Change 10f7dc075b87361d42d0145b74bfd75f28cc4545 ("sched/fair: dont run
energy calculation unless necessary") then assumed that the cpus
returned from find_best_target() would be used for energy evaluation.
Later, code was added to use the target cpu returned by fbt() in certain
latency sensitive situations. This lead to skipping prev_cpu, since it
could have been the target cpu had the above mentioned change not
dropped it.
So revert 10f7dc075b87361d42d0145b74bfd75f28cc4545 ("sched/fair: dont run
energy calculation unless necessary")
Change-Id: I06827ce9339f698dd3927a103b4c1a1ddbe123e2
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
This change is for general scheduler improvement.
Change-Id: Idb019614ff31b3cfcde023cd92f8d791f2650dee
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
The purpose of is_packing_eligible() test is to check if
we should throwaway the active cpu and use the idle CPU.
Refactor this function such a way that, the throw away
CPU (i.e active_cpu) is marked -1. The follow up patch
use this refactoring to add more checks in this function.
Change-Id: I0ae829dcc9b62da3ae69796b28abebd950b34462
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
[adharmap: made adjust cpus for packing inline]
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Extend sched_load_balance trace event to print overutilization
status of the current balancing sched domain.
Change-Id: Iae14e580294a0ab38394b4e8d660d0c5dd14327d
Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
This change is for general scheduler improvement.
Change-Id: I50d41aa3338803cbd45ff6314b2bb3978c59282b
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
It is quite a common occurrence that we have a misfit task (a task
running on a CPU which is close to the capacity limit of that CPU)
in the system AND we have recently done a load balance operation.
Normal load-balance rate limiting means we can leave those tasks on
the restrictive CPU for a few hundred ms before we move them.
Inspired by the work Leo Yan did on balancing misfits.
This patch reduces the balance interval to one jiffy when there are such
tasks running.
There appears to be no impact on interactive performance from this
patch.
Change-Id: I71d74734d3bd6bca80476c0f466ba48914b7ac02
Signed-off-by: Chris Redpath <chris.redpath@arm.com>
Git-commit: 84c88fc37932db51b87af8df3cedc621cb5c6e97
Git-repo: https://android.googlesource.com/kernel/common
[clingutla@codeaurora.org: Took only partial part from commit
'84c88fc37932("ANDROID: Combined EAS Load Balance Tweaks")',
and skip using 1 jiffy for busy balancing]
Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
* refs/heads/tmp-0b065cd:
Linux 4.19.33
platform: x86: intel_cht_int33fe: Remove the old connections for the muxes
usb: typec: class: Don't use port parent for getting mux handles
platform: x86: intel_cht_int33fe: Add connections for the USB Type-C port
platform: x86: intel_cht_int33fe: Add connection for the DP alt mode
platform: x86: intel_cht_int33fe: Register all connections at once
drivers: base: Helpers for adding device connection descriptions
bpf: do not restore dst_reg when cur_state is freed
staging: erofs: keep corrupted fs from crashing kernel in erofs_readdir()
staging: erofs: fix error handling when failed to read compresssed data
KVM: x86: Emulate MSR_IA32_ARCH_CAPABILITIES on AMD hosts
KVM: x86: update %rip after emulating IO
KVM: Reject device ioctls from processes other than the VM's creator
x86/smp: Enforce CONFIG_HOTPLUG_CPU when SMP=y
cpu/hotplug: Prevent crash when CPU bringup fails on CONFIG_HOTPLUG_CPU=n
watchdog: Respect watchdog cpumask on CPU hotplug
powerpc/64: Fix memcmp reading past the end of src/dest
powerpc/pseries/energy: Use OF accessor functions to read ibm,drc-indexes
objtool: Query pkg-config for libelf location
