[dpdk-dev,1/4] testpmd: fix LACP ports to work with idle links
Commit Message
Problem: When there is little or no TX traffic on an LACP port
(bonding mode 4), we don't call its tx burst function in a timely
manner, and thus we don't transmit LACPDUs when we should.
Solution: Add and maintain an "lacp_master" flag in rte_port struct.
In the main packet forwarding loop, if port is an LACP master, in 1
out of N loops force an empty call to the tx burst API.
Signed-off-by: Robert Sanford <rsanford@akamai.com>
---
app/test-pmd/cmdline.c | 9 +++++++++
app/test-pmd/testpmd.c | 37 +++++++++++++++++++++++++++++++++++++
app/test-pmd/testpmd.h | 4 ++++
3 files changed, 50 insertions(+), 0 deletions(-)
Comments
> -----Original Message-----
> From: rsanford2 at gmail.com
> Sent: Tuesday, August 2, 2016 4:43 AM
> Subject: [dpdk-dev] [PATCH 1/4] testpmd: fix LACP ports to work with idle links
>
> Problem: When there is little or no TX traffic on an LACP port (bonding mode 4),
> we don't call its tx burst function in a timely manner, and thus we don't
> transmit LACPDUs when we should.
>
> Solution: Add and maintain an "lacp_master" flag in rte_port struct.
> In the main packet forwarding loop, if port is an LACP master, in 1 out of N
> loops force an empty call to the tx burst API.
>
> Signed-off-by: Robert Sanford <rsanford at akamai.com>
> ---
> app/test-pmd/cmdline.c | 9 +++++++++
> app/test-pmd/testpmd.c | 37 +++++++++++++++++++++++++++++++++++++
> app/test-pmd/testpmd.h | 4 ++++
> 3 files changed, 50 insertions(+), 0 deletions(-)
>
> diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c index
> f90befc..2a629ee 100644
> --- a/app/test-pmd/cmdline.c
> +++ b/app/test-pmd/cmdline.c
> @@ -3975,6 +3975,10 @@ static void cmd_set_bonding_mode_parsed(void
> *parsed_result,
> /* Set the bonding mode for the relevant port. */
> if (0 != rte_eth_bond_mode_set(port_id, res->value))
> printf("\t Failed to set bonding mode for port = %d.\n", port_id);
> + else if (res->value == BONDING_MODE_8023AD)
> + set_port_lacp_master(port_id);
> + else
> + clear_port_lacp_master(port_id);
> }
>
> cmdline_parse_token_string_t cmd_setbonding_mode_set = @@ -4408,6
> +4412,11 @@ static void cmd_create_bonded_device_parsed(void
> *parsed_result,
> reconfig(port_id, res->socket);
> rte_eth_promiscuous_enable(port_id);
> ports[port_id].enabled = 1;
> +
> + if (res->mode == BONDING_MODE_8023AD)
> + set_port_lacp_master(port_id);
> + else
> + clear_port_lacp_master(port_id);
> }
>
> }
> diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
> 1428974..806667e 100644
> --- a/app/test-pmd/testpmd.c
> +++ b/app/test-pmd/testpmd.c
> @@ -920,12 +920,28 @@ run_pkt_fwd_on_lcore(struct fwd_lcore *fc,
> packet_fwd_t pkt_fwd)
> struct fwd_stream **fsm;
> streamid_t nb_fs;
> streamid_t sm_id;
> + unsigned int loop_count = 0;
>
> fsm = &fwd_streams[fc->stream_idx];
> nb_fs = fc->stream_nb;
> do {
> for (sm_id = 0; sm_id < nb_fs; sm_id++)
> (*pkt_fwd)(fsm[sm_id]);
> +
> + /*
> + * Per rte_eth_bond.h, we must invoke LACP master's tx
> + * burst function at least once every 100 ms.
> + */
> + loop_count++;
> + if (likely(loop_count % 1024 != 0))
> + continue;
> + for (sm_id = 0; sm_id < nb_fs; sm_id++) {
> + struct fwd_stream *fs = fsm[sm_id];
> +
> + if (port_is_lacp_master(fs->tx_port))
> + rte_eth_tx_burst(fs->tx_port, fs->tx_queue,
> + NULL, 0);
> + }
By add tx_burst here will change default logic. Some fwd engine are usually used to measure
Performance . Additional cycle in fast path may impact that. It would be better to move additional
sending to your fwd engine.
> } while (! fc->stopped);
> }
>
> @@ -1881,6 +1897,27 @@ uint8_t port_is_bonding_slave(portid_t slave_pid)
> return port->slave_flag;
> }
>
> +void set_port_lacp_master(portid_t pid) {
> + struct rte_port *port = &ports[pid];
> +
You may need to check the pid is valid or not.
