#include "bcachefs.h"
#include "checksum.h"
-#include "counters.h"
#include "disk_groups.h"
#include "ec.h"
#include "error.h"
#include "replicas.h"
#include "quota.h"
#include "sb-clean.h"
+#include "sb-counters.h"
+#include "sb-downgrade.h"
#include "sb-errors.h"
#include "sb-members.h"
#include "super-io.h"
struct bch2_metadata_version {
u16 version;
const char *name;
- u64 recovery_passes;
};
static const struct bch2_metadata_version bch2_metadata_versions[] = {
-#define x(n, v, _recovery_passes) { \
+#define x(n, v) { \
.version = v, \
.name = #n, \
- .recovery_passes = _recovery_passes, \
},
BCH_METADATA_VERSIONS()
#undef x
return v;
}
-u64 bch2_upgrade_recovery_passes(struct bch_fs *c,
- unsigned old_version,
- unsigned new_version)
-{
- u64 ret = 0;
-
- for (const struct bch2_metadata_version *i = bch2_metadata_versions;
- i < bch2_metadata_versions + ARRAY_SIZE(bch2_metadata_versions);
- i++)
- if (i->version > old_version && i->version <= new_version) {
- if (i->recovery_passes & RECOVERY_PASS_ALL_FSCK)
- ret |= bch2_fsck_recovery_passes();
- ret |= i->recovery_passes;
- }
-
- return ret &= ~RECOVERY_PASS_ALL_FSCK;
-}
-
const char * const bch2_sb_fields[] = {
#define x(name, nr) #name,
BCH_SB_FIELDS()
struct bch_sb_field *bch2_sb_field_get_id(struct bch_sb *sb,
enum bch_sb_field_type type)
{
- struct bch_sb_field *f;
-
/* XXX: need locking around superblock to access optional fields */
vstruct_for_each(sb, f)
void bch2_free_super(struct bch_sb_handle *sb)
{
kfree(sb->bio);
- if (!IS_ERR_OR_NULL(sb->bdev))
- blkdev_put(sb->bdev, sb->holder);
+ if (!IS_ERR_OR_NULL(sb->bdev_handle))
+ bdev_release(sb->bdev_handle);
kfree(sb->holder);
kfree(sb->sb_name);
u64 max_bytes = 512 << sb->sb->layout.sb_max_size_bits;
if (new_bytes > max_bytes) {
- pr_err("%pg: superblock too big: want %zu but have %llu",
- sb->bdev, new_bytes, max_bytes);
+ struct printbuf buf = PRINTBUF;
+
+ prt_bdevname(&buf, sb->bdev);
+ prt_printf(&buf, ": superblock too big: want %zu but have %llu", new_bytes, max_bytes);
+ pr_err("%s", buf.buf);
+ printbuf_exit(&buf);
return -BCH_ERR_ENOSPC_sb;
}
}
if (sb->fs_sb) {
struct bch_fs *c = container_of(sb, struct bch_fs, disk_sb);
- struct bch_dev *ca;
- unsigned i;
lockdep_assert_held(&c->sb_lock);
/* XXX: we're not checking that offline device have enough space */
- for_each_online_member(ca, c, i) {
+ for_each_online_member(c, ca) {
struct bch_sb_handle *dev_sb = &ca->disk_sb;
if (bch2_sb_realloc(dev_sb, le32_to_cpu(dev_sb->sb->u64s) + d)) {
return f;
}
+struct bch_sb_field *bch2_sb_field_get_minsize_id(struct bch_sb_handle *sb,
+ enum bch_sb_field_type type,
+ unsigned u64s)
+{
+ struct bch_sb_field *f = bch2_sb_field_get_id(sb->sb, type);
+
+ if (!f || le32_to_cpu(f->u64s) < u64s)
+ f = bch2_sb_field_resize_id(sb, type, u64s);
