Re: [PATCH v3 4/4] PCI: tegra194: Add PCI_ENDPOINT dependency

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [Medium] PCI: tegra194: host Kconfig depends on PCI_ENDPOINT instead of EPC stub
--

commit b82fa7e55a14cc1575d1ffe0c356f50c758354e7
Author: Manikanta Maddireddy <mmaddireddy@nvidia.com>

PCI: tegra194: Add PCI_ENDPOINT dependency

This commit adds a PCI_ENDPOINT dependency to PCIE_TEGRA194_HOST in Kconfig so
that pci_epc_linkdown(), called from the PERST# assertion handler, is declared
when building the Tegra194 PCIe driver.

Closes: https://lore.kernel.org/oe-kbuild-all/202610230923.2lf3BsLs-lkp@intel.com/

> diff --git a/drivers/pci/controller/dwc/Kconfig b/drivers/pci/controller/dwc/Kconfig
> index 0a0d56fcf0ac9..7041346849002 100644
> --- a/drivers/pci/controller/dwc/Kconfig
> +++ b/drivers/pci/controller/dwc/Kconfig
> @@ -245,6 +245,7 @@ config PCIE_TEGRA194_HOST
>  	tristate "NVIDIA Tegra194 (and later) PCIe controller (host mode)"
>  	depends on (ARCH_TEGRA && ARM64) || COMPILE_TEST
>  	depends on PCI_MSI
> +	depends on PCI_ENDPOINT

[Severity: Medium]
Does adding a PCI_ENDPOINT dependency to PCIE_TEGRA194_HOST disable Root
Complex support on existing host-only configurations where CONFIG_PCI_ENDPOINT
is not enabled (such as during olddefconfig), or force host-only kernels to
build the unused PCI Endpoint framework?

In drivers/pci/controller/dwc/Kconfig, Tegra194 PCIe support is split into
PCIE_TEGRA194_HOST and PCIE_TEGRA194_EP, where PCIE_TEGRA194_EP already
depends on PCI_ENDPOINT, and none of the other dual-mode DesignWare host
Kconfig symbols depend on PCI_ENDPOINT.

In drivers/pci/controller/dwc/pcie-tegra194.c, pex_ep_event_pex_rst_deassert()
already calls pci_epc_deinit_notify() and pci_epc_init_notify()
unconditionally, and those build when CONFIG_PCI_ENDPOINT=n because commit
473b2cf9c4d1 ("PCI: endpoint: Introduce 'epc_deinit' event and notify the EPF
drivers") added static inline no-op stubs in include/linux/pci-epc.h:

#ifdef CONFIG_PCI_ENDPOINT
void pci_epc_init_notify(struct pci_epc *epc);
void pci_epc_deinit_notify(struct pci_epc *epc);
#else
static inline void pci_epc_init_notify(struct pci_epc *epc)
{
}

static inline void pci_epc_deinit_notify(struct pci_epc *epc)
{
}
#endif /* CONFIG_PCI_ENDPOINT */

Would it be better to provide a matching static inline no-op stub for
pci_epc_linkdown() under #ifdef CONFIG_PCI_ENDPOINT in
include/linux/pci-epc.h, ordered before or squashed into commit 9fcf1c9f540e
("PCI: tegra194: Notify Endpoint functions on PERST# assert") (referenced in
the commit message via the stale SHA bf644226574e) so that host-only configs
and git bisect are not broken?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261102111450.4010818-1-mmaddireddy@nvidia.com?part=4