Message ID | 20180830174225.2200-13-josef@toxicpanda.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | My current patch queue | expand |
On 30.08.2018 20:42, Josef Bacik wrote: > + if (flush_state == ALLOC_CHUNK_FORCE && !commit_cycles) > + flush_state++; This is a bit obscure. So if we allocated a chunk and !commit_cycles just break from the loop? What's the reasoning behind this ?
On Mon, Sep 03, 2018 at 05:19:19PM +0300, Nikolay Borisov wrote: > > > On 30.08.2018 20:42, Josef Bacik wrote: > > + if (flush_state == ALLOC_CHUNK_FORCE && !commit_cycles) > > + flush_state++; > > This is a bit obscure. So if we allocated a chunk and !commit_cycles > just break from the loop? What's the reasoning behind this ? I'll add a comment, but it doesn't break the loop, it just goes to COMMIT_TRANS. The idea is we don't want to force a chunk allocation if we're experiencing a little bit of pressure, because we could end up with a drive full of empty metadata chunks. We want to try committing the transaction first, and then if we still have issues we can force a chunk allocation. Thanks, Josef
On 4.09.2018 20:57, Josef Bacik wrote: > On Mon, Sep 03, 2018 at 05:19:19PM +0300, Nikolay Borisov wrote: >> >> >> On 30.08.2018 20:42, Josef Bacik wrote: >>> + if (flush_state == ALLOC_CHUNK_FORCE && !commit_cycles) >>> + flush_state++; >> >> This is a bit obscure. So if we allocated a chunk and !commit_cycles >> just break from the loop? What's the reasoning behind this ? > > I'll add a comment, but it doesn't break the loop, it just goes to COMMIT_TRANS. > The idea is we don't want to force a chunk allocation if we're experiencing a > little bit of pressure, because we could end up with a drive full of empty > metadata chunks. We want to try committing the transaction first, and then if > we still have issues we can force a chunk allocation. Thanks, I think it will be better if this check is moved up somewhere before the the if (flush_state > commit trans). > > Josef >
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h index 0a4e55703d48..791e287c2292 100644 --- a/fs/btrfs/ctree.h +++ b/fs/btrfs/ctree.h @@ -2731,7 +2731,8 @@ enum btrfs_flush_state { FLUSH_DELALLOC = 5, FLUSH_DELALLOC_WAIT = 6, ALLOC_CHUNK = 7, - COMMIT_TRANS = 8, + ALLOC_CHUNK_FORCE = 8, + COMMIT_TRANS = 9, }; int btrfs_alloc_data_chunk_ondemand(struct btrfs_inode *inode, u64 bytes); diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c index 783341e3653e..22e1f9f55f4f 100644 --- a/fs/btrfs/extent-tree.c +++ b/fs/btrfs/extent-tree.c @@ -4907,6 +4907,7 @@ static void flush_space(struct btrfs_fs_info *fs_info, btrfs_end_transaction(trans); break; case ALLOC_CHUNK: + case ALLOC_CHUNK_FORCE: trans = btrfs_join_transaction(root); if (IS_ERR(trans)) { ret = PTR_ERR(trans); @@ -4914,7 +4915,9 @@ static void flush_space(struct btrfs_fs_info *fs_info, } ret = do_chunk_alloc(trans, btrfs_metadata_alloc_profile(fs_info), - CHUNK_ALLOC_NO_FORCE); + (state == ALLOC_CHUNK) ? + CHUNK_ALLOC_NO_FORCE : + CHUNK_ALLOC_FORCE); btrfs_end_transaction(trans); if (ret > 0 || ret == -ENOSPC) ret = 0; @@ -5060,6 +5063,8 @@ static void btrfs_async_reclaim_metadata_space(struct work_struct *work) } } spin_unlock(&space_info->lock); + if (flush_state == ALLOC_CHUNK_FORCE && !commit_cycles) + flush_state++; } while (flush_state <= COMMIT_TRANS); } diff --git a/include/trace/events/btrfs.h b/include/trace/events/btrfs.h index 7d205e50b09c..fdb23181b5b7 100644 --- a/include/trace/events/btrfs.h +++ b/include/trace/events/btrfs.h @@ -1051,6 +1051,7 @@ TRACE_EVENT(btrfs_trigger_flush, { FLUSH_DELAYED_REFS_NR, "FLUSH_DELAYED_REFS_NR"}, \ { FLUSH_DELAYED_REFS, "FLUSH_ELAYED_REFS"}, \ { ALLOC_CHUNK, "ALLOC_CHUNK"}, \ + { ALLOC_CHUNK_FORCE, "ALLOC_CHUNK_FORCE"}, \ { COMMIT_TRANS, "COMMIT_TRANS"}) TRACE_EVENT(btrfs_flush_space,
With my change to no longer take into account the global reserve for metadata allocation chunks we have this side-effect for mixed block group fs'es where we are no longer allocating enough chunks for the data/metadata requirements. To deal with this add a ALLOC_CHUNK_FORCE step to the flushing state machine. This will only get used if we've already made a full loop through the flushing machinery and tried committing the transaction. If we have then we can try and force a chunk allocation since we likely need it to make progress. This resolves the issues I was seeing with the mixed bg tests in xfstests with my previous patch. Signed-off-by: Josef Bacik <josef@toxicpanda.com> --- fs/btrfs/ctree.h | 3 ++- fs/btrfs/extent-tree.c | 7 ++++++- include/trace/events/btrfs.h | 1 + 3 files changed, 9 insertions(+), 2 deletions(-)