Develop Bugs

attempt to extend ril/rilgsm functionality

see http://www.xs4all.nl/~itsme/projects/xda/rilhook.html
in subdirectory 'rilhook'.
I tried to add ril ioctl functions to the rilgsm.dll, by patching a jump in ram to my extension dll for a specific version of rilgsm.dll.
all code seems to execute at the expected moments.
just the result never arrives in my test application.
any ideas?
extendril.cpp
* contains code that inserts the hook in rilgsm.dll
* contains the code to be executed for our ioctl.
loadrilextension.cpp
* inserts the extendril.dll hook in the device.exe process memory
rilhook.S
* small assembler stub, that calls our ioctl
riltest.cpp
* program to test it all

registry settings help... what do you have on your machine?

On your cavaliers, what do you have for the following registry settings?
* HKEY_LOCAL_MACHINE\MUI\SysLang
* HKEY_LOCAL_MACHINE\nls\DefaultLCID
* HKEY_LOCAL_MACHINE\nls\SystemLCID
* HKEY_CURRENT_USER\MUI\CurLang
thanks

my Cavalier's is followings
* HKEY_LOCAL_MACHINE\MUI\SysLang : 1033
* HKEY_LOCAL_MACHINE\nls\DefaultLCID : 1041
* HKEY_LOCAL_MACHINE\nls\SystemLCID : 1041
* HKEY_CURRENT_USER\MUI\CurLang : 1033
I hope to help you.
ikkoku said:
On your cavaliers, what do you have for the following registry settings?
* HKEY_LOCAL_MACHINE\MUI\SysLang
* HKEY_LOCAL_MACHINE\nls\DefaultLCID
* HKEY_LOCAL_MACHINE\nls\SystemLCID
* HKEY_CURRENT_USER\MUI\CurLang
thanks
Click to expand...
Click to collapse

Thanks... seems like the formatting for the various languages is different on different machines. my settings are
* HKEY_LOCAL_MACHINE\MUI\SysLang : 1033
* HKEY_LOCAL_MACHINE\nls\DefaultLCID : 1033
* HKEY_LOCAL_MACHINE\nls\SystemLCID : 1028
* HKEY_CURRENT_USER\MUI\CurLang : 1033

Kernel scheduler tweaks (non-BFS)

