[dts] dts can not support one more different nics test in one configuration file

[dts] dts can not support one more different nics test in one configuration file

[Tang, HaifengX]

Hi all:
Recently , I found that our dts  cannot support  one more different nics  test  in one configuration file.
Such as the following configuration:

[Execution1]
crbs=10.239.128.128
drivername=igb_uio
test_suites=
    pmd,
    ipfrag,

targets=
    x86_64-native-linuxapp-gcc
parameters=nic_type=kawela_4:func=true

[Execution2]
crbs=10.239.128.128
drivername=igb_uio
test_suites=
    pmd,
    ipfrag,
targets=
    x86_64-native-linuxapp-gcc
parameters=nic_type=springville:func=true
Thanks,
Haifeng

Re: [dts] dts can not support one more different nics test in one configuration file

[Liu, Yong]

Haifeng, thanks for your report. We will tracking this issue and it should be fixed soon.

> -----Original Message-----
> From: dts [mailto:dts-bounces@dpdk.org] On Behalf Of Tang, HaifengX
> Sent: Thursday, January 22, 2015 3:22 PM
> To: dts@dpdk.org
> Subject: [dts] dts can not support one more different nics test in one
> configuration file
> 
> Hi all:
> Recently , I found that our dts  cannot support  one more different nics
> test  in one configuration file.
> Such as the following configuration:
> 
> [Execution1]
> crbs=10.239.128.128
> drivername=igb_uio
> test_suites=
>     pmd,
>     ipfrag,
> 
> targets=
>     x86_64-native-linuxapp-gcc
> parameters=nic_type=kawela_4:func=true
> 
> [Execution2]
> crbs=10.239.128.128
> drivername=igb_uio
> test_suites=
>     pmd,
>     ipfrag,
> targets=
>     x86_64-native-linuxapp-gcc
> parameters=nic_type=springville:func=true
> Thanks,
> Haifeng

Re: [dts] dts can not support one more different nics test in one configuration file

[Qiu, Michael]

On 1/22/2015 3:23 PM, Tang, HaifengX wrote:
> Hi all:
> Recently , I found that our dts  cannot support  one more different nics  test  in one configuration file.
> Such as the following configuration:
>
> [Execution1]
> crbs=10.239.128.128

You'd better to remove this private info in public maillist :)

> drivername=igb_uio
> test_suites=
>     pmd,
>     ipfrag,
>
> targets=
>     x86_64-native-linuxapp-gcc
> parameters=nic_type=kawela_4:func=true
>
> [Execution2]
> crbs=10.239.128.128

Here

> drivername=igb_uio
> test_suites=
>     pmd,
>     ipfrag,
> targets=
>     x86_64-native-linuxapp-gcc
> parameters=nic_type=springville:func=true

BTW, you'd better to supply the error log for others to debug.

Thanks,
Michael
> Thanks,
> Haifeng
>
>

Re: [dts] dts can not support one more different nics test in one configuration file

[Zhang, XiaonanX]

Hi all,
   When I tested l3fwd performance on the fortville_spirit and fortville_eagle nic on Federal 21 and Federal20,
   I also met this problem.
   I think the question is about framework  restore_interfaces function of framework crb.py file and 
   I have been modified for fortvill testing due to Fedora20 kernel has been supported i40e driver.
   Suggestion:
   Modifiy /framework/crb.py:

   def restore_interfaces(self):
        """
        Restore Linux interfaces.
        """
        if dts.drivername == "vfio-pci":
            self.send_expect("rmmod vfio_iommu_type1", "# ", 10)
            self.send_expect("rmmod vfio_pci", "# ", 10)
            self.send_expect("rmmod vfio", "# ", 10)
        else:
            self.send_expect("rmmod igb_uio", "# ", 10)
        self.send_expect("modprobe igb", "# ", 20)
        self.send_expect("modprobe ixgbe", "# ", 20)
        self.send_expect("modprobe e1000e", "# ", 20)
        self.send_expect("modprobe e1000", "# ", 20)
        self.send_expect("modprobe virtio_net", "# ", 20)
+       self.send_expect("`insmod /usr/lib/modules/`uname -r`/kernel/drivers/net/ethernet/intel/i40e/i40e.ko`", "# ", 50)

   It has been verified in our testing performance l3fwd, That's ok.
   But, If we do this We found Fortville testing only run one case, for example L3fwd.

   If we run other cases in execution.cfg, such as pmd port map cannot map always -1.
   

