| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAJZ5v0gotNhMKe8EiGZNFf_WTqoEHVj4wXruU3ZiHSZ2DUK0NA@mail.gmail.com>
Date: Thu, 24 Oct 2024 14:22:46 +0200
From: "Rafael J. Wysocki" <rafael@...nel.org>
To: Lukasz Luba <lukasz.luba@....com>
Cc: "Rafael J. Wysocki" <rjw@...ysocki.net>, LKML <linux-kernel@...r.kernel.org>,
Linux PM <linux-pm@...r.kernel.org>, Daniel Lezcano <daniel.lezcano@...aro.org>,
Zhang Rui <rui.zhang@...el.com>,
Srinivas Pandruvada <srinivas.pandruvada@...ux.intel.com>
Subject: Re: [PATCH v1 09/10] thermal: core: Use trip lists for trip crossing detection
On Thu, Oct 24, 2024 at 1:47 PM Lukasz Luba <lukasz.luba@....com> wrote:
>
>
>
> On 10/16/24 12:35, Rafael J. Wysocki wrote:
> > From: Rafael J. Wysocki <rafael.j.wysocki@...el.com>
> >
> > Modify the thermal core to use three lists of trip points:
> >
> > trips_high, containing trips with thresholds strictly above the current
> > thermal zone temperature,
> >
> > trips_reached, containing trips with thresholds at or below the current
> > zone temperature,
> >
> > trips_invalid, containing trips with temperature equal to
> > THERMAL_ZONE_INVALID,
> >
> > where the first two lists are always sorted by the current trip
> > threshold.
> >
> > For each trip in trips_high, there is no mitigation under way and
> > the trip threshold is equal to its temperature. In turn, for each
> > trip in trips_reached, there is mitigation under way and the trip
> > threshold is equal to its low temperature. The trips in trips_invalid,
> > of course, need not be taken into consideration.
> >
> > The idea is to make __thermal_zone_device_update() walk trips_high and
> > trips_reached instead of walking the entire table of trip points in a
> > thermal zone. Usually, it will only need to walk a few entries in one
> > of the lists and check one entry in the other list, depending on the
> > direction of the zone temperature changes, because crossing many trips
> > by the zone temperature in one go between two consecutive temperature
> > checks should be unlikely (if it occurs often, the thermal zone
> > temperature should probably be checked more often either or there
> > are too many trips).
> >
> > This also helps to eliminate one temporary trip list used for trip
> > crossing notification (only one temporary list is needed for this
> > purpose instead of two) and the remaining temporary list may be sorted
> > by the current trip threshold value, like the trips_reached list, so
> > the additional notify_temp field in struct thermal_trip_desc is not
> > necessary any more.
> >
> > Moreover, since the trips_reached and trips_high lists are sorted,
> > the "low" and "high" values needed by thermal_zone_set_trips() can be
> > determined in a straightforward way by looking at one end of each list.
> >
> > Of course, additional work is needed in some places in order to
> > maintain the ordering of the lists, but it is limited to situations
> > that should be rare, like updating a trip point temperature or
> > hysteresis, thermal zone initialization, or system resume.
>
> Good point I agree. Even the trips from userspace that Daniel
> proposed shouldn't impact that design decision IMHO.
> Any disagreement (or am I missing something)?
No disagreement AFAIK.