And the function is simple, could you just assign the lacp_master
in the caller instead of defining a new one?
> + port->lacp_master = 1;
> +}
> +
> +void clear_port_lacp_master(portid_t pid) {
> + struct rte_port *port = &ports[pid];
> +
Same comment as above.
> + port->lacp_master = 0;
> +}
> +
> +uint8_t port_is_lacp_master(portid_t pid) {
> + struct rte_port *port = &ports[pid];
> +
> + return port->lacp_master;
> +}
> +
> const uint16_t vlan_tags[] = {
> 0, 1, 2, 3, 4, 5, 6, 7,
> 8, 9, 10, 11, 12, 13, 14, 15,
> diff --git a/app/test-pmd/testpmd.h b/app/test-pmd/testpmd.h index
> 2b281cc..0898194 100644
> --- a/app/test-pmd/testpmd.h
> +++ b/app/test-pmd/testpmd.h
> @@ -170,6 +170,7 @@ struct rte_port {
> struct ether_addr *mc_addr_pool; /**< pool of multicast addrs */
> uint32_t mc_addr_nb; /**< nb. of addr. in mc_addr_pool */
> uint8_t slave_flag; /**< bonding slave port */
> + uint8_t lacp_master;/**< bonding LACP master */
> };
>
> extern portid_t __rte_unused
> @@ -542,6 +543,9 @@ void init_port_config(void); void
> set_port_slave_flag(portid_t slave_pid); void clear_port_slave_flag(portid_t
> slave_pid); uint8_t port_is_bonding_slave(portid_t slave_pid);
> +void set_port_lacp_master(portid_t pid); void
> +clear_port_lacp_master(portid_t pid); uint8_t
> +port_is_lacp_master(portid_t pid);
>
> int init_port_dcb_config(portid_t pid, enum dcb_mode_enable dcb_mode,
> enum rte_eth_nb_tcs num_tcs,
> --
> 1.7.1
>
On 6/21/2017 6:25 PM, jingjing.wu at intel.com (Wu, Jingjing) wrote:
>> -----Original Message-----
>> From: rsanford2 at gmail.com
>> Sent: Tuesday, August 2, 2016 4:43 AM
>> Subject: [dpdk-dev] [PATCH 1/4] testpmd: fix LACP ports to work with idle links
>>
>> Problem: When there is little or no TX traffic on an LACP port (bonding mode 4),
>> we don't call its tx burst function in a timely manner, and thus we don't
>> transmit LACPDUs when we should.
>>
>> Solution: Add and maintain an "lacp_master" flag in rte_port struct.
>> In the main packet forwarding loop, if port is an LACP master, in 1 out of N
>> loops force an empty call to the tx burst API.
>>
>> Signed-off-by: Robert Sanford <rsanford at akamai.com>
>> ---
>> app/test-pmd/cmdline.c | 9 +++++++++
>> app/test-pmd/testpmd.c | 37 +++++++++++++++++++++++++++++++++++++
>> app/test-pmd/testpmd.h | 4 ++++
>> 3 files changed, 50 insertions(+), 0 deletions(-)
>>
>> diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c index
>> f90befc..2a629ee 100644
>> --- a/app/test-pmd/cmdline.c
>> +++ b/app/test-pmd/cmdline.c
>> @@ -3975,6 +3975,10 @@ static void cmd_set_bonding_mode_parsed(void
>> *parsed_result,
>> /* Set the bonding mode for the relevant port. */
>> if (0 != rte_eth_bond_mode_set(port_id, res->value))
>> printf("\t Failed to set bonding mode for port = %d.\n", port_id);
>> + else if (res->value == BONDING_MODE_8023AD)
>> + set_port_lacp_master(port_id);
>> + else
>> + clear_port_lacp_master(port_id);
>> }
>>
>> cmdline_parse_token_string_t cmd_setbonding_mode_set = @@ -4408,6
>> +4412,11 @@ static void cmd_create_bonded_device_parsed(void
>> *parsed_result,
>> reconfig(port_id, res->socket);
>> rte_eth_promiscuous_enable(port_id);
>> ports[port_id].enabled = 1;
>> +
>> + if (res->mode == BONDING_MODE_8023AD)
>> + set_port_lacp_master(port_id);
>> + else
>> + clear_port_lacp_master(port_id);
>> }
>>
>> }
>> diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
>> 1428974..806667e 100644
>> --- a/app/test-pmd/testpmd.c
>> +++ b/app/test-pmd/testpmd.c
>> @@ -920,12 +920,28 @@ run_pkt_fwd_on_lcore(struct fwd_lcore *fc,
>> packet_fwd_t pkt_fwd)
>> struct fwd_stream **fsm;
>> streamid_t nb_fs;
>> streamid_t sm_id;
>> + unsigned int loop_count = 0;
>>
>> fsm = &fwd_streams[fc->stream_idx];
>> nb_fs = fc->stream_nb;
>> do {
>> for (sm_id = 0; sm_id < nb_fs; sm_id++)
>> (*pkt_fwd)(fsm[sm_id]);
>> +
>> + /*
>> + * Per rte_eth_bond.h, we must invoke LACP master's tx
>> + * burst function at least once every 100 ms.
