Re: [PATCH] PCI: Add cast when assigning PCI_ROM_ADDRESS_MASK to a 32-bit variable

[Mason]

On 11/04/2017 03:24, Bjorn Helgaas wrote:

> On Mon, Apr 10, 2017 at 03:24:57PM -0700, Matthias Kaehlcke wrote:
>
>> This fixes a clang warning about "implicit conversion from 'unsigned
>> long' to 'u32'"
>>
>> Signed-off-by: Matthias Kaehlcke <mka@xxxxxxxxxxxx>
>> ---
>>  drivers/pci/probe.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/pci/probe.c b/drivers/pci/probe.c
>> index dfc9a2794141..148e80d5caf1 100644
>> --- a/drivers/pci/probe.c
>> +++ b/drivers/pci/probe.c
>> @@ -180,7 +180,7 @@ int __pci_read_base(struct pci_dev *dev, enum pci_bar_type type,
>>  	u16 orig_cmd;
>>  	struct pci_bus_region region, inverted_region;
>>  
>> -	mask = type ? PCI_ROM_ADDRESS_MASK : ~0;
>> +	mask = type ? (u32)PCI_ROM_ADDRESS_MASK : ~0;
> 
> Can we put the cast in the PCI_ROM_ADDRESS_MASK #define so we don't have to
> repeat it in all the uses?

Fixing these "implicit conversion" warnings, especially for
unsigned types, is a slippery slope. (The behavior of the
conversion is well-defined.)

How about changing the type of PCI_ROM_ADDRESS_MASK instead?
It's defined as ~0x7ffUL but it's only used in the context
of u32.

So make it an unsigned int:
#define PCI_ROM_ADDRESS_MASK (~0x7ffU)

AFAIU, unsigned int is 32 bits on all platforms supported
by Linux.

Regards.