[PATCH 0/3] testpmd support stop specify lcore

[PATCH 0/3] testpmd support stop specify lcore

[Chengwen Feng]

This patch support stop specify lcore for testpmd, and also include
two bugfix which found when test DCB function.

Chengwen Feng (3):
  app/testpmd: fix invalid txp when setup DCB forward
  app/testpmd: fix wrong Rx queues when setup DCB forward
  app/testpmd: support stop specify lcore

 app/test-pmd/cmdline.c                      | 63 +++++++++++++++++++++
 app/test-pmd/config.c                       |  9 +--
 doc/guides/testpmd_app_ug/testpmd_funcs.rst |  7 +++
 3 files changed, 73 insertions(+), 6 deletions(-)

-- 
2.17.1

[PATCH 1/3] app/testpmd: fix invalid txp when setup DCB forward

[Chengwen Feng]

The txp maybe invalid (e.g. start with only one port but set with 1),
this commit fix it by get txp from fwd_topology_tx_port_get() function.

An added benefit is that the DCB test also supports '--port-topology'
parameter.

Fixes: 1a572499beb6 ("app/testpmd: setup DCB forwarding based on traffic class")
Cc: stable@dpdk.org

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
---
 app/test-pmd/config.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
index 3ce2a14a1b..0f687018c7 100644
--- a/app/test-pmd/config.c
+++ b/app/test-pmd/config.c
@@ -5187,7 +5187,7 @@ dcb_fwd_config_setup(void)
 	/* reinitialize forwarding streams */
 	init_fwd_streams();
 	sm_id = 0;
-	txp = 1;
+	txp = fwd_topology_tx_port_get(rxp);
 	/* get the dcb info on the first RX and TX ports */
 	(void)rte_eth_dev_get_dcb_info(fwd_ports_ids[rxp], &rxp_dcb_info);
 	(void)rte_eth_dev_get_dcb_info(fwd_ports_ids[txp], &txp_dcb_info);
@@ -5235,11 +5235,8 @@ dcb_fwd_config_setup(void)
 			rxp++;
 		if (rxp >= nb_fwd_ports)
 			return;
+		txp = fwd_topology_tx_port_get(rxp);
 		/* get the dcb information on next RX and TX ports */
-		if ((rxp & 0x1) == 0)
-			txp = (portid_t) (rxp + 1);
-		else
-			txp = (portid_t) (rxp - 1);
 		rte_eth_dev_get_dcb_info(fwd_ports_ids[rxp], &rxp_dcb_info);
 		rte_eth_dev_get_dcb_info(fwd_ports_ids[txp], &txp_dcb_info);
 	}
-- 
2.17.1

[PATCH 2/3] app/testpmd: fix wrong Rx queues when setup DCB forward

[Chengwen Feng]

The nb_rx_queue should get from rxp_dcb_info not txp_dcb_info, this
commit fix it.

Fixes: 1a572499beb6 ("app/testpmd: setup DCB forwarding based on traffic class")
Cc: stable@dpdk.org

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
---
 app/test-pmd/config.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
index 0f687018c7..8557371488 100644
--- a/app/test-pmd/config.c
+++ b/app/test-pmd/config.c
@@ -5205,7 +5205,7 @@ dcb_fwd_config_setup(void)
 				fwd_lcores[lc_id]->stream_idx;
 			rxq = rxp_dcb_info.tc_queue.tc_rxq[i][tc].base;
 			txq = txp_dcb_info.tc_queue.tc_txq[i][tc].base;
-			nb_rx_queue = txp_dcb_info.tc_queue.tc_rxq[i][tc].nb_queue;
+			nb_rx_queue = rxp_dcb_info.tc_queue.tc_rxq[i][tc].nb_queue;
 			nb_tx_queue = txp_dcb_info.tc_queue.tc_txq[i][tc].nb_queue;
 			for (j = 0; j < nb_rx_queue; j++) {
 				struct fwd_stream *fs;
-- 
2.17.1

[PATCH 3/3] app/testpmd: support stop specify lcore

[Chengwen Feng]

This commit supports stop specify lcore, the command:

  stop fwd_core (lcore_id)

The background of this command:
1. Only some TCs are expected to generate traffic when the DCB function
   is tested based on txonly forwarding.
2. Because each lcore will process all the traffic of one TC, therefore,
   we could stop the lcores which process unexpected TCs by this
   command.

