Re: [PATCH v3] PM / devfreq: Add debugfs support with devfreq_summary file

From: Chanwoo Choi
Date: Mon Jan 13 2020 - 22:58:06 EST

Next message: Arvind Sankar: "Re: [PATCH v3] x86/vmlinux: Fix vmlinux.lds.S with pre-2.23 binutils"
Previous message: Chanwoo Choi: "Re: [PATCH v3] PM / devfreq: Add debugfs support with devfreq_summary file"
In reply to: Chanwoo Choi: "Re: [PATCH v3] PM / devfreq: Add debugfs support with devfreq_summary file"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

On 1/14/20 1:02 PM, Chanwoo Choi wrote:
> On 1/14/20 11:51 AM, Dmitry Osipenko wrote:
>> 14.01.2020 05:23, Chanwoo Choi ÐÐÑÐÑ:
>>> On 1/14/20 10:45 AM, Dmitry Osipenko wrote:
>>>> 14.01.2020 04:45, Chanwoo Choi ÐÐÑÐÑ:
>>>>> Add debugfs interface to provide debugging information of devfreq device.
>>>>> It contains 'devfreq_summary' entry to show the summary of registered
>>>>> devfreq devices as following and the additional debugfs file will be added.
>>>>> - /sys/kernel/debug/devfreq/devfreq_summary
>>>>>
>>>>> [Detailed description of each field of 'devfreq_summary' debugfs file]
>>>>> - dev_name : Device name of h/w.
>>>>> - dev : Device name made by devfreq core.
>>>>> - parent_dev : If devfreq device uses the passive governor,
>>>>> show parent devfreq device name. Otherwise, show 'null'.
>>>>> - governor : Devfreq governor.
>>>>> - polling_ms : If devfreq device uses the simple_ondemand governor,
>>>>> polling_ms is necessary for the period. (unit: millisecond)
>>>>> - cur_freq_Hz : Current Frequency (unit: Hz)
>>>>> - old_freq_Hz : Frequency before changing. (unit: Hz)
>>>>> - new_freq_Hz : Frequency after changed. (unit: Hz)
>>>>>
>>>>> [For example on Exynos5422-based Odroid-XU3 board]
>>>>> $ cat /sys/kernel/debug/devfreq/devfreq_summary
>>>>> dev_name dev parent_dev governor polling_ms cur_freq_Hz min_freq_Hz max_freq_Hz
>>>>> ------------------------------ ---------- ---------- --------------- ---------- ------------ ------------ ------------
>>>>> 10c20000.memory-controller devfreq0 null simple_ondemand 0 165000000 165000000 825000000
>>>>> soc:bus_wcore devfreq1 null simple_ondemand 50 532000000 88700000 532000000
>>>>> soc:bus_noc devfreq2 devfreq1 passive 0 111000000 66600000 111000000
>>>>> soc:bus_fsys_apb devfreq3 devfreq1 passive 0 222000000 111000000 222000000
>>>>> soc:bus_fsys devfreq4 devfreq1 passive 0 200000000 75000000 200000000
>>>>> soc:bus_fsys2 devfreq5 devfreq1 passive 0 200000000 75000000 200000000
>>>>> soc:bus_mfc devfreq6 devfreq1 passive 0 333000000 83250000 333000000
>>>>> soc:bus_gen devfreq7 devfreq1 passive 0 266000000 88700000 266000000
>>>>> soc:bus_peri devfreq8 devfreq1 passive 0 66600000 66600000 66600000
>>>>> soc:bus_g2d devfreq9 devfreq1 passive 0 333000000 83250000 333000000
>>>>> soc:bus_g2d_acp devfreq10 devfreq1 passive 0 266000000 66500000 266000000
>>>>> soc:bus_jpeg devfreq11 devfreq1 passive 0 300000000 75000000 300000000
>>>>> soc:bus_jpeg_apb devfreq12 devfreq1 passive 0 166500000 83250000 166500000
>>>>> soc:bus_disp1_fimd devfreq13 devfreq1 passive 0 200000000 120000000 200000000
>>>>> soc:bus_disp1 devfreq14 devfreq1 passive 0 300000000 120000000 300000000
>>>>> soc:bus_gscl_scaler devfreq15 devfreq1 passive 0 300000000 150000000 300000000
>>>>> soc:bus_mscl devfreq16 devfreq1 passive 0 666000000 84000000 666000000
