[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 "opts.h"
#include "replicas.h"
#include "sb-members.h"
#include "super-io.h"

#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) > LONG_MAX) {
		prt_printf(err, "device %u: too many buckets (got %llu, max %lu)",
			   i, le64_to_cpu(m.nbuckets), LONG_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_exists(&m))
		return;

	prt_printf(out, "Device:");
	prt_tab(out);
	prt_printf(out, "%u", i);
	prt_newline(out);

	printbuf_indent_add(out, 2);

	prt_printf(out, "Label:");
	prt_tab(out);
	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:");
	prt_tab(out);
	pr_uuid(out, m.uuid.b);
	prt_newline(out);

	prt_printf(out, "Size:");
	prt_tab(out);
	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:", bch2_member_error_strs[i]);
		prt_tab(out);
		prt_u64(out, le64_to_cpu(m.errors[i]));
		prt_newline(out);
	}

	for (unsigned i = 0; i < BCH_IOPS_NR; i++) {
		prt_printf(out, "%s iops:", bch2_iops_measurements[i]);
		prt_tab(out);
		prt_printf(out, "%u", le32_to_cpu(m.iops[i]));
		prt_newline(out);
	}

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

	prt_printf(out, "First bucket:");
	prt_tab(out);
	prt_printf(out, "%u", le16_to_cpu(m.first_bucket));
	prt_newline(out);

	prt_printf(out, "Buckets:");
	prt_tab(out);
	prt_printf(out, "%llu", le64_to_cpu(m.nbuckets));
	prt_newline(out);

	prt_printf(out, "Last mount:");
	prt_tab(out);
	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:");
	prt_tab(out);
	prt_u64(out, le64_to_cpu(m.seq));
	prt_newline(out);

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

	prt_printf(out, "Data allowed:");
	prt_tab(out);
	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:");
	prt_tab(out);
	if (data_have)
		prt_bitflags(out, __bch2_data_types, data_have);
	else
		prt_printf(out, "(none)");
	prt_newline(out);

	prt_str(out, "Durability:");
	prt_tab(out);
	prt_printf(out, "%llu", BCH_MEMBER_DURABILITY(&m) ? BCH_MEMBER_DURABILITY(&m) - 1 : 1);
	prt_newline(out);

	prt_printf(out, "Discard:");
	prt_tab(out);
	prt_printf(out, "%llu", BCH_MEMBER_DISCARD(&m));
	prt_newline(out);

	prt_printf(out, "Freespace initialized:");
	prt_tab(out);
	prt_printf(out, "%llu", BCH_MEMBER_FREESPACE_INITIALIZED(&m));
	prt_newline(out);

	printbuf_indent_sub(out, 2);
}

static int bch2_sb_members_v1_validate(struct bch_sb *sb,
				    struct bch_sb_field *f,
				    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,
				       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:", bch2_member_error_strs[i]);
		prt_tab(out);
		prt_u64(out, atomic64_read(&ca->errors[i]));
		prt_newline(out);
	}
	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:", bch2_member_error_strs[i]);
		prt_tab(out);
		prt_u64(out, atomic64_read(&ca->errors[i]) - le64_to_cpu(m.errors_at_reset[i]));
		prt_newline(out);
	}
	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)
{
	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr)
		if (!bch2_dev_btree_bitmap_marked_sectors(bch_dev_bkey_exists(c, ptr->dev),
							  ptr->offset, btree_sectors(c)))
			return false;
	return true;
}

