Re: [PATCH 2/2] misc: tc9564: introduce base PCI driver

On 8/13/26 8:59 PM, [email protected] wrote:
> Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
> - [High] Missing PCI IRQ vector allocation prevents MSI/MSI-X usage for high-speed child devices.
> --
> 
> --- Patch [2]: [PATCH 2/2] misc: tc9564: introduce base PCI driver ---
> commit e66bfe4041c21f48829a22eb54cde73756806870
> Author: Alex Elder <[email protected]>
> misc: tc9564: introduce base PCI driver
> 
> This commit introduces a base PCI driver for the Toshiba TC9564 SoC which
> implements a PCIe switch and Ethernet AVB/TSN bridge. The driver binds to the
> embedded PCI endpoint functions, performs basic PCI initialization, and uses
> devicetree to populate any endpoint bus children.
> 
> [ ... ]
> 
>> diff --git a/drivers/misc/tc9564-pci.c b/drivers/misc/tc9564-pci.c
>> new file mode 100644
>> index 0000000000000..d3ffcac1c522f
>> --- /dev/null
>> +++ b/drivers/misc/tc9564-pci.c
> 
> [ ... ]
> 
>> +static int
>> +tc9564_function_probe(struct pci_dev *pdev, const struct pci_device_id *id)
>> +{
>> +	struct device *dev = &pdev->dev;
>> +	struct device_node *np;
>> +	int ret;
>> +
>> +	/* Despite being a PCI device, we require devicetree */
>> +	np = dev_of_node(dev);
>> +	if (!np)
>> +		return dev_err_probe(dev, -EINVAL, "no devicetree node\n");
>> +
>> +	ret = pcim_enable_device(pdev);
>> +	if (ret)
>> +		return ret;
>> +
>> +	pci_set_master(pdev);
>> +
>> +	/* Scan for pci-ep-bus nodes and probe their sub-devices */
> 
> [Severity: High]
> Does this setup sequence omit PCI IRQ vector allocation?
> 
> Without calling pci_alloc_irq_vectors() in tc9564_function_probe() before
> populating the child devices, the Linux PCI core won't configure the device's
> MSI/MSI-X capabilities with host addresses and data payloads.

I will investigate this.  I don't have a good answer at the moment.

					-Alex

> Since child platform drivers cannot safely allocate PCI IRQ vectors on behalf
> of their parent PCI device, will the high-speed child devices (like the 10Gbps
> Ethernet XGMACs) be forced to share a single legacy INTx interrupt, or fail
> entirely if the host platform does not support legacy PCI interrupts?
> 
>> +	ret = of_platform_default_populate(np, NULL, dev);
>> +	if (ret)
>> +		goto err_clear_master;
>> +
>> +	return 0;
>> +
>> +err_clear_master:
>> +	pci_clear_master(pdev);
>> +
>> +	return dev_err_probe(dev, ret, "failed to populate platform bus\n");
>> +}
> 
> [ ... ]
>