numam-dpdk/drivers/event/octeontx2/otx2_evdev_adptr.c
Pavan Nikhilesh cb4261e0bf event/octeontx2: improve datapath memory locality
When event device is transmitting packet on OCTEONTX2 it needs to access
the destined ethernet device TXq data.
Currently, we get the TXq data through rte_eth_devices global array.
Instead save the TXq address inside event port memory.

Cc: stable@dpdk.org

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
2020-06-30 07:41:26 +02:00

556 lines
15 KiB
C

/* SPDX-License-Identifier: BSD-3-Clause
* Copyright(C) 2019 Marvell International Ltd.
*/
#include "otx2_evdev.h"
int
otx2_sso_rx_adapter_caps_get(const struct rte_eventdev *event_dev,
const struct rte_eth_dev *eth_dev, uint32_t *caps)
{
int rc;
RTE_SET_USED(event_dev);
rc = strncmp(eth_dev->device->driver->name, "net_octeontx2", 13);
if (rc)
*caps = RTE_EVENT_ETH_RX_ADAPTER_SW_CAP;
else
*caps = RTE_EVENT_ETH_RX_ADAPTER_CAP_INTERNAL_PORT |
RTE_EVENT_ETH_RX_ADAPTER_CAP_MULTI_EVENTQ;
return 0;
}
static inline int
sso_rxq_enable(struct otx2_eth_dev *dev, uint16_t qid, uint8_t tt, uint8_t ggrp,
uint16_t eth_port_id)
{
struct otx2_mbox *mbox = dev->mbox;
struct nix_aq_enq_req *aq;
int rc;
aq = otx2_mbox_alloc_msg_nix_aq_enq(mbox);
aq->qidx = qid;
aq->ctype = NIX_AQ_CTYPE_CQ;
aq->op = NIX_AQ_INSTOP_WRITE;
aq->cq.ena = 0;
aq->cq.caching = 0;
otx2_mbox_memset(&aq->cq_mask, 0, sizeof(struct nix_cq_ctx_s));
aq->cq_mask.ena = ~(aq->cq_mask.ena);
aq->cq_mask.caching = ~(aq->cq_mask.caching);
rc = otx2_mbox_process(mbox);
if (rc < 0) {
otx2_err("Failed to disable cq context");
goto fail;
}
aq = otx2_mbox_alloc_msg_nix_aq_enq(mbox);
aq->qidx = qid;
aq->ctype = NIX_AQ_CTYPE_RQ;
aq->op = NIX_AQ_INSTOP_WRITE;
aq->rq.sso_ena = 1;
aq->rq.sso_tt = tt;
aq->rq.sso_grp = ggrp;
aq->rq.ena_wqwd = 1;
/* Mbuf Header generation :
* > FIRST_SKIP is a super set of WQE_SKIP, dont modify first skip as
* it already has data related to mbuf size, headroom, private area.
* > Using WQE_SKIP we can directly assign
* mbuf = wqe - sizeof(struct mbuf);
* so that mbuf header will not have unpredicted values while headroom
* and private data starts at the beginning of wqe_data.
*/
aq->rq.wqe_skip = 1;
aq->rq.wqe_caching = 1;
aq->rq.spb_ena = 0;
aq->rq.flow_tagw = 20; /* 20-bits */
/* Flow Tag calculation :
*
* rq_tag <31:24> = good/bad_tag<8:0>;
* rq_tag <23:0> = [ltag]
*
* flow_tag_mask<31:0> = (1 << flow_tagw) - 1; <31:20>
* tag<31:0> = (~flow_tag_mask & rq_tag) | (flow_tag_mask & flow_tag);
*
* Setup :
* ltag<23:0> = (eth_port_id & 0xF) << 20;
* good/bad_tag<8:0> =
* ((eth_port_id >> 4) & 0xF) | (RTE_EVENT_TYPE_ETHDEV << 4);
*
* TAG<31:0> on getwork = <31:28>(RTE_EVENT_TYPE_ETHDEV) |
* <27:20> (eth_port_id) | <20:0> [TAG]
*/
aq->rq.ltag = (eth_port_id & 0xF) << 20;
aq->rq.good_utag = ((eth_port_id >> 4) & 0xF) |
(RTE_EVENT_TYPE_ETHDEV << 4);
