* [dpdk-dev] [PATCH 0/2] netvsc pmd fixes for 19.05
@ 2019-04-29 20:33 Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
` (3 more replies)
0 siblings, 4 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-29 20:33 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger
Netvsc needed to reset incoming port as was recently done
for failsafe. Implement the new close semantics.
Stephen Hemminger (2):
net/netvsc: reset mbuf port on VF receive
net/netvsc: free all queues on close
drivers/net/netvsc/hn_ethdev.c | 8 +++-
drivers/net/netvsc/hn_rxtx.c | 86 ++++++++++++++++++++++++----------
drivers/net/netvsc/hn_var.h | 1 +
3 files changed, 68 insertions(+), 27 deletions(-)
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH 0/2] netvsc pmd fixes for 19.05
2019-04-29 20:33 [dpdk-dev] [PATCH 0/2] netvsc pmd fixes for 19.05 Stephen Hemminger
@ 2019-04-29 20:33 ` Stephen Hemminger
2019-04-29 20:33 ` [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive Stephen Hemminger
` (2 subsequent siblings)
3 siblings, 0 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-29 20:33 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger
Netvsc needed to reset incoming port as was recently done
for failsafe. Implement the new close semantics.
Stephen Hemminger (2):
net/netvsc: reset mbuf port on VF receive
net/netvsc: free all queues on close
drivers/net/netvsc/hn_ethdev.c | 8 +++-
drivers/net/netvsc/hn_rxtx.c | 86 ++++++++++++++++++++++++----------
drivers/net/netvsc/hn_var.h | 1 +
3 files changed, 68 insertions(+), 27 deletions(-)
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive
2019-04-29 20:33 [dpdk-dev] [PATCH 0/2] netvsc pmd fixes for 19.05 Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
@ 2019-04-29 20:33 ` Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
` (2 more replies)
2019-04-29 20:33 ` [dpdk-dev] [PATCH 2/2] net/netvsc: free all queues on close Stephen Hemminger
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Stephen Hemminger
3 siblings, 3 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-29 20:33 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger, Stephen Hemminger
Redo the receive logic to set m->port on packets received on VF.
When using VF, still need to check for packets and completions
arriving on the VMBus path even if application is not doing bursting
(ie n_rx == 0).
Also, fix comment.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
---
drivers/net/netvsc/hn_rxtx.c | 47 +++++++++++++++++++++---------------
1 file changed, 28 insertions(+), 19 deletions(-)
diff --git a/drivers/net/netvsc/hn_rxtx.c b/drivers/net/netvsc/hn_rxtx.c
index c67e9ae25376..7856f7e6ec48 100644
--- a/drivers/net/netvsc/hn_rxtx.c
+++ b/drivers/net/netvsc/hn_rxtx.c
@@ -1393,6 +1393,24 @@ hn_xmit_pkts(void *ptxq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
return nb_tx;
}
+static uint16_t
+hn_recv_vf(uint16_t vf_port, const struct hn_rx_queue *rxq,
+ struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+ uint16_t i, n;
+
+ if (unlikely(nb_pkts == 0))
+ return 0;
+
+ n = rte_eth_rx_burst(vf_port, rxq->queue_id, rx_pkts, nb_pkts);
+
+ /* relabel the received mbufs */
+ for (i = 0; i < n; i++)
+ rx_pkts[i]->port = rxq->port_id;
+
+ return n;
+}
+
uint16_t
hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
{
@@ -1404,30 +1422,21 @@ hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
if (unlikely(hv->closed))
return 0;
- /* Transmit over VF if present and up */
+ /* Receive from VF if present and up */
vf_dev = hn_get_vf_dev(hv);
- if (vf_dev && vf_dev->data->dev_started) {
- /* Normally, with SR-IOV the ring buffer will be empty */
+ /* Check for new completions */
+ if (likely(rte_ring_count(rxq->rx_ring) < nb_pkts))
hn_process_events(hv, rxq->queue_id, 0);
- /* Get mbufs some bufs off of staging ring */
- nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
- (void **)rx_pkts,
- nb_pkts / 2, NULL);
- /* And rest off of VF */
- nb_rcv += rte_eth_rx_burst(vf_dev->data->port_id,
- rxq->queue_id,
- rx_pkts + nb_rcv, nb_pkts - nb_rcv);
- } else {
- /* If receive ring is not full then get more */
- if (rte_ring_count(rxq->rx_ring) < nb_pkts)
- hn_process_events(hv, rxq->queue_id, 0);
+ /* Always check the vmbus path for multicast and new flows */
+ nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
+ (void **)rx_pkts, nb_pkts, NULL);
- nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
- (void **)rx_pkts,
- nb_pkts, NULL);
- }
+ /* If VF is available, check that as well */
+ if (vf_dev && vf_dev->data->dev_started)
+ nb_rcv += hn_recv_vf(vf_dev->data->port_id, rxq,
+ rx_pkts + nb_rcv, nb_pkts - nb_rcv);
return nb_rcv;
}
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive
2019-04-29 20:33 ` [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive Stephen Hemminger
@ 2019-04-29 20:33 ` Stephen Hemminger
2019-04-30 8:01 ` Ferruh Yigit
2019-04-30 15:05 ` Stephen Hemminger
2 siblings, 0 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-29 20:33 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger, Stephen Hemminger
Redo the receive logic to set m->port on packets received on VF.
When using VF, still need to check for packets and completions
arriving on the VMBus path even if application is not doing bursting
(ie n_rx == 0).
