| Message ID | cover.1498026827.git.viresh.kumar@linaro.org |
|---|---|
| Headers | show |
| Series | PM / Domains: Power domain performance states | expand |
On 17-07-17, 14:38, Ulf Hansson wrote: > On 21 June 2017 at 09:10, Viresh Kumar <viresh.kumar@linaro.org> wrote: > > Some platforms have the capability to configure the performance state of > > their Power Domains. The performance levels are identified by positive > > integer values, a lower value represents lower performance state. > > > > This patch adds a new genpd API: pm_genpd_update_performance_state(). > > The caller passes the affected device and the frequency representing its > > next DVFS state. > > > > The power domains get two new callbacks: > > > > - get_performance_state(): This is called by the genpd core to retrieve > > the performance state (integer value) corresponding to a target > > frequency for the device. This state is used by the genpd core to find > > the highest requested state by all the devices powered by a domain. > > Please clarify this a bit more. > > I guess what you want to say is that genpd aggregates all requested > performance state of all its devices and its subdomains, to be able to > set a correct (highest requested) performance state. Right. > Moreover, could you perhaps explain a bit on *when* this callback > becomes invoked. Sure, its done when pm_genpd_update_performance_state() is called for a device. On such an event, the genpd core first calls get_performance_state() and gets the device's target state. It then aggregates the states of all devices/subdomains of the parent domain of this device and finally calls set_performance_state() for the genpd. > > > > - set_performance_state(): The highest state retrieved from above > > interface is then passed to this callback to finally program the > > performance state of the power domain. > > When will this callback be invoked? See above. > What happens when a power domain gets powered off and then on. Is the > performance state restored? Please elaborate a bit on this. Can this happen while the genpd is still in use? If not then we wouldn't have a problem here as the users of it would have revoked their constraints by now. > > The power domains can avoid supplying these callbacks, if they don't > > support setting performance-states. > > > > A power domain may have only get_performance_state() callback, if it > > doesn't have the capability of changing the performance state itself but > > someone in its parent hierarchy has. > > > > A power domain may have only set_performance_state(), if it doesn't have > > any direct devices below it but subdomains. And so the > > get_performance_state() will never get called from the core. > > > > It seems like the ->get_perfomance_state() is a device specific > callback, while the ->set_performance_state() is a genpd domain > callback. Yes. > I am wondering whether we should try to improve the names of the > callbacks to reflect this. What about dev_get_performance_state() and genpd_set_performance_state() ? > > The more common case would be to have both the callbacks set. > > > > Another API, pm_genpd_has_performance_state(), is also added to let > > other parts of the kernel check if the power domain of a device supports > > performance-states or not. This could have been done from > > I think a better name of this function is: > dev_pm_genpd_has_performance_state(). What do you think? Sure. > We might even want to decide to explicitly stay using the terminology > "DVFS" instead. In such case, perhaps converting the names of the > callbacks/API to use "dvfs" instead. For the API added here, maybe > dev_pm_genpd_can_dvfs(). I am not sure about that really. Because in most of the cases genpd wouldn't do any freq switching, but only voltage. > > pm_genpd_update_performance_state() as well, but that routine gets > > called every time we do DVFS for the device and it wouldn't be optimal > > in that case. > > So pm_genpd_update_performance_state() is also a new API added in > $subject patch. But there is no information about it in the changelog, > besides the above. Please add that. Yeah, I missed that and will include like what I said in the beginning of the reply on how this leads to call of other callbacks. > Moreover, perhaps we should rename the function to dev_pm_genpd_set_dvfs() Not sure as earlier said. > > Note that, the performance level as returned by > > ->get_performance_state() for the parent domain of a device is used for > > all domains in parent hierarchy. > > Please clarify a bit on this. What exactly does this mean? For a hierarchy like this: PPdomain 0 PPdomain 1 | | -------------------------- | Pdomain | device ->dev_get_performance_state(dev) would be called for the device and it will return a single value (X) representing the performance index of its parent ("Pdomain"). But the direct parent domain may not support setting of performance index and so we need to propagate the call to parents of Pdomain. And that would be PPdomain 0 and 1. Now the paragraph in the commit says that the same performance index value X will be used for both these PPdomains, as we don't want to make things more complex to begin with. > > Tested-by: Rajendra Nayak <rnayak@codeaurora.org> > > Signed-off-by: Viresh Kumar <viresh.kumar@linaro.org> > > --- > > drivers/base/power/domain.c | 223 ++++++++++++++++++++++++++++++++++++++++++++ > > include/linux/pm_domain.h | 22 +++++ > > 2 files changed, 245 insertions(+) > > > > diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c > > index 71c95ad808d5..d506be9ff1f7 100644 > > --- a/drivers/base/power/domain.c > > +++ b/drivers/base/power/domain.c > > @@ -466,6 +466,229 @@ static int genpd_dev_pm_qos_notifier(struct notifier_block *nb, > > return NOTIFY_DONE; > > } > > > > +/* > > + * Returns true if anyone in genpd's parent hierarchy has > > + * set_performance_state() set. > > + */ > > +static bool genpd_has_set_performance_state(struct generic_pm_domain *genpd) > > +{ > > So this function will be become in-directly called by generic drivers > that supports DVFS of the genpd for their devices. > > I think the data you validate here would be better to be pre-validated > at pm_genpd_init() and at pm_genpd_add|remove_subdomain() and the > result stored in a variable in the genpd struct. Especially when a > subdomain is added, that is a point when you can verify the > *_performance_state() callbacks, and thus make sure it's a correct > setup from the topology point of view. Something like this ? > > + > > + /* Traverse all subdomains within the domain */ > > + list_for_each_entry(link, &genpd->master_links, master_node) { > > + if (link->performance_state > state) > > + state = link->performance_state; > > + } > > + > > From a locking point of view we always traverse the topology from > bottom an up. In other words, we walk the genpd's ->slave_links, and > lock the masters in the order the are defined via the slave_links > list. The order is important to avoid deadlocks. I don't think you > should walk the master_links as being done above, especially not > without using locks. So we need to look at the performance states of the subdomains of a master. The way it is done in this patch with help of link->performance_state, we don't need that locking while traversing the master_links list. Here is how: - Master's (genpd) master_links list is only updated under master's lock, which we have already taken here. So master_links list can't get updated concurrently. - The link->performance_state field of a subdomain (or slave) is only updated from within the master's lock. And we are reading it here from the same lock. AFAIU, there shouldn't be any deadlocks or locking issues here. Can you describe some case that may blow up ? > > + if (genpd->performance_state == state) > > + return 0; > > + > > + if (genpd->set_performance_state) { > > + ret = genpd->set_performance_state(genpd, state); > > + if (!ret) > > + genpd->performance_state = state; > > + > > + return ret; > > This looks wrong. > > I think you should continue to walk upwards in the domain topology, as > there may be some other master than needs to get its performance state > updated. I can do that. > > + } > > + > > + /* > > + * Not all domains support updating performance state. Move on to their > > + * parent domains in that case. > > /s/parent/master > > > + */ > > + prev = genpd->performance_state; > > + > > + list_for_each_entry(link, &genpd->slave_links, slave_node) { > > + master = link->master; > > + > > + genpd_lock_nested(master, depth + 1); > > + > > + link->performance_state = state; > > + ret = genpd_update_performance_state(master, depth + 1); > > + if (ret) > > + link->performance_state = prev; > > + > > + genpd_unlock(master); > > + > > + if (ret) > > + goto err; > > + } > > + > > A general comment is that I think you should look more closely in the > code of genpd_power_off|on(). And also how it calls the > ->power_on|off() callbacks. > > Depending whether you want to update the performance state of the > master domain before the subdomain or the opposite, you will find one > of them being suited for this case as well. Isn't it very much similar to that already ? The only major difference is link->performance_state and I already explained why is it required to be done that way to avoid deadlocks. > > + /* > > + * The parent domains are updated now, lets get genpd performance_state > > + * in sync with those. > > + */ > > + genpd->performance_state = state; > > + return 0; > > + > > +err: > > + list_for_each_entry_continue_reverse(link, &genpd->slave_links, > > + slave_node) { > > + master = link->master; > > + > > + genpd_lock_nested(master, depth + 1); > > + link->performance_state = prev; > > + if (genpd_update_performance_state(master, depth + 1)) > > + pr_err("%s: Failed to roll back to %d performance state\n", > > + genpd->name, prev); > > + genpd_unlock(master); > > + } > > + > > + return ret; > > +} > > + > > +static int __dev_update_performance_state(struct device *dev, int state) > > Please use the prefix genpd, _genpd_ or __genpd for static functions. > > > +{ > > + struct generic_pm_domain_data *gpd_data; > > + int ret; > > + > > + spin_lock_irq(&dev->power.lock); > > Actually there is no need to use this lock. > > Because you hold the genpd lock here, then the device can't be removed > from its genpd and thus there is always a valid gpd_data. I am afraid we still need this lock. genpd_free_dev_data() is called from genpd_remove_device() without genpd lock and so it is possible that we reach here after that lock is dropped from genpd_remove_device() but before genpd_free_dev_data() is called. Right ? > > + > > + if (!dev->power.subsys_data || !dev->power.subsys_data->domain_data) { > > + ret = -ENODEV; > > + goto unlock; > > + } > > + > > + gpd_data = to_gpd_data(dev->power.subsys_data->domain_data); > > + > > + ret = gpd_data->performance_state; > > + gpd_data->performance_state = state; > > + > > +unlock: > > + spin_unlock_irq(&dev->power.lock); > > + > > + return ret; > > +} > > + > > +/** > > + * pm_genpd_update_performance_state - Update performance state of device's > > + * parent power domain for the target frequency for the device. > > + * > > + * @dev: Device for which the performance-state needs to be adjusted. > > + * @rate: Device's next frequency. This can be set as 0 when the device doesn't > > + * have any performance state constraints left (And so the device wouldn't > > + * participate anymore to find the target performance state of the genpd). > > + * > > + * This must be called by the user drivers (as many times as they want) only > > + * after pm_genpd_has_performance_state() is called (only once) and that > > + * returned "true". > > + * > > + * It is assumed that the user driver guarantees that the genpd wouldn't be > > + * detached while this routine is getting called. > > + * > > + * Returns 0 on success and negative error values on failures. > > + */ > > +int pm_genpd_update_performance_state(struct device *dev, unsigned long rate) > > +{ > > + struct generic_pm_domain *genpd = dev_to_genpd(dev); > > + int ret, state; > > + > > + if (IS_ERR(genpd)) > > + return -ENODEV; > > + > > + genpd_lock(genpd); > > + > > + state = genpd->get_performance_state(dev, rate); > > + if (state < 0) { > > + ret = state; > > + goto unlock; > > + } > > + > > + state = __dev_update_performance_state(dev, state); > > + if (state < 0) { > > + ret = state; > > + goto unlock; > > + } > > + > > + ret = genpd_update_performance_state(genpd, 0); > > + if (ret) > > + __dev_update_performance_state(dev, state); > > + > > +unlock: > > + genpd_unlock(genpd); > > + > > + return ret; > > +} > > +EXPORT_SYMBOL_GPL(pm_genpd_update_performance_state); > > + > > /** > > * genpd_power_off_work_fn - Power off PM domain whose subdomain count is 0. > > * @work: Work structure used for scheduling the execution of this function. > > diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h > > index b7803a251044..bf90177208a2 100644 > > --- a/include/linux/pm_domain.h > > +++ b/include/linux/pm_domain.h > > @@ -63,8 +63,12 @@ struct generic_pm_domain { > > unsigned int device_count; /* Number of devices */ > > unsigned int suspended_count; /* System suspend device counter */ > > unsigned int prepared_count; /* Suspend counter of prepared devices */ > > + unsigned int performance_state; /* Max requested performance state */ > > int (*power_off)(struct generic_pm_domain *domain); > > int (*power_on)(struct generic_pm_domain *domain); > > + int (*get_performance_state)(struct device *dev, unsigned long rate); > > + int (*set_performance_state)(struct generic_pm_domain *domain, > > + unsigned int state); > > struct gpd_dev_ops dev_ops; > > s64 max_off_time_ns; /* Maximum allowed "suspended" time. */ > > bool max_off_time_changed; > > @@ -99,6 +103,9 @@ struct gpd_link { > > struct list_head master_node; > > struct generic_pm_domain *slave; > > struct list_head slave_node; > > + > > + /* Sub-domain's per-parent domain performance state */ > > + unsigned int performance_state; > > How about aggregate_dvfs_state, and move it to the struct generic_pm_domain. > > Because I think you only need to keep track one aggregated state per > domain and per each subdomain, right? This is required to solve the deadlock you showed in the previous version of this patch. As I explained earlier. > Huh, I hope my comments make some sense. Sure, and thanks for such a detailed review :) > It's a bit of tricky code, > when walking the domain topology and I think there are couple > optimization we can do while doing that. Hopefully you can understand > most of my comments at least. :-) > > Feel free to ping me on IRC, if further explanations is needed. I would let you go through this reply and then we can talk tomorrow on IRC as we don't agree on some