linux-gen: align: Fix alignment for typedef definitions

	* typedef struct { } ODP_ALIGN() test_t; is correct
	* typedef struct ODP_ALIGN() {} test_t; is correct and
	preferred from gcc standards
	* typedef struct {} test_t ODP_ALIGN() discards align
	Had segfauls on gcc-7 using that

Signed-off-by: Ilias Apalodimas <ilias.apalodimas@linaro.org>
Reviewed-by: Bill Fischofer <bill.fischofer@linaro.org>
Signed-off-by: Maxim Uvarov <maxim.uvarov@linaro.org>

35 files changed, 160 insertions, 116 deletions

diff --git a/example/ipfragreass/odp_ipfragreass.c b/example/ipfragreass/odp_ipfragreass.c
index 6ad0db9fd..6668263b0 100644
--- a/example/ipfragreass/odp_ipfragreass.c
+++ b/example/ipfragreass/odp_ipfragreass.c
@@ -47,9 +47,9 @@ static odp_queue_t reassembled_pkts;
 static odp_atomic_u32_t packets_reassembled;
 
 /** Number of fragments processed per thread in reassembly (for printing) */
-static struct {
+static struct ODP_ALIGNED_CACHE {
 	uint32_t frags;
-} ODP_ALIGNED_CACHE thread_stats[MAX_WORKERS];
+} thread_stats[MAX_WORKERS];
 
 /** Shared hash map structure for reassembly */
 static union fraglist *fraglists;
diff --git a/example/l3fwd/odp_l3fwd_db.h b/example/l3fwd/odp_l3fwd_db.h
index 8df2999df..015f38325 100644
--- a/example/l3fwd/odp_l3fwd_db.h
+++ b/example/l3fwd/odp_l3fwd_db.h
@@ -39,7 +39,7 @@ typedef struct ip_addr_range_s {
 /**
  * TCP/UDP flow
  */
-typedef struct ipv4_tuple5_s {
+typedef struct ODP_ALIGNED_CACHE ipv4_tuple5_s {
 	union {
 		struct {
 			int32_t src_ip;
@@ -55,7 +55,7 @@ typedef struct ipv4_tuple5_s {
 			int64_t lo64;
 		};
 	};
-} ipv4_tuple5_t ODP_ALIGNED_CACHE;
+} ipv4_tuple5_t;
 
 /**
  * Forwarding data base entry
diff --git a/example/switch/odp_switch.c b/example/switch/odp_switch.c
index 3e52970be..954c667ca 100644
--- a/example/switch/odp_switch.c
+++ b/example/switch/odp_switch.c
@@ -66,7 +66,7 @@ static int exit_threads;   /**< Break workers loop if set to 1 */
 /**
  * Statistics
  */
-typedef union {
+typedef union ODP_ALIGNED_CACHE {
 	struct {
 		/** Number of received packets */
 		uint64_t rx_packets;
@@ -79,7 +79,7 @@ typedef union {
 	} s;
 
 	uint8_t padding[ODP_CACHE_LINE_SIZE];
-} stats_t ODP_ALIGNED_CACHE;
+} stats_t;
 
 /**
  * Packet buffer
diff --git a/helper/cuckootable.c b/helper/cuckootable.c
index 4707191d7..29298d42d 100644
--- a/helper/cuckootable.c
+++ b/helper/cuckootable.c
@@ -93,12 +93,12 @@ struct cuckoo_table_key_value {
  *  into a bucket, and each bucket has at most HASH_BUCKET_ENTRIES
  *  elements.
  */
-struct cuckoo_table_bucket {
+struct ODP_ALIGNED_CACHE cuckoo_table_bucket {
 	struct cuckoo_table_signatures signatures[HASH_BUCKET_ENTRIES];
 	/* Includes dummy key index that always contains index 0 */
 	odp_buffer_t key_buf[HASH_BUCKET_ENTRIES + 1];
 	uint8_t flag[HASH_BUCKET_ENTRIES];
-} ODP_ALIGNED_CACHE;
+};
 
 /* More efficient access to a map of single ullong */
 #define ULLONG_FOR_EACH_1(IDX, MAP)	\
@@ -106,7 +106,7 @@ struct cuckoo_table_bucket {
 		 MAP = (MAP & (MAP - 1)))
 
 /** A hash table structure. */
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/**< for check */
 	uint32_t magicword;
 	/**< Name of the hash. */
@@ -126,7 +126,7 @@ typedef struct {
 	/** Table with buckets storing all the hash values and key indexes
 	  to the key table*/
 	struct cuckoo_table_bucket *buckets;
-} odph_cuckoo_table_impl ODP_ALIGNED_CACHE;
+} odph_cuckoo_table_impl;
 
 /**
  * Aligns input parameter to the next power of 2
diff --git a/helper/iplookuptable.c b/helper/iplookuptable.c
index 7ca68de27..7b1faa292 100644
--- a/helper/iplookuptable.c
+++ b/helper/iplookuptable.c
@@ -96,7 +96,7 @@ typedef enum {
 } cache_type_t;
 
 /** A IP lookup table structure. */
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/**< for check */
 	uint32_t magicword;
 	/** Name of the hash. */
@@ -118,7 +118,7 @@ typedef struct {
 	odp_queue_t free_slots[2];
 	/** The number of pool used by each queue. */
 	uint32_t cache_count[2];
-} odph_iplookup_table_impl ODP_ALIGNED_CACHE;
+} odph_iplookup_table_impl;
 
