[PATCH V5 1/3] pci: Add PCIe driver for Samsung Exynos

Arnd Bergmann arnd at arndb.de
Mon Jun 17 22:44:40 EST 2013


On Monday 17 June 2013 18:45:52 Jingoo Han wrote:
> On Friday, June 14, 2013 9:54 PM, Arnd Bergmann wrote:
> > 
> > Please look up the documentation about inbound viewport and describe
> > in a code comment what it does. I /assume/ that this is how DMA accesses
> > from the bus get translated into AXI bus transactions. If so, you have
> > to let the window translate addresses from RAM. If it's something else,
> > then you should document what it is and how it needs to be set up.
> 
> One of our hardware engineer confirmed it.
> He said that these inbound functions are unnecessary.
> Also, I checked that PCIe works properly without these functions.
> So, I will remove these inbound functions.

Ok, good. So DMA just gets translated 1:1 independent of the
inbound viewport? Have you tested this with PCI device using
DMA?

> static int exynos_pcie_setup(int nr, struct pci_sys_data *sys)
> {
> 	struct pcie_port *pp;
> 
> 	pp = sys_to_pcie(sys);
> 
> 	if (!pp)
> 		return 0;
> 
> 	if (global_io_offset < SZ_1M && pp->config.io_size > 0) {
> 		sys->io_offset = global_io_offset - pp->config.io_bus_addr; /* normally 0 */
> 		pci_ioremap_io(sys->io_offset, pp->io.start);
> 		global_io_offset += SZ_64K;
> 	}
> 
> 	sys->mem_offset = pp->mem.start - pp->config.mem_bus_addr; /* normally 0 */
> 
> 	pci_add_resource_offset(&sys->resources, &pp->io, sys->io_offset);
> 	pci_add_resource_offset(&sys->resources, &pp->mem, sys->mem_offset);
> 
>         return 1;
> }

This is what I meant, yes.

> In this case, boot message is as below:
> 
> PCI host bridge to bus 0000:00
> pci_bus 0000:00: root bus resource [io  0x40001000-0x40010fff]
> pci_bus 0000:00: root bus resource [mem 0x40011000-0x5fffffff]
> pci_bus 0000:00: No busn resource found for root bus, will use [bus 00-ff]
> pci 0000:00:00.0: [144d:a549] type 01 class 0x060400
> [.....]
> PCI host bridge to bus 0001:00
> pci_bus 0001:00: root bus resource [io  0x60001000-0x60010fff] (bus address [0x5fff1000-0x6000
> 0fff])
> pci_bus 0001:00: root bus resource [mem 0x60011000-0x7fffffff]
> pci_bus 0001:00: No busn resource found for root bus, will use [bus 00-ff]
> pci 0001:00:00.0: [144d:a549] type 01 class 0x060400

The io resources here look wrong. I would have expected

pci_bus 0000:00: root bus resource [io  0x001000-0x00ffff]
pci_bus 0001:00: root bus resource [io  0x010000-0x01ffff] (bus address [0x000000-0x00ffff])

Please have a look at the pci-mvebu driver and how it calculates its
'realio' resource.

> > > > > +static int __exit exynos_pcie_remove(struct platform_device *pdev)
> > > > > +{
> > > > > +	struct pcie_port *pp = platform_get_drvdata(pdev);
> > > > > +
> > > > > +	clk_disable_unprepare(pp->bus_clk);
> > > > > +	clk_disable_unprepare(pp->clk);
> > > > > +
> > > > > +	return 0;
> > > > > +}
> > > >
> > > > You also don't remove the PCI devices here, as mentioned in an earlier
> > > > review.
> > >
> > > I reviewed Marvell PCIe driver and Tegra PCIe driver; however,
> > > I cannot know what you mean.
> > >
> > > Could let me know which additional functions are needed?
> > 
> > The mvebu driver does not allow module unload. I haven't looked at the
> > tegra driver. If you allow unloading the driver and provide a 'remove'
> > callback, that callback needs to clean up the entire bus and remove
> > all child devices that were added as well as undo everything the
> > probe function did. I think it would be great if you can do that, although
> > it might not be easy. The simplest solution would be to not support
> > unloading though.
> 
> As the mvebu driver uses platform_driver_probe(), the Exynos driver uses
> platform_driver_probe(). Thus, I will not provide a 'remove' callback.

Well, the important part is not to provide a module_exit() function, which
will ensure the driver cannot be unloaded.

	Arnd


More information about the devicetree-discuss mailing list