of the cases here for sure :) -- vireshdiff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c index 4a898e095a1d..182c1911ea9c 100644 --- a/drivers/base/power/domain.c +++ b/drivers/base/power/domain.c @@ -466,25 +466,6 @@ static int genpd_dev_pm_qos_notifier(struct notifier_block *nb, return NOTIFY_DONE; } -/* - * Returns true if anyone in genpd's parent hierarchy has - * set_performance_state() set. - */ -static bool genpd_has_set_performance_state(struct generic_pm_domain *genpd) -{ - struct gpd_link *link; - - if (genpd->set_performance_state) - return true; - - list_for_each_entry(link, &genpd->slave_links, slave_node) { - if (genpd_has_set_performance_state(link->master)) - return true; - } - - return false; -} - /** * pm_genpd_has_performance_state - Checks if power domain does performance * state management. @@ -507,7 +488,7 @@ bool pm_genpd_has_performance_state(struct device *dev) /* The parent domain must have set get_performance_state() */ if (!IS_ERR(genpd) && genpd->get_performance_state) { - if (genpd_has_set_performance_state(genpd)) + if (genpd->can_set_performance_state) return true; /* @@ -1594,6 +1575,8 @@ static int genpd_add_subdomain(struct generic_pm_domain *genpd, if (genpd_status_on(subdomain)) genpd_sd_counter_inc(genpd); + subdomain->can_set_performance_state += genpd->can_set_performance_state; + out: genpd_unlock(genpd); genpd_unlock(subdomain); @@ -1654,6 +1637,8 @@ int pm_genpd_remove_subdomain(struct generic_pm_domain *genpd, if (genpd_status_on(subdomain)) genpd_sd_counter_dec(genpd); + subdomain->can_set_performance_state -= genpd->can_set_performance_state; + ret = 0; break; } @@ -1721,6 +1706,7 @@ int pm_genpd_init(struct generic_pm_domain *genpd, genpd->max_off_time_changed = true; genpd->provider = NULL; genpd->has_provider = false; + genpd->can_set_performance_state = !!genpd->set_performance_state; genpd->domain.ops.runtime_suspend = genpd_runtime_suspend; genpd->domain.ops.runtime_resume = genpd_runtime_resume; genpd->domain.ops.prepare = pm_genpd_prepare; diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h index bf90177208a2..995d0cb1bc14 100644 --- a/include/linux/pm_domain.h +++ b/include/linux/pm_domain.h @@ -64,6 +64,7 @@ struct generic_pm_domain { unsigned int suspended_count; /* System suspend device counter */ unsigned int prepared_count; /* Suspend counter of prepared devices */ unsigned int performance_state; /* Max requested performance state */ + unsigned int can_set_performance_state; /* Number of parent domains supporting set state */ int (*power_off)(struct generic_pm_domain *domain); int (*power_on)(struct generic_pm_domain *domain); int (*get_performance_state)(struct device *dev, unsigned long rate); > > + struct gpd_link *link; > > + > > + if (genpd->set_performance_state) > > + return true; > > + > > + list_for_each_entry(link, &genpd->slave_links, slave_node) { > > + if (genpd_has_set_performance_state(link->master)) > > + return true; > > + } > > + > > + return false; > > +} > > + > > +/** > > + * pm_genpd_has_performance_state - Checks if power domain does performance > > + * state management. > > + * > > + * @dev: Device whose power domain is getting inquired. > > + * > > + * This must be called by the user drivers, before they start calling > > + * pm_genpd_update_performance_state(), to guarantee that all dependencies are > > + * met and the device's genpd supports performance states. > > + * > > + * It is assumed that the user driver guarantees that the genpd wouldn't be > > + * detached while this routine is getting called. > > + * > > + * Returns "true" if device's genpd supports performance states, "false" > > + * otherwise. > > + */ > > +bool pm_genpd_has_performance_state(struct device *dev) > > +{ > > + struct generic_pm_domain *genpd = genpd_lookup_dev(dev); > > + > > + /* The parent domain must have set get_performance_state() */ > > This comment is wrong. This is not about *parent* domains. > > Instead I think it should say: "The genpd must have the > ->get_performance_state() assigned." ... I was (wrongly) calling power domain of a device as its parent power domain and so yeah I agree with your feedback. > > + if (!IS_ERR(genpd) && genpd->get_performance_state) { > > + if (genpd_has_set_performance_state(genpd)) > > + return true; > > ... while this is about verifying that some of the domains (genpds) in > domain topology has a ->set_performance_state() callback. > > In other words, either the genpd or some of its masters must have a > ->set_performance_state() callback. Right. > That makes me wonder: what will happen if there are more than one > master having a ->set_perfomance_state() callback assigned? I guess > that is non-allowed configuration? This patch supports them at least. Device's domain can have multiple masters which require this configuration, but the same performance index will be used for all of them (for simplicity unless we have a real example to serve). > > + > > + /* > > + * A genpd with ->get_performance_state() callback must also > > + * allow setting performance state. > > + */ > > + dev_err(dev, "genpd doesn't support setting performance state\n"); > > + } > > + > > + return false; > > +} > > +EXPORT_SYMBOL_GPL(pm_genpd_has_performance_state); > > + > > +/* > > + * Re-evaluate performance state of a power domain. Finds the highest requested > > + * performance state by the devices and subdomains within the power domain and > > + * then tries to change its performance state. If the power domain doesn't have > > + * a set_performance_state() callback, then we move the request to its parent > > + * power domain. > > + * > > + * Locking: Access (or update) to device's "pd_data->performance_state" field > > + * happens only with parent domain's lock held. Subdomains have their > > What is a *parent* domain here? Same crappy wording I have been using. Its about device's genpd. > In genpd we try to use the terminology of master- and sub-domains. > Could you re-phrase this to get some clarity on what you try to > explain from the above? Yeah, sure. So do we call a device's power domain as its master domain? I thought that master is only used in context of sub-domains, right ? > > + * "genpd->performance_state" protected with their own lock (and they are the > > + * only user of this field) and their per-parent-domain > > + * "link->performance_state" field is protected with individual parent power > > + * domain's lock and is only accessed/updated with that lock held. > > + */ > > I recall we discussed this off-list, but I am not sure this was the > conclusion on how to solve the locking. :-) More comments below. There was no conclusion there :) > > +static int genpd_update_performance_state(struct generic_pm_domain *genpd, > > + int depth) > > +{ > > + struct generic_pm_domain_data *pd_data; > > + struct generic_pm_domain *master; > > + struct pm_domain_data *pdd; > > + unsigned int state = 0, prev; > > + struct gpd_link *link; > > + int ret; > > + > > + /* Traverse all devices within the domain */ > > + list_for_each_entry(pdd, &genpd->dev_list, list_node) { > > + pd_data = to_gpd_data(pdd); > > + > > + if (pd_data->performance_state > state) > > + state = pd_data->performance_state; > > + } > > I don't think walking the list of devices for each master domain is > necessary, unless the aggregated performance state from the subdomain > was increased. Actually we can skip both ^^ and below loop in few