From: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Date: Sat, 13 May 2017 09:27:25 +0000 (+0530)
Subject: eal: introduce macro for always inline
X-Git-Tag: spdx-start~3186
X-Git-Url: http://git.droids-corp.org/?a=commitdiff_plain;h=c0583d98a9153549383d236e5cc5464bbbe8dd89;p=dpdk.git

eal: introduce macro for always inline

Different drivers use internal macros like force_inline for compiler
always inline feature.
Standardizing it through __rte_always_inline macro.

Verified the change by comparing the output binary file.
No difference found in the output binary file with this change.

Signed-off-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>
Acked-by: Bruce Richardson <bruce.richardson@intel.com>
---

diff --git a/drivers/crypto/dpaa2_sec/hw/compat.h b/drivers/crypto/dpaa2_sec/hw/compat.h
index 11fdaa8e37..ab95ce6bbd 100644
--- a/drivers/crypto/dpaa2_sec/hw/compat.h
+++ b/drivers/crypto/dpaa2_sec/hw/compat.h
@@ -49,7 +49,9 @@
 #include <stdlib.h>
 #include <stdio.h>
 #include <stdbool.h>
+
 #include <rte_byteorder.h>
+#include <rte_common.h>
 
 #ifndef __BYTE_ORDER__
 #error "Undefined endianness"
@@ -60,7 +62,7 @@
 #endif
 
 #ifndef __always_inline
-#define __always_inline (inline __attribute__((always_inline)))
+#define __always_inline __rte_always_inline
 #endif
 
 #ifndef __always_unused
diff --git a/drivers/crypto/scheduler/scheduler_failover.c b/drivers/crypto/scheduler/scheduler_failover.c
index 2471a5f14f..162a29bb67 100644
--- a/drivers/crypto/scheduler/scheduler_failover.c
+++ b/drivers/crypto/scheduler/scheduler_failover.c
@@ -48,7 +48,7 @@ struct fo_scheduler_qp_ctx {
 	uint8_t deq_idx;
 };
 
-static inline uint16_t __attribute__((always_inline))
+static __rte_always_inline uint16_t
 failover_slave_enqueue(struct scheduler_slave *slave, uint8_t slave_idx,
 		struct rte_crypto_op **ops, uint16_t nb_ops)
 {
diff --git a/drivers/crypto/scheduler/scheduler_pmd_private.h b/drivers/crypto/scheduler/scheduler_pmd_private.h
index 421dae3717..05a5916c37 100644
--- a/drivers/crypto/scheduler/scheduler_pmd_private.h
+++ b/drivers/crypto/scheduler/scheduler_pmd_private.h
@@ -105,7 +105,7 @@ struct scheduler_session {
 			RTE_CRYPTODEV_SCHEDULER_MAX_NB_SLAVES];
 };
 
-static inline uint16_t __attribute__((always_inline))
+static __rte_always_inline uint16_t
 get_max_enqueue_order_count(struct rte_ring *order_ring, uint16_t nb_ops)
 {
 	uint32_t count = rte_ring_free_count(order_ring);
@@ -113,7 +113,7 @@ get_max_enqueue_order_count(struct rte_ring *order_ring, uint16_t nb_ops)
 	return count > nb_ops ? nb_ops : count;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 scheduler_order_insert(struct rte_ring *order_ring,
 		struct rte_crypto_op **ops, uint16_t nb_ops)
 {
@@ -125,7 +125,7 @@ scheduler_order_insert(struct rte_ring *order_ring,
 	op = ring[(order_ring->cons.head + pos) & order_ring->mask]; \
 } while (0)
 
-static inline uint16_t __attribute__((always_inline))
+static __rte_always_inline uint16_t
 scheduler_order_drain(struct rte_ring *order_ring,
 		struct rte_crypto_op **ops, uint16_t nb_ops)
 {
diff --git a/drivers/event/octeontx/ssovf_worker.c b/drivers/event/octeontx/ssovf_worker.c
index ad3fe684d9..fcb5f316cc 100644
--- a/drivers/event/octeontx/ssovf_worker.c
+++ b/drivers/event/octeontx/ssovf_worker.c
@@ -32,7 +32,7 @@
 
 #include "ssovf_worker.h"
 
-static force_inline void
+static __rte_always_inline void
 ssows_new_event(struct ssows *ws, const struct rte_event *ev)
 {
 	const uint64_t event_ptr = ev->u64;
@@ -43,7 +43,7 @@ ssows_new_event(struct ssows *ws, const struct rte_event *ev)
 	ssows_add_work(ws, event_ptr, tag, new_tt, grp);
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_fwd_swtag(struct ssows *ws, const struct rte_event *ev, const uint8_t grp)
 {
 	const uint8_t cur_tt = ws->cur_tt;
@@ -72,7 +72,7 @@ ssows_fwd_swtag(struct ssows *ws, const struct rte_event *ev, const uint8_t grp)
 
 #define OCT_EVENT_TYPE_GRP_FWD (RTE_EVENT_TYPE_MAX - 1)
 
-static force_inline void
+static __rte_always_inline void
 ssows_fwd_group(struct ssows *ws, const struct rte_event *ev, const uint8_t grp)
 {
 	const uint64_t event_ptr = ev->u64;
@@ -95,7 +95,7 @@ ssows_fwd_group(struct ssows *ws, const struct rte_event *ev, const uint8_t grp)
 	ssows_add_work(ws, event_ptr, tag, new_tt, grp);
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_forward_event(struct ssows *ws, const struct rte_event *ev)
 {
 	const uint8_t grp = ev->queue_id;
@@ -112,14 +112,14 @@ ssows_forward_event(struct ssows *ws, const struct rte_event *ev)
 		ssows_fwd_group(ws, ev, grp);
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_release_event(struct ssows *ws)
 {
 	if (likely(ws->cur_tt != SSO_SYNC_UNTAGGED))
 		ssows_swtag_untag(ws);
 }
 
-force_inline uint16_t __hot
+__rte_always_inline uint16_t __hot
 ssows_deq(void *port, struct rte_event *ev, uint64_t timeout_ticks)
 {
 	struct ssows *ws = port;
@@ -135,7 +135,7 @@ ssows_deq(void *port, struct rte_event *ev, uint64_t timeout_ticks)
 	}
 }
 
-force_inline uint16_t __hot
+__rte_always_inline uint16_t __hot
 ssows_deq_timeout(void *port, struct rte_event *ev, uint64_t timeout_ticks)
 {
 	struct ssows *ws = port;
@@ -171,7 +171,7 @@ ssows_deq_timeout_burst(void *port, struct rte_event ev[], uint16_t nb_events,
 	return ssows_deq_timeout(port, ev, timeout_ticks);
 }
 
-force_inline uint16_t __hot
+__rte_always_inline uint16_t __hot
 ssows_enq(void *port, const struct rte_event *ev)
 {
 	struct ssows *ws = port;
diff --git a/drivers/event/octeontx/ssovf_worker.h b/drivers/event/octeontx/ssovf_worker.h
index 300dfae839..40c5c5531f 100644
--- a/drivers/event/octeontx/ssovf_worker.h
+++ b/drivers/event/octeontx/ssovf_worker.h
@@ -42,17 +42,13 @@ enum {
 	SSO_SYNC_EMPTY
 };
 
-#ifndef force_inline
-#define force_inline inline __attribute__((always_inline))
-#endif
-
 #ifndef __hot
 #define __hot	__attribute__((hot))
 #endif
 
 /* SSO Operations */
 
-static force_inline uint16_t
+static __rte_always_inline uint16_t
 ssows_get_work(struct ssows *ws, struct rte_event *ev)
 {
 	uint64_t get_work0, get_work1;
@@ -70,7 +66,7 @@ ssows_get_work(struct ssows *ws, struct rte_event *ev)
 	return !!get_work1;
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_add_work(struct ssows *ws, const uint64_t event_ptr, const uint32_t tag,
 			const uint8_t new_tt, const uint8_t grp)
 {
@@ -80,7 +76,7 @@ ssows_add_work(struct ssows *ws, const uint64_t event_ptr, const uint32_t tag,
 	ssovf_store_pair(add_work0, event_ptr, ws->grps[grp]);
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_swtag_full(struct ssows *ws, const uint64_t event_ptr, const uint32_t tag,
 			const uint8_t new_tt, const uint8_t grp)
 {
@@ -92,7 +88,7 @@ ssows_swtag_full(struct ssows *ws, const uint64_t event_ptr, const uint32_t tag,
 				SSOW_VHWS_OP_SWTAG_FULL0));
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_swtag_desched(struct ssows *ws, uint32_t tag, uint8_t new_tt, uint8_t grp)
 {
 	uint64_t val;
@@ -101,7 +97,7 @@ ssows_swtag_desched(struct ssows *ws, uint32_t tag, uint8_t new_tt, uint8_t grp)
 	ssovf_write64(val, ws->base + SSOW_VHWS_OP_SWTAG_DESCHED);
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_swtag_norm(struct ssows *ws, uint32_t tag, uint8_t new_tt)
 {
 	uint64_t val;
@@ -110,27 +106,27 @@ ssows_swtag_norm(struct ssows *ws, uint32_t tag, uint8_t new_tt)
 	ssovf_write64(val, ws->base + SSOW_VHWS_OP_SWTAG_NORM);
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_swtag_untag(struct ssows *ws)
 {
 	ssovf_write64(0, ws->base + SSOW_VHWS_OP_SWTAG_UNTAG);
 	ws->cur_tt = SSO_SYNC_UNTAGGED;
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_upd_wqp(struct ssows *ws, uint8_t grp, uint64_t event_ptr)
 {
 	ssovf_store_pair((uint64_t)grp << 34, event_ptr, (ws->base +
 				SSOW_VHWS_OP_UPD_WQP_GRP0));
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_desched(struct ssows *ws)
 {
 	ssovf_write64(0, ws->base + SSOW_VHWS_OP_DESCHED);
 }
 
-static force_inline void
+static __rte_always_inline void
 ssows_swtag_wait(struct ssows *ws)
 {
 	/* Wait for the SWTAG/SWTAG_FULL operation */
diff --git a/drivers/event/sw/event_ring.h b/drivers/event/sw/event_ring.h
index cdaee95d33..734a3b4b17 100644
--- a/drivers/event/sw/event_ring.h
+++ b/drivers/event/sw/event_ring.h
@@ -61,10 +61,6 @@ struct qe_ring {
 	struct rte_event ring[0] __rte_cache_aligned;
 };
 
-#ifndef force_inline
-#define force_inline inline __attribute__((always_inline))
-#endif
-
 static inline struct qe_ring *
 qe_ring_create(const char *name, unsigned int size, unsigned int socket_id)
 {
@@ -91,19 +87,19 @@ qe_ring_destroy(struct qe_ring *r)
 	rte_free(r);
 }
 
-static force_inline unsigned int
+static __rte_always_inline unsigned int
 qe_ring_count(const struct qe_ring *r)
 {
 	return r->write_idx - r->read_idx;
 }
 
