From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga09.intel.com (mga09.intel.com [134.134.136.24]) by dpdk.org (Postfix) with ESMTP id ED3FB8E56 for ; Sun, 6 Sep 2015 10:40:49 +0200 (CEST) Received: from orsmga003.jf.intel.com ([10.7.209.27]) by orsmga102.jf.intel.com with ESMTP; 06 Sep 2015 01:40:50 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.17,478,1437462000"; d="scan'208";a="639532668" Received: from pgsmsx106.gar.corp.intel.com ([10.221.44.98]) by orsmga003.jf.intel.com with ESMTP; 06 Sep 2015 01:40:48 -0700 Received: from shsmsx103.ccr.corp.intel.com (10.239.110.14) by PGSMSX106.gar.corp.intel.com (10.221.44.98) with Microsoft SMTP Server (TLS) id 14.3.224.2; Sun, 6 Sep 2015 16:40:46 +0800 Received: from shsmsx102.ccr.corp.intel.com ([169.254.2.206]) by SHSMSX103.ccr.corp.intel.com ([169.254.4.248]) with mapi id 14.03.0224.002; Sun, 6 Sep 2015 16:40:45 +0800 From: "Ouyang, Changchun" To: Stephen Hemminger , "Xie, Huawei" Thread-Topic: [PATCH 3/4] virtio: use indirect ring elements Thread-Index: AQHQ51TG5SiLKqn2L0KGjDKr+OzoSZ4vMGcw Date: Sun, 6 Sep 2015 08:40:44 +0000 Message-ID: References: <1441400308-5725-1-git-send-email-stephen@networkplumber.org> <1441400308-5725-4-git-send-email-stephen@networkplumber.org> In-Reply-To: <1441400308-5725-4-git-send-email-stephen@networkplumber.org> Accept-Language: zh-CN, en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-originating-ip: [10.239.127.40] Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Cc: "dev@dpdk.org" Subject: Re: [dpdk-dev] [PATCH 3/4] virtio: use indirect ring elements X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sun, 06 Sep 2015 08:40:50 -0000 > -----Original Message----- > From: Stephen Hemminger [mailto:stephen@networkplumber.org] > Sent: Saturday, September 5, 2015 4:58 AM > To: Xie, Huawei; Ouyang, Changchun > Cc: dev@dpdk.org; Stephen Hemminger > Subject: [PATCH 3/4] virtio: use indirect ring elements >=20 > The virtio ring in QEMU/KVM is usually limited to 256 entries and the nor= mal > way that virtio driver was queuing mbufs required nsegs + 1 ring elements= . > By using the indirect ring element feature if available, each packet will= take > only one ring slot even for multi-segment packets. >=20 > Signed-off-by: Stephen Hemminger > --- > drivers/net/virtio/virtio_ethdev.c | 11 +++++--- > drivers/net/virtio/virtio_ethdev.h | 3 ++- > drivers/net/virtio/virtio_rxtx.c | 51 ++++++++++++++++++++++++++++++- > ------- > drivers/net/virtio/virtqueue.h | 8 ++++++ > 4 files changed, 57 insertions(+), 16 deletions(-) >=20 > diff --git a/drivers/net/virtio/virtio_ethdev.c > b/drivers/net/virtio/virtio_ethdev.c > index 465d3cd..bcfb87b 100644 > --- a/drivers/net/virtio/virtio_ethdev.c > +++ b/drivers/net/virtio/virtio_ethdev.c > @@ -359,12 +359,15 @@ int virtio_dev_queue_setup(struct