Re: [RFC PATCH v3 5/7] devfreq: exynos-bus: Add interconnect functionality to exynos-bus

[Georgi Djakov]

Hi Artur,

On 1/24/20 13:22, Artur ÅwigoÅ wrote:
> Hi Georgi,
> 
> On Wed, 2020-01-22 at 19:02 +0200, Georgi Djakov wrote:
>> Hi Artur,
>>
>> On 12/20/19 13:56, Artur ÅwigoÅ wrote:
>>> This patch adds interconnect functionality to the exynos-bus devfreq
>>> driver.
>>>
>>> The SoC topology is a graph (or, more specifically, a tree) and its
>>> edges are specified using the 'exynos,interconnect-parent-node' in the
>>> DT. Due to unspecified relative probing order, -EPROBE_DEFER may be
>>> propagated to ensure that the parent is probed before its children.
>>>
>>> Each bus is now an interconnect provider and an interconnect node as well
>>> (cf. Documentation/interconnect/interconnect.rst), i.e. every bus registers
>>> itself as a node. Node IDs are not hardcoded but rather assigned at
>>
>> Just to note that usually the provider consists of multiple nodes and each node
>> represents a single master or slave port on the AXI bus for example. I am not
>> sure whether this represents correctly the Exynos hardware, so it's up to
>> you.
>>
>>> runtime, in probing order (subject to the above-mentioned exception
>>> regarding relative order). This approach allows for using this driver with
>>> various Exynos SoCs.
>>
>> This sounds good. I am wondering whether such dynamic probing would be useful
>> for other platforms too. Then maybe it would make sense to even have a common DT
>> property, but we will see.
>>
>> Is this going to be used only together with devfreq?
> 
> Yes, this functions solely as an extension to devfreq, hence the slightly
> unusual architecture (one icc_provider/icc_node per devfreq).

Ok, thanks for clarifying.

> (Compared to a singleton icc_provider, this approach yields less code with
> a very simple xlate()).
> 
> With exactly one icc_node for every devfreq device, I think I will actually
> reuse the devfreq ID (as seen in the device name, e.g. the "3" in "devfreq3")
> for the node ID. The devfreq framework already does the dynamic numbering
> thing that I do in this patch using IDR.
> 

Sounds good.

>>> Frequencies requested via the interconnect API for a given node are
>>> propagated to devfreq using dev_pm_qos_update_request(). Please note that
>>> it is not an error when CONFIG_INTERCONNECT is 'n', in which case all
>>> interconnect API functions are no-op.
>>
>> How about the case where CONFIG_INTERCONNECT=m. Looks like the build will fail
>> if CONFIG_ARM_EXYNOS_BUS_DEVFREQ=y, so this dependency should be expressed in
>> Kconfig.
> 
> I think adding:
> 	depends on INTERCONNECT || !INTERCONNECT

Yes, exactly.

