On Fri, Nov 08, 2019 at 02:05:03PM -0500, Josef Bacik wrote:
> On Wed, Oct 23, 2019 at 06:52:57PM -0400, Dennis Zhou wrote:
> > Async discard will use the free space cache as backing knowledge for
> > which extents to discard. This patch plumbs knowledge about which
> > extents need to be discarded into the free space cache from
> > unpin_extent_range().
> >
> > An untrimmed extent can merge with everything as this is a new region.
> > Absorbing trimmed extents is a tradeoff to for greater coalescing which
> > makes life better for find_free_extent(). Additionally, it seems the
> > size of a trim isn't as problematic as the trim io itself.
> >
> > When reading in the free space cache from disk, if sync is set, mark all
> > extents as trimmed. The current code ensures at transaction commit that
> > all free space is trimmed when sync is set, so this reflects that.
> >
> > Signed-off-by: Dennis Zhou <dennis@xxxxxxxxxx>
> > ---
> > fs/btrfs/extent-tree.c | 15 +++++++---
> > fs/btrfs/free-space-cache.c | 60 ++++++++++++++++++++++++++++++++-----
> > fs/btrfs/free-space-cache.h | 17 ++++++++++-
> > fs/btrfs/inode-map.c | 13 ++++----
> > 4 files changed, 87 insertions(+), 18 deletions(-)
> >
> > diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
> > index 77a5904756c5..6a40bba3cb19 100644
> > --- a/fs/btrfs/extent-tree.c
> > +++ b/fs/btrfs/extent-tree.c
> > @@ -2783,6 +2783,7 @@ fetch_cluster_info(struct btrfs_fs_info *fs_info,
> >
> > static int unpin_extent_range(struct btrfs_fs_info *fs_info,
> > u64 start, u64 end,
> > + enum btrfs_trim_state trim_state,
> > const bool return_free_space)
> > {
> > struct btrfs_block_group_cache *cache = NULL;
> > @@ -2816,7 +2817,9 @@ static int unpin_extent_range(struct btrfs_fs_info *fs_info,
> > if (start < cache->last_byte_to_unpin) {
> > len = min(len, cache->last_byte_to_unpin - start);
> > if (return_free_space)
> > - btrfs_add_free_space(cache, start, len);
> > + __btrfs_add_free_space(fs_info,
> > + cache->free_space_ctl,
> > + start, len, trim_state);
> > }
> >
> > start += len;
> > @@ -2894,6 +2897,7 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans)
> >
> > while (!trans->aborted) {
> > struct extent_state *cached_state = NULL;
> > + enum btrfs_trim_state trim_state = BTRFS_TRIM_STATE_UNTRIMMED;
> >
> > mutex_lock(&fs_info->unused_bg_unpin_mutex);
> > ret = find_first_extent_bit(unpin, 0, &start, &end,
> > @@ -2903,12 +2907,14 @@ int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans)
> > break;
> > }
> >
> > - if (btrfs_test_opt(fs_info, DISCARD_SYNC))
> > + if (btrfs_test_opt(fs_info, DISCARD_SYNC)) {
> > ret = btrfs_discard_extent(fs_info, start,
> > end + 1 - start, NULL);
> > + trim_state = BTRFS_TRIM_STATE_TRIMMED;
> > + }
> >
> > clear_extent_dirty(unpin, start, end, &cached_state);
> > - unpin_extent_range(fs_info, start, end, true);
> > + unpin_extent_range(fs_info, start, end, trim_state, true);
> > mutex_unlock(&fs_info->unused_bg_unpin_mutex);
> > free_extent_state(cached_state);
> > cond_resched();
> > @@ -5512,7 +5518,8 @@ u64 btrfs_account_ro_block_groups_free_space(struct btrfs_space_info *sinfo)
> > int btrfs_error_unpin_extent_range(struct btrfs_fs_info *fs_info,
> > u64 start, u64 end)
> > {
> > - return unpin_extent_range(fs_info, start, end, false);
> > + return unpin_extent_range(fs_info, start, end,
> > + BTRFS_TRIM_STATE_UNTRIMMED, false);
> > }
> >
> > /*
> > diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
> > index d54dcd0ab230..d7f0cb961496 100644
> > --- a/fs/btrfs/free-space-cache.c
> > +++ b/fs/btrfs/free-space-cache.c
> > @@ -747,6 +747,14 @@ static int __load_free_space_cache(struct btrfs_root *root, struct inode *inode,
> > goto free_cache;
> > }
> >
> > + /*
> > + * Sync discard ensures that the free space cache is always
> > + * trimmed. So when reading this in, the state should reflect
> > + * that.
> > + */
> > + if (btrfs_test_opt(fs_info, DISCARD_SYNC))
> > + e->trim_state = BTRFS_TRIM_STATE_TRIMMED;
> > +
>
> BTRFS_TRIM_STATE_TRIMMED == 0, so if we don't have DISCARD_SYNC set, we'll still
> end up with TRIMMED. Which means we'll end up with weirdness later because
> unpinned extents will come back as UNTRIMMED in the !DISCARD_SYNC case. I'm
> thinking about spinning rust here where we'll never have discard, so we should
> either always treat the free space as trimmed, or always treat it as untrimmed,
> so we don't end up with different allocator behavior for the undiscardable
> devices. Thanks,
>
Ah, that is a good catch. I think having everything be treated as
untrimmed should be fine. I moved BTRFS_TRIM_STATE_UNTRIMMED to be 0 in
the enum.
Thanks,
Dennis
