Hey all. This is my first post here on XDA.
I've been thinking of compiling my own kernel image for my HTC Legend. I've spent hours googling and reading different forums and blog's. But without greater success, most of the guidelines are not spot on and some things are not cristal. So I have a few questions regarding this.
I have read that a specific radio image is just "comptabile" for compilation with a specific kernel version. So let's say I have HTC_LEGEND_RADIO_7_083521_sign.zip, how do I know what kernel version this is made for? In my ears this sounds pretty strange . I would not be able to build a kernel from that radio image and the 2.6.35.5 Linux kernel?
I've also tried to figure out what exactly the radio image is from various boguos posts on different forums, and what I belive it is from the tiny bits of information I've found is that It's modules for the phones hardware?
Oh, and what is the latest radio image version available for the HTC Legend?
Hope someone can help me clear some of my questions
I can only answer the very last question lol, it's 7.08.35.21
Latest RUU:
RUU_Legend_Vodafone_AU_2.10.178.1_Radio_47.39.35.09_7.08.35.21_release_138238_signed.exe
First seen in:
RUU_Legend_HTC_WWE_2.03.405.3_Radio_47.39.35.09_7.08.35.21_release_130330_signed.exe
Hi tonper,
First... where did you get that radio has something to do with kernel??? Radio image is basically a firmware (operating system of its own that interacts with radio hw) and has nothing to do with Android kernel which is in fact patched Linux kernel. One can access all functions of radio through userspace Android apps that communicate with radio via native libraries. That's all that it is to say about radio in conjunction with kernel.
To be able to build a kernel one should first master basics of Unix/Linux system principles and first try to build own kernel for some Linux distro. One need to know also how to boot this kernel and use it with the rest of the operating system. Last thing to master is cross-compiling as you will be building kernel for ARM architecture most possibly on your x86 PC. In short the topics to search for would be:
* knowing Linux (principles)
* booting Linux (process)
* compiling Linux kernel
* cross-compiling
Android powah!
Thanks for your answer BlaY0!
Will check into it some more this weekend with this new information.
BlaY0 said:
* knowing Linux (principles)
* booting Linux (process)
* compiling Linux kernel
* cross-compiling
Click to expand...
Click to collapse
Been running Linux for 10 years and Debian for 7 years. So got the basic knowledge I just can't find any good documentation on this topic
Cyanogenmod wiki have some guides
http://wiki.cyanogenmod.com/index.php?title=Building_Kernel_from_source
snakehult said:
Cyanogenmod wiki have some guides
Click to expand...
Click to collapse
How could I have missed that page? Have been googling like a maniac. That was pretty much exactly what I was looking for. Thanks alot snake
tonper said:
Been running Linux for 10 years and Debian for 7 years. So got the basic knowledge I just can't find any good documentation on this topic
Click to expand...
Click to collapse
OK, then you're set m8 you just need to grab some ARM cross-compiling toolchain for x86. Your first bet would be Android NDK. You can also use CodeSourcery or even build your own toolchain with Buildroot... but for compiling kernel really doesn't matter which one U use.
Happy compiling
I just searched ebay for RK3066 Development Board, and found one there, I have played the Beagle board.
this one seem good for play with RK3066 firmware and debug rom.
wy6688 said:
I just searched ebay for RK3066 Development Board, and found one there, I have played the Beagle board.
this one seem good for play with RK3066 firmware and debug rom.
Click to expand...
Click to collapse
Not used that one, but I'd caution you away from the RK3066 in general, as I have used a device with RK3066 chipset...
Rockchip don't respect or follow the GPL and give sources with binaries included. This means you cannot compile entirely from source, and can be problematic. Some core drivers like the clock driver are blobbed.
There may be better development boards out there, the ODroid X2 is one I've heard good things about.
I have checked the Latest Linux Kernel source from Kernel.org, all their source code included, using new ARM DTS system (device tree source), so you can compile all your kernel directly from mainstream kernel, except some board level chip driver, which you need to customize from Driver fold, /linux/arch/arm/boot/dts/rk3066a.dtsi, rk3066a-clocks.dtsi
wy6688 said:
I have checked the Latest Linux Kernel source from Kernel.org, all their source code included, using new ARM DTS system (device tree source), so you can compile all your kernel directly from mainstream kernel, except some board level chip driver, which you need to customize from Driver fold, /linux/arch/arm/boot/dts/rk3066a.dtsi, rk3066a-clocks.dtsi
Click to expand...
Click to collapse
Have you got NAND drivers for the rk3066 nand? clock drivers, ddr drivers?