+ return f;
+}
+
/* Superblock validate: */
static int validate_sb_layout(struct bch_sb_layout *layout, struct printbuf *out)
int rw)
{
struct bch_sb *sb = disk_sb->sb;
- struct bch_sb_field *f;
struct bch_sb_field_members_v1 *mi;
enum bch_opt_id opt_id;
u16 block_size;
/* device open: */
+static unsigned long le_ulong_to_cpu(unsigned long v)
+{
+ return sizeof(unsigned long) == 8
+ ? le64_to_cpu(v)
+ : le32_to_cpu(v);
+}
+
+static void le_bitvector_to_cpu(unsigned long *dst, unsigned long *src, unsigned nr)
+{
+ BUG_ON(nr & (BITS_PER_TYPE(long) - 1));
+
+ for (unsigned i = 0; i < BITS_TO_LONGS(nr); i++)
+ dst[i] = le_ulong_to_cpu(src[i]);
+}
+
static void bch2_sb_update(struct bch_fs *c)
{
struct bch_sb *src = c->disk_sb.sb;
- struct bch_dev *ca;
- unsigned i;
lockdep_assert_held(&c->sb_lock);
c->sb.features = le64_to_cpu(src->features[0]);
c->sb.compat = le64_to_cpu(src->compat[0]);
- for_each_member_device(ca, c, i) {
- struct bch_member m = bch2_sb_member_get(src, i);
+ memset(c->sb.errors_silent, 0, sizeof(c->sb.errors_silent));
+
+ struct bch_sb_field_ext *ext = bch2_sb_field_get(src, ext);
+ if (ext)
+ le_bitvector_to_cpu(c->sb.errors_silent, (void *) ext->errors_silent,
+ sizeof(c->sb.errors_silent) * 8);
+
+ for_each_member_device(c, ca) {
+ struct bch_member m = bch2_sb_member_get(src, ca->dev_idx);
ca->mi = bch2_mi_to_cpu(&m);
}
}
dst->time_base_lo = src->time_base_lo;
dst->time_base_hi = src->time_base_hi;
dst->time_precision = src->time_precision;
+ dst->write_time = src->write_time;
memcpy(dst->flags, src->flags, sizeof(dst->flags));
memcpy(dst->features, src->features, sizeof(dst->features));
static int read_one_super(struct bch_sb_handle *sb, u64 offset, struct printbuf *err)
{
- struct bch_csum csum;
size_t bytes;
int ret;
reread:
if (!uuid_equal(&sb->sb->magic, &BCACHE_MAGIC) &&
!uuid_equal(&sb->sb->magic, &BCHFS_MAGIC)) {
- prt_printf(err, "Not a bcachefs superblock");
+ prt_str(err, "Not a bcachefs superblock (got magic ");
+ pr_uuid(err, sb->sb->magic.b);
+ prt_str(err, ")");
return -BCH_ERR_invalid_sb_magic;
}
goto reread;
}
- if (BCH_SB_CSUM_TYPE(sb->sb) >= BCH_CSUM_NR) {
+ enum bch_csum_type csum_type = BCH_SB_CSUM_TYPE(sb->sb);
+ if (csum_type >= BCH_CSUM_NR) {
prt_printf(err, "unknown checksum type %llu", BCH_SB_CSUM_TYPE(sb->sb));
return -BCH_ERR_invalid_sb_csum_type;
}
/* XXX: verify MACs */
- csum = csum_vstruct(NULL, BCH_SB_CSUM_TYPE(sb->sb),
- null_nonce(), sb->sb);
-
+ struct bch_csum csum = csum_vstruct(NULL, csum_type, null_nonce(), sb->sb);
if (bch2_crc_cmp(csum, sb->sb->csum)) {
- prt_printf(err, "bad checksum");
+ bch2_csum_err_msg(err, csum_type, sb->sb->csum, csum);
return -BCH_ERR_invalid_sb_csum;
}
if (!opt_get(*opts, nochanges))
sb->mode |= BLK_OPEN_WRITE;
