Re: [PATCH 2/5] cpufreq: governor: Create separate sysfs-ops

[Juri Lelli <juri.lelli@arm.com>]

Hi Viresh,

On 02/02/16 16:27, Viresh Kumar wrote:
> Until now, governors (ondemand/conservative) were using the
> 'global-attr' or 'freq-attr', depending on the sysfs location where we
> want to create governor's directory.
> 
> The problem is that, in case of 'freq-attr', we are forced to use
> show()/store() present in cpufreq.c, which always take policy->rwsem.
> 
> And because of that we were facing some ABBA lockups during governor
> callback event CPUFREQ_GOV_POLICY_EXIT. And so we were dropping the
> rwsem right before calling governor callback for CPUFREQ_GOV_POLICY_EXIT
> event.
> 
> That caused further problems and it never worked perfectly.
> 
> This patch attempts to fix that by creating separate sysfs-ops for
> cpufreq governors.
> 
> Because things got much simplified now, we don't need separate
> show/store callbacks for governor-for-system and governor-per-policy
> cases.
> 
> Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org>

This patch cleans things up a lot, that's good.

One thing I'm still concerned about, though: don't we need some locking
in place for some of the store operations on governors attributes? Are
store_{ignore_nice_load, sampling_down_fact, etc} safe without locking?
It seems that we can call them from different cpus concurrently.

