The recent introduction of the schedtune cgroup controller has provided
the scheduler with added flexibility in terms of some of it's placement
features. In particular each cgroup under the schedtune controller can
now specify:
1) Whether it needs co-location along with other cgroups
2) Whether it is eligible for scheduler boost (sched_boost_enabled)
3) Whether the kernel can override the boost eligibility when necessary
(sched_boost_no_override)
The scheduler now creates a reserved co-location group at boot. This
group is used to co-locate all tasks that form part of any one of the
cgroups that have co-location enabled. This reserved group can neither
be destroyed nor reused for other purposes. Furthermore, cgroups are
only allowed to indicate their co-location preference once at boot.
Further updates are disallowed.
Since we are now creating co-location groups for an extended period of
time, there are a few other factors to consider when determining the
preferred cluster for the group. We first exclude any tasks in the
group that have not been observed to be running for a significant
amount of time. Secondly we introduce the notion of group up and down
migrate tunables to allow different migration policies than individual
tasks. Lastly we break co-location if a single task in a group exceeds
up-migrate but the total load of the group does not exceed group
up-migrate.
In terms of sched_boost, the scheduler now supports multiple types of
boost. These are:
1) FULL_THROTTLE : Force up-migrate tasks belonging any cgroup that
has the sched_boost_enabled flag turned on. Little
CPUs will only be used when big CPUs can no longer
accommodate tasks. Also up-migrate all RT tasks.
2) CONSERVATIVE : Override the sched_boost_enabled flag for all cgroups
except those that have the sched_boost_no_override
flag set. Force up-migrate all tasks belonging to only
those cgroups that still remain eligible for boost.
RT tasks do not get force up migrated.
3) RESTRAINED : Start frequency aggregation for co-located tasks. This
type of boost does not force up-migrate any task.
Finally the boost API removes ref-counting. This means that there can
only be a single entity using boost at any given time. If multiple
entities are managing boost, they are required to be well behaved so
that they don't interfere with one another. Even for a single client,
it is not possible to switch directly from one boost type to another.
Boost must be first turned off before switching over to a new type.
Change-Id: I8d224a70cbef162f27078b62b73acaa22670861d
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>
Add support for printing the voltage voting info to the
clock_state ftrace events.
CRs-Fixed: 1082843
Change-Id: I6ab3992958a659995b7d5020287fd6e47e28f2a4
Signed-off-by: Deepak Katragadda <dkatraga@codeaurora.org>
More than one client may call the core_ctl_set_boost api. Add support
for this.
Also add a new trace event that is emitted when this api is called.
Change-Id: Iad0a9fc45f1ce87433995e8e549bfca80e8b9cb2
Signed-off-by: Olav Haugan <ohaugan@codeaurora.org>
The previous patches in this rewrite of scheduler guided frequency
selection reintroduces the part-picture problem that we addressed in
our initial implementation. In that, when tasks migrate across CPUs
within a cluster, we end up losing the complete picture of the
sequential nature of the workload.
This patch aims to solve that problem slightly differently. We track
the top task on every CPU within a window. Top task is defined as the
task that runs the most in a given window. This enhances our ability
to detect the sequential nature of workloads. A single migrating task
executing for an entire window will cause 100% load to be reported
for frequency guidance instead of the maximum footprint left on any
individual CPU in the task's trail. There are cases, that this new
approach does not address. Namely, cases where the sum of two or more
tasks accurately reflects the true sequential nature of the workload.
Future optimizations might aim to tackle that problem.
To track top tasks, we first realize that there is no strict need to
maintain the task struct itself as long as we know the load exerted by
the top task. We also realize that to maintain top tasks on every CPU
we have to track the execution of every single task that runs during
the window. The load associated with a task needs to be migrated when
the task migrates from one CPU to another. When the top task migrates
away, we need to locate the second top task and so on.
Given the above realizations, we use hashmaps to track top task load
both for the current and the previous window. This hashmap is
implemented as an array of fixed size. The key of the hashmap is given
by task_execution_time_in_a_window / array_size. The size of the array
(number of buckets in the hashmap) dictate the load granularity of each
bucket. The value stored in each bucket is a refcount of all the tasks
that executed long enough to be in that bucket.
