On 18/12/2020 19:36:08-0400, Jason Gunthorpe wrote: > On Fri, Dec 18, 2020 at 10:16:58PM +0100, Alexandre Belloni wrote: > > > But then again, what about non-enumerable devices on the PCI device? I > > feel this would exactly fit MFD. This is a collection of IPs that exist > > as standalone but in this case are grouped in a single device. > > So, if mfd had a mfd_device and a mfd bus_type then drivers would need > to have both a mfd_driver and a platform_driver to bind. Look at > something like drivers/char/tpm/tpm_tis.c to see how a multi-probe > driver is structured > > See Mark's remarks about the old of_platform_device, to explain why we > don't have a 'dt_device' today > So, what would that mfd_driver have that the platform_driver doesn't already provide? > > Note that I then have another issue because the kernel doesn't support > > irq controllers on PCI and this is exactly what my SoC has. But for now, > > I can just duplicate the irqchip driver in the MFD driver. > > I think Thomas fixed that recently on x86 at least.. > > Having to put dummy irq chip drivers in MFD anything sounds scary :| > This isn't a dummy driver it is a real irqchip, what issue is there to register an irqchip from MFD ? > > Let me point to drivers/net/ethernet/cadence/macb_pci.c which is a > > fairly recent example. It does exactly that and I'm not sure you could > > do it otherwise while still not having to duplicate most of macb_probe. > > Creating a platform_device to avoid restructuring the driver's probe > and device logic to be generic is a *really* horrible reason to use a > platform device. > Definitively but it made it in and seemed reasonable at the time it seems. I stumbled upon that a while ago because I wanted to remove platform_data support from the macb driver and this is the last user. I never got the time to tackle that. -- Alexandre Belloni, Bootlin Embedded Linux and Kernel engineering https://bootlin.com