 /***********************************************************
  *****************   Cache management   ********************
diff --git a/include/odp/api/abi-default/cpumask.h b/include/odp/api/abi-default/cpumask.h
index 69f4c1620..4c3508228 100644
--- a/include/odp/api/abi-default/cpumask.h
+++ b/include/odp/api/abi-default/cpumask.h
@@ -33,14 +33,14 @@ extern "C" {
  *
  * Don't access directly, use access functions.
  */
-typedef struct odp_cpumask_t {
+typedef struct ODP_ALIGNED(8) odp_cpumask_t {
 	/** @private CPU mask storage
 	  *
 	  * This is private to the implementation.
 	  * Don't access directly, use access functions.
 	  */
 	uint8_t _u8[ODP_CPUMASK_SIZE / 8];
-} odp_cpumask_t ODP_ALIGNED(8);
+} odp_cpumask_t;
 
 /**
  * @}
diff --git a/platform/linux-generic/include/odp/api/plat/static_inline.h b/platform/linux-generic/include/odp/api/plat/static_inline.h
new file mode 100644
index 000000000..633e2f273
--- /dev/null
+++ b/platform/linux-generic/include/odp/api/plat/static_inline.h
@@ -0,0 +1,43 @@
+/* Copyright (c) 2016, Linaro Limited
+ * All rights reserved.
+ *
+ * SPDX-License-Identifier: BSD-3-Clause
+ */
+
+/**
+ * @file
+ *
+ * Macro for static inline functions
+ */
+
+#ifndef ODP_PLAT_STATIC_INLINE_H_
+#define ODP_PLAT_STATIC_INLINE_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/**
+ * @internal
+ * @def ODP_ABI_COMPAT
+ * Control ABI compatibility
+ */
+
+/**
+ * @internal
+ * @def _ODP_INLINE
+ * Define a function as inlined or not inlined (for ABI compatibility)
+ */
+#if 1
+#define ODP_ABI_COMPAT 1
+#define _ODP_INLINE
+#else
+#define ODP_ABI_COMPAT 0
+#define _ODP_INLINE static inline
+#endif
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif
diff --git a/platform/linux-generic/include/odp_atomic_internal.h b/platform/linux-generic/include/odp_atomic_internal.h
index f3696a994..43c29a761 100644
--- a/platform/linux-generic/include/odp_atomic_internal.h
+++ b/platform/linux-generic/include/odp_atomic_internal.h
@@ -29,10 +29,9 @@ extern "C" {
 /**
  * Pointer atomic type
  */
-typedef struct {
+typedef struct ODP_ALIGNED(sizeof(void *)) {
 	void *v; /**< Actual storage for the atomic variable */
-} _odp_atomic_ptr_t
-ODP_ALIGNED(sizeof(void *)); /* Enforce alignement! */
+} _odp_atomic_ptr_t;
 
 /**
  * Atomic flag (boolean) type
@@ -709,9 +708,9 @@ static inline void _odp_atomic_flag_clear(_odp_atomic_flag_t *flag)
 typedef __int128 _uint128_t;
 
 /** Atomic 128-bit type */
-typedef struct {
+typedef struct ODP_ALIGNED(16) {
 	_uint128_t v; /**< Actual storage for the atomic variable */
-} _odp_atomic_u128_t ODP_ALIGNED(16);
+} _odp_atomic_u128_t;
 
 /**
  * 16-byte atomic exchange operation
diff --git a/platform/linux-generic/include/odp_buffer_internal.h b/platform/linux-generic/include/odp_buffer_internal.h
index c56c5b01b..b52669387 100644
--- a/platform/linux-generic/include/odp_buffer_internal.h
+++ b/platform/linux-generic/include/odp_buffer_internal.h
@@ -42,7 +42,7 @@ typedef struct seg_entry_t {
 } seg_entry_t;
 
 /* Common buffer header */
-struct odp_buffer_hdr_t {
+struct ODP_ALIGNED_CACHE odp_buffer_hdr_t {
 
 	/* Buffer index in the pool */
 	uint32_t  index;
@@ -110,7 +110,7 @@ struct odp_buffer_hdr_t {
 
 	/* Data or next header */
 	uint8_t data[0];
-} ODP_ALIGNED_CACHE;
+};
 
 ODP_STATIC_ASSERT(CONFIG_PACKET_SEGS_PER_HDR < 256,
 		  "CONFIG_PACKET_SEGS_PER_HDR_TOO_LARGE");
diff --git a/platform/linux-generic/include/odp_ipsec_internal.h b/platform/linux-generic/include/odp_ipsec_internal.h
index 70a583c5b..dbdcbb917 100644
--- a/platform/linux-generic/include/odp_ipsec_internal.h
+++ b/platform/linux-generic/include/odp_ipsec_internal.h
@@ -92,7 +92,7 @@ int _odp_ipsec_status_send(odp_queue_t queue,
 #define ODP_CONFIG_IPSEC_SAS	8
 
 struct ipsec_sa_s {
-	odp_atomic_u32_t state ODP_ALIGNED_CACHE;
+	odp_atomic_u32_t ODP_ALIGNED_CACHE state;
 
 	uint32_t	ipsec_sa_idx;
 	odp_ipsec_sa_t	ipsec_sa_hdl;
diff --git a/platform/linux-generic/include/odp_packet_dpdk.h b/platform/linux-generic/include/odp_packet_dpdk.h
index 05f00ad34..242926555 100644
--- a/platform/linux-generic/include/odp_packet_dpdk.h
+++ b/platform/linux-generic/include/odp_packet_dpdk.h
@@ -41,13 +41,13 @@ struct pkt_cache_t {
 	unsigned count;			  /**< packets in cache */
 };
 
