Skip to contentSkip to conversation

updates for testpmd

PATCH Superseded Wei Hu (Xavier) 4 participants 11 messages Sep 8, 2020 – Sep 28, 2020

lore
Share & tools
lore
#0 lore raw

cover updates for testpmd

This series are minor updates for testpmd application.

Huisong Li (2):
app/testpmd: update Rx RSS HASH offload when setting MQ RSS app/testpmd: retain original FDIR mode when configuring DCB

 app/test-pmd/testpmd.c | 7 +++++--
 1 file changed, 5 insertions(+), 2 deletions(-)
Signature · 1 line
2.9.5
#1 lore raw

PATCH 1/2 app/testpmd: update Rx RSS HASH offload when setting MQ RSS

From: Huisong Li <[email protected]>

Currently, when starting testpmd application without '--disable-rss' and the number of Rx queue configured is greater than 1, ETH_MQ_RX_RSS flag is set in port->dev_conf.rxmode.mq_mode in testpmd application, and DEV_RX_OFFLOAD_RSS_HASH flag is set in rx_offloads (dev->data->dev_conf.rxmode.offloads) according to the ETH_MQ_RX_RSS flag of rxmode.mq_mode in PMD drivers.

However, DEV_RX_OFFLOAD_RSS_HASH is not set to rx_offloads maintained in testpmd application, this will cause the inconsistent problem that rx_offloads is different for testpmd and PMD drivers.

---
 app/test-pmd/testpmd.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)
app/test-pmd/testpmd.c +4 -2
index 7842c3b..73543bb 100644
@@ -3356,11 +3356,13 @@ init_port_config(void)
 		}
 
 		if (port->dcb_flag == 0) {
-			if( port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0)
+			if (port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0) {
 				port->dev_conf.rxmode.mq_mode =
 					(enum rte_eth_rx_mq_mode)
 						(rx_mq_mode & ETH_MQ_RX_RSS);
-			else
+				port->dev_conf.rxmode.offloads |=
+							DEV_RX_OFFLOAD_RSS_HASH;
+			} else
 				port->dev_conf.rxmode.mq_mode = ETH_MQ_RX_NONE;
 		}
 
Signature · 1 line
2.9.5
#2 lore raw

Re: [dpdk-dev] [PATCH 1/2] app/testpmd: update Rx RSS HASH offload when setting MQ RSS

Quoted 9 lines · Wei Hu (Xavier)

From: Huisong Li <[email protected]>

Currently, when starting testpmd application without '--disable-rss' and the number of Rx queue configured is greater than 1, ETH_MQ_RX_RSS flag is set in port->dev_conf.rxmode.mq_mode in testpmd application, and DEV_RX_OFFLOAD_RSS_HASH flag is set in rx_offloads (dev->data->dev_conf.rxmode.offloads) according to the ETH_MQ_RX_RSS flag of rxmode.mq_mode in PMD drivers.

Quoted 3 lines

However, DEV_RX_OFFLOAD_RSS_HASH is not set to rx_offloads maintained in testpmd application, this will cause the inconsistent problem that rx_offloads is different for testpmd and PMD drivers.

Yes for DEV_RX_OFFLOAD_RSS_HASH, application rx_offload config and PMD one diverges, for *some* PMDs.

This is done to have the backward compatibility of the PMD behavior.

The PMDs that would like to write the provide the calculated hash value back, overwrites the offload config to enable 'DEV_RX_OFFLOAD_RSS_HASH'. And does more work than user requested, this shouldn't have any side affect. Some doesn't provide the hash unless user explicitly requests it.

Applications shouldn't set it blindly, at least first check if PMD supports it but even that case, unless it is needed I think HASH offload shouldn't be requested by default.

