On Thu, 8 May 2025 at 00:24, Andrew Lunn [off-list ref] wrote:

On Wed, May 07, 2025 at 09:13:05AM +0200, Herve Codina wrote:

quoted

Only one device-tree overlay (lan966x_evb_lan9662_nic.dtbo) is handled
and this overlay is directly referenced in lan966x_pci_load_overlay().

This avoid to use the code for an other board.

In order to be more generic and to allow support for other boards (PCI
Vendor/Device IDs), introduce the lan966x_pci_info structure and attach
it to PCI Vendor/Device IDs handled by the driver.

This structure contains information related to the PCI board such as
information related to the dtbo describing the board we have to load.

Signed-off-by: Herve Codina <herve.codina@bootlin.com>

How big is the dtbo ?

This is going in the right direction. I'm just wondering if each dtbo
should be wrapped in its own very slim PCI driver, which simply
registers its lan966x_pci_info structure to a core driver. Only the
needed dtbo will then be loaded into memory as a module, not them all.

Alternatively, the dtbo could be loaded through request_firmware().
That could lead to a generic support option in the PCI core, which would
fallback to loading pci-<vid>-<pid>.dtbo when no driver is available.

Pretty much all the pieces are here, so it can be done later.

Exactly.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds