lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <5a82bf8a-d925-ba54-a26f-98b64bedc6e1@samsung.com>
Date:   Mon, 29 Jul 2019 10:52:28 +0900
From:   Chanwoo Choi <cw00.choi@...sung.com>
To:     Artur Świgoń <a.swigon@...tner.samsung.com>,
        devicetree@...r.kernel.org, linux-arm-kernel@...ts.infradead.org,
        linux-samsung-soc@...r.kernel.org, linux-kernel@...r.kernel.org,
        linux-pm@...r.kernel.org, dri-devel@...ts.freedesktop.org
Cc:     krzk@...nel.org, myungjoo.ham@...sung.com, inki.dae@...sung.com,
        sw0312.kim@...sung.com, georgi.djakov@...aro.org,
        m.szyprowski@...sung.com
Subject: Re: [RFC PATCH 09/11] devfreq: exynos-bus: Add interconnect
 functionality to exynos-bus

Hi,

On 19. 7. 23. 오후 9:20, Artur Świgoń wrote:
> This patch adds interconnect functionality to the exynos-bus devfreq
> driver.
> 
> The SoC topology is a graph (or, more specifically, a tree) and most of its
> edges are taken from the devfreq parent-child hierarchy (cf.
> Documentation/devicetree/bindings/devfreq/exynos-bus.txt). The previous
> patch adds missing edges to the DT (under the name 'parent'). Due to
> unspecified relative probing order, -EPROBE_DEFER may be propagated to
> guarantee that a child is probed before its parent.
> 
> 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
> runtime, in probing order (subject to the above-mentioned exception
> regarding relative order). This approach allows for using this driver with
> various Exynos SoCs.
> 
> The devfreq target() callback provided by exynos-bus now selects either the
> frequency calculated by the devfreq governor or the frequency requested via
> the interconnect API for the given node, whichever is higher.

Basically, I agree to support the QoS requirement between devices.
But, I think that need to consider the multiple cases.


1. When changing the devfreq governor by user,
For example of the connection between bus_dmc/leftbus/display on patch8,
there are possible multiple cases with various devfreq governor
which is changed on the runtime by user through sysfs interface.

If users changes the devfreq governor as following:
Before,
- bus_dmc (simple_ondemand, available frequency 100/200/300/400 MHz)
--> bus_leftbus(simple_ondemand, available frequency 100/200/300/400 MHz)
----> bus_display(passive)

After changed governor of bus_dmc,
if the min_freq by interconnect requirement is 400Mhz,
- bus_dmc (powersave) : min_freq and max_freq and cur_freq is 100MHz
--> bus_leftbus(simple_ondemand) : cur_freq is 400Mhz
----> bus_display(passive)

The final frequency is 400MHz of bus_dmc
even if the min_freq/max_freq/cur_freq is 100MHz.
It cannot show the correct min_freq/max_freq through
devfreq sysfs interface.


2. When disabling the some frequency by devfreq-thermal throttling,
This patch checks the min_freq of interconnect requirement
in the exynos_bus_target() and exynos_bus_passive_target().
Also, it cannot show the correct min_freq/max_freq through
devfreq sysfs interface.

For example of bus_dmc bus,
- The available frequencies are 100MHz, 200MHz, 300MHz, 400MHz
- Disable 400MHz by devfreq-thermal throttling 
- min_freq is 100MHz
- max_freq is 300MHz
- min_freq of interconnect is 400MHz

In result, the final frequency is 400MHz by exynos_bus_target()
There are no problem for working. But, the user cannot know
reason why cur_freq is 400MHz even if max_freq is 300MHz.

Basically, update_devfreq() considers the all constraints
of min_freq/max_freq to decide the proper target frequency.


3.
I think that the exynos_bus_passive_target() is used for devfreq device
using 'passive' governor. The frequency already depends on the parent device.

If already the parent devfreq device like bus_leftbus consider
the minimum frequency of QoS requirement like interconnect,
it is not necessary. The next frequency of devfreq device
with 'passive' governor, it will apply the QoS requirement
without any additional code.