Quoted 28 lines
---
  app/test-pmd/testpmd.c | 6 ++++--
  1 file changed, 4 insertions(+), 2 deletions(-)
app/test-pmd/testpmd.c +4 -2
index 7842c3b..73543bb 100644
@@ -3356,11 +3356,13 @@ init_port_config(void)
  		}
 
  		if (port->dcb_flag == 0) {
-			if( port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0)
+			if (port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0) {
  				port->dev_conf.rxmode.mq_mode =
  					(enum rte_eth_rx_mq_mode)
  						(rx_mq_mode & ETH_MQ_RX_RSS);
-			else
+				port->dev_conf.rxmode.offloads |=
+							DEV_RX_OFFLOAD_RSS_HASH;
+			} else
  				port->dev_conf.rxmode.mq_mode = ETH_MQ_RX_NONE;
  		}
 
#3 lore raw

Re: [dpdk-dev] [PATCH 1/2] app/testpmd: update Rx RSS HASH offload when setting MQ RSS

Hi, Ferruh Yigit

Quoted 14 lines · Ferruh Yigit
Quoted 9 lines · Wei Hu (Xavier)

From: Huisong Li <[email protected]>

Currently, when starting testpmd application without '--disable-rss' and the number of Rx queue configured is greater than 1, ETH_MQ_RX_RSS flag is set in port->dev_conf.rxmode.mq_mode in testpmd application, and DEV_RX_OFFLOAD_RSS_HASH flag is set in rx_offloads (dev->data->dev_conf.rxmode.offloads) according to the ETH_MQ_RX_RSS flag of rxmode.mq_mode in PMD drivers.

Quoted 3 lines

However, DEV_RX_OFFLOAD_RSS_HASH is not set to rx_offloads maintained in testpmd application, this will cause the inconsistent problem that rx_offloads is different for testpmd and PMD drivers.

Quoted 8 lines

Yes for DEV_RX_OFFLOAD_RSS_HASH, application rx_offload config and PMD one diverges, for *some* PMDs.

This is done to have the backward compatibility of the PMD behavior.

The PMDs that would like to write the provide the calculated hash value back, overwrites the offload config to enable 'DEV_RX_OFFLOAD_RSS_HASH'. And does more work than user requested, this shouldn't have any side affect. Some doesn't provide the hash unless user explicitly requests it.

Applications shouldn't set it blindly, at least first check if PMD supports it but even that case, unless it is needed I think HASH offload shouldn't be requested by default.

OK, we are going to do modification to add check if PMD support DEV_RX_OFFLOAD_RSS_HASH before setting it as below:

@@ -3356,11 +3356,13 @@ init_port_config(void)

           }
           if (port->dcb_flag == 0) {
              if (port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0) {
                   port->dev_conf.rxmode.mq_mode =
                       (enum rte_eth_rx_mq_mode)
                           (rx_mq_mode & ETH_MQ_RX_RSS);
 		  if (port->dev_info.rx_offload_capa &
		      DEV_RX_OFFLOAD_RSS_HASH)
 			port->dev_conf.rxmode.offloads |=
						DEV_RX_OFFLOAD_RSS_HASH;
             } else
                   port->dev_conf.rxmode.mq_mode = ETH_MQ_RX_NONE;
           }

Thanks, Xavier

Quoted 27 lines
Quoted 27 lines

---
  app/test-pmd/testpmd.c | 6 ++++--

  1 file changed, 4 insertions(+), 2 deletions(-)
app/test-pmd/testpmd.c +1 -1
index 7842c3b..73543bb 100644
@@ -3356,11 +3356,13 @@ init_port_config(void)
           }
           if (port->dcb_flag == 0) {
-            if( port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0)
+            if (port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0) {

                  port->dev_conf.rxmode.mq_mode =
                      (enum rte_eth_rx_mq_mode) (rx_mq_mode & ETH_MQ_RX_RSS); -            else
+                port->dev_conf.rxmode.offloads |=
+                            DEV_RX_OFFLOAD_RSS_HASH;
+            } else
                  port->dev_conf.rxmode.mq_mode = ETH_MQ_RX_NONE; }

#4 lore raw

Re: [dpdk-dev] [PATCH 1/2] app/testpmd: update Rx RSS HASH offload when setting MQ RSS

Quoted 28 lines · Wei Hu (Xavier)

