Hello,
I have tried an application on T4240rdb. I applied fmc for steering forwarding packets to just one core. After I applied fmc, packets going through on same core as I expected; but there is a problem there. If I try to ping from/to other ethernet interfaces packets are timed out, I think they are dropped by fman. When I checked the cpu stats from "top" application, just one core %0 idle, other 23 cores almost free like %99 idle. How can it be possible ?
Any help will be appreciated.
Thanks.
Hello ömer faruk er ,
In order to forward all packets to just one core, you need to modify fmc policy to specify the the queue number binging to the dedicated channel which is allocated to a specified core.
You could refer to the following method to get the relationship between queue and dedicate channel, then refer to DPAA user manual to find the relationship between queue number and the specific core.
oot@localhost:~# cat /sys/devices/platform/soc/soc\:fsl\,dpaa/soc\:fsl\,dpaa\:ethernet\@1/net/fm1-mac2/fqids
Rx error: 391
Rx default: 392
Rx PCD: 14464 - 14591
Tx confirmation (mq): 393 - 456
Tx error: 457
Tx default confirmation: 458
Tx: 459 - 511
Tx: 32768 - 32778
root@localhost:~#
root@localhost:~#
root@localhost:~# echo 14464 > /sys/kernel/debug/qman/query_fq_fields
root@localhost:~# cat /sys/kernel/debug/qman/query_fq_fields
Query FQ Programmable Fields Result fqid 0x3880
orprws: 0
oa: 0
olws: 0
cgid: 3
fq_ctrl:
Prefer in cache
Avoid Blocking
Context-A stashing
Congestion Group Enable
dest_channel: 9
dest_wq: 5
ics_cred: 0
td_mant: 0
td_exp: 0
ctx_b: 0x10a
ctx_a: 0x0
ctx_a_stash_exclusive:
FQ Ctx Stash
Frame Annotation Stash
ctx_a_stash_annotation_cl: 1
ctx_a_stash_data_cl: 2
ctx_a_stash_context_cl: 2
If your problem remains, please provide your fmc file.
Thanks,
Yiping