Also, fix comment.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
---
drivers/net/netvsc/hn_rxtx.c | 47 +++++++++++++++++++++---------------
1 file changed, 28 insertions(+), 19 deletions(-)
diff --git a/drivers/net/netvsc/hn_rxtx.c b/drivers/net/netvsc/hn_rxtx.c
index c67e9ae25376..7856f7e6ec48 100644
--- a/drivers/net/netvsc/hn_rxtx.c
+++ b/drivers/net/netvsc/hn_rxtx.c
@@ -1393,6 +1393,24 @@ hn_xmit_pkts(void *ptxq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
return nb_tx;
}
+static uint16_t
+hn_recv_vf(uint16_t vf_port, const struct hn_rx_queue *rxq,
+ struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+ uint16_t i, n;
+
+ if (unlikely(nb_pkts == 0))
+ return 0;
+
+ n = rte_eth_rx_burst(vf_port, rxq->queue_id, rx_pkts, nb_pkts);
+
+ /* relabel the received mbufs */
+ for (i = 0; i < n; i++)
+ rx_pkts[i]->port = rxq->port_id;
+
+ return n;
+}
+
uint16_t
hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
{
@@ -1404,30 +1422,21 @@ hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
if (unlikely(hv->closed))
return 0;
- /* Transmit over VF if present and up */
+ /* Receive from VF if present and up */
vf_dev = hn_get_vf_dev(hv);
- if (vf_dev && vf_dev->data->dev_started) {
- /* Normally, with SR-IOV the ring buffer will be empty */
+ /* Check for new completions */
+ if (likely(rte_ring_count(rxq->rx_ring) < nb_pkts))
hn_process_events(hv, rxq->queue_id, 0);
- /* Get mbufs some bufs off of staging ring */
- nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
- (void **)rx_pkts,
- nb_pkts / 2, NULL);
- /* And rest off of VF */
- nb_rcv += rte_eth_rx_burst(vf_dev->data->port_id,
- rxq->queue_id,
- rx_pkts + nb_rcv, nb_pkts - nb_rcv);
- } else {
- /* If receive ring is not full then get more */
- if (rte_ring_count(rxq->rx_ring) < nb_pkts)
- hn_process_events(hv, rxq->queue_id, 0);
+ /* Always check the vmbus path for multicast and new flows */
+ nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
+ (void **)rx_pkts, nb_pkts, NULL);
- nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
- (void **)rx_pkts,
- nb_pkts, NULL);
- }
+ /* If VF is available, check that as well */
+ if (vf_dev && vf_dev->data->dev_started)
+ nb_rcv += hn_recv_vf(vf_dev->data->port_id, rxq,
+ rx_pkts + nb_rcv, nb_pkts - nb_rcv);
return nb_rcv;
}
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH 2/2] net/netvsc: free all queues on close
2019-04-29 20:33 [dpdk-dev] [PATCH 0/2] netvsc pmd fixes for 19.05 Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
2019-04-29 20:33 ` [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive Stephen Hemminger
@ 2019-04-29 20:33 ` Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
2019-04-30 8:11 ` Ferruh Yigit
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Stephen Hemminger
3 siblings, 2 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-29 20:33 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger, Stephen Hemminger
When dev_close is called, the netvsc driver will clean up all
queues including the primary ring buffer.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
---
drivers/net/netvsc/hn_ethdev.c | 8 +++++--
drivers/net/netvsc/hn_rxtx.c | 39 ++++++++++++++++++++++++++++------
drivers/net/netvsc/hn_var.h | 1 +
3 files changed, 40 insertions(+), 8 deletions(-)
diff --git a/drivers/net/netvsc/hn_ethdev.c b/drivers/net/netvsc/hn_ethdev.c
index 407ee484935a..553cb06f6e33 100644
--- a/drivers/net/netvsc/hn_ethdev.c
+++ b/drivers/net/netvsc/hn_ethdev.c
@@ -112,6 +112,9 @@ eth_dev_vmbus_allocate(struct rte_vmbus_device *dev, size_t private_data_size)
eth_dev->data->dev_flags |= RTE_ETH_DEV_INTR_LSC;
eth_dev->intr_handle = &dev->intr_handle;
+ /* allow ethdev to remove on close */
+ eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
+
return eth_dev;
}
@@ -632,11 +635,12 @@ hn_dev_stop(struct rte_eth_dev *dev)
}
static void
-hn_dev_close(struct rte_eth_dev *dev __rte_unused)
+hn_dev_close(struct rte_eth_dev *dev)
{
- PMD_INIT_LOG(DEBUG, "close");
+ PMD_INIT_FUNC_TRACE();
hn_vf_close(dev);
+ hn_dev_free_queues(dev);
}
static const struct eth_dev_ops hn_eth_dev_ops = {
diff --git a/drivers/net/netvsc/hn_rxtx.c b/drivers/net/netvsc/hn_rxtx.c
index 7856f7e6ec48..1bebae7cddd4 100644
--- a/drivers/net/netvsc/hn_rxtx.c
+++ b/drivers/net/netvsc/hn_rxtx.c
@@ -840,12 +840,9 @@ hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
return error;
}
-void
-hn_dev_rx_queue_release(void *arg)
+static void
+hn_rx_queue_free(struct hn_rx_queue *rxq, bool keep_primary)
{
- struct hn_rx_queue *rxq = arg;
-
- PMD_INIT_FUNC_TRACE();
if (!rxq)
return;
@@ -857,12 +854,22 @@ hn_dev_rx_queue_release(void *arg)
hn_vf_rx_queue_release(rxq->hv, rxq->queue_id);
/* Keep primary queue to allow for control operations */
- if (rxq != rxq->hv->primary) {
+ if (keep_primary && rxq != rxq->hv->primary) {
rte_free(rxq->event_buf);
rte_free(rxq);
}
}
+void
+hn_dev_rx_queue_release(void *arg)
+{
+ struct hn_rx_queue *rxq = arg;
+
+ PMD_INIT_FUNC_TRACE();
+
+ hn_rx_queue_free(rxq, true);
+}
+
int
hn_dev_tx_done_cleanup(void *arg, uint32_t free_cnt)
{
@@ -1440,3 +1447,23 @@ hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
return nb_rcv;
}
+
+void
+hn_dev_free_queues(struct rte_eth_dev *dev)
+{
+ unsigned int i;
+
+ for (i = 0; i < dev->data->nb_rx_queues; i++) {
+ struct hn_rx_queue *rxq = dev->data->rx_queues[i];
+
+ hn_rx_queue_free(rxq, false);
+ dev->data->rx_queues[i] = NULL;
+ }
+ dev->data->nb_rx_queues = 0;
+
+ for (i = 0; i < dev->data->nb_tx_queues; i++) {
+ hn_dev_tx_queue_release(dev->data->tx_queues[i]);
+ dev->data->tx_queues[i] = NULL;
+ }
+ dev->data->nb_tx_queues = 0;
+}
diff --git a/drivers/net/netvsc/hn_var.h b/drivers/net/netvsc/hn_var.h
index 8383f3246ca4..de885d898e6a 100644
--- a/drivers/net/netvsc/hn_var.h
+++ b/drivers/net/netvsc/hn_var.h
@@ -173,6 +173,7 @@ int hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
const struct rte_eth_rxconf *rx_conf,
struct rte_mempool *mp);