-static force_inline unsigned int
+static __rte_always_inline unsigned int
 qe_ring_free_count(const struct qe_ring *r)
 {
 	return r->size - qe_ring_count(r);
 }
 
-static force_inline unsigned int
+static __rte_always_inline unsigned int
 qe_ring_enqueue_burst(struct qe_ring *r, const struct rte_event *qes,
 		unsigned int nb_qes, uint16_t *free_count)
 {
@@ -130,7 +126,7 @@ qe_ring_enqueue_burst(struct qe_ring *r, const struct rte_event *qes,
 	return nb_qes;
 }
 
-static force_inline unsigned int
+static __rte_always_inline unsigned int
 qe_ring_enqueue_burst_with_ops(struct qe_ring *r, const struct rte_event *qes,
 		unsigned int nb_qes, uint8_t *ops)
 {
@@ -157,7 +153,7 @@ qe_ring_enqueue_burst_with_ops(struct qe_ring *r, const struct rte_event *qes,
 	return nb_qes;
 }
 
-static force_inline unsigned int
+static __rte_always_inline unsigned int
 qe_ring_dequeue_burst(struct qe_ring *r, struct rte_event *qes,
 		unsigned int nb_qes)
 {
diff --git a/drivers/event/sw/iq_ring.h b/drivers/event/sw/iq_ring.h
index d480d1560f..64cf6784ca 100644
--- a/drivers/event/sw/iq_ring.h
+++ b/drivers/event/sw/iq_ring.h
@@ -56,10 +56,6 @@ struct iq_ring {
 	struct rte_event ring[QID_IQ_DEPTH];
 };
 
-#ifndef force_inline
-#define force_inline inline __attribute__((always_inline))
-#endif
-
 static inline struct iq_ring *
 iq_ring_create(const char *name, unsigned int socket_id)
 {
@@ -81,19 +77,19 @@ iq_ring_destroy(struct iq_ring *r)
 	rte_free(r);
 }
 
-static force_inline uint16_t
+static __rte_always_inline uint16_t
 iq_ring_count(const struct iq_ring *r)
 {
 	return r->write_idx - r->read_idx;
 }
 
-static force_inline uint16_t
+static __rte_always_inline uint16_t
 iq_ring_free_count(const struct iq_ring *r)
 {
 	return QID_IQ_MASK - iq_ring_count(r);
 }
 
-static force_inline uint16_t
+static __rte_always_inline uint16_t
 iq_ring_enqueue_burst(struct iq_ring *r, struct rte_event *qes, uint16_t nb_qes)
 {
 	const uint16_t read = r->read_idx;
@@ -112,7 +108,7 @@ iq_ring_enqueue_burst(struct iq_ring *r, struct rte_event *qes, uint16_t nb_qes)
 	return nb_qes;
 }
 
-static force_inline uint16_t
+static __rte_always_inline uint16_t
 iq_ring_dequeue_burst(struct iq_ring *r, struct rte_event *qes, uint16_t nb_qes)
 {
 	uint16_t read = r->read_idx;
@@ -132,7 +128,7 @@ iq_ring_dequeue_burst(struct iq_ring *r, struct rte_event *qes, uint16_t nb_qes)
 }
 
 /* assumes there is space, from a previous dequeue_burst */
-static force_inline uint16_t
+static __rte_always_inline uint16_t
 iq_ring_put_back(struct iq_ring *r, struct rte_event *qes, uint16_t nb_qes)
 {
 	uint16_t i, read = r->read_idx;
@@ -144,19 +140,19 @@ iq_ring_put_back(struct iq_ring *r, struct rte_event *qes, uint16_t nb_qes)
 	return nb_qes;
 }
 
-static force_inline const struct rte_event *
+static __rte_always_inline const struct rte_event *
 iq_ring_peek(const struct iq_ring *r)
 {
 	return &r->ring[r->read_idx & QID_IQ_MASK];
 }
 
-static force_inline void
+static __rte_always_inline void
 iq_ring_pop(struct iq_ring *r)
 {
 	r->read_idx++;
 }
 
-static force_inline int
+static __rte_always_inline int
 iq_ring_enqueue(struct iq_ring *r, const struct rte_event *qe)
 {
 	const uint16_t read = r->read_idx;
diff --git a/drivers/event/sw/sw_evdev_scheduler.c b/drivers/event/sw/sw_evdev_scheduler.c
index a333a6f0a0..35f8f175ad 100644
--- a/drivers/event/sw/sw_evdev_scheduler.c
+++ b/drivers/event/sw/sw_evdev_scheduler.c
@@ -362,7 +362,7 @@ sw_schedule_reorder(struct sw_evdev *sw, int qid_start, int qid_end)
 	return pkts_iter;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 sw_refill_pp_buf(struct sw_evdev *sw, struct sw_port *port)
 {
 	RTE_SET_USED(sw);
@@ -372,7 +372,7 @@ sw_refill_pp_buf(struct sw_evdev *sw, struct sw_port *port)
 			RTE_DIM(port->pp_buf));
 }
 
-static inline uint32_t __attribute__((always_inline))
+static __rte_always_inline uint32_t
 __pull_port_lb(struct sw_evdev *sw, uint32_t port_id, int allow_reorder)
 {
 	static struct reorder_buffer_entry dummy_rob;
diff --git a/drivers/net/fm10k/fm10k_rxtx_vec.c b/drivers/net/fm10k/fm10k_rxtx_vec.c
index 411bc44503..03f6fd70e6 100644
--- a/drivers/net/fm10k/fm10k_rxtx_vec.c
+++ b/drivers/net/fm10k/fm10k_rxtx_vec.c
@@ -738,7 +738,7 @@ vtx(volatile struct fm10k_tx_desc *txdp,
 		vtx1(txdp, *pkt, flags);
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 fm10k_tx_free_bufs(struct fm10k_tx_queue *txq)
 {
 	struct rte_mbuf **txep;
@@ -794,7 +794,7 @@ fm10k_tx_free_bufs(struct fm10k_tx_queue *txq)
 	return txq->rs_thresh;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 tx_backlog_entry(struct rte_mbuf **txep,
 		 struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 {
diff --git a/drivers/net/i40e/i40e_rxtx.c b/drivers/net/i40e/i40e_rxtx.c
index 351cb94dd0..0aefb2f468 100644
--- a/drivers/net/i40e/i40e_rxtx.c
+++ b/drivers/net/i40e/i40e_rxtx.c
@@ -1257,7 +1257,7 @@ end_of_tx:
 	return nb_tx;
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 i40e_tx_free_bufs(struct i40e_tx_queue *txq)
 {
 	struct i40e_tx_entry *txep;
diff --git a/drivers/net/i40e/i40e_rxtx_vec_common.h b/drivers/net/i40e/i40e_rxtx_vec_common.h
index 692096684b..39a6da061c 100644
--- a/drivers/net/i40e/i40e_rxtx_vec_common.h
+++ b/drivers/net/i40e/i40e_rxtx_vec_common.h
@@ -102,7 +102,7 @@ reassemble_packets(struct i40e_rx_queue *rxq, struct rte_mbuf **rx_bufs,
 	return pkt_idx;
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 i40e_tx_free_bufs(struct i40e_tx_queue *txq)
 {
 	struct i40e_tx_entry *txep;
@@ -159,7 +159,7 @@ i40e_tx_free_bufs(struct i40e_tx_queue *txq)
 	return txq->tx_rs_thresh;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 tx_backlog_entry(struct i40e_tx_entry *txep,
 		 struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 {
diff --git a/drivers/net/ixgbe/ixgbe_rxtx.c b/drivers/net/ixgbe/ixgbe_rxtx.c
index 1e0789595f..ee8ad9626b 100644
--- a/drivers/net/ixgbe/ixgbe_rxtx.c
+++ b/drivers/net/ixgbe/ixgbe_rxtx.c
@@ -126,7 +126,7 @@ uint16_t ixgbe_xmit_fixed_burst_vec(void *tx_queue, struct rte_mbuf **tx_pkts,
  * Check for descriptors with their DD bit set and free mbufs.
  * Return the total number of buffers freed.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 ixgbe_tx_free_bufs(struct ixgbe_tx_queue *txq)
 {
 	struct ixgbe_tx_entry *txep;
diff --git a/drivers/net/ixgbe/ixgbe_rxtx_vec_common.h b/drivers/net/ixgbe/ixgbe_rxtx_vec_common.h
index 1c34bb5f3e..9fc112b1c3 100644
--- a/drivers/net/ixgbe/ixgbe_rxtx_vec_common.h
+++ b/drivers/net/ixgbe/ixgbe_rxtx_vec_common.h
@@ -101,7 +101,7 @@ reassemble_packets(struct ixgbe_rx_queue *rxq, struct rte_mbuf **rx_bufs,
 	return pkt_idx;
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 ixgbe_tx_free_bufs(struct ixgbe_tx_queue *txq)
 {
 	struct ixgbe_tx_entry_v *txep;
@@ -158,7 +158,7 @@ ixgbe_tx_free_bufs(struct ixgbe_tx_queue *txq)
 	return txq->tx_rs_thresh;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 tx_backlog_entry(struct ixgbe_tx_entry_v *txep,
 		 struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
 {
diff --git a/drivers/net/mlx5/mlx5_rxtx.c b/drivers/net/mlx5/mlx5_rxtx.c
index de6e0fa4a5..53b5c68bdd 100644
--- a/drivers/net/mlx5/mlx5_rxtx.c
+++ b/drivers/net/mlx5/mlx5_rxtx.c
@@ -69,34 +69,28 @@
 #include "mlx5_defs.h"
 #include "mlx5_prm.h"
 
-static inline int
+static __rte_always_inline int
 check_cqe(volatile struct mlx5_cqe *cqe,
-	  unsigned int cqes_n, const uint16_t ci)
-	  __attribute__((always_inline));
+	  unsigned int cqes_n, const uint16_t ci);
 
-static inline void
-txq_complete(struct txq *txq) __attribute__((always_inline));
+static __rte_always_inline void
+txq_complete(struct txq *txq);
 
-static inline uint32_t
-txq_mp2mr(struct txq *txq, struct rte_mempool *mp)
-	__attribute__((always_inline));
+static __rte_always_inline uint32_t
+txq_mp2mr(struct txq *txq, struct rte_mempool *mp);
 
-static inline void
-mlx5_tx_dbrec(struct txq *txq, volatile struct mlx5_wqe *wqe)
-	__attribute__((always_inline));
+static __rte_always_inline void
+mlx5_tx_dbrec(struct txq *txq, volatile struct mlx5_wqe *wqe);
 
-static inline uint32_t
-rxq_cq_to_pkt_type(volatile struct mlx5_cqe *cqe)
-	__attribute__((always_inline));
+static __rte_always_inline uint32_t
+rxq_cq_to_pkt_type(volatile struct mlx5_cqe *cqe);
 
-static inline int
+static __rte_always_inline int
 mlx5_rx_poll_len(struct rxq *rxq, volatile struct mlx5_cqe *cqe,
-		 uint16_t cqe_cnt, uint32_t *rss_hash)
-		 __attribute__((always_inline));
+		 uint16_t cqe_cnt, uint32_t *rss_hash);
 
