Messages in this thread | | | From | Ulf Hansson <> | Date | Fri, 28 Dec 2018 16:36:52 +0100 | Subject | Re: [RFC PATCH 1/1] PM / Domains: Add multi PM domains support for attach_dev |
| |
On Thu, 27 Dec 2018 at 18:14, Aisheng Dong <aisheng.dong@nxp.com> wrote: > > Currently attach_dev() in power domain infrastructure still does > not support multi domains case as the struct device *dev passed > down from genpd_dev_pm_attach_by_id() is a virtual PD device, it > does not help for parsing the real device information from device > tree, e.g. Device/Power IDs, Clocks and it's unware of which real > power domain the device should attach.
Thanks for working on this!
I would appreciate if the changelog could clarify the problem a bit. Perhaps something along the lines of the below.
"A genpd provider's ->attach_dev() callback may be invoked with a so called virtual device, which is created by genpd, at the point when a device is being attached to one of its corresponding multiple PM domains.
In these cases, the genpd provider fails to look up any resource, by a clk_get() for example, for the virtual device in question. This is because, the virtual device that was created by genpd, does not have the virt_dev->of_node assigned."
> > Extend the framework a bit to store the multi PM domains information > in per-device struct generic_pm_domain_data, then power domain driver > could retrieve it for necessary operations during attach_dev(). > > Two new APIs genpd_is_mpd_device() and dev_gpd_mpd_data() are also > introduced to ease the driver operation. > > Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> > Cc: Kevin Hilman <khilman@kernel.org> > Cc: Ulf Hansson <ulf.hansson@linaro.org> > Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org> > Signed-off-by: Dong Aisheng <aisheng.dong@nxp.com> > --- > This patch is a follow-up work of the earlier discussion with Ulf Hansson > about the multi PM domains support for the attach_dev() function [1]. > After a bit more thinking, this is a less intrusive implementation with > the mininum impact on the exist function definitions and calling follows. > One known little drawback is that we have to use the device driver private > data (device.drvdata) to pass down the multi domains information in a > earlier time. However, as multi PD devices are created by domain framework, > this seems to be safe to use it in domain core code as device driver > is not likely going to use it. > Anyway, if any better ideas, please let me know. > > With the two new APIs, the using can be simply as: > static int xxx_attach_dev(struct generic_pm_domain *domain, > struct device *dev) > { > ... > if (genpd_is_mpd_device(dev)) { > mpd_data = dev_gpd_mpd_data(dev); > np = mpd_data->parent->of_node; > idx = mpd_data->index; > //dts parsing > ... > } > ...
I think we can make this a lot less complicated. Just assign virt_dev->of_node = of_node_get(dev->of_node), somewhere in genpd_dev_pm_attach_by_id() and before calling __genpd_dev_pm_attach().
Doing that, would mean the genpd provider's ->attach_dev() callback, don't have to distinguish between virtual and non-virtual devices. Instead they should be able to look up resources in the same way as they did before.
Kind regards Uffe
> } > > [1] https://patchwork.kernel.org/patch/10658669/ > --- > drivers/base/power/domain.c | 31 +++++++++++++++++++++++++++++++ > include/linux/pm_domain.h | 23 +++++++++++++++++++++++ > 2 files changed, 54 insertions(+) > > diff --git a/drivers/base/power/domain.c b/drivers/base/power/domain.c > index 7f38a92..1aa0918 100644 > --- a/drivers/base/power/domain.c > +++ b/drivers/base/power/domain.c > @@ -1343,6 +1343,9 @@ static struct generic_pm_domain_data *genpd_alloc_dev_data(struct device *dev, > gpd_data->td.effective_constraint_ns = PM_QOS_RESUME_LATENCY_NO_CONSTRAINT_NS; > gpd_data->nb.notifier_call = genpd_dev_pm_qos_notifier; > > + if (genpd_is_mpd_device(dev)) > + gpd_data->mpd_data = dev_get_drvdata(dev); > + > spin_lock_irq(&dev->power.lock); > > if (dev->power.subsys_data->domain_data) { > @@ -2179,6 +2182,7 @@ EXPORT_SYMBOL_GPL(of_genpd_remove_last); > > static void genpd_release_dev(struct device *dev) > { > + kfree(dev->driver_data); > kfree(dev); > } > > @@ -2320,6 +2324,20 @@ int genpd_dev_pm_attach(struct device *dev) > EXPORT_SYMBOL_GPL(genpd_dev_pm_attach); > > /** > + * genpd_is_mpd_device - Check if a device is associated with multi PM domains > + * @dev: Device to check. > + */ > + > +bool genpd_is_mpd_device(struct device *dev) > +{ > + if (!dev || (dev && !dev->bus)) > + return false; > + > + return dev->bus == &genpd_bus_type; > +}; > +EXPORT_SYMBOL_GPL(genpd_is_mpd_device); > + > +/** > * genpd_dev_pm_attach_by_id - Associate a device with one of its PM domains. > * @dev: The device used to lookup the PM domain. > * @index: The index of the PM domain. > @@ -2338,6 +2356,7 @@ EXPORT_SYMBOL_GPL(genpd_dev_pm_attach); > struct device *genpd_dev_pm_attach_by_id(struct device *dev, > unsigned int index) > { > + struct pm_domain_mpd_data *mpd_data; > struct device *genpd_dev; > int num_domains; > int ret; > @@ -2366,6 +2385,18 @@ struct device *genpd_dev_pm_attach_by_id(struct device *dev, > return ERR_PTR(ret); > } > > + /* Allocate multi power domains data */ > + mpd_data = kzalloc(sizeof(*mpd_data), GFP_KERNEL); > + if (!mpd_data) { > + device_unregister(genpd_dev); > + return ERR_PTR(-ENOMEM); > + } > + > + mpd_data->parent = dev; > + mpd_data->index = index; > + > + dev_set_drvdata(genpd_dev, mpd_data); > + > /* Try to attach the device to the PM domain at the specified index. */ > ret = __genpd_dev_pm_attach(genpd_dev, dev->of_node, index, false); > if (ret < 1) { > diff --git a/include/linux/pm_domain.h b/include/linux/pm_domain.h > index 3b5d728..106d4e7 100644 > --- a/include/linux/pm_domain.h > +++ b/include/linux/pm_domain.h > @@ -144,6 +144,11 @@ struct gpd_timing_data { > bool cached_suspend_ok; > }; > > +struct pm_domain_mpd_data { > + struct device *parent; > + unsigned int index; > +}; > + > struct pm_domain_data { > struct list_head list_node; > struct device *dev; > @@ -151,6 +156,7 @@ struct pm_domain_data { > > struct generic_pm_domain_data { > struct pm_domain_data base; > + struct pm_domain_mpd_data *mpd_data; > struct gpd_timing_data td; > struct notifier_block nb; > unsigned int performance_state; > @@ -262,10 +268,17 @@ unsigned int of_genpd_opp_to_performance_state(struct device *dev, > struct device_node *np); > > int genpd_dev_pm_attach(struct device *dev); > +bool genpd_is_mpd_device(struct device *dev); > struct device *genpd_dev_pm_attach_by_id(struct device *dev, > unsigned int index); > struct device *genpd_dev_pm_attach_by_name(struct device *dev, > char *name); > + > +static inline struct pm_domain_mpd_data *dev_gpd_mpd_data(struct device *dev) > +{ > + return dev_gpd_data(dev)->mpd_data; > +} > + > #else /* !CONFIG_PM_GENERIC_DOMAINS_OF */ > static inline int of_genpd_add_provider_simple(struct device_node *np, > struct generic_pm_domain *genpd) > @@ -311,6 +324,11 @@ static inline int genpd_dev_pm_attach(struct device *dev) > return 0; > } > > +static bool genpd_is_mpd_device(struct device *dev) > +{ > + return false; > +} > + > static inline struct device *genpd_dev_pm_attach_by_id(struct device *dev, > unsigned int index) > { > @@ -323,6 +341,11 @@ static inline struct device *genpd_dev_pm_attach_by_name(struct device *dev, > return NULL; > } > > +static inline struct pm_domain_mpd_data *dev_gpd_mpd_data(struct device *dev) > +{ > + return ERR_PTR(-ENOSYS); > +} > + > static inline > struct generic_pm_domain *of_genpd_remove_last(struct device_node *np) > { > -- > 2.7.4 >
| |