Some doubts about mt7621-pci mainline place

Some doubts about mt7621-pci mainline place

[Sergio Paracuellos]

Hi,

I'd like to mainline for the next merge window the current mt7621-pci
driver which is in staging now. I am in the process of validating
current bindings to make first all the need changes in staging before.
This SoC is based on mips ralink platform. MediaTek MT7621 PCIe subsys
supports single Root complex (RC) with 3 Root Ports. Each Root Ports
supports a Gen1 1-lane Link. I have some concerns that I am not sure
how to achieve before getting this an attempt to put in its correct
place of the kernel. This driver has been cleaned during its long time
in staging and is properly using the pci generic APIs instead of the
PCI_LEGACY which seems that all other drivers in 'arch/mips/pci' are
using. The driver is generic enough but it needs mips iocu region
configuration that is done in the driver itself. So here it is where
my questions come. Is 'arch/mips/pci' the place to move this driver?
Is the driver the correct place to configure the iocu related with
pci?

There is also one issue from a while now that sometimes is reported by
the kernel test robot [0] that I'd like also to fix correctly. please
see the link and my comments there.

Thanks in advance for your comments and help.

Best regards,
    Sergio Paracuellos

[0]: https://www.spinics.net/lists/kernel/msg3617511.html

Re: Some doubts about mt7621-pci mainline place

[Thomas Bogendoerfer]

On Mon, May 10, 2021 at 08:43:51AM +0200, Sergio Paracuellos wrote:
> in staging and is properly using the pci generic APIs instead of the
> PCI_LEGACY which seems that all other drivers in 'arch/mips/pci' are
> using. The driver is generic enough but it needs mips iocu region
> configuration that is done in the driver itself. So here it is where
> my questions come. Is 'arch/mips/pci' the place to move this driver?

moving to arch/mips/pci would work for me. Could you look at converting
the driver to PCI_DRIVERS_GENERIC instead of LEGACY ? I want to
get rid of PCI_LEGACY...

> Is the driver the correct place to configure the iocu related with
> pci?

I see no problem with this at the moment.

Thomas.

-- 
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea.                                                [ RFC1925, 2.3 ]

Re: Some doubts about mt7621-pci mainline place

[Sergio Paracuellos]

Hi Thomas,

On Tue, May 11, 2021 at 10:48 PM Thomas Bogendoerfer
<tsbogend@alpha.franken.de> wrote:
>
> On Mon, May 10, 2021 at 08:43:51AM +0200, Sergio Paracuellos wrote:
> > in staging and is properly using the pci generic APIs instead of the
> > PCI_LEGACY which seems that all other drivers in 'arch/mips/pci' are
> > using. The driver is generic enough but it needs mips iocu region
> > configuration that is done in the driver itself. So here it is where
> > my questions come. Is 'arch/mips/pci' the place to move this driver?
>
> moving to arch/mips/pci would work for me. Could you look at converting
> the driver to PCI_DRIVERS_GENERIC instead of LEGACY ? I want to
> get rid of PCI_LEGACY...

Driver is already using PCI_DRIVERS_GENERIC, so as soon as I get
bindings validated I will send patches to move into 'arch/mips/pci'.

>
> > Is the driver the correct place to configure the iocu related with
> > pci?
>
> I see no problem with this at the moment.

Ok, good to know :).

>
> Thomas.
>

Thanks,
    Sergio Paracuellos
> --
> Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
> good idea.                                                [ RFC1925, 2.3 ]