netdev - Re: 答复: [PATCH][v2] page_pool: handle page recycle for NUMA_NO

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

Date:   Tue, 10 Dec 2019 19:56:50 +0000
From:   Saeed Mahameed <saeedm@...lanox.com>
To:     "linyunsheng@...wei.com" <linyunsheng@...wei.com>,
        Li Rongqing <lirongqing@...du.com>,
        "brouer@...hat.com" <brouer@...hat.com>
CC:     "jonathan.lemon@...il.com" <jonathan.lemon@...il.com>,
        "ilias.apalodimas@...aro.org" <ilias.apalodimas@...aro.org>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: 答复: [PATCH][v2] page_pool: handle page recycle for NUMA_NO_NODE condition

On Tue, 2019-12-10 at 09:39 +0000, Li,Rongqing wrote:
> > static int mvneta_create_page_pool(struct mvneta_port *pp,
> > 				   struct mvneta_rx_queue *rxq, int
> > size) {
> > 	struct bpf_prog *xdp_prog = READ_ONCE(pp->xdp_prog);
> > 	struct page_pool_params pp_params = {
> > 		.order = 0,
> > 		.flags = PP_FLAG_DMA_MAP | PP_FLAG_DMA_SYNC_DEV,
> > 		.pool_size = size,
> > 		.nid = cpu_to_node(0),
> 
> This kind of device should only be installed to vendor's platform
> which did not support numa
> 
> But as you say , Saeed advice maybe cause that recycle always fail,
> if nid is configured like upper, and different from running NAPI node
> id
> 

I don't see an issue here, By definition recycling must fail in such
case :).

> And maybe we can catch this case by the below
> 
> diff --git a/net/core/page_pool.c b/net/core/page_pool.c
> index 3c8b51ccd1c1..973235c09487 100644
> --- a/net/core/page_pool.c
> +++ b/net/core/page_pool.c
> @@ -328,6 +328,11 @@ static bool pool_page_reusable(struct page_pool
> *pool, struct page *page)
>  void __page_pool_put_page(struct page_pool *pool, struct page *page,
>                           unsigned int dma_sync_size, bool
> allow_direct)
>  {
> +       allow_direct = allow_direct && in_serving_softirq();
> +
> +       if (allow_direct)
> +               WARN_ON_ONCE((pool->p.nid != NUMA_NO_NODE) &&
> +                                    (pool->p.nid != numa_mem_id()));
>         /* This allocator is optimized for the XDP mode that uses
>          * one-frame-per-page, but have fallbacks that act like the
>          * regular page allocator APIs.
> @@ -342,7 +347,7 @@ void __page_pool_put_page(struct page_pool *pool,
> struct page *page,
>                         page_pool_dma_sync_for_device(pool, page,
>                                                       dma_sync_size);
>  
> -               if (allow_direct && in_serving_softirq())
> +               if (allow_direct)
>                         if (__page_pool_recycle_direct(page, pool))
>                                 return;
> 
> 

too much data path complications for my taste, we need to establish
some assumptions.
1) user know what he is doing, MSIX/NAPI affinity should be as hinted
by driver, pool nid should correspond to the affinity hint.
2) if necessary mitigate NAPI numa migration via page_pool_update_nid()