_sb_ext;
1173 	}
1174 
1175 	return 0;
1176 }
1177 
1178 static void bch2_sb_ext_to_text(struct printbuf *out, struct bch_sb *sb,
1179 				struct bch_sb_field *f)
1180 {
1181 	struct bch_sb_field_ext *e = field_to_type(f, ext);
1182 
1183 	prt_printf(out, "Recovery passes required:\t");
1184 	prt_bitflags(out, bch2_recovery_passes,
1185 		     bch2_recovery_passes_from_stable(le64_to_cpu(e->recovery_passes_required[0])));
1186 	prt_newline(out);
1187 
1188 	unsigned long *errors_silent = kmalloc(sizeof(e->errors_silent), GFP_KERNEL);
1189 	if (errors_silent) {
1190 		le_bitvector_to_cpu(errors_silent, (void *) e->errors_silent, sizeof(e->errors_silent) * 8);
1191 
1192 		prt_printf(out, "Errors to silently fix:\t");
1193 		prt_bitflags_vector(out, bch2_sb_error_strs, errors_silent,
1194 				    min(BCH_FSCK_ERR_MAX, sizeof(e->errors_silent) * 8));
1195 		prt_newline(out);
1196 
1197 		kfree(errors_silent);
1198 	}
1199 
1200 	prt_printf(out, "Btrees with missing data:\t");
1201 	prt_bitflags(out, __bch2_btree_ids, le64_to_cpu(e->btrees_lost_data));
1202 	prt_newline(out);
1203 }
1204 
1205 static const struct bch_sb_field_ops bch_sb_field_ops_ext = {
1206 	.validate	= bch2_sb_ext_validate,
1207 	.to_text	= bch2_sb_ext_to_text,
1208 };
1209 
1210 static const struct bch_sb_field_ops *bch2_sb_field_ops[] = {
1211 #define x(f, nr)					\
1212 	[BCH_SB_FIELD_##f] = &bch_sb_field_ops_##f,
1213 	BCH_SB_FIELDS()
1214 #undef x
1215 };
1216 
1217 static const struct bch_sb_field_ops bch2_sb_field_null_ops;
1218 
1219 static const struct bch_sb_field_ops *bch2_sb_field_type_ops(unsigned type)
1220 {
1221 	return likely(type < ARRAY_SIZE(bch2_sb_field_ops))
1222 		? bch2_sb_field_ops[type]
1223 		: &bch2_sb_field_null_ops;
1224 }
1225 
1226 static int bch2_sb_field_validate(struct bch_sb *sb, struct bch_sb_field *f,
1227 				  enum bch_validate_flags flags, struct printbuf *err)
1228 {
1229 	unsigned type = le32_to_cpu(f->type);
1230 	struct printbuf field_err = PRINTBUF;
1231 	const struct bch_sb_field_ops *ops = bch2_sb_field_type_ops(type);
1232 	int ret;
1233 
1234 	ret = ops->validate ? ops->validate(sb, f, flags, &field_err) : 0;
1235 	if (ret) {
1236 		prt_printf(err, "Invalid superblock section %s: %s",
1237 			   bch2_sb_fields[type], field_err.buf);
1238 		prt_newline(err);
1239 		bch2_sb_field_to_text(err, sb, f);
1240 	}
1241 
1242 	printbuf_exit(&field_err);
1243 	return ret;
1244 }
1245 
1246 void __bch2_sb_field_to_text(struct printbuf *out, struct bch_sb *sb,
1247 			     struct bch_sb_field *f)
1248 {
1249 	unsigned type = le32_to_cpu(f->type);
1250 	const struct bch_sb_field_ops *ops = bch2_sb_field_type_ops(type);
1251 
1252 	if (!out->nr_tabstops)
1253 		printbuf_tabstop_push(out, 32);
1254 
1255 	if (ops->to_text)
1256 		ops->to_text(out, sb, f);
1257 }
1258 
1259 void bch2_sb_field_to_text(struct printbuf *out, struct bch_sb *sb,
1260 			   struct bch_sb_field *f)
1261 {
1262 	unsigned type = le32_to_cpu(f->type);
1263 
1264 	if (type < BCH_SB_FIELD_NR)
1265 		prt_printf(out, "%s", bch2_sb_fields[type]);
1266 	else
1267 		prt_printf(out, "(unknown field %u)", type);
1268 
1269 	prt_printf(out, " (size %zu):", vstruct_bytes(f));
1270 	prt_newline(out);
1271 
1272 	__bch2_sb_field_to_text(out, sb, f);
1273 }
1274 
1275 void bch2_sb_layout_to_text(struct printbuf *out, struct bch_sb_layout *l)
1276 {
1277 	unsigned i;
1278 
1279 	prt_printf(out, "Type:                    %u", l->layout_type);
1280 	prt_newline(out);
1281 
1282 	prt_str(out, "Superblock max size:     ");
1283 	prt_units_u64(out, 512 << l->sb_max_size_bits);
1284 	prt_newline(out);
1285 
1286 	prt_printf(out, "Nr superblocks:          %u", l->nr_superblocks);
1287 	prt_newline(out);
1288 
1289 	prt_str(out, "Offsets:                 ");
