DragonHeart kernel

Release Name: rc1

Release Date: 2021-11-11

Release Type: Stable

Build Number: 193

Download: DragonHeart-op7-rc1.zip

MD5 Checksum: 67571b1c0b06e0e51616195293ed6567

Changelog

• rebased off LA.UM.9.1.r1-11300-SMxxx0.0
• imported wlan and techpack off LA.UM.9.1.r1-11300-SMxxx0.0
• build rmnet extensions
• switch to power_efficient workingqueue for techpack
• optimize techpack for size and nuke debug information on it
• reduce debugging in qcacld
• optimize qcacld for size and nuke debug information on it
• switch to power efficient workingqueue for qca-wifi-host-cmn
• upstreamed dtc to v1.4.6-9
• added z order fod commit
• added cpuidle patches and backports for better deep sleep
• added -pipes optimization
• nuked trace_printk()
• switched to -O3 treewide
• switched to -O3 for LLD
• optimized flags for kryo 485
• optimized for crypto and crc
• optimized arch specfic code for kryo 485, crypto and crc
• nuke auditing
• enabled store motion pass for global common subexpression elimination(GCSE)
• imported wireguard v1.0.20210606
• enabled TTL & HL fixations
• picked gulch’s treewide/oneplus: massive code removal commit
• cleaned up unneeded synaptic drivers
• inlined i2c interfaces for touch drivers
• disable auto test for s3706 driver
• conditionally compile out auto_test functions
• enforce msm drm notifier for touch drivers
• nuke logging in touch driver
• init touch drivers earlier
• imported vDSO32
• switched to -O3 for vDSO32
• picked masahiro’s kbuild commits for quality of life
• enabled lse_atomics
• inlined spinlocks
• enabled DCE
• picked fp inlining commits for faster fp
• enforce msm drm notifier for fp
• enabled OPTIMIZE_INLINING
• picked power efficient workingqueues treewide
• picked & enabled BBR
• switched to WESTWOOD for default tcp congestion algorithm
• imported srandom
• switched to srandom over hw_random
• backported cgroup commits
• nuked debugging configs
• upstreamed lz4
• switched to lz4 as default crypto compression algorithm
• switched to zswap over zram
• used vbswap as a fake block device for zswap
• nuke additional ext4 tracing by caf
• add haptic level adjustment
• drop samsung_oneplus_dsc 1080p timings to fix google maps
• introduced tri-cluster api to affine IRQs
• introduced cpumasks for big, LITTLE and prime cpu masks
• affined IRQs treewide
• introduced sched api to migrate current process to a given cpumask
• Improved the scheduler and its mechanism(walt) by grabbing patches from CAF & RenderBroken
• checkout sched/energy to android-4.14-stable to move way from caf’s mess
• introduced devfreq_boost to boost DDR bus on custom events
• affined hwcomposter to big CPUs
• imported simple_lmk and tune it for our needs
• disable CONFIG_DEBUG_KERNEL for lesser debugging
• introduce f2fs rapid gc
• introduce cpu_input_boost to boost CPU on custom events
• speed up mremap by 20x
• reduce time taken to enter sleep
• nuked more debugging and tracing
• switched to cfq as default iosched
• reduced devfreq wake boost duration to 500ms
• introduce kprofiles 1.0.0
• inline active_mode() for faster execution for kprofiles
• switch to case statement over if/else to reduce latency for kprofiles
• misc improvement to kprofiles codebase
• implement display notifiers for auto kprofiles
• bump kprofiles to 2.0.0
• take mode=0 into account for active_mode() for kprofiles
• introduce 2 new api for changing modes on custom events for kprofiles
• nuke redundant break statement in kprofiles
• export sysfs to allow user to disable auto kprofiles
• bump kprofiles to 3.0.0
• mark more kprofiles functions as inline
• refactor common notifier of kprofiles to use the new API
• rename mode 0 of kprofiles to normal for clarity
• fix build when AUTO_KPROFILES is disabled
• hardcode battery mode as default
• boost DDR bus and CPU according to profile set for zygote forking (app launches)
• boost DDR bus and CPU according to profile set for frame rendering (smoothness)
• force set freqs to max/min depending on profile set for performance governor
• increase gpu data processing limit
• forbid init from changing iosched
• introduce STREEBOG algorithm for crypto operations
• reduce logging treewide
• reduce logging in binder
• silence log spam when cpu is brought up
• skip OOM_KILLER when compiled for android
• nuke rmnet logging
• disable NL80211_TESTMODE
• use TCP_NODELAY to improve network latency
• use an efficient frequency table with an optimized energy model
• configured minimum idle frequency for better idle drain
• confugured min freq fallback after boosting for better overall drains
• switch to 50hz kernel tickrate
• force gpu idle timeout to 58ms
• implement rhel’s low latency cmdline for less latency
• power off dsi phy during idle to save more power
• remove pm qos active latency override
• nuke display ramdump mem region
• nuke unused 36 MiB memdump region for less ram usage
• disable broken IRQ detection
• enable freq-energy-model to correlate with checkout of sched/energy to android-4.14-stable
• add new GPU step from CAF
• enable some panel tweaks for oneplus displays
• cleanup dtbo building
• suppress verbose output for cmdline
• configure frame transfer time to 800us
• misc improvements to dtsi
• spoof fake kernel version to fsck
• mm patches from sultan for overall improvement of mm
• nuke pm qos implementation in adsp to save power
• enable RELR relocation packing for smaller image
• compile out ipa_v3 wakelock code
• add timeout to wakelocks globally
• nuke pm qos usage in vidc for less power used during media playback
• fix misconfigured irq in rpmsg
• revert logging of interrupts for GICv3
• built with EvaGCC 12.0.0
• used LLD 13.0.0 as linker

Credits

@osm0sis for anykernel

@topjohnwu for magisk

@nem0-z, @mvaisakh, @kerneltoast, and many more for their guidance, commits and improvements towards this kernel

@dakkshesh07 for developing kprofiles and giving helpful pointers about issues in my kernel

My testers for vigorously testing my kernel