[thirdparty/linux.git] / fs / bcachefs / sb-members.c

// SPDX-License-Identifier: GPL-2.0

#include "bcachefs.h"
#include "btree_cache.h"
#include "disk_groups.h"
#include "error.h"
#include "opts.h"
#include "replicas.h"
#include "sb-members.h"
#include "super-io.h"

void bch2_dev_missing(struct bch_fs *c, unsigned dev)
{
	bch2_fs_inconsistent(c, "pointer to nonexistent device %u", dev);
}

void bch2_dev_bucket_missing(struct bch_fs *c, struct bpos bucket)
{
	bch2_fs_inconsistent(c, "pointer to nonexistent bucket %llu:%llu", bucket.inode, bucket.offset);
}

#define x(t, n, ...) [n] = #t,
static const char * const bch2_iops_measurements[] = {
	BCH_IOPS_MEASUREMENTS()
	NULL
};

char * const bch2_member_error_strs[] = {
	BCH_MEMBER_ERROR_TYPES()
	NULL
};
#undef x

/* Code for bch_sb_field_members_v1: */

struct bch_member *bch2_members_v2_get_mut(struct bch_sb *sb, int i)
{
	return __bch2_members_v2_get_mut(bch2_sb_field_get(sb, members_v2), i);
}

static struct bch_member members_v2_get(struct bch_sb_field_members_v2 *mi, int i)
{
	struct bch_member ret, *p = __bch2_members_v2_get_mut(mi, i);
	memset(&ret, 0, sizeof(ret));
	memcpy(&ret, p, min_t(size_t, le16_to_cpu(mi->member_bytes), sizeof(ret)));
	return ret;
}

static struct bch_member *members_v1_get_mut(struct bch_sb_field_members_v1 *mi, int i)
{
	return (void *) mi->_members + (i * BCH_MEMBER_V1_BYTES);
}

static struct bch_member members_v1_get(struct bch_sb_field_members_v1 *mi, int i)
{
	struct bch_member ret, *p = members_v1_get_mut(mi, i);
	memset(&ret, 0, sizeof(ret));
	memcpy(&ret, p, min_t(size_t, BCH_MEMBER_V1_BYTES, sizeof(ret)));
	return ret;
}

struct bch_member bch2_sb_member_get(struct bch_sb *sb, int i)
{
	struct bch_sb_field_members_v2 *mi2 = bch2_sb_field_get(sb, members_v2);
	if (mi2)
		return members_v2_get(mi2, i);
	struct bch_sb_field_members_v1 *mi1 = bch2_sb_field_get(sb, members_v1);
	return members_v1_get(mi1, i);
}

static int sb_members_v2_resize_entries(struct bch_fs *c)
{
	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);

	if (le16_to_cpu(mi->member_bytes) < sizeof(struct bch_member)) {
		unsigned u64s = DIV_ROUND_UP((sizeof(*mi) + sizeof(mi->_members[0]) *
					      c->disk_sb.sb->nr_devices), 8);

		mi = bch2_sb_field_resize(&c->disk_sb, members_v2, u64s);
		if (!mi)
			return -BCH_ERR_ENOSPC_sb_members_v2;

		for (int i = c->disk_sb.sb->nr_devices - 1; i >= 0; --i) {
			void *dst = (void *) mi->_members + (i * sizeof(struct bch_member));
			memmove(dst, __bch2_members_v2_get_mut(mi, i), le16_to_cpu(mi->member_bytes));
			memset(dst + le16_to_cpu(mi->member_bytes),
			       0, (sizeof(struct bch_member) - le16_to_cpu(mi->member_bytes)));
		}
		mi->member_bytes = cpu_to_le16(sizeof(struct bch_member));
	}
	return 0;
}

int bch2_sb_members_v2_init(struct bch_fs *c)
{
	struct bch_sb_field_members_v1 *mi1;
	struct bch_sb_field_members_v2 *mi2;

	if (!bch2_sb_field_get(c->disk_sb.sb, members_v2)) {
		mi2 = bch2_sb_field_resize(&c->disk_sb, members_v2,
				DIV_ROUND_UP(sizeof(*mi2) +
					     sizeof(struct bch_member) * c->sb.nr_devices,
					     sizeof(u64)));
		mi1 = bch2_sb_field_get(c->disk_sb.sb, members_v1);
		memcpy(&mi2->_members[0], &mi1->_members[0],
		       BCH_MEMBER_V1_BYTES * c->sb.nr_devices);
		memset(&mi2->pad[0], 0, sizeof(mi2->pad));
		mi2->member_bytes = cpu_to_le16(BCH_MEMBER_V1_BYTES);
	}

	return sb_members_v2_resize_entries(c);
}

int bch2_sb_members_cpy_v2_v1(struct bch_sb_handle *disk_sb)
{
	struct bch_sb_field_members_v1 *mi1;
	struct bch_sb_field_members_v2 *mi2;

	mi1 = bch2_sb_field_resize(disk_sb, members_v1,
			DIV_ROUND_UP(sizeof(*mi1) + BCH_MEMBER_V1_BYTES *
				     disk_sb->sb->nr_devices, sizeof(u64)));
	if (!mi1)
		return -BCH_ERR_ENOSPC_sb_members;

	mi2 = bch2_sb_field_get(disk_sb->sb, members_v2);

