examples/ethtool: remove useless device info query
cover examples/ethtool: remove useless device info query
Thomas Monjalon (2):
examples/ethtool: remove query of default config examples/ethtool: allocate only one mempool
examples/ethtool/ethtool-app/main.c | 36 ++++++++--------------------- 1 file changed, 10 insertions(+), 26 deletions(-)
Signature · 1 line
2.20.1
PATCH 1/2 examples/ethtool: remove query of default config
The default config is used if the setup parameter is NULL. No need to query the default config with rte_eth_dev_info_get(). The function call will be removed with another useless info.
--- examples/ethtool/ethtool-app/main.c | 4 +--- 1 file changed, 1 insertion(+), 3 deletions(-)
examples/ethtool/ethtool-app/main.c +1 -3
@@ -95,7 +95,6 @@ static void setup_ports(struct app_config *app_cfg, int cnt_ports)char str_name[16 ];uint16_t nb_rxd =PORT_RX_QUEUE_SIZE ;uint16_t nb_txd =PORT_TX_QUEUE_SIZE ; -struct rte_eth_txconf txconf;memset (&cfg_port,0 ,sizeof (cfg_port)); cfg_port.txmode.mq_mode =ETH_MQ_TX_NONE ;
@@ -140,10 +139,9 @@ static void setup_ports(struct app_config *app_cfg, int cnt_ports)rte_exit (EXIT_FAILURE ,"rte_eth_rx_queue_setup failed" ); - txconf = dev_info.default_txconf;if (rte_eth_tx_queue_setup ( idx_port,0 , nb_txd, -rte_eth_dev_socket_id (idx_port), &txconf) <0 ) +rte_eth_dev_socket_id (idx_port),NULL ) <0 )rte_exit (EXIT_FAILURE ,"rte_eth_tx_queue_setup failed" );
Signature · 1 line
2.20.1
Re: [PATCH 1/2] examples/ethtool: remove query of default config
Quoted 5 lines · Thomas Monjalon
The default config is used if the setup parameter is NULL. No need to query the default config with
rte_eth_dev_info_get(). The function call will be removed with another useless info.Signed-off-by Thomas Monjalon <[email protected]>
Rami Rosen
Reviewed-byRe: [PATCH 1/2] examples/ethtool: remove query of default config
Quoted 8 lines · Thomas Monjalon
Quoted 5 lines
The default config is used if the setup parameter is NULL. No need to query the default config with
rte_eth_dev_info_get(). The function call will be removed with another useless info.Signed-off-by Thomas Monjalon <[email protected]>Acked-by Bruce Richardson <[email protected]>
Ferruh Yigit
AppliedRe: [PATCH 1/2] examples/ethtool: remove query of default config
Quoted 5 lines · Rami Rosen
Quoted 4 lines · Thomas Monjalon
Quoted 4 lines
The default config is used if the setup parameter is NULL. No need to query the default config with
rte_eth_dev_info_get(). The function call will be removed with another useless info.
Quoted 7 lines
Quoted 4 lines
Quoted 1 line
Signed-off-by Thomas Monjalon <[email protected]>Acked-by Bruce Richardson <[email protected]>Reviewed-by Rami Rosen <[email protected]>
Applied to dpdk-next-net/master, thanks.
(only this patch merged, not series. Other patch seems waiting for comment.)
Re: [PATCH 1/2] examples/ethtool: remove query of default config
20/03/2019 14:46, Ferruh Yigit:
Quoted 3 lines
Applied to dpdk-next-net/master, thanks.
(only this patch merged, not series. Other patch seems waiting for comment.)
Applying half of patchset makes tracking of patches less obvious. I prefer we avoid such practice.
Re: [PATCH 1/2] examples/ethtool: remove query of default config
Quoted 7 lines · Thomas Monjalon
20/03/2019 14:46, Ferruh Yigit:
Quoted 3 lines
Applied to dpdk-next-net/master, thanks.
(only this patch merged, not series. Other patch seems waiting for comment.)
Applying half of patchset makes tracking of patches less obvious. I prefer we avoid such practice.
Generally I agree, this case the patches in patchset looks can be independent and merged one was trivial, I tend to this for similar cases.
Do you prefer me drop the applied patch?
Re: [PATCH 1/2] examples/ethtool: remove query of default config
20/03/2019 14:56, Ferruh Yigit:
Quoted 11 lines · Thomas Monjalon
Quoted 7 lines
20/03/2019 14:46, Ferruh Yigit:
Quoted 3 lines
Applied to dpdk-next-net/master, thanks.