cases. Here is the diff: diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c index 599b731fcffc..55bbfdabab53 100644 --- a/drivers/base/power/domain.c +++ b/drivers/base/power/domain.c @@ -509,6 +509,9 @@ EXPORT_SYMBOL_GPL(pm_genpd_has_performance_state); * a set_performance_state() callback, then we move the request to its parent * power domain. * + * The state parameter is the newly requested performance state of the device or + * subdomain for which this routine is called. + * * Locking: Access (or update) to device's "pd_data->performance_state" field * happens only with parent domain's lock held. Subdomains have their * "genpd->performance_state" protected with their own lock (and they are the @@ -517,15 +520,23 @@ EXPORT_SYMBOL_GPL(pm_genpd_has_performance_state); * domain's lock and is only accessed/updated with that lock held. */ static int genpd_update_performance_state(struct generic_pm_domain *genpd, - int depth) + int state, int depth) { struct generic_pm_domain_data *pd_data; struct generic_pm_domain *master; struct pm_domain_data *pdd; - unsigned int state = 0, prev; + unsigned int prev; struct gpd_link *link; int ret; + /* New requested state is same as Max requested state */ + if (state == genpd->performance_state) + return 0; + + /* New requested state is higher than Max requested state */ + if (state > genpd->performance_state) + goto update_state; + /* Traverse all devices within the domain */ list_for_each_entry(pdd, &genpd->dev_list, list_node) { pd_data = to_gpd_data(pdd); @@ -540,9 +551,7 @@ static int genpd_update_performance_state(struct generic_pm_domain *genpd, state = link->performance_state; } - if (genpd->performance_state == state) - return 0; - +update_state: if (genpd->set_performance_state) { ret = genpd->set_performance_state(genpd, state); if (!ret) @@ -563,7 +572,7 @@ static int genpd_update_performance_state(struct generic_pm_domain *genpd, genpd_lock_nested(master, depth + 1); link->performance_state = state; - ret = genpd_update_performance_state(master, depth + 1); + ret = genpd_update_performance_state(master, state, depth + 1); if (ret) link->performance_state = prev; @@ -587,7 +596,7 @@ static int genpd_update_performance_state(struct generic_pm_domain *genpd, genpd_lock_nested(master, depth + 1); link->performance_state = prev; - if (genpd_update_performance_state(master, depth + 1)) + if (genpd_update_performance_state(master, prev, depth + 1)) pr_err("%s: Failed to roll back to %d performance state\n", genpd->name, prev); genpd_unlock(master); @@ -659,7 +668,7 @@ int pm_genpd_update_performance_state(struct device *dev, unsigned long rate) goto unlock; } - ret = genpd_update_performance_state(genpd, 0); + ret = genpd_update_performance_state(genpd, state, 0); if (ret) __dev_update_performance_state(dev, state);
On 21-07-17, 10:35, Ulf Hansson wrote: > This depends on how drivers are dealing with runtime PM in conjunction > with the new pm_genpd_update_performance_state(). > > In case you don't want to manage some of this in genpd, then each > driver will have to drop their constraints every time they are about > to runtime suspend its device. And restore them at runtime resume. > > To me, that's seems like a bad idea. Then it's better to make genpd > deal with this - somehow. Right. So we should call the ->set_performance_state() from off/on as well. Will do that. > Yes! > > On top of that change, you could also add some validation if the > get/set callbacks is there are any constraints on how they must be > assigned. I am not sure if I understood that, sorry. What other constraints are you talking about ? > >> From a locking point of view we always traverse the topology from > >> bottom an up. In other words, we walk the genpd's ->slave_links, and > >> lock the masters in the order the are defined via the slave_links > >> list. The order is important to avoid deadlocks. I don't think you > >> should walk the master_links as being done above, especially not > >> without using locks. > > > > So we need to look at the performance states of the subdomains of a > > master. The way it is done in this patch with help of > > link->performance_state, we don't need that locking while traversing > > the master_links list. Here is how: > > > > - Master's (genpd) master_links list is only updated under master's > > lock, which we have already taken here. So master_links list can't > > get updated concurrently. > > > > - The link->performance_state field of a subdomain (or slave) is only > > updated from within the master's lock. And we are reading it here > > from the same lock. > > > > AFAIU, there shouldn't be any deadlocks or locking issues here. Can > > you describe some case that may blow up ? > > My main concern is the order of how you take the looks. We never take > a masters lock before the current domain lock. Right and this patch doesn't break that. > And when walking the topology, we use the slave links and locks the > first master from that list. Continues with that tree, then get back > to slave list and pick the next master. Again, that's how this patch does it. > If you change that order, we could end getting deadlocks. And because that order isn't changed at all, we shouldn't have deadlocks. > >> A general comment is that I think you should look more closely in the > >> code of genpd_power_off|on(). And also how it calls the > >> ->power_on|off() callbacks. > >> > >> Depending whether you want to update the performance state of the > >> master domain before the subdomain or the opposite, you will find one > >> of them being suited for this case as well. > > > > Isn't it very much similar to that already ? The only major difference > > is link->performance_state and I already explained why is it required > > to be done that way to avoid deadlocks. > > No, because you walk the master lists. Thus getting a different order or locks. > > I did some drawing of this, using the slave links, and I don't see any > issues why you can't use that instead. Damn, I am confused on which part are you talking about. Let me paste the code here once again and clarify how this is supposed to work just fine :) >> +static int genpd_update_performance_state(struct generic_pm_domain *genpd, >> + int depth) >> +{ genpd is already locked. >> + struct generic_pm_domain_data *pd_data; >> + struct generic_pm_domain *master; >> + struct pm_domain_data *pdd; >> + unsigned int state = 0, prev; >> + struct gpd_link *link; >> + int ret; >> + >> + /* Traverse all devices within the domain */ >> + list_for_each_entry(pdd, &genpd->dev_list, list_node) { >> + pd_data = to_gpd_data(pdd); >> + >> + if (pd_data->performance_state > state) >> + state = pd_data->performance_state; >> + } >> + >> + /* Traverse all subdomains within the domain */ >> + list_for_each_entry(link, &genpd->master_links, master_node) { This is the only place where we look at all the sub-domains, but we don't need locking here at all as link->performance_state is only accessed while "genpd" is locked. It doesn't need sub-domain's lock. >> + if (link->performance_state > state) >> + state = link->performance_state; >> + } >> + >> + if (genpd->performance_state == state) >> + return 0; >> + >> + if (genpd->set_performance_state) { >> + ret = genpd->set_performance_state(genpd, state); >> + if (!ret) >> + genpd->performance_state = state; >> + >> + return ret; >> + } >> + >> + /* >> + * Not all domains support updating