[RFC 2/4] PCI: generic: Add support for ARM64 and MSI(x)

[Lorenzo Pieralisi]

On Wed, Oct 01, 2014 at 10:38:45AM +0100, Arnd Bergmann wrote:

[...]

> The arm32 implementations of pci_domain_nr/pci_proc_domain can probably be
> removed if we change the arm32 pcibios_init_hw function to call the new
> interfaces that set the domain number.

I wished, but it is a bit more complicated than I thought unfortunately,
mostly because some drivers, eg cns3xxx set the domain numbers
statically in pci_sys_data and this sets a chain of dependency that is
not easy to untangle. I think cns3xxx is the only legacy driver that "uses"
the domain number (in pci_sys_data) in a way that clashes with the
generic domain_nr implementation, I need to give it more thought.

> pci_mmap_page_range could either get generalized some more in an attempt
> to have a __weak default implementation that works on ARM, or it could
> be changed to lose the dependency on pci_sys_data instead. In either
> case, the change would involve using the generic pci_host_bridge_window
> list.

I need to repost my series, but I *think* we can consider the dependency on
pci_sys_data gone in pci_mmap_page_range().

> pcibios_align_resource should probably be per host, and we could move
> that into a pointer in pci_host_bridge, something like this:

Yes, and that's likely to be true for add_bus too. I wonder what's the
best course of action. Putting together all the bits and pieces required
to remove PCI bios dependency from this patch can take a while, I wonder
whether we should aim for merging this driver (rebased on top of my port to the
new parse ranges API) with the ARM/ARM64 ifdeffery and clean it up later
or aim for the whole thing at once, I am just worried it can take us a while.

Lorenzo

> 
> diff --git a/drivers/pci/setup-res.c b/drivers/pci/setup-res.c
> index b7c3a5ea1fca..d9cb6c916d54 100644
> --- a/drivers/pci/setup-res.c
> +++ b/drivers/pci/setup-res.c
> @@ -200,11 +200,15 @@ static int pci_revert_fw_address(struct resource *res, struct pci_dev *dev,
>  static int __pci_assign_resource(struct pci_bus *bus, struct pci_dev *dev,
>  		int resno, resource_size_t size, resource_size_t align)
>  {
> +	struct pci_host_bridge *host = find_pci_host_bridge(bus);
> +	resource_size_t (*alignf)(void *, const struct resource *,
> +				  resource_size_t, resource_size_t),
>  	struct resource *res = dev->resource + resno;
>  	resource_size_t min;
>  	int ret;
>  
>  	min = (res->flags & IORESOURCE_IO) ? PCIBIOS_MIN_IO : PCIBIOS_MIN_MEM;
> +	alignf = host->align_resource ?: pcibios_align_resource;
>  
>  	/*
>  	 * First, try exact prefetching match.  Even if a 64-bit
> @@ -215,7 +219,7 @@ static int __pci_assign_resource(struct pci_bus *bus, struct pci_dev *dev,
>  	 */
>  	ret = pci_bus_alloc_resource(bus, res, size, align, min,
>  				     IORESOURCE_PREFETCH | IORESOURCE_MEM_64,
> -				     pcibios_align_resource, dev);
> +				     alignf, dev);
>  	if (ret == 0)
>  		return 0;
>  
> @@ -227,7 +231,7 @@ static int __pci_assign_resource(struct pci_bus *bus, struct pci_dev *dev,
>  	     (IORESOURCE_PREFETCH | IORESOURCE_MEM_64)) {
>  		ret = pci_bus_alloc_resource(bus, res, size, align, min,
>  					     IORESOURCE_PREFETCH,
> -					     pcibios_align_resource, dev);
> +					     alignf, dev);
>  		if (ret == 0)
>  			return 0;
>  	}
> @@ -240,7 +244,7 @@ static int __pci_assign_resource(struct pci_bus *bus, struct pci_dev *dev,
>  	 */
>  	if (res->flags & (IORESOURCE_PREFETCH | IORESOURCE_MEM_64))
>  		ret = pci_bus_alloc_resource(bus, res, size, align, min, 0,
> -					     pcibios_align_resource, dev);
> +					     alignf, dev);
>  
>  	return ret;
>  }
> 
> 
> If we decide constantly calling find_pci_host_bridge() is too expensive, we can
> be more clever about it.
> 
> 	Arnd
> 
>