Introduce a tunable to exclude special class buffers from L2ARC

Special allocation class or dedup vdevs may have roughly the same
performance as L2ARC vdevs. Introduce a new tunable to exclude those
buffers from being cacheable on L2ARC.

Reviewed-by: Don Brady <don.brady@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: George Amanakis <gamanakis@gmail.com>
Closes #11761 
Closes #12285 

8 files changed, 119 insertions, 21 deletions

diff --git a/include/sys/arc.h b/include/sys/arc.h
index afbe65bb1..9041939e5 100644
--- a/include/sys/arc.h
+++ b/include/sys/arc.h
@@ -85,6 +85,7 @@ typedef void arc_prune_func_t(int64_t bytes, void *priv);
 
 /* Shared module parameters */
 extern int zfs_arc_average_blocksize;
+extern int l2arc_exclude_special;
 
 /* generic arc_done_func_t's which you can use */
 arc_read_done_func_t arc_bcopy_func;
diff --git a/include/sys/dbuf.h b/include/sys/dbuf.h
index 89422659d..9017cf345 100644
--- a/include/sys/dbuf.h
+++ b/include/sys/dbuf.h
@@ -442,16 +442,7 @@ dbuf_find_dirty_eq(dmu_buf_impl_t *db, uint64_t txg)
 	(dbuf_is_metadata(_db) &&					\
 	((_db)->db_objset->os_primary_cache == ZFS_CACHE_METADATA)))
 
-#define	DBUF_IS_L2CACHEABLE(_db)					\
-	((_db)->db_objset->os_secondary_cache == ZFS_CACHE_ALL ||	\
-	(dbuf_is_metadata(_db) &&					\
-	((_db)->db_objset->os_secondary_cache == ZFS_CACHE_METADATA)))
-
-#define	DNODE_LEVEL_IS_L2CACHEABLE(_dn, _level)				\
-	((_dn)->dn_objset->os_secondary_cache == ZFS_CACHE_ALL ||	\
-	(((_level) > 0 ||						\
-	DMU_OT_IS_METADATA((_dn)->dn_handle->dnh_dnode->dn_type)) &&	\
-	((_dn)->dn_objset->os_secondary_cache == ZFS_CACHE_METADATA)))
+boolean_t dbuf_is_l2cacheable(dmu_buf_impl_t *db);
 
 #ifdef ZFS_DEBUG
 
diff --git a/include/sys/dmu_objset.h b/include/sys/dmu_objset.h
index e89ee64ea..7ade2dc91 100644
--- a/include/sys/dmu_objset.h
+++ b/include/sys/dmu_objset.h
@@ -200,10 +200,6 @@ struct objset {
 #define	DMU_GROUPUSED_DNODE(os)	((os)->os_groupused_dnode.dnh_dnode)
 #define	DMU_PROJECTUSED_DNODE(os) ((os)->os_projectused_dnode.dnh_dnode)
 
-#define	DMU_OS_IS_L2CACHEABLE(os)				\
-	((os)->os_secondary_cache == ZFS_CACHE_ALL ||		\
-	(os)->os_secondary_cache == ZFS_CACHE_METADATA)
-
 /* called from zpl */
 int dmu_objset_hold(const char *name, void *tag, objset_t **osp);
 int dmu_objset_hold_flags(const char *name, boolean_t decrypt, void *tag,
diff --git a/man/man4/zfs.4 b/man/man4/zfs.4
index a136690c7..9c2168870 100644
--- a/man/man4/zfs.4
+++ b/man/man4/zfs.4
@@ -109,6 +109,11 @@ A value of
 .Sy 100
 disables this feature.
 .
+.It Sy l2arc_exclude_special Ns = Ns Sy 0 Ns | Ns 1 Pq int
+Controls whether buffers present on special vdevs are eligibile for caching
+into L2ARC.
+If set to 1, exclude dbufs on special vdevs from being cached to L2ARC.
+.
 .It Sy l2arc_mfuonly Ns = Ns Sy 0 Ns | Ns 1 Pq  int
 Controls whether only MFU metadata and data are cached from ARC into L2ARC.
 This may be desired to avoid wasting space on L2ARC when reading/writing large
diff --git a/module/zfs/arc.c b/module/zfs/arc.c
index f0330150f..79e2d4381 100644
--- a/module/zfs/arc.c
+++ b/module/zfs/arc.c
@@ -870,6 +870,14 @@ static void l2arc_hdr_arcstats_update(arc_buf_hdr_t *hdr, boolean_t incr,
 	l2arc_hdr_arcstats_update((hdr), B_FALSE, B_TRUE)
 
