[dpdk-dev] [PATCH 0/2] net/netvsc: fix secondary process issues

[dpdk-dev] [PATCH 0/2] net/netvsc: fix secondary process issues

[Stephen Hemminger]

First patch is a bug fix for crash in info_get in secondary process.
Second one is modification to prevent similar issues in future.

Stephen Hemminger (2):
  net/netvsc: fix crash in secondary process
  net/netvsc: eliminate numa specific allocation for control messages

 drivers/net/netvsc/hn_ethdev.c | 17 +++++++++--------
 drivers/net/netvsc/hn_rndis.c  | 11 +++++------
 2 files changed, 14 insertions(+), 14 deletions(-)

-- 
2.20.1

[dpdk-dev] [PATCH 1/2] net/netvsc: fix crash in secondary process

[Stephen Hemminger]

The secondary process would crash trying to get the offload
capability of the device. Since the device capabilities were
already set by the primary process this was unnecessary.

Fixes: 4e9c73e96e83 ("net/netvsc: add Hyper-V network device")
Cc: stable@dpdk.org
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/netvsc/hn_ethdev.c | 17 +++++++++--------
 1 file changed, 9 insertions(+), 8 deletions(-)

diff --git a/drivers/net/netvsc/hn_ethdev.c b/drivers/net/netvsc/hn_ethdev.c
index 159116f7f691..c79f924379fe 100644
--- a/drivers/net/netvsc/hn_ethdev.c
+++ b/drivers/net/netvsc/hn_ethdev.c
@@ -257,15 +257,16 @@ static int hn_dev_info_get(struct rte_eth_dev *dev,
 	dev_info->max_rx_queues = hv->max_queues;
 	dev_info->max_tx_queues = hv->max_queues;
 
-	rc = hn_rndis_get_offload(hv, dev_info);
-	if (rc != 0)
-		return rc;
+	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+		return 0;
 
-	rc = hn_vf_info_get(hv, dev_info);
+	/* fills in rx and tx offload capability */
+	rc = hn_rndis_get_offload(hv, dev_info);
 	if (rc != 0)
 		return rc;
 
-	return 0;
+	/* merges the offload and queues of vf */
+	return hn_vf_info_get(hv, dev_info);
 }
 
 static int hn_rss_reta_update(struct rte_eth_dev *dev,
@@ -929,6 +930,9 @@ eth_hn_dev_init(struct rte_eth_dev *eth_dev)
 	eth_dev->tx_pkt_burst = &hn_xmit_pkts;
 	eth_dev->rx_pkt_burst = &hn_recv_pkts;
 
+	/* Since Hyper-V only supports one MAC address, just use local data */
+	eth_dev->data->mac_addrs = &hv->mac_addr;
+
 	/*
 	 * for secondary processes, we don't initialize any further as primary
 	 * has already done this work.
@@ -936,9 +940,6 @@ eth_hn_dev_init(struct rte_eth_dev *eth_dev)
 	if (rte_eal_process_type() != RTE_PROC_PRIMARY)
 		return 0;
 
-	/* Since Hyper-V only supports one MAC address, just use local data */
-	eth_dev->data->mac_addrs = &hv->mac_addr;
-
 	hv->vmbus = vmbus;
 	hv->rxbuf_res = &vmbus->resource[HV_RECV_BUF_MAP];
 	hv->chim_res  = &vmbus->resource[HV_SEND_BUF_MAP];
-- 
2.20.1

[dpdk-dev] [PATCH 2/2] net/netvsc: eliminate numa specific allocation for control messages

[Stephen Hemminger]

The control messages do not need NUMA specific allocation.
Numa node is not set anyway in most kernels anyway.

Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 drivers/net/netvsc/hn_rndis.c | 11 +++++------
 1 file changed, 5 insertions(+), 6 deletions(-)

diff --git a/drivers/net/netvsc/hn_rndis.c b/drivers/net/netvsc/hn_rndis.c
index 2b4714042e04..7947ca2331e2 100644
--- a/drivers/net/netvsc/hn_rndis.c
+++ b/drivers/net/netvsc/hn_rndis.c
@@ -60,10 +60,9 @@ hn_rndis_rid(struct hn_data *hv)
 	return rid;
 }
 
-static void *hn_rndis_alloc(struct hn_data *hv, size_t size)
+static void *hn_rndis_alloc(size_t size)
 {
-	return rte_zmalloc_socket("RNDIS", size, PAGE_SIZE,
-				 hv->vmbus->device.numa_node);
+	return rte_zmalloc("RNDIS", size, PAGE_SIZE);
 }
 
 #ifdef RTE_LIBRTE_NETVSC_DEBUG_DUMP
@@ -442,7 +441,7 @@ hn_rndis_query(struct hn_data *hv, uint32_t oid,
 	uint32_t rid;
 
 	reqlen = sizeof(*req) + idlen;
-	req = hn_rndis_alloc(hv, reqlen);
+	req = hn_rndis_alloc(reqlen);
 	if (req == NULL)
 		return -ENOMEM;
 
@@ -517,7 +516,7 @@ hn_rndis_halt(struct hn_data *hv)
 {
 	struct rndis_halt_req *halt;
 
-	halt = hn_rndis_alloc(hv, sizeof(*halt));
+	halt = hn_rndis_alloc(sizeof(*halt));
 	if (halt == NULL)
 		return -ENOMEM;
 
@@ -1004,7 +1003,7 @@ static int hn_rndis_init(struct hn_data *hv)
 	uint32_t comp_len, rid;
 	int error;
 
-	req = hn_rndis_alloc(hv, sizeof(*req));
+	req = hn_rndis_alloc(sizeof(*req));
 	if (!req) {
 		PMD_DRV_LOG(ERR, "no memory for RNDIS init");
 		return -ENXIO;
-- 
2.20.1

Re: [dpdk-dev] [PATCH 0/2] net/netvsc: fix secondary process issues

[Ferruh Yigit]

On 1/13/2020 7:12 PM, Stephen Hemminger wrote:
> First patch is a bug fix for crash in info_get in secondary process.
> Second one is modification to prevent similar issues in future.
> 
> Stephen Hemminger (2):
>   net/netvsc: fix crash in secondary process
>   net/netvsc: eliminate numa specific allocation for control messages
> 

Series applied to dpdk-next-net/master, thanks.