> 
> Please note that it is not an error when CONFIG_INTERCONNECT is 'n', in
> which case all interconnect API functions are no-op.
> 
> Signed-off-by: Artur Świgoń <a.swigon@...tner.samsung.com>
> ---
>  drivers/devfreq/exynos-bus.c | 145 +++++++++++++++++++++++++++++++++++
>  1 file changed, 145 insertions(+)
> 
> diff --git a/drivers/devfreq/exynos-bus.c b/drivers/devfreq/exynos-bus.c
> index 412511ca7703..12fb7c84ae50 100644
> --- a/drivers/devfreq/exynos-bus.c
> +++ b/drivers/devfreq/exynos-bus.c
> @@ -14,6 +14,7 @@
>  #include <linux/devfreq-event.h>
>  #include <linux/device.h>
>  #include <linux/export.h>
> +#include <linux/interconnect-provider.h>
>  #include <linux/module.h>
>  #include <linux/of.h>
>  #include <linux/pm_opp.h>
> @@ -23,6 +24,8 @@
>  #define DEFAULT_SATURATION_RATIO	40
>  #define DEFAULT_VOLTAGE_TOLERANCE	2
>  
> +#define icc_units_to_hz(x) ((x) * 1000UL / 8)
> +
>  struct exynos_bus {
>  	struct device *dev;
>  
> @@ -31,12 +34,17 @@ struct exynos_bus {
>  	unsigned int edev_count;
>  	struct mutex lock;
>  
> +	unsigned long min_freq;
>  	unsigned long curr_freq;
>  
>  	struct regulator *regulator;
>  	struct clk *clk;
>  	unsigned int voltage_tolerance;
>  	unsigned int ratio;
> +
> +	/* One provider per bus, one node per provider */
> +	struct icc_provider provider;
> +	struct icc_node *node;
>  };
>  
>  /*
> @@ -61,6 +69,13 @@ exynos_bus_ops_edev(enable_edev);
>  exynos_bus_ops_edev(disable_edev);
>  exynos_bus_ops_edev(set_event);
>  
> +static int exynos_bus_next_id(void)
> +{
> +	static int exynos_bus_node_id;
> +
> +	return exynos_bus_node_id++;
> +}
> +
>  static int exynos_bus_get_event(struct exynos_bus *bus,
>  				struct devfreq_event_data *edata)
>  {
> @@ -98,6 +113,8 @@ static int exynos_bus_target(struct device *dev, unsigned long *freq, u32 flags)
>  	unsigned long old_freq, new_freq, new_volt, tol;
>  	int ret = 0;
>  
> +	*freq = max(*freq, bus->min_freq);
> +
>  	/* Get new opp-bus instance according to new bus clock */
>  	new_opp = devfreq_recommended_opp(dev, freq, flags);
>  	if (IS_ERR(new_opp)) {
> @@ -208,6 +225,8 @@ static int exynos_bus_passive_target(struct device *dev, unsigned long *freq,
>  	unsigned long old_freq, new_freq;
>  	int ret = 0;
>  
> +	*freq = max(*freq, bus->min_freq);
> +
>  	/* Get new opp-bus instance according to new bus clock */
>  	new_opp = devfreq_recommended_opp(dev, freq, flags);
>  	if (IS_ERR(new_opp)) {
> @@ -251,6 +270,35 @@ 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;
> +
> +	src_bus->min_freq = icc_units_to_hz(src->peak_bw);
> +	dst_bus->min_freq = icc_units_to_hz(dst->peak_bw);
> +
> +	return 0;
> +}
> +
> +static int exynos_bus_icc_aggregate(struct icc_node *node, u32 avg_bw,
> +				    u32 peak_bw, u32 *agg_avg, u32 *agg_peak)
> +{
> +	*agg_peak = *agg_avg = peak_bw;
> +
> +	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)
>  {
> @@ -469,6 +517,95 @@ static int exynos_bus_profile_init_passive(struct exynos_bus *bus,
>  	return ret;
>  }
>  
> +static int exynos_bus_icc_connect(struct exynos_bus *bus)
> +{
> +	struct device_node *np = bus->dev->of_node;
> +	struct devfreq *parent_devfreq;
> +	struct icc_node *parent_node = NULL;
> +	struct of_phandle_args args;
> +	int ret = 0;
> +
> +	parent_devfreq = devfreq_get_devfreq_by_phandle(bus->dev, 0);
> +	if (!IS_ERR(parent_devfreq)) {
> +		struct exynos_bus *parent_bus;
> +
> +		parent_bus = dev_get_drvdata(parent_devfreq->dev.parent);
> +		parent_node = parent_bus->node;
> +	} else {
> +		/* Look for parent in DT */
> +		int num = of_count_phandle_with_args(np, "parent",
> +						     "#interconnect-cells");
> +		if (num != 1)
> +			goto out;
> +
> +		ret = of_parse_phandle_with_args(np, "parent",
> +						 "#interconnect-cells",
> +						 0, &args);
> +		if (ret < 0)
> +			goto out;
> +
> +		of_node_put(args.np);
> +
> +		parent_node = of_icc_get_from_provider(&args);
> +		if (IS_ERR(parent_node)) {
> +			/* May be -EPROBE_DEFER */
> +			ret = PTR_ERR(parent_node);
> +			goto out;
> +		}
> +	}
> +
> +	ret = icc_link_create(bus->node, parent_node->id);
> +
> +out:
> +	return ret;
> +}
> +
> +static int exynos_bus_icc_init(struct exynos_bus *bus)
> +{
> +	struct device *dev = bus->dev;
> +	struct icc_provider *provider = &bus->provider;
> +	struct icc_node *node;
> +	int id, ret;
> +
> +	/* Initialize the interconnect provider */
> +	provider->set = exynos_bus_icc_set;
> +	provider->aggregate = exynos_bus_icc_aggregate;
> +	provider->xlate = exynos_bus_icc_xlate;
> +	provider->dev = dev;
> +	provider->data = bus;
> +
> +	ret = icc_provider_add(provider);
> +	if (ret < 0)
> +		goto out;
> +
> +	id = exynos_bus_next_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);
> +
> +	ret = exynos_bus_icc_connect(bus);
> +	if (ret < 0)
> +		goto err_connect;
> +
> +out:
> +	return ret;
> +
> +err_connect:
> +	icc_node_del(node);
> +	icc_node_destroy(id);
> +err_node:
> +	icc_provider_del(provider);
> +
> +	return ret;
> +}
> +
>  static int exynos_bus_probe(struct platform_device *pdev)
>  {
>  	struct device *dev = &pdev->dev;
> @@ -517,6 +654,14 @@ static int exynos_bus_probe(struct platform_device *pdev)
>  			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)
> +		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);
> 


-- 
Best Regards,
Chanwoo Choi
Samsung Electronics

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