First, let me say I am deeply sorry to open yet another thread on this subject. But I have made some investigations of my own and would like to make some clarifications which I think should not be buried on page 6 of some long thread....
This thread continues the discussion started by these threads:
http://forum.xda-developers.com/showthread.php?t=689829
http://forum.xda-developers.com/showthread.php?t=785988
From what I gather there about 5 different parameters which are purposed for tweaking:
- NEW_FAIR_SLEEPERS (default ON) - This is about the only parameter which I do feel some difference when disabling it (on Floyo 0.8) and for the worse. Hence currently I intend to leave it on.
- NORMALIZED_SLEEPER (default ON) - I found various recommendations to turn this off. I don't feel much difference but maybe there is a slight improvement. Currently I intend to turn it off.
- sched_latency_ns:
* Targeted preemption latency for CPU-bound tasks:
* (default: 20ms * (1 + ilog(ncpus)), units: nanoseconds)
*
* NOTE: this latency value is not the same as the concept of
* 'timeslice length' - timeslices in CFS are of variable length
* and have no persistent notion like in traditional, time-slice
* based scheduling concepts.
*
* (to see the precise effective timeslice length of your workload,
* run vmstat and monitor the context-switches (cs) field)
- sched_min_granularity_ns -
* Minimal preemption granularity for CPU-bound tasks:
* (default: 4 msec * (1 + ilog(ncpus)), units: nanoseconds)
- sched_wakeup_granularity_ns (not sure if this one was mentioned in previous threads but also related)
* SCHED_OTHER wake-up granularity.
* (default: 5 msec * (1 + ilog(ncpus)), units: nanoseconds)
*
* This option delays the preemption effects of decoupled workloads
* and reduces their over-scheduling. Synchronous workloads will still
* have immediate wakeup/sleep latencies.
Note the descriptions are from the kernel sources and default values are a bit different and also dependent on the init.rc in the ramdisk which is dependent on the ROM.
I tryed to play with the last 3 parameters in a very wide range and did not feel much difference. Based on my experience and searches over these forums and such I currently have these values:
sched_latency_ns 5000000
sched_wakeup_granularity_ns 1000000
sched_min_granularity_ns 1000000
They are somewhat arbitrary and if someone can bring up a scenario where the difference is noticeable I will happy to hear it.
Some technical notes:
- I have seen this "numerical" tweak in multiple places:
kernel.sched_features = 24188
This is not smart using a numerical value, as there are some new backports which I would like to incorporate into my kernel and they actually change the "correct" numerical value here.
The preferred way to tweak these settings was brought by androcheck here. Just a reminder:
Code:
# Step 1: Mount debugfs (which is a virtual filesystem like /proc/ or /sys/)
mount -t debugfs none /sys/kernel/debug
# Step 2 (optional): Display the current scheduler flags
cat /sys/kernel/debug/sched_features
# Step 3: Set the flag NO_NEW_FAIR_SLEEPERS
echo "NO_NEW_FAIR_SLEEPERS" > /sys/kernel/debug/sched_features
# Step 4: Unmount debugfs again
umount /sys/kernel/debug
I would actually skip the unmounting step since it doesn't bother to have it there and then you can play with the different parameters at /sys/kernel/debug directly.
I would like to emphasize my point about the kernel features tweaking, using an example with my current settings (only NORMALIZED_SLEEPER disabled):
Code:
# mount -t debugfs none /sys/kernel/debug
# echo NO_NORMALIZED_SLEEPER > /sys/kernel/debug/sched_features
# cat /sys/kernel/debug/sched_features | sed "s/ /\n/g"
NEW_FAIR_SLEEPERS
NO_NORMALIZED_SLEEPER
WAKEUP_PREEMPT
START_DEBIT
AFFINE_WAKEUPS
CACHE_HOT_BUDDY
SYNC_WAKEUPS
NO_HRTICK
NO_DOUBLE_TICK
ASYM_GRAN
LB_BIAS
LB_WAKEUP_UPDATE
ASYM_EFF_LOAD
NO_WAKEUP_OVERLAP
LAST_BUDDY
# cat /proc/sys/kernel/sched_features
24189
So I would highly recommend tweaking the sched_features only through the /sys/kernel/debug/sched_features where you actually know what you are doing.
Then to make it permanent you can grab the numerical value from /proc/sys/kernel/sched_features and write it, for example, to the sysctl.conf.
- One final technical note: f_padia in his post here detailed nicely how to make the changes permanent using sysctl.conf.
I find sysctl also very useful for temporary tweaking of the values, for example:
Code:
sysctl -w kernel.sched_latency_ns=5000000
sysctl -w kernel.sched_min_granularity_ns=1000000
sysctl -w kernel.sched_wakeup_granularity_ns=1000000
Alternatively you could also write them to the appropriate "file" at /proc/sys/kernel. Both these methods do not require the mount of the debug_fs but neither gives you the literal interface to the kernel features (which the debug_fs does - see above).

Removed the part about the OWNER_SPIN, apparently was using an experimental version of my kernel which has an additional OWNER_SPIN sched_feature which makes the numerical value of 24188 incorrect.
Most likely you have no idea what I am talking this was just one epic fail on my side - you can ignore it. Just read the first post it is now updated.

looks nice.. will give it a go

Thanks for your research and additional information to this topic!

this IS excellent work.. on my 2.1 dext these hacks, included in one .sh in init.d then chmod +x, improved a lot speed and I can even think there's a faster boot
you all rule man!