> under ARM_EXYNOS_BUS_DEVFREQ does the trick.
> 
>>>
>>> Signed-off-by: Artur ÅwigoÅ <a.swigon@xxxxxxxxxxx>
>>> ---
>>>  drivers/devfreq/exynos-bus.c | 144 +++++++++++++++++++++++++++++++++++
>>>  1 file changed, 144 insertions(+)
>>>
>>> diff --git a/drivers/devfreq/exynos-bus.c b/drivers/devfreq/exynos-bus.c
>>> index 9fdb188915e8..694a9581dcdb 100644
>>> --- a/drivers/devfreq/exynos-bus.c
>>> +++ b/drivers/devfreq/exynos-bus.c
>>> @@ -14,14 +14,19 @@
>>>  #include <linux/devfreq-event.h>
>>>  #include <linux/device.h>
>>>  #include <linux/export.h>
>>> +#include <linux/idr.h>
>>> +#include <linux/interconnect-provider.h>
>>>  #include <linux/module.h>
>>>  #include <linux/of.h>
>>>  #include <linux/pm_opp.h>
>>> +#include <linux/pm_qos.h>
>>>  #include <linux/platform_device.h>
>>>  #include <linux/regulator/consumer.h>
>>>  
>>>  #define DEFAULT_SATURATION_RATIO	40
>>>  
>>> +#define kbps_to_khz(x) ((x) / 8)
>>> +
>>>  struct exynos_bus {
>>>  	struct device *dev;
>>>  
>>> @@ -35,6 +40,12 @@ struct exynos_bus {
>>>  	struct opp_table *opp_table;
>>>  	struct clk *clk;
>>>  	unsigned int ratio;
>>> +
>>> +	/* One provider per bus, one node per provider */
>>> +	struct icc_provider provider;
>>> +	struct icc_node *node;
>>> +
>>> +	struct dev_pm_qos_request qos_req;
>>>  };
>>>  
>>>  /*
>>> @@ -205,6 +216,39 @@ static void exynos_bus_passive_exit(struct device *dev)
>>>  	clk_disable_unprepare(bus->clk);
>>>  }
>>>  
>>> +static int exynos_bus_icc_set(struct icc_node *src, struct icc_node *dst)
>>> +{
>>> +	struct exynos_bus *src_bus = src->data, *dst_bus = dst->data;
>>> +	s32 src_freq = kbps_to_khz(src->avg_bw);
>>> +	s32 dst_freq = kbps_to_khz(dst->avg_bw);
>>> +	int ret;
>>> +
>>> +	ret = dev_pm_qos_update_request(&src_bus->qos_req, src_freq);
>>> +	if (ret < 0) {
>>> +		dev_err(src_bus->dev, "failed to update PM QoS request");
>>> +		return ret;
>>> +	}
>>> +
>>> +	ret = dev_pm_qos_update_request(&dst_bus->qos_req, dst_freq);
>>> +	if (ret < 0) {
>>> +		dev_err(dst_bus->dev, "failed to update PM QoS request");
>>> +		return ret;
>>> +	}
>>> +
>>> +	return 0;
>>> +}
>>> +
>>> +static struct icc_node *exynos_bus_icc_xlate(struct of_phandle_args *spec,
>>> +					     void *data)
>>> +{
>>> +	struct exynos_bus *bus = data;
>>> +
>>> +	if (spec->np != bus->dev->of_node)
>>> +		return ERR_PTR(-EINVAL);
>>> +
>>> +	return bus->node;
>>> +}
>>> +
>>>  static int exynos_bus_parent_parse_of(struct device_node *np,
>>>  					struct exynos_bus *bus)
>>>  {
>>> @@ -419,6 +463,96 @@ static int exynos_bus_profile_init_passive(struct exynos_bus *bus,
>>>  	return 0;
>>>  }
>>>  
>>> +static struct icc_node *exynos_bus_icc_get_parent(struct exynos_bus *bus)
>>> +{
>>> +	struct device_node *np = bus->dev->of_node;
>>> +	struct of_phandle_args args;
>>> +	int num, ret;
>>> +
>>> +	num = of_count_phandle_with_args(np, "exynos,interconnect-parent-node",
>>> +					"#interconnect-cells");
>>> +	if (num != 1)
>>> +		return NULL; /* parent nodes are optional */
>>> +
>>> +	ret = of_parse_phandle_with_args(np, "exynos,interconnect-parent-node",
>>> +					"#interconnect-cells", 0, &args);
>>> +	if (ret < 0)
>>> +		return ERR_PTR(ret);
>>> +
>>> +	of_node_put(args.np);
>>> +
>>> +	return of_icc_get_from_provider(&args);
>>> +}
>>> +
>>> +static int exynos_bus_icc_init(struct exynos_bus *bus)
>>> +{
>>> +	static DEFINE_IDA(ida);
>>> +
>>> +	struct device *dev = bus->dev;
>>> +	struct icc_provider *provider = &bus->provider;
>>> +	struct icc_node *node, *parent_node;
>>> +	int id, ret;
>>> +
>>> +	/* Initialize the interconnect provider */
>>> +	provider->set = exynos_bus_icc_set;
>>> +	provider->aggregate = icc_std_aggregate;
>>> +	provider->xlate = exynos_bus_icc_xlate;
>>> +	provider->dev = dev;
>>> +	provider->inter_set = true;
>>> +	provider->data = bus;
>>> +
>>> +	ret = icc_provider_add(provider);
>>> +	if (ret < 0)
>>> +		return ret;
>>> +
>>> +	ret = id = ida_alloc(&ida, GFP_KERNEL);
>>> +	if (ret < 0)
>>> +		goto err_id;
>>> +
>>> +	node = icc_node_create(id);
>>> +	if (IS_ERR(node)) {
>>> +		ret = PTR_ERR(node);
>>> +		goto err_node;
>>> +	}
>>> +
>>> +	bus->node = node;
>>> +	node->name = dev->of_node->name;
>>> +	node->data = bus;
>>> +	icc_node_add(node, provider);
>>> +
>>> +	parent_node = exynos_bus_icc_get_parent(bus);
>>> +	if (IS_ERR(parent_node)) {
>>> +		ret = PTR_ERR(parent_node);
>>> +		goto err_parent;
>>> +	}
>>> +
>>> +	if (parent_node) {
>>> +		ret = icc_link_create(node, parent_node->id);
>>> +		if (ret < 0)
>>> +			goto err_parent;
>>> +	}
>>> +
>>> +	ret = dev_pm_qos_add_request(bus->devfreq->dev.parent, &bus->qos_req,
>>> +					DEV_PM_QOS_MIN_FREQUENCY, 0);
>>> +	if (ret < 0)
>>> +		goto err_request;
>>> +
>>> +	return 0;
>>> +
>>> +err_request:
>>> +	if (parent_node)
>>> +		icc_link_destroy(node, parent_node);
>>> +err_parent:
>>> +	icc_node_del(node);
>>> +	icc_node_destroy(id);
>>> +err_node:
>>> +	ida_free(&ida, id);
>>> +err_id:
>>> +	icc_provider_del(provider);
>>> +
>>> +	return ret;
>>> +}
>>> +
>>>  static int exynos_bus_probe(struct platform_device *pdev)
>>>  {
>>>  	struct device *dev = &pdev->dev;
>>> @@ -468,6 +602,16 @@ static int exynos_bus_probe(struct platform_device *pdev)
>>>  	if (ret < 0)
>>>  		goto err;
>>>  
>>> +	/*
>>> +	 * Initialize interconnect provider. A return value of -ENOTSUPP means
>>> +	 * that CONFIG_INTERCONNECT is disabled.
>>> +	 */
>>> +	ret = exynos_bus_icc_init(bus);
>>> +	if (ret < 0 && ret != -ENOTSUPP) {

I have been also thinking that all the code that you add in this patch would fit
nicely into a separate interconnect provider driver. Then instead of the above
icc_init() you can create a sub-device (platform_device_register_data() maybe?).
The separate driver will be handling just the interconnect functionality and
could be enabled on top of this devfreq driver.

Thanks,
Georgi

P.S. I think that grouping all the related patches into a single patch-set would
make reviewing them easier. Then we can decide about the merge path of each.

>>> +		dev_err(dev, "failed to initialize the interconnect provider");
>>> +		goto err;
>>> +	}
>>> +
>>>  	max_state = bus->devfreq->profile->max_state;
>>>  	min_freq = (bus->devfreq->profile->freq_table[0] / 1000);
>>>  	max_freq = (bus->devfreq->profile->freq_table[max_state - 1] / 1000);
>>>
>>
>>