diff mbox series

[12/23] btrfs: add the data transaction commit logic into may_commit_transaction

Message ID 20200131223613.490779-13-josef@toxicpanda.com (mailing list archive)
State New, archived
Headers show
Series Convert data reservations to the ticketing infrastructure | expand

Commit Message

Josef Bacik Jan. 31, 2020, 10:36 p.m. UTC 
Data space flushing currently unconditionally commits the transaction
twice in a row, and the last time it checks if there's enough pinned
extents to satisfy it's reservation before deciding to commit the
transaction for the 3rd and final time.

Encode this logic into may_commit_transaction().  In the next patch we
will pass in U64_MAX for bytes_needed the first two times, and the final
time we will pass in the actual bytes we need so the normal logic will
apply.

This patch exists soley to make the logical changes I will make to the
flushing state machine separate to make it easier to bisect any
performance related regressions.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
---
 fs/btrfs/space-info.c | 24 +++++++++++++++++++-----
 1 file changed, 19 insertions(+), 5 deletions(-)

Comments

Nikolay Borisov Feb. 3, 2020, 2 p.m. UTC | #1 
On 1.02.20 г. 0:36 ч., Josef Bacik wrote:
> Data space flushing currently unconditionally commits the transaction
> twice in a row, and the last time it checks if there's enough pinned
> extents to satisfy it's reservation before deciding to commit the
> transaction for the 3rd and final time.
> 
> Encode this logic into may_commit_transaction().  In the next patch we

This is incorrect since the next patch simply introduces some states. So
this patch and the next one should be transposed. I guess even David can
do this but he needs to be explicitly aware of this.

> will pass in U64_MAX for bytes_needed the first two times, and the final
> time we will pass in the actual bytes we need so the normal logic will
> apply.
> 
> This patch exists soley to make the logical changes I will make to the
> flushing state machine separate to make it easier to bisect any
> performance related regressions.
> 
> Signed-off-by: Josef Bacik <josef@toxicpanda.com>

Other than the nit in the changelog this LGTM:

Reviewed-by: Nikolay Borisov <nborisov@suse.com>


<snip>
diff mbox series

Patch

diff --git a/fs/btrfs/space-info.c b/fs/btrfs/space-info.c
index 955f59f4b1d0..9f8ef2a09ad9 100644
--- a/fs/btrfs/space-info.c
+++ b/fs/btrfs/space-info.c
@@ -412,20 +412,32 @@  static void shrink_delalloc(struct btrfs_fs_info *fs_info,
  * will return -ENOSPC.
  */
 static int may_commit_transaction(struct btrfs_fs_info *fs_info,
-				  struct btrfs_space_info *space_info)
+				  struct btrfs_space_info *space_info,
+				  u64 bytes_needed)
 {
 	struct reserve_ticket *ticket = NULL;
 	struct btrfs_block_rsv *delayed_rsv = &fs_info->delayed_block_rsv;
 	struct btrfs_block_rsv *delayed_refs_rsv = &fs_info->delayed_refs_rsv;
 	struct btrfs_trans_handle *trans;
-	u64 bytes_needed;
 	u64 reclaim_bytes = 0;
 	u64 cur_free_bytes = 0;
+	bool do_commit = false;
 
 	trans = (struct btrfs_trans_handle *)current->journal_info;
 	if (trans)
 		return -EAGAIN;
 
+	/*
+	 * If we are data and have passed in U64_MAX we just want to
+	 * unconditionally commit the transaction to match the previous data
+	 * flushing behavior.
+	 */
+	if ((space_info->flags & BTRFS_BLOCK_GROUP_DATA) &&
+	   bytes_needed == U64_MAX) {
+		do_commit = true;
+		goto check_pinned;
+	}
+
 	spin_lock(&space_info->lock);
 	cur_free_bytes = btrfs_space_info_used(space_info, true);
 	if (cur_free_bytes < space_info->total_bytes)
@@ -439,7 +451,7 @@  static int may_commit_transaction(struct btrfs_fs_info *fs_info,
 	else if (!list_empty(&space_info->tickets))
 		ticket = list_first_entry(&space_info->tickets,
 					  struct reserve_ticket, list);
-	bytes_needed = (ticket) ? ticket->bytes : 0;
+	bytes_needed = (ticket) ? ticket->bytes : bytes_needed;
 
 	if (bytes_needed > cur_free_bytes)
 		bytes_needed -= cur_free_bytes;
@@ -450,6 +462,7 @@  static int may_commit_transaction(struct btrfs_fs_info *fs_info,
 	if (!bytes_needed)
 		return 0;
 
+check_pinned:
 	trans = btrfs_join_transaction(fs_info->extent_root);
 	if (IS_ERR(trans))
 		return PTR_ERR(trans);
@@ -459,7 +472,8 @@  static int may_commit_transaction(struct btrfs_fs_info *fs_info,
 	 * we have block groups that are going to be freed, allowing us to
 	 * possibly do a chunk allocation the next loop through.
 	 */
-	if (test_bit(BTRFS_TRANS_HAVE_FREE_BGS, &trans->transaction->flags) ||
+	if (do_commit ||
+	    test_bit(BTRFS_TRANS_HAVE_FREE_BGS, &trans->transaction->flags) ||
 	    __percpu_counter_compare(&space_info->total_bytes_pinned,
 				     bytes_needed,
 				     BTRFS_TOTAL_BYTES_PINNED_BATCH) >= 0)
@@ -570,7 +584,7 @@  static void flush_space(struct btrfs_fs_info *fs_info,
 		btrfs_wait_on_delayed_iputs(fs_info);
 		break;
 	case COMMIT_TRANS:
-		ret = may_commit_transaction(fs_info, space_info);
+		ret = may_commit_transaction(fs_info, space_info, num_bytes);
 		break;
 	default:
 		ret = -ENOSPC;