HD2 Android board ID

I have been reading quite a lot regarding android-porting and notes to OEMs on writing bootloaders for android. However, due to the current situation of port on HD2, i have a small problem that is, stock android init recommends the bootloader to pass a hardware identifier string which it will use to do init. I know i cannot explain perfectly as english is not my native language, here is a quote from android-porting group:
After the kernel boot is complete, the init program in the Android root file system gets started. This program parses and executes the scripts init.rc and init.XYZ.rc, where XYZ is the name given by the hardware vendor. The XYZ has to be specified as a value for the kernel parameter “androidboot.hardware=XYZ”. The init program uses this parameter to know the name of the script it has to start. On the goldfish target, XYZ is goldfish. On HTC Desire, XYZ is bravo. ie. "androidboot.hardware=bravo" is included in the kernel cmdline, so init.bravo.rc is executed alongside.
Click to expand...
Click to collapse
I have seen roms using bravo, htcleo, leo as their hardware identifier tags. However using either one of them as static string in cLK would be impartial for either. As of MAGLDR, it passes no cmdline (as far as i've seen) since it uses the yaffs2 boot structure and there is no cmdline it attaches statically. This could also mean that the current kernels instead of depending on the bootloader supplied info, use spaghetti code which makes assumption that it is running on a leo board, which is a deterrent if we are to merge the leo codebase with mainline code such as cm-kernel.
The current possibile solutions for cLK i have brainstormed over are:
1) Let it go as it is and do not include the hardware parameter.
2) Cleanup kernel space, script file names in roms, include the hardware parameter.
3) Leave the script names as it is, cleanup kernel code and let the user supply the hardware string through fastboot.
Q) What about MAGLDR ? it is closed source, we cannot change the kernel commandline in it !
A) Ever since we had NAND boot on HD2, the devs included a special kernel atag that would signal that the kernel is booting through NAND, so if this particular atag is found then the kernel would use a static hardware string.
Code:
*ptr++ = 4; // Size of this atag. atag[0]
*ptr++ = 0x4C47414D; // NAND boot atag. atag[1]
*ptr++ = 0x004b4c63; // cLK sign tag. atag[2]
*ptr++ = 16; // cLK version tag. atag[3]
All this logic could be wrapped inside something like "#ifdef TARGET_HTCLEO" to make it more clean till i notify cotulla about this on irc and he does the needful before the mag 2.0 release.
This exact thing was put into light some time ago by Charansingh but had no proper outcome, the thread stands dead with no actual agreement between chefs.
I tried my best and read as much as i could before writting down this post, if you have any suggestions, insights, solution, correction. Please PM me and i will update this post.

I just don't get it sorry

Hi Rick,
I'm not a dev, I only try to understand how Android works by reading your sources of cLK
But in order to be "Android compliant" it should be better to include a hardware parameter.
Now if I have correctly understood your post, this hardware parameter is stored by the boot structure on native Android devices.
In HD2, yaffs2 is used and there is no need to specify any parameters with it.
But on native Android devices, how is the boot structure? yaffs2 is also used? or this an other?

I dont really know too much but when im porting bravo roms with the the get prop=bravo i change it to leo to make it boot.

john_matrix said:
Hi Rick,
I'm not a dev, I only try to understand how Android works by reading your sources of cLK
But in order to be "Android compliant" it should be better to include a hardware parameter.
Now if I have correctly understood your post, this hardware parameter is stored by the boot structure on native Android devices.
In HD2, yaffs2 is used and there is no need to specify any parameters with it.
But on native Android devices, how is the boot structure? yaffs2 is also used? or this an other?
Click to expand...
Click to collapse
The boot structure on native devices including on cLK is as follows:
Code:
#define BOOT_MAGIC "ANDROID!"
#define BOOT_MAGIC_SIZE 8
#define BOOT_NAME_SIZE 16
#define BOOT_ARGS_SIZE 512
struct boot_img_hdr
{
unsigned char magic[BOOT_MAGIC_SIZE];
unsigned kernel_size; /* size in bytes */
unsigned kernel_addr; /* physical load addr */
unsigned ramdisk_size; /* size in bytes */
unsigned ramdisk_addr; /* physical load addr */
unsigned second_size; /* size in bytes */
unsigned second_addr; /* physical load addr */
unsigned tags_addr; /* physical addr for kernel tags */
unsigned page_size; /* flash page size we assume */
unsigned unused[2]; /* future expansion: should be 0 */
unsigned char name[BOOT_NAME_SIZE]; /* asciiz product name */
unsigned char cmdline[BOOT_ARGS_SIZE]; /* kernel cmdline */
unsigned id[8]; /* timestamp / checksum / sha1 / etc */
};
/*
** +-----------------+
** | boot header | 1 page
** +-----------------+
** | kernel | n pages
** +-----------------+
** | ramdisk | m pages
** +-----------------+
** | second stage | o pages
** +-----------------+
**
** n = (kernel_size + page_size - 1) / page_size
** m = (ramdisk_size + page_size - 1) / page_size
** o = (second_size + page_size - 1) / page_size
**
*/
The cmdline argument is stored in the boot.img (for native devices and cLK) but cotulla felt that using yaffs2 is easier for testing many kernels, however you need a cwr zip package to flash a kernel in magldr, as using dwi.exe will format the phone while for cLK you can do "fastboot erase boot" and "fastboot boot zImage initrd.gz". However for cLK, the initramfs has to be reflashed with the new kernel.