	for (unsigned i = 0; i < disk_sb->sb->nr_devices; i++)
		memcpy(members_v1_get_mut(mi1, i), __bch2_members_v2_get_mut(mi2, i), BCH_MEMBER_V1_BYTES);

	return 0;
}

static int validate_member(struct printbuf *err,
			   struct bch_member m,
			   struct bch_sb *sb,
			   int i)
{
	if (le64_to_cpu(m.nbuckets) > BCH_MEMBER_NBUCKETS_MAX) {
		prt_printf(err, "device %u: too many buckets (got %llu, max %u)",
			   i, le64_to_cpu(m.nbuckets), BCH_MEMBER_NBUCKETS_MAX);
		return -BCH_ERR_invalid_sb_members;
	}

	if (le64_to_cpu(m.nbuckets) -
	    le16_to_cpu(m.first_bucket) < BCH_MIN_NR_NBUCKETS) {
		prt_printf(err, "device %u: not enough buckets (got %llu, max %u)",
			   i, le64_to_cpu(m.nbuckets), BCH_MIN_NR_NBUCKETS);
		return -BCH_ERR_invalid_sb_members;
	}

	if (le16_to_cpu(m.bucket_size) <
	    le16_to_cpu(sb->block_size)) {
		prt_printf(err, "device %u: bucket size %u smaller than block size %u",
			   i, le16_to_cpu(m.bucket_size), le16_to_cpu(sb->block_size));
		return -BCH_ERR_invalid_sb_members;
	}

	if (le16_to_cpu(m.bucket_size) <
	    BCH_SB_BTREE_NODE_SIZE(sb)) {
		prt_printf(err, "device %u: bucket size %u smaller than btree node size %llu",
			   i, le16_to_cpu(m.bucket_size), BCH_SB_BTREE_NODE_SIZE(sb));
		return -BCH_ERR_invalid_sb_members;
	}

	return 0;
}

static void member_to_text(struct printbuf *out,
			   struct bch_member m,
			   struct bch_sb_field_disk_groups *gi,
			   struct bch_sb *sb,
			   int i)
{
	unsigned data_have = bch2_sb_dev_has_data(sb, i);
	u64 bucket_size = le16_to_cpu(m.bucket_size);
	u64 device_size = le64_to_cpu(m.nbuckets) * bucket_size;

	if (!bch2_member_alive(&m))
		return;

	prt_printf(out, "Device:\t%u\n", i);

	printbuf_indent_add(out, 2);

	prt_printf(out, "Label:\t");
	if (BCH_MEMBER_GROUP(&m)) {
		unsigned idx = BCH_MEMBER_GROUP(&m) - 1;

		if (idx < disk_groups_nr(gi))
			prt_printf(out, "%s (%u)",
				   gi->entries[idx].label, idx);
		else
			prt_printf(out, "(bad disk labels section)");
	} else {
		prt_printf(out, "(none)");
	}
	prt_newline(out);

	prt_printf(out, "UUID:\t");
	pr_uuid(out, m.uuid.b);
	prt_newline(out);

	prt_printf(out, "Size:\t");
	prt_units_u64(out, device_size << 9);
	prt_newline(out);

	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++)
		prt_printf(out, "%s errors:\t%llu\n", bch2_member_error_strs[i], le64_to_cpu(m.errors[i]));

	for (unsigned i = 0; i < BCH_IOPS_NR; i++)
		prt_printf(out, "%s iops:\t%u\n", bch2_iops_measurements[i], le32_to_cpu(m.iops[i]));

	prt_printf(out, "Bucket size:\t");
	prt_units_u64(out, bucket_size << 9);
	prt_newline(out);

	prt_printf(out, "First bucket:\t%u\n", le16_to_cpu(m.first_bucket));
	prt_printf(out, "Buckets:\t%llu\n", le64_to_cpu(m.nbuckets));

	prt_printf(out, "Last mount:\t");
	if (m.last_mount)
		bch2_prt_datetime(out, le64_to_cpu(m.last_mount));
	else
		prt_printf(out, "(never)");
	prt_newline(out);

	prt_printf(out, "Last superblock write:\t%llu\n", le64_to_cpu(m.seq));

	prt_printf(out, "State:\t%s\n",
		   BCH_MEMBER_STATE(&m) < BCH_MEMBER_STATE_NR
		   ? bch2_member_states[BCH_MEMBER_STATE(&m)]
		   : "unknown");

	prt_printf(out, "Data allowed:\t");
	if (BCH_MEMBER_DATA_ALLOWED(&m))
		prt_bitflags(out, __bch2_data_types, BCH_MEMBER_DATA_ALLOWED(&m));
	else
		prt_printf(out, "(none)");
	prt_newline(out);

	prt_printf(out, "Has data:\t");
	if (data_have)
		prt_bitflags(out, __bch2_data_types, data_have);
	else
		prt_printf(out, "(none)");
	prt_newline(out);

	prt_printf(out, "Btree allocated bitmap blocksize:\t");
	prt_units_u64(out, 1ULL << m.btree_bitmap_shift);
	prt_newline(out);

	prt_printf(out, "Btree allocated bitmap:\t");
	bch2_prt_u64_base2_nbits(out, le64_to_cpu(m.btree_allocated_bitmap), 64);
	prt_newline(out);

	prt_printf(out, "Durability:\t%llu\n", BCH_MEMBER_DURABILITY(&m) ? BCH_MEMBER_DURABILITY(&m) - 1 : 1);

