[dpdk-users] Does any thing wrong with the DPDK sample l2fwd-jobstats

[dpdk-users] Does any thing wrong with the DPDK sample l2fwd-jobstats

[Caianning]

Hi,
        I noticed that when the pktgen(version 3.0.09) port 1 rx rate drop to half of normal when l2fwd-jobstat shows Port statistics. (while l2fwd dropped packet is 0). And port 0 kept same as normal.
         DPDK version 16.07.0

         My cmd line as following:
         Pktgen:
pktgen -c 0x1ff -n 3 -- -P -m "[1-2:3-4].0,[5-6:7-8].1"
           with port 0 and port 1 started

         lwfwd-jobstats:
         l2fwd-jobstats -c 580000000 -- -p 3

         but the l2fwd does not. Maybe some other processes disturbed the forwarding processing(alarm trig task re-schedule?), and Shall these cores be isolated?

         Thanks.

Re: [dpdk-users] Does any thing wrong with the DPDK sample l2fwd-jobstats

[Wiles, Keith]

Comments inline.
Regards,
Keith

> On Aug 17, 2016, at 7:06 AM, Caianning <caianning@huawei.com> wrote:
> 
> Hi,
>        I noticed that when the pktgen(version 3.0.09) port 1 rx rate drop to half of normal when l2fwd-jobstat shows Port statistics. (while l2fwd dropped packet is 0). And port 0 kept same as normal.
>         DPDK version 16.07.0
> 
>         My cmd line as following:
>         Pktgen:
> pktgen -c 0x1ff -n 3 -- -P -m "[1-2:3-4].0,[5-6:7-8].1"
>           with port 0 and port 1 started

The Pktgen command line looks find to me and I assume the lcores being used are full physical core?

Is the 

> 
>         lwfwd-jobstats:
>         l2fwd-jobstats -c 580000000 -- -p 3

BTW, it maybe easier to see which cores are being used here by using the -l <lcores-numbers> instead of -c. Eg. -l 0-8 and -l 31-32,34 this also assumes the lcores are per physical core.

If you run the tools/cpu_info.py script you will be able to see the lcore/core relationship.

> 
>         but the l2fwd does not. Maybe some other processes disturbed the forwarding processing(alarm trig task re-schedule?), and Shall these cores be isolated?
> 
>         Thanks.
>