This approach has a few benefits. Firstly, any top task stats update
now take O(1) time. While task migration is also O(1), it does still
involve going through up to the size of the array to find the second
top task. Further patches will aim to optimize this behavior. Secondly,
and more importantly, not having to store the task struct itself saves
a lot of memory usage in that 1) there is no need to retrieve task
structs later causing cache misses and 2) we don't have to unnecessarily
hold up task memory for up to 2 full windows by calling get_task_struct()
after a task exits.
Change-Id: I004dba474f41590db7d3f40d9deafe86e71359ac
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>
Keeping a track of the load footprint of each task on every CPU
that it executed on gives the scheduler much more flexibility in
terms of the number of frequency guidance policies. These new fields
will be used in subsequent patches as we alter the load fixup
mechanism upon task migration. We still need to maintain the
curr/prev_window sums as they will also be required in subsequent
patches as we start to track top tasks based on cumulative load.
Also, we need to call init_new_task_load() for the idle task. This
is an existing harmless bug as load tracking for the idle task is
irrelevant. However, in this patch we are adding pointers to the
ravg structure. These pointers have to be initialized even for the
idle task.
Finally move init_new_task_load() to sched_fork(). This was always
the more appropriate place, however, following the introduction of
new pointers in the ravg struct, this is necessary to avoid races
with functions such as reset_all_task_stats().
Change-Id: Ib584372eb539706da4319973314e54dae04e5934
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>
Consider recent history (residencies) of the cluster and core
low power modes while the cluster level low power mode to enter
is selected.
Change-Id: Ifdc847a71136f55aded8a758b92179bb9aebfdcb
Signed-off-by: Srinivas Rao L <lsrao@codeaurora.org>
Consider recent history (residencies) of the low power modes per
core while the next low power mode to enter is selected. If most
of the history says the pattern of residencies is repeating with
minimal deviation then use the average of these for predicting
the next mode to enter.
If the pattern is not repeating then if more than 50 percent of
the samples out of history have exited a low power mode earlier
than the minumim residency of that mode, restrict it and also low
power modes deeper than that.
In any of the above case, trigger a hrtimer to wakeup cpu with
timeout as predicted+delta or max residency of the mode selected
if a deeper state can be selected after waking up incase if
prediction goes wrong.
Change-Id: I902a06939e19ac51dfd8c2db6b727b203ebfda14
Signed-off-by: Srinivas Rao L <lsrao@codeaurora.org>
Move the core control trace events to scheduler trace event file.
Change-Id: I65943d8e4a9eac1f9f5a40ad5aaf166679215f48
Signed-off-by: Olav Haugan <ohaugan@codeaurora.org>
Add tracepoint to capture the cpu isolation event including KPI for
time it took to isolate.
Change-Id: If2d30000f068afc50db953940f4636ef6a089b24
Signed-off-by: Olav Haugan <ohaugan@codeaurora.org>
* tmp-bab1564:
ANDROID: mmc: Add CONFIG_MMC_SIMULATE_MAX_SPEED
android: base-cfg: Add CONFIG_INET_DIAG_DESTROY
cpufreq: interactive: only apply interactive boost when enabled
cpufreq: interactive: fix policy locking
ANDROID: dm verity fec: add sysfs attribute fec/corrected
ANDROID: android: base-cfg: enable CONFIG_DM_VERITY_FEC
UPSTREAM: dm verity: add ignore_zero_blocks feature
UPSTREAM: dm verity: add support for forward error correction
UPSTREAM: dm verity: factor out verity_for_bv_block()
UPSTREAM: dm verity: factor out structures and functions useful to separate object
UPSTREAM: dm verity: move dm-verity.c to dm-verity-target.c