(see files with a .uu extension, it's a uuencoded .o binary)
https://github.com/AndrewDB/rk3066-kernel/tree/master/arch/arm/mach-rk30 (ddr.uu, ddr_freq.uu, clock_data.uu)
https://github.com/AndrewDB/rk3066-kernel/tree/master/drivers/video (fb.uu)
I believe there's also some missing drivers for the nand (https://github.com/AndrewDB/rk3066-kernel/tree/master/drivers/mtd/rknand)
I just got the development broad from ebay and checked the source code, it do missing these source file and although you can build the kernel without issue for 3.0.8.
I believe the latest 3.10.x kernel from kernel org, which included the RK3066 DTS files, can be used to build the generic kernel that can run on RK3066 board and should have no other source code needed.
I'm trying to build this kernel based on 3.10.x, anyone know detail steps that build the generic kernel based on RK3066 DTS?
Thank advance.
I have a GALAXY E5 (SM-E500H) with CM12.1 from jackeagle.
I'd like to know if there's a way (no matter how hard) to convert the OS to 64 bits.
A 64 bits rom should considerably improve performance of the device (but may reduce apps compatibility).
To check if device have is 64 bit, just go to system/lib and system/vendor look for a folder called "lib64". If still not sure, decompile boot.img file and go to ramdisk (or initrd) folder and look for init.zygote64.rc . You may want to check in system/bin for app_proccess64 and dalvikvm64
If one of these files is present then it's 64 bits (especially init.zygote64.rc)
Nonta72 said:
I have a GALAXY E5 (SM-E500H) with CM12.1 from jackeagle.
I'd like to know if there's a way (no matter how hard) to convert the OS to 64 bits.
A 64 bits rom should considerably improve performance of the device (but may reduce apps compatibility).
To check if device have is 64 bit, just go to system/lib and system/vendor look for a folder called "lib64". If still not sure, decompile boot.img file and go to ramdisk (or initrd) folder and look for init.zygote64.rc . You may want to check in system/bin for app_proccess64 and dalvikvm64
If one of these files is present then it's 64 bits (especially init.zygote64.rc)
Click to expand...
Click to collapse
Didnt find it in mine. I think mine is 32 bit
Both E5 and E7 has Snapdragon 420 which Supportshould 64-bit but Samsung made only 32 bit os
Hoardseeker said:
Both E5 and E7 has Snapdragon 420 which Supportshould 64-bit but Samsung made only 32 bit os
Click to expand...
Click to collapse
Correction: It's Snapdragon 410
Well, The Qualcomm Snapdragon 410 equipped on the system chips of E-Series, supports 64-bit architecture, but Samsung didn't enable that on the Lollipop.
I don't know but I think it's hard to enable 64-bit on Lollipop, but for sure, If It's enabled the E-Series performance would be much more smoother.
Hassan 4 said:
Well, The Qualcomm Snapdragon 410 equipped on the system chips of E-Series, supports 64-bit architecture, but Samsung didn't enable that on the Lollipop.
I don't know but I think it's hard to enable 64-bit on Lollipop, but for sure, If It's enabled the E-Series performance would be much more smoother.
Click to expand...
Click to collapse
It can be done.
If we compile a kernel with ARCH=arm64 and a 64 bit CROSS_COMPILER (aarch64); we should be good to go. But the problem is that I'm not sure if apps will keep working this case, because we need 64bit libs and some binaries (app_proccess64, dalvikvm64, debugerred64 etc.)
I will try when my exams finish.
Currently on hold due to working on Nougat first.
I've got a booting or should I say bootlooping build of Lineage 15.0 for I9000. (galaxysmtd)
I've had to use crazy hacks like adb binary from 7.1 in ramdisk.
Just to get `adb logcat` working.
For now it's stuck at bootlogo. I've attached the logcat here.
I'm looking into it to figure out what needs to be done.
Sources:
manifests and patches I've used.
https://github.com/galaxys1-resurrected/local_manifests
https://github.com/galaxys1-resurrected/android_patches
Kernel:
https://github.com/galaxys1-resurrected/android_kernel_samsung_aries
Device Tree:
https://github.com/galaxys1-resurrected/android_device_samsung_aries-common
https://github.com/galaxys1-resurrected/android_device_samsung_galaxysmtd
Thanks:
@rINanDO for backporting kernel side of things to 3.0
@xc-racer99 and @Coldwindofnowhere for getting the device upto android 7.1
And all others who had worked from beginning till now on this device.
Is there anyone still working on this?
I was curious if anyone this was still being developed? I'm totally newbie in the android scene but have some knowledge of operating systems and am interested in resurrecting my i9000.