void hn_dev_rx_queue_release(void *arg);
+void hn_dev_free_queues(struct rte_eth_dev *dev);
/* Check if VF is attached */
static inline bool
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH 2/2] net/netvsc: free all queues on close
2019-04-29 20:33 ` [dpdk-dev] [PATCH 2/2] net/netvsc: free all queues on close Stephen Hemminger
@ 2019-04-29 20:33 ` Stephen Hemminger
2019-04-30 8:11 ` Ferruh Yigit
1 sibling, 0 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-29 20:33 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger, Stephen Hemminger
When dev_close is called, the netvsc driver will clean up all
queues including the primary ring buffer.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
---
drivers/net/netvsc/hn_ethdev.c | 8 +++++--
drivers/net/netvsc/hn_rxtx.c | 39 ++++++++++++++++++++++++++++------
drivers/net/netvsc/hn_var.h | 1 +
3 files changed, 40 insertions(+), 8 deletions(-)
diff --git a/drivers/net/netvsc/hn_ethdev.c b/drivers/net/netvsc/hn_ethdev.c
index 407ee484935a..553cb06f6e33 100644
--- a/drivers/net/netvsc/hn_ethdev.c
+++ b/drivers/net/netvsc/hn_ethdev.c
@@ -112,6 +112,9 @@ eth_dev_vmbus_allocate(struct rte_vmbus_device *dev, size_t private_data_size)
eth_dev->data->dev_flags |= RTE_ETH_DEV_INTR_LSC;
eth_dev->intr_handle = &dev->intr_handle;
+ /* allow ethdev to remove on close */
+ eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
+
return eth_dev;
}
@@ -632,11 +635,12 @@ hn_dev_stop(struct rte_eth_dev *dev)
}
static void
-hn_dev_close(struct rte_eth_dev *dev __rte_unused)
+hn_dev_close(struct rte_eth_dev *dev)
{
- PMD_INIT_LOG(DEBUG, "close");
+ PMD_INIT_FUNC_TRACE();
hn_vf_close(dev);
+ hn_dev_free_queues(dev);
}
static const struct eth_dev_ops hn_eth_dev_ops = {
diff --git a/drivers/net/netvsc/hn_rxtx.c b/drivers/net/netvsc/hn_rxtx.c
index 7856f7e6ec48..1bebae7cddd4 100644
--- a/drivers/net/netvsc/hn_rxtx.c
+++ b/drivers/net/netvsc/hn_rxtx.c
@@ -840,12 +840,9 @@ hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
return error;
}
-void
-hn_dev_rx_queue_release(void *arg)
+static void
+hn_rx_queue_free(struct hn_rx_queue *rxq, bool keep_primary)
{
- struct hn_rx_queue *rxq = arg;
-
- PMD_INIT_FUNC_TRACE();
if (!rxq)
return;
@@ -857,12 +854,22 @@ hn_dev_rx_queue_release(void *arg)
hn_vf_rx_queue_release(rxq->hv, rxq->queue_id);
/* Keep primary queue to allow for control operations */
- if (rxq != rxq->hv->primary) {
+ if (keep_primary && rxq != rxq->hv->primary) {
rte_free(rxq->event_buf);
rte_free(rxq);
}
}
+void
+hn_dev_rx_queue_release(void *arg)
+{
+ struct hn_rx_queue *rxq = arg;
+
+ PMD_INIT_FUNC_TRACE();
+
+ hn_rx_queue_free(rxq, true);
+}
+
int
hn_dev_tx_done_cleanup(void *arg, uint32_t free_cnt)
{
@@ -1440,3 +1447,23 @@ hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
return nb_rcv;
}
+
+void
+hn_dev_free_queues(struct rte_eth_dev *dev)
+{
+ unsigned int i;
+
+ for (i = 0; i < dev->data->nb_rx_queues; i++) {
+ struct hn_rx_queue *rxq = dev->data->rx_queues[i];
+
+ hn_rx_queue_free(rxq, false);
+ dev->data->rx_queues[i] = NULL;
+ }
+ dev->data->nb_rx_queues = 0;
+
+ for (i = 0; i < dev->data->nb_tx_queues; i++) {
+ hn_dev_tx_queue_release(dev->data->tx_queues[i]);
+ dev->data->tx_queues[i] = NULL;
+ }
+ dev->data->nb_tx_queues = 0;
+}
diff --git a/drivers/net/netvsc/hn_var.h b/drivers/net/netvsc/hn_var.h
index 8383f3246ca4..de885d898e6a 100644
--- a/drivers/net/netvsc/hn_var.h
+++ b/drivers/net/netvsc/hn_var.h
@@ -173,6 +173,7 @@ int hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
const struct rte_eth_rxconf *rx_conf,
struct rte_mempool *mp);
void hn_dev_rx_queue_release(void *arg);
+void hn_dev_free_queues(struct rte_eth_dev *dev);
/* Check if VF is attached */
static inline bool
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive
2019-04-29 20:33 ` [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
@ 2019-04-30 8:01 ` Ferruh Yigit
2019-04-30 8:01 ` Ferruh Yigit
2019-04-30 15:05 ` Stephen Hemminger
2 siblings, 1 reply; 20+ messages in thread
From: Ferruh Yigit @ 2019-04-30 8:01 UTC (permalink / raw)
To: Stephen Hemminger, dev; +Cc: Stephen Hemminger
On 4/29/2019 9:33 PM, Stephen Hemminger wrote:
> Redo the receive logic to set m->port on packets received on VF.
>
> When using VF, still need to check for packets and completions
> arriving on the VMBus path even if application is not doing bursting
> (ie n_rx == 0).
>
> Also, fix comment.
>
> Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
If this is a fix, can you please describe what is fixed and put a fixes line,
that will help for backporting the fix.
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive
2019-04-30 8:01 ` Ferruh Yigit
@ 2019-04-30 8:01 ` Ferruh Yigit
0 siblings, 0 replies; 20+ messages in thread
From: Ferruh Yigit @ 2019-04-30 8:01 UTC (permalink / raw)
To: Stephen Hemminger, dev; +Cc: Stephen Hemminger
On 4/29/2019 9:33 PM, Stephen Hemminger wrote:
> Redo the receive logic to set m->port on packets received on VF.
>
> When using VF, still need to check for packets and completions
> arriving on the VMBus path even if application is not doing bursting
> (ie n_rx == 0).
>
> Also, fix comment.
>
> Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
If this is a fix, can you please describe what is fixed and put a fixes line,
that will help for backporting the fix.
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [dpdk-dev] [PATCH 2/2] net/netvsc: free all queues on close
2019-04-29 20:33 ` [dpdk-dev] [PATCH 2/2] net/netvsc: free all queues on close Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
@ 2019-04-30 8:11 ` Ferruh Yigit
2019-04-30 8:11 ` Ferruh Yigit
1 sibling, 1 reply; 20+ messages in thread
From: Ferruh Yigit @ 2019-04-30 8:11 UTC (permalink / raw)
To: Stephen Hemminger, dev; +Cc: Stephen Hemminger
On 4/29/2019 9:33 PM, Stephen Hemminger wrote:
> When dev_close is called, the netvsc driver will clean up all
> queues including the primary ring buffer.
