Pool allocation classes

Allocation Classes add the ability to have allocation classes in a
pool that are dedicated to serving specific block categories, such
as DDT data, metadata, and small file blocks. A pool can opt-in to
this feature by adding a 'special' or 'dedup' top-level VDEV.

Reviewed by: Pavel Zakharov <[email protected]>
Reviewed-by: Richard Laager <[email protected]>
Reviewed-by: Alek Pinchuk <[email protected]>
Reviewed-by: Håkan Johansson <[email protected]>
Reviewed-by: Andreas Dilger <[email protected]>
Reviewed-by: DHE <[email protected]>
Reviewed-by: Richard Elling <[email protected]>
Reviewed-by: Gregor Kopka <[email protected]>
Reviewed-by: Kash Pande <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Reviewed-by: Matthew Ahrens <[email protected]>
Signed-off-by: Don Brady <[email protected]>
Closes #5182 

10 files changed, 65 insertions, 8 deletions

diff --git a/include/sys/dmu.h b/include/sys/dmu.h
index 67f4be1c2..bc7046fdc 100644
--- a/include/sys/dmu.h
+++ b/include/sys/dmu.h
@@ -20,12 +20,13 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
  * Copyright 2014 HybridCluster. All rights reserved.
  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
  * Copyright 2013 Saso Kiselkov. All rights reserved.
+ * Copyright (c) 2017, Intel Corporation.
  */
 
 /* Portions Copyright 2010 Robert Milkowski */
@@ -136,6 +137,16 @@ typedef enum dmu_object_byteswap {
 	((ot) & DMU_OT_METADATA) : \
 	DMU_OT_IS_METADATA_IMPL(ot))
 
+#define	DMU_OT_IS_DDT(ot) \
+	((ot) == DMU_OT_DDT_ZAP)
+
+#define	DMU_OT_IS_ZIL(ot) \
+	((ot) == DMU_OT_INTENT_LOG)
+
+/* Note: ztest uses DMU_OT_UINT64_OTHER as a proxy for file blocks */
+#define	DMU_OT_IS_FILE(ot) \
+	((ot) == DMU_OT_PLAIN_FILE_CONTENTS || (ot) == DMU_OT_UINT64_OTHER)
+
 #define	DMU_OT_IS_ENCRYPTED(ot) (((ot) & DMU_OT_NEWTYPE) ? \
 	((ot) & DMU_OT_ENCRYPTED) : \
 	DMU_OT_IS_ENCRYPTED_IMPL(ot))
@@ -226,7 +237,7 @@ typedef enum dmu_object_type {
 	 * values.
 	 *
 	 * The DMU_OTN_* types do not have entries in the dmu_ot table,
-	 * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead
+	 * use the DMU_OT_IS_METADATA() and DMU_OT_BYTESWAP() macros instead
 	 * of indexing into dmu_ot directly (this works for both DMU_OT_* types
 	 * and DMU_OTN_* types).
 	 */
diff --git a/include/sys/dmu_objset.h b/include/sys/dmu_objset.h
index fd5afc0bc..c0650bcde 100644
--- a/include/sys/dmu_objset.h
+++ b/include/sys/dmu_objset.h
@@ -20,7 +20,7 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2012, 2017 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2018 by Delphix. All rights reserved.
  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
  */
@@ -137,6 +137,11 @@ struct objset {
 	uint64_t os_normalization;
 	uint64_t os_utf8only;
 	uint64_t os_casesensitivity;
+	/*
+	 * The largest zpl file block allowed in special class.
+	 * cached here instead of zfsvfs for easier access.
+	 */
+	int os_zpl_special_smallblock;
 