- sb->bdev = blkdev_get_by_path(path, sb->mode, sb->holder, &bch2_sb_handle_bdev_ops);
- if (IS_ERR(sb->bdev) &&
- PTR_ERR(sb->bdev) == -EACCES &&
+ sb->bdev_handle = bdev_open_by_path(path, sb->mode, sb->holder, &bch2_sb_handle_bdev_ops);
+ if (IS_ERR(sb->bdev_handle) &&
+ PTR_ERR(sb->bdev_handle) == -EACCES &&
opt_get(*opts, read_only)) {
sb->mode &= ~BLK_OPEN_WRITE;
- sb->bdev = blkdev_get_by_path(path, sb->mode, sb->holder, &bch2_sb_handle_bdev_ops);
- if (!IS_ERR(sb->bdev))
+ sb->bdev_handle = bdev_open_by_path(path, sb->mode, sb->holder, &bch2_sb_handle_bdev_ops);
+ if (!IS_ERR(sb->bdev_handle))
opt_set(*opts, nochanges, true);
}
- if (IS_ERR(sb->bdev)) {
- ret = PTR_ERR(sb->bdev);
+ if (IS_ERR(sb->bdev_handle)) {
+ ret = PTR_ERR(sb->bdev_handle);
goto out;
}
+ sb->bdev = sb->bdev_handle->bdev;
ret = bch2_sb_realloc(sb, 0);
if (ret) {
int bch2_write_super(struct bch_fs *c)
{
struct closure *cl = &c->sb_write;
- struct bch_dev *ca;
struct printbuf err = PRINTBUF;
- unsigned i, sb = 0, nr_wrote;
+ unsigned sb = 0, nr_wrote;
struct bch_devs_mask sb_written;
bool wrote, can_mount_without_written, can_mount_with_written;
unsigned degraded_flags = BCH_FORCE_IF_DEGRADED;
le64_add_cpu(&c->disk_sb.sb->seq, 1);
+ struct bch_sb_field_members_v2 *mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
+ for_each_online_member(c, ca)
+ __bch2_members_v2_get_mut(mi, ca->dev_idx)->seq = c->disk_sb.sb->seq;
+ c->disk_sb.sb->write_time = cpu_to_le64(ktime_get_real_seconds());
+
if (test_bit(BCH_FS_error, &c->flags))
SET_BCH_SB_HAS_ERRORS(c->disk_sb.sb, 1);
if (test_bit(BCH_FS_topology_error, &c->flags))
bch2_sb_members_from_cpu(c);
bch2_sb_members_cpy_v2_v1(&c->disk_sb);
bch2_sb_errors_from_cpu(c);
+ bch2_sb_downgrade_update(c);
- for_each_online_member(ca, c, i)
+ for_each_online_member(c, ca)
bch2_sb_from_fs(c, ca);
- for_each_online_member(ca, c, i) {
+ for_each_online_member(c, ca) {
printbuf_reset(&err);
ret = bch2_sb_validate(&ca->disk_sb, &err, WRITE);
if (!BCH_SB_INITIALIZED(c->disk_sb.sb))
goto out;
- for_each_online_member(ca, c, i) {
+ if (le16_to_cpu(c->disk_sb.sb->version) > bcachefs_metadata_version_current) {
+ struct printbuf buf = PRINTBUF;
+ prt_printf(&buf, "attempting to write superblock that wasn't version downgraded (");
+ bch2_version_to_text(&buf, le16_to_cpu(c->disk_sb.sb->version));
+ prt_str(&buf, " > ");
+ bch2_version_to_text(&buf, bcachefs_metadata_version_current);
+ prt_str(&buf, ")");
+ bch2_fs_fatal_error(c, "%s", buf.buf);
+ printbuf_exit(&buf);
+ return -BCH_ERR_sb_not_downgraded;
+ }
+
+ for_each_online_member(c, ca) {
__set_bit(ca->dev_idx, sb_written.d);
ca->sb_write_error = 0;
}
- for_each_online_member(ca, c, i)
+ for_each_online_member(c, ca)
read_back_super(c, ca);
closure_sync(cl);
- for_each_online_member(ca, c, i) {
+ for_each_online_member(c, ca) {