aq->rq.bad_utag = aq->rq.good_utag;
aq->rq.ena = 0; /* Don't enable RQ yet */
aq->rq.pb_caching = 0x2; /* First cache aligned block to LLC */
aq->rq.xqe_imm_size = 0; /* No pkt data copy to CQE */
otx2_mbox_memset(&aq->rq_mask, 0, sizeof(struct nix_rq_ctx_s));
/* mask the bits to write. */
aq->rq_mask.sso_ena = ~(aq->rq_mask.sso_ena);
aq->rq_mask.sso_tt = ~(aq->rq_mask.sso_tt);
aq->rq_mask.sso_grp = ~(aq->rq_mask.sso_grp);
aq->rq_mask.ena_wqwd = ~(aq->rq_mask.ena_wqwd);
aq->rq_mask.wqe_skip = ~(aq->rq_mask.wqe_skip);
aq->rq_mask.wqe_caching = ~(aq->rq_mask.wqe_caching);
aq->rq_mask.spb_ena = ~(aq->rq_mask.spb_ena);
aq->rq_mask.flow_tagw = ~(aq->rq_mask.flow_tagw);
aq->rq_mask.ltag = ~(aq->rq_mask.ltag);
aq->rq_mask.good_utag = ~(aq->rq_mask.good_utag);
aq->rq_mask.bad_utag = ~(aq->rq_mask.bad_utag);
aq->rq_mask.ena = ~(aq->rq_mask.ena);
aq->rq_mask.pb_caching = ~(aq->rq_mask.pb_caching);
aq->rq_mask.xqe_imm_size = ~(aq->rq_mask.xqe_imm_size);
rc = otx2_mbox_process(mbox);
if (rc < 0) {
otx2_err("Failed to init rx adapter context");
goto fail;
}
return 0;
fail:
return rc;
}
static inline int
sso_rxq_disable(struct otx2_eth_dev *dev, uint16_t qid)
{
struct otx2_mbox *mbox = dev->mbox;
struct nix_aq_enq_req *aq;
int rc;
aq = otx2_mbox_alloc_msg_nix_aq_enq(mbox);
aq->qidx = qid;
aq->ctype = NIX_AQ_CTYPE_CQ;
aq->op = NIX_AQ_INSTOP_WRITE;
aq->cq.ena = 1;
aq->cq.caching = 1;
otx2_mbox_memset(&aq->cq_mask, 0, sizeof(struct nix_cq_ctx_s));
aq->cq_mask.ena = ~(aq->cq_mask.ena);
aq->cq_mask.caching = ~(aq->cq_mask.caching);
rc = otx2_mbox_process(mbox);
if (rc < 0) {
otx2_err("Failed to enable cq context");
goto fail;
}
aq = otx2_mbox_alloc_msg_nix_aq_enq(mbox);
aq->qidx = qid;
aq->ctype = NIX_AQ_CTYPE_RQ;
aq->op = NIX_AQ_INSTOP_WRITE;
aq->rq.sso_ena = 0;
aq->rq.sso_tt = SSO_TT_UNTAGGED;
aq->rq.sso_grp = 0;
aq->rq.ena_wqwd = 0;
aq->rq.wqe_caching = 0;
aq->rq.wqe_skip = 0;
aq->rq.spb_ena = 0;
aq->rq.flow_tagw = 0x20;
aq->rq.ltag = 0;
aq->rq.good_utag = 0;
aq->rq.bad_utag = 0;
aq->rq.ena = 1;
aq->rq.pb_caching = 0x2; /* First cache aligned block to LLC */
aq->rq.xqe_imm_size = 0; /* No pkt data copy to CQE */
otx2_mbox_memset(&aq->rq_mask, 0, sizeof(struct nix_rq_ctx_s));
/* mask the bits to write. */
aq->rq_mask.sso_ena = ~(aq->rq_mask.sso_ena);
aq->rq_mask.sso_tt = ~(aq->rq_mask.sso_tt);
aq->rq_mask.sso_grp = ~(aq->rq_mask.sso_grp);
aq->rq_mask.ena_wqwd = ~(aq->rq_mask.ena_wqwd);
aq->rq_mask.wqe_caching = ~(aq->rq_mask.wqe_caching);
aq->rq_mask.wqe_skip = ~(aq->rq_mask.wqe_skip);
aq->rq_mask.spb_ena = ~(aq->rq_mask.spb_ena);
aq->rq_mask.flow_tagw = ~(aq->rq_mask.flow_tagw);
aq->rq_mask.ltag = ~(aq->rq_mask.ltag);
aq->rq_mask.good_utag = ~(aq->rq_mask.good_utag);
aq->rq_mask.bad_utag = ~(aq->rq_mask.bad_utag);
aq->rq_mask.ena = ~(aq->rq_mask.ena);
aq->rq_mask.pb_caching = ~(aq->rq_mask.pb_caching);