UPSTREAM: dm verity: separate function for parsing opt args
UPSTREAM: dm verity: clean up duplicate hashing code
UPSTREAM: dm: don't save and restore bi_private
mm: Export do_munmap
sdcardfs: remove unneeded __init and __exit
sdcardfs: Remove unused code
fs: Export d_absolute_path
sdcardfs: remove effectless config option
inotify: Fix erroneous update of bit count
fs: sdcardfs: Declare LOOKUP_CASE_INSENSITIVE unconditionally
trace: cpufreq: fix typo in min/max cpufreq
sdcardfs: Add support for d_canonical_path
vfs: add d_canonical_path for stacked filesystem support
sdcardfs: Bring up to date with Android M permissions:
Changed type-casting in packagelist management
Port of sdcardfs to 4.4
Included sdcardfs source code for kernel 3.0
ANDROID: usb: gadget: Add support for MTP OS desc
CHROMIUM: usb: gadget: f_accessory: add .raw_request callback
CHROMIUM: usb: gadget: audio_source: add .free_func callback
CHROMIUM: usb: gadget: f_mtp: fix usb_ss_ep_comp_descriptor
CHROMIUM: usb: gadget: f_mtp: Add SuperSpeed support
FROMLIST: mmc: block: fix ABI regression of mmc_blk_ioctl
FROMLIST: mm: ASLR: use get_random_long()
FROMLIST: drivers: char: random: add get_random_long()
FROMLIST: pstore-ram: fix NULL reference when used with pdata
usb: u_ether: Add missing rx_work init
ANDROID: dm-crypt: run in a WQ_HIGHPRI workqueue
misc: uid_stat: Include linux/atomic.h instead of asm/atomic.h
hid-sensor-hub.c: fix wrong do_div() usage
power: Provide dummy log_suspend_abort_reason() if SUSPEND is disabled
PM / suspend: Add dependency on RTC_LIB
drivers: power: use 'current' instead of 'get_current()'
video: adf: Set ADF_MEMBLOCK to boolean
video: adf: Fix modular build
net: ppp: Fix modular build for PPPOLAC and PPPOPNS
net: pppolac/pppopns: Replace msg.msg_iov with iov_iter_kvec()
ANDROID: mmc: sdio: Disable retuning in sdio_reset_comm()
ANDROID: mmc: Move tracepoint creation and export symbols
ANDROID: kernel/watchdog: fix unused variable warning
ANDROID: usb: gadget: f_mtp: don't use le16 for u8 field
ANDROID: lowmemorykiller: fix declaration order warnings
ANDROID: net: fix 'const' warnings
net: diag: support v4mapped sockets in inet_diag_find_one_icsk()
net: tcp: deal with listen sockets properly in tcp_abort.
tcp: diag: add support for request sockets to tcp_abort()
net: diag: Support destroying TCP sockets.
net: diag: Support SOCK_DESTROY for inet sockets.
net: diag: Add the ability to destroy a socket.
net: diag: split inet_diag_dump_one_icsk into two
Revert "mmc: Extend wakelock if bus is dead"
Revert "mmc: core: Hold a wake lock accross delayed work + mmc rescan"
ANDROID: mmc: move to a SCHED_FIFO thread
Conflicts:
drivers/cpufreq/cpufreq_interactive.c
drivers/misc/uid_stat.c
drivers/mmc/card/block.c
drivers/mmc/card/queue.c
drivers/mmc/card/queue.h
drivers/mmc/core/core.c
drivers/mmc/core/sdio.c
drivers/staging/android/lowmemorykiller.c
drivers/usb/gadget/function/f_mtp.c
kernel/watchdog.c
Signed-off-by: Runmin Wang <runminw@codeaurora.org>
Change-Id: Ibb4db11c57395f67dee86211a110c462e6181552
The structures being moved around are only used for trace events
defined under CONFIG_SCHED_HMP. Move code to hmp.c to reflect
the same.
Change-Id: Ib959355264405ab779b24948f111a2ca61d367de
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>
Adding source and destination port number info
in the gso trace events to differentiate between
the flows.
CRs-Fixed: 982871
Change-Id: Idbae7f95dfd56293805b58e3c6626f5f6e07d08a
Acked-by: Ashwanth Goli <ashwanth@qti.qualcomm.com>
Signed-off-by: Ravinder Konka <rkonka@codeaurora.org>
[subashab@codeaurora.org: resolve trivial merge conflicts]
Signed-off-by: Subash Abhinov Kasiviswanathan <subashab@codeaurora.org>
Add a flag to the trace event that indicates why we picked a particular
CPU. This is very useful information/statistic that can be used to
analyse the effectiveness of the scheduler.
