*/
#include <rte_string_fns.h>
-#include <rte_ether.h>
-#include <ethdev_driver.h>
-#include <ethdev_pci.h>
-#include <rte_tcp.h>
-#include <rte_atomic.h>
-#include <rte_dev.h>
#include <rte_errno.h>
#include <rte_version.h>
#include <rte_net.h>
#include <ena_eth_io_defs.h>
#define DRV_MODULE_VER_MAJOR 2
-#define DRV_MODULE_VER_MINOR 3
+#define DRV_MODULE_VER_MINOR 4
#define DRV_MODULE_VER_SUBMINOR 0
-#define ENA_IO_TXQ_IDX(q) (2 * (q))
-#define ENA_IO_RXQ_IDX(q) (2 * (q) + 1)
-/*reverse version of ENA_IO_RXQ_IDX*/
-#define ENA_IO_RXQ_IDX_REV(q) ((q - 1) / 2)
-
#define __MERGE_64B_H_L(h, l) (((uint64_t)h << 32) | l)
-#define TEST_BIT(val, bit_shift) (val & (1UL << bit_shift))
#define GET_L4_HDR_LEN(mbuf) \
((rte_pktmbuf_mtod_offset(mbuf, struct rte_tcp_hdr *, \
mbuf->l3_len + mbuf->l2_len)->data_off) >> 4)
-#define ENA_RX_RSS_TABLE_LOG_SIZE 7
-#define ENA_RX_RSS_TABLE_SIZE (1 << ENA_RX_RSS_TABLE_LOG_SIZE)
-#define ENA_HASH_KEY_SIZE 40
#define ETH_GSTRING_LEN 32
#define ARRAY_SIZE(x) RTE_DIM(x)
static void ena_tx_queue_release_bufs(struct ena_ring *ring);
static int ena_link_update(struct rte_eth_dev *dev,
int wait_to_complete);
-static int ena_create_io_queue(struct ena_ring *ring);
+static int ena_create_io_queue(struct rte_eth_dev *dev, struct ena_ring *ring);
static void ena_queue_stop(struct ena_ring *ring);
static void ena_queue_stop_all(struct rte_eth_dev *dev,
enum ena_ring_type ring_type);
-static int ena_queue_start(struct ena_ring *ring);
+static int ena_queue_start(struct rte_eth_dev *dev, struct ena_ring *ring);
static int ena_queue_start_all(struct rte_eth_dev *dev,
enum ena_ring_type ring_type);
static void ena_stats_restart(struct rte_eth_dev *dev);
static int ena_infos_get(struct rte_eth_dev *dev,
struct rte_eth_dev_info *dev_info);
-static int ena_rss_reta_update(struct rte_eth_dev *dev,
- struct rte_eth_rss_reta_entry64 *reta_conf,
- uint16_t reta_size);
-static int ena_rss_reta_query(struct rte_eth_dev *dev,
- struct rte_eth_rss_reta_entry64 *reta_conf,
- uint16_t reta_size);
static void ena_interrupt_handler_rte(void *cb_arg);
static void ena_timer_wd_callback(struct rte_timer *timer, void *arg);
static void ena_destroy_device(struct rte_eth_dev *eth_dev);
static int ena_parse_devargs(struct ena_adapter *adapter,
struct rte_devargs *devargs);
static int ena_copy_eni_stats(struct ena_adapter *adapter);
+static int ena_setup_rx_intr(struct rte_eth_dev *dev);
+static int ena_rx_queue_intr_enable(struct rte_eth_dev *dev,
+ uint16_t queue_id);
+static int ena_rx_queue_intr_disable(struct rte_eth_dev *dev,
+ uint16_t queue_id);
static const struct eth_dev_ops ena_dev_ops = {
.dev_configure = ena_dev_configure,
.dev_reset = ena_dev_reset,
.reta_update = ena_rss_reta_update,
.reta_query = ena_rss_reta_query,
+ .rx_queue_intr_enable = ena_rx_queue_intr_enable,
+ .rx_queue_intr_disable = ena_rx_queue_intr_disable,