-static inline uint32_t
-rxq_cq_to_ol_flags(struct rxq *rxq, volatile struct mlx5_cqe *cqe)
-		   __attribute__((always_inline));
+static __rte_always_inline uint32_t
+rxq_cq_to_ol_flags(struct rxq *rxq, volatile struct mlx5_cqe *cqe);
 
 #ifndef NDEBUG
 
diff --git a/drivers/net/xenvirt/virtqueue.h b/drivers/net/xenvirt/virtqueue.h
index 350eae3ec7..1bb6877cd1 100644
--- a/drivers/net/xenvirt/virtqueue.h
+++ b/drivers/net/xenvirt/virtqueue.h
@@ -123,7 +123,7 @@ void virtqueue_dump(struct virtqueue *vq);
  */
 struct rte_mbuf * virtqueue_detatch_unused(struct virtqueue *vq);
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 virtqueue_full(const struct virtqueue *vq)
 {
 	return vq->vq_free_cnt == 0;
@@ -131,7 +131,7 @@ virtqueue_full(const struct virtqueue *vq)
 
 #define VIRTQUEUE_NUSED(vq) ((uint16_t)((vq)->vq_ring.used->idx - (vq)->vq_used_cons_idx))
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 vq_ring_update_avail(struct virtqueue *vq, uint16_t desc_idx)
 {
 	uint16_t avail_idx;
@@ -148,7 +148,7 @@ vq_ring_update_avail(struct virtqueue *vq, uint16_t desc_idx)
 	vq->vq_ring.avail->idx++;
 }
 
-static inline void  __attribute__((always_inline))
+static __rte_always_inline void
 vq_ring_free_chain(struct virtqueue *vq, uint16_t desc_idx)
 {
 	struct vring_desc *dp;
@@ -171,7 +171,7 @@ vq_ring_free_chain(struct virtqueue *vq, uint16_t desc_idx)
 	vq->vq_desc_head_idx = desc_idx;
 }
 
-static inline int  __attribute__((always_inline))
+static __rte_always_inline int
 virtqueue_enqueue_recv_refill(struct virtqueue *rxvq, struct rte_mbuf *cookie)
 {
 	const uint16_t needed = 1;
@@ -201,7 +201,7 @@ virtqueue_enqueue_recv_refill(struct virtqueue *rxvq, struct rte_mbuf *cookie)
 	return 0;
 }
 
-static inline int  __attribute__((always_inline))
+static __rte_always_inline int
 virtqueue_enqueue_xmit(struct virtqueue *txvq, struct rte_mbuf *cookie)
 {
 
@@ -242,7 +242,7 @@ virtqueue_enqueue_xmit(struct virtqueue *txvq, struct rte_mbuf *cookie)
 	return 0;
 }
 
-static inline uint16_t  __attribute__((always_inline))
+static __rte_always_inline uint16_t
 virtqueue_dequeue_burst(struct virtqueue *vq, struct rte_mbuf **rx_pkts, uint32_t *len, uint16_t num)
 {
 	struct vring_used_elem *uep;
diff --git a/examples/ip_pipeline/pipeline/pipeline_passthrough_be.c b/examples/ip_pipeline/pipeline/pipeline_passthrough_be.c
index 7ab0afedbe..8cb2f0c716 100644
--- a/examples/ip_pipeline/pipeline/pipeline_passthrough_be.c
+++ b/examples/ip_pipeline/pipeline/pipeline_passthrough_be.c
@@ -76,7 +76,7 @@ static pipeline_msg_req_handler handlers[] = {
 		pipeline_msg_req_invalid_handler,
 };
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 pkt_work_dma(
 	struct rte_mbuf *pkt,
 	void *arg,
@@ -121,7 +121,7 @@ pkt_work_dma(
 	}
 }
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 pkt4_work_dma(
 	struct rte_mbuf **pkts,
 	void *arg,
@@ -217,7 +217,7 @@ pkt4_work_dma(
 	}
 }
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 pkt_work_swap(
 	struct rte_mbuf *pkt,
 	void *arg)
@@ -241,7 +241,7 @@ pkt_work_swap(
 	}
 }
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 pkt4_work_swap(
 	struct rte_mbuf **pkts,
 	void *arg)
diff --git a/examples/ip_pipeline/pipeline/pipeline_routing_be.c b/examples/ip_pipeline/pipeline/pipeline_routing_be.c
index 21ac7888fb..78317165d4 100644
--- a/examples/ip_pipeline/pipeline/pipeline_routing_be.c
+++ b/examples/ip_pipeline/pipeline/pipeline_routing_be.c
@@ -191,7 +191,7 @@ struct layout {
 	dst->c = src->c;					\
 }
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 pkt_work_routing(
 	struct rte_mbuf *pkt,
 	struct rte_pipeline_table_entry *table_entry,
@@ -317,7 +317,7 @@ pkt_work_routing(
 	}
 }
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 pkt4_work_routing(
 	struct rte_mbuf **pkts,
 	struct rte_pipeline_table_entry **table_entries,
diff --git a/examples/l3fwd/l3fwd_em.h b/examples/l3fwd/l3fwd_em.h
index 2284bbd5c6..d509a1fcdc 100644
--- a/examples/l3fwd/l3fwd_em.h
+++ b/examples/l3fwd/l3fwd_em.h
@@ -34,7 +34,7 @@
 #ifndef __L3FWD_EM_H__
 #define __L3FWD_EM_H__
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 l3fwd_em_simple_forward(struct rte_mbuf *m, uint8_t portid,
 		struct lcore_conf *qconf)
 {
diff --git a/examples/l3fwd/l3fwd_em_hlm_sse.h b/examples/l3fwd/l3fwd_em_hlm_sse.h
index 7714a20ce6..d272f11211 100644
--- a/examples/l3fwd/l3fwd_em_hlm_sse.h
+++ b/examples/l3fwd/l3fwd_em_hlm_sse.h
@@ -36,7 +36,7 @@
 
 #include "l3fwd_sse.h"
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 em_get_dst_port_ipv4x8(struct lcore_conf *qconf, struct rte_mbuf *m[8],
 		uint8_t portid, uint16_t dst_port[8])
 {
@@ -160,7 +160,7 @@ get_ipv6_5tuple(struct rte_mbuf *m0, __m128i mask0,
 	key->xmm[2] = _mm_and_si128(tmpdata2, mask1);
 }
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 em_get_dst_port_ipv6x8(struct lcore_conf *qconf, struct rte_mbuf *m[8],
 		uint8_t portid, uint16_t dst_port[8])
 {
@@ -232,7 +232,7 @@ em_get_dst_port_ipv6x8(struct lcore_conf *qconf, struct rte_mbuf *m[8],
 
 }
 
-static inline __attribute__((always_inline)) uint16_t
+static __rte_always_inline uint16_t
 em_get_dst_port(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
 		uint8_t portid)
 {
diff --git a/examples/l3fwd/l3fwd_em_sse.h b/examples/l3fwd/l3fwd_em_sse.h
index c0a9725a6d..6c794b6a51 100644
--- a/examples/l3fwd/l3fwd_em_sse.h
+++ b/examples/l3fwd/l3fwd_em_sse.h
@@ -45,7 +45,7 @@
 
 #include "l3fwd_sse.h"
 
-static inline __attribute__((always_inline)) uint16_t
+static __rte_always_inline uint16_t
 em_get_dst_port(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
 		uint8_t portid)
 {
diff --git a/examples/l3fwd/l3fwd_lpm.h b/examples/l3fwd/l3fwd_lpm.h
index 258a82fec8..4d77b5807a 100644
--- a/examples/l3fwd/l3fwd_lpm.h
+++ b/examples/l3fwd/l3fwd_lpm.h
@@ -58,7 +58,7 @@ lpm_get_ipv6_dst_port(void *ipv6_hdr,  uint8_t portid, void *lookup_struct)
 			&next_hop) == 0) ?  next_hop : portid);
 }
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 l3fwd_lpm_simple_forward(struct rte_mbuf *m, uint8_t portid,
 		struct lcore_conf *qconf)
 {
diff --git a/examples/l3fwd/l3fwd_lpm_sse.h b/examples/l3fwd/l3fwd_lpm_sse.h
index e92814b10a..fa1b90299f 100644
--- a/examples/l3fwd/l3fwd_lpm_sse.h
+++ b/examples/l3fwd/l3fwd_lpm_sse.h
@@ -36,7 +36,7 @@
 
 #include "l3fwd_sse.h"
 