> >
> > Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@...el.com>
> > ---
> > drivers/thermal/thermal_core.c | 211 +++++++++++++++++++++++++----------------
> > drivers/thermal/thermal_core.h | 7 +
> > 2 files changed, 136 insertions(+), 82 deletions(-)
> >
> > Index: linux-pm/drivers/thermal/thermal_core.h
> > ===================================================================
> > --- linux-pm.orig/drivers/thermal/thermal_core.h
> > +++ linux-pm/drivers/thermal/thermal_core.h
> > @@ -33,7 +33,6 @@ struct thermal_trip_desc {
> > struct thermal_trip_attrs trip_attrs;
> > struct list_head list_node;
> > struct list_head thermal_instances;
> > - int notify_temp;
> > int threshold;
> > };
> >
> > @@ -78,6 +77,9 @@ struct thermal_governor {
> > * @device: &struct device for this thermal zone
> > * @removal: removal completion
> > * @resume: resume completion
> > + * @trips_high: trips above the current zone temperature
> > + * @trips_reached: trips below or at the current zone temperature
> > + * @trips_invalid: trips with invalid temperature
> > * @mode: current mode of this thermal zone
> > * @devdata: private pointer for device private data
> > * @num_trips: number of trip points the thermal zone supports
> > @@ -118,6 +120,9 @@ struct thermal_zone_device {
> > struct completion removal;
> > struct completion resume;
> > struct attribute_group trips_attribute_group;
> > + struct list_head trips_high;
> > + struct list_head trips_reached;
> > + struct list_head trips_invalid;
> > enum thermal_device_mode mode;
> > void *devdata;
> > int num_trips;
> > Index: linux-pm/drivers/thermal/thermal_core.c
> > ===================================================================
> > --- linux-pm.orig/drivers/thermal/thermal_core.c
> > +++ linux-pm/drivers/thermal/thermal_core.c
> > @@ -421,7 +421,7 @@ static void move_trip_to_sorted_list(str
> >
> > /* Assume that the new entry is likely to be the last one. */
> > list_for_each_entry_reverse(entry, list, list_node) {
> > - if (entry->notify_temp <= td->notify_temp) {
> > + if (entry->threshold <= td->threshold) {
> > list_add(&td->list_node, &entry->list_node);
> > return;
> > }
> > @@ -429,53 +429,6 @@ static void move_trip_to_sorted_list(str
> > list_add(&td->list_node, list);
> > }
> >
> > -static void handle_thermal_trip(struct thermal_zone_device *tz,
> > - struct thermal_trip_desc *td,
> > - struct list_head *way_up_list,
> > - struct list_head *way_down_list)
> > -{
> > - const struct thermal_trip *trip = &td->trip;
> > - int old_threshold;
> > -
> > - if (trip->temperature == THERMAL_TEMP_INVALID)
> > - return;
> > -
> > - /*
> > - * If the trip temperature or hysteresis has been updated recently,
> > - * the threshold needs to be computed again using the new values.
> > - * However, its initial value still reflects the old ones and that
> > - * is what needs to be compared with the previous zone temperature
> > - * to decide which action to take.
> > - */
> > - old_threshold = td->threshold;
> > - td->threshold = trip->temperature;
> > -
> > - if (tz->last_temperature >= old_threshold &&
> > - tz->last_temperature != THERMAL_TEMP_INIT) {
> > - /*
> > - * Mitigation is under way, so it needs to stop if the zone
> > - * temperature falls below the low temperature of the trip.
> > - * In that case, the trip temperature becomes the new threshold.
> > - */
> > - if (tz->temperature < trip->temperature - trip->hysteresis) {
> > - td->notify_temp = trip->temperature - trip->hysteresis;
> > - move_trip_to_sorted_list(td, way_down_list);
> > - } else {
> > - td->threshold -= trip->hysteresis;
> > - }
> > - } else if (tz->temperature >= trip->temperature) {
> > - /*
> > - * There is no mitigation under way, so it needs to be started
> > - * if the zone temperature exceeds the trip one. The new
> > - * threshold is then set to the low temperature of the trip.
> > - */
> > - td->notify_temp = trip->temperature;
> > - move_trip_to_sorted_list(td, way_up_list);
> > -
> > - td->threshold -= trip->hysteresis;
> > - }
> > -}
> > -
> > static void thermal_zone_device_check(struct work_struct *work)
> > {
> > struct thermal_zone_device *tz = container_of(work, struct
> > @@ -484,9 +437,30 @@ static void thermal_zone_device_check(st
> > thermal_zone_device_update(tz, THERMAL_EVENT_UNSPECIFIED);
> > }
> >
> > +static void move_to_trips_high(struct thermal_zone_device *tz,
> > + struct thermal_trip_desc *td)
> > +{
> > + td->threshold = td->trip.temperature;
> > + move_trip_to_sorted_list(td, &tz->trips_high);
> > +}
> > +
> > +static void move_to_trips_reached(struct thermal_zone_device *tz,
> > + struct thermal_trip_desc *td)
> > +{
> > + td->threshold = td->trip.temperature - td->trip.hysteresis;
> > + move_trip_to_sorted_list(td, &tz->trips_reached);
> > +}
> > +
> > +static void move_to_trips_invalid(struct thermal_zone_device *tz,
> > + struct thermal_trip_desc *td)
> > +{
> > + td->threshold = INT_MAX;
> > + list_move(&td->list_node, &tz->trips_invalid);
> > +}
> > +
> > static void thermal_zone_device_init(struct thermal_zone_device *tz)
> > {
> > - struct thermal_trip_desc *td;
> > + struct thermal_trip_desc *td, *next;
> >
> > INIT_DELAYED_WORK(&tz->poll_queue, thermal_zone_device_check);
> >
> > @@ -500,6 +474,21 @@ static void thermal_zone_device_init(str
> > list_for_each_entry(instance, &td->thermal_instances, trip_node)
> > instance->initialized = false;
> > }
> > + /*
> > + * At this point, all valid trips need to be moved to trips_high so that
> > + * mitigation can be started if the zone temperature is above them.