>>>>>
>>>>> [lkp: Reported the build error]
>>>>> Reported-by: kbuild test robot <lkp@xxxxxxxxx>
>>>>> Signed-off-by: Chanwoo Choi <cw00.choi@xxxxxxxxxxx>
>>>>> ---
>>>>> Changes from v2:
>>>>> - Show 'null' at 'parent_dev' field when governor of devfreq device
>>>>> is not passive
>>>>> Changes from v1:
>>>>> - Drop the patch about 'devfreq_transitions' debugfs file
>>>>> - Modify from 'hz' to 'Hz'
>>>>> - Edit the indentation of 'devfreq_summary' when show summary
>>>>> - Exchange sequence between PTR_ERR and IS_ERR when debugfs_create_dir
>>>>>
>>>>> drivers/devfreq/devfreq.c | 84 +++++++++++++++++++++++++++++++++++++++
>>>>> 1 file changed, 84 insertions(+)
>>>>>
>>>>> diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c
>>>>> index 89260b17598f..c5ef2d194b1b 100644
>>>>> --- a/drivers/devfreq/devfreq.c
>>>>> +++ b/drivers/devfreq/devfreq.c
>>>>> @@ -10,6 +10,7 @@
>>>>> #include <linux/kernel.h>
>>>>> #include <linux/kmod.h>
>>>>> #include <linux/sched.h>
>>>>> +#include <linux/debugfs.h>
>>>>> #include <linux/errno.h>
>>>>> #include <linux/err.h>
>>>>> #include <linux/init.h>
>>>>> @@ -33,6 +34,7 @@
>>>>> #define HZ_PER_KHZ 1000
>>>>>
>>>>> static struct class *devfreq_class;
>>>>> +static struct dentry *devfreq_debugfs;
>>>>>
>>>>> /*
>>>>> * devfreq core provides delayed work based load monitoring helper
>>>>> @@ -1643,6 +1645,79 @@ static struct attribute *devfreq_attrs[] = {
>>>>> };
>>>>> ATTRIBUTE_GROUPS(devfreq);
>>>>>
>>>>> +/**
>>>>> + * devfreq_summary_show() - Show the summary of the devfreq devices
>>>>> + * @s: seq_file instance to show the summary of devfreq devices
>>>>> + * @data: not used
>>>>> + *
>>>>> + * Show the summary of the devfreq devices via 'devfreq_summary' debugfs file.
>>>>> + * It helps that user can know the detailed information of the devfreq devices.
>>>>> + *
>>>>> + * Return 0 always because it shows the information without any data change.
>>>>> + */
>>>>> +static int devfreq_summary_show(struct seq_file *s, void *data)
>>>>> +{
>>>>> + struct devfreq *devfreq;
>>>>> + struct devfreq *p_devfreq = NULL;
>>
>> Looks like there is no need to pre-initialize the p_devfreq, please see
>> below.
>>
>>>>> + unsigned long cur_freq, min_freq, max_freq;
>>>>> + unsigned int polling_ms;
>>>>> +
>>>>> + seq_printf(s, "%-30s %-10s %-10s %-15s %10s %12s %12s %12s\n",
>>>>> + "dev_name",
>>>>> + "dev",
>>>>> + "parent_dev",
>>>>> + "governor",
>>>>> + "polling_ms",
>>>>> + "cur_freq_Hz",
>>>>> + "min_freq_Hz",
>>>>> + "max_freq_Hz");
>>>>> + seq_printf(s, "%30s %10s %10s %15s %10s %12s %12s %12s\n",
>>>>> + "------------------------------",
>>>>> + "----------",
>>>>> + "----------",
>>>>> + "---------------",
>>>>> + "----------",
>>>>> + "------------",
>>>>> + "------------",
>>>>> + "------------");
>>>>> +
>>>>> + mutex_lock(&devfreq_list_lock);
>>>>> +
>>>>> + list_for_each_entry_reverse(devfreq, &devfreq_list, node) {
>
> #if IS_ENABLED(CONFIG_DEVFREQ_GOV_PASSIVE)
>
>>>>> + if (!strncmp(devfreq->governor_name, DEVFREQ_GOV_PASSIVE,
>>>>> + DEVFREQ_NAME_LEN)) {
>>>>> + struct devfreq_passive_data *data = devfreq->data;
>>>>> +
>>>>> + if (data)
>>>>> + p_devfreq = data->parent;
>>
>> else
>> p_devfreq = NULL;
>>
>> Otherwise p_devfreq will be reused from a previous devfreq device if