>
> Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
> ---
> drivers/net/netvsc/hn_ethdev.c | 8 +++++--
> drivers/net/netvsc/hn_rxtx.c | 39 ++++++++++++++++++++++++++++------
> drivers/net/netvsc/hn_var.h | 1 +
> 3 files changed, 40 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/net/netvsc/hn_ethdev.c b/drivers/net/netvsc/hn_ethdev.c
> index 407ee484935a..553cb06f6e33 100644
> --- a/drivers/net/netvsc/hn_ethdev.c
> +++ b/drivers/net/netvsc/hn_ethdev.c
> @@ -112,6 +112,9 @@ eth_dev_vmbus_allocate(struct rte_vmbus_device *dev, size_t private_data_size)
> eth_dev->data->dev_flags |= RTE_ETH_DEV_INTR_LSC;
> eth_dev->intr_handle = &dev->intr_handle;
>
> + /* allow ethdev to remove on close */
> + eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
> +
> return eth_dev;
> }
>
> @@ -632,11 +635,12 @@ hn_dev_stop(struct rte_eth_dev *dev)
> }
>
> static void
> -hn_dev_close(struct rte_eth_dev *dev __rte_unused)
> +hn_dev_close(struct rte_eth_dev *dev)
> {
> - PMD_INIT_LOG(DEBUG, "close");
> + PMD_INIT_FUNC_TRACE();
>
> hn_vf_close(dev);
> + hn_dev_free_queues(dev);
> }
>
> static const struct eth_dev_ops hn_eth_dev_ops = {
> diff --git a/drivers/net/netvsc/hn_rxtx.c b/drivers/net/netvsc/hn_rxtx.c
> index 7856f7e6ec48..1bebae7cddd4 100644
> --- a/drivers/net/netvsc/hn_rxtx.c
> +++ b/drivers/net/netvsc/hn_rxtx.c
> @@ -840,12 +840,9 @@ hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
> return error;
> }
>
> -void
> -hn_dev_rx_queue_release(void *arg)
> +static void
> +hn_rx_queue_free(struct hn_rx_queue *rxq, bool keep_primary)
> {
> - struct hn_rx_queue *rxq = arg;
> -
> - PMD_INIT_FUNC_TRACE();
>
> if (!rxq)
> return;
> @@ -857,12 +854,22 @@ hn_dev_rx_queue_release(void *arg)
> hn_vf_rx_queue_release(rxq->hv, rxq->queue_id);
>
> /* Keep primary queue to allow for control operations */
> - if (rxq != rxq->hv->primary) {
> + if (keep_primary && rxq != rxq->hv->primary) {
Isn't something wrong in this logic, when 'keep_primary' is 'false', it will
prevent taking this branch and it will prevent freeing _any_ 'rxq->event_buf' &
'rxq'. There is already a primary queue check after '&&', this flag looks like
working more like enable/disable freeing queues more than being related to primary.
And int he '.dev_close()' path [1] this function is called with 'false', so
queue resources are not cleaned properly.
[1]
hn_dev_close
hn_dev_free_queues
foreach rx_queue
hn_rx_queue_free(rxq, false);
> rte_free(rxq->event_buf);
> rte_free(rxq);
> }
> }
>
> +void
> +hn_dev_rx_queue_release(void *arg)
> +{
> + struct hn_rx_queue *rxq = arg;
> +
> + PMD_INIT_FUNC_TRACE();
> +
> + hn_rx_queue_free(rxq, true);
> +}
> +
> int
> hn_dev_tx_done_cleanup(void *arg, uint32_t free_cnt)
> {
> @@ -1440,3 +1447,23 @@ hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
>
> return nb_rcv;
> }
> +
> +void
> +hn_dev_free_queues(struct rte_eth_dev *dev)
> +{
> + unsigned int i;
> +
> + for (i = 0; i < dev->data->nb_rx_queues; i++) {
> + struct hn_rx_queue *rxq = dev->data->rx_queues[i];
> +
> + hn_rx_queue_free(rxq, false);
> + dev->data->rx_queues[i] = NULL;
> + }
> + dev->data->nb_rx_queues = 0;
> +
> + for (i = 0; i < dev->data->nb_tx_queues; i++) {
> + hn_dev_tx_queue_release(dev->data->tx_queues[i]);
> + dev->data->tx_queues[i] = NULL;
> + }
> + dev->data->nb_tx_queues = 0;
> +}
> diff --git a/drivers/net/netvsc/hn_var.h b/drivers/net/netvsc/hn_var.h
> index 8383f3246ca4..de885d898e6a 100644
> --- a/drivers/net/netvsc/hn_var.h
> +++ b/drivers/net/netvsc/hn_var.h
> @@ -173,6 +173,7 @@ int hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
> const struct rte_eth_rxconf *rx_conf,
> struct rte_mempool *mp);
> void hn_dev_rx_queue_release(void *arg);
> +void hn_dev_free_queues(struct rte_eth_dev *dev);
>
> /* Check if VF is attached */
> static inline bool
>
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [dpdk-dev] [PATCH 2/2] net/netvsc: free all queues on close
2019-04-30 8:11 ` Ferruh Yigit
@ 2019-04-30 8:11 ` Ferruh Yigit
0 siblings, 0 replies; 20+ messages in thread
From: Ferruh Yigit @ 2019-04-30 8:11 UTC (permalink / raw)
To: Stephen Hemminger, dev; +Cc: Stephen Hemminger
On 4/29/2019 9:33 PM, Stephen Hemminger wrote:
> When dev_close is called, the netvsc driver will clean up all
> queues including the primary ring buffer.
>
> Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
> ---
> drivers/net/netvsc/hn_ethdev.c | 8 +++++--
> drivers/net/netvsc/hn_rxtx.c | 39 ++++++++++++++++++++++++++++------
> drivers/net/netvsc/hn_var.h | 1 +
> 3 files changed, 40 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/net/netvsc/hn_ethdev.c b/drivers/net/netvsc/hn_ethdev.c
> index 407ee484935a..553cb06f6e33 100644
> --- a/drivers/net/netvsc/hn_ethdev.c
> +++ b/drivers/net/netvsc/hn_ethdev.c
> @@ -112,6 +112,9 @@ eth_dev_vmbus_allocate(struct rte_vmbus_device *dev, size_t private_data_size)
> eth_dev->data->dev_flags |= RTE_ETH_DEV_INTR_LSC;
> eth_dev->intr_handle = &dev->intr_handle;
>
> + /* allow ethdev to remove on close */
> + eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
> +
> return eth_dev;
> }
>
> @@ -632,11 +635,12 @@ hn_dev_stop(struct rte_eth_dev *dev)
> }
>
> static void
> -hn_dev_close(struct rte_eth_dev *dev __rte_unused)
> +hn_dev_close(struct rte_eth_dev *dev)
> {
> - PMD_INIT_LOG(DEBUG, "close");
> + PMD_INIT_FUNC_TRACE();
>
> hn_vf_close(dev);
> + hn_dev_free_queues(dev);
> }
>
> static const struct eth_dev_ops hn_eth_dev_ops = {
> diff --git a/drivers/net/netvsc/hn_rxtx.c b/drivers/net/netvsc/hn_rxtx.c
> index 7856f7e6ec48..1bebae7cddd4 100644
> --- a/drivers/net/netvsc/hn_rxtx.c
> +++ b/drivers/net/netvsc/hn_rxtx.c
> @@ -840,12 +840,9 @@ hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
> return error;
> }
>
> -void
> -hn_dev_rx_queue_release(void *arg)
> +static void
> +hn_rx_queue_free(struct hn_rx_queue *rxq, bool keep_primary)
> {
> - struct hn_rx_queue *rxq = arg;
> -
> - PMD_INIT_FUNC_TRACE();
>
> if (!rxq)
> return;
> @@ -857,12 +854,22 @@ hn_dev_rx_queue_release(void *arg)
> hn_vf_rx_queue_release(rxq->hv, rxq->queue_id);
>
> /* Keep primary queue to allow for control operations */
> - if (rxq != rxq->hv->primary) {
> + if (keep_primary && rxq != rxq->hv->primary) {
Isn't something wrong in this logic, when 'keep_primary' is 'false', it will
prevent taking this branch and it will prevent freeing _any_ 'rxq->event_buf' &
'rxq'. There is already a primary queue check after '&&', this flag looks like
working more like enable/disable freeing queues more than being related to primary.