performance state. Move on to their >> + * parent domains in that case. >> + */ >> + prev = genpd->performance_state; >> + The below part is what I assumed you were commenting on and this is very similar to how on/off are implemented today. >> + list_for_each_entry(link, &genpd->slave_links, slave_node) { i.e. we traverse the list of masters from slave_links list. >> + master = link->master; >> + >> + genpd_lock_nested(master, depth + 1); Take master's lock (and "genpd" is already locked when this function is called.) >> + >> + link->performance_state = state; >> + ret = genpd_update_performance_state(master, depth + 1); Do recursive calling and so the master tree will finish first before moving to next master. >> + if (ret) >> + link->performance_state = prev; >> + The above can actually be done outside of this lock as we are only concerned about "genpd" lock here. Where do you think the order of locking is screwed up ? >> + genpd_unlock(master); >> + >> + if (ret) >> + goto err; >> + } >> + >> + /* >> + * The parent domains are updated now, lets get genpd performance_state >> + * in sync with those. >> + */ >> + genpd->performance_state = state; >> + return 0; >> + >> +err: >> + list_for_each_entry_continue_reverse(link, &genpd->slave_links, >> + slave_node) { >> + master = link->master; >> + >> + genpd_lock_nested(master, depth + 1); >> + link->performance_state = prev; >> + if (genpd_update_performance_state(master, depth + 1)) >> + pr_err("%s: Failed to roll back to %d performance state\n", >> + genpd->name, prev); >> + genpd_unlock(master); >> + } >> + >> + return ret; >> +} >> + -- viresh
On 21 July 2017 at 11:05, Viresh Kumar <viresh.kumar@linaro.org> wrote: > On 21-07-17, 10:35, Ulf Hansson wrote: >> This depends on how drivers are dealing with runtime PM in conjunction >> with the new pm_genpd_update_performance_state(). >> >> In case you don't want to manage some of this in genpd, then each >> driver will have to drop their constraints every time they are about >> to runtime suspend its device. And restore them at runtime resume. >> >> To me, that's seems like a bad idea. Then it's better to make genpd >> deal with this - somehow. > > Right. So we should call the ->set_performance_state() from off/on as > well. Will do that. > >> Yes! >> >> On top of that change, you could also add some validation if the >> get/set callbacks is there are any constraints on how they must be >> assigned. > > I am not sure if I understood that, sorry. What other constraints are > you talking about ? Just thinking that if a genpd is about to be added as a subdomain, and it has ->get_performance_state(), but not ->set_performance_state(), perhaps we should require its master to have ->set_performance_state(). Anyway, I let you do the thinking of what is and what is not needed here. [...] >> >> My main concern is the order of how you take the looks. We never take >> a masters lock before the current domain lock. > > Right and this patch doesn't break that. > >> And when walking the topology, we use the slave links and locks the >> first master from that list. Continues with that tree, then get back >> to slave list and pick the next master. > > Again, that's how this patch does it. > >> If you change that order, we could end getting deadlocks. > > And because that order isn't changed at all, we shouldn't have > deadlocks. True. Trying to clarify more below... > >> >> A general comment is that I think you should look more closely in the >> >> code of genpd_power_off|on(). And also how it calls the >> >> ->power_on|off() callbacks. >> >> >> >> Depending whether you want to update the performance state of the >> >> master domain before the subdomain or the opposite, you will find one >> >> of them being suited for this case as well. >> > >> > Isn't it very much similar to that already ? The only major difference >> > is link->performance_state and I already explained why is it required >> > to be done that way to avoid deadlocks. >> >> No, because you walk the master lists. Thus getting a different order or locks. >> >> I did some drawing of this, using the slave links, and I don't see any >> issues why you can't use that instead. > > Damn, I am confused on which part are you talking about. Let me paste > the code here once again and clarify how this is supposed to work just fine :) I should have been more clear. Walking the master list, then checking each link without using locks - why is that safe? Then even if you think it's safe, then please explain in detail why its needed. Walking the slave list as being done for power off/on should work perfectly okay for your case as well. No? [...] Kind regards Uffe
On 23-07-17, 09:20, Ulf Hansson wrote: > I should have been more clear. Walking the master list, then checking > each link without using locks - why is that safe? > > Then even if you think it's safe, then please explain in detail why its needed. > > Walking the slave list as being done for power off/on should work > perfectly okay for your case as well. No? I got it. I will try to explain why it is done this way with the help of two versions of genpd_update_performance_state() routine. The first one is from a previous version and second one is from the current series. Just as a note, the problem is not with traversing slave_links list but the master_links list. 1.) Previous Version (Have deadlock issues, as you reported then). >> +static int genpd_update_performance_state(struct generic_pm_domain *genpd, >> + int depth) >> +{ The genpd is already locked here.. >> + struct generic_pm_domain_data *pd_data; >> + struct generic_pm_domain *subdomain; >> + struct pm_domain_data *pdd; >> + unsigned int state = 0, prev; >> + struct gpd_link *link; >> + int ret; >> + >> + /* Traverse all devices within the domain */ >> + list_for_each_entry(pdd, &genpd->dev_list, list_node) { >> + pd_data = to_gpd_data(pdd); >> + >> + if (pd_data->performance_state > state) >> + state = pd_data->performance_state; >> + } Above is fine as we traversed list of devices that are powered by the PM domain. No additional locks are required. >> + /* Traverse all subdomains within the domain */ >> + list_for_each_entry(link, &genpd->master_links, master_node) { >> + subdomain = link->slave; >> + >> + if (subdomain->performance_state > state) >> + state = subdomain->performance_state; >> + } But this is not fine at all. subdomain->performance_state might get updated from another thread for another genpd. And so we need locking here, but we can't do that as we need to take locks starting from slaves to masters. This is what you correctly pointed out in earlier versions. >> + if (genpd->performance_state == state) >> + return 0; >> + >> + /* >> + * Not all domains support updating performance state. Move on to their >> + * parent domains in that case. >> + */ >> + if (genpd->set_performance_state) { >> + ret = genpd->set_performance_state(genpd, state); >> + if (!ret) >> + genpd->performance_state = state; >> + >> + return ret; >> + } >> + >> + prev = genpd->performance_state; >> + genpd->performance_state = state; This is racy as well (because of the earlier traversal of master-list), as genpd->performance_state might be getting read for one of its masters currently (from another instance of this same routine). >> + >> + list_for_each_entry(link, &genpd->slave_links, slave_node) { >> + struct generic_pm_domain *master = link->master; >> + >> + genpd_lock_nested(master, depth + 1); >> + ret = genpd_update_performance_state(master, depth + 1); >> + genpd_unlock(master); >> >> ... Handle errors here. >> + } >> + >> + return 0; >> +} So the conclusion was that we surely can't lock the subdomains while running genpd_update_performance_state() for a master genpd. And that's what the below latest code tried to address. 