>> + */
>> + loop_count++;
>> + if (likely(loop_count % 1024 != 0))
>> + continue;
>> + for (sm_id = 0; sm_id < nb_fs; sm_id++) {
>> + struct fwd_stream *fs = fsm[sm_id];
>> +
>> + if (port_is_lacp_master(fs->tx_port))
>> + rte_eth_tx_burst(fs->tx_port, fs->tx_queue,
>> + NULL, 0);
>> + }
> By add tx_burst here will change default logic. Some fwd engine are usually used to measure
> Performance . Additional cycle in fast path may impact that. It would be better to move additional
> sending to your fwd engine.
Hi Robert,
This patchset is older than a year and not responded back to latest comments.
Is this still a valid patch?
Do you have any objection to remove this from patchwork?
Thanks,
ferruh
On 10/30/2017 6:07 PM, Ferruh Yigit wrote:
> On 6/21/2017 6:25 PM, jingjing.wu at intel.com (Wu, Jingjing) wrote:
>>> -----Original Message-----
>>> From: rsanford2 at gmail.com
>>> Sent: Tuesday, August 2, 2016 4:43 AM
>>> Subject: [dpdk-dev] [PATCH 1/4] testpmd: fix LACP ports to work with idle links
>>>
>>> Problem: When there is little or no TX traffic on an LACP port (bonding mode 4),
>>> we don't call its tx burst function in a timely manner, and thus we don't
>>> transmit LACPDUs when we should.
>>>
>>> Solution: Add and maintain an "lacp_master" flag in rte_port struct.
>>> In the main packet forwarding loop, if port is an LACP master, in 1 out of N
>>> loops force an empty call to the tx burst API.
>>>
>>> Signed-off-by: Robert Sanford <rsanford at akamai.com>
>>> ---
>>> app/test-pmd/cmdline.c | 9 +++++++++
>>> app/test-pmd/testpmd.c | 37 +++++++++++++++++++++++++++++++++++++
>>> app/test-pmd/testpmd.h | 4 ++++
>>> 3 files changed, 50 insertions(+), 0 deletions(-)
>>>
>>> diff --git a/app/test-pmd/cmdline.c b/app/test-pmd/cmdline.c index
>>> f90befc..2a629ee 100644
>>> --- a/app/test-pmd/cmdline.c
>>> +++ b/app/test-pmd/cmdline.c
>>> @@ -3975,6 +3975,10 @@ static void cmd_set_bonding_mode_parsed(void
>>> *parsed_result,
>>> /* Set the bonding mode for the relevant port. */
>>> if (0 != rte_eth_bond_mode_set(port_id, res->value))
>>> printf("\t Failed to set bonding mode for port = %d.\n", port_id);
>>> + else if (res->value == BONDING_MODE_8023AD)
>>> + set_port_lacp_master(port_id);
>>> + else
>>> + clear_port_lacp_master(port_id);
>>> }
>>>
>>> cmdline_parse_token_string_t cmd_setbonding_mode_set = @@ -4408,6
>>> +4412,11 @@ static void cmd_create_bonded_device_parsed(void
>>> *parsed_result,
>>> reconfig(port_id, res->socket);
>>> rte_eth_promiscuous_enable(port_id);
>>> ports[port_id].enabled = 1;
>>> +
>>> + if (res->mode == BONDING_MODE_8023AD)
>>> + set_port_lacp_master(port_id);
>>> + else
>>> + clear_port_lacp_master(port_id);
>>> }
>>>
>>> }
>>> diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c index
>>> 1428974..806667e 100644
>>> --- a/app/test-pmd/testpmd.c
>>> +++ b/app/test-pmd/testpmd.c
>>> @@ -920,12 +920,28 @@ run_pkt_fwd_on_lcore(struct fwd_lcore *fc,
>>> packet_fwd_t pkt_fwd)
>>> struct fwd_stream **fsm;
>>> streamid_t nb_fs;
>>> streamid_t sm_id;
>>> + unsigned int loop_count = 0;
>>>
>>> fsm = &fwd_streams[fc->stream_idx];
>>> nb_fs = fc->stream_nb;
>>> do {
>>> for (sm_id = 0; sm_id < nb_fs; sm_id++)
>>> (*pkt_fwd)(fsm[sm_id]);
>>> +
>>> + /*
>>> + * Per rte_eth_bond.h, we must invoke LACP master's tx
>>> + * burst function at least once every 100 ms.
