mach/io.h cleanup and removal question

[Rob Herring]

On 06/08/2012 06:43 AM, Russell King - ARM Linux wrote:
> On Fri, Jun 08, 2012 at 12:20:40PM +0200, Andrew Lunn wrote:
>> Your patchset for mach/io.h cleanup and remove, causes problems on
>> Orion5x.
>>
>> mach/io.h for orion5x had a:
>>
>> #define IO_SPACE_LIMIT		0xffffffff
>>
>> which got removed. This results in a panic at boot:
>>
>>         /*
>>          * IORESOURCE_IO
>>          */
>>         sys->io_offset = 0;
>>         res[0].name = "PCIe I/O Space";
>>         res[0].flags = IORESOURCE_IO;
>>         res[0].start = ORION5X_PCIE_IO_BUS_BASE;
>>         res[0].end = res[0].start + ORION5X_PCIE_IO_SIZE - 1;
>>         if (request_resource(&ioport_resource, &res[0]))
>>                 panic("Request PCIe IO resource failed\n");
>>
>> ORION5X_PCIE_IO_SIZE is 1MB, so the allocation fails because of the
>> 64K default.
>>
>> arch/arm/include/asm/io.h has the comment:
>>
>> /*
>>  * This is the limit of PC card/PCI/ISA IO space, which is by default
>>  * 64K if we have PC card, PCI or ISA support.  Otherwise, default to
>>  * zero to prevent ISA/PCI drivers claiming IO space (and potentially
>>  * oopsing.)
>>  *
>>  * Only set this larger if you really need inb() et.al. to operate over
>>  * a larger address space.  Note that SOC_COMMON ioremaps each sockets
>>  * IO space area, and so inb() et.al. must be defined to operate as per
>>  * readb() et.al. on such platforms.
>>  */
>>
>> Now, i know nothing about how PCI works... So i have a question:
>>
>> Which is better, put back parts of io.h so allowing the 1MB
>> request_resource, or reduce ORION5X_PCIE_IO_SIZE to 64KB, since from
>> the comment it is unlikely an PCI card needs more than 64KB?
> 
> Well, the comment about SOC_COMMON doesn't apply as you're not using that.
> 
> Probably shrinking ORION5X_PCIE_IO_SIZE to 64K, unless we really have a
> requirement to support a larger IO space.

Agreed. The next step is moving all i/o space to a fixed virtual address
and for this we want to make i/o windows 64K.

However, I think you may also have a problem with
ORION5X_PCIE_IO_BUS_BASE and the resource start. The i/o resource start
should really be 0 and then the __io() macro should add the virtual i/o
base address. However, Russell has mentioned previously that some chips
may not work correctly with i/o space at 0 (PCI bus addresses, not host).

Do you have cards with i/o that you can test or know what devices are
used with this platform?

Rob