2.) New code, which shouldn't have any of those deadlock issues. >> +static int genpd_update_performance_state(struct generic_pm_domain *genpd, genpd is still locked from its caller. >> + int depth) >> +{ >> + struct generic_pm_domain_data *pd_data; >> + struct generic_pm_domain *master; >> + struct pm_domain_data *pdd; >> + unsigned int state = 0, prev; >> + struct gpd_link *link; >> + int ret; >> + >> + /* Traverse all devices within the domain */ >> + list_for_each_entry(pdd, &genpd->dev_list, list_node) { >> + pd_data = to_gpd_data(pdd); >> + >> + if (pd_data->performance_state > state) >> + state = pd_data->performance_state; >> + } >> + Above remains the same and shouldn't have any issues. >> + /* Traverse all subdomains within the domain */ >> + list_for_each_entry(link, &genpd->master_links, master_node) { >> + if (link->performance_state > state) >> + state = link->performance_state; >> + } >> + Instead of a single performance_state field for the entire subdomain structure, we store a performance_state value for each master <-> subdomain pair. And this field is protected by the master lock, always. Since the genpd was already locked, the link->performance_state field of all its subdomains can be accessed without further locking. >> + if (genpd->performance_state == state) >> + return 0; >> + >> + if (genpd->set_performance_state) { >> + ret = genpd->set_performance_state(genpd, state); >> + if (!ret) >> + genpd->performance_state = state; >> + >> + return ret; >> + } >> + >> + /* >> + * Not all domains support updating performance state. Move on to their >> + * parent domains in that case. >> + */ >> + prev = genpd->performance_state; >> + Lets look at the below one now. >> + list_for_each_entry(link, &genpd->slave_links, slave_node) { >> + master = link->master; >> + >> + genpd_lock_nested(master, depth + 1); >> + >> + link->performance_state = state; (I incorrectly mentioned in my last reply to you that this can be present outside of the lock, this can't be.) So here we have taken the master's lock and updated the link shared between master <-> subdomain (genpd here). >> + ret = genpd_update_performance_state(master, depth + 1); So when this recursive call is made for the "master" domain, it will read link->performance_state of all its subdomains (from the earlier for-loop) and the current "genpd" domain will be part of that master-list. And that's why we updated link->performance_state before calling the genpd_update_performance_state() routine above, so that its new value can be available. IOW, we can't have a race now where link->performance_state for any of the subdomains of genpd is read/updated in parallel. >> + if (ret) >> + link->performance_state = prev; >> + >> + genpd_unlock(master); >> + >> + if (ret) >> + goto err; >> + } >> + >> + /* >> + * The parent domains are updated now, lets get genpd performance_state >> + * in sync with those. >> + */ And once we are done with all the master domains, we can update the genpd->performance_state safely. >> + genpd->performance_state = state; >> + return 0; >> + >> +err: >> ... >> +} >> + Was I able to clarify your doubts ? -- viresh
On 21-07-17, 10:35, Ulf Hansson wrote: > >> > +/* > >> > + * Returns true if anyone in genpd's parent hierarchy has > >> > + * set_performance_state() set. > >> > + */ > >> > +static bool genpd_has_set_performance_state(struct generic_pm_domain *genpd) > >> > +{ > >> > >> So this function will be become in-directly called by generic drivers > >> that supports DVFS of the genpd for their devices. > >> > >> I think the data you validate here would be better to be pre-validated > >> at pm_genpd_init() and at pm_genpd_add|remove_subdomain() and the > >> result stored in a variable in the genpd struct. Especially when a > >> subdomain is added, that is a point when you can verify the > >> *_performance_state() callbacks, and thus make sure it's a correct > >> setup from the topology point of view. Looks like I have to keep this routine as is and your solution may not work well. :( > > Something like this ? > > > > diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c > > index 4a898e095a1d..182c1911ea9c 100644 > > --- a/drivers/base/power/domain.c > > +++ b/drivers/base/power/domain.c > > @@ -466,25 +466,6 @@ static int genpd_dev_pm_qos_notifier(struct notifier_block *nb, > > return NOTIFY_DONE; > > } > > > > -/* > > - * Returns true if anyone in genpd's parent hierarchy has > > - * set_performance_state() set. > > - */ > > -static bool genpd_has_set_performance_state(struct generic_pm_domain *genpd) > > -{ > > - struct gpd_link *link; > > - > > - if (genpd->set_performance_state) > > - return true; > > - > > - list_for_each_entry(link, &genpd->slave_links, slave_node) { > > - if (genpd_has_set_performance_state(link->master)) > > - return true; > > - } > > - > > - return false; > > -} > > - > > /** > > * pm_genpd_has_performance_state - Checks if power domain does performance > > * state management. > > @@ -507,7 +488,7 @@ bool pm_genpd_has_performance_state(struct device *dev) > > > > /* The parent domain must have set get_performance_state() */ > > if (!IS_ERR(genpd) && genpd->get_performance_state) { > > - if (genpd_has_set_performance_state(genpd)) > > + if (genpd->can_set_performance_state) > > return true; > > > > /* > > @@ -1594,6 +1575,8 @@ static int genpd_add_subdomain(struct generic_pm_domain *genpd, > > if (genpd_status_on(subdomain)) > > genpd_sd_counter_inc(genpd); > > > > + subdomain->can_set_performance_state += genpd->can_set_performance_state; > > + > > out: > > genpd_unlock(genpd); > > genpd_unlock(subdomain); > > @@ -1654,6 +1637,8 @@ int pm_genpd_remove_subdomain(struct generic_pm_domain *genpd, > > if (genpd_status_on(subdomain)) > > genpd_sd_counter_dec(genpd); > > > > + subdomain->can_set_performance_state -= genpd->can_set_performance_state; > > + > > ret = 0; > > break; > > } > > @@ -1721,6 +1706,7 @@ int pm_genpd_init(struct generic_pm_domain *genpd, > > genpd->max_off_time_changed = true; > > genpd->provider = NULL; > > genpd->has_provider = false; > > + genpd->can_set_performance_state = !!genpd->set_performance_state; > > genpd->domain.ops.runtime_suspend = genpd_runtime_suspend; > > genpd->domain.ops.runtime_resume = genpd_runtime_resume; > > genpd->domain.ops.prepare = pm_genpd_prepare; > > diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h > > index bf90177208a2..995d0cb1bc14 100644 > > --- a/include/linux/pm_domain.h > > +++ b/include/linux/pm_domain.h > > @@ -64,6 +64,7 @@ struct generic_pm_domain { > > unsigned int suspended_count; /* System suspend device counter */ > > unsigned int prepared_count; /* Suspend counter of prepared devices */ > > unsigned int performance_state; /* Max requested performance state */ > > + unsigned int can_set_performance_state; /* Number of parent domains supporting set state */ > > int (*power_off)(struct generic_pm_domain *domain); > > int (*power_on)(struct generic_pm_domain *domain); > > int (*get_performance_state)(struct device *dev, unsigned long rate); > > > > Yes! The