aq->rq_mask.xqe_imm_size = ~(aq->rq_mask.xqe_imm_size);
rc = otx2_mbox_process(mbox);
if (rc < 0) {
otx2_err("Failed to clear rx adapter context");
goto fail;
}
return 0;
fail:
return rc;
}
void
sso_updt_xae_cnt(struct otx2_sso_evdev *dev, void *data, uint32_t event_type)
{
int i;
switch (event_type) {
case RTE_EVENT_TYPE_ETHDEV:
{
struct otx2_eth_rxq *rxq = data;
uint64_t *old_ptr;
for (i = 0; i < dev->rx_adptr_pool_cnt; i++) {
if ((uint64_t)rxq->pool == dev->rx_adptr_pools[i])
return;
}
dev->rx_adptr_pool_cnt++;
old_ptr = dev->rx_adptr_pools;
dev->rx_adptr_pools = rte_realloc(dev->rx_adptr_pools,
sizeof(uint64_t) *
dev->rx_adptr_pool_cnt, 0);
if (dev->rx_adptr_pools == NULL) {
dev->adptr_xae_cnt += rxq->pool->size;
dev->rx_adptr_pools = old_ptr;
dev->rx_adptr_pool_cnt--;
return;
}
dev->rx_adptr_pools[dev->rx_adptr_pool_cnt - 1] =
(uint64_t)rxq->pool;
dev->adptr_xae_cnt += rxq->pool->size;
break;
}
case RTE_EVENT_TYPE_TIMER:
{
struct otx2_tim_ring *timr = data;
uint16_t *old_ring_ptr;
uint64_t *old_sz_ptr;
for (i = 0; i < dev->tim_adptr_ring_cnt; i++) {
if (timr->ring_id != dev->timer_adptr_rings[i])
continue;
if (timr->nb_timers == dev->timer_adptr_sz[i])
return;
dev->adptr_xae_cnt -= dev->timer_adptr_sz[i];
dev->adptr_xae_cnt += timr->nb_timers;
dev->timer_adptr_sz[i] = timr->nb_timers;
return;
}
dev->tim_adptr_ring_cnt++;
old_ring_ptr = dev->timer_adptr_rings;
old_sz_ptr = dev->timer_adptr_sz;
dev->timer_adptr_rings = rte_realloc(dev->timer_adptr_rings,
sizeof(uint16_t) *
dev->tim_adptr_ring_cnt,
0);
if (dev->timer_adptr_rings == NULL) {
dev->adptr_xae_cnt += timr->nb_timers;
dev->timer_adptr_rings = old_ring_ptr;
dev->tim_adptr_ring_cnt--;
return;
}
dev->timer_adptr_sz = rte_realloc(dev->timer_adptr_sz,
sizeof(uint64_t) *
dev->tim_adptr_ring_cnt,
0);
if (dev->timer_adptr_sz == NULL) {
dev->adptr_xae_cnt += timr->nb_timers;
dev->timer_adptr_sz = old_sz_ptr;
dev->tim_adptr_ring_cnt--;
return;
}
dev->timer_adptr_rings[dev->tim_adptr_ring_cnt - 1] =
timr->ring_id;
dev->timer_adptr_sz[dev->tim_adptr_ring_cnt - 1] =
timr->nb_timers;
dev->adptr_xae_cnt += timr->nb_timers;
break;
}
default:
break;
}
}
static inline void
sso_updt_lookup_mem(const struct rte_eventdev *event_dev, void *lookup_mem)
{
struct otx2_sso_evdev *dev = sso_pmd_priv(event_dev);
int i;
for (i = 0; i < dev->nb_event_ports; i++) {
if (dev->dual_ws) {
struct otx2_ssogws_dual *ws = event_dev->data->ports[i];
ws->lookup_mem = lookup_mem;
} else {
struct otx2_ssogws *ws = event_dev->data->ports[i];
ws->lookup_mem = lookup_mem;
}
}
}
int
otx2_sso_rx_adapter_queue_add(const struct rte_eventdev *event_dev,
const struct rte_eth_dev *eth_dev,
int32_t rx_queue_id,
const struct rte_event_eth_rx_adapter_queue_conf *queue_conf)
{
struct otx2_eth_dev *otx2_eth_dev = eth_dev->data->dev_private;
struct otx2_sso_evdev *dev = sso_pmd_priv(event_dev);