And int he '.dev_close()' path [1] this function is called with 'false', so
queue resources are not cleaned properly.
[1]
hn_dev_close
hn_dev_free_queues
foreach rx_queue
hn_rx_queue_free(rxq, false);
> rte_free(rxq->event_buf);
> rte_free(rxq);
> }
> }
>
> +void
> +hn_dev_rx_queue_release(void *arg)
> +{
> + struct hn_rx_queue *rxq = arg;
> +
> + PMD_INIT_FUNC_TRACE();
> +
> + hn_rx_queue_free(rxq, true);
> +}
> +
> int
> hn_dev_tx_done_cleanup(void *arg, uint32_t free_cnt)
> {
> @@ -1440,3 +1447,23 @@ hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
>
> return nb_rcv;
> }
> +
> +void
> +hn_dev_free_queues(struct rte_eth_dev *dev)
> +{
> + unsigned int i;
> +
> + for (i = 0; i < dev->data->nb_rx_queues; i++) {
> + struct hn_rx_queue *rxq = dev->data->rx_queues[i];
> +
> + hn_rx_queue_free(rxq, false);
> + dev->data->rx_queues[i] = NULL;
> + }
> + dev->data->nb_rx_queues = 0;
> +
> + for (i = 0; i < dev->data->nb_tx_queues; i++) {
> + hn_dev_tx_queue_release(dev->data->tx_queues[i]);
> + dev->data->tx_queues[i] = NULL;
> + }
> + dev->data->nb_tx_queues = 0;
> +}
> diff --git a/drivers/net/netvsc/hn_var.h b/drivers/net/netvsc/hn_var.h
> index 8383f3246ca4..de885d898e6a 100644
> --- a/drivers/net/netvsc/hn_var.h
> +++ b/drivers/net/netvsc/hn_var.h
> @@ -173,6 +173,7 @@ int hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
> const struct rte_eth_rxconf *rx_conf,
> struct rte_mempool *mp);
> void hn_dev_rx_queue_release(void *arg);
> +void hn_dev_free_queues(struct rte_eth_dev *dev);
>
> /* Check if VF is attached */
> static inline bool
>
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive
2019-04-29 20:33 ` [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
2019-04-30 8:01 ` Ferruh Yigit
@ 2019-04-30 15:05 ` Stephen Hemminger
2019-04-30 15:05 ` Stephen Hemminger
2 siblings, 1 reply; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-30 15:05 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger
On Mon, 29 Apr 2019 13:33:24 -0700
Stephen Hemminger <stephen@networkplumber.org> wrote:
> Redo the receive logic to set m->port on packets received on VF.
>
> When using VF, still need to check for packets and completions
> arriving on the VMBus path even if application is not doing bursting
> (ie n_rx == 0).
>
> Also, fix comment.
>
> Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Fixes: dc7680e8597c ("net/netvsc: support integrated VF")
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive
2019-04-30 15:05 ` Stephen Hemminger
@ 2019-04-30 15:05 ` Stephen Hemminger
0 siblings, 0 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-30 15:05 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger
On Mon, 29 Apr 2019 13:33:24 -0700
Stephen Hemminger <stephen@networkplumber.org> wrote:
> Redo the receive logic to set m->port on packets received on VF.
>
> When using VF, still need to check for packets and completions
> arriving on the VMBus path even if application is not doing bursting
> (ie n_rx == 0).
>
> Also, fix comment.
>
> Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Fixes: dc7680e8597c ("net/netvsc: support integrated VF")
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05
2019-04-29 20:33 [dpdk-dev] [PATCH 0/2] netvsc pmd fixes for 19.05 Stephen Hemminger
` (2 preceding siblings ...)
2019-04-29 20:33 ` [dpdk-dev] [PATCH 2/2] net/netvsc: free all queues on close Stephen Hemminger
@ 2019-04-30 18:12 ` Stephen Hemminger
2019-04-30 18:12 ` Stephen Hemminger
` (3 more replies)
3 siblings, 4 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-30 18:12 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger
Netvsc needed to reset incoming port as was recently done
for failsafe. Implement the new close semantics.
Stephen Hemminger (2):
net/netvsc: reset mbuf port on VF receive
net/netvsc: free all queues on close
drivers/net/netvsc/hn_ethdev.c | 8 ++-
drivers/net/netvsc/hn_rxtx.c | 93 ++++++++++++++++++++++++----------
drivers/net/netvsc/hn_var.h | 1 +
drivers/net/netvsc/hn_vf.c | 11 +++-
4 files changed, 82 insertions(+), 31 deletions(-)
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Stephen Hemminger
@ 2019-04-30 18:12 ` Stephen Hemminger
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 1/2] net/netvsc: reset mbuf port on VF receive Stephen Hemminger
` (2 subsequent siblings)
3 siblings, 0 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-30 18:12 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger
Netvsc needed to reset incoming port as was recently done
for failsafe. Implement the new close semantics.
Stephen Hemminger (2):
net/netvsc: reset mbuf port on VF receive
net/netvsc: free all queues on close
drivers/net/netvsc/hn_ethdev.c | 8 ++-
drivers/net/netvsc/hn_rxtx.c | 93 ++++++++++++++++++++++++----------
drivers/net/netvsc/hn_var.h | 1 +
drivers/net/netvsc/hn_vf.c | 11 +++-
4 files changed, 82 insertions(+), 31 deletions(-)
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH v2 1/2] net/netvsc: reset mbuf port on VF receive
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Stephen Hemminger
2019-04-30 18:12 ` Stephen Hemminger
@ 2019-04-30 18:12 ` Stephen Hemminger
2019-04-30 18:12 ` Stephen Hemminger
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 2/2] net/netvsc: free all queues on close Stephen Hemminger
2019-05-02 18:24 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Ferruh Yigit
3 siblings, 1 reply; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-30 18:12 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger
From: Stephen Hemminger <sthemmin@microsoft.com>
Redo the receive logic to set m->port on packets received on
VF.
When using VF, still need to check for packets and completions
arriving on the VMBus path even if application is not doing bursting
(ie n_rx == 0).
