Re: [PATCH 2/2] PM / devfreq: Add devfreq_transitions debugfs file

From: Dmitry Osipenko
Date: Wed Jan 08 2020 - 09:10:47 EST


08.01.2020 14:22, Chanwoo Choi ÐÐÑÐÑ:
> On 1/8/20 6:56 AM, Dmitry Osipenko wrote:
>> 07.01.2020 12:05, Chanwoo Choi ÐÐÑÐÑ:
>>> Add new devfreq_transitions debugfs file to track the frequency transitions
>>> of all devfreq devices for the simple profiling as following:
>>> - /sys/kernel/debug/devfreq/devfreq_transitions
>>>
>>> And the user can decide the storage size (CONFIG_NR_DEVFREQ_TRANSITIONS)
>>> in Kconfig in order to save the transition history.
>>>
>>> [Detailed description of each field of 'devfreq_transitions' debugfs file]
>>> - time_ms : Change time of frequency transition. (unit: millisecond)
>>> - 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.
>>> - load_% : If devfreq device uses the simple_ondemand governor,
>>> load is used by governor whene deciding the new frequency.
>>> (unit: percentage)
>>> - 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_transitions
>>> time_ms dev_name dev parent_dev load_% old_freq_hz new_freq_hz
>>> ---------- ------------------------------ ---------- ---------- ---------- ------------ ------------
>>> 14600 soc:bus_noc devfreq2 devfreq1 0 100000000 67000000
>>> 14600 soc:bus_fsys_apb devfreq3 devfreq1 0 200000000 100000000
>>> 14600 soc:bus_fsys devfreq4 devfreq1 0 200000000 100000000
>>> 14600 soc:bus_fsys2 devfreq5 devfreq1 0 150000000 75000000
>>> 14602 soc:bus_mfc devfreq6 devfreq1 0 222000000 96000000
>>> 14602 soc:bus_gen devfreq7 devfreq1 0 267000000 89000000
>>> 14602 soc:bus_g2d devfreq9 devfreq1 0 300000000 84000000
>>> 14602 soc:bus_g2d_acp devfreq10 devfreq1 0 267000000 67000000
>>> 14602 soc:bus_jpeg devfreq11 devfreq1 0 300000000 75000000
>>> 14602 soc:bus_jpeg_apb devfreq12 devfreq1 0 167000000 84000000
>>> 14603 soc:bus_disp1_fimd devfreq13 devfreq1 0 200000000 120000000
>>> 14603 soc:bus_disp1 devfreq14 devfreq1 0 300000000 120000000
>>> 14606 soc:bus_gscl_scaler devfreq15 devfreq1 0 300000000 150000000
>>> 14606 soc:bus_mscl devfreq16 devfreq1 0 333000000 84000000
>>> 14608 soc:bus_wcore devfreq1 9 333000000 84000000
>>> 14783 10c20000.memory-controller devfreq0 35 825000000 633000000
>>> 15873 soc:bus_wcore devfreq1 41 84000000 400000000
>>> 15873 soc:bus_noc devfreq2 devfreq1 0 67000000 100000000
>>> [snip]
>>>
>>> Signed-off-by: Chanwoo Choi <cw00.choi@xxxxxxxxxxx>
>>> ---
>>> drivers/devfreq/Kconfig | 13 +++
>>> drivers/devfreq/devfreq.c | 126 +++++++++++++++++++++++++++++
>>> drivers/devfreq/governor.h | 3 +
>>> drivers/devfreq/governor_passive.c | 2 +
>>> include/linux/devfreq.h | 1 +
>>> 5 files changed, 145 insertions(+)
>>>
>>> diff --git a/drivers/devfreq/Kconfig b/drivers/devfreq/Kconfig
>>> index 0b1df12e0f21..84936eec0ef9 100644
>>> --- a/drivers/devfreq/Kconfig
>>> +++ b/drivers/devfreq/Kconfig
>>> @@ -74,6 +74,19 @@ config DEVFREQ_GOV_PASSIVE
>>> through sysfs entries. The passive governor recommends that
>>> devfreq device uses the OPP table to get the frequency/voltage.
>>>
>>> +comment "DEVFREQ Debugging"
>>> +
>>> +config NR_DEVFREQ_TRANSITIONS
>>> + int "Maximum storage size to save DEVFREQ Transitions (10-1000)"
>>> + depends on DEBUG_FS
>>> + range 10 1000
>>> + default "100"
>>> + help
>>> + Show the frequency transitions of all devfreq devices via
>>> + '/sys/kernel/debug/devfreq/devfreq_transitions' for the simple
>>> + profiling. It needs to decide the storage size to save transition
>>> + history of all devfreq devices.
>>> +
>>> comment "DEVFREQ Drivers"
>>>
>>> config ARM_EXYNOS_BUS_DEVFREQ
>>> diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c
>>> index c7f5e4e06420..7abaae06fa65 100644
>>> --- a/drivers/devfreq/devfreq.c
>>> +++ b/drivers/devfreq/devfreq.c
>>> @@ -268,6 +268,57 @@ int devfreq_update_status(struct devfreq *devfreq, unsigned long freq)
>>> }
>>> EXPORT_SYMBOL(devfreq_update_status);
>>>
>>> +/**
>>> + * devfreq_update_transitions() - Update frequency transitions for debugfs file
>>> + * @devfreq: the devfreq instance
>>> + * @old_freq: the previous frequency before changing the frequency
>>> + * @new_freq: the new frequency after frequency is changed
>>> + */
>>> +struct devfreq_transitions {
>>> + struct devfreq *devfreq;
>>> + struct devfreq_freqs freqs;
>>> + unsigned long load;
>>> +} debugfs_transitions[CONFIG_NR_DEVFREQ_TRANSITIONS];
>>> +
>>> +static spinlock_t devfreq_debugfs_lock;
>>> +static int debugfs_transitions_index;
>>> +
>>> +void devfreq_update_transitions(struct devfreq *devfreq,
>>> + unsigned long old_freq, unsigned long new_freq,
>>> + unsigned long busy_time, unsigned long total_time)
>>> +{
>>> + unsigned long load;
>>> + int i;
>>> +
>>> + if (!devfreq_debugfs || !devfreq || (old_freq == new_freq))
>>> + return;
>>> +
>>> + spin_lock_nested(&devfreq_debugfs_lock, SINGLE_DEPTH_NESTING);
>>> +
>>> + i = debugfs_transitions_index;
>>> +
>>> + /*
>>> + * Calculate the load and if load is larger than 100,
>>> + * initialize to 100 because the unit of load is percentage.
>>> + */
>>> + load = (total_time == 0 ? 0 : (100 * busy_time) / total_time);
>>> + if (load > 100)
>>> + load = 100;
>>> +
>>> + debugfs_transitions[i].devfreq = devfreq;
>>> + debugfs_transitions[i].freqs.time = ktime_to_ms(ktime_get());
>>> + debugfs_transitions[i].freqs.old = old_freq;
>>> + debugfs_transitions[i].freqs.new = new_freq;
>>> + debugfs_transitions[i].load = load;
>>> +
>>> + if (++i == CONFIG_NR_DEVFREQ_TRANSITIONS)
>>> + i = 0;
>>> + debugfs_transitions_index = i;
>>> +
>>> + spin_unlock(&devfreq_debugfs_lock);
>>> +}
>>> +EXPORT_SYMBOL(devfreq_update_transitions);
>>> +
>>> /**
>>> * find_devfreq_governor() - Find devfreq governor from name
>>> * @name: name of the governor
>>> @@ -401,6 +452,10 @@ static int set_target(struct devfreq *devfreq,
>>> return err;
>>> }
>>>
>>> + devfreq_update_transitions(devfreq, cur_freq, new_freq,
>>> + devfreq->last_status.busy_time,
>>> + devfreq->last_status.total_time);
>>> +
>>> freqs.new = new_freq;
>>> notify_transition(devfreq, &freqs, DEVFREQ_POSTCHANGE);
>>>
>>> @@ -1787,6 +1842,72 @@ static int devfreq_summary_show(struct seq_file *s, void *data)
>>> }
>>> DEFINE_SHOW_ATTRIBUTE(devfreq_summary);
>>>
>>> +/**
>>> + * devfreq_transitions_show() - Show the frequency transitions of the registered
>>> + * devfreq devices via 'devfreq_transitions' debugfs file.
>>> + */
>>> +static int devfreq_transitions_show(struct seq_file *s, void *data)
>>> +{
>>> + struct devfreq *devfreq = NULL;
>>> + struct devfreq *p_devfreq = NULL;
>>> + struct devfreq_freqs *freqs = NULL;
>>> + unsigned long load;
>>> + int i = debugfs_transitions_index;
>>> + int count;
>>> +
>>> + seq_printf(s, "%-10s %-30s %-10s %-10s %-6s %-12s %-12s\n",
>>> + "time_ms",
>>> + "dev_name",
>>> + "dev",
>>> + "parent_dev",
>>> + "load_%",
>>> + "old_freq_hz",
>>> + "new_freq_hz");
>>> + seq_printf(s, "%-10s %-30s %-10s %-10s %-6s %-12s %-12s\n",
>>> + "----------",
>>> + "------------------------------",
>>> + "----------",
>>> + "----------",
>>> + "----------",
>>> + "------------",
>>> + "------------");
>>
>> Isn't this needed here?
>>
>> mutex_lock(&devfreq_list_lock);
>
> It doesn't touch the devfreq instance of devfreq_list.
> So, it is not necessary locked of devfreq_list_lock.