perf intel-pt: Fix TSC slip
perf pmu: Fix parser error for uncore event alias
mm/migrate.c: add missing flush_dcache_page for non-mapped page migrate
mm: mempolicy: make mbind() return -EIO when MPOL_MF_STRICT is specified
iommu/io-pgtable-arm-v7s: request DMA32 memory, and improve debugging
mm: add support for kmem caches in DMA32 zone
usb: cdc-acm: fix race during wakeup blocking TX traffic
xhci: Don't let USB3 ports stuck in polling state prevent suspend
usb: xhci: dbc: Don't free all memory with spinlock held
xhci: Fix port resume done detection for SS ports with LPM enabled
usb: host: xhci-rcar: Add XHCI_TRUST_TX_LENGTH quirk
usb: common: Consider only available nodes for dr_mode
USB: gadget: f_hid: fix deadlock in f_hidg_write()
usb: mtu3: fix EXTCON dependency
phy: sun4i-usb: Support set_mode to USB_HOST for non-OTG PHYs
gpio: adnp: Fix testing wrong value in adnp_gpio_direction_input
gpio: exar: add a check for the return value of ida_simple_get fails
drm/i915/gvt: Fix MI_FLUSH_DW parsing with correct index check
drm/vkms: fix use-after-free when drm_gem_handle_create() fails
drm/vgem: fix use-after-free when drm_gem_handle_create() fails
fs/proc/proc_sysctl.c: fix NULL pointer dereference in put_links
Disable kgdboc failed by echo space to /sys/module/kgdboc/parameters/kgdboc
USB: serial: option: add Olicard 600
USB: serial: option: add support for Quectel EM12
USB: serial: option: set driver_info for SIM5218 and compatibles
USB: serial: mos7720: fix mos_parport refcount imbalance on error path
USB: serial: ftdi_sio: add additional NovaTech products
USB: serial: cp210x: add new device id
serial: sh-sci: Fix setting SCSCR_TIE while transferring data
serial: mvebu-uart: Fix to avoid a potential NULL pointer dereference
serial: max310x: Fix to avoid potential NULL pointer dereference
staging: erofs: fix to handle error path of erofs_vmap()
staging: vt6655: Fix interrupt race condition on device start up.
staging: vt6655: Remove vif check from vnt_interrupt
staging: speakup_soft: Fix alternate speech with other synths
staging: comedi: ni_mio_common: Fix divide-by-zero for DIO cmdtest
tty: serial: qcom_geni_serial: Initialize baud in qcom_geni_console_setup
tty: atmel_serial: fix a potential NULL pointer dereference
tty: mxs-auart: fix a potential NULL pointer dereference
drm/rockchip: vop: reset scale mode when win is disabled
scsi: zfcp: fix scsi_eh host reset with port_forced ERP for non-NPIV FCP devices
scsi: zfcp: fix rport unblock if deleted SCSI devices on Scsi_Host
scsi: sd: Quiesce warning if device does not report optimal I/O size
scsi: sd: Fix a race between closing an sd device and sd I/O
ocfs2: fix inode bh swapping mixup in ocfs2_reflink_inodes_lock
fs/open.c: allow opening only regular files during execve()
kbuild: modversions: Fix relative CRC byte order interpretation
ALSA: hda/realtek - Fix speakers on Acer Predator Helios 500 Ryzen laptops
ALSA: hda/realtek: Enable headset MIC of ASUS X430UN and X512DK with ALC256
ALSA: hda/realtek: Enable headset mic of ASUS P5440FF with ALC256
ALSA: hda/realtek: Enable ASUS X441MB and X705FD headset MIC with ALC256
ALSA: hda/realtek - Add support for Acer Aspire E5-523G/ES1-432 headset mic
ALSA: hda/realtek: Enable headset MIC of Acer Aspire Z24-890 with ALC286
ALSA: hda/realtek: Enable headset MIC of Acer AIO with ALC286
ALSA: hda/realtek - Add support headset mode for New DELL WYSE NB
ALSA: hda/realtek - Add support headset mode for DELL WYSE AIO
ALSA: pcm: Don't suspend stream in unrecoverable PCM state
ALSA: pcm: Fix possible OOB access in PCM oss plugins
ALSA: seq: oss: Fix Spectre v1 vulnerability
ALSA: rawmidi: Fix potential Spectre v1 vulnerability
net: dsa: qca8k: remove leftover phy accessors
NFSv4.1 don't free interrupted slot on open
NFS: fix mount/umount race in nlmclnt.
