bcachefs: moving_context now owns a btree_trans

btree_trans and moving_context are used together, and having the
moving_context owns the transaction object reduces some plumbing.

Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
This commit is contained in:
Kent Overstreet 2023-10-20 13:32:42 -04:00
parent a0bfe3b065
commit 633169035a
4 changed files with 70 additions and 88 deletions

View File

@ -487,7 +487,7 @@ int bch2_data_update_init(struct btree_trans *trans,
if (c->opts.nocow_enabled) {
if (ctxt) {
move_ctxt_wait_event(ctxt, trans,
move_ctxt_wait_event(ctxt,
(locked = bch2_bucket_nocow_trylock(&c->nocow_locks,
PTR_BUCKET_POS(c, &p.ptr), 0)) ||
!atomic_read(&ctxt->read_sectors));

View File

@ -157,13 +157,11 @@ static void move_read_endio(struct bio *bio)
closure_put(&ctxt->cl);
}
void bch2_moving_ctxt_do_pending_writes(struct moving_context *ctxt,
struct btree_trans *trans)
void bch2_moving_ctxt_do_pending_writes(struct moving_context *ctxt)
{
struct moving_io *io;
if (trans)
bch2_trans_unlock(trans);
bch2_trans_unlock(ctxt->trans);
while ((io = bch2_moving_ctxt_next_pending_write(ctxt))) {
list_del(&io->read_list);
@ -171,21 +169,20 @@ void bch2_moving_ctxt_do_pending_writes(struct moving_context *ctxt,
}
}
void bch2_move_ctxt_wait_for_io(struct moving_context *ctxt,
struct btree_trans *trans)
void bch2_move_ctxt_wait_for_io(struct moving_context *ctxt)
{
unsigned sectors_pending = atomic_read(&ctxt->write_sectors);
move_ctxt_wait_event(ctxt, trans,
move_ctxt_wait_event(ctxt,
!atomic_read(&ctxt->write_sectors) ||
atomic_read(&ctxt->write_sectors) != sectors_pending);
}
void bch2_moving_ctxt_exit(struct moving_context *ctxt)
{
struct bch_fs *c = ctxt->c;
struct bch_fs *c = ctxt->trans->c;
move_ctxt_wait_event(ctxt, NULL, list_empty(&ctxt->reads));
move_ctxt_wait_event(ctxt, list_empty(&ctxt->reads));
closure_sync(&ctxt->cl);
EBUG_ON(atomic_read(&ctxt->write_sectors));
@ -203,6 +200,9 @@ void bch2_moving_ctxt_exit(struct moving_context *ctxt)
mutex_lock(&c->moving_context_lock);
list_del(&ctxt->list);
mutex_unlock(&c->moving_context_lock);
bch2_trans_put(ctxt->trans);
memset(ctxt, 0, sizeof(*ctxt));
}
void bch2_moving_ctxt_init(struct moving_context *ctxt,
@ -214,7 +214,7 @@ void bch2_moving_ctxt_init(struct moving_context *ctxt,
{
memset(ctxt, 0, sizeof(*ctxt));
ctxt->c = c;
ctxt->trans = bch2_trans_get(c);
ctxt->fn = (void *) _RET_IP_;
ctxt->rate = rate;
ctxt->stats = stats;
@ -287,14 +287,14 @@ static int bch2_extent_drop_ptrs(struct btree_trans *trans,
bch2_trans_commit(trans, NULL, NULL, BTREE_INSERT_NOFAIL);
}
int bch2_move_extent(struct btree_trans *trans,
struct btree_iter *iter,
struct moving_context *ctxt,
int bch2_move_extent(struct moving_context *ctxt,
struct move_bucket_in_flight *bucket_in_flight,
struct bch_io_opts io_opts,
struct btree_iter *iter,
struct bkey_s_c k,
struct bch_io_opts io_opts,