 /*
+ * l2arc_exclude_special : A zfs module parameter that controls whether buffers
+ * 		present on special vdevs are eligibile for caching in L2ARC. If
+ * 		set to 1, exclude dbufs on special vdevs from being cached to
+ * 		L2ARC.
+ */
+int l2arc_exclude_special = 0;
+
+/*
  * l2arc_mfuonly : A ZFS module parameter that controls whether only MFU
  * 		metadata and data are cached from ARC into L2ARC.
  */
@@ -11097,6 +11105,10 @@ ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, rebuild_blocks_min_l2size, ULONG, ZMOD_RW,
 ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, mfuonly, INT, ZMOD_RW,
 	"Cache only MFU data from ARC into L2ARC");
 
+ZFS_MODULE_PARAM(zfs_l2arc, l2arc_, exclude_special, INT, ZMOD_RW,
+	"If set to 1 exclude dbufs on special vdevs from being cached to "
+	"L2ARC.");
+
 ZFS_MODULE_PARAM_CALL(zfs_arc, zfs_arc_, lotsfree_percent, param_set_arc_int,
 	param_get_int, ZMOD_RW, "System free memory I/O throttle in bytes");
 
diff --git a/module/zfs/dbuf.c b/module/zfs/dbuf.c
index 289247c6e..fe54da425 100644
--- a/module/zfs/dbuf.c
+++ b/module/zfs/dbuf.c
@@ -53,6 +53,7 @@
 #include <cityhash.h>
 #include <sys/spa_impl.h>
 #include <sys/wmsum.h>
+#include <sys/vdev_impl.h>
 
 kstat_t *dbuf_ksp;
 
@@ -599,6 +600,68 @@ dbuf_is_metadata(dmu_buf_impl_t *db)
 	}
 }
 
+/*
+ * We want to exclude buffers that are on a special allocation class from
+ * L2ARC.
+ */
+boolean_t
+dbuf_is_l2cacheable(dmu_buf_impl_t *db)
+{
+	vdev_t *vd = NULL;
+	zfs_cache_type_t cache = db->db_objset->os_secondary_cache;
+	blkptr_t *bp = db->db_blkptr;
+
+	if (bp != NULL && !BP_IS_HOLE(bp)) {
+		uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
+		vdev_t *rvd = db->db_objset->os_spa->spa_root_vdev;
+
+		if (vdev < rvd->vdev_children)
+			vd = rvd->vdev_child[vdev];
+
+		if (cache == ZFS_CACHE_ALL ||
+		    (dbuf_is_metadata(db) && cache == ZFS_CACHE_METADATA)) {
+			if (vd == NULL)
+				return (B_TRUE);
+
+			if ((vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
+			    vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) ||
+			    l2arc_exclude_special == 0)
+				return (B_TRUE);
+		}
+	}
+
+	return (B_FALSE);
+}
+
+static inline boolean_t
+dnode_level_is_l2cacheable(blkptr_t *bp, dnode_t *dn, int64_t level)
+{
+	vdev_t *vd = NULL;
+	zfs_cache_type_t cache = dn->dn_objset->os_secondary_cache;
+
+	if (bp != NULL && !BP_IS_HOLE(bp)) {
+		uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
+		vdev_t *rvd = dn->dn_objset->os_spa->spa_root_vdev;
+
+		if (vdev < rvd->vdev_children)
+			vd = rvd->vdev_child[vdev];
+
+		if (cache == ZFS_CACHE_ALL || ((level > 0 ||
+		    DMU_OT_IS_METADATA(dn->dn_handle->dnh_dnode->dn_type)) &&
+		    cache == ZFS_CACHE_METADATA)) {
+			if (vd == NULL)
+				return (B_TRUE);
+
+			if ((vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
+			    vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) ||
+			    l2arc_exclude_special == 0)
+				return (B_TRUE);
+		}
+	}
+
+	return (B_FALSE);
+}
+
 
 /*
  * This function *must* return indices evenly distributed between all
@@ -1527,7 +1590,7 @@ dbuf_read_impl(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags,
 	DTRACE_SET_STATE(db, "read issued");
 	mutex_exit(&db->db_mtx);
 
-	if (DBUF_IS_L2CACHEABLE(db))
+	if (dbuf_is_l2cacheable(db))
 		aflags |= ARC_FLAG_L2CACHE;
 
 	dbuf_add_ref(db, NULL);
@@ -3370,7 +3433,7 @@ dbuf_prefetch_impl(dnode_t *dn, int64_t level, uint64_t blkid,
 	dpa->dpa_arg = arg;
 
 	/* flag if L2ARC eligible, l2arc_noprefetch then decides */
-	if (DNODE_LEVEL_IS_L2CACHEABLE(dn, level))
+	if (dnode_level_is_l2cacheable(&bp, dn, level))
 		dpa->dpa_aflags |= ARC_FLAG_L2CACHE;
 