Thank you Rick to reply me
Is the HD2 boot "trick" can be explain by the utilization of YAFFS2 for the whole HD2 NAND instead of EXT for native Android devices?
For example, maybe you need to specify the hardware parameter if you use EXT on the HD2 NAND (I don't know if this is possible?)
Correct me if I'm wrong (certainly )
Edit: for me, this is better to use fastboot commands in order to "manipulate" my HD2 but for the other users, I don't know. Maybe they can say to us.

[Kernel] 4.4.2 [Stock ROM] JassyKernel Release 1 [5 April 2014]

D801 for now only!
Features
Code:
* Fully customizable MPDecision
* MPDecision TouchBoost
* 3-phase thermal throttle
* Westwood+ TCP congestion control default
* 2.8Ghz OC
* UV capable
* Wheatley added
* ROW/FIOPS/SIO/VR/ZEN/Deadline/Noop/CFQ
* Dynamic File Sync
* Asynchronous Fsync
* Power Suspend
* Color Control
* F2FS Support (Planning to see about at least making a recovery to format /data partition with this)
* Debug code reduced
* -O2
* 4.6 Google Toolchain
* Optimized
* Other stuff I probably forgot
Here's a kernel I've been working on. You might know me from the Moto Cliq, HTC Sensation, Nexus 4, or Nexus 7 FHD and like always this kernel is just something I throw together from other devs work. I don't know how to code... at all. But I do like certain features from some devs and some from others and I like to make one that has those.
As it stands the kernel is very usable without modifying anything, but I am playing around to see what works best. Wheatley with 4 cores min seems to be doing pretty well. Give your suggestions to see what will become default.
FauxClock, TricksterMOD, and Kcontrol all work with this kernel. You may want to use FauxClock + one of the other two to take advantage of all the features.
Downloads
https://mega.co.nz/#!QRRWBSaC!CgnrXVBHyikQICjYCglr4-Bhgv6Dz0Jcjy9HQbAjI8c
It asks for a login but just hit cancel. Trying to figure this out, I've never run a server before.
SOURCE

