#include "bkey_sort.h"
#include "btree_cache.h"
#include "btree_gc.h"
+#include "btree_journal_iter.h"
#include "btree_key_cache.h"
#include "btree_update_interior.h"
#include "btree_io.h"
+#include "btree_write_buffer.h"
#include "buckets_waiting_for_journal.h"
#include "chardev.h"
#include "checksum.h"
#include "clock.h"
#include "compress.h"
+#include "counters.h"
#include "debug.h"
#include "disk_groups.h"
#include "ec.h"
#include "error.h"
#include "fs.h"
#include "fs-io.h"
+#include "fs-io-buffered.h"
+#include "fs-io-direct.h"
#include "fsck.h"
#include "inode.h"
#include "io.h"
#include "move.h"
#include "migrate.h"
#include "movinggc.h"
+#include "nocow_locking.h"
#include "quota.h"
#include "rebalance.h"
#include "recovery.h"
#include "replicas.h"
+#include "sb-clean.h"
+#include "snapshot.h"
#include "subvolume.h"
#include "super.h"
#include "super-io.h"
#include "sysfs.h"
-#include "counters.h"
+#include "trace.h"
#include <linux/backing-dev.h>
#include <linux/blkdev.h>
#include <linux/idr.h>
#include <linux/module.h>
#include <linux/percpu.h>
-#include <linux/pretty-printers.h>
#include <linux/random.h>
#include <linux/sysfs.h>
#include <crypto/hash.h>
-#include <trace/events/bcachefs.h>
-
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Kent Overstreet <kent.overstreet@gmail.com>");
static LIST_HEAD(bch_fs_list);
static DEFINE_MUTEX(bch_fs_list_lock);
-static DECLARE_WAIT_QUEUE_HEAD(bch_read_only_wait);
+DECLARE_WAIT_QUEUE_HEAD(bch2_read_only_wait);
static void bch2_dev_free(struct bch_dev *);
static int bch2_dev_alloc(struct bch_fs *, unsigned);
return c;
}
-static struct bch_fs *__bch2_uuid_to_fs(uuid_le uuid)
+static struct bch_fs *__bch2_uuid_to_fs(__uuid_t uuid)
{
struct bch_fs *c;
lockdep_assert_held(&bch_fs_list_lock);
list_for_each_entry(c, &bch_fs_list, list)
- if (!memcmp(&c->disk_sb.sb->uuid, &uuid, sizeof(uuid_le)))
+ if (!memcmp(&c->disk_sb.sb->uuid, &uuid, sizeof(uuid)))
return c;
return NULL;
}
-struct bch_fs *bch2_uuid_to_fs(uuid_le uuid)
+struct bch_fs *bch2_uuid_to_fs(__uuid_t uuid)
{
struct bch_fs *c;
unsigned i, clean_passes = 0;
u64 seq = 0;
+ bch2_fs_ec_stop(c);
+ bch2_open_buckets_stop(c, NULL, true);
bch2_rebalance_stop(c);
bch2_copygc_stop(c);
bch2_gc_thread_stop(c);
+ bch2_fs_ec_flush(c);
- bch_verbose(c, "flushing journal and stopping allocators");
+ bch_verbose(c, "flushing journal and stopping allocators, journal seq %llu",
+ journal_cur_seq(&c->journal));
do {
clean_passes++;
}
} while (clean_passes < 2);
- bch_verbose(c, "flushing journal and stopping allocators complete");
+ bch_verbose(c, "flushing journal and stopping allocators complete, journal seq %llu",
+ journal_cur_seq(&c->journal));
if (test_bit(JOURNAL_REPLAY_DONE, &c->journal.flags) &&
!test_bit(BCH_FS_EMERGENCY_RO, &c->flags))
bch2_dev_allocator_remove(c, ca);
}
+#ifndef BCH_WRITE_REF_DEBUG
static void bch2_writes_disabled(struct percpu_ref *writes)
{
struct bch_fs *c = container_of(writes, struct bch_fs, writes);
set_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags);
- wake_up(&bch_read_only_wait);
+ wake_up(&bch2_read_only_wait);
}
+#endif
void bch2_fs_read_only(struct bch_fs *c)
{
* Block new foreground-end write operations from starting - any new
* writes will return -EROFS:
*/
+ set_bit(BCH_FS_GOING_RO, &c->flags);
+#ifndef BCH_WRITE_REF_DEBUG
percpu_ref_kill(&c->writes);
-
- cancel_work_sync(&c->ec_stripe_delete_work);
+#else
+ for (unsigned i = 0; i < BCH_WRITE_REF_NR; i++)
+ bch2_write_ref_put(c, i);
+#endif
/*
* If we're not doing an emergency shutdown, we want to wait on
* we do need to wait on them before returning and signalling
* that going RO is complete:
*/
- wait_event(bch_read_only_wait,
+ wait_event(bch2_read_only_wait,
test_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags) ||
test_bit(BCH_FS_EMERGENCY_RO, &c->flags));
__bch2_fs_read_only(c);
- wait_event(bch_read_only_wait,
+ wait_event(bch2_read_only_wait,
test_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags));
clear_bit(BCH_FS_WRITE_DISABLE_COMPLETE, &c->flags);
+ clear_bit(BCH_FS_GOING_RO, &c->flags);
if (!bch2_journal_error(&c->journal) &&
!test_bit(BCH_FS_ERROR, &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);
+
bch_verbose(c, "marking filesystem clean");
bch2_fs_mark_clean(c);
}
bch2_journal_halt(&c->journal);
bch2_fs_read_only_async(c);
- wake_up(&bch_read_only_wait);
+ wake_up(&bch2_read_only_wait);
return ret;
}
{
int ret;
+ /*
+ * Data move operations can't run until after check_snapshots has
+ * completed, and bch2_snapshot_is_ancestor() is available.