-typedef union {
+typedef union ODP_ALIGNED_CACHE {
 	struct pkt_cache_t s;
 	uint8_t pad[ROUNDUP_CACHE_LINE(sizeof(struct pkt_cache_t))];
-} pkt_cache_t ODP_ALIGNED_CACHE;
+} pkt_cache_t;
 
 /** Packet IO using DPDK interface */
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	odp_pool_t pool;		  /**< pool to alloc packets from */
 	struct rte_mempool *pkt_pool;	  /**< DPDK packet pool */
 	uint32_t data_room;		  /**< maximum packet length */
@@ -64,6 +64,6 @@ typedef struct {
 	odp_ticketlock_t tx_lock[PKTIO_MAX_QUEUES];  /**< TX queue locks */
 	/** cache for storing extra RX packets */
 	pkt_cache_t rx_cache[PKTIO_MAX_QUEUES];
-} pkt_dpdk_t ODP_ALIGNED_CACHE;
+} pkt_dpdk_t;
 
 #endif
diff --git a/platform/linux-generic/include/odp_packet_internal.h b/platform/linux-generic/include/odp_packet_internal.h
index 712799d71..01e0cae17 100644
--- a/platform/linux-generic/include/odp_packet_internal.h
+++ b/platform/linux-generic/include/odp_packet_internal.h
@@ -156,7 +156,7 @@ typedef struct {
 	/* Type of extra data */
 	uint8_t extra_type;
 	/* Extra space for packet descriptors. E.g. DPDK mbuf  */
-	uint8_t extra[PKT_EXTRA_LEN] ODP_ALIGNED_CACHE;
+	uint8_t ODP_ALIGNED_CACHE extra[PKT_EXTRA_LEN];
 #endif
 
 	/* Context for IPsec */
diff --git a/platform/linux-generic/include/odp_packet_io_ring_internal.h b/platform/linux-generic/include/odp_packet_io_ring_internal.h
index d044f9319..5a60a4ee3 100644
--- a/platform/linux-generic/include/odp_packet_io_ring_internal.h
+++ b/platform/linux-generic/include/odp_packet_io_ring_internal.h
@@ -132,26 +132,26 @@ typedef struct _ring {
 	int flags;
 
 	/** @private Producer */
-	struct _prod {
+	struct ODP_ALIGNED_CACHE _prod {
 		uint32_t watermark;      /* Maximum items */
 		uint32_t sp_enqueue;     /* True, if single producer. */
 		uint32_t size;           /* Size of ring. */
 		uint32_t mask;           /* Mask (size-1) of ring. */
 		volatile uint32_t head;  /* Producer head. */
 		volatile uint32_t tail;  /* Producer tail. */
-	} prod ODP_ALIGNED_CACHE;
+	} prod;
 
 	/** @private Consumer */
-	struct _cons {
+	struct ODP_ALIGNED_CACHE _cons {
 		uint32_t sc_dequeue;     /* True, if single consumer. */
 		uint32_t size;           /* Size of the ring. */
 		uint32_t mask;           /* Mask (size-1) of ring. */
 		volatile uint32_t head;  /* Consumer head. */
 		volatile uint32_t tail;  /* Consumer tail. */
-	} cons ODP_ALIGNED_CACHE;
+	} cons;
 
 	/** @private Memory space of ring starts here. */
-	void *ring[0] ODP_ALIGNED_CACHE;
+	void ODP_ALIGNED_CACHE *ring[0];
 } _ring_t;
 
 /* The default enqueue is "single-producer".*/
diff --git a/platform/linux-generic/include/odp_packet_netmap.h b/platform/linux-generic/include/odp_packet_netmap.h
index 876c7e465..616541b5e 100644
--- a/platform/linux-generic/include/odp_packet_netmap.h
+++ b/platform/linux-generic/include/odp_packet_netmap.h
@@ -30,10 +30,10 @@ struct netmap_ring_t {
 	odp_ticketlock_t lock;  /**< Queue lock */
 };
 
-typedef union {
+typedef union ODP_ALIGNED_CACHE {
 	struct netmap_ring_t s;
 	uint8_t pad[ROUNDUP_CACHE_LINE(sizeof(struct netmap_ring_t))];
-} netmap_ring_t ODP_ALIGNED_CACHE;
+} netmap_ring_t;
 
 /** Netmap ring slot */
 typedef struct  {
diff --git a/platform/linux-generic/include/odp_packet_socket.h b/platform/linux-generic/include/odp_packet_socket.h
index 0e61f6f0c..52d2776b7 100644
--- a/platform/linux-generic/include/odp_packet_socket.h
+++ b/platform/linux-generic/include/odp_packet_socket.h
@@ -67,11 +67,11 @@ ODP_STATIC_ASSERT(offsetof(struct ring, mm_space) <= ODP_CACHE_LINE_SIZE,
 /** Packet socket using mmap rings for both Rx and Tx */
 typedef struct {
 	/** Packet mmap ring for Rx */
-	struct ring rx_ring ODP_ALIGNED_CACHE;
+	struct ring ODP_ALIGNED_CACHE rx_ring;
 	/** Packet mmap ring for Tx */
-	struct ring tx_ring ODP_ALIGNED_CACHE;
+	struct ring ODP_ALIGNED_CACHE tx_ring;
 
-	int sockfd ODP_ALIGNED_CACHE;
+	int ODP_ALIGNED_CACHE sockfd;
 	odp_pool_t pool;
 	size_t frame_offset; /**< frame start offset from start of pkt buf */
 	uint8_t *mmap_base;
diff --git a/platform/linux-generic/include/odp_pool_internal.h b/platform/linux-generic/include/odp_pool_internal.h
index 61ec5ccc5..8284bcd7d 100644
--- a/platform/linux-generic/include/odp_pool_internal.h
+++ b/platform/linux-generic/include/odp_pool_internal.h
@@ -26,27 +26,27 @@ extern "C" {
 #include <odp_ring_internal.h>
 #include <odp/api/plat/strong_types.h>
 
