btrfs: handle empty block_group removal for async discard

block_group removal is a little tricky. It can race with the extent
allocator, the cleaner thread, and balancing. The current path is for a
block_group to be added to the unused_bgs list. Then, when the cleaner
thread comes around, it starts a transaction and then proceeds with
removing the block_group. Extents that are pinned are subsequently
removed from the pinned trees and then eventually a discard is issued
for the entire block_group.

Async discard introduces another player into the game, the discard
workqueue. While it has none of the racing issues, the new problem is
ensuring we don't leave free space untrimmed prior to forgetting the
block_group.  This is handled by placing fully free block_groups on a
separate discard queue. This is necessary to maintain discarding order
as in the future we will slowly trim even fully free block_groups. The
ordering helps us make progress on the same block_group rather than say
the last fully freed block_group or needing to search through the fully
freed block groups at the beginning of a list and insert after.

The new order of events is a fully freed block group gets placed on the
unused discard queue first. Once it's processed, it will be placed on
the unusued_bgs list and then the original sequence of events will
happen, just without the final whole block_group discard.

The mount flags can change when processing unused_bgs, so when flipping
from DISCARD to DISCARD_ASYNC, the unused_bgs must be punted to the
discard_list to be trimmed. If we flip off DISCARD_ASYNC, we punt
free block groups on the discard_list to the unused_bg queue which will
do the final discard for us.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

1 files changed, 112 insertions, 3 deletions

diff --git a/fs/btrfs/discard.c b/fs/btrfs/discard.c
index 5924e757471b..0f1c6d01aab0 100644
--- a/fs/btrfs/discard.c
+++ b/fs/btrfs/discard.c
@@ -13,6 +13,7 @@
 
 /* This is an initial delay to give some chance for block reuse */
 #define BTRFS_DISCARD_DELAY		(120ULL * NSEC_PER_SEC)
+#define BTRFS_DISCARD_UNUSED_DELAY	(10ULL * NSEC_PER_SEC)
 
 static struct list_head *get_discard_list(struct btrfs_discard_ctl *discard_ctl,
 					  struct btrfs_block_group *block_group)
@@ -30,9 +31,13 @@ static void add_to_discard_list(struct btrfs_discard_ctl *discard_ctl,
 		return;
 	}
 
-	if (list_empty(&block_group->discard_list))
+	if (list_empty(&block_group->discard_list) ||
+	    block_group->discard_index == BTRFS_DISCARD_INDEX_UNUSED) {
+		if (block_group->discard_index == BTRFS_DISCARD_INDEX_UNUSED)
+			block_group->discard_index = BTRFS_DISCARD_INDEX_START;
 		block_group->discard_eligible_time = (ktime_get_ns() +
 						      BTRFS_DISCARD_DELAY);
+	}
 
 	list_move_tail(&block_group->discard_list,
 		       get_discard_list(discard_ctl, block_group));
@@ -40,6 +45,27 @@ static void add_to_discard_list(struct btrfs_discard_ctl *discard_ctl,
 	spin_unlock(&discard_ctl->lock);
 }
 
+static void add_to_discard_unused_list(struct btrfs_discard_ctl *discard_ctl,
+				       struct btrfs_block_group *block_group)
+{
+	spin_lock(&discard_ctl->lock);
+
+	if (!btrfs_run_discard_work(discard_ctl)) {
+		spin_unlock(&discard_ctl->lock);
+		return;
+	}
+
+	list_del_init(&block_group->discard_list);
+
+	block_group->discard_index = BTRFS_DISCARD_INDEX_UNUSED;
+	block_group->discard_eligible_time = (ktime_get_ns() +
+					      BTRFS_DISCARD_UNUSED_DELAY);
+	list_add_tail(&block_group->discard_list,
+		      &discard_ctl->discard_list[BTRFS_DISCARD_INDEX_UNUSED]);
+
+	spin_unlock(&discard_ctl->lock);
+}
+
 static bool remove_from_discard_list(struct btrfs_discard_ctl *discard_ctl,
 				     struct btrfs_block_group *block_group)
 {
@@ -154,7 +180,10 @@ void btrfs_discard_queue_work(struct btrfs_discard_ctl *discard_ctl,
 	if (!block_group || !btrfs_test_opt(block_group->fs_info, DISCARD_ASYNC))
 		return;
 
-	add_to_discard_list(discard_ctl, block_group);
+	if (block_group->used == 0)
+		add_to_discard_unused_list(discard_ctl, block_group);
+	else
+		add_to_discard_list(discard_ctl, block_group);
 
 	if (!delayed_work_pending(&discard_ctl->work))
 		btrfs_discard_schedule_work(discard_ctl, false);
@@ -199,6 +228,29 @@ out:
 }
 
