[v2,12/29] drivers: optimize thread local storage for dpaa
diff mbox series

Message ID 20200707092244.12791-13-hemant.agrawal@nxp.com
State Accepted
Delegated to: Ferruh Yigit
Headers show
Series
  • NXP DPAAx enhancements
Related show

Checks

Context Check Description
ci/Intel-compilation success Compilation OK
ci/checkpatch success coding style OK

Commit Message

Hemant Agrawal July 7, 2020, 9:22 a.m. UTC
From: Rohit Raj <rohit.raj@nxp.com>

Minimize the number of different thread variables

Add all the thread specific variables in dpaa_portal
structure to optimize TLS Usage.

Signed-off-by: Rohit Raj <rohit.raj@nxp.com>
Acked-by: Akhil Goyal <akhil.goyal@nxp.com>
---
 doc/guides/rel_notes/release_20_08.rst    |  6 ++++
 drivers/bus/dpaa/dpaa_bus.c               | 24 ++++++-------
 drivers/bus/dpaa/rte_bus_dpaa_version.map |  1 -
 drivers/bus/dpaa/rte_dpaa_bus.h           | 42 ++++++++++++++---------
 drivers/crypto/dpaa_sec/dpaa_sec.c        | 11 +++---
 drivers/event/dpaa/dpaa_eventdev.c        |  4 +--
 drivers/mempool/dpaa/dpaa_mempool.c       |  6 ++--
 drivers/net/dpaa/dpaa_ethdev.c            |  2 +-
 drivers/net/dpaa/dpaa_rxtx.c              |  4 +--
 9 files changed, 54 insertions(+), 46 deletions(-)

Patch
diff mbox series

diff --git a/doc/guides/rel_notes/release_20_08.rst b/doc/guides/rel_notes/release_20_08.rst
index d915fce12..b1e039d03 100644
--- a/doc/guides/rel_notes/release_20_08.rst
+++ b/doc/guides/rel_notes/release_20_08.rst
@@ -119,6 +119,12 @@  New Features
   See the :doc:`../sample_app_ug/l2_forward_real_virtual` for more
   details of this parameter usage.
 
+* **Updated NXP dpaa ethdev PMD.**
+
+  Updated the NXP dpaa ethdev  with new features and improvements, including:
+
+  * Added support to use datapath APIs from non-EAL pthread
+
 * **Updated NXP dpaa2 ethdev PMD.**
 
   Updated the NXP dpaa2 ethdev  with new features and improvements, including:
diff --git a/drivers/bus/dpaa/dpaa_bus.c b/drivers/bus/dpaa/dpaa_bus.c
index 6770fbc52..aa906c34e 100644
--- a/drivers/bus/dpaa/dpaa_bus.c
+++ b/drivers/bus/dpaa/dpaa_bus.c
@@ -52,8 +52,7 @@  unsigned int dpaa_svr_family;
 
 #define FSL_DPAA_BUS_NAME	dpaa_bus
 
-RTE_DEFINE_PER_LCORE(bool, dpaa_io);
-RTE_DEFINE_PER_LCORE(struct dpaa_portal_dqrr, held_bufs);
+RTE_DEFINE_PER_LCORE(struct dpaa_portal *, dpaa_io);
 
 struct fm_eth_port_cfg *
 dpaa_get_eth_port_cfg(int dev_id)
@@ -253,7 +252,6 @@  int rte_dpaa_portal_init(void *arg)
 {
 	unsigned int cpu, lcore = rte_lcore_id();
 	int ret;
-	struct dpaa_portal *dpaa_io_portal;
 
 	BUS_INIT_FUNC_TRACE();
 
@@ -288,20 +286,21 @@  int rte_dpaa_portal_init(void *arg)
 	DPAA_BUS_LOG(DEBUG, "QMAN thread initialized - CPU=%d lcore=%d",
 		     cpu, lcore);
 
-	dpaa_io_portal = rte_malloc(NULL, sizeof(struct dpaa_portal),
+	DPAA_PER_LCORE_PORTAL = rte_malloc(NULL, sizeof(struct dpaa_portal),
 				    RTE_CACHE_LINE_SIZE);
-	if (!dpaa_io_portal) {
+	if (!DPAA_PER_LCORE_PORTAL) {
 		DPAA_BUS_LOG(ERR, "Unable to allocate memory");
 		bman_thread_finish();
 		qman_thread_finish();
 		return -ENOMEM;
 	}
 