	prt_printf(out, "Discard:\t%llu\n", BCH_MEMBER_DISCARD(&m));
	prt_printf(out, "Freespace initialized:\t%llu\n", BCH_MEMBER_FREESPACE_INITIALIZED(&m));

	printbuf_indent_sub(out, 2);
}

static int bch2_sb_members_v1_validate(struct bch_sb *sb, struct bch_sb_field *f,
				enum bch_validate_flags flags, struct printbuf *err)
{
	struct bch_sb_field_members_v1 *mi = field_to_type(f, members_v1);
	unsigned i;

	if ((void *) members_v1_get_mut(mi, sb->nr_devices) > vstruct_end(&mi->field)) {
		prt_printf(err, "too many devices for section size");
		return -BCH_ERR_invalid_sb_members;
	}

	for (i = 0; i < sb->nr_devices; i++) {
		struct bch_member m = members_v1_get(mi, i);

		int ret = validate_member(err, m, sb, i);
		if (ret)
			return ret;
	}

	return 0;
}

static void bch2_sb_members_v1_to_text(struct printbuf *out, struct bch_sb *sb,
				       struct bch_sb_field *f)
{
	struct bch_sb_field_members_v1 *mi = field_to_type(f, members_v1);
	struct bch_sb_field_disk_groups *gi = bch2_sb_field_get(sb, disk_groups);
	unsigned i;

	for (i = 0; i < sb->nr_devices; i++)
		member_to_text(out, members_v1_get(mi, i), gi, sb, i);
}

const struct bch_sb_field_ops bch_sb_field_ops_members_v1 = {
	.validate	= bch2_sb_members_v1_validate,
	.to_text	= bch2_sb_members_v1_to_text,
};

static void bch2_sb_members_v2_to_text(struct printbuf *out, struct bch_sb *sb,
				       struct bch_sb_field *f)
{
	struct bch_sb_field_members_v2 *mi = field_to_type(f, members_v2);
	struct bch_sb_field_disk_groups *gi = bch2_sb_field_get(sb, disk_groups);
	unsigned i;

	for (i = 0; i < sb->nr_devices; i++)
		member_to_text(out, members_v2_get(mi, i), gi, sb, i);
}

static int bch2_sb_members_v2_validate(struct bch_sb *sb, struct bch_sb_field *f,
				enum bch_validate_flags flags, struct printbuf *err)
{
	struct bch_sb_field_members_v2 *mi = field_to_type(f, members_v2);
	size_t mi_bytes = (void *) __bch2_members_v2_get_mut(mi, sb->nr_devices) -
		(void *) mi;

	if (mi_bytes > vstruct_bytes(&mi->field)) {
		prt_printf(err, "section too small (%zu > %zu)",
			   mi_bytes, vstruct_bytes(&mi->field));
		return -BCH_ERR_invalid_sb_members;
	}

	for (unsigned i = 0; i < sb->nr_devices; i++) {
		int ret = validate_member(err, members_v2_get(mi, i), sb, i);
		if (ret)
			return ret;
	}

	return 0;
}

const struct bch_sb_field_ops bch_sb_field_ops_members_v2 = {
	.validate	= bch2_sb_members_v2_validate,
	.to_text	= bch2_sb_members_v2_to_text,
};

void bch2_sb_members_from_cpu(struct bch_fs *c)
{
	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);

	rcu_read_lock();
	for_each_member_device_rcu(c, ca, NULL) {
		struct bch_member *m = __bch2_members_v2_get_mut(mi, ca->dev_idx);

		for (unsigned e = 0; e < BCH_MEMBER_ERROR_NR; e++)
			m->errors[e] = cpu_to_le64(atomic64_read(&ca->errors[e]));
	}
	rcu_read_unlock();
}

void bch2_dev_io_errors_to_text(struct printbuf *out, struct bch_dev *ca)
{
	struct bch_fs *c = ca->fs;
	struct bch_member m;

	mutex_lock(&ca->fs->sb_lock);
	m = bch2_sb_member_get(c->disk_sb.sb, ca->dev_idx);
	mutex_unlock(&ca->fs->sb_lock);

	printbuf_tabstop_push(out, 12);

	prt_str(out, "IO errors since filesystem creation");
	prt_newline(out);

	printbuf_indent_add(out, 2);
	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++)
		prt_printf(out, "%s:\t%llu\n", bch2_member_error_strs[i], atomic64_read(&ca->errors[i]));
	printbuf_indent_sub(out, 2);

	prt_str(out, "IO errors since ");
	bch2_pr_time_units(out, (ktime_get_real_seconds() - le64_to_cpu(m.errors_reset_time)) * NSEC_PER_SEC);
	prt_str(out, " ago");
	prt_newline(out);

	printbuf_indent_add(out, 2);
	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++)
		prt_printf(out, "%s:\t%llu\n", bch2_member_error_strs[i],
			   atomic64_read(&ca->errors[i]) - le64_to_cpu(m.errors_at_reset[i]));
	printbuf_indent_sub(out, 2);
}

void bch2_dev_errors_reset(struct bch_dev *ca)
{
	struct bch_fs *c = ca->fs;
	struct bch_member *m;

	mutex_lock(&c->sb_lock);
	m = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
	for (unsigned i = 0; i < ARRAY_SIZE(m->errors_at_reset); i++)
		m->errors_at_reset[i] = cpu_to_le64(atomic64_read(&ca->errors[i]));
	m->errors_reset_time = cpu_to_le64(ktime_get_real_seconds());

