Development [Kernel][04.05.2023][Android 13.0.0 Stable]Kirisakura_Raviantah 1.0.4 for Pixel 7/Pro aka "Pantah"
- Thread starter Freak07
- Start date
Are you missing some commas there? Do you mean "stable, canary, or canary 25206"? I'm on regular magisk 25.2. Can I flash this over pantah 1.3.0?
Holy hell. You weren't kidding when you said this update resolved dropped frames.
I thought my phone was smooth before. Buttery smooth. But the animations and scrolling are absolutely insanely smooth now. Apps that were notoriously bad are now perfectly fine. The Ankermake 3d printing app was stuttery garbage and now is just fine.
Thanks again for all the hard work.
I thought my phone was smooth before. Buttery smooth. But the animations and scrolling are absolutely insanely smooth now. Apps that were notoriously bad are now perfectly fine. The Ankermake 3d printing app was stuttery garbage and now is just fine.
Thanks again for all the hard work.
you probably solved this bug, which Google can't handle since 7 series launched 
issuetracker.google.com
Google Issue Tracker
JakeDHS07
Senior Member
Anyone getting random reboots since installing the merged kernel? Wasn't happening yesterday and I don't think I've made any other changes that could be the culprit. I've had it happen twice now while watching a video with the YT app. Dunno if that's related to the frames change?
Didn't face a reboot on my end, and I'm running the release for longer already.
Zip up the contents/logs of sys/fs/pstore once the device booted after the crash so I can check them.
JakeDHS07
Senior Member
Thanks, i can see the problem. it's a panic in f2fs driver.
Interesting thing is there were absolutely no changes between the previous release and this one that touch f2fs, so I wonder if maybe an update to an app triggers the codepath that causes the panic so you see it frequently.
I'll try to get it resolved.
JakeDHS07
Senior Member
I'm just glad it wasn't something I did as I was very confused 
. Thanks for taking a look and the quick response! Will look out for the update thanks!
. Thanks for taking a look and the quick response! Will look out for the update thanks!
JakeDHS07
Senior Member
After rebooting maybe like 4 more times? Now it's been running 15 mins with no reboot and literally I have made no changes. Idk if thats normal where it can just fix itself? I'm not as knowledgeable on the specifics.
I'm just glad it wasn't something I did as I was very confused
It doesn't fix itself, the codepath that causes the panic is just not used and therefore the crash not triggered. Seems you were "lucky" and it happened a few times in short succession. If it's always during video playback or a certain scenario that might help to reproduce after testing a fix, but no problems with YouTube on my end.
But it looks like it's a rare thing to happen generally otherwise I'd have experienced the crash too.
I'll get in contact with the f2fs devs. It's rare such a bug slips through and finds its way into f2fs stable.
JakeDHS07
Senior Member
So that's something that's unrelated to this kernel specifically and actually in the source? Sorry just trying to learn and understand
JakeDHS07
Senior Member
Glad I could help someone else avoid it in the future! It certainly was annoying and freaking me out but I'm glad it's not happening now! Let me know what they say. I'm curious now and always looking to expand my knowledge with this platform
HipKat
Recognized Contributor
JakeDHS07
Senior Member
Apparently it was something with the actual driver. Freak looked into it earlier today. It was temporary and oddly random it seems. Hasn't happened in awhile.
As I'm out of the root scene since quite a lot, I miss some new things.
Even if I'm rooted with Magisk, I need to disable vbmeta flags to flash a custom kernel, right?
Even if I'm rooted with Magisk, I need to disable vbmeta flags to flash a custom kernel, right?
HipKat
Recognized Contributor
I'm experiencing the random reboots now too. After phone reboots, I can maybe use if for 1 minute before it reboots. Gonna flash stock kernel to see if it helps.
Top Liked Posts
-
There are no posts matching your filters.
-
19
"Massive" as in benchmarks. Small in non benchmark scenarios, scoped storage eats up those gains for a lot of apps too.
Opposing are the risk of data loss, the greater risk of data corruption.
Because Android keeps fsync strictly enabled for very good reasons and it´s harder to disable than on PC. Only a very small number of users actually unlock their phones, and an even smaller number play with fsync.
I´ve seen cases of corrupted phones because fsync was disabled on a crash. I don´t want this happen when running my kernel.
You´re understanding is wrong. Android is as sensitive (more sensitive due to android being android) to data loss/corruption in case of a crash as PCs are.
If you disabled fsync and never faced data loss/corruption on corruption you were just "lucky", the same can happen for anyone else in the other "unlucky" direction.
Same as with the GPU stuff below. If you´re fine with those consequences that´s one thing, but posting in an advertising manner when there are a lot of users not fully aware of the risks is another. Especially if the risks are downplayed, because one never faced an issue. This gives a wrong impression for many users not familiar with those things.