I went through the logs and a couple of things jumped out:
1) Surface flinger returning non zero exit status because it needs OpenGL ES v2.0 or greater. I believe i9000's GPU PowerVR SGX540 supports OpenGL ES 2.0, so this issue could be solved.
2) Media extractor crash: /system/bin/mediaextractor: libminijail[1291]: prctl(PR_SET_NO_NEW_PRIVS): Invalid argument, whatever the heck it means.
3) activity_recognition HAL is deprecated, so ActivityRecognitionHardware class's init does not do anything.
For 3 , I got to android_hardware_location_ActivityRecognitionHardware.cpp's source where it comments out activity_recognition.h with the following comment:
Code:
// #include <hardware/activity_recognition.h>
// The activity recognition HAL is being deprecated. This means -
// i) Android framework code shall not depend on activity recognition
// being provided through the activity_recognition.h interface.
// ii) activity recognition HAL will not be binderized as the other HALs.
I believe more work has been done since this post based on git commits lasting upto Nov'17. Would be great if someone could post logs for an updated build. I feel that android oreo with go optimizations would be a really good fit for i9000 and uphold this device's legendary support. I mean a device running from eclair all the way to oreo would be amazing.
Even if this might not work out, I would like to thank @(°_o), @xc-racer99 , @Coldwindofnowhere and @rINanDO for bringing i9000 upto nougat. I believe even i9000's nexus sibling nexus s does not have a working nougat rom.
a1shakes said:
I was curious if anyone this was still being developed? I'm totally newbie in the android scene but have some knowledge of operating systems and am interested in resurrecting my i9000.
I went through the logs and a couple of things jumped out:
1) Surface flinger returning non zero exit status because it needs OpenGL ES v2.0 or greater. I believe i9000's GPU PowerVR SGX540 supports OpenGL ES 2.0, so this issue could be solved.
2) Media extractor crash: /system/bin/mediaextractor: libminijail[1291]: prctl(PR_SET_NO_NEW_PRIVS): Invalid argument, whatever the heck it means.
3) activity_recognition HAL is deprecated, so ActivityRecognitionHardware class's init does not do anything.
For 3 , I got to android_hardware_location_ActivityRecognitionHardware.cpp's source where it comments out activity_recognition.h with the following comment:
Code:
// #include <hardware/activity_recognition.h>
// The activity recognition HAL is being deprecated. This means -
// i) Android framework code shall not depend on activity recognition
// being provided through the activity_recognition.h interface.
// ii) activity recognition HAL will not be binderized as the other HALs.
I believe more work has been done since this post based on git commits lasting upto Nov'17. Would be great if someone could post logs for an updated build. I feel that android oreo with go optimizations would be a really good fit for i9000 and uphold this device's legendary support. I mean a device running from eclair all the way to oreo would be amazing.
Even if this might not work out, I would like to thank @(°_o), @xc-racer99 , @Coldwindofnowhere and @rINanDO for bringing i9000 upto nougat. I believe even i9000's nexus sibling nexus s does not have a working nougat rom.
Click to expand...
Click to collapse
Unfortunately, no one is really actively working on Oreo. As you've found out, it's an issue with the graphics drivers that is holding everything back. No device (that I've found) that uses a PowerVR graphics chip (we use the PowerVR SGX 540) has working graphics drivers on Oreo. There were rumours that someone had found newer working blobs, but weren't able to release them publicly due to intellectual property laws that they were trying to figure out (but this was months ago).
Our GPU does support sufficient enough OpenGL, but only using BGRA8888 as opposed to RGBA8888. BGRA hasn't officially been supported in Android since ~4.2, but there's been a hack used to make things work. Come Oreo, things have changed and the hack no longer applies cleanly. However, I think the really issue is that the gralloc blobs was extended by PowerVR (see https://github.com/xc-racer99/andro...6.0/exynos3/s5pc110/include/hal_public.h#L119) but with the binderized HALs/VNDK/other low-level Oreo changes something has broken. I had a go at trying to work around things, but failed too.
There are a few ways I can think of getting working graphics:
1) Someone finds some updated blobs for the PowerVR SGX 540 for ARM (I've found x86 ones, but they don't work for obvious reasons)
2) Someone hacks around the source code so that the blobs work - but I'm not sure if it's PowerVR "extension" of the gralloc interface that is causing issues or not...