struct data_update_opts data_opts)
{
struct btree_trans *trans = ctxt->trans;
struct bch_fs *c = trans->c;
struct bkey_ptrs_c ptrs = bch2_bkey_ptrs_c(k);
struct moving_io *io;
@ -499,14 +499,13 @@ int bch2_move_get_io_opts_one(struct btree_trans *trans,
return 0;
}
int bch2_move_ratelimit(struct btree_trans *trans,
struct moving_context *ctxt)
int bch2_move_ratelimit(struct moving_context *ctxt)
{
struct bch_fs *c = trans->c;
struct bch_fs *c = ctxt->trans->c;
u64 delay;
if (ctxt->wait_on_copygc) {
bch2_trans_unlock(trans);
bch2_trans_unlock(ctxt->trans);
wait_event_killable(c->copygc_running_wq,
!c->copygc_running ||
kthread_should_stop());
@ -516,7 +515,7 @@ int bch2_move_ratelimit(struct btree_trans *trans,
delay = ctxt->rate ? bch2_ratelimit_delay(ctxt->rate) : 0;
if (delay) {
bch2_trans_unlock(trans);
bch2_trans_unlock(ctxt->trans);
set_current_state(TASK_INTERRUPTIBLE);
}
@ -529,7 +528,7 @@ int bch2_move_ratelimit(struct btree_trans *trans,
schedule_timeout(delay);
if (unlikely(freezing(current))) {
move_ctxt_wait_event(ctxt, trans, list_empty(&ctxt->reads));
move_ctxt_wait_event(ctxt, list_empty(&ctxt->reads));
try_to_freeze();
}
} while (delay);
@ -538,7 +537,7 @@ int bch2_move_ratelimit(struct btree_trans *trans,
* XXX: these limits really ought to be per device, SSDs and hard drives
* will want different limits
*/
move_ctxt_wait_event(ctxt, trans,
move_ctxt_wait_event(ctxt,
atomic_read(&ctxt->write_sectors) < c->opts.move_bytes_in_flight >> 9 &&
atomic_read(&ctxt->read_sectors) < c->opts.move_bytes_in_flight >> 9 &&
atomic_read(&ctxt->write_ios) < c->opts.move_ios_in_flight &&
@ -547,14 +546,14 @@ int bch2_move_ratelimit(struct btree_trans *trans,
return 0;
}
static int bch2_move_data_btree(struct btree_trans *trans,
struct moving_context *ctxt,
struct bpos start,
struct bpos end,
move_pred_fn pred, void *arg,
enum btree_id btree_id)
static int bch2_move_data_btree(struct moving_context *ctxt,
struct bpos start,
struct bpos end,
move_pred_fn pred, void *arg,
enum btree_id btree_id)
{
struct bch_fs *c = ctxt->c;
struct btree_trans *trans = ctxt->trans;
struct bch_fs *c = trans->c;
struct per_snapshot_io_opts snapshot_io_opts;
struct bch_io_opts *io_opts;
struct bkey_buf sk;
@ -579,7 +578,7 @@ static int bch2_move_data_btree(struct btree_trans *trans,
if (ctxt->rate)
bch2_ratelimit_reset(ctxt->rate);
while (!bch2_move_ratelimit(trans, ctxt)) {
while (!bch2_move_ratelimit(ctxt)) {
bch2_trans_begin(trans);
k = bch2_btree_iter_peek(&iter);
@ -617,15 +616,14 @@ static int bch2_move_data_btree(struct btree_trans *trans,
bch2_bkey_buf_reassemble(&sk, c, k);
k = bkey_i_to_s_c(sk.k);
ret2 = bch2_move_extent(trans, &iter, ctxt, NULL,
*io_opts, k, data_opts);
ret2 = bch2_move_extent(ctxt, NULL, &iter, k, *io_opts, data_opts);
if (ret2) {
if (bch2_err_matches(ret2, BCH_ERR_transaction_restart))
continue;
if (ret2 == -ENOMEM) {