-	dpaa_io_portal->qman_idx = qman_get_portal_index();
-	dpaa_io_portal->bman_idx = bman_get_portal_index();
-	dpaa_io_portal->tid = syscall(SYS_gettid);
+	DPAA_PER_LCORE_PORTAL->qman_idx = qman_get_portal_index();
+	DPAA_PER_LCORE_PORTAL->bman_idx = bman_get_portal_index();
+	DPAA_PER_LCORE_PORTAL->tid = syscall(SYS_gettid);
 
-	ret = pthread_setspecific(dpaa_portal_key, (void *)dpaa_io_portal);
+	ret = pthread_setspecific(dpaa_portal_key,
+				  (void *)DPAA_PER_LCORE_PORTAL);
 	if (ret) {
 		DPAA_BUS_LOG(ERR, "pthread_setspecific failed on core %u"
 			     " (lcore=%u) with ret: %d", cpu, lcore, ret);
@@ -310,8 +309,6 @@  int rte_dpaa_portal_init(void *arg)
 		return ret;
 	}
 
-	RTE_PER_LCORE(dpaa_io) = true;
-
 	DPAA_BUS_LOG(DEBUG, "QMAN thread initialized");
 
 	return 0;
@@ -324,7 +321,7 @@  rte_dpaa_portal_fq_init(void *arg, struct qman_fq *fq)
 	u32 sdqcr;
 	int ret;
 
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		ret = rte_dpaa_portal_init(arg);
 		if (ret < 0) {
 			DPAA_BUS_LOG(ERR, "portal initialization failure");
@@ -367,8 +364,7 @@  dpaa_portal_finish(void *arg)
 
 	rte_free(dpaa_io_portal);
 	dpaa_io_portal = NULL;
-
-	RTE_PER_LCORE(dpaa_io) = false;
+	DPAA_PER_LCORE_PORTAL = NULL;
 }
 
 static int
diff --git a/drivers/bus/dpaa/rte_bus_dpaa_version.map b/drivers/bus/dpaa/rte_bus_dpaa_version.map
index 8069b05af..2defa7992 100644
--- a/drivers/bus/dpaa/rte_bus_dpaa_version.map
+++ b/drivers/bus/dpaa/rte_bus_dpaa_version.map
@@ -48,7 +48,6 @@  INTERNAL {
 	netcfg_acquire;
 	netcfg_release;
 	per_lcore_dpaa_io;
-	per_lcore_held_bufs;
 	qman_alloc_cgrid_range;
 	qman_alloc_pool_range;
 	qman_clear_irq;
diff --git a/drivers/bus/dpaa/rte_dpaa_bus.h b/drivers/bus/dpaa/rte_dpaa_bus.h
index 2a186d83f..25aff2d30 100644
--- a/drivers/bus/dpaa/rte_dpaa_bus.h
+++ b/drivers/bus/dpaa/rte_dpaa_bus.h
@@ -35,8 +35,6 @@ 
 
 extern unsigned int dpaa_svr_family;
 
-extern RTE_DEFINE_PER_LCORE(bool, dpaa_io);
-
 struct rte_dpaa_device;
 struct rte_dpaa_driver;
 
@@ -90,12 +88,38 @@  struct rte_dpaa_driver {
 	rte_dpaa_remove_t remove;
 };
 
+/* Create storage for dqrr entries per lcore */
+#define DPAA_PORTAL_DEQUEUE_DEPTH	16
+struct dpaa_portal_dqrr {
+	void *mbuf[DPAA_PORTAL_DEQUEUE_DEPTH];
+	uint64_t dqrr_held;
+	uint8_t dqrr_size;
+};
+
 struct dpaa_portal {
 	uint32_t bman_idx; /**< BMAN Portal ID*/
 	uint32_t qman_idx; /**< QMAN Portal ID*/
+	struct dpaa_portal_dqrr dpaa_held_bufs;
+	struct rte_crypto_op **dpaa_sec_ops;
+	int dpaa_sec_op_nb;
 	uint64_t tid;/**< Parent Thread id for this portal */
 };
 
