Re: [PATCH net-next v18 07/14] memory-provider: dmabuf devmem memory provider

[Pavel Begunkov <asml.silence@gmail.com>]

On 8/13/24 01:15, Jakub Kicinski wrote:
> On Mon, 12 Aug 2024 20:04:41 +0100 Pavel Begunkov wrote:
>>>> Also don't see the upside of the explicit "non-capable" flag,
>>>> but I haven't thought of that. Is there any use?
>>
>> Or maybe I don't get what you're asking, I explained
>> why to have that "PP_IGNORE_PROVIDERS" on top of the flag
>> saying that it's supported.
>>
>> Which "non-capable" flag you have in mind? A page pool create
>> flag or one facing upper layers like devmem tcp?
> 
> Let me rephrase - what's the point of having both PP_PROVIDERS_SUPPORTED
> and PP_IGNORE_PROVIDERS at the page pool level? PP_CAP_NET(MEM|IOV),
> and it's either there or it's not.

The second flag solves a problem with initializing page pools
with headers, but let's forget about it for now, it's rather a
small nuance, would probably reappear when someone would try to
use pp_params->queue for purposes different from memory providers.

> If you're thinking about advertising the support all the way to the
> user, I'm not sure if page pool is the right place to do so. It's more
> of a queue property.

Nope. Only the first "SUPPORTED" flag serves that purpose in a way
by failing setup like netlink devmem dmabuf binding and returning
the error back to user.

> BTW, Mina, the core should probably also check that XDP isn't installed
> before / while the netmem is bound to a queue.

-- 
Pavel Begunkov