X-Git-Url: http://git.droids-corp.org/?a=blobdiff_plain;f=drivers%2Fnet%2Fdpaa%2Fdpaa_rxtx.c;h=5dba1db8b367d06fe0e4c9ed2a0f89452e6dcd39;hb=b6197a7beb511f273aeca829c59355d993353261;hp=148f2652e4efa92e8361cb918cc3d4455823605b;hpb=5e0789e93507205a86f28a92efed76e454b49be6;p=dpdk.git diff --git a/drivers/net/dpaa/dpaa_rxtx.c b/drivers/net/dpaa/dpaa_rxtx.c index 148f2652e4..5dba1db8b3 100644 --- a/drivers/net/dpaa/dpaa_rxtx.c +++ b/drivers/net/dpaa/dpaa_rxtx.c @@ -1,7 +1,7 @@ /* SPDX-License-Identifier: BSD-3-Clause * * Copyright 2016 Freescale Semiconductor, Inc. All rights reserved. - * Copyright 2017 NXP + * Copyright 2017,2019 NXP * */ @@ -26,23 +26,25 @@ #include #include #include -#include +#include #include #include #include #include #include #include +#include #include "dpaa_ethdev.h" #include "dpaa_rxtx.h" #include #include +#include #include #include #include -#include +#include #include #define DPAA_MBUF_TO_CONTIG_FD(_mbuf, _fd, _bpid) \ @@ -57,7 +59,7 @@ } while (0) #if (defined RTE_LIBRTE_DPAA_DEBUG_DRIVER) -void dpaa_display_frame(const struct qm_fd *fd) +static void dpaa_display_frame(const struct qm_fd *fd) { int ii; char *ptr; @@ -88,21 +90,14 @@ static inline void dpaa_slow_parsing(struct rte_mbuf *m __rte_unused, /*TBD:XXX: to be implemented*/ } -static inline void dpaa_eth_packet_info(struct rte_mbuf *m, - uint64_t fd_virt_addr) +static inline void dpaa_eth_packet_info(struct rte_mbuf *m, void *fd_virt_addr) { struct annotations_t *annot = GET_ANNOTATIONS(fd_virt_addr); - uint64_t prs = *((uint64_t *)(&annot->parse)) & DPAA_PARSE_MASK; + uint64_t prs = *((uintptr_t *)(&annot->parse)) & DPAA_PARSE_MASK; DPAA_DP_LOG(DEBUG, " Parsing mbuf: %p with annotations: %p", m, annot); switch (prs) { - case DPAA_PKT_TYPE_NONE: - m->packet_type = 0; - break; - case DPAA_PKT_TYPE_ETHER: - m->packet_type = RTE_PTYPE_L2_ETHER; - break; case DPAA_PKT_TYPE_IPV4: m->packet_type = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV4; @@ -111,6 +106,9 @@ static inline void dpaa_eth_packet_info(struct rte_mbuf *m, m->packet_type = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6; break; + case DPAA_PKT_TYPE_ETHER: + m->packet_type = RTE_PTYPE_L2_ETHER; + break; case DPAA_PKT_TYPE_IPV4_FRAG: case DPAA_PKT_TYPE_IPV4_FRAG_UDP: case DPAA_PKT_TYPE_IPV4_FRAG_TCP: @@ -173,6 +171,9 @@ static inline void dpaa_eth_packet_info(struct rte_mbuf *m, m->packet_type = RTE_PTYPE_L2_ETHER | RTE_PTYPE_L3_IPV6 | RTE_PTYPE_L4_SCTP; break; + case DPAA_PKT_TYPE_NONE: + m->packet_type = 0; + break; /* More switch cases can be added */ default: dpaa_slow_parsing(m, prs); @@ -183,12 +184,11 @@ static inline void dpaa_eth_packet_info(struct rte_mbuf *m, << DPAA_PKT_L3_LEN_SHIFT; /* Set the hash values */ - m->hash.rss = (uint32_t)(rte_be_to_cpu_64(annot->hash)); - m->ol_flags = PKT_RX_RSS_HASH; + m->hash.rss = (uint32_t)(annot->hash); /* All packets