if (ca->sb_write_error)
continue;
do {
wrote = false;
- for_each_online_member(ca, c, i)
+ for_each_online_member(c, ca)
if (!ca->sb_write_error &&
sb < ca->disk_sb.sb->layout.nr_superblocks) {
write_one_super(c, ca, sb);
sb++;
} while (wrote);
- for_each_online_member(ca, c, i) {
+ for_each_online_member(c, ca) {
if (ca->sb_write_error)
__clear_bit(ca->dev_idx, sb_written.d);
else
can_mount_with_written =
bch2_have_enough_devs(c, sb_written, degraded_flags, false);
- for (i = 0; i < ARRAY_SIZE(sb_written.d); i++)
+ for (unsigned i = 0; i < ARRAY_SIZE(sb_written.d); i++)
sb_written.d[i] = ~sb_written.d[i];
can_mount_without_written =
}
/* Downgrade if superblock is at a higher version than currently supported: */
-void bch2_sb_maybe_downgrade(struct bch_fs *c)
+bool bch2_check_version_downgrade(struct bch_fs *c)
{
+ bool ret = bcachefs_metadata_version_current < c->sb.version;
+
lockdep_assert_held(&c->sb_lock);
/*
* Downgrade, if superblock is at a higher version than currently
* supported:
+ *
+ * c->sb will be checked before we write the superblock, so update it as
+ * well:
*/
- if (BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb) > bcachefs_metadata_version_current)
+ if (BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb) > bcachefs_metadata_version_current) {
SET_BCH_SB_VERSION_UPGRADE_COMPLETE(c->disk_sb.sb, bcachefs_metadata_version_current);
- if (c->sb.version > bcachefs_metadata_version_current)
+ c->sb.version_upgrade_complete = bcachefs_metadata_version_current;
+ }
+ if (c->sb.version > bcachefs_metadata_version_current) {
c->disk_sb.sb->version = cpu_to_le16(bcachefs_metadata_version_current);
- if (c->sb.version_min > bcachefs_metadata_version_current)
+ c->sb.version = bcachefs_metadata_version_current;
+ }
+ if (c->sb.version_min > bcachefs_metadata_version_current) {
c->disk_sb.sb->version_min = cpu_to_le16(bcachefs_metadata_version_current);
+ c->sb.version_min = bcachefs_metadata_version_current;
+ }
c->disk_sb.sb->compat[0] &= cpu_to_le64((1ULL << BCH_COMPAT_NR) - 1);
+ return ret;
}
void bch2_sb_upgrade(struct bch_fs *c, unsigned new_version)
{
lockdep_assert_held(&c->sb_lock);
+ if (BCH_VERSION_MAJOR(new_version) >
+ BCH_VERSION_MAJOR(le16_to_cpu(c->disk_sb.sb->version)))
+ bch2_sb_field_resize(&c->disk_sb, downgrade, 0);
+
c->disk_sb.sb->version = cpu_to_le16(new_version);
c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALL);
}
+static int bch2_sb_ext_validate(struct bch_sb *sb, struct bch_sb_field *f,
+ struct printbuf *err)
+{
+ if (vstruct_bytes(f) < 88) {
+ prt_printf(err, "field too small (%zu < %u)", vstruct_bytes(f), 88);
+ return -BCH_ERR_invalid_sb_ext;
+ }
+
+ return 0;
+}
+
+static void bch2_sb_ext_to_text(struct printbuf *out, struct bch_sb *sb,
+ struct bch_sb_field *f)