 	/*
@@ -3388,7 +3451,7 @@ dbuf_prefetch_impl(dnode_t *dn, int64_t level, uint64_t blkid,
 		zbookmark_phys_t zb;
 
 		/* flag if L2ARC eligible, l2arc_noprefetch then decides */
-		if (DNODE_LEVEL_IS_L2CACHEABLE(dn, level))
+		if (dnode_level_is_l2cacheable(&bp, dn, level))
 			iter_aflags |= ARC_FLAG_L2CACHE;
 
 		SET_BOOKMARK(&zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET,
@@ -4986,7 +5049,7 @@ dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx)
 			children_ready_cb = dbuf_write_children_ready;
 
 		dr->dr_zio = arc_write(pio, os->os_spa, txg,
-		    &dr->dr_bp_copy, data, DBUF_IS_L2CACHEABLE(db),
+		    &dr->dr_bp_copy, data, dbuf_is_l2cacheable(db),
 		    &zp, dbuf_write_ready,
 		    children_ready_cb, dbuf_write_physdone,
 		    dbuf_write_done, db, ZIO_PRIORITY_ASYNC_WRITE,
diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c
index f12c5eda8..eee3e70bb 100644
--- a/module/zfs/dmu.c
+++ b/module/zfs/dmu.c
@@ -1839,7 +1839,7 @@ dmu_sync(zio_t *pio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd)
 	dsa->dsa_tx = NULL;
 
 	zio_nowait(arc_write(pio, os->os_spa, txg,
-	    zgd->zgd_bp, dr->dt.dl.dr_data, DBUF_IS_L2CACHEABLE(db),
+	    zgd->zgd_bp, dr->dt.dl.dr_data, dbuf_is_l2cacheable(db),
 	    &zp, dmu_sync_ready, NULL, NULL, dmu_sync_done, dsa,
 	    ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, &zb));
 
diff --git a/module/zfs/dmu_objset.c b/module/zfs/dmu_objset.c
index af107fb8a..b30a9d619 100644
--- a/module/zfs/dmu_objset.c
+++ b/module/zfs/dmu_objset.c
@@ -63,6 +63,8 @@
 #include <sys/dmu_recv.h>
 #include <sys/zfs_project.h>
 #include "zfs_namecheck.h"
+#include <sys/vdev_impl.h>
+#include <sys/arc.h>
 
 /*
  * Needed to close a window in dnode_move() that allows the objset to be freed
@@ -411,6 +413,34 @@ dnode_multilist_index_func(multilist_t *ml, void *obj)
 	    multilist_get_num_sublists(ml));
 }
 
+static inline boolean_t
+dmu_os_is_l2cacheable(objset_t *os)
+{
+	vdev_t *vd = NULL;
+	zfs_cache_type_t cache = os->os_secondary_cache;
+	blkptr_t *bp = os->os_rootbp;
+
+	if (bp != NULL && !BP_IS_HOLE(bp)) {
+		uint64_t vdev = DVA_GET_VDEV(bp->blk_dva);
+		vdev_t *rvd = os->os_spa->spa_root_vdev;
+
+		if (vdev < rvd->vdev_children)
+			vd = rvd->vdev_child[vdev];
+
+		if (cache == ZFS_CACHE_ALL || cache == ZFS_CACHE_METADATA) {
+			if (vd == NULL)
+				return (B_TRUE);
+
+			if ((vd->vdev_alloc_bias != VDEV_BIAS_SPECIAL &&
+			    vd->vdev_alloc_bias != VDEV_BIAS_DEDUP) ||
+			    l2arc_exclude_special == 0)
+				return (B_TRUE);
+		}
+	}
+
+	return (B_FALSE);
+}
+
 /*
  * Instantiates the objset_t in-memory structure corresponding to the
  * objset_phys_t that's pointed to by the specified blkptr_t.
@@ -453,7 +483,7 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
 		SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET,
 		    ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID);
 
-		if (DMU_OS_IS_L2CACHEABLE(os))
+		if (dmu_os_is_l2cacheable(os))
 			aflags |= ARC_FLAG_L2CACHE;
 
 		if (ds != NULL && ds->ds_dir->dd_crypto_obj != 0) {
@@ -1663,7 +1693,7 @@ dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx)
 	}
 
 	zio = arc_write(pio, os->os_spa, tx->tx_txg,
-	    blkptr_copy, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os),
+	    blkptr_copy, os->os_phys_buf, dmu_os_is_l2cacheable(os),
 	    &zp, dmu_objset_write_ready, NULL, NULL, dmu_objset_write_done,
 	    os, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb);