netdev - Re: [PATCH v3 net-next 2/5] net: page_pool: add bulk support for ptr

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20201104144950.GA310976@lore-desk>
Date:   Wed, 4 Nov 2020 15:49:50 +0100
From:   Lorenzo Bianconi <lorenzo.bianconi@...hat.com>
To:     Jesper Dangaard Brouer <brouer@...hat.com>
Cc:     Lorenzo Bianconi <lorenzo@...nel.org>, netdev@...r.kernel.org,
        bpf@...r.kernel.org, davem@...emloft.net, kuba@...nel.org,
        ilias.apalodimas@...aro.org
Subject: Re: [PATCH v3 net-next 2/5] net: page_pool: add bulk support for
 ptr_ring

> > diff --git a/net/core/page_pool.c b/net/core/page_pool.c
> > index ef98372facf6..236c5ed3aa66 100644
> > --- a/net/core/page_pool.c
> > +++ b/net/core/page_pool.c
> [...]
> > @@ -408,6 +410,39 @@ void page_pool_put_page(struct page_pool *pool, struct page *page,
> >  }
> >  EXPORT_SYMBOL(page_pool_put_page);
> >  
> > +void page_pool_put_page_bulk(struct page_pool *pool, void **data,
> > +			     int count)
> > +{
> > +	int i, len = 0;
> > +
> > +	for (i = 0; i < count; i++) {
> > +		struct page *page = virt_to_head_page(data[i]);
> > +
> > +		if (likely(page_ref_count(page) == 1 &&
> > +			   pool_page_reusable(pool, page))) {
> > +			if (pool->p.flags & PP_FLAG_DMA_SYNC_DEV)
> > +				page_pool_dma_sync_for_device(pool, page, -1);
> > +
> > +			/* bulk pages for ptr_ring cache */
> > +			data[len++] = page;
> > +		} else {
> > +			page_pool_release_page(pool, page);
> > +			put_page(page);
> > +		}
> > +	}
> > +
> > +	/* Grab the producer spinlock for concurrent access to
> > +	 * ptr_ring page_pool cache
> > +	 */
> > +	page_pool_ring_lock(pool);
> > +	for (i = 0; i < len; i++) {
> > +		if (__ptr_ring_produce(&pool->ring, data[i]))
> > +			page_pool_return_page(pool, data[i]);
> > +	}
> > +	page_pool_ring_unlock(pool);
> > +}
> > +EXPORT_SYMBOL(page_pool_put_page_bulk);
> 
> I don't like that you are replicating the core logic from
> page_pool_put_page() in this function.  This means that we as
> maintainers need to keep both of this places up-to-date.
> 
> Let me try to re-implement this, while sharing the refcnt logic:
> (completely untested, not even compiled)

ack, I like the approach below, I will integrate it in v4

> 
> ---
>  net/core/page_pool.c |   58 +++++++++++++++++++++++++++++++++++++++++++-------
>  2 files changed, 51 insertions(+), 9 deletions(-)
> 
> diff --git a/net/core/page_pool.c b/net/core/page_pool.c
> index ef98372facf6..c785e9825a0d 100644
> --- a/net/core/page_pool.c
> +++ b/net/core/page_pool.c
> @@ -362,8 +362,9 @@ static bool pool_page_reusable(struct page_pool *pool, struct page *page)
>   * If the page refcnt != 1, then the page will be returned to memory
>   * subsystem.

[...]

>  
> +/* Caller must not use data area after call, as this function overwrites it */
> +void page_pool_put_page_bulk(struct page_pool *pool, void **data, int count)
> +{
> +	int i, len = 0, len2 = 0;
> +
> +	for (i = 0; i < count; i++) {
> +		struct page *page = virt_to_head_page(data[i]);
> +
> +		page = __page_pool_put_page(pool, page, -1 , false);
> +
> +		/* Approved for recycling for ptr_ring cache */
> +		if (page)
> +			data[len++] = page;
> +	}

I guess we just return here if len is 0 since we will avoid to grab the
ptr_ring lock, agree?

Regards,
Lorenzo

> +
> +	/* Bulk producer into ptr_ring page_pool cache */
> +	page_pool_ring_lock(pool);
> +	for (i = 0; i < len; i++) {
> +		if (__ptr_ring_produce(&pool->ring, data[i]))
> +			data[len2++] = data[i];
> +	}
> +	page_pool_ring_unlock(pool);
> +
> +	/* Unlikely case of ptr_ring cache full, free pages outside producer
> +	 * lock, given put_page() with refcnt==1 can be an expensive operation.
> +	 */
> +	for (i = 0; i < len2; i++) {
> +		page_pool_return_page(pool, data[i]);
> +	}
> +}
> +EXPORT_SYMBOL(page_pool_put_page_bulk);
> +
>  static void page_pool_empty_ring(struct page_pool *pool)
>  {
>  	struct page *page;
> 
> 
> -- 
> Best regards,
>   Jesper Dangaard Brouer
>   MSc.CS, Principal Kernel Engineer at Red Hat
>   LinkedIn: http://www.linkedin.com/in/brouer
> 

Download attachment "signature.asc" of type "application/pgp-signature" (229 bytes)