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;
	}

	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)
		__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"
40f7914e	6	#include "opts.h"
fb8e5b4c KO	7	#include "replicas.h"
	8	#include "sb-members.h"
	9	#include "super-io.h"
	10
94119eeb KO	11	#define x(t, n, ...) [n] = #t,
	12	static const char * const bch2_iops_measurements[] = {
	13	BCH_IOPS_MEASUREMENTS()
	14	NULL
	15	};
	16
	17	char * const bch2_member_error_strs[] = {
	18	BCH_MEMBER_ERROR_TYPES()
	19	NULL
	20	};
	21	#undef x
	22
9af26120	23	/* Code for bch_sb_field_members_v1: */
3f7b9713	24
3f7b9713 HS	25	struct bch_member bch2_members_v2_get_mut(struct bch_sb sb, int i)
3f7b9713 HS	26	{
0f0fc312	27	return __bch2_members_v2_get_mut(bch2_sb_field_get(sb, members_v2), i);
3f7b9713 HS	28	}
	29
	30	static struct bch_member members_v2_get(struct bch_sb_field_members_v2 *mi, int i)
	31	{
0f0fc312	32	struct bch_member ret, *p = __bch2_members_v2_get_mut(mi, i);
3f7b9713 HS	33	memset(&ret, 0, sizeof(ret));
	34	memcpy(&ret, p, min_t(size_t, le16_to_cpu(mi->member_bytes), sizeof(ret)));
	35	return ret;
	36	}
	37
9af26120	38	static struct bch_member members_v1_get_mut(struct bch_sb_field_members_v1 mi, int i)
1241df58	39	{
3f7b9713	40	return (void ) mi->_members + (i BCH_MEMBER_V1_BYTES);
1241df58 HS	41	}
1241df58 HS	42
9af26120	43	static struct bch_member members_v1_get(struct bch_sb_field_members_v1 *mi, int i)
1241df58 HS	44	{
	45	struct bch_member ret, *p = members_v1_get_mut(mi, i);
	46	memset(&ret, 0, sizeof(ret));
5c1ab40e KO	47	memcpy(&ret, p, min_t(size_t, BCH_MEMBER_V1_BYTES, sizeof(ret)));
5c1ab40e KO	48	return ret;
1241df58 HS	49	}
	50
	51	struct bch_member bch2_sb_member_get(struct bch_sb *sb, int i)
	52	{
4637429e	53	struct bch_sb_field_members_v2 *mi2 = bch2_sb_field_get(sb, members_v2);
3f7b9713 HS	54	if (mi2)
3f7b9713 HS	55	return members_v2_get(mi2, i);
4637429e	56	struct bch_sb_field_members_v1 *mi1 = bch2_sb_field_get(sb, members_v1);
1241df58 HS	57	return members_v1_get(mi1, i);
1241df58 HS	58	}
fb8e5b4c	59
3f7b9713 HS	60	static int sb_members_v2_resize_entries(struct bch_fs *c)
3f7b9713 HS	61	{
4637429e	62	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
3f7b9713 HS	63
	64	if (le16_to_cpu(mi->member_bytes) < sizeof(struct bch_member)) {
	65	unsigned u64s = DIV_ROUND_UP((sizeof(mi) + sizeof(mi->_members[0])
	66	c->disk_sb.sb->nr_devices), 8);
	67
4637429e	68	mi = bch2_sb_field_resize(&c->disk_sb, members_v2, u64s);
3f7b9713 HS	69	if (!mi)
	70	return -BCH_ERR_ENOSPC_sb_members_v2;
	71
	72	for (int i = c->disk_sb.sb->nr_devices - 1; i >= 0; --i) {
	73	void dst = (void ) mi->_members + (i * sizeof(struct bch_member));
0f0fc312	74	memmove(dst, __bch2_members_v2_get_mut(mi, i), le16_to_cpu(mi->member_bytes));
3f7b9713 HS	75	memset(dst + le16_to_cpu(mi->member_bytes),
	76	0, (sizeof(struct bch_member) - le16_to_cpu(mi->member_bytes)));
	77	}
	78	mi->member_bytes = cpu_to_le16(sizeof(struct bch_member));
	79	}
	80	return 0;
	81	}
	82
f5d26fa3	83	int bch2_sb_members_v2_init(struct bch_fs *c)
3f7b9713	84	{
9af26120	85	struct bch_sb_field_members_v1 *mi1;
3f7b9713 HS	86	struct bch_sb_field_members_v2 *mi2;
3f7b9713 HS	87
4637429e KO	88	if (!bch2_sb_field_get(c->disk_sb.sb, members_v2)) {