Change-Id: Ic9462fef751f9442ae504c09fbf4418e08f018b0
Signed-off-by: Olav Haugan <ohaugan@codeaurora.org>
CONFIG_SCHED_FREQ_INPUT was created to keep parts of the scheduler
dealing with frequency separate from other parts of the scheduler
that deal with task placement. However, overtime the two features
have become intricately linked whereby SCHED_FREQ_INPUT cannot be
turned on without having SCHED_HMP turned on as well. Given this
complex inter-dependency and the fact that all old, existing and
future targets use both config options, remove this unnecessary
feature separation. It will aid in making kernel upgrades a lot
simpler and faster.
Change-Id: Ia20e40d8a088d50909cc28f5be758fa3e9a4af6f
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>
Add trace events to control aborting CoreSight trace
dynamically based on module parameter.
Coresight driver will dump any trace present in the current sink
in case we hit a kernel panic, user fault or an undefined instruction.
Change-Id: Iee1ccf5cbd7b767753a3115c0570e63fbe2aa8f3
Signed-off-by: Satyajit Desai <sadesai@codeaurora.org>
With this patch, anon pages of inactive tasks can be reclaimed,
depending on memory pressure. Memory pressure is detected
using vmpressure events. 'N' best tasks in terms of anon
size is selected and pages proportional to their tasksize
is reclaimed. The total number of pages reclaimed at each
run of the swap work, can be tuned from userspace, the
default being SWAP_CLUSTER_MAX * 32.
The patch also adds tracepoints to debug and tune the
feature.
echo 1 > /sys/module/process_reclaim/parameters/enable_process_reclaim
to enable the feature.
echo <pages> > /sys/module/process_reclaim/parameters/per_swap_size,
to set the number of pages reclaimed in each scan.
/sys/module/process_reclaim/parameters/reclaim_avg_efficiency, provides
the average efficiency (scan to reclaim ratio) of the algorithm.
/sys/module/process_reclaim/parameters/swap_eff_win, to set the window
period (in unit of number of times reclaim is triggered) to detect
low efficiency runs.
/sys/module/process_reclaim/parameters/swap_opt_eff, to set the optimal
efficiency threshold for low efficiency detection.
Change-Id: I895986f10c997d1715761eaaadc4bbbee60db9d2
Signed-off-by: Vinayak Menon <vinmenon@codeaurora.org>
The function parameter cpu isn't used anymore by cpu_cycles_to_freq().
So remove it.
Change-Id: Ide19321206dacb88fedca97e1b689d740f872866
Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>
LMH DCVSh driver receives interrupt from hardware whenever there is a
new mitigation frequency decision is made in hardware.
Add ftrace event to print the hardware mitigation frequency value from
driver.
Change-Id: Ib357ee3c3a461613bfd1268ec8f98973c2982c10
Signed-off-by: Ram Chandrasekar <rkumbako@codeaurora.org>
This merge brings MMC/SD card changes from 'dev/msm-4.4-mmc_port' to
msm-4.4.
* origin/dev/msm-4.4-mmc_port: (472 commits)
mmc: sdhci-msm: fix few compilation issues
mmc: cmdq_hci: fix compilation issue
mmc: host: fix compilation issue when clk_gating config is disabled
mmc: sdhci: clean up legacy adma related variables
mmc: sdhci: enable 64-bit DMA support only if chipset supports 64-bit
mmc: sdhci: Replace SDHCI_USE_ADMA_64BIT flag
mmc: auto bkops fixes
mmc: card: fix quirk bit map
Revert "mmc: core: get drive types supported by eMMC cards"
mmc: host: sdhci: don't queue zero length descriptor
mmc: core: fix deadlock between runtime-suspend and devfreq
mmc: block: Add quirk and increase read data timeout for hynix emmc
mmc: card: Fix broken clock gating
mmc: core: postpone runtime suspend in case BKOPS is required
mmc: core: update AUTO_EN in BKOPS_EN field on runtime resume
mmc: revert runtime idle state
mmc: host: Set max frequency when disabling clock scaling
mmc: queue: Fix queue_lock spinlock bug from CMDQ shutdown path
mmc: core: fix issue with devfreq clock scaling
mmc: core: set REL_WR_SEC_C register to 0x1 per eMMC5.0 spec
...