+ .rss_hash_update = ena_rss_hash_update,
+ .rss_hash_conf_get = ena_rss_hash_conf_get,
};
-void ena_rss_key_fill(void *key, size_t size)
-{
- static bool key_generated;
- static uint8_t default_key[ENA_HASH_KEY_SIZE];
- size_t i;
-
- RTE_ASSERT(size <= ENA_HASH_KEY_SIZE);
-
- if (!key_generated) {
- for (i = 0; i < ENA_HASH_KEY_SIZE; ++i)
- default_key[i] = rte_rand() & 0xff;
- key_generated = true;
- }
-
- rte_memcpy(key, default_key, size);
-}
-
static inline void ena_rx_mbuf_prepare(struct rte_mbuf *mbuf,
- struct ena_com_rx_ctx *ena_rx_ctx)
+ struct ena_com_rx_ctx *ena_rx_ctx,
+ bool fill_hash)
{
uint64_t ol_flags = 0;
uint32_t packet_type = 0;
else
ol_flags |= PKT_RX_L4_CKSUM_GOOD;
- if (likely((packet_type & ENA_PTYPE_HAS_HASH) && !ena_rx_ctx->frag)) {
+ if (fill_hash &&
+ likely((packet_type & ENA_PTYPE_HAS_HASH) && !ena_rx_ctx->frag)) {
ol_flags |= PKT_RX_RSS_HASH;
mbuf->hash.rss = ena_rx_ctx->hash;
}
host_info->num_cpus = rte_lcore_count();
host_info->driver_supported_features =
- ENA_ADMIN_HOST_INFO_RX_OFFSET_MASK;
+ ENA_ADMIN_HOST_INFO_RX_OFFSET_MASK |
+ ENA_ADMIN_HOST_INFO_RSS_CONFIGURABLE_FUNCTION_KEY_MASK;
rc = ena_com_set_host_attributes(ena_dev);
if (rc) {
return rc;
}
-static int ena_rss_reta_update(struct rte_eth_dev *dev,
- struct rte_eth_rss_reta_entry64 *reta_conf,
- uint16_t reta_size)
-{
- struct ena_adapter *adapter = dev->data->dev_private;
- struct ena_com_dev *ena_dev = &adapter->ena_dev;
- int rc, i;
- u16 entry_value;
- int conf_idx;
- int idx;
-
- if ((reta_size == 0) || (reta_conf == NULL))
- return -EINVAL;
-
- if (reta_size > ENA_RX_RSS_TABLE_SIZE) {
- PMD_DRV_LOG(WARNING,
- "Requested indirection table size (%d) is bigger than supported: %d\n",
- reta_size, ENA_RX_RSS_TABLE_SIZE);
- return -EINVAL;
- }
-
- for (i = 0 ; i < reta_size ; i++) {
- /* each reta_conf is for 64 entries.
- * to support 128 we use 2 conf of 64
- */
- conf_idx = i / RTE_RETA_GROUP_SIZE;
- idx = i % RTE_RETA_GROUP_SIZE;
- if (TEST_BIT(reta_conf[conf_idx].mask, idx)) {
- entry_value =
- ENA_IO_RXQ_IDX(reta_conf[conf_idx].reta[idx]);
-
- rc = ena_com_indirect_table_fill_entry(ena_dev,
- i,
- entry_value);
- if (unlikely(rc && rc != ENA_COM_UNSUPPORTED)) {
- PMD_DRV_LOG(ERR,
- "Cannot fill indirect table\n");
- return rc;
- }
- }
- }
-
- rte_spinlock_lock(&adapter->admin_lock);
- rc = ena_com_indirect_table_set(ena_dev);
- rte_spinlock_unlock(&adapter->admin_lock);
- if (unlikely(rc && rc != ENA_COM_UNSUPPORTED)) {
- PMD_DRV_LOG(ERR, "Cannot flush the indirect table\n");
- return rc;
- }
-
- PMD_DRV_LOG(DEBUG, "RSS configured %d entries for port %d\n",
- reta_size, dev->data->port_id);
-
- return 0;
-}
-
-/* Query redirection table. */
-static int ena_rss_reta_query(struct rte_eth_dev *dev,
- struct rte_eth_rss_reta_entry64 *reta_conf,
- uint16_t reta_size)
-{
- struct ena_adapter *adapter = dev->data->dev_private;
