#include "fs-io-direct.h"
#include "fsck.h"
#include "inode.h"
-#include "io.h"
+#include "io_read.h"
+#include "io_write.h"
#include "journal.h"
#include "journal_reclaim.h"
#include "journal_seq_blacklist.h"
#include "recovery.h"
#include "replicas.h"
#include "sb-clean.h"
+#include "sb-errors.h"
+#include "sb-members.h"
+#include "snapshot.h"
#include "subvolume.h"
#include "super.h"
#include "super-io.h"
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Kent Overstreet <kent.overstreet@gmail.com>");
+MODULE_DESCRIPTION("bcachefs filesystem");
+MODULE_SOFTDEP("pre: crc32c");
+MODULE_SOFTDEP("pre: crc64");
+MODULE_SOFTDEP("pre: sha256");
+MODULE_SOFTDEP("pre: chacha20");
+MODULE_SOFTDEP("pre: poly1305");
+MODULE_SOFTDEP("pre: xxhash");
+
+const char * const bch2_fs_flag_strs[] = {
+#define x(n) #n,
+ BCH_FS_FLAGS()
+#undef x
+ NULL
+};
+
+void __bch2_print(struct bch_fs *c, const char *fmt, ...)
+{
+ struct log_output *output = c->output;
+ va_list args;
+
+ if (c->output_filter && c->output_filter != current)
+ output = NULL;
+
+ va_start(args, fmt);
+ if (likely(!output)) {
+ vprintk(fmt, args);
+ } else {
+ unsigned long flags;
+
+ if (fmt[0] == KERN_SOH[0])
+ fmt += 2;
+
+ spin_lock_irqsave(&output->lock, flags);
+ prt_vprintf(&output->buf, fmt, args);
+ spin_unlock_irqrestore(&output->lock, flags);
+
+ wake_up(&output->wait);
+ }
+ va_end(args);
+}
#define KTYPE(type) \
static const struct attribute_group type ## _group = { \
struct bch_fs *bch2_dev_to_fs(dev_t dev)
{
struct bch_fs *c;
- struct bch_dev *ca;
- unsigned i;
mutex_lock(&bch_fs_list_lock);
rcu_read_lock();
list_for_each_entry(c, &bch_fs_list, list)
- for_each_member_device_rcu(ca, c, i, NULL)
+ for_each_member_device_rcu(c, ca, NULL)
if (ca->disk_sb.bdev && ca->disk_sb.bdev->bd_dev == dev) {
closure_get(&c->cl);
goto found;
static void bch2_dev_usage_journal_reserve(struct bch_fs *c)
{
- struct bch_dev *ca;
- unsigned i, nr = 0, u64s =
+ unsigned nr = 0, u64s =
((sizeof(struct jset_entry_dev_usage) +
sizeof(struct jset_entry_dev_usage_type) * BCH_DATA_NR)) /
sizeof(u64);
rcu_read_lock();
- for_each_member_device_rcu(ca, c, i, NULL)
+ for_each_member_device_rcu(c, ca, NULL)
nr++;
rcu_read_unlock();
static void __bch2_fs_read_only(struct bch_fs *c)
{
- struct bch_dev *ca;
- unsigned i, clean_passes = 0;
+ unsigned clean_passes = 0;
u64 seq = 0;
bch2_fs_ec_stop(c);
journal_cur_seq(&c->journal));
if (test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags) &&
- !test_bit(BCH_FS_EMERGENCY_RO, &c->flags))
- set_bit(BCH_FS_CLEAN_SHUTDOWN, &c->flags);
+ !test_bit(BCH_FS_emergency_ro, &c->flags))
+ set_bit(BCH_FS_clean_shutdown, &c->flags);
bch2_fs_journal_stop(&c->journal);
/*
* After stopping journal:
*/
- for_each_member_device(ca, c, i)
+ for_each_member_device(c, ca)
bch2_dev_allocator_remove(c, ca);
}
{
struct bch_fs *c = container_of(writes, struct bch_fs, writes);
- set_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags);
+ set_bit(BCH_FS_write_disable_complete, &c->flags);
wake_up(&bch2_read_only_wait);
}
#endif
void bch2_fs_read_only(struct bch_fs *c)
{
- if (!test_bit(BCH_FS_RW, &c->flags)) {
+ if (!test_bit(BCH_FS_rw, &c->flags)) {
bch2_journal_reclaim_stop(&c->journal);
return;
}
- BUG_ON(test_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags));
+ BUG_ON(test_bit(BCH_FS_write_disable_complete, &c->flags));
+
+ bch_verbose(c, "going read-only");
/*
* Block new foreground-end write operations from starting - any new
* writes will return -EROFS:
*/
- set_bit(BCH_FS_GOING_RO, &c->flags);
+ set_bit(BCH_FS_going_ro, &c->flags);
#ifndef BCH_WRITE_REF_DEBUG
percpu_ref_kill(&c->writes);
#else
* that going RO is complete:
*/
wait_event(bch2_read_only_wait,
- test_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags) ||
- test_bit(BCH_FS_EMERGENCY_RO, &c->flags));
+ test_bit(BCH_FS_write_disable_complete, &c->flags) ||