Change-Id: I702a72fbbecba520f429bf1149106e684335e2a5
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Instrument the sdhci-msm platform driver with tracepoints to aid in
debugging issues and identifying latencies in the following
paths:
* System suspend/resume
* Runtime suspend/resume
Change-Id: I4fed1c2ccba7d5d7f978f161e7985c98e869d1d8
Signed-off-by: Konstantin Dorfman <kdorfman@codeaurora.org>
Instrument the mmc core layer with tracepoints to aid in
debugging issues and identifying latencies in the following
paths:
* System suspend/resume
* Runtime suspend/resume
Change-Id: I1e0fa7d3f8b54c102b4055f910b58a42412748da
Signed-off-by: Konstantin Dorfman <kdorfman@codeaurora.org>
[subhashj@codeaurora.org: fixed trivial merge conflicts]
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Use kernel's ftrace support to capture MMC clock frequency
transitions which can be useful for debugging issues related
to power consumption.
Usage:
mount -t debugfs none /sys/kernel/debug
echo 1 > /sys/kernel/debug/tracing/events/mmc/mmc_clk/enable
cat /sys/kernel/debug/tracing/trace_pipe
Change-Id: I25c4ee39dcbe30e7665902a9f723a5a421b55ca3
Signed-off-by: Sujit Reddy Thumma <sthumma@codeaurora.org>
Instrument the sdhci driver with tracepoints to aid in
debugging issues and identifying latencies in the following
path:
* CMD completion
* DATA completion
* DMA preparation
* Post DMA cleanup
Change-Id: Ie8cd0c2fb6c1bd6ab13883123be021081f8b8f78
Signed-off-by: Venkat Gopalakrishnan <venkatg@codeaurora.org>
[subhashj@codeaurora.org: fixed minor merge conflict]
Signed-off-by: Subhash Jadavani <subhashj@codeaurora.org>
Related threads in a group could execute on different CPUs and hence
present a split-demand picture to cpufreq governor. IOW the governor
fails to see the net cpu demand of all related threads in a given
window if the threads's execution were to be split across CPUs. That
could result in sub-optimal frequency chosen in comparison to the
ideal frequency at which the aggregate work (taken up by related
threads) needs to be run.
This patch aggregates cpu execution stats in a window for all related
threads in a group. This helps present cpu busy time to governor as if
all related threads were part of the same thread and thus help select
the right frequency required by related threads. This aggregation
is done per-cluster.
Change-Id: I71e6047620066323721c6d542034ddd4b2950e7f
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>
[joonwoop@codeaurora.org: Fixed notify_migration() to hold rcu read
lock as this version of Linux doesn't hold p->pi_lock when the
function gets called while keeping use of rcu_access_pointer() since
we never dereference return value.]
Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>
Dup ftrace event traffic and writes to trace_marker file from
userspace to STM. Also dup trace printk traffic to STM. This
allows Linux tracing and log data to be correlated with other
data transported over STM.
Change-Id: I4fcb42f2e97ab963fdc85853f4f3ea1f208bfc3c
Signed-off-by: Pratik Patel <pratikp@codeaurora.org>
[spjoshi@codeaurora.org: 3.18 code fixup]
Signed-off-by: Sarangdhar Joshi <spjoshi@codeaurora.org>
[mittals@codeaurora.org: 4.4 code fixup]
Signed-off-by: Shashank Mittal <mittals@codeaurora.org>
At present scheduler calculates task's demand with the task's execution
time weighted over CPU frequency. The CPU frequency is given by
governor's CPU frequency transition notification. Such notification
may not be available.
Provide an API for CPU clock driver to register callback functions so
in order for scheduler to access CPU's cycle counter to estimate CPU's
frequency without notification. At time point scheduler assumes the
cycle counter increases always even when cluster is idle which might
not be true. This will be fixed by subsequent change for more accurate
I/O wait time accounting.