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
---
 app/test-pmd/cmdline.c                      | 63 +++++++++++++++++++++
 doc/guides/testpmd_app_ug/testpmd_funcs.rst |  7 +++
 2 files changed, 70 insertions(+)

diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 22afbdbad3..e03f6e6181 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -150,6 +150,9 @@ static void cmd_help_long_parsed(void *parsed_result,
 			"    Stop packet forwarding, and display accumulated"
 			" statistics.\n\n"
 
+			"stop fwd_core (lcore_id)\n"
+			"    Stop specify lcore's forwarding.\n\n"
+
 			"quit\n"
 			"    Quit to prompt.\n\n"
 		);
@@ -3975,6 +3978,65 @@ static cmdline_parse_inst_t cmd_stop = {
 	},
 };
 
+/* *** stop specify forward core *** */
+struct cmd_stop_fwd_core_result {
+	cmdline_fixed_string_t stop;
+	cmdline_fixed_string_t fwd_core;
+	uint32_t lcore_id;
+};
+
+static void
+cmd_stop_fwd_core_parsed(void *parsed_result,
+			 __rte_unused struct cmdline *cl,
+			 __rte_unused void *data)
+{
+	struct cmd_stop_fwd_core_result *res = parsed_result;
+	struct fwd_lcore *fc = lcore_to_fwd_lcore(res->lcore_id);
+
+	if (test_done) {
+		fprintf(stderr, "Packet forwarding not started\n");
+		return;
+	}
+
+	if (fc == NULL) {
+		fprintf(stderr, "core: %u not in the forward corelist.\n", res->lcore_id);
+		return;
+	}
+
+	if (fc->stopped) {
+		fprintf(stderr, "core: %u already stopped!\n", res->lcore_id);
+		return;
+	}
+
+	printf("Telling core: %u to stop...", res->lcore_id);
+	fc->stopped = 1;
+	printf("\nWaiting for core: %u to finish...\n", res->lcore_id);
+	rte_eal_wait_lcore(res->lcore_id);
+	printf("Done.\n");
+}
+
+static cmdline_parse_token_string_t cmd_stop_fwd_core_stop =
+	TOKEN_STRING_INITIALIZER(struct cmd_stop_fwd_core_result,
+			stop, "stop");
+static cmdline_parse_token_string_t cmd_stop_fwd_core_fwd_core =
+	TOKEN_STRING_INITIALIZER(struct cmd_stop_fwd_core_result,
+			fwd_core, "fwd_core");
+static cmdline_parse_token_num_t cmd_stop_fwd_core_lcore_id =
+	TOKEN_NUM_INITIALIZER(struct cmd_stop_fwd_core_result,
+			lcore_id, RTE_UINT32);
+
+static cmdline_parse_inst_t cmd_stop_fwd_core = {
+	.f = cmd_stop_fwd_core_parsed,
+	.data = NULL,
+	.help_str = "stop fwd_core <lcore_id>: stop specify lcore's forwarding.",
+	.tokens = {
+		(void *)&cmd_stop_fwd_core_stop,
+		(void *)&cmd_stop_fwd_core_fwd_core,
+		(void *)&cmd_stop_fwd_core_lcore_id,
+		NULL,
+	},
+};
+
 static unsigned int
 get_ptype(char *value)
 {
@@ -14051,6 +14113,7 @@ static cmdline_parse_ctx_t builtin_ctx[] = {
 	&cmd_config_dcb,
 	&cmd_read_rxd_txd,
 	&cmd_stop,
+	&cmd_stop_fwd_core,
 	&cmd_mac_addr,
 	&cmd_set_fwd_eth_peer,
 	&cmd_set_xstats_hide_zero,
diff --git a/doc/guides/testpmd_app_ug/testpmd_funcs.rst b/doc/guides/testpmd_app_ug/testpmd_funcs.rst
index e423abd40e..0a362d06b8 100644
--- a/doc/guides/testpmd_app_ug/testpmd_funcs.rst
+++ b/doc/guides/testpmd_app_ug/testpmd_funcs.rst
@@ -143,6 +143,13 @@ Stop packet forwarding, and display accumulated statistics::
 