vfio: ccw: only free cp on final interrupt
powerpc: bpf: Fix generation of load/store DW instructions
ARM: imx6q: cpuidle: fix bug that CPU might not wake up at expected time
Btrfs: fix assertion failure on fsync with NO_HOLES enabled
btrfs: Avoid possible qgroup_rsv_size overflow in btrfs_calculate_inode_block_rsv_size
btrfs: raid56: properly unmap parity page in finish_parity_scrub()
btrfs: don't report readahead errors and don't update statistics
btrfs: remove WARN_ON in log_dir_items
Btrfs: fix incorrect file size after shrinking truncate and fsync
powerpc/security: Fix spectre_v2 reporting
powerpc/fsl: Fix the flush of branch predictor.
powerpc/fsl: Fixed warning: orphan section `__btb_flush_fixup'
powerpc/fsl: Update Spectre v2 reporting
powerpc/fsl: Enable runtime patching if nospectre_v2 boot arg is used
powerpc/fsl: Flush branch predictor when entering KVM
powerpc/fsl: Flush the branch predictor at each kernel entry (32 bit)
powerpc/fsl: Flush the branch predictor at each kernel entry (64bit)
powerpc/fsl: Add nospectre_v2 command line argument
powerpc/fsl: Emulate SPRN_BUCSR register
powerpc/fsl: Add macro to flush the branch predictor
powerpc/fsl: Add infrastructure to fixup branch predictor flush
tun: add a missing rcu_read_unlock() in error path
thunderx: eliminate extra calls to put_page() for pages held for recycling
thunderx: enable page recycling for non-XDP case
net: sched: fix cleanup NULL pointer exception in act_mirr
ila: Fix rhashtable walker list corruption
vxlan: Don't call gro_cells_destroy() before device is unregistered
vrf: prevent adding upper devices
tun: properly test for IFF_UP
tipc: fix cancellation of topology subscriptions
tipc: change to check tipc_own_id to return in tipc_net_stop
tipc: allow service ranges to be connect()'ed on RDM/DGRAM
tcp: do not use ipv6 header for ipv4 flow
sctp: use memdup_user instead of vmemdup_user
sctp: get sctphdr by offset in sctp_compute_cksum
rhashtable: Still do rehash when we get EEXIST
packets: Always register packet sk in the same order
net-sysfs: call dev_hold if kobject_init_and_add success
net: stmmac: fix memory corruption with large MTUs
net: rose: fix a possible stack overflow
net: phy: meson-gxl: fix interrupt support
net/packet: Set __GFP_NOWARN upon allocation in alloc_pg_vec
net: datagram: fix unbounded loop in __skb_try_recv_datagram()
net: aquantia: fix rx checksum offload for UDP/TCP over IPv6
mISDN: hfcpci: Test both vendor & device ID for Digium HFC4S
mac8390: Fix mmio access size probe
ipv6: make ip6_create_rt_rcu return ip6_null_entry instead of NULL
gtp: change NET_UDP_TUNNEL dependency to select
genetlink: Fix a memory leak on error path
dccp: do not use ipv6 header for ipv4 flow
ipmi_si: Fix crash when using hard-coded device
Bluetooth: Verify that l2cap_get_conf_opt provides large enough buffer
Bluetooth: Check L2CAP option sizes returned from l2cap_get_conf_opt
UPSTREAM: virt_wifi: Remove REGULATORY_WIPHY_SELF_MANAGED
Change-Id: I681a654a3079e9b243bdc46d1e11d5abd6e9d774
Signed-off-by: Ivaylo Georgiev <irgeorgiev@codeaurora.org>
This change is for general scheduler improvement.
Change-Id: Ib963aef88d85e15fcd19cda3d3f0944b530239ab
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
This change is for general scheduler improvement.