CRs-fixed: 1006303
Change-Id: I93b187efd7bc225db80da0184683694f5ab99738
Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>
There were issues reported, where page cache thrashing was
observed because of LMK not killing tasks when required,
resulting in sluggishness and higher app launch latency.
LMK does not kill a task for the following reasons.
1. The free and file pages are above the LMK thresholds
2. LMK tries to pick task with an adj level corresponding
to current thresholds, but fails to do so because of the
absence of tasks in that level.
But sometimes it is better to kill a lower adj task, than thrashing.
And there are cases where the number of file pages are huge, though
we dont thrash, the reclaim process becomes time consuming, since
LMK triggers will be delayed because of higher number of file
pages. Even in such cases, when reclaim path finds it difficult
to reclaim pages, it is better to trigger lmk to free up some memory
faster.
The basic idea here is to make LMK more aggressive dynamically
when such a thrashing scenario is detected.
To detect thrashing, this patch uses vmpressure events.
The values of vmpressure upon which an action has to be taken,
was derived empirically.
This patch also adds tracepoints to validate this feature,
almk_shrink and almk_vmpressure.
Two knobs are available for the user to tune adaptive lmk
behaviour.
/sys/module/lowmemorykiller/parameters/adaptive_lmk - Write
1 to enable the feature, 0 to disable. By default disabled.
/sys/module/lowmemorykiller/parameters/vmpressure_file_min -
This parameter controls the behaviour of LMK when vmpressure
is in the range of 90-94. Adaptive lmk triggers based on number file
pages wrt vmpressure_file_min, when vmpressure is in the range of
90-94. Usually this is a pseudo minfree value, higher than the
highest configured value in minfree array.
Change-Id: I1a08160c35d3e33bdfd1d2c789c288fc07d0f0d3
Signed-off-by: Vinayak Menon <vinmenon@codeaurora.org>
With modification in scheduler, governor now gets predicted
instantaneous demand waiting to run in addition to demand from
previous window for each CPU. Make use of this information since
prediction from scheduler could be more accurate than just looking at
past few windows.
Governor calculates two frequencies during each sampling period: one based
on demand in previous sampling period (f_prev), and the other based on
prediction provided by scheduler (f_pred). Max of both will be selected
as final frequency. Hispeed related logic, including both frequency
selection and delay is ignored when prediction is enabled. If only
f_pred but not f_prev picked policy->max, max_freq_hysteresis period is
not started/extended. This is to reduce power cost of mis-prediction
if it happens.
One use case prediction could dramatically help is when a heavy task
wakes up after sleeping for a long time. With prediction, governor
could ramp up to frequency the task needs much faster than before.
To enable prediction, echo 1 to enable_prediction file in
cpufreq interactive sysfs directory.
Change-Id: I27396785886e43ea01c9000c651c8bd142172273
Suggested-by: Saravana Kannan <skannan@codeaurora.org>
Signed-off-by: Junjie Wu <junjiew@codeaurora.org>
Current window based load tracking only saves history for five
windows. A historically heavy task's heavy load will be completely
forgotten after five windows of light load. Even before the five
window expires, a heavy task wakes up on same CPU it used to run won't
trigger any frequency change until end of the window. It would starve
for the entire window. It also adds one "small" load window to
history because it's accumulating load at a low frequency, further
reducing the tracked load for this heavy task.
Ideally, scheduler should be able to identify such tasks and notify
governor to increase frequency immediately after it wakes up.
Add a histogram for each task to track a much longer load history. A
prediction will be made based on runtime of previous or current
window, histogram data and load tracked in recent windows. Prediction
of all tasks that is currently running or runnable on a CPU is
aggregated and reported to CPUFreq governor in sched_get_cpus_busy().
sched_get_cpus_busy() now returns predicted busy time in addition
to previous window busy time and new task busy time, scaled to
the CPU maximum possible frequency.
Tunables:
- /proc/sys/kernel/sched_gov_alert_freq (KHz)
This tunable can be used to further filter the notifications.
Frequency alert notification is sent only when the predicted
load exceeds previous window load by sched_gov_alert_freq converted to
load.