with Bad checksum are dropped by interface (and * corresponding notification issued to RX error queues). */ - m->ol_flags |= PKT_RX_IP_CKSUM_GOOD; + m->ol_flags = PKT_RX_RSS_HASH | PKT_RX_IP_CKSUM_GOOD; /* Check if Vlan is present */ if (prs & DPAA_PARSE_VLAN_MASK) @@ -198,44 +198,45 @@ static inline void dpaa_eth_packet_info(struct rte_mbuf *m, static inline void dpaa_checksum(struct rte_mbuf *mbuf) { - struct ether_hdr *eth_hdr = rte_pktmbuf_mtod(mbuf, struct ether_hdr *); + struct rte_ether_hdr *eth_hdr = + rte_pktmbuf_mtod(mbuf, struct rte_ether_hdr *); char *l3_hdr = (char *)eth_hdr + mbuf->l2_len; - struct ipv4_hdr *ipv4_hdr = (struct ipv4_hdr *)l3_hdr; - struct ipv6_hdr *ipv6_hdr = (struct ipv6_hdr *)l3_hdr; + struct rte_ipv4_hdr *ipv4_hdr = (struct rte_ipv4_hdr *)l3_hdr; + struct rte_ipv6_hdr *ipv6_hdr = (struct rte_ipv6_hdr *)l3_hdr; DPAA_DP_LOG(DEBUG, "Calculating checksum for mbuf: %p", mbuf); if (((mbuf->packet_type & RTE_PTYPE_L3_MASK) == RTE_PTYPE_L3_IPV4) || ((mbuf->packet_type & RTE_PTYPE_L3_MASK) == RTE_PTYPE_L3_IPV4_EXT)) { - ipv4_hdr = (struct ipv4_hdr *)l3_hdr; + ipv4_hdr = (struct rte_ipv4_hdr *)l3_hdr; ipv4_hdr->hdr_checksum = 0; ipv4_hdr->hdr_checksum = rte_ipv4_cksum(ipv4_hdr); } else if (((mbuf->packet_type & RTE_PTYPE_L3_MASK) == RTE_PTYPE_L3_IPV6) || ((mbuf->packet_type & RTE_PTYPE_L3_MASK) == RTE_PTYPE_L3_IPV6_EXT)) - ipv6_hdr = (struct ipv6_hdr *)l3_hdr; + ipv6_hdr = (struct rte_ipv6_hdr *)l3_hdr; if ((mbuf->packet_type & RTE_PTYPE_L4_MASK) == RTE_PTYPE_L4_TCP) { - struct tcp_hdr *tcp_hdr = (struct tcp_hdr *)(l3_hdr + + struct rte_tcp_hdr *tcp_hdr = (struct rte_tcp_hdr *)(l3_hdr + mbuf->l3_len); tcp_hdr->cksum = 0; - if (eth_hdr->ether_type == htons(ETHER_TYPE_IPv4)) + if (eth_hdr->ether_type == htons(RTE_ETHER_TYPE_IPV4)) tcp_hdr->cksum = rte_ipv4_udptcp_cksum(ipv4_hdr, tcp_hdr); - else /* assume ethertype == ETHER_TYPE_IPv6 */ + else /* assume ethertype == RTE_ETHER_TYPE_IPV6 */ tcp_hdr->cksum = rte_ipv6_udptcp_cksum(ipv6_hdr, tcp_hdr); } else if ((mbuf->packet_type & RTE_PTYPE_L4_MASK) == RTE_PTYPE_L4_UDP) { - struct udp_hdr *udp_hdr = (struct udp_hdr *)(l3_hdr + + struct rte_udp_hdr *udp_hdr = (struct rte_udp_hdr *)(l3_hdr + mbuf->l3_len); udp_hdr->dgram_cksum = 0; - if (eth_hdr->ether_type == htons(ETHER_TYPE_IPv4)) + if (eth_hdr->ether_type == htons(RTE_ETHER_TYPE_IPV4)) udp_hdr->dgram_cksum = rte_ipv4_udptcp_cksum(ipv4_hdr, udp_hdr); - else /* assume ethertype == ETHER_TYPE_IPv6 */ + else /* assume ethertype == RTE_ETHER_TYPE_IPV6 */ udp_hdr->dgram_cksum = rte_ipv6_udptcp_cksum(ipv6_hdr, udp_hdr); } @@ -297,7 +298,7 @@ dpaa_unsegmented_checksum(struct rte_mbuf *mbuf, struct qm_fd *fd_arr) } struct rte_mbuf * -dpaa_eth_sg_to_mbuf(struct qm_fd *fd, uint32_t ifid) +dpaa_eth_sg_to_mbuf(const struct qm_fd *fd, uint32_t ifid) { struct dpaa_bp_info *bp_info = DPAA_BPID_TO_POOL_INFO(fd->bpid); struct