    testpmd> stop
 
+stop fwd_core
+~~~~~~~~~~~~~
+
+Stop specify lcore's forwarding::
+
+   testpmd> stop fwd_core (lcore_id)
+
 quit
 ~~~~
 
-- 
2.17.1

Re: [PATCH 0/3] testpmd support stop specify lcore

[Stephen Hemminger]

On Tue, 4 Nov 2025 12:09:13 +0800
Chengwen Feng <fengchengwen@huawei.com> wrote:

> This patch support stop specify lcore for testpmd, and also include
> two bugfix which found when test DCB function.
> 
> Chengwen Feng (3):
>   app/testpmd: fix invalid txp when setup DCB forward
>   app/testpmd: fix wrong Rx queues when setup DCB forward
>   app/testpmd: support stop specify lcore
> 
>  app/test-pmd/cmdline.c                      | 63 +++++++++++++++++++++
>  app/test-pmd/config.c                       |  9 +--
>  doc/guides/testpmd_app_ug/testpmd_funcs.rst |  7 +++
>  3 files changed, 73 insertions(+), 6 deletions(-)
> 

If there is a stop lcore, why is there not a command to restart the lcore?
Looks like this just an optimization for DCB.
Could it be handled better by having testpmd not spawn lcores that
have no queues? 

Re: [PATCH 0/3] testpmd support stop specify lcore

[fengchengwen]

On 11/5/2025 8:21 AM, Stephen Hemminger wrote:
> On Tue, 4 Nov 2025 12:09:13 +0800
> Chengwen Feng <fengchengwen@huawei.com> wrote:
> 
>> This patch support stop specify lcore for testpmd, and also include
>> two bugfix which found when test DCB function.
>>
>> Chengwen Feng (3):
>>   app/testpmd: fix invalid txp when setup DCB forward
>>   app/testpmd: fix wrong Rx queues when setup DCB forward
>>   app/testpmd: support stop specify lcore
>>
>>  app/test-pmd/cmdline.c                      | 63 +++++++++++++++++++++
>>  app/test-pmd/config.c                       |  9 +--
>>  doc/guides/testpmd_app_ug/testpmd_funcs.rst |  7 +++
>>  3 files changed, 73 insertions(+), 6 deletions(-)
>>
> 
> If there is a stop lcore, why is there not a command to restart the lcore?

The restart lcore is more complex which include data prepare (init).

But I think could impl such pause / restart function by config fc->stopped a special value:

Current:
	run_pkt_fwd_on_lcore(struct fwd_lcore *fc, packet_fwd_t pkt_fwd)
	{
		do {
			forwarding logic
		} while(!fc->stopped);
	}

We could adjust the logic to support pause / restart function:
	run_pkt_fwd_on_lcore(struct fwd_lcore *fc, packet_fwd_t pkt_fwd)
	{
	_restart:
		do {
			forwarding logic
		} while(!fc->stopped);
 		while (fc->stopped == 2) {
			nop
		}
		if (fc->stopped == 1)
			return;
		if (fc->stopped == 0)
			// means restart
			goto _restart;
	}
So we could configure fc->stopped = 2 to pause forwarding, and config it to 0 to restart forwarding.

> Looks like this just an optimization for DCB.

Yes, mainly for test the Tx direction's DCB function.

> Could it be handled better by having testpmd not spawn lcores that
> have no queues? 

The lcore spawned by testpmd always has queues. In this patchset, we just stop some queue's forwarding.

> 

[PATCH v2 0/3] testpmd support pause/resume specify lcore's fwd

[Chengwen Feng]

This patch support pause/resume specify lcore's forwarding for testpmd,
and also include two bugfix which found when test DCB function.

Chengwen Feng (3):
  app/testpmd: fix invalid txp when setup DCB forward
  app/testpmd: fix wrong Rx queues when setup DCB forward
