[dpdk-dev] [PKTGEN] rte_kni.ko with lo_mode=lo_mode_ring

[dpdk-dev] [PKTGEN] rte_kni.ko with lo_mode=lo_mode_ring

[Mussar, Gary]

The pktgen docs state that the rte_kni.ko should be loaded with lo_mode=lo_mode_ring however the source in dpdk master does not appear to understand this value. This wasn't an issue in the past since the kni code would simply disable lo_mode if you passed in an unknown value.

The Ubuntu 4.4.0-36-generic kernel no longer loads the module if the passed value is not in the table of known values (the 4.4.0-34-generic kernel would still load the module).

I have 2 questions:
1)
Why does pktgen specify using a value for lo_mode that the source does not support? Is lo_mode_ring something that is coming but not yet submitted?

2)
Is the Ubuntu kernel being over aggressive in parameter checking or is this a "good" thing? If it is good, then should the pktgen docs be indicating that an unsupported value be used for lo_mode?

Gary

Re: [dpdk-dev] rte_kni.ko with lo_mode=lo_mode_ring

[Wiles, Keith]

HI Gary,

Regards,
Keith

> On Sep 2, 2016, at 8:39 AM, Mussar, Gary <gmussar@ciena.com> wrote:
> 
> The pktgen docs state that the rte_kni.ko should be loaded with lo_mode=lo_mode_ring however the source in dpdk master does not appear to understand this value. This wasn't an issue in the past since the kni code would simply disable lo_mode if you passed in an unknown value.
> 
> The Ubuntu 4.4.0-36-generic kernel no longer loads the module if the passed value is not in the table of known values (the 4.4.0-34-generic kernel would still load the module).
> 
> I have 2 questions:
> 1)
> Why does pktgen specify using a value for lo_mode that the source does not support? Is lo_mode_ring something that is coming but not yet submitted?

The KNI support and kernel module is more of a DPDK issue then Pktgen, which is the application on top of DPDK. I do not know the reason lo_mode_ring is not working correctly. Look in the Maintainers file and see who maintains the KNI code.

> 
> 2)
> Is the Ubuntu kernel being over aggressive in parameter checking or is this a "good" thing? If it is good, then should the pktgen docs be indicating that an unsupported value be used for lo_mode?

I assume the Ubuntu Kernel is attempting to protect itself. Each time the kernel is updated it is a good think to recompile DPDK and the KNI kernel module or the module may not load because of a version mis-match.

> 
> Gary

Re: [dpdk-dev] rte_kni.ko with lo_mode=lo_mode_ring

[Ferruh Yigit]

On 9/2/2016 3:16 PM, Wiles, Keith wrote:
> HI Gary,
> 
> Regards,
> Keith
> 
>> On Sep 2, 2016, at 8:39 AM, Mussar, Gary <gmussar@ciena.com> wrote:
>>
>> The pktgen docs state that the rte_kni.ko should be loaded with lo_mode=lo_mode_ring however the source in dpdk master does not appear to understand this value. This wasn't an issue in the past since the kni code would simply disable lo_mode if you passed in an unknown value.
>>
>> The Ubuntu 4.4.0-36-generic kernel no longer loads the module if the passed value is not in the table of known values (the 4.4.0-34-generic kernel would still load the module).
>>
>> I have 2 questions:
>> 1)
>> Why does pktgen specify using a value for lo_mode that the source does not support? Is lo_mode_ring something that is coming but not yet submitted?
> 
> The KNI support and kernel module is more of a DPDK issue then Pktgen, which is the application on top of DPDK. I do not know the reason lo_mode_ring is not working correctly. Look in the Maintainers file and see who maintains the KNI code.

"lo_mode_ring" is not supported loopback mode, supported ones are:
lo_mode_none
lo_mode_fifo
lo_mode_fifo_skb

Please point the document with incorrect usage.

> 
>>
>> 2)
>> Is the Ubuntu kernel being over aggressive in parameter checking or is this a "good" thing? If it is good, then should the pktgen docs be indicating that an unsupported value be used for lo_mode?
> 
> I assume the Ubuntu Kernel is attempting to protect itself. Each time the kernel is updated it is a good think to recompile DPDK and the KNI kernel module or the module may not load because of a version mis-match.

Yes this looks like the issue,
instead of kernel version or parameter protection, kernel-module version
mismatch. Please re-compile kni module.

> 
>>
>> Gary
> 

Re: [dpdk-dev] rte_kni.ko with lo_mode=lo_mode_ring

[Mussar, Gary]

> -----Original Message-----
> From: Ferruh Yigit [mailto:ferruh.yigit@intel.com]
> Sent: Friday, September 02, 2016 10:55 AM
> To: Wiles, Keith; Mussar, Gary
> Cc: dev@dpdk.org
> Subject: Re: [dpdk-dev] rte_kni.ko with lo_mode=lo_mode_ring
> 
> On 9/2/2016 3:16 PM, Wiles, Keith wrote:
> > HI Gary,
> >
> > Regards,
> > Keith
> >
> >> On Sep 2, 2016, at 8:39 AM, Mussar, Gary <gmussar@ciena.com> wrote:
> >>
> >> The pktgen docs state that the rte_kni.ko should be loaded with
> lo_mode=lo_mode_ring however the source in dpdk master does not appear to
> understand this value. This wasn't an issue in the past since the kni code would
> simply disable lo_mode if you passed in an unknown value.
> >>
> >> The Ubuntu 4.4.0-36-generic kernel no longer loads the module if the passed
> value is not in the table of known values (the 4.4.0-34-generic kernel would still
> load the module).
> >>
> >> I have 2 questions:
> >> 1)
> >> Why does pktgen specify using a value for lo_mode that the source does not
> support? Is lo_mode_ring something that is coming but not yet submitted?
> >
> > The KNI support and kernel module is more of a DPDK issue then Pktgen, which
> is the application on top of DPDK. I do not know the reason lo_mode_ring is not
> working correctly. Look in the Maintainers file and see who maintains the KNI
> code.
> 
> "lo_mode_ring" is not supported loopback mode, supported ones are:
> lo_mode_none
> lo_mode_fifo
> lo_mode_fifo_skb
> 
> Please point the document with incorrect usage.

The files with the incorrect usage are:
pktgen-dpdk/README.md
pktgen-dpdk/setup.sh

> 
> >
> >>
> >> 2)
> >> Is the Ubuntu kernel being over aggressive in parameter checking or is this a
> "good" thing? If it is good, then should the pktgen docs be indicating that an
> unsupported value be used for lo_mode?
> >
> > I assume the Ubuntu Kernel is attempting to protect itself. Each time the
> kernel is updated it is a good think to recompile DPDK and the KNI kernel module
> or the module may not load because of a version mis-match.
> 
> Yes this looks like the issue,
> instead of kernel version or parameter protection, kernel-module version
> mismatch. Please re-compile kni module.

Recompiling the kni module makes it happier. Thanks.

> 
> >
> >>
> >> Gary
> >
>