Change-Id: I556b873cc46911953204792d60c60bf151345b1e
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
The system is killed when the memory allocation is failed for
the curr/prev window CPU arrays. Since these are small allocations
and they never fail except in scenarios like
(1) Testing with random memory failures induced
(2) When a fatal signal is pending for the task which is doing
allocations.
Instead of killing the system, use __GFP_NOFAIL flag.
Change-Id: If284814db5a6504b2f039053856ff30c2842808b
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
* refs/heads/tmp-6f994bf:
Revert "ANDROID: sched: Disable find_best_target() by default"
ANDROID: cpufreq: times: don't copy invalid freqs from freq table
UPSTREAM: filemap: add a comment about FAULT_FLAG_RETRY_NOWAIT behavior
BACKPORT: filemap: drop the mmap_sem for all blocking operations
BACKPORT: filemap: kill page_cache_read usage in filemap_fault
UPSTREAM: filemap: pass vm_fault to the mmap ra helpers
ANDROID: binder: remove extra declaration left after backport
FROMGIT: binder: fix BUG_ON found by selinux-testsuite
ANDROID: sched: Disable find_best_target() by default
ANDROID: sched/fair: Make the EAS wake-up prefer-idle aware
Linux 4.19.32
power: supply: charger-manager: Fix incorrect return value
ALSA: hda - Enforces runtime_resume after S3 and S4 for each codec
ALSA: hda - Record the current power state before suspend/resume calls
locking/lockdep: Add debug_locks check in __lock_downgrade()
x86/unwind: Add hardcoded ORC entry for NULL
x86/unwind: Handle NULL pointer calls better in frame unwinder
loop: access lo_backing_file only when the loop device is Lo_bound
netfilter: ebtables: remove BUGPRINT messages
f2fs: fix to avoid deadlock of atomic file operations
RDMA/cma: Rollback source IP address if failing to acquire device
drm: Reorder set_property_atomic to avoid returning with an active ww_ctx
Bluetooth: hci_ldisc: Postpone HCI_UART_PROTO_READY bit set in hci_uart_set_proto()
Bluetooth: hci_ldisc: Initialize hci_dev before open()
Bluetooth: Fix decrementing reference count twice in releasing socket
Bluetooth: hci_uart: Check if socket buffer is ERR_PTR in h4_recv_buf()
media: v4l2-ctrls.c/uvc: zero v4l2_event
ext4: brelse all indirect buffer in ext4_ind_remove_space()
ext4: fix data corruption caused by unaligned direct AIO
ext4: fix NULL pointer dereference while journal is aborted
ALSA: ac97: Fix of-node refcount unbalance
ALSA: hda/ca0132 - make pci_iounmap() call conditional
ALSA: x86: Fix runtime PM for hdmi-lpe-audio
SMB3: Fix SMB3.1.1 guest mounts to Samba
irqchip/gic-v3-its: Fix comparison logic in lpi_range_cmp
objtool: Move objtool_file struct off the stack
perf probe: Fix getting the kernel map
cifs: allow guest mounts to work for smb3.11
futex: Ensure that futex address is aligned in handle_futex_death()
scsi: ibmvscsi: Fix empty event pool access during host removal
scsi: ibmvscsi: Protect ibmvscsi_head from concurrent modificaiton
powerpc/vdso64: Fix CLOCK_MONOTONIC inconsistencies across Y2038
MIPS: Fix kernel crash for R6 in jump label branch function
MIPS: Ensure ELF appended dtb is relocated
mips: loongson64: lemote-2f: Add IRQF_NO_SUSPEND to "cascade" irqaction.