What stops devfreq device to be removed by another CPU thread while this
function is in a process of execution?

This condition is unlikely to happen in practice ever, but technically
it should be possible to happen.

>>> + spin_lock(&devfreq_debugfs_lock);
>>> + for (count = 0; count < CONFIG_NR_DEVFREQ_TRANSITIONS; count++) {
>>> + devfreq = debugfs_transitions[i].devfreq;
>>> + freqs = &debugfs_transitions[i].freqs;
>>> + load = debugfs_transitions[i].load;
>>> +
>>> + i = (CONFIG_NR_DEVFREQ_TRANSITIONS == ++i) ? 0 : i;
>>> + if (!devfreq)
>>> + continue;
>>
>> I suppose debugfs_transitions[i].devfreq should be set to NULL when
>> devfreq device is removed, but I don't see it happening anywhere in this
>> patch.
>
> When debugfs_transitions[] array is not fully filled out
> by devfreq_update_transitions(), debugfs_transitions[i].devfreq is NULL.
> In this case, if user execute 'cat /sys/kernel/debug/devfreq/devfreq_transitions',
> devfreq_transitions_show() need to check the debugfs_transitions[i].devfreq
> is NULL or not.
>
> After filled out the debugfs_transitions[] array,
> actually, 'if(!devfreq)' is not necessary. Maybe, this style is inefficient
> It need to rework. I'll think again.

Imagine this situation:

1. there is a devfreq device, let's name it defreq123

2. the debugfs_transitions array is getting filled and now it has this
entry:

debugfs_transitions[0].devfreq = defreq123

3. user removes defreq123 driver module

# rmmod defreq123

4. the defreq123 is released now

5. at what memory location debugfs_transitions[0].devfreq is pointing now?