rte_mbuf *first_seg, *prev_seg, *cur_seg, *temp; @@ -306,9 +307,7 @@ dpaa_eth_sg_to_mbuf(struct qm_fd *fd, uint32_t ifid) int i = 0; uint8_t fd_offset = fd->offset; - DPAA_DP_LOG(DEBUG, "Received an SG frame"); - - vaddr = rte_dpaa_mem_ptov(qm_fd_addr(fd)); + vaddr = DPAA_MEMPOOL_PTOV(bp_info, qm_fd_addr(fd)); if (!vaddr) { DPAA_PMD_ERR("unable to convert physical address"); return NULL; @@ -317,7 +316,7 @@ dpaa_eth_sg_to_mbuf(struct qm_fd *fd, uint32_t ifid) sg_temp = &sgt[i++]; hw_sg_to_cpu(sg_temp); temp = (struct rte_mbuf *)((char *)vaddr - bp_info->meta_data_size); - sg_vaddr = rte_dpaa_mem_ptov(qm_sg_entry_get64(sg_temp)); + sg_vaddr = DPAA_MEMPOOL_PTOV(bp_info, qm_sg_entry_get64(sg_temp)); first_seg = (struct rte_mbuf *)((char *)sg_vaddr - bp_info->meta_data_size); @@ -333,7 +332,8 @@ dpaa_eth_sg_to_mbuf(struct qm_fd *fd, uint32_t ifid) while (i < DPAA_SGT_MAX_ENTRIES) { sg_temp = &sgt[i++]; hw_sg_to_cpu(sg_temp); - sg_vaddr = rte_dpaa_mem_ptov(qm_sg_entry_get64(sg_temp)); + sg_vaddr = DPAA_MEMPOOL_PTOV(bp_info, + qm_sg_entry_get64(sg_temp)); cur_seg = (struct rte_mbuf *)((char *)sg_vaddr - bp_info->meta_data_size); cur_seg->data_off = sg_temp->offset; @@ -348,41 +348,41 @@ dpaa_eth_sg_to_mbuf(struct qm_fd *fd, uint32_t ifid) } prev_seg = cur_seg; } + DPAA_DP_LOG(DEBUG, "Received an SG frame len =%d, num_sg =%d", + first_seg->pkt_len, first_seg->nb_segs); - dpaa_eth_packet_info(first_seg, (uint64_t)vaddr); + dpaa_eth_packet_info(first_seg, vaddr); rte_pktmbuf_free_seg(temp); return first_seg; } -static inline struct rte_mbuf *dpaa_eth_fd_to_mbuf(struct qm_fd *fd, - uint32_t ifid) +static inline struct rte_mbuf * +dpaa_eth_fd_to_mbuf(const struct qm_fd *fd, uint32_t ifid) { - struct dpaa_bp_info *bp_info = DPAA_BPID_TO_POOL_INFO(fd->bpid); struct rte_mbuf *mbuf; + struct dpaa_bp_info *bp_info = DPAA_BPID_TO_POOL_INFO(fd->bpid); void *ptr; uint8_t format = (fd->opaque & DPAA_FD_FORMAT_MASK) >> DPAA_FD_FORMAT_SHIFT; - uint16_t offset = - (fd->opaque & DPAA_FD_OFFSET_MASK) >> DPAA_FD_OFFSET_SHIFT; - uint32_t length = fd->opaque & DPAA_FD_LENGTH_MASK; - - DPAA_DP_LOG(DEBUG, " FD--->MBUF"); + uint16_t offset; + uint32_t length; if (unlikely(format == qm_fd_sg)) return dpaa_eth_sg_to_mbuf(fd, ifid); + offset = (fd->opaque & DPAA_FD_OFFSET_MASK) >> DPAA_FD_OFFSET_SHIFT; + length = fd->opaque & DPAA_FD_LENGTH_MASK; + + DPAA_DP_LOG(DEBUG, " FD--->MBUF off %d len = %d", offset, length); + /* Ignoring case when format != qm_fd_contig */ dpaa_display_frame(fd); - ptr = rte_dpaa_mem_ptov(fd->addr); - /* Ignoring case when ptr would be NULL. That is only possible incase - * of a corrupted packet - */ + ptr = DPAA_MEMPOOL_PTOV(bp_info, qm_fd_addr(fd)); mbuf = (struct rte_mbuf *)((char *)ptr - bp_info->meta_data_size); /* Prefetch the Parse results and packet data to L1 */ rte_prefetch0((void *)((uint8_t *)ptr + DEFAULT_RX_ICEOF)); - rte_prefetch0((void *)((uint8_t *)ptr + offset)); mbuf->data_off = offset; mbuf->data_len = length; @@ -393,11 +393,203 @@ static inline struct rte_mbuf *dpaa_eth_fd_to_mbuf(struct qm_fd *fd, mbuf->ol_flags = 0; mbuf->next = NULL; rte_mbuf_refcnt_set(mbuf, 1); - dpaa_eth_packet_info(mbuf, (uint64_t)mbuf->buf_addr); + dpaa_eth_packet_info(mbuf, mbuf->buf_addr); return mbuf; } +/* Specific for LS1043 */ +void +dpaa_rx_cb_no_prefetch(struct qman_fq **fq, struct qm_dqrr_entry **dqrr, + void **bufs, int num_bufs) +{ + struct rte_mbuf *mbuf; + struct dpaa_bp_info *bp_info; + const struct qm_fd *fd; + void *ptr; + struct dpaa_if *dpaa_intf; + uint16_t offset, i; + uint32_t length; + uint8_t format; + + bp_info = DPAA_BPID_TO_POOL_INFO(dqrr[0]->fd.bpid); + ptr = rte_dpaa_mem_ptov(qm_fd_addr(&dqrr[0]->fd)); + rte_prefetch0((void *)((uint8_t *)ptr + DEFAULT_RX_ICEOF)); + bufs[0] = (struct rte_mbuf *)((char *)ptr - bp_info->meta_data_size); + + for (i = 0; i < num_bufs; i++) { + if (i < num_bufs - 1) { + bp_info = DPAA_BPID_TO_POOL_INFO(dqrr[i + 1]->fd.bpid); + ptr = rte_dpaa_mem_ptov(qm_fd_addr(&dqrr[i + 1]->fd)); + rte_prefetch0((void *)((uint8_t *)ptr + + DEFAULT_RX_ICEOF)); + bufs[i + 1] = (struct rte_mbuf *)((char *)ptr - + bp_info->meta_data_size); + } + + fd = &dqrr[i]->fd; + dpaa_intf = fq[0]->dpaa_intf; + + format = (fd->opaque & DPAA_FD_FORMAT_MASK) >> + DPAA_FD_FORMAT_SHIFT; + if (unlikely(format == qm_fd_sg)) { + bufs[i] = dpaa_eth_sg_to_mbuf(fd, dpaa_intf->ifid); + continue; + } + + offset = (fd->opaque & DPAA_FD_OFFSET_MASK) >> + DPAA_FD_OFFSET_SHIFT; + length = fd->opaque & DPAA_FD_LENGTH_MASK; + + mbuf = bufs[i]; + mbuf->data_off = offset; + mbuf->data_len = length; + mbuf->pkt_len = length; + mbuf->port = dpaa_intf->ifid; + + mbuf->nb_segs = 1; + mbuf->ol_flags = 0; + mbuf->next = NULL; + rte_mbuf_refcnt_set(mbuf, 1); + dpaa_eth_packet_info(mbuf, mbuf->buf_addr); + } +} + +void +dpaa_rx_cb(struct qman_fq **fq, struct qm_dqrr_entry **dqrr, + void **bufs, int num_bufs) +{ + struct rte_mbuf *mbuf; + const struct qm_fd *fd; + struct dpaa_if *dpaa_intf; + uint16_t offset, i; + uint32_t length; + uint8_t format; + + for (i = 0; i < num_bufs; i++) { + fd = &dqrr[i]->fd; + dpaa_intf = fq[0]->dpaa_intf; + + format = (fd->opaque & DPAA_FD_FORMAT_MASK) >> + DPAA_FD_FORMAT_SHIFT; + if (unlikely(format == qm_fd_sg)) { + bufs[i] = dpaa_eth_sg_to_mbuf(fd, dpaa_intf->ifid); + continue; + } + + offset = (fd->opaque & DPAA_FD_OFFSET_MASK) >> + DPAA_FD_OFFSET_SHIFT; + length = fd->opaque & DPAA_FD_LENGTH_MASK; + + mbuf = bufs[i]; + mbuf->data_off = offset; + mbuf->data_len = length; + mbuf->pkt_len = length; + mbuf->port = dpaa_intf->ifid; + + mbuf->nb_segs = 1; + mbuf->ol_flags = 0; + mbuf->next = NULL; + rte_mbuf_refcnt_set(mbuf, 1); + dpaa_eth_packet_info(mbuf, mbuf->buf_addr); + } +} + +void dpaa_rx_cb_prepare(struct qm_dqrr_entry *dq, void **bufs) +{ + struct dpaa_bp_info *bp_info = DPAA_BPID_TO_POOL_INFO(dq->fd.bpid); + void *ptr = rte_dpaa_mem_ptov(qm_fd_addr(&dq->fd)); + + /* In case of LS1046, annotation stashing is disabled due to L2 cache + * being bottleneck in case of multicore scanario for this platform. + * So we prefetch the annoation beforehand, so that it is available + * in cache when accessed. + */ + rte_prefetch0((void *)((uint8_t *)ptr + DEFAULT_RX_ICEOF)); + + *bufs = (struct rte_mbuf *)((char *)ptr - bp_info->meta_data_size); +} + +static uint16_t +dpaa_eth_queue_portal_rx(struct qman_fq *fq, + struct rte_mbuf **bufs, + uint16_t nb_bufs) +{ + int ret; + + if (unlikely(!fq->qp_initialized)) { + ret = rte_dpaa_portal_fq_init((void *)0, fq); + if (ret) { + DPAA_PMD_ERR("Failure in affining portal %d", ret); + return 0; + } + fq->qp_initialized = 1; + } + + return qman_portal_poll_rx(nb_bufs, (void **)bufs, fq->qp); +} + +enum qman_cb_dqrr_result +dpaa_rx_cb_parallel(void *event, + struct qman_portal *qm __always_unused, + struct qman_fq *fq, + const struct qm_dqrr_entry *dqrr, + void **bufs) +{ + u32 ifid = ((struct dpaa_if *)fq->dpaa_intf)->ifid; + struct rte_mbuf *mbuf; + struct rte_event *ev = (struct rte_event *)event; + + mbuf = dpaa_eth_fd_to_mbuf(&dqrr->fd, ifid); + ev->event_ptr = (void *)mbuf; + ev->flow_id = fq->ev.flow_id; + ev->sub_event_type = fq->ev.sub_event_type; + ev->event_type = RTE_EVENT_TYPE_ETHDEV; + ev->op = RTE_EVENT_OP_NEW; + ev->sched_type = fq->ev.sched_type; + ev->queue_id = fq->ev.queue_id; + ev->priority = fq->ev.priority; + ev->impl_opaque = (uint8_t)DPAA_INVALID_MBUF_SEQN; + mbuf->seqn = DPAA_INVALID_MBUF_SEQN; + *bufs = mbuf; + + return qman_cb_dqrr_consume; +} + +enum qman_cb_dqrr_result +dpaa_rx_cb_atomic(void *event, + struct qman_portal *qm __always_unused, + struct qman_fq *fq, + const struct qm_dqrr_entry *dqrr, + void **bufs) +{ + u8 index; + u32 ifid = ((struct dpaa_if *)fq->dpaa_intf)->ifid; + struct rte_mbuf *mbuf; + struct rte_event *ev = (struct rte_event *)event; + + mbuf = dpaa_eth_fd_to_mbuf(&dqrr->fd, ifid); + ev->event_ptr = (void *)mbuf; + ev->flow_id = fq->ev.flow_id; + ev->sub_event_type = fq->ev.sub_event_type; + ev->event_type = RTE_EVENT_TYPE_ETHDEV; + ev->op = RTE_EVENT_OP_NEW; + ev->sched_type = fq->ev.sched_type; + ev->queue_id = fq->ev.queue_id; + ev->priority = fq->ev.priority; + + /* Save active dqrr entries */ + index = DQRR_PTR2IDX(dqrr); + DPAA_PER_LCORE_DQRR_SIZE++; + DPAA_PER_LCORE_DQRR_HELD |= 1 << index; + DPAA_PER_LCORE_DQRR_MBUF(index) = mbuf; + ev->impl_opaque = index + 1; + mbuf->seqn = (uint32_t)index + 1; + *bufs = mbuf; + + return qman_cb_dqrr_defer; +} + uint16_t dpaa_eth_queue_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) @@ -405,16 +597,37 @@ uint16_t dpaa_eth_queue_rx(void *q, struct qman_fq *fq = q; struct qm_dqrr_entry *dq; uint32_t num_rx = 0, ifid = ((struct dpaa_if *)fq->dpaa_intf)->ifid; - int ret; + int num_rx_bufs, ret; + uint32_t vdqcr_flags = 0; - ret = rte_dpaa_portal_init((void *)0); - if (ret) { - DPAA_PMD_ERR("Failure in affining portal"); - return 0; + if (unlikely(rte_dpaa_bpid_info == NULL && + rte_eal_process_type() == RTE_PROC_SECONDARY)) + rte_dpaa_bpid_info = fq->bp_array; + + if (likely(fq->is_static)) + return dpaa_eth_queue_portal_rx(fq, bufs, nb_bufs); + + if (unlikely(!RTE_PER_LCORE(dpaa_io))) { + ret = rte_dpaa_portal_init((void *)0); + if (ret) { + DPAA_PMD_ERR("Failure in affining portal"); + return 0; + } } - ret = qman_set_vdq(fq, (nb_bufs > DPAA_MAX_DEQUEUE_NUM_FRAMES) ? - DPAA_MAX_DEQUEUE_NUM_FRAMES : nb_bufs); + /* Until request for four buffers, we provide exact number of buffers. + * Otherwise we do not set the QM_VDQCR_EXACT flag. + * Not setting QM_VDQCR_EXACT flag can provide two more buffers than + * requested, so we request two less in this case. + */ + if (nb_bufs < 4) { + vdqcr_flags = QM_VDQCR_EXACT; + num_rx_bufs = nb_bufs; + } else { + num_rx_bufs = nb_bufs > DPAA_MAX_DEQUEUE_NUM_FRAMES ? + (DPAA_MAX_DEQUEUE_NUM_FRAMES - 2) : (nb_bufs - 2); + } + ret = qman_set_vdq(fq, num_rx_bufs, vdqcr_flags); if (ret) return 0; @@ -429,52 +642,6 @@ uint16_t dpaa_eth_queue_rx(void *q, return num_rx; } -static void *dpaa_get_pktbuf(struct dpaa_bp_info *bp_info) -{ - int ret; - uint64_t buf = 0; - struct bm_buffer bufs; - - ret = bman_acquire(bp_info->bp, &bufs, 1, 0); - if (ret <= 0) { - DPAA_PMD_WARN("Failed to allocate buffers %d", ret); - return (void *)buf; - } - - DPAA_DP_LOG(DEBUG, "got buffer 0x%lx from pool %d", - (uint64_t)bufs.addr, bufs.bpid); - - buf = (uint64_t)rte_dpaa_mem_ptov(bufs.addr) - bp_info->meta_data_size; - if (!buf) - goto out; - -out: - return (void *)buf; -} - -static struct rte_mbuf *dpaa_get_dmable_mbuf(struct rte_mbuf *mbuf, - struct dpaa_if *dpaa_intf) -{ - struct rte_mbuf *dpaa_mbuf; - - /* allocate pktbuffer on bpid for dpaa port */ - dpaa_mbuf = dpaa_get_pktbuf(dpaa_intf->bp_info); - if (!dpaa_mbuf) - return NULL; - - memcpy((uint8_t *)(dpaa_mbuf->buf_addr) + mbuf->data_off, (void *) - ((uint8_t *)(mbuf->buf_addr) + mbuf->data_off), mbuf->pkt_len); - - /* Copy only the required fields */ - dpaa_mbuf->data_off = mbuf->data_off; - dpaa_mbuf->pkt_len = mbuf->pkt_len; - dpaa_mbuf->ol_flags = mbuf->ol_flags; - dpaa_mbuf->packet_type = mbuf->packet_type; - dpaa_mbuf->tx_offload = mbuf->tx_offload; - rte_pktmbuf_free(mbuf); - return dpaa_mbuf; -} - int dpaa_eth_mbuf_to_sg_fd(struct rte_mbuf *mbuf, struct qm_fd *fd, @@ -648,24 +815,84 @@ tx_on_dpaa_pool(struct rte_mbuf *mbuf, } /* Handle all mbufs on an external pool (non-dpaa) */ -static inline uint16_t -tx_on_external_pool(struct qman_fq *txq, struct rte_mbuf *mbuf, - struct qm_fd *fd_arr) +static inline struct rte_mbuf * +reallocate_mbuf(struct qman_fq *txq, struct rte_mbuf *mbuf) { struct dpaa_if *dpaa_intf = txq->dpaa_intf; - struct rte_mbuf *dmable_mbuf; + struct dpaa_bp_info *bp_info = dpaa_intf->bp_info; + struct rte_mbuf *new_mbufs[DPAA_SGT_MAX_ENTRIES + 1] = {0}; + struct rte_mbuf *temp_mbuf; + int num_new_segs, mbuf_greater, ret, extra_seg = 0, i = 0; + uint64_t mbufs_size, bytes_to_copy, offset1 = 0, offset2 = 0; + char *data; + + DPAA_DP_LOG(DEBUG, "Reallocating transmit buffer"); + + mbufs_size = bp_info->size - + bp_info->meta_data_size - RTE_PKTMBUF_HEADROOM; + extra_seg = !!(mbuf->pkt_len % mbufs_size); + num_new_segs = (mbuf->pkt_len / mbufs_size) + extra_seg; + + ret = rte_pktmbuf_alloc_bulk(bp_info->mp, new_mbufs, num_new_segs); + if (ret != 0) { + DPAA_DP_LOG(DEBUG, "Allocation for new buffers failed"); + return NULL; + } - DPAA_DP_LOG(DEBUG, "Non-BMAN offloaded buffer." - "Allocating an offloaded buffer"); - dmable_mbuf = dpaa_get_dmable_mbuf(mbuf, dpaa_intf); - if (!dmable_mbuf) { - DPAA_DP_LOG(DEBUG, "no dpaa buffers."); - return 1; + temp_mbuf = mbuf; + + while (temp_mbuf) { + /* If mbuf data is less than new mbuf remaining memory */ + if ((temp_mbuf->data_len - offset1) < (mbufs_size - offset2)) { + bytes_to_copy = temp_mbuf->data_len - offset1; + mbuf_greater = -1; + /* If mbuf data is greater than new mbuf remaining memory */ + } else if ((temp_mbuf->data_len - offset1) > + (mbufs_size - offset2)) { + bytes_to_copy = mbufs_size - offset2; + mbuf_greater = 1; + /* if mbuf data is equal to new mbuf remaining memory */ + } else { + bytes_to_copy = temp_mbuf->data_len - offset1; + mbuf_greater = 0; + } + + /* Copy the data */ + data = rte_pktmbuf_append(new_mbufs[0], bytes_to_copy); + + rte_memcpy((uint8_t *)data, rte_pktmbuf_mtod_offset(mbuf, + void *, offset1), bytes_to_copy); + + /* Set new offsets and the temp buffers */ + if (mbuf_greater == -1) { + offset1 = 0; + offset2 += bytes_to_copy; + temp_mbuf = temp_mbuf->next; + } else if (mbuf_greater == 1) { + offset2 = 0; + offset1 += bytes_to_copy; + new_mbufs[i]->next = new_mbufs[i + 1]; + new_mbufs[0]->nb_segs++; + i++; + } else { + offset1 = 0; + offset2 = 0; + temp_mbuf = temp_mbuf->next; + new_mbufs[i]->next = new_mbufs[i + 1]; + if (new_mbufs[i + 1]) + new_mbufs[0]->nb_segs++; + i++; + } } - DPAA_MBUF_TO_CONTIG_FD(dmable_mbuf, fd_arr, dpaa_intf->bp_info->bpid); + /* Copy other required fields */ + new_mbufs[0]->ol_flags = mbuf->ol_flags; + new_mbufs[0]->packet_type = mbuf->packet_type; + new_mbufs[0]->tx_offload = mbuf->tx_offload; - return 0; + rte_pktmbuf_free(mbuf); + + return new_mbufs[0]; } uint16_t @@ -677,12 +904,15 @@ dpaa_eth_queue_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) struct qm_fd fd_arr[DPAA_TX_BURST_SIZE]; uint32_t frames_to_send, loop, sent = 0; uint16_t state; - int ret; - - ret = rte_dpaa_portal_init((void *)0); - if (ret) { - DPAA_PMD_ERR("Failure in affining portal"); - return 0; + int ret, realloc_mbuf = 0; + uint32_t seqn, index, flags[DPAA_TX_BURST_SIZE] = {0}; + + if (unlikely(!RTE_PER_LCORE(dpaa_io))) { + ret = rte_dpaa_portal_init((void *)0); + if (ret) { + DPAA_PMD_ERR("Failure in affining portal"); + return 0; + } } DPAA_DP_LOG(DEBUG, "Transmitting %d buffers on queue: %p", nb_bufs, q); @@ -692,12 +922,33 @@ dpaa_eth_queue_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) DPAA_TX_BURST_SIZE : nb_bufs; for (loop = 0; loop < frames_to_send; loop++) { mbuf = *(bufs++); + /* In case the data offset is not multiple of 16, + * FMAN can stall because of an errata. So reallocate + * the buffer in such case. + */ + if (dpaa_svr_family == SVR_LS1043A_FAMILY && + (mbuf->data_off & 0x7F) != 0x0) + realloc_mbuf = 1; + seqn = mbuf->seqn; + if (seqn != DPAA_INVALID_MBUF_SEQN) { + index = seqn - 1; + if (DPAA_PER_LCORE_DQRR_HELD & (1 << index)) { + flags[loop] = + ((index & QM_EQCR_DCA_IDXMASK) << 8); + flags[loop] |= QMAN_ENQUEUE_FLAG_DCA; + DPAA_PER_LCORE_DQRR_SIZE--; + DPAA_PER_LCORE_DQRR_HELD &= + ~(1 << index); + } + } + if (likely(RTE_MBUF_DIRECT(mbuf))) { mp = mbuf->pool; bp_info = DPAA_MEMPOOL_TO_POOL_INFO(mp); if (likely(mp->ops_index == bp_info->dpaa_ops_index && mbuf->nb_segs == 1 && + realloc_mbuf == 0 && rte_mbuf_refcnt_read(mbuf) == 1)) { DPAA_MBUF_TO_CONTIG_FD(mbuf, &fd_arr[loop], bp_info->bpid); @@ -713,22 +964,12 @@ dpaa_eth_queue_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) } bp_info = DPAA_MEMPOOL_TO_POOL_INFO(mp); - if (likely(mp->ops_index == bp_info->dpaa_ops_index)) { - state = tx_on_dpaa_pool(mbuf, bp_info, - &fd_arr[loop]); - if (unlikely(state)) { - /* Set frames_to_send & nb_bufs so - * that packets are transmitted till - * previous frame. - */ - frames_to_send = loop; - nb_bufs = loop; - goto send_pkts; - } - } else { - state = tx_on_external_pool(q, mbuf, - &fd_arr[loop]); - if (unlikely(state)) { + if (unlikely(mp->ops_index != bp_info->dpaa_ops_index || + realloc_mbuf == 1)) { + struct rte_mbuf *temp_mbuf; + + temp_mbuf = reallocate_mbuf(q, mbuf); + if (!temp_mbuf) { /* Set frames_to_send & nb_bufs so * that packets are transmitted till * previous frame. @@ -737,6 +978,20 @@ dpaa_eth_queue_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs) nb_bufs = loop; goto send_pkts; } + mbuf = temp_mbuf; + realloc_mbuf = 0; + } + + state = tx_on_dpaa_pool(mbuf, bp_info, + &fd_arr[loop]); + if (unlikely(state)) { + /* Set frames_to_send & nb_bufs so + * that packets are transmitted till + * previous frame. + */ + frames_to_send = loop; + nb_bufs = loop; + goto send_pkts; } } @@ -744,7 +999,8 @@ send_pkts: loop = 0; while (loop < frames_to_send) { loop += qman_enqueue_multi(q, &fd_arr[loop], - frames_to_send - loop); + &flags[loop], + frames_to_send - loop); } nb_bufs -= frames_to_send; sent += frames_to_send;