Re: [PATCH v2] PCI/portdrv: Allow probing even without child services

[Brian Norris]

On Thu, Apr 09, 2026 at 04:20:27PM -0700, Brian Norris wrote:
> So is pci_set_master() a one-way operation done in child services (never
> call pci_clear_master()), and the only clear is via
> pcie_portdrv_remove() -> pci_disable_device()?

One other thought about this: it's actually a non-trivial question as to
whether the port is actually going to use MSI (and therefore, need bus
mastering). It depends on the result of pcie_init_service_irqs(),
whether we succeed with MSI or INTx.

We don't actually attempt to answer that question today. But if we did,
we'd have to be scattering that code across all our IRQ-utilizing child
services.

I wonder if all that churn is worthwhile, when it's a relatively
rare/strange case that a port really does *not* want bus mastering
enabled.

Brian