+ test_bit(BCH_FS_emergency_ro, &c->flags));
+
+ bool writes_disabled = test_bit(BCH_FS_write_disable_complete, &c->flags);
+ if (writes_disabled)
+ bch_verbose(c, "finished waiting for writes to stop");
__bch2_fs_read_only(c);
wait_event(bch2_read_only_wait,
- test_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags));
+ test_bit(BCH_FS_write_disable_complete, &c->flags));
+
+ if (!writes_disabled)
+ bch_verbose(c, "finished waiting for writes to stop");
- clear_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags);
- clear_bit(BCH_FS_GOING_RO, &c->flags);
+ clear_bit(BCH_FS_write_disable_complete, &c->flags);
+ clear_bit(BCH_FS_going_ro, &c->flags);
+ clear_bit(BCH_FS_rw, &c->flags);
if (!bch2_journal_error(&c->journal) &&
- !test_bit(BCH_FS_ERROR, &c->flags) &&
- !test_bit(BCH_FS_EMERGENCY_RO, &c->flags) &&
- test_bit(BCH_FS_STARTED, &c->flags) &&
- test_bit(BCH_FS_CLEAN_SHUTDOWN, &c->flags) &&
+ !test_bit(BCH_FS_error, &c->flags) &&
+ !test_bit(BCH_FS_emergency_ro, &c->flags) &&
+ test_bit(BCH_FS_started, &c->flags) &&
+ test_bit(BCH_FS_clean_shutdown, &c->flags) &&
!c->opts.norecovery) {
BUG_ON(c->journal.last_empty_seq != journal_cur_seq(&c->journal));
BUG_ON(atomic_read(&c->btree_cache.dirty));
BUG_ON(atomic_long_read(&c->btree_key_cache.nr_dirty));
- BUG_ON(c->btree_write_buffer.state.nr);
+ BUG_ON(c->btree_write_buffer.inc.keys.nr);
+ BUG_ON(c->btree_write_buffer.flushing.keys.nr);
bch_verbose(c, "marking filesystem clean");
bch2_fs_mark_clean(c);
+ } else {
+ bch_verbose(c, "done going read-only, filesystem not clean");
}
-
- clear_bit(BCH_FS_RW, &c->flags);
}
static void bch2_fs_read_only_work(struct work_struct *work)
bool bch2_fs_emergency_read_only(struct bch_fs *c)
{
- bool ret = !test_and_set_bit(BCH_FS_EMERGENCY_RO, &c->flags);
+ bool ret = !test_and_set_bit(BCH_FS_emergency_ro, &c->flags);
bch2_journal_halt(&c->journal);
bch2_fs_read_only_async(c);
static int __bch2_fs_read_write(struct bch_fs *c, bool early)
{
- struct bch_dev *ca;
- unsigned i;
int ret;
- if (test_bit(BCH_FS_INITIAL_GC_UNFIXED, &c->flags)) {
+ if (test_bit(BCH_FS_initial_gc_unfixed, &c->flags)) {
bch_err(c, "cannot go rw, unfixed btree errors");
return -BCH_ERR_erofs_unfixed_errors;
}
- if (test_bit(BCH_FS_RW, &c->flags))
+ if (test_bit(BCH_FS_rw, &c->flags))
return 0;
if (c->opts.norecovery)
bch_info(c, "going read-write");
+ ret = bch2_sb_members_v2_init(c);
+ if (ret)
+ goto err;
+
ret = bch2_fs_mark_dirty(c);
if (ret)
goto err;
- clear_bit(BCH_FS_CLEAN_SHUTDOWN, &c->flags);
+ clear_bit(BCH_FS_clean_shutdown, &c->flags);
/*
* First journal write must be a flush write: after a clean shutdown we
*/
set_bit(JOURNAL_NEED_FLUSH_WRITE, &c->journal.flags);
- for_each_rw_member(ca, c, i)
+ for_each_rw_member(c, ca)
bch2_dev_allocator_add(c, ca);
bch2_recalc_capacity(c);
+ set_bit(BCH_FS_rw, &c->flags);
+ set_bit(BCH_FS_was_rw, &c->flags);
+
+#ifndef BCH_WRITE_REF_DEBUG
+ percpu_ref_reinit(&c->writes);
+#else
+ for (unsigned i = 0; i < BCH_WRITE_REF_NR; i++) {
+ BUG_ON(atomic_long_read(&c->writes[i]));
+ atomic_long_inc(&c->writes[i]);
+ }
+#endif
+
ret = bch2_gc_thread_start(c);
if (ret) {
bch_err(c, "error starting gc thread");
return ret;
}
+ ret = bch2_journal_reclaim_start(&c->journal);
+ if (ret)
+ goto err;
+
if (!early) {
ret = bch2_fs_read_write_late(c);
if (ret)
goto err;
}
-#ifndef BCH_WRITE_REF_DEBUG
- percpu_ref_reinit(&c->writes);
-#else
- for (unsigned i = 0; i < BCH_WRITE_REF_NR; i++) {
- BUG_ON(atomic_long_read(&c->writes[i]));
- atomic_long_inc(&c->writes[i]);
- }
-#endif
- set_bit(BCH_FS_RW, &c->flags);
- set_bit(BCH_FS_WAS_RW, &c->flags);
-
bch2_do_discards(c);
bch2_do_invalidates(c);
bch2_do_stripe_deletes(c);
bch2_do_pending_node_rewrites(c);
return 0;
err:
- __bch2_fs_read_only(c);
+ if (test_bit(BCH_FS_rw, &c->flags))
+ bch2_fs_read_only(c);
+ else
+ __bch2_fs_read_only(c);
return ret;
}
static void __bch2_fs_free(struct bch_fs *c)
