netdev - Re: [RFC net-next 3/3] drivers/vdpa: add NFP devices vDPA driver

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]

Message-ID: <20240807134831.GB2991391@kernel.org>
Date: Wed, 7 Aug 2024 14:48:31 +0100
From: Simon Horman <horms@...nel.org>
To: Louis Peens <louis.peens@...igine.com>
Cc: David Miller <davem@...emloft.net>, Jakub Kicinski <kuba@...nel.org>,
	"Michael S. Tsirkin" <mst@...hat.com>,
	Jason Wang <jasowang@...hat.com>, eperezma@...hat.com,
	Kyle Xu <zhenbing.xu@...igine.com>, netdev@...r.kernel.org,
	virtualization@...ts.linux.dev, oss-drivers@...igine.com
Subject: Re: [RFC net-next 3/3] drivers/vdpa: add NFP devices vDPA driver

On Fri, Aug 02, 2024 at 11:59:31AM +0200, Louis Peens wrote:
> From: Kyle Xu <zhenbing.xu@...igine.com>
> 
> Add a new kernel module ‘nfp_vdpa’ for the NFP vDPA networking driver.
> 
> The vDPA driver initializes the necessary resources on the VF and the
> data path will be offloaded. It also implements the ‘vdpa_config_ops’
> and the corresponding callback interfaces according to the requirement
> of kernel vDPA framework.
> 
> Signed-off-by: Kyle Xu <zhenbing.xu@...igine.com>
> Signed-off-by: Louis Peens <louis.peens@...igine.com>

...

> diff --git a/drivers/vdpa/netronome/nfp_vdpa_main.c b/drivers/vdpa/netronome/nfp_vdpa_main.c

...

> +static int nfp_vdpa_map_resources(struct nfp_vdpa_net *ndev,
> +				  struct pci_dev *pdev,
> +				  const struct nfp_dev_info *dev_info)
> +{
> +	unsigned int bar_off, bar_sz, tx_bar_sz, rx_bar_sz;
> +	unsigned int max_tx_rings, max_rx_rings, txq, rxq;
> +	u64 tx_bar_off, rx_bar_off;
> +	resource_size_t map_addr;
> +	void __iomem  *tx_bar;
> +	void __iomem  *rx_bar;

Hi Kyle and Louis,

A minor nit from my side: rx_bar is set but otherwise unused in this function.

> +	int err;
> +
> +	/* Map CTRL BAR */
> +	ndev->ctrl_bar = ioremap(pci_resource_start(pdev, NFP_NET_CTRL_BAR),
> +				 NFP_NET_CFG_BAR_SZ);
> +	if (!ndev->ctrl_bar)
> +		return -EIO;
> +
> +	/* Find out how many rings are supported */
> +	max_tx_rings = readl(ndev->ctrl_bar + NFP_NET_CFG_MAX_TXRINGS);
> +	max_rx_rings = readl(ndev->ctrl_bar + NFP_NET_CFG_MAX_RXRINGS);
> +	/* Currently, only one ring is supported */
> +	if (max_tx_rings != NFP_VDPA_QUEUE_RING_MAX || max_rx_rings != NFP_VDPA_QUEUE_RING_MAX) {
> +		err = -EINVAL;
> +		goto ctrl_bar_unmap;
> +	}
> +
> +	/* Map Q0_BAR as a single overlapping BAR mapping */
> +	tx_bar_sz = NFP_QCP_QUEUE_ADDR_SZ * max_tx_rings * NFP_VDPA_QUEUE_SPACE_STRIDE;
> +	rx_bar_sz = NFP_QCP_QUEUE_ADDR_SZ * max_rx_rings * NFP_VDPA_QUEUE_SPACE_STRIDE;
> +
> +	txq = readl(ndev->ctrl_bar + NFP_NET_CFG_START_TXQ);
> +	tx_bar_off = nfp_qcp_queue_offset(dev_info, txq);
> +	rxq = readl(ndev->ctrl_bar + NFP_NET_CFG_START_RXQ);
> +	rx_bar_off = nfp_qcp_queue_offset(dev_info, rxq);
> +
> +	bar_off = min(tx_bar_off, rx_bar_off);
> +	bar_sz = max(tx_bar_off + tx_bar_sz, rx_bar_off + rx_bar_sz);
> +	bar_sz -= bar_off;
> +
> +	map_addr = pci_resource_start(pdev, NFP_NET_Q0_BAR) + bar_off;
> +	ndev->q_bar = ioremap(map_addr, bar_sz);
> +	if (!ndev->q_bar) {
> +		err = -EIO;
> +		goto ctrl_bar_unmap;
> +	}
> +
> +	tx_bar = ndev->q_bar + (tx_bar_off - bar_off);
> +	rx_bar = ndev->q_bar + (rx_bar_off - bar_off);
> +
> +	/* TX queues */
> +	ndev->vring[txq].kick_addr = ndev->ctrl_bar + NFP_VDPA_NOTIFY_AREA_BASE
> +				     + txq * NFP_VDPA_QUEUE_NOTIFY_OFFSET;
> +	/* RX queues */
> +	ndev->vring[rxq].kick_addr = ndev->ctrl_bar + NFP_VDPA_NOTIFY_AREA_BASE
> +				     + rxq * NFP_VDPA_QUEUE_NOTIFY_OFFSET;
> +	/* Stash the re-configuration queue away. First odd queue in TX Bar */
> +	ndev->qcp_cfg = tx_bar + NFP_QCP_QUEUE_ADDR_SZ;
> +
> +	return 0;
> +
> +ctrl_bar_unmap:
> +	iounmap(ndev->ctrl_bar);
> +	return err;
> +}

...