4637429e KO	89	mi2 = bch2_sb_field_resize(&c->disk_sb, members_v2,
3f7b9713 HS	90	DIV_ROUND_UP(sizeof(*mi2) +
	91	sizeof(struct bch_member) * c->sb.nr_devices,
	92	sizeof(u64)));
4637429e	93	mi1 = bch2_sb_field_get(c->disk_sb.sb, members_v1);
3f7b9713 HS	94	memcpy(&mi2->_members[0], &mi1->_members[0],
	95	BCH_MEMBER_V1_BYTES * c->sb.nr_devices);
	96	memset(&mi2->pad[0], 0, sizeof(mi2->pad));
	97	mi2->member_bytes = cpu_to_le16(BCH_MEMBER_V1_BYTES);
	98	}
	99
	100	return sb_members_v2_resize_entries(c);
	101	}
	102
94119eeb	103	int bch2_sb_members_cpy_v2_v1(struct bch_sb_handle *disk_sb)
3f7b9713	104	{
9af26120	105	struct bch_sb_field_members_v1 *mi1;
3f7b9713 HS	106	struct bch_sb_field_members_v2 *mi2;
3f7b9713 HS	107
4637429e	108	mi1 = bch2_sb_field_resize(disk_sb, members_v1,
3f7b9713 HS	109	DIV_ROUND_UP(sizeof(mi1) + BCH_MEMBER_V1_BYTES
	110	disk_sb->sb->nr_devices, sizeof(u64)));
	111	if (!mi1)
	112	return -BCH_ERR_ENOSPC_sb_members;
	113
4637429e	114	mi2 = bch2_sb_field_get(disk_sb->sb, members_v2);
3f7b9713 HS	115
3f7b9713 HS	116	for (unsigned i = 0; i < disk_sb->sb->nr_devices; i++)
0f0fc312	117	memcpy(members_v1_get_mut(mi1, i), __bch2_members_v2_get_mut(mi2, i), BCH_MEMBER_V1_BYTES);
3f7b9713 HS	118
	119	return 0;
	120	}
	121
	122	static int validate_member(struct printbuf *err,
	123	struct bch_member m,
	124	struct bch_sb *sb,
	125	int i)
	126	{
	127	if (le64_to_cpu(m.nbuckets) > LONG_MAX) {
	128	prt_printf(err, "device %u: too many buckets (got %llu, max %lu)",
	129	i, le64_to_cpu(m.nbuckets), LONG_MAX);
	130	return -BCH_ERR_invalid_sb_members;
	131	}
	132
	133	if (le64_to_cpu(m.nbuckets) -
	134	le16_to_cpu(m.first_bucket) < BCH_MIN_NR_NBUCKETS) {
	135	prt_printf(err, "device %u: not enough buckets (got %llu, max %u)",
	136	i, le64_to_cpu(m.nbuckets), BCH_MIN_NR_NBUCKETS);
	137	return -BCH_ERR_invalid_sb_members;
	138	}
	139
	140	if (le16_to_cpu(m.bucket_size) <
	141	le16_to_cpu(sb->block_size)) {
	142	prt_printf(err, "device %u: bucket size %u smaller than block size %u",
	143	i, le16_to_cpu(m.bucket_size), le16_to_cpu(sb->block_size));
	144	return -BCH_ERR_invalid_sb_members;
	145	}
	146
	147	if (le16_to_cpu(m.bucket_size) <
	148	BCH_SB_BTREE_NODE_SIZE(sb)) {
	149	prt_printf(err, "device %u: bucket size %u smaller than btree node size %llu",
	150	i, le16_to_cpu(m.bucket_size), BCH_SB_BTREE_NODE_SIZE(sb));
	151	return -BCH_ERR_invalid_sb_members;
	152	}
	153
	154	return 0;
	155	}
	156
	157	static void member_to_text(struct printbuf *out,
	158	struct bch_member m,
	159	struct bch_sb_field_disk_groups *gi,
	160	struct bch_sb *sb,
	161	int i)
	162	{
	163	unsigned data_have = bch2_sb_dev_has_data(sb, i);
	164	u64 bucket_size = le16_to_cpu(m.bucket_size);
	165	u64 device_size = le64_to_cpu(m.nbuckets) * bucket_size;
	166
dc7a15fb KO	167	if (!bch2_member_exists(&m))
	168	return;
	169
3f7b9713 HS	170	prt_printf(out, "Device:");
	171	prt_tab(out);
	172	prt_printf(out, "%u", i);
	173	prt_newline(out);
	174
	175	printbuf_indent_add(out, 2);
	176
94119eeb KO	177	prt_printf(out, "Label:");
	178	prt_tab(out);
	179	if (BCH_MEMBER_GROUP(&m)) {
	180	unsigned idx = BCH_MEMBER_GROUP(&m) - 1;
	181
	182	if (idx < disk_groups_nr(gi))
	183	prt_printf(out, "%s (%u)",
	184	gi->entries[idx].label, idx);
	185	else
