Re: unable to capture packets

["Van Haaren, Harry" <harry.van.haaren@intel.com>]

[-- Attachment #1: Type: text/plain, Size: 2782 bytes --]


> From: Lokesh Chakka <lvenkatakumarchakka@gmail.com>
> Sent: Wednesday, October 9, 2024 1:15 PM
> To: Stephen Hemminger <stephen@networkplumber.org>
> Cc: Pathak, Pravin <pravin.pathak@intel.com>; Bing Zhao <bingz@nvidia.com>; users <users@dpdk.org>
> Subject: Re: unable to capture packets
>
> hi,

Hi Chakka,

Please don't "top post" on mailing lists; a reply "inline" with context is a lot easier to follow for all current (and future!) readers.

> did certain modifications as per your suggestions. still the same problem. not able to capture any packets....!!!

Have you tried to run the DPDK example applications? Specifically, the skeleton/basicfwd.c has a "known good" setup routine,
and forwards packets on a single core. Perhaps it is a good place to compare your setup code to, as it its known working.

> I replaced 4096 with 512. rte_pktmbuf_pool_create is giving an error. for the time being i've left it as 4K only.
> I feel it should not be a problem.

There is a problem somewhere - and currently in your code it is not root caused. This is a time to re-check
code that "seemed ok" before, because somewhere something is not behaving as you expect.

> PFA for the revised code.

Attaching code to mailing lists is not an easy/good way to review; perhaps create a git repo and push the code there?

Then provide a link, and future patches could easily show what improvements/fixes occur. (If you email another
version of the "pmd.c" file, readers cannot know what changes were made -> leads to lots of duplication of review effort)


> Output is as follows :
>
> ==============================================================================================
> EAL: Detected CPU lcores: 40
> EAL: Detected NUMA nodes: 1
> EAL: Detected shared linkage of DPDK
> EAL: Multi-process socket /var/run/dpdk/rte/mp_socket
> EAL: Selected IOVA mode 'VA'
> EAL: VFIO support initialized
> EAL: Using IOMMU type 1 (Type 1)
> ice_dev_init(): Failed to read device serial number
>
> ice_load_pkg_type(): Active package is: 1.3.39.0, ICE OS Default Package (double VLAN mode)
> main 53 port id: 0 num_of_pkts_per_queue: 4096
> ice_set_rx_function(): Using AVX2 Vector Rx (port 0).
> Received Interrupt Signal SIGINT (2). Exiting...
> main 98 time : 4 total pkts rcvd: 0 bandwidth: 0
> ==============================================================================================

<snip> previous conversation, as top-posted answers, and discussion was hard to follow.

It seems "num_of_pkts_per_queue" is used everywhere (mempool size, mbuf batch array size, rx_burst size, etc)
This is almost certainly impacting things somehow. Review skeleton/basicfwd.c for better sizes/values.

Hope that helps! Regards, -Harry



[-- Attachment #2: Type: text/html, Size: 13098 bytes --]