/* memory allocation failure, wait for some IO to finish */
bch2_move_ctxt_wait_for_io(ctxt, trans);
bch2_move_ctxt_wait_for_io(ctxt);
continue;
}
@ -646,13 +644,12 @@ next_nondata:
return ret;
}
int __bch2_move_data(struct btree_trans *trans,
struct moving_context *ctxt,
int __bch2_move_data(struct moving_context *ctxt,
struct bbpos start,
struct bbpos end,
move_pred_fn pred, void *arg)
{
struct bch_fs *c = trans->c;
struct bch_fs *c = ctxt->trans->c;
enum btree_id id;
int ret = 0;
@ -665,7 +662,7 @@ int __bch2_move_data(struct btree_trans *trans,
!bch2_btree_id_root(c, id)->b)
continue;
ret = bch2_move_data_btree(trans, ctxt,
ret = bch2_move_data_btree(ctxt,
id == start.btree ? start.pos : POS_MIN,
id == end.btree ? end.pos : POS_MAX,
pred, arg, id);
@ -686,26 +683,23 @@ int bch2_move_data(struct bch_fs *c,
move_pred_fn pred, void *arg)
{
struct btree_trans *trans;
struct moving_context ctxt;
int ret;
bch2_moving_ctxt_init(&ctxt, c, rate, stats, wp, wait_on_copygc);
trans = bch2_trans_get(c);
ret = __bch2_move_data(trans, &ctxt, start, end, pred, arg);
bch2_trans_put(trans);
ret = __bch2_move_data(&ctxt, start, end, pred, arg);
bch2_moving_ctxt_exit(&ctxt);
return ret;
}
int __bch2_evacuate_bucket(struct btree_trans *trans,
struct moving_context *ctxt,
int __bch2_evacuate_bucket(struct moving_context *ctxt,
struct move_bucket_in_flight *bucket_in_flight,
struct bpos bucket, int gen,
struct data_update_opts _data_opts)
{
struct bch_fs *c = ctxt->c;
struct btree_trans *trans = ctxt->trans;
struct bch_fs *c = trans->c;
struct bch_io_opts io_opts = bch2_opts_to_inode_opts(c->opts);
struct btree_iter iter;
struct bkey_buf sk;
@ -750,7 +744,7 @@ int __bch2_evacuate_bucket(struct btree_trans *trans,
goto err;
}
while (!(ret = bch2_move_ratelimit(trans, ctxt))) {
while (!(ret = bch2_move_ratelimit(ctxt))) {
bch2_trans_begin(trans);
ret = bch2_get_next_backpointer(trans, bucket, gen,
@ -800,16 +794,15 @@ int __bch2_evacuate_bucket(struct btree_trans *trans,
i++;
}
ret = bch2_move_extent(trans, &iter, ctxt,
bucket_in_flight,
io_opts, k, data_opts);
ret = bch2_move_extent(ctxt, bucket_in_flight,
&iter, k, io_opts, data_opts);
bch2_trans_iter_exit(trans, &iter);
if (bch2_err_matches(ret, BCH_ERR_transaction_restart))
continue;
if (ret == -ENOMEM) {
/* memory allocation failure, wait for some IO to finish */
bch2_move_ctxt_wait_for_io(ctxt, trans);
bch2_move_ctxt_wait_for_io(ctxt);
continue;
}
if (ret)
@ -865,14 +858,12 @@ int bch2_evacuate_bucket(struct bch_fs *c,
struct write_point_specifier wp,
bool wait_on_copygc)
{
struct btree_trans *trans = bch2_trans_get(c);
struct moving_context ctxt;
int ret;
bch2_moving_ctxt_init(&ctxt, c, rate, stats, wp, wait_on_copygc);
ret = __bch2_evacuate_bucket(trans, &ctxt, NULL, bucket, gen, data_opts);
ret = __bch2_evacuate_bucket(&ctxt, NULL, bucket, gen, data_opts);
bch2_moving_ctxt_exit(&ctxt);
bch2_trans_put(trans);
return ret;
}

