Re: [PATCH v5 2/2] PCI: qcom: Add support for multi-root port

[Manivannan Sadhasivam]

On Thu, Jul 03, 2025 at 09:48:17AM GMT, Krishna Chaitanya Chundru wrote:
> 
> 
> On 7/2/2025 9:03 PM, Manivannan Sadhasivam wrote:
> > On Wed, Jul 02, 2025 at 04:50:42PM GMT, Krishna Chaitanya Chundru wrote:
> > 
> > [...]
> > 
> > > -	ret = phy_init(pcie->phy);
> > > -	if (ret)
> > > -		goto err_pm_runtime_put;
> > > +	for_each_available_child_of_node(dev->of_node, of_port) {
> > > +		ret = qcom_pcie_parse_port(pcie, of_port);
> > > +		of_node_put(of_port);
> > > +		if (ret) {
> > > +			if (ret != -ENOENT) {
> > > +				dev_err_probe(pci->dev, ret,
> > > +					      "Failed to parse port nodes %d\n",
> > > +					      ret);
> > > +				goto err_port_del;
> > > +			}
> > > +			break;
> > > +		}
> > > +	}
> > > +
> > > +	/*
> > > +	 * In the case of properties not populated in root port, fallback to the
> > > +	 * legacy method of parsing the host bridge node. This is to maintain DT
> > > +	 * backwards compatibility.
> > > +	 */
> > > +	if (ret) {
> > > +		pcie->phy = devm_phy_optional_get(dev, "pciephy");
> > > +		if (IS_ERR(pcie->phy)) {
> > > +			ret = PTR_ERR(pcie->phy);
> > > +			goto err_pm_runtime_put;
> > 
> > Shouldn't this and below be err_port_del?
> > 
> This is a legacy way of parsing property, if the execution
> comes here means the port parsing has failed and ports are not created.
> so err_port_del will not have any effect.
> 

Oops. I got confused by the if (ret) flow. It would be more clear if a goto is
used to indicate that the legacy codeblock is skipped. I'll just incorporate it
while applying.

- Mani

-- 
மணிவண்ணன் சதாசிவம்