Skip to contentSkip to conversation

net/null: fix secondary burst function selection

PATCH In review Ferruh Yigit 1 participant 8 messages Mar 2, 2020 – Apr 10, 2020 Network drivers

lore
Share & tools
lore
#0 lore raw

PATCH 1/6 net/null: fix secondary burst function selection

Secondary process uses the primary process device and while setting the Rx/Tx functions it uses the device arguments from the secondary process instead of the primary ones.

This may cause primary and secondary process use different Rx/Tx functions unintentionally.

Fixes bccc77a6a74a ("net/null: fix multi-process Rx and Tx")

---
Cc: [email protected]

---
 drivers/net/null/rte_eth_null.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/net/null/rte_eth_null.c +3 -1
index 025b73acb..87a29b853 100644
@@ -584,6 +584,7 @@ rte_pmd_null_probe(struct rte_vdev_device *dev)
 	PMD_LOG(INFO, "Initializing pmd_null for %s", name);
 
 	if (rte_eal_process_type() == RTE_PROC_SECONDARY) {
+		struct pmd_internals *internals;
 		eth_dev = rte_eth_dev_attach_secondary(name);
 		if (!eth_dev) {
 			PMD_LOG(ERR, "Failed to probe %s", name);
@@ -592,7 +593,8 @@ rte_pmd_null_probe(struct rte_vdev_device *dev)
 		/* TODO: request info from primary to set up Rx and Tx */
 		eth_dev->dev_ops = &ops;
 		eth_dev->device = &dev->device;
-		if (packet_copy) {
+		internals = eth_dev->data->dev_private;
+		if (internals->packet_copy) {
 			eth_dev->rx_pkt_burst = eth_null_copy_rx;
 			eth_dev->tx_pkt_burst = eth_null_copy_tx;
 		} else {
Signature · 1 line
2.24.1
#1 lore raw

PATCH 2/6 doc: add net null PMD guide

Net null PMD was missing documentation, adding it.

---
 doc/guides/nics/index.rst |  1 +
 doc/guides/nics/null.rst  | 39 +++++++++++++++++++++++++++++++++++++++
 2 files changed, 40 insertions(+)
 create mode 100644 doc/guides/nics/null.rst
doc/guides/nics/index.rst +1 -0
index 6d88028ef..2b78fcfee 100644
@@ -46,6 +46,7 @@ Network Interface Controller Drivers
     netvsc
     nfb
     nfp
+    null
     octeontx
     octeontx2
     pfe
doc/guides/nics/null.rst new file +39 -0
new file mode 100644
index 000000000..405edf07b
@@ -0,0 +1,39 @@
+..  SPDX-License-Identifier: BSD-3-Clause
+    Copyright(c) 2020 Intel Corporation.
+
+NULL Poll Mode Driver
+=====================
+
+NULL PMD is a simple virtual driver mainly for testing. It always returns success for all packets for Rx/Tx.
+
+On Rx it returns requested number of empty packets (all zero). On Tx it just frees all sent packets.
+
+
+Usage
+-----
+
+.. code-block:: console
+
+   $RTE_TARGET/app/testpmd -l 0-3 -n 4 --vdev net_null0 --vdev net_null1 -- -i
+
+
+Runtime Config Options
+----------------------
+
+- ``copy`` [optional, default disabled]
+
+ It copies data of the packet before Rx/Tx. For Rx it uses another empty dummy mbuf for this.
+
+.. code-block:: console
+
+   $RTE_TARGET/app/testpmd -l 0-3 -n 4 --vdev "net_null0,copy=1" -- -i
+
+- ``size`` [optional, default=64 bytes]
+
+ Custom packet length value to use.r
+ If ``copy`` is enabled, this is the length of copy operation.
+
+.. code-block:: console
+
+   $RTE_TARGET/app/testpmd -l 0-3 -n 4 --vdev "net_null0,size=256" -- -i
+
Signature · 1 line
2.24.1
#2 lore raw

PATCH 3/6 net/null: remove redundant check

In reply to Ferruh Yigit

There is no need to check if the argument exist or not, rte_kvargs_process returns success if the argument is not provided at all.

