diff mbox series

app/testpmd: fix MTU after device configure

Message ID 20201113114400.3356527-1-ferruh.yigit@intel.com (mailing list archive)
State Accepted, archived
Delegated to: Ferruh Yigit
Headers
Series app/testpmd: fix MTU after device configure |

Checks

Context Check Description
ci/checkpatch success coding style OK
ci/Intel-compilation success Compilation OK
ci/iol-broadcom-Performance success Performance Testing PASS
ci/iol-broadcom-Functional success Functional Testing PASS
ci/iol-intel-Functional fail Functional Testing issues
ci/iol-testing success Testing PASS
ci/iol-intel-Performance success Performance Testing PASS
ci/travis-robot success Travis build: passed
ci/iol-mellanox-Performance success Performance Testing PASS

Commit Message

Ferruh Yigit Nov. 13, 2020, 11:44 a.m. UTC 
  In 'rte_eth_dev_configure()', if 'DEV_RX_OFFLOAD_JUMBO_FRAME' is not set
the max frame size is limited to 'RTE_ETHER_MAX_LEN' (1518).
This is mistake because for the PMDs that has frame size bigger than
"RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN" (18 bytes), the MTU becomes
less than 1500, causing a valid frame with 1500 bytes payload to be
dropped.

Since 'rte_eth_dev_set_mtu()' works as expected, it is called after
'rte_eth_dev_configure()' to fix the MTU.
It may look redundant to set MTU after 'rte_eth_dev_configure()', both
with default values, but it is not, the resulting MTU config can be
different in the device based on frame overhead of the PMD.

And instead of setting the MTU to default value, it is first get via
'rte_eth_dev_get_mtu()' and set again, this is to cover cases MTU
changed from testpmd command line.

'rte_eth_dev_set_mtu()', '-ENOTSUP' error is ignored to prevent
irrelevant warning messages for the virtual PMDs.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
Reviewed-by: Qi Zhang <qi.z.zhang@intel.com>
---
Cc: Steve Yang <stevex.yang@intel.com>
Cc: Thomas Monjalon <thomas@monjalon.net>
Cc: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Cc: Konstantin Ananyev <konstantin.ananyev@intel.com>
Cc: Olivier Matz <olivier.matz@6wind.com>
Cc: Lance Richardson <lance.richardson@broadcom.com>
---
 app/test-pmd/testpmd.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

Comments

Andrew Rybchenko Nov. 13, 2020, 2:53 p.m. UTC | #1 
On 11/13/20 2:44 PM, Ferruh Yigit wrote:
> In 'rte_eth_dev_configure()', if 'DEV_RX_OFFLOAD_JUMBO_FRAME' is not set
> the max frame size is limited to 'RTE_ETHER_MAX_LEN' (1518).
> This is mistake because for the PMDs that has frame size bigger than
> "RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN" (18 bytes), the MTU becomes
> less than 1500, causing a valid frame with 1500 bytes payload to be
> dropped.
> 
> Since 'rte_eth_dev_set_mtu()' works as expected, it is called after
> 'rte_eth_dev_configure()' to fix the MTU.
> It may look redundant to set MTU after 'rte_eth_dev_configure()', both
> with default values, but it is not, the resulting MTU config can be
> different in the device based on frame overhead of the PMD.
> 
> And instead of setting the MTU to default value, it is first get via
> 'rte_eth_dev_get_mtu()' and set again, this is to cover cases MTU
> changed from testpmd command line.
> 
> 'rte_eth_dev_set_mtu()', '-ENOTSUP' error is ignored to prevent
> irrelevant warning messages for the virtual PMDs.
> 
> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
> Reviewed-by: Qi Zhang <qi.z.zhang@intel.com>

Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
Tested-by: Igor Romanov <igor.romanov@oktetlabs.ru>
Ferruh Yigit Nov. 13, 2020, 4:01 p.m. UTC | #2 
On 11/13/2020 2:53 PM, Andrew Rybchenko wrote:
> On 11/13/20 2:44 PM, Ferruh Yigit wrote:
>> In 'rte_eth_dev_configure()', if 'DEV_RX_OFFLOAD_JUMBO_FRAME' is not set
>> the max frame size is limited to 'RTE_ETHER_MAX_LEN' (1518).
>> This is mistake because for the PMDs that has frame size bigger than
>> "RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN" (18 bytes), the MTU becomes
>> less than 1500, causing a valid frame with 1500 bytes payload to be
>> dropped.
>>
>> Since 'rte_eth_dev_set_mtu()' works as expected, it is called after
>> 'rte_eth_dev_configure()' to fix the MTU.
>> It may look redundant to set MTU after 'rte_eth_dev_configure()', both
>> with default values, but it is not, the resulting MTU config can be
>> different in the device based on frame overhead of the PMD.
>>
>> And instead of setting the MTU to default value, it is first get via
>> 'rte_eth_dev_get_mtu()' and set again, this is to cover cases MTU
>> changed from testpmd command line.
>>
>> 'rte_eth_dev_set_mtu()', '-ENOTSUP' error is ignored to prevent
>> irrelevant warning messages for the virtual PMDs.
>>
>> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
>> Reviewed-by: Qi Zhang <qi.z.zhang@intel.com>
> 
> Acked-by: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
> Tested-by: Igor Romanov <igor.romanov@oktetlabs.ru>
> 

