#include "alloc_foreground.h"
#include "btree_iter.h"
#include "btree_update.h"
+#include "btree_write_buffer.h"
#include "buckets.h"
#include "clock.h"
-#include "disk_groups.h"
#include "errcode.h"
#include "error.h"
-#include "extents.h"
-#include "eytzinger.h"
-#include "io.h"
-#include "keylist.h"
+#include "lru.h"
#include "move.h"
#include "movinggc.h"
-#include "super-io.h"
+#include "trace.h"
-#include <trace/events/bcachefs.h>
#include <linux/freezer.h>
#include <linux/kthread.h>
#include <linux/math64.h>
#include <linux/sched/task.h>
-#include <linux/sort.h>
#include <linux/wait.h>
-static inline int fragmentation_cmp(copygc_heap *heap,
- struct copygc_heap_entry l,
- struct copygc_heap_entry r)
+struct buckets_in_flight {
+ struct rhashtable table;
+ struct move_bucket_in_flight *first;
+ struct move_bucket_in_flight *last;
+ size_t nr;
+ size_t sectors;
+};
+
+static const struct rhashtable_params bch_move_bucket_params = {
+ .head_offset = offsetof(struct move_bucket_in_flight, hash),
+ .key_offset = offsetof(struct move_bucket_in_flight, bucket.k),
+ .key_len = sizeof(struct move_bucket_key),
+};
+
+static struct move_bucket_in_flight *
+move_bucket_in_flight_add(struct buckets_in_flight *list, struct move_bucket b)
{
- return cmp_int(l.fragmentation, r.fragmentation);
+ struct move_bucket_in_flight *new = kzalloc(sizeof(*new), GFP_KERNEL);
+ int ret;
+
+ if (!new)
+ return ERR_PTR(-ENOMEM);
+
+ new->bucket = b;
+
+ ret = rhashtable_lookup_insert_fast(&list->table, &new->hash,
+ bch_move_bucket_params);
+ if (ret) {
+ kfree(new);
+ return ERR_PTR(ret);
+ }
+
+ if (!list->first)
+ list->first = new;
+ else
+ list->last->next = new;
+
+ list->last = new;
+ list->nr++;
+ list->sectors += b.sectors;
+ return new;
}
-static int find_buckets_to_copygc(struct bch_fs *c)
+static int bch2_bucket_is_movable(struct btree_trans *trans,
+ struct move_bucket *b, u64 time)
{
- copygc_heap *h = &c->copygc_heap;
- struct btree_trans trans;
struct btree_iter iter;
struct bkey_s_c k;
+ struct bch_alloc_v4 _a;
+ const struct bch_alloc_v4 *a;
int ret;
- bch2_trans_init(&trans, c, 0, 0);
-
- /*
- * Find buckets with lowest sector counts, skipping completely
- * empty buckets, by building a maxheap sorted by sector count,
- * and repeatedly replacing the maximum element until all
- * buckets have been visited.
