Re: drivers/pci/probe.c compile warnings on -tip

[Jesse Barnes <jbarnes@virtuousgeek.org>]

On Wednesday, October 1, 2008 3:47 am Mikael Pettersson wrote:
> Steven Noonan writes:
>  > I was hunting down some warnings I got when compiling -tip, and with
>  > one of them, I'm not sure what would be a proper way to handle it.
>  >
>  > If CONFIG_PHYS_ADDR_T_64BIT is not enabled, these warnings show up:
>  >
>  > drivers/pci/probe.c: In function '__pci_read_base':
>  > drivers/pci/probe.c:308: warning: format '%llx' expects type 'long
>  > long unsigned int', but argument 4 has type 'resource_size_t'
>  > drivers/pci/probe.c:308: warning: format '%llx' expects type 'long
>  > long unsigned int', but argument 5 has type 'resource_size_t'
>  > drivers/pci/probe.c:320: warning: format '%llx' expects type 'long
>  > long unsigned int', but argument 5 has type 'resource_size_t'
>  > drivers/pci/probe.c:320: warning: format '%llx' expects type 'long
>  > long unsigned int', but argument 6 has type 'resource_size_t'
>  > drivers/pci/probe.c: In function 'pci_read_bridge_bases':
>  > drivers/pci/probe.c:392: warning: format '%llx' expects type 'long
>  > long unsigned int', but argument 3 has type 'resource_size_t'
>  > drivers/pci/probe.c:392: warning: format '%llx' expects type 'long
>  > long unsigned int', but argument 4 has type 'resource_size_t'
>  > drivers/pci/probe.c:405: warning: format '%llx' expects type 'long
>  > long unsigned int', but argument 3 has type 'resource_size_t'
>  > drivers/pci/probe.c:405: warning: format '%llx' expects type 'long
>  > long unsigned int', but argument 4 has type 'resource_size_t'
>  > drivers/pci/probe.c:443: warning: format '%llx' expects type 'long
>  > long unsigned int', but argument 4 has type 'resource_size_t'
>  > drivers/pci/probe.c:443: warning: format '%llx' expects type 'long
>  > long unsigned int', but argument 5 has type 'resource_size_t'
>  >
>  >
>  > Each of the lines is something like this:
>  >
>  > printk(KERN_DEBUG "PCI: %s reg %x 64bit mmio: [%llx, %llx]\n",
>  > pci_name(dev), pos, res->start, res->end);
>  >
>  > res->start and res->end are resource_size_t (which is phys_addr_t),
>  > and are sized either 32-bit or 64-bit, based on whether
>  > CONFIG_PHYS_ADDR_T_64BIT is set.
>
> Oh not those again. Similar warnings in 27-rc's drivers/pci
> were fixed recently, but apparently someone out there likes
> to create new ones.
>
> The fix is to cast the resource_size_t arguments to unsigned long long.
> With "%llx" that's the /only/ valid argument type.
>
> Also note that these warnings aren't harmless. On 32-bit machines
> with parameters passed on the stack the mismatch in size between
> 32-bit resource_size_t and the 64-bit %llx causes undefined
> behaviour as printk mis-enumerates its parameters.

AFAICT, these are fixed in Linus' tree, maybe the x86 tree needs an update?

-- 
Jesse Barnes, Intel Open Source Technology Center