- struct ena_com_dev *ena_dev = &adapter->ena_dev;
- int rc;
- int i;
- u32 indirect_table[ENA_RX_RSS_TABLE_SIZE] = {0};
- int reta_conf_idx;
- int reta_idx;
-
- if (reta_size == 0 || reta_conf == NULL ||
- (reta_size > RTE_RETA_GROUP_SIZE && ((reta_conf + 1) == NULL)))
- return -EINVAL;
-
- rte_spinlock_lock(&adapter->admin_lock);
- rc = ena_com_indirect_table_get(ena_dev, indirect_table);
- rte_spinlock_unlock(&adapter->admin_lock);
- if (unlikely(rc && rc != ENA_COM_UNSUPPORTED)) {
- PMD_DRV_LOG(ERR, "Cannot get indirection table\n");
- return -ENOTSUP;
- }
-
- for (i = 0 ; i < reta_size ; i++) {
- reta_conf_idx = i / RTE_RETA_GROUP_SIZE;
- reta_idx = i % RTE_RETA_GROUP_SIZE;
- if (TEST_BIT(reta_conf[reta_conf_idx].mask, reta_idx))
- reta_conf[reta_conf_idx].reta[reta_idx] =
- ENA_IO_RXQ_IDX_REV(indirect_table[i]);
- }
-
- return 0;
-}
-
-static int ena_rss_init_default(struct ena_adapter *adapter)
-{
- struct ena_com_dev *ena_dev = &adapter->ena_dev;
- uint16_t nb_rx_queues = adapter->edev_data->nb_rx_queues;
- int rc, i;
- u32 val;
-
- rc = ena_com_rss_init(ena_dev, ENA_RX_RSS_TABLE_LOG_SIZE);
- if (unlikely(rc)) {
- PMD_DRV_LOG(ERR, "Cannot init indirection table\n");
- goto err_rss_init;
- }
-
- for (i = 0; i < ENA_RX_RSS_TABLE_SIZE; i++) {
- val = i % nb_rx_queues;
- rc = ena_com_indirect_table_fill_entry(ena_dev, i,
- ENA_IO_RXQ_IDX(val));
- if (unlikely(rc && (rc != ENA_COM_UNSUPPORTED))) {
- PMD_DRV_LOG(ERR, "Cannot fill indirection table\n");
- goto err_fill_indir;
- }
- }
-
- rc = ena_com_fill_hash_function(ena_dev, ENA_ADMIN_CRC32, NULL,
- ENA_HASH_KEY_SIZE, 0xFFFFFFFF);
- if (unlikely(rc && (rc != ENA_COM_UNSUPPORTED))) {
- PMD_DRV_LOG(INFO, "Cannot fill hash function\n");
- goto err_fill_indir;
- }
-
- rc = ena_com_set_default_hash_ctrl(ena_dev);
- if (unlikely(rc && (rc != ENA_COM_UNSUPPORTED))) {
- PMD_DRV_LOG(INFO, "Cannot fill hash control\n");
- goto err_fill_indir;
- }
-
- rc = ena_com_indirect_table_set(ena_dev);
- if (unlikely(rc && (rc != ENA_COM_UNSUPPORTED))) {
- PMD_DRV_LOG(ERR, "Cannot flush indirection table\n");
- goto err_fill_indir;
- }
- PMD_DRV_LOG(DEBUG, "RSS configured for port %d\n",
- adapter->edev_data->port_id);
-
- return 0;
-
-err_fill_indir:
- ena_com_rss_destroy(ena_dev);
-err_rss_init:
-
- return rc;
-}
-
static void ena_rx_queue_release_all(struct rte_eth_dev *dev)
{
struct ena_ring **queues = (struct ena_ring **)dev->data->rx_queues;
"Inconsistent state of Tx queues\n");
}
- rc = ena_queue_start(&queues[i]);
+ rc = ena_queue_start(dev, &queues[i]);
if (rc) {
PMD_INIT_LOG(ERR,
if (rc)
return rc;
+ rc = ena_setup_rx_intr(dev);
+ if (rc)
+ return rc;
+
rc = ena_queue_start_all(dev, ENA_RING_TYPE_RX);
if (rc)
return rc;
if (rc)
goto err_start_tx;
- if (adapter->edev_data->dev_conf.rxmode.mq_mode &
- ETH_MQ_RX_RSS_FLAG && adapter->edev_data->nb_rx_queues > 0) {
