[dpdk-dev] [PATCHv5 04/33] bus/fslmc: introducing fsl-mc bus driver

[Shreyansh Jain]

On Friday 20 January 2017 12:38 AM, Ferruh Yigit wrote:
> On 1/19/2017 1:23 PM, Hemant Agrawal wrote:
>> The fslmc bus driver is a rte_bus driver which scans the fsl-mc bus
>> for NXP DPAA2 SoCs.
>>
>> Signed-off-by: Hemant Agrawal <hemant.agrawal at nxp.com>
>> ---
>
> <...>
>
>> +#
>> +# library name
>> +#
>> +LIB = librte_pmd_fslmcbus.a
>
> Since now there is a bus folder/driver, what do you think nameming
> library with librte_bus_ prefix, like: librte_bus_fslmc.a
>
> <...>
>
>> +
>> +static int
>> +rte_fslmc_probe(void)
>> +{
>> +	int ret = -1;
>
> If any bus->probe() fails, rte_bus_probe() breaks and returns error,
> which cause app to exit.
> Here if there is no device or driver in the bus, function is returning
> error, I guess it should be returning zero for this case.

It is a nice point of discussion (even in the bus patch). Should Bus 
iteration for scan/probe fail if any bus implementation fails?

In the initial series I had placed a 'TODO' in the bus patch to get some 
comments - I couldn't make a decision so the final bus scan/probe loop 
'fails if any bus fails whether in scan or probe'.

I think that EAL should continue looping over buses irrespective of bus 
failure - specially removing such dependencies on bus implementations to 
return a valid code compatible with EAL's design.

>
>> +	struct rte_dpaa2_device *dev;
>> +	struct rte_dpaa2_driver *drv;
>> +
>> +	TAILQ_FOREACH(dev, &rte_fslmc_bus.device_list, next) {
>> +		TAILQ_FOREACH(drv, &rte_fslmc_bus.driver_list, next) {
>> +			ret = rte_fslmc_match(drv, dev);
>> +			if (ret)
>> +				continue;
>> +
>> +			if (!drv->probe)
>> +				continue;
>> +
>> +			ret = drv->probe(drv, dev);
>> +			if (ret)
>> +				FSLMC_BUS_LOG(ERR, "Unable to probe.\n");
>> +			break;
>> +		}
>> +	}
>> +	return ret;
>> +}
>
> <...>
>