 /**
+ * btrfs_finish_discard_pass - determine next step of a block_group
+ * @discard_ctl: discard control
+ * @block_group: block_group of interest
+ *
+ * This determines the next step for a block group after it's finished going
+ * through a pass on a discard list.  If it is unused and fully trimmed, we can
+ * mark it unused and send it to the unused_bgs path.  Otherwise, pass it onto
+ * the appropriate filter list or let it fall off.
+ */
+static void btrfs_finish_discard_pass(struct btrfs_discard_ctl *discard_ctl,
+				      struct btrfs_block_group *block_group)
+{
+	remove_from_discard_list(discard_ctl, block_group);
+
+	if (block_group->used == 0) {
+		if (btrfs_is_free_space_trimmed(block_group))
+			btrfs_mark_bg_unused(block_group);
+		else
+			add_to_discard_unused_list(discard_ctl, block_group);
+	}
+}
+
+/**
  * btrfs_discard_workfn - discard work function
  * @work: work
  *
@@ -219,7 +271,7 @@ static void btrfs_discard_workfn(struct work_struct *work)
 	btrfs_trim_block_group(block_group, &trimmed, block_group->start,
 			       btrfs_block_group_end(block_group), 0);
 
-	remove_from_discard_list(discard_ctl, block_group);
+	btrfs_finish_discard_pass(discard_ctl, block_group);
 	btrfs_discard_schedule_work(discard_ctl, false);
 }
 
@@ -239,6 +291,60 @@ bool btrfs_run_discard_work(struct btrfs_discard_ctl *discard_ctl)
 		test_bit(BTRFS_FS_DISCARD_RUNNING, &fs_info->flags));
 }
 
+/**
+ * btrfs_discard_punt_unused_bgs_list - punt unused_bgs list to discard lists
+ * @fs_info: fs_info of interest
+ *
+ * The unused_bgs list needs to be punted to the discard lists because the
+ * order of operations is changed.  In the normal sychronous discard path, the
+ * block groups are trimmed via a single large trim in transaction commit.  This
+ * is ultimately what we are trying to avoid with asynchronous discard.  Thus,
+ * it must be done before going down the unused_bgs path.
+ */
+void btrfs_discard_punt_unused_bgs_list(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_block_group *block_group, *next;
+
+	spin_lock(&fs_info->unused_bgs_lock);
+	/* We enabled async discard, so punt all to the queue */
+	list_for_each_entry_safe(block_group, next, &fs_info->unused_bgs,
+				 bg_list) {
+		list_del_init(&block_group->bg_list);
+		btrfs_discard_queue_work(&fs_info->discard_ctl, block_group);
+	}
+	spin_unlock(&fs_info->unused_bgs_lock);
+}
+
+/**
+ * btrfs_discard_purge_list - purge discard lists
+ * @discard_ctl: discard control
+ *
+ * If we are disabling async discard, we may have intercepted block groups that
+ * are completely free and ready for the unused_bgs path.  As discarding will
+ * now happen in transaction commit or not at all, we can safely mark the
+ * corresponding block groups as unused and they will be sent on their merry
+ * way to the unused_bgs list.
+ */
+static void btrfs_discard_purge_list(struct btrfs_discard_ctl *discard_ctl)
+{
+	struct btrfs_block_group *block_group, *next;
+	int i;
+
+	spin_lock(&discard_ctl->lock);
+	for (i = 0; i < BTRFS_NR_DISCARD_LISTS; i++) {
+		list_for_each_entry_safe(block_group, next,
+					 &discard_ctl->discard_list[i],
+					 discard_list) {
+			list_del_init(&block_group->discard_list);
+			spin_unlock(&discard_ctl->lock);
+			if (block_group->used == 0)
+				btrfs_mark_bg_unused(block_group);
+			spin_lock(&discard_ctl->lock);
+		}
+	}
+	spin_unlock(&discard_ctl->lock);
+}
+
 void btrfs_discard_resume(struct btrfs_fs_info *fs_info)
 {
 	if (!btrfs_test_opt(fs_info, DISCARD_ASYNC)) {
@@ -246,6 +352,8 @@ void btrfs_discard_resume(struct btrfs_fs_info *fs_info)
 		return;
 	}
 
+	btrfs_discard_punt_unused_bgs_list(fs_info);
+
 	set_bit(BTRFS_FS_DISCARD_RUNNING, &fs_info->flags);
 }
 
@@ -270,4 +378,5 @@ void btrfs_discard_cleanup(struct btrfs_fs_info *fs_info)
 {
 	btrfs_discard_stop(fs_info);
 	cancel_delayed_work_sync(&fs_info->discard_ctl.work);
+	btrfs_discard_purge_list(&fs_info->discard_ctl);
 }