(only this patch merged, not series. Other patch seems waiting for comment.)
Applying half of patchset makes tracking of patches less obvious. I prefer we avoid such practice.
Generally I agree, this case the patches in patchset looks can be independent and merged one was trivial, I tend to this for similar cases.
Even if it is independent (it is not really in this case), it looks weird to see half of patchset pending in patchwork. We may spend some time to understand what happened.
Quoted 1 line
Do you prefer me drop the applied patch?
You can keep it now it's applied.
PATCH 2/2 examples/ethtool: allocate only one mempool
In reply to Thomas MonjalonNo need to allocate one mempool per port.
The number of mbufs is fixed for simplicity.
--- examples/ethtool/ethtool-app/main.c | 32 ++++++++--------------------- 1 file changed, 9 insertions(+), 23 deletions(-)
examples/ethtool/ethtool-app/main.c +9 -23
@@ -22,7 +22,7 @@#define MAX_BURST_LENGTH 32#define PORT_RX_QUEUE_SIZE 1024#define PORT_TX_QUEUE_SIZE 1024 -#define PKTPOOL_EXTRA_SIZE 512 +#define NUM_MBUFS 8191#define PKTPOOL_CACHE 32
@@ -38,7 +38,6 @@ struct app_port {int port_active;int port_dirty;int idx_port; -struct rte_mempool *pkt_pool; };struct app_config {
@@ -89,36 +88,23 @@ void mark_port_newmac(int idx_port)static void setup_ports (struct app_config *app_cfg,int cnt_ports) {int idx_port; -int size_pktpool;struct rte_eth_conf cfg_port; -struct rte_eth_dev_info dev_info; -char str_name[16 ];uint16_t nb_rxd =PORT_RX_QUEUE_SIZE ;uint16_t nb_txd =PORT_TX_QUEUE_SIZE ; +struct rte_mempool *pkt_pool;memset (&cfg_port,0 ,sizeof (cfg_port)); cfg_port.txmode.mq_mode =ETH_MQ_TX_NONE ; + pkt_pool =rte_pktmbuf_pool_create ("mbuf pool" , +NUM_MBUFS ,PKTPOOL_CACHE ,0 ,RTE_MBUF_DEFAULT_BUF_SIZE , +rte_socket_id ()); +if (pkt_pool ==NULL ) +rte_exit (EXIT_FAILURE ,"rte_pktmbuf_pool_create failed" ); +for (idx_port =0 ; idx_port < cnt_ports; idx_port++) {struct app_port *ptr_port = &app_cfg->ports[idx_port]; -rte_eth_dev_info_get (idx_port, &dev_info); - size_pktpool = dev_info.rx_desc_lim.nb_max + - dev_info.tx_desc_lim.nb_max +PKTPOOL_EXTRA_SIZE ; - -snprintf (str_name,16 ,"pkt_pool%i" , idx_port); - ptr_port->pkt_pool =rte_pktmbuf_pool_create ( - str_name, - size_pktpool,PKTPOOL_CACHE , -0 , -RTE_MBUF_DEFAULT_BUF_SIZE , -rte_socket_id () - ); -if (ptr_port->pkt_pool ==NULL ) -rte_exit (EXIT_FAILURE , -"rte_pktmbuf_pool_create failed" - ); -printf ("Init port %i..\n" , idx_port); ptr_port->port_active =1 ; ptr_port->port_dirty =0 ;
@@ -135,7 +121,7 @@ static void setup_ports(struct app_config *app_cfg, int cnt_ports)if (rte_eth_rx_queue_setup ( idx_port,0 , nb_rxd,rte_eth_dev_socket_id (idx_port),NULL , - ptr_port->pkt_pool) <0 ) + pkt_pool) <0 )rte_exit (EXIT_FAILURE ,"rte_eth_rx_queue_setup failed" );
Signature · 1 line
2.20.1
Re: [PATCH 2/2] examples/ethtool: allocate only one mempool
Quoted 3 lines · Thomas Monjalon
No need to allocate one mempool per port.
The number of mbufs is fixed for simplicity.
While it's simpler to use a fixed number of mbufs, having a fixed number of mbufs multiplied by the number of ports is safer IMHO, and not really that much more complicated.
Otherwise, fully agree with the principle of the change.
/Bruce