net/sfc: support event queue
authorAndrew Rybchenko <arybchenko@solarflare.com>
Tue, 29 Nov 2016 16:19:09 +0000 (16:19 +0000)
committerFerruh Yigit <ferruh.yigit@intel.com>
Tue, 17 Jan 2017 18:39:26 +0000 (19:39 +0100)
Signed-off-by: Andrew Rybchenko <arybchenko@solarflare.com>
Reviewed-by: Andy Moreton <amoreton@solarflare.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
drivers/net/sfc/Makefile
drivers/net/sfc/sfc.c
drivers/net/sfc/sfc.h
drivers/net/sfc/sfc_ev.c [new file with mode: 0644]
drivers/net/sfc/sfc_ev.h [new file with mode: 0644]

index d19497c..3286fd8 100644 (file)
@@ -83,6 +83,7 @@ SRCS-$(CONFIG_RTE_LIBRTE_SFC_EFX_PMD) += sfc_kvargs.c
 SRCS-$(CONFIG_RTE_LIBRTE_SFC_EFX_PMD) += sfc.c
 SRCS-$(CONFIG_RTE_LIBRTE_SFC_EFX_PMD) += sfc_mcdi.c
 SRCS-$(CONFIG_RTE_LIBRTE_SFC_EFX_PMD) += sfc_intr.c
+SRCS-$(CONFIG_RTE_LIBRTE_SFC_EFX_PMD) += sfc_ev.c
 
 VPATH += $(SRCDIR)/base
 
index 8d18def..38a7e52 100644 (file)
@@ -36,6 +36,7 @@
 
 #include "sfc.h"
 #include "sfc_log.h"
+#include "sfc_ev.h"
 
 
 int
@@ -261,10 +262,17 @@ sfc_start(struct sfc_adapter *sa)
        if (rc != 0)
                goto fail_intr_start;
 
+       rc = sfc_ev_start(sa);
+       if (rc != 0)
+               goto fail_ev_start;
+
        sa->state = SFC_ADAPTER_STARTED;
        sfc_log_init(sa, "done");
        return 0;
 
+fail_ev_start:
+       sfc_intr_stop(sa);
+
 fail_intr_start:
        efx_nic_fini(sa->nic);
 
@@ -297,6 +305,7 @@ sfc_stop(struct sfc_adapter *sa)
 
        sa->state = SFC_ADAPTER_STOPPING;
 
+       sfc_ev_stop(sa);
        sfc_intr_stop(sa);
        efx_nic_fini(sa->nic);
 
@@ -324,10 +333,17 @@ sfc_configure(struct sfc_adapter *sa)
        if (rc != 0)
                goto fail_intr_init;
 
+       rc = sfc_ev_init(sa);
+       if (rc != 0)
+               goto fail_ev_init;
+
        sa->state = SFC_ADAPTER_CONFIGURED;
        sfc_log_init(sa, "done");
        return 0;
 
+fail_ev_init:
+       sfc_intr_fini(sa);
+
 fail_intr_init:
 fail_check_conf:
        sa->state = SFC_ADAPTER_INITIALIZED;
@@ -345,6 +361,7 @@ sfc_close(struct sfc_adapter *sa)
        SFC_ASSERT(sa->state == SFC_ADAPTER_CONFIGURED);
        sa->state = SFC_ADAPTER_CLOSING;
 
+       sfc_ev_fini(sa);
        sfc_intr_fini(sa);
 
        sa->state = SFC_ADAPTER_INITIALIZED;
index 44a98f4..f6d0b81 100644 (file)
@@ -116,6 +116,8 @@ struct sfc_intr {
        efx_intr_type_t                 type;
 };
 
+struct sfc_evq_info;
+
 /* Adapter private data */
 struct sfc_adapter {
        /*
@@ -140,6 +142,11 @@ struct sfc_adapter {
 
        unsigned int                    rxq_max;
        unsigned int                    txq_max;
+
+       unsigned int                    evq_count;
+       struct sfc_evq_info             *evq_info;
+
+       unsigned int                    mgmt_evq_index;
 };
 