> > + */
> > + list_for_each_entry_safe(td, next, &tz->trips_invalid, list_node) {
> > + if (td->trip.temperature != THERMAL_TEMP_INVALID)
> > + move_to_trips_high(tz, td);
> > + }
> > + /* The trips_reached list may not be empty during system resume. */
> > + list_for_each_entry_safe(td, next, &tz->trips_reached, list_node) {
> > + if (td->trip.temperature == THERMAL_TEMP_INVALID)
> > + move_to_trips_invalid(tz, td);
> > + else
> > + move_to_trips_high(tz, td);
> > + }
> > }
> >
> > static void thermal_governor_trip_crossed(struct thermal_governor *governor,
> > @@ -544,45 +533,120 @@ static void thermal_trip_crossed(struct
> > void thermal_zone_set_trip_hyst(struct thermal_zone_device *tz,
> > struct thermal_trip *trip, int hyst)
> > {
> > + struct thermal_trip_desc *td = trip_to_trip_desc(trip);
> > +
> > WRITE_ONCE(trip->hysteresis, hyst);
> > thermal_notify_tz_trip_change(tz, trip);
> > + /*
> > + * If the zone temperature is above or at the trip tmperature, the trip
> > + * is in the trips_reached list and its threshold is equal to its low
> > + * temperature. It needs to stay in that list, but its threshold needs
> > + * to be updated and the list ordering may need to be restored.
> > + */
> > + if (tz->temperature >= td->threshold)
> > + move_to_trips_reached(tz, td);
> > }
> >
> > void thermal_zone_set_trip_temp(struct thermal_zone_device *tz,
> > struct thermal_trip *trip, int temp)
> > {
> > - if (trip->temperature == temp)
> > + struct thermal_trip_desc *td = trip_to_trip_desc(trip);
> > + int old_temp = trip->temperature;
> > +
> > + if (old_temp == temp)
> > return;
> >
> > WRITE_ONCE(trip->temperature, temp);
> > thermal_notify_tz_trip_change(tz, trip);
> >
> > - if (temp == THERMAL_TEMP_INVALID) {
> > - struct thermal_trip_desc *td = trip_to_trip_desc(trip);
> > + if (old_temp == THERMAL_TEMP_INVALID) {
> > + /*
> > + * The trip was invalid before the change, so move it to the
> > + * trips_high list regardless of the new temperature value
> > + * because there is no mitigation under way for it. If a
> > + * mitigation needs to be started, the trip will be moved to the
> > + * trips_reached list later.
> > + */
>
> Good comment
>
> > + move_to_trips_high(tz, td);
> > + return;
> > + }
> >
> > + if (temp == THERMAL_TEMP_INVALID) {
> > /*
> > - * If the trip has been crossed on the way up, some adjustments
> > - * are needed to compensate for the lack of it going forward.
> > + * If the trip is in the trips_reached list, mitigation is under
> > + * way for it and it needs to be stopped because the trip is
> > + * effectively going away.
> > */
> > if (tz->temperature >= td->threshold)
> > thermal_trip_crossed(tz, td, thermal_get_tz_governor(tz), false);
> >
> > + move_to_trips_invalid(tz, td);
> > + return;
> > + }
> > +
> > + /*
> > + * The trip stays on its current list, but its threshold needs to be
> > + * updated due to the temperature change and the list ordering may need
> > + * to be restored.
> > + */
> > + if (tz->temperature >= td->threshold)
> > + move_to_trips_reached(tz, td);
> > + else
> > + move_to_trips_high(tz, td);
> > +}
> > +EXPORT_SYMBOL_GPL(thermal_zone_set_trip_temp);
> > +
> > +static void thermal_zone_handle_trips(struct thermal_zone_device *tz,
> > + struct thermal_governor *governor,
> > + int *low, int *high)
> > +{
> > + struct thermal_trip_desc *td, *next;
> > + LIST_HEAD(way_down_list);
> > +
> > + /* Check the trips that were below or at the zone temperature. */
> > + list_for_each_entry_safe_reverse(td, next, &tz->trips_reached, list_node) {
> > + if (td->threshold <= tz->temperature)
> > + break;
> > +
> > + thermal_trip_crossed(tz, td, governor, false);
> > /*
> > - * Invalidate the threshold to avoid triggering a spurious
> > - * trip crossing notification when the trip becomes valid.