3) We simply use software rendering, but this would be so slow with our ancient CPU that I haven't bothered to try
4) We work on porting a newer kernel so we have the Samsung DRM kernel driver, use the Linux PowerVR blobs coupled with drm_gralloc/drm_hwcomposer and maybe a wrapper like https://github.com/TexasInstruments/dri3wsegl and somehow cobble together working support
In terms of the mediaextractor crash, that's due to the kernel missing seccomp support. There's a whole bunch of different backports, some more successful than others. Due to our ancient kernel, backporting is no longer very easy...
If we could somehow get the graphics drivers working, we'd have a pretty good base as there are free implementations of all HALs/drivers except for GPS and TV-Out (and, of course, graphics....).
Are you really working on porting oreo on the i9000?
How do you deal with the small amount of ram?
Are you using the 'low end device' oreo feature?
nailyk said:
Are you really working on porting oreo on the i9000?
How do you deal with the small amount of ram?
Are you using the 'low end device' oreo feature?
Click to expand...
Click to collapse
No, no one (that I know of) is actively working on Oreo for the first-gen Galaxy S devices. There were attempts, the kernel got in good enough shape that everything wasn't immediately crashing, but due to the graphics driver issues described a couple posts ago nobody has managed to get a fully booting build.
xc-racer99 said:
No, no one (that I know of) is actively working on Oreo for the first-gen Galaxy S devices. There were attempts, the kernel got in good enough shape that everything wasn't immediately crashing, but due to the graphics driver issues described a couple posts ago nobody has managed to get a fully booting build.
Click to expand...
Click to collapse
Thanks for fast anwser. Yes, the graphic driver problem exist on another of my exynos device.
Anyway I wasn't able to boot the 7.1 (not able to boot something else than 2.3.6 )
Will attempt to see that post you are talking about but am probably not smart enough to deal with graphics drivers
Thanks for your time.
nailyk said:
Thanks for fast anwser. Yes, the graphic driver problem exist on another of my exynos device.
Anyway I wasn't able to boot the 7.1 (not able to boot something else than 2.3.6 )
Will attempt to see that post you are talking about but am probably not smart enough to deal with graphics drivers
Thanks for your time.
Click to expand...
Click to collapse
If you're serious about trying to mess with graphics drivers, it might be interesting to check out the blobs from https://www.renesas.com/pt-br/produ...ion-boards/renesas-starter-kit-for-rzg1e.html as it's an ARM-based device with the SGX540. It's possible that they're new enough to not run into the same issues as the older blobs (but equally possible that even the kernel part is closed source). The binary blobs are only semi-SoC specific as I've managed to use the OMAP blobs with only having hardware decoding being broken.
Is it for real???
I9000 !!
Apparently, some new SGX540 and SGX544 DDK blobs for OMAP4 have appeared:
https://gerrit.unlegacy-android.org/#/c/Unlegacy-Android/proprietary_vendor_ti/+/10525/
https://gerrit.unlegacy-android.org/#/q/topic:omap-ddk-1.14+(status:open+OR+status:merged
In fact, (Barnes and Noble's) hummingburd and ovation are both based on SGX544 and have gotten an Oreo ROM (using the new blobs).
https://forum.xda-developers.com/showpost.php?p=77526206&postcount=2490
Use android go it will be better.
MYEUHD said:
Apparently, some new SGX540 and SGX544 DDK blobs for OMAP4 have appeared:
https://gerrit.unlegacy-android.org/#/c/Unlegacy-Android/proprietary_vendor_ti/+/10525/
https://gerrit.unlegacy-android.org/#/q/topic:omap-ddk-1.14+(status:open+OR+status:merged
In fact, (Barnes and Noble's) hummingburd and ovation are both based on SGX544 and have gotten an Oreo ROM (using the new blobs).
https://forum.xda-developers.com/showpost.php?p=77526206&postcount=2490
Click to expand...
Click to collapse
Yep, I've seen the blobs, they've been there for awhile now. Just haven't had a chance to run a build with the blobs to see if they work. It's on my to-do list when I find the time
xc-racer99 said:
Yep, I've seen the blobs, they've been there for awhile now. Just haven't had a chance to run a build with the blobs to see if they work. It's on my to-do list when I find the time
Click to expand...
