netdev - Re: [PATCH 09/33] xsk: xsk_buff_pool add callback for dma

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:   Thu, 2 Feb 2023 13:51:20 +0100
From:   Magnus Karlsson <magnus.karlsson@...il.com>
To:     Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
Cc:     netdev@...r.kernel.org, "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>,
        "Michael S. Tsirkin" <mst@...hat.com>,
        Jason Wang <jasowang@...hat.com>,
        Björn Töpel <bjorn@...nel.org>,
        Magnus Karlsson <magnus.karlsson@...el.com>,
        Maciej Fijalkowski <maciej.fijalkowski@...el.com>,
        Jonathan Lemon <jonathan.lemon@...il.com>,
        Alexei Starovoitov <ast@...nel.org>,
        Daniel Borkmann <daniel@...earbox.net>,
        Jesper Dangaard Brouer <hawk@...nel.org>,
        John Fastabend <john.fastabend@...il.com>,
        Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
        Menglong Dong <imagedong@...cent.com>,
        Kuniyuki Iwashima <kuniyu@...zon.com>,
        Petr Machata <petrm@...dia.com>,
        virtualization@...ts.linux-foundation.org, bpf@...r.kernel.org
Subject: Re: [PATCH 09/33] xsk: xsk_buff_pool add callback for dma_sync

On Thu, 2 Feb 2023 at 12:05, Xuan Zhuo <xuanzhuo@...ux.alibaba.com> wrote:
>
> Use callback to implement dma sync to simplify subsequent support for
> virtio dma sync.
>
> Signed-off-by: Xuan Zhuo <xuanzhuo@...ux.alibaba.com>
> ---
>  include/net/xsk_buff_pool.h |  6 ++++++
>  net/xdp/xsk_buff_pool.c     | 24 ++++++++++++++++++++----
>  2 files changed, 26 insertions(+), 4 deletions(-)
>
> diff --git a/include/net/xsk_buff_pool.h b/include/net/xsk_buff_pool.h
> index 3e952e569418..53b681120354 100644
> --- a/include/net/xsk_buff_pool.h
> +++ b/include/net/xsk_buff_pool.h
> @@ -75,6 +75,12 @@ struct xsk_buff_pool {
>         u32 chunk_size;
>         u32 chunk_shift;
>         u32 frame_len;
> +       void (*dma_sync_for_cpu)(struct device *dev, dma_addr_t addr,
> +                                unsigned long offset, size_t size,
> +                                enum dma_data_direction dir);
> +       void (*dma_sync_for_device)(struct device *dev, dma_addr_t addr,
> +                                   unsigned long offset, size_t size,
> +                                   enum dma_data_direction dir);

If we put these two pointers here, the number of cache lines required
in the data path for this struct will be increased from 2 to 3 which
will likely affect performance negatively. These sync operations are
also not used on most systems. So how about we put them in the first
section of this struct labeled "Members only used in the control path
first." instead. There is a 26-byte hole at the end of it that can be
used.

>         u8 cached_need_wakeup;
>         bool uses_need_wakeup;
>         bool dma_need_sync;
> diff --git a/net/xdp/xsk_buff_pool.c b/net/xdp/xsk_buff_pool.c
> index ed6c71826d31..78e325e195fa 100644
> --- a/net/xdp/xsk_buff_pool.c
> +++ b/net/xdp/xsk_buff_pool.c
> @@ -403,6 +403,20 @@ static int xp_init_dma_info(struct xsk_buff_pool *pool, struct xsk_dma_map *dma_
>         return 0;
>  }
>
> +static void dma_sync_for_cpu(struct device *dev, dma_addr_t addr,
> +                            unsigned long offset, size_t size,
> +                            enum dma_data_direction dir)
> +{
> +       dma_sync_single_range_for_cpu(dev, addr, offset, size, dir);
> +}
> +
> +static void dma_sync_for_device(struct device *dev, dma_addr_t addr,
> +                               unsigned long offset, size_t size,
> +                               enum dma_data_direction dir)
> +{
> +       dma_sync_single_range_for_device(dev, addr, offset, size, dir);
> +}
> +
>  int xp_dma_map(struct xsk_buff_pool *pool, struct device *dev,
>                unsigned long attrs, struct page **pages, u32 nr_pages)
>  {
> @@ -421,6 +435,9 @@ int xp_dma_map(struct xsk_buff_pool *pool, struct device *dev,
>                 return 0;
>         }
>
> +       pool->dma_sync_for_cpu = dma_sync_for_cpu;
> +       pool->dma_sync_for_device = dma_sync_for_device;
> +
>         dma_map = xp_create_dma_map(dev, pool->netdev, nr_pages, pool->umem);
>         if (!dma_map)
>                 return -ENOMEM;
> @@ -667,15 +684,14 @@ EXPORT_SYMBOL(xp_raw_get_dma);
>
>  void xp_dma_sync_for_cpu_slow(struct xdp_buff_xsk *xskb)
>  {
> -       dma_sync_single_range_for_cpu(xskb->pool->dev, xskb->dma, 0,
> -                                     xskb->pool->frame_len, DMA_BIDIRECTIONAL);
> +       xskb->pool->dma_sync_for_cpu(xskb->pool->dev, xskb->dma, 0,
> +                                    xskb->pool->frame_len, DMA_BIDIRECTIONAL);
>  }
>  EXPORT_SYMBOL(xp_dma_sync_for_cpu_slow);
>
>  void xp_dma_sync_for_device_slow(struct xsk_buff_pool *pool, dma_addr_t dma,
>                                  size_t size)
>  {
> -       dma_sync_single_range_for_device(pool->dev, dma, 0,
> -                                        size, DMA_BIDIRECTIONAL);
> +       pool->dma_sync_for_device(pool->dev, dma, 0, size, DMA_BIDIRECTIONAL);
>  }
>  EXPORT_SYMBOL(xp_dma_sync_for_device_slow);
> --
> 2.32.0.3.g01195cf9f
>