DPDK patches and discussions
 help / color / mirror / Atom feed
From: Maxime Coquelin <maxime.coquelin@redhat.com>
To: dev@dpdk.org, david.marchand@redhat.com, chenbox@nvidia.com
Cc: Maxime Coquelin <maxime.coquelin@redhat.com>
Subject: [PATCH v3 1/2] vhost: add logging mechanism for reconnection
Date: Mon, 23 Sep 2024 21:51:50 +0200	[thread overview]
Message-ID: <20240923195151.73527-2-maxime.coquelin@redhat.com> (raw)
In-Reply-To: <20240923195151.73527-1-maxime.coquelin@redhat.com>

This patch introduces a way for backend to keep track
of the needed information to be able to reconnect without
frontend cooperation.

It will be used for VDUSE, which does not provide interface
for the backend to save and later recover local virtqueues
metadata needed to reconnect.

Vhost-user support could also be added for improved packed
ring reconnection support.

Signed-off-by: Maxime Coquelin <maxime.coquelin@redhat.com>
Reviewed-by: Chenbo Xia <chenbox@nvidia.com>
Reviewed-by: David Marchand <david.marchand@redhat.com>
---
 lib/vhost/vhost.c           |  2 ++
 lib/vhost/vhost.h           | 41 ++++++++++++++++++++++++++++++++++---
 lib/vhost/vhost_user.c      |  4 ++++
 lib/vhost/virtio_net.c      |  8 ++++++++
 lib/vhost/virtio_net_ctrl.c |  2 ++
 5 files changed, 54 insertions(+), 3 deletions(-)

diff --git a/lib/vhost/vhost.c b/lib/vhost/vhost.c
index ac71d17784..5a50a06f8d 100644
--- a/lib/vhost/vhost.c
+++ b/lib/vhost/vhost.c
@@ -1712,9 +1712,11 @@ rte_vhost_set_vring_base(int vid, uint16_t queue_id,
 		vq->avail_wrap_counter = !!(last_avail_idx & (1 << 15));
 		vq->last_used_idx = last_used_idx & 0x7fff;
 		vq->used_wrap_counter = !!(last_used_idx & (1 << 15));
+		vhost_virtqueue_reconnect_log_packed(vq);
 	} else {
 		vq->last_avail_idx = last_avail_idx;
 		vq->last_used_idx = last_used_idx;
+		vhost_virtqueue_reconnect_log_split(vq);
 	}
 
 	return 0;
diff --git a/lib/vhost/vhost.h b/lib/vhost/vhost.h
index cd3fa55f1b..1f4192f5d1 100644
--- a/lib/vhost/vhost.h
+++ b/lib/vhost/vhost.h
@@ -269,6 +269,24 @@ struct vhost_async {
 	};
 };
 
+#define VHOST_RECONNECT_VERSION		0x0
+#define VHOST_MAX_VRING			0x100
+#define VHOST_MAX_QUEUE_PAIRS		0x80
+
+struct __rte_cache_aligned vhost_reconnect_vring {
+	uint16_t last_avail_idx;
+	bool avail_wrap_counter;
+};
+
+struct vhost_reconnect_data {
+	uint32_t version;
+	uint64_t features;
+	uint8_t status;
+	struct virtio_net_config config;
+	uint32_t nr_vrings;
+	struct vhost_reconnect_vring vring[VHOST_MAX_VRING];
+};
+
 /**
  * Structure contains variables relevant to RX/TX virtqueues.
  */
@@ -351,6 +369,7 @@ struct __rte_cache_aligned vhost_virtqueue {
 	struct virtqueue_stats	stats;
 
 	RTE_ATOMIC(bool) irq_pending;
+	struct vhost_reconnect_vring *reconnect_log;
 };
 
 /* Virtio device status as per Virtio specification */
@@ -362,9 +381,6 @@ struct __rte_cache_aligned vhost_virtqueue {
 #define VIRTIO_DEVICE_STATUS_DEV_NEED_RESET	0x40
 #define VIRTIO_DEVICE_STATUS_FAILED		0x80
 
-#define VHOST_MAX_VRING			0x100
-#define VHOST_MAX_QUEUE_PAIRS		0x80
-
 /* Declare IOMMU related bits for older kernels */
 #ifndef VIRTIO_F_IOMMU_PLATFORM
 