Hi, Ferruh Yigit

Quoted 14 lines · Ferruh Yigit
Quoted 9 lines · Wei Hu (Xavier)

From: Huisong Li <[email protected]>

Currently, when starting testpmd application without '--disable-rss' and the number of Rx queue configured is greater than 1, ETH_MQ_RX_RSS flag is set in port->dev_conf.rxmode.mq_mode in testpmd application, and DEV_RX_OFFLOAD_RSS_HASH flag is set in rx_offloads (dev->data->dev_conf.rxmode.offloads) according to the ETH_MQ_RX_RSS flag of rxmode.mq_mode in PMD drivers.

Quoted 3 lines

However, DEV_RX_OFFLOAD_RSS_HASH is not set to rx_offloads maintained in testpmd application, this will cause the inconsistent problem that rx_offloads is different for testpmd and PMD drivers.

Quoted 8 lines

Yes for DEV_RX_OFFLOAD_RSS_HASH, application rx_offload config and PMD one diverges, for *some* PMDs.

This is done to have the backward compatibility of the PMD behavior.

The PMDs that would like to write the provide the calculated hash value back, overwrites the offload config to enable 'DEV_RX_OFFLOAD_RSS_HASH'. And does more work than user requested, this shouldn't have any side affect. Some doesn't provide the hash unless user explicitly requests it.

Applications shouldn't set it blindly, at least first check if PMD supports it but even that case, unless it is needed I think HASH offload shouldn't be requested by default.

OK, we are going to do modification to add check if PMD support DEV_RX_OFFLOAD_RSS_HASH before setting it as below:

Quoted 15 lines

 @@ -3356,11 +3356,13 @@ init_port_config(void)
           }
           if (port->dcb_flag == 0) {
              if (port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0) { port->dev_conf.rxmode.mq_mode =
                       (enum rte_eth_rx_mq_mode) (rx_mq_mode & ETH_MQ_RX_RSS); if (port->dev_info.rx_offload_capa & DEV_RX_OFFLOAD_RSS_HASH) port->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
             } else
                   port->dev_conf.rxmode.mq_mode = ETH_MQ_RX_NONE; }

Hi Xavier,

The capability check is correct thing to do, but even in that case I am not sure if we should set the config by default.

The reason to extract the 'DEV_RX_OFFLOAD_RSS_HASH' offload is to gain performance for some NICs, enabling it by default defeats the purpose.

The offload should be enabled when application needs the provided hash value.

I can see you are trying to remove the divergence between PMD and application config, this can be fixed when all PMDs take this user offload request into account instead of overwriting, but for some PMDs it is cheaper to provide the hash value instead of checks, so this divergence is not easy to address.

#5 lore raw

Re: [dpdk-dev] [PATCH 1/2] app/testpmd: update Rx RSS HASH offload when setting MQ RSS

Hi, Ferruh Yigit

Quoted 47 lines · Ferruh Yigit
Quoted 28 lines · Wei Hu (Xavier)

Hi, Ferruh Yigit

Quoted 14 lines · Ferruh Yigit
Quoted 9 lines · Wei Hu (Xavier)

From: Huisong Li <[email protected]>

Currently, when starting testpmd application without '--disable-rss' and the number of Rx queue configured is greater than 1, ETH_MQ_RX_RSS flag is set in port->dev_conf.rxmode.mq_mode in testpmd application, and DEV_RX_OFFLOAD_RSS_HASH flag is set in rx_offloads (dev->data->dev_conf.rxmode.offloads) according to the ETH_MQ_RX_RSS flag of rxmode.mq_mode in PMD drivers.

Quoted 3 lines

However, DEV_RX_OFFLOAD_RSS_HASH is not set to rx_offloads maintained in testpmd application, this will cause the inconsistent problem that rx_offloads is different for testpmd and PMD drivers.

Quoted 8 lines

Yes for DEV_RX_OFFLOAD_RSS_HASH, application rx_offload config and PMD one diverges, for *some* PMDs.

This is done to have the backward compatibility of the PMD behavior.