Best,

- Juri

> ---
>  drivers/cpufreq/cpufreq_conservative.c | 71 +++++++++++++---------------------
>  drivers/cpufreq/cpufreq_governor.c     | 50 +++++++++++++++++++-----
>  drivers/cpufreq/cpufreq_governor.h     | 31 +++++++++++++--
>  drivers/cpufreq/cpufreq_ondemand.c     | 71 +++++++++++++---------------------
>  4 files changed, 122 insertions(+), 101 deletions(-)
> 
> diff --git a/drivers/cpufreq/cpufreq_conservative.c b/drivers/cpufreq/cpufreq_conservative.c
> index 57750367bd26..980145da796a 100644
> --- a/drivers/cpufreq/cpufreq_conservative.c
> +++ b/drivers/cpufreq/cpufreq_conservative.c
> @@ -275,51 +275,35 @@ static ssize_t store_freq_step(struct dbs_data *dbs_data, const char *buf,
>  	return count;
>  }
>  
> -show_store_one(cs, sampling_rate);
> -show_store_one(cs, sampling_down_factor);
> -show_store_one(cs, up_threshold);
> -show_store_one(cs, down_threshold);
> -show_store_one(cs, ignore_nice_load);
> -show_store_one(cs, freq_step);
> -show_one(cs, min_sampling_rate);
> -
> -gov_sys_pol_attr_rw(sampling_rate);
> -gov_sys_pol_attr_rw(sampling_down_factor);
> -gov_sys_pol_attr_rw(up_threshold);
> -gov_sys_pol_attr_rw(down_threshold);
> -gov_sys_pol_attr_rw(ignore_nice_load);
> -gov_sys_pol_attr_rw(freq_step);
> -gov_sys_pol_attr_ro(min_sampling_rate);
> -
> -static struct attribute *dbs_attributes_gov_sys[] = {
> -	&min_sampling_rate_gov_sys.attr,
> -	&sampling_rate_gov_sys.attr,
> -	&sampling_down_factor_gov_sys.attr,
> -	&up_threshold_gov_sys.attr,
> -	&down_threshold_gov_sys.attr,
> -	&ignore_nice_load_gov_sys.attr,
> -	&freq_step_gov_sys.attr,
> +gov_show_one(cs, sampling_rate);
> +gov_show_one(cs, sampling_down_factor);
> +gov_show_one(cs, up_threshold);
> +gov_show_one(cs, down_threshold);
> +gov_show_one(cs, ignore_nice_load);
> +gov_show_one(cs, freq_step);
> +gov_show_one(cs, min_sampling_rate);
> +
> +gov_attr_rw(sampling_rate);
> +gov_attr_rw(sampling_down_factor);
> +gov_attr_rw(up_threshold);
> +gov_attr_rw(down_threshold);
> +gov_attr_rw(ignore_nice_load);
> +gov_attr_rw(freq_step);
> +gov_attr_ro(min_sampling_rate);
> +
> +static struct attribute *dbs_attributes[] = {
> +	&min_sampling_rate.attr,
> +	&sampling_rate.attr,
> +	&sampling_down_factor.attr,
> +	&up_threshold.attr,
> +	&down_threshold.attr,
> +	&ignore_nice_load.attr,
> +	&freq_step.attr,
>  	NULL
>  };
>  
> -static struct attribute_group cs_attr_group_gov_sys = {
> -	.attrs = dbs_attributes_gov_sys,
> -	.name = "conservative",
> -};
> -
> -static struct attribute *dbs_attributes_gov_pol[] = {
> -	&min_sampling_rate_gov_pol.attr,
> -	&sampling_rate_gov_pol.attr,
> -	&sampling_down_factor_gov_pol.attr,
> -	&up_threshold_gov_pol.attr,
> -	&down_threshold_gov_pol.attr,
> -	&ignore_nice_load_gov_pol.attr,
> -	&freq_step_gov_pol.attr,
> -	NULL
> -};
> -
> -static struct attribute_group cs_attr_group_gov_pol = {
> -	.attrs = dbs_attributes_gov_pol,
> +static struct attribute_group cs_attr_group = {
> +	.attrs = dbs_attributes,
>  	.name = "conservative",
>  };
>  
> @@ -365,8 +349,7 @@ define_get_cpu_dbs_routines(cs_cpu_dbs_info);
>  
>  static struct common_dbs_data cs_dbs_cdata = {
>  	.governor = GOV_CONSERVATIVE,
> -	.attr_group_gov_sys = &cs_attr_group_gov_sys,
> -	.attr_group_gov_pol = &cs_attr_group_gov_pol,
> +	.attr_group = &cs_attr_group,
>  	.get_cpu_cdbs = get_cpu_cdbs,
>  	.get_cpu_dbs_info_s = get_cpu_dbs_info_s,
>  	.gov_dbs_timer = cs_dbs_timer,
> diff --git a/drivers/cpufreq/cpufreq_governor.c b/drivers/cpufreq/cpufreq_governor.c
> index 9a7edc91ad57..e785a118cbdc 100644
> --- a/drivers/cpufreq/cpufreq_governor.c
> +++ b/drivers/cpufreq/cpufreq_governor.c
> @@ -22,14 +22,37 @@
>  
>  #include "cpufreq_governor.h"
>  
> -static struct attribute_group *get_sysfs_attr(struct dbs_data *dbs_data)