-static inline __attribute__((always_inline)) uint16_t
+static __rte_always_inline uint16_t
 lpm_get_dst_port(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
 		uint8_t portid)
 {
@@ -75,7 +75,7 @@ lpm_get_dst_port(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
  * precalculated. If packet is ipv6 dst_addr is taken directly from packet
  * header and dst_ipv4 value is not used.
  */
-static inline __attribute__((always_inline)) uint16_t
+static __rte_always_inline uint16_t
 lpm_get_dst_port_with_ipv4(const struct lcore_conf *qconf, struct rte_mbuf *pkt,
 	uint32_t dst_ipv4, uint8_t portid)
 {
diff --git a/examples/l3fwd/l3fwd_sse.h b/examples/l3fwd/l3fwd_sse.h
index fa9c4829d6..6cb1a6408b 100644
--- a/examples/l3fwd/l3fwd_sse.h
+++ b/examples/l3fwd/l3fwd_sse.h
@@ -57,7 +57,7 @@
  * If we encounter invalid IPV4 packet, then set destination port for it
  * to BAD_PORT value.
  */
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 rfc1812_process(struct ipv4_hdr *ipv4_hdr, uint16_t *dp, uint32_t ptype)
 {
 	uint8_t ihl;
@@ -314,7 +314,7 @@ process_packet(struct rte_mbuf *pkt, uint16_t *dst_port)
 	_mm_storeu_si128((__m128i *)eth_hdr, te);
 }
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 send_packetsx4(struct lcore_conf *qconf, uint8_t port, struct rte_mbuf *m[],
 		uint32_t num)
 {
@@ -401,7 +401,7 @@ send_packetsx4(struct lcore_conf *qconf, uint8_t port, struct rte_mbuf *m[],
 /**
  * Send packets burst from pkts_burst to the ports in dst_port array
  */
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 send_packets_multi(struct lcore_conf *qconf, struct rte_mbuf **pkts_burst,
 		uint16_t dst_port[MAX_PKT_BURST], int nb_rx)
 {
diff --git a/examples/performance-thread/common/lthread_pool.h b/examples/performance-thread/common/lthread_pool.h
index fb0c578b03..315a2e21e3 100644
--- a/examples/performance-thread/common/lthread_pool.h
+++ b/examples/performance-thread/common/lthread_pool.h
@@ -174,7 +174,7 @@ _qnode_pool_create(const char *name, int prealloc_size) {
 /*
  * Insert a node into the pool
  */
-static inline void __attribute__ ((always_inline))
+static __rte_always_inline void
 _qnode_pool_insert(struct qnode_pool *p, struct qnode *n)
 {
 	n->next = NULL;
@@ -198,7 +198,7 @@ _qnode_pool_insert(struct qnode_pool *p, struct qnode *n)
  * last item from the queue incurs the penalty of an atomic exchange. Since the
  * pool is maintained with a bulk pre-allocation the cost of this is amortised.
  */
-static inline struct qnode *__attribute__ ((always_inline))
+static __rte_always_inline struct qnode *
 _pool_remove(struct qnode_pool *p)
 {
 	struct qnode *head;
@@ -239,7 +239,7 @@ _pool_remove(struct qnode_pool *p)
  * This adds a retry to the _pool_remove function
  * defined above
  */
-static inline struct qnode *__attribute__ ((always_inline))
+static __rte_always_inline struct qnode *
 _qnode_pool_remove(struct qnode_pool *p)
 {
 	struct qnode *n;
@@ -259,7 +259,7 @@ _qnode_pool_remove(struct qnode_pool *p)
  * Allocate a node from the pool
  * If the pool is empty add mode nodes
  */
-static inline struct qnode *__attribute__ ((always_inline))
+static __rte_always_inline struct qnode *
 _qnode_alloc(void)
 {
 	struct qnode_pool *p = (THIS_SCHED)->qnode_pool;
@@ -304,7 +304,7 @@ _qnode_alloc(void)
 /*
 * free a queue node to the per scheduler pool from which it came
 */
-static inline void __attribute__ ((always_inline))
+static __rte_always_inline void
 _qnode_free(struct qnode *n)
 {
 	struct qnode_pool *p = n->pool;
diff --git a/examples/performance-thread/common/lthread_queue.h b/examples/performance-thread/common/lthread_queue.h
index 4fc2074e42..833ed92b52 100644
--- a/examples/performance-thread/common/lthread_queue.h
+++ b/examples/performance-thread/common/lthread_queue.h
@@ -154,7 +154,7 @@ _lthread_queue_create(const char *name)
 /**
  * Return true if the queue is empty
  */
-static inline int __attribute__ ((always_inline))
+static __rte_always_inline int
 _lthread_queue_empty(struct lthread_queue *q)
 {
 	return q->tail == q->head;
@@ -185,7 +185,7 @@ RTE_DECLARE_PER_LCORE(struct lthread_sched *, this_sched);
  * Insert a node into a queue
  * this implementation is multi producer safe
  */
-static inline struct qnode *__attribute__ ((always_inline))
+static __rte_always_inline struct qnode *
 _lthread_queue_insert_mp(struct lthread_queue
 							  *q, void *data)
 {
@@ -219,7 +219,7 @@ _lthread_queue_insert_mp(struct lthread_queue
  * Insert an node into a queue in single producer mode
  * this implementation is NOT mult producer safe
  */
-static inline struct qnode *__attribute__ ((always_inline))
+static __rte_always_inline struct qnode *
 _lthread_queue_insert_sp(struct lthread_queue
 							  *q, void *data)
 {
@@ -247,7 +247,7 @@ _lthread_queue_insert_sp(struct lthread_queue
 /*
  * Remove a node from a queue
  */
-static inline void *__attribute__ ((always_inline))
+static __rte_always_inline void *
 _lthread_queue_poll(struct lthread_queue *q)
 {
 	void *data = NULL;
@@ -278,7 +278,7 @@ _lthread_queue_poll(struct lthread_queue *q)
 /*
  * Remove a node from a queue
  */
-static inline void *__attribute__ ((always_inline))
+static __rte_always_inline void *
 _lthread_queue_remove(struct lthread_queue *q)
 {
 	void *data = NULL;
diff --git a/examples/performance-thread/common/lthread_sched.c b/examples/performance-thread/common/lthread_sched.c
index c64c21ffb2..98291478ec 100644
--- a/examples/performance-thread/common/lthread_sched.c
+++ b/examples/performance-thread/common/lthread_sched.c
@@ -369,8 +369,8 @@ void lthread_scheduler_shutdown_all(void)
 /*
  * Resume a suspended lthread
  */
-static inline void
-_lthread_resume(struct lthread *lt) __attribute__ ((always_inline));
+static __rte_always_inline void
+_lthread_resume(struct lthread *lt);
 static inline void _lthread_resume(struct lthread *lt)
 {
 	struct lthread_sched *sched = THIS_SCHED;
diff --git a/examples/performance-thread/common/lthread_sched.h b/examples/performance-thread/common/lthread_sched.h
index 7cddda9c58..aa2f0c4887 100644
--- a/examples/performance-thread/common/lthread_sched.h
+++ b/examples/performance-thread/common/lthread_sched.h
@@ -112,8 +112,8 @@ static inline uint64_t _sched_now(void)
 	return 1;
 }
 
-static inline void
-_affinitize(void) __attribute__ ((always_inline));
+static __rte_always_inline void
+_affinitize(void);
 static inline void
 _affinitize(void)
 {
@@ -123,8 +123,8 @@ _affinitize(void)
 	ctx_switch(&(THIS_SCHED)->ctx, &lt->ctx);
 }
 
-static inline void
-_suspend(void) __attribute__ ((always_inline));
+static __rte_always_inline void
+_suspend(void);
 static inline void
 _suspend(void)
 {
@@ -136,8 +136,8 @@ _suspend(void)
 	(THIS_SCHED)->nb_blocked_threads--;
 }
 
-static inline void
-_reschedule(void) __attribute__ ((always_inline));
+static __rte_always_inline void
+_reschedule(void);
 static inline void
 _reschedule(void)
 {
diff --git a/examples/performance-thread/l3fwd-thread/main.c b/examples/performance-thread/l3fwd-thread/main.c
index ac85a369fd..22047872b3 100644
--- a/examples/performance-thread/l3fwd-thread/main.c
+++ b/examples/performance-thread/l3fwd-thread/main.c
@@ -720,7 +720,7 @@ send_single_packet(struct rte_mbuf *m, uint8_t port)
 
 #if ((APP_LOOKUP_METHOD == APP_LOOKUP_LPM) && \
 	(ENABLE_MULTI_BUFFER_OPTIMIZE == 1))
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 send_packetsx4(uint8_t port,
 	struct rte_mbuf *m[], uint32_t num)
 {
@@ -1287,7 +1287,7 @@ simple_ipv6_fwd_8pkts(struct rte_mbuf *m[8], uint8_t portid)
 }
 #endif /* APP_LOOKUP_METHOD */
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 l3fwd_simple_forward(struct rte_mbuf *m, uint8_t portid)
 {
 	struct ether_hdr *eth_hdr;
@@ -1375,7 +1375,7 @@ l3fwd_simple_forward(struct rte_mbuf *m, uint8_t portid)
  * If we encounter invalid IPV4 packet, then set destination port for it
  * to BAD_PORT value.
  */
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 rfc1812_process(struct ipv4_hdr *ipv4_hdr, uint16_t *dp, uint32_t ptype)
 {
 	uint8_t ihl;
@@ -1403,7 +1403,7 @@ rfc1812_process(struct ipv4_hdr *ipv4_hdr, uint16_t *dp, uint32_t ptype)
 #if ((APP_LOOKUP_METHOD == APP_LOOKUP_LPM) && \
 	(ENABLE_MULTI_BUFFER_OPTIMIZE == 1))
 
-static inline __attribute__((always_inline)) uint16_t
+static __rte_always_inline uint16_t
 get_dst_port(struct rte_mbuf *pkt, uint32_t dst_ipv4, uint8_t portid)
 {
 	uint32_t next_hop;
diff --git a/examples/tep_termination/main.c b/examples/tep_termination/main.c
index cd6e3f1cfa..83c2189fff 100644
--- a/examples/tep_termination/main.c
+++ b/examples/tep_termination/main.c
@@ -559,7 +559,7 @@ check_ports_num(unsigned max_nb_ports)
  * This function routes the TX packet to the correct interface. This may be a local device
  * or the physical port.
  */
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m)
 {
 	struct mbuf_table *tx_q;
diff --git a/examples/vhost/main.c b/examples/vhost/main.c
index e07f866933..b625c52a0c 100644
--- a/examples/vhost/main.c
+++ b/examples/vhost/main.c
@@ -691,7 +691,7 @@ static unsigned check_ports_num(unsigned nb_ports)
 	return valid_num_ports;
 }
 
-static inline struct vhost_dev *__attribute__((always_inline))
+static __rte_always_inline struct vhost_dev *
 find_vhost_dev(struct ether_addr *mac)
 {
 	struct vhost_dev *vdev;
@@ -791,7 +791,7 @@ unlink_vmdq(struct vhost_dev *vdev)
 	}
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 virtio_xmit(struct vhost_dev *dst_vdev, struct vhost_dev *src_vdev,
 	    struct rte_mbuf *m)
 {
@@ -815,7 +815,7 @@ virtio_xmit(struct vhost_dev *dst_vdev, struct vhost_dev *src_vdev,
  * Check if the packet destination MAC address is for a local device. If so then put
  * the packet on that devices RX queue. If not then return.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 virtio_tx_local(struct vhost_dev *vdev, struct rte_mbuf *m)
 {
 	struct ether_hdr *pkt_hdr;
@@ -851,7 +851,7 @@ virtio_tx_local(struct vhost_dev *vdev, struct rte_mbuf *m)
  * Check if the destination MAC of a packet is one local VM,
  * and get its vlan tag, and offset if it is.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 find_local_dest(struct vhost_dev *vdev, struct rte_mbuf *m,
 	uint32_t *offset, uint16_t *vlan_tag)
 {
@@ -919,7 +919,7 @@ free_pkts(struct rte_mbuf **pkts, uint16_t n)
 		rte_pktmbuf_free(pkts[n]);
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 do_drain_mbuf_table(struct mbuf_table *tx_q)
 {
 	uint16_t count;
@@ -936,7 +936,7 @@ do_drain_mbuf_table(struct mbuf_table *tx_q)
  * This function routes the TX packet to the correct interface. This
  * may be a local device or the physical port.
  */
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m, uint16_t vlan_tag)
 {
 	struct mbuf_table *tx_q;
@@ -1024,7 +1024,7 @@ queue2nic:
 }
 
