On Tue, Oct 29, 2024 at 4:06 PM David Wei <dw@xxxxxxxxxxx> wrote:
>
> From: Pavel Begunkov <asml.silence@xxxxxxxxx>
>
> Add a helper that allows a page pool memory provider to efficiently
> return a netmem off the allocation callback.
>
> Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
> Signed-off-by: David Wei <dw@xxxxxxxxxxx>
> ---
> include/net/page_pool/memory_provider.h | 4 ++++
> net/core/page_pool.c | 19 +++++++++++++++++++
> 2 files changed, 23 insertions(+)
>
> diff --git a/include/net/page_pool/memory_provider.h b/include/net/page_pool/memory_provider.h
> index 83d7eec0058d..352b3a35d31c 100644
> --- a/include/net/page_pool/memory_provider.h
> +++ b/include/net/page_pool/memory_provider.h
> @@ -1,3 +1,5 @@
> +/* SPDX-License-Identifier: GPL-2.0-or-later */
> +
> #ifndef _NET_PAGE_POOL_MEMORY_PROVIDER_H
> #define _NET_PAGE_POOL_MEMORY_PROVIDER_H
>
> @@ -7,4 +9,6 @@ int page_pool_mp_init_paged_area(struct page_pool *pool,
> void page_pool_mp_release_area(struct page_pool *pool,
> struct net_iov_area *area);
>
> +void page_pool_mp_return_in_cache(struct page_pool *pool, netmem_ref netmem);
> +
> #endif
> diff --git a/net/core/page_pool.c b/net/core/page_pool.c
> index 8bd4a3c80726..9078107c906d 100644
> --- a/net/core/page_pool.c
> +++ b/net/core/page_pool.c
> @@ -1213,3 +1213,22 @@ void page_pool_mp_release_area(struct page_pool *pool,
> page_pool_release_page_dma(pool, net_iov_to_netmem(niov));
> }
> }
> +
> +/*
> + * page_pool_mp_return_in_cache() - return a netmem to the allocation cache.
> + * @pool: pool from which pages were allocated
> + * @netmem: netmem to return
> + *
> + * Return already allocated and accounted netmem to the page pool's allocation
> + * cache. The function doesn't provide synchronisation and must only be called
> + * from the napi context.
Maybe add:
/* Caller must verify that there is room in the cache */
> + */
> +void page_pool_mp_return_in_cache(struct page_pool *pool, netmem_ref netmem)
> +{
> + if (WARN_ON_ONCE(pool->alloc.count >= PP_ALLOC_CACHE_REFILL))
> + return;
The caller must verify this anyway, right? so maybe this WARN_ON_ONCE
is too defensive.
> +
> + page_pool_dma_sync_for_device(pool, netmem, -1);
> + page_pool_fragment_netmem(netmem, 1);
> + pool->alloc.cache[pool->alloc.count++] = netmem;
> +}
> --
> 2.43.5
>
--
Thanks,
Mina