  app/testpmd: support pause/resume specify lcore

 app/test-pmd/cmdline.c                      | 126 ++++++++++++++++++++
 app/test-pmd/config.c                       |   9 +-
 app/test-pmd/testpmd.c                      |   5 +
 app/test-pmd/testpmd.h                      |   2 +-
 doc/guides/testpmd_app_ug/testpmd_funcs.rst |  14 +++
 5 files changed, 149 insertions(+), 7 deletions(-)

-- 
2.17.1

[PATCH v2 1/3] app/testpmd: fix invalid txp when setup DCB forward

[Chengwen Feng]

The txp maybe invalid (e.g. start with only one port but set with 1),
this commit fix it by get txp from fwd_topology_tx_port_get() function.

An added benefit is that the DCB test also supports '--port-topology'
parameter.

Fixes: 1a572499beb6 ("app/testpmd: setup DCB forwarding based on traffic class")
Cc: stable@dpdk.org

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
---
 app/test-pmd/config.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)

diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
index 3ce2a14a1b..0f687018c7 100644
--- a/app/test-pmd/config.c
+++ b/app/test-pmd/config.c
@@ -5187,7 +5187,7 @@ dcb_fwd_config_setup(void)
 	/* reinitialize forwarding streams */
 	init_fwd_streams();
 	sm_id = 0;
-	txp = 1;
+	txp = fwd_topology_tx_port_get(rxp);
 	/* get the dcb info on the first RX and TX ports */
 	(void)rte_eth_dev_get_dcb_info(fwd_ports_ids[rxp], &rxp_dcb_info);
 	(void)rte_eth_dev_get_dcb_info(fwd_ports_ids[txp], &txp_dcb_info);
@@ -5235,11 +5235,8 @@ dcb_fwd_config_setup(void)
 			rxp++;
 		if (rxp >= nb_fwd_ports)
 			return;
+		txp = fwd_topology_tx_port_get(rxp);
 		/* get the dcb information on next RX and TX ports */
-		if ((rxp & 0x1) == 0)
-			txp = (portid_t) (rxp + 1);
-		else
-			txp = (portid_t) (rxp - 1);
 		rte_eth_dev_get_dcb_info(fwd_ports_ids[rxp], &rxp_dcb_info);
 		rte_eth_dev_get_dcb_info(fwd_ports_ids[txp], &txp_dcb_info);
 	}
-- 
2.17.1

[PATCH v2 2/3] app/testpmd: fix wrong Rx queues when setup DCB forward

[Chengwen Feng]

The nb_rx_queue should get from rxp_dcb_info not txp_dcb_info, this
commit fix it.

Fixes: 1a572499beb6 ("app/testpmd: setup DCB forwarding based on traffic class")
Cc: stable@dpdk.org

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
---
 app/test-pmd/config.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/app/test-pmd/config.c b/app/test-pmd/config.c
index 0f687018c7..8557371488 100644
--- a/app/test-pmd/config.c
+++ b/app/test-pmd/config.c
@@ -5205,7 +5205,7 @@ dcb_fwd_config_setup(void)
 				fwd_lcores[lc_id]->stream_idx;
 			rxq = rxp_dcb_info.tc_queue.tc_rxq[i][tc].base;
 			txq = txp_dcb_info.tc_queue.tc_txq[i][tc].base;
-			nb_rx_queue = txp_dcb_info.tc_queue.tc_rxq[i][tc].nb_queue;
+			nb_rx_queue = rxp_dcb_info.tc_queue.tc_rxq[i][tc].nb_queue;
 			nb_tx_queue = txp_dcb_info.tc_queue.tc_txq[i][tc].nb_queue;
 			for (j = 0; j < nb_rx_queue; j++) {
 				struct fwd_stream *fs;
-- 
2.17.1

[PATCH v2 3/3] app/testpmd: support pause/resume specify lcore

[Chengwen Feng]

This commit supports pause/resume specify lcore, the command:

  pause  fwd_core (lcore_id)
  resume fwd_core (lcore_id)

The background of this command:
1. Only some TCs are expected to generate traffic when the DCB function
   is tested based on txonly forwarding.
2. Because each lcore will process all the traffic of one TC, therefore,
