Re: [PATCH v5 00/40] support setting and querying RSS algorithms

[Thomas Monjalon <thomas@monjalon.net>]

24/10/2023 14:57, lihuisong (C):
> 在 2023/10/17 22:26, Ferruh Yigit 写道:
> > On 10/17/2023 3:06 PM, Thomas Monjalon wrote:
> >> 11/10/2023 11:27, Jie Hai:
> >>>   55 files changed, 395 insertions(+), 106 deletions(-)
> >> 
> >> Changing all drivers is suspicious.
> >> It shows that something is missing in ethdev.
> >> Please can you add the checks in ethdev only?
> >>
> > That is kind of request from me, let me try to summarize what is going on,
> >
> > there is a new config item added to "struct rte_eth_rss_conf" introduced
> > in this set, which is RSS hashing algorithm to use.
> >
> > Problem is none of the existing drivers are taking account this new
> > config item, so application will request it but drivers silently ignore it.
> >
> > This is a generic problem when adding a new config item to existing
> > config struct.
> >
> > So my request was if drivers not supporting it, and it is requested by
> > the application, driver should return an error to let application know
> > that it is not supported, that is why bunch of drivers updated.
> >
> >
> > One option can be adding a new, specific API and dev_ops for this, for
> > this case new config item is related to the existing RSS API, so I think
> > it can be part of the existing API.
> >
> > Other can be to have some kind of capability reporting by drivers, and
> > application will detect it and won't request this new config item, I
> > think Stephen already suggested something like this. This capability
> > flag is again a generic requirement, and `rte_eth_dev_info_get()`
> > partially used for this purpose. I think it will be odd from application
> > perspective to have a capability for just one config item of a feature set.
> >
> >
> > Anyway, I think updating drivers to report they are not supporting new
> > config item is best option to me, but also I think we should discuss
> > this capability reporting in ethdev in a wider context.
> IMO, it is more better to report RSS algorithm capability.
> It can avoid the later ABI break successfully as Stephen said.

Yes we should add a capability for RSS algorithm.