Click to collapse
Alright, I've had a chance to look at the blobs now. I have a build, but unfortunately it looks as if we need to adjust our hwcomposer as well We use a relatively old hwc 1.0 but the new gralloc blob doesn't appear to keep the framebuffer open which is a requirement for a hwcomposer this old. There is a prebuilt blob that is used by omap4 devices but it doesn't work on s5pc110 due to the fact that it uses some DSS stuff which is OMAP-specific. Still plenty of work to do, without even trying to figure out all the Oreo/Pie changes (I'm testing on KitKat as that's the build environment I have setup right now).
xc-racer99 said:
Alright, I've had a chance to look at the blobs now. I have a build, but unfortunately it looks as if we need to adjust our hwcomposer as well We use a relatively old hwc 1.0 but the new gralloc blob doesn't appear to keep the framebuffer open which is a requirement for a hwcomposer this old. There is a prebuilt blob that is used by omap4 devices but it doesn't work on s5pc110 due to the fact that it uses some DSS stuff which is OMAP-specific. Still plenty of work to do, without even trying to figure out all the Oreo/Pie changes (I'm testing on KitKat as that's the build environment I have setup right now).
Click to expand...
Click to collapse
We need a newer hwc anyway, as Pie requires at least hwc 1.3:
ChronoMonochrome said:
In 9.0, to get graphics to work, device is required to support HWC2 (or use either HWC2on1 or HWC2onFb adapters).
Click to expand...
Click to collapse
ChronoMonochrome said:
Yes, HWC has to be at least 1.3, to work with one of aforementioned adapters. With one of those adapters it will work like it was HWC 2 (but actually not exactly same).
Click to expand...
Click to collapse
As a reference, the galaxy S3's hwc was updated from 1.0 to 1.4: Thread
hardware/samsung
MYEUHD said:
We need a newer hwc anyway, as Pie requires at least hwc 1.3:
As a reference, the galaxy S3's hwc was updated from 1.0 to 1.4: Thread
hardware/samsung
Click to expand...
Click to collapse
Was unaware of the fact. Are you volunteering to make the patches?
I've uploaded my changes to https://github.com/xc-racer99/proprietary_vendor_samsung/tree/ddk-1.14 https://github.com/xc-racer99/android_hardware_samsung/tree/ddk-1.14 https://github.com/xc-racer99/android_kernel_samsung_aries/tree/ddk-1.14 https://github.com/xc-racer99/android_device_samsung_telusgalaxys4gmtd/tree/ddk-1.14 https://github.com/xc-racer99/android_device_samsung_aries-common/tree/ddk-1.14 but I think this might be the last I work on this as I don't really have the motivation to work on it anymore. Note the patches are against a custom version of Unlegacy Android 4.4 so you'll need to cherry pick the changes to your ROM of choice if desired.
The changes build, the EGL appears to initialize, but I always get
Code:
E/libEGL ( 471): validate_display:254 error 3008 (EGL_BAD_DISPLAY)
And in dmesg:
Code:
[ 8.509291] init: computing context for service '/system/vendor/bin/pvrsrvinit'
[ 8.509601] init: starting 'pvrsrvinit'
...
[ 8.601890] PVR_K: UM DDK-(4081762) and KM DDK-(4081762) match. [ OK ]
...
[ 8.765955] init: process 'pvrsrvinit', pid 99 exited
...
[ 55.560021] PVR_K:(Error): PVRSyncIOCTLCreate: Failed to find unused fd (-24)
[ 55.563491] PVR_K:(Error): PVRSyncIOCTLCreate: Failed to find unused fd (-24)
[ 55.597577] s3cfb s3cfb: [fb0] video memory released
Whether the issue is in the HWC or the gralloc blob that we've stolen from OMAP, I have no idea.
xc-racer99 said:
Was unaware of the fact. Are you volunteering to make the patches?
Click to expand...
Click to collapse
Will try to do my best!
BTW, do I really need jdk-7 to compile kitkat? or does it simply work with jdk-8??
MYEUHD said:
Will try to do my best!
BTW, do I really need jdk-7 to compile kitkat? or does it simply work with jdk-8??
Click to expand...
Click to collapse
You really do need jdk-7... I used the "reference implementation" available at http://jdk.java.net/java-se-ri/7 and made sure the java executables were in the PATH before the java I actually have installed.
Note that my Unlegacy Android trees will not work for the i9000 (well, they might, but you'd need to install u-boot as well at a bare minimum...)
It's kinda Insane that people are trying to get an 8 year phone to run oreo
@xc-racer99 Do you still have the AOSP 7.1 source code on your computer?
MYEUHD said:
@xc-racer99 Do you still have the AOSP 7.1 source code on your computer?
Click to expand...
Click to collapse
I've got the .repo folder, but don't have the individual files expanded as I don't have the disk space Could run a repo sync and look at things but don't have the disk space for a full build.
The_Pacific_gamer said:
It's kinda Insane that people are trying to get an 8 year phone to run oreo
Click to expand...
Click to collapse
its kinda insane that people are still using this phone.