The PMDs that would like to write the provide the calculated hash value back, overwrites the offload config to enable 'DEV_RX_OFFLOAD_RSS_HASH'. And does more work than user requested, this shouldn't have any side affect. Some doesn't provide the hash unless user explicitly requests it.

Applications shouldn't set it blindly, at least first check if PMD supports it but even that case, unless it is needed I think HASH offload shouldn't be requested by default.

OK, we are going to do modification to add check if PMD support DEV_RX_OFFLOAD_RSS_HASH before setting it as below:

Quoted 15 lines

  @@ -3356,11 +3356,13 @@ init_port_config(void)
            }
            if (port->dcb_flag == 0) {
               if (port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0) { port->dev_conf.rxmode.mq_mode =
                        (enum rte_eth_rx_mq_mode) (rx_mq_mode & ETH_MQ_RX_RSS); if (port->dev_info.rx_offload_capa & DEV_RX_OFFLOAD_RSS_HASH) port->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
              } else
                    port->dev_conf.rxmode.mq_mode = ETH_MQ_RX_NONE; }

Hi Xavier,

Quoted 8 lines

The capability check is correct thing to do, but even in that case I am not sure if we should set the config by default.

The reason to extract the 'DEV_RX_OFFLOAD_RSS_HASH' offload is to gain performance for some NICs, enabling it by default defeats the purpose.

The offload should be enabled when application needs the provided hash value.

I can see you are trying to remove the divergence between PMD and application config, this can be fixed when all PMDs take this user offload request into account instead of overwriting, but for some PMDs it is cheaper to provide the hash value instead of checks, so this divergence is not easy to address.

OK, please ignore this patch and the V2 patch.

http://patches.dpdk.org/patch/78502/

Thanks.

Regards

Xavier

#6 lore raw

PATCH 2/2 app/testpmd: retain original FDIR mode when configuring DCB

In reply to Wei Hu (Xavier)

From: Huisong Li <[email protected]>

Sometimes, we have to start testpmd application with --pkt-filter-mode to test flow table feature. When using 'port config 0 dcb vt off 4 pfc on' command, the dcb information are configured in init_port_dcb_config function by get_eth_dcb_conf function, And then rte_eth_dev_configure API function will be called to re-configure PMD driver.

The values of rxmode and txmode in rte_eth_conf struct used to configure PMD driver, come from the current values maintained in testpmd. However, the fdir_conf.mode in rte_eth_conf struct is not set, which may cause that some PMD driver cannot test the flow table feature when multiple TCs are enabled.

Fixes 1a572499beb6 ("app/testpmd: setup DCB forwarding based on traffic class")
---
 app/test-pmd/testpmd.c | 1 +
 1 file changed, 1 insertion(+)
app/test-pmd/testpmd.c +1 -0
index 73543bb..19bf972 100644
@@ -3521,6 +3521,7 @@ init_port_dcb_config(portid_t pid,
 
 	port_conf.rxmode = rte_port->dev_conf.rxmode;
 	port_conf.txmode = rte_port->dev_conf.txmode;
+	port_conf.fdir_conf.mode = rte_port->dev_conf.fdir_conf.mode;
 
 	/*set configuration of DCB in vt mode and DCB in non-vt mode*/
 	retval = get_eth_dcb_conf(pid, &port_conf, dcb_mode, num_tcs, pfc_en);
Signature · 1 line
2.9.5
#7 lore raw

Re: [dpdk-dev] [PATCH 2/2] app/testpmd: retain original FDIR mode when configuring DCB

Quoted 29 lines · Wei Hu (Xavier)

From: Huisong Li <[email protected]>

Sometimes, we have to start testpmd application with --pkt-filter-mode to test flow table feature. When using 'port config 0 dcb vt off 4 pfc on' command, the dcb information are configured in init_port_dcb_config function by get_eth_dcb_conf function, And then rte_eth_dev_configure API function will be called to re-configure PMD driver.