Yes, there´s the possibility of "dynamic" fsync. That means you might only lose data from when your screen was last enabled and fsync will happen when the screen is off in the best case.
This does not prevent the possibility of data corruption at all when the phone crashes while the screen is on.
It´s not about being comfortable.
Some options are just not worth the risk, given the real world gains. Especially since there are snippets around the internet and XDA (like the part in your post above) that completely understate the risk of disabling fsync.
It´ll appear as a simple option in kernel manager apps once added, accessible to everyone including those not aware of the risk. I can change the sysfs endpoint so it´s a bit more hidden, but again there´s no benefit in adding the option in the first place.
You might run the kernel that supports disabling fsync just fine, device crashes or shuts down unclean, corruption happens without you even noticing it for a week or longer.
Then switch back to stock or any other kernel and slowly begin to face random issues after a week or two due to that corruption.
I know XDA is partly about providing options to the user. At least it is for me. But some things are better be left alone. That´s why I´d appreciate if this option is never provided in any kernel.
If you find anything interesting about the exynos roots let me know.
Some of the exynos parts are just there and not actually enabled, because the firmware disables or just simply has no support for certain exynos parts, because google doesn´t use it.
There´s not much controversy about benchmarks for me. Benchmarks are simply tools for testing very specific scenarios.
You like to chase high benchmark scores, which is completely fine with me.
Benchmark scores may translate to improvements in real world usage, but must not which is most often the case.
As for the GPU topic:
I´ve seen enough dead devices due to playing with the GPU freqs/voltages too. It happens rarely, but when it happens there´s a dead phone with a fried mobo. Not worth it either.
If you´re fine with those consequences that´s one thing, but posting in an advertising manner when there are a lot of users not fully aware of the risks is another.
Additionally to that filesystems are a lot further than 5-10 years ago, where this idea originates from.
As I´m aware asynchronous fsync improves when copying extremely large numbers of files (rather what happens on servers). I wonder how often that happens on regular usage in Android.
It´s not worth adapting that dynamic fsync driver, because most kernel devs value file system integrity higher than boosted benchmark scores or small real world improvements.
As everybody can see this became quite a lengthy post, but this topic really matters to me.
I once suffered filesystem corruption due to a buggy f2fs patch and it was ugly. I never wish that to happen to anyone!65
It's not worth it to disable fsync or disable write barriers. The performance gain is nowhere near the benefit of having your data be intact.
Android is not stable enough to run without fsync, as when Android begins crashing it'll come down hard and not in any graceful manner. When things start going awry there is no way to easily sync, there is no guarantee that the battery can sustain for long enough to sync writes to storage, or that the data can be written without potentially clobbering the file system.
The device is fast enough, the storage speed is fast enough, the penalty for safety is negligible.
All a toggle does is guarantee a user will enable it for "free performance." It's not free.
You pay the safety tax on your data, because if you don't then it only takes one (1) crash for it to be gone.
I formatted my Pixel XL with F2FS and set nobarrier, the performance benefit was only about ~20% on writes, and then it corrupted on the first crash.
It's never worth it.5
I couldn´t have said it better myself than @Namelesswonder did.
It´s also not only about f2fs stability, every crash, even non clean reboot, can cause corrupted data if fsync is disabled.
Another example: An app with root privileges that is able to reboot the device. If this app does a forced reboot and isn´t triggering a clean reboot, disabled fsync might cause data corruption. It´s really ugly if that happens.
Didn´t find time to reply earlier as I didn´t want to provide a single "No" without any explanation and I´m currently really short on time. So a huge thanks for @Namelesswonder for having my back here!
Additionally I remembered this post of mine on the Pixel 4 forum a few years back:
[Kernel][25.03.2023][4.14.311][A12][A13] Kirisakura 9.1.4 for Pixel 4/XL aka FLORAL
Kirisakura-Kernel for the Pixel 4 and Pixel 4...
forum.xda-developers.com
5
Did you update PixelFlasher lately? I updated PF to the latest thinking all my settings would still be set. But unless you reactivate Advanced option, your advanced settings will not be used and that includes disable verity and disable-verification options in PF.
Your corrupt message is from not having verity and verification disabled. Theres no 2 ways around it. You can say you did it right but the photo says it all. Wipe and start over is your only option. -
83Kirisakura-Kernel for the Pixel 7/Pro
Hello everyone,
To keep it short: Here is Kirisakura - Kernel for the Google Pixel 7 Pro aka Cheetah and the Pixel 7 aka Panther, together Pantah.
I would appreciate if everybody that flashes the kernel, reads at least once through this opening post and the following ones.
The kernel aims to keep most of the subsystems updated, way ahead of the stock kernel, thereby improving security, stability and performance!
This includes Linux-Stable, F2FS-Stable and kernel/common!