+RTE_DECLARE_PER_LCORE(struct dpaa_portal *, dpaa_io);
+
+#define DPAA_PER_LCORE_PORTAL \
+	RTE_PER_LCORE(dpaa_io)
+#define DPAA_PER_LCORE_DQRR_SIZE \
+	RTE_PER_LCORE(dpaa_io)->dpaa_held_bufs.dqrr_size
+#define DPAA_PER_LCORE_DQRR_HELD \
+	RTE_PER_LCORE(dpaa_io)->dpaa_held_bufs.dqrr_held
+#define DPAA_PER_LCORE_DQRR_MBUF(i) \
+	RTE_PER_LCORE(dpaa_io)->dpaa_held_bufs.mbuf[i]
+#define DPAA_PER_LCORE_RTE_CRYPTO_OP \
+	RTE_PER_LCORE(dpaa_io)->dpaa_sec_ops
+#define DPAA_PER_LCORE_DPAA_SEC_OP_NB \
+	RTE_PER_LCORE(dpaa_io)->dpaa_sec_op_nb
+
 /* Various structures representing contiguous memory maps */
 struct dpaa_memseg {
 	TAILQ_ENTRY(dpaa_memseg) next;
@@ -200,20 +224,6 @@  RTE_INIT(dpaainitfn_ ##nm) \
 } \
 RTE_PMD_EXPORT_NAME(nm, __COUNTER__)
 
-/* Create storage for dqrr entries per lcore */
-#define DPAA_PORTAL_DEQUEUE_DEPTH	16
-struct dpaa_portal_dqrr {
-	void *mbuf[DPAA_PORTAL_DEQUEUE_DEPTH];
-	uint64_t dqrr_held;
-	uint8_t dqrr_size;
-};
-
-RTE_DECLARE_PER_LCORE(struct dpaa_portal_dqrr, held_bufs);
-
-#define DPAA_PER_LCORE_DQRR_SIZE       RTE_PER_LCORE(held_bufs).dqrr_size
-#define DPAA_PER_LCORE_DQRR_HELD       RTE_PER_LCORE(held_bufs).dqrr_held
-#define DPAA_PER_LCORE_DQRR_MBUF(i)    RTE_PER_LCORE(held_bufs).mbuf[i]
-
 __rte_internal
 struct fm_eth_port_cfg *dpaa_get_eth_port_cfg(int dev_id);
 
diff --git a/drivers/crypto/dpaa_sec/dpaa_sec.c b/drivers/crypto/dpaa_sec/dpaa_sec.c
index d9fa8bb36..8fcd57373 100644
--- a/drivers/crypto/dpaa_sec/dpaa_sec.c
+++ b/drivers/crypto/dpaa_sec/dpaa_sec.c
@@ -45,9 +45,6 @@ 
 
 static uint8_t cryptodev_driver_id;
 
-static __thread struct rte_crypto_op **dpaa_sec_ops;
-static __thread int dpaa_sec_op_nb;
-
 static int
 dpaa_sec_attach_sess_q(struct dpaa_sec_qp *qp, dpaa_sec_session *sess);
 
@@ -143,7 +140,7 @@  dqrr_out_fq_cb_rx(struct qman_portal *qm __always_unused,
 	struct dpaa_sec_job *job;
 	struct dpaa_sec_op_ctx *ctx;
 
-	if (dpaa_sec_op_nb >= DPAA_SEC_BURST)
+	if (DPAA_PER_LCORE_DPAA_SEC_OP_NB >= DPAA_SEC_BURST)
 		return qman_cb_dqrr_defer;
 
 	if (!(dqrr->stat & QM_DQRR_STAT_FD_VALID))
@@ -174,7 +171,7 @@  dqrr_out_fq_cb_rx(struct qman_portal *qm __always_unused,
 		}
 		mbuf->data_len = len;
 	}
-	dpaa_sec_ops[dpaa_sec_op_nb++] = ctx->op;
+	DPAA_PER_LCORE_RTE_CRYPTO_OP[DPAA_PER_LCORE_DPAA_SEC_OP_NB++] = ctx->op;
 	dpaa_sec_op_ending(ctx);
 
 	return qman_cb_dqrr_consume;
@@ -2301,7 +2298,7 @@  dpaa_sec_attach_sess_q(struct dpaa_sec_qp *qp, dpaa_sec_session *sess)
 		DPAA_SEC_ERR("Unable to prepare sec cdb");
 		return ret;
 	}
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		ret = rte_dpaa_portal_init((void *)0);
 		if (ret) {
 			DPAA_SEC_ERR("Failure in affining portal");
@@ -3463,7 +3460,7 @@  cryptodev_dpaa_sec_probe(struct rte_dpaa_driver *dpaa_drv __rte_unused,
 		}
 	}
 