 	/*
 	 * Pointer is constant; the blkptr it points to is protected by
diff --git a/include/sys/fs/zfs.h b/include/sys/fs/zfs.h
index c8c7cebae..6bbf84346 100644
--- a/include/sys/fs/zfs.h
+++ b/include/sys/fs/zfs.h
@@ -26,6 +26,7 @@
  * Copyright (c) 2013, 2017 Joyent, Inc. All rights reserved.
  * Copyright (c) 2014 Integros [integros.com]
  * Copyright (c) 2017 Datto Inc.
+ * Copyright (c) 2017, Intel Corporation.
  */
 
 /* Portions Copyright 2010 Robert Milkowski */
@@ -181,6 +182,7 @@ typedef enum {
 	ZFS_PROP_KEY_GUID,
 	ZFS_PROP_KEYSTATUS,
 	ZFS_PROP_REMAPTXG,		/* not exposed to the user */
+	ZFS_PROP_SPECIAL_SMALL_BLOCKS,
 	ZFS_NUM_PROPS
 } zfs_prop_t;
 
@@ -713,6 +715,7 @@ typedef struct zpool_load_policy {
 #define	ZPOOL_CONFIG_MMP_TXG		"mmp_txg"	/* not stored on disk */
 #define	ZPOOL_CONFIG_MMP_HOSTNAME	"mmp_hostname"	/* not stored on disk */
 #define	ZPOOL_CONFIG_MMP_HOSTID		"mmp_hostid"	/* not stored on disk */
+#define	ZPOOL_CONFIG_ALLOCATION_BIAS	"alloc_bias"	/* not stored on disk */
 
 /*
  * The persistent vdev state is stored as separate values rather than a single
@@ -759,6 +762,14 @@ typedef struct zpool_load_policy {
 #define	VDEV_TOP_ZAP_POOL_CHECKPOINT_SM \
 	"com.delphix:pool_checkpoint_sm"
 
+#define	VDEV_TOP_ZAP_ALLOCATION_BIAS \
+	"org.zfsonlinux:allocation_bias"
+
+/* vdev metaslab allocation bias */
+#define	VDEV_ALLOC_BIAS_LOG		"log"
+#define	VDEV_ALLOC_BIAS_SPECIAL		"special"
+#define	VDEV_ALLOC_BIAS_DEDUP		"dedup"
+
 /*
  * This is needed in userland to report the minimum necessary device size.
  */
diff --git a/include/sys/metaslab.h b/include/sys/metaslab.h
index 545bcafa5..fca233a38 100644
--- a/include/sys/metaslab.h
+++ b/include/sys/metaslab.h
@@ -21,6 +21,7 @@
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
+ * Copyright (c) 2017, Intel Corporation.
  */
 
 #ifndef _SYS_METASLAB_H
@@ -57,13 +58,17 @@ void metaslab_sync_done(metaslab_t *, uint64_t);
 void metaslab_sync_reassess(metaslab_group_t *);
 uint64_t metaslab_block_maxsize(metaslab_t *);
 
+/*
+ * metaslab alloc flags
+ */
 #define	METASLAB_HINTBP_FAVOR		0x0
 #define	METASLAB_HINTBP_AVOID		0x1
 #define	METASLAB_GANG_HEADER		0x2
 #define	METASLAB_GANG_CHILD		0x4
 #define	METASLAB_ASYNC_ALLOC		0x8
 #define	METASLAB_DONT_THROTTLE		0x10
-#define	METASLAB_FASTWRITE		0x20
+#define	METASLAB_MUST_RESERVE		0x20
+#define	METASLAB_FASTWRITE		0x40
 
 int metaslab_alloc(spa_t *, metaslab_class_t *, uint64_t,
     blkptr_t *, int, uint64_t, blkptr_t *, int, zio_alloc_list_t *, zio_t *,
@@ -96,8 +101,6 @@ boolean_t metaslab_class_throttle_reserve(metaslab_class_t *, int, int,
     zio_t *, int);
 void metaslab_class_throttle_unreserve(metaslab_class_t *, int, int, zio_t *);
 