If that got you curious, have a read about linux-stable and why it is important here. The stable-process is not the same for every subsystem, but the general idea, rule of thumb and benefits are applicable for other subsystems as well.
By merging linux-stable regularly we get most security updates months ahead of the stock kernel and also before they even end up in the android security bulletin.
The kernel includes a lot of improvements and contributions from other developers as well. Without those contributtion this kernel would not exist as it is.
A big part of improvements originate from @arter97´s, @kdrag0n´s and @Sultanxda´s work. Many others contributed in some way or another to this kernel.
A big thanks to all of them at this place!
Now lets continue with a list of features in the next paragraph!
Features:
Main Features:
- Based on latest A13 kernel sources from Google, Kernel is made for Android 13 Stable
- Linux-Stable-Upstream included to 5.10.179
- Compiled with prebuilt Google clang 17.0.1
- Backport entire RCU subsystem to linux 6.0
- Lazy RCU which should result in power-savings while the device is lightly-loaded or idling, more information here
- Backport Maple tree from Linux 6.1
- Per VMA-locks in conjunction with Maple Tree RCU-Mode (improve app launch time, this feature in general benefits greatly from lazy rcu!)
- Dynamic Energy Models depending on device state
- make TEO-Cpuidle util-aware (improves latency, performance and decreases energy consumption for certain workloads, more information here)
- improve CPU throttling behaviour
- improvements to uclamp, prevent capacity inversion (reduce missed frames)
- greatly speed up camera launch time!
- merged kernel/common (improvements to android-common-kernel straight from google)
- MM subsystem reworked (more info and some patchsets linked in this post)
- Multi-gen LRU backported/reworked and enabled (more info here, here as well and here) to improve mm and reduce cpu cycles, latest V15 state
- pelt multiplier tied into powerhal to speed up scheduler during interaction (more info here)
- prevent frequency spikes caused by small transient tasks when the device is idle(more info here)
- tie mechanism to prevent frequency spikes caused by small tasks also into powerhal
- scheduler improvements for RT (realtime) tasks
- introduce and setup PMU limiter (prevents CPU from spiking to max when it isn´t needed, based on PMU reads, more information here)
- improve camera performance by tuning the powerhal during recording
- bias tasks of rt, sf and ta groups to prefer high capacity cpus during app launches, interactions
- improve app launches via powerhal
- restrict maximum CPU-Freqs during screen off/ idle to 1.4GHZ for all clusters to save power
- introduce unfair f2fs rwsems to prevent writer starvation and improve IO perf under heavy load
- fuse: give wakeup hints to scheduler to speed up compress/decompress in internal storage (details)
- enable RCU_BOOST (details here), also fix RCU_BOOST behaviour
- F2FS-Stable updated
- merge and enable f2fs block_age-based extent cache (improve the accuracy for data temperature classification, reduce the garbage collection overhead after long-term data updates)
- TCP backports from mainline
- SSG IO scheduler for reduced overhead and less CPU cycles (more lightweight and android optimized)
- scheduler backports from linux-main
- mm updates from linux-main
- locking updates frm linux-main
- improve memory performance by tweaking google´s eh zram implementation
- fix sysfs endpoint for googles eh zram
- affine IRQS to CPU 7 during camera usage for improved performance as it tends to overload the little cores
- use bbr as default TCP congestion algorithm (fasted algo according to this excellent research from @kdrag0n found here )
- include bbrv2 from google, more info here
- Enable support for TTL spoofing
- important patches from kernel/common for 5.10 (here are more details)
- increase priority of GPU work
- several updates from QPR beta, and a few from Pixel 6 that googles missed to include in the pixel 7 kernel
- improve thermals
- change compaction proactiveness based on device state for improveed memory management
- improve ZRAM usage
- flashing the kernel will preserve root
CleanSlate Features
- CleanSlate Features from @tbalden, big applause here! (s2s, notification booster, battery saver, flashlight notifications. Please note: cleanslate features that work otherwise with rooted devices like kadaway (adblocking) are not implemented on this kernel since I´m running rooted)
- Check the CleanSlate thread for more information about those features
- The CleanSlate thread also contains the apps that allows you to control those features!
- If you consider those features useful, maybe consider purchasing the apps in the PlayStore or a small donation to @tbalden, he deserves it! Don´t forget to switch to beta channel for those apps, as this is needed for device running A13!
Various Optimizations:
- update several drivers to use power efficient workingqueues (for example wlan driver)
- f2fs: reduce timeout for uncongestion
- f2fs: Demote GC thread to idle scheduler class
- f2fs: set ioprio of GC kthread to idle
- tcp: enable advanced tcp, give user more options for tcp alorithm
- mm: vmstat: use power efficient workingqueues
Wakelock Blocker:
- advanced wakelock blocker with the ability to block kernel wakelocks (dangerous, use with caution, please read this for further info). You should only block kernel wakelocks in case you face uncontrollable scenarios, like company wifi causing a wakelock to be active 100% of the time! Blocking kernel wakelocks to "improve" battery life almost always backfires and causes issues!
