Re: [PATCH 5/7] regmap: Check if a register is writable instead of readable in regcache_read

[Lars-Peter Clausen <lars@metafoo.de>]

On 11/16/2011 05:38 PM, Mark Brown wrote:
> On Wed, Nov 16, 2011 at 05:34:33PM +0100, Lars-Peter Clausen wrote:
>> On 11/16/2011 05:16 PM, Mark Brown wrote:
> 
>>> This logic doesn't entirely follow - one can have registers which are
>>> volatile but could be read once at startup.  Plus...
> 
>> Hm? The use case here is chips which do not support readback. So we never
>> want to fallback to a hardware read but still want to be able to do a cached
>> read.
> 
> This code will be run on every chip, including chips with read/write
> access.  Caches are useful for all chips.

Of course. And it still works for chips with read/write support with this
patch, but it doesn't work for chips without read support without this patch.

> 
>>>> @@ -206,7 +206,7 @@ int regcache_read(struct regmap *map,
> 
>>>>  	BUG_ON(!map->cache_ops);
> 
>>>> -	if (!regmap_readable(map, reg))
>>>> +	if (!regmap_writeable(map, reg))
>>>>  		return -EIO;
> 
>>> ...the code winds up just looking like an obvious bug.
> 
>> Why? If a register is not writable we won't have anything in the cache for
>> it. So reading from the cache for a register which is not writable doesn't
>> make any sense.
> 
> If you're looking at the read function and it's checking to see if the
> register is writeable the first thought would be that this is a
> cut'n'paste error.  The above code is at best *way* too cute.

We can of course add a comment explaining why it is regmap_writable instead
of regmap_readable.