-typedef struct pool_cache_t {
+typedef struct ODP_ALIGNED_CACHE pool_cache_t {
 	uint32_t num;
 	uint32_t buf_index[CONFIG_POOL_CACHE_SIZE];
 
-} pool_cache_t ODP_ALIGNED_CACHE;
+} pool_cache_t;
 
 /* Buffer header ring */
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/* Ring header */
 	ring_t   hdr;
 
 	/* Ring data: buffer handles */
 	uint32_t buf[CONFIG_POOL_MAX_NUM];
 
-} pool_ring_t ODP_ALIGNED_CACHE;
+} pool_ring_t;
 
 /* Callback function for pool destroy */
 typedef void (*pool_destroy_cb_fn)(void *pool);
 
 typedef struct pool_t {
-	odp_ticketlock_t lock ODP_ALIGNED_CACHE;
+	odp_ticketlock_t ODP_ALIGNED_CACHE lock;
 
 	char             name[ODP_POOL_NAME_LEN];
 	odp_pool_param_t params;
diff --git a/platform/linux-generic/include/odp_queue_internal.h b/platform/linux-generic/include/odp_queue_internal.h
index df1b29ff8..158b1d5f1 100644
--- a/platform/linux-generic/include/odp_queue_internal.h
+++ b/platform/linux-generic/include/odp_queue_internal.h
@@ -37,13 +37,13 @@ extern "C" {
 #define QUEUE_STATUS_SCHED        4
 
 struct queue_entry_s {
-	odp_ticketlock_t  lock ODP_ALIGNED_CACHE;
+	odp_ticketlock_t  ODP_ALIGNED_CACHE lock;
 
 	odp_buffer_hdr_t *head;
 	odp_buffer_hdr_t *tail;
 	int               status;
 
-	queue_enq_fn_t       enqueue ODP_ALIGNED_CACHE;
+	queue_enq_fn_t       ODP_ALIGNED_CACHE enqueue;
 	queue_deq_fn_t       dequeue;
 	queue_enq_multi_fn_t enqueue_multi;
 	queue_deq_multi_fn_t dequeue_multi;
diff --git a/platform/linux-generic/include/odp_queue_scalable_internal.h b/platform/linux-generic/include/odp_queue_scalable_internal.h
index 58d64aa5b..605d87502 100644
--- a/platform/linux-generic/include/odp_queue_scalable_internal.h
+++ b/platform/linux-generic/include/odp_queue_scalable_internal.h
@@ -34,10 +34,10 @@ extern "C" {
 struct queue_entry_s {
 	sched_elem_t     sched_elem;
 
-	odp_ticketlock_t lock ODP_ALIGNED_CACHE;
+	odp_ticketlock_t ODP_ALIGNED_CACHE lock;
 	int              status;
 
-	queue_enq_fn_t       enqueue ODP_ALIGNED_CACHE;
+	queue_enq_fn_t       ODP_ALIGNED_CACHE enqueue;
 	queue_deq_fn_t       dequeue;
 	queue_enq_multi_fn_t enqueue_multi;
 	queue_deq_multi_fn_t dequeue_multi;
diff --git a/platform/linux-generic/include/odp_ring_internal.h b/platform/linux-generic/include/odp_ring_internal.h
index 44b83c603..c2eedb8e2 100644
--- a/platform/linux-generic/include/odp_ring_internal.h
+++ b/platform/linux-generic/include/odp_ring_internal.h
@@ -25,7 +25,7 @@ extern "C" {
  * must be a power of two. Also ring size must be larger than the maximum
  * number of data items that will be stored on it (there's no check against
  * overwriting). */
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/* Writer head and tail */
 	odp_atomic_u32_t w_head;
 	odp_atomic_u32_t w_tail;
@@ -36,7 +36,7 @@ typedef struct {
 	odp_atomic_u32_t r_tail;
 
 	uint32_t data[0];
-} ring_t ODP_ALIGNED_CACHE;
+} ring_t;
 
 /* Initialize ring */
 static inline void ring_init(ring_t *ring)
diff --git a/platform/linux-generic/include/odp_schedule_scalable.h b/platform/linux-generic/include/odp_schedule_scalable.h
index ac357b015..92d60276b 100644
--- a/platform/linux-generic/include/odp_schedule_scalable.h
+++ b/platform/linux-generic/include/odp_schedule_scalable.h
@@ -29,7 +29,7 @@
 #define ODP_SCHED_PRIO_PKTIN 8
 #define ODP_SCHED_PRIO_NUM  9
 
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	union {
 		struct {
 			struct llqueue llq;
@@ -37,16 +37,16 @@ typedef struct {
 		};
 		char line[ODP_CACHE_LINE_SIZE];
 	};
-} sched_queue_t ODP_ALIGNED_CACHE;
+} sched_queue_t;
 
 #define TICKET_INVALID (uint16_t)(~0U)
 
-typedef struct {
+typedef struct ODP_ALIGNED(sizeof(uint64_t)) {
 	int32_t numevts;
 	uint16_t wrr_budget;
 	uint8_t cur_ticket;
 	uint8_t nxt_ticket;
-} qschedstate_t ODP_ALIGNED(sizeof(uint64_t));
+} qschedstate_t;
 
 typedef uint32_t ringidx_t;
 