Powerhint Module:
- restrict little cluster to 1,4ghz mid cluster to 1,4ghz and big cluster to 1,4ghz during screen off, to reduce battery usage for example during music playback
- only use little cores during screen off/device suspend
- account for scheduler and other changes done to the kernel
- boost memory interface controller during interaction (decreased missed frames, improved soc efficiency according to google docs)
- tie pelt multiplier into the powerhal (more info here)
- prevent frequency spikes caused by small transient tasks during idle operation (more info here)
- boost scheduler using the pelt multiplier during fingerprint unlock operation
- setup and control PMU limiter via powerhal (more info here)
- switch between default and custom idle energy model on device is being interacted with / device idle.
- reduce missed frames during scrolling/device interaction by adjusting uclamp boosts
- dynamically adjust target load for memory interface during interaction.
- remove dynamic GPU policy change as it´s causing instability and is unneeded due to GVFS period
- implement handling during video recording of CPU like Google did on QPR A13
- more small improvements
DOWNLOAD:
Downloads are attached to the release posts linked below in the changelog section.
Please note that androidfilehost is currently not working properly.
Link to AFH:
Deprecated afh download, still here for old releases:
Changelog:
Android 13.0.0 Stable (not QPR beta!)
1.0.0 Initial Release | Mirror for Download in case AFH is not working
1.1.0 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-87743627
1.2.0 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-87822161
1.2.1 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-87878321
1.2.2 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-87944505
1.2.3 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-87960183
1.2.4 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-88108123
1.3.0 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-88286027
Raviantah_1.0.0 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-88316285
Raviantah_1.0.1 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-88323527
Raviantah_1.0.2 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-88399915
Raviantah_1.0.4 https://forum.xda-developers.com/t/...-pixel-7-pro-aka-pantah.4509795/post-88497279
Requirements
- vbmeta flags for verity/verification disabled (this requires a wipe if you´re coming from stock with those flags enabled), check the FAQ for information on how to do this
- unlocked Bootloader
- USB-Debugging in developer options enabled
- latest adb and fastboot binaries
- working adb and fastboot environment so you can flash back to stock in case something goes wrong
- working magisk environment (a device rooted with latest magisk, use stable in case you want to be absolutely safe)
- IMPORTANT: Unrelated to the kernel, but update both slots of your phone to A13 before you flash anything to your device!
- it´s handy to have a backup and a way to return back to stock (check the faq for that)
- make sure to download the correct kernel version for your firmware. (you can´t flash a kernel made for stable firmware on a QPR beta or dev preview and expect there to be no isses, the release posts mention which firmware the respective kernel releases are compiled for)
- this kernel is tested on stock google firmware. If you want to use it on custom roms you´re kind of on your own as I can´t account and give support for any changes custom roms apply that could render the kernel incompatible. (e.g. if you´re custom rom is still on the previous months update, there may be issues flashing an updated kernel for the next month on top or custom roms requiring specific kernel changes to work correctly )
- it´s always best to start on stock images and don´t have another custom kernel flashed when flashing this kernel for the first time. consecutive flashes can just be installed on top, if not I´ll mention it in the release post
- read the first and second post, there's a lot of valuable information
How to flash the Kernel:
1a. Make sure you tick all the requirements above
1b. Make sure to flash the release made for your firmware! Usually the release post contains information on that
1. Download the kernel.zip (unified for both devices) and the correct powerhint module depending on your device (Pixel 7 = panther || Pixel 7 Pro = cheetah)
2. Flash the powerhint module via magisk manager as you would any other magisk module! Make sure to flash the latest one. It will be mentioned in the release post if there´s an update.
3. Flash the kernel.zip via EXKM, FKM or kernel flasher. Root will be preserved.
4. Reboot and profit.
Donations:
Donations are not mandatory but very welcome if you want to support development or just buy me a coffee/tea/beer
If you like my work: http://paypal.me/freak07
Credits:
@osm0sis for all his work on AK3.
@tbalden for being the best HTC, Pixel, OnePlus and Asus wingman!
@capntrips for all his work on the pixels, avb, hashtree patcher, the counsel and the kernel flasher!
@LeeDroid and @mwilky for their awesome roms and work I used on multiple devices!
@Captain_Throwback for all the mentoring and guidance!
@Eliminater74 for bringing me into the game and the Inspiration
@nathanchance for his upstream guidance and assistance
@RenderBroken for helping me out
@flar2 for all his work
@joshuous for all the help he provided to me in the past!
@arter97 for giving me advice
@kdrag0n for his help and advices!
@topjohnwu for magisk and his entire work!