Also, fix comment.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Fixes: dc7680e8597c ("net/netvsc: support integrated VF")
---
v2 - add fixes
drivers/net/netvsc/hn_rxtx.c | 47 +++++++++++++++++++++---------------
1 file changed, 28 insertions(+), 19 deletions(-)
diff --git a/drivers/net/netvsc/hn_rxtx.c b/drivers/net/netvsc/hn_rxtx.c
index c67e9ae25376..7856f7e6ec48 100644
--- a/drivers/net/netvsc/hn_rxtx.c
+++ b/drivers/net/netvsc/hn_rxtx.c
@@ -1393,6 +1393,24 @@ hn_xmit_pkts(void *ptxq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
return nb_tx;
}
+static uint16_t
+hn_recv_vf(uint16_t vf_port, const struct hn_rx_queue *rxq,
+ struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+ uint16_t i, n;
+
+ if (unlikely(nb_pkts == 0))
+ return 0;
+
+ n = rte_eth_rx_burst(vf_port, rxq->queue_id, rx_pkts, nb_pkts);
+
+ /* relabel the received mbufs */
+ for (i = 0; i < n; i++)
+ rx_pkts[i]->port = rxq->port_id;
+
+ return n;
+}
+
uint16_t
hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
{
@@ -1404,30 +1422,21 @@ hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
if (unlikely(hv->closed))
return 0;
- /* Transmit over VF if present and up */
+ /* Receive from VF if present and up */
vf_dev = hn_get_vf_dev(hv);
- if (vf_dev && vf_dev->data->dev_started) {
- /* Normally, with SR-IOV the ring buffer will be empty */
+ /* Check for new completions */
+ if (likely(rte_ring_count(rxq->rx_ring) < nb_pkts))
hn_process_events(hv, rxq->queue_id, 0);
- /* Get mbufs some bufs off of staging ring */
- nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
- (void **)rx_pkts,
- nb_pkts / 2, NULL);
- /* And rest off of VF */
- nb_rcv += rte_eth_rx_burst(vf_dev->data->port_id,
- rxq->queue_id,
- rx_pkts + nb_rcv, nb_pkts - nb_rcv);
- } else {
- /* If receive ring is not full then get more */
- if (rte_ring_count(rxq->rx_ring) < nb_pkts)
- hn_process_events(hv, rxq->queue_id, 0);
+ /* Always check the vmbus path for multicast and new flows */
+ nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
+ (void **)rx_pkts, nb_pkts, NULL);
- nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
- (void **)rx_pkts,
- nb_pkts, NULL);
- }
+ /* If VF is available, check that as well */
+ if (vf_dev && vf_dev->data->dev_started)
+ nb_rcv += hn_recv_vf(vf_dev->data->port_id, rxq,
+ rx_pkts + nb_rcv, nb_pkts - nb_rcv);
return nb_rcv;
}
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH v2 1/2] net/netvsc: reset mbuf port on VF receive
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 1/2] net/netvsc: reset mbuf port on VF receive Stephen Hemminger
@ 2019-04-30 18:12 ` Stephen Hemminger
0 siblings, 0 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-30 18:12 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger
From: Stephen Hemminger <sthemmin@microsoft.com>
Redo the receive logic to set m->port on packets received on
VF.
When using VF, still need to check for packets and completions
arriving on the VMBus path even if application is not doing bursting
(ie n_rx == 0).
Also, fix comment.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
Fixes: dc7680e8597c ("net/netvsc: support integrated VF")
---
v2 - add fixes
drivers/net/netvsc/hn_rxtx.c | 47 +++++++++++++++++++++---------------
1 file changed, 28 insertions(+), 19 deletions(-)
diff --git a/drivers/net/netvsc/hn_rxtx.c b/drivers/net/netvsc/hn_rxtx.c
index c67e9ae25376..7856f7e6ec48 100644
--- a/drivers/net/netvsc/hn_rxtx.c
+++ b/drivers/net/netvsc/hn_rxtx.c
@@ -1393,6 +1393,24 @@ hn_xmit_pkts(void *ptxq, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
return nb_tx;
}
+static uint16_t
+hn_recv_vf(uint16_t vf_port, const struct hn_rx_queue *rxq,
+ struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+ uint16_t i, n;
+
+ if (unlikely(nb_pkts == 0))
+ return 0;
+
+ n = rte_eth_rx_burst(vf_port, rxq->queue_id, rx_pkts, nb_pkts);
+
+ /* relabel the received mbufs */
+ for (i = 0; i < n; i++)
+ rx_pkts[i]->port = rxq->port_id;
+
+ return n;
+}
+
uint16_t
hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
{
@@ -1404,30 +1422,21 @@ hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
if (unlikely(hv->closed))
return 0;
- /* Transmit over VF if present and up */
+ /* Receive from VF if present and up */
vf_dev = hn_get_vf_dev(hv);
- if (vf_dev && vf_dev->data->dev_started) {
- /* Normally, with SR-IOV the ring buffer will be empty */
+ /* Check for new completions */
+ if (likely(rte_ring_count(rxq->rx_ring) < nb_pkts))
hn_process_events(hv, rxq->queue_id, 0);
- /* Get mbufs some bufs off of staging ring */
- nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
- (void **)rx_pkts,
- nb_pkts / 2, NULL);
- /* And rest off of VF */
- nb_rcv += rte_eth_rx_burst(vf_dev->data->port_id,
- rxq->queue_id,
- rx_pkts + nb_rcv, nb_pkts - nb_rcv);
- } else {
- /* If receive ring is not full then get more */
- if (rte_ring_count(rxq->rx_ring) < nb_pkts)
- hn_process_events(hv, rxq->queue_id, 0);
+ /* Always check the vmbus path for multicast and new flows */
+ nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
+ (void **)rx_pkts, nb_pkts, NULL);
- nb_rcv = rte_ring_sc_dequeue_burst(rxq->rx_ring,
- (void **)rx_pkts,
- nb_pkts, NULL);
- }
+ /* If VF is available, check that as well */
+ if (vf_dev && vf_dev->data->dev_started)