Release 1
Code:
* 2f59497 (Paul Reioux) : fs/Kconfig: fix warning related to AYSNC_FSYNC patch
* b803126 (Paul Reioux) : OVERCLOCK: add ultimate overclock to 2.8 GHz
* f246915 (Paul Reioux) : f2fs: data: AIO optimization compatibility fixup
* 723f678 (Paul Reioux) : drivers/Makefile: add exFat support
* fd4441c (Paul Reioux) : drivers/exfat: add Samsung's open source exFat module version 1.2.4
* faa56ee (faux123) : scripts: remove localversion "+" from kernel version strings
* 1561c3f (Jose Salgado) : Release 1
* b8bdb8c (Jose Salgado) : Revert "fastcharge: initial Nexus 5 adaptation"
* e00a8f2 (Paul Reioux) : init/Kconfig: decouple DEBUG_KERNEL from EXPERT settings
* a865b5c (Jose Salgado) : revert a61aafc
* dbdddf8 (dr87) : Overclocking: if -> ifdef
* a61aafc (Paul Reioux) : sched/fair: improve fairness efficiency
* 2cee7d6 (dr87) : Overclocking: Enforce stock freq on boot rather than table limit
* cbb5a5a (Paul Reioux) : CPU OVERCLOCKING: initial CPU Overclocking for Snapdragon 800 Socs
* c28e477 (dr87) : cpufreq: Remove CPUBoost
* 772b0e0 (Paul Reioux) : Asynchronous Fsync: initial extraction of Async Fsync from HTC
* 612a6c8 (Paul Reioux) : msm: kgsl: show current power level for 8974 SOCs.
* 5aab668 (Paul Reioux) : fs/dyn_sync_cntrl: dynamic sync control
* bc3e94b (Paul Reioux) : fastcharge: initial Nexus 5 adaptation
* 8fcbf8c (Girish Mahadevan) : ARM: dts: msm: Allow L2 to be in GDHS during Standalone PC for msm8974
* 39ed978 (Paul Reioux) : msm8974-regulator.dtsi: further reduce CPU retention voltage
* eaed48b (Paul Reioux) : msm8974-regulator.dtsi: adjust GFX regulator parameters
* 68a18c0 (flar2) : Update ARM topology and add cpu_power driver
* b4828f1 (morfic) : remove tracing, fix piggy, tune build
* 9426ce0 (dr87) : machmsm: fix up rq stats
* 405c2c0 (Dennis Rassmann) : kbuild: drop cpuinit/cpuexit checks in modpost
* da4cae5 (Jose Salgado) : Revert "random: bringup"
* c502d32 (dennes544) : FS: import F2FS
* 60d3af0 (dr87) : bricked: Add kconfig and makefile
* 78f55e7 (dr87) : Revert "cpufreq: enable sibling cpu policy sharing"
* 15b1119 (dr87) : Revert "drivers/cpufreq/cpufreq.c: re-enable cpu policy restore"
* 657a382 (Paul Reioux) : drivers/cpufreq/cpufreq.c: re-enable cpu policy restore
* 691de92 (Paul Reioux) : cpufreq: enable sibling cpu policy sharing
* 25284d4 (franciscofranco) : sched: set sched_mc_power_savings to 2, which is the highest power saving possible value for this tunable
* 586bbc6 (Paul Reioux) : drivers/cpufreq/cpufreq.c: re-enable cpu policy restore
* d79f17e (Pranav Vashi) : msm: cpufreq: Prevent going above max freq when adding new device
* 0657604 (dr87) : lge: update board files
* 1946f87 (dr87) : bricked: add thermal_qcom
* efa211a (dr87) : bricked: update thermal
* ceed881 (Paul Reioux) : kernel/power/powersuspend: default to userspace for now
* 4e1dff6 (yank555-lu) : kernel/power/powersuspend: add back userpace control w/ default kernel control
* 4b2d961 (yank555-lu) : kernel/power/powersuspend: remove userspace dependency from powersuspend
* 3ae0cb6 (Paul Reioux) : powersuspend: new PM kernel driver for Android w/o early_suspend
* 9001717 (dr87) : bricked: add mpd_qcom
* b60fcf9 (dr87) : bricked: fix up mpdecision
* 4958d9b (dr87) : Port Bricked mpd in one commit
* f99197b (Jose Salgado) : just getting rid of it real quick
* c09c4d5 (Jose Salgado) : add .gitignore
* b84f632 (Dennis Rassmann) : kernel/sysctl: fix: 'warning: 'kbuf' may be used uninitialized in this function'
* eefb404 (showp1984) : Makefile: Optimizing build-options for linaro
* c2e5c62 (showp1984) : Makefile: Added Snapdragon optimizations
* 12b6430 (showp1984) : Makefile: adding config variable for Os/O2/O3 compilation
* 23e5630 (xcstacy) : Add /dev/frandom
* 1b38317 (Ezekeel) : Added CPUfreq governor 'wheatley'
* dd042dd (morfic) : Tune deadline, readahead, ondemand, interactive.
* c834f9e (savoca) : Add BFQ/SIO/FIOPS/ZEN/VR IO Schedulers
* e51f133 (houst0nn) : random: bringup
* c8d665d (Ricardo Cerqueira) : kcal_ctrl: Remove the lower limits from the RGB values
* 94bf2c6 (franciscofranco) : Enable pipe flag.
* a45e550 (imoseyon) : cpufreq: force cpuN policy to match cpu0 when changing freq or gov
* 9f38330 (Viresh Kumar) : cpufreq: Manage only online cpus
* 42b0f79 (faux123) : lib/string: use glibc version
* f750383 (faux123) : lib/memcopy: use glibc version
* 124f788 (Francisco Franco) : audit: kiss goodbye you stupid piece of crap logging messages.
* 305a3aa (WhatHub) : use of swp{b} is deprecated for ARMv6 and ARMv7
* 55c7b59 (Paul Reioux) : arch/arm/mach-msm/cpufreq.c: reduce dmesg log spam
* b2bcbf0 (Francisco Franco) : msm: krait: add CPU Voltage control.
* a8267bc (dr87) : scripts: add dtb tool
* 7e4dcc5 (dr87) : Make no external modules, build them into the kernel
* affd673 (Jackpot Clavin) : Add bull**** device tree stubs.
* 6b3b755 (Jose Salgado) : Initial commit