Source Code: https://github.com/freak0751Update to 1.2.4
Hey guys and girls,
So here´s the next release. February source did basically not contain anything new for this kernel (as all patches were already included since months due to merging linux-stable and kernel/common in advance) except a stability and a security fix for the GPU driver (which is vendor specific so of course not impacted by linux-stable etc).
Nonetheless this is a very big release with large underlying changes.
Short explanation follows below.
Please note that this release is not for the QPR Beta firmware, but the stable android 13 firmware!
Make sure to update/flash the new powerhint module attached to this post via magisk manager as it´s been updated!
Maple Tree / Per VMA-locks via Maple Tree in RCU-Mode
Linux 6.1 finally merged the Maple Tree patchset. I had my eyes on that patchset for quite a while now. I backported it a few months ago (back in november iirc) but at that time I still ran into a few problems with it so it never made it into a release on XDA.
However after discovering work done on per vma-locks, which builds directly on the benefits that maple-tree provides interest was sparked again. Decided to pull it in again and fix the remaining issues.
If you´re interested in details read the linked topics.
Basically those patches reduce app launch time as well as lock contention. Additionally per vma-locks directly utilizes lazy-rcu, which was backported in 1.1.0. Usually I´m very hesitant when it comes to kernel changes actually translating to real world power consumption improvements, but lazy-rcu is one of the few patchsets that translate to a small improvement for everyone regardless of the usage.
So the new per VMA-lock patchset directly benefits from the previous backport of lazy-rcu.
App launch times in ms for cold launching slack app with and without the Per-VMA lock patchset:
F2FS block_age-based extent cache
Latest f2fs-stable update contained an interesting new feature called block_age-based extent cache. It´s another optimization done in hindsight to android needs. It´s always nice to see different OEMs (in this regards xiaomi) cooperating and improving android as a whole and not just keep those patches to themselves. Since it´s pushed to f2fs-stable it will sooner or later (probably rather later as trickling through until it ends up on actual production devices always takes a lot of time) find its way to the majority of android devices.
The features allows to record data block update frequency of the extent per inode, in order to provide better temperature hints for data block allocation thereby improving the accuracy for data temperature classification and reducing the garbage collection overhead after long-term data updates.
Memory Interface Controller Boosting
Boost Memory Interface Controller slightly to a fixed min_freq value during interaction, instead of relying solely on the governor taking the decision as it doesn´t ramp up fast enough during interaction in some cases.
Research of exynos/tensor characteristic shows that the SOC does not perform optimally when cpu freqs ramp up quickly (during interaction, scrolling, touch the screen) and memory interface controller not following suite quite fast enough. This makes the soc overall perform less efficient. The boost aims to correct this behavior.
The boost also results in a tiny few less missed frames during interaction as a bonus.
There are other small improvements as well. Please check my github for those.
The kernel is made for stable A13 firmware, not beta.
I wish everyone a nice day.
Changelog:
- Linux-Stable bumped to 5.10.167
- Backport Maple tree from Linux 6.1
- Per VMA-locks in conjunction with Maple Tree RCU-Mode (this benefits greatly from lazy rcu!)
- bump f2fs to latest f2f2 stable state
- merge and enable f2fs block_age-based extent cache (improve the accuracy for data temperature classification, reduce the garbage collection overhead after long-term data updates)
- improvements to the scheduler from QPR2 Beta 2 and Beta 3
- improvements to memory management from QPR2 Beta 2
- improve ZRAM usage
- boost memory interface controller during interaction (decreased missed frames, improved soc efficiency according to google docs)
- import security fixes and performance improvments to camera driver from QPR2 Beta
- fixes/improvements to to memory management
- security fixes/ improvements from kernel/common
- other fixes please check github
Download:
Attached to release post as AFH is currently down.
Make sure to update and keep the powerhint module installed via Magisk Manager!
If you´re coming from another kernel restore stock boot.img, dtbo.img, vendor_kernel_boot.img and vendor_dlkm.img before flashing. Thank you.
Make sure to meet the requirements and read the OP as well as the FAQ before flashing.
I wish everybody a great day/evening!
Have fun, enjoy the kernel and your phone.
If you like my work please consider a donation.
Donations are not mandatory but very welcome.
If you like my work and want to buy me a coffee/green tea: http://paypal.me/freak0749Update to 1.2.1
Hey guys and girls,
So here´s the next release. As already hinted it includes some of the kernel changes from the QPR2 Beta. Please note the kernel is made for the stable firmware! You need to be on December stable firmware release to be able to install and run this kernel properly!
Make sure to flash both, the kernel.zip and the powerhint magisk module as both are updated.
Google did some significant updates in that latest qpr2 update, which is great to see.
The GPU driver is now based on version r38p1. This fixes several vulnerabilities that made the news recently.
AOC driver and WiFi driver were also updated and most of those improvements ended up in this kernel release as well.