	bch2_write_super(c);
	mutex_unlock(&c->sb_lock);
}

/*
 * Per member "range has btree nodes" bitmap:
 *
 * This is so that if we ever have to run the btree node scan to repair we don't
 * have to scan full devices:
 */

bool bch2_dev_btree_bitmap_marked(struct bch_fs *c, struct bkey_s_c k)
{
	bool ret = true;
	rcu_read_lock();
	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr) {
		struct bch_dev *ca = bch2_dev_rcu(c, ptr->dev);
		if (!ca)
			continue;

		if (!bch2_dev_btree_bitmap_marked_sectors(ca, ptr->offset, btree_sectors(c))) {
			ret = false;
			break;
		}
	}
	rcu_read_unlock();
	return ret;
}

static void __bch2_dev_btree_bitmap_mark(struct bch_sb_field_members_v2 *mi, unsigned dev,
				u64 start, unsigned sectors)
{
	struct bch_member *m = __bch2_members_v2_get_mut(mi, dev);
	u64 bitmap = le64_to_cpu(m->btree_allocated_bitmap);

	u64 end = start + sectors;

	int resize = ilog2(roundup_pow_of_two(end)) - (m->btree_bitmap_shift + 6);
	if (resize > 0) {
		u64 new_bitmap = 0;

		for (unsigned i = 0; i < 64; i++)
			if (bitmap & BIT_ULL(i))
				new_bitmap |= BIT_ULL(i >> resize);
		bitmap = new_bitmap;
		m->btree_bitmap_shift += resize;
	}

	BUG_ON(m->btree_bitmap_shift > 57);
	BUG_ON(end > 64ULL << m->btree_bitmap_shift);

	for (unsigned bit = start >> m->btree_bitmap_shift;
	     (u64) bit << m->btree_bitmap_shift < end;
	     bit++)
		bitmap |= BIT_ULL(bit);

	m->btree_allocated_bitmap = cpu_to_le64(bitmap);
}

void bch2_dev_btree_bitmap_mark(struct bch_fs *c, struct bkey_s_c k)
{
	lockdep_assert_held(&c->sb_lock);

	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr) {
		if (!bch2_member_exists(c->disk_sb.sb, ptr->dev))
			continue;

