From patchwork Tue Nov 28 10:54:47 2017 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: junjie.j.chen@intel.com X-Patchwork-Id: 31700 X-Patchwork-Delegate: yuanhan.liu@linux.intel.com Return-Path: X-Original-To: patchwork@dpdk.org Delivered-To: patchwork@dpdk.org Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id B742B2BF2; Tue, 28 Nov 2017 04:13:18 +0100 (CET) Received: from mga04.intel.com (mga04.intel.com [192.55.52.120]) by dpdk.org (Postfix) with ESMTP id C06AF271 for ; Tue, 28 Nov 2017 04:13:16 +0100 (CET) Received: from orsmga005.jf.intel.com ([10.7.209.41]) by fmsmga104.fm.intel.com with ESMTP/TLS/DHE-RSA-AES256-GCM-SHA384; 27 Nov 2017 19:13:15 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.44,466,1505804400"; d="scan'208";a="178541651" Received: from pppoe-test.sh.intel.com ([10.67.111.147]) by orsmga005.jf.intel.com with ESMTP; 27 Nov 2017 19:13:13 -0800 From: junjie.j.chen@intel.com To: yliu@fridaylinux.org, maxime.coquelin@redhat.com, tiwei.bie@intel.com, jianfeng.tan@intel.com, dev@dpdk.org Cc: Junjie Chen Date: Tue, 28 Nov 2017 05:54:47 -0500 Message-Id: <20171128105447.59381-1-junjie.j.chen@intel.com> X-Mailer: git-send-email 2.15.0 Subject: [dpdk-dev] [PATCH v3] vhost: support virtqueue interrupt/notification suppression X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" From: Junjie Chen The driver can suppress interrupt when VIRTIO_F_EVENT_IDX feature bit is negotiated. The driver set vring flags to 0, and MAY use used_event in available ring to advise device interrupt util reach an index specified by used_event. The device ignore the lower bit of vring flags, and send an interrupt when index reach used_event. The device can suppress notification in a manner analogous to the ways driver suppress interrupt. The device manipulates flags or avail_event in the used ringin the same way the driver manipulates flags or used_event in available ring. This patch is to enable this feature in vhost. Signed-off-by: Junjie Chen --- lib/librte_vhost/Makefile | 1 + lib/librte_vhost/vhost.h | 7 ++++++ lib/librte_vhost/virtio_net.c | 56 +++++++++++++++++++++++++++++++------------ 3 files changed, 49 insertions(+), 15 deletions(-) diff --git a/lib/librte_vhost/Makefile b/lib/librte_vhost/Makefile index be182798a..85695d443 100644 --- a/lib/librte_vhost/Makefile +++ b/lib/librte_vhost/Makefile @@ -40,6 +40,7 @@ LIBABIVER := 4 CFLAGS += $(WERROR_FLAGS) -I$(SRCDIR) -O3 -D_FILE_OFFSET_BITS=64 CFLAGS += -I vhost_user +CFLAGS += -fno-strict-aliasing LDLIBS += -lpthread ifeq ($(CONFIG_RTE_LIBRTE_VHOST_NUMA),y) diff --git a/lib/librte_vhost/vhost.h b/lib/librte_vhost/vhost.h index 1cc81c17c..0741a2399 100644 --- a/lib/librte_vhost/vhost.h +++ b/lib/librte_vhost/vhost.h @@ -103,6 +103,8 @@ struct vhost_virtqueue { uint16_t last_avail_idx; uint16_t last_used_idx; + /* Last used index we notify to front end. */ + uint16_t signalled_used; #define VIRTIO_INVALID_EVENTFD (-1) #define VIRTIO_UNINITIALIZED_EVENTFD (-2) @@ -195,6 +197,10 @@ struct vhost_msg { #define VHOST_USER_F_PROTOCOL_FEATURES 30 +#ifndef VIRTIO_F_EVENT_IDX + #define VIRTIO_F_EVENT_IDX 29 +#endif + /* Features supported by this builtin vhost-user net driver. */ #define VIRTIO_NET_SUPPORTED_FEATURES ((1ULL << VIRTIO_NET_F_MRG_RXBUF) | \ (1ULL << VIRTIO_NET_F_CTRL_VQ) | \ @@ -211,6 +217,7 @@ struct vhost_msg { (1ULL << VIRTIO_NET_F_GUEST_TSO4) | \ (1ULL << VIRTIO_NET_F_GUEST_TSO6) | \ (1ULL << VIRTIO_RING_F_INDIRECT_DESC) | \ + (1ULL << VIRTIO_RING_F_EVENT_IDX) | \ (1ULL << VIRTIO_NET_F_MTU) | \ (1ULL << VIRTIO_F_IOMMU_PLATFORM)) diff --git a/lib/librte_vhost/virtio_net.c b/lib/librte_vhost/virtio_net.c index 6fee16e55..a3895658f 100644 --- a/lib/librte_vhost/virtio_net.c +++ b/lib/librte_vhost/virtio_net.c @@ -52,6 +52,35 @@ #define MAX_BATCH_LEN 256 +#define vhost_used_event(vr) ((vr)->avail->ring[(vr)->size]) +#define vhost_avail_event(vr) \ + (*(volatile uint16_t *)&(vr)->used->ring[(vr)->size]) + +static __rte_always_inline void +vhost_notify(struct virtio_net *dev, struct vhost_virtqueue *vq) +{ + /* Don't notify guest if we don't reach index specified by guest. */ + if (dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX)) { + uint16_t old = vq->signalled_used; + uint16_t new = vq->last_used_idx; + + LOG_DEBUG(VHOST_DATA, "%s: used_event_idx=%d, old=%d, new=%d\n", + __func__, + vhost_used_event(vq), + old, new); + if (vring_need_event(vhost_used_event(vq), new, old) + && (vq->callfd >= 0)) { + vq->signalled_used = vq->last_used_idx; + eventfd_write(vq->callfd, (eventfd_t) 1); + } + } else { + /* Kick the guest if necessary. */ + if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT) + && (vq->callfd >= 0)) + eventfd_write(vq->callfd, (eventfd_t)1); + } +} + static bool is_valid_virt_queue_idx(uint32_t idx, int is_tx, uint32_t nr_vring) { @@ -410,11 +439,7 @@ virtio_dev_rx(struct virtio_net *dev, uint16_t queue_id, /* flush used->idx update before we read avail->flags. */ rte_mb(); - - /* Kick the guest if necessary. */ - if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT) - && (vq->callfd >= 0)) - eventfd_write(vq->callfd, (eventfd_t)1); + vhost_notify(dev, vq); out: if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) vhost_user_iotlb_rd_unlock(vq); @@ -704,11 +729,7 @@ virtio_dev_merge_rx(struct virtio_net *dev, uint16_t queue_id, /* flush used->idx update before we read avail->flags. */ rte_mb(); - - /* Kick the guest if necessary. */ - if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT) - && (vq->callfd >= 0)) - eventfd_write(vq->callfd, (eventfd_t)1); + vhost_notify(dev, vq); } out: @@ -1106,11 +1127,7 @@ update_used_idx(struct virtio_net *dev, struct vhost_virtqueue *vq, vq->used->idx += count; vhost_log_used_vring(dev, vq, offsetof(struct vring_used, idx), sizeof(vq->used->idx)); - - /* Kick guest if required. */ - if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT) - && (vq->callfd >= 0)) - eventfd_write(vq->callfd, (eventfd_t)1); + vhost_notify(dev, vq); } static __rte_always_inline struct zcopy_mbuf * @@ -1352,6 +1369,15 @@ rte_vhost_dequeue_burst(int vid, uint16_t queue_id, update_used_idx(dev, vq, i); } + /* + * Update avail event index to last_avail_idx to expect + * front end notify us after next buffer available. + */ + if (dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX)) { + vhost_avail_event(vq) = vq->last_avail_idx; + rte_mb(); + } + out: if (dev->features & (1ULL << VIRTIO_F_IOMMU_PLATFORM)) vhost_user_iotlb_rd_unlock(vq);