View File

@ -12,7 +12,7 @@
struct bch_read_bio;
struct moving_context {
struct bch_fs *c;
struct btree_trans *trans;
struct list_head list;
void *fn;
@ -38,10 +38,10 @@ struct moving_context {
wait_queue_head_t wait;
};
#define move_ctxt_wait_event(_ctxt, _trans, _cond) \
#define move_ctxt_wait_event(_ctxt, _cond) \
do { \
bool cond_finished = false; \
bch2_moving_ctxt_do_pending_writes(_ctxt, _trans); \
bch2_moving_ctxt_do_pending_writes(_ctxt); \
\
if (_cond) \
break; \
@ -60,11 +60,9 @@ void bch2_moving_ctxt_init(struct moving_context *, struct bch_fs *,
struct bch_ratelimit *, struct bch_move_stats *,
struct write_point_specifier, bool);
struct moving_io *bch2_moving_ctxt_next_pending_write(struct moving_context *);
void bch2_moving_ctxt_do_pending_writes(struct moving_context *,
struct btree_trans *);
void bch2_move_ctxt_wait_for_io(struct moving_context *,
struct btree_trans *);
int bch2_move_ratelimit(struct btree_trans *, struct moving_context *);
void bch2_moving_ctxt_do_pending_writes(struct moving_context *);
void bch2_move_ctxt_wait_for_io(struct moving_context *);
int bch2_move_ratelimit(struct moving_context *);
/* Inodes in different snapshots may have different IO options: */
struct snapshot_io_opts_entry {
@ -95,16 +93,14 @@ int bch2_move_get_io_opts_one(struct btree_trans *, struct bch_io_opts *, struct
int bch2_scan_old_btree_nodes(struct bch_fs *, struct bch_move_stats *);
int bch2_move_extent(struct btree_trans *,
struct btree_iter *,
struct moving_context *,
int bch2_move_extent(struct moving_context *,
struct move_bucket_in_flight *,
struct bch_io_opts,
struct btree_iter *,
struct bkey_s_c,
struct bch_io_opts,
struct data_update_opts);
int __bch2_move_data(struct btree_trans *,
struct moving_context *,
int __bch2_move_data(struct moving_context *,
struct bbpos,
struct bbpos,
move_pred_fn, void *);
@ -117,8 +113,7 @@ int bch2_move_data(struct bch_fs *,
bool,
move_pred_fn, void *);
int __bch2_evacuate_bucket(struct btree_trans *,
struct moving_context *,
int __bch2_evacuate_bucket(struct moving_context *,
struct move_bucket_in_flight *,
struct bpos, int,
struct data_update_opts);