above diff will work fine only for the case where the master domain has all its masters set properly before genpd_add_subdomain() is called for the subdomain, as the genpd->can_set_performance_state count wouldn't change after that. But if the masters of the master are linked to the master after genpd_add_subdomain() is called for the subdomain, then we wouldn't be update the subdomain->can_set_performance_state field later. For example, consider this scenario: Domain A (has set_performance_state()) Domain B Domain C (both don't have set_performance_state()) Domain D Domain E (both don't have set_performance_state(), but have get_performance_state()) and here is the call sequence: genpd_add_subdomain(B, D); can_set_performance_state of B and D = 0; genpd_add_subdomain(C, E); ... C and E = 0; genpd_add_subdomain(A, B); ... A = 1, B = 1; genpd_add_subdomain(A, C); ... A = 1, C = 1; While the count is set properly for A, B and C, it isn't propagated to C and E. :( Though everything would have worked fine if we had this sequence: genpd_add_subdomain(A, B); ... A = 1, B = 1; genpd_add_subdomain(A, C); ... A = 1, C = 1; genpd_add_subdomain(B, D); ... D = 1 ; genpd_add_subdomain(C, E); ... E = 1; How to fix it? I tried solving that by propagating the count to all the subdomains of the subdomain getting added here. But that requires locking and we can't do that in the reverse direction :( Anyway, genpd_has_set_performance_state() is supposed to be called only ONCE by the drivers and so its fine if we have to traverse the list of subdomains there. I will keep the original code unless you suggest a good way of getting around that. -- viresh
On 28 July 2017 at 13:00, Viresh Kumar <viresh.kumar@linaro.org> wrote: > On 21-07-17, 10:35, Ulf Hansson wrote: >> >> > +/* >> >> > + * Returns true if anyone in genpd's parent hierarchy has >> >> > + * set_performance_state() set. >> >> > + */ >> >> > +static bool genpd_has_set_performance_state(struct generic_pm_domain *genpd) >> >> > +{ >> >> >> >> So this function will be become in-directly called by generic drivers >> >> that supports DVFS of the genpd for their devices. >> >> >> >> I think the data you validate here would be better to be pre-validated >> >> at pm_genpd_init() and at pm_genpd_add|remove_subdomain() and the >> >> result stored in a variable in the genpd struct. Especially when a >> >> subdomain is added, that is a point when you can verify the >> >> *_performance_state() callbacks, and thus make sure it's a correct >> >> setup from the topology point of view. > > Looks like I have to keep this routine as is and your solution may not > work well. :( > >> > Something like this ? >> > >> > diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c >> > index 4a898e095a1d..182c1911ea9c 100644 >> > --- a/drivers/base/power/domain.c >> > +++ b/drivers/base/power/domain.c >> > @@ -466,25 +466,6 @@ static int genpd_dev_pm_qos_notifier(struct notifier_block *nb, >> > return NOTIFY_DONE; >> > } >> > >> > -/* >> > - * Returns true if anyone in genpd's parent hierarchy has >> > - * set_performance_state() set. >> > - */ >> > -static bool genpd_has_set_performance_state(struct generic_pm_domain *genpd) >> > -{ >> > - struct gpd_link *link; >> > - >> > - if (genpd->set_performance_state) >> > - return true; >> > - >> > - list_for_each_entry(link, &genpd->slave_links, slave_node) { >> > - if (genpd_has_set_performance_state(link->master)) >> > - return true; >> > - } >> > - >> > - return false; >> > -} >> > - >> > /** >> > * pm_genpd_has_performance_state - Checks if power domain does performance >> > * state management. >> > @@ -507,7 +488,7 @@ bool pm_genpd_has_performance_state(struct device *dev) >> > >> > /* The parent domain must have set get_performance_state() */ >> > if (!IS_ERR(genpd) && genpd->get_performance_state) { >> > - if (genpd_has_set_performance_state(genpd)) >> > + if (genpd->can_set_performance_state) >> > return true; >> > >> > /* >> > @@ -1594,6 +1575,8 @@ static int genpd_add_subdomain(struct generic_pm_domain *genpd, >> > if (genpd_status_on(subdomain)) >> > genpd_sd_counter_inc(genpd); >> > >> > + subdomain->can_set_performance_state += genpd->can_set_performance_state; >> > + >> > out: >> > genpd_unlock(genpd); >> > genpd_unlock(subdomain); >> > @@ -1654,6 +1637,8 @@ int pm_genpd_remove_subdomain(struct generic_pm_domain *genpd, >> > if (genpd_status_on(subdomain)) >> > genpd_sd_counter_dec(genpd); >> > >> > + subdomain->can_set_performance_state -= genpd->can_set_performance_state; >> > + >> > ret = 0; >> > break; >> > } >> > @@ -1721,6 +1706,7 @@ int pm_genpd_init(struct generic_pm_domain *genpd, >> > genpd->max_off_time_changed = true; >> > genpd->provider = NULL; >> > genpd->has_provider = false; >> > + genpd->can_set_performance_state = !!genpd->set_performance_state; >> > genpd->domain.ops.runtime_suspend = genpd_runtime_suspend; >> > genpd->domain.ops.runtime_resume = genpd_runtime_resume; >> > genpd->domain.ops.prepare = pm_genpd_prepare; >> > diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h >> > index bf90177208a2..995d0cb1bc14 100644 >> > --- a/include/linux/pm_domain.h >> > +++ b/include/linux/pm_domain.h >> > @@ -64,6 +64,7 @@ struct generic_pm_domain { >> > unsigned int suspended_count; /* System suspend device counter */ >> > unsigned int prepared_count; /* Suspend counter of prepared devices */ >> > unsigned int performance_state; /* Max requested performance state */ >> > + unsigned int can_set_performance_state; /* Number of parent domains supporting set state */ >> > int (*power_off)(struct generic_pm_domain *domain); >> > int (*power_on)(struct generic_pm_domain *domain); >> > int (*get_performance_state)(struct device *dev, unsigned long rate); >> > >> >> Yes! > > The above diff will work fine only for the case where the master > domain has all its masters set properly before genpd_add_subdomain() > is called for the subdomain, as the genpd->can_set_performance_state > count wouldn't change after that. But if the masters of the > master are linked to the master after genpd_add_subdomain() is called > for the subdomain, then we wouldn't be update the > subdomain->can_set_performance_state field later. > > For example, consider this scenario: > > Domain A (has set_performance_state()) > > Domain B Domain C (both don't have set_performance_state()) > > Domain D Domain E (both don't have set_performance_state(), but have get_performance_state()) > > > and here is the call sequence: > > genpd_add_subdomain(B, D); can_set_performance_state of B and D = 0; > genpd_add_subdomain(C, E); ... C and E = 0; > genpd_add_subdomain(A, B); ... A = 1, B = 1; > genpd_add_subdomain(A, C); ... A = 1, C = 1; > > While the count is set properly for A, B and C, it isn't propagated to > C and E. :( > > Though everything would have worked fine if we had this sequence: > > genpd_add_subdomain(A, B); ... A = 1, B = 1; > genpd_add_subdomain(A, C); ... A = 1, C = 1; > genpd_add_subdomain(B, D); ... D = 1 ; > genpd_add_subdomain(C, E); ... E = 1; > > How to fix it? I tried solving that by propagating the count to all > the subdomains of the subdomain getting added here. But that requires > locking and we can't do that in the reverse direction :( Yeah, you are right! > > Anyway, genpd_has_set_performance_state() is supposed to be called > only ONCE by the drivers and so its fine if we have to traverse the > list of subdomains there. > > I will keep the original code unless you suggest a good way of getting > around that. Let's invent a new genpd flag, GENPD_FLAG_PERF_STATE! The creator of the genpd then needs to set this before calling pm_genpd_init(). Similar as we are dealing with GENPD_FLAG_PM_CLK. The requirement for GENPD_FLAG_PERF_STATES, is to have the ->get_performance_state() assigned. This shall be verified during pm_genpd_init(). The pm_genpd_has_performance_state() then only need to return true, in cases the device's genpd has GENPD_FLAG_PERF_STATE set - else false. Regarding ->set_performance_state(), let's just make it optional - and when trying to set a new performance state, just walk the genpd hierarchy, from bottom to up, then invoke the callback when it's assigned. Kind regards Uffe
