#include <string.h>
#include <errno.h>
-#include <tmmintrin.h>
-
#include <rte_cycles.h>
#include <rte_memory.h>
-#include <rte_memzone.h>
#include <rte_branch_prediction.h>
#include <rte_mempool.h>
#include <rte_malloc.h>
#include <rte_errno.h>
#include <rte_byteorder.h>
-#include "virtio_logs.h"
-#include "virtio_ethdev.h"
-#include "virtqueue.h"
-#include "virtio_rxtx.h"
-
-#define RTE_VIRTIO_VPMD_RX_BURST 32
-#define RTE_VIRTIO_DESC_PER_LOOP 8
-#define RTE_VIRTIO_VPMD_RX_REARM_THRESH RTE_VIRTIO_VPMD_RX_BURST
+#include "virtio_rxtx_simple.h"
#ifndef __INTEL_COMPILER
#pragma GCC diagnostic ignored "-Wcast-qual"
struct vring_desc *start_dp;
uint16_t desc_idx;
+ cookie->port = vq->rxq.port_id;
+
desc_idx = vq->vq_avail_idx & (vq->vq_nentries - 1);
dxp = &vq->vq_descx[desc_idx];
dxp->cookie = (void *)cookie;
vq->sw_ring[desc_idx] = cookie;
start_dp = vq->vq_ring.desc;
- start_dp[desc_idx].addr = (uint64_t)((uintptr_t)cookie->buf_physaddr +
- RTE_PKTMBUF_HEADROOM - sizeof(struct virtio_net_hdr));
+ start_dp[desc_idx].addr =
+ VIRTIO_MBUF_ADDR(cookie, vq) +
+ RTE_PKTMBUF_HEADROOM - vq->hw->vtnet_hdr_size;
start_dp[desc_idx].len = cookie->buf_len -
- RTE_PKTMBUF_HEADROOM + sizeof(struct virtio_net_hdr);
+ RTE_PKTMBUF_HEADROOM + vq->hw->vtnet_hdr_size;
vq->vq_free_cnt--;
vq->vq_avail_idx++;
return 0;
}
-static inline void
-virtio_rxq_rearm_vec(struct virtqueue *rxvq)
-{
- int i;
- uint16_t desc_idx;
- struct rte_mbuf **sw_ring;
- struct vring_desc *start_dp;
- int ret;
-
- desc_idx = rxvq->vq_avail_idx & (rxvq->vq_nentries - 1);
- sw_ring = &rxvq->sw_ring[desc_idx];
- start_dp = &rxvq->vq_ring.desc[desc_idx];
-
- ret = rte_mempool_get_bulk(rxvq->mpool, (void **)sw_ring,
- RTE_VIRTIO_VPMD_RX_REARM_THRESH);
- if (unlikely(ret)) {
- rte_eth_devices[rxvq->port_id].data->rx_mbuf_alloc_failed +=
- RTE_VIRTIO_VPMD_RX_REARM_THRESH;
- return;
- }
-
- for (i = 0; i < RTE_VIRTIO_VPMD_RX_REARM_THRESH; i++) {
- uintptr_t p;
-
- p = (uintptr_t)&sw_ring[i]->rearm_data;
- *(uint64_t *)p = rxvq->mbuf_initializer;
-
- start_dp[i].addr =
- (uint64_t)((uintptr_t)sw_ring[i]->buf_physaddr +
- RTE_PKTMBUF_HEADROOM - sizeof(struct virtio_net_hdr));
- start_dp[i].len = sw_ring[i]->buf_len -
- RTE_PKTMBUF_HEADROOM + sizeof(struct virtio_net_hdr);
- }
-
- rxvq->vq_avail_idx += RTE_VIRTIO_VPMD_RX_REARM_THRESH;
- rxvq->vq_free_cnt -= RTE_VIRTIO_VPMD_RX_REARM_THRESH;
- vq_update_avail_idx(rxvq);
-}
-
-/* virtio vPMD receive routine, only accept(nb_pkts >= RTE_VIRTIO_DESC_PER_LOOP)
- *
- * This routine is for non-mergeable RX, one desc for each guest buffer.
- * This routine is based on the RX ring layout optimization. Each entry in the
- * avail ring points to the desc with the same index in the desc ring and this
- * will never be changed in the driver.
- *
- * - nb_pkts < RTE_VIRTIO_DESC_PER_LOOP, just return no packet
- */
uint16_t
-virtio_recv_pkts_vec(void *rx_queue, struct rte_mbuf **rx_pkts,
+virtio_xmit_pkts_simple(void *tx_queue, struct rte_mbuf **tx_pkts,
uint16_t nb_pkts)
{
- struct virtqueue *rxvq = rx_queue;
+ struct virtnet_tx *txvq = tx_queue;
+ struct virtqueue *vq = txvq->vq;
+ struct virtio_hw *hw = vq->hw;
uint16_t nb_used;
uint16_t desc_idx;
- struct vring_used_elem *rused;
- struct rte_mbuf **sw_ring;
- struct rte_mbuf **sw_ring_end;
- uint16_t nb_pkts_received;
- __m128i shuf_msk1, shuf_msk2, len_adjust;
-
- shuf_msk1 = _mm_set_epi8(
- 0xFF, 0xFF, 0xFF, 0xFF,
- 0xFF, 0xFF, /* vlan tci */
- 5, 4, /* dat len */
- 0xFF, 0xFF, 5, 4, /* pkt len */
- 0xFF, 0xFF, 0xFF, 0xFF /* packet type */
-
- );
-
- shuf_msk2 = _mm_set_epi8(
- 0xFF, 0xFF, 0xFF, 0xFF,
- 0xFF, 0xFF, /* vlan tci */
- 13, 12, /* dat len */
- 0xFF, 0xFF, 13, 12, /* pkt len */
- 0xFF, 0xFF, 0xFF, 0xFF /* packet type */
- );
-
- /* Subtract the header length.
- * In which case do we need the header length in used->len ?
- */
- len_adjust = _mm_set_epi16(
- 0, 0,
- 0,
- (uint16_t) -sizeof(struct virtio_net_hdr),
- 0, (uint16_t) -sizeof(struct virtio_net_hdr),
- 0, 0);
-
- if (unlikely(nb_pkts < RTE_VIRTIO_DESC_PER_LOOP))
- return 0;
+ struct vring_desc *start_dp;
+ uint16_t nb_tail, nb_commit;
+ int i;
+ uint16_t desc_idx_max = (vq->vq_nentries >> 1) - 1;
+ uint16_t nb_tx = 0;
- nb_used = *(volatile uint16_t *)&rxvq->vq_ring.used->idx -
- rxvq->vq_used_cons_idx;
+ if (unlikely(hw->started == 0))
+ return nb_tx;
+ nb_used = VIRTQUEUE_NUSED(vq);
rte_compiler_barrier();
- if (unlikely(nb_used == 0))
- return 0;
-
- nb_pkts = RTE_ALIGN_FLOOR(nb_pkts, RTE_VIRTIO_DESC_PER_LOOP);
- nb_used = RTE_MIN(nb_used, nb_pkts);
-
- desc_idx = (uint16_t)(rxvq->vq_used_cons_idx & (rxvq->vq_nentries - 1));
- rused = &rxvq->vq_ring.used->ring[desc_idx];
- sw_ring = &rxvq->sw_ring[desc_idx];
- sw_ring_end = &rxvq->sw_ring[rxvq->vq_nentries];
-
- _mm_prefetch((const void *)rused, _MM_HINT_T0);
+ if (nb_used >= VIRTIO_TX_FREE_THRESH)
+ virtio_xmit_cleanup(vq);
- if (rxvq->vq_free_cnt >= RTE_VIRTIO_VPMD_RX_REARM_THRESH) {
- virtio_rxq_rearm_vec(rxvq);
- if (unlikely(virtqueue_kick_prepare(rxvq)))
- virtqueue_notify(rxvq);
+ nb_commit = nb_pkts = RTE_MIN((vq->vq_free_cnt >> 1), nb_pkts);
+ desc_idx = (uint16_t)(vq->vq_avail_idx & desc_idx_max);
+ start_dp = vq->vq_ring.desc;
+ nb_tail = (uint16_t) (desc_idx_max + 1 - desc_idx);
+
+ if (nb_commit >= nb_tail) {
+ for (i = 0; i < nb_tail; i++)
+ vq->vq_descx[desc_idx + i].cookie = tx_pkts[i];
+ for (i = 0; i < nb_tail; i++) {
+ start_dp[desc_idx].addr =
+ VIRTIO_MBUF_DATA_DMA_ADDR(*tx_pkts, vq);
+ start_dp[desc_idx].len = (*tx_pkts)->pkt_len;
+ tx_pkts++;
+ desc_idx++;
+ }
+ nb_commit -= nb_tail;
+ desc_idx = 0;
+ }
+ for (i = 0; i < nb_commit; i++)
+ vq->vq_descx[desc_idx + i].cookie = tx_pkts[i];
+ for (i = 0; i < nb_commit; i++) {
+ start_dp[desc_idx].addr =
+ VIRTIO_MBUF_DATA_DMA_ADDR(*tx_pkts, vq);
+ start_dp[desc_idx].len = (*tx_pkts)->pkt_len;
+ tx_pkts++;
+ desc_idx++;
}
- for (nb_pkts_received = 0;
- nb_pkts_received < nb_used;) {
- __m128i desc[RTE_VIRTIO_DESC_PER_LOOP / 2];
- __m128i mbp[RTE_VIRTIO_DESC_PER_LOOP / 2];
- __m128i pkt_mb[RTE_VIRTIO_DESC_PER_LOOP];
-
- mbp[0] = _mm_loadu_si128((__m128i *)(sw_ring + 0));
- desc[0] = _mm_loadu_si128((__m128i *)(rused + 0));
- _mm_storeu_si128((__m128i *)&rx_pkts[0], mbp[0]);
-
- mbp[1] = _mm_loadu_si128((__m128i *)(sw_ring + 2));
- desc[1] = _mm_loadu_si128((__m128i *)(rused + 2));
- _mm_storeu_si128((__m128i *)&rx_pkts[2], mbp[1]);
-
- mbp[2] = _mm_loadu_si128((__m128i *)(sw_ring + 4));
- desc[2] = _mm_loadu_si128((__m128i *)(rused + 4));
- _mm_storeu_si128((__m128i *)&rx_pkts[4], mbp[2]);
-
- mbp[3] = _mm_loadu_si128((__m128i *)(sw_ring + 6));
- desc[3] = _mm_loadu_si128((__m128i *)(rused + 6));
- _mm_storeu_si128((__m128i *)&rx_pkts[6], mbp[3]);
-
- pkt_mb[1] = _mm_shuffle_epi8(desc[0], shuf_msk2);
- pkt_mb[0] = _mm_shuffle_epi8(desc[0], shuf_msk1);
- pkt_mb[1] = _mm_add_epi16(pkt_mb[1], len_adjust);
- pkt_mb[0] = _mm_add_epi16(pkt_mb[0], len_adjust);
- _mm_storeu_si128((void *)&rx_pkts[1]->rx_descriptor_fields1,
- pkt_mb[1]);
- _mm_storeu_si128((void *)&rx_pkts[0]->rx_descriptor_fields1,
- pkt_mb[0]);
-
- pkt_mb[3] = _mm_shuffle_epi8(desc[1], shuf_msk2);
- pkt_mb[2] = _mm_shuffle_epi8(desc[1], shuf_msk1);
- pkt_mb[3] = _mm_add_epi16(pkt_mb[3], len_adjust);
- pkt_mb[2] = _mm_add_epi16(pkt_mb[2], len_adjust);
- _mm_storeu_si128((void *)&rx_pkts[3]->rx_descriptor_fields1,
