Re: [dts] [PATCH 0/2] adapt to CVL NIC

Re: [dts] [PATCH 0/2] adapt to CVL NIC

[Xie, WeiX]

[-- Attachment #1: Type: text/plain, Size: 239 bytes --]

> -----Original Message-----
> From: Xie, WeiX
> Sent: Wednesday, March 24, 2021 7:25 PM
> To: dts@dpdk.org
> Cc: Xie, WeiX <weix.xie@intel.com>
> Subject: [dts][PATCH 0/2] adapt to CVL NIC

Tested-by: Xie Wei <weix.xie@intel.com>

[-- Attachment #2: TestRxTx_Offload.log --]
[-- Type: application/octet-stream, Size: 2126441 bytes --]

[dts] [PATCH 0/2] adapt to CVL NIC

[xiewei]

modify suite and test_plan to support CVL.

xiewei (2):
  tests/rxtx_offload: adapt to CVL NIC
  test_plans/rxtx_offload_test_plan: adapt to CVL NIC

 test_plans/rxtx_offload_test_plan.rst | 29 +++++++++++++++++++++++++++
 tests/TestSuite_rxtx_offload.py       | 23 ++++++++++++++++-----
 2 files changed, 47 insertions(+), 5 deletions(-)

-- 
2.17.1

[dts] [PATCH 1/2] tests/rxtx_offload: adapt to CVL NIC

[xiewei]

for CVL NIC,
a.rxtx_offload/rxoffload_port: disable jumboframe per-queue will change the RSS key, then pkt can be distributed to queues by RSS.
b.rxtx_offload/txoffload_queue: it failed to enable mbuf_fast_free per_queue.

Signed-off-by: xiewei <weix.xie@intel.com>
---
 tests/TestSuite_rxtx_offload.py | 23 ++++++++++++++++++-----
 1 file changed, 18 insertions(+), 5 deletions(-)

diff --git a/tests/TestSuite_rxtx_offload.py b/tests/TestSuite_rxtx_offload.py
index 0e70526f..f7481089 100644
--- a/tests/TestSuite_rxtx_offload.py
+++ b/tests/TestSuite_rxtx_offload.py
@@ -314,7 +314,10 @@ class TestRxTx_Offload(TestCase):
             scanner = re.compile(result_scanner, re.DOTALL)
             m = scanner.search(outstring)
             queue_id = m.group(1)
-            self.verify(int(expect_queue) == int(queue_id), "the offload setting doesn't take effect.")
+            if self.nic in ['columbiaville_25g', 'columbiaville_100g']:
+                self.verify(int(queue_id) in [0, 1, 2, 3], "the offload setting doesn't take effect.")
+            else:
+                self.verify(int(expect_queue) == int(queue_id), "the offload setting doesn't take effect.")
 
     def start_tcpdump(self, rxItf):
 
@@ -346,10 +349,16 @@ class TestRxTx_Offload(TestCase):
         pkt2_queue = self.get_queue_number(self.jumbo_pkt2)
 
         # Failed to disable jumboframe per_queue, foxvillee 2.5g not support
-        if self.nic != 'foxville':
+        if self.nic in ['columbiaville_25g', 'columbiaville_100g']:
             self.dut.send_expect("port stop 0", "testpmd> ")
-            self.dut.send_expect("port 0 rxq %s rx_offload jumbo_frame off" % pkt1_queue, "testpmd> ")
+            for i in range(4):
+                self.dut.send_expect("port 0 rxq %s rx_offload jumbo_frame off" % i, "testpmd> ")
             self.verify_result(self.jumbo_pkt1, 1, pkt1_queue)
+        else:
+            if self.nic != 'foxville':
+                self.dut.send_expect("port stop 0", "testpmd> ")
+                self.dut.send_expect("port 0 rxq %s rx_offload jumbo_frame off" % pkt1_queue, "testpmd> ")
+                self.verify_result(self.jumbo_pkt1, 1, pkt1_queue)
 