On 29-07-17, 10:24, Ulf Hansson wrote: > Let's invent a new genpd flag, GENPD_FLAG_PERF_STATE! > > The creator of the genpd then needs to set this before calling > pm_genpd_init(). Similar as we are dealing with GENPD_FLAG_PM_CLK. > > The requirement for GENPD_FLAG_PERF_STATES, is to have the > ->get_performance_state() assigned. This shall be verified during > pm_genpd_init(). > > The pm_genpd_has_performance_state() then only need to return true, in > cases the device's genpd has GENPD_FLAG_PERF_STATE set - else false. > > Regarding ->set_performance_state(), let's just make it optional - and > when trying to set a new performance state, just walk the genpd > hierarchy, from bottom to up, then invoke the callback when it's > assigned. Sounds good. -- viresh
Hi, Some platforms have the capability to configure the performance state of their power domains. The process of configuring the performance state is pretty much platform dependent and we may need to work with a wide range of configurables. For some platforms, like Qcom, it can be a positive integer value alone, while in other cases it can be voltage levels, etc. The power-domain framework until now was only designed for the idle state management of the device and this needs to change in order to reuse the power-domain framework for active state management of the devices. The first patch updates the genpd framework to supply new APIs to support active state management and the second patch uses them from the OPP core. Rest of the patches [3-6/6] are included just to show how user drivers would end up using the new APIs and these patches aren't there to get merged and are marked clearly like that. The ideal way is still to get the relation between device and domain states via the DT instead of platform code, but that can be done incrementally later once we have some users for it upstream. This is currently tested by: - me by hacking the kernel a bit with virtual power-domains for the dual A15 exynos platform. I have also tested the complex cases where the device's parent power domain doesn't have set_performance_state() callback set, but parents of that domains have it. Lockdep configs were enabled for these tests. - Rajendra Nayak, on msm8996 platform (Qcom) with MMC controller. Thanks Rajendra for helping me testing this out. Pushed here as well: https://git.linaro.org/people/viresh.kumar/linux.git/log/?h=opp/genpd-performance-state Rebased on: 4.12-rc6 + some OPP core changes [1] and [2]. V7->V8: - Ulf helped a lot in reviewing V7 and pointed out couple of issues, specially in locking while dealing with a hierarchy of power domains. - All those locking issues are sorted out now, even for the complex cases. - genpd_lookup_dev() is used in pm_genpd_has_performance_state() to make sure we have a valid genpd available for the device. - Validation of performance state callbacks isn't done anymore in pm_genpd_init() as it gets called very early and the binding of subdomains to their parent domains happens later. This is handled in pm_genpd_has_performance_state() now, which is called from user drivers. - User driver changes (not to be merged) are included for the first time here, to demonstrate how changes would look finally. V6->V7: - Almost a rewrite, only two patches against 9 in earlier version. - No bindings updated now and domain's performance state aren't passed via DT for now (until we know how users are going to use it). - We also skipped the QoS framework completely and new APIs are provided directly by genpd. V5->V6: - Use freq/voltage in OPP table as it is for power domain and don't create "domain-performance-level" property - Create new "power-domain-opp" property for the devices. - Take care of domain providers that provide multiple domains and extend "operating-points-v2" property to contain a list of phandles - Update code according to those bindings. V4->V5: - Only 3 patches were resent and 2 of them are Acked from Ulf. V3->V4: - Use OPP table for genpd devices as well. - Add struct device to genpd, in order to reuse OPP infrastructure. - Based over: https://marc.info/?l=linux-kernel&m=148972988002317&w=2 - Fixed examples in DT document to have voltage in target,min,max order. V2->V3: - Based over latest pm/linux-next - Bindings and code are merged together - Lots of updates in bindings - the performance-states node is present within the power-domain now, instead of its phandle. - performance-level property is replaced by "reg". - domain-performance-state property of the consumers contain an integer value now instead of phandle. - Lots of updates to the code as well - Patch "PM / QOS: Add default case to the switch" is merged with other patches and the code is changed a bit as well. - Don't pass 'type' to dev_pm_qos_add_notifier(), rather handle all notifiers with a single list. A new patch is added for that. - The OPP framework patch can be applied now and has proper SoB from me. - Dropped "PM / domain: Save/restore performance state at runtime suspend/resume". - Drop all WARN(). - Tested-by Rajendra nayak. V1->V2: - Based over latest pm/linux-next - It is mostly a resend of what is sent earlier as this series hasn't got any reviews so far and Rafael suggested that its better I resend it. - Only the 4/6 patch got an update, which was shared earlier as reply to V1 as well. It has got several fixes for taking care of power domain hierarchy, etc. -- viresh [1] https://marc.info/?l=linux-kernel&m=149499607030364&w=2 [2] https://marc.info/?l=linux-kernel&m=149795317123259&w=2 Rajendra Nayak (4): soc: qcom: rpmpd: Add a powerdomain driver to model cx/mx powerdomains soc: qcom: rpmpd: Add support for get/set performance state mmc: sdhci-msm: Adapt the driver to use OPPs to set clocks/performance state remoteproc: qcom: q6v5: Vote for proxy powerdomain performance state Viresh Kumar (2): PM / Domains: Add support to select performance-state of domains PM / OPP: Support updating performance state of device's power domains .../devicetree/bindings/power/qcom,rpmpd.txt | 10 + arch/arm64/boot/dts/qcom/msm8996.dtsi | 39 ++ drivers/base/power/domain.c | 223 +++++++++++ drivers/base/power/opp/core.c | 48 ++- drivers/base/power/opp/opp.h | 2 + drivers/clk/qcom/gcc-msm8996.c | 8 +- drivers/mmc/host/sdhci-msm.c | 39 +- drivers/remoteproc/qcom_q6v5_pil.c | 20 +- drivers/soc/qcom/Kconfig | 9 + drivers/soc/qcom/Makefile | 1 + drivers/soc/qcom/rpmpd.c | 412 +++++++++++++++++++++ include/linux/pm_domain.h | 22 ++ 12 files changed, 811 insertions(+), 22 deletions(-) create mode 100644 Documentation/devicetree/bindings/power/qcom,rpmpd.txt create mode 100644 drivers/soc/qcom/rpmpd.c -- 2.13.0.71.gd7076ec9c9cb