+ nb_rcv += hn_recv_vf(vf_dev->data->port_id, rxq,
+ rx_pkts + nb_rcv, nb_pkts - nb_rcv);
return nb_rcv;
}
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH v2 2/2] net/netvsc: free all queues on close
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Stephen Hemminger
2019-04-30 18:12 ` Stephen Hemminger
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 1/2] net/netvsc: reset mbuf port on VF receive Stephen Hemminger
@ 2019-04-30 18:12 ` Stephen Hemminger
2019-04-30 18:12 ` Stephen Hemminger
2019-05-02 18:24 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Ferruh Yigit
3 siblings, 1 reply; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-30 18:12 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger
From: Stephen Hemminger <sthemmin@microsoft.com>
When dev_close is called, the netvsc driver will clean up all
queues including the primary ring buffer.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
---
v2 - fix logic on close of primary
avoid double free if close is called twice
drivers/net/netvsc/hn_ethdev.c | 8 ++++--
drivers/net/netvsc/hn_rxtx.c | 46 +++++++++++++++++++++++++++-------
drivers/net/netvsc/hn_var.h | 1 +
drivers/net/netvsc/hn_vf.c | 11 +++++++-
4 files changed, 54 insertions(+), 12 deletions(-)
diff --git a/drivers/net/netvsc/hn_ethdev.c b/drivers/net/netvsc/hn_ethdev.c
index 407ee484935a..553cb06f6e33 100644
--- a/drivers/net/netvsc/hn_ethdev.c
+++ b/drivers/net/netvsc/hn_ethdev.c
@@ -112,6 +112,9 @@ eth_dev_vmbus_allocate(struct rte_vmbus_device *dev, size_t private_data_size)
eth_dev->data->dev_flags |= RTE_ETH_DEV_INTR_LSC;
eth_dev->intr_handle = &dev->intr_handle;
+ /* allow ethdev to remove on close */
+ eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
+
return eth_dev;
}
@@ -632,11 +635,12 @@ hn_dev_stop(struct rte_eth_dev *dev)
}
static void
-hn_dev_close(struct rte_eth_dev *dev __rte_unused)
+hn_dev_close(struct rte_eth_dev *dev)
{
- PMD_INIT_LOG(DEBUG, "close");
+ PMD_INIT_FUNC_TRACE();
hn_vf_close(dev);
+ hn_dev_free_queues(dev);
}
static const struct eth_dev_ops hn_eth_dev_ops = {
diff --git a/drivers/net/netvsc/hn_rxtx.c b/drivers/net/netvsc/hn_rxtx.c
index 7856f7e6ec48..7d7b557785fa 100644
--- a/drivers/net/netvsc/hn_rxtx.c
+++ b/drivers/net/netvsc/hn_rxtx.c
@@ -840,12 +840,9 @@ hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
return error;
}
-void
-hn_dev_rx_queue_release(void *arg)
+static void
+hn_rx_queue_free(struct hn_rx_queue *rxq, bool keep_primary)
{
- struct hn_rx_queue *rxq = arg;
-
- PMD_INIT_FUNC_TRACE();
if (!rxq)
return;
@@ -857,10 +854,21 @@ hn_dev_rx_queue_release(void *arg)
hn_vf_rx_queue_release(rxq->hv, rxq->queue_id);
/* Keep primary queue to allow for control operations */
- if (rxq != rxq->hv->primary) {
- rte_free(rxq->event_buf);
- rte_free(rxq);
- }
+ if (keep_primary && rxq == rxq->hv->primary)
+ return;
+
+ rte_free(rxq->event_buf);
+ rte_free(rxq);
+}
+
+void
+hn_dev_rx_queue_release(void *arg)
+{
+ struct hn_rx_queue *rxq = arg;
+
+ PMD_INIT_FUNC_TRACE();
+
+ hn_rx_queue_free(rxq, true);
}
int
@@ -1440,3 +1448,23 @@ hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
return nb_rcv;
}
+
+void
+hn_dev_free_queues(struct rte_eth_dev *dev)
+{
+ unsigned int i;
+
+ for (i = 0; i < dev->data->nb_rx_queues; i++) {
+ struct hn_rx_queue *rxq = dev->data->rx_queues[i];
+
+ hn_rx_queue_free(rxq, false);
+ dev->data->rx_queues[i] = NULL;
+ }
+ dev->data->nb_rx_queues = 0;
+
+ for (i = 0; i < dev->data->nb_tx_queues; i++) {
+ hn_dev_tx_queue_release(dev->data->tx_queues[i]);
+ dev->data->tx_queues[i] = NULL;
+ }
+ dev->data->nb_tx_queues = 0;
+}
diff --git a/drivers/net/netvsc/hn_var.h b/drivers/net/netvsc/hn_var.h
index 8383f3246ca4..de885d898e6a 100644
--- a/drivers/net/netvsc/hn_var.h
+++ b/drivers/net/netvsc/hn_var.h
@@ -173,6 +173,7 @@ int hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
const struct rte_eth_rxconf *rx_conf,
struct rte_mempool *mp);
void hn_dev_rx_queue_release(void *arg);
+void hn_dev_free_queues(struct rte_eth_dev *dev);
/* Check if VF is attached */
static inline bool
diff --git a/drivers/net/netvsc/hn_vf.c b/drivers/net/netvsc/hn_vf.c
index 883272ff498c..b980bb8a4d8a 100644
--- a/drivers/net/netvsc/hn_vf.c
+++ b/drivers/net/netvsc/hn_vf.c
@@ -362,7 +362,16 @@ void hn_vf_reset(struct rte_eth_dev *dev)
void hn_vf_close(struct rte_eth_dev *dev)
{
- VF_ETHDEV_FUNC(dev, rte_eth_dev_close);
+ struct hn_data *hv = dev->data->dev_private;
+ uint16_t vf_port;
+
+ rte_spinlock_lock(&hv->vf_lock);
+ vf_port = hv->vf_port;
+ if (vf_port != HN_INVALID_PORT)
+ rte_eth_dev_close(vf_port);
+
+ hv->vf_port = HN_INVALID_PORT;
+ rte_spinlock_unlock(&hv->vf_lock);
}
void hn_vf_stats_reset(struct rte_eth_dev *dev)
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* [dpdk-dev] [PATCH v2 2/2] net/netvsc: free all queues on close
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 2/2] net/netvsc: free all queues on close Stephen Hemminger
@ 2019-04-30 18:12 ` Stephen Hemminger
0 siblings, 0 replies; 20+ messages in thread
From: Stephen Hemminger @ 2019-04-30 18:12 UTC (permalink / raw)
To: dev; +Cc: Stephen Hemminger
From: Stephen Hemminger <sthemmin@microsoft.com>
When dev_close is called, the netvsc driver will clean up all
queues including the primary ring buffer.