+{
+ struct bch_sb_field_ext *e = field_to_type(f, ext);
+
+ prt_printf(out, "Recovery passes required:");
+ prt_tab(out);
+ prt_bitflags(out, bch2_recovery_passes,
+ bch2_recovery_passes_from_stable(le64_to_cpu(e->recovery_passes_required[0])));
+ prt_newline(out);
+
+ unsigned long *errors_silent = kmalloc(sizeof(e->errors_silent), GFP_KERNEL);
+ if (errors_silent) {
+ le_bitvector_to_cpu(errors_silent, (void *) e->errors_silent, sizeof(e->errors_silent) * 8);
+
+ prt_printf(out, "Errors to silently fix:");
+ prt_tab(out);
+ prt_bitflags_vector(out, bch2_sb_error_strs, errors_silent, sizeof(e->errors_silent) * 8);
+ prt_newline(out);
+
+ kfree(errors_silent);
+ }
+}
+
+static const struct bch_sb_field_ops bch_sb_field_ops_ext = {
+ .validate = bch2_sb_ext_validate,
+ .to_text = bch2_sb_ext_to_text,
+};
+
static const struct bch_sb_field_ops *bch2_sb_field_ops[] = {
#define x(f, nr) \
[BCH_SB_FIELD_##f] = &bch_sb_field_ops_##f,
return ret;
}
-void bch2_sb_field_to_text(struct printbuf *out, struct bch_sb *sb,
- struct bch_sb_field *f)
+void __bch2_sb_field_to_text(struct printbuf *out, struct bch_sb *sb,
+ struct bch_sb_field *f)
{
unsigned type = le32_to_cpu(f->type);
const struct bch_sb_field_ops *ops = bch2_sb_field_type_ops(type);
if (!out->nr_tabstops)
printbuf_tabstop_push(out, 32);
+ if (ops->to_text)
+ ops->to_text(out, sb, f);
+}
+
+void bch2_sb_field_to_text(struct printbuf *out, struct bch_sb *sb,
+ struct bch_sb_field *f)
+{
+ unsigned type = le32_to_cpu(f->type);
+
if (type < BCH_SB_FIELD_NR)
prt_printf(out, "%s", bch2_sb_fields[type]);
else
prt_printf(out, " (size %zu):", vstruct_bytes(f));
prt_newline(out);
- if (ops->to_text) {
- printbuf_indent_add(out, 2);
- ops->to_text(out, sb, f);
- printbuf_indent_sub(out, 2);
- }
+ __bch2_sb_field_to_text(out, sb, f);
}
void bch2_sb_layout_to_text(struct printbuf *out, struct bch_sb_layout *l)
void bch2_sb_to_text(struct printbuf *out, struct bch_sb *sb,
bool print_layout, unsigned fields)
{
- struct bch_sb_field *f;
u64 fields_have = 0;
unsigned nr_devices = 0;
pr_uuid(out, sb->uuid.b);
prt_newline(out);
+ prt_printf(out, "Magic number:");
+ prt_tab(out);
+ pr_uuid(out, sb->magic.b);
+ prt_newline(out);
+
prt_str(out, "Device index:");
prt_tab(out);
prt_printf(out, "%u", sb->dev_idx);
prt_printf(out, "%llu", le64_to_cpu(sb->seq));
prt_newline(out);
+ prt_printf(out, "Time of last write:");
+ prt_tab(out);
+ bch2_prt_datetime(out, le64_to_cpu(sb->write_time));
+ prt_newline(out);
+
prt_printf(out, "Superblock size:");
prt_tab(out);
- prt_printf(out, "%zu", vstruct_bytes(sb));
+ prt_units_u64(out, vstruct_bytes(sb));
+ prt_str(out, "/");
+ prt_units_u64(out, 512ULL << sb->layout.sb_max_size_bits);
prt_newline(out);
prt_printf(out, "Clean:");
projects / bcachefs-tools-debian / blobdiff