bcachefs: Ignore cached data when calculating fragmentation

Previously, bucket fragmentation was considered to be bucket size -
total amount of live data, both dirty and cached.

This meant that if a bucket was full but only a small amount of data in
it was dirty - the rest cached, we'd get stuck: copygc wouldn't move the
dirty data out of the bucket and the allocator wouldn't be able to
invalidate and drop the cached data.

This changes fragmentation to exclude cached data, so that copygc will
evacuate these buckets and copygc/the allocator will always be able to
make forward progress.

Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>

diff --git a/fs/bcachefs/alloc_background.c b/fs/bcachefs/alloc_background.c
index 0cf7112..1353e72 100644 (file)
--- a/fs/bcachefs/alloc_background.c
+++ b/fs/bcachefs/alloc_background.c
@@ -544,7 +544,7 @@ static bool bch2_can_invalidate_bucket(struct bch_dev *ca, size_t b,
 static unsigned bucket_sort_key(struct bucket *g, struct bucket_mark m,
                                u64 now, u64 last_seq_ondisk)
 {
-       unsigned used = bucket_sectors_used(m);
+       unsigned used = m.cached_sectors;
 
        if (used) {
                /*

diff --git a/fs/bcachefs/buckets.c b/fs/bcachefs/buckets.c
index fb833d8..acdc95d 100644 (file)
--- a/fs/bcachefs/buckets.c
+++ b/fs/bcachefs/buckets.c
@@ -291,8 +291,8 @@ static inline int is_unavailable_bucket(struct bucket_mark m)
 static inline int bucket_sectors_fragmented(struct bch_dev *ca,
                                            struct bucket_mark m)
 {
-       return bucket_sectors_used(m)
-               ? max(0, (int) ca->mi.bucket_size - (int) bucket_sectors_used(m))
+       return m.dirty_sectors
+               ? max(0, (int) ca->mi.bucket_size - (int) m.dirty_sectors)
                : 0;
 }
 

diff --git a/fs/bcachefs/buckets.h b/fs/bcachefs/buckets.h
index 4b53766..483c8b2 100644 (file)
--- a/fs/bcachefs/buckets.h
+++ b/fs/bcachefs/buckets.h
@@ -149,11 +149,6 @@ static inline u8 ptr_stale(struct bch_dev *ca,
 
 /* bucket gc marks */
 
-static inline unsigned bucket_sectors_used(struct bucket_mark mark)
-{
-       return mark.dirty_sectors + mark.cached_sectors;
-}
-
 static inline bool is_available_bucket(struct bucket_mark mark)
 {
        return !mark.dirty_sectors && !mark.stripe;

diff --git a/fs/bcachefs/movinggc.c b/fs/bcachefs/movinggc.c
index 4791e50..64cb10c 100644 (file)
--- a/fs/bcachefs/movinggc.c
+++ b/fs/bcachefs/movinggc.c
@@ -69,10 +69,14 @@ static enum data_cmd copygc_pred(struct bch_fs *c, void *arg,
                        .dev    = p.ptr.dev,
                        .offset = p.ptr.offset,
                };
+               ssize_t i;
 
-               ssize_t i = eytzinger0_find_le(h->data, h->used,
-                                              sizeof(h->data[0]),
-                                              bucket_offset_cmp, &search);
+               if (p.ptr.cached)
+                       continue;
+
+               i = eytzinger0_find_le(h->data, h->used,
+                                      sizeof(h->data[0]),
+                                      bucket_offset_cmp, &search);
 #if 0
                /* eytzinger search verify code: */
                ssize_t j = -1, k;
@@ -185,8 +189,7 @@ static int bch2_copygc(struct bch_fs *c)
 
                        if (m.owned_by_allocator ||
                            m.data_type != BCH_DATA_user ||
-                           !bucket_sectors_used(m) ||
-                           bucket_sectors_used(m) >= ca->mi.bucket_size)
+                           m.dirty_sectors >= ca->mi.bucket_size)
                                continue;
 
                        WARN_ON(m.stripe && !g->stripe_redundancy);
@@ -195,9 +198,9 @@ static int bch2_copygc(struct bch_fs *c)
                                .dev            = dev_idx,
                                .gen            = m.gen,
                                .replicas       = 1 + g->stripe_redundancy,
-                               .fragmentation  = bucket_sectors_used(m) * (1U << 15)
+                               .fragmentation  = m.dirty_sectors * (1U << 15)
                                        / ca->mi.bucket_size,
-                               .sectors        = bucket_sectors_used(m),
+                               .sectors        = m.dirty_sectors,
                                .offset         = bucket_to_sector(ca, b),
                        };
                        heap_add_or_replace(h, e, -fragmentation_cmp, NULL);
@@ -263,8 +266,8 @@ static int bch2_copygc(struct bch_fs *c)
                        m = READ_ONCE(buckets->b[b].mark);
 
                        if (i->gen == m.gen &&
-                           bucket_sectors_used(m)) {
-                               sectors_not_moved += bucket_sectors_used(m);
+                           m.dirty_sectors) {
+                               sectors_not_moved += m.dirty_sectors;
                                buckets_not_moved++;
                        }
                }