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 drain_mbuf_table(struct mbuf_table *tx_q)
 {
 	static uint64_t prev_tsc;
@@ -1044,7 +1044,7 @@ drain_mbuf_table(struct mbuf_table *tx_q)
 	}
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 drain_eth_rx(struct vhost_dev *vdev)
 {
 	uint16_t rx_count, enqueue_count;
@@ -1088,7 +1088,7 @@ drain_eth_rx(struct vhost_dev *vdev)
 	free_pkts(pkts, rx_count);
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 drain_virtio_tx(struct vhost_dev *vdev)
 {
 	struct rte_mbuf *pkts[MAX_PKT_BURST];
diff --git a/examples/vhost/virtio_net.c b/examples/vhost/virtio_net.c
index 5e1ed44a5d..1ab57f5263 100644
--- a/examples/vhost/virtio_net.c
+++ b/examples/vhost/virtio_net.c
@@ -80,7 +80,7 @@ vs_vhost_net_remove(struct vhost_dev *dev)
 	free(dev->mem);
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 enqueue_pkt(struct vhost_dev *dev, struct rte_vhost_vring *vr,
 	    struct rte_mbuf *m, uint16_t desc_idx)
 {
@@ -217,7 +217,7 @@ vs_enqueue_pkts(struct vhost_dev *dev, uint16_t queue_id,
 	return count;
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 dequeue_pkt(struct vhost_dev *dev, struct rte_vhost_vring *vr,
 	    struct rte_mbuf *m, uint16_t desc_idx,
 	    struct rte_mempool *mbuf_pool)
diff --git a/examples/vhost_xen/main.c b/examples/vhost_xen/main.c
index d9ef140f79..f837891764 100644
--- a/examples/vhost_xen/main.c
+++ b/examples/vhost_xen/main.c
@@ -510,7 +510,7 @@ static unsigned check_ports_num(unsigned nb_ports)
  * Function to convert guest physical addresses to vhost virtual addresses. This
  * is used to convert virtio buffer addresses.
  */
-static inline uint64_t __attribute__((always_inline))
+static __rte_always_inline uint64_t
 gpa_to_vva(struct virtio_net *dev, uint64_t guest_pa)
 {
 	struct virtio_memory_regions *region;
@@ -537,7 +537,7 @@ gpa_to_vva(struct virtio_net *dev, uint64_t guest_pa)
  * count is returned to indicate the number of packets that were succesfully
  * added to the RX queue.
  */
-static inline uint32_t __attribute__((always_inline))
+static __rte_always_inline uint32_t
 virtio_dev_rx(struct virtio_net *dev, struct rte_mbuf **pkts, uint32_t count)
 {
 	struct vhost_virtqueue *vq;
@@ -662,7 +662,7 @@ virtio_dev_rx(struct virtio_net *dev, struct rte_mbuf **pkts, uint32_t count)
 /*
  * Compares a packet destination MAC address to a device MAC address.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 ether_addr_cmp(struct ether_addr *ea, struct ether_addr *eb)
 {
 	return ((*(uint64_t *)ea ^ *(uint64_t *)eb) & MAC_ADDR_CMP) == 0;
@@ -757,7 +757,7 @@ unlink_vmdq(struct virtio_net *dev)
  * Check if the packet destination MAC address is for a local device. If so then put
  * the packet on that devices RX queue. If not then return.
  */
-static inline unsigned __attribute__((always_inline))
+static __rte_always_inline unsigned
 virtio_tx_local(struct virtio_net *dev, struct rte_mbuf *m)
 {
 	struct virtio_net_data_ll *dev_ll;
@@ -814,7 +814,7 @@ virtio_tx_local(struct virtio_net *dev, struct rte_mbuf *m)
  * This function routes the TX packet to the correct interface. This may be a local device
  * or the physical port.
  */
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 virtio_tx_route(struct virtio_net* dev, struct rte_mbuf *m, struct rte_mempool *mbuf_pool, uint16_t vlan_tag)
 {
 	struct mbuf_table *tx_q;
@@ -883,7 +883,7 @@ virtio_tx_route(struct virtio_net* dev, struct rte_mbuf *m, struct rte_mempool *
 	return;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 virtio_dev_tx(struct virtio_net* dev, struct rte_mempool *mbuf_pool)
 {
 	struct rte_mbuf m;
diff --git a/lib/librte_acl/acl_run_altivec.h b/lib/librte_acl/acl_run_altivec.h
index 7d329bcf3c..62fd6a22fc 100644
--- a/lib/librte_acl/acl_run_altivec.h
+++ b/lib/librte_acl/acl_run_altivec.h
@@ -104,13 +104,13 @@ resolve_priority_altivec(uint64_t transition, int n,
 /*
  * Check for any match in 4 transitions
  */
-static inline __attribute__((always_inline)) uint32_t
+static __rte_always_inline uint32_t
 check_any_match_x4(uint64_t val[])
 {
 	return (val[0] | val[1] | val[2] | val[3]) & RTE_ACL_NODE_MATCH;
 }
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 acl_match_check_x4(int slot, const struct rte_acl_ctx *ctx, struct parms *parms,
 	struct acl_flow_data *flows, uint64_t transitions[])
 {
diff --git a/lib/librte_acl/acl_run_avx2.h b/lib/librte_acl/acl_run_avx2.h
index b01a46a5ca..804e45afae 100644
--- a/lib/librte_acl/acl_run_avx2.h
+++ b/lib/librte_acl/acl_run_avx2.h
@@ -86,7 +86,7 @@ static const rte_ymm_t ymm_range_base = {
  * tr_hi contains high 32 bits for 8 transition.
  * next_input contains up to 4 input bytes for 8 flows.
  */
-static inline __attribute__((always_inline)) ymm_t
+static __rte_always_inline ymm_t
 transition8(ymm_t next_input, const uint64_t *trans, ymm_t *tr_lo, ymm_t *tr_hi)
 {
 	const int32_t *tr;
diff --git a/lib/librte_acl/acl_run_neon.h b/lib/librte_acl/acl_run_neon.h
index d233ff0070..dfa38f5ebb 100644
--- a/lib/librte_acl/acl_run_neon.h
+++ b/lib/librte_acl/acl_run_neon.h
@@ -99,13 +99,13 @@ resolve_priority_neon(uint64_t transition, int n, const struct rte_acl_ctx *ctx,
 /*
  * Check for any match in 4 transitions
  */
-static inline __attribute__((always_inline)) uint32_t
+static __rte_always_inline uint32_t
 check_any_match_x4(uint64_t val[])
 {
 	return (val[0] | val[1] | val[2] | val[3]) & RTE_ACL_NODE_MATCH;
 }
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 acl_match_check_x4(int slot, const struct rte_acl_ctx *ctx, struct parms *parms,
 		   struct acl_flow_data *flows, uint64_t transitions[])
 {
@@ -124,7 +124,7 @@ acl_match_check_x4(int slot, const struct rte_acl_ctx *ctx, struct parms *parms,
 /*
  * Process 4 transitions (in 2 NEON Q registers) in parallel
  */
-static inline __attribute__((always_inline)) int32x4_t
+static __rte_always_inline int32x4_t
 transition4(int32x4_t next_input, const uint64_t *trans, uint64_t transitions[])
 {
 	int32x4x2_t tr_hi_lo;
diff --git a/lib/librte_acl/acl_run_sse.h b/lib/librte_acl/acl_run_sse.h
index ad40a67451..72f66e4fcc 100644
--- a/lib/librte_acl/acl_run_sse.h
+++ b/lib/librte_acl/acl_run_sse.h
@@ -149,7 +149,7 @@ acl_process_matches(xmm_t *indices, int slot, const struct rte_acl_ctx *ctx,
 /*
  * Check for any match in 4 transitions (contained in 2 SSE registers)
  */
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 acl_match_check_x4(int slot, const struct rte_acl_ctx *ctx, struct parms *parms,
 	struct acl_flow_data *flows, xmm_t *indices1, xmm_t *indices2,
 	xmm_t match_mask)
@@ -176,7 +176,7 @@ acl_match_check_x4(int slot, const struct rte_acl_ctx *ctx, struct parms *parms,
 /*
  * Process 4 transitions (in 2 XMM registers) in parallel
  */
-static inline __attribute__((always_inline)) xmm_t
+static __rte_always_inline xmm_t
 transition4(xmm_t next_input, const uint64_t *trans,
 	xmm_t *indices1, xmm_t *indices2)
 {
diff --git a/lib/librte_eal/common/include/arch/arm/rte_io_64.h b/lib/librte_eal/common/include/arch/arm/rte_io_64.h
index 0402125bba..e59e22a0b2 100644
--- a/lib/librte_eal/common/include/arch/arm/rte_io_64.h
+++ b/lib/librte_eal/common/include/arch/arm/rte_io_64.h
@@ -44,7 +44,7 @@ extern "C" {
 #include "generic/rte_io.h"
 #include "rte_atomic_64.h"
 
-static inline uint8_t __attribute__((always_inline))
+static __rte_always_inline uint8_t
 rte_read8_relaxed(const volatile void *addr)
 {
 	uint8_t val;
@@ -56,7 +56,7 @@ rte_read8_relaxed(const volatile void *addr)
 	return val;
 }
 
-static inline uint16_t __attribute__((always_inline))
+static __rte_always_inline uint16_t
 rte_read16_relaxed(const volatile void *addr)
 {
 	uint16_t val;
@@ -68,7 +68,7 @@ rte_read16_relaxed(const volatile void *addr)
 	return val;
 }
 
-static inline uint32_t __attribute__((always_inline))
+static __rte_always_inline uint32_t
 rte_read32_relaxed(const volatile void *addr)
 {
 	uint32_t val;
@@ -80,7 +80,7 @@ rte_read32_relaxed(const volatile void *addr)
 	return val;
 }
 
-static inline uint64_t __attribute__((always_inline))
+static __rte_always_inline uint64_t
 rte_read64_relaxed(const volatile void *addr)
 {
 	uint64_t val;
@@ -92,7 +92,7 @@ rte_read64_relaxed(const volatile void *addr)
 	return val;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write8_relaxed(uint8_t val, volatile void *addr)
 {
 	asm volatile(
@@ -101,7 +101,7 @@ rte_write8_relaxed(uint8_t val, volatile void *addr)
 		    : [val] "r" (val), [addr] "r" (addr));
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write16_relaxed(uint16_t val, volatile void *addr)
 {
 	asm volatile(
@@ -110,7 +110,7 @@ rte_write16_relaxed(uint16_t val, volatile void *addr)
 		    : [val] "r" (val), [addr] "r" (addr));
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write32_relaxed(uint32_t val, volatile void *addr)
 {
 	asm volatile(
@@ -119,7 +119,7 @@ rte_write32_relaxed(uint32_t val, volatile void *addr)
 		    : [val] "r" (val), [addr] "r" (addr));
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write64_relaxed(uint64_t val, volatile void *addr)
 {
 	asm volatile(
@@ -128,7 +128,7 @@ rte_write64_relaxed(uint64_t val, volatile void *addr)
 		    : [val] "r" (val), [addr] "r" (addr));
 }
 
-static inline uint8_t __attribute__((always_inline))
+static __rte_always_inline uint8_t
 rte_read8(const volatile void *addr)
 {
 	uint8_t val;
@@ -137,7 +137,7 @@ rte_read8(const volatile void *addr)
 	return val;
 }
 
-static inline uint16_t __attribute__((always_inline))
+static __rte_always_inline uint16_t
 rte_read16(const volatile void *addr)
 {
 	uint16_t val;
@@ -146,7 +146,7 @@ rte_read16(const volatile void *addr)
 	return val;
 }
 
-static inline uint32_t  __attribute__((always_inline))
+static __rte_always_inline uint32_t
 rte_read32(const volatile void *addr)
 {
 	uint32_t val;
@@ -155,7 +155,7 @@ rte_read32(const volatile void *addr)
 	return val;
 }
 
