<!DOCTYPE html>
<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<p><br>
</p>
<div class="moz-cite-prefix">On 2/10/2026 9:12 PM, Russell King
(Oracle) wrote:<br>
</div>
<blockquote type="cite"
cite="mid:aYsurPBMnzWDZCco@shell.armlinux.org.uk">
<pre wrap="" class="moz-quote-pre">Please note that net-next is currently closed, so you should be sending
net-next patches with "RFC". See section 1.4 of:
<a class="moz-txt-link-freetext" href="https://docs.kernel.org/process/maintainer-netdev.html">https://docs.kernel.org/process/maintainer-netdev.html</a>
On Tue, Feb 10, 2026 at 01:22:26PM +0800, Joey Lu wrote:
</pre>
<blockquote type="cite">
<pre wrap="" class="moz-quote-pre">+static int nvt_gmac_setup(struct platform_device *pdev)
+{
+ struct device *dev = &pdev->dev;
+ phy_interface_t phy_mode;
+ struct regmap *regmap;
+ u32 macid, miscr, reg;
+ int ret;
+
+ regmap = syscon_regmap_lookup_by_phandle_args(dev->of_node, "nuvoton,sys", 1, &macid);
+ if (IS_ERR(regmap))
+ ret = dev_err_probe(dev, PTR_ERR(regmap), "Failed to get sys register\n");
+
+ if (macid > 1)
+ ret = dev_err_probe(dev, -EINVAL, "Invalid sys arguments\n");
+
+ if (of_get_phy_mode(pdev->dev.of_node, &phy_mode))
+ ret = dev_err_probe(dev, -EINVAL, "Missing phy mode property\n");
</pre>
</blockquote>
<pre wrap="" class="moz-quote-pre">
If you pass in the plat_dat to this function, then you have access to
plat_dat->phy_interface. I went through all the dwmac glue code and
removed such PHY interface gets, so please don't introduce new
instances.
</pre>
</blockquote>
I'll drop it.
<blockquote type="cite"
cite="mid:aYsurPBMnzWDZCco@shell.armlinux.org.uk">
<pre wrap="" class="moz-quote-pre">
</pre>
<blockquote type="cite">
<pre wrap="" class="moz-quote-pre">+
+ miscr = (macid == 0) ? NVT_REG_SYS_GMAC0MISCR : NVT_REG_SYS_GMAC1MISCR;
</pre>
</blockquote>
<pre wrap="" class="moz-quote-pre">
I find the use of "miscr" to be the register offset, and "reg" to be
the register value is confusing. Normally they're the other way around.
</pre>
</blockquote>
I'll update the naming to follow the usual convention, for example,
using reg for offset and val for value.
<blockquote type="cite"
cite="mid:aYsurPBMnzWDZCco@shell.armlinux.org.uk">
<pre wrap="" class="moz-quote-pre">
</pre>
<blockquote type="cite">
<pre wrap="" class="moz-quote-pre">+
+ switch (phy_mode) {
+ case PHY_INTERFACE_MODE_RGMII:
+ case PHY_INTERFACE_MODE_RGMII_ID:
+ case PHY_INTERFACE_MODE_RGMII_RXID:
+ case PHY_INTERFACE_MODE_RGMII_TXID:
+ ret = nvt_gmac_get_delay(dev, "rx-internal-delay-ps");
+ if (ret < 0)
+ return ret;
+ reg = FIELD_PREP(NVT_RX_DELAY_MASK, ret);
+
+ ret = nvt_gmac_get_delay(dev, "tx-internal-delay-ps");
+ if (ret < 0)
+ return ret;
+ reg |= FIELD_PREP(NVT_TX_DELAY_MASK, ret);
+ break;
+ case PHY_INTERFACE_MODE_RMII:
+ reg = NVT_MISCR_RMII;
+ break;
+ default:
+ return dev_err_probe(dev, -EINVAL, "Unsupported phy-mode (%d)\n", phy_mode);
+ }
+
+ regmap_update_bits(regmap, miscr,
+ NVT_RX_DELAY_MASK | NVT_TX_DELAY_MASK | NVT_MISCR_RMII, reg);
</pre>
</blockquote>
<pre wrap="" class="moz-quote-pre">
Please consider programming the NVT_MISCR_RMII bit via the
plat_dat->set_phy_intf_sel() method.</pre>
</blockquote>
I'll switch to using the <span style="white-space: pre-wrap">plat_dat->set_phy_intf_sel() callback and pass the required info through private structure.</span>
<blockquote type="cite"
cite="mid:aYsurPBMnzWDZCco@shell.armlinux.org.uk">
<pre wrap="" class="moz-quote-pre">
Is this register preserved over suspend/resume ?</pre>
</blockquote>
<p>It is not preserved by the GMAC suspend/resume flow and therefore
needs to be treated as a system register rather than a GMAC local
context.</p>
<blockquote type="cite"
cite="mid:aYsurPBMnzWDZCco@shell.armlinux.org.uk">
<pre wrap="" class="moz-quote-pre">
Thanks.
</pre>
</blockquote>
</body>
</html>