 /*
diff --git a/drivers/net/sfc/sfc_ev.c b/drivers/net/sfc/sfc_ev.c
new file mode 100644 (file)
index 0000000..2ee579a
--- /dev/null
@@ -0,0 +1,482 @@
+/*-
+ * Copyright (c) 2016 Solarflare Communications Inc.
+ * All rights reserved.
+ *
+ * This software was jointly developed between OKTET Labs (under contract
+ * for Solarflare) and Solarflare Communications, Inc.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+ * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+ * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
+ * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
+ * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
+ * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
+ * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <rte_debug.h>
+#include <rte_cycles.h>
+
+#include "efx.h"
+
+#include "sfc.h"
+#include "sfc_debug.h"
+#include "sfc_log.h"
+#include "sfc_ev.h"
+
+
+/* Initial delay when waiting for event queue init complete event */
+#define SFC_EVQ_INIT_BACKOFF_START_US  (1)
+/* Maximum delay between event queue polling attempts */
+#define SFC_EVQ_INIT_BACKOFF_MAX_US    (10 * 1000)
+/* Event queue init approx timeout */
+#define SFC_EVQ_INIT_TIMEOUT_US                (2 * US_PER_S)
+
+
+static boolean_t
+sfc_ev_initialized(void *arg)
+{
+       struct sfc_evq *evq = arg;
+
+       /* Init done events may be duplicated on SFN7xxx (SFC bug 31631) */
+       SFC_ASSERT(evq->init_state == SFC_EVQ_STARTING ||
+                  evq->init_state == SFC_EVQ_STARTED);
+
+       evq->init_state = SFC_EVQ_STARTED;
+
+       return B_FALSE;
+}
+
+static boolean_t
+sfc_ev_rx(void *arg, __rte_unused uint32_t label, __rte_unused uint32_t id,
+         __rte_unused uint32_t size, __rte_unused uint16_t flags)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected Rx event", evq->evq_index);
+       return B_TRUE;
+}
+
+static boolean_t
+sfc_ev_tx(void *arg, __rte_unused uint32_t label, __rte_unused uint32_t id)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected Tx event", evq->evq_index);
+       return B_TRUE;
+}
+
+static boolean_t
+sfc_ev_exception(void *arg, __rte_unused uint32_t code,
+                __rte_unused uint32_t data)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected exception event",
+               evq->evq_index);
+       return B_TRUE;
+}
+
+static boolean_t
+sfc_ev_rxq_flush_done(void *arg, __rte_unused uint32_t rxq_hw_index)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected Rx flush done event",
+               evq->evq_index);
+       return B_TRUE;
+}
+
+static boolean_t
+sfc_ev_rxq_flush_failed(void *arg, __rte_unused uint32_t rxq_hw_index)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected Rx flush failed event",
+               evq->evq_index);
+       return B_TRUE;
+}
+
+static boolean_t
+sfc_ev_txq_flush_done(void *arg, __rte_unused uint32_t txq_hw_index)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected Tx flush done event",
+               evq->evq_index);
+       return B_TRUE;
+}
+
+static boolean_t
+sfc_ev_software(void *arg, uint16_t magic)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected software event magic=%#.4x",
+               evq->evq_index, magic);
+       return B_TRUE;
+}
+
+static boolean_t
+sfc_ev_sram(void *arg, uint32_t code)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected SRAM event code=%u",