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		retval = rte_dpaa_portal_init((void *)1);
 		if (retval) {
 			DPAA_SEC_ERR("Unable to initialize portal");
diff --git a/drivers/event/dpaa/dpaa_eventdev.c b/drivers/event/dpaa/dpaa_eventdev.c
index e78728b7e..a3c138b7a 100644
--- a/drivers/event/dpaa/dpaa_eventdev.c
+++ b/drivers/event/dpaa/dpaa_eventdev.c
@@ -179,7 +179,7 @@  dpaa_event_dequeue_burst(void *port, struct rte_event ev[],
 	struct dpaa_port *portal = (struct dpaa_port *)port;
 	struct rte_mbuf *mbuf;
 
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		/* Affine current thread context to a qman portal */
 		ret = rte_dpaa_portal_init((void *)0);
 		if (ret) {
@@ -251,7 +251,7 @@  dpaa_event_dequeue_burst_intr(void *port, struct rte_event ev[],
 	struct dpaa_port *portal = (struct dpaa_port *)port;
 	struct rte_mbuf *mbuf;
 
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		/* Affine current thread context to a qman portal */
 		ret = rte_dpaa_portal_init((void *)0);
 		if (ret) {
diff --git a/drivers/mempool/dpaa/dpaa_mempool.c b/drivers/mempool/dpaa/dpaa_mempool.c
index 8d1da8028..e6b06f057 100644
--- a/drivers/mempool/dpaa/dpaa_mempool.c
+++ b/drivers/mempool/dpaa/dpaa_mempool.c
@@ -53,7 +53,7 @@  dpaa_mbuf_create_pool(struct rte_mempool *mp)
 
 	MEMPOOL_INIT_FUNC_TRACE();
 
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		ret = rte_dpaa_portal_init((void *)0);
 		if (ret) {
 			DPAA_MEMPOOL_ERR(
@@ -169,7 +169,7 @@  dpaa_mbuf_free_bulk(struct rte_mempool *pool,
 	DPAA_MEMPOOL_DPDEBUG("Request to free %d buffers in bpid = %d",
 			     n, bp_info->bpid);
 
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		ret = rte_dpaa_portal_init((void *)0);
 		if (ret) {
 			DPAA_MEMPOOL_ERR("rte_dpaa_portal_init failed with ret: %d",
@@ -224,7 +224,7 @@  dpaa_mbuf_alloc_bulk(struct rte_mempool *pool,
 		return -1;
 	}
 
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		ret = rte_dpaa_portal_init((void *)0);
 		if (ret) {
 			DPAA_MEMPOOL_ERR("rte_dpaa_portal_init failed with ret: %d",
diff --git a/drivers/net/dpaa/dpaa_ethdev.c b/drivers/net/dpaa/dpaa_ethdev.c
index 6c94fd396..c9f828a7c 100644
--- a/drivers/net/dpaa/dpaa_ethdev.c
+++ b/drivers/net/dpaa/dpaa_ethdev.c
@@ -1707,7 +1707,7 @@  rte_dpaa_probe(struct rte_dpaa_driver *dpaa_drv __rte_unused,
 		is_global_init = 1;
 	}
 
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		ret = rte_dpaa_portal_init((void *)1);
 		if (ret) {
 			DPAA_PMD_ERR("Unable to initialize portal");
diff --git a/drivers/net/dpaa/dpaa_rxtx.c b/drivers/net/dpaa/dpaa_rxtx.c
index 819cad7c6..5303c9b76 100644
--- a/drivers/net/dpaa/dpaa_rxtx.c
+++ b/drivers/net/dpaa/dpaa_rxtx.c
@@ -670,7 +670,7 @@  uint16_t dpaa_eth_queue_rx(void *q,
 	if (likely(fq->is_static))
 		return dpaa_eth_queue_portal_rx(fq, bufs, nb_bufs);
 
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		ret = rte_dpaa_portal_init((void *)0);
 		if (ret) {
 			DPAA_PMD_ERR("Failure in affining portal");
@@ -970,7 +970,7 @@  dpaa_eth_queue_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
 	int ret, realloc_mbuf = 0;
 	uint32_t seqn, index, flags[DPAA_TX_BURST_SIZE] = {0};
 
-	if (unlikely(!RTE_PER_LCORE(dpaa_io))) {
+	if (unlikely(!DPAA_PER_LCORE_PORTAL)) {
 		ret = rte_dpaa_portal_init((void *)0);
 		if (ret) {
 			DPAA_PMD_ERR("Failure in affining portal");