@@ -58,7 +58,7 @@ typedef uint32_t ringidx_t;
 
 #define ODP_NO_SCHED_QUEUE (ODP_SCHED_SYNC_ORDERED + 1)
 
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	struct llnode node;
 	sched_queue_t *schedq;
 #ifdef CONFIG_QSCHST_LOCK
@@ -89,7 +89,7 @@ typedef struct {
 #define cons_type qschst_type
 #endif
 	odp_schedule_group_t sched_grp;
-} sched_elem_t ODP_ALIGNED_CACHE;
+} sched_elem_t;
 
 /* Number of scheduling groups */
 #define MAX_SCHED_GROUP (sizeof(sched_group_mask_t) * CHAR_BIT)
@@ -98,7 +98,7 @@ typedef bitset_t sched_group_mask_t;
 
 typedef struct {
 	/* Threads currently associated with the sched group */
-	bitset_t thr_actual[ODP_SCHED_PRIO_NUM] ODP_ALIGNED_CACHE;
+	bitset_t ODP_ALIGNED_CACHE thr_actual[ODP_SCHED_PRIO_NUM];
 	bitset_t thr_wanted;
 	/* Used to spread queues over schedq's */
 	uint32_t xcount[ODP_SCHED_PRIO_NUM];
@@ -106,13 +106,13 @@ typedef struct {
 	uint32_t xfactor;
 	char name[ODP_SCHED_GROUP_NAME_LEN];
 	/* ODP_SCHED_PRIO_NUM * xfactor. Must be last. */
-	sched_queue_t schedq[1] ODP_ALIGNED_CACHE;
+	sched_queue_t ODP_ALIGNED_CACHE schedq[1];
 } sched_group_t;
 
 /* Number of reorder contexts per thread */
 #define TS_RVEC_SIZE 16
 
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/* Atomic queue currently being processed or NULL */
 	sched_elem_t *atomq;
 	/* Schedq the currently processed queue was popped from */
@@ -135,10 +135,10 @@ typedef struct {
 	sched_group_mask_t sg_wanted[ODP_SCHED_PRIO_NUM];
 	bitset_t priv_rvec_free;
 	/* Bitset of free entries in rvec[] */
-	bitset_t rvec_free ODP_ALIGNED_CACHE;
+	bitset_t ODP_ALIGNED_CACHE rvec_free;
 	/* Reordering contexts to allocate from */
-	reorder_context_t rvec[TS_RVEC_SIZE] ODP_ALIGNED_CACHE;
-} sched_scalable_thread_state_t ODP_ALIGNED_CACHE;
+	reorder_context_t ODP_ALIGNED_CACHE rvec[TS_RVEC_SIZE];
+} sched_scalable_thread_state_t;
 
 void sched_update_enq(sched_elem_t *q, uint32_t actual);
 void sched_update_enq_sp(sched_elem_t *q, uint32_t actual);
diff --git a/platform/linux-generic/include/odp_schedule_scalable_ordered.h b/platform/linux-generic/include/odp_schedule_scalable_ordered.h
index c493dead4..14e411e23 100644
--- a/platform/linux-generic/include/odp_schedule_scalable_ordered.h
+++ b/platform/linux-generic/include/odp_schedule_scalable_ordered.h
@@ -57,12 +57,12 @@
  * We need a point of synchronisation where this knowledge and potential state
  * change can be transferred between threads.
  */
-typedef struct hc {
+typedef struct ODP_ALIGNED(sizeof(uint64_t)) hc {
 	/* First missing context */
 	uint32_t head;
 	/* Change indicator */
 	uint32_t chgi;
-} hc_t ODP_ALIGNED(sizeof(uint64_t));
+} hc_t;
 
 /* Number of reorder contects in the reorder window.
  * Should be at least one per CPU.
@@ -90,7 +90,7 @@ typedef struct reorder_window {
  */
 #define RC_EVT_SIZE 18
 
-struct reorder_context {
+struct ODP_ALIGNED_CACHE reorder_context {
 	/* Reorder window to which this context belongs */
 	reorder_window_t *rwin;
 	/* Pointer to TS->rvec_free */
@@ -109,7 +109,7 @@ struct reorder_context {
 	/* Events stored in this context */
 	odp_buffer_hdr_t *events[RC_EVT_SIZE];
 	queue_entry_t *destq[RC_EVT_SIZE];
-} ODP_ALIGNED_CACHE;
+};
 
 reorder_window_t *rwin_alloc(_odp_ishm_pool_t *pool,
 			     unsigned lock_count);
diff --git a/platform/linux-generic/odp_name_table.c b/platform/linux-generic/odp_name_table.c
index 3ff46b347..e9a84ef32 100644
--- a/platform/linux-generic/odp_name_table.c
+++ b/platform/linux-generic/odp_name_table.c
@@ -56,7 +56,7 @@ typedef struct name_tbl_entry_s name_tbl_entry_t;
  /* It is important for most platforms that the following struct fit within
  * one cacheline.
  */
-struct name_tbl_entry_s {
+struct ODP_ALIGNED_CACHE name_tbl_entry_s {
 	name_tbl_entry_t *next_entry;
 	uint64_t          user_data;
 	_odp_int_name_t   name_tbl_id;
@@ -64,9 +64,9 @@ struct name_tbl_entry_s {
 	uint8_t           name_kind;
 	uint8_t           name_len;
 	char              name[_ODP_INT_NAME_LEN + 1];
-} ODP_ALIGNED_CACHE;
+};
 