The values of rxmode and txmode in rte_eth_conf struct used to configure PMD driver, come from the current values maintained in testpmd. However, the fdir_conf.mode in rte_eth_conf struct is not set, which may cause that some PMD driver cannot test the flow table feature when multiple TCs are enabled.

Fixes 1a572499beb6 ("app/testpmd: setup DCB forwarding based on traffic class")
---
 app/test-pmd/testpmd.c | 1 +
 1 file changed, 1 insertion(+)
app/test-pmd/testpmd.c +0 -0
index 73543bb..19bf972 100644

app/test-pmd/testpmd.c init_port_dcb_config(portid_t pid, · line 3524

Quoted 4 lines
app/test-pmd/testpmd.c +1 -0
@@ -3521,6 +3521,7 @@ init_port_dcb_config(portid_t pid,
 
        port_conf.rxmode = rte_port->dev_conf.rxmode;
        port_conf.txmode = rte_port->dev_conf.txmode;
+       port_conf.fdir_conf.mode = rte_port->dev_conf.fdir_conf.mode;

This is probably not necessary. FDIR is due for deprecation [1].
https://doc.dpdk.org/guides/rel_notes/deprecation.html

Quoted 6 lines
app/test-pmd/testpmd.c +0 -1
        /*set configuration of DCB in vt mode and DCB in non-vt mode*/
        retval = get_eth_dcb_conf(pid, &port_conf, dcb_mode, num_tcs, pfc_en);
--

2.9.5

#8 lore raw

Re: [dpdk-dev] [PATCH 2/2] app/testpmd: retain original FDIR mode when configuring DCB

Hi, Ajit Khaparde

Quoted 31 lines · Ajit Khaparde
Quoted 29 lines · Wei Hu (Xavier)

From: Huisong Li <[email protected]>

Sometimes, we have to start testpmd application with --pkt-filter-mode to test flow table feature. When using 'port config 0 dcb vt off 4 pfc on' command, the dcb information are configured in init_port_dcb_config function by get_eth_dcb_conf function, And then rte_eth_dev_configure API function will be called to re-configure PMD driver.

The values of rxmode and txmode in rte_eth_conf struct used to configure PMD driver, come from the current values maintained in testpmd. However, the fdir_conf.mode in rte_eth_conf struct is not set, which may cause that some PMD driver cannot test the flow table feature when multiple TCs are enabled.

Fixes 1a572499beb6 ("app/testpmd: setup DCB forwarding based on traffic class")
---
  app/test-pmd/testpmd.c | 1 +
  1 file changed, 1 insertion(+)
app/test-pmd/testpmd.c +0 -0
index 73543bb..19bf972 100644
@@ -3521,6 +3521,7 @@ init_port_dcb_config(portid_t pid,
 
Quoted 5 lines

app/test-pmd/testpmd.c init_port_dcb_config(portid_t pid, · line 3524

Quoted 3 lines
app/test-pmd/testpmd.c +1 -0
         port_conf.rxmode = rte_port->dev_conf.rxmode;
         port_conf.txmode = rte_port->dev_conf.txmode;
+       port_conf.fdir_conf.mode = rte_port->dev_conf.fdir_conf.mode;

This is probably not necessary. FDIR is due for deprecation [1].
https://doc.dpdk.org/guides/rel_notes/deprecation.html

Thanks your for pointing out it, this patch is really necessary.

Thanks, Xavier

Quoted 5 lines
Quoted 5 lines
app/test-pmd/testpmd.c +0 -1
         /*set configuration of DCB in vt mode and DCB in non-vt mode*/
         retval = get_eth_dcb_conf(pid, &port_conf, dcb_mode, num_tcs, pfc_en);
--

2.9.5

#9 lore raw

Re: [dpdk-dev] [PATCH 2/2] app/testpmd: retain original FDIR mode when configuring DCB

Hi, Ajit Khaparde

Quoted 35 lines · Wei Hu (Xavier)

Hi, Ajit Khaparde

Quoted 32 lines · Ajit Khaparde
Quoted 30 lines · Wei Hu (Xavier)

From: Huisong Li <[email protected]>

Sometimes, we have to start testpmd application with --pkt-filter-mode to
test flow table feature. When using 'port config 0 dcb vt off 4 pfc on' command, the dcb information are configured in init_port_dcb_config function by get_eth_dcb_conf function, And then rte_eth_dev_configure API function will be called to re-configure PMD driver.

