Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Wed, 19 Jul 2023 20:45:53 -0700
From:   Jakub Kicinski <kuba@kernel.org>
To:     Wei Fang <wei.fang@nxp.com>
Cc:     davem@davemloft.net, edumazet@google.com, pabeni@redhat.com,
        ast@kernel.org, daniel@iogearbox.net, hawk@kernel.org,
        john.fastabend@gmail.com, xiaoning.wang@nxp.com,
        shenwei.wang@nxp.com, netdev@vger.kernel.org, linux-imx@nxp.com,
        linux-kernel@vger.kernel.org, bpf@vger.kernel.org
Subject: Re: [PATCH net-next] net: fec: add XDP_TX feature support
Message-ID: <20230719204553.46856b29@kernel.org>
In-Reply-To: <20230717103709.2629372-1-wei.fang@nxp.com>
References: <20230717103709.2629372-1-wei.fang@nxp.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Mon, 17 Jul 2023 18:37:09 +0800 Wei Fang wrote:
> -			xdp_return_frame(xdpf);
> +			if (txq->tx_buf[index].type == FEC_TXBUF_T_XDP_NDO)
> +				xdp_return_frame(xdpf);
> +			else
> +				xdp_return_frame_rx_napi(xdpf);

Are you taking budget into account? When NAPI is called with budget 
of 0 we are *not* in napi / softirq context. You can't be processing
any XDP tx under such conditions (it may be a netpoll call from IRQ
context).

> +static int fec_enet_xdp_tx_xmit(struct net_device *ndev,
> +				struct xdp_buff *xdp)
> +{
> +	struct xdp_frame *xdpf = xdp_convert_buff_to_frame(xdp);
> +	struct fec_enet_private *fep = netdev_priv(ndev);
> +	struct fec_enet_priv_tx_q *txq;
> +	int cpu = smp_processor_id();
> +	struct netdev_queue *nq;
> +	int queue, ret;
> +
> +	queue = fec_enet_xdp_get_tx_queue(fep, cpu);
> +	txq = fep->tx_queue[queue];
> +	nq = netdev_get_tx_queue(fep->netdev, queue);
> +
> +	__netif_tx_lock(nq, cpu);
> +
> +	ret = fec_enet_txq_xmit_frame(fep, txq, xdpf, false);
> +
> +	__netif_tx_unlock(nq);

If you're reusing the same queues as the stack you need to call
txq_trans_cond_update() at some point, otherwise the stack may
print a splat complaining the queue got stuck.
-- 
pw-bot: cr