- */
- h->used = 0;
-
- for_each_btree_key(&trans, iter, BTREE_ID_alloc, POS_MIN,
- BTREE_ITER_PREFETCH, k, ret) {
- struct bch_dev *ca = bch_dev_bkey_exists(c, iter.pos.inode);
- struct copygc_heap_entry e;
- struct bch_alloc_v4 a_convert;
- const struct bch_alloc_v4 *a;
-
- a = bch2_alloc_to_v4(k, &a_convert);
-
- if ((a->data_type != BCH_DATA_btree &&
- a->data_type != BCH_DATA_user) ||
- a->dirty_sectors >= ca->mi.bucket_size ||
- bch2_bucket_is_open(c, iter.pos.inode, iter.pos.offset))
- continue;
+ if (bch2_bucket_is_open(trans->c,
+ b->k.bucket.inode,
+ b->k.bucket.offset))
+ return 0;
- e = (struct copygc_heap_entry) {
- .dev = iter.pos.inode,
- .gen = a->gen,
- .replicas = 1 + a->stripe_redundancy,
- .fragmentation = div_u64((u64) a->dirty_sectors * (1ULL << 31),
- ca->mi.bucket_size),
- .sectors = a->dirty_sectors,
- .bucket = iter.pos.offset,
- };
- heap_add_or_replace(h, e, -fragmentation_cmp, NULL);
+ k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_alloc,
+ b->k.bucket, BTREE_ITER_CACHED);
+ ret = bkey_err(k);
+ if (ret)
+ return ret;
- }
- bch2_trans_iter_exit(&trans, &iter);
+ a = bch2_alloc_to_v4(k, &_a);
+ b->k.gen = a->gen;
+ b->sectors = a->dirty_sectors;
+
+ ret = data_type_movable(a->data_type) &&
+ a->fragmentation_lru &&
+ a->fragmentation_lru <= time;
- bch2_trans_exit(&trans);
+ bch2_trans_iter_exit(trans, &iter);
return ret;
}
-static int bch2_copygc(struct bch_fs *c)
+static void move_buckets_wait(struct moving_context *ctxt,
+ struct buckets_in_flight *list,
+ bool flush)
{
- copygc_heap *h = &c->copygc_heap;
- struct copygc_heap_entry e;
- struct bch_move_stats move_stats;
- struct bch_dev *ca;
- unsigned dev_idx;
- size_t heap_size = 0;
- struct moving_context ctxt;
- struct data_update_opts data_opts = {
- .btree_insert_flags = BTREE_INSERT_USE_RESERVE|JOURNAL_WATERMARK_copygc,
- };
- int ret = 0;
+ struct move_bucket_in_flight *i;
+ int ret;
- bch2_move_stats_init(&move_stats, "copygc");
+ while ((i = list->first)) {
+ if (flush)
+ move_ctxt_wait_event(ctxt, !atomic_read(&i->count));
- for_each_rw_member(ca, c, dev_idx)
- heap_size += ca->mi.nbuckets >> 7;
+ if (atomic_read(&i->count))
+ break;
- if (h->size < heap_size) {
- free_heap(&c->copygc_heap);
- if (!init_heap(&c->copygc_heap, heap_size, GFP_KERNEL)) {
- bch_err(c, "error allocating copygc heap");
- return 0;
- }
+ list->first = i->next;
+ if (!list->first)
+ list->last = NULL;
+
+ list->nr--;
+ list->sectors -= i->bucket.sectors;
+
+ ret = rhashtable_remove_fast(&list->table, &i->hash,
+ bch_move_bucket_params);
+ BUG_ON(ret);
+ kfree(i);
}
- ret = find_buckets_to_copygc(c);
- if (ret) {
- bch2_fs_fatal_error(c, "error walking buckets to copygc!");
+ bch2_trans_unlock(ctxt->trans);
+}
+
+static bool bucket_in_flight(struct buckets_in_flight *list,
+ struct move_bucket_key k)
+{
+ return rhashtable_lookup_fast(&list->table, &k, bch_move_bucket_params);
+}
+
+typedef DARRAY(struct move_bucket) move_buckets;
+
+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;
+ size_t nr_to_get = max_t(size_t, 16U, buckets_in_flight->nr / 4);
+ size_t saw = 0, in_flight = 0, not_movable = 0, sectors = 0;
+ int ret;
+
+ 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()",