Change-Id: If29098cd2c5499163ceaff18668639db76ee8504
Suggested-by: Saravana Kannan <skannan@codeaurora.org>
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>
Signed-off-by: Junjie Wu <junjiew@codeaurora.org>
[joonwoop@codeaurora.org: fixed merge conflicts around __migrate_task()
and removed changes for CONFIG_SCHED_QHMP.]
Provide userspace interface for tasks to be grouped together as
"related" threads. For example, all threads involved in updating
display buffer could be tagged as related.
Scheduler will attempt to provide special treatment for group of
related threads such as:
1) Colocation of related threads in same "preferred" cluster
2) Aggregation of demand towards determination of cluster frequency
This patch extends scheduler to provide best-effort colocation support
for a group of related threads.
Change-Id: Ic2cd769faf5da4d03a8f3cb0ada6224d0101a5f5
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>
[joonwoop@codeaurora.org: fixed minor merge conflicts. removed ifdefry
for CONFIG_SCHED_QHMP.]
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>
Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>
A cluster is set of CPUs sharing some power controls and an L2 cache.
This patch buids a list of clusters at bootup which are sorted by
their max_power_cost. Many cluster-shared attributes like cur_freq,
max_freq etc are needlessly maintained in per-cpu 'struct rq' currently.
Consolidate them in a cluster structure.
Change-Id: I0567672ad5fb67d211d9336181ceb53b9f6023af
Signed-off-by: Srivatsa Vaddagiri <vatsa@codeaurora.org>
Signed-off-by: Joonwoo Park <joonwoop@codeaurora.org>
[joonwoop@codeaurora.org: fixed minor conflict in
arch/arm64/kernel/topology.c. fixed conflict due to ommited changes for
CONFIG_SCHED_QHMP.]
Signed-off-by: Syed Rameez Mustafa <rameezmustafa@codeaurora.org>
Add snapshot for Video driver source for MSM targets. The code is
migrated from msm-3.18 kernel at the below commit level -
d5809484bb1bf5864dad2f081b0145224762963a.
Signed-off-by: Arun Menon <avmenon@codeaurora.org>
Detect Perf cluster peak loads near FMAX based on the trigger
thresholds set. On meeting the peak load criteria, the userspace
is notified to take action by applying parameters to enhance performance.
CRs-Fixed: 969499
Change-Id: Ie9687bf1aa832434dc61d20056f91a096d7be4f0
Signed-off-by: Vijay Ganti <viganti@codeaurora.org>
Add support for enter/exit cycle sysfs nodes for io detection
There are some usecases which may benefit from different enter/exit
cycle load criteria for IO load. This change adds support for
that.
Change-Id: Iff135ed11b92becc374ace4578e0efc212d2b731
Signed-off-by: Tapas Kumar Kundu <tkundu@codeaurora.org>
Add support for multi_enter_cycles/multi_exit_cycles per cluster
There are some usecases which may benefit from different enter/exit
cycle load criteria for multimode cpu load. This change adds support for
that.
Change-Id: I3408405307ca03b9bba3f03e216ef59b98f29832
Signed-off-by: Tapas Kumar Kundu <tkundu@codeaurora.org>
Certain governors may stop sending out notifications once CPUs enter
idle at min frequency.If governor's notifications stop then single mode
will not exit for long time. It can happen only if the exit conditions are
set in such a way that the time taken to exit single mode exceeds the time
for the governor to ramp down, idle out and hence stop sending
notifications leaving the system in single mode indefinitely.
This change adds separate enter/exit cycle sysfs nodes along with a per
cluster non-deferrable timer for single mode exit. The timer is armed only
when the load starts falling below the exit load threshold and is
cancelled when either the load starts going up or SINGLE mode is exited
due to exceeding exit cycle count. On expiry the timer resets SINGLE mode
and the enter/exit cycle counts.
Change-Id: I13552b2f4085c435b917833a2993f8c64ff4ed2f
Signed-off-by: Tapas Kumar Kundu <tkundu@codeaurora.org>
Detect single and multi threaded heavy workloads based on loads
received from interactive governor.
- If the max load across all the CPUs is greater than a
user-specified threshold for certain number of governor windows
then the load is detected as a single-threaded workload.