		__bch2_dev_btree_bitmap_mark(mi, ptr->dev, ptr->offset, btree_sectors(c));
	}
}
Commit	Line	Data
fb8e5b4c KO	1	// SPDX-License-Identifier: GPL-2.0
	2
	3	#include "bcachefs.h"
27c15ed2	4	#include "btree_cache.h"
fb8e5b4c	5	#include "disk_groups.h"
23f308ae	6	#include "error.h"
40f7914e	7	#include "opts.h"
fb8e5b4c KO	8	#include "replicas.h"
	9	#include "sb-members.h"
	10	#include "super-io.h"
	11
b07eb825 KO	12	void bch2_dev_missing(struct bch_fs *c, unsigned dev)
	13	{
	14	bch2_fs_inconsistent(c, "pointer to nonexistent device %u", dev);
	15	}
	16
9b3059a1 KO	17	void bch2_dev_bucket_missing(struct bch_fs *c, struct bpos bucket)
	18	{
	19	bch2_fs_inconsistent(c, "pointer to nonexistent bucket %llu:%llu", bucket.inode, bucket.offset);
	20	}
	21
94119eeb KO	22	#define x(t, n, ...) [n] = #t,
	23	static const char * const bch2_iops_measurements[] = {
	24	BCH_IOPS_MEASUREMENTS()
	25	NULL
	26	};
	27
	28	char * const bch2_member_error_strs[] = {
	29	BCH_MEMBER_ERROR_TYPES()
	30	NULL
	31	};
	32	#undef x
	33
9af26120	34	/* Code for bch_sb_field_members_v1: */
3f7b9713	35
3f7b9713 HS	36	struct bch_member bch2_members_v2_get_mut(struct bch_sb sb, int i)
3f7b9713 HS	37	{
0f0fc312	38	return __bch2_members_v2_get_mut(bch2_sb_field_get(sb, members_v2), i);
3f7b9713 HS	39	}
	40
	41	static struct bch_member members_v2_get(struct bch_sb_field_members_v2 *mi, int i)
	42	{
0f0fc312	43	struct bch_member ret, *p = __bch2_members_v2_get_mut(mi, i);
3f7b9713 HS	44	memset(&ret, 0, sizeof(ret));
	45	memcpy(&ret, p, min_t(size_t, le16_to_cpu(mi->member_bytes), sizeof(ret)));
	46	return ret;
	47	}
	48
9af26120	49	static struct bch_member members_v1_get_mut(struct bch_sb_field_members_v1 mi, int i)
1241df58	50	{
3f7b9713	51	return (void ) mi->_members + (i BCH_MEMBER_V1_BYTES);
1241df58 HS	52	}
1241df58 HS	53
9af26120	54	static struct bch_member members_v1_get(struct bch_sb_field_members_v1 *mi, int i)
1241df58 HS	55	{
	56	struct bch_member ret, *p = members_v1_get_mut(mi, i);
	57	memset(&ret, 0, sizeof(ret));
5c1ab40e KO	58	memcpy(&ret, p, min_t(size_t, BCH_MEMBER_V1_BYTES, sizeof(ret)));
5c1ab40e KO	59	return ret;
1241df58 HS	60	}
	61
	62	struct bch_member bch2_sb_member_get(struct bch_sb *sb, int i)
	63	{
4637429e	64	struct bch_sb_field_members_v2 *mi2 = bch2_sb_field_get(sb, members_v2);
3f7b9713 HS	65	if (mi2)
3f7b9713 HS	66	return members_v2_get(mi2, i);
4637429e	67	struct bch_sb_field_members_v1 *mi1 = bch2_sb_field_get(sb, members_v1);
1241df58 HS	68	return members_v1_get(mi1, i);
1241df58 HS	69	}
fb8e5b4c	70
3f7b9713 HS	71	static int sb_members_v2_resize_entries(struct bch_fs *c)
3f7b9713 HS	72	{
4637429e	73	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
3f7b9713 HS	74
	75	if (le16_to_cpu(mi->member_bytes) < sizeof(struct bch_member)) {
	76	unsigned u64s = DIV_ROUND_UP((sizeof(mi) + sizeof(mi->_members[0])
	77	c->disk_sb.sb->nr_devices), 8);
	78
4637429e	79	mi = bch2_sb_field_resize(&c->disk_sb, members_v2, u64s);
3f7b9713 HS	80	if (!mi)
	81	return -BCH_ERR_ENOSPC_sb_members_v2;
	82
	83	for (int i = c->disk_sb.sb->nr_devices - 1; i >= 0; --i) {
	84	void dst = (void ) mi->_members + (i * sizeof(struct bch_member));
0f0fc312	85	memmove(dst, __bch2_members_v2_get_mut(mi, i), le16_to_cpu(mi->member_bytes));
3f7b9713 HS	86	memset(dst + le16_to_cpu(mi->member_bytes),
	87	0, (sizeof(struct bch_member) - le16_to_cpu(mi->member_bytes)));
	88	}
	89	mi->member_bytes = cpu_to_le16(sizeof(struct bch_member));
	90	}
	91	return 0;
	92	}
	93
f5d26fa3	94	int bch2_sb_members_v2_init(struct bch_fs *c)
3f7b9713	95	{
9af26120	96	struct bch_sb_field_members_v1 *mi1;
3f7b9713 HS	97	struct bch_sb_field_members_v2 *mi2;
3f7b9713 HS	98
4637429e KO	99	if (!bch2_sb_field_get(c->disk_sb.sb, members_v2)) {
4637429e KO	100	mi2 = bch2_sb_field_resize(&c->disk_sb, members_v2,
3f7b9713 HS	101	DIV_ROUND_UP(sizeof(*mi2) +
	102	sizeof(struct bch_member) * c->sb.nr_devices,
	103	sizeof(u64)));
4637429e	104	mi1 = bch2_sb_field_get(c->disk_sb.sb, members_v1);
3f7b9713 HS	105	memcpy(&mi2->_members[0], &mi1->_members[0],
	106	BCH_MEMBER_V1_BYTES * c->sb.nr_devices);
	107	memset(&mi2->pad[0], 0, sizeof(mi2->pad));
	108	mi2->member_bytes = cpu_to_le16(BCH_MEMBER_V1_BYTES);
	109	}
	110
	111	return sb_members_v2_resize_entries(c);
	112	}
	113