The values of rxmode and txmode in rte_eth_conf struct used to configure PMD driver, come from the current values maintained in testpmd. However, the fdir_conf.mode in rte_eth_conf struct is not set, which may cause that some PMD driver cannot test the flow table feature when multiple TCs are enabled.

Fixes 1a572499beb6 ("app/testpmd: setup DCB forwarding based on traffic class")

---
  app/test-pmd/testpmd.c | 1 +

  1 file changed, 1 insertion(+)
app/test-pmd/testpmd.c +0 -0
index 73543bb..19bf972 100644
@@ -3521,6 +3521,7 @@ init_port_dcb_config(portid_t pid,
 
Quoted 8 lines
Quoted 5 lines

app/test-pmd/testpmd.c init_port_dcb_config(portid_t pid, · line 3524

Quoted 3 lines
app/test-pmd/testpmd.c +1 -0
          port_conf.rxmode = rte_port->dev_conf.rxmode;
          port_conf.txmode = rte_port->dev_conf.txmode;
+       port_conf.fdir_conf.mode = rte_port->dev_conf.fdir_conf.mode;

This is probably not necessary. FDIR is due for deprecation [1].
https://doc.dpdk.org/guides/rel_notes/deprecation.html

Thanks your for pointing out it, this patch is really necessary.

Sorry, it is a typo.

This patch is not really necessary.

Thanks, Xavier

Quoted 8 lines

Thanks, Xavier

Quoted 5 lines
Quoted 5 lines
app/test-pmd/testpmd.c +0 -0
          /*set configuration of DCB in vt mode and DCB in non-vt mode*/
          retval = get_eth_dcb_conf(pid, &port_conf, dcb_mode, num_tcs, pfc_en);
Signature · 1 line
2.9.5
#10 lore raw

PATCH v2 1/1 app/testpmd: update Rx RSS HASH offload when setting MQ RSS

In reply to Wei Hu (Xavier)

From: Huisong Li <[email protected]>

Currently, when starting testpmd application without '--disable-rss' and the number of Rx queue configured is greater than 1, ETH_MQ_RX_RSS flag is set in port->dev_conf.rxmode.mq_mode in testpmd application, and DEV_RX_OFFLOAD_RSS_HASH flag is set in rx_offloads (dev->data->dev_conf.rxmode.offloads) according to the ETH_MQ_RX_RSS flag of rxmode.mq_mode in PMD drivers.

However, DEV_RX_OFFLOAD_RSS_HASH is not set to rx_offloads maintained in testpmd application, this will cause the inconsistent problem that rx_offloads is different for testpmd and PMD drivers.

---
v1 -> v2: add check if PMD support DEV_RX_OFFLOAD_RSS_HASH before
setting it.

---
 app/test-pmd/testpmd.c | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)
app/test-pmd/testpmd.c +6 -2
index fe6450c..a36db76 100644
@@ -3353,11 +3353,15 @@ init_port_config(void)
 		}
 
 		if (port->dcb_flag == 0) {
-			if( port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0)
+			if (port->dev_conf.rx_adv_conf.rss_conf.rss_hf != 0) {
 				port->dev_conf.rxmode.mq_mode =
 					(enum rte_eth_rx_mq_mode)
 						(rx_mq_mode & ETH_MQ_RX_RSS);
-			else
+				if (port->dev_info.rx_offload_capa &
+				    DEV_RX_OFFLOAD_RSS_HASH)
+					port->dev_conf.rxmode.offloads |=
+							DEV_RX_OFFLOAD_RSS_HASH;
+			} else
 				port->dev_conf.rxmode.mq_mode = ETH_MQ_RX_NONE;
 		}
 
Signature · 1 line
2.9.5