+               evq->evq_index, code);
+       return B_TRUE;
+}
+
+static boolean_t
+sfc_ev_wake_up(void *arg, uint32_t index)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected wake up event index=%u",
+               evq->evq_index, index);
+       return B_TRUE;
+}
+
+static boolean_t
+sfc_ev_timer(void *arg, uint32_t index)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected timer event index=%u",
+               evq->evq_index, index);
+       return B_TRUE;
+}
+
+static boolean_t
+sfc_ev_link_change(void *arg, __rte_unused efx_link_mode_t link_mode)
+{
+       struct sfc_evq *evq = arg;
+
+       sfc_err(evq->sa, "EVQ %u unexpected link change",
+               evq->evq_index);
+       return B_TRUE;
+}
+
+static const efx_ev_callbacks_t sfc_ev_callbacks = {
+       .eec_initialized        = sfc_ev_initialized,
+       .eec_rx                 = sfc_ev_rx,
+       .eec_tx                 = sfc_ev_tx,
+       .eec_exception          = sfc_ev_exception,
+       .eec_rxq_flush_done     = sfc_ev_rxq_flush_done,
+       .eec_rxq_flush_failed   = sfc_ev_rxq_flush_failed,
+       .eec_txq_flush_done     = sfc_ev_txq_flush_done,
+       .eec_software           = sfc_ev_software,
+       .eec_sram               = sfc_ev_sram,
+       .eec_wake_up            = sfc_ev_wake_up,
+       .eec_timer              = sfc_ev_timer,
+       .eec_link_change        = sfc_ev_link_change,
+};
+
+
+void
+sfc_ev_qpoll(struct sfc_evq *evq)
+{
+       SFC_ASSERT(evq->init_state == SFC_EVQ_STARTED ||
+                  evq->init_state == SFC_EVQ_STARTING);
+
+       /* Synchronize the DMA memory for reading not required */
+
+       efx_ev_qpoll(evq->common, &evq->read_ptr, &sfc_ev_callbacks, evq);
+
+       /* Poll-mode driver does not re-prime the event queue for interrupts */
+}
+
+int
+sfc_ev_qprime(struct sfc_evq *evq)
+{
+       SFC_ASSERT(evq->init_state == SFC_EVQ_STARTED);
+       return efx_ev_qprime(evq->common, evq->read_ptr);
+}
+
+int
+sfc_ev_qstart(struct sfc_adapter *sa, unsigned int sw_index)
+{
+       const struct sfc_evq_info *evq_info;
+       struct sfc_evq *evq;
+       efsys_mem_t *esmp;
+       unsigned int total_delay_us;
+       unsigned int delay_us;
+       int rc;
+
+       sfc_log_init(sa, "sw_index=%u", sw_index);
+
+       evq_info = &sa->evq_info[sw_index];
+       evq = evq_info->evq;
+       esmp = &evq->mem;
+
+       /* Clear all events */
+       (void)memset((void *)esmp->esm_base, 0xff,
+                    EFX_EVQ_SIZE(evq_info->entries));
+
+       /* Create the common code event queue */
+       rc = efx_ev_qcreate(sa->nic, sw_index, esmp, evq_info->entries,
+                           0 /* unused on EF10 */, 0,
+                           EFX_EVQ_FLAGS_TYPE_THROUGHPUT |
+                           EFX_EVQ_FLAGS_NOTIFY_DISABLED,
+                           &evq->common);
+       if (rc != 0)
+               goto fail_ev_qcreate;
+
+       evq->init_state = SFC_EVQ_STARTING;
+
+       /* Wait for the initialization event */
+       total_delay_us = 0;
+       delay_us = SFC_EVQ_INIT_BACKOFF_START_US;
+       do {
+               (void)sfc_ev_qpoll(evq);
+
+               /* Check to see if the initialization complete indication
+                * posted by the hardware.
+                */
+               if (evq->init_state == SFC_EVQ_STARTED)
