#include "error.h"
#include "io.h"
#include "journal.h"
+#include "journal_io.h"
#include "journal_seq_blacklist.h"
#include "replicas.h"
#include "quota.h"
unsigned old_u64s = f ? le32_to_cpu(f->u64s) : 0;
unsigned sb_u64s = le32_to_cpu(sb->sb->u64s) + u64s - old_u64s;
- BUG_ON(get_order(__vstruct_bytes(struct bch_sb, sb_u64s)) >
- sb->page_order);
+ BUG_ON(__vstruct_bytes(struct bch_sb, sb_u64s) > sb->buffer_size);
if (!f && !u64s) {
/* nothing to do: */
if (!IS_ERR_OR_NULL(sb->bdev))
blkdev_put(sb->bdev, sb->mode);
- free_pages((unsigned long) sb->sb, sb->page_order);
+ kfree(sb->sb);
memset(sb, 0, sizeof(*sb));
}
int bch2_sb_realloc(struct bch_sb_handle *sb, unsigned u64s)
{
size_t new_bytes = __vstruct_bytes(struct bch_sb, u64s);
- unsigned order = get_order(new_bytes);
+ size_t new_buffer_size;
struct bch_sb *new_sb;
struct bio *bio;
- if (sb->sb && sb->page_order >= order)
+ if (sb->bdev)
+ new_bytes = max_t(size_t, new_bytes, bdev_logical_block_size(sb->bdev));
+
+ new_buffer_size = roundup_pow_of_two(new_bytes);
+
+ if (sb->sb && sb->buffer_size >= new_buffer_size)
return 0;
if (sb->have_layout) {
}
}
- if (sb->page_order >= order && sb->sb)
+ if (sb->buffer_size >= new_buffer_size && sb->sb)
return 0;
if (dynamic_fault("bcachefs:add:super_realloc"))
return -ENOMEM;
if (sb->have_bio) {
- bio = bio_kmalloc(GFP_KERNEL, 1 << order);
+ bio = bio_kmalloc(GFP_KERNEL,
+ DIV_ROUND_UP(new_buffer_size, PAGE_SIZE));
if (!bio)
return -ENOMEM;
sb->bio = bio;
}
- new_sb = (void *) __get_free_pages(GFP_NOFS|__GFP_ZERO, order);
+ new_sb = krealloc(sb->sb, new_buffer_size, GFP_NOFS|__GFP_ZERO);
if (!new_sb)
return -ENOMEM;
- if (sb->sb)
- memcpy(new_sb, sb->sb, PAGE_SIZE << sb->page_order);
-
- free_pages((unsigned long) sb->sb, sb->page_order);
sb->sb = new_sb;
-
- sb->page_order = order;
+ sb->buffer_size = new_buffer_size;
return 0;
}
c->sb.uuid = src->uuid;
c->sb.user_uuid = src->user_uuid;
c->sb.version = le16_to_cpu(src->version);
+ c->sb.version_min = le16_to_cpu(src->version_min);
c->sb.nr_devices = src->nr_devices;
c->sb.clean = BCH_SB_CLEAN(src);
c->sb.encryption_type = BCH_SB_ENCRYPTION_TYPE(src);
c->sb.encoded_extent_max= 1 << BCH_SB_ENCODED_EXTENT_MAX_BITS(src);
- c->sb.time_base_lo = le64_to_cpu(src->time_base_lo);
+
+ c->sb.nsec_per_time_unit = le32_to_cpu(src->time_precision);
+ c->sb.time_units_per_sec = NSEC_PER_SEC / c->sb.nsec_per_time_unit;
+
+ /* XXX this is wrong, we need a 96 or 128 bit integer type */
+ c->sb.time_base_lo = div_u64(le64_to_cpu(src->time_base_lo),
+ c->sb.nsec_per_time_unit);
c->sb.time_base_hi = le32_to_cpu(src->time_base_hi);
- c->sb.time_precision = le32_to_cpu(src->time_precision);
+
c->sb.features = le64_to_cpu(src->features[0]);
c->sb.compat = le64_to_cpu(src->compat[0]);
ca->mi = bch2_mi_to_cpu(mi->members + i);
}
-/* doesn't copy member info */
static void __copy_super(struct bch_sb_handle *dst_handle, struct bch_sb *src)
{
struct bch_sb_field *src_f, *dst_f;
__copy_super(&c->disk_sb, src);
+ if (BCH_SB_HAS_ERRORS(c->disk_sb.sb))
+ set_bit(BCH_FS_ERROR, &c->flags);
+ if (BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb))
