Re: [PATCH v2 2/4] mfd: ioc3: Add driver for SGI IOC3 chip

[Esben Haabendal]

On Tue, 09 Apr 2019, Thomas Bogendoerfer wrote:
> 
> diff --git a/drivers/net/ethernet/sgi/ioc3-eth.c b/drivers/net/ethernet/sgi/ioc3-eth.c
> index 358e66b81926..21fe722ebcd8 100644
> --- a/drivers/net/ethernet/sgi/ioc3-eth.c
> +++ b/drivers/net/ethernet/sgi/ioc3-eth.c
>  
>  [ ... ]
>  
> -	err = pci_request_regions(pdev, "ioc3");

Why are you dropping the call to pci_request_regions()?  Shouldn't you
do something similar in the new mfd driver?

When you are use the the BAR 0 resource as mem_base argument to
mfd_add_devices() later on, it will be split into child resources for
the child devices, but they will not be related to the IORESOURCE_MEM
root tree (iomem_resource) anymore.  I don't think that is how it is
supposed to be done, as it will allow random other drivers to request
the exact same memory area.

How/where is the memory resources inserted in the root IORESOURCE_MEM
resource (iomem_resource)?  Or is it allowed to use resources without
inserting it into the root tree?

> +	SET_NETDEV_DEV(dev, &pdev->dev);
> +	ip = netdev_priv(dev);
> +	r = platform_get_resource(pdev, IORESOURCE_MEM, 0);
> +	ip->regs = ioremap(r->start, resource_size(r));
> +	r = platform_get_resource(pdev, IORESOURCE_MEM, 1);
> +	ip->ssram = ioremap(r->start, resource_size(r));

Maybe use devm_platform_ioremap_resource() instead, which handles both
platform_get_resource() and ioremap() in one call..

/Esben