lru_cache: allow multiple changes per transaction

Allow multiple changes to the active set of elements in lru_cache.
The only current user of lru_cache, drbd, is driving this generalisation.
Signed-off-by: Philipp Reisner <philipp.reisner@linbit.com>
Signed-off-by: Lars Ellenberg <lars.ellenberg@linbit.com>

drivers/block/drbd/drbd_actlog.c

@@ -175,7 +175,6 @@ static struct lc_element *_al_get(struct drbd_conf *mdev, unsigned int enr)
 {
 	struct lc_element *al_ext;
 	struct lc_element *tmp;
-	unsigned long     al_flags = 0;
 	int wake;
 
 	spin_lock_irq(&mdev->al_lock);
@@ -191,18 +190,7 @@ static struct lc_element *_al_get(struct drbd_conf *mdev, unsigned int enr)
 		}
 	}
 	al_ext = lc_get(mdev->act_log, enr);
-	al_flags = mdev->act_log->flags;
 	spin_unlock_irq(&mdev->al_lock);
-
-	/*
-	if (!al_ext) {
-		if (al_flags & LC_STARVING)
-			dev_warn(DEV, "Have to wait for LRU element (AL too small?)\n");
-		if (al_flags & LC_DIRTY)
-			dev_warn(DEV, "Ongoing AL update (AL device too slow?)\n");
-	}
-	*/
-
 	return al_ext;
 }
 
@@ -235,7 +223,7 @@ void drbd_al_begin_io(struct drbd_conf *mdev, sector_t sector)
 		mdev->al_writ_cnt++;
 
 		spin_lock_irq(&mdev->al_lock);
-		lc_changed(mdev->act_log, al_ext);
+		lc_committed(mdev->act_log);
 		spin_unlock_irq(&mdev->al_lock);
 		wake_up(&mdev->al_wait);
 	}
@@ -601,7 +589,7 @@ void drbd_al_shrink(struct drbd_conf *mdev)
 	struct lc_element *al_ext;
 	int i;
 
-	D_ASSERT(test_bit(__LC_DIRTY, &mdev->act_log->flags));
+	D_ASSERT(test_bit(__LC_LOCKED, &mdev->act_log->flags));
 
 	for (i = 0; i < mdev->act_log->nr_elements; i++) {
 		al_ext = lc_element_by_index(mdev->act_log, i);
@@ -708,7 +696,9 @@ static void drbd_try_clear_on_disk_bm(struct drbd_conf *mdev, sector_t sector,
 			}
 			ext->rs_left = rs_left;
 			ext->rs_failed = success ? 0 : count;
-			lc_changed(mdev->resync, &ext->lce);
+			/* we don't keep a persistent log of the resync lru,
+			 * we can commit any change right away. */
+			lc_committed(mdev->resync);
 		}
 		lc_put(mdev->resync, &ext->lce);
 		/* no race, we are within the al_lock! */
@@ -892,7 +882,7 @@ struct bm_extent *_bme_get(struct drbd_conf *mdev, unsigned int enr)
 		if (bm_ext->lce.lc_number != enr) {
 			bm_ext->rs_left = drbd_bm_e_weight(mdev, enr);
 			bm_ext->rs_failed = 0;
-			lc_changed(mdev->resync, &bm_ext->lce);
+			lc_committed(mdev->resync);
 			wakeup = 1;
 		}
 		if (bm_ext->lce.refcnt == 1)
@@ -908,7 +898,7 @@ struct bm_extent *_bme_get(struct drbd_conf *mdev, unsigned int enr)
 		if (rs_flags & LC_STARVING)
 			dev_warn(DEV, "Have to wait for element"
 			     " (resync LRU too small?)\n");
-		BUG_ON(rs_flags & LC_DIRTY);
+		BUG_ON(rs_flags & LC_LOCKED);
 	}
 
 	return bm_ext;
@@ -916,26 +906,12 @@ struct bm_extent *_bme_get(struct drbd_conf *mdev, unsigned int enr)
 
 static int _is_in_al(struct drbd_conf *mdev, unsigned int enr)
 {
-	struct lc_element *al_ext;
-	int rv = 0;
+	int rv;
 
 	spin_lock_irq(&mdev->al_lock);
-	if (unlikely(enr == mdev->act_log->new_number))
-		rv = 1;
-	else {
-		al_ext = lc_find(mdev->act_log, enr);
-		if (al_ext) {
-			if (al_ext->refcnt)
-				rv = 1;
-		}
-	}
+	rv = lc_is_used(mdev->act_log, enr);
 	spin_unlock_irq(&mdev->al_lock);
 