+ __func__, bch2_err_str(ret)))
return ret;
- }
- if (!h->used) {
- s64 wait = S64_MAX, dev_wait;
- u64 dev_min_wait_fragmented = 0;
- u64 dev_min_wait_allowed = 0;
- int dev_min_wait = -1;
-
- for_each_rw_member(ca, c, dev_idx) {
- struct bch_dev_usage usage = bch2_dev_usage_read(ca);
- s64 allowed = ((__dev_buckets_available(ca, usage, RESERVE_none) *
- ca->mi.bucket_size) >> 1);
- s64 fragmented = usage.d[BCH_DATA_user].fragmented;
-
- dev_wait = max(0LL, allowed - fragmented);
-
- if (dev_min_wait < 0 || dev_wait < wait) {
- dev_min_wait = dev_idx;
- dev_min_wait_fragmented = fragmented;
- dev_min_wait_allowed = allowed;
- }
+ ret = for_each_btree_key2_upto(trans, iter, BTREE_ID_lru,
+ lru_pos(BCH_LRU_FRAGMENTATION_START, 0, 0),
+ lru_pos(BCH_LRU_FRAGMENTATION_START, U64_MAX, LRU_TIME_MAX),
+ 0, k, ({
+ struct move_bucket b = { .k.bucket = u64_to_bucket(k.k->p.offset) };
+ int ret2 = 0;
+
+ saw++;
+
+ if (!bch2_bucket_is_movable(trans, &b, lru_pos_time(k.k->p)))
+ not_movable++;
+ else if (bucket_in_flight(buckets_in_flight, b.k))
+ in_flight++;
+ else {
+ ret2 = darray_push(buckets, b) ?: buckets->nr >= nr_to_get;
+ if (ret2 >= 0)
+ sectors += b.sectors;
}
+ ret2;
+ }));
- bch_err_ratelimited(c, "copygc requested to run but found no buckets to move! dev %u fragmented %llu allowed %llu",
- dev_min_wait, dev_min_wait_fragmented, dev_min_wait_allowed);
- return 0;
- }
+ pr_debug("have: %zu (%zu) saw %zu in flight %zu not movable %zu got %zu (%zu)/%zu buckets ret %i",
+ buckets_in_flight->nr, buckets_in_flight->sectors,
+ saw, in_flight, not_movable, buckets->nr, sectors, nr_to_get, ret);
- heap_resort(h, fragmentation_cmp, NULL);
+ return ret < 0 ? ret : 0;
+}
- bch2_moving_ctxt_init(&ctxt, c, NULL, &move_stats,
- writepoint_ptr(&c->copygc_write_point),
- false);
+noinline
+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,
+ };
+ move_buckets buckets = { 0 };
+ struct move_bucket_in_flight *f;
+ struct move_bucket *i;
+ u64 moved = atomic64_read(&ctxt->stats->sectors_moved);
+ int ret = 0;
+
+ ret = bch2_copygc_get_buckets(ctxt, buckets_in_flight, &buckets);
+ if (ret)
+ goto err;
+
+ darray_for_each(buckets, i) {
+ if (unlikely(freezing(current)))
+ break;
- /* not correct w.r.t. device removal */
- while (h->used && !ret) {
- BUG_ON(!heap_pop(h, e, -fragmentation_cmp, NULL));
- ret = __bch2_evacuate_bucket(&ctxt, POS(e.dev, e.bucket), e.gen,
- data_opts);
+ f = move_bucket_in_flight_add(buckets_in_flight, *i);
+ ret = PTR_ERR_OR_ZERO(f);
+ if (ret == -EEXIST) { /* rare race: copygc_get_buckets returned same bucket more than once */
+ ret = 0;
+ continue;
+ }
+ if (ret == -ENOMEM) { /* flush IO, continue later */
+ ret = 0;
+ break;
+ }
+
+ ret = __bch2_evacuate_bucket(ctxt, f, f->bucket.k.bucket,
+ f->bucket.k.gen, data_opts);
+ if (ret)
+ goto err;
}
+err:
+ darray_exit(&buckets);
- bch2_moving_ctxt_exit(&ctxt);
+ /* no entries in LRU btree found, or got to end: */
+ if (bch2_err_matches(ret, ENOENT))
+ ret = 0;
if (ret < 0 && !bch2_err_matches(ret, EROFS))
- bch_err(c, "error from bch2_move_data() in copygc: %s", bch2_err_str(ret));
+ bch_err_msg(c, ret, "from bch2_move_data()");
- trace_and_count(c, copygc, c, atomic64_read(&move_stats.sectors_moved), 0, 0, 0);