>>> + */
>>> + loop_count++;
>>> + if (likely(loop_count % 1024 != 0))
>>> + continue;
>>> + for (sm_id = 0; sm_id < nb_fs; sm_id++) {
>>> + struct fwd_stream *fs = fsm[sm_id];
>>> +
>>> + if (port_is_lacp_master(fs->tx_port))
>>> + rte_eth_tx_burst(fs->tx_port, fs->tx_queue,
>>> + NULL, 0);
>>> + }
>> By add tx_burst here will change default logic. Some fwd engine are usually used to measure
>> Performance . Additional cycle in fast path may impact that. It would be better to move additional
>> sending to your fwd engine.
>
> Hi Robert,
>
> This patchset is older than a year and not responded back to latest comments.
>
> Is this still a valid patch?
> Do you have any objection to remove this from patchwork?
Patches are updated as rejected.
>
> Thanks,
> ferruh
>
@@ -3975,6 +3975,10 @@ static void cmd_set_bonding_mode_parsed(void *parsed_result,
/* Set the bonding mode for the relevant port. */
if (0 != rte_eth_bond_mode_set(port_id, res->value))
printf("\t Failed to set bonding mode for port = %d.\n", port_id);
+ else if (res->value == BONDING_MODE_8023AD)
+ set_port_lacp_master(port_id);
+ else
+ clear_port_lacp_master(port_id);
}
cmdline_parse_token_string_t cmd_setbonding_mode_set =
@@ -4408,6 +4412,11 @@ static void cmd_create_bonded_device_parsed(void *parsed_result,
reconfig(port_id, res->socket);
rte_eth_promiscuous_enable(port_id);
ports[port_id].enabled = 1;
+
+ if (res->mode == BONDING_MODE_8023AD)
+ set_port_lacp_master(port_id);
+ else
+ clear_port_lacp_master(port_id);
}
}
@@ -920,12 +920,28 @@ run_pkt_fwd_on_lcore(struct fwd_lcore *fc, packet_fwd_t pkt_fwd)
struct fwd_stream **fsm;
streamid_t nb_fs;
streamid_t sm_id;
+ unsigned int loop_count = 0;
fsm = &fwd_streams[fc->stream_idx];
nb_fs = fc->stream_nb;
do {
for (sm_id = 0; sm_id < nb_fs; sm_id++)
(*pkt_fwd)(fsm[sm_id]);
+
+ /*
+ * Per rte_eth_bond.h, we must invoke LACP master's tx
+ * burst function at least once every 100 ms.
+ */
+ loop_count++;
+ if (likely(loop_count % 1024 != 0))
+ continue;
+ for (sm_id = 0; sm_id < nb_fs; sm_id++) {
+ struct fwd_stream *fs = fsm[sm_id];
+
+ if (port_is_lacp_master(fs->tx_port))
+ rte_eth_tx_burst(fs->tx_port, fs->tx_queue,
+ NULL, 0);
+ }
} while (! fc->stopped);
}
@@ -1881,6 +1897,27 @@ uint8_t port_is_bonding_slave(portid_t slave_pid)
return port->slave_flag;
}
+void set_port_lacp_master(portid_t pid)
+{
+ struct rte_port *port = &ports[pid];
+
+ port->lacp_master = 1;
+}
+
+void clear_port_lacp_master(portid_t pid)
+{
+ struct rte_port *port = &ports[pid];
+
+ port->lacp_master = 0;
+}
+
+uint8_t port_is_lacp_master(portid_t pid)
+{
+ struct rte_port *port = &ports[pid];
+
+ return port->lacp_master;
+}
+
const uint16_t vlan_tags[] = {
0, 1, 2, 3, 4, 5, 6, 7,
8, 9, 10, 11, 12, 13, 14, 15,
@@ -170,6 +170,7 @@ struct rte_port {
struct ether_addr *mc_addr_pool; /**< pool of multicast addrs */
uint32_t mc_addr_nb; /**< nb. of addr. in mc_addr_pool */
uint8_t slave_flag; /**< bonding slave port */
+ uint8_t lacp_master;/**< bonding LACP master */
};
extern portid_t __rte_unused
@@ -542,6 +543,9 @@ void init_port_config(void);
void set_port_slave_flag(portid_t slave_pid);
void clear_port_slave_flag(portid_t slave_pid);
uint8_t port_is_bonding_slave(portid_t slave_pid);
+void set_port_lacp_master(portid_t pid);
+void clear_port_lacp_master(portid_t pid);
+uint8_t port_is_lacp_master(portid_t pid);
int init_port_dcb_config(portid_t pid, enum dcb_mode_enable dcb_mode,
enum rte_eth_nb_tcs num_tcs,