Best Regards
Xiaonan 

   

-----Original Message-----
From: dts [mailto:dts-bounces@dpdk.org] On Behalf Of Qiu, Michael
Sent: Thursday, January 22, 2015 3:44 PM
To: Tang, HaifengX; dts@dpdk.org
Subject: Re: [dts] dts can not support one more different nics test in one configuration file

On 1/22/2015 3:23 PM, Tang, HaifengX wrote:
> Hi all:
> Recently , I found that our dts  cannot support  one more different nics  test  in one configuration file.
> Such as the following configuration:
>
> [Execution1]
> crbs=10.239.128.128

You'd better to remove this private info in public maillist :)

> drivername=igb_uio
> test_suites=
>     pmd,
>     ipfrag,
>
> targets=
>     x86_64-native-linuxapp-gcc
> parameters=nic_type=kawela_4:func=true
>
> [Execution2]
> crbs=10.239.128.128

Here

> drivername=igb_uio
> test_suites=
>     pmd,
>     ipfrag,
> targets=
>     x86_64-native-linuxapp-gcc
> parameters=nic_type=springville:func=true

BTW, you'd better to supply the error log for others to debug.

Thanks,
Michael
> Thanks,
> Haifeng
>
>

Re: [dts] dts can not support one more different nics test in one configuration file

[Qiu, Michael]

On 1/22/2015 6:31 PM, Zhang, XiaonanX wrote:
> Hi all,
>    When I tested l3fwd performance on the fortville_spirit and fortville_eagle nic on Federal 21 and Federal20,
>    I also met this problem.
>    I think the question is about framework  restore_interfaces function of framework crb.py file and 
>    I have been modified for fortvill testing due to Fedora20 kernel has been supported i40e driver.
>    Suggestion:
>    Modifiy /framework/crb.py:
>
>    def restore_interfaces(self):
>         """
>         Restore Linux interfaces.
>         """
>         if dts.drivername == "vfio-pci":
>             self.send_expect("rmmod vfio_iommu_type1", "# ", 10)
>             self.send_expect("rmmod vfio_pci", "# ", 10)
>             self.send_expect("rmmod vfio", "# ", 10)
>         else:
>             self.send_expect("rmmod igb_uio", "# ", 10)
>         self.send_expect("modprobe igb", "# ", 20)
>         self.send_expect("modprobe ixgbe", "# ", 20)
>         self.send_expect("modprobe e1000e", "# ", 20)
>         self.send_expect("modprobe e1000", "# ", 20)
>         self.send_expect("modprobe virtio_net", "# ", 20)
> +       self.send_expect("`insmod /usr/lib/modules/`uname -r`/kernel/drivers/net/ethernet/intel/i40e/i40e.ko`", "# ", 50)

If have supported, you should just exec "modprobe i40e" .

Also you need to send out as a patch.

Thanks,
Michael
>    It has been verified in our testing performance l3fwd, That's ok.
>    But, If we do this We found Fortville testing only run one case, for example L3fwd.
>
>    If we run other cases in execution.cfg, such as pmd port map cannot map always -1.
>    
>
> Best Regards
> Xiaonan 
>
>    
>
> -----Original Message-----
> From: dts [mailto:dts-bounces@dpdk.org] On Behalf Of Qiu, Michael
> Sent: Thursday, January 22, 2015 3:44 PM
> To: Tang, HaifengX; dts@dpdk.org
> Subject: Re: [dts] dts can not support one more different nics test in one configuration file
>
> On 1/22/2015 3:23 PM, Tang, HaifengX wrote:
>> Hi all:
>> Recently , I found that our dts  cannot support  one more different nics  test  in one configuration file.
>> Such as the following configuration:
>>
>> [Execution1]
>> crbs=10.239.128.128
> You'd better to remove this private info in public maillist :)
>
>> drivername=igb_uio
>> test_suites=
>>     pmd,
>>     ipfrag,
>>
>> targets=
>>     x86_64-native-linuxapp-gcc
>> parameters=nic_type=kawela_4:func=true
>>
>> [Execution2]
>> crbs=10.239.128.128
> Here
>
>> drivername=igb_uio
>> test_suites=
>>     pmd,
>>     ipfrag,
>> targets=
>>     x86_64-native-linuxapp-gcc
>> parameters=nic_type=springville:func=true
> BTW, you'd better to supply the error log for others to debug.
>
> Thanks,
> Michael
>> Thanks,
>> Haifeng
>>
>>
>