Re: [RFC PATCH net-next] page_pool: Track DMA-mapped pages and unmap them when destroying the pool

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

On 3/12/25 12:27, Toke Høiland-Jørgensen wrote:
> Yunsheng Lin <linyunsheng@huawei.com> writes:
...
>>>>> cases where it's absolutely needed.
>>>>
>>>> The above can also be done for using page_pool_item too as the
>>>> lower 2 bits can be used to indicate the pointer in 'struct page'
>>>> is 'page_pool_item' or 'page_pool', I just don't think it is
>>>> necessary yet as it might add more checking in the fast path.
>>>
>>> Yup, did think about using the lower bits to distinguish if it does turn
>>> out that we can't avoid an indirection. See above; it's not actually the
>>
>> The 'memdesc' seems like an indirection to me when using that to shrink
>> 'struct page' to a smaller size.
> 
> Yes, it does seem like we'll end up with an indirection of some kind
> eventually. But let's cross that bridge when we get to it...

At which point it might be easier to avoid all the "bump"s business,
fully embrace net_iov / netmem format, wrap all pp pages into a
structure on the page pool side, and pass that around. That would
remove the indirection for most of the accesses, and the allocation
can be easily cached.

-- 
Pavel Begunkov