{
unsigned i;
- int cpu;
for (i = 0; i < BCH_TIME_STAT_NR; i++)
bch2_time_stats_exit(&c->times[i]);
bch2_free_pending_node_rewrites(c);
+ bch2_fs_sb_errors_exit(c);
bch2_fs_counters_exit(c);
bch2_fs_snapshots_exit(c);
bch2_fs_quota_exit(c);
bch2_fs_fsio_exit(c);
bch2_fs_ec_exit(c);
bch2_fs_encryption_exit(c);
- bch2_fs_io_exit(c);
+ bch2_fs_nocow_locking_exit(c);
+ bch2_fs_io_write_exit(c);
+ bch2_fs_io_read_exit(c);
bch2_fs_buckets_waiting_for_journal_exit(c);
bch2_fs_btree_interior_update_exit(c);
bch2_fs_btree_iter_exit(c);
bch2_io_clock_exit(&c->io_clock[WRITE]);
bch2_io_clock_exit(&c->io_clock[READ]);
bch2_fs_compress_exit(c);
- bch2_journal_keys_free(&c->journal_keys);
- bch2_journal_entries_free(c);
+ bch2_journal_keys_put_initial(c);
+ BUG_ON(atomic_read(&c->journal_keys.ref));
bch2_fs_btree_write_buffer_exit(c);
percpu_free_rwsem(&c->mark_lock);
free_percpu(c->online_reserved);
- if (c->btree_paths_bufs)
- for_each_possible_cpu(cpu)
- kfree(per_cpu_ptr(c->btree_paths_bufs, cpu)->path);
-
darray_exit(&c->btree_roots_extra);
- free_percpu(c->btree_paths_bufs);
free_percpu(c->pcpu);
mempool_exit(&c->large_bkey_pool);
mempool_exit(&c->btree_bounce_pool);
void __bch2_fs_stop(struct bch_fs *c)
{
- struct bch_dev *ca;
- unsigned i;
-
bch_verbose(c, "shutting down");
- set_bit(BCH_FS_STOPPING, &c->flags);
+ set_bit(BCH_FS_stopping, &c->flags);
cancel_work_sync(&c->journal_seq_blacklist_gc_work);
bch2_fs_read_only(c);
up_write(&c->state_lock);
- for_each_member_device(ca, c, i)
+ for_each_member_device(c, ca)
if (ca->kobj.state_in_sysfs &&
ca->disk_sb.bdev)
sysfs_remove_link(bdev_kobj(ca->disk_sb.bdev), "bcachefs");
bch2_fs_debug_exit(c);
bch2_fs_chardev_exit(c);
+ bch2_ro_ref_put(c);
+ wait_event(c->ro_ref_wait, !refcount_read(&c->ro_ref));
+
kobject_put(&c->counters_kobj);
kobject_put(&c->time_stats);
kobject_put(&c->opts_dir);
/* btree prefetch might have kicked off reads in the background: */
bch2_btree_flush_all_reads(c);
- for_each_member_device(ca, c, i)
+ for_each_member_device(c, ca)
cancel_work_sync(&ca->io_error_work);
cancel_work_sync(&c->read_only_work);
-
- for (i = 0; i < c->sb.nr_devices; i++) {
- struct bch_dev *ca = rcu_dereference_protected(c->devs[i], true);
-
- if (ca)
- bch2_free_super(&ca->disk_sb);
- }
}
void bch2_fs_free(struct bch_fs *c)
closure_sync(&c->cl);
closure_debug_destroy(&c->cl);
- for (i = 0; i < c->sb.nr_devices; i++)
- if (c->devs[i])
- bch2_dev_free(rcu_dereference_protected(c->devs[i], 1));
+ for (i = 0; i < c->sb.nr_devices; i++) {
+ struct bch_dev *ca = rcu_dereference_protected(c->devs[i], true);
+
+ if (ca) {
+ bch2_free_super(&ca->disk_sb);
+ bch2_dev_free(ca);
+ }
+ }
bch_verbose(c, "shutdown complete");
static int bch2_fs_online(struct bch_fs *c)
{
- struct bch_dev *ca;
- unsigned i;
int ret = 0;
lockdep_assert_held(&bch_fs_list_lock);
ret = kobject_add(&c->kobj, NULL, "%pU", c->sb.user_uuid.b) ?:
kobject_add(&c->internal, &c->kobj, "internal") ?:
kobject_add(&c->opts_dir, &c->kobj, "options") ?:
+#ifndef CONFIG_BCACHEFS_NO_LATENCY_ACCT
kobject_add(&c->time_stats, &c->kobj, "time_stats") ?:
+#endif
kobject_add(&c->counters_kobj, &c->kobj, "counters") ?:
bch2_opts_create_sysfs_files(&c->opts_dir);
if (ret) {
down_write(&c->state_lock);
- for_each_member_device(ca, c, i) {
+ for_each_member_device(c, ca) {
ret = bch2_dev_sysfs_online(c, ca);
if (ret) {
bch_err(c, "error creating sysfs objects");
static struct bch_fs *bch2_fs_alloc(struct bch_sb *sb, struct bch_opts opts)
{
- struct bch_sb_field_members *mi;
struct bch_fs *c;
struct printbuf name = PRINTBUF;
unsigned i, iter_size;
goto out;
}
+ c->output = (void *)(unsigned long) opts.log_output;
+
__module_get(THIS_MODULE);
closure_init(&c->cl, NULL);
mutex_init(&c->btree_root_lock);
INIT_WORK(&c->read_only_work, bch2_fs_read_only_work);
+ refcount_set(&c->ro_ref, 1);
+ init_waitqueue_head(&c->ro_ref_wait);