One more for me

Very good, waiting for d802 release
Sent from my LG-D805 using XDA Premium 4 mobile app

Sweet d801 on it
Sent from my LG-D801 using XDA Premium 4 mobile app

waiting for d802

No access for me.
Sent from my LG-D801 using XDA Premium 4 mobile app

razrlover said:
No access for me.
Sent from my LG-D801 using XDA Premium 4 mobile app
Click to expand...
Click to collapse
Just his cancel and then you should be able to download. Works for me when I downloaded from my phone
Sent from my LG-D801 using xda app-developers app

razrlover said:
No access for me.
Sent from my LG-D801 using XDA Premium 4 mobile app
Click to expand...
Click to collapse
pluginre said:
Just his cancel and then you should be able to download. Works for me when I downloaded from my phone
Sent from my LG-D801 using xda app-developers app
Click to expand...
Click to collapse
like plug said.
Sorry, trying to fix that. It's also causing issues when adding FauxClock Download support.
Working on other variants also. Will keep you guys posted.

Will this work for cloudy pro 2 based rom?
Sent from my LG-D801 using Tapatalk

jeyd02 said:
Will this work for cloudy pro 2 based rom?
Sent from my LG-D801 using Tapatalk
Click to expand...
Click to collapse
That's what I'm currently using.

jassycliq said:
That's what I'm currently using.
Click to expand...
Click to collapse
Great sorry to Annoy lol but how is this one different from simple kernel 2?
Sent from my LG-D801 using Tapatalk

@dr87 seems to be using a lot of @franciscofranco's work with that kernel. It's also suppose to be leaner with less features and less "garbage".
It's a good route but I prefer to use @show-p1984's hotplug and thermal, it allows users full control over hotplugging. I've also added "wheatley" because I wanted to see how it would do running 4 cores always on; have yet to tune it so its not completely smooth. (The gov Wheatley is not smooth IMO, the kernel with its defaults is IMO)
A lot of it is in the changelog. dynamic fsync, RGB control, etc. etc.

Thanks a lot I'll certainly give it a try
Sent from my LG-D801 using Tapatalk

F2FS! I'd really like to have this file format on my G2 storage (but on an AOSP ROM)
 @jassycliq Could You maybe try to make resizing partitions possible? /system partition is waaay too big. (Almost 2GB is wasted on AOSP rom)

reas0n said:
F2FS! I'd really like to have this file format on my G2 storage (but on an AOSP ROM)
@jassycliq Could You maybe try to make resizing partitions possible? /system partition is waaay too big. (Almost 2GB is wasted on AOSP rom)
Click to expand...
Click to collapse
Resizing partitions = brick

I can use toolbox pro for cpu tunings?
Sent from my LG-D801 using Tapatalk

jeyd02 said:
I can use toolbox pro for cpu tunings?
Sent from my LG-D801 using Tapatalk
Click to expand...
Click to collapse
Pretty sure you can, I don't use it though so I can't promise anything. Some things require FauxClock seeing that it's the only app that supports those features. I believe that its only intelli-ksm, power suspend and dynamic filesync though.
As far as mpdecision and 3phase thermal, it requires trickstermod, kcontrol, or anyother app that has support for it (don't know all those out there)

Awwee Ok. Why requiring various apps. It's okay for it to running all those app at background.?
Sent from my LG-D801 using Tapatalk

jeyd02 said:
Awwee Ok. Why requiring various apps. It's okay for it to running all those app at background.?
Sent from my LG-D801 using Tapatalk
Click to expand...
Click to collapse
I don't require any apps at all.
The defaults ran great for me, but if you want to take more control I was recommending one of those.
They tend to run fine, AFAIK they don't usually continue running and applying settings over and over (only time i experienced that was with Mako and it would only apply rgb values on screen on) Most of the times these apps write to a file and that's it. Other ones just read it and modify as needed.