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	uint32_t          num_allocd;
 	uint32_t          num_used;
 	uint32_t          num_added_to_free_list;
@@ -74,7 +74,7 @@ typedef struct {
 	uint32_t          base_id;
 	name_tbl_entry_t *free_list_head;
 	name_tbl_entry_t  entries[0];
-} ODP_ALIGNED_CACHE name_tbl_t;
+} name_tbl_t;
 
 typedef struct {
 	name_tbl_t *tbls[NUM_NAME_TBLS];
diff --git a/platform/linux-generic/odp_packet.c b/platform/linux-generic/odp_packet.c
index bd9e77c0a..77765fb89 100644
--- a/platform/linux-generic/odp_packet.c
+++ b/platform/linux-generic/odp_packet.c
@@ -26,7 +26,7 @@
 #include <odp/visibility_begin.h>
 
 /* Fill in packet header field offsets for inline functions */
-const _odp_packet_inline_offset_t _odp_packet_inline ODP_ALIGNED_CACHE = {
+const _odp_packet_inline_offset_t ODP_ALIGNED_CACHE _odp_packet_inline = {
 	.data           = offsetof(odp_packet_hdr_t, buf_hdr.seg[0].data),
 	.seg_len        = offsetof(odp_packet_hdr_t, buf_hdr.seg[0].len),
 	.frame_len      = offsetof(odp_packet_hdr_t, frame_len),
diff --git a/platform/linux-generic/odp_pkt_queue.c b/platform/linux-generic/odp_pkt_queue.c
index 4f6a1eb61..172191931 100644
--- a/platform/linux-generic/odp_pkt_queue.c
+++ b/platform/linux-generic/odp_pkt_queue.c
@@ -22,15 +22,16 @@
 
 #define NUM_PKTS     7
 
-typedef struct /* Must be exactly 64 bytes long AND cacheline aligned! */ {
+/* Must be exactly 64 bytes long AND cacheline aligned! */
+typedef struct ODP_ALIGNED_CACHE {
 	uint32_t next_queue_blk_idx;
 	uint32_t tail_queue_blk_idx;
 	odp_packet_t pkts[NUM_PKTS];
-} ODP_ALIGNED_CACHE queue_blk_t;
+} queue_blk_t;
 
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	queue_blk_t blks[0];
-} ODP_ALIGNED_CACHE queue_blks_t;
+} queue_blks_t;
 
 /* The queue_num_tbl is used to map from a queue_num to a queue_num_desc.
  * The reason is based on the assumption that usually only a small fraction
diff --git a/platform/linux-generic/odp_pool.c b/platform/linux-generic/odp_pool.c
index 7e3e156f9..e5ba8982a 100644
--- a/platform/linux-generic/odp_pool.c
+++ b/platform/linux-generic/odp_pool.c
@@ -61,7 +61,7 @@ static __thread pool_local_t local;
 #include <odp/visibility_begin.h>
 
 /* Fill in pool header field offsets for inline functions */
-const _odp_pool_inline_offset_t _odp_pool_inline ODP_ALIGNED_CACHE = {
+const _odp_pool_inline_offset_t ODP_ALIGNED_CACHE _odp_pool_inline = {
 	.pool_hdl          = offsetof(pool_t, pool_hdl),
 	.uarea_size        = offsetof(pool_t, params.pkt.uarea_size)
 };
diff --git a/platform/linux-generic/odp_schedule.c b/platform/linux-generic/odp_schedule.c
index 1a9a2a8f1..118f9eb17 100644
--- a/platform/linux-generic/odp_schedule.c
+++ b/platform/linux-generic/odp_schedule.c
@@ -177,24 +177,24 @@ typedef struct {
 } sched_local_t;
 
 /* Priority queue */
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/* Ring header */
 	ring_t ring;
 
 	/* Ring data: queue indexes */
 	uint32_t queue_index[PRIO_QUEUE_RING_SIZE];
 
-} prio_queue_t ODP_ALIGNED_CACHE;
+} prio_queue_t;
 
 /* Packet IO queue */
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/* Ring header */
 	ring_t ring;
 
 	/* Ring data: pktio poll command indexes */
 	uint32_t cmd_index[PKTIO_RING_SIZE];
 
-} pktio_queue_t ODP_ALIGNED_CACHE;
+} pktio_queue_t;
 
 /* Packet IO poll command */
 typedef struct {
@@ -205,17 +205,17 @@ typedef struct {
 } pktio_cmd_t;
 
 /* Order context of a queue */
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/* Current ordered context id */
-	odp_atomic_u64_t  ctx ODP_ALIGNED_CACHE;
+	odp_atomic_u64_t ODP_ALIGNED_CACHE ctx;
 
 	/* Next unallocated context id */
-	odp_atomic_u64_t  next_ctx;
+	odp_atomic_u64_t next_ctx;
 
 	/* Array of ordered locks */
-	odp_atomic_u64_t  lock[CONFIG_QUEUE_MAX_ORD_LOCKS];
+	odp_atomic_u64_t lock[CONFIG_QUEUE_MAX_ORD_LOCKS];
 
-} order_context_t ODP_ALIGNED_CACHE;
+} order_context_t;
 
 typedef struct {
 	pri_mask_t     pri_mask[NUM_PRIO];
diff --git a/platform/linux-generic/odp_schedule_iquery.c b/platform/linux-generic/odp_schedule_iquery.c
index 3d794b140..33767c1f8 100644
--- a/platform/linux-generic/odp_schedule_iquery.c
+++ b/platform/linux-generic/odp_schedule_iquery.c
@@ -93,13 +93,13 @@ typedef struct {
 #define MAX_DEQ CONFIG_BURST_SIZE
 