+ set_bit(BCH_FS_TOPOLOGY_ERROR, &c->flags);
+
ret = bch2_sb_replicas_to_cpu_replicas(c);
if (ret)
return ret;
bio_set_dev(sb->bio, sb->bdev);
sb->bio->bi_iter.bi_sector = offset;
bio_set_op_attrs(sb->bio, REQ_OP_READ, REQ_SYNC|REQ_META);
- bch2_bio_map(sb->bio, sb->sb, PAGE_SIZE << sb->page_order);
+ bch2_bio_map(sb->bio, sb->sb, sb->buffer_size);
if (submit_bio_wait(sb->bio))
return "IO error";
if (bytes > 512 << sb->sb->layout.sb_max_size_bits)
return "Bad superblock: too big";
- if (get_order(bytes) > sb->page_order) {
+ if (bytes > sb->buffer_size) {
if (bch2_sb_realloc(sb, le32_to_cpu(sb->sb->u64s)))
return "cannot allocate memory";
goto reread;
sb->offset = sb->layout.sb_offset[idx];
- SET_BCH_SB_CSUM_TYPE(sb, c->opts.metadata_checksum);
+ SET_BCH_SB_CSUM_TYPE(sb, bch2_csum_opt_to_type(c->opts.metadata_checksum, false));
sb->csum = csum_vstruct(c, BCH_SB_CSUM_TYPE(sb),
null_nonce(), sb);
const char *err;
struct bch_devs_mask sb_written;
bool wrote, can_mount_without_written, can_mount_with_written;
+ unsigned degraded_flags = BCH_FORCE_IF_DEGRADED;
int ret = 0;
+ if (c->opts.very_degraded)
+ degraded_flags |= BCH_FORCE_IF_LOST;
+
lockdep_assert_held(&c->sb_lock);
closure_init_stack(cl);
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))
+ SET_BCH_SB_HAS_TOPOLOGY_ERRORS(c->disk_sb.sb, 1);
+
+ SET_BCH_SB_BIG_ENDIAN(c->disk_sb.sb, CPU_BIG_ENDIAN);
for_each_online_member(ca, c, i)
bch2_sb_from_fs(c, ca);
nr_wrote = dev_mask_nr(&sb_written);
can_mount_with_written =
- bch2_have_enough_devs(c, sb_written, BCH_FORCE_IF_DEGRADED, false);
+ bch2_have_enough_devs(c, sb_written, degraded_flags, false);
for (i = 0; i < ARRAY_SIZE(sb_written.d); i++)
sb_written.d[i] = ~sb_written.d[i];
can_mount_without_written =
- bch2_have_enough_devs(c, sb_written, BCH_FORCE_IF_DEGRADED, false);
+ bch2_have_enough_devs(c, sb_written, degraded_flags, false);
/*
* If we would be able to mount _without_ the devices we successfully
/* BCH_SB_FIELD_clean: */
-void bch2_sb_clean_renumber(struct bch_sb_field_clean *clean, int write)
+int bch2_sb_clean_validate(struct bch_fs *c, struct bch_sb_field_clean *clean, int write)
{
struct jset_entry *entry;
+ int ret;
for (entry = clean->start;
entry < (struct jset_entry *) vstruct_end(&clean->field);
- entry = vstruct_next(entry))
- bch2_bkey_renumber(BKEY_TYPE_BTREE, bkey_to_packed(entry->start), write);
+ entry = vstruct_next(entry)) {
+ ret = bch2_journal_entry_validate(c, "superblock", entry,
+ le16_to_cpu(c->disk_sb.sb->version),
+ BCH_SB_BIG_ENDIAN(c->disk_sb.sb),
+ write);
+ if (ret)
+ return ret;
+ }
+
+ return 0;
}
int bch2_fs_mark_dirty(struct bch_fs *c)
mutex_lock(&c->sb_lock);
SET_BCH_SB_CLEAN(c->disk_sb.sb, false);
- c->disk_sb.sb->features[0] |= 1ULL << BCH_FEATURE_new_extent_overwrite;
- c->disk_sb.sb->features[0] |= 1ULL << BCH_FEATURE_extents_above_btree_updates;
- c->disk_sb.sb->features[0] |= 1ULL << BCH_FEATURE_btree_updates_journalled;
+ c->disk_sb.sb->features[0] |= cpu_to_le64(BCH_SB_FEATURES_ALWAYS);
ret = bch2_write_super(c);
mutex_unlock(&c->sb_lock);
* The u64s field counts from the start of data, ignoring the shared
* fields.