Applied to dpdk-next-net/main, thanks.
Ferruh Yigit Nov. 16, 2020, 6:50 p.m. UTC | #3 
On 11/13/2020 11:44 AM, Ferruh Yigit wrote:
> In 'rte_eth_dev_configure()', if 'DEV_RX_OFFLOAD_JUMBO_FRAME' is not set
> the max frame size is limited to 'RTE_ETHER_MAX_LEN' (1518).
> This is mistake because for the PMDs that has frame size bigger than
> "RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN" (18 bytes), the MTU becomes
> less than 1500, causing a valid frame with 1500 bytes payload to be
> dropped.
> 
> Since 'rte_eth_dev_set_mtu()' works as expected, it is called after
> 'rte_eth_dev_configure()' to fix the MTU.
> It may look redundant to set MTU after 'rte_eth_dev_configure()', both
> with default values, but it is not, the resulting MTU config can be
> different in the device based on frame overhead of the PMD.
> 
> And instead of setting the MTU to default value, it is first get via
> 'rte_eth_dev_get_mtu()' and set again, this is to cover cases MTU
> changed from testpmd command line.
> 
> 'rte_eth_dev_set_mtu()', '-ENOTSUP' error is ignored to prevent
> irrelevant warning messages for the virtual PMDs.
> 
> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
> Reviewed-by: Qi Zhang <qi.z.zhang@intel.com>
> ---
> Cc: Steve Yang <stevex.yang@intel.com>
> Cc: Thomas Monjalon <thomas@monjalon.net>
> Cc: Andrew Rybchenko <andrew.rybchenko@oktetlabs.ru>
> Cc: Konstantin Ananyev <konstantin.ananyev@intel.com>
> Cc: Olivier Matz <olivier.matz@6wind.com>
> Cc: Lance Richardson <lance.richardson@broadcom.com>
> ---
>   app/test-pmd/testpmd.c | 19 +++++++++++++++++++
>   1 file changed, 19 insertions(+)
> 
> diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c
> index 33fc0fddf5..48e9647fc7 100644
> --- a/app/test-pmd/testpmd.c
> +++ b/app/test-pmd/testpmd.c
> @@ -2537,6 +2537,8 @@ start_port(portid_t pid)
>   		}
>   
>   		if (port->need_reconfig > 0) {
> +			uint16_t mtu = RTE_ETHER_MTU;
> +
>   			port->need_reconfig = 0;
>   
>   			if (flow_isolate_all) {
> @@ -2570,6 +2572,23 @@ start_port(portid_t pid)
>   				port->need_reconfig = 1;
>   				return -1;
>   			}
> +
> +			/*
> +			 * Workaround for rte_eth_dev_configure(), max_rx_pkt_len
> +			 * set MTU wrong for the PMDs that have frame overhead
> +			 * bigger than RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN.
> +			 * For a PMD that has 26 bytes overhead, rte_eth_dev_configure()
> +			 * can set MTU to max 1492, not to expected 1500 bytes.
> +			 * Using rte_eth_dev_set_mtu() to be able to set MTU correctly,
> +			 * default MTU value is 1500.
> +			 */
> +			diag = rte_eth_dev_get_mtu(pi, &mtu);
> +			if (diag)
> +				printf("Failed to get MTU for port %d\n", pi);
> +			diag = rte_eth_dev_set_mtu(pi, mtu);
> +			if (diag != 0 && diag != -ENOTSUP)
> +				printf("Failed to set MTU to %u for port %d\n",
> +						mtu, pi);
>   		}
>   		if (port->need_reconfig_queues > 0) {
>   			port->need_reconfig_queues = 0;
> 

@David highlighted that 'scatter' tests are failing in the lab with this commit,
https://lab.dpdk.org/results/dashboard/patchsets/14492/

With above commit only 'mtu' is taken into account, so in testpmd both 
"--max-pkt-len=N" parameter and "port config all max-pkt-len #" command are no 
more working as expected. This seems the reason of the failure.

