mail archive of the barebox mailing list
 help / color / mirror / Atom feed
From: Sascha Hauer <s.hauer@pengutronix.de>
To: Steffen Trumtrar <s.trumtrar@pengutronix.de>
Cc: barebox@lists.infradead.org
Subject: Re: [PATCH 06/10] net: add support for Designware XGMAC (10gb) ethernet
Date: Mon, 4 Nov 2024 12:14:50 +0100	[thread overview]
Message-ID: <ZyisqoVwDIUO6bJj@pengutronix.de> (raw)
In-Reply-To: <20241029-v2024-10-0-topic-socfpga-agilex5-v1-6-96df2d7dadf4@pengutronix.de>

On Tue, Oct 29, 2024 at 09:42:36AM +0100, Steffen Trumtrar wrote:
> +static int xgmac_mdio_read(struct mii_bus *bus, int mdio_addr, int mdio_reg)
> +{
> +	struct xgmac_priv *xgmac = bus->priv;
> +	u32 val;
> +	u32 hw_addr;
> +	u32 idle;
> +	int ret;
> +
> +	ret = readl_poll_timeout(&xgmac->mac_regs->mdio_data, idle,
> +				 !(idle & XGMAC_MAC_MDIO_ADDRESS_SBUSY),
> +				 XGMAC_TIMEOUT_100MS);
> +	if (ret) {
> +		pr_err("MDIO not idle at entry: %d\n", ret);

dev_err() please. Same for other pr_ style functions in this file.

> +		return ret;
> +	}
> +
> +	/* Set clause 22 format */
> +	val = BIT(mdio_addr);
> +	writel(val, &xgmac->mac_regs->mdio_clause_22_port);
> +
> +	hw_addr = (mdio_addr << XGMAC_MAC_MDIO_ADDRESS_PA_SHIFT) |
> +		   (mdio_reg & XGMAC_MAC_MDIO_REG_ADDR_C22P_MASK);
> +
> +	val = xgmac->config->config_mac_mdio <<
> +	      XGMAC_MAC_MDIO_ADDRESS_CR_SHIFT;
> +
> +	val |= XGMAC_MAC_MDIO_ADDRESS_SADDR |
> +	       XGMAC_MDIO_SINGLE_CMD_ADDR_CMD_READ |
> +	       XGMAC_MAC_MDIO_ADDRESS_SBUSY;
> +
> +	ret = readl_poll_timeout(&xgmac->mac_regs->mdio_data, idle,
> +				 !(idle & XGMAC_MAC_MDIO_ADDRESS_SBUSY),
> +				 XGMAC_TIMEOUT_100MS);
> +	if (ret) {
> +		pr_err("MDIO not idle at entry: %d\n", ret);
> +		return ret;
> +	}
> +
> +	writel(hw_addr, &xgmac->mac_regs->mdio_address);
> +	writel(val, &xgmac->mac_regs->mdio_data);
> +
> +	ret = readl_poll_timeout(&xgmac->mac_regs->mdio_data, idle,
> +				 !(idle & XGMAC_MAC_MDIO_ADDRESS_SBUSY),
> +				 XGMAC_TIMEOUT_100MS);
> +	if (ret) {
> +		pr_err("MDIO read didn't complete: %d\n", ret);
> +		return ret;
> +	}
> +
> +	val = readl(&xgmac->mac_regs->mdio_data);
> +	val &= XGMAC_MAC_MDIO_DATA_GD_MASK;
> +
> +	return val;
> +}
> +
> +static int xgmac_send(struct eth_device *edev, void *packet, int length)
> +{
> +	struct xgmac_priv *xgmac = edev->priv;
> +	struct xgmac_desc *tx_desc;
> +	dma_addr_t dma;
> +	u32 des3_prev, des3;
> +	int ret;
> +
> +	tx_desc = &xgmac->tx_descs[xgmac->tx_desc_idx];
> +	xgmac->tx_desc_idx++;
> +	xgmac->tx_desc_idx %= XGMAC_DESCRIPTORS_NUM;
> +
> +	dma = dma_map_single(edev->parent, packet, length, DMA_TO_DEVICE);
> +	if (dma_mapping_error(edev->parent, dma))
> +		return -EFAULT;
> +
> +	tx_desc->des0 = dma;

lower_32_bits()

> +	tx_desc->des1 = 0;

upper_32_bits() go here?