There were several improvements to memory management done by google, which should result in improved efficiency, improved reclaim and also reduced jank.
There are a few other nice patches that should improve performance and generally improve system responsiveness. For specifics please check those on github.
Scheduler patches from linux-main are also included.
The lazy rcu patchset was updated to the latest state that was also recently merged into the linux kernel. (more info about lazy rcu check here)
Kernel is now compiled with latest 16.0.2 prebuilt Clang from Google.
There are other small improvements as well. Please check my github for those.
I wish everyone a nice day.
Changelog:
- Linux-Stable bumped to 5.10.159
- Bump prebuilt clang to latest 16.0.2 from Google
- include most of the QPR2 improvements so we can enjoy them ahead of time
- fixes/improvements to to memory management
- import all improvments from google QPR2 to powerhint module
- other fixes please check github
Download:
Attached to release post as AFH is currently down.
I´ll push to AFH once it´s back up.
Make sure to update the powerhint module via Magisk Manager!
If you´re coming from another kernel restore stock boot.img, dtbo.img, vendor_kernel_boot.img and vendor_dlkm.img before flashing. Thank you.
Make sure to meet the requirements and read the OP as well as the FAQ before flashing.
I wish everybody a great day/evening!
Have fun, enjoy the kernel and your phone.
If you like my work please consider a donation.
Donations are not mandatory but very welcome.
If you like my work and want to buy me a coffee/green tea: http://paypal.me/freak0742Update to Raviantha 1.0.4
Hey guys and girls,
So here´s the next release for stable May firmware! (although technically it should still work on April as there were no changes that weren´t already present.)
This release is not including the bigger changes I was alluding to during my last post in this thread, but it still includes some big improvements that warrant a new release.
More information below.
Important:
Make sure to flash the powerhint module attached to this post via magisk manager if you haven´t already done so.
Please note that this release is not for the QPR Beta firmware or dev preview, but the stable android 13 firmware!
Please make use magisk stable if you don´t know your way around bugs and don´t want to actively help debugging magisk development.
You should download magisk stable only from official magisk release site!
Latest platform tools have issues flashing super.img, please use platform tools 33.0.3 when updating your firmware
There are other small improvements as well. Please check my github for those.
The kernel is made for stable A13 firmware, not beta.
I wish everyone a nice day.
Changelog:
- Linux-Stable bumped to 5.10.179
- bump clang to 17.0.1 stable
- bump f2fs-stable to latest
- loads of improvements, bugfixes etc to memory management from latest linux main
- improvements to memory management from 6.1 kernel common
- improvements to MGLRU (check first post for more info about MGLRU)
- updates/improvements to RCU from latest linux main
- updates to lazy RCU from latest linux main
- fixed screen-off max cpufreq restriction (thanks to @elesbb for making me aware)
- other fixes please check github
Download:
Attached to release post as AFH is currently down.
Make sure to update and keep the powerhint module installed via Magisk Manager!
If you´re coming from another kernel restore stock boot.img, dtbo.img, vendor_kernel_boot.img and vendor_dlkm.img before flashing. Thank you.
Make sure to meet the requirements and read the OP as well as the FAQ before flashing.
I wish everybody a great day/evening!
Have fun, enjoy the kernel and your phone.
If you like my work please consider a donation.
Donations are not mandatory but very welcome.
If you like my work and want to buy me a coffee/green tea: http://paypal.me/freak0741Update to 1.1.0
Hey guys and girls,
So here´s the next release. I finally feel it´s ready to be publicly released as there was not a single issue encountered over the last two weeks.
Make sure to flash both, the kernel.zip and the powerhint magisk module!
The kernel is now at 5.10.155 way ahead of the stock kernel. To further emphasize the importance of this. About 90% of kernel security issues are solved in linux stable.
5.10.148 and 5.10.149 fixed the CVEs discussed here and here. Those were fixed on kernel common on November 2nd. It might take up to a few months until they trickly down into any officially shipped android kernel!
The initial 1.0.0 release, had those CVEs fixed already as 5.10.149 was merged.
What also needs to be kept in mind is that a big part of those security issues are resolved in linux-stable, sometimes months ago, before those patches end up in a monthly security patch. Additionally to that, the security patches on android devices are about 3 months behind the discovery of those vulnerabilities.
This release features a few major changes, which I want to explain in greater details, so excuse the longer release post. I´d appreciate if everyone takes the time to at least read it.
Introduction of Lazy RCU
1.1.0 has the entire(!) RCU subsystem updated to latest linux 6.0 kernel state. This also allowed for Lazy RCU to be merged into the kernel, which should result in power-savings while the device is lightly-loaded or idling (which is basically the case all the time a smartphone isn´t being interacted with (nothing touching the screen and no other workload such as video editing is done).
If you´re interested in details check the slides, which are also linked in the article I linked above.