+ sema_init(&c->online_fsck_mutex, 1);
+
init_rwsem(&c->gc_lock);
mutex_init(&c->gc_gens_lock);
+ atomic_set(&c->journal_keys.ref, 1);
+ c->journal_keys.initial_ref_held = true;
for (i = 0; i < BCH_TIME_STAT_NR; i++)
bch2_time_stats_init(&c->times[i]);
bch2_fs_copygc_init(c);
bch2_fs_btree_key_cache_init_early(&c->btree_key_cache);
+ bch2_fs_btree_iter_init_early(c);
bch2_fs_btree_interior_update_init_early(c);
bch2_fs_allocator_background_init(c);
bch2_fs_allocator_foreground_init(c);
bch2_fs_quota_init(c);
bch2_fs_ec_init_early(c);
bch2_fs_move_init(c);
+ bch2_fs_sb_errors_init_early(c);
INIT_LIST_HEAD(&c->list);
mutex_init(&c->bio_bounce_pages_lock);
mutex_init(&c->snapshot_table_lock);
+ init_rwsem(&c->snapshot_create_lock);
spin_lock_init(&c->btree_write_error_lock);
INIT_LIST_HEAD(&c->journal_iters);
- INIT_LIST_HEAD(&c->fsck_errors);
- mutex_init(&c->fsck_error_lock);
+ INIT_LIST_HEAD(&c->fsck_error_msgs);
+ mutex_init(&c->fsck_error_msgs_lock);
seqcount_init(&c->gc_pos_lock);
c->journal.flush_write_time = &c->times[BCH_TIME_journal_flush_write];
c->journal.noflush_write_time = &c->times[BCH_TIME_journal_noflush_write];
- c->journal.blocked_time = &c->times[BCH_TIME_blocked_journal];
c->journal.flush_seq_time = &c->times[BCH_TIME_journal_flush_seq];
bch2_fs_btree_cache_init_early(&c->btree_cache);
c->btree_key_cache_btrees |= 1U << BTREE_ID_alloc;
if (c->opts.inodes_use_key_cache)
c->btree_key_cache_btrees |= 1U << BTREE_ID_inodes;
+ c->btree_key_cache_btrees |= 1U << BTREE_ID_logged_ops;
c->block_bits = ilog2(block_sectors(c));
c->btree_foreground_merge_threshold = BTREE_FOREGROUND_MERGE_THRESHOLD(c);
BIOSET_NEED_BVECS) ||
!(c->pcpu = alloc_percpu(struct bch_fs_pcpu)) ||
!(c->online_reserved = alloc_percpu(u64)) ||
- !(c->btree_paths_bufs = alloc_percpu(struct btree_path_buf)) ||
mempool_init_kvpmalloc_pool(&c->btree_bounce_pool, 1,
btree_bytes(c)) ||
mempool_init_kmalloc_pool(&c->large_bkey_pool, 1, 2048) ||
}
ret = bch2_fs_counters_init(c) ?:
+ bch2_fs_sb_errors_init(c) ?:
bch2_io_clock_init(&c->io_clock[READ]) ?:
bch2_io_clock_init(&c->io_clock[WRITE]) ?:
bch2_fs_journal_init(&c->journal) ?:
bch2_fs_buckets_waiting_for_journal_init(c) ?:
bch2_fs_btree_write_buffer_init(c) ?:
bch2_fs_subvolumes_init(c) ?:
- bch2_fs_io_init(c) ?:
+ bch2_fs_io_read_init(c) ?:
+ bch2_fs_io_write_init(c) ?:
bch2_fs_nocow_locking_init(c) ?:
bch2_fs_encryption_init(c) ?:
bch2_fs_compress_init(c) ?:
bch2_fs_ec_init(c) ?:
bch2_fs_fsio_init(c) ?:
- bch2_fs_fs_io_buffered_init(c);
+ bch2_fs_fs_io_buffered_init(c) ?:
bch2_fs_fs_io_direct_init(c);
if (ret)
goto err;
- mi = bch2_sb_get_members(c->disk_sb.sb);
for (i = 0; i < c->sb.nr_devices; i++)
- if (bch2_dev_exists(c->disk_sb.sb, mi, i) &&
+ if (bch2_dev_exists(c->disk_sb.sb, i) &&
bch2_dev_alloc(c, i)) {
ret = -EEXIST;
goto err;
int bch2_fs_start(struct bch_fs *c)
{
- struct bch_sb_field_members *mi;
- struct bch_dev *ca;
time64_t now = ktime_get_real_seconds();
- unsigned i;
int ret;
print_mount_opts(c);
down_write(&c->state_lock);
- BUG_ON(test_bit(BCH_FS_STARTED, &c->flags));
+ BUG_ON(test_bit(BCH_FS_started, &c->flags));
mutex_lock(&c->sb_lock);
- for_each_online_member(ca, c, i)
- bch2_sb_from_fs(c, ca);
+ ret = bch2_sb_members_v2_init(c);
+ if (ret) {
+ mutex_unlock(&c->sb_lock);
+ goto err;
+ }
- mi = bch2_sb_get_members(c->disk_sb.sb);
- for_each_online_member(ca, c, i)
- mi->members[ca->dev_idx].last_mount = cpu_to_le64(now);
+ for_each_online_member(c, ca)
+ bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx)->last_mount = cpu_to_le64(now);
mutex_unlock(&c->sb_lock);
- for_each_rw_member(ca, c, i)
+ for_each_rw_member(c, ca)
bch2_dev_allocator_add(c, ca);
bch2_recalc_capacity(c);
- for (i = 0; i < BCH_TRANSACTIONS_NR; i++) {
- mutex_lock(&c->btree_transaction_stats[i].lock);
- bch2_time_stats_init(&c->btree_transaction_stats[i].lock_hold_times);
- mutex_unlock(&c->btree_transaction_stats[i].lock);
