X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libbcachefs%2Fmovinggc.c;h=fd239a261aca054b22072716e6d2ef7d49791e6a;hb=3bd4653767bb5eadd253e382fe8503ceb6bef5d1;hp=80f922767da94d20998be44f6b54b1e6a4c9e9b7;hpb=b0c9ad15f4e5cee60973a8f5f6dc49acfeec9755;p=bcachefs-tools-debian

diff --git a/libbcachefs/movinggc.c b/libbcachefs/movinggc.c
index 80f9227..fd239a2 100644
--- a/libbcachefs/movinggc.c
+++ b/libbcachefs/movinggc.c
@@ -13,29 +13,64 @@
 #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/bsearch.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>
 
+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)
+{
+	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 bch2_bucket_is_movable(struct btree_trans *trans,
-				  struct bpos bucket, u64 time, u8 *gen)
+				  struct move_bucket *b, u64 time)
 {
 	struct btree_iter iter;
 	struct bkey_s_c k;
@@ -43,185 +78,179 @@ static int bch2_bucket_is_movable(struct btree_trans *trans,
 	const struct bch_alloc_v4 *a;
 	int ret;
 
-	if (bch2_bucket_is_open(trans->c, bucket.inode, bucket.offset))
+	if (bch2_bucket_is_open(trans->c,
+				b->k.bucket.inode,
+				b->k.bucket.offset))
 		return 0;
 
-	bch2_trans_iter_init(trans, &iter, BTREE_ID_alloc, bucket, 0);
-	k = bch2_btree_iter_peek_slot(&iter);
+	k = bch2_bkey_get_iter(trans, &iter, BTREE_ID_alloc,
+			       b->k.bucket, BTREE_ITER_CACHED);
 	ret = bkey_err(k);
-	bch2_trans_iter_exit(trans, &iter);
-
 	if (ret)
 		return ret;
 
 	a = bch2_alloc_to_v4(k, &_a);
-	*gen = a->gen;
-	ret = (a->data_type == BCH_DATA_btree ||
-	       a->data_type == BCH_DATA_user) &&
+	b->k.gen	= a->gen;
+	b->sectors	= bch2_bucket_sectors_dirty(*a);
+
+	ret = data_type_movable(a->data_type) &&
 		a->fragmentation_lru &&
 		a->fragmentation_lru <= time;
 
-	if (ret) {
-		struct printbuf buf = PRINTBUF;
-
-		bch2_bkey_val_to_text(&buf, trans->c, k);
-		pr_debug("%s", buf.buf);
-		printbuf_exit(&buf);
-	}
-
+	bch2_trans_iter_exit(trans, &iter);
 	return ret;
 }
 
-struct copygc_bucket_in_flight {
-	struct bpos		bucket;
-	u8			gen;
-	struct moving_context	ctxt;
-};
+static void move_buckets_wait(struct moving_context *ctxt,
+			      struct buckets_in_flight *list,
+			      bool flush)
+{
+	struct move_bucket_in_flight *i;
+	int ret;
 
-typedef FIFO(struct copygc_bucket_in_flight) copygc_buckets_in_flight;
+	while ((i = list->first)) {
+		if (flush)
+			move_ctxt_wait_event(ctxt, !atomic_read(&i->count));
 
-struct copygc_bucket {
-	struct bpos		bucket;
-	u8			gen;
-};
+		if (atomic_read(&i->count))
+			break;
 
-typedef DARRAY(struct copygc_bucket) copygc_buckets;
+		list->first = i->next;
+		if (!list->first)
+			list->last = NULL;
 
-static int copygc_bucket_cmp(const void *_l, const void *_r)
-{
-	const struct copygc_bucket *l = _l;
-	const struct copygc_bucket *r = _r;
+		list->nr--;
+		list->sectors -= i->bucket.sectors;
 
-	return bpos_cmp(l->bucket, r->bucket) ?: cmp_int(l->gen, r->gen);
-}
+		ret = rhashtable_remove_fast(&list->table, &i->hash,
+					     bch_move_bucket_params);
+		BUG_ON(ret);
+		kfree(i);
+	}
 
-static bool bucket_in_flight(copygc_buckets *buckets_sorted, struct copygc_bucket b)
-{
-	return bsearch(&b,
-		       buckets_sorted->data,
-		       buckets_sorted->nr,
-		       sizeof(buckets_sorted->data[0]),
-		       copygc_bucket_cmp) != NULL;
+	bch2_trans_unlock_long(ctxt->trans);
 }
 