A very simplified explanation: RCU functionality can be called 1000s of times a second, batching RCU calls can save power by not calling as often.
That´s one of the changes I wanted to be tested extensively, as there are a few 10000 lines of code changes.
Please keep in mind, those are kernel changes and battery life will not be improved magically by huge amounts, due to such a change, even if the change itself is massive.
Example: If you raise or lower the display brightness by around 10% over an entire day, you´ll see a bigger impact.
Reading the OP gives you a good idea what this project is about!
Dynamic Energy Model
Google introduced the possibility to use different Energy Models "on the fly" a while back in the Pixel 6 and Pixel 7 kernel.
I played around with this during the last few weeks and ironically Google did something similar in the QPR A13 Beta when I checked the firmware dump. Google switches to a different Energy Module while taking Videos. Based on my understanding to control heat better during video recordings.
However I had another idea, but the mechanism I use to switch the profiles taps into the same system.
While the device is idle, a different Energy Model is used, compared to when interaction is happening.
While interaction is happening there are many boosts in play already to bias tasks to get executed as quickly as possible anyway. But switching to a different energy model allows us to basically bias tasks to get even more prominently scheduled to the little and middle cores. Since the mid cores are a78 on p7 and way more energy efficient, this should give as a double positive effect. As during idle state there are no boosts active the scheduler can now freely decide how to use EAS to the best to place the tasks as efficiently as possible.
The energy module was tuned, so the scheduler prefers little and mid cores during idle. (nothing touching the screen, watching videos, listening to music, reading websites etc.)
Additionally to this I decided to include the changes Google did on A13 QPR Beta while taking video to also include already in the kernel.
Powerhal
Switch between default and idle energy model on device is being interacted with / device idle.
Reduce missed frames during scrolling/device interaction by adjust uclamp boosts.
Dynamically adjust target load for memory interface during interaction.
Remove dynamic GPU policy change as it´s causing instability and is unneeded due to GVFS period
Implement handling during video recording of CPU like Google did on QPR A13 stable
Raise minfreqs during screen off to 1,4ghz
Playback over USB C to 3.5mm Dongles
Some users were experiencing issues while using usb-c to 3.5mm dongles while running my kernel. After a long time I finally discovered how to reproduce the bug as the logs did not point towards any kernel error. It happens only if default usb configuration in developer options is set to anything else than "no data transfer". If "no data transfer" is selected there´s no issue at all.
A few commits from linux-stable for the usb dwc3 driver breaks playback when screen is shut off or cycling through the "default usb configuration" settings, which causes the aoc driver that´s responsible for playback to shut down.
Reverting those commits fixes the issue.
Thanks to @WhoIsJohnGalt1979 and @Hurt Copain for nagging me about the issue and providing logs, I finally found a way to reproduce the issue so I was able to debug it.
Given that playback is "glitchy" (a brief audio stutter) on stock kernel as well, when either cycling through different "default usb configuration" settings or turning screen on/off, that driver is generally very wonky. A user told me however that this glitch is solved in QPR Beta. So that´s good to hear if that´s the case.
Other improvements
Several other improvements from kernel/common (for higher branches than 5.10) and linux-main were backported to this branch so we can benefit from those improvements. For details please check my git.
I´d like to write all of this with more detail, but at the moment I lack the time to do so.
Kernel is compiled for stable A13, not A13 QPR Beta!
I wish everyone a nice day.
Changelog:
- Linux-Stable bumped to 5.10.155
- Backport entire RCU subsystem to linux 6.0
- - Lazy RCU which should result in power-savings while the device is lightly-loaded or idling, more information here
- Dynamic Energy Models depending on device state
- Fix audio playback stopping for some users when turning screen off (thanks for @WhoIsJohnGalt1979 and @Hurt Copain for nagging me about the issue and providing logs)
- improve CPU throttling behaviour by backporting a QPR Beta change
- improvements to MGLRU
- improvements from linux-mainline for several subsystems
- scheduler
-
02595c9fb4ec sched/core: Fix comparison in sched_group_cookie_match()
407e4ccbb158 sched/fair: Move call to list_last_entry() in detach_tasks