+ *
+ * Ideally we'd start copygc/rebalance earlier instead of waiting for
+ * all of recovery/fsck to complete:
+ */
+ ret = bch2_copygc_start(c);
+ if (ret) {
+ bch_err(c, "error starting copygc thread");
+ return ret;
+ }
+
ret = bch2_rebalance_start(c);
if (ret) {
bch_err(c, "error starting rebalance thread");
if (test_bit(BCH_FS_INITIAL_GC_UNFIXED, &c->flags)) {
bch_err(c, "cannot go rw, unfixed btree errors");
- return -EROFS;
+ return -BCH_ERR_erofs_unfixed_errors;
}
if (test_bit(BCH_FS_RW, &c->flags))
return 0;
+ if (c->opts.norecovery)
+ return -BCH_ERR_erofs_norecovery;
+
/*
* nochanges is used for fsck -n mode - we have to allow going rw
* during recovery for that to work:
*/
- if (c->opts.norecovery ||
- (c->opts.nochanges &&
- (!early || c->opts.read_only)))
- return -EROFS;
+ if (c->opts.nochanges && (!early || c->opts.read_only))
+ return -BCH_ERR_erofs_nochanges;
bch_info(c, "going read-write");
clear_bit(BCH_FS_CLEAN_SHUTDOWN, &c->flags);
+ /*
+ * First journal write must be a flush write: after a clean shutdown we
+ * don't read the journal, so the first journal write may end up
+ * overwriting whatever was there previously, and there must always be
+ * at least one non-flush write in the journal or recovery will fail:
+ */
+ set_bit(JOURNAL_NEED_FLUSH_WRITE, &c->journal.flags);
+
for_each_rw_member(ca, c, i)
bch2_dev_allocator_add(c, ca);
bch2_recalc_capacity(c);
return ret;
}
- ret = bch2_copygc_start(c);
- if (ret) {
- bch_err(c, "error starting copygc thread");
- return ret;
- }
-
- schedule_work(&c->ec_stripe_delete_work);
-
- bch2_do_discards(c);
- bch2_do_invalidates(c);
-
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);
for (i = 0; i < BCH_TIME_STAT_NR; i++)
bch2_time_stats_exit(&c->times[i]);
+ bch2_free_pending_node_rewrites(c);
bch2_fs_counters_exit(c);
bch2_fs_snapshots_exit(c);
bch2_fs_quota_exit(c);
+ bch2_fs_fs_io_direct_exit(c);
+ bch2_fs_fs_io_buffered_exit(c);
bch2_fs_fsio_exit(c);
bch2_fs_ec_exit(c);
bch2_fs_encryption_exit(c);
bch2_fs_compress_exit(c);
bch2_journal_keys_free(&c->journal_keys);
bch2_journal_entries_free(c);
+ 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);
- free_percpu(c->online_reserved);
+ 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);
bioset_exit(&c->btree_bio);
mempool_exit(&c->fill_iter);
+#ifndef BCH_WRITE_REF_DEBUG
percpu_ref_exit(&c->writes);
+#endif
kfree(rcu_dereference_protected(c->disk_groups, 1));
kfree(c->journal_seq_blacklist_table);
kfree(c->unused_inode_hints);
- free_heap(&c->copygc_heap);
+ if (c->write_ref_wq)
+ destroy_workqueue(c->write_ref_wq);
if (c->io_complete_wq)
destroy_workqueue(c->io_complete_wq);
if (c->copygc_wq)
cancel_work_sync(&ca->io_error_work);
cancel_work_sync(&c->read_only_work);
-
- for (i = 0; i < c->sb.nr_devices; i++)
- if (c->devs[i])
- bch2_free_super(&c->devs[i]->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");
unsigned i, iter_size;
int ret = 0;
- pr_verbose_init(opts, "");
-
c = kvpmalloc(sizeof(struct bch_fs), GFP_KERNEL|__GFP_ZERO);
if (!c) {
- c = ERR_PTR(-ENOMEM);
+ c = ERR_PTR(-BCH_ERR_ENOMEM_fs_alloc);
goto out;
}
bch2_fs_copygc_init(c);
bch2_fs_btree_key_cache_init_early(&c->btree_key_cache);
+ bch2_fs_btree_interior_update_init_early(c);