1290 	for (i = 0; i < l->nr_superblocks; i++) {
1291 		if (i)
1292 			prt_str(out, ", ");
1293 		prt_printf(out, "%llu", le64_to_cpu(l->sb_offset[i]));
1294 	}
1295 	prt_newline(out);
1296 }
1297 
1298 void bch2_sb_to_text(struct printbuf *out, struct bch_sb *sb,
1299 		     bool print_layout, unsigned fields)
1300 {
1301 	if (!out->nr_tabstops)
1302 		printbuf_tabstop_push(out, 44);
1303 
1304 	prt_printf(out, "External UUID:\t");
1305 	pr_uuid(out, sb->user_uuid.b);
1306 	prt_newline(out);
1307 
1308 	prt_printf(out, "Internal UUID:\t");
1309 	pr_uuid(out, sb->uuid.b);
1310 	prt_newline(out);
1311 
1312 	prt_printf(out, "Magic number:\t");
1313 	pr_uuid(out, sb->magic.b);
1314 	prt_newline(out);
1315 
1316 	prt_printf(out, "Device index:\t%u\n", sb->dev_idx);
1317 
1318 	prt_printf(out, "Label:\t");
1319 	if (!strlen(sb->label))
1320 		prt_printf(out, "(none)");
1321 	else
1322 		prt_printf(out, "%.*s", (int) sizeof(sb->label), sb->label);
1323 	prt_newline(out);
1324 
1325 	prt_printf(out, "Version:\t");
1326 	bch2_version_to_text(out, le16_to_cpu(sb->version));
1327 	prt_newline(out);
1328 
1329 	prt_printf(out, "Version upgrade complete:\t");
1330 	bch2_version_to_text(out, BCH_SB_VERSION_UPGRADE_COMPLETE(sb));
1331 	prt_newline(out);
1332 
1333 	prt_printf(out, "Oldest version on disk:\t");
1334 	bch2_version_to_text(out, le16_to_cpu(sb->version_min));
1335 	prt_newline(out);
1336 
1337 	prt_printf(out, "Created:\t");
1338 	if (sb->time_base_lo)
1339 		bch2_prt_datetime(out, div_u64(le64_to_cpu(sb->time_base_lo), NSEC_PER_SEC));
1340 	else
1341 		prt_printf(out, "(not set)");
1342 	prt_newline(out);
1343 
1344 	prt_printf(out, "Sequence number:\t");
1345 	prt_printf(out, "%llu", le64_to_cpu(sb->seq));
1346 	prt_newline(out);
1347 
1348 	prt_printf(out, "Time of last write:\t");
1349 	bch2_prt_datetime(out, le64_to_cpu(sb->write_time));
1350 	prt_newline(out);
1351 
1352 	prt_printf(out, "Superblock size:\t");
1353 	prt_units_u64(out, vstruct_bytes(sb));
1354 	prt_str(out, "/");
1355 	prt_units_u64(out, 512ULL << sb->layout.sb_max_size_bits);
1356 	prt_newline(out);
1357 
1358 	prt_printf(out, "Clean:\t%llu\n", BCH_SB_CLEAN(sb));
1359 	prt_printf(out, "Devices:\t%u\n", bch2_sb_nr_devices(sb));
1360 
1361 	prt_printf(out, "Sections:\t");
1362 	u64 fields_have = 0;
1363 	vstruct_for_each(sb, f)
1364 		fields_have |= 1 << le32_to_cpu(f->type);
1365 	prt_bitflags(out, bch2_sb_fields, fields_have);
1366 	prt_newline(out);
1367 
1368 	prt_printf(out, "Features:\t");
1369 	prt_bitflags(out, bch2_sb_features, le64_to_cpu(sb->features[0]));
1370 	prt_newline(out);
1371 
1372 	prt_printf(out, "Compat features:\t");
1373 	prt_bitflags(out, bch2_sb_compat, le64_to_cpu(sb->compat[0]));
1374 	prt_newline(out);
1375 
1376 	prt_newline(out);
1377 	prt_printf(out, "Options:");
1378 	prt_newline(out);
1379 	printbuf_indent_add(out, 2);
1380 	{
1381 		enum bch_opt_id id;
1382 
1383 		for (id = 0; id < bch2_opts_nr; id++) {
1384 			const struct bch_option *opt = bch2_opt_table + id;
1385 
1386 			if (opt->get_sb != BCH2_NO_SB_OPT) {
1387 				u64 v = bch2_opt_from_sb(sb, id);
1388 
1389 				prt_printf(out, "%s:\t", opt->attr.name);
1390 				bch2_opt_to_text(out, NULL, sb, opt, v,
1391 						 OPT_HUMAN_READABLE|OPT_SHOW_FULL_LIST);
1392 				prt_newline(out);
1393 			}
1394 		}
1395 	}
1396 
1397 	printbuf_indent_sub(out, 2);
1398 
1399 	if (print_layout) {
1400 		prt_newline(out);
1401 		prt_printf(out, "layout:");
1402 		prt_newline(out);
1403 		printbuf_indent_add(out, 2);
1404 		bch2_sb_layout_to_text(out, &sb->layout);
1405 		printbuf_indent_sub(out, 2);
1406 	}
1407 
1408 	vstruct_for_each(sb, f)
1409 		if (fields & (1 << le32_to_cpu(f->type))) {
1410 			prt_newline(out);
1411 			bch2_sb_field_to_text(out, sb, f);
1412 		}
1413 }
1414