- pkt_mb[3]);
- _mm_storeu_si128((void *)&rx_pkts[2]->rx_descriptor_fields1,
- pkt_mb[2]);
-
- pkt_mb[5] = _mm_shuffle_epi8(desc[2], shuf_msk2);
- pkt_mb[4] = _mm_shuffle_epi8(desc[2], shuf_msk1);
- pkt_mb[5] = _mm_add_epi16(pkt_mb[5], len_adjust);
- pkt_mb[4] = _mm_add_epi16(pkt_mb[4], len_adjust);
- _mm_storeu_si128((void *)&rx_pkts[5]->rx_descriptor_fields1,
- pkt_mb[5]);
- _mm_storeu_si128((void *)&rx_pkts[4]->rx_descriptor_fields1,
- pkt_mb[4]);
-
- pkt_mb[7] = _mm_shuffle_epi8(desc[3], shuf_msk2);
- pkt_mb[6] = _mm_shuffle_epi8(desc[3], shuf_msk1);
- pkt_mb[7] = _mm_add_epi16(pkt_mb[7], len_adjust);
- pkt_mb[6] = _mm_add_epi16(pkt_mb[6], len_adjust);
- _mm_storeu_si128((void *)&rx_pkts[7]->rx_descriptor_fields1,
- pkt_mb[7]);
- _mm_storeu_si128((void *)&rx_pkts[6]->rx_descriptor_fields1,
- pkt_mb[6]);
+ rte_compiler_barrier();
- if (unlikely(nb_used <= RTE_VIRTIO_DESC_PER_LOOP)) {
- if (sw_ring + nb_used <= sw_ring_end)
- nb_pkts_received += nb_used;
- else
- nb_pkts_received += sw_ring_end - sw_ring;
- break;
- } else {
- if (unlikely(sw_ring + RTE_VIRTIO_DESC_PER_LOOP >=
- sw_ring_end)) {
- nb_pkts_received += sw_ring_end - sw_ring;
- break;
- } else {
- nb_pkts_received += RTE_VIRTIO_DESC_PER_LOOP;
+ vq->vq_free_cnt -= (uint16_t)(nb_pkts << 1);
+ vq->vq_avail_idx += nb_pkts;
+ vq->vq_ring.avail->idx = vq->vq_avail_idx;
+ txvq->stats.packets += nb_pkts;
- rx_pkts += RTE_VIRTIO_DESC_PER_LOOP;
- sw_ring += RTE_VIRTIO_DESC_PER_LOOP;
- rused += RTE_VIRTIO_DESC_PER_LOOP;
- nb_used -= RTE_VIRTIO_DESC_PER_LOOP;
- }
- }
+ if (likely(nb_pkts)) {
+ if (unlikely(virtqueue_kick_prepare(vq)))
+ virtqueue_notify(vq);
}
- rxvq->vq_used_cons_idx += nb_pkts_received;
- rxvq->vq_free_cnt += nb_pkts_received;
- rxvq->packets += nb_pkts_received;
- return nb_pkts_received;
+ return nb_pkts;
}
int __attribute__((cold))
-virtio_rxq_vec_setup(struct virtqueue *rxq)
+virtio_rxq_vec_setup(struct virtnet_rx *rxq)
{
uintptr_t p;
struct rte_mbuf mb_def = { .buf_addr = 0 }; /* zeroed mbuf */
return 0;
}
+
+/* Stub for linkage when arch specific implementation is not available */
+uint16_t __attribute__((weak))
+virtio_recv_pkts_vec(void *rx_queue __rte_unused,
+ struct rte_mbuf **rx_pkts __rte_unused,
+ uint16_t nb_pkts __rte_unused)
+{
+ rte_panic("Wrong weak function linked by linker\n");
+ return 0;
+}
git.droids-corp.org / dpdk.git / blobdiff