+                       goto done;
+
+               /* Give event queue some time to init */
+               rte_delay_us(delay_us);
+
+               total_delay_us += delay_us;
+
+               /* Exponential backoff */
+               delay_us *= 2;
+               if (delay_us > SFC_EVQ_INIT_BACKOFF_MAX_US)
+                       delay_us = SFC_EVQ_INIT_BACKOFF_MAX_US;
+
+       } while (total_delay_us < SFC_EVQ_INIT_TIMEOUT_US);
+
+       rc = ETIMEDOUT;
+       goto fail_timedout;
+
+done:
+       return 0;
+
+fail_timedout:
+       evq->init_state = SFC_EVQ_INITIALIZED;
+       efx_ev_qdestroy(evq->common);
+
+fail_ev_qcreate:
+       sfc_log_init(sa, "failed %d", rc);
+       return rc;
+}
+
+void
+sfc_ev_qstop(struct sfc_adapter *sa, unsigned int sw_index)
+{
+       const struct sfc_evq_info *evq_info;
+       struct sfc_evq *evq;
+
+       sfc_log_init(sa, "sw_index=%u", sw_index);
+
+       SFC_ASSERT(sw_index < sa->evq_count);
+
+       evq_info = &sa->evq_info[sw_index];
+       evq = evq_info->evq;
+
+       if (evq == NULL || evq->init_state != SFC_EVQ_STARTED)
+               return;
+
+       evq->init_state = SFC_EVQ_INITIALIZED;
+       evq->read_ptr = 0;
+       evq->exception = B_FALSE;
+
+       efx_ev_qdestroy(evq->common);
+}
+
+int
+sfc_ev_start(struct sfc_adapter *sa)
+{
+       int rc;
+
+       sfc_log_init(sa, "entry");
+
+       rc = efx_ev_init(sa->nic);
+       if (rc != 0)
+               goto fail_ev_init;
+
+       /*
+        * Rx/Tx event queues are started/stopped when corresponding queue
+        * is started/stopped.
+        */
+
+       return 0;
+
+fail_ev_init:
+       sfc_log_init(sa, "failed %d", rc);
+       return rc;
+}
+
+void
+sfc_ev_stop(struct sfc_adapter *sa)
+{
+       unsigned int sw_index;
+
+       sfc_log_init(sa, "entry");
+
+       /* Make sure that all event queues are stopped */
+       sw_index = sa->evq_count;
+       while (sw_index-- > 0)
+               sfc_ev_qstop(sa, sw_index);
+
+       efx_ev_fini(sa->nic);
+}
+
+int
+sfc_ev_qinit(struct sfc_adapter *sa, unsigned int sw_index,
+            unsigned int entries, int socket_id)
+{
+       struct sfc_evq_info *evq_info;
+       struct sfc_evq *evq;
+       int rc;
+
+       sfc_log_init(sa, "sw_index=%u", sw_index);
+
+       evq_info = &sa->evq_info[sw_index];
+
+       SFC_ASSERT(rte_is_power_of_2(entries));
+       SFC_ASSERT(entries <= evq_info->max_entries);
+       evq_info->entries = entries;
+
+       evq = rte_zmalloc_socket("sfc-evq", sizeof(*evq), RTE_CACHE_LINE_SIZE,
+                                socket_id);
+       if (evq == NULL)
+               return ENOMEM;
+
+       evq->sa = sa;
+       evq->evq_index = sw_index;
+
+       /* Allocate DMA space */
+       rc = sfc_dma_alloc(sa, "evq", sw_index, EFX_EVQ_SIZE(evq_info->entries),
+                          socket_id, &evq->mem);
+       if (rc != 0)
+               return rc;
+
+       evq->init_state = SFC_EVQ_INITIALIZED;
+
+       evq_info->evq = evq;
+
+       return 0;
+}
+
+void
+sfc_ev_qfini(struct sfc_adapter *sa, unsigned int sw_index)
+{
+       struct sfc_evq *evq;
+
+       sfc_log_init(sa, "sw_index=%u", sw_index);
+
+       evq = sa->evq_info[sw_index].evq;
+
+       SFC_ASSERT(evq->init_state == SFC_EVQ_INITIALIZED);
+
+       sa->evq_info[sw_index].evq = NULL;
+
+       sfc_dma_free(sa, &evq->mem);
+
+       rte_free(evq);