-static void copygc_buckets_wait(struct btree_trans *trans,
-				copygc_buckets_in_flight *buckets_in_flight,
-				size_t nr, bool verify_evacuated)
+static bool bucket_in_flight(struct buckets_in_flight *list,
+			     struct move_bucket_key k)
 {
-	while (!fifo_empty(buckets_in_flight)) {
-		struct copygc_bucket_in_flight *i = &fifo_peek_front(buckets_in_flight);
-
-		if (fifo_used(buckets_in_flight) <= nr &&
-		    closure_nr_remaining(&i->ctxt.cl) != 1)
-			break;
-
-		/*
-		 * moving_ctxt_exit calls bch2_write as it flushes pending
-		 * reads, which inits another btree_trans; this one must be
-		 * unlocked:
-		 */
-		bch2_trans_unlock(trans);
-		bch2_moving_ctxt_exit(&i->ctxt);
-		if (verify_evacuated)
-			bch2_verify_bucket_evacuated(trans, i->bucket, i->gen);
-		buckets_in_flight->front++;
-	}
+	return rhashtable_lookup_fast(&list->table, &k, bch_move_bucket_params);
 }
 
-static int bch2_copygc_get_buckets(struct btree_trans *trans,
-			copygc_buckets_in_flight *buckets_in_flight,
-			copygc_buckets *buckets)
+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;
-	copygc_buckets buckets_sorted = { 0 };
-	struct copygc_bucket_in_flight *i;
 	struct bkey_s_c k;
-	size_t fifo_iter;
+	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;
 
-	copygc_buckets_wait(trans, buckets_in_flight, buckets_in_flight->size / 2, true);
+	move_buckets_wait(ctxt, buckets_in_flight, false);
 
-	fifo_for_each_entry_ptr(i, buckets_in_flight, fifo_iter) {
-		ret = darray_push(&buckets_sorted, ((struct copygc_bucket) {i->bucket, i->gen}));
-		if (ret) {
-			bch_err(trans->c, "error allocating copygc_buckets_sorted");
-			goto err;
-		}
-	}
+	ret = bch2_btree_write_buffer_tryflush(trans);
+	if (bch2_err_matches(ret, EROFS))
+		return ret;
 
-	sort(buckets_sorted.data,
-	     buckets_sorted.nr,
-	     sizeof(buckets_sorted.data[0]),
-	     copygc_bucket_cmp,
-	     NULL);
+	if (bch2_fs_fatal_err_on(ret, c, "%s: error %s from bch2_btree_write_buffer_tryflush()",
+				 __func__, bch2_err_str(ret)))
+		return ret;
 
 	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 copygc_bucket b = { .bucket = u64_to_bucket(k.k->p.offset) };
-		int ret = 0;
+		struct move_bucket b = { .k.bucket = u64_to_bucket(k.k->p.offset) };
+		int ret2 = 0;
 
-		if (!bucket_in_flight(&buckets_sorted, b) &&
-		    bch2_bucket_is_movable(trans, b.bucket, lru_pos_time(k.k->p), &b.gen))
-			ret = darray_push(buckets, b) ?: buckets->nr >= fifo_free(buckets_in_flight);
+		saw++;
 
-		ret;
-	}));
+		ret2 = bch2_bucket_is_movable(trans, &b, lru_pos_time(k.k->p));
+		if (ret2 < 0)
+			goto err;
+
+		if (!ret2)
+			not_movable++;
+		else if (bucket_in_flight(buckets_in_flight, b.k))
+			in_flight++;
+		else {
+			ret2 = darray_push(buckets, b);
+			if (ret2)
+				goto err;
+			sectors += b.sectors;
+		}
+
+		ret2 = buckets->nr >= nr_to_get;
 err:
-	darray_exit(&buckets_sorted);
+		ret2;
+	}));
+
+	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);
 
 	return ret < 0 ? ret : 0;
 }
 
-static int bch2_copygc(struct btree_trans *trans,
-		       copygc_buckets_in_flight *buckets_in_flight,
-		       struct bch_move_stats *stats)
+noinline
+static int bch2_copygc(struct moving_context *ctxt,
+		       struct buckets_in_flight *buckets_in_flight,
+		       bool *did_work)
 {
+	struct btree_trans *trans = ctxt->trans;
 	struct bch_fs *c = trans->c;
 	struct data_update_opts data_opts = {
-		.btree_insert_flags = BTREE_INSERT_USE_RESERVE|JOURNAL_WATERMARK_copygc,
+		.btree_insert_flags = BCH_WATERMARK_copygc,
 	};
-	copygc_buckets buckets = { 0 };
-	struct copygc_bucket_in_flight *f;
-	struct copygc_bucket *i;
+	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_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;
-
-	ret = bch2_copygc_get_buckets(trans, buckets_in_flight, &buckets);
+	ret = bch2_copygc_get_buckets(ctxt, buckets_in_flight, &buckets);
 	if (ret)
 		goto err;
 
 	darray_for_each(buckets, i) {
-		if (unlikely(freezing(current)))
+		if (kthread_should_stop() || freezing(current))
 			break;
 