>> that device has the data->parent.
>
> This path has the problem. I'm missing that.
>
> 'struct devfreq_passive_data' is only defined
> when CONFIG_DEVFREQ_GOV_PASSIVE is enabled.
>
> In case of when CONFIG_DEVFREQ_GOV_PASSIVE is disabled,
> p_devfreq should be initialized with NULL as following:
> I'll modify this patch as following:
>
> struct devfreq *p_devfreq = NULL;
>
> #if IS_ENABLED(CONFIG_DEVFREQ_GOV_PASSIVE)
> if (!strncmp(devfreq->governor_name, DEVFREQ_GOV_PASSIVE,
> DEVFREQ_NAME_LEN)) {
> struct devfreq_passive_data *data = devfreq->data;
> if (data)
> p_devfreq = data->parent;
> else
> p_devfreq = NULL;
> #endif
>
>
>>
>>>>> + } else {
>>>>> + p_devfreq = NULL;
>>>>> + }
>>>>> +
>>>>> + mutex_lock(&devfreq->lock);
>>>>> + cur_freq = devfreq->previous_freq,
>>>>> + get_freq_range(devfreq, &min_freq, &max_freq);
>>>>> + polling_ms = devfreq->profile->polling_ms,
>>>>> + mutex_unlock(&devfreq->lock);
>>>>> +
>>>>> + seq_printf(s,
>>>>> + "%-30s %-10s %-10s %-15s %10d %12ld %12ld %12ld\n",
>>>>> + dev_name(devfreq->dev.parent),
>>>>> + dev_name(&devfreq->dev),
>>>>> + p_devfreq ? dev_name(&p_devfreq->dev) : "null",
>>>>> + devfreq->governor_name,
>>>>> + polling_ms,
>>>>> + cur_freq,
>>>>> + min_freq,
>>>>> + max_freq);
>>>>> + }
>>>>> +
>>>>> + mutex_unlock(&devfreq_list_lock);
>>>>> +
>>>>> + return 0;
>>>>> +}
>>>>> +DEFINE_SHOW_ATTRIBUTE(devfreq_summary);
>>>>> +
>>>>> static int __init devfreq_init(void)
>>>>> {
>>>>> devfreq_class = class_create(THIS_MODULE, "devfreq");
>>>>> @@ -1659,6 +1734,15 @@ static int __init devfreq_init(void)
>>>>> }
>>>>> devfreq_class->dev_groups = devfreq_groups;
>>>>>
>>>>> +
>>>> If you'll you'll take a look at [2], you may notice notice that it
>>>> checks whether devfreq_debugfs IS_ERR [3] and then bails out.
>>>>
>>>> [2] https://protect2.fireeye.com/url?k=a5047d02-f8ca7cd1-a505f64d-000babff317b-ed64c541cd573190&u=https://elixir.bootlin.com/linux/v5.5-rc6/source/fs/debugfs/inode.c#L432
>>>> [3] https://protect2.fireeye.com/url?k=0a43fffa-578dfe29-0a4274b5-000babff317b-1c1c1062bfaeb0fc&u=https://elixir.bootlin.com/linux/v5.5-rc6/source/fs/debugfs/inode.c#L316
>>>>
>>>> Thus you could simply remove the above error handling, making code to
>>>> look cleaner.
>>>>
>>>>
>>>
>>> OK. I'll modify it as following: Do you agree?
>>
>> Looks good, thanks.
>>
>>> devfreq_debugfs = debugfs_create_dir("devfreq", NULL);
>>
>> I'd also add a newline here, to ease reading of the code.
>
> ok.
>
>>
>>> debugfs_create_file("devfreq_summary", 0444,
>>> devfreq_debugfs, NULL,
>>> &devfreq_summary_fops);

Because there is not checking of return value,
I think that it is more proper instead of adding the newline
between debugfs_create_dir and debugfs_create_file.

devfreq_debugfs = debugfs_create_dir("devfreq", NULL);
debugfs_create_file("devfreq_summary", 0444,
devfreq_debugfs, NULL,
&devfreq_summary_fops);

>>>
>>>
>>
>>
>>
>
>

--
Best Regards,
Chanwoo Choi
Samsung Electronics

Next message: Arvind Sankar: "Re: [PATCH v3] x86/vmlinux: Fix vmlinux.lds.S with pre-2.23 binutils"
Previous message: Chanwoo Choi: "Re: [PATCH v3] PM / devfreq: Add debugfs support with devfreq_summary file"
In reply to: Chanwoo Choi: "Re: [PATCH v3] PM / devfreq: Add debugfs support with devfreq_summary file"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]