	186	prt_printf(out, "(bad disk labels section)");
	187	} else {
	188	prt_printf(out, "(none)");
	189	}
	190	prt_newline(out);
	191
3f7b9713 HS	192	prt_printf(out, "UUID:");
	193	prt_tab(out);
	194	pr_uuid(out, m.uuid.b);
	195	prt_newline(out);
	196
	197	prt_printf(out, "Size:");
	198	prt_tab(out);
	199	prt_units_u64(out, device_size << 9);
	200	prt_newline(out);
	201
94119eeb KO	202	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++) {
	203	prt_printf(out, "%s errors:", bch2_member_error_strs[i]);
	204	prt_tab(out);
	205	prt_u64(out, le64_to_cpu(m.errors[i]));
	206	prt_newline(out);
	207	}
	208
40f7914e HS	209	for (unsigned i = 0; i < BCH_IOPS_NR; i++) {
	210	prt_printf(out, "%s iops:", bch2_iops_measurements[i]);
	211	prt_tab(out);
	212	prt_printf(out, "%u", le32_to_cpu(m.iops[i]));
	213	prt_newline(out);
	214	}
	215
3f7b9713 HS	216	prt_printf(out, "Bucket size:");
	217	prt_tab(out);
	218	prt_units_u64(out, bucket_size << 9);
	219	prt_newline(out);
	220
	221	prt_printf(out, "First bucket:");
	222	prt_tab(out);
	223	prt_printf(out, "%u", le16_to_cpu(m.first_bucket));
	224	prt_newline(out);
	225
	226	prt_printf(out, "Buckets:");
	227	prt_tab(out);
	228	prt_printf(out, "%llu", le64_to_cpu(m.nbuckets));
	229	prt_newline(out);
	230
	231	prt_printf(out, "Last mount:");
	232	prt_tab(out);
	233	if (m.last_mount)
59154f2c	234	bch2_prt_datetime(out, le64_to_cpu(m.last_mount));
3f7b9713 HS	235	else
	236	prt_printf(out, "(never)");
	237	prt_newline(out);
6b00de06 KO	238
	239	prt_printf(out, "Last superblock write:");
	240	prt_tab(out);
	241	prt_u64(out, le64_to_cpu(m.seq));
	242	prt_newline(out);
3f7b9713 HS	243
	244	prt_printf(out, "State:");
	245	prt_tab(out);
	246	prt_printf(out, "%s",
	247	BCH_MEMBER_STATE(&m) < BCH_MEMBER_STATE_NR
	248	? bch2_member_states[BCH_MEMBER_STATE(&m)]
	249	: "unknown");
	250	prt_newline(out);
	251
3f7b9713 HS	252	prt_printf(out, "Data allowed:");
	253	prt_tab(out);
	254	if (BCH_MEMBER_DATA_ALLOWED(&m))
e58f963c	255	prt_bitflags(out, __bch2_data_types, BCH_MEMBER_DATA_ALLOWED(&m));
3f7b9713 HS	256	else
	257	prt_printf(out, "(none)");
	258	prt_newline(out);
	259
	260	prt_printf(out, "Has data:");
	261	prt_tab(out);
	262	if (data_have)
e58f963c	263	prt_bitflags(out, __bch2_data_types, data_have);
3f7b9713 HS	264	else
	265	prt_printf(out, "(none)");
	266	prt_newline(out);
	267
d05db127 KO	268	prt_str(out, "Durability:");
d05db127 KO	269	prt_tab(out);
f60250de	270	prt_printf(out, "%llu", BCH_MEMBER_DURABILITY(&m) ? BCH_MEMBER_DURABILITY(&m) - 1 : 1);
d05db127 KO	271	prt_newline(out);
d05db127 KO	272
3f7b9713 HS	273	prt_printf(out, "Discard:");
	274	prt_tab(out);
	275	prt_printf(out, "%llu", BCH_MEMBER_DISCARD(&m));
	276	prt_newline(out);
	277
	278	prt_printf(out, "Freespace initialized:");
	279	prt_tab(out);
	280	prt_printf(out, "%llu", BCH_MEMBER_FREESPACE_INITIALIZED(&m));
	281	prt_newline(out);
	282
	283	printbuf_indent_sub(out, 2);
	284	}
	285
	286	static int bch2_sb_members_v1_validate(struct bch_sb *sb,
fb8e5b4c KO	287	struct bch_sb_field *f,
	288	struct printbuf *err)
	289	{
9af26120	290	struct bch_sb_field_members_v1 *mi = field_to_type(f, members_v1);
fb8e5b4c KO	291	unsigned i;
fb8e5b4c KO	292
5c1ab40e	293	if ((void *) members_v1_get_mut(mi, sb->nr_devices) > vstruct_end(&mi->field)) {