Fixes c743e50c475f ("null: new poll mode driver")

---
Cc: [email protected]

---
 drivers/net/null/rte_eth_null.c | 29 ++++++++++++-----------------
 1 file changed, 12 insertions(+), 17 deletions(-)
drivers/net/null/rte_eth_null.c +12 -17
index 87a29b853..beedd5f4b 100644
@@ -610,23 +610,18 @@ rte_pmd_null_probe(struct rte_vdev_device *dev)
 		if (kvlist == NULL)
 			return -1;
 
-		if (rte_kvargs_count(kvlist, ETH_NULL_PACKET_SIZE_ARG) == 1) {
-
-			ret = rte_kvargs_process(kvlist,
-					ETH_NULL_PACKET_SIZE_ARG,
-					&get_packet_size_arg, &packet_size);
-			if (ret < 0)
-				goto free_kvlist;
-		}
-
-		if (rte_kvargs_count(kvlist, ETH_NULL_PACKET_COPY_ARG) == 1) {
-
-			ret = rte_kvargs_process(kvlist,
-					ETH_NULL_PACKET_COPY_ARG,
-					&get_packet_copy_arg, &packet_copy);
-			if (ret < 0)
-				goto free_kvlist;
-		}
+		ret = rte_kvargs_process(kvlist,
+				ETH_NULL_PACKET_SIZE_ARG,
+				&get_packet_size_arg, &packet_size);
+		if (ret < 0)
+			goto free_kvlist;
+
+
+		ret = rte_kvargs_process(kvlist,
+				ETH_NULL_PACKET_COPY_ARG,
+				&get_packet_copy_arg, &packet_copy);
+		if (ret < 0)
+			goto free_kvlist;
 	}
 
 	PMD_LOG(INFO, "Configure pmd_null: packet size is %d, "
Signature · 1 line
2.24.1
#3 lore raw

PATCH 4/6 net/null: prefer unsigned int

In reply to Ferruh Yigit

Prefer 'unsigned int' storage type keyword against 'unsigned', this also silence the checkpatch warnings.

---
 drivers/net/null/rte_eth_null.c | 46 ++++++++++++++++-----------------
 1 file changed, 23 insertions(+), 23 deletions(-)
drivers/net/null/rte_eth_null.c +23 -23
index beedd5f4b..7fa3b9e45 100644
@@ -15,8 +15,8 @@
 #define ETH_NULL_PACKET_SIZE_ARG	"size"
 #define ETH_NULL_PACKET_COPY_ARG	"copy"
 
-static unsigned default_packet_size = 64;
-static unsigned default_packet_copy;
+static unsigned int default_packet_size = 64;
+static unsigned int default_packet_copy;
 
 static const char *valid_arguments[] = {
 	ETH_NULL_PACKET_SIZE_ARG,
@@ -37,8 +37,8 @@ struct null_queue {
 };
 
 struct pmd_internals {
-	unsigned packet_size;
-	unsigned packet_copy;
+	unsigned int packet_size;
+	unsigned int packet_copy;
 	uint16_t port_id;
 
 	struct null_queue rx_null_queues[RTE_MAX_QUEUES_PER_PORT];
@@ -74,7 +74,7 @@ eth_null_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
 {
 	int i;
 	struct null_queue *h = q;
-	unsigned packet_size;
+	unsigned int packet_size;
 
 	if ((q == NULL) || (bufs == NULL))
 		return 0;
@@ -99,7 +99,7 @@ eth_null_copy_rx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
 {
 	int i;
 	struct null_queue *h = q;
-	unsigned packet_size;
+	unsigned int packet_size;
 
 	if ((q == NULL) || (bufs == NULL))
 		return 0;
@@ -143,7 +143,7 @@ eth_null_copy_tx(void *q, struct rte_mbuf **bufs, uint16_t nb_bufs)
 {
 	int i;
 	struct null_queue *h = q;
-	unsigned packet_size;
+	unsigned int packet_size;
 
 	if ((q == NULL) || (bufs == NULL))
 		return 0;
@@ -194,7 +194,7 @@ eth_rx_queue_setup(struct rte_eth_dev *dev, uint16_t rx_queue_id,
 {
 	struct rte_mbuf *dummy_packet;
 	struct pmd_internals *internals;
-	unsigned packet_size;
+	unsigned int packet_size;
 
 	if ((dev == NULL) || (mb_pool == NULL))
 		return -EINVAL;
@@ -228,7 +228,7 @@ eth_tx_queue_setup(struct rte_eth_dev *dev, uint16_t tx_queue_id,
 {
 	struct rte_mbuf *dummy_packet;
 	struct pmd_internals *internals;
-	unsigned packet_size;
+	unsigned int packet_size;
 
 	if (dev == NULL)
 		return -EINVAL;
@@ -283,7 +283,7 @@ eth_dev_info(struct rte_eth_dev *dev,
 static int
 eth_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *igb_stats)
 {
-	unsigned i, num_stats;
+	unsigned int i, num_stats;
 	unsigned long rx_total = 0, tx_total = 0;
 	const struct pmd_internals *internal;
 