- rc = ena_rss_init_default(adapter);
+ if (adapter->edev_data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG) {
+ rc = ena_rss_configure(adapter);
if (rc)
goto err_rss_init;
}
{
struct ena_adapter *adapter = dev->data->dev_private;
struct ena_com_dev *ena_dev = &adapter->ena_dev;
+ struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+ struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
int rc;
/* Cannot free memory in secondary process */
PMD_DRV_LOG(ERR, "Device reset failed, rc: %d\n", rc);
}
+ rte_intr_disable(intr_handle);
+
+ rte_intr_efd_disable(intr_handle);
+ if (intr_handle->intr_vec != NULL) {
+ rte_free(intr_handle->intr_vec);
+ intr_handle->intr_vec = NULL;
+ }
+
+ rte_intr_enable(intr_handle);
+
++adapter->dev_stats.dev_stop;
adapter->state = ENA_ADAPTER_STATE_STOPPED;
dev->data->dev_started = 0;
return 0;
}
-static int ena_create_io_queue(struct ena_ring *ring)
+static int ena_create_io_queue(struct rte_eth_dev *dev, struct ena_ring *ring)
{
- struct ena_adapter *adapter;
- struct ena_com_dev *ena_dev;
+ struct ena_adapter *adapter = ring->adapter;
+ struct ena_com_dev *ena_dev = &adapter->ena_dev;
+ struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+ struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
struct ena_com_create_io_ctx ctx =
/* policy set to _HOST just to satisfy icc compiler */
{ ENA_ADMIN_PLACEMENT_POLICY_HOST,
unsigned int i;
int rc;
- adapter = ring->adapter;
- ena_dev = &adapter->ena_dev;
-
+ ctx.msix_vector = -1;
if (ring->type == ENA_RING_TYPE_TX) {
ena_qid = ENA_IO_TXQ_IDX(ring->id);
ctx.direction = ENA_COM_IO_QUEUE_DIRECTION_TX;
} else {
ena_qid = ENA_IO_RXQ_IDX(ring->id);
ctx.direction = ENA_COM_IO_QUEUE_DIRECTION_RX;
+ if (rte_intr_dp_is_en(intr_handle))
+ ctx.msix_vector = intr_handle->intr_vec[ring->id];
for (i = 0; i < ring->ring_size; i++)
ring->empty_rx_reqs[i] = i;
}
ctx.queue_size = ring->ring_size;
ctx.qid = ena_qid;
- ctx.msix_vector = -1; /* interrupts not used */
ctx.numa_node = ring->numa_socket_id;
rc = ena_com_create_io_queue(ena_dev, &ctx);
if (ring->type == ENA_RING_TYPE_TX)
ena_com_update_numa_node(ring->ena_com_io_cq, ctx.numa_node);
+ /* Start with Rx interrupts being masked. */
+ if (ring->type == ENA_RING_TYPE_RX && rte_intr_dp_is_en(intr_handle))
+ ena_rx_queue_intr_disable(dev, ring->id);
+
return 0;
}
ena_queue_stop(&queues[i]);
}
-static int ena_queue_start(struct ena_ring *ring)
+static int ena_queue_start(struct rte_eth_dev *dev, struct ena_ring *ring)
{
int rc, bufs_num;
ena_assert_msg(ring->configured == 1,
"Trying to start unconfigured queue\n");
- rc = ena_create_io_queue(ring);
+ rc = ena_create_io_queue(dev, ring);
if (rc) {
PMD_INIT_LOG(ERR, "Failed to create IO queue\n");
return rc;
uint16_t queue_idx,
uint16_t nb_desc,
unsigned int socket_id,
- __rte_unused const struct rte_eth_rxconf *rx_conf,
+ const struct rte_eth_rxconf *rx_conf,
