Re: [PATCH net-next v3 2/4] net: ethernet: altera-tse: Convert to mdio-regmap and use PCS Lynx

Hello Simon,

On Fri, 26 May 2023 10:39:08 +0200
Simon Horman [off-list ref] wrote:

On Fri, May 26, 2023 at 09:42:50AM +0200, Maxime Chevallier wrote:

quoted

The newly introduced regmap-based MDIO driver allows for an easy
mapping of an mdiodevice onto the memory-mapped TSE PCS, which is
actually a Lynx PCS.

Convert Altera TSE to use this PCS instead of the pcs-altera-tse,
which is nothing more than a memory-mapped Lynx PCS.

Signed-off-by: Maxime Chevallier <maxime.chevallier@bootlin.com>  

Hi Maxime,

I have some concerns about the error paths in this patch.

Heh I didn't do a very good job in that regard indeed :/ I'll address
these, but I'll wait for Russell series to make it through though.

Thanks for spotting this.

...

quoted

@@ -1134,13 +1136,21 @@ static int altera_tse_probe(struct

platform_device *pdev) const struct of_device_id *of_id = NULL;
 	struct altera_tse_private *priv;
 	struct resource *control_port;
+	struct regmap *pcs_regmap;
 	struct resource *dma_res;
 	struct resource *pcs_res;
+	struct mii_bus *pcs_bus;
 	struct net_device *ndev;
 	void __iomem *descmap;
-	int pcs_reg_width = 2;
 	int ret = -ENODEV;
 
+	struct regmap_config pcs_regmap_cfg;  

nit: this probably belongs in with the bunch of declarations above it.

quoted

+
+	struct mdio_regmap_config mrc = {
+		.parent = &pdev->dev,
+		.valid_addr = 0x0,
+	};  

nit: maybe this too.

quoted

+
 	ndev = alloc_etherdev(sizeof(struct altera_tse_private));
 	if (!ndev) {
 		dev_err(&pdev->dev, "Could not allocate network
device\n"); @@ -1258,10 +1268,29 @@ static int
altera_tse_probe(struct platform_device *pdev) ret =
request_and_map(pdev, "pcs", &pcs_res, &priv->pcs_base);
 	if (ret) {
+		/* If we can't find a dedicated resource for the
PCS, fallback
+		 * to the internal PCS, that has a different
address stride
+		 */
 		priv->pcs_base = priv->mac_dev +
tse_csroffs(mdio_phy0);
-		pcs_reg_width = 4;
+		pcs_regmap_cfg.reg_bits = 32;
+		/* Values are MDIO-like values, on 16 bits */
+		pcs_regmap_cfg.val_bits = 16;
+		pcs_regmap_cfg.reg_shift = REGMAP_UPSHIFT(2);
+	} else {
+		pcs_regmap_cfg.reg_bits = 16;
+		pcs_regmap_cfg.val_bits = 16;
+		pcs_regmap_cfg.reg_shift = REGMAP_UPSHIFT(1);
 	}
 
+	/* Create a regmap for the PCS so that it can be used by
the PCS driver */
+	pcs_regmap = devm_regmap_init_mmio(&pdev->dev,
priv->pcs_base,
+					   &pcs_regmap_cfg);
+	if (IS_ERR(pcs_regmap)) {
+		ret = PTR_ERR(pcs_regmap);
+		goto err_free_netdev;
+	}
+	mrc.regmap = pcs_regmap;
+
 	/* Rx IRQ */
 	priv->rx_irq = platform_get_irq_byname(pdev, "rx_irq");
 	if (priv->rx_irq == -ENXIO) {

@@ -1384,7 +1413,20 @@ static int altera_tse_probe(struct

platform_device *pdev) (unsigned long) control_port->start,
priv->rx_irq, priv->tx_irq);
 
-	priv->pcs = alt_tse_pcs_create(ndev, priv->pcs_base,
pcs_reg_width);
+	snprintf(mrc.name, MII_BUS_ID_SIZE, "%s-pcs-mii",
ndev->name);
+	pcs_bus = devm_mdio_regmap_register(&pdev->dev, &mrc);
+	if (IS_ERR(pcs_bus)) {
+		ret = PTR_ERR(pcs_bus);
+		goto err_init_phy;
+	}
+
+	priv->pcs_mdiodev = mdio_device_create(pcs_bus, 0);  

mdio_device_create() can fail. Should that be handled here?

quoted

+
+	priv->pcs = lynx_pcs_create(priv->pcs_mdiodev);
+	if (!priv->pcs) {
+		ret = -ENODEV;
+		goto err_init_phy;  

Does this leak priv->pcs_mdiodev?

quoted

+	}
 
 	priv->phylink_config.dev = &ndev->dev;
 	priv->phylink_config.type = PHYLINK_NETDEV;

@@ -1407,11 +1449,12 @@ static int altera_tse_probe(struct

platform_device *pdev) if (IS_ERR(priv->phylink)) {
 		dev_err(&pdev->dev, "failed to create phylink\n");
 		ret = PTR_ERR(priv->phylink);
-		goto err_init_phy;
+		goto err_pcs;  

Does this leak priv->pcs ?

quoted

 	}
 
 	return 0;
-
+err_pcs:
+	mdio_device_free(priv->pcs_mdiodev);
 err_init_phy:
 	unregister_netdev(ndev);
 err_register_netdev:

@@ -1433,6 +1476,8 @@ static int altera_tse_remove(struct

platform_device *pdev) altera_tse_mdio_destroy(ndev);
 	unregister_netdev(ndev);
 	phylink_destroy(priv->phylink);
+	mdio_device_free(priv->pcs_mdiodev);
+
 	free_netdev(ndev);
 
 	return 0;

diff --git a/include/linux/mdio/mdio-regmap.h

b/include/linux/mdio/mdio-regmap.h index b8508f152552..679d9069846b
100644 --- a/include/linux/mdio/mdio-regmap.h

+++ b/include/linux/mdio/mdio-regmap.h

@@ -7,6 +7,8 @@
 #ifndef MDIO_REGMAP_H
 #define MDIO_REGMAP_H
 
+#include <linux/phy.h>
+
 struct device;
 struct regmap;
   

This hunk doesn't seem strictly related to the patch.
Perhaps the include belongs elsewhere.
Or the hunk belongs in another patch.

Indeed, I had the same issue in another patch in this series. I'll
re-split everything correctly.

Thank you for the review,

Maxime