- }
-
ret = BCH_SB_INITIALIZED(c->disk_sb.sb)
? bch2_fs_recovery(c)
: bch2_fs_initialize(c);
goto err;
}
- set_bit(BCH_FS_STARTED, &c->flags);
+ set_bit(BCH_FS_started, &c->flags);
if (c->opts.read_only || c->opts.nochanges) {
bch2_fs_read_only(c);
} else {
- ret = !test_bit(BCH_FS_RW, &c->flags)
+ ret = !test_bit(BCH_FS_rw, &c->flags)
? bch2_fs_read_write(c)
: bch2_fs_read_write_late(c);
if (ret)
up_write(&c->state_lock);
return ret;
err:
- bch_err(c, "error starting filesystem: %s", bch2_err_str(ret));
+ bch_err_msg(c, ret, "starting filesystem");
goto out;
}
static int bch2_dev_may_add(struct bch_sb *sb, struct bch_fs *c)
{
- struct bch_sb_field_members *sb_mi;
-
- sb_mi = bch2_sb_get_members(sb);
- if (!sb_mi)
- return -BCH_ERR_member_info_missing;
+ struct bch_member m = bch2_sb_member_get(sb, sb->dev_idx);
if (le16_to_cpu(sb->block_size) != block_sectors(c))
return -BCH_ERR_mismatched_block_size;
- if (le16_to_cpu(sb_mi->members[sb->dev_idx].bucket_size) <
+ if (le16_to_cpu(m.bucket_size) <
BCH_SB_BTREE_NODE_SIZE(c->disk_sb.sb))
return -BCH_ERR_bucket_size_too_small;
{
struct bch_sb *newest =
le64_to_cpu(fs->seq) > le64_to_cpu(sb->seq) ? fs : sb;
- struct bch_sb_field_members *mi = bch2_sb_get_members(newest);
if (!uuid_equal(&fs->uuid, &sb->uuid))
return -BCH_ERR_device_not_a_member_of_filesystem;
- if (!bch2_dev_exists(newest, mi, sb->dev_idx))
+ if (!bch2_dev_exists(newest, sb->dev_idx))
return -BCH_ERR_device_has_been_removed;
if (fs->block_size != sb->block_size)
struct bch_member *member)
{
struct bch_dev *ca;
+ unsigned i;
ca = kzalloc(sizeof(*ca), GFP_KERNEL);
if (!ca)
bch2_time_stats_init(&ca->io_latency[WRITE]);
ca->mi = bch2_mi_to_cpu(member);
+
+ for (i = 0; i < ARRAY_SIZE(member->errors); i++)
+ atomic64_set(&ca->errors[i], le64_to_cpu(member->errors[i]));
+
ca->uuid = member->uuid;
ca->nr_btree_reserve = DIV_ROUND_UP(BTREE_NODE_RESERVE,
static int bch2_dev_alloc(struct bch_fs *c, unsigned dev_idx)
{
- struct bch_member *member =
- bch2_sb_get_members(c->disk_sb.sb)->members + dev_idx;
+ struct bch_member member = bch2_sb_member_get(c->disk_sb.sb, dev_idx);
struct bch_dev *ca = NULL;
int ret = 0;
if (bch2_fs_init_fault("dev_alloc"))
goto err;
- ca = __bch2_dev_alloc(c, member);
+ ca = __bch2_dev_alloc(c, &member);
if (!ca)
goto err;
/* Commit: */
ca->disk_sb = *sb;
- if (sb->mode & FMODE_EXCL)
- ca->disk_sb.bdev->bd_holder = ca;
memset(sb, 0, sizeof(*sb));
ca->dev = ca->disk_sb.bdev->bd_dev;
enum bch_member_state new_state, int flags)
{
struct bch_devs_mask new_online_devs;
- struct bch_dev *ca2;
- int i, nr_rw = 0, required;
+ int nr_rw = 0, required;
lockdep_assert_held(&c->state_lock);
return true;
/* do we have enough devices to write to? */
- for_each_member_device(ca2, c, i)
+ for_each_member_device(c, ca2)
if (ca2 != ca)
nr_rw += ca2->mi.state == BCH_MEMBER_STATE_rw;
static bool bch2_fs_may_start(struct bch_fs *c)
{
- struct bch_sb_field_members *mi;
struct bch_dev *ca;
unsigned i, flags = 0;
if (!c->opts.degraded &&
!c->opts.very_degraded) {
mutex_lock(&c->sb_lock);
- mi = bch2_sb_get_members(c->disk_sb.sb);
for (i = 0; i < c->disk_sb.sb->nr_devices; i++) {
- if (!bch2_dev_exists(c->disk_sb.sb, mi, i))
+ if (!bch2_dev_exists(c->disk_sb.sb, i))
continue;
ca = bch_dev_locked(c, i);
int __bch2_dev_set_state(struct bch_fs *c, struct bch_dev *ca,
enum bch_member_state new_state, int flags)
{
- struct bch_sb_field_members *mi;
+ struct bch_member *m;
int ret = 0;
if (ca->mi.state == new_state)
bch_notice(ca, "%s", bch2_member_states[new_state]);
mutex_lock(&c->sb_lock);
- mi = bch2_sb_get_members(c->disk_sb.sb);
- SET_BCH_MEMBER_STATE(&mi->members[ca->dev_idx], new_state);
+ m = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
+ SET_BCH_MEMBER_STATE(m, new_state);
bch2_write_super(c);
mutex_unlock(&c->sb_lock);
BTREE_TRIGGER_NORUN, NULL) ?:
bch2_btree_delete_range(c, BTREE_ID_bucket_gens, start, end,
BTREE_TRIGGER_NORUN, NULL);
- if (ret)
- bch_err(c, "error removing dev alloc info: %s", bch2_err_str(ret));
-
+ bch_err_msg(c, ret, "removing dev alloc info");
return ret;
}
int bch2_dev_remove(struct bch_fs *c, struct bch_dev *ca, int flags)
{
- struct bch_sb_field_members *mi;
+ struct bch_member *m;
unsigned dev_idx = ca->dev_idx, data;
int ret;
__bch2_dev_read_only(c, ca);
ret = bch2_dev_data_drop(c, ca->dev_idx, flags);
- if (ret) {
- bch_err(ca, "Remove failed: error dropping data: %s", bch2_err_str(ret));
+ bch_err_msg(ca, ret, "dropping data");
+ if (ret)
goto err;
- }
ret = bch2_dev_remove_alloc(c, ca);
- if (ret) {
- bch_err(ca, "Remove failed, error deleting alloc info");
+ bch_err_msg(ca, ret, "deleting alloc info");
+ if (ret)
goto err;
- }
ret = bch2_journal_flush_device_pins(&c->journal, ca->dev_idx);
- if (ret) {
- bch_err(ca, "Remove failed: error flushing journal: %s", bch2_err_str(ret));
+ bch_err_msg(ca, ret, "flushing journal");
+ if (ret)
goto err;
- }
ret = bch2_journal_flush(&c->journal);
- if (ret) {
- bch_err(ca, "Remove failed, journal error");
+ bch_err(ca, "journal error");
+ if (ret)
goto err;
- }
ret = bch2_replicas_gc2(c);
- if (ret) {
- bch_err(ca, "Remove failed: error from replicas gc: %s", bch2_err_str(ret));
+ bch_err_msg(ca, ret, "in replicas_gc2()");
+ if (ret)
goto err;
- }
data = bch2_dev_has_data(c, ca);
if (data) {
* this device must be gone:
*/
mutex_lock(&c->sb_lock);
- mi = bch2_sb_get_members(c->disk_sb.sb);
- memset(&mi->members[dev_idx].uuid, 0, sizeof(mi->members[dev_idx].uuid));
+ m = bch2_members_v2_get_mut(c->disk_sb.sb, dev_idx);
+ memset(&m->uuid, 0, sizeof(m->uuid));
bch2_write_super(c);
struct bch_opts opts = bch2_opts_empty();
struct bch_sb_handle sb;
struct bch_dev *ca = NULL;
- struct bch_sb_field_members *mi;
+ struct bch_sb_field_members_v2 *mi;
struct bch_member dev_mi;
unsigned dev_idx, nr_devices, u64s;
struct printbuf errbuf = PRINTBUF;
int ret;
ret = bch2_read_super(path, &opts, &sb);
- if (ret) {
- bch_err(c, "device add error: error reading super: %s", bch2_err_str(ret));
+ bch_err_msg(c, ret, "reading super");
+ if (ret)
goto err;
- }
- dev_mi = bch2_sb_get_members(sb.sb)->members[sb.sb->dev_idx];
+ dev_mi = bch2_sb_member_get(sb.sb, sb.sb->dev_idx);
if (BCH_MEMBER_GROUP(&dev_mi)) {
- bch2_disk_path_to_text(&label, sb.sb, BCH_MEMBER_GROUP(&dev_mi) - 1);
+ bch2_disk_path_to_text_sb(&label, sb.sb, BCH_MEMBER_GROUP(&dev_mi) - 1);
if (label.allocation_failure) {
ret = -ENOMEM;
goto err;
}
ret = bch2_dev_may_add(sb.sb, c);
- if (ret) {
- bch_err(c, "device add error: %s", bch2_err_str(ret));
+ if (ret)
goto err;
- }
ca = __bch2_dev_alloc(c, &dev_mi);
if (!ca) {
- bch2_free_super(&sb);
ret = -ENOMEM;
goto err;
}
bch2_dev_usage_init(ca);
ret = __bch2_dev_attach_bdev(ca, &sb);
- if (ret) {
- bch2_dev_free(ca);
+ if (ret)
goto err;
- }
ret = bch2_dev_journal_alloc(ca);
- if (ret) {
- bch_err(c, "device add error: journal alloc failed");
+ bch_err_msg(c, ret, "allocating journal");
+ if (ret)
goto err;
- }
down_write(&c->state_lock);
mutex_lock(&c->sb_lock);
ret = bch2_sb_from_fs(c, ca);
- if (ret) {
- bch_err(c, "device add error: new device superblock too small");
+ bch_err_msg(c, ret, "setting up new superblock");
+ if (ret)
goto err_unlock;
- }
-
- mi = bch2_sb_get_members(ca->disk_sb.sb);
-
- if (!bch2_sb_resize_members(&ca->disk_sb,
- le32_to_cpu(mi->field.u64s) +
- sizeof(dev_mi) / sizeof(u64))) {
- bch_err(c, "device add error: new device superblock too small");
- ret = -BCH_ERR_ENOSPC_sb_members;
- goto err_unlock;
- }
if (dynamic_fault("bcachefs:add:no_slot"))
goto no_slot;
- mi = bch2_sb_get_members(c->disk_sb.sb);
for (dev_idx = 0; dev_idx < BCH_SB_MEMBERS_MAX; dev_idx++)
- if (!bch2_dev_exists(c->disk_sb.sb, mi, dev_idx))
+ if (!bch2_dev_exists(c->disk_sb.sb, dev_idx))
goto have_slot;
no_slot:
- bch_err(c, "device add error: already have maximum number of devices");