udf: Fix crash on IO error during truncate
libceph: wait for latest osdmap in ceph_monc_blacklist_add()
iommu/amd: fix sg->dma_address for sg->offset bigger than PAGE_SIZE
drm/vmwgfx: Return 0 when gmrid::get_node runs out of ID's
drm/vmwgfx: Don't double-free the mode stored in par->set_mode
mmc: renesas_sdhi: limit block count to 16 bit for old revisions
mmc: mxcmmc: "Revert mmc: mxcmmc: handle highmem pages"
mmc: pxamci: fix enum type confusion
ALSA: firewire-motu: use 'version' field of unit directory to identify model
ALSA: hda - add Lenovo IdeaCentre B550 to the power_save_blacklist
ANDROID: dm-bow: Fix 32 bit compile errors
UPSTREAM: sched/pelt: Skip updating util_est when utilization is higher than CPU's capacity
UPSTREAM: sched/fair: Update scale invariance of PELT
UPSTREAM: sched/fair: Move the rq_of() helper function
UPSTREAM: sched/fair: Remove setting task's se->runnable_weight during PELT update
ANDROID: Add dm-bow to cuttlefish configuration
UPSTREAM: binder: fix handling of misaligned binder object
UPSTREAM: binder: fix sparse issue in binder_alloc_selftest.c
BACKPORT: binder: use userspace pointer as base of buffer space
UPSTREAM: binder: fix kerneldoc header for struct binder_buffer
BACKPORT: binder: remove user_buffer_offset
UPSTREAM: binder: remove kernel vm_area for buffer space
UPSTREAM: binder: avoid kernel vm_area for buffer fixups
BACKPORT: binder: add function to copy binder object from buffer
BACKPORT: binder: add functions to copy to/from binder buffers
UPSTREAM: binder: create userspace-to-binder-buffer copy function
ANDROID: dm-bow: Add dm-bow feature
f2fs: set pin_file under CAP_SYS_ADMIN
f2fs: fix to avoid deadlock in f2fs_read_inline_dir()
f2fs: fix to adapt small inline xattr space in __find_inline_xattr()
f2fs: fix to do sanity check with inode.i_inline_xattr_size
f2fs: give some messages for inline_xattr_size
f2fs: don't trigger read IO for beyond EOF page
f2fs: fix to add refcount once page is tagged PG_private
f2fs: remove wrong comment in f2fs_invalidate_page()
f2fs: fix to use kvfree instead of kzfree
f2fs: print more parameters in trace_f2fs_map_blocks
f2fs: trace f2fs_ioc_shutdown
f2fs: fix to avoid deadlock of atomic file operations
f2fs: fix to dirty inode for i_mode recovery
f2fs: give random value to i_generation
f2fs: no need to take page lock in readdir
f2fs: fix to update iostat correctly in IPU path
f2fs: fix encrypted page memory leak
f2fs: make fault injection covering __submit_flush_wait()
f2fs: fix to retry fill_super only if recovery failed
f2fs: silence VM_WARN_ON_ONCE in mempool_alloc
f2fs: correct spelling mistake
f2fs: fix wrong #endif
f2fs: don't clear CP_QUOTA_NEED_FSCK_FLAG
f2fs: don't allow negative ->write_io_size_bits
f2fs: fix to check inline_xattr_size boundary correctly
Revert "f2fs: fix to avoid deadlock of atomic file operations"
Revert "f2fs: fix to check inline_xattr_size boundary correctly"
f2fs: do not use mutex lock in atomic context
f2fs: fix potential data inconsistence of checkpoint
f2fs: fix to avoid deadlock of atomic file operations
f2fs: fix to check inline_xattr_size boundary correctly
f2fs: jump to label 'free_node_inode' when failing from d_make_root()
f2fs: fix to document inline_xattr_size option
f2fs: fix to data block override node segment by mistake
f2fs: fix typos in code comments
f2fs: use xattr_prefix to wrap up
f2fs: sync filesystem after roll-forward recovery
f2fs: flush quota blocks after turnning it off
f2fs: avoid null pointer exception in dcc_info
f2fs: don't wake up too frequently, if there is lots of IOs
f2fs: try to keep CP_TRIMMED_FLAG after successful umount
f2fs: add quick mode of checkpoint=disable for QA
f2fs: run discard jobs when put_super
f2fs: fix to set sbi dirty correctly
f2fs: fix to initialize variable to avoid UBSAN/smatch warning
f2fs: UBSAN: set boolean value iostat_enable correctly
f2fs: add brackets for macros
f2fs: check if file namelen exceeds max value
f2fs: fix to trigger fsck if dirent.name_len is zero
f2fs: no need to check return value of debugfs_create functions
f2fs: export FS_NOCOW_FL flag to user
f2fs: check inject_rate validity during configuring
f2fs: remove set but not used variable 'err'
f2fs: fix compile warnings: 'struct *' declared inside parameter list
f2fs: change error code to -ENOMEM from -EINVAL
Conflicts:
drivers/md/Kconfig
kernel/sched/fair.c
Change-Id: I2c6ba055f1160864446c87507a7fd7c8249ad885
Signed-off-by: Ivaylo Georgiev <irgeorgiev@codeaurora.org>
This reverts commit 549020c814.