 /* Instantiate a RING data structure as pktio command queue */
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/* Ring header */
 	ring_t ring;
 
 	/* Ring data: pktio poll command indexes */
 	uint32_t cmd_index[PKTIO_RING_SIZE];
-} pktio_cmd_queue_t ODP_ALIGNED_CACHE;
+} pktio_cmd_queue_t;
 
 /* Packet IO poll command */
 typedef struct {
@@ -122,17 +122,17 @@ typedef struct {
 typedef struct sched_thread_local sched_thread_local_t;
 
 /* Order context of a queue */
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/* Current ordered context id */
-	odp_atomic_u64_t  ctx ODP_ALIGNED_CACHE;
+	odp_atomic_u64_t ODP_ALIGNED_CACHE ctx;
 
 	/* Next unallocated context id */
-	odp_atomic_u64_t  next_ctx;
+	odp_atomic_u64_t next_ctx;
 
 	/* Array of ordered locks */
-	odp_atomic_u64_t  lock[CONFIG_QUEUE_MAX_ORD_LOCKS];
+	odp_atomic_u64_t lock[CONFIG_QUEUE_MAX_ORD_LOCKS];
 
-} order_context_t ODP_ALIGNED_CACHE;
+} order_context_t;
 
 typedef struct {
 	odp_shm_t selfie;
diff --git a/platform/linux-generic/odp_schedule_sp.c b/platform/linux-generic/odp_schedule_sp.c
index 7f0404b1c..017eaee34 100644
--- a/platform/linux-generic/odp_schedule_sp.c
+++ b/platform/linux-generic/odp_schedule_sp.c
@@ -61,20 +61,20 @@ struct sched_cmd_s {
 	int                pktin_idx[NUM_PKTIN];
 };
 
-typedef struct sched_cmd_t {
+typedef struct ODP_ALIGNED_CACHE sched_cmd_t {
 	struct sched_cmd_s s;
 	uint8_t            pad[ROUNDUP_CACHE_LINE(sizeof(struct sched_cmd_s)) -
 			       sizeof(struct sched_cmd_s)];
-} sched_cmd_t ODP_ALIGNED_CACHE;
+} sched_cmd_t;
 
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	/* Ring header */
 	ring_t ring;
 
 	/* Ring data: queue indexes */
 	uint32_t ring_idx[RING_SIZE];
 
-} prio_queue_t ODP_ALIGNED_CACHE;
+} prio_queue_t;
 
 typedef struct thr_group_t {
 	/* A generation counter for fast comparison if groups have changed */
@@ -88,7 +88,7 @@ typedef struct thr_group_t {
 
 } thr_group_t;
 
-typedef struct sched_group_t {
+typedef struct ODP_ALIGNED_CACHE sched_group_t {
 	struct {
 		odp_ticketlock_t  lock;
 
@@ -104,7 +104,7 @@ typedef struct sched_group_t {
 
 	} s;
 
-} sched_group_t ODP_ALIGNED_CACHE;
+} sched_group_t;
 
 typedef struct {
 	sched_cmd_t   queue_cmd[NUM_QUEUE];
diff --git a/platform/linux-generic/odp_timer.c b/platform/linux-generic/odp_timer.c
index 72f9505dd..bdb25c60e 100644
--- a/platform/linux-generic/odp_timer.c
+++ b/platform/linux-generic/odp_timer.c
@@ -98,7 +98,11 @@ static odp_timeout_hdr_t *timeout_hdr(odp_timeout_t tmo)
  * odp_timer abstract datatype
  *****************************************************************************/
 
-typedef struct tick_buf_s {
+typedef struct
+#ifdef ODP_ATOMIC_U128
+ODP_ALIGNED(16) /* 16-byte atomic operations need properly aligned addresses */
+#endif
+tick_buf_s {
 #if __GCC_ATOMIC_LLONG_LOCK_FREE < 2
 	/* No atomics support for 64-bit variables, will use separate lock */
 	/* Use the same layout as odp_atomic_u64_t but without lock variable */
@@ -112,11 +116,7 @@ typedef struct tick_buf_s {
 #ifdef TB_NEEDS_PAD
 	uint32_t pad;/* Need to be able to access padding for successful CAS */
 #endif
-} tick_buf_t
-#ifdef ODP_ATOMIC_U128
-ODP_ALIGNED(16) /* 16-byte atomic operations need properly aligned addresses */
-#endif
-;
+} tick_buf_t;
 
 #if __GCC_ATOMIC_LLONG_LOCK_FREE >= 2
 /* Only assert this when we perform atomic operations on tick_buf_t */
diff --git a/platform/linux-generic/pktio/socket_mmap.c b/platform/linux-generic/pktio/socket_mmap.c
index ac764b3af..b83486778 100644
--- a/platform/linux-generic/pktio/socket_mmap.c
+++ b/platform/linux-generic/pktio/socket_mmap.c
@@ -61,9 +61,10 @@ static int set_pkt_sock_fanout_mmap(pkt_sock_mmap_t *const pkt_sock,
 
 union frame_map {
 	struct {
-		struct tpacket2_hdr tp_h ODP_ALIGNED(TPACKET_ALIGNMENT);
-		struct sockaddr_ll s_ll
-		ODP_ALIGNED(TPACKET_ALIGN(sizeof(struct tpacket2_hdr)));
+		struct tpacket2_hdr ODP_ALIGNED(TPACKET_ALIGNMENT) tp_h;
+		struct sockaddr_ll
+			ODP_ALIGNED(TPACKET_ALIGN(sizeof(struct tpacket2_hdr)))
+			s_ll;
 	} *v2;
 
 	void *raw;
diff --git a/test/performance/odp_l2fwd.c b/test/performance/odp_l2fwd.c
index 2353e2414..09a12b050 100644
--- a/test/performance/odp_l2fwd.c
+++ b/test/performance/odp_l2fwd.c
@@ -123,7 +123,7 @@ static void sig_handler(int signo ODP_UNUSED)
 /**
  * Statistics
  */
-typedef union {
+typedef union ODP_ALIGNED_CACHE {
 	struct {
 		/** Number of forwarded packets */
 		uint64_t packets;
@@ -134,7 +134,7 @@ typedef union {
 	} s;
 
 	uint8_t padding[ODP_CACHE_LINE_SIZE];
-} stats_t ODP_ALIGNED_CACHE;
+} stats_t;
 