uint16_t port = eth_dev->data->port_id;
struct otx2_eth_rxq *rxq;
int i, rc;
rc = strncmp(eth_dev->device->driver->name, "net_octeontx2", 13);
if (rc)
return -EINVAL;
if (rx_queue_id < 0) {
for (i = 0 ; i < eth_dev->data->nb_rx_queues; i++) {
rxq = eth_dev->data->rx_queues[i];
sso_updt_xae_cnt(dev, rxq, RTE_EVENT_TYPE_ETHDEV);
rc = sso_xae_reconfigure((struct rte_eventdev *)
(uintptr_t)event_dev);
rc |= sso_rxq_enable(otx2_eth_dev, i,
queue_conf->ev.sched_type,
queue_conf->ev.queue_id, port);
}
rxq = eth_dev->data->rx_queues[0];
sso_updt_lookup_mem(event_dev, rxq->lookup_mem);
} else {
rxq = eth_dev->data->rx_queues[rx_queue_id];
sso_updt_xae_cnt(dev, rxq, RTE_EVENT_TYPE_ETHDEV);
rc = sso_xae_reconfigure((struct rte_eventdev *)
(uintptr_t)event_dev);
rc |= sso_rxq_enable(otx2_eth_dev, (uint16_t)rx_queue_id,
queue_conf->ev.sched_type,
queue_conf->ev.queue_id, port);
sso_updt_lookup_mem(event_dev, rxq->lookup_mem);
}
if (rc < 0) {
otx2_err("Failed to configure Rx adapter port=%d, q=%d", port,
queue_conf->ev.queue_id);
return rc;
}
dev->rx_offloads |= otx2_eth_dev->rx_offload_flags;
dev->tstamp = &otx2_eth_dev->tstamp;
sso_fastpath_fns_set((struct rte_eventdev *)(uintptr_t)event_dev);
return 0;
}
int
otx2_sso_rx_adapter_queue_del(const struct rte_eventdev *event_dev,
const struct rte_eth_dev *eth_dev,
int32_t rx_queue_id)
{
struct otx2_eth_dev *dev = eth_dev->data->dev_private;
int i, rc;
RTE_SET_USED(event_dev);
rc = strncmp(eth_dev->device->driver->name, "net_octeontx2", 13);
if (rc)
return -EINVAL;
if (rx_queue_id < 0) {
for (i = 0 ; i < eth_dev->data->nb_rx_queues; i++)
rc = sso_rxq_disable(dev, i);
} else {
rc = sso_rxq_disable(dev, (uint16_t)rx_queue_id);
}
if (rc < 0)
otx2_err("Failed to clear Rx adapter config port=%d, q=%d",
eth_dev->data->port_id, rx_queue_id);
return rc;
}
int
otx2_sso_rx_adapter_start(const struct rte_eventdev *event_dev,
const struct rte_eth_dev *eth_dev)
{
RTE_SET_USED(event_dev);
RTE_SET_USED(eth_dev);
return 0;
}
int
otx2_sso_rx_adapter_stop(const struct rte_eventdev *event_dev,
const struct rte_eth_dev *eth_dev)
{
RTE_SET_USED(event_dev);
RTE_SET_USED(eth_dev);
return 0;
}
int
otx2_sso_tx_adapter_caps_get(const struct rte_eventdev *dev,
const struct rte_eth_dev *eth_dev, uint32_t *caps)
{
int ret;
RTE_SET_USED(dev);
ret = strncmp(eth_dev->device->driver->name, "net_octeontx2,", 13);
if (ret)
*caps = 0;
else
*caps = RTE_EVENT_ETH_TX_ADAPTER_CAP_INTERNAL_PORT;
return 0;
}
static int
sso_sqb_aura_limit_edit(struct rte_mempool *mp, uint16_t nb_sqb_bufs)
{
struct otx2_npa_lf *npa_lf = otx2_intra_dev_get_cfg()->npa_lf;
struct npa_aq_enq_req *aura_req;
aura_req = otx2_mbox_alloc_msg_npa_aq_enq(npa_lf->mbox);
aura_req->aura_id = npa_lf_aura_handle_to_aura(mp->pool_id);
aura_req->ctype = NPA_AQ_CTYPE_AURA;
aura_req->op = NPA_AQ_INSTOP_WRITE;
aura_req->aura.limit = nb_sqb_bufs;