+ moved = atomic64_read(&ctxt->stats->sectors_moved) - moved;
+ trace_and_count(c, copygc, c, moved, 0, 0, 0);
return ret;
}
struct bch_dev *ca;
unsigned dev_idx;
s64 wait = S64_MAX, fragmented_allowed, fragmented;
+ unsigned i;
for_each_rw_member(ca, c, dev_idx) {
struct bch_dev_usage usage = bch2_dev_usage_read(ca);
- fragmented_allowed = ((__dev_buckets_available(ca, usage, RESERVE_none) *
+ fragmented_allowed = ((__dev_buckets_available(ca, usage, BCH_WATERMARK_stripe) *
ca->mi.bucket_size) >> 1);
- fragmented = usage.d[BCH_DATA_user].fragmented;
+ fragmented = 0;
+
+ for (i = 0; i < BCH_DATA_NR; i++)
+ if (data_type_movable(i))
+ fragmented += usage.d[i].fragmented;
wait = min(wait, max(0LL, fragmented_allowed - fragmented));
}
return wait;
}
+void bch2_copygc_wait_to_text(struct printbuf *out, struct bch_fs *c)
+{
+ prt_printf(out, "Currently waiting for: ");
+ prt_human_readable_u64(out, max(0LL, c->copygc_wait -
+ atomic64_read(&c->io_clock[WRITE].now)) << 9);
+ prt_newline(out);
+
+ prt_printf(out, "Currently waiting since: ");
+ prt_human_readable_u64(out, max(0LL,
+ atomic64_read(&c->io_clock[WRITE].now) -
+ c->copygc_wait_at) << 9);
+ prt_newline(out);
+
+ prt_printf(out, "Currently calculated wait: ");
+ prt_human_readable_u64(out, bch2_copygc_wait_amount(c));
+ prt_newline(out);
+}
+
static int bch2_copygc_thread(void *arg)
{
struct bch_fs *c = arg;
+ struct moving_context ctxt;
+ struct bch_move_stats move_stats;
struct io_clock *clock = &c->io_clock[WRITE];
+ struct buckets_in_flight buckets;
u64 last, wait;
int ret = 0;
+ memset(&buckets, 0, sizeof(buckets));
+
+ ret = rhashtable_init(&buckets.table, &bch_move_bucket_params);
+ if (ret) {
+ bch_err_msg(c, ret, "allocating copygc buckets in flight");
+ return ret;
+ }
+
set_freezable();
+ bch2_move_stats_init(&move_stats, "copygc");
+ bch2_moving_ctxt_init(&ctxt, c, NULL, &move_stats,
+ writepoint_ptr(&c->copygc_write_point),
+ false);
+
while (!ret && !kthread_should_stop()) {
+ bch2_trans_unlock(ctxt.trans);
cond_resched();
- if (kthread_wait_freezable(c->copy_gc_enabled))
- break;
+ if (!c->copy_gc_enabled) {
+ move_buckets_wait(&ctxt, &buckets, true);
+ kthread_wait_freezable(c->copy_gc_enabled);
+ }
+
+ if (unlikely(freezing(current))) {
+ move_buckets_wait(&ctxt, &buckets, true);
+ __refrigerator(false);
+ continue;
+ }
last = atomic64_read(&clock->now);
wait = bch2_copygc_wait_amount(c);
if (wait > clock->max_slop) {
- trace_and_count(c, copygc_wait, c, wait, last + wait);
+ c->copygc_wait_at = last;
c->copygc_wait = last + wait;
+ 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);
continue;
c->copygc_wait = 0;
c->copygc_running = true;
- ret = bch2_copygc(c);
+ ret = bch2_copygc(&ctxt, &buckets);
c->copygc_running = false;
wake_up(&c->copygc_running_wq);
}
+ move_buckets_wait(&ctxt, &buckets, true);
+ rhashtable_destroy(&buckets.table);
+ bch2_moving_ctxt_exit(&ctxt);
+ bch2_move_stats_exit(&move_stats, c);
+
return 0;
}
t = kthread_create(bch2_copygc_thread, c, "bch-copygc/%s", c->name);
ret = PTR_ERR_OR_ZERO(t);
if (ret) {
- bch_err(c, "error creating copygc thread: %s", bch2_err_str(ret));
+ bch_err_msg(c, ret, "creating copygc thread");
return ret;
}
projects / bcachefs-tools-debian / blobdiff