> +#define to_dbs_data(k) container_of(k, struct dbs_data, kobj)
> +#define to_attr(a) container_of(a, struct governor_attr, attr)
> +
> +static ssize_t show(struct kobject *kobj, struct attribute *attr, char *buf)
>  {
> -	if (have_governor_per_policy())
> -		return dbs_data->cdata->attr_group_gov_pol;
> -	else
> -		return dbs_data->cdata->attr_group_gov_sys;
> +	struct dbs_data *dbs_data = to_dbs_data(kobj);
> +	struct governor_attr *gattr = to_attr(attr);
> +
> +	if (gattr->show)
> +		return gattr->show(dbs_data, buf);
> +
> +	return -EIO;
> +}
> +
> +static ssize_t store(struct kobject *kobj, struct attribute *attr,
> +		     const char *buf, size_t count)
> +{
> +	struct dbs_data *dbs_data = to_dbs_data(kobj);
> +	struct governor_attr *gattr = to_attr(attr);
> +
> +	if (gattr->store)
> +		return gattr->store(dbs_data, buf, count);
> +
> +	return -EIO;
>  }
>  
> +static const struct sysfs_ops sysfs_ops = {
> +	.show	= show,
> +	.store	= store,
> +};
> +
>  void dbs_check_cpu(struct dbs_data *dbs_data, int cpu)
>  {
>  	struct cpu_dbs_info *cdbs = dbs_data->cdata->get_cpu_cdbs(cpu);
> @@ -354,6 +377,7 @@ static int cpufreq_governor_init(struct cpufreq_policy *policy,
>  				 struct dbs_data *dbs_data,
>  				 struct common_dbs_data *cdata)
>  {
> +	struct attribute_group *attr_group;
>  	int ret;
>  
>  	/* State should be equivalent to EXIT */
> @@ -395,10 +419,17 @@ static int cpufreq_governor_init(struct cpufreq_policy *policy,
>  
>  	policy->governor_data = dbs_data;
>  
> -	ret = sysfs_create_group(get_governor_parent_kobj(policy),
> -				 get_sysfs_attr(dbs_data));
> -	if (ret)
> +	attr_group = dbs_data->cdata->attr_group;
> +	dbs_data->kobj_type.sysfs_ops = &sysfs_ops;
> +	dbs_data->kobj_type.default_attrs = attr_group->attrs;
> +
> +	ret = kobject_init_and_add(&dbs_data->kobj, &dbs_data->kobj_type,
> +				   get_governor_parent_kobj(policy),
> +				   attr_group->name);
> +	if (ret) {
> +		pr_err("%s: failed to init dbs_data kobj: %d\n", __func__, ret);
>  		goto reset_gdbs_data;
> +	}
>  
>  	return 0;
>  
> @@ -426,8 +457,7 @@ static int cpufreq_governor_exit(struct cpufreq_policy *policy,
>  		return -EBUSY;
>  
>  	if (!--dbs_data->usage_count) {
> -		sysfs_remove_group(get_governor_parent_kobj(policy),
> -				   get_sysfs_attr(dbs_data));
> +		kobject_put(&dbs_data->kobj);
>  
>  		policy->governor_data = NULL;
>  
> diff --git a/drivers/cpufreq/cpufreq_governor.h b/drivers/cpufreq/cpufreq_governor.h
> index ad44a8546a3a..59b28133dd68 100644
> --- a/drivers/cpufreq/cpufreq_governor.h
> +++ b/drivers/cpufreq/cpufreq_governor.h
> @@ -108,6 +108,31 @@ static ssize_t store_##file_name##_gov_pol				\
>  show_one(_gov, file_name);						\
>  store_one(_gov, file_name)
>  
> +/* Governor's specific attributes */
> +struct dbs_data;
> +struct governor_attr {
> +	struct attribute attr;
> +	ssize_t (*show)(struct dbs_data *dbs_data, char *buf);
> +	ssize_t (*store)(struct dbs_data *dbs_data, const char *buf,
> +			 size_t count);
> +};
> +
> +#define gov_show_one(_gov, file_name)					\
> +static ssize_t show_##file_name						\
> +(struct dbs_data *dbs_data, char *buf)					\
> +{									\
> +	struct _gov##_dbs_tuners *tuners = dbs_data->tuners;		\
> +	return sprintf(buf, "%u\n", tuners->file_name);			\
> +}
> +
> +#define gov_attr_ro(_name)						\
> +static struct governor_attr _name =					\
> +__ATTR(_name, 0444, show_##_name, NULL)
> +
> +#define gov_attr_rw(_name)						\
> +static struct governor_attr _name =					\
> +__ATTR(_name, 0644, show_##_name, store_##_name)
> +
>  /* create helper routines */
>  #define define_get_cpu_dbs_routines(_dbs_info)				\
>  static struct cpu_dbs_info *get_cpu_cdbs(int cpu)			\
> @@ -197,14 +222,12 @@ struct cs_dbs_tuners {