bch2_fs_allocator_background_init(c);
bch2_fs_allocator_foreground_init(c);
bch2_fs_rebalance_init(c);
bch2_fs_quota_init(c);
bch2_fs_ec_init_early(c);
+ bch2_fs_move_init(c);
INIT_LIST_HEAD(&c->list);
INIT_LIST_HEAD(&c->fsck_errors);
mutex_init(&c->fsck_error_lock);
- INIT_LIST_HEAD(&c->ec_stripe_head_list);
- mutex_init(&c->ec_stripe_head_lock);
-
- INIT_LIST_HEAD(&c->ec_stripe_new_list);
- mutex_init(&c->ec_stripe_new_lock);
-
- INIT_LIST_HEAD(&c->data_progress_list);
- mutex_init(&c->data_progress_lock);
-
- spin_lock_init(&c->ec_stripes_heap_lock);
-
seqcount_init(&c->gc_pos_lock);
seqcount_init(&c->usage_lock);
+ sema_init(&c->io_in_flight, 128);
+
+ INIT_LIST_HEAD(&c->vfs_inodes_list);
+ mutex_init(&c->vfs_inodes_lock);
+
c->copy_gc_enabled = 1;
c->rebalance.enabled = 1;
c->promote_whole_extents = true;
strscpy(c->name, name.buf, sizeof(c->name));
printbuf_exit(&name);
- ret = name.allocation_failure ? -ENOMEM : 0;
+ ret = name.allocation_failure ? -BCH_ERR_ENOMEM_fs_name_alloc : 0;
if (ret)
goto err;
/* Compat: */
- if (sb->version <= bcachefs_metadata_version_inode_v2 &&
+ if (le16_to_cpu(sb->version) <= bcachefs_metadata_version_inode_v2 &&
!BCH_SB_JOURNAL_FLUSH_DELAY(sb))
SET_BCH_SB_JOURNAL_FLUSH_DELAY(sb, 1000);
- if (sb->version <= bcachefs_metadata_version_inode_v2 &&
+ if (le16_to_cpu(sb->version) <= bcachefs_metadata_version_inode_v2 &&
!BCH_SB_JOURNAL_RECLAIM_DELAY(sb))
SET_BCH_SB_JOURNAL_RECLAIM_DELAY(sb, 100);
bch2_opts_apply(&c->opts, opts);
- /* key cache currently disabled for inodes, because of snapshots: */
- c->opts.inodes_use_key_cache = 0;
-
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->inode_shard_bits = ilog2(roundup_pow_of_two(num_possible_cpus()));
if (!(c->btree_update_wq = alloc_workqueue("bcachefs",
- WQ_FREEZABLE|WQ_MEM_RECLAIM|WQ_CPU_INTENSIVE, 1)) ||
+ WQ_FREEZABLE|WQ_UNBOUND|WQ_MEM_RECLAIM, 512)) ||
!(c->btree_io_complete_wq = alloc_workqueue("bcachefs_btree_io",
- WQ_FREEZABLE|WQ_MEM_RECLAIM|WQ_CPU_INTENSIVE, 1)) ||
+ WQ_FREEZABLE|WQ_MEM_RECLAIM, 1)) ||
!(c->copygc_wq = alloc_workqueue("bcachefs_copygc",
WQ_FREEZABLE|WQ_MEM_RECLAIM|WQ_CPU_INTENSIVE, 1)) ||
!(c->io_complete_wq = alloc_workqueue("bcachefs_io",
WQ_FREEZABLE|WQ_HIGHPRI|WQ_MEM_RECLAIM, 1)) ||
+ !(c->write_ref_wq = alloc_workqueue("bcachefs_write_ref",
+ WQ_FREEZABLE, 0)) ||
+#ifndef BCH_WRITE_REF_DEBUG
percpu_ref_init(&c->writes, bch2_writes_disabled,
PERCPU_REF_INIT_DEAD, GFP_KERNEL) ||
+#endif
mempool_init_kmalloc_pool(&c->fill_iter, 1, iter_size) ||
bioset_init(&c->btree_bio, 1,
max(offsetof(struct btree_read_bio, bio),
offsetof(struct btree_write_bio, wbio.bio)),
BIOSET_NEED_BVECS) ||
!(c->pcpu = alloc_percpu(struct bch_fs_pcpu)) ||
- !(c->btree_paths_bufs = alloc_percpu(struct btree_path_buf)) ||
!(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) ||
!(c->unused_inode_hints = kcalloc(1U << c->inode_shard_bits,
sizeof(u64), GFP_KERNEL))) {
- ret = -ENOMEM;
+ ret = -BCH_ERR_ENOMEM_fs_other_alloc;
goto err;
}
bch2_fs_btree_iter_init(c) ?:
bch2_fs_btree_interior_update_init(c) ?:
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_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_fsio_init(c) ?:
+ bch2_fs_fs_io_buffered_init(c);
+ bch2_fs_fs_io_direct_init(c);