94119eeb	114	int bch2_sb_members_cpy_v2_v1(struct bch_sb_handle *disk_sb)
3f7b9713	115	{
9af26120	116	struct bch_sb_field_members_v1 *mi1;
3f7b9713 HS	117	struct bch_sb_field_members_v2 *mi2;
3f7b9713 HS	118
4637429e	119	mi1 = bch2_sb_field_resize(disk_sb, members_v1,
3f7b9713 HS	120	DIV_ROUND_UP(sizeof(mi1) + BCH_MEMBER_V1_BYTES
	121	disk_sb->sb->nr_devices, sizeof(u64)));
	122	if (!mi1)
	123	return -BCH_ERR_ENOSPC_sb_members;
	124
4637429e	125	mi2 = bch2_sb_field_get(disk_sb->sb, members_v2);
3f7b9713 HS	126
3f7b9713 HS	127	for (unsigned i = 0; i < disk_sb->sb->nr_devices; i++)
0f0fc312	128	memcpy(members_v1_get_mut(mi1, i), __bch2_members_v2_get_mut(mi2, i), BCH_MEMBER_V1_BYTES);
3f7b9713 HS	129
	130	return 0;
	131	}
	132
	133	static int validate_member(struct printbuf *err,
	134	struct bch_member m,
	135	struct bch_sb *sb,
	136	int i)
	137	{
db42549d KO	138	if (le64_to_cpu(m.nbuckets) > BCH_MEMBER_NBUCKETS_MAX) {
	139	prt_printf(err, "device %u: too many buckets (got %llu, max %u)",
	140	i, le64_to_cpu(m.nbuckets), BCH_MEMBER_NBUCKETS_MAX);
3f7b9713 HS	141	return -BCH_ERR_invalid_sb_members;
	142	}
	143
	144	if (le64_to_cpu(m.nbuckets) -
	145	le16_to_cpu(m.first_bucket) < BCH_MIN_NR_NBUCKETS) {
	146	prt_printf(err, "device %u: not enough buckets (got %llu, max %u)",
	147	i, le64_to_cpu(m.nbuckets), BCH_MIN_NR_NBUCKETS);
	148	return -BCH_ERR_invalid_sb_members;
	149	}
	150
	151	if (le16_to_cpu(m.bucket_size) <
	152	le16_to_cpu(sb->block_size)) {
	153	prt_printf(err, "device %u: bucket size %u smaller than block size %u",
	154	i, le16_to_cpu(m.bucket_size), le16_to_cpu(sb->block_size));
	155	return -BCH_ERR_invalid_sb_members;
	156	}
	157
	158	if (le16_to_cpu(m.bucket_size) <
	159	BCH_SB_BTREE_NODE_SIZE(sb)) {
	160	prt_printf(err, "device %u: bucket size %u smaller than btree node size %llu",
	161	i, le16_to_cpu(m.bucket_size), BCH_SB_BTREE_NODE_SIZE(sb));
	162	return -BCH_ERR_invalid_sb_members;
	163	}
	164
	165	return 0;
	166	}
	167
	168	static void member_to_text(struct printbuf *out,
	169	struct bch_member m,
	170	struct bch_sb_field_disk_groups *gi,
	171	struct bch_sb *sb,
	172	int i)
	173	{
	174	unsigned data_have = bch2_sb_dev_has_data(sb, i);
	175	u64 bucket_size = le16_to_cpu(m.bucket_size);
	176	u64 device_size = le64_to_cpu(m.nbuckets) * bucket_size;
	177
2f724563	178	if (!bch2_member_alive(&m))
dc7a15fb KO	179	return;
dc7a15fb KO	180
7423330e	181	prt_printf(out, "Device:\t%u\n", i);
3f7b9713 HS	182
	183	printbuf_indent_add(out, 2);
	184
7423330e	185	prt_printf(out, "Label:\t");
94119eeb KO	186	if (BCH_MEMBER_GROUP(&m)) {
	187	unsigned idx = BCH_MEMBER_GROUP(&m) - 1;
	188
	189	if (idx < disk_groups_nr(gi))
	190	prt_printf(out, "%s (%u)",
	191	gi->entries[idx].label, idx);
	192	else
	193	prt_printf(out, "(bad disk labels section)");
	194	} else {
	195	prt_printf(out, "(none)");
	196	}
	197	prt_newline(out);
	198
7423330e	199	prt_printf(out, "UUID:\t");
3f7b9713 HS	200	pr_uuid(out, m.uuid.b);
	201	prt_newline(out);
	202
7423330e	203	prt_printf(out, "Size:\t");
3f7b9713 HS	204	prt_units_u64(out, device_size << 9);
	205	prt_newline(out);
	206
7423330e KO	207	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++)
7423330e KO	208	prt_printf(out, "%s errors:\t%llu\n", bch2_member_error_strs[i], le64_to_cpu(m.errors[i]));
94119eeb	209
7423330e KO	210	for (unsigned i = 0; i < BCH_IOPS_NR; i++)
7423330e KO	211	prt_printf(out, "%s iops:\t%u\n", bch2_iops_measurements[i], le32_to_cpu(m.iops[i]));
40f7914e	212
7423330e	213	prt_printf(out, "Bucket size:\t");
3f7b9713 HS	214	prt_units_u64(out, bucket_size << 9);
	215	prt_newline(out);
	216
7423330e KO	217	prt_printf(out, "First bucket:\t%u\n", le16_to_cpu(m.first_bucket));
7423330e KO	218	prt_printf(out, "Buckets:\t%llu\n", le64_to_cpu(m.nbuckets));
3f7b9713	219
7423330e	220	prt_printf(out, "Last mount:\t");
3f7b9713	221	if (m.last_mount)
59154f2c	222	bch2_prt_datetime(out, le64_to_cpu(m.last_mount));
3f7b9713 HS	223	else
	224	prt_printf(out, "(never)");
	225	prt_newline(out);
6b00de06	226
7423330e	227	prt_printf(out, "Last superblock write:\t%llu\n", le64_to_cpu(m.seq));
3f7b9713	228
7423330e	229	prt_printf(out, "State:\t%s\n",
3f7b9713 HS	230	BCH_MEMBER_STATE(&m) < BCH_MEMBER_STATE_NR
	231	? bch2_member_states[BCH_MEMBER_STATE(&m)]