+}
+
+static int
+sfc_ev_qinit_info(struct sfc_adapter *sa, unsigned int sw_index)
+{
+       struct sfc_evq_info *evq_info = &sa->evq_info[sw_index];
+       unsigned int max_entries;
+
+       sfc_log_init(sa, "sw_index=%u", sw_index);
+
+       max_entries = sfc_evq_max_entries(sa, sw_index);
+       SFC_ASSERT(rte_is_power_of_2(max_entries));
+
+       evq_info->max_entries = max_entries;
+
+       return 0;
+}
+
+static void
+sfc_ev_qfini_info(struct sfc_adapter *sa, unsigned int sw_index)
+{
+       sfc_log_init(sa, "sw_index=%u", sw_index);
+
+       /* Nothing to cleanup */
+}
+
+int
+sfc_ev_init(struct sfc_adapter *sa)
+{
+       int rc;
+       unsigned int sw_index;
+
+       sfc_log_init(sa, "entry");
+
+       sa->evq_count = sfc_ev_qcount(sa);
+       sa->mgmt_evq_index = 0;
+
+       /* Allocate EVQ info array */
+       rc = ENOMEM;
+       sa->evq_info = rte_calloc_socket("sfc-evqs", sa->evq_count,
+                                        sizeof(struct sfc_evq_info), 0,
+                                        sa->socket_id);
+       if (sa->evq_info == NULL)
+               goto fail_evqs_alloc;
+
+       for (sw_index = 0; sw_index < sa->evq_count; ++sw_index) {
+               rc = sfc_ev_qinit_info(sa, sw_index);
+               if (rc != 0)
+                       goto fail_ev_qinit_info;
+       }
+
+       /*
+        * Rx/Tx event queues are created/destroyed when corresponding
+        * Rx/Tx queue is created/destroyed.
+        */
+
+       return 0;
+
+fail_ev_qinit_info:
+       while (sw_index-- > 0)
+               sfc_ev_qfini_info(sa, sw_index);
+
+       rte_free(sa->evq_info);
+       sa->evq_info = NULL;
+
+fail_evqs_alloc:
+       sa->evq_count = 0;
+       sfc_log_init(sa, "failed %d", rc);
+       return rc;
+}
+
+void
+sfc_ev_fini(struct sfc_adapter *sa)
+{
+       int sw_index;
+
+       sfc_log_init(sa, "entry");
+
+       /* Cleanup all event queues */
+       sw_index = sa->evq_count;
+       while (--sw_index >= 0) {
+               if (sa->evq_info[sw_index].evq != NULL)
+                       sfc_ev_qfini(sa, sw_index);
+               sfc_ev_qfini_info(sa, sw_index);
+       }
+
+       rte_free(sa->evq_info);
+       sa->evq_info = NULL;
+       sa->evq_count = 0;
+}
diff --git a/drivers/net/sfc/sfc_ev.h b/drivers/net/sfc/sfc_ev.h
new file mode 100644 (file)
index 0000000..11478e9
--- /dev/null
@@ -0,0 +1,139 @@
+/*-
+ * Copyright (c) 2016 Solarflare Communications Inc.
+ * All rights reserved.
+ *
+ * This software was jointly developed between OKTET Labs (under contract
+ * for Solarflare) and Solarflare Communications, Inc.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ * 1. Redistributions of source code must retain the above copyright notice,
+ *    this list of conditions and the following disclaimer.
+ * 2. Redistributions in binary form must reproduce the above copyright notice,
+ *    this list of conditions and the following disclaimer in the documentation
+ *    and/or other materials provided with the distribution.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO,
+ * THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
+ * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR
+ * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
+ * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
+ * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS;
+ * OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY,
+ * WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR
+ * OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE,
+ * EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef _SFC_EV_H_
+#define _SFC_EV_H_
+
+#include "efx.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* Number of entries in the management event queue */
+#define SFC_MGMT_EVQ_ENTRIES   (EFX_EVQ_MINNEVS)
+
+struct sfc_adapter;
+
+enum sfc_evq_state {
+       SFC_EVQ_UNINITIALIZED = 0,
+       SFC_EVQ_INITIALIZED,
+       SFC_EVQ_STARTING,
+       SFC_EVQ_STARTED,
+
+       SFC_EVQ_NSTATES
+};
+
+struct sfc_evq {
+       /* Used on datapath */
+       efx_evq_t               *common;
+       unsigned int            read_ptr;
+       boolean_t               exception;
+       efsys_mem_t             mem;
+
+       /* Not used on datapath */
+       struct sfc_adapter      *sa;
+       unsigned int            evq_index;
+       enum sfc_evq_state      init_state;
+};
+
+struct sfc_evq_info {
+       /* Maximum number of EVQ entries taken into account when buffer
+        * table space is allocated.
+        */
+       unsigned int            max_entries;
+       /* Real number of EVQ entries, less or equal to max_entries */
+       unsigned int            entries;
+       /* NUMA-aware EVQ data structure used on datapath */
+       struct sfc_evq          *evq;
+};
+
+/*
+ * Functions below define event queue to transmit/receive queue and vice
+ * versa mapping.
+ */
+
+static inline unsigned int
+sfc_ev_qcount(struct sfc_adapter *sa)
+{
+       const struct rte_eth_dev_data *dev_data = sa->eth_dev->data;
+
+       /*
+        * One management EVQ for global events.
+        * Own EVQ for each Tx and Rx queue.
+        */
+       return 1 + dev_data->nb_rx_queues + dev_data->nb_tx_queues;
+}
+
+static inline unsigned int
+sfc_evq_max_entries(struct sfc_adapter *sa, unsigned int sw_index)
+{
+       unsigned int max_entries;
+
+       if (sw_index == sa->mgmt_evq_index)
+               max_entries = SFC_MGMT_EVQ_ENTRIES;
+       else if (sw_index <= sa->eth_dev->data->nb_rx_queues)
+               max_entries = EFX_RXQ_MAXNDESCS;
+       else
+               max_entries = efx_nic_cfg_get(sa->nic)->enc_txq_max_ndescs;
+
+       return max_entries;
+}
+
+static inline unsigned int
+sfc_evq_index_by_rxq_sw_index(__rte_unused struct sfc_adapter *sa,
+                             unsigned int rxq_sw_index)
+{
+       return 1 + rxq_sw_index;
+}
+
+static inline unsigned int
+sfc_evq_index_by_txq_sw_index(struct sfc_adapter *sa, unsigned int txq_sw_index)
+{
+       return 1 + sa->eth_dev->data->nb_rx_queues + txq_sw_index;
+}
+
+int sfc_ev_init(struct sfc_adapter *sa);
+void sfc_ev_fini(struct sfc_adapter *sa);
+int sfc_ev_start(struct sfc_adapter *sa);
+void sfc_ev_stop(struct sfc_adapter *sa);
+
+int sfc_ev_qinit(struct sfc_adapter *sa, unsigned int sw_index,
+                unsigned int entries, int socket_id);
+void sfc_ev_qfini(struct sfc_adapter *sa, unsigned int sw_index);
+int sfc_ev_qstart(struct sfc_adapter *sa, unsigned int sw_index);
+void sfc_ev_qstop(struct sfc_adapter *sa, unsigned int sw_index);
+
+int sfc_ev_qprime(struct sfc_evq *evq);
+void sfc_ev_qpoll(struct sfc_evq *evq);
+
+#ifdef __cplusplus
+}
+#endif
+#endif /* _SFC_EV_H_ */