fb8e5b4c KO	294	prt_printf(err, "too many devices for section size");
	295	return -BCH_ERR_invalid_sb_members;
	296	}
	297
	298	for (i = 0; i < sb->nr_devices; i++) {
3f7b9713	299	struct bch_member m = members_v1_get(mi, i);
fb8e5b4c	300
3f7b9713 HS	301	int ret = validate_member(err, m, sb, i);
	302	if (ret)
	303	return ret;
fb8e5b4c KO	304	}
	305
	306	return 0;
	307	}
	308
3f7b9713 HS	309	static void bch2_sb_members_v1_to_text(struct printbuf out, struct bch_sb sb,
3f7b9713 HS	310	struct bch_sb_field *f)
fb8e5b4c	311	{
9af26120	312	struct bch_sb_field_members_v1 *mi = field_to_type(f, members_v1);
4637429e	313	struct bch_sb_field_disk_groups *gi = bch2_sb_field_get(sb, disk_groups);
fb8e5b4c KO	314	unsigned i;
fb8e5b4c KO	315
dc7a15fb KO	316	for (i = 0; i < sb->nr_devices; i++)
dc7a15fb KO	317	member_to_text(out, members_v1_get(mi, i), gi, sb, i);
3f7b9713	318	}
fb8e5b4c	319
9af26120	320	const struct bch_sb_field_ops bch_sb_field_ops_members_v1 = {
3f7b9713 HS	321	.validate = bch2_sb_members_v1_validate,
	322	.to_text = bch2_sb_members_v1_to_text,
	323	};
fb8e5b4c	324
3f7b9713 HS	325	static void bch2_sb_members_v2_to_text(struct printbuf out, struct bch_sb sb,
	326	struct bch_sb_field *f)
	327	{
	328	struct bch_sb_field_members_v2 *mi = field_to_type(f, members_v2);
4637429e	329	struct bch_sb_field_disk_groups *gi = bch2_sb_field_get(sb, disk_groups);
3f7b9713	330	unsigned i;
fb8e5b4c	331
dc7a15fb KO	332	for (i = 0; i < sb->nr_devices; i++)
dc7a15fb KO	333	member_to_text(out, members_v2_get(mi, i), gi, sb, i);
3f7b9713 HS	334	}
	335
	336	static int bch2_sb_members_v2_validate(struct bch_sb *sb,
	337	struct bch_sb_field *f,
	338	struct printbuf *err)
	339	{
	340	struct bch_sb_field_members_v2 *mi = field_to_type(f, members_v2);
0f0fc312	341	size_t mi_bytes = (void *) __bch2_members_v2_get_mut(mi, sb->nr_devices) -
3f7b9713	342	(void *) mi;
fb8e5b4c	343
3f7b9713 HS	344	if (mi_bytes > vstruct_bytes(&mi->field)) {
	345	prt_printf(err, "section too small (%zu > %zu)",
	346	mi_bytes, vstruct_bytes(&mi->field));
	347	return -BCH_ERR_invalid_sb_members;
	348	}
fb8e5b4c	349
3f7b9713 HS	350	for (unsigned i = 0; i < sb->nr_devices; i++) {
	351	int ret = validate_member(err, members_v2_get(mi, i), sb, i);
	352	if (ret)
	353	return ret;
fb8e5b4c	354	}
3f7b9713 HS	355
3f7b9713 HS	356	return 0;
fb8e5b4c KO	357	}
fb8e5b4c KO	358
3f7b9713 HS	359	const struct bch_sb_field_ops bch_sb_field_ops_members_v2 = {
	360	.validate = bch2_sb_members_v2_validate,
	361	.to_text = bch2_sb_members_v2_to_text,
fb8e5b4c	362	};
94119eeb KO	363
	364	void bch2_sb_members_from_cpu(struct bch_fs *c)
	365	{
	366	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
94119eeb KO	367
94119eeb KO	368	rcu_read_lock();
41b84fb4 KO	369	for_each_member_device_rcu(c, ca, NULL) {
41b84fb4 KO	370	struct bch_member *m = __bch2_members_v2_get_mut(mi, ca->dev_idx);
94119eeb	371
41b84fb4	372	for (unsigned e = 0; e < BCH_MEMBER_ERROR_NR; e++)
94119eeb KO	373	m->errors[e] = cpu_to_le64(atomic64_read(&ca->errors[e]));
	374	}
	375	rcu_read_unlock();
	376	}
	377
	378	void bch2_dev_io_errors_to_text(struct printbuf out, struct bch_dev ca)
	379	{
	380	struct bch_fs *c = ca->fs;
	381	struct bch_member m;
	382
	383	mutex_lock(&ca->fs->sb_lock);