This is a preparation change for merging android-4.19.32 into
msm-4.19 branch.
This change affects scheduler behavior that can lead to boot time
regression.
Change-Id: Iedc341fb5d07e7ba1dd37e4b9b4474a84a3afe11
Signed-off-by: Ivaylo Georgiev <irgeorgiev@codeaurora.org>
The current implementation of overutilization, aborts energy aware
scheduling if any cpu in the system is over-utilized. This patch introduces
over utilization flag per sched domain level instead of a single flag
system wide. Load balancing is done at the sched domain where any
of the cpu is over utilized. If energy aware scheduling is
enabled and no cpu in a sched domain is overuttilized,
load balancing is skipped for that sched domain and energy aware
scheduling continues at that level.
The implementation takes advantage of the shared sched_domain structure
that is common across all the sched domains at a level. The new flag
introduced is placed in this structure so that all the sched domains the
same level share the flag. In case of an overutilized cpu, the flag gets
set at level1 sched_domain. The flag at the parent sched_domain level gets
set in either of the two following scenarios.
1. There is a misfit task in one of the cpu's in this sched_domain.
2. The total utilization of the domain is greater than the domain capacity
The flag is cleared if no cpu in a sched domain is overutilized.
This implementation still can have corner scenarios with respect to
misfit tasks. For example consider a sched group with n cpus and
n+1 70%utilized tasks. Ideally this is a case for load balance to happen
in a parent sched domain. But neither the total group utilization is
high enough for the load balance to be triggered in the parent domain nor
there is a cpu with a single overutilized task so that aload balance is
triggered in a parent domain. But again this could be a purely academic
sceanrio, as during task wake up these tasks will be placed
more appropriately.
Signed-off-by: Thara Gopinath <thara.gopinath@linaro.org>
Change-Id: I3f327cff4080096a3e58208dd72c9b7f7913cdb2
Signed-off-by: Chris Redpath <chris.redpath@arm.com>
Git-commit: addef37808728c719d8c095a75bcf81befdacdaf
Git-repo: https://android.googlesource.com/kernel/common/
[clingutla@codeaurora.org: Resolved trivial merge conflicts]
Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
When tasks move across cpusets, the current affinity settings
are lost. Cache the task affinity and restore it during cpuset
migration. The restoring happens only when the cached affinity
is subset of the current cpuset settings.
Change-Id: I6c2ec1d5e3d994e176926d94b9e0cc92418020cc
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
This change is for general scheduler improvement.
Change-Id: I102a9d2a7a12a3ffe08795825eecf73cdb3ef3e8
Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
This change is for general scheduler improvement.
Change-Id: Ie37ab752a4d69569bce506b0a12715bb45ece79e
Co-developed-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
When energy delta is same as the chosen cpu, the energy model simply
rejects the new cpu.
This is incorrect behaviour, because it ends up selecting the cpu it
visits first. It should break ties intelligently - fix it such that
it prefers cpu in a shallower idle state.
Implement this.
Change-Id: I82f39163e71037eca05ea52c69800a71b1fdcc27
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
As freq increases freq/power should decrease, indicating nonlinearly
higher power is required for running at higher frequencies.
If freq/power increases, the cost associated with that
freq will be lower than its previous one, causing the energy
calculations to choose a cpu with that frequency over a cpu with
the previous lesser freq.