- If the total load across all the CPUs is greater than a
user-specified threshold for certain number of governor windows
then the load is detected as a multi-threaded workload.
If one of these is detected then a notification is sent to the
userspace so that an entity can read the nodes exposed to get an
idea of the nature of workload running.
Change-Id: Iba75d26fb3981886b3a8460d5f8999a632bbb73a
Signed-off-by: Rohit Gupta <rohgup@codeaurora.org>
Some workloads spend a lot of time in IO activity and need higher
performance from system resources (for eg. CPU/DDR frequencies)to
complete with decent performance. Unfortunately cpufreq governors and
other system resources crucial for IO are tuned for general usecases
and hence might be slower to react to such demanding IO workloads.
This patch adds functionality to detect IO workloads and then send
hints to userspace of the detected activity so that userspace can
take necessary tuning action to prepare the system for such activity.
IO activity is tracked every interactive governor timer boundary and
if the percentage of iowait time in each cycle exceeds certain
threshold continuously for certain number of cycles then heavy IO
activity is detected.
Change-Id: I73859517cb436e50340ef14739183e61fc62f90f
Signed-off-by: Rohit Gupta <rohgup@codeaurora.org>
Sometimes for power saving reasons we might want to keep fewer CPUs
online without adversely affecting performance for certain real world
usecases. This module helps to provide that hotplug support to the
userspace such that it tries to make a best effort in keeping a certain
number of CPUs online as specified by the userspace.
It allows any userspace entity to specify the CPUs that it wants to
manage with this module and of those, the number of CPUs that should be
kept online.
Change-Id: I82c6d6e998d3740ad6f8c67b47344ce87f328b8b
Signed-off-by: Rohit Gupta <rohgup@codeaurora.org>
Create a scm group to enable profiling time spent in a scm call.
This will help determine which scm call is spending how much time in
a higher execution level.
To enable "echo 1 > /sys/kernel/debug/tracing/events/scm/enable".
It is disabled by default. If enabled, traces can be found in
Ftrace logs.
Ftrace Output Example:
PROCESS CPU TIME SCM ID, X0, Number of args, args[0-2], X5, return values [0-2]
kworker/u8:4-329 [002] 128.201129: scm_call_start: func id=0x42000904 (args: 0x6, 0x2, 0x200000000, 0x65b8000000019, 0x142e0f000)
kworker/u8:4-329 [002] 128.201383: scm_call_end: ret: 0, 0, 0x4a07e00000001
kworker/u8:4-329 [002] 128.201464: scm_call_start: func id=0x42000904 (args: 0x6, 0x3, 0x1312d0000000000, 0x17900000000, 0x142e0f000)
kworker/u8:4-329 [002] 128.201542: scm_call_end: ret: 0x1bf03dddddd, 0x2f72656b726f776b, 0x343a32
CRs-Fixed: 969770
Change-Id: I4e5aaff796dbc9457c55fa529114dcb57780b7ec
Signed-off-by: Sanrio Alvares <salvares@codeaurora.org>
Add cma and migrate trace events to enable CMA allocation
performance to be measured via ftrace.
Change-Id: I1e471e9e21f1a14ce2ed167d8515ccb5f83eb88c
Signed-off-by: Liam Mark <lmark@codeaurora.org>
This patch adds trace events to help with debug for gso feature
by identifying the packets(and their lenghts) that are using
the segmentation offload feature.
Change-Id: Ibfe1194cc63e74c75047040b0c540713d539992e
Acked-by: Ashwanth Goli <ashwanth@qti.qualcomm.com>
Signed-off-by: Ravinder Konka <rkonka@codeaurora.org>
Use performance counters to detect the memory latency sensitivity
of CPU workloads and vote for higher DDR frequency if required.
Change-Id: Ie77a3523bc5713fc0315bd0abc3913f485a96e0e
Signed-off-by: Rohit Gupta <rohgup@codeaurora.org>
Suggested-by: Saravana Kannan <skannan@codeaurora.org>
[junjiew@codeaurora.org: dropped changes in arch/arm64/Kconfig]
Signed-off-by: Junjie Wu <junjiew@codeaurora.org>