-void metaslab_class_space_update(metaslab_class_t *, int64_t, int64_t,
-    int64_t, int64_t);
 uint64_t metaslab_class_get_alloc(metaslab_class_t *);
 uint64_t metaslab_class_get_space(metaslab_class_t *);
 uint64_t metaslab_class_get_dspace(metaslab_class_t *);
diff --git a/include/sys/spa.h b/include/sys/spa.h
index 82fe2c18c..b86c65557 100644
--- a/include/sys/spa.h
+++ b/include/sys/spa.h
@@ -20,13 +20,14 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
  * Copyright 2013 Saso Kiselkov. All rights reserved.
  * Copyright (c) 2014 Integros [integros.com]
  * Copyright 2017 Joyent, Inc.
  * Copyright (c) 2017 Datto Inc.
+ * Copyright (c) 2017, Intel Corporation.
  */
 
 #ifndef _SYS_SPA_H
@@ -976,6 +977,11 @@ extern uint64_t spa_version(spa_t *spa);
 extern boolean_t spa_deflate(spa_t *spa);
 extern metaslab_class_t *spa_normal_class(spa_t *spa);
 extern metaslab_class_t *spa_log_class(spa_t *spa);
+extern metaslab_class_t *spa_special_class(spa_t *spa);
+extern metaslab_class_t *spa_dedup_class(spa_t *spa);
+extern metaslab_class_t *spa_preferred_class(spa_t *spa, uint64_t size,
+    dmu_object_type_t objtype, uint_t level, uint_t special_smallblk);
+
 extern void spa_evicting_os_register(spa_t *, objset_t *os);
 extern void spa_evicting_os_deregister(spa_t *, objset_t *os);
 extern void spa_evicting_os_wait(spa_t *spa);
@@ -1040,6 +1046,7 @@ extern void spa_set_missing_tvds(spa_t *spa, uint64_t missing);
 extern boolean_t spa_top_vdevs_spacemap_addressable(spa_t *spa);
 extern boolean_t spa_multihost(spa_t *spa);
 extern unsigned long spa_get_hostid(void);
+extern void spa_activate_allocation_classes(spa_t *, dmu_tx_t *);
 
 extern int spa_mode(spa_t *spa);
 extern uint64_t zfs_strtonum(const char *str, char **nptr);
diff --git a/include/sys/spa_impl.h b/include/sys/spa_impl.h
index 1b8e48180..676e8b8a2 100644
--- a/include/sys/spa_impl.h
+++ b/include/sys/spa_impl.h
@@ -20,12 +20,13 @@
  */
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
  * Copyright 2011 Nexenta Systems, Inc.  All rights reserved.
  * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
  * Copyright 2013 Saso Kiselkov. All rights reserved.
  * Copyright (c) 2016 Actifio, Inc. All rights reserved.
  * Copyright (c) 2017 Datto Inc.
+ * Copyright (c) 2017, Intel Corporation.
  */
 
 #ifndef _SYS_SPA_IMPL_H
@@ -220,6 +221,8 @@ struct spa {
 	boolean_t	spa_is_initializing;	/* true while opening pool */
 	metaslab_class_t *spa_normal_class;	/* normal data class */
 	metaslab_class_t *spa_log_class;	/* intent log data class */
+	metaslab_class_t *spa_special_class;	/* special allocation class */
+	metaslab_class_t *spa_dedup_class;	/* dedup allocation class */
 	uint64_t	spa_first_txg;		/* first txg after spa_open() */
 	uint64_t	spa_final_txg;		/* txg of export/destroy */
 	uint64_t	spa_freeze_txg;		/* freeze pool at this txg */
diff --git a/include/sys/vdev.h b/include/sys/vdev.h
index 6d31d61b5..b37b60bdd 100644
--- a/include/sys/vdev.h
+++ b/include/sys/vdev.h
@@ -22,6 +22,7 @@
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
+ * Copyright (c) 2017, Intel Corporation.
  */
 