rte_eth_dev > *dev, > if (queue_type =3D=3D VTNET_TQ) { > /* > * For each xmit packet, allocate a virtio_net_hdr > + * and indirect ring elements > */ > snprintf(vq_name, sizeof(vq_name), > "port%d_tvq%d_hdrzone", > - dev->data->port_id, queue_idx); > - vq->virtio_net_hdr_mz =3D > rte_memzone_reserve_aligned(vq_name, > - vq_size * hw->vtnet_hdr_size, > - socket_id, 0, RTE_CACHE_LINE_SIZE); > + dev->data->port_id, queue_idx); > + > + vq->virtio_net_hdr_mz =3D > + rte_memzone_reserve_aligned(vq_name, > + vq_size * sizeof(struct > virtio_tx_region), > + socket_id, 0, > RTE_CACHE_LINE_SIZE); > if (vq->virtio_net_hdr_mz =3D=3D NULL) { > if (rte_errno =3D=3D EEXIST) > vq->virtio_net_hdr_mz =3D > diff --git a/drivers/net/virtio/virtio_ethdev.h > b/drivers/net/virtio/virtio_ethdev.h > index 9026d42..07a9265 100644 > --- a/drivers/net/virtio/virtio_ethdev.h > +++ b/drivers/net/virtio/virtio_ethdev.h > @@ -64,7 +64,8 @@ > 1u << VIRTIO_NET_F_CTRL_VQ | \ > 1u << VIRTIO_NET_F_CTRL_RX | \ > 1u << VIRTIO_NET_F_CTRL_VLAN | \ > - 1u << VIRTIO_NET_F_MRG_RXBUF) > + 1u << VIRTIO_NET_F_MRG_RXBUF | \ > + 1u << VIRTIO_RING_F_INDIRECT_DESC) >=20 > /* > * CQ function prototype > diff --git a/drivers/net/virtio/virtio_rxtx.c b/drivers/net/virtio/virtio= _rxtx.c > index dbe6665..8979695 100644 > --- a/drivers/net/virtio/virtio_rxtx.c > +++ b/drivers/net/virtio/virtio_rxtx.c > @@ -199,14 +199,15 @@ virtqueue_enqueue_recv_refill(struct virtqueue > *vq, struct rte_mbuf *cookie) } >=20 > static int > -virtqueue_enqueue_xmit(struct virtqueue *txvq, struct rte_mbuf *cookie) > +virtqueue_enqueue_xmit(struct virtqueue *txvq, struct rte_mbuf *cookie, > + int use_indirect) > { > struct vq_desc_extra *dxp; > struct vring_desc *start_dp; > uint16_t seg_num =3D cookie->nb_segs; > - uint16_t needed =3D 1 + seg_num; > + uint16_t needed =3D use_indirect ? 1 : 1 + seg_num; > uint16_t head_idx, idx; > - uint16_t head_size =3D txvq->hw->vtnet_hdr_size; > + unsigned long offs; >=20 > if (unlikely(txvq->vq_free_cnt =3D=3D 0)) > return -ENOSPC; > @@ -220,11 +221,26 @@ virtqueue_enqueue_xmit(struct virtqueue *txvq, > struct rte_mbuf *cookie) > dxp =3D &txvq->vq_descx[idx]; > dxp->cookie =3D (void *)cookie; > dxp->ndescs =3D needed; > - > start_dp =3D txvq->vq_ring.desc; > - start_dp[idx].addr =3D > - txvq->virtio_net_hdr_mem + idx * head_size; > - start_dp[idx].len =3D (uint32_t)head_size; > + > + if (use_indirect) { > + offs =3D offsetof(struct virtio_tx_region, tx_indir) > + + idx * sizeof(struct virtio_tx_region); > + > + start_dp[idx].addr =3D txvq->virtio_net_hdr_mem + offs; > + start_dp[idx].len =3D sizeof(struct vring_desc); Should the length be N * sizeof(struct vring_desc)? > + start_dp[idx].flags =3D VRING_DESC_F_INDIRECT; > + > + start_dp =3D (struct vring_desc *) > + ((char *)txvq->virtio_net_hdr_mz->addr + offs); > + idx =3D 0; > + } > + > + offs =3D offsetof(struct virtio_tx_region, tx_hdr) > + + idx * sizeof(struct virtio_tx_region); > + > + start_dp[idx].addr =3D txvq->virtio_net_hdr_mem + offs; > + start_dp[idx].len =3D txvq->hw->vtnet_hdr_size; > start_dp[idx].flags =3D VRING_DESC_F_NEXT; >=20 > for (; ((seg_num > 0) && (cookie !