ret = -BCH_ERR_ENOSPC_sb_members;
+ bch_err_msg(c, ret, "setting up new superblock");
goto err_unlock;
have_slot:
nr_devices = max_t(unsigned, dev_idx + 1, c->sb.nr_devices);
- u64s = (sizeof(struct bch_sb_field_members) +
- sizeof(struct bch_member) * nr_devices) / sizeof(u64);
- mi = bch2_sb_resize_members(&c->disk_sb, u64s);
+ mi = bch2_sb_field_get(c->disk_sb.sb, members_v2);
+ u64s = DIV_ROUND_UP(sizeof(struct bch_sb_field_members_v2) +
+ le16_to_cpu(mi->member_bytes) * nr_devices, sizeof(u64));
+
+ mi = bch2_sb_field_resize(&c->disk_sb, members_v2, u64s);
if (!mi) {
- bch_err(c, "device add error: no room in superblock for member info");
ret = -BCH_ERR_ENOSPC_sb_members;
+ bch_err_msg(c, ret, "setting up new superblock");
goto err_unlock;
}
+ struct bch_member *m = bch2_members_v2_get_mut(c->disk_sb.sb, dev_idx);
/* success: */
- mi->members[dev_idx] = dev_mi;
- mi->members[dev_idx].last_mount = cpu_to_le64(ktime_get_real_seconds());
+ *m = dev_mi;
+ m->last_mount = cpu_to_le64(ktime_get_real_seconds());
c->disk_sb.sb->nr_devices = nr_devices;
ca->disk_sb.sb->dev_idx = dev_idx;
if (BCH_MEMBER_GROUP(&dev_mi)) {
ret = __bch2_dev_group_set(c, ca, label.buf);
- if (ret) {
- bch_err(c, "device add error: error setting label");
+ bch_err_msg(c, ret, "creating new label");
+ if (ret)
goto err_unlock;
- }
}
bch2_write_super(c);
bch2_dev_usage_journal_reserve(c);
ret = bch2_trans_mark_dev_sb(c, ca);
- if (ret) {
- bch_err(c, "device add error: error marking new superblock: %s", bch2_err_str(ret));
+ bch_err_msg(ca, ret, "marking new superblock");
+ if (ret)
goto err_late;
- }
ret = bch2_fs_freespace_init(c);
- if (ret) {
- bch_err(c, "device add error: error initializing free space: %s", bch2_err_str(ret));
+ bch_err_msg(ca, ret, "initializing free space");
+ if (ret)
goto err_late;
- }
ca->new_fs_bucket_idx = 0;
bch2_free_super(&sb);
printbuf_exit(&label);
printbuf_exit(&errbuf);
+ bch_err_fn(c, ret);
return ret;
err_late:
up_write(&c->state_lock);
{
struct bch_opts opts = bch2_opts_empty();
struct bch_sb_handle sb = { NULL };
- struct bch_sb_field_members *mi;
struct bch_dev *ca;
unsigned dev_idx;
int ret;
dev_idx = sb.sb->dev_idx;
ret = bch2_dev_in_fs(c->disk_sb.sb, sb.sb);
- if (ret) {
- bch_err(c, "error bringing %s online: %s", path, bch2_err_str(ret));
+ bch_err_msg(c, ret, "bringing %s online", path);
+ if (ret)
goto err;
- }
ret = bch2_dev_attach_bdev(c, &sb);
if (ret)
ca = bch_dev_locked(c, dev_idx);
ret = bch2_trans_mark_dev_sb(c, ca);
- if (ret) {
- bch_err(c, "error bringing %s online: error from bch2_trans_mark_dev_sb: %s",
- path, bch2_err_str(ret));
+ bch_err_msg(c, ret, "bringing %s online: error from bch2_trans_mark_dev_sb", path);
+ if (ret)
goto err;
- }
if (ca->mi.state == BCH_MEMBER_STATE_rw)
__bch2_dev_read_write(c, ca);
- mutex_lock(&c->sb_lock);
- mi = bch2_sb_get_members(c->disk_sb.sb);
+ if (!ca->mi.freespace_initialized) {
+ ret = bch2_dev_freespace_init(c, ca, 0, ca->mi.nbuckets);
+ bch_err_msg(ca, ret, "initializing free space");
+ if (ret)
+ goto err;
+ }
- mi->members[ca->dev_idx].last_mount =
- cpu_to_le64(ktime_get_real_seconds());
+ if (!ca->journal.nr) {
+ ret = bch2_dev_journal_alloc(ca);
+ bch_err_msg(ca, ret, "allocating journal");
+ if (ret)
+ goto err;
+ }
+ mutex_lock(&c->sb_lock);
+ bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx)->last_mount =
+ cpu_to_le64(ktime_get_real_seconds());
bch2_write_super(c);
mutex_unlock(&c->sb_lock);
- ret = bch2_fs_freespace_init(c);
- if (ret)
- bch_err(c, "device add error: error initializing free space: %s", bch2_err_str(ret));
-
up_write(&c->state_lock);
return 0;
err:
int bch2_dev_resize(struct bch_fs *c, struct bch_dev *ca, u64 nbuckets)
{
- struct bch_member *mi;
+ struct bch_member *m;
+ u64 old_nbuckets;
int ret = 0;
down_write(&c->state_lock);
+ old_nbuckets = ca->mi.nbuckets;
if (nbuckets < ca->mi.nbuckets) {