	384	m = bch2_sb_member_get(c->disk_sb.sb, ca->dev_idx);
	385	mutex_unlock(&ca->fs->sb_lock);
	386
	387	printbuf_tabstop_push(out, 12);
	388
	389	prt_str(out, "IO errors since filesystem creation");
	390	prt_newline(out);
	391
	392	printbuf_indent_add(out, 2);
	393	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++) {
	394	prt_printf(out, "%s:", bch2_member_error_strs[i]);
	395	prt_tab(out);
	396	prt_u64(out, atomic64_read(&ca->errors[i]));
	397	prt_newline(out);
	398	}
	399	printbuf_indent_sub(out, 2);
	400
	401	prt_str(out, "IO errors since ");
	402	bch2_pr_time_units(out, (ktime_get_real_seconds() - le64_to_cpu(m.errors_reset_time)) * NSEC_PER_SEC);
	403	prt_str(out, " ago");
	404	prt_newline(out);
	405
	406	printbuf_indent_add(out, 2);
	407	for (unsigned i = 0; i < BCH_MEMBER_ERROR_NR; i++) {
	408	prt_printf(out, "%s:", bch2_member_error_strs[i]);
	409	prt_tab(out);
	410	prt_u64(out, atomic64_read(&ca->errors[i]) - le64_to_cpu(m.errors_at_reset[i]));
	411	prt_newline(out);
	412	}
	413	printbuf_indent_sub(out, 2);
	414	}
	415
	416	void bch2_dev_errors_reset(struct bch_dev *ca)
	417	{
	418	struct bch_fs *c = ca->fs;
	419	struct bch_member *m;
	420
	421	mutex_lock(&c->sb_lock);
	422	m = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
	423	for (unsigned i = 0; i < ARRAY_SIZE(m->errors_at_reset); i++)
	424	m->errors_at_reset[i] = cpu_to_le64(atomic64_read(&ca->errors[i]));
04eb5793	425	m->errors_reset_time = cpu_to_le64(ktime_get_real_seconds());
94119eeb KO	426
	427	bch2_write_super(c);
	428	mutex_unlock(&c->sb_lock);
	429	}
27c15ed2 KO	430
	431	/*
	432	* Per member "range has btree nodes" bitmap:
	433	*
	434	* This is so that if we ever have to run the btree node scan to repair we don't
	435	* have to scan full devices:
	436	*/
	437
	438	bool bch2_dev_btree_bitmap_marked(struct bch_fs *c, struct bkey_s_c k)
	439	{
	440	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr)
	441	if (!bch2_dev_btree_bitmap_marked_sectors(bch_dev_bkey_exists(c, ptr->dev),
	442	ptr->offset, btree_sectors(c)))
	443	return false;
	444	return true;
	445	}
	446
	447	static void __bch2_dev_btree_bitmap_mark(struct bch_sb_field_members_v2 *mi, unsigned dev,
	448	u64 start, unsigned sectors)
	449	{
	450	struct bch_member *m = __bch2_members_v2_get_mut(mi, dev);
	451	u64 bitmap = le64_to_cpu(m->btree_allocated_bitmap);
	452
	453	u64 end = start + sectors;
	454
	455	int resize = ilog2(roundup_pow_of_two(end)) - (m->btree_bitmap_shift + 6);
	456	if (resize > 0) {
	457	u64 new_bitmap = 0;
	458
	459	for (unsigned i = 0; i < 64; i++)
	460	if (bitmap & BIT_ULL(i))
	461	new_bitmap \|= BIT_ULL(i >> resize);
	462	bitmap = new_bitmap;
	463	m->btree_bitmap_shift += resize;
	464	}
	465
fa845c73 KO	466	for (unsigned bit = start >> m->btree_bitmap_shift;
fa845c73 KO	467	(u64) bit << m->btree_bitmap_shift < end;
27c15ed2 KO	468	bit++)
	469	bitmap \|= BIT_ULL(bit);
	470
	471	m->btree_allocated_bitmap = cpu_to_le64(bitmap);
	472	}
	473
	474	void bch2_dev_btree_bitmap_mark(struct bch_fs *c, struct bkey_s_c k)
	475	{
	476	lockdep_assert_held(&c->sb_lock);
	477
	478	struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
	479	bkey_for_each_ptr(bch2_bkey_ptrs_c(k), ptr)
	480	__bch2_dev_btree_bitmap_mark(mi, ptr->dev, ptr->offset, btree_sectors(c));
	481	}