> +	tx_desc->des2 = length;
> +	/*
> +	 * Make sure that if HW sees the _OWN write below, it will see all the
> +	 * writes to the rest of the descriptor too.
> +	 */
> +	barrier();
> +
> +	des3_prev = XGMAC_DESC3_OWN | XGMAC_DESC3_FD | XGMAC_DESC3_LD | length;
> +	writel(des3_prev, &tx_desc->des3);
> +	writel((ulong)(tx_desc + 1), &xgmac->dma_regs->ch0_txdesc_tail_pointer); // <-- TODO
> +
> +	ret = readl_poll_timeout(&tx_desc->des3, des3,
> +				 !(des3 & XGMAC_DESC3_OWN),
> +				 100 * USEC_PER_MSEC);
> +
> +	dma_unmap_single(edev->parent, dma, length, DMA_TO_DEVICE);
> +
> +	if (ret == -ETIMEDOUT)
> +		debug("%s: TX timeout 0x%08x\n", __func__, des3);
> +
> +	return ret;
> +}
> +
> +static void xgmac_recv(struct eth_device *edev)
> +{
> +	struct xgmac_priv *xgmac = edev->priv;
> +	struct xgmac_desc *rx_desc;
> +	dma_addr_t dma;
> +	void *pkt;
> +	int length;
> +
> +	rx_desc = &xgmac->rx_descs[xgmac->rx_desc_idx];
> +
> +	if (rx_desc->des3 & XGMAC_DESC3_OWN)
> +		return;
> +
> +	dma = xgmac->dma_rx_buf[xgmac->rx_desc_idx];
> +	pkt = phys_to_virt(dma);

You should store the virtual address somewhere in your private data and
use it here rather than converting back the dma address.

> +	length = rx_desc->des3 & XGMAC_RDES3_PKT_LENGTH_MASK;
> +
> +	dma_sync_single_for_cpu(edev->parent, (unsigned long)pkt, length,
> +				DMA_FROM_DEVICE);

Use the dma address directly here.

> +	net_receive(edev, pkt, length);
> +	dma_sync_single_for_device(edev->parent, (unsigned long)pkt,
> +				   length, DMA_FROM_DEVICE);
> +
> +	/* Read Format RX descriptor */
> +	rx_desc = &xgmac->rx_descs[xgmac->rx_desc_idx];
> +	rx_desc->des0 = dma;
> +	rx_desc->des1 = 0;
> +	rx_desc->des2 = 0;
> +	/*
> +	 * Make sure that if HW sees the _OWN write below, it will see all the
> +	 * writes to the rest of the descriptor too.
> +	 */
> +	rx_desc->des3 = XGMAC_DESC3_OWN;
> +	barrier();
> +
> +	writel((ulong)rx_desc, &xgmac->dma_regs->ch0_rxdesc_tail_pointer);

calculate the address based on xgmac->rx_descs_phys rather than assuming
a 1:1 mapping.

> +
> +	xgmac->rx_desc_idx++;
> +	xgmac->rx_desc_idx %= XGMAC_DESCRIPTORS_NUM;
> +}
> +
> +int xgmac_probe(struct device *dev)
> +{
> +	struct device_node *np = dev->of_node;
> +	struct mii_bus *miibus;
> +	struct xgmac_priv *xgmac;
> +	struct resource *iores;
> +	struct eth_device *edev;
> +	int ret = 0;
> +
> +	xgmac = xzalloc(sizeof(*xgmac));
> +
> +	xgmac->dev = dev;
> +	ret = dev_get_drvdata(dev, (const void **)&xgmac->config);

Use device_get_match_data() instead.