-static inline uint64_t __attribute__((always_inline))
+static __rte_always_inline uint64_t
 rte_read64(const volatile void *addr)
 {
 	uint64_t val;
@@ -164,28 +164,28 @@ rte_read64(const volatile void *addr)
 	return val;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write8(uint8_t value, volatile void *addr)
 {
 	rte_io_wmb();
 	rte_write8_relaxed(value, addr);
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write16(uint16_t value, volatile void *addr)
 {
 	rte_io_wmb();
 	rte_write16_relaxed(value, addr);
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write32(uint32_t value, volatile void *addr)
 {
 	rte_io_wmb();
 	rte_write32_relaxed(value, addr);
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write64(uint64_t value, volatile void *addr)
 {
 	rte_io_wmb();
diff --git a/lib/librte_eal/common/include/arch/x86/rte_memcpy.h b/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
index b9785e85ee..74c280c2cf 100644
--- a/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
+++ b/lib/librte_eal/common/include/arch/x86/rte_memcpy.h
@@ -44,6 +44,7 @@
 #include <stdint.h>
 #include <string.h>
 #include <rte_vect.h>
+#include <rte_common.h>
 
 #ifdef __cplusplus
 extern "C" {
@@ -64,8 +65,8 @@ extern "C" {
  * @return
  *   Pointer to the destination data.
  */
-static inline void *
-rte_memcpy(void *dst, const void *src, size_t n) __attribute__((always_inline));
+static __rte_always_inline void *
+rte_memcpy(void *dst, const void *src, size_t n);
 
 #ifdef RTE_MACHINE_CPUFLAG_AVX512F
 
diff --git a/lib/librte_eal/common/include/generic/rte_io.h b/lib/librte_eal/common/include/generic/rte_io.h
index d82ee6951e..477e7b5928 100644
--- a/lib/librte_eal/common/include/generic/rte_io.h
+++ b/lib/librte_eal/common/include/generic/rte_io.h
@@ -264,55 +264,55 @@ rte_write64(uint64_t value, volatile void *addr);
 
 #ifndef RTE_OVERRIDE_IO_H
 
-static inline uint8_t __attribute__((always_inline))
+static __rte_always_inline uint8_t
 rte_read8_relaxed(const volatile void *addr)
 {
 	return *(const volatile uint8_t *)addr;
 }
 
-static inline uint16_t __attribute__((always_inline))
+static __rte_always_inline uint16_t
 rte_read16_relaxed(const volatile void *addr)
 {
 	return *(const volatile uint16_t *)addr;
 }
 
-static inline uint32_t __attribute__((always_inline))
+static __rte_always_inline uint32_t
 rte_read32_relaxed(const volatile void *addr)
 {
 	return *(const volatile uint32_t *)addr;
 }
 
-static inline uint64_t __attribute__((always_inline))
+static __rte_always_inline uint64_t
 rte_read64_relaxed(const volatile void *addr)
 {
 	return *(const volatile uint64_t *)addr;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write8_relaxed(uint8_t value, volatile void *addr)
 {
 	*(volatile uint8_t *)addr = value;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write16_relaxed(uint16_t value, volatile void *addr)
 {
 	*(volatile uint16_t *)addr = value;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write32_relaxed(uint32_t value, volatile void *addr)
 {
 	*(volatile uint32_t *)addr = value;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write64_relaxed(uint64_t value, volatile void *addr)
 {
 	*(volatile uint64_t *)addr = value;
 }
 
-static inline uint8_t __attribute__((always_inline))
+static __rte_always_inline uint8_t
 rte_read8(const volatile void *addr)
 {
 	uint8_t val;
@@ -321,7 +321,7 @@ rte_read8(const volatile void *addr)
 	return val;
 }
 
-static inline uint16_t __attribute__((always_inline))
+static __rte_always_inline uint16_t
 rte_read16(const volatile void *addr)
 {
 	uint16_t val;
@@ -330,7 +330,7 @@ rte_read16(const volatile void *addr)
 	return val;
 }
 
-static inline uint32_t  __attribute__((always_inline))
+static __rte_always_inline uint32_t
 rte_read32(const volatile void *addr)
 {
 	uint32_t val;
@@ -339,7 +339,7 @@ rte_read32(const volatile void *addr)
 	return val;
 }
 
-static inline uint64_t __attribute__((always_inline))
+static __rte_always_inline uint64_t
 rte_read64(const volatile void *addr)
 {
 	uint64_t val;
@@ -348,28 +348,28 @@ rte_read64(const volatile void *addr)
 	return val;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write8(uint8_t value, volatile void *addr)
 {
 	rte_io_wmb();
 	rte_write8_relaxed(value, addr);
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write16(uint16_t value, volatile void *addr)
 {
 	rte_io_wmb();
 	rte_write16_relaxed(value, addr);
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write32(uint32_t value, volatile void *addr)
 {
 	rte_io_wmb();
 	rte_write32_relaxed(value, addr);
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_write64(uint64_t value, volatile void *addr)
 {
 	rte_io_wmb();
diff --git a/lib/librte_eal/common/include/rte_common.h b/lib/librte_eal/common/include/rte_common.h
index e057f6e213..a9a749488b 100644
--- a/lib/librte_eal/common/include/rte_common.h
+++ b/lib/librte_eal/common/include/rte_common.h
@@ -102,6 +102,11 @@ typedef uint16_t unaligned_uint16_t;
  */
 #define RTE_SET_USED(x) (void)(x)
 
+/**
+ * Force a function to be inlined
+ */
+#define __rte_always_inline inline __attribute__((always_inline))
+
 /*********** Macros for pointer arithmetic ********/
 
 /**
diff --git a/lib/librte_ether/rte_ethdev.h b/lib/librte_ether/rte_ethdev.h
index 0f38b45f83..121058c120 100644
--- a/lib/librte_ether/rte_ethdev.h
+++ b/lib/librte_ether/rte_ethdev.h
@@ -3266,7 +3266,7 @@ rte_eth_tx_buffer_flush(uint8_t port_id, uint16_t queue_id,
  *     causing N packets to be sent, and the error callback to be called for
  *     the rest.
  */
-static inline uint16_t __attribute__((always_inline))
+static __rte_always_inline uint16_t
 rte_eth_tx_buffer(uint8_t port_id, uint16_t queue_id,
 		struct rte_eth_dev_tx_buffer *buffer, struct rte_mbuf *tx_pkt)
 {
diff --git a/lib/librte_mbuf/rte_mbuf.h b/lib/librte_mbuf/rte_mbuf.h
index 1cb03109c1..fe605c7a48 100644
--- a/lib/librte_mbuf/rte_mbuf.h
+++ b/lib/librte_mbuf/rte_mbuf.h
@@ -840,7 +840,7 @@ static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
  * @param m
  *   The mbuf to be freed.
  */
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_mbuf_raw_free(struct rte_mbuf *m)
 {
 	RTE_ASSERT(RTE_MBUF_DIRECT(m));
@@ -1287,8 +1287,7 @@ static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
  *   - (m) if it is the last reference. It can be recycled or freed.
  *   - (NULL) if the mbuf still has remaining references on it.
  */
-__attribute__((always_inline))
-static inline struct rte_mbuf *
+static __rte_always_inline struct rte_mbuf *
 rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
 {
 	__rte_mbuf_sanity_check(m, 0);
@@ -1339,7 +1338,7 @@ __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
  * @param m
  *   The packet mbuf segment to be freed.
  */
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_pktmbuf_free_seg(struct rte_mbuf *m)
 {
 	m = rte_pktmbuf_prefree_seg(m);
diff --git a/lib/librte_mempool/rte_mempool.h b/lib/librte_mempool/rte_mempool.h
index 48bc8ea3c7..76b5b3b152 100644
--- a/lib/librte_mempool/rte_mempool.h
+++ b/lib/librte_mempool/rte_mempool.h
@@ -993,7 +993,7 @@ rte_mempool_cache_free(struct rte_mempool_cache *cache);
  * @param mp
  *   A pointer to the mempool.
  */
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_mempool_cache_flush(struct rte_mempool_cache *cache,
 			struct rte_mempool *mp)
 {
@@ -1011,7 +1011,7 @@ rte_mempool_cache_flush(struct rte_mempool_cache *cache,
  * @return
  *   A pointer to the mempool cache or NULL if disabled or non-EAL thread.
  */
-static inline struct rte_mempool_cache *__attribute__((always_inline))
+static __rte_always_inline struct rte_mempool_cache *
 rte_mempool_default_cache(struct rte_mempool *mp, unsigned lcore_id)
 {
 	if (mp->cache_size == 0)
@@ -1038,7 +1038,7 @@ rte_mempool_default_cache(struct rte_mempool *mp, unsigned lcore_id)
  *   The flags used for the mempool creation.
  *   Single-producer (MEMPOOL_F_SP_PUT flag) or multi-producers.
  */
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 __mempool_generic_put(struct rte_mempool *mp, void * const *obj_table,
 		      unsigned n, struct rte_mempool_cache *cache)
 {
@@ -1100,7 +1100,7 @@ ring_enqueue:
  *   The flags used for the mempool creation.
  *   Single-producer (MEMPOOL_F_SP_PUT flag) or multi-producers.
  */
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_mempool_generic_put(struct rte_mempool *mp, void * const *obj_table,
 			unsigned n, struct rte_mempool_cache *cache,
 			__rte_unused int flags)
@@ -1123,7 +1123,7 @@ rte_mempool_generic_put(struct rte_mempool *mp, void * const *obj_table,
  * @param n
  *   The number of objects to add in the mempool from obj_table.
  */
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
 		     unsigned n)
 {
@@ -1144,7 +1144,7 @@ rte_mempool_put_bulk(struct rte_mempool *mp, void * const *obj_table,
  * @param obj
  *   A pointer to the object to be added.
  */
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 rte_mempool_put(struct rte_mempool *mp, void *obj)
 {
 	rte_mempool_put_bulk(mp, &obj, 1);
@@ -1167,7 +1167,7 @@ rte_mempool_put(struct rte_mempool *mp, void *obj)
  *   - >=0: Success; number of objects supplied.
  *   - <0: Error; code of ring dequeue function.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 __mempool_generic_get(struct rte_mempool *mp, void **obj_table,
 		      unsigned n, struct rte_mempool_cache *cache)
 {
@@ -1248,7 +1248,7 @@ ring_dequeue:
  *   - 0: Success; objects taken.
  *   - -ENOENT: Not enough entries in the mempool; no object is retrieved.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_mempool_generic_get(struct rte_mempool *mp, void **obj_table, unsigned n,
 			struct rte_mempool_cache *cache, __rte_unused int flags)
 {
@@ -1281,7 +1281,7 @@ rte_mempool_generic_get(struct rte_mempool *mp, void **obj_table, unsigned n,
  *   - 0: Success; objects taken
  *   - -ENOENT: Not enough entries in the mempool; no object is retrieved.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_mempool_get_bulk(struct rte_mempool *mp, void **obj_table, unsigned n)
 {
 	struct rte_mempool_cache *cache;
@@ -1309,7 +1309,7 @@ rte_mempool_get_bulk(struct rte_mempool *mp, void **obj_table, unsigned n)
  *   - 0: Success; objects taken.
  *   - -ENOENT: Not enough entries in the mempool; no object is retrieved.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_mempool_get(struct rte_mempool *mp, void **obj_p)
 {
 	return rte_mempool_get_bulk(mp, obj_p, 1);
diff --git a/lib/librte_net/net_crc_sse.h b/lib/librte_net/net_crc_sse.h
index 8bce522a79..ac93637bf1 100644
--- a/lib/librte_net/net_crc_sse.h
+++ b/lib/librte_net/net_crc_sse.h
@@ -73,7 +73,7 @@ struct crc_pclmulqdq_ctx crc16_ccitt_pclmulqdq __rte_aligned(16);
  * @return
  *   New 16 byte folded data
  */
-static inline __attribute__((always_inline)) __m128i
+static __rte_always_inline __m128i
 crcr32_folding_round(__m128i data_block,
 		__m128i precomp,
 		__m128i fold)
@@ -96,7 +96,7 @@ crcr32_folding_round(__m128i data_block,
  *  64 bits reduced data
  */
 
-static inline __attribute__((always_inline)) __m128i
+static __rte_always_inline __m128i
 crcr32_reduce_128_to_64(__m128i data128, __m128i precomp)
 {
 	__m128i tmp0, tmp1, tmp2;
@@ -125,7 +125,7 @@ crcr32_reduce_128_to_64(__m128i data128, __m128i precomp)
  *   reduced 32 bits data
  */
 
-static inline __attribute__((always_inline)) uint32_t
+static __rte_always_inline uint32_t
 crcr32_reduce_64_to_32(__m128i data64, __m128i precomp)
 {
 	static const uint32_t mask1[4] __rte_aligned(16) = {
@@ -171,7 +171,7 @@ static const uint8_t crc_xmm_shift_tab[48] __rte_aligned(16) = {
  *   reg << (num * 8)
  */
 
-static inline __attribute__((always_inline)) __m128i
+static __rte_always_inline __m128i
 xmm_shift_left(__m128i reg, const unsigned int num)
 {
 	const __m128i *p = (const __m128i *)(crc_xmm_shift_tab + 16 - num);
@@ -179,7 +179,7 @@ xmm_shift_left(__m128i reg, const unsigned int num)
 	return _mm_shuffle_epi8(reg, _mm_loadu_si128(p));
 }
 
-static inline __attribute__((always_inline)) uint32_t
+static __rte_always_inline uint32_t
 crc32_eth_calc_pclmulqdq(
 	const uint8_t *data,
 	uint32_t data_len,
diff --git a/lib/librte_net/rte_net_crc.c b/lib/librte_net/rte_net_crc.c
index 9d1ee63faf..0391c7209a 100644
--- a/lib/librte_net/rte_net_crc.c
+++ b/lib/librte_net/rte_net_crc.c
@@ -116,7 +116,7 @@ crc32_eth_init_lut(uint32_t poly,
 	}
 }
 
-static inline __attribute__((always_inline)) uint32_t
+static __rte_always_inline uint32_t
 crc32_eth_calc_lut(const uint8_t *data,
 	uint32_t data_len,
 	uint32_t crc,
diff --git a/lib/librte_port/rte_port_ring.c b/lib/librte_port/rte_port_ring.c
index 64bd965f53..a4e709c96b 100644
--- a/lib/librte_port/rte_port_ring.c
+++ b/lib/librte_port/rte_port_ring.c
@@ -293,7 +293,7 @@ rte_port_ring_multi_writer_tx(void *port, struct rte_mbuf *pkt)
 	return 0;
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_port_ring_writer_tx_bulk_internal(void *port,
 		struct rte_mbuf **pkts,
 		uint64_t pkts_mask,
@@ -609,7 +609,7 @@ rte_port_ring_multi_writer_nodrop_tx(void *port, struct rte_mbuf *pkt)
 	return 0;
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_port_ring_writer_nodrop_tx_bulk_internal(void *port,
 		struct rte_mbuf **pkts,
 		uint64_t pkts_mask,
diff --git a/lib/librte_ring/rte_ring.h b/lib/librte_ring/rte_ring.h
index 97f025a1fb..e4e910b4f6 100644
--- a/lib/librte_ring/rte_ring.h
+++ b/lib/librte_ring/rte_ring.h
@@ -345,7 +345,7 @@ void rte_ring_dump(FILE *f, const struct rte_ring *r);
 	} \
 } while (0)
 
-static inline __attribute__((always_inline)) void
+static __rte_always_inline void
 update_tail(struct rte_ring_headtail *ht, uint32_t old_val, uint32_t new_val,
 		uint32_t single)
 {
@@ -383,7 +383,7 @@ update_tail(struct rte_ring_headtail *ht, uint32_t old_val, uint32_t new_val,
  *   Actual number of objects enqueued.
  *   If behavior == RTE_RING_QUEUE_FIXED, this will be 0 or n only.
  */
-static inline __attribute__((always_inline)) unsigned int
+static __rte_always_inline unsigned int
 __rte_ring_move_prod_head(struct rte_ring *r, int is_sp,
 		unsigned int n, enum rte_ring_queue_behavior behavior,
 		uint32_t *old_head, uint32_t *new_head,
@@ -443,7 +443,7 @@ __rte_ring_move_prod_head(struct rte_ring *r, int is_sp,
  *   Actual number of objects enqueued.
  *   If behavior == RTE_RING_QUEUE_FIXED, this will be 0 or n only.
  */
-static inline __attribute__((always_inline)) unsigned int
+static __rte_always_inline unsigned int
 __rte_ring_do_enqueue(struct rte_ring *r, void * const *obj_table,
 		 unsigned int n, enum rte_ring_queue_behavior behavior,
 		 int is_sp, unsigned int *free_space)
@@ -489,7 +489,7 @@ end:
  *   - Actual number of objects dequeued.
  *     If behavior == RTE_RING_QUEUE_FIXED, this will be 0 or n only.
  */
-static inline __attribute__((always_inline)) unsigned int
+static __rte_always_inline unsigned int
 __rte_ring_move_cons_head(struct rte_ring *r, int is_sc,
 		unsigned int n, enum rte_ring_queue_behavior behavior,
 		uint32_t *old_head, uint32_t *new_head,
@@ -548,7 +548,7 @@ __rte_ring_move_cons_head(struct rte_ring *r, int is_sc,
  *   - Actual number of objects dequeued.
  *     If behavior == RTE_RING_QUEUE_FIXED, this will be 0 or n only.
  */
-static inline __attribute__((always_inline)) unsigned int
+static __rte_always_inline unsigned int
 __rte_ring_do_dequeue(struct rte_ring *r, void **obj_table,
 		 unsigned int n, enum rte_ring_queue_behavior behavior,
 		 int is_sc, unsigned int *available)
@@ -590,7 +590,7 @@ end:
  * @return
  *   The number of objects enqueued, either 0 or n
  */
-static inline unsigned int __attribute__((always_inline))
+static __rte_always_inline unsigned int
 rte_ring_mp_enqueue_bulk(struct rte_ring *r, void * const *obj_table,
 			 unsigned int n, unsigned int *free_space)
 {
@@ -613,7 +613,7 @@ rte_ring_mp_enqueue_bulk(struct rte_ring *r, void * const *obj_table,
  * @return
  *   The number of objects enqueued, either 0 or n
  */
-static inline unsigned int __attribute__((always_inline))
+static __rte_always_inline unsigned int
 rte_ring_sp_enqueue_bulk(struct rte_ring *r, void * const *obj_table,
 			 unsigned int n, unsigned int *free_space)
 {
@@ -640,7 +640,7 @@ rte_ring_sp_enqueue_bulk(struct rte_ring *r, void * const *obj_table,
  * @return
  *   The number of objects enqueued, either 0 or n
  */
-static inline unsigned int __attribute__((always_inline))
+static __rte_always_inline unsigned int
 rte_ring_enqueue_bulk(struct rte_ring *r, void * const *obj_table,
 		      unsigned int n, unsigned int *free_space)
 {
@@ -662,7 +662,7 @@ rte_ring_enqueue_bulk(struct rte_ring *r, void * const *obj_table,
  *   - 0: Success; objects enqueued.
  *   - -ENOBUFS: Not enough room in the ring to enqueue; no object is enqueued.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_ring_mp_enqueue(struct rte_ring *r, void *obj)
 {
 	return rte_ring_mp_enqueue_bulk(r, &obj, 1, NULL) ? 0 : -ENOBUFS;
@@ -679,7 +679,7 @@ rte_ring_mp_enqueue(struct rte_ring *r, void *obj)
  *   - 0: Success; objects enqueued.
  *   - -ENOBUFS: Not enough room in the ring to enqueue; no object is enqueued.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_ring_sp_enqueue(struct rte_ring *r, void *obj)
 {
 	return rte_ring_sp_enqueue_bulk(r, &obj, 1, NULL) ? 0 : -ENOBUFS;
@@ -700,7 +700,7 @@ rte_ring_sp_enqueue(struct rte_ring *r, void *obj)
  *   - 0: Success; objects enqueued.
  *   - -ENOBUFS: Not enough room in the ring to enqueue; no object is enqueued.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_ring_enqueue(struct rte_ring *r, void *obj)
 {
 	return rte_ring_enqueue_bulk(r, &obj, 1, NULL) ? 0 : -ENOBUFS;
@@ -724,7 +724,7 @@ rte_ring_enqueue(struct rte_ring *r, void *obj)
  * @return
  *   The number of objects dequeued, either 0 or n
  */
-static inline unsigned int __attribute__((always_inline))
+static __rte_always_inline unsigned int
 rte_ring_mc_dequeue_bulk(struct rte_ring *r, void **obj_table,
 		unsigned int n, unsigned int *available)
 {
@@ -748,7 +748,7 @@ rte_ring_mc_dequeue_bulk(struct rte_ring *r, void **obj_table,
  * @return
  *   The number of objects dequeued, either 0 or n
  */
-static inline unsigned int __attribute__((always_inline))
+static __rte_always_inline unsigned int
 rte_ring_sc_dequeue_bulk(struct rte_ring *r, void **obj_table,
 		unsigned int n, unsigned int *available)
 {
@@ -775,7 +775,7 @@ rte_ring_sc_dequeue_bulk(struct rte_ring *r, void **obj_table,
  * @return
  *   The number of objects dequeued, either 0 or n
  */
-static inline unsigned int __attribute__((always_inline))
+static __rte_always_inline unsigned int
 rte_ring_dequeue_bulk(struct rte_ring *r, void **obj_table, unsigned int n,
 		unsigned int *available)
 {
@@ -798,7 +798,7 @@ rte_ring_dequeue_bulk(struct rte_ring *r, void **obj_table, unsigned int n,
  *   - -ENOENT: Not enough entries in the ring to dequeue; no object is
  *     dequeued.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_ring_mc_dequeue(struct rte_ring *r, void **obj_p)
 {
 	return rte_ring_mc_dequeue_bulk(r, obj_p, 1, NULL)  ? 0 : -ENOBUFS;
@@ -816,7 +816,7 @@ rte_ring_mc_dequeue(struct rte_ring *r, void **obj_p)
  *   - -ENOENT: Not enough entries in the ring to dequeue, no object is
  *     dequeued.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_ring_sc_dequeue(struct rte_ring *r, void **obj_p)
 {
 	return rte_ring_sc_dequeue_bulk(r, obj_p, 1, NULL) ? 0 : -ENOBUFS;
@@ -838,7 +838,7 @@ rte_ring_sc_dequeue(struct rte_ring *r, void **obj_p)
  *   - -ENOENT: Not enough entries in the ring to dequeue, no object is
  *     dequeued.
  */
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 rte_ring_dequeue(struct rte_ring *r, void **obj_p)
 {
 	return rte_ring_dequeue_bulk(r, obj_p, 1, NULL) ? 0 : -ENOENT;
@@ -962,7 +962,7 @@ struct rte_ring *rte_ring_lookup(const char *name);
  * @return
  *   - n: Actual number of objects enqueued.
  */
-static inline unsigned __attribute__((always_inline))
+static __rte_always_inline unsigned
 rte_ring_mp_enqueue_burst(struct rte_ring *r, void * const *obj_table,
 			 unsigned int n, unsigned int *free_space)
 {
@@ -985,7 +985,7 @@ rte_ring_mp_enqueue_burst(struct rte_ring *r, void * const *obj_table,
  * @return
  *   - n: Actual number of objects enqueued.
  */
-static inline unsigned __attribute__((always_inline))
+static __rte_always_inline unsigned
 rte_ring_sp_enqueue_burst(struct rte_ring *r, void * const *obj_table,
 			 unsigned int n, unsigned int *free_space)
 {
@@ -1012,7 +1012,7 @@ rte_ring_sp_enqueue_burst(struct rte_ring *r, void * const *obj_table,
  * @return
  *   - n: Actual number of objects enqueued.
  */
-static inline unsigned __attribute__((always_inline))
+static __rte_always_inline unsigned
 rte_ring_enqueue_burst(struct rte_ring *r, void * const *obj_table,
 		      unsigned int n, unsigned int *free_space)
 {
@@ -1040,7 +1040,7 @@ rte_ring_enqueue_burst(struct rte_ring *r, void * const *obj_table,
  * @return
  *   - n: Actual number of objects dequeued, 0 if ring is empty
  */
-static inline unsigned __attribute__((always_inline))
+static __rte_always_inline unsigned
 rte_ring_mc_dequeue_burst(struct rte_ring *r, void **obj_table,
 		unsigned int n, unsigned int *available)
 {
@@ -1065,7 +1065,7 @@ rte_ring_mc_dequeue_burst(struct rte_ring *r, void **obj_table,
  * @return
  *   - n: Actual number of objects dequeued, 0 if ring is empty
  */
-static inline unsigned __attribute__((always_inline))
+static __rte_always_inline unsigned
 rte_ring_sc_dequeue_burst(struct rte_ring *r, void **obj_table,
 		unsigned int n, unsigned int *available)
 {
@@ -1092,7 +1092,7 @@ rte_ring_sc_dequeue_burst(struct rte_ring *r, void **obj_table,
  * @return
  *   - Number of objects dequeued
  */
-static inline unsigned __attribute__((always_inline))
+static __rte_always_inline unsigned
 rte_ring_dequeue_burst(struct rte_ring *r, void **obj_table,
 		unsigned int n, unsigned int *available)
 {
diff --git a/lib/librte_vhost/rte_vhost.h b/lib/librte_vhost/rte_vhost.h
index 605e47cbf0..22d0db23db 100644
--- a/lib/librte_vhost/rte_vhost.h
+++ b/lib/librte_vhost/rte_vhost.h
@@ -120,7 +120,7 @@ struct vhost_device_ops {
  * @return
  *  the host virtual address on success, 0 on failure
  */
-static inline uint64_t __attribute__((always_inline))
+static __rte_always_inline uint64_t
 rte_vhost_gpa_to_vva(struct rte_vhost_memory *mem, uint64_t gpa)
 {
 	struct rte_vhost_mem_region *reg;
diff --git a/lib/librte_vhost/vhost.h b/lib/librte_vhost/vhost.h
index ddd8a9c436..0f294f3959 100644
--- a/lib/librte_vhost/vhost.h
+++ b/lib/librte_vhost/vhost.h
@@ -201,13 +201,13 @@ struct virtio_net {
 