         # Succeed to disable jumboframe per_port
         self.dut.send_expect("port stop 0", "testpmd> ")
@@ -698,8 +707,12 @@ class TestRxTx_Offload(TestCase):
         self.dut.send_expect("port 0 txq 1 tx_offload mbuf_fast_free on", "testpmd> ")
         self.dut.send_expect("port 0 txq 2 tx_offload mbuf_fast_free on", "testpmd> ")
         self.dut.send_expect("port 0 txq 3 tx_offload mbuf_fast_free on", "testpmd> ")
-        offload = ["mbuf_fast_free", "mbuf_fast_free", "mbuf_fast_free", "mbuf_fast_free"]
-        self.check_queue_config("tx", offload)
+        if self.nic in ['columbiaville_25g', 'columbiaville_100g']:
+            out = self.dut.send_expect("port start 0", "testpmd> ")
+            self.verify("Fail to configure port 0 tx queues" in out, "mbuf_fast_free can be set by queue.")
+        else:
+            offload = ["mbuf_fast_free", "mbuf_fast_free", "mbuf_fast_free", "mbuf_fast_free"]
+            self.check_queue_config("tx", offload)
 
         # Disable mbuf_fast_free per_queue.
         self.dut.send_expect("port stop 0", "testpmd> ")
-- 
2.17.1

[dts] [PATCH 2/2] test_plans/rxtx_offload_test_plan: adapt to CVL NIC

[xiewei]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset=y, Size: 2090 bytes --]

add CVL NIC supported.

Signed-off-by: xiewei <weix.xie@intel.com>
---
 test_plans/rxtx_offload_test_plan.rst | 29 +++++++++++++++++++++++++++
 1 file changed, 29 insertions(+)

diff --git a/test_plans/rxtx_offload_test_plan.rst b/test_plans/rxtx_offload_test_plan.rst
index 246e1e16..8b034d67 100644
--- a/test_plans/rxtx_offload_test_plan.rst
+++ b/test_plans/rxtx_offload_test_plan.rst
@@ -146,6 +146,21 @@ Test case: Rx offload per-port setting
    The port can be started normally, but the setting doesn't take effect.
    Pkt1 still can be distributed to queue 1.
 
+Note:
+
+But for ice NIC, disable jumboframe per-queue will change the RSS key.
+So if set jumboframe per-queue, the hash value of received packet will be changed, and we can’t judge the test result by queue number.
+So we disable jumboframe per-queue in all the queues, and if the packet still can be received in any queue, the setting doesn't take effect::
+
+    testpmd> port stop 0
+    testpmd> port 0 rxq 0 rx_offload jumbo_frame off
+    testpmd> port 0 rxq 1 rx_offload jumbo_frame off
+    testpmd> port 0 rxq 2 rx_offload jumbo_frame off
+    testpmd> port 0 rxq 3 rx_offload jumbo_frame off
+    testpmd> port start 0
+
+   Pkt1 can be distributed to queues by RSS.
+
 4. Succeed to disable jumboframe per_port::
 
     testpmd> port stop 0
@@ -643,6 +658,20 @@ Test case: FVL Tx offload per-queue setting
 
    The port fwd can be started normally.
 
+Note:
+
+But for ice NIC, it failed to enable mbuf_fast_free per_queue::
+
+    testpmd> port stop 0
+    testpmd> port 0 txq 0 tx_offload mbuf_fast_free on
+    testpmd> port 0 txq 1 tx_offload mbuf_fast_free on
+    testpmd> port 0 txq 2 tx_offload mbuf_fast_free on
+    testpmd> port 0 txq 3 tx_offload mbuf_fast_free on
+    testpmd> port start 0
+    Configuring Port 0 (socket 0)
+    Ethdev port_id=0 tx_queue_id=0, new added offloads 0x10000 must be within per-queue offload capabilities 0x0 in rte_eth_tx_queue_setup()
+    Fail to configure port 0 tx queues
+
 4. Disable mbuf_fast_free per_queue::
 
     testpmd> port stop 0
-- 
2.17.1

Re: [dts] [PATCH 0/2] adapt to CVL NIC

[Peng, Yuan]

-----Original Message-----
From: dts <dts-bounces@dpdk.org> On Behalf Of xiewei
Sent: Wednesday, March 24, 2021 7:25 PM
To: dts@dpdk.org
Cc: Xie, WeiX <weix.xie@intel.com>
Subject: [dts] [PATCH 0/2] adapt to CVL NIC

modify suite and test_plan to support CVL.

xiewei (2):
  tests/rxtx_offload: adapt to CVL NIC
  test_plans/rxtx_offload_test_plan: adapt to CVL NIC

 test_plans/rxtx_offload_test_plan.rst | 29 +++++++++++++++++++++++++++
 tests/TestSuite_rxtx_offload.py       | 23 ++++++++++++++++-----
 2 files changed, 47 insertions(+), 5 deletions(-)