-		f = fifo_push_ref(buckets_in_flight);
-		f->bucket	= i->bucket;
-		f->gen		= i->gen;
-		bch2_moving_ctxt_init(&f->ctxt, c, NULL, NULL, //stats,
-				      writepoint_ptr(&c->copygc_write_point),
-				      false);
+		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(trans, &f->ctxt, f->bucket, f->gen, data_opts);
+		ret = bch2_evacuate_bucket(ctxt, f, f->bucket.k.bucket,
+					     f->bucket.k.gen, data_opts);
 		if (ret)
 			goto err;
+
+		*did_work = true;
 	}
 err:
 	darray_exit(&buckets);
 
 	/* no entries in LRU btree found, or got to end: */
-	if (ret == -ENOENT)
+	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(&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;
 }
 
@@ -244,13 +273,18 @@ unsigned long bch2_copygc_wait_amount(struct bch_fs *c)
 	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_stripe) *
+		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));
 	}
@@ -265,6 +299,12 @@ void bch2_copygc_wait_to_text(struct printbuf *out, struct bch_fs *c)
 					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);
@@ -273,33 +313,44 @@ 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];
-	copygc_buckets_in_flight copygc_buckets;
+	struct buckets_in_flight *buckets;
 	u64 last, wait;
 	int ret = 0;
 
-	if (!init_fifo(&copygc_buckets, 1 << 14, GFP_KERNEL)) {
-		bch_err(c, "error allocating copygc buckets in flight");
+	buckets = kzalloc(sizeof(struct buckets_in_flight), GFP_KERNEL);
+	if (!buckets)
 		return -ENOMEM;
+	ret = rhashtable_init(&buckets->table, &bch_move_bucket_params);
+	if (ret) {
+		kfree(buckets);
+		bch_err_msg(c, ret, "allocating copygc buckets in flight");
+		return ret;
 	}
 
 	set_freezable();
+
 	bch2_move_stats_init(&move_stats, "copygc");
-	bch2_trans_init(&trans, c, 0, 0);
+	bch2_moving_ctxt_init(&ctxt, c, NULL, &move_stats,
+			      writepoint_ptr(&c->copygc_write_point),
+			      false);
 
 	while (!ret && !kthread_should_stop()) {
-		bch2_trans_unlock(&trans);
+		bool did_work = false;
 
-		try_to_freeze();
+		bch2_trans_unlock_long(ctxt.trans);
 		cond_resched();
 
-		kthread_wait(freezing(current) || c->copy_gc_enabled);
+		if (!c->copy_gc_enabled) {
+			move_buckets_wait(&ctxt, buckets, true);
+			kthread_wait_freezable(c->copy_gc_enabled ||
+					       kthread_should_stop());
+		}
 
 		if (unlikely(freezing(current))) {
-			copygc_buckets_wait(&trans, &copygc_buckets, 0, true);
-			bch2_trans_unlock(&trans);
+			move_buckets_wait(&ctxt, buckets, true);
 			__refrigerator(false);
 			continue;
 		}
@@ -308,8 +359,10 @@ static int bch2_copygc_thread(void *arg)
 		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;
@@ -318,15 +371,29 @@ static int bch2_copygc_thread(void *arg)
 		c->copygc_wait = 0;
 
 		c->copygc_running = true;
-		ret = bch2_copygc(&trans, &copygc_buckets, &move_stats);
+		ret = bch2_copygc(&ctxt, buckets, &did_work);
 		c->copygc_running = false;
 
 		wake_up(&c->copygc_running_wq);
+
+		if (!wait && !did_work) {
+			u64 min_member_capacity = bch2_min_rw_member_capacity(c);
+
+			if (min_member_capacity == U64_MAX)
+				min_member_capacity = 128 * 2048;
+
+			bch2_trans_unlock_long(ctxt.trans);
+			bch2_kthread_io_clock_wait(clock, last + (min_member_capacity >> 6),
+					MAX_SCHEDULE_TIMEOUT);
+		}
 	}
 
-	copygc_buckets_wait(&trans, &copygc_buckets, 0, !ret);
-	free_fifo(&copygc_buckets);
-	bch2_trans_exit(&trans);
+	move_buckets_wait(&ctxt, buckets, true);
+
+	rhashtable_destroy(&buckets->table);
+	kfree(buckets);
+	bch2_moving_ctxt_exit(&ctxt);
+	bch2_move_stats_exit(&move_stats, c);
 
 	return 0;
 }
@@ -357,7 +424,7 @@ int bch2_copygc_start(struct bch_fs *c)
 	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;
 	}