@@ -291,7 +291,7 @@ eth_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *igb_stats)
 		return -EINVAL;
 
 	internal = dev->data->dev_private;
-	num_stats = RTE_MIN((unsigned)RTE_ETHDEV_QUEUE_STAT_CNTRS,
+	num_stats = RTE_MIN((unsigned int)RTE_ETHDEV_QUEUE_STAT_CNTRS,
 			RTE_MIN(dev->data->nb_rx_queues,
 				RTE_DIM(internal->rx_null_queues)));
 	for (i = 0; i < num_stats; i++) {
@@ -300,7 +300,7 @@ eth_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *igb_stats)
 		rx_total += igb_stats->q_ipackets[i];
 	}
 
-	num_stats = RTE_MIN((unsigned)RTE_ETHDEV_QUEUE_STAT_CNTRS,
+	num_stats = RTE_MIN((unsigned int)RTE_ETHDEV_QUEUE_STAT_CNTRS,
 			RTE_MIN(dev->data->nb_tx_queues,
 				RTE_DIM(internal->tx_null_queues)));
 	for (i = 0; i < num_stats; i++) {
@@ -318,7 +318,7 @@ eth_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *igb_stats)
 static int
 eth_stats_reset(struct rte_eth_dev *dev)
 {
-	unsigned i;
+	unsigned int i;
 	struct pmd_internals *internal;
 
 	if (dev == NULL)
@@ -463,11 +463,11 @@ static const struct eth_dev_ops ops = {
 
 static int
 eth_dev_null_create(struct rte_vdev_device *dev,
-		unsigned packet_size,
-		unsigned packet_copy)
+		unsigned int packet_size,
+		unsigned int packet_copy)
 {
-	const unsigned nb_rx_queues = 1;
-	const unsigned nb_tx_queues = 1;
+	const unsigned int nb_rx_queues = 1;
+	const unsigned int nb_tx_queues = 1;
 	struct rte_eth_dev_data *data;
 	struct pmd_internals *internals = NULL;
 	struct rte_eth_dev *eth_dev = NULL;
@@ -537,12 +537,12 @@ get_packet_size_arg(const char *key __rte_unused,
 		const char *value, void *extra_args)
 {
 	const char *a = value;
-	unsigned *packet_size = extra_args;
+	unsigned int *packet_size = extra_args;
 
 	if ((value == NULL) || (extra_args == NULL))
 		return -EINVAL;
 
-	*packet_size = (unsigned)strtoul(a, NULL, 0);
+	*packet_size = (unsigned int)strtoul(a, NULL, 0);
 	if (*packet_size == UINT_MAX)
 		return -1;
 
@@ -554,12 +554,12 @@ get_packet_copy_arg(const char *key __rte_unused,
 		const char *value, void *extra_args)
 {
 	const char *a = value;
-	unsigned *packet_copy = extra_args;
+	unsigned int *packet_copy = extra_args;
 
 	if ((value == NULL) || (extra_args == NULL))
 		return -EINVAL;
 
-	*packet_copy = (unsigned)strtoul(a, NULL, 0);
+	*packet_copy = (unsigned int)strtoul(a, NULL, 0);
 	if (*packet_copy == UINT_MAX)
 		return -1;
 
@@ -570,8 +570,8 @@ static int
 rte_pmd_null_probe(struct rte_vdev_device *dev)
 {
 	const char *name, *params;
-	unsigned packet_size = default_packet_size;
-	unsigned packet_copy = default_packet_copy;
+	unsigned int packet_size = default_packet_size;
+	unsigned int packet_copy = default_packet_copy;
 	struct rte_kvargs *kvlist = NULL;
 	struct rte_eth_dev *eth_dev;
 	int ret;
Signature · 1 line
2.24.1
#4 lore raw

PATCH 5/6 net/null: group device arguments

In reply to Ferruh Yigit

Group device argument to the struct, to increase code readability.

---
 drivers/net/null/rte_eth_null.c | 31 ++++++++++++++++++-------------
 1 file changed, 18 insertions(+), 13 deletions(-)
drivers/net/null/rte_eth_null.c +18 -13
index 7fa3b9e45..2c08ebf8c 100644
@@ -36,6 +36,11 @@ struct null_queue {
 	rte_atomic64_t tx_pkts;
 };
 
+struct pmd_options {
+	unsigned int packet_copy;
+	unsigned int packet_size;
+};
+
 struct pmd_internals {
 	unsigned int packet_size;
 	unsigned int packet_copy;
@@ -462,9 +467,7 @@ static const struct eth_dev_ops ops = {
 };
 
 static int
-eth_dev_null_create(struct rte_vdev_device *dev,
-		unsigned int packet_size,
-		unsigned int packet_copy)
+eth_dev_null_create(struct rte_vdev_device *dev, struct pmd_options *args)
 {
 	const unsigned int nb_rx_queues = 1;
 	const unsigned int nb_tx_queues = 1;
@@ -499,8 +502,8 @@ eth_dev_null_create(struct rte_vdev_device *dev,
 	 * so the nulls are local per-process */
 
 	internals = eth_dev->data->dev_private;
-	internals->packet_size = packet_size;
-	internals->packet_copy = packet_copy;
+	internals->packet_size = args->packet_size;
+	internals->packet_copy = args->packet_copy;
 	internals->port_id = eth_dev->data->port_id;
 	rte_eth_random_addr(internals->eth_addr.addr_bytes);
 