-- 
Acked-by Peng, Yuan <yuan.peng@intel.com>

2.17.1

Re: [dts] [PATCH 0/2] adapt to CVL NIC

[Peng, Yuan]

Acked-by Peng, Yuan <yuan.peng@intel.com>

-----Original Message-----
From: dts <dts-bounces@dpdk.org> On Behalf Of xiewei
Sent: Wednesday, March 24, 2021 7:25 PM
To: dts@dpdk.org
Cc: Xie, WeiX <weix.xie@intel.com>
Subject: [dts] [PATCH 0/2] adapt to CVL NIC

modify suite and test_plan to support CVL.

xiewei (2):
  tests/rxtx_offload: adapt to CVL NIC
  test_plans/rxtx_offload_test_plan: adapt to CVL NIC

 test_plans/rxtx_offload_test_plan.rst | 29 +++++++++++++++++++++++++++
 tests/TestSuite_rxtx_offload.py       | 23 ++++++++++++++++-----
 2 files changed, 47 insertions(+), 5 deletions(-)

-- 
2.17.1

Re: [dts] [PATCH 2/2] test_plans/rxtx_offload_test_plan: adapt to CVL NIC

[Tu, Lijuan]

> -----Original Message-----
> From: dts <dts-bounces@dpdk.org> On Behalf Of xiewei
> Sent: 2021年3月24日 19:25
> To: dts@dpdk.org
> Cc: Xie, WeiX <weix.xie@intel.com>
> Subject: [dts] [PATCH 2/2] test_plans/rxtx_offload_test_plan: adapt to CVL NIC
> 
> add CVL NIC supported.
> 
> Signed-off-by: xiewei <weix.xie@intel.com>
> ---
>  test_plans/rxtx_offload_test_plan.rst | 29 +++++++++++++++++++++++++++
>  1 file changed, 29 insertions(+)
> 
> diff --git a/test_plans/rxtx_offload_test_plan.rst
> b/test_plans/rxtx_offload_test_plan.rst
> index 246e1e16..8b034d67 100644
> --- a/test_plans/rxtx_offload_test_plan.rst
> +++ b/test_plans/rxtx_offload_test_plan.rst
> @@ -146,6 +146,21 @@ Test case: Rx offload per-port setting
>     The port can be started normally, but the setting doesn't take effect.
>     Pkt1 still can be distributed to queue 1.
> 
> +Note:
> +
> +But for ice NIC, disable jumboframe per-queue will change the RSS key.
> +So if set jumboframe per-queue, the hash value of received packet will be
> changed, and we can’t judge the test result by queue number.
> +So we disable jumboframe per-queue in all the queues, and if the packet still
> can be received in any queue, the setting doesn't take effect::
> +
> +    testpmd> port stop 0
> +    testpmd> port 0 rxq 0 rx_offload jumbo_frame off
> +    testpmd> port 0 rxq 1 rx_offload jumbo_frame off
> +    testpmd> port 0 rxq 2 rx_offload jumbo_frame off
> +    testpmd> port 0 rxq 3 rx_offload jumbo_frame off
> +    testpmd> port start 0
> +
> +   Pkt1 can be distributed to queues by RSS.
> +

How about use a fixed RSS key., it will keep cases consistency in different cards.

>  4. Succeed to disable jumboframe per_port::
> 
>      testpmd> port stop 0
> @@ -643,6 +658,20 @@ Test case: FVL Tx offload per-queue setting
> 
>     The port fwd can be started normally.
> 
> +Note:
> +
> +But for ice NIC, it failed to enable mbuf_fast_free per_queue::
> +
> +    testpmd> port stop 0
> +    testpmd> port 0 txq 0 tx_offload mbuf_fast_free on
> +    testpmd> port 0 txq 1 tx_offload mbuf_fast_free on
> +    testpmd> port 0 txq 2 tx_offload mbuf_fast_free on
> +    testpmd> port 0 txq 3 tx_offload mbuf_fast_free on
> +    testpmd> port start 0
> +    Configuring Port 0 (socket 0)
> +    Ethdev port_id=0 tx_queue_id=0, new added offloads 0x10000 must be
> within per-queue offload capabilities 0x0 in rte_eth_tx_queue_setup()
> +    Fail to configure port 0 tx queues
> +
>  4. Disable mbuf_fast_free per_queue::
> 
>      testpmd> port stop 0
> --
> 2.17.1