*/
- entry->u64s = u64s - 1;
+ entry->u64s = cpu_to_le16(u64s - 1);
*end = vstruct_next(*end);
return entry;
struct bch_dev *ca;
unsigned i, dev;
- percpu_down_write(&c->mark_lock);
+ percpu_down_read(&c->mark_lock);
if (!journal_seq) {
for (i = 0; i < ARRAY_SIZE(c->usage); i++)
}
}
- percpu_up_write(&c->mark_lock);
+ percpu_up_read(&c->mark_lock);
for (i = 0; i < 2; i++) {
struct jset_entry_clock *clock =
clock->entry.type = BCH_JSET_ENTRY_clock;
clock->rw = i;
- clock->time = atomic64_read(&c->io_clock[i].now);
+ clock->time = cpu_to_le64(atomic64_read(&c->io_clock[i].now));
}
}
struct bch_sb_field_clean *sb_clean;
struct jset_entry *entry;
unsigned u64s;
+ int ret;
mutex_lock(&c->sb_lock);
if (BCH_SB_CLEAN(c->disk_sb.sb))
SET_BCH_SB_CLEAN(c->disk_sb.sb, true);
- c->disk_sb.sb->compat[0] |= 1ULL << BCH_COMPAT_FEAT_ALLOC_INFO;
- c->disk_sb.sb->compat[0] |= 1ULL << BCH_COMPAT_FEAT_ALLOC_METADATA;
- c->disk_sb.sb->features[0] &= ~(1ULL << BCH_FEATURE_extents_above_btree_updates);
- c->disk_sb.sb->features[0] &= ~(1ULL << BCH_FEATURE_btree_updates_journalled);
+ c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_alloc_info);
+ c->disk_sb.sb->compat[0] |= cpu_to_le64(1ULL << BCH_COMPAT_alloc_metadata);
+ c->disk_sb.sb->features[0] &= cpu_to_le64(~(1ULL << BCH_FEATURE_extents_above_btree_updates));
+ c->disk_sb.sb->features[0] &= cpu_to_le64(~(1ULL << BCH_FEATURE_btree_updates_journalled));
u64s = sizeof(*sb_clean) / sizeof(u64) + c->journal.entry_u64s_reserved;
memset(entry, 0,
vstruct_end(&sb_clean->field) - (void *) entry);
- if (le16_to_cpu(c->disk_sb.sb->version) <
- bcachefs_metadata_version_bkey_renumber)
- bch2_sb_clean_renumber(sb_clean, WRITE);
+ /*
+ * this should be in the write path, and we should be validating every
+ * superblock section:
+ */
+ ret = bch2_sb_clean_validate(c, sb_clean, WRITE);
+ if (ret) {
+ bch_err(c, "error writing marking filesystem clean: validate error");
+ goto out;
+ }
bch2_write_super(c);
out:
projects / bcachefs-tools-debian / blobdiff