#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 "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 "io_read.h"
+#include "io_write.h"
#include "journal.h"
#include "journal_reclaim.h"
#include "journal_seq_blacklist.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"
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Kent Overstreet <kent.overstreet@gmail.com>");
+MODULE_DESCRIPTION("bcachefs filesystem");
#define KTYPE(type) \
static const struct attribute_group type ## _group = { \
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;
{
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");
return ret;
}
- ret = bch2_copygc_start(c);
- if (ret) {
- bch_err(c, "error starting copygc thread");
- return ret;
- }
+ ret = bch2_journal_reclaim_start(&c->journal);
+ if (ret)
+ goto err;
if (!early) {
ret = bch2_fs_read_write_late(c);
#ifndef BCH_WRITE_REF_DEBUG
percpu_ref_reinit(&c->writes);
#else
- for (unsigned i = 0; i < BCH_WRITE_REF_NR; i++) {
+ for (i = 0; i < BCH_WRITE_REF_NR; i++) {
BUG_ON(atomic_long_read(&c->writes[i]));
atomic_long_inc(&c->writes[i]);
}
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_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_io_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);
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->btree_paths_bufs);
+ darray_exit(&c->btree_roots_extra);
free_percpu(c->pcpu);
mempool_exit(&c->large_bkey_pool);
mempool_exit(&c->btree_bounce_pool);
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(-BCH_ERR_ENOMEM_fs_alloc);
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);
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);
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) ||
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_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_printf(&p, "mounted version=%s", bch2_metadata_versions[c->sb.version]);
+ prt_str(&p, "mounting version ");
+ bch2_version_to_text(&p, c->sb.version);
if (c->opts.read_only) {
prt_str(&p, " opts=");
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);
return ret;
err:
- bch_err(c, "error starting filesystem: %s", bch2_err_str(ret));
+ bch_err_msg(c, ret, "starting filesystem");
goto out;
}
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 = -BCH_ERR_ENOMEM_dev_alloc;
- goto out;
+ return -BCH_ERR_ENOMEM_dev_alloc;
}
static int __bch2_dev_attach_bdev(struct bch_dev *ca, struct bch_sb_handle *sb)
/* 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;
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;
}
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");
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");
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");
goto err;
}
ret = bch2_journal_flush(&c->journal);
if (ret) {
- bch_err(ca, "Remove failed, journal error");
+ bch_err(ca, "journal error");
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()");
goto err;
}
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:
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");
goto err;
}
ret = bch2_dev_may_add(sb.sb, c);
if (ret) {
- bch_err(c, "device add error: %s", bch2_err_str(ret));
+ bch_err_fn(c, 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");
goto err;
}
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");
goto err_unlock;
}
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;
+ bch_err_msg(c, ret, "setting up new superblock");
goto err_unlock;
}
if (!bch2_dev_exists(c->disk_sb.sb, mi, 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:
mi = bch2_sb_resize_members(&c->disk_sb, 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;
}
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");
goto err_unlock;
}
}
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(c, ret, "marking new superblock");
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(c, ret, "initializing free space");
goto err_late;
}
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);
goto err;
}
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);
goto err;
}
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(c, ret, "initializing free space");
up_write(&c->state_lock);
return 0;
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");
goto err;
}
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 %s (%i)", bch2_err_str(PTR_ERR_OR_ZERO(c)),
- 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
-unsigned bch2_metadata_version = bcachefs_metadata_version_current;
+__maybe_unused
+static unsigned bch2_metadata_version = bcachefs_metadata_version_current;
module_param_named(version, bch2_metadata_version, uint, 0400);
module_exit(bcachefs_exit);