bch_err(ca, "Cannot shrink yet");
}
ret = bch2_dev_buckets_resize(c, ca, nbuckets);
- if (ret) {
- bch_err(ca, "Resize error: %s", bch2_err_str(ret));
+ bch_err_msg(ca, ret, "resizing buckets");
+ if (ret)
goto err;
- }
ret = bch2_trans_mark_dev_sb(c, ca);
if (ret)
goto err;
mutex_lock(&c->sb_lock);
- mi = &bch2_sb_get_members(c->disk_sb.sb)->members[ca->dev_idx];
- mi->nbuckets = cpu_to_le64(nbuckets);
+ m = bch2_members_v2_get_mut(c->disk_sb.sb, ca->dev_idx);
+ m->nbuckets = cpu_to_le64(nbuckets);
bch2_write_super(c);
mutex_unlock(&c->sb_lock);
+ if (ca->mi.freespace_initialized) {
+ ret = bch2_dev_freespace_init(c, ca, old_nbuckets, nbuckets);
+ if (ret)
+ goto err;
+
+ /*
+ * XXX: this is all wrong transactionally - we'll be able to do
+ * this correctly after the disk space accounting rewrite
+ */
+ ca->usage_base->d[BCH_DATA_free].buckets += nbuckets - old_nbuckets;
+ }
+
bch2_recalc_capacity(c);
err:
up_write(&c->state_lock);
/* return with ref on ca->ref: */
struct bch_dev *bch2_dev_lookup(struct bch_fs *c, const char *name)
{
- struct bch_dev *ca;
- unsigned i;
-
rcu_read_lock();
- for_each_member_device_rcu(ca, c, i, NULL)
- if (!strcmp(name, ca->name))
- goto found;
- ca = ERR_PTR(-BCH_ERR_ENOENT_dev_not_found);
-found:
+ for_each_member_device_rcu(c, ca, NULL)
+ if (!strcmp(name, ca->name)) {
+ rcu_read_unlock();
+ return ca;
+ }
rcu_read_unlock();
-
- return ca;
+ return ERR_PTR(-BCH_ERR_ENOENT_dev_not_found);
}
/* Filesystem open: */
struct bch_fs *bch2_fs_open(char * const *devices, unsigned nr_devices,
struct bch_opts opts)
{
- struct bch_sb_handle *sb = NULL;
+ DARRAY(struct bch_sb_handle) sbs = { 0 };
struct bch_fs *c = NULL;
- struct bch_sb_field_members *mi;
- unsigned i, best_sb = 0;
+ struct bch_sb_handle *best = NULL;
struct printbuf errbuf = PRINTBUF;
int ret = 0;
goto err;
}
- sb = kcalloc(nr_devices, sizeof(*sb), GFP_KERNEL);
- if (!sb) {
- ret = -ENOMEM;
+ ret = darray_make_room(&sbs, nr_devices);
+ if (ret)
goto err;
- }
- for (i = 0; i < nr_devices; i++) {
- ret = bch2_read_super(devices[i], &opts, &sb[i]);
+ for (unsigned i = 0; i < nr_devices; i++) {
+ struct bch_sb_handle sb = { NULL };
+
+ ret = bch2_read_super(devices[i], &opts, &sb);
if (ret)
goto err;
+ BUG_ON(darray_push(&sbs, sb));
}
- for (i = 1; i < nr_devices; i++)
- if (le64_to_cpu(sb[i].sb->seq) >
- le64_to_cpu(sb[best_sb].sb->seq))
- best_sb = i;
-
- mi = bch2_sb_get_members(sb[best_sb].sb);
+ darray_for_each(sbs, sb)
+ if (!best || le64_to_cpu(sb->sb->seq) > le64_to_cpu(best->sb->seq))
+ best = sb;
- i = 0;
- while (i < nr_devices) {
- if (i != best_sb &&
- !bch2_dev_exists(sb[best_sb].sb, mi, sb[i].sb->dev_idx)) {
- pr_info("%pg has been removed, skipping", sb[i].bdev);
- bch2_free_super(&sb[i]);
- array_remove_item(sb, nr_devices, i);
+ darray_for_each_reverse(sbs, sb) {
+ if (sb != best && !bch2_dev_exists(best->sb, sb->sb->dev_idx)) {
+ pr_info("%pg has been removed, skipping", sb->bdev);
+ bch2_free_super(sb);
+ darray_remove_item(&sbs, sb);
+ best -= best > sb;
continue;
}
- ret = bch2_dev_in_fs(sb[best_sb].sb, sb[i].sb);
+ ret = bch2_dev_in_fs(best->sb, sb->sb);
if (ret)
goto err_print;
- i++;
}
- c = bch2_fs_alloc(sb[best_sb].sb, opts);
- if (IS_ERR(c)) {
- ret = PTR_ERR(c);
+ c = bch2_fs_alloc(best->sb, opts);
+ ret = PTR_ERR_OR_ZERO(c);
+ if (ret)
goto err;
- }
down_write(&c->state_lock);
- for (i = 0; i < nr_devices; i++) {
- ret = bch2_dev_attach_bdev(c, &sb[i]);
+ darray_for_each(sbs, sb) {
+ ret = bch2_dev_attach_bdev(c, sb);
if (ret) {
up_write(&c->state_lock);
goto err;
goto err;
}
out:
- kfree(sb);
+ darray_for_each(sbs, sb)
+ bch2_free_super(sb);
+ darray_exit(&sbs);
printbuf_exit(&errbuf);
module_put(THIS_MODULE);
return c;
err:
if (!IS_ERR_OR_NULL(c))
bch2_fs_stop(c);
- if (sb)
- for (i = 0; i < nr_devices; i++)
- bch2_free_super(&sb[i]);
c = ERR_PTR(ret);
goto out;
}
projects / bcachefs-tools-debian / blobdiff