Technically it is possible to fix dts testcase by adding following commands:
port stop all
port config mtu 0 9000
port start all

But, there may be other side affects from "max-pkt-len" is not working in 
testpmd as expected. Reverting this one too can be safest option.

For now we need to live with the issue this patch is fixing, hopefully we can 
fix it in next release with fixing all testpmd, ethdev and drivers, there is a 
question about ethdev change if it will be an ABI break or not, we will see it.

And there is a longer term target to deprecate 'max_rx_pkt_len' and 'mtu' to 
unify them:
https://patches.dpdk.org/patch/81591/
Thomas Monjalon Nov. 16, 2020, 8:24 p.m. UTC | #4 
16/11/2020 19:50, Ferruh Yigit:
> On 11/13/2020 11:44 AM, Ferruh Yigit wrote:
> > In 'rte_eth_dev_configure()', if 'DEV_RX_OFFLOAD_JUMBO_FRAME' is not set
> > the max frame size is limited to 'RTE_ETHER_MAX_LEN' (1518).
> > This is mistake because for the PMDs that has frame size bigger than
> > "RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN" (18 bytes), the MTU becomes
> > less than 1500, causing a valid frame with 1500 bytes payload to be
> > dropped.
> > 
> > Since 'rte_eth_dev_set_mtu()' works as expected, it is called after
> > 'rte_eth_dev_configure()' to fix the MTU.
> > It may look redundant to set MTU after 'rte_eth_dev_configure()', both
> > with default values, but it is not, the resulting MTU config can be
> > different in the device based on frame overhead of the PMD.
> > 
> > And instead of setting the MTU to default value, it is first get via
> > 'rte_eth_dev_get_mtu()' and set again, this is to cover cases MTU
> > changed from testpmd command line.
> > 
> > 'rte_eth_dev_set_mtu()', '-ENOTSUP' error is ignored to prevent
> > irrelevant warning messages for the virtual PMDs.
> > 
> > Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
> > Reviewed-by: Qi Zhang <qi.z.zhang@intel.com>
> > ---
> 
> @David highlighted that 'scatter' tests are failing in the lab with this commit,
> https://lab.dpdk.org/results/dashboard/patchsets/14492/

To be more precise, this error showed up in the CI even before
the patch was merged: https://patches.dpdk.org/patch/84132/
We should be more careful with the CI results before merging.

> With above commit only 'mtu' is taken into account, so in testpmd both 
> "--max-pkt-len=N" parameter and "port config all max-pkt-len #" command are no 
> more working as expected. This seems the reason of the failure.
> 
> Technically it is possible to fix dts testcase by adding following commands:
> port stop all
> port config mtu 0 9000
> port start all
> 
> But, there may be other side affects from "max-pkt-len" is not working in 
> testpmd as expected. Reverting this one too can be safest option.

Next fix will be the right one ;)
diff mbox series

Patch

diff --git a/app/test-pmd/testpmd.c b/app/test-pmd/testpmd.c
index 33fc0fddf5..48e9647fc7 100644
--- a/app/test-pmd/testpmd.c
+++ b/app/test-pmd/testpmd.c
@@ -2537,6 +2537,8 @@  start_port(portid_t pid)
 		}
 
 		if (port->need_reconfig > 0) {
+			uint16_t mtu = RTE_ETHER_MTU;
+
 			port->need_reconfig = 0;
 
 			if (flow_isolate_all) {
@@ -2570,6 +2572,23 @@  start_port(portid_t pid)
 				port->need_reconfig = 1;
 				return -1;
 			}
+
+			/*
+			 * Workaround for rte_eth_dev_configure(), max_rx_pkt_len
+			 * set MTU wrong for the PMDs that have frame overhead
+			 * bigger than RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN.
+			 * For a PMD that has 26 bytes overhead, rte_eth_dev_configure()
+			 * can set MTU to max 1492, not to expected 1500 bytes.
+			 * Using rte_eth_dev_set_mtu() to be able to set MTU correctly,
+			 * default MTU value is 1500.
+			 */
+			diag = rte_eth_dev_get_mtu(pi, &mtu);
+			if (diag)
+				printf("Failed to get MTU for port %d\n", pi);
+			diag = rte_eth_dev_set_mtu(pi, mtu);
+			if (diag != 0 && diag != -ENOTSUP)
+				printf("Failed to set MTU to %u for port %d\n",
+						mtu, pi);
 		}
 		if (port->need_reconfig_queues > 0) {
 			port->need_reconfig_queues = 0;