 #define VHOST_LOG_PAGE	4096
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 vhost_log_page(uint8_t *log_base, uint64_t page)
 {
 	log_base[page / 8] |= 1 << (page % 8);
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 vhost_log_write(struct virtio_net *dev, uint64_t addr, uint64_t len)
 {
 	uint64_t page;
@@ -229,7 +229,7 @@ vhost_log_write(struct virtio_net *dev, uint64_t addr, uint64_t len)
 	}
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 vhost_log_used_vring(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		     uint64_t offset, uint64_t len)
 {
@@ -272,7 +272,7 @@ extern uint64_t VHOST_FEATURES;
 extern struct virtio_net *vhost_devices[MAX_VHOST_DEVICE];
 
 /* Convert guest physical address to host physical address */
-static inline phys_addr_t __attribute__((always_inline))
+static __rte_always_inline phys_addr_t
 gpa_to_hpa(struct virtio_net *dev, uint64_t gpa, uint64_t size)
 {
 	uint32_t i;
diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c
index 48219e0509..b5d809676a 100644
--- a/lib/librte_vhost/virtio_net.c
+++ b/lib/librte_vhost/virtio_net.c
@@ -55,7 +55,7 @@ is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t nr_vring)
 	return (is_tx ^ (idx & 1)) == 0 && idx < nr_vring;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 do_flush_shadow_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			  uint16_t to, uint16_t from, uint16_t size)
 {
@@ -67,7 +67,7 @@ do_flush_shadow_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			size * sizeof(struct vring_used_elem));
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 flush_shadow_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq)
 {
 	uint16_t used_idx = vq->last_used_idx & (vq->size - 1);
@@ -95,7 +95,7 @@ flush_shadow_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq)
 		sizeof(vq->used->idx));
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 update_shadow_used_ring(struct vhost_virtqueue *vq,
 			 uint16_t desc_idx, uint16_t len)
 {
@@ -153,7 +153,7 @@ virtio_enqueue_offload(struct rte_mbuf *m_buf, struct virtio_net_hdr *net_hdr)
 	}
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 copy_mbuf_to_desc(struct virtio_net *dev, struct vring_desc *descs,
 		  struct rte_mbuf *m, uint16_t desc_idx, uint32_t size)
 {
@@ -237,7 +237,7 @@ copy_mbuf_to_desc(struct virtio_net *dev, struct vring_desc *descs,
  * added to the RX queue. This function works when the mbuf is scattered, but
  * it doesn't support the mergeable feature.
  */
-static inline uint32_t __attribute__((always_inline))
+static __rte_always_inline uint32_t
 virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id,
 	      struct rte_mbuf **pkts, uint32_t count)
 {
@@ -335,7 +335,7 @@ virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id,
 	return count;
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 fill_vec_buf(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			 uint32_t avail_idx, uint32_t *vec_idx,
 			 struct buf_vector *buf_vec, uint16_t *desc_chain_head,
@@ -424,7 +424,7 @@ reserve_avail_buf_mergeable(struct virtio_net *dev, struct vhost_virtqueue *vq,
 	return 0;
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct rte_mbuf *m,
 			    struct buf_vector *buf_vec, uint16_t num_buffers)
 {
@@ -512,7 +512,7 @@ copy_mbuf_to_desc_mergeable(struct virtio_net *dev, struct rte_mbuf *m,
 	return 0;
 }
 
-static inline uint32_t __attribute__((always_inline))
+static __rte_always_inline uint32_t
 virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id,
 	struct rte_mbuf **pkts, uint32_t count)
 {
@@ -655,7 +655,7 @@ parse_ethernet(struct rte_mbuf *m, uint16_t *l4_proto, void **l4_hdr)
 	}
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 vhost_dequeue_offload(struct virtio_net_hdr *hdr, struct rte_mbuf *m)
 {
 	uint16_t l4_proto = 0;
@@ -743,13 +743,13 @@ make_rarp_packet(struct rte_mbuf *rarp_mbuf, const struct ether_addr *mac)
 	return 0;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 put_zmbuf(struct zcopy_mbuf *zmbuf)
 {
 	zmbuf->in_use = 0;
 }
 
-static inline int __attribute__((always_inline))
+static __rte_always_inline int
 copy_desc_to_mbuf(struct virtio_net *dev, struct vring_desc *descs,
 		  uint16_t max_desc, struct rte_mbuf *m, uint16_t desc_idx,
 		  struct rte_mempool *mbuf_pool)
@@ -899,7 +899,7 @@ copy_desc_to_mbuf(struct virtio_net *dev, struct vring_desc *descs,
 	return 0;
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 update_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		 uint32_t used_idx, uint32_t desc_idx)
 {
@@ -910,7 +910,7 @@ update_used_ring(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			sizeof(vq->used->ring[used_idx]));
 }
 
-static inline void __attribute__((always_inline))
+static __rte_always_inline void
 update_used_idx(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		uint32_t count)
 {
@@ -930,7 +930,7 @@ update_used_idx(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		eventfd_write(vq->callfd, (eventfd_t)1);
 }
 
-static inline struct zcopy_mbuf *__attribute__((always_inline))
+static __rte_always_inline struct zcopy_mbuf *
 get_zmbuf(struct vhost_virtqueue *vq)
 {
 	uint16_t i;
@@ -961,7 +961,7 @@ again:
 	return NULL;
 }
 
-static inline bool __attribute__((always_inline))
+static __rte_always_inline bool
 mbuf_is_consumed(struct rte_mbuf *m)
 {
 	while (m) {
diff --git a/test/test/test_xmmt_ops.h b/test/test/test_xmmt_ops.h
index 42174d2c92..ef014818bb 100644
--- a/test/test/test_xmmt_ops.h
+++ b/test/test/test_xmmt_ops.h
@@ -44,7 +44,7 @@
 #define vect_loadu_sil128(p) vld1q_s32((const int32_t *)p)
 
 /* sets the 4 signed 32-bit integer values and returns the xmm_t variable */
-static inline xmm_t  __attribute__((always_inline))
+static __rte_always_inline xmm_t
 vect_set_epi32(int i3, int i2, int i1, int i0)
 {
 	int32_t data[4] = {i0, i1, i2, i3};
@@ -70,7 +70,7 @@ vect_set_epi32(int i3, int i2, int i1, int i0)
 #define vect_loadu_sil128(p) vec_ld(0, p)
 
 /* sets the 4 signed 32-bit integer values and returns the xmm_t variable */
-static inline xmm_t  __attribute__((always_inline))
+static __rte_always_inline xmm_t
 vect_set_epi32(int i3, int i2, int i1, int i0)
 {
 	xmm_t data = (xmm_t){i0, i1, i2, i3};