struct rte_mempool *mp)
{
struct ena_adapter *adapter = dev->data->dev_private;
for (i = 0; i < nb_desc; i++)
rxq->empty_rx_reqs[i] = i;
+ rxq->offloads = rx_conf->offloads | dev->data->dev_conf.rxmode.offloads;
+
/* Store pointer to this queue in upper layer */
rxq->configured = 1;
dev->data->rx_queues[queue_idx] = rxq;
adapter->offloads.rx_csum_supported =
(get_feat_ctx.offload.rx_supported &
ENA_ADMIN_FEATURE_OFFLOAD_DESC_RX_L4_IPV4_CSUM_MASK) != 0;
+ adapter->offloads.rss_hash_supported =
+ (get_feat_ctx.offload.rx_supported &
+ ENA_ADMIN_FEATURE_OFFLOAD_DESC_RX_HASH_MASK) != 0;
/* Copy MAC address and point DPDK to it */
eth_dev->data->mac_addrs = (struct rte_ether_addr *)adapter->mac_addr;
get_feat_ctx.dev_attr.mac_addr,
(struct rte_ether_addr *)adapter->mac_addr);
+ rc = ena_com_rss_init(ena_dev, ENA_RX_RSS_TABLE_LOG_SIZE);
+ if (unlikely(rc != 0)) {
+ PMD_DRV_LOG(ERR, "Failed to initialize RSS in ENA device\n");
+ goto err_delete_debug_area;
+ }
+
adapter->drv_stats = rte_zmalloc("adapter stats",
sizeof(*adapter->drv_stats),
RTE_CACHE_LINE_SIZE);
PMD_DRV_LOG(ERR,
"Failed to allocate memory for adapter statistics\n");
rc = -ENOMEM;
- goto err_delete_debug_area;
+ goto err_rss_destroy;
}
rte_spinlock_init(&adapter->admin_lock);
return 0;
+err_rss_destroy:
+ ena_com_rss_destroy(ena_dev);
err_delete_debug_area:
ena_com_delete_debug_area(ena_dev);
if (adapter->state != ENA_ADAPTER_STATE_CLOSED)
ena_close(eth_dev);
+ ena_com_rss_destroy(ena_dev);
+
ena_com_delete_debug_area(ena_dev);
ena_com_delete_host_info(ena_dev);
if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
+ dev->data->dev_conf.txmode.offloads |= DEV_TX_OFFLOAD_MULTI_SEGS;
adapter->tx_selected_offloads = dev->data->dev_conf.txmode.offloads;
adapter->rx_selected_offloads = dev->data->dev_conf.rxmode.offloads;
DEV_RX_OFFLOAD_TCP_CKSUM;
rx_feat |= DEV_RX_OFFLOAD_JUMBO_FRAME;
+ tx_feat |= DEV_TX_OFFLOAD_MULTI_SEGS;
/* Inform framework about available features */
dev_info->rx_offload_capa = rx_feat;
- dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_RSS_HASH;
+ if (adapter->offloads.rss_hash_supported)
+ dev_info->rx_offload_capa |= DEV_RX_OFFLOAD_RSS_HASH;
dev_info->rx_queue_offload_capa = rx_feat;
dev_info->tx_offload_capa = tx_feat;
dev_info->tx_queue_offload_capa = tx_feat;
- dev_info->flow_type_rss_offloads = ETH_RSS_IP | ETH_RSS_TCP |
- ETH_RSS_UDP;
+ dev_info->flow_type_rss_offloads = ENA_ALL_RSS_HF;
+ dev_info->hash_key_size = ENA_HASH_KEY_SIZE;
dev_info->min_rx_bufsize = ENA_MIN_FRAME_LEN;
dev_info->max_rx_pktlen = adapter->max_mtu;
uint16_t completed;
struct ena_com_rx_ctx ena_rx_ctx;
int i, rc = 0;
+ bool fill_hash;
#ifdef RTE_ETHDEV_DEBUG_RX
/* Check adapter state */
}
#endif
+ fill_hash = rx_ring->offloads & DEV_RX_OFFLOAD_RSS_HASH;
+
descs_in_use = rx_ring->ring_size -
ena_com_free_q_entries(rx_ring->ena_com_io_sq) - 1;