   we could pause the lcores which process unexpected TCs by this
   command.

Signed-off-by: Chengwen Feng <fengchengwen@huawei.com>
---
 app/test-pmd/cmdline.c                      | 126 ++++++++++++++++++++
 app/test-pmd/testpmd.c                      |   5 +
 app/test-pmd/testpmd.h                      |   2 +-
 doc/guides/testpmd_app_ug/testpmd_funcs.rst |  14 +++
 4 files changed, 146 insertions(+), 1 deletion(-)

diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c
index 22afbdbad3..dacc1921bd 100644
--- a/app/test-pmd/cmdline.c
+++ b/app/test-pmd/cmdline.c
@@ -150,6 +150,12 @@ static void cmd_help_long_parsed(void *parsed_result,
 			"    Stop packet forwarding, and display accumulated"
 			" statistics.\n\n"
 
+			"pause fwd_core (lcore_id)\n"
+			"    Pause specify lcore's forwarding.\n\n"
+
+			"resume fwd_core (lcore_id)\n"
+			"    Resume specify lcore's forwarding.\n\n"
+
 			"quit\n"
 			"    Quit to prompt.\n\n"
 		);
@@ -3975,6 +3981,124 @@ static cmdline_parse_inst_t cmd_stop = {
 	},
 };
 
+/* *** pause specify forward core *** */
+struct cmd_pause_fwd_core_result {
+	cmdline_fixed_string_t pause;
+	cmdline_fixed_string_t fwd_core;
+	uint32_t lcore_id;
+};
+
+static void
+cmd_pause_fwd_core_parsed(void *parsed_result,
+			  __rte_unused struct cmdline *cl,
+			  __rte_unused void *data)
+{
+	struct cmd_pause_fwd_core_result *res = parsed_result;
+	struct fwd_lcore *fc = lcore_to_fwd_lcore(res->lcore_id);
+
+	if (test_done) {
+		fprintf(stderr, "Packet forwarding not started\n");
+		return;
+	}
+
+	if (fc == NULL) {
+		fprintf(stderr, "core: %u not in the forward corelist.\n", res->lcore_id);
+		return;
+	}
+
+	if (fc->stopped == 2) {
+		fprintf(stderr, "core: %u already paused!\n", res->lcore_id);
+		return;
+	}
+
+	printf("Telling core: %u to pause...", res->lcore_id);
+	fc->stopped = 2;
+	printf("\nWaiting for core: %u to finish...\n", res->lcore_id);
+	rte_delay_ms(10);
+	printf("Done.\n");
+}
+
+static cmdline_parse_token_string_t cmd_pause_fwd_core_pause =
+	TOKEN_STRING_INITIALIZER(struct cmd_pause_fwd_core_result,
+			pause, "pause");
+static cmdline_parse_token_string_t cmd_pause_fwd_core_fwd_core =
+	TOKEN_STRING_INITIALIZER(struct cmd_pause_fwd_core_result,
+			fwd_core, "fwd_core");
+static cmdline_parse_token_num_t cmd_pause_fwd_core_lcore_id =
+	TOKEN_NUM_INITIALIZER(struct cmd_pause_fwd_core_result,
+			lcore_id, RTE_UINT32);
+
+static cmdline_parse_inst_t cmd_pause_fwd_core = {
+	.f = cmd_pause_fwd_core_parsed,
+	.data = NULL,
+	.help_str = "pause fwd_core <lcore_id>: pause specify lcore's forwarding.",
+	.tokens = {
+		(void *)&cmd_pause_fwd_core_pause,
+		(void *)&cmd_pause_fwd_core_fwd_core,
+		(void *)&cmd_pause_fwd_core_lcore_id,
+		NULL,
+	},
+};
+
+/* *** resume specify forward core *** */
+struct cmd_resume_fwd_core_result {
+	cmdline_fixed_string_t resume;
+	cmdline_fixed_string_t fwd_core;
+	uint32_t lcore_id;
+};
+
+static void
+cmd_resume_fwd_core_parsed(void *parsed_result,
+			   __rte_unused struct cmdline *cl,
+			   __rte_unused void *data)
+{
+	struct cmd_resume_fwd_core_result *res = parsed_result;
+	struct fwd_lcore *fc = lcore_to_fwd_lcore(res->lcore_id);
+
+	if (test_done) {
+		fprintf(stderr, "Packet forwarding not started\n");
+		return;
+	}
+
+	if (fc == NULL) {