0ea56a65b686 sched/fair: Cleanup loop_max and loop_break
0461c9736ec2 sched/fair: Make sure to try to detach at least one movable task
f2489ca21b11 sched/fair: Cleanup for SIS_PROP
5a3074a7ac79 sched/fair: Default to false in test_idle_cores()
bb95868debb0 sched/fair: Remove useless check in select_idle_core()
50112e040e6a sched/fair: Avoid double search on same cpu
2457687f35d3 sched/fair: Remove redundant check in select_idle_smt()
-
- memory management
-
6bf69138a12b mm: rename p4d_page_vaddr to p4d_pgtable and make it return pud_t *
9819fcefb782 mm: rename pud_page_vaddr to pud_pgtable and make it return pmd_t *
997cdba71861 mm: add vma_lookup(), update find_vma_intersection() comments
eb1eccad20cb mm/page_alloc: fix obsolete comment in deferred_pfn_valid()
1f30d4f3fb20 mm/page_alloc: remove obsolete gfpflags_normal_context()
17dd220de848 mm/page_alloc: use costly_order in WARN_ON_ONCE_GFP()
4d7be8d8d6d4 mm/page_alloc: init local variable buddy_pfn
521b49f4067c mm/page_alloc: use helper macro SZ_1{K,M}
9838ea6feaee mm/page_alloc: use local variable zone_idx directly
d049e3821e8b mm/page_alloc: add missing is_migrate_isolate() check in set_page_guard()
5dc1c3c60469 mm: remove obsolete pgdat_is_empty()
55e89c7708e3 mm/page_alloc: fix freeing static percpu memory
bd801ed8eace mm/page_alloc: add __init annotations to init_mem_debugging_and_hardening()
ce00e122335e mm/page_alloc: remove obsolete comment in zone_statistics()
cc20df37215b mm: remove obsolete macro NR_PCP_ORDER_MASK and NR_PCP_ORDER_WIDTH
97e15d20dd79 mm/page_alloc: make zone_pcp_update() static
f7425d856d30 mm/page_alloc: ensure kswapd doesn't accidentally go to sleep
236a9831f7e4 mm: add merging after mremap resize
bd1537226f30 mm: mremap: fix sign for EFAULT error return value
4fe39974c254 mm/mremap: avoid unneeded do_munmap call
3983a53d85a6 mm/mremap:: use vma_lookup() instead of find_vma()
b832f38c3630 mm, hugepages: add mremap() support for hugepage backed vma
23aa8d328a4c mm/mremap: allow arch runtime override
b029910a3ec8 mm/mremap: use pmd/pud_poplulate to update page table entries
53093a2f034d mm/mremap: don't enable optimized PUD move if page table levels is 2
c7b8580c66ff mm/mremap: convert huge PUD move to separate helper
d7e1ce9e582e mm/mremap: fix memory account on do_munmap() failure
a2a195324435 mm/mremap: use vma_lookup() in vma_to_resize()
7f87797e053b mm/mremap: don't account pages in vma_to_resize()
bbebb360b5e2 mm: forbid splitting special mappings
9fce5b920e9d mremap: check if it's possible to split original vma
089557da4960 vm_ops: rename .split() callback to .may_split()
c180d1867972 mremap: don't allow MREMAP_DONTUNMAP on special_mappings and aio
e7404be1ba61 mm/mremap: for MREMAP_DONTUNMAP check security_vm_enough_memory_mm()
5bae8c6fbfff mm/mremap: account memory on do_munmap() failure
281904d82593 mm: refactor of vma_merge()
-
- PSI
-
d95535a6cabf sched/psi: Fix periodic aggregation shut off
73abf1a6b993 psi: dont alloc memory for psi by default
UPSTREAM: psi: Fix psi state corruption when schedule() races with cgroup move
df3995427bf1 sched/psi: Remove unused parameter nbytes of psi_trigger_create()
feb99cdbd69e psi: Fix PSI_MEM_FULL state when tasks are in memstall and doing reclaim
83fc90cd1a63 psi: Reduce calls to sched_clock() in psi
f6af1db49cbd psi: Optimize task switch inside shared cgroups
317f29565dd3 psi: Pressure states are unlikely
7e17bae683a3 psi: Use ONCPU state tracking machinery to detect reclaim
118d212fd15e psi: Add PSI_CPU_FULL state
-
- more details please check github
- increase priority of GPU work
- binder fix
- updates from kernel/common to several subsystems
- tweak powerhal for improved performance/efficiency
- Switch between default and idle energy model on device is being interacted with / device idle.
- Reduce missed frames during scrolling/device interaction by adjust uclamp boosts.
- Dynamically adjust target load for memory interface during interaction.
- Remove dynamic GPU policy change as it´s causing instability and is unneeded due to GVFS period
- Implement handling during video recording of CPU like Google did on QPR A13 stable
- Raise minfreqs during screen off to 1,4ghz
Download:
Attached to release post as AFH is currently down.
I´ll push to AFH once it´s back up.
Make sure to update the powerhint module via Magisk Manager!
If you´re coming from another kernel restore stock boot.img, dtbo.img, vendor_kernel_boot.img and vendor_dlkm.img before flashing. Thank you.
Make sure to meet the requirements and read the OP as well as the FAQ before flashing.
I wish everybody a great day/evening!
Have fun, enjoy the kernel and your phone.
If you like my work please consider a donation.
Donations are not mandatory but very welcome.
If you like my work and want to buy me a coffee/green tea: http://paypal.me/freak07 - scheduler