-	/*
-	if (unlikely(rv)) {
-		dev_info(DEV, "Delaying sync read until app's write is done\n");
-	}
-	*/
 	return rv;
 }
 
@@ -1065,13 +1041,13 @@ int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector)
 			if (rs_flags & LC_STARVING)
 				dev_warn(DEV, "Have to wait for element"
 				     " (resync LRU too small?)\n");
-			BUG_ON(rs_flags & LC_DIRTY);
+			BUG_ON(rs_flags & LC_LOCKED);
 			goto try_again;
 		}
 		if (bm_ext->lce.lc_number != enr) {
 			bm_ext->rs_left = drbd_bm_e_weight(mdev, enr);
 			bm_ext->rs_failed = 0;
-			lc_changed(mdev->resync, &bm_ext->lce);
+			lc_committed(mdev->resync);
 			wake_up(&mdev->al_wait);
 			D_ASSERT(test_bit(BME_LOCKED, &bm_ext->flags) == 0);
 		}
@@ -1082,8 +1058,6 @@ int drbd_try_rs_begin_io(struct drbd_conf *mdev, sector_t sector)
 	}
 check_al:
 	for (i = 0; i < AL_EXT_PER_BM_SECT; i++) {
-		if (unlikely(al_enr+i == mdev->act_log->new_number))
-			goto try_again;
 		if (lc_is_used(mdev->act_log, al_enr+i))
 			goto try_again;
 	}

drivers/block/drbd/drbd_nl.c

@@ -760,7 +760,7 @@ static int drbd_check_al_size(struct drbd_conf *mdev)
 
 	in_use = 0;
 	t = mdev->act_log;
-	n = lc_create("act_log", drbd_al_ext_cache,
+	n = lc_create("act_log", drbd_al_ext_cache, 1,
 		mdev->sync_conf.al_extents, sizeof(struct lc_element), 0);
 
 	if (n == NULL) {
@@ -1016,7 +1016,7 @@ static int drbd_nl_disk_conf(struct drbd_conf *mdev, struct drbd_nl_cfg_req *nlp
 	}
 
 	resync_lru = lc_create("resync", drbd_bm_ext_cache,
-			61, sizeof(struct bm_extent),
+			1, 61, sizeof(struct bm_extent),
 			offsetof(struct bm_extent, lce));
 	if (!resync_lru) {
 		retcode = ERR_NOMEM;

include/linux/lru_cache.h

@@ -166,9 +166,11 @@ struct lc_element {
 	/* if we want to track a larger set of objects,
 	 * it needs to become arch independend u64 */
 	unsigned lc_number;
-
 	/* special label when on free list */
 #define LC_FREE (~0U)
+
+	/* for pending changes */
+	unsigned lc_new_number;
 };
 
 struct lru_cache {
@@ -176,6 +178,7 @@ struct lru_cache {
 	struct list_head lru;
 	struct list_head free;
 	struct list_head in_use;
+	struct list_head to_be_changed;
 
 	/* the pre-created kmem cache to allocate the objects from */
 	struct kmem_cache *lc_cache;
@@ -194,18 +197,19 @@ struct lru_cache {
 	 * 8 high bits of .lc_index to be overloaded with flags in the future. */
 #define LC_MAX_ACTIVE	(1<<24)
 
+	/* allow to accumulate a few (index:label) changes,
+	 * but no more than max_pending_changes */
+	unsigned int max_pending_changes;
+	/* number of elements currently on to_be_changed list */
+	unsigned int pending_changes;
+
 	/* statistics */
-	unsigned used; /* number of lelements currently on in_use list */
-	unsigned long hits, misses, starving, dirty, changed;
+	unsigned used; /* number of elements currently on in_use list */
+	unsigned long hits, misses, starving, locked, changed;
 
 	/* see below: flag-bits for lru_cache */
 	unsigned long flags;
 