 #ifndef _SYS_VDEV_H
@@ -108,6 +109,8 @@ extern boolean_t vdev_children_are_offline(vdev_t *vd);
 extern void vdev_space_update(vdev_t *vd,
     int64_t alloc_delta, int64_t defer_delta, int64_t space_delta);
 
+extern int64_t vdev_deflated_space(vdev_t *vd, int64_t space);
+
 extern uint64_t vdev_psize_to_asize(vdev_t *vd, uint64_t psize);
 
 extern int vdev_fault(spa_t *spa, uint64_t guid, vdev_aux_t aux);
diff --git a/include/sys/vdev_impl.h b/include/sys/vdev_impl.h
index 701328ea6..e055161e8 100644
--- a/include/sys/vdev_impl.h
+++ b/include/sys/vdev_impl.h
@@ -21,6 +21,7 @@
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
+ * Copyright (c) 2017, Intel Corporation.
  */
 
 #ifndef _SYS_VDEV_IMPL_H
@@ -140,6 +141,14 @@ struct vdev_queue {
 	kmutex_t	vq_lock;
 };
 
+typedef enum vdev_alloc_bias {
+	VDEV_BIAS_NONE,
+	VDEV_BIAS_LOG,		/* dedicated to ZIL data (SLOG) */
+	VDEV_BIAS_SPECIAL,	/* dedicated to ddt, metadata, and small blks */
+	VDEV_BIAS_DEDUP		/* dedicated to dedup metadata */
+} vdev_alloc_bias_t;
+
+
 /*
  * On-disk indirect vdev state.
  *
@@ -236,6 +245,7 @@ struct vdev {
 	boolean_t	vdev_ishole;	/* is a hole in the namespace	*/
 	kmutex_t	vdev_queue_lock; /* protects vdev_queue_depth	*/
 	uint64_t	vdev_top_zap;
+	vdev_alloc_bias_t vdev_alloc_bias; /* metaslab allocation bias	*/
 
 	/* pool checkpoint related */
 	space_map_t	*vdev_checkpoint_sm;	/* contains reserved blocks */
diff --git a/include/sys/zio.h b/include/sys/zio.h
index d32a42833..322006649 100644
--- a/include/sys/zio.h
+++ b/include/sys/zio.h
@@ -353,6 +353,7 @@ typedef struct zio_prop {
 	uint8_t			zp_salt[ZIO_DATA_SALT_LEN];
 	uint8_t			zp_iv[ZIO_DATA_IV_LEN];
 	uint8_t			zp_mac[ZIO_DATA_MAC_LEN];
+	uint32_t		zp_zpl_smallblk;
 } zio_prop_t;
 
 typedef struct zio_cksum_report zio_cksum_report_t;
@@ -473,6 +474,7 @@ struct zio {
 	vdev_t		*io_vd;
 	void		*io_vsd;
 	const zio_vsd_ops_t *io_vsd_ops;
+	metaslab_class_t *io_metaslab_class;	/* dva throttle class */
 
 	uint64_t	io_offset;
 	hrtime_t	io_timestamp;	/* submitted at */
diff --git a/include/zfeature_common.h b/include/zfeature_common.h
index 8d31309e8..84bc7f816 100644
--- a/include/zfeature_common.h
+++ b/include/zfeature_common.h
@@ -23,6 +23,7 @@
  * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
  * Copyright (c) 2013, Joyent, Inc. All rights reserved.
+ * Copyright (c) 2017, Intel Corporation.
  */
 
 #ifndef _ZFEATURE_COMMON_H
@@ -63,6 +64,7 @@ typedef enum spa_feature {
 	SPA_FEATURE_OBSOLETE_COUNTS,
 	SPA_FEATURE_POOL_CHECKPOINT,
 	SPA_FEATURE_SPACEMAP_V2,
+	SPA_FEATURE_ALLOCATION_CLASSES,
 	SPA_FEATURES
 } spa_feature_t;