@@ -538,8 +554,26 @@ struct __rte_cache_aligned virtio_net {
 	struct rte_vhost_user_extern_ops extern_ops;
 
 	struct vhost_backend_ops *backend_ops;
+
+	struct vhost_reconnect_data *reconnect_log;
 };
 
+static __rte_always_inline void
+vhost_virtqueue_reconnect_log_split(struct vhost_virtqueue *vq)
+{
+	if (vq->reconnect_log != NULL)
+		vq->reconnect_log->last_avail_idx = vq->last_avail_idx;
+}
+
+static __rte_always_inline void
+vhost_virtqueue_reconnect_log_packed(struct vhost_virtqueue *vq)
+{
+	if (vq->reconnect_log != NULL) {
+		vq->reconnect_log->last_avail_idx = vq->last_avail_idx;
+		vq->reconnect_log->avail_wrap_counter = vq->avail_wrap_counter;
+	}
+}
+
 static inline void
 vq_assert_lock__(struct virtio_net *dev, struct vhost_virtqueue *vq, const char *func)
 	__rte_assert_exclusive_lock(&vq->access_lock)
@@ -584,6 +618,7 @@ vq_inc_last_avail_packed(struct vhost_virtqueue *vq, uint16_t num)
 		vq->avail_wrap_counter ^= 1;
 		vq->last_avail_idx -= vq->size;
 	}
+	vhost_virtqueue_reconnect_log_packed(vq);
 }
 
 void __vhost_log_cache_write(struct virtio_net *dev,
diff --git a/lib/vhost/vhost_user.c b/lib/vhost/vhost_user.c
index 5f470da38a..d20b9e8497 100644
--- a/lib/vhost/vhost_user.c
+++ b/lib/vhost/vhost_user.c
@@ -954,6 +954,7 @@ translate_ring_addresses(struct virtio_net **pdev, struct vhost_virtqueue **pvq)
 			vq->last_used_idx, vq->used->idx);
 		vq->last_used_idx  = vq->used->idx;
 		vq->last_avail_idx = vq->used->idx;
+		vhost_virtqueue_reconnect_log_split(vq);
 		VHOST_CONFIG_LOG(dev->ifname, WARNING,
 			"some packets maybe resent for Tx and dropped for Rx");
 	}
@@ -1039,9 +1040,11 @@ vhost_user_set_vring_base(struct virtio_net **pdev,
 		 */
 		vq->last_used_idx = vq->last_avail_idx;
 		vq->used_wrap_counter = vq->avail_wrap_counter;
+		vhost_virtqueue_reconnect_log_packed(vq);
 	} else {
 		vq->last_used_idx = ctx->msg.payload.state.num;
 		vq->last_avail_idx = ctx->msg.payload.state.num;
+		vhost_virtqueue_reconnect_log_split(vq);
 	}
 
 	VHOST_CONFIG_LOG(dev->ifname, INFO,
@@ -1997,6 +2000,7 @@ vhost_check_queue_inflights_split(struct virtio_net *dev,
 	}
 
 	vq->last_avail_idx += resubmit_num;
+	vhost_virtqueue_reconnect_log_split(vq);
 
 	if (resubmit_num) {
 		resubmit = rte_zmalloc_socket("resubmit", sizeof(struct rte_vhost_resubmit_info),
diff --git a/lib/vhost/virtio_net.c b/lib/vhost/virtio_net.c
index 370402d849..f66a0c82f8 100644
--- a/lib/vhost/virtio_net.c
+++ b/lib/vhost/virtio_net.c
@@ -1445,6 +1445,7 @@ virtio_dev_rx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		}
 
 		vq->last_avail_idx += num_buffers;
+		vhost_virtqueue_reconnect_log_split(vq);
 	}
 
 	do_data_copy_enqueue(dev, vq);
@@ -1857,6 +1858,7 @@ virtio_dev_rx_async_submit_split(struct virtio_net *dev, struct vhost_virtqueue
 		pkts_info[slot_idx].mbuf = pkts[pkt_idx];
 
 		vq->last_avail_idx += num_buffers;
+		vhost_virtqueue_reconnect_log_split(vq);
 	}
 
 	if (unlikely(pkt_idx == 0))
@@ -1885,6 +1887,7 @@ virtio_dev_rx_async_submit_split(struct virtio_net *dev, struct vhost_virtqueue
 		/* recover shadow used ring and available ring */
 		vq->shadow_used_idx -= num_descs;
 		vq->last_avail_idx -= num_descs;
+		vhost_virtqueue_reconnect_log_split(vq);
 	}
 