> > + * The current trips_high list needs to be processed before
> > + * adding new entries to it, so put them on a temporary list.
> > */
> > - td->threshold = INT_MAX;
> > + list_move(&td->list_node, &way_down_list);
> > + }
> > + /* Check the trips that were previously above the zone temperature. */
> > + list_for_each_entry_safe(td, next, &tz->trips_high, list_node) {
> > + if (td->threshold > tz->temperature)
> > + break;
> > +
> > + thermal_trip_crossed(tz, td, governor, true);
> > + move_to_trips_reached(tz, td);
> > + }
>
> nit:
> I would add and extra empty line here, since this is
> quite separate from the loop above and has important
> meaning.
It is though closely related to the first loop.
I'll add a one-line comment here when applying the patch.
> > + list_for_each_entry_safe(td, next, &way_down_list, list_node)
> > + move_to_trips_high(tz, td);
> > +
> > + if (!list_empty(&tz->trips_reached)) {
> > + td = list_last_entry(&tz->trips_reached,
> > + struct thermal_trip_desc, list_node);
> > + /*
> > + * Set the "low" value below the current trip threshold in case
> > + * the zone temperature is at that threshold and stays there,
> > + * which would trigger a new interrupt immediately in vain.
> > + */
> > + *low = td->threshold - 1;
> > + }
> > + if (!list_empty(&tz->trips_high)) {
> > + td = list_first_entry(&tz->trips_high,
> > + struct thermal_trip_desc, list_node);
> > + *high = td->threshold;
> > }
> > }
> > -EXPORT_SYMBOL_GPL(thermal_zone_set_trip_temp);
> >
> > void __thermal_zone_device_update(struct thermal_zone_device *tz,
> > enum thermal_notify_event event)
> > {
> > struct thermal_governor *governor = thermal_get_tz_governor(tz);
> > - struct thermal_trip_desc *td, *next;
> > - LIST_HEAD(way_down_list);
> > - LIST_HEAD(way_up_list);
> > int low = -INT_MAX, high = INT_MAX;
> > int temp, ret;
> >
> > @@ -614,25 +678,7 @@ void __thermal_zone_device_update(struct
> >
> > tz->notify_event = event;
> >
> > - for_each_trip_desc(tz, td) {
> > - handle_thermal_trip(tz, td, &way_up_list, &way_down_list);
> > -
> > - if (td->threshold <= tz->temperature && td->threshold > low)
> > - low = td->threshold;
> > -
> > - if (td->threshold >= tz->temperature && td->threshold < high)
> > - high = td->threshold;
> > - }
> > -
> > - list_for_each_entry_safe(td, next, &way_up_list, list_node) {
> > - thermal_trip_crossed(tz, td, governor, true);
> > - list_del_init(&td->list_node);
> > - }
> > -
> > - list_for_each_entry_safe_reverse(td, next, &way_down_list, list_node) {
> > - thermal_trip_crossed(tz, td, governor, false);
> > - list_del_init(&td->list_node);
> > - }
> > + thermal_zone_handle_trips(tz, governor, &low, &high);
> >
> > thermal_thresholds_handle(tz, &low, &high);
> >
> > @@ -1507,6 +1553,9 @@ thermal_zone_device_register_with_trips(
> > }
> >
> > INIT_LIST_HEAD(&tz->node);
> > + INIT_LIST_HEAD(&tz->trips_high);
> > + INIT_LIST_HEAD(&tz->trips_reached);
> > + INIT_LIST_HEAD(&tz->trips_invalid);
> > ida_init(&tz->ida);
> > mutex_init(&tz->lock);
> > init_completion(&tz->removal);
> > @@ -1536,7 +1585,7 @@ thermal_zone_device_register_with_trips(
> > * this only matters for the trips that start as invalid and
> > * become valid later.
> > */
> > - td->threshold = INT_MAX;
> > + move_to_trips_invalid(tz, td);
> > }
> >
> > tz->polling_delay_jiffies = msecs_to_jiffies(polling_delay);
> >
> >
> >
>
> It took me a while to get the new code flow and idea.
>
> Anyway, not that much controversial changes, which might
> look like, on the first glance, LGTM.
>
> Reviewed-by: Lukasz Luba <lukasz.luba@....com>
Thank you and thanks for all of the reviews!
Powered by blists - more mailing lists