=3D NULL)); seg_num--) { @@ -236,7 > +252,12 @@ virtqueue_enqueue_xmit(struct virtqueue *txvq, struct > rte_mbuf *cookie) > } >=20 > start_dp[idx].flags &=3D ~VRING_DESC_F_NEXT; > - idx =3D start_dp[idx].next; > + > + if (use_indirect) > + idx =3D txvq->vq_ring.desc[head_idx].next; > + else > + idx =3D start_dp[idx].next; > + > txvq->vq_desc_head_idx =3D idx; > if (txvq->vq_desc_head_idx =3D=3D VQ_RING_DESC_CHAIN_END) > txvq->vq_desc_tail_idx =3D idx; > @@ -760,7 +781,15 @@ virtio_xmit_pkts(void *tx_queue, struct rte_mbuf > **tx_pkts, uint16_t nb_pkts) >=20 > for (nb_tx =3D 0; nb_tx < nb_pkts; nb_tx++) { > struct rte_mbuf *txm =3D tx_pkts[nb_tx]; > - int need =3D txm->nb_segs - txvq->vq_free_cnt + 1; > + int use_indirect, slots, need; > + > + use_indirect =3D vtpci_with_feature(txvq->hw, > + > VIRTIO_RING_F_INDIRECT_DESC) > + && (txm->nb_segs < VIRTIO_MAX_TX_INDIRECT); > + > + /* How many ring entries are needed to this Tx? */ > + slots =3D use_indirect ? 1 : 1 + txm->nb_segs; > + need =3D slots - txvq->vq_free_cnt; >=20 > /* Positive value indicates it need free vring descriptors */ > if (need > 0) { > @@ -769,7 +798,7 @@ virtio_xmit_pkts(void *tx_queue, struct rte_mbuf > **tx_pkts, uint16_t nb_pkts) > need =3D RTE_MIN(need, (int)nb_used); >=20 > virtio_xmit_cleanup(txvq, need); > - need =3D txm->nb_segs - txvq->vq_free_cnt + 1; > + need =3D slots - txvq->vq_free_cnt; > if (unlikely(need > 0)) { > PMD_TX_LOG(ERR, > "No free tx descriptors to transmit"); > @@ -787,7 +816,7 @@ virtio_xmit_pkts(void *tx_queue, struct rte_mbuf > **tx_pkts, uint16_t nb_pkts) > } >=20 > /* Enqueue Packet buffers */ > - error =3D virtqueue_enqueue_xmit(txvq, txm); > + error =3D virtqueue_enqueue_xmit(txvq, txm, use_indirect); > if (unlikely(error)) { > if (error =3D=3D ENOSPC) > PMD_TX_LOG(ERR, "virtqueue_enqueue > Free count =3D 0"); diff --git a/drivers/net/virtio/virtqueue.h > b/drivers/net/virtio/virtqueue.h index 7789411..a9410b4 100644 > --- a/drivers/net/virtio/virtqueue.h > +++ b/drivers/net/virtio/virtqueue.h > @@ -237,6 +237,14 @@ struct virtio_net_hdr_mrg_rxbuf { > uint16_t num_buffers; /**< Number of merged rx buffers */ }; >=20 > +/* Region reserved to allow for transmit header and indirect ring */ > +#define VIRTIO_MAX_TX_INDIRECT 8 struct virtio_tx_region { > + struct virtio_net_hdr_mrg_rxbuf tx_hdr; > + struct vring_desc tx_indir[VIRTIO_MAX_TX_INDIRECT] > + __attribute__((__aligned__(16))); }; > + > /** > * Tell the backend not to interrupt us. > */ > -- > 2.1.4