if (ret)
goto err;
if (ret)
goto err;
out:
- pr_verbose_init(opts, "ret %i", PTR_ERR_OR_ZERO(c));
return c;
err:
bch2_fs_free(c);
struct printbuf p = PRINTBUF;
bool first = true;
+ prt_str(&p, "mounting version ");
+ bch2_version_to_text(&p, c->sb.version);
+
if (c->opts.read_only) {
- prt_printf(&p, "ro");
+ prt_str(&p, " opts=");
first = false;
+ prt_printf(&p, "ro");
}
for (i = 0; i < bch2_opts_nr; i++) {
if (v == bch2_opt_get_by_id(&bch2_opts_default, i))
continue;
- if (!first)
- prt_printf(&p, ",");
+ prt_str(&p, first ? " opts=" : ",");
first = false;
bch2_opt_to_text(&p, c, c->disk_sb.sb, opt, v, OPT_SHOW_MOUNT_STYLE);
}
- if (!p.pos)
- prt_printf(&p, "(null)");
-
- bch_info(c, "mounted version=%s opts=%s", bch2_metadata_versions[c->sb.version], p.buf);
+ bch_info(c, "%s", p.buf);
printbuf_exit(&p);
}
unsigned i;
int ret;
+ print_mount_opts(c);
+
down_write(&c->state_lock);
BUG_ON(test_bit(BCH_FS_STARTED, &c->flags));
goto err;
}
- print_mount_opts(c);
ret = 0;
out:
up_write(&c->state_lock);
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_le_cmp(fs->uuid, sb->uuid))
+ 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))
struct bch_dev *ca = NULL;
int ret = 0;
- pr_verbose_init(c->opts, "");
-
if (bch2_fs_init_fault("dev_alloc"))
goto err;
ca->fs = c;
bch2_dev_attach(c, ca, dev_idx);
-out:
- pr_verbose_init(c->opts, "ret %i", ret);
return ret;
err:
if (ca)
bch2_dev_free(ca);
- ret = -ENOMEM;
- goto out;
+ return -BCH_ERR_ENOMEM_dev_alloc;
}
static int __bch2_dev_attach_bdev(struct bch_dev *ca, struct bch_sb_handle *sb)
bch2_btree_delete_range(c, BTREE_ID_backpointers, start, end,
BTREE_TRIGGER_NORUN, NULL) ?:
bch2_btree_delete_range(c, BTREE_ID_alloc, start, end,
+ 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));
bch2_dev_free(ca);
+ /*
+ * At this point the device object has been removed in-core, but the
+ * on-disk journal might still refer to the device index via sb device
+ * usage entries. Recovery fails if it sees usage information for an
+ * invalid device. Flush journal pins to push the back of the journal
+ * past now invalid device index references before we update the
+ * superblock, but after the device object has been removed so any
+ * further journal writes elide usage info for the device.
+ */
+ bch2_journal_flush_all_pins(&c->journal);
+
/*
* Free this device's slot in the bch_member array - all pointers to
* this device must be gone:
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:
for_each_member_device_rcu(ca, c, i, NULL)
if (!strcmp(name, ca->name))
goto found;
- ca = ERR_PTR(-ENOENT);
+ ca = ERR_PTR(-BCH_ERR_ENOENT_dev_not_found);
found:
rcu_read_unlock();
if (!try_module_get(THIS_MODULE))
return ERR_PTR(-ENODEV);
- pr_verbose_init(opts, "");
-
if (!nr_devices) {
ret = -EINVAL;
goto err;
kfree(sb);
printbuf_exit(&errbuf);
module_put(THIS_MODULE);
- pr_verbose_init(opts, "ret %i", PTR_ERR_OR_ZERO(c));
return c;
err_print:
pr_err("bch_fs_open err opening %s: %s",
BCH_DEBUG_PARAMS()
#undef BCH_DEBUG_PARAM
+__maybe_unused
+static unsigned bch2_metadata_version = bcachefs_metadata_version_current;
+module_param_named(version, bch2_metadata_version, uint, 0400);
+
module_exit(bcachefs_exit);
module_init(bcachefs_init);