For some frequencies, if their voltage increment from the
previous ones is very small (or is the same), we could end up with
higher freq/power ratio. This is primarily because
of_dev_pm_opp_get_cpu_power() returns power in mW and looses precision.
But instead of addressing it, enforce the same cost as the previous
frequency.
The energy evaluation code prefers the previous cpu when the energy costs
are same as other candidates. By keeping the energy costs same in these
situations we are increasing the likely hood of selecting prev cpu even
if it results in a slight freq bump. In general, selecting prev cpu is
beneficial because it avoids warming up the caches at a different cpu.
Change-Id: Ic66ab18ba65f2917b73d9fbe92a39b743b9839a0
Signed-off-by: Abhijeet Dharmapurikar <adharmap@codeaurora.org>
Add ftrace events to capture per-zone memory stats when a
trigger threshold is hit.
Change-Id: Ic52a27713acd34090c8f99543408b277877015c7
Signed-off-by: Patrick Daly <pdaly@codeaurora.org>
Sched_cpu_isolate takes cpu number as input. The input is not
being checked against the NR_CPUS which could potentially
overflow the buffer cpu_isolation_vote. Fix it by checking the
boundary.
Change-Id: If5e07b2481454b539ccdca15a302722478d22df4
Signed-off-by: Runmin Wang <runminw@codeaurora.org>
If the FBT takes the fast path for prev_cpu, a duplicate trace_print
would occur and shares no additional information. Clean this addtional
trace_print.
Change-Id: I3257ffab03b6ba27bec226609b553ff25297e513
Signed-off-by: Runmin Wang <runminw@codeaurora.org>
When a task is waking and both evaluation CPU and the destination
CPU are the task's previous CPU, the task utilization must be
added to the evaluation CPU's utilization. But it is not done.
Due to this the task may get placed on the previous CPU incorrectly
and results in suboptimal power and performance.
Change-Id: I62b5eec3046b7fef31321125dd76f67afe8f567d
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
Add sched_compute_energy trace point to aid the task placement
issues debugging. It prints the energy of all the evaluated
CPUs and the best_energy for each iteration.
Change-Id: I965a950de5006bdc601a9fcc146d8d619409dd8f
Signed-off-by: Pavankumar Kondeti <pkondeti@codeaurora.org>
This change is for general scheduler improvement.
Change-Id: I4c56a1111d983b2a690cbce03787ff1d82f6f506
Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
This change is for general scheduler improvement.
Change-Id: I8a65125f4f5a123f5a9551a6a4876e81d1b77dcf
Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
This change is for general scheduler improvement.
Change-Id: I090ed8f91592646f11dd65bc83357945f02702cd
Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
The current fbt starts the CPU search from the suitable
sched_group/cluster. If an eligible CPU (either busy or idle) is found
in this primary cluster, we break and use this eligible CPU. The search
is expanded to the other cluster only when either an eligible CPU is
not found or boost is enabled. This works for the existing topologies
like 6 + 2 or 4 + 3 + 1 systems. However with a topology like 6 + 1 + 1
where the mid capacity cluster has only 1 CPU, this policy results in
excessive packing on that single CPU. So expand the search to the higher
capacity CPU as well when the start cluster is mid capacity cluster.
The exception is when the mid capacity CPU is idle. This is a power
optimization to not to use the higher capacity CPU more.
Change-Id: I15c591dc2f61fbef203545c86e038a0ff918806c
Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
This change is for general scheduler improvement.
Change-Id: I7b080f2acc4ae978a0086b08ddc81a0ef98f24e6
Signed-off-by: Lingutla Chandrasekhar <clingutla@codeaurora.org>
When new cpufreq min/max limits are set, the expectation is that the CPU
frequency will be updated appropriately. This happens for targets which do
not utilize the cpufreq fast switch APIs, but the code to do this is not
present by default for the fast switch path. Add the necessary code to
update frequency to reflect the new cpufreq limits for targets with fast
switch enabled.
Change-Id: I211b6117005df9d340dfe0d825032cd7600cbffa
Signed-off-by: Jonathan Avila <avilaj@codeaurora.org>