@@ -520,7 +523,7 @@ eth_dev_null_create(struct rte_vdev_device *dev,
 	eth_dev->dev_ops = &ops;
 
 	/* finally assign rx and tx ops */
-	if (packet_copy) {
+	if (internals->packet_copy) {
 		eth_dev->rx_pkt_burst = eth_null_copy_rx;
 		eth_dev->tx_pkt_burst = eth_null_copy_tx;
 	} else {
@@ -570,8 +573,10 @@ static int
 rte_pmd_null_probe(struct rte_vdev_device *dev)
 {
 	const char *name, *params;
-	unsigned int packet_size = default_packet_size;
-	unsigned int packet_copy = default_packet_copy;
+	struct pmd_options args = {
+		.packet_copy = default_packet_copy,
+		.packet_size = default_packet_size,
+	};
 	struct rte_kvargs *kvlist = NULL;
 	struct rte_eth_dev *eth_dev;
 	int ret;
@@ -612,23 +617,23 @@ rte_pmd_null_probe(struct rte_vdev_device *dev)
 
 		ret = rte_kvargs_process(kvlist,
 				ETH_NULL_PACKET_SIZE_ARG,
-				&get_packet_size_arg, &packet_size);
+				&get_packet_size_arg, &args.packet_size);
 		if (ret < 0)
 			goto free_kvlist;
 
 
 		ret = rte_kvargs_process(kvlist,
 				ETH_NULL_PACKET_COPY_ARG,
-				&get_packet_copy_arg, &packet_copy);
+				&get_packet_copy_arg, &args.packet_copy);
 		if (ret < 0)
 			goto free_kvlist;
 	}
 
 	PMD_LOG(INFO, "Configure pmd_null: packet size is %d, "
-			"packet copy is %s", packet_size,
-			packet_copy ? "enabled" : "disabled");
+			"packet copy is %s", args.packet_size,
+			args.packet_copy ? "enabled" : "disabled");
 
-	ret = eth_dev_null_create(dev, packet_size, packet_copy);
+	ret = eth_dev_null_create(dev, &args);
 
 free_kvlist:
 	if (kvlist)
Signature · 1 line
2.24.1
#5 lore raw

PATCH 6/6 net/null: add argument for no Rx

In reply to Ferruh Yigit

Add an new device argument 'no-rx', which will prevent PMD receiving packets.

This is useful for testing when a PMD is needed only to send packets to.

---
 doc/guides/nics/null.rst        |  4 +++
 drivers/net/null/rte_eth_null.c | 55 ++++++++++++++++++++++++++++++++-
 2 files changed, 58 insertions(+), 1 deletion(-)
doc/guides/nics/null.rst +4 -0
index 405edf07b..c68d0d605 100644
@@ -37,3 +37,7 @@ Runtime Config Options
 
    $RTE_TARGET/app/testpmd -l 0-3 -n 4 --vdev "net_null0,size=256" -- -i
 
+- ``no-rx`` [optional, default disabled]
+
+ Makes PMD more like ``/dev/null``. On Rx no packets received, on Tx all packets are freed.
+ This option can't co-exist with ``copy`` option.
drivers/net/null/rte_eth_null.c +54 -1

127 lines not on this page · raw message

Signature · 1 line
2.24.1
#6 lore raw
In reply to Ferruh Yigit
Quoted 11 lines · Ferruh Yigit

Secondary process uses the primary process device and while setting the Rx/Tx functions it uses the device arguments from the secondary process instead of the primary ones.

This may cause primary and secondary process use different Rx/Tx functions unintentionally.

Fixes bccc77a6a74a ("net/null: fix multi-process Rx and Tx")

Is there any review on this set? If there is no objection I will merge it soon.

#7 lore raw
Quoted 8 lines · Ferruh Yigit
Quoted 7 lines · Ferruh Yigit

Secondary process uses the primary process device and while setting the Rx/Tx functions it uses the device arguments from the secondary process instead of the primary ones.

This may cause primary and secondary process use different Rx/Tx functions unintentionally.

Quoted 7 lines
Quoted 4 lines
Fixes bccc77a6a74a ("net/null: fix multi-process Rx and Tx")

Is there any review on this set? If there is no objection I will merge it soon.

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