	232	: "unknown");
3f7b9713	233
7423330e	234	prt_printf(out, "Data allowed:\t");
3f7b9713	235	if (BCH_MEMBER_DATA_ALLOWED(&m))
e58f963c	236	prt_bitflags(out, __bch2_data_types, BCH_MEMBER_DATA_ALLOWED(&m));
3f7b9713 HS	237	else
	238	prt_printf(out, "(none)");
	239	prt_newline(out);
	240
7423330e	241	prt_printf(out, "Has data:\t");
3f7b9713	242	if (data_have)
e58f963c	243	prt_bitflags(out, __bch2_data_types, data_have);
3f7b9713 HS	244	else
	245	prt_printf(out, "(none)");
	246	prt_newline(out);
	247
018b32a6 KO	248	prt_printf(out, "Btree allocated bitmap blocksize:\t");
	249	prt_units_u64(out, 1ULL << m.btree_bitmap_shift);
	250	prt_newline(out);
	251
	252	prt_printf(out, "Btree allocated bitmap:\t");
	253	bch2_prt_u64_base2_nbits(out, le64_to_cpu(m.btree_allocated_bitmap), 64);
	254	prt_newline(out);
	255
7423330e	256	prt_printf(out, "Durability:\t%llu\n", BCH_MEMBER_DURABILITY(&m) ? BCH_MEMBER_DURABILITY(&m) - 1 : 1);
d05db127	257
7423330e KO	258	prt_printf(out, "Discard:\t%llu\n", BCH_MEMBER_DISCARD(&m));
7423330e KO	259	prt_printf(out, "Freespace initialized:\t%llu\n", BCH_MEMBER_FREESPACE_INITIALIZED(&m));
3f7b9713 HS	260
	261	printbuf_indent_sub(out, 2);
	262	}
	263
a5c3e265 KO	264	static int bch2_sb_members_v1_validate(struct bch_sb sb, struct bch_sb_field f,
a5c3e265 KO	265	enum bch_validate_flags flags, struct printbuf *err)
fb8e5b4c	266	{
9af26120	267	struct bch_sb_field_members_v1 *mi = field_to_type(f, members_v1);
fb8e5b4c KO	268	unsigned i;
fb8e5b4c KO	269
5c1ab40e	270	if ((void *) members_v1_get_mut(mi, sb->nr_devices) > vstruct_end(&mi->field)) {
fb8e5b4c KO	271	prt_printf(err, "too many devices for section size");
	272	return -BCH_ERR_invalid_sb_members;
	273	}
	274
	275	for (i = 0; i < sb->nr_devices; i++) {
3f7b9713	276	struct bch_member m = members_v1_get(mi, i);
fb8e5b4c	277
3f7b9713 HS	278	int ret = validate_member(err, m, sb, i);
	279	if (ret)
	280	return ret;
fb8e5b4c KO	281	}
	282
	283	return 0;
	284	}
	285
3f7b9713 HS	286	static void bch2_sb_members_v1_to_text(struct printbuf out, struct bch_sb sb,
3f7b9713 HS	287	struct bch_sb_field *f)
fb8e5b4c	288	{
9af26120	289	struct bch_sb_field_members_v1 *mi = field_to_type(f, members_v1);
4637429e	290	struct bch_sb_field_disk_groups *gi = bch2_sb_field_get(sb, disk_groups);
fb8e5b4c KO	291	unsigned i;
fb8e5b4c KO	292
dc7a15fb KO	293	for (i = 0; i < sb->nr_devices; i++)
dc7a15fb KO	294	member_to_text(out, members_v1_get(mi, i), gi, sb, i);
3f7b9713	295	}
fb8e5b4c	296
9af26120	297	const struct bch_sb_field_ops bch_sb_field_ops_members_v1 = {
3f7b9713 HS	298	.validate = bch2_sb_members_v1_validate,
	299	.to_text = bch2_sb_members_v1_to_text,
	300	};
fb8e5b4c	301
3f7b9713 HS	302	static void bch2_sb_members_v2_to_text(struct printbuf out, struct bch_sb sb,
	303	struct bch_sb_field *f)
	304	{
	305	struct bch_sb_field_members_v2 *mi = field_to_type(f, members_v2);
4637429e	306	struct bch_sb_field_disk_groups *gi = bch2_sb_field_get(sb, disk_groups);
3f7b9713	307	unsigned i;
fb8e5b4c	308
dc7a15fb KO	309	for (i = 0; i < sb->nr_devices; i++)
dc7a15fb KO	310	member_to_text(out, members_v2_get(mi, i), gi, sb, i);
3f7b9713 HS	311	}
3f7b9713 HS	312
a5c3e265 KO	313	static int bch2_sb_members_v2_validate(struct bch_sb sb, struct bch_sb_field f,
a5c3e265 KO	314	enum bch_validate_flags flags, struct printbuf *err)
3f7b9713 HS	315	{
3f7b9713 HS	316	struct bch_sb_field_members_v2 *mi = field_to_type(f, members_v2);
0f0fc312	317	size_t mi_bytes = (void *) __bch2_members_v2_get_mut(mi, sb->nr_devices) -
3f7b9713	318	(void *) mi;
fb8e5b4c	319
3f7b9713 HS	320	if (mi_bytes > vstruct_bytes(&mi->field)) {
	321	prt_printf(err, "section too small (%zu > %zu)",
	322	mi_bytes, vstruct_bytes(&mi->field));
	323	return -BCH_ERR_invalid_sb_members;
	324	}
fb8e5b4c	325
3f7b9713 HS	326	for (unsigned i = 0; i < sb->nr_devices; i++) {
	327	int ret = validate_member(err, members_v2_get(mi, i), sb, i);
	328	if (ret)
	329	return ret;
fb8e5b4c	330	}
3f7b9713 HS	331
3f7b9713 HS	332	return 0;
fb8e5b4c KO	333	}
fb8e5b4c KO	334
3f7b9713 HS	335	const struct bch_sb_field_ops bch_sb_field_ops_members_v2 = {
	336	.validate = bch2_sb_members_v2_validate,
	337	.to_text = bch2_sb_members_v2_to_text,
fb8e5b4c	338	};
94119eeb KO	339
	340	void bch2_sb_members_from_cpu(struct bch_fs *c)