 	/* keep used descriptors */
@@ -2100,6 +2103,7 @@ dma_error_handler_packed(struct vhost_virtqueue *vq, uint16_t slot_idx,
 		vq->last_avail_idx = vq->last_avail_idx + vq->size - descs_err;
 		vq->avail_wrap_counter ^= 1;
 	}
+	vhost_virtqueue_reconnect_log_packed(vq);
 
 	if (async->buffer_idx_packed >= buffers_err)
 		async->buffer_idx_packed -= buffers_err;
@@ -3182,6 +3186,7 @@ virtio_dev_tx_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
 
 	if (likely(vq->shadow_used_idx)) {
 		vq->last_avail_idx += vq->shadow_used_idx;
+		vhost_virtqueue_reconnect_log_split(vq);
 		do_data_copy_dequeue(vq);
 		flush_shadow_used_ring_split(dev, vq);
 		vhost_vring_call_split(dev, vq);
@@ -3854,6 +3859,7 @@ virtio_dev_tx_async_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		async->desc_idx_split++;
 
 		vq->last_avail_idx++;
+		vhost_virtqueue_reconnect_log_split(vq);
 	}
 
 	if (unlikely(dropped))
@@ -3872,6 +3878,7 @@ virtio_dev_tx_async_split(struct virtio_net *dev, struct vhost_virtqueue *vq,
 		pkt_idx = n_xfer;
 		/* recover available ring */
 		vq->last_avail_idx -= pkt_err;
+		vhost_virtqueue_reconnect_log_split(vq);
 
 		/**
 		 * recover async channel copy related structures and free pktmbufs
@@ -4153,6 +4160,7 @@ virtio_dev_tx_async_packed(struct virtio_net *dev, struct vhost_virtqueue *vq,
 			vq->last_avail_idx += vq->size - descs_err;
 			vq->avail_wrap_counter ^= 1;
 		}
+		vhost_virtqueue_reconnect_log_packed(vq);
 	}
 
 	async->pkts_idx += pkt_idx;
diff --git a/lib/vhost/virtio_net_ctrl.c b/lib/vhost/virtio_net_ctrl.c
index 8f78122361..b8ee94018e 100644
--- a/lib/vhost/virtio_net_ctrl.c
+++ b/lib/vhost/virtio_net_ctrl.c
@@ -169,6 +169,7 @@ virtio_net_ctrl_pop(struct virtio_net *dev, struct vhost_virtqueue *cvq,
 	cvq->last_avail_idx++;
 	if (cvq->last_avail_idx >= cvq->size)
 		cvq->last_avail_idx -= cvq->size;
+	vhost_virtqueue_reconnect_log_split(cvq);
 
 	if (dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX))
 		vhost_avail_event(cvq) = cvq->last_avail_idx;
@@ -181,6 +182,7 @@ virtio_net_ctrl_pop(struct virtio_net *dev, struct vhost_virtqueue *cvq,
 	cvq->last_avail_idx++;
 	if (cvq->last_avail_idx >= cvq->size)
 		cvq->last_avail_idx -= cvq->size;
+	vhost_virtqueue_reconnect_log_split(cvq);
 
 	if (dev->features & (1ULL << VIRTIO_RING_F_EVENT_IDX))
 		vhost_avail_event(cvq) = cvq->last_avail_idx;
-- 
2.46.0


  reply	other threads:[~2024-09-23 19:52 UTC|newest]

Thread overview: 4+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-09-23 19:51 [PATCH v3 0/2] vhost: add VDUSE reconnection support Maxime Coquelin
2024-09-23 19:51 ` Maxime Coquelin [this message]
2024-09-23 19:51 ` [PATCH v3 2/2] vhost: add reconnection support to VDUSE Maxime Coquelin
2024-09-24 18:22 ` [PATCH v3 0/2] vhost: add VDUSE reconnection support Maxime Coquelin

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20240923195151.73527-2-maxime.coquelin@redhat.com \
    --to=maxime.coquelin@redhat.com \
    --cc=chenbox@nvidia.com \
    --cc=david.marchand@redhat.com \
    --cc=dev@dpdk.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).