On 08/02/16 10:44, Antoine Tenart wrote:
> On Mon, Feb 08, 2016 at 10:32:08AM +0000, Marc Zyngier wrote:
>> On 08/02/16 10:26, Antoine Tenart wrote:
>>>>> +static int alpine_msix_init(struct device_node *node,
>>>>> + struct device_node *parent)
>>>>> +{
>>>>> + struct alpine_msix_data *priv;
>>>>> + struct resource res;
>>>>> + int ret;
>>>>> +
>>>>> + priv = kzalloc(sizeof(*priv), GFP_KERNEL);
>>>>> + if (!priv)
>>>>> + return -ENOMEM;
>>>>> +
>>>>> + spin_lock_init(&priv->msi_map_lock);
>>>>> +
>>>>> + ret = of_address_to_resource(node, 0, &res);
>>>>> + if (ret) {
>>>>> + pr_err("Failed to allocate resource\n");
>>>>> + goto err_priv;
>>>>> + }
>>>>> +
>>>>> + priv->addr_high = upper_32_bits((u64)res.start);
>>>>> + priv->addr_low = lower_32_bits(res.start) +
>>>>> ALPINE_MSIX_SPI_TARGET_CLUSTER0;
>>>>
>>>> This is a bit odd. If you always set bit 16, why isn't that reflected in
>>>> the base address coming from the DT?
>>>
>>> The 20 least significant bits of addr_low provide direct information
>>> regarding the interrupt destination, so I thought it would be clearer
>>> to have this explicitly in the driver so that we know what those bits
>>> mean.
>>
>> So what is this information? TARGET_CLUSTER0 is not very expressive, and
>> doesn't show what the alternatives are. Could you please elaborate a bit
>> on that front?
>
> For now lots of bits are reserved, so there aren't many alternatives.
> Bits [18:17] are used to set the GIC to which to route the MSI and bit
> 16 must be set when this target GIC is the primary GIC (bits [18:17] set
> to 0x0). There aren't other options available for now (that I'm aware
> of) for the target GIC configuration.
OK. So maybe add that as a comment, so that people know what is
happening there. And if the code gets updated to include new
functionalities, it will be easier to track the changes.
Thanks,
M.
--
Jazz is not dead. It just smells funny...