aura_req->aura_mask.limit = ~(aura_req->aura_mask.limit);
return otx2_mbox_process(npa_lf->mbox);
}
static int
sso_add_tx_queue_data(const struct rte_eventdev *event_dev,
uint16_t eth_port_id, uint16_t tx_queue_id,
struct otx2_eth_txq *txq)
{
struct otx2_sso_evdev *dev = sso_pmd_priv(event_dev);
int i;
for (i = 0; i < event_dev->data->nb_ports; i++) {
dev->max_port_id = RTE_MAX(dev->max_port_id, eth_port_id);
if (dev->dual_ws) {
struct otx2_ssogws_dual *old_dws;
struct otx2_ssogws_dual *dws;
old_dws = event_dev->data->ports[i];
dws = rte_realloc_socket(old_dws,
sizeof(struct otx2_ssogws_dual)
+ (sizeof(uint64_t) *
(dev->max_port_id + 1) *
RTE_MAX_QUEUES_PER_PORT),
RTE_CACHE_LINE_SIZE,
event_dev->data->socket_id);
if (dws == NULL)
return -ENOMEM;
((uint64_t (*)[RTE_MAX_QUEUES_PER_PORT]
)&dws->tx_adptr_data)[eth_port_id][tx_queue_id] =
(uint64_t)txq;
event_dev->data->ports[i] = dws;
} else {
struct otx2_ssogws *old_ws;
struct otx2_ssogws *ws;
old_ws = event_dev->data->ports[i];
ws = rte_realloc_socket(old_ws,
sizeof(struct otx2_ssogws_dual)
+ (sizeof(uint64_t) *
(dev->max_port_id + 1) *
RTE_MAX_QUEUES_PER_PORT),
RTE_CACHE_LINE_SIZE,
event_dev->data->socket_id);
if (ws == NULL)
return -ENOMEM;
((uint64_t (*)[RTE_MAX_QUEUES_PER_PORT]
)&ws->tx_adptr_data)[eth_port_id][tx_queue_id] =
(uint64_t)txq;
event_dev->data->ports[i] = ws;
}
}
return 0;
}
int
otx2_sso_tx_adapter_queue_add(uint8_t id, const struct rte_eventdev *event_dev,
const struct rte_eth_dev *eth_dev,
int32_t tx_queue_id)
{
struct otx2_eth_dev *otx2_eth_dev = eth_dev->data->dev_private;
struct otx2_sso_evdev *dev = sso_pmd_priv(event_dev);
struct otx2_eth_txq *txq;
int i, ret;
RTE_SET_USED(id);
if (tx_queue_id < 0) {
for (i = 0 ; i < eth_dev->data->nb_tx_queues; i++) {
txq = eth_dev->data->tx_queues[i];
sso_sqb_aura_limit_edit(txq->sqb_pool,
OTX2_SSO_SQB_LIMIT);
ret = sso_add_tx_queue_data(event_dev,
eth_dev->data->port_id, i,
txq);
if (ret < 0)
return ret;
}
} else {
txq = eth_dev->data->tx_queues[tx_queue_id];
sso_sqb_aura_limit_edit(txq->sqb_pool, OTX2_SSO_SQB_LIMIT);
ret = sso_add_tx_queue_data(event_dev, eth_dev->data->port_id,
tx_queue_id, txq);
if (ret < 0)
return ret;
}
dev->tx_offloads |= otx2_eth_dev->tx_offload_flags;
sso_fastpath_fns_set((struct rte_eventdev *)(uintptr_t)event_dev);
return 0;
}
int
otx2_sso_tx_adapter_queue_del(uint8_t id, const struct rte_eventdev *event_dev,
const struct rte_eth_dev *eth_dev,
int32_t tx_queue_id)
{
struct otx2_eth_txq *txq;
int i;
RTE_SET_USED(id);
RTE_SET_USED(eth_dev);
RTE_SET_USED(event_dev);
if (tx_queue_id < 0) {
for (i = 0 ; i < eth_dev->data->nb_tx_queues; i++) {
txq = eth_dev->data->tx_queues[i];
sso_sqb_aura_limit_edit(txq->sqb_pool,
txq->nb_sqb_bufs);
}
} else {
txq = eth_dev->data->tx_queues[tx_queue_id];
sso_sqb_aura_limit_edit(txq->sqb_pool, txq->nb_sqb_bufs);
}
return 0;
}