	341	{
	342	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
94119eeb KO	343
94119eeb KO	344	rcu_read_lock();
41b84fb4 KO	345	for_each_member_device_rcu(c, ca, NULL) {
41b84fb4 KO	346	struct bch_member *m = __bch2_members_v2_get_mut(mi, ca->dev_idx);
94119eeb	347
41b84fb4	348	for (unsigned e = 0; e < BCH_MEMBER_ERROR_NR; e++)
94119eeb KO	349	m->errors[e] = cpu_to_le64(atomic64_read(&ca->errors[e]));
	350	}
	351	rcu_read_unlock();
	352	}
	353
	354	void bch2_dev_io_errors_to_text(struct printbuf out, struct bch_dev ca)
	355	{
	356	struct bch_fs *c = ca->fs;
	357	struct bch_member m;
	358
	359	mutex_lock(&ca->fs->sb_lock);
	360	m = bch2_sb_member_get(c->disk_sb.sb, ca->dev_idx);
	361	mutex_unlock(&ca->fs->sb_lock);
	362
	363	printbuf_tabstop_push(out, 12);
	364
	365	prt_str(out, "IO errors since filesystem creation");
	366	prt_newline(out);
	367
	368	printbuf_indent_add(out, 2);
7423330e KO	369	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++)
7423330e KO	370	prt_printf(out, "%s:\t%llu\n", bch2_member_error_strs[i], atomic64_read(&ca->errors[i]));
94119eeb KO	371	printbuf_indent_sub(out, 2);
	372
	373	prt_str(out, "IO errors since ");
	374	bch2_pr_time_units(out, (ktime_get_real_seconds() - le64_to_cpu(m.errors_reset_time)) * NSEC_PER_SEC);
	375	prt_str(out, " ago");
	376	prt_newline(out);
	377
	378	printbuf_indent_add(out, 2);
7423330e KO	379	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++)
	380	prt_printf(out, "%s:\t%llu\n", bch2_member_error_strs[i],
	381	atomic64_read(&ca->errors[i]) - le64_to_cpu(m.errors_at_reset[i]));
94119eeb KO	382	printbuf_indent_sub(out, 2);
	383	}
	384
	385	void bch2_dev_errors_reset(struct bch_dev *ca)
	386	{
	387	struct bch_fs *c = ca->fs;
	388	struct bch_member *m;
	389
	390	mutex_lock(&c->sb_lock);
	391	m = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
	392	for (unsigned i = 0; i < ARRAY_SIZE(m->errors_at_reset); i++)
	393	m->errors_at_reset[i] = cpu_to_le64(atomic64_read(&ca->errors[i]));
04eb5793	394	m->errors_reset_time = cpu_to_le64(ktime_get_real_seconds());
94119eeb KO	395
	396	bch2_write_super(c);
	397	mutex_unlock(&c->sb_lock);
	398	}
27c15ed2 KO	399
	400	/*
	401	* Per member "range has btree nodes" bitmap:
	402	*
	403	* This is so that if we ever have to run the btree node scan to repair we don't
	404	* have to scan full devices:
	405	*/
	406
	407	bool bch2_dev_btree_bitmap_marked(struct bch_fs *c, struct bkey_s_c k)
	408	{
13a16dab KO	409	bool ret = true;
	410	rcu_read_lock();
	411	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr) {
	412	struct bch_dev *ca = bch2_dev_rcu(c, ptr->dev);
	413	if (!ca)
	414	continue;
	415
	416	if (!bch2_dev_btree_bitmap_marked_sectors(ca, ptr->offset, btree_sectors(c))) {
	417	ret = false;
	418	break;
	419	}
	420	}
	421	rcu_read_unlock();
	422	return ret;
27c15ed2 KO	423	}
	424
	425	static void __bch2_dev_btree_bitmap_mark(struct bch_sb_field_members_v2 *mi, unsigned dev,
	426	u64 start, unsigned sectors)
	427	{
	428	struct bch_member *m = __bch2_members_v2_get_mut(mi, dev);
	429	u64 bitmap = le64_to_cpu(m->btree_allocated_bitmap);
	430
	431	u64 end = start + sectors;
	432
	433	int resize = ilog2(roundup_pow_of_two(end)) - (m->btree_bitmap_shift + 6);
	434	if (resize > 0) {
	435	u64 new_bitmap = 0;
	436
	437	for (unsigned i = 0; i < 64; i++)
	438	if (bitmap & BIT_ULL(i))
	439	new_bitmap \|= BIT_ULL(i >> resize);
	440	bitmap = new_bitmap;
	441	m->btree_bitmap_shift += resize;
	442	}
	443
60f2b1bc KO	444	BUG_ON(m->btree_bitmap_shift > 57);
	445	BUG_ON(end > 64ULL << m->btree_bitmap_shift);
	446
fa845c73 KO	447	for (unsigned bit = start >> m->btree_bitmap_shift;
fa845c73 KO	448	(u64) bit << m->btree_bitmap_shift < end;
27c15ed2 KO	449	bit++)
	450	bitmap \|= BIT_ULL(bit);
	451
	452	m->btree_allocated_bitmap = cpu_to_le64(bitmap);
	453	}
	454
	455	void bch2_dev_btree_bitmap_mark(struct bch_fs *c, struct bkey_s_c k)
	456	{
	457	lockdep_assert_held(&c->sb_lock);
	458
	459	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
13a16dab KO	460	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr) {
	461	if (!bch2_member_exists(c->disk_sb.sb, ptr->dev))
	462	continue;
	463
27c15ed2	464	__bch2_dev_btree_bitmap_mark(mi, ptr->dev, ptr->offset, btree_sectors(c));
13a16dab	465	}
27c15ed2	466	}