+		fprintf(stderr, "core: %u not in the forward corelist.\n", res->lcore_id);
+		return;
+	}
+
+	if (fc->stopped != 2) {
+		fprintf(stderr, "core: %u not in pause state!\n", res->lcore_id);
+		return;
+	}
+
+	printf("Telling core: %u to resume...", res->lcore_id);
+	fc->stopped = 0;
+	printf("\nWaiting for core: %u to resume...\n", res->lcore_id);
+	rte_delay_ms(10);
+	printf("Done.\n");
+}
+
+static cmdline_parse_token_string_t cmd_resume_fwd_core_resume =
+	TOKEN_STRING_INITIALIZER(struct cmd_resume_fwd_core_result,
+			resume, "resume");
+static cmdline_parse_token_string_t cmd_resume_fwd_core_fwd_core =
+	TOKEN_STRING_INITIALIZER(struct cmd_resume_fwd_core_result,
+			fwd_core, "fwd_core");
+static cmdline_parse_token_num_t cmd_resume_fwd_core_lcore_id =
+	TOKEN_NUM_INITIALIZER(struct cmd_resume_fwd_core_result,
+			lcore_id, RTE_UINT32);
+
+static cmdline_parse_inst_t cmd_resume_fwd_core = {
+	.f = cmd_resume_fwd_core_parsed,
+	.data = NULL,
+	.help_str = "resume fwd_core <lcore_id>: resume specify lcore's forwarding.",
+	.tokens = {
+		(void *)&cmd_resume_fwd_core_resume,
+		(void *)&cmd_resume_fwd_core_fwd_core,
+		(void *)&cmd_resume_fwd_core_lcore_id,
+		NULL,
+	},
+};
+
 static unsigned int
 get_ptype(char *value)
 {
@@ -14051,6 +14175,8 @@ static cmdline_parse_ctx_t builtin_ctx[] = {
 	&cmd_config_dcb,
 	&cmd_read_rxd_txd,
 	&cmd_stop,
+	&cmd_pause_fwd_core,
+	&cmd_resume_fwd_core,
 	&cmd_mac_addr,
 	&cmd_set_fwd_eth_peer,
 	&cmd_set_xstats_hide_zero,
diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c
index 2360da3a48..05549be1c5 100644
--- a/app/test-pmd/testpmd.c
+++ b/app/test-pmd/testpmd.c
@@ -2247,6 +2247,7 @@ run_pkt_fwd_on_lcore(struct fwd_lcore *fc, packet_fwd_t pkt_fwd)
 #endif
 	fsm = &fwd_streams[fc->stream_idx];
 	nb_fs = fc->stream_nb;
+_resume:
 	prev_tsc = rte_rdtsc();
 	do {
 		for (sm_id = 0; sm_id < nb_fs; sm_id++) {
@@ -2287,6 +2288,10 @@ run_pkt_fwd_on_lcore(struct fwd_lcore *fc, packet_fwd_t pkt_fwd)
 			prev_tsc = tsc;
 		}
 	} while (! fc->stopped);
+	while (fc->stopped == 2)
+		rte_pause();
+	if (fc->stopped == 0)
+		goto _resume;
 }
 
 static int
diff --git a/app/test-pmd/testpmd.h b/app/test-pmd/testpmd.h
index fa46865c67..c3c4373af7 100644
--- a/app/test-pmd/testpmd.h
+++ b/app/test-pmd/testpmd.h
@@ -399,7 +399,7 @@ struct fwd_lcore {
 	streamid_t stream_idx;   /**< index of 1st stream in "fwd_streams" */
 	streamid_t stream_nb;    /**< number of streams in "fwd_streams" */
 	lcoreid_t  cpuid_idx;    /**< index of logical core in CPU id table */
-	volatile char stopped;   /**< stop forwarding when set */
+	volatile char stopped;   /**< stop forwarding when set to 1, pause when set to 2 */
 	uint64_t total_cycles;   /**< used with --record-core-cycles */
 };
 
diff --git a/doc/guides/testpmd_app_ug/testpmd_funcs.rst b/doc/guides/testpmd_app_ug/testpmd_funcs.rst
index e423abd40e..be125ff35c 100644
--- a/doc/guides/testpmd_app_ug/testpmd_funcs.rst
+++ b/doc/guides/testpmd_app_ug/testpmd_funcs.rst
@@ -143,6 +143,20 @@ Stop packet forwarding, and display accumulated statistics::
 
    testpmd> stop
 
+pause fwd_core
+~~~~~~~~~~~~~~
+
+Pause specify lcore's forwarding::
+
+   testpmd> pause fwd_core (lcore_id)
+
+resume fwd_core
+~~~~~~~~~~~~~~~
+
+Resume specify lcore's forwarding::
+
+   testpmd> resume fwd_core (lcore_id)
+
 quit
 ~~~~
 
