Re: [PATCH 1/1] - sgiioc4: fixup use of mmio ops

[Jeff Garzik <jeff@garzik.org>]

John Keller wrote:
>>>> -	if (!request_region(base, IOC4_CMD_CTL_BLK_SIZE, hwif->name)) {
>>>> +	cmd_phys_base = bar0 + IOC4_CMD_OFFSET;
>>>> +	if (!request_mem_region(cmd_phys_base, IOC4_CMD_CTL_BLK_SIZE,
>>>> +	    hwif->name)) {
>>>>  		printk(KERN_ERR
>>>> -			"%s : %s -- ERROR, Port Addresses "
>>>> +			"%s : %s -- ERROR, Addresses "
>>>>  			"0x%p to 0x%p ALREADY in use\n",
>>>> -		       __FUNCTION__, hwif->name, (void *) base,
>>>> -		       (void *) base + IOC4_CMD_CTL_BLK_SIZE);
>>>> +		       __FUNCTION__, hwif->name, (void *) cmd_phys_base,
>>>> +		       (void *) cmd_phys_base + IOC4_CMD_CTL_BLK_SIZE);
>> If 'void __iomem *' were used, no casts would be needed here
> 
> So, 'void __iomem *' should also be used for physical (non-mapped)
> addresses, as in this case?


Ooops, no, just cookies returned from ioremap() and ioremap_nocache().

	Jeff