>  };
>  
>  /* Common Governor data across policies */
> -struct dbs_data;
>  struct common_dbs_data {
>  	/* Common across governors */
>  	#define GOV_ONDEMAND		0
>  	#define GOV_CONSERVATIVE	1
>  	int governor;
> -	struct attribute_group *attr_group_gov_sys; /* one governor - system */
> -	struct attribute_group *attr_group_gov_pol; /* one governor - policy */
> +	struct attribute_group *attr_group; /* one governor - system */
>  
>  	/*
>  	 * Common data for platforms that don't set
> @@ -234,6 +257,8 @@ struct dbs_data {
>  	struct common_dbs_data *cdata;
>  	int usage_count;
>  	void *tuners;
> +	struct kobject kobj;
> +	struct kobj_type kobj_type;
>  };
>  
>  /* Governor specific ops, will be passed to dbs_data->gov_ops */
> diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c
> index b31f64745232..b7983dd02e24 100644
> --- a/drivers/cpufreq/cpufreq_ondemand.c
> +++ b/drivers/cpufreq/cpufreq_ondemand.c
> @@ -436,51 +436,35 @@ static ssize_t store_powersave_bias(struct dbs_data *dbs_data, const char *buf,
>  	return count;
>  }
>  
> -show_store_one(od, sampling_rate);
> -show_store_one(od, io_is_busy);
> -show_store_one(od, up_threshold);
> -show_store_one(od, sampling_down_factor);
> -show_store_one(od, ignore_nice_load);
> -show_store_one(od, powersave_bias);
> -show_one(od, min_sampling_rate);
> -
> -gov_sys_pol_attr_rw(sampling_rate);
> -gov_sys_pol_attr_rw(io_is_busy);
> -gov_sys_pol_attr_rw(up_threshold);
> -gov_sys_pol_attr_rw(sampling_down_factor);
> -gov_sys_pol_attr_rw(ignore_nice_load);
> -gov_sys_pol_attr_rw(powersave_bias);
> -gov_sys_pol_attr_ro(min_sampling_rate);
> -
> -static struct attribute *dbs_attributes_gov_sys[] = {
> -	&min_sampling_rate_gov_sys.attr,
> -	&sampling_rate_gov_sys.attr,
> -	&up_threshold_gov_sys.attr,
> -	&sampling_down_factor_gov_sys.attr,
> -	&ignore_nice_load_gov_sys.attr,
> -	&powersave_bias_gov_sys.attr,
> -	&io_is_busy_gov_sys.attr,
> +gov_show_one(od, sampling_rate);
> +gov_show_one(od, io_is_busy);
> +gov_show_one(od, up_threshold);
> +gov_show_one(od, sampling_down_factor);
> +gov_show_one(od, ignore_nice_load);
> +gov_show_one(od, powersave_bias);
> +gov_show_one(od, min_sampling_rate);
> +
> +gov_attr_rw(sampling_rate);
> +gov_attr_rw(io_is_busy);
> +gov_attr_rw(up_threshold);
> +gov_attr_rw(sampling_down_factor);
> +gov_attr_rw(ignore_nice_load);
> +gov_attr_rw(powersave_bias);
> +gov_attr_ro(min_sampling_rate);
> +
> +static struct attribute *dbs_attributes[] = {
> +	&min_sampling_rate.attr,
> +	&sampling_rate.attr,
> +	&up_threshold.attr,
> +	&sampling_down_factor.attr,
> +	&ignore_nice_load.attr,
> +	&powersave_bias.attr,
> +	&io_is_busy.attr,
>  	NULL
>  };
>  
> -static struct attribute_group od_attr_group_gov_sys = {
> -	.attrs = dbs_attributes_gov_sys,
> -	.name = "ondemand",
> -};
> -
> -static struct attribute *dbs_attributes_gov_pol[] = {
> -	&min_sampling_rate_gov_pol.attr,
> -	&sampling_rate_gov_pol.attr,
> -	&up_threshold_gov_pol.attr,
> -	&sampling_down_factor_gov_pol.attr,
> -	&ignore_nice_load_gov_pol.attr,
> -	&powersave_bias_gov_pol.attr,
> -	&io_is_busy_gov_pol.attr,
> -	NULL
> -};
> -
> -static struct attribute_group od_attr_group_gov_pol = {
> -	.attrs = dbs_attributes_gov_pol,
> +static struct attribute_group od_attr_group = {
> +	.attrs = dbs_attributes,
>  	.name = "ondemand",
>  };
>  
> @@ -542,8 +526,7 @@ static struct od_ops od_ops = {
>  
>  static struct common_dbs_data od_dbs_cdata = {
>  	.governor = GOV_ONDEMAND,
> -	.attr_group_gov_sys = &od_attr_group_gov_sys,
> -	.attr_group_gov_pol = &od_attr_group_gov_pol,
> +	.attr_group = &od_attr_group,
>  	.get_cpu_cdbs = get_cpu_cdbs,
>  	.get_cpu_dbs_info_s = get_cpu_dbs_info_s,
>  	.gov_dbs_timer = od_dbs_timer,
> -- 
> 2.7.0.79.gdc08a19
>