-- 
2.17.1

Re: [PATCH v2 0/3] testpmd support pause/resume specify lcore's fwd

[Stephen Hemminger]

On Wed, 5 Nov 2025 17:47:45 +0800
Chengwen Feng <fengchengwen@huawei.com> wrote:

> This patch support pause/resume specify lcore's forwarding for testpmd,
> and also include two bugfix which found when test DCB function.
> 
> Chengwen Feng (3):
>   app/testpmd: fix invalid txp when setup DCB forward
>   app/testpmd: fix wrong Rx queues when setup DCB forward
>   app/testpmd: support pause/resume specify lcore
> 
>  app/test-pmd/cmdline.c                      | 126 ++++++++++++++++++++
>  app/test-pmd/config.c                       |   9 +-
>  app/test-pmd/testpmd.c                      |   5 +
>  app/test-pmd/testpmd.h                      |   2 +-
>  doc/guides/testpmd_app_ug/testpmd_funcs.rst |  14 +++
>  5 files changed, 149 insertions(+), 7 deletions(-)
> 

Please split bug fix from enhancements.

The whole scheme looks like an optimization for power usage in testpmd.
Since testpmd is solely focused on performance, I don't think you have
provided enough justification to add the new feature.

With DCB if there are TC's that could receive packets, it is better
not to never poll those queues. 

That said, longer term it would make sense to bring of the power
saving logic that is in examples into testpmd.