[dpdk-stable] [EXT] RE: [dpdk-dev] [PATCH 2/2] examples/ip_pipeline: fix to use sock id instead of cpuid

[Jun Yang]

Hi Cristian,
If my recall is correct, the issue comes from following commands:

ip_pipleline running on DPAA1:
#NOTE: rss ip_pipeline is not supported due to the limitation of DPAA1 driver.
1) mkdir -p /mnt/hugepages
2) mount -t hugetlbfs none /mnt/hugepages

3) export DPAA_FMCLESS_MODE=1

4) #firewall ip_pipeline demo:
   ./ip_pipeline -c 0x3 -n 4 -- -s ./dpaa2_examples/firewall.cli
5) #flow ip_pipeline demo:
   ./ip_pipeline -c 0xf -n 4 -- -s ./dpaa2_examples/flow.cli
6) #flow crypto ip_pipeline demo:
   ./ip_pipeline -c 0x7 -n 4 -- -s ./dpaa2_examples/flow_crypto.cli
7) #l2fwd ip_pipeline demo:
   ./ip_pipeline -c 0x3 -n 4 -- -s ./dpaa2_examples/l2fwd.cli
8) #route ip_pipeline demo:
   ./ip_pipeline -c 0x3 -n 4 -- -s ./dpaa2_examples/route.cli
9) #route ecmp ip_pipeline demo:
   ./ip_pipeline -c 0x3 -n 4 -- -s ./dpaa2_examples/route_ecmp.cli

-----Original Message-----
From: Dumitrescu, Cristian [mailto:cristian.dumitrescu at intel.com] 
Sent: Tuesday, February 18, 2020 6:48 PM
To: Jun Yang <jun.yang at nxp.com>; Hemant Agrawal <hemant.agrawal at nxp.com>; dev at dpdk.org
Cc: Stable at dpdk.org
Subject: RE: [EXT] RE: [dpdk-dev] [PATCH 2/2] examples/ip_pipeline: fix to use sock id instead of cpuid

Caution: EXT Email

> -----Original Message-----
> From: Jun Yang <jun.yang at nxp.com>
> Sent: Friday, February 14, 2020 3:19 PM
> To: Dumitrescu, Cristian <cristian.dumitrescu at intel.com>; Hemant 
> Agrawal <hemant.agrawal at nxp.com>; dev at dpdk.org
> Cc: Stable at dpdk.org
> Subject: RE: [EXT] RE: [dpdk-dev] [PATCH 2/2] examples/ip_pipeline: 
> fix to use sock id instead of cpuid
>
> Hi Cristian,
> Per my debug log, the CPU ID is actually DPDK current core ID on NXP 
> ARM64 platform whose socket ID is always 0.
>

Hi Jun,

The params->cpu_id is vendor independent, it is always the CPU socket ID and is consistently used as the NUMA node to allocate memory on. If your system has a single CPU socket, then this parameter should be equal to 0. As clearly documented and stated before, this parameter is never the CPU core ID, which is typically called core_id or lcore_id consistently in the code. Again, not sure where this confusion is coming from.

Regards,
Cristian
-------------- next part --------------
A non-text attachment was scrubbed...
Name: dpaa1_examples.tar.gz
Type: application/x-gzip
Size: 3188 bytes
Desc: dpaa1_examples.tar.gz
URL: <http://mails.dpdk.org/archives/stable/attachments/20200219/acb1dbab/attachment.bin>