On Mon, Jul 26, 2010 at 4:26 PM, Sam Ravnborg <sam@xxxxxxxxxxxx> wrote:
> On Mon, Jul 26, 2010 at 04:04:55PM -0600, Grant Likely wrote:
>> of_node_to_nid() is only relevant in a few architectures. Don't force
>> everyone to implement it anyway. This patch also adds asm-generic/of.h
>> which will be used to contain other overrideable symbols.
>>
>> Signed-off-by: Grant Likely <grant.likely@xxxxxxxxxxxx>
>> ---
>>
>> Changes in v3: don't use asm-generic, just keep macros in of.h
>> Changes in v2: address comments from sfr, add asm-generic/of.h
>
> The use of asm-generic makes perfect sense.
> This is how we usually deal with arch specific stuff.
>
> With v3 of your patch we have a different result depending
> on if we do:
> #include <linux/of.h>
>
> or we do:
> #include <asm/prom.h>
>
> This is undesireable.
The patch does maintain consistency. Including only asm/prom.h may
mean that of_node_to_nid is undefined, but it will never result in a
different definition. linux/of.h includes asm/prom.h before doing the
#ifdef test.
> I suggest to go back to v2 of your patch where you use asm-generic/of.h.
Stephen suggested dropping asm-generic/of.h. I'm happy to do it either way.
> linux/of.h shall include asm/of.h
> Then all archs shall have a of.h that may
> include the asm-generic variant.
>
>
> One patch to introduce of.h all over.
> And a second patch to do the of_node_to_nid stuff would be appropriate.
>
>
>> diff --git a/arch/powerpc/include/asm/prom.h b/arch/powerpc/include/asm/prom.h
>> index da7dd63..dca25a5 100644
>> --- a/arch/powerpc/include/asm/prom.h
>> +++ b/arch/powerpc/include/asm/prom.h
>> @@ -103,6 +103,11 @@ struct device_node *of_find_next_cache_node(struct device_node *np);
>> /* Get the MAC address */
>> extern const void *of_get_mac_address(struct device_node *np);
>>
>
> This shall go in asm/of.h
>> +#ifdef CONFIG_NUMA
>> +extern int of_node_to_nid(struct device_node *device);
>> +#define of_node_to_nid of_node_to_nid
>
> This define is used to tell asm-generic/of.h that the arch has
> a local definition - OK.
>
>> +#endif
>> +
>> /**
>> * of_irq_map_pci - Resolve the interrupt for a PCI device
>> * @pdev: the device whose interrupt is to be resolved
>
>> diff --git a/include/linux/of.h b/include/linux/of.h
>> index b0756f3..cc936ca 100644
>> --- a/include/linux/of.h
>> +++ b/include/linux/of.h
>> @@ -146,6 +146,11 @@ static inline unsigned long of_read_ulong(const __be32 *cell, int size)
>>
>> #define OF_BAD_ADDR ((u64)-1)
>>
>> +#ifndef of_node_to_nid
>> +static inline int of_node_to_nid(struct device_node *np) { return 0; }
>> +#define of_node_to_nid of_node_to_nid
>
> But I fail to see the purpose of this define.
It protects against some later include file doing a #define
of_node_to_nid and thus resulting in an inconsistent definition. If
some code tries to do this then the preprocessor will complain. This
is the pattern that Stephen suggested.
g.
--
To unsubscribe from this list: send the line "unsubscribe sparclinux" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
