Re: [net-next v3 2/2] net: sched: support hash/classid/cpuid selecting tx queue

On Sat, Dec 11, 2021 at 6:34 PM Tonghao Zhang [off-list ref] wrote:

On Sun, Dec 12, 2021 at 10:19 AM Cong Wang [off-list ref] wrote:

quoted

On Thu, Dec 9, 2021 at 6:36 PM [off-list ref] wrote:

quoted

From: Tonghao Zhang <redacted>

This patch allows users to select queue_mapping, range
from A to B. And users can use skb-hash, cgroup classid
and cpuid to select Tx queues. Then we can load balance
packets from A to B queue. The range is an unsigned 16bit
value in decimal format.

$ tc filter ... action skbedit queue_mapping hash-type normal A B

"skbedit queue_mapping QUEUE_MAPPING" (from "man 8 tc-skbedit") is
enhanced with flags:
* SKBEDIT_F_QUEUE_MAPPING_HASH
* SKBEDIT_F_QUEUE_MAPPING_CLASSID
* SKBEDIT_F_QUEUE_MAPPING_CPUID

With act_bpf you can do all of them... So why do you have to do it
in skbedit?

Hi Cong
This idea is inspired by skbedit queue_mapping, and skbedit is
enhanced by this patch.

This is exactly my question. ;)

We support this in skbedit firstly in production. act_bpf can do more
things than this. Anyway we
can support this in both act_skbedit/acc_bpf. 1/2 is changed from
skip_tx_queue in skb to per-cpu var suggested-by Eric. We need another
patch which can change the
per-cpu var in bpf. I will post this patch later.

The point is if act_bpf can do it, you don't need to bother skbedit at
all. More importantly, you are enforcing policies in kernel, which is
not encouraged. So unless you provide more details, this patch is not
needed at all.

Thanks.