nb_pkts = RTE_MIN(descs_in_use, nb_pkts);
}
/* fill mbuf attributes if any */
- ena_rx_mbuf_prepare(mbuf, &ena_rx_ctx);
+ ena_rx_mbuf_prepare(mbuf, &ena_rx_ctx, fill_hash);
if (unlikely(mbuf->ol_flags &
(PKT_RX_IP_CKSUM_BAD | PKT_RX_L4_CKSUM_BAD))) {
return rc;
}
+static int ena_setup_rx_intr(struct rte_eth_dev *dev)
+{
+ struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+ struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+ int rc;
+ uint16_t vectors_nb, i;
+ bool rx_intr_requested = dev->data->dev_conf.intr_conf.rxq;
+
+ if (!rx_intr_requested)
+ return 0;
+
+ if (!rte_intr_cap_multiple(intr_handle)) {
+ PMD_DRV_LOG(ERR,
+ "Rx interrupt requested, but it isn't supported by the PCI driver\n");
+ return -ENOTSUP;
+ }
+
+ /* Disable interrupt mapping before the configuration starts. */
+ rte_intr_disable(intr_handle);
+
+ /* Verify if there are enough vectors available. */
+ vectors_nb = dev->data->nb_rx_queues;
+ if (vectors_nb > RTE_MAX_RXTX_INTR_VEC_ID) {
+ PMD_DRV_LOG(ERR,
+ "Too many Rx interrupts requested, maximum number: %d\n",
+ RTE_MAX_RXTX_INTR_VEC_ID);
+ rc = -ENOTSUP;
+ goto enable_intr;
+ }
+
+ intr_handle->intr_vec = rte_zmalloc("intr_vec",
+ dev->data->nb_rx_queues * sizeof(*intr_handle->intr_vec), 0);
+ if (intr_handle->intr_vec == NULL) {
+ PMD_DRV_LOG(ERR,
+ "Failed to allocate interrupt vector for %d queues\n",
+ dev->data->nb_rx_queues);
+ rc = -ENOMEM;
+ goto enable_intr;
+ }
+
+ rc = rte_intr_efd_enable(intr_handle, vectors_nb);
+ if (rc != 0)
+ goto free_intr_vec;
+
+ if (!rte_intr_allow_others(intr_handle)) {
+ PMD_DRV_LOG(ERR,
+ "Not enough interrupts available to use both ENA Admin and Rx interrupts\n");
+ goto disable_intr_efd;
+ }
+
+ for (i = 0; i < vectors_nb; ++i)
+ intr_handle->intr_vec[i] = RTE_INTR_VEC_RXTX_OFFSET + i;
+
+ rte_intr_enable(intr_handle);
+ return 0;
+
+disable_intr_efd:
+ rte_intr_efd_disable(intr_handle);
+free_intr_vec:
+ rte_free(intr_handle->intr_vec);
+ intr_handle->intr_vec = NULL;
+enable_intr:
+ rte_intr_enable(intr_handle);
+ return rc;
+}
+
+static void ena_rx_queue_intr_set(struct rte_eth_dev *dev,
+ uint16_t queue_id,
+ bool unmask)
+{
+ struct ena_adapter *adapter = dev->data->dev_private;
+ struct ena_ring *rxq = &adapter->rx_ring[queue_id];
+ struct ena_eth_io_intr_reg intr_reg;
+
+ ena_com_update_intr_reg(&intr_reg, 0, 0, unmask);
+ ena_com_unmask_intr(rxq->ena_com_io_cq, &intr_reg);
+}
+
+static int ena_rx_queue_intr_enable(struct rte_eth_dev *dev,
+ uint16_t queue_id)
+{
+ ena_rx_queue_intr_set(dev, queue_id, true);
+
+ return 0;
+}
+
+static int ena_rx_queue_intr_disable(struct rte_eth_dev *dev,
+ uint16_t queue_id)
+{
+ ena_rx_queue_intr_set(dev, queue_id, false);
+
+ return 0;
+}
+
/*********************************************************************
* PMD configuration
*********************************************************************/