> +	if (ret < 0) {
> +		pr_err("xgmac_probe() failed to get driver data: %d\n", ret);
> +		return ret;
> +	}
> +
> +	iores = dev_request_mem_resource(dev, 0);
> +	if (IS_ERR(iores))
> +		return PTR_ERR(iores);
> +	xgmac->regs = IOMEM(iores->start);
> +
> +	xgmac->mac_regs = (void *)(xgmac->regs + XGMAC_MAC_REGS_BASE);
> +	xgmac->mtl_regs = (void *)(xgmac->regs + XGMAC_MTL_REGS_BASE);
> +	xgmac->dma_regs = (void *)(xgmac->regs + XGMAC_DMA_REGS_BASE);

xgmac->regs already is of type void *, no need to cast-

> +
> +/* DMA Registers */
> +
> +#define XGMAC_DMA_REGS_BASE 0x3000
> +
> +struct xgmac_dma_regs {
> +	u32 mode;					/* 0x3000 */
> +	u32 sysbus_mode;				/* 0x3004 */
> +	u32 unused_3008[(0x3100 - 0x3008) / 4];		/* 0x3008 */
> +	u32 ch0_control;				/* 0x3100 */
> +	u32 ch0_tx_control;				/* 0x3104 */
> +	u32 ch0_rx_control;				/* 0x3108 */
> +	u32 slot_func_control_status;			/* 0x310c */
> +	u32 ch0_txdesc_list_haddress;			/* 0x3110 */
> +	u32 ch0_txdesc_list_address;			/* 0x3114 */
> +	u32 ch0_rxdesc_list_haddress;			/* 0x3118 */
> +	u32 ch0_rxdesc_list_address;			/* 0x311c */
> +	u32 unused_3120;				/* 0x3120 */

Is this really unused or should it be the upper 32 bit of
ch0_txdesc_tail_pointer?

> +	u32 ch0_txdesc_tail_pointer;			/* 0x3124 */
> +	u32 unused_3128;				/* 0x3128 */

Same for ch0_rxdesc_tail_pointer.

> +	u32 ch0_rxdesc_tail_pointer;			/* 0x312c */
> +	u32 ch0_txdesc_ring_length;			/* 0x3130 */
> +	u32 ch0_rxdesc_ring_length;			/* 0x3134 */
> +	u32 unused_3138[(0x3160 - 0x3138) / 4];		/* 0x3138 */
> +	u32 ch0_status;					/* 0x3160 */
> +};
> +

Sascha

-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |



  reply	other threads:[~2024-11-04 11:25 UTC|newest]

Thread overview: 15+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-10-29  8:42 [PATCH 00/10] ARM: SoCFPGA: Add initial support for Agilex5 Steffen Trumtrar
2024-10-29  8:42 ` [PATCH 01/10] ARM: socfpga: kconfig: sort entries Steffen Trumtrar
2024-10-29  8:42 ` [PATCH 02/10] mach: socfpga: debug_ll: rework putc_ll Steffen Trumtrar
2024-10-29  8:42 ` [PATCH 03/10] reset: reset-socfpga: build only for 32-bit socfpga Steffen Trumtrar
2024-10-29  8:42 ` [PATCH 04/10] arm: socfgpa: add support for SoCFPGA Agilex5 Steffen Trumtrar
2024-11-04 10:31   ` Sascha Hauer
2024-10-29  8:42 ` [PATCH 05/10] ARM: socfpga: add Arrow AXE5 Agilex5 board Steffen Trumtrar
2024-11-04 10:48   ` Sascha Hauer
2024-10-29  8:42 ` [PATCH 06/10] net: add support for Designware XGMAC (10gb) ethernet Steffen Trumtrar
2024-11-04 11:14   ` Sascha Hauer [this message]
2024-10-29  8:42 ` [PATCH 07/10] net: phy: add Analog Devices ADIN1300 Steffen Trumtrar
2024-10-29  8:42 ` [PATCH 08/10] linux: clk: add clk_parent_data Steffen Trumtrar
2024-10-29  8:42 ` [PATCH 09/10] clk: support init->parent_data Steffen Trumtrar
2024-10-29  8:42 ` [PATCH 10/10] clk: socfpga: add agilex5 clock support Steffen Trumtrar
2024-11-04 11:23   ` Sascha Hauer

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ZyisqoVwDIUO6bJj@pengutronix.de \
    --to=s.hauer@pengutronix.de \
    --cc=barebox@lists.infradead.org \
    --cc=s.trumtrar@pengutronix.de \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox