[cdi-devel] [PATCH v2 5/9] cdi.h: Add cdi_handle_bus_device()

Max Reitz ·

On 16.04.2015 23:50, Kevin Wolf wrote:
> On Thu, Apr 16, 2015 at 12:13:54AM +0200, Max Reitz wrote:
>> + Currently, all devices provided on a bus have to be passed to the
>>    init_device() function of all drivers handling any device type on that
>>    bus. cdi_handle_bus_device() can be used by the driver to specify a
>>    pattern (struct cdi_bus_device_pattern) which allows the CDI
>>    implementation to speed up the selection of the correct driver for a
>>    new device.
>>
>> Signed-off-by: Max Reitz <max at tyndur.org>
> As you still haven't convinced me that you only speak Saxionian and no
> German, bilingual documentation, please. ;-)

Bah. Ähmal Dogumendazschioun zu schraim war mir ganz bersöhnlisch nu or 
schon fast ä bissel zu viel, da wäsch nu ni, obsch das ohch noch dobblt 
hinkriesche. ;-)

>> diff --git a/include/cdi.h b/include/cdi.h
>> index 998e080..9c1c1b1 100644
>> --- a/include/cdi.h
>> +++ b/include/cdi.h
>> @@ -141,6 +141,17 @@ struct cdi_bus_data {
>>   };
>>   
>>   /**
>> + * This structure, which is extended by any bus type supporting it, contains
>> + * information about which devices can be handled by a certain driver.
>> + * Typically, it will contain some of the fields provided by the appropriate
>> + * cdi_bus_data structure (the ones which are used to identify a device like
>> + * device/vendor ID) where negative values signify a "don't care".
>> + */
>> +struct cdi_bus_device_pattern {
>> +    cdi_device_type_t bus_type;
>> +};
>> +
>> +/**
>>    * \german
>>    * Beschreibt ein Gerät
>>    * \endgerman
>> @@ -406,6 +417,19 @@ int cdi_provide_device(struct cdi_bus_data* device);
>>   int cdi_provide_device_internal_drv(struct cdi_bus_data* device,
>>                                       struct cdi_driver* driver);
>>   
>> +/**
>> + * Tells the CDI implementation that the driver @drv is able to handle devices
>> + * with the pattern @pattern (appearing on the bus @pattern->bus_type). This
>> + * function can be called multiple times. Any device appearing on the bus not
>> + * fitting the pattern will not be passed to the @drv->init_device() function.
>> + * That function may however return NULL even after the driver has used this
>> + * function, that is, it is valid to call this function with a @pattern that
>> + * does not only fit devices handled by @drv but others as well (which can then
>> + * be rejected by letting @drv->init_device() return NULL).
>> + */
>> +void cdi_handle_bus_device(struct cdi_driver* drv,
>> +                           struct cdi_bus_device_pattern* pattern);
> Is this supposed to be called in drv->init_driver()? If so, could be
> worth mentioning.

I didn't intend any constraints for it, but limiting it to init_driver() 
seems reasonable.

> Ideally it would just be part of the static driver declaration instead
> of a separate function, but I imagine this is hard to achieve with both
> the driver and device pattern being subclassed.

Initially, I made it part of the static driver declaration, but then I 
didn't really want to go into having a variable-length array for 
multiple bus device patterns as part of the driver structure...

Max