netdev - Re: [PATCH net-next 8/9] igc: Enable RX via AF

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 for Android: free password hash cracker in your pocket

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

Date:   Fri, 9 Apr 2021 17:36:04 -0700
From:   Jakub Kicinski <kuba@...nel.org>
To:     Tony Nguyen <anthony.l.nguyen@...el.com>
Cc:     davem@...emloft.net, Andre Guedes <andre.guedes@...el.com>,
        netdev@...r.kernel.org, sassmann@...hat.com, bjorn.topel@...el.com,
        magnus.karlsson@...el.com, maciej.fijalkowski@...el.com,
        sasha.neftin@...el.com, vitaly.lifshits@...el.com,
        Vedang Patel <vedang.patel@...el.com>,
        Jithu Joseph <jithu.joseph@...el.com>,
        Dvora Fuxbrumer <dvorax.fuxbrumer@...ux.intel.com>
Subject: Re: [PATCH net-next 8/9] igc: Enable RX via AF_XDP zero-copy

On Fri,  9 Apr 2021 09:43:50 -0700 Tony Nguyen wrote:
> From: Andre Guedes <andre.guedes@...el.com>
> 
> Add support for receiving packets via AF_XDP zero-copy mechanism.
> 
> Add a new flag to 'enum igc_ring_flags_t' to indicate the ring has
> AF_XDP zero-copy enabled so proper ring setup is carried out during ring
> configuration in igc_configure_rx_ring().
> 
> RX buffers can now be allocated via the shared pages mechanism (default
> behavior of the driver) or via xsk pool (when AF_XDP zero-copy is
> enabled) so a union is added to the 'struct igc_rx_buffer' to cover both
> cases.
> 
> When AF_XDP zero-copy is enabled, rx buffers are allocated from the xsk
> pool using the new helper igc_alloc_rx_buffers_zc() which is the
> counterpart of igc_alloc_rx_buffers().
> 
> Likewise other Intel drivers that support AF_XDP zero-copy, in igc we
> have a dedicated path for cleaning up rx irqs when zero-copy is enabled.
> This avoids adding too many checks within igc_clean_rx_irq(), resulting
> in a more readable and efficient code since this function is called from
> the hot-path of the driver.

> +static struct sk_buff *igc_construct_skb_zc(struct igc_ring *ring,
> +					    struct xdp_buff *xdp)
> +{
> +	unsigned int metasize = xdp->data - xdp->data_meta;
> +	unsigned int datasize = xdp->data_end - xdp->data;
> +	struct sk_buff *skb;
> +
> +	skb = __napi_alloc_skb(&ring->q_vector->napi,
> +			       xdp->data_end - xdp->data_hard_start,
> +			       GFP_ATOMIC | __GFP_NOWARN);
> +	if (unlikely(!skb))
> +		return NULL;
> +
> +	skb_reserve(skb, xdp->data - xdp->data_hard_start);
> +	memcpy(__skb_put(skb, datasize), xdp->data, datasize);
> +	if (metasize)
> +		skb_metadata_set(skb, metasize);

But you haven't actually copied the matadata into the skb,
the metadata is before xdp->data, right?

> +	return skb;
> +}

> +static int igc_xdp_enable_pool(struct igc_adapter *adapter,
> +			       struct xsk_buff_pool *pool, u16 queue_id)
> +{
> +	struct net_device *ndev = adapter->netdev;
> +	struct device *dev = &adapter->pdev->dev;
> +	struct igc_ring *rx_ring;
> +	struct napi_struct *napi;
> +	bool needs_reset;
> +	u32 frame_size;
> +	int err;
> +
> +	if (queue_id >= adapter->num_rx_queues)
> +		return -EINVAL;
> +
> +	frame_size = xsk_pool_get_rx_frame_size(pool);
> +	if (frame_size < ETH_FRAME_LEN + VLAN_HLEN * 2) {
> +		/* When XDP is enabled, the driver doesn't support frames that
> +		 * span over multiple buffers. To avoid that, we check if xsk
> +		 * frame size is big enough to fit the max ethernet frame size
> +		 * + vlan double tagging.
> +		 */
> +		return -EOPNOTSUPP;
> +	}
> +
> +	err = xsk_pool_dma_map(pool, dev, IGC_RX_DMA_ATTR);
> +	if (err) {
> +		netdev_err(ndev, "Failed to map xsk pool\n");
> +		return err;
> +	}
> +
> +	needs_reset = netif_running(adapter->netdev) && igc_xdp_is_enabled(adapter);
> +
> +	rx_ring = adapter->rx_ring[queue_id];
> +	napi = &rx_ring->q_vector->napi;
> +
> +	if (needs_reset) {
> +		igc_disable_rx_ring(rx_ring);
> +		napi_disable(napi);
> +	}
> +
> +	set_bit(IGC_RING_FLAG_AF_XDP_ZC, &rx_ring->flags);
> +
> +	if (needs_reset) {
> +		napi_enable(napi);
> +		igc_enable_rx_ring(rx_ring);
> +
> +		err = igc_xsk_wakeup(ndev, queue_id, XDP_WAKEUP_RX);
> +		if (err)
> +			return err;

No need for an unwind path here?
Does something call XDP_SETUP_XSK_POOL(NULL) on failure automagically?

> +	}
> +
> +	return 0;
> +}