-	/* when changing the label of an index element */
-	unsigned int  new_number;
-
-	/* for paranoia when changing the label of an index element */
-	struct lc_element *changing_element;
 
 	void  *lc_private;
 	const char *name;
@@ -221,10 +225,15 @@ enum {
 	/* debugging aid, to catch concurrent access early.
 	 * user needs to guarantee exclusive access by proper locking! */
 	__LC_PARANOIA,
-	/* if we need to change the set, but currently there is a changing
-	 * transaction pending, we are "dirty", and must deferr further
-	 * changing requests */
+
+	/* annotate that the set is "dirty", possibly accumulating further
+	 * changes, until a transaction is finally triggered */
 	__LC_DIRTY,
+
+	/* Locked, no further changes allowed.
+	 * Also used to serialize changing transactions. */
+	__LC_LOCKED,
+
 	/* if we need to change the set, but currently there is no free nor
 	 * unused element available, we are "starving", and must not give out
 	 * further references, to guarantee that eventually some refcnt will
@@ -236,9 +245,11 @@ enum {
 };
 #define LC_PARANOIA (1<<__LC_PARANOIA)
 #define LC_DIRTY    (1<<__LC_DIRTY)
+#define LC_LOCKED   (1<<__LC_LOCKED)
 #define LC_STARVING (1<<__LC_STARVING)
 
 extern struct lru_cache *lc_create(const char *name, struct kmem_cache *cache,
+		unsigned max_pending_changes,
 		unsigned e_count, size_t e_size, size_t e_off);
 extern void lc_reset(struct lru_cache *lc);
 extern void lc_destroy(struct lru_cache *lc);
@@ -249,7 +260,7 @@ extern struct lc_element *lc_try_get(struct lru_cache *lc, unsigned int enr);
 extern struct lc_element *lc_find(struct lru_cache *lc, unsigned int enr);
 extern struct lc_element *lc_get(struct lru_cache *lc, unsigned int enr);
 extern unsigned int lc_put(struct lru_cache *lc, struct lc_element *e);
-extern void lc_changed(struct lru_cache *lc, struct lc_element *e);
+extern void lc_committed(struct lru_cache *lc);
 
 struct seq_file;
 extern size_t lc_seq_printf_stats(struct seq_file *seq, struct lru_cache *lc);
@@ -258,31 +269,40 @@ extern void lc_seq_dump_details(struct seq_file *seq, struct lru_cache *lc, char
 				void (*detail) (struct seq_file *, struct lc_element *));
 
 /**
- * lc_try_lock - can be used to stop lc_get() from changing the tracked set
+ * lc_try_lock_for_transaction - can be used to stop lc_get() from changing the tracked set
  * @lc: the lru cache to operate on
  *
- * Note that the reference counts and order on the active and lru lists may
- * still change.  Returns true if we acquired the lock.
+ * Allows (expects) the set to be "dirty".  Note that the reference counts and
+ * order on the active and lru lists may still change.  Used to serialize
+ * changing transactions.  Returns true if we aquired the lock.
  */
-static inline int lc_try_lock(struct lru_cache *lc)
+static inline int lc_try_lock_for_transaction(struct lru_cache *lc)
 {
-	return !test_and_set_bit(__LC_DIRTY, &lc->flags);
+	return !test_and_set_bit(__LC_LOCKED, &lc->flags);
 }
 
+/**
+ * lc_try_lock - variant to stop lc_get() from changing the tracked set
+ * @lc: the lru cache to operate on
+ *
+ * Note that the reference counts and order on the active and lru lists may
+ * still change.  Only works on a "clean" set.  Returns true if we aquired the
+ * lock, which means there are no pending changes, and any further attempt to
+ * change the set will not succeed until the next lc_unlock().
+ */
+extern int lc_try_lock(struct lru_cache *lc);
+
 /**
  * lc_unlock - unlock @lc, allow lc_get() to change the set again
  * @lc: the lru cache to operate on
  */
 static inline void lc_unlock(struct lru_cache *lc)
 {
-	clear_bit_unlock(__LC_DIRTY, &lc->flags);
+	clear_bit(__LC_DIRTY, &lc->flags);
+	clear_bit_unlock(__LC_LOCKED, &lc->flags);
 }
 
-static inline int lc_is_used(struct lru_cache *lc, unsigned int enr)
-{
-	struct lc_element *e = lc_find(lc, enr);
-	return e && e->refcnt;
-}
+extern bool lc_is_used(struct lru_cache *lc, unsigned int enr);
 
 #define lc_entry(ptr, type, member) \
 	container_of(ptr, type, member)