 /**
  * Thread specific data
diff --git a/test/performance/odp_pktio_ordered.c b/test/performance/odp_pktio_ordered.c
index 2248eb2fd..db9bfe764 100644
--- a/test/performance/odp_pktio_ordered.c
+++ b/test/performance/odp_pktio_ordered.c
@@ -179,7 +179,7 @@ ODP_STATIC_ASSERT(sizeof(flow_t) <= PKT_UAREA_SIZE,
 /**
  * Statistics
  */
-typedef union {
+typedef union ODP_ALIGNED_CACHE {
 	struct {
 		/** Number of forwarded packets */
 		uint64_t packets;
@@ -192,7 +192,7 @@ typedef union {
 	} s;
 
 	uint8_t padding[ODP_CACHE_LINE_SIZE];
-} stats_t ODP_ALIGNED_CACHE;
+} stats_t;
 
 /**
  * IPv4 5-tuple
diff --git a/test/performance/odp_sched_latency.c b/test/performance/odp_sched_latency.c
index fc2860518..ae7a5914d 100644
--- a/test/performance/odp_sched_latency.c
+++ b/test/performance/odp_sched_latency.c
@@ -97,11 +97,11 @@ typedef struct {
 } test_stat_t;
 
 /** Performance test statistics (per core) */
-typedef union {
+typedef union ODP_ALIGNED_CACHE {
 	test_stat_t prio[NUM_PRIOS]; /**< Test statistics per priority */
 
 	uint8_t pad[CACHE_ALIGN_ROUNDUP(NUM_PRIOS * sizeof(test_stat_t))];
-} core_stat_t ODP_ALIGNED_CACHE;
+} core_stat_t;
 
 /** Test global variables */
 typedef struct {
diff --git a/test/performance/odp_scheduling.c b/test/performance/odp_scheduling.c
index 03078a5b3..8d2574b73 100644
--- a/test/performance/odp_scheduling.c
+++ b/test/performance/odp_scheduling.c
@@ -55,12 +55,12 @@ typedef struct {
 	int fairness;   /**< Check fairness */
 } test_args_t;
 
-typedef struct {
+typedef struct ODP_ALIGNED_CACHE {
 	uint64_t num_ev;
 
 	/* Round up the struct size to cache line size */
 	uint8_t pad[ODP_CACHE_LINE_SIZE - sizeof(uint64_t)];
-} queue_context_t ODP_ALIGNED_CACHE;
+} queue_context_t;
 
 /** Test global variables */
 typedef struct {
diff --git a/test/validation/api/chksum/chksum.c b/test/validation/api/chksum/chksum.c
index de98b74a0..f8687955e 100644
--- a/test/validation/api/chksum/chksum.c
+++ b/test/validation/api/chksum/chksum.c
@@ -14,7 +14,7 @@
 #define NUM_UDP      4
 #define MAX_UDP_LEN  128
 
-static uint8_t ip_hdr_test_vect[NUM_IP_HDR][IP_HDR_LEN] ODP_ALIGNED(4) = {
+static uint8_t ODP_ALIGNED(4) ip_hdr_test_vect[NUM_IP_HDR][IP_HDR_LEN] = {
 	{	0x45, 0x00, 0x00, 0x2E, 0x00, 0x00, 0x00, 0x00, 0x40, 0x11,
 		0xAB, 0x33, 0xC0, 0xA8, 0x2C, 0xA2, 0xC0, 0xA8, 0x21, 0x99
 	},
@@ -38,7 +38,7 @@ struct udp_test_vect_s {
 	uint8_t data[MAX_UDP_LEN];
 };
 
-static struct udp_test_vect_s udp_test_vect[NUM_UDP] ODP_ALIGNED(4) = {
+static struct udp_test_vect_s ODP_ALIGNED(4) udp_test_vect[NUM_UDP] = {
 	{.len = 38,
 	.data = { 0x00, 0x11, 0x00, 0x1A, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
 		  0x00, 0x00, 0x00, 0x3F, 0x00, 0x3F, 0x00, 0x1A, 0xFF, 0x3C,
@@ -86,7 +86,7 @@ static struct udp_test_vect_s udp_test_vect[NUM_UDP] ODP_ALIGNED(4) = {
 /* Long UDP packet with pseudo header. Checksum field is set to zero.
  * The array contains padding, so that a possible overrun is more likely
  * detected (overrun bytes are not all zeros). */
-static uint8_t udp_test_vect_long[] ODP_ALIGNED(4) = {
+static uint8_t ODP_ALIGNED(4) udp_test_vect_long[] = {
 	0x00, 0x11, 0x05, 0xED, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
 	0x00, 0x00, 0x00, 0x3F, 0x00, 0x3F, 0x05, 0xED, 0x00, 0x00,
 	0x0B, 0x2C, 0x9C, 0x06, 0x07, 0xF3, 0x51, 0x05, 0xF7, 0xA7,