Signed-off-by: Stephen Hemminger <sthemmin@microsoft.com>
---
v2 - fix logic on close of primary
avoid double free if close is called twice
drivers/net/netvsc/hn_ethdev.c | 8 ++++--
drivers/net/netvsc/hn_rxtx.c | 46 +++++++++++++++++++++++++++-------
drivers/net/netvsc/hn_var.h | 1 +
drivers/net/netvsc/hn_vf.c | 11 +++++++-
4 files changed, 54 insertions(+), 12 deletions(-)
diff --git a/drivers/net/netvsc/hn_ethdev.c b/drivers/net/netvsc/hn_ethdev.c
index 407ee484935a..553cb06f6e33 100644
--- a/drivers/net/netvsc/hn_ethdev.c
+++ b/drivers/net/netvsc/hn_ethdev.c
@@ -112,6 +112,9 @@ eth_dev_vmbus_allocate(struct rte_vmbus_device *dev, size_t private_data_size)
eth_dev->data->dev_flags |= RTE_ETH_DEV_INTR_LSC;
eth_dev->intr_handle = &dev->intr_handle;
+ /* allow ethdev to remove on close */
+ eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
+
return eth_dev;
}
@@ -632,11 +635,12 @@ hn_dev_stop(struct rte_eth_dev *dev)
}
static void
-hn_dev_close(struct rte_eth_dev *dev __rte_unused)
+hn_dev_close(struct rte_eth_dev *dev)
{
- PMD_INIT_LOG(DEBUG, "close");
+ PMD_INIT_FUNC_TRACE();
hn_vf_close(dev);
+ hn_dev_free_queues(dev);
}
static const struct eth_dev_ops hn_eth_dev_ops = {
diff --git a/drivers/net/netvsc/hn_rxtx.c b/drivers/net/netvsc/hn_rxtx.c
index 7856f7e6ec48..7d7b557785fa 100644
--- a/drivers/net/netvsc/hn_rxtx.c
+++ b/drivers/net/netvsc/hn_rxtx.c
@@ -840,12 +840,9 @@ hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
return error;
}
-void
-hn_dev_rx_queue_release(void *arg)
+static void
+hn_rx_queue_free(struct hn_rx_queue *rxq, bool keep_primary)
{
- struct hn_rx_queue *rxq = arg;
-
- PMD_INIT_FUNC_TRACE();
if (!rxq)
return;
@@ -857,10 +854,21 @@ hn_dev_rx_queue_release(void *arg)
hn_vf_rx_queue_release(rxq->hv, rxq->queue_id);
/* Keep primary queue to allow for control operations */
- if (rxq != rxq->hv->primary) {
- rte_free(rxq->event_buf);
- rte_free(rxq);
- }
+ if (keep_primary && rxq == rxq->hv->primary)
+ return;
+
+ rte_free(rxq->event_buf);
+ rte_free(rxq);
+}
+
+void
+hn_dev_rx_queue_release(void *arg)
+{
+ struct hn_rx_queue *rxq = arg;
+
+ PMD_INIT_FUNC_TRACE();
+
+ hn_rx_queue_free(rxq, true);
}
int
@@ -1440,3 +1448,23 @@ hn_recv_pkts(void *prxq, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
return nb_rcv;
}
+
+void
+hn_dev_free_queues(struct rte_eth_dev *dev)
+{
+ unsigned int i;
+
+ for (i = 0; i < dev->data->nb_rx_queues; i++) {
+ struct hn_rx_queue *rxq = dev->data->rx_queues[i];
+
+ hn_rx_queue_free(rxq, false);
+ dev->data->rx_queues[i] = NULL;
+ }
+ dev->data->nb_rx_queues = 0;
+
+ for (i = 0; i < dev->data->nb_tx_queues; i++) {
+ hn_dev_tx_queue_release(dev->data->tx_queues[i]);
+ dev->data->tx_queues[i] = NULL;
+ }
+ dev->data->nb_tx_queues = 0;
+}
diff --git a/drivers/net/netvsc/hn_var.h b/drivers/net/netvsc/hn_var.h
index 8383f3246ca4..de885d898e6a 100644
--- a/drivers/net/netvsc/hn_var.h
+++ b/drivers/net/netvsc/hn_var.h
@@ -173,6 +173,7 @@ int hn_dev_rx_queue_setup(struct rte_eth_dev *dev,
const struct rte_eth_rxconf *rx_conf,
struct rte_mempool *mp);
void hn_dev_rx_queue_release(void *arg);
+void hn_dev_free_queues(struct rte_eth_dev *dev);
/* Check if VF is attached */
static inline bool
diff --git a/drivers/net/netvsc/hn_vf.c b/drivers/net/netvsc/hn_vf.c
index 883272ff498c..b980bb8a4d8a 100644
--- a/drivers/net/netvsc/hn_vf.c
+++ b/drivers/net/netvsc/hn_vf.c
@@ -362,7 +362,16 @@ void hn_vf_reset(struct rte_eth_dev *dev)
void hn_vf_close(struct rte_eth_dev *dev)
{
- VF_ETHDEV_FUNC(dev, rte_eth_dev_close);
+ struct hn_data *hv = dev->data->dev_private;
+ uint16_t vf_port;
+
+ rte_spinlock_lock(&hv->vf_lock);
+ vf_port = hv->vf_port;
+ if (vf_port != HN_INVALID_PORT)
+ rte_eth_dev_close(vf_port);
+
+ hv->vf_port = HN_INVALID_PORT;
+ rte_spinlock_unlock(&hv->vf_lock);
}
void hn_vf_stats_reset(struct rte_eth_dev *dev)
--
2.20.1
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Stephen Hemminger
` (2 preceding siblings ...)
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 2/2] net/netvsc: free all queues on close Stephen Hemminger
@ 2019-05-02 18:24 ` Ferruh Yigit
2019-05-02 18:24 ` Ferruh Yigit
3 siblings, 1 reply; 20+ messages in thread
From: Ferruh Yigit @ 2019-05-02 18:24 UTC (permalink / raw)
To: Stephen Hemminger, dev
On 4/30/2019 7:12 PM, Stephen Hemminger wrote:
> Netvsc needed to reset incoming port as was recently done
> for failsafe. Implement the new close semantics.
>
> Stephen Hemminger (2):
> net/netvsc: reset mbuf port on VF receive
> net/netvsc: free all queues on close
Series applied to dpdk-next-net/master, thanks.
^ permalink raw reply [flat|nested] 20+ messages in thread
* Re: [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05
2019-05-02 18:24 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Ferruh Yigit
@ 2019-05-02 18:24 ` Ferruh Yigit
0 siblings, 0 replies; 20+ messages in thread
From: Ferruh Yigit @ 2019-05-02 18:24 UTC (permalink / raw)
To: Stephen Hemminger, dev
On 4/30/2019 7:12 PM, Stephen Hemminger wrote:
> Netvsc needed to reset incoming port as was recently done
> for failsafe. Implement the new close semantics.
>
> Stephen Hemminger (2):
> net/netvsc: reset mbuf port on VF receive
> net/netvsc: free all queues on close
Series applied to dpdk-next-net/master, thanks.
^ permalink raw reply [flat|nested] 20+ messages in thread
end of thread, other threads:[~2019-05-02 18:24 UTC | newest]
Thread overview: 20+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-04-29 20:33 [dpdk-dev] [PATCH 0/2] netvsc pmd fixes for 19.05 Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
2019-04-29 20:33 ` [dpdk-dev] [PATCH 1/2] net/netvsc: reset mbuf port on VF receive Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
2019-04-30 8:01 ` Ferruh Yigit
2019-04-30 8:01 ` Ferruh Yigit
2019-04-30 15:05 ` Stephen Hemminger
2019-04-30 15:05 ` Stephen Hemminger
2019-04-29 20:33 ` [dpdk-dev] [PATCH 2/2] net/netvsc: free all queues on close Stephen Hemminger
2019-04-29 20:33 ` Stephen Hemminger
2019-04-30 8:11 ` Ferruh Yigit
2019-04-30 8:11 ` Ferruh Yigit
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Stephen Hemminger
2019-04-30 18:12 ` Stephen Hemminger
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 1/2] net/netvsc: reset mbuf port on VF receive Stephen Hemminger
2019-04-30 18:12 ` Stephen Hemminger
2019-04-30 18:12 ` [dpdk-dev] [PATCH v2 2/2] net/netvsc: free all queues on close Stephen Hemminger
2019-04-30 18:12 ` Stephen Hemminger
2019-05-02 18:24 ` [dpdk-dev] [PATCH v2 0/2] netvsc PMD fixes for 19.05 Ferruh Yigit
2019-05-02 18:24 ` Ferruh Yigit
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).