View File

@ -101,8 +101,7 @@ static int bch2_bucket_is_movable(struct btree_trans *trans,
return ret;
}
static void move_buckets_wait(struct btree_trans *trans,
struct moving_context *ctxt,
static void move_buckets_wait(struct moving_context *ctxt,
struct buckets_in_flight *list,
bool flush)
{
@ -111,7 +110,7 @@ static void move_buckets_wait(struct btree_trans *trans,
while ((i = list->first)) {
if (flush)
move_ctxt_wait_event(ctxt, trans, !atomic_read(&i->count));
move_ctxt_wait_event(ctxt, !atomic_read(&i->count));
if (atomic_read(&i->count))
break;
@ -129,7 +128,7 @@ static void move_buckets_wait(struct btree_trans *trans,
kfree(i);
}
bch2_trans_unlock(trans);
bch2_trans_unlock(ctxt->trans);
}
static bool bucket_in_flight(struct buckets_in_flight *list,
@ -140,11 +139,11 @@ static bool bucket_in_flight(struct buckets_in_flight *list,
typedef DARRAY(struct move_bucket) move_buckets;
static int bch2_copygc_get_buckets(struct btree_trans *trans,
struct moving_context *ctxt,
static int bch2_copygc_get_buckets(struct moving_context *ctxt,
struct buckets_in_flight *buckets_in_flight,
move_buckets *buckets)
{
struct btree_trans *trans = ctxt->trans;
struct bch_fs *c = trans->c;
struct btree_iter iter;
struct bkey_s_c k;
@ -152,7 +151,7 @@ static int bch2_copygc_get_buckets(struct btree_trans *trans,
size_t saw = 0, in_flight = 0, not_movable = 0, sectors = 0;
int ret;
move_buckets_wait(trans, ctxt, buckets_in_flight, false);
move_buckets_wait(ctxt, buckets_in_flight, false);
ret = bch2_btree_write_buffer_flush(trans);
if (bch2_fs_fatal_err_on(ret, c, "%s: error %s from bch2_btree_write_buffer_flush()",
@ -188,10 +187,10 @@ static int bch2_copygc_get_buckets(struct btree_trans *trans,
}
noinline
static int bch2_copygc(struct btree_trans *trans,
struct moving_context *ctxt,
static int bch2_copygc(struct moving_context *ctxt,
struct buckets_in_flight *buckets_in_flight)
{
struct btree_trans *trans = ctxt->trans;
struct bch_fs *c = trans->c;
struct data_update_opts data_opts = {
.btree_insert_flags = BCH_WATERMARK_copygc,
@ -202,7 +201,7 @@ static int bch2_copygc(struct btree_trans *trans,
u64 moved = atomic64_read(&ctxt->stats->sectors_moved);
int ret = 0;
ret = bch2_copygc_get_buckets(trans, ctxt, buckets_in_flight, &buckets);
ret = bch2_copygc_get_buckets(ctxt, buckets_in_flight, &buckets);
if (ret)
goto err;
@ -221,7 +220,7 @@ static int bch2_copygc(struct btree_trans *trans,
break;
}
ret = __bch2_evacuate_bucket(trans, ctxt, f, f->bucket.k.bucket,
ret = __bch2_evacuate_bucket(ctxt, f, f->bucket.k.bucket,
f->bucket.k.gen, data_opts);
if (ret)
goto err;
@ -300,7 +299,6 @@ void bch2_copygc_wait_to_text(struct printbuf *out, struct bch_fs *c)
static int bch2_copygc_thread(void *arg)
{
struct bch_fs *c = arg;
struct btree_trans *trans;
struct moving_context ctxt;
struct bch_move_stats move_stats;
struct io_clock *clock = &c->io_clock[WRITE];
@ -317,7 +315,6 @@ static int bch2_copygc_thread(void *arg)
}
set_freezable();
trans = bch2_trans_get(c);
bch2_move_stats_init(&move_stats, "copygc");
bch2_moving_ctxt_init(&ctxt, c, NULL, &move_stats,
@ -325,16 +322,16 @@ static int bch2_copygc_thread(void *arg)
false);
while (!ret && !kthread_should_stop()) {
bch2_trans_unlock(trans);
bch2_trans_unlock(ctxt.trans);
cond_resched();
if (!c->copy_gc_enabled) {
move_buckets_wait(trans, &ctxt, &buckets, true);
move_buckets_wait(&ctxt, &buckets, true);
kthread_wait_freezable(c->copy_gc_enabled);
}
if (unlikely(freezing(current))) {
move_buckets_wait(trans, &ctxt, &buckets, true);
move_buckets_wait(&ctxt, &buckets, true);
__refrigerator(false);
continue;
}
@ -345,7 +342,7 @@ static int bch2_copygc_thread(void *arg)
if (wait > clock->max_slop) {
c->copygc_wait_at = last;
c->copygc_wait = last + wait;
move_buckets_wait(trans, &ctxt, &buckets, true);
move_buckets_wait(&ctxt, &buckets, true);
trace_and_count(c, copygc_wait, c, wait, last + wait);
bch2_kthread_io_clock_wait(clock, last + wait,
MAX_SCHEDULE_TIMEOUT);
@ -355,15 +352,14 @@ static int bch2_copygc_thread(void *arg)
c->copygc_wait = 0;
c->copygc_running = true;
ret = bch2_copygc(trans, &ctxt, &buckets);
ret = bch2_copygc(&ctxt, &buckets);
c->copygc_running = false;
wake_up(&c->copygc_running_wq);
}
move_buckets_wait(trans, &ctxt, &buckets, true);
move_buckets_wait(&ctxt, &buckets, true);
rhashtable_destroy(&buckets.table);
bch2_trans_put(trans);
bch2_moving_ctxt_exit(&ctxt);
return 0;